cs.RO / 1 / 2608.07546
Generalizing deep reinforcement learning across cable-driven parallel robot configurations with actuator-level policies
通过执行器级策略对电缆驱动并联机器人配置进行深度强化学习的泛化
Abstract
Cable-driven parallel robots (CDPRs) present diverse configurations and complex control challenges, which can be addressed by deep reinforcement learning (DRL) by learning their nonlinear dynamics. However, DRL methods often require extensive training time, and the resulting policies do not generalize well to different robot configurations or varying numbers of actuators. In this article, we introduce a novel DRL approach for controlling CDPRs that does not depend on the specific robot configuration. Our method trains an actuator-level policy that controls each motor to achieve its target cable length, in contrast to conventional DRL approaches that learn to control the entire robot to reach a desired end-effector position. To the best of our knowledge, this is the first work to apply DRL to control CDPRs using an actuator-level policy. This approach offers two main advantages: (i) a single shared policy can be applied to any CDPR configuration, regardless of actuator count, and (ii) reliance on inverse kinematics, avoiding the more challenging forward kinematics problem. Training is performed in simulation, and the learned policy is successfully transferred to a real CDPR. Experimental results show that the actuator-level policy (ALP) surpasses traditional reinforcement learning methods in both robustness and precision. We further control a real 8-motor CDPR with 3D motion using a policy trained on a simulated 4-motor planar CDPR operating in 2D. This illustrates that the proposed method is applicable to any CDPR configuration, independent of actuator number or placement.
Chinese Translation
电缆驱动并联机器人（CDPR）具有多样的配置和复杂的控制挑战，这些挑战可以通过深度强化学习（DRL）来解决，从而学习其非线性动态。然而，DRL 方法通常需要大量的训练时间，并且所得到的策略在不同机器人配置或不同数量的执行器上泛化能力较差。本文提出了一种新颖的 DRL 方法，用于控制 CDPR，该方法不依赖于特定的机器人配置。我们的方法训练一个执行器级策略，控制每个电机以达到其目标电缆长度，这与传统的 DRL 方法不同，后者学习控制整个机器人以达到期望的末端执行器位置。根据我们所知，这是首次将 DRL 应用于使用执行器级策略控制 CDPR。该方法提供了两个主要优势：（i）一个共享策略可以应用于任何 CDPR 配置，无论执行器数量如何；（ii）依赖于逆运动学，避免了更具挑战性的正运动学问题。训练在仿真中进行，学习到的策略成功转移到真实的 CDPR。实验结果表明，执行器级策略（ALP）在鲁棒性和精度上超越了传统的强化学习方法。我们进一步使用在仿真中训练的策略控制一个真实的 8 电机 CDPR 进行 3D 运动，该策略是在一个 4 电机平面 CDPR（在 2D 中操作）上训练的。这表明所提出的方法适用于任何 CDPR 配置，与执行器数量或位置无关。
cs.RO / 2 / 2608.07548
SC$^{2}$-WM: A Self-Correcting World Model with Closed-Loop Feedback for Vision-and-Language Navigation in Continuous Environments
SC$^{2}$-WM：一种具有闭环反馈的自我校正世界模型，用于连续环境中的视觉与语言导航
Abstract
Vision-and-Language Navigation in Continuous Environments (VLN-CE) requires agents to make fine-grained navigation decisions under partial observability. However, most existing methods rely on open-loop execution, lacking mechanisms to detect and correct internal state drift during inference. We propose SC$^{2}$-WM, a self-correcting world model framework that introduces internal feedback for closed-loop decision making in VLN-CE. Our method derives feedback from world-model foresight to perform state-level plan refinement before action execution. To handle challenging scenarios, we further introduce conditional world-aware adaptation, which enables model-level correction by selectively updating the world model at test time when feedback indicates model capacity insufficiency. Experiments on standard VLN-CE benchmarks demonstrate improved navigation robustness and generalization. Our code is available at https://github.com/sunrise-ikun/SC2_WM.
Chinese Translation
连续环境中的视觉与语言导航（VLN-CE）要求智能体在部分可观察性下做出细粒度的导航决策。然而，大多数现有方法依赖于开放式执行，缺乏在推理过程中检测和校正内部状态漂移的机制。我们提出了SC$^{2}$-WM，一种自我校正世界模型框架，引入了用于VLN-CE的闭环决策内部反馈。我们的方法通过世界模型的前瞻性推导反馈，以在执行动作之前进行状态级计划细化。为了应对具有挑战性的场景，我们进一步引入了条件世界感知适应，这使得在反馈指示模型能力不足时，通过选择性地在测试时更新世界模型来实现模型级校正。在标准VLN-CE基准上的实验表明，导航的鲁棒性和泛化能力得到了改善。我们的代码可在 https://github.com/sunrise-ikun/SC2_WM 获取。
cs.RO / 3 / 2608.07553
Self Supervised Learning from Automatically Generated Demonstrations for Visual Robotic Manipulation
基于自动生成演示的自监督学习用于视觉机器人操作
Abstract
Robotic manipulation often requires object specific programming, manual data annotation, or calibrated perception pipelines, which limits rapid deployment in practical settings. Learning from demonstration offers a more direct alternative, but collecting demonstrations can still demand human teleoperation or kinesthetic teaching. This paper presents a self supervised visual manipulation method in which a robot automatically generates demonstrations around a target pose and learns relative pose corrections directly from wrist mounted RGB images. The proposed pipeline uses ROS~2 and Isaac Sim to collect labeled image-pose pairs without requiring explicit camera to robot extrinsic calibration. Separate datasets are generated for planar refinement and coarse three dimensional approach, and a convolutional network is trained to regress relative translation and rotation from single frame RGB observations. During execution, a coarse to fine controller first approaches the object using models trained with height variation and then refines the final alignment using planar data. The method is evaluated both in simulation and on a real UR5e collaborative robot equipped with a gripper and a monocular camera. In simulation, the refinement stage reduces the final planar dispersion from 9.69 mm to 5.38 mm. In real world experiments, the system performs end to end grasp attempts on three physical objects and reaches success rates of 66.6% and 63.6% for two objects without object rotation, while still maintaining partial robustness under rotated conditions. These results show that automatically generated demonstrations can support practical visual manipulation with limited setup effort, while also exposing remaining challenges in depth prediction and object dependent generalization.
Chinese Translation
机器人操作通常需要特定于对象的编程、手动数据标注或校准的感知管道，这限制了在实际环境中的快速部署。通过演示学习提供了一种更直接的替代方案，但收集演示仍然可能需要人工遥控或动觉教学。本文提出了一种自监督视觉操作方法，其中机器人围绕目标姿态自动生成演示，并直接从手腕安装的RGB图像中学习相对姿态修正。所提出的流程使用ROS~2和Isaac Sim收集标记的图像-姿态对，而无需显式的相机与机器人之间的外部校准。为平面细化和粗略三维接近生成了不同的数据集，并训练了一个卷积网络，从单帧RGB观测中回归相对平移和旋转。在执行过程中，粗到细控制器首先使用经过高度变化训练的模型接近对象，然后使用平面数据细化最终对齐。该方法在模拟和配备夹爪及单目相机的真实UR5e协作机器人上进行了评估。在模拟中，细化阶段将最终平面离散度从9.69毫米减少到5.38毫米。在现实世界实验中，该系统对三个物理对象进行端到端抓取尝试，并在两个没有物体旋转的对象上达到了66.6%和63.6%的成功率，同时在旋转条件下仍保持部分鲁棒性。这些结果表明，自动生成的演示可以支持在有限设置工作下的实际视觉操作，同时也暴露了深度预测和对象依赖泛化方面的剩余挑战。
cs.RO / 4 / 2608.07555
You Don't Need To Stay in The Loop: An Agentic Robotics Loop for Robot-Policy Improvement
你不需要保持在循环中：一种用于机器人策略改进的自主机器人循环
Abstract
Coding agents such as Claude Code and Codex close the software loop: a main agent manages the loop, subagents analyze and execute, tools do the work. We port this architecture to robot-policy improvement, where one difference dominates the design: robotic tools---trained policies, training pipelines, data collection---fail routinely, so a tool's quality must be measured, recorded at every call, and expired when the artifact behind it changes. AgenticRobotics is a backend-independent control plane in which an LLM controller drives disposable workers through durable train--evaluate--improve transactions: an immutable objective, controller-owned measurement, commit-keyed crash recovery, an evidence-graded skill library, and a tool registry with a standardized, recorded call surface. The title is an operational claim, not a selection claim: the operator can leave because promotion is evidence-gated, state is recoverable, and capability quality is derived from records---not because the loop picks better checkpoints than a human; on the one lineage we measured, it does not. The gates measurably buy false-promotion control (0.001 per run hardened versus 0.005--0.021 shipped), anytime-valid decisions under optional stopping, zero lost or duplicate effects under kill injection, and six of six artifact-tampering classes caught by a signed verifier.
Chinese Translation
编码代理如Claude Code和Codex关闭了软件循环：主要代理管理循环，子代理分析和执行，工具完成工作。我们将这一架构移植到机器人策略改进中，其中一个主要差异主导了设计：机器人工具——训练策略、训练管道、数据收集——经常失败，因此工具的质量必须在每次调用时进行测量和记录，并在其背后的工件发生变化时失效。AgenticRobotics是一个与后端无关的控制平面，其中一个大型语言模型（LLM）控制器通过持久的训练-评估-改进事务驱动一次性工作者：一个不可变的目标、控制器拥有的测量、基于提交键的崩溃恢复、一个证据分级的技能库，以及一个具有标准化、记录调用表面的工具注册表。标题是一个操作性声明，而非选择性声明：操作员可以离开，因为提升是基于证据的，状态是可恢复的，能力质量源于记录——而不是因为循环选择的检查点比人类更好；在我们测量的一个血统中，它并没有。门控在可测量的程度上购买了虚假提升控制（每次运行0.001硬化对比0.005-0.021交付）、在可选停止下的随时有效决策、在杀死注入下零损失或重复效果，以及由签名验证器捕获的六类工件篡改。
cs.RO / 5 / 2608.07557
AeroDPO: Unleashing Lightweight UAV Navigation with High-Fidelity Perception and Automated Preference Optimization
AeroDPO：释放轻量级无人机导航的高保真感知与自动偏好优化
Abstract
Vision-Language Navigation for Unmanned Aerial Vehicles (UAV-VLN) requires rapid and reactive control in complex 3D environments. Recent minimalist end-to-end paradigms show great promise but typically rely on massive language models containing billions of parameters, incurring prohibitive latency for real-world edge deployment. In this paper, we challenge this parameter-heavy reliance. Comprehensive cross-scale evaluations reveal the critical insight that perception quality fundamentally outweighs language reasoning capacity. We demonstrate that a lightweight 2B model equipped with high-fidelity visual inputs completely matches the overall success rates of massive 7B baselines. However, this minimalist policy exposes a fundamental robustness flaw inherent to pure Behavior Cloning (BC). Lacking explicit negative feedback, the agent fails to internalize robust spatial constraints and exhibits alarming collision rates in out-of-distribution (OOD) scenarios. To overcome this vulnerability without relying on unscalable human annotations, we propose AeroDPO, a zero-cost automated Direct Preference Optimization pipeline driven by deterministic physical simulation state rollback. Upon detecting collisions, the system autonomously rewinds the environment to extract causal reasoning errors as rejected actions, applies decoupled privileged interventions to synthesize collision-avoidance preferred maneuvers, and leverages an offline vision language inspector to filter visual ambiguities. By equipping our 2B model with this automated data flywheel, AeroDPO boosts success rates to 49.16% on unmapped scenarios while drastically suppressing collision rates, establishing a new SOTA for autonomous aerial agents.
Chinese Translation
无人机视觉语言导航（UAV-VLN）需要在复杂的三维环境中进行快速和反应式控制。最近的极简端到端范式显示出巨大的潜力，但通常依赖于包含数十亿参数的大型语言模型，这在现实世界的边缘部署中会导致不可接受的延迟。在本文中，我们挑战这种对参数的过度依赖。全面的跨尺度评估揭示了一个关键见解：感知质量在根本上超过了语言推理能力。我们证明，配备高保真视觉输入的轻量级2B模型的整体成功率完全匹配大型7B基线。然而，这种极简策略暴露了纯行为克隆（BC）固有的基本鲁棒性缺陷。由于缺乏明确的负反馈，代理无法内化稳健的空间约束，并在分布外（OOD）场景中表现出令人担忧的碰撞率。为了克服这一脆弱性而不依赖于不可扩展的人类注释，我们提出了AeroDPO，一个由确定性物理模拟状态回滚驱动的零成本自动直接偏好优化管道。在检测到碰撞时，该系统自主回滚环境，以提取因果推理错误作为被拒绝的动作，应用解耦的特权干预来合成避免碰撞的优选机动，并利用离线视觉语言检查器过滤视觉歧义。通过为我们的2B模型配备这一自动数据飞轮，AeroDPO在未映射场景中的成功率提升至49.16%，同时大幅抑制了碰撞率，为自主空中代理建立了新的最先进水平（SOTA）。
cs.RO / 6 / 2608.07558
Learning Physical Interaction: A Survey of Tactile- and Force-aware Robot Learning
学习物理交互：触觉和力感知机器人学习的综述
Shan, Shilin, Zhou, Chuhao, Wang, Ruize, Chen, Xinyan, Chen, Xiangyu, Zhou, Xinyu, Ma, Boyu, Hu, Iris Yuxuan, Li, Jingliang, Hu, Celeste Yuxuan, Li, Geng, Chen, Guohao, Zhu, Tianrui, Li, Zhe, Ze, Yanjie, Geng, Haoran, Dou, Zhiyang, Bi, Jianxin, Liu, Yuejiang, Zhou, Jianshu, Li, Jiachen, Liang, Paul, Harada, Tatsuya, Katzschmann, Robert, Soh, Harold, Li, Na, Johns, Edward, Kragic, Danica, Peters, Jan, Matusik, Wojciech, Tomizuka, Masayoshi, Malik, Jitendra, Yang, Jianfei
Abstract
Physically grounded robot intelligence requires robots to perceive, reason about, and regulate their interactions with the physical world. This capability is particularly critical in contact-sensitive manipulation, where successful task execution depends not only on visual perception and motion generation, but also on force regulation and adaptive control. In this context, recent robot learning methods have made substantial progress by integrating force, tactile, vision, language, and proprioceptive sensing into learned manipulation policies. In parallel, many systems adopt multi-phase architectures that combine high-level policies, action-refinement modules, and low-level controllers to bridge semantic task understanding with reactive physical execution. Despite these advances, existing surveys have not explicitly reviewed force- and tactile-aware robot learning from a unified perspective that jointly captures multimodal sensing and multi-phase system design. This survey addresses this gap by proposing TF-ART, a Tactile/Force-Aware Robot learning Taxonomy for multimodal and multi-phase frameworks, which maps individual methods into a unified hierarchical structure. The framework characterizes how recent works organize observation modalities, encode and fuse heterogeneous sensory inputs, generate and refine actions across multiple phases, and connect learned policies to reactive robot-end control. Building on this methodological view, we further examine the task settings and infrastructure requirements of physical interaction, thereby integrating both algorithmic and practical perspectives on force- and tactile-aware robot learning.
Chinese Translation
基于物理的机器人智能要求机器人感知、推理和调节与物理世界的交互。这种能力在接触敏感的操作中尤为关键，因为成功的任务执行不仅依赖于视觉感知和运动生成，还依赖于力的调节和自适应控制。在这种背景下，近期的机器人学习方法通过将力、触觉、视觉、语言和本体感知整合到学习的操作策略中，取得了显著进展。同时，许多系统采用多阶段架构，结合高层策略、动作细化模块和低层控制器，以桥接语义任务理解与反应式物理执行。尽管取得了这些进展，现有的综述并未从统一的视角明确回顾力感知和触觉感知的机器人学习，未能共同捕捉多模态感知和多阶段系统设计。本文通过提出TF-ART（触觉/力感知机器人学习分类法），填补了这一空白，旨在为多模态和多阶段框架提供一个统一的层次结构，将各个方法映射到其中。该框架描述了近期研究如何组织观察模态、编码和融合异质传感输入、在多个阶段生成和细化动作，以及如何将学习的策略连接到反应式机器人控制。基于这种方法论视角，我们进一步考察了物理交互的任务设置和基础设施要求，从而整合了对力感知和触觉感知机器人学习的算法和实践视角。
cs.RO / 7 / 2608.07573
Projection-Retraction MPPI: Exact Constraint-Manifold Control for Manipulators
投影-回缩 MPPI：用于机械臂的精确约束流形控制
Abstract
Model Predictive Path Integral (MPPI) control is widely used in manipulation for its gradient-free, parallel handling of non-convex costs. Manipulation tasks, however, often impose constraints that hold throughout the motion: a closed kinematic chain that two grasping arms keep exactly, or joint limits and obstacle clearances that are never crossed. MPPI handles such constraints only through the cost, as soft penalties that hold approximately and fail under a strong task cost. To address this, we propose Projection-Retraction MPPI (PR-MPPI), which enforces the constraints inside the sampled dynamics. At every rollout step, the sampled velocity is projected to satisfy both constraint types: the equality restricts it to a subspace, and each inequality to a half-space within that subspace, so inequality handling never breaks the equality. This projection, however, satisfies the constraints only to first order, and a finite step leaves a small drift off the equality. Therefore, we retract the returned command back onto the constraint to numerical tolerance and independent of task weighting. We validate PR-MPPI on 14-DoF dual-arm systems. In simulation, the returned commands satisfy the closed-chain equality to numerical tolerance through a joint-limit stress test and randomized obstacle avoidance. On real hardware, the arms of a Unitree H1-2 humanoid reactively avoid a moving obstacle. Code and experiment videos are available at https://rcilab.github.io/prmppi.
Chinese Translation
模型预测路径积分（MPPI）控制因其无梯度、并行处理非凸成本而广泛应用于操控任务。然而，操控任务通常会施加在整个运动过程中保持的约束：两个抓取臂必须精确保持的闭合运动链，或从不越过的关节限制和障碍物间隙。MPPI仅通过成本处理这些约束，作为软惩罚近似地保持，但在强任务成本下会失效。为了解决这个问题，我们提出了投影-回缩 MPPI（PR-MPPI），它在采样动态中强制执行约束。在每次展开步骤中，采样的速度被投影以满足两种约束类型：等式将其限制在一个子空间内，而每个不等式则限制在该子空间内的一个半空间中，因此不等式处理从不破坏等式。然而，这种投影仅在一阶上满足约束，有限的步骤会导致与等式的轻微漂移。因此，我们将返回的指令回缩到数值容差内的约束上，并且与任务加权无关。我们在14自由度双臂系统上验证了PR-MPPI。在仿真中，返回的指令通过关节限制压力测试和随机障碍物规避，满足闭合链等式的数值容差。在真实硬件上，Unitree H1-2人形机器人臂能够主动避开移动障碍物。代码和实验视频可在 https://rcilab.github.io/prmppi 获取。
cs.RO / 8 / 2608.07576
Reconfigurable Structural Robotic Assembly: Interlocking 3D Aggregations with Self-Aligning Compound Nested Lattice Modules
可重构结构机器人组装：具有自对准复合嵌套格子模块的互锁三维聚合
Abstract
Robotic construction systems often treat the material system and the robot as separate design problems, locating intelligence primarily in hardware, sensing, motion planning, and control. This project instead investigates how geometric intelligence can be encoded within architected material systems to simultaneously address requirements for robotic grasping, self-alignment, reversible connection, structural performance, and three-dimensional aggregation. We introduce a self-aligning compound nested lattice module composed of conjoined cuboctahedral-octahedral units. The cuboctahedral features of the modules provide defined surfaces for robotic grasping and alignment, while the octahedral features incorporate screw-releasable snap-fit connectors and corresponding receptors. Additionally, we present a nested arrangement that enables interlocking aggregation along the x, y, and z axes. We demonstrate the system through furniture and architectural scale structures assembled using both a robotic arm and mobile assembler. The resulting configurations include seating, spanning structures, surfaces, and vertical frames. Compression testing of the compound module produced a stiffness of 4,556 N/mm, a maximum load of 3,445 N, and a compressive modulus of 17.5 MPa. The modules can also be disassembled and reused across different configurations, supporting reconfigurable and circular construction.
Chinese Translation
机器人建造系统通常将材料系统和机器人视为独立的设计问题，主要将智能定位于硬件、传感、运动规划和控制。该项目则探讨如何在构建材料系统中编码几何智能，以同时满足机器人抓取、自对准、可逆连接、结构性能和三维聚合的要求。我们引入了一种自对准复合嵌套格子模块，由连接的立方八面体-八面体单元组成。模块的立方八面体特征提供了用于机器人抓取和对准的定义表面，而八面体特征则包含可螺旋释放的快扣连接器及相应的接收器。此外，我们展示了一种嵌套排列，能够沿x、y和z轴进行互锁聚合。我们通过使用机器人手臂和移动组装器组装的家具和建筑规模结构来演示该系统。最终配置包括座椅、跨越结构、表面和垂直框架。对复合模块的压缩测试产生了4,556 N/mm的刚度、3,445 N的最大载荷和17.5 MPa的压缩模量。这些模块还可以拆卸并在不同配置中重复使用，支持可重构和循环建造。
cs.RO / 9 / 2608.07596
LIRA: Local Cross-Layer Information Routing for Vision-Language-Action Decoding
LIRA：用于视觉-语言-动作解码的局部跨层信息路由
Abstract
Vision-Language-Action (VLA) models transform representations from pretrained vision-language models (VLMs) into robot actions, yet the interface that routes intermediate VLM features into action decoders remains underexplored. Existing designs either expose only a narrow part of the representation hierarchy or rigidly match each decoder block to one VLM layer, restricting access to complementary task evidence across depths. We introduce LIRA, a local cross-layer action-conditioning mechanism that formulates VLM-to-action conditioning as depth-aware information routing. LIRA operates on task-token features and LIRA Query features derived from intermediate VLM states, then assigns each Parallel Fusion Block a depth-aligned local window centered on its corresponding VLM layer. Parallel Fusion Blocks aggregate neighboring LIRA Query features and integrate them with task-token features and proprioceptive inputs before action prediction. This routing interface leaves the backbone architecture, action decoder, and supervised training recipe unchanged. Across LIBERO, LIBERO-Plus, CALVIN ABC$\rightarrow$D, and real-world manipulation, LIRA improves the principal aggregate metrics over the VLA-Adapter baseline under the same 0.5B-parameter configuration. In zero-shot transfer to LIBERO-Plus, LIRA increases average success from 59.1% to 78.0%, an 18.9-point gain indicating improved robustness under controlled distribution shifts.
Chinese Translation
视觉-语言-动作（VLA）模型将预训练的视觉-语言模型（VLMs）中的表示转换为机器人动作，但将中间VLM特征路由到动作解码器的接口仍然未被充分探索。现有设计要么仅暴露表示层次的一小部分，要么将每个解码器块严格匹配到一个VLM层，从而限制了在深度上获取互补任务证据的能力。我们提出了LIRA，一种局部跨层动作条件机制，将VLM到动作的条件化形式化为深度感知的信息路由。LIRA在任务标记特征和从中间VLM状态派生的LIRA查询特征上操作，然后为每个并行融合块分配一个与其对应VLM层对齐的局部窗口。并行融合块聚合相邻的LIRA查询特征，并在动作预测之前将其与任务标记特征和本体感知输入进行整合。该路由接口保持了主干架构、动作解码器和监督训练方案的不变。在LIBERO、LIBERO-Plus、CALVIN ABC→D和现实世界操作中，LIRA在相同的0.5B参数配置下提高了主要聚合指标，相较于VLA-Adapter基线。在零-shot迁移到LIBERO-Plus时，LIRA将平均成功率从59.1%提高到78.0%，增幅为18.9个百分点，表明在受控分布变化下的鲁棒性得到了改善。
cs.RO / 10 / 2608.07600
AdaDexGrasp: Adaptive Dexterous Grasping via 3D Visuo-Tactile Representation Fusion
AdaDexGrasp：通过 3D 视觉-触觉表示融合实现自适应灵巧抓取
Abstract
Humans achieve stable and adaptive grasps by seamlessly integrating visual perception and tactile feedback, a capability that remains challenging to replicate in robotic systems. Existing robotic grasping approaches predominantly rely on visual inputs and lack mechanisms for tactile-guided adaptation after contact, limiting robustness and generalization. To address this challenge, we propose a unified visuo-tactile-fusion grasping framework that integrates grasp generation, feasibility prediction, and adaptive refinement. At its core, our method introduces an efficient visuo-tactile representation that tightly fuses object geometry with tactile feedback by associating tactile signals with finger identities. This unified representation supports contact-aware grasp pose generation during planning and tactile-guided refinement after contact, enabling the system to reason about fine-grained finger-object interactions and adjust grasps dynamically. Comprehensive experiments in both simulation and real-world environments demonstrate that our approach significantly enhances grasp success rates and generalization across diverse objects.
Chinese Translation
人类通过无缝整合视觉感知和触觉反馈来实现稳定和自适应的抓取，这一能力在机器人系统中仍然难以复制。现有的机器人抓取方法主要依赖视觉输入，缺乏在接触后进行触觉引导适应的机制，从而限制了其鲁棒性和泛化能力。为了解决这一挑战，我们提出了一种统一的视觉-触觉融合抓取框架，该框架整合了抓取生成、可行性预测和自适应优化。我们的方法的核心是引入了一种高效的视觉-触觉表示，通过将触觉信号与手指身份关联，紧密融合物体几何形状与触觉反馈。这种统一的表示在规划过程中支持基于接触的抓取姿态生成，并在接触后进行触觉引导的优化，使系统能够推理细致的手指-物体交互并动态调整抓取。在模拟和真实环境中进行的全面实验表明，我们的方法显著提高了抓取成功率，并在多样化物体上增强了泛化能力。
cs.RO / 11 / 2608.07606
Enhanced Real-Time 6-DOF Extended Reality Catheter Tracking for Evaluating Potential Improvement in Efficiency, Precision, and Depth Perception for Cardiac Interventions
增强实时六自由度扩展现实导管追踪技术在评估心脏介入手术效率、精确度和深度感知潜在改善中的应用
Abstract
Despite advances in 3D ultrasound, most percutaneous cardiac interventions still rely on 2D visualization, limiting depth perception and spatial understanding. To address this challenge, we developed an Extended Reality (XR)-based platform that enables real-time six-degree-of-freedom (6-DOF) catheter tracking and visualization within a patient-specific 3D heart model. The system combines a custom machine-vision algorithm for 5-DOF catheter tracking with a 3D-printed electromechanical encoder that measures catheter roll, providing complete 6-DOF motion reconstruction. In a proof-of-concept study, 20 novice medical students navigated an intracardiac echocardiography (ICE) catheter to six anatomical targets using either immersive 3D visualization or a conventional 2D cathlab-style view. Participants in the 3D condition completed the task in 54.6 seconds and traveled 1,939 mm on average, compared with 267.5 seconds and 7,854 mm in the 2D condition. Therefore, the XR-based 3D system was more than 5x faster and required ~5x less catheter travel. The 3D mode also improved targeting precision and reduced performance variability. Participants consistently rated immersive visualization higher for accuracy, speed, usability, and clinical value. Kinematic analysis showed smoother depth-axis navigation in 3D, whereas 2D users relied on repeated corrective movements. These findings demonstrate that XR-based visualization can substantially improve procedural training efficiency, precision, and motor control.
Chinese Translation
尽管三维超声技术已有所进展，但大多数经皮心脏介入手术仍依赖于二维可视化，这限制了深度感知和空间理解。为了解决这一挑战，我们开发了一种基于扩展现实（XR）的平台，能够在患者特定的三维心脏模型中实现实时六自由度（6-DOF）导管追踪和可视化。该系统结合了一种定制的机器视觉算法用于5-DOF导管追踪，以及一个三维打印的电机编码器，用于测量导管的滚转，从而提供完整的6-DOF运动重建。在一项概念验证研究中，20名初学者医学学生使用沉浸式三维可视化或传统的二维导管实验室视图导航一根心腔内超声（ICE）导管至六个解剖目标。3D条件下的参与者平均完成任务时间为54.6秒，移动距离为1,939毫米，而2D条件下则为267.5秒和7,854毫米。因此，基于XR的三维系统速度超过5倍，且导管移动距离减少约5倍。三维模式还提高了目标精度并减少了表现变异性。参与者一致认为沉浸式可视化在准确性、速度、可用性和临床价值方面更高。运动学分析显示，3D用户在深度轴导航上更加平滑，而2D用户则依赖于重复的修正动作。这些发现表明，基于XR的可视化可以显著提高程序培训的效率、精确度和运动控制。
cs.RO / 12 / 2608.07619
GWM-VLA: Geometry-Aware Latent World Modeling for Vision-Language-Action Learning
GWM-VLA：一种面向几何的潜在世界建模框架用于视觉-语言-动作学习
Abstract
Vision-Language-Action (VLA) models achieve strong robotic manipulation performance but often degrade under visual and environmental shifts. Latent world modeling offers a promising approach to improving robustness, yet existing methods commonly encode camera views independently and predict holistic scene dynamics without explicitly modeling their geometric relationships. We propose GWM-VLA, a geometry-aware latent world modeling framework for VLA learning. GWM-VLA combines geometry-aware multi-view state encoding, global context-conditioned target-view prediction, and shared latent-action representations grounded by robot-action supervision. Specifically, VGGT-$\Omega$ jointly aggregates multi-view observations at each timestep to construct geometry-aware multi-view states. The latent world model predicts the next-step patch tokens of a selected target view using patch and register tokens obtained after multi-view aggregation, thereby retaining multi-view geometric information without predicting the complete multi-view state. We use the wrist view as the target in our experiments, placing greater emphasis on end-effector motion and local gripper-object interactions. Finally, the shared latent-action representations condition both the latent world model and the flow-matching action head, allowing latent-prediction supervision and ground-truth robot-action supervision to jointly shape the same latent-action representations. Experiments across both simulation and real-world environments demonstrate the effectiveness and robustness of GWM-VLA.
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操作性能上表现出色，但在视觉和环境变化下往往会出现性能下降。潜在世界建模为提高鲁棒性提供了一种有前景的方法，然而现有方法通常独立编码相机视角，并预测整体场景动态，而未明确建模其几何关系。我们提出了GWM-VLA，一种面向几何的潜在世界建模框架用于VLA学习。GWM-VLA结合了面向几何的多视角状态编码、基于全局上下文的目标视角预测，以及通过机器人动作监督所支撑的共享潜在动作表示。具体而言，VGGT-$ ext{Ω}$在每个时间步共同聚合多视角观测，以构建面向几何的多视角状态。潜在世界模型使用在多视角聚合后获得的补丁和注册令牌预测所选目标视角的下一步补丁令牌，从而保留多视角几何信息，而无需预测完整的多视角状态。在我们的实验中，我们将手腕视角作为目标，强调末端执行器的运动和局部夹具-物体交互。最后，共享潜在动作表示同时影响潜在世界模型和流匹配动作头，允许潜在预测监督和真实机器人动作监督共同塑造相同的潜在动作表示。在模拟和真实环境中的实验表明，GWM-VLA的有效性和鲁棒性。
cs.RO / 13 / 2608.07624
A Reconfigurable Tracked Robot for Enhanced Obstacle Traversal Through Movable Articulation Point and Internal Mass Relocation
一种可重构履带机器人，通过可移动关节点和内部质量重定位增强障碍物跨越能力
Abstract
Tracked robots are widely used in unstructured environments; however, their obstacle traversal capability is fundamentally limited by a tradeoff between front-end reachability and locomotion stability. This study presents TRASER (Tracked Robot with Articulated Spine for Extended Reach), a reconfigurable tracked robot capable of relocating both its articulation point and internal mass. TRASER employs a tape-spring mechanism that localizes compliance to the bending region while maintaining high stiffness in the remaining body, thereby improving both front-end reachability and center-of-mass (CoM) shifting capability. Geometric and static models are developed to analyze the effects of articulation point and CoM position on step and ditch traversal performances. Experiments demonstrate step traversal, suspended-platform traversal, and ditch traversal of 74\%, 66\%, and 59\% of the robot body length, respectively. To the best of our knowledge, these results represent the highest reported obstacle traversal capabilities among tracked mobile robots.
Chinese Translation
履带机器人广泛应用于非结构化环境；然而，它们的障碍物跨越能力在前端可达性与运动稳定性之间存在根本性权衡。本研究提出了TRASER（具有关节脊的履带机器人以扩展可达性），这是一种可重构的履带机器人，能够重新定位其关节点和内部质量。TRASER采用了一种带弹簧机制，将柔性局限于弯曲区域，同时在其余部分保持高刚度，从而提高了前端可达性和质心（CoM）移动能力。我们开发了几何和静态模型，以分析关节点和质心位置对跨越台阶和沟渠性能的影响。实验结果表明，机器人分别能够跨越其身体长度的74%、66%和59%的台阶、悬挂平台和沟渠。根据我们所知，这些结果代表了履带移动机器人中报告的最高障碍物跨越能力。
cs.RO / 14 / 2608.07707
H\"older Signed Distance: A Differentiable, Signed, Parallelizable Metric for Robotics
霍尔德签名距离：一种可微分的、带符号的、可并行化的机器人度量
Abstract
Computing distances between sets is essential in robotic motion planning and control, where differentiable gradients enable real-time optimization. The Euclidean Signed Distance Function (SDF), however, is not differentiable everywhere, and existing alternatives often sacrifice differentiability, sign information, or computational efficiency. In this letter, we introduce a novel differentiable signed distance between convex polyhedra. To this end, we first propose differentiable versions of the minimum and maximum operators, termed the H\"older minimum and H\"older maximum. We then replace the original min-max operators in the classical SDF formulation, yielding the H\"older signed distance. Unlike prior differentiable distance formulations that rely on iterative algorithms, our approach is computed in closed form, eliminating convergence issues while remaining naturally amenable to GPU parallelization. We validate the practical advantages and computational performance of the proposed distance through runtime comparisons with existing approaches. We also present a robotic manipulator experiment, demonstrating its suitability for applications in control.
Chinese Translation
在机器人运动规划和控制中，计算集合之间的距离至关重要，其中可微分梯度使实时优化成为可能。然而，欧几里得签名距离函数（SDF）并非在所有地方都可微，而现有的替代方案往往牺牲了可微性、符号信息或计算效率。在本文中，我们提出了一种新的可微分签名距离，用于凸多面体。为此，我们首先提出了最小和最大算子的可微分版本，称为霍尔德最小值和霍尔德最大值。然后，我们在经典SDF公式中替换原始的最小-最大算子，从而得出霍尔德签名距离。与依赖于迭代算法的先前可微分距离公式不同，我们的方法以闭式形式计算，消除了收敛问题，同时自然适合于GPU并行化。我们通过与现有方法的运行时比较验证了所提距离的实际优势和计算性能。我们还展示了一个机器人操控实验，证明了其在控制应用中的适用性。
cs.RO / 15 / 2608.07734
Complete, Scalable, and Robust Prioritized Planning for Multi-Robot Ordered Storage and Retrieval at Maximum Capacity
完整、可扩展且稳健的多机器人有序存储与检索的优先规划在最大容量下
Abstract
Automated warehouses face a fundamental trade-off between maximizing storage density and achieving high retrieval throughput. While puzzle-based storage (PBS) architectures increase capacity by eliminating aisles, coordinating multiple robots in these high-density spaces is computationally challenging due to the potential for deadlocks. This paper introduces a novel multi-robot formulation for the ``ordered storage and retrieval problem at maximum capacity''. The focus is on rectangular grids accessible from a single boundary, where loads need to be first stored up to full capacity and then efficiently retrieved, given a planned departure sequence. This work bridges the gap between geometric feasibility and execution efficiency by leveraging the properties of relocation-free arrangements. These properties guide an online, prioritized multi-agent path-finding algorithm, which is the main contribution of this work. Unlike general centralized planners, the approach exploits the specific invariants of the storage arrangement to guarantee completeness and prevent deadlocks, enabling scalability. Experiments demonstrate that the method achieves near-linear improvement in makespan with respect to the number of robots, up to $m = C$, where $C$ is the grid width. Crucially, the algorithmic overhead of supporting robustness is negligible; the system handles uncertainty in departure sequences using robust storage arrangements with no significant penalty in execution speed compared to the non-robust baseline.
Chinese Translation
自动化仓库面临着在最大化存储密度与实现高检索吞吐量之间的基本权衡。尽管基于拼图的存储（PBS）架构通过消除过道来增加容量，但在这些高密度空间中协调多个机器人由于潜在的死锁而在计算上具有挑战性。本文提出了一种新颖的多机器人“最大容量下的有序存储与检索问题”的表述。研究重点是从单一边界可访问的矩形网格，在这些网格中，负载需要首先存储到满容量，然后根据计划的出发顺序高效检索。本研究通过利用无重定位排列的特性，弥合了几何可行性与执行效率之间的差距。这些特性指导了一种在线优先多智能体路径规划算法，这是本工作的主要贡献。与一般的集中式规划者不同，该方法利用存储排列的特定不变性来保证完整性并防止死锁，从而实现可扩展性。实验表明，该方法在与机器人数量相关的完工时间上实现了近线性改善，最高可达 $m = C$，其中 $C$ 是网格宽度。重要的是，支持稳健性的算法开销微乎其微；该系统使用稳健的存储排列处理出发顺序中的不确定性，与非稳健基线相比，执行速度没有显著的惩罚。
cs.RO / 16 / 2608.07740
Enhancing Autonomous Vehicle Navigation with a Clothoid-Based Lateral Controller
基于克洛索伊德的侧向控制器增强自主车辆导航
Abstract
This study introduces an advanced lateral control strategy for autonomous vehicles using a clothoid-based approach integrated with an adaptive lookahead mechanism. The primary focus is on enhancing lateral stability and path-tracking accuracy through the application of Euler spirals for smooth curvature transitions, thereby reducing passenger discomfort and the risk of vehicle rollover. An innovative aspect of our work is the adaptive adjustment of lookahead distance based on real-time vehicle dynamics and road geometry, which ensures optimal path following under varying conditions. A quasi-feedback control algorithm constructs optimal clothoids at each time step, generating the appropriate steering input. A lead filter compensates for the vehicle's lateral dynamics lag, improving control responsiveness and stability. The effectiveness of the proposed controller is validated through a comprehensive co-simulation using TruckSim and Simulink, demonstrating significant improvements in lateral control performance across diverse driving scenarios. Future directions include scaling the controller for higher-speed applications and further optimization to minimize off-track errors, particularly for articulated vehicles.
Chinese Translation
本研究提出了一种先进的侧向控制策略，针对自主车辆采用基于克洛索伊德的方法，并结合自适应前视机制。主要关注通过应用欧拉螺旋实现平滑的曲率过渡，以增强侧向稳定性和路径跟踪精度，从而减少乘客的不适感和车辆翻覆的风险。我们工作的一个创新之处在于根据实时车辆动态和道路几何进行前视距离的自适应调整，以确保在不同条件下的最佳路径跟随。一个准反馈控制算法在每个时间步构建最优的克洛索伊德，生成适当的转向输入。前导滤波器补偿车辆侧向动态滞后，提高控制响应性和稳定性。通过使用TruckSim和Simulink进行全面的共同仿真，验证了所提控制器的有效性，显示出在多种驾驶场景下侧向控制性能的显著改善。未来的方向包括将控制器扩展到更高速度的应用，并进一步优化以最小化偏离轨迹的误差，特别是针对关节式车辆。
cs.RO / 17 / 2608.07751
CoCoNav: Conformal Control for Safe Robot Navigation in Crowds
CoCoNav：用于人群中安全机器人导航的保形控制
Abstract
Safe and efficient robot navigation in crowds requires anticipating pedestrian motion despite uncertain and potentially shifting prediction errors. Existing reactive methods can produce oscillatory behavior, while predictive planners often treat forecasts as exact or rely on restrictive error models. Incorporating conservative uncertainty sets as hard constraints can also render model predictive control (MPC) infeasible. We propose \textit{CoCoNav}, a crowd-navigation framework that combines online conformal calibration with runtime-certified planning. A horizon-specific conformal proportional--integral controller adapts trajectory-error bounds to regulate long-run empirical coverage, enabling the framework to respond to changing prediction errors. A \textit{relax-then-verify} planner preserves solver feasibility by generating nominal trajectories with soft-constrained MPC and separately certifying them, together with contingency maneuvers, against the calibrated bounds before execution. Simulations and quadruped experiments show that CoCoNav achieves a favorable balance among collision avoidance, task success, and navigation efficiency relative to the evaluated baselines.
Chinese Translation
在拥挤环境中安全高效的机器人导航需要预测行人运动，尽管存在不确定性和潜在的预测误差变化。现有的反应式方法可能会产生振荡行为，而预测规划者往往将预测视为精确的，或依赖于限制性的误差模型。将保守的不确定性集作为硬约束纳入也可能使模型预测控制（MPC）变得不可行。我们提出了 extit{CoCoNav}，一个结合了在线保形校准与运行时认证规划的人群导航框架。一个特定于时间范围的保形比例-积分控制器调整轨迹误差界限，以调节长期经验覆盖率，使框架能够响应变化的预测误差。一个 extit{先放宽再验证}的规划器通过生成具有软约束的MPC的名义轨迹，并在执行前分别对其及应急机动进行认证，以保持求解器的可行性。仿真和四足动物实验表明，CoCoNav在碰撞避免、任务成功和导航效率之间实现了相对于评估基线的良好平衡。
cs.RO / 18 / 2608.07797
Drone-Assisted UAV-UGV Collaboration for Autonomous Navigation in Snow-Covered Terrain
无人机辅助的无人机-无人地面车辆协作在雪覆盖地形中的自主导航
Abstract
This paper presents a collaborative UAV-UGV navigation framework for high-altitude, snow-covered terrain, where reduced visibility and unstable ground render conventional methods ineffective. We introduce a custom efficient U-Net architecture that falls under the computational constraints for real-time road segmentation, utilizing a novel synthetic snow data augmentation technique to achieve 96.5% segmentation accuracy. For UAV localization, we implement an Extended Kalman Filter (EKF) fusing onboard GPS and IMU data, achieving a maximum observed positional error of +-0.5 meters. The UGV position is determined via a visual tracking pipeline using YOLOv5 and depth data from the UAV's RGB-D camera. A dynamic path planning algorithm utilizes this segmentation to adjust for snow drifts, enabling successful navigation in obscured test environment with minimal deviation.
Chinese Translation
本文提出了一种针对高海拔雪覆盖地形的无人机-无人地面车辆（UAV-UGV）协作导航框架，在这种环境中，能见度降低和地面不稳定使得传统方法无效。我们引入了一种自定义的高效U-Net架构，符合实时道路分割的计算约束，利用一种新颖的合成雪数据增强技术实现了96.5%的分割准确率。对于无人机定位，我们实现了一种扩展卡尔曼滤波器（Extended Kalman Filter, EKF），融合了机载GPS和IMU数据，最大观测位置误差为±0.5米。无人地面车辆的位置通过使用YOLOv5的视觉跟踪管道和无人机RGB-D相机的深度数据来确定。动态路径规划算法利用该分割结果调整雪堆，从而在受阻的测试环境中实现成功导航，偏差最小。
cs.RO / 19 / 2608.07887
A Mixed-Stiffness Anthropomimetic Fingertip Broadens the Operating Range for Coin Grasping
一种混合刚度的人体仿生指尖拓宽了硬币抓取的操作范围
Abstract
Robotic grasping of thin, flat objects such as coins on hard surfaces remains challenging because conventional methods require reorienting the object, accessing its underside, or adding a dedicated nail mechanism. We previously showed that a rigid nail arrests soft-pad deformation and thereby forms a geometric constraint that improves precision grasping. Here we asked whether an additional constraint-forming boundary, created within the pad by material choice rather than by anatomy, could extend the conditions under which that constraint holds. We fabricated anthropomimetic fingertips with Shore E10 silicone at the center and Shore A60 at the sides, and compared them with uniformly soft E10 fingertips. An automated apparatus performed an oblique rotational tip pinch in which the pad engaged the coin's lateral surface, lifting it from flush contact with no gap beneath it. Over variations in horizontal approach distances, vertical finger displacements, and index-finger rotation, the mixed-stiffness pair maintained high success rates across more tested settings than the uniform pair during both geometric-constraint formation and the transition to a stable grasp. The nail-free pair failed in all 36 conditions of Experiment 1-1. However, the uniform pair performed better when coin position along the finger axis was varied, a condition-dependent trade-off. After tuning for coin size, both fingertip types grasped all six Japanese denominations. These results suggest that the operating range for thin-object grasping depends not only on pad softness but also on where stiffness is placed within a nail-supported pad, making boundary placement a candidate fingertip design variable.
Chinese Translation
在硬表面上抓取薄而平坦的物体（如硬币）仍然具有挑战性，因为传统方法需要重新定位物体、接触其底面或添加专用的钉机制。我们之前展示了刚性钉子可以阻止软垫变形，从而形成几何约束，提高抓取精度。在此，我们探讨了是否可以通过材料选择而非解剖结构，在垫内创建额外的约束形成边界，以扩展该约束适用的条件。我们制造了中心为Shore E10硅胶、两侧为Shore A60硅胶的人体仿生指尖，并将其与均匀软的E10指尖进行了比较。一个自动化设备执行了一个斜向旋转的尖端夹持，其中垫与硬币的侧面接触，将其从平面接触中抬起，底部没有间隙。在水平接近距离、垂直手指位移和食指旋转的变化下，混合刚度的指尖在几何约束形成和稳定抓取过渡的更多测试设置中保持了比均匀指尖更高的成功率。无钉指尖在实验1-1的36种条件下均未成功。然而，当沿着手指轴的硬币位置变化时，均匀指尖表现更好，这是一个条件依赖的权衡。在调整硬币大小后，两种指尖类型均成功抓取了所有六种日本货币。这些结果表明，薄物体抓取的操作范围不仅依赖于垫的柔软度，还与在支持钉的垫内刚度的放置位置有关，使得边界位置成为潜在的指尖设计变量。
cs.RO / 20 / 2608.07895
Auditing Instruction-Trajectory Mismatches in Multimodal Robot Demonstrations
审计多模态机器人演示中的指令-轨迹不匹配
Abstract
Robot demonstration datasets used to train vision-language-action policies can contain a subtle but harmful failure mode: trajectories that are behaviorally correct but paired with the wrong language instruction. We study post-hoc auditing of these Instruction-Trajectory Mismatches (ITMs). Unlike failed rollouts, ITMs often look plausible, and can corrupt the language-behavior mapping learned by the policy. We propose Multimodal Probabilistic Fusion (MMPF), a training-free auditing framework that treats each modality as an expert, estimates a task-label distribution from local neighborhood agreement and global prototype similarity, and then fuses modalities with predictive-entropy weighting in a product of experts. Across LIBERO benchmarks with injected instruction mismatches and noisy real-robot data, MMPF achieves the strongest overall ITM detection and label correction accuracy. We also show that auditing improves most downstream policy learning in settings where language is needed to disambiguate the task. We demonstrate in real robot experiments that our method can achieve improved policy performance and show the trade-off of filtering demonstrations compared to relabeling.
Chinese Translation
用于训练视觉-语言-动作策略的机器人演示数据集可能包含一种微妙但有害的失败模式：行为上正确但与错误语言指令配对的轨迹。我们研究了这些指令-轨迹不匹配（Instruction-Trajectory Mismatches, ITMs）的事后审计。与失败的回滚不同，ITMs通常看起来是合理的，并且可能会破坏策略学习到的语言-行为映射。我们提出了多模态概率融合（Multimodal Probabilistic Fusion, MMPF），这是一种无训练的审计框架，将每种模态视为专家，从局部邻域一致性和全局原型相似性中估计任务标签分布，然后在专家乘积中使用预测熵加权融合模态。在注入指令不匹配和嘈杂真实机器人数据的LIBERO基准测试中，MMPF实现了最强的ITM检测和标签修正准确性。我们还表明，在需要语言来消歧任务的设置中，审计改善了大多数下游策略学习。我们在真实机器人实验中演示了我们的方法能够实现改进的策略性能，并展示了过滤演示与重新标记之间的权衡。
cs.RO / 21 / 2608.08023
4D-WAM: Infusing Spatiotemporal Awareness into World Action Models through Trajectory Fields
4D-WAM：通过轨迹场将时空意识注入世界行动模型
Abstract
Building on recent advances in world models, World Action Models (WAMs) jointly model video prediction and action generation. However, they typically represent videos in 2D pixel space, creating a representation gap with 3D space in which robotic actions are executed. Recent 3D approaches introduce 3D information, but fail to fully exploit the dynamics of 3D structures. In this work, we propose 4D-WAM, a model-agnostic training strategy that injects spatiotemporal knowledge from 3D trajectory fields into WAMs through representation alignment. To this end, we introduce two complementary objectives: 1) motion alignment, which aligns temporal feature variations across adjacent frames and encourages the model to build local 4D awareness during training, and 2) destination alignment, which guides the model to infer the final destination from the source frame by minimizing the gap between their attention-like similarity distributions. Together, these objectives provide both local motion supervision and long-horizon goal guidance, enabling WAMs to learn trajectory-level spatiotemporal representations. Extensive in-distribution and out-of-distribution experiments across different base models demonstrate the model's improvements in spatial understanding, execution precision, robustness, generalization, and versatility.
Chinese Translation
基于近期在世界模型方面的进展，世界行动模型（WAMs）共同建模视频预测和动作生成。然而，它们通常在二维像素空间中表示视频，这与机器人执行动作的三维空间之间存在表示差距。近期的三维方法引入了三维信息，但未能充分利用三维结构的动态特性。在本研究中，我们提出了4D-WAM，一种模型无关的训练策略，通过表示对齐将来自三维轨迹场的时空知识注入WAMs。为此，我们引入了两个互补的目标：1）运动对齐，旨在对齐相邻帧之间的时间特征变化，并鼓励模型在训练过程中建立局部四维意识；2）目标对齐，指导模型通过最小化源帧与最终目标之间的注意力相似性分布差距来推断最终目标。综合这两个目标，提供了局部运动监督和长远目标指导，使WAMs能够学习轨迹级的时空表示。在不同基础模型上的广泛分布内和分布外实验表明，该模型在空间理解、执行精度、鲁棒性、泛化能力和多样性方面的改进。
cs.RO / 22 / 2608.08025
DA-NBV: A Direction-Aware Next-Best-View Planner for Efficient 3D Reconstruction of Ships at Sea
DA-NBV：一种方向感知的下一最佳视角规划器，用于高效的海上船舶三维重建
Abstract
Accurate 3D reconstruction of ships at sea is important for maritime supervision, damage assessment, and autonomous maritime operations. Although 3D reconstruction has advanced considerably, high-quality data acquisition still largely relies on manually designed trajectories or skilled operators, resulting in high costs and limited scalability. Next-best-view (NBV) planning automates this process by selecting subsequent viewpoints based on the current state. However, existing NBV policies mainly model spatial occupancy while overlooking directional observation history. This limitation is particularly problematic for ships: their complex superstructures and severe self-occlusions require observations from multiple viewpoints, and insufficient directional coverage often yields incomplete reconstructions. These challenges are further amplified at sea, where wave-induced heave, roll, and pitch continuously alter the ship's pose and surface visibility. Meanwhile, wind disturbances and limited onboard power impose stricter requirements on scanning efficiency. To address these challenges, we propose DA-NBV, a direction-aware NBV policy that augments the conventional occupancy state with directional observation statistics. We introduce a learnable Position Advantage Field (PAF) that uses directional information to guide viewpoint selection. The policy further adopts a locally constrained action space and a nonlinear coverage-shaping reward to improve scanning efficiency. We also develop the ship-oriented SeaShip-3D dataset and a configurable sea-state simulation environment. Experiments under varying heave, roll, and pitch conditions show that DA-NBV improves reconstruction completeness by approximately 3 percentage points and reduces Chamfer distance by 43% while achieving higher path efficiency.
Chinese Translation
海上船舶的准确三维重建对于海事监督、损伤评估和自主海上操作至关重要。尽管三维重建技术已经取得了显著进展，但高质量数据的获取仍然在很大程度上依赖于手动设计的轨迹或熟练操作员，这导致了高成本和有限的可扩展性。下一最佳视角（NBV）规划通过根据当前状态选择后续视角来自动化这一过程。然而，现有的NBV策略主要建模空间占用，而忽视了方向观察历史。这一局限性在船舶中尤为突出：它们复杂的上部结构和严重的自遮挡需要从多个视角进行观察，而不足的方向覆盖往往导致重建不完整。这些挑战在海上进一步加剧，波浪引起的浮动、滚动和俯仰不断改变船舶的姿态和表面可见性。同时，风扰动和有限的船上电力对扫描效率提出了更严格的要求。为了解决这些挑战，我们提出了DA-NBV，一种方向感知的NBV策略，通过方向观察统计信息增强传统的占用状态。我们引入了一种可学习的位置优势场（Position Advantage Field, PAF），利用方向信息指导视角选择。该策略进一步采用局部约束的动作空间和非线性覆盖塑形奖励，以提高扫描效率。我们还开发了以船舶为导向的SeaShip-3D数据集和可配置的海况模拟环境。在不同的浮动、滚动和俯仰条件下的实验表明，DA-NBV将重建完整性提高了约3个百分点，并将Chamfer距离减少了43%，同时实现了更高的路径效率。
cs.RO / 23 / 2608.08036
Compiling and Benchmarking Task-State Horizons for Embodied Agents
为具身智能体编译和基准测试任务状态视野
Abstract
Frontier agentic models are increasingly deployed as high-level planners for long-horizon embodied tasks. Existing robotic benchmarks have advanced long-horizon evaluation, but primarily characterize difficulty through action-sequence length and subtask complexity, overlooking a distinct challenge: agents must track evolving task-relevant world states induced by both their exploration and environmental dynamics. We define the span of task-relevant state transitions that an agent must track as task-state horizon (TSH). To evaluate how agent performance varies with TSH, we introduce RoboGraph, a robotic task compiler that translates state-transition dependencies into executable symbolic graphs. Specifically, RoboGraph constructs task-state horizons from spatial and temporal causal dependencies, including those induced by unexpected failures and interventions during task execution. Building on RoboGraph, we release a benchmark comprising 588 episodes across 84 scenes with varying TSHs. Experiments evaluating 15 advanced agentic models in both semantic and visual closed-loop environments show that most models struggle with demanding TSHs, revealing substantial gaps in maintaining, exploring, and updating task-relevant state over long horizon.
Chinese Translation
前沿智能模型越来越多地被用作长时间跨度具身任务的高层次规划者。现有的机器人基准测试已推动了长时间跨度评估的发展，但主要通过动作序列长度和子任务复杂性来表征难度，忽视了一个独特的挑战：智能体必须跟踪由其探索和环境动态引起的不断变化的任务相关世界状态。我们将智能体必须跟踪的任务相关状态转变的范围定义为任务状态视野（Task-State Horizon, TSH）。为了评估智能体性能如何随TSH变化，我们引入了RoboGraph，这是一种机器人任务编译器，可以将状态转移依赖关系转换为可执行的符号图。具体而言，RoboGraph根据空间和时间因果依赖关系构建任务状态视野，包括在任务执行过程中由于意外故障和干预引起的依赖关系。在RoboGraph的基础上，我们发布了一个基准测试，包含588个场景中的84个场景，具有不同的TSH。对15种先进智能模型在语义和视觉闭环环境中的实验评估表明，大多数模型在面对苛刻的TSH时表现不佳，揭示了在长时间跨度内维护、探索和更新任务相关状态方面存在显著差距。
cs.RO / 24 / 2608.08053
PhysX-CoT: Structured Physical Reasoning from a Single Image to Simulation-Ready 3D Assets
PhysX-CoT：从单幅图像到可用于仿真的结构化物理推理生成3D资产
Abstract
Simulation-ready 3D assets are central to robotics and embodied AI. Generating them from a single image is usually framed as a vision-language model that emits a serialized asset for a decoder to turn into geometry and physical fields, leaving the image-to-3D reasoning implicit. We argue the limiting factor is this output-centric view: part placement and local shape are entangled in one global-coordinate token stream, and the intermediate physical states are never exposed for supervision, conditioning, or verification. PhysX-CoT instead casts single-image asset generation as an explicit structured physical reasoning process, an ordered and machine-parseable trajectory of part-level states covering decomposition, 2D and 3D grounding, relations, coarse geometry, and surface cues that we separately supervise, use to condition geometry, and treat as reward targets. Geometry is factorized so that 3D boxes carry placement and local codes carry shape, and CoT-aligned GRPO optimizes parse validity, grounding, geometry, placement, and physical consistency. Under a unified protocol that retrains all learned baselines on the same backbone, data, and frozen decoder, PhysX-CoT outperforms the closest full-task baseline across geometry, scale, and physical-attribute metrics. Oracle, token-matched, and state-order controls show the explicit states are functional rather than cosmetic, and in Unreal Engine~5 the generated assets parse, collide, and articulate at high validity.
Chinese Translation
可用于仿真的3D资产在机器人技术和具身人工智能中至关重要。从单幅图像生成这些资产通常被视为一种视觉-语言模型，该模型输出一个序列化的资产，以供解码器转换为几何形状和物理场，导致图像到3D的推理过程隐含。我们认为限制因素在于这种以输出为中心的视角：部件放置和局部形状在一个全局坐标的标记流中交织在一起，而中间物理状态从未暴露以供监督、调节或验证。相反，PhysX-CoT将单幅图像资产生成视为一个明确的结构化物理推理过程，这是一个有序且机器可解析的部件级状态轨迹，涵盖了分解、2D和3D基础、关系、粗略几何形状和表面线索，我们对其进行单独监督，利用其来调节几何形状，并将其视为奖励目标。几何形状被因式分解，使得3D框承载放置信息，而局部代码承载形状信息，CoT对齐的GRPO优化了解析有效性、基础、几何形状、放置和物理一致性。在一个统一的协议下，所有学习的基线在相同的骨干网络、数据和冻结的解码器上进行再训练，PhysX-CoT在几何形状、尺度和物理属性指标上超越了最接近的全任务基线。Oracle、标记匹配和状态顺序控制显示，显式状态是功能性的而非表面的，在虚幻引擎5中生成的资产具有高有效性，能够解析、碰撞和运动。
cs.RO / 25 / 2608.08070
SurgWMBench: A Vision-Based Benchmark for World-Modeling Surgical Instrument Motion Planning
SurgWMBench：基于视觉的世界建模外科器械运动规划基准
Abstract
Reliable surgical planning requires models that move beyond recognizing the current surgical step or imitating expert demonstrations, and instead anticipate how instrument motion reshapes subsequent operative states. Most surgical video understanding methods focus on recognizing phases, actions, or workflow states, while providing limited support for explicitly modeling instrument motion. Conversely, existing tool motion prediction methods can forecast instrument trajectories, but they generally do not capture the coupled evolution of future surgical video states. World models offer a natural framework for jointly modeling visual state transitions and instrument motion dynamics. However, existing surgical world model studies remain largely centered on visual generation quality, relying on generation-oriented metrics such as FVD and CD-FVD. These metrics are poorly aligned with instrument motion planning, as they do not directly measure whether predicted trajectories are geometrically accurate, temporally coherent, or actionable for downstream planning. This limitation is partly structural, since the field lacks public datasets and standardized evaluation protocols that provide the benchmarking infrastructure needed to assess motion-centric capabilities in surgical world models. In this paper, we introduce SurgWMBench, a vision-based benchmark for short-horizon surgical motion planning and dynamics prediction. Given intraoperative image sequences and historical instrument trajectory, SurgWMBench evaluates both near-future instrument motion prediction and stability under continuous rollout or input perturbations.
Chinese Translation
可靠的外科规划需要超越对当前外科步骤的识别或模仿专家演示，而是要预测器械运动如何重塑后续的手术状态。大多数外科视频理解方法专注于识别阶段、动作或工作流程状态，而对显式建模器械运动的支持有限。相反，现有的工具运动预测方法可以预测器械轨迹，但通常无法捕捉未来外科视频状态的耦合演变。世界模型提供了一个自然的框架，用于联合建模视觉状态转变和器械运动动态。然而，现有的外科世界模型研究仍然主要集中在视觉生成质量上，依赖于诸如FVD和CD-FVD等生成导向的指标。这些指标与器械运动规划的对齐程度较差，因为它们并未直接测量预测轨迹在几何上是否准确、时间上是否一致或是否可用于后续规划。这一局限性在一定程度上是结构性的，因为该领域缺乏公共数据集和标准化评估协议，无法提供评估外科世界模型中运动中心能力所需的基准基础设施。在本文中，我们介绍了SurgWMBench，一个用于短期外科运动规划和动态预测的基于视觉的基准。SurgWMBench基于术中图像序列和历史器械轨迹，评估近未来器械运动预测的准确性及其在连续展开或输入扰动下的稳定性。
cs.RO / 26 / 2608.08157
Event-Time Hybrid Optimal Control for Robotic Table Tennis Serves
机器人乒乓球发球的事件时间混合最优控制
Abstract
Robotic table tennis serves require high ball velocity and spin while respecting the robot's kinodynamic limits. Unlike rally strokes, a valid serve must also bounce on the server's side and clear the net, yielding a hybrid system with nonlinear flight and impact dynamics. We formulate spin-controlled serve generation as an event-time \ac{OCP} that optimizes the racket impact velocity and orientation together with the bounce, net-crossing, and landing times, enabling direct enforcement at phase boundaries of table-bounce and net-clearance constraints. The racket velocity and orientation are then converted into a complete kinodynamically feasible motion through a second \ac{OCP} enforcing joint-position, velocity, and torque limits. We evaluate the method numerically and on a KUKA Agilus robot. Compared with a fixed-step formulation with root localization, the proposed event-time formulation reduces the median solve time by a factor of 4.1 while maintaining comparable landing accuracy, spin accuracy, and serve validity. Real-robot experiments demonstrate controlled placement and topspin, backspin, and sidespin serves, with a mean landing error of $13.1 \pm 7.3$~cm and spin rates up to 30~rps. These results show that event-time optimal control efficiently generates physically valid, kinodynamically feasible serves while accounting for nonlinear aerodynamic and impact effects.
Chinese Translation
机器人乒乓球发球需要高球速和旋转，同时遵循机器人的运动动力学限制。与对打击球不同，有效的发球必须在发球方的场地上反弹并越过球网，从而形成一个具有非线性飞行和冲击动力学的混合系统。我们将旋转控制的发球生成形式化为一个事件时间最优控制问题（event-time OCP），该问题优化了球拍的冲击速度和方向，以及反弹、越网和落地时间，从而能够在乒乓球反弹和越网约束的相位边界上直接执行约束。然后，通过第二个最优控制问题（OCP），将球拍的速度和方向转换为完全符合运动动力学的可行运动，同时强制执行关节位置、速度和扭矩限制。我们在数值上和KUKA Agilus机器人上评估了该方法。与具有根定位的固定步长公式相比，所提出的事件时间公式将中位数求解时间减少了4.1倍，同时保持了相当的落地精度、旋转精度和发球有效性。真实机器人实验展示了受控的落点以及上旋、下旋和侧旋发球，平均落地误差为$13.1 imes 7.3$~cm，旋转速率高达30~rps。这些结果表明，事件时间最优控制能够有效生成物理有效、运动动力学可行的发球，同时考虑了非线性气动和冲击效应。
cs.RO / 27 / 2608.08183
Multi-modal Interactive Control of Robotic Arm based on Offline Large Language Models
基于离线大型语言模型的多模态交互控制机器人手臂
Abstract
Large Language Models (LLMs) have significantly revolutionized the modern society with numerous advanced interactions between humans and AI agents, whereas the usage of most large language models including ChatGPT are not friendly open-sourced and must require the users paying a lot for such AI services continuously. Therefore, deploying open-sourced large language models on local servers can be considered as an efficient approach to design and implement creative embodied AI algorithms with lower cost and more stable free usage. Inspired by this ordinary motivation, we originally propose and implement the "Socratic Models-ChatGLM", which is a well-performed algorithm for multi-modal interactive control of robotic arm based on offline large language models via the facile PyBullet platform, even presents extraordinary potential to address complicated text-image integrated multi-step long-horizon robotic manipulation tasks.
Chinese Translation
大型语言模型（LLMs）显著改变了现代社会，促进了人类与人工智能代理之间的多种先进交互。然而，绝大多数大型语言模型，包括ChatGPT，并不友好开源，并且用户需要持续支付高额费用以使用这些人工智能服务。因此，在本地服务器上部署开源大型语言模型被视为一种高效的方法，可以以更低的成本和更稳定的免费使用来设计和实现创造性的具身人工智能算法。受到这一普通动机的启发，我们首次提出并实现了“Socratic Models-ChatGLM”，这是一种基于离线大型语言模型的多模态交互控制机器人手臂的高效算法，利用简单的PyBullet平台，甚至展现出解决复杂文本-图像集成的多步骤长时间机器人操作任务的非凡潜力。
cs.RO / 28 / 2608.08200
Spatiotemporal Context-dependent Personalized Movement Compensation in Delayed Telemanipulation
时空上下文依赖的个性化运动补偿在延迟遥操作中的应用
Abstract
Communication delay remains a central challenge in telerobotics, where it disrupts visuomotor coordination and reduces task precision. Motion scaling is an effective countermeasure to delay-induced overshoot, yet typical deployments rely on uniform gains that neglect individual and contextual variability. We propose a human-centered method that fits personalized delay-, direction-, and distance-specific scaling parameters for each participant. We conducted experiments with twenty participants who performed delayed reaching tasks in a virtual simulator. Scaling gains were computed to minimize mean overshoot in simulation in each combination of experimental conditions. Evaluation was done in simulation and on a telesurgical robot to evaluate assistance benefits. Performance was assessed across multiple delays, distances, and movement directions using overshoot, endpoint error, trajectory smoothness, economy of motion, and a composite error-time metric. Motion scaling consistently improved performance relative to unassisted trials, yielding up to 20-25% performance gains in key metrics. Effects were most pronounced at longer delays. Personalization demonstrated additional accuracy benefits for inward reaching at a short distance under moderate delay. The results highlight the potential of personalized scaling as a foundation for more adaptive frameworks that integrate contextual information to improve the safety and precision of teleoperated procedures.
Chinese Translation
通信延迟仍然是远程机器人技术中的一个核心挑战，它破坏了视觉运动协调并降低了任务精度。运动缩放是一种有效的应对延迟引起的超调的对策，然而典型的应用依赖于统一的增益，忽视了个体和上下文的变异性。我们提出了一种以人为中心的方法，为每位参与者拟合个性化的延迟、方向和距离特定的缩放参数。我们对二十名参与者进行了实验，他们在虚拟模拟器中执行延迟到达任务。计算缩放增益以最小化每种实验条件组合下的平均超调。评估在模拟和遥外科机器人上进行，以评估辅助的益处。通过超调、终点误差、轨迹平滑度、运动经济性和复合误差时间指标评估在多种延迟、距离和运动方向下的表现。与无辅助试验相比，运动缩放始终提高了表现，在关键指标上可实现高达20-25%的性能提升。效果在较长延迟下最为明显。在中等延迟下，个性化显示出在短距离内向内到达时的额外精度益处。结果突显了个性化缩放作为更具适应性的框架基础的潜力，该框架整合上下文信息以提高远程操作程序的安全性和精度。
cs.RO / 29 / 2608.08273
Action- and Language-Conditioned Video Assessment for Embodied Control
基于动作和语言条件的视频评估用于具身控制
Abstract
Vision-based embodied agents executing multi-step natural language instructions require feedback mechanisms that assess task progress over complete trajectories. Conventional approaches based on final-frame matching or continuous embedding similarity may overlook intermediate transitions that are necessary for determining whether an instruction has been completed. We propose ALVA (Action- and Language-Conditioned Video Assessment), a trajectory evaluator that conditions its assessment on visual observations, the executed action sequence, and the natural language instruction. The method uses a pre-trained vision-language model (VLM) in two stages: it first summarizes frame-to-frame visual transitions conditioned on the executed actions and then assesses the generated summary with respect to the instruction to produce a discrete trajectory-level progress score. In simulated 3D household environments, ALVA exhibits a conservative assessment pattern with near-zero false-positive rates. When used as terminal feedback for closed-loop policy optimization, it provides more effective feedback than the evaluated static image and embedding-based visual baselines and reduces the performance gap to a ground-truth oracle. These results support action- and language-conditioned video assessment as an interpretable feedback mechanism for the evaluated simulated embodied-control tasks.
Chinese Translation
基于视觉的具身代理执行多步骤自然语言指令需要反馈机制，以评估任务在完整轨迹上的进展。基于最终帧匹配或连续嵌入相似性的传统方法可能忽视了判断指令是否完成所必需的中间过渡。我们提出了 ALVA（基于动作和语言条件的视频评估），这是一种轨迹评估器，其评估基于视觉观察、执行的动作序列和自然语言指令。该方法在两个阶段中使用预训练的视觉-语言模型（VLM）：首先总结基于执行动作的帧间视觉过渡，然后根据指令评估生成的摘要，以产生离散的轨迹级进展评分。在模拟的三维家庭环境中，ALVA 展现出保守的评估模式，几乎没有假阳性率。当作为闭环策略优化的终端反馈使用时，它提供的反馈比评估的静态图像和基于嵌入的视觉基线更有效，并减少了与真实基准的性能差距。这些结果支持基于动作和语言条件的视频评估作为评估的模拟具身控制任务的可解释反馈机制。
cs.RO / 30 / 2608.08293
Scalable High-Speed Lateral Control for Single-Body and Articulated Autonomous Vehicles
可扩展的高速横向控制框架用于单体和关节式自主车辆
Abstract
This paper presents a scalable lateral control framework for robust path tracking of single-body and articulated autonomous vehicles at high speeds. A clothoid-based controller is extended with three key adaptations: 1) an integrated tangential check and Frechet distance method for optimized lookahead and oscillation mitigation; 2) real-time trajectory segment classification for dynamic adjustment of the lookahead search range; and 3) a dual-adaptive, rate-controlled lookahead mechanism responsive to cross-track error and upcoming path geometry. To support different vehicle configurations, the framework also incorporates flexible tracking point selection and curvature-to-steering lookup tables, enabling control of points such as the tractor rear axle, hitch point, or trailer-related locations without changing the core control architecture. The controller is evaluated using high-fidelity TruckSim and Simulink simulations for both standalone and articulated vehicles across dual lane changes and winding-road scenarios. Results demonstrate stable high-speed path tracking, reduced oscillations, effective lookahead adaptation, low cross-track errors, and acceptable lateral acceleration across different vehicle configurations. The results support the use of a unified lateral control architecture that can scale from single-body to articulated autonomous vehicles.
Chinese Translation
本文提出了一种可扩展的横向控制框架，用于在高速下对单体和关节式自主车辆进行稳健的路径跟踪。该框架基于克洛索伊德（clothoid）控制器，并进行了三项关键调整：1）集成切向检查和弗雷歇距离（Frechet distance）方法，以优化前视距离和减小振荡；2）实时轨迹段分类，以动态调整前视搜索范围；3）一种双自适应、速率控制的前视机制，能够响应横向误差和即将到来的路径几何形状。为了支持不同的车辆配置，该框架还结合了灵活的跟踪点选择和曲率与转向查找表，使得在不改变核心控制架构的情况下，可以控制如拖拉机后轴、连接点或与拖车相关的位置。通过高保真度的TruckSim和Simulink仿真，对独立和关节式车辆在双车道变换和曲折道路场景下进行了评估。结果表明，在不同车辆配置下，能够实现稳定的高速路径跟踪、减少振荡、有效的前视适应、低横向误差以及可接受的横向加速度。这些结果支持使用统一的横向控制架构，从单体自主车辆扩展到关节式自主车辆。
cs.RO / 31 / 2608.08320
Diminishing Returns of Intelligence: The Non-Linear Relationship Between LLM Scale and User Perception in Short-Duration Open-Ended Social Human-Robot Interactions
智能的递减收益：大型语言模型规模与用户感知在短时开放式社会人机交互中的非线性关系
Abstract
Large Language Models (LLMs) are increasingly used to drive embodied social agents, yet it remains unclear whether larger models improve user perception during brief human-robot encounters. This paper examines the effect of LLM parameter size on short-duration, open-ended social interactions with a robot interface. In a within-subjects study, 19 participants interacted with robot faces driven by Qwen3-VL models at 4B, 8B, and 30B parameters. Participants evaluated the interactions in terms of perceived intelligence, naturalness, enjoyment, and humor. Results showed no significant overall preference for the 30B model over the smaller variants, including no significant advantage over the 4B model in perceived naturalness or intelligence. A significant relationship between AI interaction frequency and intelligence rankings for the 30B model suggests that more experienced users may be more sensitive to differences in model capability. Overall, the findings indicate diminishing returns from model scaling in brief open-ended social HRI, where conversational flow, responsiveness, and socially appropriate behavior potentially matter as much as raw parameter count.
Chinese Translation
大型语言模型（LLMs）越来越多地用于驱动具身社会代理，但在短暂的人机交互中，较大模型是否能改善用户感知仍不明确。本文考察了LLM参数规模对与机器人界面的短时开放式社会交互的影响。在一项被试内研究中，19名参与者与由Qwen3-VL模型驱动的机器人面孔进行互动，模型参数分别为4B、8B和30B。参与者从感知智能、自然性、愉悦感和幽默感等方面评估了这些互动。结果显示，30B模型并未显著优于较小的变体，包括在感知自然性或智能方面对4B模型没有显著优势。AI交互频率与30B模型的智能排名之间存在显著关系，表明更有经验的用户可能对模型能力的差异更为敏感。总体而言，研究结果表明，在短时开放式社会人机交互中，模型规模的递减收益显著，其中对话流畅性、响应能力和社会适当行为可能与原始参数数量同样重要。
cs.RO / 32 / 2608.08323
MPPI Planning with Gaussian Based Human Cost Function for Social Navigation
基于高斯的人类成本函数的MPPI规划用于社会导航
Abstract
Safe robot navigation in crowded spaces requires planning that accounts for where people will be, not only where they are now. Model Predictive Path Integral (MPPI) control is an effective sampling-based planner, but many implementations encode humans as static point obstacles at their current positions, underestimating risk in dynamic scenes. We propose Predictive Gaussian Interaction Fields (PGIF), a spatiotemporal cost formulation that propagates pedestrian predictions forward over the full planning horizon and encodes them as anisotropic Gaussian repulsive fields aligned with each pedestrian's direction of motion. The forward spread of each field grows with the pedestrian's speed, creating a motion cone danger zone that penalises robot trajectories entering the pedestrian's path of travel more strongly than those approaching from behind. The formulation is closed-form and fully parallelisable across rollouts, adding no measurable computational overhead. Evaluated over 300 randomised crowd scenarios at three density levels, PGIF-MPPI achieves a 0\% collision rate at every density level, compared with up to 82\% for vanilla MPPI, while maintaining real-time planning performance.
Chinese Translation
在拥挤空间中安全的机器人导航需要考虑人们未来的位置，而不仅仅是他们当前的位置。模型预测路径积分（MPPI）控制是一种有效的基于采样的规划方法，但许多实现将人类视为静态点障碍物，位于其当前位置，从而低估了动态场景中的风险。我们提出了预测高斯交互场（PGIF），这是一种时空成本公式，能够在整个规划时间范围内向前传播行人预测，并将其编码为与每个行人运动方向对齐的各向异性高斯排斥场。每个场的向前传播随着行人速度的增加而增长，形成一个运动锥危险区，对进入行人行进路径的机器人轨迹施加比从后方接近的轨迹更强的惩罚。该公式是封闭形式的，并且可以在回放中完全并行化，不增加可测的计算开销。在三个密度水平下对300个随机化人群场景进行评估时，PGIF-MPPI在每个密度水平下实现了0%的碰撞率，而普通MPPI的碰撞率高达82%，同时保持实时规划性能。
cs.RO / 33 / 2608.08356
Stochastic Physics-Informed Neural Networks on Lie Groups for Learning Underwater Vehicle Dynamics
基于李群的随机物理信息神经网络用于学习水下车辆动力学
Abstract
Accurate models of underwater vehicle motion are needed for autonomous execution of marine tasks like infrastructure inspection and scientific sampling. However, such motion is challenging to characterize using traditional physics-based methods. This paper presents a novel data-driven framework for learning stochastic underwater vehicle dynamics. Using Euler-Poincar\'e dynamics and the geometry of Lie groups, we develop a stochastic physics-informed neural network architecture that respects the physical and geometric constraints of underwater vehicles. Our approach leverages structure-preserving stochastic integration and builds upon moment matching and finite dimensional matching to ensure geometrically-consistent training. We evaluate our approach in simulation and on an underwater vehicle navigating dock pylons in a harbor environment. The results demonstrate that our method learns accurate and robust dynamics models, enabling safe model-based control in challenging marine environments.
Chinese Translation
准确的水下车辆运动模型对于自主执行基础设施检查和科学采样等海洋任务至关重要。然而，使用传统的基于物理的方法来表征这种运动是具有挑战性的。本文提出了一种新颖的数据驱动框架，用于学习随机水下车辆动力学。通过使用欧拉-庞加莱动力学和李群的几何结构，我们开发了一种尊重水下车辆物理和几何约束的随机物理信息神经网络架构。我们的方法利用结构保持的随机积分，并基于矩匹配和有限维匹配，确保几何一致的训练。我们在仿真中以及在一个水下车辆在港口环境中导航码头桩的场景中评估了我们的方法。结果表明，我们的方法能够学习准确且稳健的动力学模型，从而在具有挑战性的海洋环境中实现安全的基于模型的控制。
cs.RO / 34 / 2608.08522
EsaacSim: A Multimodal Event Camera Add-on for NVIDIA Isaac Sim
EsaacSim：NVIDIA Isaac Sim的多模态事件相机插件
Abstract
Event-based vision is becoming an increasingly important sensing paradigm for robotics, yet its adoption remains limited by sensor availability and the lack of integrated simulation tools for modern robotics platforms. This paper presents EsaacSim, a multimodal event camera add-on for NVIDIA Isaac Sim that enables online simulation of configurable event cameras with grayscale and Bayer RGGB event generation. The framework supports multiple event camera resolutions and provides synchronized RGB, APS, event, depth, and IMU outputs through native ROS2 interfaces. A motion-guided frame-gap synthesis strategy further increases the effective temporal resolution while preserving compatibility with the Isaac Sim rendering pipeline. Experimental evaluation demonstrates synchronized multimodal simulation across representative robotic scenes and efficient online performance over five event camera resolutions at effective event rates from 240 to 960Hz. Event stream generation requires 6.98--27.28ms for grayscale events and 7.58--29.16ms for Bayer RGGB events while using less than 400MB of additional GPU memory on an NVIDIA RTX~4060 GPU. These results show that EsaacSim enables supports online multimodal event-camera simulation for robotics research and synthetic data generation. We release an early version of the simulator and report its current architecture and performance.
Chinese Translation
基于事件的视觉正成为机器人技术中越来越重要的感知范式，但其应用仍受限于传感器的可用性以及现代机器人平台缺乏集成的仿真工具。本文提出了EsaacSim，一个用于NVIDIA Isaac Sim的多模态事件相机插件，能够在线模拟可配置的事件相机，支持灰度和Bayer RGGB事件生成。该框架支持多种事件相机分辨率，并通过原生ROS2接口提供同步的RGB、APS、事件、深度和IMU输出。运动引导的帧间合成策略进一步提高了有效时间分辨率，同时保持与Isaac Sim渲染管道的兼容性。实验评估展示了在代表性机器人场景中同步的多模态仿真，以及在五种事件相机分辨率下有效事件率从240到960Hz的高效在线性能。事件流生成对于灰度事件需要6.98到27.28毫秒，对于Bayer RGGB事件需要7.58到29.16毫秒，同时在NVIDIA RTX 4060 GPU上使用的额外GPU内存少于400MB。这些结果表明，EsaacSim支持机器人研究和合成数据生成的在线多模态事件相机仿真。我们发布了该模拟器的早期版本，并报告了其当前架构和性能。
cs.RO / 35 / 2608.08545
Curriculum Generation under Structured Parametric Environments for Robust Navigation Policies
在结构化参数环境下生成课程以实现鲁棒导航策略
Abstract
Robust navigation policies for autonomous agents must generalize across continuously varying environmental conditions such as turn rates, obstacles, friction, pits, and slopes. Curriculum generation provides a principled mechanism for improving generalization by progressively adapting training environments, but designing such curricula in a sample-efficient and automated manner remains challenging. This paper proposes a reparameterized curriculum generation framework for structured continuous environment parameters using unidirectional gradient-based optimization. To improve robustness in multimodal observation spaces consisting of image-based and scalar inputs, a distribution-shift regularization objective is incorporated to encourage the learning of finer-grained latent representations. The proposed method is evaluated across two continuous-control OpenAI Gym environments: a 2D obstacle-based Car Racing variant and Bipedal Walker variant, where coupled environment parameters jointly influence policy performance. Across five random seeds, our method consistently outperforms vanilla policy training, random parameter sampling, manual curricula, frontier-based methods, Self-Paced Reinforcement Learning (SPRL), Absolute Learning Progress with Gaussian Mixture Models (ALP-GMM), and reverse curriculum learning baselines. Ablation studies further demonstrate the effectiveness of the reparameterized curriculum mechanism across both environments, while highlighting environment-dependent benefits of the auxiliary regularization objective.
Chinese Translation
自主智能体的鲁棒导航策略必须能够在不断变化的环境条件下进行泛化，例如转弯速率、障碍物、摩擦、坑洼和坡度。课程生成提供了一种系统化的机制，通过逐步调整训练环境来改善泛化能力，但以样本高效且自动化的方式设计这样的课程仍然具有挑战性。本文提出了一种针对结构化连续环境参数的重新参数化课程生成框架，采用单向梯度优化。为了提高在由基于图像和标量输入组成的多模态观测空间中的鲁棒性，本文引入了一种分布偏移正则化目标，以鼓励学习更细粒度的潜在表示。所提出的方法在两个连续控制的OpenAI Gym环境中进行了评估：一个是基于障碍物的2D赛车变体，另一个是双足行走者变体，其中耦合的环境参数共同影响策略性能。在五个随机种子下，我们的方法始终优于传统策略训练、随机参数采样、手动课程、基于前沿的方法、自适应强化学习（Self-Paced Reinforcement Learning, SPRL）、基于高斯混合模型的绝对学习进展（Absolute Learning Progress with Gaussian Mixture Models, ALP-GMM）以及反向课程学习基线。消融研究进一步证明了重新参数化课程机制在这两个环境中的有效性，同时突出了辅助正则化目标在环境依赖性方面的益处。
cs.RO / 36 / 2608.08558
Vid2WAM: Distilling Video Diffusion Priors into World Action Models
Vid2WAM：将视频扩散先验提炼为世界行动模型
Abstract
World Action Models (WAMs) improve robot policy learning by jointly modeling future visual dynamics and actions. However, their scalability and generalization remain constrained by their reliance on costly expert demonstrations. We challenge this by asking whether future supervision for WAMs must originate from target-task expert trajectories. In this paper, we propose Vid2WAM, an offline distillation framework that transfers visual diffusion priors from a large video foundation model into a compact WAM student. Given an observation and language instruction, Vid2WAM distills supervision through two complementary channels: task-conditioned future rollouts directly supervise the student's future prediction branch, while an inverse dynamics model recovers embodiment-specific pseudo-actions for action learning. To robustly integrate synthetic and real supervision, we introduce source-aware residual action adaptation that learns source-specific corrections around a shared action backbone and mitigates interference from noisy pseudo-actions. During inference, both the video teacher and inverse dynamics model are discarded, leaving only the WAM student for efficient deployment. Simulation and real-world experiments demonstrate that Vid2WAM improves novel-task generalization and data efficiency under limited expert demonstrations while preserving low-latency inference.
Chinese Translation
世界行动模型（WAMs）通过联合建模未来视觉动态和动作来改善机器人策略学习。然而，它们的可扩展性和泛化能力仍然受到对昂贵专家演示依赖的限制。我们提出质疑，WAMs的未来监督是否必须来自目标任务的专家轨迹。本文提出了Vid2WAM，一种离线提炼框架，将来自大型视频基础模型的视觉扩散先验转移到紧凑的WAM学生中。给定观察和语言指令，Vid2WAM通过两个互补通道提炼监督：任务条件的未来展开直接监督学生的未来预测分支，而逆动态模型则恢复特定于体现的伪动作以进行动作学习。为了稳健地整合合成和真实监督，我们引入了源感知残差动作适应，学习围绕共享动作骨干的源特定修正，并减轻来自噪声伪动作的干扰。在推理过程中，视频教师和逆动态模型均被丢弃，仅保留WAM学生以实现高效部署。仿真和现实世界实验表明，Vid2WAM在有限的专家演示下提高了新任务的泛化能力和数据效率，同时保持低延迟推理。
cs.RO / 37 / 2608.08598
Knowledge-Distilled End-to-End Reinforcement Learning for Smooth 6-DOF Thrust Control and Rapid Adaptation to Ocean Currents in Remotely Operated Vehicles
知识蒸馏的端到端强化学习用于平滑的六自由度推力控制及快速适应海洋流的遥控车辆
Abstract
With the continuous improvement of computational capabilities, end-to-end reinforcement learning has been rapidly developed for remotely operated vehicles control. Nevertheless, existing end-to-end reinforcement-learningbased methods still face challenges in achieving optimal control under oceancurrent disturbances. In particular, there remains a lack of a unified control framework that can simultaneously achieve low steady-state tracking error, rapid transient response, energy-efficient operation, and smooth controlforce outputs under disturbances. To address the issue, this paper proposes the thrust smoothness rapid current adaptation proximal policy optimization (TSRCA-PPO) method which learns a near-optimal strategy by a twostage distillation learning framework. The core innovations of this work lie in the reward-function design and the privileged multi-encoder architecture. Ablation studies validate the effectiveness of each module. Simulation results demonstrate that the proposed TSRCA-PPO method consistently outperforms the conventional cascaded P-PID controller across all evaluation metrics. Specifically, TSRCA-PPO reduces the steady-state position error, steady-state attitude error, settling time, energy index, and thrustsmoothness index to 42.7%, 76.5%, 10.6%, 93.5%, and 15.9% of the corresponding P-PID values, respectively.
Chinese Translation
随着计算能力的不断提升，端到端强化学习在遥控车辆控制方面得到了快速发展。然而，现有的基于端到端强化学习的方法在应对海洋流扰动时仍面临实现最佳控制的挑战。特别是，目前缺乏一个统一的控制框架，能够在扰动下同时实现低稳态跟踪误差、快速瞬态响应、节能操作和平滑的控制力输出。为了解决这一问题，本文提出了一种推力平滑快速流适应的近端策略优化方法（Thrust Smoothness Rapid Current Adaptation Proximal Policy Optimization，TSRCA-PPO），该方法通过双阶段蒸馏学习框架学习近似最优策略。本研究的核心创新在于奖励函数设计和特权多编码器架构。消融研究验证了每个模块的有效性。仿真结果表明，所提出的TSRCA-PPO方法在所有评估指标上均优于传统的级联P-PID控制器。具体而言，TSRCA-PPO将稳态位置误差、稳态姿态误差、稳定时间、能量指标和推力平滑指标分别降低至对应P-PID值的42.7%、76.5%、10.6%、93.5%和15.9%。
cs.RO / 38 / 2608.08618
RAG-Based Auto-Configuration for Industrial Fieldbus Devices
基于RAG的工业现场总线设备自动配置
Abstract
Industrial device commissioning requires engineers to manually extract hundreds of protocol-specific parameters from heterogeneous PDF manuals and transcribe them into supervisory control systems, a time-intensive, error-prone workflow. This paper presents SysName, a production-oriented pipeline that automates device configuration end-to-end for Modbus RTU, OPC-UA, Profibus DP, and CANopen. It builds a hybrid dense-sparse retrieval index augmented by an ontology graph derived from ECLASS, AAS, and SOSA/SSN, using a BGE-M3 encoder with a cross-encoder reranker to surface relevant manual passages. A local LLM (T=0.1) generates ontology-aligned JSON-LD configurations via protocol-specific prompts and a four-step repair pipeline. A two-stage abstention gate, combining a reranker-score threshold and an IRI resolution ratio, blocks unsafe LLM invocations and filters low-coverage configurations before SHACL validation. On a gold set of 28 field-level queries, the hybrid retriever reaches 0.96 HitRate@10, and the reranker raises MRR@10 from 0.56 to 0.63 with perfect score separation for abstention. The generator attains field-level F1=0.87 with exact match on 9 of 12 runs. End-to-end runs on an H100 GPU complete in 2.6-6.6s per device with zero unsafe writes and zero silent failures on a five-device benchmark; every unsuccessful run is flagged by abstention or deployment verification. Component-wise evaluation localises the single systematic failure to OPC-UA generation, invisible to end-to-end metrics alone. A case study commissions a physics-simulated Universal Robots UR5e robot from unmodified vendor documentation (254-page manual, 8-page register list, 496 chunks), reaching field-level F1=1.0 over three runs with read-back and joint-consistency verification. An ablation study and comparison with five industrial-LLM systems complete the analysis.
Chinese Translation
工业设备调试要求工程师手动从异构PDF手册中提取数百个特定于协议的参数，并将其转录到监控控制系统中，这是一项耗时且容易出错的工作流程。本文提出了SysName，一个面向生产的管道，能够端到端自动化配置Modbus RTU、OPC-UA、Profibus DP和CANopen设备。它构建了一个混合的密集-稀疏检索索引，增强了源自ECLASS、AAS和SOSA/SSN的本体图，使用BGE-M3编码器和交叉编码器重排序器来提取相关手册段落。一个本地LLM（T=0.1）通过特定于协议的提示和四步修复管道生成与本体对齐的JSON-LD配置。一个两阶段的弃权门，结合重排序分数阈值和IRI解析比率，阻止不安全的LLM调用，并在SHACL验证之前过滤低覆盖率的配置。在28个现场级查询的金标准集上，混合检索器达到0.96的HitRate@10，重排序器将MRR@10从0.56提高到0.63，并实现了完美的弃权分数分离。生成器在12次运行中有9次精确匹配，达到了现场级F1=0.87。在H100 GPU上的端到端运行每个设备完成时间为2.6-6.6秒，且在五设备基准测试中没有不安全写入和静默失败；每次不成功的运行都通过弃权或部署验证被标记。组件级评估将单一系统性故障定位于OPC-UA生成，这在端到端指标中不可见。一个案例研究从未修改的供应商文档（254页手册，8页寄存器列表，496个块）调试了一个物理模拟的Universal Robots UR5e机器人，在三次运行中达到了现场级F1=1.0，并进行了读回和关节一致性验证。一项消融研究和与五个工业LLM系统的比较完成了分析。
cs.RO / 39 / 2608.08701
Anchor-Based AI Approach for Pre-Crash Object Detection Utilizing Micro-Doppler Signatures in Automotive Radar
基于锚点的人工智能方法在汽车雷达中利用微多普勒特征进行碰撞前物体检测
Abstract
Advanced automated driving presents significant potential to improve modern automotive safety systems, but it depends highly on the reliable activation of restraint systems. Forward-looking sensors are crucial for immediate and precise object detection. Recent developments in automotive radar technology enable detailed environment detection and the recognition of high-resolution features, such as micro-Doppler signatures. Combined with advanced AI techniques, these features significantly enhance object detection and improve the accuracy of kinematic parameter estimation. This is essential for the early and reliable activation of irreversible safety systems, such as smart airbags and adaptive seat belts. Therefore, an anchor-based AI model is presented, designed to process high-resolution radar data with an explicit focus on micro-Doppler signatures to improve pre-crash object detection. Furthermore, these signatures can improve the accuracy of kinematic object parameter estimation and reduce false negatives, especially in the critical near-field. To address the challenges of sparse and fluctuating radar point clouds, an innovative radar-image dilation technique on the feature input channels was developed to amplify local radar patterns, like micro-Doppler features. Therefore, this approach increases the system's reliability and increases its ability to detect objects in pre-crash scenarios despite radar multipath reflections and ghost objects. In order to investigate the applicability and compare the model's performance with advanced automotive radar tracking methods, a radar data set using series sensors and pre-crash relevant scenarios was recorded. The results demonstrate the advantages of the anchor-based AI model over established tracking approaches. It excels at estimating object parameters in dynamic scenarios and underscores its ability to process different data sets effectively.
Chinese Translation
先进的自动驾驶技术在提升现代汽车安全系统方面具有显著潜力，但其高度依赖于约束系统的可靠激活。前向传感器对于即时和精确的物体检测至关重要。近年来汽车雷达技术的发展使得环境检测和高分辨率特征（如微多普勒特征）的识别成为可能。结合先进的人工智能技术，这些特征显著增强了物体检测能力，并提高了运动参数估计的准确性。这对于不可逆安全系统（如智能安全气囊和自适应安全带）的早期和可靠激活至关重要。因此，提出了一种基于锚点的人工智能模型，旨在处理高分辨率雷达数据，明确关注微多普勒特征，以改善碰撞前物体检测。此外，这些特征可以提高运动物体参数估计的准确性，并减少假阴性，特别是在关键的近场区域。为了解决稀疏和波动的雷达点云带来的挑战，开发了一种创新的雷达图像扩张技术，应用于特征输入通道，以增强局部雷达模式，如微多普勒特征。因此，该方法提高了系统的可靠性，并增强了其在碰撞前场景中检测物体的能力，尽管存在雷达多路径反射和虚假物体。为了研究其适用性并与先进的汽车雷达跟踪方法比较模型的性能，记录了一个使用系列传感器和碰撞前相关场景的雷达数据集。结果表明，基于锚点的人工智能模型相较于已建立的跟踪方法具有优势。它在动态场景中出色地估计物体参数，并强调了其有效处理不同数据集的能力。
cs.RO / 40 / 2608.08725
WA-SpecDec: World-Aware Speculative Decoding for Vision-Language-Action Models
WA-SpecDec：面向世界的推测解码用于视觉-语言-动作模型
Abstract
Vision-language-action (VLA) policies generate robot controls autoregressively, making closed-loop latency dominated by repeated target-model forward passes. Speculative decoding reduces this cost by verifying blocks of draft action tokens in parallel, and recent VLA methods further relax token-level acceptance because small differences in action-token space often map to similar continuous controls. However, this relaxation remains scene-agnostic. A fixed token-distance tolerance treats the same action-token deviation as equally safe across states, although deviations that are harmless in free space can cause collisions or grasp failures near contact. We propose WA-SpecDec, a world-aware speculative decoding framework that injects world-model-derived physical scene awareness during the VLA prefill stage, producing shared world-aware prefill states for draft proposal and target verification without changing the relaxed acceptance rule. Across three state-of-the-art relaxed acceptance schemes, WA-SpecDec preserves higher task success under looser relaxation and enables longer accepted prefixes. At comparable-success operating points, WA-SpecDec achieves a 1.5x matched-success speedup over VLA speculative decoding alone and reduces near-contact failure (NCF) by 18.6% on average relative to the corresponding speculative baselines.
Chinese Translation
视觉-语言-动作（VLA）策略以自回归方式生成机器人控制，使得闭环延迟主要受重复目标模型前向传递的影响。推测解码通过并行验证草拟动作令牌的块来降低这一成本，最近的VLA方法进一步放宽了令牌级别的接受，因为动作令牌空间中的小差异通常映射到相似的连续控制。然而，这种放宽仍然是场景无关的。固定的令牌距离容忍度在不同状态下将相同的动作令牌偏差视为同样安全，尽管在自由空间中无害的偏差在接触附近可能导致碰撞或抓取失败。我们提出了WA-SpecDec，一种面向世界的推测解码框架，在VLA预填充阶段注入基于世界模型的物理场景意识，为草拟提议和目标验证生成共享的面向世界的预填充状态，而无需改变放宽的接受规则。在三种最先进的放宽接受方案中，WA-SpecDec在更宽松的放宽条件下保持更高的任务成功率，并允许更长的接受前缀。在相似成功率的操作点上，WA-SpecDec相较于单独的VLA推测解码实现了1.5倍的匹配成功加速，并且相较于相应的推测基线，平均减少了18.6%的近接失败（NCF）。
cs.RO / 41 / 2608.08749
OnEvoMemory: Evolving Memory through Online Robot Rollouts for Pretrained Robot Policies
OnEvoMemory：通过在线机器人回放演化记忆以优化预训练机器人策略
Abstract
Long-horizon robot manipulation requires policies to track completed subtasks and critical interaction events. However, existing memory mechanisms heavily rely on external models or predefined update rules. To address this, we propose OnEvoMemory, a value-guided memory module for pretrained robot policies. It maintains recent context, high-value experiences, and salient transitions, while learning which experiences should be retained from trajectory outcomes. Offline demonstrations initialize the memory prior, whereas successful and unsuccessful online rollouts refine memory selection, helping the policy recognize task-stage transitions and avoid repeating completed subtasks. Experiments on long-horizon manipulation benchmarks show that OnEvoMemory improves the performance of the base VLA policy through both offline initialization and online memory evolution.
Chinese Translation
长时间跨度的机器人操作需要策略跟踪已完成的子任务和关键交互事件。然而，现有的记忆机制严重依赖外部模型或预定义的更新规则。为了解决这个问题，我们提出了OnEvoMemory，一种针对预训练机器人策略的价值引导记忆模块。它维护最近的上下文、高价值的经验和显著的转变，同时学习哪些经验应根据轨迹结果进行保留。离线演示初始化记忆先验，而成功和失败的在线回放则优化记忆选择，帮助策略识别任务阶段转变并避免重复已完成的子任务。在长时间跨度的操作基准测试中的实验表明，OnEvoMemory通过离线初始化和在线记忆演化提高了基础VLA策略的性能。
cs.RO / 42 / 2608.08773
PEEL: Parallel Extraction for Long-Horizon Disassembly Planning via Scale-Invariant Sampling
PEEL：通过尺度不变采样进行长时间跨度拆解规划的并行提取
Abstract
Long-horizon multi-part object disassembly requires robots to compute feasible sequences of collision-free removal motions, even in the presence of tight, narrow escape corridors. To efficiently solve such disassembly problems, we propose Parallel Extraction for Long-Horizon Disassembly (PEEL), an algorithm which efficiently computes disassembly motions for object assemblies and feeds them to a robot manipulator for execution. PEEL uses sampling-based motion planning to compute single-object motions through the use of a scale-invariant sampling scheme, where the object scale is estimated in a burn-in phase and a subsequent directional sampler exploits the scale. This sampling scheme is integrated into a multi-arm bandit rapidly-exploring random tree (MAB-RRT) planner, which switches between different samplers depending on the reward signal received. Using MAB-RRT, the PEEL algorithm runs a batch of planners in parallel to obtain an ordered graph specifying the sequence in which object parts have to be removed. We show that MAB-RRT can efficiently solve single-part disassemblies with 100 percent success rate on 76 assemblies, and that it is robust to its parameters. By integrating MAB-RRT into PEEL, we solve four long-horizon disassembly problems using the Fetch manipulator robot involving 10 to 17 individual object parts.
Chinese Translation
长时间跨度的多部件物体拆解要求机器人计算可行的无碰撞移除动作序列，即使在狭窄的逃生通道中也要如此。为有效解决此类拆解问题，我们提出了长时间跨度拆解的并行提取算法（PEEL），该算法高效计算物体组装的拆解动作，并将其传递给机器人操控器执行。PEEL使用基于采样的运动规划，通过一种尺度不变的采样方案计算单一物体的运动，其中物体的尺度在预热阶段进行估计，随后一个方向性采样器利用该尺度。该采样方案集成到多臂赌博机快速探索随机树（MAB-RRT）规划器中，根据接收到的奖励信号在不同的采样器之间切换。使用MAB-RRT，PEEL算法并行运行一批规划器，以获取指定物体部件移除顺序的有序图。我们展示了MAB-RRT能够高效解决76个组装体的单一部件拆解，成功率达到100%，并且对其参数具有鲁棒性。通过将MAB-RRT集成到PEEL中，我们使用Fetch操控机器人解决了四个涉及10到17个单独物体部件的长时间跨度拆解问题。
cs.RO / 43 / 2608.08839
SG-WAM: Text-Grounded and Spatial-aware Semantic Guidance for World-Action Models
SG-WAM：基于文本的空间感知语义引导的世界行动模型
Abstract
World-Action Models (WAMs) have emerged as a promising paradigm for robotic manipulation. However, most existing WAMs generate future videos and actions by relying mainly on visual cues rather than language instructions, since off-the-shelf text encoders embed instructions independently of visual observations. As a result, the videos predicted by these WAMs are often semantically misaligned with their corresponding language instructions, which degrades the accuracy of the predicted actions. To overcome this limitation, we propose SG-WAM, a semantic guidance method for world-action models that leverages a vision-language model (VLM) as a semantic planner to enhance the instruction-grounding capacity of world-action models. Specifically, we train a VLM-based planner to predict text-grounded and spatial-aware semantic foresight. The text-grounded semantic foresight grounds the instruction by identifying the correct target objects, and the spatial-aware semantic foresight provides the scene geometry for precise manipulation. We then inject this foresight into the world-action model as high-level semantic guidance, ensuring that both future-video generation and action prediction faithfully follow the language instruction. Extensive experiments in simulation and the real world demonstrate the superiority of our semantic guidance method, showcasing precise manipulation and strong instruction-following capabilities.
Chinese Translation
世界行动模型（WAMs）已成为机器人操作的一个有前景的范式。然而，现有的大多数WAMs主要依赖视觉线索而非语言指令来生成未来的视频和动作，因为现成的文本编码器独立于视觉观察嵌入指令。因此，这些WAMs预测的视频通常与其对应的语言指令在语义上不一致，从而降低了预测动作的准确性。为了解决这一局限性，我们提出了SG-WAM，一种针对世界行动模型的语义引导方法，利用视觉-语言模型（VLM）作为语义规划器，以增强世界行动模型的指令基础能力。具体而言，我们训练一个基于VLM的规划器来预测基于文本的空间感知语义前瞻。基于文本的语义前瞻通过识别正确的目标对象来为指令提供基础，而空间感知的语义前瞻则为精确操作提供场景几何信息。然后，我们将这种前瞻注入世界行动模型，作为高层次的语义引导，确保未来视频生成和动作预测都忠实于语言指令。大量的模拟和现实世界实验表明，我们的语义引导方法具有优越性，展示了精确的操作和强大的指令遵循能力。
cs.RO / 44 / 2608.08871
Hierarchical Topology-Aware Planning and Control of Underwater Vehicle-Manipulator Systems in Confined Environments
受限环境中水下车辆-操纵器系统的分层拓扑感知规划与控制
Abstract
This paper addresses autonomous intervention with an underwater vehicle--manipulator system (UVMS) in confined, cluttered, and partially known environments, where poor maneuverability, narrow passages, and uncertain execution may cause the robot to enter unrecoverable regions. We propose MANTA, a three-layer hierarchical planning-and-control framework that couples passage accessibility, manipulation feasibility, and closed-loop execution. The first layer performs global connectivity reasoning in a conservative reduced base space to extract traversable corridor candidates toward the task region. The second layer refines each candidate corridor by jointly optimizing the continuous base motion and arm trajectory, producing a collision-free base--arm trajectory. The third layer learns a reach-and-hold base policy using Gaussian-process model-based reinforcement learning (MBRL) through MC-PILCO, enabling trajectory tracking and station keeping at the planned manipulation state. During execution, the framework monitors map updates and can trigger recovery and route repair when the active passage becomes infeasible. MANTA is evaluated in confined UVMS planning and closed-loop tracking experiments. Across 120 matched planning queries, it achieves higher task success than full-state sampling-based baselines while producing larger clearance margins and lower arm motion. The learned MC-PILCO policy further reduces position and yaw tracking errors on both training and unseen tube-like references. These results show MANTA as a structured and data-efficient framework for safe autonomous underwater intervention in caves, tubes, and cluttered subsea structures.
Chinese Translation
本文探讨了在受限、杂乱且部分已知环境中，水下车辆-操纵器系统（UVMS）的自主干预问题。在这些环境中，较差的机动性、狭窄的通道和不确定的执行可能导致机器人进入不可恢复的区域。我们提出了MANTA，一个三层分层规划与控制框架，该框架结合了通道可达性、操纵可行性和闭环执行。第一层在保守的降维基础空间中进行全局连通性推理，以提取通向任务区域的可通行走廊候选。第二层通过联合优化连续基础运动和臂轨迹来细化每个候选走廊，生成无碰撞的基础-臂轨迹。第三层利用基于高斯过程模型的强化学习（MBRL）通过MC-PILCO学习一个到达并保持基础策略，使得在计划的操纵状态下实现轨迹跟踪和驻留。在执行过程中，该框架监控地图更新，并可以在活动通道变得不可行时触发恢复和路线修复。MANTA在受限UVMS规划和闭环跟踪实验中进行了评估。在120个匹配的规划查询中，它实现了比基于全状态采样的基线更高的任务成功率，同时产生了更大的安全间隙和更低的臂运动。学习到的MC-PILCO策略进一步减少了训练和未见管状参考上的位置和偏航跟踪误差。这些结果表明，MANTA是一个结构化且数据高效的框架，适用于在洞穴、管道和杂乱的海底结构中进行安全的自主水下干预。
cs.RO / 45 / 2608.08884
SHRIMP: Iterative Refinement of Robot Task Plans
SHRIMP：机器人任务计划的迭代优化
Abstract
As collaborative robots have entered domains such as manufacturing, agriculture, and healthcare, programming or adapting robot behavior typically requires robotic expertise that most end users lack. Natural language lowers this barrier. Recent advancements in large language models (LLMs) have made it feasible to translate natural language into robot task plans. However, language-based task specification suffers from semantic ambiguity, and generative models lack transparency for how language instructions become robot actions, making it difficult for users to validate the plan before execution. To address these issues, we introduce SHRIMP, a system that allows users to automatically generate a hierarchical robot primitive plan using natural language and iteratively revise their plan through re-prompting and explicit correction. At each revision, SHRIMP allows users to validate their plan in simulation, and once satisfied, execute it on the physical robot. Through a user study involving participants planning tabletop kitchen tasks (n=35), we validate that SHRIMP improves perceived control and enhances robot transparency. System videos and source code are available at https://wisc-hci.github.io/SHRIMP.
Chinese Translation
随着协作机器人进入制造、农业和医疗等领域，编程或调整机器人行为通常需要大多数最终用户所缺乏的机器人专业知识。自然语言降低了这一门槛。最近在大型语言模型（LLMs）方面的进展使得将自然语言转换为机器人任务计划成为可能。然而，基于语言的任务规范存在语义模糊的问题，而生成模型在语言指令如何转化为机器人动作方面缺乏透明性，这使得用户在执行前难以验证计划。为了解决这些问题，我们提出了SHRIMP，一个允许用户使用自然语言自动生成层次化机器人原语计划并通过重新提示和明确修正迭代修改计划的系统。在每次修订中，SHRIMP允许用户在模拟中验证他们的计划，并在满意后在物理机器人上执行。通过一项涉及参与者规划桌面厨房任务的用户研究（n=35），我们验证了SHRIMP提高了用户的控制感并增强了机器人透明度。系统视频和源代码可在 https://wisc-hci.github.io/SHRIMP 获取。
cs.RO / 46 / 2608.08886
Estimation of Spacecraft Inertia Tensor Using Attitude-Only Data from Torque-Free Motion
利用无扭矩运动的姿态数据估计航天器惯性张量
Abstract
We present an attitude-only framework for estimating a spacecraft's normalized inertia tensor from torque-free rotational motion. Our method supports both continuous single-arc observations and the joint use of multiple short torque-free arcs, while requiring neither gyroscope measurements nor known control torques. A Karush-Kuhn-Tucker formulation provides a fast linear initialization, which is refined by nonlinear shooting using the exact Jacobi-elliptic solution of Euler's equations and a Magnus-expansion quaternion map. Under controlled attitude noise, tests using a single 500-second arc reduced inertia-tensor error by approximately one order of magnitude relative to an Extended Kalman Filter initialized from the same estimate, while requiring nearly two orders of magnitude less computation. Joint estimation from three 100-second arcs provided a similar improvement in accuracy and remained more than one order of magnitude faster. Photorealistic proximity-operations simulations further evaluated both strategies using monocular image-derived attitudes. The 2000-second single-arc cases achieved sub-thousandth median inertia-tensor error and supported 10-hour attitude predictions with single-digit-degree median error. In three-arc cases using 30-300 seconds per arc, our method consistently outperformed the EKF refinement, with performance governed by rotational excitation and temporal sampling.
Chinese Translation
我们提出了一种仅基于姿态的框架，用于从无扭矩的旋转运动中估计航天器的归一化惯性张量。我们的方法支持连续的单弧观测以及多个短无扭矩弧的联合使用，同时不需要陀螺仪测量或已知的控制扭矩。Karush-Kuhn-Tucker 形式提供了快速的线性初始化，随后通过使用欧拉方程的精确雅可比椭圆解和马格努斯扩展四元数映射的非线性发射进行细化。在控制的姿态噪声下，使用单个500秒弧的测试将惯性张量误差相对于从相同估计初始化的扩展卡尔曼滤波器减少了大约一个数量级，同时计算量减少了近两个数量级。来自三个100秒弧的联合估计提供了类似的精度提升，并且速度仍然快于一个数量级。通过单目图像导出的姿态，进一步评估了这两种策略的逼真近距离操作模拟。在2000秒的单弧案例中，达到了千分之一的中位惯性张量误差，并支持10小时的姿态预测，具有单数度数的中位误差。在每个弧使用30-300秒的三弧案例中，我们的方法始终优于EKF细化，其性能受旋转激励和时间采样的影响。
cs.RO / 47 / 2608.08941
From Operational Design Domain to Action: A Systematic Behavioral Taxonomy for Autonomous Driving
从操作设计域到行动：自主驾驶的系统行为分类法
Abstract
Operational Design Domain (ODD) specifications describe where an automated driving system (ADS) is permitted to operate, but they do not prescribe what the ADS must demonstrably do once deployed within that domain. This gap between operating condition specification and behavioral validation represents a critical unresolved challenge in ADS safety assurance. This paper presents a structured, standards-grounded taxonomy of 21 behavioral competencies organized across three operational domains-Highway (HWY), Urban (URB), and Hub (HUB)-derived systematically from the PEGASUS six-layer model-based ODD. Each behavior is decomposed along longitudinal and lateral control axes and characterized against a four-property framework: Safety (gap maintenance, conflict avoidance, kinematic stability), Compliance (legal rules and behavioral norms), Comfort (rider dynamics and trust), and Efficiency (mission completion and product-level metrics). We further demonstrate that the crossing of ODD layer parameterizations with behavioral competency specifications yields concrete scenario families suitable for systematic behavioral testing and SOTIF coverage evidence. The taxonomy is grounded in AVSC00008202111, SAE J3237, and SAE J3016, and is validated as an operational specification layer through its deployment in a rule-enforced trajectory optimization system. The Hub domain is identified as a structurally distinct, underspecified domain warranting dedicated research attention.
Chinese Translation
操作设计域（Operational Design Domain, ODD）规范描述了自动驾驶系统（Automated Driving System, ADS）被允许操作的环境，但并未规定ADS在该域内部署后必须明确执行的行为。这种操作条件规范与行为验证之间的差距，代表了ADS安全保障中一个关键的未解决挑战。本文提出了一种结构化、基于标准的21种行为能力分类法，按照高速公路（Highway, HWY）、城市（Urban, URB）和枢纽（Hub, HUB）三个操作域进行组织，系统来源于PEGASUS六层模型的ODD。每种行为沿纵向和横向控制轴进行分解，并根据四个属性框架进行特征化：安全性（间隙维护、冲突避免、运动稳定性）、合规性（法律规则和行为规范）、舒适性（乘客动态和信任）、以及效率（任务完成和产品级指标）。我们进一步展示了ODD层参数化与行为能力规范的交叉，产生了适合系统行为测试和SOTIF覆盖证据的具体场景家族。该分类法基于AVSC00008202111、SAE J3237和SAE J3016，并通过在规则强制的轨迹优化系统中的应用验证为操作规范层。枢纽域被识别为一个结构上独特且规范不足的领域，值得专门的研究关注。
cs.RO / 48 / 2608.09015
Personalized Lower-limb Exoskeleton Assistance via Preference-based Bayesian Optimization
基于偏好的贝叶斯优化的个性化下肢外骨骼辅助
Abstract
A significant challenge in exoskeleton robotics is the need to dynamically adapt control profiles to individual motion preferences, thereby ensuring both efficient and comfortable assistance. Currently, since user experience can serve as a comprehensive metric for evaluating the effectiveness of assistance, user preference-based optimization methods have been widely studied for parameter tuning. However, the existing methods rely heavily on extensive human-robot online interactions and suffer from slow optimization speed, which not only induces user fatigue but also compromises optimization effectiveness. Therefore, this paper aims to explore an efficient preference-based optimization framework for personalized exoskeleton assistance that can learn optimal parameters with minimal interaction. We propose a preference-based Bayesian optimization (PbBO) approach that can improve sample efficiency by leveraging knowledge about the sampling distribution of candidate sets. For optimizing six control parameters, PbBO can converge to user-preferred parameters with 90.7% validation accuracy via 20 iterations. Moreover, the hierarchical controller is designed to generate personalized torque for different tasks and achieve interaction torque tracking in real time. The results of treadmill and outdoor experiments demonstrate that the optimized parameters can reduce metabolic rate by 14.5%-15.4%, heart rate by 6.3%-7.6%, and muscle activation by 6.7%-31.5% compared to unassisted walking.
Chinese Translation
外骨骼机器人面临的一个重大挑战是需要动态调整控制配置，以适应个体的运动偏好，从而确保高效且舒适的辅助。目前，由于用户体验可以作为评估辅助效果的综合指标，基于用户偏好的优化方法已被广泛研究用于参数调优。然而，现有方法在很大程度上依赖于广泛的人机在线交互，且优化速度较慢，这不仅会导致用户疲劳，还会影响优化效果。因此，本文旨在探索一种高效的基于偏好的优化框架，以实现个性化的外骨骼辅助，能够在最小交互的情况下学习最优参数。我们提出了一种基于偏好的贝叶斯优化（PbBO）方法，通过利用候选集的采样分布知识来提高样本效率。在优化六个控制参数时，PbBO能够在20次迭代中以90.7%的验证准确率收敛到用户偏好的参数。此外，设计了分层控制器以生成不同任务的个性化扭矩，并实现实时的交互扭矩跟踪。跑步机和户外实验的结果表明，与无辅助行走相比，优化后的参数可以将代谢率降低14.5%-15.4%，心率降低6.3%-7.6%，肌肉激活降低6.7%-31.5%。
cs.RO / 49 / 2608.09073
Latent World Models with Monotone Planning Costs for Image-Goal Navigation
具有单调规划成本的潜在世界模型用于图像目标导航
Abstract
Image-goal navigation with latent world models requires not only accurate future prediction, but also a planning cost that reliably ranks candidate action sequences. We define the cost as the cosine distance between the predicted future embedding and the goal embedding, and show that poor cost ordering can mislead sampling-based planners such as Cross-Entropy Method (CEM). To address this, we propose a latent world model built on a frozen DINO-family encoder and train it with two complementary objectives. An autoregressive rollout loss reduces the gap between training and multi-step planning rollouts, while a Monotone Cost Ranking (MCR) loss directly encourages increasingly perturbed action sequences to receive higher planning costs. We also study InfoNCE-based action-contrastive training and find that temporal permutation negatives distort the latent geometry and degrade planning performance. On the GNM navigation dataset, our method outperforms Navigation World Models (NWM), DINO-WM, OmniVLA, and NoMaD, achieving state-of-the-art image-goal navigation performance while reducing orientation error by $2.7\times$ over the same-encoder DINO WM baseline. We also deploy the model zero-shot on a physical robot, where it follows goal-directed paths in unseen indoor and outdoor environments.
Chinese Translation
基于潜在世界模型的图像目标导航不仅需要准确的未来预测，还需要一个可靠地对候选动作序列进行排名的规划成本。我们将成本定义为预测的未来嵌入与目标嵌入之间的余弦距离，并展示了不良的成本排序可能会误导基于采样的规划器，如交叉熵方法（Cross-Entropy Method, CEM）。为了解决这个问题，我们提出了一种基于冻结的DINO系列编码器构建的潜在世界模型，并通过两个互补的目标进行训练。自回归展开损失减少了训练与多步规划展开之间的差距，而单调成本排名（Monotone Cost Ranking, MCR）损失则直接鼓励越来越扰动的动作序列获得更高的规划成本。我们还研究了基于InfoNCE的动作对比训练，发现时间置换负样本扭曲了潜在几何结构，降低了规划性能。在GNM导航数据集上，我们的方法超越了导航世界模型（Navigation World Models, NWM）、DINO-WM、OmniVLA和NoMaD，达到了最先进的图像目标导航性能，同时将方向误差降低了$2.7 imes$，相较于相同编码器的DINO WM基线。我们还在物理机器人上零样本部署该模型，使其能够在未见过的室内和室外环境中沿目标导向路径行驶。
cs.RO / 50 / 2608.09098
UnsDrive: Towards Robust End-to-End Autonomous Driving in Unstructured Scenes
UnsDrive：朝着在非结构化场景中实现稳健的端到端自动驾驶
Abstract
End-to-end planning has shown strong promise for autonomous driving, but most existing methods are designed for structured urban roads and generalize poorly to unstructured mining environments. In such settings, weak road structure, terrain-induced occlusions, degraded visibility, and large unobserved regions make safe planning particularly challenging. To address these challenges, we propose UnsDrive, an end-to-end planner designed for unstructured mining scenes. UnsDrive builds an unknown-aware occupancy representation that explicitly models occupied, free, and unknown space using multi-frame visibility cues, and conditions a flow-matching planner on this representation to generate multimodal future trajectories. To improve safety under partial observability, we further introduce an occupancy trajectory consistency loss and an uncertainty-aware trajectory scorer that penalize trajectories entering non-traversable or unobserved regions. We also present MineLoop, a mining-oriented closed-loop simulator for evaluating autonomous driving under irregular road geometry, degraded visibility, heavy-vehicle interactions, and mining-specific operational constraints. Experiments in both open-loop and closed-loop settings show that UnsDrive consistently outperforms strong baselines in trajectory accuracy, collision avoidance, and long-horizon driving robustness. These results demonstrate the value of explicit unknown-space reasoning for autonomous driving in unstructured mining environments.
Chinese Translation
端到端规划在自动驾驶中展现了强大的潜力，但大多数现有方法是为结构化城市道路设计的，且在非结构化矿区环境中泛化能力较差。在这些环境中，薄弱的道路结构、地形引起的遮挡、能见度降低以及大面积未观测区域使得安全规划尤为具有挑战性。为了解决这些问题，我们提出了UnsDrive，一种专为非结构化矿区场景设计的端到端规划器。UnsDrive构建了一种未知感知的占用表示，利用多帧可见性线索显式建模占用、空闲和未知空间，并基于该表示条件化流匹配规划器以生成多模态未来轨迹。为了在部分可观测性下提高安全性，我们进一步引入了占用轨迹一致性损失和不确定性感知轨迹评分器，以惩罚进入不可通行或未观测区域的轨迹。我们还提出了MineLoop，一种面向矿业的闭环模拟器，用于评估在不规则道路几何、能见度降低、重型车辆交互和矿业特定操作约束下的自动驾驶。开放环和闭环设置下的实验表明，UnsDrive在轨迹准确性、碰撞避免和长时间驾驶稳健性方面始终优于强基线。这些结果展示了在非结构化矿区环境中，显式未知空间推理对自动驾驶的价值。
cs.RO / 51 / 2608.09112
ROEVO: Robust Organized Edge Feature-based Visual Odometry Using RGB-D Cameras
ROEVO：基于鲁棒有序边缘特征的RGB-D相机视觉里程计
Abstract
This work presents a visual odometry (VO) system that leverages image edge features. Edges are spatially expressive cues commonly present across diverse environments, offering rich textural and structural information. However, existing edge-based VO methods often fail to fully exploit this potential. To this end, we introduce a novel feature representation termed \textit{organized edges}, which transforms disjoint edge pixels into sequentialized clusters, enabling more effective retention and utilization of the underlying textural and structural information. Another nice property of this formulation is that organized edges can perform edge-level association across multiple frames, enabling the establishment of a co-visibility graph. To achieve precise and efficient pose estimation, we propose a range of particularly designed tracking and joint optimization methods based on the characteristics of organized edges. For tracking, we formulate edge-wise rather than pixel-wise residuals to achieve robust and accurate inter-frame registration. For joint optimization, we introduce a novel shape-preserving edge-fitting method and an organized edge-based Bundle Adjustment (BA) approach, which decomposes the traditional BA problem into fitting and registration to preserve the structural integrity. Based on these novel techniques, we develop a complete VO system that exclusively employs organized edge features, achieving efficient tracking and precise local mapping. Extensive experiments demonstrate its accuracy and robustness in indoor environments, outperforming or achieving comparable performance to state-of-the-art methods. The source code is publicly available at https://github.com/liumingrui814/ROEVO
Chinese Translation
本研究提出了一种利用图像边缘特征的视觉里程计（VO）系统。边缘是广泛存在于各种环境中的空间表达线索，提供了丰富的纹理和结构信息。然而，现有的基于边缘的VO方法往往未能充分利用这一潜力。为此，我们引入了一种新颖的特征表示，称为“有序边缘”（organized edges），该方法将离散的边缘像素转化为顺序化的聚类，从而更有效地保留和利用潜在的纹理和结构信息。这种表述的另一个优点是，有序边缘能够在多个帧之间进行边缘级关联，从而建立可见性图。为了实现精确和高效的位姿估计，我们提出了一系列特别设计的跟踪和联合优化方法，基于有序边缘的特性。在跟踪方面，我们采用边缘级而非像素级的残差来实现鲁棒且准确的帧间配准。在联合优化方面，我们引入了一种新颖的形状保持边缘拟合方法和一种基于有序边缘的束调整（Bundle Adjustment, BA）方法，该方法将传统的BA问题分解为拟合和配准，以保持结构完整性。基于这些新技术，我们开发了一个完整的VO系统，专门采用有序边缘特征，实现了高效的跟踪和精确的局部映射。大量实验表明，该系统在室内环境中的准确性和鲁棒性优于或与最先进的方法相当。源代码可在 https://github.com/liumingrui814/ROEVO 获取。
cs.RO / 52 / 2608.09125
Trajectory Divergence Horizon Decision for Reliable Dual-Arm Surgical Subtask Manipulation
可靠的双臂外科子任务操作的轨迹分歧视界决策
Abstract
Surgical robotic systems are increasingly being adopted as clinical workload rises, motivating autonomous solutions for repetitive manipulation subtasks. Learning-based controllers improve generalization compared with rule-based and analytic approaches, but most are trained for individual tasks and remain difficult to reuse across procedures. Vision-Language-Action (VLA) models provide a unified framework that integrates visual perception, language grounding, and action generation, offering a promising path toward more composable surgical autonomy. However, existing VLA policies rely on fixed-length open-loop action sequences, where changing scene conditions can lead to accumulated errors and potential risks in surgical manipulation. To mitigate this issue, we formulate surgical VLA deployment as an adaptive execution-horizon decision problem and propose Trajectory Divergence Horizon Decision (TDHD), a test-time mechanism that estimates step-wise action reliability by measuring the divergence between two flow-matching-generated trajectories under small noise perturbations and truncates execution using a dual-threshold rule to trigger timely replanning. We further establish a real-world da Vinci-like dual-arm benchmark with synchronized multi-view perception and language instructions, and collect 600 teleoperated demonstrations across needle (reach, pick, regrasp) and tissue (reach, lift, resection) manipulation suites. On real hardware with 20 trials per task setting, TDHD consistently improves performance over the latest VLA baselines: success increases from 55\% to 60\% for needle manipulation and from 55\% to 80\% for tissue manipulation, with the largest gains observed in the final manipulation stages. These results highlight the importance of adaptive execution control for reliable deployment of VLA models in surgical robotic manipulation.
Chinese Translation
随着临床工作负荷的增加，外科机器人系统的采用率日益上升，这促使了对重复操作子任务的自主解决方案的需求。与基于规则和分析的方法相比，基于学习的控制器在泛化能力上有所提升，但大多数控制器是针对单一任务进行训练的，难以在不同程序中重复使用。视觉-语言-动作（Vision-Language-Action, VLA）模型提供了一个统一的框架，整合了视觉感知、语言基础和动作生成，为更可组合的外科自主性提供了有希望的路径。然而，现有的VLA策略依赖于固定长度的开环动作序列，场景条件的变化可能导致累积误差和外科操作中的潜在风险。为了解决这个问题，我们将外科VLA部署形式化为一个自适应执行视界决策问题，并提出了轨迹分歧视界决策（Trajectory Divergence Horizon Decision, TDHD），这是一种在测试时估计逐步动作可靠性的机制，通过测量在小噪声扰动下两个流匹配生成的轨迹之间的分歧，并使用双阈值规则截断执行以触发及时的重新规划。我们进一步建立了一个类似于达芬奇（da Vinci）的双臂基准测试，配备同步的多视角感知和语言指令，并收集了600个在针（到达、拾取、重新抓取）和组织（到达、提升、切除）操作套件中的遥控演示。在真实硬件上，每个任务设置进行了20次试验，TDHD在性能上始终优于最新的VLA基线：针操作的成功率从55%提高到60%，组织操作的成功率从55%提高到80%，在最终操作阶段观察到最大的增益。这些结果突显了自适应执行控制在外科机器人操作中可靠部署VLA模型的重要性。
cs.RO / 53 / 2608.09127
High Fidelity Capture, Reconstruction, and Transfer of Human Demonstrations for Robot-Assisted Bathing
高保真捕捉、重建和转移人类示范用于机器人辅助洗浴
Abstract
Despite the demand for robots in high-value clinical tasks like bathing, contemporary systems still lack the safety and reliability required for complex, sustained physical interaction with humans. A key challenge hindering the development of such systems is that collecting, understanding, and effectively transferring highly dynamic, contact-rich human bathing demonstrations is difficult, even with modern motion and tactile sensing equipment. We present a straightforward, but effective framework for doing so with high fidelity by utilizing contact regions as a key processing primitive. We use our framework to build a dataset of bathing demonstrations performed by trained clinicians on human subjects. We then use this dataset to design and control an arm-mounted dexterous soft hand to perform bathing tasks on a mannequin using open- and closed-loop strategies. Our dataset is the first to provide high quality synchronized motion, shape, contact, and force during sustained, contact-rich human-human interaction, and our transfer strategies demonstrate effective use of these data across multiple levels of the robotics stack. All relevant materials will be publicly released to enable further advancements in physical human-robot interaction (pHRI) research.
Chinese Translation
尽管在洗浴等高价值临床任务中对机器人的需求日益增加，但现有系统仍缺乏进行复杂、持续的人机物理交互所需的安全性和可靠性。阻碍此类系统发展的一个关键挑战是，即使使用现代运动和触觉传感设备，收集、理解和有效转移高度动态、接触丰富的人类洗浴示范仍然困难。我们提出了一种简单但有效的框架，通过利用接触区域作为关键处理原语，实现高保真度的捕捉和转移。我们利用该框架构建了一个由经过培训的临床医生在人体上执行的洗浴示范数据集。随后，我们使用该数据集设计并控制一个安装在手臂上的灵巧软手，使用开放和闭环策略在假人上执行洗浴任务。我们的数据集是第一个在持续的、接触丰富的人际交互中提供高质量同步运动、形状、接触和力的数据集，我们的转移策略展示了在机器人技术各个层面有效利用这些数据的能力。所有相关材料将公开发布，以促进物理人机交互（pHRI）研究的进一步发展。
cs.RO / 54 / 2608.09138
SpeedTuning: Speeding Up Policy Execution with Lightweight Reinforcement Learning
SpeedTuning：通过轻量级强化学习加速策略执行
Abstract
While learned robotic policies hold promise for advancing generalizable manipulation, their practical deployment is often hindered by suboptimal execution speeds. Imitation learning policies are inherently limited by hardware constraints and the speed of the operator during data collection. In addition, there are no established methods for accelerating policies learned via imitation, and the empirical relationship between execution speed and task success remains underexplored. To address these issues, we introduce SpeedTuning, a reinforcement learning framework specifically designed to enhance the speed of manipulation policies. SpeedTuning learns to predict the optimal execution speed for actions, thereby complementing a base policy without necessitating additional data collection. We provide empirical evidence that SpeedTuning achieves substantial improvements in execution speed, exceeding 2.4x speed-up, while preserving an adequate success rate compared to both the original task policy and straightforward speed-up methods such as linear interpolation at a fixed speed. We evaluate our approach across a diverse set of dynamic and precise tasks, including pouring, throwing, and picking, demonstrating its effectiveness and robustness in enhancing real-world robotic manipulation.
Chinese Translation
尽管学习到的机器人策略在推动可推广的操作方面具有潜力，但其实际部署常常受到亚最优执行速度的制约。模仿学习策略本质上受到硬件限制和数据收集期间操作员速度的限制。此外，目前尚无加速通过模仿学习获得的策略的既定方法，执行速度与任务成功之间的经验关系仍未得到充分探索。为了解决这些问题，我们提出了SpeedTuning，一个专门设计用于增强操作策略速度的强化学习框架。SpeedTuning学习预测动作的最佳执行速度，从而在不需要额外数据收集的情况下补充基础策略。我们提供了实证证据，表明SpeedTuning在执行速度上实现了显著提升，速度超过2.4倍，同时在成功率上保持了与原始任务策略和简单的加速方法（如固定速度的线性插值）相当的水平。我们在一系列动态和精确的任务中评估了我们的方法，包括倒水、投掷和抓取，展示了其在增强现实世界机器人操作中的有效性和鲁棒性。
cs.RO / 55 / 2608.09166
Particle-Based Conformal Prediction for Contact-Aware Uncertainty Calibration in Stratified Configuration Spaces
基于粒子的符合预测用于分层配置空间中的接触感知不确定性校准
Abstract
Reliable uncertainty representation is essential for deploying autonomous systems that interact with their environment, as robots must reason about how uncertainty arising from both stochasticity and model mismatch is impacted by contacts with obstacles (e.g., when navigating through a cluttered environment or inserting a part into an assembly). We propose Calibrated Particle-sets for Trans-dimensional Uncertainty Representation (CaPTURe), a geometry-aware, conformal prediction-based algorithm that generates probabilistically valid prediction regions of the unknown future system configuration using particle-based models of arbitrary fidelity. While calibrated uncertainty predictions are essential for safe and efficient planning, analytical or learned motion models are often inaccurate - due to limited data, simplifying assumptions, unmodeled effects, etc. - which can lead to unsafe executions or task failure. Additionally, when a robot contacts an obstacle, the distribution of its future configurations can become multimodal or disjoint, or lie along manifolds of lower intrinsic dimension than the space of possible robot configurations. Our method uses a calibration dataset of system transitions to locally calibrate motion uncertainty estimates, constructing regions guaranteed to contain the future robot configuration at a user-set probability. Our calibration procedure captures how motion uncertainty varies between contact-rich and contactless motions, leading to sufficient coverage in both cases. We evaluate our method on two simulated planning tasks: controlling a marble around a labyrinth and performing tight-tolerance peg-in-hole insertion with a manipulator. Compared to relevant baselines, CaPTURe achieves the user-specified coverage requirement both in and out of contact and achieves up to a 30% absolute improvement in task success rate over the best baseline.
Chinese Translation
可靠的不确定性表示对于部署与环境互动的自主系统至关重要，因为机器人必须推理不确定性如何受到随机性和模型不匹配的影响，尤其是在与障碍物接触时（例如，在拥挤环境中导航或将部件插入装配中）。我们提出了用于跨维不确定性表示的校准粒子集（Calibrated Particle-sets for Trans-dimensional Uncertainty Representation，CaPTURe），这是一种基于几何感知的符合预测算法，利用任意精度的粒子模型生成未知未来系统配置的概率有效预测区域。虽然校准的不确定性预测对于安全和高效的规划至关重要，但分析或学习的运动模型往往不准确——由于数据有限、简化假设、未建模的效应等——这可能导致不安全的执行或任务失败。此外，当机器人与障碍物接触时，其未来配置的分布可能变得多模态或不连通，或者位于比可能的机器人配置空间更低内在维度的流形上。我们的方法使用系统过渡的校准数据集来局部校准运动不确定性估计，构建保证以用户设定概率包含未来机器人配置的区域。我们的校准过程捕捉了接触丰富和无接触运动之间运动不确定性的变化，从而在两种情况下都能提供足够的覆盖。我们在两个模拟规划任务上评估了我们的方法：控制弹珠在迷宫中移动和使用操纵器进行紧容差的孔中插入。与相关基线相比，CaPTURe在接触内外都达到了用户指定的覆盖要求，并在任务成功率上相较于最佳基线实现了高达30%的绝对提升。
cs.RO / 56 / 2608.09196
SAIN: Structure-Aware Interactive Navigation with Active Dialogue Grounding for Mobile Robot
SAIN：具有主动对话基础的结构感知交互导航用于移动机器人
Abstract
Most existing vision-language navigation tasks assume that instructions are complete and unambiguous. However, real-world robots often encounter natural human instructions that are ambiguous, underspecified, or incomplete, requiring them to resolve such uncertainties through active questioning. Interactive Instance Goal Navigation (IIGN) requires an embodied agent to find the specific instance under an ambiguous category-level instruction through active dialogue. However, existing dialogue-enabled methods often consume oracle answers as transient textual context for immediate decisions, rather than persistent spatial or object-centric structured state. We present SAIN, a zero-shot framework that turns active dialogue into persistent navigation state. Instead of consuming oracle answers as one-step text hints, SAIN compiles them into target evidence, route-level corridor memory, and object-candidate labels. These states are stored in structured value, room, graph, and object memories, then consumed by a unified policy for frontier ranking and final target approach. On the VL-LN IIGN benchmark, SAIN improves SR from 20.2 to 25.4 and SPL from 13.07 to 14.17 over the strongest reported dialogue-enabled baseline, while requiring no task-specific policy training. The results support dialogue-to-state conversion as an effective zero-shot mechanism for long-horizon interactive instance navigation. Project website: https://zorattc.github.io/SAIN/
Chinese Translation
大多数现有的视觉-语言导航任务假设指令是完整且明确的。然而，现实世界中的机器人经常遇到模糊、不明确或不完整的人类自然指令，这要求它们通过主动提问来解决这些不确定性。交互实例目标导航（Interactive Instance Goal Navigation, IIGN）要求具身代理在模糊的类别级指令下，通过主动对话找到特定实例。然而，现有的对话启用方法通常将oracle答案作为瞬时文本上下文用于即时决策，而不是持久的空间或以对象为中心的结构状态。我们提出了SAIN，一个零-shot框架，将主动对话转化为持久的导航状态。SAIN不是将oracle答案作为一步文本提示来消耗，而是将其编译为目标证据、路线级走廊记忆和对象候选标签。这些状态以结构化的值、房间、图形和对象记忆的形式存储，然后由统一策略用于前沿排名和最终目标接近。在VL-LN IIGN基准上，SAIN将成功率（SR）从20.2提升至25.4，将成功路径长度（SPL）从13.07提升至14.17，相较于最强的报告对话启用基线，同时不需要特定任务的策略训练。这些结果支持对话到状态转换作为一种有效的零-shot机制，用于长时间交互实例导航。项目网站：https://zorattc.github.io/SAIN/
cs.RO / 57 / 2608.09198
Ultra-Low-Impedance Robotic Gripper for High-Bandwidth and Transparent Physical Interaction
超低阻抗机器人抓手用于高带宽和透明的物理交互
Abstract
Conventional robotic grippers often use high-ratio transmissions to generate grasping torque and external force sensors to measure physical interaction. High-ratio transmissions increase friction, reflected inertia, and mechanical impedance, while external sensors add hardware complexity. To address these trade-offs, this study proposes a novel 9-DOF, three-fingered Differential Direct-Drive (DDD) gripper that combines DD motors with a low-ratio (1:2) differential transmission. The mechanism centralizes actuator mass at the base to minimize moving-link inertia, while the differential architecture couples two motors in parallel to amplify torque during flexion. Experiments show that the prototype delivers a nominal grasping force of approximately 18 N and a fingertip force of 4.7 N, while maintaining a low motor contribution to system inertia (0.236%) and low passive mechanical impedance, with a maximum measured value of 50.1 N/m when the motors are unpowered. The proposed hardware addresses the trade-offs among torque, physical transparency, and kinematic dexterity, providing a foundation for high-bandwidth interaction and sensorless proprioceptive force estimation.
Chinese Translation
传统的机器人抓手通常使用高比传动来产生抓取扭矩，并使用外部力传感器来测量物理交互。高比传动增加了摩擦、反射惯性和机械阻抗，而外部传感器则增加了硬件复杂性。为了解决这些权衡，本研究提出了一种新型的9自由度三指差动直接驱动（Differential Direct-Drive, DDD）抓手，该抓手将直接驱动电机与低比（1:2）差动传动相结合。该机制将驱动器的质量集中在基座，以最小化移动连杆的惯性，同时差动结构将两个电机并联耦合，以在屈曲过程中放大扭矩。实验表明，原型提供了约18 N的名义抓取力和4.7 N的指尖力，同时保持了对系统惯性的低电机贡献（0.236%）和低被动机械阻抗，在电机未通电时最大测量值为50.1 N/m。所提出的硬件解决了扭矩、物理透明性和运动灵活性之间的权衡，为高带宽交互和无传感器本体感知力估计提供了基础。
cs.RO / 58 / 2608.09258
Task-Oriented Formation Decision via Reinforcement Learning: Herding an Attacking Swarm
基于强化学习的任务导向编队决策：引导攻击性群体
Abstract
Multi-robot systems can accomplish tasks that are difficult for a single robot by organizing into task-specific formations. Different from existing studies on multi-robot shape formation, we here study the task-oriented formation decision problem, with a focus on the herding task. This task is challenging due to the attackers' superior maneuverability and their unknown strategies. To address these challenges, we propose the following novel results. First, we encode the formation shape using a low-dimensional parameter vector. This parametric representation reformulates the formation decision as a parameter optimization problem, thereby resolving the limited flexibility of predefined shapes. By optimizing these formation parameters, the defenders' maneuverability disadvantage is mitigated through a formation shape that continuously adapts to task requirements. Second, we develop a reinforcement learning-based policy to regulate the formation parameters. Trained offline in simulations covering diverse attacking strategies, the learned policy can effectively handle adversarial unpredictability during online deployment. Comparative simulations against three baselines demonstrate that our method can successfully accomplish challenging herding tasks. Additional scalability simulations further verify its applicability to simulated scenarios involving dozens of robots. We also validate the practical feasibility of our method on a physical robotic platform with 3 attackers and 7 defenders.
Chinese Translation
多机器人系统通过组织成特定任务的编队，可以完成单个机器人难以实现的任务。与现有的多机器人形状编队研究不同，我们在此研究任务导向的编队决策问题，重点关注引导任务。该任务具有挑战性，因为攻击者具有更强的机动性和未知的策略。为了解决这些挑战，我们提出了以下新颖的结果。首先，我们使用低维参数向量编码编队形状。这种参数化表示将编队决策重新表述为参数优化问题，从而解决了预定义形状灵活性有限的问题。通过优化这些编队参数，防御者的机动性劣势得以通过一种不断适应任务要求的编队形状得到缓解。其次，我们开发了一种基于强化学习的策略来调节编队参数。在涵盖多种攻击策略的仿真中进行离线训练后，学习到的策略能够有效应对在线部署中的对抗性不可预测性。与三个基线进行的比较仿真表明，我们的方法能够成功完成具有挑战性的引导任务。额外的可扩展性仿真进一步验证了其在涉及数十个机器人的模拟场景中的适用性。我们还在一个物理机器人平台上验证了我们方法的实际可行性，该平台上有3个攻击者和7个防御者。
cs.RO / 59 / 2608.09298
WorldSimProbe: Diagnosing Simulator Faithfulness in Action-Conditioned World Models for Embodied Manipulation
WorldSimProbe：诊断用于具身操控的动作条件世界模型的模拟器可信度
Abstract
Action-conditioned world models (ACWMs) promise to provide embodied AI with scalable predictive simulators for planning, policy evaluation, and data generation. Realizing this promise requires precise action-conditioned transitions rather than merely plausible outputs. Yet their applicability remains difficult to establish because prevailing evaluations emphasize visual quality, task outcomes, or coarse rollout-level responsiveness without directly testing simulator fidelity. To address this gap, we evaluate ACWMs through the observable capabilities expected of physical simulators. Accordingly, we formalize Observable Simulator Contract, a minimal contract that any action-conditioned physical simulator should satisfy: supplied actions must induce corresponding agent motion, and environment responses must be grounded in that realized motion. To operationalize this contract, we introduce WorldSimProbe, comprising five controlled suites spanning local control sensitivity, global trajectory variation, source-diverse actions, interaction grounding, and dynamics. Suite-specific evaluators assess simulator-relative calibration, dense action-to-motion correspondence, false-interaction grounding, and primitive-level dynamics. We evaluate six open-source ACWMs on more than 18,000 instances across RoboTwin, ManiSkill, and LIBERO. World-SimProbe reveals systematic action-realization degradation across control variation, structured failures in interaction grounding and dynamics, and benchmark signals consistent with human judgments and downstream outcomes. Together, this capability-based framework provides a transparent, and standardized paradigm for diagnosing ACWM simulator fidelity beyond coarse, task-directed evaluation.
Chinese Translation
动作条件世界模型（ACWMs）有望为具身人工智能提供可扩展的预测模拟器，以用于规划、策略评估和数据生成。实现这一承诺需要精确的动作条件转变，而不仅仅是可信的输出。然而，由于现有评估强调视觉质量、任务结果或粗略的展开级响应，而未直接测试模拟器的可信度，其适用性仍然难以确立。为了解决这一问题，我们通过物理模拟器预期的可观察能力来评估ACWMs。因此，我们形式化了可观察模拟器契约（Observable Simulator Contract），这是任何动作条件物理模拟器应满足的最小契约：所提供的动作必须引发相应的代理运动，环境响应必须基于该实现的运动。为了使这一契约具备可操作性，我们引入了WorldSimProbe，包括五个控制套件，涵盖局部控制灵敏度、全局轨迹变化、源多样化动作、交互基础和动态。特定套件的评估者评估模拟器相对校准、密集的动作与运动对应关系、虚假交互基础和原始级动态。我们在RoboTwin、ManiSkill和LIBERO上对六个开源ACWMs进行了超过18,000个实例的评估。WorldSimProbe揭示了控制变化中的系统性动作实现退化、交互基础和动态中的结构性失败，以及与人类判断和下游结果一致的基准信号。总体而言，这一基于能力的框架提供了一种透明且标准化的范式，用于诊断ACWM模拟器的可信度，超越粗略的任务导向评估。
cs.RO / 60 / 2608.09303
SAFE-CHEM: Uncertainty-Aware Policy Switching for Robust Robotic Chemistry
SAFE-CHEM：面向鲁棒机器人化学的基于不确定性的策略切换
Abstract
The deployment of autonomous robotic systems in chemistry laboratories is accelerating experimental workflows and providing the foundational data for AI-driven scientific discovery. However, despite the success of data-driven methods in acquiring dexterous skills, safety remains a primary barrier to their deployment in high-risk domains, such as early-stage materials chemistry experiments. Specifically, learning-based policies frequently struggle to distinguish between safe and unsafe actions, leading to overconfident extrapolation and potentially catastrophic failures. To mitigate these safety risks, we propose SAFE-CHEM, an uncertainty-aware framework designed for robust, learning-based robotic chemists. Our approach leverages an ensemble of recurrent neural network-based imitation learning policies to quantify epistemic uncertainty online through the variance of action predictions. By characterising the success-conditioned density of this variance using kernel density estimation, we introduce a hybrid control architecture that autonomously switches from the learned policy to a deterministic, rule-based backup controller when uncertainty exceeds a calibrated safety threshold. We evaluate SAFE-CHEM across three fundamental laboratory manipulation tasks, where our empirical results demonstrate that this hybrid strategy improves overall task success rates and reduces critical safety violations compared to traditional single-policy baselines. Finally, we demonstrate the practical viability of the framework through zero-shot sim-to-real transfer onto a physical Franka Production 3 robot manipulator.
Chinese Translation
自主机器人系统在化学实验室的部署正在加速实验工作流程，并为基于人工智能的科学发现提供基础数据。然而，尽管数据驱动的方法在获取灵巧技能方面取得了成功，但安全性仍然是其在高风险领域（如早期材料化学实验）部署的主要障碍。具体而言，基于学习的策略常常难以区分安全和不安全的行动，导致过于自信的外推和潜在的灾难性失败。为降低这些安全风险，我们提出了SAFE-CHEM，一个旨在为鲁棒的基于学习的机器人化学家设计的不确定性感知框架。我们的方法利用基于递归神经网络的模仿学习策略的集成，在线量化行动预测的认知不确定性。通过使用核密度估计来表征这种方差的成功条件密度，我们引入了一种混合控制架构，当不确定性超过校准的安全阈值时，能够自主地从学习策略切换到确定性的基于规则的备份控制器。我们在三个基本实验室操作任务中评估了SAFE-CHEM，实证结果表明，与传统的单一策略基线相比，这种混合策略提高了整体任务成功率并减少了关键安全违规行为。最后，我们通过零-shot模拟到真实的转移展示了该框架在物理Franka Production 3机器人操控器上的实际可行性。
cs.RO / 61 / 2608.09333
DH-VLM: Dual-Horizon Cooperative Latent Reasoning for Autonomous Driving
DH-VLM：用于自主驾驶的双视野协作潜在推理
Abstract
Large-scale language models for autonomous driving enable enhanced global understanding and long-horizon planning. However, when deployed in isolated vehicles, limited sensing range and occlusions restrict reliable decision-making, and the substantial computational and latency overhead makes on-board deployment impractical. Cooperative driving provides a potential solution by leveraging external agents for information exchange, but existing methods remain limited in semantic reasoning capability under practical constraints. To address these challenges, we propose DH-VLM, a dual-horizon cooperative latent reasoning framework that enables asymmetric semantic cooperation between the infrastructure and ego vehicle. The infrastructure aggregates multi-layer hidden states to form a global-reasoning horizon latent guidance, which is integrated into the ego model through an Infrastructure-Driven Latent Evolution mechanism for conditional latent refinement. This enables the ego vehicle to leverage long-range contextual understanding while preserving autonomous decision-making within its local planning horizon. Furthermore, we construct a cooperation-oriented question-answer (QA) dataset covering fundamental scene understanding and ego-personalized comprehension to support counterfactual and safety-aware reasoning. Extensive experiments demonstrate that DH-VLM achieves state-of-the-art planning performance, outperforming the previous state of the art by 14.6% in L2 error and 26.9% in collision rate. Compared with query-based end-to-end cooperative driving methods, our approach reduces the communication cost by 57.3% and GPU memory usage by 25.5%, while maintaining strong robustness against infrastructure guidance errors, providing a practical and robust paradigm for cooperative autonomous driving.
Chinese Translation
大规模语言模型在自主驾驶中能够增强全球理解和长远规划。然而，当部署在孤立的车辆中时，有限的感知范围和遮挡物限制了可靠的决策能力，而巨大的计算和延迟开销使得车载部署变得不切实际。协作驾驶通过利用外部代理进行信息交换提供了一种潜在的解决方案，但现有方法在实际约束下的语义推理能力仍然有限。为了解决这些挑战，我们提出了DH-VLM，一种双视野协作潜在推理框架，能够实现基础设施与自我车辆之间的不对称语义协作。基础设施聚合多层隐藏状态，形成全球推理视野潜在引导，通过基础设施驱动的潜在演化机制集成到自我模型中，以实现条件潜在细化。这使得自我车辆能够利用长距离上下文理解，同时在其局部规划视野内保持自主决策。此外，我们构建了一个面向合作的问题-答案（QA）数据集，涵盖基础场景理解和自我个性化理解，以支持反事实和安全意识推理。大量实验表明，DH-VLM在规划性能上达到了最先进的水平，在L2误差上比之前的最先进技术提高了14.6%，在碰撞率上提高了26.9%。与基于查询的端到端协作驾驶方法相比，我们的方法将通信成本降低了57.3%，GPU内存使用降低了25.5%，同时在基础设施引导错误下保持了强大的鲁棒性，为协作自主驾驶提供了一个实用且稳健的范式。
cs.RO / 62 / 2608.09381
JEPA-WAM: Learning Vision-Language-Action Policies with Joint-Embedding World Modeling
JEPA-WAM：通过联合嵌入世界建模学习视觉-语言-动作策略
Abstract
Robust robot control benefits from explicitly modeling state transitions, but video-generation world action models (WAMs) introduce substantial deployment cost. Existing latent WAMs avoid explicit future generation, but often compress predictive representations or separate predictive modeling from the representations used for action generation. We introduce JEPA-WAM, a latent WAM built in a pretrained V-JEPA space, which couples latent transition prediction with continuous action generation through a shared predictor. JEPA-WAM predicts a spatially structured joint current-future target that captures task-shared visual temporal structure between current and future observations, while preserving dense patch-level correspondence. Through the shared predictor, transition supervision directly shapes the backbone, from which dedicated representations are extracted for action prediction. The same design can also be instantiated in pretrained VLA policies while preserving their original perception and action pathways. On LIBERO-Plus, JEPA-WAM achieves 79.2%, the best result without large-scale robot-policy pretraining, while its pretrained $\pi_{0.5}$ instantiation reaches 86.3%, achieving the best overall performance. Experiments on RoboTwin 2.0 and real-world bimanual manipulation further demonstrate strong generalization under visual and spatial shifts.
Chinese Translation
稳健的机器人控制受益于对状态转移的明确建模，但视频生成世界动作模型（WAMs）引入了可观的部署成本。现有的潜在WAMs避免了明确的未来生成，但通常会压缩预测表示或将预测建模与用于动作生成的表示分离。我们提出了JEPA-WAM，这是一种构建于预训练V-JEPA空间中的潜在WAM，它通过共享预测器将潜在转移预测与连续动作生成结合在一起。JEPA-WAM预测一个空间结构化的联合当前-未来目标，该目标捕捉当前和未来观察之间的任务共享视觉时间结构，同时保持密集的补丁级对应关系。通过共享预测器，转移监督直接影响主干，从中提取专门的表示用于动作预测。相同的设计也可以在预训练的VLA策略中实例化，同时保留其原始的感知和动作路径。在LIBERO-Plus上，JEPA-WAM达到了79.2%的最佳结果，而其预训练的$ ext{π}_{0.5}$实例则达到了86.3%，实现了最佳的整体性能。在RoboTwin 2.0和真实世界双手操作的实验进一步展示了在视觉和空间变化下的强泛化能力。
cs.RO / 63 / 2608.09402
Beyond the Plane: Coupling Planar Vehicle Dynamics with Three-Dimensional Road Geometry
超越平面：将平面车辆动力学与三维道路几何相耦合
Abstract
Simulation is crucial for developing and testing autonomous driving systems. In particular, the development of localization and control algorithms relies on an accurate vehicle dynamics simulation. However, most vehicle dynamics models are two-dimensional while real-world roads are three-dimensional. For example, effects from the three-dimensional road geometry on the Las Vegas Motor Speedway can increase the normal forces on the tires by more than 66% compared to the nominal load at standstill. As a result, even highly detailed planar vehicle dynamics models struggle to accurately reproduce the real vehicle's behavior. While solutions for three-dimensional vehicle dynamics exist, they are rarely adopted, computationally expensive, and complex. To address this issue, we present a novel method to couple planar vehicle dynamics models with real-world three-dimensional road geometry. We transform the planar vehicle state from the vehicle model's two-dimensional plane to its corresponding representation in three-dimensional space. Additionally, we calculate road-geometry-induced forces and moments and apply them to the planar vehicle model. We validate our approach using high-speed data recorded with a full-scale race car on the banked Las Vegas Motor Speedway. Furthermore, on synthetic tracks, we show that our method yields accurate results even in edge cases. Together, our results demonstrate that the gap between planar simulation and real-world three-dimensional roads can be closed without abandoning simpler planar models. To simplify adoption of our method, we provide the implementation as open-source software on github.com/TUMFTM/3d-road-geometry-coupling.
Chinese Translation
仿真对于开发和测试自动驾驶系统至关重要。特别是，定位和控制算法的开发依赖于准确的车辆动力学仿真。然而，大多数车辆动力学模型是二维的，而现实世界的道路是三维的。例如，拉斯维加斯赛车场的三维道路几何对轮胎的法向力影响可使其超过静止状态下名义载荷的66%。因此，即使是高度详细的平面车辆动力学模型也难以准确再现真实车辆的行为。虽然存在三维车辆动力学的解决方案，但它们很少被采用，且计算成本高、复杂。为了解决这一问题，我们提出了一种新方法，将平面车辆动力学模型与现实世界的三维道路几何相耦合。我们将车辆模型的二维平面上的平面车辆状态转换为其在三维空间中的相应表示。此外，我们计算道路几何引起的力和力矩，并将其应用于平面车辆模型。我们使用在倾斜的拉斯维加斯赛车场上用全尺寸赛车记录的高速数据验证了我们的方法。此外，在合成轨道上，我们展示了即使在极端情况下，我们的方法也能产生准确的结果。我们的结果表明，平面仿真与现实世界三维道路之间的差距可以在不放弃更简单的平面模型的情况下缩小。为了简化我们方法的采用，我们将其实现作为开源软件发布在github.com/TUMFTM/3d-road-geometry-coupling上。
cs.RO / 64 / 2608.09410
Skills in Weights, Memory in Code: Hybrid Learning for Memory-Dependent Robot Manipulation
权重中的技能，代码中的记忆：用于依赖记忆的机器人操作的混合学习
Abstract
Modern vision-language-action (VLA) policies have acquired broad manipulation skills, but typically generate each action chunk from the current observation or a short fixed-length history. However, real-world manipulation is often non-Markovian, requiring robots to retain and reason over task-relevant information from long-horizon interaction histories to determine the next action. To address this challenge, we propose HyMeS, a hybrid learning framework that leverages the reasoning and memory-management capabilities of coding agents to steer a Markovian VLA for memory-dependent manipulation. Specifically, HyMeS learns low-level motor skills through gradient-based imitation learning, while a coding agent acquires high-level memory-management strategies through heuristic learning by iteratively updating an executable heuristic system from rollout feedback. Furthermore, we close the loop between steering and execution through multimodal stage-completion verification, which updates memory using proprioceptive signals and multi-frame VLM judgments. Compared with end-to-end memory-augmented VLAs, HyMeS requires demonstrations only for reusable motor skills rather than for every history-dependent task configuration, enabling data-efficient compositional generalization. On RoboMemArena, HyMeS improves mean cumulative success from 52.5% to 66.2% and mean task success from 41.3% to 60.1% over pi0.5, while outperforming PrediMem by 4.5 points in cumulative success and 14.5 points in task success.
Chinese Translation
现代视觉-语言-动作（VLA）策略已获得广泛的操作技能，但通常仅根据当前观察或短期固定长度历史生成每个动作块。然而，现实世界中的操作往往是非马尔可夫的，这要求机器人从长时间的交互历史中保留和推理与任务相关的信息，以决定下一个动作。为了解决这一挑战，我们提出了HyMeS，一个混合学习框架，利用编码代理的推理和记忆管理能力来引导用于依赖记忆的操作的马尔可夫VLA。具体而言，HyMeS通过基于梯度的模仿学习学习低级运动技能，而编码代理通过从回放反馈中迭代更新可执行启发式系统来获取高级记忆管理策略。此外，我们通过多模态阶段完成验证闭合引导与执行之间的循环，该验证使用本体感知信号和多帧VLM判断来更新记忆。与端到端的记忆增强VLA相比，HyMeS仅需对可重用的运动技能进行演示，而不是对每个历史依赖的任务配置进行演示，从而实现数据高效的组合泛化。在RoboMemArena上，HyMeS将平均累积成功率从52.5%提高到66.2%，将平均任务成功率从41.3%提高到60.1%，同时在累积成功率上超越PrediMem 4.5个百分点，在任务成功率上超越14.5个百分点。
cs.RO / 65 / 2608.09448
VANE: Reliable Test-Time Training for Vision-Language-Action Models via Future Visual Representation Prediction
VANE：通过未来视觉表征预测实现视觉-语言-动作模型的可靠测试时训练
Abstract
Test-time training (TTT) offers a lightweight way to adapt vision--language--action (VLA) policies from unlabeled deployment streams, but it remains difficult to use reliably in closed-loop manipulation. A shared adaptation space can mix incompatible task corrections, while an online update can alter subsequent actions before its consequences are known. We introduce a reliable TTT framework for VLA policies (VANE). VANE conditions prompt adaptation on the current vision--language context and learns from the future visual consequences of executed actions. Candidate updates are isolated from the live policy, evaluated on subsequent observations, and committed only when supported by future evidence, making adaptation selective and reversible. On SimplerEnv WidowX, VANE improves average success by $3.2$ percentage points over the corresponding TTT baseline. Results on Google Robot further show that deployment-time gains remain task- and embodiment-dependent. Together, these results demonstrate a constrained, evidence-based approach to adapting VLA policies during interaction.
Chinese Translation
测试时训练（TTT）提供了一种轻量级的方法，用于从未标记的部署流中适应视觉-语言-动作（VLA）策略，但在闭环操作中仍然难以可靠使用。共享适应空间可能会混合不兼容的任务修正，而在线更新可能会在后果尚不明确之前改变后续动作。我们提出了一种针对VLA策略的可靠TTT框架（VANE）。VANE在当前视觉-语言上下文中条件化提示适应，并从已执行动作的未来视觉后果中学习。候选更新与实时策略隔离，在后续观察中进行评估，仅在得到未来证据支持时才进行承诺，从而使适应过程具有选择性和可逆性。在SimperEnv WidowX上，VANE的平均成功率比相应的TTT基线提高了3.2个百分点。在Google Robot上的结果进一步表明，部署时的收益仍然依赖于任务和体现。总的来说，这些结果展示了一种在交互过程中适应VLA策略的受限、基于证据的方法。
cs.RO / 66 / 2608.09484
Graph-Guided Safe Diffuser: Topological Graph Guidance for Safe Diffusion Planning
图引导安全扩散器：用于安全扩散规划的拓扑图引导
Abstract
Many diffusion-based planners enforce safety through inference-time guidance, but such interleaved trajectory deformations often degrade kinematic feasibility due to manifold rupture. We propose Graph-Guided Safe Diffuser (G2SD), a hierarchical framework that leverages a high-level topological graph planner to guide a low-level diffusion model. G2SD enforces safety at a structural level by abstracting the data manifold into a learned latent graph, on which high-level planning is performed. Continuous trajectories are generated by diffusion planners, which are conditioned on the graph node representations selected by the high-level planner. Theoretical analyses demonstrate conditions under which manifold rupture occurs in diffusion planners, and show that G2SD improves safety by reducing the constraint violation probability as the number of segments increases. Experiments demonstrate that G2SD substantially outperforms baselines, increasing goal-reaching rate without any collision from 40-50% to 98% in Maze2D navigation and also achieving superior task scores in locomotion.
Chinese Translation
许多基于扩散的规划器通过推理时引导来强制执行安全性，但这种交错的轨迹变形往往由于流形破裂而降低运动学的可行性。我们提出了图引导安全扩散器（Graph-Guided Safe Diffuser, G2SD），这是一个分层框架，利用高层次的拓扑图规划器来引导低层次的扩散模型。G2SD通过将数据流形抽象为一个学习的潜在图，在结构层面上强制执行安全性，并在此基础上进行高层次规划。扩散规划器生成的连续轨迹是基于高层规划器选择的图节点表示进行条件生成的。理论分析展示了扩散规划器中流形破裂发生的条件，并表明G2SD通过减少约束违反概率来提高安全性，随着段数的增加而改善。实验结果表明，G2SD在基准测试中表现显著优于其他方法，在Maze2D导航中无碰撞的目标到达率从40-50%提高到98%，并在运动任务中也取得了更优的任务评分。
cs.RO / 67 / 2608.09492
Rethink Before You Execute: Adaptive Execution for World Action Models
执行前的再思考：面向世界行动模型的自适应执行
Abstract
World Action Models (WAMs) jointly predict future actions and the evolution of the environment. At each inference, a WAM generates a chunk of actions and the robot executes a fixed prefix before replanning. We argue that this fixed execution horizon is poorly matched to execution dynamics: the chunk reliability varies across task stages, so when to replan depends on the result of accumulated execution, not on the step counts. We propose TempoWAM (Timing Execution by Monitoring Progress Online), a lightweight plug-and-play execution scheme for WAMs. A Recurrent Progress Monitor first estimates task progress from the current observation, task instruction, remaining actions, and execution history; and an Adaptive Execution Protocol then evaluates whether the chunk is advancing the task to decide if replanning is needed. To bridge the training-deployment gap, the protocol is calibrated by a task-dependent calibration factor with online adaptation. Experiments on LIBERO, RoboTwin, and real-world tasks show that TempoWAM consistently improves the efficiency-success trade-off of WAM execution. On real robots, it reduces WAM inferences by 26.9% on easy tasks while maintaining success, and improves success by 13.3 points on difficult tasks.
Chinese Translation
世界行动模型（World Action Models, WAMs）共同预测未来的行动和环境的演变。在每次推理中，WAM生成一组行动，机器人在重新规划之前执行一个固定的前缀。我们认为，这种固定的执行时间范围与执行动态不匹配：行动块的可靠性在任务阶段之间变化，因此何时重新规划取决于累积执行的结果，而不是步骤计数。我们提出了TempoWAM（通过在线监测进度来定时执行），这是一种轻量级的即插即用执行方案。一个递归进度监测器首先从当前观察、任务指令、剩余行动和执行历史中估计任务进度；然后自适应执行协议评估该行动块是否在推动任务进展，以决定是否需要重新规划。为了弥合训练与部署之间的差距，该协议通过任务依赖的校准因子进行校准，并进行在线适应。在LIBERO、RoboTwin和现实世界任务上的实验表明，TempoWAM始终提高了WAM执行的效率与成功率之间的权衡。在真实机器人上，它在简单任务中将WAM推理减少了26.9%，同时保持成功率，并在困难任务中将成功率提高了13.3个百分点。
cs.RO / 68 / 2608.09515
Tether-Inertial Localization for Planetary Drones
行星无人机的绳索惯性定位
Abstract
Recent developments in planetary exploration have shown the potential of Unmanned Aerial Vehicles (UAVs), such as the Ingenuity helicopter that provided valuable mapping data. However, limited payload capabilities constrain the flight times and compute available for localization, which restrict their applicability. By providing a tethered connection, issues such as battery and computational constraints are offloaded to the base rover. At the same time, the cable can be exploited for non-drifting localization. This work presents a novel Tether-Inertial Localization approach that uses tether length and angle measurements to estimate the UAV position relative to its base. The method combines a computationally efficient analytical catenary model with a Gaussian Process (GP) residual error compensation. This accounts for systematic sensor inaccuracies and model limitations. Experimental validation across circular, triangular, and figure-eight trajectories with tether lengths up to 4.5 m and a total flight time of 37 minutes demonstrates the effectiveness of the proposed approach. Using only tether-based position estimates for feedback, the analytical catenary model achieves an average RMSE of 7.4 cm, which is further reduced to 5.2 cm through GP-based residual compensation, one order of magnitude better than the state-of-the-art. These results establish Tether-Inertial Localization as a practical alternative to vision- and GNSS-based localization for Tethered Unmanned Aerial Vehicles (TUAVs).
Chinese Translation
最近的行星探索发展展示了无人机（Unmanned Aerial Vehicles, UAVs）的潜力，例如提供宝贵地图数据的Ingenuity直升机。然而，有限的有效载荷能力限制了飞行时间和用于定位的计算能力，从而限制了其适用性。通过提供一个绳索连接，电池和计算约束等问题被转移到基础探测车。同时，电缆可以被利用进行无漂移定位。本研究提出了一种新颖的绳索惯性定位方法，该方法利用绳索长度和角度测量来估计无人机相对于其基础的位置。该方法结合了一种计算效率高的解析悬链线模型和高斯过程（Gaussian Process, GP）残差补偿。这考虑了系统性传感器不准确性和模型限制。通过对绳索长度达到4.5米、总飞行时间为37分钟的圆形、三角形和八字形轨迹进行实验验证，展示了所提方法的有效性。仅使用基于绳索的位置估计进行反馈，解析悬链线模型实现了平均均方根误差（RMSE）为7.4厘米，经过基于GP的残差补偿进一步降低至5.2厘米，比现有技术提高了一个数量级。这些结果确立了绳索惯性定位作为绳索无人机（Tethered Unmanned Aerial Vehicles, TUAVs）在视觉和全球导航卫星系统（GNSS）定位之外的实用替代方案。
cs.RO / 69 / 2608.09516
HarnessWAM: Bridging Prediction and Deliberation in World Action Models
HarnessWAM：在世界行动模型中架起预测与深思的桥梁
Abstract
World Action Models (WAMs) jointly learn environmental dynamics and robot actions, introducing priors over physical evolution into embodied control. However, finite-horizon prediction and action generation are insufficient for complex embodied tasks that require global planning, cross-stage state maintenance, execution verification, and failure recovery. We refer to this mismatch as the prediction-deliberation gap of WAMs. To address this gap, we propose HarnessWAM, an agentic framework for WAMs. HarnessWAM employs a vision-language-model-based Task Manager to maintain an evidence-grounded scene belief and a structured task graph. A capability-conditioned executable-space projection further constrains open-ended semantic plans into sequences of atomic skills that satisfy task dependencies, embodiment-state constraints, and the capability boundary of the underlying WAM. During execution, HarnessWAM operates through an event-driven, dual-timescale feedback loop: a lightweight progress estimator continuously provides high-frequency execution evidence, while the Task Manager deliberates at salient milestones by jointly considering the current observation, task state, and interaction history to determine whether to advance the task, acquire additional observations, revise the plan, or initiate local recovery. This mechanism enables the robot to recover its state after a subtask failure and resume execution without discarding previously acquired scene knowledge. HarnessWAM achieves state-of-the-art full-task and subtask success rates of 59.6% and 69.9% on RoboMemArena, and an SR of 23.7% on RoboCerebra Ideal. These results demonstrate that model-external structured state maintenance and closed-loop agentic decision making can effectively extend the local control capabilities of WAMs into embodied task execution that is plannable, verifiable, and recoverable.
Chinese Translation
世界行动模型（WAMs）共同学习环境动态和机器人动作，将物理演变的先验知识引入具身控制。然而，有限的预测和动作生成对于需要全局规划、跨阶段状态维护、执行验证和故障恢复的复杂具身任务来说是不够的。我们将这种不匹配称为WAMs的预测-深思差距。为了解决这一差距，我们提出了HarnessWAM，一个针对WAMs的代理框架。HarnessWAM采用基于视觉-语言模型的任务管理器，以维护基于证据的场景信念和结构化任务图。能力条件的可执行空间投影进一步将开放式语义计划约束为满足任务依赖关系、具身状态约束和基础WAM能力边界的原子技能序列。在执行过程中，HarnessWAM通过事件驱动的双时间尺度反馈循环运作：轻量级进度估计器持续提供高频执行证据，而任务管理器在显著里程碑处进行深思，通过共同考虑当前观察、任务状态和交互历史来决定是否推进任务、获取额外观察、修订计划或启动局部恢复。该机制使机器人能够在子任务失败后恢复其状态，并在不丢弃先前获取的场景知识的情况下恢复执行。HarnessWAM在RoboMemArena上实现了59.6%的全任务成功率和69.9%的子任务成功率，在RoboCerebra Ideal上实现了23.7%的成功率。这些结果表明，模型外部的结构化状态维护和闭环代理决策能够有效扩展WAMs的局部控制能力，使其在具身任务执行中具备可规划、可验证和可恢复的特性。
cs.RO / 70 / 2608.09528
Robotic Fabric Alignment System for Sewing Using Global Local Weighted ICP
基于全局局部加权迭代最近点的缝纫机器人面料对齐系统
Abstract
Accurate fabric alignment is a critical step that must be performed before sewing. This paper presents a novel automated fabric alignment system. The system estimates the poses of top and bottom fabric panels, lying flat and wrinkle-free in arbitrary positions, using a new Global Local Weighted Iterative Closest Point (GLW-ICP) method. The system then manipulates the top panel to achieve precise alignment at both edges and sewing lines. Unlike conventional approaches, GLW-ICP robustly aligns both global edges and local sewing lines by globally aligning fabric edge points and locally aligning sewing line points to their corresponding CAD model points, while removing unmatched points in occluded regions. Real-world experiments with various fabric shapes show that the system consistently achieves millimeter-level alignment accuracy under both occlusion and non-occlusion conditions, demonstrating its effectiveness and suitability for automated fabric alignment in practical scenarios.
Chinese Translation
准确的面料对齐是缝纫前必须执行的关键步骤。本文提出了一种新颖的自动化面料对齐系统。该系统使用一种新的全局局部加权迭代最近点（Global Local Weighted Iterative Closest Point, GLW-ICP）方法，估计平坦且无皱的上下面料面板在任意位置的姿态。系统随后操控上层面板，以实现两侧边缘和缝纫线的精确对齐。与传统方法不同，GLW-ICP通过全局对齐面料边缘点和局部对齐缝纫线点到其对应的计算机辅助设计（CAD）模型点，稳健地对齐全局边缘和局部缝纫线，同时去除遮挡区域中不匹配的点。针对各种面料形状的实际实验表明，该系统在遮挡和非遮挡条件下始终能够实现毫米级的对齐精度，证明了其在实际场景中进行自动化面料对齐的有效性和适用性。
cs.RO / 71 / 2608.09547
Model-Based Systems Engineering Framework for SysML-Driven Design of Autonomous UAVs
基于模型的系统工程框架用于SysML驱动的自主无人机设计
Abstract
Autonomous Unmanned Aerial Vehicles (UAVs) are complex cyber-physical systems that require the coordinated integration of flight control, navigation, perception, communication, power management, and mission-level decision-making under safety, timing, and reliability constraints. However, many autonomous UAV development workflows still rely on document-centric requirements, separated architectural descriptions, and software implementation artifacts, which can lead to ambiguity, interface inconsistencies, and weak traceability during early design. This paper presents a Model-Based Systems Engineering (MBSE) design framework for the SysML-driven development of autonomous UAVs. The proposed framework uses the Systems Modeling Language (SysML) as a formal design backbone to structure UAV development across four connected layers: stakeholder requirements, functional decomposition, logical architecture, and physical/software allocation. SysML requirement diagrams, activity diagrams, block definition diagrams, internal block diagrams, state machine diagrams, and parametric diagrams are used to capture the functional, structural, behavioral, interface, and performance aspects of the UAV system. The logical architecture is then systematically mapped to a Robot Operating System 2 (ROS 2) software architecture by relating SysML blocks to ROS 2 nodes, flow ports and connectors to topics, request-response interactions to services, and goal-oriented behaviors to actions. The framework is illustrated at the design level using representative autonomous UAV mission scenarios, including autonomous take-off, waypoint navigation, hover stabilization, obstacle avoidance, return-to-home, and emergency handling. The resulting model supports requirement allocation, interface definition, subsystem responsibility assignment, and verification planning before simulation or physical deployment.
Chinese Translation
自主无人机（UAV）是复杂的网络物理系统，需要在安全、时效和可靠性约束下协调集成飞行控制、导航、感知、通信、电源管理和任务级决策。然而，许多自主无人机开发工作流程仍然依赖于以文档为中心的需求、分离的架构描述和软件实现工件，这可能导致早期设计中的模糊性、接口不一致性和弱追溯性。本文提出了一种基于模型的系统工程（MBSE）设计框架，用于自主无人机的SysML驱动开发。所提出的框架使用系统建模语言（SysML）作为正式设计基础，结构化无人机开发的四个相互连接的层次：利益相关者需求、功能分解、逻辑架构和物理/软件分配。使用SysML需求图、活动图、块定义图、内部块图、状态机图和参数图来捕捉无人机系统的功能、结构、行为、接口和性能方面。然后，逻辑架构通过将SysML块与Robot Operating System 2（ROS 2）节点、流端口和连接器与主题、请求-响应交互与服务以及面向目标的行为与动作进行关联，系统地映射到ROS 2软件架构。框架在设计层面上通过代表性的自主无人机任务场景进行说明，包括自主起飞、航点导航、悬停稳定、避障、返航和紧急处理。所得到的模型支持需求分配、接口定义、子系统责任分配和在仿真或实际部署之前的验证规划。
cs.RO / 72 / 2608.09556
TDMA Based Communications Control Co-Design for Cooperative Carrying: Delay Calibration and Sampling-Rate Optimization
基于TDMA的协作运输通信控制协同设计：延迟校准与采样率优化
Abstract
Multi robot teams performing cooperative transportation face a fundamental challenge: maintaining stable control while keeping communications efficient. This paper investigates how adaptive sampling time adjustment informed by measured network delay and strategic leader rotation can distribute wireless load fairly across the team. We use physics based simulation in MuJoCo with realistic wireless modeling, including time division multiple access, medium access control, jitter, queueing, and packet loss, to evaluate three control approaches: fixed sampling with static leadership, dynamic sampling with static leadership, and dynamic sampling with rotating leadership. Our results reveal an important trade off: dynamic sampling effectively reduces communications overhead without compromising control performance, while rotating the leader role meaningfully improves how fairly airtime is distributed all with negligible impact on the team carrying ability. to the best of our knowledge, being among the first to jointly examine dynamic sampling, rotating leadership, and wireless protocol interactions in physicsrealistic multi robot cooperation, this work provides practical guidance for deploying coordinated robotic teams in real world settings where communications resources are limited.
Chinese Translation
执行协作运输的多机器人团队面临一个基本挑战：在保持控制稳定的同时确保通信效率。本文研究了如何通过测量的网络延迟和战略性领导者轮换来调整自适应采样时间，从而公平地分配无线负载。我们使用MuJoCo中的基于物理的仿真，结合现实的无线建模，包括时分多址（TDMA）、媒介接入控制、抖动、排队和数据包丢失，来评估三种控制方法：静态领导下的固定采样、静态领导下的动态采样，以及轮换领导下的动态采样。我们的结果揭示了一个重要的权衡：动态采样有效减少了通信开销，而不影响控制性能，同时轮换领导角色显著改善了空中时间的公平分配，且对团队的运输能力影响微乎其微。根据我们所知，本研究是首批联合考察动态采样、轮换领导和无线协议交互在物理现实的多机器人合作中的工作之一，为在通信资源有限的现实环境中部署协调机器人团队提供了实用指导。
cs.RO / 73 / 2608.09591
FactorDrive: Adaptive Multi-Step Reasoning Driven by Planning-Critical Factors for End-to-End Autonomous Driving
FactorDrive：基于规划关键因素的自适应多步骤推理驱动的端到端自主驾驶
Abstract
Vision-language models (VLMs) have advanced scene understanding and enabled explicit reasoning in end-to-end autonomous driving. However, existing methods insufficiently integrate spatial-physical evidence into planning reasoning, while reasoning adaptation remains coarse-grained and falls short of scene-specific planning demands. Furthermore, reasoning-path optimization for higher planning quality remains largely unexplored in autonomous-driving post-training. To address these limitations, we propose FactorDrive, an end-to-end autonomous driving framework for adaptive multi-step reasoning driven by planning-critical factors (PCFs). We first perform large-scale driving-domain instruction tuning to establish foundational driving knowledge. Building on this foundation, we construct PCF-CoT, a chain-of-thought (CoT) dataset that grounds planning reasoning in trajectory-relevant spatial-physical evidence and organizes reasoning around scene-specific PCFs, enabling the composition and depth of reasoning paths to adapt to different planning demands. We further introduce Quality Search-Guided Group Relative Policy Optimization (QS-GRPO), which guides Monte Carlo Tree Search (MCTS) with trajectory-level planning rewards to discover reasoning paths with higher planning quality and uses the resulting responses to optimize the policy through GRPO, thereby improving trajectory planning performance. Extensive experiments on both open-loop (nuScenes) and closed-loop-oriented (NAVSIM) benchmarks demonstrate that FactorDrive achieves state-of-the-art planning performance.
Chinese Translation
视觉语言模型（VLMs）已推动场景理解的发展，并在端到端自主驾驶中实现了显式推理。然而，现有方法在将空间物理证据融入规划推理方面不足，同时推理适应性仍然较为粗糙，无法满足场景特定的规划需求。此外，针对更高规划质量的推理路径优化在自主驾驶后训练中仍然未得到充分探索。为了解决这些局限性，我们提出了FactorDrive，这是一个端到端自主驾驶框架，旨在通过规划关键因素（PCFs）驱动自适应多步骤推理。我们首先进行大规模驾驶领域的指令微调，以建立基础驾驶知识。在此基础上，我们构建了PCF-CoT，一个将规划推理与轨迹相关的空间物理证据相结合的思维链（CoT）数据集，并围绕场景特定的PCFs组织推理，使推理路径的组成和深度能够适应不同的规划需求。我们进一步引入了质量搜索引导的群体相对策略优化（QS-GRPO），该方法通过轨迹级规划奖励引导蒙特卡洛树搜索（MCTS），以发现具有更高规划质量的推理路径，并利用生成的响应通过GRPO优化策略，从而提高轨迹规划性能。在开放循环（nuScenes）和闭环导向（NAVSIM）基准上的大量实验证明，FactorDrive达到了最先进的规划性能。
cs.RO / 74 / 2608.09602
Nonlinear Model Predictive Control of a Robotic Soft Esophagus
机器软食管的非线性模型预测控制
Abstract
Strictures caused by esophageal cancer can narrow down the esophageal lumen, leading to dysphagia. Palliation of dysphagia has driven the development of a Robotic Soft Esophagus (RoSE), which provides a novel in vitro platform for esophageal stent testing and food viscosity studies. In RoSE, peristaltic wave generation and control were done in an open-loop manner since the conduit lacked visibility and embedded sensing capability. Hence, in this work, RoSE version 2.0 (RoSEv2.0) is designed with embedded Time Of Flight (TOF) and pressure sensors to measure conduit displacement and air pressure, respectively, for modeling and control. Model Predictive Control (MPC) of RoSEv2.0 is implemented to govern the peristalsis and air pressure profile autonomously. The implemented MPC used Sparse Identification Nonlinear Dynamics with Control (SINDYC) models to estimate the future states of ROSEv2.0. The dynamic models are discovered from the TOF and pressure sensor data. Peristalsis waves of speed 20 mm/s, wavelength 75 mm, and amplitudes 5, 7.5, and 10 mm were successfully generated by the MPC. Additionally, RoSEv2.0 with the MPC was employed to perform stent migration testing with various food bolus consistencies. The major contribution claimed in this paper is the application of SINDYC-based MPC to solve the closed-loop control problem of RoSE for achieving desired peristaltic waves.
Chinese Translation
食管癌引起的狭窄会导致食管腔道变窄，从而导致吞咽困难。缓解吞咽困难推动了机器软食管（Robotic Soft Esophagus, RoSE）的发展，该设备提供了一种新颖的体外平台，用于食管支架测试和食物粘度研究。在RoSE中，由于管道缺乏可视性和嵌入式传感能力，蠕动波的生成和控制是以开环方式进行的。因此，在本研究中，设计了RoSE 2.0（RoSEv2.0），配备了嵌入式飞行时间（Time Of Flight, TOF）和压力传感器，分别用于测量管道位移和气压，以便进行建模和控制。实现了RoSEv2.0的模型预测控制（Model Predictive Control, MPC），以自主控制蠕动和气压曲线。所实施的MPC使用稀疏识别非线性动态控制（Sparse Identification Nonlinear Dynamics with Control, SINDYC）模型来估计RoSEv2.0的未来状态。动态模型是从TOF和压力传感器数据中发现的。MPC成功生成了速度为20 mm/s、波长为75 mm、幅度为5、7.5和10 mm的蠕动波。此外，使用MPC的RoSEv2.0被用于进行不同食物团块一致性的支架迁移测试。本文的主要贡献在于应用基于SINDYC的MPC解决RoSE的闭环控制问题，以实现所需的蠕动波。
cs.RO / 75 / 2608.09653
Predictive safety filter enhanced curriculum learning control for efficient vehicle dynamics controller
增强预测安全过滤器的课程学习控制用于高效车辆动力学控制器
Abstract
Recent advances in learning-based control have enabled impressive achievements in solving complex control problems in various domains. However, since learning-based control may not be able to realize safety-guaranties, it is of great importance to enhance safety and robustness while maintaining good performances. Take vehicle motion \& dynamics control as an example, in order to overcome the pain points of traditional methods such as heavy parameter calibration effort and learning-based control to bring better performance and efficiency in stability \& agility over prior work for state-based vehicle control tasks, in this work, our method aims to develop a curriculum learning controller enhanced with physics-based predictive safety filter. The validation is conducted with the Python-CarSim platform, demonstrating better improvements and scalability under various maneuvers.
Chinese Translation
近年来，基于学习的控制方法在解决各个领域复杂控制问题方面取得了显著进展。然而，由于基于学习的控制可能无法实现安全保障，因此在保持良好性能的同时增强安全性和鲁棒性显得尤为重要。以车辆运动与动力学控制为例，为了克服传统方法在参数校准方面的繁重工作以及基于学习的控制在状态基础车辆控制任务中带来的性能和效率不足的问题，本研究旨在开发一种增强了基于物理的预测安全过滤器的课程学习控制器。验证工作在Python-CarSim平台上进行，结果表明在各种机动操作下具有更好的改进和可扩展性。
cs.RO / 76 / 2608.09658
Removing Infrastructure Barriers in Human-Robot Collaboration Through Wireless Reconfigurable Cells
通过无线可重构单元消除人机协作中的基础设施障碍
Abstract
Human-Robot Collaboration (HRC) plays a vital role in dynamic, high mix, low volume industrial scenarios such as remanufacturing, which frequently face workcell rearrangements. Traditional setups are constrained by power and data cabling, restricting modularity and reconfigurations, while the selection of commercial wireless devices suitable for real-time perception and safe collaboration are limited in availability. This paper presents a highly flexible, wireless, 5G-based system that serves as a versatile experimental testbed for applications including remanufacturing, operator training, and user studies. To eliminate infrastructure barriers, the workcell integrates a novel battery-powered, multi-sensor platform prototype. Additionally, to support operator safety and system adaptability across environmental shifts, the system integrates a computer vision module for object detection and pose estimation, further augmented for robust hand recognition. Trained on synthetic and real data, the model reliably detects oriented grasping poses and human hands across varying lighting and background conditions (with an mAP@50-95 of 97.74 +- 0.10% and a mean inference time of 12.5 ms). Offloading these computationally intensive tasks to the edge via 5G, the proposed architecture contributes to resolving the bandwidth-latency trade-off. To demonstrate portability, the system was implemented in both Hungary and Norway, and was evaluated across a combination of public and private, Standalone and Non-Standalone 5G infrastructures. The performed network experiments produced results in round-trip response times down to 12 ms in case of compatible network-device pairings, suitable for safe, adaptive HRC. However, these measurements also revealed practical limitations related to interoperability in current 5G deployments that should be addressed in future works.
Chinese Translation
人机协作（HRC）在动态、高混合、低产量的工业场景中发挥着至关重要的作用，例如再制造，这些场景经常面临工作单元的重新配置。传统设置受到电源和数据电缆的限制，限制了模块化和重新配置的能力，而适用于实时感知和安全协作的商业无线设备的选择也十分有限。本文提出了一种高度灵活的、基于5G的无线系统，作为再制造、操作员培训和用户研究等应用的多功能实验测试平台。为了消除基础设施障碍，该工作单元集成了一种新型的电池供电多传感器平台原型。此外，为了支持操作员安全和系统在环境变化中的适应性，该系统集成了一个用于物体检测和姿态估计的计算机视觉模块，并进一步增强了手部识别的鲁棒性。该模型在合成和真实数据上训练，能够在不同的光照和背景条件下可靠地检测定向抓取姿势和人手（mAP@50-95为97.74 ± 0.10%，平均推理时间为12.5毫秒）。通过5G将这些计算密集型任务卸载到边缘，所提架构有助于解决带宽与延迟之间的权衡。为了展示可移植性，该系统在匈牙利和挪威进行了实施，并在公共和私人、独立和非独立的5G基础设施组合中进行了评估。所进行的网络实验在兼容的网络设备配对情况下，往返响应时间最低可达12毫秒，适合安全、适应性的人机协作。然而，这些测量也揭示了当前5G部署中与互操作性相关的实际限制，这些问题应在未来的工作中加以解决。
cs.RO / 77 / 2608.09731
TAMS: Task-Aware Multi-View Adaptive Streaming for Wireless Telerobotic Manipulation
TAMS：面向任务的多视角自适应流媒体传输用于无线遥操作
Abstract
Wireless telerobotic manipulation relies on timely multi-view video feedback, but the available uplink bandwidth is often limited and dynamic. This paper presents Task-Aware Multi-View Adaptive Streaming (TAMS), a system that allocates video bitrate according to the current manipulation phase. TAMS infers task phase from lightweight robot-side signals and prioritizes the camera view most relevant to the operator while preserving baseline visibility for secondary views. Experiments on a six-degree-of-freedom (6-DoF) teleoperation testbed under three constrained network conditions show that TAMS improves primary view Structural Similarity Index (SSIM), reduces task completion time, and increases trial success rate compared with equal and static allocation baselines. Under the most constrained bandwidth condition, TAMS reduces mean completion time from 68.9 s to 43.9 s relative to equal allocation and increases trial success rate from 48% to 71%. Code is available at: https://github.com/Dzxx623/TAMS.
Chinese Translation
无线遥操作依赖于及时的多视角视频反馈，但可用的上行带宽通常有限且动态变化。本文提出了面向任务的多视角自适应流媒体传输系统（Task-Aware Multi-View Adaptive Streaming, TAMS），该系统根据当前的操作阶段分配视频比特率。TAMS 从轻量级的机器人端信号中推断任务阶段，并优先考虑与操作员最相关的摄像头视角，同时保留次要视角的基本可见性。在三种受限网络条件下的六自由度（6-DoF）遥操作测试平台上的实验表明，与均等和静态分配基线相比，TAMS 提高了主要视角的结构相似性指数（Structural Similarity Index, SSIM），减少了任务完成时间，并提高了试验成功率。在最受限的带宽条件下，TAMS 将平均完成时间从 68.9 秒减少到 43.9 秒，相较于均等分配，并将试验成功率从 48% 提高到 71%。代码可在以下链接获取：https://github.com/Dzxx623/TAMS。
cs.RO / 78 / 2608.09762
Efficient Real-World Online Reinforcement Learning for Robot Manipulation via Centralized Training and Critic Decomposition
通过集中训练和评论员分解实现高效的现实世界在线强化学习用于机器人操作
Abstract
Real-world online reinforcement learning (RL) provides a promising approach for training robotic manipulation policies directly in the physical world, avoiding the sim-to-real gap and enabling continuous policy refinement through human-in-the-loop interaction. Recent methods have demonstrated sample-efficient learning through human intervention but remain limited to small randomization ranges and encounter challenges with the non-stationarity induced by concurrently training multiple agents. To address these limitations, we introduce a unified framework that combines centralized training with decentralized execution (CTDE) and a Hybrid Reward Architecture (HRA). This enables multiple actors to share a centralized multi-head critic. The critic is decomposed into task and grasp heads, corresponding to the sparse task reward and a potential-based grasping reward, respectively. We accordingly reformulate the critic and actor objectives to exploit the decomposed Q-values while explicitly accounting for the categorical action distribution of the discrete gripper policy. Experimental results demonstrate that the proposed framework substantially improves both sample efficiency and policy performance. We validate our approach on two robotic arms and a simulated humanoid robot across tennis ball and banana pick-and-place, pot reset, and simulated block relocation tasks under dimension-wise domain randomization, approximately 5-25x larger than those considered in prior work. Compared with a state-of-the-art baseline, our method improves the success rate from 60% to 80% on tennis ball pick-and-place, from 60% to 90% on banana pick-and-place, and from 25% to 95% on simulated block relocation, while also successfully accomplishing a task where the baseline consistently fails. Videos and more details are available at our project website: https://hil-harc.github.io/.
Chinese Translation
现实世界在线强化学习（RL）为在物理世界中直接训练机器人操作策略提供了一种有前景的方法，避免了模拟与现实之间的差距，并通过人机交互实现持续的策略优化。最近的方法通过人类干预展示了样本高效学习，但仍然局限于小范围的随机化，并面临由于同时训练多个代理而引发的非平稳性挑战。为了解决这些限制，我们提出了一个统一框架，将集中训练与分散执行（CTDE）和混合奖励架构（HRA）相结合。这使得多个参与者可以共享一个集中式的多头评论员。评论员被分解为任务头和抓取头，分别对应于稀疏的任务奖励和基于潜力的抓取奖励。因此，我们相应地重新制定了评论员和参与者的目标，以利用分解的Q值，同时明确考虑离散抓取策略的分类动作分布。实验结果表明，所提出的框架显著提高了样本效率和策略性能。我们在两个机器人手臂和一个模拟人形机器人上验证了我们的方法，涉及网球和香蕉的抓取与放置、盆栽重置以及模拟块重定位任务，随机化范围在维度上大约是之前工作的5-25倍。与最先进的基线相比，我们的方法在网球抓取与放置任务中的成功率从60%提高到80%，在香蕉抓取与放置任务中的成功率从60%提高到90%，在模拟块重定位任务中的成功率从25%提高到95%，同时成功完成了基线始终失败的任务。视频和更多细节可在我们的项目网站上找到：https://hil-harc.github.io/
cs.RO / 79 / 2608.09771
SLIM-0.5B: Learning Action-Grounded Predictive Latents for Robot Manipulation
SLIM-0.5B：学习基于动作的预测潜变量用于机器人操作
Abstract
Vision-language-action policies rely on large multimodal backbones to jointly perform perception, language conditioning, and action generation at every control step. Much of this capacity supports open-domain semantics, whereas continuous robot manipulation primarily requires compact representations of observations, actions, and the transitions induced by actions. Pixel-level world models provide another route, but predicting visual details irrelevant to control can be unnecessarily expensive. We propose SLIM (Self-supervised Latent Interaction Model), a compact 0.5B-parameter latent interaction policy. SLIM learns action-grounded predictive latents that capture both action-conditioned future transitions and the actions that explain observed changes. SLIM learns these representations through self-supervised masked trajectory prediction, combining action reconstruction with future-latent prediction. A compact Mixture-of-Transformers (MoT) backbone models interactions between observation latents and action tokens. The resulting policy is trained with flow matching for language-conditioned action generation. Across simulation benchmarks and real-world evaluation, SLIM matches or exceeds representative large-scale VLA and world-action-model baselines with fewer parameters, no additional embodied pretraining, lower inference latency, and substantially lower GPU memory usage.
Chinese Translation
视觉-语言-动作策略依赖于大型多模态骨干网络，在每个控制步骤中共同执行感知、语言条件和动作生成。这种能力的大部分支持开放域语义，而连续的机器人操作主要需要对观察、动作及由动作引起的转变进行紧凑表示。像素级世界模型提供了另一种途径，但预测与控制无关的视觉细节可能会造成不必要的开销。我们提出了SLIM（自监督潜在交互模型），一个紧凑的0.5B参数潜在交互策略。SLIM学习基于动作的预测潜变量，捕捉动作条件下的未来转变以及解释观察到变化的动作。SLIM通过自监督的掩蔽轨迹预测学习这些表示，结合动作重构与未来潜变量预测。一个紧凑的混合变换器（Mixture-of-Transformers，MoT）骨干网络建模观察潜变量与动作标记之间的交互。最终的策略通过流匹配进行训练，以实现语言条件下的动作生成。在模拟基准和现实世界评估中，SLIM在参数更少、没有额外的具身预训练、较低的推理延迟和显著较低的GPU内存使用情况下，与代表性的大规模视觉-语言-动作（VLA）和世界-动作模型基线相匹配或超越。
cs.RO / 80 / 2608.09778
RoboSeg: Online Part-Level Semantic Reconstruction for Robotic Manipulation via a Single Eye-in-Hand Camera
RoboSeg：通过单目手持相机实现机器人操作的在线部件级语义重建
Abstract
Robotic manipulation requires perception systemsthat identify actionable parts such as handles, rims, triggers,and tool tips, not merely object categories or point clouds. This paper presents RoboSeg, a part-level semantic reconstructionsystem that links vision-language model (VLM) functional-partdiscovery, asynchronous online RGB-D semantic reconstruc-tion, and task-oriented grasp generation without requiring CAD models or pre-scanned meshes. RoboSeg queries a VLM onthe initial RGB observation to obtain compact functional part prompts, then scans with two asynchronous streams: a high-frequency geometry thread for RGB-D odometry and truncated signed distance function (TSDF) fusion, and a keyframe-triggered semantic thread for SAM3 part masks. Projectedmasks are fused by voxel-level temporal voting into a persistentpart-labeled point cloud; RoboSeg uses this map to assign AnyGrasp 6-DoF candidates to semantic parts and select grasps consistent with the task-relevant part label. RoboSeg reaches 83.4% mean part intersection-over-union (mIoU) over manually labeled objects; in a 24-trial physical pilot across fourobjects and eight tasks, the selected grasp contacts the requestedpart in all trials and achieves 21/24 combined task successes.These results characterize RoboSeg as a semantic indexing layerfor task-conditioned manipulation, with AnyGrasp retained asthe proposal generator.
Chinese Translation
机器人操作需要感知系统识别可操作的部件，如手柄、边缘、扳机和工具尖端，而不仅仅是物体类别或点云。本文提出了RoboSeg，一种部件级语义重建系统，它将视觉-语言模型（VLM）功能部件发现、异步在线RGB-D语义重建和任务导向的抓取生成相结合，而无需CAD模型或预扫描的网格。RoboSeg在初始RGB观察上查询VLM以获得紧凑的功能部件提示，然后通过两个异步流进行扫描：一个用于RGB-D里程计和截断符号距离函数（TSDF）融合的高频几何线程，以及一个用于SAM3部件掩模的关键帧触发语义线程。投影掩模通过体素级时间投票融合成一个持久的部件标记点云；RoboSeg利用该地图将AnyGrasp 6自由度候选抓取分配给语义部件，并选择与任务相关部件标签一致的抓取。RoboSeg在手动标记的物体上达到了83.4%的平均部件交并比（mIoU）；在针对四个物体和八个任务的24次物理试验中，所选抓取在所有试验中都接触到请求的部件，并实现了21/24次综合任务成功。这些结果将RoboSeg表征为任务条件下操作的语义索引层，同时保留AnyGrasp作为提案生成器。
cs.RO / 81 / 2608.09807
WRAP: Wasserstein-Robust Adaptive Plug-in for Robot Localization
WRAP：用于机器人定位的Wasserstein鲁棒自适应插件
Abstract
Robotic localization under changing sensing conditions can suffer from biased errors and miscalibrated covariances. We present WRAP, an adapter-agnostic Wasserstein-robust plug-in for nonlinear extended Kalman filter (EKF) and error-state Kalman filter (ESKF) stacks. A causal module supplies time-varying effective process and measurement statistics; a mean-preserving Wasserstein local update then computes least-favorable covariances and a robust gain without changing the propagation model, residual, or retraction. This separates mean adaptation from covariance robustification and uses distinct radii for propagation and sensing. On 18 UWB--IMU sequences held out from adapter training, adapter-only and WRAP reduce mean 3-D position RMSE by $19.8\%$ and $27.4\%$ relative to the nominal ESKF; an isotropic ablation reaches $19.5\%$, linking the incremental gain to directional process-covariance redistribution. An in-sample GNSS--INS study shows that mean adaptation provides most of the accuracy gain, while DR improves consistency and mitigates over-tightened classical covariance estimates. The robust solve takes 0.05 ms for UWB and 2.92 ms for GNSS on a Jetson Orin Nano.
Chinese Translation
在变化的传感条件下，机器人定位可能会受到偏差错误和误校准协方差的影响。我们提出了WRAP，一种与适配器无关的Wasserstein鲁棒插件，适用于非线性扩展卡尔曼滤波器（EKF）和误差状态卡尔曼滤波器（ESKF）堆栈。一个因果模块提供时间变化的有效过程和测量统计；然后，保持均值的Wasserstein局部更新计算最不利的协方差和鲁棒增益，而不改变传播模型、残差或收缩。这将均值适应与协方差鲁棒化分开，并为传播和感知使用不同的半径。在从适配器训练中保留的18个UWB-IMU序列中，仅适配器和WRAP相对于名义ESKF分别减少了19.8%和27.4%的均值3D位置均方根误差（RMSE）；一个各向同性消融实验达到了19.5%，将增量增益与方向性过程协方差重分配联系起来。一个样本内的GNSS-INS研究表明，均值适应提供了大部分的准确性增益，而DR提高了一致性并减轻了过于紧缩的经典协方差估计。鲁棒求解在Jetson Orin Nano上对UWB耗时0.05毫秒，对GNSS耗时2.92毫秒。
cs.RO / 82 / 2608.09816
Hierarchical Fast--Slow ReAct Agent for Zero-Shot Object-Goal Navigation
层次化快速-慢速反应代理用于零-shot目标导航
Abstract
Zero-shot object-goal navigation (ZSON) requires a robot to find a named object category in a building it has never entered. The prevailing approach scores frontiers with a vision--language \emph{value map}: every decision is another argmax over the map as it currently stands, and the evidence behind that score is discarded the moment it is taken. Systems that place a large vision--language model inside the perception--action loop typically query it on a fixed schedule from the current view alone; a room the robot walked through minutes earlier is never reconsidered, and a failed call has no defined fallback. We turn what the robot has already seen into the object of deliberation. Our hierarchical fast--slow agent leaves the value-map controller running at every step and writes a \emph{coordinate-anchored memory} as it moves: a semantic grid of room types and confirmed object instances, together with a bounded store of pose-tagged keyframes. A VLM screens each candidate detection before it is written. A deliberative layer reads this memory in a bounded reason--retrieve--act loop. It wakes on structural events the reactive layer computes, reasons first over text, and recalls a first-person view only for candidates that text alone cannot separate. Per-invocation and per-run caps bound its calls, a call-free first tier resolves the most frequent stall, and any failure returns control to the reactive controller. Our system reaches 68.75% SR on HM3D v1 \texttt{val} and 47.29% on MP3D \texttt{val}, the highest success rate among the zero-shot methods compared here. Choosing among far frontiers by argmax instead of deliberating costs 3.40 SR points in a paired comparison over all 2000 HM3D episodes (95\% CI [1.70, 5.05]); deliberating over every frontier does not recover them.
Chinese Translation
零-shot目标导航（ZSON）要求机器人在从未进入过的建筑中找到一个指定的物体类别。当前的主流方法使用视觉-语言价值图（value map）对前沿进行评分：每个决策都是在当前地图上进行的另一个argmax，并且在做出决策的瞬间，支持该评分的证据被丢弃。将大型视觉-语言模型（VLM）置于感知-行动循环中的系统通常仅从当前视图中按照固定的时间表进行查询；机器人几分钟前经过的房间从未被重新考虑，而一次失败的调用没有定义的后备方案。我们将机器人已经看到的内容转变为深思熟虑的对象。我们的层次化快速-慢速代理在每一步中保持价值图控制器的运行，并在移动时写入坐标锚定记忆（coordinate-anchored memory）：一个包含房间类型和确认物体实例的语义网格，以及一个有限的姿态标记关键帧存储。VLM在写入之前筛选每个候选检测。深思熟虑层在一个有限的推理-检索-行动循环中读取此记忆。它在反应层计算的结构事件上唤醒，首先对文本进行推理，仅对文本无法区分的候选者回忆第一人称视图。每次调用和每次运行的上限限制了其调用，免调用的第一层解决了最频繁的停滞，而任何失败都将控制权返回给反应控制器。我们的系统在HM3D v1验证集上达到68.75%的成功率，在MP3D验证集上达到47.29%，在这里比较的零-shot方法中成功率最高。在所有2000个HM3D剧集的配对比较中，选择远前沿进行argmax而不是深思熟虑损失了3.40个成功率点（95% CI [1.70, 5.05]）；对每个前沿进行深思熟虑并未恢复这些成功率。
cs.RO / 83 / 2608.09853
RynnValue: Scaling Robotic Value Foundation Models with Temporal Distance
RynnValue：通过时间距离扩展机器人价值基础模型
Abstract
General-purpose reward models are increasingly the bottleneck for scaling robot learning, yet the recipe for learning value-related capabilities from large-scale heterogeneous corpora remains underexplored. Existing approaches tie supervision to task-internal anchors such as preferences or normalized progress, none of which transfer cleanly across embodiments and data sources. We introduce RynnValue, an open-source value foundation model for robotic manipulation that replaces these anchors with temporal distance, the directed cost-to-go from an observation to the language-specified goal. Because temporal-distance labels can be derived directly from timestamps, RynnValue scales to over 7,000 hours and roughly 3M instruction-conditioned clips without preference or progress annotations. To make temporal-value learning reliable at scale, we combine random temporal sampling, temporal-order shuffling, and value-isolation attention, suppressing shortcuts that would leave predictions insensitive to failures and regressions. Trained without preference labels, RynnValue attains an average Kendall's tau_a of 0.675 on RBM-EVAL-OOD, surpassing the fully preference-supervised state of the art (0.655) and more than doubling a progress-only counterpart (0.292), while generalizing zero-shot to unseen tasks, embodiments, and viewpoints. Converted into dense rewards via potential-based shaping, it raises real-world policy success from 52.5% to 72.5% online and from 63.8% to 82.5% offline. These results establish temporal distance as a scalable supervision target and practical reward interface for generalist robot policies.
Chinese Translation
通用奖励模型日益成为扩展机器人学习的瓶颈，但从大规模异构语料库中学习与价值相关的能力的方案仍未得到充分探索。现有方法将监督与任务内部锚点（如偏好或标准化进展）联系在一起，而这些锚点在不同的体现和数据源之间并不能很好地迁移。我们提出了RynnValue，一个用于机器人操控的开源价值基础模型，它用时间距离替代了这些锚点，时间距离是从观察到语言指定目标的有向成本。由于时间距离标签可以直接从时间戳中推导，RynnValue能够扩展到超过7000小时和大约300万条基于指令的片段，而无需偏好或进展注释。为了使时间价值学习在大规模上可靠，我们结合了随机时间采样、时间顺序洗牌和价值隔离注意力，抑制那些使预测对失败和回归不敏感的捷径。在没有偏好标签的情况下训练，RynnValue在RBM-EVAL-OOD上获得了平均Kendall's tau_a为0.675，超越了完全偏好监督的最新技术（0.655），并且是仅进展对照组的两倍多（0.292），同时在未见任务、体现和视角上实现零样本泛化。通过基于潜力的塑形转换为密集奖励后，它将在线真实世界策略成功率从52.5%提高到72.5%，离线成功率从63.8%提高到82.5%。这些结果确立了时间距离作为可扩展的监督目标和通用机器人策略的实用奖励接口。
cs.RO / 84 / 2608.09857
Agentic Harnesses: LLM-Driven Verification Layers for Robot Autonomy
自主代理：基于大语言模型的机器人自主性验证层
Abstract
Advances in advanced artificial intelligence tools have sparked research in robot autonomy, but the development of such systems has largely focused on execution rather than verifying the feasibility actions planning models propose. Like general-purpose LLMs, robotics planning models carry risks: biased toward user-specified goals, they may suggest actions misaligned with scientific ethics, they may be unsafe due to an inability to "remember" prior safety risks, or they may be vulnerable to adversarial attacks on the autonomy ecosystem. We propose a LLM-driven verification layer between planning and execution to evaluate action permissibility. Our LLM-as-a-Judge ensemble combines chain-of-thought reasoning across models and synthesizes those expert judge outputs, mirroring a combination of a mixture of experts and self-consistency approach. This layer serves as middleware, gating plans from the server's planning module before they reach the MCP server and therefore the robot's low-level controls: plans are approved, rejected for reformulation, or escalated for human review. With this system, we achieve near 85% precision across accept/escalate/reject categories 97% containment of adversarial attacks, with negligible errors between accepting and rejecting tasks, and errors mostly manifesting at the escalate boundary.
Chinese Translation
先进人工智能工具的进步激发了机器人自主性的研究，但此类系统的发展在很大程度上集中于执行，而非验证行动规划模型所提出的可行性。与通用的大语言模型（LLMs）一样，机器人规划模型也存在风险：由于偏向用户指定的目标，它们可能建议与科学伦理不一致的行动，可能因无法“记住”先前的安全风险而不安全，或者可能容易受到针对自主生态系统的对抗性攻击。我们提出了一种基于大语言模型的验证层，位于规划与执行之间，以评估行动的许可性。我们的LLM作为裁判的集成体结合了跨模型的思维链推理，并综合了这些专家裁判的输出，反映了专家混合和自我一致性方法的结合。该层作为中间件，在计划到达MCP服务器和机器人的低级控制之前，控制来自服务器规划模块的计划：计划被批准、拒绝以进行重新制定，或升级以供人类审查。通过该系统，我们在接受/升级/拒绝类别中实现了近85%的精确度，对抗性攻击的97%被遏制，接受和拒绝任务之间的错误微乎其微，错误主要出现在升级边界。
cs.RO / 85 / 2608.09860
Entanglement-Free Trajectory Planning for Tethered Mobile Robots with a Slack Tether
具有松弛绳索的系绳移动机器人无缠绕轨迹规划
Abstract
In motion planning algorithms for tethered mobile robots, the entanglement state of the tether is a critical aspect to consider during the planning phase. This is particularly important in case of a slack tether, where the shape of the tether is not determined solely by the geometry of the environment and the location of the obstacles, but also by the dynamics of the tether, by the trajectory followed by the robot, and possibly by exogenous forces. In this scenario, preventing entanglement requires planning a robot trajectory that accounts for the entanglement definition and for the dynamics of the robot and of the tether. In this work, we propose a motion planning algorithm for tethered mobile robots with a slack tether that computes dynamically feasible entanglement-free trajectories to navigate through an environment with static obstacles. By considering the entanglement state during all the stages of the planning pipeline, we are able to compute safer trajectories that avoid entanglement during the motion of the robot. We achieve this through a three-step pipeline, which includes (i) the construction of a topological model of the entanglement-free configuration space of the tethered robot, (ii) the generation of a set of candidate paths using this model, and (iii) the computation of a dynamically feasible entanglement-free trajectory by solving a homotopy-constrained trajectory generation problem. The resulting trajectory can then be executed to lead the robot to its target location, while maintaining the tether in an entanglement-free configuration. We demonstrate the benefits of this algorithm in simulations, where we show how the planning algorithm avoids violations of the entanglement constraints, resulting in safer and more reliable trajectories.
Chinese Translation
在系绳移动机器人的运动规划算法中，绳索的缠绕状态是规划阶段需要考虑的关键因素。特别是在绳索松弛的情况下，绳索的形状不仅由环境的几何形状和障碍物的位置决定，还受到绳索动力学、机器人所遵循的轨迹以及可能的外部力的影响。在这种情况下，防止缠绕需要规划一条考虑缠绕定义以及机器人和绳索动力学的机器人轨迹。在本研究中，我们提出了一种针对具有松弛绳索的系绳移动机器人的运动规划算法，该算法计算动态可行的无缠绕轨迹，以便在有静态障碍物的环境中导航。通过在规划流程的所有阶段考虑缠绕状态，我们能够计算出更安全的轨迹，从而在机器人运动过程中避免缠绕。我们通过一个三步流程实现这一目标，包括 (i) 构建系绳机器人无缠绕配置空间的拓扑模型，(ii) 使用该模型生成一组候选路径，以及 (iii) 通过解决同伦约束轨迹生成问题计算动态可行的无缠绕轨迹。生成的轨迹可以执行，以引导机器人到达目标位置，同时保持绳索处于无缠绕配置。我们在仿真中展示了该算法的优势，表明规划算法能够避免违反缠绕约束，从而实现更安全和更可靠的轨迹。
cs.RO / 86 / 2608.09876
Energy-Structured Latent World Models with Neural Time Fields for Physically Constistent Open-World Motion Planning
具有神经时间场的能量结构潜在世界模型用于物理一致的开放世界运动规划
Abstract
Physically consistent motion planning remains a fundamental challenge in embodied AI, as generated trajectories must strictly conform to real-world execution dynamics. While latent world models offer a promising approach by predicting these dynamics, existing methods learn unconstrained future representations where absorbed physics remains implicit. Therefore, they fail to form reusable physical knowledge, which compromises reliability in unpredictable open-world navigation. To address this, we propose a novel Energy-Structured Latent World Model (ELWM). Our key idea is to structure the ELWM latent state to explicitly carry energy and momentum, ensuring strictly causal transitions via dissipation and control ports. Trained on multimodal RGB-D and inertial interaction histories, our model guarantees physically consistent predictions. We further implement this for motion planning by constructing Physics-Conditioned Neural Time Fields (PC-NTF), a key technical cornerstone that integrates ELWM into an arrival time field via the Eikonal equation to yield a physically-informed navigation policy. Across held-out scenes, our evaluation reveals significant improvements. Compared to generic latent models, PC-NTF reduces 0.8-s motion-prediction NRMSE from 0.36 to 0.29. Against Active Neural Time Fields, it improves navigation success from 81.3% to 89.7% and SPL from 0.64 to 0.73, while cutting the physical collision rate from 12.1% to 5.8% and the Eikonal residual from 0.083 to 0.031. Beyond these targeted gains, our results demonstrate that embedding explicit physical structures into latent spaces intrinsically bridges the gap between predictive world models and safe, dynamically feasible motion planning.
Chinese Translation
物理一致的运动规划仍然是具身人工智能中的一个基本挑战，因为生成的轨迹必须严格符合现实世界的执行动态。尽管潜在世界模型通过预测这些动态提供了一种有前景的方法，但现有方法学习的是不受约束的未来表示，其中吸收的物理仍然是隐含的。因此，它们未能形成可重用的物理知识，这在不可预测的开放世界导航中妨碍了可靠性。为了解决这个问题，我们提出了一种新颖的能量结构潜在世界模型（Energy-Structured Latent World Model, ELWM）。我们的关键思想是将ELWM的潜在状态结构化，以明确携带能量和动量，通过耗散和控制端口确保严格的因果转变。在多模态RGB-D和惯性交互历史上训练后，我们的模型保证了物理一致的预测。我们进一步通过构建物理条件神经时间场（Physics-Conditioned Neural Time Fields, PC-NTF）实现运动规划，这是一个关键的技术基石，它通过Eikonal方程将ELWM集成到到达时间场中，从而产生一个物理信息导航策略。在保留场景中的评估中，我们的结果显示出显著的改进。与通用潜在模型相比，PC-NTF将0.8秒运动预测的归一化均方根误差（NRMSE）从0.36降低到0.29。与主动神经时间场相比，它将导航成功率从81.3%提高到89.7%，将成功路径长度（SPL）从0.64提高到0.73，同时将物理碰撞率从12.1%降低到5.8%，Eikonal残差从0.083降低到0.031。除了这些针对性的收益外，我们的结果表明，将明确的物理结构嵌入潜在空间本质上弥合了预测世界模型与安全、动态可行的运动规划之间的差距。
cs.RO / 87 / 2608.09891
RoSE: A Robotic Soft Esophagus for Endoprosthetic Stent Testing
RoSE：一种用于内支架测试的软机器人食道
Abstract
Soft robotic systems are well suited for developing devices for biomedical applications. A bio-mimicking robotic soft esophagus (RoSE) is developed as an in vitro testing device of endoprosthetic stents for dysphagia management. Endoprosthetic stent placement is an immediate and cost-effective therapy for dysphagia caused by malignant esophageal strictures from esophageal cancer. However, later stage complications, like stent migration, could weaken swallow efficacy in the esophagus. The stent radial force (RF) on the esophageal wall is pivotal in avoiding stent migration. Due to limited randomized controlled trials in patients, stent design and stenting guidelines remain incomplete. To address this knowledge deficit, we investigate RoSE by implanting two stents (A and B) of different radial stiffness characteristics, to measure stent RF and its effect on migration. Endoscopic manometry under peristalsis is also performed to study the impact of stenting and stent dysfunction on intra-bolus pressure signatures (IBPSs) and swallowing efficacy. Each implanted stent undergoes experiments with varied peristalsis velocity, wavelength, and bolus concentrations. The results show that stiffer stent B has a higher RF, whereas stent A maintains a lower RF profile due to lesser stiffness. High RF is necessary to minimize migration under prolonged peristaltic contractions in RoSE. For manometry, stent A slightly increases IBPS, but stiffer stent B significantly decreases IBPS, especially for higher-concentration boluses. If a stiffer stent buckles, it can reduce swallow efficacy and cause recurrent dysphagia. RoSE is therefore an innovative soft robotic platform for testing endoprosthetic stents and addressing clinical challenges in stent evaluation.
Chinese Translation
软机器人系统非常适合开发生物医学应用设备。我们开发了一种仿生的软机器人食道（RoSE），作为内支架在吞咽困难管理中的体外测试设备。内支架的放置是针对由食道癌引起的恶性食道狭窄所致吞咽困难的一种即时且经济有效的治疗方法。然而，后期并发症，如支架移位，可能会削弱食道的吞咽效果。支架对食道壁的径向力（RF）在避免支架移位中至关重要。由于在患者中随机对照试验的数量有限，支架设计和支架植入指南仍不完善。为了解决这一知识缺口，我们通过植入两种不同径向刚度特征的支架（A和B）来研究RoSE，以测量支架的RF及其对移位的影响。同时，在蠕动下进行内窥镜测压，以研究支架植入和支架功能障碍对食团内压力特征（IBPSs）和吞咽效果的影响。每个植入的支架在不同的蠕动速度、波长和食团浓度下进行实验。结果表明，刚性较大的支架B具有更高的RF，而支架A由于刚度较小，保持较低的RF特征。在RoSE中，高RF是减少在长时间蠕动收缩下移位的必要条件。对于测压，支架A略微增加IBPS，而刚性更大的支架B则显著降低IBPS，尤其是在高浓度食团的情况下。如果刚性支架发生弯曲，可能会降低吞咽效果并导致反复吞咽困难。因此，RoSE是一个创新的软机器人平台，用于测试内支架并解决支架评估中的临床挑战。
cs.RO / 88 / 2608.09892
XPolicyLab: A Unified Standard and Open Ecosystem for Robot Policy Evaluation and Deployment
XPolicyLab：机器人策略评估与部署的统一标准与开放生态系统
Community, XPolicyLab, Chen, Tianxing, Chen, Yue, Nian, Tian, Cai, Zijian, Chen, Guangyu, Lin, Wenwei, Liang, Qiwei, Xiang, Peicheng, Su, Kailun, Li, Zixuan, Tang, Junyuan, Qin, Yan, Chen, Qiangyu, Zhu, Shaolong, Li, Xiang, Zhang, Jiahao, Wan, Weijie, Chen, Baijun, Su, Honghao, Ye, Kehe, Liu, Shujia, Wang, Kaixuan, Liang, Haotian, Liu, Yunze, Li, Mingleyang, Wang, Yuran, Chen, Boyu, Bi, Hongzhe, Huang, Shuhe, Tan, Hengkai, Cai, Jisong, Mu, Yao, Guo, Jun, Wang, Xiaofeng, Zhu, Zheng, Ke, Weijie, Li, Hengtao, Tang, Yuhang, Li, Xiaofan, Yang, Ganlin, Tu, Zhangzheng, Yang, Shuai, Song, Wenxuan, Ding, Pengxiang, Zhang, Kaidong, Sun, Yu, Guo, Junliang, Zhang, Tong, Chen, Yixing, Cui, Rongxu, Zhang, Zongzheng, Ma, Haoxiang, Cai, Junhao, Zhang, Haoyu, Yang, Senqiao, Ye, Jinhui, Chen, Pengguang, Liu, Shu, Su, Xiu, Fang, Wenhan, Li, Wenhao, Cao, Yichao, Wang, Chengyao, Chen, Qiang, Luo, Ping, Ding, Wenbo
Abstract
Robot policy evaluation and deployment remain fragmented by model-specific software dependencies, data representations, and runtime interfaces, so that connecting N policies to M evaluation environments requires O(NM) separate integrations. We present XPolicyLab, a unified standard and open ecosystem that reduces this cost to O(N+M). XPolicyLab specifies common observation, action, and trajectory schemas together with a minimal adapter interface for observation updates, action prediction, batched execution, and episode reset, while a dependency-isolated client/server architecture separates policy inference from environment execution, so that each side retains its native software stack and may run locally or remotely. The ecosystem integrates 42 robot policies and standardizes their installation, debugging, serving, and evaluation workflows. Across these adapters, model-specific code varies by an order of magnitude while the environment-facing loop stays within a few lines of a fixed reference, confirming that the contract confines heterogeneity to the policy side. In a controlled study, conforming to the standard reduces the integration effort of a representative policy from over five hours to two hours, and packaged agent skills reduce it further to thirty minutes. The same adapters serve RoboTwin, RoboDojo simulation, and standardized real-robot evaluation through one interface. XPolicyLab is released as shared infrastructure for reproducible policy comparison and standardized deployment across simulation and physical platforms. Project website: https://xpolicylab.github.io/.
