arXiv Daily Digest

589

Papers

Seeing What the Vehicle Sees: Video-Augmented Virtual Reality for Physical Autonomous Vehicles

观察车辆所见：用于物理自主车辆的视频增强虚拟现实

Islam, Md Tanjemul, Shafin, Mohammad, Kabir, Md Rafiul

Abstract

Autonomous vehicles are expected to improve road safety and efficiency, but passengers often remain uncertain about what the vehicle perceives and why it acts as it does. Virtual reality (VR) offers a safe and repeatable medium for presenting this information, yet most passenger-facing VR studies rely on fully simulated vehicles or pre-scripted scenarios, so the motion and perception shown to the user do not originate from a physically operating autonomous system. This paper presents a video-augmented VR framework that couples a physical ROS 2 autonomous robot vehicle to a Unity 6 application deployed on a Meta Quest 3S headset. The vehicle state and live onboard camera stream are transmitted over two independent communication channels, allowing the virtual vehicle to mirror the physical robot's motion while the passenger simultaneously views the vehicle's first-person camera feed and its navigation decisions through an in-vehicle dashboard interface. We evaluate the framework over 20 repeated closed-loop navigation trials. The system achieves a mean state-update latency of 29.63 ms, a mean relative route-progress error of 2.28% between the physical and virtual vehicles, and video delivery at 10.006 frames per second with 0.25% frame loss. All monitored navigation decisions were correctly reflected in the VR interface with no missed or incorrect notifications. The results indicate that the framework can support temporally synchronized, semantically consistent, and accurate route-progress representation for immersive observation of physical autonomous-vehicle behavior.

Chinese Translation

自主车辆有望提高道路安全性和效率，但乘客常常对车辆的感知及其行为原因感到不确定。虚拟现实（VR）提供了一种安全且可重复的信息呈现媒介，但大多数面向乘客的VR研究依赖于完全模拟的车辆或预先编写的场景，因此展示给用户的运动和感知并不源于实际操作的自主系统。本文提出了一种视频增强虚拟现实框架，将物理ROS 2自主机器人车辆与部署在Meta Quest 3S头显上的Unity 6应用程序相结合。车辆状态和实时车载摄像头流通过两个独立的通信通道传输，使虚拟车辆能够反映物理机器人运动，同时乘客可以通过车内仪表盘界面同时查看车辆的第一人称摄像头画面及其导航决策。我们在20次重复的闭环导航试验中评估了该框架。系统实现了29.63毫秒的平均状态更新延迟，物理和虚拟车辆之间的平均相对路线进度误差为2.28%，视频传输速率为每秒10.006帧，帧丢失率为0.25%。所有监测到的导航决策在VR界面中均得到了正确反映，没有遗漏或错误的通知。结果表明，该框架能够支持时间同步、语义一致和准确的路线进度表示，以便沉浸式观察物理自主车辆的行为。

View on arXiv Download PDF AI Translation

cs.RO / 2 / 2609.13231

ShieldVLA: Feasibility-Aware Safety Alignment for Vision-Language-Action Models

ShieldVLA：面向可行性意识的视觉-语言-动作模型安全对齐

Tayal, Manan, Nambi, Akshay

Abstract

Vision-Language-Action (VLA) models demonstrate strong generalization in robotic manipulation and navigation, but existing fine-tuning methods provide limited safety guarantees. Current approaches primarily rely on Lagrangian optimization that enforces safety through soft penalties on expected cumulative cost, often resulting in residual constraint violations or overly conservative behavior. Moreover, learning safety in visual domains is challenging due to the absence of dense per-step safety annotations. We propose ShieldVLA, a safety-aligned fine-tuning framework for VLA models based on Hamilton-Jacobi (HJ) reachability. ShieldVLA learns a model-free approximation of the HJ reachability value function directly from visual observations to estimate the safe operating region. The learned safety critic gates policy optimization by separating reward maximization within feasible regions from recovery near unsafe states, avoiding persistent reward-cost trade-offs. To enable scalable supervision in visual environments, we introduce rubric-based VLM safety scores that convert semantic safety feedback into structured critic targets without requiring manual cost labels. Across five navigation and manipulation benchmarks spanning multiple VLA backbones, ShieldVLA reduces cumulative safety cost by 57% on average and improves task success rate by +0.13 over SafeVLA.

Chinese Translation

视觉-语言-动作（VLA）模型在机器人操作和导航中表现出强大的泛化能力，但现有的微调方法提供的安全保障有限。目前的方法主要依赖拉格朗日优化，通过对预期累积成本施加软惩罚来强制执行安全性，通常导致残余约束违反或过于保守的行为。此外，由于缺乏密集的逐步安全注释，在视觉领域学习安全性是具有挑战性的。我们提出了ShieldVLA，这是一个基于汉密尔顿-雅可比（HJ）可达性的VLA模型安全对齐微调框架。ShieldVLA直接从视觉观测中学习HJ可达性值函数的无模型近似，以估计安全操作区域。学习到的安全评估器通过将可行区域内的奖励最大化与接近不安全状态的恢复分开，从而在策略优化中起到门控作用，避免了持续的奖励-成本权衡。为了在视觉环境中实现可扩展的监督，我们引入了基于评分标准的VLM安全评分，将语义安全反馈转换为结构化的评估目标，而无需手动成本标签。在涵盖多个VLA骨干网络的五个导航和操作基准测试中，ShieldVLA平均减少了57%的累积安全成本，并使任务成功率提高了0.13。

View on arXiv Download PDF AI Translation

cs.RO / 3 / 2609.13234

Harnessing human expertise for high-precision robotic assembly in industrialized construction: A sample-efficient installer-in-the-loop interactive reinforcement learning framework

利用人类专业知识实现工业化建筑中的高精度机器人组装：一种样本高效的安装者循环交互强化学习框架

Jin, Zekai, Wang, Huiguang, Sun, Xiaoning, Shao, Yi

Abstract

Industrialized construction imposes stringent precision requirements on robotic assembly of modular components such as prefabricated window units. In tolerance-critical operations, the central bottleneck is not only mechanical clearance but also converting tacit installer expertise into data-efficient autonomy under sparse acceptance feedback, contact variability, and millimeter-scale constraints. We present an installer-in-the-loop interactive reinforcement learning framework that acquires expertise through offline teleoperated demonstrations, sparse event-driven binary takeovers at contact-failure boundaries, and acceptance-aligned terminal rewards, logged under a unified schema for traceable offline-to-online adaptation. A temporally abstract action-sequence policy built on Q-chunking with Flow Q-Learning captures multimodal recovery maneuvers under sparse terminal rewards, while a non-updating warm-start phase stabilizes the offline-to-online transition. The framework is evaluated in MuJoCo across the workflow from suction acquisition through clearance-limited seating, under structured staging and end-to-end randomized placement. Within a defined stress-test regime with 2 mm per-side clearance, bounded pose perturbations, and friction randomization, the pipeline attains 100\% autonomous seating with 12--15 min of cumulative installer supervision over 3.0 h of online training, and reaches the 95\% success milestone in approximately 0.5 h and 1.5 h in the two experiments. We also report wall-clock adaptation time, cumulative takeover minutes, intervention-rate decay, and stage-wise failure attribution to inform supervision budgeting. Ablations isolate the complementary contributions of temporal abstraction, installer intervention, and warm-start value calibration.

Chinese Translation

工业化建筑对模块化组件（如预制窗单元）的机器人组装提出了严格的精度要求。在容差关键操作中，中心瓶颈不仅在于机械间隙，还在于如何在稀疏的接受反馈、接触变异性和毫米级约束下，将隐性安装者专业知识转化为数据高效的自主性。我们提出了一种安装者循环交互强化学习框架，通过离线遥操作演示、在接触故障边界的稀疏事件驱动二元接管以及与接受对齐的终端奖励来获取专业知识，这些数据在统一的模式下记录，以便于可追溯的离线到在线适应。基于Q分块和流Q学习的时间抽象动作序列策略捕捉了在稀疏终端奖励下的多模态恢复操作，而非更新的热启动阶段则稳定了离线到在线的过渡。该框架在MuJoCo中进行了评估，涵盖了从吸取到限位座位的工作流程，采用结构化分阶段和端到端随机放置。在定义的压力测试条件下，具有每侧2毫米的间隙、有限的姿态扰动和摩擦随机化，该管道在3.0小时的在线训练中实现了100%的自主座位，累计安装者监督时间为12至15分钟，并在两个实验中分别在约0.5小时和1.5小时内达到95%的成功里程碑。我们还报告了实际适应时间、累计接管分钟、干预率衰减和阶段性失败归因，以指导监督预算。消融实验隔离了时间抽象、安装者干预和热启动价值校准的互补贡献。

View on arXiv Download PDF AI Translation

cs.RO / 4 / 2609.13235

Learning Manipulation-Sufficient Representations via Outcome Bottlenecks

通过结果瓶颈学习足够的操作表示

Sarowar, Md Selim, Kim, Sungho

Abstract

Networked manipulation endpoints couple perception to actuation across compute- and bandwidth-limited links, yet commonly exchange dense geometric states optimized for fidelity rather than action outcomes. A stochastic representation is learned with a policy-free, action-conditioned outcome bottleneck: marginal outcome log-loss supplies distortion and a KL term regularizes rate. The construction is motivated by the minimal statistic that preserves the outcome distribution of every admissible action, while the implemented finite model is evaluated as a rate-regularized mixture predictor. The same encoder and outcome head support grasp selection, singleton conformal filtering, active viewpoint selection, and latent test-time adaptation. A finite-probe theorem identifies the local level-set tangent space with the null space of an outcome Jacobian. The synthetic oracle verifies this result; on scanned objects, an analytic surrogate agrees with measured simulator invariances within $0.56^\circ$. Across 11,979 simulated grasps on 13 objects, a reconstructed-geometry wrench score attains 0.542 AUC against lift success and falls below chance on curved objects, while our representation attains 0.876. At 25\% commitment, executed-grasp success is 0.503 versus 0.984. The 512-byte interface is $288\times$ smaller than one RGB-D frame and runs at 16\,ms per CPU decision. Independent synthetic points track the tested conformal levels; scanned-object all-pair coverage is reported as a clustered empirical diagnostic. On unseen objects, within-scene AUC falls to 0.569, and a full-feedback update raises empirical mean pairwise coverage from 0.728 to 0.883.

Chinese Translation

网络化的操作端点将感知与执行连接在一起，跨越计算和带宽受限的链接，然而通常交换的是为保真度而优化的密集几何状态，而非动作结果。通过无策略、基于动作条件的结果瓶颈学习到了一种随机表示：边际结果对数损失提供了失真，而KL项则对速率进行正则化。这一构建的动机在于保留每个可接受动作的结果分布的最小统计量，而实现的有限模型则作为速率正则化的混合预测器进行评估。相同的编码器和结果头支持抓取选择、单例共形滤波、主动视角选择和潜在测试时适应。有限探测定理识别了局部水平集切空间与结果雅可比矩阵的零空间。合成神谕验证了这一结果；在扫描物体上，解析替代与测量的模拟器不变性一致，误差在0.56度以内。在13个物体上进行的11,979次模拟抓取中，重建几何的扭矩得分在提升成功率方面达到了0.542 AUC，而在曲面物体上则低于随机水平，而我们的表示达到了0.876。在25%的承诺下，执行抓取成功率为0.503，而在0.984。512字节的接口比一个RGB-D帧小288倍，并且每个CPU决策的运行时间为16毫秒。独立的合成点跟踪测试的共形水平；扫描物体的全对覆盖率被报告为聚类经验诊断。在未见物体上，场景内的AUC降至0.569，而全反馈更新将经验均值成对覆盖率从0.728提高到0.883。

View on arXiv Download PDF AI Translation

cs.RO / 5 / 2609.13236

Self-Evolving AI for Humanoids: Mechanisms, Safety, and Evaluation of Post-Deployment Self-Improvement

自我进化的类人机器人人工智能：机制、安全性及部署后自我改进的评估

Nguyen, Loc X., Raha, Avi Deb, Le, Huy Q., Huh, Eui-Nam, Niyato, Dusit, Hong, Choong Seon

Abstract

Humanoid robots are becoming an important part of embodied artificial intelligence, driven by advances in reinforcement learning for locomotion, world models for prediction, and vision-language-action models for general control. However, most of these systems remain static after deployment. A policy is trained offline for a fixed objective and then frozen, even though the tasks, environments, and robot bodies keep drifting over time. An emerging paradigm of self-evolving agents aims to address this problem by allowing systems to improve from their own post-deployment experience. Since most existing studies focus on disembodied software agents, this survey examines how self-evolution changes when an agent has a physical body. We first define self-evolution for humanoids and represent a deployed robot using a state tuple that includes its policy, perception, memory, workflow, and body. This state is updated by an evolution operator in a slow outer loop with a lifelong objective. We then organize the literature into four complementary mechanisms of self-evolution, presented in increasing order of autonomy: self-learning, self-adaptation, self-optimization, and self-generation. Since changes to a humanoid can introduce physical hazards, we treat safety and uncertainty as key design dimensions of the evolution operator, and further formulate admissible evolution as a constraint enforced by a world-model verification gate within a human-oversight envelope. Finally, we present that evaluation should track the robot's evolving trajectory rather than a fixed checkpoint, and we identify the lack of a benchmark designed specifically for self-evolving humanoids. Moreover, we outline open challenges spanning AI algorithms, on-board systems, and governance.

Chinese Translation

类人机器人正成为具身人工智能的重要组成部分，这得益于在运动强化学习、用于预测的世界模型以及用于一般控制的视觉-语言-动作模型方面的进展。然而，大多数这些系统在部署后仍然是静态的。策略是在离线情况下为固定目标训练的，然后被冻结，尽管任务、环境和机器人身体随着时间的推移不断变化。自我进化代理的新兴范式旨在通过允许系统从自身的部署后经验中改进来解决这一问题。由于现有研究大多集中于无身体的软件代理，本调查研究了当代理具有物理身体时自我进化的变化。我们首先为类人机器人定义自我进化，并使用状态元组表示已部署的机器人，该元组包括其策略、感知、记忆、工作流程和身体。这个状态通过在一个以终身目标为导向的缓慢外循环中的进化算子进行更新。然后，我们将文献组织为四种互补的自我进化机制，按自主性递增的顺序呈现：自我学习、自我适应、自我优化和自我生成。由于对类人的改变可能引入物理危险，我们将安全性和不确定性视为进化算子的关键设计维度，并进一步将可接受的进化形式化为由人类监督范围内的世界模型验证门强制执行的约束。最后，我们提出评估应跟踪机器人的进化轨迹，而不是固定的检查点，并指出缺乏专门为自我进化类人机器人设计的基准。此外，我们概述了跨越人工智能算法、机载系统和治理的开放挑战。

View on arXiv Download PDF AI Translation

cs.RO / 6 / 2609.13243

GzDRL: Reproducible and Scalable Deep Reinforcement Learning with Gazebo

GzDRL：具有可重复性和可扩展性的深度强化学习框架与Gazebo

Haridevan, Amal Dev, Kang, Junjie, Shan, Jinjun

Abstract

We present GzDRL, a novel single-process reinforcement learning (RL) framework for Gazebo that overcomes longstanding bottlenecks in scalable, reproducible robotics experimentation. Unlike conventional middleware-based RL-Gazebo integrations that suffer from nondeterminism and irreproducibility, GzDRL introduces a systematic, middleware-free environment-stepping mechanism that directly synchronizes agent actions and physics updates. This design enables deterministic, high-throughput data collection, efficient vectorization, and reproducible RL training and evaluation. Comprehensive benchmarks demonstrate that GzDRL achieves the highest workstation throughput among the evaluated frameworks while remaining competitive with GPU-accelerated simulators on laptop hardware, and maintains precise agent-environment synchronization, multi-agent scalability, and experiment-level reproducibility. We further validate sim-to-real transfer by deploying learned policies directly onto a physical quadrotor, without fine-tuning. Our results establish GzDRL as an accessible and reproducible platform for advancing RL in robotics and automation.

Chinese Translation

我们提出了GzDRL，这是一个新颖的单进程强化学习（RL）框架，旨在解决可扩展、可重复的机器人实验中的长期瓶颈。与传统的基于中间件的RL-Gazebo集成方法不同，后者存在非确定性和不可重复性的问题，GzDRL引入了一种系统化的无中间件环境步进机制，直接同步智能体的动作和物理更新。该设计实现了确定性、高吞吐量的数据收集、高效的向量化，以及可重复的RL训练和评估。全面的基准测试表明，GzDRL在评估的框架中实现了最高的工作站吞吐量，同时在笔记本硬件上与GPU加速的模拟器保持竞争力，并保持精确的智能体-环境同步、多智能体可扩展性和实验级可重复性。我们进一步通过将学习到的策略直接部署到物理四旋翼上来验证仿真到现实的转移，而无需微调。我们的结果确立了GzDRL作为一个可访问且可重复的平台，以推动机器人和自动化领域的强化学习发展。

View on arXiv Download PDF AI Translation

cs.RO / 7 / 2609.13244

Physical Kernel: Structured Visual Latents for Dark Manipulation

物理核：用于暗操作的结构化视觉潜变量

Hang, Jinting, Li, Hong, Cai, Zhenhui, Zhao, Zhihao, He, Jian

Abstract

We study dark manipulation: after a brief lit Write encodes z0 = Enc(rgb), a policy pi(z) and open-loop dynamics f(z,a) complete contact-rich skills without further pixels (dark_f). On ManiSkill StackCube (n=160; seed packs 0/1000), dark_f attains 68.1% stacked on the five-rung chain (near_A -> grasped -> lifted -> on_B -> stacked), compared with 35.6% for per-step lit_reenc and 0% for freeze/encode_black. On a shared Write->HOLD protocol (n=40), occlusion and camera-aligned GT contact-neighbor masks drive lit lift from 43% to 0%, while dark_f holds 82.5%; shuffling actions inflates dynamics MSE by ~9.4x; write-time appearance shifts break encoding (night: 0% stacked), yet the same shifts during HOLD leave dark_f lift unchanged; Write length Tw is flat once the stop phase is reached, while earlier stops and write-time blur/JPEG sharply cut stacked. A dedicated pi_write reaches 35% vision-budget stacked (n=80); matched Dreamer-style/pixel nulls without privileged geom stay at 0%. Privileged state-RSSM MPC reaches ~35% stacked with 9D dark observations -- a stronger-observation null, not a matched visual baseline.

Chinese Translation

我们研究暗操作：在简短的光照写入编码 z0 = Enc(rgb) 后，策略 pi(z) 和开环动态 f(z,a) 完成了丰富接触技能，而无需进一步的像素（dark_f）。在 ManiSkill StackCube（n=160；种子包 0/1000）上，dark_f 在五级链（near_A -> grasped -> lifted -> on_B -> stacked）上达到了 68.1%，而逐步的 lit_reenc 仅为 35.6%，freeze/encode_black 则为 0%。在共享的 Write->HOLD 协议（n=40）中，遮挡和与相机对齐的 GT 接触邻居掩码使得光照提升从 43% 降至 0%，而 dark_f 保持在 82.5%；打乱动作使动态均方误差（MSE）增加约 9.4 倍；写入时间的外观变化破坏了编码（夜间：0% 堆叠），然而在 HOLD 期间相同的变化使 dark_f 的提升保持不变；一旦达到停止阶段，写入长度 Tw 变得平坦，而较早的停止和写入时间模糊/JPEG 显著减少了堆叠。专用的 pi_write 达到 35% 的视觉预算堆叠（n=80）；匹配的 Dreamer 风格/像素空值在没有特权几何的情况下保持在 0%。特权状态-RSSM MPC 在 9D 暗观察下达到 ~35% 堆叠——这是一个更强的观察空值，而不是匹配的视觉基线。

View on arXiv Download PDF AI Translation

cs.RO / 8 / 2609.13270

Conflict-Predictive Variable Horizons in Multi-Drone Distributed Model Predictive Control

多无人机分布式模型预测控制中的冲突预测可变预测范围

Mümken, Linda, Schwung, Michael, Lier, Stefan, Schwung, Andreas

Abstract

In distributed model predictive control for multi-drone collision avoidance, a fixed prediction horizon forces a compromise: a short horizon is inexpensive but reacts late to approaching neighbors, whereas a long one anticipates conflicts at a per-step cost that grows superlinearly with its length. We propose a conflict-predictive variable horizon that each drone sets locally, leaving the distributed model predictive control itself unchanged. From a short history of observed positions, a drone extrapolates the flight lines of its neighbors, tests each against its own using confidence funnels that narrow with prediction range, and obtains each time to conflict in closed form. The horizon is then the smallest admissible value whose planning window covers the farthest predicted conflict. It collapses to its minimum in clear airspace and grows only when a conflict lies ahead. Provided this minimum meets a single computable feasibility bound, we prove that recursive feasibility and asymptotic stability are preserved for every horizon the policy can select. These guarantees hold for a linear model, and a cascaded inner loop reduces each quadrotor's translational dynamics to a perturbed double integrator, so they carry over to the linearized quadrotor model and, as practical stability, to the full nonlinear one. In simulation on dense antipodal-swap benchmarks, the variable horizon reduces both per-step solver cost and total computation well below those of a long fixed horizon, and it maintains separation in every run, which a short fixed horizon of comparable per-step cost does not.

Chinese Translation

在多无人机碰撞避免的分布式模型预测控制中，固定的预测范围迫使我们进行妥协：短预测范围成本低，但对接近的邻居反应迟缓，而长预测范围则以超线性增长的每步成本来预判冲突。我们提出了一种冲突预测可变预测范围，每架无人机在本地设置，保持分布式模型预测控制本身不变。无人机从短期观察到的位置中推断邻居的飞行轨迹，利用随着预测范围缩小的置信漏斗对每条轨迹进行测试，并以封闭形式获得每次冲突时间。然后，预测范围是覆盖最远预测冲突的最小可接受值。在清晰的空域中，它收缩到最小值，仅在前方存在冲突时才增长。只要这个最小值满足单个可计算的可行性边界，我们证明了该策略可以选择的每个预测范围都保持递归可行性和渐近稳定性。这些保证适用于线性模型，并且级联的内部循环将每个四旋翼的平移动态简化为受扰动的双重积分器，因此它们可以转移到线性化的四旋翼模型，并作为实际稳定性转移到完整的非线性模型。在密集的对立交换基准测试中的仿真中，可变预测范围将每步求解器成本和总计算量降低到远低于长固定预测范围的水平，并且在每次运行中保持分离，而与之相当的短固定预测范围则无法做到这一点。

View on arXiv Download PDF AI Translation

cs.RO / 9 / 2609.13289

Extending the Speed Limit of Quadrupedal Locomotion via Refined Actuator Modeling and Adaptive Command Scheduling

通过精细化执行器建模和自适应指令调度扩展四足运动的速度极限

Tao, Yucheng, Cheng, Shaowen, Lan, Guorong, Yuan, Yanyan, Jin, Yongbin, Wang, Hongtao

Abstract

Achieving high-speed locomotion in quadrupedal robots remains highly challenging, as actuators operate near their physical limits and exhibit pronounced nonlinearities. However, many existing methods neglect actuator nonlinearities and physical constraints during training, leading to a significant sim-to-real gap under highly dynamic motions and limiting achievable performance. To address this issue, we propose a high-speed locomotion framework that reduces sim-to-real discrepancies and stabilizes learning over a wide command distribution. A refined actuator model explicitly captures high-speed voltage coupling and magnetic saturation, enabling a more accurate representation of the torque-speed envelope. In addition, a reinforcement learning framework incorporating a two-stage curriculum and adaptive command scheduling (ACS) ensures stable training. Experiments on the 36.5 kg quadruped BlackPanther2 (BP2) demonstrate speeds of up to 13.2 m/s on a treadmill and 11.65 m/s outdoors, establishing a new state-of-the-art and, to the best of our knowledge, a world record for quadrupedal robot locomotion. The results further highlight the importance of accurate actuator modeling in preventing non-physical policy exploitation, and show that ACS improves robustness without sacrificing performance.

Chinese Translation

在四足机器人中实现高速运动仍然面临巨大挑战，因为执行器在其物理极限附近运行，并表现出明显的非线性特征。然而，许多现有方法在训练过程中忽视了执行器的非线性和物理约束，导致在高度动态的运动下出现显著的仿真与现实差距，从而限制了可实现的性能。为了解决这个问题，我们提出了一种高速运动框架，该框架减少了仿真与现实之间的差异，并在广泛的指令分布上稳定学习。精细化的执行器模型明确捕捉了高速电压耦合和磁饱和现象，从而能够更准确地表示扭矩-速度包络。此外，结合两阶段课程和自适应指令调度（ACS）的强化学习框架确保了稳定的训练。在36.5公斤的四足机器人BlackPanther2（BP2）上的实验表明，该机器人在跑步机上达到了13.2米/秒的速度，在户外达到了11.65米/秒，建立了新的技术水平，并且据我们所知，创造了四足机器人运动的世界纪录。结果进一步强调了准确的执行器建模在防止非物理策略利用中的重要性，并显示ACS在不牺牲性能的情况下提高了鲁棒性。

View on arXiv Download PDF AI Translation

cs.RO / 10 / 2609.13290

Breaking speed scaling in quadrupedal robots via Huygens' coupled-pendulum dynamics

通过惠根斯耦合摆动力学突破四足机器人速度缩放限制

Tao, Yucheng, Jin, Yongbin, Cheng, Shaowen, Liu, Xianwei, Yuan, Yanyan, Liang, Yanhong, Su, Chengkai, Fu, Chaojie, Lan, Guorong, Yang, Wei, Wang, Hongtao

Abstract

Achieving biological-level running speeds has largely been pursued through advances in control algorithms, which improve the utilization of existing hardware. However, the ultimate speed limits remain governed by the underlying force and torque requirements of rapid locomotion, which are typically addressed through increased actuator capacity. Inspired by Huygens' coupled pendulums, we demonstrate that superior locomotion can emerge from principled exploitation of intrinsic dynamics rather than brute-force hardware scaling. Inter-limb inertial coupling redistributes energy across the gait cycle and reduces peak joint torque required for rapid periodic motion, thereby expanding the achievable speed without proportional increases in actuator capability. Incorporating hardware parameters as additional design variables further extends this analysis into a co-optimization framework, enabling the systematic utilization of inertial coupling in robot design. Guided by this framework, a quadruped robot achieves a running speed of 10.74 m/s (Froude number 21.4) and completes a 100-meter sprint in 12.2 seconds, representing the first legged robot to surpass 10 m/s. These results establish inertial coupling as an underlying mechanism governing high-speed legged locomotion and highlight its role in reducing force requirements, offering new insights into the design of agile robotic systems.

Chinese Translation

实现生物水平的奔跑速度主要依赖于控制算法的进步，这些算法提高了现有硬件的利用效率。然而，最终的速度限制仍然受到快速运动所需的基本力和扭矩要求的制约，这通常通过增加执行器的能力来解决。受到惠根斯耦合摆的启发，我们展示了优越的运动性能可以通过对内在动力学的原则性利用而非单纯的硬件扩展而实现。肢体间的惯性耦合在步态周期中重新分配能量，并减少了快速周期性运动所需的峰值关节扭矩，从而在不成比例增加执行器能力的情况下扩展了可实现的速度。将硬件参数作为额外设计变量进一步将此分析扩展到一个协同优化框架，使得在机器人设计中系统性地利用惯性耦合成为可能。在该框架的指导下，一种四足机器人实现了10.74 m/s的奔跑速度（弗劳德数21.4），并在12.2秒内完成了100米的冲刺，成为首个超过10 m/s的腿式机器人。这些结果确立了惯性耦合作为高速度腿式运动的基本机制，并突显了其在降低力要求方面的作用，为灵活机器人系统的设计提供了新的见解。

View on arXiv Download PDF AI Translation

cs.RO / 11 / 2609.13295

Ergodic Control and Controlled Diffusion for Robot Learning: Review and Tutorial

机器人学习的遍历控制与受控扩散：综述与教程

Sun, Max Muchen, Bilaloglu, Cem, Rao, Ananya, Ivic, Stefan, Sartoretti, Guillaume, Fitzsimons, Kathleen, Abraham, Ian, Calinon, Sylvain, Murphey, Todd

Abstract

Diffusion learning leverages the statistical mechanism of diffusion processes for learning, reasoning, and inferring complex distributions from data. Recent advances in diffusion learning have been transformative, with robot learning emerging as a key opportunity area, with applications spanning perception, control, and decision-making. At the same time, the statistical mechanism of diffusion processes can be controlled to shape the temporal evolution of the state distribution underlying robot trajectories, inducing ergodic behavior in robotic systems. The frameworks of controlled diffusion and ergodic control were developed around the same time as diffusion learning, and their theories and algorithms have increasingly converged. Ergodicity induced by controlled diffusion has several significant implications for robot learning, distinct from applying diffusion learning to robotics problems: it formally enforces statistical properties required to ensure optimality of robot learning, enables non-myopic search over uncertain information landscapes for data collection, and enables behavior specification based on spatial rather than temporal characteristics of trajectories. This survey introduces the intuition behind controlled diffusion for robot learning, explores its connection to diffusion learning, presents theoretical foundations and numerical tutorials for solving controlled diffusion and ergodic control problems, and reviews applications across robotics. Finally, we discuss key challenges and future opportunities in leveraging controlled diffusion for robot learning.

Chinese Translation

扩散学习利用扩散过程的统计机制进行学习、推理和从数据中推断复杂分布。近年来，扩散学习的进展具有变革性，机器人学习作为一个关键的机会领域应运而生，其应用涵盖感知、控制和决策等方面。同时，扩散过程的统计机制可以被控制，以塑造机器人轨迹下状态分布的时间演变，从而在机器人系统中引发遍历行为。受控扩散和遍历控制的框架与扩散学习几乎同时发展，其理论和算法逐渐趋于一致。受控扩散引发的遍历性对机器人学习具有几个重要的影响，这与将扩散学习应用于机器人问题有所不同：它正式强制执行确保机器人学习最优性所需的统计性质，使得在不确定信息空间上进行非短视搜索以进行数据收集成为可能，并基于轨迹的空间特征而非时间特征来指定行为。本综述介绍了受控扩散在机器人学习中的直觉，探讨了其与扩散学习的联系，呈现了解决受控扩散和遍历控制问题的理论基础和数值教程，并回顾了在机器人领域的应用。最后，我们讨论了利用受控扩散进行机器人学习的关键挑战和未来机会。

View on arXiv Download PDF AI Translation

cs.RO / 12 / 2609.13307

IMM-based Multiple Object Tracking using a State Prediction Neural Network

基于IMM的多目标跟踪方法：一种状态预测神经网络

Lim, Chan-Bin, Paek, Dong-Hee, Kong, Seung-Hyun

Abstract

Object tracking is essential for autonomous vehicles to avoid obstacles and plan routes. Radar maintains detection performance even in adverse weather and can measure relative velocity through the Doppler effect, making it well suited for object tracking. In this paper, we propose a data-driven state PRedictor-based Interacting Multiple Model tracking method (PR-IMM) that improves nonlinear object-motion representation while preserving the stability and interpretability of physics-based motion models. The proposed method employs a transformer-based PRediction model (PR) that incorporates radar Doppler measurements to predict object displacement. The PR model is integrated into the IMM as a mode alongside the CV, CA, and CT motion models, and their prior positions are dynamically combined according to the mode probabilities. Experimental results show that PR-IMM reduces position-estimation error by 57.3% over the IMM and by 16.5% over the PR, while reducing ID switches by 25.3% and improving IDF1 by 9.6% over the IMM.

Chinese Translation

目标跟踪对于自主车辆避开障碍物和规划路线至关重要。雷达在恶劣天气条件下仍能保持检测性能，并通过多普勒效应测量相对速度，使其非常适合用于目标跟踪。本文提出了一种基于数据驱动的状态预测器交互多模型跟踪方法（PR-IMM），该方法在保持物理基础运动模型的稳定性和可解释性的同时，改善了非线性目标运动的表示。所提出的方法采用了一种基于变换器的预测模型（PR），该模型结合了雷达多普勒测量来预测目标位移。PR模型作为一种模式与恒速（CV）、加速（CA）和恒加速度（CT）运动模型一起集成到IMM中，并根据模式概率动态组合它们的先前位置。实验结果表明，PR-IMM在位置估计误差上比IMM减少了57.3%，比PR减少了16.5%，同时ID切换减少了25.3%，IDF1相较于IMM提高了9.6%。

View on arXiv Download PDF AI Translation

cs.RO / 13 / 2609.13318

Attention-DP3: Spatially Object-aware 3D Diffusion Policy via Geometry-aligned Attentional Conditioning

Attention-DP3：通过几何对齐的注意力调节实现空间对象感知的3D扩散策略

Yan, Changbo, Zhang, Zhongbo, Zhang, Zaibin, Wang, Yifan, Wang, Lijun, Lu, Huchuan

Abstract

3D point-cloud observations are inherently ambiguous in complex, cluttered manipulation scenes, where target objects may be partially occluded or tightly intermingled with visually similar distractors. As a result, standard 3D diffusion policies often struggle to localize and exploit task-relevant geometry as scene complexity grows. We propose \textbf{Attention-DP3}, a spatially object-aware 3D diffusion policy that injects object-level geometric cues via attention while keeping the DP3 diffusion backbone unchanged. Our pipeline performs open-vocabulary 2D segmentation on RGB images, then lifts predicted target masks into 3D using calibrated camera geometry to obtain object-centric geometric priors. We incorporate these cues through Tri-field Attentional Conditioning, which constructs three complementary fields: (i) a targetness field to anchor the target object, (ii) an intra-target saliency field to emphasize task-relevant geometry within the target, and (iii) a backgroundness field to suppress distractors and clutter. Experiments on Adroit, DexArt, MetaWorld, and the real-world SO101 platform show consistent improvements over DP3, achieving state-of-the-art performance across benchmarks. Notably, as distractor objects increase, DP3 drops sharply, whereas Attention-DP3 remains stable and outperforms DP3 by up to 31\% under heavy clutter. The code is publicly available at https://github.com/zhangzhongbo2213/Attention-DP3.

Chinese Translation

在复杂且杂乱的操作场景中，3D点云观测本质上是模糊的，目标物体可能部分被遮挡或与视觉上相似的干扰物紧密交织。因此，标准的3D扩散策略在场景复杂性增加时，往往难以定位和利用与任务相关的几何信息。我们提出了 extbf{Attention-DP3}，一种空间对象感知的3D扩散策略，通过注意力机制注入对象级几何线索，同时保持DP3扩散骨干网络不变。我们的流程对RGB图像进行开放词汇的2D分割，然后利用标定的相机几何将预测的目标掩膜提升至3D，以获得以对象为中心的几何先验。我们通过三场注意力调节（Tri-field Attentional Conditioning）整合这些线索，该方法构建了三个互补的场： (i) 目标性场，用于锚定目标物体；(ii) 目标内显著性场，用于强调目标内与任务相关的几何信息；(iii) 背景性场，用于抑制干扰物和杂乱。我们在Adroit、DexArt、MetaWorld和真实世界的SO101平台上的实验显示，Attention-DP3在各基准测试中均显著优于DP3，达到了最先进的性能。值得注意的是，随着干扰物体数量的增加，DP3的性能急剧下降，而Attention-DP3保持稳定，在重度杂乱情况下比DP3提高了多达31 ext%。代码已公开发布在https://github.com/zhangzhongbo2213/Attention-DP3。

View on arXiv Download PDF AI Translation

cs.RO / 14 / 2609.13335

Bridging Thought and Action: Taming Long-Horizon Instability in Open-Source LLM Agents with a MetaTool-Enhanced ROS Framework

桥接思维与行动：通过增强MetaTool的ROS框架驯服开源LLM代理中的长时间不稳定性

Mahmud, Kazi Abrar, Dhurubo, Nilotpaul Kundu, Kirttonia, Tamal, Ujjal, Sabbir Hossain, Haque, Mohammad Ariful

Abstract

Large Language Models (LLMs) have enabled more natural human-robot interaction, but open-source models often exhibit unstable long-horizon reasoning and inefficient action execution when deployed in agentic robotic frameworks. This paper presents an enhanced ROS-Agent based architecture that improves task reliability and execution efficiency for agentic robotic systems using open-source LLMs. The proposed system introduces a novel intermediate mechanism, termed the MetaTool, which enforces structured planning prior to action execution. Given a natural-language command, the MetaTool induces the LLM to generate a pseudo-code plan of intended tool invocations, which is stored in the ROS-Agent's scratchpad and persists throughout execution. By explicitly separating planning from execution, the proposed approach reduces execution loops and improves deterministic behavior. The architecture is validated on a custom mobile robotic platform with multimodal perception and motion control capabilities. Experimental results on real-world interactive tasks demonstrate improved task completion and contextual consistency, with up to ~24% gains on complex tasks compared to the baseline framework.

Chinese Translation

大型语言模型（LLMs）使人机交互变得更加自然，但开源模型在代理机器人框架中部署时，往往表现出不稳定的长时间推理和低效的行动执行。本文提出了一种基于增强ROS-Agent的架构，利用开源LLMs提高代理机器人系统的任务可靠性和执行效率。所提出的系统引入了一种新颖的中介机制，称为MetaTool，该机制在行动执行之前强制进行结构化规划。在接收到自然语言命令后，MetaTool促使LLM生成预期工具调用的伪代码计划，该计划存储在ROS-Agent的临时记忆中，并在整个执行过程中保持有效。通过明确将规划与执行分离，所提出的方法减少了执行循环并改善了确定性行为。该架构在具有多模态感知和运动控制能力的定制移动机器人平台上进行了验证。针对真实世界交互任务的实验结果表明，任务完成率和上下文一致性得到了改善，与基线框架相比，复杂任务的性能提升可达约24%。

View on arXiv Download PDF AI Translation

cs.RO / 15 / 2609.13428

Chance-Constrained Belief-Space Maneuver Planning for Autonomous Collision Avoidance Under Uncertainty

基于机会约束的信念空间机动规划用于不确定条件下的自主避碰

Kim, Grace Ra, Eddy, Duncan, Kochenderfer, Mykel J.

Abstract

Increasing conjunction frequency in low Earth orbit places growing pressure on spacecraft operators to determine not only whether an encounter requires mitigation, but whether sufficient information is available to commit to a maneuver. This work formulates this information-action tradeoff as a belief-space planning problem for conjunctions between a maneuverable spacecraft and an unmaneuverable secondary object. The planner represents the uncertain orbital states as Gaussian beliefs and uses a chance-constrained belief-space Monte Carlo tree search framework to reason over possible future tracking updates before time of closest approach (TCA). A terminal chance constraint limits the probability of reaching TCA above a prescribed collision-risk threshold, allowing the planner to wait for informative tracking while intervening when deferral becomes too risky. We evaluate the approach on eight historical conjunctions from NASA's Conjunction Assessment Risk Analysis dataset. By varying the secondary-object measurement quality and tracking cadence, we generate a total of 96 distinct evaluation scenarios. Across the evaluated conditions, the planner reaches TCA without maneuvering in approximately 40% of episodes while maintaining no terminal collision-risk violations. In contrast, fixed-time rule-based maneuver policies resolve more encounters without maneuvering when intervention is deferred closer to TCA, but at the expense of increasing terminal risk violations. The fraction of episodes reaching TCA without maneuvering depends strongly on tracking quality and measurement cadence, ranging from 76% under accurate, frequent measurements to approximately 18%-20% under the poorest tracking conditions. These results show that tracking quality and frequency are not only inputs to collision-risk estimation: they can determine when intervention becomes necessary.

Chinese Translation

低地球轨道中接近频率的增加给航天器操作员带来了越来越大的压力，不仅需要判断遭遇是否需要干预，还需要确定是否有足够的信息来决定进行机动。本研究将这一信息与行动的权衡形式化为一个信念空间规划问题，涉及可机动航天器与不可机动次要物体之间的接近。规划者将不确定的轨道状态表示为高斯信念，并使用基于机会约束的信念空间蒙特卡洛树搜索框架，在最近接触时间（TCA）之前对可能的未来跟踪更新进行推理。终端机会约束限制了在超过规定碰撞风险阈值的情况下达到TCA的概率，使规划者能够在等待有用的跟踪信息的同时，在延迟变得过于风险时进行干预。我们在NASA的接近评估风险分析数据集中评估了八个历史接近案例。通过改变次要物体的测量质量和跟踪频率，我们生成了总共96个不同的评估场景。在评估的条件下，规划者在大约40%的情境中无需机动即可达到TCA，同时保持没有终端碰撞风险违规。相比之下，基于固定时间规则的机动策略在干预延迟接近TCA时能够解决更多的遭遇，但以增加终端风险违规为代价。达到TCA而无需机动的情境比例在很大程度上依赖于跟踪质量和测量频率，从准确、频繁的测量下的76%到最差跟踪条件下的约18%-20%。这些结果表明，跟踪质量和频率不仅是碰撞风险估计的输入：它们可以决定何时需要进行干预。

View on arXiv Download PDF AI Translation

cs.RO / 16 / 2609.13458

STAGE: Diagnosing Semantic Transfer at Grounded Execution in Embodied Agents

STAGE：在具身智能体的落地执行中诊断语义迁移问题

Jin, Baosheng, Liang, Yushen, Shen, Hua

Abstract

Embodied language grounding requires more than identifying the referent of an instruction: recovered semantics must also control the action an agent exposes. We study this missing link as a semantic-action gap, where instruction semantics are recoverable but weakly expressed in native continuous actions. We introduce SAT-Bench, a fixed-observation counterfactual benchmark that holds the visual scene and agent state fixed while changing only instruction semantics. On LIBERO target-name and pixel-grounded relation swaps, target recovery reaches 100.0% and 95.8%, whereas OpenVLA action sensitivity remains only 6.8% and 7.7%. The gap persists across 1,000 additional compositional and temporal/procedural counterfactuals, with overall action sensitivity of 6.1%. Hidden-state, threshold-free, cross-policy, and rollout diagnostics further support this semantic-action transfer failure. We introduce VISA, a lightweight execution-time interface that converts recovered semantics into ALLOW, DEFER, target-consistency, and verified-selection decisions. VISA reduces invalid-instruction blind execution from 92.7% to 2.8% while preserving 94.0% of normal commands, and verified selection further improves target-consistent action exposure without updating the underlying policy. Overall, embodied language evaluation should measure semantic-action transfer, not semantic parsing alone.