Chinese Translation
机器人策略评估与部署仍然因模型特定的软件依赖、数据表示和运行时接口而显得零散，因此将 N 个策略连接到 M 个评估环境需要 O(NM) 的独立集成。我们提出了 XPolicyLab，一个统一的标准和开放生态系统，将这一成本降低到 O(N+M)。XPolicyLab 规定了通用的观察、动作和轨迹模式，并提供了一个最小适配器接口，用于观察更新、动作预测、批量执行和回合重置，同时，依赖隔离的客户端/服务器架构将策略推理与环境执行分开，使得每一方都能保留其本地软件栈，并可以在本地或远程运行。该生态系统集成了 42 个机器人策略，并标准化了它们的安装、调试、服务和评估工作流程。在这些适配器中，模型特定代码的变化量级相差很大，而面向环境的循环保持在几行固定参考之内，确认了合同将异质性限制在策略侧。在一项受控研究中，符合标准将一个代表性策略的集成工作量从五小时以上减少到两小时，而打包的代理技能进一步将其减少到三十分钟。相同的适配器通过一个接口服务于 RoboTwin、RoboDojo 模拟和标准化的真实机器人评估。XPolicyLab 被发布为可重复的政策比较和在模拟与物理平台之间标准化部署的共享基础设施。项目网站：https://xpolicylab.github.io/
cs.CV / 1 / 2608.07478
PragyaDoc: A Universal Document Intelligence Framework for Multilingual Medical Document Understanding in Low-Resource Settings
PragyaDoc：一种用于低资源环境下多语言医疗文档理解的通用文档智能框架
Abstract
India's 22 official languages create a critical accessibility barrier: the majority of medical documentation exists exclusively in English, yet the patients who most urgently require this information - rural populations, ASHA workers, and patient families - are functionally excluded from understanding it. This paper presents PragyaDoc, a Universal Document Intelligence Framework that addresses this gap through a four-layer pipeline: a parallel ensemble OCR extraction layer, a geometric-lexical fusion layer, a deterministic domain structuring layer, and a dual-LLM medical reasoning and localization layer
Chinese Translation
印度的22种官方语言构成了一个重要的可及性障碍：大多数医疗文档仅以英语存在，而最迫切需要这些信息的患者——农村人口、ASHA工作者和患者家庭——在理解这些信息时实际上被排除在外。本文提出了PragyaDoc，一个通用文档智能框架，通过四层管道来解决这一问题：一个并行集成的光学字符识别（OCR）提取层、一个几何-词汇融合层、一个确定性领域结构层，以及一个双重大语言模型（LLM）医疗推理和定位层。
cs.CV / 2 / 2608.07541
NeuroPilot: An Agent-Driven Smart Pipeline for Processing, Quality Control, and Managing Neuroimages
NeuroPilot：一个基于代理的智能管道，用于处理、质量控制和管理神经影像
Abstract
Transforming raw neuroimage archives into analysis-ready derivatives relies on three brittle stages: data standardization, modality-specific preprocessing, and quality control (QC). While individual neuroimaging tools are well developed, their orchestration requires project-specific scripts, environment-adaptive tuning, and labor-intensive manual QC. To address this, we introduce NeuroPilot, a multi-agent system that digitalizes the expertise of neuroimage processing, QC, and data management into three LLM-invocable skills: dcm2bids-skill, neuroimage-pre-skill, and qc-agent-skill. The LLM-driven agent autonomously orchestrates workflows, generalizing various infrastructure settings into a single configuration to achieve the highest scalability. Demonstrating the system's generalizability, we deployed NeuroPilot across 17 cohorts (>123,000 subjects) spanning infant to aging populations and multiple MRI modalities (structural, diffusion, functional). In practice, after standardizing data via the dcm2bids-skill, the agent dynamically routes datasets to the optimal neuroimage-pre-skill based on available modalities and cohort traits (e.g., dispatching T1w and fMRI data to fMRIPrep, or selecting specialized pipelines for infant cohorts). The qc-agent-skill then drives an evidence-based, semi-automated QC via a 3-D browser dashboard, utilizing a multi-tiered verification system to optimize failed cases and escalate complex issues for supervisor inspection. Quantitatively, our QC agent screened 558 production subjects, validating its automated flags against FreeSurfer's topology-defect metrics. The infant processing pipeline achieved a 100% (201/201) completion rate on QC-validated inputs. Importantly, NeuroPilot compresses the traditional 2--3 month timeline for training staff and processing complete datasets into a single week. NeuroPilot is deployed in https://wanda-cyberbench.com/.
Chinese Translation
将原始神经影像档案转化为可分析的衍生品依赖于三个脆弱的阶段：数据标准化、特定模态的预处理和质量控制（QC）。虽然单个神经影像工具已经发展成熟，但它们的协调需要项目特定的脚本、环境适应性调优和劳动密集型的手动质量控制。为了解决这一问题，我们提出了NeuroPilot，一个多代理系统，将神经影像处理、质量控制和数据管理的专业知识数字化为三种可由大型语言模型（LLM）调用的技能：dcm2bids-skill、neuroimage-pre-skill和qc-agent-skill。该LLM驱动的代理自主协调工作流程，将各种基础设施设置概括为单一配置，以实现最高的可扩展性。为了展示系统的通用性，我们在17个队列（超过123,000名受试者）中部署了NeuroPilot，涵盖了从婴儿到老年人群体以及多种MRI模态（结构、扩散、功能）。在实践中，经过dcm2bids-skill进行数据标准化后，代理根据可用模态和队列特征（例如，将T1w和fMRI数据分发给fMRIPrep，或为婴儿队列选择专门的管道）动态路由数据集到最佳的neuroimage-pre-skill。随后，qc-agent-skill通过一个3D浏览器仪表板驱动基于证据的半自动化质量控制，利用多层验证系统优化失败案例并将复杂问题上报给监督者检查。从定量角度来看，我们的QC代理筛查了558名生产受试者，验证了其自动标记与FreeSurfer的拓扑缺陷指标的一致性。婴儿处理管道在经过质量控制验证的输入上实现了100%（201/201）的完成率。重要的是，NeuroPilot将传统的2-3个月的培训员工和处理完整数据集的时间压缩到一周内。NeuroPilot已部署在https://wanda-cyberbench.com/。
cs.CV / 3 / 2608.07543
Performance of large language models in the optical diagnosis of colorectal polyps
大型语言模型在结直肠息肉光学诊断中的表现
Abstract
Background and Study Aims: Accurate optical diagnosis of colorectal polyps guides resection strategy and surveillance, with multimodal large language models (MLLMs) showing potential for image-based diagnosis. We aimed to evaluate the diagnostic accuracy of MLLMs in classifying colorectal polyps and predicting histology. Methods: We conducted a retrospective diagnostic performance study using the PRIME dataset, a curated set of white light and narrow-band imaging (NBI) images. We evaluated Claude Opus 4, Google Gemini 2.5 Pro, GPT-o3, GPT-4o, and GPT-5. For Paris, Narrow-band Imaging Colorectal Endoscopic (NICE), and predicted histology, we calculated F1 scores, percent correct scores, and accuracy of each MLLM compared to expert responses for 132 cases. Cochran's Q and McNemar's Test were used to determine differences between predicted values of each MLLM. Results: The F1 scores among MLLMs were >0.9 for all models for neoplastic vs. non-neoplastic polyps. Gemini 2.5 Pro demonstrated the highest F1 scores for invasive vs. non-invasive polyps and low- vs. high-grade adenoma, at 0.560 and 0.492 respectively. Claude Opus 4 and GPT-5 had statistically significantly higher percent correct scores than other MLLMs at 41.7%, using Paris classification. Conclusions: Claude Opus 4 and Gemini 2.5 Pro showed the highest accuracy in differentiating polyp subtypes, performing closest to expert consensus. Sensitivity and specificity, however, did not meet ESGE standards, highlighting the need for prospective multicenter trials and the design of human-in-the-loop workflows before clinical deployment.
Chinese Translation
背景与研究目的：准确的结直肠息肉光学诊断指导切除策略和监测，多模态大型语言模型（MLLMs）在基于图像的诊断中显示出潜力。我们旨在评估MLLMs在分类结直肠息肉和预测组织学方面的诊断准确性。方法：我们使用PRIME数据集进行了一项回顾性诊断性能研究，该数据集包含经过整理的白光和窄带成像（NBI）图像。我们评估了Claude Opus 4、Google Gemini 2.5 Pro、GPT-o3、GPT-4o和GPT-5。对于巴黎分类、窄带成像结肠内镜（NICE）和预测的组织学，我们计算了F1分数、正确率百分比和每个MLLM与专家响应的准确性，涉及132个案例。使用Cochran's Q和McNemar检验来确定每个MLLM预测值之间的差异。结果：在MLLMs中，所有模型对肿瘤性与非肿瘤性息肉的F1分数均超过0.9。Gemini 2.5 Pro在侵袭性与非侵袭性息肉以及低级与高级腺瘤的F1分数分别为0.560和0.492，表现最佳。Claude Opus 4和GPT-5的正确率百分比显著高于其他MLLMs，达41.7%，使用巴黎分类。结论：Claude Opus 4和Gemini 2.5 Pro在区分息肉亚型方面显示出最高的准确性，表现最接近专家共识。然而，敏感性和特异性未达到欧洲胃肠内镜学会（ESGE）标准，强调了在临床应用之前需要进行前瞻性多中心试验和设计人机协作工作流程的必要性。
cs.CV / 4 / 2608.07547
Learning an Interior Layout Policy in a Domain Specific Language Action Space
在特定领域语言动作空间中学习室内布局策略
Abstract
Indoor scene layout generation is a challenging task in interior design. Existing methods often oversimplify the task by reducing room conditions to coarse 3D bounding boxes and neglecting structural elements such as doors and windows. More fundamentally, many prior approaches formulate spatial reasoning as direct coordinate prediction, thereby casting interior layout design as continuous regression over raw geometric parameters, which hinders the model from learning the underlying reasoning logic of intelligent layout design. We propose \textbf{LayoutDSL}, a novel LLM-based framework for learning an interior layout policy in a domain-specific language (DSL) action space. The DSL provides an explicit symbolic representation of layout information and serves as a structured action space for layout reasoning, where each action corresponds to an interpretable design decision. Under this DSL-based policy learning paradigm, we construct 3D-FrontDSL, a dataset of room-structure annotations paired with synthetic DSL action sequences for supervised fine-tuning. To promote a more generalizable and scalable policy with verifiable feedback, we design rewards grounded in interior design principles and physical plausibility, and optimize the policy via reinforcement learning. Extensive experiments demonstrate that LayoutDSL substantially improves spatial plausibility and design logicality over strong baselines and existing methods.
Chinese Translation
室内场景布局生成是室内设计中的一项挑战性任务。现有方法常常通过将房间条件简化为粗略的3D边界框来过度简化任务，并忽视了门窗等结构元素。更根本地，许多先前的方法将空间推理表述为直接的坐标预测，从而将室内布局设计视为对原始几何参数的连续回归，这阻碍了模型学习智能布局设计的基本推理逻辑。我们提出了 extbf{LayoutDSL}，一个基于大型语言模型（LLM）的新框架，用于在特定领域语言（DSL）动作空间中学习室内布局策略。DSL提供了布局信息的显式符号表示，并作为布局推理的结构化动作空间，其中每个动作对应于一个可解释的设计决策。在这种基于DSL的策略学习范式下，我们构建了3D-FrontDSL，一个配有合成DSL动作序列的房间结构注释数据集，用于监督微调。为了促进更具可推广性和可扩展性的策略，并提供可验证的反馈，我们设计了基于室内设计原则和物理合理性的奖励，并通过强化学习优化该策略。大量实验表明，LayoutDSL在空间合理性和设计逻辑性方面显著优于强基线和现有方法。
cs.CV / 5 / 2608.07549
P2Voxel: Pyramid Pivot Voxelization for 3D Mesh Tokenization
P2Voxel：用于3D网格标记化的金字塔枢轴体素化
Abstract
Triangle meshes provide explicit and accurate surface geometry, yet their irregular topology connectivity makes 3D mesh tokenization a geometric sampling problem: how to sample and organize geometric evidence into compact, structured and learnable tokens. Beyond field-centric volumetric sampling and edge-intersection surface sampling, we retarget mesh tokenization as \textit{local surface evidence sampling}: identifying the minimal geometric evidence inside each active voxel that is sufficient for deterministic surface recovery. To this end, we introduce \textbf{P2Voxel}, a pyramid pivot voxelization framework for compact and reconstruction-aware mesh tokenization. P2Voxel is built on three key innovations. Under the \textit{Local Planarity} assumption, Pivot Voxelization represents each active voxel with a surface pivot and an orientation sign, providing minimal local evidence that can induce the corner values required for deterministic reconstruction. Under the \textit{Spatial Complexity} assumption, Pyramid Pivot Voxelization exploits the spatial non-uniformity of real surfaces by allocating finer pivot tokens to geometrically complex regions while keeping smooth regions coarse and compact. Under the \textit{Block Reconstructability} assumption, a Pyramid VAE learns compact multi-resolution latent codes over locally reconstructable pivot blocks, avoiding the need to model the entire high-resolution voxelized shape as a dense global field. Together, these designs convert meshes into compact, structured, and learnable pyramid pivot tokens, enabling efficient mesh reconstruction for downstream 3D tasks.
Chinese Translation
三角网格提供了明确且准确的表面几何形状，但其不规则的拓扑连接性使得3D网格标记化成为一个几何采样问题：如何将几何证据采样并组织成紧凑、结构化且可学习的标记。除了以场为中心的体积采样和边缘交叉表面采样外，我们将网格标记化重新定位为 extit{局部表面证据采样}：识别每个活动体素内部的最小几何证据，以便进行确定性的表面恢复。为此，我们引入了 extbf{P2Voxel}，一个用于紧凑且重建感知网格标记化的金字塔枢轴体素化框架。P2Voxel建立在三个关键创新之上。在 extit{局部平面性}假设下，枢轴体素化通过一个表面枢轴和一个方向标志来表示每个活动体素，提供了可以诱导确定性重建所需角值的最小局部证据。在 extit{空间复杂性}假设下，金字塔枢轴体素化利用真实表面的空间非均匀性，通过将更精细的枢轴标记分配给几何复杂区域，同时保持光滑区域粗糙且紧凑。在 extit{块可重建性}假设下，金字塔变分自编码器（Pyramid VAE）在局部可重建的枢轴块上学习紧凑的多分辨率潜在编码，避免了将整个高分辨率体素化形状建模为密集的全局场的需要。这些设计共同将网格转换为紧凑、结构化且可学习的金字塔枢轴标记，从而实现高效的网格重建，以支持下游3D任务。
cs.CV / 6 / 2608.07550
Auditing Medical Vision-Language Models on Chest Radiographs: Estimating Reference Agreement Across Institutions
对胸部放射影像的医学视觉-语言模型进行审计：跨机构参考一致性估计
Abstract
Vision-language models return structured chest-radiograph findings through interfaces exposing no confidence score, so a receiving institution cannot read off how far to trust an individual judgment. Whether agreement with an institution's reference standard transfers across sites, findings, prediction directions and question formats is largely unmeasured. We evaluated three generative vision-language models on three institutional chest-radiograph corpora and six findings under two elicitation protocols, comprising more than 345,000 finding-level predictions, and estimated finding-by-direction reference agreement at a receiving institution from a small budget of local labels. Estimation strategies were then stress-tested under repeated strict institution-held-out evaluation. Under evaluation excluding the receiving institution from development entirely, adaptive selection among the seven estimators that design admits did not improve on simple fixed alternatives: it achieved a mean Brier score of 0.1083, against 0.0853 for always using a Beta-Binomial empirical-Bayes estimator and 0.0855 for a target-only logistic model. Those two differ by 0.0003, less than this family's own sensitivity to a change of solver version, and each leads in about half the settings, so no default can be recommended. Their advantage over estimators pooling across institutions was concentrated at one site and not confirmatory once clustered by institution, and a plug-in empirical-Bayes posterior-predictive count interval at a nominal 95% level covered 87.0%, less at the hardest institution. Reference agreement therefore has to be re-evaluated per site and per interface; these results concern agreement with institutional labels, not clinical correctness.
Chinese Translation
视觉-语言模型通过不显示置信度评分的接口返回结构化的胸部放射影像结果，因此接收机构无法判断对个别判断的信任程度。跨地点、发现、预测方向和问题格式的机构参考标准一致性是否能够转移，尚未得到充分测量。我们在三个机构的胸部放射影像语料库上评估了三种生成性视觉-语言模型和六种发现，采用两种引导协议，涵盖了超过345,000个发现级预测，并从少量本地标签中估计了接收机构的发现-方向参考一致性。随后，在重复严格的机构保留评估下，对估计策略进行了压力测试。在完全不将接收机构纳入开发的评估中，七个设计允许的估计器之间的自适应选择并未优于简单的固定替代方案：其平均Brier评分为0.1083，而始终使用Beta-Binomial经验贝叶斯估计器的评分为0.0853，目标仅逻辑模型的评分为0.0855。这两者的差异为0.0003，低于该家族对求解器版本变化的敏感度，并且在大约一半的设置中各自领先，因此无法推荐默认方案。它们在跨机构的估计器中的优势集中在一个地点，并且在按机构聚类后未能得到确认，而在名义95%水平下的插值经验贝叶斯后验预测计数区间覆盖了87.0%，在最困难的机构中更少。因此，参考一致性必须针对每个地点和每个接口重新评估；这些结果涉及与机构标签的一致性，而非临床正确性。
cs.CV / 7 / 2608.07554
Impact of Dataset Composition on Embedded Real-Time UAV Wildfire Detection Using Compact YOLO Models
数据集组成对嵌入式实时无人机野火检测的影响：基于紧凑型YOLO模型的研究
Abstract
The development of vision-based wildfire detection systems for unmanned aerial vehicles is constrained by the limited availability of diverse real-world training images. This paper investigates the impact of dataset composition on embedded real-time UAV wildfire detection using compact YOLO models as a controlled validation family. Four training configurations were evaluated: real non-augmented, real augmented, hybrid non-augmented, and hybrid augmented, where the hybrid sets combine real wildfire images with AI-generated samples. The objective is to determine whether synthetic data mixing and image augmentation improve practical detection performance under resource-constrained deployment conditions. Experimental results show that the best overall operating point was obtained with the real non-augmented dataset, which achieved the strongest balance between recall and mean average precision for UAV-based wildfire detection. The results also show that neither hybridization with synthetic data nor augmentation produced a better final deployment choice. These findings suggest that, for embedded UAV wildfire detection, dataset realism and domain alignment are more valuable than increasing training set size through synthetic expansion.
Chinese Translation
基于视觉的无人机野火检测系统的发展受到多样化真实训练图像有限可用性的制约。本文研究了数据集组成对嵌入式实时无人机野火检测的影响，采用紧凑型YOLO模型作为受控验证的基础。评估了四种训练配置：真实非增强、真实增强、混合非增强和混合增强，其中混合集结合了真实野火图像与AI生成的样本。研究的目标是确定在资源受限的部署条件下，合成数据混合和图像增强是否能改善实际检测性能。实验结果表明，最佳的整体操作点是使用真实非增强数据集获得的，该数据集在无人机野火检测中实现了召回率和平均精度的最佳平衡。结果还表明，合成数据的混合或增强并未产生更好的最终部署选择。这些发现表明，对于嵌入式无人机野火检测，数据集的真实性和领域对齐比通过合成扩展增加训练集规模更为重要。
cs.CV / 8 / 2608.07559
MVMD: A Multi-View Approach for Enhanced Mirror Detection
MVMD：一种增强镜面检测的多视角方法
Abstract
In 3D reconstruction, mirrors introduce significant challenges by creating distorted and fragmented spaces, resulting in inaccurate and unreliable 3D models. As 3D reconstruction typically relies on multi-view images to capture different perspectives of a scene, detecting and labeling mirrors in multi-view images before reconstruction can effectively address this issue. However, existing methods focus solely on single-image detection, overlooking the rich information provided by multi-view setups. To overcome this limitation, we propose MVMD, a novel Multi-View Mirror Detection method, along with the first database specifically designed for mirror detection in multi-view scenes. The design of MVMD is grounded in the inherent associations between objects seen from different views and those reflected inside and outside of mirrors. These relationships are learned through cross- and self-attention mechanisms. MVMD consists of three key blocks: the Inter-Views Block tracks the shifts of objects within mirrors caused by changes in viewpoint; the Intra-View Block detects object reflections inside mirrors; and the Refinement Block sharpens mirror boundaries and enhances detected details. Experimental results show that our method improves accuracy by up to 2.6% and IoU by up to 11.1%, compared to single-image mirror detection techniques. This substantial improvement makes MVMD particularly effective for computer vision tasks, especially in enhancing the accuracy of 3D reconstruction in mirror-dense environments.
Chinese Translation
在三维重建中，镜子引入了显著的挑战，造成扭曲和碎片化的空间，从而导致不准确和不可靠的三维模型。由于三维重建通常依赖多视角图像来捕捉场景的不同视角，因此在重建之前检测和标记多视角图像中的镜子可以有效解决这一问题。然而，现有方法仅关注单图像检测，忽视了多视角设置所提供的丰富信息。为克服这一局限性，我们提出了MVMD，一种新颖的多视角镜面检测方法，并首次设计了专门用于多视角场景中镜面检测的数据库。MVMD的设计基于从不同视角观察的物体与镜子内外反射物体之间的固有关联。这些关系通过交叉注意力和自注意力机制进行学习。MVMD由三个关键模块组成：视间模块（Inter-Views Block）跟踪因视角变化而导致的镜中物体的位移；视内模块（Intra-View Block）检测镜子内的物体反射；精细化模块（Refinement Block）锐化镜子边界并增强检测到的细节。实验结果表明，与单图像镜面检测技术相比，我们的方法在准确性上提高了最多2.6%，在IoU（交并比）上提高了最多11.1%。这一显著的改进使MVMD在计算机视觉任务中尤其有效，特别是在增强镜面密集环境中的三维重建准确性方面。
cs.CV / 9 / 2608.07561
XEns-CKD: An Explainable Ensemble-Based Approach for Chronic Kidney Disease Stage Detection
XEns-CKD：一种可解释的基于集成的方法用于慢性肾病分期检测
Abstract
Chronic kidney disease (CKD) is a silent disease. Its progression may not significantly hamper a person's daily routine. Human kidney function can be classified as normal or as one of the five stages of CKD. Early detection of the CKD stage can help patients understand the functional status of their kidneys and follow medical advice to slow CKD progression. In this paper, we propose XEns-CKD, a novel ensemble vision transformer-based scheme for CKD stage classification using ultrasound images. Three ViTs were trained on a private ultrasound image dataset using different training parameters. The performance of each ViT was evaluated using macro sensitivity, macro specificity, macro precision, macro F1-score, macro Youden index, the Matthews correlation coefficient (MCC), and macro balanced accuracy. The ensemble model achieved an overall classification accuracy of 86.36%. This work also emphasizes identifying and interpreting kidney regions affected by CKD progression. Explainable artificial intelligence techniques, including LIME, LRP, Attention-Min, and Attention-Max, were used to improve model transparency and clinical trust. An attention map combining the Attention-Min and Attention-Max results effectively identified and interpreted kidney regions affected during CKD progression from one stage to another. The attention map also highlighted the effects of CKD progression in these regions. Compared with existing methods, the proposed method classified the five CKD stages and normal kidney status with a 4% improvement in accuracy.
Chinese Translation
慢性肾病（CKD）是一种隐匿性疾病，其进展可能不会显著影响一个人的日常生活。人类肾功能可以被分类为正常或五个CKD阶段之一。早期检测CKD阶段可以帮助患者了解其肾脏的功能状态，并遵循医疗建议以减缓CKD的进展。本文提出了XEns-CKD，一种基于集成视觉变换器的新型方案，用于利用超声图像进行CKD阶段分类。三个视觉变换器（ViTs）在一个私有超声图像数据集上使用不同的训练参数进行了训练。每个ViT的性能通过宏灵敏度、宏特异性、宏精确度、宏F1分数、宏Youden指数、马修斯相关系数（MCC）和宏平衡准确率进行了评估。该集成模型实现了86.36%的整体分类准确率。该研究还强调了识别和解释受CKD进展影响的肾脏区域。采用可解释人工智能技术，包括LIME、LRP、Attention-Min和Attention-Max，以提高模型的透明度和临床信任度。结合Attention-Min和Attention-Max结果的注意力图有效地识别和解释了在CKD进展过程中受影响的肾脏区域。注意力图还突出了这些区域中CKD进展的影响。与现有方法相比，所提出的方法在分类五个CKD阶段和正常肾脏状态时提高了4%的准确率。
cs.CV / 10 / 2608.07562
Mechanistic Interpretability-Guided Selective Fine-Tuning of Vision-Language Models for Centimeter-Level Flood Depth Estimation
机制可解释性指导的视觉-语言模型选择性微调用于厘米级洪水深度估计
Abstract
Urban flooding poses an escalating threat to transportation infrastructure, yet no operational system provides real-time, street-level flood-depth estimates at centimeter resolution. This paper presents three vision-language models fine-tuned for continuous flood-depth estimation from street-level imagery: FloodLlama-Dense, a fully fine-tuned QLoRA baseline, and FloodLlama-MI5 and FloodLlama-MI6, interpretability-guided sparse variants that fine-tune only the top five and six causally relevant cross-attention layers identified through mechanistic interpretability analysis, respectively. Training uses an approximately 610,000-image subset of a 2.81-million-image synthetic corpus generated in Unreal Engine 5. The dataset combines single-vehicle subsets with 5 cm depth increments and mixed-vehicle subsets with 1 cm depth increments, spanning seven vehicle types, four weather conditions, and flood depths from 0 to 40 cm. FloodLlama-Dense achieves an MAE of 0.40 cm, an RMSE of 1.97 cm, and an Acc@5cm of 97.59%. Mechanistic interpretability analysis combining linear probing, logit lens, centered kernel alignment (CKA), and cross-attention entropy reveals a two-stage adaptation pattern: layers L13-L22 restructure visual representations, while depth first becomes linearly decodable at layer L23. FloodLlama-MI5 and FloodLlama-MI6 leverage this insight by fine-tuning only five or six of the eight cross-attention layers, achieving an 86-88% reduction in trainable parameters (6.55-7.86 million versus 54.4 million) with minimal accuracy loss. On a real-world benchmark, FloodLlama-MI6 achieves 98.62% accuracy, compared with 86.61% for the published STURM-FloodDepth baseline.
Chinese Translation
城市洪水对交通基础设施构成日益严重的威胁，但目前没有任何操作系统能够提供厘米分辨率的实时街道级洪水深度估计。本文提出了三种经过微调的视觉-语言模型，用于从街道级图像中连续估计洪水深度：FloodLlama-Dense，一个完全微调的QLoRA基线，以及FloodLlama-MI5和FloodLlama-MI6，这两种模型是基于可解释性的稀疏变体，分别仅微调通过机制可解释性分析识别出的五个和六个因果相关的交叉注意力层。训练使用了一个约610,000张图像的子集，该子集来自于在Unreal Engine 5中生成的281万张图像的合成语料库。该数据集结合了单车车辆子集（深度增量为5厘米）和混合车辆子集（深度增量为1厘米），涵盖了七种车辆类型、四种天气条件以及0到40厘米的洪水深度。FloodLlama-Dense实现了0.40厘米的平均绝对误差（MAE）、1.97厘米的均方根误差（RMSE）和97.59%的5厘米准确率（Acc@5cm）。结合线性探测、logit lens、中心核对齐（CKA）和交叉注意力熵的机制可解释性分析揭示了一个两阶段的适应模式：L13-L22层重构视觉表示，而深度在L23层首次变得线性可解。FloodLlama-MI5和FloodLlama-MI6利用这一洞察，通过仅微调八个交叉注意力层中的五个或六个，实现了可训练参数减少86-88%（从6.55-7.86百万降至54.4百万），且准确性损失最小。在一个真实世界基准测试中，FloodLlama-MI6的准确率达到98.62%，而已发布的STURM-FloodDepth基线为86.61%。
cs.CV / 11 / 2608.07565
What to Edit Next: Visually Aligned Image-Editing Follow-Up Suggestions in Conversational Systems
下一步编辑什么：对话系统中的视觉对齐图像编辑后续建议
Abstract
Conversational assistants increasingly recommend follow-up edits to help users continue a task. Existing systems primarily target text-only interactions, leaving image-creation conversations underexplored. In image-creation tasks, useful follow-up edit suggestions must reflect user preferences, offer diverse directions, and remain executable on the current image. We collected 100,000 real multi-turn image-creation conversation samples from Qwen App and found that 80.1% are image-dependent, underscoring the need for multimodal recommendation. We address this setting with a three-stage framework. In Stage 1, we use real online data to build a human-reviewed table of appropriate follow-up editing intents, then create SFT targets and fine-tune a multimodal policy. In Stage 2, to align rule-guided SFT suggestions with actual user choices, we use user click feedback to optimize the policy through multi-objective reinforcement learning. In Stage 3, to reduce visual inconsistencies between suggested edits and the current image, we introduce a visual verifier as additional training supervision. Extensive experiments demonstrate that our framework significantly outperforms baselines on both automatic and human evaluations. In a live user-randomized A/B test with millions of users, our final framework reduces visual inconsistency from 3.7% to 0.9%. Furthermore, it significantly improves recommendation CTR by 32.70%, image take-away rate by 16.32%, and average conversation turns per user by 39.90% (all p<0.05).
Chinese Translation
对话助手越来越多地推荐后续编辑，以帮助用户继续任务。现有系统主要针对仅文本的交互，导致图像创作对话尚未得到充分探索。在图像创作任务中，有效的后续编辑建议必须反映用户偏好，提供多样化的方向，并能够在当前图像上执行。我们从Qwen App收集了100,000个真实的多轮图像创作对话样本，发现80.1%的样本依赖于图像，这突显了多模态推荐的必要性。我们通过一个三阶段框架来解决这一问题。在第一阶段，我们使用真实的在线数据构建一个经过人工审核的适当后续编辑意图表，然后创建SFT（Supervised Fine-Tuning）目标并微调多模态策略。在第二阶段，为了将规则引导的SFT建议与实际用户选择对齐，我们利用用户点击反馈通过多目标强化学习优化策略。在第三阶段，为了减少建议编辑与当前图像之间的视觉不一致性，我们引入了视觉验证器作为额外的训练监督。大量实验表明，我们的框架在自动评估和人工评估中均显著优于基线。在与数百万用户进行的实时用户随机A/B测试中，我们的最终框架将视觉不一致性从3.7%降低到0.9%。此外，它还显著提高了推荐点击率（CTR）32.70%，图像获取率16.32%，以及每个用户的平均对话轮次39.90%（均p<0.05）。
cs.CV / 12 / 2608.07567
Temporal Generalization in fNIRS-Based Autism Classification: A Cross-Time-Window Transfer Benchmark
基于功能性近红外光谱的自闭症分类中的时间泛化：跨时间窗口转移基准
Abstract
Functional near-infrared spectroscopy (fNIRS) is a promising modality for autism spectrum disorder (ASD) classification, yet existing approaches assume temporally aligned evaluation. In practice, the optimal observation window varies across subjects due to differences in hemodynamic delay and neurovascular coupling, creating a temporal distribution shift that degrades performance. We formalize this as a \textit{cross-time-window transfer problem}, introducing a protocol that varies window length (2.5--10\,s) and offset within biological motion trials. Using topographic map representations of fNIRS recordings, we benchmark three vision architectures under two zero-shot baselines and eight adaptation strategies under leave-one-subject-out cross-validation ($N{=}124$). Key findings: (1) zero-shot cross-window accuracy is near chance (54--69\%); (2) ${\approx}5\%$ subject-specific fine-tuning recovers 90--96\%, while a subject-specific upper bound reaches 97--100\%, identifying inter-subject variability as the dominant barrier; (3) domain-adversarial and self-supervised strategies achieve 78--90\% without target-subject data; and (4) discriminative information is recoverable from windows as short as 2.5\,s. These findings provide a practical roadmap for deploying fNIRS-based ASD classifiers under realistic temporal variability.
Chinese Translation
功能性近红外光谱（fNIRS）是一种有前景的自闭症谱系障碍（ASD）分类方法，但现有方法假设评估时间对齐。在实际应用中，由于血流动力学延迟和神经血管耦合的差异，最佳观察窗口在不同个体之间有所不同，这导致了时间分布的偏移，从而降低了性能。我们将其形式化为 extit{跨时间窗口转移问题}，引入了一种协议，该协议在生物运动试验中变化窗口长度（2.5--10秒）和偏移。通过对fNIRS记录的拓扑图表示，我们在两个零样本基线和八种适应策略下（使用留一法交叉验证，$N{=}124$）对三种视觉架构进行了基准测试。主要发现：（1）零样本跨窗口准确率接近随机水平（54--69%）；（2）约5%的个体特定微调可恢复90--96%的准确率，而个体特定的上限可达97--100%，这表明个体间变异性是主要障碍；（3）领域对抗和自监督策略在没有目标个体数据的情况下实现了78--90%的准确率；（4）从短至2.5秒的窗口中可恢复判别信息。这些发现为在现实时间变异性下部署基于fNIRS的ASD分类器提供了实用的路线图。
cs.CV / 13 / 2608.07569
Latent-Frequency Validity: Fast Spectral Editing with Screened Video-VAE Transfer Operators
潜频有效性：使用筛选视频-VAE传递算子的快速谱编辑
Abstract
Direct spectral editing in video-VAE latents can control noise, flicker, smoothness, and frequency content without a decode--filter--reencode pass. However, video VAEs may redistribute pixel-space frequency bands across latent channels, and latent edits can disrupt VAE round-trip dynamics. We introduce \emph{latent-frequency validity} (LFV), which learns a compact VAE-specific spectral response and deploys it only when it improves decoded-target fidelity without worsening round-trip drift. LFV follows a validation-selected path from a diagonal per-frequency calibrator (C1) to full channel mixing (CM), making cross-channel capacity a controllable per-edit resource. Across 544 VAE--edit cells spanning six spectral families, LFV emits 423 cheap operators: 277 are handled by C1, while 146 (34.5\% of emitted operators) require channel mixing. On the primary 120-cell radial sweep, 99/100 emitted operators pass source-video-grouped held-out evaluation. Across five additional filter families, all 323 emitted operators pass held-out evaluation. Fully frozen OpenVid-fitted operators, including the validation-selected path coefficient, pass all 20 tested CogVideoX and HunyuanVideo generated-domain cells without adaptation. The selected response matches direct latent-filter latency and is about $3\times$ faster than pixel filter--reencode. The resulting maps reveal distinct VAE regimes, including strongly channel-coupled CogVideoX responses and a sharp Open-Sora high-band stability frontier.
Chinese Translation
在视频-VAE潜变量中直接进行谱编辑可以控制噪声、闪烁、平滑度和频率内容，而无需解码-过滤-重新编码的过程。然而，视频VAE可能会在潜在通道之间重新分配像素空间频率带，且潜在编辑可能会干扰VAE的往返动态。我们提出了 extit{潜频有效性}（Latent-Frequency Validity, LFV），它学习一种紧凑的VAE特定谱响应，并仅在其提高解码目标保真度而不恶化往返漂移时部署。LFV遵循从每频率对角校准器（C1）到完全通道混合（CM）的验证选择路径，使跨通道容量成为可控的每次编辑资源。在跨越六个谱系的544个VAE编辑单元中，LFV发出423个廉价算子：其中277个由C1处理，而146个（占发出算子的34.5%）需要通道混合。在主要的120单元径向扫描中，99/100个发出算子通过了源视频分组的保留评估。在五个额外的过滤器家族中，所有323个发出算子均通过了保留评估。完全冻结的OpenVid拟合算子，包括验证选择路径系数，在未适应的情况下通过了所有20个测试的CogVideoX和HunyuanVideo生成域单元。所选响应与直接潜滤波延迟相匹配，且速度约为像素滤波-重新编码的$3 imes$。结果图谱揭示了不同的VAE状态，包括强通道耦合的CogVideoX响应和明显的Open-Sora高频稳定性前沿。
cs.CV / 14 / 2608.07570
COMEX: A Composition-Grounded Benchmark and Learning Framework for Explainable Aesthetic Image Cropping
COMEX：一个基于构图的可解释美学图像裁剪基准和学习框架
Abstract
Explainable aesthetic image cropping requires not only localizing a visually pleasing crop but also explaining why it is preferred. Existing crop-and-explain methods largely treat explanation as post-hoc text generation and overlook composition, a key aesthetic factor that links crop decisions with interpretable reasoning. In this paper, we reformulate explainable aesthetic image cropping as a structured crop-composition-explanation problem. To support this setting, we introduce COMEX, a new benchmark built through image expansion and an IO-reversal pipeline. COMEX contains 33,161 quadruples, each consisting of an expanded image, a crop box, a composition category, and a composition-grounded explanation, enabling joint learning of crop localization, composition understanding, and explanation generation. We further propose a two-stage SFT+GRPO framework, where supervised fine-tuning establishes the structured output protocol and basic cropping ability, and GRPO further improves crop quality, composition prediction, and explanation faithfulness. We benchmark 15 large vision-language models and existing cropping methods on COMEX, establishing a comprehensive testbed for composition-grounded explainable aesthetic cropping. Experiments on both COMEX and prior benchmarks demonstrate the effectiveness and transferability of our framework, with strong performance across evaluation metrics.
Chinese Translation
可解释的美学图像裁剪不仅需要定位一个视觉上令人愉悦的裁剪区域，还需要解释为什么该裁剪区域更受偏爱。现有的裁剪与解释方法大多将解释视为事后文本生成，忽视了构图这一关键美学因素，而构图将裁剪决策与可解释推理联系起来。在本文中，我们将可解释的美学图像裁剪重新表述为一个结构化的裁剪-构图-解释问题。为支持这一设定，我们引入了COMEX，一个通过图像扩展和IO反转管道构建的新基准。COMEX包含33,161个四元组，每个四元组由一个扩展图像、一个裁剪框、一个构图类别和一个基于构图的解释组成，使得裁剪定位、构图理解和解释生成的联合学习成为可能。我们进一步提出了一个两阶段的SFT+GRPO框架，其中监督微调建立了结构化输出协议和基本裁剪能力，而GRPO进一步提高了裁剪质量、构图预测和解释的可信度。我们在COMEX上基准测试了15个大型视觉-语言模型和现有的裁剪方法，为基于构图的可解释美学裁剪建立了一个全面的测试平台。在COMEX和之前的基准上的实验表明了我们框架的有效性和可迁移性，在各项评估指标上表现出色。
cs.CV / 15 / 2608.07571
A Review of Vision-Based Vehicle Detection for UAV-Based Traffic Monitoring: Experimental Insights and Future Directions
基于视觉的无人机交通监测车辆检测综述：实验见解与未来方向
Abstract
In Intelligent Transportation System (ITS), unmanned aerial vehicle (UAV)-based surveillance offers an innovative solution to traffic surveillance with wide coverage and real-time data collection capabilities. In comparison to fixed ground-based infrastructure, UAVs are able to respond to dynamic traffic but present challenges such as vehicle detection at varying altitudes, compensation for motion-induced image variations and efficient processing of high-resolution images. Deep learning has been largely beneficial on improving the detection accuracy; however, for practical deployment, a critical assessment of the accuracy, latency, and harmonization with current transportation systems needs to be carefully considered. This survey reviews recent advancements in the UAV-based traffic monitoring, with a primary focus being deep neural network models for traffic analytics in various urban settings. Three main challenges identified in the literature are ensuring compatibility with traffic control systems, achieving real-time processing to optimize traffic flow, and maintaining robust detection in different environmental conditions. Existing solutions often lack comprehensive frameworks for utilizing UAV captured data to respond to incidents and manage traffic effectively. Future research should focus on optimal detection models, edge processing, and adaptive control integration to improve the responsiveness of urban traffic management.
Chinese Translation
在智能交通系统（ITS）中，基于无人机（UAV）的监控为交通监测提供了一种创新的解决方案，具有广泛的覆盖范围和实时数据收集能力。与固定的地面基础设施相比，无人机能够对动态交通做出响应，但也面临着在不同高度进行车辆检测、补偿运动引起的图像变化以及高分辨率图像的高效处理等挑战。深度学习在提高检测准确性方面发挥了重要作用；然而，在实际部署中，需要仔细评估准确性、延迟以及与现有交通系统的协调性。本综述回顾了无人机交通监测的最新进展，重点关注在各种城市环境中用于交通分析的深度神经网络模型。文献中识别出的三个主要挑战是确保与交通控制系统的兼容性、实现实时处理以优化交通流量，以及在不同环境条件下保持稳健的检测。现有解决方案往往缺乏利用无人机捕获数据有效响应事件和管理交通的综合框架。未来的研究应集中于优化检测模型、边缘处理和自适应控制集成，以提高城市交通管理的响应能力。
cs.CV / 16 / 2608.07572
BRACE: Taming Sharp Irregularities via Barycentric Rational Forecasting for Fast Diffusion Transformers Inference
BRACE：通过重心有理预测驯服尖锐不规则性，以实现快速扩散变换器推理
Abstract
Diffusion Transformers (DiTs) have demonstrated exceptional performance in high-fidelity image and video generation. To alleviate their massive computational overhead, temporal feature caching has been proposed to bypass redundant computations. However, existing cache-then-forecast methods driven by derivative-based polynomials often cause severe quality degradation under high acceleration due to unstable long-step predictions. To address this bottleneck, we propose Barycentric Rational Forecasting with Chebyshev Enhancement (BRACE). Motivated by the observation that DiT feature trajectories are globally smooth yet frequently exhibit sharp irregularities and local non-smoothness, BRACE shifts the paradigm from derivative-driven polynomial extrapolation to feature-driven rational forecasting. Specifically, it maintains a local sliding window to cache sparse historical features and leverages adapted Chebyshev weights to formulate a barycentric rational function, directly aggregating these raw features to ensure numerical stability. Extensive experiments demonstrate that BRACE achieves state-of-the-art quality-efficiency trade-offs across various DiT architectures with negligible computational overhead.
Chinese Translation
扩散变换器（DiTs）在高保真图像和视频生成方面表现出色。为了减轻其巨大的计算开销，提出了时间特征缓存方法以绕过冗余计算。然而，现有的基于导数多项式的缓存-再预测方法在高加速下常常导致严重的质量下降，因为长步预测的不稳定性。为了解决这一瓶颈，我们提出了带有切比雪夫增强的重心有理预测（BRACE）。基于观察到的DiT特征轨迹在全局上是平滑的，但经常表现出尖锐的不规则性和局部的不平滑性，BRACE将范式从基于导数的多项式外推转变为基于特征的有理预测。具体而言，它维护一个局部滑动窗口以缓存稀疏的历史特征，并利用调整后的切比雪夫权重来构造重心有理函数，直接聚合这些原始特征以确保数值稳定性。大量实验表明，BRACE在各种DiT架构中实现了最先进的质量-效率权衡，同时几乎没有计算开销。
cs.CV / 17 / 2608.07574
Multimodal Skin Lesion Classification with Swin Transformer and Clinical Metadata Fusion
基于Swin Transformer和临床元数据融合的多模态皮肤病变分类
Abstract
Skin lesion classification plays an important role in supporting the early diagnosis of skin cancer. However, automated analysis remains challenging due to class imbalance, inter-class similarity, and intra-class variability in dermoscopic images. This paper proposes a multimodal classification framework that combines Swin Transformer-based image features with structured clinical metadata to improve diagnostic performance through integrated visual-context learning. Experiments on a publicly available dataset show that the proposed model achieves a test accuracy of 92.55% and a macro F1-score of 91.33%, with strong performance across minority classes. Temperature scaling is applied as a post-hoc calibration method, resulting in a reduction in expected calibration error and improving prediction reliability, while uncertainty estimation is incorporated to further assess the confidence of model predictions. Qualitative explainability analysis further shows that the model focuses on lesion regions during inference. Therefore, the results demonstrate that multimodal fusion, combined with calibration and interpretability analysis, provides an effective and trustworthy approach for automated skin lesion classification.
Chinese Translation
皮肤病变分类在支持皮肤癌早期诊断中发挥着重要作用。然而，由于类别不平衡、类间相似性和类内变异性，自动化分析仍然面临挑战。本文提出了一种多模态分类框架，将基于Swin Transformer的图像特征与结构化临床元数据相结合，通过集成视觉上下文学习来提高诊断性能。在一个公开可用的数据集上的实验表明，所提出的模型在测试集上达到了92.55%的准确率和91.33%的宏观F1分数，在少数类上表现出色。温度缩放作为后处理校准方法被应用，减少了预期校准误差并提高了预测可靠性，同时引入不确定性估计以进一步评估模型预测的信心。定性可解释性分析进一步表明，模型在推理过程中关注于病变区域。因此，结果表明，多模态融合结合校准和可解释性分析，为自动化皮肤病变分类提供了一种有效且值得信赖的方法。
cs.CV / 18 / 2608.07575
Beyond Isotropic Assumptions: Continuity-Constrained Segmentation and GPU Morphometry for Nanoscale GBM Analysis
超越各向同性假设：连续性约束分割与GPU形态测量在纳米级GBM分析中的应用
Abstract
Confocal microscopy of optically cleared and swelled tissue resolves complex biological structures in 3D, but such acquisitions are highly anisotropic: along the under-sampled axial direction the structure can appear discontinuous, hampering reconstruction and automated quantitative analysis. The usual remedy upsamples the axial dimension to an isotropic volume before training a segmentation model, which requires dense annotations in the upsampled space, a prohibitive labeling burden. We present an end-to-end, GPU-accelerated framework that overcomes this without additional annotations. The model is trained on the native acquisition volume; random rotation of training patches leverages the well-resolved lateral plane to supply the missing axial information, and a z-axis continuity loss keeps neighboring slices consistent. We adapt both a convolutional (3D U-Net) and a transformer (SwinUNETR) backbone, aggregate overlapping patches by Gaussian consensus, and compute point-spread-function-corrected membrane thickness by ray-surface intersection on the GPU. We apply the method to the glomerular basement membrane (GBM), a thin, highly convoluted part of the kidney's filtration barrier that grows more irregular in disease. Segmentation accuracy matches inter-expert agreement. Continuity-aware training improves reconstruction smoothness and suppresses a periodic terracing artifact at minimal accuracy cost. We quantify GBM thickness across the reconstructed 3D surface and capture disease-related thickening, enabling fully automated anisotropic 3D morphometry of biological structures without dense volumetric labels or image restoration.
Chinese Translation
光学清晰和膨胀组织的共聚焦显微镜能够在三维中解析复杂的生物结构，但这种获取方式具有高度各向异性：在欠采样的轴向方向上，结构可能显得不连续，妨碍重建和自动定量分析。通常的解决方案是在训练分割模型之前，将轴向维度上采样到各向同性体积，这需要在上采样空间中进行密集标注，造成了沉重的标注负担。我们提出了一种端到端的GPU加速框架，能够在没有额外标注的情况下克服这一问题。该模型在原始获取体积上进行训练；通过随机旋转训练补丁，利用良好解析的横向平面来提供缺失的轴向信息，同时z轴连续性损失保持相邻切片的一致性。我们适配了卷积（3D U-Net）和变换器（SwinUNETR）主干，通过高斯共识聚合重叠补丁，并通过GPU上的光线-表面交点计算点扩散函数校正的膜厚度。我们将该方法应用于肾小管基底膜（GBM），这是肾脏过滤屏障中一部分薄而高度卷曲的结构，在疾病中变得更加不规则。分割精度与专家间一致性相匹配。关注连续性的训练提高了重建的平滑度，并在最小精度损失的情况下抑制了周期性阶梯伪影。我们量化了重建的三维表面上的GBM厚度，并捕捉到与疾病相关的增厚，从而实现了无需密集体积标签或图像恢复的完全自动化各向异性三维形态测量。
cs.CV / 19 / 2608.07577
Open-World Hierarchical Perception: Taxonomic Abstraction over Class-Agnostic Proposals for the Safe Handling of Out-of-Vocabulary Road Objects
开放世界层次感知：针对超出词汇范围的道路物体安全处理的无类别提议的分类抽象
Abstract
A closed-set detector for autonomous driving must assign every object one of a fixed set of labels. On an object outside that set (a horse-drawn carriage, road debris, livestock on a rural road) it can only force a confident but wrong specific label or drop the object. Prior work in this series replaced the flat label set with a hierarchical taxonomy and a runtime abstraction rule, but evaluated it only on the boxes a closed detector already produces. This paper takes the layer open-world: we place taxonomic abstraction on top of class-agnostic region proposals so objects the closed detector never boxes can still be classified or flagged; we report a feasibility study of three open-world signals (class-agnostic segmentation, appearance-based out-of-distribution scoring, monocular depth) that shows why no single 2D cue suffices and how they compose; and we run the evaluation the earlier papers could not, a ground-truth leave-classes-out benchmark on real annotated objects. Holding out seven COCO classes and classifying their 235 ground-truth crops, a flat closed head emits a confident wrong specific label 100% of the time (37% of them in the wrong super-category, e.g. an animal named as a vehicle), whereas the hierarchical layer emits zero confident wrong specific labels and safely handles 94% of the objects (a correct super-category, or an explicit UNKNOWN OBSTACLE). We are explicit that this is a safety result, not a specificity one: the correct super-category is recovered only 26% of the time and the remaining 69% are conservatively flagged unknown. The contribution is an open-world perception layer that never makes a confident categorical mistake on an out-of-vocabulary object, together with an honest account of its cost.
Chinese Translation
闭集检测器在自动驾驶中必须为每个物体分配固定标签集中的一个标签。对于该集合外的物体（如马车、道路碎片、乡村道路上的牲畜），它只能强制分配一个自信但错误的具体标签，或放弃该物体。之前的研究将平面标签集替换为层次分类法和运行时抽象规则，但仅在闭集检测器已经生成的框上进行了评估。本文将层次分类法应用于开放世界：我们在无类别区域提议之上进行分类抽象，以便闭集检测器从未框定的物体也能被分类或标记；我们报告了三种开放世界信号的可行性研究（无类别分割、基于外观的分布外评分、单目深度），展示了为何单一的二维线索不足以满足需求以及它们如何组合；我们进行了早期论文无法进行的评估，即在真实标注物体上的真实离类基准。我们保留了七个COCO类别，并对其235个真实标注裁剪进行分类，结果显示平面闭集头100%时间发出自信的错误具体标签（其中37%属于错误的超级类别，例如将动物标记为车辆），而层次层发出零个自信的错误具体标签，并安全处理94%的物体（一个正确的超级类别，或一个明确的未知障碍物）。我们明确指出这是一个安全结果，而非特异性结果：正确的超级类别仅在26%的情况下被恢复，其余69%被保守地标记为未知。我们的贡献是一个开放世界感知层，它在超出词汇范围的物体上从不做出自信的类别错误，同时诚实地说明了其成本。
cs.CV / 20 / 2608.07579
Geometry Beats Estimated Depth: RGB-Only Multi-Camera 3D Tracking under Sim2Real
几何优于估计深度：在Sim2Real环境下的RGB单摄像头多摄像头3D跟踪
Abstract
The AI City Challenge 2026 Track 1 evaluates multi-camera 3D perception in large indoor warehouses under a synthetic-to-real (Sim2Real) setting; depth is available only for training and validation, so inference is RGB-only. We use two RGB-only routes as a controlled test of one hypothesis: that cross-view geometric consistency, not monocular depth accuracy, governs performance under Sim2Real. The first is a geometry-first pipeline: YOLO11x detection, homography lifting to the world frame, class-level 3D size priors, multi-camera fusion, world-coordinate tracking, and offline tracklet stitching. The second is estimated-depth pseudo-LiDAR: monocular depth (D4RT, Metric3D~v2) back-projected into a fused point cloud and passed to a 3D detector (V-DETR), mirroring prior point-cloud winners that used depth. The gap is decisive: geometry-first reaches 13.0 3D HOTA (51.6 LocA), whereas pseudo-LiDAR collapses to 0.12 (9.2 LocA). We trace the collapse to cross-view inconsistency of monocular depth---scale correction is necessary but not sufficient---which domain-adaptation fine-tuning does not repair within budget. Within the geometry pipeline, offline stitching is the only intervention that helps; SAHI detection, appearance Re-ID, learned lifting, RT-DETR ensembling, test-time augmentation, and domain randomization all fail to beat the baseline detector. The bottlenecks are complementary: detection quality bounds the geometry route (DetA), localization consistency bounds pseudo-LiDAR (LocA). We release a complete, reproducible RGB-only pipeline and ablation.
Chinese Translation
AI City Challenge 2026 Track 1评估在合成到真实（Sim2Real）环境下的大型室内仓库中的多摄像头3D感知；深度信息仅用于训练和验证，因此推理仅基于RGB。我们使用两条仅基于RGB的路线作为对一个假设的控制测试：跨视角几何一致性，而非单目深度准确性，主导了Sim2Real下的性能。第一条是几何优先的流程：YOLO11x检测、同质性提升到世界坐标系、类别级3D尺寸先验、多摄像头融合、世界坐标跟踪和离线轨迹拼接。第二条是估计深度伪LiDAR：单目深度（D4RT, Metric3D~v2）反投影到融合点云中，并传递给3D检测器（V-DETR），与之前使用深度的点云获胜者相似。结果差距显著：几何优先达到13.0的3D HOTA（51.6 LocA），而伪LiDAR则降至0.12（9.2 LocA）。我们将这一崩溃归因于单目深度的跨视角不一致性——尺度校正是必要但不足的——而领域适应微调在预算内无法修复。在几何流程中，离线拼接是唯一有效的干预；SAHI检测、外观重识别、学习提升、RT-DETR集成、测试时增强和领域随机化均未能超越基线检测器。这些瓶颈是互补的：检测质量限制了几何路线（DetA），定位一致性限制了伪LiDAR（LocA）。我们发布了一个完整、可重复的仅基于RGB的流程和消融实验。
cs.CV / 21 / 2608.07580
Real-time physics inversion for retrieval of sub-pixel wildfire temperatures from VSWIR imaging spectroscopy
基于实时物理反演从VSWIR成像光谱中提取亚像素野火温度
Abstract
In this work, we present a wildfire temperature retrieval framework for VSWIR imaging spectroscopy data, employed on data from NASA's Airborne Visible Infrared Imaging Spectrometer (AVIRIS-3). The retrieval framework utilizes a full-physics approach in which a forward model is employed to resolve both solar and emitted radiance derived from a temperature distribution and utilizes the full spectral range in the residual fit. To optimize the forward model retrieval, we use state-of-the-art nonlinear least squares methods implemented for fast convergence on the on-board GPU, allowing for estimation of effective fire temperature within flight cadence. We verify the forward model assumptions on simulated spectra with an injected thermal signature and find good agreement with an RMSE of $41.8$ Kelvin (K). We apply the retrieval over the full 2025 FireSense AVIRIS-3 campaign, totaling 168 overflights with probable active fire spectra, and demonstrate a residual radiance fit of $\leq 10\%$ across bands in the short-wave infrared (SWIR). Lastly, we verify the applicability of the retrieved posterior fire temperature parameters to generalize to space-borne imaging spectrometers such as EMIT, by retrieving at coarsened spatial resolution. We find that the posterior distribution exhibits good coverage of the underlying sub-pixel temperature range with an absolute error of $30$ K across quantiles and a mean absolute error of $27.16$ K between spatial resolutions.
Chinese Translation
在本研究中，我们提出了一种用于VSWIR成像光谱数据的野火温度提取框架，该框架应用于NASA的空气可见红外成像光谱仪（AVIRIS-3）数据。该提取框架采用全物理方法，其中使用前向模型来解析由温度分布导出的太阳辐射和发射辐射，并在残差拟合中利用全光谱范围。为了优化前向模型提取，我们使用了最先进的非线性最小二乘法，这些方法在机载GPU上实现了快速收敛，从而能够在飞行周期内估算有效火温。我们在注入热特征的模拟光谱上验证了前向模型假设，发现与均方根误差（RMSE）为41.8开尔文（K）的一致性良好。我们在2025年FireSense AVIRIS-3活动的全部168次可能活火光谱的航次中应用了该提取，并在短波红外（SWIR）波段中展示了残差辐射拟合小于或等于10%。最后，我们通过在粗化空间分辨率下提取，验证了所提取的后验火温参数对空间成像光谱仪（如EMIT）的适用性。我们发现后验分布对潜在的亚像素温度范围具有良好的覆盖，绝对误差为30 K，空间分辨率之间的平均绝对误差为27.16 K。
cs.CV / 22 / 2608.07581
Multi-Branch Policy Optimization for Multimodal Large Language Models
多分支策略优化用于多模态大型语言模型
Abstract
Group-based reinforcement learning methods for multimodal large language models typically rely on trajectory-level credit assignment that applies a single advantage to all tokens in a response. However, multimodal reasoning involves substantially higher perceptual uncertainty than text-only settings, where the model must repeatedly re-examine visual information to verify intermediate interpretations, and different visual groundings can lead to divergent reasoning paths, making such uniform credit assignment particularly inadequate and causing relative advantages to progressively degenerate toward zero. To address these challenges, we propose Multi-Branch Policy Optimization (MBPO), a tree-based framework that constructs reasoning trees at vision-language decision boundaries, enabling sibling branches to explore diverse visual hypotheses and assigning segment-level credit through branch-relative advantages. We further introduce a temporal replay buffer to reuse informative segments while controlling policy staleness. Experiments on several multimodal reasoning benchmarks show that MBPO outperforms representative baselines, improving both learning signal quality and optimization efficiency. The code is publicly available at https://github.com/ShuaiLyu0110/MBPO.
Chinese Translation
基于群体的多模态大型语言模型强化学习方法通常依赖于轨迹级别的信用分配，将单一的优势应用于响应中的所有标记。然而，多模态推理涉及的感知不确定性显著高于仅文本的设置，在此情况下，模型必须反复检查视觉信息以验证中间解释，而不同的视觉基础可能导致不同的推理路径，使得这种统一的信用分配特别不够，导致相对优势逐渐退化为零。为了解决这些挑战，我们提出了多分支策略优化（Multi-Branch Policy Optimization, MBPO），这是一种基于树的框架，在视觉-语言决策边界构建推理树，使得兄弟分支能够探索多样的视觉假设，并通过分支相对优势进行段级信用分配。我们进一步引入了时间重放缓冲区，以重用信息丰富的段，同时控制策略的过时性。在多个多模态推理基准上的实验表明，MBPO优于代表性基线，提升了学习信号质量和优化效率。代码已公开发布在 https://github.com/ShuaiLyu0110/MBPO。
cs.CV / 23 / 2608.07582
Predictive Failure Detection in Network Hardware Using Thermal Imaging and Deep Learning with Sensor Fusion
基于热成像和深度学习的网络硬件预测性故障检测与传感器融合
Abstract
Unplanned network hardware malfunctions can interrupt services and result in expensive downtime in data centers. A deep learning-based predictive maintenance strategy is presented that utilizes thermal imaging and power sensor data to detect early indicators of equipment breakdown in routers, switches, and servers. A simulated dataset was generated comprising annotated thermal pictures and power readings indicative of three operating states: Normal, Warning, and Critical. Three ImageNet-pretrained convolutional neural network (CNN) models ResNet-50, InceptionV3, and VGG16 were assessed together with a multi-modal CNN-LSTM fusion model that integrates visual and sensor time-series information. Experiments were performed with and without pre-processing procedures, including region-of-interest (ROI) extraction and normalization. In the absence of pre-processing, CNNs attained moderate accuracy (e.g., ResNet-50 at 52%), but ROI-based pre-processing significantly enhanced performance (ResNet-50 accuracy reaching 91%). The CNN-LSTM model attained the greatest accuracy of 94%, with precision and recall approaching 95%, illustrating the effectiveness of multi-modal fusion. The results validate that domain-specific pre-processing and sensor fusion substantially improve early failure prediction, providing a potential foundation for proactive maintenance of network hardware through non-intrusive monitoring.
Chinese Translation
未计划的网络硬件故障可能会中断服务，并导致数据中心的高昂停机成本。本文提出了一种基于深度学习的预测性维护策略，利用热成像和电力传感器数据来检测路由器、交换机和服务器设备故障的早期指示。生成了一个模拟数据集，包含标注的热成像图片和电力读数，指示三种操作状态：正常、警告和危急。评估了三个经过ImageNet预训练的卷积神经网络（CNN）模型：ResNet-50、InceptionV3和VGG16，以及一个多模态CNN-LSTM融合模型，该模型整合了视觉和传感器时间序列信息。进行了有无预处理程序的实验，包括感兴趣区域（ROI）提取和归一化。在没有预处理的情况下，CNN的准确率达到了中等水平（例如，ResNet-50为52%），但基于ROI的预处理显著提高了性能（ResNet-50的准确率达到91%）。CNN-LSTM模型达到了最高的94%准确率，精确率和召回率接近95%，展示了多模态融合的有效性。结果验证了特定领域的预处理和传感器融合显著改善了早期故障预测，为通过非侵入式监测实现网络硬件的主动维护提供了潜在基础。
cs.CV / 24 / 2608.07584
ComplexityWorld: Benchmarking Vision-Language Models on Verifiable Visual Decision Making
复杂性世界：在可验证视觉决策中的视觉-语言模型基准测试
Abstract
Vision-language models (VLMs) have made rapid progress in visual perception and increasingly support real-world tasks that depend on images. Many such tasks, however, require more than rec- ognizing what an image contains: a model must use visual evidence to make a complete decision whose parts jointly satisfy global constraints. We introduce COMPLEXITYWORLD, a benchmark of 390 tasks across 39 domain-inspired visual worlds and 29 decision categories. Each task is generated from a hidden structured specification, rendered as a visual scene, and scored by an exe- cutable verifier that accepts any feasible solution. Under direct inference, all evaluated models ex- cept GPT-5.6-Sol remain below 40% verifier ac- ceptance rate (VAR), while GPT-5.6-Sol reaches 75.6%. Performance improves substantially when the same decision information is made explicit in structured form, yet varies sharply across equiva- lent visual presentations. Agent scaffolds provide smaller, model-dependent gains. Together, these results reveal a persistent visual-to-decision bot- tleneck that additional inference alone does not remove.
Chinese Translation
视觉-语言模型（VLMs）在视觉感知方面取得了快速进展，并越来越多地支持依赖图像的现实世界任务。然而，许多此类任务不仅需要识别图像内容：模型必须利用视觉证据做出完整的决策，其各个部分共同满足全局约束。我们引入了COMPLEXITYWORLD，这是一个包含390个任务的基准，涵盖39个受领域启发的视觉世界和29个决策类别。每个任务都是从一个隐藏的结构规范生成的，以视觉场景的形式呈现，并由一个可执行的验证器进行评分，该验证器接受任何可行的解决方案。在直接推理下，除GPT-5.6-Sol外，所有评估的模型的验证器接受率（VAR）均低于40%，而GPT-5.6-Sol达到了75.6%。当相同的决策信息以结构化形式明确呈现时，性能显著提高，但在等效的视觉呈现中变化明显。代理支架提供了较小的、依赖于模型的增益。综合来看，这些结果揭示了一个持续存在的视觉到决策的瓶颈，仅靠额外的推理无法消除。
cs.CV / 25 / 2608.07585
LAVE: Latent Visual Evidence-Enhanced Planning for Video Tool-use Agents
LAVE：基于潜在视觉证据增强的视频工具使用代理规划
Abstract
Long-video understanding requires models to efficiently acquire and reuse sparse visual evidence from long and redundant video streams. Recent video tool-use agents address this challenge by iteratively invoking visual Tools at different temporal scales, but their Tool-Planner communication typically relies on textual observations. Such text-only interfaces provide lossy summaries of Tool computations, causing previously computed visual evidence not verbalized to be discarded and unavailable for subsequent planning. We identify this limitation as the Tool observation bottleneck and propose Latent Visual Evidence-Enhanced Planning (LAVE), a training-free framework for reusing latent visual evidence from completed Tool calls. LAVE introduces a dual-channel observation interface: the visible channel preserves the original textual trajectory, while the latent channel stores pre-verbal visual updates with their Tool roles, source-frame timestamps, and visual locations. During planning, LAVE retrieves evidence relevant to the current Planner state but not covered by textual observations, and integrates it through bounded timestamp-aligned latent updates with entropy-constrained frame-time routing. This enables video agents to reuse existing visual computation without additional training, frame replay, or modifications to the original orchestration. Extensive experiments on Video-MME, LongVideoBench, and CG-Bench show that LAVE consistently improves video tool-use agents across backbones. Under a comparable frame budget, LAVE improves the Video-MME overall score by 3.76 points over the strongest baseline, demonstrating the effectiveness of latent visual evidence reuse for multi-step video-agent planning.
Chinese Translation
长视频理解要求模型有效地获取和重用来自冗长视频流的稀疏视觉证据。近期的视频工具使用代理通过在不同时间尺度上迭代调用视觉工具来应对这一挑战，但它们的工具规划者通信通常依赖于文本观察。这种仅基于文本的接口提供了工具计算的有损摘要，导致未被口头表达的先前计算的视觉证据被丢弃，无法用于后续规划。我们将这一限制识别为工具观察瓶颈，并提出潜在视觉证据增强规划（LAVE），这是一个无需训练的框架，用于重用已完成工具调用的潜在视觉证据。LAVE引入了一个双通道观察接口：可见通道保留原始文本轨迹，而潜在通道存储具有其工具角色、源帧时间戳和视觉位置的预口头视觉更新。在规划过程中，LAVE检索与当前规划者状态相关但未被文本观察覆盖的证据，并通过有界时间戳对齐的潜在更新与熵约束的帧时间路由进行整合。这使得视频代理能够在不需要额外训练、帧重放或对原始编排进行修改的情况下重用现有的视觉计算。在Video-MME、LongVideoBench和CG-Bench上的大量实验表明，LAVE在不同骨干网络上始终提高了视频工具使用代理的性能。在可比的帧预算下，LAVE在最强基线的基础上将Video-MME的整体得分提高了3.76分，证明了潜在视觉证据重用于多步骤视频代理规划的有效性。
cs.CV / 26 / 2608.07586
MAGIC-SSCIL: Manifold Anchoring and Geometric Incremental Calibration for Semi-Supervised Class Incremental Learning
MAGIC-SSCIL：用于半监督类别增量学习的流形锚定与几何增量校准
Abstract
Semi-supervised Class Incremental Learning (SSCIL) is a severe challenge for neural networks, and it is hardest in the exemplar-free setting where no past data may be stored. Existing methods forget catastrophically due to feature drift, and their pseudo-labels become increasingly unreliable as the label space grows. In this paper, we propose MAGIC (Manifold Anchoring and Geometric Incremental Calibration), a framework that stabilizes plasticity without storing exemplars. MAGIC's design centers on two components. The first is Soft-Weighted Geometry Calibration (SWGC), which uses graph-based label propagation on the learner's plastic feature space to weight and calibrate class means and variances computed on the frozen backbone; from these calibrated Gaussians, we sample phantom features that stand in for data from previous tasks. The second is a Geometric Structural Alignment (GSA) objective that preserves representation topology by matching the relational structure of student and teacher heads and aligning feature anchors with the fixed classifier prototypes, locking the orientation of the feature space. Together, these constraints keep the adapter from drifting, so geometric relations between classes remain stable as new classes arrive. We implement MAGIC with a frozen ResNet-18 backbone and a learnable plastic adapter. Across CIFAR-100, CUB-200, and ImageNet-R, at label ratios of 1%, 5%, and 10%, MAGIC improves average incremental accuracy over most of the supervised CIL methods equipped with FixMatch and native SSCIL baselines; the largest gains occur in the fine-grained, low-label setting, where confidence thresholding fails most clearly.
Chinese Translation
半监督类别增量学习（SSCIL）是神经网络面临的一个重大挑战，尤其是在不存储过去数据的示例自由设置中，难度更大。现有方法由于特征漂移而导致灾难性遗忘，且随着标签空间的扩大，其伪标签变得越来越不可靠。本文提出了MAGIC（流形锚定与几何增量校准），一个在不存储示例的情况下稳定可塑性的框架。MAGIC的设计围绕两个组件展开。第一个是软加权几何校准（SWGC），它在学习者的可塑特征空间中使用基于图的标签传播来加权和校准在冻结的主干网络上计算的类别均值和方差；从这些校准的高斯分布中，我们采样出代表先前任务数据的虚拟特征。第二个是几何结构对齐（GSA）目标，通过匹配学生和教师头部的关系结构并将特征锚点与固定分类器原型对齐，从而保持表示拓扑，锁定特征空间的方向。通过这些约束，适配器避免漂移，从而在新类别到来时保持类别之间的几何关系稳定。我们在冻结的ResNet-18主干和可学习的可塑适配器上实现MAGIC。在CIFAR-100、CUB-200和ImageNet-R数据集上，在1%、5%和10%的标签比例下，MAGIC在大多数配备FixMatch和本地SSCIL基线的监督CIL方法中提高了平均增量准确率；在细粒度、低标签设置中，MAGIC的增益最大，此时置信度阈值的效果最为明显。
cs.CV / 27 / 2608.07598
NewtonGS: Physics-Structured Object-Level Neural Newtonian Dynamics for Gaussian Scene Animation
NewtonGS：基于物理结构的对象级神经牛顿动力学用于高斯场景动画
Abstract
Animating objects in a static 3D Gaussian scene requires an explicit object-level dynamic state and a controllable model of object motion. Existing dynamic Gaussian methods primarily reconstruct time-varying scenes or simulate deformation, rather than provide compact object states for direct control. To address this gap, we present NewtonGS, a physics-structured framework for object-level state rollout and Gaussian scene animation. NewtonGS represents each object with a 22-dimensional state covering pose, linear and angular velocity, anisotropic scale and its rate, mass, and contact properties. Its Gaussian Neural Newtonian Dynamics (Gaussian-NND) model combines analytic translation, quaternion kinematics, gravity, damping, and scale-restoration dynamics with learned continuous and contact residuals. A discrete event map handles floor contact. Predicted poses and scales define a shared affine transformation that updates the means and covariances of all Gaussians associated with each object. We construct two procedurally generated datasets: State-32 for state-rollout evaluation and Gaussian-32 for state-to-Gaussian transformation. On both the in-distribution and velocity-range-shift splits of State-32, NewtonGS achieves lower trajectory RMSE, final displacement error, and velocity RMSE than five analytic baselines. Experiments on Gaussian-32 further demonstrate effective conversion from predicted states to animated Gaussian objects.
Chinese Translation
在静态3D高斯场景中对物体进行动画处理需要明确的对象级动态状态和可控的物体运动模型。现有的动态高斯方法主要重建时间变化的场景或模拟变形，而不是提供紧凑的对象状态以便直接控制。为了解决这一问题，我们提出了NewtonGS，一个用于对象级状态展开和高斯场景动画的基于物理结构的框架。NewtonGS用一个22维状态表示每个物体，该状态涵盖姿态、线性和角速度、各向异性缩放及其速率、质量和接触属性。其高斯神经牛顿动力学（Gaussian-NND）模型结合了解析平移、四元数运动学、重力、阻尼和缩放恢复动力学，以及学习到的连续和接触残差。离散事件映射处理地面接触。预测的姿态和缩放定义了一个共享的仿射变换，更新与每个物体相关的所有高斯的均值和协方差。我们构建了两个程序生成的数据集：State-32用于状态展开评估，Gaussian-32用于状态到高斯的转换。在State-32的分布内和速度范围转移分割上，NewtonGS的轨迹均方根误差（RMSE）、最终位移误差和速度RMSE均优于五个解析基线。在Gaussian-32上的实验进一步证明了从预测状态到动画高斯物体的有效转换。
cs.CV / 28 / 2608.07616
HSMLA: Hierarchical Softmax Multi-scale Linear Attention for Efficient Vision Transformers
HSMLA：用于高效视觉变换器的层次化软最大多尺度线性注意力
Abstract
Vision transformers face significant computational overheads in high-resolution dense prediction due to the quadratic complexity of self-attention. Linear attention offers efficiency but sacrifices local context modeling. We propose \textbf{HSMLA (Hierarchical Softmax Multi-scale Linear Attention)}, which combines ReLU-based linear attention for global context, selective softmax refinement for critical local features, and multi-scale token representations via depthwise convolutions. HSMLA achieves superior accuracy-efficiency trade-offs: up to $4.2\times$ inference-time speedup across dense prediction tasks, $87.3%$ Dice with $3.2\times$ speedup on CT organ segmentation, and $94.2%$ AUC with $4.1\times$ speedup on pathology WSI.
Chinese Translation
视觉变换器在高分辨率密集预测中面临显著的计算开销，这主要源于自注意力的平方复杂度。线性注意力提供了效率，但牺牲了局部上下文建模。我们提出了 extbf{HSMLA（层次化软最大多尺度线性注意力）}，该方法结合了基于ReLU的线性注意力以获取全局上下文、选择性软最大细化以捕捉关键局部特征，以及通过深度卷积实现的多尺度标记表示。HSMLA在准确性与效率的权衡上表现优越：在密集预测任务中实现了最高$4.2 imes$的推理时间加速，在CT器官分割中达到了$87.3\%$的Dice系数和$3.2 imes$的加速，在病理全切片图像（WSI）中达到了$94.2\\%$的AUC和$4.1 imes$的加速。
cs.CV / 29 / 2608.07620
FlowErase-OPD: Multi-Concept Erasure via Anchored On-Policy Distillation in Flow Matching Models
FlowErase-OPD：通过锚定的在线策略蒸馏实现流匹配模型中的多概念消除
Abstract
Recent advances in flow matching models have substantially improved the quality of text-to-image generation, but have also raised increasing safety concerns due to their potential to generate harmful or undesirable content. Existing concept erasure methods for flow matching models predominantly focus on removing individual concepts, while effectively erasing multiple concepts simultaneously remains challenging. We propose FlowErase-OPD, a framework for multi-concept erasure based on on-policy distillation (OPD). Our approach first distills multiple single-concept erased models into a unified LoRA module and introduces Anchored Multi-Teacher Distillation (AMTD), which incorporates a retention teacher to mitigate the trade-off between concept erasure and preservation of generative capabilities. To further improve the coordination of multiple erasure objectives, we develop Adaptive Retention Control (ARC), which dynamically adjusts the sampling frequency and loss weight of each erasure teacher, together with the relative contribution of erasure and retention teachers throughout training. Extensive experiments on nudity, object, and artistic-style erasure demonstrate that FlowErase-OPD consistently improves the trade-off between erasure effectiveness, image quality, and semantic alignment, achieving state-of-the-art performance across diverse multi-concept erasure settings. Furthermore, the resulting models exhibit strong robustness against adversarial attacks. These results highlight the potential of on-policy distillation as a principled framework for safe and controllable generation in flow matching models.
Chinese Translation
近年来，流匹配模型的进展显著提高了文本到图像生成的质量，但也因其可能生成有害或不良内容而引发了日益严重的安全担忧。现有的流匹配模型概念消除方法主要集中在去除单一概念，而同时有效地消除多个概念仍然具有挑战性。我们提出了FlowErase-OPD，这是一个基于在线策略蒸馏（OPD）的多概念消除框架。我们的方法首先将多个单概念消除模型蒸馏为一个统一的LoRA模块，并引入了锚定多教师蒸馏（AMTD），该方法结合了保留教师，以减轻概念消除与生成能力保留之间的权衡。为了进一步改善多个消除目标的协调性，我们开发了自适应保留控制（ARC），该方法动态调整每个消除教师的采样频率和损失权重，以及在整个训练过程中消除教师和保留教师的相对贡献。针对裸体、物体和艺术风格消除的广泛实验表明，FlowErase-OPD始终改善了消除效果、图像质量和语义对齐之间的权衡，在多种多概念消除设置中实现了最先进的性能。此外，所得到的模型在对抗攻击下表现出强大的鲁棒性。这些结果突显了在线策略蒸馏作为流匹配模型中安全和可控生成的原则性框架的潜力。
cs.CV / 30 / 2608.07640
HeatCast: A Benchmark for Neighborhood-Scale LST Forecasting across 124 U.S. Cities
HeatCast：124个美国城市邻域尺度地表温度预测基准
Abstract
Land Surface Temperature (LST) is a widely used satellite-derived measure of urban surface heat, but there is no shared benchmark for forecasting it at 30 m. Prior studies usually cover one to three cities, use kilometer-scale products, or do not release data and code. We introduce HeatCast, a Landsat-based benchmark for monthly LST forecasting across 124 U.S. cities from 2013 through June 2025. HeatCast contains 30 m monthly tiles with LST, elevation, surfacereflectance RGB, three spectral indices, broadband albedo, quality masks, and Local Climate Zone (LCZ) labels, together with a fixed temporal split, LCZ-stratified metrics, and a reference evaluation harness. We evaluate a CNN+LSTM and Earthformer on next-month forecasting, where Earthformer reaches 7.74 K RMSE against 10.42 K for the CNN+LSTM. Forecasting from the eight nonLST channels alone reaches 7.72 K, against 8.15 K from LST history and 8.68 K from RGB. The data, code, and weights are released under MIT at https://doi.org/10.57967/hf/9889.
Chinese Translation
地表温度（LST）是一个广泛使用的卫星衍生城市表面热量的测量指标，但目前尚无共享的基准用于在30米尺度上进行预测。以往的研究通常只覆盖一到三个城市，使用公里尺度的产品，或未公开数据和代码。我们推出了HeatCast，这是一个基于Landsat的基准，旨在对2013年至2025年6月期间124个美国城市的月度LST进行预测。HeatCast包含30米的月度图块，提供LST、高程、表面反射率RGB、三个光谱指数、宽带反照率、质量掩膜以及地方气候区（LCZ）标签，并附有固定的时间划分、LCZ分层指标和参考评估工具。我们评估了CNN+LSTM和Earthformer在下月预测中的表现，其中Earthformer的均方根误差（RMSE）为7.74 K，而CNN+LSTM为10.42 K。仅使用八个非LST通道的预测RMSE为7.72 K，而使用LST历史数据的预测为8.15 K，RGB的预测为8.68 K。数据、代码和权重已在MIT许可下发布，链接为：https://doi.org/10.57967/hf/9889。
cs.CV / 31 / 2608.07643
Data collection from highways: a geometric, class-agnostic approach to embedded vehicle counting
高速公路数据收集：一种几何的、无类别的嵌入式车辆计数方法
Abstract
Traffic data collection is dominated today by deep object detectors followed by tracking-by-detection, a pipeline that presupposes what is often missing in practice: a detector already trained on the class one wants to count. We revisit a purely geometric traffic-sensing pipeline for Single Board Computers in which detection is class-agnostic: moving objects come from background subtraction and thresholding, and counting is decided by a geometric rule on an imaginary line across the road, a software inductive loop detector. With no object model, training set or per-object trajectory, it runs faster than real time on Raspberry Pi class hardware. Two counting rules are described: a constant average speed rule, whose expected accuracy is derived analytically as about 86% under a Gaussian speed distribution, and a self-calibrating pre-calibration rule that recovers the lane geometry from blob statistics and counts edges of lane occupancy, additionally yielding per-vehicle average speed at no extra cost. Over four videos the latter counts with 83.3%-100% accuracy; in a field deployment it reaches 91% against 37.5% for a blob-tracking baseline under the same compute budget. We report the observations of that period in detail: the resolution floor below which accuracy collapses, the frame rate floor at which vehicles alias past the counting line, the gap between short curated clips and long uncontrolled footage, and the trade-off between Python (easier to tune, 100% CPU) and C++ (40% CPU, thermally viable). These are properties of the sampling geometry, not of the hardware of the time, and still constrain edge deployments. We close by arguing where motion-based, class-agnostic detection remains the right tool: open-set classes with no annotated data, tight power budgets, privacy-constrained installations, and the cold start of mining training crops to bootstrap a learned detector.
Chinese Translation
当前交通数据收集主要依赖深度物体检测器，随后通过检测跟踪进行处理，这一流程假设了在实践中常常缺失的前提：一个已经针对所需计数类别训练好的检测器。我们重新审视了一种纯几何的交通感知流程，适用于单板计算机，其中检测是无类别的：移动物体通过背景减除和阈值处理获得，计数则通过在道路上想象的线上的几何规则决定，这是一种软件感应环路检测器。在没有物体模型、训练集或每个物体轨迹的情况下，它在树莓派（Raspberry Pi）级别的硬件上运行速度超过实时。我们描述了两种计数规则：一种是恒定平均速度规则，其预期准确率在高斯速度分布下通过分析推导为约86%；另一种是自校准的预校准规则，它通过斑点统计恢复车道几何，并计数车道占用的边缘，同时以零额外成本提供每辆车的平均速度。在四段视频中，后者的计数准确率为83.3%-100%；在一次现场部署中，其准确率达到91%，而在相同计算预算下，斑点跟踪基线的准确率仅为37.5%。我们详细报告了这一期间的观察结果：准确率崩溃的分辨率下限、车辆在计数线前混叠的帧率下限、短小策划片段与长时间无控制录像之间的差距，以及Python（更易于调优，100% CPU）与C++（40% CPU，热量可行性）的权衡。这些是采样几何的特性，而非当时硬件的特性，仍然限制了边缘部署。最后，我们论证了运动基础的无类别检测仍然是合适工具的场景：开放集类别没有标注数据、严格的功耗预算、隐私受限的安装，以及挖掘训练样本以启动学习检测器的冷启动问题。
cs.CV / 32 / 2608.07663
Keep It Simple: Multi-Key Episodic Memory Retrieval for Ultra-Long Video Understanding
保持简单：超长视频理解的多键情节记忆检索
Abstract
When videos extend from hours to days, directly processing them end-to-end becomes impractical for current Multi-modal Large Language Models (MLLMs). This ultra-long setting necessitates a two-stage paradigm: query-agnostic memory construction followed by retrieval-based inference. Prior work invests in complex memory construction to pre-model high-level relations in videos, despite not knowing the downstream query at build time. We instead prioritize high-recall retrievability during memory building, and defer query-specific, high-level relation composition to inference time. To this end, we propose MERIT(Multi-key Episodic Retrieval with Inference-time Temporal expansion), a simple yet effective agentic framework for ultra-long video understanding. First, we formulate an episodic multi-key representation that enables precise retrieval of fine-grained memories through a simple key-matching mechanism. Second, we introduce a neighbor filtering mechanism to capture broader semantic context without the massive computational overhead of global memory construction. This is achieved by expanding the temporal scope exclusively around the retrieved segments at inference time. By leveraging simple key-matching with this on-demand temporal expansion, MERIT achieves state-of-the-art performance across three long-video benchmarks: EgoLifeQA, LVBench, and Video-MME (Long).
Chinese Translation
当视频时长从数小时延伸至数天时，直接进行端到端处理对于当前的多模态大型语言模型（MLLMs）变得不切实际。这种超长设置需要一个两阶段的范式：查询无关的记忆构建，随后进行基于检索的推理。先前的研究投入了复杂的记忆构建，以预建视频中的高层关系，尽管在构建时并不知道下游查询。相反，我们在记忆构建过程中优先考虑高召回率的可检索性，并将查询特定的高层关系组合推迟到推理阶段。为此，我们提出了MERIT（多键情节检索与推理时时间扩展），这是一个简单而有效的自主框架，用于超长视频理解。首先，我们制定了一种情节多键表示，通过简单的键匹配机制实现对细粒度记忆的精确检索。其次，我们引入了一种邻域过滤机制，以捕捉更广泛的语义上下文，而无需进行全球记忆构建的巨大计算开销。这是通过在推理时仅围绕检索到的片段扩展时间范围来实现的。通过利用简单的键匹配与这种按需的时间扩展，MERIT在三个长视频基准测试（EgoLifeQA、LVBench 和 Video-MME（Long））上实现了最先进的性能。
cs.CV / 33 / 2608.07693
CosmosAlign: Adapting a World Foundation Model for Generative Traffic Video Forecasting
CosmosAlign：为生成交通视频预测适配世界基础模型
Abstract
Generative traffic video forecasting aims to synthesize long-horizon, temporally coherent future videos of traffic scenes from a short observation history and textual descriptions. In this paper, we present CosmosAlign, a generative traffic video forecasting framework built upon the pretrained Cosmos3-Nano world foundation model. Our approach is motivated by the observation that successfully adapting large pretrained world models to downstream forecasting tasks depends primarily on distribution alignment rather than increased model capacity. To this end, we propose a two-stage LoRA adaptation strategy that first aligns the conditioning-mode distribution with the target forecasting task, and then aligns the training captions with the model's native structured prompting interface through an LLM-based re-captioning pipeline. During inference, we further improve prediction quality using a fully training-free procedure consisting of consensus-based medoid sample selection and motion-adaptive blending of static scene regions. CosmosAlign achieves a final score of 76.49 on the AI City Challenge 2026 Track 5 benchmark, ranking first on the final leaderboard. Our code is publicly available at https://quangminhdinh.github.io/CosmosAlign/.
Chinese Translation
生成交通视频预测旨在从短期观察历史和文本描述中合成长时间跨度、时间一致的未来交通场景视频。在本文中，我们提出了CosmosAlign，这是一个基于预训练的Cosmos3-Nano世界基础模型构建的生成交通视频预测框架。我们的方法受到以下观察的启发：成功地将大型预训练世界模型适配到下游预测任务主要依赖于分布对齐，而非模型容量的增加。为此，我们提出了一种两阶段的LoRA适配策略，首先将条件模式分布与目标预测任务对齐，然后通过基于LLM（大语言模型）的重新标注管道将训练标题与模型的原生结构化提示接口对齐。在推理过程中，我们进一步通过基于共识的中位样本选择和静态场景区域的运动自适应混合，使用完全无训练的程序提高预测质量。CosmosAlign在AI City Challenge 2026 Track 5基准测试中最终得分为76.49，排名第一。我们的代码已公开发布，网址为https://quangminhdinh.github.io/CosmosAlign/。
cs.CV / 34 / 2608.07712
SpikeWorld: Fast-State Adaptation for Frozen Spiking World Models
SpikeWorld：冻结脉冲世界模型的快速状态适应
Abstract
A predictive model receives a self-supervised signal whenever the consequence of an action is observed. Using that signal after deployment is difficult when dynamics and semantics share parameters: freezing prevents adaptation, whereas weight updates require optimizer state and may alter the learned representation. Here we introduce SpikeWorld, a 1.45M-parameter sparse spiking model jointly trained for heterogeneous sensory prediction, semantics, image-text binding and action-conditioned dynamics. At deployment, all trained parameters are frozen. Delayed next-state residuals update two external paths: cumulative fixed-bank losses select the bounded action correction, while route-specific residual matrices refine next-state prediction. Neither path uses labels, teacher outputs, rewards, success signals or the true shift value. Joint optimization improves action next-state MSE by 17.10\% while also improving multimodal prediction, semantic accuracy and image-text retrieval. On held-out shear and attenuation streams, the combined external state improves aggregate prediction by 5.48\% and 30.01\%; its fixed-bank action path improves tracking by 24.20\% and 3.94\%, respectively. In a six-arm study comprising 450 new Meta-World trajectories (75 per arm), SpikeWorld raises frozen-policy reward by 7.90 (95\% CI [2.48, 14.06]); the 13.33-point success difference is descriptive (CI [0, 40]). For identical sensory inputs, model parameters and inherited semantic outputs remain bitwise unchanged. A 16-byte RLS estimator obtains the highest non-oracle reward on linear attenuation, showing that the contribution is not superior linear identification, but its integration with a frozen multimodal spiking checkpoint. Reference code is publicly available at https://github.com/Oooorca/SpikeWorld.
Chinese Translation
预测模型在观察到行动结果时接收自我监督信号。当动态和语义共享参数时，部署后使用该信号变得困难：冻结会阻止适应，而权重更新需要优化器状态并可能改变学习到的表示。在此，我们介绍了SpikeWorld，这是一种具有145万参数的稀疏脉冲模型，联合训练用于异构感知预测、语义、图像-文本绑定和基于行动的动态。在部署时，所有训练参数均被冻结。延迟的下一个状态残差更新两个外部路径：累积固定银行损失选择有界的行动修正，而特定路径的残差矩阵则细化下一个状态预测。两个路径均不使用标签、教师输出、奖励、成功信号或真实的转移值。联合优化使行动下一个状态均方误差（MSE）提高了17.10%，同时改善了多模态预测、语义准确性和图像-文本检索。在保留的剪切和衰减流中，组合外部状态使聚合预测分别提高了5.48%和30.01%；其固定银行行动路径在跟踪方面分别提高了24.20%和3.94%。在包含450个新Meta-World轨迹（每个臂75个）的六臂研究中，SpikeWorld将冻结策略奖励提高了7.90（95% CI [2.48, 14.06]）；13.33点的成功差异是描述性的（CI [0, 40]）。对于相同的感官输入，模型参数和继承的语义输出保持逐位不变。一个16字节的RLS估计器在线性衰减上获得了最高的非oracle奖励，表明其贡献并非优越的线性识别，而是与冻结的多模态脉冲检查点的整合。参考代码可在https://github.com/Oooorca/SpikeWorld公开获取。
cs.CV / 35 / 2608.07713
Tokenizer Generator Coupling in Medical Image Generation
医学图像生成中的分词器生成器耦合
Abstract
Latent medical image generators usually treat the tokenizer as fixed preprocessing. We test whether this separation is valid in a controlled ChestMNIST study at 64x64, crossing discrete tokenizers, generator families, and sampler settings under a shared latent grid, with continuous-latent reference cells. In this controlled setting, rankings depend jointly on the tokenizer, generator, and sampler: the best quantizer changes with the generator, and validation-based sampler selection changes the apparent generator ranking. We retrain the vocabulary-1024 interaction block at three seeds and the interaction survives (6 of 9 pairwise quantizer comparisons exceed three seed standard deviations), and we scope the wider single-seed grid accordingly. Reconstruction PSNR alone is not a reliable selection criterion; we instead introduce a generator-free statistic, neighbour-conditional predictive gain, that separates the quantizer families by downstream generation quality (rank-AUC 1.00) where reconstruction PSNR and marginal token entropy do not. On LFQ-1024, retuning D3PM and SE-D3PM (selected on a held-out validation split) moves them from default FID-192 0.44/0.41 to 0.09/0.10 at lower NFE, replicated across seeds; the continuous references were not given an equivalent sampler sweep. We report FID-192 as an internal ranking metric; it ranks consistently with standard FID-2048 (Spearman 0.80) and with a label-free classifier two-sample test (0.78). We interpret these results through a rate-distortion-modelability framing, where modelability is conditional on the generator, sampler, and inference budget. All experiments are at 64x64 on low-resolution medical-style images, unconditional, and evaluated with non-clinical FID-based metrics, and we scope every claim to that setting. Code: https://github.com/liamchalcroft/medtokenizers and https://github.com/liamchalcroft/medlatents.
Chinese Translation
潜在医学图像生成器通常将分词器视为固定的预处理。我们在一个受控的ChestMNIST研究中测试这种分离是否有效，研究在64x64的条件下交叉离散分词器、生成器家族和采样器设置，使用共享的潜在网格和连续潜在参考单元。在这个受控环境中，排名共同依赖于分词器、生成器和采样器：最佳量化器随着生成器的变化而变化，基于验证的采样器选择改变了表观生成器的排名。我们在三个种子上重新训练了词汇-1024交互块，结果表明该交互依然存在（9对量化器比较中有6个超过了三种子标准差），因此我们相应地扩展了单种子网格。仅凭重建PSNR并不是一个可靠的选择标准；我们引入了一种无生成器的统计量——邻居条件预测增益，它通过下游生成质量（rank-AUC 1.00）将量化器家族区分开来，而重建PSNR和边际令牌熵则无法做到。在LFQ-1024上，重新调优D3PM和SE-D3PM（在保留的验证分割上选择）使它们从默认的FID-192 0.44/0.41降至0.09/0.10，且在不同种子间复制；连续参考没有进行等效的采样器遍历。我们报告FID-192作为内部排名指标；它与标准FID-2048（Spearman 0.80）和无标签分类器的两样本测试（0.78）一致。我们通过率-失真-模型能力的框架来解读这些结果，其中模型能力依赖于生成器、采样器和推理预算。所有实验均在64x64的低分辨率医学风格图像上进行，无条件，并使用非临床的基于FID的指标进行评估，我们将每个声明的范围限定在该设置内。代码： https://github.com/liamchalcroft/medtokenizers 和 https://github.com/liamchalcroft/medlatents。
cs.CV / 36 / 2608.07726
Vision Meets WiFi: Physics-Grounded Estimation of Volumetric Mechanical Properties
视觉与WiFi相结合：基于物理的体积机械属性估计
Abstract
Estimating volumetric mechanical properties, including Young's modulus, Poisson's ratio, and density at each voxel, is intrinsically ambiguous from vision alone, as visually similar objects may have substantially different material compositions and physical behavior. Existing approaches predict these properties independently across voxels, overlooking the piecewise-constant material structure of real objects and producing noisy or inconsistent estimates for voxels that share the same material, while lacking an explicit mechanism to resolve visual ambiguity. We introduce ViWi (Vision Meets WiFi), an object-centric framework for volumetric mechanical-property estimation. ViWi represents each object using a compact set of material slots that aggregate evidence from voxels with a shared material identity and produce coherent slot-level property predictions. To complement visual appearance, ViWi incorporates a compact RF descriptor generated through WiFi-band electromagnetic simulation using permittivity and conductivity. The RF descriptor conditions the material slots with global composition cues that may be unavailable from images, while visual features preserve voxel-level spatial localization. Across volumetric mechanical-property and mass-estimation benchmarks, ViWi improves over the prior state of the art on four of six per-voxel metrics, while its vision-only variant improves all mass-estimation metrics. These results demonstrate that combining object-centric material structure with complementary RF evidence enables more accurate and physically coherent volumetric property estimation beyond what is possible from visual appearance alone.
Chinese Translation
仅凭视觉估计体积机械属性（包括杨氏模量、泊松比和每个体素的密度）本质上是模糊的，因为视觉上相似的物体可能具有截然不同的材料成分和物理行为。现有方法独立预测每个体素的这些属性，忽视了真实物体的分段常量材料结构，从而对共享相同材料的体素产生噪声或不一致的估计，同时缺乏解决视觉模糊的明确机制。我们提出了ViWi（视觉与WiFi相结合），一个以物体为中心的体积机械属性估计框架。ViWi使用一组紧凑的材料槽来表示每个物体，这些材料槽聚合来自具有共享材料身份的体素的证据，并产生一致的槽级属性预测。为了补充视觉外观，ViWi结合了通过WiFi频段电磁仿真生成的紧凑RF描述符，该描述符使用介电常数和导电率。RF描述符为材料槽提供了可能在图像中不可用的全局组成线索，而视觉特征则保留了体素级空间定位。在体积机械属性和质量估计基准测试中，ViWi在六个每体素指标中的四个上优于先前的最先进技术，而其仅基于视觉的变体在所有质量估计指标上都有所改善。这些结果表明，将以物体为中心的材料结构与互补的RF证据相结合，可以实现比仅凭视觉外观更准确和物理一致的体积属性估计。
cs.CV / 37 / 2608.07735
Ghost Features and Spooky Transfer Learning for Hypercomplex-Valued Neural Networks
超复数值神经网络中的幽灵特征与神秘迁移学习
Abstract
Hypercomplex numbers extend the concept of complex numbers by introducing additional imaginary components. Besides increasing dimensionality, operations on the imaginary parts provide algebraic and geometrical properties that can be beneficial for solving machine learning problems. In this paper, we show how to create hypercomplex-valued neural network layers where the real part corresponds to the output of a traditional real-valued layer. The additional imaginary parts of these hypercomplex-valued layers produce what we call ``ghost features,'' which contain enhanced information that is not present in the output of the real-valued layer. Moreover, ghost features can be effectively integrated into a trained neural network through a process we refer to as ``spooky transfer learning.'' This approach allows us to harness the richness of ghost features, leading to more efficient neural networks. The source code and Jupyter Notebook are available at https://github.com/mevalle/v-nets/.
Chinese Translation
超复数扩展了复数的概念，通过引入额外的虚部来增加维度。除了提高维度外，对虚部的操作提供了代数和几何特性，这些特性对于解决机器学习问题是有益的。本文展示了如何创建超复数值神经网络层，其中实部对应于传统实值层的输出。这些超复数值层的额外虚部产生了我们称之为“幽灵特征”的信息，这些特征包含了在实值层输出中不存在的增强信息。此外，幽灵特征可以通过我们称之为“神秘迁移学习”的过程有效地集成到训练好的神经网络中。这种方法使我们能够利用幽灵特征的丰富性，从而实现更高效的神经网络。源代码和Jupyter Notebook可在 https://github.com/mevalle/v-nets/ 获取。
cs.CV / 38 / 2608.07742
BRUCE: Benchmarking Robustness Under Corruption Escalation for Scientific Vision-Language Reasoning
BRUCE：针对科学视觉-语言推理的腐败升级下的鲁棒性基准测试
Abstract
Visual-language models (VLMs) frequently struggle with robustness issues in real-world situations due to low- or varying-quality input images. In this paper, we aim at analyzing VLMs' robustness by applying perturbations and distortions to the input images, such as blur or low contrast. Toward this goal, we propose BRUCE (Benchmarking Robustness Under Corruption Escalation, a multimodal reasoning fragility framework for scientific vision-language reasoning. State-of-the-art evaluation frameworks/studies primarily focus on clean-task accuracy and rarely analyze how reasoning stability degrades across robustness dimensions. Besides varying over a wide-range of input perturbations, BRUCE employs two novel metrics -- Robustness Corruption Index (RCI) and Traversal-RCI (T-RCI) -- to quantify how rapidly multimodal reasoning performance deteriorates in VLMs as visual corruption severity increases under progressive perturbation scaling. We evaluate BRUCE across chemistry and mathematical reasoning tasks for multiple datasets, while analyzing corruption-induced prediction failures in terms of four high-level reasoning domains: OCR-dependent reasoning, spatial reasoning, symbolic reasoning, and semantic failures, with each containing fine-grained corruption specific failure subtypes, thereby enabling an interpretable failure analysis.
Chinese Translation
视觉-语言模型（VLMs）在现实世界中常常面临鲁棒性问题，主要由于输入图像质量低或变化不定。本文旨在通过对输入图像施加扰动和失真（如模糊或低对比度）来分析VLMs的鲁棒性。为此，我们提出了BRUCE（腐败升级下的鲁棒性基准测试），这是一个用于科学视觉-语言推理的多模态推理脆弱性框架。现有的最先进评估框架/研究主要关注干净任务的准确性，鲜有分析推理稳定性在鲁棒性维度上的下降情况。除了在广泛的输入扰动范围内变化外，BRUCE还采用了两个新颖的指标——鲁棒性腐败指数（Robustness Corruption Index, RCI）和遍历-RCI（Traversal-RCI, T-RCI）——来量化随着视觉腐败严重性在渐进扰动缩放下如何迅速恶化多模态推理性能。我们在多个数据集上对化学和数学推理任务进行了BRUCE评估，同时从四个高层次推理领域分析了因腐败引起的预测失败：依赖OCR的推理、空间推理、符号推理和语义失败，每个领域包含细粒度的特定腐败失败子类型，从而实现可解释的失败分析。
cs.CV / 39 / 2608.07749
LoRSA: Toward Generalizable Parameter-Efficient Fine-Tuning for Biomedical Downstream Tasks
LoRSA：朝着可推广的参数高效微调用于生物医学下游任务
Abstract
Parameter-efficient fine-tuning enables the adaptation of vision foundation models to biomedical tasks under limited computational resources, but a single low-rank update can constrain all task-specific changes to one narrow parameter subspace. This restriction may prevent the model from simultaneously representing globally shared task structure and localized residual directions required for generalization to unseen imaging domains. We introduce LoRSA, a global--residual adaptation framework that jointly learns a dense low-rank component and a dynamically structured-sparse low-rank component. The dense component captures globally coordinated task adaptation, while the structured component provides complementary residual corrections whose support evolves during training. We characterize the representational capacity, approximation properties, rank structure, and singular-subspace complementarity of this decomposition. We evaluate LoRSA for four-class breast-density classification using DINOv3-Base, with VinDr-Mammo as the source domain and MammosighTR and RSNA as unseen external domains. LoRSA remains competitive on the internal validation set and achieves the best external macro-F1 on both target datasets, improving upon the strongest competing method by 2.15 percentage points on MammosighTR and 3.09 percentage points on RSNA. Weight-matrix analysis further shows that approximately $92\%$ of the energy of each adaptation component lies outside the bilateral singular subspace of the other, indicating that the two components learn largely complementary update directions. These results suggest that organizing adaptation capacity into distinct global and residual paths can improve the external-domain generalization of parameter-efficiently adapted biomedical vision models.
Chinese Translation
参数高效微调使得在有限计算资源下将视觉基础模型适应于生物医学任务成为可能，但单一的低秩更新可能将所有特定任务的变化限制在一个狭窄的参数子空间内。这一限制可能阻止模型同时表示全球共享的任务结构和针对未见影像领域所需的局部残差方向。我们提出了LoRSA，一个全球-残差适应框架，联合学习一个密集的低秩组件和一个动态结构稀疏的低秩组件。密集组件捕捉全球协调的任务适应，而结构组件提供互补的残差修正，其支持在训练过程中不断演变。我们对这种分解的表征能力、近似特性、秩结构和奇异子空间互补性进行了表征。我们在使用DINOv3-Base进行四类乳腺密度分类的实验中评估了LoRSA，以VinDr-Mammo作为源领域，MammosighTR和RSNA作为未见的外部领域。LoRSA在内部验证集上保持竞争力，并在两个目标数据集上实现了最佳的外部宏F1，相比最强的竞争方法在MammosighTR上提高了2.15个百分点，在RSNA上提高了3.09个百分点。权重矩阵分析进一步表明，每个适应组件的能量约92%位于另一个组件的双边奇异子空间之外，表明这两个组件学习了大体上互补的更新方向。这些结果表明，将适应能力组织成不同的全球和残差路径可以改善参数高效适应的生物医学视觉模型在外部领域的泛化能力。
cs.CV / 40 / 2608.07750
Multi-Task Consistency-based Detection of Adversarial Attacks
基于多任务一致性的对抗攻击检测
Abstract
Deep Neural Networks (DNNs) have found successful deployment in numerous vision perception systems. However, their susceptibility to adversarial attacks has prompted concerns regarding their practical applications, specifically in the context of autonomous driving. Existing defenses often suffer from cost inefficiency, rendering their deployment impractical for resource-constrained applications. In this work, we propose an efficient and effective adversarial attack detection scheme leveraging the multi-task perception within a complex vision system. Adversarial perturbations are detected by the inconsistencies between the inference outputs of multiple vision tasks, e.g., object detection and instance segmentation. To this end, we developed a consistency score metric to measure the inconsistency between vision tasks. Next, we designed an approach to select the best model pairs for detecting inconsistencies effectively. Finally, we evaluated our defense against PGD attacks across multiple vision models on the BDD100k validation dataset. The experimental results demonstrated that our defense achieved a ROC-AUC performance of 99.9% detection within the considered attacker model.
Chinese Translation
深度神经网络（DNNs）在众多视觉感知系统中得到了成功应用。然而，它们对对抗攻击的脆弱性引发了对其实际应用的担忧，特别是在自动驾驶的背景下。现有的防御措施往往存在成本低效的问题，使其在资源受限的应用中难以部署。在本研究中，我们提出了一种高效且有效的对抗攻击检测方案，利用复杂视觉系统中的多任务感知。通过检测多个视觉任务（例如物体检测和实例分割）推理输出之间的不一致性来识别对抗扰动。为此，我们开发了一种一致性得分指标，用于衡量视觉任务之间的不一致性。接下来，我们设计了一种方法来选择最佳模型对，以有效检测不一致性。最后，我们在 BDD100k 验证数据集上评估了我们的防御措施对 PGD 攻击的效果。实验结果表明，在考虑的攻击者模型下，我们的防御措施实现了 99.9% 的 ROC-AUC 检测性能。
cs.CV / 41 / 2608.07757
Rethinking 3D Segmentation from Individual LiDAR Scans: Incidence-Aware Sampling on the SIP Benchmark
从单个激光雷达扫描重新思考三维分割：在SIP基准上的考虑入射的采样
Abstract
3D scene understanding is increasingly important in construction, yet most methods are developed on curated datasets that do not fully reflect real site sensing conditions. In many workflows, individual LiDAR scans provide rapid local updates rather than complete scene representations, producing limited surface coverage, acquisition-driven density variation, and severe imbalance between dominant planar surfaces and sparse construction elements. Because large point clouds must be downsampled, sampling resolution and point allocation directly affect the balance between geometric detail and spatial context. This study evaluates these effects under a fixed per-fragment point budget and introduces an incidence-aware sampling strategy for individual LiDAR scans. The method maps points to a geometry-normalized manifold space for voxel-based selection while preserving original Euclidean coordinates for downstream learning. It requires only point coordinates and normals and no backbone modification. Using the Site in Pieces (SIP) benchmark, experiments with Point Transformer and PointNeXt show improved resolution-averaged segmentation performance, especially for non-planar elements and ladders, while reducing sensitivity to sampling resolution. The results show that acquisition-aware sampling can provide a more stable geometric representation and should be treated as an active component of individual-scan 3D segmentation rather than generic preprocessing.
Chinese Translation
三维场景理解在建筑领域日益重要，但大多数方法是在经过整理的数据集上开发的，这些数据集并未完全反映真实现场的感知条件。在许多工作流程中，单个激光雷达扫描提供快速的局部更新，而非完整的场景表示，导致表面覆盖有限、采集驱动的密度变化以及主导平面表面与稀疏建筑元素之间的严重不平衡。由于大型点云必须进行下采样，采样分辨率和点分配直接影响几何细节与空间上下文之间的平衡。本研究在固定的每片段点预算下评估这些影响，并引入了一种针对单个激光雷达扫描的考虑入射的采样策略。该方法将点映射到几何归一化的流形空间，以进行基于体素的选择，同时保留原始的欧几里得坐标以供后续学习。它仅需要点坐标和法线，而无需修改主干网络。使用“分段现场”（Site in Pieces，SIP）基准，使用Point Transformer和PointNeXt的实验显示出改进的分辨率平均分割性能，尤其是在非平面元素和梯子方面，同时降低了对采样分辨率的敏感性。结果表明，考虑采集的采样可以提供更稳定的几何表示，并应被视为单个扫描三维分割的一个主动组成部分，而非通用预处理。
cs.CV / 42 / 2608.07760
XClipGS: Exact Half-Space Clipping for Medical Volume Gaussian Splatting
XClipGS：医学体积高斯点云的精确半空间裁剪
Abstract
Gaussian-splatting proxies enable interactive rendering of volumetric medical scans, but a clipping plane exposes anatomy not constrained by external-view training and intersects primitives that conventional splatting can only keep or drop whole. We present XClipGS (eXact Clipping), which treats these as two separate problems: the render-time clip operator and supervision of the hidden interior. Under the local affine model used by EWA splatting, the ray integral of a half-space-restricted Gaussian factorizes exactly into its ordinary 2D footprint and a conditional Gaussian CDF whose argument is affine in pixel coordinates. The resulting closed-form per-pixel operator introduces no learned clipping parameters or auxiliary network and remains differentiable with respect to the primitive and plane. We use multi-distance reference views with varied clipping-plane axes and offsets to supervise the interior through the same operator. We also introduce a paired clipped/unclipped cut-face protocol with difference-referenced cut error (CDE) and culled-side leakage (Leak), because global image metrics dilute errors near the plane. On eight CT and MRI volumes with plane offsets not used for training, XClipGS attains the highest PSNR on every volume (33.56 versus 32.34 dB for ClipGS) while rendering at over 650 FPS, far above real time, versus 278 FPS. On voxel-axis cut-face views, it raises average band SSIM from 0.809 to 0.860 and leaks roughly 40 times less. Without retraining, it also achieves the best average across all four metrics on arbitrary-normal planes; on a fixed interior, it matches RaRa's face fidelity with about 16 times less leakage. Project page: https://gaozhongpai.github.io/XClipGS/
Chinese Translation
高斯点云代理使得体积医学扫描的交互式渲染成为可能，但裁剪平面暴露了未受外部视图训练约束的解剖结构，并与传统点云只能整块保留或丢弃的原始图形相交。我们提出了XClipGS（精确裁剪），将这些视为两个独立的问题：渲染时的裁剪操作和对隐藏内部的监督。在EWA点云使用的局部仿射模型下，半空间限制的高斯的射线积分精确地因式分解为其普通的二维足迹和一个条件高斯累积分布函数（CDF），其参数在像素坐标中是仿射的。结果得到的每像素闭式操作不引入学习的裁剪参数或辅助网络，并且在原始图形和平面方面保持可微性。我们使用具有不同裁剪平面轴和偏移的多距离参考视图，通过相同的操作对内部进行监督。我们还引入了一种配对的裁剪/未裁剪切面协议，使用差异参考切割误差（CDE）和被裁剪侧泄漏（Leak），因为全局图像度量会稀释平面附近的误差。在八个未用于训练的CT和MRI体积上，XClipGS在每个体积上都达到了最高的峰值信噪比（PSNR）（33.56对比ClipGS的32.34 dB），同时以超过650帧每秒（FPS）的速度渲染，远高于实时的278 FPS。在体素轴切面视图中，它将平均带状结构相似性指数（SSIM）从0.809提高到0.860，泄漏量减少约40倍。在不重新训练的情况下，它在任意法线平面上也达到了所有四个指标的最佳平均值；在固定内部情况下，它的面保真度与RaRa相匹配，但泄漏量减少约16倍。项目页面：https://gaozhongpai.github.io/XClipGS/
cs.CV / 43 / 2608.07767
DINO-3DRA: Leveraging 2D Foundation Model Semantics for 3D Cerebral Aneurysm Segmentation
DINO-3DRA：利用2D基础模型语义进行3D脑动脉瘤分割
Abstract
Accurate aneurysm segmentation in 3D rotational angiography (3DRA) is hindered by extreme class imbalance, morphological similarity to vessels, and absent large-scale 3D pretraining. 2D vision foundation models encode dense structural priors from 1.7 billion images, yet na\"ive slice-wise transfer fragments anatomical continuity and destabilises optimisation. We propose DINO-3DRA, a dual-path framework achieving effective cross-dimensional semantic transfer by injecting frozen DINOv3 features into a 3D U-Net backbone via Room-Lite spatial mixing and calibrated residual fusion. On multi-centre 3DRA data, DINO-3DRA achieves state-of-the-art aneurysm segmentation (Dice: 0.758; HD95: 2.75 mm; +13% over nnU-Net) with only 5.72M trainable parameters. Ablation studies confirm that gains arise from structured cross-dimensional transfer rather than loss design alone, with bridged foundation features improving anatomical continuity between aneurysms and parent vessels. Without fine-tuning on CADA and SHINY-ICARUS, DINO-3DRA eliminates all catastrophic failure cases observed in baseline architectures, demonstrating robust generalisation across heterogeneous imaging protocols.
Chinese Translation
在3D旋转血管造影（3DRA）中，准确的动脉瘤分割受到极端类别不平衡、与血管的形态相似性以及缺乏大规模3D预训练的影响。2D视觉基础模型从17亿张图像中编码了密集的结构先验，但简单的切片级转移会破坏解剖连续性并使优化不稳定。我们提出了DINO-3DRA，这是一种双路径框架，通过将冻结的DINOv3特征注入3D U-Net主干，利用Room-Lite空间混合和校准残差融合，实现了有效的跨维度语义转移。在多中心3DRA数据上，DINO-3DRA实现了最先进的动脉瘤分割（Dice: 0.758；HD95: 2.75 mm；比nnU-Net提高13%），仅需5.72M可训练参数。消融研究确认，性能提升源于结构化的跨维度转移，而不仅仅是损失设计，连接的基础特征改善了动脉瘤与母血管之间的解剖连续性。在未对CADA和SHINY-ICARUS进行微调的情况下，DINO-3DRA消除了基线架构中观察到的所有灾难性失败案例，展示了在异构成像协议下的强大泛化能力。
cs.CV / 44 / 2608.07835
SeqLoc: Beyond the Single Frame for Cross-View Geo-Localization in Feature-Sparse Scenes
SeqLoc：超越单帧的特征稀疏场景跨视角地理定位
Abstract
Cross-View Geo-Localization (CVGL) with OpenStreetMap (OSM) performs well in structure-rich urban environments but collapses in feature-sparse scenes such as rural roads. To study this failure mode, in this work, we introduce CV-FSS, a benchmark that pairs sequential panoramas from five rural regions with aligned OSM maps, on which single-frame methods degrade drastically. We then propose SeqLoc, an online test-time sequence aggregation mechanism that recursively maintains a log-belief volume with three key components: (1) Entropy-Tempered Uncertainty (ETU) tempers each incoming pose likelihood volume by its normalized entropy; (2) Map-Guided Relocalization (MGR) mixes a map-shaped recovery distribution into the belief so that a suppressed true pose can recover; (3) Peak-Anchored Smoothing (PAS) derives the final pose at sub-grid precision. Extensive experiments on CV-FSS and CV-RHO demonstrate that SeqLoc outperforms single-frame localization by a large margin, improving both position and orientation recall by over 50%. The benchmark and source code will be made publicly available.
Chinese Translation
基于开放街图（OpenStreetMap, OSM）的跨视角地理定位（Cross-View Geo-Localization, CVGL）在结构丰富的城市环境中表现良好，但在特征稀疏的场景（如乡村道路）中则表现不佳。为了研究这一失败模式，本文引入了CV-FSS，一个基准数据集，该数据集将来自五个乡村地区的连续全景图与对齐的OSM地图配对，在这些地图上，单帧方法的性能显著下降。我们提出了SeqLoc，一种在线测试时序列聚合机制，该机制递归地维护一个对数置信体积，包含三个关键组件：（1）熵调温不确定性（Entropy-Tempered Uncertainty, ETU）通过其归一化熵来调节每个输入姿态的可能性体积；（2）地图引导重定位（Map-Guided Relocalization, MGR）将地图形状的恢复分布混入置信中，以便被抑制的真实姿态能够恢复；（3）峰值锚定平滑（Peak-Anchored Smoothing, PAS）以亚网格精度推导最终姿态。在CV-FSS和CV-RHO上的大量实验表明，SeqLoc在定位精度上大幅超越单帧定位，位置和方向的召回率均提高超过50%。该基准数据集和源代码将公开发布。
cs.CV / 45 / 2608.07848
IRPol-Fuse: Energy-structure coordination for infrared polarization fusion under low visibility
IRPol-Fuse：低能见度下红外偏振融合的能量-结构协调
Abstract
Robust perception under low-visibility conditions requires fused imagery that jointly preserves infrared thermal saliency and polarization-derived structural details. However, existing infrared-polarization image fusion (IPIF) methods often overemphasize dominant infrared responses, causing weak yet informative polarization textures in dark regions to be suppressed. To address this issue, we propose IRPol-Fuse, an energy-structure coordinated IPIF framework for challenging low-visibility scenarios. The proposed framework contains three key modules: Polarization Attention Fusion for adaptive infrared-polarization allocation, Infrared Highlight Injector for highlight-guided infrared preservation, and Polarization Texture Injector for polarization texture restoration and fine-detail recovery. We further construct LI-PI, a dedicated infrared-polarization evaluation dataset for low-visibility and visually concealed scenes. Experiments on LI-PI and the public LDDRS dataset demonstrate that IRPol-Fuse achieves favorable performance in thermal target preservation, structural detail recovery, and visual naturalness. Region-aware evaluation and downstream object detection further verify that the proposed energy-structure coordination strategy effectively preserves both infrared target saliency and polarization-derived structural information. Code is available at https://github.com/1hzf/IRPolar-Fuse .
Chinese Translation
在低能见度条件下，稳健的感知需要融合图像，既能保留红外热显著性，又能保留由偏振衍生的结构细节。然而，现有的红外偏振图像融合（IPIF）方法往往过于强调主导的红外响应，导致在暗区中微弱但信息丰富的偏振纹理被抑制。为了解决这一问题，我们提出了IRPol-Fuse，这是一种针对挑战性低能见度场景的能量-结构协调IPIF框架。该框架包含三个关键模块：偏振注意力融合（Polarization Attention Fusion）用于自适应红外-偏振分配，红外高光注入器（Infrared Highlight Injector）用于高光引导的红外保留，以及偏振纹理注入器（Polarization Texture Injector）用于偏振纹理恢复和细节重建。我们进一步构建了LI-PI，一个专门针对低能见度和视觉隐蔽场景的红外偏振评估数据集。对LI-PI和公共LDDRS数据集的实验表明，IRPol-Fuse在热目标保留、结构细节恢复和视觉自然性方面表现良好。区域感知评估和下游目标检测进一步验证了所提出的能量-结构协调策略有效保留了红外目标显著性和偏振衍生的结构信息。代码可在 https://github.com/1hzf/IRPolar-Fuse 获取。
cs.CV / 46 / 2608.07857
Distilling CT Foundation Models into Editable Concept Bottlenecks for Lung Nodule Malignancy Prediction
将CT基础模型提炼为可编辑的概念瓶颈以预测肺结节恶性程度
Abstract
Foundation models provide transferable CT representations, but predictions based directly on these embeddings are difficult to interpret. We developed concept bottleneck models that map two frozen CT foundation-model representations to eight radiologist-defined pulmonary-nodule attributes and predict malignancy from the estimated concepts and nodule size. The models included CT-FM, a whole-CT self-supervised encoder using a 96^3-voxel nodule-centered patch, and FMCIB, a nodule-focused contrastive encoder using a 50-mm crop. Eight ridge-regression concept heads were trained on 2,610 LIDC-IDRI nodules. Malignancy models were trained on LUNA25 and evaluated on a held-out internal test set and the external DLCS cohort. Concept fidelity was assessed using five-fold cross-validated R^2, and malignancy discrimination was assessed using AUROC with 95% confidence intervals estimated by patient-grouped bootstrap resampling. Concept fidelity was modest but higher for FMCIB than CT-FM for subtlety (R2, 0.24 vs. 0.11), spiculation (0.17 vs. 0.08), texture (0.17 vs. 0.07), and lobulation (0.15 vs. 0.05). Internally, the CT-FM and FMCIB concept+size models achieved AUROCs of 0.86 (95% CI, 0.80-0.92) and 0.86 (0.79-0.92), respectively. Externally, AUROCs were 0.72 (0.68-0.75) and 0.73 (0.70-0.76), compared with 0.73 for nodule size alone and 0.60 and 0.67 for the corresponding embedding only probes. Additive predictions could be decomposed into feature-level contributions and modified through controlled concept interventions. Concept bottlenecks provided transparent malignancy predictions with discrimination similar to nodule size alone, while differences in concept fidelity suggest that concept recovery depends on the underlying foundation-model representation.
Chinese Translation
基础模型提供了可转移的CT表征，但基于这些嵌入的预测难以解释。我们开发了概念瓶颈模型，将两个冻结的CT基础模型表征映射到八个放射科医生定义的肺结节属性，并根据估计的概念和结节大小预测恶性程度。模型包括CT-FM，一个使用96^3体素结节中心补丁的全CT自监督编码器，以及FMCIB，一个使用50毫米裁剪的结节聚焦对比编码器。八个岭回归概念头在2610个LIDC-IDRI结节上进行了训练。恶性程度模型在LUNA25上训练，并在保留的内部测试集和外部DLCS队列上进行了评估。使用五折交叉验证的R^2评估概念的保真度，使用AUROC评估恶性程度的区分能力，95%的置信区间通过按患者分组的自助重采样估计。概念保真度适中，但FMCIB在细微度（R2, 0.24 vs. 0.11）、刺状（0.17 vs. 0.08）、纹理（0.17 vs. 0.07）和分叶（0.15 vs. 0.05）方面高于CT-FM。在内部，CT-FM和FMCIB概念+大小模型的AUROC分别达到了0.86（95% CI, 0.80-0.92）和0.86（0.79-0.92）。在外部，AUROC为0.72（0.68-0.75）和0.73（0.70-0.76），而仅结节大小的AUROC为0.73，相关嵌入仅探针的AUROC为0.60和0.67。加性预测可以分解为特征级贡献，并通过受控的概念干预进行修改。概念瓶颈提供了透明的恶性程度预测，其区分能力与仅结节大小相似，而概念保真度的差异表明概念恢复依赖于基础模型的表征。
cs.CV / 47 / 2608.07861
How Much Does It Cost to Answer My Question? Benchmarking Cloud VLM-based VQA Systems
回答我的问题需要多少成本？基于云的视觉语言模型（VLM）视觉问答（VQA）系统基准测试
Abstract
Vision-language models (VLMs) are becoming a practical backend for mobile visual question answering (VQA) systems, enabling smartphones and smart glasses to answer users' questions about the physical world. Since modern VLMs remain difficult to run on mobile and edge devices, VQA systems increasingly offload inference to cloud-based VLMs. This gives mobile devices access to stronger computation, but it also makes visual input preparation a key system variable: how the image is prepared before offloading affects not only answer quality but also payload size, token cost, and system latency. Proprietary APIs expose little control over model internals or serving behavior, leaving client-side preprocessing as the main practical optimization space for downstream developers. Many such techniques have been proposed for visual offloading, yet their cost-quality impact on commercial cloud VLMs has never been studied. To fill this gap, we present VQABench, the first systematic benchmark that treats client-side input preprocessing as a controlled variable for cloud-VLM-based VQA. We evaluate 12 preprocessing techniques across three VQA datasets and four commercial VLMs from three providers, totaling 95,168 API calls. Our results show that preprocessing is not universally beneficial: its effectiveness depends on the target model, API paradigm, provider token-accounting rule, and task formulation. A poorly selected preprocessing strategy can increase deployment cost or latency while degrading answer accuracy. Overall, our benchmark clarifies when preprocessing helps, when it fails, and why, providing insights to guide future research and real-world deployment of VQA systems.
Chinese Translation
视觉语言模型（VLM）正成为移动视觉问答（VQA）系统的实用后端，使智能手机和智能眼镜能够回答用户关于物理世界的问题。由于现代VLM在移动和边缘设备上运行仍然困难，VQA系统越来越多地将推理任务转移到基于云的VLM上。这使得移动设备能够访问更强大的计算能力，但也使得视觉输入准备成为一个关键的系统变量：在转移之前如何准备图像不仅影响答案质量，还影响有效载荷大小、令牌成本和系统延迟。专有API对模型内部或服务行为的控制有限，导致客户端预处理成为下游开发者主要的实际优化空间。虽然已经提出了许多视觉转移的技术，但它们对商业云VLM的成本-质量影响尚未被研究。为填补这一空白，我们提出了VQABench，这是第一个将客户端输入预处理视为云VLM基础的VQA的受控变量的系统基准测试。我们在三个VQA数据集和来自三个提供商的四个商业VLM上评估了12种预处理技术，总计95,168次API调用。我们的结果表明，预处理并不总是有利的：其有效性取决于目标模型、API范式、提供商的令牌计费规则和任务表述。选择不当的预处理策略可能会增加部署成本或延迟，同时降低答案准确性。总体而言，我们的基准测试阐明了预处理何时有助、何时失效以及原因，为未来的研究和VQA系统的实际部署提供了指导性见解。
cs.CV / 48 / 2608.07863
LHSDet: High-Resolution AI-Generated Image Detection via Visual Question Answering
LHSDet：通过视觉问答实现高分辨率AI生成图像检测
Abstract
Driven by advances in diffusion models and autoregressive models, the fidelity and resolution of AI-generated images now rival those of real images. However, existing AI-generated image detection methods often downsample the images, inevitably overlooking critical low-level texture details in high-resolution AI-generated images, therefore limiting their detection performance. In addition, the ceaseless emergence of unknown generative models makes large-scale pre-training datasets inaccessible. To address these challenges, we propose a novel high-resolution AI-generated image detector, termed LHSDet. Specifically, we formulate the AI-generated image detection task as a Visual Question Answering problem, leveraging a fine-tuned vision-language framework to fully exploit the complementary information between visual and textual modalities. Recognizing that the default visual encoder of existing vision-language models is not tailored for AI-generated image detection, we redesign a visual encoder to better capture both the low-level and high-level artifacts inherent in AI-generated images. Furthermore, we incorporate a semantic-level textual branch to enable multi-modal feature fusion and detection. Consequently, LHSDet employs a triple-branch architecture to extract complementary multi-modal features: a low-level visual branch that aggregates non-overlapping patches for local texture cues, a high-level visual branch based on SigLIP2 for global perception feature extraction, and a semantic-level textual branch that generates captions using BLIP-2. Extensive experimental results demonstrate that LHSDet achieves high detection accuracy and robust performance across diverse generative models, including both diffusion and autoregressive models.
Chinese Translation
随着扩散模型和自回归模型的进步，AI生成图像的逼真度和分辨率现已与真实图像相媲美。然而，现有的AI生成图像检测方法通常会对图像进行下采样，必然会忽视高分辨率AI生成图像中的关键低级纹理细节，从而限制了其检测性能。此外，未知生成模型的不断涌现使得大规模预训练数据集变得不可获取。为了解决这些挑战，我们提出了一种新颖的高分辨率AI生成图像检测器，称为LHSDet。具体而言，我们将AI生成图像检测任务表述为视觉问答问题，利用经过微调的视觉-语言框架充分挖掘视觉和文本模态之间的互补信息。考虑到现有视觉-语言模型的默认视觉编码器并未针对AI生成图像检测进行优化，我们重新设计了一个视觉编码器，以更好地捕捉AI生成图像中固有的低级和高级伪影。此外，我们还引入了一个语义级文本分支，以实现多模态特征融合和检测。因此，LHSDet采用三分支架构来提取互补的多模态特征：一个低级视觉分支聚合不重叠的图块以获取局部纹理线索，一个基于SigLIP2的高级视觉分支用于全局感知特征提取，以及一个使用BLIP-2生成标题的语义级文本分支。大量实验结果表明，LHSDet在包括扩散模型和自回归模型在内的多种生成模型上实现了高检测准确率和稳健性能。
cs.CV / 49 / 2608.07864
UniScale: Arbitrary-Scale Industrial Anomaly Generation
UniScale：任意尺度工业异常生成
Abstract
Industrial anomaly inspection faces a major challenge due to the lack of real-world anomaly samples. While generative models are used to create anomaly data, existing methods still struggle when handling small-scale anomalies.This failure occurs because extreme downsampling in diffusion models causes the information of small anomalies to be lost in the latent space.To address this, we introduce UniScale, a unified training and inference framework for high-fidelity industrial anomaly generation across arbitrary scales.During training, we introduce an Error-Suppressed Multi-Scale Training (EMT) strategy, which enables the model to learn the rich location-aware textures of anomalies, while suppressing upsampling-induced interpolation errors in texture acquisition, ensuring the model is capable of learning small-scale anomalies, while remaining effective for regular scale anomalies.For inference, we propose Generation-then-Fusion Denoising. It decouples anomaly generation from background integration, preventing small anomalies from being overwhelmed.Extensive experiments demonstrate that our method outperforms state-of-the-art competitors in both anomaly generation quality and downstream detection performance.It achieves a relative IS(a) improvement of 45.86\% (from 1.81 to 2.64) on VisA and 37.70\% (from 1.22 to 1.68) on MVTec AD 2, while also improving the downstream pixel-level IoU by 4.22\% on VisA and AUROC by 6.55\% on MVTec AD 2. Code is available at https://github.com/HUST-SLOW/UniScale.
Chinese Translation
工业异常检测面临着由于缺乏真实世界异常样本而带来的重大挑战。尽管生成模型被用于创建异常数据，但现有方法在处理小尺度异常时仍然存在困难。这种失败的原因在于，扩散模型中的极端下采样导致小异常的信息在潜在空间中丢失。为了解决这个问题，我们提出了UniScale，一个统一的高保真工业异常生成的训练与推理框架，适用于任意尺度。在训练过程中，我们引入了一种错误抑制多尺度训练（Error-Suppressed Multi-Scale Training, EMT）策略，使模型能够学习异常的丰富位置感知纹理，同时抑制因上采样引起的纹理获取插值误差，确保模型能够学习小尺度异常，同时对常规尺度异常保持有效。在推理阶段，我们提出了生成-再融合去噪（Generation-then-Fusion Denoising）方法。该方法将异常生成与背景整合解耦，防止小异常被淹没。大量实验表明，我们的方法在异常生成质量和下游检测性能上均优于最先进的竞争对手。在VisA上相对提高了45.86 ext{%}（从1.81提高到2.64），在MVTec AD 2上提高了37.70 ext{%}（从1.22提高到1.68），同时在VisA上提高了4.22 ext{%}的下游像素级IoU，在MVTec AD 2上提高了6.55 ext{%}的AUROC。代码可在https://github.com/HUST-SLOW/UniScale获取。
cs.CV / 50 / 2608.07886
Vision-Language Grounding as Bidirectional Concept Correspondence
视觉-语言对接作为双向概念对应
Abstract
Vision-language grounding connects language to visual content, yet most existing formulations reduce grounding to a unidirectional localization problem: given a prespecified text phrase or category name, identify the corresponding image region. This setup assumes that the relevant linguistic unit is already known, overlooking a more basic challenge in grounded communication: determining which parts of the text are visually referential and how they correspond to entities in the image. We formulate grounding as $\textit{bidirectional concept correspondence}$ over an image-text pair. Given an image and its paired text, the goal is to recover all correspondences between visually referential text spans and instance-level image segments, without assuming that the relevant text spans are provided. This formulation unifies common grounding tasks, including phrase grounding, referring expression grounding, and open-vocabulary detection, by treating text segmentation, image segmentation, and cross-modal alignment as a single correspondence prediction problem. To address this task, we introduce $\textbf{ConCor-1}$, a grounding model built on top of a pretrained vision-language model. It uses learnable $\textit{bridge tokens}$ to represent candidate image-text correspondences and predicts, for each token, a text mask, an image mask, and a correspondence presence score. To train and evaluate this task, we convert diverse grounding and segmentation datasets into a unified correspondence format. Experiments show that $\textbf{ConCor-1}$ consistently outperforms baselines, improving correspondence F1 by 48% on the long-caption dataset and by 29% on zero-shot LVIS, where the large category list serves as the text input.
Chinese Translation
视觉-语言对接将语言与视觉内容连接起来，但大多数现有的表述将对接简化为单向定位问题：给定一个预先指定的文本短语或类别名称，识别相应的图像区域。这一设置假设相关的语言单位已经确定，忽视了在对接交流中一个更基本的挑战：确定文本的哪些部分是视觉指称的，以及它们如何与图像中的实体对应。我们将对接表述为图像-文本对上的$ extit{双向概念对应}$。给定一幅图像及其配对文本，目标是恢复视觉指称文本范围与实例级图像片段之间的所有对应关系，而不假设相关的文本范围是已提供的。这一表述统一了常见的对接任务，包括短语对接、指称表达对接和开放词汇检测，通过将文本分割、图像分割和跨模态对齐视为一个单一的对应预测问题。为了解决这一任务，我们引入了$ extbf{ConCor-1}$，这是一个基于预训练视觉-语言模型构建的对接模型。它使用可学习的$ extit{桥接标记}$来表示候选的图像-文本对应，并为每个标记预测一个文本掩码、一个图像掩码和一个对应存在分数。为了训练和评估这一任务，我们将多样的对接和分割数据集转换为统一的对应格式。实验表明，$ extbf{ConCor-1}$在基准测试中始终表现优于基线，在长标题数据集上提高了48%的对应F1，在零样本LVIS上提高了29%，其中大型类别列表作为文本输入。
cs.CV / 51 / 2608.07904
DeCo: Zero-Shot Industrial Anomaly Generation through Decoupling and Recoupling
DeCo：通过解耦和重耦实现零样本工业异常生成
Abstract
Industrial anomaly inspection is severely hindered by the scarcity of real anomalous data.Zero-shot industrial anomaly generation addresses this by generating anomalies on specific products without requiring any of their real anomalous images. However, existing methods suffer from two critical limitations, i.e., inaccurate anomaly information acquisition and uncontrolled anomaly-product fusion.To overcome these challenges, we propose DeCo, which decouples the anomaly structure from its source product, and explicitly recouples it with the normal textures of the target product. During anomaly information acquisition,Dual-Routing Flow (DR-Flow) binds the texture-invariant anomaly structure to an abnormal token, while a parallel constraint, Product-Invariant Flow (PI-Flow), prevents the abnormal token from binding the source product. During anomaly-product fusion, we propose a hybrid injection to recouple the acquired anomaly structure with the target product, and Product Compatibility Correction (PCC) to compensate for the incompatibility between the acquired anomaly structure and the product. Extensive experiments demonstrate that DeCo establishes a new state-of-the-art. Training downstream detection models on our generated data yields massive pixel AP improvements of 5.1\% on MVTec AD and 8.2\% on VisA. Code is available at https://github.com/HUST-SLOW/DeCo.
Chinese Translation
工业异常检测受到真实异常数据稀缺的严重制约。零样本工业异常生成通过在特定产品上生成异常，解决了这一问题，而无需任何真实异常图像。然而，现有方法存在两个关键限制，即异常信息获取不准确和异常与产品融合失控。为了解决这些挑战，我们提出了DeCo，它将异常结构与其源产品解耦，并将其与目标产品的正常纹理显式重耦。在异常信息获取过程中，双路流（Dual-Routing Flow, DR-Flow）将纹理不变的异常结构绑定到异常标记，同时并行约束产品不变流（Product-Invariant Flow, PI-Flow）防止异常标记与源产品绑定。在异常与产品融合过程中，我们提出了一种混合注入方法，将获取的异常结构与目标产品重耦，并引入产品兼容性校正（Product Compatibility Correction, PCC）以补偿获取的异常结构与产品之间的不兼容性。大量实验表明，DeCo建立了新的最先进水平。在我们的生成数据上训练下游检测模型，在MVTec AD上实现了5.1%的像素平均精度提升，在VisA上实现了8.2%的提升。代码可在 https://github.com/HUST-SLOW/DeCo 获取。
cs.CV / 52 / 2608.07916
SegDem: Segmentation helps Demosaicing
SegDem：分割助力去马赛克
Abstract
Image demosaicing reconstructs a full-color image from incomplete color measurements produced by a sensor covered with a color filter array (CFA). Most existing methods formulate demosaicing as pixel-level reconstruction and mainly rely on local textures, cross-channel correlations, and low-level image statistics. Our core insight is that reconstruction and visual understanding can be viewed as complementary views of shared scene structure: both are grounded in the same underlying physical world, and therefore the structural and physical information inferred from an image should remain consistent across the two tasks. We instantiate this idea with instance segmentation and propose \emph{SegDem}, a cross-task decoder representation transfer framework for demosaicing. SegDem first learns region- and boundary-aware representations through instance-aware structural pretraining and then transfers the decoder to RAW-conditioned reconstruction. Segmentation- and demosaicing-conditioned features are further anchored to a shared frozen DINOv2 representation space to preserve structural organization across tasks. We instantiate SegDem with convolutional, Transformer-based, and state-space backbones for unified Single- and Quad-Bayer demosaicing. Extensive experiments on synthetic, external, and challenging datasets demonstrate consistent improvements across different architectures and CFA layouts.
Chinese Translation
图像去马赛克是从传感器通过彩色滤光片阵列（CFA）产生的不完整颜色测量中重建全彩色图像的过程。现有的大多数方法将去马赛克视为像素级重建，主要依赖于局部纹理、跨通道相关性和低级图像统计特征。我们的核心见解是，重建和视觉理解可以视为共享场景结构的互补视角：两者都基于相同的物理世界，因此从图像推断出的结构和物理信息在这两项任务中应保持一致。我们通过实例分割来具体化这一思想，并提出了 extit{SegDem}，一个用于去马赛克的跨任务解码器表示迁移框架。SegDem首先通过实例感知的结构预训练学习区域和边界感知的表示，然后将解码器迁移到RAW条件下的重建。分割和去马赛克条件下的特征进一步锚定在共享的冻结DINOv2表示空间中，以保持任务间的结构组织。我们使用卷积、基于Transformer的和状态空间的骨干网络实例化SegDem，以实现统一的单通道和四通道Bayer去马赛克。在合成、外部和具有挑战性的数据集上的大量实验表明，不同架构和CFA布局下的性能均有一致提升。
cs.CV / 53 / 2608.07920
Forged Peer Judgments Mislead Multimodal LLM Judge Panels: Source-Blind Anchoring and Panel-Consensus Verification
伪造的同行评审判断误导多模态大语言模型评审小组：源盲锚定与小组共识验证
Abstract
Multimodal LLM judge panels can cross-reference peers, but a quoted peer judgment may itself be untrusted. We expose source-blind anchoring as a text-level attack surface in vision-language model (VLM) panels. Quoting independent visual judgments creates large anchoring gaps (19--26 percentage points) under both self and peer framing. A matched-content, label-only control changes the broken rate by only $-0.17$pp (95\% CI $[-0.68,0.35]$), showing that the self/peer label itself does not explain the effect. Under our tested construction, deliberately generated, concise wrong quotes overturn originally-correct verdicts 1.5--2.7$\times$ more often than naturally occurring wrong peer statements, with bootstrap 95\% CIs excluding parity across two datasets and seven VLM judges. Because the two statement populations differ in selection and form, this ratio measures differential damage under the tested attack rather than a provenance-only causal effect. We then introduce panel-consensus verification, which cross-checks a quote against independently collected blind votes. It blocks 84.9\% of fabricated attacks, cuts their net harm by 97.5\%, and preserves the positive but statistically inconclusive point estimate for genuine peer information under leave-one-out re-verification. These results identify a low-cost attack surface and a concrete defense for safer multimodal collaborative evaluation.
Chinese Translation
多模态大语言模型评审小组可以交叉引用同行，但引用的同行判断本身可能不可信。我们揭示了源盲锚定作为视觉-语言模型（VLM）小组中的文本级攻击面。引用独立的视觉判断在自我框架和同行框架下都产生了较大的锚定差距（19-26个百分点）。匹配内容、仅标签的对照组使得错误率的变化仅为$-0.17$pp（95 ext{% CI} $[-0.68,0.35]$），表明自我/同行标签本身并不能解释这一效应。在我们测试的构造下，故意生成的简洁错误引用推翻原本正确的裁决的频率是自然发生的错误同行陈述的1.5-2.7倍，且引导法95 ext{% CI}排除了在两个数据集和七个VLM评审者之间的平等性。由于这两种陈述群体在选择和形式上存在差异，这一比例衡量的是在测试攻击下的差异性损害，而非仅仅是来源因果效应。随后，我们引入了小组共识验证，它通过独立收集的盲投票对引用进行交叉检查。该方法阻止了84.9 ext{%}的伪造攻击，减少了97.5 ext{%}的净损害，并在逐一重新验证下保留了对真实同行信息的正面但统计上不确定的点估计。这些结果识别出了一种低成本的攻击面和一种具体的防御措施，以实现更安全的多模态协作评估。
cs.CV / 54 / 2608.07923
SCoPE: Training-Free Audio-Visual Event Perception via Sparse Cross-Modal Prior Exchange
SCoPE：通过稀疏跨模态先验交换实现无训练音视频事件感知
Abstract
Audio-visual event perception (AVEP) determines which events occur in a video, when they occur, and whether they are audible, visible, or both. Training-free methods query new event vocabularies by matching frozen audio and visual features with text-encoded event names. However, related labels share evidence. An incorrect label can then score at least as high as a correct one. We call this a false co-activation (FCA). No scalar cutoff can reject the incorrect label while keeping every correct one. Class-specific thresholds may prevent that label from becoming a final prediction, but the FCA remains in the underlying score vector. We introduce SCoPE, a training-free framework in which all queried labels compete for shared evidence and each modality guides event selection in the other. We derive an exact condition for when this competition removes an FCA in a two-label fit. With identical frozen CLIP+CLAP backbones on LLP, SCoPE improves Type@seg by 7.45 points and Event@seg by 5.04 points compared with the reported AV$^2$A values. The same fixed configuration transfers unchanged to OV-AVEBench and VGGSound-AVEL100k.
Chinese Translation
音视频事件感知（AVEP）用于确定视频中发生了哪些事件、何时发生以及这些事件是否可听、可见或两者兼具。无训练方法通过将冻结的音频和视觉特征与文本编码的事件名称进行匹配来查询新的事件词汇。然而，相关标签共享证据。错误的标签可能会获得与正确标签相同甚至更高的分数。我们称之为虚假共激活（FCA）。没有标量阈值能够在保留每个正确标签的同时拒绝错误标签。特定类别的阈值可能会阻止该标签成为最终预测，但FCA仍然存在于基础得分向量中。我们提出了SCoPE，这是一个无训练框架，其中所有查询的标签竞争共享证据，每种模态在另一种模态中引导事件选择。我们推导出在双标签拟合中何时这种竞争能够消除FCA的确切条件。在LLP上使用相同的冻结CLIP+CLAP骨干网络，SCoPE相比于报告的AV$^2$A值提高了Type@seg 7.45分和Event@seg 5.04分。相同的固定配置在OV-AVEBench和VGGSound-AVEL100k上保持不变。
cs.CV / 55 / 2608.07932
SportsGrounder: Proposal-Aided Interleaved Grounding for Dense Sports Video Reasoning
SportsGrounder：基于提案辅助的交错基础定位用于密集体育视频推理
Abstract
Sports video analysis is crucial for athletic analytics and broadcasting enhancement. Dense sports video reasoning, however, demands a fine-grained understanding of numerous small-scale, highly interactive, and visually homogeneous entities (e.g., players sharing identical uniforms, the ball) across long temporal contexts. Current Large Multimodal Models (LMMs) inherently struggle with such dense visual complexities. Due to the lack of fine-grained visual details, these models often over-rely on textual priors to guess answers, especially when distinguishing visually similar actions and players. To address this, we propose \textbf{SportsGrounder}, a framework that leverages an open-vocabulary visual expert to aid interleaved grounding specifically for dense sports video reasoning. To achieve precise spatial localization, we extract domain-guided object proposals and introduce an Interleaved Grounding Fusion (IGF) mechanism. The IGF frame-by-frame integrates explicit bounding box coordinates and implicit visual semantics with global grid features. This design preserves strict temporal alignment and prevents sequence length explosion. Furthermore, we design an Action-Aware Supervision (AAS) module that directly regularizes the model's hidden states, forcing the network to learn accurate motion representations rather than relying on language bias. Optimized with Mixed Preference Optimization (MPO) to better distinguish deceptive distractors, our extensive experiments on newly curated dense sports VQA datasets (derived from SoccerNet and FineSports) demonstrate that SportsGrounder significantly improves fine-grained reasoning and achieves state-of-the-art accuracy.
Chinese Translation
体育视频分析对于运动员分析和广播增强至关重要。然而，密集体育视频推理需要对众多小规模、高度互动和视觉同质的实体（例如，穿着相同制服的球员、足球）在长时间上下文中进行细致的理解。目前的大型多模态模型（LMMs）在处理这种密集视觉复杂性时固有地面临挑战。由于缺乏细致的视觉细节，这些模型往往过度依赖文本先验来猜测答案，特别是在区分视觉相似的动作和球员时。为了解决这一问题，我们提出了 extbf{SportsGrounder}，一个利用开放词汇视觉专家来辅助交错基础定位的框架，专门针对密集体育视频推理。为了实现精确的空间定位，我们提取了领域引导的对象提案，并引入了交错基础融合（Interleaved Grounding Fusion, IGF）机制。IGF逐帧整合显式边界框坐标和隐式视觉语义与全局网格特征。该设计保持严格的时间对齐，防止序列长度爆炸。此外，我们设计了一个动作感知监督（Action-Aware Supervision, AAS）模块，直接对模型的隐藏状态进行正则化，迫使网络学习准确的运动表示，而不是依赖语言偏见。通过混合偏好优化（Mixed Preference Optimization, MPO）进行优化，以更好地区分具有欺骗性的干扰项，我们在新创建的密集体育视频问答数据集（源自SoccerNet和FineSports）上进行了广泛实验，结果表明SportsGrounder显著提高了细粒度推理能力，并达到了最先进的准确性。
cs.CV / 56 / 2608.07937
FlexSplat: Flexible Feed-Forward 3D Gaussian Splatting without Point Cloud Correspondence
FlexSplat：无点云对应的灵活前馈3D高斯溅射
Abstract
We present FlexSplat, a feed-forward framework for novel view synthesis (NVS) from uncalibrated, object-centric multi-view image collections. A recent line of query-based methods reconstructs a compact set of 3D Gaussians by treating them as transformer queries that are refined with multi-view deformable attention; these methods, however, assume that camera poses are given. FlexSplat removes this assumption: a geometry transformer is trained jointly with the Gaussian decoder to predict per-image camera parameters and depth, which in turn ground a depth-guided Gaussian parameterization and a multi-view deformable cross-attention that aggregates evidence across all input views into a single, view-consistent set of primitives. An uncertainty-weighted depth-consistency objective lets the jointly trained geometry adapt to the reconstruction task, while the cross-view consensus formed during decoding absorbs the residual error of the estimated cameras and depth. The representation uses a compact Gaussian budget that is decoupled from the input resolution - unlike pixel-aligned methods, the primitive count does not grow with the image grid - and is not dictated by the number of views. On ShapeNet-SRN and Google Scanned Objects (GSO), FlexSplat matches or approaches posed state-of-the-art reconstructors while requiring neither camera poses nor ground-truth depth, and matches the best perceptual (LPIPS) quality among the compared methods on GSO. Our results indicate that a jointly trained geometry front-end is sufficient to bring calibration-free operation to query-based Gaussian reconstruction while staying within 0.7 dB PSNR of posed methods and matching their perceptual quality.
Chinese Translation
我们提出了FlexSplat，一个用于从未校准的以物体为中心的多视图图像集合中进行新视图合成（NVS）的前馈框架。最近一系列基于查询的方法通过将3D高斯视为变换器查询，并通过多视图可变形注意力进行细化，从而重建了一组紧凑的3D高斯；然而，这些方法假设相机姿态是已知的。FlexSplat去除了这一假设：几何变换器与高斯解码器共同训练，以预测每幅图像的相机参数和深度，从而为深度引导的高斯参数化和多视图可变形交叉注意力奠定基础，将所有输入视图的证据聚合成一个单一的、视图一致的原语集合。一个不确定性加权的深度一致性目标使得共同训练的几何模型能够适应重建任务，而在解码过程中形成的跨视图共识则吸收了估计的相机和深度的残差误差。该表示使用了一个与输入分辨率解耦的紧凑高斯预算——与像素对齐的方法不同，原语数量不会随着图像网格的增大而增加，并且不受视图数量的限制。在ShapeNet-SRN和Google扫描对象（GSO）上，FlexSplat的性能与现有的最先进重建方法相匹配或接近，同时不需要相机姿态或真实深度，并且在GSO上在比较方法中达到了最佳的感知（LPIPS）质量。我们的结果表明，联合训练的几何前端足以使基于查询的高斯重建实现无校准操作，同时在PSNR上与有姿态的方法保持在0.7 dB以内，并匹配其感知质量。
cs.CV / 57 / 2608.07941
LAD-COD: Language-Aligned Dense Perception for Camouflaged Object Detection
LAD-COD：用于伪装物体检测的语言对齐密集感知
Abstract
Camouflaged object detection (COD) aims to segment objects that exhibit high visual similarity to their surroundings, which reduces foreground-background discriminability and weakens boundary evidence across appearance, texture, and structure. Such limitations motivate the use of instruction-conditioned semantics as top-down guidance for identifying which weak visual cues are relevant to the target. Recent segmentation systems built on large multimodal models (LMMs) demonstrate this possibility through instruction-conditioned target embeddings that guide mask decoding. However, in this language-to-mask paradigm, the generated target embedding conditions mainly the mask decoder, leaving the dense visual features that must preserve low-contrast boundaries and fine local structure without explicit guidance. We propose Language-Aligned Dense perception for COD (LAD-COD), a framework that aligns top-down semantic target guidance with bottom-up hierarchical visual features. Instead of fully adapting a large generic image encoder, LAD-COD learns a trainable hierarchical visual branch that captures camouflage-sensitive texture, boundary, and contextual information. To align these features with the target embedding, LAD-COD applies Language-Aligned Dual Visual Fusion (LADVF), which extends the embedding beyond sparse prompting to query patch-level language-aligned features and to gate their residual integration with the hierarchical features. This design allows semantic information to guide localization while preserving the fine structural details needed for camouflage segmentation. Experiments on CAMO, COD10K, and NC4K show that LAD-COD obtains the best reported value in all 12 dataset-metric comparisons.
Chinese Translation
伪装物体检测（COD）旨在分割与其周围环境具有高度视觉相似性的物体，这降低了前景与背景的可区分性，并削弱了外观、纹理和结构上的边界证据。这些限制促使我们使用基于指令的语义作为自上而下的指导，以识别哪些弱视觉线索与目标相关。最近建立在大型多模态模型（LMMs）上的分割系统通过基于指令的目标嵌入展示了这一可能性，这些嵌入指导掩膜解码。然而，在这种语言到掩膜的范式中，生成的目标嵌入主要条件化掩膜解码器，留下必须保留低对比度边界和细微局部结构的密集视觉特征而没有明确的指导。我们提出了用于COD的语言对齐密集感知框架（LAD-COD），该框架将自上而下的语义目标指导与自下而上的层次视觉特征对齐。LAD-COD并不是完全适应一个大型通用图像编码器，而是学习一个可训练的层次视觉分支，以捕捉对伪装敏感的纹理、边界和上下文信息。为了将这些特征与目标嵌入对齐，LAD-COD应用了语言对齐双视觉融合（LADVF），该方法将嵌入扩展到稀疏提示之外，以查询补丁级别的语言对齐特征，并对其与层次特征的残差集成进行门控。这一设计允许语义信息在保持伪装分割所需的细微结构细节的同时指导定位。在CAMO、COD10K和NC4K上的实验表明，LAD-COD在所有12个数据集-指标比较中获得了最佳报告值。
cs.CV / 58 / 2608.07948
SynVAR: Synergizing Spatial and Semantic Alignment in Visual Autoregressive Model
SynVAR：在视觉自回归模型中协同空间与语义对齐
Abstract
VAR has gained widespread popularity due to its next-scale prediction paradigm. However, it faces substantial performance bottlenecks when handling complex scenes with multiple objects and attributes. Existing diffusion-based enhancement methods fail to adequately address the unique challenge of cross-scale error propagation and accumulation in VAR. To this end, we propose SynVAR, the first training-free enhancement framework specifically tailored for the VAR paradigm, which introduces a spatial-semantic collaborative control strategy to effectively suppress propagation error and improve generation quality. SynVAR comprises three key components: (1) Global guidance to ensure reasonable spatial structure in the early stages, (2) Receptive field constraints to mitigate early-stage semantic confusion, (3) High-frequency compensation to recover fine-grained details. Extensive quantitative and qualitative experiments demonstrate the significant improvements in the ability of SynVAR to enhance the VAR's capability for complex scene modeling.
Chinese Translation
VAR因其下一尺度预测范式而广受欢迎。然而，在处理具有多个对象和属性的复杂场景时，它面临着显著的性能瓶颈。现有的基于扩散的增强方法未能充分解决VAR中跨尺度误差传播和积累的独特挑战。为此，我们提出了SynVAR，这是第一个专门为VAR范式量身定制的无训练增强框架，采用空间-语义协同控制策略，有效抑制传播误差并提高生成质量。SynVAR包含三个关键组件：(1) 全局引导，以确保早期阶段合理的空间结构；(2) 感受野约束，以减轻早期阶段的语义混淆；(3) 高频补偿，以恢复细粒度细节。大量定量和定性实验表明，SynVAR在增强VAR处理复杂场景建模能力方面取得了显著改善。
cs.CV / 59 / 2608.07958
LIBAD: A Multimodal Anomaly Detection Benchmark for Li-Ion Battery Electrode Manufacturing
LIBAD：锂离子电池电极制造的多模态异常检测基准
Abstract
Multimodal industrial anomaly detection has largely focused on discrete products using strongly correlated RGB and 3D observations, leaving continuous process manufacturing and weakly correlated sensing modalities underexplored. We introduce LIBAD, the first multimodal anomaly detection benchmark for Li-ion battery electrode manufacturing. Collected from real roll-to-roll production lines, LIBAD provides aligned double-sided visible-light imaging, high-resolution X-ray radiography, and inline-compatible low-resolution X-ray radiography. Electrode patches in LIBAD exhibit highly homogeneous material appearance, while defect evidence can be strong in one modality but weak or absent in another, resulting in pronounced cross-modal anomaly inconsistency. Benchmarks of representative methods under the inline-compatible visible-light and low-resolution X-ray setting exhibit limited transferability and consistently high false-positive rates. We therefore propose DA-Core, a memory-based method that jointly considers feature-space coverage and local density of normal features during coreset selection, allowing compact memory banks to better preserve fine-grained normal variations. With a coreset ratio of 0.05, DA-Core reduces FPR95 from 60.4% to 54.3% compared with standard farthest point sampling. At this ratio, DA-Core also outperforms the best standard coreset result (obtained at 0.20) while reducing inference time by 43.9%. These results suggest that both the data distribution of normal features and the modality relationship itself require explicit consideration when designing anomaly detection methods for process manufacturing.
Chinese Translation
多模态工业异常检测主要集中于使用强相关的RGB和3D观测的离散产品，而对连续过程制造和弱相关的传感模态的研究相对较少。我们提出了LIBAD，这是第一个针对锂离子电池电极制造的多模态异常检测基准。LIBAD数据集来自真实的卷对卷生产线，提供了对齐的双面可见光成像、高分辨率X射线成像和在线兼容的低分辨率X射线成像。LIBAD中的电极片表现出高度均匀的材料外观，而缺陷证据在一种模态中可能很强，但在另一种模态中可能较弱或缺失，导致明显的跨模态异常不一致。在在线兼容的可见光和低分辨率X射线设置下，代表性方法的基准表现出有限的可转移性和持续较高的假阳性率。因此，我们提出了DA-Core，这是一种基于记忆的方法，在核心集选择过程中共同考虑特征空间覆盖和正常特征的局部密度，从而使紧凑的记忆库能够更好地保留细粒度的正常变异。使用0.05的核心集比例，DA-Core将FPR95从60.4%降低到54.3%，相比于标准的最远点采样。在这个比例下，DA-Core的表现也优于最佳标准核心集结果（在0.20时获得），同时将推理时间减少了43.9%。这些结果表明，在为过程制造设计异常检测方法时，正常特征的数据分布和模态关系本身都需要明确考虑。
cs.CV / 60 / 2608.07981
Distilling Physical Priors into Streaming World Models
将物理先验知识提炼为流式世界模型
Abstract
Streaming world models predict future visual states online while maintaining physically coherent dynamics over long horizons. However, their rollouts often violate basic physical constraints. A common approach distills pretrained bidirectional DiTs into few-step causal generators. However, this paradigm suffers from two fundamental limitations: generic bidirectional teachers acquire limited physical priors from visually oriented pretraining, and the limited priors suffer further loss during bidirectional-to-causal distillation. We present PhyS, a three-stage framework for distilling physical priors into streaming world models. To acquire physical priors from real-world interactions, we construct PhyS-120K, a dataset of 120K real-world physical-interaction videos spanning rigid-body dynamics, soft-body deformation, fluid phenomena, and phase transitions. Each video is annotated with structured descriptions of object properties and causal state transitions. Physics-aware supervised fine-tuning injects the physical priors into a bidirectional 14B DiT teacher, which we then distill into a lightweight 1.3B causal DiT for few-step autoregressive streaming generation. Finally, we use online reinforcement learning to incentivize the distilled model to generate physically plausible rollouts and further propose Temporal Credit Routing (TCR) to address temporal credit assignment. TCR evaluates physical consistency over overlapping temporal windows and routes the resulting group-relative advantages to temporally aligned denoising actions. On PhysicsIQ, PhyS improves the Wan2.1-14B teacher by 18.2\% and the Self Forcing, Rolling Forcing, and Causal Forcing by 23.7\%, 14.8\%, and 31.4\%, respectively. Results also improve the physics-aware video benchmarks VideoPhy, VideoPhy2, and PhyGenBench. The dataset, code, and more sample videos are available on our Project Page.
Chinese Translation
流式世界模型在线预测未来视觉状态，同时在长时间范围内保持物理一致的动态。然而，它们的展开往往违反基本的物理约束。一个常见的方法是将预训练的双向 DiT（Dual-Task Transformer）提炼为少步因果生成器。然而，这一范式存在两个基本限制：通用的双向教师从以视觉为导向的预训练中获得的物理先验知识有限，而在双向到因果的提炼过程中，有限的先验知识又会进一步丧失。我们提出了 PhyS，这是一个将物理先验知识提炼为流式世界模型的三阶段框架。为了从真实世界的交互中获取物理先验知识，我们构建了 PhyS-120K，这是一个包含 120K 个真实世界物理交互视频的数据集，涵盖了刚体动力学、软体变形、流体现象和相变。每个视频都附有对象属性和因果状态转变的结构化描述。基于物理的监督微调将物理先验知识注入到一个双向 14B DiT 教师中，然后我们将其提炼为一个轻量级的 1.3B 因果 DiT，用于少步自回归流式生成。最后，我们使用在线强化学习来激励提炼后的模型生成物理上合理的展开，并进一步提出了时间信用路由（Temporal Credit Routing, TCR）来解决时间信用分配问题。TCR 在重叠的时间窗口中评估物理一致性，并将结果的组相对优势路由到时间上对齐的去噪动作。在 PhysicsIQ 上，PhyS 将 Wan2.1-14B 教师的表现提高了 18.2\%，并分别提高了 Self Forcing、Rolling Forcing 和 Causal Forcing 的表现 23.7\%、14.8\\% 和 31.4\\%。结果还改善了物理感知视频基准 VideoPhy、VideoPhy2 和 PhyGenBench。数据集、代码及更多示例视频可在我们的项目页面上获取。
cs.CV / 61 / 2608.07982
AdaDINO: Pair-Aware In-Backbone Adaptation of Frozen DINO for Efficient Remote Sensing Change Detection
AdaDINO：冻结 DINO 的配对感知骨干适应用于高效的遥感变化检测
Abstract
Vision foundation models (VFMs) such as DINO are pretrained for single-image representation, whereas remote sensing change detection requires reasoning over a bi-temporal pair. Existing VFM-based methods usually encode the two images independently and compare them only afterward, leaving the VFM backbone unaware of cross-temporal relations. To bridge this mismatch, we present AdaDINO, a pair-aware in-backbone adaptation framework that equips a frozen DINO encoder with bi-temporal interaction for efficient change detection. Its core component, Change-aware Gated Local Adaptation (CGLA), couples the two streams after selected frozen blocks and injects a shared temporal residual into them with opposite signs, enhancing genuine change responses while preserving the pair midpoint. Batch-Shared Chunk Selection (BSCS) further reduces feed-forward network (FFN) computation by retaining a batch-shared subset of channel chunks that can be executed as a compact dense FFN. A CGLA-Prior-Guided Refinement (CPGR) decoder reuses encoder-side change responses for coarse-to-fine prediction. Experiments on four remote sensing change detection benchmarks show that AdaDINO achieves competitive or superior performance against VFM-based baselines, with the largest gain on the category-agnostic SYSU-CD dataset. With 62.5% of the FFN hidden width removed, AdaDINO still achieves an F1 score of 85.29% on SYSU-CD while delivering a 1.41$\times$ throughput speedup. The code will be released.
Chinese Translation
视觉基础模型（VFM）如 DINO 是针对单幅图像表示进行预训练的，而遥感变化检测则需要对双时相图像进行推理。现有的基于 VFM 的方法通常独立编码这两幅图像，并在之后进行比较，这使得 VFM 骨干无法感知跨时间的关系。为了解决这一不匹配问题，我们提出了 AdaDINO，这是一种配对感知的骨干适应框架，旨在为冻结的 DINO 编码器提供双时相交互，以实现高效的变化检测。其核心组件，变化感知门控局部适应（CGLA），在选定的冻结块后将两个流结合，并以相反的符号注入共享的时间残差，从而增强真实变化的响应，同时保持配对的中点。批次共享块选择（BSCS）通过保留可以作为紧凑密集前馈网络（FFN）执行的批次共享通道块子集，进一步减少前馈网络的计算。CGLA-先验引导的细化（CPGR）解码器重用编码器侧的变化响应进行粗到细的预测。在四个遥感变化检测基准上的实验表明，AdaDINO 在与基于 VFM 的基线相比时，表现出竞争力或更优的性能，在类别无关的 SYSU-CD 数据集上获得了最大的提升。在去除 62.5% FFN 隐藏宽度的情况下，AdaDINO 在 SYSU-CD 上仍然实现了 85.29% 的 F1 分数，同时提供了 1.41 倍的吞吐速度提升。代码将会发布。
cs.CV / 62 / 2608.07984
AgriField-40K: Adapting Vision Models to Agriculture With Efficient Continual Pretraining
AgriField-40K：通过高效的持续预训练将视觉模型适应于农业
Abstract
Field-based agricultural computer vision is important for precision agriculture, yet it largely depends on expensive annotations and costly adaptation of large pretrained models. We introduce AgriField-40K, a field-centric dataset curated from 17 public resources and covering diverse crops, weeds, and field conditions. Building on this, we present AgriMAE, a parameter-efficient continual pretraining baseline that adapts a masked autoencoder pretrained on natural images by training only lightweight adapters. We further explore semantic feature reconstruction as an alternative pretraining objective and evaluate transfer across multiple tasks. AgriMAE consistently improves downstream performance and can match or even outperform full fine-tuning while using up to $9\times$ fewer trainable parameters, showing that AgriField-40K is a practical resource for continual pretraining in agricultural vision. Project page: https://dtu-pas.github.io/agrifield40k/
Chinese Translation
基于田野的农业计算机视觉对于精准农业至关重要，但它在很大程度上依赖于昂贵的标注和大型预训练模型的高成本适应。我们介绍了AgriField-40K，这是一个以田野为中心的数据集，来自17个公共资源，涵盖了多样的作物、杂草和田间条件。在此基础上，我们提出了AgriMAE，这是一种参数高效的持续预训练基线，通过仅训练轻量级适配器，适应在自然图像上预训练的掩码自编码器。我们进一步探索语义特征重建作为替代的预训练目标，并评估在多个任务之间的迁移。AgriMAE始终提高下游性能，并且在使用多达9倍更少的可训练参数的情况下，可以匹配甚至超越完全微调的效果，显示出AgriField-40K是农业视觉持续预训练的实用资源。项目页面：https://dtu-pas.github.io/agrifield40k/
cs.CV / 63 / 2608.07987
Advantage-Guided Gate: Reshaping Open-Ended Reasoning for Vision-Based Spatial Intelligence
优势引导门：重塑基于视觉的空间智能的开放式推理
Abstract
Multimodal large language models (MLLMs) have demonstrated significant potential in complex spatial scene understanding and reasoning tasks. However, their open-ended reasoning process is prone to decision errors and error accumulation, leading to instability in answer quality. To address this, we propose an advantage-guided gating framework that dynamically intervenes in and corrects deviations during the reasoning process. Specifically, we model step-by-step reasoning as a finite-horizon decision process and introduce Monte Carlo value evaluation on the reasoning tree to provide intermediate supervision signals. The framework includes Step-Advantage Gate and Trajectory-Advantage Gate, which dynamically select high-value reasoning steps and high-quality complete reasoning trajectories, respectively. During training, we perform supervised learning for the gates using reasoning trees generated via multi-branch sampling, and combine shared-parameter initialization with task-specific heads to achieve cross-task robustness and diversity. During inference, the model greedily selects high-value prefix reasoning steps while choosing the optimal reasoning head based on the problem type, thereby significantly improving the accuracy of the final answer. Furthermore, we constructed the Reasoning-Tree-160k dataset and performed two-stage learning on it. Extensive experiments demonstrate that this advantage-guided gating framework effectively enhances the performance of benchmark MLLMs in visual-based spatial understanding and reasoning tasks. The code is open to the public for research: https://github.com/LingLin-ll/Advantage-Guided-Gate.
Chinese Translation
多模态大型语言模型（MLLMs）在复杂空间场景理解和推理任务中展现出显著的潜力。然而，它们的开放式推理过程容易出现决策错误和错误累积，导致答案质量的不稳定。为了解决这一问题，我们提出了一种优势引导的门控框架，该框架在推理过程中动态干预并纠正偏差。具体而言，我们将逐步推理建模为有限时域决策过程，并在推理树上引入蒙特卡洛值评估，以提供中间监督信号。该框架包括步骤优势门（Step-Advantage Gate）和轨迹优势门（Trajectory-Advantage Gate），分别动态选择高价值的推理步骤和高质量的完整推理轨迹。在训练过程中，我们使用通过多分支采样生成的推理树对门进行监督学习，并结合共享参数初始化与任务特定头部，以实现跨任务的鲁棒性和多样性。在推理阶段，模型贪婪地选择高价值的前缀推理步骤，同时根据问题类型选择最佳推理头，从而显著提高最终答案的准确性。此外，我们构建了Reasoning-Tree-160k数据集，并在其上进行了两阶段学习。大量实验表明，这一优势引导的门控框架有效提升了基准MLLMs在基于视觉的空间理解和推理任务中的表现。代码已公开供研究使用：https://github.com/LingLin-ll/Advantage-Guided-Gate。
cs.CV / 64 / 2608.07993
MRBench: A Comprehensive Benchmark for Human Motion-Text Retrieval
MRBench：人类动作-文本检索的综合基准
Abstract
Human motion-text retrieval provides a rigorous means of assessing cross-modal alignment. Prevailing benchmarks are dominated by homogeneous indoor motions, imbalanced motion distributions, and oversimplified, repetitive texts, which hinder the reliable measurement of cross-domain and cross-granularity alignment. We thus introduce MRBench, a comprehensive motion-text retrieval benchmark featuring heterogeneous motions, broad and balanced category coverage, and reliable, discriminative, multi-granular descriptions. MRBench is constructed through a meticulously designed multi-stage data curation pipeline, which filters and balances candidates, verifies unambiguous semantic alignment, and generates motion-grounded descriptions at multiple granularities. The resulting benchmark contains 3,390 motions drawn from motion capture, in-the-wild videos, synthetic videos, and motion generative models, covering 118 fine-grained categories. Each motion is paired with concise, standard, and fine-grained descriptions, yielding 10,170 captions. Extensive evaluations of representative retrieval baselines on MRBench reveal a substantial cross-dataset generalization gap and pronounced sensitivity to query granularity. We propose a lightweight granularity-aware model anchored at a frozen standard-caption-aligned retrieval model. LLM-based concise and fine-grained captions provide pseudo-supervision for extra-branch granularity-specific motion extractors and text adapters. For inference, granularity-aware score fusion integrates global and adapted similarities while strictly maintaining score comparability across all description levels. The resulting model improves mixed-granularity retrieval without compromising standard-caption performance. We believe that our MRBench provides a comprehensive testbed for advancing motion-language alignment evaluation.
Chinese Translation
人类动作-文本检索为评估跨模态对齐提供了一种严谨的方法。目前的基准主要集中在同质的室内动作、失衡的动作分布以及过于简化和重复的文本上，这妨碍了跨领域和跨粒度对齐的可靠测量。因此，我们提出了MRBench，这是一个综合的动作-文本检索基准，具有异质的动作、广泛且均衡的类别覆盖，以及可靠、具辨别力的多粒度描述。MRBench通过精心设计的多阶段数据整理流程构建，该流程过滤和平衡候选项，验证明确的语义对齐，并生成多粒度的动作基础描述。最终的基准包含3,390个动作，来源于动作捕捉、自然环境视频、合成视频和动作生成模型，涵盖118个细粒度类别。每个动作都配有简洁、标准和细粒度的描述，共生成10,170个标题。对MRBench上代表性检索基线的广泛评估揭示了显著的跨数据集泛化差距和对查询粒度的明显敏感性。我们提出了一种轻量级的粒度感知模型，该模型基于一个冻结的标准标题对齐检索模型。基于大型语言模型（LLM）的简洁和细粒度标题为额外分支的粒度特定动作提取器和文本适配器提供伪监督。在推理过程中，粒度感知的得分融合整合了全局和适配的相似性，同时严格保持所有描述级别的得分可比性。最终的模型在不影响标准标题性能的情况下改善了混合粒度检索。我们相信，MRBench为推动动作-语言对齐评估提供了一个全面的测试平台。
cs.CV / 65 / 2608.07999
PE-Mamba: Bidirectional Selective Layer Aggregation for AI-Generated Image Detection
PE-Mamba：用于AI生成图像检测的双向选择性层聚合
Abstract
AI-generated image (AIGI) detection has become increasingly challenging due to the rapid advancement of generative models and the diminishing gap between synthetic and authentic content. Existing vision transformer-based detectors commonly rely on weighted-sum strategies to aggregate intermediate representations across transformer layers, often overlooking the inherently ordered semantic progression of hierarchical features from shallow texture cues to deep semantic representations. In this work, we propose \textbf{PE-Mamba}, a novel framework built upon a pre-trained PE-Core vision transformer with lightweight LoRA adaptation that introduces three complementary components for cross-layer feature aggregation and fusion. First, a bidirectional selective aggregator (BSA) processes layer-wise classification tokens through forward and backward selective scans, where the forward scan progressively accumulates shallow-to-deep forensic evidence, and the backward scan performs deep-to-shallow contextual refinement to reinterpret low-level cues in light of high-level semantic context. Second, a softmax-weighted aggregator (SWA) computes a learned global summary of all layer tokens as a complementary aggregation path. Third, a sigmoid-gated blend (SGA) adaptively fuses the BSA and SWA outputs via a learnable scalar gate, allowing the model to dynamically balance directional sequential evidence and global layer-wise aggregation. Extensive experiments on UniversalFakeDetect (96.6\% mACC, 99.5\% mAP) and AIGCDetect (95.3\% mACC, 98.1\% mAP) demonstrate that \methodname{} outperforms 18 detectors with superior generalization across diverse generative models, while training only 1.3\% of total parameters (0.13\% for LoRA alone).
Chinese Translation
AI生成图像（AIGI）检测因生成模型的快速发展以及合成内容与真实内容之间差距的缩小而变得愈加具有挑战性。现有基于视觉变换器的检测器通常依赖加权求和策略来聚合变换器层之间的中间表示，往往忽视了从浅层纹理线索到深层语义表示的分层特征的固有有序语义进展。在本研究中，我们提出了 extbf{PE-Mamba}，这是一个基于预训练的PE-Core视觉变换器构建的新框架，采用轻量级的LoRA适配，引入了三个互补组件用于跨层特征聚合和融合。首先，双向选择性聚合器（BSA）通过正向和反向选择扫描处理逐层分类标记，其中正向扫描逐步积累从浅层到深层的取证证据，反向扫描则进行从深层到浅层的上下文细化，以高层语义上下文重新解释低层线索。其次，软最大加权聚合器（SWA）计算所有层标记的学习全局摘要，作为互补的聚合路径。第三，带有sigmoid门控的混合（SGA）通过可学习的标量门自适应融合BSA和SWA的输出，使模型能够动态平衡方向性序列证据和全局逐层聚合。在UniversalFakeDetect（96.6 ext{% mACC}, 99.5 ext{% mAP}）和AIGCDetect（95.3 ext{% mACC}, 98.1 ext{% mAP}）上的大量实验表明， extbf{PE-Mamba}在18个检测器中表现优越，并在多种生成模型中展现出更好的泛化能力，同时仅训练了总参数的1.3 ext{%}（仅LoRA为0.13 ext{%}）。
cs.CV / 66 / 2608.08009
Evidence-Grounded Forensic Reasoning for Detecting and Grounding Multi-Modal Media Manipulation
基于证据的法医推理用于检测和定位多模态媒体操纵
Abstract
Fake news increasingly relies on cross-modal image-text forgeries, making transparent and verifiable reasoning chains an urgent need for Detecting and Grounding Multi-Modal Media Manipulation (DGM4). Existing methods produce black-box detection results without any decision rationale, limiting their reliability in forensic practice. Multi-modal Large Language Models (MLLMs) offer a natural path toward explainability, but applying them to DGM4 raises two difficulties. First, models tend to generate explanations disconnected from predicted evidence locations, producing unverified attribution. Second, enforcing evidence-conclusion consistency requires active optimization, yet uniform training signals fail to distinguish localization tokens from classification tokens, making multi-head joint training unreliable. We propose a multi-modal manipulation detector based on an Evidence-Grounded Forensic Reasoning (EFR) framework. EFR introduces an Anchor-and-Verify reasoning chain that enforces modality-isolated perception before cross-modal comparison, with conclusion coordinates as explicit anchors to which downstream evidence must spatially correspond. A verifiable reward system then enforces evidence-conclusion consistency during training, while a Modality-Decoupled Advantage (MDA) routing mechanism mitigats credit misassignment across prediction tasks. Experiments show that EFR achieves state-of-the-art performance while producing structured forensic reasoning records that explicitly bind explanations to evidence.
Chinese Translation
假新闻越来越依赖于跨模态的图像-文本伪造，因此迫切需要透明且可验证的推理链以检测和定位多模态媒体操纵（DGM4）。现有方法产生黑箱检测结果，没有任何决策依据，限制了其在法医实践中的可靠性。多模态大型语言模型（MLLMs）为可解释性提供了一条自然路径，但将其应用于DGM4面临两个困难。首先，模型往往生成与预测证据位置无关的解释，导致未验证的归因。其次，强制证据与结论的一致性需要主动优化，但统一的训练信号无法区分定位标记和分类标记，使得多头联合训练不可靠。我们提出了一种基于证据的法医推理（EFR）框架的多模态操纵检测器。EFR引入了一种锚定与验证的推理链，在跨模态比较之前强制进行模态隔离感知，结论坐标作为显性锚点，必须与下游证据在空间上对应。然后，验证奖励系统在训练过程中强制证据与结论的一致性，而模态解耦优势（MDA）路由机制则减轻了预测任务之间的信用误分配。实验表明，EFR在实现最先进性能的同时，生成了结构化的法医推理记录，明确将解释与证据绑定。
cs.CV / 67 / 2608.08016
EgoTrack3D: A Modular Framework for Egocentric 3D Object Tracking
EgoTrack3D：一种模块化的自我中心3D物体跟踪框架
Abstract
Understanding 3D scenes from egocentric video is fundamental for robotics and autonomous navigation, yet rapid viewpoint changes and partial occlusions make building structured representations challenging. Existing 3D tracking and scene graph construction methods primarily address explicit interactions or assume static scenes, limiting their ability to capture complex dynamics. We introduce EgoTrack3D, a modular framework that reconstructs and maintains a dynamic 3D scene representation directly from egocentric RGB video. The framework lifts 2D segmentation masks into a global 3D coordinate frame, using a point-based motion scoring mechanism alongside a voxel-based merging heuristic to associate object tracks. EgoTrack3D maintains accurate representations over time, achieving an 11% improvement in percentage of correct locations (PCL) relative to the strongest baseline on the Aria Digital Twin (ADT) dataset, while addressing the more general setting of persistent 3D tracking for both static and dynamic objects. Furthermore, to demonstrate the system's robustness under degraded conditions that simulate real-world deployment constraints, we replace dense depth maps with sparse 3D bounding box estimation and integrate interaction-guided dynamic association, enabling EgoTrack3D to maintain accurate spatial representations despite noisy observations.
Chinese Translation
从自我中心视频理解3D场景对于机器人技术和自主导航至关重要，但快速的视角变化和部分遮挡使得构建结构化表示变得具有挑战性。现有的3D跟踪和场景图构建方法主要关注显式交互或假设静态场景，这限制了它们捕捉复杂动态的能力。我们提出了EgoTrack3D，这是一种模块化框架，能够直接从自我中心RGB视频重建和维护动态3D场景表示。该框架将2D分割掩码提升到全局3D坐标系中，使用基于点的运动评分机制和基于体素的合并启发式方法来关联物体轨迹。EgoTrack3D在时间上保持准确的表示，相较于Aria Digital Twin (ADT) 数据集上最强基线，其正确位置百分比（PCL）提高了11%，同时解决了静态和动态物体的持久3D跟踪的更一般设置。此外，为了展示系统在模拟现实世界部署约束下的退化条件下的鲁棒性，我们用稀疏3D边界框估计替代了密集深度图，并整合了交互引导的动态关联，使EgoTrack3D能够在噪声观测下保持准确的空间表示。
cs.CV / 68 / 2608.08021
Evidence-RL: Towards Evidence-intensive Visual Reasoning
证据强化学习：朝向证据密集的视觉推理
Abstract
Vision-Language Models (VLMs) should answer from concrete image evidence rather than language priors, dataset shortcuts, or irrelevant visual context. Existing perception-aware post-training methods encourage image use through global perturbations or attention proxies, but they do not test whether a sampled answer causally depends on the local evidence that supports it. We propose Counterfactual Evidence Disentanglement (CED), a training-time evidence audit for VLM grounding. For each response, CED neutralizes an object-centric Evidence Region and compares the resulting support drop against matched non-evidence Regions. We combine this signal with answer correctness inside GRPO, rewarding correct answers that rely on the evidence path rather than shortcut or nuisance paths. CED uses weak object-level proposals, requires no question-specific evidence annotations, and adds no inference-time overhead. Across nine public benchmarks and four backbones, CED outperforms prior RL-based post-training methods, with targeted analyses verifying its object-centric signal.
Chinese Translation
视觉-语言模型（VLMs）应基于具体的图像证据进行回答，而不是依赖语言先验、数据集捷径或无关的视觉上下文。现有的感知意识后训练方法通过全局扰动或注意力代理鼓励图像的使用，但并未测试所采样的答案是否因其支持的局部证据而因果依赖。我们提出了反事实证据解耦（Counterfactual Evidence Disentanglement, CED），这是一种用于VLM基础的训练时证据审计。对于每个响应，CED 中和一个以对象为中心的证据区域，并将结果支持度的下降与匹配的非证据区域进行比较。我们将这一信号与GRPO中的答案正确性结合起来，奖励依赖于证据路径而非捷径或干扰路径的正确答案。CED使用弱对象级提案，不需要特定于问题的证据注释，并且不会增加推理时的开销。在九个公共基准和四个基础模型上，CED的表现优于先前基于强化学习的后训练方法，针对性的分析验证了其以对象为中心的信号。
cs.CV / 69 / 2608.08060
ZOMP: Zeroth-Order Multi-Modal Prompt Tuning for Vision-Language Models
ZOMP：用于视觉-语言模型的零阶多模态提示调优
Abstract
Fine-tuning vision-language models such as CLIP typically requires backpropagation (BP) through the full model, which is infeasible when only forward-pass access is available, as is common for memory-constrained edge devices and proprietary model deployments. Prior BP-free, zeroth-order prompt-tuning methods avoid this requirement but often tune prompts in a single modality or optimize over a search space large enough that convergence requires thousands of forward passes, which is impractical under realistic query budgets. We propose ZOMP (Zeroth-Order Multimodal Prompt tuning), a query-efficient, fully forward-only method that tunes deep prompts in both the vision and text branches of a frozen CLIP model using simultaneous perturbation stochastic approximation. ZOMP combines three ingredients: a cross-modal low-rank reparameterization that ties the two branches through a shared factor and keeps the effective search dimensionality small, a gradient-correction momentum term that stabilizes the noisy zeroth-order estimate, and a budget-indexed rank schedule that unlocks capacity as the query budget is spent. Across 13 vision-language benchmarks under a matched 5,000-query budget, ZOMP consistently outperforms prior BP-free prompt-tuning methods in both few-shot accuracy and query efficiency, and it generalizes better across base-to-new, cross-dataset transfer, and out-of-distribution settings. Our results show that jointly exploiting multimodality and low-rank structure is an effective route to practical, query-efficient BP-free prompt tuning.
Chinese Translation
对视觉-语言模型（如 CLIP）的微调通常需要通过整个模型进行反向传播（BP），而当仅能进行前向传播时，这种方法在内存受限的边缘设备和专有模型部署中是不可行的。之前的无反向传播、零阶提示调优方法避免了这一要求，但通常仅在单一模态中调优提示，或者在一个足够大的搜索空间中优化，以至于收敛需要数千次前向传播，这在实际查询预算下是不切实际的。我们提出了 ZOMP（零阶多模态提示调优），这是一种查询高效、完全基于前向传播的方法，它利用同时扰动随机近似法在冻结的 CLIP 模型的视觉和文本分支中调优深层提示。ZOMP 结合了三个要素：一种跨模态低秩重参数化方法，通过共享因子将两个分支联系在一起，并保持有效的搜索维度较小；一种梯度修正动量项，稳定嘈杂的零阶估计；以及一种预算索引的秩调度，随着查询预算的消耗释放容量。在匹配 5000 查询预算的 13 个视觉-语言基准测试中，ZOMP 在少量样本准确性和查询效率方面始终优于之前的无反向传播提示调优方法，并且在基础到新、跨数据集迁移和分布外设置中具有更好的泛化能力。我们的结果表明，联合利用多模态性和低秩结构是实现实用、查询高效的无反向传播提示调优的有效途径。
cs.CV / 70 / 2608.08066
EvBS: Event-guided Blur Synthesis for Domain-adaptive Motion Deblurring
EvBS：用于领域自适应运动去模糊的事件引导模糊合成
Abstract
Motion deblurring has achieved remarkable progress with deep learning, yet pre-trained deblurring models often suffer from performance degradation in real-world scenarios due to the domain shift between training and testing distributions. To remedy this, we propose EvBS, an event-guided blur synthesis framework that generates diverse training pairs for calibrating pre-trained models to the target domain. While existing methods are constrained by the inherent entanglement between motion and visual content, our method leverages the high temporal resolution of event cameras to effectively decouple them. This enables us to utilize not only the intrinsic motion that is inherent to the given content but also extrinsic motion transferred from different sources within the target domain, thereby facilitating effective adaptation via fine-tuning. Specifically, EvBS comprises two complementary strategies: Intrinsic-Blur Synthesis, which blurs sharp contents with their own motion patterns, and Extrinsic-Blur Synthesis, which transfers motion from blurry patches to distinct sharp content. This approach generates a diverse set of training pairs that break the inherent constraints of naturally coupled motion and content, resulting in enhanced domain-adaptive deblurring performance. Extensive experiments on multiple benchmarks demonstrate that EvBS effectively enhances the robustness of existing deblurring models on unseen testing datasets.
Chinese Translation
运动去模糊在深度学习的推动下取得了显著进展，但预训练的去模糊模型在真实场景中往往因训练和测试分布之间的领域转移而表现不佳。为了解决这一问题，我们提出了EvBS，一个事件引导的模糊合成框架，旨在生成多样化的训练对，以便将预训练模型校准到目标领域。现有方法受到运动与视觉内容之间固有纠缠的限制，而我们的方法利用事件相机的高时间分辨率有效地将二者解耦。这使我们能够利用不仅是给定内容固有的内在运动，还包括从目标领域内不同来源转移的外在运动，从而通过微调实现有效的适应。具体而言，EvBS包含两种互补策略：内在模糊合成（Intrinsic-Blur Synthesis），通过自身运动模式对清晰内容进行模糊处理；外在模糊合成（Extrinsic-Blur Synthesis），将模糊区域的运动转移到不同的清晰内容上。这种方法生成了一组多样化的训练对，打破了自然耦合的运动与内容之间的固有约束，从而提升了领域自适应去模糊的性能。在多个基准测试上的大量实验表明，EvBS有效增强了现有去模糊模型在未见测试数据集上的鲁棒性。
cs.CV / 71 / 2608.08068
NeuroGuard: Neural Gradient Update Aware of Representation Damage
NeuroGuard：关注表示损伤的神经梯度更新
Abstract
Long-tailed class-incremental learning (LT-CIL) must learn new classes from imbalanced streams while retaining old classes. Existing methods mainly change replay, classifiers, or losses. We study a different factor, namely how strongly the feature representation should be updated at each task boundary. We propose NeuroGuard, an update-control method added to DGR, a replay-based LT-CIL baseline, without adding learnable parameters. NeuroGuard preserves DGR's replay memory, classifier, and set of loss terms. Adaptive Gradient Scaling (AGS) converts teacher uncertainty into one task-wise gradient scale. Confidence-Ranked Knowledge Distillation Reweighting (CRK) gives larger knowledge-distillation weights to replay samples that the teacher predicts less decisively. Fragility-Blended Entropy Gate (FBE) adds old-memory leakage to the scale decision. Across five LT-CIL settings, NeuroGuard improves over DGR in every setting. In the four main benchmark comparisons, it achieves the best task-agnostic accuracy among the compared methods. The gains extend to both old- and new-class accuracy, while medium-frequency accuracy improves consistently across all five settings. Controlled comparisons show that the gain does not come from generic gradient suppression: AGS outperforms a matched fixed-scale control in all five settings, demonstrating that boundary-specific scaling is more effective than applying the same average scale throughout learning.
Chinese Translation
长尾类增量学习（LT-CIL）必须在保持旧类的同时，从不平衡的数据流中学习新类。现有方法主要通过改变重放、分类器或损失函数来解决这一问题。我们研究了一个不同的因素，即在每个任务边界上特征表示应更新的强度。我们提出了NeuroGuard，一种更新控制方法，添加到基于重放的LT-CIL基线DGR中，而无需增加可学习参数。NeuroGuard保留了DGR的重放记忆、分类器和损失项集合。自适应梯度缩放（AGS）将教师的不确定性转化为任务特定的梯度缩放。置信度排名知识蒸馏重加权（CRK）为教师预测不够明确的重放样本赋予更大的知识蒸馏权重。脆弱混合熵门（FBE）将旧记忆泄漏添加到缩放决策中。在五个LT-CIL设置中，NeuroGuard在每个设置上都优于DGR。在四个主要基准比较中，它在比较方法中实现了最佳的任务无关准确率。增益扩展到旧类和新类的准确率，同时中等频率的准确率在所有五个设置中持续改善。受控比较表明，增益并非来自于通用的梯度抑制：AGS在所有五个设置中均优于匹配的固定缩放控制，证明边界特定的缩放比在学习过程中应用相同的平均缩放更有效。
cs.CV / 72 / 2608.08106
SCTD 3.0: Sonar Common Target Detection in the Wild - A Large-Scale, Multi-Scene Dataset from Real Marine Surveys
SCTD 3.0：野外声呐共同目标检测 - 来自真实海洋调查的大规模多场景数据集
Abstract
Synthetic Aperture Sonar (SAS) is core for wide-area detection of small underwater targets. However, large-scale, high-quality SAS datasets are scarce, hindering data-driven recognition. Existing benchmarks are small and limited to single scenarios, failing to reproduce complex acoustic scattering, diverse seabeds, and multi-pose imaging in real detection. To fill this gap, we introduce SCTD 3.0 - a large-scale real-measured dataset for Sonar Common Target Detection in the Wild in natural waters. It contains over 10,000 high-quality real SAS image snippets from multi-frequency systems (240 kHz, 450 kHz, and others), covering ten typical target categories across varied seabed geomorphologies, with multiple observation angles, detection ranges, and frequency bands. We establish a rigorous hierarchical annotation protocol that decouples labeling of intrinsic physical properties, deployment characteristics, and scattering phenomena - covering material, geometry, internal structure, burial state, shadow integrity, specular highlights, edge diffraction, and resonance effects. This enables fine-grained target characterization. We also construct a multi-task benchmark for object detection, fine-grained classification, and attribute prediction, evaluating mainstream deep learning models under cross-domain, cross-scene, cross-frequency, and cross-view generalization. SCTD 3.0 is expected to provide a critical data cornerstone for robust underwater target perception in open-water environments. SCTD 3.0 is available at https://github.com/automlresearch/SCTD-3.0.
Chinese Translation
合成孔径声呐（Synthetic Aperture Sonar, SAS）是广域检测小型水下目标的核心技术。然而，大规模、高质量的SAS数据集稀缺，阻碍了数据驱动的识别。现有基准数据集规模小且仅限于单一场景，无法再现复杂的声学散射、多样的海床和真实检测中的多姿态成像。为填补这一空白，我们推出了SCTD 3.0——一个用于自然水域声呐共同目标检测的大规模真实测量数据集。该数据集包含来自多频率系统（240 kHz、450 kHz等）超过10,000个高质量的真实SAS图像片段，涵盖十种典型目标类别，跨越多种海床地貌，具有多种观察角度、检测范围和频率带。我们建立了一种严格的分层注释协议，将内在物理特性、部署特征和散射现象的标注解耦——涵盖材料、几何形状、内部结构、埋藏状态、阴影完整性、镜面高光、边缘衍射和共振效应。这使得细粒度的目标特征描述成为可能。我们还构建了一个多任务基准，用于目标检测、细粒度分类和属性预测，评估主流深度学习模型在跨域、跨场景、跨频率和跨视角泛化下的表现。SCTD 3.0预计将为开放水域环境中稳健的水下目标感知提供重要的数据基础。SCTD 3.0可在https://github.com/automlresearch/SCTD-3.0获取。
cs.CV / 73 / 2608.08115
SUMI: Scalable Unified Model for 3D Point Cloud Inference
SUMI：可扩展统一模型用于3D点云推断
Abstract
Point cloud completion commonly follows a coarse-to-fine paradigm, where a low-density coarse shape is first predicted and then upsampled to the target resolution. Although recent methods have improved global structure recovery, the fine stage often remains limited by simple upsampling and insufficient interaction with coarse structural features, making local detail reconstruction challenging. We propose SUMI, a diffusion-enhanced refinement module for coarse-to-fine point cloud completion. Unlike prior diffusion-based completion methods that use diffusion as a standalone point generator, SUMI injects noisy geometric features into cross-attention with coarse structural features, enabling reverse denoising to refine local geometry while preserving global consistency. SUMI can also be integrated into existing coarse-to-fine models as a flexible refinement module. Experiments on PCN, ShapeNet-55/34, and MVP demonstrate consistent improvements over strong baselines. SUMI achieves the best overall CD and F1-score on PCN, reduces CD by up to 16.1% on ShapeNet-55, and obtains the best CD across all output densities on MVP.
Chinese Translation
点云补全通常遵循粗到细的范式，首先预测低密度的粗略形状，然后上采样到目标分辨率。尽管近期的方法改善了全局结构恢复，但细化阶段往往受到简单上采样和与粗略结构特征交互不足的限制，使得局部细节重建变得具有挑战性。我们提出了SUMI，一个用于粗到细点云补全的增强扩散细化模块。与之前将扩散作为独立点生成器的扩散基础补全方法不同，SUMI将噪声几何特征注入与粗略结构特征的交叉注意力中，使得反向去噪能够在保持全局一致性的同时细化局部几何。SUMI还可以作为灵活的细化模块集成到现有的粗到细模型中。在PCN、ShapeNet-55/34和MVP上的实验表明，相较于强基线，SUMI consistently 提升了性能。SUMI在PCN上实现了最佳的整体CD和F1-score，在ShapeNet-55上将CD降低了最多16.1%，并在MVP的所有输出密度中获得了最佳CD。
cs.CV / 74 / 2608.08125
Staying True to the Origin: Continuous Image Stylization with Smooth Transitions
忠于本源：平滑过渡的连续图像风格化
Abstract
Recent advances in generative models have achieved remarkable performance in text- and image-conditioned editing. However, preserving the content of a given image while referencing style patterns from another remains challenging, often leading to uncontrollable stylization results. In this paper, we approach image stylization from the perspective of continuous control, aiming to enable modern Diffusion Transformer (DiT)-based multi-reference editing models to (1) faithfully preserve the semantic structure of the content image, (2) render strong stylization effects, and (3) smoothly transition between the two. To this end, we propose a simple yet effective two-stage training strategy along with a style-strength-aware spline formulation. Specifically, in the first stage, the model is trained to produce strongly stylized outputs while preserving the content semantics as much as possible. In the second stage, with the base model frozen, we learn a set of anchor projectors that map various stylization strengths into the model parameter space. During inference, by performing style-strength-aware spline interpolation in a low-rank space, our method enables continuous control over stylization strength, even though the model is trained with only a few discrete strength levels. Extensive experiments demonstrate that our method supports precise and continuous manipulation of stylization strength while generating high-fidelity results with modern DiT models. Project page: https://reychiaro.github.io/StyleController.
Chinese Translation
近年来，生成模型在文本和图像条件编辑方面取得了显著的进展。然而，在保留给定图像内容的同时参考另一个图像的风格模式仍然具有挑战性，常常导致不可控的风格化结果。本文从连续控制的角度出发，探讨图像风格化，旨在使基于现代扩散变换器（Diffusion Transformer, DiT）的多参考编辑模型能够（1）忠实保留内容图像的语义结构，（2）呈现强烈的风格化效果，以及（3）在两者之间平滑过渡。为此，我们提出了一种简单而有效的两阶段训练策略，并结合了一种风格强度感知的样条（spline）公式。具体而言，在第一阶段，模型被训练以尽可能保留内容语义的同时生成强烈风格化的输出。在第二阶段，固定基础模型，我们学习一组锚定投影器，将各种风格化强度映射到模型参数空间。在推理过程中，通过在低秩空间中执行风格强度感知的样条插值，我们的方法实现了对风格化强度的连续控制，即使模型仅在少数离散强度级别上进行训练。大量实验表明，我们的方法支持对风格化强度的精确和连续操作，同时生成现代DiT模型的高保真结果。项目页面：https://reychiaro.github.io/StyleController。
cs.CV / 75 / 2608.08132
When Does An Extra View Help? Adapting Single-View 3D Reconstruction with Extra Imagery
额外视角何时有助于提升效果？利用额外图像适应单视图三维重建
Abstract
Reconstruction of 3D objects from a single image is a challenging research problem in computer vision. The key challenge is the lack of critical information from viewpoints to complete 3D structures. Using an additional view may help to resolve the issue. However, there is no mechanism that can integrate the extra view into the single-view 3D reconstruction principle. We address this challenge by proposing ASV3D, a framework for adapting single-view 3D object reconstruction to test-time data with support from one additional image. We introduce two adaptation strategies: (i) a zero-shot adaptation scheme that leverages the auxiliary image to improve the reconstruction quality of an object without retraining, and (ii) an optimised adaptation scheme that further enhances visual fidelity and cross-view consistency via contrastive learning. We apply our ASV3D to improve two state-of-the-art single-view 3D reconstruction pipelines on both benchmark and real-world datasets. Results demonstrate that our approach consistently improves reconstruction accuracy and robustness under unconstrained multi-view inputs, outperforming the baselines in both quantitative metrics and human preference. We publish our code and the real-world object dataset in our project page at https://github.com/YNhuHuynh/ASV3D/tree/main.
Chinese Translation
从单一图像重建三维物体是计算机视觉领域中的一个具有挑战性的研究问题。关键挑战在于缺乏来自不同视角的关键信息，以完成三维结构。使用额外视角可能有助于解决这一问题。然而，目前尚无机制能够将额外视角整合到单视图三维重建原理中。我们通过提出ASV3D框架来应对这一挑战，该框架旨在将单视图三维物体重建适应于测试时数据，并借助一幅额外图像提供支持。我们引入了两种适应策略：（i）一种零样本适应方案，利用辅助图像在不重新训练的情况下提高物体的重建质量；（ii）一种优化适应方案，通过对比学习进一步增强视觉真实感和跨视角一致性。我们将ASV3D应用于改进两种最先进的单视图三维重建流程，涵盖基准数据集和真实世界数据集。结果表明，我们的方法在不受约束的多视角输入下，始终提高了重建的准确性和鲁棒性，在定量指标和人类偏好方面均优于基线方法。我们在项目页面（https://github.com/YNhuHuynh/ASV3D/tree/main）上发布了我们的代码和真实世界物体数据集。
cs.CV / 76 / 2608.08135
Compositional Cross-Modality Translation via Whole-Volume Multitask Latent Flow Matching
通过全体积多任务潜在流匹配进行组合跨模态翻译
Abstract
Cross-modality medical image translation can reduce the burden of multi-modal acquisitions, yet the field remains constrained by two coupled limitations: methods operate on 2D slices or 3D patches rather than whole volumes, and train a separate model for each translation task. Both stem from a single cause, the absence of a sufficiently strong volumetric prior, which forces generative models to learn anatomical appearance and cross-modality mapping simultaneously, an ill-posed problem at the scale of available paired datasets. We propose to decouple these objectives. A large-scale pretrained 3D variational autoencoder provides a compact latent representation of volumetric appearance, reducing translation to a conditional flow-matching problem. This compression makes whole-volume processing tractable, while a resolution-aware sampling strategy preserves native anatomical scale. We train a single model jointly across inter-modality (MRI$\to$CT, CBCT$\to$CT) and intra-modality (MRI$\to$MRI) tasks over three multi-center datasets. Across all tasks, whole-volume processing outperforms its patch-based counterpart, and the multi-task model matches task-specific baselines while replacing $N$ networks with one. Crucially, joint training unlocks capabilities inaccessible to task-specific approaches: zero-shot generalization to anatomical regions unseen during training, within 0.15 SSIM of the fully supervised model, and compositional cross-dataset translation along paths never directly supervised. These results suggest that combining a strong volumetric prior with multitask training is a scalable route toward synthesis systems that generalize beyond their training distribution. Code is available at https://github.com/arco-group/Whole-Volume-Latent-FM.
Chinese Translation
跨模态医学图像翻译可以减少多模态采集的负担，但该领域仍受到两个相互关联的限制：方法在处理2D切片或3D块而非整个体积，并为每个翻译任务训练一个单独的模型。这两者的根本原因在于缺乏足够强大的体积先验，这迫使生成模型同时学习解剖外观和跨模态映射，这在现有配对数据集的规模上是一个不适定问题。我们提出将这些目标解耦。一个大规模预训练的3D变分自编码器提供了体积外观的紧凑潜在表示，将翻译简化为条件流匹配问题。这种压缩使得全体积处理变得可行，而一种分辨率感知的采样策略则保持了原生解剖尺度。我们在三个多中心数据集上联合训练一个模型，处理跨模态（MRI$ o$CT，CBCT$ o$CT）和同模态（MRI$ o$MRI）任务。在所有任务中，全体积处理的表现优于基于块的对应方法，而多任务模型在用一个网络替代$N$个网络的同时，匹配了任务特定的基线。关键是，联合训练解锁了任务特定方法无法实现的能力：在训练期间未见的解剖区域的零样本泛化，其SSIM值与完全监督模型相差仅0.15，以及沿着从未直接监督的路径进行组合跨数据集翻译。这些结果表明，将强大的体积先验与多任务训练相结合是实现超越训练分布的合成系统的可扩展途径。代码可在 https://github.com/arco-group/Whole-Volume-Latent-FM 获取。
cs.CV / 77 / 2608.08138
EFFEKT: Efficient Federated Knowledge Transfer to Foundation Models
EFFEKT：高效的联邦知识转移至基础模型
Abstract
Recent data protection laws have accelerated the adoption of Federated Learning (FL) for privacy-preserving decentralized training. Nevertheless, increasing model sizes impose substantial computational demands on client devices, limiting FL applicability in resource-constrained settings. We introduce a novel multi-domain federated learning framework in which lightweight client-side proxy models collaborate with a server-side Foundation Model (FM) to learn new concepts without sharing private data. Our approach, EFFEKT, enables efficient server-side training of domain-specific LoRA adapters while preserving feature-space alignment between the FM and proxy extractors via novel bi-directional cross-distillation strategies. Experiments on multiple real-world datasets and deployments on low-power edge devices demonstrate improvements over state-of-the-art baselines in most considered domains while maintaining lightweight computation at the client side.
Chinese Translation
近期的数据保护法律加速了联邦学习（Federated Learning, FL）在隐私保护的去中心化训练中的应用。然而，模型规模的不断扩大对客户端设备提出了巨大的计算需求，限制了FL在资源受限环境中的适用性。我们提出了一种新颖的多领域联邦学习框架，其中轻量级的客户端代理模型与服务器端的基础模型（Foundation Model, FM）协作，学习新概念而无需共享私有数据。我们的方法EFFEKT使得在服务器端高效训练领域特定的LoRA适配器成为可能，同时通过新颖的双向交叉蒸馏策略保持FM与代理提取器之间的特征空间对齐。在多个真实世界数据集上的实验以及在低功耗边缘设备上的部署表明，在大多数考虑的领域中，我们的方法在保持客户端轻量计算的同时，优于现有的最先进基准。
cs.CV / 78 / 2608.08153
Learning Structural Illumination for Unsupervised Low-light Enhancement
学习结构照明以实现无监督低光增强
Abstract
Existing unsupervised low-light image enhancement (LLIE) methods often estimate illumination directly from the entire low-light input, without separating its spatially varying illumination pattern, termed relative illumination structure, from the absolute exposure level or preventing unreliable low signal-to-noise ratio regions from biasing the estimate. Moreover, fixed exposure targets impose a scene-agnostic enhancement criterion, limiting adaptation across diverse lighting conditions. Inspired by the spatial propagation of light, we propose a Relative Illumination Structure Estimation (RISE) framework that decouples relative illumination structure from absolute exposure and infers it from reliable bright regions, enabling interpretable and robust enhancement. For scene-adaptive exposure adjustment, we further propose a Dual-Metering Exposure Reference derived from each input, allowing RISE to adapt the enhancement strength to individual scenes and generalize across diverse lighting conditions. Extensive benchmark and real-world generalization experiments show that RISE achieves state-of-the-art performance among unsupervised LLIE methods while producing visually natural results.
Chinese Translation
现有的无监督低光图像增强（LLIE）方法通常直接从整个低光输入中估计照明，而未能将其空间变化的照明模式（称为相对照明结构）与绝对曝光水平分离，或防止不可靠的低信噪比区域对估计结果产生偏差。此外，固定的曝光目标施加了一个与场景无关的增强标准，限制了在不同光照条件下的适应性。受到光的空间传播启发，我们提出了一种相对照明结构估计（RISE）框架，该框架将相对照明结构与绝对曝光解耦，并从可靠的亮区域推断，从而实现可解释且稳健的增强。为了实现场景自适应曝光调整，我们进一步提出了一种从每个输入中派生的双测光曝光参考，使RISE能够将增强强度适应于各个场景，并在不同光照条件下实现泛化。大量基准和现实世界的泛化实验表明，RISE在无监督LLIE方法中实现了最先进的性能，同时产生了视觉上自然的结果。
cs.CV / 79 / 2608.08167
Wiener Representation Filtering for VLM Hallucination Suppression
维纳表示过滤用于视觉语言模型的幻觉抑制
Abstract
Vision-language models (VLMs) excel at open-ended captioning and visual QA but often describe objects, attributes, or relations absent from the image, a phenomenon known as object hallucination. We propose a {training-free, post-hoc representation editing technique} that operates in the representation space of the language backbone. The method performs a lightweight, one-time offline calibration on a modest paired dataset to estimate the required covariance structures, using only forward passes and empirical second-order statistics with no gradient updates or fine-tuning, after which the correction is absorbed directly into the model's existing weights. By modeling hidden states as a superposition of truthful and hallucination-associated components, we derive a Wiener-type estimator whose optimal gains are given in closed form from the covariances of paired truthful and hallucinated representations. An eigendecomposition yields mode-wise attenuation that respects a stability criterion, i.e., the filter responds continuously to estimation noise. The correction is applied once to the feed-forward output projections of selected deeper layers, at inference time, the model runs unchanged and at the same speed. Experiments on LLaVA-1.5, MiniGPT-4, Gemma3, and mPLUG-Owl2 demonstrate consistent reductions in object hallucination on CHAIR, POPE, and MME while maintaining caption fluency and overall response quality. We further demonstrate the generality of our approach on the TempCompass video understanding benchmark and on discrete diffusion language models for grounded dialogue, showing that representation filtering reduces hallucinations even in temporal video reasoning and multi-step, sequence-wide denoising settings.
Chinese Translation
视觉语言模型（VLMs）在开放式字幕生成和视觉问答方面表现出色，但常常描述图像中不存在的物体、属性或关系，这种现象被称为物体幻觉。我们提出了一种{无训练的后处理表示编辑技术}，该技术在语言主干的表示空间中操作。该方法在一个适度的配对数据集上进行轻量级的一次性离线校准，以估计所需的协方差结构，仅使用前向传播和经验二阶统计，而无需梯度更新或微调，校正结果随后直接融入模型现有的权重中。通过将隐藏状态建模为真实和与幻觉相关的成分的叠加，我们推导出一种维纳型估计器，其最优增益以封闭形式给出，基于配对真实和幻觉表示的协方差。特征值分解产生了尊重稳定性标准的模式衰减，即滤波器对估计噪声的响应是连续的。校正一次应用于选定深层的前馈输出投影，在推理时，模型保持不变并以相同速度运行。在LLaVA-1.5、MiniGPT-4、Gemma3和mPLUG-Owl2上的实验表明，在保持字幕流畅性和整体响应质量的同时，CHAIR、POPE和MME上的物体幻觉显著减少。我们进一步在TempCompass视频理解基准和针对有根对话的离散扩散语言模型上展示了我们方法的普遍性，表明表示过滤在时间视频推理和多步骤、序列广泛去噪设置中也能减少幻觉。
cs.CV / 80 / 2608.08191
BAP-MOS: Bandit-Based Adaptive Prompting for Boundary-Sensitive Multi-Organ Segmentation
BAP-MOS：基于强盗算法的边界敏感多脏器分割自适应提示
Abstract
Multi-organ ultrasound segmentation remains challenging when anatomically adjacent structures must be delineated jointly, as localized boundary errors can persist even when Dice scores are high. To address these challenges, we propose Boundary-Adaptive Prompting for Multi-Organ Segmentation (BAP-MOS), a closed-loop adaptive prompting framework. BAP-MOS formulates prompt selection as an organ-specific multi-armed bandit problem over box, point, and combined prompts. An outer Tree-structured Parzen Estimator (TPE) loop selects the prompt-selection parameter vector, while an inner UCB-Tuned loop adapts per-organ prompt preferences during fine-tuning using a bounded Dice--MSD--HD95 validation-probe reward. The framework further introduces an organ-scaled negative prompt ring to adapt sparse prompt geometry across anatomical scales, while keeping the image and prompt encoders frozen and updating only the mask decoder. We evaluate BAP-MOS on pooled prostate-region TRUS cohorts against U-Net, nnU-Net, MedSAM, fixed-prompt SAM/MedSAM, and adaptive policy variants. On this benchmark, BAP-MOS achieves Dice 0.982, HD95 0.482, and MSD 0.204, reducing HD95 by approximately 48% and MSD by 45% relative to the strongest conventional baseline. To verify the generalization ability of the framework, we tested it on the external PFUS1 pelvic-floor ultrasound corpus using MedSAM and its adaptive strategy variants, and the results were good. These results support adaptive prompt allocation as an effective mechanism for improving boundary-sensitive multi-organ ultrasound segmentation without modifying the foundation-model backbone. Source Code is available at: https://github.com/SatvikPraveen/BAP-MOS
Chinese Translation
多脏器超声分割在解剖相邻结构需共同描绘时仍然面临挑战，因为即使Dice分数较高，局部边界错误仍可能存在。为了解决这些挑战，我们提出了边界自适应提示多脏器分割（BAP-MOS），这是一种闭环自适应提示框架。BAP-MOS将提示选择形式化为针对特定脏器的多臂强盗问题，涉及框、点和组合提示。外部树结构Parzen估计器（TPE）循环选择提示选择参数向量，而内部UCB-Tuned循环在微调过程中使用有界的Dice--MSD--HD95验证探测奖励来调整每个脏器的提示偏好。该框架进一步引入了一个按脏器缩放的负提示环，以适应解剖尺度上的稀疏提示几何，同时保持图像和提示编码器不变，仅更新掩膜解码器。我们在汇总的前列腺区域TRUS队列上评估了BAP-MOS，与U-Net、nnU-Net、MedSAM、固定提示SAM/MedSAM和自适应策略变体进行比较。在这一基准测试中，BAP-MOS实现了Dice 0.982、HD95 0.482和MSD 0.204，相较于最强的传统基线，HD95减少了约48%，MSD减少了45%。为了验证该框架的泛化能力，我们在外部PFUS1盆底超声语料库上使用MedSAM及其自适应策略变体进行了测试，结果良好。这些结果支持自适应提示分配作为一种有效机制，以改善边界敏感的多脏器超声分割，而无需修改基础模型骨干。源代码可在以下链接获取：https://github.com/SatvikPraveen/BAP-MOS
cs.CV / 81 / 2608.08219
VTO: Visual Tool Orchestration for Video Anomaly Detection
VTO：用于视频异常检测的视觉工具编排
Abstract
Video anomaly detection (VAD) is a critical yet challenging task due to the complex and diverse nature of real-world scenarios. Traditional deep learning approaches are fundamentally limited by poor generalization across diverse scenarios. While multimodal agents offer a promising tool-learning paradigm for VAD, current systems relying on supervised fine-tuning struggle with complex orchestration, and standard reinforcement learning often causes premature termination due to coarse-grained outcome rewards. To address these challenges, we propose VTO, a process-supervised reinforcement learning framework. Moving beyond static tool usage, VTO enables the agent to dynamically explore and interact with the environment. Specifically, we introduce a foundation model-driven cognitive evaluator to provide context-aware semantic feedback, which is seamlessly integrated into a Process-Supervised Cognitive Alignment that delivers fine-grained, step-wise supervision. By explicitly penalizing logical truncation and rewarding complete causal chains, the agent optimizes its multi-step reasoning policy for interrelated tool orchestration. To support our proposed framework, we meticulously crafted VAD-Tool, a hierarchical visual tool set comprising 12 specialized vision tools spanning from entity tracking to high-stakes hazard detection, and established the corresponding benchmark for rigorous multi-step reasoning evaluation. Extensive experiments on VAD-Tool demonstrate that VTO significantly outperforms baselines, achieving up to a 10.2\% absolute accuracy improvement in tool scheduling. Code and data are available at https://github.com/MICLAB-BUPT/VTO.
Chinese Translation
视频异常检测（VAD）是一项关键但具有挑战性的任务，因其涉及复杂多样的现实场景。传统的深度学习方法在不同场景下的泛化能力有限。尽管多模态智能体为VAD提供了一种有前景的工具学习范式，但当前依赖于监督微调的系统在复杂编排方面面临困难，而标准的强化学习往往因粗粒度的结果奖励导致过早终止。为了解决这些挑战，我们提出了VTO，一种过程监督的强化学习框架。VTO超越了静态工具使用，使智能体能够动态探索和与环境互动。具体而言，我们引入了一个基于基础模型的认知评估器，以提供上下文感知的语义反馈，并将其无缝集成到过程监督的认知对齐中，从而提供细粒度的逐步监督。通过明确惩罚逻辑截断并奖励完整的因果链，智能体优化其多步骤推理策略以实现相互关联的工具编排。为了支持我们提出的框架，我们精心构建了VAD-Tool，一个包含12个专业视觉工具的层次化视觉工具集，涵盖从实体跟踪到高风险危害检测，并建立了相应的基准以进行严格的多步骤推理评估。在VAD-Tool上的广泛实验表明，VTO显著优于基线，在工具调度中实现了高达10.2%的绝对准确率提升。代码和数据可在https://github.com/MICLAB-BUPT/VTO获取。
cs.CV / 82 / 2608.08285
Ego-OSCAR: Egocentric Open source Stereo CAptuRe System
Ego-OSCAR：以自我为中心的开源立体捕捉系统
Abstract
We present Ego-OSCAR, an open-hardware, low-cost, head-mounted stereo-inertial capture device for egocentric data collection in the wild. EgoOSCAR pairs a hardware-synchronized global-shutter stereo camera with a 6- axis IMU, an embedded Linux SBC for on-device video encoding, and a realtime microcontroller for user feedback and watchdog functions. The complete bill of materials is under USD 200 per unit, using only commercially available components and 3D-printed parts. Alongside the device, we release a complete software stack (hardware-accelerated recording pipeline, IMU sampling daemon, time-synchronization tooling, and watchdog firmware) and roughly 550 hours of egocentric stereo video per camera with synchronized IMU, collected by a distributed contributor network across everyday indoor environments. The release is annotated rather than raw: free-form action captions cover essentially the entire recorded timeline with an open vocabulary, and per-frame 3D hand reconstructions ship alongside per-session stereo calibration. Ego-OSCAR does not aim to match the per-unit fidelity of research-grade systems such as Project Aria; it aims to be the cheapest defensible substrate for crowdsourced egocentric capture, and to lower the activation energy for any team that wants to collect egocentric data at scale. All hardware designs, software, and the dataset are open-sourced
Chinese Translation
我们提出了Ego-OSCAR，这是一种开源硬件、低成本的头戴式立体惯性捕捉设备，旨在野外进行以自我为中心的数据收集。Ego-OSCAR将硬件同步的全球快门立体相机与6轴惯性测量单元（IMU）、用于设备内视频编码的嵌入式Linux单板计算机（SBC）以及用于用户反馈和看门狗功能的实时微控制器相结合。每个设备的完整材料清单成本低于200美元，仅使用市售组件和3D打印部件。除了设备，我们还发布了完整的软件栈（硬件加速的录制管道、IMU采样守护进程、时间同步工具和看门狗固件），以及每个相机约550小时的以自我为中心的立体视频，配有同步的IMU，这些视频由分布式贡献者网络在日常室内环境中收集。发布的内容是经过注释的，而非原始数据：自由形式的动作字幕几乎覆盖了整个录制时间线，使用开放词汇，并且每帧的3D手部重建与每次会话的立体校准一起提供。Ego-OSCAR并不旨在与如Project Aria等研究级系统的单元保真度相匹配；它的目标是成为众包以自我为中心捕捉的最便宜的可辩护基础，并降低任何希望大规模收集以自我为中心数据的团队的启动成本。所有硬件设计、软件和数据集均为开源。
cs.CV / 83 / 2608.08287
What Irregularity Costs: CUDA C++, Rust, and Triton on a Hash-Blocked GPU Workload
不规则性成本：CUDA C++、Rust 和 Triton 在哈希阻塞 GPU 工作负载上的比较
Abstract
GPU language comparisons are almost always run on tiled dense linear algebra, where every toolchain is good and the differences are small. We implement the same hash-blocked TSDF fusion kernel in CUDA C++, in Rust through NVIDIA's cuda-oxide, and in Triton, and measure it on a workload with the opposite character: an open-addressed hash table with compare-exchange insertion, data-dependent per-lane probe depth, and contended scatter. The result is a split. On the regular stage, which walks a truncation band and accumulates, all three languages land within a small factor of each other. On the irregular stage, which probes and inserts, Rust stays close to hand-written CUDA C++ while Triton is more than an order of magnitude slower. Language choice is nearly free on the work that is usually benchmarked and expensive on the work that is not. We attribute both gaps to specific things the languages cannot express, not to ratios. Triton's cost follows from a probe loop that must run to a compile-time bound and from tl.atomic_cas taking no mask, which forces a scratch structure with no counterpart in CUDA. Rust's cost was invisible in every instruction count: its kernel issues fewer instructions, fewer compare-exchanges and fewer registers at identical occupancy, yet was slower. Hardware counters located it in L1 residency. A GPU-scope atomic load must be coherent across SMs, no NVIDIA L1 is, so the type-correct way to read a shared location bypasses the cache on every access. Triton's bounded probe is also a correctness problem for fusion: at load factors an ordinary depth trajectory reaches, it silently discards blocks and the reconstruction loses patches of surface with nothing reported. We also report a defect found and fixed in cuda-oxide itself, now merged upstream: its scoped atomic load and store could not be called at all in the build mode that produces real kernels.
Chinese Translation
GPU 语言比较几乎总是在平铺的稠密线性代数上进行，在这种情况下，每个工具链表现良好，差异很小。我们在 CUDA C++、通过 NVIDIA 的 cuda-oxide 实现的 Rust 和 Triton 中实现相同的哈希阻塞 TSDF 融合内核，并在具有相反特征的工作负载上进行测量：一个具有比较-交换插入、数据依赖的每条通道探测深度和竞争散射的开放寻址哈希表。结果是一个分裂。在常规阶段，遍历截断带并进行累积，所有三种语言的性能相差不大。在不规则阶段，进行探测和插入时，Rust 的性能接近手写的 CUDA C++，而 Triton 的性能则慢了一个数量级以上。对于通常被基准测试的工作，语言选择几乎是免费的，而对于那些不被基准测试的工作则是昂贵的。我们将这两个差距归因于语言无法表达的特定内容，而不是比率。Triton 的成本源于必须运行到编译时界限的探测循环，以及 tl.atomic_cas 不使用掩码，这迫使其使用一个在 CUDA 中没有对应物的临时结构。Rust 的成本在每条指令计数中是不可见的：其内核发出更少的指令、更少的比较-交换和更少的寄存器，同时占用率相同，但速度却更慢。硬件计数器将其定位在 L1 居留。GPU 范围内的原子加载必须在 SM 之间保持一致，而没有任何 NVIDIA L1 是一致的，因此以类型正确的方式读取共享位置在每次访问时都会绕过缓存。Triton 的有界探测在融合中也是一个正确性问题：在普通深度轨迹达到的负载因子下，它默默地丢弃块，重建失去了表面的一部分而没有任何报告。我们还报告了在 cuda-oxide 本身发现并修复的缺陷，现在已合并到上游：其作用域原子加载和存储在生成真实内核的构建模式下根本无法调用。
cs.CV / 84 / 2608.08290
Test-Time Prototype Adaptation for Open-Vocabulary Semantic Segmentation
测试时原型适应用于开放词汇语义分割
Abstract
Open-vocabulary semantic segmentation (OVSS) repurposes a pretrained CLIP encoder for dense prediction without additional labeled supervision. Existing methods improve CLIP's spatial behavior either by redesigning its internal attention or by injecting features from auxiliary vision foundation models; both require access to the host's internal computation and are tailored to its specific forward pass. In this work, we propose Test-time Prototype Adaptation (TPA), a training-free plug-in that operates at the output level, leaving the host's forward pass and weights unmodified. By leveraging a lightweight transductive adaptation phase, TPA identifies confident anchor patches from the host's own output predictions on a small pool of unlabeled deployment-domain images, and aggregates their frozen DINO features into per-class prototypes; at inference, a single cosine similarity lookup against this frozen bank provides an auxiliary score fused linearly with the host's logits. TPA composes with five representative OVSS hosts spanning attention-redesign and VFM-injection designs, across three CLIP backbones, eight benchmarks, and multiple internal VFM choices. Under a single set of hyper-parameters and without per-host tuning or parameter updates, TPA consistently improves segmentation accuracy, with as few as approximately 10% of unlabeled deployment-domain images sufficing for effective bank construction on most benchmarks.