Chinese Translation

具身语言落地不仅需要识别指令的指称对象，还要求恢复的语义能够控制智能体所执行的动作。我们将这一缺失环节定义为语义-动作鸿沟，即指令语义虽可恢复，但在原生连续动作中表达较弱。为此，我们提出SAT-Bench，一种固定观察条件下的反事实基准测试，保持视觉场景和智能体状态不变，仅改变指令语义。在LIBERO的目标名称和像素级关系替换测试中，目标恢复率分别达到100.0%和95.8%，而OpenVLA的动作敏感度仅为6.8%和7.7%。这一鸿沟在1000个额外的组合性及时间/程序性反事实测试中依然存在，总体动作敏感度为6.1%。隐藏状态、无阈值、跨策略及展开诊断进一步支持了语义-动作迁移失败的结论。我们提出VISA，一种轻量级执行时接口，将恢复的语义转化为ALLOW（允许）、DEFER（延迟）、目标一致性及验证选择决策。VISA将无效指令的盲目执行率从92.7%降至2.8%，同时保留94.0%的正常指令执行，且验证选择在不更新底层策略的情况下进一步提升了目标一致的动作暴露。总体而言，具身语言的评估应关注语义-动作迁移能力，而非仅仅语义解析。

View on arXiv Download PDF AI Translation

cs.RO / 17 / 2609.13516

Constraint-Grounded Reinforcement Learning for Variable Impedance Control in Contact-Rich Robotic Insertion

基于约束的强化学习在接触丰富的机器人插入中的可变阻抗控制

He, Lin, Deng, Min

Abstract

In robotic insertion under uncertain contact, the axial force limit and the appropriate controller gain vary across tasks. As a result, a single fixed gain is unlikely to remain suitable across different task conditions, making conventional impedance controllers reliant on manual retuning. To eliminate manual retuning, we propose Constraint-Grounded Reinforcement Learning (CG-RL), a variable impedance framework for online gain adaptation. Conditioned on the force limit and contact feedback, the policy outputs a residual motion, an insertion rate, and a requested gain. The controller projects this gain into the admissible range without exposing the range itself to the policy. This separation allows a single policy to operate under different force limits without retraining or manual retuning. We evaluate CG-RL on simulated oblique insertion across five training seeds. CG-RL achieves an $85.8\pm7.7\%$ (mean $\pm$ SD) success rate of insertions without violating the force limit, while keeping the applied gain within the admissible range. As a comparison, a fixed-gain baseline using the midpoint gain achieves a success rate of $50.1\%$. The policy adapts its insertion rate continuously to the specified force limit and further generalizes to more permissive force limits above the training range. In contrast, the same actor without force-limit input does not exhibit this adaptation. The applied gain is guaranteed to remain within the admissible range, while force-limit satisfaction is validated empirically rather than guaranteed formally.

Chinese Translation

在不确定接触下的机器人插入中，轴向力限制和适当的控制增益在不同任务中有所不同。因此，单一固定增益不太可能在不同任务条件下保持适用，使得传统的阻抗控制器依赖于手动重新调节。为了消除手动重新调节，我们提出了基于约束的强化学习（Constraint-Grounded Reinforcement Learning, CG-RL），这是一个用于在线增益自适应的可变阻抗框架。在力限制和接触反馈的条件下，策略输出残余运动、插入速率和请求增益。控制器将该增益投影到可接受范围内，而不将范围本身暴露给策略。这种分离使得单一策略能够在不同的力限制下运行，而无需重新训练或手动调节。我们在五个训练种子上对CG-RL进行了模拟斜插入的评估。CG-RL在不违反力限制的情况下，实现了$85.8 ext{±}7.7 ext{%}$（均值$ ext{±}$标准差）的插入成功率，同时保持施加的增益在可接受范围内。作为对比，使用中点增益的固定增益基线成功率为$50.1 ext{%}$。该策略持续适应指定的力限制的插入速率，并进一步推广到训练范围以上的更宽松力限制。相比之下，未输入力限制的同一演员没有表现出这种适应性。施加的增益被保证保持在可接受范围内，而力限制的满足是通过经验验证的，而不是形式上保证的。

View on arXiv Download PDF AI Translation

cs.RO / 18 / 2609.13545

Runtime-Incremental Transformer for Reinforcement-Learning-Based Adaptive Control

基于强化学习的自适应控制的运行时增量变换器

Cirrincione, Giansalvo, Fagiolini, Adriano

Abstract

Learning-based adaptive control of robotic manipulators with non-observable friction memory has been addressed by attention- based meta-controllers whose number of attention heads is fixed before training and is tuned by costly offline search. At long memory horizons, such fixed-capacity controllers are prone to catastrophic failures on a sizeable fraction of training seeds. The present paper introduces a runtime mechanism that grows and prunes the heads of the attention block during reinforcement learning, governed by two signals: the effective rank of the on-policy context distribution, which triggers growth when representational capacity becomes insufficient, and the per-head output magnitude, which flags redundant heads for removal. Policy continuity at growth events and a quantitative bound at prune events are established analytically. On a two- link manipulator with Stribeck friction, the proposed mechanism attains full success across all memory regimes, eliminating the long-horizon failure mode and removing the need for offline tuning of the head count.

Chinese Translation

针对具有不可观测摩擦记忆的机器人操纵器的基于学习的自适应控制，已经通过注意力机制的元控制器得到了研究，这些元控制器在训练之前固定了注意力头的数量，并通过代价高昂的离线搜索进行调整。在较长的记忆时间范围内，这种固定容量的控制器在相当一部分训练种子上容易出现灾难性失败。本文提出了一种运行时机制，在强化学习过程中动态地增长和修剪注意力块的头部，该机制由两个信号控制：在表示能力不足时触发增长的在政策上下文分布的有效秩，以及标记冗余头部以便移除的每头输出幅度。通过分析建立了增长事件的策略连续性和修剪事件的定量界限。在具有斯特里贝克摩擦的双连杆操纵器上，所提出的机制在所有记忆模式下均取得了完全成功，消除了长时间范围内的失败模式，并去除了对头部数量的离线调优的需求。

View on arXiv Download PDF AI Translation

cs.RO / 19 / 2609.13572

An MRI-Guided Robotic System to Improve Hippocampal Access for Epilepsy Interventions

一种MRI引导的机器人系统以改善癫痫干预中的海马体获取

Peters, John E., Grillo, Abby M., Mansouri, Mahshid, Esser, Daniel S., Garrow, Sarah, Kumar, Nithin S., Englot, Dario J., Neimat, Joseph S., Grissom, William A., Webster III, Robert J., Barth, Eric J.

Abstract

This paper presents an MRI-guided robotic system that improves hippocampal access by delivering a curved needle-like laser ablator through the foramen ovale, a natural opening in the base of the skull. Both the delivery path and the curved nature of the needle improve upon current clinical straight-line laser interstitial thermal therapy (LITT) as measured by hippocampal cannulation percentage. We describe the design of the robotic system which includes positioning, aiming, and curved needle deployment stages, followed by experimental results assessing cannulation percentages using MRI images in phantoms. In three curvilinear, single-insertion experiments, we achieved cannulation percentages of 93.5%, 96.5%, and 60.4%, exceeding reported clinical averages of 50-60%. Seizure control is believed by physicians to be a function of hippocampal volume treated, and our system provides a means of treating a greater volume of the hippocampus with LITT-based interventions.

Chinese Translation

本文提出了一种MRI引导的机器人系统，通过将弯曲的针状激光消融器通过卵圆孔（foramen ovale）送入，改善了对海马体的获取，这是颅底的一个自然开口。与当前的临床直线激光间质热疗（LITT）相比，该系统在送达路径和针的弯曲特性上有所改进，具体通过海马体插管百分比进行测量。我们描述了该机器人系统的设计，包括定位、瞄准和弯曲针部署阶段，并随后展示了使用MRI图像在模型中评估插管百分比的实验结果。在三次曲线单次插入实验中，我们达到了93.5%、96.5%和60.4%的插管百分比，超过了报告的临床平均值50-60%。医生认为，癫痫控制与治疗的海马体体积有关，而我们的系统提供了一种使用基于LITT的干预手段来治疗更大海马体体积的方法。

View on arXiv Download PDF AI Translation

cs.RO / 20 / 2609.13588

Differential Realizability of Static Control Allocation in Multirotors: An Impossibility under Nonredundant Full Actuation and a Pseudoinverse Obstruction under Redundant Actuation

多旋翼静态控制分配的差异可实现性：在非冗余全驱动下的不可能性与冗余驱动下的伪逆障碍

Franchi, Antonio, Mizzoni, Mirko

Abstract

Control allocation for multirotors with bidirectional propellers is commonly formulated in signed-thrust variables, where the wrench map is linear. The signed-quadratic map from physical rotor speed to thrust, however, is not a local diffeomorphism at zero speed. This work derives two distinct consequences. Under nonredundant full actuation, a single-propeller reversal removes one instantaneous task direction; hence, no global continuously differentiable exact static allocator exists over the complete task space. Under redundant actuation, the physical task map may remain regular, yet a transverse pseudoinverse zero crossing requires an unbounded rotor-speed derivative. We define differential realizability as regularity of the physical lift of an actuator-output section, derive exact and first-order validity conditions, and distinguish structural rank loss from an allocator- induced rate singularity. A local nullspace deformation repairs isolated pseudoinverse reversals, while a global fixed-orthant construction establishes existence of regular sections at the cost of persistent task-preserving internal actuation.

Chinese Translation

对于具有双向螺旋桨的多旋翼，控制分配通常在带符号的推力变量中进行公式化，其中扭矩映射是线性的。然而，物理转子速度到推力的带符号二次映射在零速度下并不是局部微分同胚。本文推导出两个不同的结果。在非冗余全驱动下，单个螺旋桨的反转会移除一个瞬时任务方向；因此，在完整任务空间内不存在全局连续可微的精确静态分配器。在冗余驱动下，物理任务映射可能保持规则，但横向伪逆零交叉需要无界的转子速度导数。我们将差异可实现性定义为驱动器输出部分的物理升力的规则性，推导出精确和一阶有效性条件，并区分结构秩损失与分配器引起的速率奇点。局部零空间变形修复孤立的伪逆反转，而全局固定象限构造则以持续的任务保持内部驱动为代价建立规则部分的存在。

View on arXiv Download PDF AI Translation

cs.RO / 21 / 2609.13606

From Vision to Harvest: Benchmarking Vision-Language Models for Multi-Arm Robotic Fruit Harvesting

从视觉到收获：多臂机器人果实采摘的视觉-语言模型基准测试

Inukollu, Vrishan, Zaman, Adyan, Kudaraya, Anvi, Lazcano, Carlos, Zhu, Yuankai, Vougioukas, Stavros, Yu, Xiaofan

Abstract

Multi-arm robotic harvesting offers a promising path to improve harvesting efficiency and reduce reliance on manual labor. However, practical deployment remains challenging because the system must generalize across diverse environments while efficiently coordinating multiple arms in a shared workspace. Existing methods often require substantial data collection in target environments or rely on simplifying assumptions that limit planning quality. In this work, we introduce the first comprehensive benchmark for evaluating pretrained Vision-Language Models (VLMs) on zero-shot multi-arm fruit harvesting planning. Our benchmark uses real-world apple and citrus orchard images and compares a VLM-based planning pipeline with a traditional perception-and-planning pipeline. The VLM pipeline directly generates harvesting sequences and waypoints for each arm, while a lightweight trajectory verifier checks for collisions. Our results show that frontier VLMs can generate effective multi-arm harvesting plans zero-shot, but a practical deployment remains limited by accurate 3D waypoint generation and collision-aware coordination. These results highlight both the promise and current limitations of pretrained VLMs for multi-arm robotic harvesting.

Chinese Translation

多臂机器人采摘为提高采摘效率和减少对人工劳动的依赖提供了有希望的途径。然而，实际部署仍然面临挑战，因为系统必须在多样化的环境中进行泛化，同时有效协调多个臂在共享工作空间中的操作。现有方法通常需要在目标环境中进行大量数据收集，或依赖于简化假设，这限制了规划质量。在本研究中，我们首次引入了一个全面的基准，用于评估预训练的视觉-语言模型（Vision-Language Models, VLMs）在零样本多臂果实采摘规划中的表现。我们的基准使用真实世界的苹果和柑橘果园图像，并将基于VLM的规划流程与传统的感知与规划流程进行了比较。VLM流程直接生成每个臂的采摘序列和航点，而轻量级轨迹验证器则检查碰撞情况。我们的结果表明，前沿的VLM能够在零样本情况下生成有效的多臂采摘计划，但实际部署仍然受到准确的3D航点生成和碰撞感知协调的限制。这些结果突显了预训练VLM在多臂机器人采摘中的潜力和当前局限性。

View on arXiv Download PDF AI Translation

cs.RO / 22 / 2609.13627

How Well Do Pseudo-Rigid Body Models Capture Real Plants? In-Field Validation of Simulated Blueberry Canes

伪刚体模型在多大程度上捕捉真实植物？蓝莓枝条的现场验证

Kolano, Hannah, VanAtter, Chelse, Yang, Wei, Grimm, Cindy, Davidson, Joseph R.

Abstract

Many labor-intensive tasks in fruit production such as pruning require physically interacting with the plant (e.g., pushing, pulling, bending limbs, etc.). Due to increasing labor shortages, there is widespread interest in the adoption of robotics in this area. When the robot must physically interact with the system, a deformable model of the plant is beneficial. Coupled, spring-loaded beams have been used in prior work to build deformable plant models in simulation for learning, planning, and control, but this approach has rarely been validated against the mechanical properties of real-world plants. In this paper, we use this rigid-body model, grounded in real material properties and beam bending theory, to simulate the deformation of blueberry canes under loading. To validate the model, we simulate the canes in MuJoCo and compare their behavior with real-world data collected from probing canes at a commercial farm with a custom testbed. We perform sensitivity analysis on several key modeling variables and show that this approach is highly sensitive to the plant diameter calculations and a priori flexural modulus estimation, which is dependent on season and blueberry variety. We also share our dataset of live blueberry cane deformation, including RGB-D images and measured forces and displacements.

Chinese Translation

在果树生产中，许多劳动密集型任务如修剪需要与植物进行物理交互（例如，推、拉、弯曲枝条等）。由于劳动力短缺日益严重，机器人在这一领域的应用受到广泛关注。当机器人必须与系统进行物理交互时，植物的可变形模型是有益的。在之前的研究中，耦合的弹簧梁被用于构建可变形植物模型，以用于学习、规划和控制，但这种方法很少与真实植物的机械特性进行验证。在本文中，我们使用这一基于真实材料特性和梁弯曲理论的刚体模型，模拟蓝莓枝条在负载下的变形。为了验证该模型，我们在MuJoCo中模拟了枝条，并将其行为与在商业农场使用自定义测试平台收集的真实数据进行比较。我们对几个关键建模变量进行了敏感性分析，结果表明该方法对植物直径计算和先验弯曲模量估计高度敏感，而后者依赖于季节和蓝莓品种。我们还分享了我们的蓝莓枝条变形的实时数据集，包括RGB-D图像以及测量的力和位移。

View on arXiv Download PDF AI Translation

cs.RO / 23 / 2609.13651

Compositional Shift Algebra: Extrapolating Mixed Robot Shifts Without Mixed Finetuning

组合转移代数：在不进行混合微调的情况下外推混合机器人转移

Hang, Jinting, Cai, Zhenhui

Abstract

Robot deployments rarely change one mechanism at a time: cameras, action interfaces, and physical dynamics often shift together. Prior adaptation recipes either finetune a new model for every mix or attempt to select which module to update. We instead learn shift operators on a modular stack z{=}E(o), a{=}g(z,u), z'{=}f(z,a) and compose them. Compositional Shift Algebra (CSA) fits single-factor observation, policy, and dynamics operators from exact-reset probes, then extrapolates held-out mixed shifts by operator composition---without mixed-shift finetuning. On ManiSkill StackCube, residual CSA matches an oracle mixed inverse on held-out mixes (success 1.0 over 10 seeds) while beating best-single / zero-shot / parameter-average baselines by { approx}67 pp. RGB-D vision-in-the-loop composition remains near oracle and far above non-compositional arms; a delay commutator stress shows ordered necessity for policy timesdelay. On a second task (PickCube), residual CSA again reaches compose 1.0 vs. 0.33 non-compositional (n{=}10), and an L1 vision controller without privileged cube/goal poses or grasp flags in the control loop retains compose 0.95 vs. 0.00. Main-track upgrades freeze PushCube (+33 pp), PegInsertion joint8 / pose7 EE (+67 pp each), and thin BC under frozen CSA (+67 pp); deeper BC and fair adapt baselines still need compose (+67 pp each), vision-localized BC needs compose (+56 pp), and delay favors ordered/few-shot deploy. We report Intervention-Gated Adaptation as a negative control.

Chinese Translation

机器人部署很少一次只更改一个机制：相机、动作接口和物理动态往往是一起变化的。之前的适应方法要么为每种混合微调一个新模型，要么尝试选择更新哪个模块。我们则在模块化堆栈 z{=}E(o), a{=}g(z,u), z'{=}f(z,a) 上学习转移算子并进行组合。组合转移代数（Compositional Shift Algebra, CSA）从精确重置探测中拟合单因子观察、策略和动态算子，然后通过算子组合外推保留的混合转移——无需混合转移微调。在 ManiSkill StackCube 上，残余 CSA 在保留的混合上匹配一个神谕混合逆（10 次实验成功率为 1.0），同时超越最佳单一/零-shot/参数平均基线约 67 个百分点。RGB-D 视觉环路中的组合接近神谕，远高于非组合臂；延迟交换子压力显示策略时间延迟的有序必要性。在第二个任务（PickCube）中，残余 CSA 再次达到组合 1.0，而非组合则为 0.33（n{=}10），且没有特权立方体/目标姿态或抓取标志的 L1 视觉控制器在控制环中保持组合 0.95，对比 0.00。主要轨道升级冻结 PushCube（+33 个百分点）、PegInsertion joint8 / pose7 EE（各 +67 个百分点），以及在冻结 CSA 下的薄 BC（+67 个百分点）；更深的 BC 和公平适应基线仍需组合（各 +67 个百分点），视觉定位的 BC 需要组合（+56 个百分点），而延迟则有利于有序/少量样本部署。我们报告干预门控适应作为负对照。

View on arXiv Download PDF AI Translation

cs.RO / 24 / 2609.13675

Does Online Gravity Estimation Matter? Revisiting a Silent Design Split in LiDAR-Inertial Odometry

在线重力估计重要吗？重新审视LiDAR-惯性里程计中的静默设计分裂

Xu, Jie, Jin, Ziyi, Yu, Kangjin, Jiang, Can, Huang, Hongjun, Jin, Tongxing, Luo, Hongkun, Xia, Zhongpu

Abstract

LiDAR-inertial odometry (LIO) systems differ in whether they continue estimating gravity after initialization. We compare four gravity-bias state configurations in each of FAST-LIO2 and LIO-SAM, then separately test a gravity-direction factor. Across 12 dataset sequences evaluated with FAST-LIO2, fixing gravity under continuous LiDAR correction produces mean paired changes in vertical and 3D position errors with 90% confidence intervals within $\pm 2\%$. Tests on 4 sequences with LIO-SAM likewise show no consistent benefit from online gravity. Multi-second LiDAR outages, unlike reduced range or field of view, reveal trajectory-dependent costs of fixing gravity. A history-matched 23D-to-21D switch places the repeatable 3D error increase after LiDAR updates resume. Under 5-s outages, a direction factor from the same IMU used for preintegration improves accuracy on Hall05 but worsens both errors with online gravity on TUHH. Dynamic-start tests also show fixed-bias failures at particular starting phases. We recommend keeping gravity and accelerometer bias online for robustness; use a direction factor only after verifying vertical and 3D accuracy gains under the intended operating conditions.

Chinese Translation

LiDAR-惯性里程计（LIO）系统在初始化后是否继续估计重力存在差异。我们比较了FAST-LIO2和LIO-SAM中四种重力偏差状态配置，然后单独测试了重力方向因子。在使用FAST-LIO2评估的12个数据集序列中，在持续的LiDAR校正下固定重力产生的垂直和三维位置误差的均值配对变化，其90%置信区间在±2%以内。对4个使用LIO-SAM的序列的测试同样未显示在线重力的持续益处。与减少范围或视场不同，多秒的LiDAR中断揭示了固定重力的轨迹依赖成本。经过历史匹配的23D到21D切换使得在LiDAR更新恢复后可重复的三维误差增加。对于5秒的中断，来自同一IMU的方向因子在Hall05上提高了准确性，但在TUHH上则使得在线重力下的两种误差都恶化。动态启动测试也显示在特定启动阶段固定偏差的失败。我们建议保持重力和加速度计偏差在线以增强鲁棒性；仅在验证了在预期操作条件下的垂直和三维准确性提升后，才使用方向因子。

View on arXiv Download PDF AI Translation

cs.RO / 25 / 2609.13679

How to Better Train VLAs: Lessons Learned From the REAL-I Challenge at ICRA 2026

如何更好地训练视觉语言行动模型：来自2026年ICRA REAL-I挑战赛的经验教训

Wang, Jiaming, Chen, Jizhuo, Liu, Diwen, Song, Wang, Wang, Qiang, Ren, Jie, Fu, Chao, Zhu, Dingkun, Ruan, Minchi, Li, Hongtong, Jiang, Yuhua, Xue, Zhiwei, Pan, Yongping, Soh, Harold

Abstract

How can robot policies learn more effectively from a fixed demonstration budget? The first Real-world Embodied AI Learning (REAL-I) Challenge at ICRA 2026 examined this question through simulation, real-robot evaluation, and an on-site final on a shared dual-arm humanoid platform. We describe the challenge tasks, data and deployment interfaces, and competition results, then compare the approaches contributed by NUS-CLEAR, RCL-Lab, and Deeptouch.ai. Their systems combined pretrained vision-language-action models and task-specific imitation policies with different strategies for data curation, staged adaptation, checkpoint selection, and action-space design. The team reports highlight the importance of adapting to the deployment environment while retaining prior capabilities, treating demonstration quality at an appropriate temporal scale, and suppressing errors in inactive robot components. They also expose the limitations of offline action-prediction metrics for forecasting closed-loop success. These observations motivate a view of fixed-data robot learning that integrates data, adaptation, evaluation, and deployment.

Chinese Translation

机器人策略如何能够更有效地从固定的演示预算中学习？2026年ICRA的首届真实世界具身人工智能学习（REAL-I）挑战赛通过仿真、真实机器人评估以及在共享双臂类人平台上的现场决赛来探讨这个问题。我们描述了挑战任务、数据和部署接口，以及比赛结果，并比较了新加坡国立大学CLEAR实验室、RCL实验室和Deeptouch.ai提供的方法。他们的系统结合了预训练的视觉-语言-行动模型和特定任务的模仿策略，并采用了不同的数据策划、分阶段适应、检查点选择和动作空间设计策略。团队报告强调了在保留先前能力的同时适应部署环境的重要性，适当地处理演示质量的时间尺度，以及抑制非活动机器人组件中的错误。他们还揭示了离线动作预测指标在预测闭环成功方面的局限性。这些观察促使我们对固定数据机器人学习的看法，整合了数据、适应、评估和部署。

View on arXiv Download PDF AI Translation

cs.RO / 26 / 2609.13695

GROOVE: Geometry-Guided Reduction of Operational-Space Jerk in VLA Execution

GROOVE：基于几何的操作空间抖动在视觉语言动作执行中的减少

Yun, Sangho, Kim, Minsoo, Cho, Minwoo, Yu, Hwanjo

Abstract

Chunked vision language action (VLA) policies execute several commands per query, but jerk within chunks and across replanning boundaries can induce oscillatory motion and sharp actuator transients. We present GROOVE, an online regulator that searches directional correction regions around the raw three dimensional end effector (EEF) path, without retraining or additional VLA inference. It optimizes the new chunk using delivered commands as boundary conditions, reducing boundary and within chunk jerk while bounding cumulative translation and local axis angle deviation from the raw plan after every command. Using quadratic programs (QPs), GROOVE generates a cube reference and thirteen directional candidates, then selects the one with the lowest command space jerk under a reference relative deviation cap. On a held out LIBERO benchmark, GROOVE achieves the largest reductions among the evaluated methods, reducing translational and rotational EEF jerk by 33.02% and 43.42%, respectively, with task success of 95.75% versus 93.75% for raw execution. Across 50 matched UR5e pairs with measured execution timing, it reduces translational and rotational tool center point (TCP) jerk by 16.39% and 19.49% and joint current slew by 29.09%.

Chinese Translation

分块视觉语言动作（VLA）策略在每个查询中执行多个命令，但在分块内和重新规划边界处的抖动可能会引起振荡运动和急剧的执行器瞬态。我们提出了GROOVE，一种在线调节器，它在原始三维末端执行器（EEF）路径周围搜索方向修正区域，而无需重新训练或额外的VLA推理。它利用已交付的命令作为边界条件来优化新的分块，减少边界和分块内的抖动，同时在每个命令后限制相对于原始计划的累积位移和局部轴角度偏差。通过使用二次规划（QPs），GROOVE生成一个立方体参考和十三个方向候选，然后选择在参考相对偏差限制下具有最低命令空间抖动的候选。在一个保留的LIBERO基准测试中，GROOVE在评估的方法中实现了最大的减少，分别将平移和旋转的EEF抖动降低了33.02%和43.42%，任务成功率为95.75%，而原始执行为93.75%。在50对匹配的UR5e中，测量了执行时间，GROOVE将平移和旋转工具中心点（TCP）抖动降低了16.39%和19.49%，关节电流变化降低了29.09%。

View on arXiv Download PDF AI Translation

cs.RO / 27 / 2609.13711

Decentralized Multi-Robot Task Allocation Under Degraded Communication: A Benchmark of Performance, Reliability, and Computation

在通信退化情况下的去中心化多机器人任务分配：性能、可靠性和计算的基准测试

Lott, James, Honary, Vahraz

Abstract

Selecting a decentralized Multi-Robot Task Allocation (MRTA) method for embedded deployment on autonomous platforms requires considering more than route performance alone. We benchmark six decentralized MRTA allocators (CBAA, ACBBA, PI, HIPC, DMCHBA, and DGA) in the Collaborative Visit (CV) scenario to characterize tradeoffs among MinMax and MinSum travel, communication robustness and demand, allocation reliability, computational burden, and scale sensitivity. The core study uses 500 paired ten-target instances across 25 ideal and degraded communication conditions spanning Bernoulli loss, Gilbert--Elliott loss, and Rayleigh fading, with additional campaigns examining pre-allocation, execution-integrated computation, and sensitivity to grid size, robot density, and target load. Across the 24 impaired core conditions, DGA and DMCHBA achieved the lowest mean MinMax travel at 24.49 and 24.78 steps, respectively. HIPC narrowly led mean MinSum travel at 66.95 steps, followed by DGA at 67.22, with both methods occupying the top two in every impaired condition. DMCHBA had the lowest publication intensity at 2.08 publications per team step. In ten-target pre-allocation, HIPC and DMCHBA remained viable and stable in every tested condition, while ACBBA, PI, and DGA lost stability or viability as communication degraded. Under ideal delivery, median full-protocol computation $\Cterm$ in the primary ten-target comparison ranged from 4.88 ms for DMCHBA to 1.346 s for DGA. Static route quality preserved DGA and DMCHBA as the leading MinMax methods, while DGA led MinSum at three of four target loads and HIPC led at 50 targets. Static and execution-integrated computation rankings diverged as task load increased. The results identify distinct allocator operating regions across route objective, communication behavior, reliability, and computational constraints.

Chinese Translation

为自主平台的嵌入式部署选择一种去中心化的多机器人任务分配（MRTA）方法，需要考虑的不仅仅是路径性能。我们在协作访问（CV）场景中对六种去中心化的MRTA分配器（CBAA、ACBBA、PI、HIPC、DMCHBA和DGA）进行了基准测试，以表征最小最大和最小总旅行、通信鲁棒性和需求、分配可靠性、计算负担以及规模敏感性之间的权衡。核心研究使用了500对十目标实例，涵盖了25种理想和退化的通信条件，包括伯努利损失、吉尔伯特-埃利奥特损失和瑞利衰落，此外还进行了额外的实验，考察了预分配、执行集成计算以及对网格大小、机器人密度和目标负载的敏感性。在24种受损的核心条件下，DGA和DMCHBA分别以24.49和24.78步达到了最低的平均最小最大旅行。HIPC在平均最小总旅行上以66.95步略微领先，DGA以67.22步紧随其后，这两种方法在每种受损条件下均占据前两名。DMCHBA的出版强度最低，为每个团队步骤2.08篇。在十目标的预分配中，HIPC和DMCHBA在每种测试条件下均保持可行和稳定，而ACBBA、PI和DGA在通信退化时失去了稳定性或可行性。在理想交付下，主要十目标比较中的中位全协议计算$ ext{Cterm}$范围从DMCHBA的4.88毫秒到DGA的1.346秒。静态路径质量使DGA和DMCHBA保持为领先的最小最大方法，而DGA在四种目标负载中的三种情况下领先于最小总旅行，HIPC在50个目标时领先。随着任务负载的增加，静态和执行集成计算的排名出现了分歧。结果识别了在路径目标、通信行为、可靠性和计算约束下不同的分配器操作区域。

View on arXiv Download PDF AI Translation

cs.RO / 28 / 2609.13761

Learning In-Hand Object Reaching to General 6D Poses

学习手中物体到达通用6D姿态

Lin, Junxiao, Wu, Tianyue, Yin, Jie, Pan, Jia, Zhang, Kaifeng, Zhi, Weiming

Abstract

In-hand manipulation allows multi-fingered dexterous hands to reconfigure grasped objects without releasing and regrasping them. This improves manipulation efficiency by reducing repeated grasp acquisition and large arm motions. However, most learning-based methods focus on reorientation, continuous rotation, or translation, whereas many tasks require joint control of object position and orientation. We formulate this capability as in-hand 6D object pose reaching: starting from an existing grasp, coordinated finger motions move the object to a palm-relative target pose. We present POISE (Palm-relative Object reaching In SE(3)), a sim-to-real reinforcement learning framework for this task. POISE combines diverse stable-grasp initialization, goal- and geometry-conditioned control, an adaptive 6D goal curriculum, and a compact reward scheme for pose reaching and grasp preservation. In simulation, diverse initialization raises held-out-grasp success from 40.1% to 51.5% and post-drop recovery from 33.8% to 72.9%; the curriculum raises full-range success from 6.2% to 59.5%. On hardware, the grasp-maintenance reward improves three-target sequence success from 20% to 80%. In real-world experiments, POISE reaches successive 6D targets without manual reset across multiple object geometries and wrist orientations, and recovers from external disturbances. To support further research in dexterous manipulation, we will release our code at https://junxiaolin.github.io/poise-website/.

Chinese Translation

手中操作允许多指灵巧手在不释放和重新抓取物体的情况下重新配置抓取的物体。这通过减少重复抓取和大幅度手臂运动来提高操作效率。然而，大多数基于学习的方法集中于重新定向、连续旋转或平移，而许多任务则需要对物体位置和方向的联合控制。我们将这种能力表述为手中6D物体姿态到达：从现有的抓取开始，协调的手指运动将物体移动到相对于手掌的目标姿态。我们提出了POISE（Palm-relative Object reaching In SE(3)），这是一个用于该任务的模拟到现实强化学习框架。POISE结合了多样的稳定抓取初始化、目标和几何条件控制、自适应的6D目标课程以及紧凑的奖励机制，以实现姿态到达和抓取保持。在模拟中，多样化的初始化将保留抓取的成功率从40.1%提高到51.5%，掉落后恢复的成功率从33.8%提高到72.9%；课程学习将全范围成功率从6.2%提高到59.5%。在硬件上，抓取保持奖励将三目标序列的成功率从20%提高到80%。在真实世界实验中，POISE在多个物体几何形状和手腕方向下，无需手动重置即可连续到达6D目标，并能够从外部干扰中恢复。为了支持灵巧操作的进一步研究，我们将在https://junxiaolin.github.io/poise-website/上发布我们的代码。

View on arXiv Download PDF AI Translation

cs.RO / 29 / 2609.13777

When Do Learned Priors Help Visual Inertial Estimation? A Controlled Study of Prior Integration, Calibration, Initialization, and Backend Consistency

学习先验何时有助于视觉惯性估计？先验融合、标定、初始化与后端一致性的受控研究

Zhang, Jinchang, Lu, Guoyu

Abstract

Learned components are increasingly integrated into geometric visual--inertial estimators to provide motion, depth, bias, uncertainty, or confidence cues. Yet it remains unclear whether gains arise from useful learned priors or from changes in the backend, calibration, initialization, temporal association, or evaluation gauge. We present a controlled framework for learning-augmented visual--inertial estimation that separates fusion gain from the incremental value of a learned prior and evaluates four evidence layers: local motion consistency, global trajectory accuracy, physical-state correctness, and numerical consistency. We instantiate the framework with a MonoViT-based monocular motion prior added as a local relative-motion factor to an unchanged VINS backend. We compare Original VINS and learned-prior VINS under matched sensor streams, timestamps, initialization, frontend/backend settings, and camera--IMU extrinsics, while probing calibration, initialization, state coupling, scale, bundle adjustment, and loop closure. On KITTI, with fixed reference extrinsics, translation APE RMSE is 31.4 m for Original VINS and 31.8 m with the learned prior. Across four recordings, the prior changes mean APE by only -0.2%, while a five-times-higher weight worsens it by 8.2%. Online extrinsic updates increase mean APE by 45.7% and 52.1%, respectively, while mean RPE changes by less than 2%. These results show that fusion performance alone cannot establish the value of learned priors. Reliable evaluation requires same-backend controls and joint analysis of prior compatibility, calibration, initialization, global drift, physical-state error, and backend consistency.

Chinese Translation

学习组件正日益被集成到几何视觉-惯性估计器中，以提供运动、深度、偏置、不确定性或置信度线索。然而，目前尚不清楚性能提升是源自有用的学习先验，还是由于后端、标定、初始化、时间关联或评估标准的变化。本文提出了一个受控框架，用于学习增强的视觉-惯性估计，该框架将融合增益与学习先验的增量价值分离开来，并评估四个证据层面：局部运动一致性、全局轨迹精度、物理状态正确性及数值一致性。我们以基于MonoViT的单目运动先验为例，将其作为局部相对运动因子添加到未更改的VINS后端中。我们在匹配的传感器数据流、时间戳、初始化、前端/后端设置及相机-IMU外参条件下，比较了原始VINS与含学习先验的VINS，并探讨了标定、初始化、状态耦合、尺度、束调整及闭环检测的影响。在KITTI数据集上，固定参考外参时，原始VINS的平移绝对位置误差（APE）均方根误差（RMSE）为31.4米，含学习先验时为31.8米。在四个录制序列中，先验仅使平均APE变化-0.2%，而权重提高五倍则使误差恶化8.2%。在线外参更新分别使平均APE增加45.7%和52.1%，而平均相对位置误差（RPE）变化不足2%。这些结果表明，仅凭融合性能无法确定学习先验的价值。可靠的评估需要相同后端的对照实验，并联合分析先验兼容性、标定、初始化、全局漂移、物理状态误差及后端一致性。

View on arXiv Download PDF AI Translation

cs.RO / 30 / 2609.13779

Force-Aware Reinforcement Learning with Hybrid Sensorless Force Estimation for Wheeled-Legged Loco-Manipulation

基于力感知的强化学习与混合无传感器力估计在轮腿式运动操控中的应用

Zeng, Xuanqi, Wang, Jiaming, Zhang, Tianlin, Zhang, Lingwei, Xu, Botian, Xia, Weipeng, Li, Zhongyu, Liu, Yun-Hui

Abstract

Force-controlled loco-manipulation requires a whole-body policy to coordinate locomotion and arm motion while regulating end-effector interaction forces. This is challenging under floating-base dynamics and changing support contacts, particularly when end-effector force/torque sensing is unavailable for control. This paper presents a force-aware reinforcement learning approach with hybrid sensorless force estimation for wheeled-legged loco-manipulation. The proposed method provides a structured estimate of the end-effector force as an explicit policy observation, enabling force-guided contact behavior without using an end-effector force/torque sensor for control. The force estimate is obtained by combining generalized momentum observation, contact-constrained wrench projection, and temporal residual learning: the model-based components extract the physically structured part of the whole-body disturbance, while the residual network compensates the remaining motion-dependent bias. The estimated force is integrated into a mode-conditioned whole-body policy with an axis-wise force/position selector, allowing free-space motion, pure force regulation, and hybrid force/position control within one controller. Simulation results demonstrate improved sensorless force estimation and force-control performance. Hardware experiments further validate the proposed controller through quantitative valve-rotation and hybrid wiping evaluations, together with force-guided door opening and zero-force human-guided motion on a real wheeled-legged platform.

Chinese Translation

力控制的运动操控需要一个全身策略来协调运动和手臂动作，同时调节末端执行器的交互力。在浮动基座动力学和变化的支撑接触下，这一任务尤其具有挑战性，特别是在无法进行末端执行器力/扭矩传感控制的情况下。本文提出了一种基于力感知的强化学习方法，结合混合无传感器力估计，应用于轮腿式运动操控。所提出的方法提供了末端执行器力的结构化估计，作为显式策略观测，使得在不使用末端执行器力/扭矩传感器的情况下能够实现力引导的接触行为。力估计通过结合广义动量观测、接触约束扭矩投影和时间残差学习获得：模型基础组件提取全身扰动的物理结构部分，而残差网络则补偿剩余的运动依赖偏差。估计的力被整合到一个模式条件的全身策略中，配备轴向力/位置选择器，允许在一个控制器内实现自由空间运动、纯力调节和混合力/位置控制。仿真结果表明，传感器无力估计和力控制性能得到了改善。硬件实验通过定量阀门旋转和混合擦拭评估进一步验证了所提出的控制器，同时在真实的轮腿式平台上实现了力引导的开门和零力人机引导运动。

View on arXiv Download PDF AI Translation

cs.RO / 31 / 2609.13812

GeomVLA: Unifying Scene, Motion, and Action in 3D

GeomVLA：统一3D场景、运动与动作

Xiong, Ziyin, Gkanatsios, Nikos, Reuss, Moritz, Fragkiadaki, Katerina

Abstract

We present GeomVLA, a Vision-Language-Action (VLA) model that unifies perception, latent scene motion prediction, and action generation within a shared robot-centric 3D coordinate frame. Our approach lifts pretrained VLM features into spatially grounded 3D scene tokens using depth and camera calibration, while retaining the semantic representations learned during VLM pretraining. We further introduce a 3D Scene Trajectory Denoiser, a task-conditioned module that learns a latent representation of how scene points are expected to move in 3D. Rather than executing the predicted trajectory as an open-loop plan, GeomVLA extracts intermediate motion tokens from the trajectory denoiser and uses them to condition a 3D flow-based action denoiser through geometry-aware attention. GeomVLA achieves state-of-the-art performance on CALVIN, competitive performance on LIBERO and RoboTwin2.0, and outperforms strong baselines in real-world manipulation settings without robot-action pretraining. Extensive ablations show that future-motion reasoning alone is insufficient: the primary gains are associated with maintaining geometric consistency among scene representation, motion prediction, and robot actions throughout the perception-to-action pipeline.