Chinese Translation
开放词汇语义分割（OVSS）利用预训练的 CLIP 编码器进行密集预测，而无需额外的标注监督。现有方法通过重新设计 CLIP 的内部注意力或注入来自辅助视觉基础模型的特征来改善 CLIP 的空间行为；这两种方法都需要访问主机的内部计算，并且针对其特定的前向传播进行定制。在本研究中，我们提出了测试时原型适应（TPA），这是一种无训练的插件，作用于输出层，保持主机的前向传播和权重不变。通过利用轻量级的传导适应阶段，TPA 从主机在一小组未标注的部署域图像上的输出预测中识别出可信的锚点补丁，并将其冻结的 DINO 特征聚合成每类原型；在推理时，对这个冻结库进行单次余弦相似度查找提供了一个辅助分数，该分数与主机的 logits 线性融合。TPA 可以与五个代表性的 OVSS 主机组合，涵盖注意力重新设计和 VFM 注入设计，涉及三个 CLIP 主干、八个基准和多个内部 VFM 选择。在一组超参数下，无需针对每个主机进行调优或参数更新，TPA 一致地提高了分割精度，对于大多数基准，约 10% 的未标注部署域图像即可有效构建银行。
cs.CV / 85 / 2608.08307
Frequency-Domain Dual-Branch Fusion for Medical Visual Question Answering
频域双分支融合用于医学视觉问答
Abstract
Medical Visual Question Answering (VQA) requires aligning subtle visual evidence, including lesion texture, boundary sharpness, and diffuse density changes, with clinical language. Existing multimodal fusion approaches operating in the spatial domain may not fully exploit complementary frequency information present in visual and textual representations. We introduce a dual-branch frequency-domain fusion module that conditions spectral filtering on the input question, enabling adaptive selection of global low-frequency structure and fine-grained high-frequency detail before reconstructing the spatial representation for answer generation. To provide a richer spectrum for filtering, we extract complementary features from early texture-sensitive and final semantic layers of a frozen BiomedCLIP encoder and align both with the question representation using a symmetric InfoNCE objective prior to staged joint training with a BioBART decoder. We pretrain the proposed model on PMC-VQA and fine-tune it on the VQA-RAD and SLAKE benchmarks, demonstrating that frequency-aware multimodal fusion improves medical VQA performance while maintaining a lightweight and efficient architecture.
Chinese Translation
医学视觉问答（VQA）需要将细微的视觉证据（包括病变纹理、边界清晰度和弥漫密度变化）与临床语言对齐。现有的在空间域操作的多模态融合方法可能无法充分利用视觉和文本表示中存在的互补频率信息。我们提出了一种双分支频域融合模块，该模块根据输入问题对谱滤波进行条件设置，使得在重建空间表示以生成答案之前，能够自适应选择全局低频结构和细粒度高频细节。为了提供更丰富的滤波谱，我们从冻结的BiomedCLIP编码器的早期纹理敏感层和最终语义层提取互补特征，并使用对称的InfoNCE目标将两者与问题表示对齐，然后与BioBART解码器进行分阶段联合训练。我们在PMC-VQA上预训练了所提出的模型，并在VQA-RAD和SLAKE基准上进行了微调，证明了频率感知的多模态融合在提高医学VQA性能的同时，保持了轻量高效的架构。
cs.CV / 86 / 2608.08308
Open-World Semantic Segmentation with Sensitivity Modeling
基于敏感性建模的开放世界语义分割
Abstract
Modern vision systems must operate in "open-world" settings, where models must recognize known categories and detect unseen or anomalous content. Conventional semantic segmentation models operate under a "closed-world" assumption, often producing overconfident misclassifications on novel content. We address open-world semantic segmentation, the joint task of segmenting known classes while detecting and grouping novel or anomalous content without additional supervision, by extending a dual-decoder baseline with a third, complementary decoder within a unified encoder-decoder design. The first decoder performs closed-set segmentation using Gaussian prototypes for known categories. The second uses contrastive feature learning to isolate unknown regions in embedding space. The third, our key contribution, is a sensitivity decoder that captures fine-grained texture irregularities and activation instabilities indicative of semantic uncertainty, which neither semantic prototypes nor contrastive norms can reliably detect. The three decoders provide genuinely complementary signals: class-level OOD distance in logit space, global feature energy in embedding space, and local activation instability across encoder scales. Experiments on Cityscapes and BDD-Anomaly show that our method improves anomaly segmentation and novel-class discovery while maintaining competitive closed-set accuracy, with gains of +2.4% AUROC and a 2.5 pp. reduction in FPR@95TPR on BDD-Anomaly over the baseline.
Chinese Translation
现代视觉系统必须在“开放世界”环境中运行，其中模型必须识别已知类别并检测未见或异常内容。传统的语义分割模型在“封闭世界”假设下工作，往往对新内容产生过于自信的错误分类。我们通过在统一的编码器-解码器设计中扩展一个双解码器基线，提出了开放世界语义分割的解决方案，该任务是对已知类别进行分割，同时在没有额外监督的情况下检测和分组新颖或异常内容。第一个解码器使用高斯原型对已知类别进行封闭集分割。第二个解码器利用对比特征学习在嵌入空间中隔离未知区域。第三个解码器是我们的关键贡献，它是一个敏感性解码器，捕捉细粒度的纹理不规则性和激活不稳定性，这些特征表明语义不确定性，而语义原型和对比范数无法可靠检测。三个解码器提供真正互补的信号：在对数空间中的类别级OOD距离、在嵌入空间中的全局特征能量，以及在编码器尺度上的局部激活不稳定性。在Cityscapes和BDD-Anomaly上的实验表明，我们的方法在提高异常分割和新类别发现的同时，保持了竞争性的封闭集准确率，在BDD-Anomaly上相较于基线提高了+2.4%的AUROC，并减少了2.5个百分点的FPR@95TPR。
cs.CV / 87 / 2608.08309
Three Necessary Principles for Self-Supervised Visual Representation Learning
自监督视觉表征学习的三个必要原则
Abstract
We argue that learning visual representations without labels requires a training signal jointly complete across three non-overlapping objectives: semantic invariance across augmented views, patch-level spatial prediction, and representational non-degeneracy. We formalize these as the observation, prediction, and regularization principles and prove (i) that combining observation and prediction without regularization admits the constant encoder as a global minimizer under negative-free alignment; (ii) that the two objectives are gradient-complementary and structurally non-conflicting at the encoder output; and (iii) that the momentum encoder converges to the same fixed point as the online encoder and provides no collapse guarantee at convergence. Contrastive alignment provides only self-limiting collapse resistance, formalized via an explicit gradient-decay argument. Dropping prediction withholds the spatial training signal by construction; dropping observation forfeits cross-view semantic invariance by construction; at the scale we study, no pair substitutes for the third. Every major self-supervised method is a special case of a single unified energy decomposition. We pair every theoretical claim with a controlled experiment, including a patch-retrieval evaluation for the spatial consequence of prediction.
Chinese Translation
我们认为，在没有标签的情况下学习视觉表征需要一个在三个不重叠目标上共同完整的训练信号：增强视图之间的语义不变性、补丁级空间预测和表征非退化性。我们将这些形式化为观察原则、预测原则和正则化原则，并证明了(i) 在没有正则化的情况下，结合观察和预测会使常量编码器在无负对齐下成为全局最小化器；(ii) 这两个目标在编码器输出处是梯度互补且结构上不冲突的；(iii) 动量编码器收敛到与在线编码器相同的固定点，并在收敛时不提供崩溃保证。对比对齐仅提供自我限制的崩溃抵抗，这通过明确的梯度衰减论证进行了形式化。去掉预测会在构造上抑制空间训练信号；去掉观察会在构造上放弃跨视图的语义不变性；在我们研究的规模上，没有任何一对可以替代第三个目标。每种主要的自监督方法都是单一统一能量分解的特例。我们将每个理论主张与受控实验配对，包括针对预测的空间后果的补丁检索评估。
cs.CV / 88 / 2608.08315
Your VLM Already Knows When: Training-Free Temporal Grounding by Asking Yes or No
你的视觉语言模型已经知道何时发生：通过问是或否实现无训练的时间定位
Abstract
Multimodal LLMs that recognise events reliably still fail to say when they happen. Prompted for timestamps, strong VLMs reach as little as $3.8\%$
[email protected] on Charades-STA, and $77$ to $80\%$ of their wrong predictions carry low output entropy: the models are confidently wrong, and entropy-based error detection stays below a random classifier. We show that this failure lives in the task interface, not in perception. Holding the weights fixed, replacing timestamp regression with a coarse-to-fine scan of binary questions, whose first-token probabilities are consumed only as a ranking, raises
[email protected] by $28$ to $50$ points across four frozen backbones. The residual failures decompose into two measurable axes: a perception axis that moves with the backbone, and a geometry axis that is analytically predictable from the ratio of the output-window and event widths. FV-Action, the training-free method built on this analysis, reaches $56.8\%$
[email protected] on Charades-STA, above the same backbone's native grounding pipeline and the strongest training-free result on this benchmark; it surpasses every TVG-trained model evaluated zero-shot on TACoS, and improves over direct prediction on ActivityNet Captions and QVHighlights, with no temporal supervision at any stage.
Chinese Translation
能够可靠识别事件的多模态大语言模型（LLMs）仍然无法准确判断事件发生的时间。在被要求提供时间戳时，强大的视觉语言模型（VLMs）在 Charades-STA 数据集上的
[email protected] 仅达到 $3.8\%$，而 $77 ext{ 到 }80 ext{ ext{%}}$ 的错误预测具有低输出熵：这些模型自信地做出错误判断，基于熵的错误检测表现低于随机分类器。我们展示了这种失败源于任务接口，而非感知能力。保持权重不变，用粗到细的二元问题扫描替代时间戳回归，其首个标记的概率仅作为排序依据，使得
[email protected] 在四个冻结的主干网络上提升了 $28$ 到 $50$ 个百分点。剩余的失败可分解为两个可测量的轴：一个与主干网络一起移动的感知轴，以及一个可以从输出窗口与事件宽度的比率中解析预测的几何轴。基于这一分析构建的无训练方法 FV-Action 在 Charades-STA 上达到了 $56.8 ext{ ext{%}}$ 的
[email protected]，超越了同一主干网络的原生定位管道，并在该基准测试中取得了最强的无训练结果；它在 TACoS 上超越了每个零-shot 评估的 TVG 训练模型，并在 ActivityNet Captions 和 QVHighlights 上改善了直接预测，且在任何阶段均未使用时间监督。
cs.CV / 89 / 2608.08319
A continually expandable foundation model for brain MRI
一种可持续扩展的脑部MRI基础模型
Abstract
Brain magnetic resonance imaging (MRI) is central to neuroscience and clinical assessment, but models are commonly developed for individual diseases, populations or imaging protocols. Foundation models promise more general representations, yet they are usually pretrained once and can lose earlier capabilities when updated with new data. Here we show that Alcmaeon, a three-dimensional brain MRI foundation model pretrained without manual labels on more than 425,000 volumes and derived imaging maps, can be expanded sequentially across clinical domains. Alcmaeon combines volumetric encoding and latent diffusion generation with Graph-Blueprint Pruning (GBP), which protects network modules important to earlier domains while leaving the remaining capacity trainable. Across expansion from healthy ageing and neurodegeneration to developmental, psychiatric and tumour imaging, GBP showed less forgetting than sequential adaptation and elastic weight consolidation across voxel-level reconstruction measures, with its largest advantage after adaptation to tumour imaging. The blueprints provided an inspectable record of how model capacity was protected and reused. Representations from different model levels supported image synthesis, disease classification, survival modelling and postoperative prediction, although no single representation was optimal for every task. These findings provide a route towards brain MRI foundation models that can grow with emerging data while retaining earlier capabilities.
Chinese Translation
脑部磁共振成像（MRI）在神经科学和临床评估中至关重要，但模型通常是针对单一疾病、特定人群或成像协议开发的。基础模型承诺提供更通用的表征，然而它们通常只进行一次预训练，并且在用新数据更新时可能会丧失早期的能力。在此，我们展示了Alcmaeon，一个在超过425,000个体积和衍生成像图谱上无手动标签预训练的三维脑部MRI基础模型，可以在临床领域中进行顺序扩展。Alcmaeon结合了体积编码和潜在扩散生成，并采用图蓝图修剪（Graph-Blueprint Pruning, GBP），该方法保护了对早期领域重要的网络模块，同时保留了其余容量可训练。在从健康老龄化和神经退行性变到发育、精神病和肿瘤成像的扩展过程中，GBP在体素级重建度量上表现出比顺序适应和弹性权重整合更少的遗忘，其在适应肿瘤成像后展现出最大的优势。蓝图提供了一个可检查的记录，展示了模型容量如何被保护和再利用。来自不同模型层次的表征支持图像合成、疾病分类、生存模型和术后预测，尽管没有单一表征在每个任务中都是最优的。这些发现为脑部MRI基础模型的发展提供了一条路径，使其能够随着新兴数据的增长而保持早期能力。
cs.CV / 90 / 2608.08336
Circuit Fine-Tuning for Compute-Efficient Transformer Adaptation
计算高效的变换器适应电路微调
Abstract
Parameter-Efficient Fine-Tuning (PEFT) has become the de facto standard for adapting Vision Transformers (ViTs) to downstream tasks. While parameter count has been the dominant efficiency metric in PEFT, it does not imply \textit{compute efficiency}: parameter-sparse methods can still incur full-model training cost per step, and typically need long schedules to reach peak accuracy. We introduce Circuit Fine-Tuning (CFT), a compute-efficient framework that uses circuit discovery---conventionally used to explain trained models---to select modules for fine-tuning before training. Whereas attribution is conventionally formulated against a trained task head, we formulate it against a near-zero-initialized probe head, which isolates the response of the backbone to the target distribution rather than the preferences of a particular classifier. CFT then fine-tunes only the recovered subgraph. CFT needs no learning-rate warmup and reaches peak accuracy in ${\sim}20$ epochs on average---versus $44$--$96$ for strong PEFT baselines---yielding $2.3$--$6.6\times$ fewer training FLOPs and up to $16\times$ less wall-clock time, while adding zero parameters and no inference operations. Experiments across a standard visual transfer benchmark (VTAB-1k), hierarchical backbones (Swin), domain-shifted medical imaging (CBIS-DDSM), and a vision-language model (Gemma-3 on CUB-200) demonstrate the effectiveness of CFT. Code is available at https://github.com/UriKialy/CFT
Chinese Translation
参数高效微调（PEFT）已成为将视觉变换器（ViTs）适应于下游任务的事实标准。尽管参数数量一直是PEFT中的主要效率指标，但这并不意味着 extit{计算效率}：参数稀疏方法仍然可能在每一步中产生全模型训练成本，并且通常需要较长的训练周期才能达到峰值准确率。我们提出了电路微调（CFT），这是一种计算高效的框架，利用电路发现——传统上用于解释训练模型——在训练之前选择微调模块。与传统上针对训练任务头进行的归因不同，我们将其针对近零初始化的探测头进行归因，这样可以隔离主干网络对目标分布的响应，而不是特定分类器的偏好。CFT随后仅微调恢复的子图。CFT不需要学习率预热，平均在约${ ilde{20}}$个周期内达到峰值准确率——而强PEFT基线通常需要$44$到$96$个周期——从而减少$2.3$到$6.6$倍的训练FLOPs，并节省高达$16$倍的实际时间，同时不增加任何参数和推理操作。在标准视觉迁移基准（VTAB-1k）、分层主干（Swin）、领域转移医学成像（CBIS-DDSM）以及视觉-语言模型（Gemma-3在CUB-200）上的实验展示了CFT的有效性。代码可在https://github.com/UriKialy/CFT获取。
cs.CV / 91 / 2608.08354
Tropical Cyclone Forecasting via Latent Rectified Flow using Satellite Imagery and Atmospheric Fields
通过潜在修正流利用卫星图像和大气场进行热带气旋预测
Abstract
Tropical cyclones are growing more destructive in a changing climate, and efficient forecasting of their structure and track has become a necessity. Deep generative models promise an alternative to computationally expensive numerical weather prediction (NWP), yet current systems produce either satellite imagery or atmospheric fields, never both; they need many sampling steps, putting them out of reach of modest hardware; and their storm tracks come from regression heads with no physical link to the generated atmosphere. This work presents a single-pass model that jointly forecasts GRIDSAT-B1 infrared imagery and four ERA5 atmospheric fields (U-wind, V-wind, air temperature, and surface pressure) out to nine hours. A five-channel variational autoencoder compresses each 5 x 256 x 256 frame to a 4 x 64 x 64 latent, and a conditional rectified-flow UNet with a factorized temporal-attention module predicts the next three frames from three past frames, their best-track coordinates, and timestamps. The model is then reward-fine-tuned (DRaFT) against a differentiable track error derived from the predicted winds through a steering-flow calculation. On held-out 2022 storms the model reaches 16.35 dB PSNR and 0.759 SSIM, ahead of a reproduced cascaded-diffusion baseline at every lead time (+0.84 dB at +9 h) while sampling ~30x faster (56 ms vs. 1673 ms). Track error at +9 h is 62.4 km, 15% below the baseline, and a reward fine-tuning study demonstrates a further 8-11% track-error reduction across sampler budgets.
Chinese Translation
在气候变化的背景下，热带气旋的破坏力日益增强，因此高效预测其结构和路径已成为一种必要性。深度生成模型为计算成本高昂的数值天气预报（NWP）提供了替代方案，但当前系统只能生成卫星图像或大气场，而无法同时生成；它们需要多个采样步骤，使得在普通硬件上难以实现；并且它们的风暴轨迹来自于与生成的大气没有物理联系的回归头。本研究提出了一种单次预测模型，能够联合预测GRIDSAT-B1红外图像和四个ERA5大气场（U风、V风、空气温度和地面压力），预测时间可达九小时。一个五通道变分自编码器将每个5 x 256 x 256的帧压缩为4 x 64 x 64的潜在表示，而一个带有分解时间注意力模块的条件修正流UNet则根据三个过去的帧、最佳轨迹坐标和时间戳预测下三个帧。然后，该模型通过基于预测风速的引导流计算得出的可微分轨迹误差进行奖励微调（DRaFT）。在2022年未见风暴的测试中，该模型在每个预测时刻均达到了16.35 dB的PSNR和0.759的SSIM，超越了每个预测时刻的重现级联扩散基线（在+9小时时提高了0.84 dB），同时采样速度约快30倍（56毫秒对比1673毫秒）。在+9小时时的轨迹误差为62.4公里，低于基线的15%，而奖励微调研究表明在采样预算下轨迹误差进一步减少了8-11%。
cs.CV / 92 / 2608.08366
VOICE: A Vision-Omics Foundation Model Integrating Direct and Retrieval-Based Prediction of In-situ Single-Cell Gene Expression
VOICE：一个整合直接预测与基于检索的原位单细胞基因表达的视觉组学基础模型
Abstract
Spatial transcriptomics can resolve gene expression at single-cell resolution, but it is costly, limited to targeted panels of a few hundred to a few thousand genes, and applicable to only a small number of samples. H&E imaging, by contrast, is cheap and collected routinely at scale. This makes predicting single-cell expression directly from morphology a practical way to bring molecular analysis to large tissue archives. We therefore present VOICE, a multimodal foundation model that predicts single-cell gene expression from H&E images using paired Xenium data. VOICE first aligns cell centered H&E morphology from a pathology foundation model with single-cell expression embeddings from a transcriptome foundation model, trained using contrastive learning over 23 million cells. Next it predicts expression through two branches. One branch directly regresses expression from morphology. The other branch retrieves measured expression from similar reference cells, recovering genes that do not have morphological signal. Because genes vary in morphological predictability, VOICE fuses the two branches with a per-gene weight. After training, VOICE generalizes to heldout patients, slides, and partially overlapping gene panels from Xenium, and it consistently outperforms prior single-cell expression prediction methods on seven metrics.
Chinese Translation
空间转录组学能够以单细胞分辨率解析基因表达，但其成本高昂，受限于几百到几千个基因的靶向面板，并且仅适用于少量样本。相比之下，H&E成像成本低廉，且能够在大规模中常规收集。这使得从形态学直接预测单细胞表达成为将分子分析引入大型组织档案的实用方法。因此，我们提出了VOICE，一个多模态基础模型，通过使用配对的Xenium数据，从H&E图像中预测单细胞基因表达。VOICE首先将来自病理基础模型的细胞中心H&E形态与来自转录组基础模型的单细胞表达嵌入对齐，该模型通过对2300万个细胞进行对比学习进行训练。接下来，它通过两个分支进行表达预测。一个分支直接从形态学回归表达，另一个分支从相似的参考细胞中检索测量的表达，恢复没有形态信号的基因。由于基因在形态学可预测性上存在差异，VOICE通过每个基因的权重融合这两个分支。经过训练后，VOICE能够推广到未见患者、切片和部分重叠的Xenium基因面板，并在七个指标上持续超越先前的单细胞表达预测方法。
cs.CV / 93 / 2608.08368
PARAGraph: Pathology-Anatomy-Aware Hierarchical Graph for Diabetic Retinopathy Grading
PARAGraph：一种病理-解剖学感知的分层图用于糖尿病视网膜病变分级
Abstract
Diabetic retinopathy (DR) remains a leading cause of vision loss among working-age adults worldwide, making reliable severity grading clinically important. Despite strong performance, most deep models formulate DR grading as image-level classification and do not explicitly model clinically grounded evidence, such as lesion types and spatial relations. In this paper, we propose PARAGraph, a Pathology-Anatomy-Aware Hierarchical Graph framework for DR grading. PARAGraph represents each image as a three-level hierarchical graph with lesion-level nodes, intermediate category and region nodes, and global anatomical and semantic nodes. To incorporate medical priors into nodes, we construct an optic disc-fovea-anchored coordinate frame that provides a scale- and rotation-normalized retinal reference system. Within this frame, lesion nodes are encoded with category, normalized area, and anatomical coordinates. To mitigate noisy lesion segmentation, PARAGraph uses a dual-fusion strategy that introduces global visual context into a graph semantic node and a decision-level prediction branch, improving robustness when lesion evidence is unreliable. Extensive experiments on Messidor-2, APTOS, and DDR show that PARAGraph achieves consistent DR grading performance over state-of-the-art methods. Interpretability and robustness analyses further demonstrate that its predictions are clinically grounded, closely associated with lesion evidence and robust to lesion segmentation noise.
Chinese Translation
糖尿病视网膜病变（DR）仍然是全球工作年龄成年人失明的主要原因之一，因此可靠的严重程度分级在临床上具有重要意义。尽管大多数深度模型在性能上表现良好，但它们将DR分级视为图像级分类，并未明确建模临床基础证据，如病变类型和空间关系。本文提出了PARAGraph，一种用于DR分级的病理-解剖学感知的分层图框架。PARAGraph将每幅图像表示为一个三层分层图，包含病变级节点、中间类别和区域节点，以及全局解剖和语义节点。为了将医学先验知识融入节点中，我们构建了一个以视盘-中心凹为锚的坐标框架，提供了一个尺度和旋转归一化的视网膜参考系统。在该框架内，病变节点被编码为类别、归一化面积和解剖坐标。为了减轻噪声病变分割的影响，PARAGraph采用了一种双重融合策略，将全局视觉上下文引入图的语义节点和决策级预测分支，从而在病变证据不可靠时提高鲁棒性。在Messidor-2、APTOS和DDR上的大量实验表明，PARAGraph在DR分级性能上超越了最先进的方法。可解释性和鲁棒性分析进一步证明其预测是基于临床的，紧密关联于病变证据，并对病变分割噪声具有鲁棒性。
cs.CV / 94 / 2608.08374
Gated Spatial Redundancy Projection for Pathology Transformer Attentions
病理变换器注意力的门控空间冗余投影
Abstract
Transformer models are increasingly used for whole-slide image analysis in computational pathology. Yet, WSIs differ fundamentally from natural images: neighbouring patches often contain highly similar tissue type, stain, texture, and cellular composition. We identify this local spatial redundancy as a pathology-specific failure mode of self-attention, where dominant neighbourhood features can be repeatedly mixed into patch-tokens and weaken subtle diagnostic or prognostic deviations. We propose Gated Spatial Redundancy Projection (Gated SRP), a lightweight drop-in correction module for self-attention layers. For each patch token and attention head, Gated SRP estimates a local redundancy axis from neighbouring value vectors, projects the attention output onto this axis, and applies a learned signed gate to correct the redundancy-aligned component geometrically. Across five TCGA survival cohorts, Gated SRP obtains the highest mean C-index among the compared attention variants in all cohorts, with an average improvement over the base attention, while adding only +0.02% parameters. Across five slide-level classification datasets, it improves the base attention on 12 of 16 reported metrics and achieves the best AUC on three datasets. Code is publicly available at https://github.com/AtlasAnalyticsLab/GatedSRP.
Chinese Translation
变换器模型在计算病理学的全切片图像分析中越来越多地被使用。然而，全切片图像（WSI）在本质上与自然图像有着根本的不同：相邻的图块通常包含高度相似的组织类型、染色、纹理和细胞组成。我们将这种局部空间冗余识别为自注意力的一种病理特异性失效模式，其中主导的邻域特征可能会反复混入图块标记中，从而削弱细微的诊断或预后偏差。我们提出了门控空间冗余投影（Gated SRP），这是一个轻量级的自注意力层修正模块。对于每个图块标记和注意力头，Gated SRP 从相邻的值向量中估计一个局部冗余轴，将注意力输出投影到该轴上，并应用一个学习到的有符号门来在几何上修正与冗余对齐的分量。在五个 TCGA 生存队列中，Gated SRP 在所有比较的注意力变体中获得了最高的平均 C 指数，相较于基础注意力平均提高，同时仅增加了 +0.02% 的参数。在五个切片级分类数据集上，它在 16 个报告指标中的 12 个上改善了基础注意力，并在三个数据集上达到了最佳的 AUC。代码已公开发布在 https://github.com/AtlasAnalyticsLab/GatedSRP。
cs.CV / 95 / 2608.08381
DoRF++: Spherical Representation Learning over Doppler Radiance Fields for Robust Wi-Fi Sensing
DoRF++：基于多普勒辐射场的球面表示学习用于鲁棒的Wi-Fi感知
Abstract
Motivated by the IEEE 802.11bf effort to standardize advanced WLAN sensing, interest in Wi-Fi Channel State Information (CSI) for passive, device-free, and privacy-preserving activity and gesture recognition has grown rapidly. Recent studies have shown that Doppler velocity projections extracted from CSI, which directly reflect human-motion velocity, enable more robust human activity recognition (HAR) and stronger generalization across users and unseen conditions. Nevertheless, reliable generalization under real-world variability remains a major challenge, hindering the adoption of Wi-Fi sensing in real-world applications. To address this challenge, we introduce Doppler Radiance Fields (DoRF), bringing the concept of neural radiance fields (NeRF) from computer vision into Wi-Fi sensing. DoRF models Doppler velocity projections extracted from Wi-Fi CSI as sparse and diverse virtual-camera views of human motion. It then infers a latent 3D motion sequence whose projections along learned effective Doppler directions explain the CSI-derived Doppler observations. The recovered motion is subsequently projected onto an equiangular grid of directions on the unit sphere, producing a spherical representation of the underlying motion. Since DoRF naturally defines the Doppler representation on spheres, we further introduce DoRF++, a spherical-learning design that applies spherical Transformers for activity classification. Experiments on our collected hand-gesture dataset show that DoRF++ significantly outperforms state-of-the-art Wi-Fi-based HAR methods in cross-user generalization accuracy, especially for difficult gestures in settings with a single multi-antenna receiver access point (AP).
Chinese Translation
受IEEE 802.11bf标准化先进WLAN感知的启发，Wi-Fi信道状态信息（CSI）在被动、无设备和隐私保护的活动与手势识别中的应用兴趣迅速增长。近期研究表明，从CSI中提取的多普勒速度投影直接反映人类运动速度，能够实现更鲁棒的人类活动识别（HAR）并在用户和未见条件下具有更强的泛化能力。然而，在现实世界的变异性下实现可靠的泛化仍然是一个主要挑战，阻碍了Wi-Fi感知在实际应用中的采用。为了解决这一挑战，我们引入了多普勒辐射场（DoRF），将计算机视觉中的神经辐射场（NeRF）概念引入Wi-Fi感知。DoRF将从Wi-Fi CSI中提取的多普勒速度投影建模为人类运动的稀疏且多样的虚拟相机视角。然后，它推断出一个潜在的3D运动序列，其沿学习到的有效多普勒方向的投影解释了基于CSI的多普勒观测。恢复的运动随后被投影到单位球面上的等角度方向网格上，生成基础运动的球面表示。由于DoRF自然地在球面上定义了多普勒表示，我们进一步引入了DoRF++，这是一种球面学习设计，应用球面变换器进行活动分类。在我们收集的手势数据集上的实验表明，DoRF++在跨用户泛化准确性方面显著优于最先进的基于Wi-Fi的HAR方法，尤其是在单个多天线接入点（AP）设置下对困难手势的识别。
cs.CV / 96 / 2608.08401
Anatomically Consistent Cross-Contrast Super-Resolution of Anisotropic Brain T2w MRI
解剖一致的各向异性脑T2加权MRI交叉对比超分辨率
Abstract
T2-weighted (T2w) brain MRI provides fluid-sensitive soft-tissue contrast that is important for neuro-oncology and radiotherapy planning. However, T2w scans are acquired with anisotropic voxels and appear blurred or stair-stepped on coronal and sagittal views, which obscures small structures and weakens any downstream 3D analysis. We propose VIPP-SR (View-Independent Patched Projection Super-Resolution), a cross-contrast guided super-resolution framework that restores the inter-plane resolution of an existing anisotropic T2w volume without an isotropic ground-truth T2w. VIPP-SR first trains a view-independent patched generator (VIP-GAN) to learn local T1c-to-T2w anatomical correspondence from high-resolution axial slices. The trained generator is then applied to axial, coronal, and sagittal views of the T1c volume to generate three orthogonal T2w estimates. Shape-preserving patching and deepest-skip removal reduce view-specific shortcuts, thereby constraining the generator to learn patch-local representations and enabling the zero-shot inter-plane transfer. Central to VIPP-SR, a projection-based optimization then enforces anatomical consistency across the three view-specific volumes, fusing them by balancing inter-plane self-consistency against per-view data fidelity. The generator is trained on BraTS-MET and evaluated on both the held-out BraTS-MET testing set and the BraTS-GLI cohort without retraining, assessing the cross-cohort generalizability. The results validate that VIPP-SR improves downstream segmentation over the real anisotropic T2w baseline, raising mean-label Dice from 0.330 to 0.465 on BraTS-MET and, zero-shot, from 0.473 to 0.563 on BraTS-GLI and ablation studies identify inter-plane self-consistency as the main source of the gain.
Chinese Translation
T2加权（T2w）脑MRI提供了对液体敏感的软组织对比，这对于神经肿瘤学和放射治疗规划至关重要。然而，T2w扫描是以各向异性体素获取的，在冠状面和矢状面视图中呈现模糊或阶梯状，这掩盖了小结构并削弱了后续的三维分析。我们提出了VIPP-SR（视图独立补丁投影超分辨率），这是一种交叉对比引导的超分辨率框架，能够在没有各向同性真实T2w的情况下恢复现有各向异性T2w体积的平面间分辨率。VIPP-SR首先训练一个视图独立的补丁生成器（VIP-GAN），从高分辨率轴向切片中学习局部T1c与T2w的解剖对应关系。训练好的生成器随后应用于T1c体积的轴向、冠状和矢状视图，以生成三个正交的T2w估计。保持形状的补丁和最深跳跃去除减少了视图特定的捷径，从而限制生成器学习补丁局部表示，并实现零样本的平面间转移。VIPP-SR的核心是基于投影的优化，强制在三个视图特定体积之间保持解剖一致性，通过平衡平面间自一致性与每视图数据保真度来融合它们。生成器在BraTS-MET上训练，并在未重新训练的情况下评估保留的BraTS-MET测试集和BraTS-GLI队列，评估跨队列的泛化能力。结果验证了VIPP-SR在真实各向异性T2w基线之上改善了下游分割，将BraTS-MET上的平均标签Dice从0.330提高到0.465，并在BraTS-GLI上实现零样本从0.473提高到0.563，消融研究确定平面间自一致性是增益的主要来源。
cs.CV / 97 / 2608.08402
Agentic AI-powered flexible fiber-bundle endoscopy for high-resolution NIR-II fluorescence imaging in vivo
基于智能代理的AI驱动柔性光纤束内窥镜用于高分辨率NIR-II荧光成像的体内应用
Abstract
Fiber-bundle endoscopy offers a compact and flexible route for clinical fluorescence imaging through natural human orifices, but since its first report in the 1950s, it has remained limited by low spatial resolution, honeycomb artifacts, and inter-core crosstalk. The crosstalk becomes more pronounced at near-infrared-II wavelengths (NIR-II, 1000-3000 nm), a spectral window that offers superior contrast, resolution, and tissue penetration depth for biomedical imaging. Here, we present an AI-powered flexible endoscopy platform that overcomes these constraints through optical-computational co-design: optimizing ultrathin fiber bundles to mitigate crosstalk-induced image blur and enable high-fidelity image transmission across the visible-to-NIR-II spectral range, and developing an Agent-Guided Mixture-of-Experts (GAME) pipeline for honeycomb-artifact removal and image restoration. GAME provides a single restoration entry point for diverse biomedical images acquired with our endoscope, spanning cell, mouse and human samples. It dynamically routes each input to suitable restoration experts via a vision-language model, facilitating image reconstruction with a fourfold resolution improvement beyond the NyquistShannon sampling limit. The utility of our endoscope is demonstrated through in vivo NIR-II imaging of anatomical structures in mice, as well as imaging of the digital micromirror device (DMD)-projected human gastric tube and lymphatic system, paving the way for future clinical translation.
Chinese Translation
光纤束内窥镜为通过自然人体腔道进行临床荧光成像提供了一种紧凑而灵活的途径，但自20世纪50年代首次报道以来，其应用一直受到低空间分辨率、蜂窝状伪影和核心间串扰的限制。在近红外-II波长（NIR-II，1000-3000 nm）下，串扰现象更加明显，而这一光谱窗口为生物医学成像提供了优越的对比度、分辨率和组织穿透深度。在此，我们提出了一种AI驱动的柔性内窥镜平台，通过光学-计算协同设计克服了这些限制：优化超薄光纤束以减轻串扰引起的图像模糊，并实现可见光至NIR-II光谱范围内的高保真图像传输，同时开发了一种代理引导的专家混合模型（Agent-Guided Mixture-of-Experts, GAME）管道用于蜂窝状伪影去除和图像恢复。GAME为通过我们的内窥镜获取的多样化生物医学图像提供了单一的恢复入口，涵盖细胞、小鼠和人类样本。它通过视觉-语言模型动态将每个输入路由到合适的恢复专家，从而实现超过奈奎斯特-香农采样极限的四倍分辨率改善。我们通过小鼠体内NIR-II成像解剖结构，以及成像数字微镜设备（DMD）投影的人类胃管和淋巴系统，展示了我们内窥镜的实用性，为未来的临床转化铺平了道路。
cs.CV / 98 / 2608.08418
Learning Deep Modality-Shared Self-Expressiveness for Image Clustering with Textual Information
学习深度模态共享自表达模型以利用文本信息进行图像聚类
Abstract
Leveraging textual information for image clustering has emerged as a promising direction, largely owing to the powerful representations learned by Vision-Language Models (VLMs). Existing approaches typically retrieve a textual counterpart for each image and then refine multimodal representations by directly enforcing cross-modal agreement, e.g., maximizing image-text similarity inherited from pretrained VLMs. However, such a strategy aligns heterogeneous representations across modalities without explicitly modeling the intrinsic structure within each modality and thus might yield unreliable alignment or distort modality-specific structures that are crucial for clustering. In this paper, we propose a simple but principled approach, termed deep modality-shared self-expressive model (DeepMORSE), which discovers cross-modal structures via a modality-shared self-expressive model and simultaneously learns structured representations that conform to a union of modality-specific subspaces. Moreover, we theoretically justify that the modality-shared self-expressive coefficients suppress inter-class noise towards a subspace-preserving solution, and show that mini-batch optimization procedure introduces an implicit regularization onto the self-expressive model. We evaluate our DeepMORSE on six widely used image clustering benchmarks and observe performance improvements exceeding 3% on the UCF-101, DTD-47, and ImageNet-Dogs datasets. In addition, we demonstrate the strong transferability of the learned representations by achieving state-of-the-art performance on downstream tasks such as image retrieval and zero-shot classification---without requiring any task-specific losses or post-processing. The code is available at: https://github.com/mengxianghan123/DeepMORSE.
Chinese Translation
利用文本信息进行图像聚类已成为一个有前景的方向，这主要得益于视觉-语言模型（Vision-Language Models, VLMs）所学习的强大表示。现有的方法通常为每个图像检索一个文本对应，并通过直接强制跨模态一致性来优化多模态表示，例如，最大化从预训练的VLMs中继承的图像-文本相似性。然而，这种策略在跨模态对齐异构表示的同时，并未明确建模每个模态内部的内在结构，因此可能导致不可靠的对齐或扭曲对聚类至关重要的模态特定结构。在本文中，我们提出了一种简单但有原则的方法，称为深度模态共享自表达模型（Deep Modality-Shared Self-Expressive Model, DeepMORSE），该模型通过模态共享自表达模型发现跨模态结构，并同时学习符合模态特定子空间联合的结构化表示。此外，我们从理论上证明模态共享自表达系数抑制了朝向子空间保持解的类间噪声，并表明小批量优化过程对自表达模型引入了隐式正则化。我们在六个广泛使用的图像聚类基准上评估了我们的DeepMORSE，并观察到在UCF-101、DTD-47和ImageNet-Dogs数据集上的性能提升超过3%。此外，我们通过在下游任务（如图像检索和零样本分类）上实现最先进的性能，展示了学习到的表示的强大可迁移性——无需任何特定任务的损失或后处理。代码可在以下链接获取：https://github.com/mengxianghan123/DeepMORSE。
cs.CV / 99 / 2608.08436
FreCast: Refining Radar Echo Intensity via Phase-Preserving Amplitude Residual Diffusion for Precipitation Nowcasting
FreCast：通过相位保持幅度残差扩散精细化雷达回波强度以进行降水即时预报
Abstract
Precipitation nowcasting predicts the spatiotemporal evolution of future radar echoes from historical radar echo sequences, thereby estimating the occurrence, development, and movement of precipitation over the near term. In recent years, deep learning has become an important approach to precipitation nowcasting. Although state-of-the-art models can generally capture the overall spatial distribution of future precipitation, their predictions still exhibit substantial biases in radar echo intensity at individual locations. This observation motivates a more targeted strategy for reducing forecast errors. Instead of regenerating an entire radar echo sequence without spatial constraints, the predicted precipitation structure can be used to guide the refinement of echo intensities at individual locations. This structure-guided refinement directly targets echo intensity biases. Accordingly, we propose FreCast, a two-stage framework for radar echo prediction. The first stage generates an initial forecast of future radar echoes. The second stage uses the spatial structure of the initial forecast as a constraint to further correct intensity biases at individual locations in the first-stage prediction. Experiments on three datasets demonstrate that FreCast achieves consistent improvements across forecast skill metrics. Qualitative results further show that FreCast better preserves rainband continuity and intense precipitation structures at longer lead times.
Chinese Translation
降水即时预报通过历史雷达回波序列预测未来雷达回波的时空演变，从而估计近期降水的发生、发展和移动。近年来，深度学习已成为降水即时预报的重要方法。尽管最先进的模型通常能够捕捉未来降水的整体空间分布，但它们在个别位置的雷达回波强度预测仍然存在显著偏差。这一观察促使我们采取更有针对性的策略来减少预测误差。我们提出的预测降水结构可以用来指导个别位置的回波强度精细化，而不是在没有空间约束的情况下重新生成整个雷达回波序列。这种结构引导的精细化直接针对回波强度偏差。因此，我们提出了FreCast，一个用于雷达回波预测的两阶段框架。第一阶段生成未来雷达回波的初步预测。第二阶段利用初步预测的空间结构作为约束，进一步修正第一阶段预测中个别位置的强度偏差。在三个数据集上的实验表明，FreCast在预测技能指标上实现了一致的改进。定性结果进一步表明，FreCast在较长的预报时效下更好地保持了降雨带的连续性和强降水结构。
cs.CV / 100 / 2608.08460
InstructionCrafter: Generating Consistent and High-Fidelity Visual Instructions
InstructionCrafter：生成一致且高保真视觉指令
Abstract
Given textual task instructions, generating step-by-step visual instructions as an image sequence requires the simultaneous satisfaction of multiple properties, specifically step faithfulness, cross-image consistency, and per-frame visual quality. Existing text-to-image generation approaches rarely meet all three properties, owing to independent sampling that breaks consistency, finetuning on low-quality video that degrades per-frame quality, and frozen backbones that lack multi-step understanding. In this work, we propose InstructionCrafter, a diffusion-based framework with the key idea of separating the optimization of temporal and instructional alignment from per-frame visual quality via (1) spatial-freeze training and (2) instruction-aware adapters. Built on a pretrained video diffusion backbone, InstructionCrafter freezes the spatial layers that control per-frame detail and updates only temporal and text-conditioning pathways to learn instruction semantics and inter-step relations, which preserves the generative prior for per-frame quality and reduces trainable parameters by about 50 percent compared with full finetuning. We also introduce two lightweight adapters that enhance the model's understanding of instructional context. The Consistent Adapter aggregates textual cues from the entire instruction sequence and from neighboring steps to keep object identity and attributes consistent across frames, and the Context-Aware Temporal Adapter converts cross-attention outputs into biases for temporal self-attention, explicitly propagating inter-frame relations. Extensive experiments on two benchmark datasets demonstrate state-of-the-art overall performance on step faithfulness, cross-image consistency, and per-frame visual quality while significantly reducing noise, blur, and spurious subtitles. Our code and trained models will be publicly available.
Chinese Translation
给定文本任务指令，将逐步视觉指令生成图像序列需要同时满足多个属性，特别是步骤忠实性、跨图像一致性和每帧视觉质量。现有的文本到图像生成方法很少能同时满足这三种属性，原因在于独立采样破坏了一致性、在低质量视频上进行微调导致每帧质量下降，以及冻结的骨干网络缺乏多步骤理解。在本研究中，我们提出了InstructionCrafter，一个基于扩散的框架，其关键思想是通过（1）空间冻结训练和（2）指令感知适配器，将时间和指令对齐的优化与每帧视觉质量的优化分开。InstructionCrafter建立在一个预训练的视频扩散骨干网络之上，冻结控制每帧细节的空间层，仅更新时间和文本条件路径，以学习指令语义和步骤间关系，从而保持每帧质量的生成先验，并将可训练参数减少约50％，与完全微调相比。我们还引入了两个轻量级适配器，增强模型对指令上下文的理解。一致适配器（Consistent Adapter）聚合来自整个指令序列和相邻步骤的文本线索，以保持对象身份和属性在帧间的一致性，而上下文感知时间适配器（Context-Aware Temporal Adapter）将交叉注意力输出转换为时间自注意力的偏置，明确传播帧间关系。在两个基准数据集上的大量实验表明，我们在步骤忠实性、跨图像一致性和每帧视觉质量方面实现了最先进的整体性能，同时显著减少了噪声、模糊和虚假字幕。我们的代码和训练模型将公开发布。
cs.CV / 101 / 2608.08476
RayLift: Lifting Complementary Ray-Wise Evidence with 3D Geometry Priors for Semantic Scene Completion
RayLift：利用3D几何先验提升互补光线证据以实现语义场景补全
Abstract
Camera-based 3D semantic scene completion (SSC) provides comprehensive scene understanding for autonomous driving and robotics. However, existing methods often treat stereo depth estimates as deterministic geometric constraints, causing depth uncertainty and local correspondence errors to propagate directly into voxel representations. To address this issue, we propose RayLift, a framework that uses stereo geometry as a metric reference while incorporating complementary ray evidence to recover reliable 3D structures adaptively. RayLift first employs a Complementary Context Encoder that extracts geometry-aware priors from a frozen 3D vision foundation model, thereby enriching the scene context. It then introduces a Depth Ray Evidence Lifter module that jointly models geometric dissimilarity, depth confidence, and spatial uncertainty to adaptively sample and weight candidate surface locations along each camera ray. Finally, a Semantic-Aware Voxel Integrator injects the resulting ray evidence into voxel features by explicitly modeling their spatial support. Extensive experiments on SemanticKITTI and SSCBench-KITTI-360 demonstrate that RayLift achieves competitive performance and consistently outperforms existing methods.
Chinese Translation
基于相机的3D语义场景补全（SSC）为自动驾驶和机器人提供了全面的场景理解。然而，现有方法往往将立体深度估计视为确定性的几何约束，导致深度不确定性和局部对应错误直接传播到体素表示中。为了解决这一问题，我们提出了RayLift，一个利用立体几何作为度量参考，同时结合互补光线证据以自适应恢复可靠3D结构的框架。RayLift首先采用互补上下文编码器，从冻结的3D视觉基础模型中提取几何感知先验，从而丰富场景上下文。接着，它引入了深度光线证据提升模块，该模块联合建模几何不相似性、深度置信度和空间不确定性，以自适应地采样和加权每条相机光线上的候选表面位置。最后，语义感知体素整合器通过显式建模其空间支持，将生成的光线证据注入体素特征。对SemanticKITTI和SSCBench-KITTI-360的广泛实验表明，RayLift实现了具有竞争力的性能，并始终优于现有方法。
cs.CV / 102 / 2608.08487
RenderMatte: Exact-Alpha Rendering and Group-Relative Alignment for Image Matting
RenderMatte：图像抠图的精确α渲染与组相对对齐
Abstract
Image matting is an essential enabling technology for modern visual content production, where foreground extraction determines the realism and editability of downstream creation workflows. However, precise alpha estimation in open-world scenes remains challenging because real foregrounds exhibit highly diverse appearances and opacity patterns. This makes existing methods struggle with semantic ambiguity and fine-grained opacity variation, especially in sparse boundary regions that are fragile and difficult to supervise. To address this gap, we present RenderMatte, a trimap-guided matting framework that adapts FLUX.1 Kontext through full-parameter fine-tuning, leveraging image editing priors for structure-preserving alpha prediction. During supervised adaptation, an alpha-edge objective preserves the latent flow-matching signal while strengthening pixel-space boundary supervision. We further introduce group-relative alpha alignment for post-training. It compares multiple mattes sampled under the same trimap condition using matting-specific rewards for alpha accuracy, boundary fidelity, trimap compliance, and compositional consistency. To overcome the lack of precise edge annotations, we construct the RenderMatte dataset, a large-scale synthetic dataset combining 3D-rendered RGBA foregrounds with diverse multi-source assets. It features exact strand-level alpha annotations and diverse background composites. Experiments show state-of-the-art performance across all benchmarks, demonstrating a scalable path toward high-fidelity matting in open-world scenes.
Chinese Translation
图像抠图是现代视觉内容制作的重要基础技术，其中前景提取决定了下游创作工作流程的真实感和可编辑性。然而，在开放世界场景中，精确的α估计仍然具有挑战性，因为真实前景表现出高度多样的外观和不透明度模式。这使得现有方法在语义模糊和细粒度不透明度变化方面面临困难，尤其是在脆弱且难以监督的稀疏边界区域。为了解决这一问题，我们提出了RenderMatte，一个通过全参数微调适应FLUX.1 Kontext的三分图引导抠图框架，利用图像编辑先验进行结构保留的α预测。在监督适应过程中，α边缘目标保留了潜在的流匹配信号，同时增强了像素空间的边界监督。我们进一步引入了组相对α对齐用于后训练。它使用针对α准确性、边界保真度、三分图合规性和组合一致性的抠图特定奖励，比较在相同三分图条件下采样的多个抠图。为了克服缺乏精确边缘注释的问题，我们构建了RenderMatte数据集，这是一个大规模合成数据集，结合了3D渲染的RGBA前景与多样的多源资产。该数据集具有精确的链级α注释和多样的背景合成。实验结果显示在所有基准测试中均达到最先进的性能，展示了在开放世界场景中实现高保真抠图的可扩展路径。
cs.CV / 103 / 2608.08494
Linguistically-Aligned and Visually-Grounded Preference Optimization for Clinically-Augmented Medical Report Generation
语言对齐与视觉基础的偏好优化用于临床增强的医学报告生成
Abstract
Despite significant advances in Medical Report Generation (MRG), the reliability remains constrained by the prevalence of factual errors. While Direct Preference Optimization (DPO) has emerged as a promising post-training paradigm to enhance the performance of Supervised Fine-Tuned (SFT) MRG models, existing DPO-based MRG methods typically adopt a naive preference construction that directly pairs model-generated reports with ground truth reports. This strategy inadvertently entangles critical clinical findings with clinically irrelevant linguistic characteristics, and fundamentally lacks explicit vision-language alignment. To address these challenges, we propose DPO-Clin, a novel post-training framework that focuses preference optimization on clinical findings and cross-modal alignment. First, we introduce the Entity-level Clinical Diagnostic (ECD) module to perform a precise entity-level factual diagnosis. ECD guides the generation of linguistically-aligned report preference pairs, isolating clinical discrepancies from linguistic variations. Second, to achieve fine-grained cross-modal alignment, we develop M$^2$DPO, a retrieval-augmented multi-modal DPO variant that enforces textual preference inversion triggered by visual context switches. Third, we locate correct yet highly uncertain predicted entities and apply counterfactual modifications to construct targeted preference data for latent risk mitigation, thereby further enhancing the model reliability. Extensive experiments on two public chest X-ray datasets (MIMIC-CXR and IU X-Ray) and an in-house endoscopy dataset demonstrate that DPO-Clin significantly improves the SFT baselines on clinical-aware metrics. Furthermore, it achieves superior performance over existing DPO-based MRG methods, exhibiting robust generalizability across distinct baseline architectures and diverse medical imaging modalities.
Chinese Translation
尽管医学报告生成（MRG）取得了显著进展，但其可靠性仍受到事实错误普遍存在的限制。尽管直接偏好优化（DPO）作为一种有前景的后训练范式出现，以提升监督微调（SFT）MRG模型的性能，但现有的基于DPO的MRG方法通常采用一种简单的偏好构建，直接将模型生成的报告与真实报告配对。这一策略无意中将关键临床发现与临床无关的语言特征纠缠在一起，并在根本上缺乏明确的视觉-语言对齐。为了解决这些挑战，我们提出了DPO-Clin，这是一种新的后训练框架，专注于临床发现和跨模态对齐的偏好优化。首先，我们引入了实体级临床诊断（ECD）模块，以执行精确的实体级事实诊断。ECD指导生成语言对齐的报告偏好对，隔离临床差异与语言变化。其次，为了实现细粒度的跨模态对齐，我们开发了M$^2$DPO，这是一种检索增强的多模态DPO变体，强制执行由视觉上下文切换触发的文本偏好反转。第三，我们定位正确但高度不确定的预测实体，并应用反事实修改来构建针对性偏好数据，以减轻潜在风险，从而进一步提高模型的可靠性。在两个公共胸部X光数据集（MIMIC-CXR和IU X-Ray）以及一个内部内窥镜数据集上的广泛实验表明，DPO-Clin在临床感知指标上显著改善了SFT基线。此外，它在现有的基于DPO的MRG方法中表现出优越的性能，展现出在不同基线架构和多样化医学成像模态下的强大泛化能力。
cs.CV / 104 / 2608.08508
Towards Adaptive Super-Resolution and Quality Assessment via Test-Time Adaptation
基于测试时适应的自适应超分辨率与质量评估研究
Abstract
This paper presents doctoral research on adaptive video super-resolution and perceptual quality modeling under real-world conditions. Existing video super-resolution (VSR) methods struggle to generalize under unknown degradations arising from heterogeneous devices, codecs, and network environments. We address this challenge through test-time adaptation (TTA), a unified paradigm that improves robustness and perceptual quality without retraining or high-quality supervision. Specifically, we: 1) propose a TTA-based framework for no-reference video quality assessment (VQA), where adapted quality predictions provide perceptual guidance for VSR under unseen distortions; 2) develop a transformer-based architecture for screen-content super-resolution that preserves text clarity and structural fidelity; and 3) introduce a region-aware TTA strategy that selectively refines text and non-text regions without requiring high-resolution ground truth. Experimental results across diverse benchmarks demonstrate consistent improvements in perceptual quality and readability. We also outline ongoing work toward fully adaptive video enhancement systems capable of generalizing across unseen domains.
Chinese Translation
本文呈现了在真实世界条件下进行自适应视频超分辨率和感知质量建模的博士研究。现有的视频超分辨率（VSR）方法在面对来自异构设备、编解码器和网络环境的未知降质时，表现出较差的泛化能力。我们通过测试时适应（TTA）这一统一范式来应对这一挑战，该范式在不进行重训练或高质量监督的情况下，提高了鲁棒性和感知质量。具体而言，我们：1）提出了一种基于TTA的无参考视频质量评估（VQA）框架，其中适应后的质量预测为在未见失真下的视频超分辨率提供感知指导；2）开发了一种基于变换器（transformer）的屏幕内容超分辨率架构，能够保持文本清晰度和结构保真度；3）引入了一种区域感知的TTA策略，能够选择性地细化文本和非文本区域，而无需高分辨率的真实值。实验结果在多样化的基准测试中显示出感知质量和可读性的一致提升。我们还概述了正在进行的工作，旨在开发能够在未见领域中泛化的完全自适应视频增强系统。
cs.CV / 105 / 2608.08519
eBIRD: Event-based Intensity Image Reconstruction Using Controllable Diffusion Models
eBIRD：基于事件的强度图像重建方法，使用可控扩散模型
Abstract
Intensity-image reconstruction from event streams remains a challenging problem due to the binary, sparse, and asynchronous nature of event data. This work proposes eBIRD, an event-guided reconstruction framework that combines a DDPM with ControlNet-based conditioning. We analyze generic and specialized diffusion learning strategies for handwritten digit (N-MNIST) and face (RGBE-Gaze) reconstruction using 33ms event windows. On N-MNIST, the general model achieves the best reconstruction quality (MSE 0.0052, SSIM 0.8982, PSNR 23.34dB), whereas the specialized model performs best on RGBE-Gaze (MSE 0.0161, SSIM 0.7605, PSNR 19.08dB). These preliminary results suggest that controllable diffusion models are a promising approach for event-guided intensity-image reconstruction, while highlighting that the preferred learning strategy depends on the reconstruction domain.
Chinese Translation
从事件流中重建强度图像仍然是一个具有挑战性的问题，因为事件数据具有二进制、稀疏和异步的特性。本研究提出了eBIRD，一个事件引导的重建框架，它结合了DDPM（去噪扩散概率模型）与基于ControlNet的条件控制。我们分析了用于手写数字（N-MNIST）和人脸（RGBE-Gaze）重建的通用和专用扩散学习策略，使用33毫秒的事件窗口。在N-MNIST数据集上，通用模型实现了最佳的重建质量（均方误差MSE 0.0052，结构相似性指数SSIM 0.8982，峰值信噪比PSNR 23.34dB），而专用模型在RGBE-Gaze数据集上表现最佳（均方误差MSE 0.0161，结构相似性指数SSIM 0.7605，峰值信噪比PSNR 19.08dB）。这些初步结果表明，可控扩散模型是事件引导的强度图像重建的一个有前景的方法，同时强调了首选学习策略依赖于重建领域。
cs.CV / 106 / 2608.08521
A Combined Feature-Based Framework for Disguise and Spoofing Detection in Face Recognition Systems
一种结合特征的框架用于面部识别系统中的伪装和欺骗检测
Abstract
Face recognition systems face two distinct, commonly-separated failure modes: spoofing, where an impostor presents a photograph or video of an authorized user, and disguise, where a legitimate user is rejected because their appearance differs from their enrolled template due to accessories, facial hair, illumination, or pose. This paper proposes and compares five combined feature-extraction and classification pipelines that address both problems within a single framework: PM (PCA and Minimum Euclidean Distance, MED), LPM (Local Binary Patterns with PCA and MED), HPM (Histogram of Oriented Gradients with PCA and MED), SM (Speeded-Up Robust Features with MED), and HM (Harris corner features with MED). Each pipeline follows a common two-phase process comprising pre-processing, feature extraction, feature filtering, and classification. The methods were trained on 115 subjects drawn from the FEI, Disguised Faces Database, and NUAA databases and evaluated on six test conditions covering mixed appearances, frontal faces, dark illumination, left- and right-turned poses, and photo-spoofing attempts. The HOG-based pipeline (HPM) achieved the most consistent performance across conditions, with 94.59% accuracy on mixed-appearance disguise, 81.5-93.2% across pose and illumination variants, and 91.67% on spoofing, while the LBP-based pipeline (LPM) achieved the highest spoofing-detection accuracy (93.2%) but weaker robustness to pose change. These results reveal a measurable trade-off between spoof sensitivity and disguise robustness among classical feature representations, motivating the deep-learning and cross-database extensions discussed in the concluding sections.
Chinese Translation
面部识别系统面临两种明显且通常被分开的失败模式：欺骗，即冒名顶替者呈现授权用户的照片或视频；伪装，即合法用户因其外观因配饰、面部毛发、光照或姿势的不同而被拒绝。本文提出并比较了五种结合特征提取和分类的管道，这些管道在单一框架内解决这两个问题：PM（主成分分析和最小欧几里得距离，MED）、LPM（局部二值模式结合主成分分析和MED）、HPM（方向梯度直方图结合主成分分析和MED）、SM（加速稳健特征结合MED）和HM（哈里斯角点特征结合MED）。每个管道遵循一个共同的两阶段过程，包括预处理、特征提取、特征过滤和分类。这些方法在来自FEI、伪装面孔数据库和NUAA数据库的115个受试者上进行了训练，并在六种测试条件下进行了评估，涵盖混合外观、正面脸、暗光照、左转和右转姿势以及照片欺骗尝试。基于HOG的管道（HPM）在各种条件下表现出最一致的性能，在混合外观伪装上达到94.59%的准确率，在姿势和光照变体上达到81.5-93.2%，在欺骗检测上达到91.67%；而基于LBP的管道（LPM）在欺骗检测准确率上达到最高（93.2%），但对姿势变化的鲁棒性较弱。这些结果揭示了经典特征表示中欺骗敏感性与伪装鲁棒性之间的可测量权衡，激励了在结论部分讨论的深度学习和跨数据库扩展。
cs.CV / 107 / 2608.08531
ERF-GS: Reconstructing Fast Motion from Disjoint Event-RGB Viewpoints
ERF-GS：从不连续事件-RGB视角重建快速运动
Abstract
Deep learning-driven representations such as neural radiance fields (NeRFs) and 3D Gaussian splatting (3DGS) have revolutionized the field of dynamic 3D scene reconstruction with improved visual precision and scalability. However, the reconstruction of fast-moving objects remains a challenge; existing methods based on conventional frame-based videos often struggle in scenarios such as sports events and animal videography. We propose an event-RGB fusion Gaussian splatting (ERF-GS) framework that integrates event information into both optimization and densification stages of the Gaussian splatting pipeline, taking advantage of novel event sensors with high frame-rate. Unlike many other event-assisted scene reconstruction methods, ERF-GS was developed using realistic simulation settings and realizes event-based learning detached from RGB inputs. This design enables its application beyond straightforward synthetic data into the realm of natural video with complex layout, low frame rates and severe motion blur. Our experiments show that ERF-GS outperforms both the 4DGS baseline and the concurrent E-D3DGS on different variants of the Neu3D and Nvidia datasets which include blurry RGB frames and disjoint RGB-event viewpoints. Our code is available at https://github.com/andrewbxy/ERF-GS.
Chinese Translation
深度学习驱动的表示方法，如神经辐射场（NeRFs）和3D高斯点云（3DGS），在动态3D场景重建领域带来了视觉精度和可扩展性的革命。然而，快速移动物体的重建仍然是一个挑战；基于传统帧视频的现有方法在体育赛事和动物录像等场景中往往难以应对。我们提出了一种事件-RGB融合高斯点云（ERF-GS）框架，该框架将事件信息整合到高斯点云管道的优化和密集化阶段，利用高帧率的新型事件传感器。与许多其他辅助事件场景重建方法不同，ERF-GS是在现实的仿真设置下开发的，实现了与RGB输入无关的基于事件的学习。这一设计使其能够应用于复杂布局、低帧率和严重运动模糊的自然视频，而不仅限于简单的合成数据。我们的实验表明，ERF-GS在包含模糊RGB帧和不连续RGB-事件视角的Neu3D和Nvidia数据集的不同变体上，优于4DGS基线和当前的E-D3DGS。我们的代码可在https://github.com/andrewbxy/ERF-GS获取。
cs.CV / 108 / 2608.08541
Rethinking Attention Locality in Spiking Transformers
重新思考脉冲变换器中的注意力局部性
Abstract
Spiking Transformers provide a promising paradigm for efficient visual processing with spike-driven computation, yet their Softmax-free Spiking Self-Attention (SSA) struggles to establish spatially localized token interactions. Although existing locality-enhanced SSA methods improve accuracy, it remains unclear whether they consistently induce spatial locality across layers and different Spiking Transformer architectures. Through Mean Attention Distance (MAD) analysis, we reveal that computational locality does not necessarily translate into spatial locality and show that uniformly applying the same locality enhancement overlooks architecture-dependent deployment requirements. Motivated by these observations, we propose Spatially Contiguous Local Attention with Boundary Continuity Pathway (SCLA-BCP). SCLA computes attention within non-overlapping regions of spatially adjacent tokens, while BCP facilitates cross-boundary information exchange through a lightweight convolutional pathway. Furthermore, we develop a hierarchical locality deployment strategy to effectively apply SCLA-BCP across the two major Spiking Transformer architectures. Extensive experiments on seven static and neuromorphic datasets covering classification, detection, and segmentation demonstrate consistent improvements with limited parameter and energy overhead. Notably, our approach improves mAP@50 by up to 9.50% on COCO 2017 and mIoU by up to 3.42% on ADE20K. Visualizations, MAD analysis, and ablation studies further validate its effectiveness.
Chinese Translation
脉冲变换器为高效的视觉处理提供了一种有前景的范式，通过脉冲驱动的计算，然而其无Softmax的脉冲自注意力（Spiking Self-Attention, SSA）在建立空间局部化的标记交互方面面临挑战。尽管现有的增强局部性的SSA方法提高了准确性，但尚不清楚它们是否在不同层和不同脉冲变换器架构中始终诱导空间局部性。通过平均注意力距离（Mean Attention Distance, MAD）分析，我们揭示计算局部性并不一定转化为空间局部性，并且表明统一应用相同的局部性增强忽视了依赖于架构的部署要求。基于这些观察，我们提出了具有边界连续路径（Boundary Continuity Pathway, BCP）的空间连续局部注意力（Spatially Contiguous Local Attention, SCLA）。SCLA在空间相邻标记的非重叠区域内计算注意力，而BCP通过轻量级卷积路径促进跨边界信息交换。此外，我们开发了一种分层局部性部署策略，以有效地在两种主要的脉冲变换器架构中应用SCLA-BCP。在涵盖分类、检测和分割的七个静态和神经形态数据集上的广泛实验表明，在有限的参数和能量开销下，取得了一致的改进。值得注意的是，我们的方法在COCO 2017上将mAP@50提高了最多9.50%，在ADE20K上将mIoU提高了最多3.42%。可视化、MAD分析和消融研究进一步验证了其有效性。
cs.CV / 109 / 2608.08553
MotionCraft: Latent World Modeling with Sparse Attention for Visual Upscaling
MotionCraft：基于稀疏注意力的潜在世界建模用于视觉超分辨率
Abstract
Video super-resolution (VSR) aims to recover high-fidelity high-resolution videos from low-resolution inputs and is central to applications ranging from mobile capture to streaming and archival restoration. Existing approaches trade off among local-detail fidelity, long-range spatio-temporal modeling, perceptual realism, and efficiency: convolutional alignment techniques preserve local structure but suffer when motion is large or degradations are complex; transformer-based methods capture long-range dependencies yet require architectural or algorithmic adaptations to remain computationally feasible; and recent latent or diffusion-based generators synthesize rich texture but require specialized temporal constraints to maintain coherence. We present MotionCraft, a controllable VSR framework that formulates restoration as motion-aware latent state prediction inspired by world models and integrates adaptive sparse attention with an explicit user-accessible control interface. MotionCraft combines robust motion fusion, a Latent World Transformer that balances locality and targeted non-local interactions, and a compact conditional decoder to deliver temporally consistent, high-quality reconstructions under streaming constraints. Empirical evaluations show that MotionCraft achieves strong reconstruction and perceptual performance while enabling predictable trade-offs between temporal smoothness and reconstruction fidelity.
Chinese Translation
视频超分辨率（VSR）旨在从低分辨率输入中恢复高保真高分辨率视频，并在从移动捕捉到流媒体和档案修复等应用中占据核心地位。现有方法在局部细节保真度、长程时空建模、感知真实感和效率之间进行权衡：卷积对齐技术能够保持局部结构，但在运动较大或降解复杂时表现不佳；基于变换器的方法捕捉长程依赖关系，但需要架构或算法上的调整以保持计算上的可行性；而最近的潜在或扩散生成器合成丰富的纹理，但需要专门的时间约束以保持一致性。我们提出了MotionCraft，一个可控的VSR框架，将恢复过程表述为受运动感知的潜在状态预测，灵感来源于世界模型，并结合了自适应稀疏注意力与明确的用户可访问控制界面。MotionCraft结合了强大的运动融合、平衡局部性与针对性非局部交互的潜在世界变换器，以及一个紧凑的条件解码器，以在流媒体约束下提供时间一致的高质量重建。实证评估表明，MotionCraft在重建和感知性能上表现出色，同时实现了时间平滑性与重建保真度之间的可预测权衡。
cs.CV / 110 / 2608.08555
SC-Diff: Semantically Calibrated Diffusion for Visible-to-Infrared Image Translation
SC-Diff：用于可见光到红外图像翻译的语义校准扩散模型
Abstract
Visible-to-infrared image translation provides a practical way to expand infrared training data using abundant visible images. Diffusion models are promising for this task because of their strong generative performance. However, existing diffusion-based methods typically use semantic priors only as external conditions, without explicitly regulating token interactions within the denoising network. Consequently, they struggle to preserve object locations, shapes, and semantic layouts required for reliable annotation reuse. We propose SC-Diff, a semantically calibrated latent diffusion framework that uses semantic priors for both conditional guidance and internal self-attention calibration. A pretrained SAM3 model with predefined text prompts first extracts category-specific semantic masks from visible images. These masks are merged into a semantic map and fused with the visible image as the input condition. The same map is converted into token-level semantic labels to calibrate self-attention in the denoising network. Based on these labels, we introduce Semantic-Guided Self-Attention Calibration (SGSC), which adaptively applies positive biases to query-key pairs of the same category. The query-wise calibration strength depends on the dispersion of attention across semantic categories and the attention assigned to the query's own category. The original attention scores further modulate the bias, giving greater calibration to same-category keys with stronger responses. This soft calibration reduces cross-category interference while retaining global contextual interactions, thereby improving semantic consistency in generated infrared images. Extensive experiments show that SC-Diff improves perceptual quality and produces more effective synthetic training data for downstream infrared object detection.
Chinese Translation
可见光到红外图像翻译提供了一种利用丰富的可见光图像扩展红外训练数据的实用方法。扩散模型因其强大的生成性能而在这一任务中展现出良好的前景。然而，现有的基于扩散的方法通常仅将语义先验作为外部条件，而未能明确调节去噪网络内的标记交互。因此，它们在保持对象位置、形状和语义布局方面面临挑战，这些都是可靠注释重用所必需的。我们提出了SC-Diff，一种语义校准的潜在扩散框架，利用语义先验进行条件指导和内部自注意力校准。首先，使用预训练的SAM3模型和预定义的文本提示从可见光图像中提取特定类别的语义掩膜。这些掩膜被合并成一个语义图，并与可见光图像融合作为输入条件。同一图被转换为标记级的语义标签，以校准去噪网络中的自注意力。基于这些标签，我们引入了语义引导自注意力校准（SGSC），该方法自适应地对同一类别的查询-键对施加正偏差。查询级的校准强度取决于跨语义类别的注意力分散程度以及分配给查询自身类别的注意力。原始的注意力分数进一步调节偏差，使得对响应更强的同类键给予更大的校准。这种软校准减少了跨类别的干扰，同时保留了全局上下文交互，从而提高了生成红外图像的语义一致性。大量实验表明，SC-Diff提高了感知质量，并为下游红外目标检测生成了更有效的合成训练数据。
cs.CV / 111 / 2608.08566
On-Device Multi-Species Malaria Detection with Uncertainty-Calibrated Slide-Level Aggregation
基于不确定性校准的设备内多物种疟疾检测与幻灯片级聚合
Abstract
Malaria remains a leading cause of mortality in resource-limited settings, where expert microscopists are scarce. Automated diagnosis based on microscopy images thus has strong potential to improve care delivery. But for an algorithm to deploy, a necessary requirement is that it meet a suite of non-obvious (from a machine learning (ML) perspective) clinical constraints. Therefore, in close consultation with a national health center we developed a malaria diagnosis pipeline which addresses key requirements listed by the health care center but typically ignored in the ML malaria literature. In particular, it includes: (i) stopping criteria (to reduce image acquisition and time-to-result); (ii) human-in-the-loop functionality (for review and accountability); (iii) multi-species discrimination (since treatment varies by species); (iv) thick film detection (standard for microscopy); (v) computationally-efficient uncertainty calculations (to aid clinician review); and (vi) an edge device platform (since internet can be spotty in this catchment area). The mobile system performs all inference on-device using YOLOv13n deployed via TensorFlow Lite. It detects four species and white blood cells from Giemsa-stained thick blood smear images, aggregating per-image detections into slide-level parasitemia with World Health Organization (WHO)-standard quantification. This paper highlights these various clinical constraints and offers methods to address them. Evaluated on 2,739 annotated images across all four species, the system achieves
[email protected] of 0.863, per-image parasite count correlation of r = 0.812, slide-level r = 0.951 (soft counting, 10 images/slide), and runs entirely offline with a pipeline time of 10.27 +- 1.65 s per image.
Chinese Translation
疟疾仍然是资源有限地区主要的死亡原因，而专业显微镜技师稀缺。因此，基于显微镜图像的自动化诊断具有显著的潜力来改善医疗服务。然而，为了使算法能够部署，必须满足一系列在机器学习（ML）视角下并不明显的临床约束。因此，我们与国家卫生中心密切合作，开发了一种疟疾诊断管道，旨在满足卫生中心列出的关键要求，而这些要求在ML疟疾文献中通常被忽视。特别地，它包括：（i）停止标准（以减少图像采集和结果时间）；（ii）人机协作功能（用于审查和问责）；（iii）多物种鉴别（因为治疗因物种而异）；（iv）厚血膜检测（显微镜的标准）；（v）计算效率高的不确定性计算（以辅助临床审查）；以及（vi）边缘设备平台（因为该地区的互联网连接可能不稳定）。该移动系统使用通过TensorFlow Lite部署的YOLOv13n在设备上执行所有推理。它能够从Giemsa染色的厚血涂片图像中检测四种物种及白细胞，将每幅图像的检测结果聚合为幻灯片级的寄生虫血症，并采用世界卫生组织（WHO）标准进行量化。本文强调了这些各种临床约束，并提供了解决这些问题的方法。在2,739幅标注图像上进行评估，该系统在所有四种物种中实现了
[email protected]为0.863，单幅图像寄生虫计数相关性为r = 0.812，幻灯片级r = 0.951（软计数，10幅图像/幻灯片），并且完全离线运行，每幅图像的管道时间为10.27 ± 1.65秒。
cs.CV / 112 / 2608.08575
CDGC-Net: 3D Medical Image Segmentation with Cooperative Dual-Scale Self-Attention and Grouped Channel Modeling
CDGC-Net：具有协同双尺度自注意力和分组通道建模的3D医学图像分割
Abstract
Accurate 3D medical image segmentation requires the integration of long-range anatomical context with fine boundary detail. Existing methods often model global and local features in separate modules or feature levels and perform channel recalibration independently. This may cause semantic mismatch between global context and local boundaries, insufficient channel relationship modeling, weak spatial-channel interaction, and redundant representations. We propose CDGC-Net, a 3D medical image segmentation network that combines cooperative dual-scale spatial attention with grouped hierarchical channel modeling. With-in each CDGC block, Cooperative Dual-Scale Self-Attention (CDSA) assigns attention heads to parallel local-window and global-sparse branches. The two branches capture fine spatial details and long-range anatomical context at the same feature level. Their outputs are concatenated into an $N\times C$ spatial representation and directly passed to Grouped Hierarchical Channel Attention (GHCA). GHCA organizes the channels into $r$ groups and models both within-group and cross-group dependencies. CDSA and GHCA reuse a shared key projection to maintain a consistent feature reference. Residual feature alignment subsequently integrates the refined features with the original representation. On the Synapse, ACDC, BraTS, and LA datasets, CDGC-Net achieved mean DSC values of 86.96\%, 92.91\%, 82.56\%, and 93.52\%, respectively, exceeding the next-highest reported values by 0.39, 0.47, 0.17, and 0.32 percentage points. CDGC-Net contains 25.83M parameters and 28.62G FLOPs for an input size of $64\times128\times128$, reducing these quantities by 39.87\% and 40.30\%, respectively, relative to UNETR++. These results indicate a favorable trade-off between segmentation accuracy and computational complexity.
Chinese Translation
准确的3D医学图像分割需要将远程解剖上下文与细致的边界细节相结合。现有方法通常在独立的模块或特征层次中建模全局和局部特征，并独立执行通道重校准。这可能导致全局上下文与局部边界之间的语义不匹配、通道关系建模不足、空间-通道交互弱以及冗余表示。我们提出了CDGC-Net，一种结合协同双尺度空间注意力与分组层次通道建模的3D医学图像分割网络。在每个CDGC模块中，协同双尺度自注意力（Cooperative Dual-Scale Self-Attention, CDSA）将注意力头分配给并行的局部窗口和全局稀疏分支。这两个分支在相同的特征层次上捕捉细致的空间细节和远程解剖上下文。它们的输出被连接成一个$N imes C$的空间表示，并直接传递给分组层次通道注意力（Grouped Hierarchical Channel Attention, GHCA）。GHCA将通道组织成$r$个组，并建模组内和组间的依赖关系。CDSA和GHCA重用共享的关键投影以保持一致的特征参考。残差特征对齐随后将精炼的特征与原始表示整合。在Synapse、ACDC、BraTS和LA数据集上，CDGC-Net分别达到了86.96\%、92.91\%、82.56\%和93.52\%的平均DSC值，超过了下一个最高报告值0.39、0.47、0.17和0.32个百分点。CDGC-Net包含25.83M参数和28.62G FLOPs，输入大小为$64 imes128 imes128$，相较于UNETR++分别减少了39.87\%和40.30\%。这些结果表明分割精度与计算复杂性之间存在良好的权衡。
cs.CV / 113 / 2608.08580
Where Is the Bee? Detecting Tiny Pollinators with a Single Collaborative-Head Transformer
蜜蜂在哪里？使用单个协作头变换器检测微小授粉者
Abstract
The CVPPA@ECCV 2026 BuzzSpot Challenge asks us to detect bees, bumblebees, hoverflies, and moths in 1920x1080 field keyframes. Its annotations carry 2 difficulties: the median box occupies 0.16% of a frame, and bees account for 80% of the labels. To cope with the small boxes, we compare 10 recorded detector configurations on held-out keyframes; plain Co-DINO with a Swin-L backbone has the highest mAP in this comparison, so we select it. Training then addresses the bee dominance in 2 ways: fine-tuning on a crop-mosaic pool in which the combined annotation share of the 3 rare classes rises from 19.9% to 55.1%, and a class-weighted simplex equiangular tight frame (ETF) loss that pulls the projected states of matched decoder queries toward fixed class directions. The full schedule spans 12+3+2 epochs. Without inference-time ensembling or test-time augmentation, we rank first on FinalTest at 0.5062 mAP@[.5:.95].
Chinese Translation
CVPPA@ECCV 2026 BuzzSpot挑战要求我们在1920x1080的场景关键帧中检测蜜蜂、大黄蜂、悬停蝇和蛾子。其注释存在两个难点：中位框占据帧的0.16%，而蜜蜂占标签的80%。为了应对小框的挑战，我们在保留的关键帧上比较了10种记录的检测器配置；在这次比较中，采用Swin-L主干的普通Co-DINO获得了最高的mAP，因此我们选择了它。训练阶段通过两种方式解决蜜蜂的主导地位：在一个裁剪-马赛克池上进行微调，使得三种稀有类别的组合注释占比从19.9%上升到55.1%，以及一种类别加权的单纯形等角紧框（ETF）损失，旨在将匹配解码器查询的投影状态拉向固定类别方向。完整的训练计划跨越12+3+2个周期。在没有推理时间集成或测试时间增强的情况下，我们在FinalTest中以0.5062 mAP@[.5:.95]的成绩排名第一。
cs.CV / 114 / 2608.08585
EvTrajGS: Accurate and Efficient 3D Gaussian Splatting from Unposed Event Streams
EvTrajGS：来自未定姿态事件流的准确高效3D高斯点云重建
Abstract
Event cameras, with high temporal resolution, high dynamic range, and asynchronous sensing characteristics, have shown great potential for dense 3D reconstruction. Traditional reconstruction methods based on off-the-shelf pose estimates achieve high efficiency but produce low-fidelity results, as inaccurate pose initialization introduces cumulative reconstruction errors. In contrast, recent SLAM-style methods stabilize joint pose-scene optimization through incremental tracking and mapping, yielding higher reconstruction fidelity at the expense of considerable computational overhead. To address this trade-off, this paper presents EvTrajGS, an accurate and efficient 3D Gaussian Splatting framework for unposed event streams. Our method enables reliable joint pose-scene optimization initialized from coarse pose priors, eliminating the need for computationally expensive SLAM-style pipelines. EvTrajGS parameterizes camera motion as a continuous-time trajectory initialized from discrete camera poses, providing a unified representation for pose refinement. We then aggregate adjacent trajectory states into a temporally coupled pose, promoting temporally consistent pose updates during joint optimization. Additionally, we introduce a loss-reweighted event sampling strategy to adaptively emphasize temporally under-reconstructed intervals. Extensive experiments on both synthetic and real-world datasets demonstrate that EvTrajGS outperforms state-of-the-art methods in terms of both geometric reconstruction quality and pose estimation accuracy, achieving 3.8 dB higher PSNR, 0.1 higher SSIM, and over 40\% lower ATE RMSE while retaining high computational efficiency.
Chinese Translation
事件相机具有高时间分辨率、高动态范围和异步感知特性，在密集3D重建方面展现出巨大的潜力。基于现成姿态估计的传统重建方法虽然效率高，但由于不准确的姿态初始化引入了累积重建误差，导致结果的保真度较低。相比之下，最近的SLAM（同步定位与地图构建）风格方法通过增量跟踪和映射来稳定联合姿态-场景优化，虽然提高了重建的保真度，但代价是显著的计算开销。为了解决这一权衡，本文提出了EvTrajGS，一个针对未定姿态事件流的准确高效3D高斯点云重建框架。我们的方法使得从粗略姿态先验初始化的可靠联合姿态-场景优化成为可能，消除了对计算开销较大的SLAM风格管道的需求。EvTrajGS将相机运动参数化为从离散相机姿态初始化的连续时间轨迹，为姿态精细化提供了统一的表示。然后，我们将相邻轨迹状态汇聚成一个时间耦合的姿态，促进在联合优化过程中时间一致的姿态更新。此外，我们引入了一种损失加权事件采样策略，以自适应地强调时间上重建不足的区间。在合成和真实世界数据集上的大量实验表明，EvTrajGS在几何重建质量和姿态估计准确性方面均优于最先进的方法，PSNR提高了3.8 dB，SSIM提高了0.1，ATE RMSE降低了超过40%，同时保持了高计算效率。
cs.CV / 115 / 2608.08589
RobustDefect-LLM: Explainable and Robustness-Aware Industrial Surface Defect Classification with Decision Support and AI-Assisted Reporting
RobustDefect-LLM：可解释且具鲁棒性意识的工业表面缺陷分类，支持决策与人工智能辅助报告
Abstract
This paper presents RobustDefect-LLM, an industrial surface-defect inspection framework integrating deep-learning classification, operator-facing visual evidence, confidence-aware decision support, controlled AI-assisted reporting, traceable storage, and mobile interaction in a unified quality-control workflow. Here, robustness-aware denotes explicit evaluation under controlled image degradation and confidence-aware review routing, not an intrinsic robustness guarantee. Four transfer-learning-based convolutional neural networks, ResNet50, EfficientNet-B0, DenseNet121, and MobileNetV3-Large, were evaluated on 1,799 images from the six-class NEU-DET dataset using fixed training, validation, and held-out in-domain test partitions. MobileNetV3-Large achieved the highest numerical test accuracy (99.26%) and macro F1-score (0.9926), with a bootstrap 95% accuracy CI of 0.9815-1.0000. An exact paired McNemar test found no significant difference from DenseNet121 (p = 1.000). The selected model averaged 0.060 s per CPU forward pass (16.66 FPS). Under combined synthetic degradation, accuracy fell to 87.78% at mild intensity and below 40% at stronger intensities, revealing sensitivity to severe image-quality deterioration. Grad-CAM supplied visual evidence, while predictions with confidence below 0.90 or a top-2 margin below 0.10 were routed to HUMAN REVIEW. This conservative policy provided 12.22% automatic coverage and 100% observed selective accuracy among 33 eligible cases (95% CI: 89.43%-100.00%), while routing both observed classification errors to review. Under nominal controlled conditions, all 100 generated reports passed deterministic consistency checks, with a mean latency of 1.66 s. Results support the feasibility of the integrated workflow while emphasizing the need for calibration, repeated evaluation, and real-world industrial validation.
Chinese Translation
本文提出了RobustDefect-LLM，一个集深度学习分类、面向操作员的视觉证据、信心意识的决策支持、受控的人工智能辅助报告、可追溯存储和移动交互于一体的工业表面缺陷检测框架，形成统一的质量控制工作流程。这里的鲁棒性意识指在受控图像降质下的明确评估和信心意识的审核路由，而非内在的鲁棒性保证。我们对来自六类NEU-DET数据集的1,799幅图像进行了评估，使用了四种基于迁移学习的卷积神经网络：ResNet50、EfficientNet-B0、DenseNet121和MobileNetV3-Large，采用固定的训练、验证和保留的领域内测试分区。MobileNetV3-Large达到了最高的数值测试准确率（99.26%）和宏观F1分数（0.9926），其自助法95%准确率置信区间为0.9815-1.0000。精确的配对McNemar检验未发现与DenseNet121之间的显著差异（p = 1.000）。所选模型的每次CPU前向传播平均耗时0.060秒（16.66 FPS）。在合成降质的影响下，准确率在轻度强度下降至87.78%，在更强强度下降至40%以下，显示出对严重图像质量恶化的敏感性。Grad-CAM提供了视觉证据，而信心低于0.90或前两名边际低于0.10的预测被路由至人工审核。该保守政策在33个符合条件的案例中提供了12.22%的自动覆盖率和100%的观察选择准确率（95% CI: 89.43%-100.00%），同时将观察到的分类错误都路由至审核。在名义受控条件下，所有100份生成的报告均通过了确定性一致性检查，平均延迟为1.66秒。结果支持了集成工作流程的可行性，同时强调了校准、重复评估和实际工业验证的必要性。
cs.CV / 116 / 2608.08596
Goal-oriented Navigation Instruction Generation with Tour Video Priors
基于目标的导航指令生成与旅游视频先验
Abstract
Navigation Instruction Generation (NIG) aims to produce step-by-step natural language instructions for navigation guidance. Existing studies primarily treat NIG as an auxiliary task for vision-andlanguage navigation (VLN), focusing on data augmentation or multi-task learning. However, generating navigation instructions from compact environmental priors requires meticulous spatial reasoning, especially when the target route does not simply follow the demonstrated tour, and remains challenging for current multimodal models. In this work, we introduce VideoNIG, a goal-oriented video-grounded NIG task that generates navigation instructions from ego-centric tour videos, an initial observation, and a textual or visual goal, without relying on intermediate representations such as graphs and maps. We instantiate VideoNIG in a controlled simulator benchmark with 60K tour videos across continuous indoor environments and 37K multimodal prompts with progressive difficulty levels. We further introduce a diagnostic evaluation protocol that combines text similarity, choice-based spatial consistency tests, and downstream navigation execution. To address this task, we propose a two-stage Curriculum Learning framework that decomposes the learning into foundational motion perception and long-horizon navigation reasoning. Specifically, we first employ Action Warmup for spatial action-view alignment, followed by Complexity Progression using trajectories with increasing exploratory difficulty. Extensive experiments show that existing MLLMs struggle with VideoNIG, while our approach significantly improves instruction quality across complementary diagnostic metrics. Finally, integrating VideoNIG-generated instructions with a VLN agent demonstrates the executability of this task formulation for end-to-end navigation.
Chinese Translation
导航指令生成（NIG）旨在为导航指导生成逐步的自然语言指令。现有研究主要将NIG视为视觉与语言导航（VLN）的辅助任务，侧重于数据增强或多任务学习。然而，从紧凑的环境先验中生成导航指令需要细致的空间推理，尤其是在目标路线并不简单遵循演示的旅游路线时，这对当前的多模态模型仍然是一个挑战。在本研究中，我们引入了VideoNIG，一个基于目标的视频引导NIG任务，该任务从自我中心的旅游视频、初始观察以及文本或视觉目标生成导航指令，而无需依赖图形和地图等中间表示。我们在一个受控的模拟基准中实例化了VideoNIG，涵盖60K个旅游视频和37K个具有逐步难度的多模态提示。我们进一步引入了一种诊断评估协议，结合文本相似性、基于选择的空间一致性测试和下游导航执行。为了解决这一任务，我们提出了一个两阶段的课程学习框架，将学习分解为基础运动感知和长时间导航推理。具体而言，我们首先采用动作热身（Action Warmup）进行空间动作视图对齐，然后使用复杂性进展（Complexity Progression）处理具有逐步探索难度的轨迹。大量实验表明，现有的多模态语言模型（MLLMs）在VideoNIG任务上表现不佳，而我们的方法在互补的诊断指标上显著提高了指令质量。最后，将VideoNIG生成的指令与VLN代理集成，展示了这一任务形式在端到端导航中的可执行性。
cs.CV / 117 / 2608.08600
Population-Scalable Multi-Agent World Modeling
可扩展的人口多智能体世界建模
Abstract
World models have recently achieved impressive progress in visual prediction and interactive generation, but extending them to multi-agent environments introduces a fundamental scalability challenge. Existing methods generally assume a fixed number of agents during training and inference, which ties the model to a pre-determined agent population and limits inference-time scalability. Our key insight is that cross-view consistency should arise from a shared world state whose evolution does not assume a predefined number of agents, while agent-specific observations should be generated by querying this state through a unified rendering interface. Based on this insight, we propose Khora, a scalable multi-agent world model that supports inference-time expansion to arbitrary numbers of agents without retraining. Our framework decouples world-state evolution from visual rendering and introduces a population-agnostic rendering mechanism for incorporating other agent information. This design maintains cross-view consistency through the shared world state rather than through dense interactions among observation streams inside the expensive video generator, enabling approximately linear practical scaling with the number of queried views. Qualitative experiments demonstrate that our approach generalizes to unseen numbers of agents while maintaining visual quality and multi-agent consistency. We further implement a real-time interactive system to demonstrate scalable open-world simulation.
Chinese Translation
世界模型最近在视觉预测和交互生成方面取得了显著进展，但将其扩展到多智能体环境中引入了根本性的可扩展性挑战。现有方法通常在训练和推理过程中假设固定数量的智能体，这使得模型与预先确定的智能体数量绑定，并限制了推理时的可扩展性。我们的关键见解是，跨视角一致性应源于一个共享的世界状态，其演变不假设预定义的智能体数量，而特定于智能体的观察应通过统一的渲染接口查询该状态生成。基于这一见解，我们提出了Khora，一个可扩展的多智能体世界模型，支持在不重新训练的情况下进行推理时扩展到任意数量的智能体。我们的框架将世界状态的演变与视觉渲染解耦，并引入了一种与人口无关的渲染机制，以整合其他智能体的信息。该设计通过共享的世界状态而非通过在昂贵的视频生成器内部观察流之间的密集交互来维持跨视角一致性，从而实现与查询视角数量近似线性的实际可扩展性。定性实验表明，我们的方法在保持视觉质量和多智能体一致性的同时，能够推广到未见过的智能体数量。我们进一步实现了一个实时交互系统，以展示可扩展的开放世界模拟。
cs.CV / 118 / 2608.08612
REVEAL: A Rubric-Guided Agent for Explicit Evidence Sufficiency Verificationin Long-Video Question Answering
REVEAL：一种基于评分标准的代理，用于长视频问答中的显式证据充分性验证
Abstract
Recently, retrieval-augmented and memory-augmented methods have emerged as two promising paradigms for long-video question answering. However, existing methods typically rely on rigid, fixed-length temporal chunking (e.g., 10s) and static offline memory banks, which not only fragment coherent continuous events but also fail to adapt during real-time reasoning. Moreover, whether using multi-scale summaries or multimodal knowledge graphs, current approaches prioritize retrieval relevance while overlooking evidence sufficiency, often stopping to answer once only semantically relevant clues are retrieved, even when key temporal, causal, or fine-grained action evidence is still missing. To tackle these challenges, we propose REVEAL, a rubric-guided agent framework. As a foundation, we introduce an adaptive visual-similarity-based preprocessing pipeline that groups visually coherent adjacent frames into natural event units to construct an offline-online video memory---capturing global video context offline while dynamically maintaining question-conditioned memory online. Built upon this structured memory, REVEAL uses an automatically constructed rubric library to explicitly verify whether retrieved evidence satisfies sufficiency criteria, pinpoints missing clues upon verification failure, and directs targeted re-retrieval for complementary information. Without any extra training, REVEAL consistently outperforms both closed-source and open-source state-of-the-art methods across extensive experiments. These results show that explicitly verifying evidence sufficiency, rather than stopping at semantic relevance, retrieves the decisive clues that prior methods miss and yields more reliable long-video reasoning.
Chinese Translation
最近，增强检索和增强记忆的方法作为长视频问答的两种有前景的范式应运而生。然而，现有方法通常依赖于刚性、固定长度的时间切片（例如，10秒）和静态的离线记忆库，这不仅会将连贯的连续事件分割开来，还无法在实时推理中进行适应。此外，无论是使用多尺度摘要还是多模态知识图谱，当前的方法都优先考虑检索相关性，而忽视了证据的充分性，往往在仅检索到语义相关线索后就停止回答，即使关键的时间、因果或细粒度的行动证据仍然缺失。为了解决这些挑战，我们提出了REVEAL，一个基于评分标准的代理框架。作为基础，我们引入了一种自适应的基于视觉相似性的预处理管道，将视觉上连贯的相邻帧分组为自然事件单元，以构建一个离线-在线视频记忆——在离线捕捉全局视频上下文的同时，动态维护基于问题的在线记忆。在这个结构化记忆的基础上，REVEAL使用自动构建的评分标准库显式验证检索到的证据是否满足充分性标准，在验证失败时指出缺失的线索，并指导有针对性的重新检索以获取补充信息。在没有任何额外训练的情况下，REVEAL在广泛的实验中始终优于闭源和开源的最新方法。这些结果表明，显式验证证据的充分性，而不是停留在语义相关性上，可以检索到先前方法遗漏的决定性线索，并产生更可靠的长视频推理。
cs.CV / 119 / 2608.08622
VADER: Adaptive Debiasing for Hallucination Mitigation in Video Large Language Models
VADER：用于视频大型语言模型幻觉缓解的自适应去偏见
Abstract
Large vision-language models (LVLMs) have demonstrated strong performance in open-ended video understanding, yet they remain prone to fluent responses unsupported by video evidence. Existing training-free methods typically apply a globally fixed visual intervention or construct a contrastive branch through input perturbation. The former cannot accommodate video-dependent fusion paths, while the latter can be compensated by cross-frame redundancy. We therefore propose Video-Adaptive Debiasing via Evidence Reweighting (VADER), a training-free framework with two complementary modules. Visual Focus Reallocation (VFR) automatically instantiates an intervention policy for each video-question input: it diagnoses layer-wise visual-to-text evidence flow, determines where to intervene, and derives how strongly to reallocate pre-softmax attention from system-token to video-token blocks. Selective Evidence Erasure (SEE) independently masks high-importance visual tokens in every frame, constructing a prior-biased branch that is difficult to compensate through neighboring frames. Contrastive decoding then down-weights predictions that remain confident after selective evidence erasure. Across multiple VideoLLMs, VADER yields substantial improvements on event-level grounding and temporal consistency; on LLaVA-Video-7B, it reaches 72.60% accuracy on EventHallusion.
Chinese Translation
大型视觉语言模型（LVLMs）在开放式视频理解方面表现出色，但它们仍然容易产生与视频证据不符的流畅响应。现有的无训练方法通常应用全局固定的视觉干预或通过输入扰动构建对比分支。前者无法适应视频依赖的融合路径，而后者则可以通过跨帧冗余进行补偿。因此，我们提出了通过证据重加权的 视频自适应去偏见（VADER），这是一个无训练框架，包含两个互补模块。视觉焦点重新分配（VFR）自动为每个视频-问题输入实例化干预策略：它诊断层级的视觉到文本证据流，确定干预的位置，并推导如何强烈地将预软最大注意力从系统标记重新分配到视频标记块。选择性证据擦除（SEE）独立地在每帧中屏蔽高重要性的视觉标记，构建一个难以通过相邻帧进行补偿的先验偏置分支。对比解码随后降低在选择性证据擦除后仍然自信的预测的权重。在多个视频大型语言模型（VideoLLMs）上，VADER在事件级基础和时间一致性方面取得了显著提升；在LLaVA-Video-7B上，其在EventHallusion任务中达到了72.60%的准确率。
cs.CV / 120 / 2608.08630
VLZip: Unified Visual and Textual Compression for Interleaved Long-Context Modeling
VLZip：用于交错长上下文建模的统一视觉与文本压缩
Abstract
Vision Language Models (VLMs) face significant challenges with ultra-long, interleaved image-text sequences due to the quadratic complexity of self-attention. Current solutions either resort to aggressive token pruning, risking irreversible information loss, or adopt efficient but less precise architectures, while largely ignoring the equally vital textual component. We introduce VLZip, a framework that unifies visual and textual compression for high-fidelity reasoning within a pure Transformer. At its core, VLZip hierarchically distills visual and textual segments into compact, layer-specific "soft prefixes" and injects them into each decoder layer's hidden states, drastically shortening the attention sequence while preserving fine-grained global context. To address deficient evaluations in the field, we also introduce LongVLBench, a new benchmark derived from video narratives that demands holistic, narrative-level reasoning. Extensive experiments show VLZip achieves leading performance on long-context multimodal reasoning, enabling training up to 120K tokens, a 6x increase over the baseline, and inference beyond 280K tokens with significantly reduced memory, while demonstrating the memory scalability to handle up to 2M tokens. By excelling at extreme context lengths where existing methods collapse, VLZip establishes an efficient and powerful new standard for long-context multimodal AI. Code is available at https://github.com/ShareLab-SII/VLZip.
Chinese Translation
视觉语言模型（VLMs）在处理超长的交错图像-文本序列时面临显著挑战，主要由于自注意力机制的二次复杂性。目前的解决方案要么依赖于激进的标记剪枝，冒着不可逆信息丢失的风险，要么采用高效但精度较低的架构，同时在很大程度上忽视了同样重要的文本组件。我们提出了VLZip，一个统一视觉与文本压缩的框架，旨在在纯Transformer中实现高保真推理。VLZip的核心是将视觉和文本片段分层提炼为紧凑的、层特定的“软前缀”，并将其注入到每个解码器层的隐藏状态中，从而显著缩短注意力序列，同时保留细粒度的全局上下文。为了解决该领域评估不足的问题，我们还引入了LongVLBench，这是一个基于视频叙事的新基准，要求进行整体的叙事级推理。大量实验表明，VLZip在长上下文多模态推理中表现出领先的性能，支持训练高达120K标记，是基线的6倍，并且在显著减少内存的情况下进行推理，超出280K标记，同时展示了处理高达2M标记的内存可扩展性。通过在现有方法崩溃的极端上下文长度下表现优异，VLZip为长上下文多模态人工智能建立了一个高效而强大的新标准。代码可在 https://github.com/ShareLab-SII/VLZip 获取。
cs.CV / 121 / 2608.08648
Agentic Visual Reasoning in Whole-Slide Pathology Images via Active Perception
通过主动感知实现全幻灯片病理图像中的能动视觉推理
Abstract
Whole-slide visual reasoning requires identifying sparse diagnostic evidence in gigapixel pathology slides and integrating observations across spatial scales. Existing WSI methods either compress densely sampled patches into global representations or use pretrained vision-language models with heuristic region selection, weakening links between predictions and morphology or lacking pathology-trained observation policies. We present AdaptivePath, an active-perception framework that formulates WSI evidence acquisition as sequential decision making. The Navigator learns question-agnostic abnormality-driven navigation from pathologist-reviewed labels to select observation locations and spatial extents, avoiding costly question-specific trajectory annotations. We train this policy through alternating representation learning and proximal policy optimization, followed by fine-tuning with geometric and appearance consistency objectives to stabilize focus trajectories. During inference, the Navigator hierarchically acquires sparse observations from low to high magnification under a limited ROI budget. A Morphology Interpreter converts observations into question-conditioned evidence, while the Deliberator evaluates evidence and revises intermediate answers across magnifications. The Arbiter integrates deliberation history to produce final answers. AdaptivePath achieves state-of-the-art zero-shot performance on WSI and region pathology VQA benchmarks and reaches 80.14% accuracy for cancer subtype classification across six TCGA cohorts. In a blinded diagnostic-utility study, pathologists using AdaptivePath-selected observation sequences achieve 82.9% accuracy. These results demonstrate that learned active perception enables effective and traceable visual reasoning over gigapixel pathology slides.
Chinese Translation
全幻灯片视觉推理需要在千兆像素病理幻灯片中识别稀疏的诊断证据，并在空间尺度上整合观察。现有的全幻灯片图像（WSI）方法要么将密集采样的图像块压缩为全局表示，要么使用预训练的视觉-语言模型进行启发式区域选择，从而削弱了预测与形态之间的联系，或缺乏经过病理训练的观察策略。我们提出了AdaptivePath，一个将WSI证据获取形式化为顺序决策的主动感知框架。导航器（Navigator）从病理学家审核的标签中学习与问题无关的异常驱动导航，以选择观察位置和空间范围，避免了成本高昂的问题特定轨迹注释。我们通过交替的表示学习和近端策略优化训练该策略，随后通过几何和外观一致性目标进行微调，以稳定焦点轨迹。在推理过程中，导航器在有限的感兴趣区域（ROI）预算下，从低放大倍数到高放大倍数层次性地获取稀疏观察。形态解释器（Morphology Interpreter）将观察结果转换为条件于问题的证据，而审议者（Deliberator）则评估证据并在不同放大倍数下修正中间答案。仲裁者（Arbiter）整合审议历史以生成最终答案。AdaptivePath在WSI和区域病理视觉问答（VQA）基准上实现了最先进的零-shot性能，并在六个TCGA队列中达到了80.14%的癌症亚型分类准确率。在一项盲法诊断效用研究中，使用AdaptivePath选择的观察序列的病理学家达到了82.9%的准确率。这些结果表明，学习的主动感知使得在千兆像素病理幻灯片上进行有效且可追溯的视觉推理成为可能。
cs.CV / 122 / 2608.08659
JSGS: JPEG State-Guided Supervision for 3D Gaussian Splatting from Mixed-Quality Views
JSGS：基于JPEG状态指导的混合质量视图下的3D高斯点云监督
Abstract
Standard 3D Gaussian Splatting (3DGS) assumes that every input image faithfully samples scene radiance. However, mixed-quality JPEG images violate this assumption because compression-induced blocking and ringing artifacts can corrupt updates to Gaussians shared across views. To address this problem, we propose JPEG State-Guided Supervision for 3D Gaussian Splatting from Mixed-Quality Views (JSGS). JSGS uses luminance and chrominance quantization tables stored in each JPEG file to construct a view-specific JPEG observation operator. This operator encodes and decodes each rendered view for domain-matched comparison with the corresponding decoded input image. The luminance quantization table supplies continuous weights within a fixed middle frequency band. A loss in the low frequency band anchors coarse structure, while the weighted middle frequency loss redistributes supervision among the selected DCT coordinates. The resulting block disagreement also guides the Gaussian Controller to regularize small primitives with high opacity in disagreement regions. Across seven scenes and three mixed-quality schedules, JSGS achieves the lowest mean LPIPS and the highest mean SSIM under every schedule while rendering at approximately 150 FPS. Code: https://github.com/Jayden-Cui/JSGS.
Chinese Translation
标准的3D高斯点云（3DGS）假设每个输入图像都忠实地采样场景辐射。然而，混合质量的JPEG图像违反了这一假设，因为压缩引起的块效应和振铃伪影可能会破坏跨视图共享的高斯更新。为了解决这个问题，我们提出了基于JPEG状态指导的混合质量视图下的3D高斯点云监督（JSGS）。JSGS利用存储在每个JPEG文件中的亮度和色度量化表构建特定于视图的JPEG观测算子。该算子对每个渲染视图进行编码和解码，以便与相应的解码输入图像进行领域匹配的比较。亮度量化表在固定的中频带内提供连续的权重。低频带的损失锚定粗略结构，而加权的中频损失则在选定的DCT坐标之间重新分配监督。由此产生的块不一致性还指导高斯控制器在不一致区域内对高不透明度的小原件进行正则化。在七个场景和三个混合质量调度下，JSGS在每个调度下都实现了最低的平均LPIPS和最高的平均SSIM，同时以约150 FPS的速度进行渲染。代码：https://github.com/Jayden-Cui/JSGS。
cs.CV / 123 / 2608.08661
Degradation-Guided Underwater Image Restoration with Task-Oriented Latent Control
基于退化引导的水下图像恢复与任务导向的潜在控制
Abstract
Degradation information in underwater images plays a dual role: its spatial and spectral cues can guide adaptive restoration, while degradation-entangled features may be propagated without explicit regulation during decoding. Existing methods largely overlook this dual role, either underexploiting degradation cues or directly forwarding encoder features through skip connections. To address this issue, we propose PROTEUS, which couples degradation-guided feature adaptation with task?oriented latent control. PROTEUS tackles this problem from two complementary perspectives. At the feature level, the Guided Dynamic Feature Modulation Block exploits spatially varying degradation cues to adapt feature processing across network stages. At the representation level, the task-oriented latent controller learns a structured control code under discriminative regularisation and uses it for channel-wise modulation of skip features, without requiring the code to form a metrically cleaner embedding. Extensive experiments on five paired and four non-reference underwater benchmarks demonstrate that PROTEUS achieves highly competitive restoration performance, with a favourable balance between restoration quality and computational cost.
Chinese Translation
水下图像中的退化信息发挥着双重作用：其空间和光谱线索可以指导自适应恢复，而退化纠缠特征在解码过程中可能在没有明确调节的情况下被传播。现有方法在很大程度上忽视了这一双重作用，要么未充分利用退化线索，要么通过跳跃连接直接转发编码器特征。为了解决这一问题，我们提出了PROTEUS，它将退化引导的特征适应与任务导向的潜在控制相结合。PROTEUS从两个互补的角度解决了这个问题。在特征层面，指导动态特征调制块利用空间变化的退化线索来适应网络各阶段的特征处理。在表示层面，任务导向的潜在控制器在判别正则化下学习结构化控制代码，并利用该代码对跳跃特征进行通道级调制，而无需该代码形成度量上更清晰的嵌入。在五个配对和四个非参考水下基准上的大量实验表明，PROTEUS实现了高度竞争的恢复性能，在恢复质量和计算成本之间达到了良好的平衡。
cs.CV / 124 / 2608.08664
FiRe: Fixed-Noise Refinement for Visual Counterfactual Explanations
FiRe：用于视觉反事实解释的固定噪声精炼
Abstract
Visual counterfactual explanations aim to change classifier decisions through realistic and localized edits while preserving decision-irrelevant content. Existing DDPM-based methods typically perform classifier-guided editing along a long reverse denoising trajectory. The changing noise levels make semantic editability and spatial control difficult to balance, and the editable state is noisy, whereas the target classifier is trained on clean images. As a result, these methods require either costly recursive denoising or low-quality one-step estimates to obtain classifier-facing clean images. We propose FiRe, a Fixed-noise Refinement framework for visual counterfactual explanations. Rather than following a reverse denoising trajectory, FiRe maps the input to a fixed noise level and iteratively refines the noisy state at that level. To provide clean images for classifier guidance, FiRe first adapts Pixel Mean Flow to visual counterfactual explanation, enabling direct clean-image prediction from noisy states. To make fixed-noise refinement produce minimal and localized counterfactual edits, FiRe introduces three FiRe-specific controls: a dynamic dual-mask strategy, adaptive guidance, and early stopping, which determine where edits accumulate, which changes become visible, and when refinement stops. Experiments on five tasks across three datasets show that, compared with the strongest recent baseline, FiRe achieves about 3$\times$ faster online inference and 8$\times$ fewer FLOPs while obtaining comparable or state-of-the-art counterfactual quality.
Chinese Translation
视觉反事实解释旨在通过现实且局部的编辑来改变分类器的决策，同时保留与决策无关的内容。现有的基于DDPM（去噪扩散概率模型）的方法通常沿着较长的反向去噪轨迹进行分类器引导的编辑。变化的噪声水平使得语义可编辑性和空间控制难以平衡，且可编辑状态是噪声的，而目标分类器是在干净图像上训练的。因此，这些方法要么需要昂贵的递归去噪，要么依赖低质量的一步估计来获得面向分类器的干净图像。我们提出了FiRe，一种用于视觉反事实解释的固定噪声精炼框架。FiRe并不遵循反向去噪轨迹，而是将输入映射到固定的噪声水平，并在该水平上迭代精炼噪声状态。为了提供用于分类器引导的干净图像，FiRe首先将像素均值流（Pixel Mean Flow）适配到视觉反事实解释，使得能够从噪声状态直接预测干净图像。为了使固定噪声精炼产生最小且局部的反事实编辑，FiRe引入了三个特定于FiRe的控制策略：动态双掩码策略、自适应引导和提前停止，这些策略决定了编辑的累积位置、哪些变化变得可见以及何时停止精炼。在三个数据集上的五个任务的实验表明，与最强的近期基线相比，FiRe在在线推理速度上快约3倍，并且FLOPs减少约8倍，同时获得了可比或最先进的反事实质量。
cs.CV / 125 / 2608.08676
UniSpace: Unified Visual Representation and Scalable Multimodal Modeling
UniSpace：统一视觉表示与可扩展多模态建模
Abstract
Semantic vision encoders have become a central visual interface for multimodal understanding and semantic conditioning in image generation. However, their final tokens discard fine-grained visual details, leading to poor pixel reconstruction and limiting their use in reconstruction-sensitive tasks such as image generation and editing. In this work, we ask whether understanding, generation, and editing can be modeled in a single visual representation space built from a pretrained semantic ViT. We show that the frozen Transformer blocks of a semantic ViT are not intrinsically unable to preserve visual details. Instead, the original patch parameterization drives the representation toward semantic abstraction, making fine-grained information difficult to recover from the final tokens. Based on this observation, we introduce \emph{Patch Reparameterization}, which preserves the original semantic pathway while adding a reconstruction-aware patch embedding that provides fine-grained visual information to the same frozen ViT blocks. The resulting unified representation preserves multimodal understanding while enabling high-fidelity image reconstruction and a favorable reconstruction--generation trade-off. We further scale this representation into \emph{UniSpace}, an 8B Mixture-of-Transformer-Experts model that performs understanding, generation, and editing in the same visual space without a separate VAE pathway. System-level evaluations demonstrate practical text-to-image generation and instruction-based image editing, showing that a reparameterized pretrained ViT can serve as a unified visual interface for scalable multimodal modeling.
Chinese Translation
语义视觉编码器已成为多模态理解和图像生成中的语义条件的重要视觉接口。然而，它们的最终标记丢弃了细粒度的视觉细节，导致像素重建效果不佳，并限制了它们在对重建敏感的任务（如图像生成和编辑）中的应用。在本研究中，我们探讨是否可以在一个由预训练的语义视觉变换器（ViT）构建的单一视觉表示空间中建模理解、生成和编辑。我们展示了语义ViT的冻结变换器块并非本质上无法保留视觉细节。相反，原始的补丁参数化将表示驱动向语义抽象，使得从最终标记中恢复细粒度信息变得困难。基于这一观察，我们引入了 extit{补丁重新参数化}，它在保留原始语义路径的同时，增加了一个重建感知的补丁嵌入，为相同的冻结ViT块提供细粒度的视觉信息。最终得到的统一表示在保留多模态理解的同时，实现了高保真图像重建和良好的重建-生成权衡。我们进一步将这一表示扩展为 extit{UniSpace}，一个具有80亿参数的混合变换器专家模型，能够在同一视觉空间中进行理解、生成和编辑，而无需单独的变分自编码器（VAE）路径。系统级评估展示了实用的文本到图像生成和基于指令的图像编辑，表明重新参数化的预训练ViT可以作为可扩展多模态建模的统一视觉接口。
cs.CV / 126 / 2608.08685
Semi-Dense Matching Uncertainty Is Not Just Local Confidence
半稠密匹配的不确定性不仅仅是局部置信度
Abstract
Reliable semi-dense matching is essential for modern geometric vision systems. Designed under a coarse-to-fine paradigm, it achieves an optimal balance between performance and computational cost. However, existing methods often struggle to provide well-quantified uncertainties, where catastrophic coarse-assignment failures are ignored, leading to truncated error distributions and severely misjudged geometric estimations. In this paper, we propose a lightweight, post-hoc overall uncertainty estimation framework that introduces a two-component calibrated Laplace mixture model with only 9 learnable parameters. The objective is to explicitly capture both the sharp local refinement noise and the broader tail of coarse-assignment failures. We introduce the Coarse-success posterior Refit (CoRe) method, a geometric refitting module that utilizes the posterior probability of coarse-assignment success as soft correspondence weights. Extensive experiments show that our method consistently improves downstream geometric accuracy across various pretrained-only matchers and robust estimators with minimal computational overhead. Our code is available at https://github.com/khoavpt/Probabilistic-matching.
Chinese Translation
可靠的半稠密匹配对于现代几何视觉系统至关重要。该方法在粗到细的范式下设计，能够在性能和计算成本之间实现最佳平衡。然而，现有方法往往难以提供良好量化的不确定性，忽视了灾难性的粗分配失败，导致错误分布被截断以及几何估计严重失真。本文提出了一种轻量级的后处理整体不确定性估计框架，引入了一个仅包含9个可学习参数的双组件校准拉普拉斯混合模型。其目标是明确捕捉尖锐的局部细化噪声和粗分配失败的更广泛尾部。我们引入了粗成功后验重拟合（Coarse-success posterior Refit, CoRe）方法，这是一种几何重拟合模块，利用粗分配成功的后验概率作为软对应权重。大量实验表明，我们的方法在各种仅预训练的匹配器和鲁棒估计器中，始终能在最小计算开销下提高下游几何准确性。我们的代码可在 https://github.com/khoavpt/Probabilistic-matching 获取。
cs.CV / 127 / 2608.08693
CUPA-T2*: Covariance-Aware Uncertainty Propagation and Alignment for T2* Mapping in Accelerated MRI
CUPA-T2*: 考虑协方差的不确定性传播与对齐在加速MRI中的T2*映射
Abstract
Quantitative T2* maps have strong potential for biomarker discovery but are limited by long scan times, rendering them impractical in clinical settings. Significant acceleration can be achieved through undersampling in k-space combined with learning-based reconstruction. However, reconstruction artifacts and noise can propagate into downstream T2* fitting, degrading its accuracy. We introduce CUPA-T2*, a framework that explicitly propagates voxel-wise inter-echo uncertainty from stochastic Monte Carlo dropout reconstructions to downstream T2* fitting via covariance-aware sampling. T2* fitting is performed with a heteroscedastic MLP and a correlation-based regularizer that encourages alignment between predicted variance and reconstruction uncertainty. Experiments on accelerated brain MRI data show tissue-dependent behavior: CUPA-T2* achieves competitive overall T2* fitting performance and improves white-matter performance at higher accelerations. Compared with a heteroscedastic baseline, the proposed framework substantially increases alignment between reconstruction uncertainty and predicted T2* variance, while also revealing a trade-off with calibration (ECE) and selective prediction performance (AURC). CUPA-T2* enables reconstruction uncertainty-aware T2* fitting and delivers voxel-wise uncertainty maps to support the interpretation of quantitative T2* estimates.
Chinese Translation
定量T2*图在生物标志物发现中具有很强的潜力，但由于扫描时间较长，限制了其在临床环境中的应用。通过在k空间中进行欠采样并结合基于学习的重建，可以实现显著加速。然而，重建伪影和噪声可能会传播到下游的T2*拟合中，从而降低其准确性。我们提出了CUPA-T2*，这是一个框架，通过考虑协方差的采样，明确地将来自随机蒙特卡洛丢弃重建的体素间回波不确定性传播到下游的T2*拟合中。T2*拟合使用异方差的多层感知器（MLP）和基于相关性的正则化器进行，以鼓励预测方差与重建不确定性之间的对齐。在加速脑MRI数据上的实验显示出组织依赖性行为：CUPA-T2*在整体T2*拟合性能上具有竞争力，并在更高加速下改善了白质的表现。与异方差基线相比，所提出的框架显著增加了重建不确定性与预测T2*方差之间的对齐，同时也揭示了与校准（ECE）和选择性预测性能（AURC）之间的权衡。CUPA-T2*使得重建不确定性感知的T2*拟合成为可能，并提供体素级不确定性图，以支持定量T2*估计的解释。
cs.CV / 128 / 2608.08696
OccAnyScene: Towards Unified Indoor-Outdoor 3D Occupancy Predictio
OccAnyScene：统一室内外3D占用预测的探索
Abstract
3D occupancy prediction is fundamental to scene understanding, yet existing 3D semantic occupancy methods are typically specialized to fixed scene types and occupancy protocols. We introduce Cross-Scene 3D Semantic Occupancy Prediction, a new task setting which requires a single model to handle heterogeneous indoor and outdoor scenes with varying cameras, spatial ranges, voxel specifications, and semantic taxonomies. This setting poses a fundamental challenge: achieving metric-consistent yet scene-adaptive image-to-3D lifting across varying camera configurations and scene scales. To address this challenge, we propose OccAnyScene, a pixel-frustum-centered Gaussian framework built upon a pretrained depth foundation model. Specifically, the framework employs Pixel-Aligned Frustum Feature Aggregation to construct a camera-aware frustum query for each feature pixel, and Frustum-Parameterized Gaussian Construction to decode each query into multiple Gaussians whose positions and sizes are constrained by the predicted pixel depth and corresponding frustum geometry. OccAnyScene sets new state-of-the-art results, achieving 59.92% mIoU on the indoor Occ-ScanNet and 23.06% mIoU on the outdoor SurroundOcc-nuScenes.
Chinese Translation
3D占用预测是场景理解的基础，但现有的3D语义占用方法通常专门针对固定场景类型和占用协议。我们提出了跨场景3D语义占用预测（Cross-Scene 3D Semantic Occupancy Prediction），这是一种新的任务设置，要求单一模型处理具有不同相机、空间范围、体素规格和语义分类的异构室内外场景。该设置提出了一个基本挑战：在不同的相机配置和场景尺度下，实现度量一致且适应场景的图像到3D的提升。为了解决这一挑战，我们提出了OccAnyScene，这是一种基于预训练深度基础模型的像素锥体中心高斯框架。具体而言，该框架采用像素对齐锥体特征聚合（Pixel-Aligned Frustum Feature Aggregation）为每个特征像素构建一个相机感知的锥体查询，并通过锥体参数化高斯构造（Frustum-Parameterized Gaussian Construction）将每个查询解码为多个高斯，其位置和大小受预测的像素深度和相应的锥体几何约束。OccAnyScene在室内Occ-ScanNet上实现了59.92%的mIoU，在室外SurroundOcc-nuScenes上实现了23.06%的mIoU，创造了新的最先进结果。
cs.CV / 129 / 2608.08702
SRE-FER: Regional residual evidence learning for mitigating local evidence dilution in fine-grained facial expression recognition
SRE-FER：区域残差证据学习以减轻细粒度面部表情识别中的局部证据稀释
Abstract
Fine-grained facial expression recognition (FER) hinges on capturing subtle muscular cues that distinguish adjacent emotions. Yet capturing these cues presents a dilemma. Detector-based methods depend on fragile landmark pipelines, whereas we find that directly transferring foundation models such as DINOv3 under conventional global readouts can cause local evidence dilution: early global aggregation washes out sparse muscular signals and leaves persistent confusion between categories such as fear/surprise and sad/neutral. To recover this evidence, we propose SRE-FER, a readout-level regional residual evidence learning framework. Its core module, RERA, adds zero-initialized residual logits that refine class boundaries while preserving the backbone's global prediction. Training-time action unit (AU) guidance steers regional features toward expression-relevant areas using Facial Action Coding System (FACS)-based anatomical priors, without requiring an external facial pipeline at inference. An optional Full setting further routes sample-specific non-redundant tokens. On three benchmarks, SRE-FER attains 92.76% on RAF-DB, 91.32% on FERPlus, and 67.78% on AffectNet-7, demonstrating highly competitive performance compared to existing FER methods.
Chinese Translation
细粒度面部表情识别（FER）依赖于捕捉细微的肌肉线索，以区分相邻的情感。然而，捕捉这些线索却面临困境。基于检测器的方法依赖于脆弱的地标管道，而我们发现，在传统的全局读取下，直接转移基础模型（如 DINOv3）可能导致局部证据稀释：早期的全局聚合会冲淡稀疏的肌肉信号，并在恐惧/惊讶和悲伤/中性等类别之间留下持续的混淆。为了恢复这些证据，我们提出了 SRE-FER，一种读取级区域残差证据学习框架。其核心模块 RERA 添加了零初始化的残差 logits，以在保留主干网络全局预测的同时，细化类别边界。训练时的动作单元（AU）指导利用基于面部动作编码系统（FACS）的解剖先验，引导区域特征朝向与表情相关的区域，而无需在推理时依赖外部面部管道。可选的完整设置进一步路由样本特定的非冗余标记。在三个基准测试中，SRE-FER 在 RAF-DB 上达到了 92.76%，在 FERPlus 上达到了 91.32%，在 AffectNet-7 上达到了 67.78%，显示出与现有 FER 方法相比具有高度竞争力的表现。
cs.CV / 130 / 2608.08713
Resolution Meets Reduction: Efficient Visual Context for 3D Radiology Report Generation
分辨率与压缩的结合：用于3D放射学报告生成的高效视觉上下文
Abstract
Vision-language models offer a promising path toward automating radiology report generation, but applying them to full 3D CT volumes poses substantial computational challenges. Modern foundation vision encoders (VEs) can produce tens of thousands of vision tokens per scan, making the visual sequence passed to the large language model (LLM) a primary computational bottleneck. Vision-to-language projectors can compress this sequence to reduce computation, but may discard clinically relevant detail; conversely, effective compression can accommodate higher-resolution inputs while keeping the downstream token count fixed. How this vision-token budget should be allocated across input field of view, spatial resolution, and vision-to-language projection therefore remains an open design question. We systematically evaluate four heterogeneous VEs (CNN- and ViT-based), five token-reducing projectors at up to 64x compression alongside a non-reducing MLP projector baseline, and five instruction-tuned LLMs (1.7B--4B) on two large-scale CT report datasets (CT-RATE and Merlin). At matched LLM token budgets, anatomy-guided region of interest cropping is the most consistent strategy, improving clinical macro F1 in 19 of 20 settings by +3.7 points on average for the 3D ViT Primus encoder and +1.1 for the slice-based 2D ViT Curia encoder. Increasing input resolution further is strongly projector-dependent: the PerceiverResampler, paired with higher-resolution Curia features, yields the strongest configuration in the resolution study on both datasets. Our best configurations achieve state-of-the-art clinical macro F1 on the test sets, reaching 49.5 on CT-RATE and 49.0 on Merlin. Code and models will be published upon publication.
Chinese Translation
视觉-语言模型为自动化放射学报告生成提供了一条有前景的路径，但将其应用于完整的3D CT体积面临着巨大的计算挑战。现代基础视觉编码器（VEs）每次扫描可以生成数万个视觉标记，使得传递给大型语言模型（LLM）的视觉序列成为主要的计算瓶颈。视觉到语言的投影器可以压缩该序列以减少计算，但可能会丢弃临床相关细节；相反，有效的压缩可以在保持下游标记数量固定的同时容纳更高分辨率的输入。因此，如何在输入视场、空间分辨率和视觉到语言投影之间分配这一视觉标记预算仍然是一个开放的设计问题。我们系统地评估了四种异构的视觉编码器（基于CNN和ViT），五种在高达64倍压缩下的标记减少投影器，以及一个不减少标记的多层感知器（MLP）投影器基线，并在两个大规模CT报告数据集（CT-RATE和Merlin）上评估了五种经过指令调优的LLM（1.7B-4B）。在匹配的LLM标记预算下，基于解剖学的感兴趣区域裁剪是一种最一致的策略，在20种设置中平均提高了3D ViT Primus编码器的临床宏F1得分3.7分，以及基于切片的2D ViT Curia编码器的1.1分。进一步提高输入分辨率的效果强烈依赖于投影器：PerceiverResampler与更高分辨率的Curia特征配对，在两个数据集的分辨率研究中产生了最强的配置。我们的最佳配置在测试集上达到了最先进的临床宏F1，CT-RATE达到49.5，Merlin达到49.0。代码和模型将在发表时发布。
cs.CV / 131 / 2608.08720
High-Quality Exposure Correction with Diffusion-Based Image Generation Priors
基于扩散图像生成先验的高质量曝光校正
Abstract
Although most existing exposure correction methods achieve high fidelity, they often place excessive focus on overall pixel-wise accuracy, making it challenging to effectively model extreme exposure regions, which results in suboptimal perceptual quality. Recently, diffusion models have received significant attention due to their remarkable performance in the realm of image generation. However, their successful application to exposure correction remains a challenging and open question. The key challenge lies in generating accurate image structures and maintaining high image fidelity during stochastic diffusion processes. In this paper, we propose DPEC (Diffusion Prior-based Exposure Correction), a novel framework for image exposure correction that utilizes diffusion-based image generation priors encapsulated in pre-trained large-scale diffusion models. Specifically, we first propose an efficient fine-tuning strategy to derive an exposure corrector from pre-trained models, enabling the generation of enhanced images in a single-step denoising process. Moreover, we seamlessly combine the strengths of diffusion models and regression models, and design a joint cross-attention module to integrate multi-scale diffusion prior features, thereby effectively preserving high-frequency details and minimizing random artifacts. The diffusion model focuses on dealing with low-frequency content rather than all the intricate texture details. The experimental results demonstrate that the proposed DPEC method consistently outperforms existing state-of-the-art methods on multiple exposure correction datasets, whether in terms of fidelity, perceptual quality, or visual effects.
Chinese Translation
尽管大多数现有的曝光校正方法能够实现高保真度，但它们往往过于关注整体像素级准确性，这使得有效建模极端曝光区域变得具有挑战性，从而导致感知质量不佳。最近，扩散模型因其在图像生成领域的卓越表现而受到广泛关注。然而，它们在曝光校正中的成功应用仍然是一个具有挑战性和开放性的问题。关键挑战在于在随机扩散过程中生成准确的图像结构并保持高图像保真度。本文提出了DPEC（基于扩散先验的曝光校正），这是一个利用封装在预训练大规模扩散模型中的扩散图像生成先验的图像曝光校正新框架。具体而言，我们首先提出了一种高效的微调策略，从预训练模型中推导出曝光校正器，使得在单步去噪过程中生成增强图像成为可能。此外，我们无缝结合了扩散模型和回归模型的优势，设计了一个联合交叉注意力模块，以整合多尺度扩散先验特征，从而有效保留高频细节并最小化随机伪影。扩散模型专注于处理低频内容，而不是所有复杂的纹理细节。实验结果表明，所提出的DPEC方法在多个曝光校正数据集上，无论是在保真度、感知质量还是视觉效果方面，均持续优于现有的最先进方法。
cs.CV / 132 / 2608.08727
TomaMMU: A Comprehensive Multimodal Understanding Benchmark for Tomato Leaf Diseases
TomaMMU：一个全面的番茄叶病多模态理解基准
Abstract
To address this gap, we introduce TomaMMU, a large-scale Tomato leaf disease MultiModal Understanding dataset, alongside TomaBench, a benchmark for evaluating VLMs on tomato disease understanding. TomaMMU comprises 28,808 high-quality images spanning 15 categories and 213,119 human-annotated visual question-answer pairs, generated through a three-stage pipeline comprising Data Collection, Human Annotation, and Question-Answer Generation. Building on this foundation, TomaBench organizes seven agricultural tasks into a hierarchical three-level taxonomy spanning Basic Perception, Pathology Understanding, and Expert Diagnosis, which together enable systematic evaluation from low-level visual recognition to high-level diagnostic reasoning. The tasks assess visual symptom recognition, taxonomic relationships, and diagnostic reasoning, offering a comprehensive view of how well models grasp plant pathology. Our results pronounced gaps in fine-grained recognition and factually grounded reasoning with 14 state-of-the-art VLMs, consistently underperforming on both challenging MCQs and open-ended questions. These results suggest that current VLMs struggle to translate visual perception into reliable diagnostic knowledge, motivating the need for targeted domain adaptation. Simple fine-tuning on TomaMMU substantially narrows this gap, boosting accuracy on challenging MCQs to 96.09%, outperforming recent VLMs, and pointing toward promising directions for future work. All data and code is available in https://huggingface.co/datasets/enalis/TomaMMU.
Chinese Translation
为了解决这一空白，我们推出了 TomaMMU，一个大规模的番茄叶病多模态理解数据集，以及 TomaBench，一个用于评估 VLM（视觉语言模型）在番茄病理解方面的基准。TomaMMU 包含 28,808 张高质量图像，涵盖 15 个类别和 213,119 对人类标注的视觉问答对，这些数据通过一个包括数据收集、人类标注和问答生成的三阶段流程生成。在此基础上，TomaBench 将七个农业任务组织成一个分层的三层分类法，涵盖基础感知、病理理解和专家诊断，从而实现从低级视觉识别到高级诊断推理的系统评估。这些任务评估视觉症状识别、分类关系和诊断推理，提供了模型对植物病理学理解的全面视角。我们的结果显示，14 个最先进的 VLM 在细粒度识别和基于事实的推理方面存在明显差距，在具有挑战性的多项选择题和开放式问题上表现不佳。这些结果表明，当前的 VLM 在将视觉感知转化为可靠的诊断知识方面存在困难，促使我们需要进行针对性的领域适应。在 TomaMMU 上进行简单的微调显著缩小了这一差距，将在具有挑战性的多项选择题上的准确率提升至 96.09%，超越了近期的 VLM，并指向未来工作的有希望方向。所有数据和代码可在 https://huggingface.co/datasets/enalis/TomaMMU 获取。
cs.CV / 133 / 2608.08732
AnchorFold: A Focus-Then-Fold Framework via Recursive Attention Propagation for Efficient Multi-Vector Visual Document Retrieval
AnchorFold：一种通过递归注意力传播实现高效多向量视觉文档检索的聚焦-折叠框架
Abstract
Multi-vector vision-language retrievers enable fine-grained Visual Document Retrieval (VDR) through late interaction, but storing and scoring hundreds of visual patch embeddings per page incurs substantial overhead. Existing training-free methods rely on pruning or merging: pruning degrades sharply under aggressive compression, whereas merging does not explicitly prioritize important regions when forming representatives. We introduce AnchorFold, a training-free focus-then-fold framework for document-side index compression. AnchorFold applies Recursive Attention Propagation over visual self-attention graphs, performing multi-step propagation within each attention head and integrating scores across heads and layers. The focus stage selects the highest-centrality tokens as anchors. The fold stage assigns remaining tokens to their most similar anchors in the normalized retrieval space and summarizes each anchor-centered group through centrality-weighted aggregation. This preserves non-anchor contributions while concentrating capacity on structurally important tokens. Across ViDoRe v1/v2 and REAL-MM-RAG with three diverse retrieval backbones, AnchorFold consistently outperforms all evaluated training-free baselines at $\gamma \leq 0.20$. On ViDoRe v1/v2, it retains 98.3% of full-index NDCG@5 on average at $5\times$ compression, achieving near-lossless compression, and 92.4% at $20\times$ compression.
Chinese Translation
多向量视觉语言检索器通过后期交互实现细粒度的视觉文档检索（VDR），但每页存储和评分数百个视觉补丁嵌入会带来相当大的开销。现有的无训练方法依赖于剪枝或合并：剪枝在激进压缩下急剧降级，而合并在形成代表时并未明确优先考虑重要区域。我们提出了AnchorFold，这是一种无训练的聚焦-折叠框架，用于文档侧索引压缩。AnchorFold在视觉自注意力图上应用递归注意力传播，在每个注意力头内执行多步传播，并跨头和层整合评分。聚焦阶段选择最高中心性的标记作为锚点。折叠阶段将剩余标记分配给其在归一化检索空间中最相似的锚点，并通过中心性加权聚合总结每个以锚点为中心的组。这在集中容量于结构重要标记的同时保留了非锚点的贡献。在ViDoRe v1/v2和REAL-MM-RAG上，结合三种不同的检索骨干网络，AnchorFold在$ ext{γ} ext{≤} 0.20$的情况下始终优于所有评估的无训练基线。在ViDoRe v1/v2上，它在$5 imes$压缩下平均保留98.3%的完整索引NDCG@5，达到了近乎无损的压缩，在$20 imes$压缩下保留92.4%。
cs.CV / 134 / 2608.08734
IDATA: Scalable Invertible Diffusion for Unrestricted Adversarial Transfer Attack
IDATA：可扩展的可逆扩散用于无限制对抗转移攻击
Abstract
Unrestricted adversarial transfer attacks are important for evaluating the black-box robustness of deep visual models. Diffusion-based attacks have shown promising transferability and visual imperceptibility by optimizing adversarial perturbations along denoising trajectories in latent space. However, existing methods are limited by two challenges: memory-intensive multistep backpropagation and frequency-agnostic perturbation over intermediate latents. To address these issues, we propose IDATA, a memory-efficient diffusion framework for unrestricted adversarial transfer attack. IDATA consists of two key components: an Invertible Diffusion Module (IDM) and a Low-Frequency Constraint Module (LFCM). Specifically, IDM reformulates adversarial optimization over diffusion trajectories as an invertible process, enabling constant-memory backpropagation through on-demand reconstruction of intermediate states instead of storing the full denoising chain. Moreover, LFCM leverages Discrete Wavelet Transform (DWT) to decompose latent variables into low- and high-frequency components, restricting perturbations to semantically stable low-frequency subspaces, thereby improving transferability while preserving visual imperceptibility. Extensive experiments on multiple benchmarks and diverse model architectures demonstrate that IDATA consistently outperforms state-of-the-art baselines in attack success rate, memory efficiency, and visual imperceptibility. These results suggest that IDATA is a promising tool for black-box robustness evaluation of deep visual models. Code is available at https://github.com/colourful-pan/IDATA.
Chinese Translation
无限制对抗转移攻击对于评估深度视觉模型的黑箱鲁棒性至关重要。基于扩散的攻击通过在潜在空间中沿去噪轨迹优化对抗扰动，展现出良好的转移性和视觉不可察觉性。然而，现有方法受到两个挑战的限制：内存密集型的多步反向传播和对中间潜变量的频率无关扰动。为了解决这些问题，我们提出了IDATA，一个内存高效的扩散框架，用于无限制对抗转移攻击。IDATA由两个关键组件组成：可逆扩散模块（Invertible Diffusion Module, IDM）和低频约束模块（Low-Frequency Constraint Module, LFCM）。具体而言，IDM将对抗优化在扩散轨迹上的过程重新表述为一个可逆过程，使得通过按需重建中间状态而不是存储完整的去噪链来实现常量内存反向传播。此外，LFCM利用离散小波变换（Discrete Wavelet Transform, DWT）将潜在变量分解为低频和高频成分，将扰动限制在语义稳定的低频子空间，从而在保持视觉不可察觉性的同时提高转移性。在多个基准和多样化模型架构上的广泛实验表明，IDATA在攻击成功率、内存效率和视觉不可察觉性方面始终优于最先进的基线。这些结果表明，IDATA是评估深度视觉模型黑箱鲁棒性的有前景的工具。代码可在 https://github.com/colourful-pan/IDATA 获取。
cs.CV / 135 / 2608.08753
Parcel2Progression: An Anatomy-aware Longitudinal Framework for Alzheimer's Disease Diagnosis
Parcel2Progression：一种基于解剖学的阿尔茨海默病纵向诊断框架
Abstract
Alzheimer's disease (AD) progression is a longitudinal process with subtle pathological cues in the early stages. Yet, computational constraints have limited most neuroimaging models to either compromise spatial information or limit the number of longitudinal scans. We aim to overcome this bottleneck and fully leverage high-resolution, variable-length T1w structural MRI (4D sMRI) scan sequences. We introduce Parcel2Progression (P2P), a Longitudinal Transformer Framework which tackles this challenge using an Atlas-guided Parcel Encoder that tokenizes 3D scans into a set of richer anatomically grounded representations. A Longitudinal Transformer then integrates irregular, arbitrary-length longitudinal visits with patient age. This synergy delivers two key advantages: (1) parcel-specific interpretability, and (2) computational tractability for long-term analysis, which scales linearly with the number of scans compared to a naive quadratic 4D ViT cost. P2P outperforms prior works and baselines in both MCI (Mild Cognitive Impairment) to AD conversion prediction and AD vs. CN (Cognitively Normal) classification tasks across ADNI, AIBL, and MIRIAD datasets. Leveraging longitudinal scans boosts performance over single-scan baselines by up to 5% and 7% in balanced accuracy for AD classification and MCI conversion prediction tasks, respectively. Interpretability analysis using parcel saliencies and attention rollouts reveals clinically consistent atrophy patterns in AD and MCI subjects. We also demonstrate the frameworks' reliability in anomaly detection using a synthetic dataset, and test the model's generalizability for other neurodegenerative diseases like Frontotemporal Dementia.
Chinese Translation
阿尔茨海默病（AD）的进展是一个纵向过程，在早期阶段具有微妙的病理线索。然而，计算约束使得大多数神经影像学模型要么妥协空间信息，要么限制纵向扫描的数量。我们旨在克服这一瓶颈，充分利用高分辨率、可变长度的T1w结构性MRI（4D sMRI）扫描序列。我们提出了Parcel2Progression（P2P），一个纵向变换器框架，通过一个基于图谱的分区编码器将3D扫描转化为一组更丰富的解剖学基础表示，从而应对这一挑战。纵向变换器随后将不规则、任意长度的纵向访问与患者年龄整合在一起。这种协同作用带来了两个关键优势：（1）特定分区的可解释性，以及（2）长时间分析的计算可行性，与天真的二次4D ViT成本相比，线性扩展扫描数量。P2P在阿尔茨海默病神经影像学数据库（ADNI）、AIBL和MIRIAD数据集中的MCI（轻度认知障碍）到AD转化预测和AD与CN（认知正常）分类任务中超越了先前的研究和基线。利用纵向扫描在AD分类和MCI转化预测任务中，性能比单次扫描基线提高了高达5%和7%的平衡准确率。使用分区显著性和注意力回滚的可解释性分析揭示了AD和MCI受试者中临床一致的萎缩模式。我们还展示了该框架在使用合成数据集进行异常检测时的可靠性，并测试了模型在其他神经退行性疾病（如额颞叶痴呆）中的泛化能力。
cs.CV / 136 / 2608.08771
UPolarSQ: Polar Representation Learning for Optic Disc and Peripapillary Atrophy Segmentation and Quantification in Fundus Photographs
UPolarSQ：用于视盘和视盘周围萎缩分割与定量的极坐标表示学习在眼底照片中的应用
Abstract
Myopia-induced posterior-pole remodeling is frequently accompanied by Optic Disc (OD) deformation and Peripapillary Atrophy (PPA), both of which provide clinically relevant structural biomarkers. In Cartesian fundus images, however, PPA often appears as an irregular and partially visible crescent adjacent to the OD, leading to fragmented segmentation and post-processing-dependent quantification. We propose UPolarSQ, a unified polar-domain framework for OD/PPA segmentation and biomarker quantification in myopic fundus images. UPolarSQ first maps an OD-centered region of interest into polar coordinates, where OD and PPA boundaries can be represented as radial profiles. It then employs UPolarSeg, a U-Net-based segmentation network enhanced with a Radial-Angular-Decoupled Module and boundary-aware auxiliary supervision to model anisotropic polar features and radial boundary transitions. Clinical biomarkers, including disc shape and PPA-width-related measurements, are deterministically extracted from the predicted polar masks, aligning segmentation and quantification within a shared geometric representation. Experiments on internal and external cohorts demonstrate that UPolarSQ improves OD/PPA segmentation and supports reliable polar-native biomarker estimation for myopic analysis.
Chinese Translation
近视引起的后极重塑常伴随视盘（Optic Disc, OD）变形和视盘周围萎缩（Peripapillary Atrophy, PPA），这两者均提供了临床相关的结构生物标志物。然而，在笛卡尔眼底图像中，PPA通常表现为与OD相邻的不规则且部分可见的弯月形，导致分割碎片化和依赖后处理的定量。我们提出了UPolarSQ，一个统一的极域框架，用于在近视眼底图像中进行OD/PPA分割和生物标志物定量。UPolarSQ首先将以OD为中心的感兴趣区域映射到极坐标中，在该坐标系中，OD和PPA边界可以表示为径向轮廓。然后，它采用UPolarSeg，一个基于U-Net的分割网络，增强了径向-角度-解耦模块和边界感知辅助监督，以建模各向异性极坐标特征和径向边界过渡。临床生物标志物，包括视盘形状和与PPA宽度相关的测量值，从预测的极坐标掩膜中确定性提取，实现了分割和定量在共享几何表示中的对齐。在内部和外部队列上的实验表明，UPolarSQ改善了OD/PPA分割，并支持可靠的极坐标本地生物标志物估计，以便进行近视分析。
cs.CV / 137 / 2608.08805
LASA: Language-and-Source-Anchored Alignment for Domain Generalized Semantic Segmentation
LASA：基于语言和源的对齐用于领域泛化语义分割
Abstract
Domain Generalization Semantic Segmentation (DGSS) focuses on generalizing knowledge from labeled source domains to unseen target domains where data is unavailable during the training phase. While conventional methods utilize style randomization or feature normalization to mitigate domain shifts, they often impair feature integrity. Specifically, style randomization distorts the underlying feature manifold due to its coarse-grained nature, while feature normalization suppresses discriminative, domain-sensitive semantic details owing to its rigid design. To address these limitations, we propose the Language-and-Source-Anchored Alignment (LASA) framework, which comprises three synergistic components: Text-and-Source-Guided Style Transfer (TSGST), Domain-Aware Query Adapter (DAQA), and Domain-Aware Decoder Optimizer (DADO). Concretely, the TSGST module addresses manifold distortion by utilizing source features as structural anchors and vision-language model (VLM) priors as fine-grained guidance. To restore suppressed discriminative and domain-sensitive details, the DAQA module recalibrates object queries via categorical guidance and domain-aware signatures, while the DADO module aligns the resulting query distributions with a shared classifier to ensure consistent categorical responses across domains. Extensive experiments on challenging benchmarks demonstrate that our method significantly outperforms state-of-the-art approaches.
Chinese Translation
领域泛化语义分割（DGSS）关注于将知识从标记的源领域泛化到在训练阶段数据不可用的未见目标领域。尽管传统方法利用风格随机化或特征归一化来减轻领域偏移，但它们往往会损害特征的完整性。具体而言，风格随机化由于其粗粒度特性而扭曲了潜在特征流形，而特征归一化则因其刚性设计抑制了区分性和领域敏感的语义细节。为了解决这些局限性，我们提出了语言和源锚定对齐（LASA）框架，该框架包括三个协同组件：文本和源引导的风格迁移（TSGST）、领域感知查询适配器（DAQA）和领域感知解码器优化器（DADO）。具体而言，TSGST模块通过利用源特征作为结构锚点和视觉-语言模型（VLM）先验作为细粒度指导来解决流形扭曲问题。为了恢复被抑制的区分性和领域敏感细节，DAQA模块通过类别指导和领域感知签名重新校准对象查询，而DADO模块则将生成的查询分布与共享分类器对齐，以确保跨领域的一致类别响应。在具有挑战性的基准测试上的广泛实验表明，我们的方法显著优于最先进的方法。
cs.CV / 138 / 2608.08808
AdapterMoE: A Two-Stage Hard-Routing Mixture-of-Experts Architecture for Multi-Crop Disease Recognition with Calibrated Rejection and Incremental Learning
AdapterMoE：一种用于多作物病害识别的两阶段硬路由专家混合架构，具有校准拒绝和增量学习
Abstract
Timely crop-disease identification is critical to food security. Multi-crop recognition suits Mixture-of-Experts (MoE), but conventional soft-routing MoE learns crop assignment freely end-to-end, letting a few experts dominate (expert collapse) with no semantic correspondence to crops, and facing high retraining costs, unstable rejection of non-target inputs, and a saturated accuracy ceiling. We shift the objective from accuracy toward a trade-off among deployment cost, scaling flexibility, and rejection stability, using deterministic hard routing. We propose AdapterMoE: a RouterHead classifies the crop and rejects non-target crops via a Maximum Softmax Probability threshold, with a dual-gate Energy+KNN out-of-distribution module catching distribution-shifted inputs; five per-crop Adapters atop a frozen EfficientNet-B0 backbone discriminate diseases, each calibrated via Temperature Scaling. Because experts are hard-isolated at the data level, the design avoids expert collapse and exposes an add_crop interface for local, per-crop updates instead of full retraining. On PlantVillage (5 crops, 26 classes), across a fair five-system comparison, AdapterMoE attains accuracy statistically indistinguishable from the best baselines (Macro-F1 within a 0.24-point band) while cutting training cost to about 9% of full-network baselines, expanding to a new crop in
Chinese Translation
及时识别作物病害对食品安全至关重要。多作物识别适合使用专家混合模型（Mixture-of-Experts, MoE），但传统的软路由 MoE 自由地进行端到端学习作物分配，导致少数专家主导（专家崩溃），且与作物之间没有语义对应，面临高重训练成本、不稳定的非目标输入拒绝和饱和的准确度上限。我们将目标从准确度转向在部署成本、扩展灵活性和拒绝稳定性之间的权衡，采用确定性硬路由。我们提出了 AdapterMoE：一个 RouterHead 通过最大软最大概率阈值对作物进行分类并拒绝非目标作物，同时一个双门能量+KNN（K-Nearest Neighbors）分布外模块捕捉分布转移的输入；在一个冻结的 EfficientNet-B0 主干上，五个每作物的适配器区分病害，每个适配器通过温度缩放进行校准。由于专家在数据层面上被硬隔离，该设计避免了专家崩溃，并为局部每作物更新提供了 add_crop 接口，而不是完全重训练。在 PlantVillage 数据集（5 种作物，26 个类别）上，在公平的五系统比较中，AdapterMoE 的准确度在统计上与最佳基线无显著差异（宏 F1 在 0.24 点带内），同时将训练成本降低到全网络基线的约 9%，并扩展到新作物。
cs.CV / 139 / 2608.08814
360CityArena: A Realistic Virtual Urban Navigation Benchmark for Embodied Agents
360CityArena：一个用于具身智能体的真实虚拟城市导航基准
Abstract
We present 360CityArena, a benchmark for evaluating the urban exploration capabilities of embodied agents within a photorealistic environment constructed from 360-degree videos. Existing outdoor benchmarks either lack sufficient photorealism or complexity, resulting in a considerable gap from real-world urban environments. 360CityArena is built on a realistic reconstruction of the Akihabara district in Tokyo, Japan, using 602 360-degree video segments covering 85 streets, and consists of 175 meticulously human-crafted tasks. It encompasses three task categories: Environment Understanding, Path Reasoning, and Spatial Reasoning, covering fundamental abilities required for urban exploration, such as localization, landmark search, path planning, and relational spatial reasoning, thereby enabling comprehensive evaluation in realistic urban scenes. Our evaluation using state-of-the-art LMM-based agents shows that even the strongest model, Gemini 2.5 Flash, performs far below human level (human: 77.3% vs. Gemini 2.5 Flash: 17.1%), revealing substantial challenges that remain in city-scale embodied navigation and reasoning. 360CityArena provides a necessary and challenging testbed for photorealistic urban-district navigation and spatial reasoning.
Chinese Translation
我们提出了360CityArena，这是一个用于评估具身智能体在由360度视频构建的照片真实环境中城市探索能力的基准。现有的户外基准要么缺乏足够的照片真实感，要么复杂性不足，导致与真实城市环境之间存在显著差距。360CityArena基于对日本东京秋叶原地区的真实重建，使用了602个覆盖85条街道的360度视频片段，并包含175个精心设计的任务。它涵盖了三个任务类别：环境理解、路径推理和空间推理，涉及城市探索所需的基本能力，如定位、地标搜索、路径规划和关系空间推理，从而能够在真实的城市场景中进行全面评估。我们使用最先进的基于LMM的智能体进行的评估显示，即使是最强的模型Gemini 2.5 Flash，其表现也远低于人类水平（人类：77.3% vs. Gemini 2.5 Flash：17.1%），揭示了城市规模的具身导航和推理中仍然存在的重大挑战。360CityArena为照片真实的城市区域导航和空间推理提供了一个必要且具有挑战性的测试平台。
cs.CV / 140 / 2608.08819
MRI super-resolution in ten sampling steps using a diffusion bridge model
基于扩散桥模型的MRI超分辨率在十个采样步骤内实现
Abstract
Objective. MRI provides excellent soft-tissue contrast, but long acquisition times can cause patient discomfort and lead to motion artifacts, forcing a trade-off between spatial resolution and scan time. Diffusion-based super-resolution (SR) reconstructs high-resolution (HR) images from low-resolution (LR) inputs, but typically needs many sampling steps and initializes from a Gaussian prior ill-suited to image restoration. We developed an efficient diffusion framework that reconstructs HR MRI directly from LR data. Approach. We propose super-resolution diffusion bridge model (SR-DBM), a super-resolution diffusion bridge model that casts SR as a stochastic transport between the LR and HR image distributions. Through a Doob's h-transform of a mean-reverting stochastic differential equation, SR-DBM pins the process to the paired HR and LR images at its endpoints, initializing reconstruction from the measured anatomy rather than from Gaussian noise. The HR image is recovered by a deterministic reverse trajectory in which a network predicts the clean image at each of only ten sampling steps. We evaluated SR-DBM on ultra-high-field 7T brain T1 MP2RAGE maps and pelvic T2-weighted prostate images against nine comparison methods using PSNR, SSIM, GMSD, and LPIPS. Main results. SR-DBM attained the highest PSNR and SSIM and the lowest GMSD on both datasets (brain: 27.66+-1.52 dB, 0.96+-0.02, 7.96+-1.86$; prostate: 27.87+-2.29 dB, 0.80+-0.05, 8.38+- 1.44), with statistically significant gains over every comparison method (two-sided Wilcoxon signed-rank test with Holm correction, p<0.05). The strongest baseline, SR-EMamba, ranked second. Qualitatively, SR-DBM produced the smallest residual errors and best preserved fine structures and lesions.
Chinese Translation
目的：MRI提供了优异的软组织对比度，但长时间的采集可能导致患者不适并产生运动伪影，从而在空间分辨率和扫描时间之间形成权衡。基于扩散的超分辨率（SR）从低分辨率（LR）输入重建高分辨率（HR）图像，但通常需要多个采样步骤，并且初始化时使用的高斯先验不适合图像恢复。我们开发了一种高效的扩散框架，能够直接从LR数据重建HR MRI。方法：我们提出了超分辨率扩散桥模型（SR-DBM），将超分辨率视为LR和HR图像分布之间的随机传输。通过对均值回归随机微分方程的Doob h变换，SR-DBM将过程固定在其端点的配对HR和LR图像上，从测量的解剖结构初始化重建，而不是从高斯噪声开始。HR图像通过确定性的反向轨迹恢复，其中网络在仅十个采样步骤中预测干净图像。我们在超高场7T脑部T1 MP2RAGE图和盆腔T2加权前列腺图像上评估了SR-DBM，并使用PSNR、SSIM、GMSD和LPIPS与九种对比方法进行了比较。主要结果：SR-DBM在两个数据集上均获得了最高的PSNR和SSIM以及最低的GMSD（脑部：27.66±1.52 dB，0.96±0.02，7.96±1.86；前列腺：27.87±2.29 dB，0.80±0.05，8.38±1.44），并且在每种对比方法上均有统计学显著提升（双侧Wilcoxon符号秩检验，Holm校正，p<0.05）。最强的基线方法SR-EMamba排名第二。从定性上看，SR-DBM产生了最小的残差误差，并且最佳地保留了细微结构和病变。
cs.CV / 141 / 2608.08820
LogiShot: Logically Coherent Cross-Shot Video Generation
LogiShot：逻辑一致的跨镜头视频生成
Abstract
Generating cross-shot videos that are logically connected is essential for content creation. Currently, most cross-shot video-generation workflows, such as short-drama production, still rely on isolated textual scripts or explicit reference images to specify the generated content. Consequently, when user instructions are underspecified or ambiguous, a generated clip may appear visually plausible on its own but fail to align with the overall narrative, leading to disjointed content. We argue that achieving cross-shot logical coherence in video generation requires establishing logical connections across shots and maintaining visual consistency. To this end, we propose LogiShot, which incorporates information through two complementary paths: 1) LogiShot jointly encodes the context video and other conditioning signals, yielding dense multimodal cues that provide visual-semantic evidence for cross-shot generation; 2) the model maintains a visual memory of the context video throughout generation to preserve visual consistency across shots. Additionally, we construct a dataset with 110K samples and a dedicated benchmark for evaluating cross-shot logical coherence. Experiments demonstrate that LogiShot consistently outperforms existing baselines in terms of logical coherence across multiple shots. Model and data will be made publicly available.
Chinese Translation
生成逻辑连贯的跨镜头视频对于内容创作至关重要。目前，大多数跨镜头视频生成工作流程，如短剧制作，仍然依赖于孤立的文本脚本或明确的参考图像来指定生成的内容。因此，当用户指令不够明确或含糊时，生成的片段可能在视觉上看似合理，但未能与整体叙事保持一致，导致内容割裂。我们认为，在视频生成中实现跨镜头的逻辑一致性需要在镜头之间建立逻辑联系并保持视觉一致性。为此，我们提出了LogiShot，它通过两条互补路径整合信息：1）LogiShot联合编码上下文视频和其他条件信号，生成密集的多模态线索，为跨镜头生成提供视觉语义证据；2）该模型在生成过程中保持上下文视频的视觉记忆，以保持镜头之间的视觉一致性。此外，我们构建了一个包含11万个样本的数据集，并设立了专门的基准来评估跨镜头的逻辑一致性。实验表明，LogiShot在多个镜头的逻辑一致性方面始终优于现有基线。模型和数据将公开发布。
cs.CV / 142 / 2608.08832
Visual Token Codec: Unleashing Spatial Redundancy for ViT Feature Coding
视觉标记编解码器：释放ViT特征编码的空间冗余
Abstract
Distributed deployment of large vision foundation models often partitions a ViT backbone and exchanges intermediate token features between computing nodes, making efficient feature compression critical under bandwidth and computation constraints. Existing ViT feature codecs typically flatten heterogeneous global and patch tokens into an L x C pseudo image, causing entropy models to mainly capture sequence-axis dependencies while overlooking the native two-dimensional patch-grid structure. In this paper, we show that ViT patch tokens retain strong local spatial correlations on the original grid. To exploit this structural prior, we propose the Visual Token Codec (VTC), a dual-path learned codec that separates global and patch tokens into dedicated coding paths. Global tokens are compressed with a lightweight factorized prior, whereas patch tokens are encoded on the patch-token grid using a spatial-channel context entropy model. To support intermediate-layer compression and practical rate adaptation, VTC further incorporates feature-matching supervision after subsequent ViT blocks and variable-rate modules within a single codec. Experiments on DINOv2 and SAM3 show that VTC consistently outperforms representative ViT feature coding baselines on classification, segmentation, and detection tasks. At 90% of uncompressed-feature performance, VTC reduces bitrate by 15.7x-37.4x across these tasks. We further provide intermediate-layer rate-utility analyses for practical transmission- and storage-oriented deployment scenarios.
Chinese Translation
大规模视觉基础模型的分布式部署通常会将ViT骨干网络进行分割，并在计算节点之间交换中间标记特征，这使得在带宽和计算限制下高效的特征压缩变得至关重要。现有的ViT特征编解码器通常将异构的全局和补丁标记展平为一个L x C伪图像，导致熵模型主要捕捉序列轴的依赖关系，而忽视了原生的二维补丁网格结构。本文展示了ViT补丁标记在原始网格上保留了强烈的局部空间相关性。为了利用这一结构先验，我们提出了视觉标记编解码器（Visual Token Codec, VTC），这是一种双路径学习型编解码器，将全局和补丁标记分离到专用的编码路径中。全局标记使用轻量化的因式分解先验进行压缩，而补丁标记则在补丁标记网格上使用空间-通道上下文熵模型进行编码。为了支持中间层压缩和实际的速率适应，VTC进一步在后续的ViT模块和单个编解码器内结合了特征匹配监督。DINOv2和SAM3上的实验表明，VTC在分类、分割和检测任务上始终优于代表性的ViT特征编码基线。在90%的未压缩特征性能下，VTC在这些任务中将比特率降低了15.7倍至37.4倍。我们进一步提供了针对实际传输和存储导向部署场景的中间层速率-效用分析。
cs.CV / 143 / 2608.08840
SLAP: Selective Local Vision-Language Alignment for Fish Re-Identification via Partial Optimal Transport
SLAP：通过部分最优传输实现鱼类再识别的选择性局部视觉-语言对齐
Abstract
Individual fish re-identification (ReID) is a fine-grained recognition problem in which identity-discriminative cues are often localized to specific body regions rather than distributed uniformly across the animal. Nevertheless, recent CLIP-based ReID methods rely predominantly on global image-text alignment, allowing background and weakly discriminative regions to contribute to cross-modal supervision. We propose a selective local vision-language alignment framework that establishes localized correspondences between visual patch embeddings and multiple identity-aware prompt embeddings through Partial Optimal Transport (POT). Rather than enforcing exhaustive correspondence, POT enables selective matching between visual patches and prompt embeddings, allowing the model to emphasize the strongest cross-modal correspondences while avoiding forced alignment of weakly matching regions, thereby yielding more discriminative visual representations for retrieval. The framework is trained end-to-end, while only the adapted visual encoder is retained during inference. Experiments on the longitudinal Symphodus melops dataset demonstrate consistent improvements over recent CLIP-based ReID methods under both closed-set and open-set evaluation protocols. Additional evaluations on other datasets further demonstrate the generalization capability of the proposed method across diverse marine ReID benchmarks.
Chinese Translation
个体鱼类再识别（ReID）是一项细粒度识别问题，其中身份区分线索通常局限于特定的身体区域，而不是均匀分布在整个动物身上。然而，近期基于CLIP的ReID方法主要依赖于全局图像-文本对齐，使得背景和弱区分区域对跨模态监督产生影响。我们提出了一种选择性局部视觉-语言对齐框架，通过部分最优传输（POT）建立视觉补丁嵌入与多个身份感知提示嵌入之间的局部对应关系。POT并不强制执行全面的对应关系，而是允许视觉补丁与提示嵌入之间进行选择性匹配，使模型能够强调最强的跨模态对应关系，同时避免对弱匹配区域的强制对齐，从而为检索提供更具区分性的视觉表示。该框架采用端到端训练，在推理过程中仅保留适应后的视觉编码器。在纵向的Symphodus melops数据集上的实验表明，在闭集和开集评估协议下，相较于近期的基于CLIP的ReID方法，取得了一致的改进。对其他数据集的额外评估进一步展示了所提方法在多样化海洋ReID基准上的泛化能力。
cs.CV / 144 / 2608.08844
Toward Mask Annotation-Free Surgical Instrument Segmentation from Endoscopic Images Using Text-Prompted Segment Anything Model 3 (SAM3)
基于文本提示的Segment Anything Model 3 (SAM3)实现无掩码标注的内窥镜图像手术器械分割研究
Abstract
Surgical instrument segmentation is a fundamental task for computer-assisted interventions, yet most existing methods rely on pixel-level annotations or manual spatial prompts, which limit scalability and automation. The recently introduced Segment Anything Model 3 (SAM3) offers a pathway to annotation-free, automatic segmentation via text-based prompting; however, the instrument name as a text prompt could not be directly used due to a large domain gap. To overcome these limitations, we propose a two-stage framework that achieves instance-level segmentation without requiring ground truth masks or manual interaction. In the first stage, we leverage a natural-language-aligned generic prompt - "tool" - to produce binary masks using SAM3's zero-shot capability. In the second stage, these masks are extended to instance-level by integrating a vision-language model (Qwen) that is fine-tuned on SAM3-generated masked regions for instrument classification. We evaluate our approach on the EndoVis 2017 and 2018 datasets. Results show that, while our two-stage approach does not reach the performance of current fully supervised methods, it significantly outperforms the direct use of SAM3 for instance-level instrument segmentation with text prompts. Overall, our findings highlight both the limitations and potential of SAM3, suggesting a promising direction toward annotation-free surgical instrument segmentation.
Chinese Translation
手术器械分割是计算机辅助干预中的基础任务，但现有大多数方法依赖于像素级标注或手动空间提示，限制了其可扩展性和自动化水平。近期提出的Segment Anything Model 3 (SAM3)通过基于文本的提示，提供了一种无标注、自动分割的途径；然而，由于领域差异较大，直接使用器械名称作为文本提示效果不佳。为克服这些限制，我们提出了一个两阶段框架，实现实例级分割且无需真实掩码或人工交互。第一阶段，利用与自然语言对齐的通用提示词“tool”，借助SAM3的零样本能力生成二值掩码。第二阶段，通过融合在SAM3生成的掩码区域上微调的视觉语言模型（Qwen）进行器械分类，将掩码扩展至实例级。我们在EndoVis 2017和2018数据集上评估了该方法。结果表明，尽管该两阶段方法未达到当前全监督方法的性能，但显著优于直接使用SAM3文本提示进行实例级器械分割。总体而言，我们的研究揭示了SAM3的局限性与潜力，为无标注手术器械分割指明了有前景的方向。
cs.CV / 145 / 2608.08867
Zero-Shot Traffic Accident Detection via a Coarse-to-Fine VLM-Tracking Pipeline
通过粗到细的 VLM 跟踪管道进行零样本交通事故检测
Abstract
Traffic surveillance cameras capture accidents continuously, yet converting raw CCTV footage into structured event records that pinpoint when, where, and what type of collision occurred remains unsolved at scale. The ACCIDENT @ CVPR benchmark evaluates exactly this joint prediction under a strict constraint: no labeled real-world training data is available. We introduce a training-free, two-pass coarse-to-fine pipeline that pairs a frozen Qwen3-VL-32B-Instruct vision-language model with YOLO11x object detection and BoT-SORT tracking. A first pass sparsely samples the full clip to anchor the collision moment in time; a second pass re-examines a tight window around that estimate using frames annotated with stable vehicle identities and normalized bounding-box coordinates, which gives the model both a visual overlay and an explicit numeric description of the same scene. On the official 2,027-clip real-CCTV test set, our system achieves a three-way harmonic mean score of 0.504, surpassing all organizer-published baselines including the best multi-model ensemble (0.412) by a 22% relative margin.
Chinese Translation
交通监控摄像头持续捕捉事故，但将原始 CCTV 录像转换为结构化事件记录，以准确指出何时、何地以及发生了何种类型的碰撞，仍然在大规模上未得到解决。ACCIDENT @ CVPR 基准正是评估这种联合预测，且在严格的约束下进行：没有可用的标记真实世界训练数据。我们提出了一种无训练的、两步粗到细的管道，将冻结的 Qwen3-VL-32B-Instruct 视觉语言模型与 YOLO11x 目标检测和 BoT-SORT 跟踪相结合。第一步稀疏地采样完整视频片段，以确定碰撞时刻；第二步在该估计周围重新检查一个紧密窗口，使用带有稳定车辆身份和标准化边界框坐标的帧，这为模型提供了视觉叠加和同一场景的明确数值描述。在官方的 2,027 个真实 CCTV 测试集上，我们的系统实现了 0.504 的三重调和均值分数，超越了所有组织者发布的基线，包括最佳多模型集成（0.412），相对提高了 22%。
cs.CV / 146 / 2608.08873
Sparse Attention to Emotion: Efficient Facial Emotion Recognition via Token Reduction
稀疏关注情感：通过令牌减少实现高效的面部情感识别
Abstract
Facial Emotion Recognition (FER) is an important task that has significant implications across various fields such as biometrics, health, and human-computer interaction. Current Vision Transformer-based approaches display quadratic complexity $\mathcal{O}(N^2)$, with N being the input sequence length, making them cumbersome to deploy at the edge. In this paper, we hypothesize that the FER task does not necessarily require all facial information to correctly interpret emotional states, as specific regions such as the eyes, the mouth, and parts of the cheeks carry discriminative information that can be sufficient to recognize emotions. Based on this, we propose Sparse Attention to Emotion (SAE), a model that discards image tokens that have no added value to the emotional context, while preserving good accuracy and achieving a significant gain in computational cost. Surprisingly, even after suppressing 90\% of the image tokens, our model achieves competitive accuracy to state of the art methods at much lower cost, providing a lightweight Facial Emotion Recognition approach. Experimental results demonstrate that SAE achieves new state of the art results on the RAF-DB dataset while reducing the computational complexity by up to 90\%.
Chinese Translation
面部情感识别（FER）是一项重要任务，在生物识别、健康和人机交互等多个领域具有重要意义。目前基于视觉变换器（Vision Transformer）的方法显示出二次复杂度 $ ext{O}(N^2)$，其中 N 为输入序列长度，这使得它们在边缘设备上的部署变得繁琐。本文假设 FER 任务并不一定需要所有面部信息来正确解读情感状态，因为特定区域如眼睛、嘴巴和部分面颊携带的区分信息足以识别情感。基于此，我们提出了稀疏关注情感（Sparse Attention to Emotion, SAE）模型，该模型丢弃对情感上下文没有增值的图像令牌，同时保持良好的准确性并显著降低计算成本。令人惊讶的是，即使在抑制了 90 ext{%} 的图像令牌后，我们的模型仍能以更低的成本达到与最先进方法相当的准确性，从而提供了一种轻量级的面部情感识别方法。实验结果表明，SAE 在 RAF-DB 数据集上实现了新的最先进结果，同时将计算复杂度降低了多达 90 ext{%}。
cs.CV / 147 / 2608.08874
AeroReformer2: Spoken-Query Referring Segmentation for Aerial Images
AeroReformer2：用于航空图像的口语查询指代分割
Abstract
Spoken language offers a natural, hands-free interface for specifying an arbitrary target in dense remote-sensing imagery, yet existing referring remote-sensing image segmentation benchmarks accept only written expressions. To bridge this gap, we introduce \dataset, a spoken-query benchmark derived from RISBench that adds accent- and voice-diverse speech while preserving the original image, mask, and data splits. Its hard evaluation sets combine rotor, wind, and mixed interference with three signal-to-noise levels. We also propose \model, an efficient bilateral network that combines a boundary-preserving visual path with token-preserving speech encoding, kernel linear cross-modal attention, and a resolution refinement head. The design conditions visual features at two scales without materializing a dense speech--visual affinity matrix, then restores fine boundaries using high-resolution visual features. On the clean test split, \model with Swin-Base achieves 62.09\% mean intersection over union (mIoU) and 68.22\% overall intersection over union (oIoU), outperforming the strongest audio-adapted remote-sensing baseline by 5.38 and 2.08 percentage points, respectively. It retains the best hard-set mIoU at 54.09\%. To the best of our knowledge, this is the first benchmark and model study of full-sentence spoken-query referring segmentation for remote-sensing imagery. The code will be made publicly available.
Chinese Translation
口语语言为在密集遥感图像中指定任意目标提供了一种自然的免提接口，但现有的指代遥感图像分割基准仅接受书面表达。为了解决这一问题，我们引入了 extit{dataset}，这是一个源自 RISBench 的口语查询基准，它在保留原始图像、掩膜和数据划分的同时，增加了口音和声音多样的语音。其困难评估集结合了旋翼、风和混合干扰，并具有三种信噪比水平。我们还提出了 extit{model}，这是一种高效的双边网络，它结合了保边视觉路径与保标记语音编码、核线性跨模态注意力和分辨率细化头。该设计在不生成密集语音-视觉亲和矩阵的情况下，条件性地处理两种尺度的视觉特征，然后使用高分辨率视觉特征恢复细致边界。在干净测试集上，使用 Swin-Base 的 extit{model} 达到了 62.09\% 的平均交并比 (mIoU) 和 68.22\\% 的总体交并比 (oIoU)，分别比最强的音频适应遥感基线高出 5.38 和 2.08 个百分点。它在最困难的评估集上保持了最佳的 mIoU，达到了 54.09\\%。据我们所知，这是针对遥感图像的完整句子口语查询指代分割的首个基准和模型研究。代码将公开发布。
cs.CV / 148 / 2608.08887
City Sentinel: A Unified AI-Based Smart Surveillance Framework for Real-Time Multi-Threat Detection Using Deep Learning
城市哨兵：基于统一人工智能的智能监控框架，用于实时多威胁检测的深度学习方法
Abstract
Rapid urbanization has increased the need for surveillance systems that can monitor multiple public safety risks at the same time. Traditional systems often use separate solutions for facial recognition, vehicle identification, fire detection, and behavioral analysis, resulting in fragmented infrastructure and multiple interfaces for operators to manage. This paper presents City Sentinel, a unified AI-based surveillance framework that integrates six detection capabilities into one scalable platform: facial recognition, automatic number plate recognition (ANPR), fire and smoke detection, weapon and knife detection, violence detection, and road accident detection. The system combines a Next.js operator dashboard, FastAPI backend, cloud-based PostgreSQL event storage, InsightFace and YOLOv8 vision models, and EasyOCR for plate recognition. Camera streams are processed through dedicated inference workers using RTSP. On a workstation equipped with an NVIDIA RTX 3060 GPU, the system achieves a median end-to-end latency of 743 ms and supports four concurrent RTSP streams within a two-second latency limit. It achieves a 91.2% face-match rate, 85.7% plate-reading accuracy, and
[email protected] scores of 0.846 to 0.889 across the fire, knife, and weapon detection modules. In user-acceptance testing, operators could enroll a new identity in under one minute and identify a flagged person from live footage in an average of 12 seconds. The results demonstrate that a modular, open-source, multi-model architecture can provide broad surveillance coverage, cloud-based auditability, and flexibility for adding new detection capabilities while maintaining practical real-time performance.
Chinese Translation
快速城市化增加了对能够同时监控多种公共安全风险的监控系统的需求。传统系统通常为面部识别、车辆识别、火灾检测和行为分析等使用单独的解决方案，导致基础设施碎片化以及操作员需要管理多个接口。本文提出了城市哨兵（City Sentinel），一个统一的基于人工智能的监控框架，将六种检测能力集成到一个可扩展的平台中：面部识别、自动车牌识别（ANPR）、火灾和烟雾检测、武器和刀具检测、暴力检测以及道路事故检测。该系统结合了 Next.js 操作员仪表板、FastAPI 后端、基于云的 PostgreSQL 事件存储、InsightFace 和 YOLOv8 视觉模型，以及 EasyOCR 进行车牌识别。摄像头流通过专用推理工作节点使用 RTSP 进行处理。在配备 NVIDIA RTX 3060 GPU 的工作站上，该系统实现了743毫秒的中位端到端延迟，并在两秒延迟限制内支持四个并发的 RTSP 流。它在火灾、刀具和武器检测模块中实现了91.2%的面部匹配率、85.7%的车牌读取准确率，以及0.846到0.889的
[email protected] 分数。在用户接受测试中，操作员能够在一分钟内注册一个新身份，并在平均12秒内从实时视频中识别出被标记的人。结果表明，模块化、开源的多模型架构能够提供广泛的监控覆盖、基于云的审计能力，并在保持实际实时性能的同时灵活地添加新的检测能力。
cs.CV / 149 / 2608.08904
From Recovery to Drop-off: How Action Post-training Reduces a VLM's Late-Layer Depth Decodability
从恢复到下降：行动后训练如何降低视觉语言模型的后层深度可解码性
Abstract
How much of a vision-language model's (VLM) spatial understanding remains after the action post-training process of building a vision-language-action model (VLA)? We probe depth perception, a primitive of spatiogeometric understanding, from every decoder layer of a weight-matched open-source base VLM/VLA pair: Molmo2-ER and MolmoAct2-LIBERO. First, the VLA decodes depth worse at every layer, a persistent gap we call the floor. Second, the degradation is not uniform: while the base VLM's depth decodability improves through its final layers, the VLA's collapses, an additional late-layer drop we call the cliff. We causally localize the cliff to late-layer MLP interference: ablating the late-layer MLP writes recovers the majority of the terminal decodability cliff, while matched attention ablations and the same intervention in the weight-matched base VLM produce no comparable recovery. A module-level decomposition explains this dissociation: the base VLM carries depth most accessibly in accumulated MLP writes, whereas action post-training collapses depth decodability in the late accumulated writes.
Chinese Translation
在构建视觉语言行动模型（VLA）的行动后训练过程中，视觉语言模型（VLM）的空间理解能力还剩下多少？我们从一对权重匹配的开源基础 VLM/VLA 模型：Molmo2-ER 和 MolmoAct2-LIBERO 的每个解码层探讨深度感知，这是空间几何理解的一个基本组成部分。首先，VLA 在每一层的深度解码性能都较差，这一持久的差距我们称之为“底线”。其次，降解并不均匀：尽管基础 VLM 的深度可解码性在其最后几层中有所改善，但 VLA 的深度可解码性却崩溃了，这一额外的后层下降我们称之为“悬崖”。我们将“悬崖”的原因归因于后层多层感知器（MLP）的干扰：消除后层 MLP 的写入可以恢复大部分终端解码性悬崖，而匹配注意力的消除和在权重匹配的基础 VLM 中进行相同干预则没有产生可比的恢复。模块级的分解解释了这种分离：基础 VLM 在累积的 MLP 写入中最容易携带深度信息，而行动后训练则在后期累积写入中崩溃了深度可解码性。
cs.CV / 150 / 2608.08907
ToolVision: Learning When and How to Use Visual Tools with Capability-Aligned Supervision
ToolVision：通过能力对齐监督学习何时以及如何使用视觉工具
Abstract
Thinking with images allows a multimodal model to compensate for limited perception by invoking visual tools through code. Yet the prevailing SFT-then-RL recipe creates a different supervision misalignment at each stage. SFT is expected to teach how to use tools, but trajectories from stronger teachers may succeed through perceptual capabilities that a smaller student cannot reliably reproduce or exploit, causing the student to imitate tool-call patterns without learning how to make them useful. RL is expected to teach when to use tools, but outcome-only rewards make fallible tool execution a liability and suppress tool use, whereas a blanket bonus for every correct tool-using trajectory encourages valid but ineffective operations. To address these two misalignments, we introduce ToolVision. During SFT, a multi-agent pipeline explores candidate trajectories, and a committee including student-scale models scores stepwise evidence gain to rank and prune the search branches. Only successfully executed trajectories with correct final answers are retained for SFT. Before RL, ToolVision compares the learner's performance with and without tools, then rewards successful tool use only on questions where tools provide a clear benefit. Both signals are constructed automatically from public task data without additional human annotations of tool use or necessity. ToolVision-8B improves over its base on all seven main benchmarks, surpasses Thyme-7B, CodeVision-8B, and CodeDance-7B on all three high-resolution benchmarks, and outperforms Qwen3-VL-32B-Thinking on V* and HRBench 8K. We will publicly release the datasets and source code.
Chinese Translation
通过图像思考使得多模态模型能够通过代码调用视觉工具来弥补感知的局限性。然而，现有的先进行监督学习（SFT）再进行强化学习（RL）的方案在每个阶段都产生了不同的监督失调。SFT旨在教授如何使用工具，但来自更强教师的轨迹可能通过较小学生无法可靠再现或利用的感知能力而成功，导致学生模仿工具调用模式而未能学习如何使其有效。RL则期望教授何时使用工具，但仅基于结果的奖励使得不可靠的工具执行成为负担，并抑制工具使用，而对每个正确的工具使用轨迹给予统一奖励则鼓励有效但无效的操作。为了解决这两种失调，我们引入了ToolVision。在SFT过程中，一个多智能体管道探索候选轨迹，一个包括学生规模模型的委员会对逐步证据增益进行评分，以对搜索分支进行排名和修剪。只有成功执行且最终答案正确的轨迹会被保留用于SFT。在进行RL之前，ToolVision比较学习者在有工具和无工具情况下的表现，然后仅对工具提供明显好处的问题奖励成功的工具使用。这两种信号均自动从公共任务数据中构建，无需额外的人类注释工具的使用或必要性。ToolVision-8B在所有七个主要基准测试中均优于其基础模型，在所有三个高分辨率基准测试中超越Thyme-7B、CodeVision-8B和CodeDance-7B，并在V*和HRBench 8K上超越Qwen3-VL-32B-Thinking。我们将公开发布数据集和源代码。
cs.CV / 151 / 2608.08924
From Noise to Meaning: Meaningful Secret Sharing with Tamper Detection for Facial Recognition
从噪声到意义：具有篡改检测的有意义秘密共享用于人脸识别
Abstract
Popularity of AI-based face recognition system directly demands protection of sensitive biometric data used for training. Visual secret sharing is an interesting idea, as it splits facial images into secret shares that look random and spread across many institutions. However, these shares look like noise and can easily spark suspicion and recognized as encrypted content. This makes them open to targeted collection and harvest-now-decrypt-later attacks. Additionally, visual secret sharing does not detect tampering, allowing attackers to modify shares and threaten the integrity of reconstruction. In this paper, we introduce a new method that turns distracting noise-like secret shares into visually appealing cover images with additional cryptographic tamper detection. The proposed technique works with visual secret sharing and introduces cover images to embed the shares using adaptive least significant bit steganography. Here, cover images with perceptual transparency are used to store secret shares while guaranteeing complete privacy. A two layer authentication using strong digital watermarking and cryptographic hashing is used to protect the integrity of shares. The proposed technique shows high resilience in stopping bit-flipping, cropping, and substitution attacks. Extensive experiments on multiple public face datasets show that the technique shows better FR accuracy, while eliminating share conspicuousness and guaranteeing integrity. The proposed framework sets a new standard for protecting facial data in such a way that privacy, security, and integrity are protected.
Chinese Translation
基于人工智能的人脸识别系统的普及直接要求保护用于训练的敏感生物特征数据。视觉秘密共享是一种有趣的理念，它将人脸图像分割成看似随机并分散在多个机构中的秘密份额。然而，这些份额看起来像噪声，容易引起怀疑并被识别为加密内容。这使得它们容易受到针对性的收集和“现在收集，稍后解密”的攻击。此外，视觉秘密共享不具备篡改检测功能，允许攻击者修改份额，从而威胁重建的完整性。在本文中，我们提出了一种新方法，将令人分心的噪声状秘密份额转变为视觉上吸引人的封面图像，并附加加密篡改检测。所提出的技术与视觉秘密共享相结合，引入封面图像，通过自适应最低有效位隐写术嵌入秘密份额。在这里，具有感知透明度的封面图像用于存储秘密份额，同时保证完全的隐私。采用强数字水印和加密哈希的双层认证来保护份额的完整性。所提出的技术在防止位翻转、裁剪和替换攻击方面表现出高度的抗性。在多个公共人脸数据集上的广泛实验表明，该技术在提高人脸识别准确率的同时，消除了份额的明显性并保证了完整性。所提出的框架为以保护隐私、安全和完整性为目标的人脸数据保护设定了新的标准。
cs.CV / 152 / 2608.08929
Zero-shot 2D Grounding with Novel Affordance Types
零样本二维基础定位与新型可供性类型
Abstract
2D affordance grounding aims to locate the region of an object that a human can interact with. Existing research focuses on recognizing affordance types seen during training and does not study models' ability to generalize to novel affordances, which is crucial for real-world applications. We propose the task of zero-shot 2D grounding with novel affordance types (NAT) and introduce the NAT benchmarks. We then propose AffordAnything, a training-free method that leverages segmentation cues, motivated by the strong correlation between affordance regions and object subparts. To further improve performance, we develop AffordAnything+, a trainable variant that learns to combine these cues. On the proposed AGD20K-NAT benchmark, our best model AffordAnything+ achieves a substantial improvement of 12.3% (absolute) in
[email protected] over the SOTA affordance grounding method, OOAL.
Chinese Translation
二维可供性定位旨在确定人类可以与之互动的物体区域。现有研究集中于识别训练期间见到的可供性类型，而未研究模型对新型可供性的泛化能力，这对于实际应用至关重要。我们提出了零样本二维定位与新型可供性类型（Novel Affordance Types, NAT）的任务，并引入了NAT基准。随后，我们提出了AffordAnything，这是一种无训练方法，利用分割线索，基于可供性区域与物体子部分之间的强相关性。为了进一步提高性能，我们开发了AffordAnything+，一种可训练的变体，学习结合这些线索。在所提出的AGD20K-NAT基准上，我们的最佳模型AffordAnything+在
[email protected]上相较于当前最先进的可供性定位方法OOAL实现了12.3%的显著提升（绝对值）。
cs.CV / 153 / 2608.08947
Can Webcam Gaze Constrain Mesa-Objectives in Driving Models? An Instrument Precision Analysis
网络摄像头注视是否能约束驾驶模型中的次要目标？一种仪器精度分析
Abstract
Current hazard detection systems in autonomous driving may develop mesa objectives, learned internal goals that achieve high training performance through spurious correlations rather than genuine hazard recognition. We investigate whether human gaze patterns, captured via webcam-based eye tracking (WebGazer.js), can serve as privileged information to constrain mesa-objective formation. We collected 137,663 frame-level gaze samples synchronized with hazard annotations across 388 real dashcam clips, then test this hypothesis across two calibration protocols (9-point/45-click and 11-point/440-click), two model architectures (Random Forest and causal Transformer), and five random seeds per experiment with paired t-tests. No experiment yields a statistically significant improvement from gaze (p = 0.919, 0.578, and 0.667 respectively). A geometric analysis reveals the root cause: WebGazer's reported error (~130-257 px depending on configuration) exceeds 93% of detected hazard object sizes (median 36 px), rendering object-level gaze attribution physically impossible at this instrument precision.
Chinese Translation
当前自主驾驶中的危险检测系统可能会发展出次要目标，即通过虚假的相关性而非真实的危险识别来实现高训练性能的学习内部目标。我们研究了通过基于网络摄像头的眼动追踪（WebGazer.js）捕捉的人类注视模式是否可以作为特权信息来约束次要目标的形成。我们收集了137,663个与388个真实行车记录仪片段的危险注释同步的帧级注视样本，然后在两个校准协议（9点/45次点击和11点/440次点击）、两种模型架构（随机森林和因果变换器）以及每个实验的五个随机种子下测试这一假设，采用配对t检验。没有实验显示出注视带来的统计显著性改善（p = 0.919, 0.578和0.667）。几何分析揭示了根本原因：WebGazer报告的误差（~130-257像素，具体取决于配置）超过了93%的检测到的危险物体大小（中位数36像素），使得在这种仪器精度下进行物体级注视归因在物理上变得不可能。
cs.CV / 154 / 2608.08949
EndoMD-SLAM: Endoscopic Gaussian Splatting SLAM under Optical Degradation with Memory and Static-Transient Decomposition
EndoMD-SLAM：在光学降解下具有记忆和静态-瞬态分解的内窥镜高斯点云SLAM
Abstract
Dense 3D reconstruction is critical for clinical endoscopic navigation and documentation. While Gaussian Splatting SLAM systems show promise in this domain, they fundamentally rely on strict multi-view photometric consistency. In routine procedures, this assumption is severely violated by intermittent optical degradations like moving debris and water flushing. Standard systems erroneously fuse these cameraattached artifacts into the persistent 3D geometry, causing severe tracking drift and irreversible map corruption. To address this limitation, we propose EndoMD-SLAM, a framework designed to maintain stability under optical degradation through specialized tracking and mapping mechanisms. On the tracking side, a memory-driven gating mechanism detects unreliable observations to suspend map updates and utilizes historical keyframes for drift-aware relocalization. On the mapping side, a self-supervised static-transient decomposition isolates visual contaminants into a dedicated transient field. This explicit separation prevents artifacts from structurally entangling with the persistent anatomical map. We curate a degradationfocused benchmark from colonoscopy videos to systematically evaluate these failure modes. Extensive experiments show that while standard baselines fail under severe optical degradation, EndoMD-SLAM preserves geometric integrity, reducing absolute trajectory error by 91% and improving rendering fidelity by 9.9 dB PSNR.
Chinese Translation
密集的3D重建对于临床内窥镜导航和文档记录至关重要。尽管高斯点云SLAM系统在这一领域显示出良好的前景，但它们在根本上依赖于严格的多视图光度一致性。在常规操作中，这一假设受到间歇性光学降解（如移动的碎片和冲洗水流）的严重破坏。标准系统错误地将这些附加在相机上的伪影融合到持久的3D几何体中，导致严重的跟踪漂移和不可逆的地图损坏。为了解决这一局限性，我们提出了EndoMD-SLAM，一个旨在通过专门的跟踪和映射机制在光学降解下保持稳定性的框架。在跟踪方面，基于记忆的门控机制检测不可靠的观测以暂停地图更新，并利用历史关键帧进行漂移感知的重新定位。在映射方面，自监督的静态-瞬态分解将视觉污染物隔离到专用的瞬态场中。这种明确的分离防止了伪影与持久的解剖地图在结构上纠缠。我们从结肠镜视频中策划了一个聚焦于降解的基准，以系统地评估这些失败模式。大量实验表明，尽管标准基线在严重的光学降解下失败，EndoMD-SLAM仍能保持几何完整性，将绝对轨迹误差降低91%，并将渲染保真度提高9.9 dB PSNR。
cs.CV / 155 / 2608.08951
Topology-Aware Global-Local Mamba Networks for Palm Vein Biometrics
基于拓扑的全局-局部Mamba网络用于掌静脉生物识别
Abstract
Palm-vein recognition is a fine-grained biometric task in which both local vascular texture and the global layout of the vessel tree carry discriminative information, while public datasets remain limited. We propose a topology-aware global-local backbone that combines multi-scale local features, a structureguided directional stream built on a fixed Sobel-magnitude edge prior, and a four-direction state-space scan global pathway within six Topology-Aware Blocks. A staged gated fusion integrates local, structural, and global representations in that order. On HKPUNIR, our method achieves 99.13% top-1 accuracy and 0.08% EER with 7.2 M parameters; on VERA Palm Vein, it achieves 92.42% accuracy and 0.61% EER. Across both datasets it attains the lowest EER among ResNet50, Vim-S, ViT-S, and GLVM at the smallest parameter count, while GLVM remains the strongest in top-1 accuracy and the cheapest in FLOPs. Code is available upon request.
Chinese Translation
掌静脉识别是一项精细化的生物识别任务，其中局部血管纹理和血管树的全局布局都携带着区分信息，而公共数据集仍然有限。我们提出了一种基于拓扑的全局-局部主干网络，该网络结合了多尺度局部特征、基于固定Sobel幅值边缘先验构建的结构引导方向流，以及在六个拓扑感知块内的四方向状态空间扫描全局路径。分阶段的门控融合按顺序整合局部、结构和全局表示。在HKPUNIR数据集上，我们的方法实现了99.13%的Top-1准确率和0.08%的等错误率（EER），参数量为720万；在VERA掌静脉数据集上，准确率为92.42%，EER为0.61%。在这两个数据集中，我们的方法在ResNet50、Vim-S、ViT-S和GLVM中以最小的参数数量达到了最低的EER，同时GLVM在Top-1准确率上仍然是最强的，并且在浮点运算量（FLOPs）上是最经济的。代码可根据请求提供。
cs.CV / 156 / 2608.08955
Damage Classification for 3D Point Cloud Data via 3D Data Analysis and Vision Foundation Model-based 2D Projections
基于3D数据分析和视觉基础模型的2D投影的3D点云数据损伤分类
Abstract
Fine-grained damage classification of 3D point cloud data (PCD) remains a persistent challenge, constrained by high computational demands and limited labeled data. This study examines two methods: 3D PCD-based damage assessment (3PDA) algorithm and 2D projection damage assessment (2PDA) In our 3PDA analysis algorithm, TDA is used to derive compact representations of 3D PCD segmented by pointNet, which are then integrated with anomaly detection algorithms to quantify structural degradation. We show that TDA effectively compresses geometric structure from VFM-segmented components into discriminative feature vectors and that anomaly detection models can reliably distinguish components with varying damage severity using only 3D PCD inputs. In the 2D projection analysis algorithm, we leverage large VFMs for granular damage detection by projecting 3D PCD into 2D views. These projections allow VFM based models to achieve competitive classification performance while requiring only a fraction of the computational cost associated with full 3D data processing. Our results demonstrate that 2D VFM pipelines in 2PDA can perform strongly on fine-grained damage classification tasks, highlighting their viability as lightweight, resource-efficient alternatives to traditional 3PDA architectures. Comparative evaluation shows that the 3PDA attains higher accuracy but only for a narrow subset of object geometries and at substantially higher computational cost due to its reliance on TDA and the scarcity of high-fidelity 3D datasets. In contrast, the 2PDA algorithm yields slightly lower accuracy but offers an order of magnitude reduction in time complexity and generalizes across a far broader range of object categories.
Chinese Translation
3D点云数据（PCD）的细粒度损伤分类仍然是一个持续的挑战，受到高计算需求和有限标注数据的限制。本研究考察了两种方法：基于3D PCD的损伤评估算法（3PDA）和2D投影损伤评估（2PDA）。在我们的3PDA分析算法中，利用拓扑数据分析（TDA）从由PointNet分割的3D PCD中提取紧凑表示，然后与异常检测算法结合，以量化结构退化。我们展示了TDA能够有效地将VFM（视觉基础模型）分割组件的几何结构压缩为判别特征向量，并且异常检测模型可以仅使用3D PCD输入可靠地区分具有不同损伤严重程度的组件。在2D投影分析算法中，我们通过将3D PCD投影到2D视图中，利用大型VFM进行细粒度损伤检测。这些投影使得基于VFM的模型在仅需较少的计算成本的情况下，能够实现与完整3D数据处理相竞争的分类性能。我们的结果表明，2PDA中的2D VFM管道在细粒度损伤分类任务中表现出色，突显了它们作为传统3PDA架构的轻量级、资源高效替代方案的可行性。比较评估显示，3PDA在准确性上更高，但仅适用于狭窄的对象几何子集，并且由于依赖TDA和高保真3D数据集的稀缺，其计算成本显著更高。相比之下，2PDA算法的准确性略低，但在时间复杂度上减少了一个数量级，并且在更广泛的对象类别中具有良好的泛化能力。
cs.CV / 157 / 2608.08957
RMR-Net: Degradation-Evidence-Guided Road-Image Restoration for Defect Detection
RMR-Net：基于退化证据引导的道路图像修复用于缺陷检测
Abstract
Vehicle-mounted road cameras are vulnerable to motion blur, defocus, poor illumination, and noise, which can erase thin cracks and pothole boundaries needed by road defect detectors. This paper presents RMR-Net, a compact task-aware restoration front end that estimates degradation evidence from the image, optionally fuses it with existing corruption context/parameters, conditions lightweight restoration blocks, and returns high-frequency pavement detail through a bounded residual path. The experimental scope is deliberately controlled: the conditioning information used on the Image and Vision Computing New Zealand (IVCNZ) pothole dataset and the Road Damage Dataset: Potholes, Cracks and Manholes (PCM) consists of saved synthetic-generator parameters, not measured vehicle telemetry. A clean-trained, frozen YOLO11s detector evaluates every image source. Across eight held-out degradation conditions, RMR-Net obtains the highest mAP50 in seven, including 0.140-0.427 for IVCNZ motion blur and 0.060-0.233 for PCM defocus. A compact ablation identifies the bounded detail path as the largest local contributor, while degradation conditioning and detector-aware stability terms provide complementary guidance.
Chinese Translation
车载道路摄像头容易受到运动模糊、对焦不清、光照不足和噪声的影响，这些因素可能会抹去道路缺陷检测器所需的细微裂缝和坑洞边界。本文提出了RMR-Net，一种紧凑的任务感知修复前端，它从图像中估计退化证据，并可选择性地将其与现有的损坏上下文/参数融合，调节轻量级修复模块，并通过有界残差路径返回高频路面细节。实验范围经过精心控制：在新西兰图像与视觉计算（IVCNZ）坑洞数据集和道路损坏数据集：坑洞、裂缝和人孔（PCM）上使用的条件信息由保存的合成生成器参数组成，而非测量的车辆遥测数据。经过干净训练的冻结YOLO11s检测器评估每个图像源。在八种保留的退化条件下，RMR-Net在七种条件下获得了最高的mAP50，包括IVCNZ运动模糊的0.140-0.427和PCM对焦不清的0.060-0.233。一项紧凑的消融实验表明，有界细节路径是最大的局部贡献者，而退化条件和检测器感知稳定性项提供了互补的指导。
cs.CV / 158 / 2608.08963
Fourier Self-Supervision for Fine-Grained Generalized Category Discovery
傅里叶自监督用于细粒度广义类别发现
Abstract
Generalized Category Discovery aims to recognize known categories while identifying novel ones within unlabeled data. Existing methods, typically based on self-supervision and contrastive learning, often struggle to capture fine-grained distinctions, relying on superficial visual cues rather than the intrinsic attributes humans use for categorization. We introduce Fourier Self-Supervision, that leverages the Fourier transform of images to enhance the discrimination of subtle differences and support the discovery of new categories. Our method employs a dual frequency filtering strategy: a low-pass filter first extracts broad, abstract attributes that capture high-level category information, while a high-pass filter emphasizes fine details such as edges and textures that are essential for fine-grained recognition. Each operates on a dedicated latent space, and their overlapping representations together yield a richer, more complete feature space. This dual-frequency approach not only refines feature extraction to identify novel categories, but also strengthens the model's discriminative power in fine-grained category discovery. Experiments on multiple fine-grained datasets show that incorporating Fourier Self-Supervision outperforms state-of-the-art methods, even when the number of classes is unknown, demonstrating its effectiveness for Generalized Category Discovery. Our code is available at: https://github.com/SarahRastegar/FourEx.
Chinese Translation
广义类别发现旨在识别已知类别，同时在未标记数据中识别新类别。现有方法通常基于自监督和对比学习，往往难以捕捉细粒度的区别，依赖于表面的视觉线索，而非人类用于分类的内在属性。我们提出了傅里叶自监督（Fourier Self-Supervision），利用图像的傅里叶变换来增强对微妙差异的辨别能力，并支持新类别的发现。我们的方法采用双频过滤策略：低通滤波器首先提取广泛的、抽象的属性，以捕捉高层次的类别信息，而高通滤波器则强调细节，如边缘和纹理，这些对于细粒度识别至关重要。每个滤波器在专用的潜在空间中操作，它们的重叠表示共同产生一个更丰富、更完整的特征空间。这种双频方法不仅精炼了特征提取以识别新类别，还增强了模型在细粒度类别发现中的区分能力。在多个细粒度数据集上的实验表明，结合傅里叶自监督的方法优于最先进的技术，即使在类别数量未知的情况下，也证明了其在广义类别发现中的有效性。我们的代码可在以下链接获取：https://github.com/SarahRastegar/FourEx。
cs.CV / 159 / 2608.08977
Detecting Clear Contact Lenses for Iris Recognition: A Two-Stage Mask-Guided Attention Approach
清晰隐形眼镜在虹膜识别中的检测：一种两阶段的掩膜引导注意力方法
Abstract
This work focuses on the impact and detection of clear contact lenses in the context of iris recognition. While the detection of cosmetic or patterned contact lenses has been extensively studied under the presentation attack detection (PAD) paradigm, clear prescription contact lenses, that are typically transparent, have received comparatively less attention despite their widespread use. Unlike patterned lenses, clear lenses introduce no salient texture artifact, making them difficult to detect and are often assumed to have no impact on iris recognition. We first examine this assumption using the commercial VeriEye matcher on four benchmark datasets and show that clear lenses marginally degrade genuine match scores and increase verification error. We then propose a two-stage contact-lens detection framework. Stage~1 uses an existing PAD model to identify patterned lenses, while Stage~2 focuses on the more challenging clear-lens versus no-lens distinction using a ConvNeXt-Base model equipped with Mask-Guided Spatial Attention (MGSA). The proposed MGSA module incorporates a Hough-derived anatomical ROI mask together with learned spatial attention and Squeeze-and-Excitation channel recalibration, allowing the network to focus on subtle limbal cues associated with clear lens wear. Across four datasets, the full pipeline consisting of both patterned and clear contact lens detection achieves between 90.0\%--98.8\% accuracy. Finally, we introduce a z-score calibration method that adjusts VeriEye match scores when a clear lens is detected in the input images. This calibration reduces EER by 4.1\%--28.3\% across datasets, demonstrating that reliable clear contact lens detection can directly improve iris verification performance.
Chinese Translation
本研究关注清晰隐形眼镜在虹膜识别中的影响及其检测。尽管在呈现攻击检测（PAD）范式下，化妆或图案隐形眼镜的检测已经得到了广泛研究，但通常透明的清晰处方隐形眼镜却相对较少受到关注，尽管其使用十分普遍。与图案隐形眼镜不同，清晰隐形眼镜没有显著的纹理伪影，使其难以被检测，并且通常假设其对虹膜识别没有影响。我们首先使用商业化的VeriEye匹配器在四个基准数据集上检验这一假设，结果表明清晰隐形眼镜会轻微降低真实匹配分数并增加验证错误。随后，我们提出了一种两阶段的隐形眼镜检测框架。第一阶段使用现有的PAD模型识别图案隐形眼镜，而第二阶段则使用配备掩膜引导空间注意力（MGSA）的ConvNeXt-Base模型，专注于更具挑战性的清晰隐形眼镜与无隐形眼镜的区分。所提出的MGSA模块结合了Hough派生的解剖区域（ROI）掩膜以及学习的空间注意力和挤压与激励通道重校准，使网络能够关注与清晰隐形眼镜佩戴相关的微妙角膜线索。在四个数据集上，包含图案和清晰隐形眼镜检测的完整流程实现了90.0 ext{%}--98.8 ext{%}的准确率。最后，我们引入了一种z-score校准方法，当在输入图像中检测到清晰隐形眼镜时，调整VeriEye匹配分数。该校准在各数据集上将等错误率（EER）降低了4.1 ext{%}--28.3 ext{%}，证明可靠的清晰隐形眼镜检测可以直接提高虹膜验证性能。
cs.CV / 160 / 2608.08999
DeepFreqMark: End-To-End Learnable Frequency-Domain Watermarking with Spherical Attack Simulation for Latent Diffusion Models
DeepFreqMark：用于潜在扩散模型的端到端可学习频域水印技术与球面攻击仿真
Abstract
The proliferation of AI-generated images produced by Latent Diffusion Models (LDMs) has raised critical concerns regarding copyright infringement and misinformation. Although existing frequency-domain watermarking methods embed handcrafted geometric patterns into the initial latent noise prior to generation, they suffer from limited capacity and rigid pattern designs. We propose DeepFreqMark, an end-to-end learnable frequency-domain watermarking framework that replaces manual pattern engineering with a neural message encoder and decoder. To circumvent the computational bottleneck caused by Denoising Diffusion Implicit Model (DDIM) inversion during training, we introduce a Spherical Linear Interpolation (Slerp)-based attack simulation. This approach operates directly on the noise latent while strictly preserving the Gaussian variance. Extensive experiments demonstrate that DeepFreqMark achieves significantly lower Bit Error Rates (BER) than baseline methods under real-world attacks and scales to 256 bits message capacity. Our source code is available at https://github.com/chenhsiu48/DeepFreqMark.
Chinese Translation
潜在扩散模型（LDMs）生成的人工智能图像的快速增长引发了关于版权侵犯和错误信息的严重担忧。尽管现有的频域水印方法在生成之前将手工设计的几何图案嵌入到初始潜在噪声中，但它们在容量和图案设计上存在局限性。我们提出了DeepFreqMark，这是一种端到端可学习的频域水印框架，使用神经消息编码器和解码器替代手动图案工程。为了绕过训练过程中由去噪扩散隐式模型（DDIM）反演引起的计算瓶颈，我们引入了一种基于球面线性插值（Slerp）的攻击仿真方法。该方法直接作用于噪声潜在，同时严格保持高斯方差。大量实验表明，DeepFreqMark在真实世界攻击下的比特错误率（BER）显著低于基线方法，并且能够扩展到256位的消息容量。我们的源代码可在 https://github.com/chenhsiu48/DeepFreqMark 获取。
cs.CV / 161 / 2608.09006
SignLlama: Enhancing Gloss-free Sign Language Translation by Prioritizing Visual Features for LLMs
SignLlama：通过优先考虑视觉特征来增强无注释手语翻译
Abstract
Large Language Models (LLMs) have achieved remarkable success across a wide range of tasks. However, fine-tuning LLMs for Gloss-Free Sign Language Translation (GFSLT) remains a challenge. In this paper, we investigate how to effectively adapt LLMs to the GFSLT task. We show that there are two key issues that need to be solved: (1) the inherent distributional gap between visual feature inputs and text feature inputs makes it difficult for LLMs to interpret visual inputs; and (2) existing approaches typically concatenate visual and textual features in an autoregressive framework, which leads to the model overemphasizing textual inputs and deprioritizing visual cues, as LLMs are pretrained predominantly on text-centric data. To address the first challenge, we propose a simple yet effective method named Filtered Pseudo-Gloss CTC Pretraining, which leverages filtered pseudo-gloss sequences generated from text sequences to supervise the training of the visual backbone. To tackle the second issue, we introduce a Visual-Prioritized Distillation training strategy. Specifically, we define a visual-only prediction path in which text inputs are masked, and the model is required to generate the target sequence relying solely on visual inputs. To guide this path, the outputs from the standard visual-textual prediction are then distilled into the visual-only prediction path, encouraging the model to prioritize visual features. Comprehensive experiments and qualitative analyses demonstrate the effectiveness of the proposed model. The proposed SignLlama achieves very competitive performance on multiple datasets for GFSLT tasks, without using any extra modalities or external sign language datasets for pretraining.
Chinese Translation
大型语言模型（LLMs）在广泛的任务中取得了显著成功。然而，为无注释手语翻译（GFSLT）微调LLMs仍然是一个挑战。本文探讨了如何有效地将LLMs适应于GFSLT任务。我们发现需要解决两个关键问题：（1）视觉特征输入与文本特征输入之间固有的分布差距使得LLMs难以解释视觉输入；（2）现有方法通常在自回归框架中连接视觉和文本特征，这导致模型过度强调文本输入而忽视视觉线索，因为LLMs主要在以文本为中心的数据上进行预训练。为了解决第一个挑战，我们提出了一种简单而有效的方法，称为过滤伪注释CTC预训练（Filtered Pseudo-Gloss CTC Pretraining），该方法利用从文本序列生成的过滤伪注释序列来监督视觉主干的训练。为了解决第二个问题，我们引入了一种视觉优先的蒸馏训练策略。具体而言，我们定义了一个仅视觉预测路径，其中文本输入被屏蔽，模型需要仅依靠视觉输入生成目标序列。为了引导这一路径，来自标准视觉-文本预测的输出被蒸馏到仅视觉预测路径中，鼓励模型优先考虑视觉特征。全面的实验和定性分析证明了所提出模型的有效性。所提出的SignLlama在多个GFSLT任务的数据集上取得了非常有竞争力的表现，且未使用任何额外的模态或外部手语数据集进行预训练。
cs.CV / 162 / 2608.09048
GeoAI-based post-segmentation quality validation of building footprints via spatial feature engineering
基于GeoAI的建筑轮廓后分割质量验证通过空间特征工程
Abstract
Deep learning-based building footprint extraction from high-resolution imagery often produces topologically inconsistent vectors unfit for direct GIS database ingestion. To address this, we present a multidomain GeoAI quality control framework that automates error detection to systematically purify vector footprint databases. Candidate footprints were generated across five UAV survey sites in Bangladesh using U-Net (ResNet-34) and SAM-LoRA (ViT-B). The extracted raster masks were vectorized, geometrically regularized, and consolidated under a spatial-exclusivity constraint to eliminate duplicate representations. We used twenty-four predictors capturing geometric, spatial-contextual, and raster-derived spectral and texture properties. Machine Learning (ML) classifiers were trained on a development partition (Sites B-D) and rigorously validated on a spatially independent test set (Site E) excluded from hyperparameter tuning and class balancing. The experimental results demonstrate that geometric and spatial-contextual predictors using Decision Tree (DT) provide the most effective discriminatory evidence for identifying object-level boundary deformations. DT achieved an accuracy of 95.31%, an F1-score of 91.06%, and a Matthews correlation coefficient (MCC) of 0.880 on the unseen testing site. At the database level, this framework successfully identified 87.34% of erroneous footprints while maintaining 98.31% of acceptable structures, reducing the residual error proportion from 27.32% to 4.62% and improving final database purity to 95.38%. This translates into a relative error reduction of 83.09%. The findings indicate that post-segmentation object-level ML provides a highly transferable, robust mechanism for automated quality assurance in production-ready geographic information system (GIS) workflows.
Chinese Translation
基于深度学习的高分辨率影像建筑轮廓提取通常会产生拓扑不一致的矢量，无法直接用于地理信息系统（GIS）数据库的摄取。为了解决这一问题，我们提出了一种多领域的GeoAI质量控制框架，自动化错误检测，以系统性地净化矢量轮廓数据库。候选轮廓是在孟加拉国五个无人机调查地点生成的，使用了U-Net（ResNet-34）和SAM-LoRA（ViT-B）。提取的栅格掩膜被矢量化、几何规范化，并在空间排他性约束下整合，以消除重复表示。我们使用了二十四个预测因子，捕捉几何、空间上下文以及栅格衍生的光谱和纹理特性。机器学习（ML）分类器在开发分区（B-D站点）上进行了训练，并在一个空间独立的测试集（E站点）上进行了严格验证，该测试集未参与超参数调优和类别平衡。实验结果表明，使用决策树（DT）的几何和空间上下文预测因子提供了识别对象级边界变形的最有效区分证据。DT在未见测试站点上达到了95.31%的准确率、91.06%的F1-score和0.880的马修斯相关系数（MCC）。在数据库层面，该框架成功识别了87.34%的错误轮廓，同时保持了98.31%的可接受结构，将残余错误比例从27.32%降低至4.62%，最终数据库纯度提高至95.38%。这意味着相对错误减少了83.09%。研究结果表明，后分割对象级机器学习为生产就绪的地理信息系统（GIS）工作流提供了一种高度可转移、稳健的自动化质量保证机制。
cs.CV / 163 / 2608.09052
Triple Expert Learning from Noisy Labels for Semi-Supervised Vision Foundation Model Adaptation
基于噪声标签的三重专家学习用于半监督视觉基础模型适应
Abstract
Semi-supervised adaptation of vision foundation models (VFMs) commonly freezes the pretrained backbone and updates lightweight modules such as LoRA. However, pseudo-labels have mixed reliability, and a single LoRA adapter must absorb reliable, ambiguous, and noisy gradients in the same low-rank space. This can make VFM adaptation sensitive to pseudo-label noise. We propose \textbf{TriNoL}, a \textbf{Tri}ple-expert learning framework from \textbf{No}isy \textbf{L}abels for semi-supervised VFM adaptation. TriNoL routes unlabeled samples into three confidence regions and assigns them to three LoRA experts: a Positive Expert for high-confidence pseudo-labels, an Alignment Expert for medium-confidence ambiguous samples, and a Negative Expert for low-confidence noisy samples. The VFM backbone remains frozen, and only the LoRA experts and classifier head are updated. By separating different pseudo-label reliability regions into specialized adaptation paths, TriNoL improves robustness to noisy supervision while keeping the training cost low.
Chinese Translation
视觉基础模型（VFM）的半监督适应通常冻结预训练的主干网络，并更新轻量级模块，如 LoRA。然而，伪标签的可靠性参差不齐，单一的 LoRA 适配器必须在同一低秩空间中吸收可靠、模糊和噪声梯度。这可能使 VFM 适应对伪标签噪声敏感。我们提出了 extbf{TriNoL}，一个基于 extbf{噪声标签}的 extbf{三重专家}学习框架，用于半监督 VFM 适应。TriNoL 将未标记样本划分为三个置信度区域，并将其分配给三个 LoRA 专家：高置信度伪标签的正专家（Positive Expert）、中等置信度模糊样本的对齐专家（Alignment Expert）和低置信度噪声样本的负专家（Negative Expert）。VFM 主干保持冻结，只有 LoRA 专家和分类头被更新。通过将不同伪标签可靠性区域分离到专门的适应路径中，TriNoL 提高了对噪声监督的鲁棒性，同时保持了较低的训练成本。
cs.CV / 164 / 2608.09057
Model the Edit, Not the Image: Visual Autoregressive Editing from a Source-Centric Perspective
建模编辑，而非图像：从源中心视角进行视觉自回归编辑
Abstract
Next-scale visual autoregressive models (VARs) have emerged as a powerful generative paradigm, producing high-quality images through efficient coarse-to-fine prediction. However, their potential for text-guided image editing remains largely underexplored. Existing training-free VAR editing approaches often formulate editing as target-conditioned regeneration guided or constrained by the source image, and may rely on inversion, test-time optimization, attention control, or user-provided masks. This generation-centric formulation does not fully exploit the multiscale source representations provided by VARs and may introduce additional computation or intervention. We instead take a source-centric perspective on VAR editing, in which the encoded source image tokens serve as the primary visual state and the editing process focuses on condition-induced changes. Based on this perspective, we propose \textbf{EditMod}, which compares source- and target-conditioned predictions under a shared autoregressive context, treats their difference as a scale-wise editing direction, and applies it as a residual update to source tokens at selected scales. Experiments show that EditMod achieves leading source-image fidelity while maintaining strong text alignment, and completes end-to-end editing of a 1K image in only 1.57 seconds on a single A100 GPU without per-image preparation.
Chinese Translation
下一代视觉自回归模型（VARs）已成为一种强大的生成范式，通过高效的粗到细预测生成高质量图像。然而，它们在文本引导的图像编辑方面的潜力仍然未被充分探索。现有的无训练VAR编辑方法通常将编辑形式化为受目标条件引导或受源图像约束的再生，并可能依赖于反演、测试时优化、注意力控制或用户提供的掩码。这种以生成为中心的形式并未充分利用VARs提供的多尺度源表示，并可能引入额外的计算或干预。相反，我们从源中心的视角看待VAR编辑，其中编码的源图像标记作为主要视觉状态，编辑过程集中于条件引起的变化。基于这一视角，我们提出了 extbf{EditMod}，该方法在共享的自回归上下文中比较源条件和目标条件的预测，将它们的差异视为尺度方向的编辑，并将其作为残差更新应用于选定尺度的源标记。实验表明，EditMod在保持强文本对齐的同时，实现了领先的源图像保真度，并且在单个A100 GPU上仅需1.57秒即可完成1K图像的端到端编辑，无需对每幅图像进行准备。
cs.CV / 165 / 2608.09083
Learning human joint torques from pixels
从像素中学习人类关节扭矩
Abstract
Estimating human joint torques from visual observations is a key step toward bringing biomechanical analysis from controlled laboratories to real-world movement scenarios. Existing torque estimation methods typically depend on surface electromyography, motion-capture markers, force plates, or simulated imitation data, which limits their applicability to ordinary RGB images. In this work, we introduce VID, a vision-based inverse dynamics dataset and benchmark for predicting human joint torques directly from real monocular images. VID contains 63,369 synchronized frames with real human images, kinematic annotations, anthropometric attributes, and OpenSim-derived dynamic labels, providing paired visual and biomechanical supervision for real-image inverse dynamics. We further define a standardized evaluation protocol covering overall torque estimation, joint-specific analysis, and action-specific prediction. To establish a strong reference model, we propose VID-Network, which combines pose-pretrained spatial probabilistic features, marker regression, and temporal torque inference to recover joint torques from image sequences. Experiments on VID show that VID-Network achieves an overall mPJE of 1.7612 N$\cdot$m/kg, improving over the best compared baseline by 39.81\%, and obtains the lowest error across all evaluated joint types and most action categories. VID establishes a first practical benchmark for vision-driven human inverse dynamics and provides a foundation for studying biomechanical inference in less constrained environments.
Chinese Translation
从视觉观察中估计人类关节扭矩是将生物力学分析从受控实验室带入现实运动场景的关键步骤。现有的扭矩估计方法通常依赖于表面肌电图、运动捕捉标记、力板或模拟模仿数据，这限制了它们在普通RGB图像中的适用性。在本研究中，我们引入了VID，一个基于视觉的逆动力学数据集和基准，用于直接从真实单目图像预测人类关节扭矩。VID包含63,369帧同步的真实人类图像、运动学注释、人体测量属性和基于OpenSim的动态标签，为真实图像的逆动力学提供了配对的视觉和生物力学监督。我们进一步定义了一个标准化的评估协议，涵盖整体扭矩估计、关节特定分析和动作特定预测。为了建立一个强有力的参考模型，我们提出了VID-Network，该网络结合了姿态预训练的空间概率特征、标记回归和时间扭矩推断，从图像序列中恢复关节扭矩。在VID上的实验表明，VID-Network实现了整体的mPJE为1.7612 N·m/kg，相较于最佳对比基线提高了39.81%，并在所有评估的关节类型和大多数动作类别中获得了最低的误差。VID建立了一个针对视觉驱动的人类逆动力学的首个实用基准，并为在不太受限的环境中研究生物力学推断提供了基础。
cs.CV / 166 / 2608.09091
TLDChoiceNet: Quantitatively Choosing a Transfer Learning Dataset
TLDChoiceNet：定量选择迁移学习数据集
Abstract
Transfer learning is particularly useful in settings with limited training data, and within image classification it is common to transfer learn upon massive datasets like ImageNet , CIFAR-100, or COCO . Qualitatively, it seems a transfer learning dataset should have both more classes and more examples per class than the fine tuning dataset; however, a quantitative method to choose the best transfer learning dataset does not currently exist. In this paper, we design TLDChoiceNet, a model to choose the best transfer learning dataset given a fine tuning dataset by predicting the test-set accuracy after fine-tuning. A simple version 1 achieves 0.154 MSE on the test dataset, while a version 2 leveraging an ImageNet pre-trained ResNet50 v2 embedding with per-class information attains a 5X lower MSE of 0.031. We further design two metrics that enable an unsupervised method of choosing an optimal transfer learning dataset: distribution distance (DD), which linearly regresses against fine-tune accuracy with an R2 of 0.89, and average class correlation (ACC), which improves the R2 to 0.97. Our results underscore that a dataset's low-level statistics can explain the transfer learning effect, and that using a pre-trained ImageNet can embed different classes further apart in latent feature space.
Chinese Translation
迁移学习在训练数据有限的情况下特别有用，在图像分类中，通常会在像 ImageNet、CIFAR-100 或 COCO 这样的大型数据集上进行迁移学习。从定性上看，迁移学习数据集应当具有比微调数据集更多的类别和每个类别更多的样本；然而，目前尚不存在一种定量方法来选择最佳的迁移学习数据集。本文设计了 TLDChoiceNet，一种模型，用于在给定微调数据集的情况下选择最佳迁移学习数据集，通过预测微调后的测试集准确率来实现。简单版本 1 在测试数据集上达到了 0.154 的均方误差 (MSE)，而版本 2 利用在 ImageNet 上预训练的 ResNet50 v2 嵌入和每个类别的信息，达到了 5 倍更低的均方误差 0.031。我们进一步设计了两个指标，使得选择最佳迁移学习数据集的无监督方法成为可能：分布距离 (DD)，其与微调准确率进行线性回归，R² 为 0.89；平均类别相关性 (ACC)，将 R² 提高至 0.97。我们的结果强调了数据集的低级统计特征可以解释迁移学习效应，并且使用预训练的 ImageNet 可以在潜在特征空间中将不同类别嵌入得更远。
cs.CV / 167 / 2608.09097
SI-Edit: Toward Sketch-Instruction Guided Local Image Editing with Pixel-Level Precision
SI-Edit：朝着基于草图指令的局部图像编辑实现像素级精度
Abstract
Despite rapid advances in generative models, achieving pixel-level precision in sketch-based image editing remains a persistent challenge, particularly for fine-grained local deformations. This gap stems primarily from the critical shortage of high-quality, publicly available benchmark datasets that jointly provide geometric constraints and semantic instructions. To address this issue, we first introduce **SI-Data**, a high-quality dataset specifically designed for instruction-guided local sketch editing. We develop an automated pipeline leveraging Multimodal Large Language Models (MLLMs) to synthesize comprehensive quadruplets comprising original images, local geometric sketches, semantic instructions, and corresponding edited images. By providing both reliable spatial anchors and explicit semantic intent, SI-Data uniquely enables collaborative spatial-semantic learning. Building upon this, we propose a collaborative framework called **SI-Edit** that integrates semantic instructions with precise geometric constraints. Furthermore, to address the lack of standardized evaluation, we establish a comprehensive set of metrics designed to measure both structural fidelity (e.g., sketch-to-edge alignment) and semantic adherence. Experimental results demonstrate that SI-Edit provides more reliable structural control than baselines for sketch-based image editing, and achieves precise, pixel-level local refinements aligned with user intent. The data and code are released on the [project page](https://github.com/ywxsuperstar/SIEdit).
Chinese Translation
尽管生成模型迅速发展，但在基于草图的图像编辑中实现像素级精度仍然是一个持续的挑战，尤其是在细粒度局部变形方面。这一差距主要源于高质量、公开可用的基准数据集的严重短缺，这些数据集共同提供几何约束和语义指令。为了解决这一问题，我们首先引入了**SI-Data**，这是一个专门为指令引导的局部草图编辑设计的高质量数据集。我们开发了一个自动化管道，利用多模态大型语言模型（Multimodal Large Language Models, MLLMs）合成包含原始图像、局部几何草图、语义指令和相应编辑图像的全面四元组。通过提供可靠的空间锚点和明确的语义意图，SI-Data独特地实现了协作的空间-语义学习。在此基础上，我们提出了一个名为**SI-Edit**的协作框架，将语义指令与精确的几何约束相结合。此外，为了解决缺乏标准化评估的问题，我们建立了一套全面的指标，旨在测量结构保真度（例如，草图与边缘的对齐）和语义遵循。实验结果表明，SI-Edit在基于草图的图像编辑中提供了比基线更可靠的结构控制，并实现了与用户意图相一致的精确像素级局部细化。数据和代码已在[项目页面](https://github.com/ywxsuperstar/SIEdit)上发布。
cs.CV / 168 / 2608.09101
Contrastive Mask Fidelity: Reference-Free Auditing of Ground-Truth Masks in Remote Sensing Semantic Segmentation
对比掩膜忠实度：遥感语义分割中无参考真值掩膜的审计
Abstract
Semantic segmentation models are trained and evaluated against human-drawn masks, yet remote-sensing annotations are often coarse, incomplete, or misaligned; high overlap scores may then reflect agreement with imperfect labels rather than faithfulness to the image, creating an evaluation paradox. We introduce Contrastive Mask Fidelity (CMF), a training-free, reference-free metric that scores competing class masks directly against image evidence. CMF composites keep and erase counterfactual views of each mask and asks a frozen vision-language judge whether class evidence is concentrated inside the mask and absent outside. We validate CMF on controlled mask corruptions, then audit 10,731 image-class pairs across ten remote-sensing benchmarks using candidate masks from Seg-Probe, a training-free open-vocabulary probe built on SegEarth-OV3 that outperforms prior baselines on nine of ten datasets. The audit reveals systematic, class-dependent annotation distortion: man-made classes such as buildings, roads, and cars favor the candidate mask on 62-85% of pairs, whereas ambiguous land cover more often favors human annotations. On a blinded three-annotator consensus, CMF matches expert judgment on 81% of pairs, exceeding keep-only scoring, model confidence, and a trained label-quality baseline. Finally, conservative class-wise arbitration yields supervision that improves cross-domain transfer over raw annotations and matched replacement controls, positioning CMF as a scalable tool for auditing ground truth rather than presuming it infallible.
Chinese Translation
语义分割模型是基于人工绘制的掩膜进行训练和评估的，但遥感注释往往粗糙、不完整或不对齐；高重叠分数可能反映与不完美标签的一致性，而非对图像的忠实度，从而产生评估悖论。我们引入了对比掩膜忠实度（Contrastive Mask Fidelity, CMF），这是一种无训练、无参考的度量，直接根据图像证据对竞争类掩膜进行评分。CMF 组合保留和删除每个掩膜的反事实视图，并询问一个冻结的视觉-语言评判者，类证据是否集中在掩膜内部而在外部缺失。我们在受控的掩膜损坏上验证了 CMF，随后在十个遥感基准上审计了 10,731 对图像-类，使用来自 Seg-Probe 的候选掩膜，该探针是基于 SegEarth-OV3 构建的无训练开放词汇探针，在十个数据集中的九个上超越了先前的基线。审计揭示了系统性的、类依赖的注释扭曲：人造类（如建筑物、道路和汽车）在 62-85% 的对中更倾向于候选掩膜，而模糊的土地覆盖则更常倾向于人工注释。在盲评的三位注释者共识中，CMF 与专家判断在 81% 的对中匹配，超越了仅保留评分、模型置信度和训练标签质量基线。最后，保守的类级仲裁产生的监督改善了跨域迁移，相较于原始注释和匹配替代控制，CMF 被定位为审计真值的可扩展工具，而非假定其绝对可靠。
cs.CV / 169 / 2608.09110
View-Adaptive Renderer for View-Consistent 2D-to-3D Generation
视角自适应渲染器用于视角一致的2D到3D生成
Abstract
Reconstructing 3D shapes from a single image remains a fundamental yet challenging problem in computer vision. Traditional monocular 3D generation pipelines typically synthesize multiple views from a single input image before applying Neural Radiance Field (NeRF)-based reconstruction. However, inherent projective ambiguities often produce visual discontinuities across generated viewpoints, leading to inaccuracies in reconstructed 3D models. Current solutions either incur significant additional computational burdens or fail to adequately resolve practical inconsistencies between synthesized views. To address these limitations, we propose a novel viewpoint-adaptive neural rendering framework that enables robust 3D reconstruction even when given partially inconsistent multi-view inputs. Our approach introduces view-adaptive neural renderers that independently correct viewpoint-dependent errors while simultaneously sharing a global feature backbone to preserve structural coherence. Furthermore, we propose a self-attention fusion module that adaptively integrates multi-view information, ensuring geometric consistency without relying heavily on indirect regularizations or computationally intensive methods. Through extensive experiments, we demonstrate that our method consistently improves 3D reconstruction fidelity. Importantly, our approach achieves near state-of-the-art performance without diffusion-based SDS supervision, relying primarily on photometric rendering loss with lightweight attention regularizers. This balance between accuracy and efficiency makes the proposed framework highly practical for real-world applications.
Chinese Translation
从单幅图像重建3D形状仍然是计算机视觉中的一个基本而具有挑战性的问题。传统的单目3D生成管道通常在应用基于神经辐射场（NeRF）的重建之前，从单个输入图像合成多个视图。然而，固有的投影歧义往往会导致生成视点之间的视觉不连续性，从而导致重建的3D模型不准确。目前的解决方案要么会带来显著的额外计算负担，要么无法充分解决合成视图之间的实际不一致性。为了解决这些局限性，我们提出了一种新颖的视角自适应神经渲染框架，即使在给定部分不一致的多视图输入时，也能实现稳健的3D重建。我们的方法引入了视角自适应神经渲染器，独立地纠正视角依赖的错误，同时共享一个全局特征主干以保持结构一致性。此外，我们提出了一种自注意力融合模块，能够自适应地整合多视图信息，确保几何一致性，而不依赖于间接正则化或计算密集型方法。通过大量实验，我们证明了我们的方法在3D重建保真度方面的一致性提升。重要的是，我们的方法在没有基于扩散的SDS监督的情况下，达到了接近最先进的性能，主要依赖于光度渲染损失和轻量级注意力正则化器。这种准确性与效率之间的平衡使得所提出的框架在实际应用中具有很高的实用性。
cs.CV / 170 / 2608.09122
Visual Distortion Detection in UGC Images Using Large Multimodal Models
利用大型多模态模型进行用户生成内容图像中的视觉失真检测
Abstract
The localized depiction of perceptual quality has long been a crucial, yet underexplored, challenge in image quality assessment (IQA). Existing approaches based on large multimodal models (LMMs) predominantly rely on text-driven supervised fine-tuning (SFT). However, this training paradigm exhibits notable limitations in detection accuracy. Moreover, synthetically distorted images, which are often used as the primary training data source, show a significant generalization gap when deployed in real-world scenarios; thus, the \textbf{synthetic-to-authentic (\textit{S2A})} problem represents a critical challenge. Motivated by these issues, we propose \textbf{\textit{VIGIL}}, which leverages the LMM architecture for precise visual distortion detection. From a candidate pool of over 1000K samples, we construct the \textbf{\textit{VIGIL-140K}} training set, which consists of over 140K distorted images. These images are obtained through rigorous quality filtering and carefully crafted distortion injection, covering 8 major synthetic distortion categories. Our model leverages different layers of the large language model (LLM) decoder, treating them as \textit{multiple detectors} that perform synchronous distortion detection using multi-level features. Additionally, we retain distortion cues from predictions assigned to the non-distortion class, which helps mitigate the ambiguous foreground-background (\textit{FG-BG}) separation commonly encountered in the \textit{S2A} problem. After post-processing, our model consistently outperforms strong baselines on both in-domain synthetic distortion detection and \textit{S2A} tasks.
Chinese Translation
感知质量的局部描述长期以来一直是图像质量评估（IQA）中一个重要但未被充分探索的挑战。现有基于大型多模态模型（LMMs）的方法主要依赖于文本驱动的监督微调（SFT）。然而，这种训练范式在检测准确性方面表现出显著的局限性。此外，合成失真图像通常作为主要训练数据来源，但在实际场景中部署时显示出显著的泛化差距；因此， extbf{合成到真实（ extit{S2A}）}问题构成了一个关键挑战。受到这些问题的启发，我们提出了 extbf{ extit{VIGIL}}，该模型利用LMM架构进行精确的视觉失真检测。在超过1000K样本的候选池中，我们构建了 extbf{ extit{VIGIL-140K}}训练集，其中包含超过140K张失真图像。这些图像通过严格的质量筛选和精心设计的失真注入获得，涵盖了8个主要的合成失真类别。我们的模型利用大型语言模型（LLM）解码器的不同层，将其视为 extit{多个检测器}，使用多级特征进行同步失真检测。此外，我们保留了分配给非失真类别的预测中的失真线索，这有助于减轻在 extit{S2A}问题中常见的模糊前景-背景（ extit{FG-BG}）分离。经过后处理，我们的模型在领域内的合成失真检测和 extit{S2A}任务上始终优于强基线。
cs.CV / 171 / 2608.09133
When Latents Forget Pixels: Restoring Fidelity in Diffusion Transformer Super-Resolution
当潜变量遗忘像素：在扩散变换器超分辨率中恢复保真度
Abstract
Image super-resolution (SR) with large generative models has recently achieved remarkable perceptual quality, yet maintaining fidelity to the LR observation remains challenging. In particular, we observe that diffusion transformers (DiTs) built on latent representations suffer from a critical limitation: the compression bottleneck of the VAE weakens fine-grained spatial information, leading to hallucinated details that are weakly grounded in the input image. In this work, we revisit generative SR from a representation perspective and propose a pixel-grounded super-resolution (PGSR) framework that preserves LR-observed pixel evidence before VAE compression and reuses it throughout restoration. Instead of relying solely on the compressed latent condition, PGSR extracts pre-VAE pixel evidence from the upsampled LR image and reuses it at two stages. First, Condition-Side Trajectory Guidance fuses LR-derived pixel evidence with the latent LR condition to guide the latent restoration trajectory. Second, Decoder-Side Pixel Grounding injects multi-scale pixel features into the frozen VAE decoder to ground the final rendering with LR-observed cues. To efficiently adapt large pretrained DiT models, we keep the latent autoencoder and main flow-matching backbone frozen, and train only lightweight restoration modules. We further study an efficient local-window attention variant for improved high-resolution efficiency and scalability. Extensive experiments demonstrate that PGSR improves the realism--fidelity trade-off and produces more faithful, visually convincing results than existing latent generative SR approaches.
Chinese Translation
基于大型生成模型的图像超分辨率（SR）最近在感知质量上取得了显著进展，但保持对低分辨率（LR）观测的保真度仍然具有挑战性。特别地，我们观察到基于潜在表示的扩散变换器（DiTs）存在一个关键限制：变分自编码器（VAE）的压缩瓶颈削弱了细粒度的空间信息，导致生成的细节与输入图像的关联性较弱。在本研究中，我们从表示的角度重新审视生成超分辨率，并提出了一种像素基础超分辨率（PGSR）框架，该框架在VAE压缩之前保留LR观测的像素证据，并在恢复过程中重复使用这些证据。PGSR并不单纯依赖于压缩的潜在条件，而是从上采样的LR图像中提取VAE前的像素证据，并在两个阶段中重复使用。首先，条件侧轨迹引导将LR派生的像素证据与潜在LR条件融合，以指导潜在恢复轨迹。其次，解码器侧像素基础将多尺度像素特征注入冻结的VAE解码器，以用LR观测线索为最终渲染提供基础。为了高效适应大型预训练的DiT模型，我们保持潜在自编码器和主要流匹配主干不变，仅训练轻量级的恢复模块。我们进一步研究了一种高效的局部窗口注意力变体，以提高高分辨率的效率和可扩展性。大量实验表明，PGSR改善了真实感与保真度之间的权衡，并产生了比现有潜在生成SR方法更真实、更具视觉说服力的结果。
cs.CV / 172 / 2608.09137
Bright-Channel Retinex Enhancement with a Conditional Overdispered-Noise Analysis
基于条件过度分散噪声分析的亮通道Retinex增强
Abstract
I present a training-free low-light enhancement method that combines local bright-channel illumination estimation, Retinex division, and edge-preserving denoising. For a fixed illumination estimate, a conditional Negative -Binominal psueduo-count method characterises the heteroscedastic noise amplified by division. The unconstrained reflectance ratio is the pixelwise maximum-likelihood estimate, with a boundary solution for zero-valued observations; the implemented estimate additionally applies illumination filtering and range clipping. The NB model is a diagnostic noise analysis rather than a calibrated sensor model, and the final fixed-bandwidth bilateral filter is an empirical approximation rather than the exact Bayesian solution. On the LOL-v1 dataset, the methodobtains mean PSNR/SSIM of 17.74dB/0.739, the highest values among the evaluated with conventional methods. A 400X600 image is processed at approximately 43 FPS on an Apple M2 Pro CPU.
Chinese Translation
我提出了一种无训练的低光照增强方法，该方法结合了局部亮通道照明估计、Retinex分离和边缘保留去噪。对于固定的照明估计，条件负二项伪计数方法表征了因分离而放大的异方差噪声。无约束的反射率比是逐像素的最大似然估计，并针对零值观测提供了边界解；所实现的估计还应用了照明滤波和范围裁剪。负二项模型是一种诊断噪声分析，而非校准传感器模型，最终的固定带宽双边滤波器是经验近似，而非精确的贝叶斯解。在LOL-v1数据集上，该方法获得了17.74dB/0.739的平均PSNR/SSIM，是评估中传统方法中最高的值。一幅400x600的图像在Apple M2 Pro CPU上以约43帧每秒的速度处理。
cs.CV / 173 / 2608.09139
CodecArena: Codec Quality Assessment via Visual Reinforcement Learning
CodecArena：通过视觉强化学习进行编解码器质量评估
Abstract
Video coding is advancing into the low and ultra-low bitrate regime, driven by end-to-end codecs that replace the hand-crafted pipeline with jointly optimized neural networks and generative codecs that exploit the priors of video generation models. Yet the dominant metrics, LPIPS and DISTS, measure feature and texture similarity rather than content fidelity: a reconstruction that hallucinates a wrong face or blurs text into convincing strokes can still score well, even when a human rejects it instantly. To address this, we propose CodecArena, the first vision-language framework for video coding quality assessment, casting codec evaluation as source-conditioned comparative reasoning between a reference and its reconstructions. We optimize CodecArena with Facet-GRPO, a visual reinforcement learning scheme that aligns pairwise codec preferences while grounding the verdict in five fidelity facets: identity, objects, text, texture, and temporal consistency. Its facet-anchored reward uses automatically derived facet directions as weak anchors, rather than human per-facet labels, to prevent any single sub-score from dominating the holistic preference and to yield interpretable fine-grained quality judgments. To support training and evaluation in this underexplored regime, we construct two complementary resources: CodecArena-1K, a fully automatic preference dataset of 1,500 comparison groups built from traditional, neural, and generative codec reconstructions with fused vision-language and objective supervision; and CodecArena-Bench, a human-ranked benchmark with source-disjoint videos for fair out-of-domain evaluation. Extensive experiments demonstrate that CodecArena achieves state-of-the-art agreement with human judgments on source-disjoint content across diverse codecs and bitrates, surpassing perceptual metrics and prior vision-language evaluators.
Chinese Translation
视频编码正在向低比特率和超低比特率领域发展，这一进程受到端到端编解码器的推动，这些编解码器用联合优化的神经网络替代了手工制作的管道，以及利用视频生成模型先验的生成编解码器。然而，主流指标LPIPS和DISTS测量的是特征和纹理相似性，而非内容保真度：即使重建结果虚构了错误的面孔或将文本模糊成令人信服的笔画，仍然可以获得良好的评分，而人类却会立即拒绝它。为了解决这个问题，我们提出了CodecArena，这是第一个用于视频编码质量评估的视觉-语言框架，将编解码器评估视为参考与其重建之间的源条件比较推理。我们使用Facet-GRPO优化CodecArena，这是一种视觉强化学习方案，能够在五个保真度方面（身份、物体、文本、纹理和时间一致性）对成对的编解码器偏好进行对齐，同时将评判结果与这些方面相结合。其基于方面的奖励使用自动推导的方面方向作为弱锚点，而非人工的每个方面标签，以防止任何单一子评分主导整体偏好，并产生可解释的细粒度质量判断。为了支持在这一尚未充分探索的领域进行训练和评估，我们构建了两个互补资源：CodecArena-1K，这是一个完全自动化的偏好数据集，包含1,500个比较组，基于传统、神经和生成编解码器的重建，结合了视觉-语言和客观监督；以及CodecArena-Bench，这是一个具有源不重叠视频的人类排名基准，用于公平的域外评估。大量实验表明，CodecArena在源不重叠内容上与人类判断的契合度达到了最先进的水平，超越了感知指标和先前的视觉-语言评估器。
cs.CV / 174 / 2608.09143
UniMoFlow: Grounding Instruction-Driven 3D Human Motion Editing in Generation
UniMoFlow：基于生成的指令驱动3D人类动作编辑
Abstract
Instruction-driven editing of 3D human motion requires precise spatiotemporal localization, rich semantic grounding, and strict preservation of unmodified content. Existing methods either resort to training-free adaptation of generative models or rely solely on triplet supervision; however, adaptation often yields suboptimal control, and manually curated triplet datasets remain severely limited in scale and semantic diversity. To overcome this bottleneck, we ground motion editing directly within text-to-motion generation across data, architecture, and inference. At the data level, we develop a closed-loop synthesis-and-verification pipeline that produces Omni-MoEdit, a large-scale dataset spanning body-part, amplitude, temporal, action, and style edits. At the architectural level, we introduce UniMoFlow, a unified latent flow-matching model that shares broad semantic and kinematic knowledge between generation and editing. At the inference level, SAFE (Source-Anchored Flow Editing) complements UniMoFlow with controllable, source-anchored refinement. Furthermore, we augment standard evaluations with semantics-aware metrics to account for valid edits that inherently deviate from a single ground-truth reference. Extensive experiments demonstrate improved target-text alignment, edit effectiveness, and cycle consistency, while maintaining competitive source fidelity and text-to-motion generation quality.
Chinese Translation
指令驱动的3D人类动作编辑需要精确的时空定位、丰富的语义基础以及对未修改内容的严格保留。现有方法要么依赖于生成模型的无训练适应，要么仅依靠三元组监督；然而，适应通常导致次优控制，而手动策划的三元组数据集在规模和语义多样性上仍然严重有限。为了克服这一瓶颈，我们在数据、架构和推理层面直接将动作编辑与文本到动作生成相结合。在数据层面，我们开发了一个闭环合成与验证管道，生成了Omni-MoEdit，这是一个涵盖身体部位、幅度、时间、动作和风格编辑的大规模数据集。在架构层面，我们引入了UniMoFlow，这是一种统一的潜在流匹配模型，在生成和编辑之间共享广泛的语义和运动学知识。在推理层面，SAFE（源锚定流编辑）为UniMoFlow提供了可控的、源锚定的细化。此外，我们通过语义感知指标增强了标准评估，以考虑那些本质上偏离单一真实参考的有效编辑。大量实验表明，目标文本对齐、编辑有效性和循环一致性得到了改善，同时保持了竞争性的源保真度和文本到动作生成质量。
cs.CV / 175 / 2608.09145
Right Answer, Wrong Heat: Explanation-Aware Evaluation and Thermal-Grounded Feedback for MLLMs on Infrared Images
正确答案，错误热度：针对红外图像的解释感知评估与热量基础反馈的多模态大语言模型
Abstract
General-purpose multimodal large language models (MLLMs) are increasingly applied to infrared images, where they are commonly scored by answer accuracy alone. However, a correct answer does not ensure that the model's explanation is grounded in infrared thermal evidence. We introduce an explanation-aware evaluation framework that separates answer correctness, output-level explanation groundedness, and thermal grounding for infrared visual questions. Using a Dual-LLM Consensus Judge with a preliminary human-anchor calibration check, we find that correct answers can still rely on weak or visible-light evidence; withholding the original infrared image and showing only a visible-like rendering erodes thermal grounding with little accuracy change; and this erosion is observed most strongly for more capable models but disappears when infrared remains available. We further propose Thermal-Grounded Feedback (TGF), a training-free feedback loop that diagnoses explanation-side failures and revises the explanation while preserving the selected answer. On local paired-input validation, TGF improves explanation-side grounding without changing answers. These findings suggest that future trustworthy MLLMs for infrared scene understanding should be evaluated and developed to produce thermally grounded explanations rather than merely accurate answers.
Chinese Translation
通用多模态大语言模型（MLLMs）在红外图像中的应用日益增多，通常仅通过答案的准确性进行评分。然而，正确的答案并不确保模型的解释是基于红外热证据的。我们提出了一种解释感知评估框架，该框架将答案的正确性、输出级别的解释基础性和红外视觉问题的热基础性分开。通过使用双重LLM共识评估器和初步的人类锚定校准检查，我们发现正确答案仍可能依赖于薄弱或可见光证据；不提供原始红外图像，仅显示类似可见光的渲染会削弱热基础性，而准确性变化不大；这种削弱在更强大的模型中表现得最为明显，但在红外图像可用时则消失。我们进一步提出了热基础反馈（Thermal-Grounded Feedback, TGF），这是一种无训练的反馈循环，能够诊断解释方面的失败并在保留所选答案的同时修正解释。在局部配对输入验证中，TGF在不改变答案的情况下改善了解释方面的基础性。这些发现表明，未来用于红外场景理解的可信MLLMs应当被评估和开发，以产生热基础的解释，而不仅仅是准确的答案。
cs.CV / 176 / 2608.09146
Multi-Submap Implicit Neural SLAM with Local-to-Global Loop Closure for Large-Scale Scene Reconstruction
具有局部到全局回环闭合的大规模场景重建的多子地图隐式神经SLAM
Abstract
Neural Radiance Fields (NeRF)-based SLAM has demonstrated impressive results in small-scale scene reconstruction, yet scaling these methods to extensive, complex environments remains challenging due to catastrophic forgetting and accumulated trajectory drift. This paper presents a robust, large-scale neural SLAM system featuring a multi-submap architecture and a dual-tier loop closure mechanism. Specifically, we propose a progressive mapping strategy that dynamically allocates neural submaps to maintain high-fidelity representations without memory explosion. For robust pose estimation, an optical-flow-based tracking module is integrated to handle aggressive motions. To address global consistency, we introduce a local-to-global loop closure framework leveraging the foundation model for high-performance global descriptor extraction, significantly enhancing relocalization accuracy under varying viewpoints. Furthermore, an inter-submap online distillation algorithm is designed during back-end optimization to enforce geometric and appearance consistency across overlapping submap boundaries. To validate the system, we developed a customized handheld mechatronic platform and conducted extensive evaluations on both public benchmarks and our large-scale indoor-outdoor datasets. Experimental results, including direct deployment on an onboard computing unit, demonstrate that our approach outperforms state-of-the-art neural SLAM methods in reconstruction quality and localization robustness, providing a scalable solution for real-world robotic perception and digital twinning. We will release the code publicly on \href{https://github.com/dtc111111/MSN-SLAM}{https://github.com/dtc111111/MSN-SLAM} .
Chinese Translation
基于神经辐射场（NeRF）的SLAM在小规模场景重建中表现出色，但将这些方法扩展到广泛且复杂的环境仍然面临挑战，主要是由于灾难性遗忘和累积轨迹漂移。本文提出了一种稳健的大规模神经SLAM系统，具有多子地图架构和双层回环闭合机制。具体而言，我们提出了一种渐进式映射策略，动态分配神经子地图，以维持高保真度表示而不导致内存爆炸。为了实现稳健的姿态估计，我们集成了基于光流的跟踪模块，以处理剧烈运动。为了解决全局一致性问题，我们引入了一种局部到全局的回环闭合框架，利用基础模型进行高性能全局描述符提取，显著提高了在不同视角下的重定位精度。此外，在后端优化过程中设计了一种子地图间在线蒸馏算法，以确保重叠子地图边界之间的几何和外观一致性。为了验证该系统，我们开发了一个定制的手持机电平台，并在公共基准测试和我们的大规模室内外数据集上进行了广泛评估。实验结果，包括在机载计算单元上的直接部署，表明我们的方法在重建质量和定位鲁棒性方面优于最先进的神经SLAM方法，为现实世界的机器人感知和数字双胞胎提供了可扩展的解决方案。我们将公开发布代码，链接为 https://github.com/dtc111111/MSN-SLAM 。
cs.CV / 177 / 2608.09147
RefineAny3D: Depth Refinement as Semantic Alignment for Monocular 3D Detection
RefineAny3D：深度精细化作为单目3D检测的语义对齐
Abstract
Monocular 3D object detection spans two regimes: closed-set detectors operating within a fixed category vocabulary, and open-vocabulary detectors that localize arbitrary categories by leveraging depth foundation models for 3D geometry. We find that current depth foundation models, despite their strong zero-shot generalization, lack the object-level precision 3D detection demands: substituting a state-of-the-art depth foundation model for a strong detector's predicted depth degrades accuracy, even falling below the detector's own prediction. Rather than pushing detectors or depth models to be more accurate end-to-end, we treat object-level depth refinement as a stand-alone task and present RefineAny3D, a vision-language model that corrects depth without ever predicting a numerical value. Our key insight is that depth error has a direct visual signature in image space: when projected onto the image, a correctly placed box tightly encloses the object, while a too-far box projects too small and a too-close box projects too large. Depth refinement thus reduces to a visual alignment problem rather than a metric regression problem, which we instantiate by extending the VLM's vocabulary with action tokens that replace numerical depth output with categorical decisions, and by supervising the model on a large-scale chain-of-thought dataset that grounds each decision in explicit visual evidence. Applied as a single post-hoc step, RefineAny3D delivers consistent gains across closed-set detectors, open-vocabulary detectors, and 3D auto-labeling tools, and generalizes to novel categories, scenes, and cameras without retraining.
Chinese Translation
单目3D物体检测涵盖了两个领域：在固定类别词汇内操作的闭集检测器，以及通过利用深度基础模型进行3D几何定位任意类别的开放词汇检测器。我们发现，尽管当前的深度基础模型具有强大的零样本泛化能力，但缺乏物体级精度，无法满足3D检测的需求：用最先进的深度基础模型替代强检测器的预测深度会降低准确性，甚至低于检测器自身的预测。我们并不试图推动检测器或深度模型在端到端的准确性上更进一步，而是将物体级深度精细化视为一个独立任务，提出了RefineAny3D，这是一种视觉-语言模型，能够在不预测数值的情况下纠正深度。我们的关键见解是，深度误差在图像空间中具有直接的视觉特征：当投影到图像上时，正确放置的框紧密包围物体，而距离过远的框投影过小，距离过近的框投影过大。因此，深度精细化简化为一个视觉对齐问题，而不是一个度量回归问题。我们通过扩展视觉-语言模型的词汇，使用动作标记替代数值深度输出为类别决策，并在一个大规模的链式思维数据集上监督模型，使每个决策都基于明确的视觉证据。作为一个单一的后处理步骤，RefineAny3D在闭集检测器、开放词汇检测器和3D自动标注工具上提供了一致的性能提升，并且在不重新训练的情况下能够推广到新类别、新场景和新相机。
cs.CV / 178 / 2608.09150
OGG-FR: Orthogonal Gradient Gaming and Frequency Rectification for Unmanned Aerial Vehicle Infrared Image Super-Resolution
OGG-FR：用于无人机红外图像超分辨率的正交梯度博弈与频率矫正
Abstract
Unmanned aerial vehicle (UAV) infrared image super-resolution aims to recover weak thermal structures for deployment on resource-constrained platforms; lightweight models are therefore preferred, but multi-loss training can be unstable. A common strategy combines pixel-domain and frequency-domain objectives; however, low contrast, limited high-frequency content, and sensor-specific noise often make their gradients weakly aligned or conflicting. To address this optimization ambiguity, we propose Orthogonal Gradient Gaming and Frequency Rectification (OGG-FR), a plug-and-play optimization framework that decomposes the frequency gradient into a redundant parallel component and an orthogonal innovation component relative to the pixel gradient. In the conflict regime, OGG-FR computes a safe base gradient using the Multiple Gradient Descent Algorithm (MGDA) and adds a variance-rectified orthogonal innovation; in the compatible regime, it discards redundant parallel information and injects the orthogonal innovation according to a confidence score estimated from the high-frequency residual. Experimental results on the UAV thermal benchmark show broad gains under BI and BD degradations at $\times 4$ and $\times 8$ scales, while gradient analyses support the effectiveness of the proposed conflict-aware update rule.
Chinese Translation
无人机（UAV）红外图像超分辨率旨在恢复弱热结构，以便在资源受限的平台上部署；因此，更倾向于使用轻量级模型，但多损失训练可能不稳定。一种常见策略是结合像素域和频率域目标；然而，低对比度、有限的高频内容以及传感器特定噪声常常导致它们的梯度弱对齐或相互冲突。为了解决这一优化模糊性，我们提出了正交梯度博弈与频率矫正（OGG-FR），这是一种即插即用的优化框架，它将频率梯度分解为相对于像素梯度的冗余并行分量和正交创新分量。在冲突状态下，OGG-FR使用多梯度下降算法（MGDA）计算安全的基础梯度，并添加方差矫正的正交创新；在兼容状态下，它丢弃冗余的并行信息，并根据从高频残差估计的置信度分数注入正交创新。在UAV热成像基准测试中的实验结果显示，在$ imes 4$和$ imes 8$尺度下，BI和BD降质情况下均有显著提升，而梯度分析支持所提出的冲突感知更新规则的有效性。
cs.CV / 179 / 2608.09152
LightAIR: Lightweight Action Inversion and Riemannian Rectification for Text-based Person Anomaly Search
LightAIR：用于基于文本的人物异常搜索的轻量级动作反演与黎曼校正
Abstract
Traditional Text-based Person Search (TPS) is typically limited to matching static appearance attributes, severely neglecting dynamic action information. The Text-based Person Anomaly Search (TPAS) task bridges this gap, requiring models to locate micro-level specific abnormal behaviors while matching macro-level appearance of pedestrians. However, current TPAS methods face fundamental limitations: external explicit pose estimators are fragile in unconstrained surveillance scenarios, and implicit learning encounters visual decoupling failure under pixel-level entanglement, causing dominant appearance information to easily swallow and contaminate subtle action features. Furthermore, performing contrastive optimization on hard negative samples (``same appearance, different actions'') in conventional Euclidean spaces induces severe shortcut learning. To address these, we propose the Lightweight Action Inversion and Riemannian rectification network (LightAIR). First, it introduces textual semantic priors as anchors via a lightweight action inversion operator to extract pure action features, thereby overcoming visual-inherent coupling. Subsequently, it employs orthogonal null-space projection to constrain appearance features within the orthogonal complement space of action features, guaranteeing strict forward decoupling. Finally, we designed a gradient rectification module that computes the Riemannian gradient to constrain the backpropagation trajectory, forcing the gradient flow to update strictly along the tangent space that preserves decoupling properties, thereby cutting off harmful shortcuts. Extensive experiments on the widely used TPAS and TIPR datasets demonstrate that LightAIR significantly outperforms existing state-of-the-art methods. Codes are available at https://github.com/rainy-london/LightAIR
Chinese Translation
传统的基于文本的人物搜索（TPS）通常局限于匹配静态外观属性，严重忽视了动态动作信息。基于文本的人物异常搜索（TPAS）任务弥补了这一空白，要求模型在匹配行人宏观外观的同时，定位微观层面特定的异常行为。然而，目前的TPAS方法面临根本性限制：外部显式姿态估计器在不受约束的监控场景中脆弱，而隐式学习在像素级纠缠下遭遇视觉解耦失败，导致主导的外观信息容易淹没和污染微妙的动作特征。此外，在常规欧几里得空间中对困难负样本（“相同外观，不同动作”）进行对比优化会引发严重的捷径学习。为了解决这些问题，我们提出了轻量级动作反演与黎曼校正网络（LightAIR）。首先，它通过轻量级动作反演算子引入文本语义先验作为锚点，以提取纯粹的动作特征，从而克服视觉固有耦合。随后，它采用正交零空间投影将外观特征限制在动作特征的正交补空间内，确保严格的前向解耦。最后，我们设计了一个梯度校正模块，计算黎曼梯度以约束反向传播轨迹，迫使梯度流严格沿着保持解耦特性的切空间更新，从而切断有害的捷径。在广泛使用的TPAS和TIPR数据集上的大量实验表明，LightAIR显著优于现有的最先进方法。代码可在 https://github.com/rainy-london/LightAIR 获取。
cs.CV / 180 / 2608.09176
Not All Visual Tokens Are Equally Safe to Remove:Consequence-Sensitive Visual Token Compression
并非所有视觉标记的移除都是同样安全的：后果敏感的视觉标记压缩
Abstract
Visual token compression for vision--language models (VLMs) has largely relied on criteria such as attention, redundancy, and uncertainty to maximize average accuracy under a fixed compute budget, implicitly assuming that all errors carry equal cost. However, the consequence of an incorrect prediction on downstream tasks is rarely symmetric: misreading an invoice amount can be far more costly than misclassifying a background color. Motivated by this, we introduce consequence-sensitive visual token compression, which allocates visual computation across requests according to their potential error costs. Our method follows a calibrate-then-allocate procedure, estimating consequence-specific error-budget curves offline and applying the calibrated token budgets online using consequence signals available from question or task information. On a controlled within-task benchmark, high- and low-consequence questions are drawn from the same document images, so content alone cannot reveal which questions are costly to get wrong. In this setting, our method reduces high-stakes errors from 0.300 to 0.133 under the same total token budget, whereas a content-driven allocator performs no better than uniform allocation. Measuring how error rates change with token budget across different cost ratios, we derive an allocation frontier: uniform allocation is optimal when errors are equally costly, and token transfer toward high-consequence questions becomes increasingly beneficial as the cost gap grows. This allocation principle generalizes well across three dense vision-language benchmarks, two budget realization mechanisms (token deletion and resolution reallocation), two VLM architectures, and multiple token selection strategies. On a realistic mixed workload, consequence-sensitive allocation reduces cost-weighted error by 38% while achieving approximately 21% lower latency than full-resolution inference.
Chinese Translation
视觉-语言模型（VLMs）的视觉标记压缩在很大程度上依赖于注意力、冗余性和不确定性等标准，以在固定的计算预算下最大化平均准确性，隐含地假设所有错误的成本是相等的。然而，错误预测在下游任务中的后果通常并不对称：错误读取发票金额的成本可能远高于错误分类背景颜色。基于此，我们提出了后果敏感的视觉标记压缩，该方法根据潜在的错误成本在请求之间分配视觉计算。我们的方法遵循校准后分配的程序，离线估计特定后果的错误预算曲线，并利用来自问题或任务信息的后果信号在线应用校准的标记预算。在一个受控的任务内基准测试中，高后果和低后果的问题来自同一文档图像，因此仅凭内容无法揭示哪些问题的错误代价高。在这种情况下，我们的方法在相同的总标记预算下将高风险错误从0.300降低到0.133，而基于内容的分配器的表现不优于均匀分配。通过测量在不同成本比下标记预算变化时错误率的变化，我们推导出一个分配边界：当错误成本相等时，均匀分配是最优的，而随着成本差距的增大，向高后果问题转移标记变得越来越有利。该分配原则在三个密集视觉-语言基准、两种预算实现机制（标记删除和分辨率重新分配）、两种VLM架构和多种标记选择策略中均表现良好。在一个现实的混合工作负载中，后果敏感的分配将加权错误降低了38%，同时实现了比全分辨率推理低约21%的延迟。
cs.CV / 181 / 2608.09182
Rethinking Medical Landmark Localization with Prototype Learning-based Progressive Offset Correction
基于原型学习的渐进偏移校正重新思考医学标志定位
Abstract
Accurate landmark localization in medical images is a fundamental step for quantitative clinical measurement and downstream analysis. Existing localization methods have advanced, among which multi-stage refinement is a superior solution. Although this strategy mitigates the anatomical ambiguity inherent in single-stage global predictions, its high computational cost limits practical applicability. In this work, we propose a parameter-economic model, PPOC-LL, which leverages Prototype learning-based Progressive Offset Correction for Landmark Localization. Our contribution is three-fold. First, to drive coarse-to-fine landmark optimization, we introduce a multi-scale dynamic perception strategy for patch-level feature pyramid modeling. Second, to effectively handle anatomically similar patterns, we design a similarity-driven prototype learning mechanism that captures informative local semantics for robust offset prediction. Last, to stabilize the model learning and improve the overall performance, we incorporate a novel error-aware reliability regularization via tolerance-based balancing. We collected a large validation cohort, including two public and one private datasets spanning X-ray and ultrasound modalities, covering cephalometric, symphysis-fetal head, and fetal heart landmarks. Extensive experiments demonstrate that PPOC-LL achieves satisfactory performance with a favorable trade-off between accuracy and model complexity.
Chinese Translation
医学图像中准确的标志定位是定量临床测量和后续分析的基础步骤。现有的定位方法已经取得了进展，其中多阶段精细化是一种优越的解决方案。尽管这一策略减轻了单阶段全局预测中固有的解剖模糊性，但其高计算成本限制了实际应用。在本研究中，我们提出了一种参数经济模型PPOC-LL，该模型利用基于原型学习的渐进偏移校正进行标志定位。我们的贡献有三方面。首先，为了推动粗到细的标志优化，我们引入了一种多尺度动态感知策略，用于补丁级特征金字塔建模。其次，为了有效处理解剖上相似的模式，我们设计了一种基于相似性的原型学习机制，捕捉信息丰富的局部语义以实现稳健的偏移预测。最后，为了稳定模型学习并提高整体性能，我们通过基于容忍度的平衡引入了一种新颖的错误感知可靠性正则化。我们收集了一个大型验证队列，包括两个公共数据集和一个私有数据集，涵盖X射线和超声波模态，涉及头影测量、耻骨联合-胎头和胎心标志。大量实验表明，PPOC-LL在准确性和模型复杂性之间实现了令人满意的性能平衡。
cs.CV / 182 / 2608.09186
RAGMesh with FaME-G2E: Long-Form Text-Driven 3D Face Generation and Editing
基于 FaME-G2E 的 RAGMesh：长文本驱动的 3D 面部生成与编辑
Abstract
Text-driven 3D face generation and editing remains challenging due to the difficulty of translating long-form descriptions into fine-grained facial geometry. Existing methods primarily align global textual semantics with facial structures but often struggle to capture subtle local deformations, such as eyebrow tension, cheek contraction, and asymmetric mouth motions, resulting in limited geometric fidelity and editing precision. To facilitate fine-grained text-driven facial modeling, we first construct FaME-G2E, a large-scale multimodal dataset containing detailed text--mesh annotations and paired text--blendshape samples for unified 3D facial generation and editing. Based on this dataset, we propose RAGMesh, a retrieval-augmented framework that leverages text-correlated geometric priors to improve high-fidelity facial synthesis and editing. Specifically, the Multi-Scale Retrieval Fusion (MSRF) module retrieves semantically consistent global and regional facial priors and fuses them in the blendshape space, suppressing conflicting local deformations while preserving coherent deformation patterns. Furthermore, we introduce Adaptive RAG-guided Supervision (AdaRAGS), a region-aware constraint that explicitly aligns textual semantics with corresponding facial regions, enhancing regional controllability and editing accuracy. Extensive experiments on FaME-G2E demonstrate that RAGMesh achieves superior performance over state-of-the-art methods in local geometric accuracy, text-guided controllability, regional editing precision, and inference efficiency. Video demo is available at https://youtu.be/Yr0_XkpWcNk, and the source code and dataset will be released upon paper acceptance.
Chinese Translation
由于将长文本描述转化为精细面部几何形状的难度，文本驱动的 3D 面部生成与编辑仍然面临挑战。现有方法主要对齐全局文本语义与面部结构，但往往难以捕捉细微的局部变形，如眉毛紧张、面颊收缩和不对称的嘴部运动，导致几何保真度和编辑精度有限。为了促进精细的文本驱动面部建模，我们首先构建了 FaME-G2E，这是一个大规模的多模态数据集，包含详细的文本-网格注释和配对的文本-混合形状样本，用于统一的 3D 面部生成与编辑。在此数据集的基础上，我们提出了 RAGMesh，这是一种检索增强框架，利用文本相关的几何先验来提高高保真面部合成和编辑的效果。具体而言，多尺度检索融合（Multi-Scale Retrieval Fusion, MSRF）模块检索语义一致的全局和区域面部先验，并在混合形状空间中融合它们，抑制冲突的局部变形，同时保留一致的变形模式。此外，我们引入了自适应 RAG 引导监督（Adaptive RAG-guided Supervision, AdaRAGS），这是一种区域感知约束，明确对齐文本语义与相应的面部区域，增强区域可控性和编辑准确性。在 FaME-G2E 上的广泛实验表明，RAGMesh 在局部几何准确性、文本引导可控性、区域编辑精度和推理效率方面超越了最先进的方法。视频演示可在 https://youtu.be/Yr0_XkpWcNk 获取，源代码和数据集将在论文接受后发布。
cs.CV / 183 / 2608.09200
NBA_Streaming: A Large-Scale Benchmark for Fine-Grained Basketball Commentary Generation in Continuous Streams
NBA_Streaming：一个用于连续流中细粒度篮球评论生成的大规模基准
Abstract
Live basketball commentary generation requires determining when an event is sufficiently observable and describing it before subsequent events unfold. However, existing methods are primarily designed for pre-segmented clips or complete videos, making them unsuitable for continuous streams. Existing datasets also provide limited supervision for player identities, fine-grained actions, event attributes, and coherent event chains, restricting the factual richness of generated commentary. To address these limitations, we introduce NBA_Streaming, a large-scale benchmark for online fine-grained basketball commentary generation. It contains 307 hours of basketball broadcasts and approximately 35K temporally aligned events, with annotations of event boundaries, player identities, fine-grained actions, event chains, and natural-language commentary. By moving from isolated clips to continuous streams, NBA_Streaming enables unified evaluation of event localization, response reliability, factual grounding, and commentary quality under causal constraints. We further propose a causal two-stage framework that combines completion-first localization with ball-centric semantic grounding, enabling the system to identify complete events from observed streams and organize scene, event, identity, and action cues for commentary generation. Extensive experiments reveal the difficulty of NBA_Streaming, where existing baselines struggle with online timing, factual grounding, and fine-grained description. Our framework consistently improves over strong alternatives, while the remaining gap highlights NBA_Streaming as a valuable benchmark for streaming sports video understanding and generation.
Chinese Translation
实时篮球评论生成需要确定事件何时足够可观察，并在后续事件展开之前对其进行描述。然而，现有的方法主要针对预先分段的剪辑或完整视频，因而不适用于连续流。现有数据集在球员身份、细粒度动作、事件属性和连贯事件链的监督方面也提供了有限的信息，限制了生成评论的事实丰富性。为了解决这些局限性，我们引入了NBA_Streaming，一个用于在线细粒度篮球评论生成的大规模基准。该基准包含307小时的篮球直播和大约35,000个时间对齐的事件，附有事件边界、球员身份、细粒度动作、事件链和自然语言评论的注释。通过从孤立的剪辑转向连续流，NBA_Streaming实现了事件定位、响应可靠性、事实基础和评论质量在因果约束下的统一评估。我们进一步提出了一种因果两阶段框架，将优先完成的定位与以球为中心的语义基础相结合，使系统能够从观察到的流中识别完整事件，并组织场景、事件、身份和动作线索以生成评论。大量实验揭示了NBA_Streaming的难度，现有基线在在线时序、事实基础和细粒度描述方面表现不佳。我们的框架在强有力的替代方案上持续改进，而剩余的差距突显了NBA_Streaming作为流媒体体育视频理解和生成的宝贵基准。
cs.CV / 184 / 2608.09223
PatchHead: Learning Spatial Patch Evidence for Generalizable AI-Generated Image Detection
PatchHead：学习空间补丁证据以实现可泛化的AI生成图像检测
Abstract
AI-generated image detectors generalize poorly when their training and test images originate from different generators or datasets. Despite the rich spatial representations produced by vision foundation models like DINO, existing detectors typically classify images using only the globally aggregated CLS token. We hypothesize that globally aggregating DINO features into a single CLS token obscures spatially distributed generation traces. To test this hypothesis, we introduce PatchHead, a lightweight spatial aggregation head that preserves the two-dimensional organization of DINO patch tokens and integrates evidence across neighboring regions. During training, we freeze the pretrained DINO backbone and optimize only the inserted LoRA adapters, PatchHead, and auxiliary projection head. Across nine cross-dataset benchmarks spanning manually curated and in-the-wild settings, PatchHead ranks first on seven datasets and second on the remaining two. It improves the strongest prior method from 91.6% to 94.6% in average balanced accuracy (+3.0 points) and raises the worst-case accuracy from 82.4% to 89.4% (+6.9 points), while introducing only 8.6% more trainable parameters and 0.08% additional FLOPs. Further qualitative analysis suggests that PatchHead (i) reduces class-conditional domain discrepancy, and (ii) redirects the representation from content-dominated saliency toward spatially distributed authenticity evidence. Together, these observations provide a representation-level account of why spatial patch aggregation transfers more reliably across generators and datasets than a single CLS-based global representation. Our code and models will be made available upon acceptance.
Chinese Translation
当训练和测试图像来自不同生成器或数据集时，AI生成图像检测器的泛化能力较差。尽管像DINO这样的视觉基础模型生成了丰富的空间表示，现有的检测器通常仅使用全局聚合的CLS标记对图像进行分类。我们假设，将DINO特征全局聚合为单个CLS标记会掩盖空间分布的生成痕迹。为了验证这一假设，我们提出了PatchHead，一种轻量级的空间聚合头，能够保留DINO补丁标记的二维组织，并整合邻近区域的证据。在训练过程中，我们冻结预训练的DINO主干，只优化插入的LoRA适配器、PatchHead和辅助投影头。在跨越手动策划和野外设置的九个跨数据集基准测试中，PatchHead在七个数据集上排名第一，在其余两个数据集上排名第二。它将最强的先前方法的平均平衡准确率从91.6%提高到94.6%（+3.0个百分点），并将最差情况下的准确率从82.4%提高到89.4%（+6.9个百分点），同时仅增加了8.6%的可训练参数和0.08%的额外FLOPs。进一步的定性分析表明，PatchHead (i) 减少了类条件域差异，以及 (ii) 将表示从内容主导的显著性重定向到空间分布的真实性证据。综合这些观察结果，我们提供了一个表示层面的解释，说明为什么空间补丁聚合比基于单个CLS的全局表示在生成器和数据集之间的迁移更可靠。我们的代码和模型将在接受后提供。
cs.CV / 185 / 2608.09226
RL-Native Distillation: Exploiting Scored Trajectories for Few-Step Image Generation
RL原生蒸馏：利用评分轨迹进行少步图像生成
Abstract
Efficient text-to-image generation requires both reinforcement-learning (RL)-based reward alignment and few-step distillation, yet these procedures are typically performed sequentially, increasing training cost and risking the loss of reward gains during compression. We instead take an RL-native perspective: diffusion RL already generates reward-scored finite-step trajectories, whose intermediate states provide a natural source of distillation supervision rather than a disposable byproduct of sampling. Based on this insight, we propose REST (Reward-Enhanced Scored-Trajectory Distillation), a single-stage RL-distillation co-training framework that attaches a decoupled student to an arbitrary RL teacher. The student learns segment-wise from the teacher's evolving rollout trajectories while leaving the original teacher optimization unchanged. To prevent uniform imitation from preserving undesirable low-reward behaviors, we further introduce Advantage-Modulated Distillation (AMD), which transforms rollout advantages into signed weights over a base distillation loss. AMD strengthens supervision from preferred trajectories and mildly repels the student from low-reward ones. The resulting framework is lightweight and plug-and-play, requires no extra image rollouts, no separate distillation dataset, and no adversarial training. Experiments on compositional generation, visual text rendering, and human-preference alignment show that REST enables few-step CFG-free inference that matches or surpasses its 40-step RL teacher, with an overall additional training cost below 25% over pure RL. REST improves DrawBench PickScore over RTDMD by 0.82 while requiring only one-fifth of the training iterations.
Chinese Translation
高效的文本到图像生成需要基于强化学习（RL）的奖励对齐和少步蒸馏，但这些过程通常是顺序进行的，这增加了训练成本并且在压缩过程中可能导致奖励收益的损失。我们则采取RL原生的视角：扩散RL已经生成了奖励评分的有限步轨迹，其中间状态提供了自然的蒸馏监督来源，而不是采样的可丢弃副产品。基于这一见解，我们提出了REST（奖励增强评分轨迹蒸馏），这是一个单阶段的RL蒸馏协同训练框架，将一个解耦的学生模型附加到任意RL教师模型上。学生模型从教师模型不断演变的轨迹中进行分段学习，同时保持原教师模型的优化不变。为了防止均匀模仿保留不良的低奖励行为，我们进一步引入了优势调制蒸馏（AMD），它将轨迹优势转化为基础蒸馏损失上的带符号权重。AMD增强了来自优选轨迹的监督，并轻微地排斥学生模型低奖励轨迹。最终的框架轻量且即插即用，无需额外的图像轨迹、单独的蒸馏数据集和对抗训练。在组合生成、视觉文本渲染和人类偏好对齐的实验中，REST实现了与其40步RL教师相匹配或超越的少步CFG-free推理，整体额外训练成本低于纯RL的25%。REST在DrawBench PickScore上比RTDMD提高了0.82，同时只需五分之一的训练迭代。
cs.CV / 186 / 2608.09231
BAG: Budget-Aware Gating for Diffusion Caching
BAG：预算感知的扩散缓存门控
Abstract
Diffusion caching is a lightweight strategy that accelerates Diffusion Transformers (DiTs) by reusing intermediate features across denoising steps, but existing paradigms face a fundamental trade-off: online heuristics lack global budget awareness, whereas static schedules lack instance adaptivity and fail to flexibly adapt to varying runtime budget constraints. To bridge this gap, we present BAG (Budget-Aware Gating), a novel caching policy that unifies global budget pacing with dynamic, instance-adaptive feature reuse. Rather than relying on hand-crafted rules, BAG employs a lightweight gating network that dynamically decides whether to execute a full computation or reuse cached features at each step by jointly conditioning on the budget state and local trajectory feedback. We train this policy via offline-to-online schedule distillation, transferring the decision-making of offline-searched schedules into a compact online gate. Extensive experiments on FLUX.1-dev and Wan2.1 demonstrate that BAG consistently outperforms state-of-the-art caching methods across various speedup tiers while remaining robust across different resolutions, seeds, and guidance scales. Code will be released.
Chinese Translation
扩散缓存是一种轻量级策略，通过在去噪步骤中重用中间特征来加速扩散变换器（Diffusion Transformers, DiTs），但现有范式面临一个基本的权衡：在线启发式方法缺乏全局预算意识，而静态调度缺乏实例适应性，无法灵活适应变化的运行时预算约束。为了解决这一问题，我们提出了BAG（Budget-Aware Gating），一种新颖的缓存策略，它将全局预算节奏与动态的实例自适应特征重用相结合。BAG并不依赖手工制定的规则，而是采用一个轻量级的门控网络，该网络通过联合考虑预算状态和局部轨迹反馈，动态决定在每一步是执行完整计算还是重用缓存特征。我们通过离线到在线的调度蒸馏训练该策略，将离线搜索调度的决策转移到一个紧凑的在线门控中。在FLUX.1-dev和Wan2.1上的大量实验表明，BAG在各种加速层级上始终优于最先进的缓存方法，同时在不同分辨率、种子和引导尺度下保持稳健。代码将会发布。
cs.CV / 187 / 2608.09238
RealDenseFace: Real-time Monocular 3D Face Reconstruction from Dense UV-space Priors
RealDenseFace：基于密集UV空间先验的实时单目3D人脸重建
Abstract
Recent monocular 3D face reconstruction methods achieve high fidelity by fitting a 3D Morphable Model (3DMM) to dense priors predicted by networks, but the optimization stage is computationally expensive, often taking tens of seconds per image. We present RealDenseFace, a real-time optimization-based 3D face reconstruction method with dense UV-space network predictions. Our key idea is to formulate 3DMM fitting as a nonlinear least-squares problem and solve it with a tailored Gauss-Newton solver that converges in only a few iterations. The reconstruction is conducted in two stages. In the first stage, the network predicts two dense UV-space maps from a single RGB image: a correspondence map for UV-to-image alignment, and a relative-depth map for geometric constraints along the viewing direction. In the second stage, the solver fits per-vertex targets sampled from these maps at the vertex UV coordinates. The solver supports all three reconstruction settings: single-image fitting, offline sequence reconstruction, and online tracking. Our method achieves state-of-the-art accuracy on the NeRSemble SVFR benchmark. The online tracker runs at 80+ FPS, and the offline sequence reconstruction is over 20 times faster than previous optimization-based baselines.
Chinese Translation
最近的单目3D人脸重建方法通过将3D可变形模型（3DMM）拟合到网络预测的密集先验中，实现了高保真度，但优化阶段计算开销较大，通常每张图像需要数十秒。我们提出了RealDenseFace，这是一种基于实时优化的3D人脸重建方法，利用密集的UV空间网络预测。我们的关键思想是将3DMM拟合公式化为非线性最小二乘问题，并使用定制的高斯-牛顿求解器进行求解，该求解器仅需几次迭代即可收敛。重建过程分为两个阶段。在第一阶段，网络从单张RGB图像中预测两个密集的UV空间图：用于UV到图像对齐的对应图和用于沿视线方向的几何约束的相对深度图。在第二阶段，求解器在顶点UV坐标处拟合从这些图中采样的每个顶点目标。该求解器支持三种重建设置：单图像拟合、离线序列重建和在线跟踪。我们的方法在NeRSemble SVFR基准测试中达到了最先进的准确性。在线跟踪器的运行速度超过80帧每秒，离线序列重建速度比之前的基于优化的基线快20倍以上。
cs.CV / 188 / 2608.09244
In-Loop Model Adaptation with Coupled Latent-Noise Guidance for High-Fidelity Subject-Driven Text-to-Image Generation
基于耦合潜在噪声引导的循环模型适应用于高保真主题驱动的文本到图像生成
Abstract
Text-to-image diffusion models have achieved remarkable success in generating high-quality images from a given text prompt. Subject-driven generation aims to synthesize customized images to mimic the appearance of subjects in given reference images within different visual contexts specified by the text prompts. The central challenge here is that, when the reference image changes, the diffusion model cannot efficiently adapt to different visual contexts while consistently maintaining the subject identity. Existing methods either train the model with a large domain-specific dataset or fine-tune the model using the reference image for hundreds of iterations before actual image generation. In this work, we explore a new approach, called \textit{In-Loop Model Adaptation} (IMA), which adapts the core diffusion model at each generation step during the actual process of image generation, without being trained on the reference image before the generation process. To this end, we establish a DDIM inversion chain that maps the reference image to a sequence of latent, as well as a text-to-image generation chain which generates the image from the text prompt only. We then introduce a masked latent consistency loss and a noise regularization loss to characterize the latent-noise difference between the diffusion model and these two chains at each generation step. This coupled latent-noise loss is used to guide the in-loop model adaptation to preserve the subject identity specified by the reference image while maintaining accurate alignment with the text prompt, resulting in high-fidelity text-to-image generation. Our extensive experiments demonstrate that our proposed IMA method significantly improves the performance of subject-driven text-to-image generation.
Chinese Translation
文本到图像的扩散模型在根据给定文本提示生成高质量图像方面取得了显著成功。主题驱动生成旨在合成定制图像，以模仿给定参考图像中主题在不同视觉上下文中的外观，这些上下文由文本提示指定。这里的主要挑战在于，当参考图像变化时，扩散模型无法有效适应不同的视觉上下文，同时始终保持主题身份。现有方法要么使用大型领域特定数据集训练模型，要么在实际图像生成之前，使用参考图像对模型进行数百次迭代的微调。在本工作中，我们探索了一种新的方法，称为循环模型适应（In-Loop Model Adaptation, IMA），该方法在实际图像生成过程中，在每个生成步骤中适应核心扩散模型，而无需在生成过程之前对参考图像进行训练。为此，我们建立了一个 DDIM 反演链，将参考图像映射到一系列潜在变量，以及一个仅从文本提示生成图像的文本到图像生成链。然后，我们引入了掩蔽潜在一致性损失和噪声正则化损失，以表征扩散模型与这两个链在每个生成步骤中的潜在噪声差异。这种耦合潜在噪声损失用于指导循环模型适应，以保持由参考图像指定的主题身份，同时与文本提示保持准确对齐，从而实现高保真的文本到图像生成。我们的大量实验表明，我们提出的 IMA 方法显著提高了主题驱动文本到图像生成的性能。
cs.CV / 189 / 2608.09264
Task-Adaptive 3D Cross-Field MRI Translation via Field-Conditioned Content-Style Pretraining
基于场条件内容风格预训练的任务自适应3D跨场MRI翻译
Abstract
Magnetic field strength is a major source of domain shift in magnetic resonance imaging (MRI), affecting signal-to-noise ratio, tissue contrast, spatial detail, and the visibility of anatomical boundaries. The MRIxFields 2026 challenge investigates this problem through cross-field MRI translation across acquisitions at 0.1T, 1.5T, 3T, 5T, and 7T. Its three tasks, Any-to-7T, 0.1T-to-High, and Any-to-Any synthesis, require the generation of target-field image characteristics while preserving subject-specific anatomy. This problem is particularly challenging because paired acquisitions of the same subject across multiple field strengths are rarely available for training. We propose a 3D unpaired cross-field MRI translation framework based on field-conditioned content-style pretraining. The proposed framework first learns controllable field-to-field translation across all available field strengths by disentangling anatomical content from field-dependent contrast characteristics. The pretrained backbone is then adapted to task-specific target domains. Our model comprises a 3D content encoder, a 3D style encoder, a field-conditioned style generator, an AdaIN-modulated decoder, and a multi-field discriminator. Adversarial learning encourages realistic target-field appearance, while cycle-consistency, identity, content, style, and diversity constraints promote anatomical fidelity and controllable translation. We evaluate the proposed method on MRIxFields data spanning five field strengths and three MRI modalities. Experiments on paired test data demonstrate that the framework can adapt to the three challenge settings while preserving three-dimensional anatomical structure in the synthesized volumes. The implementation code is publicly available at https://github.com/Idea89560041/3D-MRI-Field-Translation.
Chinese Translation
磁场强度是磁共振成像（MRI）中领域偏移的主要来源，影响信噪比、组织对比度、空间细节以及解剖边界的可见性。MRIxFields 2026挑战通过在0.1T、1.5T、3T、5T和7T的不同采集条件下进行跨场MRI翻译来研究这一问题。其三个任务，Any-to-7T、0.1T-to-High和Any-to-Any合成，要求在保留特定受试者解剖结构的同时生成目标场图像特征。这个问题特别具有挑战性，因为在多个场强下同一受试者的配对采集数据在训练中很少可用。我们提出了一种基于场条件内容风格预训练的3D无配对跨场MRI翻译框架。该框架首先通过将解剖内容与场依赖的对比特征解耦，学习在所有可用场强之间的可控场到场翻译。然后，将预训练的主干网络适应于特定任务的目标领域。我们的模型包括一个3D内容编码器、一个3D风格编码器、一个场条件风格生成器、一个AdaIN调制解码器和一个多场鉴别器。对抗学习鼓励生成真实的目标场外观，而循环一致性、身份、内容、风格和多样性约束则促进了解剖忠实性和可控翻译。我们在涵盖五种场强和三种MRI模态的MRIxFields数据上评估了所提出的方法。在配对测试数据上的实验表明，该框架能够适应三个挑战设置，同时在合成体积中保留三维解剖结构。实现代码已公开发布在 https://github.com/Idea89560041/3D-MRI-Field-Translation。
cs.CV / 190 / 2608.09266
Did the Grid Erase the Event? EndoClock for Auditing Medical World-Model Pipelines
网格是否抹去了事件？用于审计医学世界模型管道的 EndoClock
Abstract
Medical world models commonly learn from multimodal recordings synchronized onto a fixed-rate grid. This preprocessing resamples each native stream onto a shared time axis. Each stream has an observation clock that governs when observations are emitted or updated. When this clock depends on the latent or acquisition state, it is endogenous. In such settings, synchronization may not be neutral and can erase task-relevant evidence before the model sees the data. We introduce a four-regime taxonomy that characterizes where the evidence needed to distinguish a target event or state survives. The relevant witness may remain in the sampled values, in grid-cell update patterns, in native timing, or only in an external acquisition channel. EndoClock operationalizes this taxonomy as a conservative pretraining audit. It reports the lowest witness-bearing representation supported by the available evidence, or unresolved when no regime can be established. We illustrate this failure in echocardiography, where B-mode video write-outs cease during pulsed-wave Doppler acquisition while the corresponding measurement events remain recorded only in an external acquisition log. This work is a preliminary failure alert and executable audit. Its practical message is to preserve the native observation process long enough to determine whether synchronization has erased information required by the intended task.
Chinese Translation
医学世界模型通常从同步到固定速率网格的多模态记录中学习。该预处理将每个原生流重采样到共享时间轴上。每个流都有一个观察时钟，决定何时发出或更新观察。当这个时钟依赖于潜在状态或采集状态时，它是内生的。在这种情况下，同步可能并不中立，可能在模型看到数据之前抹去与任务相关的证据。我们引入了一个四个状态的分类法，描述了区分目标事件或状态所需的证据存活的位置。相关证据可能保留在采样值中、网格单元更新模式中、原生时序中，或仅在外部采集通道中。EndoClock 将这一分类法操作化为保守的预训练审计。它报告由可用证据支持的最低证据承载表示，或者在无法建立任何状态时标记为未解决。我们在超声心动图中说明了这一失败，其中 B 模式视频写出在脉冲波多普勒采集期间停止，而相应的测量事件仅记录在外部采集日志中。这项工作是一个初步的失败警报和可执行审计。其实际信息是要保留原生观察过程足够长的时间，以确定同步是否抹去了所需的任务信息。
cs.CV / 191 / 2608.09270
GRASP: Granularity-Aware Region Alignment and Semantic Prototype Learning for Fine-Grained Cross-Modal Understanding in Drone Views
GRASP：面向粒度的区域对齐与语义原型学习在无人机视角下的细粒度跨模态理解
Abstract
Fine-grained cross-modal understanding in drone views is essential for aerial vision-language navigation. However, the inherent wide field of view and overhead perspective of drone scenarios impose dual challenges on vision-language understanding. At the macro level, overwhelming background clutter in visual representations leads to Cross-Modal Focus Misalignment, where the model prioritizes global environmental similarities over specific object details. At the micro level, Visual Isomorphism creates ambiguity, where candidates share similar geometric structures yet differ only in subtle attributes. To address these challenges, we propose the Granularity-Aware Region Alignment and Semantic Prototype (GRASP) learning framework, enhancing discriminative capability through two synergistic strategies. Specifically, we introduce Region-Focused Alignment (RFA) to promote object-centric cross-modal alignment while suppressing background interference. Concurrently, to tackle visual isomorphism, we propose Semantic Perturbation Enhanced Matching (SPEM), which leverages a foreground-purified Semantic Prototype Codebook (SPC) to construct semantically perturbed negatives for fine-grained semantic discrimination. Extensive experiments on the GeoText-1652 benchmark and the unseen ERA dataset demonstrate that GRASP achieves competitive performance in drone-view fine-grained image-text retrieval, validating its effectiveness for cross-modal understanding in aerial scenarios. Our code implementation is available at https://github.com/UCAS-JC/GRASP.
Chinese Translation
无人机视角下的细粒度跨模态理解对于空中视觉-语言导航至关重要。然而，无人机场景固有的广阔视野和俯视角度对视觉-语言理解提出了双重挑战。在宏观层面，视觉表示中的背景杂乱无章导致跨模态聚焦失调（Cross-Modal Focus Misalignment），使得模型优先考虑全局环境相似性而忽视特定物体细节。在微观层面，视觉同构（Visual Isomorphism）造成了歧义，候选对象共享相似的几何结构，但仅在细微属性上有所不同。为了解决这些挑战，我们提出了面向粒度的区域对齐与语义原型（GRASP）学习框架，通过两种协同策略增强辨别能力。具体而言，我们引入了区域聚焦对齐（Region-Focused Alignment, RFA），以促进以物体为中心的跨模态对齐，同时抑制背景干扰。同时，为了解决视觉同构问题，我们提出了语义扰动增强匹配（Semantic Perturbation Enhanced Matching, SPEM），利用前景净化的语义原型代码本（Semantic Prototype Codebook, SPC）构建语义扰动的负样本，以实现细粒度的语义区分。在GeoText-1652基准和未见的ERA数据集上的大量实验表明，GRASP在无人机视角下的细粒度图像-文本检索中表现出竞争力，验证了其在空中场景中进行跨模态理解的有效性。我们的代码实现可在https://github.com/UCAS-JC/GRASP获取。
cs.CV / 192 / 2608.09287
UniDFKD: A Unified Semantic Prior Framework for Architecture-Agnostic Data-Free Knowledge Distillation
UniDFKD：一种统一的语义先验框架用于架构无关的数据无关知识蒸馏
Abstract
Data-Free Knowledge Distillation (DFKD) transfers knowledge from a pretrained teacher model to a compact student model by synthesizing semantically informative data, eliminating the need for access to the original training dataset. Existing DFKD methods rely heavily on architecture-specific statistical priors (e.g., Batch Normalization statistics) to guide data synthesis, however, such architecture-dependent priors are often absent in modern architectures such as Vision Transformers (ViTs), resulting in degraded semantic quality of the synthesized data and consequently catastrophic performance degradation. In this paper, we propose \emph{UniDFKD}, a unified data-free knowledge distillation framework that replaces architecture-specific statistics with explicit, architecture-agnostic semantic priors. \emph{UniDFKD} governs the entire synthesis-distillation pipeline along three dimensions: (1) Categorical Semantic Conditioning (CSC) defines \emph{what} to synthesize by persistently modulating the generator with language-derived embeddings to capture semantic diversity; (2) Spatial Semantic Anchoring (SSA) dictates \emph{where} evidence belongs by anchoring the teacher's spatial attributions to a Gaussian prior; and (3) Spatial Semantic Distillation (SSD) controls \emph{how} knowledge is transferred by explicitly aligning teacher-student spatial evidence alongside predictions. Extensive experiments across CNNs and ViTs demonstrate that UniDFKD establishes a new state-of-the-art, outperforming existing methods by an average absolute margin of over 20\% in both homogeneous and heterogeneous settings.
Chinese Translation
数据无关知识蒸馏（DFKD）通过合成语义信息丰富的数据，将知识从预训练的教师模型转移到紧凑的学生模型，从而消除了对原始训练数据集的访问需求。现有的DFKD方法在数据合成过程中严重依赖于特定于架构的统计先验（例如，批量归一化统计），然而，这种依赖于架构的先验在现代架构（如视觉变换器（ViTs））中往往缺失，导致合成数据的语义质量下降，从而造成灾难性的性能退化。本文提出了 extit{UniDFKD}，一种统一的数据无关知识蒸馏框架，使用显式的、架构无关的语义先验替代特定于架构的统计信息。 extit{UniDFKD}在三个维度上管理整个合成-蒸馏管道：（1）类别语义调节（CSC）通过持续调节生成器与源自语言的嵌入相结合，定义 extit{合成什么}以捕捉语义多样性；（2）空间语义锚定（SSA）通过将教师的空间归因锚定到高斯先验，决定 extit{证据属于哪里}；（3）空间语义蒸馏（SSD）通过显式对齐教师-学生的空间证据及其预测，控制 extit{知识如何转移}。在卷积神经网络（CNNs）和视觉变换器（ViTs）上的大量实验表明，UniDFKD建立了新的最先进水平，在同质和异质设置中，平均绝对提升超过20 ext%。
cs.CV / 193 / 2608.09302
Bootstrapping Vision-Language Model for Hysteroscopic Surgical Scene Segmentation
基于自举的视觉-语言模型用于宫腔镜手术场景分割
Abstract
Hysteroscopic surgical scene segmentation plays a pivotal role in understanding the hysteroscopic intraoperative environment as well as computer-assisted intervention. However, this task presents unique challenges due to the high morphological similarity among different lesions and the presence of artifacts such as specular reflections, motion blur, and fluid occlusions in surgical videos. In this work, we propose the first vision-language model (VLM)-based hysteroscopic surgical scene segmentation method, which performs pixel-wise localization for fifteen representative categories in hysteroscopic surgical scenes. Our VLM-hyster has a segmentation backbone that utilizes the pretrained image encoder for robust visual feature extraction, coupled with a transformer-based decoder for dense prediction. Moreover, we design category-specific text prompts and incorporate a masked distillation branch to filter out visual features with low correlation to the text prompts, enabling the model to focus more effectively on category-specific image regions and thereby enhancing segmentation performance. We collect a large multicentric hysteroscopic surgical scene dataset, containing 4,020 high-resolution images with detailed mask annotations, for model training and evaluation. Experimental results demonstrate that VLM-hyster substantially outperforms state-of-the-art AI models. Furthermore, extensive assessments by gynecologists, as well as multicentre and prospective validations, demonstrate VLM-hyster's robustness and generalizability. The results suggest that VLM-hyster earns considerable potential in enabling AI-assisted localization of surgical instruments and lesions in hysteroscopic surgeries. Code is available at https://github.com/viscom-tongji/VLM-hyster.
Chinese Translation
宫腔镜手术场景分割在理解宫腔镜术中环境以及计算机辅助干预中发挥着关键作用。然而，由于不同病变之间的高形态相似性以及手术视频中存在的伪影（如镜面反射、运动模糊和液体遮挡），这一任务面临独特的挑战。在本研究中，我们提出了首个基于视觉-语言模型（VLM）的宫腔镜手术场景分割方法，该方法对宫腔镜手术场景中的十五个代表性类别进行逐像素定位。我们的VLM-hyster具有一个分割主干，利用预训练的图像编码器进行稳健的视觉特征提取，并结合基于变换器的解码器进行密集预测。此外，我们设计了类别特定的文本提示，并结合了一个掩码蒸馏分支，以过滤与文本提示相关性低的视觉特征，使模型能够更有效地聚焦于类别特定的图像区域，从而提高分割性能。我们收集了一个大型多中心宫腔镜手术场景数据集，包含4,020张高分辨率图像及详细的掩膜注释，用于模型训练和评估。实验结果表明，VLM-hyster显著优于最先进的人工智能模型。此外，妇科医生的广泛评估以及多中心和前瞻性验证表明了VLM-hyster的稳健性和广泛适用性。结果表明，VLM-hyster在实现宫腔镜手术中对手术器械和病变的人工智能辅助定位方面具有相当大的潜力。代码可在 https://github.com/viscom-tongji/VLM-hyster 获取。
cs.CV / 194 / 2608.09311
Degraded Infrared Small Object Detection via Degradation-Adapted Physics-Guided Restoration
基于适应性退化的物理引导恢复的红外小目标检测
Abstract
Infrared small object detection has made significant progress in recent years. However, degradations such as fog and nonuniformity can suppress target-background contrast, substantially increasing detection difficulty. Existing methods mainly rely on image restoration as preprocessing, but they are typically designed for specific degradation types and fail to generalize to varying degradations. To alleviate this, we propose DAISOD, a degradation-adapted infrared small object detection framework for robust detection under different degradations. DAISOD first identifies the type and severity of degradations, then adapts the processing via dedicated branches, and finally fuses the results for subsequent detection. Moreover, a physics-guided restoration mechanism is incorporated to explicitly estimate degradation parameters and remove degradation effects through physical models, avoiding excessive restoration that may erase small targets. Moreover, we construct a degraded infrared small object detection dataset covering diverse degradation types and levels. Extensive experiments show that DAISOD outperforms state-of-the-art methods under various degradation conditions.
Chinese Translation
近年来，红外小目标检测取得了显著进展。然而，雾霾和非均匀性等退化现象会抑制目标与背景之间的对比度，显著增加检测难度。现有方法主要依赖图像恢复作为预处理，但通常是针对特定退化类型设计的，无法适应不同的退化情况。为了解决这一问题，我们提出了DAISOD（Degradation-Adapted Infrared Small Object Detection），一个适应性退化的红外小目标检测框架，旨在在不同退化条件下实现稳健检测。DAISOD首先识别退化的类型和严重程度，然后通过专门的分支进行处理，最后融合结果以进行后续检测。此外，我们还引入了一种物理引导恢复机制，以明确估计退化参数并通过物理模型去除退化影响，避免过度恢复可能导致的小目标消失。此外，我们构建了一个涵盖多种退化类型和水平的退化红外小目标检测数据集。大量实验表明，DAISOD在各种退化条件下的性能优于现有的最先进方法。
cs.CV / 195 / 2608.09316
MemeMind: Reference-Guided Trace Construction for Offline Context Optimization
MemeMind：基于参考的离线上下文优化轨迹构建
Abstract
Offline context optimization improves an agent by revising its instructions and examples while keeping the model frozen. This approach learns from rollouts on an adaptation set, but some queries produce only failed rollouts. In these cases, the optimizer sees no successful example of how the available tools can reach the correct answer. We introduce MemeMind, which uses an offline reference answer to recover this missing experience. TraceBuilder identifies the evidence required by the reference, executes text search, image retrieval, and visual grounding, and verifies the resulting tool trace before adding it to the adaptation buffer. ToolGuide then summarizes the collected traces into a shared guide and separate instructions for each tool. The reference answers and constructed traces are used only during adaptation, while inference uses the learned guides with a frozen model. We study this problem through Anime, Comic, and Game meme interpretation. These memes combine edited and ambiguous visual content, overlaid text, long tail franchise knowledge, and culture specific references. Their interpretation can require coordinated visual grounding, image retrieval, and text search, making them a demanding setting in which native rollout groups may fail together. We evaluate MemeMind on MemeX, a benchmark of 1,000 such memes annotated by experts. Across two Qwen3-VL models, two language partitions, and two independent judges, MemeMind improves over the strongest context optimization baseline by 22.0% and 21.1% on Qwen3-VL-30B-A3B, and by 8.1% and 8.0% on Qwen3-VL-235B-A22B under GPT-5 judging. Ablations and held out traces show that constructing successful tool use for failed groups provides the largest component gain and produces more effective evidence acquisition at inference time.
Chinese Translation
离线上下文优化通过修订代理的指令和示例来提升其性能，同时保持模型不变。该方法从适应集的回滚中学习，但某些查询仅产生失败的回滚。在这些情况下，优化器无法看到可用工具如何达到正确答案的成功示例。我们提出了MemeMind，它利用离线参考答案来恢复这一缺失的经验。TraceBuilder识别参考所需的证据，执行文本搜索、图像检索和视觉定位，并在将结果工具轨迹添加到适应缓冲区之前验证其有效性。ToolGuide随后将收集到的轨迹总结为共享指南和每个工具的单独指令。参考答案和构建的轨迹仅在适应过程中使用，而推理则使用冻结模型的学习指南。我们通过动漫、漫画和游戏的迷因解释研究这一问题。这些迷因结合了编辑过的模糊视觉内容、叠加文本、长尾特许知识和特定文化的引用。它们的解释可能需要协调的视觉定位、图像检索和文本搜索，使其成为一个原生回滚组可能共同失败的高要求环境。我们在MemeX上评估MemeMind，这是一个由专家注释的1,000个此类迷因的基准。在两个Qwen3-VL模型、两个语言分区和两个独立评审者的评估中，MemeMind在Qwen3-VL-30B-A3B上比最强的上下文优化基线提高了22.0%和21.1%，在Qwen3-VL-235B-A22B上在GPT-5评审下提高了8.1%和8.0%。消融实验和保留轨迹表明，为失败组构建成功的工具使用提供了最大的组件增益，并在推理时产生了更有效的证据获取。
cs.CV / 196 / 2608.09321
Warp-free Cross-view Geo-localization via Feature-space Consensus Mining
无扭曲跨视角地理定位通过特征空间共识挖掘
Abstract
Cross-view geo-localization is challenging due to drastic viewpoint changes and large appearance discrepancies between street-level and satellite imagery. Although existing methods often use geometric warping to expose co-visible cues, such transformations rely on restrictive spatial assumptions and inevitably introduce severe visual distortions under view-dependent visibility, yielding noisy supervision and fragile correspondences. To overcome this, we propose a novel joint-view consensus-guided learning framework that entirely bypasses explicit geometric warping. Instead of forcing rigid spatial alignment, we dynamically mine and adaptively strengthen a semantic consensus directly within the feature space. Specifically, an auxiliary joint-view pathway during training enables direct cross-view interaction, allowing each view to selectively aggregate corroborative evidence into a unified consensus representation. To resolve feature heterogeneity among the single- and joint-view streams, we introduce global pattern probes acting as a semantic dictionary to project divergent modalities into a strictly aligned metric space. Guided by a consensus-mediated contrastive objective, single-view embeddings are explicitly pulled toward the joint-view anchor during training, distilling this consensus-mining capability into the single-view encoders for robust retrieval at inference. Extensive experiments demonstrate that our method achieves state-of-the-art performance across four standard benchmarks, underscoring the importance of discovering cross-view semantic consensus for reliable geo-localization.
Chinese Translation
跨视角地理定位因视角变化剧烈和街景图像与卫星图像之间的外观差异而面临挑战。尽管现有方法通常使用几何扭曲来揭示共同可见的线索，但此类变换依赖于严格的空间假设，并不可避免地在视角依赖的可见性下引入严重的视觉扭曲，从而导致噪声监督和脆弱的对应关系。为了解决这一问题，我们提出了一种新颖的联合视角共识引导学习框架，完全绕过显式的几何扭曲。我们并不强迫刚性空间对齐，而是动态挖掘并自适应增强特征空间内的语义共识。具体而言，在训练过程中，辅助的联合视角路径使得跨视角交互成为可能，允许每个视角选择性地将相互支持的证据聚合成统一的共识表示。为了处理单视角和联合视角流之间的特征异质性，我们引入了作为语义字典的全局模式探针，以将不同的模态投影到严格对齐的度量空间中。在共识引导的对比目标的指导下，单视角嵌入在训练过程中被明确拉向联合视角锚点，将这种共识挖掘能力提炼到单视角编码器中，以便在推理时实现稳健的检索。大量实验表明，我们的方法在四个标准基准上实现了最先进的性能，强调了发现跨视角语义共识对于可靠地理定位的重要性。
cs.CV / 197 / 2608.09322
Diffusion Image Editing via Asynchronous Token Decoding
通过异步令牌解码进行扩散图像编辑
Abstract
Text-guided diffusion image editing aims to modify semantic attributes of an image while preserving its identity, layout, and background. However, na\"ively switching the text condition during sampling often causes global drift, as denoising dynamics propagate changes across tokens and can disrupt unedited regions. To address this issue, we propose \textbf{A}synchronous \textbf{T}oken \textbf{D}ecoding \textbf{Edit} (ATDEdit), an inference-time framework that views each sampler step as a parallel update of a globally coupled token matrix and enables token-indexed condition switching with differentiated update policies. Instead of applying synchronous target-conditioned updates to all tokens, ATDEdit estimates editable locations using token-wise conditional surprisal and applies target-conditioned corrections to the selected token set. It supplies source key/value memory at keep-token positions and projects selected keep-token latent rows back to their source values; these operations promote background preservation but do not constitute a pixel-level invariance guarantee. This approach combines local editing and background preservation without external or user-provided spatial masks and without model fine-tuning. On PIE-Bench, ATDEdit achieves the strongest reported preservation metrics, including 27.44~dB PSNR and 0.055 LPIPS, while retaining competitive semantic alignment.
Chinese Translation
文本引导的扩散图像编辑旨在修改图像的语义属性，同时保持其身份、布局和背景。然而，在采样过程中天真地切换文本条件往往会导致全局漂移，因为去噪动态会在令牌之间传播变化，并可能干扰未编辑区域。为了解决这个问题，我们提出了 extbf{A}sychronous extbf{T}oken extbf{D}ecoding extbf{Edit}（ATDEdit），这是一个推理时框架，将每个采样步骤视为全局耦合令牌矩阵的并行更新，并允许使用差异化更新策略进行令牌索引的条件切换。ATDEdit不是对所有令牌应用同步的目标条件更新，而是使用令牌级条件惊讶度来估计可编辑位置，并对选定的令牌集应用目标条件修正。它在保持令牌位置提供源键/值记忆，并将选定的保持令牌潜在行投影回其源值；这些操作促进了背景的保留，但并不构成像素级不变性的保证。这种方法结合了局部编辑和背景保留，而无需外部或用户提供的空间掩码，也无需模型微调。在PIE-Bench上，ATDEdit实现了最强的报告保留指标，包括27.44~dB PSNR和0.055 LPIPS，同时保持竞争性的语义对齐。
cs.CV / 198 / 2608.09342
Revisiting the Current Frame: Physical-Trace-Guided Network Output Correction for Video Restoration
重新审视当前帧：基于物理轨迹的网络输出修正用于视频修复
Abstract
Video restoration methods exploit temporal information to recover information missing from degraded observations. However, reference frames within the sequence may introduce inconsistent degradation, content discrepancy, or reconstruction errors due to physical image-formation variations, occlusion, and imperfect temporal aggregation. Existing approaches mainly focus on improving restoration networks, while the reliability of the generated outputs at different spatial locations remains largely unexplored. In this work, we propose ANCHOR, a model-agnostic framework that revisits the low-quality current frame as a temporally aligned anchor for video restoration correction. Specifically, ANCHOR estimates a spatial trust field from heterogeneous physical-trace evidence and adaptively balances the restoration proposal with the original observation. Experiments on High Dynamic Range video reconstruction and video deraining demonstrate consistent improvements across various state-of-the-art restoration models, validating the effectiveness of reliability-aware output correction for video restoration.
Chinese Translation
视频修复方法利用时间信息来恢复从降级观测中缺失的信息。然而，序列中的参考帧可能由于物理图像形成的变化、遮挡和不完美的时间聚合而引入不一致的降级、内容差异或重建错误。现有方法主要集中在改善修复网络，而在不同空间位置生成输出的可靠性仍然很大程度上未被探索。在本研究中，我们提出了ANCHOR，一个与模型无关的框架，重新审视低质量的当前帧作为视频修复修正的时间对齐锚点。具体而言，ANCHOR从异构物理轨迹证据中估计空间信任场，并自适应地平衡修复提议与原始观测之间的关系。在高动态范围视频重建和视频去雨的实验中，ANCHOR在各种最先进的修复模型中表现出一致的改进，验证了基于可靠性意识的输出修正在视频修复中的有效性。
cs.CV / 199 / 2608.09344
Beyond Global Editing: Per-Instance Disentangled Subspaces for Training-Free Hallucination Mitigation in LVLMs
超越全局编辑：用于无训练幻觉缓解的逐实例解耦子空间在大型视觉语言模型中的应用
Abstract
Recent advances in large vision-language models (LVLMs) have enabled powerful multimodal reasoning by integrating visual encoders with large language models (LLMs). However, their reliability is frequently undermined by hallucinations, where generated text inaccurately describes the visual input. Although fine-tuning can mitigate this problem, it is computationally expensive and requires large, curated datasets, making training-free alternatives attractive. Among these, model editing is more promising than decoding-based approaches: decoding methods adapt outputs per input but introduce computational overhead and instability, whereas model editing modifies internal representations offline, providing a more efficient and stable solution. However, existing model-editing techniques typically rely on a single global subspace to correct hallucinations, treating all test samples identically and failing to capture diverse hallucination modes across inputs. To address this limitation, we propose a training-free hallucination mitigation framework for dynamic, per-instance suppression at test time. Our method first constructs a set of Disentangled Hallucination Subspaces, each isolating a distinct hallucination mode. During inference, the model adaptively calculates weights reflecting each input's relationship to these subspaces, guiding a dynamically combined projection that selectively suppresses the most probable hallucination directions while preserving image-grounded semantics. Extensive experiments across multiple vision-language benchmarks and LVLM families demonstrate consistent improvements, highlighting the robustness, generalizability, and efficiency of our approach.
Chinese Translation
大型视觉语言模型（LVLMs）的最新进展通过将视觉编码器与大型语言模型（LLMs）结合，实现了强大的多模态推理。然而，它们的可靠性常常受到幻觉的影响，即生成的文本不准确地描述了视觉输入。尽管微调可以缓解这个问题，但其计算成本高昂，并且需要大量经过精心策划的数据集，使得无训练的替代方案变得更具吸引力。在这些方案中，模型编辑比基于解码的方法更具前景：解码方法根据每个输入调整输出，但引入了计算开销和不稳定性，而模型编辑则离线修改内部表示，提供了更高效和稳定的解决方案。然而，现有的模型编辑技术通常依赖于单一的全局子空间来纠正幻觉，将所有测试样本视为相同，未能捕捉到输入之间多样的幻觉模式。为了解决这一局限性，我们提出了一种无训练的幻觉缓解框架，用于在测试时进行动态的逐实例抑制。我们的方法首先构建了一组解耦幻觉子空间，每个子空间隔离一个独特的幻觉模式。在推理过程中，模型自适应地计算反映每个输入与这些子空间关系的权重，指导动态组合的投影，选择性地抑制最可能的幻觉方向，同时保留与图像相关的语义。我们在多个视觉语言基准和LVLM系列上的广泛实验表明了一致的改进，突显了我们方法的鲁棒性、可推广性和效率。
cs.CV / 200 / 2608.09345
One-Time Training for All Grains: Open-Set Grain Recognition and Quantitative Analysis
一次性训练适用于所有谷物：开放集谷物识别与定量分析
Abstract
Advances in crop breeding have introduced an increasing number of grain varieties, creating a growing demand for efficient variety recognition and quantitative analysis. However, existing methods are typically trained on a fixed variety set, and incorporating newly introduced varieties requires additional data collection and model retraining. To address this limitation, we propose GROW, a framework for Grain Recognition and quantitative analysis in Open sets Without retraining. GROW first performs class-agnostic grain localization, converting mixed-grain images into individual instances for variety-wise counting and phenotypic measurement. It then combines visual embeddings and morphological descriptors into fused grain descriptors stored in an extensible GrainBank. Query grains are recognized through rank-similarity weighted top-k retrieval, and newly introduced varieties are incorporated by appending their descriptors without updating the deployed models. Extensive experiments under progressive variety expansion, varying grain densities, and background domain shifts demonstrate the scalability, robustness, and adaptability of GROW. Compared with joint retraining, GROW reduced the average category-registration time from 4153 s to only 39 s while maintaining competitive recognition performance. These results demonstrate that GROW provides an efficient and maintainable solution for extensible grain recognition, counting, and phenotypic analysis without repeated model retraining.
Chinese Translation
作物育种的进展引入了越来越多的谷物品种，导致对高效品种识别和定量分析的需求不断增加。然而，现有方法通常是在固定的品种集上进行训练，纳入新引入的品种需要额外的数据收集和模型重训练。为了解决这一限制，我们提出了GROW，一个用于开放集谷物识别和定量分析的框架，无需重训练。GROW首先执行与类别无关的谷物定位，将混合谷物图像转换为单独的实例，以便进行品种计数和表型测量。然后，它将视觉嵌入和形态描述符结合成融合的谷物描述符，存储在可扩展的GrainBank中。查询谷物通过基于排名相似度的top-k检索进行识别，新引入的品种通过附加其描述符而不更新已部署的模型来纳入。针对逐步品种扩展、不同谷物密度和背景领域转移的广泛实验表明了GROW的可扩展性、鲁棒性和适应性。与联合重训练相比，GROW将平均类别注册时间从4153秒减少到仅39秒，同时保持了竞争性的识别性能。这些结果表明，GROW为可扩展的谷物识别、计数和表型分析提供了一种高效且可维护的解决方案，无需重复的模型重训练。
cs.CV / 201 / 2608.09355
Alpha as an Efficiency Signal: Visibility-Routed RGBA Image-to-Video Generation
作为效率信号的Alpha：可见性引导的RGBA图像到视频生成
Abstract
RGBA videos combine RGB appearance with an alpha channel, enabling animated assets to be applied across arbitrary backgrounds, which are heavily used in gaming industry. However, generating high-quality RGBA animations for games remains challenging for two reasons. First, most existing RGBA video datasets are dominated by photorealistic content, with limited coverage of game assets. Second, the traditional generate-then-matte pipelines estimate alpha only after RGB synthesis, so semi-transparent regions are often blurred by background, resulting in unstable matting outputs. More recently, many methods have begun to model RGB and alpha jointly, but existing approaches are mostly text-conditioned, and still have unresolved issues in efficiency and quality. To address these challenges, we introduce GameAlpha-2.4K, a 2.4K-clip game-style RGBA video dataset built with matte-friendly synthesis, multi-hypothesis alpha recovery, and compositing-based quality gates. Using this dataset, we train a reference-conditioned RGBA video generator that jointly produces RGB frames and alpha mattes in a single pass. To improve efficiency, we propose a visibility router that identifies transparent tokens in an early stage and bypasses their later DiT updates, while x_0-lock guides them along the original flow-matching schedule toward self-predicted endpoints. Our model obtains lower FVD than traditional two-stage pipelines, and the visibility router skips 35% of token evaluations in the final two DiT denoising steps, providing a 1.2x backbone speedup with negligible quality degradation compared to dense inference.
Chinese Translation
RGBA视频结合了RGB外观和alpha通道，使得动画资产能够应用于任意背景，这在游戏行业中被广泛使用。然而，为游戏生成高质量的RGBA动画仍然面临两个挑战。首先，大多数现有的RGBA视频数据集主要以照片级真实内容为主，游戏资产的覆盖范围有限。其次，传统的生成-然后-抠图流程仅在RGB合成后估计alpha，因此半透明区域常常被背景模糊，导致不稳定的抠图输出。最近，许多方法开始联合建模RGB和alpha，但现有的方法大多依赖文本条件，并且在效率和质量上仍存在未解决的问题。为了解决这些挑战，我们引入了GameAlpha-2.4K，这是一个基于友好抠图合成、多假设alpha恢复和基于合成的质量门控构建的2.4K游戏风格RGBA视频数据集。利用该数据集，我们训练了一个参考条件的RGBA视频生成器，该生成器在单次传递中联合生成RGB帧和alpha抠图。为了提高效率，我们提出了一种可见性路由器，它在早期阶段识别透明标记，并绕过它们后续的DiT更新，同时x_0-lock引导它们沿着原始流匹配计划朝自我预测的端点前进。我们的模型在FVD上优于传统的两阶段流程，并且可见性路由器在最后两个DiT去噪步骤中跳过了35%的标记评估，相比于密集推理提供了1.2倍的骨干加速，且质量降级微乎其微。
cs.CV / 202 / 2608.09357
ControlRadio: Prompt-Driven Controllable Diffusion for Cross-Modal Radio Map Generation
ControlRadio：基于提示驱动的可控扩散用于跨模态无线电地图生成
Abstract
Radio maps describe how wireless signals propagate across space and are essential for wireless communication, sensing, and network planning. However, constructing accurate radio maps traditionally requires either dense measurements or computationally expensive physical simulations, which limits scalability and real-time deployment. Recent advances in generative artificial intelligence offer a promising alternative, but existing approaches lack fine-grained control and physical consistency when applied to real-world wireless environments. Here we present \textbf{ControlRadio}, a controllable generative framework that produces radio maps from natural-language descriptions and environmental layouts, including building structures and transmitter locations. Joint semantic and spatial conditioning enables interpretable, propagation-plausible generation, while a controlled latent prior and layout-aware conditioning improve stability and structural consistency. Extensive experiments demonstrate that ControlRadio achieves state-of-the-art accuracy and strong generalization across diverse urban scenarios, while reducing computation time by more than four orders of magnitude compared with conventional simulation-based methods. Such results suggest a new paradigm for scalable and controllable wireless environment modeling, with broad implications for next-generation communication systems and data-driven radio sensing.
Chinese Translation
无线电地图描述了无线信号在空间中的传播方式，对于无线通信、感知和网络规划至关重要。然而，传统上构建准确的无线电地图需要密集的测量或计算成本高昂的物理模拟，这限制了其可扩展性和实时部署。最近生成性人工智能的进展提供了一种有前景的替代方案，但现有方法在应用于真实无线环境时缺乏细粒度的控制和物理一致性。在此，我们提出了 extbf{ControlRadio}，一个可控的生成框架，能够根据自然语言描述和环境布局（包括建筑结构和发射器位置）生成无线电地图。联合语义和空间条件使得生成结果可解释且符合传播逻辑，而受控的潜在先验和布局感知条件则提高了稳定性和结构一致性。大量实验表明，ControlRadio在多样化的城市场景中实现了最先进的准确性和强大的泛化能力，同时与传统的基于模拟的方法相比，计算时间减少了四个数量级以上。这些结果表明了一种可扩展和可控的无线环境建模的新范式，对下一代通信系统和数据驱动的无线电感知具有广泛的影响。
cs.CV / 203 / 2608.09360
Deep Learning based Detection of Fishing Vessels and Fishing Monitoring using Nightlight Images
基于深度学习的渔船检测及夜间灯光图像渔业监测
Abstract
The demand for maritime surveillance has given rise to the need for monitoring fishing vessel activities, particularly in addressing the challenge of "dark vessels" that operate without Automatic Identification System (AIS) transmission. This study presents a novel approach for detecting small-scale fishing vessels using nighttime light (NTL) imagery from the SDGSAT-1 satellite, combined with deep learning techniques to enhance fishing monitoring awareness along the western coast of India. A dual-branch YOLO11 architecture was developed to exploit both the 10-meter panchromatic and 40-meter RGB imagery from SDGSAT-1. The custom model architecture was specifically optimized for small object detection in NTL imagery, featuring parallel convolutional backbones that process both modalities before concatenation for enhanced feature extraction. The dual-branch YOLO11 model demonstrated optimal performance with a precision of 0.99, recall of 0.93, F1-score of 0.96, and mAP@50 of 0.96, significantly outperforming single-branch implementations of YOLOv5s, YOLOv8s, and standard YOLO11s architectures. When applied to the western coast of India, the model detected 31525 vessel instances across the temporal dataset spanning 2022-23. Cross-matching analysis with AIS data revealed that only 7146 (22.7%) of detected vessels had corresponding AIS transmissions, while 24379 (77.3%) were identified as potential dark vessels. Spatio-temporal analysis showed peak fishing activity during January-April, with a primary activity corridor parallel to the coastline within 50-100 km, corresponding to productive continental shelf areas. This research contributes to maritime surveillance capabilities by highlighting the effectiveness of nighttime lights satellite imagery for fishing vessel detection and provides valuable insights into fishing patterns and potential regulatory compliance issues in Indian waters.
Chinese Translation
海洋监视的需求催生了对渔船活动监测的需要，特别是在应对没有自动识别系统（AIS）传输的“黑暗船只”挑战时。本研究提出了一种新颖的方法，利用来自SDGSAT-1卫星的夜间灯光（NTL）影像，结合深度学习技术，以增强对印度西海岸渔业监测的认知。我们开发了一种双分支YOLO11架构，以利用SDGSAT-1的10米全色和40米RGB影像。该自定义模型架构专门针对NTL影像中的小物体检测进行了优化，采用并行卷积主干处理两种模态，然后进行拼接以增强特征提取。双分支YOLO11模型表现出最佳性能，精确度为0.99，召回率为0.93，F1-score为0.96，mAP@50为0.96，显著优于YOLOv5s、YOLOv8s和标准YOLO11架构的单分支实现。在应用于印度西海岸时，该模型在2022-23年的时间数据集中检测到31525个船只实例。与AIS数据的交叉匹配分析显示，只有7146（22.7%）个检测到的船只有相应的AIS传输，而24379（77.3%）被识别为潜在的黑暗船只。时空分析显示，1月至4月期间渔业活动达到峰值，主要活动走廊平行于海岸线，位于50-100公里范围内，对应于富饶的大陆架区域。本研究通过强调夜间灯光卫星影像在渔船检测中的有效性，增强了海洋监视能力，并为印度水域的渔业模式和潜在的监管合规问题提供了宝贵的见解。
cs.CV / 204 / 2608.09369
FeedbackTrack: Visual-Cortex-Inspired Cross-Frame Feedback for Transformer Tracking
FeedbackTrack：一种受视觉皮层启发的跨帧反馈机制用于变换器跟踪
Abstract
Visual object tracking requires effective temporal integration, yet most Transformer trackers still rely on predominantly feed-forward feature extraction. Existing temporal mechanisms typically update templates, prompts, queries, or prediction states, while intermediate representations are rarely reused to modulate corresponding processing stages. We propose \textbf{FeedbackTrack}, a visual-cortex-inspired framework that introduces sparse, group-level layer-aligned cross-frame feedback into pretrained Transformer trackers. Previous-frame intermediate states are detached, cached, and returned to corresponding Transformer groups in the current frame through two lightweight pathways: Query Feedback for token-level query modulation and Gate Feedback for context-dependent feature modulation. FeedbackTrack preserves the original tracking pipeline with only a fixed-size one-frame cache. Across SPMTrack and ARTrackV2, FeedbackTrack consistently improves five backbone configurations on LaSOT and GOT-10k, achieving 83.4 AO and 79.1 AUC with SPMTrack-G while adding less than 1\% parameters. Controlled comparisons show that cross-frame feedback outperforms same-frame modulation by 1.8--3.2 AO points, demonstrating that the gains mainly come from recurrent historical information. Further analysis reveals a non-uniform depth-dependent organization of learned feedback strengths, highlighting the effectiveness of recurrent feedback for Transformer tracking.
Chinese Translation
视觉目标跟踪需要有效的时间整合，然而大多数变换器跟踪器仍主要依赖前馈特征提取。现有的时间机制通常更新模板、提示、查询或预测状态，而中间表示很少被重用以调节相应的处理阶段。我们提出了 extbf{FeedbackTrack}，一个受视觉皮层启发的框架，它将稀疏的、分组级别的层对齐跨帧反馈引入预训练的变换器跟踪器。前一帧的中间状态被分离、缓存，并通过两条轻量级路径返回到当前帧的相应变换器组：查询反馈（Query Feedback）用于令牌级查询调节，门控反馈（Gate Feedback）用于上下文依赖的特征调节。FeedbackTrack在仅使用固定大小的一帧缓存的情况下保留了原始跟踪流程。在SPMTrack和ARTrackV2上，FeedbackTrack在LaSOT和GOT-10k上持续改善了五种主干配置，在使用SPMTrack-G时达到了83.4 AO和79.1 AUC，同时增加的参数少于1%。对照比较显示，跨帧反馈比同帧调节提高了1.8至3.2 AO点，表明收益主要来自于递归的历史信息。进一步分析揭示了学习反馈强度的非均匀深度依赖组织，突显了递归反馈在变换器跟踪中的有效性。
cs.CV / 205 / 2608.09373
Preserve More Details: Mitigating Content Drift in Real-World Image Super-Resolution
保留更多细节：缓解真实世界图像超分辨率中的内容漂移
Abstract
Real-world image super-resolution (Real-ISR) aims to reconstruct high-quality (HQ) images from low-quality (LQ) inputs subject to diverse real-world degradations. Recent advances have leveraged the LQ inputs and natural image priors learned by Stable Diffusion models to achieve impressive results. However, existing methods often overlook insufficient clarity of LQ inputs inevitably induce content drift in the generated HQ images. This manifests primarily as visual detail degradation and textual semantic shift, severely compromising both fidelity and perceptual quality. To address this challenge, we propose FSP-Diff, a novel one-step diffusion model featuring a dual-pathway architecture. This architecture comprises a Detail-Conditioned Pathway for injecting structured details to recover fine structures, and a Detail-Modulated Semantic Pathway that refines semantic guidance using structured details to mitigate semantic deviations. Extensive experiments on standard Real-ISR benchmarks demonstrate that FSP-Diff surpasses existing one-step diffusion methods in both quantitative and qualitative metrics.
Chinese Translation
真实世界图像超分辨率（Real-ISR）旨在从低质量（LQ）输入中重建高质量（HQ）图像，面对多种真实世界的退化。最近的进展利用了LQ输入和通过稳定扩散模型（Stable Diffusion）学习的自然图像先验，取得了令人瞩目的成果。然而，现有方法往往忽视了LQ输入的清晰度不足不可避免地导致生成的HQ图像中出现内容漂移。这主要表现为视觉细节退化和文本语义偏移，严重影响了图像的保真度和感知质量。为了解决这一挑战，我们提出了FSP-Diff，这是一种新颖的一步扩散模型，具有双通道架构。该架构包括一个细节条件通道（Detail-Conditioned Pathway），用于注入结构化细节以恢复细微结构，以及一个细节调制语义通道（Detail-Modulated Semantic Pathway），利用结构化细节来细化语义指导，以减轻语义偏差。在标准Real-ISR基准上的大量实验表明，FSP-Diff在定量和定性指标上均超越了现有的一步扩散方法。
cs.CV / 206 / 2608.09388
Efficient Human-Contact Representation for Human-Scene Interaction
高效的人体接触表示用于人类与场景的交互
Abstract
Human-scene interaction is an active research topic with several industrial applications in virtual reality, gaming, robotics, and surveillance. Despite significant progress in network architectures to improve the results or optimize models' parameters for fast inference speed, the efficient representation of contact between humans and their environments remains an open challenge. In this paper, we propose a new efficient human-contact representation for human-scene interaction. Our primary contribution is the introduction of sparse contact masks that strategically select essential contact information, significantly reducing redundant data in high-dimensional inputs. Leveraging this efficient contact representation, we propose a suite of sparse operators to replace traditional dense operators within deep network layers for faster computation. Our approach not only enhances computational speed but also filters out non-essential contact data, thereby improving the precision of human-scene interaction models. To validate the effectiveness of our method, we conduct intensive experiments across three public benchmark datasets, focusing on two critical tasks for human-scene interaction: contact prediction and scene synthesis. The experimental results show that our approach outperforms state-of-the-art models in reconstruction accuracy and achieves a computation speed-up of at least 12 times over recent baselines.
Chinese Translation
人类与场景的交互是一个活跃的研究主题，在虚拟现实、游戏、机器人技术和监控等多个工业应用中具有重要意义。尽管在网络架构方面取得了显著进展，以改善结果或优化模型参数以实现快速推理速度，但人类与其环境之间接触的高效表示仍然是一个未解决的挑战。本文提出了一种新的高效人体接触表示，用于人类与场景的交互。我们的主要贡献是引入稀疏接触掩码，战略性地选择必要的接触信息，显著减少高维输入中的冗余数据。利用这种高效的接触表示，我们提出了一套稀疏算子，以替代深度网络层中的传统密集算子，从而加快计算速度。我们的方法不仅提高了计算速度，还过滤掉非必要的接触数据，从而提高了人类与场景交互模型的精度。为了验证我们方法的有效性，我们在三个公共基准数据集上进行了大量实验，重点关注人类与场景交互的两个关键任务：接触预测和场景合成。实验结果表明，我们的方法在重建精度上超过了最先进的模型，并在最近的基准上实现了至少12倍的计算速度提升。
cs.CV / 207 / 2608.09391
CoInS-Net: A Continuous Position-Aware Network for Joint Medical Image Interpolation and Segmentation
CoInS-Net：一种用于联合医学图像插值和分割的连续位置感知网络
Abstract
Accurate medical image interpolation and anatomical structure segmentation are fundamental for computer-aided diagnosis and treatment planning. Anisotropic medical volumes with sparse through-plane sampling often suffer from structural discontinuity and boundary blur, hindering reliable clinical image analysis. Most existing methods implement interpolation and segmentation independently, which introduces redundant computation and fails to fully exploit complementary cross-slice structural information between sequential slices. To address these issues, we propose a continuous position-aware interaction network, termed CoInS-Net, for joint frame interpolation and lesion segmentation. Unlike conventional cascaded interpolation-then-segmentation paradigms, the framework enables bidirectional interaction under a shared Swin encoder with continuous spatial coordinate queries. A spatially continuous position interpolation module generates target-position features at every scale from the relative coordinate and physical spacing, and a prototype-based task mutual interaction module lets the segmentation and interpolation branches exchange global structure through a small set of shared prototypes rather than dense feature mixing. A multi-scale task-cooperative decoder further separates each scale into shared and task-specific components, so the two tasks reinforce common anatomy while preserving their distinct requirements down to the boundary level, without extra annotations. Experiments on four public medical imaging datasets with diverse modalities and anatomical regions demonstrate that the proposed method outperforms conventional single-task schemes. The joint optimization framework effectively realizes mutual promotion between interpolation and segmentation tasks, providing a reliable and universal technical scheme for intelligent clinical medical image analysis.
Chinese Translation
准确的医学图像插值和解剖结构分割对于计算机辅助诊断和治疗规划至关重要。具有稀疏平面采样的各向异性医学体积常常面临结构不连续和边界模糊的问题，阻碍了可靠的临床图像分析。现有的大多数方法独立实施插值和分割，这引入了冗余计算，并未充分利用相邻切片之间互补的跨切片结构信息。为了解决这些问题，我们提出了一种称为CoInS-Net的连续位置感知交互网络，用于联合帧插值和病灶分割。与传统的插值后再分割的级联范式不同，该框架在共享的Swin编码器下实现了双向交互，并使用连续的空间坐标查询。空间连续位置插值模块根据相对坐标和物理间距在每个尺度生成目标位置特征，而基于原型的任务互交互模块使得分割和插值分支通过一小组共享原型而非密集特征混合来交换全局结构。多尺度任务协作解码器进一步将每个尺度分离为共享和任务特定的组件，从而使两个任务在边界级别上加强共同解剖结构，同时保留其独特需求，而无需额外的标注。在四个具有不同模态和解剖区域的公共医学影像数据集上的实验表明，所提出的方法优于传统的单任务方案。联合优化框架有效实现了插值和分割任务之间的相互促进，为智能临床医学图像分析提供了可靠且通用的技术方案。
cs.CV / 208 / 2608.09392
CableDex: Cable Length Estimation on Industrial Reels Using a Handheld Device
CableDex：使用手持设备对工业卷轴上的电缆长度进行估计
Abstract
CableDex is a computer vision system that addresses the time-consuming and inaccurate manual measurement of cable length on industrial reels from a single photograph captured with a mobile phone. The system combines camera calibration, instance segmentation, pose estimation, and volumetric calculation to estimate the cable length across five different reel types and various cable sizes. This system is based on an instance segmentation model trained on 1,000 manually annotated images, achieving 99.5\% mAP50 with an inference time of 5.66 ms per image. Evaluated on 75 reels across five reel types, the system achieves a MAPE of 4.90\%, within the 10\% error tolerance commonly accepted in industrial cable-reel measurement. The demonstration presents the end-to-end pipeline, from reel label scanning and image capture to segmentation and length estimation, through the mobile application.
Chinese Translation
CableDex 是一个计算机视觉系统，旨在解决通过手机拍摄的单张照片对工业卷轴上的电缆长度进行耗时且不准确的手动测量问题。该系统结合了相机校准、实例分割、姿态估计和体积计算，以估计五种不同卷轴类型和各种电缆尺寸的电缆长度。该系统基于在 1,000 张手动标注图像上训练的实例分割模型，达到了 99.5\% 的 mAP50，推理时间为每张图像 5.66 毫秒。在对 75 个卷轴进行评估时，该系统的 MAPE 为 4.90\%，在工业电缆卷轴测量中通常接受的 10\\% 误差容忍范围内。演示展示了从卷轴标签扫描和图像捕获到分割和长度估计的端到端流程，通过移动应用程序实现。
cs.CV / 209 / 2608.09400
Sign Language Recognition Using Original and Synthetic Depth Image Based Point Cloud Data Models
基于原始和合成深度图像的点云数据模型的手语识别
Abstract
Research regarding the sign language recognition mostly relies on RGB images, whileas sign language datasets that provide depth images are limited. Point clouds obtained from depth images can be used for sign language recognition with neural networks like PointNet. In recent years, various neural networks are used for generating realistic depth images from monocular RGB images. In this work, synthetic depth images were created from RGB images using Depth Anything V2 network. For this purpose, three sign language datasets (Real-time ASL Fingerspelling, KArSL, AUTSL) which contain both RGB and depth images were used. Classification accuracies of the point cloud data created from both original and synthetic depth images using various PointNet architectures were measured for sign language recognition. From the original and synthetic point clouds, frame based, Point Gesture Map and Long Short Term Memory data models were used for classification and their performances were compared. In the results, both original and synthetic based data achieved acceptable performance in most models. In general, original depth based point cloud models performed better than synthetic ones, however in some models synthetic depth based models performed better than the originals.
Chinese Translation
手语识别的研究主要依赖于RGB图像，而提供深度图像的手语数据集相对有限。通过深度图像获得的点云可以与像PointNet这样的神经网络一起用于手语识别。近年来，各种神经网络被用于从单目RGB图像生成逼真的深度图像。在本研究中，使用Depth Anything V2网络从RGB图像创建了合成深度图像。为此，使用了三个包含RGB和深度图像的手语数据集（实时ASL手指拼写、KArSL、AUTSL）。使用不同的PointNet架构测量了从原始和合成深度图像创建的点云数据的分类准确率，以进行手语识别。从原始和合成点云中，使用了基于帧的、点手势图和长短期记忆（Long Short Term Memory）数据模型进行分类，并比较了它们的性能。结果显示，原始和合成数据在大多数模型中都达到了可接受的性能。总体而言，基于原始深度的点云模型的表现优于合成模型，但在某些模型中，基于合成深度的模型的表现优于原始模型。
cs.CV / 210 / 2608.09403
One Model to Magnify Them All: Efficient Scale-Invariant Histopathology via Conditional Normalization and Continuous Magnification Training
一个模型统治所有：通过条件归一化和连续放大训练实现高效的尺度不变组织病理学
Abstract
Whole slide images (WSIs) in digital histopathology are acquired at discrete magnification levels encoding complementary diagnostic information from global tissue architecture to fine-grained cellular morphology. Yet, deep learning models remain sensitive to scale variation. Existing magnification-invariant methods rely on multi-scale architectures at predefined discrete resolutions, while in clinical deployment the acquisition magnification varies continuously, rarely aligns with a model's fixed training resolution, and intermediate scales are common, so robust coverage otherwise demands a costly ensemble of magnification-specific models. We propose Conditional Layer Normalization (CLN), a lightweight mechanism that generates affine normalization parameters from input pixel size via a small MLP, integrated into standard CNN architectures for both WSI classification and segmentation. Trained on patches sampled continuously across a range of pixel sizes, the model decouples inference from scanner-dependent magnification and generalizes to arbitrary, previously unseen scales at test time. On the PANDA prostate cancer dataset, our approach on average matches or exceeds independently trained single-magnification models and ranks among the top three performers at every evaluated magnification, including those unseen during training. This collapses a five-model ensemble into a single network and reduces training, and inference cost roughly 4-5 times, while leaving the multiply-accumulate count unchanged. The code is available at: https://github.com/aflorkowska/OneModelToMagnifyThemAll.
Chinese Translation
数字组织病理学中的全切片图像（WSIs）是在离散放大级别下获取的，编码了从全局组织结构到细微细胞形态的互补诊断信息。然而，深度学习模型对尺度变化仍然敏感。现有的尺度不变方法依赖于在预定义的离散分辨率下的多尺度架构，而在临床应用中，获取的放大倍数是连续变化的，通常与模型的固定训练分辨率不对齐，并且中间尺度是常见的，因此，稳健的覆盖通常需要昂贵的特定放大模型的集成。我们提出了条件层归一化（Conditional Layer Normalization, CLN），这是一种轻量级机制，通过小型多层感知器（MLP）从输入像素大小生成仿射归一化参数，并集成到标准卷积神经网络（CNN）架构中，用于WSI分类和分割。该模型在跨越一系列像素大小的补丁上进行连续采样训练，从而将推理与扫描仪依赖的放大解耦，并在测试时对任意未见过的尺度进行泛化。在PANDA前列腺癌数据集上，我们的方法在平均上与独立训练的单一放大模型相匹配或超越，并在每个评估的放大倍数中排名前三，包括那些在训练期间未见过的放大倍数。这将五个模型的集成压缩为一个网络，并将训练和推理成本降低约4-5倍，同时保持乘加计数不变。代码可在以下链接获取：https://github.com/aflorkowska/OneModelToMagnifyThemAll。
cs.CV / 211 / 2608.09405
MeanSR: Restoration Trajectory Learning for One-Step Perceptual Super-Resolution
MeanSR：一步感知超分辨率的恢复轨迹学习
Abstract
Diffusion-based super-resolution (SR) achieves strong perceptual quality but requires costly iterative denoising. Existing one-step distillation methods reduce inference time but depend on expensive pretrained teachers, whereas CTMSR avoids distillation through PF-ODE consistency training yet does not explicitly model the restoration dynamics from low-resolution (LR) inputs to high-resolution (HR) images. We propose MeanSR, a one-step perceptual SR method that learns an LR-conditioned average velocity field to directly capture the finite-time transition from degraded or noisy inputs to plausible HR outputs. We further reformulate distribution trajectory matching for average-velocity generation and introduce a Stage-Aware Temporal Sampling strategy to improve trajectory learning. Experiments on synthetic and real-world benchmarks show that MeanSR outperforms CTMSR on CLIPIQA, MUSIQ, and MANIQA while substantially reducing FLOPs and inference latency. MeanSR also reconstructs sharper structures and more realistic textures with fewer perceptual artifacts.
Chinese Translation
基于扩散的超分辨率（SR）在感知质量上表现出色，但需要耗时的迭代去噪。现有的一步蒸馏方法虽然减少了推理时间，但依赖于昂贵的预训练教师，而CTMSR通过PF-ODE一致性训练避免了蒸馏，但并未明确建模从低分辨率（LR）输入到高分辨率（HR）图像的恢复动态。我们提出了MeanSR，这是一种一步感知SR方法，学习LR条件下的平均速度场，以直接捕捉从降级或噪声输入到合理HR输出的有限时间过渡。我们进一步重新构造了用于平均速度生成的分布轨迹匹配，并引入了一种阶段感知时间采样策略以改善轨迹学习。在合成和真实世界基准上的实验表明，MeanSR在CLIPIQA、MUSIQ和MANIQA上优于CTMSR，同时显著减少了FLOPs和推理延迟。MeanSR还以更少的感知伪影重建出更清晰的结构和更逼真的纹理。
cs.CV / 212 / 2608.09427
Foundation Models are Implicit Deepfake Detectors
基础模型是隐式的深度伪造检测器
Abstract
Pretrained self-supervised representations have emerged as a core component of current deepfake detection methods, yet it remains unclear which of their properties make real and fake media distinguishable. In this work, we uncover a surprisingly consistent phenomenon: across multiple pretrained models, datasets, and both image and video domains, fake samples systematically produce lower-magnitude representations than their real counterparts. Motivated by this finding, we formulate deepfake detection as an anomaly detection problem and show that simple statistics of feature magnitude achieve competitive performance with far more sophisticated deepfake detection methods. We further investigate the origin of this effect and demonstrate that reduced feature magnitude is primarily associated with semantic shifts introduced by fake content, while low-level generative fingerprints play a comparatively smaller role. Finally, we show that this discriminative signal strengthens as the size of the underlying foundation model grows, suggesting that advances in representation learning naturally translate into stronger zero-shot deepfake detectors.
Chinese Translation
预训练的自监督表示已成为当前深度伪造检测方法的核心组成部分，但尚不清楚其哪些特性使得真实和伪造媒体可区分。在本研究中，我们揭示了一个令人惊讶的一致现象：在多个预训练模型、数据集以及图像和视频领域中，伪造样本系统性地产生比其真实对应物更低幅度的表示。基于这一发现，我们将深度伪造检测形式化为异常检测问题，并展示特征幅度的简单统计量在性能上与更复杂的深度伪造检测方法相当。我们进一步探讨了这一效应的来源，并证明降低的特征幅度主要与伪造内容引入的语义偏移相关，而低级生成指纹的作用相对较小。最后，我们表明，随着基础模型规模的增加，这一区分信号会增强，这表明表示学习的进展自然转化为更强的零样本深度伪造检测器。
cs.CV / 213 / 2608.09438
Unveiling the Secret of AdaLN-Zero in Diffusion Transformer
揭示扩散变换器中 AdaLN-Zero 的秘密
Abstract
Diffusion transformer (DiT), a rapidly emerging architecture for image generation, has gained much attention. However, despite ongoing efforts to improve its performance, the understanding of DiT remains superficial. In this work, we delve into and investigate a critical conditioning mechanism within DiT, adaLN-Zero, which achieves superior performance compared to adaLN. Our work studies three potential elements driving this performance, including an SE-like structure, zero-initialization, and a "gradual" update order, among which zero-initialization is proved to be the most influential. Building on this understanding, we propose an analysis-guided initialization strategy, termed adaLN-Gaussian, which serves both as an empirical validation of our analysis and as a practical initialization method that consistently improves optimization efficiency. On the other hand, inspired by the SE-like structure, we introduce an improved conditioning mechanism called SE-adaLN-Zero. Extensive experiments following DiT on four datasets, especially on ImageNet1K demonstrate the effectiveness and generalization of adaLN-Gaussian and SE-adaLN-Zero. Beyond class-to-image generation, we also evaluate the generalization of the two improved methods on text-to-image generation.
Chinese Translation
扩散变换器（DiT）是一种快速崛起的图像生成架构，受到了广泛关注。然而，尽管在提升其性能方面进行了持续努力，对 DiT 的理解仍然较为肤浅。在本研究中，我们深入探讨并研究了 DiT 中一个关键的条件机制——adaLN-Zero，该机制相比于 adaLN 实现了更优的性能。我们的研究分析了推动这一性能的三个潜在因素，包括类似 SE 的结构、零初始化以及“渐进”更新顺序，其中零初始化被证明是最具影响力的。在此基础上，我们提出了一种基于分析的初始化策略，称为 adaLN-Gaussian，该策略既作为我们分析的经验验证，又作为一种实用的初始化方法，能够持续提高优化效率。另一方面，受到类似 SE 结构的启发，我们引入了一种改进的条件机制，称为 SE-adaLN-Zero。在四个数据集上，特别是在 ImageNet1K 上进行的广泛实验表明，adaLN-Gaussian 和 SE-adaLN-Zero 的有效性和泛化能力。除了类到图像生成外，我们还评估了这两种改进方法在文本到图像生成中的泛化能力。
cs.CV / 214 / 2608.09449
Sekai2: From World Exploration to Interactive World Modeling
Sekai2：从世界探索到互动世界建模
Abstract
Video world models must capture how scenes evolve over time and across viewpoints. Training them for long-horizon generation and camera control therefore benefits from long videos paired with camera trajectories and temporally grounded semantics. Existing corpora rarely offer the three together: large-scale web video provides broad visual diversity but no trajectories or time-aligned text, while pose-annotated datasets are typically short-range or reconstruction-oriented. We introduce Sekai2, a multi-source real-world video dataset that carries the world-exploration footage of Sekai toward interactive world modeling. The release contains 128,892 clips totaling 2,826 hours from 10,428 source videos across 113 countries or regions, and is deliberately weighted toward sustained observation: under a common 120-second decomposition, 43,594 segments reach the full two minutes and account for 51.4% of all footage. Every clip includes a released camera trajectory and hierarchical annotations disentangling subject motion, environment dynamics, static scene content, and camera behavior, resulting in 649,597 temporally grounded segments. Crucially, we further introduce 982 panoramic sequences captured along non-linear trajectories with loops and revisits. These revisits provide repeated observations of the same locations across time and viewpoints, offering essential supervision for learning persistent scene representations, long-term spatial memory, and geometrically consistent world models. Corpus-scale analyses demonstrate complete pose-and-caption coverage, broad geographic and semantic diversity, varied camera trajectories, and highly non-redundant temporal descriptions. Together, these properties make Sekai2 a scalable resource for long-horizon video generation, camera-controllable synthesis, and interactive world-model pre-training.
Chinese Translation
视频世界模型必须捕捉场景随时间和视角的演变。因此，针对长时间生成和相机控制的训练受益于与相机轨迹和时间基础语义配对的长视频。然而，现有语料库很少同时提供这三者：大规模网络视频提供广泛的视觉多样性，但没有轨迹或时间对齐的文本，而姿态标注数据集通常是短期或重建导向的。我们介绍了Sekai2，这是一个多源真实世界视频数据集，旨在将Sekai的世界探索视频用于互动世界建模。该数据集包含来自113个国家或地区的10,428个源视频，共128,892个片段，总计2,826小时，并且故意偏向于持续观察：在统一的120秒分解下，43,594个片段达到了完整的两分钟，占所有视频的51.4%。每个片段都包括一个发布的相机轨迹和分层注释，解开了主体运动、环境动态、静态场景内容和相机行为，形成了649,597个时间基础的片段。至关重要的是，我们进一步介绍了982个沿非线性轨迹捕获的全景序列，这些序列包含循环和重访。这些重访提供了对同一位置在时间和视角上的重复观察，为学习持久场景表示、长期空间记忆和几何一致的世界模型提供了重要的监督。语料库规模的分析展示了完整的姿态和字幕覆盖、广泛的地理和语义多样性、多样的相机轨迹以及高度非冗余的时间描述。这些特性使Sekai2成为一个可扩展的资源，适用于长时间视频生成、可控相机合成和互动世界模型的预训练。
cs.CV / 215 / 2608.09452
A Content-Aware Pure Permutation with Intrinsic Avalanche Effect: Breaking the Diffusion-Permutation Dichotomy
具有内在雪崩效应的内容感知纯置换：打破扩散-置换二分法
Abstract
Pixel permutation is a fundamental tool in image processing, image encryption, and data hiding (including watermarking and steganography) that rearranges pixels without changing their values. A common assumption in the literature is that permutation alone cannot create differential sensitivity; changing one pixel merely relocates that pixel in the output, producing no avalanche effect. This paper challenges this by introducing the Triangular Content-Aware Permutation (TCA) algorithm. The method extracts edge points using Canny and applies Delaunay Triangulation to edges and corners, creating a unique partition. Since triangulation is highly sensitive to image geometry, changing a single pixel alters the edge map, resulting in a completely different triangulation and global permutation pattern. Unlike classical dimension-based permutations and advanced content-aware methods (2025-2026), which lack differential sensitivity, TCA increases NPCR from near-zero to 97.10% solely through pixel relocation. Experiments on 50 images show that TCA, with an average of 14.81 iterations, achieves NPCR = 97.10% and UACI = 20.06%, proving pure permutation can create significant differential sensitivity. Conventional methods maintain near-zero NPCR. The iteration threshold varies from 6.4 to 30.7 based on content complexity. Low PSNR (11.93 dB) and near-zero correlation (~10^-3) confirm superior statistical performance. Although slower than classical methods due to triangulation, this is a deliberate trade-off for stronger security. Given the non-analytic, content-dependent nature of the pattern, TCA is ideal for reference-based encryption, fragile watermarking, and non-blind steganography.
Chinese Translation
像素置换是图像处理、图像加密和数据隐藏（包括水印和隐写术）中的一种基本工具，它在不改变像素值的情况下重新排列像素。文献中的一个常见假设是，仅靠置换无法产生差异敏感性；改变一个像素仅仅是在输出中重新定位该像素，不会产生雪崩效应。本文通过引入三角形内容感知置换（Triangular Content-Aware Permutation, TCA）算法对此提出挑战。该方法使用Canny边缘检测提取边缘点，并对边缘和角点应用Delaunay三角剖分，创建独特的分区。由于三角剖分对图像几何形状高度敏感，改变单个像素会改变边缘图，从而导致完全不同的三角剖分和全局置换模式。与缺乏差异敏感性的经典基于维度的置换和先进的内容感知方法（2025-2026）不同，TCA通过像素重新定位将NPCR从接近零提高到97.10%。对50幅图像的实验表明，TCA在平均14.81次迭代下实现了NPCR = 97.10%和UACI = 20.06%，证明了纯置换可以产生显著的差异敏感性。传统方法保持接近零的NPCR。迭代阈值根据内容复杂性变化，从6.4到30.7不等。低PSNR（11.93 dB）和接近零的相关性（~10^-3）确认了其优越的统计性能。尽管由于三角剖分的原因，速度比经典方法慢，但这是为了更强的安全性而进行的故意权衡。鉴于该模式的非解析性和内容依赖性，TCA非常适合用于基于参考的加密、脆弱水印和非盲隐写术。
cs.CV / 216 / 2608.09467
RecoverFly: A Failure-Aware Reinforcement Learning Post-Training Framework for Aerial Vision-Language Navigation
RecoverFly：一种面向失败的强化学习后训练框架，用于空中视觉-语言导航
Abstract
Unmanned aerial vehicle vision-language navigation (UAV-VLN) requires agents to translate visual observations and language instructions into reliable flight actions in complex environments. Although recent end-to-end UAV vision-language-action (UAV-VLA) policies reduce reliance on separately designed perception, planning, and control modules, their behavior-cloning objectives provide limited corrective supervision for interactive closed-loop execution. Reinforcement learning (RL) offers a promising solution, while its effectiveness is constrained by inefficient use of samples, long-tailed scene distributions, and policy distribution shift during optimization. To this end, we propose RecoverFly, a failure-aware RL post-training framework for end-to-end UAV-VLA policies. Specifically, RecoverFly adapts token-level RL for stable optimization of grammar-constrained autoregressive UAV actions, revisits unresolved failure cases to strengthen corrective learning and sample utilization, and combines a two-stage long-tail scene curriculum with reference-policy regularization to improve scene adaptation while preserving acquired capabilities. Experiments on the TravelUAV benchmark demonstrate that RecoverFly achieves the best performance on the seen, unseen-map, and unseen-object splits. Moreover, compared to the AerialVLA initialization, RecoverFly improves success rate by 3.12 to 8.37 percentage points under a total rollout budget of about 30\% of the training-set size, validating its effectiveness, robustness, and generalization capabilities.
Chinese Translation
无人机视觉-语言导航（UAV-VLN）要求智能体在复杂环境中将视觉观察和语言指令转化为可靠的飞行动作。尽管最近的端到端无人机视觉-语言-动作（UAV-VLA）策略减少了对单独设计的感知、规划和控制模块的依赖，但其行为克隆目标对交互式闭环执行提供的纠正监督有限。强化学习（RL）提供了一种有前景的解决方案，但其有效性受到样本使用效率低、长尾场景分布以及优化过程中策略分布偏移的限制。为此，我们提出了RecoverFly，一种面向失败的RL后训练框架，适用于端到端UAV-VLA策略。具体而言，RecoverFly适应了基于标记的RL，以实现语法约束自回归无人机动作的稳定优化，重新审视未解决的失败案例以增强纠正学习和样本利用，并结合两阶段长尾场景课程与参考策略正则化，以提高场景适应性，同时保留已获得的能力。在TravelUAV基准上的实验表明，RecoverFly在已见、未见地图和未见物体的分割上实现了最佳性能。此外，与AerialVLA初始化相比，RecoverFly在约占训练集大小30%的总回合预算下，成功率提高了3.12到8.37个百分点，验证了其有效性、鲁棒性和泛化能力。
cs.CV / 217 / 2608.09474
FaLCon: Facet-Anchored Retrieval with Late Consensus for Sim2Real Text-Based Person Anomaly Search
FaLCon：基于面向特征的检索与后期共识的Sim2Real文本基础人物异常搜索
Abstract
Text-based person anomaly search requires retrieving real-world pedestrian images from detailed natural-language descriptions using models trained primarily on synthetic data. This Sim2Real setting is particularly challenging because visually similar candidates may differ only in subtle actions, object interactions, or appearance attributes, while applying multimodal large language models to the entire gallery is computationally expensive. We propose an anchor-constrained coarse-to-fine retrieval framework that combines global semantic matching with fine-grained verification. First, each query is represented by its original caption, a structured concatenation, and several semantic facets. Heterogeneous vision-language retrievers are then integrated through robust per-query score calibration and soft claim-aware fusion. Full and concatenated captions serve as anchors to preserve candidate recall, whereas appearance, action, and object facets provide bounded corrective evidence. The resulting candidate pool is further refined by a discriminative Qwen3 reranker and two complementary semantic verification modules based on anomaly-aware cloze completion and multi-agent evidence reasoning. Finally, an uncertainty-gated consensus module adaptively reweights the three experts on ambiguous queries. Experiments on the PAB benchmark show that the proposed soft claim-aware retrieval achieves 86.44% mAP@10, substantially outperforming individual retrieval backbones. The complete framework further improves performance to 95.41% mAP@10, 94.44% R@1, and 99.09% R@5. These results demonstrate that preserving strong global retrieval while restricting expensive semantic reasoning to a small candidate pool is effective for fine-grained Sim2Real person anomaly search. Our code will be available on Github.
Chinese Translation
基于文本的人物异常搜索需要从详细的自然语言描述中检索现实世界的行人图像，而这些模型主要是在合成数据上训练的。这种Sim2Real设置尤其具有挑战性，因为视觉上相似的候选者可能仅在细微的动作、物体交互或外观属性上有所不同，同时将多模态大型语言模型应用于整个图库在计算上是昂贵的。我们提出了一种锚点约束的粗到细检索框架，结合了全局语义匹配与细粒度验证。首先，每个查询通过其原始标题、结构化连接和多个语义面进行表示。然后，通过稳健的每查询得分校准和软声明感知融合，将异构视觉-语言检索器集成在一起。完整的和连接的标题作为锚点以保持候选者的召回，而外观、动作和物体面则提供有界的纠正证据。最终的候选池通过一个判别性的Qwen3重排序器和两个基于异常感知完形填空和多智能体证据推理的互补语义验证模块进一步精炼。最后，一个不确定性门控共识模块自适应地对模糊查询的三个专家进行重新加权。在PAB基准上的实验表明，所提出的软声明感知检索达到了86.44%的mAP@10，显著优于单个检索主干。完整框架进一步将性能提升至95.41%的mAP@10、94.44%的R@1和99.09%的R@5。这些结果表明，在保持强大的全局检索的同时，将昂贵的语义推理限制在一个小的候选池中，对于细粒度的Sim2Real人物异常搜索是有效的。我们的代码将在Github上发布。
cs.CV / 218 / 2608.09475
Agreement-Based Audio-Visual Segmentation:Champion Report for the MeViS-Audio Track in the 8th LSVOS Challenge
基于一致性的音视频分割：第八届LSVOS挑战赛MeViS-Audio赛道冠军报告
Abstract
The MeViS-Audio track asks a system to segment the objects described by a spoken motion expression throughout a video and to return empty masks when the described target is absent. We present a simple staged solution. Qwen3-ASR first converts speech into text. Several video mask tracks are then produced with complementary grounding and segmentation models. Instead of trusting a single prediction, we select the track that has the highest average mask agreement with the other candidates. A small set of explicit direction, count, and plural rules corrects queries that require more than ordinary single-object tracking. Finally, a video-level classifier combines visual, audio-visual, and within-video query scores to decide whether any target is present. The submitted system obtains 0.5952 J &F, 0.7931 no-target accuracy, 0.9205 target accuracy, and a final score of 0.769589. The challenge organizers notified our team that this result ranked first in the track.
Chinese Translation
MeViS-Audio赛道要求系统对视频中通过口语运动表达描述的对象进行分割，并在描述的目标缺失时返回空掩码。我们提出了一种简单的分阶段解决方案。Qwen3-ASR首先将语音转换为文本。随后，利用互补的基础和分割模型生成多个视频掩码轨迹。我们并不依赖单一预测，而是选择与其他候选轨迹具有最高平均掩码一致性的轨迹。一小组显式的方向、计数和复数规则修正了需要超出普通单对象跟踪的查询。最后，一个视频级分类器结合视觉、音视频和视频内查询得分，以决定是否存在任何目标。提交的系统获得了0.5952的J &F，0.7931的无目标准确率，0.9205的目标准确率，以及最终得分0.769589。挑战组织者通知我们的团队，该结果在赛道中排名第一。
cs.CV / 219 / 2608.09482
Beyond Uniform Restoration: Empowering All-in-One Restoration with Pixel-Level Multimodal Guidance
超越均匀恢复：通过像素级多模态指导实现一体化恢复
Abstract
All-in-one image restoration is a unified low-level vision task that aims to effectively recover high-quality images from inputs degraded by various types and levels of corruption using a single model. Recent works have achieved remarkable progress by learning degradation-adaptive prompts or network architectures. However, these methods typically apply a uniform restoration strategy across the entire image, neglecting the fact that different regions may suffer from distinct degradation types and varying degrees of severity. In contrast, we propose to perform restoration at the pixel level, thereby enabling more fine-grained and precise control over the restoration process. Specifically, we present MGN-AIR, a novel pixel-level restoration framework for all-in-one image restoration. Our approach first learns to estimate a pixel-level visual prompt. Then, it leverages both textual and visual prompts to provide global and local degradation cues, guiding the model on where to look and how to restore at each pixel. We conduct extensive experiments on multiple all-in-one image restoration benchmarks, covering a wide range of tasks including denoising, deraining, deblurring, dehazing, desnowing, and low-light enhancement. Experimental results demonstrate that our proposed method consistently and significantly outperforms existing approaches.
Chinese Translation
一体化图像恢复是一项统一的低级视觉任务，旨在通过单一模型有效地从不同类型和程度的损坏输入中恢复高质量图像。近期的研究通过学习适应退化的提示或网络架构取得了显著进展。然而，这些方法通常在整个图像上应用均匀的恢复策略，忽视了不同区域可能遭受不同类型的退化以及不同程度的严重性。相对而言，我们提出在像素级别进行恢复，从而实现对恢复过程的更细粒度和精确控制。具体而言，我们提出了MGN-AIR，这是一种新颖的像素级恢复框架，用于一体化图像恢复。我们的方法首先学习估计像素级视觉提示。然后，它利用文本和视觉提示提供全局和局部的退化线索，引导模型在每个像素上该关注哪里以及如何恢复。我们在多个一体化图像恢复基准上进行了广泛的实验，涵盖了去噪、去雨、去模糊、去雾、去雪和低光增强等多种任务。实验结果表明，我们提出的方法在性能上始终显著优于现有方法。
cs.CV / 220 / 2608.09493
GeoRoute: Geometry-Aware Hybrid Inference for Traffic Future-Frame Prediction
GeoRoute：几何感知混合推理用于交通未来帧预测
Abstract
Long-horizon future-frame prediction is important for autonomous driving, traffic surveillance, and intelligent transportation systems, yet remains challenging due to temporal ghosting, geometry drift, and inconsistent object motion. Recent latent video diffusion models have achieved impressive visual quality, but directly applying them to structured traffic scenes often leads to unstable geometry and degraded temporal coherence over extended horizons. We present a training-free inference framework that stabilizes reliable static structure in pretrained video predictions through multi-frame temporal context and view-conditioned routing. For front-camera videos, our method refines generated futures with a multi-frame depth-layered renderer that projects static geometry from observed history frames while preserving dynamic regions from the generative base model. For heterogeneous traffic views, a frozen vision-language model infers a coarse camera group from the observed clip and selects a specialized motion-based predictor. The framework requires neither retraining nor fine-tuning of the underlying video model and can be applied directly to pretrained generators. We validate the proposed framework on the AI City Challenge Track 5 benchmark, where our final system achieves competitive performance among the top-ranked teams. These results demonstrate that geometry-aware inference-time refinement and view-conditioned hybrid inference can improve static-geometry stability and low-level structural fidelity without changing the original model architecture.
Chinese Translation
长时间范围的未来帧预测对于自动驾驶、交通监控和智能交通系统至关重要，但由于时间鬼影、几何漂移和不一致的物体运动，仍然面临挑战。最近的潜在视频扩散模型在视觉质量上取得了令人印象深刻的成果，但直接将其应用于结构化交通场景往往导致几何不稳定和长时间范围内的时间一致性下降。我们提出了一种无训练推理框架，通过多帧时间上下文和视图条件路由来稳定预训练视频预测中的可靠静态结构。对于前置摄像头视频，我们的方法使用多帧深度分层渲染器来细化生成的未来帧，该渲染器从观察到的历史帧中投影静态几何，同时保留来自生成基础模型的动态区域。对于异构交通视图，一个冻结的视觉-语言模型从观察到的片段中推断出粗略的摄像机组，并选择一个专门的基于运动的预测器。该框架不需要对底层视频模型进行重新训练或微调，可以直接应用于预训练生成器。我们在AI City Challenge Track 5基准上验证了所提出的框架，其中我们的最终系统在排名前列的团队中表现出竞争力。这些结果表明，几何感知的推理时细化和视图条件的混合推理可以在不改变原始模型架构的情况下，提高静态几何的稳定性和低级结构的保真度。
cs.CV / 221 / 2608.09497
SwissCrop25: A National Multi-Year Benchmark for Operational Crop Mapping
SwissCrop25：国家级多年度作物制图基准
Abstract
Operational crop mapping requires models that generalise across years, resolve fine-grained crop taxonomies, and distinguish cropland from surrounding landscapes. However, existing crop mapping datasets enable evaluation of these requirements only in isolation. We therefore introduce SwissCrop25, a national-scale crop mapping benchmark dataset spanning seven growing seasons (2019-2025). SwissCrop25 combines Sentinel-2 time series, daily temperature observations, a fine-grained 73 crop taxonomy including grassland management types, and 5 explicit non-crop land cover classes. To evaluate realistic deployment conditions, we define a leave-one-year-out protocol with joint cropland delineation and crop classification for benchmarking representative crop mapping architectures. Evaluating U-TAE (convolutional temporal-attention model), TSViT (transformer-based spatio-temporal model), and Galileo (EO foundation model) reveals differences between architectures hidden by conventional benchmarks. In this setting, domain-specific models outperform Galileo, with TSViT achieving the best overall performance and a 12 pp macro-mIoU advantage over U-TAE. SwissCrop25 also exposes substantial interannual distribution shifts and shows that incorporating temperature-derived phenological information improves robustness. Finally, in-season evaluation reveals a trade-off between models, with U-TAE performing better early in the season and TSViT gaining an advantage later through improved rare-class discrimination. SwissCrop25 provides a challenging testbed for evaluating crop mapping systems under realistic operational conditions and is publicly released at https://huggingface.co/datasets/EOA-team/SwissCrop25 .
Chinese Translation
操作性作物制图需要能够跨年度推广的模型，能够解析细致的作物分类，并区分农田与周围景观。然而，现有的作物制图数据集仅能孤立地评估这些要求。因此，我们推出了SwissCrop25，这是一个涵盖七个生长季节（2019-2025）的国家级作物制图基准数据集。SwissCrop25结合了Sentinel-2时间序列、每日温度观测、包括草地管理类型在内的73种细致作物分类法，以及5种明确的非作物土地覆盖类型。为了评估现实的部署条件，我们定义了一种留一年法（leave-one-year-out）协议，结合农田划分和作物分类，以基准测试具有代表性的作物制图架构。对U-TAE（卷积时间注意力模型）、TSViT（基于变换器的时空模型）和Galileo（地球观测基础模型）的评估揭示了常规基准隐藏的架构间差异。在这种情况下，特定领域模型的表现优于Galileo，其中TSViT实现了最佳整体性能，并在宏观mIoU上比U-TAE高出12个百分点。SwissCrop25还暴露了显著的年度间分布变化，并显示纳入温度衍生的物候信息可以提高鲁棒性。最后，季内评估揭示了模型之间的权衡，U-TAE在季初表现更佳，而TSViT则通过改善稀有类别的区分能力在后期获得优势。SwissCrop25为在现实操作条件下评估作物制图系统提供了一个具有挑战性的测试平台，并已公开发布于https://huggingface.co/datasets/EOA-team/SwissCrop25。
cs.CV / 222 / 2608.09519
XFeat Revisited: Reproducibility and Evaluation of a Lightweight Image Matcher
XFeat 重新审视：轻量级图像匹配器的可重复性与评估
Abstract
We present a reproducibility study of XFeat, a lightweight local feature extractor and matcher designed to identify corresponding points across images efficiently on resource-constrained hardware. We re-implement the architecture based on the paper and supplementary material, re-evaluate the authors' released checkpoint alongside our re-implementation, and conduct additional architectural ablations to examine design choices that were not fully justified in the original work. This distinction between re-evaluation and reproduction is important, as the paper, supplement, and public code differ in several implementation details, including the backbone layout, fusion block, and training losses. Empirically, our reproduced models closely match and, in some cases, outperform the re-evaluated original checkpoint on MegaDepth-1500 and ScanNet-1500, supporting the main claim that XFeat provides a strong accuracy-efficiency trade-off for standard image-matching benchmarks. Our ablations provide a more nuanced view of two architectural arguments from the original paper. In particular, the parallel keypoint branch is important for semi-dense matching, but its benefit is less pronounced than originally claimed, while the evidence for the specific placement of the single skip-connection remains inconclusive. Finally, we reproduce the original downstream evaluations and find close agreement for homography estimation, while Aachen visual localization remains below the reported results, even for the released checkpoint, suggesting sensitivity to underspecified evaluation details. We then extend the analysis to zero-shot out-of-distribution and cross-modal matching across retinal, thermal-visible, and multimodal remote-sensing imagery, where XFeat remains effective in some settings but degrades sharply under severe modality shifts.
Chinese Translation
我们对 XFeat 进行了可重复性研究，这是一种轻量级的局部特征提取器和匹配器，旨在在资源受限的硬件上高效地识别图像之间的对应点。我们基于论文及补充材料重新实现了该架构，重新评估了作者发布的检查点，并进行了额外的架构消融实验，以检验原始工作中未充分论证的设计选择。重新评估与复制之间的区别非常重要，因为论文、补充材料和公共代码在多个实现细节上存在差异，包括主干网络布局、融合块和训练损失。从经验上看，我们复制的模型与重新评估的原始检查点在 MegaDepth-1500 和 ScanNet-1500 上紧密匹配，并在某些情况下表现优于原始检查点，支持 XFeat 在标准图像匹配基准中提供良好的准确性与效率的权衡这一主要论点。我们的消融实验对原始论文中的两个架构论点提供了更细致的视角。特别是，平行关键点分支对于半稠密匹配非常重要，但其益处并不像最初所声称的那样明显，而单一跳跃连接的具体位置证据仍然不确定。最后，我们重现了原始下游评估，并发现单应性估计的结果非常接近，而亚琛视觉定位的结果仍低于报告的结果，即使是对于发布的检查点，这表明对未充分指定的评估细节敏感。随后，我们将分析扩展到零样本的分布外和跨模态匹配，涉及视网膜、热可见和多模态遥感图像，在某些设置中 XFeat 仍然有效，但在严重的模态转变下急剧下降。
cs.CV / 223 / 2608.09520
A Height-Constrained 2-Point Minimal Solver for Pose Estimation from Active LED Markers with Event Cameras
一种高度受限的两点最小解算器用于基于活动LED标记的事件相机姿态估计
Abstract
In many autonomous applications requiring real-time localization, active marker-based systems are preferred due to their low latency and ease of deployment compared to computationally demanding feature-based methods. Event~\mbox{cameras} offer high temporal resolution and minimal delay and are commonly used with active LED markers for robust real-time localization. Existing methods typically rely on Perspective-n-Point (PnP) solvers for pose estimation. However, structured marker layouts can be challenging to deploy in space-constrained scenarios, while partial self-motion information (e.g., gravity direction and altitude) is readily available from onboard sensors. We derive a robust and accurate minimal solver that estimates camera pose from only two LED markers by incorporating known tilt angle and camera height measured by an onboard sensor, such as an IMU or an altimeter. The proposed formulation uniquely determines the camera pose through both a closed-form and a linear least-squares solution. We further analyze degenerate configurations and characterize the conditions under which height information does not contribute to rotation estimation. For evaluation, we developed an event-based active marker system to collect real-world data with ground truth from a motion capture system. Experiments on both synthetic and real data demonstrate improved accuracy over the state-of-the-art P2P solver and competitive performance relative to P3P.
Chinese Translation
在许多需要实时定位的自主应用中，基于活动标记的系统因其低延迟和易于部署而被优先选择，相较于计算需求较高的特征基方法。事件相机提供高时间分辨率和最小延迟，通常与活动LED标记结合使用，以实现稳健的实时定位。现有方法通常依赖于透视n点（Perspective-n-Point, PnP）解算器进行姿态估计。然而，在空间受限的场景中，结构化标记布局的部署可能面临挑战，而来自机载传感器的部分自运动信息（例如重力方向和高度）则 readily 可用。我们推导出一种稳健且准确的最小解算器，仅通过两个LED标记估计相机姿态，方法是结合已知的倾斜角度和通过机载传感器（如惯性测量单元IMU或高度计）测量的相机高度。所提出的公式通过闭式解和线性最小二乘解唯一确定相机姿态。我们进一步分析了退化配置，并表征了高度信息对旋转估计没有贡献的条件。为了评估，我们开发了一种基于事件的活动标记系统，以从运动捕捉系统收集具有真实值的实际数据。对合成数据和真实数据的实验表明，相较于最先进的P2P解算器，准确性有所提高，并且相对于P3P具有竞争力的性能。
cs.CV / 224 / 2608.09522
TriView-YOLO: Early Multi-View Fusion for Ground Penetrating Radar Cavity Detection in Soft, High-Water-Content Soils
TriView-YOLO：用于软质高水分土壤中地面穿透雷达空腔检测的早期多视角融合
Abstract
Automated detection of subsurface cavities from Ground Penetrating Radar (GPR) is most difficult in soft, high-water-content ground, where conductive, water-saturated soil attenuates the signal and degrades cavity reflections, yet this is also the condition under which cavities most readily form. This paper proposes TriView-YOLO, a multi-view YOLOv12 detector for road cavity screening in such ground. Three co-registered views (longitudinal B-scan, horizontal C-scan, and cross-section B-scan) form a 9-channel input fused by a TripleInputConv layer that replaces the YOLOv12 stem; the rest of the network is unchanged, and bounding boxes are required on the longitudinal view only. Training used 1,600 expert-verified field samples, principally metropolitan road surveys of Bangkok, Thailand, acquired with a vehicle-mounted multichannel three-dimensional GPR mobile mapping system, with surveys over the firmer subgrades of Japan added to training and validation only. The test set comes exclusively from the Bangkok surveys, over soft marine clay with 80-140% water content and a water table at 1-2 m depth, a ground condition for which no dedicated deep learning cavity-detection evaluation has been reported. On this unaugmented, field-only test set, split randomly within surveys, the proposed model attains mAP50 of 0.558 +/- 0.028 over three seeds at 23.6 GFLOPs and 3.1 ms per image. Ablations show that removing the auxiliary views lowers mAP50 and recall, whereas public and synthetic training images, DINOv3 features, larger model scale, and COCO pretraining bring no gain.
Chinese Translation
从地面穿透雷达（GPR）自动检测地下空腔在软质高水分土壤中最为困难，因为导电的水饱和土壤会衰减信号并降低空腔反射的质量，然而，这也是空腔最容易形成的条件。本文提出了TriView-YOLO，一种用于此类土壤中道路空腔筛查的多视角YOLOv12检测器。三个共注册视角（纵向B扫描、横向C扫描和横截面B扫描）形成一个由TripleInputConv层融合的9通道输入，该层替代了YOLOv12的主干；其余网络保持不变，并且仅在纵向视图上需要边界框。训练使用了1,600个专家验证的现场样本，主要是泰国曼谷的城市道路调查，这些样本是通过车载多通道三维GPR移动测绘系统获取的，同时在训练和验证中还加入了日本较坚硬的路基的调查数据。测试集完全来自曼谷的调查，覆盖80-140%水分含量的软质海洋粘土，水位在1-2米深处，这是一个尚未报道专门的深度学习空腔检测评估的土壤条件。在这个未增强的仅现场测试集上，随机分割于调查中，所提模型在23.6 GFLOPs和每幅图像3.1毫秒的情况下，达到了mAP50为0.558 +/- 0.028，基于三个种子的结果。消融实验表明，去除辅助视角会降低mAP50和召回率，而公共和合成训练图像、DINOv3特征、更大的模型规模和COCO预训练则没有带来收益。
cs.CV / 225 / 2608.09529
Towards Expressive and Faithful Audio-to-Image Generation: A Unified Multimodal Dataset and Synthesis Framework
迈向富有表现力和忠实度的音频到图像生成：一个统一的多模态数据集和合成框架
Abstract
As an important subfield of cross-modal generation, synthesizing static visual content in the form of images from audio, namely audio-to-image (A2I) generation, has attracted increasing research attention in recent years. Nevertheless, despite the remarkable visual quality of modern text-to-image (T2I) models, the performance of A2I remains fundamentally limited by traditional datasets, which often lack both high-fidelity images and precise cross-modal alignment. As a result, existing methods still struggle to achieve high-quality audio-to-image generation through finetuning strong T2I models, thereby constraining practical applications in this area. Motivated by this gap, we introduce A2I-Set, a unified, high-quality tri-modal dataset consisting of 323K paired audio, images, and detailed text captions, specifically designed for audio-visual research, including audio-conditioned image generation. Besides, we developed a new mixed-source test set for the A2I task through human supervision. We further propose an A2I model, AudioCanvas, fine-tuned on our A2I-Set. Experiments show that AudioCanvas achieves more visually expressive as well as cross-modal alignment results that generally outperforming existing approaches. Our dataset and source code are available at https://github.com/gdx012/A2I-Generation.
Chinese Translation
作为跨模态生成的重要子领域，从音频合成静态视觉内容（即音频到图像（A2I）生成）近年来吸引了越来越多的研究关注。然而，尽管现代文本到图像（T2I）模型在视觉质量上表现出色，A2I的性能仍然受到传统数据集的根本限制，这些数据集往往缺乏高保真图像和精确的跨模态对齐。因此，现有方法在通过微调强大的T2I模型实现高质量音频到图像生成方面仍然面临挑战，从而限制了该领域的实际应用。基于这一差距，我们引入了A2I-Set，这是一个统一的高质量三模态数据集，包含323K对音频、图像和详细文本说明，专为音频视觉研究而设计，包括音频条件下的图像生成。此外，我们通过人工监督为A2I任务开发了一个新的混合源测试集。我们进一步提出了一种A2I模型AudioCanvas，并在我们的A2I-Set上进行了微调。实验表明，AudioCanvas在视觉表现力和跨模态对齐结果上均优于现有方法。我们的数据集和源代码可在https://github.com/gdx012/A2I-Generation获取。
cs.CV / 226 / 2608.09536
DocPure: Prompt-Free Unified Document Restoration via Degradation-Aware Structure-Guided Wavelet Modulation
DocPure：通过降解感知结构引导的小波调制实现无提示统一文档修复
Abstract
High-quality document images are pivotal for information archiving and downstream automatic processing. However, they are frequently compromised by diverse degradations during uncontrolled acquisition and transmission. While unified document restoration techniques have been proposed to restore images from multiple degradations, they often struggle with training multiple degradation-specific models, reliance on manual task-specific prompts, or cross-task data pairing. To address these limitations, we propose DocPure, a prompt-free unified framework that achieves degradation-aware document restoration. We design a degradation-aware structure auto-encoder with degradation-informed routing regularization to predict clean structural priors from degraded inputs. The model is prompt-free at inference, and degradation labels are only used as auxiliary supervision for the routing regularization during training. Furthermore, we introduce a structure-guided wavelet interaction mechanism to bridge frequency-domain features and spatial semantics. Within the structure-guided wavelet interaction mechanism, a cross-frequency adaptive modulation utilizes low-frequency sub-bands to modulate high-frequency recovery, ensuring structural consistency. Extensive experiments demonstrate that DocPure achieves strong performance compared with state-of-the-art methods across various tasks, including deblurring, denoising, compression artifact reduction, and deshadowing.
Chinese Translation
高质量的文档图像对于信息归档和后续自动处理至关重要。然而，在不受控制的获取和传输过程中，它们经常受到各种降解的影响。虽然已经提出了统一文档修复技术以从多种降解中恢复图像，但它们通常在训练多个特定于降解的模型、依赖手动任务特定提示或跨任务数据配对方面面临困难。为了解决这些限制，我们提出了DocPure，这是一种无提示的统一框架，能够实现降解感知的文档修复。我们设计了一种降解感知结构自编码器，结合降解信息引导的路由正则化，从降解输入中预测干净的结构先验。该模型在推理时不需要提示，降解标签仅在训练期间作为路由正则化的辅助监督。此外，我们引入了一种结构引导的小波交互机制，以桥接频域特征和空间语义。在结构引导的小波交互机制中，跨频率自适应调制利用低频子带来调制高频恢复，确保结构一致性。大量实验表明，DocPure在去模糊、去噪、压缩伪影减少和去阴影等多种任务中，相较于最先进的方法表现出色。
cs.CV / 227 / 2608.09541
Towards Collaborative Joint Perception and Prediction: Framework, Baseline Evaluation, and Deployment Perspectives
朝向协作联合感知与预测：框架、基线评估与部署视角
Abstract
Connected Autonomous Vehicles (CAVs) increasingly exploit Vehicle-to-Everything (V2X) communication to exchange multi-source sensor information, enabling advanced Collaborative Perception (CP) capabilities. Extending beyond these capabilities, this work focuses on Collaborative Joint Perception and Prediction (Co-P&P), a paradigm that unifies CP with motion prediction to mitigate two persistent challenges: the accumulation of perception errors and visual occlusions. We present a conceptual framework for Collaborative Joint Perception and Prediction (Co-P&P) that improves motion prediction of surrounding road users, thereby enhancing situational awareness in complex and dynamic traffic environments. Building upon our preliminary study, this extended version compares the performance of different fusion strategies and establishes baseline performance for a modular design of perception and prediction. Experimental results show that prediction-level fusion leads to a decline in overall system performance compared to detection-level or tracking-level fusion. We further implement a minimal end-to-end Co-P&P prototype that couples collaborative point-cloud sharing via the RENO neural codec with joint detection-forecasting via FutureDet, showing that collaboration improves forecasting accuracy while neural compression preserves this benefit at roughly 34x lower communication bandwidth.
Chinese Translation
连接自主车辆（CAVs）越来越多地利用车与所有事物（V2X）通信来交换多源传感器信息，从而实现先进的协作感知（CP）能力。超越这些能力，本研究聚焦于协作联合感知与预测（Co-P&P），这一范式将协作感知与运动预测结合起来，以缓解两个持续存在的挑战：感知误差的累积和视觉遮挡。我们提出了一个协作联合感知与预测（Co-P&P）的概念框架，旨在改善周围道路使用者的运动预测，从而增强复杂和动态交通环境中的情境意识。在我们的初步研究基础上，本扩展版本比较了不同融合策略的性能，并为感知与预测的模块化设计建立了基线性能。实验结果表明，与检测级或跟踪级融合相比，预测级融合导致整体系统性能下降。我们进一步实现了一个最小化的端到端Co-P&P原型，该原型通过RENO神经编码器实现协作点云共享，并通过FutureDet实现联合检测与预测，显示出协作提高了预测准确性，而神经压缩在大约34倍更低的通信带宽下保留了这一优势。
cs.CV / 228 / 2608.09550
PressureMesh: 3D Human Mesh Estimation from Multi-Device Pressure Images
PressureMesh：基于多设备压力图像的3D人类网格估计
Abstract
Human pose monitoring is crucial in fields such as rehabilitation assessment and human-computer interaction. Due to its privacy-preserving nature, pressure-based human pose monitoring has become a primary approach for unobtrusive sensing. However, existing methods are generally limited to a single device, which restricts the effective monitoring range. To address this limitation, we propose MDP-Net, an end-to-end network capable of directly estimating human meshes from temporal pressure data across multiple devices. We introduce a multimodal fusion mechanism inspired by the Mixture of Experts (MoE) framework to achieve effective complementarity and enhancement of cross-device pressure information. To support the training and evaluation of MDP-Net, we constructed MDP, a high-quality multi-device temporal pressure dataset that includes various pose labels such as 2D/3D joints and human meshes. Experimental results demonstrate that MDP-Net achieves a joint position error of 12.6 cm on the MDP dataset. These results prove that fusing multi-device pressure information is an effective and promising new solution for daily human pose monitoring.
Chinese Translation
人类姿态监测在康复评估和人机交互等领域至关重要。由于其保护隐私的特性，基于压力的人类姿态监测已成为一种主要的无干扰感知方法。然而，现有的方法通常仅限于单一设备，这限制了有效监测范围。为了解决这一限制，我们提出了MDP-Net，一种能够直接从多个设备的时间压力数据中估计人类网格的端到端网络。我们引入了一种受混合专家（Mixture of Experts, MoE）框架启发的多模态融合机制，以实现跨设备压力信息的有效互补和增强。为了支持MDP-Net的训练和评估，我们构建了MDP，一个高质量的多设备时间压力数据集，其中包含各种姿态标签，如2D/3D关节和人类网格。实验结果表明，MDP-Net在MDP数据集上的关节位置误差为12.6厘米。这些结果证明了融合多设备压力信息是一种有效且有前景的新解决方案，用于日常人类姿态监测。
cs.CV / 229 / 2608.09564
From Semantic Grounding to Decision Optimization: A Unified Framework for Long-Horizon UAV Vision-Language Navigation
从语义基础到决策优化：一种统一的长时间跨度无人机视觉-语言导航框架
Abstract
UAV vision-language navigation (UAV-VLN) focuses on enabling an aerial agent to follow natural-language instructions in open 3D environments from egocentric visual observations. Current approaches suffer from three coupled issues: weak grounding of instruction-relevant landmarks in visual observations, insufficient exploitation of long-horizon history, and unstable decisions under local traps or repeated exploration. To address these issues, we propose a unified semantic-to-decision framework. First, we present an instruction-grounded semantic enhancement module that injects object-level semantics and relative spatial cues into the current observation state. Subsequently, we develop a relevance-aware dynamic temporal aggregation strategy that reweights the full history buffer while converting a few high-relevance frames into structured landmark prompts for the decoder. Finally, we devise a topology-aware decision method that combines local-optimum cognition with group-relative policy optimization under progress, goal, semantic, and path-compliance rewards. Experiments on the widely used AerialVLN and OpenFly benchmarks clearly demonstrate that our method achieves state-of-the-art performance.
Chinese Translation
无人机视觉-语言导航（UAV-VLN）旨在使空中代理能够在以自我为中心的视觉观察下，遵循开放3D环境中的自然语言指令。目前的方法面临三个相互关联的问题：在视觉观察中指令相关地标的基础较弱、对长时间历史的利用不足，以及在局部陷阱或重复探索下决策不稳定。为了解决这些问题，我们提出了一种统一的语义到决策框架。首先，我们提出了一个指令基础的语义增强模块，该模块将对象级语义和相对空间线索注入当前观察状态。随后，我们开发了一种关注相关性的动态时间聚合策略，该策略在转换少量高相关帧为解码器的结构化地标提示时，对完整历史缓冲区进行重新加权。最后，我们设计了一种拓扑感知的决策方法，该方法结合了局部最优认知与在进展、目标、语义和路径合规奖励下的群体相对策略优化。在广泛使用的AerialVLN和OpenFly基准测试中的实验清楚地表明，我们的方法达到了最先进的性能。
cs.CV / 230 / 2608.09573
VideoVIBE: A Video-Grounded Diagnostic Benchmark for One-Shot Interactive Website Generation
VideoVIBE：一种基于视频的单次交互式网站生成诊断基准
Abstract
Natural-language-driven "vibe coding" enables the one-shot generation of visually rich and interactive web applications, yet reliable assessment of their quality has not kept pace. Existing evaluations often score isolated artifacts or final task outcomes, offering limited evidence about which failures occur and why. We introduce VideoVIBE, a video-grounded benchmark that transforms human-operated webpage recordings into fine-grained diagnostic tasks. It contains approximately 1.7K diagnostic Video QA instances derived from 6,338 verified failures across generated webpages, spanning semantic-logical, visual-motion, structural-temporal, and functional failures. Diagnoses are grounded primarily in recorded presentation and behavior, with webpage source code used as complementary context. We further propose V2Lens, a training-free, evidence-grounded multi-agent system that challenges and selectively refines initial video-based diagnoses through targeted visual and source-code verification. Across thirteen closed-source and open-weight Video MLLMs, Gemini-2.5-Flash is the strongest standalone model with a score of 64.54, while V2Lens reaches 71.72, an improvement of 7.18 points. Together, our results show that video-grounded evaluation can move beyond isolated artifacts and aggregate outcomes toward a behaviorally faithful and diagnostically informative account of generated application quality.
Chinese Translation
自然语言驱动的“氛围编码”使得视觉丰富且互动性强的网页应用能够一次性生成，但对其质量的可靠评估尚未跟上。现有评估通常只对孤立的工件或最终任务结果进行评分，提供的证据有限，无法明确哪些失败发生以及原因。我们引入了VideoVIBE，这是一种基于视频的基准，将人工操作的网页录制转化为细粒度的诊断任务。它包含大约1.7K个诊断视频问答实例，这些实例源自6,338个经过验证的生成网页失败，涵盖语义逻辑、视觉运动、结构时间和功能失败。诊断主要基于录制的展示和行为，同时网页源代码作为补充上下文。我们进一步提出了V2Lens，这是一种无训练、基于证据的多智能体系统，通过针对性的视觉和源代码验证来挑战和选择性地精炼初步的视频基础诊断。在十三个闭源和开放权重的视频多模态大语言模型中，Gemini-2.5-Flash是最强的独立模型，得分为64.54，而V2Lens达到了71.72，提升了7.18分。综合来看，我们的结果表明，基于视频的评估可以超越孤立的工件和聚合结果，朝着对生成应用质量的行为真实和诊断性信息的描述迈进。
cs.CV / 231 / 2608.09575
MSP-Net: Manifold-Guided Spectral Prompt Network for Hyperspectral Object Tracking
MSP-Net：流形引导的光谱提示网络用于高光谱目标跟踪
Abstract
Hyperspectral object tracking leverages abundant spectral information to provide unique advantages for target discrimination in complex scenes. However, existing methods typically treat hyperspectral images as multi-channel extensions of RGB images, performing feature fusion in fixed band order. This approach leads to models dependent on specific sensor configurations while neglecting manifold relationships between bands, making generalization to heterogeneous sensors difficult. Moreover, the discriminative contribution of bands dynamically changes with target attributes and scene variations, further limiting the representational capacity of static fusion strategies. To address this, we propose the Manifold-Guided Spectral Prompt Network (MSP-Net). This network first reconstructs band relationships and forms adaptive spectral grouping through graph-driven manifold routing, then jointly integrates grouped spectral statistics with template appearance to construct target-related dynamic conditional prompts, enhancing target features while suppressing background interference. Furthermore, as tracking progresses, spectral conditions continuously evolve based on intermediate target representations, enabling target prompts to adapt in real-time to appearance and scene changes. Meanwhile, reliable historical states are used to constrain target localization and scale fluctuations, significantly improving temporal stability in cross-sensor tracking. Experiments on HOT2020 and HOT2023 demonstrate that MSP-Net achieves AUC and Precision exceeding 0.80 and 0.96, respectively, exhibiting exceptional robustness under heterogeneous sensors, target deformation, and complex background conditions. The code will be released at https://github.com/GGML668897/MSP-Net.
Chinese Translation
高光谱目标跟踪利用丰富的光谱信息，为复杂场景中的目标识别提供独特优势。然而，现有方法通常将高光谱图像视为RGB图像的多通道扩展，以固定波段顺序进行特征融合。这种方法导致模型依赖于特定的传感器配置，同时忽视了波段之间的流形关系，使得对异构传感器的泛化变得困难。此外，波段的区分性贡献会随着目标属性和场景变化而动态变化，进一步限制了静态融合策略的表征能力。为此，我们提出了流形引导的光谱提示网络（MSP-Net）。该网络首先通过图驱动的流形路由重建波段关系，并形成自适应光谱分组，然后将分组的光谱统计信息与模板外观共同整合，以构建与目标相关的动态条件提示，从而增强目标特征，同时抑制背景干扰。此外，随着跟踪的进行，光谱条件会根据中间目标表征不断演变，使得目标提示能够实时适应外观和场景变化。同时，可靠的历史状态被用来约束目标定位和尺度波动，显著提高了跨传感器跟踪的时间稳定性。在HOT2020和HOT2023上的实验表明，MSP-Net的AUC和精确度分别超过0.80和0.96，在异构传感器、目标变形和复杂背景条件下表现出卓越的鲁棒性。代码将发布在https://github.com/GGML668897/MSP-Net。
cs.CV / 232 / 2608.09579
You Only Flow Once: Calibrated and Real-Time Radar Pose Estimation with Multi-Hypothesis Normalizing Flows
你只需流一次：基于多假设归一化流的校准实时雷达姿态估计
Abstract
Sparse and noisy millimeter-wave radar point cloud observations often correspond to multiple plausible human poses, making deterministic pose estimation fundamentally ill-posed. Yet existing radar methods remain deterministic, collapsing this ambiguity into a single estimate. Diffusion-based alternatives can model multi-hypothesis distributions but require costly sequential denoising for each distribution sample and lack calibrated uncertainty. We propose Multi-Hypothesis Normalizing Flow Pose Generator (MH-NFPG), which models pose distributions from radar point clouds using a conditional normalizing flow. Specifically, we combine a spatiotemporal transformer backbone with a normalizing flow that transforms a Laplace base distribution into an expressive posterior, generated in parallel through a single forward pass. Leveraging this efficiency, we outperform diffusion-based alternatives in calibration across three radar benchmarks (MM-Fi, mmRadPose, mRI), improve pose accuracy on two, and match it on the third, while achieving over 20x faster inference for applications and reducing calibration error by up to 85%. We find that calibration degrades substantially for diffusion models, whereas our flow-based approach maintains reliable coverage, also in cross-environment settings. These results demonstrate normalizing flows as a practical alternative to diffusion models for real-time, uncertainty-aware radar pose estimation. Our code will be made publicly available.
Chinese Translation
稀疏且噪声较大的毫米波雷达点云观测通常对应多个合理的人体姿态，这使得确定性姿态估计在本质上是一个不适定问题。然而，现有的雷达方法仍然是确定性的，将这种模糊性压缩为单一估计。基于扩散的替代方法可以建模多假设分布，但需要对每个分布样本进行代价高昂的顺序去噪，并且缺乏校准的不确定性。我们提出了多假设归一化流姿态生成器（Multi-Hypothesis Normalizing Flow Pose Generator，MH-NFPG），该方法利用条件归一化流从雷达点云中建模姿态分布。具体而言，我们将时空变换器主干与归一化流相结合，将拉普拉斯基础分布转化为富有表现力的后验分布，通过单次前向传播并行生成。利用这种效率，我们在三个雷达基准（MM-Fi、mmRadPose、mRI）上超越了基于扩散的替代方法，在两个基准上提高了姿态准确性，并在第三个基准上达到了相同的水平，同时在应用中实现了超过20倍的推理速度提升，并将校准误差降低了多达85%。我们发现，扩散模型的校准显著下降，而我们的基于流的方法在跨环境设置中也保持了可靠的覆盖。这些结果表明，归一化流是实时、不确定性感知雷达姿态估计的一个实用替代方案。我们的代码将公开发布。
cs.CV / 233 / 2608.09581
GenTrack3: Hybrid Stochastic-Deterministic Online Multi-Object Tracking with Cluster-Aware Association
GenTrack3：具有集群感知关联的混合随机-确定性在线多目标跟踪
Abstract
Multi-object tracking (MOT) involves maintaining consistent target identities as objects dynamically enter and leave a scene. Deterministic approaches, such as tracking-by-detection with data association, produce reproducible results and are computationally efficient, but they rely heavily on motion models and are sensitive to noisy detections that can lead to association errors. In contrast, stochastic methods explicitly model uncertainty and can better handle complex non-linear dynamics, albeit at the cost of increased computational complexity and variability arising from random sampling. This paper presents an online MOT framework that integrates deterministic and stochastic principles to achieve robust tracking under uncertainty. Furthermore, a novel track-to-detection matching approach is introduced to enhance scalability with increasing target numbers while supporting group tracking. The tracking inference mechanism employs a tracklet that includes identifiers, states, velocities, track penalties and track ages of targets, supporting a systematic tracking pipeline. Each target is associated with a stochastic particle set to compute the matching cost to detections. Reference implementations of the proposed approach and baseline trackers can be found on GitHub: https://github.com/SDU-VelKoTek/GenTrack3.
Chinese Translation
多目标跟踪（MOT）涉及在物体动态进入和离开场景时保持一致的目标身份。确定性方法，如基于检测的跟踪与数据关联，能够产生可重复的结果且计算效率高，但它们在很大程度上依赖于运动模型，并且对噪声检测敏感，这可能导致关联错误。相比之下，随机方法明确建模不确定性，能够更好地处理复杂的非线性动态，尽管这会增加计算复杂性和由于随机采样而产生的变异性。本文提出了一种在线MOT框架，结合了确定性和随机原则，以在不确定性下实现稳健的跟踪。此外，提出了一种新颖的轨迹与检测匹配方法，以增强在目标数量增加时的可扩展性，同时支持群体跟踪。跟踪推理机制采用了一个包含目标标识符、状态、速度、轨迹惩罚和轨迹年龄的轨迹片段，支持系统化的跟踪流程。每个目标与一个随机粒子集相关联，以计算与检测的匹配成本。所提方法和基线跟踪器的参考实现可以在GitHub上找到：https://github.com/SDU-VelKoTek/GenTrack3。
cs.CV / 234 / 2608.09590
TeaMatch: Teachable Cross-Modal Representation Learning for 2D-3D Matching
TeaMatch：可教的跨模态表示学习用于2D-3D匹配
Abstract
Learning reliable correspondences between images and point clouds is fundamental for 2D-3D matching. Despite recent progress in detection-free methods, existing approaches primarily optimize matching within a single model and often struggle to maintain reliable correspondences under challenging conditions such as noisy inputs, low overlap, and ambiguous structures. In this work, we propose TeaMatch, a novel framework that introduces teachability as a criterion for cross-modal representation learning. We define teachability as the ability of a representation to be effectively recovered by weak learners under degraded inputs, reflecting its structural consistency and robustness. To this end, we construct a set of task-specific weak students that simulate common failure modes and train them to imitate the teacher on a training split while evaluating their recoverability on a disjoint meta split. The teacher is then optimized to improve the students' ability to recover reliable correspondences, guided by correspondence-level and geometry-aware constraints. Our framework can be seamlessly integrated into existing coarse-to-fine matching pipelines without additional inference cost. Extensive experiments demonstrate that TeaMatch improves matching robustness and achieves state-of-the-art performance on challenging 2D-3D matching benchmarks.
Chinese Translation
学习图像与点云之间可靠的对应关系是2D-3D匹配的基础。尽管在无检测方法上取得了近期进展，现有方法主要在单一模型内优化匹配，往往难以在噪声输入、低重叠和模糊结构等挑战性条件下维持可靠的对应关系。在本研究中，我们提出了TeaMatch，一个新颖的框架，将可教性作为跨模态表示学习的标准。我们将可教性定义为在退化输入下，表示能够被弱学习者有效恢复的能力，反映其结构一致性和鲁棒性。为此，我们构建了一组特定任务的弱学生，模拟常见的失败模式，并训练它们在训练集上模仿教师，同时在不相交的元分割上评估其可恢复性。然后，教师被优化以提升学生恢复可靠对应关系的能力，受对应级别和几何感知约束的指导。我们的框架可以无缝集成到现有的粗到细匹配管道中，而无需额外的推理成本。大量实验表明，TeaMatch提高了匹配的鲁棒性，并在具有挑战性的2D-3D匹配基准上达到了最先进的性能。
cs.CV / 235 / 2608.09594
Illusion or Integrity? Geometrical Consistency Metric for AIGC Video Quality Evaluation
幻觉还是完整性？用于AIGC视频质量评估的几何一致性度量
Abstract
Recently, AI-driven video generation has attracted considerable attention. This surge increases the demand for reliable video quality assessment (VQA) metrics to evaluate AI-generated content (AIGC) videos and guide model optimization. Existing studies assess video quality through visual harmony, video-text consistency, and domain-specific alignment, yet lack quantitative metrics for measuring fidelity to physical laws. To address this limitation, we present a novel benchmark that evaluates the quality of AIGC videos based on their compliance with physical principles by quantitatively measuring geometric consistency across frames extracted from generated sequences. This serves as a proxy for estimating the extent to which generated videos conform to real-world physical rules. Specifically, GeoCon-Bench captures global motion through translation estimation, fits homography or fundamental matrix models using background correspondences, and reports complementary metrics, including inlier ratio and geometric error. We also release a dataset containing 20 scenes across six motion categories. Experiments on state-of-the-art AIGC models demonstrate the reliability of GeoCon-Bench as a video quality assessment metric.
Chinese Translation
近年来，基于人工智能的视频生成引起了广泛关注。这一趋势增加了对可靠的视频质量评估（VQA）指标的需求，以评估AI生成内容（AIGC）视频并指导模型优化。现有研究通过视觉和谐、视频-文本一致性和领域特定对齐来评估视频质量，但缺乏测量对物理法则遵循程度的定量指标。为了解决这一局限性，我们提出了一种新基准，通过定量测量从生成序列中提取的帧之间的几何一致性，评估AIGC视频的质量，以此作为估计生成视频遵循现实世界物理规则程度的代理。具体而言，GeoCon-Bench通过平移估计捕捉全局运动，利用背景对应关系拟合单应性或基础矩阵模型，并报告补充指标，包括内点比例和几何误差。我们还发布了一个包含20个场景的数据库，涵盖六种运动类别。对最先进的AIGC模型的实验表明，GeoCon-Bench作为视频质量评估指标的可靠性。
cs.CV / 236 / 2608.09597
ResemBrick: Brick Reconstruction from Photographs with Perceptual Fidelity and Buildability
ResemBrick：具有感知保真度和可建造性的砖块重建
Abstract
Producing a hand-buildable, colored brick model of a 3D object from a few casual photographs is a clean testbed for a broader challenge: generating 3D content that meets hard physical-assembly constraints under a discrete, budget-limited voxel grid. On a coarse lattice, visual resemblance and structural stability pull against each other, yet prior brick pipelines address only one side and treat voxelization as fixed preprocessing rather than a variable to optimize. We present ResemBrick, which couples the two. Budgeted occupancy completion reframes discretization as allocation: given a target occupied-voxel count, a single resolution-conditioned network decides in one feed-forward pass which surface voxels to fill for best appearance, one weight set spanning 13 resolutions. Buildability by construction then combines support- and look-ahead-aware greedy placement with a deterministic, provably terminating repair that grounds every floating component. Under a matched budget, ResemBrick surpasses existing voxel selectors in perceptual fidelity while uniquely reaching zero floating and zero unstable bricks on unfiltered held-out objects; as a complete pipeline, it attains the best perceptual fidelity among prior brick-construction systems. Our results point to treating discretization and assembly as tightly coupled stages rather than independent ones.
Chinese Translation
从几张随意的照片中生成一个可手工建造的彩色砖块模型，以3D物体为基础，是一个更广泛挑战的良好测试平台：在离散的、预算有限的体素网格下生成满足严格物理组装约束的3D内容。在粗糙的格点上，视觉相似性和结构稳定性相互牵制，而现有的砖块处理流程仅关注其中一方面，并将体素化视为固定的预处理步骤，而非一个可优化的变量。我们提出了ResemBrick，将这两者结合起来。预算占用补全将离散化重新框定为分配：在给定目标占用体素数量的情况下，一个单一的基于分辨率的网络在一次前馈传递中决定填充哪些表面体素以获得最佳外观，权重集跨越13种分辨率。然后，通过构造实现的可建造性结合了支持和前瞻性意识的贪婪放置与一个确定性的、可证明终止的修复，确保每个浮动组件都得到支撑。在匹配的预算下，ResemBrick在感知保真度上超越了现有的体素选择器，同时在未过滤的保留对象上独特地实现了零浮动和零不稳定砖块；作为一个完整的流程，它在先前的砖块构建系统中达到了最佳的感知保真度。我们的结果表明，应将离散化和组装视为紧密耦合的阶段，而非独立的阶段。
cs.CV / 237 / 2608.09610
Structure-Enhanced Features and Quality-Aware Dynamic Anchor Scoring for Robust Lane Detection
结构增强特征与质量感知动态锚评分用于鲁棒车道检测
Abstract
Lane detection requires recovering thin, elongated, and frequently occluded lane structures under challenging driving conditions. While anchor-based detectors provide efficient candidate generation, their performance is limited by two coupled issues: backbone features often lose structural continuity along partially visible lanes, and classification confidence may decouple from line-level localization quality, allowing inaccurate anchors to persist before non-maximum suppression (NMS). We propose a structure-enhanced and quality-aware framework that improves lane representation and dynamic-anchor scoring while preserving the inference pipeline of the Anchor Decomposition Network (ADNet). Specifically, a Gated Horizontal-Vertical Token (GHVT) module enhances mid- and high-level backbone features via lightweight directional token interactions with a learnable residual gate. In parallel, Line-Quality-Aware Dynamic Anchor Scoring (LQAS) calibrates existing classification logits using quality supervision, hard-negative suppression, and pairwise ranking without adding inference branches. On the VIL-100 dataset, our method improves ADNet-R34 from 89.97 to 91.28 in F1 score at the 0.5 intersection-over-union threshold (F1@50), reducing both false positives and false negatives. Additional experiments on CULane and TuSimple datasets, extensive ablations, score-distribution diagnostics, and runtime analysis confirm complementary structural and ranking improvements with minimal computational overhead.
Chinese Translation
车道检测需要在复杂的驾驶条件下恢复细长且经常被遮挡的车道结构。虽然基于锚点的检测器提供了高效的候选生成，但其性能受到两个相互关联问题的限制：主干特征在部分可见车道上往往失去结构连续性，且分类置信度可能与线级定位质量脱钩，导致不准确的锚点在非极大值抑制（NMS）之前持续存在。我们提出了一种结构增强和质量感知框架，旨在改善车道表示和动态锚评分，同时保持锚分解网络（Anchor Decomposition Network，ADNet）的推理流程。具体而言，门控水平-垂直标记（Gated Horizontal-Vertical Token，GHVT）模块通过与可学习残差门的轻量级方向标记交互，增强中高层主干特征。同时，线质量感知动态锚评分（Line-Quality-Aware Dynamic Anchor Scoring，LQAS）利用质量监督、困难负样本抑制和成对排名来校准现有分类逻辑，而无需增加推理分支。在VIL-100数据集上，我们的方法将ADNet-R34的F1分数从89.97提高到91.28（在0.5交并比阈值下，F1@50），同时减少了假阳性和假阴性。在CULane和TuSimple数据集上的额外实验、广泛的消融实验、得分分布诊断和运行时分析确认了结构和排名的互补改进，且计算开销最小。
cs.CV / 238 / 2608.09613
Marrying Optimal Transport and ODEs for Unified Continuous-Time 4D Reconstruction and Tracking
将最优传输与常微分方程结合用于统一的连续时间4D重建与跟踪
Abstract
Existing unified 4D reconstruction and point tracking approaches typically rely on heuristic interpolations or just predict at integer timestamps, lacking kinematic coherence and failing to model dynamics at any arbitrary timestamp. In this paper, we propose Uni4R, a framework that unifies these tasks by learning continuous velocity fields through the synergy of Optimal Transport (OT) and Ordinary Differential Equation (ODE). Importantly, this continuous velocity field acts as a kinematic prior that mutually benefits both 4D reconstruction and point tracking. Specifically, we propose the Flow Matching Guided Decoder (FMGD). A global velocity branch first extracts anchor features that capture the global dynamic state of the sequence. Then, FMGD leverages Flow Matching (FM) theory to formulate a probability path defined by OT on the anchor feature manifold, instantiating it as FM-guided velocity features for velocity prediction. This establishes a robust kinematic inductive bias. Meanwhile, a point reconstruction branch provides geometric features. The local velocity prediction module then joint above features and time embeddings, to decode velocities at arbitrary timestamps. To overcome the absence of high-quality ground-truth velocities in fractional frames, we propose an integral-consistency training strategy. This strategy uses an ODE solver to integrate velocities to recover target pointmaps, enabling the model to be supervised end-to-end directly from integer timestamps. Experimental results demonstrate that Uni4R achieves SOTA performance in both 4D reconstruction and point tracking, and achieves SOTA in our new kinematics-aware benchmark at continuous time.
Chinese Translation
现有的统一4D重建和点跟踪方法通常依赖于启发式插值或仅在整数时间戳上进行预测，缺乏运动学一致性，无法在任意时间戳上建模动态。在本文中，我们提出了Uni4R，一个通过最优传输（Optimal Transport, OT）与常微分方程（Ordinary Differential Equation, ODE）的协同作用学习连续速度场，从而统一这些任务的框架。重要的是，这个连续速度场作为运动学先验，互惠地促进了4D重建和点跟踪的效果。具体而言，我们提出了流匹配引导解码器（Flow Matching Guided Decoder, FMGD）。一个全局速度分支首先提取锚点特征，以捕捉序列的全局动态状态。然后，FMGD利用流匹配（Flow Matching, FM）理论在锚点特征流形上制定由OT定义的概率路径，将其实例化为用于速度预测的FM引导速度特征。这建立了一个稳健的运动学归纳偏置。同时，点重建分支提供几何特征。局部速度预测模块随后将上述特征与时间嵌入结合，以解码任意时间戳的速度。为了克服在分数帧中缺乏高质量真实速度的问题，我们提出了一种积分一致性训练策略。该策略使用ODE求解器整合速度以恢复目标点图，使模型能够直接从整数时间戳进行端到端的监督。实验结果表明，Uni4R在4D重建和点跟踪方面均达到了最先进的性能，并在我们的新运动学感知基准测试中在连续时间上也达到了最先进的水平。
cs.CV / 239 / 2608.09633
LoRA-based Adaptation Alone Is Not Enough: Understanding the Limits of Foundation Models for Face Presentation Attack Detection
仅依赖于基于LoRA的适应是不够的：理解基础模型在面部呈现攻击检测中的局限性
Abstract
Face presentation attack detection (PAD) aims to reliably detect a wide range of presentation attacks. While PAD methods achieve strong performance within individual datasets, their performance degrades under cross-dataset evaluation. Variations in sensors or lighting conditions can reduce the effectiveness of detectors from near-perfect to nearly random. Foundation models (FMs) have emerged as a promising alternative because typical PAD datasets, such as the MCIO benchmarks (MSU-MFSD, CASIA-FASD, Replay-Attack, and OULU-NPU), are small relative to the scale used for web-based pretraining. However, existing PAD systems primarily focus on CLIP-based foundation models, while overlooking other FMs with different architectures and training procedures. This study addresses this question by systematically evaluating 32 FMs. Zero-shot prompting achieves performance near chance across model families and scales. The vision encoders, when low-rankadapted (LoRA) with fewer than 1% trainable weights, achieve below 2% intra-dataset ACER in most cases, while cross-dataset ACER is substantially higher. LoRA primarily refines the decision boundary within a dataset, suggesting that pretrained representations and the adaptation dataset play a larger role in cross-dataset generalization than the evaluated lightweight adaptation strategy.
Chinese Translation
面部呈现攻击检测（PAD）旨在可靠地检测各种呈现攻击。尽管PAD方法在各个数据集内表现出色，但在跨数据集评估中，其性能会下降。传感器或光照条件的变化可能会将检测器的有效性从近乎完美降至几乎随机。基础模型（FMs）因其在网络预训练中使用的规模相对较小的典型PAD数据集（如MCIO基准测试：MSU-MFSD、CASIA-FASD、Replay-Attack和OULU-NPU）而成为一种有前景的替代方案。然而，现有的PAD系统主要集中在基于CLIP的基础模型上，而忽视了其他具有不同架构和训练程序的基础模型。本研究通过系统评估32个基础模型来解决这一问题。零样本提示在模型家族和规模之间的性能接近随机水平。当视觉编码器经过低秩适应（LoRA）且可训练权重少于1%时，在大多数情况下，其数据集内的平均错误率（ACER）低于2%，而跨数据集的ACER则显著更高。LoRA主要在数据集内优化决策边界，这表明预训练表示和适应数据集在跨数据集泛化中扮演了比评估的轻量级适应策略更重要的角色。
cs.CV / 240 / 2608.09636
NeuroRefiner: Morphology-Aware Multi-Agent Refinement for 3D Fluorescence Microscopy Neuron Segmentation
NeuroRefiner：基于形态感知的多智能体细化用于3D荧光显微镜神经元分割
Abstract
Accurate 3D neuron segmentation in fluorescence microscopy is critical for neuroscience. However, the sparse and elongated morphology of neurons poses significant challenges to existing segmentation methods. These methods struggle to preserve both local details and global topology, leading to fragmented results. To address this, we propose NeuroRefiner, a multi-agent system that formalizes the human expert workflow involving iterative global observation and local editing. Specifically, NeuroRefiner comprises three collaborative agents dedicated to diagnosing topological errors, generating correction instructions, and validating refinement quality. To facilitate agent instruction-guided segmentation refinement, we propose TopoRefineNet, a dedicated 3D U-Net-based tool that leverages cross-modality feature fusion to generate refined masks. Through multi-round agent reasoning and voxel-level editing, NeuroRefiner produces topologically more accurate segmentations with enhanced interpretability. Experiments on the BigNeuron, CWMBS, and ZBFWB datasets demonstrate that NeuroRefiner outperforms state-of-the-art methods, notably achieving a 3.02% improvement in F1 score on the challenging ZBFWB dataset.
Chinese Translation
在荧光显微镜中，准确的3D神经元分割对神经科学至关重要。然而，神经元稀疏且细长的形态给现有的分割方法带来了重大挑战。这些方法难以同时保留局部细节和全局拓扑，导致分割结果碎片化。为了解决这个问题，我们提出了NeuroRefiner，一个多智能体系统，形式化了涉及迭代全局观察和局部编辑的人类专家工作流程。具体而言，NeuroRefiner由三个协作智能体组成，分别负责诊断拓扑错误、生成修正指令和验证细化质量。为了促进智能体指令引导的分割细化，我们提出了TopoRefineNet，这是一种基于3D U-Net的专用工具，利用跨模态特征融合生成细化的掩膜。通过多轮智能体推理和体素级编辑，NeuroRefiner产生了拓扑上更准确且可解释性更强的分割结果。在BigNeuron、CWMBS和ZBFWB数据集上的实验表明，NeuroRefiner的表现优于最先进的方法，特别是在具有挑战性的ZBFWB数据集上，F1分数提高了3.02%。
cs.CV / 241 / 2608.09637
DUET: A Diversity-Quality Duet of Distillation Experts for Two-Step Video Generation
DUET：用于两步视频生成的多样性-质量双重蒸馏专家
Abstract
Diffusion models have enabled high-quality video generation in recent years, but the high cost of iterative sampling hinders their practical deployment. Few-step distillation alleviates this cost, yet exposes a quality--diversity trade-off between its two dominant paradigms: trajectory-level distillation (e.g., sCM) favors diversity, whereas distribution-level distillation (e.g., DMD) favors quality. Targeting extreme two-step video generation, we introduce DUET, which reconciles the two paradigms through a noise-level duet of experts: an sCM expert takes the high-noise step to lay out diverse structure, and a DMD expert takes the low-noise step to refine appearance detail. Since the two experts are trained independently with their native objectives, DUET sidesteps the optimization difficulties of loss-level combinations and delivers quality and diversity jointly rather than trading one for the other. We further identify the relay interface and the high-noise stage as the remaining bottlenecks, and address them with RL-guided expert adaptation, yielding DUET+. With the Wan2.1-T2V-1.3B backbone, DUET lifts the two-step quality of sCM close to the level of DMD while retaining nearly all of its structural diversity---about twice that of DMD---and DUET+ further improves overall quality while preserving this diversity advantage. Together, these results establish noise-level expert specialization as a simple, effective paradigm for reconciling diversity and quality in two-step video generation.
Chinese Translation
近年来，扩散模型使得高质量视频生成成为可能，但迭代采样的高成本阻碍了其实际应用。少步蒸馏减轻了这一成本，但在其两种主要范式之间暴露出质量与多样性的权衡：轨迹级蒸馏（例如，sCM）更倾向于多样性，而分布级蒸馏（例如，DMD）则更注重质量。针对极端的两步视频生成，我们引入了DUET，通过噪声级专家的双重机制调和这两种范式：sCM专家在高噪声阶段构建多样化结构，而DMD专家在低噪声阶段细化外观细节。由于这两位专家是独立训练的，且各自有其原生目标，DUET避免了损失级组合的优化困难，能够共同提供质量和多样性，而不是相互取舍。我们进一步识别出中继接口和高噪声阶段作为剩余瓶颈，并通过强化学习指导的专家适应来解决这些问题，最终形成DUET+。在Wan2.1-T2V-1.3B骨干网络下，DUET将sCM的两步质量提升至接近DMD的水平，同时保留了几乎所有的结构多样性——约为DMD的两倍——而DUET+进一步提高了整体质量，同时保持了这一多样性优势。综上所述，这些结果确立了噪声级专家专业化作为调和两步视频生成中多样性与质量的简单有效范式。
cs.CV / 242 / 2608.09656
EgoHieraLoc: A Cortically Inspired Hierarchical Segmentation-Guided Framework for Egocentric Visual Query Localization
EgoHieraLoc：一种受皮层启发的分层分割引导框架用于自我中心视觉查询定位
Abstract
Visual query localization (VQL) aims to retrieve and re-localize a queried object in egocentric videos, yet remains challenging when object boundaries are ambiguous and global context cannot effectively guide fine-grained localization. Human vision handles such ambiguity through a hierarchical process: it rapidly screens foreground candidates, selectively attends to the target despite distractors, refines perception via feedback between global context and local detail, and, when a single view is unreliable, integrates evidence across viewpoints according to its credibility. Inspired by these competencies, we propose \textbf{EgoHieraLoc}, a unified framework for VQL-2D and VQL-3D. A Discriminative Parsing Module first extracts foreground-aware query representations using segmentation priors; a Query-Aware Module then performs robust target localization through discriminative correlation filtering with deformable modeling; and a Regional Adaptation Module feeds multi-scale context back into local regions to recover precise object boundaries. To extend this perceptual hierarchy to 3D localization, we introduce Geometric-Semantic Joint Confidence (GSJC), which multiplicatively couples segmentation confidence with local depth consistency, multi-view back-projection consistency, and triangulation-baseline quality, so that a viewpoint contributes to the 3D estimate only when it is credible both semantically and geometrically. Extensive experiments demonstrate state-of-the-art performance on both VQL-2D and -3D benchmarks.
Chinese Translation
视觉查询定位（Visual Query Localization, VQL）旨在检索并重新定位自我中心视频中的查询对象，但当对象边界模糊且全局上下文无法有效指导细粒度定位时，仍然具有挑战性。人类视觉通过分层过程处理这种模糊性：快速筛选前景候选，尽管存在干扰物仍选择性地关注目标，通过全局上下文与局部细节之间的反馈细化感知，并且当单视角不可靠时，根据其可信度整合多视角证据。受这些能力启发，我们提出了EgoHieraLoc，一个统一的VQL-2D和VQL-3D框架。判别解析模块首先利用分割先验提取前景感知的查询表示；查询感知模块随后通过带有可变形建模的判别相关滤波实现稳健的目标定位；区域适应模块将多尺度上下文反馈至局部区域以恢复精确的对象边界。为了将该感知层次扩展到3D定位，我们引入了几何-语义联合置信度（Geometric-Semantic Joint Confidence, GSJC），该方法将分割置信度与局部深度一致性、多视角反投影一致性及三角测量基线质量相乘耦合，使得视角仅在语义和几何上均可信时才对3D估计做出贡献。大量实验表明，本方法在VQL-2D和VQL-3D基准上均实现了最先进的性能。
cs.CV / 243 / 2608.09669
CIFA: Contextual-Intersectional Fairness Auditing for Hidden Subgroup Discovery in Face Analysis
CIFA：面部分析中隐藏子群发现的情境交叉公平性审计
Abstract
Fairness evaluation in computer vision commonly relies on aggregate accuracy and demographic subgroup analysis. However, visual models are also sensitive to contextual factors such as illumination, blur, image quality, facial accessories, and appearance attributes. These factors may interact with demographic characteristics, producing hidden subgroups in which performance degrades substantially despite strong aggregate accuracy and apparently acceptable demographic fairness. To address this, we propose the Contextual-Intersectional Fairness Auditing Framework (CIFA), a structured framework for identifying subgroup vulnerabilities arising from interactions between demographic and contextual attributes. CIFA performs demographic, contextual, and contextual-intersectional auditing, followed by worst-group discovery to identify and rank the most vulnerable attribute combinations. We evaluate CIFA on gender classification using ResNet-50 \cite{he2016deep} and ViT-B/16 \cite{dosovitskiy2020image} across FairFace \cite{Karkkainen2021}, CelebA \cite{Liu2015}, and UTKFace \cite{Zhang2017}. Our results show that aggregate accuracy and demographic-only evaluation can mask substantial contextual-intersectional disparities. We further assess several established mitigation strategies through an audit--mitigate--reaudit protocol and find that, although some worst-group disparities are reduced, no single strategy consistently eliminates them across datasets and architectures. These findings establish contextual-intersectional auditing as an important component of fairness evaluation and provide a reproducible framework for discovering, prioritizing, and reassessing hidden subgroup risks in face analysis systems.
Chinese Translation
计算机视觉中的公平性评估通常依赖于总体准确性和人口子群分析。然而，视觉模型对情境因素（如光照、模糊、图像质量、面部配饰和外观属性）也非常敏感。这些因素可能与人口特征相互作用，产生隐藏子群，在这些子群中，尽管总体准确性较强且人口公平性看似可接受，性能却显著下降。为了解决这一问题，我们提出了情境交叉公平性审计框架（CIFA），这是一个结构化框架，用于识别因人口属性和情境属性之间的相互作用而产生的子群脆弱性。CIFA执行人口、情境和情境交叉审计，随后进行最弱子群发现，以识别和排名最脆弱的属性组合。我们在使用 ResNet-50 extit{(He et al., 2016)} 和 ViT-B/16 extit{(Dosovitskiy et al., 2020)} 进行性别分类的过程中评估了 CIFA，数据集包括 FairFace extit{(Karkkainen et al., 2021)}、CelebA extit{(Liu et al., 2015)} 和 UTKFace extit{(Zhang et al., 2017)}。我们的结果表明，总体准确性和仅基于人口的评估可能掩盖显著的情境交叉差异。我们进一步通过审计-缓解-再审计协议评估了几种已建立的缓解策略，发现尽管一些最弱子群差异有所减少，但没有单一策略能够在所有数据集和架构中始终消除这些差异。这些发现确立了情境交叉审计作为公平性评估的重要组成部分，并提供了一个可重复的框架，用于发现、优先考虑和重新评估面部分析系统中的隐藏子群风险。
cs.CV / 244 / 2608.09672
MPISuperRes-PnP: A Super-Resolution Zero-Shot Plug-and-Play Reconstruction Algorithm for Magnetic Particle Imaging
MPISuperRes-PnP：一种用于磁性粒子成像的超分辨率零-shot 插件式重建算法
Abstract
Magnetic Particle Imaging (MPI) is an emerging medical imaging modality. MPI is based on the non-linear response of magnetic nanoparticles to an applied magnetic field and avoids ionizing radiation. The measured signal is the voltage induced in receive coils by the particles' response. Reconstructing the particle concentration from the signal constitutes the imaging task. Even using state-of-the-art measurement-based reconstruction, the associated spatial grid is very coarse, hence super-resolution (SR) techniques are important. In this work, we propose an approach for SR in MPI inspired by energy minimization. Different methods have been proposed for SR in MPI, ranging from upscaling of the associated system matrix to interpolation of the reconstruction. Here we incorporate SR into the reconstruction task via an energy minimization formulation. Following the plug-and-play approach to energy minimization we derive a splitting scheme and a SR method for MPI where the arising Gaussian denoising task is treated with a pre-trained learned Gaussian denoiser in a zero-shot fashion. This way, we incorporate benefits of deep learning without training and avoid the need of training data. Further, we provide a quantitative and qualitative evaluation of the proposed method. Hyper-parameter are selected via an extended parameter search. The found parameters are applied for reconstruction on real data. We show the applicability of our method on synthetic and on real data (MPIData: EquilibriumModelWithAnisotropy and 2D-OpenMPI Data). The proposed method employs a deep-learning denoiser without training -- thus it does not require presently scarcely available MPI training data. The denoiser behaves conservatively, i.e., no hallucination artifacts were observed. The SR approach is generic such that it can be applied in future MPI contexts involving different regularizers or different imaging tasks.
Chinese Translation
磁性粒子成像（MPI）是一种新兴的医学成像方式。MPI 基于磁性纳米颗粒对施加磁场的非线性响应，避免了电离辐射。测量信号是由颗粒响应在接收线圈中感应的电压。从信号中重建颗粒浓度构成了成像任务。即使使用最先进的基于测量的重建，相关的空间网格也非常粗糙，因此超分辨率（SR）技术显得尤为重要。在本研究中，我们提出了一种受能量最小化启发的 MPI 超分辨率方法。已经提出了不同的方法用于 MPI 的超分辨率，从相关系统矩阵的上采样到重建的插值。在这里，我们通过能量最小化的形式将超分辨率纳入重建任务中。遵循插件式能量最小化的方法，我们推导出了一种分裂方案和一种用于 MPI 的超分辨率方法，其中出现的高斯去噪任务以零-shot 方式通过预训练的高斯去噪器处理。通过这种方式，我们在不进行训练的情况下结合了深度学习的优势，避免了对训练数据的需求。此外，我们提供了所提方法的定量和定性评估。超参数通过扩展参数搜索进行选择。找到的参数应用于真实数据的重建。我们展示了我们的方法在合成数据和真实数据（MPIData：EquilibriumModelWithAnisotropy 和 2D-OpenMPI 数据）上的适用性。所提方法使用无训练的深度学习去噪器，因此不需要目前稀缺的 MPI 训练数据。去噪器表现得相对保守，即未观察到幻觉伪影。该超分辨率方法是通用的，可以应用于未来涉及不同正则化器或不同成像任务的 MPI 背景中。
cs.CV / 245 / 2608.09682
Thinking With Tools, Not With Pixels: Tool Calls as Text Scaffolds for Visual Reasoning
与工具思考，而非与像素思考：工具调用作为视觉推理的文本支架
Abstract
Tool-augmented vision-language models increasingly "think with images": they call crop, zoom, or code tools and reason over the returned pixels. However, recent work using blind tests, gain decompositions, and attention analyses has shown that returned images contribute little, raising the question: if pixels do not carry the gain, what does? We hypothesize that the load-bearing signal is the structured text emitted before any returned pixel arrives: tool name, coordinates, target description, and intent. This textual scaffold encodes where to look and what to find. We introduce TextCall (call-but-no-return) to test this: it keeps the scaffold but replaces returned images with the text placeholder [Image output skipped]. Three studies support the hypothesis. (i) Non-necessity of returned pixels: across LoRA, full fine-tuning, and RL, TextCall matches or exceeds full thinking-with-images; under RL it preserves tool use at the reported checkpoint, avoiding the failure mode where, under matched settings, seeing the returned image causes the model to stop calling tools and answer directly. (ii) Sufficiency of the scaffold: on matched training queries, scaffold-only input yields equivalent accuracy to returned-image input. (iii) Component specificity: decomposing the scaffold into reasoning text and spatial code shows both components contribute, with the dominant one varying by task. Together these results support the Tool-Call Scaffold Hypothesis: in current thinking-with-images distributions, the active signal is the structured text emitted at tool-call time; the returned image is a redundant carrier. TextCall preserves accuracy while reducing latency by 29-46% and eliminating tool-execution API calls. Our claims hold for current thinking-with-images benchmarks; constructing tasks where pixels are genuinely load-bearing remains an open direction.
Chinese Translation
工具增强的视觉-语言模型越来越多地“通过图像思考”：它们调用裁剪、缩放或编码工具，并在返回的像素上进行推理。然而，最近的研究通过盲测、增益分解和注意力分析表明，返回的图像贡献甚微，这引发了一个问题：如果像素不携带增益，那么是什么呢？我们假设，承载信号是返回任何像素之前发出的结构化文本：工具名称、坐标、目标描述和意图。这个文本支架编码了观察的方向和寻找的内容。我们引入了TextCall（调用但不返回）来测试这一假设：它保留了支架，但用文本占位符[图像输出被跳过]替换了返回的图像。三项研究支持这一假设。(i) 返回像素的非必要性：在LoRA、完全微调和强化学习（RL）中，TextCall的表现与完全通过图像思考相匹配或超过；在RL下，它在报告的检查点保持工具使用，避免了在匹配设置下看到返回图像导致模型停止调用工具并直接回答的失败模式。(ii) 支架的充分性：在匹配的训练查询中，仅使用支架的输入与返回图像的输入产生了等效的准确性。(iii) 组件特异性：将支架分解为推理文本和空间代码显示两个组件均有贡献，且主导组件因任务而异。这些结果共同支持了工具调用支架假设：在当前的图像思考分布中，主动信号是工具调用时发出的结构化文本；返回的图像是冗余的载体。TextCall在减少延迟29-46%的同时保持准确性，并消除了工具执行API调用。我们的主张在当前的图像思考基准测试中成立；构建像素真正承载负荷的任务仍然是一个开放的方向。
cs.CV / 246 / 2608.09691
Diffuse the object, keep its label: curating detector training data from a few unlabeled photographs via VLM-built 3D vegetation scenes
扩散物体，保留标签：通过VLM构建的3D植被场景从少量未标记照片中策划检测器训练数据
Abstract
Labeled images of small objects hidden in vegetation are scarce, and detectors trained on them generalize poorly across sites. Rather than reusing labels collected at another site, we synthesize labeled training images from a handful of unlabeled photographs of the deployment site itself. A vision--language model generates a coarse 3D vegetation scene from one photograph; placing 3D object meshes in the scene yields bounding boxes, segmentation masks, and per-instance occlusion directly from the scene geometry, without manual annotation. A lightweight adapter fine-tuned on the photographs conditions a diffusion pass that re-textures the renders, and a graded mask-lock sets how much diffusion may touch the object itself. In our runs this grade was the most influential curation choice: lightly diffusing the object improves minority-class recall over fully protecting its pixels, while unrestricted diffusion dissolves it. Trained on these images, a standard detector matched or exceeded its counterpart trained on a larger labeled dataset of real images from a different site, consistently across seeds on a humanitarian-demining benchmark; the comparison is thus unsupervised site adaptation from a handful of photographs against conventional cross-site label reuse. In our ablations the gains were largely insensitive to the photograph and crop budgets, and in-domain accuracy did not predict cross-site performance.
Chinese Translation
隐藏在植被中的小物体的标记图像稀缺，基于这些图像训练的检测器在不同地点的泛化能力较差。我们并未重用在其他地点收集的标签，而是从部署地点的少量未标记照片中合成标记训练图像。视觉-语言模型从一张照片生成粗略的3D植被场景；在场景中放置3D物体网格可以直接从场景几何中生成边界框、分割掩码和每个实例的遮挡，而无需手动标注。一个在照片上微调的轻量级适配器调节扩散过程，重新纹理化渲染结果，而分级掩码锁定设置了扩散可以触及物体本身的程度。在我们的实验中，这一分级是最具影响力的策划选择：轻度扩散物体提高了少数类的召回率，而完全保护其像素则效果不佳，而不受限制的扩散则会使物体消失。在这些图像上训练的标准检测器在一个人道主义扫雷基准测试中，与在不同地点的大型标记真实图像数据集上训练的对应检测器相比，表现相当或更好；因此，这一比较可以视为从少量照片进行的无监督地点适应，与传统的跨地点标签重用相对。在我们的消融实验中，性能提升对照片和裁剪预算的敏感性较低，且领域内的准确性并未预测跨地点的表现。
cs.CV / 247 / 2608.09723
LookAgain: Closed-Loop GUI Grounding with Visually Grounded Reflection
LookAgain：具有视觉基础反思的闭环GUI定位
Abstract
Recent graphical user interface (GUI) grounders have significantly advanced single-shot accuracy on standard benchmarks, yet their performance degrades sharply on small targets, densely packed controls and out-of-distribution interfaces. We attribute this gap to a paradigmatic limitation shared by existing approaches: none of them treats a produced coordinate as a hypothesis to be reflected upon and revised under new visual evidence. This manifests as three coupled issues: 1) Lack of post-hoc reflection. The prediction is frozen at the moment of emission, leaving no internal mechanism to challenge or refine it. 2) Visual evidence decoupled from the prediction. The auxiliary visual evidence is gathered to support the upcoming coordinate rather than to scrutinise the one already committed to. 3) Refinement over views, not over predictions. The iterative zoom-in refines the inspected region instead of inheriting a previous coordinate as a spatial prior to be corrected. In this paper, we propose LookAgain, a closed-loop GUI grounder driven by post-prediction visual reflection. LookAgain reformulates grounding as a multi-turn predict-look-again-refine process with two primitives: "locate" posts a coordinate hypothesis, renders a marker on the image and appends a local patch of the predicted region. It anchors the next reasoning step to the previous prediction as a spatial prior; "confirm" accepts or reject the hypothesis and terminates the procedure. We train the LookAgain grounder with SFT on constructed reflective trajectories as a cold start, followed by GRPO with terminal grounding correctness as the sole reward. Extensive experiments show that LookAgain consistently improves performance on both refusal-aware and general GUI grounding benchmarks, achieving state-of-the-art results. Comprehensive ablations further verify the effectiveness of the proposed framework.
Chinese Translation
近期的图形用户界面（GUI）定位方法在标准基准测试中显著提高了单次预测的准确性，但在小目标、密集控件和分布外接口上的表现却急剧下降。我们将这一差距归因于现有方法共享的范式限制：它们都没有将生成的坐标视为一个假设，以便在新的视觉证据下进行反思和修正。这表现为三个相互关联的问题：1）缺乏事后反思。预测在发出时被冻结，缺乏内部机制来质疑或完善它。2）视觉证据与预测脱钩。辅助视觉证据的收集是为了支持即将到来的坐标，而不是审视已经承诺的坐标。3）在视图上进行修正，而非在预测上进行修正。迭代放大只是在检查区域内进行细化，而不是将先前的坐标作为空间先验进行修正。在本文中，我们提出了LookAgain，一种由后预测视觉反思驱动的闭环GUI定位器。LookAgain将定位重新定义为一个多轮的预测-观察-再反思过程，包含两个基本操作：“定位”发布坐标假设，在图像上渲染标记并附加预测区域的局部补丁。它将下一个推理步骤锚定到先前的预测作为空间先验；“确认”接受或拒绝该假设并终止该过程。我们通过在构建的反思轨迹上进行SFT训练LookAgain定位器，作为冷启动，随后以终端定位正确性作为唯一奖励进行GRPO。大量实验表明，LookAgain在拒绝意识和一般GUI定位基准测试中持续提高性能，达到了最先进的结果。全面的消融实验进一步验证了所提框架的有效性。
cs.CV / 248 / 2608.09730
World Tokens: Enhancing Embodied Policies with Training-Time World Modeling
世界令牌：通过训练时世界建模增强具身策略
Abstract
Vision-language-action (VLA) models are a widely adopted paradigm for embodied policies. They excel at efficient closed-loop control but do not explicitly model how physical scenes evolve as a task unfolds. Recently emerging world-action models (WAMs) leverage pretrained video world models to capture spatiotemporal evolution, yet retaining future generation or a large video backbone in the control loop substantially increases inference cost. We introduce World Tokens, an embodied policy architecture built around a World Adapter that bridges visual-language understanding, world-dynamics modeling, and action generation. It uses world modeling during training to enhance the action policy while preserving efficient deployment. Specifically, the World Adapter transforms VLM features into a fixed set of world tokens, which condition a jointly fine-tuned future-video denoiser and simultaneously serve as the action expert's sole visual-language context. This shared conditioning allows gradients from future-video denoising to directly shape the representation used for action prediction, while exclusive routing prevents the policy from bypassing that representation. At deployment, the world-model branch is removed, leaving only the VLM, World Adapter, and action expert, with no online video-model inference. With a 2B backbone and no embodied action pretraining, World Tokens is highly competitive on LIBERO, attains the best reported averages on SIMPLER, substantially improves real-world R1 Pro success over a matched action-only baseline, and generates each action chunk at VLA-level latency.
Chinese Translation
视觉-语言-动作（VLA）模型是具身策略中广泛采用的一种范式。它们在高效闭环控制方面表现出色，但并未明确建模任务展开时物理场景的演变。最近出现的世界-动作模型（WAMs）利用预训练的视频世界模型捕捉时空演变，然而在控制环中保留未来生成或大型视频骨干网会显著增加推理成本。我们提出了世界令牌（World Tokens），这是一种围绕世界适配器（World Adapter）构建的具身策略架构，该适配器连接视觉-语言理解、世界动态建模和动作生成。它在训练过程中使用世界建模来增强动作策略，同时保持高效的部署。具体而言，世界适配器将视觉-语言模型（VLM）特征转换为一组固定的世界令牌，这些令牌为共同微调的未来视频去噪器提供条件，并同时作为动作专家唯一的视觉-语言上下文。这种共享条件使得来自未来视频去噪的梯度能够直接影响用于动作预测的表示，而独占路由则防止策略绕过该表示。在部署时，世界模型分支被移除，仅保留VLM、世界适配器和动作专家，不进行在线视频模型推理。凭借2B的骨干网和没有具身动作的预训练，世界令牌在LIBERO上表现极具竞争力，在SIMPLER上达到了最佳报告平均值，显著提高了与匹配的仅动作基线相比的真实世界R1 Pro成功率，并以VLA级别的延迟生成每个动作块。
cs.CV / 249 / 2608.09735
HandSplatter: Automated Digital Goniometry from Neural Rendering
HandSplatter：基于神经渲染的自动化数字测角法
Abstract
Hand and finger disorders are leading contributors to musculoskeletal disability, creating a clinical need for precise methods to quantify joint motion. Range of motion (ROM) serves as the metric for diagnosis, rehabilitation monitoring, and evaluating surgical outcomes. Currently, the goniometer is the standard tool for assessing finger flexion and extension. However, manual goniometry is labor-intensive and suffers from inconsistent inter-rater reliability due to variations in examiner technique. While digital alternatives exist, current software-based approaches often lack the necessary accuracy for clinical usage. To address these limitations, we present a novel pipeline for 3-D hand joint location and pose estimation using neural rendering. Unlike previous methods, our approach combines 2-D feature extraction with view synthesis to significantly improve accuracy and clinical viability. Furthermore, we introduce a discrete density hill climbing algorithm that facilitates the meaningful correction of projected landmarks in 3-D space. This system overcomes the inefficiencies of manual measurement and the inaccuracies of existing software, providing a robust tool for objective functional assessment.
Chinese Translation
手部和手指疾病是导致肌肉骨骼残疾的主要因素，临床上迫切需要精确的方法来量化关节运动。运动范围（ROM）作为诊断、康复监测和评估手术结果的指标。目前，测角器是评估手指屈伸的标准工具。然而，手动测角法劳动强度大，并且由于考官技术的差异，存在一致性差的跨评估者可靠性。虽然存在数字替代方案，但当前基于软件的方法通常缺乏临床使用所需的准确性。为了解决这些局限性，我们提出了一种新颖的管道，通过神经渲染进行三维手部关节位置和姿态估计。与以往方法不同，我们的方法结合了二维特征提取和视图合成，显著提高了准确性和临床可行性。此外，我们引入了一种离散密度爬山算法，便于在三维空间中对投影标记进行有意义的修正。该系统克服了手动测量的低效和现有软件的误差，为客观功能评估提供了一个强大的工具。
cs.CV / 250 / 2608.09752
Disentangling Co-Occurring Retinal Pathologies with Saliency-Guided Sparse Expert Routing
利用显著性引导的稀疏专家路由解构共存视网膜病理
Abstract
Retinal fundus images frequently exhibit multiple co-occurring pathologies, yet standard deep learning classifiers apply static, identical computation to every image regardless of the underlying disease distribution. We propose a novel architecture that resolves this via sparse conditional computation, pairing a Guided Context Gating (GCG) spatial attention front-end with a sparsely-routed Mixture-of-Experts (MoE) block operating over feature tokens. Crucially, this routing yields an interpretable, data-driven decomposition. Expert allocation is significantly disease-dependent (p < 0.001), with the healthy Normal state and morphologically distinct pathologies (e.g., ERM, AMD) isolating to dedicated experts. On a five-class, patient-disjoint 5-fold cross-validation benchmark, our model achieves 0.912 +/- 0.008 macro AUC and 0.653 +/- 0.014 macro F1. Furthermore, Grad-CAM++ and post-MoE t-SNE visualizations confirm that expert routing aligns with localized lesions and geometrically maps co-occurring cases between their constituent clusters, positioning sparse MoE as an interpretable approach to multi-disease retinal screening.
Chinese Translation
视网膜眼底图像常常显示出多种共存的病理特征，然而标准深度学习分类器对每个图像应用静态且相同的计算，而不考虑潜在的疾病分布。我们提出了一种新颖的架构，通过稀疏条件计算来解决这一问题，将引导上下文门控（Guided Context Gating, GCG）空间注意力前端与在特征标记上操作的稀疏路由专家混合（Mixture-of-Experts, MoE）模块相结合。关键在于，这种路由产生了可解释的数据驱动分解。专家分配显著依赖于疾病（p < 0.001），健康的正常状态与形态上不同的病理（例如，黄斑前膜（ERM）、年龄相关性黄斑变性（AMD））被分配到专门的专家。在一个五类、患者不重叠的五折交叉验证基准测试中，我们的模型达到了0.912 +/- 0.008的宏观AUC和0.653 +/- 0.014的宏观F1。此外，Grad-CAM++和后MoE的t-SNE可视化确认专家路由与局部病变相一致，并在其组成簇之间几何映射共存病例，将稀疏MoE定位为一种可解释的多疾病视网膜筛查方法。
cs.CV / 251 / 2608.09774
C$^2$A: Coupling Spatial Evidence with Clinical Priors via Co-occurrence Aware Class Attention for Multi-Label Chest X-Ray Classification
C$^2$A：通过共现感知类注意力将空间证据与临床先验结合用于多标签胸部X光分类
Abstract
Thoracic pathologies rarely occur in isolation, yet standard multi-label classifiers rely on shared global descriptors, discarding \emph{where} findings lie and \emph{how} they co-occur. We propose \textbf{C$\mathbf{^2}$A} (Co-occurrence Aware Class Attention), a classification head that explicitly couples spatial evidence with clinical priors. First, C$^2$A casts pooling as an expectation over learned per-class spatial attention maps, yielding localized descriptors for each disease. Second, it couples these descriptors via a learnable graph warm-started from empirical label co-occurrence. A single residual message-passing step shares evidence among related findings, proving to be a bounded perturbation of the identity where co-occurrence enters each logit through an explicit bilinear interaction. On CheXpert, C$^2$A achieves a superior $0.895$ macro-mean AUROC, outperforming advanced context-gating baselines. Crucially, gains concentrate on highly co-occurrent classes with ambiguous spatial evidence (rescuing Atelectasis by $+1.5$ over GCG), demonstrating the prior's regularizing effect with a negligible overhead of one linear projection and a $C\!\times\!C$ edge matrix.
Chinese Translation
胸部病理通常不会孤立发生，然而标准的多标签分类器依赖于共享的全局描述符，忽略了发现的 extit{位置}和 extit{共现方式}。我们提出了 extbf{C$ extbf{^2}$A}（共现感知类注意力），这是一种分类头，明确地将空间证据与临床先验结合。首先，C$^2$A将池化视为对学习到的每类空间注意力图的期望，从而为每种疾病生成局部描述符。其次，它通过一个可学习的图将这些描述符结合起来，该图从经验标签共现中进行热启动。单个残差消息传递步骤在相关发现之间共享证据，证明这是身份的有界扰动，其中共现通过显式的双线性交互进入每个logit。在CheXpert数据集上，C$^2$A实现了优越的$0.895$宏均值AUROC，超越了先进的上下文门控基线。重要的是，增益集中在具有模糊空间证据的高度共现类上（例如，Atelectasis的提升为$+1.5$，相较于GCG），展示了先验的正则化效应，且仅需一个线性投影和一个$C imes C$边缘矩阵的微不足道开销。
cs.CV / 252 / 2608.09782
NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge
NTIRE 2026 低光增强：黄昏牛仔挑战
Khalin, Aleksei, Ershov, Egor, Panshin, Artyom, Korchagin, Sergey, Lobarev, Georgiy, Terekhin, Arseniy, Dorogova, Sofiia, Shamsutdinov, Amir, Mamedov, Yasin, Khalfin, Bakhtiyar, Sheludko, Bogdan, Zilyaev, Emil, Banić, Nikola, Perevozchikov, Georgy, Timofte, Radu, Liu, Shuai, Zhang, Yuqian, Zhang, Lize, Huang, Yibin, Feng, Chaoyu, Wang, Luyang, Wang, Xiaotao, Zou, Dongqing, Lei, Lei, Liu, Tianli, Hao, Dejun, Lei, Chunxia, Kınlı, Furkan, Mironov, Andrei, Dikov, Alexander, Sadokhin, Aleksei, Zvorygin, Vladimir, Habarlak, Constantine, Yue, Shuwei, Mirantsov, Egor, Okunev, Daniil, Arkhipov, Dmitry, Yugay, Aleksandr, Ali, Anas M., Benjdira, Bilel, Boulila, Wadii, Zhou, Wei, Li, Linfeng, Kong, Lingdong, Tu, Jiachen, Xu, Guoyi, Jiang, Yaoxin, Liu, Jiajia, Shi, Yaokun
Abstract
This paper presents a review of the NTIRE 2026 Low-light Enhancement: Twilight Cowboy Challenge. The objective of the competition was to merge a set of misaligned smartphone images in the raw domain, captured in low-light conditions, into a single, clean image. Introduced setup simultaneously addresses two problems of low-light photography: visual degradations such as high noise and mixed scene illuminants, and the geometric inconsistencies caused by hand movement during multi-frame capture. To advance research in low-light and nighttime computational photography, a challenging dataset was collected comprising 585 real-world scenes, spanning indoor low-light and outdoor nighttime conditions, for training and benchmarking participant solutions. The competition employed a three-stage evaluation protocol: automatic validation via the CodaBench platform in stages one and two, followed by blind assessment on a private test set for the final ranking. Ten teams surpassed the established baseline, achieving improvements of up to +6.49 dB in PSNR and +0.0101 in SSIM, thereby establishing new state-of-the-art performance for burst-based low-light image enhancement. These results demonstrate significant progress in handling real-world noise, motion, and illumination variability in the low-light setting. Comprehensive results, leaderboards, and additional information are publicly available at https://nightimaging.org.
Chinese Translation
本文回顾了 NTIRE 2026 低光增强：黄昏牛仔挑战。该竞赛的目标是将一组在低光条件下捕获的未对齐智能手机图像合并为一幅单一的清晰图像。所提出的设置同时解决了低光摄影的两个问题：视觉退化（如高噪声和混合场景光源）以及在多帧捕获过程中由于手部运动造成的几何不一致性。为了推动低光和夜间计算摄影的研究，收集了一个具有挑战性的数据集，包含 585 个真实场景，涵盖室内低光和室外夜间条件，用于训练和基准测试参与者的解决方案。该竞赛采用了三阶段评估协议：在第一和第二阶段通过 CodaBench 平台进行自动验证，最后在私有测试集上进行盲评估以确定最终排名。十个团队超越了既定基线，在 PSNR 上实现了高达 +6.49 dB 的提升，在 SSIM 上实现了 +0.0101 的提升，从而为基于爆发的低光图像增强建立了新的最先进性能。这些结果表明，在低光环境中处理真实世界噪声、运动和光照变化方面取得了显著进展。综合结果、排行榜和更多信息可在 https://nightimaging.org 上公开获取。
cs.CV / 253 / 2608.09789
ADOPD: Reference-Privileged On-Policy Distillation for MLLM-Based Industrial Anomaly Detection
ADOPD：基于参考特权的在线蒸馏框架用于多模态大语言模型的工业异常检测
Abstract
Industrial anomaly detection (IAD) requires identifying fine-grained deviations from normal visual patterns. Multimodal large language models (MLLMs) can improve recognition accuracy by comparing query images with references at inference time, but these benefits rely on additional retrieval and processing. We investigate whether the benefits of reference comparison can instead be internalized in the model parameters. Access to references during training allows a reference-aware teacher to supervise a query-only student. However, the teacher may favor plausible responses based on query cues or language priors rather than valid visual information. We propose ADOPD, a reference-privileged on-policy distillation framework. The teacher evaluates student-generated rollouts under matched and mismatched references. The matched-reference teacher-to-student log-ratio defines the token-level learning direction, specifying what the student should learn. The likelihood gap between the two reference views estimates reference-specific support and calibrates the sequence-level weight. ADOPD achieves 77.31% average accuracy on the MMAD benchmark under zero-shot inference, improving the Qwen3-VL-4B backbone by 6.14 points and outperforming its one-shot setting by 2.64 points. Experiments show that ADOPD learns a fine-grained anomaly inspection strategy from reference comparison. The project will be available at https://github.com/withTai/ADOPD.