Chinese Translation

我们提出了GeomVLA，一个视觉-语言-动作（VLA）模型，它在共享的以机器人为中心的3D坐标系中统一了感知、潜在场景运动预测和动作生成。我们的方法利用深度和相机标定将预训练的视觉语言模型（VLM）特征提升为空间上固定的3D场景标记，同时保留在VLM预训练过程中学习的语义表示。我们进一步引入了3D场景轨迹去噪器，这是一个任务条件模块，学习场景点在3D中预期移动的潜在表示。GeomVLA并不是将预测的轨迹作为开放循环计划来执行，而是从轨迹去噪器中提取中间运动标记，并通过几何感知注意力将其用于条件化3D基于流的动作去噪器。GeomVLA在CALVIN上实现了最先进的性能，在LIBERO和RoboTwin2.0上表现出竞争力，并在真实世界的操作环境中超越了强基线，而无需机器人动作的预训练。大量消融实验表明，仅仅依靠未来运动推理是不够的：主要的提升与在感知到动作的整个流程中保持场景表示、运动预测和机器人动作之间的几何一致性有关。

View on arXiv Download PDF AI Translation

cs.RO / 32 / 2609.13845

LePlanner: An Iterative Amortized Controller For World Models

LePlanner：一种用于世界模型的迭代摊销控制器

Bansal, Saksham, Naphade, Om, Aggarwal, Chayan, M, Vrishin

Abstract

World models trained with joint-embedding predictive architectures learn compact, structured latent representations from physical interaction, yet planning in these latent spaces typically relies on one of two costly approaches. Search-based planners such as CEM, MPPI, and iCEM optimize action sequences through many predictor rollouts, achieving strong performance at the cost of high per-decision compute and latency. Policy-based methods amortize inference into a single forward pass but can degrade on contact-rich tasks where the demonstration distribution is multimodal. We propose LePlanner, an amortized iterative controller that learns to construct and refine latent action sequences through a frozen world-model predictor. LePlanner is trained with an arrival-and-hold objective that encourages the controller to reach the goal at the earliest feasible horizon and remain there. This addresses horizon-reset procrastination, a failure mode in which repeated receding-horizon replanning continually postpones goal arrival. An additional action-Gaussian loss keeps generated actions near the support of the offline dataset. Across navigation, contact-rich manipulation, and continuous-control environments, LePlanner matches or exceeds search-based planners while requiring an order of magnitude fewer predictor evaluations and 3-49x lower wall-clock time per decision. It achieves success rates of 98% on PushT, 100% on Reacher, 100% on TwoRooms, and 92% on the OGBench Cube task. These results show that much of the structure discovered through online search can be amortized into a lightweight iterative policy, enabling fast, horizon-aware, nonlinear physical control without online optimization.

Chinese Translation

使用联合嵌入预测架构训练的世界模型能够从物理交互中学习紧凑的、结构化的潜在表示，但在这些潜在空间中的规划通常依赖于两种成本高昂的方法。基于搜索的规划器，如 CEM、MPPI 和 iCEM，通过许多预测器的回滚来优化动作序列，虽然在性能上表现出色，但代价是每次决策需要高计算量和延迟。基于策略的方法将推理摊销为单次前向传递，但在接触丰富的任务中，演示分布呈多模态时可能会退化。我们提出了 LePlanner，一种摊销的迭代控制器，通过一个冻结的世界模型预测器学习构建和优化潜在动作序列。LePlanner 的训练目标是到达并保持（arrival-and-hold），鼓励控制器在最早可行的时间内到达目标并保持在那里。这解决了时间范围重置拖延的问题，这是一种失败模式，其中重复的递归时间范围重新规划不断推迟目标到达。额外的动作高斯损失使生成的动作保持在离线数据集的支持附近。在导航、接触丰富的操作和连续控制环境中，LePlanner 的表现与基于搜索的规划器相当或更优，同时所需的预测器评估次数减少了一个数量级，每次决策的墙钟时间降低了 3-49 倍。在 PushT 上成功率为 98%，在 Reacher 上为 100%，在 TwoRooms 上为 100%，在 OGBench Cube 任务上为 92%。这些结果表明，通过在线搜索发现的许多结构可以被摊销到一个轻量级的迭代策略中，从而实现快速、具有时间范围意识的非线性物理控制，而无需在线优化。

View on arXiv Download PDF AI Translation

cs.RO / 33 / 2609.13851

ReWeight: Leveraging Human Data for VLA Post-Training via Demonstration Retrieval and Sample Weighting

ReWeight：通过示范检索和样本加权利用人类数据进行视觉-语言-动作（VLA）后训练

Wang, Chenwei, Huang, Dianye, Ko, Match W. L., Bai, Chenjia, Jiang, Zhongliang

Abstract

Post-training vision-language-action (VLA) models for specific robots and tasks requires in-domain demonstrations, yet collecting diverse robot data is costly. Egocentric human demonstrations provide a scalable alternative, but directly mixing human and robot data can introduce cross-embodiment discrepancies and degrade policy performance. To address this challenge, we introduce ReWeight, a framework that incorporates human data into VLA post-training through demonstration-level retrieval and sample-level weighting. ReWeight learns a cross-embodiment visuomotor representation that combines visual observations with future actions to measure behavioral similarity between human and robot demonstrations. Based on optimal transport, it retrieves human demonstrations relevant to the target robot data and assigns larger weights to samples with smaller cross-embodiment discrepancies. We evaluate ReWeight using $\pi_{0.5}$ across eight simulation tasks and four real-world tasks under both clean and randomized settings. In simulation, ReWeight improves the average success rate of post-trained $\pi_{0.5}$ from 39% with only robot data and 44% with randomly mixed human-robot data to 57%. In the physical experimental setting, it achieves an average success rate of 68.8%, outperforming the baselines by 28.8% and 13.8%, respectively. Overall, ReWeight provides an effective paradigm for transforming abundant egocentric human experience into transferable supervision for robot learning. (Project webpage: https://reweight-vla.github.io/)

Chinese Translation

针对特定机器人和任务的视觉-语言-动作（VLA）模型的后训练需要领域内的示范数据，但收集多样化的机器人数据成本高昂。以自我为中心的人类示范提供了一种可扩展的替代方案，但直接混合人类和机器人数据可能会引入跨体现差异，从而降低策略性能。为了解决这一挑战，我们提出了ReWeight，一个通过示范级检索和样本级加权将人类数据纳入VLA后训练的框架。ReWeight学习了一种跨体现的视觉运动表示，将视觉观察与未来动作结合，以衡量人类和机器人示范之间的行为相似性。基于最优传输，它检索与目标机器人数据相关的人类示范，并对具有较小跨体现差异的样本分配更大的权重。我们在八个仿真任务和四个现实任务中使用$ ext{π}_{0.5}$评估ReWeight，在干净和随机化设置下进行测试。在仿真中，ReWeight将后训练的$ ext{π}_{0.5}$的平均成功率从仅使用机器人数据的39%和随机混合人类-机器人数据的44%提高到57%。在物理实验设置中，其平均成功率达到68.8%，分别比基线提高了28.8%和13.8%。总体而言，ReWeight为将丰富的自我中心人类经验转化为可转移的机器人学习监督提供了一种有效的范式。（项目网页：https://reweight-vla.github.io/）

View on arXiv Download PDF AI Translation

cs.RO / 34 / 2609.13876

Co-Speech with You: Training-Free Personalization of Robot Co-Speech Gestures

与您共同发言：无训练个性化机器人共语手势

Ding, Bosong, Ancel, Selma, Spigler, Giacomo, Kirtay, Murat

Abstract

Personal robots should adapt their co-speech gesture style to a new user without requiring model retraining. We present a training-free personalization pipeline that combines a frozen audio-conditioned diffusion prior with a gesture style encoder and lightweight conditioning adapters. The encoder is first trained to discriminate speaker identities and then jointly refined with the adapters using the diffusion objective, enabling a reusable style embedding to be extracted from approximately 10 seconds of enrollment motion through a single forward pass. To support this setting, we also release a Quest~3 capture application and a dataset of spontaneous co-speech motion from ten participants. We evaluate the system on held-out speakers using Style Recognition Accuracy (SRA) and Fr'echet Gesture Distance (FGD) to measure personalization and motion quality. Our approach improves SRA from 27.6\% for the frozen prior to 69.5\% while preserving motion quality (FGD 34.2 versus 34.8), and replacing the enrollment embedding with another person's reduces SRA to 11.4\%. The generated gestures are retargeted to a physical NAO robot, and this improvement also transfers to the robot deployment setting, where speech is synthesized using five TTS voices, retaining 67.3\% SRA.

Chinese Translation

个人机器人应能够在不需要模型重新训练的情况下，调整其共语手势风格以适应新用户。我们提出了一种无训练个性化管道，该管道结合了冻结的音频条件扩散先验、手势风格编码器和轻量级条件适配器。编码器首先被训练以区分说话者身份，然后与适配器共同使用扩散目标进行精细调整，从而能够通过一次前向传播从大约10秒的注册动作中提取可重用的风格嵌入。为了支持这一设置，我们还发布了一个Quest~3捕捉应用程序和一个来自十位参与者的自发共语动作数据集。我们使用风格识别准确率（Style Recognition Accuracy, SRA）和弗雷歇手势距离（Fr'echet Gesture Distance, FGD）在保留的说话者上评估系统，以测量个性化和动作质量。我们的方法将冻结先验的SRA从27.6\%提高到69.5\\%，同时保持动作质量（FGD从34.2降至34.8），而用另一个人的注册嵌入替换后，SRA降至11.4\\%。生成的手势被重新定向到物理NAO机器人，这一改进也转移到机器人部署设置中，其中使用五种文本转语音（TTS）声音合成语音，保持67.3\\%的SRA。

View on arXiv Download PDF AI Translation

cs.RO / 35 / 2609.13982

Rethinking the Implications of Human Feedback for Preference Learning in Human-Robot Collaboration

重新思考人类反馈对人机协作中偏好学习的影响

Zhang, Qiping, Candon, Kate, Ghose, Debasmita, Vázquez, Marynel

Abstract

In Human-Robot Interaction, the standard approach to learn a reward model that represents human preferences for robot behavior consists of three steps. First, the robot collects limited direct evidence from human feedback (e.g., positive or negative binary feedback). Then, the robot utilizes the direct evidence to derive accepted or rejected labels to feasible but unchosen actions using fixed implication rules. Finally, the robot updates the reward model with both the direct and derived evidence. Unfortunately, the fixed rule can hinder preference learning: in a user study with two collaborative simulation environments, human-provided implication labels often differed from the standard fixed rule, and using the human labels substantially improved reward learning with the Preference Learning from Implicit and Explicit Feedback (PIE) algorithm. Consequently, we propose IMPLIED, an implication modeling method that treats fixed-rule implications as an initial guide while learning to infer and revise accepted and rejected action labels over time. Across evaluations on recorded human-robot interaction trajectories and a physical robot pizza-making study, IMPLIED predicts human implications more accurately than the fixed rule approach and LLM baselines, approaching the performance of a human-label oracle. In turn, IMPLIED reduces preference-estimation error and leads to robot actions that are more often rational with respect to a combined reward (which includes the true preference reward and a task-specific reward) compared to baselines. By learning to reason about the implications of human feedback, this work enables more faithful and efficient robot behavior adaptation during human-robot collaboration.

Chinese Translation

在人机交互中，学习一个代表人类对机器人行为偏好的奖励模型的标准方法包括三个步骤。首先，机器人从人类反馈中收集有限的直接证据（例如，正面或负面的二元反馈）。然后，机器人利用这些直接证据通过固定的推理规则推导出可行但未选择的动作的接受或拒绝标签。最后，机器人使用直接证据和推导证据更新奖励模型。不幸的是，固定规则可能会阻碍偏好学习：在一个包含两个协作模拟环境的用户研究中，人类提供的推理标签往往与标准固定规则不同，而使用人类标签显著改善了使用隐性和显性反馈的偏好学习（Preference Learning from Implicit and Explicit Feedback, PIE）算法的奖励学习。因此，我们提出了IMPLIED，这是一种推理建模方法，它将固定规则的推理视为初步指导，同时学习推断和修正接受和拒绝的动作标签。在对记录的人机交互轨迹和一个物理机器人制作披萨的研究进行评估时，IMPLIED比固定规则方法和大型语言模型（LLM）基线更准确地预测人类的推理，接近人类标签神谕的表现。反过来，IMPLIED减少了偏好估计误差，并使得机器人行为在与基线相比时，更加理性地符合综合奖励（包括真实偏好奖励和任务特定奖励）。通过学习推理人类反馈的含义，本研究使得机器人在与人类协作时能够更真实和高效地适应行为。

View on arXiv Download PDF AI Translation

cs.RO / 36 / 2609.13984

What Makes an Efficient VLA? Navigating Action-Head Design, Scaling, and Latency

什么使得高效的视觉-语言-动作（VLA）模型？探索动作头设计、规模和延迟

Sun, Luoyang, Xia, Guoyang, Li, Fengfa, Ren, Lei, Cui, Xinyu, Zhang, Haifeng, Feng, Fangxiang, Zhang, Kaike, Zhan, Kun, Xie, Yan, Wang, Jun, Deng, Cheng

Abstract

Vision-Language-Action (VLA) models combine a pretrained vision encoder, a language backbone, and an action head, but their relative contribution has not been established under controlled, latency-paired conditions. We fix the backbone families (SigLIP2 and Qwen2.5) and the training pipeline, sweep action-head design and module scale, and pair each configuration with measured on-device latency. The study yields three findings. First, action-head performance is governed primarily by initialization rather than decoder architecture, loss, or inference budget: copying the last transformer layers of the language backbone into the head is the single largest lever, at no latency cost, and the only axis that helps at every module scale. Alignment also explains the other axes: flow matching and a heavier decoder pay off only while the head is misaligned and reverse once it is aligned, and extra inference passes give no measurable benefit; expressiveness appears to substitute for missing alignment. We read this as representation transfer: the aligned head keeps attending to the instruction's object nouns and stays close to the backbone in weight space rather than relearning to act from scratch. Because we reach alignment only through initialization, we offer this as the account that best organizes the measurements, not a demonstrated cause, and name the control that would settle it. Second, capacity pays only after alignment: the aligned action head is the highest-return module to scale. Third, those returns diminish sharply near the size today's $\pi$-series VLAs already use, so further growth buys little in-domain accuracy for its latency. These specify EffVLA, a compact model matching the strongest open-source VLAs on standard LIBERO, leading on most LIBERO-Plus perturbation axes at lower latency, and transferring to a real SO-ARM101 arm with the recipe unchanged.

Chinese Translation

视觉-语言-动作（VLA）模型结合了预训练的视觉编码器、语言主干和动作头，但在受控的、配对延迟的条件下，它们的相对贡献尚未确立。我们固定主干家族（SigLIP2 和 Qwen2.5）和训练流程，调整动作头设计和模块规模，并将每个配置与测量的设备延迟配对。研究得出三个发现。首先，动作头的性能主要受初始化的影响，而非解码器架构、损失或推理预算：将语言主干的最后一层变换器复制到头部是最大的杠杆，且没有延迟成本，并且是每个模块规模上唯一有帮助的轴。对齐也解释了其他轴：流匹配和更重的解码器仅在头部未对齐时才有收益，一旦对齐则反转，而额外的推理传递没有可测量的好处；表现力似乎可以替代缺失的对齐。我们将其视为表征转移：对齐的头部持续关注指令的对象名词，并在权重空间中保持接近主干，而不是从头开始重新学习如何行动。由于我们仅通过初始化达到对齐，我们将其视为最佳组织测量的解释，而非证明的原因，并命名出能够解决这一问题的控制。第二，容量仅在对齐后才有回报：对齐的动作头是扩展的最高回报模块。第三，这些回报在今天的 $ ext{π}$-系列 VLA 已经使用的规模附近急剧减少，因此进一步增长在其延迟下几乎没有带来领域内的准确性。这些结果具体化了EffVLA，一个紧凑的模型，在标准LIBERO上与最强的开源VLA匹配，在大多数LIBERO-Plus扰动轴上以更低的延迟领先，并在配方不变的情况下转移到真实的SO-ARM101臂上。

View on arXiv Download PDF AI Translation

cs.RO / 37 / 2609.14058

Mobile Multi-Robot Navigation under Runtime Uncertainty via Koopman Operator Learning and Nonlinear Model Predictive Control

基于库普曼算子学习和非线性模型预测控制的移动多机器人导航在运行时不确定性下的研究

Zhang, Xiaobin, Karydis, Konstantinos

Abstract

In this work, we developed a nonlinear model predictive control (NMPC) framework that employs learned dynamics via the Koopman Operator theory for mobile multi-robot navigation. We formulated and solved NMPC problems using a lifted bilinear Koopman-based model that accurately predicts affine input systems affected by perturbations and uncertainties. Two exemplary multi-robot navigation problems are considered: target reaching and formation control. The output of our method enables closed-loop multi-robot navigation and formation control in environments populated with obstacles, whereby the Koopman operator-based model used in the NMPC formulation addresses runtime uncertainties, namely, various degrees of random wheel slipping. We validated the effectiveness of our method for both problems via extensive numerical simulations in different environments with wheeled robots affected by different amounts of slip and without knowledge of their true dynamic models.

Chinese Translation

在本研究中，我们开发了一种非线性模型预测控制（NMPC）框架，该框架利用库普曼算子理论学习的动态模型进行移动多机器人导航。我们使用提升的双线性库普曼模型来制定和解决NMPC问题，该模型能够准确预测受扰动和不确定性影响的仿射输入系统。我们考虑了两个典型的多机器人导航问题：目标到达和编队控制。我们方法的输出使得在充满障碍物的环境中实现闭环多机器人导航和编队控制成为可能，其中在NMPC公式中使用的基于库普曼算子的模型解决了运行时的不确定性，即各种程度的随机轮滑。我们通过在不同环境中进行广泛的数值仿真，验证了我们方法在这两个问题上的有效性，这些环境中轮式机器人受到不同程度的滑移影响，并且对其真实动态模型没有先验知识。

View on arXiv Download PDF AI Translation

cs.RO / 38 / 2609.14087

Gradient-Free Neural Hamilton-Jacobi Reachability for Scalable Safety-Critical Control

无梯度神经汉密尔顿-雅可比可达性用于可扩展的安全关键控制

Feng, Zeyuan, Sahin, Ali Fuat, Thorup, Santiago, Bansal, Somil

Abstract

Hamilton-Jacobi (HJ) reachability provides a principled framework for synthesizing safety certificates and robust controllers for safety-critical robotic systems. However, applying reachability analysis to high-dimensional nonlinear systems remains challenging: classical grid-based solvers suffer from the curse of dimensionality, continuous-time neural solvers require accurate spatial value gradients, and reinforcement-learning-based approaches often suffer from weak boundary anchoring and non-stationary adversarial policy optimization. We propose a discrete-time neural reachability framework for control-disturbance-affine systems that learns backward reachable tubes (BRTs) and backward reach-avoid tubes (BRATs) through Bellman-Isaacs value propagation. Our key idea is to combine equation-driven self-supervision with structured policy learning: rather than computing explicit PDE-gradients, we exploit the bang-bang structure of optimal safety interventions to construct approximate teacher actions from gradient-free value probes, converting adversarial actor learning into supervised policy learning. To stabilize long-horizon value propagation, we leverage the learned actor to train the value function backward from the terminal boundary using a windowed temporal curriculum, where each window is used as the boundary condition for the next window. Across benchmark problems up to 80 dimensions, our method learns accurate reachability value functions while improving stability over existing learning-based solvers. We further demonstrate observation-space scalability on F1-tenth racing with over 16,000-dimensional egocentric inputs. The learned safety filter generalizes zero-shot to unseen tracks and transfers to a physical RC car, achieving real-time robust collision avoidance.

Chinese Translation

汉密尔顿-雅可比（HJ）可达性为合成安全证书和鲁棒控制器提供了一个有原则的框架，适用于安全关键的机器人系统。然而，将可达性分析应用于高维非线性系统仍然具有挑战性：经典的基于网格的求解器受到维度诅咒的影响，连续时间神经求解器需要准确的空间值梯度，而基于强化学习的方法往往面临边界锚定弱和非平稳对抗策略优化的问题。我们提出了一种离散时间神经可达性框架，适用于控制-干扰-仿射系统，通过贝尔曼-伊萨克斯值传播学习向后可达管（BRTs）和向后避达管（BRATs）。我们的关键思想是将方程驱动的自我监督与结构化策略学习相结合：我们不计算显式的偏微分方程（PDE）梯度，而是利用最优安全干预的bang-bang结构，从无梯度的值探测中构建近似教师动作，将对抗性演员学习转化为监督策略学习。为了稳定长时间值传播，我们利用学习到的演员，从终端边界向后训练值函数，使用窗口化的时间课程，其中每个窗口作为下一个窗口的边界条件。在高达80维的基准问题中，我们的方法学习到准确的可达性值函数，同时在稳定性上优于现有的基于学习的求解器。我们进一步展示了在F1-tenth赛车中观察空间的可扩展性，处理超过16,000维的自我中心输入。学习到的安全过滤器在未见轨道上实现零-shot泛化，并成功转移到物理遥控车上，实现实时鲁棒的碰撞避免。

View on arXiv Download PDF AI Translation

cs.RO / 39 / 2609.14133

Vision-Force Admittance Learning for Peg Insertion into a Movable Hole

用于插入可移动孔的视觉-力适应学习

Chen, Yuzhong, Liang, Yongqing, Xu, Yunzhi, Fang, Irving, Kidder, Chase, Wang, Hui-ping, Haque, Raihan, Zhang, Yubiao, Feng, Chen

Abstract

Precise manipulation in dynamic environments, whether induced by a mobile robot base or a target with unknown motion, remains a major challenge in robotics. Manipulation in dynamic environments introduces substantial uncertainty, which fundamentally conflicts with the tight precision requirement of precise tasks such as peg-in-the-hole. We propose a Vision-Force Admittance Learning (VFAL) framework that fuses asynchronous visual feedback with a high-frequency force-based model, using visual pose estimations as a regularization term. VFAL adapts insertion strategies online to dynamic motion while maintaining millimeter-level precision. To obtain robust, low-frequency pose information, we employ state-of-the-art vision foundation models for visual pose estimation. Additionally, we incorporate failure recovery mechanisms to enhance overall robustness. We validate our approach in real-world experiments, demonstrating high success rates and strong adaptability to various pegs and dynamic environments.