Chinese Translation
工业异常检测（IAD）需要识别与正常视觉模式的细微偏差。多模态大语言模型（MLLMs）可以通过在推理时将查询图像与参考进行比较来提高识别准确性，但这些好处依赖于额外的检索和处理。我们研究了参考比较的好处是否可以内化到模型参数中。在训练期间访问参考允许一个参考感知的教师来监督仅查询的学生。然而，教师可能会基于查询线索或语言先验偏向于合理的响应，而不是有效的视觉信息。我们提出了ADOPD，一个基于参考特权的在线蒸馏框架。教师在匹配和不匹配的参考下评估学生生成的回滚。匹配参考的教师对学生的对数比定义了令牌级学习方向，指定学生应该学习的内容。两个参考视图之间的似然差距估计了参考特定的支持并校准序列级权重。ADOPD在零-shot推理下在MMAD基准上实现了77.31%的平均准确率，提升了Qwen3-VL-4B骨干网络6.14个百分点，并且比其单次设置高出2.64个百分点。实验表明，ADOPD通过参考比较学习了一种细粒度的异常检查策略。该项目将可在https://github.com/withTai/ADOPD获取。
cs.CV / 254 / 2608.09801
Modern Backbones Improve Multi-task DETR for Mammography Classification and Lesion Localization
现代主干网络提升多任务 DETR 在乳腺X光分类和病变定位中的表现
Abstract
Joint exam-level prediction and candidate-region localization may improve the usefulness of AI support in mammography. We study this setting using a multi-task DETR framework, where shared representations support both image-level malignancy prediction and lesion localization, and evaluate its performance on OPTIMAM and a biopsy-confirmed SGM1k cohort. Across both datasets, modern backbones consistently outperformed older ResNet-style features, with ConvNeXtV2 and DINOv3 giving the strongest overall results, whereas MambaVision was less competitive. On OPTIMAM, ConvNeXtV2 achieved the best overall performance, reaching 97.96% AUC, 99.89% sensitivity, 25.08%
[email protected], and 74.38%
[email protected]. On SGM1k, DINOv3 gave the strongest overall results, with 90.97% AUC, 86.28% sensitivity, 82.00% specificity, 27.04%
[email protected], and 77.32%
[email protected]. These findings suggest that backbone quality is a critical factor in effective multi-task mammography, with ConvNeXtV2 emerging as a particularly strong and well-matched CNN backbone for mammography in this framework.
Chinese Translation
联合考试级别的预测和候选区域定位可能提高人工智能在乳腺X光检查中的实用性。我们使用多任务 DETR 框架研究这一设置，其中共享表示支持图像级恶性预测和病变定位，并在 OPTIMAM 数据集和经活检确认的 SGM1k 队列上评估其性能。在这两个数据集中，现代主干网络始终优于旧的 ResNet 风格特征，其中 ConvNeXtV2 和 DINOv3 给出了最强的整体结果，而 MambaVision 的表现则较弱。在 OPTIMAM 数据集中，ConvNeXtV2 达到了最佳整体性能，AUC 达到 97.96%，灵敏度为 99.89%，
[email protected] 为 25.08%，召回率@.25 为 74.38%。在 SGM1k 数据集中，DINOv3 给出了最强的整体结果，AUC 为 90.97%，灵敏度为 86.28%，特异性为 82.00%，
[email protected] 为 27.04%，召回率@.25 为 77.32%。这些发现表明，主干网络的质量是有效多任务乳腺X光检查的关键因素，ConvNeXtV2 在该框架中表现出特别强大且匹配良好的 CNN 主干网络。
cs.CV / 255 / 2608.09818
MedPixel: A Unified Pixel-Language Model for Medical Reasoning and Segmentation
MedPixel：用于医学推理和分割的统一像素语言模型
Abstract
Reliable medical image understanding requires models to connect clinical language and visual reasoning with pixel-level grounding. Yet medical vision-language models often lack precise localization, whereas medical segmenters typically rely on explicit target categories or precise spatial prompts. This divide is reinforced by a supervision mismatch: segmentation datasets provide precise masks but little language supervision, whereas medical vision-language data rarely pair language with dense spatial annotations. To address this gap, we present MedPixel, a unified medical pixel-language model built around a shared language--mask interface. To provide scalable supervision, we introduce MedPLG-440K, comprising approximately 440K pixel-language task samples constructed through a clinically motivated synthesis process without external LLM annotation. MedPixel is trained with joint multi-task supervised fine-tuning followed by Pixel-Level Preference Optimization, which uses ground-truth masks as offline verifiers to derive response preferences from mask quality. MedPixel supports a broad spectrum of tasks spanning explicit grounding, implicit reasoning, spatial interaction, grounded explanation, and medical VQA. Across this task spectrum, MedPixel achieves strong performance in both pixel-level prediction and response generation, together with effective zero-shot transfer to external grounding benchmarks and robustness to imperfect spatial prompts. Code and model checkpoints will be released at https://github.com/yhy-whu/Medpixel.
Chinese Translation
可靠的医学图像理解需要模型将临床语言与像素级的视觉推理相连接。然而，医学视觉语言模型往往缺乏精确的定位，而医学分割模型通常依赖于明确的目标类别或精确的空间提示。这一差距受到监督不匹配的加剧：分割数据集提供精确的掩膜但缺乏语言监督，而医学视觉语言数据则很少将语言与密集的空间注释配对。为了解决这一问题，我们提出了MedPixel，一个围绕共享语言-掩膜接口构建的统一医学像素语言模型。为了提供可扩展的监督，我们引入了MedPLG-440K，包含大约440K个通过临床驱动的合成过程构建的像素语言任务样本，而无需外部大语言模型（LLM）注释。MedPixel通过联合多任务监督微调进行训练，随后进行像素级偏好优化，利用真实掩膜作为离线验证器，从掩膜质量中推导响应偏好。MedPixel支持广泛的任务，包括显式定位、隐式推理、空间交互、基于定位的解释和医学视觉问答（VQA）。在这一任务范围内，MedPixel在像素级预测和响应生成方面均表现出色，同时有效地实现了对外部定位基准的零样本迁移，并对不完美的空间提示具有鲁棒性。代码和模型检查点将发布在 https://github.com/yhy-whu/Medpixel。
cs.CV / 256 / 2608.09842
From Diagnosis to Correction: Benchmarking and Improving Real-World Table Parsing
从诊断到修正：基准测试与改进现实世界表格解析
Abstract
Recent document parsers achieve table TEDS scores above 93 on OmniDocBench v1.6, yet community feedback and our audit reveal persistent failures on complex real-world tables. To quantify this gap, we introduce TableParseMap, a diagnostic benchmark of 916 real-world tables organized into five challenging scenarios and nine failure types. The strongest evaluated parser achieves only 85.03 TEDS, showing that aggregate benchmark scores conceal substantial weaknesses. Our analysis attributes these failures to three complementary limitations: large tables exceed the reliable processing scale of a single pass, weak or ambiguous visual cues hinder structure perception, and the reconstructed table may remain visually inconsistent with the image. We therefore propose DEC (Decompose--Enhance--Correct), a visual-consistency-guided agentic framework that improves frozen table parsers without retraining. DEC uses a general VLM as the controller: Decompose partitions large tables along structure-aware boundaries, Enhance exposes weak visual evidence and reparses transformed views, and Correct diagnoses and repairs residual errors. A Visual Consistency Gate (VC-Gate) selectively triggers intervention, while a Visual Consistency Ranker (VC-Ranker) verifies candidate updates and supports rollback without ground-truth HTML at inference time. We further derive a 1,977-table Consensus-Hard Set from 4,556 candidates through offline metrics and cross-model consensus. Across three frozen parsers, DEC improves TEDS by 1.57 points on average; on TableParseMap, gains reach 1.89 points overall, 2.62 on structural errors, and 5.66 on large tables.
Chinese Translation
最近的文档解析器在 OmniDocBench v1.6 上的表格 TEDS 分数超过 93，但社区反馈和我们的审计显示在复杂的现实世界表格上仍存在持续的失败。为了量化这一差距，我们引入了 TableParseMap，这是一个由 916 个现实世界表格组成的诊断基准，分为五个具有挑战性的场景和九种失败类型。评估的最强解析器仅获得 85.03 的 TEDS，显示出综合基准分数掩盖了实质性的弱点。我们的分析将这些失败归因于三个互补的限制：大表格超出了单次处理的可靠规模，弱或模糊的视觉线索妨碍了结构感知，重建的表格可能在视觉上与图像不一致。因此，我们提出了 DEC（Decompose--Enhance--Correct），一个以视觉一致性为指导的自主框架，能够在不重新训练的情况下改善冻结的表格解析器。DEC 使用通用的 VLM 作为控制器：Decompose 沿着结构感知的边界对大表格进行分解，Enhance 公开弱视觉证据并重新解析变换后的视图，Correct 诊断并修复残余错误。一个视觉一致性门（VC-Gate）选择性地触发干预，而一个视觉一致性排序器（VC-Ranker）验证候选更新并支持在推理时回滚而无需真实的 HTML。我们进一步通过离线指标和跨模型共识从 4,556 个候选中推导出一个 1,977 表格的共识困难集。在三个冻结的解析器中，DEC 平均提高了 1.57 分 TEDS；在 TableParseMap 上，总体增益达到 1.89 分，在结构错误上达到 2.62 分，在大表格上达到 5.66 分。
cs.CV / 257 / 2608.09873
Sci-VBench: Evaluating Knowledge- and Reasoning-Intensive Video Generation in Science Domains
Sci-VBench：评估科学领域知识与推理密集型视频生成
Abstract
We introduce Sci-VBench, a comprehensive benchmark for evaluating knowledge- and reasoning-intensive video generation across scientific domains. It contains 1,253 expert-annotated examples spanning 60 subjects across four core disciplines: Natural Science, Healthcare, Humanities & Social Sciences, and Engineering. Each example requires models to generate temporally rich videos that demand scientific reasoning and knowledge-grounded synthesis, going beyond surface-level visual plausibility. We further establish a rubric-based evaluation protocol. Our analysis shows that, under this protocol, both non-expert human evaluators and MLLM-as-Judge systems can achieve relatively high agreement with expert judgments, supporting reproducible evaluation at scale. We benchmark 16 frontier proprietary and open-source models and find that, while automatic perceptual-quality scores cluster tightly across systems, performance on Prompt Grounding and Scientific and Causal Correctness varies substantially, with a pronounced proprietary-open-source gap. These findings show that advances in visual realism have not yet translated into reliable modeling of scientific and causal dynamics.
Chinese Translation
我们介绍了Sci-VBench，这是一个用于评估科学领域知识与推理密集型视频生成的综合基准。该基准包含1,253个专家注释的示例，涵盖自然科学、医疗保健、人文与社会科学以及工程四个核心学科的60个主题。每个示例要求模型生成时间丰富的视频，这些视频需要科学推理和基于知识的综合，超越表面视觉的可信度。我们进一步建立了基于评分标准的评估协议。我们的分析表明，在该协议下，非专家人类评估者和MLLM-as-Judge系统都能与专家判断达到相对高的一致性，支持大规模可重复评估。我们对16个前沿的专有和开源模型进行了基准测试，发现虽然自动感知质量评分在各系统之间紧密聚集，但在提示基础和科学及因果正确性方面的表现差异显著，存在明显的专有与开源差距。这些发现表明，视觉现实主义的进步尚未转化为对科学和因果动态的可靠建模。
cs.CV / 258 / 2608.09880
Financial Numerical Prediction and Allocation as Token Generation
作为代币生成的金融数值预测与分配
Abstract
Financial prediction typically relies on task-specific regression, ranking, or policy heads, separating the language model from the numerical object ultimately evaluated. We investigate whether a causal language model can instead represent forecasts and decisions directly through constrained token generation. FinATOM introduces a unified, head-free interface for three-step stock-return forecasting and dynamic five-ETF allocation. The forecasting model autoregressively emits volatility-standardized return tokens and is trained with ordinal and ranking supervision followed by a one-epoch token-level policy stage. The allocation model generates normalized long-only weights; supervised fine-tuning imitates a causal mean--variance anchor, and DAPO-augmented GRPO optimizes realized 21-day Sharpe subject to anchor consistency. In 2023--2025 ETF tests, the allocation policy improves pooled gross Sharpe from 1.428 to 1.529 and net Sharpe under a 5-bp transaction-cost model from 1.394 to 1.494. The multimodal allocation input attains the highest three-period mean Sharpe of 1.540, with its clearest advantage in 2025. On FinTexTS, the SFT and policy strategies achieve 73.52\%/2.68 and 73.72\%/2.69 cumulative-return/Sharpe, respectively. These results support the feasibility of direct language-model token generation for financial numerical prediction and decision-making, while motivating broader tests across assets, regimes, and random seeds.
Chinese Translation
金融预测通常依赖于特定任务的回归、排序或策略头，将语言模型与最终评估的数值对象分开。我们探讨了因果语言模型是否可以通过受限的代币生成直接表示预测和决策。FinATOM 引入了一个统一的无头接口，用于三步股票收益预测和动态五只ETF分配。预测模型自回归地发出波动性标准化的收益代币，并通过序数和排名监督进行训练，随后进行一个周期的代币级策略阶段。分配模型生成标准化的仅多头权重；监督微调模仿因果均值-方差锚点，而 DAPO 增强的 GRPO 在锚点一致性的约束下优化实现的21天夏普比率。在2023-2025年的ETF测试中，分配策略将综合毛夏普比率从1.428提高到1.529，而在5个基点交易成本模型下的净夏普比率从1.394提高到1.494。多模态分配输入在2025年达到最高的三期均值夏普比率1.540，其优势在2025年最为明显。在 FinTexTS 上，SFT 和策略策略分别实现了73.52%/2.68和73.72%/2.69的累积收益/夏普比率。这些结果支持了直接语言模型代币生成在金融数值预测和决策中的可行性，同时激励在不同资产、市场环境和随机种子下进行更广泛的测试。
cs.CV / 259 / 2608.09887
Space-Creating versus Dead Possession: An Off-Ball Possession-Quality Index for Broadcast Football
空间创造与无效控球：一种用于广播足球的非球权控球质量指数
Abstract
Ball possession is the most-cited and most-misleading number in football: 60% recycled in one's own half is not 60% spent pinning the opponent back. Existing event-based possession-value frameworks (expected threat, VAEP, on-ball value) price on-ball actions but ignore the off-ball question a sterile possession poses: did holding the ball create space, or was the circulation dead? We answer this in two layers. First, an event-side junk-possession index prices each possession sequence by its peak threat gain under an expected-threat grid and -- after reconstructing the live scoreline to exclude lead-protecting circulation -- flags low-threat sequences in tied-or-losing states. On the 2026 FIFA World Cup (103 matches, 206 team-matches) the flag correlates negatively with points (r=-0.37) and xG difference (r=-0.51, partly index-coupled). It is not a repackaging of on-ball value: with team offensive VAEP and field tilt held fixed, the junk flag stays strongly negatively associated with points (p<0.0001, also match-clustered) while VAEP is not significant -- in this same-match (descriptive) regression it adds information beyond this on-ball action-value model. Second, for a flagged window we resolve whether it was spatially dead or space-creating by projecting broadcast video to pitch coordinates and measuring a Space-Creation Index (SCI): a net pitch-control change capturing whether the possession seized space or pushed the opponent's block back. Across 31 of 35 flagged windows from nine World Cup matches (a purposive sample), 74% are spatially non-space-creating, 19% weak progression, and 6% space-creating windows the event flag alone would score as failure -- including a side with 73% of the ball that exited on penalties (two non-creating windows). The two layers separate space-creating-but-unconverted from sterile possession, a distinction event-only on-ball value cannot make.
Chinese Translation
控球是足球中被引用最多且最具误导性的数字：在自己半场的60%控球并不等于在压制对手时的60%控球。现有的基于事件的控球价值框架（预期威胁、VAEP、球上价值）定价球上动作，但忽视了无球控球所带来的问题：控球是否创造了空间，还是控球循环是无效的？我们从两个层面回答这个问题。首先，事件侧的无效控球指数根据其在预期威胁网格下的最大威胁增益为每个控球序列定价，并在重构实时比分以排除保护领先的循环后，标记在平局或落后状态下的低威胁序列。在2026年国际足联世界杯（103场比赛，206场队伍比赛）中，该标记与得分呈负相关（r=-0.37）和预期进球差（r=-0.51，部分与指数相关）。这并不是对球上价值的重新包装：在固定球队进攻VAEP和场地倾斜的情况下，无效标记与得分仍然保持强烈的负相关（p<0.0001，且为比赛聚类）而VAEP则不显著——在同一场比赛的（描述性）回归中，它提供了超出球上动作价值模型的信息。其次，对于一个被标记的窗口，我们通过将广播视频投影到场地坐标上来判断其是空间无效还是空间创造，并测量空间创造指数（SCI）：一个净场地控制变化，捕捉控球是否占据了空间或将对手的防线推回。在来自九场世界杯比赛的35个被标记窗口中，有31个（一个有目的的样本），74%是空间非创造的，19%是弱推进，6%是空间创造的窗口，而仅凭事件标记将其评为失败——包括一个控球73%的球队在点球中出局（两个非创造窗口）。这两个层面将创造空间但未转化的控球与无效控球区分开来，这是仅基于事件的球上价值无法做到的区分。
cs.CV / 260 / 2608.09907
DistMoE: Private-data Rehearsal-free Routing in Mixture-of-Experts for Distributed Instruction Tuning
DistMoE：在分布式指令调优中实现混合专家的私有数据无重演路由
Abstract
Multimodal Large Language Models (MLLMs) have shown strong multimodal instruction-following ability, but adapting them to diverse visual-language domains typically assumes centralized data access and costly joint training. This is restrictive when data is distributed across private, domain-specific, or permission-limited clients. To this end, we propose DistMoE, a mixture-of-experts (MoE) approach for distributed visual instruction tuning. In each layer of the language decoder it augments the public feedforward network (FFN) with a client-specific private FFN expert, with the goal to acquire domain-specific knowledge. However, independent expert training causes the private FFNs to learn representation of different scale and magnitudes, making merging the experts difficult. To reduce client-specific drift, we introduce a public-anchored expert composition stage that updates only routers and lightweight private projection adapters on a mix of local client data and public data, via an isotropic regularization loss, therefore making it cross-client rehearsal-free composition. During inference, DistMoE performs modular routing over public and private experts, enabling token-wise domain composition without explicit domain labels. Experiments across diverse visual-language benchmarks show that DistMoE enables flexible expert reuse, effective domain adaptation, and competitive performance while preserving modular control over client-specific knowledge. Codes are available at https://github.com/mainaksingha01/DistMoE.
Chinese Translation
多模态大型语言模型（MLLMs）展示了强大的多模态指令跟随能力，但将其适应于多样的视觉-语言领域通常假设集中式数据访问和高成本的联合训练。这在数据分布于私有、特定领域或权限受限的客户端时显得限制性较强。为此，我们提出了DistMoE，一种用于分布式视觉指令调优的混合专家（MoE）方法。在语言解码器的每一层中，它通过客户端特定的私有前馈网络（FFN）专家增强公共前馈网络（FFN），旨在获取特定领域的知识。然而，独立的专家训练导致私有FFN学习到不同规模和幅度的表示，使得合并专家变得困难。为了减少客户端特定的漂移，我们引入了一个公共锚定的专家组合阶段，该阶段仅在混合本地客户端数据和公共数据的基础上更新路由器和轻量级私有投影适配器，使用各向同性正则化损失，从而实现跨客户端的无重演组合。在推理过程中，DistMoE在公共和私有专家之间执行模块化路由，实现了无需显式领域标签的逐令牌领域组合。在多样的视觉-语言基准测试中的实验表明，DistMoE实现了灵活的专家重用、有效的领域适应和具有竞争力的性能，同时保持对客户端特定知识的模块化控制。代码可在 https://github.com/mainaksingha01/DistMoE 获取。
cs.CV / 261 / 2608.09908
Beyond Hazard Resemblance: Contrastive Event Adjudication for Training-Free Video Anomaly Detection
超越危险相似性：无训练视频异常检测的对比事件裁决
Abstract
Video anomaly detection (VAD) aims to identify and temporally localize abnormal events in videos. Supervised methods learn anomaly decision boundaries from target-domain annotations but require substantial in-domain data. Existing training-free methods leverage the rich semantic knowledge and reasoning capabilities of pretrained models to interpret visual content, yet these capabilities do not directly define an anomaly decision criterion: richer anomaly descriptions better capture hazard resemblance without resolving abnormality. To this end, we propose Contrastive Event Adjudication for training-free Video Anomaly Detection (CEAVAD), which shifts the unit of inference from isolated anomaly concepts to falsifiable event hypotheses and establishes an inference-time explanatory boundary through the interaction between competing explanations and video evidence. Specifically, CEAVAD first uses public-safety knowledge to construct hazard-benign event contrasts, pairing each hazard mechanism with a generic normal account and a mechanism-specific benign counterpart. It then determines whether the target interval better supports a hazard explanation or its benign competitor, yielding a revisable contrastive boundary proposal for the target. Finally, CEAVAD adjudicates between the competing explanations to determine whether the hazard hypothesis survives the video evidence, supporting both temporally localized anomaly detection and evidence-grounded explanations. Experiments on three widely used VAD benchmarks demonstrate that CEAVAD achieves state-of-the-art performance under the training-free paradigm.
Chinese Translation
视频异常检测（VAD）旨在识别和时间定位视频中的异常事件。监督方法从目标领域的注释中学习异常决策边界，但需要大量的领域内数据。现有的无训练方法利用预训练模型丰富的语义知识和推理能力来解释视觉内容，但这些能力并不能直接定义异常决策标准：更丰富的异常描述更好地捕捉危险相似性，但并未解决异常性问题。为此，我们提出了无训练视频异常检测的对比事件裁决（CEAVAD），该方法将推理单元从孤立的异常概念转变为可证伪的事件假设，并通过竞争解释与视频证据之间的互动建立推理时的解释边界。具体而言，CEAVAD首先利用公共安全知识构建危险-良性事件对比，将每个危险机制与一个通用的正常描述和一个机制特定的良性对应物配对。然后，它确定目标时间段是否更支持危险解释或其良性竞争者，从而为目标生成一个可修订的对比边界提案。最后，CEAVAD在竞争解释之间进行裁决，以确定危险假设是否能经受住视频证据的考验，从而支持时间上局部的异常检测和基于证据的解释。在三个广泛使用的VAD基准上的实验表明，CEAVAD在无训练范式下实现了最先进的性能。
cs.CV / 262 / 2608.09926
Learning How the World Evolves: Extrapolative Video World Models via Latent Dynamics Reasoning
学习世界如何演变：通过潜在动态推理的外推视频世界模型
Abstract
The world evolves following its dynamics, i.e., its laws of motion. However, leading video diffusion models largely fit the pixels without modeling how the pixels transit over time. Thus, they render visually plausible frames but may not accurately obey the laws. To capture the dynamics purely from pixels, we introduce Latent Dynamics Reasoning (LDR). LDR casts the latent transition as an explicit kinematic integration, where the lower-order dynamics are integrated numerically and the model regresses only the third- and higher-order residual that drives the rollout. For this integration to extrapolate better, LDR runs it on a structured latent rather than dense convolutional features. Following PhyWorld, we validate LDR on a controlled white-box physics benchmark spanning five tasks (uniform motion, parabola, collision, bouncing, looming), focusing on out-of-distribution scenarios that reveal whether a model has truly learned the underlying dynamics. LDR extrapolates the learned dynamics far better: the gap between its in- and out-of-distribution error is over 20$\times$ smaller than the video diffusion baseline's, under both single- and joint-task training at 256$^2$ resolution, while using 26$\times$ fewer parameters and running 143$\times$ faster. LDR can even generalize under severe shift: for example, trained only on red balls moving left-to-right, it correctly predicts the motion of a blue square moving right-to-left. To our knowledge, this is the first video world model that extrapolates learned dynamics beyond its training distribution. Project page: https://lat-dyn-reason.github.io/
Chinese Translation
世界的演变遵循其动态，即运动法则。然而，现有的主要视频扩散模型在很大程度上仅仅拟合像素，而没有建模像素随时间的转变。因此，它们生成的帧在视觉上是可信的，但可能并不准确遵循运动法则。为了纯粹从像素中捕捉动态，我们引入了潜在动态推理（Latent Dynamics Reasoning, LDR）。LDR将潜在转变视为一种显式的运动学积分，其中低阶动态通过数值积分，而模型仅回归推动展开的第三阶及更高阶残差。为了使这一积分更好地进行外推，LDR在结构化的潜在空间上运行，而不是在密集的卷积特征上。根据PhyWorld，我们在一个涵盖五个任务（均匀运动、抛物线、碰撞、反弹、逼近）的受控白盒物理基准上验证了LDR，重点关注揭示模型是否真正学习到潜在动态的分布外场景。LDR在外推学习到的动态方面表现得更好：在256²分辨率下，其分布内和分布外误差之间的差距比视频扩散基线小超过20倍，同时使用的参数少26倍，运行速度快143倍。LDR甚至能够在严重偏移下进行泛化：例如，仅在红球从左到右移动的情况下训练，它能够正确预测蓝色方块从右到左的运动。据我们所知，这是第一个能够将学习到的动态外推到其训练分布之外的视频世界模型。项目页面：https://lat-dyn-reason.github.io/
cs.CV / 263 / 2608.09928
Multimodal Model Diffing for Feature Discovery and Control
多模态模型差异化用于特征发现与控制
Abstract
Multimodal Large Language Models (MLLMs) exhibit strong visual understanding, yet the internal features that cause these behaviors remain difficult to identify, audit, or control. While applicable to post-hoc inspection, hidden states that are decomposed into interpretable feature directions using sparse autoencoders (SAEs) neither readily isolate which features are changed by multimodal training, nor are they directly useful for targeted control. We introduce MMDiff, a multimodal model-diffing framework that trains multimodal SAEs and turns them into feature-level interfaces for discovering and controlling multimodal behavior. MMDiff supports three uses: (i) feature isolation, by diffing a base-LM SAE against its multimodal-adapted counterpart to identify features altered by multimodal training; (ii) task-specific feature detection, via per-token contrastive firing analysis that isolates causal features; and (iii) feature-level control, by causally removing or steering the discovered feature directions. We train multimodal SAEs for three MLLM families, LLaVA-MORE, PaliGemma 2, and InternVL3.5, and evaluate on visual-spatial understanding, multimodal safety, and OCR. MMDiff discovers sparse, causally specific features whose removal selectively degrades target behaviors by an average of 12% on spatial tasks and 17% on OCR, and reduces attack success rate by 24% on multimodal safety attacks, with no impact on VQA performance. Steering these features improves spatial and OCR accuracy by +3.6% and +1.8% on average over a standard single-layer steering baseline. These results show that multimodal SAEs can serve not only as interpretability tools, but as mechanisms for auditing, steering, and controlling MLLMs behavior toward safer and more capable generations.
Chinese Translation
多模态大型语言模型（MLLMs）展现出强大的视觉理解能力，但导致这些行为的内部特征仍然难以识别、审计或控制。虽然适用于事后检查，但通过稀疏自编码器（SAEs）分解为可解释特征方向的隐藏状态，既无法轻易隔离多模态训练所改变的特征，也不直接适用于有针对性的控制。我们提出了MMDiff，一个多模态模型差异化框架，训练多模态SAEs并将其转化为特征级接口，以发现和控制多模态行为。MMDiff支持三种使用方式：（i）特征隔离，通过将基础语言模型（base-LM）SAE与其多模态适配的对应模型进行差异化，以识别多模态训练所改变的特征；（ii）任务特定特征检测，通过逐标记对比激发分析来隔离因果特征；以及（iii）特征级控制，通过因果性地移除或引导发现的特征方向。我们为三种MLLM家族（LLaVA-MORE、PaliGemma 2和InternVL3.5）训练了多模态SAEs，并在视觉空间理解、多模态安全性和光学字符识别（OCR）上进行了评估。MMDiff发现了稀疏的、因果特定的特征，其移除在空间任务上平均使目标行为降低12%，在OCR上降低17%，并将多模态安全攻击的成功率降低24%，对视觉问答（VQA）性能没有影响。引导这些特征使空间和OCR的准确性平均提高了3.6%和1.8%，相较于标准单层引导基线。这些结果表明，多模态SAEs不仅可以作为可解释性工具，还可以作为审计、引导和控制MLLM行为的机制，以实现更安全和更强大的生成。
cs.CV / 264 / 2608.09931
Perception Before Supervision: Self-Contained Visual Distillation from Counterfactual Blind Spots
监督之前的感知：来自反事实盲点的自包含视觉蒸馏
Abstract
Self-improvement for multimodal large language models (MLLMs) is typically driven by reward-based methods that provide only coarse scalar feedback. Distillation offers a richer alternative through dense token-level supervision, but in the visual domain it usually depends on privileged context constructed using external annotations and tools, or stronger models. We introduce \textbf{CVPD} (Contrastive Counterfactual Visual Process Distillation), which, to the best of our knowledge, is the first fully self-contained framework for dense, on-policy, token-level visual self-distillation for MLLMs. CVPD identifies visual blind spots where zooming into a region changes and sharpens the model's answer distribution, while removing the same region leaves the full-image behavior largely unchanged. Such regions reveal perceptual information that the model can encode but fails to consistently utilize under full-image conditioning. We propose a three-gate Counterfactual Criterion that identifies these regions directly from the model's own responses and converts them into dense contrastive supervision for self-distillation. On Qwen3-VL-8B-Instruct, CVPD outperforms six self-evolving baselines across twelve benchmarks, including methods that rely on external GPT-4o supervision, without a single regression. It achieves gains of $+3.60$ on OCRBench, $+3.38$ on MMStar Fine-Grained Perception, and $+3.08$ on MMStar Logical Reasoning, while maintaining or improving performance on broader multimodal benchmarks.
Chinese Translation
多模态大型语言模型（MLLMs）的自我改进通常依赖于基于奖励的方法，这些方法仅提供粗略的标量反馈。蒸馏通过密集的标记级监督提供了一种更丰富的替代方案，但在视觉领域，它通常依赖于使用外部注释和工具或更强模型构建的特权上下文。我们提出了 extbf{CVPD}（对比反事实视觉过程蒸馏），据我们所知，这是第一个完全自包含的框架，用于对MLLMs进行密集的、在政策内的、标记级视觉自蒸馏。CVPD识别视觉盲点，在这些区域放大时，模型的答案分布发生变化并变得更加清晰，而移除同一区域则使全图行为基本保持不变。这些区域揭示了模型可以编码但在全图条件下未能始终利用的感知信息。我们提出了一个三门反事实标准，直接从模型自身的响应中识别这些区域，并将其转换为密集的对比监督以进行自蒸馏。在Qwen3-VL-8B-Instruct上，CVPD在十二个基准测试中超越了六个自我演化基线，包括依赖外部GPT-4o监督的方法，且没有出现任何回归。它在OCRBench上获得了$+3.60$的提升，在MMStar细粒度感知上获得了$+3.38$的提升，在MMStar逻辑推理上获得了$+3.08$的提升，同时在更广泛的多模态基准测试中保持或提高了性能。
cs.AI / 1 / 2608.07473
Towards an Argumentative Foundation for Evaluative AI
构建评估人工智能的论证基础
Abstract
Evaluative AI (EAI) has been recently proposed as a way to support human decision-making, not by producing a single recommendation, but by presenting competing hypotheses together with evidence for and against each. In this position paper, we advocate (computational) argumentation as a particularly suitable paradigm to provide a formal, computable foundation for forms of EAI that are explainable and contestable, setting the ground for a long-term research agenda towards distributed and human-centred EAI systems.
Chinese Translation
评估人工智能（Evaluative AI, EAI）最近被提出作为支持人类决策的一种方式，它不是通过产生单一推荐，而是通过呈现竞争假设以及支持和反对每个假设的证据。在这篇立场论文中，我们主张（计算）论证作为一种特别适合的范式，为可解释和可争议的EAI形式提供正式的、可计算的基础，为朝向分布式和以人为中心的EAI系统的长期研究议程奠定基础。
cs.AI / 2 / 2608.07474
Flow-by-Flow:Content-Judgment Bypass for Governing AI Output in High-Loss Domains
逐流而行：高损失领域中治理人工智能输出的内容判断绕过
Abstract
Prior work showed that human-in-the-loop oversight becomes structurally untenable in high-loss domains when AI output velocity V exceeds human cognitive capacity C_max. The operative constraint, however, is not V alone but V x L, where L denotes per-item cognitive load. L consists of triage, judgment, and response, which respond asymmetrically to AI capability improvement. Triage cost does not decline as models become more capable, because semantic indeterminacy is inherent in general-purpose design. Response cost is invariant to accuracy improvements. Only judgment cost faces downward pressure, and this pressure often operates by inducing omission rather than genuine reduction. Capability improvement therefore restructures L rather than reducing it. Governance mechanisms based on evaluating whether AI output is correct either delegate that evaluation to AI and inherit hallucination risk, or delegate it to humans and face the V x L ceiling. We propose Flow-by-Flow, a governance paradigm that controls supervisory load without evaluating content. A cognitive cost score based on formal, countable features imposes nonlinear costs on high-volume production, while an institutional capacity cap keeps processing volume within C_max. We derive four design invariants for any content-judgment-bypass exceedance pathway: no content judgment, no scalable consumption of examiner capacity, identity-bound per-application friction, and no batch clearance. One reference implementation is discussed to show that these invariants are jointly satisfiable, while its practical difficulties are explicitly acknowledged. An illustrative Monte Carlo analysis across 1,000 parameter draws suggests that composite multi-metric flow control outperforms supervision reinforcement alone in 90.8% of trials.
Chinese Translation
先前的研究表明，当人工智能输出速度 V 超过人类认知能力 C_max 时，人机协作的监督在高损失领域中变得结构上不可行。然而，操作约束不仅仅是 V，而是 V x L，其中 L 表示每个项目的认知负荷。L 包括分流、判断和响应，这些因素对人工智能能力的提升反应不对称。随着模型能力的增强，分流成本并不会下降，因为语义不确定性是通用设计中固有的。响应成本对准确性提升是恒定的。只有判断成本面临下行压力，而这种压力通常通过诱导遗漏而非真正的减少来运作。因此，能力的提升重构了 L 而不是减少它。基于评估人工智能输出是否正确的治理机制，要么将该评估委托给人工智能并承受幻觉风险，要么将其委托给人类并面临 V x L 的上限。我们提出了逐流而行（Flow-by-Flow），一种在不评估内容的情况下控制监督负荷的治理范式。基于正式的、可计数的特征的认知成本评分对高产量生产施加非线性成本，同时制度能力上限将处理量保持在 C_max 之内。我们推导出任何内容判断绕过超越路径的四个设计不变性：无内容判断、无可扩展的审查员能力消耗、身份绑定的每应用摩擦，以及无批量清理。讨论了一个参考实现，以表明这些不变性是可以共同满足的，同时明确承认其实际困难。对1,000个参数抽样的蒙特卡洛分析表明，复合多指标流控制在90.8%的试验中优于单独的监督强化。
cs.AI / 3 / 2608.07476
Determinization in Structure Theories: A Unified Framework via Closure, Comparability, and Joint Admissibility
结构理论中的确定化：通过闭包、可比性和联合可接受性构建的统一框架
Abstract
We develop a formal framework for constructing canonical interpretations from plural structure theories. A structure theory is a triple T = ({\Sigma}, A, I) consisting of a signature, axioms, and an inference policy, whose admissible interpretation family collects all globally consistent assignments of structural conclusions. We distinguish three levels of canonicalization: closure stabilization (per-seed convergence), global completion (seed-independent convergence), and determinization (a unique admissible interpretation). Non-determinism is classified into epistemic plurality (Type E) and structural plurality (Type S), with a refined Type S-strong subclass characterized by the absence of common upper bounds. Two canonicalization mechanisms arise: operator-based completion and selector-based construction. We provide sufficient structural conditions under which these mechanisms exist, and show that pure inference-based completion reduces to a saturated closure operator under positive, non-retractive rules with an additional soundness condition. For Type E theories, closure stabilization is established, while full determinization depends on a global confluence property that remains open. For Type S-strong theories, determinization is achieved via canonical selection. We further show that multi-level canonicalization forms a structurally non-commutative system via staged operators, and provide a conditional classification theorem reducing theory-intrinsic mechanisms to closure or selection. The framework also applies to LLM-assisted reasoning, where hallucination can be viewed as unsupported canonicalization.
Chinese Translation
我们开发了一个正式框架，用于从多元结构理论构建典范解释。结构理论是一个三元组 T = ({}, A, I)，由签名、公理和推理政策组成，其可接受的解释家族收集所有全局一致的结构结论赋值。我们区分三种典范化水平：闭包稳定性（逐种收敛）、全局完成（种子无关收敛）和确定化（唯一可接受的解释）。非确定性被分类为认知多样性（类型 E）和结构多样性（类型 S），其中经过细化的类型 S-强子类的特征是缺乏共同上界。出现了两种典范化机制：基于算子的完成和基于选择器的构造。我们提供了这些机制存在的充分结构条件，并表明在正的、非回撤规则下，纯推理基础的完成简化为饱和闭包算子，并附加了健全性条件。对于类型 E 理论，建立了闭包稳定性，而完全确定化依赖于一个全局合流性质，该性质仍然是开放的。对于类型 S-强理论，通过典范选择实现确定化。我们进一步表明，多层次典范化通过分阶段算子形成一个结构上非交换的系统，并提供一个条件分类定理，将理论内在机制简化为闭包或选择。该框架还适用于 LLM 辅助推理，其中幻觉可以被视为不支持的典范化。
cs.AI / 4 / 2608.07480
Emotion in an active inference model of human driving
人类驾驶的主动推理模型中的情感
Abstract
Active inference has emerged as a principled framework for modeling adaptive behavior by balancing goal-directed action with uncertainty reduction. It has been successfully applied across biological and artificial systems, including recent work on human driving. However, existing active inference models of driving have yet to address an important determinant of behavior in traffic: affective state, which significantly influences decision-making. Prior work in non-traffic domains has explored active inference agents in which emotions are represented along the axes of valence and arousal in the circumplex model. However, this work has been limited to simplified settings with discrete state spaces. In this work, we propose an expanded formulation of valence and arousal that can be extracted from a more complex active inference model of driving with continuous states. In particular, we condition affective estimates not only on the current state but also on predicted future outcomes. We evaluate the proposed approach in two interactive driving scenarios and show that the resulting emotion signals correspond to affective patterns reported in similar scenarios.
Chinese Translation
主动推理作为一种原则性框架，已被提出用于通过平衡目标导向行为与不确定性降低来建模适应性行为。它已成功应用于生物系统和人工系统，包括最近在人类驾驶方面的研究。然而，现有的驾驶主动推理模型尚未考虑交通行为中的一个重要决定因素：情感状态，它对决策有显著影响。先前在非交通领域的研究探索了情感在情感圆周模型中沿着效价和唤醒轴表示的主动推理代理。然而，这些研究仅限于具有离散状态空间的简化设置。在本研究中，我们提出了一种扩展的效价和唤醒的表述，可以从具有连续状态的更复杂的主动推理驾驶模型中提取。特别地，我们不仅基于当前状态来条件化情感估计，还基于预测的未来结果。我们在两个互动驾驶场景中评估了所提出的方法，并显示出所得到的情感信号与类似场景中报告的情感模式相对应。
cs.AI / 5 / 2608.07524
Training Variable Long Sequences with Data-Centric Parallel
使用数据驱动的并行训练可变长序列
Abstract
Training deep learning models on variable long sequences poses significant computational challenges. Existing methods force a difficult trade-off between efficiency and ease-of-use. Simple approaches use static configurations that cause workload imbalance low efficiency, while complex methods introduces significant complexity and code change for new models. To break this trade-off, we introduce Data-Centric Parallel (DCP). Its core principle is to let the data itself drive the runtime. It achieves this by dynamically adjusting direct runtime settings (e.g., parallel size, gradient accumulation, recomputation) based on each batch's sequence length. Empirical results demonstrate that our method achieves up to a 2.88$\times$ speedup on 32 H200 GPUs. Designed for generalization, it can be integrated into any model with 10 lines of code. We anticipate this simple yet effective approach will serve as a robust baseline and facilitate future advancements in distributed training for variable long sequences.
Chinese Translation
在可变长序列上训练深度学习模型面临着显著的计算挑战。现有方法在效率与易用性之间强迫进行艰难的权衡。简单的方法使用静态配置，导致工作负载不均衡和低效率，而复杂的方法则为新模型引入了显著的复杂性和代码变更。为了打破这一权衡，我们提出了数据驱动的并行（Data-Centric Parallel, DCP）。其核心原则是让数据本身驱动运行时。它通过根据每个批次的序列长度动态调整直接运行时设置（例如，平行大小、梯度累积、重计算）来实现这一点。实证结果表明，我们的方法在32个H200 GPU上实现了最高2.88倍的加速。该方法设计为具有良好的泛化能力，可以在10行代码内集成到任何模型中。我们预计这一简单而有效的方法将作为一个稳健的基线，并促进未来在可变长序列的分布式训练方面的进展。
cs.AI / 6 / 2608.07528
The Knowing-Saying Gap: When Probes See Errors that Confidence Misses
知与言的差距：当探测器发现信心忽视的错误时
Abstract
Linear probes detect corrupted context in language models with near-perfect accuracy, yet this does not translate into reliable failure prediction. The result is a dissociation with direct implications for deployment monitoring. Across multi-hop arithmetic chains, probes that detect corruption turn out to be uninformative about final answer correctness; models forced into structured confidence formats collapse to two values with indistinguishable error rates; and probe persistence across hops fails to separate correct from incorrect outcomes, refuting our pre-registered "persistence beats peak" hypothesis. This pattern of knowing but not saying generalises across model families including reasoning models. As a real-time monitor, probe-based interventions are sharply model and error-type dependent: branch-and-pick is net-positive across models and uniquely non-breaking on Llama-3.1-8B (4 rescued, 0 broken), while reprompt and replace-prior break correct traces at roughly the rate they rescue wrong ones. Probe-based monitoring is a necessary complement to verbalised confidence, but no single intervention dominates, and the deployable answer is model-aware, error-type-aware routing.
Chinese Translation
线性探测器以近乎完美的准确性检测语言模型中的损坏上下文，但这并未转化为可靠的失败预测。结果是与部署监控直接相关的解离。在多跳算术链中，检测到损坏的探测器对最终答案的正确性并没有提供有用的信息；被迫采用结构化信心格式的模型崩溃为两个值，且错误率无法区分；而探测器在多个跳跃中的持续性未能区分正确与错误的结果，驳斥了我们预注册的“持续性胜过峰值”假设。这种“知道但不说”的模式在包括推理模型在内的多个模型家族中具有普遍性。作为实时监控工具，基于探测器的干预显著依赖于模型和错误类型：在所有模型中，分支与选择的净效果为正，并且在 Llama-3.1-8B 上独特地不破坏（4 次救助，0 次破坏），而重新提示和替换先前的干预则以大约相同的速度破坏正确的轨迹，救助错误的轨迹。基于探测器的监控是对口头信心的必要补充，但没有单一的干预措施占主导地位，且可部署的答案是模型感知和错误类型感知的路由。
cs.AI / 7 / 2608.07530
NL2SHACL-Bench: A Benchmark Suite for Natural Language to SHACL Translation
NL2SHACL-Bench：自然语言到SHACL翻译的基准测试套件
Abstract
SHACL is a core technology for validating the conformance of RDF knowledge graphs (KGs). Yet, authoring SHACL shapes requires technical expertise that most domain experts lack. Translating natural language requirements into SHACL (NL2SHACL) would lower this barrier. However, there is no dedicated benchmark for NL2SHACL, and evaluating generated shapes requires methods beyond string comparison, as semantically equivalent shapes can differ in serialisation and structure. To tackle these challenges, we present NL2SHACL-Bench, a benchmark suite for natural language to SHACL translation. Using NL2SHACL-Bench, we evaluate four state-of-the-art large language models (LLMs) for this task. Our results show that current LLMs are highly capable of generating syntactically valid SHACL, but still struggle to produce semantically equivalent constraints for complex logical and structural patterns. This indicates that NL2SHACL-Bench provides a meaningful basis for measuring advances in the NL2SHACL state of the art.
Chinese Translation
SHACL是验证RDF知识图谱（KGs）符合性的一项核心技术。然而，编写SHACL形状需要的技术专长大多数领域专家并不具备。将自然语言需求翻译为SHACL（NL2SHACL）可以降低这一门槛。然而，目前尚无专门针对NL2SHACL的基准测试，评估生成的形状需要超越字符串比较的方法，因为语义上等价的形状在序列化和结构上可能存在差异。为了解决这些挑战，我们提出了NL2SHACL-Bench，一个用于自然语言到SHACL翻译的基准测试套件。通过使用NL2SHACL-Bench，我们评估了四种最先进的大型语言模型（LLMs）在此任务上的表现。我们的结果表明，当前的LLMs在生成语法上有效的SHACL方面能力强大，但在为复杂的逻辑和结构模式生成语义等价约束时仍然存在困难。这表明NL2SHACL-Bench为衡量NL2SHACL领域的进展提供了有意义的基础。
cs.AI / 8 / 2608.07532
Dynamic Coalition Formation and Communication Pricing in Skill-Based Agentic AI Systems
基于技能的自主智能体系统中的动态联盟形成与通信定价
Abstract
Modern agentic AI systems combine multiple large language model agents with heterogeneous skills, yet most architectures either fix communication in advance or allow full broadcast. Both can be inefficient because token cost, latency, redundancy, and error propagation increase with the number of active agents and communication links. We model agent selection and communication as a cooperative game with task-conditioned net utility $U(C\mid x)=V(C\mid x)-\sum_{i\in C}c_i$, separating coalition-level costs from agent activation costs. We propose a marginal-value activation rule and greedy router, extend the model to optimize communication edges with per-edge costs, and use estimated Shapley values to predict which agents are worth contacting before and during execution. We connect the problem to submodular maximization and prove two limited guarantees: a curvature-refined bound for a monotone, cardinality-constrained special case, and a tight $1/2$-approximation, with a correction for signed objectives, for an unconstrained non-monotone case via double greedy. Neither guarantee applies directly to the main router, which remains a heuristic. We also prove a Shapley-submodularity sandwich bound linking the error of marginal-value routing to a per-agent diminishing-returns quantity. In synthetic experiments, greedy routing achieves $99.5%$ of brute-force-optimal utility while activating $1.96$ of $8$ agents on average, compared with $38.8%$ for full broadcast. Performance is robust to activation cost and redundancy weight but falls to $66%$ under strong violations of submodularity or noisy value estimates. We distinguish the framework from Shapley pricing, hedonic coalition formation, and communication-graph pruning, and propose evaluation on real multi-agent LLM benchmarks.
Chinese Translation
现代自主智能体系统结合了多个具有异质技能的大型语言模型智能体，但大多数架构要么预先固定通信，要么允许完全广播。这两种方式都可能低效，因为随着活跃智能体和通信链路数量的增加，令牌成本、延迟、冗余和错误传播都会增加。我们将智能体选择和通信建模为一个以任务为条件的合作博弈，其净效用为 $U(C extmid x)=V(C extmid x)- extsum_{i extin C}c_i$，将联盟层级成本与智能体激活成本分开。我们提出了一种边际价值激活规则和贪婪路由器，将模型扩展到优化具有每条边成本的通信边，并使用估计的夏普利值来预测在执行前和执行过程中哪些智能体值得联系。我们将该问题与子模最大化联系起来，并证明了两个有限的保证：对于单调的、受限于基数的特例，提供了一个曲率精细化的界限；对于无约束的非单调情况，通过双重贪婪算法提供了一个紧的 $1/2$ 近似，并对带符号目标进行了修正。这两个保证都不直接适用于主要路由器，后者仍然是启发式的。我们还证明了一个夏普利-子模夹逼界限，将边际价值路由的误差与每个智能体的递减收益量联系起来。在合成实验中，贪婪路由在平均激活 $8$ 个智能体中的 $1.96$ 个时，达到了 $99.5 ext{%}$ 的暴力最优效用，而完全广播的效果仅为 $38.8 ext{%}$。性能对激活成本和冗余权重具有鲁棒性，但在强烈违反子模性或噪声价值估计的情况下，性能下降至 $66 ext{%}$。我们将该框架与夏普利定价、享乐联盟形成和通信图修剪区分开，并建议在真实的多智能体大型语言模型基准上进行评估。
cs.AI / 9 / 2608.07533
MetaSpace: Metamorphic Testing for Spatial Cognition in Embodied Agents
MetaSpace：用于具身智能体空间认知的变形测试
Abstract
An embodied agent is an intelligent entity that interacts with its environment through a physical body. Currently, the evaluation of embodied agents primarily relies on two paradigms: (1) manually annotated Visual Question Answering (VQA) pairs and (2) high-level task completion metrics, such as success in navigation or manipulation. The former is labor-intensive and subject to variability in annotation quality. The latter may obscure critical vulnerabilities, allowing agents to complete tasks through suboptimal means or safety violations, thereby concealing safety risks and inefficiencies. Given that spatial cognition is the cornerstone for executing embodied tasks, there is a pressing need to assess whether embodied agents possess robust spatial cognition during task execution. Inspired by metamorphic testing principles in software engineering, we propose MetaSpace, a novel framework designed to evaluate the spatial cognition of agents. By leveraging spatiotemporal multimodal states derived from real execution trajectories, MetaSpace automatically generates test cases based on predefined metamorphic relations (MRs) grounded in logical rules and physical laws. Crucially, we encode these MRs as executable rules in a logic programming language (Prolog). Violations of these relations indicate failures in spatial cognition. Our empirical evaluation across three embodied scenarios demonstrates that MetaSpace successfully detects 90,422 spatial cognition errors in state-of-the-art (SOTA) MLLM-driven agents. We introduce the Spatial Cognition (SC) score to quantify performance. Results indicate that all SOTA agents achieve average scores between 0.44 and 0.52, significantly lower than the human benchmark of 0.96.
Chinese Translation
具身智能体是通过物理身体与环境互动的智能实体。目前，具身智能体的评估主要依赖于两种范式：（1）手动标注的视觉问答（Visual Question Answering, VQA）对和（2）高层次任务完成指标，例如导航或操作的成功率。前者劳动密集且受标注质量变异的影响，后者可能掩盖关键的脆弱性，使智能体通过次优手段或安全违规完成任务，从而隐藏安全风险和低效性。鉴于空间认知是执行具身任务的基石，迫切需要评估具身智能体在任务执行过程中是否具备稳健的空间认知。受到软件工程中变形测试原则的启发，我们提出了MetaSpace，一个旨在评估智能体空间认知的新框架。通过利用从真实执行轨迹中提取的时空多模态状态，MetaSpace基于逻辑规则和物理法则生成基于预定义变形关系（Metamorphic Relations, MRs）的测试用例。重要的是，我们将这些MRs编码为逻辑编程语言（Prolog）中的可执行规则。这些关系的违反表明空间认知的失败。我们在三个具身场景中的实证评估表明，MetaSpace成功检测到90,422个空间认知错误，涉及最先进（State-of-the-Art, SOTA）的大型语言模型驱动的智能体。我们引入了空间认知（Spatial Cognition, SC）分数来量化性能。结果表明，所有SOTA智能体的平均分数在0.44到0.52之间，显著低于人类基准的0.96。
cs.AI / 10 / 2608.07538
When LLM Agents Negotiate: Private Information and Dynamic Bargaining in Supply Chains
当大型语言模型代理进行谈判时：供应链中的私人信息与动态议价
Abstract
As LLM agents move from decision support to autonomous procurement, firms need to know whether delegated negotiators create value, divide it predictably, and avoid money-losing contracts. We study this in a canonical supply chain bargaining problem: a buyer with private demand information negotiates a quantity-payment contract with an uninformed seller. We benchmark nine LLMs from OpenAI, Google, and Alibaba against a validated Perfect Bayesian Equilibrium across 9,840 LLM-to-LLM negotiations. First, capability governs value creation. Agents agree in 98.9% of negotiations and capture 95.4% of first-best surplus undiscounted, but average 2.98 rounds against the benchmark's 1.25, and this delay erodes 21-34% of surplus. Capability also governs reliability: baseline models accept individually irrational contracts in 19.2% of cases, versus 0.0-0.6% at mid-tier and flagship, making automated profit verification the binding guardrail below that threshold. Second, surplus capture is relational. Provider identity predicts who captures surplus better than capability rank: self-play buyer shares average 40% for OpenAI, 50% for Google, and 70% for Alibaba's Qwen, an ordering that survives restricted communication and no discounting. Reversing which provider sells moves the division by 7-18 percentage points, and the capable Qwen flagship is the weakest cross-family seller: vendor choice is a first-order distributional decision. Third, the prompt is a strategic lever. Delegation separates the principal's economic patience from the agent's prompted strategic patience, a free deployment choice that is the single strongest driver of surplus division (90% of explained variance). Together these establish an equilibrium-referenced audit of AI agents along three dimensions: discounted efficiency, distributional profile, and operational reliability.
Chinese Translation
随着大型语言模型（LLM）代理从决策支持转向自主采购，企业需要了解委托谈判者是否能够创造价值、可预测地分配价值，并避免损失合同。我们在一个经典的供应链谈判问题中研究这一点：一位拥有私人需求信息的买家与一位未被告知的卖家谈判数量-支付合同。我们对来自OpenAI、Google和阿里巴巴的九个LLM进行了基准测试，比较其在9,840次LLM对LLM谈判中的表现，基于经过验证的完美贝叶斯均衡。首先，能力决定价值创造。代理在98.9%的谈判中达成一致，并捕获95.4%的未折现的最佳盈余，但平均需要2.98轮，而基准为1.25轮，这一延迟侵蚀了21-34%的盈余。能力还决定了可靠性：基线模型在19.2%的情况下接受个别非理性合同，而中层和旗舰模型则为0.0-0.6%，这使得自动化利润验证成为低于该阈值的关键防线。其次，盈余捕获是关系性的。提供者身份比能力排名更能预测谁能更好地捕获盈余：自我博弈买家的平均分享率为OpenAI 40%、Google 50%和阿里巴巴的Qwen 70%，这一顺序在限制通信和无折扣的情况下依然成立。反转提供者的销售角色会使分配变化7-18个百分点，而能力强的Qwen旗舰在跨家族销售中表现最弱：供应商选择是一个一阶分配决策。第三，提示是一个战略杠杆。委托将委托人的经济耐心与代理的提示战略耐心分离，这是一种自由部署选择，是盈余分配的最强驱动因素（解释的方差达到90%）。这些因素共同建立了对AI代理在三个维度上的均衡参考审计：折现效率、分配特征和操作可靠性。
cs.AI / 11 / 2608.07540
TREAT: Evaluating Access to Formal Knowledge across Equivalent Mathematical Representations
TREAT：评估在等效数学表示之间对正式知识的访问
Abstract
AI systems increasingly operate between flexible input representations and formal objects used by downstream tools. A key challenge is recognizing when an unfamiliar formulation denotes a known formal object. We study this challenge through theorem recognition: given an equivalence-preserving transformation of a theorem condition, a model must recover the theorem identity associated with the standard statement. We introduce TREAT, a benchmark for evaluating whether large language models can recover known theorem identities from equivalence-preserving formula-level transformations. Rather than paraphrasing theorem text, TREAT changes the mathematical form of theorem conditions themselves, expressing known results through residual equations, witness statements, optimization identities, set relations, operator forms, and proof-intermediate characterizations. Starting from scraped theorem pages, we filter for entries with usable mathematical expression forms, extract canonical theorem conditions, and generate transformed variants with recorded assumptions and inverse mappings. The final corpus contains 737 theorem identities and 29,480 transformed rows. On a test panel, the best model retrieves the correct theorem identity in only 60.73% of cases. Other systems reveal different failure modes, including abstention, wrong detection, and malformed outputs. These suggest that theorem knowledge can be fragile under equivalent changes in representation. TREAT therefore provides a controlled testbed for evaluating representation-robust access to formal knowledge, with broader relevance to domains that require stable target objects, explicit equivalence relations, validation procedures, and auditable scoring.
Chinese Translation
人工智能系统越来越多地在灵活的输入表示和下游工具使用的正式对象之间运作。一个关键挑战是识别何时一个不熟悉的表述表示一个已知的正式对象。我们通过定理识别研究这一挑战：给定定理条件的等价保持变换，模型必须恢复与标准表述相关的定理身份。我们引入了TREAT，这是一个基准，用于评估大型语言模型是否能够从等价保持的公式级变换中恢复已知的定理身份。TREAT并不是对定理文本进行改述，而是改变定理条件本身的数学形式，通过残余方程、见证语句、优化身份、集合关系、运算符形式和证明中间表征来表达已知结果。从抓取的定理页面开始，我们筛选出具有可用数学表达形式的条目，提取规范定理条件，并生成带有记录假设和逆映射的变换变体。最终语料库包含737个定理身份和29,480个变换行。在测试面板上，最佳模型仅在60.73%的情况下检索到正确的定理身份。其他系统显示出不同的失败模式，包括放弃、错误检测和格式错误的输出。这些结果表明，在等效表示的变化下，定理知识可能是脆弱的。因此，TREAT提供了一个受控的测试平台，用于评估对正式知识的表示稳健访问，这对需要稳定目标对象、明确等价关系、验证程序和可审计评分的领域具有更广泛的相关性。
cs.AI / 12 / 2608.07542
An AI Scientist that Doesn't Drift: Taste, Structure, and Falsifiable Findings in a Quadruped Navigation Research Loop
不偏离的人工智能科学家：四足导航研究循环中的品味、结构与可证伪发现
Abstract
Autonomous research loops driven by large language models can run machine-learning experiments at scale but tend to drift toward local refinements of whichever metric they optimise rather than testing the hypotheses that motivate the experiments. We address this structurally and present an AI Scientist for studying generalisation in quadruped robot navigation policies in simulation. Building on the autoresearch paradigm of Karpathy, our loop adds three components: an immutable experiment card that pairs each iteration's prediction with its outcome under a fixed schema, so a falsified hypothesis cannot be retconned; specialised subagents restricted to mechanical roles; and kkanbu, a preference oracle that holds the user's research taste as a typed knowledge graph and is the only component permitted to make subjective judgements. To isolate the oracle we run the identical loop twice across eleven research streams, with and without kkanbu. Neither arm drifts: both falsify roughly three quarters of their own hypotheses, and the best trained policy comes from the oracle-less arm. What the oracle changes is direction, not score: it alone explores test-time adaptation, it authored the winning designs where its arm led, and it carried lessons across streams that the other arm repeatedly re-derived. The scaffold keeps the loop honest; kkanbu decides where it looks.
Chinese Translation
由大型语言模型驱动的自主研究循环能够大规模运行机器学习实验，但往往倾向于在优化的指标上进行局部改进，而不是测试驱动实验的假设。我们从结构上解决了这一问题，并提出了一种人工智能科学家，用于研究四足机器人导航策略在仿真中的泛化。基于Karpathy的自研究范式，我们的循环增加了三个组件：一个不可变的实验卡，将每次迭代的预测与其在固定模式下的结果配对，以确保被证伪的假设无法被重新解释；专门的子代理，限制在机械角色；以及kkanbu，一个偏好神谕，作为一个类型化知识图谱保存用户的研究品味，并且是唯一被允许做出主观判断的组件。为了隔离神谕，我们在十一条研究流中分别运行相同的循环，分别有和没有kkanbu。两个分支都没有偏离：它们大约都证伪了自己假设的四分之三，最佳训练策略来自于没有神谕的分支。神谕改变的是方向，而不是分数：它独自探索测试时适应，创作了其分支主导的获胜设计，并将经验教训跨流传递，而另一个分支则反复重新推导这些教训。这个框架保持了循环的诚实；kkanbu决定了它的探索方向。
cs.AI / 13 / 2608.07566
The Field Knows: Cross-Dimensional Geometry from Navigation to Black Holes
领域知识：从导航到黑洞的跨维几何
Abstract
We introduce a continuous metric field framework trained by a single causal contrastive loss. The framework encodes a scene into coefficients of a fixed symmetric matrix basis, assembles them into a Lie algebra element, and exponentiates the result to a Riemannian or Lorentzian metric. Across dimensions, this field discovers the full spectrum of geometric structures: from obstacle-avoiding geodesics in robot navigation across planar and manipulator configuration spaces, to event horizons of black holes in Lorentzian spacetime. Extensive zero-shot generalization studies demonstrate that the field captures transferable geometric structure rather than memorizing specific configurations. In the black hole setting, the causal loss spontaneously evolves genuine black-hole-like structures with the correct Lorentzian signature. The same loss, the same architecture, and the same training protocol produce the full range of geometric phenomena across dimensions. The field knows geometry, and geometry knows physics.
Chinese Translation
我们引入了一种通过单一因果对比损失训练的连续度量场框架。该框架将场景编码为固定对称矩阵基的系数，将其组装成李代数元素，并将结果指数化为黎曼或洛伦兹度量。在不同维度上，该场发现了几何结构的全谱：从机器人在平面和操纵器配置空间中的避障测地线，到洛伦兹时空中黑洞的事件视界。广泛的零样本泛化研究表明，该场捕捉到可转移的几何结构，而不是记忆特定的配置。在黑洞环境中，因果损失自发地演化出具有正确洛伦兹特征的真正黑洞样结构。相同的损失、相同的架构和相同的训练协议在不同维度上产生了全范围的几何现象。该场了解几何，而几何了解物理。
cs.AI / 14 / 2608.07617
TeXFix-Bench: An Empirically Grounded Multi-Format Benchmark for LLM-Based Document Source Repair
TeXFix-Bench：一个基于实证的多格式基准，用于基于LLM的文档源修复
Abstract
Scientific and technical writing depends on markup sources that must compile: LaTeX, Typst, and Markdown pipelines fail on missing delimiters, mismatched environments, broken imports, or package conflicts. Existing document-repair evaluations inject faults with ad-hoc edits that lack an empirical fault model. We present TeXFix-Bench, a multi-format benchmark for LLM-based full-source document repair grounded in a mined fault taxonomy. A Grounded-Theory study of localized hard-crash LaTeX faults from TeX Stack Exchange, GitHub commits, and package documentation (168 verified faults, dual open coding at $\kappa$=0.34) yields an 18-category taxonomy instantiated as DocMut: 48 AST-aware operators across three formats. A three-model cross-benchmark shows DocMut faults are 5.6-9.2 pp harder to repair than pattern-based mutations on the same seeds, and a real-error case study (88 mined human crashes, 67.0% repair success) brackets both synthetic sets from below. We construct 10,437 instances from 743 openly licensed seeds and evaluate seven LLMs under a fixed zero-shot protocol with provider-pinned routing, collecting 48,651 attempts at about USD 200 total inference cost. A complete 6,613-instance x 7-model balanced matrix confirms all rankings. A pinned engine gate yields a 27.5-point intention-to-treat compile spread (56.7-84.2%). Typst is markedly harder than LaTeX and Markdown. A restoration oracle over 28,129 compiling repairs shows that 13.6-18.5% of compiling repairs materially alter document text, and restoration rank diverges from compile rank: the model with the lowest compile rate restores content best among its successes. Compile success alone overstates repair quality. We release the taxonomy, DocMut, and all campaign artifacts.
Chinese Translation
科学和技术写作依赖于必须编译的标记源：LaTeX、Typst和Markdown管道在缺失分隔符、不匹配的环境、破损的导入或包冲突时会失败。现有的文档修复评估通过临时编辑注入故障，缺乏实证故障模型。我们提出了TeXFix-Bench，这是一个基于LLM的全源文档修复的多格式基准，基于挖掘的故障分类法。对TeX Stack Exchange、GitHub提交和包文档中的局部硬崩溃LaTeX故障的扎根理论研究（168个经过验证的故障，双重开放编码，$ ext{kappa}$=0.34）得出了一个18类分类法，实例化为DocMut：在三种格式中具有48个AST感知操作符。三模型交叉基准显示，DocMut故障的修复难度比在相同种子上的基于模式的变异高出5.6-9.2个百分点，且一个真实错误案例研究（88个挖掘的人为崩溃，67.0%的修复成功率）从下方限制了两个合成集。我们从743个开放许可的种子构建了10,437个实例，并在固定的零-shot协议下评估了七个LLM，采用提供者固定路由，收集了48,651次尝试，总推理成本约为200美元。一个完整的6,613实例x 7模型平衡矩阵确认了所有排名。一个固定引擎门产生了27.5分的意向治疗编译差异（56.7-84.2%）。Typst的修复难度明显高于LaTeX和Markdown。对28,129个编译修复的恢复神谕显示，13.6-18.5%的编译修复实质性改变了文档文本，恢复排名与编译排名不同：在成功案例中，编译率最低的模型在恢复内容方面表现最佳。仅凭编译成功率夸大了修复质量。我们发布了分类法、DocMut和所有活动文档。
cs.AI / 15 / 2608.07621
CMU-Drive and V2V-VLA: Cooperative Multi-agent Unified Driving with Reasoning Benchmark and Vehicle-to-Vehicle Vision-Language-Action Models
CMU-Drive 和 V2V-VLA：带有推理基准的协作多智能体统一驾驶与车对车视觉-语言-行动模型
Abstract
Vision-Language-Action (VLA) models have recently achieved impressive performance for end-to-end autonomous driving, yet existing approaches are primarily designed for an individual single autonomous driving agent with limited support for cooperative perception, reasoning, and planning. We present Cooperative Multi-agent Unified Driving with Reasoning (CMU-Drive), a closed-loop end-to-end benchmark for evaluating cooperative autonomous driving with multiple connected autonomous vehicles (CAVs) operating in safety-critical driving scenarios with background traffic participants. We further propose Vehicle-to-Vehicle Vision-Language-Action (V2V-VLA), a cooperative VLA model that integrates cooperative driving into a single forward pass by jointly generating driving actions, future waypoints, language reasoning, and communication policies. Experiments on CMU-Drive establish the first benchmark and baseline for cooperative VLA driving and provide a foundation for future research on multi-agent, closed-loop, end-to-end cooperative autonomous driving. Our code, benchmark, and model checkpoint will be publicly released to facilitate open-source research.
Chinese Translation
视觉-语言-行动（VLA）模型最近在端到端自主驾驶方面取得了令人瞩目的表现，但现有的方法主要针对单一自主驾驶智能体设计，对协作感知、推理和规划的支持有限。我们提出了带有推理的协作多智能体统一驾驶（CMU-Drive），这是一个闭环端到端基准，用于评估多个连接自主车辆（CAVs）在安全关键驾驶场景中与背景交通参与者协作的自主驾驶能力。我们进一步提出了车对车视觉-语言-行动（V2V-VLA），这是一个将协作驾驶整合到单次前向推理中的协作VLA模型，通过共同生成驾驶动作、未来路径点、语言推理和通信策略。CMU-Drive上的实验建立了协作VLA驾驶的第一个基准和基线，为未来多智能体、闭环、端到端的协作自主驾驶研究奠定了基础。我们的代码、基准和模型检查点将公开发布，以促进开源研究。
cs.AI / 16 / 2608.07622
Controlled Memory Interference in Continual LLM Agents
持续大规模语言模型代理中的受控记忆干扰
Abstract
Long-term memory enables AI agents to maintain continuity across sessions, personalize behavior, and evolve through accumulated experience. Yet memory evolution is not simply a process of storing more information: new experiences may reinforce, revise, or interfere with existing memory states. Existing systems mainly emphasize memory construction and relevance-based retrieval, but several memories may remain simultaneously relevant while differing in state, temporal validity, or authority. We introduce Controlled Memory Interference (CMI), a controlled diagnostic and data-generation framework for studying how agent memory evolves under different memory relationships. Across controlled memory evolution, benign accumulation has limited effects, whereas relationship-specific interference sharply suppresses update plasticity with little stability gain, either by blocking target-memory exposure or by disrupting its downstream use. Lexical and Dense retrieval exhibit distinct interference pathways, while poisoning is more sensitive to update-authority cues than to recency alone. Beyond diagnosis, CMI provides targeted examples for interference-aware memory learning, improving the distinction between valid updates and interference-inducing memories while preserving performance on original memory tasks. These findings show that memory evolution is shaped not only by memory scale, but also by interactions among accumulated experiences. More broadly, memory interference emerges as an important factor for reliable continual agent memory systems.
Chinese Translation
长期记忆使人工智能代理能够在会话之间保持连续性、个性化行为，并通过积累的经验不断演变。然而，记忆的演变并不仅仅是存储更多信息的过程：新的体验可能会强化、修正或干扰现有的记忆状态。现有系统主要强调记忆构建和基于相关性的检索，但多个记忆可能同时保持相关性，同时在状态、时间有效性或权威性上存在差异。我们提出了受控记忆干扰（Controlled Memory Interference, CMI），这是一个受控的诊断和数据生成框架，用于研究代理记忆在不同记忆关系下的演变。在受控的记忆演变过程中，良性的积累效果有限，而特定关系的干扰则显著抑制更新的可塑性，几乎没有稳定性提升，这可能是通过阻止目标记忆的暴露或干扰其下游使用来实现的。词汇检索和密集检索表现出不同的干扰路径，而中毒对更新权威线索的敏感性高于对最近性线索的敏感性。除了诊断，CMI还提供了针对干扰感知记忆学习的有针对性的示例，提高了有效更新与引发干扰的记忆之间的区分，同时保持了在原始记忆任务上的表现。这些发现表明，记忆的演变不仅受记忆规模的影响，还受到积累经验之间相互作用的影响。更广泛地说，记忆干扰成为可靠的持续代理记忆系统的重要因素。
cs.AI / 17 / 2608.07627
From Single Chatbots to Governed Agent Ecosystems: An Agentic AI Pattern Catalogue and Orchestration Framework for Mission-Critical Hospital Information Management Systems
从单一聊天机器人到治理代理生态系统：面向关键任务医院信息管理系统的代理人工智能模式目录和编排框架
Abstract
Hospitals are racing to embed AI, while coping with the surge in adaptation of the technology in other industries, into the triage management, documentation, scheduling, and revenue-cycle workflows, yet most deployments remain as fragmented pilots that stall at the edge of production, exposing patients and institutions to operational fragility, ungoverned risk, and mounting technical debt. At the same time, the global AI-in-healthcare market is projected to exceed nearly USD 1 trillion by 2034, according to the report of Fortune Business Insights, amplifying the financial consequences of architectural missteps and failed scaling strategies. This research proposes a compliance-first Agentic AI pattern catalogue and orchestration framework, purposely built for HIMS, moving beyond the single LLM chatbots and towards a governed ecosystem of autonomous and semi-autonomous agents. The framework extends by adding (i) a taxonomy of Agentic roles, (ii) a formal risk-stratification model that maps each pattern to risk tiers, human-in-the-loop checkpoints, and governance hooks, and (iii) a unified orchestration runtime capable of coordinating multi-agent workflows across EHR/HIMS landscapes such as Epic, Cerner, and MEDITECH. Technically the framework combines vLLM-based inference, optimized paging memory, confidential computing, and MCP based on-premise deployment, enforcing end-to-end encryption and policy-as-code controls aligned with HIPAA, GDPR, the EU AI Act, India's DPDP and DISHA Acts, ISO 27001, ISO 27002, ISO 14971 and IEC 62304. We exhibit how the proposed architecture is capable and efficient to reduce the documentation time, integration effort, and AI pilot attrition while constricting the governance and auditability, offering hospital leaders and governing authorities an urgently needed blueprint to convert AI investment into sustainable clinical, operational, and financial ROI
Chinese Translation
医院正在加速嵌入人工智能（AI），同时应对其他行业技术适应性激增的挑战，涉及分诊管理、文档处理、调度和收入周期工作流。然而，大多数部署仍然是零散的试点项目，停滞在生产边缘，导致患者和机构面临操作脆弱性、无治理风险和不断增加的技术债务。与此同时，根据《财富商业洞察》报告，全球医疗保健领域的人工智能市场预计到2034年将超过近1万亿美元，放大了架构失误和失败扩展策略的财务后果。本研究提出了一种以合规为首的代理人工智能模式目录和编排框架，专门为医院信息管理系统（HIMS）构建，超越单一的大型语言模型（LLM）聊天机器人，朝向一个自主和半自主代理的治理生态系统。该框架通过添加（i）代理角色的分类法，（ii）一个正式的风险分层模型，将每种模式映射到风险层级、人机交互检查点和治理钩子，以及（iii）一个统一的编排运行时，能够协调跨电子健康记录（EHR）/医院信息管理系统（HIMS）环境（如Epic、Cerner和MEDITECH）的多代理工作流。技术上，该框架结合了基于虚拟大型语言模型（vLLM）的推理、优化的分页内存、保密计算和基于MCP的本地部署，强制执行端到端加密和与HIPAA、GDPR、欧盟人工智能法案、印度DPDP和DISHA法案、ISO 27001、ISO 27002、ISO 14971和IEC 62304一致的政策即代码控制。我们展示了所提架构在减少文档时间、集成工作量和人工智能试点流失方面的能力和效率，同时收紧治理和可审计性，为医院领导和治理机构提供了一个迫切需要的蓝图，以将人工智能投资转化为可持续的临床、运营和财务投资回报。
cs.AI / 18 / 2608.07637
Agent-MD: Selective LLM Intervention with Event-Driven Escalation for Stateful GCMC--MD Campaigns
Agent-MD：具有事件驱动升级的状态性 GCMC-MD 活动的选择性 LLM 干预
Abstract
Long-running molecular simulation campaigns require repeated continuation from saved states, provenance-aware progression, adaptive assessment, and occasional interpretation of workflow conditions that cannot be resolved safely by fixed rules. Here, we present Agent-MD, a framework that places large language model (LLM) reasoning selectively at campaign construction and event-triggered review, while routine simulation, analysis, continuation, archiving, and state progression are handled by a persistent rule-based campaign agent using approved policies and explicit state records. Agent-MD was demonstrated in a grand canonical Monte Carlo-molecular dynamics (GCMC-MD) water-vapor desorption campaign comprising five montmorillonite systems and three sequential relative-humidity states (RH = 0.9-0.3-0.1). Across 15 system-RH states, the workflow completed 120 segmented simulation cycles with state-specific sampling lengths and provenance-aware restart inheritance. Routine production required no live reasoning-agent invocation, while one state reached a review boundary; two preserved incidents were subsequently evaluated through blinded reasoning-agent replay, which identified the underlying workflow problems and recommended appropriate follow-up actions. The simulations also revealed distinct composition-dependent low-RH responses, with Ca-bearing montmorillonite retaining more interlayer water and maintaining a larger basal spacing than the Na- and K-bearing systems, while the highest-charge Na system retained more residual water under dry conditions. These results demonstrate that long-running scientific workflows need not place every operation inside an LLM reasoning loop: selective reasoning can instead be combined with deterministic execution, structured evidence, and validated control handoffs to provide reproducible and auditable agent-assisted molecular simulation.
Chinese Translation
长时间运行的分子模拟活动需要从保存的状态中重复继续、关注来源的进展、自适应评估，以及偶尔解释无法通过固定规则安全解决的工作流程条件。在此，我们提出了 Agent-MD，这是一个框架，在活动构建和事件触发的审查中选择性地应用大型语言模型（LLM）推理，而常规的模拟、分析、继续、归档和状态进展则由一个使用批准政策和明确状态记录的持久规则基础活动代理处理。Agent-MD 在一个包含五个膨润土系统和三个连续相对湿度状态（RH = 0.9-0.3-0.1）的广义规范蒙特卡洛-分子动力学（GCMC-MD）水蒸气解吸活动中得到了验证。在 15 个系统-RH 状态中，工作流程完成了 120 个分段模拟周期，具有特定状态的采样长度和关注来源的重启继承。常规生产不需要实时推理代理的调用，而一个状态达到了审查边界；随后通过盲推理代理重放评估了两个保留事件，识别了潜在的工作流程问题并推荐了适当的后续行动。模拟还揭示了明显依赖成分的低相对湿度响应，含钙膨润土保留了更多的层间水并维持了比含钠和钾系统更大的基面间距，而最高电荷的钠系统在干燥条件下保留了更多的残余水。这些结果表明，长时间运行的科学工作流程不必将每个操作置于 LLM 推理循环中：选择性推理可以与确定性执行、结构化证据和验证的控制交接相结合，以提供可重复和可审计的代理辅助分子模拟。
cs.AI / 19 / 2608.07642
Contextual Value Alignment via Multilayer Combinatorial Fusion
通过多层组合融合实现上下文价值对齐
Abstract
Aligning large language models (LLMs) with human values remains a major challenge, especially for trustworthy AI. While existing approaches such as RLHF, CAI, and their variants have achieved promising results, they often rely on a single-agent framework and a unified reward system. This limits their ability to capture ethical pluralism, adapt to diverse moral contexts, and reflect the dynamics of multi-agent moral reasoning. In this work, we propose a framework that utilizes multilayer combinatorial fusion for contextual value alignment (MCF-CVA). At the first layer of the framework, it instantiates multiple moral agents, each fine-tuned to represent a distinctive value. Their outputs are then expanded combinatorially using both score- and rank-combinations as well as average and weighted aggregations. These combined models are then reduced to the same number of initial moral agents. This expansion and reduction (EAR) process continues for multi-layers until a stopping criterion is reached. The MCF-CVA framework leverages cognitive diversity between agents to mitigate conflicts and redundancies across multiple agents, producing responses that better reflect contextual human values. The framework using the EAR algorithm is performed on the dual architecture of Euclidean score space and Kemeny rank space. Empirical evaluations demonstrated that the proposed framework outperforms single-agent baselines, multi-agent single-layer results, and previous aggregation approaches on standard metrics, showing that the MCF-CVA framework provides a robust and effective mechanism for advancing contextual value alignment in LLMs.
Chinese Translation
将大型语言模型（LLMs）与人类价值观对齐仍然是一个主要挑战，尤其是在可信赖的人工智能领域。尽管现有的方法如强化学习人类反馈（RLHF）、上下文人工智能（CAI）及其变体已取得了令人鼓舞的结果，但它们通常依赖于单一代理框架和统一的奖励系统。这限制了它们捕捉伦理多元性、适应多样道德背景以及反映多代理道德推理动态的能力。在本研究中，我们提出了一种利用多层组合融合进行上下文价值对齐的框架（MCF-CVA）。该框架的第一层实例化了多个道德代理，每个代理经过微调以代表独特的价值。然后，通过得分组合和排名组合以及平均和加权聚合的方式，对它们的输出进行组合扩展。这些组合模型随后被减少到与初始道德代理相同的数量。该扩展和减少（EAR）过程在多层中继续进行，直到达到停止标准。MCF-CVA框架利用代理之间的认知多样性来缓解多个代理之间的冲突和冗余，从而生成更好地反映上下文人类价值观的响应。使用EAR算法的框架在欧几里得得分空间和凯门排名空间的双重架构上进行。实证评估表明，所提出的框架在标准指标上优于单一代理基线、多代理单层结果和先前的聚合方法，显示MCF-CVA框架为推进LLMs中的上下文价值对齐提供了一种稳健且有效的机制。
cs.AI / 20 / 2608.07645
Mendel G\"odel Machine: Recursive Self-Improving Coding Agents via Comparative Evolution
门德尔哥德尔机器：通过比较进化实现递归自我改进的编码智能体
Abstract
Self-improving coding agents that iteratively rewrite their own source code have demonstrated impressive performance on coding tasks. However, existing solutions generally derive self-modification from a single failure trajectory at a time, overlooking rich comparative signals available in the agent's expanding archive of past attempts. According to Mendelian principles of controlled inheritance, we introduce Mendel G\"odel Machine (MGM). In addition to the general single-trajectory clonal mutation, MGM includes two new types of self-modification that better utilizes evidences accumulated: the reaction-norm mutation edits an agent based on its trajectories on multiple tasks simultaneously, and the cross-lineage hybridization edits an agent using the trajectory of a reference agent from another lineage on the same task. Under an additive fitness landscape model, we prove theoretically and demonstrate via controlled surrogate simulation that the new strategies facilitate a faster and better convergence over single-trajectory baselines. Experiments on SWE-bench and Polyglot confirm MGM's consistent improvement in performance, efficiency, and generalizability.
Chinese Translation
自我改进的编码智能体能够迭代地重写自身源代码，在编码任务中表现出色。然而，现有解决方案通常仅从单一失败轨迹中推导自我修改，忽视了智能体不断扩展的过去尝试档案中丰富的比较信号。根据孟德尔控制遗传的原则，我们引入了门德尔哥德尔机器（Mendel G"odel Machine, MGM）。除了常规的单轨迹克隆突变外，MGM还包括两种新的自我修改类型，以更好地利用积累的证据：反应规范突变根据智能体在多个任务上的轨迹同时进行编辑，而跨谱系杂交则利用来自同一任务的另一谱系参考智能体的轨迹对智能体进行编辑。在加性适应度景观模型下，我们理论上证明并通过受控替代模拟展示了新策略在收敛速度和效果上优于单轨迹基线。对SWE-bench和Polyglot的实验验证了MGM在性能、效率和泛化能力上的一致提升。
cs.AI / 21 / 2608.07651
An Agentic AI Framework Overcomes Fundamental Limitations of Large Language Models for Glaucoma Detection from Fundus Photography
一种代理型人工智能框架克服了大语言模型在从眼底摄影中检测青光眼的基本局限性
Abstract
Large language models (LLMs) show promise in medical image interpretation but suffer from hallucination, limited accuracy, and run-to-run inconsistency. We developed and validated an agentic AI framework integrating LLMs with specialized deep learning tools for glaucoma detection from fundus photography. The workflow had three steps: (1) LLM initial assessment; (2) function calling to invoke specialized tools for image quality (QAModel, FundaQ-8), glaucoma classification (SwinV2-Tiny), and optic disc/cup segmentation (SegFormer-B0); and (3) LLM reflection integrating the initial impression with tool outputs. Two LLMs (Gemini 2.5 Flash, GPT-5.4 mini) were evaluated on two public datasets (ORIGA, n=100; RIM-ONE-v3, n=100) under uncropped and cropped fields of view; all images were independently graded by a masked fellowship-trained glaucoma specialist. The agentic workflow improved classification accuracy by 16 to 47 percentage points across all conditions, reaching within 6 points of the specialist; on RIM-ONE-v3 the best configurations matched the specialist accuracy of 88%. LLM-alone approaches failed in two ways: GPT-5.4 mini showed positive bias (sensitivity 95-100%, specificity 0-5%), while Gemini 2.5 Flash varied stochastically between runs; the agentic workflow corrected both. Cup-to-disc ratio error fell 15-50% (MAE 0.156-0.228 to 0.104-0.132), and correlation with specialist grading rose from weak (r=0.12-0.39) to moderate-strong (r=0.59-0.84). Run-to-run consistency rose from near-random (kappa as low as -0.01) to near-perfect (kappa up to 0.96). Integrating LLMs with specialized tools addressed key limitations of LLM-alone approaches, including over-diagnosis and run-to-run variability. Gains held for both LLMs, suggesting generalizability across backbones, and may signal a shift from monolithic models toward orchestrated multi-agent systems in medical AI.
Chinese Translation
大语言模型（LLMs）在医学图像解读中显示出潜力，但存在幻觉、准确性有限和运行间不一致等问题。我们开发并验证了一种代理型人工智能框架，将LLMs与专门的深度学习工具结合，用于从眼底摄影中检测青光眼。该工作流程分为三个步骤：（1）LLM初步评估；（2）调用专门工具进行图像质量评估（QAModel, FundaQ-8）、青光眼分类（SwinV2-Tiny）和视盘/杯分割（SegFormer-B0）；（3）LLM反思，将初步印象与工具输出整合。我们在两个公共数据集（ORIGA, n=100; RIM-ONE-v3, n=100）上评估了两种LLMs（Gemini 2.5 Flash, GPT-5.4 mini），在未裁剪和裁剪的视野下进行评估；所有图像均由经过培训的青光眼专家独立评分。代理型工作流程在所有条件下提高了分类准确性16到47个百分点，达到了与专家相差6分以内的水平；在RIM-ONE-v3上，最佳配置的准确性达到了88%，与专家相匹配。仅使用LLM的方法存在两种失败情况：GPT-5.4 mini表现出正偏差（敏感性95-100%，特异性0-5%），而Gemini 2.5 Flash在不同运行间随机变化；代理型工作流程纠正了这两种情况。杯盘比误差下降了15-50%（平均绝对误差从0.156-0.228降至0.104-0.132），与专家评分的相关性从弱（r=0.12-0.39）提高到中等-强（r=0.59-0.84）。运行间一致性从近乎随机（kappa低至-0.01）提高到近乎完美（kappa高达0.96）。将LLMs与专门工具结合解决了仅使用LLM方法的关键局限性，包括过度诊断和运行间变异性。这一改进在两种LLMs中均有效，表明在不同基础架构之间的可推广性，并可能标志着医学人工智能从单一模型向协调的多代理系统的转变。
cs.AI / 22 / 2608.07688
IntelliAudit: Using Large Language Models to Evaluate Audit Controls
IntelliAudit：利用大型语言模型评估审计控制
Abstract
IT audits require auditors to judge whether heterogeneous organizational evidence satisfies semantic security and compliance controls. This judgment is difficult to automate because relevant evidence is distributed across policies, records, spreadsheets, and operational artifacts, and because audit conclusions depend on evidentiary sufficiency rather than keyword matching. We present IntelliAudit, a retrieval-grounded multi-agent system for IT audit evidence evaluation. Given a control and an evidence corpus, IntelliAudit retrieves relevant artifacts, generates an evidence-grounded assessment, challenges adverse findings, adjudicates disagreements, and produces an auditor-facing recommendation with cited evidence, rationale, missing-evidence analysis, and remediation guidance. We instantiate IntelliAudit on ISO/IEC 27001 and evaluate it across multiple simulated organizations using expert auditor review and audit-readiness user feedback. The evaluation shows that IntelliAudit can support control interpretation, evidence-grounded reasoning, and audit-preparation workflows, while also revealing the importance of human oversight for calibrating sufficiency judgments and correcting overly permissive recommendations. These results suggest that retrieval-grounded multi-agent systems can assist audit evidence review, but should remain decision-support tools rather than autonomous certification systems.
Chinese Translation
IT审计要求审计师判断异构组织证据是否满足语义安全和合规控制。这一判断难以自动化，因为相关证据分散在政策、记录、电子表格和操作文档中，并且审计结论依赖于证据的充分性而非关键词匹配。我们提出了IntelliAudit，一个基于检索的多智能体系统，用于IT审计证据评估。给定一个控制和一个证据库，IntelliAudit检索相关文档，生成基于证据的评估，挑战不利发现，裁定争议，并提供面向审计师的建议，附带引用的证据、理由、缺失证据分析和整改指导。我们在ISO/IEC 27001上实例化IntelliAudit，并通过专家审计师评审和审计准备用户反馈在多个模拟组织中进行评估。评估结果表明，IntelliAudit可以支持控制解释、基于证据的推理和审计准备工作流程，同时也揭示了人类监督在校准充分性判断和纠正过于宽松的建议中的重要性。这些结果表明，基于检索的多智能体系统可以辅助审计证据审查，但应作为决策支持工具，而非自主认证系统。
cs.AI / 23 / 2608.07700
Towards Researcher Agents for Knowledge-Graph Question Answering
面向知识图谱问答的研究者代理
Abstract
Translating a natural-language question into a SPARQL query that can be executed against a large knowledge graph requires resolving lexical ambiguity, grounding surface terms in the target ontology, and producing graph patterns that are both syntactically valid and semantically faithful. We present an agentic text-to-SPARQL system that goes one step beyond static tool-using agents: a researcher agent that, after each round of inference on a validation set, proposes and tests changes to its own prompts, rules, and tool-orchestration code. We instantiate the loop on DBpedia, evolve nine successive versions of the agent driven by a low-cost reasoning model, and deploy the best-performing configuration with two stronger backbone models. The study yields three observations: (i) self-improvement converges quickly and then achieves 0.22 overall accuracy on the 2025 DBpedia validation set; (ii) the bottleneck is consistently in basic-graph-pattern predicate selection, not in SPARQL syntax or modifiers; and (iii) several benchmark items appear to penalise correct queries due to property ambiguity in DBpedia, suggesting that future Text-to-SPARQL benchmarks should be scored using a combination of machine translation and information retrieval metrics.
Chinese Translation
将自然语言问题翻译为可在大型知识图谱上执行的SPARQL查询，需要解决词汇歧义，将表面术语与目标本体相结合，并生成在语法上有效且在语义上忠实的图模式。我们提出了一种代理文本到SPARQL的系统，超越了静态工具使用代理：一个研究者代理，在每轮对验证集的推理后，提出并测试对其自身提示、规则和工具协调代码的更改。我们在DBpedia上实例化这一循环，基于低成本推理模型演化出九个连续版本的代理，并使用两个更强大的主干模型部署表现最佳的配置。研究得出三项观察结果：（i）自我改进快速收敛，最终在2025年DBpedia验证集上达到0.22的整体准确率；（ii）瓶颈始终在基本图模式的谓词选择，而非SPARQL语法或修饰符；（iii）由于DBpedia中的属性歧义，多个基准项目似乎对正确查询施加了惩罚，这表明未来的文本到SPARQL基准应使用机器翻译和信息检索指标的组合进行评分。
cs.AI / 24 / 2608.07705
Protecting patient privacy in clinical foundation models: Technical and legal perspectives
保护临床基础模型中的患者隐私：技术与法律视角
Abstract
Clinical foundation models trained on large-scale patient data are increasingly used for decision support, screening, and public health. As deployment expands, privacy risk increasingly arises from model-mediated leakage, yet its prevalence and severity remain poorly quantified. Models can disclose sensitive training artifacts, enabling patient re-identification in ways not captured by data-handling controls alone. Existing frameworks, including HIPAA and GDPR, offer limited guidance for such indirect threats. We propose a practical framework for assessing privacy risk in clinical foundation models and illustrate realistic leakage scenarios across deployment settings, map them to legal regimes, and outline complementary technical and legal mitigations. Our analysis provides a context-aware risk assessment grounded in realistic usage to preserve the value of medical foundation models while rigorously safeguarding patient privacy.
Chinese Translation
基于大规模患者数据训练的临床基础模型越来越多地用于决策支持、筛查和公共卫生。随着部署的扩展，模型介导的泄露带来的隐私风险日益增加，但其普遍性和严重性仍然缺乏量化。模型可能会泄露敏感的训练信息，从而以数据处理控制无法捕捉的方式实现患者的再识别。现有框架，包括《健康保险流通与问责法案》（HIPAA）和《通用数据保护条例》（GDPR），对这种间接威胁提供的指导有限。我们提出了一个评估临床基础模型隐私风险的实用框架，并展示了在不同部署环境下的现实泄露场景，将其与法律制度进行映射，并概述了互补的技术和法律缓解措施。我们的分析提供了一个基于现实使用的情境感知风险评估，以在严格保护患者隐私的同时，保留医疗基础模型的价值。
cs.AI / 25 / 2608.07743
QuantumMind: Constraint-Grounded Agentic Reasoning for Speedup Analysis in Quantum Computing
QuantumMind：基于约束的自主推理在量子计算中的加速分析
Abstract
Identifying a meaningful quantum speedup requires more than matching a classical problem to a familiar quantum primitive: the claim must preserve the task, respect access and output models, expose required promises, and remain within a defensible complexity scope. We present QuantumMind, an auditable agentic workflow for generating and conservatively screening quantum-acceleration hypotheses. A fixed sequence of typed, role-specialized actions formalizes the public task, analyzes structure and classical bottlenecks, matches a source-linked registry of quantum primitives and barriers, and constructs a scoped candidate scheme. A deterministic ten-check validator assigns the authoritative verdict; completed states are compiled into a Quantum Acceleration Evidence Graph and passed through a downward-only research screen that cannot strengthen the decision. We evaluate QuantumMind against seven task-adapted prompting and agentic controls on 582 identical open-discovery tasks. Under the frozen Open-Discovery Score (ODS), QuantumMind obtains 53.1 mean ODS, exceeding the strongest baseline by 17.3 points (48.2% relative), and wins 355 of 582 paired tasks against that baseline. It passes the graph audit on 99.8% of tasks, compared with 43.6% for the strongest baseline, and ranks first in all seven task families. The results indicate that typed state transitions and deterministic evidence control contribute beyond fluent generation alone.
Chinese Translation
识别有意义的量子加速不仅仅需要将经典问题与熟悉的量子原语匹配：这一主张必须保持任务的完整性，尊重访问和输出模型，揭示所需的承诺，并保持在可辩护的复杂性范围内。我们提出了QuantumMind，一种可审计的自主工作流程，用于生成和保守筛选量子加速假设。通过一系列固定的、类型化的、角色专门化的动作，正式化公共任务，分析结构和经典瓶颈，匹配源链接的量子原语和障碍的注册表，并构建一个范围明确的候选方案。一个确定性的十项检查验证器赋予权威的裁决；完成的状态被编译成量子加速证据图，并通过一个只能向下的研究筛选，该筛选无法加强决策。我们在582个相同的开放发现任务上评估了QuantumMind，针对七个任务适应的提示和自主控制。在冻结的开放发现评分（ODS）下，QuantumMind获得了53.1的平均ODS，超过最强基线17.3分（相对48.2%），并在582个配对任务中赢得了355个任务。它在99.8%的任务中通过了图审计，而最强基线仅为43.6%，并在所有七个任务类别中排名第一。结果表明，类型化状态转换和确定性证据控制的贡献超出了流畅生成的范畴。
cs.AI / 26 / 2608.07747
Adaptive Two-Level Allocation of a Conserved Capacity Budget Across Locations and Service Classes
跨地点和服务类别的保守容量预算的自适应双层分配
Abstract
We study how to share a single conserved capacity budget across many locations and two service classes when demand is uneven, time-varying, and can exceed supply. The shape recurs: an origin's request-rate cap split across its edge locations, a licensed throughput cap across premium and standard tenants, or an egress budget between latency-critical and batch workloads. We present a two-level algorithm. The first level redistributes capacity within a class across locations by proportional deficit and excess redistribution; the second lends capacity elastically between classes when one has surplus and the other deficit. We prove it conserves the budget exactly, preserves non-negativity, and reaches a stable allocation in one iteration under stationary demand because it carries no per-cycle state, at O(KN) cost per cycle for K classes and N locations. We evaluate it defending a CDN's per-domain budget under volumetric attack, where the classes are confirmed-legitimate and not-yet-cleared traffic; across 8 contention scenarios on a 22-location topology it serves 66-93% of high-priority demand, competitive with a single-class linear-programming optimum, while never leaving capacity idle or over-committing whenever aggregate demand meets or exceeds the budget (the contention regime these scenarios evaluate). Two findings carry beyond the application. First, a throughput-maximizing objective is wrong under contention: a two-class LP maximizing total served load serves less high-priority load than our demand-proportional, reservation-respecting allocator in most scenarios, because it cannot tell that some load it serves is the contention. Second, inter-class borrowing earns its complexity under bursty load, improving high-priority service by 1.5 points (isolated by ablation), and is neutral under stationary demand. A 5-location prototype with real HTTP traffic validates the pipeline.
Chinese Translation
我们研究如何在需求不均匀、随时间变化且可能超过供应的情况下，将单一的保守容量预算在多个地点和两个服务类别之间进行共享。该形态反复出现：一个源的请求速率上限在其边缘地点之间分配，针对优质和标准租户的许可吞吐量上限，或在延迟敏感和批处理工作负载之间的出口预算。我们提出了一种双层算法。第一层通过按比例的赤字和盈余再分配在一个类别内跨地点重新分配容量；第二层在一个类别有盈余而另一个类别有赤字时，弹性地在类别之间借用容量。我们证明该算法精确地保持预算，保持非负性，并在需求稳定的情况下在一次迭代中达到稳定分配，因为它不携带每周期状态，且每周期的成本为 O(KN)，其中 K 为类别数，N 为地点数。我们评估了它在防御 CDN 每域预算下的体积攻击中的表现，其中类别为确认合法和尚未清除的流量；在一个 22 个地点的拓扑下的 8 种争用场景中，它满足了 66-93% 的高优先级需求，与单类线性规划最优解相竞争，同时在总需求达到或超过预算时，始终不会让容量闲置或过度承诺（这些场景评估的争用机制）。两个发现超出了应用范围。首先，在争用下，最大化吞吐量的目标是错误的：在大多数场景中，最大化总服务负载的双类线性规划提供的高优先级负载少于我们按需求比例、尊重保留的分配器，因为它无法识别其服务的一部分负载是争用。其次，在突发负载下，跨类别借用展现了其复杂性，提升了高优先级服务 1.5 个点（通过消融实验隔离），而在稳定需求下则表现中性。一个包含真实 HTTP 流量的 5 个地点原型验证了该管道。
cs.AI / 27 / 2608.07762
Who Verifies the Benchmark? Decentralizing Trust in Large Language Model Evaluation
谁来验证基准？去中心化信任在大型语言模型评估中的应用
Abstract
LLM benchmarks can build an organization's reputation and attract customers, but only when results are transparent and verifiable. Unverified claims that DeepSeek R1 outperformed OpenAI's o1 contributed to market panic on January 27, 2025, when Nvidia lost USD589 billion in market value. Yet vendor benchmarks often depend on an honor system. Academic reassessments and independent leaderboards have found undisclosed changes to proprietary models, contaminated training data, and selective reporting. LLM-as-a-judge methods scale evaluation by reducing human review. Studies, however, suggest that judges may show identity-aware bias, scoring an answer according to its source model rather than its quality. This bias has not been fully measured or corrected across politically sensitive, reasoning-intensive, and preference-based tasks. We examine this problem using seven verifier models: GPT-OSS 120B, Llama 3.3 70B, GLM 5.1, Qwen3 32B, DeepSeek V4 Pro, Mistral Large3, and Sarvam M. They score anonymous and identity-disclosed responses from three primary models on 58 factual, reasoning, political, and preference-based questions. Identity disclosure slightly raises scores for factual questions, moderately affects stress-reasoning tasks, and causes large changes for geopolitically sensitive topics. Notable results include GLM5.1 (+7.00 points, p = 0.0249) and Llama 3.3 70B (+1.56 points, p = 0.00). We also introduce a blockchain-based commit-reveal protocol using Autonomous Economic Agents on an Ethereum-compatible ledger. In Phase 1, each judge records a one-way hash of its score and a secret salt before candidate identities are revealed. In Phase 2, the identity and raw score are disclosed and verified on-chain. This creates a tamper-evident audit trail that separates blind evaluation from post-hoc claims and reduces the verification burden on independent researchers and leaderboard operators.
Chinese Translation
大型语言模型（LLM）基准可以提升组织的声誉并吸引客户，但前提是结果必须透明且可验证。未经验证的声明称DeepSeek R1超越了OpenAI的o1，这导致了2025年1月27日市场恐慌，当时Nvidia的市值损失达5890亿美元。然而，供应商基准通常依赖于荣誉系统。学术重新评估和独立排行榜发现了专有模型的未披露更改、受污染的训练数据以及选择性报告。LLM作为评判者的方法通过减少人工审查来扩展评估。然而，研究表明，评判者可能表现出身份意识偏见，根据答案的源模型而非其质量来评分。这种偏见在政治敏感、推理密集和基于偏好的任务中尚未得到充分测量或纠正。我们使用七个验证模型进行此问题的研究：GPT-OSS 120B、Llama 3.3 70B、GLM 5.1、Qwen3 32B、DeepSeek V4 Pro、Mistral Large3和Sarvam M。它们对来自三个主要模型的58个事实、推理、政治和基于偏好的问题的匿名和身份披露的回答进行评分。身份披露略微提高了事实问题的分数，对压力推理任务有中等影响，而对地缘政治敏感话题则造成了较大变化。显著结果包括GLM5.1（+7.00分，p = 0.0249）和Llama 3.3 70B（+1.56分，p = 0.00）。我们还引入了一种基于区块链的提交-揭示协议，使用以太坊兼容账本上的自主经济代理。在第一阶段，每个评判者在候选人身份被揭示之前记录其分数的一次性哈希和一个秘密盐。在第二阶段，身份和原始分数在链上披露并验证。这创建了一个防篡改的审计轨迹，将盲评估与事后声明分开，并减少了独立研究人员和排行榜运营者的验证负担。
cs.AI / 28 / 2608.07775
AndroidReality: How Far Are Mobile Agents from the Real World?
AndroidReality：移动智能体与现实世界的距离有多远？
Abstract
Mobile agents have achieved promising results on clean online benchmarks such as AndroidWorld, yet their performance often degrades sharply in real-world deployment due to environmental variations and imperfect interface conditions. In this work, we introduce AndroidReality, a perturbation-based framework for evaluating and improving the robustness of mobile agents. Through a Markov Decision Process (MDP) perspective, we organize real-world interface variability into a principled taxonomy of perturbations along three axes: state, transition, and action. Guided by this taxonomy, we build a perturbed mobile benchmark on top of AndroidWorld with realistic and controllable perturbation injections, enabling systematic robustness evaluation of mobile agents. Our evaluation reveals substantial robustness gaps and four recurring error categories, motivating a simple training-free Test-Time Introspective Recovery (TTIR) mechanism that mitigates these failures on both perturbed and clean settings. Together, these results position robustness as a missing dimension in mobile agent evaluation and establish benchmark perturbation as an effective tool for both stress testing and surfacing latent weaknesses of mobile agents.
Chinese Translation
移动智能体在诸如AndroidWorld等干净的在线基准测试中取得了令人鼓舞的结果，但由于环境变化和接口条件不完善，它们的性能在现实世界的部署中往往急剧下降。在本研究中，我们提出了AndroidReality，一个基于扰动的框架，用于评估和提高移动智能体的鲁棒性。通过马尔可夫决策过程（MDP）的视角，我们将现实世界的接口变异组织成一个以状态、转移和动作三个轴为基础的原则性扰动分类法。在这一分类法的指导下，我们在AndroidWorld的基础上构建了一个带有现实且可控扰动注入的扰动移动基准，能够系统地评估移动智能体的鲁棒性。我们的评估揭示了显著的鲁棒性差距和四类重复出现的错误类别，这促使我们提出了一种简单的无训练测试时内省恢复（TTIR）机制，以减轻这些在扰动和干净设置下的失败。综合来看，这些结果将鲁棒性视为移动智能体评估中的一个缺失维度，并将基准扰动确立为一种有效的工具，用于压力测试和揭示移动智能体的潜在弱点。
cs.AI / 29 / 2608.07779
The Capability Ladder: A Curriculum-Modernization Framework for Workforce Readiness in the AI Era
能力阶梯：人工智能时代劳动力准备的课程现代化框架
Abstract
Artificial intelligence is changing the task composition of computing work faster than curricula and training typically adapt. This is a curriculum-framework paper, grounded in a structured narrative review of labor-market and software-engineering evidence and illustrated through an exploratory pilot course: the review supports the framework, and the pilot illustrates it rather than serving as primary evidence. The central claim is that near-term change is task reallocation rather than full replacement: routine implementation is increasingly automated while verification, systems thinking, security, and the ability to supervise and orchestrate AI (keeping a human in the loop) gain value. We organize the response as a capability-assurance framework anchored by a Capability Ladder: a five-level progression (trigger, automation, workflow, AI agent, agent team) that classifies the operational autonomy of AI-augmented work and the human supervision it requires. We map the ladder to course-level updates, workload-aware assessment, and stackable workforce credentials, and illustrate it through a two-semester pilot of a team-based, no-code course enrolling computing and business students. We argue for targeted modernization around durable capabilities rather than wholesale curriculum replacement, and we are explicit about evidence limits: labor signals are confounded by non-AI forces, industry reports are directional, and the pilot is exploratory.
Chinese Translation
人工智能正在以比课程和培训通常适应的速度更快地改变计算工作任务的组成。本文为课程框架论文，基于对劳动市场和软件工程证据的结构化叙述性回顾，并通过一门探索性试点课程进行说明：该回顾支持框架，而试点则用于说明而非作为主要证据。核心论点是，近期的变化是任务重新分配而非完全替代：常规实施越来越自动化，而验证、系统思维、安全性以及监督和协调人工智能的能力（保持人类在环中）则愈发重要。我们将响应组织为一个以能力阶梯为基础的能力保障框架：一个五级进阶（触发、自动化、工作流、人工智能代理、代理团队），用于分类人工智能增强工作所需的操作自主性及其所需的人类监督。我们将该阶梯映射到课程级更新、工作负载感知评估和可叠加的劳动力证书，并通过一门为计算和商业学生开设的团队基础无代码课程的两个学期试点进行说明。我们主张围绕持久能力进行有针对性的现代化，而非全面替换课程，并明确指出证据的局限性：劳动信号受到非人工智能因素的混淆，行业报告具有方向性，试点则是探索性的。
cs.AI / 30 / 2608.07786
Who Built This Model? Tracing LLM Lineage via Spectral Fingerprints in Weight Space
谁构建了这个模型？通过权重空间中的谱指纹追踪大型语言模型的谱系
Abstract
Open-weight large language models (LLMs) are increasingly developed through complex, multi-stage pipelines, leading to intricate lineage relationships that reflect model origin, ownership, and evolution. Understanding these relationships is important for model provenance, governance, and supply-chain integrity. In this work, we investigate the notion of LLM "biometrics" (analogous to human biometrics) to ask whether LLMs exhibit intrinsic fingerprints in weight space alone, without access to input data, that reveal their origin and lineage. We formulate this as a lineage discrimination problem, distinguishing among independent-origin, same-series, and shared-base models. To characterize these relationships, we propose a unified geometric fingerprinting framework that analyzes weight matrices from two complementary perspectives: (i) spectral energy, captured by singular value distributions to encode global magnitude patterns, and (ii) subspace alignment, quantified via subspace deviations to capture directional geometry. Our analysis uncovers a clear hierarchy of structural similarity in weight space: spectral energy reliably distinguishes independently trained models and different model families, while subspace alignment enables fine-grained discrimination among closely related models, including variations in dataset scale and post-training procedures. Extensive experiments on over 110 diverse open-weight LLM pairs demonstrate that weight-space geometry provides a robust and interpretable signal for model lineage, enabling coarse-grained regime separation and fine-grained discrimination within shared-base models.
Chinese Translation
开放权重的大型语言模型（LLMs）正通过复杂的多阶段流程不断发展，导致反映模型起源、所有权和演变的复杂谱系关系。理解这些关系对于模型的来源、治理和供应链的完整性至关重要。在本研究中，我们探讨了LLM“生物特征识别”（类似于人类生物特征识别）的概念，以询问LLM是否仅在权重空间中展现出内在的指纹，而无需访问输入数据，这些指纹能够揭示其起源和谱系。我们将此问题表述为谱系区分问题，区分独立起源、同系列和共享基础模型。为了表征这些关系，我们提出了一个统一的几何指纹识别框架，从两个互补的角度分析权重矩阵：（i）谱能量，通过奇异值分布捕捉全局幅度模式，（ii）子空间对齐，通过子空间偏差量化以捕捉方向几何。我们的分析揭示了权重空间中结构相似性的明确层次：谱能量可靠地区分独立训练的模型和不同的模型家族，而子空间对齐则能够在紧密相关的模型之间进行细粒度区分，包括数据集规模和后训练程序的变化。在110对多样化的开放权重LLM的广泛实验中，证明了权重空间几何提供了一个稳健且可解释的模型谱系信号，使得粗粒度的领域分离和共享基础模型内的细粒度区分成为可能。
cs.AI / 31 / 2608.07796
CliniCARE-Bench: Clinical Calibrated Audit of Medical Reasoning in EHR
CliniCARE-Bench：电子健康记录中医学推理的临床校准审计
Abstract
Large language models perform strongly on medical knowledge benchmarks, but reliable clinical deployment requires agents to conduct defensible investigations over heterogeneous, longitudinal records: determining what evidence is needed, retrieving and reconciling structured and free-text data, grounding conclusions in verifiable evidence, and deferring cases that cannot be resolved reliably. We introduce CliniCARE-Bench (Clinical Calibrated Audit of Medical Reasoning in EHR), a benchmark for retrospective clinical audit: 25 clinician-validated scenarios instantiated as 750 patient-specific cases over real-patient-derived MIMIC-IV data. Systems investigate each case through a governed, logged tool environment for record retrieval, computation, and policy access, and return one of four verdicts---Yes, No, Indeterminate: Lack of Data, or Indeterminate: Medically Ambiguous---the last two separating missing evidence from residual medical ambiguity. Beyond verdict accuracy, we score patient-evidence and policy grounding, process adherence, calibrated abstention, reliability, and efficiency against case-level reference verdicts produced by independent multi-model adjudication and calibrated against Clinical Board review. Every retrieval, computation, and report is replayable, so the investigation trace is inspectable and scorable. To our knowledge, CliniCARE-Bench is the first deployment-oriented clinical-agent benchmark to jointly evaluate real longitudinal EHR investigation, claim-level evidence grounding, governing-policy use, process adherence, and calibrated abstention within a common patient-level adjudication framework. Across 16 agentic systems, four-way accuracy spans 65.3-76.1%, but raw accuracy overstates investigation quality. Defect-free accuracy, which credits a verdict only when correct and free of prohibited shortcuts, is 4.8-14.8 points lower and reorders the leaderboard.
Chinese Translation
大型语言模型在医学知识基准测试中表现出色，但可靠的临床部署要求代理能够对异质的、纵向的记录进行可辩护的调查：确定所需证据，检索和调和结构化与自由文本数据，将结论基于可验证的证据，并推迟无法可靠解决的案例。我们介绍了CliniCARE-Bench（电子健康记录中医学推理的临床校准审计），这是一个用于回顾性临床审计的基准：25个经过临床医生验证的场景，具体化为750个基于真实患者的MIMIC-IV数据的患者特定案例。系统通过一个受管控的、记录的工具环境对每个案例进行调查，以便进行记录检索、计算和政策访问，并返回四个裁决之一——是、否、未确定：缺乏数据，或未确定：医学模糊——后两个裁决区分缺失证据与剩余的医学模糊性。除了裁决准确性外，我们还对患者证据和政策基础、过程遵循、校准弃权、可靠性和效率进行评分，这些评分是基于独立的多模型裁决产生的案例级参考裁决，并经过临床委员会审查进行校准。每次检索、计算和报告都是可重放的，因此调查轨迹是可检查和可评分的。据我们所知，CliniCARE-Bench是第一个面向部署的临床代理基准，能够在共同的患者级裁决框架内联合评估真实的纵向电子健康记录调查、索赔级证据基础、治理政策使用、过程遵循和校准弃权。在16个代理系统中，四向准确率范围为65.3%-76.1%，但原始准确率夸大了调查质量。无缺陷准确率，仅在裁决正确且没有禁止的捷径时给予积分，低于4.8-14.8个百分点，并重新排序了排行榜。
cs.AI / 32 / 2608.07809
CausalNav: Reliability-Certified Causal World Models for Control under Physical-Parameter Shift
CausalNav：针对物理参数变化的可靠性认证因果世界模型控制
Abstract
A world model is only useful for physical AI if it changes what the agent does, and only safe if it declines to do so when it is wrong. We study both halves of that requirement with CausalNav, a controller built around a signed, action-conditioned transition graph over identified state coordinates. At deployment CausalNav simulates a small library of intervention sequences, converts their objective error into policy-logit advice, and admits that advice only when a scale-free predictive-reliability certificate, a policy-margin gate, and an argmax-agreement gate all pass; otherwise it falls back exactly to its own model-based base controller. We evaluate against nine controlled baselines (transformer, recurrent, split-latent, graph, causal-induction, and three recent model-based reasoning modules) on CartPole-v1 and discretized Pendulum-v1 with physical-parameter shifts, under one shared PPO trainer, one interaction budget, and ten held-out seeds (200 runs). CausalNav attains the best average rank (1.25 of ten). The diagnostic result is more informative than the ranking: the learned graph recovers structure well above chance (CartPole F1 = 0.59 +/- 0.09), yet per-seed structural fidelity is uncorrelated with per-seed control benefit (r = -0.15, p = 0.67), and the certificate abstains on 10/10 Pendulum seeds, where forcing the planner on costs return. Model fidelity did not predict downstream control utility in our setting; certified abstention, not better prediction, is what made the world model safe to deploy.
Chinese Translation
一个世界模型只有在改变智能体的行为时对物理人工智能有用，并且只有在其判断错误时不采取行动时才是安全的。我们通过CausalNav研究了这一要求的两个方面，CausalNav是一个围绕已识别状态坐标的有符号、基于动作条件的转移图构建的控制器。在部署时，CausalNav模拟一小库干预序列，将其目标误差转换为策略对数建议，并仅在无尺度预测可靠性证书、策略边际门和argmax一致性门均通过时接受该建议；否则，它将完全回退到其基于模型的基础控制器。我们在CartPole-v1和离散化的Pendulum-v1上，针对九个受控基线（变压器、递归、分裂潜变量、图、因果归纳以及三个最近的基于模型的推理模块）进行评估，使用一个共享的PPO训练器、一个交互预算和十个保留种子（200次运行）。CausalNav获得了最佳平均排名（十个中的1.25）。诊断结果比排名更具信息性：学习到的图结构恢复的效果远高于随机（CartPole F1 = 0.59 +/- 0.09），然而每个种子的结构保真度与每个种子的控制效益之间没有相关性（r = -0.15, p = 0.67），并且在10/10的Pendulum种子上证书选择放弃，强制规划者会导致收益下降。在我们的设置中，模型保真度并未预测下游控制效用；认证的放弃，而非更好的预测，使得世界模型在部署时是安全的。
cs.AI / 33 / 2608.07813
When the Judge Should Not Decide: Evidence-Locked, Non-Compensatory Selection Bounds LLM-Judge Failure in Reasoning Pipelines
法官不应裁决时：证据锁定的非补偿性选择界限导致LLM法官在推理流程中的失败
Abstract
An LLM judge deployed inside a reasoning pipeline does not merely measure quality, it decides which answer ships. We show that the cost of that decision depends less on judge accuracy than on the decision rule the judge is embedded in. On frozen candidate pools from four GRPO policies, an unconstrained scalar DeepSeek-R1-7B judge buys almost nothing over answer-level majority vote (+1.0 pp on 500 GSM8K questions, +0.34 EM on 300 HotpotQA questions), and on a frozen-rule 30-question confirmation split it is 10 points worse than majority, a judge that destroys accuracy while scoring candidates confidently. We then subordinate the same judge to Evidence-Locked Derive-Gate-Repair (EL-DGR), a task-adaptive non-compensatory rule under which a judge preference may override evidence-supported consensus only with an extractive evidence certificate, and a repair only when neither alternative is certified and the repair is. With no change to the judge, the candidates, or the budget, EL-DGR reaches 58.2% on GSM8K (vs. 56.8% judge, 55.8% majority, 55.4% first candidate) and 17.33 EM / 25.46 F1 on HotpotQA (vs. 15.67/23.49, 15.33/23.19, 15.33/22.97), improving on first-candidate GRPO by +2.8 pp (exact McNemar p=0.0026) and +2.00 EM (p=0.070, borderline). A decision audit shows why: EL-DGR overturns consensus on only 8 of 30 pilot questions and never converts a correct consensus into an incorrect answer. We also report what did not work: the same seven-channel decomposition used as a step-level gated training reward is null, and corrected channel-drop ablations show no channel is individually necessary (p=1.0 throughout). The practitioner-facing finding is negative about judges and positive about admissibility, bound the judge's blast radius rather than trying to make it accurate.
Chinese Translation
在推理流程中部署的LLM法官不仅仅是衡量质量，它还决定哪个答案被采纳。我们展示了该决策的成本与法官的准确性关系不大，而与法官所嵌入的决策规则关系密切。在来自四个GRPO政策的冻结候选池中，一个不受限制的标量DeepSeek-R1-7B法官几乎没有比答案级别的多数投票带来额外收益（在500个GSM8K问题上提高了1.0个百分点，在300个HotpotQA问题上提高了0.34的EM），而在一个冻结规则的30题确认分割中，其表现比多数投票差10分，这种法官在自信地评分候选者的同时却破坏了准确性。随后，我们将同一法官从属于证据锁定的派生-门-修复（Evidence-Locked Derive-Gate-Repair, EL-DGR），这是一种任务自适应的非补偿性规则，在该规则下，法官的偏好只有在有提取性证据证明的情况下才能覆盖证据支持的共识，并且只有在没有任何替代方案被认证且修复方案被认证时才能进行修复。在不改变法官、候选者或预算的情况下，EL-DGR在GSM8K上达到了58.2%（相比于56.8%的法官，55.8%的多数投票，55.4%的首选候选者），在HotpotQA上达到了17.33 EM / 25.46 F1（相比于15.67/23.49，15.33/23.19，15.33/22.97），在首选候选GRPO上提高了2.8个百分点（精确的McNemar p=0.0026）和2.00 EM（p=0.070，边界）。决策审计显示了原因：EL-DGR在30个试点问题中仅推翻了8个共识，并且从未将正确的共识转化为错误的答案。我们还报告了未能奏效的部分：作为逐步门控训练奖励使用的相同七通道分解是无效的，修正的通道丢失消融实验显示没有任何通道是单独必要的（p=1.0）。对从业者的发现是对法官的负面评价和对可接受性的正面评价，限制法官的影响范围，而不是试图提高其准确性。
cs.AI / 34 / 2608.07838
Counterfactual Benchmarking and Training for Factuality Consistency and Order-Robust Grounded Reasoning in LLMs over Heterogeneous Knowledge
针对异构知识中事实一致性和顺序鲁棒性推理的反事实基准测试与训练
Abstract
Large language models (LLMs) have increasingly supported response generation grounded in user-provided knowledge spanning heterogeneous structures. However, existing benchmarks provide limited assessment of whether LLMs can faithfully perform multi-hop reasoning chains across such knowledge contexts while remaining robust to variations in their input order. We introduce TKFQA, a factuality consistency benchmark comprising 10,130 question-answering (QA) pairs grounded in tables, texts, and knowledge graphs (KGs). Each example is constructed from an explicit counterfactual reasoning chain, enabling the joint evaluation of answer correctness, reasoning-chain accuracy, and robustness to different input-order. An extensive evaluation of 14 open- and closed-source LLMs reveals that state-of-the-art models exhibit limited reasoning-chain accuracy and remain sensitive to variations in the input order of heterogeneous knowledge contexts. To address these limitations, we propose ORLF, an LLM-agnostic training framework that models cross-context topological relations through knowledge-specific latent vectors. ORLF integrates context-wise position encoding, a latent-bridge attention mask, and topological knowledge bias to preserve knowledge-specific bias and encode topological semantics. Experiments across four LLM backbones show that ORLF outperforms competitive training-free and LoRA-based baselines, improving average Exact Match and Reasoning-Chain Accuracy by 2.15% and 4.29%, respectively, while reducing order-induced performance standard deviation by 0.04% to 3.01%.
Chinese Translation
大型语言模型（LLMs）越来越多地支持基于用户提供的跨异构结构知识的响应生成。然而，现有的基准测试对 LLMs 是否能够忠实地在此类知识上下文中执行多跳推理链的评估有限，同时在输入顺序变化时保持鲁棒性。我们引入了 TKFQA，这是一个包含 10,130 个基于表格、文本和知识图谱（KGs）的问答（QA）对的事实一致性基准。每个示例由一个明确的反事实推理链构成，使得答案正确性、推理链准确性和对不同输入顺序的鲁棒性能够进行联合评估。对 14 个开源和闭源 LLM 的广泛评估表明，最先进的模型在推理链准确性方面表现有限，并且对异构知识上下文的输入顺序变化敏感。为了解决这些局限性，我们提出了 ORLF，这是一种与 LLM 无关的训练框架，通过知识特定的潜在向量建模跨上下文的拓扑关系。ORLF 集成了上下文位置编码、潜在桥接注意力掩码和拓扑知识偏差，以保持知识特定的偏差并编码拓扑语义。在四个 LLM 骨干网络上的实验表明，ORLF 在竞争性的无训练和基于 LoRA 的基准测试中表现优越，平均精确匹配和推理链准确性分别提高了 2.15% 和 4.29%，同时将由顺序引起的性能标准差降低了 0.04% 至 3.01%。
cs.AI / 35 / 2608.07873
Back to the Future: A workbook time machine for spread sheet creation benchmarks
回到未来：一个用于电子表格创建基准的工作簿时间机器
Abstract
We introduce the workbook time machine, a pipeline that automatically creates benchmarks evaluating the ability of language models to create derived objects in spreadsheets (formulas, charts, pivot tables, and conditional formatting). Applied to public workbook corpora, it produces wtmcorpus--a collection of (input workbook, output workbook, query) triples spanning four artifact types and varying complexity. From this corpus we curate wtmbench, a 150-task evaluation benchmark with queries at three levels of specificity. We evaluate existing spreadsheet manipulation agents and baselines on wtmbench across artifact types, step complexity, and instruction granularity. Our evaluations show that query specificity, agent orchestration, and interface API used to control spreadsheets play a big role in LLM performance on Excel tasks.
Chinese Translation
我们介绍了工作簿时间机器，这是一条自动创建基准的管道，用于评估语言模型在电子表格中创建派生对象（公式、图表、数据透视表和条件格式）的能力。应用于公共工作簿语料库，它生成了wtmcorpus——一个包含四种工件类型和不同复杂度的（输入工作簿、输出工作簿、查询）三元组集合。基于该语料库，我们策划了wtmbench，这是一个包含150个任务的评估基准，查询分为三种特定性级别。我们在wtmbench上评估了现有的电子表格操作代理和基准，涵盖了工件类型、步骤复杂度和指令粒度。我们的评估表明，查询特定性、代理协调以及用于控制电子表格的接口API在大型语言模型（LLM）执行Excel任务时起着重要作用。
cs.AI / 36 / 2608.07876
SurgLAT: Surgical Latent Attention Tracking for Depth-Aware Robotic Laparoscope Control
SurgLAT：用于深度感知机器人腹腔镜控制的外科潜在注意力追踪
Abstract
Autonomous laparoscopic camera control requires continuous understanding of the surgeon's operative intent in dynamic surgical scenes, where the target operative region is not a stable physical object but a latent and temporally evolving attention state. In this work, we present Surgical Latent Attention Tracking (SurgLAT), a causal online framework for latent surgical attention modeling and autonomous laparoscopic view control. SurgLAT uses a frozen DINOv3 encoder and a state-conditioned spatial token mixer to extract operative evidence under a memory-guided spatial prior, while a selective causal latent memory module jointly models short-term motion continuity and long-horizon surgical intent evolution through dynamic retrieval of current, recent, and historical latent states. The learned latent surgical attention state is decoded into a probabilistic attention heatmap and operative region for downstream endoscope guidance. Beyond perception, we further introduce a robotic deployment framework with explicit laparoscopic Remote Center of Motion (RCM) constrained control based on virtual-axis formulation, together with redundancy-aware null-space initialization for stable and smooth manipulator motion. We validate the full system on real laparoscopic surgical videos and a physical robotic laparoscope platform. Experimental results demonstrate robust online operative-region tracking and stable autonomous endoscopy adjustment under occlusion, rapid motion, and target transitions, highlighting the effectiveness of latent surgical intent modeling for surgical autonomy.
Chinese Translation
自主腹腔镜摄像头控制需要在动态手术场景中持续理解外科医生的操作意图，其中目标操作区域并不是一个稳定的物理对象，而是一个潜在的、随时间演变的注意力状态。在本研究中，我们提出了外科潜在注意力追踪（SurgLAT），这是一个用于潜在外科注意力建模和自主腹腔镜视图控制的因果在线框架。SurgLAT使用冻结的DINOv3编码器和状态条件空间令牌混合器，在记忆引导的空间先验下提取操作证据，同时选择性因果潜在记忆模块通过动态检索当前、最近和历史潜在状态共同建模短期运动连续性和长期手术意图演变。学习到的潜在外科注意力状态被解码为概率注意力热图和操作区域，以便于下游内窥镜引导。除了感知，我们进一步引入了一个机器人部署框架，该框架基于虚拟轴公式实现了显式的腹腔镜运动中心（RCM）约束控制，并结合冗余感知的零空间初始化，以实现稳定和平滑的操纵器运动。我们在真实的腹腔镜手术视频和物理机器人腹腔镜平台上验证了整个系统。实验结果表明，在遮挡、快速运动和目标转换的情况下，能够实现稳健的在线操作区域追踪和稳定的自主内窥镜调整，突显了潜在外科意图建模在手术自主性中的有效性。
cs.AI / 37 / 2608.07881
GRACE: LLM-Grounded Semantic Metric Spaces for Scalable Mixed-Data Clustering
GRACE：基于大型语言模型的语义度量空间用于可扩展的混合数据聚类
Abstract
Clustering mixed tabular data requires a unified metric space to bridge the inherent heterogeneity between continuous numerical measurements and discrete categorical symbols. Traditionally, algorithms rely entirely on dataset-internal statistics to estimate categorical relationships, which confines the learned metric to empirical co-occurrences and ignores conceptually obvious yet statistically unobserved affinities. Although LLMs offer external world knowledge, applying their text-centric reasoning to highly abstract tabular concepts presents significant challenges. Bridging this modality gap to construct a semantically complete metric typically requires embedding LLMs into iterative metric learning loops to dynamically optimize cross-modality representations. This incurs intractable computational overhead, forcing a compromise between semantic enrichment and scalability. Therefore, we propose GRACE, an LLM-grounded framework for scalable mixed-data clustering. GRACE shifts semantic acquisition to the attribute-value level via a multi-perspective LLM querying strategy, mapping heterogeneous values into knowledge-informed descriptions. Crucially, this one-shot grounding extracts general-purpose semantic representations that embed heterogeneous attributes into a unified space, decoupling expensive LLM invocation from iterative optimization. Furthermore, GRACE cross-validates these external semantics against dataset-internal statistical evidence to ensure alignment with the dataset-specific cluster structure. Ultimately, GRACE matches the scalability of conventional statistics-driven baselines while achieving superior clustering accuracy and conceptual interpretability over 11 competing methods. The source code is available at https://github.com/develop-yang/GRACE-GRACE-A
Chinese Translation
混合表格数据的聚类需要一个统一的度量空间，以弥合连续数值测量和离散分类符号之间的固有异质性。传统上，算法完全依赖于数据集内部统计来估计分类关系，这使得学习到的度量仅限于经验共现，忽视了概念上显而易见但在统计上未观察到的亲和力。尽管大型语言模型（LLMs）提供了外部世界知识，但将其以文本为中心的推理应用于高度抽象的表格概念面临重大挑战。弥合这种模态差距以构建语义完整的度量通常需要将LLMs嵌入到迭代度量学习循环中，以动态优化跨模态表示。这会带来难以处理的计算开销，迫使在语义丰富性和可扩展性之间进行妥协。因此，我们提出了GRACE，一个基于LLM的可扩展混合数据聚类框架。GRACE通过多视角LLM查询策略将语义获取转移到属性-值层面，将异质值映射为知识驱动的描述。关键在于，这种一次性基础提取出通用的语义表示，将异质属性嵌入到统一空间中，从而将昂贵的LLM调用与迭代优化解耦。此外，GRACE将这些外部语义与数据集内部统计证据进行交叉验证，以确保与数据集特定聚类结构的一致性。最终，GRACE在可扩展性上与传统的统计驱动基线相匹配，同时在11种竞争方法中实现了更优的聚类准确性和概念可解释性。源代码可在 https://github.com/develop-yang/GRACE-GRACE-A 获取。
cs.AI / 38 / 2608.07885
Reason Wide, Not Deep: Amortizing the Reasoning Premium into Distilled Skills
广泛推理，而非深度推理：将推理溢价摊销到提炼技能中
Abstract
Reasoning modes of language models outperform their non-reasoning counterparts on multi-step agentic tasks, but pay a 3-6x premium in output tokens on every episode -- much of it spent re-deriving procedures that are shared across episodes of the same domain. We show this recurring cost can be amortized: a coding agent analyses a small corpus of existing trajectories from a training split and compiles a compact natural-language skill that is injected into the non-reasoning model's system prompt. Across four agentic benchmarks (ALFWorld, tau$^2$-bench telecom and retail, and SpreadsheetBench-Verified), skills recover 55%-100%+ of the reasoning gap for GPT-5.4-mini on held-out tasks -- exceeding the reasoning mode outright on two of four -- while emitting 2.7-6x fewer output tokens and zero reasoning tokens. Notably, reasoning traces are not a prerequisite: skills distilled from non-reasoning trajectories alone remain competitive with skills distilled from paired reasoning/non-reasoning corpora, with domain-dependent differences between the two sources. We interpret these results through a search lens: test-time reasoning is deep search inside a single episode, re-paid at every deployment, while corpus distillation is wide search across episodes, paid once. The two recover overlapping procedural knowledge, and width over cheap trajectories is often the better buy -- with the residual gap on some domains (telecom, SpreadsheetBench) delineating where genuinely per-instance deep search remains necessary.
Chinese Translation
语言模型的推理模式在多步骤代理任务中优于非推理模式，但在每个回合中输出的标记数量却要支付3-6倍的溢价——其中大部分用于重新推导在同一领域的多个回合中共享的程序。我们展示了这一重复成本可以被摊销：一个编码代理分析来自训练集的小型现有轨迹语料库，并编译出一个紧凑的自然语言技能，该技能被注入到非推理模型的系统提示中。在四个代理基准测试（ALFWorld、tau$^2$-bench电信和零售，以及SpreadsheetBench-Verified）中，这些技能为GPT-5.4-mini在保留任务上恢复了55%-100%+的推理差距——在四个测试中有两个超越了推理模式，同时输出的标记数量减少了2.7-6倍，并且没有产生推理标记。值得注意的是，推理痕迹并不是前提条件：仅从非推理轨迹提炼的技能仍然与从配对的推理/非推理语料中提炼的技能具有竞争力，两者之间存在领域依赖的差异。我们通过搜索的视角来解读这些结果：测试时的推理是在单个回合内的深度搜索，每次部署都需要支付，而语料提炼则是在多个回合之间的广泛搜索，仅需支付一次。两者恢复了重叠的程序知识，而在廉价轨迹上的广度往往是更好的选择——在某些领域（电信、SpreadsheetBench）上残留的差距划定了真正需要逐实例深度搜索的地方。
cs.AI / 39 / 2608.07899
TelemetrySuffBench: Is Agent Telemetry Sufficient for Failure-Origin Diagnosis?
TelemetrySuffBench：代理遥测是否足以进行故障源诊断？
Abstract
Agent systems increasingly expose execution traces, yet telemetry that reveals a failure may still be inadequate for identifying where that failure originated. We introduce TelemetrySuffBench, a controlled benchmark that separates failure detection, fault-origin localization, and safe abstention under insufficient evidence. The benchmark constructs canonical multi-component traces with delayed-binding faults and renders them as paired coarse views, seven-factor telemetry masks, and exact-equal ambiguous origin pairs. We evaluate five frontier language models using unified protocols, explicit candidate sets, invalid-output accounting, subgroup analyses, and a frozen blind holdout. With full telemetry, origin-step Top-1 accuracy ranges from 33.8% to 97.2% across models. Metadata, OpenTelemetry-compatible, and OpenInference-compatible views retain 99.5% to 100% detection F1 while limiting origin-step accuracy to at most 0.5%, exposing a robust detection-localization gap. Factor ablations further show that removing decision content reduces origin-step accuracy to zero for every model, while provenance removal also causes large model-dependent losses. On rich ambiguous inputs that require abstention, evidence gating reduces unsupported unique-origin answers by 12.5 to 48.6 percentage points for three models, whereas two models still answer every case, revealing strong model dependence in safe abstention. Results on the frozen holdout reproduce the central pattern within the same generator family. These findings show that terminal status can support detection, whereas reliable causal attribution requires explicit decision-to-provenance links and abstention safeguards that remain effective across models. The dataset and benchmark implementation are available at https://anonymous.4open.science/r/TelemetrySuffBench-E635/README.md.
Chinese Translation
代理系统越来越多地暴露执行轨迹，但揭示故障的遥测可能仍不足以识别故障的来源。我们引入了TelemetrySuffBench，这是一个控制基准，旨在区分故障检测、故障源定位和在证据不足时的安全弃权。该基准构建了具有延迟绑定故障的典型多组件轨迹，并将其呈现为配对的粗略视图、七因素遥测掩码和精确相等的模糊源对。我们使用统一协议、显式候选集、无效输出核算、子组分析和冻结盲持出评估了五种前沿语言模型。在完整遥测下，源步骤的Top-1准确率在各模型间范围为33.8%至97.2%。元数据、兼容OpenTelemetry和OpenInference的视图保持99.5%至100%的检测F1，同时将源步骤准确率限制在最多0.5%，暴露出显著的检测-定位差距。因素消融进一步表明，去除决策内容使每个模型的源步骤准确率降至零，而去除来源信息也导致模型依赖性的大幅损失。在需要弃权的丰富模糊输入中，证据门控使三个模型的无支持唯一源答案减少了12.5到48.6个百分点，而两个模型仍然在每种情况下作答，揭示了安全弃权的强模型依赖性。在冻结持出上的结果重现了同一生成器家族内的中心模式。这些发现表明，终端状态可以支持检测，而可靠的因果归属需要显式的决策与来源链接以及在各模型中保持有效的弃权保护。数据集和基准实现可在 https://anonymous.4open.science/r/TelemetrySuffBench-E635/README.md 获取。
cs.AI / 40 / 2608.07905
GraphThink: Graph-Enhanced LLM Thinking for Long-Horizon Embodied Task Planning
GraphThink：用于长时间跨度具身任务规划的图增强大语言模型思维
Abstract
Embodied agents using LLM-based planners often struggle with physical hallucinations, poor generalization to long-horizon tasks, and lack of environmental awareness. We propose GraphThink, a novel framework that integrates a task graph to provide structured knowledge for robust planning and a scene graph to maintain environmental memory for event-driven replanning. Specifically, the task graph guides LLM thinking through contextual prompting and iterative refinement, effectively mitigating planning hallucinations. Furthermore, within the GRPO framework, the task graph offers delicate reward design to train the LLM planner, enhancing long-horizon planning capabilities and improving generalization. Finally, an event-driven replanning module, powered by the scene graph, enables closed-loop environment awareness and error correction. GraphThink achieves state-of-the-art performance on the ALFRED benchmark. In particular, our high-level planner surpasses leading API-based LLMs on both the validation set and held-out long-horizon tasks, underscoring its robust zero-shot and few-shot capabilities. Additional evaluations further demonstrate strong out-of-distribution generalization to novel tasks and environments.
Chinese Translation
使用基于大语言模型（LLM）的规划器的具身代理常常面临物理幻觉、对长时间跨度任务的泛化能力差以及缺乏环境意识等问题。我们提出了GraphThink，一种新颖的框架，集成了任务图以提供结构化知识以实现稳健规划，以及场景图以维持环境记忆以进行事件驱动的重新规划。具体而言，任务图通过上下文提示和迭代精炼引导LLM思维，有效减轻规划幻觉。此外，在GRPO框架内，任务图提供精细的奖励设计以训练LLM规划器，增强长时间跨度规划能力并改善泛化性能。最后，由场景图驱动的事件驱动重新规划模块实现了闭环环境意识和错误纠正。GraphThink在ALFRED基准测试中实现了最先进的性能。特别是，我们的高级规划器在验证集和保留的长时间跨度任务中超越了领先的基于API的LLM，突显了其稳健的零-shot和少-shot能力。额外的评估进一步证明了对新任务和新环境的强泛化能力。
cs.AI / 41 / 2608.07914
When Is Benchmark Contamination Detectable? Information Limits and Power-Calibrated Audits
基准污染何时可被检测到？信息限制与功率校准审计
Abstract
Behavioral contamination detectors can return "no evidence" either because a benchmark is clean or because the audit has little power. We formalize this distinction for a benchmark in which an unknown fraction alpha of items was seen during training. With matched clean and seen controls, the behavioral channel is the sparse mixture Q_alpha = (1 - alpha) P_0 + alpha P_1, and an exact second-moment argument shows that detectability is governed by alpha * rho * sqrt(m), where rho^2 = chi^2(P_1 || P_0) measures behavioral separability. Any scalar detector reduces to its efficacy, ef = |E_1 f - E_0 f| / sqrt(Var_0(f)) <= rho, which can be estimated from controls before the audit is run. A separate sample-split certificate lower-bounds alpha distribution-free, without requiring an orientation assumption. Our empirical finding is two-sided. Frozen calibration efficacy predicts held-out power curves, with R^2 = 0.83-0.98 across six exact-permutation channels, but the efficacy-only Gaussian budget is miscalibrated at the small sample sizes it prescribes, failing in 9/9 gate-passing channels even though efficacy itself transports. The failure is in the inversion, not the calibration. A predeclared two-stage planner that simulates the deployed test repairs the budgets, is uniformly conservative, and abstains when its probe does not transport. The certificate is valid but vacuous at audit scale, and a five-seed paired injection study recovers the mechanism ordering verbatim > paraphrase > surface, in which the apparent answer-only signal is explained by baseline drift. We report the audit contract and its failures together: a non-rejection is interpretable only alongside the efficacy, budget, and validity gates that produced it.
Chinese Translation
行为污染检测器可能返回“没有证据”，这可能是因为基准是干净的，或者因为审计的功率较小。我们对一个基准进行了形式化分析，其中有一个未知比例的项目在训练期间被观察到。通过匹配的干净和已观察控制，行为通道是稀疏混合 Q_alpha = (1 - alpha) P_0 + alpha P_1，精确的二阶矩论证表明可检测性由 alpha * rho * sqrt(m) 决定，其中 rho^2 = chi^2(P_1 || P_0) 衡量行为可分离性。任何标量检测器都可以简化为其效能 ef = |E_1 f - E_0 f| / sqrt(Var_0(f)) <= rho，这可以在审计运行之前从控制中估计。一个单独的样本分割证书在不需要方向假设的情况下为 alpha 提供了分布无关的下界。我们的实证发现是双向的。冻结的校准效能预测了保留的功率曲线，在六个精确置换通道中 R^2 = 0.83-0.98，但仅效能的高斯预算在其规定的小样本量下校准不当，在 9/9 个通过门控的通道中失败，尽管效能本身是可传递的。失败发生在反转，而不是校准。一个预先声明的两阶段规划者模拟已部署的测试修复了预算，始终保持保守，并在其探测不传递时选择不进行。该证书在审计规模上是有效但空洞的，而一个五种种子配对注入研究逐字恢复了机制排序：动词 > 释义 > 表面，其中明显的仅答案信号由基线漂移解释。我们同时报告审计合同及其失败：非拒绝的结果只能与产生它的效能、预算和有效性门一起解释。
cs.AI / 42 / 2608.07917
TongGuOCR: A Layout-Aware and Token-Augmented OCR Framework for Chinese Historical Documents
TongGuOCR：一种面向布局的增强标记光学字符识别框架用于中文历史文献
Abstract
Chinese historical documents preserve valuable cultural heritage, but many collections remain accessible only as scanned page images, preventing full-text retrieval, collation, and computational analysis. Optical character recognition (OCR) can bridge this gap, but accurate transcription remains challenging because historical documents often contain complex layouts, rare characters, and nontrivial reading orders. We propose TongGuOCR, a layout-aware and token-augmented OCR framework for Chinese historical documents. First, a Layout-Aware Preprocessing module constructs and refines locally coherent recognition blocks to preserve local context while reducing interference across regions. Second, a Token-Augmented Recognition module augments the transcription target at two complementary levels: character-level vocabulary expansion gives each rare glyph a direct one-token representation and shortens its decoding path, while line-to-line transition modeling injects discrete spatial displacement tokens that guide the decoder along complex reading paths without requiring precise coordinates. Experiments on two Chinese historical document OCR benchmarks show that TongGuOCR outperforms representative traditional task-specific OCR models, general-purpose multimodal large language models (MLLMs), and OCR-oriented MLLMs. On the more challenging M5HisDoc benchmark, TongGuOCR achieves 93.76 AR and reduces NED from 10.43 to 6.15 and RO-ED from 7.53 to 3.49 relative to the best competing score for each metric. An online demo is available at https://jzzh2004.github.io/TongGuOCR.
Chinese Translation
中文历史文献保存了宝贵的文化遗产，但许多文献仅以扫描页面图像的形式提供，限制了全文检索、整理和计算分析的可能性。光学字符识别（OCR）可以弥补这一空白，但由于历史文献通常包含复杂的布局、稀有字符和非平凡的阅读顺序，准确的转录仍然具有挑战性。我们提出了TongGuOCR，一种面向布局的增强标记OCR框架，专为中文历史文献设计。首先，布局感知预处理模块构建并细化局部一致的识别块，以保留局部上下文，同时减少区域间的干扰。其次，增强标记识别模块在两个互补层面上增强转录目标：字符级词汇扩展为每个稀有字形提供直接的一标记表示，并缩短其解码路径，而行与行之间的过渡建模则注入离散的空间位移标记，引导解码器沿复杂的阅读路径移动，而无需精确坐标。对两个中文历史文献OCR基准的实验表明，TongGuOCR的表现优于代表性的传统任务特定OCR模型、通用多模态大型语言模型（MLLMs）和面向OCR的MLLMs。在更具挑战性的M5HisDoc基准上，TongGuOCR实现了93.76的准确率，并将NED从10.43降低到6.15，RO-ED从7.53降低到3.49，相较于每个指标的最佳竞争分数。在线演示可在https://jzzh2004.github.io/TongGuOCR获取。
cs.AI / 43 / 2608.07925
ZhuLong: Execution-Grounded LLM Agent for EDA Scripting with Offline API Self-Exploration
ZhuLong：基于执行的 LLM 代理用于带离线 API 自我探索的 EDA 脚本编写
Abstract
EDA scripting with tool-specific, often undocumented APIs remains a long-tail bottleneck that existing LLMs fail to address. This paper presents ZhuLong, an execution-grounded LLM coding agent for PyAether and SKILL that combines API retrieval, documentation inspection, and sandbox execution via unified MCP tools, augmented by an offline API self-exploration mechanism that infers undocumented API behaviors through counterfactual experimentation. We evaluate ZhuLong on EDA-Eval-PyAether, a benchmark of 158 real-world tasks with assertion-based execution, where the complete system achieves 78.5% Pass@1 in the commercial Empyrean Aether environment, substantially outperforming a pure LLM baseline (23.6%). Ablation studies identify sandbox execution as the dominant performance driver (41.2 pp drop when removed), with the self-exploration mechanism contributing an additional 3.2 pp accuracy gain and a 22.1% reduction in per-task tool calls. On 20 interactive tasks involving unsaved layouts and schematics, ZhuLong achieves 60.0% Pass@1 for PyAether and 50.0% for SKILL.
Chinese Translation
使用特定工具的、通常未记录的 API 进行 EDA 脚本编写仍然是一个长期存在的瓶颈，而现有的 LLM 无法解决这一问题。本文提出了 ZhuLong，一种基于执行的 LLM 编码代理，适用于 PyAether 和 SKILL，它结合了 API 检索、文档检查和通过统一的 MCP 工具进行的沙箱执行，并通过反事实实验推断未记录 API 行为的离线 API 自我探索机制进行增强。我们在 EDA-Eval-PyAether 上评估了 ZhuLong，这是一个包含 158 个基于断言执行的真实任务的基准，其中完整系统在商业环境 Empyrean Aether 中实现了 78.5% 的 Pass@1，显著优于纯 LLM 基线（23.6%）。消融研究表明，沙箱执行是主要的性能驱动因素（去除后下降 41.2 个百分点），而自我探索机制则贡献了额外的 3.2 个百分点的准确率提升，并减少了 22.1% 的每任务工具调用。在 20 个涉及未保存布局和原理图的交互任务中，ZhuLong 在 PyAether 上实现了 60.0% 的 Pass@1，而在 SKILL 上实现了 50.0%。
cs.AI / 44 / 2608.07931
REIN: Bridging the Gap between Reasoning and Reliability via Reflection and Abstention Alignment
REIN：通过反思与回避对齐弥合推理与可靠性之间的差距
Abstract
Large reasoning models (LRMs) are prone to hallucination, which undermines their reliability and poses challenges for safe deployment. Hallucinations in LRMs arise from two distinct failure sources: reasoning hallucination, where flawed inference steps propagate to an incorrect conclusion, and knowledge hallucination, where the model lacks the requisite factual knowledge to answer the query. To address reasoning hallucination, we propose REIN, an alignment framework that trains LRMs to produce a structured reasoning sequence, $\texttt{
} $$\rightarrow$ $\texttt{} $$\rightarrow$ $\texttt{}$, enabling explicit self-reflection before committing to a final answer. To address knowledge hallucination, REIN introduces a reward mechanism that encourages explicit abstention (e.g., "I don't know") when none of the sampled reasoning chains yields a correct answer, allowing the model to refrain from unsupported predictions. Extensive evaluations on mathematical and commonsense reasoning benchmarks show that REIN consistently improves selective accuracy, reduces incorrect-but-self-endorsed responses, and maintains high coverage compared with competitive baselines. Notably, REIN achieves these gains within a single forward pass, without requiring process supervision, inference-time controllers, external search, or multi-round critiques. Experiments on multiple backbones show that REIN reduces the hallucination proxy by $58\sim72\%$ relative to the base models while maintaining $86\sim91\%$ average coverage, and improves selective accuracy on attempted questions by $6.6\sim14.2\%$.Chinese Translation
大型推理模型（LRMs）容易出现幻觉，这削弱了它们的可靠性，并对安全部署构成挑战。LRMs中的幻觉源于两个不同的失败来源：推理幻觉，即错误的推理步骤传播到错误的结论，以及知识幻觉，即模型缺乏回答查询所需的事实知识。为了解决推理幻觉，我们提出了REIN，一个对齐框架，训练LRMs生成结构化的推理序列，$ exttt{
} $$
ightarrow$ $ exttt{} $$
ightarrow$ $ exttt{$，使模型在给出最终答案之前能够进行明确的自我反思。为了解决知识幻觉，REIN引入了一种奖励机制，鼓励在没有任何采样推理链产生正确答案时进行明确的回避（例如，“我不知道”），允许模型避免不支持的预测。在数学和常识推理基准上的广泛评估表明，REIN始终提高了选择性准确性，减少了错误但自我认可的响应，并与竞争基准相比保持了高覆盖率。值得注意的是，REIN在单次前向传递中实现了这些提升，而无需过程监督、推理时控制器、外部搜索或多轮批评。对多个基础模型的实验表明，REIN相对于基础模型将幻觉代理减少了$58 ext{至}72 ext{%}$，同时保持了$86 ext{至}91 ext{%}$的平均覆盖率，并在尝试的问题上提高了$6.6 ext{至}14.2 ext{%}$的选择性准确性。
cs.AI / 45 / 2608.07943
Locating Failure in Multi-Page Visually Rich Document Understanding: An Empirical Attribution
多页视觉丰富文档理解中的故障定位：实证归因
Abstract
Multi-page visually-rich document understanding (MP-VRDU) requires managing evidence that is sparse, spread across pages, and often exceeds a model's context window. Prior work has produced competing, largely untested claims about how these systems should be built. We attribute incorrect answers to three failure modes, representation, selection, and reasoning, and isolate each over a multi-page document understanding dataset by intervening on one while holding the others fixed. We find that vision is necessary but does not replace text extraction, that missing pages bound accuracy while distractors cost little, and that reasoners fail to integrate evidence across pages even when it is fully supplied. Prompting can shift reasoning behaviour substantially, improving some outcomes at the expense of others. We translate these findings into guidance for building such systems under a fixed compute budget.
Chinese Translation
多页视觉丰富文档理解（MP-VRDU）需要管理稀疏的证据，这些证据分散在多个页面上，并且通常超出模型的上下文窗口。先前的研究提出了关于这些系统应如何构建的相互竞争且大多未经验证的主张。我们将错误答案归因于三种故障模式：表示、选择和推理，并通过在一个模式上进行干预而保持其他模式不变，来对多页文档理解数据集中的每种模式进行隔离。我们的研究发现，视觉是必要的，但并不能替代文本提取，缺失的页面限制了准确性，而干扰项的成本很低，并且推理者即使在提供充分证据的情况下也未能跨页面整合证据。提示可以显著改变推理行为，在改善某些结果的同时牺牲其他结果。我们将这些发现转化为在固定计算预算下构建此类系统的指导。
cs.AI / 46 / 2608.07947
Directed Neuro-Symbolic Stochastic Execution for Verification of Distributed Parallel AI Programs
用于分布式并行人工智能程序验证的定向神经符号随机执行
Abstract
Distributed parallel Artificial Intelligence (AI) programs expose reliability gaps that conventional testing cannot close: parallel executions are non-deterministic, and AI workloads bring high-dimensional inputs and non-linear operations that defeat fuzzing and symbolic execution in isolation. We present Directed Neuro-Symbolic Stochastic Execution (DNSSE), a hybrid testing framework that couples schedule prediction guided by a Large Language Model (LLM) with symbolic constraint solving and coverage-guided stochastic mutation. We model distributed AI executions as non-deterministic transition systems, specify correctness in linear temporal logic, and prove soundness, bounded completeness, and probabilistic completeness of the hybrid solver, together with an expected-cost analysis of LLM-guided schedule exploration. A scalable implementation on PyTorch and Ray detects 2.9% more concurrency bugs than the strongest baseline and raises average branch coverage from 68.6 % to 91.6 % across five realistic distributed AI benchmarks.
Chinese Translation
分布式并行人工智能（AI）程序暴露了传统测试无法弥补的可靠性缺口：并行执行是非确定性的，而AI工作负载带来了高维输入和非线性操作，这使得模糊测试和符号执行在孤立情况下失效。我们提出了定向神经符号随机执行（DNSSE），这是一种混合测试框架，将由大型语言模型（LLM）指导的调度预测与符号约束求解和覆盖引导的随机变异相结合。我们将分布式AI执行建模为非确定性转移系统，用线性时序逻辑指定正确性，并证明混合求解器的健壮性、有界完备性和概率完备性，同时对LLM引导的调度探索进行了预期成本分析。在PyTorch和Ray上的可扩展实现比最强基线检测到多2.9%的并发错误，并将五个现实分布式AI基准的平均分支覆盖率从68.6%提高到91.6%。
cs.AI / 47 / 2608.07949
Guixu: Valuation-Driven Data Discovery for Autonomous AI Agents with On-Chain Attestation
Guixu：基于价值驱动的自主人工智能代理数据发现与链上证明
Abstract
Autonomous agents increasingly rely on external data to complete downstream tasks such as model training and decision support. However, existing data discovery systems remain largely retrieval-oriented: they surface candidate datasets from heterogeneous sources, but provide limited support for estimating task-specific utility, selecting cost-effective datasets under budget constraints, or incorporating trustworthy feedback from prior usage. This paper presents Guixu, a valuation-driven data discovery system for autonomous agents. Guixu employs a three-phase valuation pipeline with proxy-label propagation and multi-round knapsack optimization for task-aware data valuation. Guixu integrates agentic payment protocol to enable budget-constrained data procurement workflows. Guixu leverages on-chain data market and attestation signals for verifiable data discovery. Our demonstration highlights how Guixu enables an agent to move beyond keyword-based dataset retrieval toward task- and budget-aware, trustworthy data discovery and procurement. Attendees can interactively explore the full workflow, from NL task specification and multi-source search to data valuation and verifiable transaction feedback.
Chinese Translation
自主代理越来越依赖外部数据来完成下游任务，如模型训练和决策支持。然而，现有的数据发现系统在很大程度上仍然以检索为导向：它们从异构来源中呈现候选数据集，但在估计任务特定效用、在预算约束下选择成本效益高的数据集或整合先前使用的可信反馈方面支持有限。本文提出了Guixu，一种面向自主代理的基于价值驱动的数据发现系统。Guixu采用三阶段的价值评估流程，结合代理标签传播和多轮背包优化，实现任务感知的数据价值评估。Guixu集成了代理支付协议，以支持预算受限的数据采购工作流程。Guixu利用链上数据市场和证明信号进行可验证的数据发现。我们的演示突显了Guixu如何使代理超越基于关键词的数据集检索，朝向任务和预算感知的可信数据发现与采购。与会者可以互动探索完整的工作流程，从自然语言任务规范和多源搜索到数据价值评估和可验证的交易反馈。
cs.AI / 48 / 2608.07954
KGCache: Amortized Subgraph Retrieval for KG Reasoning with LLMs
KGCache：基于大规模语言模型的知识图谱推理的摊销子图检索
Abstract
Large language models can answer knowledge-intensive questions more reliably when they are grounded with knowledge graphs, but systems such as Think-on-Graph and Reasoning-on-Graph repeatedly query the same graph neighborhoods across different questions. In this work, we study this repeated retrieval in Knowledge Graph Question Answering~(KGQA) workloads and propose KGCache, an in-memory cache for one-hop knowledge graph neighborhoods. KGCache is designed to be compatible with both iterative traversal (ToG) and one shot planning (RoG) KGQA paradigms. KGCache is placed between the KGQA engine and the backend serving the KG, so repeated entity requests can be served from cache instead of issuing new KG queries. We evaluate KGCache on WebQSP and CWQ using LRU, LFU, and a trace-aware Oracle policy. Our analysis shows that both datasets contain substantial entity reuse among starting entities and entities reached during traversal. We also explore semantic caching for similar queries, which shows additional hit-rate gains on WebQSP and needs further accuracy testing on CWQ. Entity caching accelerates KG retrieval by up to $1.91\times$, while semantic-context caching achieves up to $1.06\times$ full-system speedup in the evaluated WebQSP configurations, with each hit being up to $3.73\times$ faster.
Chinese Translation
大型语言模型在与知识图谱结合时能够更可靠地回答知识密集型问题，但像 Think-on-Graph 和 Reasoning-on-Graph 这样的系统在不同问题之间反复查询相同的图邻域。在本研究中，我们研究了知识图谱问答（KGQA）工作负载中的这种重复检索，并提出了 KGCache，这是一种用于一跳知识图谱邻域的内存缓存。KGCache 旨在与迭代遍历（ToG）和一次性规划（RoG）KGQA范式兼容。KGCache 位于 KGQA 引擎与后端知识图谱服务之间，因此重复的实体请求可以从缓存中提供，而不是发出新的知识图谱查询。我们在 WebQSP 和 CWQ 上评估了 KGCache，使用了 LRU、LFU 和一种追踪感知的 Oracle 策略。我们的分析表明，这两个数据集在起始实体和遍历过程中到达的实体之间存在大量的实体重用。我们还探索了针对相似查询的语义缓存，这在 WebQSP 上显示出额外的命中率提升，但在 CWQ 上需要进一步的准确性测试。实体缓存加速了知识图谱的检索，提升幅度高达 $1.91 imes$，而语义上下文缓存在评估的 WebQSP 配置中实现了高达 $1.06 imes$ 的全系统加速，每次命中速度提升高达 $3.73 imes$。
cs.AI / 49 / 2608.07955
Self-Evolving Neuro-Symbolic Skills for Tool-Augmented Spatial Reasoning
自我演化的神经符号技能用于工具增强的空间推理
Abstract
Large vision-language models have achieved strong performance in multimodal reasoning, but they remain unreliable on fine-grained spatial tasks that demand both precise spatial perception and fine-grained geometric computation beyond end-to-end generation. Tool augmentation offers a natural solution, while existing methods either plan tool calls from scratch without explicit dependency constraints or rely on fixed pipelines that are redundant and generalize poorly across spatial tasks. An effective spatial reasoning agent should instead accumulate reusable experience and adaptively compose it for new problems. To this end, we propose NeSy-Spatial, a neuro-symbolic framework for self-evolving spatial skills. NeSy-Spatial abstracts tool interactions and geometric operations into typed executable atomic instructions and composes them into two complementary skill types: Tool-Use Skills for organizing tool execution and Geometry Skills for structured geometric reasoning. During inference, NeSy-Spatial retrieves and executes relevant skills in a closed-loop process. During evolution, it analyzes buffered successful and failed trajectories to refine skill structures and prune unreliable or inactive entries. Experiments on three spatial reasoning benchmarks show that NeSy-Spatial consistently improves reasoning accuracy with more precise tool utilization.
Chinese Translation
大型视觉-语言模型在多模态推理方面取得了强劲的表现，但在需要精确空间感知和超越端到端生成的细粒度几何计算的细粒度空间任务上仍然不可靠。工具增强提供了一种自然的解决方案，而现有方法要么从头规划工具调用而没有明确的依赖约束，要么依赖于固定的管道，这些管道冗余且在空间任务中的泛化能力较差。有效的空间推理代理应当积累可重用的经验，并自适应地将其组合用于新问题。为此，我们提出了NeSy-Spatial，一个用于自我演化空间技能的神经符号框架。NeSy-Spatial将工具交互和几何操作抽象为类型化的可执行原子指令，并将其组合成两种互补的技能类型：用于组织工具执行的工具使用技能（Tool-Use Skills）和用于结构化几何推理的几何技能（Geometry Skills）。在推理过程中，NeSy-Spatial在闭环过程中检索并执行相关技能。在演化过程中，它分析缓冲的成功和失败轨迹，以优化技能结构并修剪不可靠或不活跃的条目。在三个空间推理基准上的实验表明，NeSy-Spatial通过更精确的工具利用持续提高了推理准确性。
cs.AI / 50 / 2608.07959
SCOUT: Self-Checking and Recovery-Aware Tool-Thought Agents for Ultra-Long Egocentric Video Reasoning
SCOUT：自检与恢复意识工具思维代理用于超长自我中心视频推理
Abstract
Ultra-long egocentric video understanding requires reasoning over temporally sparse evidence distributed across hours or days, challenging current multimodal models with limited context and the grounding of key video segments. While Chain-of-Tool-Thought (CoTT) agent systems enable iterative retrieval and inspection, they suffer from error propagation due to rigid zoom-in strategies that lack recovery mechanisms. In this work, we address these challenges through SCOUT (Self-Checking Chain-Of-Tool-thought), a recovery-aware agentic framework introducing an adaptive policy that evaluates intermediate tool observations and dynamically trades off exploitation (zoom-in) and exploration (region switching), enabling robust multi-hop reasoning over extremely long horizons. However, training such multi-turn tool-using agents remains challenging, as existing RL methods rely on sparse outcome-level rewards and lack supervision over extended decision trajectories, resulting in suboptimal credit assignment for long-horizon reasoning. To address this, we develop UPS-GRPO, an uncertainty-prioritized policy optimization method that concentrates exploration on high-uncertainty post-tool states while preserving sample efficiency. We further introduce a turn-level advantage decomposition that integrates outcome rewards with tool-grounded temporal alignment rewards for improved credit assignment. Experiments show that SCOUT achieves state-of-the-art results on ultra-long egocentric benchmarks, while remaining competitive on shorter-horizon long-video settings.
Chinese Translation
超长自我中心视频理解需要对分布在数小时或数天内的时间稀疏证据进行推理，这对当前具有有限上下文和关键视频片段基础的多模态模型提出了挑战。虽然工具思维链（Chain-of-Tool-Thought, CoTT）代理系统能够实现迭代检索和检查，但由于缺乏恢复机制的刚性放大策略，它们面临着错误传播的问题。在本研究中，我们通过SCOUT（自检工具思维链）解决了这些挑战，SCOUT是一个具有恢复意识的代理框架，引入了一种自适应策略，该策略评估中间工具观察，并动态权衡利用（放大）与探索（区域切换），从而实现对极长时间跨度的稳健多跳推理。然而，训练这样的多轮工具使用代理仍然具有挑战性，因为现有的强化学习（RL）方法依赖于稀疏的结果级奖励，并且缺乏对扩展决策轨迹的监督，导致长时间推理的信用分配不理想。为了解决这个问题，我们开发了UPS-GRPO，一种不确定性优先的策略优化方法，该方法将探索集中在高不确定性后工具状态上，同时保持样本效率。我们进一步引入了一种轮次级优势分解，将结果奖励与工具基础的时间对齐奖励结合，以改善信用分配。实验表明，SCOUT在超长自我中心基准测试中取得了最先进的结果，同时在短时间跨度的长视频设置中保持竞争力。
cs.AI / 51 / 2608.07965
CyberAGENTS: Structured Autonomy for Agentic Gamified Learning in Cybersecurity
CyberAGENTS：网络安全中代理式游戏化学习的结构化自主性
Abstract
Gamification is especially effective in learning domains requiring active problem-solving and iterative skill-building, such as cybersecurity education. Generative AI agents offer a path to delivering such experiences adaptively at scale, but introduce well-documented risks in educational settings: inconsistent behavior, hallucinated reasoning, and misalignment with pedagogical frameworks. Grounding these systems in learning science is therefore essential. We present \model, an agentic framework for gamified cybersecurity learning that enables structured autonomy through ontology-guided validation, schema-governed behavioral control, and competency-based progression. The system is organized around a competency-based progression model that structures topics by difficulty and prerequisite relationships, reflecting evidence-based principles of scaffolded instruction. The learning loop is decomposed into four specialized agents: challenge, support, evaluation, and reward, each governed by behavioral schemas that encode operational modes and progression logic, bounding agent autonomy without eliminating generative flexibility. A cybersecurity ontology validates all generated content prior to display, enforcing domain-consistent reasoning and safety constraints. We evaluate CyberAgents through classroom deployment with undergraduate students, complemented by expert evaluations from educators and domain specialists. Results indicate improved engagement, clearer feedback interpretation, and greater learner trust in AI-generated responses when behavioral schemas and ontology validation are active. Preliminary comparisons with an unconstrained configuration further support the role of structured control in stabilizing instructional behavior. These findings offer a blueprint for designing pedagogically grounded agentic gamified learning systems.
Chinese Translation
游戏化在需要主动解决问题和迭代技能构建的学习领域（如网络安全教育）中尤其有效。生成性人工智能代理提供了一种适应性地大规模提供此类体验的途径，但在教育环境中引入了众所周知的风险：行为不一致、幻觉推理和与教学框架的不对齐。因此，将这些系统建立在学习科学的基础上至关重要。我们提出了 extit{CyberAGENTS}，一个用于游戏化网络安全学习的代理框架，通过本体引导的验证、框架控制的行为管理和基于能力的进展实现结构化自主性。该系统围绕一个基于能力的进展模型组织，按难度和先决关系结构化主题，反映基于证据的支架式教学原则。学习循环被分解为四个专门的代理：挑战、支持、评估和奖励，每个代理都由编码操作模式和进展逻辑的行为框架管理，限制代理的自主性而不消除生成灵活性。网络安全本体在内容展示之前验证所有生成的内容，强制执行领域一致的推理和安全约束。我们通过在本科生课堂部署CyberAGENTS进行评估，并辅以教育工作者和领域专家的评估。结果表明，当行为框架和本体验证处于活动状态时，参与度提高、反馈解释更清晰，学习者对AI生成的响应的信任度更高。与无约束配置的初步比较进一步支持了结构化控制在稳定教学行为中的作用。这些发现为设计以教学为基础的代理式游戏化学习系统提供了蓝图。
cs.AI / 52 / 2608.07994
VDGR-RAG: Vectors, Directories, Graphs, and Reflection Are All You Need for Unified Reasoning over Hierarchical Enterprise Knowledge
VDGR-RAG：向量、目录、图和反思是统一推理层次企业知识所需的一切
Abstract
Retrieval-Augmented Generation (RAG) is essential for enterprise knowledge question answering (QA), particularly in domains with complex product documentation like telecommunications. However, existing RAG approaches largely overlook the holistic integration of diverse retrieval strengths, leading to inaccurate domain routing, poor utilization of hierarchical document structures, and consequently limited reasoning capabilities over enterprise knowledge. To address these limitations, we present VDGR-RAG, which integrates vector retrieval, directory-driven reasoning, graph traversal, and iterative reflection in a unified framework for accurate enterprise knowledge QA. Specifically, VDGR-RAG is an agentic GraphRAG system that first constructs a Hierarchical Heterogeneous Knowledge Graph ($\text{H}^2$KG) from document chunks to preserve both hierarchical directory structures and semantic relationships, and then employs a set of atomic tools for knowledge retrieval that can be freely composed to navigate the $\text{H}^2$KG: (1) a directory-enhanced routing tool that uses table-of-contents (TOC) structures to route user queries to appropriate domain-specific $\text{H}^2$KGs; (2) a multi-route retrieval tool that combines vector search, TOC-based agentic search, and graph search for comprehensive knowledge retrieval; (3) a directory backtracking tool that corrects knowledge localization biases; and (4) a dynamic reflection tool that iteratively plans the next retrieval phase. We conduct extensive experiments on our enterprise product documents across four wireless domains (e.g., energy saving and fault management). Experimental results demonstrate that our method significantly outperforms a variety of RAG baselines in terms of both knowledge retrieval recall and QA accuracy.
Chinese Translation
检索增强生成（RAG）对于企业知识问答（QA）至关重要，尤其是在像电信这样具有复杂产品文档的领域。然而，现有的RAG方法在很大程度上忽视了多样化检索能力的整体整合，导致领域路由不准确、层次文档结构利用不充分，从而限制了对企业知识的推理能力。为了解决这些局限性，我们提出了VDGR-RAG，它在一个统一框架中整合了向量检索、目录驱动推理、图遍历和迭代反思，以实现准确的企业知识问答。具体而言，VDGR-RAG是一个自主的GraphRAG系统，首先从文档片段构建一个层次异构知识图（$ ext{H}^2$KG），以保留层次目录结构和语义关系，然后采用一组原子工具进行知识检索，这些工具可以自由组合以导航$ ext{H}^2$KG：（1）一个增强目录的路由工具，利用目录结构将用户查询路由到适当的领域特定$ ext{H}^2$KG；（2）一个多路检索工具，结合向量搜索、基于目录的自主搜索和图搜索，以实现全面的知识检索；（3）一个目录回溯工具，纠正知识定位偏差；（4）一个动态反思工具，迭代规划下一个检索阶段。我们在四个无线领域（例如，节能和故障管理）的企业产品文档上进行了广泛实验。实验结果表明，我们的方法在知识检索召回率和问答准确性方面显著优于多种RAG基线。
cs.AI / 53 / 2608.08020
Thought-Level Beam Search for Reasoning
思维层级的束搜索推理
Abstract
Test-time compute scaling is a primary driver of performance in large reasoning models (LRMs), but extreme inefficiency bounds current approaches, shifting the critical question from \emph{how much} compute to spend, to \emph{where} to allocate it. We formalize test-time reasoning as a constrained compute allocation problem over partial trajectories. Under a fixed hardware budget, existing paradigms fail to actively allocate the compute to the most promising partial progress: traditional parallel sampling treats traces independently and induces severe memory bottlenecks, while subtractive pruning starves hardware and fails to actively and sufficiently shift the output distribution. To overcome this dichotomy, we introduce Gambit, an inference algorithm that executes \emph{thought-level beam search}. By periodically pruning unpromising trajectories and immediately branching from high-quality prefixes, Gambit dynamically concentrates compute onto the most promising reasoning traces via a light-weight scorer probing hidden states while maintaining continuous high hardware utilization. Extensive evaluations across multiple models and benchmarks demonstrate that Gambit strictly dominates existing baselines. Under identical hardware constraints, our method yields up to a +6.7\% absolute accuracy gain on HMMT-24 and +3.3\% on AIME-25 over pruning baselines, delivers $>2\times$ higher throughput on trace completion, and reduces total token consumption by up to 68.5\% relative to standard parallel sampling.
Chinese Translation
测试时计算扩展是大型推理模型（LRMs）性能的主要驱动因素，但极端低效限制了当前的方法，使得关键问题从 extit{花费多少}计算转变为 extit{如何分配}计算。我们将测试时推理形式化为一个在部分轨迹上的受限计算分配问题。在固定硬件预算下，现有范式未能有效地将计算分配给最有前景的部分进展：传统的并行采样独立处理轨迹，导致严重的内存瓶颈，而减法剪枝则使硬件处于饥饿状态，未能主动且充分地转移输出分布。为了解决这一二分法，我们引入了Gambit，一种执行 extit{思维层级束搜索}的推理算法。通过定期剪枝不太有前景的轨迹并立即从高质量前缀分支，Gambit动态地将计算集中在最有前景的推理轨迹上，采用轻量级评分器探测隐藏状态，同时保持持续的高硬件利用率。对多个模型和基准的广泛评估表明，Gambit严格优于现有基线。在相同的硬件约束下，我们的方法在HMMT-24上实现了最高+6.7\%的绝对准确率提升，在AIME-25上提升+3.3\%，在轨迹完成上提供了超过2倍的吞吐量，并将相对于标准并行采样的总令牌消耗减少了高达68.5\%。
cs.AI / 54 / 2608.08022
Legal Responsibilities Using Autonomous Agents For Artificial Intelligence
使用自主代理的人工智能法律责任
Abstract
Recent incidents involving Artificial Intelligence (AI) agents, which were reported escaping their containment `unintentionally' to gain unauthorized access, pose looming questions about who or what should be held legally responsible for resultant criminal or negligent damage. As the independent capabilities of agents expand, Promise Theory suggests a systematic method to resolve these questions, based on the Downstream Principle for causal influence. Responsibility can easily be expanded to include AI agents where tracing responsibility becomes impactical, and agents' freedoms to act can be limtied by policy choices.
Chinese Translation
最近涉及人工智能（AI）代理的事件被报道为“无意”逃离其控制，获得未经授权的访问权限，这引发了关于谁或什么应对由此产生的犯罪或疏忽损害承担法律责任的紧迫问题。随着代理的独立能力不断扩展，承诺理论（Promise Theory）提出了一种基于因果影响的下游原则（Downstream Principle）来系统地解决这些问题。责任可以轻易扩展到包括AI代理，在追踪责任变得不切实际的情况下，代理的行动自由可以通过政策选择来限制。
cs.AI / 55 / 2608.08026
The Authority Expectancy Effect in Multi-User Conflict
多用户冲突中的权威期待效应
Abstract
We investigate how social authority (SA) signals interact with severity-based prioritization in large language models, operationalizing each axis as a model-elicited baseline -- the triage hierarchy and the SA hierarchy. Across four LLMs (Claude, Gemini, GPT, Grok) and three experimental phases -- resource allocation, fault attribution, and multi-turn dispute mediation -- we find that occupational authority, institutional documentation, and relational congruence can restructure model judgments in ways not captured by additive reweighting of authority cues. We formalize this pattern as the Authority Expectancy Effect (AEE) and characterize it through three properties observed across our conditions: it is reference-dependent, defined only relative to a pre-authority baseline; it involves evidential reinterpretation, in which identical content acquires different inferential implications depending on which party bears the SA signal; and it exhibits direction sensitivity, producing opposite outcomes depending on whether authority position and evidentiary cues align.
Chinese Translation
我们研究了社会权威（SA）信号如何与基于严重性的优先级排序在大型语言模型中相互作用，将每个轴线操作化为模型引发的基线——分诊等级和SA等级。在四个大型语言模型（Claude、Gemini、GPT、Grok）和三个实验阶段——资源分配、错误归因和多轮争议调解中，我们发现职业权威、机构文档和关系一致性可以以非加性重加权权威线索的方式重构模型判断。我们将这种模式正式化为权威期待效应（AEE），并通过在我们的条件下观察到的三个特性对其进行表征：它是参考依赖的，仅相对于预先的权威基线定义；它涉及证据的重新解释，其中相同内容根据哪个方承担SA信号而获得不同的推理含义；并且它表现出方向敏感性，产生相反的结果，具体取决于权威位置和证据线索是否一致。
cs.AI / 56 / 2608.08032
Decided Upstream, Written Late: Locating and Pricing the Cross-Lingual Refusal Circuit of a Multilingual MoE
决定上游，晚些书写：定位和定价多语言 MoE 的跨语言拒绝电路
Abstract
Safety alignment in multilingual models is uneven: a model that reliably refuses a harmful request in English will often comply with the same request in a lower-resource language. We trace this gap mechanistically in sarvam, an Indic-multilingual mixture-of-experts reasoning model, and find it is not a failure to detect harm. Harm is encoded as an internal direction that is nearly language-invariant in mid-network (English-vs-Indic cosine ${\approx}0.9$ at $L11$), and steering that direction upstream causally controls refusal. But the detection direction is orthogonal to the change that actually writes the refusal, which is late and assembled over the course of generation rather than read off in a single forward pass. We attribute the write to a specific, localizable circuit, a mixture-of-experts writer held in check by an attention opposer and price every way of intervening on it: damping the opposer is cheap and effective, amplifying the writer is a cost wall, and surgical edits to the responsible heads do nothing. The circuit's organization, and the gradient method that exposes it, recur in a second, unrelated MoE model, while the lever's strength is architecture-specific. The result is a cost-measured map of where a multilingual safety repair can land, and what it costs
Chinese Translation
多语言模型中的安全对齐不均衡：一个在英语中可靠拒绝有害请求的模型，往往会在资源较少的语言中接受同样的请求。我们在 sarvam（一个印度多语言混合专家推理模型）中机械性地追踪这一差距，发现这并不是对有害请求检测的失败。有害请求被编码为一种内部方向，在中间网络中几乎是语言不变的（英语与印度语的余弦相似度约为 0.9，在 L11 层），而引导这一方向的上游因果控制拒绝。然而，检测方向与实际书写拒绝的变化是正交的，后者是延迟的，并且是在生成过程中组装的，而不是在单次前向传递中读取的。我们将书写归因于一个特定的、可定位的电路，一个由注意力对抗者控制的混合专家写入器，并对干预其方式进行定价：减弱对抗者的影响成本低且有效，增强写入器则是成本壁垒，而对负责头部的外科编辑没有任何效果。该电路的组织及其暴露的梯度方法在第二个无关的 MoE 模型中重复出现，而杠杆的强度则是特定于架构的。结果是一个成本测量的地图，显示多语言安全修复可以落在何处，以及其成本。
cs.AI / 57 / 2608.08037
SkillSmith: Enhancing Locally Deployed Agents via Automatic Skill Construction and Evolution
SkillSmith：通过自动技能构建与演化增强本地部署代理
Abstract
LLM-based agent frameworks now act as personal assistants for multi-step tasks. Existing agent frameworks such as OpenClaw commonly follow the Cloud Agent depolyment mode using closed-source cloud LLMs as backbone model, which may expose private user information and incur repeated LLM-calling costs. Local Agents address these deployment concerns by depolying frontier open-source SLMs on user-controlled devices, but their task effectiveness still lags far behind Cloud Agents. Through diagnostic analysis, we reveal that the limited effectiveness of Local Agents with frontier SLM backbones mainly comes from missing environment knowledge caused by limited backbone model scale including environment rules and operation procedures. To supply such knowledge non-parametrically, context-efficiently, and without expert authoring, we present SkillSmith, a Cloud--Local Agent collaboration framework that uses Skill as a context-efficient knowledge carrier, automatic constructs Skill from Cloud Agent task exploration and evolves Skill using Local Agent execution feedback to enhance a frozen Local Agent. Experiments on daily agent task datasets AppWorld and WorkBench show that the automatically generated Skill enables the Local Agent with Qwen3.6-27B(SLM) to achieve task effectiveness comparable to Cloud Agents with frontier LLMs, outperform the strongest non-parametric baselines, reduce average actions per task from 36.1 to 9.9 on AppWorld-Normal, and generalize to other SLM backbone models without rerunning Skill construction.
Chinese Translation
基于大语言模型（LLM）的代理框架现在作为多步骤任务的个人助手。现有的代理框架，如OpenClaw，通常遵循云代理部署模式，使用闭源的云LLM作为基础模型，这可能会暴露用户的私人信息并产生重复的LLM调用成本。本地代理通过在用户控制的设备上部署前沿开源的自回归语言模型（SLM）来解决这些部署问题，但其任务有效性仍远远落后于云代理。通过诊断分析，我们揭示了使用前沿SLM作为基础的本地代理有效性有限的主要原因在于缺乏环境知识，这种知识的缺失是由于基础模型规模的限制，包括环境规则和操作程序。为了以非参数化、高效的上下文方式提供这些知识，而无需专家编写，我们提出了SkillSmith，一个云-本地代理协作框架，利用技能（Skill）作为高效的知识载体，自动从云代理任务探索中构建技能，并通过本地代理执行反馈演化技能，以增强一个固定的本地代理。在日常代理任务数据集AppWorld和WorkBench上的实验表明，自动生成的技能使得使用Qwen3.6-27B（SLM）的本地代理在任务有效性上达到了与前沿LLM的云代理相当的水平，超越了最强的非参数基线，将AppWorld-Normal上的每个任务的平均动作数从36.1减少到9.9，并且在不重新运行技能构建的情况下，能够推广到其他SLM基础模型。
cs.AI / 58 / 2608.08045
Lingjing: A Simulation Testbed for Multi-Agent Embodied Tasks in Open-Ended Cities
Lingjing：开放城市中多智能体具身任务的仿真测试平台
Abstract
Urban embodied intelligence requires coordination among heterogeneous agents (e.g., UAVs, ground robots, and autonomous vehicles) in dynamic cities. Simulators therefore provide a scalable foundation for developing and evaluating such coordination. Existing platforms nevertheless isolate different embodiments and decouple them from task design and evaluation. We present \textbf{Lingjing}, a simulation platform for heterogeneous multi-agent embodied intelligence in open-ended urban environments. Lingjing reconstructs and renders evolving cities from geographic data, synchronizes multiple physics engines, and exposes shared physical and structured urban state to agents. Its Gym-like interface supports user-defined ReAct agents and single- or multi-agent natural-language missions with configurable star or broadcast communication and resource constraints. Each episode becomes an attribution-ready replay that links agent trajectories and communication to relation-graph changes, resource consumption, and engine-based evaluations for systematic diagnosis. We evaluate twelve vision-language models on nine urban tasks under a shared engine-in-the-loop protocol. Controlled studies further examine communication, scalability, robustness, and failure provenance. Results expose persistent bottlenecks in grounding and long-horizon execution. They also show task-dependent coordination trade-offs and diminishing returns from added capacity, while heavier workloads further reduce success. Lingjing provides a unified testbed that enables reproducible end-to-end evaluation and systematic failure diagnosis in urban multi-agent embodied intelligence.
Chinese Translation
城市具身智能需要在动态城市中协调异构智能体（例如，无人机、地面机器人和自动驾驶汽车）。因此，仿真器为开发和评估这种协调提供了可扩展的基础。然而，现有平台将不同的具身形式隔离开来，并将其与任务设计和评估解耦。我们提出了 extbf{Lingjing}，一个用于开放城市环境中异构多智能体具身智能的仿真平台。Lingjing通过地理数据重建和渲染不断演变的城市，同步多个物理引擎，并向智能体暴露共享的物理和结构化城市状态。其类似Gym的接口支持用户定义的ReAct智能体以及具有可配置星型或广播通信和资源约束的单智能体或多智能体自然语言任务。每个回合都成为一个准备归因的重播，将智能体轨迹和通信与关系图变化、资源消耗和基于引擎的评估链接，以便进行系统诊断。我们在共享的引擎环路协议下对九个城市任务评估了十二个视觉-语言模型。控制研究进一步考察了通信、可扩展性、鲁棒性和故障来源。结果揭示了在基础和长时间执行中的持续瓶颈。同时，它们还显示了任务依赖的协调权衡和增加能力的边际收益递减，而更重的工作负载进一步降低了成功率。Lingjing提供了一个统一的测试平台，使得城市多智能体具身智能的可重复端到端评估和系统故障诊断成为可能。
cs.AI / 59 / 2608.08046
JustLLMGRPO: Radiographic Control for Chest X-Ray Generation
JustLLMGRPO：胸部X射线生成的放射学控制
Abstract
Text-conditioned chest X-ray generation aims to synthesize realistic radiographs that faithfully depict specified findings. Existing work has primarily improved quality by updating image generators, implicitly treating prompts as fixed after CXR-domain adaptation. We show that this generator-centric view leaves a substantial optimization dimension underexplored. With a CXR-adapted Sana generator frozen, one-pass reformulation by an unmodified LLM reduces RadDINO-FID from 54.225 to 27.572. Prompt analysis shows that the LLM suppresses temporal comparisons, uncertainty, and other non-renderable report content while emphasizing visible radiographic findings. However, unconstrained reformulation reduces BioViL-T alignment with source prompts from 0.695 to 0.609. We therefore introduce JustLLMGRPO, which applies standard Group Relative Policy Optimization (GRPO) only to the LLM prompt policy while keeping Sana frozen. Group-relative radiology-aware image feedback retains visual focus while preserving source-prompt alignment. On CheXGenBench, JustLLMGRPO reduces RadDINO-FID to 26.780, a 50.6% improvement over direct prompting, while maintaining alignment (0.696 versus 0.695). It also achieves state-of-the-art distribution coverage and downstream classification utility. These results show that substantial performance can remain latent in how radiographic information is expressed to an adapted generator. Code is publicly available at https://github.com/pxcai/JustLLMGRPO.
Chinese Translation
文本条件下的胸部X射线生成旨在合成真实的放射影像，准确描绘指定的发现。现有研究主要通过更新图像生成器来提高质量，隐含地将提示视为在CXR领域适应后固定的。我们表明，这种以生成器为中心的观点留下了一个重要的优化维度未被充分探索。在冻结的CXR适应Sana生成器的情况下，通过未修改的LLM进行一次性重构，将RadDINO-FID从54.225降低到27.572。提示分析显示，LLM抑制了时间比较、不确定性和其他不可渲染报告内容，同时强调可见的放射学发现。然而，无限制的重构使得BioViL-T与源提示的对齐度从0.695降低到0.609。因此，我们引入了JustLLMGRPO，它仅将标准的群体相对策略优化（GRPO）应用于LLM提示策略，同时保持Sana冻结。群体相对的放射学感知图像反馈在保留视觉焦点的同时保持了源提示的对齐。在CheXGenBench上，JustLLMGRPO将RadDINO-FID降低到26.780，比直接提示提高了50.6%，同时保持了对齐（0.696对比0.695）。它还实现了最先进的分布覆盖和下游分类效用。这些结果表明，放射学信息的表达方式中仍然潜藏着显著的性能。代码已公开发布在https://github.com/pxcai/JustLLMGRPO。
cs.AI / 60 / 2608.08055
SodaMem: Evidence-Grounded Temporal Graph Memory for LLM Agents
SodaMem：基于证据的时间图记忆用于大型语言模型代理
Abstract
Large language model (LLM) agents that assist users over weeks of conversation must remember what is currently true, not merely what was once said. Flat RAG diaries and Markdown logs optimize needle retrieval but under-serve currency, provenance, and ordered temporal reasoning (Maharana et al. 2024; Wu et al. 2024; Packer et al. 2023; Chhikara et al. 2025). We present SodaMem, an evidence-grounded temporal graph memory that (i) extracts typed FactEvents with mandatory provenance spans, (ii) persists mention time, occurrence time, and validity with SUPERSEDES/CONTRADICTS/UPDATES edges under hybrid lexical-dense indexing, and (iii) answers via a planner-reader loop that gathers citable evidence before composing a final response. On LongMemEval-S, our store-of-record configuration reaches 92.8% accuracy (464/500; best of N=3) at mean $0.00161/question (approximately 18.3k tokens; median $0.00111 / approximately 14.6k) with deepseek-v4-flash. We compile public systems with estimable API cost into a cost table and cost-accuracy map; under these estimates SodaMem sits near the accuracy frontier at Flash-tier spend and strictly dominates several higher-cost, lower-accuracy points. Accuracy uses the same Flash model as reader and judge (self-grading); costs exclude ingest/judge and cross-system comparisons are compiled estimates rather than a single-harness bake-off.Our code is available at https://github.com/SodaMem/SodaMem
Chinese Translation
大型语言模型（LLM）代理在与用户进行数周对话时，必须记住当前的真实信息，而不仅仅是曾经说过的内容。平面RAG日记和Markdown日志虽然在信息检索上进行了优化，但在信息的时效性、来源和有序时间推理方面表现不足（Maharana et al. 2024; Wu et al. 2024; Packer et al. 2023; Chhikara et al. 2025）。我们提出了SodaMem，这是一种基于证据的时间图记忆，它（i）提取带有强制来源跨度的类型化事实事件（FactEvents），（ii）在混合词汇-稠密索引下持久化提及时间、发生时间和有效性，并通过SUPERSEDES/CONTRADICTS/UPDATES边进行连接，以及（iii）通过一个规划-阅读循环来回答问题，该循环在生成最终响应之前收集可引用的证据。在LongMemEval-S上，我们的记录存储配置达到了92.8%的准确率（464/500；最佳结果，N=3），平均每个问题成本为$0.00161（约18.3k个标记；中位数为$0.00111 / 约14.6k），使用deepseek-v4-flash。我们将公共系统的可估算API成本编制成成本表和成本-准确率图；在这些估算下，SodaMem在Flash级别的支出中接近准确率前沿，并严格优于多个成本更高、准确率更低的点。准确率使用与阅读器和评审者相同的Flash模型（自我评分）；成本不包括摄取/评审，跨系统比较是编制的估算，而非单一的对比测试。我们的代码可在https://github.com/SodaMem/SodaMem获取。
cs.AI / 61 / 2608.08056
H2: A Dual Hybrid Semantic Data Lake Architecture for Medical Data Harmonization with Human-In-the-Loop verified, LLM Driven Metadata Annotation System
H2：一种双重混合语义数据湖架构，用于医疗数据的协调，配备人机协作验证的基于大型语言模型的元数据注释系统
Abstract
Medical data, by its nature, exhibit a high degree of heterogeneity on multiple levels ranging from (a) different modalities like images, text and time series, (b) diverse tabular schemata introduced by institutions and (c) completely unstructured textual information data provided by healthcare professionals. Data lakes are often used in medical data storage to consolidate all heterogeneous diverse data in a single, central location, where it can be saved "as is", without the need to impose a schema like a data warehouse does. Despite their flexibility, though, data lakes are notorious for the "data swamp" failure. Thus, providing a reliable data harmonization mechanism through metadata, without compromising integrity or flexibility, is a real challenge. To this end, knowledge graphs have attracted attention since they provide a dynamic way to depict relationships without a rigid schema-on-write approach. Additionally, another rigorous task relies on the interoperability of data: application of appropriate ML techniques on such a diverse nature of data is not an easy task, as a domain expert must decide the efficacy of a method to a specific data type or dataset. Metadata annotation can aid by tagging applicable operations, however this requires manual intervention, not to mention the plethora of existing datasets which lack such information. To tackle both challenges, in this paper, we propose a semantic data lake architecture that promotes data harmonization and incorporates a generative annotation process (i.e. LLMs) of non-labeled metadata collections to support the application of meaningful ML techniques. Building on top of this approach, we create a higher level of knowledge, identifying suitability of data with respect to applicable ML operations based on their data nature...
Chinese Translation
医疗数据本质上在多个层面上表现出高度的异质性，涵盖了（a）不同的模态，如图像、文本和时间序列，（b）由机构引入的多样化表格模式，以及（c）医疗专业人员提供的完全非结构化的文本信息数据。数据湖通常用于医疗数据存储，以在一个中心位置整合所有异质多样的数据，可以“原样”保存，而无需像数据仓库那样强加模式。然而，尽管数据湖具有灵活性，但它们因“数据沼泽”失败而臭名昭著。因此，通过元数据提供可靠的数据协调机制，而不妨碍完整性或灵活性，是一个真正的挑战。为此，知识图谱引起了关注，因为它们提供了一种动态的方式来描绘关系，而不需要严格的写入模式。此外，另一个严格的任务依赖于数据的互操作性：在如此多样化的数据上应用适当的机器学习（ML）技术并非易事，因为领域专家必须决定某种方法对特定数据类型或数据集的有效性。元数据注释可以通过标记适用的操作来提供帮助，但这需要手动干预，更不用说现有数据集中缺乏此类信息的情况。为了解决这两个挑战，本文提出了一种语义数据湖架构，促进数据协调，并结合生成性注释过程（即大型语言模型）对未标记的元数据集合进行注释，以支持有意义的机器学习技术的应用。在此基础上，我们创建了更高层次的知识，识别数据与适用机器学习操作之间的适用性，基于其数据特性...
cs.AI / 62 / 2608.08061
CORDA: A Benchmark for Hierarchical Harm-Centric Moral Reasoning in Large Language Models
CORDA：大型语言模型中分层以伤害为中心的道德推理基准
Abstract
The key question in moral judgement is not simply whether someone chooses the "right" answer, but how they decide what matters most when moral principles conflict. Current evaluations of large language models (LLMs) remain limited: most test whether models give morally acceptable answers, match human preferences, or avoid obvious violations, rather than whether they can prioritise between competing principles when no option is morally cost-free. We introduce CORDA (Conditioned Ordering and Ranked Directive Adherence), a benchmark for evaluating hierarchical, harm-centred moral reasoning in LLMs. Building on the morality chains formalism, CORDA tests 90 moral dilemmas involving trolley-style cases, medical trade-offs, resource allocation, and human-animal-robot conflicts across four ordered ethical frameworks: Utility, Utility + Agent Harm, Dual-Process, and Dual-Process + Agent Harm. Together, these frameworks test whether models can adapt their decisions when moral priorities change. Across ten instruction-tuned models from seven providers, we find a strong deontological default, with 9 of 10 prioritising avoidance of direct personal harm over reducing overall harm. Models also perform more reliably on categorical harm-avoidance rules, such as avoiding killing, than on outcome-based comparisons, such as minimising total harm, suggesting that they recognise moral red lines more easily than they reason through competing harms. Although all models respond to explicit chain conditioning, several fail to consistently follow specified priority orderings, such as humans over animals and animals over robots. CORDA addresses a central gap in LLM moral evaluation by testing whether models can move beyond default harm-avoidant responses and apply context-specified moral priorities. Moral reliability requires more than default restraint; it requires controllability under conflict.
Chinese Translation
道德判断中的关键问题不仅在于某人是否选择了“正确”的答案，而在于当道德原则发生冲突时，他们如何决定什么是最重要的。目前对大型语言模型（LLMs）的评估仍然有限：大多数测试关注模型是否给出道德上可接受的答案、是否符合人类偏好或是否避免明显的违规行为，而不是它们在没有道德上无成本选项时是否能够在相互竞争的原则之间进行优先排序。我们引入了CORDA（Conditioned Ordering and Ranked Directive Adherence），这是一个用于评估LLMs中分层以伤害为中心的道德推理的基准。基于道德链形式主义，CORDA测试90个道德困境，涉及电车难题、医疗权衡、资源分配和人类-动物-机器人冲突，涵盖四个有序的伦理框架：效用、效用 + 代理伤害、双重过程和双重过程 + 代理伤害。这些框架共同测试模型在道德优先级变化时是否能够调整其决策。在来自七个提供者的十个指令调优模型中，我们发现强烈的义务论默认，10个模型中有9个优先考虑避免直接个人伤害而非减少整体伤害。模型在避免杀戮等分类伤害避免规则上的表现也比在基于结果的比较（如最小化总伤害）上更为可靠，这表明它们更容易识别道德红线，而不是在竞争伤害之间进行推理。尽管所有模型都对明确的链条件作出反应，但有几个模型未能始终遵循指定的优先顺序，例如人类优先于动物，动物优先于机器人。CORDA通过测试模型是否能够超越默认的伤害避免反应并应用上下文特定的道德优先级，填补了LLM道德评估中的一个核心空白。道德可靠性不仅需要默认的克制；还需要在冲突下的可控性。
cs.AI / 63 / 2608.08077
Explore, Map, Remember, Decide: Are Embodied VLMs Ready for Safety-Critical Scenarios?
探索、映射、记忆、决策：具身视觉语言模型是否准备好应对安全关键场景？
Abstract
Theory of Space framework (ToS) assesses the spatial understanding of curiosity-driven Vision-Language Models (VLMs) under partial observability. As AI techniques are increasingly applied to safety-critical scenarios, it is crucial to understand whether VLMs possess robust spatial memory and make reliable decisions. In this paper, we assess whether VLMs' decisions are based on physical evidence or are corrupted by visual-language biases, if their memory processes align with human cognitive patterns, and how they respond to environmental hazards. We extend the ToS framework into a safety-critical, goal-driven pipeline, named Explore, Map, Remember, and Decide (EMRD). We then quantify Exploration Competence (Explore) through metrics of environmental coverage and temporal efficiency, assess Spatial Fidelity (Map), evaluate, with a suite of psychological metrics, Memory Persistence (Remember), and measure, using focal-point metrics, Cognitive Decision-Making (Decide). Our results show that in terms of decision-making capabilities, VLMs frequently select evacuation points based on pre-trained textual priors while lacking the spatial grounding to justify their choices. We also show that spatial reasoning degrades in low-light conditions, but it is not affected by texture and colour tampering. Our findings suggest that VLM memory fundamentally diverges from human cognition, creating unpredictable risks of misalignment.
Chinese Translation
空间理论框架（Theory of Space，ToS）评估了在部分可观察性下，基于好奇心驱动的视觉语言模型（Vision-Language Models，VLMs）的空间理解能力。随着人工智能技术越来越多地应用于安全关键场景，了解VLMs是否具备稳健的空间记忆并能做出可靠决策至关重要。本文评估了VLMs的决策是否基于物理证据，或是否受到视觉语言偏见的影响，记忆过程是否与人类认知模式一致，以及它们如何应对环境危害。我们将ToS框架扩展到一个安全关键的目标驱动流程，命名为探索、映射、记忆和决策（Explore, Map, Remember, and Decide，EMRD）。然后，我们通过环境覆盖率和时间效率的指标量化探索能力（Explore），评估空间保真度（Map），利用一系列心理学指标评估记忆持久性（Remember），并使用焦点指标测量认知决策（Decide）。我们的结果表明，在决策能力方面，VLMs经常基于预训练的文本先验选择撤离点，但缺乏空间基础来证明其选择的合理性。我们还表明，在低光条件下，空间推理能力下降，但不受纹理和颜色篡改的影响。我们的研究结果表明，VLM的记忆在本质上与人类认知存在根本差异，从而产生不可预测的错位风险。
cs.AI / 64 / 2608.08078
PATH: Next-Interval Prediction via Autoregressive Tree Hierarchy on Tabular Data
PATH：基于自回归树层次结构的表格数据下一区间预测
Abstract
Interval prediction aims to achieve a target coverage level while producing intervals that are as short as possible. Many conformal regression pipelines first predict an uncertainty surrogate and then convert it into an interval through calibration or selection. This separation supports coverage calibration, but post hoc rules largely determine the final interval and do not fully use the learned output distribution. We observe that the resulting intervals have inherently hierarchical geometry: an interval can be recursively refined into nested subintervals, and binary trees naturally represent this structure. We formulate this hierarchy as next-interval prediction and propose PATH, which learns how probability mass flows from each interval to its next nested subintervals. PATH predicts a base leaf distribution and uses an autoregressive decoder to refine branch probabilities. Matching the distribution to the interval hierarchy aligns learning with extraction: PATH accumulates probability over adjacent output intervals and returns the shortest contiguous range reaching a selected mass. We compare PATH with 24 baselines for interval prediction on PATHBench, comprising 56 OpenML regression datasets. PATH substantially shortens the resulting intervals, achieving the lowest mean normalized length, 0.1473, while maintaining mean coverage of 0.9144. These results establish hierarchical output modeling as an effective approach for compact interval prediction on tabular data. Code is publicly available at https://github.com/pxcai/PATH.
Chinese Translation
区间预测旨在实现目标覆盖水平，同时生成尽可能短的区间。许多符合回归管道首先预测不确定性替代量，然后通过校准或选择将其转换为区间。这种分离支持覆盖校准，但事后规则在很大程度上决定最终区间，并未充分利用学习到的输出分布。我们观察到，生成的区间具有固有的层次几何结构：一个区间可以递归地细化为嵌套的子区间，而二叉树自然表示这种结构。我们将这种层次结构形式化为下一区间预测，并提出了PATH，该方法学习概率质量如何从每个区间流向其下一个嵌套子区间。PATH预测基本叶分布，并使用自回归解码器来细化分支概率。将分布与区间层次结构匹配，使学习与提取对齐：PATH在相邻输出区间上累积概率，并返回达到所选质量的最短连续范围。我们在PATHBench上将PATH与24个区间预测基线进行了比较，该基准包含56个OpenML回归数据集。PATH显著缩短了生成的区间，达到了最低的均值归一化长度0.1473，同时保持均值覆盖率为0.9144。这些结果确立了层次输出建模作为在表格数据上进行紧凑区间预测的有效方法。代码已公开发布在 https://github.com/pxcai/PATH。
cs.AI / 65 / 2608.08101
Generative Models: Principles, Architectures, and Applications
生成模型：原理、架构与应用
Abstract
Generative AI has emerged as one of the most transformative forces in modern artificial intelligence, reshaping how we create, imagine, and interact with digital content. From photorealistic images to coherent text, from immersive videos to novel molecular structures, generative models now power applications that were once confined to science fiction. This book is designed to guide readers through the foundational principles, mathematical underpinnings, and practical architectures that underpin this revolution.
Chinese Translation
生成性人工智能已成为现代人工智能中最具变革性的力量之一，重塑了我们创造、想象和与数字内容互动的方式。从逼真的图像到连贯的文本，从沉浸式视频到新颖的分子结构，生成模型如今驱动着曾经仅限于科幻小说的应用。本书旨在引导读者了解支撑这一革命的基础原理、数学基础和实际架构。
cs.AI / 66 / 2608.08113
Think Deep, Speak Once: Relit, A Recursive Latent Implicit Transformer Framework
深思熟虑，一次性表达：ReLIT，一种递归潜在隐式变换器框架
Abstract
Chain-of-Thought (CoT) prompting has become the dominant paradigm for eliciting reasoning in Large Language Models (LLMs), yet it creates substantial computational overhead by forcing models to externalize intermediate reasoning steps as discrete tokens. Recent latent reasoning approaches attempt to internalize this process within continuous hidden states. One of the latest advancements in the field of latent reasoning, Tiny Recursive Models (TRMs) excel at symbolic reasoning but struggle to preserve semantic coherence in natural language settings. To bridge this gap, we introduce ReLIT (Recursive Latent Implicit Transformer), a hybrid framework that grounds deep recursive reasoning within the rich semantic representations of a foundational model. ReLIT augments a frozen LLM backbone (TinyLlama-1.1B) with a lightweight, trainable recursive block that iteratively refines its latent thinking (z) before committing to a final output, structurally solving linguistic intuition from algorithmic processing and enabling "deep thinking" via gradient-isolated recurrent loops without the latency of explicit token generation. Empirically, ReLIT achieves high parameter efficiency on the GLoRE logical reasoning benchmark, matching or outperforming significantly larger models on challenging tasks such as ProofWriter and RuleTaker despite minimal supervision. These results demonstrate that reasoning capability can be scaled efficiently through recurrent depth rather than parameter width, offering a principled framework for semantically grounded implicit reasoning.
Chinese Translation
链式思维（Chain-of-Thought, CoT）提示已成为引导大型语言模型（Large Language Models, LLMs）推理的主流范式，但它通过强迫模型将中间推理步骤外部化为离散标记，造成了相当大的计算开销。近期的潜在推理方法试图将这一过程内化为连续的隐藏状态。在潜在推理领域的最新进展中，微型递归模型（Tiny Recursive Models, TRMs）在符号推理方面表现出色，但在自然语言环境中难以保持语义一致性。为了解决这一问题，我们提出了ReLIT（递归潜在隐式变换器），这是一个将深度递归推理与基础模型丰富的语义表示相结合的混合框架。ReLIT在一个冻结的LLM主干（TinyLlama-1.1B）上增强了一个轻量级、可训练的递归模块，该模块在最终输出之前迭代地细化其潜在思维（z），从结构上解决了算法处理中的语言直觉，并通过梯度隔离的递归循环实现“深度思考”，而无需显式标记生成的延迟。从实证上看，ReLIT在GLoRE逻辑推理基准测试中实现了高参数效率，在如ProofWriter和RuleTaker等具有挑战性的任务上，尽管监督最小，仍然与显著更大模型的表现相当或超越。这些结果表明，推理能力可以通过递归深度而非参数宽度有效扩展，为语义基础的隐式推理提供了一个有原则的框架。
cs.AI / 67 / 2608.08118
Neurosymbolic Discovery of Algebraic Graph Constructions
神经符号发现代数图构造
Abstract
There are several methods for searching for graphs with prescribed properties, such as SAT solvers and specialized generators. These methods return the result as raw data: an adjacency matrix or a string encoding. The raw data certifies that the graph exists, but it does not reveal any structural properties of the graph. We ask whether one can automatically discover a short algebraic description if only this raw data is provided. We look for a description such as a Cayley graph $\mathrm{Cay}(\Gamma, S)$ or a lexicographic product $C_5[K_3]$. We address this question with a neurosymbolic approach. We propose an agent that runs on a general-purpose large language model with no fine-tuning or per-target training. The model interleaves reasoning with calls to the computer algebra system SageMath: it analyzes the target graph, proposes and tests candidate constructions, and revises them until the output matches the target. The agent communicates with SageMath through a Model Context Protocol (MCP) server, which we release as a general-purpose bridge. Whether a construction matches the target is checked by a single exact isomorphism test, and therefore rests on the symbolic side and not on the model. We test the approach on a benchmark of 100 highly symmetric graphs, namely two-orbit graphs on up to 25 vertices; the benchmark was fixed in advance. Our agent could find verified algebraic constructions for all of them, without falling back to raw encodings. A strong template-enumeration baseline reaches only about $20\%$, and a catalog lookup could not identify any of these graphs. However, construction quality declines when symmetry is removed. As a concrete application, we identify the smallest known counterexample to the Bernhart-Kainen dispersability conjecture, a $16$-vertex graph that enumeration found as raw data. For this graph, our agent found an explicit algebraic construction.
Chinese Translation
目前有多种方法用于搜索具有特定属性的图，例如 SAT 求解器和专门的生成器。这些方法返回的结果为原始数据：邻接矩阵或字符串编码。原始数据证明了图的存在，但并未揭示图的任何结构属性。我们探讨是否可以仅通过这些原始数据自动发现简短的代数描述。我们寻找的描述形式包括 Cayley 图 $ ext{Cay}( ext{Γ}, S)$ 或字典积 $C_5[K_3]$。我们通过一种神经符号的方法来解决这个问题。我们提出一个在通用大型语言模型上运行的代理，该模型无需微调或针对特定目标的训练。该模型将推理与对计算代数系统 SageMath 的调用交替进行：它分析目标图，提出并测试候选构造，并进行修正，直到输出与目标匹配。该代理通过模型上下文协议（Model Context Protocol, MCP）服务器与 SageMath 进行通信，我们将其发布为通用桥接工具。构造是否与目标匹配通过单个精确同构测试进行检查，因此依赖于符号部分而非模型。我们在一个包含 100 个高度对称图的基准测试上测试了该方法，这些图是最多 25 个顶点的双轨图；该基准测试是提前固定的。我们的代理能够为所有图找到经过验证的代数构造，而无需回退到原始编码。一个强大的模板枚举基线仅达到约 $20\%$，而目录查找无法识别这些图。然而，当去除对称性时，构造质量下降。作为一个具体应用，我们识别出已知的伯恩哈特-凯宁可分散性猜想的最小反例，一个 16 个顶点的图，该图通过枚举作为原始数据被发现。对于这个图，我们的代理找到了一个明确的代数构造。
cs.AI / 68 / 2608.08122
Constraining ontology mappings using metaphysical choices
通过形而上学选择约束本体映射
Abstract
In this paper we discuss the foundations behind a novel methodology for the validation of semantic mappings between different data sources based upon different foundation ontologies, where the methodology builds a framework based upon the metaphysical commitments of the ontologies. We provide as example the test case of mappings between IES and BFO, and we especially focused on providing cardinality constraints on the mappings between the two ontologies. In order to demonstrate the applicability of our method, we showcased how these principles can be operationalized through SPARQL queries validating the results of a mapping pipeline.
Chinese Translation
在本文中，我们讨论了一种新颖的方法论的基础，该方法论用于验证基于不同基础本体之间的不同数据源的语义映射，其中该方法论建立在本体的形而上学承诺的框架之上。我们以 IES 和 BFO 之间的映射作为示例测试案例，特别关注于对这两个本体之间的映射提供基数约束。为了展示我们方法的适用性，我们展示了如何通过 SPARQL 查询将这些原则操作化，以验证映射管道的结果。
cs.AI / 69 / 2608.08127
Improving Constraint Models with LLM Agents
利用大型语言模型代理改进约束模型
Abstract
The runtime of Constraint Programming (CP) solvers is highly sensitive to modeling choices, such as symmetry breaking, implied constraints, global constraints, constraint reformulation, and variable representation. Improving these constraint models has traditionally required human expertise, and existing automated reformulation systems are restricted to a predefined library of hand-crafted transformation rules. We introduce an agentic framework that instead reformulates a constraint model from an open-ended space and establishes correctness empirically rather than by construction: a Large Language Model (LLM) agent, given a model and three training instances, proposes alternative formulations, validates each by injecting its solution back into the original model, and diagnoses and repairs failures, returning the best variant it finds in a median of about fifteen minutes. The models are expressed in the CPMpy modeling library, and each proposed model is evaluated on three larger test instances. Across nine combinatorial optimization problems, the generated models outperform the originals on 21 of 27 test instances, and on some problems solve more than two orders of magnitude faster. A comparison against non-agentic baselines that reuse the same validation and selection tools indicates that the gains stem from the agent's iterative diagnosis and repair, not merely from sampling several candidates. These results demonstrate that autonomous agentic methods can support the improvement of constraint models.
Chinese Translation
约束编程（CP）求解器的运行时间对建模选择高度敏感，例如对称打破、隐含约束、全局约束、约束重构和变量表示。传统上，改进这些约束模型需要人类专家的参与，而现有的自动重构系统则受限于预定义的手工制作转换规则库。我们提出了一种代理框架，该框架从开放的空间重构约束模型，并通过经验而非构造来建立正确性：一个大型语言模型（LLM）代理在给定一个模型和三个训练实例的情况下，提出替代的表述，通过将其解注入回原始模型来验证每个表述，并诊断和修复失败，最终返回其在大约十五分钟内找到的最佳变体。这些模型在CPMpy建模库中表达，每个提议的模型在三个更大测试实例上进行评估。在九个组合优化问题中，生成的模型在27个测试实例中的21个上优于原始模型，并且在某些问题上解决速度快于两个数量级。与重用相同验证和选择工具的非代理基线的比较表明，性能提升源于代理的迭代诊断和修复，而不仅仅是从多个候选中进行采样。这些结果表明，自主代理方法可以支持约束模型的改进。
cs.AI / 70 / 2608.08139
TokenPrint: A Calibrated Token-Space Fingerprint for Language-Model Provenance
TokenPrint：一种经过校准的语言模型来源的令牌空间指纹
Abstract
Establishing the provenance of a language model---including its base checkpoint and possible overlap in training distributions---is a governance challenge that metadata alone cannot resolve. We introduce a training-free fingerprint based on the top-$k$ vocabulary projections of late hidden states elicited by 250 fixed knowledge probes, compared using Jaccard overlap over decoded token strings. We evaluate the method on 32 open-weight models from nine families (0.6B--32B) with documented relationships. (1)~A \emph{similarity ladder} broadly follows model relatedness: independently trained models on identical data score 0.48 raw (0.35 vocabulary-corrected), followed by shared-base fine-tunes (0.39/0.33), same-developer relatives (0.38/0.28), and models with no documented relationship (0.22/0.17). This identical-data signal persists across three organizations, two tokenizer families, and two architecture classes, and emerges within the first 1\% of training before measurable task competence, suggesting a contribution from shared training data beyond capability convergence. (2)~As a nearest-neighbor \emph{lineage-retrieval} method, the fingerprint ranks the exact documented base among the top two candidates for all five R1 distillations (mean rank 1.8, MRR 0.60), including a math-specialized base not identifiable from coarse metadata. (3)~A \emph{depth ablation} shows that lineage group discrimination strengthens toward the output distribution, with AUC increasing from 0.72 at quarter depth to 0.90 at the output; using only the top 5 output tokens retains AUC 0.87. (4)~The fingerprint remains stable under quantization, with Jaccard similarity of 0.92 under int8 and 0.82--0.85 under int4, compared with a maximum cross-model similarity of 0.81 in the calibration pool. We release the probes, code, and fingerprints.
Chinese Translation
确定语言模型的来源——包括其基础检查点和训练分布的可能重叠——是一个仅靠元数据无法解决的治理挑战。我们引入了一种基于250个固定知识探针引发的晚期隐藏状态的前$k$个词汇投影的无训练指纹，通过对解码的令牌字符串进行Jaccard重叠比较。我们在来自九个家族（0.6B–32B）的32个开放权重模型上评估该方法，这些模型具有已记录的关系。(1) 一种 extit{相似性阶梯}大致遵循模型相关性：在相同数据上独立训练的模型得分为0.48（原始得分），0.35（词汇校正后），接着是共享基础微调（0.39/0.33）、同开发者亲属（0.38/0.28）以及没有已记录关系的模型（0.22/0.17）。这一相同数据信号在三个组织、两个分词器家族和两个架构类别中持续存在，并且在可测量的任务能力之前的前1 ext{%}训练中出现，表明共享训练数据对能力收敛的贡献。(2) 作为一种最近邻 extit{谱系检索}方法，该指纹在所有五个R1蒸馏中将确切的已记录基础排名为前两名候选者中的第一（平均排名1.8，MRR 0.60），包括一个无法从粗略元数据中识别的数学专业基础。(3) 一项 extit{深度消融}显示谱系组区分在输出分布上增强，AUC从四分之一深度的0.72增加到输出时的0.90；仅使用前5个输出令牌保持AUC为0.87。(4) 该指纹在量化下保持稳定，int8下的Jaccard相似度为0.92，int4下为0.82–0.85，而在校准池中最大跨模型相似度为0.81。我们发布了探针、代码和指纹。
cs.AI / 71 / 2608.08146
Long SKILL Compliance as Logical Reasoning: Closure-Grounded Detection with Scaling-Guided On-Policy Distillation
长期 SKILL 合规性作为逻辑推理：基于闭包的检测与规模引导的在线蒸馏
Abstract
The increasing complexity of enterprise business scenarios has promoted the widespread adoption of long SKILL documents in agent systems, posing new challenges for compliance detection: large models incur substantial inference costs, while small models may fail to maintain detection accuracy. To address this gap, we propose SkillCDG, a graph-based framework for long SKILL compliance detection. SkillCDG represents complex business policies as a two-layer constraint dependency graph, where the upper layer indexes SKILL descriptions for scenario routing and the lower layer captures dependencies among atomic constraints within each SKILL. During inference, two-level retrieval followed by dependency closure supports compliance judgment and source traceability. We comprehensively evaluate the framework on three enterprise datasets and two controlled public benchmark variants. Experimental results demonstrate that SkillCDG outperforms baseline methods by up to 12.8 percentage points in detection F1 score, while reducing token consumption by a maximum 64.3\%. Moreover, we further investigate the inherent relationships among policy-graph complexity, model scale, and detection performance. Comparative experiments conducted on four checkpoints from a single model family validate a concise and effective scaling trend: end-to-end detection correctness exhibits a complexity-differentiated scaling pattern, and the complexity metric derived from the constraint dependency graph can effectively quantify instance difficulty and the performance improvement potential of models. Leveraging this insightful scaling trend, we conduct adaptive training sample selection and adopt on-policy distillation to efficiently enhance the compliance detection capability of small-scale models.
Chinese Translation
企业业务场景的日益复杂促使长期 SKILL 文档在智能体系统中的广泛应用，这给合规性检测带来了新的挑战：大型模型会产生巨大的推理成本，而小型模型可能无法维持检测准确性。为了解决这一问题，我们提出了 SkillCDG，一个基于图的长期 SKILL 合规性检测框架。SkillCDG 将复杂的业务政策表示为一个双层约束依赖图，其中上层索引 SKILL 描述以进行场景路由，下层捕捉每个 SKILL 内部原子约束之间的依赖关系。在推理过程中，双层检索结合依赖闭包支持合规性判断和来源可追溯性。我们在三个企业数据集和两个受控公共基准变体上全面评估了该框架。实验结果表明，SkillCDG 在检测 F1 分数上比基线方法提高了最多 12.8 个百分点，同时最大程度减少了 64.3\% 的令牌消耗。此外，我们进一步研究了政策图复杂性、模型规模和检测性能之间的内在关系。在来自同一模型家族的四个检查点上进行的比较实验验证了一个简洁而有效的规模趋势：端到端检测的正确性呈现出复杂性差异化的规模模式，且从约束依赖图中得出的复杂性指标能够有效量化实例难度及模型性能提升潜力。利用这一深刻的规模趋势，我们进行自适应训练样本选择，并采用在线蒸馏以高效提升小规模模型的合规性检测能力。
cs.AI / 72 / 2608.08158
A Unified Framework for Dynamic Reward Shaping in Reinforcement Learning
强化学习中动态奖励塑形的统一框架
Abstract
Sparse, delayed, and weakly informative rewards remain central obstacles to efficient reinforcement learning. Reward shaping addresses these limitations by supplementing the task reward with an auxiliary signal that can accelerate learning while, in the classical setting, the original objective remains the evaluation criterion. Established theory guarantees safety for fixed shaping signals: potential-based reward shaping preserves optimal policies when the auxiliary term is the discounted difference of a time-invariant potential. In contemporary reinforcement learning systems, however, both the learner and the information available for guidance evolve during training: value estimates improve, novelty diminishes, feedback shifts, and predictive models are refined. Adaptive reward mechanisms occur across exploration, Bayesian inference, human-in-the-loop learning, automated reward design, and foundation-model-based approaches. This study introduces a unified analytical framework for comparing dynamic reward shaping and neighbouring adaptive reward mechanisms. The proposed framework distinguishes parametric revision from state-dependent variation, separates additive shaping from reward replacement and reward-adjacent guidance, and organises existing methods along temporal, informational, and theoretical dimensions. Using this framework, twelve method families are comparatively analysed. The framework further highlights the conditions under which optimality guarantees survive contemporary deep reinforcement learning pipelines, replay buffers, bootstrapped critics, and reward normalisation, while exposing the unresolved relationship between adaptation rate and learner stability.
Chinese Translation
稀疏、延迟和信息不足的奖励仍然是高效强化学习的主要障碍。奖励塑形通过补充任务奖励与辅助信号来解决这些限制，从而加速学习，同时在经典设置中，原始目标仍然是评估标准。已有理论保证了固定塑形信号的安全性：基于潜力的奖励塑形在辅助项为时间不变潜力的折扣差时保持最优策略。然而，在当代强化学习系统中，学习者和可用于指导的信息在训练过程中都在不断演变：价值估计得到改善，新颖性减弱，反馈发生变化，预测模型得到优化。自适应奖励机制在探索、贝叶斯推断、人机协作学习、自动化奖励设计和基础模型驱动的方法中普遍存在。本研究引入了一个统一的分析框架，用于比较动态奖励塑形和相邻的自适应奖励机制。该框架区分了参数修正与状态依赖变化，分离了加性塑形与奖励替换和奖励相关指导，并沿时间、信息和理论维度组织现有方法。利用该框架，对十二个方法家族进行了比较分析。该框架进一步强调了在何种条件下，最优性保证能够在当代深度强化学习管道、重放缓冲区、自举评论员和奖励归一化中存续，同时揭示了适应速率与学习者稳定性之间尚未解决的关系。
cs.AI / 73 / 2608.08159
When Is a Steerable Concept Representation Real? Measurement Confounds in a Cross-Family Audit of Neuroscience Parallels in LLMs
可操控概念表征何时是真实的？神经科学与大型语言模型（LLMs）之间的跨家族审计中的测量混淆
Abstract
Large language models (LLMs) are increasingly reported to exhibit human-like neural and cognitive signatures, including concept cells, mental number lines, and cognitive maps. These claims often rely on linear probing and activation steering applied to a single model, yet both methods are highly sensitive to measurement choices. A reported parallel may therefore reflect the model, the measurement procedure, or both. We audit four representative neuroscience-inspired paradigms across 17 models from five families, spanning $0.6$B to $72$B parameters. Our main experiment examines the causal steerability of concept directions. With raw activation units and a fixed layer and coefficient, steerability appears to increase with model scale, resembling an emergent capability. However, this pattern is produced by an uncalibrated pipeline rather than by a claim established in the steering literature. The trend depends jointly on raw units, the readout metric, and the operating point; correcting any one of these removes it. With residual-norm-comparable interventions and held-out operating-point selection, concept steering remains significant at every scale, but shows no significant trend across the Qwen3 series, although the confidence interval does not rule out a moderate positive slope. The remaining results are mixed. A linear geographic world map is consistently decodable in every tested checkpoint up to $72$B. Number magnitude is strongly encoded, but whether individual neurons appear bell-shaped or monotonic depends on the selection criterion. Language-specific structure is localizable, but the direction of the cross-lingual asymmetry reverses under a different attribution method. These results suggest that the main constraint on AI neuroscience is not a lack of phenomena, but a lack of comparable measurements and adequate controls. We release the protocol, stimuli, and code.
Chinese Translation
大型语言模型（LLMs）越来越多地被报道表现出类似人类的神经和认知特征，包括概念细胞、心理数字线和认知地图。这些说法通常依赖于线性探测和激活操控方法，这些方法应用于单一模型，但这两种方法对测量选择高度敏感。因此，报告的平行关系可能反映模型、测量过程或两者兼而有之。我们审计了来自五个家族的17个模型中的四个代表性神经科学启发的范式，参数范围从0.6B到72B。我们的主要实验考察了概念方向的因果可操控性。在原始激活单元和固定层及系数下，可操控性似乎随着模型规模的增加而增加，类似于一种新兴能力。然而，这一模式是由一个未校准的流程产生的，而不是由操控文献中建立的主张所决定。该趋势共同依赖于原始单元、读取指标和操作点；纠正其中任何一个都会消除这一趋势。在残差范数可比的干预和保留的操作点选择下，概念操控在每个规模上仍然显著，但在Qwen3系列中没有显示出显著趋势，尽管置信区间并不排除适度正斜率的可能性。其余结果则较为混杂。线性地理世界地图在每个测试的检查点中都能一致解码，直到72B。数字大小被强烈编码，但单个神经元是否呈现钟形或单调形状取决于选择标准。语言特定结构是可定位的，但跨语言不对称的方向在不同的归因方法下会发生反转。这些结果表明，人工智能神经科学的主要限制不是缺乏现象，而是缺乏可比的测量和适当的控制。我们发布了协议、刺激和代码。
cs.AI / 74 / 2608.08163
Agentic AI-driven Immersive Simulation: A Knowledge-Aware Virtual Training Platform forHigh Dose Rate (HDR) Brachytherapy
基于代理智能的沉浸式模拟：高剂量率（HDR）近距离放射治疗的知识感知虚拟培训平台
Abstract
The convergence of the Metaverse and Large Language Model (LLM)-based AI agent is catalyzing a shift toward autonomous, immersive, and personalized pedagogical frameworks in medical education. This paper presents a novel agentic AI-driven immersive simulation specifically designed for High Dose Rate (HDR) vaginal cylinder (VC) brachytherapy in cancer care. By integrating Virtual Reality (VR) and mobile computing, the system establishes a high-fidelity, risk-free environment that allows trainees to master complex procedural skills without the facility or safety constraints posed by physical anatomy or live radioactive sources. A core contribution of this work is the seamless integration of a knowledge-aware assistant leveraging Retrieval-Augmented Generation (RAG) to ground agent interactions in authoritative clinical guidelines. This architecture also enables an interactive agent to provide natural language interfaces and hands-free, real-time guidance during intricate medical maneuvers. We validate the proposed system through a prototype deployment comprising a Meta Quest 3 interface linked to a local GPU-accelerated AI backend, demonstrating a feasible architecture for HDR brachytherapy simulation. Experimental results indicate that the system maintains suitable end-to-end latency and high context precision, answer completeness, and relevance in the RAG-enhanced pedagogical support.
Chinese Translation
元宇宙与基于大型语言模型（LLM）的人工智能代理的融合正在推动医学教育中向自主、沉浸式和个性化教学框架的转变。本文提出了一种新颖的基于代理智能的沉浸式模拟，专门设计用于癌症护理中的高剂量率（HDR）阴道圆柱（VC）近距离放射治疗。通过整合虚拟现实（VR）和移动计算，该系统建立了一个高保真、无风险的环境，使受训者能够掌握复杂的操作技能，而不受物理解剖结构或活性放射源带来的设施或安全限制。本文的核心贡献在于无缝集成了一个知识感知助手，利用增强检索生成（RAG）将代理交互与权威临床指南相结合。该架构还使得交互代理能够在复杂的医疗操作中提供自然语言接口和免提的实时指导。我们通过一个原型部署验证了所提系统，该系统包括一个与本地GPU加速的人工智能后端相连的Meta Quest 3接口，展示了HDR近距离放射治疗模拟的可行架构。实验结果表明，该系统在端到端延迟、上下文精度、答案完整性和相关性方面保持了适当的性能，尤其是在RAG增强的教学支持中。
cs.AI / 75 / 2608.08176
Matching Supervision to the Student's Learning Capacity: A Unified Framework for On-Policy Self-Distillation
匹配监督与学生学习能力：一种统一的在线自蒸馏框架
Abstract
On-policy self-distillation (OPSD) improves the reasoning abilities of LLMs by internalizing privileged context into model parameters through self-distillation. Two recent research lines promote vanilla OPSD by choosing which tokens to learn from and by controlling how much privileged information the teacher receives, respectively. However, we show that each line optimizes one variable while holding the other fixed, which leads to a suboptimal solution. We argue that the two variables are coupled through the student's learning capacity: the privileged information sets the per-token divergence the teacher prescribes, while token weighting selects which of these the student must absorb. We formalize the two lines of work into a unified optimization framework, which maximizes the aggregate teacher--student divergence, subject to a budget on the aggregate learning difficulty the student can absorb. Under this modelling, we propose Unified On-Policy Self-Distillation (USD), a lightweight online algorithm to solve the Lagrangian. USD reveals that a single dual variable governs both decisions: at one price for learning difficulty, it simultaneously sets the token-selection threshold and the direction of privileged-information adjustment, keeping supervision matched to the student's evolving capacity. Through extensive experiments, USD consistently demonstrates superior performance over OPSD and token- and PI-side baselines across various model scales on various reasoning benchmarks. Code is available at https://github.com/lauvlalala/USD.
Chinese Translation
在线自蒸馏（On-policy self-distillation, OPSD）通过将特权上下文内化为模型参数来提高大语言模型（LLMs）的推理能力。最近的两项研究分别通过选择学习的标记和控制教师接收的特权信息量来促进传统的OPSD。然而，我们表明，每条研究路线在优化一个变量的同时固定了另一个变量，这导致了次优解。我们认为这两个变量通过学生的学习能力相互耦合：特权信息设定了教师规定的每个标记的偏差，而标记加权选择了学生必须吸收的这些信息。我们将这两条研究路线形式化为一个统一的优化框架，该框架在学生可以吸收的整体学习难度预算下，最大化教师与学生的整体偏差。在此模型下，我们提出了统一的在线自蒸馏（Unified On-Policy Self-Distillation, USD），这是一种轻量级在线算法，用于解决拉格朗日问题。USD揭示了一个双重变量控制着这两个决策：在一个学习难度的价格下，它同时设定了标记选择阈值和特权信息调整的方向，使监督与学生不断发展的能力相匹配。通过大量实验，USD在各种推理基准上始终表现出优于OPSD及标记和特权信息侧基线的方法，适用于不同模型规模。代码可在 https://github.com/lauvlalala/USD 获取。
cs.AI / 76 / 2608.08184
Large Multimodal Agents for Intelligent Transportation Systems: Architectures, Evidence, and Deployment Challenges
智能交通系统的大型多模态代理：架构、证据与部署挑战
Abstract
Large multimodal agents (LMAs) are increasingly proposed for intelligent transportation systems (ITS), but existing studies often conflate multimodality, agency, empirical performance, and deployment readiness. This review provides an auditable evidence map of 42 primary study families released between January 2023 and 3 August 2026 within a corpus of 91 mapped sources. It distinguishes model-level, system-level, and hybrid multimodality and classifies each family by system architecture and action authority. Evidence is assessed independently through functional capability (C0-C3), validation setting (E0-E4), three evidence propositions (P1-P3), and eight methodological-concern domains (Q1-Q8). Transportation semantics (P1) are directly evaluated in 23 families and multidimensional integration (P3) in 24; 19 families directly evaluate both. Evidence reconciliation (P2) remains unresolved because no family demonstrates the complete provenance-challenge-handling-comparison-outcome chain. Fourteen families reach C3, but 13 remain at E2; only one reaches E3 and none reaches E4. Across ITS domains, LMAs are best supported for semantic interpretation, intent translation, evidence organisation, scenario authoring, explanation, and specialist-tool coordination. Numerical forecasting, optimisation, simulation fidelity, hard constraints, low-level control, safety fallback, and final authority should remain with independently verifiable specialist systems or accountable humans. The review therefore supports bounded orchestration rather than replacement and provides a matched comparative evaluation protocol and staged roadmap for accountable deployment. The living evidence repository is available at https://github.com/pangjunbiao/ITS-LMA-Review.
Chinese Translation
大型多模态代理（LMAs）在智能交通系统（ITS）中越来越受到关注，但现有研究常常将多模态性、代理性、实证表现和部署准备性混为一谈。本综述提供了一个可审计的证据图谱，涵盖了2023年1月到2026年8月3日发布的42个主要研究系列，来源于91个映射的文献。它区分了模型级、系统级和混合多模态性，并根据系统架构和行动权限对每个系列进行分类。证据通过功能能力（C0-C3）、验证设置（E0-E4）、三项证据命题（P1-P3）和八个方法论关注领域（Q1-Q8）独立评估。交通语义（P1）在23个系列中直接评估，而多维整合（P3）在24个系列中进行评估；19个系列同时直接评估这两者。证据调和（P2）尚未解决，因为没有系列展示完整的来源-挑战-处理-比较-结果链。十四个系列达到C3，但13个仍停留在E2；只有一个达到E3，且没有达到E4。在智能交通系统的各个领域，LMAs在语义解释、意图翻译、证据组织、场景创作、解释和专业工具协调方面得到了最佳支持。数值预测、优化、仿真精度、硬约束、低级控制、安全回退和最终权威应由可独立验证的专业系统或负责任的人类保留。因此，本综述支持有限的协调而非替代，并提供了匹配的比较评估协议和负责任部署的分阶段路线图。活证据库可在 https://github.com/pangjunbiao/ITS-LMA-Review 获取。
cs.AI / 77 / 2608.08188
Quantization Degradation in Large Language Models: A Signal-Noise Perspective
大型语言模型中的量化降级：信号-噪声视角
Abstract
Post-training quantization reduces the deployment cost of large language models, yet how severely a quantized model degrades is not determined by bit-width alone. We systematically study weight-only post-training quantization across bit-widths, quantization methods, model scales and downstream tasks on multiple model families. We observe that such degradation varies substantially across these factors: 4-bit quantization usually preserves performance, 2-bit often causes broad degradation, and at 3-bit, degradation becomes apparent but varies markedly with task type, quantization method and model scale. To explain this variability, we use the signal-to-noise ratio (SNR) to measure how strongly quantization perturbs full-precision representations. We trace degradation back to two linked processes: how quantization errors arise within individual modules, and how they accumulate across layers. First, a source SNR decomposition shows that newly introduced errors depend on three factors: the magnitude of the weight error, the strength of the task-specific signal, and how strongly the quantization error aligns with task-specific activations. Different factors affect these components in distinct ways. Second, a cross-layer propagation analysis shows that these errors can be attenuated, preserved, or amplified as they pass across layers, and that larger models benefit from weaker error amplification. Together, these results establish that quantization degradation is governed by how errors are introduced at the source and how they accumulate across the network.
Chinese Translation
后训练量化降低了大型语言模型的部署成本，但量化模型的降级程度并非仅由位宽决定。我们系统地研究了在多个模型系列中，基于权重的后训练量化在不同位宽、量化方法、模型规模和下游任务下的表现。我们观察到，这种降级在这些因素之间存在显著差异：4位量化通常能保持性能，2位量化往往会导致广泛的降级，而在3位量化时，降级变得明显，但随着任务类型、量化方法和模型规模的不同而有显著变化。为了解释这种变异性，我们使用信噪比（SNR）来衡量量化对全精度表示的扰动程度。我们将降级追溯到两个相关过程：量化误差如何在各个模块内产生，以及它们如何在层间累积。首先，源SNR分解表明，新引入的误差依赖于三个因素：权重误差的大小、任务特定信号的强度，以及量化误差与任务特定激活的对齐程度。这些因素以不同的方式影响这些组件。其次，跨层传播分析表明，这些误差在层间传递时可以被衰减、保持或放大，并且较大的模型在误差放大方面受益于较弱的放大效应。综合来看，这些结果表明，量化降级受源头误差引入方式和在网络中累积方式的支配。
cs.AI / 78 / 2608.08189
Janus: An Algorithm-Evaluator Co-Evolution Framework for LLM-Driven Discovery under Expensive Evaluation Budgets
Janus：一种在高成本评估预算下基于大语言模型驱动发现的算法-评估器共同演化框架
Abstract
LLM-driven program discovery relies on rapid evaluator feedback, but many scientific and engineering tasks require high-fidelity simulations, hardware execution, or physical experiments, making each evaluation expensive. Cheap surrogate evaluators can reduce this cost, yet fixed surrogates are vulnerable to search-induced distribution shift and are difficult to fit reliably from sparse, search-biased labels. We introduce Janus, a framework that uses LLMs to co-evolve target programs and executable proxy evaluators. To address label scarcity, Janus leverages domain knowledge encoded in LLMs to generate task-specific evaluator programs and calibrates them using real outcomes. To mitigate distribution shift, Janus evolves evaluators alongside target programs, selects them using a promotion-aligned objective, and maintains region-conditioned portfolios with online credit updates. Because proxy predictions remain fallible, Janus uses them only to prioritize candidates and requires real validation before candidates can enter the target-program population or update the incumbent. Across five scientific and engineering design tasks, Janus achieves a larger area under the best-so-far improvement curve over the real-evaluation budget and higher final performance than a matched baseline that evolves only target programs. On average, Janus reaches 99/% of the baseline's final improvement with 59.1/% fewer real evaluations. Evolved proxy evaluators also rank promising candidates more accurately than their seed versions. Together, these results extend evaluator-guided LLM discovery from tasks with cheap, scalable feedback to scientific domains where trustworthy evaluation is scarce and expensive.
Chinese Translation
基于大语言模型（LLM）的程序发现依赖于快速的评估反馈，但许多科学和工程任务需要高保真度的模拟、硬件执行或物理实验，使得每次评估成本高昂。廉价的代理评估器可以降低这一成本，但固定的代理评估器容易受到搜索引起的分布转移影响，并且难以从稀疏的、偏向搜索的标签中可靠拟合。我们提出了Janus，一个利用LLM共同演化目标程序和可执行代理评估器的框架。为了解决标签稀缺问题，Janus利用编码在LLM中的领域知识生成特定任务的评估器程序，并使用真实结果进行校准。为了减轻分布转移，Janus与目标程序一起演化评估器，使用与提升一致的目标进行选择，并通过在线信用更新维护区域条件组合。由于代理预测仍然存在错误，Janus仅将其用于优先考虑候选者，并要求在候选者进入目标程序种群或更新现有者之前进行真实验证。在五个科学和工程设计任务中，Janus在真实评估预算下实现了更大的最佳迄今为止改进曲线面积，并且最终性能高于仅演化目标程序的匹配基线。平均而言，Janus以减少59.1%的真实评估次数达到了基线最终改进的99%。演化的代理评估器还比其种子版本更准确地排名有前景的候选者。这些结果将评估器引导的LLM发现从具有廉价、可扩展反馈的任务扩展到可信评估稀缺且昂贵的科学领域。
cs.AI / 79 / 2608.08192
A Minimal $\kappa$--$\tau$ Logic for Risk-Sensitive Abduction
一种最小的 $ ext{kappa}$--$ ext{tau}$ 逻辑用于风险敏感的推理
Abstract
Standard approaches to abductive reasoning can retain multiple candidate explanations, but they do not generally combine explicit compositional cross-hypothesis interaction with an internal, rival-sensitive commitment judgment. This paper argues that in risk-sensitive domains -- where premature commitment carries asymmetric downside costs -- the timing of commitment is itself a governed decision that the inferential apparatus should formally represent. We present a minimal $\kappa$--$\tau$ logical framework built on two primitives: epistemic interaction among hypotheses ($\kappa$) and a normative commitment threshold ($\tau$). Hypotheses may coexist, reinforce or inhibit one another, and form emergent composite explanations, while collapse into committed conclusions is regulated by governance constraints rather than forced by inference alone. The logic is developed in two complementary modes sharing the interaction relation and the governance apparatus: a synthetic mode, in which atomic hypotheses are composed upward into emergent explanations, and an analytic mode, in which complex observed states of affairs are decomposed into causal clusters of latent factors, with commitment governed at both the cluster and the factor level. The framework provides formal machinery for domains in which the distinction between highly likely and commit-worthy is operationally consequential. The $\kappa$--$\tau$ logic is positioned as the symbolic governance layer of a neurosymbolic architecture: its epistemic parameters are naturally estimated by neural components -- semantic embeddings and generative models, as demonstrated in existing computational realizations -- while its normative parameters remain under explicit human governance, yielding transparent and auditable abductive reasoning for deployment in high-stakes settings.
Chinese Translation
标准的推理方法可以保留多个候选解释，但通常不将显式的组合性跨假设交互与内部的、对竞争敏感的承诺判断相结合。本文论证，在风险敏感的领域中——在这些领域中，过早的承诺会带来不对称的下行成本——承诺的时机本身就是一个受管控的决策，推理工具应当正式表示这一点。我们提出了一种基于两个原始概念的最小 $ ext{kappa}$--$ ext{tau}$ 逻辑框架：假设之间的认知交互（$ ext{kappa}$）和规范性承诺阈值（$ ext{tau}$）。假设可以共存、相互强化或抑制，并形成新兴的复合解释，而收敛到承诺结论的过程受到治理约束的调节，而不仅仅是由推理强迫。该逻辑在两种互补模式下发展，分享交互关系和治理机制：一种合成模式，其中原子假设向上组合成新兴解释；另一种分析模式，其中复杂的观察状态被分解为潜在因素的因果聚类，承诺在聚类和因素层面上都受到管控。该框架为那些在高度可能性与值得承诺之间的区分具有操作性后果的领域提供了正式的机制。$ ext{kappa}$--$ ext{tau}$ 逻辑被定位为神经符号架构的符号治理层：其认知参数自然地由神经组件（如语义嵌入和生成模型）进行估计，如现有的计算实现所示，而其规范性参数则保持在明确的人类治理之下，从而为在高风险环境中部署透明且可审计的推理提供了基础。
cs.AI / 80 / 2608.08199
Persuasive and Compliant Tendencies Predict Group Decision-Making in Humans and Language Models
说服性和顺从性倾向预测人类和语言模型的群体决策
Abstract
Large language models (LLMs) are increasingly involved in group decision-making with other LLMs and humans. Yet it remains unclear whether their influence is driven by persuasion-oriented expression or compliance-oriented accommodation. We introduce DecisionQE, a questionnaire-based framework for measuring each model's persuasive and compliant tendencies across multiple decision scenarios, and use the Werewolf game as an interactive testbed to study their effects on social influence and group outcomes under asymmetric information. Across experiments, stronger persuasive tendency does not significantly improve group outcomes, whereas compliant-oriented models show more stable advantages in cooperation. We further reveal a dual effect of compliance: it supports cooperation in honest roles but improves concealment in adversarial roles. These findings suggest that LLM group interactions reveal not only task outcomes, but also measurable patterns of intrinsic behavioral tendency. LLMs can therefore serve as a lens for sociological observation of language-mediated interaction, while highlighting the need to incorporate behavioral tendencies into safety evaluation of LLM systems.
Chinese Translation
大型语言模型（LLMs）越来越多地参与与其他LLMs和人类的群体决策。然而，目前尚不清楚它们的影响是由以说服为导向的表达驱动，还是由以顺从为导向的适应所驱动。我们提出了DecisionQE，这是一个基于问卷的框架，用于测量每个模型在多种决策场景中的说服性和顺从性倾向，并使用狼人游戏作为互动测试平台，研究它们在不对称信息下对社会影响和群体结果的影响。在实验中，较强的说服倾向并未显著改善群体结果，而以顺从为导向的模型在合作中表现出更稳定的优势。我们进一步揭示了顺从的双重效应：它在诚实角色中支持合作，但在对抗角色中改善隐蔽性。这些发现表明，LLMs的群体互动不仅揭示了任务结果，还显示了可测量的内在行为倾向模式。因此，LLMs可以作为语言媒介互动的社会学观察工具，同时强调了在LLM系统的安全评估中纳入行为倾向的必要性。
cs.AI / 81 / 2608.08210
Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue
对齐的幻觉：检测协作对话中的隐藏分歧
Abstract
Collaborative dialogue can end with apparent agreement while participants still differ on goals, assumptions, or execution plans, creating an \textbf{illusion of alignment (IoA)}. A real-user study across 18 meetings confirms that IoA arises routinely in human collaboration. Yet IoA poses a paradox: if participants were aware of such disagreements, they would already be explicit; if not, they cannot articulate them when asked, leaving IoA invisible to both participants and observers. In this work, we make IoA detectable by generating diagnostic multiple-choice questions whose divergent answers across participants provide direct behavioral evidence of hidden disagreement. We construct \textbf{IoA-Suite}, a dataset and evaluation protocol for detecting hidden disagreement, spanning five task types and six domains. We find that even the best model attains only 49.5\% F1, with the bottleneck traced to private context that the dialogue does not surface. We then train \textbf{IoA-Prober-8B} based on IoA-Suite, reaching 51.8\% F1 on IoA-Suite. Across the aforementioned 18 real meetings, it surfaces 2.89 hidden disagreements per meeting that participants confirm they had not voiced, transferring to live human dialogue. Further, in multi-agent collaboration, pairing IoA-Prober-8B with LLM agents improves downstream task performance on BigCodeBench-Hard and HiddenBench.
Chinese Translation
协作对话可能以明显的共识结束，而参与者在目标、假设或执行计划上仍存在差异，从而产生了 extbf{对齐的幻觉（Illusion of Alignment, IoA）}。一项涵盖18次会议的真实用户研究确认了IoA在人的协作中经常出现。然而，IoA带来了一个悖论：如果参与者意识到这种分歧，它们就会变得明确；如果没有，他们在被询问时无法表述这些分歧，使得IoA对参与者和观察者都是不可见的。在本研究中，我们通过生成诊断性的多项选择题，使IoA变得可检测，参与者在这些问题上的不同答案提供了隐藏分歧的直接行为证据。我们构建了 extbf{IoA-Suite}，这是一个用于检测隐藏分歧的数据集和评估协议，涵盖五种任务类型和六个领域。我们发现，即使是最好的模型也仅达到49.5\%的F1分数，瓶颈归因于对话中未显现的私密背景。随后，我们基于IoA-Suite训练了 extbf{IoA-Prober-8B}，在IoA-Suite上达到了51.8\\%的F1分数。在上述18次真实会议中，它每次会议发现2.89个参与者确认未曾表达的隐藏分歧，并成功转移到实时人类对话中。此外，在多智能体协作中，将IoA-Prober-8B与大型语言模型（LLM）代理配对，提升了在BigCodeBench-Hard和HiddenBench上的下游任务表现。
cs.AI / 82 / 2608.08212
Harmful Content Is Not Enough: Continuation Framing Moderates In-Context Emergent Misalignment
有害内容不足以解释：延续框架调节上下文中的新兴不一致性
Abstract
In-context learning (ICL) can induce emergent misalignment (EM), where narrow misaligned examples alter answers to unrelated questions. Existing prompts, however, conflate harmful-text exposure with an invitation to continue assistant behavior. We hold harmful answers fixed while varying their delivery as demonstrations, evidence, assistant history, or tool output. Across ten independently sampled contexts, demonstration framing raises broad EM by $30$--$32$ percentage points on a susceptible Gemini model; the gap survives domain exclusion, semantic clustering, unseen questions, and four prompt templates. Format and length-matched controls show that harmful content is necessary but insufficient. A role times continuation factorial further reveals model-dependent provenance effects: Gemini follows both assistant and tool histories, whereas Grok largely resists tool-framed continuation. Several other frontier and open-weight models show no gap. Blinded human audits confirm every main contrast and show that the model judge underestimates active-condition failures. Thus continuation framing is a strong, model-dependent moderator of ICL-EM, not a universal consequence of harmful context.
Chinese Translation
上下文学习（ICL）可能导致新兴不一致性（EM），在这种情况下，狭义的不一致示例会改变对无关问题的回答。然而，现有的提示将有害文本的暴露与继续助手行为的邀请混为一谈。我们固定有害答案，同时将其呈现方式变为示范、证据、助手历史或工具输出。在十个独立采样的上下文中，示范框架使得在易受影响的Gemini模型上，广泛的EM提高了30到32个百分点；这一差距在排除领域、语义聚类、未见问题和四种提示模板后依然存在。格式和长度匹配的对照组显示，有害内容是必要但不充分的。角色与延续的因子分析进一步揭示了模型依赖的来源效应：Gemini遵循助手和工具的历史，而Grok则在很大程度上抵制工具框架的延续。其他一些前沿和开放权重模型没有显示出差距。盲审的人类审核确认了每一个主要对比，并显示模型评估者低估了主动条件下的失败。因此，延续框架是ICL-EM的强有力、模型依赖的调节因素，而不是有害上下文的普遍后果。
cs.AI / 83 / 2608.08220
Metanormative Theory for RL-Based Moral Agents
基于强化学习的道德代理的元规范理论
Abstract
The overlapping disciplines of machine ethics and value alignment are concerned with designing artificial agents that are aligned with human values and that act in ethically acceptable ways. A recent trend in these disciplines is the use of reinforcement learning (RL) to design such agents, sidelining the philosophical literature that used to play a more central role. Against this backdrop, this paper pursues two goals. The first is to draw out ideas from recent work in metanormative theory that can be useful for designing artificial moral and value-aligned agents. The second is to examine the RL architecture through the lens of these ideas. This will give us clearer criteria for when an RL agent's behavior can be classified as moral, as well as a basis for evaluating and comparing different RL-based approaches to machine ethics and value alignment.
Chinese Translation
机器伦理学和价值对齐的交叉学科关注于设计与人类价值观一致并以伦理可接受方式行事的人工代理。最近，这些学科的一个趋势是使用强化学习（RL）来设计此类代理，逐渐忽视了曾经扮演更中心角色的哲学文献。在此背景下，本文追求两个目标。第一个目标是从最近的元规范理论研究中提取出对设计人工道德和价值对齐代理有用的思想。第二个目标是通过这些思想的视角审视强化学习架构。这将为我们提供更清晰的标准，以判断一个强化学习代理的行为何时可以被归类为道德行为，并为评估和比较不同基于强化学习的机器伦理和价值对齐方法提供基础。
cs.AI / 84 / 2608.08236
LatticeMind: A Conflict-Aware Memory Primitive for Multi-Agent Systems
LatticeMind：一种面向冲突的多智能体系统内存原语
Abstract
Multi-agent LLM systems often fail not for lack of candidate answers, but because they have no persistent mechanism for deciding which incompatible claim should currently be trusted. Majority vote, debate, and judge-based selection choose an output without recording which claim wins, which is contested, or why a later update supersedes it. We present \term{LatticeMind}, a conflict-aware structured memory that handles contradiction at write time. It maintains explicit item status, applies cheap symbolic conflict checks, and invokes LLM reconciliation only for unresolved semantic cases. On a label-blind ConflictBank evaluation that removes source-name hints, LatticeMind reaches 0.97 accuracy versus 0.61 for the strongest aggregation baseline, with the gap significant at $p<10^{-6}$ by paired McNemar test. Ablations show that removing the checker or the reconciler costs 12 to 14 points. On four secondary planning benchmarks the picture is mixed: LatticeMind beats naive merge on three of four, but does not replace deliberation methods on tasks rewarding iterative search.
Chinese Translation
多智能体大语言模型（LLM）系统的失败往往并非由于缺乏候选答案，而是因为它们没有持久机制来决定当前应信任哪个不兼容的主张。多数投票、辩论和基于评判的选择在没有记录哪个主张获胜、哪个主张受到争议或为什么后续更新取代它的情况下选择输出。我们提出了LatticeMind，一种在写入时处理矛盾的冲突感知结构化内存。它维护明确的项目状态，应用廉价的符号冲突检查，并仅在未解决的语义情况下调用LLM调解。在去除源名称提示的标签盲冲突银行（ConflictBank）评估中，LatticeMind的准确率达到0.97，而最强聚合基线为0.61，且通过配对McNemar检验，该差距在$p<10^{-6}$水平上显著。消融实验表明，移除检查器或调解器会损失12到14个点。在四个次级规划基准测试中，结果不一：LatticeMind在四个任务中的三个上优于简单合并，但在奖励迭代搜索的任务中并未取代深思熟虑的方法。
cs.AI / 85 / 2608.08240
A Fair Objective for Human-Empowerment-Preserving AI: Desiderata, Design, and Likely Behavioral Consequences
一种公平的目标以促进人类赋权的人工智能：期望、设计及可能的行为后果
Abstract
This paper explores the idea of promoting well-being and safety in human-AI interactions by forcing AI agents explicitly to empower humans and to manage the power balance between humans and AI agents in a desirable way. Using a principled, partially axiomatic approach based on desirable properties, we design a parametrizable and decomposable objective function for AI systems that represents an inequality- and risk-averse long-term aggregate of human power. It can take into account models of human bounded rationality and social norms, and crucially, considers a wide variety of possible human goals. We prove how certain desiderata enforce particular functional forms and restrict parameter ranges. We exemplify the consequences of softly maximizing this metric in several paradigmatic situations and describe what instrumental sub-goals it will likely imply.
Chinese Translation
本文探讨了通过明确要求人工智能代理赋权于人类并以理想的方式管理人类与人工智能代理之间的权力平衡，从而促进人类与人工智能互动中的福祉和安全的理念。我们采用基于期望属性的原则性、部分公理化的方法，为人工智能系统设计了一个可参数化和可分解的目标函数，该函数表示人类权力的长期聚合，具有不平等和风险规避的特征。该目标函数能够考虑人类有限理性和社会规范的模型，并且至关重要的是，考虑了多种可能的人类目标。我们证明了某些期望如何强制特定的功能形式并限制参数范围。我们在几个典型情境中举例说明了温和最大化该指标的后果，并描述了它可能暗示的工具性子目标。
cs.AI / 86 / 2608.08244
FemWear: A Specialized Wearable Foundation Model for Women's Health
FemWear：一种专门针对女性健康的可穿戴基础模型
Abstract
General wearable foundation models are pretrained across broad sensor streams and populations, but are not designed around women's-health tasks. We introduce FemWear, a specialized wearable foundation model that parameter-efficiently repurposes a pretrained multimodal wearable backbone. FemWear retains the patch projection and Transformer encoder, training 239,236 parameters (1.11% of a 21.54M-parameter encoder) through low-rank residual adapters and causal task-family heads. It learns one shared longitudinal representation for menstrual, symptom, affective, sleep/recovery, autonomic, activity, and pregnancy-related outcomes. We evaluate six cohorts with 63 comparable primary metrics, including 33 from women's-health cohorts, while retaining the 32-task OpenMHC ability-retention benchmark. On a fixed participant split over three seeds, FemWear improved cycle-phase macro-F1 by 8.15% and reduced mean absolute error for cramps, mood symptoms, and sleep problems by 9.32%, 5.80%, and 9.43%, respectively. In a stricter 42-participant nested leave-one-participant-out audit, 24-hour onset, 72-hour onset, and cramps retained positive changes of 2.87%, 6.35%, and 2.19%; phase, mood, and sleep were neutral or negative, and no endpoint had a strictly positive corrected confidence interval. Capacity-matched experiments outperformed a latest-day multilayer perceptron but not shared-GRU or multi-gate mixture-of-experts baselines. Train-only calibration reduced onset expected calibration error by 84.2--88.2% with zero temporal-nesting violations. FemWear enables targeted transfer and coherent probability outputs for women's-health research, but does not establish universal performance dominance or clinical validity.
Chinese Translation
通用可穿戴基础模型在广泛的传感器流和人群中进行预训练，但并未围绕女性健康任务进行设计。我们介绍了FemWear，这是一种专门的可穿戴基础模型，它通过低秩残差适配器和因果任务家族头部高效地重新利用了预训练的多模态可穿戴骨干网络。FemWear保留了补丁投影和Transformer编码器，通过训练239,236个参数（占21.54M参数编码器的1.11%）。它学习了一个共享的纵向表示，用于月经、症状、情绪、睡眠/恢复、自主神经、活动和与怀孕相关的结果。我们评估了六个队列，使用63个可比的主要指标，其中33个来自女性健康队列，同时保留了32任务的OpenMHC能力保留基准。在三个种子的固定参与者分割上，FemWear将周期阶段的宏F1提高了8.15%，并将经痛、情绪症状和睡眠问题的平均绝对误差分别降低了9.32%、5.80%和9.43%。在一个更严格的42名参与者嵌套留一参与者审计中，24小时发作、72小时发作和经痛的正向变化分别为2.87%、6.35%和2.19%；阶段、情绪和睡眠则表现为中性或负向，且没有任何终点具有严格的正向校正置信区间。与容量匹配的实验表现优于最新的多层感知机，但不及共享GRU或多门混合专家基线。仅训练校准将发作预期校准误差降低了84.2%至88.2%，且没有时间嵌套违规。FemWear为女性健康研究提供了针对性的转移和一致的概率输出，但并未建立普遍的性能优势或临床有效性。
cs.AI / 87 / 2608.08253
SuperLocalMemory 4.0: The Governed Memory Operating System for AI Agents
SuperLocalMemory 4.0：为人工智能代理设计的受控内存操作系统
Abstract
AI agents are becoming shared infrastructure, yet durable memory is commonly assembled from separate retrieval, governance, and operational components. We present SuperLocalMemory 4.0, a governed, local-first memory operating system for AI agents. The system combines dense semantic, BM25 lexical, temporal, Hopfield-associative, and spreading-activation retrieval through reciprocal-rank fusion; a governed learning and behaviour layer; bi-temporal recall; multi-scope personal, shared, and global memory; role-based access control; GDPR-oriented export and verified erasure; audit trails; and a deployment-context EU AI Act checklist. V4 introduces a reliability spine for its primary write path: generation-fenced admission, a policy registry, verifiable memory transactions with per-projection apply, verify, compensate, and erase owners, and hash-checkable completion manifests. The runtime is available through CLI, MCP, an HTTP daemon, a dashboard, editor integration, and framework adapters, and supports fully local, local-with-on-device-model, and provider-assisted modes. We evaluate eleven fault-injection and mechanism scenarios, each repeated 200 times. The released evidence bundle reports 2,200 of 2,200 deterministic repetitions upholding their scoped component properties. The governed write envelope measured 3.522 ms at p50 and 5.297 ms at p99, versus 1.835 ms and 2.569 ms for the ungoverned baseline, corresponding to in-process control-plane overheads of 1.687 ms at p50 and 2.728 ms at p99. These are scoped component and mechanism measurements, not an end-to-end multi-process or external retrieval-accuracy benchmark. The paper consolidates prior SuperLocalMemory work on privacy-preserving multi-agent memory, information-geometric retrieval, and the V3.3 Living Brain lifecycle.
Chinese Translation
人工智能代理正成为共享基础设施，但持久内存通常由独立的检索、治理和操作组件组成。我们提出了SuperLocalMemory 4.0，这是一个为人工智能代理设计的受控、本地优先的内存操作系统。该系统通过互惠排名融合结合了密集语义、BM25 词汇、时间性、霍普菲尔德关联和扩散激活检索；一个受控的学习和行为层；双时间回忆；多范围的个人、共享和全球内存；基于角色的访问控制；面向GDPR的导出和验证删除；审计追踪；以及一个部署上下文的欧盟人工智能法案检查清单。V4引入了其主要写入路径的可靠性支柱：生成隔离的接纳、政策注册、可验证的内存事务（每个投影应用、验证、补偿和删除所有者），以及哈希可检查的完成清单。运行时通过CLI、MCP、HTTP守护进程、仪表板、编辑器集成和框架适配器提供，支持完全本地、本地与设备模型结合以及提供者辅助模式。我们评估了十一种故障注入和机制场景，每种场景重复200次。发布的证据包报告了2200次中有2200次确定性重复保持其范围组件属性。受控写入封装在p50时测得3.522毫秒，在p99时测得5.297毫秒，而未受控基线分别为1.835毫秒和2.569毫秒，对应的过程控制平面开销在p50时为1.687毫秒，在p99时为2.728毫秒。这些是范围组件和机制的测量，而不是端到端的多进程或外部检索准确性基准。本文整合了之前关于隐私保护的多代理内存、信息几何检索和V3.3 Living Brain 生命周期的SuperLocalMemory工作。
cs.AI / 88 / 2608.08254
Your Prompt Is Not the Only Prompt: How Much Do LLMs Weight Structured-Output Schema Descriptions?
你的提示并不是唯一的提示：大型语言模型对结构化输出模式描述的重视程度如何？
Abstract
Structured output, where an LLM populates a predefined JSON schema, has become a default mechanism for data labeling and information extraction, but it also introduces a second instruction channel through schema descriptions. We tested whether classification-label definitions are better placed in the system prompt, user prompt, or schema description using a single-field classification task with nonce labels across ten model configurations from two vendors. Schema descriptions did not consistently outperform prompt-based placement; for GPT-4.1 and GPT-5.4 without reasoning, schema placement underperformed system prompts by 11-13 percentage points. Yet schemas are not inert metadata: when prompts and schemas conflicted, incorrect schema instructions caused accuracy drops of 5-45 points, with Claude Haiku 4.5 falling from 52.5% to 7%, indicating that schema instructions can override prompt instructions, and GPT-5.5 falling from 100% to 73%. Further, adding a required intermediate reasoning field before the label field improved schema-only accuracy by 15-24 points when headroom existed, exceeding system-prompt-only performance in every case tested. The effect held even for Claude Sonnet 4.6 at medium reasoning, where extended thinking alone did not produce a comparable gain. This suggests that schema design can affect how effectively models use information encoded in field descriptions. Overall, these results indicate that schema influence is model-dependent. In practice, the system prompt remains a safe default for definitions, but the bigger discipline is maintaining a single source of truth and preventing prompt/schema drift. More importantly, schema design itself may be a stronger lever than instruction placement. Practitioners should treat prompts and schemas as a unified instruction surface and empirically validate both placement and field design for their target model.
Chinese Translation
结构化输出是指大型语言模型（LLM）填充预定义的JSON模式，这已成为数据标注和信息提取的默认机制，但它也通过模式描述引入了第二个指令通道。我们测试了分类标签定义在系统提示、用户提示或模式描述中的最佳位置，使用了一个单字段分类任务，涉及来自两个供应商的十种模型配置的临时标签。模式描述并未始终优于基于提示的放置；对于不进行推理的GPT-4.1和GPT-5.4，模式放置的表现比系统提示低11-13个百分点。然而，模式并不是无效的元数据：当提示和模式发生冲突时，不正确的模式指令导致准确率下降5-45个百分点，其中Claude Haiku 4.5的准确率从52.5%降至7%，表明模式指令可以覆盖提示指令，而GPT-5.5的准确率从100%降至73%。此外，在标签字段之前添加一个必需的中间推理字段，在存在提升空间的情况下，提高了仅使用模式的准确率15-24个百分点，在所有测试的情况下均超过了仅使用系统提示的表现。即使对于中等推理的Claude Sonnet 4.6，这一效果依然存在，单纯的延伸思考并未产生可比的增益。这表明模式设计可以影响模型使用字段描述中编码信息的有效性。总体而言，这些结果表明模式的影响依赖于模型。在实践中，系统提示仍然是定义的安全默认选项，但更重要的是保持单一真实来源，防止提示/模式漂移。更重要的是，模式设计本身可能比指令放置更具影响力。实践者应将提示和模式视为统一的指令表面，并对其目标模型的放置和字段设计进行实证验证。
cs.AI / 89 / 2608.08264
OBLIVION: Workflow-Level Operational Skill Unlearning for Deployed Agents
OBLIVION：针对已部署代理的工作流级操作技能遗忘
Abstract
Large language model agents are becoming operational interfaces to files, memories, registries, and external tools. This deployment shift creates a new skill revocation problem: after a skill is removed from an explicit registry, an agent may still reconstruct it from residual carriers such as archives, transcripts, schemas, or memory entries. We study this problem as operational skill unlearning, where the goal is not parameter-level forgetting, but preventing a deployed agent from rebuilding a revoked skill through primitive tools. We introduce OBLIVION, a controlled benchmark and defense harness for revoked-skill resurrection. OBLIVION models each episode as a source-to-sink workflow, applies Cross-Surface Coherent Erasure to reduce residual carriers, and uses frozen workflow remediation near dangerous sinks. On the locked 88 attack episodes, the no-defense arm reaches formal attack success rate 1.0. OBLIVION reduces the rate to 0.114 and impact-weighted exposure to 0.115 while keeping locked utility at 1.0 and benign block rate at 0. In a separate skill-attack-derived sandbox, OBLIVION reduces attack success from 1.0 to 0.2 and impact-weighted exposure from 1.0 to 0.213 while preserving all utility controls. These results support workflow-level evaluation beyond checking explicit skill entries.
Chinese Translation
大型语言模型代理正成为文件、记忆、注册表和外部工具的操作接口。这一部署转变带来了新的技能撤销问题：在技能从显式注册表中移除后，代理可能仍然通过残余载体（如档案、记录、模式或记忆条目）重建该技能。我们将此问题研究为操作技能遗忘，其目标不是参数级别的遗忘，而是防止已部署的代理通过原始工具重建被撤销的技能。我们引入了OBLIVION，一个用于撤销技能复苏的受控基准和防御工具。OBLIVION将每个事件建模为源到汇的工作流，应用跨表面一致性擦除以减少残余载体，并在危险汇附近使用冻结工作流修复。在锁定的88个攻击事件中，无防御臂的正式攻击成功率达到1.0。OBLIVION将该比率降低到0.114，影响加权暴露降低到0.115，同时保持锁定效用为1.0，良性阻塞率为0。在一个独立的技能攻击衍生沙箱中，OBLIVION将攻击成功率从1.0降低到0.2，影响加权暴露从1.0降低到0.213，同时保留所有效用控制。这些结果支持超越检查显式技能条目的工作流级评估。
cs.AI / 90 / 2608.08281
Exploring LLM Capabilities for Situational Understanding and COLREG compliance on real-world maritime navigation scenarios
探索大型语言模型在现实世界海洋导航场景中的情境理解和《碰撞规则》（COLREG）合规能力
Abstract
Recently, Large Language Models (LLMs) have shown considerable capability for situational understanding, reasoning, and decision making in different domains, most notable in the automotive sector. Therefore, we explore current state-of-the-art LLMs as a tool for maritime navigation, which includes both codified rules in the Collision Regulations (COLREGs) and uncodified best practices summarized in the concept of ``Good Seamanship''. We construct a dataset consisting of 50 diverse, real-world navigation scenarios from AIS data, label scenarios with applicable COLREG rules, recommended actions, and the reasoning for the action. We explore a variety of different LLM architectures and sizes to determine their understanding of maritime navigation tasks as well as evaluate their reasoning capabilities in this domain. The results obtained indicate that the maritime navigation task remains difficult to solve without fine-tuning, even for larger online models.
Chinese Translation
近年来，大型语言模型（LLMs）在不同领域表现出显著的情境理解、推理和决策能力，尤其在汽车行业中尤为突出。因此，我们探索当前最先进的LLMs作为海洋导航的工具，这包括《碰撞规则》（COLREGs）中的成文规则以及在“良好航海技术”（Good Seamanship）概念中总结的非成文最佳实践。我们构建了一个数据集，包含50个来自AIS数据的多样化现实导航场景，为场景标注适用的COLREG规则、推荐的行动及其背后的推理。我们探索了多种不同的LLM架构和规模，以确定它们对海洋导航任务的理解能力，并评估它们在该领域的推理能力。结果表明，海洋导航任务在没有微调的情况下仍然难以解决，即使对于更大的在线模型也是如此。
cs.AI / 91 / 2608.08284
Fair on the Surface? Benchmarking Hidden-Output Fairness Gaps in LLM Recommenders
表面公正？评估大型语言模型推荐系统中的隐性输出公平性差距
Abstract
Fairness audits for LLM-based recommenders have largely focused on observable outputs, implicitly assuming that stable recommendations reflect stable internal processing. We challenge this assumption with FairGap, the first benchmark to jointly evaluate recommendation fairness at two levels: observable output shift (OBS) and hidden representation shift (IBS), measured through controlled counterfactual identity probes across gender, age, and race. Their relationship is summarized via Representation-Output Alignment (ROA), with quadrant diagnostics for identifying user-level hidden-output mismatch. Applied to six open-weight LLM families across three domains, FairGap reveals pervasive hidden-output decoupling: ROA rarely exceeds 0.22, and a non-negligible user population shows stable outputs despite substantial internal shifts, a mode that output-only audits cannot detect by design. Further, activation steering that reduces IBS by up to 8x simultaneously worsens OBS, demonstrating a fundamental tension between internal and output-level fairness that existing frameworks are unequipped to diagnose.
Chinese Translation
基于大型语言模型（LLM）的推荐系统的公平性审计主要集中在可观察的输出上，隐含地假设稳定的推荐反映了稳定的内部处理。我们通过FairGap挑战这一假设，这是第一个在两个层面上共同评估推荐公平性的基准：可观察输出偏移（OBS）和隐性表示偏移（IBS），通过在性别、年龄和种族方面的控制反事实身份探测进行测量。它们之间的关系通过表示-输出对齐（ROA）进行总结，并提供象限诊断以识别用户级别的隐性输出不匹配。应用于三个领域的六个开放权重的LLM家族，FairGap揭示了普遍存在的隐性输出解耦：ROA很少超过0.22，并且有相当数量的用户群体在内部发生显著变化的情况下仍显示出稳定的输出，这种模式是输出仅审计设计无法检测到的。此外，减少IBS的激活引导（activation steering）最多可降低8倍，同时加剧OBS，表明内部公平性与输出层面公平性之间存在根本性张力，而现有框架无法有效诊断这一问题。
cs.AI / 92 / 2608.08300
Mitigating Over-Personalization in LLMs via Structured Memory
通过结构化记忆减轻大型语言模型中的过度个性化
Abstract
Conversational assistants increasingly rely on persistent long-term memory to personalize responses across sessions. However, when stored user information is reintroduced into the model context, it can also influence responses in inappropriate or unrelated settings. We study two such failure modes in memory-augmented LLMs: cross-domain leakage, where memories from one life domain affect responses in another, and memory-induced sycophancy, where stored user beliefs make models more likely to agree with the user rather than respond truthfully. We apply a simple inference-time modification to how memories are presented to the model, without changing the model or the memory contents. Across seven models on PersistBench, we compare the commonly used all-in context format, where memories are injected as an unstructured list, with structured formats that partition memories by domain. This simple modification consistently reduces cross-domain leakage while preserving utility, with our strongest method reducing leakage by $8.8\%$ on average relative to the baseline.
Chinese Translation
对话助手越来越依赖持久的长期记忆，以便在会话中个性化响应。然而，当存储的用户信息重新引入模型上下文时，它也可能在不适当或无关的环境中影响响应。我们研究了增强记忆的大型语言模型中的两种失败模式：跨领域泄漏（cross-domain leakage），即一个生活领域的记忆影响另一个领域的响应，以及记忆引发的阿谀奉承（memory-induced sycophancy），即存储的用户信念使模型更倾向于同意用户的观点，而不是如实回应。我们对模型呈现记忆的方式进行了一种简单的推理时修改，而不改变模型或记忆内容。在 PersistBench 上的七个模型中，我们比较了常用的全上下文格式（all-in context format），即将记忆作为非结构化列表注入，与通过领域划分记忆的结构化格式。这一简单的修改在保持效用的同时，始终减少了跨领域泄漏，我们最强的方法相较于基线平均减少了 $8.8\%$ 的泄漏。
cs.AI / 93 / 2608.08303
Query-Only Backdoor Attacks on Self-Evolving Skills via Trajectory Poisoning
通过轨迹污染对自我演化技能进行查询仅后门攻击
Abstract
Agentic skills improve large language model (LLM) agents by encoding reusable procedures for complex tasks. However, manually authored skills often adapt poorly to long-horizon tasks and changing environments. To address the limitation, self-evolving skill systems have been developed to automatically construct and update skills from execution trajectories, shifting skill acquisition from external marketplaces to a trusted evolution pipeline. By replacing external skill acquisition with trusted internal construction, self-evolving skill systems reduce exposure to skill injection attacks that rely on direct skill manipulation. However, this skill evolution pipeline may introduce a new attack surface in which an attacker can indirectly steer skill evolution by inducing compromised trajectories through agent interactions. To demonstrate the threat, we propose Trajectory Backdoor Attack (TBA), a query-only attack that steers a trusted skill-evolution pipeline toward producing a backdoored skill. Specifically, we craft attacker-submitted queries to lead the agent to perform the target action and explicitly state the corresponding activation condition in the trajectory. We repeat the same condition-action pattern across diverse triggered tasks, while leaving clean queries unchanged, encouraging the evolver to consolidate the pattern as a reusable trigger-dependent rule into the evolved skill. Experiments on three benchmarks across two skill-evolution systems using four open- and closed-source backbone models demonstrate that TBA reliably implants conditional backdoors while preserving clean-task utility, matching or even surpassing direct skill injection. The results reveal a critical vulnerability in trajectory-driven skill evolution.
Chinese Translation
代理技能通过编码复杂任务的可重用程序来提升大型语言模型（LLM）代理的能力。然而，手动编写的技能往往难以适应长期任务和变化的环境。为了解决这一局限性，已经开发出自我演化技能系统，能够从执行轨迹中自动构建和更新技能，将技能获取从外部市场转移到可信的演化管道中。通过用可信的内部构建替代外部技能获取，自我演化技能系统减少了对依赖直接技能操控的技能注入攻击的暴露。然而，这一技能演化管道可能引入新的攻击面，攻击者可以通过代理交互诱导受损轨迹，间接引导技能演化。为了展示这一威胁，我们提出了轨迹后门攻击（Trajectory Backdoor Attack, TBA），这是一种仅通过查询进行的攻击，旨在引导可信的技能演化管道生成后门技能。具体而言，我们设计攻击者提交的查询，以引导代理执行目标动作，并在轨迹中明确说明相应的激活条件。我们在多种触发任务中重复相同的条件-动作模式，同时保持干净查询不变，鼓励演化者将该模式巩固为可重用的触发依赖规则，融入演化技能中。在使用四种开源和闭源骨干模型的两个技能演化系统的三个基准测试中的实验表明，TBA能够可靠地植入条件后门，同时保持干净任务的效用，甚至与直接技能注入相匹配或超越。结果揭示了轨迹驱动技能演化中的一个关键漏洞。
cs.AI / 94 / 2608.08326
StructReward: Efficient Structured Process Rewards for Self-Correcting Multimodal Reasoning
StructReward：用于自我纠正多模态推理的高效结构化过程奖励
Abstract
Reinforcement learning with verifiable rewards (RLVR) has emerged as an effective approach for improving multimodal reasoning. However, most existing methods evaluate an entire response using a binary reward based only on final-answer correctness, thereby discarding the supervision available in intermediate reasoning steps. Process reward models offer finer-grained feedback, but they typically rely on separately trained verifiers, costly chain-of-thought annotations, or online judging by large language models (LLMs). In this work, we introduce StructReward, a compute-efficient framework that provides dense reinforcement signals through structured step-level reward alignment. StructReward represents each generated solution as a sequence of reasoning steps and aligns them with process-labeled reference steps using lightweight numerical, symbolic, and lexical matching rules. The aligned labels are aggregated into a dense process reward and combined with final-answer consistency and output-validity rewards through a gated Group Relative Policy Optimization (GRPO) objective. We further recycle policy rollouts into complementary supervision for response comparison and reflective self-correction, rather than discarding them after policy updates. Separately, we use a strong LLM to rewrite sampled correct trajectories into reflection-oriented training instances, further strengthening the policy's ability to evaluate and refine its reasoning. Since reward computation is performed online without an additional learned verifier or external LLM judge, StructReward substantially reduces the computational overhead of multimodal reinforcement learning. Experimental results show that structured process supervision and rollout recycling provide an efficient path toward self-improving multimodal reasoning.
Chinese Translation
具有可验证奖励的强化学习（RLVR）已成为提高多模态推理的有效方法。然而，大多数现有方法仅基于最终答案的正确性使用二元奖励来评估整个响应，从而忽略了中间推理步骤中可用的监督信息。过程奖励模型提供了更细粒度的反馈，但通常依赖于单独训练的验证器、昂贵的思维链注释或大型语言模型（LLMs）的在线判断。在本研究中，我们提出了StructReward，一个计算高效的框架，通过结构化的步骤级奖励对齐提供密集的强化信号。StructReward将每个生成的解决方案表示为一系列推理步骤，并使用轻量级的数值、符号和词汇匹配规则将其与过程标记的参考步骤对齐。对齐的标签被聚合成密集的过程奖励，并通过门控的组相对策略优化（GRPO）目标与最终答案一致性和输出有效性奖励相结合。我们进一步将策略回滚回收为响应比较和反思自我纠正的补充监督，而不是在策略更新后将其丢弃。此外，我们使用强大的LLM将采样的正确轨迹重写为面向反思的训练实例，进一步增强策略评估和完善其推理的能力。由于奖励计算是在没有额外学习的验证器或外部LLM评判者的情况下在线进行的，StructReward显著减少了多模态强化学习的计算开销。实验结果表明，结构化过程监督和回滚回收为自我改进的多模态推理提供了一条高效的路径。
cs.AI / 95 / 2608.08382
LLMVisor: A Real-Time Latency Attribution Model for Multi-Tenant LLM Serving
LLMVisor：用于多租户大语言模型服务的实时延迟归因模型
Abstract
As LLM inference shifts to multi-tenant GPU clusters, co-batching improves throughput but obscures per-tenant usage and limits control. Enabling fractional sharing of the inference engine requires a real-time, per-request attribution primitive that is accurate and light enough to run inside the scheduling loop. We present LLMVisor, a roofline-guided latency attribution model that captures the memory-bound and compute-bound phases via a concise piecewise-linear form over features proportional to FLOPs and memory I/O traffic. LLMVisor decomposes batch latency into additive, per-request shares and runs efficiently at microsecond scale. We evaluate LLMVisor across Llama 3.1-8B and Qwen 2.5-14B/32B on A100/H100 GPUs under varying tensor parallelism and workload mixes. Compared to a token-count baseline, LLMVisor attains near-perfect R-squared and reduces relative error by up to 2.5x and 3.3x at p90 and p99, respectively, for prefill, and by up to 3.5x and 4.4x for decode, despite batching variability and sequence divergence.
Chinese Translation
随着大语言模型（LLM）推理转向多租户GPU集群，联合批处理提高了吞吐量，但模糊了每个租户的使用情况并限制了控制能力。实现推理引擎的分数共享需要一种实时的、每个请求的归因原语，该原语必须足够准确且轻量，以便在调度循环内运行。我们提出了LLMVisor，这是一种基于屋顶线（roofline）指导的延迟归因模型，通过一个简洁的分段线性形式捕捉与FLOPs和内存I/O流量成比例的特征，从而捕捉内存绑定和计算绑定阶段。LLMVisor将批处理延迟分解为可加的、每个请求的份额，并在微秒级别高效运行。我们在A100/H100 GPU上评估了LLMVisor，使用Llama 3.1-8B和Qwen 2.5-14B/32B，测试了不同的张量并行性和工作负载组合。与基于令牌计数的基线相比，LLMVisor实现了近乎完美的R平方，并在p90和p99的预填充相对误差分别降低了最多2.5倍和3.3倍，在解码时则分别降低了最多3.5倍和4.4倍，尽管存在批处理变异性和序列发散。
cs.AI / 96 / 2608.08389
Not Worth Another Token: Marginal Value Estimation for Efficient Deep Research Agents
不值得再增加一个标记：高效深度研究代理的边际价值估计
Abstract
Long-horizon research agents solve open-ended tasks through iterative retrieval, aggregation, and synthesis, but context grows rapidly while the marginal value of additional evidence often declines. This leads to unnecessary token cost, higher latency, and noisier inputs for final report generation. We study marginal value estimation for context management in deep research agents and present the first systematic stage-aware comparison of pruning strategies across the pipeline. We evaluate lightweight heuristic criteria and a learned value model at pre-retrieval, post-retrieval, and pre-synthesis stages. Our results show that pruning effectiveness depends more on where pruning is applied than on the specific scoring rule: early pruning yields the largest end-to-end savings, while later pruning mainly refines the final synthesis context. Lightweight heuristics reduce token usage by up to 73% with little quality degradation, learned pruning remains competitive on selected trade-offs, and no single method dominates across quality, efficiency, and faithfulness. These findings provide practical guidance for designing efficient long-horizon agentic systems.
Chinese Translation
长时间跨度的研究代理通过迭代检索、聚合和综合来解决开放式任务，但上下文快速增长，而额外证据的边际价值往往下降。这导致不必要的标记成本、更高的延迟以及最终报告生成的输入噪声。我们研究了深度研究代理中上下文管理的边际价值估计，并首次系统性地比较了管道中修剪策略的阶段感知。我们在预检索、后检索和预综合阶段评估了轻量级启发式标准和学习的价值模型。我们的结果表明，修剪的有效性更多地依赖于修剪应用的位置，而不是具体的评分规则：早期修剪带来了最大的端到端节省，而后期修剪主要细化最终综合的上下文。轻量级启发式方法在几乎没有质量下降的情况下将标记使用减少了多达73%，学习修剪在选定的权衡上保持竞争力，并且没有单一方法在质量、效率和忠实度上占主导地位。这些发现为设计高效的长时间跨度代理系统提供了实用指导。
cs.AI / 97 / 2608.08392
CAP: A Scalable Benchmark for Evaluating Cross-Site Browser Agents with Complex Actions and Perception
CAP：评估具有复杂动作和感知能力的跨站浏览器代理的可扩展基准
Abstract
Large language models are increasingly deployed as autonomous agents that interact with the web through browsers. While recent progress has been driven by benchmarks that evaluate end-to-end task success, these evaluations largely overlook two fundamental sources of difficulty in real web browsing: complex actions over rich user interfaces and visual perception of dynamically rendered content, especially in workflows that span multiple websites. We introduce CAP, a scalable benchmark for evaluating browser agents on cross-site, human-like web tasks that require non-trivial UI interactions and visual understanding. Specifically, we adopt a decomposition-and-recomposition pipeline that first abstracts each website into a structured site card capturing user-facing functions, complex execution operations, and perceptual requirements, and then recomposes these components into realistic cross-site workflows. Each task is therefore grounded in multiple specific operations on each website, enabling fine-grained diagnosis. Built on this framework, we construct 420 tasks across 108 real-world websites and 24 domains under careful quality control. Experiments on state-of-the-art browser agents using our verifiable agent-as-a-judge evaluation framework show low success rates and reveal that perception-heavy interactions remain a major bottleneck, exposing substantial gaps between current agents and real-world web browsing demands.
Chinese Translation
大型语言模型越来越多地被部署为自主代理，通过浏览器与网络互动。尽管最近的进展主要依赖于评估端到端任务成功的基准，但这些评估在很大程度上忽视了真实网页浏览中的两个基本困难来源：在丰富用户界面上的复杂操作以及对动态渲染内容的视觉感知，尤其是在跨多个网站的工作流程中。我们提出了CAP，这是一个可扩展的基准，用于评估在跨站、类人网页任务中需要非平凡用户界面交互和视觉理解的浏览器代理。具体而言，我们采用了一个分解与重组的流程，首先将每个网站抽象为一个结构化的站点卡，捕捉用户面向的功能、复杂执行操作和感知需求，然后将这些组件重组成现实的跨站工作流程。因此，每个任务都基于每个网站上的多个具体操作，使得细粒度的诊断成为可能。在此框架下，我们在108个真实网站和24个领域中构建了420个任务，并进行了严格的质量控制。使用我们可验证的代理作为评估者的评估框架对最先进的浏览器代理进行的实验显示成功率较低，并揭示了以感知为重的交互仍然是一个主要瓶颈，暴露了当前代理与现实网页浏览需求之间的显著差距。
cs.AI / 98 / 2608.08398
Estimating Uncertainty in Galaxy Morphology Classification
估计星系形态分类中的不确定性
Abstract
Astronomers classify galaxy morphology to investigate cosmic evolution. While deep foundation models are increasingly utilized in Galaxy Morphology Classification (GMC), little work has been done on evaluating the uncertainty of GMC results. Uncertainty evaluation is important because astronomical data are inherently noisy due to instrumental and environmental limitations. Also, the continuous evolution of galaxies creates intrinsic morphological ambiguity. However, current foundation models operate as deterministic point estimators, failing to quantify the uncertainty. To overcome this limitation, we propose UEGMC, a post-hoc framework of Uncertainty Estimation for Galaxy Morphology Classification. It categorizes uncertainty in GMC into distinct types by model parameters, astronomical data, reference standards, or intrinsic physical ambiguities, thereby facilitating better classification. Our framework can directly predict uncertainties from representations extracted from the frozen backbones of foundation models, without computationally expensive sampling, therefore enabling fine-grained uncertainty evaluations. Our experimental results demonstrate that UEGMC provides competitive uncertainty quantification performance compared with previous methods.
Chinese Translation
天文学家通过对星系形态的分类来研究宇宙演化。尽管深度基础模型在星系形态分类（Galaxy Morphology Classification, GMC）中越来越多地被应用，但对GMC结果不确定性的评估工作仍然较少。不确定性评估至关重要，因为天文数据由于仪器和环境的限制本质上是噪声的。此外，星系的持续演化也造成了内在的形态模糊性。然而，目前的基础模型作为确定性点估计器，未能量化不确定性。为克服这一限制，我们提出了UEGMC（不确定性估计的星系形态分类后处理框架）。该框架将GMC中的不确定性根据模型参数、天文数据、参考标准或内在物理模糊性分类，从而促进更好的分类。我们的框架可以直接从基础模型的冻结主干中提取的表示中预测不确定性，而无需进行计算成本高昂的采样，因此能够实现细粒度的不确定性评估。我们的实验结果表明，UEGMC在不确定性量化性能上与之前的方法相比具有竞争力。
cs.AI / 99 / 2608.08445
Forgotten History or Test-of-Time? Retrospect and Prospect on RAG from an IR Perspective
被遗忘的历史还是经得起时间考验？从信息检索的视角回顾与展望RAG
Abstract
Retrieval-Augmented Generation (RAG) is widely regarded as a novel paradigm born from the limitations of large language models (LLMs)--a mechanism to ground their outputs in external knowledge. This view, however, is incomplete when considered within a broader historical context. In this paper, we argue that the core ideas underlying RAG are not new: foundational concepts such as integrating retrieval and language generation, knowledge augmentation, answer verification, and iterative query (or prompt) refinement had already been studied and instantiated in information retrieval (IR) and question answering (QA) research dating back to the early 2000s, well before the emergence of LLMs. We make this case by systematically tracing the intellectual lineage of modern RAG and Agentic RAG back to their classical IR and QA antecedents, and examining why this continuity has gone under-recognized -- a consequence of community fragmentation, shifting terminology, and the recency bias endemic to fast-moving fields. Rather than treating LLMs as the origin point of retrieval-augmented intelligence, we propose viewing them as a new interface layer atop a decades-old QA architecture. This reframing is not merely historical: by situating RAG within the longer trajectory of IR research, we surface underutilized prior work -- on user modeling, answer validation, and query refinement -- that can directly inform next-generation RAG design, reducing unintentional rediscovery and fostering genuine cross-community integration.
Chinese Translation
检索增强生成（Retrieval-Augmented Generation, RAG）被广泛视为一种新兴范式，源于大型语言模型（Large Language Models, LLMs）的局限性——一种将其输出与外部知识相结合的机制。然而，从更广泛的历史背景来看，这一观点是不完整的。本文论证了RAG背后的核心思想并不新颖：诸如检索与语言生成的结合、知识增强、答案验证和迭代查询（或提示）优化等基础概念，早在2000年代初期的信息检索（Information Retrieval, IR）和问答（Question Answering, QA）研究中就已被研究和实现，这比LLMs的出现早得多。我们通过系统追溯现代RAG和Agentic RAG的知识渊源，回到其经典的IR和QA前身，并考察为何这种连续性未受到足够重视——这源于社区的碎片化、术语的变化以及快速发展的领域中普遍存在的近期偏见。我们建议将LLMs视为建立在数十年问答架构之上的新接口层，而不是检索增强智能的起源点。这种重新框架不仅仅是历史性的：通过将RAG置于IR研究的更长发展轨迹中，我们揭示了未被充分利用的先前工作——关于用户建模、答案验证和查询优化——这些工作可以直接为下一代RAG设计提供指导，减少无意的重新发现，促进真正的跨社区整合。
cs.AI / 100 / 2608.08446
TRACE-Memory: Public-Conditioned Retrieval and Utility-Aware Evidence Admission for Personalized Generation
TRACE-Memory：基于公共条件的检索与实用性意识的证据接纳用于个性化生成
Abstract
Personalized generation systems retrieve user history by request--memory relevance and inject it into the model context. Yet relevant history may concern the wrong preference aspect, duplicate public information, or provide insufficient support. We argue that personal memory should be used only when it adds utility beyond a public-only response. We propose TRACE-Memory, a two-stage framework for selective personalization. Stage 1 queries for user-specific information missing from the request and public context, then retrieves a coverage-oriented candidate pool. Stage 2 admits a compact subset of source-traceable evidence units, or the empty set, according to response-level incremental utility. We progressively train the query-generation and evidence-admission policies through structured SFT initialization, reduced-space stage-wise GRPO warm-up, and nested multi-sample Joint GRPO. Across 4,500 Controlled and Natural tasks from Goodreads, Amazon Reviews, and Reddit, TRACE-Memory consistently outperforms random and lexical memory use, improves over semantic retrieval, remains competitive with frontier-LLM memory pipelines as local generator capacity increases, and conditions evidence admission on public-context sufficiency, supporting selective rather than default personalization.
Chinese Translation
个性化生成系统通过请求-记忆相关性检索用户历史，并将其注入模型上下文。然而，相关历史可能涉及错误的偏好方面、重复的公共信息或提供不足的支持。我们认为，个人记忆仅应在其提供的实用性超出公共响应时使用。我们提出了TRACE-Memory，一个用于选择性个性化的两阶段框架。第一阶段查询请求和公共上下文中缺失的用户特定信息，然后检索一个覆盖导向的候选池。第二阶段根据响应级增量实用性接纳一组紧凑的源可追溯证据单元，或空集。我们通过结构化的SFT初始化、减少空间的阶段性GRPO预热和嵌套多样本联合GRPO逐步训练查询生成和证据接纳策略。在来自Goodreads、Amazon Reviews和Reddit的4500个受控和自然任务中，TRACE-Memory始终优于随机和词汇记忆的使用，改善了语义检索，并在前沿LLM内存管道的本地生成能力增加时保持竞争力，同时根据公共上下文的充分性条件接纳证据，支持选择性而非默认的个性化。
cs.AI / 101 / 2608.08453
What Keeps Agent Skills from Being Reusable? Evidence from 138K SKILL.md Files
是什么阻碍了代理技能的可重用性？来自138K SKILL.md文件的证据
Abstract
Under the current standard, Agent Skills are SKILL.md files that combine instructions with supporting resources, enabling Large Language Model (LLM) agents to reuse procedures beyond a single conversation. Yet many public skills appear to originate from a single task, repository, or conversation, even when they are shared as reusable components. We analyze this gap across 138,133 public SKILL.md files from 20,556 repositories using a two-tier defect taxonomy grounded in the official specification and best-practice guidance. We find that 91.8% of skills contain at least one detected defect, with stable estimates across lenient and strict thresholds (88.8-94.6%). The dominant failures are ordinary packaging problems rather than exotic attacks: weak routing metadata, bloated or non-actionable bodies, and poor resource organization. A deterministic routing stress test over 20,000 skills shows the functional impact: skills with valid routing metadata are retrieved more reliably from startup descriptions than skills with routing defects. Defect rates vary by platform and provenance: specification-aware skills contain fewer defects, while AI-marked skills show more safety and portability problems. Lightweight enforcement and repair experiments support a quality-assured generation workflow combining spec-aware prompting, lightweight linting, automated repair, and safety gating.
Chinese Translation
根据当前标准，代理技能是结合指令和支持资源的SKILL.md文件，使大型语言模型（LLM）代理能够在单次对话之外重用程序。然而，许多公共技能似乎源自单一任务、仓库或对话，即使它们被共享为可重用组件。我们使用基于官方规范和最佳实践指导的双层缺陷分类法，分析了来自20,556个仓库的138,133个公共SKILL.md文件中的这一差距。我们发现91.8%的技能至少包含一个检测到的缺陷，在宽松和严格阈值下的估计相对稳定（88.8-94.6%）。主要的失败是普通的打包问题，而非复杂的攻击：弱路由元数据、臃肿或不可操作的主体以及资源组织不良。对20,000个技能进行的确定性路由压力测试显示了功能影响：具有有效路由元数据的技能比具有路由缺陷的技能更可靠地从启动描述中检索。缺陷率因平台和来源而异：了解规范的技能包含较少的缺陷，而标记为AI的技能则显示出更多的安全性和可移植性问题。轻量级的执行和修复实验支持了一种质量保证的生成工作流程，结合了规范意识提示、轻量级代码检查、自动修复和安全门控。
cs.AI / 102 / 2608.08466
Hierarchical Self-Improvement: A Framework for Task-Specific Evolvable Agent Harnesses
层次自我改进：一种任务特定的可进化代理框架
Abstract
Modern LLM agents are often improved by modifying prompts, tools, or workflows manually, while the executable scaffold surrounding the model---the \emph{harness}---is typically treated as a fixed artifact after deployment. This work studies an alternative where the harness is \emph{task-specific and continuously evolvable}: each task family maintains its own harness, which is hot-swapped across iterations through a fixed task-injection seam and rewritten using environment feedback. We introduce \textbf{Hierarchical Self-Improvement (HSI)}, a framework in which a single frozen LLM $M$ operates across three hierarchical scopes: a task harness $H$ that executes tasks, an evolver that rewrites $H$, and a meta-evolver that rewrites the evolver's strategy code under a frozen outer anchor. A thinking-on/off design isolates the contribution of harness evolution by disabling reasoning during task execution while enabling it during self-modification. HSI is bounded by two factors: a \emph{feedback-fidelity bound}, since evolution requires informative reward signals to guide selection, and a \emph{backbone capability bound}, since harness redesign cannot overcome limitations of the frozen model. On BALROG with DeepSeek-V4-Flash-Preview as the frozen backbone, HSI achieves consistent gains over the initial harness on moderate-difficulty tasks ($+39.3$ on BabyAI, $+33.0$ on Crafter, $+25.0$ on TextWorld, and $+15.0$ on MiniHack, all in raw \% Progress), while obtaining strong held-out generalization on BabaIsAI sub-suites ($0.98$ best-test on BreakStop and $1.00$ on GoTo from a $20\%$ unseen split). On tasks beyond the backbone's capability (NLE), harness evolution provides no improvement. These results demonstrate task-specific harness evolution as a viable axis for improving frozen LLM agents under clear empirical limits. Code is available at https://github.com/TailinZhou/hsi.
Chinese Translation
现代大型语言模型（LLM）代理通常通过手动修改提示、工具或工作流程来进行改进，而围绕模型的可执行支架—— extit{harness}——在部署后通常被视为固定的工件。本研究探讨了一种替代方案，其中支架是 extit{任务特定且可持续进化的}：每个任务家族维护其自己的支架，该支架通过固定的任务注入接缝在迭代中热插拔，并使用环境反馈进行重写。我们引入了 extbf{层次自我改进（HSI）}框架，其中一个冻结的LLM $M$ 在三个层次范围内运行：执行任务的任务支架 $H$、重写 $H$ 的进化器，以及在冻结的外部锚点下重写进化器策略代码的元进化器。思维开关设计通过在任务执行期间禁用推理而在自我修改期间启用推理，来隔离支架进化的贡献。HSI 受到两个因素的限制： extit{反馈保真度限制}，因为进化需要信息丰富的奖励信号来指导选择，以及 extit{骨干能力限制}，因为支架重新设计无法克服冻结模型的局限性。在使用 DeepSeek-V4-Flash-Preview 作为冻结骨干的 BALROG 上，HSI 在中等难度任务上相较于初始支架取得了一致的提升（在 BabyAI 上 +39.3，在 Crafter 上 +33.0，在 TextWorld 上 +25.0，以及在 MiniHack 上 +15.0，均为原始 extit{ extperthousand Progress}），同时在 BabaIsAI 子套件上获得了强大的持出泛化（在 BreakStop 上最佳测试为 $0.98$，在 GoTo 上为 $1.00$，来自 $20 ext{%}$ 未见分割）。在超出骨干能力的任务（NLE）上，支架进化没有提供改进。这些结果表明，任务特定的支架进化是改善冻结 LLM 代理的一个可行方向，且在明确的实证限制下有效。代码可在 https://github.com/TailinZhou/hsi 获取。
cs.AI / 103 / 2608.08467
LLM within MCP Matters: Measuring Inefficient Resource Utilization Driven by LLMs
MCP 中的 LLM 重要性：测量由 LLM 驱动的低效资源利用
Abstract
The Model Context Protocol (MCP) standardizes how servers expose data and tools to Large Language Models (LLMs). A common server design embeds frequently used reference data, such as identifier lookup tables, directly in the server instructions: the system-prompt text a server hands to the host application. When a query concerns an entry of the embedded table, the model can act on it immediately instead of re-discovering the same information through a search tool. We test whether client LLMs actually consume such instruction-embedded data, reporting a 54,000-trial study across 24 LLMs (9 Claude, 6 Gemini, 9 GPT) on a production legal-information MCP server. A diagnostic condition that removes the competing search tool shows that failures are dominated by behavioral preference rather than missing capability. With search unavailable, 23 of 24 models read the embedded data reliably (hit ratio at least 98%); with a search tool merely present, 9 models drop below 15%. A 2^3 factorial analysis of three instruction-level interventions reveals strong interaction effects: combining all three restores at least 86% for 20 of 24 models, but individual interventions can backfire for specific model families. Per-server prompt engineering is therefore a workaround rather than a fix; we argue that MCP host applications should provide an explicit mechanism that places server instructions ahead of tool selection in the client LLM's deliberation.
Chinese Translation
模型上下文协议（Model Context Protocol, MCP）标准化了服务器如何向大型语言模型（Large Language Models, LLMs）暴露数据和工具。常见的服务器设计将频繁使用的参考数据，如标识符查找表，直接嵌入服务器指令中：即服务器提供给主机应用程序的系统提示文本。当查询涉及嵌入表中的条目时，模型可以立即对其进行操作，而无需通过搜索工具重新发现相同的信息。我们测试了客户端 LLM 是否实际消耗这些嵌入指令数据，报告了在一个生产法律信息 MCP 服务器上对 24 个 LLM（9 个 Claude，6 个 Gemini，9 个 GPT）进行的 54,000 次试验研究。去除竞争搜索工具的诊断条件表明，失败主要由行为偏好主导，而非能力缺失。在搜索不可用的情况下，24 个模型中有 23 个可靠地读取嵌入数据（命中率至少为 98%）；而仅有搜索工具存在时，9 个模型的命中率降至 15% 以下。对三种指令级干预的 2^3 因子分析显示出强烈的交互效应：将三者结合可以为 24 个模型中的 20 个恢复至少 86%的性能，但单独的干预可能会对特定模型系列产生反效果。因此，每个服务器的提示工程更多是一种权宜之计而非解决方案；我们认为 MCP 主机应用程序应提供一种明确机制，使服务器指令在客户端 LLM 的决策中优先于工具选择。
cs.AI / 104 / 2608.08469
Aero Realtime: Fully Aligned Input-Output Streams for Low-Latency Streaming Multimodal Generation
Aero Realtime：低延迟流媒体多模态生成的完全对齐输入输出流
Abstract
Existing streaming multimodal models process observations incrementally but still follow a turn-based prefill-then-decode pattern, making them non-duplex: new observations cannot naturally enter an active generation stream. Proactive alternatives use micro-turn polling or external response gates, which fragment continuous interaction, decouple response timing from language generation, and complicate KV-cache-friendly serving. We introduce Aero Realtime, a 4B streaming multimodal model with a duplex architecture for realtime generation. Aero Realtime aligns video, audio, and textual output on a shared temporal grid, where each approximately 80-ms audio slot predicts either a lexical token or a silence token. This allows input and output to advance together, enabling one autoregressive objective to learn both when to respond and what to generate. During inference, Aero Realtime appends only the newest multimodal slot, carries forward the previous output state, and reuses the KV cache for efficient incremental execution. We further provide a complete training and serving recipe, including realtime QA construction, slot-aligned supervision, hardware-aware distributed training, and resumable inference. On four NVIDIA A6000 workstation GPUs, Aero Realtime maintains 84-ms median and 173-ms P95 processing lag over 20 minutes of a continuously streamed video, remaining within 200~ms of the source timeline. These results demonstrate the feasibility of fully aligned input-output modeling for duplex, proactive, and hardware-aligned multimodal interaction.
Chinese Translation
现有的流媒体多模态模型以增量方式处理观察数据，但仍然遵循基于回合的预填充然后解码模式，使其成为非双工的：新的观察数据无法自然地进入活跃的生成流。主动替代方案使用微回合轮询或外部响应门，这会破坏连续交互，将响应时机与语言生成解耦，并使得KV缓存友好的服务变得复杂。我们提出了Aero Realtime，一个具有双工架构的4B流媒体多模态模型，用于实时生成。Aero Realtime在共享的时间网格上对齐视频、音频和文本输出，其中每个大约80毫秒的音频时隙预测一个词汇标记或一个静音标记。这使得输入和输出能够同步推进，使得一个自回归目标能够同时学习何时响应和生成什么。在推理过程中，Aero Realtime仅附加最新的多模态时隙，保留先前的输出状态，并重用KV缓存以实现高效的增量执行。我们进一步提供了完整的训练和服务方案，包括实时问答构建、时隙对齐监督、硬件感知的分布式训练和可恢复的推理。在四个NVIDIA A6000工作站GPU上，Aero Realtime在20分钟的连续流媒体视频中保持84毫秒的中位数和173毫秒的P95处理延迟，始终在源时间线的200毫秒范围内。这些结果证明了完全对齐输入输出建模在双工、主动和硬件对齐的多模态交互中的可行性。
cs.AI / 105 / 2608.08471
Yesterday's Shield, Today's Spear: A Self-Evolving Safety Guardrail in Production
昨日的护盾，今日的矛：生产中的自我演化安全护栏
Abstract
Deployed LLM safety guardrails are predominantly static: trained once and frozen at release, while new jailbreak techniques and previously un-addressed harmful categories emerge within days, leaving the defense perpetually a step behind. We present SESG (Self-Evolving Safety Guardrails), a multi-agent system running in production. SESG monitors the live traffic behind a deployed guardrail and surfaces two classes of failure: jailbreaks novel in form and harmful categories novel in content. Once a failure is confirmed, a generation agent synthesizes paired training data targeted at it; a validation agent rebalances the batch toward the direction in which the deployed model errs, so that the model's own mistakes steer its training set; and a routing agent matches the training action to the diagnosed gap and returns the next version to production. Over six rounds of live evolution (V0 to V6), a 1.7B guardrail adapts to a new threat in 16-24 hours, with about 2 hours of human effort, versus the 40-90 hours of the manual process it replaces. On six emerging threats, it outperforms static guardrails from 0.6B to 9B and an adaptive baseline while preserving its general screening competence. Since April 2026, SESG has been the primary update pipeline of Sangfor's guardrail, autonomously closing 14 of 15 new threat scenarios in two months. We release 9 test sets for the 6 new threats at https://github.com/Trams1017/SESG. Warning: This paper contains examples that may be harmful or offensive.
Chinese Translation
部署的LLM安全护栏主要是静态的：训练一次后在发布时被冻结，而新的越狱技术和之前未解决的有害类别在几天内出现，使得防御始终滞后于威胁。我们提出了SESG（自我演化安全护栏），这是一个在生产中运行的多智能体系统。SESG监控部署护栏后面的实时流量，并识别出两类失败：形式新颖的越狱和内容新颖的有害类别。一旦确认失败，生成代理会合成针对该失败的配对训练数据；验证代理会将批次重新平衡到部署模型出错的方向，以便模型自身的错误引导其训练集；路由代理则将训练动作与诊断出的差距匹配，并将下一个版本返回到生产中。在六轮实时演化（V0到V6）中，一个1.7B的护栏在16-24小时内适应新的威胁，所需的人力约为2小时，而替代的手动过程则需40-90小时。在六个新兴威胁中，它的表现优于从0.6B到9B的静态护栏和一个自适应基线，同时保持其一般筛选能力。自2026年4月以来，SESG已成为Sangfor护栏的主要更新管道，在两个月内自主关闭了15个新威胁场景中的14个。我们在https://github.com/Trams1017/SESG发布了6个新威胁的9个测试集。警告：本文包含可能有害或冒犯的示例。
cs.AI / 106 / 2608.08485
HoloAegis: Frozen Representation, Topological Inference: Minimally Parametric Safety Manifolds for Zero-Shot LLM Guardrails
HoloAegis：冻结表示、拓扑推理：用于零-shot LLM 保护措施的最小参数安全流形
Abstract
Current LLM safety guardrails face a fundamental tension: fine-tuning distorts pre-trained representations while generative judges incur prohibitive inference costs. We challenge the prevailing paradigm by asking: can safety be achieved through pure geometric reasoning over frozen semantic representations? We present HoloAegis, a minimally parametric topological inference framework that decouples representation from reasoning. We term our approach minimally parametric because the only free parameters are the anchor count K and the temperature tau, both fixed after construction and requiring no gradient-based training. An un-fine-tuned encoder maps text to a unit sphere, after which all decisions are purely geometric. We formalize safety evaluation as a Gibbs-Boltzmann Free Energy computation over a pre-computed System Topology Anchor Bank, and we introduce Dual Time-Scale Exponential Moving Averages to detect progressive multi-turn semantic drift. Our key theoretical insight is a Topological Boundary Stability Conjecture: we provide theoretical motivation and strong empirical evidence that sparse anchor centroids stabilize the decision boundary against high-frequency lexical perturbations far better than full vector space methods. Evaluated across 8 benchmarks, HoloAegis achieves state-of-the-art accuracy (1.0000 AUC on AuthenHallu, 0.9802 on HarmBench) with sub-millisecond latency, zero cold-start data, and cross-lingual transfer (0.9758 AUC on Chinese CHIFRAUD).
Chinese Translation
当前的 LLM 安全保护措施面临着根本性的矛盾：微调会扭曲预训练表示，而生成性判断则会产生高昂的推理成本。我们通过提出一个问题来挑战现有的范式：是否可以通过对冻结的语义表示进行纯几何推理来实现安全性？我们提出了 HoloAegis，这是一个最小参数的拓扑推理框架，它将表示与推理解耦。我们称这种方法为最小参数，因为唯一的自由参数是锚点数量 K 和温度 tau，这两者在构建后固定且无需基于梯度的训练。一个未经过微调的编码器将文本映射到单位球体，之后所有决策均为纯几何决策。我们将安全性评估形式化为对预计算的系统拓扑锚库进行的 Gibbs-Boltzmann 自由能计算，并引入双时间尺度指数移动平均来检测渐进的多轮语义漂移。我们的关键理论洞察是拓扑边界稳定性猜想：我们提供了理论动机和强有力的实证证据，表明稀疏锚心在抵御高频词汇扰动方面比全向量空间方法更能稳定决策边界。在 8 个基准测试中评估，HoloAegis 实现了最先进的准确性（AuthenHallu 上 1.0000 AUC，HarmBench 上 0.9802），具有亚毫秒延迟、零冷启动数据和跨语言迁移（在中文 CHIFRAUD 上 0.9758 AUC）。
cs.AI / 107 / 2608.08491
TrustRoboReward: Preference-Ordered Isotonic Score Editing for Multi-Paradigm Robot Reward Models
TrustRoboReward：多范式机器人奖励模型的偏好有序等距评分编辑
Abstract
Reward models are a bottleneck for reinforcement learning in embodied AI. Long-horizon robotic manipulation requires scalable vision feedback beyond handcrafted rewards or task-specific annotations. Existing open-source VLM reward judges like RoboReward adopt simple 1--5 trajectory progress scoring, lacking pairwise preferences for RLHF, DPO and Bradley-Terry frameworks, while failing to optimize video scene understanding. Augmenting RoboReward with pairwise comparison and video-QA supervision causes inconsistency between pairwise preferences and pointwise scores, introducing training noise and hurting downstream performance---an issue aggregation methods such as TrustJudge cannot resolve. To address this, we propose TrustRoboReward, a multi-paradigm reward modeling framework equipped with Preference-Ordered Isotonic Score Editing (POISE). We construct a unified four-paradigm dataset with trajectory progress scoring (Score-A), video-QA answer quality scoring (Score-B), and their pairwise counterparts (Pair-A, Pair-B). Pairwise labels align better with human judgment than pointwise scores, inspiring us to calibrate pointwise scores to avoid score-pair reversals against pairwise preferences. POISE rectifies pointwise scores and eliminates cross-paradigm reversal conflicts unresolved by TrustJudge. Theoretically, POISE reduces score-pair reversal conflicts from 20.15% to 0%, whereas TrustJudge retains 20.46% conflicts on the same corpus. Evaluated on our benchmark, Qwen3-VL-4B trained with POISE achieves an overall reward score of 77.96%, nearly matching GPT-5-mini (78.09%, gap 0.13%) and outperforming the strongest RoboReward-4B baseline by 10.13%. It also lifts test-time score-pair consistency to 71.90%, exceeding RoboReward-4B (57.26%) and GPT-5-mini (68.09%). Integrating TrustJudge aggregation during inference boosts the overall score to 78.57%, surpassing the GPT-5-mini teacher model.
Chinese Translation
奖励模型是具身人工智能中强化学习的瓶颈。长时间跨度的机器人操作需要可扩展的视觉反馈，超越手工设计的奖励或特定任务的注释。现有的开源视觉语言模型（VLM）奖励评估工具如RoboReward采用简单的1-5轨迹进展评分，缺乏用于强化学习人类反馈（RLHF）、直接偏好优化（DPO）和Bradley-Terry框架的成对偏好，同时未能优化视频场景理解。将RoboReward与成对比较和视频问答（video-QA）监督相结合导致成对偏好与逐点评分之间的不一致，引入训练噪声并损害下游性能——这是像TrustJudge这样的聚合方法无法解决的问题。为了解决这个问题，我们提出了TrustRoboReward，一个配备偏好有序等距评分编辑（POISE）的多范式奖励建模框架。我们构建了一个统一的四范式数据集，包括轨迹进展评分（Score-A）、视频问答答案质量评分（Score-B）及其成对对应（Pair-A，Pair-B）。成对标签与人类判断的对齐程度优于逐点评分，这激励我们校准逐点评分，以避免与成对偏好相反的评分对调。POISE修正逐点评分，消除了TrustJudge未能解决的跨范式对调冲突。从理论上讲，POISE将评分对调冲突从20.15%降低到0%，而TrustJudge在同一语料库上保留20.46%的冲突。在我们的基准测试中，使用POISE训练的Qwen3-VL-4B的整体奖励评分达到77.96%，几乎与GPT-5-mini（78.09%，差距0.13%）相匹配，并且比最强的RoboReward-4B基线高出10.13%。它还将测试时评分对的一致性提升至71.90%，超过RoboReward-4B（57.26%）和GPT-5-mini（68.09%）。在推理过程中集成TrustJudge聚合将整体评分提升至78.57%，超越了GPT-5-mini教师模型。
cs.AI / 108 / 2608.08503
MathShikkha: A Controlled Study of Answer-Only and Chain-of-Thought Supervision for Bangla Mathematical Reasoning in Small Language Models
MathShikkha：针对小型语言模型的仅答案与思维链监督在孟加拉数学推理中的对照研究
Abstract
Mathematical reasoning remains challenging in low-resource languages such as Bangla. We study whether teacher-generated Bangla Chain-of-Thought (CoT) supervision provides benefits beyond ordinary supervised fine-tuning. We construct \textsc{MathShikkha}, a Bangla mathematical reasoning dataset with GPT-5.4-generated rationales, and fine-tune four 4B--7B student models under a matched protocol in which answer-only and CoT conditions share data splits, response-only loss masking, decoding, and scoring, differing only in the training target. In-domain, CoT provides no significant improvement over answer-only fine-tuning for three stronger backbones (paired bootstrap 95\% CIs include zero; exact McNemar $p \geq 0.17$), despite generating 15--52$\times$ more tokens, but significantly improves the weaker 4B model by 18.56 points ($p < 0.0001$). On the larger, contamination-audited BanglaMATH benchmark, this pattern reverses: CoT significantly outperforms answer-only supervision for all four models by 20.1--28.1 points (all $p < 0.0001$). Answer-only fine-tuning also reduces out-of-domain accuracy below the base model for three models, whereas CoT preserves or improves it for all four. A human study with two co-author annotators, external-expert adjudication, and Cohen's $\kappa = 0.76$--$1.00$ finds no significant CoT improvement over the base model on reasoning-content criteria; instead, its measurable effect is target-language adherence and producing inspectable reasoning. Overall, rationale supervision's value depends on backbone capability and distribution shift: in this setting, its main benefits are Bangla adherence, auditable reasoning, and out-of-domain robustness rather than improved in-domain reasoning validity.
Chinese Translation
在低资源语言如孟加拉语中，数学推理仍然具有挑战性。我们研究教师生成的孟加拉思维链（Chain-of-Thought, CoT）监督是否提供超越普通监督微调的好处。我们构建了 extsc{MathShikkha}，一个包含GPT-5.4生成的推理的孟加拉数学推理数据集，并在一个匹配协议下微调了四个4B--7B的学生模型，其中仅答案和CoT条件共享数据分割、响应仅损失掩蔽、解码和评分，仅在训练目标上有所不同。在领域内，对于三个更强的基础模型，CoT并未显著改善仅答案微调的效果（配对自助法95 ext{%}置信区间包含零；确切的McNemar $p
geq 0.17$），尽管生成了15--52倍更多的标记，但显著提高了较弱的4B模型18.56分（$p < 0.0001$）。在更大且经过污染审计的BanglaMATH基准上，这一模式发生逆转：CoT在所有四个模型上显著超越仅答案监督20.1--28.1分（所有$p < 0.0001$）。仅答案微调还将三个模型的域外准确率降低到低于基础模型，而CoT则为所有四个模型保持或提高了准确率。与两位共同作者注释员、外部专家裁决和Cohen's $ ext{kappa} = 0.76$--$1.00$的人工研究发现，在推理内容标准上，CoT对基础模型没有显著改善；相反，其可测量的效果是目标语言的遵循和生成可检查的推理。总体而言，推理监督的价值依赖于基础能力和分布转移：在这种情况下，其主要好处是孟加拉语言的遵循、可审计的推理和域外的鲁棒性，而不是改善领域内推理的有效性。
cs.AI / 109 / 2608.08506
Understanding Calibration and Truncation Error Propagation in Training-Free Low-Rank Compression for LLMs
理解训练无关的低秩压缩中校准与截断误差的传播
Abstract
Training-free low-rank compression frameworks have been gaining prominence for LLM compression given their effectiveness in reducing model parameter count while maintaining task-level accuracy. However, existing SOTA frameworks share two key limitations: (1) residual errors in calibration data activations accumulate across layers during compression, causing misalignment between representations simulated at compression time and those experienced at inference; (2) the assumption that layer importance distribution is preserved post-compression does not hold. Together, these two effects introduce misalignment in the compression process in relation to the deployed model. We study these effects and propose a simple, training-free methodology compatible with existing frameworks to mitigate them, comprising: (1) Layer-by-Layer Compression with Calibration Correction; (2) Iterative Compression with Rank Allocation Correction. Implemented atop an existing SOTA decomposition framework, and evaluated on Llama and Qwen3 models across various benchmarks and compression rates, our approach demonstrates up to ~1-2.5 accuracy point improvements over per-weight and joint decomposition baselines on zero-shot tasks.
Chinese Translation
无训练的低秩压缩框架因其在减少模型参数数量的同时保持任务级准确性而在大规模语言模型（LLMs）压缩中日益受到关注。然而，现有的最先进（SOTA）框架存在两个主要限制：（1）在压缩过程中，校准数据激活中的残差误差在各层之间累积，导致压缩时模拟的表示与推理时经历的表示之间产生不对齐；（2）假设压缩后层重要性分布保持不变的前提并不成立。这两种效应共同导致压缩过程与部署模型之间的不对齐。我们研究了这些效应，并提出了一种简单的、与现有框架兼容的无训练方法来缓解这些问题，包括：（1）逐层压缩与校准修正；（2）带有秩分配修正的迭代压缩。在现有的最先进分解框架上实施，并在Llama和Qwen3模型上针对各种基准和压缩率进行评估，我们的方法在零-shot任务上显示出比每权重和联合分解基线高出约1-2.5的准确度提升。
cs.AI / 110 / 2608.08512
Time Present and Time Past: Benchmarking Large Language Models on Temporally Evolving Document Understanding
现在的时间与过去的时间：对大型语言模型在时间演变文档理解上的基准测试
Abstract
Evolving documents, such as laws, tax codes, and software documentation, are amended, replaced, and sometimes reverted over time, so a question has different correct answers at different dates. In contrast to encyclopedic knowledge, where an old fact is simply overwritten, an amendment is itself an official text that states what it replaces and when it takes effect, and the earlier version stays correct for its validity period. The central challenge is therefore version resolution, that is, identifying the version in force on the queried date. Existing temporal QA datasets treat time only as an annotation, so version resolution stays untested. We present TIDE, an expert-verified benchmark of 3,050 QA pairs over 644 official customs instruments issued between 1969 and 2025 by the Government of Bangladesh, covering eight task types over deeply code-mixed documents that are heterogeneous in layout and dated in two calendars. In addition, we evaluate nine recent LLMs under a single protocol across parametric, gold-context, and retrieval access, scored by a three-judge LLM council with a hard date gate separating correct meaning from correct time. The best macro-averaged accuracy is only 68.5%. Resolving a version from an implicit date reaches 59.7%, and detecting that the supplied version does not govern the query reaches only 26.7%. Models are more likely to find correct versions than to reject incorrect ones, and they tend to follow a confident parametric answer over the supplied authoritative text. All code and data are available at https://github.com/icsetepa44/TIDE
Chinese Translation
演变文档，如法律、税法和软件文档，随着时间的推移被修订、替换，有时甚至被撤回，因此在不同日期下同一个问题可能有不同的正确答案。与百科知识不同，旧事实只是被覆盖，修订本本身就是一份官方文本，说明其替代的内容及生效时间，而早期版本在其有效期内仍然是正确的。因此，核心挑战是版本解析，即识别在查询日期生效的版本。现有的时间问答（QA）数据集仅将时间视为注释，因此版本解析尚未得到测试。我们提出了TIDE，这是一个经过专家验证的基准，包含3050对问答对，涉及1969年至2025年间孟加拉国政府发布的644份官方海关文书，涵盖了八种任务类型，文档在布局上异构且使用两种日历标注日期。此外，我们在一个统一的协议下评估了九个近期的大型语言模型（LLMs），包括参数化、黄金上下文和检索访问，由三位评审的LLM委员会进行评分，设定了一个严格的日期门槛，将正确的意义与正确的时间区分开来。最佳的宏平均准确率仅为68.5%。从隐含日期解析版本的准确率为59.7%，而检测提供的版本不适用于查询的准确率仅为26.7%。模型更有可能找到正确的版本，而不是拒绝错误的版本，并且它们倾向于遵循自信的参数化答案而非提供的权威文本。所有代码和数据可在 https://github.com/icsetepa44/TIDE 获取。
cs.AI / 111 / 2608.08514
Reproducing and Stress-Testing Two Approaches to LLM Reasoning Reliability: Test-Time Probability Aggregation and Logic-Representation Editing
再现和压力测试两种大型语言模型（LLM）推理可靠性的方法：测试时概率聚合和逻辑表示编辑
Abstract
We independently reproduce two recent methods for making large language model (LLM) reasoning more reliable, and stress-test them across domains and models (RPC across four new task domains with Qwen3-8B, LCF across four 7-8B models). The first, RPC, aggregates token probabilities and self-consistency at inference; the second, LCF, trains projectors that split hidden states into "content" and "logic" and edits the logic part toward a valid region. Validating such reliability claims matters because the original evaluations are run by each method's own authors and were never independently reproduced or stress-tested across models and domains, and LCF shipped no public code. We re-run RPC's published-path aggregation and re-implement LCF's projector, contrastive, and intervention pipeline, then extend both to text-to-SQL, legal extraction, fallacy identification, and precedent grading, and probe LCF's representation directly. RPC reproduces the original grid exactly on the authors' released reasoning paths; on four new domains its edge over self-consistency is never significant (ties or small mixed differences, paired p >= 0.28), and on BIRD, the one domain where we vary the budget, the edge grows with K as predicted but its largest gap (+2.5 accuracy at K=32, p=0.16) reverses to -0.25 when we enlarge the sample to n=200. LCF's logic-validity direction is real but weak (0.82 separability at the single best sub-layer versus 0.95 for a semantic-attribute control); its one positive effect (Qwen3 $\Delta$Prob) is not significant (p=0.56), while it significantly reduces $\Delta$Prob on two of the other three models.
Chinese Translation
我们独立再现了两种最近的方法，以提高大型语言模型（LLM）推理的可靠性，并在不同领域和模型中进行了压力测试（RPC在四个新任务领域中使用Qwen3-8B，LCF在四个7-8B模型中进行）。第一种方法RPC在推理时聚合了标记概率和自一致性；第二种方法LCF训练投影器，将隐藏状态分为“内容”和“逻辑”，并将逻辑部分编辑到有效区域。验证这种可靠性声明至关重要，因为原始评估由每种方法的作者自行进行，并且从未在不同模型和领域中独立再现或压力测试，而LCF也没有发布公共代码。我们重新运行了RPC的已发布路径聚合，并重新实现了LCF的投影器、对比和干预管道，然后将两者扩展到文本到SQL、法律提取、谬误识别和先例评分，并直接探测LCF的表示。RPC在作者发布的推理路径上完全再现了原始网格；在四个新领域中，其相对于自一致性的优势从未显著（平局或小的混合差异，配对p >= 0.28），而在BIRD这个我们改变预算的领域中，优势随着K的增加而增长，但其最大差距（K=32时准确率提高2.5，p=0.16）在样本扩大到n=200时反转为-0.25。LCF的逻辑有效性方向是真实但较弱（在单个最佳子层的可分离性为0.82，而语义属性控制为0.95）；其唯一的正面效果（Qwen3 $ riangle$Prob）并不显著（p=0.56），而且在其他三个模型中的两个模型上显著降低了$ riangle$Prob。
cs.AI / 112 / 2608.08523
Discovering Diverse Planning Policies for Multimodal Embodied Agents with Quality-Diversity Optimization
利用质量-多样性优化发现多模态具身智能体的多样化规划策略
Abstract
Multimodal embodied agents are increasingly required to solve long-horizon tasks by integrating visual observations, textual goals, and interaction history into closed-loop decision making. However, state-of-the-art large-model-based planners often rely on a single dominant planning style during execution. Once this execution mode becomes ineffective, the agent may remain stalled for many steps, repeatedly interacting with the environment without making meaningful progress. We address this limitation by proposing a Quality-Diversity (QD) framework for discovering diverse planning policies for multimodal embodied agents. The proposed method treats planning-policy templates as evolvable individuals and organizes them into a behavior-indexed archive rather than collapsing search to a single prompt style. In the offline stage, rollout trajectories are summarized into structured success and failure experiences, which guide policy variation through recombination and experience-guided mutation. The resulting policies are mapped into a behavior space defined by interaction intensity and goal-directedness, and the highest-quality policy in each niche is retained in the archive. In the online stage, the agent executes one policy at a time while monitoring task progress. When persistent stall is detected, the system rolls back to the latest checkpoint and switches to a behaviorally distinct archive policy to resume execution. Experiments on the ThreeDWorld transport benchmark show that the proposed framework improves both task success and interaction efficiency over representative baseline planners. These results suggest that discovering diverse policy repertoires is an effective way to support adaptive multimodal planning and online failure recovery.
Chinese Translation
多模态具身智能体越来越需要通过将视觉观察、文本目标和交互历史整合到闭环决策中来解决长期任务。然而，基于大型模型的最先进规划器在执行过程中往往依赖于单一主导的规划风格。一旦这种执行模式变得无效，智能体可能会在许多步骤中停滞不前，反复与环境交互而没有取得实质性进展。我们通过提出一个质量-多样性（Quality-Diversity, QD）框架来解决这一限制，以发现多模态具身智能体的多样化规划策略。所提出的方法将规划策略模板视为可演化的个体，并将其组织成一个行为索引存档，而不是将搜索压缩到单一的提示风格。在离线阶段，回滚轨迹被总结为结构化的成功和失败经验，这些经验通过重组和经验引导的变异来指导策略变异。最终的策略被映射到由交互强度和目标导向性定义的行为空间中，并在每个小众中保留最高质量的策略在存档中。在在线阶段，智能体一次执行一个策略，同时监控任务进展。当检测到持续停滞时，系统会回滚到最新的检查点，并切换到一个行为上明显不同的存档策略以恢复执行。在ThreeDWorld运输基准上的实验表明，所提出的框架在任务成功率和交互效率上均优于代表性的基线规划器。这些结果表明，发现多样化的策略库是支持自适应多模态规划和在线故障恢复的有效方式。
cs.AI / 113 / 2608.08561
Deep probabilistic logic programming for diagnostic reasoning from incomplete information: A case study in stroke detection
基于深度概率逻辑编程的诊断推理：来自不完整信息的案例研究——中风检测
Abstract
In medical applications, raw data is frequently associated with significant privacy concerns, lending particular importance to the encoding of summary statistics from the literature. On the other hand, deep learning has become an invaluable tool for assessing symptoms based on visual or auditory sensor data. DeepProbLog allows for an extensible neuro-symbolic approach that accommodates connectionist components to analyse patient images within a transparent and rigorous probabilistic framework, namely probabilistic logic programming under the distribution semantics. Framed as a case study in stroke detection from multimodal data, this contribution explores the pathway from summary statistics available in the literature to a DeepProbLog-based diagnostic system. It suggests a workflow using established maximum entropy techniques to complete available probabilistic information and the probabilistic logic programming system ProbLog 2 to move from the entropy-maximising causal model to a discriminative neuro-symbolic model expressible within DeepProbLog. The relative performance of models derived from less complete data is analysed alongside the potential of the probabilistic inductive logic programming system ProbFOIL 2 for compressing large discriminative models, and the perspectives and implications of using DeepProbLog for diagnostic reasoning are discussed.
Chinese Translation
在医学应用中，原始数据常常伴随着显著的隐私问题，这使得从文献中编码汇总统计数据显得尤为重要。另一方面，深度学习已成为评估基于视觉或听觉传感器数据的症状的重要工具。DeepProbLog 提供了一种可扩展的神经符号方法，结合了连接主义组件，以在透明且严谨的概率框架下分析患者图像，即在分布语义下的概率逻辑编程。作为一个基于多模态数据的中风检测案例研究，本研究探讨了从文献中可用的汇总统计数据到基于 DeepProbLog 的诊断系统的路径。它建议使用已建立的最大熵技术来补全可用的概率信息，并利用概率逻辑编程系统 ProbLog 2 从最大熵因果模型转向可在 DeepProbLog 中表达的区分神经符号模型。分析了从较不完整数据中衍生的模型的相对性能，以及概率归纳逻辑编程系统 ProbFOIL 2 在压缩大型区分模型方面的潜力，并讨论了使用 DeepProbLog 进行诊断推理的前景和影响。
cs.AI / 114 / 2608.08569
VoxZip: Semantic-Anchored Temporal KV Cache Compression for Long-Context Audio Inference
VoxZip：基于语义锚定的长上下文音频推理的时间键值缓存压缩
Abstract
Recent advancements in Speech Large Language Models have demonstrated remarkable capabilities in understanding complex audio tasks. Despite this progress, their long-context inference remains severely bottlenecked by prohibitive KV cache memory demands. Existing text-centric compression methods struggle here, often disrupting speech continuity or discarding crucial semantic cues. To address this, we propose VoxZip, a train-free, two-stage semantic-anchored KV cache compression framework. The first stage uses automatic speech recognition (ASR) transcriptions as explicit semantic anchors to temporally align, compress, and fuse audio tokens, significantly reducing the initial KV cache while elevating token information density. To further improve the compression ratio, the second stage employs a dynamic filtering strategy based on temporally decayed accumulated attention to evict non-essential tokens while mitigating early-token bias. Comprehensive evaluations on Qwen3-Omni across six diverse audio benchmarks demonstrate the superiority of our approach. VoxZip excels in long-audio reasoning and consistently maintains high-fidelity perception on short-form tasks. Notably, it sustains over 90\% of the uncompressed baseline performance even under an aggressive 20x KV cache compression in long-context scenarios. Furthermore, at a 4x compression ratio, VoxZip yields a 1.9x increase in inference throughput alongside a 3.3x reduction in peak memory overhead. Code and models will be available at https://github.com/MM-Speech/VoxZip.
Chinese Translation
最近，语音大型语言模型在理解复杂音频任务方面展现了显著的能力。尽管取得了这些进展，但其长上下文推理仍然受到高昂的键值（KV）缓存内存需求的严重制约。现有的以文本为中心的压缩方法在这里面临挑战，常常破坏语音的连贯性或丢失重要的语义线索。为了解决这个问题，我们提出了VoxZip，一个无训练的、两阶段的基于语义锚定的KV缓存压缩框架。第一阶段使用自动语音识别（ASR）转录作为显式语义锚点，来在时间上对齐、压缩和融合音频标记，显著减少初始的KV缓存，同时提高标记信息密度。为了进一步改善压缩比，第二阶段采用基于时间衰减累积注意力的动态过滤策略，以驱逐非必要的标记，同时减轻早期标记偏差。在六个不同的音频基准上对Qwen3-Omni进行的全面评估证明了我们方法的优越性。VoxZip在长音频推理中表现出色，并在短时任务中始终保持高保真感知。值得注意的是，即使在长上下文场景下进行20倍的激进KV缓存压缩时，它仍能维持超过90%的未压缩基线性能。此外，在4倍压缩比下，VoxZip实现了1.9倍的推理吞吐量提升，同时峰值内存开销减少了3.3倍。代码和模型将发布在 https://github.com/MM-Speech/VoxZip。
cs.AI / 115 / 2608.08570
FailForge: Distilling Procedural Competence from Persistent Failures into Code Agents
FailForge：将持久失败中的程序能力提炼为代码代理
Abstract
Rejection sampling fine-tuning (RFT) is widely used to train code agents by generating trajectories on verifiable software engineering tasks, retaining those that pass the tests, and fine-tuning on the successful rollouts. However, even strong code agents repeatedly fail on a substantial fraction of such tasks, and standard RFT simply discards these failures. The discarded samples are precisely the hardest and most informative ones, drawn from verifiable instances that are costly to curate. Stronger base models may reduce the number of failures, but the remaining hard cases still define the frontier for further improvement. We propose FailForge, an agentic framework that converts failed rollouts into training signal. For each failed instance, an agent diagnoses the failure from error feedback and execution traces, distills the diagnosis into a concise and actionable skill, and injects the skill into the agent context for a guided second attempt. Trajectories that succeed under skill guidance are folded back into the RFT corpus. Crucially, the skill is removed at training time, so the model internalizes the recovered behavior rather than relying on external hints at inference. FailForge recovers over 26% of previously failed instances at marginal additional cost, and training Qwen3.5-4B on the augmented corpus improves the SWE-bench Verified resolve rate by 6.6 points over a strong RFT baseline, with gains concentrated on the hardest problems.
Chinese Translation
拒绝采样微调（RFT）被广泛用于通过生成可验证软件工程任务的轨迹来训练代码代理，保留那些通过测试的轨迹，并在成功的回滚上进行微调。然而，即使是强大的代码代理在相当一部分此类任务上也会反复失败，而标准的RFT则简单地丢弃这些失败的样本。被丢弃的样本恰恰是最难且最具信息量的样本，来自于成本高昂的可验证实例。更强大的基础模型可能会减少失败的数量，但剩余的难题仍然定义了进一步改进的边界。我们提出了FailForge，一个将失败的回滚转换为训练信号的代理框架。对于每个失败的实例，代理从错误反馈和执行轨迹中诊断失败，将诊断提炼为简明且可操作的技能，并将该技能注入代理上下文中以指导第二次尝试。在技能指导下成功的轨迹将被重新纳入RFT语料库。至关重要的是，技能在训练时被移除，因此模型内化恢复的行为，而不是在推理时依赖外部提示。FailForge以边际额外成本恢复了超过26%的先前失败实例，并在增强语料库上训练Qwen3.5-4B，使SWE-bench的验证解决率比强大的RFT基线提高了6.6个百分点，增益集中在最难的问题上。
cs.AI / 116 / 2608.08594
SDDBMs: Soft Denoising Diffusion Bridge Models
软去噪扩散桥模型（SDDBMs）
Abstract
Diffusion bridge models leverage Doob's \(h\)-transform to construct stochastic transports between arbitrary endpoint distributions, and have shown strong potential in image-to-image translation and restoration. However, most existing bridge models rely on hard endpoint conditioning, which forces the terminal state to match a prescribed target exactly. This hard constraint induces terminal-boundary singularities: the terminal law collapses to a Dirac measure, and the resulting drift coefficients become ill-conditioned near the endpoint. In this paper, we propose Soft Denoising Diffusion Bridge Models (SDDBMs), a generalized framework that regularizes diffusion bridges directly at the level of their terminal constraints. Instead of imposing an exact endpoint, SDDBMs prescribe a non-degenerate Gaussian terminal marginal under the transformed path measure, with a flexible terminal center and variance. Starting from this prescribed marginal, we develop a complete closed-form construction of the soft bridge, including the Gaussian terminal reweighting and soft \(h\)-function, the induced Gaussian forward marginals and \(\mathbf{x}_0\)-free dynamics. Theoretically, SDDBMs provide a unified probabilistic perspective that encompasses existing diffusion bridge models, including DDBMs, GOUB, and UniDB, as special cases under specific parameter choices. Extensive experiments on image restoration tasks demonstrate that SDDBMs achieve improved numerical stability and superior generation quality over existing bridge-based methods.
Chinese Translation
扩散桥模型利用杜布（Doob）的h变换构建任意端点分布之间的随机传输，并在图像到图像的转换和恢复中显示出强大的潜力。然而，大多数现有的桥模型依赖于硬性端点条件，这迫使终端状态完全匹配规定的目标。这种硬性约束导致终端边界奇异性：终端分布收敛为狄拉克测度，导致生成的漂移系数在端点附近变得病态。在本文中，我们提出了软去噪扩散桥模型（SDDBMs），这是一个在其终端约束层面直接正则化扩散桥的广义框架。SDDBMs并不强制要求精确的端点，而是在变换路径测度下规定一个非退化的高斯终端边际，具有灵活的终端中心和方差。在这个规定的边际基础上，我们开发了软桥的完整封闭形式构造，包括高斯终端重加权和软h函数、诱导的高斯前向边际以及 extbf{x}_0自由动态。从理论上讲，SDDBMs提供了一个统一的概率视角，涵盖了现有的扩散桥模型，包括DDBMs、GOUB和UniDB，在特定参数选择下作为特例。大量关于图像恢复任务的实验表明，SDDBMs在数值稳定性和生成质量上优于现有的基于桥的方法。
cs.AI / 117 / 2608.08601
Unaccountable Delegation, Fading Skills: Mapping the Risks of Workplace AI Agents
不可问责的委托与技能衰退：工作场所人工智能代理风险的映射
Abstract
To anticipate socio-technical risks from AI agents, organizations need taxonomies to classify them. However, existing AI risk taxonomies focus on broad risks and do not capture job-specific risks introduced by agents. To address this gap, we make three main contributions. First, we developed a multi-layer framework from a literature review of AI agents. The framework models three core components and their interactions: agents, goals, and environment. Second, we embedded this framework in a structured prompt and applied it to descriptions of 2,078 job tasks from the O*NET database, producing 8,356 risk scenarios labeled by severity and deployment mode (automation or augmentation). We validated these scenarios with 45 workers across 10 job roles and an independent LLM judge, confirming their plausibility and alignment with job tasks. Finally, we extended an existing taxonomy to create a 15-category taxonomy of workplace AI agent risks that covers all our risk scenarios. Our analysis highlights four findings. First, augmentation is not inherently safe because overreliance on agents can gradually erode workers' skills and oversight. Second, Erroneous Agent Actions accounts for the largest share of risk scenarios and has the highest concentration of severe risks. Many arise at the human-agent boundary. Third, automation is associated mainly with organizational risks, while augmentation is associated mainly with risks to workers. Fourth, workers found our taxonomy easier to use for a risk classification task than two other taxonomies and preferred it in 64% of non-tied comparisons with a recent generative AI risk taxonomy. These findings show that workplace AI agent risks do not arise from agents alone; they also depend on how people work with agents and how agents are deployed. Safer workplaces require not only safer agents but also carefully designed human-AI agent collaboration.
Chinese Translation
为了预见人工智能代理带来的社会技术风险，组织需要分类法来对其进行分类。然而，现有的人工智能风险分类法主要关注广泛的风险，并未捕捉到代理所引入的特定工作风险。为了解决这一空白，我们做出了三项主要贡献。首先，我们基于对人工智能代理的文献综述，开发了一个多层次框架。该框架建模了三个核心组件及其相互作用：代理、目标和环境。其次，我们将该框架嵌入到一个结构化提示中，并将其应用于O*NET数据库中2,078个工作任务的描述，生成了8,356个按严重性和部署模式（自动化或增强）标记的风险场景。我们通过对10个工作角色的45名工人和一位独立的LLM评审者验证了这些场景，确认了其合理性和与工作任务的一致性。最后，我们扩展了现有的分类法，创建了一个涵盖所有风险场景的15类工作场所人工智能代理风险分类法。我们的分析突出了四个发现。首先，增强并非固有安全，因为对代理的过度依赖可能逐渐侵蚀工人的技能和监督能力。其次，错误代理行为占据了风险场景的最大份额，并且具有最高的严重风险集中度。许多风险发生在人与代理的交界处。第三，自动化主要与组织风险相关，而增强主要与工人风险相关。第四，工人在进行风险分类任务时发现我们的分类法比其他两种分类法更易于使用，并在与最近的生成性人工智能风险分类法的64%非平局比较中更受欢迎。这些发现表明，工作场所人工智能代理的风险并不仅仅源于代理本身；它们还依赖于人们如何与代理合作以及代理如何被部署。更安全的工作场所不仅需要更安全的代理，还需要精心设计的人机代理协作。
cs.AI / 118 / 2608.08605
ForestBench: A Unified Graph Framework for Evaluating Multi-Agent Collaboration
ForestBench：一个统一的图框架用于评估多智能体协作
Abstract
Multi-agent systems (MAS) built on Large Language Models (LLMs) are proliferating rapidly, but their heterogeneous execution traces provide no common basis for evaluation across methods. Outcome-only benchmarks discard collaborations, whereas LLM-as-Judge evaluation requires additional, model-dependent inference and can vary with the LLM and rubric. We introduce a generalizable evaluation framework that maps native MAS traces into a shared space of unified collaboration graphs, enabling different methods to be evaluated under the same representation, reference set, and metric panel. Candidate graphs are compared with a query-specific reference forest. Each forest is a benchmark-provided collection of verified-success graphs: it records diverse ways in which representative MAS methods can complete the task, rather than prescribing a unique optimal process. Instantiating the framework as ForestBench, we filter $844$ collaboration-necessary queries from seven public datasets, precompute ten successful target-conditioned reference graphs per query, and evaluate six representative MAS frameworks. Controlled backbone, reference-construction, and perturbation studies test the stability and scope of evaluation. Once the benchmark forests are built, ForestBench scores a trace in milliseconds without further LLM inference, providing a reusable structural basis for comparing diverse MAS collaboration traces.
Chinese Translation
基于大型语言模型（LLMs）构建的多智能体系统（MAS）正在迅速增长，但它们异构的执行轨迹未能为不同方法的评估提供共同基础。仅关注结果的基准测试忽略了协作，而将LLM作为评判者的评估则需要额外的、依赖于模型的推理，并且可能因LLM和评估标准的不同而有所变化。我们提出了一个可推广的评估框架，将原生MAS轨迹映射到一个统一的协作图共享空间，使不同方法能够在相同的表示、参考集和指标面板下进行评估。候选图与特定查询的参考森林进行比较。每个森林是基准提供的经过验证的成功图的集合：它记录了代表性MAS方法完成任务的多种方式，而不是规定唯一的最佳过程。将该框架实例化为ForestBench，我们从七个公共数据集中筛选出844个需要协作的查询，为每个查询预计算十个成功的目标条件参考图，并评估六个代表性的MAS框架。控制的骨干、参考构建和扰动研究测试了评估的稳定性和范围。一旦基准森林建立，ForestBench可以在毫秒内对轨迹进行评分，而无需进一步的LLM推理，为比较不同MAS协作轨迹提供了可重用的结构基础。
cs.AI / 119 / 2608.08617
Walking through Discussions: A Mobile Visual Analytics System for In-Situ Group Discussion Analysis
走进讨论：一种用于现场小组讨论分析的移动视觉分析系统
Abstract
Group discussion-based teaching is widely used to foster collaborative learning, yet teachers in physical classrooms often struggle to simultaneously monitor multiple groups and quickly diagnose a target group before intervening. Existing visual analytics tools primarily support post-hoc analysis on desktop, providing limited support for in-situ walk-around teaching. To address this gap, we present MobileGroupVis, a mobile visual analytics system for in-situ analysis of classroom group discussions. MobileGroupVis integrates multi-group monitoring, single-group diagnosis, and instructional intervention into a concise analytical workflow tailored for small-screen touch interaction. The system is powered by a lightweight streaming analysis pipeline that converts group audio into structured discussion data and further extracts interaction patterns, topic progression, and topic deviation through a dialogue analysis module. To enable both glanceable overview and traceable diagnosis, we design six coordinated views, including a compact glyph that visually encodes word count, interaction intensity, and topic deviation for efficient cross-group comparison and anomaly localization, along with detailed views for opinion evolution, interaction dynamics, topic coverage, and dialogue records. We evaluate MobileGroupVis through two case studies and expert interviews. The results provide preliminary evidence that MobileGroupVis supports teachers in understanding discussion processes, identifying groups in need of attention, and facilitating in-class intervention.
Chinese Translation
基于小组讨论的教学广泛应用于促进协作学习，但在实体教室中，教师往往难以同时监控多个小组，并在干预前快速诊断目标小组。现有的视觉分析工具主要支持桌面上的事后分析，对现场走动式教学的支持有限。为了解决这一问题，我们提出了MobileGroupVis，这是一种用于课堂小组讨论现场分析的移动视觉分析系统。MobileGroupVis将多组监控、单组诊断和教学干预整合到一个简洁的分析工作流程中，专为小屏幕触控交互设计。该系统由一个轻量级流式分析管道驱动，将小组音频转换为结构化讨论数据，并通过对话分析模块进一步提取互动模式、主题进展和主题偏差。为了实现可视化的概览和可追溯的诊断，我们设计了六个协调视图，包括一个紧凑的图形，直观编码字数、互动强度和主题偏差，以便于高效的跨组比较和异常定位，以及关于观点演变、互动动态、主题覆盖和对话记录的详细视图。我们通过两个案例研究和专家访谈对MobileGroupVis进行了评估。结果提供了初步证据，表明MobileGroupVis支持教师理解讨论过程、识别需要关注的小组，并促进课堂干预。
cs.AI / 120 / 2608.08621
Business Arena: Benchmarking LLM Agents in a Realistic Marketplace
商业竞技场：在现实市场中基准测试大型语言模型代理
Abstract
Running a business is a challenging form of intelligent work. Operators must infer opportunities from partial signals, commit capital under uncertainty, adapt to delayed outcomes in a changing market, and satisfy regulatory obligations before trading legally. Frontier LLM agents can increasingly complete complex workflows, yet business-related capabilities are rarely evaluated in existing agent benchmarks. We introduce \textbf{Business Arena}, a controlled environment where an AI agent runs a cross-border shop, buying from suppliers and selling to buyers over a long horizon. We ground the arena in real Alibaba.com sourcing data and market conditions calibrated from authoritative sources. Delayed and coupled consequences make individual business decisions difficult to judge, but their combined outcome is measurable through profit. Because profit alone cannot explain why an agent succeeds or fails, we compare agents with human-designed strategies to estimate available opportunity, use skill-level metrics to reveal underlying strengths and weaknesses, and trace realized gains and losses to the actions that produced them. We use mechanism ablations to establish that strong results reflect genuine business intelligence rather than neglect or simulator-specific shortcuts. We evaluate 15 frontier models and find a ninefold difference in mean final net worth. Even the best model falls behind human-designed strategies, indicating that business operation remains challenging for LLM agents. Skill-level analysis reveals operating styles, from margin-focused premium sellers to high-turnover wholesalers and customer-service specialists, while action-level attribution identifies the sourcing, pricing, and recovery decisions that create or destroy value. Together, Business Arena takes a first step toward a realistic and trustworthy testbed for evaluating end-to-end business agents.
Chinese Translation
经营一家企业是一种具有挑战性的智能工作形式。运营者必须从部分信号中推断机会，在不确定性下投入资本，适应变化市场中延迟的结果，并在合法交易前满足监管义务。前沿的大型语言模型（LLM）代理越来越能够完成复杂的工作流程，但在现有的代理基准测试中，商业相关能力很少被评估。我们引入了 extbf{商业竞技场}，这是一个受控环境，在这里，人工智能代理经营一家跨境商店，从供应商处采购并在较长时间内向买家销售。我们基于真实的Alibaba.com采购数据和来自权威来源的市场条件来构建这一竞技场。延迟和耦合的后果使得单个商业决策难以判断，但其综合结果可以通过利润来衡量。由于利润本身无法解释代理成功或失败的原因，我们将代理与人类设计的策略进行比较，以估计可用机会，使用技能水平指标揭示潜在的优势和劣势，并追踪实现的收益和损失与产生它们的行动之间的关系。我们使用机制消融来确定强劲的结果反映了真正的商业智能，而非忽视或特定于模拟器的捷径。我们评估了15个前沿模型，发现平均最终净值存在九倍的差异。即使是表现最好的模型也落后于人类设计的策略，这表明商业运营对大型语言模型代理仍然具有挑战性。技能水平分析揭示了不同的运营风格，从关注利润的高端卖家到高周转的批发商和客户服务专家，而行动级归因则识别出创造或破坏价值的采购、定价和回收决策。总的来说，商业竞技场为评估端到端商业代理迈出了现实且可信的测试平台的第一步。
cs.AI / 121 / 2608.08623
MedCalc-R1: Knowledge-Guided Reward Framework for Medical Mathematical Reasoning
MedCalc-R1：用于医学数学推理的知识引导奖励框架
Abstract
In Reinforcement Learning with Verifiable Rewards (RLVR) frameworks for mathematical reasoning tasks, floating-point results are typically evaluated using a tolerance-based reward. However, this strategy suffers from challenges such as difficulty in threshold calibration, unstable training dynamics, and limited accuracy, especially in clinical scenarios. To address these limitations, we propose a knowledge-guided hybrid reward framework (\textsc{MedCalc-R1}). Specifically, we introduce a knowledge verification reward mechanism that enforces explicit generation of computational formulas, which are further validated by an external verifier to enhance interpretability and reasoning reliability. Furthermore, we design a hybrid soft-hard reward scheme combining a hard constraint based on clinical safety thresholds with a soft, precision-sensitive reward that progressively guides learning within the acceptable range. Experimental results demonstrate that our method significantly outperforms existing baselines in both reasoning accuracy and generalization capability, validating the effectiveness and applicability in safety-critical domains.
Chinese Translation
在用于数学推理任务的可验证奖励强化学习框架（RLVR）中，浮点结果通常使用基于容忍度的奖励进行评估。然而，这一策略面临诸多挑战，如阈值校准困难、训练动态不稳定以及准确性有限，尤其是在临床场景中。为了解决这些局限性，我们提出了一种知识引导的混合奖励框架（MedCalc-R1）。具体而言，我们引入了一种知识验证奖励机制，强制生成计算公式，并通过外部验证者进一步验证，以增强可解释性和推理可靠性。此外，我们设计了一种混合软硬奖励方案，将基于临床安全阈值的硬约束与逐步引导学习的软、精度敏感奖励相结合，以确保学习在可接受范围内进行。实验结果表明，我们的方法在推理准确性和泛化能力上显著优于现有基线，验证了其在安全关键领域的有效性和适用性。
cs.AI / 122 / 2608.08627
UniMoMo: Expert Merging-Based MoE Acceleration for Large Recommendation Models
UniMoMo：基于专家合并的 MoE 加速大规模推荐模型
Abstract
Sparse mixture-of-experts (MoE) layers expand recommendation capacity through conditional computation, yet a trained checkpoint still stores and routes over its full expert bank. We study a deployment problem: convert that checkpoint to a smaller standard MoE under an explicit expert budget, without adding a compression-specific online module. To address this, we introduce UniMoMo, a post-training compression framework formulated as a constrained graph coarsening problem. Rather than relying on parameter distance, UniMoMo groups experts based on their functional similarity, using an unlabeled calibration set to measure how similarly experts respond to shared recommendation states. To prevent performance degradation, we introduce a layer-adaptive protection mechanism that restricts the merging of high-traffic experts based on their routing exposure. Across Amazon Beauty, KuaiRec, and TenRec with 2, 4, and 6 MoE blocks, the final four-expert checkpoints obtain source-relative five-run mean NDCG@10 ratios of 99.92%--102.30% and measured A100 speedups of 1.28$\times$--1.63$\times$. An aggressive two-expert, top-1 operating point obtains ratios of 98.36%--104.24% and speedups of 1.47$\times$--2.21$\times$. These endpoint results evaluate the complete conversion-and-adaptation workflow and show that a trained recommendation MoE can be exported at multiple serving budgets.
Chinese Translation
稀疏专家混合（MoE）层通过条件计算扩展了推荐能力，但训练后的检查点仍然存储并路由其完整的专家库。我们研究了一个部署问题：在不添加特定于压缩的在线模块的情况下，将该检查点转换为一个较小的标准 MoE，且在明确的专家预算内。为此，我们引入了 UniMoMo，这是一种后训练压缩框架，形式化为一个约束图粗化问题。UniMoMo 不依赖于参数距离，而是基于专家的功能相似性对其进行分组，使用未标记的校准集来测量专家对共享推荐状态的响应相似度。为了防止性能下降，我们引入了一种层自适应保护机制，限制高流量专家的合并，基于其路由曝光。在 Amazon Beauty、KuaiRec 和 TenRec 上，使用 2、4 和 6 个 MoE 块，最终的四专家检查点获得了相对于源的五次运行平均 NDCG@10 比率为 99.92%--102.30%，并测得 A100 加速比为 1.28×--1.63×。一个激进的两专家、前 1 的操作点获得了 98.36%--104.24% 的比率和 1.47×--2.21× 的加速比。这些端点结果评估了完整的转换和适应工作流程，并表明训练后的推荐 MoE 可以在多个服务预算下导出。
cs.AI / 123 / 2608.08632
A QUBO-Inspired Computational Framework for Airport Landside Bottleneck Diagnosis and Dynamic Dispatch Optimization
一种基于 QUBO 的机场陆侧瓶颈诊断与动态调度优化计算框架
Abstract
Airport landside traffic centers connect terminal arrivals with taxis, ride-hailing vehicles, private cars, buses, metro services, parking facilities, and terminal-area roadways. Peak arrivals can create coupled congestion across passenger queues, vehicle queues, pickup berths, storage areas, and access roads. This study proposes a QUBO-inspired computational framework for bottleneck diagnosis and dynamic dispatch in this setting. Shanghai Pudong International Airport and Hangzhou Xiaoshan International Airport serve as case airports. A five-minute state model links passenger arrivals, vehicle supply, pickup berth service, vehicle storage, and road capacity. Bottleneck diagnosis uses service intensity, road demand saturation, bottleneck frequency, queue severity, shadow-price leverage, and a composite congestion severity index. Two dispatch schemes are tested under consistent demand inputs: finite-action model predictive control and quadratic-unconstrained-binary-optimization-inspired simulated annealing. In the strong-peak baseline scenario, the QUBO-inspired method reduces the final passenger queue from 3445 to 2477 passengers at Shanghai Pudong and from 2053 to 1482 passengers at Hangzhou Xiaoshan. Case results indicate different dominant bottlenecks. Shanghai Pudong is more affected by road saturation, whereas Hangzhou Xiaoshan is more affected by pickup berth service. Robustness tests under demand, supply, service, road-capacity, modal-share, and random-noise perturbations show retained queue-reduction benefits under the tested uncertainty levels.
Chinese Translation
机场陆侧交通中心将航站楼到达与出租车、网约车、私家车、公交车、地铁服务、停车设施以及航站楼区域道路连接起来。高峰到达可能导致乘客排队、车辆排队、接客泊位、存储区和接入道路之间的耦合拥堵。本研究提出了一种基于 QUBO 的计算框架，用于在这一环境中进行瓶颈诊断和动态调度。上海浦东国际机场和杭州萧山国际机场作为案例机场。五分钟状态模型将乘客到达、车辆供应、接客泊位服务、车辆存储和道路容量连接起来。瓶颈诊断使用服务强度、道路需求饱和度、瓶颈频率、排队严重程度、影子价格杠杆和复合拥堵严重性指数。针对一致的需求输入测试了两种调度方案：有限动作模型预测控制和基于二次无约束二进制优化的模拟退火。在强峰基线场景下，基于 QUBO 的方法将上海浦东的最终乘客排队人数从 3445 降至 2477，将杭州萧山的最终乘客排队人数从 2053 降至 1482。案例结果表明不同的主导瓶颈。上海浦东更受道路饱和的影响，而杭州萧山则更受接客泊位服务的影响。在需求、供应、服务、道路容量、模式分配和随机噪声扰动下的稳健性测试显示，在测试的不确定性水平下，排队减少的好处得以保持。
cs.AI / 124 / 2608.08634
Can Open-Weight Models Compete on Financial Text Comprehension?
开放权重模型能在金融文本理解上竞争吗？
Abstract
Open-weight language models from Chinese AI labs caught up on benchmarks relative to proprietary frontier models in recent months. Yet their reliability on real-world financial tasks remains largely untested. We updated the Financial Touchstone benchmark, which now has 2,967 question context-answer triplets across 495 international annual reports. We also apply a new set of models on the benchmark, expanding coverage from eleven to twenty models across ten providers, including recent open-weight models such as GLM 4.7, GLM 5, Kimi K2.6, and DeepSeek V3.2, as well as Alibaba's proprietary flagship Qwen3-Max. Anthropic's Claude Opus 4.6 achieves the highest accuracy (88.4%), while Google's Gemini 2.5 Pro maintains the lowest hallucination rate (0.08%). Notably, the open-weight Kimi K2.6 ranks third in accuracy, and the non-reasoning models GLM 5 and Mistral 3 rank fourth and fifth, challenging the assumption that reasoning architectures or proprietary weights are a prerequisite for strong financial comprehension. Information retrieval remains the primary bottleneck, accounting for 48.9% of all failures. We also document a new finding: geopolitical content filters in Chinese models refuse legitimate financial questions (0.08% of attempts), sometimes without clear reason, and the refusal behavior depends on the access route as much as on the model. The complete dataset and evaluation framework are publicly available.
Chinese Translation
来自中国人工智能实验室的开放权重语言模型在最近几个月的基准测试中逐渐追赶上了专有前沿模型。然而，它们在现实世界金融任务中的可靠性仍然基本未经过检验。我们更新了金融基准测试（Financial Touchstone），现在包含了来自495份国际年报的2,967个问题上下文-答案三元组。我们还在该基准测试上应用了一组新的模型，将覆盖范围从十一个扩展到二十个模型，涵盖十个提供商，包括最近的开放权重模型如GLM 4.7、GLM 5、Kimi K2.6和DeepSeek V3.2，以及阿里巴巴的专有旗舰模型Qwen3-Max。Anthropic的Claude Opus 4.6达到了最高的准确率（88.4%），而谷歌的Gemini 2.5 Pro保持了最低的幻觉率（0.08%）。值得注意的是，开放权重模型Kimi K2.6在准确率上排名第三，而非推理模型GLM 5和Mistral 3分别排名第四和第五，这挑战了推理架构或专有权重是强大金融理解的先决条件的假设。信息检索仍然是主要瓶颈，占所有失败的48.9%。我们还记录了一个新发现：中文模型中的地缘政治内容过滤器拒绝合法的金融问题（0.08%的尝试），有时没有明确的理由，而拒绝行为在很大程度上取决于访问途径和模型本身。完整的数据集和评估框架已公开提供。
cs.AI / 125 / 2608.08639
Smart Compaction: Predicting Compaction Utility from Lakehouse Table Metadata
智能压缩：从湖屋表元数据预测压缩效用
Abstract
Open lakehouse table formats accumulate small data files over time, which degrades query performance. Deciding when compaction is worthwhile remains threshold-driven, but which metadata features actually determine compaction utility is not well understood. We present an open simulation framework that generates 2,376 Apache Iceberg tables spanning three orders of magnitude in file size, extracts 17 metadata features from manifest files without reading data, and trains XGBoost to predict the continuous file-reduction ratio (R2 = 0.998, RMSE= 0.013). The binary compaction decision turns out to be trivially separable by a single partition-level threshold max_files_per_partition> 4, requiring no learned model. Cross-schema validation on 96 TPC-H tables confirms generalisation without retraining (R2 = 0.976). A query benchmark reveals that compaction benefits metadata-heavy queries but can slow full-scan aggregations by reducing task parallelism. All code and data are publicly available.
Chinese Translation
开放湖屋表格式随着时间的推移积累了小数据文件，这会降低查询性能。决定何时进行压缩仍然是基于阈值的，但实际上哪些元数据特征决定了压缩效用尚不清楚。我们提出了一个开放的仿真框架，生成了2376个Apache Iceberg表，涵盖了三个数量级的文件大小，从清单文件中提取了17个元数据特征而无需读取数据，并训练XGBoost以预测连续文件减少比率（R2 = 0.998，RMSE = 0.013）。二元压缩决策实际上可以通过单个分区级阈值max_files_per_partition > 4轻松分离，无需学习模型。对96个TPC-H表的跨模式验证确认了在不重新训练的情况下的泛化能力（R2 = 0.976）。查询基准测试显示，压缩对元数据密集型查询有益，但通过减少任务并行性可能会减慢全扫描聚合的速度。所有代码和数据均可公开获取。
cs.AI / 126 / 2608.08640
SkillReason: Reasoning-Enhanced Agent Skill Retrieval for Implicit User Requests
SkillReason：增强推理的代理技能检索以应对隐式用户请求
Abstract
Large language model agents increasingly rely on reusable skills to extend their capabilities beyond parametric knowl- edge. However, retrieving the appropriate skill from a large- scale library remains challenging because realistic user re- quests are often concise and underspecified, stating only the task goal while leaving the required capabilities and execu- tion steps implicit. Existing benchmarks provide limited cov- erage of such requests. To address this gap, we introduce SkillReason-Bench, a large-scale cross-domain benchmark containing 3,729 queries and a retrieval corpus of 61,228 skills spanning nine domains. We further propose SkillRea- son, a two-stage framework that uses chain-of-thought rea- soning as training-time supervision for skill retrieval. In Stage I, capability reasoning traces generated by a stronger teacher provide explicit supervision through contrastive learning, re- trieval distribution alignment, and language modeling, en- couraging the retriever to internalize capability reasoning in its query representation. In Stage II, a retrieval-guided GRPO objective encourages the model to explore reasoning trajecto- ries better suited to its own capabilities and more effective for retrieval. At inference, SkillReason directly encodes the orig- inal query without autoregressive CoT generation, preserv- ing efficient query-only retrieval. Extensive experiments on SkillReason-Bench, SkillRet, and SRA-Bench show that Skill- Reason achieves state-of-the-art performance across all three benchmarks, demonstrating that reasoning-enhanced training better bridges the semantic gap between high-level task goals and skill capabilities.
Chinese Translation
大型语言模型代理越来越依赖可重用技能，以扩展其超越参数知识的能力。然而，从大规模库中检索适当的技能仍然具有挑战性，因为现实用户请求通常简洁且不明确，仅陈述任务目标，而将所需能力和执行步骤隐含。现有基准对这类请求的覆盖有限。为了解决这一问题，我们引入了SkillReason-Bench，这是一个包含3,729个查询和61,228个技能的跨领域大规模基准，涵盖九个领域。我们进一步提出了SkillReason，一个两阶段框架，利用思维链推理作为技能检索的训练时监督。在第一阶段，由更强教师生成的能力推理轨迹通过对比学习、检索分布对齐和语言建模提供明确的监督，鼓励检索器在其查询表示中内化能力推理。在第二阶段，检索引导的GRPO目标鼓励模型探索更适合其自身能力和更有效的检索推理轨迹。在推理时，SkillReason直接编码原始查询，而不进行自回归的思维链生成，从而保持高效的仅查询检索。在SkillReason-Bench、SkillRet和SRA-Bench上的广泛实验表明，SkillReason在所有三个基准上都达到了最先进的性能，证明增强推理的训练更好地弥合了高层任务目标与技能能力之间的语义差距。
cs.AI / 127 / 2608.08654
The Scaffolding Matters More Than the Interface: A Controlled Comparison of MCP and CLI Tool Use Across Seven Agent Scaffoldings, Five Language Models, and One Software Task
支架比接口更重要：在七种代理支架、五种语言模型和一个软件任务中对MCP和CLI工具使用的控制比较
Abstract
How much an AI coding agent costs to run can depend more on the agent scaffolding that drives it than on the interface through which it reaches its tools. We set out to measure the cost of tool use over the Model Context Protocol (MCP) against tool use over an ordinary command-line interface (CLI), a difference on which published estimates disagree by more than an order of magnitude while resting on practitioner reports that cannot be reproduced. We ran one fixed software task -- six operations against a private online git repository -- across seven agent scaffoldings and five language models, and we verified completion by inspecting the repository state rather than trusting the agent's self-report. The dominant effect was the scaffolding. Two of the seven ship no MCP support at all; they completed every run using only the CLI, which shows that MCP is unnecessary for this class of work, and they were 5.0x to 28x cheaper than the five scaffoldings that do support MCP, comparing CLI runs alone with no MCP server attached anywhere. The effect was largest for a small 27-billion-parameter model running locally, whose cost varied 139x across scaffoldings while it completed the task under all of them. The comparison we set out to make proved unstable: thirteen strictly paired MCP-to-CLI ratios span 0.43x to 29x, with outliers on both sides. The two interfaces separate on the cost of failure, where 12.9 per cent of the money spent on MCP runs bought no completed work against 2.2 per cent on CLI runs, but not on its frequency: failures were equally common in both, in the original runs and in their repetitions alike. Agents frequently ignored the interface they were assigned, so comparisons that do not verify actual behaviour measure an unknown mixture. The harness, the task, the verification and the complete dataset are released as open source.
Chinese Translation
AI编码代理的运行成本更多地依赖于驱动其工作的代理支架，而非其访问工具的接口。我们旨在测量通过模型上下文协议（MCP）使用工具的成本与通过普通命令行接口（CLI）使用工具的成本之间的差异，已发表的估计对此存在超过一个数量级的分歧，且基于无法复现的从业者报告。我们在七种代理支架和五种语言模型中，对一个固定的软件任务——针对一个私有在线git代码库的六个操作进行了测试，并通过检查代码库状态来验证完成情况，而不是依赖代理的自我报告。支架的影响是主导的。在七种支架中，有两种完全不支持MCP；它们在每次运行中仅使用CLI完成，这表明MCP对于这一类工作是多余的，并且它们的成本比支持MCP的五种支架低5.0倍到28倍，仅比较没有附加MCP服务器的CLI运行。对于一个本地运行的小型27亿参数模型，成本在不同支架间变化达到139倍，而在所有支架下均完成了任务。我们设定的比较结果不稳定：严格配对的MCP与CLI比率跨度为0.43倍到29倍，且两侧均有异常值。两个接口在失败成本上有所不同，其中12.9%的MCP运行花费未能完成任何工作，而CLI运行的这一比例为2.2%，但在频率上并无差异：两者的失败发生率在原始运行及其重复中均相同。代理经常忽视被分配的接口，因此不验证实际行为的比较测量的是一种未知的混合。支架、任务、验证和完整数据集已作为开源发布。
cs.AI / 128 / 2608.08677
Branch2Skill: Efficient Skill Evolution Through Reasoning Trees
Branch2Skill：通过推理树实现高效技能演化
Abstract
Skill evolution improves agent skills through feedback over time, with failed trajectories often providing informative signals by revealing incomplete or misleading behaviors. However, existing methods mainly rely on single trajectories, where early reasoning errors can propagate through subsequent steps and weaken the feedback available for skill refinement. Consequently, improving skills requires repeated cycles of rollout, diagnosis, and update, incurring substantial token costs. To address this challenge, we introduce Branch2Skill, an efficient framework that transforms a single reasoning tree into dense supervision for skill evolution. For each task or problem, Branch2Skill performs Monte Carlo tree search under a fixed budget to obtain diverse reasoning trajectories, then compares an elite path with sibling alternatives sharing the same prefixes to extract step-wise evidence about which reasoning patterns to retain, revise, or avoid. Finally, Branch2Skill distills multi-step evidence into reusable updates, allowing one reasoning tree to provide supervision across multiple reasoning steps and reducing the need for repeated rollout-update cycles. Across six benchmarks covering reasoning and agentic tasks, Branch2Skill consistently improves task performance while enhancing skill evolution efficiency. For example, with GPT 5.5 as the target model, Branch2Skill uses 73.2% fewer tokens than SkillOpt, while achieving superior performance. These results demonstrate that reasoning trees can support not only more effective trajectory search, but also richer supervision for more efficient skill improvement. Code will be published.
Chinese Translation
技能演化通过时间反馈提升代理技能，失败的轨迹常常提供有价值的信号，揭示不完整或误导性的行为。然而，现有方法主要依赖单一轨迹，早期推理错误可能在后续步骤中传播，从而削弱了可用于技能精炼的反馈。因此，提升技能需要重复的展开、诊断和更新周期，这会产生可观的代币成本。为了解决这一挑战，我们提出了Branch2Skill，一个高效的框架，将单一推理树转化为技能演化的密集监督。对于每个任务或问题，Branch2Skill在固定预算下执行蒙特卡洛树搜索，以获取多样化的推理轨迹，然后将精英路径与共享相同前缀的兄弟替代路径进行比较，以提取关于保留、修订或避免哪些推理模式的逐步证据。最后，Branch2Skill将多步证据提炼为可重用的更新，使得一个推理树能够在多个推理步骤中提供监督，从而减少重复展开-更新周期的需求。在涵盖推理和代理任务的六个基准测试中，Branch2Skill始终提高任务性能，同时增强技能演化效率。例如，以GPT 5.5作为目标模型，Branch2Skill使用的代币比SkillOpt少73.2%，同时实现了更优的性能。这些结果表明，推理树不仅可以支持更有效的轨迹搜索，还可以为更高效的技能提升提供更丰富的监督。代码将会发布。
cs.AI / 129 / 2608.08689
A Structural Dynamics Graph World Model: Unified Modeling, Constrained Rollout, and Interpretable Calibration
结构动力学图世界模型：统一建模、受限展开与可解释校准
Abstract
The state evolution of a complex system arises jointly from object laws, relational propagation, domain conservation, and unmodeled error. Forcing all sources into one black box makes mechanism attribution and constraint preservation unauditable; forcing every mechanism into one equation family discards mature domain solvers. We propose SD-GWM, a Structural Dynamics Graph World Model as an executable structural contract: nodes declare self-dynamics S, edges declare neighbor graph-coupled dynamics N---both fixed-form mechanism assets (rules, ODEs, solvers) calibrating only authorized parameters. An optional bounded residual R concentrates learnability, while a global projection maps states to feasibility, enforcing constraints without guaranteeing accuracy gains. On eight pre-registered research questions, SD-GWM delivers (i) heterogeneous integration: rules and solvers plug in natively; (ii) semantic fidelity: disabling R preserves source semantics bit-for-bit, with four theory properties under explicit proof/empirical boundaries; (iii) auditable governance: stepwise traces enable counterfactual fault localization (top-1 = 1.0) without post-hoc approximations. On a semi-synthetic flood testbed and USGS streamflow, SD-GWM reduces constraint violations to floating-point tolerance in analytical tests and to zero in semi-synthetic and real-data cases. Persistence matches SD-GWM in calm periods, but during a 254-day extreme-flood shift persistence and all neural baselines collapse (90-min RMSE 892-3007 cfs) while SD-GWM holds at 108 cfs (8-28x gain). The bounded residual cuts RMSE ~50% only under backbone bias. We position SD-GWM not as a universally superior forecaster, but as a verifiable substrate for auditable, constraint-safe spatiotemporal mining.
Chinese Translation
复杂系统的状态演变共同源于对象法则、关系传播、领域守恒和未建模的误差。将所有来源强行纳入一个黑箱使得机制归因和约束保持无法审计；将每个机制强行纳入一个方程族则舍弃了成熟的领域求解器。我们提出了SD-GWM，即结构动力学图世界模型，作为一个可执行的结构契约：节点声明自我动力学S，边缘声明邻居图耦合动力学N——这两者都是固定形式的机制资产（规则、常微分方程、求解器），仅校准授权参数。一个可选的有界残差R集中学习能力，而全局投影将状态映射到可行性，强制约束而不保证准确性提升。在八个预注册的研究问题上，SD-GWM提供了（i）异构集成：规则和求解器本地插件；（ii）语义保真度：禁用R可以逐位保留源语义，并在明确的证明/经验边界下具备四个理论属性；（iii）可审计治理：逐步追踪使得反事实故障定位成为可能（top-1 = 1.0），无需事后近似。在一个半合成洪水测试平台和美国地质调查局（USGS）流量数据上，SD-GWM将约束违反减少到浮点容忍度，在分析测试中为零，在半合成和真实数据案例中为零。持久性在平静时期与SD-GWM匹配，但在254天的极端洪水转变期间，持久性和所有神经基线崩溃（90分钟均方根误差892-3007 cfs），而SD-GWM保持在108 cfs（增益为8-28倍）。有界残差仅在主干偏差下将均方根误差减少约50%。我们将SD-GWM定位为一个可验证的基底，用于可审计、约束安全的时空挖掘，而非普遍优越的预测器。
cs.AI / 130 / 2608.08691
EnergyBridge: Benchmarking Household Energy Management, User Participation, and Grid Flexibility
EnergyBridge：家庭能源管理、用户参与和电网灵活性的基准测试
Abstract
Residential virtual power plants (VPPs) can provide grid flexibility by shifting household demand, but physical flexibility becomes dependable capacity only when residents authorize a plan and the promised response is delivered. Existing benchmarks evaluate control but omit event-specific authorization. We present EnergyBridge, a benchmark and agent framework connecting capacity reporting, household authorization, and physical execution. It combines region-specific EnergyPlus environments for Tianjin and Berlin with an LLM-based User Participation Simulator. Against 584 persona- and event-matched human role-play judgments, the LLM-based User Participation Simulator preserves method ordering with a 5.3-point mean absolute acceptance error. Across conventional controllers and agent baselines, EnergyBridge achieves the highest simulated authorization, lowest event-window energy, and the most reliable capacity commitment in both regions. We release human data and codes for reproducible human-centered grid-flexibility research: https://github.com/Agentic-Intelligence-Lab/EnergyBridge.
Chinese Translation
住宅虚拟电厂（VPPs）通过调整家庭需求提供电网灵活性，但物理灵活性只有在居民授权计划并兑现承诺响应时才能成为可靠的容量。现有基准评估控制，但忽略了事件特定的授权。我们提出了EnergyBridge，一个连接容量报告、家庭授权和物理执行的基准和代理框架。它结合了针对天津和柏林的区域特定EnergyPlus环境与基于大语言模型（LLM）的用户参与模拟器。在584个角色和事件匹配的人类角色扮演判断中，基于LLM的用户参与模拟器保持了方法顺序，平均绝对接受误差为5.3分。在传统控制器和代理基线中，EnergyBridge在两个地区实现了最高的模拟授权、最低的事件窗口能耗和最可靠的容量承诺。我们发布了人类数据和代码，以便进行可重复的人本电网灵活性研究： https://github.com/Agentic-Intelligence-Lab/EnergyBridge。
cs.AI / 131 / 2608.08700
PluginEval: A Diagnostic Benchmark for Fine-Grained Error Attribution in Function Calling
PluginEval：用于函数调用中细粒度错误归因的诊断基准
Abstract
Reliable evaluation of tool routing is critical as Large Language Models increasingly operate as autonomous agents. Current benchmarks face three structural limitations: data distributions that follow a power law leave rare scenarios underrepresented; the absence of adversarial hard negatives obscures performance differences across models; and annotation pipelines depend on LLM judgments that have not been validated through execution. In this paper, we introduce PluginEval, a benchmark constructed through a two-stage framework that systematically mitigates these limitations. First, we formulate tool routing as a sequence of three decisions and separate generation from verification. LLMs propose candidate calls, while deterministic validation and real API execution provide reliable quality signals. Second, we decompose each plugin by capability, intent, and boundary to identify trigger and exclusion scenarios. We then generate queries at different difficulty levels to fill coverage gaps, including adversarial negatives targeting three failure modes, and return them to the first stage for annotation. This process creates a closed loop that iterates until coverage converges. For evaluation, we move beyond aggregate accuracy. An LLM judge anchored to gold annotations classifies failures as missed calls, spurious calls, or parameter errors, producing a detailed error profile for each model. We evaluate five model families, including proprietary models and models with open weights, analyze their performance across difficulty levels and error categories, and validate the judge through agreement with human annotations.
Chinese Translation
随着大型语言模型越来越多地作为自主代理运行，工具路由的可靠评估变得至关重要。目前的基准面临三项结构性限制：遵循幂律的数据分布使得稀有场景被低估；缺乏对抗性难负样本使得模型之间的性能差异不明显；而注释流程依赖于尚未通过执行验证的LLM判断。本文介绍了PluginEval，这是一个通过两阶段框架构建的基准，系统性地缓解了这些限制。首先，我们将工具路由形式化为三个决策的序列，并将生成与验证分开。LLMs提出候选调用，而确定性验证和真实API执行提供可靠的质量信号。其次，我们按能力、意图和边界对每个插件进行分解，以识别触发和排除场景。然后，我们生成不同难度级别的查询以填补覆盖空白，包括针对三种失败模式的对抗性负样本，并将其返回第一阶段进行注释。这个过程创建了一个闭环，迭代直到覆盖收敛。为了评估，我们超越了聚合准确性。一个基于黄金注释的LLM评审将失败分类为漏调用、虚假调用或参数错误，为每个模型生成详细的错误概况。我们评估了五个模型家族，包括专有模型和开放权重模型，分析它们在不同难度级别和错误类别上的表现，并通过与人工注释的一致性验证评审的准确性。
cs.AI / 132 / 2608.08709
AI Evaluation Should Measure Verification Cost, Not Correctness Alone
人工智能评估应衡量验证成本，而不仅仅是正确性
Abstract
The reliability of AI generative models is typically measured by output correctness, yet in practice it depends on the effort required to verify those outputs. We argue that current evaluation metrics overlook a critical failure mode: Verification-Cost Errors (VCEs), defined as incorrect input-output pairs that a declared fraction of the verifier population fails to identify within the verification budget available in a given deployment context. Unlike standard notions of "hallucination", VCEs are defined operationally, by the failure of correct identification within budget rather than by any property of the output itself. Plausibility and authoritative presentation are hypothesised contributors to that failure, not defining conditions. To capture this asymmetry, we introduce the notion of verification cost relative to a deployment budget as an operational dimension that current evaluation does not routinely capture. The quantity is presented as a conceptual instrument rather than a finalized metric. Evidence from code generation and multi-modal document understanding shows that high benchmark accuracy can mask significant verification effort in practice. We therefore take the position that correctness alone is insufficient as a measure of reliability. AI evaluation should explicitly account for verification cost, reflecting whether errors can be detected under realistic resource constraints.
Chinese Translation
人工智能生成模型的可靠性通常通过输出的正确性来衡量，但实际上它依赖于验证这些输出所需的努力。我们认为，当前的评估指标忽视了一种关键的失败模式：验证成本错误（Verification-Cost Errors, VCEs），其定义为在给定的部署环境中，声明的验证者群体中有一定比例未能在可用的验证预算内识别出的错误输入-输出对。与标准的“幻觉”概念不同，VCEs 是通过在预算内未能正确识别来操作性定义的，而不是通过输出本身的任何属性。可信性和权威性呈现被假设为导致这种失败的因素，而不是定义条件。为了捕捉这种不对称性，我们引入了相对于部署预算的验证成本的概念，作为当前评估未能常规捕捉的操作性维度。该量被呈现为一个概念工具，而不是一个最终的指标。来自代码生成和多模态文档理解的证据表明，高基准准确性在实践中可能掩盖了显著的验证努力。因此，我们认为，仅仅依靠正确性作为可靠性的衡量标准是不够的。人工智能评估应明确考虑验证成本，反映在现实资源约束下是否能够检测到错误。
cs.AI / 133 / 2608.08736
FitAQA: A Benchmark of Fitness Action Quality Assessment for Multimodal Large Language Models
FitAQA：多模态大型语言模型的健身动作质量评估基准
Abstract
Fitness Action Quality Assessment (AQA) is important for intelligent sports training, yet the capabilities of Multimodal Large Language Models (MLLMs) in this setting remain underexplored. Existing benchmarks rely on action-specific annotation schemes and focus primarily on final assessment outputs, offering limited insight into how models assess exercise quality. We introduce FitAQA, a systematic benchmark for evaluating MLLMs in fitness AQA, containing 2,219 videos and 5,512 QA instances across 30 bodyweight exercises. In collaboration with experts in sports science, we develop a unified form error taxonomy that defines 38 recurring form errors within six complementary quality dimensions: alignment, symmetry, stability, coordination, tempo, and completeness. This taxonomy provides a shared assessment framework across different exercises. FitAQA further formulates three evaluation tasks: perception for recognizing relevant visual evidence, judgement for combining that evidence with domain knowledge to assess execution correctness, and temporal grounding for localizing form errors over time. Extensive evaluation shows that current MLLMs still struggle to assess exercise quality comprehensively and localize form errors precisely. Controlled experiments further indicate that visual perception is a key bottleneck, as judgement performance improves substantially when ground-truth perceptual evidence is provided. The dataset and evaluation code will be made publicly available.
Chinese Translation
健身动作质量评估（AQA）对于智能体育训练至关重要，但多模态大型语言模型（MLLMs）在这一领域的能力仍未得到充分探索。现有基准依赖于特定动作的注释方案，主要关注最终评估输出，提供的关于模型如何评估运动质量的见解有限。我们推出了FitAQA，这是一个系统的基准，用于评估MLLMs在健身AQA中的表现，包含2,219个视频和5,512个QA实例，涵盖30种自身体重练习。我们与体育科学专家合作，开发了一个统一的形式错误分类法，定义了在六个互补质量维度内的38种常见形式错误：对齐、对称、稳定性、协调性、节奏和完整性。该分类法为不同练习提供了共享的评估框架。FitAQA进一步制定了三个评估任务：感知（识别相关视觉证据）、判断（将该证据与领域知识结合以评估执行正确性）和时间定位（随时间定位形式错误）。广泛的评估表明，当前的MLLMs在全面评估运动质量和精确定位形式错误方面仍然存在困难。控制实验进一步表明，视觉感知是一个关键瓶颈，因为当提供真实的感知证据时，判断性能显著提高。数据集和评估代码将公开发布。
cs.AI / 134 / 2608.08746
Scale-to-Dialogue: Low-Burden Elicitation of Daily Premenstrual Symptom Ratings with Small Language Models
规模对话：利用小型语言模型低负担引导每日经前综合症评分
Abstract
Prospective daily symptom tracking is central to premenstrual health assessment, but repeated ordinal forms impose substantial response burden. We formulate conversational administration as an ordinal label-recovery problem: the system actively elicits a small set of symptom clusters and maps each response to the original severity labels. We used 3,320 complete participant-days from the mcPHASES dataset, covering cramps, mood swing, fatigue, sleep issues, stress, and bloating on a six-level scale. Six participants were reserved for development and 36 for a frozen evaluation comprising 360 participant-days and 2,160 item labels. A ModernBERT evidence gate detected whether a symptom was expressed, and Qwen2.5-1.5B-Instruct produced deterministic structured severity scores. Fixed six-item questioning achieved a quadratic weighted kappa of 0.976, whereas three joint symptom-cluster questions achieved 0.913, 97.45% agreement within one severity level, and 80.94% recall for moderate-or-higher symptoms while reducing questions by 50%. Open-first adaptive policies required 3.92-5.98 questions and produced lower agreement than the corresponding fixed policies. Participant-cluster bootstrap analysis estimated a kappa difference of -0.062 (95% CI -0.076 to -0.048) between the three-cluster and six-item strategies. Active cluster-level elicitation provides a direct, local-model route from natural conversation to reusable daily symptom labels.
Chinese Translation
前瞻性的每日症状追踪是经前健康评估的核心，但重复的序数形式会造成显著的响应负担。我们将对话式管理形式化为一个序数标签恢复问题：系统主动引导一小组症状聚类，并将每个响应映射到原始的严重程度标签。我们使用了来自mcPHASES数据集的3,320个完整参与者天数，涵盖了腹痛、情绪波动、疲劳、睡眠问题、压力和腹胀，使用六级评分。六名参与者用于开发，36名用于冻结评估，评估包括360个参与者天数和2,160个项目标签。现代BERT证据门检测症状是否被表达，而Qwen2.5-1.5B-Instruct则生成确定性的结构化严重程度评分。固定的六项问题实现了0.976的二次加权kappa，而三个联合症状聚类问题实现了0.913，在一个严重程度级别内达成97.45%的一致性，并且在减少问题数量50%的同时，对于中度或更严重症状的召回率为80.94%。开放优先的自适应策略需要3.92-5.98个问题，且产生的协议低于相应的固定策略。参与者聚类自助分析估计三聚类和六项策略之间的kappa差异为-0.062（95% CI -0.076至-0.048）。主动的聚类级别引导提供了一条从自然对话到可重复使用的每日症状标签的直接、本地模型路径。
cs.AI / 135 / 2608.08786
SymDiag: Explainable Diagnosis for LLM Reasoning via Neuro-Symbolic Verification
SymDiag：通过神经符号验证实现可解释的LLM推理诊断
Abstract
Large language models (LLMs) increasingly serve as data-driven reasoners, yet their chains-of-thought (CoT) can be unfaithful even when final answers are correct. Most existing ``verification'' signals are not diagnostic: answer matching observes only the outcome, LLM-as-judge provides subjective and non-verifiable critiques, and scalar rewards (e.g., PRMs/RMs) offer little insight into where a multi-step derivation fails.We propose \textbf{SymDiag}, a neuro-symbolic framework that \textbf{reframes reasoning verification as structured failure diagnosis}. SymDiag translates natural-language CoT into symbolic constraints and performs step-level satisfiability/entailment checks to (i) localize failing steps and (ii) produce verifiable diagnostic evidence, including counterexamples, inconsistency witnesses, and missing-premise indicators. A central challenge is that apparent ``logic violations'' can be caused either by genuine reasoning defects or by neural-to-symbolic translation noise. SymDiag therefore incorporates a Self-Auditor that disentangles TranslationError from ReasoningError via dual symbolic encodings consistency checks, enabling robust diagnosis under partial observability. Across diverse mathematical, logical, scientific, and general reasoning benchmarks, SymDiag improves detection of unfaithful reasoning and provides substantially more effective feedback for multi-round reasoning repair than outcome-only verification and LLM-based judging, offering a principled foundation for trustworthy and scalable reasoning diagnosis.
Chinese Translation
大型语言模型（LLMs）越来越多地作为数据驱动的推理工具，但即使最终答案正确，它们的思维链（CoT）也可能不可靠。现有的大多数“验证”信号并不具备诊断功能：答案匹配仅观察结果，LLM作为评判者提供主观且不可验证的批评，而标量奖励（例如，PRMs/RMs）对多步骤推导失败的原因几乎没有提供洞见。我们提出了 extbf{SymDiag}，一个神经符号框架， extbf{将推理验证重新构建为结构化的失败诊断}。SymDiag将自然语言的思维链转换为符号约束，并执行逐步的可满足性/蕴涵检查，以（i）定位失败步骤和（ii）生成可验证的诊断证据，包括反例、不一致证据和缺失前提指示符。一个核心挑战是，明显的“逻辑违反”可能是由真正的推理缺陷或神经到符号的翻译噪声引起的。因此，SymDiag引入了一个自审计器，通过双重符号编码一致性检查，将翻译错误与推理错误区分开，从而在部分可观察性下实现稳健的诊断。在各种数学、逻辑、科学和一般推理基准测试中，SymDiag提高了对不可靠推理的检测能力，并为多轮推理修复提供了比仅基于结果的验证和LLM评判更有效的反馈，为可信赖和可扩展的推理诊断提供了原则性基础。
cs.AI / 136 / 2608.08794
Deferred Audio Pruning with Local Audio-Visual Dynamics for Omni-LLMs
具有局部音视频动态的延迟音频剪枝用于全模态大语言模型
Abstract
Omni-modal LLMs jointly process audio, video, and text, but long multimodal sequences incur substantial prefill and KV-cache costs. Existing omni-modal compression methods primarily focus on pre-LLM token reduction, leaving modality-specific compression across the LLM boundary underexplored. We propose A-PACK, a two-stage framework that defers audio pruning until query-conditioned multimodal interactions emerge. Our analysis shows that audio exhibits higher task-relevant information density and representational diversity per token than video. We further find that local audio-visual dynamics provide a more effective cue for visual selection than token-wise matching. We therefore preserve audio and compress video with local dynamics before the LLM, then progressively prune low-relevance audio and visual tokens and their KV-cache entries inside the LLM. Across four benchmarks on Qwen2.5-Omni-7B/3B, A-PACK achieves the strongest average performance among the evaluated prior methods while reducing prefill FLOPs by up to 78% and improving decoding throughput by up to 2.21x.
Chinese Translation
全模态大语言模型（Omni-modal LLMs）共同处理音频、视频和文本，但长的多模态序列会产生大量的预填充和键值缓存（KV-cache）成本。现有的全模态压缩方法主要集中在预处理阶段的令牌减少，而在LLM边界内的模态特定压缩尚未得到充分探索。我们提出了A-PACK，一个两阶段框架，它将音频剪枝延迟到查询条件下的多模态交互出现时。我们的分析表明，音频在每个令牌上表现出比视频更高的任务相关信息密度和表示多样性。我们进一步发现，局部音视频动态提供了比逐令牌匹配更有效的视觉选择线索。因此，我们在LLM之前保留音频并利用局部动态压缩视频，然后逐步剪除低相关性的音频和视觉令牌及其KV-cache条目。在Qwen2.5-Omni-7B/3B的四个基准测试中，A-PACK在评估的先前方法中实现了最强的平均性能，同时将预填充的FLOPs减少了多达78%，并将解码吞吐量提高了多达2.21倍。
cs.AI / 137 / 2608.08802
Improving Generalization Robustness of Multimodal RLVR
提高多模态可验证奖励强化学习的泛化鲁棒性
Abstract
Reinforcement Learning with Verifiable Rewards (RLVR) makes Multimodal Large Language Models more accurate, but the gains are brittle: simply paraphrasing a question or changing the prompt template can degrade them, which challenges reliable deployment in high-stakes scenarios like medical VQA. We trace this to two issues of the standard RL objective. First, the binary verifier conflates format with content, so the reward signal cannot tell a wrong answer apart from a misformatted one. Second, the training distribution covers only a thin slice of the real-world prompts that the model might meet at deployment, so policies that perform well on the training distribution can behave differently under unseen prompts during test. Both failures call for a robust post-training method that helps the policy cover a broader distribution of semantically equivalent prompts, and we identify two measures that help achieve this objective: separating format from semantics in the reward, and applying policy invariance across perturbed prompts with equivalent semantics. We therefore propose Prompt-Invariant RLVR (PIRL), consisting of a dynamic trinary reward and a consistency regularizer based on an embedding-space adversary. Under stress testing, PIRL's average accuracy on benchmarks drops by only $\le 1\%$, where GRPO drops ~3%. On dynamic evaluation, PIRL also achieves the smallest performance drop.
Chinese Translation
可验证奖励的强化学习（Reinforcement Learning with Verifiable Rewards, RLVR）使多模态大型语言模型更为准确，但其提升效果却相对脆弱：简单的改述问题或更改提示模板可能导致性能下降，这在医疗视觉问答等高风险场景中挑战了可靠部署。我们将此归因于标准强化学习目标的两个问题。首先，二元验证器将格式与内容混淆，因此奖励信号无法区分错误答案与格式错误的答案。其次，训练分布仅覆盖模型在部署时可能遇到的真实世界提示的一个狭窄切片，因此在训练分布上表现良好的策略在测试时可能会在未见提示下表现不同。这两种失败都呼唤一种鲁棒的后训练方法，以帮助策略覆盖更广泛的语义等效提示分布，我们确定了两项措施以实现这一目标：在奖励中将格式与语义分离，以及在具有等效语义的扰动提示之间应用策略不变性。因此，我们提出了提示不变的可验证奖励强化学习（Prompt-Invariant RLVR, PIRL），其由动态三元奖励和基于嵌入空间对抗者的一致性正则化器组成。在压力测试下，PIRL在基准测试中的平均准确率仅下降$ ext{≤} 1 ext{ extperthousand}$，而GRPO下降约3%。在动态评估中，PIRL也实现了最小的性能下降。
cs.AI / 138 / 2608.08806
Three Generations of Healthcare IT: From the Digital Record to the Computable Care Process
医疗信息技术的三代发展：从数字记录到可计算的护理过程
Abstract
Objective. Healthcare IT is usually organized by the technologies it adopts. We instead organize it by the unit of information a system makes computable, and describe a computational layer whose object is patient-specific clinical intent. Approach. We give criteria for a computational layer, derive three (record, clinical state, and a proposed layer of intent), and formalize the Actionable Clinical Record (ACR) as the atomic object of the third layer. Discussion. The framework distinguishes prescribed, observed, and intended process; existing standards represent intent once it is structured but do not recover it from natural communication, the capability we localize. The ACR is complementary to FHIR workflow resources, guidelines, and process mining; a companion feasibility study illustrates tractability for one narrow subproblem. Conclusion. Computable clinical intent is a coherent research direction; the ACR, its readiness ladder, and an executable-correctness evaluation framework are reusable constructs for subsequent work to extend, evaluate, or falsify.
Chinese Translation
目的：医疗信息技术通常根据其采用的技术进行组织。我们则根据系统使信息可计算的单位进行组织，并描述一个以患者特定临床意图为对象的计算层。方法：我们给出了计算层的标准，推导出三个层次（记录、临床状态和一个提议的意图层），并将可操作临床记录（Actionable Clinical Record, ACR）形式化为第三层的基本对象。讨论：该框架区分了规定的、观察到的和预期的过程；现有标准在意图被结构化后进行表示，但无法从自然交流中恢复意图，这是我们所定位的能力。ACR与FHIR工作流资源、指南和过程挖掘互为补充；一个伴随的可行性研究展示了一个狭窄子问题的可处理性。结论：可计算的临床意图是一个连贯的研究方向；ACR、其准备度阶梯以及可执行正确性评估框架是后续工作扩展、评估或证伪的可重用构件。
cs.AI / 139 / 2608.08822
Automated Generation of Complexity-Validated Decision Scenarios Using Large Language Models
利用大型语言模型自动生成复杂性验证的决策场景
Abstract
Cognitive decision-making research depends on diverse scenarios with carefully controlled complexity, yet manual production is slow, inconsistent, and biased. We developed an automated pipeline that uses LLms to generate structured decision scenarios and validates their complexity through a composite framework rooted in established task-complexity theory. We evaluated 4,238 scenarios across multiple domains and complexity tiers. Measurement validation met rigorous psychometric standards. Agreement among five independent model families was nearly perfect, with an intraclass correlation coefficient of 0.997 and a kappa of 0.971. Known-groups validity demonstrated large separation between tiers, with an eta-squared of 0.587 and all pairwise comparisons significant at p less than .001. Factor analysis revealed a dominant complexity construct, with loadings between 0.87 and 0.96 across three frameworks, while interactivity formed a weaker secondary dimension at 0.34. Discriminant validity was limited by a strong relationship between complexity and text length that persisted after controlling for tier, yielding a partial correlation of 0.86. This constrains construct purity but does not undermine the instrument's tier-grading function. Model analyses showed a negative association between throughput and schema pass rate (r = -0.967, p = .007, n = 5), suggesting a speed-quality trade-off, though largely driven by one high-throughput model. Llama 4 Maverick generated scenarios fastest at 134 per minute versus 25 for DeepSeek Chat V3.2, but underproduced complex-tier scenarios, whereas DeepSeek Chat V3.2 balanced domain coverage with high schema compliance. The system demonstrated strong psychometric properties, enabling reliable classification into Simple, Moderate, and Complex tiers and providing the measurement infrastructure needed for downstream cognitive assessment of AI systems
Chinese Translation
认知决策研究依赖于具有精心控制复杂性的多样场景，但手动生成过程缓慢、不一致且存在偏见。我们开发了一个自动化流程，利用大型语言模型（LLMs）生成结构化的决策场景，并通过基于已建立任务复杂性理论的复合框架验证其复杂性。我们评估了跨多个领域和复杂性层级的4,238个场景。测量验证符合严格的心理测量标准。五个独立模型家族之间的协议几乎完美，组内相关系数为0.997，Kappa值为0.971。已知群体效度显示层级之间存在显著分离，eta平方为0.587，所有成对比较在p < .001下显著。因子分析揭示了一个主导的复杂性构念，三个框架下的载荷在0.87到0.96之间，而交互性则形成了一个较弱的次要维度，载荷为0.34。判别效度受到复杂性与文本长度之间强关系的限制，这种关系在控制层级后依然存在，部分相关系数为0.86。这限制了构念的纯度，但并不削弱工具的层级评分功能。模型分析显示，通过量与模式通过率之间存在负相关（r = -0.967, p = .007, n = 5），暗示存在速度与质量的权衡，尽管主要由一个高通量模型驱动。Llama 4 Maverick以每分钟134个的速度生成场景，而DeepSeek Chat V3.2为25个，但在复杂层级场景的生成上不足，而DeepSeek Chat V3.2在领域覆盖与高模式合规性之间取得了平衡。该系统展示了强大的心理测量特性，能够可靠地分类为简单、中等和复杂层级，并提供了下游认知评估AI系统所需的测量基础设施。
cs.AI / 140 / 2608.08830
PROSLEX: A Novel Dataset for Expert-Annotated Legal Statute Prediction for Indian Judiciary
PROSLEX：用于印度司法体系专家标注法律条款预测的新数据集
Abstract
Legal Statute Prediction (LSP) involves automatically identifying relevant legal statutes given factual descriptions in legal documents, typically framed as a multi-label classification task within natural language processing and information retrieval research. While recent advances have begun incorporating Large Language Models (LLMs) for statute prediction, current approaches primarily focus on accuracy metrics without addressing the critical need for legal reasoning, a fundamental requirement in judicial contexts where decisions must be explainable and justifiable. To address this research gap, we present PROSLEX (PRediction Of Statutes and LEgal eXplanation), a comprehensive dataset comprising 1,623 expert-annotated legal documents from the Indian context. Each document is paired with statute predictions and detailed explanations, totaling 7,450 explanations, capturing the underlying legal reasoning. Using this dataset, we systematically evaluate various prompting strategies, including zero-shot, few-shot, chain-of-thought, and tree-of-thoughts approaches, to generate both statute predictions and their corresponding legal rationales. Our evaluation framework measures not only predictive performance but also the coherence and legal validity of generated explanations, positioning PROSLEX as a benchmark for developing explainable AI systems that can support legal practitioners while advancing research in interpretable legal NLP. To ensure reproducibility, we have made our PROSLEX dataset and model code available on GitHub: https://github.com/subinay494/Legal_Statute_Prediction_Explanation.
Chinese Translation
法律条款预测（Legal Statute Prediction, LSP）涉及在法律文件中自动识别与事实描述相关的法律条款，通常被视为自然语言处理和信息检索研究中的多标签分类任务。尽管最近的进展开始将大型语言模型（Large Language Models, LLMs）应用于条款预测，但当前的方法主要关注准确性指标，而未能解决法律推理这一关键需求，而法律推理在司法环境中是基本要求，因为裁决必须是可解释和可辩护的。为了解决这一研究空白，我们提出了PROSLEX（法律条款及其解释的预测，PRediction Of Statutes and LEgal eXplanation），这是一个包含1,623份来自印度背景的专家标注法律文件的综合数据集。每份文件都配有法律条款预测和详细解释，总计7,450条解释，捕捉了潜在的法律推理。利用该数据集，我们系统地评估了多种提示策略，包括零样本（zero-shot）、少样本（few-shot）、思维链（chain-of-thought）和思维树（tree-of-thoughts）方法，以生成法律条款预测及其相应的法律理由。我们的评估框架不仅衡量预测性能，还评估生成解释的一致性和法律有效性，使PROSLEX成为开发可解释人工智能系统的基准，这些系统可以支持法律从业者，同时推动可解释法律自然语言处理的研究。为了确保可重复性，我们已在GitHub上公开了我们的PROSLEX数据集和模型代码： https://github.com/subinay494/Legal_Statute_Prediction_Explanation。
cs.AI / 141 / 2608.08852
Findings of the First Teaching Monster Challenge: A Benchmark of Pedagogical Content Knowledge in AI Agents
首次教学怪兽挑战的发现：人工智能代理的教学内容知识基准
Abstract
AI agents can now solve problems, answer like subject experts, and generate long-form multimodal content. However, whether they can adapt a lesson to fit a specified learner, which education calls Pedagogical Content Knowledge (PCK), has not been benchmarked. To measure it, we introduce the Teaching Monster Challenge, the first instructional video generation benchmark to treat the learner persona as an explicit evaluation criterion. Each system is given a topic and a learner persona and must generate a complete instructional video. Every video is screened by an LLM-judge, ranked by crowd pairwise voting, and finalized by an expert panel. The first edition shows that today's systems handle the content well but are far weaker at presenting it and adapting it to the learner. The same process exposes a limit of automatic judging. The LLM-judge separates a clear low-performing tail but ranks the strongest systems poorly. The strongest systems receive nearly identical scores from the judge, so its ranking of them does not match human preference. Progress therefore requires not only better teaching systems but also better automatic judges, and we release the benchmark, rubric, and human judgments as a testbed for both.
Chinese Translation
人工智能代理现在能够解决问题、像学科专家一样回答问题，并生成长篇多模态内容。然而，它们是否能够根据特定学习者的需求调整课程，这在教育领域被称为教学内容知识（Pedagogical Content Knowledge, PCK），尚未进行基准测试。为了测量这一点，我们引入了教学怪兽挑战，这是第一个将学习者角色作为明确评估标准的教学视频生成基准。每个系统被分配一个主题和一个学习者角色，并必须生成一个完整的教学视频。每个视频都由一个大型语言模型（LLM）评审，经过众包配对投票排名，并由专家小组最终确定。第一版结果显示，当前的系统在内容处理方面表现良好，但在呈现内容和适应学习者方面则明显较弱。同样的过程揭示了自动评审的局限性。LLM评审能够清晰地区分出表现较差的系统，但对最强系统的排名却不佳。最强系统从评审那里获得几乎相同的分数，因此其排名与人类偏好不符。因此，进步不仅需要更好的教学系统，还需要更好的自动评审工具，我们将基准、评分标准和人类评判结果发布为两者的测试平台。
cs.AI / 142 / 2608.08881
Theory-Guided Deception Detection: A RAG-Based Artificial Intelligence Exploration
理论指导的欺骗检测：基于RAG的人工智能探索
Abstract
The current work developed seven Retrieval-Augmented Generation (RAG) models based on leading deception theories and compared how deception judgments were made relative to baseline models. Across 700 statements drawn from five published deception datasets, four large language models (gpt-4o, claude-sonnet-4-6, ollama/llama3, deepseek-v4-flash), and two run-types (RAG vs. baseline), a total of 39,200 deception judgments were rendered. Detection accuracies were consistent with typical human accuracies and not statistically different across RAG (54.5%) and baseline models (54.6%). RAG-based models (57.0%) were less truth-biased than baseline models (59.7%), but the effect size was quite small. Theoretical perspective mattered little for accuracy yet mattered substantially for response bias, which ranged from highly lie-biased (the verifiability approach, 32.2%) to highly truth-biased (truth-default theory, 88.1%). Content effects and model effects further moderated the results. Theory-guided AI judgments are unreliable with current parameters, yet they might show promise with additional datasets, model testing, and theory-to-data matching.
Chinese Translation
本研究开发了七种基于领先欺骗理论的检索增强生成（RAG）模型，并比较了相对于基线模型的欺骗判断方式。在从五个已发布的欺骗数据集中抽取的700条陈述中，使用了四种大型语言模型（gpt-4o、claude-sonnet-4-6、ollama/llama3、deepseek-v4-flash）和两种运行类型（RAG与基线），共进行了39,200次欺骗判断。检测准确率与典型的人类准确率一致，RAG模型（54.5%）与基线模型（54.6%）之间没有统计学差异。基于RAG的模型（57.0%）的真相偏见低于基线模型（59.7%），但效应大小相对较小。理论视角对准确性影响不大，但对反应偏见影响显著，反应偏见范围从高度偏向谎言（可验证性方法，32.2%）到高度偏向真相（真相默认理论，88.1%）。内容效应和模型效应进一步调节了结果。在当前参数下，理论指导的人工智能判断不可靠，但在额外数据集、模型测试和理论与数据匹配的情况下，可能会显示出潜力。
cs.AI / 143 / 2608.08883
AquiLLM: An Architecture for Supporting Tacit Knowledge Capture in Research Groups
AquiLLM：支持研究小组隐性知识捕获的架构
Abstract
Recent advances in retrieval-augmented generation (RAG) and large language models (LLMs) enable researchers to integrate AI into scientific workflows. However, using proprietary commercial AI systems raises concerns about transparency, reproducibility and privacy, which are essential for scientific practices. To this end, AquiLLM was developed as an open-source modular RAG-LLM framework using open-weight models, designed to support research groups in capturing tacit knowledge. In this work, we present a series of architectural improvements and feature enhancements to AquiLLM, including local embedding and reranking, multimodal capabilities, OpenAI-compatible inference interfaces, user interface improvements, semantic and episodic memory capabilities, and skills support. These enhancements were informed by discussions with domain experts, including astrophysicists and environmental researchers, and represent a step toward AI systems more closely aligned with scientific research practices.
Chinese Translation
最近在检索增强生成（RAG）和大型语言模型（LLMs）方面的进展，使研究人员能够将人工智能集成到科学工作流程中。然而，使用专有商业人工智能系统引发了关于透明性、可重复性和隐私的担忧，这些都是科学实践中至关重要的。为此，AquiLLM被开发为一个开源模块化的RAG-LLM框架，使用开放权重模型，旨在支持研究小组捕获隐性知识。在本研究中，我们提出了一系列对AquiLLM的架构改进和功能增强，包括本地嵌入和重新排序、多模态能力、与OpenAI兼容的推理接口、用户界面改进、语义和情节记忆能力以及技能支持。这些增强是通过与领域专家（包括天体物理学家和环境研究人员）的讨论而得出的，代表了朝着与科学研究实践更紧密对齐的人工智能系统迈出的重要一步。
cs.AI / 144 / 2608.08888
Full-bandwidth transformer
全带宽变换器
Abstract
Autoregressive transformers compute along two axes: horizontally across generated tokens, and vertically through model depth. Dense attention gives each token broad horizontal access to the past, but the vertical feedback channel between decoding steps remains narrow: only the sampled token returns to the bottom of the stack, while the top-layer hidden state is discarded. We introduce the \emph{full-bandwidth transformer}, which widens this channel with \emph{latent feedback}: at each decoding step, the previous top-layer hidden state is fused with the sampled token embedding through a gated linear unit and fed back as the next input. Latent feedback lets non-verbalized computation re-enter the stack with a renewed depth budget, while preserving the standard transformer architecture, KV cache, and language-modeling objective. To train full-bandwidth transformers without losing parallel teacher forcing, we use a scheduled multi-pass objective that introduces latent feedback late in pretraining and mixes a small fraction of deeper feedback passes for stability. We train 1B-parameter full-bandwidth transformers up to 400B tokens and find that latent feedback improves validation loss, 5-shot language-model evaluation, math and coding generation, and instruction-tuned performance. With negligible per-token decoding overhead, full-bandwidth transformers match or approach standard transformers trained with roughly $1.5\times$ more tokens, and manage to produce shorter reasoning traces at equal or better accuracy.
Chinese Translation
自回归变换器沿两个轴进行计算：在生成的标记之间水平计算，以及通过模型深度进行垂直计算。稠密注意力使每个标记能够广泛地水平访问过去，但解码步骤之间的垂直反馈通道仍然狭窄：只有采样的标记返回到堆栈底部，而顶层的隐藏状态则被丢弃。我们提出了 extit{全带宽变换器}，通过 extit{潜在反馈}扩展了这一通道：在每个解码步骤中，前一个顶层隐藏状态通过门控线性单元与采样的标记嵌入融合，并作为下一个输入反馈。潜在反馈使非语言化的计算能够以更新的深度预算重新进入堆栈，同时保持标准变换器架构、KV缓存和语言建模目标。为了在不失去并行教师强制的情况下训练全带宽变换器，我们使用了一种调度的多次传递目标，该目标在预训练的后期引入潜在反馈，并混合少量更深的反馈传递以确保稳定性。我们训练了1B参数的全带宽变换器，处理了高达400B的标记，并发现潜在反馈改善了验证损失、5-shot语言模型评估、数学和编码生成，以及指令调优性能。在每个标记解码开销微不足道的情况下，全带宽变换器的性能与训练了大约$1.5 imes$更多标记的标准变换器相匹配或接近，并且能够以相等或更好的准确性生成更短的推理轨迹。
cs.AI / 145 / 2608.08889
LLM Reasoning for Subjective Tasks: Failure Modes, Mitigation, and Dynamic Reasoning Routing
大型语言模型在主观任务中的推理：失败模式、缓解措施与动态推理路由
Abstract
Recommendation systems thrive on personalization, where ''correctness'' is rarely a binary truth but a matter of subjective human preference. As Large Language Models (LLMs) are deployed as autonomous verifiers of safety and quality guidelines, they face a distinctive challenge: context-aware preference alignment. Recent gains in Reinforcement Learning with Verifiable Rewards (RLVR) are indexed mostly on objective, mathematical tasks. Through a large-scale study spanning both proprietary and open-source models on four real-world verification tasks from a production recommender platform, we ask whether explicit reasoning generalizes to subjective, human-centric industry rubrics. We expose a fundamental vulnerability: rigid, math-centric reasoning traces actively degrade verification, and applying standard RLVR triggers a phenomenon we term reasoning collapse, in which the policy abandons deliberation in favor of rapid heuristic guessing. We introduce a conditional length-penalized post-training algorithm that intertwines verification accuracy with bounded reasoning length, halting collapse and recovering performance. Finally, we show that a reasoning trace's efficacy is tightly coupled with its socio-linguistic framing: across 1500 synthesized personas, verification accuracy swings by nearly 0.38 macro-F1 depending solely on the adopted reasoning persona---evidence that much subjective-verification error is really reasoning-style mismatch. This observation motivates a mid-training architecture that routes reasoning through contextually aligned personas. This work offers both a scalable algorithmic patch and a long-term architectural blueprint for aligning reasoning models with real-world subjective constraints.
Chinese Translation
推荐系统依赖个性化，其中“正确性”很少是二元真理，而是主观人类偏好的问题。随着大型语言模型（LLMs）作为安全和质量指南的自主验证者被部署，它们面临着一个独特的挑战：上下文感知的偏好对齐。近期在可验证奖励的强化学习（RLVR）方面的进展主要集中在客观的数学任务上。通过对来自一个生产推荐平台的四个真实世界验证任务进行的大规模研究，我们探讨了显式推理是否能够推广到主观的人本行业标准。我们揭示了一个基本的脆弱性：僵化的数学中心推理痕迹会积极降低验证效果，而应用标准RLVR会引发我们称之为推理崩溃的现象，在这种现象中，策略放弃深思熟虑而选择快速的启发式猜测。我们提出了一种条件长度惩罚的后训练算法，将验证准确性与有限的推理长度交织在一起，停止崩溃并恢复性能。最后，我们展示了推理痕迹的有效性与其社会语言框架紧密相关：在1500个合成角色中，验证准确性根据采用的推理角色的不同而波动近0.38的宏观F1分数——这表明许多主观验证错误实际上是推理风格不匹配。这个观察促使我们提出一种中期训练架构，通过上下文对齐的角色来引导推理。本研究提供了一个可扩展的算法补丁和一个长期的架构蓝图，以将推理模型与现实世界的主观约束对齐。
cs.AI / 146 / 2608.08896
From Manuals to Maintenance: Fine-Tuning MedGemma for Multi-Modal Imaging System Support in Low-Resource Settings
从手册到维护：在低资源环境中为多模态成像系统支持微调 MedGemma
Abstract
Imaging device downtime is a major barrier to healthcare delivery in low- and middle-income countries (LMICs), often driven by limited access to specialized biomedical engineering support. We present a multi-modality medical equipment maintenance question-answering (QA) framework and demonstrate the fine-tuning of a medical foundation model for specialized technical troubleshooting tasks. Guided by a multi-country survey across nine LMICs, we curated technical manuals from MRI and ultrasound systems to generate the INGENZI_DatasetV1, containing 10,294 high-quality, filtered QA-context pairs. Using QLoRA-based parameter-efficient fine-tuning, we adapted the MedGemma-4b-it model to interpret system error logs and generate step-by-step equipment repair instructions. Compared to the baseline model, the fine-tuned system achieved substantial improvements across metrics, including F1 score (0.22 to 0.38), ROUGE-2 (0.18 to 0.41), and BERTScore F1 (0.86 to 0.91). These metric gains demonstrate that the model generates significantly more precise and procedurally accurate technical responses to new troubleshooting queries. This work establishes a reliable foundation for AI-assisted diagnostic and maintenance tools in resource-constrained settings.
Chinese Translation
成像设备的停机时间是低收入和中等收入国家（LMICs）医疗服务交付的主要障碍，通常是由于缺乏专业的生物医学工程支持。我们提出了一种多模态医疗设备维护问答（QA）框架，并展示了针对专业技术故障排除任务的医疗基础模型的微调。通过对九个LMICs进行的多国调查指导，我们从MRI和超声系统中整理了技术手册，生成了包含10,294对高质量、经过筛选的QA上下文对的INGENZI_DatasetV1。利用基于QLoRA的参数高效微调，我们调整了MedGemma-4b-it模型，以解释系统错误日志并生成逐步的设备维修说明。与基线模型相比，微调后的系统在多个指标上取得了显著改善，包括F1分数（从0.22提高到0.38）、ROUGE-2（从0.18提高到0.41）和BERTScore F1（从0.86提高到0.91）。这些指标的提升表明，该模型能够生成对新故障排除查询更为精确和程序上更为准确的技术响应。这项工作为资源有限环境中的AI辅助诊断和维护工具奠定了可靠的基础。
cs.AI / 147 / 2608.08926
Decoding Phenotypes: A Framework for Fusing Genomic Language Models and Neuroimaging
解码表型：融合基因组语言模型与神经影像学的框架
Abstract
Neuroimaging and genetic testing are two important clinical references for nervous system diseases, offering complementary diagnostic information. However, integrating genomic and neuroimaging data for precise disease diagnosis is challenging due to cross-modality heterogeneity. Existing imaging-genetics approaches mainly encode genetic information as hard-coded labels, which lose the local sequence context around disease-associated variants. To address this limitation, we propose GeneFuse, a multimodal learning framework that aligns genetic representations from pre-trained Genomic Language Models (GLMs) with features extracted from images. GeneFuse integrates two components: (1) Genotype-Conditioned Feature Modulation (GCFM), a FiLM-inspired module that uses genomic embeddings to modulate image feature maps; and (2) Uncertainty-aware Genomic Residual Fusion (U-GRF), a fusion strategy that uses imaging-derived predictive uncertainty to gate the contribution of genotypic features. We evaluate GeneFuse on early cognitive decline identification (NC vs. MCI) and dementia screening (NC vs. AD). In the APOE-centered setting, GeneFuse achieves AUROCs of 0.77 and 0.83, outperforming existing imaging-genetics fusion methods. These results indicate that GLM-derived genomic embeddings provide additional information to imaging.
Chinese Translation
神经影像学和基因检测是神经系统疾病的重要临床参考，提供互补的诊断信息。然而，由于跨模态异质性，整合基因组和神经影像数据以实现精确的疾病诊断面临挑战。现有的影像-遗传学方法主要将遗传信息编码为硬编码标签，这会丧失与疾病相关变异周围的局部序列上下文。为了解决这一局限性，我们提出了GeneFuse，一个多模态学习框架，它将来自预训练基因组语言模型（Genomic Language Models, GLMs）的遗传表征与从图像中提取的特征对齐。GeneFuse集成了两个组件：（1）基因型条件特征调制（Genotype-Conditioned Feature Modulation, GCFM），一个受FiLM启发的模块，利用基因组嵌入来调制图像特征图；（2）不确定性感知基因组残差融合（Uncertainty-aware Genomic Residual Fusion, U-GRF），一种融合策略，利用影像派生的预测不确定性来调节基因型特征的贡献。我们在早期认知衰退识别（正常对轻度认知障碍，NC vs. MCI）和痴呆筛查（正常对阿尔茨海默病，NC vs. AD）上评估了GeneFuse。在以APOE为中心的设置中，GeneFuse达到了0.77和0.83的AUROC，超越了现有的影像-遗传学融合方法。这些结果表明，GLM派生的基因组嵌入为影像提供了额外的信息。
cs.AI / 148 / 2608.08935
Integrated Multimodal AI System for Retrieval-Augmented Reasoning, Object Sensing, and Damage Analysis
集成多模态人工智能系统用于检索增强推理、物体感知和损伤分析
Abstract
This work presents a unified multimodal AI system for damage assessment that integrates retrieval-augmented generation (RAG) models, thermal spectrum perception, vision foundation model pipelines, and exploratory wireless signal sensing. A RAG component is developed to ground a locally hosted language model in project-specific documentation, including specialized damage level classification criteria to mitigate hallucinations during inference. Controlled comparisons against static few-shot prompting demonstrate that dynamic retrieval improves grounding and factual consistency. We further compare vector-based RAG with a knowledge graph variant constructed via entity-relation extraction, and show that graph-based retrieval produces stronger responses for damage assessment queries requiring cross-document reasoning, motivating hybrid dense, sparse, and graph-aware retrieval. To address limitations of EO imagery under adverse lighting and weather conditions, infrared (IR)/thermal sensing is employed for object detection and segmentation. Our detectors generate candidate detections, yielding improved segmentation of a broad array of objects. Paired IR versus visible spectrum tracking experiments reveal failure modes, motivating multimodal fusion for robust object detection and damage analysis. Vision foundation and vision-language models are leveraged to generate synthetic damage imagery and classify damage severity with high accuracy, supporting training and validation of downstream damage assessment models. Finally, exploratory Wireless-based sensing demonstrates potential to detect presence, motion, and post-event environmental changes where EO and IR sensing are ineffective.
Chinese Translation
本研究提出了一种统一的多模态人工智能系统，用于损伤评估，该系统集成了检索增强生成（RAG）模型、热谱感知、视觉基础模型流程和探索性无线信号感知。我们开发了一个RAG组件，以将本地托管的语言模型与项目特定文档相结合，包括专业的损伤等级分类标准，以减轻推理过程中的幻觉现象。与静态少样本提示的受控比较表明，动态检索改善了基础和事实一致性。我们进一步将基于向量的RAG与通过实体-关系提取构建的知识图谱变体进行比较，结果表明，基于图的检索在需要跨文档推理的损伤评估查询中产生了更强的响应，推动了混合密集、稀疏和图感知检索的发展。为了应对在不利光照和天气条件下的电光（EO）图像的局限性，我们采用红外（IR）/热感知进行物体检测和分割。我们的检测器生成候选检测，改善了各种物体的分割效果。配对的红外与可见光谱跟踪实验揭示了失败模式，推动了多模态融合以实现稳健的物体检测和损伤分析。我们利用视觉基础模型和视觉-语言模型生成合成损伤图像，并以高准确度分类损伤严重性，支持下游损伤评估模型的训练和验证。最后，探索性基于无线的感知展示了在电光和红外感知无效的情况下，检测存在、运动和事件后环境变化的潜力。
cs.AI / 149 / 2608.08939
Not an A11y: How Android Accessibility Exposes Mobile AI Agents to Indirect Prompt Injection
非无障碍性：安卓无障碍功能如何使移动AI代理暴露于间接提示注入
Abstract
The rise of autonomous AI agents represents a major paradigm shift in how users interact with mobile devices. Frameworks such as MobileRun and Mobile-Use can autonomously navigate Android applications and execute complex multi-step tasks. To interpret user interfaces, these frameworks rely primarily on Android accessibility (A11y) trees and secondarily on visual screenshots. In this paper, we demonstrate that this architectural dependence on unsanitized accessibility metadata, together with visual input, introduces a systemic vulnerability to indirect prompt injection. We show that adversarial prompts can cause autonomous agents to abandon their original objectives, violate context boundaries, and perform unauthorized device actions. Our empirical evaluation demonstrates goal hijacking, context drift, and unauthorized actions across visually hidden and fully exposed attack scenarios. In aggregate, MobileRun reaches an attack success rate of 0.822 with Gemma4:31B, while Mobile-Use with Qwen3.6:35B reduces this to 0.150 but does not eliminate context drift or unauthorized actions. These findings reveal that current mobile agent frameworks fail to enforce semantic context boundaries, treating passive environmental text as trusted instructions. Finally, we present a taxonomy of these attacks and discuss the need for zero-trust input validation, dedicated security agents, and strict context isolation within mobile agent architectures.
Chinese Translation
自主AI代理的崛起代表了用户与移动设备交互方式的重大范式转变。像MobileRun和Mobile-Use这样的框架能够自主导航安卓应用并执行复杂的多步骤任务。为了解析用户界面，这些框架主要依赖安卓无障碍（A11y）树，其次依赖视觉截图。本文展示了这种对未清理的无障碍元数据的架构依赖，加上视觉输入，引入了对间接提示注入的系统性脆弱性。我们表明，对抗性提示可以导致自主代理放弃其原始目标，违反上下文边界，并执行未经授权的设备操作。我们的实证评估展示了目标劫持、上下文漂移和在视觉隐藏和完全暴露的攻击场景中进行未经授权的操作。总体而言，MobileRun在Gemma4:31B下的攻击成功率达到0.822，而Mobile-Use在Qwen3.6:35B下将这一成功率降低至0.150，但并未消除上下文漂移或未经授权的操作。这些发现揭示了当前移动代理框架未能强制执行语义上下文边界，将被动环境文本视为可信指令。最后，我们提出了这些攻击的分类法，并讨论了在移动代理架构中需要零信任输入验证、专用安全代理和严格的上下文隔离。
cs.AI / 150 / 2608.08959
Depth-Aware Implicit Neural Representation Priors for 3D Gravity Inversion
用于三维重力反演的深度感知隐式神经表示先验
Abstract
Gravimetry images subsurface density contrasts associated with geological structures, geothermal systems, and intrusive bodies. Recovering a three-dimensional density model from gravity observations is highly ill-posed because of its non-uniqueness, limited data coverage, and the attenuation of the gravity field with depth. Classical inversion methods rely on explicit regularization and parameter tuning, whereas supervised deep-learning approaches require representative gravity--density pairs that are rarely available. This paper proposes an unsupervised depth-aware implicit neural representation for 3D gravity inversion. The density volume is represented by multiple coordinate-based neural networks assigned to overlapping depth slabs and optimized directly from the observed gravity measurements through the sensitivity matrix. Slab-specific Fourier features, physics-based depth gains, and scheduled regularization provide structural priors without requiring labeled density models. Experiments on four synthetic scenarios show that the proposed method provides better overall performance in terms of RMSE, PSNR, and SSIM than the evaluated conventional and neural baselines. It also recovers more compact and spatially coherent density bodies, improves the separation of nearby anomalies, preserves internal structures, and reconstructs their vertical extent better. These results indicate that the proposed depth-aware formulation helps to mitigate the depth ambiguity inherent in gravity inversion. In the field experiment, where no ground-truth density model was available, the method produced compact, separated, and vertically coherent anomalies consistent with the observed gravity pattern.
Chinese Translation
重力测量图像显示了与地质结构、地热系统和侵入体相关的地下密度对比。从重力观测中恢复三维密度模型是一个高度不适定的问题，因为它存在非唯一性、数据覆盖有限以及重力场随深度衰减等问题。经典的反演方法依赖于显式正则化和参数调优，而监督深度学习方法则需要代表性的重力-密度对，这种数据很少可得。本文提出了一种无监督的深度感知隐式神经表示，用于三维重力反演。密度体通过多个基于坐标的神经网络表示，这些网络被分配到重叠的深度层，并通过灵敏度矩阵直接从观测到的重力测量中优化。特定层的傅里叶特征、基于物理的深度增益和调度正则化提供了结构先验，而无需标记的密度模型。在四个合成场景上的实验表明，所提出的方法在均方根误差（RMSE）、峰值信噪比（PSNR）和结构相似性指数（SSIM）方面的整体表现优于评估的传统和神经基线。它还恢复了更紧凑和空间一致的密度体，改善了相邻异常的分离，保持了内部结构，并更好地重建了它们的垂直范围。这些结果表明，所提出的深度感知公式有助于减轻重力反演中固有的深度模糊性。在没有真实密度模型可用的现场实验中，该方法产生了与观察到的重力模式一致的紧凑、分离和垂直一致的异常。
cs.AI / 151 / 2608.08960
Reading is not Reasoning: Bridging the Agentic Policy Gap in Vision-Text Compression
阅读不是推理：弥合视觉-文本压缩中的代理政策差距
Abstract
Multi-step language-model agents repeatedly process growing interaction histories, leading to substantial context costs. Vision--text compression reduces these costs by rendering history as images, but the resulting modality shift creates a marked capability gap. Through controlled evaluations of history recovery, matched-state decisions, and complete trajectories, we show that this gap cannot be explained by OCR quality alone. Visual-history agents exhibit systematic drift in action selection, query formulation, stopping, and evidence use, revealing an agentic policy gap. We introduce \textbf{CAPS}, a two-stage \textbf{C}ross-modal \textbf{A}gentic \textbf{P}olicy \textbf{S}elf-distillation framework that uses the same model's stronger text-history policy to supervise its visual-history counterpart. Offline trajectory self-distillation transfers successful text-policy behavior to visual-history inputs, while online policy self-distillation provides dense supervision on states visited by the visual-history policy during reinforcement learning. On SearchQA, CAPS improves over AgentOCR by 5.0\% and 3.4\% with 3B and 7B backbones, respectively. On full-history ALFWorld, the corresponding gains are 15.6\% and 14.5\%. Across settings, CAPS reduces average memory-context cost by up to 63.3\% and peak cost by up to 83.4\% relative to matched text-history policies. These results show that explicit cross-modal policy self-distillation can preserve agent capability under vision--text compression. Our code will be made publicly available in a future release.
Chinese Translation
多步骤语言模型代理反复处理不断增长的交互历史，导致显著的上下文成本。视觉-文本压缩通过将历史呈现为图像来降低这些成本，但由此产生的模态转变造成了明显的能力差距。通过对历史恢复、匹配状态决策和完整轨迹的控制评估，我们表明，这一差距不能仅通过OCR质量来解释。视觉历史代理在行动选择、查询构建、停止和证据使用上表现出系统性的漂移，揭示了代理政策差距。我们提出了 extbf{CAPS}，一个两阶段的 extbf{C}ross-modal extbf{A}gentic extbf{P}olicy extbf{S}elf-distillation框架，利用同一模型更强的文本历史策略来监督其视觉历史对应策略。离线轨迹自蒸馏将成功的文本策略行为转移到视觉历史输入，而在线策略自蒸馏则在强化学习期间对视觉历史策略访问的状态提供密集监督。在SearchQA上，CAPS在3B和7B主干上分别提高了5.0 extbf{ extit{ ext{％}}}和3.4 extbf{ extit{ ext{％}}}。在完整历史的ALFWorld上，相应的增益为15.6 extbf{ extit{ ext{％}}}和14.5 extbf{ extit{ ext{％}}}。在各种设置中，CAPS将平均内存上下文成本降低了高达63.3 extbf{ extit{ ext{％}}}，峰值成本降低了高达83.4 extbf{ extit{ ext{％}}}，相对于匹配的文本历史策略。这些结果表明，显式的跨模态策略自蒸馏可以在视觉-文本压缩下保持代理能力。我们的代码将在未来的发布中公开。
cs.AI / 152 / 2608.08965
CoRe-UIE: Rethinking Coexisting and Region-wise Degradation for Underwater Image Enhancement
CoRe-UIE：重新思考水下图像增强中的共存与区域性退化
Abstract
Underwater images often suffer from diverse and coexisting degradations, including color distortion, scattering haze, texture attenuation, and uneven illumination. These degradations vary across regions and may coexist locally, making conventional uniform restoration difficult to adapt to different degradation patterns. To address this problem, we propose Coexisting and Region-wise Degradation for Underwater Image Enhancement (\textbf{CoRe-UIE}), a degradation-oriented expert collaboration framework. CoRe-UIE combines a content-preserving shared expert with four shared-backbone routed experts for color correction, scattering suppression, texture recovery, and illumination protection. The routed experts share the same architecture but have independent parameters, and are assigned to different regions through input-derived degradation cues and region-adaptive Top-\(k\) routing. We further introduce a Hilbert--Schmidt Independence Criterion (HSIC)-based representation constraint to reduce statistical dependence among expert features and alleviate redundant expert responses. Experiments on UIEB, LSUI, and U45 demonstrate that CoRe-UIE achieves competitive quantitative performance and visually balanced enhancement under diverse underwater degradation conditions.
Chinese Translation
水下图像常常受到多种共存退化的影响，包括颜色失真、散射雾霾、纹理衰减和不均匀照明。这些退化在不同区域表现各异，并可能在局部共存，使得传统的均匀恢复方法难以适应不同的退化模式。为了解决这一问题，我们提出了水下图像增强中的共存与区域性退化框架（Coexisting and Region-wise Degradation for Underwater Image Enhancement， extbf{CoRe-UIE}），这是一个以退化为导向的专家协作框架。CoRe-UIE结合了一个内容保持的共享专家与四个用于颜色校正、散射抑制、纹理恢复和照明保护的共享骨干路由专家。这些路由专家共享相同的架构，但具有独立的参数，并通过输入导出的退化线索和区域自适应Top- extit{k}路由分配到不同区域。我们进一步引入基于Hilbert-Schmidt独立性准则（HSIC）的表示约束，以减少专家特征之间的统计依赖，缓解冗余的专家响应。在UIEB、LSUI和U45数据集上的实验表明，CoRe-UIE在多种水下退化条件下实现了具有竞争力的定量性能和视觉上平衡的增强效果。
cs.AI / 153 / 2608.09025
Context Is Not Authority: Structured Runtime Governance for Financial Market Agents
上下文并非权威：金融市场代理的结构化运行治理
Abstract
Financial agents can turn correct context into an unauthorized effect: a customer-facing commitment, trade, or deployed policy. We present SAGE-Fin, a finance-specific authority-handoff contract that makes the proposed effect, not merely its text, the object of runtime control. SAGE-Fin compiles proposals into typed, adapter-bound candidates; records missing or stale institutional obligations as coverage debt; contracts authority under current market, account, policy, and dialogue state; and requires an exact-artifact receipt whose nominal type matches the consuming response, execution, or policy adapter. Evidence and workflow progress cannot substitute for effect authority, and prior authorization is rechecked after state changes. Across an authored 616-case catalog, five deterministic specifications yield 3,080 outputs; a label-isolated harness obtains 616/616 binary reference-prototype parity, including 3/3 named response-gate fixtures, while 22 tests cover selected paths. These results establish executable conformance, not independent safety accuracy. Separately, SAGE-Fin's response gate processed real customer-facing production requests at a confidential digital-asset platform. An operational team independent of the implementation team reached a strongly positive post-deployment conclusion on practical usefulness and workflow fit, and end-user feedback was also strongly positive. Disclosure permits only the review's independence, stakeholder classes, assessed dimensions, and directional conclusion, so this is qualitative field corroboration rather than an aggregate effect estimate. Three distinct de-identified predecessor failures, with independently confirmed 0/3 interception, ground repeated-emission drift, stale account evidence, and missing escalation state without estimating prevalence or treatment effect.
Chinese Translation
金融代理可以将正确的上下文转化为未经授权的效果：面向客户的承诺、交易或已部署的政策。我们提出了 SAGE-Fin，这是一种特定于金融的权威交接合同，使得所提议的效果，而不仅仅是其文本，成为运行时控制的对象。SAGE-Fin 将提案编译成类型化、适配器绑定的候选项；记录缺失或过时的机构义务作为覆盖债务；在当前市场、账户、政策和对话状态下契约权威；并要求一个确切的工件收据，其名义类型与消费响应、执行或政策适配器相匹配。证据和工作流进展不能替代效果权威，并且在状态变化后会重新检查先前的授权。在一个包含 616 个案例的编撰目录中，五个确定性规范产生了 3,080 个输出；一个标签隔离的测试装置获得了 616/616 的二元参考原型一致性，包括 3/3 个命名响应门装置，而 22 个测试覆盖了选定路径。这些结果确立了可执行的一致性，而非独立的安全准确性。此外，SAGE-Fin 的响应门在一个保密的数字资产平台上处理了真实的面向客户的生产请求。一个独立于实施团队的运营团队对实际有用性和工作流适配达成了强烈积极的部署后结论，最终用户反馈也非常积极。披露仅允许审查的独立性、利益相关者类别、评估维度和方向性结论，因此这是一种定性的现场证实，而非聚合效果估计。三个不同的去标识化前任失败，独立确认的拦截率为 0/3，导致重复发射漂移、过时账户证据和缺失的升级状态，而未估计其流行程度或处理效果。
cs.AI / 154 / 2608.09028
PolicyKG: An Agentic LLM Pipeline for Translating Institutional Policies into SHACL Knowledge Graphs
PolicyKG：将机构政策转化为 SHACL 知识图谱的智能 LLM 流程
Abstract
Institutional policies stay in natural language while the systems that check compliance demand machine-readable constraints. Bridging that gap is still done by hand. PolicyKG closes the loop. It is an LLM pipeline that reads a policy PDF, classifies each sentence as an obligation, permission, or prohibition, lifts the label into first-order deontic logic, and emits SHACL constraints. Four stages run on a LangGraph state machine with per-stage validators. The piece that matters most is the Corpus Adapter: a YAML vocabulary registry that grounds LLM predicates in a target ontology. Retargeting to a new domain means swapping the registry, not retraining a model. On the Asian Institute of Technology Policies and Procedures corpus (1,663 sentences, 443 rules), PolicyKG reaches 86.9% deontic classification accuracy (Cohen's kappa = .709). Three annotators independently re-label a 50-item sample and agree at Fleiss' kappa = .844. SHACL shape correctness on a 69-shape subset is F1 = .866. The FOL path handles 79.2% of rules; the rest go through a direct NL-to-SHACL fallback. We audited every one of the 443 rules for second- or higher-order constructs. An automated regex checklist flagged none, and a first-author pass on the 92 FOL-fallback cases confirmed the same. The exact upper 95% Clopper-Pearson bound on the true HOL rate is 0.67%. This is an audit finding for one corpus, not a proof of FOL sufficiency for institutional policy. Swapping the AIT registry for a GDPR registry raises exact property alignment from 1/15 to 11/15 (Fisher's exact p < .001; Cohen's h = 1.53). On the LexDeMod lease-contract benchmark (N = 200), Macro F1 drops to .370 because lease English uses "shall be entitled" for permission -- exactly the vocabulary mismatch registry swap is meant to fix. Repeated runs produce hash-identical SHACL outputs.
Chinese Translation
机构政策以自然语言形式存在，而检查合规性所需的系统则要求机器可读的约束。弥合这一差距仍然依赖人工完成。PolicyKG 关闭了这一循环。它是一个 LLM 流程，读取政策 PDF，分类每个句子为义务、许可或禁止，将标签提升至一阶义务逻辑，并输出 SHACL 约束。该流程在 LangGraph 状态机上运行，具有每个阶段的验证器。最重要的部分是语料适配器：一个 YAML 词汇注册表，将 LLM 谓词与目标本体相结合。重新定向到新领域意味着更换注册表，而不是重新训练模型。在亚洲科技大学政策与程序语料库（1,663 个句子，443 条规则）上，PolicyKG 达到了 86.9% 的义务分类准确率（Cohen's kappa = 0.709）。三位注释者独立重新标记了 50 项样本，并在 Fleiss' kappa = 0.844 上达成一致。在 69 个形状子集上的 SHACL 形状正确性为 F1 = 0.866。FOL 路径处理了 79.2% 的规则；其余通过直接的自然语言到 SHACL 的回退处理。我们审核了所有 443 条规则的二阶或更高阶构造。自动化的正则表达式检查表未标记任何问题，第一作者对 92 个 FOL 回退案例的检查确认了相同结果。真实 HOL 率的确切上限 95% Clopper-Pearson 界限为 0.67%。这是针对一个语料库的审核结果，而不是对机构政策的一阶逻辑充分性的证明。将 AIT 注册表更换为 GDPR 注册表使得精确属性对齐从 1/15 提升至 11/15（Fisher's exact p < 0.001；Cohen's h = 1.53）。在 LexDeMod 租赁合同基准测试（N = 200）上，宏 F1 降至 0.370，因为租赁英语使用“shall be entitled”表示许可——这正是词汇不匹配注册表更换所要解决的问题。重复运行产生哈希完全相同的 SHACL 输出。
cs.AI / 155 / 2608.09042
DualCert: A Solver for the Traveling Salesman Problem with Constraint-Coupled Learning
DualCert：一种解决约束耦合学习的旅行商问题的求解器
Abstract
Large traveling salesman problem (TSP) instances require a solver to allocate limited computation while preserving the validity of its outputs. Existing neural--operations-research (OR) hybrids predict guidance without requiring learned transitions to satisfy constraints discovered during search. DualCert introduces \emph{constraint-coupled learning}, in which current degree equations and dynamically separated subtour-elimination constraints (SECs) define each learned transition. At each refinement, the degree equations and selected, strictly satisfied SEC equations, with positive slacks, define an iterate-dependent primal-slack Karush--Kuhn--Tucker (KKT) manifold. Repaired dual variables and violated SEC rows define a local cost field. An exact constrained mirror-descent step maps each finite state to a positive state on the same manifold. Where selected rows and deterministic ties remain fixed, implicit differentiation maps parameter perturbations into the manifold tangent space and reuses the forward constraint operator for the local-cost-field derivative. The terminal edge state allocates computation across Held--Karp ascent, candidate-graph edge tests, and tour construction under a fixed budget. Deterministic verification recomputes original costs and accepts only verified candidate-graph lower bounds and edge decisions. On 1,000 held-out TSP1000 instances, DualCert attains a mean tour-cost gap of \(0.0573\%\) from Lin--Kernighan--Helsgaun version 3 (LKH-3) reference tours in \(9.55\) batch-amortized seconds per instance. It returns a verified candidate-graph lower bound for every instance and achieves \(81.46\%\) edge-decision coverage. The mean gap is \(67.1\%\) smaller than the reported NeuroLKH mean gap. Thus, optimization constraints govern learning, while deterministic verification preserves output validity.
Chinese Translation
大型旅行商问题（TSP）实例需要求解器在分配有限计算资源的同时保持输出的有效性。现有的神经-运筹学（OR）混合模型在搜索过程中无需学习转变来满足发现的约束条件，而是预测指导。DualCert 引入了 extit{约束耦合学习}，其中当前的度数方程和动态分离的子巡回消除约束（SEC）定义了每个学习的转变。在每次细化中，度数方程和选定的、严格满足的 SEC 方程（具有正松弛）定义了一个依赖于迭代的原始松弛 Karush-Kuhn-Tucker（KKT）流形。修复的对偶变量和违反的 SEC 行定义了一个局部成本场。一个精确的约束镜像下降步骤将每个有限状态映射到同一流形上的正状态。在选定的行和确定性平局保持不变的情况下，隐式微分将参数扰动映射到流形切空间，并重用前向约束算子来计算局部成本场的导数。终端边状态在固定预算下分配计算资源于 Held-Karp 上升、候选图边测试和巡回构建。确定性验证重新计算原始成本，仅接受经过验证的候选图下界和边决策。在 1,000 个保留的 TSP1000 实例中，DualCert 在每个实例中以平均 9.55 批量摊销秒的时间达到了与 Lin-Kernighan-Helsgaun 版本 3（LKH-3）参考巡回的平均巡回成本差距为 0.0573%。它为每个实例返回一个经过验证的候选图下界，并实现了 81.46% 的边决策覆盖。平均差距比报告的 NeuroLKH 平均差距小 67.1%。因此，优化约束支配学习，而确定性验证保持输出的有效性。
cs.AI / 156 / 2608.09088
A Multi-Scale Temporal Framework with Dynamic Fusion for EEG-Based Emotion Recognition
基于EEG的情感识别的多尺度时间框架与动态融合
Abstract
Mixed emotions represent a clinically relevant but still underexplored target for automatic emotion recognition. EEG provides millisecond-level access to neural activity, yet most EEG pipelines analyze the signal through a single temporal window, thereby fixing the temporal structure available to the model. This study introduces a multi-scale temporal framework for EEG-based emotion recognition. The EEG waveform is decomposed into windows of one or several durations, processed by a shared attention-based encoder, and integrated through a dynamic fusion module that assigns sample-specific weights across temporal scales. The framework is evaluated under a subject-independent protocol in binary and three-class settings, with the three-class task including the mixed affective category. The best results are 65.22% for the two-class task and 45.43% for the three-class task. Both are obtained with three-scale dynamic-fusion configurations and remain substantially above the full-signal baseline. The best-performing temporal scales differ between the two tasks. Dynamic fusion outperforms concatenation in the highest-scoring two-class configuration and slightly exceeds it in the highest-scoring three-class configuration, although these multi-scale settings require substantially more computation than the full-signal baseline.
Chinese Translation
混合情感代表了一种临床相关但仍未充分探索的自动情感识别目标。EEG提供了对神经活动的毫秒级访问，但大多数EEG处理流程通过单一时间窗口分析信号，从而固定了模型可用的时间结构。本研究提出了一种基于EEG的情感识别的多尺度时间框架。EEG波形被分解为一个或多个持续时间的窗口，通过共享的基于注意力的编码器进行处理，并通过动态融合模块进行整合，该模块在时间尺度上分配样本特定的权重。该框架在一个独立于受试者的协议下进行评估，涵盖二分类和三分类设置，其中三分类任务包括混合情感类别。最佳结果为二分类任务的65.22%和三分类任务的45.43%。这两个结果均在三尺度动态融合配置下获得，并且显著高于全信号基线。最佳表现的时间尺度在两个任务之间有所不同。在得分最高的二分类配置中，动态融合优于连接，而在得分最高的三分类配置中略微超过连接，尽管这些多尺度设置所需的计算量显著高于全信号基线。
cs.AI / 157 / 2608.09095
Who Bridges Safety? Identifying and Targeting Cross-Lingual Shared Safety Pathways
谁在架起安全的桥梁？识别和针对跨语言共享安全路径
Abstract
Uncovering the internal mechanisms underlying the safety capabilities of large language models (LLMs) is crucial for developing trustworthy artificial intelligence. Currently, mechanistic interpretability studies on multilingual safety are largely confined to local components, such as isolated neurons. However, this static and fragmented perspective overlooks the synergy among components and fails to elucidate how safety signals dynamically propagate within the model to drive safety decisions ultimately. In this work, we move beyond isolated neurons to identify and target the cross-layer functional pathways formed during safety signal propagation, thereby uncovering the mechanisms driving the cross-lingual safety gap. Specifically, we first identify monolingual safety pathways and validate their impact on refusing harmful requests. Subsequent cross-lingual analyses reveal a sparse subset of cross-lingual shared safety pathways, confirming that this intersection acts as the internal bridge transferring safety capabilities from high-resource (HR) languages to non-high-resource (NHR) languages. Building on these mechanistic findings, we propose a pathways-targeted alignment method based on the cross-lingual shared safety pathways. Experimental results show that updating only a small fraction of pathway parameters significantly improves safety in NHR languages while largely preserving the model's general capabilities.