Chinese Translation

在动态环境中进行精确操作，无论是由移动机器人底座引起的，还是由具有未知运动的目标引起的，仍然是机器人技术中的一大挑战。动态环境中的操作引入了显著的不确定性，这与诸如插销入孔等精确任务的严格精度要求根本上存在冲突。我们提出了一种视觉-力适应学习（Vision-Force Admittance Learning, VFAL）框架，该框架融合了异步视觉反馈与高频基于力的模型，利用视觉位姿估计作为正则化项。VFAL 在线适应动态运动的插入策略，同时保持毫米级的精度。为了获得稳健的低频位姿信息，我们采用了最先进的视觉基础模型进行视觉位姿估计。此外，我们还结合了故障恢复机制，以增强整体的鲁棒性。我们在真实世界实验中验证了我们的方法，展示了高成功率和对各种插销及动态环境的强适应性。

View on arXiv Download PDF AI Translation

cs.RO / 40 / 2609.14146

When Faster VLA Deployment Changes Closed-Loop Behavior: Task Success-Latency Analysis of SmolVLA Across PyTorch and ONNX Variants

快速 VLA 部署如何改变闭环行为：SmolVLA 在 PyTorch 和 ONNX 变体中的任务成功-延迟分析

Islam, Rafiqul

Abstract

Vision-language-action (VLA) deployment can reduce inference latency while changing closed-loop task behavior. We evaluate HuggingFaceVLA/smolvla_libero on an RTX 2060 (6 GB) in LIBERO Spatial and Object (MuJoCo 3.3.2, LeRobot 0.6.1, seed 42), comparing PyTorch+AMP with ONNX Runtime CUDA Execution Provider (CUDA EP). The main evaluation uses 100 episodes/suite; a paired rollout uses 300 episodes/suite. PyTorch+AMP reaches 70.0%/88.0% Spatial/Object success at 1181 ms p99. Requested-FP16 and requested-INT8 ONNX reduce tether-inspect p99 to 601 ms and 532 ms, while Spatial success falls to 41.0% and 40.0% and Object remains at 89.0%. A graph audit shows those artifacts are byte-identical FP32 graphs, so the requested-INT8 row is not operator-level INT8 quantization. A static language-width ablation (16/24/32 tokens) yields Spatial success of 41.0%, 75.0%, and 71.0%; widths 24 and 32 recover much of the Spatial drop while Object success and uniform-bench latency stay approximately stable. Width-24 ONNX Spatial success is comparable to the PyTorch+AMP baseline at roughly half the latency (Wilson intervals overlap; two-proportion chi-squared p=0.53). Context width is an important contributor in this stack; it does not account for every PyTorch-vs-ONNX difference. Deployment evaluation should jointly report latency, artifact inspection, interface constraints, and closed-loop success. Code: https://github.com/rafiqul713/smolvla-libero-onnx.

Chinese Translation

视觉-语言-动作（VLA）部署可以在改变闭环任务行为的同时减少推理延迟。我们在 RTX 2060（6 GB）上评估了 HuggingFaceVLA/smolvla_libero，在 LIBERO 空间和物体（MuJoCo 3.3.2, LeRobot 0.6.1, seed 42）中，比较了 PyTorch+AMP 和 ONNX Runtime CUDA 执行提供者（CUDA EP）。主要评估使用 100 个回合/套件；配对回合使用 300 个回合/套件。PyTorch+AMP 在 1181 毫秒的 p99 下达到了 70.0%/88.0% 的空间/物体成功率。请求的 FP16 和请求的 INT8 ONNX 将 tether-inspect 的 p99 降低到 601 毫秒和 532 毫秒，而空间成功率降至 41.0% 和 40.0%，物体成功率保持在 89.0%。图形审计显示这些工件是字节相同的 FP32 图，因此请求的 INT8 行并不是操作级别的 INT8 量化。静态语言宽度消融（16/24/32 个标记）产生了 41.0%、75.0% 和 71.0% 的空间成功率；宽度为 24 和 32 时，空间成功率恢复了大部分下降，而物体成功率和均匀基准延迟保持大致稳定。宽度为 24 的 ONNX 空间成功率与 PyTorch+AMP 基线相当，但延迟大约减半（Wilson 区间重叠；两比例卡方检验 p=0.53）。上下文宽度是该堆栈中的一个重要因素；它并不能解释每个 PyTorch 与 ONNX 之间的差异。部署评估应共同报告延迟、工件检查、接口约束和闭环成功。代码： https://github.com/rafiqul713/smolvla-libero-onnx.

View on arXiv Download PDF AI Translation

cs.RO / 41 / 2609.14156

Visible Touch: Rendering Contact for Visuomotor Policies

可见触觉：为视觉运动策略渲染接触

Dogan, Metin Alp, Sun, Edward, Xu, Feng, Wu, Daniel, Peng, Allen, Hong, Dennis, Cui, Yuchen

Abstract

Integrating contact information into visuomotor policies remains an open problem. Touch is essential to robust manipulation, yet most modern policies, including pretrained vision-language-action (VLA) models, operate from vision and proprioception alone. Existing approaches to closing this gap require specialized tactile hardware, add separate tactile encoders, or commit to non-image policy backbones, all incompatible with the modern paradigm of image-conditioned policies built on pretrained 2D visual representations. Our key insight is that the bottleneck is not the contact information itself, but how it is delivered: when contact signals are exposed in the same spatial frame as the scene the policy already attends to, they become directly usable by any image-conditioned policy without architectural changes. We operationalize this insight in Visible Touch, paired with a custom low-cost magnetic contact sensor that is open-sourced and fabricated from off-the-shelf parts via a parametric CAD-to-mold pipeline. Across the LIBERO benchmark, Visible Touch improves BC-Transformer success by 15.7 percentage points on average in the 2-view setting, with similar gains in the 1-view setting; controlled comparisons show that the contact-integration strategy strongly affects how effectively tactile information is used. The pattern holds when fine-tuning pretrained VLAs: miniVLA on LIBERO gains 25 percentage points on average, and $\pi_{0.5}$ on four real-world contact-rich tasks gains 30 percentage points with our custom sensor.

Chinese Translation

将接触信息整合到视觉运动策略中仍然是一个未解决的问题。触觉对于稳健的操控至关重要，然而大多数现代策略，包括预训练的视觉-语言-动作（VLA）模型，仅依赖于视觉和本体感觉。现有的解决方案需要专用的触觉硬件、增加独立的触觉编码器，或采用与现代基于预训练二维视觉表示的图像条件策略不兼容的非图像策略骨干网络。我们的关键见解是，瓶颈并不在于接触信息本身，而在于其传递方式：当接触信号在与策略已经关注的场景相同的空间框架中暴露时，它们可以被任何图像条件策略直接使用，而无需架构上的改变。我们在可见触觉（Visible Touch）中实现了这一见解，并配备了一个定制的低成本磁性接触传感器，该传感器是开源的，并通过参数化的CAD到模具管道使用现成部件制造而成。在LIBERO基准测试中，可见触觉在2视图设置下平均提高了BC-Transformer成功率15.7个百分点，在1视图设置下也有类似的提升；控制比较显示，接触整合策略对触觉信息的有效利用有显著影响。当微调预训练的VLA时，miniVLA在LIBERO上平均提高了25个百分点，而在四个真实世界的接触丰富任务中，使用我们的定制传感器的$ ext{π}_{0.5}$提高了30个百分点。

View on arXiv Download PDF AI Translation

cs.RO / 42 / 2609.14173

GIFT: Glove-Inferred Force Transfer: Force-Aware Human-to-Robot Skill Transfer from a Wearable Sensing Glove to a Robot Hand Without Tactile Sensors

GIFT：手套推断力传递：从可穿戴传感手套到无触觉传感器的机器人手的力感知人机技能传递

Sarusi, Tzah

Abstract

Human-to-robot skill transfer from sensing gloves has so far relied on shared hardware: the same tactile glove worn by the demonstrator and the robot, or a learned alignment between two tactile sensors. We present GIFT (Glove-Inferred Force Transfer), a pipeline in which the interface between human and robot is a physical unit rather than a shared sensor: fingertip force is measured in newtons on the human side and estimated in newtons on the robot side. A wearable glove records finger flexion, calibrated fingertip force, and wrist orientation, while a head-mounted camera records the demonstration; no robot is present. At deployment, the robot estimates force from actuator-current residuals relative to a free-space baseline, through a calibrated mapping to newtons, so any position-controlled hand that reports motor current can serve as the deployment platform. The policy uses a glove-space state and predicts finger-position targets; the robot enters only through two calibrated adapters, a retargeting decoder and a force estimator. We evaluate GIFT on a cup grasp-and-hold task with two action-chunking policies trained on the same demonstrations, with fingertip-force inputs retained in one and zeroed in the other. In a 50-rollout evaluation with sample size and metrics fixed before scoring, both policies succeeded in all 25 rollouts. The median of the per-rollout hold-phase grip-force estimates was 53% lower with force inputs: 1.20 N versus 2.55 N (one-sided Mann-Whitney U, p<0.0001). In an observation ablation, a vision-only policy achieved 0/15 grasps, policies given hand-command state acquired the grasp, and the force inputs determined how hard the policy held. A force channel measured on the human hand thus transfers to a robot hand with no tactile hardware, through a retargeting map from five glove channels to seven robot actuators, with no sensor shared between the two.

Chinese Translation

迄今为止，从传感手套到机器人的人机技能传递依赖于共享硬件：演示者和机器人佩戴相同的触觉手套，或两个触觉传感器之间的学习对齐。我们提出了GIFT（手套推断力传递），这是一种管道，其中人机之间的接口是一个物理单元，而不是共享传感器：人侧的指尖力以牛顿为单位进行测量，机器人侧则进行估算。可穿戴手套记录手指的弯曲、校准的指尖力和手腕方向，而头戴式摄像头记录演示；在此过程中没有机器人存在。在部署时，机器人通过相对于自由空间基线的驱动器电流残差估算力，通过校准映射到牛顿，因此任何报告电机电流的位置控制手都可以作为部署平台。该策略使用手套空间状态并预测手指位置目标；机器人仅通过两个校准适配器进入，一个是重定向解码器，另一个是力估算器。我们在一个杯子抓握与保持任务上评估了GIFT，使用两个在相同演示上训练的动作分块策略，其中一个保留指尖力输入，另一个则将其归零。在一次50次回合的评估中，样本量和指标在评分前固定，两个策略在所有25次回合中均成功。每次回合保持阶段的握持力估算中位数在有力输入的情况下比无力输入低53%：1.20 N对比2.55 N（单侧Mann-Whitney U检验，p<0.0001）。在一次观察消融实验中，仅使用视觉的策略实现了0/15次抓握，而给定手部指令状态的策略成功抓握，且力输入决定了策略的握持力度。因此，在没有触觉硬件的情况下，人手上测量的力通道通过从五个手套通道到七个机器人驱动器的重定向映射转移到机器人手上，二者之间没有共享传感器。

View on arXiv Download PDF AI Translation

cs.RO / 43 / 2609.14183

DreamSat-Bench: Development and Initial Testing of a Testbed for AI-Based Pose Estimation from 3D Reconstruction

DreamSat-Bench：基于AI的三维重建姿态估计测试平台的开发与初步测试

Posadas-Nava, Alex, Berne, August, Lavezzi, Giovanni, Shah, Kareena, Carrasco, Alejandro, Uwumukiza, Josiane, Battaglia, Giacomo, Panicucci, Paolo, Wijayatunga, Minduli C., Rodriguez-Fernandez, Victor, Linares, Richard

Abstract

This paper presents the development and initial testing of DreamSat-Bench, a modular rendezvous and proximity operation testbed designed to benchmark AI-based relative navigation techniques. By integrating a software- and hardware-in-the-loop robotic pipeline, the platform enables a seamless transition from digital simulation to physical reality. DreamSat-Bench unifies state-of-the-art robotic learning tools such as MuJoCo, Isaac Lab, and LeRobot into a single benchmarking platform, utilizing robotic arms to trace 3D trajectories. The platform allows for extensive customization of orbital environments and lighting to evaluate the simulation-to-reality gap. We demonstrate the testbed's utility by evaluating an end-to-end vision-based navigation pipeline that pairs DreamSat, a generative AI framework for single-view 3D reconstruction, with FoundationPose for zero-shot 6-DoF tracking of unseen spacecraft. Initial testing explores mission-representative orbital segments, including fixed-point station-keeping and fly-around characterization. Through a series of parametric studies, we quantify the impacts of reconstruction latency, mesh resolution, orbital range, and illumination geometry on pose estimation accuracy. Finally, a preliminary hardware-in-the-loop campaign qualitatively validates the physical deployment of the pipeline, identifying target symmetry and accumulated tracking drift as critical factors for robust navigation. DreamSat-Bench provides a rigorous framework for maturing autonomous navigation with unprepared space assets in the absence of prior geometric models.

Chinese Translation

本文介绍了DreamSat-Bench的开发与初步测试，这是一种模块化的交会与近距离操作测试平台，旨在基准测试基于AI的相对导航技术。通过集成软件和硬件环路的机器人流程，该平台实现了从数字仿真到物理现实的无缝过渡。DreamSat-Bench将MuJoCo、Isaac Lab和LeRobot等最先进的机器人学习工具统一到一个基准测试平台中，利用机器人手臂描绘三维轨迹。该平台允许对轨道环境和照明进行广泛的定制，以评估仿真与现实之间的差距。我们通过评估一个端到端的基于视觉的导航流程展示了该测试平台的实用性，该流程将DreamSat（一个用于单视图三维重建的生成AI框架）与FoundationPose（用于未见航天器的零样本6自由度跟踪）相结合。初步测试探索了具有代表性的任务轨道段，包括固定点保持和飞行绕行特性。通过一系列参数研究，我们量化了重建延迟、网格分辨率、轨道范围和照明几何对姿态估计精度的影响。最后，初步的硬件环路实验定性验证了该流程的物理部署，识别出目标对称性和累积跟踪漂移作为稳健导航的关键因素。DreamSat-Bench为在缺乏先前几何模型的情况下，成熟自主导航与未准备的太空资产提供了一个严格的框架。

View on arXiv Download PDF AI Translation

cs.RO / 44 / 2609.14198

Novel Ex-vivo Calf Brain Model with Integrated Sub-Skull Force Sensors to Access Simulated Neurosurgical Procedures

新型体外小牛脑模型与集成的颅下力传感器以访问模拟神经外科手术

Binhammad, Hamad, Ballestero, Matheus, Babgi, Mohammed, Shaka, Seana, Hemati, Nima, Saeedi, Rothaina, Mazidi, Aiden, Giglio, Bianca, Dou, Rukun, Gueziri, Houssem-Eddine, Hooshiar, Amir, Del Maestro, Rolando F.

Abstract

Surgical tissue manipulation demands precision; however, tool-tissue manipulation force magnitudes under realistic conditions are rarely quantified. To address this gap, we proposed and validated a portable ex-vivo force-sensing platform that measures tool-tissue interaction forces across the skull-brain interface during simulated neurosurgery. The system involves fresh calf brain tissue, used as a biological surrogate for brain parenchyma, placed in a 3D-printed human skull model equipped with a 6 degree-of-freedom force/torque sensor and a real-time data acquisition system. Five validation protocols assessed the accuracy and dynamic fidelity of the platform against ground-truth measurement, static accuracy and linearity using calibrated weights (0.5-50 g), minimum detectable force, spatial consistency across different anatomical regions, effect of surgical draping, and long-duration stability. Across protocols, measured forces showed excellent agreement with reference loads (correlation R = 0.9997), with RMSE < 0.005 N and mean relative error under 2%. The platform reliably detected low-magnitude forces down to 1 g (9.8 mN), while surgical drapes introduced no meaningful signal distortion and prolonged recordings exhibited minimal drift. Overall, the proposed framework provides objective, high-fidelity force quantification for skill training and performance assessment using fresh calf brain tissue and may serve as a foundation for force-based evaluation across other surgical procedures. Future work will integrate clinically used surgical instruments to increase procedural realism and will progress toward clinical trials to evaluate usability, educational impact, and translational relevance in practice-adjacent settings.

Chinese Translation

外科组织操作要求精确性；然而，在现实条件下，工具与组织之间的操作力大小很少被量化。为了解决这一问题，我们提出并验证了一种便携式体外力传感平台，该平台在模拟神经外科手术过程中测量工具与组织在颅脑界面上的相互作用力。该系统涉及新鲜的小牛脑组织，作为脑实质的生物替代物，放置在一个配备有六自由度力/扭矩传感器和实时数据采集系统的3D打印人头模型中。五个验证协议评估了该平台的准确性和动态保真度，与真实测量、静态准确性和线性度（使用校准重量0.5-50克）、最小可检测力、不同解剖区域的空间一致性、手术覆盖物的影响以及长时间稳定性进行了比较。在各个协议中，测量的力与参考负载表现出极好的一致性（相关性R = 0.9997），均方根误差（RMSE）小于0.005 N，平均相对误差低于2%。该平台可靠地检测到低至1克（9.8毫牛顿）的低幅度力，而手术覆盖物未引入显著的信号失真，长时间记录表现出最小的漂移。总体而言，所提出的框架为使用新鲜小牛脑组织进行技能训练和表现评估提供了客观、高保真的力量化基础，并可能为其他外科手术中的基于力的评估奠定基础。未来的工作将整合临床使用的外科工具，以提高程序的真实感，并将向临床试验推进，以评估其在实践相关环境中的可用性、教育影响和转化相关性。

View on arXiv Download PDF AI Translation

cs.RO / 45 / 2609.14208

ITA-LaCAM: A Complete and Scalable TAPF Solver via Assignment-Aware Configuration-Space Search

ITA-LaCAM：一种通过赋值感知配置空间搜索的完整且可扩展的目标分配与路径规划求解器

Tang, Yimin, Zhang, Han, Chan, Shao-Hung, Kim, Junsoo, Bıyık, Erdem, Koenig, Sven, Chen, Jingkai

Abstract

Combined Target Assignment and Path Finding (TAPF) requires assigning targets for agents while simultaneously planning collision-free paths. We present ITA-LaCAM, a complete and scalable TAPF solver inspired by LaCAM and ITA-CBS. In ITA-LaCAM, each joint-configuration node carries an agent-to-target matching. When a successor is generated, ITA-LaCAM incrementally repairs the matching for the agents that moved and uses the targets to guide PIBT successor generation. This design enables adaptive reassignment without explicitly enumerating the combinatorial assignment space, while preserving LaCAM's completeness and scalability. Across 9,760 benchmark instances on eight maps with 5--200 agents, ITA-LaCAM solved 100% of the instances, compared with 95.6% for IR-TAPF configured with DBS-Hungarian. ITA-LaCAM found an initial solution faster in 84.0% of the comparisons and achieved a lower sum of costs in 65.0% of the instances solved by both methods.