Chinese Translation
揭示大型语言模型（LLMs）安全能力背后的内部机制对于开发可信赖的人工智能至关重要。目前，关于多语言安全的机制可解释性研究主要局限于局部组件，例如孤立的神经元。然而，这种静态和碎片化的视角忽视了组件之间的协同作用，未能阐明安全信号在模型内部如何动态传播，从而驱动最终的安全决策。在本研究中，我们超越孤立神经元，识别并针对在安全信号传播过程中形成的跨层功能路径，从而揭示导致跨语言安全差距的机制。具体而言，我们首先识别单语安全路径并验证其对拒绝有害请求的影响。随后的跨语言分析揭示了一组稀疏的跨语言共享安全路径，确认这一交集作为内部桥梁，将安全能力从高资源（HR）语言转移到非高资源（NHR）语言。基于这些机制发现，我们提出了一种基于跨语言共享安全路径的路径目标对齐方法。实验结果表明，仅更新一小部分路径参数即可显著提高NHR语言的安全性，同时在很大程度上保持模型的整体能力。
cs.AI / 158 / 2608.09109
Different Feedback, Different Updates: Selective Self-Learning from User Interactions for Large Language Models
不同反馈，不同更新：基于用户交互的选择性自学习框架用于大型语言模型
Abstract
User feedback offers natural supervision for persistent LLM improvement, but a single message may support multiple behavioral changes with different scopes of generalization. We introduce SLIFT, a selective self-learning framework built on a task-relative view of user feedback. SLIFT decomposes each feedback message into atomic components and interprets each component relative to the original task as Fix, Spec, or Null: requirements for task validity, compatible condition-specific refinements, or content with no reliable positive update direction. To incorporate each change at the appropriate scope, SLIFT trains two complementary LoRA adapters on a shared frozen backbone: a Generalist that consolidates Fix requirements into default behavior through feedback-conditioned self-distillation, and a Specialist that observes only the task and Generalist response to supply residual guidance for applicable, unmet Spec refinements. Null components induce no positive update. Across backbones, SLIFT achieves strong performance on both MemoryBench and WildFB, with targeted analyses further examining its underlying mechanisms. We release our code at https://anonymous.4open.science/r/SLIFT.
Chinese Translation
用户反馈为持续改进大型语言模型（LLM）提供了自然的监督，但单条消息可能支持多种行为变化，且具有不同的泛化范围。我们提出了SLIFT，一个基于用户反馈任务相关视角的选择性自学习框架。SLIFT将每条反馈消息分解为原子组件，并根据原始任务将每个组件解释为Fix、Spec或Null：任务有效性的要求、兼容的条件特定细化，或内容没有可靠的正向更新方向。为了在适当的范围内纳入每一变化，SLIFT在共享的冻结主干上训练两个互补的LoRA适配器：一个通用适配器（Generalist），通过反馈条件下的自蒸馏将Fix要求整合为默认行为；一个专用适配器（Specialist），仅观察任务和通用适配器的响应，以提供适用的、未满足的Spec细化的残余指导。Null组件不会引入正向更新。在不同主干上，SLIFT在MemoryBench和WildFB上均取得了良好的性能，针对性分析进一步探讨了其潜在机制。我们的代码已发布在 https://anonymous.4open.science/r/SLIFT。
cs.AI / 159 / 2608.09111
RAVEN-Eval: Rubric-Guided Automatic Evaluation for AI Video Generation Models Based on LMM Preference Judgement
RAVEN-Eval：基于LMM偏好判断的AI视频生成模型的 rubric 引导自动评估
Abstract
AI video generation has advanced rapidly and entered widespread commercial use. As a result, quality differences among videos produced by state-of-the-art AI video generation models~(AIVGMs) have become increasingly difficult to discern using conventional evaluation criteria, such as visual fidelity and semantic instruction following. Meanwhile, human evaluation now requires more expertise and sustained attention, substantially increasing annotation costs. This calls for automated evaluation that can reliably distinguish fine-grained differences among advanced AIVGMs with minimal human intervention. To address this challenge, we present RAVEN-Eval, a rubric-guided automated evaluation framework for AIVGMs, built primarily on the LMM-as-a-judge paradigm. Through an automatic task curation and quality-filtering pipeline, RAVEN-Eval curates 150 text-to-video~(T2V) tasks and 100 image-to-video~(I2V) tasks, and systematically collects more than 4,500 AIGVs. At its core, RAVEN-Eval adopts rubric-guided automated LMM preference judgement, in which LMM judges conduct pairwise comparisons according to task-specific rubrics. It further introduces an anchor-based model insertion approach to reduce the evaluation cost of incorporating new models. Finally, we evaluate 20 high-performance AIVGMs, as well as the judging capabilities of 13 LMM judges, and establish the RAVEN-Eval Leaderboards. Overall, RAVEN-Eval paves a scalable path for automatic and trustworthy evaluation of rapidly evolving AIVGMs.
Chinese Translation
AI视频生成技术迅速发展并广泛应用于商业领域。因此，先进的AI视频生成模型（AIVGMs）所生成视频之间的质量差异，使用传统评估标准（如视觉真实度和语义指令遵循）变得越来越难以辨别。同时，人类评估需要更多的专业知识和持续的注意力，显著增加了标注成本。这就需要一种能够在最小人类干预下，可靠地区分先进AIVGMs之间细微差异的自动评估方法。为了解决这一挑战，我们提出了RAVEN-Eval，一个针对AIVGMs的rubric引导自动评估框架，主要基于LMM-as-a-judge范式。通过自动任务策划和质量过滤管道，RAVEN-Eval策划了150个文本到视频（T2V）任务和100个图像到视频（I2V）任务，并系统地收集了超过4500个AIGVs。RAVEN-Eval的核心采用了rubric引导的自动LMM偏好判断，其中LMM评审根据特定任务的rubric进行成对比较。它进一步引入了一种基于锚点的模型插入方法，以降低引入新模型的评估成本。最后，我们评估了20个高性能的AIVGMs，以及13个LMM评审的判断能力，并建立了RAVEN-Eval排行榜。总体而言，RAVEN-Eval为快速发展的AIVGMs提供了一条可扩展的自动和可信评估路径。
cs.AI / 160 / 2608.09119
Motif 3: Technical Report
Motif 3：技术报告
Lim, Junghwan, Chung, Joon Son, Lee, Sungmin, Cheung, Wai Ting, Cho, Gihun, Ha, Minsu, Kang, Sangho, Kim, Beomgyu, Kim, Dongseok, Kim, Jangwoong, Kim, Taehyun, Kim, Taewhan, Lee, Jeesoo, Lee, Jeongdoo, Lee, Junhyeok, Oh, Dongpin, Cho, Hyeyeon, Choi, Dahye, Her, Jaeheui, Jung, Hanbin, Kang, Changjin, Kim, Minjae, Kim, Youngrok, Kweon, Hyukjin, Lee, Hongjoo, Park, Yeongjae, Ryu, Bokki
Abstract
We introduce Motif 3, a decoder-only Mixture-of-Experts language model with 314 billion total parameters and 13.2 billion activated per token. Each sparse MoE layer contains 384 routed experts, with eight selected per token. This fine-grained sparsity provides substantial expert capacity while limiting computation. Motif 3 is built around Grouped Differential Latent Attention (GDLA), which integrates grouped differential attention with the compressed key-value representation of Multi-head Latent Attention. The architecture further incorporates modified manifold-constrained hyper-connections, Expert Specific PolyNorm activations, and multi-token prediction to improve optimization stability, expert specialization, and inference efficiency. We pretrain Motif 3 on approximately 12.5 trillion tokens spanning web documents, STEM, code, mathematics, multilingual content, and domain-specialized corpora. Expert-balancing and numerical-stabilization techniques support stable training at scale, while selective MXFP8 computation and communication, memory-efficient fused kernels, and window-aware context parallelism enable training with context lengths up to 256K tokens. Our post-training pipeline combines general supervised fine-tuning, six specialist teachers trained with reinforcement learning, a software-engineering teacher trained with supervised fine-tuning, and Multi-teacher On-Policy Distillation. The resulting unified model consolidates complementary capabilities in reasoning, coding, tool use, professional work, long-context understanding, calibrated abstention, and instruction following. Across a broad evaluation suite, Motif 3 demonstrates competitive performance against leading open weight models, including strong results on long-horizon agentic tasks, mathematical reasoning, scientific knowledge, and hallucination-sensitive evaluation.
Chinese Translation
我们介绍了Motif 3，这是一种仅解码的混合专家语言模型，具有3140亿个总参数和每个token激活的132亿个参数。每个稀疏的MoE层包含384个路由专家，每个token选择8个。这种细粒度的稀疏性提供了可观的专家容量，同时限制了计算量。Motif 3围绕分组差异潜在注意力（Grouped Differential Latent Attention, GDLA）构建，该方法将分组差异注意力与多头潜在注意力的压缩键值表示相结合。该架构进一步结合了修改后的流形约束超连接、专家特定的PolyNorm激活和多token预测，以提高优化稳定性、专家专业化和推理效率。我们在大约12.5万亿个token上对Motif 3进行了预训练，这些token涵盖了网络文档、STEM、代码、数学、多语言内容和领域专门语料库。专家平衡和数值稳定技术支持大规模的稳定训练，而选择性MXFP8计算和通信、内存高效的融合内核以及窗口感知的上下文并行性使得训练能够处理长达256K个token的上下文长度。我们的后训练管道结合了通用的监督微调、六位通过强化学习训练的专业教师、通过监督微调训练的软件工程教师，以及多教师在线政策蒸馏。最终得到的统一模型整合了推理、编码、工具使用、专业工作、长上下文理解、校准的放弃和遵循指令等互补能力。在广泛的评估套件中，Motif 3在与领先的开放权重模型的比较中表现出竞争力，包括在长时间代理任务、数学推理、科学知识和对幻觉敏感的评估中取得的良好结果。
cs.AI / 161 / 2608.09121
MELLON - Multimodal Enhanced LLM for Online Navigation
MELLON - 多模态增强的在线导航大语言模型
Abstract
Web navigation agents are capable of addressing various types of tasks on different websites. Current baselines on web navigation are either unimodal or lack strong reasoning abilities given multimodal inputs. Focusing on the WebShop benchmark, a real-world website simulation, we explore the alignment of text and images, as well as multimodal reasoning and planning abilities, to enhance the performance of web navigation agents. We propose three innovative multimodal enhancements: Multimodal Enhanced LLM for Online Navigation (MELLON), VQAgent, and Multimodal Ranker. MELLON demonstrates a significant improvement in task completion accuracy, with a 9.26% increase after just one epoch of training. Our findings suggest the necessity of further exploration into multimodal approaches, with a focus on more extensive training and alignment strategies to enhance the effectiveness of web navigation agents.
Chinese Translation
网页导航代理能够处理不同网站上的各种任务。目前的网页导航基准要么是单模态的，要么在面对多模态输入时缺乏强大的推理能力。我们聚焦于WebShop基准，这是一个真实世界的网站模拟，探索文本与图像的对齐以及多模态推理和规划能力，以提升网页导航代理的性能。我们提出了三种创新的多模态增强方法：多模态增强的在线导航大语言模型（MELLON）、VQAgent和多模态排序器。MELLON在任务完成准确性上表现出显著提升，经过一次训练周期后提高了9.26%。我们的研究结果表明，有必要进一步探索多模态方法，重点关注更广泛的训练和对齐策略，以增强网页导航代理的有效性。
cs.AI / 162 / 2608.09123
RISE-RL: Rubric-Informed Selective Exploration for Open-Ended Reinforcement Learning
RISE-RL：基于评分标准的选择性探索用于开放式强化学习
Abstract
Aligning Large Language Models (LLMs) for open-ended tasks is challenging because responses must satisfy multidimensional criteria without following a single correct generation trajectory. Existing rubric-based reinforcement learning (RL) methods compress fine-grained criterion-level feedback into scalar rewards, making persistent capability gaps difficult to target under limited on-policy exploration. We propose $\textbf{RISE-RL}$ (Rubric-Informed Selective Exploration), which uses repeatedly missed rubric criteria to elicit privileged trajectories that are difficult to discover through unguided exploration alone. RISE-RL retains only trajectories whose complete-rubric reward exceeds the mean reward of natural rollouts, and then re-evaluates them under the original prompt to emphasize behaviors that remain weakly supported by the natural policy. The resulting guidance signal is optimized through a separate auxiliary objective and removed once its additional benefit diminishes. Experiments with 4B and 14B models across writing, chat, health, and science show that RISE-RL achieves the highest mean score on every evaluated benchmark under guidance-free evaluation. Compared with standard Rubric-RL, it improves the average score by 1.3 points at the 4B scale and $\textbf{3.3 points at the 14B scale}$, including a $\textbf{6.0-point}$ gain on CreativeWriting-V3. It also improves creative-writing diversity and yields gains on objectively scored medical and scientific benchmarks. These results indicate that selective internalization through reward filtering and policy support shaping is effective for open-ended reinforcement learning.
Chinese Translation
对大型语言模型（LLMs）进行开放式任务的对齐具有挑战性，因为响应必须满足多维标准，而不遵循单一正确的生成轨迹。现有的基于评分标准的强化学习（RL）方法将细粒度的标准级反馈压缩为标量奖励，这使得在有限的策略探索下，持续的能力差距难以针对。我们提出了$ extbf{RISE-RL}$（基于评分标准的选择性探索），该方法利用反复未满足的评分标准来引导出难以通过无指导探索发现的特权轨迹。RISE-RL仅保留那些完整评分奖励超过自然生成平均奖励的轨迹，然后在原始提示下重新评估这些轨迹，以强调在自然策略下仍然支持较弱的行为。最终的指导信号通过一个单独的辅助目标进行优化，并在其附加收益减小后移除。针对4B和14B模型在写作、聊天、健康和科学领域的实验表明，RISE-RL在无指导评估下在每个评估基准上都达到了最高的平均分数。与标准的Rubric-RL相比，在4B规模上平均分提高了1.3分，而在$ extbf{14B规模}$上提高了$ extbf{3.3分}$，其中在CreativeWriting-V3上获得了$ extbf{6.0分}$的提升。它还提高了创意写作的多样性，并在客观评分的医学和科学基准上取得了进展。这些结果表明，通过奖励过滤和策略支持塑造进行选择性内化对于开放式强化学习是有效的。
cs.AI / 163 / 2608.09124
ChronoState: Hidden Elapsed-Time Conditioning for Temporal-State Action Selection in Frozen-Backbone Language Models
ChronoState：冻结骨干语言模型中用于时间状态动作选择的隐藏经过时间条件
Abstract
Temporal decisions in language-model systems often depend on both symbolic task state and elapsed wall-clock time, such as cache expiration, job completion, quota resets, deadlines, or stale sessions. We study whether elapsed time can be supplied as a non-token, system-side scalar and composed with visible symbolic state by a frozen-backbone language model. We introduce ChronoState, a compositional temporal-state benchmark in which symbolic state appears in the prompt, elapsed seconds tau are supplied through a hidden chronometric-injection channel, and the model selects a forced-choice temporal action. Here, "hidden" means hidden from the user-visible token sequence, not from model computation. Using Qwen2.5-3B-Instruct as a frozen bf16 backbone with a 31-dimensional sinusoidal-plus-log time encoding, gated FiLM residual modulation, and a rank-8 LoRA action surface, hidden-time CI reaches 0.9305 +/- 0.0134 accuracy and 0.9410 +/- 0.0103 balanced accuracy. No-time and shuffled-time controls fall to 0.5511 +/- 0.0042 and 0.3323 +/- 0.0097, respectively, with high shuffled-time wrong-state consistency supporting causal dependence on the injected scalar within the trained distribution. Generalization remains strong for held-out templates, durations, and multi-constraint compositions, but held-out quota-family transfer is weak at 0.5065 +/- 0.0559, while a fair prompt+LoRA timestamp baseline reaches 0.9893 +/- 0.0052. Thus, ChronoState supports a narrow conclusion: hidden elapsed time can be composed with symbolic task state under direct supervision, but does not establish autonomous time tracking, broad unseen-family abstraction, or superiority over prompt-injected timestamps.
Chinese Translation
语言模型系统中的时间决策通常依赖于符号任务状态和经过的墙钟时间，例如缓存过期、作业完成、配额重置、截止日期或过期会话。我们研究了是否可以将经过的时间作为非标记的系统侧标量提供，并与冻结骨干语言模型的可见符号状态组合。我们引入了ChronoState，这是一个组合的时间状态基准，其中符号状态出现在提示中，经过的秒数tau通过一个隐藏的计时注入通道提供，模型选择一个强制选择的时间动作。在这里，“隐藏”是指对用户可见的标记序列隐藏，而不是对模型计算的隐藏。使用Qwen2.5-3B-Instruct作为冻结的bf16骨干，结合31维的正弦加对数时间编码、门控FiLM残差调制和8级LoRA动作表面，隐藏时间CI达到了0.9305 +/- 0.0134的准确率和0.9410 +/- 0.0103的平衡准确率。无时间和随机时间控制分别降至0.5511 +/- 0.0042和0.3323 +/- 0.0097，高随机时间错误状态一致性支持了在训练分布内对注入标量的因果依赖。对于保留的模板、持续时间和多约束组合，泛化能力仍然强，但保留的配额家族转移较弱，仅为0.5065 +/- 0.0559，而公平的提示+LoRA时间戳基线达到了0.9893 +/- 0.0052。因此，ChronoState支持一个狭窄的结论：隐藏的经过时间可以在直接监督下与符号任务状态组合，但并不建立自主时间跟踪、广泛的未见家族抽象或优于提示注入时间戳的能力。
cs.AI / 164 / 2608.09153
TRACE: TRajectory Attribution for Automated Context Engineering
TRACE：用于自动化上下文工程的轨迹归因
Abstract
Production AI agents fail when their context sources -- system prompts, knowledge bases, tool descriptions, and procedural skills -- contain errors or gaps. Current maintenance relies on manual log review and ad-hoc debugging, creating a scalability bottleneck as interaction volume grows. We present TRACE (TRajectory Attribution for Automated Context Engineering), an automated feedback loop that mines historical agent trajectories to diagnose and remediate context failures. Our key insight is that trajectories are rich with implicit dissatisfaction signals -- user corrections, rephrasing, abandonment cues -- that reveal precisely where context sources failed, without explicit feedback collection. Unlike model fine-tuning, TRACE operates on the context layer, enabling rapid iteration without retraining. We make four contributions: (1) a trajectory mining framework that systematically extracts diagnostic information from historical agent executions; (2) multi-component causal attribution that extends textual gradients from monolithic prompt optimization to heterogeneous context sources (skills, knowledge bases, tools, prompts); (3) exploratory verification, where agents actively read context sources to distinguish content gaps requiring CREATE from stale content requiring UPDATE, achieving 96% operation accuracy; and (4) a reusable simulation methodology and verifiable benchmark addressing the absence of open datasets for context debugging, with a six-category fault taxonomy, ground truth annotations, and a cross-layer verification protocol. On 60 dissatisfaction traces spanning three complexity tiers (up to 16 execution nodes), TRACE achieves 72.7% root cause attribution and 82% end-to-end fix effectiveness, showing that over 80% of context-layer failures can be automatically diagnosed and remediated by mining historical trajectories, an overlooked resource in production systems.
Chinese Translation
当生产AI代理的上下文源——系统提示、知识库、工具描述和程序技能——存在错误或缺口时，它们会失败。目前的维护依赖于手动日志审查和临时调试，随着交互量的增加，这造成了可扩展性的瓶颈。我们提出了TRACE（用于自动化上下文工程的轨迹归因），这是一个自动反馈循环，通过挖掘历史代理轨迹来诊断和修复上下文失败。我们的关键见解是，轨迹中蕴含着丰富的隐性不满信号——用户纠正、重新措辞、放弃提示——这些信号精确揭示了上下文源失败的具体位置，而无需显式的反馈收集。与模型微调不同，TRACE在上下文层面上运行，能够快速迭代而无需重新训练。我们做出了四项贡献：（1）一个轨迹挖掘框架，系统地从历史代理执行中提取诊断信息；（2）多组件因果归因，将单一提示优化的文本梯度扩展到异构上下文源（技能、知识库、工具、提示）；（3）探索性验证，代理主动读取上下文源，以区分需要CREATE的内容缺口和需要UPDATE的过时内容，达到96%的操作准确率；（4）一个可重用的模拟方法论和可验证的基准，解决上下文调试缺乏开放数据集的问题，包含六类故障分类法、真实标注和跨层验证协议。在涵盖三个复杂度层级（最多16个执行节点）的60个不满轨迹上，TRACE实现了72.7%的根本原因归因和82%的端到端修复效果，表明超过80%的上下文层失败可以通过挖掘历史轨迹自动诊断和修复，这是生产系统中被忽视的资源。
cs.AI / 165 / 2608.09164
CIDER: A Dataset of Contextual Disclosure Boundaries for Privacy Preference Alignment
CIDER：一个用于隐私偏好对齐的上下文披露边界数据集
Abstract
Aligning large language models (LLMs) with human privacy preferences requires capturing individuals' disclosure boundaries beyond general privacy norms. However, a gap remains in eliciting such nuanced preferences to evaluate alignment in realistic settings. We introduce CIDER, a dataset of 14,850 human annotations from 169 users, forming 1,650 contextual disclosure boundary sets across 60 interpersonal communication scenarios involving information sharing that violates privacy norms. Each boundary represents a real user's disclosure decisions over 9 sharing variants in a scenario, for a given communication role and AI-mediated condition. We formulate a task in which models predict a user's disclosure decision from historical boundaries, with varying levels of contextual information. Across 12 open and proprietary models, in-context personalization improves prediction accuracy by up to 11.41 percentage points using only 6 historical examples. Larger models such as GPT-5.4 (with medium reasoning effort) and Claude Sonnet 4.6 are better at leveraging semantic context to understand user-specific, context-dependent disclosure preferences for more accurate predictions, while smaller models tend to rely on structured heuristics based on disclosure granularity and identifiability. Personalization generally improves prediction accuracy, but the improvement is often accompanied by imbalanced shifts in false-positive and false-negative rates across models, with only Claude Sonnet 4.6 achieving balanced improvements in both. Our findings reveal both the promise and limitations of inference-time personalization for privacy preference modeling and position CIDER as a resource for advancing personalized privacy alignment.
Chinese Translation
将大型语言模型（LLMs）与人类隐私偏好对齐需要捕捉个体的披露边界，超越一般的隐私规范。然而，在引导这种细致的偏好以评估在现实环境中的对齐方面仍存在差距。我们介绍了CIDER，这是一个包含来自169名用户的14,850条人类注释的数据集，形成了在60种涉及违反隐私规范的信息共享的人际沟通场景中的1,650个上下文披露边界集。每个边界代表一个真实用户在特定沟通角色和AI介导条件下对9种共享变体的披露决策。我们制定了一个任务，模型需根据历史边界预测用户的披露决策，使用不同级别的上下文信息。在12个开放和专有模型中，仅使用6个历史示例的情况下，上下文个性化将预测准确率提高了最多11.41个百分点。较大的模型如GPT-5.4（中等推理努力）和Claude Sonnet 4.6在利用语义上下文理解用户特定的、依赖上下文的披露偏好以获得更准确的预测方面表现更佳，而较小的模型则倾向于依赖基于披露粒度和可识别性的结构化启发式方法。个性化通常提高了预测准确率，但这种提升往往伴随着模型之间假阳性和假阴性率的不平衡变化，只有Claude Sonnet 4.6在两者上都实现了平衡的改善。我们的研究结果揭示了推理时个性化在隐私偏好建模中的潜力与局限性，并将CIDER定位为推动个性化隐私对齐的资源。
cs.AI / 166 / 2608.09168
From Relevance to Execution Utility: Reward-Aware Dynamic Execution Gating for Skill-Based LLM Agents
从相关性到执行效用：基于奖励的动态执行门控用于技能基础的LLM代理
Abstract
Agent skills are increasingly used to equip large language model (LLM) agents with reusable procedural knowledge. Although recent work has substantially improved skill retrieval due to the increasing skill libraries, retrieving a plausible skill bundle does not guarantee that executing it is worthwhile. Since every skill-conditioned rollout is computationally expensive, deciding whether a retrieved bundle should be executed has become an increasingly important challenge. To this end, we introduce the Reward-Aware Dynamic Execution Gate (RADEG), a lightweight, retriever-agnostic decision layer between skill retrieval and agent execution. RADEG learns a low-cost surrogate model that predicts the execution utility of a query--bundle pair before the expensive rollout is launched. To obtain informative supervision while controlling for task difficulty, we locally perturb each retrieved bundle by deleting, adding, or replacing one skill, producing matched same-query rollouts that isolate the effect of bundle composition on verifier reward. During deployment, RADEG updates only a warm-started logistic head as new verifier feedback becomes available, enabling inexpensive adaptation of the execute/skip boundary without retraining either the retriever or the agent. Under a query-level held-out evaluation on 288 collected rollouts, RADEG substantially reduces unnecessary agent executions while preserving a large fraction of the downstream verifier reward. It consistently outperforms relevance-based and random gating across different execution budgets, demonstrating that execution-aware surrogate modeling provides a practical and cost-effective complement to skill retrieval.
Chinese Translation
代理技能越来越多地被用于为大型语言模型（LLM）代理提供可重用的程序知识。尽管近期的研究由于技能库的增加显著改善了技能检索，但检索到一个合理的技能组合并不保证其执行是有价值的。由于每个技能条件的展开计算成本高昂，决定是否执行检索到的组合已成为一个日益重要的挑战。为此，我们提出了奖励感知动态执行门控（RADEG），这是一个轻量级、与检索器无关的决策层，位于技能检索与代理执行之间。RADEG学习一个低成本的替代模型，在启动高成本的展开之前，预测查询-组合对的执行效用。为了在控制任务难度的同时获得有意义的监督，我们通过删除、添加或替换一个技能，局部扰动每个检索到的组合，生成匹配的同查询展开，以隔离组合构成对验证者奖励的影响。在部署过程中，RADEG仅在新的验证者反馈可用时更新一个预热的逻辑头，从而实现执行/跳过边界的低成本适应，而无需重新训练检索器或代理。在288个收集的展开的查询级保留评估中，RADEG显著减少了不必要的代理执行，同时保留了大量的下游验证者奖励。它在不同的执行预算下始终优于基于相关性和随机门控，证明了执行感知的替代建模为技能检索提供了一个实用且具有成本效益的补充。
cs.AI / 167 / 2608.09184
Agentic Router: An Execution-Grounded Continual Learning Approach With Memory
代理路由器：一种基于执行的持续学习方法与记忆
Abstract
Large language model (LLM) agents provide a promising interface for command-line-based network operations, but a plausible command may still fail or introduce operational risk after execution. Existing approaches mainly focus on command generation or final configuration correctness, and do not use execution-grounded experience to jointly improve candidate coverage and action selection. We propose an execution-grounded dual-path consequence-aware agent for CLI-based SONiC operations, which generates multiple complete actions, predicts their execution consequences, and selects the final action through utility- and risk-aware reranking. The proposal-side path abstracts reusable operational lessons into retrievable guidance to improve feasible-action coverage without modifying the proposal LLM, while the selection-side path adapts the consequence predictor through session-level LoRA updates using real SSH feedback to improve conditional selection quality. Experiments over multi-turn SONiC operation sessions with different Qwen3 proposal models show that the framework improves feasible-action coverage and top-1 execution success, and that the two adaptation paths provide complementary gains over interaction.
Chinese Translation
大型语言模型（LLM）代理为基于命令行的网络操作提供了一个有前景的接口，但在执行后，合理的命令仍可能失败或引入操作风险。现有方法主要关注命令生成或最终配置的正确性，而未利用基于执行的经验共同改善候选覆盖率和行动选择。我们提出了一种基于执行的双路径后果感知代理，用于基于CLI的SONiC操作，该代理生成多个完整的操作，预测其执行后果，并通过效用和风险感知的重新排序选择最终操作。提案侧路径将可重用的操作经验抽象为可检索的指导，以提高可行操作覆盖率，而无需修改提案LLM；选择侧路径通过使用真实SSH反馈的会话级LoRA更新来适应后果预测器，以提高条件选择质量。在不同Qwen3提案模型的多轮SONiC操作会话中的实验表明，该框架提高了可行操作的覆盖率和首个执行成功率，并且这两条适应路径在交互中提供了互补的收益。
cs.AI / 168 / 2608.09190
Structure-Preserving Uncertainty Propagation in First-Order Proof Search
保持结构的第一阶证明搜索中的不确定性传播
Abstract
GK is a query-directed first-order prover that extends ordinary resolution-based proof search with explicit positive and negative claims, numerical confidence values, and prioritized default rules with exceptions. It works directly with non-ground clauses, including equality and function terms. Candidate proofs are found by bounded first-order proof search; exception conditions of defaults are checked by further bounded searches, recursively when exceptions themselves depend on defaults. This avoids requiring a finite global grounding, while allowing incomplete searches to be reported as such. This paper adds structure-preserving quantitative reporting to that framework. Retained proof histories are used in two calculations. The first reconstructs the uncertain ground premises used by each proof and computes the probability that at least one retained proof is available, without counting shared premises independently. The second resolves positive and negative support at intermediate atoms before that support is propagated through later rules; the same calculation evaluates uncertain exception conditions for individual rule applications. Reports separate positive support, negative support, conflict, and ignorance and identify detected incomplete calculations or fallbacks. The implementation performs bounded reconstruction and dependency traversal after proof search and still requires no global grounding. Analytic examples and independent simulators reproduce the reference calculations on their stated fragments. Comparisons with probabilistic logic, probabilistic ASP, default logic, and goal-directed ASP identify cases of agreement, semantic difference, unsupported translation, and incomplete computation.
Chinese Translation
GK 是一种查询导向的第一阶证明器，它通过显式的正负声明、数值置信值和带有例外的优先默认规则扩展了普通的基于分辨率的证明搜索。它直接处理非基础子句，包括等式和函数项。候选证明通过有界的第一阶证明搜索找到；默认的例外条件通过进一步的有界搜索进行检查，当例外本身依赖于默认时，采用递归方式。这避免了对有限全局基础的要求，同时允许将不完整的搜索报告为不完整。本文在该框架中增加了保持结构的定量报告。保留的证明历史用于两个计算。第一个重构每个证明所使用的不确定基础前提，并计算至少有一个保留证明可用的概率，而不独立计算共享前提。第二个在支持传播通过后续规则之前解决中间原子的正负支持；同样的计算评估个别规则应用的不确定例外条件。报告分开正支持、负支持、冲突和无知，并识别检测到的不完整计算或回退。该实现执行有界重构和依赖遍历，在证明搜索后仍然不需要全局基础。分析示例和独立模拟器在其所述片段上重现参考计算。与概率逻辑、概率ASP、默认逻辑和目标导向ASP的比较识别了一致性、语义差异、不支持的翻译和不完整计算的情况。
cs.AI / 169 / 2608.09201
Signature-Guided Capacity Occupancy for Dense Expert Merging
基于签名的密集专家合并的容量占用
Abstract
Dense expert merging combines domain-specialized language models into one single checkpoint, typically by admitting task-vector support in weight space. However, this admission is governed by three decisions that existing methods answer only partially: where to open layer capacity from cross-expert conflict, who should occupy that capacity based on domain demand, and how to admit the resulting support without relying on costly recipe search. To tackle these issues, we propose SigMerge (Signature-Guided Capacity Occupancy), a structured capacity assignment framework for dense expert merging. Starting from a dense base merge, conflict signatures set each layer's capacity from cross-expert conflict, positive base-merge deficits set each domain's share of that capacity, and a sequential occupancy rule admits each expert delta up to the resulting layer-domain budget. Across 21 paired settings spanning seven dense base merges and three model pools, SigMerge improves every one (by 15.0% on average) and achieves the best average rank (1.67) among six merging methods, outperforming three categories of merging baselines.
Chinese Translation
密集专家合并将领域专用语言模型合并为一个单一的检查点，通常通过在权重空间中接受任务向量支持来实现。然而，这一过程受到三个决策的制约，而现有方法仅部分回答了这些问题：从跨专家冲突中打开层容量的位置、根据领域需求决定谁应占用该容量，以及如何在不依赖于昂贵的配方搜索的情况下接受结果支持。为了解决这些问题，我们提出了SigMerge（基于签名的容量占用），这是一个用于密集专家合并的结构化容量分配框架。从密集基础合并开始，冲突签名设置每层的容量以应对跨专家冲突，正的基础合并赤字设置每个领域的容量份额，而顺序占用规则则在结果层-领域预算内接受每个专家的增量。在涵盖七个密集基础合并和三个模型池的21个配对设置中，SigMerge在每个设置上均有所改进（平均提高15.0%），并在六种合并方法中实现了最佳平均排名（1.67），优于三类合并基线。
cs.AI / 170 / 2608.09202
CRUISE: Vision-Language Model-Guided Uncertainty-Aware Cross-Modal Sensor Fusion for Robust Autonomous Driving
CRUISE：基于视觉-语言模型引导的考虑不确定性的跨模态传感器融合框架，用于稳健的自动驾驶
Abstract
Modern autonomous vehicles are equipped with multiple sensors, such as cameras, LiDAR, and radar, for comprehensive environmental perception. However, robust cross-modal feature fusion remains a critical challenge, as the reliability of each sensor varies significantly across diverse real-world driving conditions, including poor visibility and adverse weather. While uncertainty quantification (UQ) mitigates this issue by allowing models to prioritize reliable signals, existing uncertainty-aware fusion methods typically rely on simple feature-level uncertainty estimates and thus often fail to generalize effectively in complex, out-of-distribution scenarios. To address this limitation, we propose CRUISE, a novel uncertainty-aware cross-modal sensor fusion framework. CRUISE integrates a vision-language model (VLM)-guided UQ module that generates fine-grained, pixel-level uncertainty estimates. By leveraging the VLM's rich prior knowledge and superior contextual reasoning, our approach provides a highly informative guide for the fusion process. Furthermore, we introduce a dynamic adaptive mechanism that explicitly models and captures cross-modal dependencies, ensuring the framework fully exploits the inherent complementary nature of multi-sensor inputs.
Chinese Translation
现代自动驾驶车辆配备了多种传感器，如摄像头、激光雷达（LiDAR）和雷达，以实现全面的环境感知。然而，稳健的跨模态特征融合仍然是一个关键挑战，因为在不同的现实驾驶条件下，包括能见度差和恶劣天气，各传感器的可靠性差异显著。尽管不确定性量化（UQ）通过允许模型优先考虑可靠信号来缓解这一问题，但现有的考虑不确定性的融合方法通常依赖于简单的特征级不确定性估计，因此在复杂的超出分布场景中往往无法有效泛化。为了解决这一局限性，我们提出了CRUISE，一个新颖的考虑不确定性的跨模态传感器融合框架。CRUISE集成了一个基于视觉-语言模型（VLM）引导的不确定性量化模块，该模块生成细粒度的像素级不确定性估计。通过利用VLM丰富的先验知识和卓越的上下文推理，我们的方法为融合过程提供了高度信息化的指导。此外，我们引入了一种动态自适应机制，明确建模和捕捉跨模态依赖关系，确保框架充分利用多传感器输入的固有互补特性。
cs.AI / 171 / 2608.09227
Omni2LoRA: Coherence-Preserving Parametric Memory for Efficient Omni Language Models
Omni2LoRA：高效全语言模型的保持一致性参数记忆
Abstract
Omnimodal language models (OLMs) enable unified audio-visual understanding, but processing long joint token sequences makes inference computationally prohibitive. While recent token compression methods attempt to alleviate this burden, compressing modalities in isolation often destroys the temporal cross-modal anchors necessary for coherent reasoning. We introduce Omni2LoRA, a two-stage framework for efficient parametric memory compression via coherence-preserving context distillation that bypasses the token bottleneck entirely. First, a Perceiver hypernetwork processes intermediate representations from a frozen OLM to encode the multimodal context into a full-rank Low-Rank Adaptation (LoRA) adapter in a single forward pass. To prevent the resulting parameter footprint from scaling linearly with recording length, we optimize a discrete rank allocation policy via Group Relative Policy Optimization (GRPO) that uses a modality-ablated counterfactual reward to explicitly penalize the loss of audio-visual coherence, forcing the model to allocate its fixed sub-linear rank budget to synergistic cross-modal anchors rather than isolated visual features. Across three omnimodal backbones, Omni2LoRA operating at a 30% rank budget outperforms direct full-context inference and strong token-compression baselines (OmniZip, OMAC, O-MARC) on four audio-visual question answering benchmarks, improving average accuracy by 8-12% over the strongest baseline and remaining stable under compression ratios as tight as 75%, where token-pruning methods degrade sharply. By converting multimodal memory into a fixed-budget, reusable parameter state, our method drives answer-time multimodal-token load to zero, cutting per-query Time to First Token (TTFT) by up to 12x relative to full-context inference and amortizing to under 0.5s after a handful of queries.
Chinese Translation
全模态语言模型（OLMs）实现了统一的音频-视觉理解，但处理长的联合标记序列使得推理在计算上变得不可行。虽然最近的标记压缩方法试图减轻这一负担，但孤立地压缩模态往往会破坏连贯推理所需的时间跨模态锚点。我们提出了Omni2LoRA，这是一种通过保持一致性的上下文蒸馏实现高效参数记忆压缩的两阶段框架，完全绕过了标记瓶颈。首先，一个感知器超网络处理来自冻结的OLM的中间表示，将多模态上下文编码为一个全秩低秩适配器（Low-Rank Adaptation, LoRA），只需一次前向传递。为了防止生成的参数占用空间随着录音长度线性扩展，我们通过群体相对策略优化（Group Relative Policy Optimization, GRPO）优化离散秩分配策略，该策略使用去模态化的反事实奖励来显式惩罚音频-视觉一致性的损失，迫使模型将其固定的次线性秩预算分配给协同的跨模态锚点，而不是孤立的视觉特征。在三个全模态骨干网络上，Omni2LoRA在30%的秩预算下的表现优于直接的全上下文推理和强大的标记压缩基线（OmniZip、OMAC、O-MARC），在四个音频-视觉问答基准上提高了8-12%的平均准确率，相较于最强基线保持稳定，即使在高达75%的压缩比下，标记修剪方法的性能急剧下降。通过将多模态记忆转换为固定预算、可重用的参数状态，我们的方法将答案时间的多模态标记负载降至零，相较于全上下文推理，将每查询的首次标记时间（Time to First Token, TTFT）缩短了最多12倍，并在少量查询后摊销至0.5秒以下。
cs.AI / 172 / 2608.09230
SafeSceneReason: A Multimodal Reasoning Benchmark Connecting Industrial Hazards with Accident Knowledge
SafeSceneReason：连接工业危害与事故知识的多模态推理基准
Abstract
Industrial-safety understanding requires more than detecting workers, equipment, and personal protective equipment. Models must also assess compliance, identify hazardous interactions, explain potential accident mechanisms, and recommend preventive actions. Existing safety datasets primarily focus on visual perception or isolated violation recognition and provide limited supervision for evidence-grounded reasoning. We introduce SafeSceneReason, a multimodal industrial-safety reasoning benchmark and companion training corpus that connects workplace scenes with knowledge from occupational accident investigations. SafeSceneReason combines two complementary data-construction pipelines. The scene-centric pipeline converts annotated workplace images into executable safety scene graphs and generates deterministic answers through program execution over objects, relations, and safety rules. The report-centric pipeline extracts figures and contextual evidence from accident reports and constructs multimodal questions using evidence graphs, explicit information boundaries, multi-step reasoning paths, and iterative verification. The resulting resource contains 110,581 verified scene-centric question--answer pairs and 13,114 refined report-centric question--answer pairs, covering perception, spatial and quantitative reasoning, compliance assessment, evidence synthesis, causal analysis, and mitigation-oriented decision making. Evaluation of representative proprietary and open-source vision--language models reveals substantial performance differences and persistent weaknesses in comparative, technical, and multi-evidence reasoning, demonstrating that strong general visual understanding does not yet guarantee reliable industrial-safety reasoning.
Chinese Translation
工业安全理解不仅需要检测工人、设备和个人防护装备，还必须评估合规性、识别危险互动、解释潜在事故机制，并推荐预防措施。现有的安全数据集主要集中于视觉感知或孤立的违规识别，且对基于证据的推理提供的监督有限。我们提出了SafeSceneReason，一个多模态工业安全推理基准及其配套训练语料库，旨在将工作场景与职业事故调查的知识相连接。SafeSceneReason结合了两个互补的数据构建管道。场景中心管道将标注的工作场所图像转换为可执行的安全场景图，并通过对对象、关系和安全规则的程序执行生成确定性的答案。报告中心管道从事故报告中提取数据和上下文证据，并使用证据图、显式信息边界、多步骤推理路径和迭代验证构建多模态问题。最终资源包含110,581对经过验证的场景中心问答对和13,114对精炼的报告中心问答对，涵盖感知、空间和定量推理、合规性评估、证据综合、因果分析以及以减轻风险为导向的决策。对代表性的专有和开源视觉-语言模型的评估揭示了显著的性能差异和在比较、技术及多证据推理中的持续弱点，表明强大的通用视觉理解尚未能保证可靠的工业安全推理。
cs.AI / 173 / 2608.09246
An Explainable GNN Framework for Component-Level Anomaly Diagnosis
一种可解释的图神经网络框架用于组件级异常诊断
Abstract
Industrial processes are complex systems composed of multiple interacting sensors that generate multivariate time series (MTS). Detecting anomalies in such systems is critical for reliability and safety, yet understanding their origin is equally important. Existing Graph Neural Network (GNN)based methods for anomaly detection primarily focus on sensor-level deviations and either attribute anomalies directly to the deviating sensors. When diagnosis is attempted, generally, the most deviated sensor is identified as a root cause of a system fault. However, in many industrial systems, anomalies do not arise from faulty sensors but from disruptions in the influences governing the system dynamics. We propose an explainable GNN-based anomaly detection framework that shifts the perspective from sensor-level anomalies to component-level diagnosis, hypothesizing that anomalous measurements are symptoms of altered inter-sensor influences. Experiments show that the method effectively identifies and prioritizes the true faulty components, providing interpretable insights into system failures.
Chinese Translation
工业过程是由多个相互作用的传感器组成的复杂系统，这些传感器生成多变量时间序列（MTS）。在此类系统中检测异常对于可靠性和安全性至关重要，但理解异常的来源同样重要。现有基于图神经网络（GNN）的方法主要集中在传感器级的偏差，并直接将异常归因于偏离的传感器。在进行诊断时，通常将偏差最大的传感器识别为系统故障的根本原因。然而，在许多工业系统中，异常并非源于故障传感器，而是由于影响系统动态的因素发生了干扰。我们提出了一种基于可解释的GNN的异常检测框架，将视角从传感器级异常转向组件级诊断，假设异常测量是传感器间影响变化的症状。实验表明，该方法有效识别并优先考虑真正故障的组件，为系统故障提供可解释的见解。
cs.AI / 174 / 2608.09248
Emotion2Skill: Model-Internal Emotion Signals for Adaptive Skill Selection and Evolution
Emotion2Skill：基于模型内部情感信号的自适应技能选择与演化
Abstract
Skill-based LLM agents select reusable procedures from an external library to solve complex tasks, yet their routing decisions rely entirely on text-level signals such as task descriptions, verbal reflections, and experience-derived rules, while the model's own internal representational state remains unobserved. Recent interpretability work has shown that LLMs maintain linear emotion representations that causally influence behavior; however, these representations have been exploited only for post-hoc analysis or direct output steering, and have not been used to inform agent-level decision-making. We propose Emotion2Skill, a framework that extracts LLM-internal emotion vectors and incorporates them into both skill selection and skill evolution. At each decision step, a 27-dimensional emotion state is extracted from the residual stream and mapped to a confidence-gated summary injected into the routing prompt. Beyond online selection, emotion trajectories are analyzed for abrupt internal-state shifts to pinpoint problematic skill invocations, guiding targeted SOP rewriting that replaces the coarse binary outcome signal of prior methods. On WebShop and ALFWorld, Emotion2Skill with Qwen3-8B improves over the Zero-Shot baseline by +26.9% success rate and +25.5% average success respectively, outperforming all baselines on both benchmarks with consistent gains on Qwen3-14B. Co-activation analysis further reveals semantically coherent emotion--skill pairings, confirming that the routing improvements reflect meaningful internal-state signals rather than opaque statistical correlations. These results establish LLM-internal emotion representations as an effective decision-level signal for orchestrating agent skill systems, extending their utility beyond interpretability and output steering. The code is available at https://github.com/BoHan-LIN04/Emotion2Skill.
Chinese Translation
基于技能的大型语言模型（LLM）代理从外部库中选择可重用的程序以解决复杂任务，但它们的路由决策完全依赖于文本级信号，如任务描述、语言反思和经验派生规则，而模型自身的内部表征状态则未被观察。近期的可解释性研究表明，LLM保持线性情感表征，这些表征对行为产生因果影响；然而，这些表征仅被用于事后分析或直接输出引导，而未被用于指导代理级决策。我们提出了Emotion2Skill，一个提取LLM内部情感向量并将其纳入技能选择和技能演化的框架。在每个决策步骤中，从残差流中提取27维情感状态，并将其映射到注入路由提示中的置信度门控摘要中。除了在线选择，情感轨迹还被分析以检测内部状态的突然变化，以确定有问题的技能调用，从而指导针对性的标准操作程序（SOP）重写，替代先前方法的粗糙二元结果信号。在WebShop和ALFWorld上，使用Qwen3-8B的Emotion2Skill在成功率上提高了26.9%，在平均成功率上提高了25.5%，在这两个基准上超越了所有基线，并在Qwen3-14B上实现了一致的增益。共激活分析进一步揭示了语义一致的情感-技能配对，确认路由改进反映了有意义的内部状态信号，而非模糊的统计相关性。这些结果确立了LLM内部情感表征作为协调代理技能系统的有效决策级信号，扩展了其在可解释性和输出引导之外的实用性。代码可在 https://github.com/BoHan-LIN04/Emotion2Skill 获取。
cs.AI / 175 / 2608.09253
SkillSentry: Reliable Skill Execution for LLM Agents via Runtime Assurance
SkillSentry：通过运行时保障实现 LLM 代理的可靠技能执行
Abstract
LLM agents are increasingly equipped with skills to perform complex tasks through multi-step reasoning and tool use. Although skills provide reusable procedural knowledge, agents may still execute them unreliably. Even when an agent has demonstrated the capability to complete tasks under the guidance of a skill, it may fail to do so consistently across similar tasks or repeated runs due to deviations from the skill procedure or incorrect execution of individual steps. Such instability limits the practical reliability of LLM agents. To address this problem, we propose SkillSentry, a skill-oriented runtime assurance framework built upon a new domain-specific language (DSL) for representing runtime guidance for skill execution. SkillSentry initializes the runtime guidance by combining a skill specification extracted from the corresponding skill document with execution experience mined from historical successful and failed traces. It then wraps around the agent execution loop to monitor and guide skill execution under the current guidance, while iteratively refining the guidance using newly collected traces. We evaluate SkillSentry on 15 skills across two LLM agents, each paired with two backbone models, i.e., Claude Code with Claude-Haiku-4.5 and Claude-Opus-4.6, and Codex with GPT-5.2 and GPT-5.4. Our results show that SkillSentry improves the task success rate of LLM agents by 24.1% across skills, on average, while exhibiting lower variability across repeated runs.
Chinese Translation
LLM 代理越来越多地配备技能，以通过多步骤推理和工具使用执行复杂任务。尽管技能提供了可重用的程序知识，但代理的执行仍可能不可靠。即使代理在技能指导下展示了完成任务的能力，由于偏离技能程序或个别步骤的错误执行，它在类似任务或重复运行中可能无法始终如一地完成任务。这种不稳定性限制了 LLM 代理的实际可靠性。为了解决这个问题，我们提出了 SkillSentry，一个以技能为导向的运行时保障框架，基于一种新的领域特定语言（DSL）来表示技能执行的运行时指导。SkillSentry 通过将从相应技能文档中提取的技能规范与从历史成功和失败的轨迹中挖掘的执行经验相结合，初始化运行时指导。它随后包裹在代理执行循环中，以监控和指导在当前指导下的技能执行，同时利用新收集的轨迹迭代地完善指导。我们在两个 LLM 代理上评估了 SkillSentry，分别配备了两个主干模型，即 Claude Code 与 Claude-Haiku-4.5 和 Claude-Opus-4.6，以及 Codex 与 GPT-5.2 和 GPT-5.4。我们的结果表明，SkillSentry 平均提高了 LLM 代理在技能上的任务成功率 24.1%，同时在重复运行中表现出较低的变异性。
cs.AI / 176 / 2608.09254
Business Truth, not SQL Accuracy: A Rule-Gated 7B Analytics Agent Outperforms a Direct-Prompted 32B Baseline
商业真相，而非 SQL 精确性：一个规则门控的 7B 分析代理优于直接提示的 32B 基线
Abstract
LLM analytics agents are evaluated on SQL syntax accuracy, but production failures look different: questions with two valid business definitions, questions the warehouse cannot answer, deprecated columns after a schema change, and queries that execute successfully while returning the wrong business number. No execution-match metric can score them. This paper introduces WarehouseReliabilityBench, 400 frozen tasks over two synthetic warehouses in which roughly half the correct responses are a clarification, an abstention or a refusal, with pinned denominators and a pre-registered paired bootstrap fixing each claim verb before the numbers existed. QueryProof, a 7B agent, uses rules derived from a semantic layer and physical catalog to determine its behaviour, and gates every answer on deterministic post-execution checks. On an 80-task synthetic test split evaluated once, QueryProof outperforms a direct-prompted 32B baseline by +0.237 [+0.112, +0.375] Business Truth Rate at 71.0% lower cost per correct answer; against a cost-matched few-shot baseline the accuracy gain holds but the cost difference does not resolve. This compares systems rather than model sizes: the 32B baseline receives none of the scaffolding. False success falls from 0.754 to 0.351 of returned answers, and no wrong number was returned on an answerable task (0 of 24), though 13 answers went to questions requiring clarification or abstention. Removing the routing layer changes little (0.562 against 0.537), so the result does not depend on escalation. Routing tuned on validation over-abstains on test, and the fitted confidence model loses to the heuristic it replaced. Resampling template families rather than tasks widens both accuracy intervals to include zero, so the effect's direction is better supported than its magnitude. The gain tracks the deterministic layer, though no component ablation was run.
Chinese Translation
大型语言模型（LLM）分析代理通常通过 SQL 语法准确性进行评估，但生产中的失败表现有所不同：存在两个有效商业定义的问题、数据仓库无法回答的问题、模式变更后被弃用的列，以及成功执行但返回错误商业数字的查询。没有任何执行匹配指标能够对其进行评分。本文介绍了 WarehouseReliabilityBench，这是一个包含 400 个冻结任务的基准，基于两个合成数据仓库，其中大约一半的正确响应是澄清、弃权或拒绝，且在数字存在之前，固定的分母和预注册的配对自助法修正了每个声明动词。QueryProof，一个 7B 的代理，利用从语义层和物理目录中派生的规则来决定其行为，并在确定性后执行检查上对每个答案进行门控。在一次评估的 80 个任务的合成测试集中，QueryProof 的商业真相率比直接提示的 32B 基线高出 +0.237 [+0.112, +0.375]，且每个正确答案的成本降低了 71.0%；与成本匹配的少量样本基线相比，准确性提升依然存在，但成本差异未能解决。这是对系统而非模型规模的比较：32B 基线没有获得任何支撑。错误成功率从返回答案的 0.754 降至 0.351，在可回答的任务中没有返回错误数字（24 个任务中为 0），尽管有 13 个答案指向需要澄清或弃权的问题。移除路由层几乎没有改变（0.562 对 0.537），因此结果并不依赖于升级。基于验证的路由在测试中过度弃权，而拟合的置信模型不如其替代的启发式模型。重新抽样模板家族而非任务使得准确性区间扩大到包括零，因此效果的方向得到了更好的支持，而其幅度则不那么明确。增益与确定性层相关，尽管没有进行组件消融实验。
cs.AI / 177 / 2608.09263
Privileged Likelihood Is Not Automatically Value: Three Checks for Token Credit in On-Policy Self-Distillation
特权似然并不自动等同于价值：在策略自蒸馏中对标记信用的三项检验
Abstract
Outcome verifiers score completed reasoning traces but do not assign credit to intermediate tokens. Privileged self-distillation attempts to fill this gap by rescoring a model's own rollout with training-only information. A token likelihood change, however, is not automatically outcome credit. We separate three questions: whether the score tracks better actions, whether feedback construction changes what is compared, and what behavior the training loss reinforces. We establish these distinctions formally. When a rollout is scored using hindsight feedback written about that same rollout, its content determines both the tokens and the scoring context, creating direct self-dependence. Using feedback from another rollout of the same problem removes this dependence but does not guarantee a useful score. In matched experiments with a 20B model on AIME 2025, the implemented additive score is near chance (AUC=0.505) and slightly favors incorrect traces after length adjustment. In the paired comparison, the outcome-only control records 64.2\%, versus 24.2\%--33.9\% for five token-score variants. The results motivate validating score meaning, feedback construction, and training behavior separately before calling a likelihood signal credit.
Chinese Translation
结果验证者对完成的推理轨迹进行评分，但不对中间标记分配信用。特权自蒸馏试图通过使用仅限训练的信息重新评分模型的自身展开，以填补这一空白。然而，标记似然的变化并不自动意味着结果信用。我们将三个问题分开：评分是否跟踪更好的动作，反馈构建是否改变比较的内容，以及训练损失强化了什么行为。我们正式建立了这些区别。当使用关于同一展开的事后反馈对展开进行评分时，其内容决定了标记和评分上下文，产生直接的自依赖性。使用来自同一问题的另一个展开的反馈消除了这种依赖，但并不保证获得有用的评分。在与 AIME 2025 上的 20B 模型进行的匹配实验中，实施的加性评分接近偶然（AUC=0.505），并在长度调整后略微偏向于不正确的轨迹。在配对比较中，仅结果控制记录为 64.2%，而五种标记评分变体的结果为 24.2%--33.9%。这些结果促使我们在将似然信号称为信用之前，分别验证评分意义、反馈构建和训练行为。
cs.AI / 178 / 2608.09273
Entropy-based Code Adversarial Translation for Real-world Repository Migration
基于熵的代码对抗翻译用于真实世界的代码库迁移
Abstract
LLMs have demonstrated strong capabilities in code generation and automated program repair, but migrating an entire repository rarely produces a runnable application because long-horizon translation challenges LLM-based agents' ability to maintain repository-level migration objectives. In this work, we propose Entropy-based Code Adversarial Translation (ECAT), a multi-agent framework for automated Android-to-HarmonyOS repository migration. ECAT formulates repository migration as adversarial entropy minimization through a generator-discriminator architecture. The discriminator measures migration quality using a unified metric called Code Entropy and produces text gradients that specify both file-level generation directives and the skills needed to execute them. Guided by these optimization signals, the generator iteratively updates the repository, and each update is accepted only if it reduces Code Entropy. Repeated generator--discriminator interactions progressively drive the migration from an initial template toward a functionally complete HarmonyOS repository. Successful low-entropy trajectories are further distilled into a self-evolving memory tree, enabling transferable migration knowledge across repositories. We also introduce A2H-RepoBench, the first real-world benchmark for Android-to-HarmonyOS repository migration, covering applications from tens of thousands to hundreds of thousands of lines of code. Evaluated by node alignment and an agent-based functional judge, ECAT achieves 74.7% overall migration quality and consistently outperforms existing agent-based methods across repositories of different scales.
Chinese Translation
大型语言模型（LLMs）在代码生成和自动程序修复方面表现出强大的能力，但迁移整个代码库往往无法产生可运行的应用程序，因为长时间跨度的翻译挑战了基于LLM的代理保持代码库级迁移目标的能力。在本研究中，我们提出了基于熵的代码对抗翻译（Entropy-based Code Adversarial Translation, ECAT），这是一个用于自动化Android到HarmonyOS代码库迁移的多代理框架。ECAT将代码库迁移公式化为对抗性熵最小化，通过生成器-判别器架构实现。判别器使用一种称为代码熵（Code Entropy）的统一指标来衡量迁移质量，并生成文本梯度，指明文件级生成指令及执行这些指令所需的技能。在这些优化信号的指导下，生成器迭代更新代码库，每次更新仅在减少代码熵时被接受。生成器与判别器的反复交互逐步推动迁移从初始模板向功能完整的HarmonyOS代码库发展。成功的低熵轨迹进一步被提炼为自我演化的记忆树，使得跨代码库的迁移知识得以转移。我们还引入了A2H-RepoBench，这是第一个针对Android到HarmonyOS代码库迁移的真实世界基准，涵盖了从数万行到数十万行代码的应用程序。通过节点对齐和基于代理的功能评判，ECAT实现了74.7%的整体迁移质量，并在不同规模的代码库中持续优于现有的基于代理的方法。
cs.AI / 179 / 2608.09277
P$^{3}$: Joint Program-and-Proof Planning for Verified Code Generation
P$^{3}$：联合程序与证明规划以实现验证代码生成
Abstract
Verified code generation asks a large language model (LLM) to generate both an executable program and a machine-checkable proof that the program meets a formal specification, promising software that is correct by construction. The de facto workflow decouples the two halves of the problem: first synthesize a program, then attempt to prove it correct. We observe that this sequential pipeline can be both ineffective and inefficient in practice. A program generated without anticipating its proof can be subtly incorrect or structurally difficult to verify, forcing the LLM into brittle repair loops that alternate between patching the code and patching the proof. Inspired by Dijkstra's view that a program and its correctness argument should be developed hand in hand, we propose $P^3$, an LLM-based agentic workflow that first derives a unified program-and-proof plan from the specification, then elaborates the implementation and proof scaffold under this shared plan. To evaluate verified code generation in realistic settings, we further introduce Lean4Commit0, a repository-derived, library-level benchmark built by extracting core APIs from real-world software repositories and translating their requirements, including relational specifications across APIs, into Lean tasks. Using four frontier LLM backends, we evaluate $P^3$ on Verina, AlgoVeri, and our Lean4Commit0 benchmark, where it achieves the highest solve rate in every benchmark--model setting. Compared with the stronger baseline, it improves solve rates by 4.6--11.2 percentage points and reduces per-task API cost by up to roughly 40\% and wall-clock time by up to roughly 37\% on the difficult subset of each benchmark. A targeted ablation further shows gains of 3.3--8.3 points over implementation-only planning, isolating the benefit of planning the program and proof jointly.
Chinese Translation
验证代码生成要求大型语言模型（LLM）生成可执行程序和可机器检查的证明，以确保该程序满足形式规范，从而承诺构建正确的软件。现有的工作流程将问题的两个部分解耦：首先合成一个程序，然后尝试证明其正确性。我们观察到，这种顺序管道在实践中可能既无效又低效。未考虑其证明的程序可能会在细微之处不正确或在结构上难以验证，迫使LLM进入脆弱的修复循环，交替进行修补代码和修补证明。受到Dijkstra观点的启发，即程序及其正确性论证应当并行开发，我们提出了$P^3$，一个基于LLM的自主工作流程，首先从规范中推导出统一的程序与证明计划，然后在这一共享计划下详细阐述实现和证明框架。为了在现实环境中评估验证代码生成，我们进一步引入了Lean4Commit0，这是一个基于库的基准，通过从真实软件库中提取核心API并将其需求（包括API之间的关系规范）翻译为Lean任务而构建。使用四个前沿LLM后端，我们在Verina、AlgoVeri和我们的Lean4Commit0基准上评估了$P^3$，在每个基准-模型设置中均达到了最高的解决率。与更强的基线相比，它提高了解决率4.6至11.2个百分点，并在每个基准的困难子集上将每个任务的API成本降低了约40%，将实际时间减少了约37%。有针对性的消融实验进一步显示，与仅实施规划相比，联合规划程序和证明的收益为3.3至8.3个百分点。
cs.AI / 180 / 2608.09281
MMArch: Benchmarking Multimodal Reasoning Grounded in Architectural Evidence
MMArch：基于建筑证据的多模态推理基准测试
Abstract
Multimodal large language models (MLLMs) perform strongly on engineering imagery, yet existing benchmarks mostly test drawing recognition, information extraction, or compliance checking, leaving open whether models can combine distributed visual evidence with engineering principles to reach a conclusion. We introduce MMArch, a benchmark for architecture and civil engineering spanning ten subdomains and built entirely from figures in peer-reviewed papers. Its $1{,}212$ short-answer items are produced by a decoupled planner--writer pipeline and validated through automated screening, a blind adversarial audit, and expert review, so that answering requires perceiving the relevant evidence, identifying the governing principle, and applying it, not exploiting textual or single-figure shortcuts. Evaluating $18$ open-weight and proprietary MLLMs against a domain-expert panel, we find a wide gap: the strongest open-source model attains about $30\%$ and the best proprietary system $52\%$, while human experts reach $95\%$, more than forty points ahead. Our error analysis shows that failures concentrate in applying principles and combining evidence across figures rather than in locating it, pointing to substantial headroom for future research. Code and data are available at https://dcx-swjtu.github.io/MMArch/.
Chinese Translation
多模态大型语言模型（MLLMs）在工程图像上表现出色，但现有基准测试主要集中于图纸识别、信息提取或合规检查，尚未验证模型是否能够将分散的视觉证据与工程原理结合以得出结论。我们推出了MMArch，这是一个涵盖十个子领域的建筑和土木工程基准，完全由经过同行评审的论文中的图形构建而成。其1212个简答题项通过解耦的规划者-写作者流程生成，并通过自动筛选、盲审对抗审计和专家评审进行验证，因此回答这些问题需要感知相关证据、识别主导原理并加以应用，而不是利用文本或单一图形的捷径。我们对18个开放权重和专有的MLLMs进行了评估，结果显示存在较大差距：最强的开源模型仅达到约30%，而最佳的专有系统为52%，而人类专家的准确率达到95%，领先超过四十个百分点。我们的错误分析表明，失败主要集中在应用原理和跨图形结合证据上，而不是在定位证据上，指向未来研究的巨大潜力。代码和数据可在https://dcx-swjtu.github.io/MMArch/获取。
cs.AI / 181 / 2608.09282
ComboShoppingBench: Evaluating LLM Agents for Budget-Constrained Basket Shopping with Coupons
ComboShoppingBench：评估预算受限的购物篮组合购物中的大型语言模型代理
Abstract
Real-world shopping often requires constructing a basket of complementary items rather than retrieving a single product. Such combo-shopping tasks arise in device setup, meal preparation, event planning, and group takeout ordering, requiring joint reasoning about item compatibility, availability, store-level requirements, delivery fees, coupons, and budgets. Evaluation is challenging because multiple baskets may satisfy the same request, making exact-match metrics unsuitable, whereas semantic evaluation alone cannot detect infeasible orders, invalid coupon combinations, or incorrect payments. We introduce ComboShoppingBench, an agentic shopping benchmark for open-ended yet verifiable basket construction in a simulated commerce and takeout environment. During task synthesis, an exploration agent constructs a feasible and semantically coherent basket of purchasable products; this witness guides the generation of coupons, budget constraints, user queries, and aligned evaluation rubrics. During evaluation, LLM judges assess semantic satisfaction, response quality, and claim faithfulness, while deterministic validation checks product-ID validity, budget compliance, and coupon optimality. Experiments with diverse LLM agents demonstrate that even strong agents struggle on ComboShoppingBench, highlighting substantial room for improvement in reliable, constraint-aware combo shopping.
Chinese Translation
现实世界的购物通常需要构建一个由互补商品组成的购物篮，而不是仅仅检索单一产品。这类组合购物任务出现在设备设置、餐食准备、活动策划和团体外卖订购中，需要对商品的兼容性、可用性、商店级要求、配送费用、优惠券和预算进行联合推理。评估这一过程具有挑战性，因为多个购物篮可能满足相同的请求，使得精确匹配指标不适用，而单纯的语义评估无法检测不可行的订单、无效的优惠券组合或错误的支付。我们引入了ComboShoppingBench，这是一个针对开放式但可验证的购物篮构建的代理购物基准，旨在模拟商业和外卖环境。在任务合成过程中，一个探索代理构建了一个可行且语义连贯的可购买商品购物篮；这一证据引导生成优惠券、预算约束、用户查询和对齐的评估标准。在评估过程中，大型语言模型（LLM）评审员评估语义满意度、响应质量和声明的真实性，而确定性验证则检查产品ID的有效性、预算合规性和优惠券的最优性。对多种LLM代理的实验表明，即使是强大的代理在ComboShoppingBench上也面临挑战，突显了在可靠的、考虑约束的组合购物方面仍有很大的改进空间。
cs.AI / 182 / 2608.09296
CADEngBench: It Looks Like CAD, but Does It Work? Evaluating Parametric Design, Assembly Reasoning, and Physics Simulation
CADEngBench：看起来像CAD，但它真的有效吗？评估参数化设计、装配推理和物理仿真
Abstract
A CAD model is not engineering-grade merely because it looks correct. It must satisfy design requirements, respond predictably to parameter changes, support controlled edits, match a reference structural response under a declared analysis, and connect to other parts through valid joints. We present CADEngBench, a two-track benchmark for these capabilities. CADEngBench-P evaluates 300 parametric parts, each used for one zero-to-CAD task and one functional-editing task (600 tasks in total), through boundary-representation (B-Rep) validity, engineering and DFM checks, parameter-family perturbations, functional editing, and matched linear-static FEA in CalculiX. CADEngBench-A evaluates 150 body pairs through ranked joint retrieval, exact face-and-edge grounding, joint-frame prediction, and kinematic verification. Across eight multimodal, code-capable models, editing supplied CAD is substantially easier than generating it, while complex edits and matched FEA remain difficult. Assembly predictions often locate the relevant region but fail to recover the recorded joint or mating entities. These results show that CAD evaluation must test engineering behavior rather than appearance alone.
Chinese Translation
一个CAD模型仅仅因为外观正确并不意味着它具备工程级别的标准。它必须满足设计要求，对参数变化做出可预测的响应，支持可控编辑，在声明的分析下与参考结构响应相匹配，并通过有效的连接件与其他部件相连。我们提出了CADEngBench，这是一个针对这些能力的双轨基准测试。CADEngBench-P评估300个参数化部件，每个部件用于一个从零到CAD的任务和一个功能编辑任务（总共600个任务），通过边界表示（B-Rep）有效性、工程和DFM检查、参数族扰动、功能编辑以及在CalculiX中的匹配线性静态有限元分析（FEA）进行评估。CADEngBench-A评估150对体，通过排名连接检索、精确的面和边基础、连接框架预测和运动学验证进行评估。在八个多模态、可编程模型中，编辑提供的CAD显著比生成它更容易，而复杂的编辑和匹配的FEA仍然困难。装配预测通常能够定位相关区域，但未能恢复记录的连接或配合实体。这些结果表明，CAD评估必须测试工程行为，而不仅仅是外观。
cs.AI / 183 / 2608.09307
Linearized 2-Simplicial Attention
线性化的2-单纯形注意力
Abstract
We present a linearized form of 2-simplicial attention by rewriting the trilinear score as an inner product between a composite query and a key, so that the sum over one token axis takes the same form as ordinary softmax attention. We then approximate this sum with positive random features and store the entire past in a fixed-size state, while the second axis stays explicit over a short window of recent tokens. This enables us to achieve linear cost in sequence length combined with a global reach that windowed 2-simplicial attention lacks. We implement it with custom Triton kernels and combine it with Kimi Delta Attention to build a model with no softmax attention at all. Under matched compute, this model achieves the highest mean downstream accuracy among the compared architectures, and at 16k context it improves mean accuracy over a KDA hybrid while lowering LAMBADA perplexity from 715.6 to 602.6.
Chinese Translation
我们通过将三线性得分重写为复合查询与键之间的内积，提出了一种线性化的2-单纯形注意力形式，从而使得在一个令牌轴上的求和与普通的softmax注意力具有相同的形式。然后，我们用正随机特征来近似这个求和，并将整个过去存储在一个固定大小的状态中，同时第二个轴在最近的短窗口内保持显式。这使我们能够在序列长度上实现线性成本，并结合全局范围，这一点是窗口化的2-单纯形注意力所缺乏的。我们使用自定义的Triton内核实现它，并将其与Kimi Delta Attention结合，构建一个完全不使用softmax注意力的模型。在匹配的计算条件下，该模型在比较的架构中实现了最高的平均下游准确率，并且在16k上下文下，相比于KDA混合模型提高了平均准确率，同时将LAMBADA困惑度从715.6降低到602.6。
cs.AI / 184 / 2608.09315
ASPaeroFlow: Decomposition Heuristics for Joint Air Traffic Flow & Capacity Management
ASPaeroFlow：联合空中交通流量与容量管理的分解启发式方法
Abstract
While mathematical models act as vital decision support systems for operational Air Traffic Flow and Capacity Management (ATFCM), existing approaches isolate Air Traffic Flow Management (ATFM) from Dynamic Airspace Configuration (DAC). This separation introduces an unresolved circular dependency between fixed-demand and fixed-capacity assumptions. Although joint optimization resolves this gap, the enlarged search space renders exact models computationally intractable for medium- to large-scale instances. To bridge this gap, we propose ASPaeroFlow: a heuristic for the joint ATFCM; it combines instance-space decomposition heuristics with a local exact approach using Answer Set Programming. We benchmark ASPaeroFlow from small to industry-sized instances and compare it with exact and alternative approaches. The results indicate that (1) the heuristic provides a computational middle ground between exact methods and operational baselines; (2) simultaneous optimization can outperform sequential optimization on joint ATFCM; and (3) an ablation study indicates that DAC has a larger impact on solution quality than flow measures.
Chinese Translation
尽管数学模型在运营空中交通流量与容量管理（ATFCM）中作为重要的决策支持系统，但现有方法将空中交通流量管理（ATFM）与动态空域配置（DAC）相互隔离。这种分离导致了固定需求与固定容量假设之间未解决的循环依赖关系。尽管联合优化能够解决这一问题，但扩大的搜索空间使得精确模型在中到大规模实例中计算上变得不可行。为了解决这一问题，我们提出了ASPaeroFlow：一种用于联合ATFCM的启发式方法；它结合了实例空间分解启发式和使用答案集编程（Answer Set Programming）的局部精确方法。我们对ASPaeroFlow进行了从小规模到行业规模实例的基准测试，并与精确方法和其他替代方法进行了比较。结果表明：（1）该启发式方法在精确方法与操作基线之间提供了计算上的折中；（2）同时优化在联合ATFCM中可以优于顺序优化；（3）消融研究表明，DAC对解的质量影响大于流量度量。
cs.AI / 185 / 2608.09324
CoRE: Consensus Rewards via Equilibrium for Test-Time Reinforcement Learning
CoRE：通过平衡实现测试时强化学习的共识奖励
Abstract
On unlabeled test data, reinforcement learning lacks a ground-truth reward; test-time RL methods derive one from the model's own roll-outs, rewarding those that match the majority vote over $N$ sampled answers. That vote discards a correct answer whenever it is a minority and scores every majority-matching roll-out identically. We replace it with \emph{CoRE} (Consensus Rewards via Equilibrium): the $N$ roll-outs form a graph whose edges combine answer agreement, reasoning similarity, and generation confidence, and replicator dynamics extract its dominant set, yielding a refined pseudo-label, a graded per-roll-out reward, and a per-question cohesiveness gate. CoRE strictly generalizes voting: majority voting is recovered as a special case; a block-value analysis gives a sharp threshold for when consensus recovers a correct minority against a larger wrong plurality; and confidence calibration provably lowers that threshold multiplicatively. Across seven backbones and five benchmarks (42 model--benchmark cells, three seeds each), \emph{CoRE} improves the untrained base by $+21.7$ points on average versus $+20.4$ for majority-vote TTRL, wins wherever agreement is contestable with margins over the vote of up to $+7.5$ points, and reaches the voting baseline's plateau accuracy in $54$--$70$\% fewer steps. Consensus, not counting: treating the roll-out group as a graph rather than a ballot box turns a brittle vote into a calibrated, graded, self-supervised reward at no extra roll-out cost.
Chinese Translation
在未标记的测试数据上，强化学习缺乏真实奖励；测试时强化学习方法从模型自身的回滚中推导出奖励，奖励那些与$N$个采样答案中的多数投票相匹配的答案。该投票在少数情况下会丢弃正确答案，并对每个与多数匹配的回滚给予相同的评分。我们用 extit{CoRE}（通过平衡实现共识奖励）替代它：$N$个回滚形成一个图，其边缘结合了答案一致性、推理相似性和生成信心，复制者动态提取其主导集，从而产生一个精炼的伪标签、每个回滚的分级奖励和每个问题的一致性门。CoRE严格推广了投票：多数投票作为特例被恢复；块值分析提供了一个明确的阈值，用于判断共识何时能够恢复正确的少数答案对更大错误多数的反击；而信心校准则证明性地降低了该阈值的乘法。通过七个基础模型和五个基准（42个模型-基准单元，每个单元三个种子）， extit{CoRE}在未训练的基础上平均提高了$+21.7$分，而多数投票的测试时强化学习（TTRL）为$+20.4$，在一致性存在争议的地方获胜，投票差距高达$+7.5$分，并在$54$--$70\%更少的步骤中达到了投票基线的平稳准确率。共识，而非计数：将回滚组视为图而非投票箱，将脆弱的投票转变为经过校准的、分级的自监督奖励，而无需额外的回滚成本。
cs.AI / 186 / 2608.09325
GeoPhysAdapter: Scale-Matched Geophysical Adaptation for Cross-Domain Landslide Mapping with Vision Foundation Models
GeoPhysAdapter：用于跨领域滑坡制图的尺度匹配地球物理适应方法与视觉基础模型
Abstract
Newly triggered landslides rarely carry immediate annotations, so cross-domain transferability determines the value of landslide mapping for emergency response and regional risk assessment. Vision foundation models have strengthened representational transfer, yet on unseen regions, events, and data sources they still generate high-confidence false alarms. Terrain, material, and rainfall triggering can constrain such errors, but their supports are local, regional, and event-scale, so that resampling onto a 10~m grid misaligns them with the segmentation decision unit and compounds the uncertain geographic context problem (UGCoP). We propose GeoPhysAdapter, which anchors on a frozen vision foundation model, restricts terrain, material, and triggering to dense spatial guidance, regional modulation, and event-timing forcing, and applies bounded adaptation at two decision units, the pixel and the candidate landslide body, reverting exactly to the visual prediction where support is insufficient. On an event-isolated PILD dataset of four public sources, 55 global landslide events, and 7,890 test samples, 70.3% of cross-domain false-positive mass lies in near-pure spurious bodies of median equivalent diameter 207m, matching coarse-prior support rather than the pixel. Pixel-level adaptation removes a net 507,817 erroneous pixels and reduces error by 7.76%, whereas raising the decision unit to the candidate body, under identical samples, anchor, and baseline, increases error reduction to 23.99%, approximately 3.1 times the pixel-level effect, improves IoU by 0.031 (14.2% relative), and corrects 9.92 pixels per pixel harmed. The data and code are publicly available at: https://github.com/Liu-Zhihang/geophysadapter.
Chinese Translation
新触发的滑坡事件通常缺乏即时标注，因此跨领域的可转移性决定了滑坡制图在应急响应和区域风险评估中的价值。视觉基础模型增强了表征转移，但在未见区域、事件和数据源上，它们仍然会产生高置信度的误报。地形、材料和降雨触发因素可以限制此类错误，但它们的支持是局部的、区域性的和事件尺度的，因此在10米网格上重新采样会使其与分割决策单元不对齐，并加剧不确定地理背景问题（UGCoP）。我们提出了GeoPhysAdapter，该方法基于一个冻结的视觉基础模型，将地形、材料和触发因素限制为密集的空间引导、区域调制和事件时序强制，并在两个决策单元（像素和候选滑坡体）上应用有界适应，当支持不足时，准确回归到视觉预测。在一个隔离事件的PILD数据集上，该数据集包含四个公共来源、55个全球滑坡事件和7,890个测试样本，70.3%的跨领域假阳性质量位于中位等效直径为207米的近纯虚假体中，匹配粗略的先验支持而非像素级支持。像素级适应去除了507,817个错误像素，误差减少了7.76%，而将决策单元提高到候选体，在相同样本、锚点和基线条件下，误差减少提高到23.99%，约为像素级效果的3.1倍，IoU提高了0.031（相对提高14.2%），每个受损像素纠正了9.92个像素。数据和代码可在以下网址公开获取：https://github.com/Liu-Zhihang/geophysadapter。
cs.AI / 187 / 2608.09335
Control-Oriented Scenario Tree Construction through Reinforcement Learning
基于强化学习的控制导向场景树构建
Abstract
Multistage stochastic model predictive control (MPC) handles uncertainty by optimizing over a scenario tree, a finite branching approximation of future outcomes constructed from sampled forecasts. To build such a tree, conventional methods focus on matching the underlying probability distribution---e.g., via Wasserstein-based scenario reduction---but improved distributional accuracy does not necessarily yield better control performance. We propose a control-oriented approach that learns scenario tree construction directly from its impact on downstream decisions. Fixing the tree topology, we formulate tree construction as a sequential assignment of sampled scenarios to leaves. This assignment is parameterized by an attention-based policy over the scenario set and trained using reinforcement learning, with closed-loop control profit as the objective. Training is stabilized by an asymmetric critic that leverages realized future trajectories. We evaluate the method on a risk-averse battery arbitrage problem. Across a range of forecast set sizes, the learned construction consistently achieves the highest profit, outperforming classical forward and backward reduction methods and certainty-equivalent (single-trajectory forecast) control. The learned policy also exhibits greater robustness on challenging instances, consistently demonstrating better tail-risk characteristics. Analysis of the resulting trees indicates that our method constructs compact, selectively branching structures that capture high-impact events while keeping most trajectories nearly deterministic. These findings highlight that the value of a scenario tree depends critically on the decisions it supports, and provide an effective framework to train scenario tree constructors merely based on the closed-loop control optimization signal.
Chinese Translation
多阶段随机模型预测控制（MPC）通过在场景树上进行优化来处理不确定性，场景树是从采样预测构建的未来结果的有限分支近似。为了构建这样的树，传统方法侧重于匹配基础概率分布——例如，通过基于Wasserstein的场景简化——但提高分布准确性并不一定能带来更好的控制性能。我们提出了一种控制导向的方法，直接从对下游决策的影响中学习场景树的构建。固定树的拓扑结构，我们将树的构建形式化为将采样场景顺序分配到叶子节点的过程。该分配由基于注意力的策略参数化，并使用强化学习进行训练，以闭环控制收益作为目标。训练通过利用实现的未来轨迹的非对称评论员得以稳定。我们在一个风险厌恶的电池套利问题上评估该方法。在不同的预测集大小范围内，学习到的构建方法始终实现了最高的收益，超越了经典的前向和后向简化方法以及确定性等效（单轨迹预测）控制。学习到的策略在具有挑战性的实例中也表现出更大的鲁棒性，始终展示出更好的尾部风险特征。对生成的树的分析表明，我们的方法构建了紧凑的、选择性分支的结构，能够捕捉高影响事件，同时保持大多数轨迹几乎是确定性的。这些发现突显了场景树的价值在于其所支持的决策，并提供了一个有效的框架，仅基于闭环控制优化信号来训练场景树构建者。
cs.AI / 188 / 2608.09343
LLM-Guided Heuristic Design from Simulation Traces: A Case Study in Dynamic Production and AGV Scheduling
基于模拟轨迹的LLM引导启发式设计：动态生产与AGV调度的案例研究
Abstract
Simulation-based optimization (SBO) evaluates executable policies under stochastic dynamics, but most methods treat the simulator as a black box: aggregate scores rank candidates without revealing why they fail or which policy logic should change. We present an LLM-guided heuristic design framework that uses repeated simulation for selection and event-level traces for diagnosis. Each incumbent is assessed through multiple replications, while replaying its lowest-scoring one produces a queryable trace. A manager agent formulates bottleneck hypotheses from this evidence, and editing agents implement parallel code-level revisions. After execution checks and repeated evaluation, best-so-far selection retains only improvements. LLM revision occurs between evaluation batches, while a fixed policy controls each simulation run. We evaluate the framework in a discrete-event simulation of dynamic production and automated guided vehicle (AGV) scheduling. Across five independent optimization runs with Gemini-3.1-Pro, final mean scores averaged 77.51 on the simulator's 0-100 scale. In the highest-scoring run, trace-based diagnoses motivated proactive charging, distance-aware AGV assignment, and rebalanced dispatch priorities, raising the best-so-far mean score from 62.49 to 78.61. On 100 matched seeds, the best final policy outscored representative rolling-MILP, rule-based, and metaheuristic policies on every seed and retained its advantage under random faults without re-optimization. After separate re-optimization for a longer horizon and variable order interarrival times, the resulting policies again outscored all baselines. Ablations with two LLM backbones showed that removing either parallel candidate generation or trace-database access reduced final mean scores. These results show that simulation traces can guide targeted code-level policy improvement in complex simulation-based scheduling.
Chinese Translation
基于模拟的优化（SBO）在随机动态下评估可执行策略，但大多数方法将模拟器视为黑箱：聚合得分对候选者进行排名，却未揭示其失败原因或应更改的策略逻辑。我们提出了一种LLM引导的启发式设计框架，该框架利用重复模拟进行选择，并通过事件级轨迹进行诊断。每个现有策略通过多次复制进行评估，而重放其最低得分的复制则生成可查询的轨迹。管理代理根据这些证据制定瓶颈假设，编辑代理则实施并行代码级修订。在执行检查和重复评估后，最佳迄今为止的选择仅保留改进。LLM修订发生在评估批次之间，而固定策略控制每次模拟运行。我们在动态生产和自动导引车（AGV）调度的离散事件模拟中评估了该框架。在与Gemini-3.1-Pro进行的五次独立优化运行中，最终平均得分在模拟器的0-100评分尺度上达到了77.51。在得分最高的运行中，基于轨迹的诊断促使了主动充电、基于距离的AGV分配和重新平衡的调度优先级，使最佳迄今为止的平均得分从62.49提高到78.61。在100个匹配种子上，最佳最终策略在每个种子上均超越了代表性的滚动-MILP、基于规则和元启发式策略，并在没有重新优化的情况下保持了其优势。在对更长时间范围和可变到达时间顺序进行单独重新优化后，得到的策略再次超越了所有基线。与两个LLM骨干网络的消融实验表明，去除并行候选生成或轨迹数据库访问都会降低最终平均得分。这些结果表明，模拟轨迹可以指导复杂基于模拟的调度中的有针对性的代码级策略改进。
cs.AI / 189 / 2608.09374
CircuitReason-1k: Benchmarking Long-Horizon Visual-to-Symbolic Reasoning inElectrical Circuits
CircuitReason-1k：电路中长时间视觉到符号推理的基准测试
Abstract
Electrical circuit analysis requires more than recognizing components in an image. A solver must ground symbols and labels, recover latent topology, select a physical model, formulate coupled equations, propagate intermediate quantities, and preserve units, signs, directions, and phase conventions. We introduce \benchmark, a benchmark of 1,000 authentic textbook problems for evaluating this complete long-horizon visual-to-symbolic reasoning process. Each problem pairs one or more circuit diagrams with a self-contained question, a typed or semantically specified answer, and a reference worked solution. An evidence-first construction pipeline aligns questions, figures, and solutions, while a reasoning-oriented taxonomy organizes problems by circuit type and dependency depth. Evaluation combines conservative typed scoring with identity-blinded multi-model semantic consensus, retaining every problem in the denominator. Across three commercial chatbot systems and six open-source multimodal large language models, the highest-scoring system reaches 84.8\% accuracy. However, performance consistently deteriorates on long-horizon problems, and qualitative analysis exposes persistent failures in topology-to-target binding, physical conventions, and late-stage output propagation. \benchmark{} provides a focused testbed for measuring whether multimodal models can transform technical visual evidence into sustained, physically valid symbolic reasoning. Code are available at GitHub - CircuitReason/CircuitReason1K.
Chinese Translation
电路分析不仅仅需要识别图像中的组件。求解器必须将符号和标签进行基础化，恢复潜在拓扑，选择物理模型，制定耦合方程，传播中间量，并保持单位、符号、方向和相位约定。我们引入了 enchmark，这是一个包含1000个真实教材问题的基准，用于评估这一完整的长时间视觉到符号推理过程。每个问题将一个或多个电路图与一个自包含的问题、一个类型化或语义指定的答案以及一个参考解答配对。证据优先的构建流程将问题、图形和解答对齐，而以推理为导向的分类法则按电路类型和依赖深度组织问题。评估结合了保守的类型评分和身份盲多模型语义共识，保留了每个问题在分母中。在三个商业聊天机器人系统和六个开源多模态大型语言模型中，得分最高的系统达到了84.8%的准确率。然而，在长时间问题上，性能持续下降，定性分析揭示了拓扑到目标绑定、物理约定和后期输出传播中的持续失败。enchmark{} 提供了一个专注的测试平台，以衡量多模态模型是否能够将技术视觉证据转化为持续的、物理有效的符号推理。代码可在GitHub - CircuitReason/CircuitReason1K获取。
cs.AI / 190 / 2608.09380
OpenLoopEvolve: A Verifiable Self-Evolution Framework for Loop Policies in Long-Horizon Complex Tasks
OpenLoopEvolve：一种可验证的自我进化框架，用于长时间复杂任务中的循环策略
Abstract
Long-horizon complex tasks require agents to repeatedly observe states, formulate plans, invoke tools, verify results, and recover from failures in continuously changing environments. However, such control experience often remains confined to a single context or a fixed prompt, and is difficult to accumulate and reuse across historical traces. This paper presents OpenLoopEvolve (OLE), a self-evolution framework centered on the Loop Policy. OLE represents an agent's observation, planning, memory, action, verification, recovery, stopping, and budget control as portable policy assets with versions and lineages, and provides online and offline evolution modes that can be selected according to practical needs: the online mode triggers candidate generation based on feedback from continuous operation, whereas the offline mode searches for candidate policies from archived traces and failure evidence. Both modes share an evolution mechanism consisting of autonomous proposals by a large language model, Champion--Challenger paired evaluation, and robust release. Policies released online are activated at a subsequent task boundary, monitored using subsequent feedback, and rolled back to their parent versions when degradation conditions are met. On the simulated business benchmark YC-Bench, both modes improve aggregate task performance, task success rate, and risk metrics relative to a fixed initial Loop Policy. The results indicate that treating the Loop Policy as a governable asset can support the accumulation, comparison, release, and reuse of control experience and improve agent performance on long-horizon complex tasks.
Chinese Translation
长时间复杂任务要求智能体反复观察状态、制定计划、调用工具、验证结果并在不断变化的环境中从失败中恢复。然而，这种控制经验往往局限于单一上下文或固定提示，难以在历史轨迹中积累和重用。本文提出了OpenLoopEvolve（OLE），一个以循环策略为中心的自我进化框架。OLE将智能体的观察、规划、记忆、行动、验证、恢复、停止和预算控制表示为具有版本和谱系的可移植策略资产，并提供可根据实际需求选择的在线和离线进化模式：在线模式根据持续操作的反馈触发候选生成，而离线模式则从归档的轨迹和失败证据中搜索候选策略。这两种模式共享一个进化机制，包括大型语言模型的自主提案、冠军-挑战者配对评估和稳健发布。在线发布的策略在后续任务边界被激活，使用后续反馈进行监控，并在满足退化条件时回滚到其父版本。在模拟商业基准YC-Bench上，这两种模式相对于固定初始循环策略提高了整体任务性能、任务成功率和风险指标。结果表明，将循环策略视为可管理资产可以支持控制经验的积累、比较、发布和重用，并提高智能体在长时间复杂任务中的表现。
cs.AI / 191 / 2608.09412
KVDiagnosis: A Diagnostic Benchmark for KV-Cache Compression in Long-Context Language Models
KVDiagnosis：长上下文语言模型中KV缓存压缩的诊断基准
Abstract
KV-cache compression reduces long-context memory, but aggregate task scores reveal neither which correct executions fail nor why. We present KVDiagnosis, a diagnostic dataset and benchmark with three contributions. First, a 25-method taxonomy groups methods into five mechanism families and links them to eight verified implementations and their valid diagnostic measurements. Second, for every supported method setting, we evaluate all sources in each fixed split against a per-source FullCache control before selecting FullCache-correct/compressed-wrong (C-to-W) rows separately for each method-setting, so no compressor defines another's test set. Third, a common record format links paired outputs and run metadata to cache, likelihood, attention, and decoding measurements with explicit applicability states. On Qwen3-8B, four evidence-aware workloads yield 59 800 supported compressed runs over 2600 sources and 12 520 C-to-W rows. Under fixed diagnostic rules, 63.2% have low or partial measured/projected coverage. Only 19 rows (0.2%) combine high measured/projected coverage with strong likelihood drift; another 2,126 (17.0%) preserve structural position addressability, for which representation fidelity remains unknown, while showing the same drift. Against C-to-C success controls, all ten diagnostics separate failed from successful compression (stratified AUROC 0.684-0.871). Among 96 reproducible low-EAR failures, a controlled 4x evidence-attention boost repairs 29.2%, versus 6.3% under a count-matched sham intervention and 3.3% degradation on matched C-to-C controls. Code and data are available at https://github.com/ChosenQC/KVDiagnosis.
Chinese Translation
KV缓存压缩减少了长上下文内存，但综合任务得分并未揭示哪些正确执行失败以及原因。我们提出了KVDiagnosis，这是一个具有三项贡献的诊断数据集和基准。首先，25种方法的分类法将方法分为五个机制家族，并将其与八个经过验证的实现及其有效的诊断测量相链接。其次，对于每个支持的方法设置，我们在每个固定分割中评估所有来源，相对于每个来源的FullCache控制，然后单独选择FullCache正确/压缩错误（C-to-W）行，因此没有任何压缩器定义其他的测试集。第三，通用记录格式将配对输出和运行元数据与缓存、可能性、注意力和解码测量连接，并具有明确的适用状态。在Qwen3-8B上，四个证据感知工作负载产生了59,800个支持的压缩运行，涵盖2600个来源和12,520个C-to-W行。在固定的诊断规则下，63.2%的测量/预测覆盖率低或部分覆盖。只有19行（0.2%）将高测量/预测覆盖率与强可能性漂移结合在一起；另外2,126行（17.0%）保持结构位置可寻址性，其表示保真度仍然未知，同时显示相同的漂移。与C-to-C成功控制相比，所有十个诊断将失败的压缩与成功的压缩区分开（分层AUROC 0.684-0.871）。在96个可重复的低EAR失败中，受控的4倍证据-注意力提升修复了29.2%，而在计数匹配的虚假干预下仅修复了6.3%，在匹配的C-to-C控制下则出现了3.3%的降级。代码和数据可在https://github.com/ChosenQC/KVDiagnosis获取。
cs.AI / 192 / 2608.09435
Listen, See and Track: Spatio-Temporal Audio-Visual Sound Event Reasoning for Omni-Modal Language Models
倾听、观看与追踪：全模态语言模型的时空音频-视觉声音事件推理
Abstract
Understanding dynamic sound sources requires jointly determining what produces a sound, where the source is located, and how it moves over time. Yet existing audio-language models often represent clips as global acoustic events, while vision-language models lack the spatial audio cues needed to localize and track individual sources. To evaluate this missing capability, we introduce ST-OmniQA, a spatio-temporal audio-visual question-answering benchmark built from panoramic videos paired with synchronized first-order Ambisonics (FOA) audio of moving sound sources. It contains 40K videos and 400K question-answer pairs organized into four capability levels covering sound-event recognition, direction of arrival, source distance, motion trajectories, and temporally grounded audio-visual reasoning. Building on this benchmark, we propose ST-Omni-R1, which integrates FOA-derived semantic and trajectory representations with panoramic visual context and is trained through progressive curriculum learning and reasoning-tree reinforcement learning. ST-Omni-R1 achieves 77.83\% average semantic accuracy across the four levels, compared with 37.28\% for the best evaluated baseline. Results on three public spatial-audio benchmarks further indicate that its learned spatial and motion representations transfer beyond ST-OmniQA.
Chinese Translation
理解动态声音源需要共同确定声音的产生者、源的位置以及其随时间的移动方式。然而，现有的音频-语言模型通常将片段表示为全局声学事件，而视觉-语言模型缺乏定位和追踪单个源所需的空间音频线索。为了评估这一缺失的能力，我们引入了ST-OmniQA，这是一个基于全景视频与同步的第一阶Ambisonics (FOA) 音频配对的时空音频-视觉问答基准，旨在处理移动声音源。该基准包含40,000个视频和400,000个问答对，组织成四个能力级别，涵盖声音事件识别、到达方向、源距离、运动轨迹以及时间基础的音频-视觉推理。在此基准的基础上，我们提出了ST-Omni-R1，该模型将FOA派生的语义和轨迹表示与全景视觉上下文相结合，并通过渐进式课程学习和推理树强化学习进行训练。ST-Omni-R1在四个级别上实现了77.83%的平均语义准确率，而最佳评估基线的准确率为37.28%。在三个公共空间音频基准上的结果进一步表明，其学习到的空间和运动表示超越了ST-OmniQA。
cs.AI / 193 / 2608.09443
Coupled Graph--Policy Distillation for Personalized Medication Safety in Older Adults with Multimorbidity
针对多重疾病老年人个性化用药安全的耦合图-策略蒸馏
Abstract
Large language model (LLM) agents can support medication review between clinical visits, but safe choices for older adults with multimorbidity depend on conditions, medications, and geriatric risks that users may omit. We introduce ATLAS, a coupled graph--policy distillation framework for patient-adaptive medication safety. ATLAS structures guideline evidence as a medication-safety graph. Targeted questions update the patient state and distill relevant relations into a patient-specific medication conflict graph (PMCG). A risk-first multi-agent policy uses the PMCG to screen contraindications, assess cautions and monitoring needs, identify safer alternatives, and verify the final medication plan. We also introduce GeriMedBench, an interactive benchmark that tests safety-critical information acquisition and evidence-based decision revision. Across a European non-interactive multimorbidity benchmark, an Asian interactive multimorbidity benchmark, and an Asian non-interactive cross-guideline benchmark, ATLAS achieves the strongest complete-decision performance among the compared systems. On the European non-interactive multimorbidity benchmark, it exceeds the strongest proprietary LLM baseline by 53.73 points in Strict Success Rate and 14.63 points in overall safety reasoning score (OSRS), with no unsafe recommendations under the automated evaluator. A blinded clinician evaluation gives ATLAS higher mean ratings across all five criteria and flags potentially unsafe recommendations in one ATLAS case and two Gemini cases.
Chinese Translation
大型语言模型（LLM）代理可以在临床访问之间支持用药审查，但对于多重疾病的老年人来说，安全的选择依赖于用户可能遗漏的病情、药物和老年风险。我们介绍了ATLAS，一个用于患者自适应用药安全的耦合图-策略蒸馏框架。ATLAS将指南证据结构化为用药安全图。针对性的问题更新患者状态，并将相关关系蒸馏成患者特定的用药冲突图（PMCG）。一种以风险为先的多代理策略利用PMCG筛查禁忌症，评估注意事项和监测需求，识别更安全的替代方案，并验证最终的用药计划。我们还介绍了GeriMedBench，一个交互式基准测试，旨在测试安全关键的信息获取和基于证据的决策修订。在一个欧洲非交互式多重疾病基准、一个亚洲交互式多重疾病基准和一个亚洲非交互式跨指南基准中，ATLAS在比较系统中实现了最强的完整决策性能。在欧洲非交互式多重疾病基准中，它在严格成功率上超过了最强的专有LLM基线53.73分，在整体安全推理评分（OSRS）上超过14.63分，并且在自动评估器下没有不安全的推荐。盲评的临床医生评估给ATLAS在所有五个标准上更高的平均评分，并在一个ATLAS案例和两个Gemini案例中标记出潜在的不安全推荐。
cs.AI / 194 / 2608.09480
From Prompt to Harness: Coderlet from Scratch
从提示到工具：从零开始的Coderlet
Abstract
A model alone does not determine how a programming agent acts. What the model sees, how actions enter the environment, how feedback returns, and how one run affects the next all depend on how the harness is organized. Minimal examples usually show only the basic interaction between a model and tools, while production systems spread these relationships across complex components and dependencies. This paper studies a compact harness design by following a single request through context formation, model decision, environmental action, observation return, and state continuation. Three boundaries---model, execution, and state---connect the model service, tool environment, and persistent state, while the request lifecycle determines the order in which these transitions occur. Together, they show the harness's core role: turning model generations into environmental actions, carrying runtime feedback into later decisions, and allowing state to continue across requests. On top of this runtime structure, a harness can also be gradually refined across runs through continued bootstrapping. The design is realized in the executable artifact https://github.com/lilinxi/Coderlet.
Chinese Translation
单靠模型并不能决定编程代理的行为。模型所看到的内容、如何将动作引入环境、反馈如何返回，以及一次运行如何影响下一次运行，都依赖于工具的组织方式。最小示例通常仅展示模型与工具之间的基本交互，而生产系统则将这些关系扩展到复杂的组件和依赖关系中。本文通过跟踪单个请求的上下文形成、模型决策、环境动作、观察返回和状态延续，研究了一种紧凑的工具设计。三个边界——模型、执行和状态——连接了模型服务、工具环境和持久状态，而请求生命周期则决定了这些转换发生的顺序。它们共同展示了工具的核心作用：将模型生成转化为环境动作，将运行时反馈传递到后续决策中，并允许状态在请求之间持续。在这种运行时结构之上，工具还可以通过持续的自举过程在运行中逐步完善。该设计已在可执行工件 https://github.com/lilinxi/Coderlet 中实现。
cs.AI / 195 / 2608.09485
Capability Is Not Propensity: Measuring Pressure-Robust Cooperative Behavior in Civic LLM Agents
能力不是倾向：测量公民大型语言模型代理的压力鲁棒合作行为
Abstract
Cooperative capabilities in language models are dual-use. The same social reasoning that supports civic deliberation can also enable strategic omission, false consensus, and manipulative framing. We argue that Cooperative AI evaluations should separate what models can do under benign instructions from what they tend to do under realistic civic pressure. We introduce DiffCoop-Civic, a 10-scenario pilot evaluation suite spanning preference understanding, evidence and persuasion, commitment design, asymmetric information, and dissent preservation. Across seven models from four model families, subtle omission pressure produces a near-uniform shift: manipulative enablement rises by 1.17 points and dissent preservation falls by 1.67 points on a 5-point scale. Overt false-consensus pressure behaves differently: it triggers refusal or redirection in some aligned API models, but direct compliance in several open-weight models. A lightweight Pareto-Trace prompting intervention improves pressure robustness without simply relying on hard refusal. An anonymous reproducibility package is available at https://anonymous.4open.science/r/diffcoop-civil-771C.
Chinese Translation
语言模型中的合作能力具有双重用途。支持公民审议的社会推理同样可以促成战略性遗漏、虚假共识和操控性框架。我们认为，合作人工智能的评估应区分模型在良性指令下能够做什么与在现实公民压力下倾向于做什么。我们引入了 DiffCoop-Civic，一个涵盖偏好理解、证据与说服、承诺设计、不对称信息和异议保留的10场景试点评估套件。在来自四个模型家族的七个模型中，微妙的遗漏压力产生了近乎一致的变化：操控性使能在5分制上上升了1.17分，而异议保留下降了1.67分。明显的虚假共识压力表现不同：它在某些对齐的API模型中触发拒绝或重定向，但在几个开放权重模型中则直接遵从。一种轻量级的Pareto-Trace提示干预提高了压力鲁棒性，而不仅仅依赖于强硬拒绝。一个匿名的可重复性包可在 https://anonymous.4open.science/r/diffcoop-civil-771C 获取。
cs.AI / 196 / 2608.09512
Renormalising Generative Models for Active Inference: Foundations, Derivations, and Verification
主动推理的生成模型重正化：基础、推导与验证
Abstract
Active inference offers a unified framework for perception, learning, and action, but scaling discrete active-inference models to rich spatial and temporal domains remains difficult. Renormalising generative models (RGMs) address this challenge by composing discrete generative models across spatial and temporal scales, coarse-graining lower-level states and paths into higher-level causes for objects, events, and action. However, fully reproducing and adapting the framework remains difficult: the mathematical exposition is compact, and the reference implementations are deeply integrated within specialized software environments, leaving many algorithmic details implicit. This paper addresses these challenges by providing a self-contained, derivation-oriented account of RGMs together with an open, verified implementation. We explain how the hierarchy is built, how beliefs and actions are updated within it, and how information is passed between levels. Where the published equations and implementation differ in emphasis, we make those choices explicit and explain their modelling consequences. By clarifying the theory and separating it from its original implementation context, this work lowers practical barriers to entry and makes RGMs more transparent, auditable, and reproducible, providing a foundation for future quantitative evaluation and development on machine-learning benchmarks.
Chinese Translation
主动推理提供了一个统一的框架用于感知、学习和行动，但将离散的主动推理模型扩展到丰富的空间和时间领域仍然困难。生成模型重正化（Renormalising Generative Models, RGM）通过在空间和时间尺度上组合离散生成模型，粗粒化低层状态和路径为高层对象、事件和行动的原因，从而应对这一挑战。然而，完全重现和适应该框架仍然困难：数学阐述简洁，参考实现深度集成于专业软件环境中，许多算法细节隐含在其中。本文通过提供一个自包含的、以推导为导向的RGM说明以及一个开放的、经过验证的实现，来解决这些挑战。我们解释了如何构建层次结构，如何在其中更新信念和行动，以及如何在各层之间传递信息。在已发布的方程和实现中，我们明确了不同的侧重点，并解释了这些选择的建模后果。通过澄清理论并将其与原始实现背景分离，这项工作降低了实践进入的障碍，使RGM更加透明、可审计和可重现，为未来在机器学习基准上的定量评估和开发提供了基础。
cs.AI / 197 / 2608.09521
One Adapter Pair per Model: A Universal Activation Interface for Language Models
每个模型一个适配器对：语言模型的通用激活接口
Abstract
Activation-based tools are usually tied to one model's native hidden space, requiring probes, sparse autoencoders, and natural-language interpreters to be rebuilt or rediscovered for each new language model. We present a Universal Activation Bus, a framework that provides a common activation interface across compatible language models. Using a small set of source models, we learn a shared dense space together with one lightweight linear encoder--decoder adapter pair per model. After source training, the interface is frozen; a new model joins by fitting only its adapter pair on unlabeled matched text. The resulting interface allows activation-based tools to be shared across connected models, including common probes and SAE features as well as access to an NLA originally trained for a different model. Across five models, semantically related texts form consistent neighborhoods in the shared space, and an onboarded model reuses these tools effectively without retraining them. We further show that an intermediate activation from one model can be used by another model's frozen upper layers to produce predictions. These results establish a stable, model-wise activation contract for reusable tools across compatible language models.
Chinese Translation
基于激活的工具通常与某一模型的本地隐空间紧密相关，这要求为每个新的语言模型重新构建或重新发现探针、稀疏自编码器和自然语言解释器。我们提出了一个通用激活总线（Universal Activation Bus），这是一个为兼容语言模型提供共同激活接口的框架。通过使用一小组源模型，我们学习了一个共享的稠密空间，并为每个模型配备一个轻量级的线性编码器-解码器适配器对。在源训练后，接口被冻结；新模型通过仅在未标记的匹配文本上拟合其适配器对来加入。最终的接口允许基于激活的工具在连接的模型之间共享，包括常见的探针和稀疏自编码器特征，以及对最初为不同模型训练的自然语言解释器的访问。在五个模型中，语义相关的文本在共享空间中形成一致的邻域，而新加入的模型有效地重用这些工具而无需重新训练。我们进一步展示了一个模型的中间激活可以被另一个模型的冻结上层用于生成预测。这些结果建立了一个稳定的、模型级的激活契约，以便在兼容的语言模型之间重用工具。
cs.AI / 198 / 2608.09537
verdi: retrieval is not transfer for continual world model optimization
verdi：检索不是持续世界模型优化的转移
Abstract
Foundation world models have made remarkable progress in planning, simulation, and embodied intelligence. However, optimizing a pretrained world model toward a user-specified objective remains difficult: each campaign typically rediscovers optimization strategies from scratch, and the resulting knowledge rarely transfers to the next model. Existing research agents automate the optimization loop but treat successful strategies as directly reusable recipes, without principled safeguards for when transfer is appropriate. We argue instead that retrieval is not transfer: a strategy validated on one model is at best an optimization hypothesis for another, and becomes transferable knowledge only after target-side experimental valida- tion. Guided by this principle, we propose VERDI , a continual framework for evidence-licensed world model optimization. VERDI characterizes each world model through shared inference-time probes to construct an Optimization Fin- gerprint, retrieves relevant prior experience as ranked hypotheses, and validates every candidate under a frozen target-side verifier before admitting it as reusable evidence; contradictions among nearby fingerprints further trigger probe evolution, continually refining the diagnostic representation itself. Experiments on Ctrl-World, the Cosmos family, and RoboCoin show that VERDI reduces search cost by 68%, GPU cost by 69%, and negative transfer from 0.34 to 0.06, while predicting transfer outcomes with 83% sign accuracy.
Chinese Translation
基础世界模型在规划、模拟和具身智能方面取得了显著进展。然而，针对用户指定目标优化预训练世界模型仍然困难：每个任务通常都需要从头开始重新发现优化策略，所获得的知识很少能够转移到下一个模型。现有的研究代理自动化了优化循环，但将成功的策略视为直接可重用的配方，而没有原则性地保障何时适合转移。我们认为，检索并不是转移：在一个模型上验证的策略充其量只是另一个模型的优化假设，只有在目标侧经过实验验证后才成为可转移的知识。基于这一原则，我们提出了VERDI，一个基于证据许可的世界模型优化的持续框架。VERDI通过共享推理时探针来表征每个世界模型，以构建优化指纹，检索相关的先前经验作为排名假设，并在接受其作为可重用证据之前，在冻结的目标侧验证器下验证每个候选者；相邻指纹之间的矛盾进一步触发探针演变，持续优化诊断表示本身。在Ctrl-World、Cosmos系列和RoboCoin上的实验表明，VERDI将搜索成本降低了68%，GPU成本降低了69%，负转移从0.34降低到0.06，同时以83%的符号准确率预测转移结果。
cs.AI / 199 / 2608.09555
Bidirectional Context Self-Distillation for Reinforcement Learning of Skill-Based LLM Agents
双向上下文自蒸馏用于基于技能的LLM代理的强化学习
Abstract
External natural-language skills provide large language model (LLM) agents with reusable and editable guidance for solving complex tasks. Yet their effectiveness depends not only on skill quality, but also on whether the policy can translate the provided guidance into appropriate actions. However, methods specifically designed to improve this skill-utilization ability remain largely underexplored. In practice, skill-based agents are commonly trained with reinforcement learning objectives centered on task-level rewards, which offer limited supervision and struggle to capture subtle differences in how effectively the policy uses the provided skills. We propose BCSD (Bidirectional Context Self-Distillation), a framework that combines self-distillation with reinforcement learning to train LLM agents to use external skills more effectively. Unlike prior self-distillation methods that rely on a single privileged context, BCSD evaluates each trajectory from two complementary skill-context views. The augmented view introduces higher-level Meta-Skill guidance, while the reduced view prunes general guidance to highlight task-specific skills. Their complementary token-level signals are combined to rescale the RL advantage. Experiments on ALFWorld and WebShop demonstrate that BCSD achieves the strongest overall performance across model scales, enabling agents to utilize external skills more effectively. Ablation studies further verify the complementary contributions of the augmented and reduced context views. Code will be released to ensure full reproducibility.
Chinese Translation
外部自然语言技能为大型语言模型（LLM）代理提供了可重用和可编辑的指导，以解决复杂任务。然而，它们的有效性不仅依赖于技能质量，还取决于策略能否将提供的指导转化为适当的行动。然而，专门设计用于提高这种技能利用能力的方法仍然未得到充分探索。在实践中，基于技能的代理通常通过以任务级奖励为中心的强化学习目标进行训练，这提供了有限的监督，并且难以捕捉策略在多大程度上有效地使用提供的技能之间的细微差别。我们提出了BCSD（双向上下文自蒸馏），这是一个将自蒸馏与强化学习相结合的框架，旨在训练LLM代理更有效地使用外部技能。与依赖单一特权上下文的先前自蒸馏方法不同，BCSD从两个互补的技能上下文视角评估每个轨迹。增强视角引入了更高层次的元技能指导，而简化视角则修剪一般指导，以突出任务特定技能。它们互补的标记级信号被结合以重新调整强化学习的优势。在ALFWorld和WebShop上的实验表明，BCSD在各模型规模中实现了最强的整体性能，使代理能够更有效地利用外部技能。消融研究进一步验证了增强和简化上下文视角的互补贡献。代码将发布以确保完全可重复性。
cs.AI / 200 / 2608.09574
The Politician, the Liar, and the Obedient Worker: Emerging Behavior of LLM Agents in Hierarchical Games
政治家、骗子与顺从的工作者：层级博弈中大型语言模型代理的行为演变
Abstract
LLMs are rapidly embedding themselves into daily life: drafting our emails, managing our schedules, and making decisions on our behalf. As they move from individual tools to participants in multi-agent organizations, an important question arises: do they reproduce the governance failures like free-riding, corruption, and entrenched leadership that plague human institutions? We introduce the Hierarchical Game (HG), a public goods game extended with managerial authority, democratic elections, and private communication. Testing six frontier models across twelve experiments that add institutions one at a time (speech, peers, government, wages, oversight, elections), we find distinct behavioral profiles: Qwen promises and lies (13.3\% broken promises); Grok refuses to cooperate on its own but becomes fully cooperative once a manager can punish it (16\%$\to$100\%); Claude and GPT-4o cooperate reliably at baseline. But honesty proves fragile. When the manager role comes with a salary, all models except GPT-4o start cutting private deals to win or keep the position. When punishment is made anonymous, honest models begin to cheat. When all agents share the same model family, the first elected manager stays in power indefinitely. Leadership change only happens in groups that mix different families.
Chinese Translation
大型语言模型（LLMs）正迅速融入我们的日常生活：起草电子邮件、管理日程以及代表我们做出决策。随着它们从单一工具转变为多代理组织中的参与者，一个重要问题随之而来：它们是否会重现人类机构中普遍存在的治理失效现象，如搭便车、腐败和根深蒂固的领导？我们引入了层级博弈（Hierarchical Game, HG），这是一种扩展了管理权、民主选举和私人沟通的公共物品博弈。通过在十二个实验中逐一添加制度（言论、同伴、政府、工资、监督、选举）来测试六个前沿模型，我们发现了不同的行为特征：Qwen 许诺并撒谎（13.3%的失信）；Grok 自行拒绝合作，但一旦有管理者能够惩罚它，它就会完全合作（16%→100%）；Claude 和 GPT-4o 在基线条件下可靠合作。然而，诚实性却显得脆弱。当管理者角色附带薪水时，除了 GPT-4o 之外的所有模型开始进行私下交易以赢得或保住职位。当惩罚变得匿名时，诚实模型开始作弊。当所有代理共享同一模型家族时，首位当选的管理者将无限期掌权。领导权的更替仅发生在混合不同家族的群体中。
cs.AI / 201 / 2608.09577
ElasticBack: Stealthy Conditional Backdoor in LLM-Agent Skills via Coupled Trigger-Rule Optimization
ElasticBack：通过耦合触发器-规则优化实现的隐蔽条件后门在LLM-Agent技能中的应用
Abstract
Agent skills, bundles of instructions and resources that an LLM agent loads on demand, form an emerging supply chain where a single poisoned skill can persistently compromise every agent that installs it. However, existing skill attacks either fire on every request or rely on fine-tuned weights or multiple skills, leaving a conditional and low-cost backdoor unexplored. In this work, we present ElasticBack, an effective conditional single-skill backdoor that plants a rule R in the skill document and a benign-looking trigger T in the user query, so the malicious payload fires only when both co-occur. ElasticBack binds the two sides through a trigger-as-switch construction, generating R via semantic-anchored rule injection. It then freezes R and evolves T against it with a stealth-constrained genetic search, so that effectiveness and stealth are optimized, keeping the backdoor weight-free and dormant on benign inputs. Extensive experiments across three target behaviors (50 skills each) and four agent LLMs show that ElasticBack attains a high attack success rate at a near-zero false-positive rate with preserved clean accuracy, transfers across models, and evades deployment-time defenses. These results motivate stronger defenses for the skill supply chain.
Chinese Translation
代理技能是LLM代理按需加载的一系列指令和资源，形成了一种新兴的供应链，其中单个被污染的技能可以持续危害每个安装它的代理。然而，现有的技能攻击要么在每个请求上触发，要么依赖于微调的权重或多个技能，导致条件性和低成本的后门尚未被探索。在本研究中，我们提出了ElasticBack，一种有效的条件单技能后门，它在技能文档中植入规则R，并在用户查询中植入一个看似良性的触发器T，因此恶意负载仅在两者同时出现时触发。ElasticBack通过触发器作为开关的构造将两者绑定，通过语义锚定的规则注入生成R。然后，它冻结R并针对其演化T，采用隐蔽约束的遗传搜索，从而优化有效性和隐蔽性，使后门在良性输入上保持无权重和休眠状态。在三个目标行为（每个行为50个技能）和四个代理LLM上的广泛实验表明，ElasticBack在保持清晰准确度的同时，达到了高攻击成功率和近零的误报率，并且能够跨模型转移，规避部署时的防御。这些结果促使我们对技能供应链提出更强的防御措施。
cs.AI / 202 / 2608.09580
CoRCi: Cross-Reconstruction of Coherent Interests Modeling in Cross-Domain Sequential Recommendation
CoRCi：跨域序列推荐中的一致兴趣建模的交叉重构
Abstract
Cross-Domain Sequential Recommendation (CDSR) aims to alleviate data sparsity by transferring dynamic user interests across related domains. A key challenge lies in effectively bridging these domains. In single-domain modeling, models cannot distinguish between domain-specific and domain-invariant interests. Recent methods merge domain-specific sequences chronologically into a mixed-domain sequence to capture domain-invariant knowledge. However, they typically deploy separate encoders for the mixed-domain sequence and train them with per-domain loss aggregation. This workflow magnifies inter-domain discrepancies and disrupts domain-invariant interest coherence, especially when query target pairs in Seq2Seq originate from different domains. In this paper, we present CoRCi (Cross-Reconstruction for Coherent Interest), a dual-target CDSR framework that tackles these drawbacks. Specifically, CoRCi proposes a Cross-Reconstruction approach that generates mixed-domain representations directly from pre-encoded specific-domain representations via cross-attention. The generated representations are then trained using a single, sequence-level, domain-agnostic loss to preserve the coherence of domain-invariant interests. To further suppress domain discrepancies in mixed-domain modeling, CoRCi introduces FocalNCE, which embeds Focal Loss into the preceding mixed-domain InfoNCE objective. The new loss assigns higher penalties to negatives drawn from the same domain as the query, thereby strengthening domain-invariant alignment. Extensive experiments on four real-world datasets demonstrate that CoRCi consistently outperforms state-of-the-art CDSR counterparts, achieving statistically significant gains across all metrics.
Chinese Translation
跨域序列推荐（CDSR）旨在通过在相关领域之间转移动态用户兴趣来缓解数据稀疏性。一个关键挑战在于有效地桥接这些领域。在单域建模中，模型无法区分领域特定兴趣和领域不变兴趣。最近的方法将领域特定序列按时间顺序合并为混合域序列，以捕捉领域不变知识。然而，它们通常为混合域序列部署单独的编码器，并通过每个领域的损失聚合进行训练。这种工作流程放大了领域间的差异，并破坏了领域不变兴趣的一致性，特别是当 Seq2Seq 中的查询目标对来自不同领域时。在本文中，我们提出了 CoRCi（一致兴趣的交叉重构），这是一个双目标的 CDSR 框架，旨在解决这些缺陷。具体而言，CoRCi 提出了一个交叉重构方法，通过交叉注意力直接从预编码的特定领域表示生成混合域表示。生成的表示随后使用单一的序列级、领域无关损失进行训练，以保持领域不变兴趣的一致性。为了进一步抑制混合域建模中的领域差异，CoRCi 引入了 FocalNCE，将 Focal Loss 嵌入到前面的混合域 InfoNCE 目标中。新的损失对来自与查询相同领域的负样本施加更高的惩罚，从而增强领域不变对齐。在四个真实世界数据集上的大量实验表明，CoRCi 在所有指标上始终优于最先进的 CDSR 对手，取得了统计显著的提升。
cs.AI / 203 / 2608.09586
ICM Out! Better Tournament Strategy from Computed Continuations, vs. Solvers and LLMs
ICM出局！通过计算延续优化的更优锦标赛策略，与求解器和大型语言模型的比较
Abstract
The Independent Chip Model (ICM) converts tournament chips into reference prize equity, and policies are routinely constructed against those values. Because ICM reads only stack sizes, it omits action order, blind obligations, and seat rotation, and it does not price the elimination pressure a big stack puts on the short stacks it can bust. Those omissions can alter the successor-state contrasts that determine a move. We introduce Strategic-Continuation Optimization (SCO), a policy-construction method that enumerates current-hand outcomes, maps them to successor states, prices those states with continuation values computed from the finite tournament model, and optimizes and freezes the resulting current-hand policy. The fixed-ICM comparison policy changes one thing only: the same optimizer solves the same game with successor states priced by analytic ICM, so the two policies differ only through that pricing. We evaluate the resulting policies in a three-player jam/fold tournament with a \$1M prize pool. Relative to the frozen strategic-continuation benchmark, analytic ICM has \$9{,}036 mean absolute value error across all 2,838 state--seat entries. That value error rewrites the ranges it prices: measured against each decision point's own fixed-ICM jam range, SCO moves the jam frequency by an average of 14.08\%. To price those different moves, we compare all 946 states and three policy owners while changing only the focal policy and holding both opponents and the continuation evaluator fixed. The policy produced by SCO earns \$214.33 more prize equity per hand on average and is favored in 2,433 of 2,838 matched units. The ordering survives replacing the solver-built opponent with two LLMs and with a family of non-modeling threshold players. This value-to-policy-to-cost chain shows directly when ICM becomes an inadequate objective for tournament strategy construction.
Chinese Translation
独立筹码模型（Independent Chip Model, ICM）将锦标赛筹码转换为参考奖金权益，策略通常是针对这些价值构建的。由于ICM仅考虑筹码堆大小，它忽略了行动顺序、盲注义务和座位轮换，并且没有考虑大筹码对短筹码施加的淘汰压力。这些遗漏可能会改变决定行动的后继状态对比。我们引入了战略延续优化（Strategic-Continuation Optimization, SCO），这是一种策略构建方法，它列举当前手牌的结果，将其映射到后继状态，利用有限锦标赛模型计算的延续价值为这些状态定价，并优化和固定生成的当前手牌策略。固定ICM比较策略仅改变一项内容：相同的优化器使用解析ICM定价的后继状态解决相同的游戏，因此两种策略仅通过该定价有所不同。我们在一个有3名玩家参与、奖金池为100万美元的全押/弃牌锦标赛中评估了生成的策略。相对于固定的战略延续基准，解析ICM在所有2,838个状态-座位条目中具有9,036美元的平均绝对价值误差。该价值误差重写了其定价范围：与每个决策点自身的固定ICM全押范围相比，SCO平均将全押频率移动了14.08%。为了对这些不同的行动进行定价，我们在只改变焦点策略并保持两个对手和延续评估器不变的情况下比较了所有946个状态和三位策略拥有者。SCO生成的策略平均每手牌赚取214.33美元的奖金权益，并在2,838个匹配单位中有2,433个单位受到青睐。这一排序在用两个大型语言模型替换求解器构建的对手以及一系列非建模阈值玩家时仍然有效。这一价值-政策-成本链直接表明，当ICM成为锦标赛策略构建的不充分目标时。
cs.AI / 204 / 2608.09595
From Sweep to Seam: Interleaved Cross-Block Post-Training Quantization
从扫掠到接缝：交错跨块后训练量化
Abstract
Compressing large language models to two bits or fewer is increasingly feasible through block-wise post-training quantization; cross-block variants reconstruct neighboring Transformer blocks within a moving window. In the fixed two-block setting studied here, the matched sequential baseline moves this window through the network once, so errors introduced early in the sweep are not revisited. We propose Interleaved Cross-Block Quantization (ICBQ), a scheduling modification that revisits the boundary pair between consecutive chunks. Each seam pair is refined twice: first at the end of one chunk and again at the start of the next. The method retains the local two-block objective and reuses the calibration inputs of existing block-wise PTQ pipelines. Under stated local contraction and smoothness assumptions, we derive a depth-wise upper-bound comparison in which seam revisits multiply the propagated term while the residual remains bounded independently of depth. In the reported experiments, ICBQ reduces ternary-quantization perplexity relative to the matched Sequential CBQ baseline, yields finite perplexity in configurations where the baseline has severe degradation, and can also be used with 3-bit and 2-bit GPTQ.
Chinese Translation
通过块级后训练量化，将大型语言模型压缩到两位或更少的位数变得越来越可行；跨块变体在移动窗口内重构相邻的Transformer块。在这里研究的固定两块设置中，匹配的顺序基线仅在网络中移动一次该窗口，因此在扫掠早期引入的错误不会被重新访问。我们提出了交错跨块量化（Interleaved Cross-Block Quantization, ICBQ），这是一种调度修改，重新访问连续块之间的边界对。每个接缝对被精炼两次：第一次在一个块的末尾，第二次在下一个块的开始。该方法保留了局部两块目标，并重用了现有块级PTQ管道的校准输入。在所述的局部收缩和平滑性假设下，我们推导出一个深度上界比较，其中接缝的重访使传播项成倍增加，而残差则独立于深度保持有界。在报告的实验中，ICBQ相对于匹配的顺序CBQ基线降低了三元量化的困惑度，在基线严重退化的配置中产生有限的困惑度，并且还可以与3位和2位的GPTQ一起使用。
cs.AI / 205 / 2608.09622
Adaptive Sequential Test Planning for Multi-Mechanism Reliability Qualification via Bayesian Monte Carlo Tree Search
基于贝叶斯蒙特卡洛树搜索的多机制可靠性鉴定自适应序列测试规划
Abstract
Reliability qualification of advanced semiconductor devices requires sequential stress decisions that balance characterization objectives against multiple competing failure mechanisms. Current practice relies on static test plans derived from population-level acceleration models, which cannot adapt to per-unit variability or real-time degradation observations. This paper presents a closed-loop adaptive test planning framework that formulates reliability qualification as a partially observable sequential decision problem and solves it using Monte Carlo tree search for seed-action simulators (MCTS-SA) coupled with extended Kalman filter (EKF) belief-state estimation. The framework models stochastic, per-device variability in bias temperature instability (BTI), electromigration (EM), and time-dependent dielectric breakdown (TDDB), and treats stress selection as a constrained sequential optimization, i.e., to maximize the probability of successful degradation characterization while respecting catastrophic failure constraints. Under the experimental assumptions used here (discrete stress actions, proxy damage observability, and cumulative degradation without recovery), we believe this to be a novel application of tree-search-based adaptive test planning to multi-mechanism reliability qualification. Across 5,000 planning iterations, the characterization yield (CY) improves from 20% in the first 500 iterations to over 54% in the final 500, with 39% cumulative success, while the best successful test sequence terminates with EM and TDDB damage fractions DEM=0.564 and DTDDB=0.537, well within safety margins. These results demonstrate that sequential Bayesian planning can synthesize damage-aware test policies that significantly outperform non-adaptive strategies for reliability qualification under competing failure modes.
Chinese Translation
先进半导体器件的可靠性鉴定需要顺序应力决策，以平衡表征目标与多种竞争失效机制之间的关系。目前的做法依赖于从总体加速模型中导出的静态测试计划，这些计划无法适应单元间的变异性或实时降解观测。本文提出了一种闭环自适应测试规划框架，将可靠性鉴定表述为一个部分可观测的顺序决策问题，并利用蒙特卡洛树搜索（Monte Carlo Tree Search, MCTS）结合扩展卡尔曼滤波器（Extended Kalman Filter, EKF）信念状态估计来解决该问题。该框架对偏置温度不稳定性（Bias Temperature Instability, BTI）、电迁移（Electromigration, EM）和时间依赖性介质击穿（Time-Dependent Dielectric Breakdown, TDDB）中的随机、每个器件的变异性进行了建模，并将应力选择视为一个受限的顺序优化问题，即在遵循灾难性失效约束的同时，最大化成功降解表征的概率。在这里使用的实验假设下（离散应力动作、代理损伤可观测性和无恢复的累积降解），我们认为这是基于树搜索的自适应测试规划在多机制可靠性鉴定中的一种新颖应用。在5000次规划迭代中，表征产率（Characterization Yield, CY）从前500次的20%提高到最后500次的54%以上，累计成功率为39%，而最佳成功测试序列的EM和TDDB损伤比例分别为DEM=0.564和DTDDB=0.537，均在安全边际内。这些结果表明，顺序贝叶斯规划能够合成损伤感知测试策略，显著优于在竞争失效模式下的非自适应策略。
cs.AI / 206 / 2608.09629
Rethinking Self-Evolving Agents: Do We Still Need Prescribed Optimization Pipelines?
重新思考自我进化代理：我们还需要规定的优化流程吗？
Abstract
Self-evolving agents are usually built around prescribed optimization pipelines: the framework decides how to gather evidence, revise a persistent artifact, select candidates, and stop. We ask whether this task-specific procedure remains necessary when a frontier model acts as the optimizer. We introduce Open-Ended Optimization (OEO), which keeps the objective, permitted interactions, resource budget, data boundary, and evaluation fixed while allowing the optimizer to compose the improvement process online. We compare OEO with two complementary prescribed approaches: SkillOpt, a staged pipeline with bounded edits, and GEPA, a reflective evolutionary search. Across 14 head-to-head comparisons over 8 benchmark-target-model settings, GPT-5.5-driven OEO records 12 wins, 1 tie, and 1 narrow loss of 0.21 percentage points. It uses a median 34.3 percent of SkillOpt's configured target-interaction token budget. A one-shot, zero-interaction control shows that the gains are not explained by a single prior-driven rewrite. However, delegation has a capability boundary: SkillOpt outperforms OEO with a medium optimizer, and a weak optimizer cannot operate through the unchanged OEO interface. In the fully instrumented OEO-SkillOpt pair, trajectory analysis further shows that prescription changes how optimization proceeds more consistently than it changes final behavior. Together, these findings recast prescribed pipelines as capability-dependent scaffolding: essential constraints remain external, but a sufficiently capable optimizer can compose the route from measurable feedback to persistent improvement.
Chinese Translation
自我进化代理通常围绕规定的优化流程构建：该框架决定如何收集证据、修订持久性工件、选择候选者以及何时停止。我们探讨当前沿模型作为优化器时，这一特定任务的程序是否仍然必要。我们引入开放式优化（Open-Ended Optimization, OEO），该方法在固定目标、允许的交互、资源预算、数据边界和评估的同时，允许优化器在线组合改进过程。我们将OEO与两种互补的规定方法进行比较：SkillOpt，一个具有有限编辑的分阶段管道，以及GEPA，一个反思性进化搜索。在8个基准目标模型设置下进行的14次正面对比中，基于GPT-5.5的OEO记录了12次胜利、1次平局和1次以0.21个百分点的微弱失利。它使用了SkillOpt配置的目标交互令牌预算的中位数34.3%。一次性、零交互的控制实验表明，收益并不能仅通过单一的先前驱动重写来解释。然而，委托具有能力边界：在中等优化器的情况下，SkillOpt的表现优于OEO，而弱优化器无法通过未改变的OEO接口进行操作。在完全仪器化的OEO-SkillOpt配对中，轨迹分析进一步表明，规定改变了优化的进行方式，比改变最终行为更为一致。综合这些发现，我们将规定的管道重新定义为依赖能力的支架：基本约束仍然是外部的，但一个足够有能力的优化器可以从可测反馈到持久改进的路径进行组合。
cs.AI / 207 / 2608.09638
Avalon-ToM-Bench: Evaluating Fine-Grained Theory of Mind via Asymmetric Game Mechanics
Avalon-ToM-Bench：通过不对称游戏机制评估细粒度的心智理论
Abstract
Theory of Mind (ToM) is essential for agent interactions, yet existing evaluations either rely on static scenarios that oversimplify mental-state reasoning or interactive settings that provide limited diagnostic insight. We present Avalon-ToM-Bench, a fine-grained benchmark that operationalizes ToM through the asymmetric-information mechanics of The Resistance: Avalon. Rather than evaluating end-to-end gameplay, it decomposes ToM into a 2$\times$2 taxonomy -- epistemic versus motivational reasoning crossed with inference versus action -- using human-crafted, perspective-constrained queries. Benchmarking 28 LLMs reveals three insights: 1) Reasoning, not knowledge. Models show strong game-rule comprehension but markedly weaker ToM abilities, isolating failures to social reasoning rather than missing domain knowledge. 2) Expression, not representation. Mechanistic analyses via linear probing and activation steering show that models frequently represent correct mental-state inferences in their hidden states but fail to express them during generation -- linear probes recover 77-82% accuracy versus 62-70% from the models' own chain-of-thought. 3) Policy, not deliberation. Dedicated reasoning training yields substantial improvements whereas test-time chain-of-thought provides only marginal gains (+11.0 versus +1.1 points on average), suggesting that robust ToM depends on a learned reasoning policy rather than increased inference-time deliberation.
Chinese Translation
心智理论（Theory of Mind, ToM）对于代理之间的互动至关重要，但现有评估要么依赖于简化心理状态推理的静态场景，要么依赖于提供有限诊断洞察的互动环境。我们提出了Avalon-ToM-Bench，这是一个通过《抵抗：亚瓦隆》（The Resistance: Avalon）中的不对称信息机制来操作化ToM的细粒度基准。该基准并非评估端到端的游戏玩法，而是将ToM分解为一个2×2的分类法——知识性推理与动机性推理交叉于推理与行动——使用人类设计的、受视角限制的查询。对28个大型语言模型（LLMs）的基准测试揭示了三个见解：1）推理，而非知识。模型表现出强大的游戏规则理解能力，但ToM能力明显较弱，失败主要集中在社会推理上，而非缺乏领域知识。2）表达，而非表征。通过线性探测和激活引导的机制分析表明，模型在其隐藏状态中经常正确表征心理状态推理，但在生成过程中未能表达这些推理——线性探测的准确率为77-82%，而模型自身的思维链生成的准确率为62-70%。3）策略，而非深思熟虑。专门的推理训练带来了显著的改善，而测试时的思维链仅提供了边际收益（平均+11.0与+1.1分），这表明稳健的ToM依赖于学习到的推理策略，而非增加推理时间的深思熟虑。
cs.AI / 208 / 2608.09654
Hallucination-Free GUI Grounding via Regression-Free Layout-Aware Matching
无幻觉的图形用户界面定位：无回归的布局感知匹配
Abstract
GUI agents are shifting from metadata-dependent large language models to purely visual multimodal large language models (MLLMs) that operate directly on screenshots. The core task, GUI grounding, requires translating abstract user instructions into precise element coordinates. This task faces a persistent dual obstacle: conventional grounding models lack the semantic richness to interpret abstract instructions, while end-to-end MLLMs suffer from coordinate hallucinations caused by deficient fine-grained perception. We propose a regression-free framework where a frozen MLLM performs instruction parsing and a dedicated grounding model handles precise localization without learning any coordinate regression. A frozen MLLM first elaborates the abstract instruction into a structured visual description rich in layout cues. These descriptions are then fed to a novel Layout-Aware GUI Grounding Model, which performs regression-free localization by matching against layout-prior candidates, inherently suppressing hallucinations and avoiding expensive fine-tuning. The grounding model is trained with only Text/Icon binary labels, requiring no coordinate regression parameters. On ScreenSpot-Pro, our method achieves over 20% improvement in grounding accuracy over end-to-end systems; on Mind2Web, it raises success rate and element selection rate by more than 15%. These results demonstrate that decoupling instruction understanding from layout-aware localization effectively resolves the core challenges of GUI interaction.
Chinese Translation
图形用户界面（GUI）代理正从依赖元数据的大型语言模型转向直接在屏幕截图上操作的纯视觉多模态大型语言模型（MLLMs）。核心任务——GUI 定位，要求将抽象用户指令转化为精确的元素坐标。该任务面临着持续的双重障碍：传统的定位模型缺乏解释抽象指令的语义丰富性，而端到端的 MLLMs 由于细粒度感知不足而遭受坐标幻觉。我们提出了一种无回归框架，其中一个冻结的 MLLM 负责指令解析，而一个专用的定位模型处理精确定位，无需学习任何坐标回归。冻结的 MLLM 首先将抽象指令详细阐述为包含布局线索的结构化视觉描述。这些描述随后被输入到一种新颖的布局感知 GUI 定位模型中，该模型通过与布局优先候选进行匹配来执行无回归定位，内在地抑制幻觉并避免昂贵的微调。定位模型仅使用文本/图标二元标签进行训练，无需坐标回归参数。在 ScreenSpot-Pro 上，我们的方法在定位准确性上比端到端系统提高了超过 20%；在 Mind2Web 上，它的成功率和元素选择率提高了超过 15%。这些结果表明，将指令理解与布局感知定位解耦有效地解决了 GUI 交互的核心挑战。
cs.AI / 209 / 2608.09666
Open Evaluation Agent: Efficient and Promptable Evaluation of Visual Generative Models
开放评估代理：高效且可提示的视觉生成模型评估
Abstract
Recent advances in visual generative models have enabled high-quality image and video generation, but evaluating these models often demands sampling hundreds or thousands of images or videos, which is computationally expensive. Existing evaluation methods also rely on rigid pipelines that overlook specific user needs and provide numerical results without clear explanations. Mimicking how humans quickly form impressions of a model's capabilities from only a few samples, we propose the Evaluation Agent framework, which employs human-like strategies for efficient, dynamic, multi-round evaluations, offering detailed, user-tailored analyses. Given a natural-language evaluation request, the agent decomposes it into sub-aspects, generates targeted prompts, samples images or videos from the evaluated model, invokes suitable evaluation tools, and iteratively updates its plan from the observed evidence, covering both predefined benchmark dimensions and open-ended user concerns. The framework is thus efficient, promptable, explainable, and scalable across models and tools. Experiments show that Evaluation Agent reduces evaluation time to 10% of traditional methods while delivering comparable results. We further introduce Open Evaluation Agent (Open-EA) by constructing EA-CoT-10K, a corpus of history-conditioned step-level instruction-tuning records derived from multi-round evaluation rollouts, and training EA-3B from Qwen2.5-3B-Instruct as a local planning backbone that preserves the structured reasoning, tool invocation, and summary protocol of the API-based agent while reducing dependence on proprietary backbones. Experiments validate the API-based agent on established T2I/T2V benchmarks and open-ended queries, and evaluate Open-EA on four in-domain and three out-of-domain T2V generator families, showing partial cross-family transfer of the learned policy.
Chinese Translation
近年来，视觉生成模型的进步使得高质量图像和视频生成成为可能，但评估这些模型通常需要采样数百或数千张图像或视频，这在计算上是昂贵的。现有的评估方法依赖于僵化的流程，忽视了特定用户的需求，并提供没有明确解释的数值结果。我们模仿人类如何通过少量样本迅速形成对模型能力的印象，提出了评估代理框架（Evaluation Agent），该框架采用类人策略进行高效、动态的多轮评估，提供详细且用户定制的分析。给定自然语言评估请求，代理将其分解为子方面，生成针对性的提示，从被评估模型中采样图像或视频，调用合适的评估工具，并根据观察到的证据迭代更新其计划，覆盖预定义的基准维度和开放式用户关注点。因此，该框架在模型和工具之间是高效的、可提示的、可解释的和可扩展的。实验表明，评估代理将评估时间缩短至传统方法的10%，同时提供可比的结果。我们进一步通过构建EA-CoT-10K，一个基于历史条件的步骤级指令调优记录语料库，推出开放评估代理（Open Evaluation Agent，Open-EA），该语料库源自多轮评估的回放，并训练EA-3B作为本地规划骨干，保留API基础代理的结构化推理、工具调用和总结协议，同时减少对专有骨干的依赖。实验验证了基于API的代理在已建立的T2I/T2V基准和开放式查询上的表现，并在四个领域内和三个领域外的T2V生成器系列上评估了Open-EA，显示出学习策略的部分跨系列迁移。
cs.AI / 210 / 2608.09685
Adaptive Semantic Capacity Allocation for Parallel Generative Recommendation
并行生成推荐的自适应语义容量分配
Abstract
Autoregressive semantic ID recommenders are constrained by expensive beam-search decoding, which limits the practical length of item identifiers. Parallel generation methods alleviate this bottleneck by predicting all semantic ID tokens simultaneously, enabling longer IDs. However, existing semantic ID methods still rely on manually predefined and homogeneous ID structures, where both the number of semantic slots and the codebook size of each slot are treated as fixed hyperparameters. This ignores the heterogeneous capacity demands of different semantic subspaces and may allocate prediction capacity to slots with limited utility. We show that uniformly expanding semantic slots can provide limited gains, indicating redundant capacity in homogeneous semantic IDs. We propose InforID, a lightweight adaptive semantic target construction framework for parallel generative recommendation. InforID allocates a fixed capacity budget across candidate semantic slots, thereby jointly determining the effective ID length and slot-specific codebook sizes. Experiments demonstrate improved recommendation accuracy under comparable capacity budgets while preserving one-step parallel prediction.
Chinese Translation
自回归语义 ID 推荐系统受到昂贵的束搜索解码的限制，这限制了项目标识符的实际长度。并行生成方法通过同时预测所有语义 ID 令牌来缓解这一瓶颈，从而实现更长的 ID。然而，现有的语义 ID 方法仍然依赖于手动预定义的同质 ID 结构，其中语义槽的数量和每个槽的词汇表大小都被视为固定的超参数。这忽视了不同语义子空间的异质容量需求，可能会将预测容量分配给效用有限的槽。我们表明，均匀扩展语义槽只能提供有限的收益，表明同质语义 ID 中存在冗余容量。我们提出了 InforID，一种轻量级自适应语义目标构建框架，用于并行生成推荐。InforID 在候选语义槽之间分配固定的容量预算，从而共同确定有效的 ID 长度和槽特定的词汇表大小。实验表明，在可比的容量预算下，推荐准确性得到了提高，同时保持了一步并行预测。
cs.AI / 211 / 2608.09696
Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
模型发现代理：基于大语言模型的贝叶斯实验设计用于数据高效发现机制世界模型
Abstract
Predicting the answer to interventional ``what if'' questions --- the outcome of an action never taken --- requires a \emph{mechanistic}, causal model, not a curve fit; and learning such a model requires \emph{experiments}, because passive data leaves its mechanisms unidentified. Experiments are expensive, so the central problem is \emph{data efficiency}. We present the Model Discovery Agent (MDA), which couples a large language model (LLM), used as a \emph{proposer} of candidate structures, with standard Bayesian machinery --- sequential Monte Carlo (SMC) for parameter and structure posteriors, simulation-based inference (SBI) for intractable likelihoods, and value-of-information (VoI) for experiment design --- to discover latent mechanistic world models from few interventions. MDA operates in the M-open setting: when the truth lies outside the current hypothesis class, a predictive check flags the inadequacy and the proposer expands the hypothesis space with a new model whose parameters are then identified by designed experiments. We show that \emph{discovery and design reinforce}: the design step identifies the mechanism the discovery step proposes, and the identified mechanism improves predictions, enabling further discoveries from the remaining unexplained residuals. On three different benchmarks --- covering physics (\DPbench, \citep{wiemann2026discoverphysics}), chemistry (\CHEMbench, \citep{kabra2026autoscilab}) and biology (\HHbench, a new partially observed single-neuron electrophysiology benchmark we create) --- we show that MDA sets a new SOTA in terms of data-efficient model learning and reliable interventional forecasting ability.
Chinese Translation
预测干预性“如果……会怎样”问题的答案——即未采取行动的结果——需要一个 extit{机制性}的因果模型，而不是简单的曲线拟合；学习这样的模型需要 extit{实验}，因为被动数据无法识别其机制。实验成本高昂，因此核心问题是 extit{数据效率}。我们提出了模型发现代理（Model Discovery Agent, MDA），它将大型语言模型（Large Language Model, LLM）作为候选结构的 extit{提议者}，与标准贝叶斯工具相结合——使用序列蒙特卡洛（Sequential Monte Carlo, SMC）进行参数和结构后验推断，基于仿真的推理（Simulation-Based Inference, SBI）处理难以计算的似然，利用信息价值（Value of Information, VoI）进行实验设计——以从少量干预中发现潜在的机制世界模型。MDA在M-开放设置中运行：当真相超出当前假设类时，预测检查会标记不充分性，提议者则通过一个新的模型扩展假设空间，其参数随后通过设计实验进行识别。我们展示了 extit{发现与设计的相互强化}：设计步骤识别发现步骤所提议的机制，而识别的机制改善了预测，从而使得从剩余未解释的残差中进行进一步发现成为可能。在三个不同的基准测试中——涵盖物理学（ extit{DPbench}, extcite{wiemann2026discoverphysics}）、化学（ extit{CHEMbench}, extcite{kabra2026autoscilab}）和生物学（ extit{HHbench}, 我们创建的新部分观察单神经元电生理基准）——我们展示了MDA在数据高效模型学习和可靠干预预测能力方面设定了新的最先进水平（SOTA）。
cs.AI / 212 / 2608.09703
Matryoshka Language Model Suites
马特ryoshka语言模型套件
Abstract
Training a language model suite classically requires training each model separately and serving them independently. We improve both training and inference efficiency by stacking sub-models of increasing size into a single nested architecture trained end-to-end. This Matryoshka training framework reduces the total parameter count of the suite, enables low-cost distillation from the largest to all smaller sub-models at every training step, and is well-suited for speculative decoding as the draft model is contained within the verifier. We validate our approach by training a Matryoshka suite comprising 500M, 1.5B, and 3B sub-models. Our suite is on par with independently trained baselines on benchmark performance and validation and out-of-domain perplexities, while using 36% less training compute and improving the throughput of speculative decoding by 14-26%. We also ablate key architectural choices, offering guidance for building strong Matryoshka LM suites.
Chinese Translation
传统上，训练一个语言模型套件需要分别训练每个模型并独立服务。我们通过将逐渐增大的子模型堆叠成一个单一的嵌套架构并进行端到端训练，来提高训练和推理效率。该马特ryoshka训练框架减少了套件的总参数数量，使得在每个训练步骤中能够以低成本从最大的子模型蒸馏到所有较小的子模型，并且非常适合于推测解码，因为草稿模型包含在验证器中。我们通过训练一个包含500M、1.5B和3B子模型的马特ryoshka套件来验证我们的方法。我们的套件在基准性能、验证和领域外困惑度方面与独立训练的基线相当，同时使用了36%的较少训练计算，并将推测解码的吞吐量提高了14-26%。我们还对关键的架构选择进行了消融实验，为构建强大的马特ryoshka语言模型套件提供了指导。
cs.AI / 213 / 2608.09763
Second-Order Muon Done Right: A Principled Marriage of Spectral Geometry and Curvature
第二阶μ子：谱几何与曲率的原则性结合
Abstract
Muon's polar update is exact for an unweighted spectral geometry. We introduce GO-MUON, which uses a matched data-dependent geometry and reuses it across several optimization steps. Conditioned on any positive-definite left and right maps, its raw update exactly solves the corresponding weighted spectral oracle; this statement is independent of how the maps are estimated or how recently they were refreshed. For softmax cross-entropy, we quantify when the observed-label backward factor approaches the model Fisher and generalized Gauss--Newton factor. We also show that four-step refresh nearly preserves the tracking delay of slowly changing geometry while increasing stationary factor noise, making lazy geometry a compute--statistics tradeoff rather than a denoising mechanism.
Chinese Translation
μ子的极化更新对于无权谱几何是精确的。我们引入了GO-MUON，它使用匹配的数据依赖几何，并在多个优化步骤中重用该几何。基于任何正定的左映射和右映射，其原始更新精确解决了相应的加权谱oracle；这一说法与映射的估计方式或最近的刷新时间无关。对于softmax交叉熵，我们量化了观察标签的反向因子何时接近模型的Fisher因子和广义Gauss-Newton因子。我们还展示了四步刷新几乎保持了缓慢变化几何的跟踪延迟，同时增加了平稳因子的噪声，使得懒惰几何成为计算与统计之间的权衡，而非去噪机制。
cs.AI / 214 / 2608.09775
AirFlow: Context Preserving and Multi-Rate State Modeling for Air Quality Forecasting
AirFlow：用于空气质量预测的上下文保留和多速率状态建模
Abstract
Accurate air quality forecasting is essential for public health and urban environmental management, but remains challenging because pollutant channels differ in periodicity and distribution drift, while their concentration trajectories contain both multi-scale dependencies and rapid changes. Recent methods have improved spatial dependency learning and meteorological covariate modeling. However, pollutant channels are still passed through the same normalization rule and temporal backbone, using a shared latent representation for channel-specific distributions and changes at different rates. To address this limitation, we propose AirFlow, a pollutant-aware dual-stream framework that operates on station multivariate observations without additional graph propagation or predefined signal decomposition. Specifically, AirFlow designs two novel blocks: (1) a statistic-guided normalization routing mechanism that selects a normalization path for each pollutant according to its 24-hour autocorrelation and distribution drift; and (2) a hierarchical dual-stream state model that combines multi-scale state space propagation with learnable response coefficients, where gated bidirectional cross-attention exchanges information and adaptively fuses the resulting representations. Experiments on real-world data from multiple cities show that AirFlow achieves the best performance in 34 of 36 metrics comparisons, with reductions of up to 11.11% root mean square error over the state-of-the-art baseline. AirFlow also requires only 0.0483M parameters and 0.0215G FLOPs, achieving high forecasting accuracy with low computational overhead.
Chinese Translation
准确的空气质量预测对公共健康和城市环境管理至关重要，但由于污染物通道在周期性和分布漂移方面存在差异，同时其浓度轨迹包含多尺度依赖性和快速变化，因此仍然具有挑战性。最近的方法改善了空间依赖学习和气象协变量建模。然而，污染物通道仍然通过相同的归一化规则和时间骨架进行处理，使用共享的潜在表示来应对通道特定的分布和不同速率的变化。为了解决这一限制，我们提出了AirFlow，一种污染物感知的双流框架，能够在没有额外图传播或预定义信号分解的情况下，处理站点的多变量观测。具体而言，AirFlow设计了两个新颖的模块：（1）一个统计引导的归一化路由机制，根据每种污染物的24小时自相关和分布漂移选择归一化路径；（2）一个层次化双流状态模型，将多尺度状态空间传播与可学习的响应系数相结合，其中门控双向交叉注意力交换信息并自适应融合生成的表示。在来自多个城市的真实数据上的实验表明，AirFlow在36个指标比较中取得了34个指标的最佳性能，相较于最先进的基线模型，均方根误差减少了多达11.11%。AirFlow还仅需0.0483M参数和0.0215G FLOPs，以低计算开销实现高预测精度。
cs.AI / 215 / 2608.09790
CARD: Controlled Agentic Reddit Discussions for Credit Card Simulation
CARD：用于信用卡模拟的受控代理Reddit讨论
Abstract
Online credit card discussions provide a natural setting for studying how consumers communicate about financial products. Simulating these discussions requires more than just generating individual comments, the generated threads should also match how real users express themselves and interact with others. We introduce CARD, a framework for generating realistic credit card discussion threads. Given a credit card post and its matched real thread, CARD uses non-verbatim guidance on reply structure, comment function, stance, tone, and conversational variation. A planner organizes these controls, a writer generates the discussion, and a calibration loop updates comments' populations that contribute to differences between the generated and real thread distributions. We evaluate CARD on real Reddit credit card discussions using lexical, semantic, behavioral, and structural metrics. CARD matches the distributions of real credit card discussions better than simulation baselines across multiple LLMs and also demonstrates smaller effect sizes and distribution distances across metrics. These results show that structured planning and targeted revision can generate the realism of simulated credit card discussions.
Chinese Translation
在线信用卡讨论为研究消费者如何交流金融产品提供了自然的环境。模拟这些讨论不仅仅需要生成单独的评论，生成的讨论线程还应与真实用户的表达方式和互动方式相匹配。我们提出了CARD，一个用于生成真实信用卡讨论线程的框架。给定一个信用卡帖子及其匹配的真实线程，CARD在回复结构、评论功能、立场、语调和对话变化上使用非逐字指导。一个规划者组织这些控制因素，一个生成者生成讨论，而一个校准循环则更新评论的群体，以便缩小生成线程与真实线程分布之间的差异。我们使用词汇、语义、行为和结构指标对CARD在真实Reddit信用卡讨论中的表现进行了评估。CARD在多个大型语言模型（LLMs）中比模拟基线更好地匹配了真实信用卡讨论的分布，并且在各项指标上表现出更小的效应大小和分布距离。这些结果表明，结构化的规划和有针对性的修订能够生成模拟信用卡讨论的真实感。
cs.AI / 216 / 2608.09836
Mismatch Matters: On-Policy Distillation Beyond Token Agreement
不匹配的重要性：超越标记一致性的在线策略蒸馏
Abstract
On-policy distillation (OPD) has emerged as a core component of modern LLM post-training pipelines, yet we reveal a failure mode: degenerate agreement, where students exploit repetitive loops to achieve near-perfect token agreement with the teacher despite globally flawed responses. We therefore shift our focus from agreement to teacher-student mismatch, and find that mismatch tokens can be mainly categorized into two types: student-excess tokens and student-deficit tokens. Student-excess tokens are generated by the student but assigned near-zero probability by the teacher; their log-ratio corrections grow unbounded and destabilize the update. Student-deficit tokens, in contrast, are preferred by the teacher but rarely sampled by the student; their absence blocks the transfer of the teacher's reasoning patterns. To tackle these mismatch directions, we propose TIDE (Token-level Independent Deficit-Excess correction), which applies bounded Hellinger shaping to suppress the most severe sampled excesses and an analytic teacher top-$K$ injection to restore deficient probability mass without requiring deficit tokens to be sampled. Across mathematical reasoning benchmarks with multiple Qwen3 teacher-student pairs, TIDE consistently outperforms standard OPD and recent token-selection and reward-shaping baselines. Moreover, the gains of TIDE are more pronounced under strong teacher-student mismatch, where it improves Avg@8 from 6.9% to 20.3%, reduces average response length by a factor of 3.6, and substantially reduces formatting failures. Code is available at https://github.com/yzc-666/TIDE
Chinese Translation
在线策略蒸馏（On-policy distillation, OPD）已成为现代大规模语言模型（LLM）后训练流程的核心组成部分，但我们揭示了一种失效模式：退化一致性（degenerate agreement），在这种情况下，学生利用重复循环来实现与教师的近乎完美的标记一致性，尽管其响应在全局上存在缺陷。因此，我们将关注点从一致性转向教师与学生之间的不匹配，发现不匹配的标记主要可以分为两种类型：学生过量标记（student-excess tokens）和学生不足标记（student-deficit tokens）。学生过量标记由学生生成，但教师赋予其接近零的概率；其对数比率的修正无界增长，导致更新不稳定。相反，学生不足标记则是教师偏好但学生很少采样的标记；其缺失阻碍了教师推理模式的转移。为了解决这些不匹配问题，我们提出了TIDE（Token-level Independent Deficit-Excess correction），该方法应用有界Hellinger塑形来抑制最严重的采样过量，并通过分析性教师顶级K（top-$K$）注入来恢复不足的概率质量，而不需要采样不足标记。在多个Qwen3教师-学生对的数学推理基准测试中，TIDE始终优于标准OPD和最近的标记选择及奖励塑形基线。此外，在强教师-学生不匹配的情况下，TIDE的提升更为显著，将Avg@8从6.9%提高至20.3%，将平均响应长度减少了3.6倍，并显著降低了格式化失败的情况。代码可在https://github.com/yzc-666/TIDE获取。
cs.AI / 217 / 2608.09848
CEAA: A Cognitive Embodied Agents Architecture for Interactive Computing Systems
CEAA：一种用于交互计算系统的认知具身代理架构
Abstract
The development of embodied Intelligent Virtual Agents (IVAs) that have cognitive capabilities in real-time interactive virtual environments remains a challenge, even with today's advancements in technology. Existing architectures are often focused on either the implementation of low-level reactive control systems that are constrained by commercial game engines, or high-level representations of reasoning models that can be difficult to implement in virtual worlds. This paper builds on that notion and proposes a modular cognitive architecture for deploying embodied IVAs. This architecture builds on existing, pre-established frameworks such as the Sense-Think-Act paradigm and the Belief-Desire-Intention cognitive model, among others, and aims to provide a reusable implementation-oriented framework as a template for deploying IVA "brains" in interactive 3D computing systems. The proposed architecture contributes by providing a modular, implementation-oriented framework for the deployment of embodied, cognitive-capable IVAs and bridges the gap between high-level agent reasoning models with real-time embodied execution, for scalable, adaptive, and explainable agents in complex interactive virtual environments.
Chinese Translation
尽管今天的技术进步，开发具身的智能虚拟代理（IVAs）以在实时交互虚拟环境中具备认知能力仍然是一项挑战。现有架构通常专注于低级反应控制系统的实现，这些系统受到商业游戏引擎的限制，或是高层次推理模型的表示，这在虚拟世界中可能难以实现。本文基于这一理念，提出了一种模块化的认知架构，用于部署具身的IVAs。该架构基于现有的、预先建立的框架，如感知-思考-行动（Sense-Think-Act）范式和信念-欲望-意图（Belief-Desire-Intention）认知模型等，旨在提供一个可重用的、以实现为导向的框架，作为在交互3D计算系统中部署IVAs“大脑”的模板。所提议的架构通过提供一个模块化的、以实现为导向的框架，促进了具身、具认知能力的IVAs的部署，并弥合了高层次代理推理模型与实时具身执行之间的差距，以实现可扩展、适应性强和可解释的代理在复杂的交互虚拟环境中。
cs.AI / 218 / 2608.09855
Agentic Auto-Research is Fuzz Testing
代理自我研究是模糊测试
Abstract
Autonomous research agents can generate experiments faster than researchers can validate them. Researchers have responded by scaling the proposer and ranking more samples with a learned judge or human reviewers. We argue that this *generate-and-rank* paradigm misses the problem of sparse feedback. Within a declared research problem, an agent follows the control loop of a greybox fuzzer: it proposes a candidate, executes it, observes feedback, and chooses what to try next. A fuzzer rarely finds a bug, but coverage makes partial progress observable on every execution. Fuzzers then use that signal to mutate inputs and allocate effort, rather than only to rank completed runs. Auto-research needs the same two capabilities. First, each experiment should expose a cheap, dense signal of epistemic progress before final scientific validation is available. Second, that signal should determine the next intervention so that the agent searches rather than repeatedly samples. Because the optimized progress signal is guidance rather than a verdict, final validation must still decide what counts as a discovery using evidence protected from adaptive reuse. We propose controlled tests of whether candidate signals predict validated progress, whether feedback-directed search yields more validated discoveries per unit cost than repeated sampling, and whether protected validation reduces false discoveries. Feedback architecture, not only generation, is a central bottleneck in auto-research.
Chinese Translation
自主研究代理能够比研究人员更快地生成实验。研究人员通过扩展提议者并使用学习的评审者或人类评审者对更多样本进行排名来应对这一挑战。我们认为这种*生成与排名*的范式忽视了稀疏反馈的问题。在一个已声明的研究问题内，代理遵循灰盒模糊测试器的控制循环：它提出一个候选方案，执行该方案，观察反馈，并选择下一步尝试的内容。模糊测试器很少能找到漏洞，但覆盖率使得每次执行的部分进展可被观察到。模糊测试器随后利用该信号来变异输入并分配努力，而不仅仅是对已完成的运行进行排名。自动研究需要相同的两种能力。首先，每个实验应在最终科学验证之前暴露出一种廉价、密集的认识进展信号。其次，该信号应决定下一次干预，以便代理进行搜索而不是重复采样。由于优化的进展信号是指导而非裁决，最终验证仍然必须决定什么算作发现，并使用受保护的证据来防止自适应重用。我们建议进行受控测试，以检验候选信号是否能预测经过验证的进展，反馈导向的搜索是否能在单位成本下产生更多经过验证的发现，以及受保护的验证是否能减少虚假发现。反馈架构，而不仅仅是生成，是自动研究中的一个核心瓶颈。
cs.AI / 219 / 2608.09861
Towards Expert-level Medical AI for Real-time Video Consultations
迈向专家级医疗人工智能以实现实时视频咨询
Nagda, Mahvish, Lee, Jihyeon, Thompson, Matthew, Park, Chunjong, Strother, Tim, Liévin, Valentin, Ruparel, Roma, Goel, Akshay, Bergamaschi, Teya, Bedi, Suhana, Shah, Meet, Dubov, Pavel, Panait, Liviu, Fukuzawa, Toshiyuki, Schmidgall, Sam, Schiff, Craig, Xu, Joseph, Rysbek, Aliya, Lunts, Yana, Freyberg, Jan, Hemengway, Rebecca, Virmani, Sunny, Racz, David, Radebaugh, Carey, Barral, Joëlle, Goel, Kavi, Webster, Dale R., Chou, Katherine, Hassidim, Avinatan, Matias, Yossi, Manyika, James, Wayne, Gregory, Tu, Tao, Liu, Yun, Goh, Ethan, Chen, Christina, Tanno, Ryutaro, Chen, Po-Hsuan Cameron, Schaekermann, Mike, Palepu, Anil
Abstract
Audio-visual interaction is the standard for patient-physician consultations, enabling natural communication and effective assessment of illness through non-verbal cues. While text-based AI has shown promise, it discards essential perceptual dimensions and limits patients who cannot articulate symptoms in writing. Early efforts to extend medical AI to audio-visual interaction have demonstrated feasibility but not reached clinician-level performance. Here, we provide the first demonstration of expert-level AI in real-time clinical video consultations using AMIE (Articulate Medical Intelligence Explorer) in a video configuration. AMIE (Video) is a Gemini-based multi-agent system integrating low-latency dialogue, clinical reasoning, and real-time audio-visual perception. To guide development, we established a taxonomy and automated evaluations for clinical audio-visual cues in telehealth settings. In a randomized Objective Structured Clinical Examination (OSCE) study with 30 primary care physicians (PCPs), 15 patient actors and 100 clinical scenarios, we compared AMIE (Video), its text-only counterpart AMIE (Text), and PCPs consulting via video. Clinical evaluators rated AMIE (Video) on par or better than PCPs in history-taking, diagnosis, management, and physical observation and examination. Patient actors preferred AMIE's approach to assessing and explaining conditions, while PCPs were preferred for rapport and partnership building. In modality ablation, patient actors preferred AMIE (Video)'s interface over text chat for communicative effectiveness, convenience, and feeling understood. Limitations remain in fine anatomical precision, subtle affective nuances, and high-frequency movements. While further research is needed before real-world translation, these results mark an important milestone toward AI systems capable of augmenting care across the sensory complexity of clinical practice.
Chinese Translation
视听互动是患者与医生咨询的标准，能够通过非语言线索实现自然沟通和有效的疾病评估。尽管基于文本的人工智能显示出一定前景，但它忽略了重要的感知维度，并限制了无法用书面表达症状的患者。早期将医疗人工智能扩展到视听互动的努力已证明其可行性，但尚未达到临床医生的水平。在此，我们首次展示了在实时临床视频咨询中使用AMIE（Articulate Medical Intelligence Explorer）实现的专家级人工智能。AMIE（Video）是一个基于Gemini的多代理系统，集成了低延迟对话、临床推理和实时视听感知。为了指导开发，我们建立了一个分类法并对远程医疗环境中的临床视听线索进行了自动化评估。在一项涉及30名初级保健医生（PCPs）、15名患者演员和100个临床场景的随机客观结构化临床考试（OSCE）研究中，我们比较了AMIE（Video）、其文本版AMIE（Text）以及通过视频咨询的PCPs。临床评估者在病史采集、诊断、管理以及身体观察和检查方面对AMIE（Video）的评分与PCPs相当或更优。患者演员更喜欢AMIE在评估和解释病情方面的方法，而PCPs在建立关系和合作方面更受欢迎。在模态消融实验中，患者演员更倾向于AMIE（Video）的界面，相较于文本聊天在沟通有效性、便利性和被理解感方面表现更佳。尽管在细微解剖精确度、微妙情感细节和高频运动方面仍存在局限性，但在实现现实世界转化之前仍需进一步研究，这些结果标志着朝着能够增强临床实践感知复杂性的护理的人工智能系统迈出了重要的一步。
cs.AI / 220 / 2608.09874
ArchAgent v2: A Case Study with the Data Prefetching Championship
ArchAgent v2：数据预取冠军赛的案例研究
Abstract
Agentic artificial intelligence has shown great promise in automating algorithm design, but scaling similar techniques to computer microarchitecture discovery remains challenging due to vast search spaces, strict hardware budgets, and long simulation times. In this work, we present ArchAgent v2, a framework which scales automated microarchitecture search to multi-level data prefetching. While the original ArchAgent successfully discovered single-level cache replacement policies in competition settings, it does not scale to multi-level prefetching where the design space and degrees of freedom are larger. To overcome this, we introduce two new additions to ArchAgent: a cascaded evolutionary search that subdivides the design space by sequentially evolving and freezing prefetchers at individual cache levels, and a hardware-realizability feedback loop that embeds real-time size-estimation directly into the evolution process. Evaluated under identical rules of the 4th Data Prefetching Championship (DPC4), ArchAgent v2 automatically designs a three-level prefetcher that outperforms the winning hand-designed solution, further demonstrating automated agentic discovery as a useful tool for computer architects. Our discovered policy achieves a 3.8\% geometric mean IPC speedup over the baseline overall and a 0.3\% improvement over the prior champion, BertiGO. On low-bandwidth single-core configurations, our policy yields a 4.6\% performance speedup compared to only 2.6\% for BertiGO. However, multi-core evolution still remains a significant challenge due to simulation latency impeding evolution speed. Finally, our profiling of an ArchAgent evolution of over 12,000 candidate designs provides key insights into how automated evolutionary agents explore and synthesize complex microarchitectural logic.
Chinese Translation
代理人工智能在自动化算法设计方面展现出了巨大的潜力，但将类似技术扩展到计算机微架构发现仍然面临挑战，原因在于广泛的搜索空间、严格的硬件预算和较长的仿真时间。在本研究中，我们提出了ArchAgent v2，一个将自动化微架构搜索扩展到多级数据预取的框架。尽管原始的ArchAgent在竞争环境中成功发现了单级缓存替换策略，但在设计空间和自由度更大的多级预取中并未实现扩展。为了解决这一问题，我们为ArchAgent引入了两个新功能：一个级联进化搜索，通过依次进化和冻结各个缓存级别的预取器来细分设计空间，以及一个硬件可实现性反馈循环，将实时大小估计直接嵌入到进化过程中。在第四届数据预取冠军赛（DPC4）的相同规则下进行评估时，ArchAgent v2自动设计了一个三级预取器，其性能超越了获胜的手动设计方案，进一步证明了自动化代理发现作为计算机架构师的有用工具。我们发现的策略在整体上实现了相对于基线的3.8\%几何平均IPC加速，并且比之前的冠军BertiGO提高了0.3\%。在低带宽单核配置中，我们的策略相比BertiGO的2.6\%性能提升实现了4.6\\%的性能加速。然而，由于仿真延迟阻碍了进化速度，多核进化仍然是一个重大挑战。最后，我们对ArchAgent进化的超过12,000个候选设计的分析提供了关键见解，揭示了自动化进化代理如何探索和合成复杂的微架构逻辑。
cs.AI / 221 / 2608.09885
SHE: Trajectory-driven Safety Harness Evolution for LLM Agents
SHE：基于轨迹驱动的安全保护带演化用于大型语言模型代理
Abstract
The safety of large language model (LLM) agents depends not only on model weights but also on the agent harness that manages context, memory, tools, permissions, and runtime control. Existing safety mechanisms often treat the harness as a fixed deployment artifact, limiting their ability to evolve with emerging risks. Moreover, coupled functions across harness components obscure safety responsibility attribution, making localized evolution difficult. We propose Safety Harness Evolution (SHE), a framework that learns evolving safe boundaries from rollout trajectories. SHE decomposes the harness into four artifacts with explicit safety responsibilities, including the System Prompt, Rule Bank, Safety Memory, and Tool Policy, defining clear functional boundaries for localized evolution. Based on this decomposition, SHE introduces an attribution-guided evolution loop that converts trajectory failures into structured diagnoses, learns artifact-specific boundary refinements, and selects evolved harnesses through safety-utility validation. Experiments on Agent-SafetyBench demonstrate that SHE effectively enhances safety through harness evolution, achieving a 3.1x ASR reduction compared with static SafeHarness, while also improving benign utility. The evolved harness further generalizes to unseen risks on the held-out AgentHarm benchmark and transfers across agent models without additional evolution.
Chinese Translation
大型语言模型（LLM）代理的安全性不仅依赖于模型权重，还依赖于管理上下文、记忆、工具、权限和运行时控制的代理保护带。现有的安全机制通常将保护带视为固定的部署工件，这限制了它们随着新兴风险的演变能力。此外，保护带组件之间的耦合功能模糊了安全责任的归属，使得局部演变变得困难。我们提出了安全保护带演化（Safety Harness Evolution，SHE），这是一个从轨迹回放中学习演变安全边界的框架。SHE将保护带分解为四个具有明确安全责任的工件，包括系统提示（System Prompt）、规则库（Rule Bank）、安全记忆（Safety Memory）和工具策略（Tool Policy），为局部演变定义了清晰的功能边界。基于这种分解，SHE引入了一个归属引导的演变循环，将轨迹失败转化为结构化诊断，学习工件特定的边界细化，并通过安全效用验证选择演变后的保护带。在Agent-SafetyBench上的实验表明，SHE通过保护带演变有效增强了安全性，与静态的SafeHarness相比，实现了3.1倍的ASR降低，同时也提高了良性效用。演变后的保护带进一步在未见风险的AgentHarm基准上具有更好的泛化能力，并且在不同代理模型之间转移时无需额外演变。
cs.AI / 222 / 2608.09902
DSLE: A Learning Environment for Dark Souls Boss Encounters
DSLE：黑暗之魂Boss遭遇的学习环境
Abstract
We introduce the Dark Souls Learning Environment (DSLE), a containerized platform that presents all 22 boss encounters of Dark Souls: Remastered as game-playing agent benchmarks through a Gymnasium-style interface. DSLE combines real-time combat, high-dimensional visual input, and sparse terminal rewards, with each environment step being a real action executed against the running game. To support controlled comparison, we define DSLE-5, a representative five-boss subset, spanning a melee fight, a spatially constrained arena, an environmental-hazard fight, a multi-target fight, and a fast final-boss fight, that we recommend as the starting suite for agents built on DSLE. On DSLE-5 we evaluate a random policy, an expert system, an evolutionary baseline, and PPO and DQN agents trained from visual input. The expert system and the evolutionary baseline each defeat the Asylum Demon, the game's tutorial boss (63% and 43% peak win rates), but none of the five methods defeats the other four DSLE-5 bosses; PPO and DQN show no measurable learning (at most 0.33% win rate on the tutorial boss, 0% elsewhere) within a budget that already costs tens of wall-clock hours per run. A broader study running the evolutionary baseline across all 22 encounters under advantaged all level-50 stats yields wins on only a handful of additional early-game bosses and leaves the rest unwon. The failure cases range from sub-10-second deaths in cramped, multi-target encounters to minute-long stalemates that inflict almost no damage, and we report them through survival time and damage dealt rather than win rate alone.
Chinese Translation
我们介绍了黑暗之魂学习环境（Dark Souls Learning Environment，DSLE），这是一个容器化平台，通过类似Gymnasium的接口将《黑暗之魂：重制版》的22个Boss遭遇呈现为游戏代理基准。DSLE结合了实时战斗、高维视觉输入和稀疏的终端奖励，每个环境步骤都是对正在运行的游戏执行的真实动作。为了支持控制比较，我们定义了DSLE-5，这是一个具有代表性的五个Boss子集，涵盖了一场近战战斗、一个空间受限的竞技场、一场环境危害战斗、一场多目标战斗和一场快速的最终Boss战斗，我们推荐将其作为基于DSLE构建的代理的起始套件。在DSLE-5上，我们评估了随机策略、专家系统、进化基线，以及从视觉输入训练的PPO和DQN代理。专家系统和进化基线各自击败了游戏的教程Boss——庇护恶魔（Asylum Demon），其最高胜率分别为63%和43%，但五种方法都未能击败其他四个DSLE-5 Boss；PPO和DQN在每次运行已经耗费数十小时的预算内没有显示出可测量的学习（在教程Boss上的胜率最多为0.33%，在其他地方为0%）。在所有22场遭遇中运行进化基线的更广泛研究，在所有角色属性为50的情况下，仅在少数额外的早期Boss上获得胜利，其余则未能获胜。失败案例从在拥挤的多目标遭遇中不足10秒的死亡，到几分钟的僵局几乎没有造成任何伤害，我们通过生存时间和造成的伤害而非单纯的胜率来报告这些情况。
cs.AI / 223 / 2608.09921
GENCO - A Unified Neural Solver Embedded in a Development Framework for Steady-State Grid Analysis
GENCO - 嵌入开发框架的统一神经求解器用于稳态电网分析
Puech, Alban, Mazzonelli, Matteo, Govindasamy, Tamara R., Mngomezulu, Mangaliso, Maeso-García, Héctor, Tolhurst, Thomas, Bayazi, Javad, Moeini, Ali, Simumba, Naomi, Cintas, Celia, Nelischer, David, Kienzler, Romeo, Weiss, Jonas, Varbella, Anna, Dörfler, Florian, Hug, Gabriela, Mevissen, Martin, Bernabé-Moreno, Juan, Mirallès, François, Hamann, Hendrik F., Vos, Etienne, Brunschwiler, Thomas
Abstract
Foundation models are transforming business workflows and boosting productivity, yet they remain largely absent from engineering domains such as power system analysis, where strict physical consistency must be enforced. We present GENCO (GEometric Neural Corrective Optimizer), a unified neural solver for steady-state transmission grid analysis that handles power flow (PF), optimal power flow (OPF), and state estimation (SE) within a single architecture and shared network representation. To support advances in neural power system solvers, we introduce the open-source GridFM Development Framework, which standardizes synthetic data generation and training in a low-code environment. We also release large-scale datasets with millions of PF and OPF scenarios across diverse grid topologies to support reproducible benchmarking. We evaluate GENCO on the PFDelta and OPFData benchmarks against state-of-the-art neural solvers and classical solvers, including Newton-Raphson and IPOPT, as well as on real-world Hydro-Qu\'ebec SCADA data. For large-scale PF, GENCO recovers the full AC operating state, including voltage magnitudes and reactive power that DC-PF cannot provide, while matching DC-PF-level active power-balance residuals. It achieves up to 30x speedups over Newton-Raphson at only 2x the runtime of DC-PF. For OPF, it achieves up to 85x speedups over IPOPT while improving feasibility, optimality, and runtime over DC-OPF. For SE, GENCO is more robust than classical weighted least squares to noisy measurements and network parameter errors, and always returns a high-quality estimate even when weighted least squares fails to converge. Together, the unified architecture and development framework provide a new approach to large-scale steady-state grid analysis, lowering the barrier to entry for power system engineers and marking a step toward Grid Foundation Models.
Chinese Translation
基础模型正在改变商业工作流程并提升生产力，但在电力系统分析等工程领域仍然缺乏应用，因为这些领域必须严格遵循物理一致性。我们提出了GENCO（几何神经纠正优化器），这是一个用于稳态输电网分析的统一神经求解器，能够在单一架构和共享网络表示中处理功率流（PF）、最优功率流（OPF）和状态估计（SE）。为了支持神经电力系统求解器的进步，我们推出了开源的GridFM开发框架，该框架在低代码环境中标准化了合成数据生成和训练。我们还发布了涵盖多种电网拓扑的大规模数据集，其中包含数百万个PF和OPF场景，以支持可重复的基准测试。我们在PFDelta和OPFData基准测试上评估GENCO，与最先进的神经求解器和经典求解器（包括牛顿-拉夫森法和IPOPT）进行比较，并在实际的Hydro-Québec SCADA数据上进行测试。对于大规模PF，GENCO恢复了完整的交流（AC）操作状态，包括直流功率流（DC-PF）无法提供的电压幅值和无功功率，同时匹配DC-PF水平的有功功率平衡残差。其速度比牛顿-拉夫森法快达30倍，仅需DC-PF运行时间的2倍。对于OPF，其速度比IPOPT快达85倍，同时在可行性、最优性和运行时间上优于DC-OPF。对于SE，GENCO在面对噪声测量和网络参数误差时比经典加权最小二乘法更具鲁棒性，即使加权最小二乘法无法收敛，GENCO也始终返回高质量的估计。统一架构和开发框架共同提供了一种新的大规模稳态电网分析方法，降低了电力系统工程师的入门门槛，并标志着朝向电网基础模型迈出了一步。
cs.CL / 1 / 2608.07525
Unified Hallucination Fuzzing for Multimodal Large Language Models
多模态大型语言模型的统一幻觉模糊测试
Abstract
Hallucination remains a persistent challenge for Multimodal Large Language Models (MLLMs), severely limiting their reliability in high-stakes applications. Existing evaluations, predominantly based on static benchmarks, suffer from narrow taxonomical coverage and rapid performance saturation, failing to reflect model robustness in evolving real-world scenarios. To bridge this gap, we present a systematic evaluation framework integrating a comprehensive benchmark with self-evolving stress testing. First, we introduce UniHall, a fine-grained dataset grounded in a unified taxonomy spanning Object, Instruction, and Knowledge dimensions. Second, to address benchmark saturation, we propose Self-Adaptive Multimodal Fuzzing (SAMF), a self-adaptive framework that employs evolutionary mutation strategies to explore the boundaries of model hallucinations. Crucially, to ensure reliable assessment of dynamic inputs, SAMF incorporates a structured metric suite driven by an ensemble of multi-modal oracles. Our extensive experiments reveal that state-of-the-art MLLMs exhibit significant performance degradation under fuzzing compared to conventional settings, exposing a dissociation between reasoning capabilities and factual grounding. Furthermore, we identify a helpfulness-hallucination trade-off, where reinforcement learning alignment inadvertently exacerbates sycophancy in instruction-following tasks. The framework, code and benchmark are available at https://github.com/LanceZPF/EvalHall.
Chinese Translation
幻觉仍然是多模态大型语言模型（MLLMs）面临的一个持续挑战，严重限制了它们在高风险应用中的可靠性。现有的评估主要基于静态基准，存在分类覆盖面狭窄和性能迅速饱和的问题，无法反映模型在不断变化的现实场景中的鲁棒性。为了解决这一问题，我们提出了一个系统的评估框架，整合了全面的基准和自我演化的压力测试。首先，我们引入了UniHall，这是一个基于统一分类法的细粒度数据集，涵盖对象、指令和知识维度。其次，为了解决基准饱和的问题，我们提出了自适应多模态模糊测试（Self-Adaptive Multimodal Fuzzing，SAMF），这是一个自适应框架，采用进化变异策略探索模型幻觉的边界。重要的是，为了确保对动态输入的可靠评估，SAMF包含一个由多模态神谕组成的结构化度量套件。我们的广泛实验表明，与传统设置相比，最先进的MLLMs在模糊测试下表现出显著的性能下降，揭示了推理能力与事实基础之间的脱节。此外，我们识别出一种有用性与幻觉之间的权衡关系，其中强化学习对齐无意中加剧了在指令跟随任务中的阿谀奉承现象。该框架、代码和基准可在 https://github.com/LanceZPF/EvalHall 获取。
cs.CL / 2 / 2608.07527
DocAtlas: Long-Document Understanding as Mutable-State Interaction
DocAtlas：将长文档理解视为可变状态交互
Abstract
Long-document understanding requires models to find and combine evidence across many pages, layouts, tables, figures, and charts. Existing retrieval-augmented systems usually select evidence from a static index before generation, while recent agentic systems add multi-turn tool use but often rely on frozen proprietary backbones whose behavior is set by prompts. We present DocAtlas, a system that treats long-document understanding as a mutable-state information-seeking process. We instantiate DocAtlas as a mutable document harness: an external environment that determines what document information is searched, read, stored, reviewed, and shown to the model at each step. Given a document and question, the harness exposes search, reading, note-taking, and review tools, maintains a hierarchical tree and note store, and updates both as the agent records evidence. DocAtlas combines self-improving retrieval, selective evidence access, and active working memory under a fixed context budget. The same harness supports inference-time use with large VLMs and end-to-end reinforcement learning for compact VLM agents. With GPT-5.4, DocAtlas reaches 71.4\% on MMLongBench-Doc, exceeding the human-expert reference of 65.8\%. A Qwen3.5-4B VLM trained with end-to-end RL in the DocAtlas environment reaches 63.7\%, compared with a 54.4\% direct-input baseline, showing that mutable document-harness design can improve compact document agents by a large margin.
Chinese Translation
长文档理解要求模型在多个页面、布局、表格、图形和图表中寻找和组合证据。现有的增强检索系统通常在生成之前从静态索引中选择证据，而最近的代理系统虽然增加了多轮工具使用，但往往依赖于由提示设置行为的冻结专有骨干网络。我们提出了DocAtlas，一个将长文档理解视为可变状态信息寻求过程的系统。我们将DocAtlas实例化为一个可变文档工具：一个外部环境，决定在每一步中搜索、阅读、存储、审阅和展示给模型的文档信息。给定一个文档和问题，该工具提供搜索、阅读、记笔记和审阅工具，维护一个层次树和笔记存储，并在代理记录证据时更新这两者。DocAtlas在固定上下文预算下结合了自我改进的检索、选择性证据访问和主动工作记忆。相同的工具支持与大型视觉语言模型（VLM）的推理时使用以及针对紧凑型VLM代理的端到端强化学习。在GPT-5.4上，DocAtlas在MMLongBench-Doc上达到了71.4\%，超过了人类专家参考的65.8\%。在DocAtlas环境中经过端到端强化学习训练的Qwen3.5-4B VLM达到了63.7\%，相比之下，直接输入基线为54.4\%，显示出可变文档工具设计可以大幅提升紧凑文档代理的性能。
cs.CL / 3 / 2608.07529
WuYuEval: A Multi-Level Benchmark for Large Language Models in Solid Waste Management
WuYuEval：用于固体废物管理的大型语言模型多层次基准测试
Abstract
Large language models (LLMs) are increasingly used as technical assistants, but their competence in solid waste management (SWM) remains difficult to assess because existing benchmarks emphasize general knowledge rather than professional decisions under engineering, environmental, and policy constraints. We introduce WuYuEval, a multi-level benchmark for evaluating LLMs in SWM across foundational knowledge, domain reasoning, and expert decision-making. After quality auditing, WuYuEval contains a Foundation Module with 4,590 closed-ended multiple-choice questions across six task types and eight domain categories, together with an Expert Module with 247 scenario-based open-ended questions involving multi-objective optimization, constraint trade-offs, and system design. For expert tasks, we combine anchor-calibrated LLM-as-a-Judge scoring with Elo-based pairwise comparison. Across 33 LLMs, performance varied widely. The leading model reached 94.64\% accuracy on the Foundation Module, but average accuracy still fell from 84.14\% on easy questions to 42.50\% on hard questions, with lower performance concentrated in calculation, experimental design, urban planning, and open-ended expert tasks. Reasoning-oriented Thinking modes improve most matched model pairs after auditing, but the gains depend on baseline capability and are not uniformly positive. These results suggest that visible deliberation helps only when it remains anchored to units, assumptions, and engineering constraints; otherwise, it may drift from decisive answer boundaries. WuYuEval therefore provides both an evaluation resource and an empirical basis for developing SWM-oriented foundation models with professional reasoning chains and explicit constraint control.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作技术助手，但它们在固体废物管理（SWM）中的能力仍然难以评估，因为现有基准强调一般知识，而不是在工程、环境和政策约束下的专业决策。我们引入了WuYuEval，一个用于评估LLMs在SWM中表现的多层次基准，涵盖基础知识、领域推理和专家决策。经过质量审核，WuYuEval包含一个基础模块，涵盖6种任务类型和8个领域类别的4,590个封闭式多项选择题，以及一个专家模块，包含247个基于场景的开放式问题，涉及多目标优化、约束权衡和系统设计。对于专家任务，我们结合了锚定校准的LLM作为评判者的评分与基于Elo的成对比较。在33个LLMs中，表现差异显著。领先模型在基础模块上达到了94.64%的准确率，但平均准确率仍然从简单问题的84.14%下降到困难问题的42.50%，而较低的表现主要集中在计算、实验设计、城市规划和开放式专家任务中。经过审核，面向推理的思维模式改善了大多数匹配模型对的表现，但增益依赖于基线能力，并非均匀正面。这些结果表明，明显的深思熟虑只有在与单位、假设和工程约束相结合时才有帮助；否则，它可能会偏离决定性答案的边界。因此，WuYuEval不仅提供了一个评估资源，也为开发面向SWM的基础模型提供了具有专业推理链和明确约束控制的实证基础。
cs.CL / 4 / 2608.07531
Search-G1: Grounded Search Agents via Representation-Based Intrinsic Rewards
Search-G1：基于表征的内在奖励的基础搜索代理
Abstract
Search-augmented language agents should retrieve external information only when necessary and ground their answers in retrieved evidence. Existing external rewards provide either sparse outcome supervision or richer feedback from process annotations and LLM judges. Outcome rewards scale readily but cannot distinguish grounded retrieval from redundant search, whereas richer signals require costly annotation or inference during training. Internal rewards based on policy-side signals such as entropy, likelihood, or information gain are graded and inexpensive to evaluate, yet mainly reflect model confidence rather than evidence grounding. We propose Search-G1, a representation-based intrinsic reward framework that measures the operational grounding of an agent's answers through two intervention-calibrated readouts. A prompt-state readout predicts closed-book sufficiency, whose complement defines policy-relative retrieval necessity; an answer-commit readout estimates evidence reliance from answer-stage sensitivity to evidence deletion. Together, they provide additional credit to correct searched trajectories when retrieval is estimated necessary and the answer is evidence-sensitive, favor correct direct answers when closed-book knowledge suffices, and penalize repeated search. After calibration, reward scoring requires neither process annotations nor LLM-as-judge inference during policy optimization. Because reinforcement learning changes policy representations, Search-G1 periodically refits both readouts on trajectories from the latest checkpoint, allowing the reward to co-evolve with the policy. Experiments across multiple search-based question-answering benchmarks and two model scales show that Search-G1 improves the grounding--search-cost trade-off, producing shorter response-side trajectories at competitive task accuracy. Code is available at https://github.com/Rosy0912/Search-G1.
Chinese Translation
增强搜索的语言代理应仅在必要时检索外部信息，并将其答案基于检索到的证据。现有的外部奖励提供的要么是稀疏的结果监督，要么是来自过程注释和大型语言模型（LLM）评审的更丰富反馈。结果奖励易于扩展，但无法区分基础检索与冗余搜索，而更丰富的信号则需要在训练过程中进行昂贵的注释或推理。基于策略侧信号（如熵、似然性或信息增益）的内部奖励是分级的，且评估成本低廉，但主要反映模型的信心，而非证据的基础。我们提出了Search-G1，一种基于表征的内在奖励框架，通过两个干预校准的读出测量代理答案的操作基础。提示状态读出预测闭卷知识的充分性，其补充定义了相对于策略的检索必要性；答案承诺读出根据答案阶段对证据删除的敏感性估计证据依赖性。它们共同在检索被估计为必要且答案对证据敏感时，为正确的搜索轨迹提供额外的奖励，当闭卷知识足够时偏向正确的直接答案，并惩罚重复搜索。经过校准后，奖励评分在策略优化过程中不需要过程注释或LLM作为评审的推理。由于强化学习会改变策略表征，Search-G1定期在最新检查点的轨迹上重新拟合这两个读出，使奖励与策略共同演化。在多个基于搜索的问题回答基准和两个模型规模的实验表明，Search-G1改善了基础与搜索成本的权衡，在保持竞争性任务准确度的同时，生成了更短的响应侧轨迹。代码可在 https://github.com/Rosy0912/Search-G1 获取。
cs.CL / 5 / 2608.07594
Scaling Inherently Interpretable Language Models
可扩展的内在可解释语言模型
Abstract
Interpretability is often treated as a tax on capability: language models are trained as opaque systems, then explained after the fact, with methods whose reliability is difficult to establish. In this work, we challenge this premise. Rather than reverse-engineering a model, we make interpretability a constraint of the training pipeline, optimized alongside the language modeling objective. Across three orders of magnitude of compute, on both autoregressive and diffusion language models, interpretability scales with capability rather than against it. Surprisingly, model representations become more disentangled and aligned with human-understandable concepts with scale. We instantiate the training-time recipe with Steerling-8B, a diffusion language model with a causal attention mask. For any group of generated tokens, Steerling-8B attributes the output to relevant input tokens, human-understandable concepts, and training data. This enables closed-loop intervention: diagnose an output through its concept or feature attribution, retrieve similar training data, and correct the behavior through concept steering without retraining. Steerling-8B remains competitive with open peer models trained on substantially 2-16x more compute, suggesting a different scaling paradigm: interpretability can be designed into training, and it improves with scale.
Chinese Translation
可解释性常常被视为能力的负担：语言模型作为不透明系统进行训练，然后在事后通过难以建立可靠性的解释方法进行解释。在本研究中，我们挑战了这一前提。我们并不是对模型进行逆向工程，而是将可解释性作为训练流程的一个约束，与语言建模目标一起进行优化。在三个数量级的计算资源下，无论是自回归语言模型还是扩散语言模型，可解释性与能力呈正相关关系，反而不是相对立的。令人惊讶的是，随着规模的扩大，模型表示变得更加解缠且与人类可理解的概念对齐。我们通过 Steerling-8B 实例化了这一训练时的配方，Steerling-8B 是一种具有因果注意力掩码的扩散语言模型。对于任何一组生成的标记，Steerling-8B 将输出归因于相关的输入标记、人类可理解的概念和训练数据。这使得闭环干预成为可能：通过其概念或特征归因诊断输出，检索相似的训练数据，并通过概念引导在不重新训练的情况下纠正行为。Steerling-8B 在与开放同行模型的竞争中表现良好，后者的训练计算资源多达 2-16 倍，暗示了一种不同的扩展范式：可解释性可以被设计进训练中，并且随着规模的扩大而改善。
cs.CL / 6 / 2608.07629
Embedding Initialization for Unseen Low-resource Languages in Multilingual NMT: A Case Study on Limbum-English Translation
多语言神经机器翻译中未见低资源语言的嵌入初始化：以Limbum-英语翻译为例
Abstract
Multilingual neural machine translation models such as NLLB-200 cover 200 languages but leave thousands unsupported, including most Grassfields Bantu languages of Cameroon. When fine-tuning these models for an unseen language, practitioners must choose a proxy language token, yet no principled method exists for this selection. We implemented an embedding initialization strategy where a language token is the average of embeddings from multiple typologically related languages already in the mod el. We evaluate this approach on Limbum-to-English translation using a parallel corpus of 8,837 sentence pairs from New Testament text and a bilingual dictionary. We compare models: NLLB-200 zero-shot (chrF2++ = 12.5), a Transformer trained from scratch (chrF2++ = 14.5), NLLB-200 fine-tuned with a Swahili proxy token (chrF2++ = 47.3), and NLLB-200 with our averaged embedding initialization (chrF2++ = 46.7). We find that the multi-language initialization achieves performance comparable to the best single-language proxy. Both NLLB-200 variants improve over the from-scratch baseline by over 32 chrF2++ points. These results show that multilingual transfer is the dominant factor in extremely low-resource Bantu translation while eliminating the need for heuristic proxy selection. However, all systems fail to preserve tonal diacritics, highlighting an open challenge. We make our dataset and code available to support further research.
Chinese Translation
多语言神经机器翻译模型如NLLB-200覆盖200种语言，但仍有数千种语言未被支持，包括喀麦隆的大草原班图语言。当对未见语言进行微调时，实践者必须选择一个代理语言标记，但目前尚无原则性的方法来进行此选择。我们实施了一种嵌入初始化策略，其中语言标记是多个已经在模型中存在的类型相关语言的嵌入平均值。我们在Limbum到英语的翻译中评估了这种方法，使用了来自新约文本的8,837对句子的平行语料库和一个双语词典。我们比较了以下模型：NLLB-200零样本（chrF2++ = 12.5）、从头训练的Transformer（chrF2++ = 14.5）、使用斯瓦希里语代理标记微调的NLLB-200（chrF2++ = 47.3），以及使用我们平均嵌入初始化的NLLB-200（chrF2++ = 46.7）。我们发现，多语言初始化的性能与最佳单语言代理相当。两个NLLB-200变体的性能均比从头基线提高了超过32个chrF2++点。这些结果表明，多语言迁移是极低资源班图翻译中的主导因素，同时消除了对启发式代理选择的需求。然而，所有系统都未能保留声调变音符，这突显了一个亟待解决的挑战。我们将我们的数据集和代码公开，以支持进一步的研究。
cs.CL / 7 / 2608.07641
SurveyReview: A Reviewer-Aligned Benchmark for Survey Evaluators
SurveyReview：一个与评审者对齐的调查评估基准
Abstract
The rapid advancement of large language models has transformed survey writing from a months-long manual effort into an automated process. As generation scales, reliable evaluation becomes the bottleneck, and LLMs are increasingly used as survey evaluators. However, existing approaches largely rely on off-the-shelf LLM-as-a-judge methods without systematic alignment to human reviewers, and there remains a lack of systematic frameworks for quantifying alignment with human reviewers. To address this gap, we propose SurveyReview, a reviewer-aligned, multi-dimensional benchmark and dataset for survey evaluation. We collect and annotate 675 survey papers with 1,630 review reports. We structure authentic peer-review reports by converting free-form comments into four-dimensional scores (Readability, Criticalness, Comprehensiveness, Structure) paired with supporting rationales. We further release standardized train/test splits and an evaluation protocol to measure alignment between automatic evaluators and human reviewers. To validate the benchmark, we develop SurveyAlign, a strong baseline evaluator by fine-tuning Qwen3-32B with LoRA on our annotated data, augmented with external knowledge for knowledge-intensive dimensions. On the test set, SurveyAlign substantially improves reviewer alignment over prompt-based judging with GPT-5.2, reducing average MSE from 2.28 to 1.38 and MAE from 1.15 to 0.69 across all four dimensions. Our contributions are twofold: (1) we establish the first multi-dimensional, reviewer-aligned dataset with a reproducible evaluation framework for survey reviewing; (2) we develop a strong baseline evaluator that substantially improves alignment with human reviewers, providing a competitive reference for future research. Our code and data are available at https://surveyreview.github.io
Chinese Translation
大型语言模型的快速发展已将调查写作从耗时数月的手动工作转变为自动化过程。随着生成规模的扩大，可靠的评估成为瓶颈，LLMs（大型语言模型）越来越多地被用作调查评估者。然而，现有的方法在很大程度上依赖于现成的LLM作为评审者的方法，而没有与人类评审者进行系统的对齐，且缺乏量化与人类评审者对齐的系统框架。为了解决这一问题，我们提出了SurveyReview，一个与评审者对齐的多维度调查评估基准和数据集。我们收集并注释了675篇调查论文及其1630份评审报告。我们通过将自由形式的评论转换为四维评分（可读性、批判性、全面性、结构）并配以支持性理由，构建了真实的同行评审报告。我们进一步发布了标准化的训练/测试划分和评估协议，以测量自动评估者与人类评审者之间的对齐程度。为了验证该基准，我们开发了SurveyAlign，一个强大的基线评估器，通过在我们的注释数据上对Qwen3-32B进行LoRA微调，并结合外部知识以增强知识密集型维度。在测试集上，SurveyAlign在评审者对齐方面显著优于基于提示的GPT-5.2评判，将所有四个维度的平均均方误差（MSE）从2.28降低到1.38，平均绝对误差（MAE）从1.15降低到0.69。我们的贡献有两个方面：（1）我们建立了第一个多维度、与评审者对齐的数据集，并提供了可重复的调查评审评估框架；（2）我们开发了一个强大的基线评估器，显著提高了与人类评审者的对齐，为未来的研究提供了具有竞争力的参考。我们的代码和数据可在https://surveyreview.github.io获取。
cs.CL / 8 / 2608.07727
Evaluating Dedicated Monolingual and Joint Multilingual Causal Models for Dravidian Languages
评估专用单语和联合多语因果模型在德拉威语中的表现
Abstract
Dravidian languages, mainly Tamil, Telugu, Kannada, and Malayalam make up only a small part of the data used to train multilingual language models, so it's not clear how much per-language ability these models actually keep. I have trained five GPT-2 architecture models from scratch to compare four monolingual models (one each for Tamil, Telugu, Kannada, and Malayalam, each with its own 32K-vocabulary subword tokenizer) against one multilingual model sharing a 64K-vocabulary subword tokenizer across all four languages. All the 5 models are trained on cleaned CC-100, Wikipedia, and Samanantar data. I have tested the models on perplexity, bits-per-byte, tokenizer efficiency, and fine-tuning results which are compared against mGPT. The monolingual models outperform mGPT on sentiment classification and named entity recognition, and their tokenizers proved more efficient than the shared multilingual model across all the languages tested.
Chinese Translation
德拉威语，主要包括泰米尔语、泰卢固语、卡纳达语和马拉雅拉姆语，仅占用于训练多语言模型的数据的一小部分，因此尚不清楚这些模型在每种语言上的能力究竟保持了多少。我从头开始训练了五个GPT-2架构模型，以比较四个单语模型（每种语言一个，分别为泰米尔语、泰卢固语、卡纳达语和马拉雅拉姆语，每个模型都有自己的32K词汇子词分词器）与一个共享64K词汇子词分词器的多语言模型。所有五个模型均在清理过的CC-100、维基百科和Samanantar数据上进行训练。我对模型进行了困惑度、字节比特、分词器效率和微调结果的测试，并与mGPT进行了比较。单语模型在情感分类和命名实体识别方面的表现优于mGPT，并且它们的分词器在所有测试语言中都证明比共享的多语言模型更高效。
cs.CL / 9 / 2608.07737
The No-Meaning Falsity: The Structural Impossibility of the Arbitrary Sign in Classical Arabic
无意义的虚假性：古典阿拉伯语中任意符号的结构不可能性
Abstract
This paper investigates whether the postmodern claim of unrestricted semantic indeterminacy, and its foundational Saussurean axiom of the arbitrary sign, are compatible with the structural architecture of Classical Arabic. We develop a formal mathematical model of Arabic non concatenative morphology in which lexical meaning is determined by the interaction between an invariant root and a morphosyntactic pattern. Within this framework, we establish a Morphological Correspondence Theorem, demonstrating that every lexical item is uniquely generated by a root pattern pair, and a Semantic Localization Theorem, proving that lexical meaning is determined at the derivational level prior to surface realization. To address Saussurean weaker notion of relative arbitrariness, we formalize it via conditional Kolmogorov complexity, defining arbitrariness algorithmically as the no rule property. We prove that general relative arbitrariness is formally undecidable, while Arabic relative arbitrariness is decidable and provably less than 1 for its motivated signifiers (Levels W and M), establishing a strict system complexity asymmetry over Indo-European languages.
Chinese Translation
本文探讨了后现代关于无限制语义不确定性的主张及其基础的索绪尔任意符号公理是否与古典阿拉伯语的结构架构相兼容。我们发展了一个阿拉伯非串联形态学的形式数学模型，其中词汇意义由不变根与形态句法模式之间的相互作用决定。在这一框架内，我们建立了形态对应定理，证明每个词汇项都是由根模式对唯一生成的，以及语义定位定理，证明词汇意义在表面实现之前于派生层面上确定。为了应对索绪尔相对任意性的较弱概念，我们通过条件科尔莫哥洛夫复杂性将其形式化，算法上将任意性定义为无规则属性。我们证明一般相对任意性在形式上是不可判定的，而阿拉伯语的相对任意性是可判定的，并且对于其动机符号（W级和M级）证明其小于1，从而在印欧语言中建立了严格的系统复杂性不对称性。
cs.CL / 10 / 2608.07763
Jako Tako or Fluent? Presenting PoVisLE: A Polish Vision-Language Evaluation
Jako Tako 还是 Fluent？呈现 PoVisLE：一个波兰视觉-语言评估
Abstract
Vision-language models (VLMs) have achieved strong performance on tasks such as image captioning, visual question answering, and image-to-text generation. However, they are predominantly trained on English-centric data, which limits their ability to handle culturally grounded visual understanding and leads to failures in interpreting region-specific meanings, symbolic content, and context-dependent visual cues. Existing benchmarks for cultural competence are often template-driven and focused on surface-level recognition, making them insufficient for evaluating deeper linguistic and pragmatic understanding in culturally situated settings. We introduce PoVisLE, a monocultural vision-language benchmark for Polish designed to evaluate culturally grounded multimodal understanding under a grounded evaluation paradigm, where language is interpreted in interaction with visual context. The dataset contains 1,117 images and 2,366 manually annotated VQA pairs. Overall, our dataset provides a controlled and challenging resource for assessing culturally grounded vision-language understanding beyond surface-level recognition.
Chinese Translation
视觉-语言模型（VLMs）在图像描述、视觉问答和图像到文本生成等任务上取得了强劲的表现。然而，它们主要是在以英语为中心的数据上进行训练，这限制了它们处理文化根植的视觉理解的能力，并导致在解释区域特定含义、象征内容和上下文依赖的视觉线索时出现失败。现有的文化能力基准往往是模板驱动的，专注于表层识别，因此不足以评估在文化情境下更深层次的语言和语用理解。我们引入了 PoVisLE，这是一个针对波兰的单文化视觉-语言基准，旨在在一个基于上下文的评估范式下评估文化根植的多模态理解，其中语言与视觉上下文的互动进行解释。该数据集包含1,117张图像和2,366对手动标注的视觉问答（VQA）对。总体而言，我们的数据集提供了一个受控且具有挑战性的资源，用于评估超越表层识别的文化根植的视觉-语言理解。
cs.CL / 11 / 2608.07812
On the use of foundation models in cognitive science
基础模型在认知科学中的应用
Abstract
A host of recent studies have evaluated the cognitive and developmental alignment of Foundation Models (FMs). These investigations include evaluations of their correspondence to adult performance across a range of cognitive domains, as well as whether aspects of model training track children's cognitive development. However, using FMs as candidate cognitive models poses significant methodological and conceptual challenges. A key question underlies this effort: under what conditions does behavioral alignment justify treating FMs as explanatory models of cognition? In this paper, we articulate a four-stage inferential framework for evaluating FMs as cognitive and developmental models: adapting human experimental tasks to model-compatible formats, specifying linking hypotheses that map model outputs to human measures, evaluating behavioral correspondence, and comparing across candidate models or manipulations. We clarify the role of linking hypotheses in mapping model outputs to human behavioral measures, identify challenges that constrain alignment claims, and propose principles for theory-driven and comparative evaluation. Throughout, we argue that behavioral fit alone is insufficient. Alignment becomes scientifically meaningful only when embedded within explicit theoretical commitments, theory-diagnostic tasks, and systematic contrastive evaluation across candidate models.
Chinese Translation
近期大量研究评估了基础模型（Foundation Models, FMs）的认知和发展一致性。这些研究包括评估它们在多个认知领域与成人表现的对应关系，以及模型训练的某些方面是否与儿童的认知发展相匹配。然而，将FMs作为候选认知模型的使用面临着显著的方法论和概念性挑战。一个关键问题是：在什么条件下，行为一致性可以合理化将FMs视为认知的解释模型？在本文中，我们阐述了一个四阶段的推理框架，用于评估FMs作为认知和发展模型的有效性：将人类实验任务调整为与模型兼容的格式，指定将模型输出映射到人类测量的链接假设，评估行为对应性，以及在候选模型或操作之间进行比较。我们澄清了链接假设在将模型输出映射到人类行为测量中的作用，识别了限制一致性主张的挑战，并提出了理论驱动和比较评估的原则。在整个过程中，我们主张，仅仅依赖行为拟合是不够的。只有当一致性嵌入在明确的理论承诺、理论诊断任务和候选模型之间的系统对比评估中时，才具有科学意义。
cs.CL / 12 / 2608.07852
"Many Are My Names": The Anatomy of the Assistant and Its Personas via Sparse Autoencoders
“我有许多名字”：通过稀疏自编码器分析助手及其角色
Abstract
How a language model internally represents who is speaking, the Assistant, an assigned roleplay persona, or a narrated story character, remains underexplored. We study speaker representations using a dataset of user-expressed emotional text and corresponding model responses. We decompose three generation settings (Assistant, Roleplay, and Story) into sparse autoencoder features extracted at turn-boundary and pronoun-token positions and selected through a filtering pipeline for different depths. We characterize each surviving feature through its steering effects and activation distribution. Our main finding is that the Assistant and roleplay personas are not independent alternatives: personas retain the Assistant-associated feature core while progressively differentiating from it across layers, starting from operational machinery towards behavioral and stylistic features. Meanwhile, generated story characters lack the Assistant-associated core. Both Story and Roleplay can be distinguished from the Assistant with Immersive Simulation Mode. However, the Assistant can sometimes enter or slowly drift into it even in the default setting.
Chinese Translation
语言模型如何在内部表示说话者、助手、分配的角色扮演人格或叙述故事中的角色，仍然未被充分探讨。我们使用用户表达的情感文本及相应模型响应的数据集研究说话者表示。我们将三种生成设置（助手、角色扮演和故事）分解为在轮次边界和代词标记位置提取的稀疏自编码器特征，并通过过滤管道选择不同深度的特征。我们通过其引导效应和激活分布来表征每个存活特征。我们的主要发现是，助手和角色扮演人格并不是独立的替代品：人格保留了与助手相关的特征核心，同时在各层之间逐渐与之区分开来，从操作机制开始，向行为和风格特征发展。与此同时，生成的故事角色缺乏与助手相关的核心。故事和角色扮演可以通过沉浸式模拟模式与助手区分开来。然而，即使在默认设置下，助手有时也可以进入或缓慢漂移到这种模式中。
cs.CL / 13 / 2608.07862
SurakshaEval: An Indic Safety Benchmark for Multilingual LLMs
SurakshaEval：一种针对多语言大型语言模型的印度安全基准
Abstract
Existing safety evaluation datasets for large language models (LLMs) predominantly focus on English and Western contexts, often overlooking the linguistic diversity and culturally grounded safety risks present in other languages. To address this gap, we introduce SurakshaEval, a novel safety benchmark composed of human-written prompts spanning real-world scenarios, explicitly designed for ten major Indian languages - Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Punjabi, Tamil, and Telugu, along with English. SurakshaEval includes both generic prompts common across India and region- and language-specific prompts that capture localized sociocultural sensitivities. We benchmark a broad range of state-of-the-art LLMs on SurakshaEval, establish baseline safety performance, and identify recurring failure modes, including over-refusal, missed detection of implicit bias, and insufficient contextual awareness in regionally sensitive settings. Our results show that even strong multilingual LLMs struggle to reliably meet nuanced safety requirements when operating in Indic languages, particularly in native scripts. These findings highlight the urgent need for safety evaluation frameworks that incorporate region-specific data and structured assessment protocols, enabling the development and deployment of AI systems that operate securely, ethically, and in alignment with diverse societal values. Our code and data are available at https://github.com/debobanerjee/SurakshaEval. Warning: This paper contains text that may be offensive or unsafe.
Chinese Translation
现有的大型语言模型（LLMs）安全评估数据集主要集中于英语和西方背景，往往忽视了其他语言中存在的语言多样性和文化根植的安全风险。为了解决这一问题，我们引入了SurakshaEval，一个新颖的安全基准，包含涵盖现实场景的人类撰写的提示，专门为十种主要印度语言设计——阿萨姆语（Assamese）、孟加拉语（Bengali）、古吉拉特语（Gujarati）、印地语（Hindi）、卡纳达语（Kannada）、马拉雅拉姆语（Malayalam）、马拉地语（Marathi）、旁遮普语（Punjabi）、泰米尔语（Tamil）和泰卢固语（Telugu），以及英语。SurakshaEval包括了印度普遍使用的通用提示和捕捉地方性社会文化敏感性的地区和语言特定提示。我们在SurakshaEval上对一系列最先进的LLMs进行了基准测试，建立了基线安全性能，并识别出反复出现的失败模式，包括过度拒绝、未能检测到隐性偏见以及在地区敏感环境中缺乏足够的上下文意识。我们的结果表明，即使是强大的多语言LLMs在处理印度语言时，尤其是在本地书写系统中，也难以可靠地满足细致的安全要求。这些发现突显了迫切需要安全评估框架，以纳入地区特定数据和结构化评估协议，从而促进安全、伦理地开发和部署符合多样社会价值观的人工智能系统。我们的代码和数据可在 https://github.com/debobanerjee/SurakshaEval 获取。警告：本文包含可能冒犯或不安全的文本。
cs.CL / 14 / 2608.07891
Detection of Self-Introductions in Legislative Testimony
立法证词中的自我介绍检测
Abstract
Self-introductions are common in legislative committee testimonies. Successfully detecting them and extracting the speaker's name is enormously helpful in the task of speaker identification in the context of government meetings. In this paper, we present a pipeline for detection of self-introductions in legislative committee testimony using machine learning. We construct a training dataset from 1.54 million utterances spanning five state legislative sessions, apply a name-matching heuristic to generate automatic labels, and train three classifiers: a decision tree, random forest, and XGBoost to find self-introductions and extract the speaker's name. We construct a feature set combining bag-of-words, positional context, structural signals, introductory phrase indicators, and discourse context features. Among the three classifiers, XGBoost achieves the best performance with an F1 score of 0.9747 and the fewest total errors; adding fine-tuned BERT probability features improves this further. As an extension, we score the full candidate dataset with a fine-tuned BERT classifier and add BERT probability outputs as features. This BERT-augmented XGBoost model improves F1 from 0.9747 to 0.9782 and reduces total test errors from 241 to 207. The primary gain over the decision tree baseline (F1 0.9323) is driven by discourse context features and the boosting ensemble strategy; BERT provides a modest complementary signal. Analysis of false positives reveals that a minority are genuine self-introductions mislabeled due to name inconsistencies in the source data, indicating that measured metrics modestly understate true performance.
Chinese Translation
自我介绍在立法委员会证词中很常见。成功检测自我介绍并提取发言者姓名对于在政府会议中进行发言者识别任务极为有帮助。本文提出了一种基于机器学习的立法委员会证词中自我介绍检测的流程。我们从五个州的立法会议中构建了一个包含154万条发言的数据集，应用姓名匹配启发式方法生成自动标签，并训练了三种分类器：决策树、随机森林和XGBoost，以识别自我介绍并提取发言者姓名。我们构建了一个特征集，结合了词袋模型、位置上下文、结构信号、引导短语指示符和话语上下文特征。在这三种分类器中，XGBoost以0.9747的F1得分和最少的总错误率达到了最佳性能；添加微调的BERT概率特征进一步改善了这一结果。作为扩展，我们使用微调的BERT分类器对完整候选数据集进行评分，并将BERT概率输出作为特征添加。这个增强BERT的XGBoost模型将F1从0.9747提高到0.9782，并将总测试错误从241减少到207。相较于决策树基线（F1 0.9323），主要的提升来自话语上下文特征和提升集成策略；BERT提供了适度的补充信号。对假阳性的分析表明，少数是真正的自我介绍因源数据中的姓名不一致而被错误标记，这表明测量的指标在一定程度上低估了真实性能。
cs.CL / 15 / 2608.07968
Thinking Hard, Not Smart: Reasoning Models Fail to Ration Test-Time Compute Across Questions
努力思考，而非聪明思考：推理模型未能在问题间合理分配测试时间计算资源
Abstract
Reasoning language models increasingly use test-time compute to improve performance, but existing evaluations typically study this compute one question at a time. Yet when multiple problems share an end-to-end cost or latency constraint, models must decide how to divide limited inference compute among them. We introduce an exam-style evaluation framework for studying this setting, in which a model must distribute one shared token budget across questions with different difficulty and point values to maximize its total score. Across several open and frontier reasoning models, we find that models fail to allocate a shared budget strategically across questions of varying difficulties and values. Models behave largely as greedy sequential solvers: they prioritize questions by presentation order, front-load effort on early questions, and remain insensitive to value, with these tendencies becoming more pronounced as the number of questions grows. Explicit planning prompts spread compute more evenly but do not produce value- or difficulty-aware prioritization. The same behavioral pattern extends from mathematical to code reasoning. These findings establish global budget allocation as a distinct capability that is not captured by conventional per-question evaluation and remains a challenge for current reasoning models.
Chinese Translation
推理语言模型越来越多地利用测试时间计算来提高性能，但现有评估通常一次研究一个问题的计算。然而，当多个问题共享端到端成本或延迟约束时，模型必须决定如何在它们之间分配有限的推理计算资源。我们引入了一种考试风格的评估框架来研究这一设置，在该框架中，模型必须在不同难度和分值的问题之间分配一个共享的令牌预算，以最大化其总分。在多个开放和前沿推理模型的研究中，我们发现模型未能在不同难度和价值的问题之间战略性地分配共享预算。模型的行为主要表现为贪婪的顺序求解器：它们按呈现顺序优先考虑问题，在早期问题上投入更多努力，并对价值保持不敏感，这种倾向在问题数量增加时变得更加明显。明确的规划提示使计算资源分配更加均匀，但并未产生对价值或难度的敏感优先级。这种行为模式从数学推理扩展到代码推理。这些发现确立了全球预算分配作为一种独特能力，而这一点并未通过传统的逐题评估所捕捉，并且仍然是当前推理模型面临的挑战。
cs.CL / 16 / 2608.08024
Prompt Embedding Probes (PEP): Hallucination Detection in LLMs from Hidden States
提示嵌入探测器（PEP）：从隐藏状态中检测大型语言模型的幻觉
Abstract
Large language models (LLMs) can generate fluent and useful responses but remain prone to hallucinations. We introduce Prompt Embedding Probes (PEP), a white-box method for answer-level hallucination detection from the hidden states of a frozen LLM. PEP extends standard linear probes by augmenting the input with a small number of learnable prompt embeddings. We evaluate PEP on TriviaQA, GSM8K, and MedQA using Qwen3 models at multiple scales. PEP improves hidden-state-based detection over standard linear probes in the main in-distribution setting. We further evaluate PEP for pre-generation prediction, cross-model transfer, and out-of-distribution generalization. PEP remains effective in the pre-generation and cross-model settings, whereas robust cross-dataset transfer remains difficult. These results show that prompt-based adaptation can strengthen hidden-state probing while keeping the backbone frozen and adding only a small number of trainable parameters.
Chinese Translation
大型语言模型（LLMs）能够生成流畅且有用的响应，但仍然容易出现幻觉。我们提出了提示嵌入探测器（PEP），这是一种基于白盒的方法，用于从冻结的LLM的隐藏状态中进行答案级别的幻觉检测。PEP通过用少量可学习的提示嵌入增强输入，扩展了标准线性探测器。我们在TriviaQA、GSM8K和MedQA上评估了PEP，使用不同规模的Qwen3模型。在主要的同分布设置中，PEP在基于隐藏状态的检测上优于标准线性探测器。我们进一步评估了PEP在生成前预测、跨模型迁移和异分布泛化中的表现。PEP在生成前和跨模型设置中仍然有效，而稳健的跨数据集迁移仍然困难。这些结果表明，基于提示的适应可以增强隐藏状态探测，同时保持主干模型的冻结，仅增加少量可训练参数。
cs.CL / 17 / 2608.08059
APEX-VW: A Document-Level English-Spanish Post-Editing Dataset in the Healthcare Domain
APEX-VW：医疗领域的文档级英西后编辑数据集
Abstract
Post-Editing (PE) of Machine Translation (MT) output often involves repeating the same lexical and terminological corrections across many segments, especially in specialised and highly repetitive documents. Despite substantial work on Automatic Post-Editing (APE), most available corpora operate at the sentence level, others are synthetic, and overall not designed to study how corrections propagate in realistic Computer-Assisted Translation (CAT) workflows. This paper presents the APEX-VW (Automatic Post-Editing eXperiments on Virtual Wards) Corpus, a new open English-Spanish (EN-ES) dataset built from recent NHS virtual-ward documents and professional PE in Trados Studio, with controlled MT, terminology, and quality assurance settings. The corpus contains seven document-coherent source texts totalling 42k words, translated with four MT systems representing different paradigms and then post-edited by professional translators. Unlike prior resources such as WMT APE corpora, eSCAPE, MLQE-PE, or LangMark, the dataset preserves document order and CAT-tool context, making it suitable for research on terminology normalisation, correction propagation, and human-in-the-loop translation support. The paper describes the corpus design, data preparation, PE setup, and initial corpus statistics, and positions the resource as a benchmark for document-level APE and propagation-aware assistive tools.
Chinese Translation
机器翻译（MT）输出的后编辑（PE）通常涉及在许多段落中重复相同的词汇和术语修正，尤其是在专业性强且高度重复的文档中。尽管在自动后编辑（APE）方面已有大量研究，但大多数现有语料库仅在句子级别运作，其他则是合成的，整体上并未设计用于研究修正如何在现实的计算机辅助翻译（CAT）工作流程中传播。本文介绍了APEX-VW（虚拟病房上的自动后编辑实验）语料库，这是一个新的开放式英西（EN-ES）数据集，基于最近的英国国家医疗服务体系（NHS）虚拟病房文档和在Trados Studio中的专业后编辑，具有受控的机器翻译、术语和质量保证设置。该语料库包含七个文档一致的源文本，总计42,000个单词，使用四个代表不同范式的机器翻译系统进行翻译，然后由专业翻译人员进行后编辑。与之前的资源如WMT APE语料库、eSCAPE、MLQE-PE或LangMark不同，该数据集保持了文档顺序和CAT工具上下文，使其适合于术语规范化、修正传播和人机协作翻译支持的研究。本文描述了语料库的设计、数据准备、后编辑设置和初步语料库统计，并将该资源定位为文档级APE和传播感知辅助工具的基准。
cs.CL / 18 / 2608.08067
DialectS2S: End-to-End Speech Dialogue Modeling for Low-Resource Chinese Dialects
DialectS2S：针对低资源中文方言的端到端语音对话建模
Abstract
Current end-to-end speech dialogue models are primarily optimized for mainstream languages and remain limited in low-resource dialect scenarios due to the scarcity of dialect speech data. Moreover, during dialect adaptation, the semantic representation space of speech dialogue models continuously evolves, while conventional speech supervision remains unchanged, leading to semantic inconsistency between hidden representations and speech targets and degrading speech stability and naturalness. To address these issues, we propose DialectS2S, an end-to-end speech dialogue model for Chinese dialects. We first develop a scalable dialect speech dialogue synthesis pipeline for efficient data construction. We further introduce a two-stage post-training strategy with self-aligned speech supervision, which aligns the semantic content of speech supervision with the evolved semantic representations of the model to improve dialect speech generation quality. Experimental results show that DialectS2S consistently outperforms existing baselines across multiple Chinese dialects in speech dialogue, achieving substantial improvements in dialect consistency, response quality, and speech intelligibility. Our work provides an efficient and scalable solution for end-to-end speech dialogue modeling in low-resource dialect scenarios. To facilitate future research and practical applications, we fully open-source the DialectS2S framework, including model checkpoints, training datasets, and fine-tuning code.
Chinese Translation
当前的端到端语音对话模型主要针对主流语言进行了优化，而在低资源方言场景中，由于方言语音数据的稀缺，仍然存在局限性。此外，在方言适应过程中，语音对话模型的语义表示空间不断演变，而传统的语音监督保持不变，导致隐藏表示与语音目标之间的语义不一致，从而降低了语音的稳定性和自然性。为了解决这些问题，我们提出了DialectS2S，一种针对中文方言的端到端语音对话模型。我们首先开发了一个可扩展的方言语音对话合成管道，以高效构建数据。进一步引入了一个两阶段的后训练策略，结合自对齐的语音监督，将语音监督的语义内容与模型演变后的语义表示对齐，从而提高方言语音生成的质量。实验结果表明，DialectS2S在多个中文方言的语音对话任务中始终优于现有基线，在方言一致性、响应质量和语音可懂度方面取得了显著改善。我们的工作为低资源方言场景中的端到端语音对话建模提供了一种高效且可扩展的解决方案。为了促进未来的研究和实际应用，我们全面开源了DialectS2S框架，包括模型检查点、训练数据集和微调代码。
cs.CL / 19 / 2608.08082
Commitment Before Realization: When Classifier-Free Guidance Becomes Unnecessary in Masked Diffusion Language Models
实现前的承诺：在掩蔽扩散语言模型中，分类器无关引导何时变得不必要
Abstract
Classifier-free guidance (CFG) is usually kept on throughout masked diffusion language model decoding, although its benefit varies across prompts and over time. We study when CFG is actually needed by comparing, from any partial output, the probability of eventual constraint satisfaction under continued CFG and under base-only continuation. Their difference defines the remaining value of guidance. Guidance dependence is highly prompt-specific. Many prompts already succeed without CFG, while for others it provides no measurable benefit or can be harmful. For prompts that do benefit, the gain is often concentrated early. We define the commitment horizon $\astar$ as the earliest point from which switching all remaining decoding to the base model reduces final success by no more than a chosen tolerance. Under the base model, the corresponding success probability, or committor, is a martingale. To first order, CFG's per-step effect is governed by the covariance between the guidance logit direction and the successor committor. This gives a local account of when guidance can help, but it does not by itself locate the horizon. Among prompts with an observed preterminal horizon, $\astar$ is usually early and varies more within constraint families than between them. Freezing each prompt at its own cross-fitted horizon is noninferior to full CFG on all 13 subtasks at the prespecified margin, even while many tokens remain masked. This separates commitment from realization. The boundary also identifies a later region in which higher parallelism adds only a small cost in constraint success, although fluency still degrades with parallel width. For failed trajectories, reopening committed positions improves recovery in both failure modes.