Chinese Translation

组合目标分配与路径规划（TAPF）需要为代理分配目标，同时规划无碰撞路径。我们提出了ITA-LaCAM，这是一种受LaCAM和ITA-CBS启发的完整且可扩展的TAPF求解器。在ITA-LaCAM中，每个联合配置节点都携带代理与目标的匹配。当生成后继节点时，ITA-LaCAM会增量地修复移动代理的匹配，并利用目标来指导PIBT后继生成。这一设计实现了自适应重新分配，而无需显式枚举组合分配空间，同时保持了LaCAM的完整性和可扩展性。在对8张地图上5到200个代理的9,760个基准实例的测试中，ITA-LaCAM解决了100%的实例，而配置为DBS-Hungarian的IR-TAPF仅解决了95.6%的实例。ITA-LaCAM在84.0%的比较中找到初始解的速度更快，并在65.0%的两个方法都解决的实例中实现了更低的成本总和。

View on arXiv Download PDF AI Translation

cs.RO / 46 / 2609.14219

Task-Specified Active Metrological Inspection with Measurement-Steered VLA Manipulation and Deterministic Evidence Gating

任务特定的主动计量检验与测量引导的VLA操作及确定性证据门控

Chen, Zhiling, Ge, Jingzhan, Chen, Ruimin, Castanier, Matthew P., Gorsich, David, Imani, Farhad

Abstract

High-mix low-volume (HMLV) manufacturing requires inspection systems to adapt to changing parts, specifications, and work orders without repeated task-specific programming. Existing inspection automation typically assumes predefined sensing sequences, while general purpose robot agents optimize task completion rather than the completeness and validity of metrological evidence. We formulate task-specified active metrological inspection and propose From Requirements to Admissible Metrological Evidence (FRAME), a hierarchical dual-arm framework that converts an inspection instruction and structured specification into traceable conformance evidence. FRAME coordinates learned manipulation with calibrated laser profilometry: a task manager grounds and schedules requirements, active surface correspondence verifies physical-to-specification localization, and evidence memory tracks measurement provenance, admissibility, and coverage. Learned components may propose inspection targets and physical access actions, but deterministic datum-grounded measurement, admissibility checks, coverage auditing, and conformance evaluation prevent incomplete or unverified evidence from authorizing PASS. A series of physical experiments shows that FRAME achieves higher end-to-end inspection reliability, fewer false accepts, and shorter task completion time.

Chinese Translation

高混合低批量（HMLV）制造要求检验系统能够适应不断变化的零件、规格和工作订单，而无需重复进行任务特定的编程。现有的检验自动化通常假设预定义的传感序列，而通用机器人代理则优化任务完成，而非计量证据的完整性和有效性。我们提出了任务特定的主动计量检验，并提出了从需求到可接受计量证据（FRAME）这一分层双臂框架，该框架将检验指令和结构化规格转换为可追溯的符合性证据。FRAME协调学习到的操作与校准的激光轮廓测量：任务管理器确定并调度需求，主动表面对应验证物理位置与规格的匹配，证据记忆跟踪测量来源、可接受性和覆盖范围。学习到的组件可以提出检验目标和物理接入动作，但确定性数据基础测量、可接受性检查、覆盖审计和符合性评估防止不完整或未经验证的证据授权通过。系列物理实验表明，FRAME实现了更高的端到端检验可靠性、更少的误接受和更短的任务完成时间。

View on arXiv Download PDF AI Translation

cs.RO / 47 / 2609.14261

VGFM: Expressive Robot Policies via Dense Value Guidance in Flow Matching

VGFM：通过流匹配中的密集价值引导实现表达性机器人策略

Koirala, Prajwal, Campbell, Mark

Abstract

Recent robot learning paradigms increasingly rely on large offline datasets of robotic interactions to train control policies. Expressive generative models enable rich and multimodal action representations, expanding the capability of this paradigm for complex robotic control. However, policy improvement with multi-step generative actors remains challenging. In offline reinforcement learning (RL), incorporating value-based objectives along generative trajectories often introduces substantial training complexity, including backpropagation through time (BPTT), auxiliary architectures, or distillation losses. We propose Value-Guided Flow Matching (VGFM), a scalable offline RL framework that enables dense value-guided shaping within a flow-based policy while avoiding BPTT and additional algorithmic overhead. VGFM parameterizes the policy as a conditional flow-matching model in action (x-prediction) space, ensuring that each intermediate flow step produces a valid robot action that can be directly evaluated by a standard offline RL critic. This design allows value guidance to be applied at randomly sampled flow times without differentiating through the entire generative trajectory, while preserving inference-time flexibility by varying the discretization of the underlying flow ODE without retraining. Evaluated on robotic locomotion and manipulation tasks in OGBench, VGFM achieves strong performance across a wide range of tasks under rigorous evaluation protocols. With minimal hyperparameter tuning, these results demonstrate that VGFM provides a simple, scalable, and effective approach for expressive policy learning in long-horizon, goal-oriented robotic control.

Chinese Translation

近期的机器人学习范式越来越依赖于大型离线数据集中的机器人交互来训练控制策略。表达性生成模型能够实现丰富且多模态的动作表示，扩展了这一范式在复杂机器人控制中的能力。然而，使用多步生成演员进行策略改进仍然具有挑战性。在离线强化学习（RL）中，将基于价值的目标纳入生成轨迹通常会引入显著的训练复杂性，包括时间反向传播（BPTT）、辅助架构或蒸馏损失。我们提出了价值引导流匹配（Value-Guided Flow Matching，VGFM），这是一个可扩展的离线RL框架，能够在基于流的策略中实现密集的价值引导塑形，同时避免BPTT和额外的算法开销。VGFM将策略参数化为条件流匹配模型，位于动作（x-预测）空间中，确保每个中间流步骤产生一个有效的机器人动作，可以被标准的离线RL评论家直接评估。这一设计允许在随机采样的流时间上应用价值引导，而无需对整个生成轨迹进行微分，同时通过改变基础流常微分方程的离散化来保持推理时的灵活性，而无需重新训练。在OGBench上评估的机器人运动和操控任务中，VGFM在严格的评估协议下在广泛的任务中表现出色。通过最小的超参数调优，这些结果表明VGFM为长时间跨度、目标导向的机器人控制中的表达性策略学习提供了一种简单、可扩展且有效的方法。

View on arXiv Download PDF AI Translation

cs.RO / 48 / 2609.14268

Learning Communication-Conditioned Generative Policies for Decentralized Multi-Agent Collision Avoidance

学习基于通信条件的生成策略以实现去中心化多智能体避碰

Koirala, Prajwal, Campbell, Mark

Abstract

In this work, we propose a decentralized communication-conditioned generative framework for multi-agent collision avoidance. Agents generate short-horizon action sequences using a flow-matching policy trained from privileged offline demonstrations with access to global state. The demonstrations do not include explicit communication signals; instead, agents learn to exchange and aggregate latent messages that encode interaction-relevant intent under partial observability. This formulation supports flexible inference at test time, where unconditioned generation corresponds to independent behavior and communication-conditioned generation enables coordinated interaction without centralized planning. The resulting policies operate in a fully decentralized manner at execution time, relying only on local observations and learned messages. Combined with a receding-horizon inference scheme, the proposed approach enables efficient single-step inference of short-horizon action sequences and degrades gracefully under communication dropouts. Extensive simulation results demonstrate near-expert collision avoidance performance and strong generalization to denser, unseen multi-agent scenarios, along with zero-shot transfer to real-robot experiments.

Chinese Translation

在本研究中，我们提出了一种去中心化的基于通信条件的生成框架，用于多智能体避碰。智能体利用从特权离线演示中训练的流匹配策略生成短期行动序列，这些演示能够访问全局状态。演示中不包括显式的通信信号；相反，智能体学习交换和聚合潜在消息，这些消息在部分可观测性下编码与交互相关的意图。这种表述支持在测试时灵活推理，其中无条件生成对应于独立行为，而基于通信条件的生成则使得在没有集中规划的情况下实现协调交互。所得到的策略在执行时以完全去中心化的方式运行，仅依赖于本地观察和学习到的消息。结合递归视野推理方案，所提方法能够高效地进行短期行动序列的单步推理，并在通信中断的情况下优雅降级。大量仿真结果表明，该方法在避碰性能上接近专家水平，并在更密集且未见过的多智能体场景中具有良好的泛化能力，同时在真实机器人实验中实现了零样本迁移。

View on arXiv Download PDF AI Translation

cs.RO / 49 / 2609.14297

Multi-Task Visual Perception Network with LLM Conditioning for Autonomous Navigation

基于大语言模型条件的多任务视觉感知网络用于自主导航

Kumar, Praveen, Guruprasad, K. R., Sandhan, Tushar

Abstract

Long-term navigation for service robots faces crit- ical challenges like the accumulation of odometry drift and sensor error, which progressively degrade 2D maps and renders traditional path planning algorithms (e.g., A*, RRT*, DiPPer, ViT-A*) ineffective over time. To address this, we propose a user-friendly, interactive framework that eliminates the reliance on globally consistent maps. Our approach integrates visual perception with Large Language Models (LLM) to interpret user commands via text or voice. Instead of relying on a drift- prone global map, the system generates a sequential action plan based on local visual cues and egocentric geometric instructions. These action plans are executed sequentially, allowing the robot to navigate known and unknown environments safely. By reset- ting localization relative to immediate targets, our framework effectively works with a minimum accumulation drift strategy, ensuring accurate, efficient, and collision-free navigation without the maintenance overhead of traditional mapping. Experiments on real-world and simulated data have shown significant improve- ments over other methods. Our source code is publicly accessible at https://github.com/PraveenSingh24/VL-Navigation.

Chinese Translation

服务机器人在长期导航中面临诸多关键挑战，如里程计漂移和传感器误差的累积，这逐渐降低了二维地图的有效性，并使传统路径规划算法（如 A*、RRT*、DiPPer、ViT-A*）随着时间的推移变得无效。为了解决这一问题，我们提出了一种用户友好且互动的框架，消除了对全球一致性地图的依赖。我们的方法将视觉感知与大语言模型（LLM）相结合，通过文本或语音解释用户命令。系统不再依赖易受漂移影响的全球地图，而是基于局部视觉线索和自我中心几何指令生成顺序行动计划。这些行动计划按顺序执行，使机器人能够安全地在已知和未知环境中导航。通过相对于即时目标重置定位，我们的框架有效地采用了最小累积漂移策略，确保了准确、高效且无碰撞的导航，而无需传统制图的维护开销。在真实世界和模拟数据上的实验显示，与其他方法相比，显著提高了性能。我们的源代码可在 https://github.com/PraveenSingh24/VL-Navigation 上公开获取。

View on arXiv Download PDF AI Translation

cs.RO / 50 / 2609.14310

VLBiMan++: Expanding the Generalization Boundary of Vision-Language Anchored One-Shot Bimanual Manipulation

VLBiMan++：扩展视觉-语言锚定的一次性双手操作的泛化边界

Zhou, Huayi, Gao, Wei, Han, Yiyang, Jia, Kui, Huang, Hui

Abstract

Generalizable bimanual robotic manipulation requires a reusable task prior that can persist across increasingly diverse tasks, objects, scenes, embodiments, and execution conditions, thus avoiding the prohibitive cost of large-scale teleoperated demonstrations and policy retraining. In this work, we present VLBiMan++, an extended framework that expands the generalization boundary of vision-language anchored one-shot bimanual manipulation. Starting from a single human demonstration, VLBiMan++ performs task-aware decomposition to identify reusable and adaptable skill components, and employs vision-language grounded geometric adaptation to transfer these skills to novel configurations without retraining. Building on this foundation, we systematically extend generalization along five dimensions: task generalization through diverse and long-horizon skill compositions; object generalization across unseen categories, varying geometries, and more complex articulated or deformable objects; scene generalization under clutter, occlusion, and dynamic interference; embodiment generalization across heterogeneous dual-arm robotic platforms; and deployment generalization through prolonged closed-loop execution under repeated external perturbations. To support this broader scope, we further introduce object-state-aware adaptation and lightweight trajectory optimization mechanisms that accommodate changes beyond simple rigid 6-DoF pose variations while preserving reliable bimanual coordination. Extensive real-world experiments demonstrate that VLBiMan++ maintains strong task success and adaptation capability across these increasingly challenging settings. Overall, VLBiMan++ advances one-shot bimanual manipulation from demonstrating isolated transferability toward a more systematic and scalable framework for generalization across tasks, objects, scenes, embodiments, and long-term deployment conditions.

Chinese Translation

可泛化的双手机器人操作需要一个可重用的任务先验，以便在日益多样化的任务、物体、场景、体现和执行条件下持续存在，从而避免大规模遥控演示和策略再训练的高昂成本。在本研究中，我们提出了VLBiMan++，一个扩展框架，旨在扩展视觉-语言锚定的一次性双手操作的泛化边界。VLBiMan++从单个人工演示开始，执行任务感知分解，以识别可重用和可适应的技能组件，并采用视觉-语言基础的几何适应性，将这些技能转移到新配置中，而无需再训练。在此基础上，我们系统地沿五个维度扩展泛化：通过多样化和长时间跨度的技能组合实现任务泛化；跨未见类别、不同几何形状以及更复杂的关节或可变形物体实现物体泛化；在杂乱、遮挡和动态干扰下实现场景泛化；在异构双臂机器人平台上实现体现泛化；以及通过在重复外部干扰下的长期闭环执行实现部署泛化。为了支持这一更广泛的范围，我们进一步引入了物体状态感知适应和轻量级轨迹优化机制，以适应超出简单刚性6自由度姿态变化的变化，同时保持可靠的双手协调。大量现实世界的实验表明，VLBiMan++在这些日益具有挑战性的环境中保持了强大的任务成功率和适应能力。总体而言，VLBiMan++将一次性双手操作从展示孤立的可转移性推进到一个更系统和可扩展的框架，以实现任务、物体、场景、体现和长期部署条件的泛化。

View on arXiv Download PDF AI Translation

cs.RO / 51 / 2609.14343

BIG-CBF: Behavior-Imagination-Guided Control Barrier Function with Shared Uncertainty for Mobile Robot Navigation

BIG-CBF：基于行为想象的共享不确定性控制障碍函数用于移动机器人导航

Li, Shibo, Wang, Zhongcheng, Cao, Jiahe, Yang, Jianhua, Wu, Ke

Abstract

Control barrier functions (CBFs) provide a mathematically grounded framework for enforcing local collision-avoidance constraints in autonomous mobile robots, commonly through optimization-based safety filters. However, a minimum-intervention CBF filter lacks task-level maneuver awareness and may fail to select a productive avoidance direction when multiple distinct maneuvers are locally viable, leading to safe but stalled behavior in geometrically ambiguous environments. This paper presents BIG-CBF, Behavior-Imagination-Guided Control Barrier Function with shared uncertainty, a two-rate navigation architecture that separates low-rate maneuver selection from high-rate safety filtering. Over a short horizon, six closed-loop feedback behaviors are imagined and evaluated using analytic CBF compatibility together with a lightweight objective accounting for task progress, freezing, smoothness, and switching. To reduce planning-execution mismatch, the imagination and execution layers share consistent uncertainty sources for relative-motion delay, obstacle prediction, zero-order-hold motion, and command-execution residuals, while a hard CBF remains the final safety authority. In a 3,600-episode comparative benchmark across nine scenarios, BIG-CBF achieves the highest overall task success rate of 99.78% while substantially reducing downstream CBF intervention. On a physical omnidirectional robot with onboard Jetson Orin Nano computation, BIG-CBF completes all 15 evaluation runs without a recorded contact event. Matched hardware comparisons against the non-shared variant further show lower CBF intervention energy and activation frequency, supporting improved consistency between maneuver selection and safety-critical execution.

Chinese Translation

控制障碍函数（CBFs）为自主移动机器人提供了一个数学基础框架，以强制执行局部避碰约束，通常通过基于优化的安全过滤器实现。然而，最小干预的CBF过滤器缺乏任务级的机动意识，当多个不同的机动在局部可行时，可能无法选择有效的避让方向，导致在几何模糊环境中出现安全但停滞的行为。本文提出了BIG-CBF，基于行为想象的共享不确定性控制障碍函数，这是一种将低频机动选择与高频安全过滤分开的双速导航架构。在短时间范围内，想象并评估六种闭环反馈行为，利用解析CBF兼容性以及一个轻量级目标，考虑任务进展、冻结、平滑性和切换。为了减少规划与执行之间的不匹配，想象层和执行层共享相对运动延迟、障碍物预测、零阶保持运动和命令执行残差的一致不确定性来源，同时硬性CBF仍然是最终的安全权威。在九个场景的3600个对比基准测试中，BIG-CBF实现了99.78%的最高整体任务成功率，同时显著减少了下游CBF干预。在一台配备Jetson Orin Nano计算的物理全向机器人上，BIG-CBF完成了所有15次评估运行，未记录任何接触事件。与非共享变体的硬件比较进一步显示了较低的CBF干预能量和激活频率，支持机动选择与安全关键执行之间的一致性改善。

View on arXiv Download PDF AI Translation

cs.RO / 52 / 2609.14376

Orientation Control of Soft Robots via Adiabatic Spectral Submanifolds

通过绝热谱子流形控制软机器人的方向

Karakai, Aron, Kaundinya, Roshan S., Michelis, Mike Yan, Katzschmann, Robert, Haller, George

Abstract

Soft robots are commonly sought for safety-critical interactions in delicate environments, where accurate position and orientation control is imperative. Model predictive control (MPC) offers a solution, but it requires a model of the robot's infinite-dimensional nonlinear dynamics that is at once accurate and computationally cheap. Recent theory on adiabatic spectral submanifolds (aSSMs) and their applications to soft robots provide data-driven model-reduction methods to construct such models. Here, we extend these methods to identify aSSMs from enlarged observable datasets and upgrade the currently available aSSM-MPC schemes. Evaluated on a high-fidelity finite-element simulation of a pressure-actuated soft arm, our controller reduces position and orientation tracking error by more than 60% compared to existing data-driven baselines.

Chinese Translation

软机器人因其在精细环境中的安全关键交互而受到广泛关注，其中精确的位置和方向控制至关重要。模型预测控制（Model Predictive Control, MPC）提供了一种解决方案，但它需要一个既准确又计算成本低的机器人无限维非线性动态模型。最近关于绝热谱子流形（adiabatic spectral submanifolds, aSSMs）的理论及其在软机器人中的应用提供了数据驱动的模型降维方法，以构建此类模型。在此，我们扩展了这些方法，以从扩展的可观察数据集中识别aSSMs，并升级当前可用的aSSM-MPC方案。在对一个高保真有限元模拟的压力驱动软臂进行评估时，我们的控制器将位置和方向跟踪误差降低了超过60%，相比现有的数据驱动基准。

View on arXiv Download PDF AI Translation

cs.RO / 53 / 2609.14426

Learning-Based Dynamic Obstacle Avoidance for a UAV Using Only Three Range Sensors

基于学习的无人机动态障碍物避让方法，仅使用三种距离传感器

Divkoti, Mohammad Reza Ranjbar, Aguiar, A. Pedro

Abstract

We present a learning-based approach to kinodynamic online motion planning for an Unmanned Aerial Vehicle (UAV) operating at a fixed altitude in unknown dynamic environments, where real-time avoidance of both static and dynamic obstacles must be achieved under conditions of extreme partial observability. The UAV is controlled with a single degree of freedom (yaw only), resulting in constrained, nonholonomic motion similar to fixed-wing platforms. The proposed framework integrates a behavior grid map representation with Deep Reinforcement Learning (DRL), using Proximal Policy Optimization (PPO) for stable policy learning in continuous control. The key idea is the co-design of a state representation and control policy that enables reliable navigation using only three low-cost directional range sensors, without reliance on dense sensing modalities such as LiDAR or vision-based systems. The behavior grid map dynamically aggregates sparse measurements into a structured local representation that supports real-time decision-making for obstacle avoidance and target reaching. Extensive simulations across environments of varying sizes and obstacle densities demonstrate that the proposed standard and enhanced methods achieve higher success rates than PPO variants and Model Predictive Control (MPC) (94\% vs. 79--90\% in small-scale high-congestion scenarios, and 83\% vs. 62--71\% in large-scale high-congestion scenarios), while maintaining real-time performance. Real-world experiments across four scenarios further confirm practical feasibility, with consistent target-reaching behaviour and no collisions under the tested conditions.

Chinese Translation

我们提出了一种基于学习的运动规划方法，用于在未知动态环境中以固定高度操作的无人机（UAV），该方法能够在极端部分可观测条件下实时避让静态和动态障碍物。无人机仅通过一个自由度（仅偏航）进行控制，导致其运动受到约束，呈现出类似于固定翼平台的非完整运动特性。所提出的框架将行为网格地图表示与深度强化学习（Deep Reinforcement Learning, DRL）相结合，使用近端策略优化（Proximal Policy Optimization, PPO）实现连续控制中的稳定策略学习。关键思想是共同设计状态表示和控制策略，使得仅依靠三种低成本方向性距离传感器即可实现可靠导航，而无需依赖于激光雷达（LiDAR）或基于视觉的系统等密集传感方式。行为网格地图动态地将稀疏测量聚合成结构化的局部表示，支持实时决策以实现障碍物避让和目标到达。在不同规模和障碍物密度的环境中进行的广泛仿真表明，所提出的标准和增强方法在成功率上优于PPO变体和模型预测控制（Model Predictive Control, MPC）（在小规模高密度场景中为94%对79-90%，在大规模高密度场景中为83%对62-71%），同时保持实时性能。在四个场景中的实际实验进一步确认了其实际可行性，在测试条件下表现出一致的目标到达行为且没有发生碰撞。

View on arXiv Download PDF AI Translation

cs.RO / 54 / 2609.14432

EMoG: Emotion-Modulated Gait Generation for Expressive Humanoid Locomotion

EMoG：情感调制步态生成用于富有表现力的人形机器人运动

Lu, Yi, Jiang, Tianhao, Tian, Honglong, Zhang, Yumeng, Zhao, Qingrui, Wang, Zhengtao, Long, Xiao-Xiao, Shen, Qiu, Cao, Xun

Abstract

Existing humanoid locomotion systems primarily focus on stability and task execution, while integrating expressiveness with explicit locomotion control remains challenging. We propose EMoG, an emotion-modulated gait generation framework for expressive humanoid locomotion. EMoG introduces an emotional-style code with continuously adjustable intensity. Conditioned on this code and physical commands, a lightweight MLP generates expressive, command-consistent periodic gait trajectories in real time, which are tracked by a unified reinforcement learning policy for physical execution. To support training, we collect a large-scale emotion-annotated gait dataset from professional performers and develop an automated pipeline to extract physically consistent periodic gait cycles. EMoG also integrates an LLM-based parser that converts free-form language into emotional style and motion parameters for interactive control. Experiments demonstrate continuous gait-style modulation with perceptible expressive cues while maintaining command tracking. EMoG provides a practical approach to parameterized emotional-style walking for human-robot interaction.