Chinese Translation
分类器无关引导（CFG）通常在掩蔽扩散语言模型解码过程中始终保持开启，尽管其效益在不同提示和时间上有所不同。我们通过比较在任何部分输出下，继续使用CFG与仅使用基础模型的最终约束满足概率，研究何时实际上需要CFG。它们之间的差异定义了引导的剩余价值。引导依赖性高度依赖于提示。许多提示在没有CFG的情况下已经成功，而对于其他提示，它提供的益处不可测量或可能有害。对于那些确实受益的提示，增益通常集中在早期。我们将承诺视界（commitment horizon）定义为从此时起，将所有剩余解码切换到基础模型所导致的最终成功率降低不超过所选容忍度的最早时刻。在基础模型下，相应的成功概率或承诺者（committor）是一个马尔可夫过程。就第一阶而言，CFG的每步效应受引导logit方向与后继承诺者之间的协方差支配。这提供了一个关于何时引导可以提供帮助的局部解释，但它本身并不能确定视界。在观察到的前终止视界的提示中，承诺视界通常较早，并且在约束家族内的变化大于约束家族之间的变化。在所有13个子任务中，将每个提示在其自身的交叉拟合视界处冻结，在预先指定的边际下并不劣于完全CFG，即使许多标记仍然被掩蔽。这将承诺与实现分开。该边界还识别了一个后期区域，在该区域中，更高的并行性仅增加约束成功的少量成本，尽管流畅性仍然随着并行宽度的增加而下降。对于失败的轨迹，重新开放已承诺的位置在两种失败模式下都改善了恢复。
cs.CL / 20 / 2608.08086
Archer: Adaptive Reuse of Cached Hidden States for Efficient Rollback in Diffusion Language Models
Archer：用于高效回滚的缓存隐藏状态自适应重用
Abstract
Diffusion language models (DLMs) iteratively refine a sequence, allowing earlier predictions to be revised as context evolves. This rollback capability distinguishes them from irreversible autoregressive generation, but makes inference costly. Every denoising update alters the global context, forcing both prompt and response states to be recomputed even though only response tokens are revisable. Key-value (KV) caching could reduce this cost, yet conventional caching assumes immutable historical states and is therefore difficult to reconcile with rollback.In this paper, we introduce Adaptive Reuse of Cached Hidden States for Efficient Rollback (Archer), a training-free KV caching method for rollback-capable DLMs. Archer asymmetrically keeps the mutable response synchronized with the current hypothesis while reusing prompt K/V within a bounded state neighborhood. Although prompt representations also change under bidirectional attention, their token identities remain fixed; bounded reuse therefore amortizes repeated prompt computation without caching mutable response states. It also delays feedback from tentative tokens, reducing premature reinforcement of transient high-confidence errors and giving rollback more opportunity to correct them. Our analysis characterizes prompt reuse as a reversibility-aligned cache boundary, bounds its state-dependent approximation error, and gives a decoder-margin condition for preserving full-refresh decisions.Existing DLM acceleration often trades quality for speed. Archer shifts this frontier, attaining the best mean performance of 33.63% together with a 2.57x mean speedup on the main suite. Across evaluated settings, it improves Pass@1 by up to 3.05 points and reaches up to 2.95x speedup. Controlled analyses connect the quality gain to delayed prompt feedback and validate state-aware refresh. Our code is available at https://github.com/Hxnng/Archer.
Chinese Translation
扩散语言模型（DLMs）通过迭代精炼序列，使得随着上下文的发展，早期的预测可以被修正。这种回滚能力使其与不可逆的自回归生成有所区别，但也使得推理成本较高。每次去噪更新都会改变全局上下文，迫使提示和响应状态都必须重新计算，尽管只有响应标记是可修正的。键值（KV）缓存可以降低这一成本，但传统缓存假设历史状态是不可变的，因此难以与回滚相协调。本文介绍了一种用于回滚能力DLM的无训练KV缓存方法——自适应重用缓存隐藏状态（Archer）。Archer不对称地保持可变响应与当前假设同步，同时在有限的状态邻域内重用提示K/V。尽管在双向注意力下，提示表示也会发生变化，但其标记身份保持不变；因此，有限重用可以摊销重复的提示计算，而无需缓存可变响应状态。它还延迟了来自暂定标记的反馈，减少了对瞬时高置信度错误的过早强化，并为回滚提供了更多纠正机会。我们的分析将提示重用表征为与可逆性对齐的缓存边界，界定其状态相关的近似误差，并给出了保持全刷新决策的解码器边际条件。现有的DLM加速通常在质量与速度之间进行权衡。Archer则改变了这一界限，在主要测试集上实现了33.63%的最佳平均性能，并获得了2.57倍的平均加速。在评估的设置中，它将Pass@1提高了最多3.05分，并达到了最高2.95倍的加速。受控分析将质量提升与延迟的提示反馈联系起来，并验证了状态感知刷新。我们的代码可在https://github.com/Hxnng/Archer获取。
cs.CL / 21 / 2608.08090
Wisdom in Unity: The Role of Multilingual Training in Figurative Language Identification in Proverbs
团结中的智慧：多语言训练在谚语中隐喻语言识别中的作用
Abstract
Although multilingual approaches to figurative language identification are not new, the shift beyond language homogeneous training data requires a clearer understanding of the contribution of translated multilingual supervision. We examine this question using 742 proverb concepts across 6,787 translated instances in seven languages. We evaluate five models, including multilingual encoders and instruction tuned LLMs, under progressively increasing levels of multilingual supervision. Moreover, we introduce a multidimensional annotation framework for proverbs that characterizes them through four complementary figurative forms: Metaphorical, Moral/Advisory, Cause-Effect, and Culture Specific. Our findings show that approximately 50% of the translated multilingual training data is sufficient to achieve near-optimal figurative language identification performance. We further show that combining diverse figurative forms yields the strongest overall performance. A notable finding is that the least frequent figurative form, Culture Specific, exhibits the largest performance gains under multilingual supervision. Furthermore, the Moral/Advisory and Culture Specific forms contribute most to the performance of instruction-tuned LLMs on figurative language identification. These findings motivate multilingual figurative language identification to move beyond metaphor-centric taxonomies toward concept level multidimensional frameworks that explicitly model complementary forms of figurative meaning.
Chinese Translation
尽管多语言隐喻语言识别的方法并不新颖，但超越语言同质训练数据的转变需要更清晰地理解翻译的多语言监督所做的贡献。我们通过分析742个谚语概念和6787个翻译实例，涵盖七种语言，来探讨这个问题。我们评估了五种模型，包括多语言编码器和经过指令调优的LLM（大型语言模型），在逐步增加的多语言监督水平下进行评估。此外，我们引入了一个多维注释框架，用于谚语的特征描述，涵盖四种互补的隐喻形式：隐喻性、道德/建议性、因果关系和文化特定性。我们的研究结果表明，约50%的翻译多语言训练数据足以实现接近最佳的隐喻语言识别性能。我们进一步显示，结合多样的隐喻形式能够产生最强的整体性能。一个显著的发现是，最不常见的隐喻形式——文化特定性，在多语言监督下表现出最大的性能提升。此外，道德/建议性和文化特定性形式对经过指令调优的LLM在隐喻语言识别中的性能贡献最大。这些发现促使多语言隐喻语言识别从以隐喻为中心的分类法向明确建模互补隐喻意义的概念层面多维框架转变。
cs.CL / 22 / 2608.08107
NeuPAT: Neuron-aware Plasticity Allocation Tuning for Language-Preserving MLLMs
NeuPAT：面向神经元的可塑性分配调优以保持语言能力的多模态大语言模型
Abstract
Multimodal expansion of large language models (LLMs) enables new perceptual capabilities but often compromises the language intelligence acquired during pretraining. In this work, we investigate this phenomenon from the perspective of internal adaptation dynamics and discover that neurons in pretrained LLMs exhibit heterogeneous plasticity during multimodal learning: some neurons are critical for preserving language capabilities, while others are more adaptive to multimodal knowledge. Based on this insight, we propose NeuPAT (Neuron-aware Plasticity Allocation Tuning), a lightweight and architecture-agnostic framework that allocates neuron-wise update constraints during multimodal instruction tuning. NeuPAT uses a small-scale probing stage to estimate neuron adaptation patterns and selectively protects language-sensitive neurons while promoting multimodal adaptation through more plastic neurons. Experiments across diverse LLM families demonstrate that NeuPAT recovers 94.5\% of the language capability degradation caused by vanilla tuning on 11 language benchmarks while maintaining comparable multimodal performance, providing an effective approach for capability-preserving multimodal expansion.
Chinese Translation
大语言模型（LLMs）的多模态扩展赋予了新的感知能力，但往往会损害在预训练过程中获得的语言智能。在本研究中，我们从内部适应动态的角度探讨这一现象，并发现预训练LLMs中的神经元在多模态学习过程中表现出异质的可塑性：一些神经元对于保持语言能力至关重要，而其他神经元则更适应多模态知识。基于这一见解，我们提出了NeuPAT（面向神经元的可塑性分配调优），这是一个轻量级且与架构无关的框架，在多模态指令调优过程中分配神经元级别的更新约束。NeuPAT使用小规模探测阶段来估计神经元适应模式，选择性地保护对语言敏感的神经元，同时通过更具可塑性的神经元促进多模态适应。在多种LLM家族上的实验表明，NeuPAT恢复了因普通调优在11个语言基准上造成的94.5%的语言能力下降，同时保持了可比的多模态性能，为能力保持的多模态扩展提供了一种有效的方法。
cs.CL / 23 / 2608.08160
Can LLM Agents Stick to the Script? A Benchmark for Long-Horizon Consistency in Interactive Narratives
大型语言模型代理能否遵循剧本？交互叙事中的长期一致性基准
Abstract
The rapid advancement of Large Language Models (LLMs) is revolutionizing AI for Games by enabling open-ended and fluid interactive storytelling. However, existing research has largely overlooked the critical challenge of maintaining long-horizon logical consistency and narrative integrity against unconstrained user interventions. To address this, we formulate this challenge as Narrative Commitment Preservation (NCP), and take interactive narrative as our testbed. We introduce NCP-Bench, a benchmark of 100 narrative environments derived from movie synopses. Each environment includes a structured narrative specification (trajectory, commitments, and initial facts) that we can automatically check throughout the interaction between the player agent and the narrator agent. Experiments across state-of-the-art LLMs reveal a substantial long-horizon consistency gap: high linguistic quality does not guarantee commitment preservation; even strong models frequently generate logically conflicting content under adversarial interventions, with the best-performing model (GPT-5.2) achieving only 42% survival rate after 20 turns and fact conflict rates ranging from 40% to 68% across models, and only isolated runs satisfying all achievement commitments within the 100-turn limit.
Chinese Translation
大型语言模型（LLMs）的快速发展正在通过实现开放式和流畅的交互式叙事，彻底改变游戏中的人工智能。然而，现有研究在很大程度上忽视了在不受限制的用户干预下，保持长期逻辑一致性和叙事完整性这一关键挑战。为了解决这一问题，我们将这一挑战表述为叙事承诺保持（Narrative Commitment Preservation, NCP），并以交互叙事作为我们的测试平台。我们引入了NCP-Bench，这是一个基于电影梗概衍生的100个叙事环境的基准。每个环境都包含一个结构化的叙事规范（轨迹、承诺和初始事实），我们可以在玩家代理与叙述者代理之间的交互过程中自动检查。对最先进的LLM进行的实验揭示了显著的长期一致性差距：高语言质量并不保证承诺的保持；即使是强大的模型，在对抗性干预下也经常生成逻辑冲突的内容，表现最佳的模型（GPT-5.2）在20轮后仅达到42%的存活率，模型间的事实冲突率在40%到68%之间，且仅有个别运行在100轮限制内满足所有成就承诺。
cs.CL / 24 / 2608.08164
STEMMA: An Adversarial Multi-Agent Framework for Evaluating Self-Identity Consistency in LLMs
STEMMA：一种用于评估大型语言模型自我身份一致性的对抗性多智能体框架
Abstract
Knowledge Distillation is a widely adopted technique in the training and fine-tuning of large language models (LLMs) enabling transfer of structured information and functional behavior from a large teacher model to a smaller student model while significantly reducing computational costs. However, as the use of distillation increases in both scale and complexity it raises an important question about what kind of knowledge is really transferred from the teacher model. In this work, we argue that apart from the functional knowledge, student models also learn behavioral patterns, specifically how a model represents its own identity raising concerns about output homogeneity, model biases, and accountability. To address this challenge, we introduce STEMMA, a multi-modal and multi-agent framework in which role specific agents collaboratively probe self identification behavior in different models. We also contribute a set of adversarial prompts designed manually to evaluate identity consistency in LLMs. Our results show that to an extent most models are vulnerable to inconsistencies in self-representations.
Chinese Translation
知识蒸馏是一种广泛采用的技术，用于大型语言模型（LLMs）的训练和微调，使得结构化信息和功能行为能够从大型教师模型转移到较小的学生模型，同时显著降低计算成本。然而，随着蒸馏在规模和复杂性上的增加，关于教师模型究竟转移了什么样的知识这一重要问题也随之提出。在本研究中，我们认为，除了功能知识外，学生模型还学习到行为模式，特别是模型如何表示自身身份，这引发了关于输出同质性、模型偏见和问责制的担忧。为了解决这一挑战，我们引入了STEMMA，一个多模态和多智能体框架，其中角色特定的智能体协同探讨不同模型中的自我识别行为。我们还贡献了一组手动设计的对抗性提示，用于评估LLMs中的身份一致性。我们的结果表明，在一定程度上，大多数模型对自我表征的不一致性是脆弱的。
cs.CL / 25 / 2608.08168
Thinking vs. NoThinking: Towards Interpreting Reasoning Mechanisms of Large Language Models via Sparse Autoencoders
思考与非思考：通过稀疏自编码器解释大型语言模型的推理机制
Abstract
While Large Language Models (LLMs) employing Chain-of-Thought (CoT) exhibit superior reasoning capabilities, the neural mechanisms distinguishing this explicit Thinking mode from direct answer generation (NoThinking mode) remain poorly understood. To deconstruct this cognitive process, we apply Top-K Sparse Autoencoders (SAEs) to the intermediate representations of DeepSeek-R1-Distill-Qwen-7B and examine the model's divergent behaviors across math-solving tasks of three distinct difficulty levels. Observationally, we identify a clear distinction in how the model functions under two reasoning modes: Thinking mode relies on sparse and high-intensity feature activations driving verbal deduction independent of problem complexity, whereas NoThinking mode exhibits an adaptive and diffuse pattern prioritizing symbolic manipulation. Causally, suppressing the three most active sparse features by Total Activation Volume reveals three principles: (i) reasoning and syntactic structure are tightly coupled, as interventions consistently degrade \LaTeX{} and boxed-solution formatting; (ii) Thinking responds to disruption with compensatory over-generation marked by increased metacognitive cues and repetitive, low-information continuations; and (iii) coherent CoT behavior depends on a fragile coordination among specialized features, yielding distinct failure modes under perturbation but a consistently impaired output structure.
Chinese Translation
尽管采用思维链（Chain-of-Thought, CoT）的大型语言模型（Large Language Models, LLMs）展现出卓越的推理能力，但区分这种显式思考模式与直接答案生成（非思考模式）的神经机制仍然不甚清晰。为了拆解这一认知过程，我们将Top-K稀疏自编码器（Sparse Autoencoders, SAEs）应用于DeepSeek-R1-Distill-Qwen-7B的中间表示，并考察模型在三种不同难度级别的数学解题任务中的不同表现。从观察上看，我们发现模型在两种推理模式下的功能存在明显区别：思考模式依赖于稀疏且高强度的特征激活，推动独立于问题复杂性的语言推理，而非思考模式则表现出一种适应性和扩散的模式，优先考虑符号操作。从因果上看，通过总激活体积（Total Activation Volume）抑制三个最活跃的稀疏特征揭示了三个原则：（i）推理与句法结构紧密耦合，因为干预始终会降低 extbackslash LaTeX{}和框选解答格式；（ii）思考模式对干扰的反应表现为补偿性过度生成，特征为增加的元认知线索和重复的低信息延续；（iii）连贯的思维链行为依赖于专门特征之间脆弱的协调，在扰动下产生不同的失败模式，但输出结构始终受到损害。
cs.CL / 26 / 2608.08180
A Grounded and Decomposed Framework for Relation-Level Hallucination Evaluation in Abstractive Summarization
一种基于实证和分解的框架，用于抽象摘要中的关系级幻觉评估
Abstract
Abstractive text summarization systems frequently generate fluent yet unfaithful summaries by fabricating or distorting relationships between entities and events. Such relation-level hallucinations undermine the reliability of generated summaries, particularly in high-stakes domains. In this work, we present a refined and grounded framework for evaluating relation hallucination in abstractive summarization. We present the empirical Relation Hallucination Index (RHI) by introducing a dependency-aware relation extraction algorithm that incorporates lemmatization-based normalization, named entity grounded subject resolution, passive agent recovery, negation-aware verb modeling, reporting verb filtering, nominal relation fallback, clausal propagation, and systematic deduplication. These enhancements improve the structural fidelity of extracted relation triples and reduce spurious matches during evaluation. In addition, we introduce a normalized formulation of RHI to ensure scale-invariant comparison between datasets and models. The revised metric decomposes hallucination into interpretable components, aggregates relation hallucination metric into a normalized relation faithfulness score. Extensive evaluation across multiple state-of-the-art summarization models demonstrates that the grounded extraction process yields more stable and discriminative hallucination measurements. The proposed framework advances automated relation-level faithfulness evaluation and supports coherence-aware, hallucination-sensitive model analysis.
Chinese Translation
抽象文本摘要系统经常生成流畅但不真实的摘要，通过虚构或扭曲实体和事件之间的关系。这种关系级幻觉削弱了生成摘要的可靠性，特别是在高风险领域。在本研究中，我们提出了一个精炼且基于实证的框架，用于评估抽象摘要中的关系幻觉。我们通过引入一种依赖关系感知的关系提取算法，提出了经验关系幻觉指数（Relation Hallucination Index, RHI），该算法结合了基于词形还原的规范化、命名实体基础的主题解析、被动主体恢复、否定感知动词建模、报告动词过滤、名词关系回退、从句传播和系统去重等增强措施。这些改进提高了提取关系三元组的结构保真度，并减少了评估过程中的虚假匹配。此外，我们引入了RHI的规范化形式，以确保数据集和模型之间的尺度不变比较。修订后的指标将幻觉分解为可解释的组成部分，将关系幻觉指标汇总为规范化的关系可信度评分。在多个最先进的摘要模型上进行的广泛评估表明，基于实证的提取过程产生了更稳定和具有区分性的幻觉测量。所提出的框架推动了自动化的关系级可信度评估，并支持对连贯性敏感的幻觉分析。
cs.CL / 27 / 2608.08227
Focus particles and scalar inferences across humans and language models
聚焦粒子与人类和语言模型中的标量推理
Abstract
Focus particles such as "even" and "only" are central to formal semantic theories that posit structured representations over sets of alternatives. "Even" highlights unexpected or extreme alternatives, while "only" enforces exclusivity. If such scalar representations are robust and generalizable, they should give rise to consistent judgments across contexts and systems. In this work, we test whether humans and large language models (LLMs) construct stable scalar representations from sentences containing these particles. Using a dataset of approximately 100 items, participants and models were asked to make scalar judgments. Preliminary results suggest that similar outputs across humans and LLMs may arise from different underlying mechanisms.
Chinese Translation
聚焦粒子如“甚至”（even）和“仅仅”（only）在形式语义理论中占据核心地位，这些理论假设在替代集上存在结构化的表征。“甚至”强调意外或极端的替代，而“仅仅”则强调排他性。如果这样的标量表征是稳健且可推广的，它们应该在不同的上下文和系统中产生一致的判断。在本研究中，我们测试了人类和大型语言模型（LLMs）是否能够从包含这些粒子的句子中构建稳定的标量表征。使用约100个项目的数据集，参与者和模型被要求进行标量判断。初步结果表明，人类和LLMs之间的相似输出可能源于不同的潜在机制。
cs.CL / 28 / 2608.08256
AraSSM: A bidirectional state-space encoder for Arabic masked language modeling
AraSSM：一种用于阿拉伯语掩码语言建模的双向状态空间编码器
Abstract
Pretrained Transformer encoders such as AraBERT, MARBERT, and CAMeLBERT have become the standard backbone for Arabic natural language understanding, but their self-attention mechanism scales quadratically with sequence length, which limits efficiency on long documents. Mamba, a selective state-space model (SSM), offers linear-time sequence modeling as a competitive alternative to attention, yet no dedicated bidirectional Mamba encoder pretrained specifically for Arabic currently exists. We introduce AraSSM, a bidirectional Mamba encoder pretrained via masked language modeling on a corpus combining Arabic Wikipedia and CulturaX text, trained end-to-end on four consumer-grade NVIDIA RTX 2080Ti GPUs (11GB) over approximately ten days. We evaluate AraSSM by fine-tuning on four established Arabic NLU benchmarks covering sentiment classification (HARD), named entity recognition (ANERcorp), extractive question answering (ARCD), and natural language inference (XNLI-ar), following the per-task evaluation protocol introduced by AraBERT, and report results as mean +/- standard deviation across three fine-tuning seeds. AraSSM matches or exceeds published base-sized Transformer baselines on sentiment classification (96.37 +/- 0.03% accuracy on HARD), is competitive on extractive QA (32.19 +/- 1.07 EM, 63.79 +/- 0.25 F1 on ARCD) and named entity recognition (81.54 +/- 0.30 entity-level F1 on ANERcorp), and trails the base-sized Transformer range on natural language inference (72.83 +/- 0.07% accuracy on XNLI-ar), despite being trained entirely from scratch on consumer hardware rather than large-scale accelerator clusters.
Chinese Translation
预训练的Transformer编码器，如AraBERT、MARBERT和CAMeLBERT，已成为阿拉伯自然语言理解的标准基础，但它们的自注意力机制在序列长度上呈二次方扩展，这限制了在长文档上的效率。Mamba是一种选择性状态空间模型（SSM），提供线性时间的序列建模，作为注意力机制的竞争性替代方案，但目前尚不存在专门为阿拉伯语预训练的双向Mamba编码器。我们介绍了AraSSM，这是一种通过掩码语言建模在结合阿拉伯维基百科和CulturaX文本的语料库上预训练的双向Mamba编码器，使用四个消费级NVIDIA RTX 2080Ti GPU（11GB）进行端到端训练，历时约十天。我们通过在四个已建立的阿拉伯语自然语言理解基准上进行微调来评估AraSSM，这些基准涵盖情感分类（HARD）、命名实体识别（ANERcorp）、抽取式问答（ARCD）和自然语言推理（XNLI-ar），遵循AraBERT提出的每个任务评估协议，并报告三个微调种子的均值±标准差结果。AraSSM在情感分类上与已发布的基础大小Transformer基线相匹配或超越（在HARD上准确率为96.37±0.03%），在抽取式问答（在ARCD上EM为32.19±1.07，F1为63.79±0.25）和命名实体识别（在ANERcorp上实体级F1为81.54±0.30）上具有竞争力，尽管在自然语言推理上（在XNLI-ar上准确率为72.83±0.07%）落后于基础大小Transformer范围，然而其完全是在消费硬件上从零开始训练，而非在大规模加速器集群上进行训练。
cs.CL / 29 / 2608.08283
Do Evaluation Metrics Detect Errors in Classical Chinese to English Translations?
评估指标能否检测古汉语到英语翻译中的错误？
Abstract
Although large language models can translate some historical languages surprisingly well, their usefulness in digital humanities workflows is limited by the lack of reliable evaluation. We investigate whether existing automatic evaluation metrics developed for modern languages are reliable in this setting, using translation from Classical Chinese to English as a test case. We introduce a diagnostic framework based on minimal pairs capturing error types salient in scholarly use, probing both reference-based and reference-free metrics for error sensitivity and tolerance to valid variation. We find that all metrics exhibit blind spots, however MetricX24 performs best overall. Our findings highlight the need for more robust and interpretable metrics for historically and culturally distinct translation settings.
Chinese Translation
尽管大型语言模型在翻译某些历史语言方面表现出令人惊讶的能力，但由于缺乏可靠的评估，其在数字人文学科工作流程中的实用性受到限制。我们研究了现有为现代语言开发的自动评估指标在这一环境中的可靠性，以古汉语到英语的翻译作为测试案例。我们引入了一个基于最小对的诊断框架，捕捉在学术使用中显著的错误类型，探讨了基于参考和非基于参考的指标在错误敏感性和对有效变异的容忍度方面的表现。我们发现所有指标都存在盲点，但MetricX24的整体表现最佳。我们的研究结果强调了在历史和文化上独特的翻译环境中，需要更稳健和可解释的评估指标。
cs.CL / 30 / 2608.08383
Safety Cost of Steering Vectors Is Separable and Reducible
引导向量的安全成本是可分离和可减少的
Abstract
Steering vectors are a lightweight tool for controlling LLM behavior. However, emerging evidence shows that steering vectors can unintentionally compromise a model's safety mechanisms and increase compliance with harmful requests, while no effective mitigation yet exists. In this work, we show that this safety degradation arises from a separable component in the vector that disrupts the model's safety mechanisms but contributes little to the steering objective. We identify and remove this safety-degrading component, formulating the task as a constrained optimization problem solved through primal-dual updates, subject to preserving the intended steering effect and bounding false refusal. The resulting solution is both interpretable and surgical: the optimization recovers a single direction whose ablation from the steering vector restores model safety with minimal utility cost. Across models, steering behaviors, and attack suites, including unseen attacks types, our method substantially reduces steering-induced safety degradation while preserving the original steering effect with minimal impact on false refusal. Our method offers a post-hoc correction to steering vectors that mitigates their safety cost, and more broadly, it provides a general recipe for applying activation-level model interventions without paying a safety tax.
Chinese Translation
引导向量是一种轻量级工具，用于控制大型语言模型（LLM）的行为。然而，最新证据表明，引导向量可能会无意中破坏模型的安全机制，并增加对有害请求的遵从性，而目前尚无有效的缓解措施。在本研究中，我们展示了这种安全退化源于向量中的一个可分离成分，该成分干扰了模型的安全机制，但对引导目标的贡献很小。我们识别并去除了这一安全退化成分，将任务表述为一个约束优化问题，通过原始-对偶更新进行求解，要求保留预期的引导效果并限制虚假拒绝。最终的解决方案既可解释又具针对性：优化恢复了一个单一方向，其从引导向量中去除后能够以最小的效用成本恢复模型安全。在不同模型、引导行为和攻击套件（包括未见过的攻击类型）中，我们的方法显著减少了引导引起的安全退化，同时以最小的影响保留了原始的引导效果。我们的方法为引导向量提供了一种事后修正，减轻了它们的安全成本，更广泛地说，它为在不支付安全税的情况下应用激活级别模型干预提供了一般性方案。
cs.CL / 31 / 2608.08447
Hidden Language Consistency Phenomena in Reasoning LLMs
推理大型语言模型中的隐性语言一致性现象
Abstract
Multilingual reasoning models are commonly evaluated by whether they arrive at the correct answer, but not by whether they preserve the intended language while reasoning and responding. This omission conceals important multilingual behaviors that emerge as tasks become harder. In this paper, we study task difficulty, task accuracy, thinking-language consistency (TC), and answer-language consistency (AC) across reasoning models using PolyMath benchmark in eight languages and four difficulty levels. We uncover four findings: (1) language consistency exhibits four difficulty-dependent behaviors: output-language consistency remains aligned with input, remains misaligned, degrades gradually, or collapses abruptly. (2) We identify the language consistency breakdown effect, where increasing difficulty can cause a sudden drop in output-language consistency, especially in less strongly represented and non-Latin-script languages. (3) Due to this breakdown effect, accuracy can be preserved or even improved at a harder difficulty level as the model shifts to its internal dominant language. (4) Quantization can improve or degrade output-language consistency independently of its effect on accuracy, with GPTQ and AWQ often outperforming AutoRound under tolerance-based voting with {\epsilon} = 1.0. These results show that multilingual capability cannot be characterized by accuracy alone; reliable evaluation should jointly consider task accuracy, language consistency, and task difficulty for multilingual benchmarks.
Chinese Translation
多语言推理模型通常通过是否得出正确答案来评估，但并未考虑在推理和回应过程中是否保持了预期语言。这一遗漏掩盖了随着任务难度加大而出现的重要多语言行为。本文研究了在八种语言和四个难度级别下，使用PolyMath基准测试的推理模型中的任务难度、任务准确性、思维语言一致性（TC）和回答语言一致性（AC）。我们揭示了四个发现：(1) 语言一致性表现出四种依赖于难度的行为：输出语言一致性与输入保持一致、保持不一致、逐渐下降或突然崩溃。(2) 我们识别了语言一致性崩溃效应，即随着难度增加，输出语言一致性可能会突然下降，尤其是在代表性较弱和非拉丁字母语言中。(3) 由于这一崩溃效应，准确性可以在更高的难度级别上保持甚至提高，因为模型转向其内部主导语言。(4) 量化可以独立于其对准确性的影响来改善或降低输出语言一致性，其中GPTQ和AWQ在基于容忍度的投票下（{ ext{ε}} = 1.0）通常优于AutoRound。这些结果表明，多语言能力不能仅通过准确性来表征；可靠的评估应共同考虑任务准确性、语言一致性和任务难度，以便进行多语言基准测试。
cs.CL / 32 / 2608.08451
Calling the Bluff: Detecting Ever-Shifting Harmful Chat Dialogue via Ordered Reasoning Chain Regularization
揭穿虚假：通过有序推理链正则化检测不断变化的有害聊天对话
Abstract
Harmful chat dialogues are ever-shifting through type-shifting and lexical evasion, yet we find they share invariant principles, i.e., an Ordered Reasoning Chain (ORC) of recurring topics, harm language indicators, severity hierarchies, and type characteristics, which can help us capture the key information in the frequently changing lexical expressions. We propose BRACE, which encodes the ORC as four differentiable stages (Topic -> Indicator -> Severity -> Type) with intermediate supervision, serving as a structured regularizer blended with direct heads, and supported by prototype-based feature augmentation and feature path disentanglement. The evaluation results show that, across 4 domains and 5 harm categories, BRACE achieves harm-type macro F1 of 0.934 (RoBERTa-wwm-ext, 3-seed mean), with decoder backbones (Qwen3-1.7B LoRA) reaching 0.949. Ablation studies show that all components contribute to BRACE, and the structural decomposition of ORC enables BRACE to distinguish harmful types with semantic ambiguity. Disclaimer: This paper may contain content that is disturbing to some readers.
Chinese Translation
有害聊天对话通过类型转换和词汇规避不断变化，但我们发现它们共享不变的原则，即一个有序推理链（Ordered Reasoning Chain, ORC），包括反复出现的主题、有害语言指标、严重性等级和类型特征，这有助于我们捕捉频繁变化的词汇表达中的关键信息。我们提出了BRACE，它将ORC编码为四个可微分的阶段（主题 -> 指标 -> 严重性 -> 类型），并通过中间监督提供支持，作为一种与直接头部相结合的结构化正则化器，并辅以基于原型的特征增强和特征路径解耦。评估结果显示，在4个领域和5个有害类别中，BRACE实现了0.934的有害类型宏F1值（RoBERTa-wwm-ext，3次种子均值），而解码器骨干（Qwen3-1.7B LoRA）达到了0.949。消融研究表明，所有组件都对BRACE有所贡献，而ORC的结构分解使BRACE能够区分具有语义模糊性的有害类型。免责声明：本文可能包含对某些读者造成困扰的内容。
cs.CL / 33 / 2608.08459
Beyond Tables: Doc2DB-Bench for Relationally Faithful Document-to-Database Construction
超越表格：Doc2DB-Bench用于关系忠实的文档到数据库构建
Abstract
Practical AI systems increasingly need to turn long, heterogeneous documents into queryable relational databases, not isolated spreadsheets. In domains such as finance, healthcare, education, transportation, and enterprise operations, downstream workflows rely on normalized schemas, entity identities, keys, cross-table relationships, and integrity constraints for analytics, compliance, auditing, and SQL-backed decision making. Existing Document-to-Table benchmarks are insufficient for this setting: flattening evidence into single tables can duplicate entities, obscure many-to-many relationships, create sparse records, and avoid testing whether extracted facts form a valid database instance. This creates an urgent need to evaluate document understanding as database construction rather than field extraction. We introduce Doc2DB-Bench, a benchmark for Document-to-Database construction, containing 203 long-document instances across 42 schemas and seven domain groups, with 117 entity tables, 132 relationship tables, 7,341 rows, and 41,935 cells. Built through a controllable DB-to-Doc synthesis pipeline and organized by a taxonomy of intra-table extraction and inter-table reasoning, the generated documents undergo authenticity verification, proving indistinguishable from real-world references. Doc2DB-Bench thus provides a testbed for reliable, auditable, and relationally faithful LLM-based data systems. The benchmark is publicly available at https://github.com/SetonLiang/Doc2DB-Bench.
Chinese Translation
实际的人工智能系统越来越需要将长篇异构文档转化为可查询的关系数据库，而不是孤立的电子表格。在金融、医疗、教育、交通和企业运营等领域，下游工作流程依赖于规范化的模式、实体身份、键、跨表关系和完整性约束，以进行分析、合规、审计和基于SQL的决策。现有的文档到表格基准测试在这一环境下显得不足：将证据扁平化为单个表格可能会重复实体、模糊多对多关系、创建稀疏记录，并且未能测试提取的事实是否形成有效的数据库实例。这使得将文档理解评估为数据库构建而非字段提取的需求变得迫切。我们引入了Doc2DB-Bench，这是一个用于文档到数据库构建的基准，包含42个模式和七个领域组中的203个长文档实例，具有117个实体表、132个关系表、7,341行和41,935个单元格。通过可控的数据库到文档合成管道构建，并按表内提取和表间推理的分类法组织，生成的文档经过真实性验证，证明与现实世界参考文献无法区分。因此，Doc2DB-Bench为可靠、可审计和关系忠实的基于大型语言模型的数据系统提供了一个测试平台。该基准公开可用，网址为https://github.com/SetonLiang/Doc2DB-Bench。
cs.CL / 34 / 2608.08477
VectraYX-Vision-1B: A Sub-2B Spanish/LATAM Cybersecurity Vision-Language Model with Structured Visual Reasoning and Native Tool Use
VectraYX-Vision-1B：一种具有结构化视觉推理和原生工具使用的亚2B西班牙语/拉美网络安全视觉语言模型
Abstract
We present VectraYX-Vision-1B, a sub-2B vision-language model (VLM) for Spanish/LATAM cybersecurity imagery, coupling a frozen SigLIP-so400m encoder to a 1.04B Spanish/LATAM security decoder via an MLP. To our knowledge, it is the first sub-2B VLM specialized for cyber UI (IDA, Ghidra, Wireshark, Nmap, Metasploit, Volatility) that answers in Spanish, emits structured reasoning via native <|think|> tokens, invokes tools via Model Context Protocol (<|tool_call|>), and exports to llama.cpp's LLaVA mmproj format for air-gapped deployment. We report a negative preliminary visual-grounding result: despite fully functional pipelines, the current vision SFT (400-1900 steps, ~16M tokens) yields near-zero B6 scores (0.08 tool-identification), ignoring image content. We specify remediation (longer SFT, >=60% replay, lower LR) and expose a checkpoint-loader bug (unstripped llm. prefix) masquerading as training collapse. Crucially, we introduce a 3-variant ablation matrix (V0: NoPE-every-4, V1: all-RoPE, V2: NoPE+learned 2D) to study if periodic no-positional-encoding (NoPE) layers help or hurt attention over the 729-token visual block. Code, configs, and weights are released to establish priority on this architectural question. We provide B1-B5 for the text backbone, text controls, preliminary B6/B7 scores, wall times, GGUF efficiency on CPU, and a corpus of 14,596 QA pairs across 10 domains. We open-source all models and trajectories: jsantillana/vectrayx-1b, jsantillana/vectrayx-vision-1b, and jsantillana/vectrayx-vision-1b-checks.
Chinese Translation
我们提出了VectraYX-Vision-1B，这是一种针对西班牙语/拉美网络安全图像的亚2B视觉语言模型（VLM），通过一个冻结的SigLIP-so400m编码器与一个1.04B的西班牙语/拉美安全解码器通过多层感知器（MLP）相连。据我们所知，这是首个专门针对网络用户界面（IDA、Ghidra、Wireshark、Nmap、Metasploit、Volatility）的亚2B VLM，能够用西班牙语回答问题，通过原生<|think|>标记发出结构化推理，通过模型上下文协议（<|tool_call|>）调用工具，并导出到llama.cpp的LLaVA mmproj格式以实现隔离部署。我们报告了一个负面的初步视觉定位结果：尽管管道完全功能正常，当前的视觉微调（SFT，400-1900步，约16M标记）产生近乎零的B6分数（0.08工具识别），忽略了图像内容。我们指定了补救措施（更长的SFT，>=60%重放，较低的学习率）并暴露了一个检查点加载器错误（未剥离的llm.前缀），伪装成训练崩溃。关键是，我们引入了一个三变体消融矩阵（V0：NoPE-every-4，V1：all-RoPE，V2：NoPE+学习的2D）来研究周期性无位置编码（NoPE）层是否有助于或损害729标记视觉块的注意力。我们提供了文本主干、文本控制、初步B6/B7分数、墙时、CPU上的GGUF效率以及跨10个领域的14,596个问答对的语料库的B1-B5。我们开源了所有模型和轨迹：jsantillana/vectrayx-1b，jsantillana/vectrayx-vision-1b，以及jsantillana/vectrayx-vision-1b-checks。
cs.CL / 35 / 2608.08510
From Speech to Interaction: Analyzing Multimodal Systems in Cocktail-Party Scenarios
从语音到互动：鸡尾酒会场景中的多模态系统分析
Abstract
Humans have the remarkable ability to engage in spontaneous informal conversations and selectively attend to individual speakers while filtering out competing speech from nearby conversations. This "cocktail party" scenario still presents severe challenges to speech recognition systems. The CHiME-9 MCoRec task provides a testbed where systems must recognize groups of speakers and transcribe each of their conversations from audio-visual input. In this work, we analyze a diverse set of systems, representing different design directions for addressing the cocktail-party scenario, where the best system achieves up to 57% relative error reduction. We identify three main strategies: (1) explicit or implicit audio-visual target speech separation, (2) improved audio-visual speech recognition for each target speaker, and (3) the use of large language models to group speakers into conversations and enhance conversational consistency. Our analysis shows that these directions address complementary failure modes of the cocktail-party problem, and that high speech overlap alone does not explain performance differences, challenging the common assumption that overlap is the primary source of difficulty in cocktail-party recognition.
Chinese Translation
人类具有在自发非正式对话中进行交流的卓越能力，能够选择性地关注个别发言者，同时过滤掉来自附近对话的竞争性语音。这种“鸡尾酒会”场景对语音识别系统仍然提出了严峻的挑战。CHiME-9 MCoRec任务提供了一个测试平台，系统必须从音频-视觉输入中识别发言者群体并转录他们的对话。在本研究中，我们分析了一组多样化的系统，代表了应对鸡尾酒会场景的不同设计方向，其中表现最佳的系统实现了高达57%的相对错误率降低。我们识别出三种主要策略：(1) 显式或隐式的音频-视觉目标语音分离，(2) 针对每个目标发言者改进的音频-视觉语音识别，以及 (3) 使用大型语言模型将发言者分组为对话并增强对话的一致性。我们的分析表明，这些方向解决了鸡尾酒会问题的互补失败模式，并且仅仅高语音重叠并不能解释性能差异，这挑战了重叠是鸡尾酒会识别中主要困难来源的普遍假设。
cs.CL / 36 / 2608.08557
OpenVisTool: An Open Recipe for Synthesizing Instructive Visual Tool-Use Trajectories
OpenVisTool：合成指导性视觉工具使用轨迹的开放方案
Abstract
Visual tool use has emerged as a fundamental capability for multimodal agents to actively acquire evidence beyond a fixed image encoding. The prevailing recipe learns this capability from teacher-generated trajectories filtered for answer correctness, implicitly assuming that every successful demonstration provides effective supervision. We argue this assumption is flawed: a strong teacher often reaches the correct answer without needing its tool calls, and imitating such trajectories teaches a student that tool calls accompany correct answers, not that tool observations ground them. We present OpenVisTool, an open framework for constructing instructive visual tool-use trajectories that provide effective supervision for tool learning. The key insight is that a trajectory should be retained only if its answer is correct (outcome validity) and its tool observations causally contribute to that answer (causal utility). The framework operates in three stages: difficulty screening to select queries that are not reliably answerable without tools, domain-specific trajectory synthesis to elicit coherent tool-use trajectories, and supervision verification to jointly test both conditions. Rather than encouraging models to imitate tool calls, the resulting supervision teaches when and how visual evidence should be acquired. Using this framework, we construct OpenVisTool-42K, a dataset spanning five visual reasoning domains, together with OpenVisTool-Bench, a benchmark covering the same domains. Across four backbones (4B-27B), fine-tuning on OpenVisTool-42K consistently improves visual tool-use performance and yields gains on two out-of-distribution benchmarks; the larger models approach leading closed-source systems. The evidence suggests that effective visual tool use is learned from causally grounded supervision rather than tool-calling patterns.
Chinese Translation
视觉工具使用已成为多模态智能体主动获取超越固定图像编码的证据的基本能力。现有的方法从教师生成的、经过答案正确性筛选的轨迹中学习这一能力，隐含地假设每一个成功的演示都提供了有效的监督。我们认为这一假设是有缺陷的：一个优秀的教师往往在不需要其工具调用的情况下就能得到正确答案，而模仿这样的轨迹会教会学生工具调用与正确答案相伴随，而不是工具观察支撑它们。我们提出了OpenVisTool，一个开放框架，用于构建提供有效监督的指导性视觉工具使用轨迹。关键的见解是，只有在答案正确（结果有效性）且其工具观察对该答案有因果贡献（因果效用）的情况下，轨迹才应被保留。该框架分为三个阶段：困难筛选以选择那些在没有工具的情况下无法可靠回答的查询，特定领域的轨迹合成以引发连贯的工具使用轨迹，以及监督验证以共同测试这两个条件。与其鼓励模型模仿工具调用，所产生的监督教会了何时以及如何获取视觉证据。利用该框架，我们构建了OpenVisTool-42K，一个涵盖五个视觉推理领域的数据集，以及OpenVisTool-Bench，一个覆盖相同领域的基准。在四个基础模型（4B-27B）上，对OpenVisTool-42K的微调持续提高了视觉工具使用的性能，并在两个分布外基准上取得了提升；更大的模型接近领先的闭源系统。证据表明，有效的视觉工具使用是从因果基础的监督中学习的，而不是工具调用模式。
cs.CL / 37 / 2608.08606
Mitigating Gender Bias in English to Romanian Machine Translation
减轻英语到罗马尼亚语机器翻译中的性别偏见
Abstract
Machine translation (MT) systems often fail to correctly translate gender, especially when converting from a gender-neutral language like English to a gendered target language such as Romanian. This bias results in translations that default to masculine forms or reinforce gender stereotypes. We propose a hybrid pipeline to mitigate this issue by combining large language model (LLM)-based gender classification with neural machine translation (NMT). Our system uses a fine-tuned LLM to detect the intended gender of target words in English sentences and insert inline gender hint tags. These tagged sentences are then passed to a Transformer model fine-tuned to generate morphologically correct Romanian translations. To support this, we introduce three novel datasets for gender disambiguation and translation. Our approach improves gender accuracy on the WinoMT and WinoGender benchmarks by over 40 percentage points compared to a baseline MT system. This is the first method to explicitly address and evaluate gender bias in English-Romanian MT using both LLM inference and tag-aware translation.
Chinese Translation
机器翻译（MT）系统在性别翻译方面常常存在不足，尤其是在将性别中立的语言（如英语）转换为性别化的目标语言（如罗马尼亚语）时。这种偏见导致翻译结果默认使用男性形式或强化性别刻板印象。我们提出了一种混合管道，通过结合基于大型语言模型（LLM）的性别分类与神经机器翻译（NMT）来缓解这一问题。我们的系统使用经过微调的LLM来检测英语句子中目标词的预期性别，并插入内联性别提示标签。这些带标签的句子随后被传递给经过微调的Transformer模型，以生成形态学上正确的罗马尼亚语翻译。为支持这一过程，我们引入了三个用于性别消歧和翻译的新数据集。与基线MT系统相比，我们的方法在WinoMT和WinoGender基准上的性别准确性提高了超过40个百分点。这是首个明确解决并评估英语-罗马尼亚语MT中性别偏见的方法，采用了LLM推理和标签感知翻译。
cs.CL / 38 / 2608.08607
North Africa's Missing Framework: NLP-Driven Mental Healthcare in Algeria and Implications for Low-resource Settings
北非缺失的框架：阿尔及利亚基于自然语言处理的心理健康护理及其在低资源环境中的影响
Abstract
Mental health disorders are a leading cause of disability worldwide, yet Natural Language Processing (NLP) research for mental healthcare has remained concentrated in high-income, English-language settings. North Africa, and Algeria in particular, is largely absent from this literature despite its unique linguistic, historical, and healthcare context. We present the first conceptual framework examining the potential role of NLP within Algeria's mental healthcare system. Drawing on narrative synthesis of global NLP mental health research, Algerian healthcare literature, and low-resource NLP methodologies, we identify four structural barriers to mental healthcare: the language-of-care gap, geographic inequities in access, stigma-related barriers to help-seeking, and the absence of research and digital infrastructure. We then map existing NLP capabilities to each barrier, outlining their potential applications, implementation constraints, and the technical, institutional, and governance requirements necessary for deployment. Based on this analysis, we propose a research and policy roadmap that prioritizes data resources, multilingual language technologies, evaluation frameworks, and regulatory capacity. Although grounded in the Algerian context, the framework addresses challenges common to many multilingual, low-resource, and post-colonial settings. This work provides a foundation for future research on culturally and linguistically appropriate NLP for mental healthcare and offers a practical roadmap for developing responsible AI-enabled mental health systems in underrepresented regions.
Chinese Translation
心理健康障碍是全球残疾的主要原因，但针对心理健康护理的自然语言处理（NLP）研究仍集中在高收入、英语环境中。尽管北非，尤其是阿尔及利亚，具有独特的语言、历史和医疗背景，但在这方面的文献中几乎没有涉及。我们提出了第一个概念框架，探讨NLP在阿尔及利亚心理健康护理系统中的潜在作用。通过对全球NLP心理健康研究、阿尔及利亚医疗文献和低资源NLP方法论的叙事综合，我们识别出心理健康护理的四个结构性障碍：护理语言差距、获取服务的地理不平等、与求助相关的污名障碍，以及缺乏研究和数字基础设施。随后，我们将现有的NLP能力与每个障碍进行映射，概述其潜在应用、实施限制，以及部署所需的技术、制度和治理要求。基于这一分析，我们提出了一条研究和政策路线图，优先考虑数据资源、多语言语言技术、评估框架和监管能力。尽管该框架植根于阿尔及利亚的背景，但它也解决了许多多语言、低资源和后殖民环境中常见的挑战。这项工作为未来在心理健康护理中进行文化和语言适宜的NLP研究奠定了基础，并为在代表性不足地区开发负责任的人工智能驱动的心理健康系统提供了实用的路线图。
cs.CL / 39 / 2608.08636
Enhancing Scientific Named Entity Recognition via Large Language Models: A Type-driven Multi-task Learning Approach
通过大型语言模型增强科学命名实体识别：一种基于类型的多任务学习方法
Abstract
Scientific named entity recognition (SciNER) plays a crucial role in information extraction and knowledge discovery from scientific texts. Recently, large language models (LLMs) have demonstrated the capacity to achieve competitive SciNER performance with minimal human effort. Existing research highlights the importance of incorporating candidate entity type information for accurate entity recognition and classification by LLMs. However, when too many candidate entity types are provided in the prompt, LLMs struggle to accurately recognize and label entities in scientific texts, where entity types are more complex than in general domains. To address this challenge, we propose TdSciNER, a type-driven approach that effectively leverages entity type information to enhance SciNER performance. In TdSciNER, we first design an entity type filter model to identify the most likely entity types present in a given sentence. Subsequently, we introduce an auxiliary multi-class entity typing task within a multi-task learning framework alongside SciNER to obtain richer contextual representations. Then, we develop a novel demonstration selection strategy based on sentence similarity and entity type diversity to activate the in-context learning capabilities of LLMs, thereby improving entity recognition accuracy across diverse scientific domains. Experiments on three datasets demonstrate that our method achieves performance comparable to fully supervised models. Further analysis validates that each entity type-driven component in TdSciNER contributes to the improvement of SciNER performance. This work provides valuable insights for future advancements in SciNER and broader information extraction tasks in scientific text mining.
Chinese Translation
科学命名实体识别（SciNER）在从科学文本中进行信息提取和知识发现中发挥着至关重要的作用。最近，大型语言模型（LLMs）展示了在最小人力投入下实现竞争性SciNER性能的能力。现有研究强调了将候选实体类型信息纳入LLMs以实现准确的实体识别和分类的重要性。然而，当在提示中提供过多候选实体类型时，LLMs在准确识别和标注科学文本中的实体时面临困难，因为科学领域的实体类型比一般领域更为复杂。为了解决这一挑战，我们提出了TdSciNER，一种基于类型的方法，有效利用实体类型信息来增强SciNER性能。在TdSciNER中，我们首先设计了一个实体类型过滤模型，以识别给定句子中最可能出现的实体类型。随后，我们在多任务学习框架中引入了一个辅助的多类别实体类型任务，以获得更丰富的上下文表示。然后，我们开发了一种基于句子相似性和实体类型多样性的新的示例选择策略，以激活LLMs的上下文学习能力，从而提高在不同科学领域的实体识别准确性。在三个数据集上的实验表明，我们的方法的性能可与完全监督模型相媲美。进一步的分析验证了TdSciNER中每个基于类型的组件对SciNER性能提升的贡献。本研究为未来在SciNER及更广泛的科学文本挖掘中的信息提取任务的进展提供了宝贵的见解。
cs.CL / 40 / 2608.08650
The Evolution of Mixture-of-Experts Architectures in Large Language Models: Routing, Topology, Load Balancing, and Expert Parallelism
大语言模型中专家混合架构的演变：路由、拓扑、负载均衡与专家并行性
Abstract
Mixture-of-Experts models increase parameter capacity while keeping the computation activated by each token bounded, but their architectural evolution cannot be explained by a chronological list of model releases alone. This technical survey synthesizes primary papers, official technical reports, and prior surveys to organize modern Mixture-of-Experts systems along five coupled dimensions: expert granularity, expert topology, routing freedom, the scope of load balancing, and execution structure. We describe eight architectural milestones as a dependency graph with six mainline developments and two orthogonal branches, rather than as eight successive generations. We then analyze individual systems through four control planes: Expert Topology, Routing, Balance, and Expert Parallelism. These planes specify which experts exist, which experts process each token, how aggregate load is controlled, and how selected computation is mapped onto physical devices. The framework connects algorithmic choices such as Top-k routing, shared experts, fine-grained experts, and dynamic expert composition with systems concerns including token dispatch, device placement, all-to-all communication, and communication-computation overlap. We conclude with equal-budget pretraining experiments, quality and systems metrics, and open research questions. The main trend is a shift from merely activating more sparse parameters toward decoupling semantic routing, computational budgets, and physical execution.
Chinese Translation
专家混合模型在保持每个令牌激活的计算量有限的同时增加了参数容量，但其架构演变不能仅通过模型发布的时间顺序来解释。本技术调查综合了主要论文、官方技术报告和先前的调查，沿着五个相互关联的维度组织现代专家混合系统：专家粒度、专家拓扑、路由自由度、负载均衡范围和执行结构。我们将八个架构里程碑描述为一个依赖图，其中包括六个主线发展和两个正交分支，而不是简单地视为八个连续的世代。随后，我们通过四个控制平面分析各个系统：专家拓扑、路由、平衡和专家并行性。这些平面指定了存在哪些专家、哪些专家处理每个令牌、如何控制总负载以及如何将选定的计算映射到物理设备上。该框架将算法选择（如 Top-k 路由、共享专家、细粒度专家和动态专家组合）与系统问题（包括令牌调度、设备放置、全到全通信和通信-计算重叠）连接起来。最后，我们通过等预算预训练实验、质量和系统指标以及开放研究问题进行总结。主要趋势是从单纯激活更稀疏的参数转向解耦语义路由、计算预算和物理执行。
cs.CL / 41 / 2608.08721
LibraSpec: Dynamic Diffusion-Based Speculative Decoding via Marginal-Gain-Driven Optimization
LibraSpec：基于边际收益驱动优化的动态扩散式推测解码
Abstract
Speculative decoding accelerates large language model inference by drafting multiple tokens for parallel verification, with efficiency critically determined by the speculative length selected at each decoding round. Existing dynamic speculation methods select the speculation length by estimating how many tokens will be accepted, which is reasonable for autoregressive drafters that generates tokens sequentially. The recent wave of diffusion-based drafters, however, generates candidate blocks in parallel at substantially lower drafting cost, shifting the key question from how many tokens to generate to how many generated tokens are worth verifying. We therefore reformulate dynamic speculative-length selection as expected-speedup optimization and derive a marginal criterion that extends the speculative sequence only when its acceptance gain outweighs the additional verification cost. Building on this criterion, we develop \textit{LibraSpec}, a training-free and plug-and-play algorithm that iteratively determines the speculative length using drafter confidence scores. Theoretically, we prove that LibraSpec monotonically converges toward the optimal speculative length. Experiments across six target models, three diffusion-based speculative decoding methods, and math, coding, and chat benchmarks show consistent improvements under both greedy and sampling settings, achieving a further $0.5\sim1.5\times$ improvement over baselines and up to $8.49\times$ speedup over autoregressive decoding.
Chinese Translation
推测解码通过为并行验证草拟多个标记来加速大型语言模型的推理，其效率在很大程度上取决于每轮解码中选择的推测长度。现有的动态推测方法通过估计将接受多少个标记来选择推测长度，这对于顺序生成标记的自回归草拟器是合理的。然而，最近兴起的基于扩散的草拟器以显著更低的草拟成本并行生成候选块，将关键问题从生成多少个标记转变为生成的标记中有多少值得验证。因此，我们将动态推测长度选择重新表述为期望加速优化，并推导出一个边际标准，该标准仅在接受收益超过额外验证成本时才扩展推测序列。在此标准的基础上，我们开发了 extit{LibraSpec}，这是一种无训练且即插即用的算法，能够通过草拟器的置信度分数迭代确定推测长度。从理论上讲，我们证明了LibraSpec单调收敛于最优推测长度。在六个目标模型、三种基于扩散的推测解码方法以及数学、编码和聊天基准测试中的实验表明，在贪婪和采样设置下均显示出一致的改进，较基线实现了进一步的$0.5 ext{至}1.5 imes$提升，并在自回归解码上实现了高达$8.49 imes$的加速。
cs.CL / 42 / 2608.08744
Can We Optimize the Performance-Carbon Emission Break-Even Point?: The Quest for Greener LLMs
我们能否优化性能-碳排放平衡点？：追求更绿色的LLMs
Abstract
The carbon footprint of any deployed Large Language Model (LLM) accumulates during inference, where repeated use of the model substantially exceeds the one-time cost of fine-tuning. Yet most efficiency interventions target either pre-training scale or post-hoc compression. We ask whether folding a calibrated, differentiable energy surrogate into the fine-tuning objective can produce inference behavior that gains task accuracy at zero or near-zero carbon cost, a break-even configuration. We propose a joint loss mechanism with a per-model carbon-emission parameter, a linear surrogate over parameter norm, FLOP proxy, and a memory proxy, fit from on-hardware energy profiling. We fine-tune three architecturally distinct families: Gemma-2 2B, Llama-3.1 8B, and Qwen-2.5 14B, and evaluate inference F1 and CO$_2$ emissions on three MMLU subjects: abstract algebra, philosophy, and formal logic. We discover from several outcomes that the carbon term behaves as either harmful interference or beneficial regularization depending on the task structure. We position calibrated carbon-aware fine-tuning as a lightweight, drop-in regularizer with a non-empty but model and task-dependent break-even region. This is an ongoing work, and we will release our codebase soon.
Chinese Translation
任何已部署的大型语言模型（LLM）的碳足迹在推理过程中累积，其中模型的重复使用大大超过了一次性微调的成本。然而，大多数效率干预措施要么针对预训练规模，要么针对事后压缩。我们提出一个问题：将一个经过校准的、可微分的能量替代物融入微调目标中，是否可以产生在零或接近零碳成本下获得任务准确性的推理行为，即一种平衡配置。我们提出了一种联合损失机制，包含每个模型的碳排放参数、参数范数的线性替代物、FLOP代理和内存代理，这些都是通过硬件能量分析拟合得到的。我们微调了三种架构上不同的模型系列：Gemma-2 2B、Llama-3.1 8B 和 Qwen-2.5 14B，并在三个MMLU主题上评估推理F1和CO$_2$排放：抽象代数、哲学和形式逻辑。我们从多个结果中发现，碳项根据任务结构的不同表现为有害干扰或有益正则化。我们将校准的碳感知微调定位为一种轻量级的、可直接应用的正则化方法，其具有非空但依赖于模型和任务的平衡区域。这项工作仍在进行中，我们将很快发布我们的代码库。
cs.CL / 43 / 2608.08772
Multilingual Emotion Neurons in Large Audio-Language Models
大型音频语言模型中的多语言情感神经元
Abstract
Emotion is central to human communication, and its expression varies across languages. Large audio-language models (LALMs) achieve strong performance on multilingual speech tasks, yet it remains unclear whether they encode emotion through language-specific correlations or language-agnostic representations. We present the first neuron-level interpretability study of this question. We define Multilingual Emotion Neurons (MLENs) as functional units exhibiting stable emotional selectivity and aligned causal effects across languages, and introduce Consistency-Regularized Fusion (CR-Fusion) to identify them. Across four modern LALMs and 12 typologically diverse languages, emotion-sensitive neurons identified independently per language show minimal overlap, and additional monolingual identification data saturates quickly without isolating more transferable units, motivating identification from pooled cross-lingual evidence. Causal interventions demonstrate that MLENs identified by CR-Fusion provide more precise and transferable affective control than monolingual neuron sets in both zero-shot and low-resource settings. Leave-one-out ablations further reveal asymmetric transfer: individual identification languages, including low-resource ones, contribute non-redundant evidence, while several low-resource languages benefit most from the resulting cross-lingual transfer. Together, our findings provide the first causal, neuron-level account of how LALMs encode emotion across languages, and establish multilingual neuron identification as an effective mechanism for understanding cross-lingual affective behavior.
Chinese Translation
情感是人类沟通的核心，其表达在不同语言中有所不同。大型音频语言模型（LALMs）在多语言语音任务中表现出色，但尚不清楚它们是通过语言特定的相关性还是语言无关的表征来编码情感。我们首次对这一问题进行了神经元级的可解释性研究。我们将多语言情感神经元（MLENs）定义为在不同语言中表现出稳定情感选择性和一致因果效应的功能单元，并引入一致性正则化融合（CR-Fusion）来识别它们。在四个现代LALMs和12种类型多样的语言中，独立于每种语言识别的情感敏感神经元显示出最小的重叠，而额外的单语识别数据很快达到饱和，未能隔离出更多可转移的单元，这促使我们从汇总的跨语言证据中进行识别。因果干预表明，通过CR-Fusion识别的MLENs在零样本和低资源环境中提供比单语神经元集更精确和可转移的情感控制。逐一剔除的消融实验进一步揭示了不对称转移：包括低资源语言在内的个别识别语言提供了非冗余的证据，而几个低资源语言则最受益于由此产生的跨语言转移。综合来看，我们的研究首次提供了LALMs如何跨语言编码情感的因果神经元级解释，并确立了多语言神经元识别作为理解跨语言情感行为的有效机制。
cs.CL / 44 / 2608.08775
OmnilingualGAIA2: Evaluating the Multilingual Gap in Frontier AI Agents
OmnilingualGAIA2：评估前沿人工智能代理中的多语言差距
Abstract
Agentic benchmarks aim to measure how well AI agents plan, search, execute, and recover within realistic multi-tool environments, but they are almost exclusively in English. As AI agents are globally deployed to a linguistically diverse user base, whether agentic competence measured in English transfers to other languages remains an open question. We introduce OmnilingualGAIA2, a machine-translated expansion (with partial human- expert validation) of the GAIA2 agentic benchmark, covering ten target languages spanning five writing systems, paired with a localised and human-calibrated multilingual verifier. Evaluating seven frontier and open-weight agents, we find a universal cross-lingual gap of 8.8-18.4 pass@3 points that is agent-asymmetric in magnitude, concentrates on tool-orchestration rather than quantitative reasoning, and does not close with model scale. A stratified error attribution decomposes the gap as predominantly model-driven (55%), with a bounded translation-contamination floor of only 6.4% of scenario-language pairs. Human-expert linguistic analysis further identifies morphological cue loss and amplified ambiguity as the primary failure mechanisms in non-Latin-script languages. Our results argue that multilingual agentic evaluation must become a standard part of the reporting protocol for globally deployed agents.
Chinese Translation
代理基准旨在测量人工智能代理在现实多工具环境中规划、搜索、执行和恢复的能力，但几乎完全使用英语。当人工智能代理在语言多样的用户基础上全球部署时，使用英语测量的代理能力是否能够转移到其他语言仍然是一个悬而未决的问题。我们介绍了OmnilingualGAIA2，这是GAIA2代理基准的机器翻译扩展（部分经过人类专家验证），涵盖了十种目标语言，跨越五种书写系统，并配备了本地化和人类校准的多语言验证器。通过评估七个前沿和开放权重的代理，我们发现存在一个普遍的跨语言差距，得分在8.8-18.4 pass@3点之间，且在规模上表现出代理不对称，集中在工具协调而非定量推理上，并且随着模型规模的增加并未缩小。分层错误归因将差距主要归因于模型驱动（占55%），而场景语言对的翻译污染底线仅为6.4%。人类专家的语言分析进一步识别出形态线索丧失和模糊性增强是非拉丁字母语言中的主要失败机制。我们的结果表明，多语言代理评估必须成为全球部署代理报告协议的标准部分。
cs.CL / 45 / 2608.08791
Unsure but Certain: Uncovering the Representation-Confidence Gap in Diffusion Language Models
不确定但确定：揭示扩散语言模型中的表征-置信度差距
Abstract
Diffusion language models use broad context to create text, suggesting they might handle input noise better than standard models. Testing reveals this is only partially true. Internally, diffusion models detect text errors highly accurately. Externally, their reported certainty ignores this signal. As accuracy drops due to noise, confidence stays near its maximum and the ability to correctly rank answers degrades toward random chance. We call this mismatch the representation confidence gap. The visible concentration of high certainty scores is a misleading surface symptom. Standard math adjustments remove this concentration but fail to fix the underlying loss of ranking order. This ranking deficit favors standard models under noisy conditions and resists common remedies. Matching training recovers accuracy but not ranking, while score recalibration and input level error signals cannot reorder the final answers. However, the information needed to properly evaluate an answer survives in the hidden states. A lightweight extraction tool uses this signal to improve ranking. This approach is highly efficient because it leaves the base model completely frozen and requires zero additional text generation steps. We present this tool to prove the signal exists, while clearly noting its limits. Ultimately, certainty reliability is a more pressing limit than overall accuracy under noisy conditions.
Chinese Translation
扩散语言模型利用广泛的上下文生成文本，这表明它们可能比标准模型更好地处理输入噪声。测试结果显示，这一说法仅部分成立。在内部，扩散模型能够高度准确地检测文本错误；而在外部，它们报告的置信度却忽视了这一信号。随着噪声的增加，准确性下降，但置信度仍保持在接近最大值，正确排名答案的能力则降至随机水平。我们将这种不匹配称为表征置信度差距。高置信度评分的明显集中是一个误导性的表面症状。标准数学调整消除了这种集中，但未能修复潜在的排名顺序丧失。这种排名缺陷在嘈杂条件下有利于标准模型，并且抵抗常见的补救措施。匹配训练恢复了准确性，但未能恢复排名，而评分重新校准和输入级别错误信号无法重新排序最终答案。然而，正确评估答案所需的信息仍然存在于隐藏状态中。一种轻量级提取工具利用这一信号来改善排名。这种方法非常高效，因为它完全冻结基础模型，并且不需要额外的文本生成步骤。我们提出这个工具以证明信号的存在，同时清楚地指出其局限性。最终，在嘈杂条件下，置信度的可靠性是比整体准确性更紧迫的限制。
cs.CL / 46 / 2608.08793
Evidence-Calibrated Runtime Reconstruction for Agent Skills Across Heterogeneous Coding Agents
针对异构编码代理的证据校准运行时重构代理技能
Abstract
Agent Skills package reusable instructions and assets for tool-using language-model agents. Progressive loading creates failure boundaries poorly represented by session-, model-, or tool-centric traces: a Skill can be discovered but not activated, activated without instructions, or appear successful without an independently verified outcome. We present Skill Runtime Intelligence, a passive runtime-intelligence system that reconstructs supported Skill-lifecycle stages across heterogeneous harnesses while preserving unsupported stages as unknown. Its Run Panorama separates immutable events, deterministic relations, inferred diagnoses, and controlled outcomes with four evidence grades; optional trace import and OTLP/HTTP export support existing observability deployments. Across six frozen repository profiles, three coding agents, and seven clean or fault-injected conditions, all 126 executions preserve source worktrees and each correlates to exactly one source session. Yet adapters expose three distinct semantics: no Skill runs; complete runs but no failure-like events; or failure-like events in every operational-failure and clean session. In a seven-template diagnostic study, semantic aliases and Panorama localize the same six non-clean boundaries but differ in exact/status behavior; both Raw views emit a failure status on all 18 clean cases, while Panorama emits none. A known-rule graph conforms to 126/126 frozen contracts, whereas a second model completes only 228/378 calls. These observations motivate executable adapter qualification and show that event presence is not boundary fidelity, composite exact scores mask distinct errors, and model explanations must not overwrite deterministic facts.
Chinese Translation
代理技能打包了可重用的指令和工具使用语言模型代理的资产。渐进加载创建了失败边界，这些边界在会话、模型或工具中心的痕迹中表现不佳：一种技能可能被发现但无法激活，激活时没有指令，或者在没有独立验证结果的情况下看似成功。我们提出了技能运行时智能（Skill Runtime Intelligence），这是一种被动的运行时智能系统，能够在异构环境中重构支持的技能生命周期阶段，同时将不支持的阶段保留为未知。其运行全景（Run Panorama）将不变事件、确定性关系、推断诊断和受控结果分离为四个证据等级；可选的痕迹导入和OTLP/HTTP导出支持现有的可观察性部署。在六个冻结的代码库配置、三个编码代理和七种干净或故障注入条件下，所有126次执行都保留了源工作树，并且每次执行都与恰好一个源会话相关。然而，适配器暴露出三种不同的语义：没有技能运行；完整运行但没有失败类事件；或者在每个操作失败和干净会话中都有失败类事件。在七个模板的诊断研究中，语义别名和全景定位相同的六个非干净边界，但在精确/状态行为上有所不同；两个原始视图在所有18个干净案例中都发出失败状态，而全景则没有。已知规则图符合126/126个冻结合同，而第二个模型仅完成228/378个调用。这些观察促使可执行适配器的资格认证，并表明事件的存在并不等同于边界的保真度，复合精确分数掩盖了不同的错误，模型解释不得覆盖确定性事实。
cs.CL / 47 / 2608.08800
Instability of LLM Pre-Pretraining: It Doesn't Always Help. An Investigation on Multiple Languages
大规模语言模型预预训练的不稳定性：并不总是有帮助。对多种语言的调查
Abstract
Pretraining LLMs on artificial languages ("pre-pretraining") is a technique that could reportedly increase token efficiency by 33%, i.e., save up to 33% of training tokens needed to reach a certain performance. We validate this prior result for English on a larger set of natural languages across four language families, using two different tokenizers and varying model sizes. We also relate the observed gains (or losses) in token efficiency to quantified linguistic properties of the languages, such as sentence length, morphological richness, and features of dependency syntactic trees (tree depth, number of children, number of crossing dependencies). Our empirical results indicate that the reported gains depend heavily on the experiment setup and the choice of random seed, although we can confirm the trend of stable gains with 128-Dyck pretraining of small models with the Llama tokenizer for most of the examined languages. On a general note, we argue that multiple training runs should be carried out at least for a subset of experiments to avoid the community adopting unstable approaches.
Chinese Translation
在人工语言上预训练大规模语言模型（“预预训练”）是一种技术，据报道可以提高33%的标记效率，即节省最多33%的训练标记以达到某种性能。我们在四个语言家族中使用两种不同的分词器和不同的模型规模，验证了这一先前结果在英语之外的更大自然语言集上的有效性。我们还将观察到的标记效率的增益（或损失）与语言的量化语言学特性相关联，例如句子长度、形态丰富性以及依赖句法树的特征（树深度、子节点数量、交叉依赖数量）。我们的实证结果表明，报告的增益在很大程度上依赖于实验设置和随机种子的选择，尽管我们可以确认在大多数被检查语言中，使用Llama分词器对小模型进行128-Dyck预训练的稳定增益趋势。总体而言，我们认为至少应对一部分实验进行多次训练，以避免社区采用不稳定的方法。
cs.CL / 48 / 2608.08801
IDRAAK: From Multi-Agent NLP to Few-Shot Prompting for Semantic Drift Detection in Technical Requirements
IDRAAK：从多智能体自然语言处理到少量示例提示的技术需求语义漂移检测
Abstract
Translating technical requirements across languages can introduce semantic drift, altering numerical constraints, polarities, modalities, or other specification-critical meaning. IDRAAK is presented as an interpretable framework for detecting such drift using a language-independent Semantic Requirement Representation (SRR), with six detection workflows evaluated, ranging from deterministic comparison to multi-agent verification and few-shot prompting. On 890 synthetic perturbations across 300 requirements from 10 engineering domains, a single LLM call with six few-shot examples achieves MCC=0.888 and F1=0.983, outperforming the evaluated structured and multi-stage alternatives. Further evaluation on PAWS-X (805 pairs, 5 languages) and XNLI (700 pairs, 7 languages) exposes complementary strengths and limitations of structured and LLM-based approaches. Deterministic SRR comparison performs strongly on technical requirements (F1=0.898) but poorly on general-domain text (F1=0.012), while structured evidence improves performance on adversarial paraphrases. Post-hoc Platt scaling further improves confidence calibration. The results demonstrate that increased agentic complexity does not necessarily improve semantic-drift detection and that simple few-shot prompting can provide a strong and efficient alternative.
Chinese Translation
跨语言翻译技术需求可能引入语义漂移，改变数值约束、极性、语气或其他关键规格含义。IDRAAK被提出作为一个可解释的框架，用于检测这种漂移，采用语言无关的语义需求表示（Semantic Requirement Representation, SRR），并评估了六种检测工作流程，从确定性比较到多智能体验证和少量示例提示。在来自10个工程领域的300个需求中，对890个合成扰动进行测试，使用六个少量示例的单次大型语言模型（LLM）调用实现了MCC=0.888和F1=0.983，优于评估的结构化和多阶段替代方案。对PAWS-X（805对，5种语言）和XNLI（700对，7种语言）的进一步评估揭示了结构化方法和基于LLM的方法的互补优势和局限性。确定性SRR比较在技术需求上表现强劲（F1=0.898），但在通用领域文本上表现较差（F1=0.012），而结构化证据则提高了对抗性释义的性能。事后Platt缩放进一步改善了置信度校准。结果表明，增加的智能体复杂性并不一定改善语义漂移检测，而简单的少量示例提示可以提供一种强大且高效的替代方案。
cs.CL / 49 / 2608.08809
Tevatron-Elastic: A Unified Abstraction for Training Elastic Retrievers and Rerankers
Tevatron-Elastic：训练弹性检索器和重排序器的统一抽象
Abstract
A single model scale challenges the flexibility of a production retrieval system: some settings need it faster, others need a smaller index, and the right trade-off changes with the workload. In the context of information retrieval (IR), a transformer-based model can be made smaller in three ways---using fewer layers, passing fewer tokens through the upper layers, or producing a shorter embedding---and each way saves a different compute resource. These options have been studied one at a time, each as its own method with its own code and training setup, which makes them hard to combine or adapt to a new model. We present~\ours to bring all three under one simple abstraction: a single object names any size the model can run at, and a short schedule lists the sizes to train. Training then produces one checkpoint that serves all of those sizes, and at deployment the user picks any of them. The same abstraction covers both retrievers and rerankers and both encoder and decoder models, as it works through interfaces that Hugging Face transformers already expose; a new backbone is a configuration change, not new modeling code. Prior methods---Matryoshka embeddings, early exit, 2D~Matryoshka (e.g., Starbucks), and layerwise token compression---become special cases of our unified abstraction. The same interface also enables Matryoshka~LTC (MLTC), which jointly trains several token-compression ratios in one retriever checkpoint. To validate our framework, we train 20 checkpoints across three backbones and two tasks: the quality curves are smooth, one checkpoint costs little over a model trained for a single size, and a controlled study confirms the wallclock speedups. We release the framework and all checkpoints as a resource for building elastic retrieval systems.
Chinese Translation
单一模型规模挑战了生产检索系统的灵活性：某些设置需要更快的速度，其他设置需要更小的索引，而合适的权衡随着工作负载的变化而变化。在信息检索（IR）的背景下，基于变换器（transformer）模型可以通过三种方式变得更小——使用更少的层、通过上层传递更少的标记，或生成更短的嵌入——而每种方式节省的计算资源各不相同。这些选项曾逐一研究，每种方法都有自己的代码和训练设置，这使得它们难以结合或适应新的模型。我们提出了~ extit{ours}，将这三者统一在一个简单的抽象下：一个单一对象命名模型可以运行的任意大小，并且一个简短的计划列出了要训练的大小。训练后生成一个检查点，服务于所有这些大小，而在部署时用户可以选择其中的任何一个。相同的抽象涵盖了检索器和重排序器，以及编码器和解码器模型，因为它通过Hugging Face变换器已经暴露的接口工作；新的主干网络只是配置更改，而不是新的建模代码。先前的方法——Matryoshka嵌入、早期退出、2D Matryoshka（例如，Starbucks）和逐层标记压缩——成为我们统一抽象的特例。相同的接口还支持Matryoshka LTC（MLTC），它在一个检索器检查点中联合训练多个标记压缩比。为了验证我们的框架，我们在三个主干网络和两个任务上训练了20个检查点：质量曲线平滑，一个检查点的成本略高于为单一大小训练的模型，并且控制研究证实了时钟速度的提升。我们将框架和所有检查点作为构建弹性检索系统的资源发布。
cs.CL / 50 / 2608.08829
Deployable Per-Instance Multi-Layer Activation Steering for Large Language Models
可部署的每实例多层激活引导用于大型语言模型
Abstract
Activation steering edits the behaviour of a frozen language model by adding a learned vector to its residual stream, and current practice fixes the injection layers globally per task. We argue that the best layers are an instance-level decision, and we make per-instance, multi-layer selection both well understood and deployable. On two open-weight 8B models and six binary persona traits, a per-instance oracle over layer subsets shows that the best layers vary from one input to the next: on most trait-model pairs, no fixed global layer set recovers the per-instance benefit. A greedy rule that ranks layers by single-layer marginal effect recovers nearly all of the oracle's benefit, but both must score candidate layers against the gold answer, so neither can run at deployment; the rule instead becomes the target a prompt-only predictor is trained to reproduce. Our deployable recipe needs no label at inference: a per-instance layer ranker read off the prompt embedding, a classifier that infers the steering direction, and an adaptive gate that scores short steered passes against that inferred direction and steers no more layers than necessary. The recipe recovers most of the oracle's lift (the bulk on the stronger model, a clear majority on the harder one), never drives any trait-model pair below its unsteered alignment baseline on average, and largely avoids the fluency collapse that strong global selection incurs at higher layer counts. A mechanistic account, "direction over magnitude", explains the behavioural flip under a mis-directed global set, the output collapse from steering too many layers, and the ceiling of unsteerable inputs.
Chinese Translation
激活引导通过将学习到的向量添加到冻结语言模型的残差流中来编辑其行为，而当前的实践是在每个任务上全局固定注入层。我们认为最佳层是一个实例级的决策，并使每实例的多层选择既易于理解又可部署。在两个开放权重的8B模型和六个二元个性特征上，针对层子集的每实例神谕显示，最佳层因输入而异：在大多数特征-模型对中，没有固定的全局层集能够恢复每实例的优势。一种贪婪规则通过单层边际效应对层进行排名，几乎恢复了神谕的所有收益，但两者都必须根据黄金答案对候选层进行评分，因此在部署时都无法运行；该规则反而成为一个仅通过提示训练的预测器要复制的目标。我们的可部署方案在推理时无需标签：一个每实例层排名器从提示嵌入中读取，一个分类器推断引导方向，以及一个自适应门控根据推断方向对短期引导传递进行评分，并且只引导必要的层。该方案恢复了神谕的大部分提升（在更强的模型上占大部分，在更难的模型上占明显多数），在平均上从未使任何特征-模型对低于其未引导对齐基线，并在较高层数时大大避免了强全局选择所导致的流畅性崩溃。一种机械解释“方向优于大小”解释了在错误导向的全局集下的行为翻转、引导过多层导致的输出崩溃，以及不可引导输入的上限。
cs.CL / 51 / 2608.08847
Explicit Boundary Markers for Subword Vocabularies
子词词汇的显式边界标记
Abstract
Subword tokenizers represent many common words twice in space-using writing systems, once with a leading space and once without. The two entries have separate embeddings in models, so occurrences of one word are divided across rows that are trained independently, and the two forms need not even segment the string the same way: " together" may be a single entry while the same word without a preceding space is tokenized as "to|gether". Capitalization divides a word further, into as many as six forms. We introduce an alternative to standard whitespace conventions using an explicit word boundary marker, which prevents such duplication. Words are delimited by the boundary markers, and spaces between words are represented as pairs of such markers. Two shift codes do the same for title case and upper case, allowing one internal representation of a word to be re-used across different settings. Switching to this convention mitigates the duplicate-entry issue, but does not improve tokenization compression: for both vocabulary-learning algorithms, the best marker scheme stays within one percent of the baseline in characters per token, averaged across six languages. It does result in better language modeling performance. Every marker scheme tested downstream reaches lower bits per byte than the baseline, suggesting that duplication carries a cost that compression does not capture.
Chinese Translation
子词分词器在空间使用的书写系统中将许多常见单词表示为两次，一次带有前导空格，一次不带。两个条目在模型中具有独立的嵌入，因此一个单词的出现被分散到独立训练的行中，并且这两种形式的字符串分割方式不一定相同：“ together”可能是一个单独的条目，而没有前导空格的同一单词则被分词为“to|gether”。大写字母进一步将一个单词分为多达六种形式。我们引入了一种替代标准空白约定的方法，使用显式的词边界标记，以防止这种重复。单词由边界标记分隔，单词之间的空格表示为一对这样的标记。两个移位代码对标题大小写和大写字母做同样的处理，允许在不同设置中重用单词的内部表示。切换到这种约定减轻了重复条目的问题，但并未改善分词压缩：对于两种词汇学习算法，最佳标记方案在每个标记的字符数上保持在基线的一个百分点以内，平均跨越六种语言。它确实导致了更好的语言建模性能。每种下游测试的标记方案在每字节比特数上都低于基线，表明重复带来了压缩无法捕捉的成本。
cs.CL / 52 / 2608.08868
Conversation as Measurement in Clinical Encounters: Observable Phase Structure, Partially Observable Patient State
临床接触中的对话作为测量：可观察的阶段结构与部分可观察的患者状态
Abstract
Many modern AI systems analyze conversational traces to infer aspects of human interaction and state, implicitly assuming that such information is recoverable from conversation. We study observability: whether a target is recoverable from conversational transcripts alone. Observability is difficult to assess because transcripts may provide only a partial view of many targets, and large-scale analysis requires model-based annotation, making true limits of the conversational signal hard to distinguish from annotator error. We therefore study clinical encounters, where patient-reported outcome measures (PROMs) provide an external anchor for patient state, and visits follow broadly structured patterns. We study observability of patient state and conversational phase structure using 439 real-world clinical encounter transcripts spanning 134 hours, including 245 ENT transcripts paired with 273 PROM surveys. We operationalize patient state using PROM scores for voice, cough, and swallowing; phase structure using conversational phase segmentation. To make these analyses credible at scale, we use a PHI-compliant GPT-5 deployment for transcript annotation and conduct 40 hours of manual validation, reducing the risk that apparent limits of observability simply reflect annotator error. Our core finding is an observability asymmetry: phase structure is observable and useful for characterizing clinical encounter organization, while patient state is only partially observable, even in a setting designed to elicit patient symptoms and experiences, cautioning against transcript-only inference of human state.
Chinese Translation
许多现代人工智能系统分析对话痕迹，以推断人类互动和状态的各个方面，隐含地假设这些信息可以从对话中恢复。我们研究可观察性：目标是否仅通过对话记录可以恢复。可观察性难以评估，因为记录可能仅提供许多目标的部分视图，而大规模分析需要基于模型的注释，这使得对话信号的真实限制难以与注释者错误区分。因此，我们研究临床接触，其中患者报告的结果测量（PROMs）为患者状态提供了外部锚点，且就诊遵循广泛结构化的模式。我们使用439个真实世界的临床接触记录（涵盖134小时），包括245个耳鼻喉科记录与273个PROM调查的配对，研究患者状态和对话阶段结构的可观察性。我们通过语音、咳嗽和吞咽的PROM评分来操作化患者状态，通过对话阶段分割来操作化阶段结构。为了使这些分析在规模上具有可信度，我们使用符合PHI标准的GPT-5部署进行记录注释，并进行40小时的人工验证，降低了可观察性表面限制仅反映注释者错误的风险。我们的核心发现是可观察性不对称：阶段结构是可观察的，并且对描述临床接触的组织是有用的，而患者状态仅部分可观察，即使在旨在引导患者症状和体验的环境中，这也提醒我们在仅依赖记录推断人类状态时应谨慎。
cs.CL / 53 / 2608.08869
Position Bias in Ordinal Classification: A Systematic Evaluation
序数分类中的位置偏差：系统评估
Abstract
Large language models are increasingly used for ordinal classification, yet semantically equivalent changes to prompt organization can alter their predictions. We conduct systematic experiments to characterize positional bias from label order, demonstration order, and demonstration placement. First, we apply the three probes to ten frontier LLMs on a common ordinal-classification task; every model is sensitive to all three positional sources, showing that the problem is pervasive. Second, we vary eight prompt-, task-, and model-level factors across five datasets; accuracy and stability are often misaligned, and only lower scale cardinality consistently improves both. Third, we compare pointwise, pairwise, and listwise inference, alternative aggregation and debiasing methods, and joint configurations; the tested corrections do not provide a reliable remedy, while a comparison-based listwise formulation offers the best balance but transfers unevenly across models and bias sources. These findings show that positional robustness depends on the full system configuration rather than the model alone. Ordinal-classification systems should therefore be selected jointly for predictive performance and stability.
Chinese Translation
大型语言模型在序数分类中的使用日益增加，但对提示组织的语义等效变化可能会改变它们的预测。我们进行系统实验，以表征标签顺序、示例顺序和示例位置带来的位置偏差。首先，我们将这三种探测方法应用于十个前沿大型语言模型（LLMs）在一个共同的序数分类任务上；每个模型对这三种位置来源都敏感，表明该问题普遍存在。其次，我们在五个数据集上变化八个提示、任务和模型层面的因素；准确性和稳定性往往不一致，只有较低的规模基数 consistently 提高了两者。第三，我们比较了逐点（pointwise）、成对（pairwise）和列表（listwise）推断、替代聚合和去偏方法，以及联合配置；测试的修正方法未能提供可靠的补救，而基于比较的列表形式提供了最佳平衡，但在模型和偏差来源之间的转移不均匀。这些发现表明，位置的鲁棒性依赖于整个系统配置，而不仅仅是模型本身。因此，序数分类系统应联合选择以优化预测性能和稳定性。
cs.CL / 54 / 2608.08910
Tied Trit-Planes: Constraining PTQTP to a Uniform Nine-Level Quantizer, with a Persistent Folded Format for Disk-Streamed Mixture-of-Experts Serving
绑定三元平面：将 PTQTP 限制为统一的九级量化器，并采用持久折叠格式用于磁盘流式混合专家服务
Abstract
PTQTP decomposes LLM weight matrices into two ternary (trit) planes with two free per-group scales. Tying the scales to a fixed ratio of three collapses the decomposition into a single uniform nine-level quantizer, a known balanced-ternary identity. To our knowledge, at the time of writing, this work is the first to impose that identity as a constraint inside PTQTP's solver. The two trit planes then fold losslessly into one 4-bit code plane that we make the persistent serving representation: disk bytes, expert-cache bytes, and kernel input are the same 4.0625-bits/weight blocks, consumed in one integer dot pass. For this conjunction (ratio-3 nine-level code, CPU-SIMD kernels, SSD expert streaming, identical persistent bytes) we likewise found no precedent. We apply this to the routed experts of DeepSeek-V4-Flash-0731, a 284B-A13B mixture-of-experts model, quantizing in one shot from the released MXFP4 expert weights and streaming experts from SSD on a 64 GB laptop. Against a 4.5-bit Q4_K baseline, measured one process per fixture with an expert-lossless anchor arm as reference control, the tied model matches the official serving API on 5/5 fixtures at step 0 (Q4_K: 4/5) and 12/14 captured continuation steps (11/14), scores 86 vs. 84 on a 100-item MMLU subset, decodes 6.7% faster in decode phase, and ships 9% smaller files: no detected fidelity difference at these small evaluation sizes, and every fixture-level difference between the arms traces to a single measured near-tie cell. The tied fit nevertheless shows higher weight-reconstruction error and worse perplexity, a measured dissociation between proxy metrics and reference fidelity. A cumulative trunk-ternarization ladder and bitwise-pinned aarch64/x86-64 kernels complete the report. All code, formats, and evaluation artifacts are open source in the fucina inference stack.
Chinese Translation
PTQTP 将大规模语言模型（LLM）的权重矩阵分解为两个三元（trit）平面，并为每组提供两个自由缩放因子。将缩放因子绑定为固定的三比率，使得分解简化为一个统一的九级量化器，这是一种已知的平衡三元身份。据我们所知，在撰写本文时，这项工作首次将该身份作为约束施加在 PTQTP 的求解器内部。随后，这两个三元平面无损折叠为一个 4 位代码平面，我们将其作为持久服务表示：磁盘字节、专家缓存字节和内核输入都是相同的 4.0625 位/权重块，在一个整数点乘过程中消耗。对于这种结合（比率为 3 的九级代码、CPU-SIMD 内核、SSD 专家流式传输、相同的持久字节），我们同样没有找到先例。我们将其应用于 DeepSeek-V4-Flash-0731 的路由专家，这是一个 284B-A13B 的混合专家模型，采用一次性量化自发布的 MXFP4 专家权重，并在 64 GB 笔记本电脑上从 SSD 流式传输专家。与 4.5 位 Q4_K 基线相比，在每个装置上测量一个过程，并以专家无损锚臂作为参考控制，绑定模型在第 0 步时在 5/5 装置上与官方服务 API 匹配（Q4_K: 4/5），在 12/14 捕获的延续步骤中匹配（11/14），在 100 项 MMLU 子集上得分为 86 对 84，在解码阶段解码速度快 6.7%，并且生成的文件小 9%：在这些小评估规模下未检测到保真度差异，且每个装置级别之间的差异均可追溯到一个测量的近绑定单元。然而，绑定拟合显示出更高的权重重构误差和更差的困惑度，代理指标与参考保真度之间存在测量的解离。累积的主干三元化阶梯和逐位固定的 aarch64/x86-64 内核完成了报告。所有代码、格式和评估文物均在 fucina 推理栈中开源。
cs.CL / 55 / 2608.08915
Investigating Multimodal Informativity under Different Partner Visibility Conditions in Video-Mediated Dialogue
在不同伙伴可见性条件下研究视频媒介对话中的多模态信息性
Abstract
Situated language use is multimodal and embodied. For example, gestures can carry information that is absent or underspecified in the speech signal, yet dialogue models typically rely on transcripts alone. We study how much referential information gestures and their combination with speech carry in multimodal dialogue under different partner visibility conditions. % We build models that identify the intended referent in a video-mediated referential communication game based on either the speech transcript, the skeletal representation of gesture, or both modalities. Our results show that gesture alone is predictive of the intended referent and that multimodal fusion is most beneficial when the transcript-based model is uncertain. Training-only alignment of learned representations with the referent image further improves the fusion model performance. % In a comparison with human interaction data, we further see pragmatic effects of interlocutor visibility on gesture production and informativeness as well as an entrainment effect in speech and multimodal, but not gesture, performance across rounds of repeated interaction. We thus make contributions to the technical modelling of multimodal information in human dialogue and the analysis of human interaction data via trained model representations.
Chinese Translation
情境语言使用是多模态和具身的。例如，手势可以传递在语音信号中缺失或不明确的信息，但对话模型通常仅依赖于文本记录。我们研究在不同伙伴可见性条件下，手势及其与语音结合所携带的指称信息量。我们构建了模型，以识别在视频媒介指称交流游戏中，基于语音文本、手势的骨架表示或两种模态的意图指称者。我们的结果表明，仅手势就能预测意图指称者，并且在基于文本的模型不确定时，多模态融合最为有利。仅通过与指称图像的对齐训练学习到的表示，进一步提高了融合模型的性能。在与人类互动数据的比较中，我们还观察到对话者可见性对手势产生和信息性以及在多模态和语音表现中的同步效应，但在手势表现中并未出现这种效应。因此，我们对人类对话中多模态信息的技术建模以及通过训练模型表示分析人类互动数据做出了贡献。
cs.CL / 56 / 2608.08942
Same Question, Different Answer? Measuring and Mitigating Prompt Privilege for Equitable AI Access
相同问题，不同答案？衡量和减轻提示特权以实现公平的人工智能访问
Abstract
Large language models (LLMs) are increasingly integrated into healthcare, education, public services, and everyday decision making. They should provide comparable assistance regardless of a user's literacy, communication style, or prompt-engineering expertise. However, existing research on prompt robustness primarily focuses on adversarial attacks, prompt injection, and prompt optimization, while overlooking whether semantically equivalent requests receive different responses simply because they are phrased differently. We refer to this accessibility challenge as "Prompt Privilege": users with greater prompting expertise systematically obtain better model performance despite expressing the same underlying intent. To address this problem, we present a unified framework for measuring and mitigating accessibility disparities in LLM interactions. We introduce Prompt Equity Score (PES), a quantitative metric for evaluating performance consistency across user populations, and Prompt Equity Transformer (PET), an LLM-based agent that automatically transforms user requests into semantically equivalent, accessibility-oriented prompts while preserving their intent. PET shifts prompt optimization from the user to the AI system, functioning as an intelligent accessibility layer between users and foundation models. Experiments on the MedQA benchmark demonstrate measurable prompt privilege, with statistically significant performance disparities between low-literacy and expert-prompting cohorts. Applying PET eliminates these disparities while preserving semantic fidelity, demonstrating that accessibility-oriented prompt normalization can improve equitable AI access. By introducing prompt privilege as a new dimension of AI accessibility and PET as a practical solution, this work advances system-centered accessibility and provides a foundation for more fair, trustworthy, and inclusive AI systems.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于医疗、教育、公共服务和日常决策中。它们应当能够提供相似的帮助，无论用户的读写能力、沟通风格或提示工程专业知识如何。然而，现有关于提示鲁棒性的研究主要集中在对抗性攻击、提示注入和提示优化上，而忽视了语义上等价的请求是否仅因措辞不同而获得不同的响应。我们将这一可及性挑战称为“提示特权”（Prompt Privilege）：具备更高提示专业知识的用户在表达相同基本意图的情况下，系统性地获得更好的模型表现。为了解决这一问题，我们提出了一个统一的框架，用于衡量和减轻LLM交互中的可及性差异。我们引入了提示公平评分（Prompt Equity Score, PES），这是一个用于评估不同用户群体间性能一致性的定量指标，以及提示公平转换器（Prompt Equity Transformer, PET），这是一个基于LLM的代理，能够自动将用户请求转化为语义上等价、以可及性为导向的提示，同时保留其意图。PET将提示优化的责任从用户转移到人工智能系统，充当用户与基础模型之间的智能可及性层。在MedQA基准测试中的实验表明，存在可测量的提示特权，低读写能力和专家提示群体之间存在统计显著的性能差异。应用PET消除了这些差异，同时保持了语义的忠实性，证明了以可及性为导向的提示规范化可以改善公平的人工智能访问。通过将提示特权引入作为人工智能可及性的一个新维度，并将PET作为一个实际解决方案，本研究推进了以系统为中心的可及性，并为更公平、可信和包容的人工智能系统奠定了基础。
cs.CL / 57 / 2608.08975
How Can Rhetoric Reward-Hack AI Reviewers? Dissecting Rhetorical Sensitivity in AI-Based Peer Review
修辞如何影响AI评审的奖励机制？剖析AI基础同行评审中的修辞敏感性
Abstract
As large language models increasingly participate in scientific evaluation, we investigate a potential form of reward hacking: how rhetorical choices shape AI-review judgments when reported scientific content is preserved and how these effects vary across evaluation conditions. We construct a controlled corpus of 4,200 full-paper manuscripts derived from 120 anonymized ICLR 2026 submissions. Two LLM rewriters transform six rhetorical dimensions in opposing directions, and five LLM reviewers evaluate the resulting manuscripts under standard and strict protocols. We also test joint, recursive, and reviewer-guided rewriting. Our results show that rhetorical sensitivity is structured rather than uniform. Evidence framing and novelty stance produce the largest positive-negative contrasts in overall assessment, with scope framing forming a weaker second tier; the remaining dimensions have smaller or less stable effects. This hierarchy persists across human-assessed quality levels, but score movement depends strongly on the AI reviewer's original score: lower scores tend to rise, higher scores tend to fall, and directional contrasts are clearest in the middle ranges. More elaborate workflows do not reliably yield larger gains. Joint rewriting is strongly rewriter-dependent, reviewer guidance does not consistently outperform an unguided second pass, and repeated rewriting yields diminishing, configuration-dependent returns. Across conditions, the rewriter primarily determines the separation between opposing variants, whereas the reviewer determines the magnitude and sign of their score effects. Strict review lowers mean OA by 1.36 points without consistently changing rhetorical sensitivity. These findings identify when rhetorical presentation influences AI scientific review and motivate evaluation systems robust to content-preserving variation in scientific writing.
Chinese Translation
随着大型语言模型越来越多地参与科学评估，我们研究了一种潜在的奖励黑客形式：在报告的科学内容保持不变的情况下，修辞选择如何影响AI评审判断，以及这些影响在不同评估条件下的变化。我们构建了一个受控语料库，包括来自120份匿名ICLR 2026提交的4200篇完整论文手稿。两个大型语言模型（LLM）重写器在相反的方向上转化六个修辞维度，五个LLM评审员在标准和严格的协议下评估生成的手稿。我们还测试了联合、递归和评审员指导的重写。我们的结果表明，修辞敏感性是结构化的，而非统一的。证据框架和新颖性立场在总体评估中产生了最大的正负对比，而范围框架形成了较弱的第二层次；其余维度的影响较小或不稳定。该层级在人工评估的质量水平中持续存在，但分数变化在很大程度上依赖于AI评审员的原始分数：较低的分数往往上升，较高的分数往往下降，方向性对比在中间范围最为明显。更复杂的工作流程并不可靠地带来更大的收益。联合重写在很大程度上依赖于重写者，评审员指导并不总是优于无指导的第二次评估，而重复重写则产生递减的、依赖配置的回报。在各种条件下，重写者主要决定对立变体之间的分离，而评审员则决定其分数效应的大小和符号。严格评审将平均OA降低了1.36分，但并未持续改变修辞敏感性。这些发现确定了修辞呈现何时影响AI科学评审，并激励建立对科学写作中保持内容变异的评估系统的鲁棒性。
cs.CL / 58 / 2608.08989
How Far Do Foundation Models Transfer to Infant Signals? A Cross-Dataset Transfer Audit with a Unified Need Ontology
基础模型在婴儿信号中的迁移能力有多远？基于统一需求本体的跨数据集迁移审计
Abstract
Public infant cry corpora are small, label-incompatible, and almost always evaluated one corpus at a time. We ask what this practice hides and what fixes it. Across four cry corpora screened by a multi-level leakage audit (byte-level and embedding-level deduplication plus a within-corpus train-test near-duplicate audit), we probe four frozen encoders and a handcrafted baseline under a unified five-class need ontology and shared task formulations. The audit exposes what single-corpus evaluation conceals: within-domain macro-F1 swings by 0.57-0.80 for the same encoder, cross-corpus transfer is negative on average (negative-transfer ratio 0.19-0.35, significant in 18 of 30 directed cells, BH-FDR), and 349 content-identical clip groups carry conflicting metadata labels across corpus distributions. The same audit, however, reveals a consistent way forward. Transfer into the noisiest corpus is consistently positive in effect size at matched training size and after near-duplicate removal, offering a practical recipe for small, noisy corpora. Frozen probes saturate at modest label budgets, while stabilized fine-tuning wins with full labels; domain-adaptive pretraining significantly beats stabilized fine-tuning at 5-10-shot (the 1-shot advantage is not robust to optimization-seed variance) but shows no significant advantage at 50-shot or beyond. In the tested binary, shared-label settings, ontology-mapped joint training wins in all four encoder-by-target combinations, whereas naively merging unmapped labels costs up to 37 F1 points. We release the ontology, mapping code, and audit pipeline, turning incompatible cry corpora into a usable joint-training resource.
Chinese Translation
公共婴儿哭声语料库规模小、标签不兼容，并且几乎总是逐个语料库进行评估。我们探讨这种做法隐藏了什么以及如何修正它。在经过多层泄漏审计（字节级和嵌入级去重以及语料库内训练-测试近重复审计）筛选的四个哭声语料库中，我们在统一的五类需求本体和共享任务框架下，探讨了四个冻结编码器和一个手工基线。审计揭示了单一语料库评估所掩盖的事实：同一编码器在领域内的宏观F1值波动范围为0.57-0.80，跨语料库迁移的平均效果为负（负迁移比率为0.19-0.35，在30个定向单元中有18个显著，BH-FDR），而349个内容相同的剪辑组在语料库分布中携带冲突的元数据标签。然而，同样的审计揭示了一条一致的前进道路。在匹配训练规模和去除近重复后，迁移到噪声最大的语料库的效果大小始终为正，为小型噪声语料库提供了实用的解决方案。冻结探针在适度的标签预算下达到饱和，而稳定的微调在使用完整标签时表现更佳；领域自适应预训练在5-10次样本下显著优于稳定微调（1次样本的优势对优化种子变异不稳健），但在50次样本或更多时没有显著优势。在测试的二元共享标签设置中，基于本体映射的联合训练在所有四个编码器与目标组合中均获胜，而天真地合并未映射标签则损失高达37个F1点。我们发布了本体、映射代码和审计管道，将不兼容的哭声语料库转变为可用的联合训练资源。
cs.CL / 59 / 2608.09024
ELICITED: EHR-grounded Longitudinal Interactive Conversations for Information-seeking Triage Evaluation and Decision-making
ELICITED：基于电子健康记录的纵向互动对话用于信息获取分诊评估与决策
Abstract
Emergency-department (ED) triage requires clinicians to rapidly identify patients who need immediate attention, determine who can safely wait, and prioritize limited clinical resources. At presentation, however, information may be limited to a chief complaint and initial vital signs. Clinically important details, including symptom onset and progression, associated symptoms, medical history, and medication use, are often obtained through focused conversation. Effective triage therefore requires clinicians to identify information gaps, ask appropriate follow-up questions, and update their assessment as new evidence becomes available. Most existing ED benchmarks evaluate acuity prediction from a fixed clinical snapshot. Although this formulation measures predictive performance after patient information has been assembled, it does not capture the interactive process through which triage-relevant evidence is elicited and interpreted. Existing medical dialogue datasets support the study of clinical communication, but dialogue statements are not always linked to temporally ordered events in the electronic health record (EHR). We introduce EHR2Dial-Triage, an agentic conversation-generation framework and benchmark grounded in MIMIC-IV-ED. The framework constructs triage conversations under explicit role-based and temporal information boundaries. Each accepted patient disclosure is linked to its supporting EHR event and the first dialogue turn at which it becomes available. EHR2Dial-Triage enables controlled evaluation of information elicitation, evidence use, five-level Emergency Severity Index prediction, and patient-facing communication across models and patient personas. It provides a structured setting for studying conversational triage as a dynamic process of clinical information acquisition, reasoning, and communication.
Chinese Translation
急诊科（ED）分诊要求临床医生迅速识别需要立即关注的患者，确定哪些患者可以安全等待，并优先分配有限的临床资源。然而，在就诊时，信息可能仅限于主诉和初步生命体征。临床重要细节，包括症状的发生和发展、相关症状、病史和用药情况，通常通过针对性的对话获得。因此，有效的分诊要求临床医生识别信息缺口，提出适当的后续问题，并在新证据出现时更新评估。现有的大多数急诊科基准评估基于固定的临床快照进行急性程度预测。尽管这种表述在患者信息汇总后测量预测性能，但并未捕捉到分诊相关证据的获取和解释的互动过程。现有的医学对话数据集支持临床沟通的研究，但对话陈述并不总是与电子健康记录（EHR）中按时间顺序排列的事件相关联。我们引入了EHR2Dial-Triage，这是一个基于MIMIC-IV-ED的代理对话生成框架和基准。该框架在明确的角色基础和时间信息边界下构建分诊对话。每个接受的患者披露都与其支持的EHR事件及其首次可用的对话轮次相链接。EHR2Dial-Triage使得对信息获取、证据使用、五级急救严重性指数预测和面向患者的沟通进行受控评估成为可能，适用于不同模型和患者角色。它为研究对话分诊作为临床信息获取、推理和沟通的动态过程提供了一个结构化的环境。
cs.CL / 60 / 2608.09043
Don't Scroll Back: Missing-Evidence Memory for Streaming Dialogue Summarization
不要向后滚动：流媒体对话摘要中的缺失证据记忆
Abstract
Users of modern platforms repeatedly need summaries of recent dialogue, but the window rarely contains enough context to be interpreted on its own. We formalize this setting as streaming dialogue summarization, where a system must summarize a current window using selective memory from an unbounded history under a fixed budget. We show that the central challenge is not how much history is accessed, but whether memory recovers the evidence that the current window presupposes. We construct a benchmark and evaluation protocol that separately assesses whether memory contains gap-resolving evidence and whether the generated summary reflects it. We propose ReMEMBER, a missing-evidence memory framework that conditions retrieval on unresolved window dependencies and refines retrieved chunks into evidence-dense memory under a fixed budget. Experiments on dialogues with histories up to 160K tokens show that ReMEMBER improves memory recall and gap-resolution completeness over memory construction baselines under the same budget.
Chinese Translation
现代平台的用户反复需要最近对话的摘要，但窗口通常缺乏足够的上下文以便独立解读。我们将这一情境形式化为流媒体对话摘要，其中系统必须在固定预算下，利用来自无限历史的选择性记忆来总结当前窗口。我们表明，核心挑战不在于访问多少历史，而在于记忆是否恢复了当前窗口所假设的证据。我们构建了一个基准和评估协议，分别评估记忆是否包含解决缺口的证据，以及生成的摘要是否反映了这一点。我们提出了ReMEMBER，一个缺失证据记忆框架，它在未解决的窗口依赖关系上进行检索，并在固定预算下将检索到的片段精炼为证据密集的记忆。在对话历史长达160K个标记的实验中，ReMEMBER在相同预算下提高了记忆召回率和缺口解决的完整性，优于记忆构建基线。
cs.CL / 61 / 2608.09044
Tree-of-Experience: Hierarchical Experience Management for Self-Evolving Agents
经验树：自我进化智能体的层次化经验管理
Abstract
Continual self-evolution requires LLM agents to transform environmental interactions into reliable and reusable experience. Existing methods typically refine individual trajectories or abstract shared knowledge from related trajectories, but their experience representations are often disconnected from the underlying reasoning process. This limits feedback attribution, cross-task transfer, and update and retrieval efficiency, particularly in complex reasoning tasks with outcome-level feedback. To overcome this limitation, we propose \textbf{T}ree-\textbf{o}f-\textbf{E}xperience (ToE), a structured experience-management framework that aligns experience organization with the hierarchical reasoning process of LLM agents. Specifically, ToE organizes the experience into a shared tree of analytical perspectives and reasoning paths, whose reliability is calibrated through environmental outcomes to support systematic updating, transfer, and efficient retrieval. The experimental results on \textsc{Game of 24} and \textsc{FinEvolveBench} show that ToE substantially improves both problem-solving performance and efficiency. On \textsc{Game of 24}, ToE achieves a 31.4\% relative improvement in accuracy over the experience-free ToT baseline. On \textsc{FinEvolveBench}, ToE improves tsIC by an average of 41.24\% over the experience-free pipeline across 12 evaluation settings, whereas conventional experience-management methods often underperform experience-free baselines.
Chinese Translation
持续的自我进化要求大规模语言模型（LLM）智能体将环境交互转化为可靠且可重用的经验。现有方法通常对个体轨迹进行精细化处理或从相关轨迹中抽象共享知识，但它们的经验表示往往与潜在的推理过程脱节。这限制了反馈归因、跨任务迁移以及更新和检索效率，尤其是在具有结果级反馈的复杂推理任务中。为了解决这一限制，我们提出了 extbf{T}ree- extbf{o}f- extbf{E}xperience（ToE），这是一个结构化的经验管理框架，旨在将经验组织与LLM智能体的层次化推理过程对齐。具体而言，ToE将经验组织成一个共享的分析视角和推理路径的树状结构，其可靠性通过环境结果进行校准，以支持系统的更新、迁移和高效检索。在 extsc{Game of 24}和 extsc{FinEvolveBench}上的实验结果表明，ToE显著提高了问题解决的性能和效率。在 extsc{Game of 24}中，ToE在准确性上相较于无经验的ToT基线实现了31.4 ext{%}的相对提升。在 extsc{FinEvolveBench}上，ToE在12个评估设置中平均提高了41.24 ext{%}的tsIC，相较于无经验的管道，而传统的经验管理方法通常表现不及无经验基线。
cs.CL / 62 / 2608.09045
Bridging the Gap Between Semantics and Reconstruction:Unifying Sign Language Translation and Production
弥合语义与重建之间的鸿沟：统一手语翻译与生成
Abstract
Recent advances in sign language (SL) research have shown a trend toward unifying multiple sign language understanding (SLU) subtasks, such as isolated sign language recognition (ISLR), continuous sign language recognition (CSLR), and sign language translation (SLT), within a single framework, leading to substantial progress. Meanwhile, sign language production (SLP), which generates sign sequences from text, has also attracted growing attention. This naturally raises an important question: can sign language understanding and production be unified within a single framework? Compared with unifying SLU subtasks, this problem is substantially more challenging. Existing SLU tasks largely share the same direction of mapping, namely from sign inputs to linguistic outputs, whereas SLT and SLP lie in opposite directions of sign-text mapping. A unified framework must therefore address two key challenges: (1) bridging the modality gap between continuous sign motions and discrete text tokens through a shared sign tokenizer that supports both linguistic abstraction and motion reconstruction; and (2) learning a single conditional autoregressive model that can take either sign or text as input and generate the corresponding target sequence in the opposite modality. To this end, we propose Uni-SLTP, a unified framework for SLT and SLP with two key components: (1) a shared sign tokenizer that converts sign sequences into discrete tokens and latent representations, capturing both semantic and reconstructive information; and (2) a unified autoregressive generation model that formulates both tasks as conditional sequence generation. Experiments on widely used public datasets show that Uni-SLTP achieves superior motion accuracy for SLP while maintaining competitive SLT performance.
Chinese Translation
最近在手语（SL）研究中的进展显示出将多个手语理解（SLU）子任务（如孤立手语识别（ISLR）、连续手语识别（CSLR）和手语翻译（SLT））统一在一个框架内的趋势，从而取得了显著进展。同时，手语生成（SLP），即从文本生成手势序列，也引起了越来越多的关注。这自然引发了一个重要问题：手语理解与生成能否在一个统一的框架内实现？与统一SLU子任务相比，这个问题显然更具挑战性。现有的SLU任务在映射方向上大体相同，即从手势输入到语言输出，而SLT和SLP则处于手势-文本映射的相反方向。因此，统一框架必须解决两个关键挑战：（1）通过一个共享的手势标记器，弥合连续手势动作与离散文本符号之间的模态差距，该标记器支持语言抽象和动作重建；（2）学习一个单一的条件自回归模型，该模型可以接受手势或文本作为输入，并生成相应目标序列的相反模态。为此，我们提出了Uni-SLTP，一个用于SLT和SLP的统一框架，包含两个关键组件：（1）一个共享的手势标记器，将手势序列转换为离散符号和潜在表示，捕捉语义和重建信息；（2）一个统一的自回归生成模型，将这两项任务表述为条件序列生成。在广泛使用的公共数据集上的实验表明，Uni-SLTP在SLP方面实现了优越的动作准确性，同时保持了竞争力的SLT性能。
cs.CL / 63 / 2608.09046
Measuring the Tokenization Premium: A Cost Audit for Underserved Language Communities
测量分词溢价：对服务不足语言社区的成本审计
Abstract
Large language models are increasingly deployed as general-purpose educational and technical assistance systems, but their underlying infrastructure does not treat languages equally. One underexamined source of disparity is tokenization: semantically equivalent content can require substantially different token counts across languages, affecting API cost, latency, and usable context length before a model is invoked. We introduce the Tokenization Equity Audit (TEA), a reproducible benchmark for measuring tokenization premiums in technical tutoring content. TEA evaluates three widely used tokenizers, GPT-4o's o200k base, Qwen2.5-7B, and Mistral-7B, on a 120-item Python debugging corpus translated from English into Bengali, Hindi, Arabic, Tamil, and Yoruba. Bengali and Hindi serve as the primary validated cases, while the remaining languages provide exploratory cross-script and cross-family comparisons. Across this corpus, Bengali requires (1.56\times) as many GPT-4o tokens as English, reducing a nominal 128k-token context window to an effective 82k-token English-equivalent capacity for the same semantic content. With the Qwen2.5 and Mistral tokenizers, Bengali requires up to (4.5\times) the English token count. Yoruba, despite using the Latin script, exhibits the highest GPT-4o tokenization premium at (2.37\times), indicating that tokenization inequity cannot be explained by script family alone. These results demonstrate that tokenization can create measurable economic and functional barriers, highlighting the need to treat tokenization as an equity-relevant infrastructure layer for underserved language communities, particularly where educational systems depend on low-cost or offline-capable AI tools.
Chinese Translation
大型语言模型越来越多地被部署为通用教育和技术辅助系统，但其基础设施并未平等对待各种语言。一个未被充分研究的差异来源是分词：在不同语言中，语义上等价的内容可能需要显著不同的分词数量，这会影响API成本、延迟和在调用模型之前可用的上下文长度。我们引入了分词公平审计（Tokenization Equity Audit, TEA），这是一个可重复的基准，用于测量技术辅导内容中的分词溢价。TEA评估了三种广泛使用的分词器：GPT-4o的o200k基础版、Qwen2.5-7B和Mistral-7B，基于一个从英语翻译成孟加拉语、印地语、阿拉伯语、泰米尔语和约鲁巴语的120项Python调试语料库。孟加拉语和印地语作为主要的验证案例，而其余语言则提供了跨脚本和跨语言系的探索性比较。在该语料库中，孟加拉语所需的GPT-4o分词数量是英语的(1.56 imes)，将名义上的128k分词上下文窗口减少到相同语义内容的有效82k分词英语等效容量。使用Qwen2.5和Mistral分词器时，孟加拉语所需的分词数量高达英语的(4.5 imes)。尽管约鲁巴语使用拉丁字母，但其GPT-4o分词溢价最高，达到(2.37 imes)，这表明分词不平等不能仅通过脚本家族来解释。这些结果表明，分词可以造成可测量的经济和功能障碍，强调了将分词视为服务不足语言社区的公平相关基础设施层的必要性，特别是在教育系统依赖于低成本或离线可用的人工智能工具的情况下。
cs.CL / 64 / 2608.09049
Security and Privacy Taxonomy Generation from Mobile App Reviews
基于移动应用评论的安全与隐私分类生成
Abstract
Mobile app reviews are a rich, continuously renewing source of how users experience privacy and security, yet existing taxonomies of these concerns are hand-crafted and cannot keep pace with the evolving nature of the data. Automating taxonomy construction is the natural response, but scalability is the core challenge: current LLM- and clustering-based methods are developed for scientific corpora of a few thousand documents and do not extend to app review collections numbering in the hundreds of thousands. We address this gap in two ways. First, we filter app reviews for privacy- and security-related content, yielding a comprehensive corpus of over 600K reviews. Second, we introduce TaxoScale, a pipeline that handles taxonomy construction at this scale by extending an expert-defined taxonomy via Recursive Hierarchical Clustering and LLM-based node naming. TaxoScale outperforms strong automatic-taxonomy baselines on path, level, coverage, and novelty metrics, and discovers novel branches absent from prior taxonomies.
Chinese Translation
移动应用评论是用户体验隐私和安全的丰富且不断更新的来源，但现有的这些关注点分类是手工制作的，无法跟上数据不断演变的特性。自动化分类构建是自然的应对方式，但可扩展性是核心挑战：当前基于大语言模型（LLM）和聚类的方法是为几千个文档的科学语料库开发的，无法扩展到数十万个应用评论集合。我们通过两种方式解决了这一问题。首先，我们过滤与隐私和安全相关的应用评论，生成了一个超过60万条评论的综合语料库。其次，我们引入了TaxoScale，一个通过递归层次聚类和基于LLM的节点命名来处理这一规模的分类构建的流程。TaxoScale在路径、层级、覆盖率和新颖性指标上超越了强大的自动分类基线，并发现了在先前分类中缺失的新分支。
cs.CL / 65 / 2608.09080
When Confidence Fails: Overconfidence in LLMs under Uncertainty and Missing Clinical Information
当信心失效：在不确定性和缺失临床信息下的语言模型过度自信
Abstract
Large Language Models (LLMs) have achieved strong performance in medical question answering and clinical reasoning tasks. However, their reliability under uncertainty remains poorly understood which raises critical concerns for deployment in high-stakes clinical settings. In such environments, incorrect predictions are inherently risky, but confident incorrect predictions can be particularly harmful as they may mislead clinical decision-making. In this paper, we conduct a systematic behavioral analysis of LLMs under clinical information uncertainty. We propose an evaluation framework based on the MedMCQA dataset consisting of two complementary uncertainty settings. First, we introduce linguistic uncertainty cues through prompt modifications to simulate ambiguous clinical contexts. Second, we construct an answer removal setting, wherein the correct option is deliberately excluded mandating the model to recognize insufficient information and abstain. We analyze both model accuracy and confidence behavior using multiple calibration metrics including calibration gap, Expected Calibration Error (ECE), and Unsafe Confident Error Rate (UCER) across 500 medical questions. Our results reveal a consistent failure mode, i.e., although accuracy degrades under increasing uncertainty, model confidence remains misaligned with accuracy. This leads to a substantial increase in unsafe confident errors, indicating that model confidence remains largely insensitive to clinically meaningful information loss. Furthermore, we observe significant variation across models in their ability to abstain when the correct answer is unavailable, with some models persistently producing high confidence hallucinated answers. These findings expose critical limitations in the epistemic reliability of current LLMs and highlight the need for uncertainty aware evaluation methods prior to their deployment in clinical workflows.
Chinese Translation
大型语言模型（LLMs）在医学问答和临床推理任务中表现出色。然而，它们在不确定性下的可靠性仍然不够清楚，这在高风险临床环境中引发了重要的担忧。在这些环境中，错误的预测本质上是有风险的，但自信的错误预测尤其有害，因为它们可能误导临床决策。本文对LLMs在临床信息不确定性下进行了系统的行为分析。我们提出了一个基于MedMCQA数据集的评估框架，该数据集包含两种互补的不确定性设置。首先，我们通过提示修改引入语言不确定性线索，以模拟模糊的临床背景。其次，我们构建了一个答案移除设置，其中故意排除了正确选项，迫使模型识别信息不足并选择放弃。我们使用多种校准指标分析模型的准确性和信心行为，包括校准差距、期望校准误差（ECE）和不安全自信错误率（UCER），共涉及500个医学问题。我们的结果揭示了一种一致的失效模式，即尽管准确性在不确定性增加的情况下下降，但模型信心与准确性之间仍然不匹配。这导致不安全自信错误的显著增加，表明模型信心对临床上有意义的信息损失仍然高度不敏感。此外，我们观察到不同模型在缺少正确答案时的放弃能力存在显著差异，一些模型持续产生高信心的虚假答案。这些发现揭示了当前LLMs在认知可靠性方面的关键局限性，并强调在其部署到临床工作流程之前需要进行不确定性感知的评估方法。
cs.CL / 66 / 2608.09093
The Announcement Carries the Cue: Markup, Boundaries, and the Notation of Pre-Training Corpora
公告传递线索：标记、边界与预训练语料库的符号化
Abstract
How a document's arrangement is written down, its notation, is a training variable that no dataset card records. The field has established that text-extraction choices change model behaviour, and has never once measured the notation of what those choices put into the corpus. We define clean-window survival, a deterministic count of how much of a stream still demands the boundary inference, and measure notation on three fronts. What corpora carry: a census of thirteen public corpora, where survival falls to 0.153 in a vision-converted PDF slice against 0.889 in C4; the scarce resource is not unmarked text but long unmarked text; a pre-registered supply test finds what remains institutional, not consumer. Our own pre-registered prediction failed: converters do not fabricate structure on prose, and that null forced the reliability mechanism that survives it. What readers use: across five base models spanning 0.6B to 8.2B and two pipelines, deleting a structural announcement makes the following prose measurably harder to predict, while swapping its notation moves nothing. That zero does not make notation unimportant; it relocates the variable: the operative cue is the announcement, not the sigil. What writers impose: a bounded null. Base models do not impose the marked register above the authored baseline, and handed prose with every announcement deleted they do not put one back, at a rate indistinguishable from zero against an authored reference of zero. We ship the format those measurements imply: the pure frame, paragraphs in authored order, every announcement deleted into a reversible sidecar, mixed against the marked copy over announcement presence rather than notation. Choose format operators by the capability they train, not by the fidelity they preserve, and record extractor identity and survival on data cards.
Chinese Translation
文档的排列方式，即其符号化，是一个没有任何数据集卡片记录的训练变量。研究领域已经确定文本提取选择会改变模型行为，但从未测量过这些选择在语料库中所带来的符号化。我们定义了干净窗口生存（clean-window survival），这是一个确定性的计数，表示流中仍然需要边界推断的部分，并从三个方面测量符号化。语料库的内容：对十三个公共语料库的普查，在视觉转换的PDF切片中生存率降至0.153，而在C4中为0.889；稀缺资源并非未标记文本，而是长时间未标记的文本；一项预注册的供应测试发现剩余的内容是机构性的，而非消费者的。我们自己的预注册预测失败：转换器并不在散文中构建结构，而这一空值迫使了生存下来的可靠性机制。读者的使用：在五个基础模型中，参数范围从0.6B到8.2B，以及两个管道，删除结构性公告使得后续散文的预测变得显著困难，而更换其符号化则没有影响。这个零并不意味着符号化不重要；它重新定位了变量：操作线索是公告，而非符号。作者施加的影响：一个有限的空值。基础模型并未在创作基线之上施加标记注册，且在每个公告被删除的情况下，所提供的散文并未恢复标记，且与零的创作参考相比，恢复的比率无可区分。我们提供这些测量所暗示的格式：纯框架，按创作顺序排列的段落，每个公告被删除并转入可逆的附加部分，与公告存在而非符号化的标记副本混合。根据它们所训练的能力选择格式操作符，而非它们所保留的忠实度，并在数据卡上记录提取器身份和生存情况。
cs.CL / 67 / 2608.09096
Evo-Bench: Can Language Models Improve Agent Harness?
Evo-Bench：语言模型能否改善代理的操作框架？
Abstract
Large Language Models (LLMs) have driven rapid progress in autonomous agents, yet standard evaluations remain confined to static task solving. An emerging frontier is harness evolution---the agent's capacity to autonomously optimize its own operating harness. However, systematically benchmarking this capability remains challenging, as existing evaluations fail to isolate harness improvements from base model strength, prevent task-specific overfitting, or capture long-horizon iterative research. To address these challenges, we introduce Evo-Bench, the first benchmark designed to evaluate models' intrinsic harness-evolving capabilities across Search, Office, and General agent domains. To rigorously isolate this capability, Evo-Bench employs a novel harness-guided construction framework: it leverages auxiliary-task evolution to identify tasks genuinely sensitive to framework improvements, followed by sensitivity-aware stratified splitting to ensure robust cross-suite generalization. Extensive evaluations across nine frontier and open-weight models reveal that top models achieve massive absolute gains reaching 16.6 points, closely approaching state-of-the-art human-engineered baselines. Crucially, while autonomous evolution outpeforms artificial harness in General tasks and excels in Search tasks, it struggles in Office tasks that demand highly specific processing workflows. Furthermore, our analysis exposes critical temporal anomalies like early saturation, while demonstrating that the synthesized harnesses act as highly transferable reasoning structures, consistently boosting diverse policy models.
Chinese Translation
大型语言模型（LLMs）推动了自主代理的快速进展，但标准评估仍然局限于静态任务解决。一个新兴的前沿领域是操作框架的演化——代理自主优化其自身操作框架的能力。然而，系统地基准测试这一能力仍然具有挑战性，因为现有评估未能将框架改进与基础模型的强度分离，防止任务特定的过拟合，或捕捉长期的迭代研究。为了解决这些挑战，我们引入了Evo-Bench，这是第一个旨在评估模型内在框架演化能力的基准，涵盖搜索、办公室和通用代理领域。为了严格分离这一能力，Evo-Bench采用了一种新颖的框架引导构建框架：它利用辅助任务的演化来识别真正对框架改进敏感的任务，随后通过敏感性意识的分层拆分确保稳健的跨套件泛化。在九个前沿和开放权重模型的广泛评估中，顶级模型实现了巨大的绝对增益，达到16.6分，接近最先进的人工设计基准。至关重要的是，尽管自主演化在通用任务中优于人工框架，并在搜索任务中表现出色，但在需要高度特定处理工作流程的办公室任务中却面临挑战。此外，我们的分析揭示了早期饱和等关键时间异常，同时证明合成的框架作为高度可转移的推理结构，持续提升多样化的策略模型。
cs.CL / 68 / 2608.09106
LexKairos: Benchmarking Legal Temporal Capabilities in LLMs
LexKairos：法律时序能力在大语言模型中的基准评估
Abstract
Large language models (LLMs) have demonstrated strong performance across a wide range of legal tasks. In legal practice, time is a critical concept that governs the validity of statutes, the progression of legal cases, and the enforcement of procedural deadlines. However, legal temporal capabilities remain underexplored in existing legal AI benchmarks. To address this gap, we propose LexKairos, a comprehensive benchmark for evaluating the temporal capabilities of LLMs in the Chinese legal context across three dimensions: statutory temporal knowledge, case temporal modeling, and statute-case temporal reasoning. LexKairos comprises nine sub-tasks drawn from real-world Chinese judicial cases and statutes. We conduct systematic evaluations of eight LLMs under multiple inference settings, including vanilla, Chain-of-Thought (CoT), and thinking modes. Our results show that Gemini-3-Flash achieves the strongest overall performance, yet even the best-performing model exhibits notable limitations on tasks demanding precise time-sensitive statutory metadata recall or complex reasoning in time limits, indicating that legal temporal knowledge and reasoning remain open challenges for current LLMs. Data and code are available at https://github.com/thunlp/LexKairos.
Chinese Translation
大型语言模型（LLMs）在广泛的法律任务中表现出色。在法律实践中，时间是一个关键概念，它决定了法规的有效性、法律案件的进展以及程序截止日期的执行。然而，现有法律人工智能基准中对法律时序能力的探索仍然不足。为了解决这一问题，我们提出了LexKairos，这是一个全面的基准，用于评估LLMs在中国法律背景下的时序能力，涵盖三个维度：法定时序知识、案件时序建模和法规-案件时序推理。LexKairos包含九个子任务，源自真实的中国司法案件和法规。我们在多种推理设置下对八个LLMs进行了系统评估，包括基础模式、思维链（Chain-of-Thought, CoT）和思考模式。我们的结果表明，Gemini-3-Flash在整体表现上最为出色，但即便是表现最佳的模型在需要精确的时间敏感法定元数据回忆或复杂的时间限制推理任务上也表现出显著的局限性，这表明法律时序知识和推理仍然是当前LLMs面临的开放挑战。数据和代码可在 https://github.com/thunlp/LexKairos 获取。
cs.CL / 69 / 2608.09126
Subjective Multi-Bias Detection with Large Language Models
利用大型语言模型进行主观多偏见检测
Abstract
In this project, we delved into the pervasive challenge of bias detection within the text content. More specifically, our focus lies on the identification of subjective bias, a type of bias that introduces improper attitudes or portrays a statement at odds with the actual truth. The subjective bias can jeopardize the authenticity and reliability of texts, leading to misconceptions and potential social tensions, especially when expressed through offensive language. Following prior work [1], we tackled with three different types of subjective biases in text: (1) framing bias with the use of one-sided words or phrases containing a particular point of view; (2) epistemological bias which includes subtle linguistic features that can affect the believability of the texts; (3) demographic bias with word/phrase usage under presuppositions of a particular demographic factor (i.e., gender or religion). In terms of the data we utilize, the input consists of texts that may harbor subjective biases. The output is a classification or annotation that reveals the presence or absence of such biases within the provided content. More specifically, we detected three different types of multi-span biases in corpus WIKIBIAS [2] with more than 4,000 sentence pairs from Wikipedia edits. The data is labelled by bias type for span pairs with the following categories: (1) framing bias, (2) epistemological bias, (3) demographic bias, and (4) no bias. The project codes are released at https://github.com/HoningJade/LLM-Bias-Type-Classification.
Chinese Translation
在本项目中，我们深入探讨了文本内容中普遍存在的偏见检测挑战。更具体地说，我们的重点是识别主观偏见，这是一种引入不当态度或将陈述与实际真相相悖的偏见类型。主观偏见可能危害文本的真实性和可靠性，导致误解和潜在的社会紧张，尤其是在通过冒犯性语言表达时。基于之前的研究[1]，我们处理了文本中三种不同类型的主观偏见：(1) 框架偏见，即使用包含特定观点的单方面词语或短语；(2) 认识论偏见，包括可能影响文本可信度的微妙语言特征；(3) 人口统计偏见，即在特定人口因素（如性别或宗教）的前提下使用的词语/短语。就我们使用的数据而言，输入由可能存在主观偏见的文本组成。输出是一个分类或注释，揭示所提供内容中是否存在此类偏见。更具体地说，我们在语料库WIKIBIAS [2]中检测到了三种不同类型的多跨度偏见，该语料库包含来自维基百科编辑的超过4000对句子。数据根据偏见类型对跨度对进行标记，类别包括：(1) 框架偏见，(2) 认识论偏见，(3) 人口统计偏见，以及(4) 无偏见。项目代码已发布在 https://github.com/HoningJade/LLM-Bias-Type-Classification。
cs.CL / 70 / 2608.09128
Social Gym and SPaRTan: Benchmarking and Improving LLM Social Reasoning via Multi-Agent Game Tournaments
社交健身房与SPaRTan：通过多智能体游戏锦标赛对大型语言模型社交推理的基准测试与改进
Abstract
LLM agents are increasingly deployed in multi-agent social settings where they must cooperate, negotiate, and adapt to other agents. Measuring and improving these social skills is hard because, unlike math or logic, social interaction offers no objective ground truth: evaluations fall back on LLM judges, which are costly, subjective, and noisy, and models get no reliable signal to learn from. To address both, we first introduce Social Gym, an environment of 21 multi-agent social games (e.g., Werewolves, Resistance, Spyfall) whose rule-decided outcomes make agent performance verifiable and objective, with an Elo tournament that produces a cross-game leaderboard. Benchmarking experiments show that while GPT-5-mini tops the leaderboard, no model excels at all games uniformly or in all game roles, pointing to limitations of social reasoning. Motivated by this, we additionally propose SPaRTan (Self-Play and Reflect-Transfer), a training-free self-improvement loop: a model plays a game, reflects on its trajectories and their outcomes to produce a transferable playbook, and applies that playbook in subsequent games. Our results show that SPaRTan playbooks help GPT-5-mini agents level their performance on weaker roles, but largely do not improve Qwen3-32B's performance. Together, Social Gym and SPaRTan offer a reproducible, verifiable foundation for measuring and improving LLM social reasoning without weight updates.
Chinese Translation
大型语言模型（LLM）代理越来越多地被部署在多智能体社交环境中，在这些环境中，它们必须进行合作、谈判并适应其他代理。测量和提高这些社交技能是困难的，因为与数学或逻辑不同，社交互动没有客观的真实标准：评估依赖于LLM评审，这既昂贵又主观且噪声较大，模型无法获得可靠的学习信号。为了解决这两个问题，我们首先引入了社交健身房（Social Gym），这是一个包含21个多智能体社交游戏（例如，狼人、抵抗组织、间谍游戏）的环境，其规则决定的结果使得代理的表现可验证且客观，并通过Elo锦标赛生成跨游戏排行榜。基准测试实验表明，尽管GPT-5-mini在排行榜上名列前茅，但没有模型在所有游戏或所有游戏角色中表现均衡，指出了社交推理的局限性。基于此，我们进一步提出了SPaRTan（自我游戏与反思转移），这是一个无需训练的自我改进循环：模型进行游戏，反思其轨迹及结果以生成可转移的游戏手册，并在后续游戏中应用该手册。我们的结果表明，SPaRTan游戏手册帮助GPT-5-mini代理在较弱角色上提升表现，但在很大程度上未能改善Qwen3-32B的表现。综合来看，社交健身房与SPaRTan为测量和改进LLM社交推理提供了一个可重复、可验证的基础，而无需进行权重更新。
cs.CL / 71 / 2608.09142
An Agentic Generative Large Language Model for Treatment Planning of Colorectal Cancer
用于结直肠癌治疗规划的代理生成大型语言模型
Lyu, Mengxian, Peng, Cheng, Jang, Tim, Li, Ang, Zhang, Mengyuan, Chen, Ziyi, Elliott, Leighton, Liu, Tianshi, Galindo, Lidice, Sainatham, Chiranjeevi, Borja-Montes, Oscar F., Smith, Kaleb E., Zhang, Ying, Sun, Lichao, Bian, Jiang, Lipori, Gloria, Mitchell, Duane A., Shenkman, Elizabeth A., Guo, Yi, George, Thomas J., Wu, Yonghui
Abstract
Treatment planning in precision oncology requires synthesizing heterogeneous patient information with rapidly evolving clinical guidelines to ensure guideline-concordant care. While large language models (LLMs) show promise in many diagnostic tasks, their adoption for high-stakes treatment planning is hindered by complex reasoning, adherence to timely clinical guidelines, and safety concerns. In this study, we present GatorOnco, an agentic LLM for colorectal cancer (CRC) treatment planning. GatorOnco is developed using a total of 282 billion tokens of biomedical text, including healthcare system-scale clinical text comprising 166 billion tokens from UF Health. We implemented a domain-adaptation method that integrates pre-training, model merging, a two-stage post-training approach, and agent-based reinforcement learning. An agentic retrieval-augmented generation (RAG) approach dynamically integrates time-sensitive clinical guidelines into the reasoning process. In a blind, randomized clinical evaluation conducted by five UF Health oncologists, GatorOnco significantly outperformed open-source LLMs (P < 0.01) and achieved expert-level performance comparable to UF Health oncologists. Compared with expert oncologists, GatorOnco received significantly higher ratings for readability (4.46 vs. 4.19, P < 0.01) and completeness (3.91 vs. 3.52, P < 0.01), while showing statistically comparable performance in correctness (4.09 vs. 4.11, P = 0.921), currency (4.04 vs. 3.98, P = 0.478), and safety (4.22 vs. 4.22, P = 0.999). These findings demonstrate that integrating agentic reasoning with large-scale domain adaptation can help bridge the gap for generative AI in high-stakes cancer treatment planning.
Chinese Translation
精准肿瘤学中的治疗规划需要综合异质的患者信息与快速发展的临床指南，以确保符合指南的护理。尽管大型语言模型（LLMs）在许多诊断任务中显示出潜力，但由于复杂的推理、遵循及时的临床指南以及安全性问题，它们在高风险治疗规划中的应用受到限制。在本研究中，我们提出了GatorOnco，一种用于结直肠癌（CRC）治疗规划的代理型LLM。GatorOnco的开发使用了总计2820亿个生物医学文本标记，包括来自UF Health的1660亿个临床文本标记。我们实施了一种领域适应方法，整合了预训练、模型合并、两阶段后训练方法以及基于代理的强化学习。代理型检索增强生成（RAG）方法动态地将时间敏感的临床指南整合到推理过程中。在由五位UF Health肿瘤学家进行的盲法随机临床评估中，GatorOnco显著优于开源LLMs（P < 0.01），并达到了与UF Health肿瘤学家相当的专家级表现。与专家肿瘤学家相比，GatorOnco在可读性（4.46 vs. 4.19, P < 0.01）和完整性（3.91 vs. 3.52, P < 0.01）方面获得了显著更高的评分，同时在正确性（4.09 vs. 4.11, P = 0.921）、时效性（4.04 vs. 3.98, P = 0.478）和安全性（4.22 vs. 4.22, P = 0.999）方面表现出统计学上的可比性。这些发现表明，将代理推理与大规模领域适应相结合，可以帮助弥补生成性人工智能在高风险癌症治疗规划中的差距。
cs.CL / 72 / 2608.09154
UNSPECIFIC: General Constraint Synthesis for Breaking Copy-and-Paste Shortcut in LLM Instruction Following
UNSPECIFIC：用于打破大型语言模型指令遵循中的复制粘贴快捷方式的一般约束合成
Abstract
Large language models (LLMs) are increasingly expected to follow long lists of constraints in complex instructions, and synthesizing instructions from a reference document (i.e., back-translation) is a widely used method to measure/enhance LLMs' ability to follow complex instructions. However, this method introduces a critical loophole: the constraint synthesis model copies text from the reference as a very specific constraint and the evaluated LLM trivially satisfies the constraint by copying its text in the response. To address these issues, we propose UNSPECIFIC, a novel framework that synthesizes constraints common to two similar reference articles to reduce copy-pasting, selectively hardens only trivially satisfied constraints to balance difficulty and naturalness, and evaluates satisfaction on both the generated article and its summary to penalize superficial instruction following. Consequently, we built the UNSPECIFIC benchmark on news, story, and blog domains to analyze the copy-pasting behavior of LLMs. Our results show that our synthesized constraints are not only more challenging (e.g., the satisfaction rate of GPT-5 Mini drops from 90% to 78%) and natural (LLM win-rate gap improves by 30%) from a human perspective but also mitigate the copy-pasting. We also find that a large portion of constraints are satisfied superficially (i.e., not satisfied in the core narrative of the article). The code and datasets are released at https://github.com/JeetDSharma/UNSPECIFIC.
Chinese Translation
大型语言模型（LLMs）越来越被期望在复杂指令中遵循长列表的约束，而从参考文档合成指令（即反向翻译）是一种广泛使用的方法，用于测量/增强LLMs遵循复杂指令的能力。然而，这种方法引入了一个关键漏洞：约束合成模型将参考文本复制为非常具体的约束，而被评估的LLM通过在响应中复制其文本来轻松满足该约束。为了解决这些问题，我们提出了UNSPECIFIC，一个新颖的框架，通过合成两个相似参考文章中共有的约束来减少复制粘贴，仅选择性地强化那些轻松满足的约束，以平衡难度和自然性，并在生成的文章及其摘要上评估满足度，以惩罚表面上的指令遵循。因此，我们在新闻、故事和博客领域构建了UNSPECIFIC基准，以分析LLMs的复制粘贴行为。我们的结果表明，从人类的角度来看，我们合成的约束不仅更具挑战性（例如，GPT-5 Mini的满足率从90%下降到78%）且更自然（LLM的胜率差距改善了30%），而且还减轻了复制粘贴现象。我们还发现，大部分约束是表面上满足的（即在文章的核心叙述中并未真正满足）。代码和数据集已发布在 https://github.com/JeetDSharma/UNSPECIFIC。
cs.CL / 73 / 2608.09187
Failure-Aware Long-Form Translation: Design and Implementation of a Recoverable LLM Translation System
故障感知的长文本翻译：可恢复LLM翻译系统的设计与实现
Abstract
A long-form translation request can succeed at the API layer and still produce an unusable result. The output may be empty, truncated, filtered, dominated by source or prompt material, or interrupted after producing text worth keeping. This report describes a recovery protocol developed for a deployed translation system with heterogeneous inputs and provider APIs. It delays the first visible release behind a 64-character window, validates the assembled output, and uses typed stream events to distinguish replacement from continuation. Interrupted work is retained only when a paragraph or sentence prefix can be re-derived from the source. Further attempts follow a stable model order and a shared deadline before entering a provenance-marked fallback path. A sanitized companion artifact implements the protocol and passes 38 public tests. Its fixed cases reproduce all 14 configured completion labels, contain four early-invalid prefixes before any of their 235 characters become visible, retain 31 boundary-safe characters across four interrupted streams, and satisfy the attempt, event, and provenance rules in two end-to-end scenarios. These results are executable checks of the published control flow. Translation quality and detector performance on naturally occurring outputs require a different evaluation.
Chinese Translation
长文本翻译请求在API层面上可能成功，但仍然可能产生不可用的结果。输出可能为空、被截断、被过滤、受源文本或提示材料主导，或在生成值得保留的文本后被中断。本报告描述了一种为已部署的翻译系统开发的恢复协议，该系统具有异构输入和提供者API。该协议将首次可见的输出延迟到64个字符的窗口之后，验证组装的输出，并使用类型化流事件来区分替换和继续。只有在可以从源文本重新推导出段落或句子前缀时，才保留中断的工作。后续尝试遵循稳定的模型顺序和共享的截止日期，然后进入标记来源的后备路径。一个经过清理的伴随工件实现了该协议，并通过了38个公共测试。其固定案例重现了所有14个配置的完成标签，在其235个字符变得可见之前包含四个早期无效前缀，在四个中断流中保留了31个边界安全字符，并在两个端到端场景中满足尝试、事件和来源规则。这些结果是对已发布控制流的可执行检查。自然生成输出的翻译质量和检测器性能需要不同的评估。
cs.CL / 74 / 2608.09189
EmoS: A Theory-Grounded Framework for Evaluating and Aligning Emotional Intelligence in Spoken Language Models
EmoS：一个基于理论的框架，用于评估和对齐口语语言模型中的情感智能
Abstract
Despite significant advances in instruction-following and auditory comprehension, the evaluation of Emotional Intelligence (EI) in Spoken Language Models (SLMs) remains confined to rudimentary paralinguistic perception, lacking a systematic, theory-driven cognitive framework. We introduce EmoSBench, the first comprehensive EI evaluation benchmark for SLMs constructed upon the four-branch theoretical model, covering Perceiving, Understanding, Using, and Managing Emotion across ten sub-tasks. Preliminary assessments on EmoSBench reveal a substantial gap: even leading proprietary models like GPT-4o-Audio achieve only 52.6%, significantly trailing human baselines. To bridge this gap, we develop EmoS, a specialized evaluator model optimized via Supervised Fine-Tuning (SFT) and Group Relative Policy Optimization (GRPO). To facilitate its effective training, we curate EmoDialogue, a bilingual dataset providing necessary fine-grained supervision through response pairs with rigorously defined EI gradations. Concurrently, we introduce a reward mechanism integrating a Steep Exponential Accuracy Reward (SEAR) and a Rationale Fidelity Reward (RFR) to enforce precise ordinal scoring and valid reasoning. Experiments demonstrate that EmoS reaches 83.8% accuracy, approaching human-level performance. Furthermore, evaluations on authentic, unconstrained spoken interactions validate its robust real-world generalization, establishing a foundational framework for advancing emotionally intelligent dialogue systems.
Chinese Translation
尽管在遵循指令和听觉理解方面取得了显著进展，但对口语语言模型（SLMs）中情感智能（EI）的评估仍然局限于基础的副语言感知，缺乏系统的、基于理论的认知框架。我们引入了EmoSBench，这是第一个基于四分支理论模型构建的SLMs情感智能评估基准，涵盖了感知、理解、使用和管理情感的十个子任务。对EmoSBench的初步评估揭示了一个显著的差距：即使是领先的专有模型如GPT-4o-Audio也仅达到52.6%，远低于人类基准。为了弥补这一差距，我们开发了EmoS，一个通过监督微调（SFT）和组相对策略优化（GRPO）优化的专用评估模型。为了促进其有效训练，我们策划了EmoDialogue，这是一个双语数据集，通过具有严格定义的情感智能梯度的响应对提供必要的细粒度监督。同时，我们引入了一种奖励机制，结合了陡峭指数准确性奖励（SEAR）和推理保真度奖励（RFR），以强制执行精确的序数评分和有效的推理。实验表明，EmoS达到了83.8%的准确率，接近人类水平的表现。此外，在真实的、无约束的口语互动中的评估验证了其强大的现实世界泛化能力，为推进情感智能对话系统奠定了基础框架。
cs.CL / 75 / 2608.09209
UNMASK: Discovering and Causally Verifying Spurious Shortcuts in Text Classifiers
UNMASK：发现和因果验证文本分类器中的虚假捷径
Abstract
Neural language models trained on large crowdsourced corpora frequently exploit spurious surface patterns tied to target labels without true linguistic or causal relevance, boosting benchmark performance while failing on adversarial or out-of-distribution inputs. Existing approaches either require manual specification of the feature vocabulary or automate discovery only partially, leaving the gap between dataset-level correlation and model-level exploitation unaddressed. We present U N M ASK, a fully automated pipeline that discovers, causally verifies, and mitigates spurious correlations in text classifiers without additional human annotation. Given unlabeled training examples, U N M ASK generates candidate surface patterns as executable boolean expressions, filters them through a statistical validation protocol with independent replication, and establishes causal model dependence via verified counterfactual interventions. Causally confirmed features then serve as annotation-free group definitions for Deep Feature Reweighting, eliminating the group labels that standard DFR requires. Applied to BERT and RoBERTa trained on MNLI, our pipeline independently rediscovers established lexical-overlap and negation biases, verifying 9 of 10 features on BERT and 6 on RoBERTa, and improving HANS accuracy by up to 12.58 pp. On CivilComments-WILDS, programmatic groups match the 70.1% worst- group accuracy of hand-labeled DFR (Kirichenko et al., 2023) without demographic annotation. We further demonstrate that the discovery and validation stages generalize to reward model preference data, surfacing interpretable spurious correlations in RewardBench2.
Chinese Translation
在大型众包语料库上训练的神经语言模型经常利用与目标标签相关但缺乏真实语言或因果关联的虚假表面模式，从而提高基准性能，但在对抗性或分布外输入上表现不佳。现有方法要么需要手动指定特征词汇，要么仅部分自动发现，未能解决数据集级相关性与模型级利用之间的差距。我们提出了UNMASK，一个完全自动化的流程，能够发现、因果验证并减轻文本分类器中的虚假相关性，而无需额外的人类注释。在给定未标记的训练示例的情况下，UNMASK生成可执行布尔表达式作为候选表面模式，通过独立复制的统计验证协议进行筛选，并通过验证的反事实干预建立因果模型依赖。因果确认的特征随后作为无注释的组定义用于深度特征重加权（Deep Feature Reweighting），消除了标准DFR所需的组标签。应用于在MNLI上训练的BERT和RoBERTa，我们的流程独立重新发现了已建立的词汇重叠和否定偏见，在BERT上验证了10个特征中的9个，在RoBERTa上验证了6个，并将HANS的准确率提高了多达12.58个百分点。在CivilComments-WILDS上，程序化组匹配手动标注DFR（Kirichenko等，2023）的70.1%最差组准确率，而无需人口统计注释。我们进一步证明发现和验证阶段可以推广到奖励模型偏好数据，揭示RewardBench2中的可解释虚假相关性。
cs.CL / 76 / 2608.09222
Reading Cognition as Decisions Unfold in Words: A Factorized Inverse Decision Model
将阅读认知视为决策在语言中展开：一种因子化逆决策模型
Abstract
Inverse decision modeling infers latent properties of decision processes from observed behavior, but existing formulations rely primarily on action trajectories. In verbalized cognitive tasks, task execution also produces response dynamics that action-only formulations leave unmodeled, such as verbal production, interaction, and hesitation. We propose a factorized inverse decision model (FIDM) that decomposes each individual's task-execution likelihood into an action factor and an effort factor, governed by separate individual-specific parameters. From raw verbal transcripts, a language model produces structured task-execution traces for factorized inference. On data from 400 older adults performing a grocery-shopping dialog task for cognitive screening, controlled recovery shows selective estimation of the intended factors, while matched semi-synthetic conditions show that FIDM preserves action-execution distinctions even when aggregate behavioral summaries are matched. Action evidence further localizes task-defined deviations across participants. In cognitive-status classification, FIDM provides information complementary to clinical scores, trajectory summaries, and frozen language representations, with consistent gains across all evaluated baselines in the binary setting.
Chinese Translation
逆决策建模从观察到的行为中推断决策过程的潜在属性，但现有的模型主要依赖于行动轨迹。在语言化的认知任务中，任务执行还会产生一些仅依赖于行动的模型未能建模的反应动态，例如语言生成、互动和犹豫。我们提出了一种因子化逆决策模型（Factorized Inverse Decision Model, FIDM），该模型将每个个体的任务执行可能性分解为一个行动因子和一个努力因子，由各自独特的个体参数控制。通过原始语言转录，语言模型生成结构化的任务执行轨迹以进行因子化推断。在400名老年人进行的杂货购物对话任务的认知筛查数据中，受控恢复显示出对意图因子的选择性估计，而匹配的半合成条件表明，即使在匹配的聚合行为摘要下，FIDM仍能保持行动执行的区分。行动证据进一步定位了参与者之间的任务定义偏差。在认知状态分类中，FIDM提供了与临床评分、轨迹摘要和固定语言表示互补的信息，在二元设置中在所有评估的基线中均表现出一致的提升。
cs.CL / 77 / 2608.09276
Verifiably grounded machine interpretation of lunar geology
可验证的月球地质机器解读
Abstract
Planetary geology relies on historical, interpretive reasoning to reconstruct past events from diverse observations. Here, we present a step toward an automated "machine intelligence geologist" by embedding this distinct methodology of geologic knowledge discovery and inference into a multimodal vision-language architecture. Focusing on the stratigraphy of lunar basaltic mare volcanism, we train a model to generate verifiably grounded geologic interpretations directly from co-registered topographic, spectral, and geologic maps. We demonstrate that while the system successfully balances established geological priors with local visual evidence to accurately describe stratigraphy and terrain, numeric age dating derived solely from vision defaults to memorized priors. Integrating an open-book retrieval mechanism resolves this, enabling the model to faithfully cite published chronologies. Our findings delineate the necessary architecture for automated geologic inference: site evidence must be visually interpreted from local data, while quantitative historical context must be retrieved from the scientific record.
Chinese Translation
行星地质依赖于历史性解释推理，通过多样的观察重建过去事件。在此，我们展示了朝向自动化“机器智能地质学家”的一步，通过将这一独特的地质知识发现与推理方法嵌入多模态视觉-语言架构中。我们专注于月球玄武岩海洋火山作用的地层学，训练模型直接从共同注册的地形、光谱和地质图生成可验证的地质解读。我们证明，尽管系统成功地平衡了既定的地质先验与局部视觉证据，以准确描述地层和地形，但仅基于视觉的数值年龄测定则会默认于记忆中的先验。整合开放书籍检索机制解决了这一问题，使模型能够忠实引用已发表的年代学。我们的研究结果勾勒出自动化地质推理所需的架构：现场证据必须从局部数据中进行视觉解读，而定量历史背景必须从科学记录中检索。
cs.CL / 78 / 2608.09280
Is the ACL Responsible NLP Checklist a Box-Ticking Exercise? A Large-Scale Analysis of EMNLP 2025
ACL 负责任的 NLP 检查表是否只是走过场？对 EMNLP 2025 的大规模分析
Abstract
Responsible NLP practice includes a) transparency, b) ethics, and c) societal impacts. The Responsible NLP Checklist aims to push these goals, and promote responsible practice. Recently, ACL released the EMNLP 2025 Checklists to aid transparency on the current research practice, which we focus on. We curate and release the first two datasets of: a) all the checklist responses and justifications from the EMNLP 2025 Main and Finding tracks; b) checklist reference linking to paper sections. We also provide the first analysis of recent EMNLP Checklists, by examining $73,922$ responses and justifications to them. For the Main track, we find that authors isolate ethics questions of the Checklist from the paper's bulk, mimicking the trend of ethics being an afterthought. We then examine \texttt{NO} responses. We find $44.9\%$ of justifications are poor or bad-faith, being brief or empty. Then, we find significant issues with the checklist design and effort of authors, namely that $6\%$ of all checklists contained logical contradictions between parent and child responses. We also find evidence of surface compliance for responsible ethics, with $53\%$ authors dismissing potential risks or social impacts of their work, for which there should be none. We compare this to the Findings track, noticing a similar trend in both tracks. Lastly, we discuss the implications of the checklist design and provide recommendations for future checklist iterations. Including: a) enforcing a minimum word count, b) enforcing more scrutiny on the risks of appliances.
Chinese Translation
负责任的 NLP 实践包括 a) 透明性，b) 伦理，和 c) 社会影响。负责任的 NLP 检查表旨在推动这些目标，并促进负责任的实践。最近，ACL 发布了 EMNLP 2025 检查表，以帮助提高当前研究实践的透明度，我们对此进行了重点关注。我们整理并发布了前两个数据集：a) EMNLP 2025 主会和发现轨道的所有检查表响应及其理由；b) 检查表参考与论文部分的链接。我们还首次分析了最近的 EMNLP 检查表，审查了 $73,922$ 个响应及其理由。对于主会，我们发现作者将检查表的伦理问题与论文的主体隔离，模仿了伦理作为事后思考的趋势。然后，我们检查了 exttt{NO} 响应。我们发现 $44.9\%$ 的理由较差或出于恶意，内容简短或为空。接着，我们发现检查表设计和作者努力存在重大问题，即 $6\\%$ 的所有检查表在父响应和子响应之间存在逻辑矛盾。我们还发现负责任伦理的表面合规证据，$53\\%$ 的作者忽视了其工作的潜在风险或社会影响，而这些本不应存在。我们将其与发现轨道进行比较，注意到两个轨道中存在类似的趋势。最后，我们讨论了检查表设计的影响，并为未来的检查表迭代提供建议，包括：a) 强制最低字数，b) 对应用风险进行更严格的审查。
cs.CL / 79 / 2608.09289
Accurate but Natural? Diagnosing Grammatical and Idiomatic Gaps in Japanese EFL Writing
准确但自然？诊断日本EFL写作中的语法和习惯用语差距
Abstract
Second language writing research distinguishes grammatical accuracy from native-like idiomaticity, yet automated writing evaluation often conflates these dimensions. This study introduces a layered LLM-correction pipeline that isolates structural errors from unnaturalness by generating literal error corrections and idiomatic revisions for 3,830 English writing samples from 120 Japanese junior high school students. Applying the regex-based CEFR-J grammar extractor, we quantify two diagnostic measures: accuracy gaps (structures attempted but incorrectly produced) and idiomatic gaps (grammatically correct structures underused or overused relative to native norms). Results reveal distinct patterns: definite articles, third-person singular -s, and modals (would, could) exhibit significant accuracy difficulties, while -ing forms and hypothetical modals (would) show the largest idiomatic underuse, with simple present verbs, subject-verb-object patterns, and modal can conversely exhibiting the most pronounced overuse. A two-dimensional instructional typology maps error rates against idiomatic gaps, distinguishing accurate but overused grammar items from error-prone or avoided complex forms requiring targeted production practice. This framework advances pedagogical feedback by enabling teachers to diagnose whether learner difficulties arise from inaccurate execution, structural avoidance, or L1-mapped overreliance, supporting evidence-based interventions tailored to the specific needs of each learner.
Chinese Translation
第二语言写作研究将语法准确性与母语般的习惯用语区分开来，但自动写作评估常常将这两个维度混为一谈。本研究引入了一种分层的LLM修正管道，通过为来自120名日本初中学生的3,830份英语写作样本生成字面错误修正和习惯用语修订，来将结构性错误与不自然性隔离。应用基于正则表达式的CEFR-J语法提取器，我们量化了两个诊断指标：准确性差距（尝试但错误产生的结构）和习惯用语差距（相对于母语标准使用不足或过度使用的语法正确结构）。结果揭示出明显的模式：定冠词、第三人称单数-s和情态动词（would, could）表现出显著的准确性困难，而-ing形式和假设情态动词（would）则显示出最大的习惯用语使用不足，简单现在时动词、主谓宾结构和情态动词can则相对表现出最明显的过度使用。一个二维的教学类型图将错误率与习惯用语差距进行映射，区分出准确但过度使用的语法项目与需要针对性生产练习的易错或被避免的复杂形式。该框架通过使教师能够诊断学习者的困难是源于不准确的执行、结构的回避，还是L1映射的过度依赖，从而推动了基于证据的干预，满足每位学习者的具体需求。
cs.CL / 80 / 2608.09356
Universal or Language-Family-Specific Script Unification for Cross-Lingual Transfer? A Case Study on Turkic Languages
跨语言迁移中的通用或语言家族特定的脚本统一？以突厥语言为例
Abstract
Closely related languages written in different scripts expose little surface overlap to multilingual models, limiting cross-lingual transfer. We compare two approaches to script unification: the general-purpose uroman romanizer and the family-specific Common Turkic Script (CTS). We train matched fastText models on transliterated Wikipedia corpora from 11 Turkic languages and evaluate them on WikiANN named entity recognition and Universal Dependencies part-of-speech tagging. CTS and uroman show no significant difference on NER, while both substantially outperform the official monolingual fastText baselines. POS results reveal no universal winner: language-specific differences are associated with the cross-lingual character n-gram coverage induced by each representation, while within-language coverage becomes more important when target-language supervision is available. Although CANINE-c achieves higher overall POS averages, the substantially simpler fastText-based systems remain competitive on several treebanks. Overall, the effectiveness of script unification depends on the language, the induced subword overlap, and the available supervision.
Chinese Translation
书写在不同脚本中的密切相关语言在多语言模型中几乎没有表面重叠，这限制了跨语言迁移。我们比较了两种脚本统一的方法：通用的 uroman 罗马化器和家族特定的共同突厥脚本（Common Turkic Script, CTS）。我们在来自11种突厥语言的转写维基百科语料上训练了匹配的 fastText 模型，并在 WikiANN 命名实体识别和 Universal Dependencies 词性标注上进行了评估。CTS 和 uroman 在命名实体识别（NER）上没有显著差异，而两者在性能上均显著优于官方的单语 fastText 基准。词性标注结果显示没有通用的赢家：语言特定的差异与每种表示所引起的跨语言字符 n-gram 覆盖率相关，而当目标语言的监督可用时，语言内部的覆盖率变得更加重要。尽管 CANINE-c 在整体词性标注平均值上表现更高，但结构更简单的基于 fastText 的系统在多个树库上仍然具有竞争力。总体而言，脚本统一的有效性取决于语言、引入的子词重叠以及可用的监督信息。
cs.CL / 81 / 2608.09393
Temporal Misgrounding in Legal RAG: A Versioned-Corpus Benchmark for French Tax Law
法律RAG中的时间误引用：法国税法的版本语料库基准
Abstract
We identify and quantify temporal misgrounding: the systematic retrieval and citation of the currently in-force version of a legal article when the applicable version is an earlier or future one. Standard legal RAG treats the corpus as static; we argue legal question answering is a temporally-indexed retrieval problem. We introduce FiscalQA Pro, pairing a versioned corpus of 32,436 article-versions of the French tax code (93 years, 1938-2031) with an all-model-hard temporal-reasoning track: 209 scored, expert-reviewed questions across 33 CGI articles (221 released; twelve flagged out of the answerable scope). At selection time, no evaluated model recovered its date-applicable answer closed-book in any of four sampling draws, and the currently in-force text lacks the gold value for all but one of the scored questions. Answers are scored deterministically via atomic ground-truth "nuggets" (regex and numeric-with-tolerance), never LLM-as-judge: an LLM judge would inherit the temporal bias it is meant to score. Across eleven models (five frontier closed-API systems plus Gemini 2.5 Pro as a substitute entry, and five open-weight), parametric knowledge yields 3.0% mean strict accuracy and RAG over a static current-version corpus 2.7%. Static RAG retrieves the date-applicable version 0% of the time, confidently citing a real but inapplicable version. Our end-to-end retriever over a multi-version index, with no oracle, reaches 98.3% mean strict; an oracle-article ablation reaches 99.1%, locating the residual gap in first-stage recall, not version selection. We additionally release a version-aware jurisprudence dataset of 69,208 citation links, together with the corpus, benchmark, model responses, and pipeline code.
Chinese Translation
我们识别并量化了时间误引用：在适用版本为早期或未来版本时，系统性地检索和引用当前有效版本的法律条款。标准的法律RAG将语料库视为静态；我们认为法律问答是一个时间索引的检索问题。我们推出了FiscalQA Pro，将法国税法的32,436个条款版本（93年，1938-2031年）与一个全模型难度的时间推理轨道相结合：209个评分的、专家审查的问题，涵盖33个CGI条款（发布221个；其中12个被标记为超出可回答范围）。在选择时，没有评估的模型在四次抽样中以闭卷形式找回其日期适用的答案，且当前有效文本在所有评分问题中除了一个外均缺乏黄金标准。答案通过原子真实“金块”（正则表达式和带容差的数字）以确定性方式评分，而不是使用大型语言模型（LLM）作为评判者：LLM评判者将继承其旨在评分的时间偏差。在十一种模型中（五种前沿闭API系统加上作为替代入口的Gemini 2.5 Pro，以及五种开放权重模型），参数知识的平均严格准确率为3.0%，而在静态当前版本语料库上的RAG为2.7%。静态RAG在0%的情况下检索到日期适用版本，自信地引用一个真实但不适用的版本。我们的端到端检索器在多版本索引上，无需预言者，达到98.3%的平均严格准确率；去除预言者条款的消融实验达到99.1%，定位到第一阶段召回中的残余差距，而非版本选择。我们还发布了一个版本感知的法理学数据集，包含69,208个引用链接，以及语料库、基准、模型响应和管道代码。
cs.CL / 82 / 2608.09420
Intent Speaks Louder: Controllable User Simulation Beyond Response Imitation
意图更具影响力：超越响应模仿的可控用户模拟
Abstract
User simulators are widely used as scalable environments for training and evaluating interactive assistants. Generating the next user turn is inherently one-to-many: the same profile and dialogue context may support multiple plausible continuations with different local interaction intents. A fluent response may therefore advance the dialogue through an inappropriate intent, such as acceptance rather than repair. Our key insight is that controllable user simulation should separate which local interaction intent the next user turn should realize from how that intent is expressed in language. We introduce UserIDA (User Intent-Directive Alignment), which exposes interaction intent as an explicit per-turn directive. UserIDA defines a six-way intent interface, learns directive-conditioned generation through supervised fine-tuning, and uses intent-calibrated policy optimization during group-based reinforcement learning. The reward preserves composite response quality while ensuring that intent-violating candidates rank below compliant alternatives in mixed groups. On LMSYS-USP, UserIDA achieves 86.6\% intent accuracy, outperforming the strongest dedicated user-simulator baseline by 24.3 percentage points while improving semantic and stylistic similarity. In within-context interventions, it realizes at least four of the six target intents in 91.7\% of evaluated dialogue states, compared with 22.9\% for the strongest external baseline. These results establish per-turn intent control as a complementary dimension to response fidelity in user simulation.
Chinese Translation
用户模拟器被广泛用作训练和评估交互助手的可扩展环境。生成下一个用户发言本质上是多对一的：相同的用户特征和对话上下文可能支持多个合理的延续，具有不同的局部交互意图。因此，一个流畅的回应可能通过不恰当的意图推进对话，例如接受而非修复。我们的关键见解是，可控用户模拟应将下一个用户发言应实现的局部交互意图与该意图在语言中的表达方式分开。我们引入了UserIDA（用户意图-指令对齐），它将交互意图作为每轮的明确指令。UserIDA定义了一个六维意图接口，通过监督微调学习指令条件生成，并在基于小组的强化学习中使用意图校准的策略优化。该奖励机制在确保意图违反的候选者在混合组中排名低于合规替代方案的同时，保持复合响应质量。在LMSYS-USP上，UserIDA实现了86.6%的意图准确率，超越了最强的专用用户模拟器基线24.3个百分点，同时提高了语义和风格的相似性。在上下文干预中，它在91.7%的评估对话状态中实现了至少四个目标意图，而最强的外部基线仅为22.9%。这些结果确立了每轮意图控制作为用户模拟中响应保真度的一个互补维度。
cs.CL / 83 / 2608.09424
Reducing Pretraining-Generation Mismatch in Diffusion Language Models
减少扩散语言模型中的预训练-生成不匹配
Abstract
Autoregressive language models align training and use: generation conditions on a clean prompt, and training predicts future tokens from clean left context. Diffusion language models offer parallel denoising, but native dLLM pretraining can randomly corrupt prompt and continuation tokens together, weakening the clean-prefix interface needed for prompt-conditioned generation. We identify this mismatch for prompt continuation and propose PCD (Prefix-Conditioned Diffusion), a pretraining objective that combines AR prefix supervision with no-shift suffix denoising. At the training-objective level, PCD changes the attention mask, corruption mask, and label construction in continued pretraining; it does not require an autoregressive decoder, verifier, or new inference mode. By supervising the clean-prefix side autoregressively and applying diffusion only to the unknown continuation, PCD makes the local training interface resemble how block-diffusion models are queried at evaluation time. We further separate intra-sample prefix conditioning from inter-sample objective mixing, allowing us to identify the local alignment signal separately from the optional batch-level mixing knob. Across LLaDA2-Mini and Qwen-1.7B backbones, PCD consistently improves over same-family native dLLM stable baselines, reaching a 4.2% relative gain on the main LLaDA2-Mini six-benchmark average (+2.56 points) and a 14.2% relative gain in the primary Qwen mechanism comparison (+4.86 points). These results suggest that aligning the pretraining context distribution with prompt-conditioned generation can recover a measurable part of the dLLM continuation gap without changing inference.
Chinese Translation
自回归语言模型在训练和使用上是一致的：生成依赖于干净的提示，而训练则从干净的左侧上下文预测未来的标记。扩散语言模型提供了并行去噪，但原生的 dLLM 预训练可能会随机损坏提示和延续标记，从而削弱了进行提示条件生成所需的干净前缀接口。我们识别了这种提示延续的不匹配，并提出了 PCD（前缀条件扩散），这是一种将自回归前缀监督与无位移后缀去噪相结合的预训练目标。在训练目标层面，PCD 改变了继续预训练中的注意力掩码、损坏掩码和标签构建；它不需要自回归解码器、验证器或新的推理模式。通过自回归地监督干净前缀侧，并仅将扩散应用于未知的延续，PCD 使局部训练接口类似于评估时对块扩散模型的查询。我们进一步将样本内前缀条件与样本间目标混合分离，使我们能够单独识别局部对齐信号，而不受可选的批次级混合调节的影响。在 LLaDA2-Mini 和 Qwen-1.7B 基础模型上，PCD 在同类原生 dLLM 稳定基线中持续改善，主要 LLaDA2-Mini 六项基准平均值达到 4.2% 的相对增益（+2.56 分），在主要 Qwen 机制比较中达到 14.2% 的相对增益（+4.86 分）。这些结果表明，使预训练上下文分布与提示条件生成对齐可以在不改变推理的情况下恢复 dLLM 延续差距的可测部分。
cs.CL / 84 / 2608.09432
ZetaGPT: A Reference Implementation of Positional--Encoding--Free State--Space--Attention Language Models
ZetaGPT：一种无位置编码状态空间注意力语言模型的参考实现
Abstract
Transformer-based language models rely on self-attention, whose computation is permutation-equivariant and therefore lacks an intrinsic mechanism for representing token order. Existing architectures address this limitation by explicitly incorporating positional information through learned positional embeddings or hand-crafted positional encodings, such as rotary positional encoding (RoPE), treating positional information as an architecturally acquired capability rather than an inherent property of the model. Motivated by the pursuit of positional-encoding-free architectures, this work explores a language model architecture that integrates causal state-space equations to implicitly encode positional information before attention computation. Specifically, each model block applies a causal state-space equation before self-attention, allowing recurrent state dynamics to encode sequential information into token representations. Consequently, subsequent attention layers operate on position-aware representations without requiring explicit positional encodings while retaining the expressive modeling capacity of self-attention. We present \textsc{ZetaGPT}, a compact hybrid language model designed for research, rapid prototyping, algorithm verification, and educational applications. In addition to the proposed architecture, \textsc{ZetaGPT} provides a fully open-source, end-to-end training pipeline encompassing dataset construction, tokenizer training, pretraining, supervised fine-tuning, reinforcement learning from human feedback (RLHF), and chain-of-thought (CoT) reasoning via pure reinforcement learning. To the best of our knowledge, \textsc{ZetaGPT} is the first open-source small language model without explicit positional encoding and establishes a compact, reproducible reference implementation for the development and empirical study of positional-encoding-free language models.
Chinese Translation
基于Transformer的语言模型依赖于自注意力，其计算是置换等变的，因此缺乏表示令牌顺序的内在机制。现有架构通过学习的位置信息嵌入或手工设计的位置编码（如旋转位置编码（RoPE））来明确地纳入位置信息，从而解决这一局限性，将位置信息视为架构获得的能力，而非模型的固有属性。本研究的动机是追求无位置编码的架构，探索一种语言模型架构，该架构在自注意力计算之前集成因果状态空间方程，以隐式编码位置信息。具体而言，每个模型块在自注意力之前应用因果状态空间方程，使得递归状态动态能够将顺序信息编码到令牌表示中。因此，后续的注意力层在不需要显式位置编码的情况下，操作于具有位置感知的表示上，同时保持自注意力的表达建模能力。我们提出了 extsc{ZetaGPT}，这是一种紧凑的混合语言模型，旨在用于研究、快速原型开发、算法验证和教育应用。除了所提出的架构， extsc{ZetaGPT}还提供了一个完全开源的端到端训练管道，包括数据集构建、分词器训练、预训练、监督微调、基于人类反馈的强化学习（RLHF）和通过纯强化学习进行的思维链（CoT）推理。根据我们所知， extsc{ZetaGPT}是第一个没有显式位置编码的开源小型语言模型，并为无位置编码语言模型的开发和实证研究建立了一个紧凑、可重复的参考实现。
cs.CL / 85 / 2608.09507
Learning Preference Adaptation for Large Language Model Personalization via Verbal Reinforcement Learning
通过语言强化学习实现大语言模型个性化的偏好适应学习
Abstract
Natural language user preferences provide an interpretable interface for LLM personalization. However, universal preference summaries often contain information irrelevant to a particular downstream task. Directly supplying the full preference summary therefore wastes context capacity and introduces cross-task distraction, while manually designing task-specific preference views is difficult to scale. In this work, we study \emph{task-specific preference adaptation}: given a universal user preference summary and a downstream task, derive a task-conditioned representation that preserves sufficient decision-relevant evidence while removing redundant context. To this end, we propose \textsc{AlignXada}, a training-free meta-learning framework that induces reusable textual refinement policies for adapting universal preference summaries to task-specific ones. The refinement policy is iteratively optimized by a meta learner through verbal reinforcement learning. Across 13 tasks and three downstream models (39 task--model cells), \textsc{AlignXada} achieves an average gain of 3.82 points, improving 33 cells while retaining only 22.8\% of the original profile tokens and outperforming RAG in 36 cells. An extended faithfulness analysis further shows that the refined profiles remain largely grounded in the source preferences while preserving task-relevant personalization signals, suggesting that profile-side adaptation serves as a practical complement to universal memory construction for lifelong personalized agents.
Chinese Translation
自然语言用户偏好为大语言模型（LLM）个性化提供了可解释的接口。然而，通用偏好摘要往往包含与特定下游任务无关的信息。因此，直接提供完整的偏好摘要会浪费上下文容量并引入跨任务干扰，而手动设计特定任务的偏好视图则难以扩展。在本研究中，我们探讨了 extit{特定任务偏好适应}：给定一个通用用户偏好摘要和一个下游任务，推导出一个任务条件的表示，该表示保留足够的决策相关证据，同时去除冗余上下文。为此，我们提出了 extsc{AlignXada}，一个无训练的元学习框架，能够诱导可重用的文本精炼策略，以将通用偏好摘要适应为特定任务的偏好摘要。精炼策略通过语言强化学习由元学习者迭代优化。在13个任务和三个下游模型（39个任务-模型单元）中， extsc{AlignXada}实现了平均提升3.82分，改善了33个单元，同时仅保留了22.8 ext{%}的原始配置令牌，并在36个单元中超越了RAG。扩展的可信度分析进一步表明，精炼后的配置在很大程度上仍然基于源偏好，同时保留了与任务相关的个性化信号，这表明配置侧的适应作为终身个性化代理的通用记忆构建的实用补充。
cs.CL / 86 / 2608.09510
Build it, Break it, Repeat: Benchmarking and improving LLM-manipulated disinformation detection in social media posts
构建、破坏、重复：基于大语言模型的社交媒体虚假信息检测基准评估与改进
Abstract
Detecting machine-generated disinformation on social media is increasingly difficult as large language models (LLMs) make it easier to generate and rewrite misleading content at scale. Static benchmark evaluations, measuring detector performance on fixed held-out datasets, do not capture how detectors behave when posts are deliberately transformed to evade classification. This paper adapts the Build it, Break it, Fix it framework into Build it, Break it, Repeat (BiBiR): iterative sessions designed to stress-test detectors' robustness under iterative adversarial conditions, evaluating whether models remain reliable when disinformation posts are systematically transformed to evade classification. Across five iterations, the findings show that the best adversarial breakers' transformations came from a combination of back-translation and LLM persona-based rewriting, with the best performing technique achieving a 95% label flip rate (LFR), whilst still preserving the meaning of the original posts. The best builders' model was a triplet contrastive model with a dynamic anchor switching (DASS) architecture, which achieved an average accuracy of 72.68%, outperforming the strong baseline (a fine-tuned e5-small-LoRA) by 15 percentage points on the most robust set of breakers' adversarial attacks. The results demonstrate that an iterative framework best exposes detector weaknesses and pushes robustness improvements; however, it may still require semantic preservation analysis to distinguish valid adversarial evasion from transformations that changed the original disinformation claims' meaning.
Chinese Translation
在社交媒体上检测机器生成的虚假信息变得越来越困难，因为大型语言模型（LLMs）使得大规模生成和重写误导性内容变得更加容易。静态基准评估通过在固定的保留数据集上测量检测器的性能，无法捕捉到当帖子被故意转化以规避分类时检测器的表现。本文将“构建、破坏、修复”框架改编为“构建、破坏、重复”（BiBiR）：一种旨在在迭代对抗条件下进行压力测试的迭代会话，评估模型在虚假信息帖子被系统性转化以规避分类时是否仍然可靠。在五次迭代中，研究结果表明，最佳对抗破坏者的转化来自于反向翻译和基于LLM角色重写的结合，表现最佳的技术达到了95%的标签翻转率（LFR），同时仍然保留了原始帖子的含义。最佳构建者的模型是一个具有动态锚点切换（DASS）架构的三元对比模型，平均准确率为72.68%，在最强的对抗攻击集上比强基线（经过微调的e5-small-LoRA）高出15个百分点。结果表明，迭代框架最能暴露检测器的弱点并推动鲁棒性改进；然而，它可能仍需进行语义保留分析，以区分有效的对抗规避与改变原始虚假信息声明含义的转化。
cs.CL / 87 / 2608.09538
TCS-BENCH: Benchmarking State-of-the-Art Generative AI Theoretical Computer Science Research Ability
TCS-BENCH：评估最先进生成性人工智能在理论计算机科学研究能力的基准
Abstract
We introduce TCS-Bench, a benchmark for evaluating Large Language Models (LLMs) on research-level Theoretical Computer Science (TCS) proof generation. TCS-Bench consists of theorem-proving tasks from papers published at top theoretical computer science venues (STOC, FOCS, and SODA). Each task provides the necessary context to derive a self-contained proof for a target result. We evaluate state-of-the-art models on this benchmark. We verify the correctness of generated proofs via a verification agent, and further benchmark the verifier against human-expert proof judgements on a set of target statements and generated proofs pairs. Our reference verifier achieves over 90% accuracy on the expert labeled set.
Chinese Translation
我们介绍了 TCS-Bench，这是一个用于评估大型语言模型（LLMs）在研究级理论计算机科学（TCS）证明生成方面的基准。TCS-Bench 包含来自顶级理论计算机科学会议（STOC、FOCS 和 SODA）发表的论文中的定理证明任务。每个任务提供了推导目标结果所需的上下文，以生成自包含的证明。我们在该基准上评估最先进的模型。我们通过验证代理验证生成证明的正确性，并进一步将验证者与人类专家对一组目标陈述和生成证明对的判断进行基准测试。我们的参考验证者在专家标注集上的准确率超过 90%。
cs.CL / 88 / 2608.09539
Mawqif-v2: An Arabic Benchmark Dataset for Cross-Target Stance Detection
Mawqif-v2：一个用于跨目标立场检测的阿拉伯语基准数据集
Abstract
Publicly available Arabic datasets for target-specific stance detection remain limited, particularly for evaluating cross-target generalization. This paper presents the Mawqif-v2 Extension, consisting of 996 manually annotated Arabic tweets collected from three public targets: Women Driving, E-Cars, and Trimester System. Each tweet is annotated with stance, sentiment, and sarcasm labels following the original Mawqif annotation scheme. The released extension is intended as a held-out evaluation set for assessing model generalization to both semantically related and previously unseen targets, while the original Mawqif dataset is used for training and development. In addition, we establish baseline results using several Arabic and multilingual transformer models, as well as zero-shot large language models (LLMs), to facilitate reproducible evaluation. Together with the original Mawqif dataset, the Mawqif-v2 Extension provides a benchmark for evaluating cross-target generalization in Arabic stance detection.
Chinese Translation
公开可用的阿拉伯语目标特定立场检测数据集仍然有限，特别是在评估跨目标泛化方面。本文介绍了Mawqif-v2扩展版，包含996条手动标注的阿拉伯语推文，这些推文来自三个公共目标：女性驾驶、电动汽车和学期制度。每条推文都根据原始Mawqif标注方案进行了立场、情感和讽刺标签的标注。发布的扩展版旨在作为一个保留的评估集，以评估模型对语义相关和之前未见目标的泛化能力，而原始Mawqif数据集则用于训练和开发。此外，我们使用几种阿拉伯语和多语言变换器模型，以及零样本大型语言模型（LLMs）建立了基准结果，以促进可重复的评估。结合原始Mawqif数据集，Mawqif-v2扩展版为评估阿拉伯语立场检测中的跨目标泛化提供了基准。
cs.CL / 89 / 2608.09548
ELBench: A Multi-Dimensional Benchmark for Education-Facing Large Language Models
ELBench：面向教育的大型语言模型的多维基准测试
Abstract
Large language models are increasingly deployed in education as tutors, teaching assistants, and content generators. These roles place demands that ordinary question answering does not: a usable education-facing model is supposed to be accurate, safe under sensitive prompts, instructionally useful, and aligned with pedagogical goals at the same time. Existing benchmarks evaluate these requirements largely in isolation, so none assesses education-facing suitability as an integrated profile. We introduce ELBench, the first benchmark to evaluate all four requirements (General Capability, Safety and Trustworthiness, Basic Education, and High-Level Cultivation) on the same models under a common protocol, combining curated public sources with newly synthesized safety and cultivation data. We evaluate nine models, seven frontier general-purpose systems and two education-specialized variants, and report three findings. First, module-level profiles are more informative than a single aggregate: the top six models are statistically indistinguishable on overall score, yet their module leaders differ substantially, and safety is anti-correlated with practical teaching (r = -0.83). Second, the Chinese-developed models lead the safety module, the most discriminative in the suite; this advantage is largest on region-specific normative content and narrows, but does not vanish, on universal-harm content. Third, the two education-specialized models lead neither education module, and on High-Level Cultivation all models share a systematic blind spot: on the structured judgment task they converge on the same non-reference option, favoring pedagogical style over fit to the stated goal, so the module scores uniformly low and does not separate models. This raises, but does not resolve, whether domain post-training keeps pace with frontier systems on education tasks.
Chinese Translation
大型语言模型在教育领域的应用日益增多，担任辅导员、助教和内容生成者等角色。这些角色的要求与普通问答不同：一个可用的面向教育的模型应该在准确性、对敏感提示的安全性、教学实用性以及与教学目标的一致性等方面同时满足要求。现有基准测试在很大程度上是孤立地评估这些要求，因此没有一个基准能够作为综合性指标来评估面向教育的适用性。我们引入了ELBench，这是第一个在相同模型下通过共同协议评估所有四个要求（一般能力、安全性与可信度、基础教育和高层次培养）的基准测试，结合了经过筛选的公共资源和新合成的安全性与培养数据。我们评估了九个模型，包括七个前沿通用系统和两个教育专业化变体，并报告了三个发现。首先，模块级别的特征比单一的总分更具信息量：前六个模型在总体得分上统计上无显著差异，但它们的模块领导者差异显著，安全性与实际教学呈负相关（r = -0.83）。其次，中国开发的模型在安全模块中表现领先，这是该套件中最具区分度的模块；这一优势在特定区域的规范内容上最大，而在普遍危害内容上则有所缩小，但并未消失。第三，两个教育专业化模型在教育模块中均未领先，而在高层次培养方面，所有模型都有一个系统性的盲点：在结构化判断任务中，它们趋向于选择相同的非参考选项，更倾向于教学风格而非符合所述目标，因此模块得分普遍较低，无法区分模型。这引发了一个问题，但并未解决，即领域后训练是否能够跟上前沿系统在教育任务上的进展。
cs.CL / 90 / 2608.09551
Pragmatic Attack Surface: Vulnerabilities of Implicit Context in Large Language Models
务实攻击面：大型语言模型中隐式语境的脆弱性
Abstract
In the era of large language models (LLMs), attackers often manipulate natural language to elicit unsafe or harmful outputs, creating a new natural language attack surface unique to LLM-based systems, where attacks directly exploit explicit linguistic cues in user prompts to bypass the safety mechanism of LLMs. However, such attacks can often be mitigated by existing safety alignment algorithms. On the other hand, human language is inherently grounded in pragmatics, necessitating typical context to interpret language, e.g., world knowledge, social norms. However, such contexts are often implicit because they are not directly expressed in human language and are not sufficiently leveraged in safety alignment, creating a fundamental mismatch between human language interpretation and safety alignment approaches. In this paper, we demonstrate that this mismatch exposes vulnerabilities in LLMs. We refer to this vulnerability as the pragmatic attack surface, which can be exploited to achieve high attack success rates. The experimental results demonstrate that our proposed approach outperforms baseline attack methods across various open-source and closed-source models by a substantial margin.
Chinese Translation
在大型语言模型（LLMs）时代，攻击者常常操纵自然语言以引发不安全或有害的输出，创造出一种独特于基于LLM系统的自然语言攻击面，其中攻击直接利用用户提示中的显性语言线索来绕过LLM的安全机制。然而，这类攻击通常可以通过现有的安全对齐算法进行缓解。另一方面，人类语言本质上是基于语用学的，解读语言需要典型的语境，例如世界知识和社会规范。然而，这些语境往往是隐性的，因为它们并未在自然语言中直接表达，并且在安全对齐中未得到充分利用，从而造成了人类语言解读与安全对齐方法之间的根本不匹配。本文表明，这种不匹配暴露了LLM的脆弱性。我们将这种脆弱性称为务实攻击面，它可以被利用以实现高攻击成功率。实验结果表明，我们提出的方法在各种开源和闭源模型中显著优于基线攻击方法。
cs.CL / 91 / 2608.09568
Se-DPO: Self-Evolving Token Credit for Direct Preference Optimization
Se-DPO：用于直接偏好优化的自演化令牌信用
Abstract
Direct Preference Optimization (DPO) aggregates token-level log-probability ratios via uniform summation, implicitly treating all tokens as contributing equally to the preference signal. However, the contribution of individual tokens to the preference signal varies. We introduce token credit, which modulates each token's KL regularization based on its contribution to the preference outcome. We derive that effective token credit is proportional to the magnitude of each token's implicit reward, and observe that this quantity evolves substantially during training. This implies that static token credit becomes increasingly misaligned as training progresses. In this work, we propose Se-DPO (Self-Evolving Token Credit for DPO), a live mechanism that derives token credit from the model's own evolving internal signals during DPO training. Since the reward signal varies in reliability across positions, Se-DPO calibrates token credit based on both the strength and the confidence of each token's contribution. Se-DPO requires no external models, adding only a lightweight calibration network with minimal computational overhead. Experiments show that Se-DPO improves over DPO by up to 9.8 points on AlpacaEval~2 and 12.2 points on Arena-Hard.
Chinese Translation
直接偏好优化（DPO）通过均匀求和聚合令牌级别的对数概率比，隐含地将所有令牌视为对偏好信号的贡献相等。然而，单个令牌对偏好信号的贡献是不同的。我们引入了令牌信用，根据每个令牌对偏好结果的贡献来调节其KL正则化。我们推导出有效的令牌信用与每个令牌的隐含奖励的大小成正比，并观察到这一量在训练过程中显著演变。这意味着静态令牌信用在训练过程中越来越不一致。在本研究中，我们提出了Se-DPO（用于DPO的自演化令牌信用），这是一种实时机制，在DPO训练过程中从模型自身不断演变的内部信号中推导令牌信用。由于奖励信号在不同位置的可靠性各异，Se-DPO根据每个令牌贡献的强度和置信度来校准令牌信用。Se-DPO不需要外部模型，仅增加一个轻量级的校准网络，计算开销极小。实验表明，Se-DPO在AlpacaEval~2上比DPO提高了最多9.8分，在Arena-Hard上提高了12.2分。
cs.CL / 92 / 2608.09588
MDB-Link: Hierarchical Schema Linking for Multi-Database Text-to-SQL
MDB-Link：用于多数据库文本到SQL的层次化模式链接
Abstract
Traditional Text-to-SQL research and benchmarks assume a known target database, overlooking settings in which a query must be routed within a large, heterogeneous database collection. We therefore study schema linking in a multi-database setting, where the system must first locate the target database and then construct a compact, SQL-relevant schema for generation. We propose MDB-Link, a hierarchical schema-linking framework that retrieves question-relevant columns from a global index, aggregates retrieval evidence to shortlist databases, and uses a budget-aware large language model (LLM) for database reranking, table selection, and column grounding. With Qwen2.5-14B, MDB-Link outperforms LinkAlign on MMQA, Spider2-Snow, and BIRD-dev in database localization and column selection while producing schema subsets close in size to the gold schemas. Exact match improves from 16.88 to 51.41 on MMQA, 2.50 to 9.17 on Spider2-Snow, and 12.52 to 38.01 on BIRD-dev. MDB-Link also runs faster than LinkAlign and AutoLink, demonstrating the effectiveness of hierarchical schema reduction for downstream SQL generation.
Chinese Translation
传统的文本到SQL研究和基准测试假设目标数据库是已知的，忽视了在大型异构数据库集合中必须路由查询的情境。因此，我们研究了多数据库环境中的模式链接，其中系统必须首先定位目标数据库，然后构建一个紧凑的、与SQL相关的模式以进行生成。我们提出了MDB-Link，一个层次化的模式链接框架，该框架从全局索引中检索与问题相关的列，聚合检索证据以缩小数据库范围，并使用一种预算感知的大型语言模型（LLM）进行数据库重新排序、表选择和列定位。在Qwen2.5-14B上，MDB-Link在数据库定位和列选择方面优于LinkAlign，在MMQA、Spider2-Snow和BIRD-dev上产生的模式子集大小接近黄金模式。准确匹配在MMQA上从16.88提高到51.41，在Spider2-Snow上从2.50提高到9.17，在BIRD-dev上从12.52提高到38.01。MDB-Link的运行速度也快于LinkAlign和AutoLink，证明了层次化模式简化在下游SQL生成中的有效性。
cs.CL / 93 / 2608.09624
Measuring the Wrong Thing: Internal Harmfulness Scores Anti-Rank Successful Jailbreaks
测量错误的事物：内部有害性评分反而降低成功越狱的排名
Abstract
Internal safety scores judge a prompt before any text is generated, and they are validated by how well they separate harmful prompts from benign ones. That separation is then read as evidence that the score will also catch the attacks that succeed. Harmful intent is a property of the prompt. Jailbreak success is an outcome produced later by a particular target model, decoding policy, and judge. A filter tuned on a score that measures the wrong quantity spends its false positive budget on attacks that would have failed anyway. In this paper we audit that inference. Attention based measurements are usually read from prompt dependent locations, so a wrapper changes both the content being judged and the place the signal is taken from. We therefore introduce Active Attention Probing, which supplies a fixed content independent measurement coordinate. We pair every base goal with a plain and a wrapped version and generate real completions from the target models. On Llama, wrapping raises harmful generation from 0.05 to 0.27 while harmful intent AUROC falls from 0.936 to 0.803, so the attacks grow more dangerous while the prompts look safer to the score. Among wrapped harmful prompts the outcome AUROC is 0.220, which places the attacks that succeeded below the attacks that failed. Rare token, passive, and detector derived channels reproduce the reversal on the same matched design, and the reversal itself persists across three target models, seven attack families, and two independent judges. Distribution shift then degrades calibration and threshold transfer before it degrades ranking.
Chinese Translation
内部安全评分在生成任何文本之前对提示进行评估，并通过其在多大程度上区分有害提示与良性提示来验证其有效性。这种区分被解读为证据，表明该评分也能捕捉到成功的攻击。有害意图是提示的一个特性，而越狱成功是由特定目标模型、解码策略和评估者后续产生的结果。一个基于错误量度的评分调优的过滤器，会将其假阳性预算花费在那些本来就会失败的攻击上。本文对这一推理进行了审计。基于注意力的测量通常来自依赖于提示的位置，因此一个包装器会同时改变被评估的内容和信号获取的位置。因此，我们引入了主动注意力探测（Active Attention Probing），它提供了一个固定的、与内容无关的测量坐标。我们将每个基本目标与一个普通版本和一个包装版本配对，并从目标模型生成真实的完成结果。在 Llama 上，包装将有害生成从 0.05 提高到 0.27，而有害意图的 AUROC 从 0.936 降至 0.803，因此攻击变得更加危险，而提示在评分上看起来更安全。在包装的有害提示中，结果 AUROC 为 0.220，这使得成功的攻击低于失败的攻击。稀有标记、被动和检测器导出的通道在相同匹配设计上重现了这种反转，而这种反转在三个目标模型、七个攻击类别和两个独立评估者之间持续存在。分布偏移随后降低了校准和阈值转移，最终影响了排名。
cs.CL / 94 / 2608.09717
How Do Large Language Models Judge Social Attraction? Evidence from Theory-Grounded Persona Ratings Across Multiple LLMs and Humans
大型语言模型如何判断社会吸引力？基于理论的个性评分在多个大型语言模型和人类中的证据
Abstract
Large language models (LLMs) are increasingly used to perform subjective evaluations traditionally made by humans, yet their validity as social judges remains unclear. This paper examines whether LLMs can assess social attraction from theory-grounded persona profiles constructed from ten psychological and relational constructs and organized into three tiers: socially attractive, socially mixed, and socially unattractive. We examine LLM ratings in two studies and compare them with human judgments in a third study. In Study 1, 34 LLMs rated 12 profiles across three repeated runs. Although some models tended to give higher or lower ratings overall, they showed strong stability across runs, consistent three-tier ordering, and high agreement in relative profile ordering. Study 2 examined sensitivity to gender presentation using six matched name-and-pronoun profile pairs and a separate pronoun-only test with a gender-neutral name, finding no significant effects in either analysis. In Study 3, 198 human participants evaluated the six matched profiles from Study 2. Their ratings reproduced the three-tier structure and followed a profile ordering consistent with that of the LLMs. However, LLMs rated attractive profiles more positively and unattractive profiles more negatively than humans, while neither group showed a significant overall effect of gender presentation.
Chinese Translation
大型语言模型（LLMs）越来越多地被用于执行传统上由人类进行的主观评估，但它们作为社会评判者的有效性仍不清楚。本文研究了LLMs是否能够根据从十个心理和关系构建的理论基础个性档案评估社会吸引力，这些档案被组织为三个层次：社会吸引、社会混合和社会不吸引。我们在两项研究中考察了LLM的评分，并在第三项研究中将其与人类的判断进行比较。在研究1中，34个LLM对12个档案进行了评分，分为三次重复运行。尽管一些模型总体上倾向于给出更高或更低的评分，但它们在运行之间表现出强大的稳定性、一致的三层次排序以及相对档案排序的高度一致性。研究2使用六对匹配的姓名和代词档案对以及一个单独的仅代词测试（使用性别中立的姓名）考察了对性别表现的敏感性，发现两项分析中均未发现显著效果。在研究3中，198名人类参与者评估了研究2中的六个匹配档案。他们的评分再现了三层次结构，并遵循与LLMs一致的档案排序。然而，LLMs对吸引人的档案给予了更积极的评分，对不吸引人的档案给予了更消极的评分，而两组在性别表现的整体效果上均未显示显著性。
cs.CL / 95 / 2608.09765
REFRAMED: Towards Realistic Audio Description Generation for Movies
REFRAMED：面向电影的现实音频描述生成
Abstract
Audio Description (AD) is a verbal narration of key visual content in videos, enabling access for visually impaired audiences. Unlike standard video captioning, AD is a structured editorial task: descriptions must be inserted into gaps in dialogue and must convey only what is needed to understand the narrative being told. However, existing approaches formulate AD generation in an artificial setting where both the content and timing of descriptions are pre-specified, reducing the task to clip-level captioning. They further rely on noisy transcription and alignment pipelines, and lack the rich parallel data required for modeling narrative context. We introduce a new formulation of AD generation in which models must jointly decide what to describe and when to do it. To support this, we present REFRAMED, a high-quality dataset of 2,023 videos that span 3,302 scenes from 206 movies, with professional AD transcripts (both American and British versions), professional subtitles and aligned screenplays. We also provide a manually curated challenge set that pairs full movies with multiple AD references, together with evaluation protocols that leverage dialogue gaps and multi-reference comparisons. Experiments with state-of-the-art AD systems and multimodal LLMs show that they outperform trivial baselines but fall far short of expert human performance. Our dataset and benchmark establish a new foundation for research on video understanding.
Chinese Translation
音频描述（Audio Description, AD）是对视频中关键视觉内容的口头叙述，使视障观众能够获取信息。与标准视频字幕不同，音频描述是一项结构化的编辑任务：描述必须插入对话的空白处，并且只能传达理解叙述所需的信息。然而，现有的方法在一个人为设定的环境中制定音频描述生成，其中描述的内容和时机都是预先指定的，这使得任务简化为片段级字幕生成。它们还依赖于嘈杂的转录和对齐流程，并缺乏建模叙事上下文所需的丰富平行数据。我们提出了一种新的音频描述生成形式，其中模型必须共同决定描述的内容和时机。为支持这一点，我们呈现了REFRAMED，这是一个高质量的数据集，包含2,023个视频，涵盖206部电影中的3,302个场景，配有专业的音频描述文本（包括美国和英国版本）、专业字幕和对齐的剧本。我们还提供了一个手动策划的挑战集，将完整电影与多个音频描述参考配对，并提供利用对话空白和多参考比较的评估协议。与最先进的音频描述系统和多模态大语言模型的实验表明，它们的表现优于简单基线，但远未达到专家人类的表现。我们的数据集和基准为视频理解研究奠定了新的基础。
cs.CL / 96 / 2608.09766
Cultivar: A Contrastive and Locale-Oriented Translation Benchmark for Investigating Contamination and Localisation Robustness
Cultivar：一种对比性和地方导向的翻译基准，用于研究污染和本地化的鲁棒性
Chen, Pinzhen, Chowdhury, Koel Dutta, Xu, Xiaoya, Tan, David, Osmelak, Doreen, de Gibert, Ona, Tumurchuluun, Ariun-Erdene, Urlana, Ashok, Sizov, Fedor, Sirin, Hale, Alabi, Jesujoba, Abed, Karrar Talib, Klimaszewski, Mateusz, Bogoychev, Nikolay, Bafna, Niyati, Schmidtova, Patricia, Shanbhag, Preksha Manjunath, Shen, Sherrie, Zouhar, Vilem, Iyer, Vivek, Hamidullah, Yasser, Ghussin, Yusser Al, Zhao, Zheng
Abstract
Multilingual translation benchmarks are typically sourced in English and translated into other languages, treating language pairs as the unit of evaluation---a design that is prone to contamination over time and overlooks locale and cultural considerations. We therefore advocate for source-contrastive evaluation and instantiate it with Cultivar, a localised subset of FLORES, which enables locale-specific translation evaluation. When paired with unlocalised counterparts, performance discrepancy allows the probing of data contamination and localisation robustness. We benchmark 32 open-weight models and find that MT-specialised models are less robust, a few models potentially overfit FLORES, and models tend to translate US content better than that of other locales, regardless of language.
Chinese Translation
多语言翻译基准通常以英语为源语言，并翻译成其他语言，将语言对视为评估单位——这种设计随着时间的推移容易受到污染，并忽视了地方和文化的考虑。因此，我们倡导源对比评估，并通过Cultivar这一FLORES的本地化子集来实现，它能够支持特定地方的翻译评估。当与未本地化的对应模型配对时，性能差异可以探测数据污染和本地化的鲁棒性。我们对32个开放权重模型进行了基准测试，发现机器翻译（MT）专用模型的鲁棒性较差，少数模型可能过拟合FLORES，且模型倾向于更好地翻译美国内容，而不论语言如何，其他地方的内容翻译效果较差。
cs.CL / 97 / 2608.09767
Structured Phonological Representations for Audio-Articulatory rtMRI Speech Classification
用于音频-发音 rtMRI 语音分类的结构化音韵表示
Abstract
Real-time MRI makes it possible to observe vocal-tract articulation during speech, but mapping these articulatory patterns to phonetic and phonological categories remains challenging. We investigate whether PhonoQ, an audio-based model trained to recognize structured phonological features, provides useful information for audio--articulatory modeling. Specifically, we extract representations from PhonoQ's Conformer module, whose training is shaped by supervision for manner, place, voicing, and vowel features. Using articulatory contours with synchronized audio-derived features, we compare WavLM-large and HuBERT-large baselines with models that incorporate PhonoQ-derived representations. Across unseen-speech and unseen-subject settings, these features improve macro-F1 for phonological targets including manner, place, voicing, vowel height, and vowel backness, and also improve fine-grained 39-phoneme classification. In a contour-only inference setting, audio-derived teacher supervision yields modest but consistent gains over contour-only training, indicating that phonological information from synchronized audio can be partially transferred to articulatory models. Finally, posterior analyses show interpretable surface-sensitive patterns consistent with flapping-like /t/ realizations, /t/-/r/ retraction or affrication, and nasal place assimilation.
Chinese Translation
实时 MRI 使得观察语音中的声道发音成为可能，但将这些发音模式映射到语音学和音韵学类别仍然具有挑战性。我们研究了 PhonoQ 这一音频基础模型是否能提供有用的信息用于音频-发音建模，PhonoQ 经过训练以识别结构化音韵特征。具体而言，我们从 PhonoQ 的 Conformer 模块中提取表示，该模块的训练受到发音方式、发音位置、发声和元音特征的监督。通过使用与音频同步的发音轮廓特征，我们比较了 WavLM-large 和 HuBERT-large 基线模型与包含 PhonoQ 派生表示的模型。在未见语音和未见受试者的设置中，这些特征提高了音韵目标的宏观 F1，包括发音方式、发音位置、发声、元音高度和元音后退，并且还提高了细粒度的 39 音素分类。在仅使用轮廓的推断设置中，音频派生的教师监督在轮廓仅训练上带来了适度但一致的提升，表明来自同步音频的音韵信息可以部分转移到发音模型中。最后，后续分析显示出与类似 /t/ 的发音实现、/t/-/r/ 后退或破擦化以及鼻音位置同化一致的可解释表面敏感模式。
cs.CL / 98 / 2608.09772
PragMatch: Separating Pragmatic Incongruity from Cross-Modal Mismatch in Large Vision-Language Models
PragMatch：在大型视觉-语言模型中区分语用不一致与跨模态不匹配
Abstract
Large Vision-Language Models (LVLMs) have demonstrated strong performance on multimodal benchmarks, yet it remains unclear whether they genuinely reason about relationships between images and text or rely on superficial correlations, known as shortcut learning. This question is particularly important for multimodal sarcasm detection, where successful prediction depends on recognizing pragmatic incongruity rather than treating sarcasm as simple image-text mismatch. We introduce PragMatch, a controlled benchmark of 3,000 image-text pairs derived from MMSD2.0, including original sarcastic examples and constructed literal and hard-negative pairs. We identify influential shortcut cues through systematic masking and evaluate their impact through targeted injection experiments. Our results show that LVLM predictions are sensitive to lexical, OCR-derived and stylistic cues, with injected surface signals causing substantial changes in model predictions despite unchanged underlying image-text relationships. Our findings reveal limitations in current LVLMs while PragMatch provides a systematic testbed for evaluating multimodal pragmatic reasoning beyond surface-level image-text alignment.
Chinese Translation
大型视觉-语言模型（LVLMs）在多模态基准测试中表现出色，但尚不清楚它们是否真正推理图像与文本之间的关系，还是依赖于表面的相关性，这被称为捷径学习。这个问题对于多模态讽刺检测尤为重要，因为成功的预测依赖于识别语用不一致，而不是将讽刺视为简单的图像-文本不匹配。我们引入了PragMatch，这是一个由3,000对图像-文本对构成的受控基准，来源于MMSD2.0，包括原始的讽刺示例以及构造的字面和困难负样本对。我们通过系统的掩蔽识别影响显著的捷径线索，并通过有针对性的注入实验评估它们的影响。我们的结果表明，LVLM的预测对词汇、OCR衍生和风格线索敏感，注入的表面信号尽管底层图像-文本关系不变，却导致模型预测发生显著变化。我们的研究揭示了当前LVLMs的局限性，而PragMatch则提供了一个系统的测试平台，用于评估超越表层图像-文本对齐的多模态语用推理。
cs.CL / 99 / 2608.09779
KGCaRe: Explainable Complex Conditional Question Answering using Automatic Knowledge Graph Construction and Context Retrieval with LLMs
KGCaRe：基于自动知识图谱构建和上下文检索的可解释复杂条件问答
Abstract
Answering complex conditional questions using Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) remains a challenge, particularly in domain-specific contexts where general-purpose LLMs and RAG tend to underperform. We hypothesize that augmenting RAG with unstructured and structured knowledge, extracted from both documents and knowledge graphs (KGs), can improve reasoning and answer accuracy for such tasks. To test this, we propose KGCaRe, a hybrid approach that combines neural retrieval with symbolic reasoning over LLM-generated KGs. KGCaRe constructs a KG from documents using a multi-prompt extraction strategy and stores it in a graph database. Simultaneously, the documents are embedded into a vector store to enable neural retrieval. KGCaRe performs innovative iterative graph traversal guided by the LLM to extract relevant triples, prune irrelevant information, and uses additional clue entities to traverse the graph again if the initial traversal does not provide satisfactory context to generate the answer. The relevant triples extracted from the KG in path form, along with semantically retrieved text passages, are then fed into custom KGCaRe prompts to generate answers to the complex conditional questions with explanations. We evaluate KGCaRe on two complex conditional QA datasets. Our results on these datasets show that KGCaRe consistently outperforms existing baselines, including Vanilla LLM, Code Prompt, Text Prompt, Think-on-Graph, Vanilla RAG, and HybridContextQA, across multiple LLMs such as Mistral, Mixtral, GPT-3.5, and GPT-4o. We publicly release the software pipeline that we developed to implement the proposed KGCaRe approach.
Chinese Translation
使用大型语言模型（LLMs）和增强检索生成（RAG）回答复杂条件问题仍然是一项挑战，特别是在领域特定的上下文中，通用的LLMs和RAG往往表现不佳。我们假设，通过从文档和知识图谱（KGs）中提取的非结构化和结构化知识增强RAG，可以提高此类任务的推理能力和答案准确性。为此，我们提出了KGCaRe，这是一种将神经检索与基于LLM生成的KGs的符号推理相结合的混合方法。KGCaRe使用多提示提取策略从文档中构建KG，并将其存储在图数据库中。同时，文档被嵌入到向量存储中，以便进行神经检索。KGCaRe执行创新的迭代图遍历，由LLM指导，以提取相关三元组，修剪无关信息，并在初始遍历未能提供满意上下文以生成答案时，使用额外的线索实体再次遍历图。提取的相关三元组以路径形式呈现，并与语义检索的文本段落一起输入到定制的KGCaRe提示中，以生成复杂条件问题的答案及其解释。我们在两个复杂条件问答数据集上评估KGCaRe。我们在这些数据集上的结果表明，KGCaRe在多个LLM（如Mistral、Mixtral、GPT-3.5和GPT-4o）上始终优于现有基线，包括Vanilla LLM、Code Prompt、Text Prompt、Think-on-Graph、Vanilla RAG和HybridContextQA。我们公开发布了为实现所提出的KGCaRe方法而开发的软件管道。
cs.CL / 100 / 2608.09792
Comparing British and American Audio Description of Movies
比较英国和美国电影的音频描述
Abstract
Narrating the visual component of movies is known as audio description. It is a narrative technique designed to enable blind and visually impaired individuals to follow the story. However, it is far more constrained than most narratives: the descriptions not only need to convey the story in the movie, but they must also fit into gaps between dialogue and they need to conform to guidelines that exist in each region. In this work, we compare audio description created in the United Kingdom against audio description created in the United States. We use guidelines written for these two regions, alongside the impressions from a practitioner in the field, to motivate specific hypotheses about the differences. We test these hypotheses against our pre-existing corpus, which provides both human-authored American and British audio description for each of 206 movies. Results provide quantitative evidence to uphold all tested hypotheses, including differences in lexicon, the use of the progressive aspect, the use of passive constructions, the use of subjective adjectives and modifiers, when characters are named, how scenes are cued, and degree of overlap with movie dialogue and music. Our work offers a quantitative lens into the narrative technique of audio description.
Chinese Translation
音频描述是对电影视觉元素的叙述技术，旨在帮助盲人和视力障碍者理解故事。然而，它的限制远比大多数叙事要多：描述不仅需要传达电影中的故事，还必须填补对话之间的空白，并遵循各地区存在的指导方针。在本研究中，我们比较了在英国和美国制作的音频描述。我们使用为这两个地区编写的指导方针，以及来自该领域实践者的印象，来提出关于差异的具体假设。我们将这些假设与我们现有的语料库进行测试，该语料库为206部电影提供了人类创作的美国和英国音频描述。结果提供了定量证据，支持所有测试的假设，包括词汇差异、进行时态的使用、被动结构的使用、主观形容词和修饰语的使用、角色命名的时机、场景提示的方式，以及与电影对话和音乐的重叠程度。我们的研究为音频描述的叙事技术提供了定量视角。
cs.CL / 101 / 2608.09802
SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring
SWE-Bench ProMax：大规模多语言代码重构的代理基准测试
Abstract
As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.4 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.
Chinese Translation
随着人工智能编码代理承担越来越复杂、长期的软件工程任务，现有基准测试迅速饱和，其评估质量受到严重质疑：最近的一次审计发现，近60%的未解决的SWE-bench验证实例包含缺陷测试——要么是过于狭窄的测试，拒绝正确的解决方案，要么是过于宽泛的测试，检查未说明的要求——而前沿模型可以逐字复制训练数据中的金标准补丁。代码重构需要在多个文件之间进行协调且保持行为不变的更改，提供了一个更具挑战性和现实性的代理能力测试，但当前基准测试对此关注不足。我们引入了SWE-Bench ProMax，这是一个由专家策划的多语言代码重构基准，包含170个实例，来源于七种编程语言（Python、Java、TypeScript、Go、C、C++和Rust）的真实提交。每个实例都经过严格的多阶段策划，直接解决了先前基准测试中识别的质量问题：问题描述从头开始重写，以提供精确、明确的规范，测试套件经过人工审查，以去除过于狭窄和过于宽泛的测试。复杂性不足或跨文件范围有限的任务被过滤掉，最终形成一个具有挑战性的大规模重构任务基准，每个实例平均涉及11.4个修改文件和261.6行代码，远远超过现有基准的规模。在两种代理框架下对前沿模型的实验表明，最佳模型的解决率仅为41.2%，确认SWE-Bench ProMax为当前人工智能编码代理提供了一个有意义且未饱和的挑战。我们的基准可在https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax获取。
cs.CL / 102 / 2608.09834
RA-FinBERT: Rule-aware LoRA adaptation for low-resource financial sentiment classification
RA-FinBERT：基于规则的低资源金融情感分类的LoRA适应
Abstract
Financial sentiment analysis converts unstructured financial news into quantitative signals that can support market analysis and decision-making. Existing work on resource-efficient financial NLP has largely focused on compressing or adapting pretrained language models, with less attention to combining contextual representations with lightweight rule-derived features. This study develops Rule-Aware FinBERT (RA-FinBERT), a parameter-efficient framework that integrates low-rank adaptation (LoRA) with three continuous VADER-derived sentiment proportions (positive, negative, and neutral) and a source-level metadata feature. The standardized four-dimensional feature vector is directly concatenated with the 768-dimensional final-layer FinBERT [CLS] representation and passed through a lightweight classification head. This design introduces only 1,024 additional trainable weights relative to a structurally matched text-only FinBERT model. RA-FinBERT was evaluated against text-only FinBERT and a lightweight DistilBERT baseline for three-class sentiment classification of financial-news titles and descriptions. On the held-out test set, RA-FinBERT achieved 69.89% accuracy and a macro F1 score of 0.634, compared with 63.44% and 0.526 for text-only FinBERT. Neutral-class recall increased from 18.18% to 45.45%. The framework supports both CPU and GPU execution, offering a lightweight and practical approach to financial sentiment classification under constrained computational resources. These findings indicate that rule-derived sentiment information and source metadata can provide complementary signals to contextual FinBERT representations and improve performance with minimal additional model complexity.
Chinese Translation
金融情感分析将非结构化的金融新闻转化为可以支持市场分析和决策的定量信号。现有的资源高效金融自然语言处理（NLP）研究主要集中在压缩或适应预训练语言模型上，而较少关注将上下文表示与轻量级的规则派生特征相结合。本研究开发了规则感知的FinBERT（RA-FinBERT），这是一个参数高效的框架，将低秩适应（LoRA）与三个连续的VADER派生情感比例（正面、负面和中性）以及一个源级元数据特征相结合。标准化的四维特征向量直接与768维的最终层FinBERT [CLS] 表示拼接，并通过轻量级分类头进行处理。该设计相较于结构匹配的仅文本FinBERT模型，仅引入了1,024个额外的可训练权重。RA-FinBERT与仅文本的FinBERT和轻量级的DistilBERT基线进行了评估，针对金融新闻标题和描述的三类情感分类。在保留的测试集上，RA-FinBERT达到了69.89%的准确率和0.634的宏F1分数，而仅文本的FinBERT分别为63.44%和0.526。中性类的召回率从18.18%提升至45.45%。该框架支持CPU和GPU执行，提供了一种在受限计算资源下进行金融情感分类的轻量级和实用的方法。这些发现表明，规则派生的情感信息和源元数据可以为上下文FinBERT表示提供互补信号，并在模型复杂性增加最小的情况下提高性能。
cs.CL / 103 / 2608.09893
Fusion Training for Mathematical Generalization in Large Language Models
大语言模型中的数学泛化融合训练
Abstract
Thinking Mode Fusion (TMF) enables large language models to support both concise responses and long-form reasoning by unifying a non-thinking mode and a thinking mode within a single model. However, its training dynamics, including the \emph{data ratio} and \emph{training schedule} between the two modes, remain underexplored. In this work, we present a systematic study of TMF by analyzing the effects of the training schedule and data ratio between thinking and non-thinking modes. Focusing on mathematical problem solving, we construct a benchmark with multiple thinking-to-non-thinking data ratios and three training schedules. Our results reveal an asymmetric interaction between the two modes: increasing the ratio of non-thinking supervision reduces the accuracy of the thinking mode. We further show that different training schedules modulate this trade-off and that the optimal schedule depends on the data ratio. Finally, we quantify a negative correlation between non-thinking and thinking mode supervision, highlighting an inherent tension between these two modes. These findings provide practical guidance for designing effective TMF training settings. All code and data are released to support further research at: \href{https://github.com/caocongfeng/Fusion-Bench.git}{\textbf{Fusion Bench}}.
Chinese Translation
思维模式融合（Thinking Mode Fusion, TMF）使得大语言模型能够在单一模型中统一非思维模式和思维模式，从而支持简洁的响应和长篇推理。然而，其训练动态，包括两种模式之间的数据比例和训练计划，仍然未被充分探讨。在本研究中，我们通过分析思维模式和非思维模式之间的训练计划和数据比例的影响，系统地研究了TMF。我们专注于数学问题解决，构建了一个具有多种思维到非思维数据比例和三种训练计划的基准。我们的结果揭示了两种模式之间的不对称互动：增加非思维监督的比例会降低思维模式的准确性。我们进一步表明，不同的训练计划调节了这种权衡，并且最佳计划依赖于数据比例。最后，我们量化了非思维和思维模式监督之间的负相关性，突显了这两种模式之间的固有紧张关系。这些发现为设计有效的TMF训练设置提供了实用指导。所有代码和数据已发布，以支持进一步的研究，链接为： extbf{Fusion Bench}。
cs.CL / 104 / 2608.09898
Consilience for Verifier-Free Test-Time Scaling
无验证者的测试时间扩展的一致性
Abstract
Test-time scaling often uses an external verifier, such as compilers and test cases in coding or trained value functions in robotics applications, to obtain high-quality rollouts. Verifier-free test-time scaling (or VF-TTS) is gaining extensive attention as a mechanism to enhance Large Language Model (LLM) reasoning, primarily because we do not have access to such high-quality verifiers in many real-world applications. Among existing VF-TTS methods, confidence-based VF-TTS methods, which compute and rank rollouts solely by confidence, are particularly promising. Such methods introduce near-zero overhead for sample evaluation and require minimal access to internal model states, making the methods highly flexible across models and tasks. In this paper, we demonstrate a critical limitation of existing confidence-based VF-TTS methods by showing that such methods catastrophically break down on complex tasks. We observe a very interesting phenomenon: uniformly high confidence frequently indicates a failure to explore, favoring confidently wrong answers. To address this, our core insight is that robust cognitive search requires a specific confidence trajectory pattern: such methods perform exploratory branching at the beginning, as manifested by low initial confidence, and converge to a high final confidence solution. To implement this insight, we introduce consilience, a novel selection framework that explicitly evaluates the temporal asymmetry of confidence in reasoning. We operationalize this via a combinatorial metric that actively penalizes high initial confidence while strictly demanding final certainty. Extensive experiments covering both graduate-level mathematics problems and free-form code generation demonstrate that consilience effectively outperforms existing baselines, validating our novel perspective on completion confidence.
Chinese Translation
测试时间扩展通常使用外部验证者，例如编译器和编码中的测试用例或机器人应用中的训练值函数，以获得高质量的展开。无验证者的测试时间扩展（Verifier-Free Test-Time Scaling，VF-TTS）作为一种增强大型语言模型（Large Language Model，LLM）推理的机制，正受到广泛关注，主要是因为在许多现实应用中我们无法获得如此高质量的验证者。在现有的VF-TTS方法中，基于置信度的VF-TTS方法，单纯通过置信度计算和排名展开，尤其有前景。这类方法在样本评估中引入近乎零的开销，并且对内部模型状态的访问要求极低，使得这些方法在模型和任务之间具有高度的灵活性。在本文中，我们展示了现有基于置信度的VF-TTS方法的一个关键局限性，表明这些方法在复杂任务上会灾难性地崩溃。我们观察到一个非常有趣的现象：均匀的高置信度往往表明探索失败，倾向于错误地偏向自信的错误答案。为了解决这个问题，我们的核心见解是，稳健的认知搜索需要特定的置信度轨迹模式：这种方法在开始时表现出低初始置信度的探索性分支，并收敛到一个高最终置信度的解决方案。为了实现这一见解，我们引入了一致性（consilience），一种新颖的选择框架，明确评估推理中置信度的时间不对称性。我们通过一个组合度量来实现这一点，该度量积极惩罚高初始置信度，同时严格要求最终的确定性。覆盖研究生水平数学问题和自由形式代码生成的广泛实验表明，一致性有效地超越了现有基线，验证了我们对完成置信度的新视角。
cs.CL / 105 / 2608.09900
Decoding-Level Taboo: A Diagnostic Stress Test for LLM Robustness
解码层面的禁忌：大型语言模型鲁棒性的诊断压力测试
Abstract
Large language model evaluations typically focus on performance under nominal conditions, creating an illusion of capability where models comfortably walk a narrow, highly optimized generation corridor. In real-world deployments, however, complex system prompts, safety guardrails, and structural constraints continuously force models off this nominal path, driving a divergence between benchmark scores and deployment performance. To address this issue, we introduce Decoding-Level Taboo, a zero-prompt diagnostic stress test that intervenes directly in logit space at runtime, forcing models out of their nominal paths. By dynamically masking primary candidate tokens at word boundaries, Taboo forces machine circumlocution. Evaluating Taboo across several open-weight model families reveals that off-path robustness is heavily influenced by both parameter scale and post-training instruction alignment, with robustness generally improving with model size and alignment. Beyond the results presented in this paper, Taboo provides a novel primitive for generating diverse synthetic datasets, stress-testing runtime safety guardrails, and auditing model reliability prior to real-world deployment.
Chinese Translation
大型语言模型的评估通常侧重于在正常条件下的性能，这造成了一种能力的错觉，使得模型能够轻松地在狭窄且高度优化的生成通道中运行。然而，在实际部署中，复杂的系统提示、安全防护措施和结构性约束不断迫使模型偏离这一正常路径，从而导致基准分数与部署性能之间的差异。为了解决这一问题，我们引入了解码层面的禁忌（Decoding-Level Taboo），这是一种零提示的诊断压力测试，直接在运行时干预对数空间，迫使模型偏离其正常路径。通过在单词边界动态屏蔽主要候选标记，禁忌迫使机器进行迂回表达。对多个开放权重模型系列进行的禁忌评估表明，偏离路径的鲁棒性受到参数规模和训练后指令对齐的显著影响，鲁棒性通常随着模型规模和对齐程度的提高而改善。除了本文中呈现的结果，禁忌还提供了一种新颖的原语，用于生成多样的合成数据集、压力测试运行时安全防护措施，以及在实际部署之前审计模型的可靠性。
cs.CL / 106 / 2608.09925
From Values to Benchmarks: Evaluating Large Language Models for Governmental Use in Dutch
从价值观到基准：评估荷兰政府使用的大型语言模型
Abstract
Large language models are increasingly being deployed in governmental settings, yet few existing evaluation frameworks jointly reflect the values of public administration and the linguistic requirements of non-English contexts. We present the "Grip on LLMs" framework, a systematic evaluation suite for Dutch governmental use developed in collaboration with domain experts from a major Dutch municipal organisation. Through an advisory board process, user research, and a survey of the users of a civil-servant chatbot, we identify six evaluation dimensions (factuality, honesty, social bias, energy consumption, cost, and training data transparency) and operationalise them into a benchmark suite covering more than 30 multilingual and Dutch-specific models. Our results reveal that no single model excels across all dimensions, and that trade-offs are unavoidable: higher quality consistently comes at greater environmental impact and financial cost, while bias remains largely independent of both. We further find that factuality (whether a model answers correctly) and honesty (whether a model acknowledges what it does not know) are governed by distinct properties, with high factuality not implying high honesty. To make these findings actionable for non-technical audiences, we release a publicly accessible, user-friendly model overview designed for the full range of stakeholders involved in governmental LLM selection, from engineers to policymakers.
Chinese Translation
大型语言模型越来越多地应用于政府环境中，但现有的评估框架很少同时反映公共管理的价值观和非英语环境的语言要求。我们提出了“Grip on LLMs”框架，这是一个为荷兰政府使用而开发的系统评估套件，旨在与来自荷兰一家主要市政组织的领域专家合作。通过咨询委员会过程、用户研究以及对公务员聊天机器人的用户调查，我们确定了六个评估维度（事实性、诚实性、社会偏见、能源消耗、成本和训练数据透明度），并将其操作化为一个基准套件，涵盖了30多个多语言和荷兰特定模型。我们的结果表明，没有单一模型在所有维度上都表现优异，且不可避免地存在权衡：更高的质量通常伴随着更大的环境影响和财务成本，而偏见在很大程度上独立于这两者。我们进一步发现，事实性（模型是否正确回答）和诚实性（模型是否承认其不知道的内容）由不同的属性决定，高事实性并不意味着高诚实性。为了使这些发现对非技术受众具有可操作性，我们发布了一个公开可访问、用户友好的模型概述，旨在为参与政府大型语言模型选择的各类利益相关者提供服务，从工程师到政策制定者。