Chinese Translation

现有的人形机器人运动系统主要关注稳定性和任务执行，而将表现力与明确的运动控制相结合仍然具有挑战性。我们提出了EMoG，一个用于富有表现力的人形机器人运动的情感调制步态生成框架。EMoG引入了一种情感风格代码，其强度可持续调整。基于该代码和物理指令，一个轻量级的多层感知器（MLP）实时生成表现力强、与指令一致的周期性步态轨迹，这些轨迹由统一的强化学习策略进行物理执行追踪。为了支持训练，我们从专业表演者那里收集了大规模的情感标注步态数据集，并开发了一个自动化管道来提取物理一致的周期性步态循环。EMoG还集成了一个基于大型语言模型（LLM）的解析器，将自由形式的语言转换为情感风格和运动参数，以实现互动控制。实验表明，在保持指令追踪的同时，步态风格可以连续调制，并且具有明显的表现力线索。EMoG为人机交互提供了一种参数化情感风格行走的实用方法。

View on arXiv Download PDF AI Translation

cs.RO / 55 / 2609.14450

EcoBoat: Design and Experimental Validation of an Autonomous Body-Board Boat For Cleaning Water Bodies

EcoBoat：一种用于清洁水体的自主体板船的设计与实验验证

Ansari, M. Aman, Khan, Saifullah, Kulkarni, Rahul, Sujit, PB

Abstract

Cleaning water bodies such as swimming pools and lakes typically demands significant manual effort or reliance on costly, sensor-intensive robotic systems. This paper presents EcoBoat, a low-cost autonomous surface vehicle built on a modified hull, designed to collect floating debris in both indoor and outdoor water bodies. For indoor environments, EcoBoat uses ultrasonic sensors to detect boundaries and obstacles, combining random-walk motion with boundary-following behavior. For outdoor environments, it relies on GPS-based geofencing paired with a random-walk strategy for area coverage. A key design feature is a Tesla-valve-inspired collection basket that allows debris intake during forward motion while preventing its escape during turning or braking. Field experiments in pools and lakes validated the design through iterative refinement, demonstrating that a minimalist sensing and computation approach can achieve effective, versatile debris collection across diverse water bodies.

Chinese Translation

清洁游泳池和湖泊等水体通常需要大量人工努力或依赖昂贵的、传感器密集型的机器人系统。本文介绍了EcoBoat，一种基于改进船体构建的低成本自主水面车辆，旨在收集室内和室外水体中的漂浮 debris。对于室内环境，EcoBoat使用超声波传感器检测边界和障碍物，结合随机行走运动与边界跟随行为。对于室外环境，它依赖基于GPS的地理围栏，并配合随机行走策略进行区域覆盖。一个关键设计特点是受特斯拉阀启发的收集篮，允许在前进运动中吸入 debris，同时在转弯或制动时防止其逃逸。通过在游泳池和湖泊中的现场实验，验证了设计的有效性，并通过迭代改进展示了简约的传感和计算方法能够在多样化的水体中实现有效且多功能的 debris 收集。

View on arXiv Download PDF AI Translation

cs.RO / 56 / 2609.14539

Fault Diagnosis for Underwater Vehicles using Moving Horizon Estimation and Gaussian Processes

基于移动视野估计和高斯过程的水下车辆故障诊断

Panetsos, Fotis, Kyriakopoulos, Kostas J.

Abstract

This work proposes a model-based fault detection and diagnosis framework for underwater vehicles subject to actuator faults that explicitly accounts for the presence of unmodeled dynamics. To this end, a Moving Horizon Estimator (MHE) is developed to estimate the lumped disturbance, capturing both unmodeled and fault effects. Gaussian Processes (GPs) are employed to approximate the unmodeled dynamics, providing predictions of the corresponding mean and uncertainty across diverse operating conditions. During online operation, the residual between the MHE lumped disturbance estimate and the GP prediction is evaluated using a Generalized Likelihood Ratio Test. By incorporating GP-based predictions within the diagnostic framework, robustness to unmodeled dynamics is achieved, enabling effective fault detection and isolation as well as accurate quantitative estimation of fault magnitude. The proposed methodology is experimentally validated in a laboratory water tank, demonstrating reliable diagnostic performance under both open-loop and closed-loop control.

Chinese Translation

本研究提出了一种基于模型的水下车辆故障检测与诊断框架，该框架针对执行器故障，明确考虑了未建模动态的存在。为此，开发了一种移动视野估计器（Moving Horizon Estimator, MHE），用于估计集中扰动，捕捉未建模效应和故障效应。采用高斯过程（Gaussian Processes, GPs）来近似未建模动态，提供在不同操作条件下相应的均值和不确定性的预测。在在线操作过程中，使用广义似然比检验（Generalized Likelihood Ratio Test）评估MHE集中扰动估计与GP预测之间的残差。通过在诊断框架中引入基于GP的预测，实现了对未建模动态的鲁棒性，从而有效地进行故障检测与隔离，并准确地定量估计故障幅度。所提出的方法在实验室水槽中进行了验证，显示出在开环和闭环控制下的可靠诊断性能。

View on arXiv Download PDF AI Translation

cs.RO / 57 / 2609.14543

NavPatch: Evidence-Guided Object-Level Costmap Correction with Vision-Language Models

NavPatch：基于证据的对象级成本图修正与视觉-语言模型

Sun, Shiji, Tao, Xingyu, Wang, Hao, Wang, Ling, Chen, Zhengyi

Abstract

Mobile robots typically rely on geometric maps for obstacle avoidance and path planning, but the resulting obstacle representation does not always match how an object should affect navigation. A low lying cable may be missed, a flexible curtain may create spurious blockage, and a traffic cone may require an exclusion region larger than its observed footprint. We present NavPatch, an object level correction layer that assigns ADD, REMOVE, or EXTEND to navigation relevant object categories through periodic scene understanding with a vision-language model. Open vocabulary grounding localizes object instances, and LiDAR and RGB-D observations provide 3D support. Observation quality filtering and cross frame maintenance determine when each correction patch is committed, replaced, or revoked. In 50 real robot trials across five layouts, NavPatch achieves an overall success rate of 86.0%. An ablation study of four configurations with 200 runs in total shows that NavPatch improves the success rate from 70.0% to 86.0% and reduces the false commit rate from 68.4% to 40.7% compared with updates based only on the current observation.

Chinese Translation

移动机器人通常依赖几何地图进行障碍物规避和路径规划，但由此产生的障碍物表示并不总是与物体对导航的影响相匹配。低悬的电缆可能会被忽略，灵活的窗帘可能会造成虚假的阻塞，而交通锥可能需要比其观察到的占地面积更大的排除区域。我们提出了NavPatch，一个对象级修正层，通过与视觉-语言模型的周期性场景理解，为与导航相关的对象类别分配添加（ADD）、移除（REMOVE）或扩展（EXTEND）。开放词汇基础定位对象实例，LiDAR和RGB-D观测提供三维支持。观测质量过滤和跨帧维护决定何时提交、替换或撤销每个修正补丁。在五个布局的50次真实机器人试验中，NavPatch的整体成功率达到了86.0%。对四种配置进行的消融研究，总共进行了200次实验，结果表明，与仅基于当前观测的更新相比，NavPatch将成功率从70.0%提高到86.0%，并将错误提交率从68.4%降低到40.7%。

View on arXiv Download PDF AI Translation

cs.RO / 58 / 2609.14558

Language-Grounded Semantic Target Navigation for Autonomous Surface Vehicles

基于语言的语义目标导航用于自主水面车辆

Lin, Yuqing, Kim, Youngroung

Abstract

Autonomous Surface Vehicles (ASVs) are increasingly expected to operate in ports and harbour environments, where operators may specify navigation targets through language-based descriptions rather than predefined coordinates or fixed target identifiers. However, existing ASV navigation methods mainly execute predefined geometric goals or task-specific objectives and give limited attention to language-grounded target specification. This study proposes Semantically Grounded Navigation (SGNav), a framework that enables an ASV to identify and approach a maritime target from an operator-provided description. SGNav integrates text-guided semantic grounding, harbour-aware candidate filtering, CLIP-based semantic verification, grounded target control-state construction, and Proximal Policy Optimisation-based closed-loop control. It grounds the target description in onboard RGB observations, suppresses visually or semantically irrelevant distractors, and converts the selected target into a compact control-oriented representation for policy execution. Experiments in simulated port environments show that SGNav achieves success rates of $97.0\pm1.2\%$, $92.0\pm1.5\%$, and $90.0\pm1.8\%$ across three representative target-reaching tasks, with semantic target accuracy above $97\%$ and wrong-target rates below $3\%$. SGNav also maintains $97.7$--$98.7\%$ success rates across held-out port layouts. In the Task~3 ablation study, removing harbour-aware filtering or semantic consistency reduces the success rate to $40.4\pm2.6\%$ and $50.4\pm3.1\%$, respectively. These findings demonstrate the importance of semantic grounding, harbour-aware filtering, and semantic verification for reliable language-grounded ASV navigation. These results indicate that the proposed perception-to-control framework can support language-grounded target approach manoeuvres of ASV under the complex port environments.

Chinese Translation

自主水面车辆（ASVs）越来越被期望在港口和海港环境中操作，在这些环境中，操作员可以通过基于语言的描述来指定导航目标，而不是使用预定义的坐标或固定目标标识符。然而，现有的ASV导航方法主要执行预定义的几何目标或特定任务目标，对基于语言的目标指定关注有限。本研究提出了语义基础导航（SGNav），这是一个框架，使ASV能够根据操作员提供的描述识别并接近海上目标。SGNav集成了文本引导的语义基础、港口感知的候选过滤、基于CLIP的语义验证、基础目标控制状态构建以及基于近端策略优化的闭环控制。它将目标描述与船载RGB观测相结合，抑制视觉或语义上不相关的干扰项，并将所选目标转换为紧凑的控制导向表示，以便执行策略。在模拟港口环境中的实验表明，SGNav在三个代表性目标到达任务中的成功率分别为$97.0 ext{±}1.2 ext{%}$、$92.0 ext{±}1.5 ext{%}$和$90.0 ext{±}1.8 ext{%}$，语义目标准确率超过$97 ext{%}$，错误目标率低于$3 ext{%}$。SGNav在保留的港口布局中也保持了$97.7$--$98.7 ext{%}$的成功率。在任务3的消融研究中，去除港口感知过滤或语义一致性分别将成功率降低至$40.4 ext{±}2.6 ext{%}$和$50.4 ext{±}3.1 ext{%}$。这些发现表明，语义基础、港口感知过滤和语义验证对于可靠的基于语言的ASV导航的重要性。这些结果表明，所提出的感知到控制框架可以支持ASV在复杂港口环境下的基于语言的目标接近机动。

View on arXiv Download PDF AI Translation

cs.RO / 59 / 2609.14561

GLAM: Training a latent world model over global spatiotemporal memory for active exploration and navigation

GLAM：基于全球时空记忆训练潜在世界模型以实现主动探索和导航

Ieong, I-Tak, Feng, Ruizhi, Lu, Zhaoyang, Cao, Yifei, Zhao, Jiayao, Li, Leon, Zhu, Senhua, Ding, Wenbo

Abstract

Active exploration and semantic navigation require an embodied agent to build memory from partial observations, predict how the evolution of observed spatial memory may support future motion, and convert that prediction into actionable plans. We present GLAM, a goal-conditioned latent world model trained over global spatiotemporal memory, and GLAM NAV, the complete navigation system built around it. Given historical map tokens, a navigation goal, and the current robot pose, GLAM jointly predicts future map representations and robot-centric waypoint latents, allowing future spatial context and navigation intent to be inferred in a shared representation space. The model follows a JEPA-like latent prediction paradigm, operates directly on map-level latent tokens rather than RGB reconstruction, and uses a pretrained waypoint encoder-decoder to supervise and decode navigation plans within GLAM NAV. Training data are collected by replaying ObjectNav expert trajectories in Habitat over HM3D v0.2 scene assets and slicing them into multi-timescale prediction samples. On a controlled HM3D-ObjectNav subset reproduction setting, GLAM NAV improves over a reproduced BSC-Nav baseline in both success rate and success weighted by path length.

Chinese Translation

主动探索和语义导航要求具身代理从部分观察中构建记忆，预测观察到的空间记忆的演变如何支持未来的运动，并将该预测转化为可执行的计划。我们提出了GLAM，一种基于全球时空记忆训练的目标条件潜在世界模型，以及围绕它构建的完整导航系统GLAM NAV。给定历史地图标记、导航目标和当前机器人姿态，GLAM联合预测未来的地图表示和以机器人为中心的航点潜在变量，从而在共享表示空间中推断未来的空间上下文和导航意图。该模型遵循类似JEPA的潜在预测范式，直接在地图级潜在标记上操作，而不是RGB重建，并使用预训练的航点编码器-解码器来监督和解码GLAM NAV中的导航计划。训练数据通过在Habitat中重放ObjectNav专家轨迹并将其切割成多时间尺度预测样本来收集。在受控的HM3D-ObjectNav子集重现设置中，GLAM NAV在成功率和按路径长度加权的成功率上均优于重现的BSC-Nav基线。

View on arXiv Download PDF AI Translation

cs.RO / 60 / 2609.14567

Learning Multi-Agent Task Assignment and Navigation in the Factory: from Simulation to Real Robots

工厂中多智能体任务分配与导航的学习：从仿真到真实机器人

Abdalwhab, Abdalwhab Bakheet Mohamed, Beltrame, Giovanni, St-Onge, David

Abstract

Reinforcement learning (RL) has shown considerable promise for robotic decision-making, yet deploying multi-agent RL (MARL) on physical multi-robot systems in industrial environments remains challenging. This paper investigates the real-world applicability of decentralized MARL for multi-robot multi-machine tending. We propose Feature-fusion Multi-Agent Proximal Policy Optimization (FMAPPO), which fuses 2D LiDAR measurements with task-specific state information to enable safe decentralized multi-robot task assignment and navigation. A complete simulation-to-reality pipeline was developed using high-fidelity robotic simulation and ROS2 and deployed on physical mobile-manipulator platforms operating under realistic real-world conditions, with the robotic arms disabled during the experiments. We further investigate the sensitivity of the learned policy to command update frequency, an important consideration for real-world deployment. Comparative evaluation in simulation demonstrated that FMAPPO significantly outperformed state-of-the-art baselines with a large effect size, achieving improvements of 106\% and 21\% in parts delivery and 48\% and 11\% in parts collection over MAPPO and SMAPPO, respectively. FMAPPO also increased machine utilization by 31 and 10 percentage points, respectively, while reducing collisions by 18\% and 15\% and increasing the safety score by 14 and 6 percentage points compared with MAPPO and SMAPPO, respectively. Furthermore, real-world experiments demonstrated that the learned decentralized policies can coordinate multiple robots to service multiple machines while maintaining safe operation under real-world sensing and control constraints. Videos of the real-world experiment are available online https://anonymouspapers123.github.io/FMAPPO/.

Chinese Translation

强化学习（RL）在机器人决策中展现出相当大的潜力，但在工业环境中将多智能体强化学习（MARL）部署到物理多机器人系统仍然面临挑战。本文研究了去中心化MARL在多机器人多机床操作中的实际应用。我们提出了特征融合多智能体近端策略优化（FMAPPO），该方法将2D激光雷达（LiDAR）测量与任务特定状态信息融合，以实现安全的去中心化多机器人任务分配与导航。我们开发了一个完整的从仿真到现实的管道，使用高保真机器人仿真和ROS2，并在真实世界条件下的物理移动操控平台上进行部署，在实验过程中机器人臂被禁用。我们进一步研究了学习到的策略对命令更新频率的敏感性，这是实际部署中的一个重要考虑因素。仿真中的比较评估表明，FMAPPO显著优于最先进的基线，效果显著，部件交付和部件收集的改进分别达到了106 ext{%}和21 ext{%}，以及48 ext{%}和11 ext{%}，相较于MAPPO和SMAPPO。FMAPPO还分别提高了机器利用率31和10个百分点，同时减少了18 ext{%}和15 ext{%}的碰撞，并与MAPPO和SMAPPO相比，安全评分提高了14和6个百分点。此外，真实世界实验表明，学习到的去中心化策略能够协调多个机器人为多个机器提供服务，同时在真实世界的感知和控制约束下保持安全操作。真实世界实验的视频可在线获取 https://anonymouspapers123.github.io/FMAPPO/.

View on arXiv Download PDF AI Translation

cs.RO / 61 / 2609.14589

Robotic Servo Tracking of Moving Targets with Dynamic Imitation Constraints

具有动态模仿约束的移动目标机器人伺服跟踪

Luo, Yazhe, Ruan, Sipu, Li, Yifei, Chen, Diansheng

Abstract

Imposing explicit trajectory constraints in robot visual servoing remains challenging. Existing tracking methods achieve fast responses by mapping visual residuals to control velocities, but they have weak constraints on the intermediate motion process, which lead to trajectory discontinuity, oscillation, or conservative behaviors. To enable constrained tracking for moving targets, this paper proposes a servo tracking method based on imitation trajectory constraints. A dynamic model describing the robot approaching a moving target is formulated and analyzed for convergence. A time-scalable deformation mechanism and a trajectory modulation incorporating shape and amplitude components are introduced to generate a series of trajectories in real time, from which tracking points are adaptively determined to form dynamic constraints. The robot velocity is then computed from target pose differentials or tracked key features to follow the constrained trajectory. Simulation and real-world experiments demonstrate that the proposed method can achieve dynamic obstacle avoidance and high-precision convergence compared with several state-of-the-art methods in complex environments.

Chinese Translation

在机器人视觉伺服中施加明确的轨迹约束仍然具有挑战性。现有的跟踪方法通过将视觉残差映射到控制速度来实现快速响应，但对中间运动过程的约束较弱，导致轨迹不连续、振荡或保守行为。为了实现对移动目标的约束跟踪，本文提出了一种基于模仿轨迹约束的伺服跟踪方法。构建并分析了描述机器人接近移动目标的动态模型，以确保其收敛性。引入了一种时间可缩放的变形机制和一种结合形状与幅度成分的轨迹调制方法，以实时生成一系列轨迹，从中自适应确定跟踪点以形成动态约束。然后，根据目标姿态差异或跟踪的关键特征计算机器人速度，以跟随约束轨迹。仿真和实际实验表明，与几种最先进的方法相比，所提出的方法能够在复杂环境中实现动态障碍物避免和高精度收敛。

View on arXiv Download PDF AI Translation

cs.RO / 62 / 2609.14633

REVOLVE: An Automated Closed-Loop Framework for Evolving Robot Manipulation with Minimal Human Intervention

REVOLVE：一种自动化闭环框架，用于在最小人类干预下演化机器人操作

Liu, Hanyu, Li, Qian, Ding, Yizhu, Wen, Jiayi, Ren, Keqiang, Ma, Yunsheng, Jian, Tao, Wang, Zhihua, Yu, Zhuofan, Li, Xinran, Song, Zhigong

Abstract

Recent advances in data-driven robot manipulation policies have substantially improved task execution and generalization. However, real-world deployment still relies heavily on humans for failure assessment, correction, and environment reset, while models often fail to continually learn from failures and corrective experience. We present REVOLVE (Robot Evolving via Orchestrated Loops, Verification, and Experience), an automated closed-loop framework for evolving robot manipulation with minimal human intervention. Built on a unified software platform, REVOLVE integrates data collection, policy training and deployment, failure recovery, and continual learning into a single closed-loop workflow. Its Automated Reset and Collection (ARC) architecture automatically resets the environment and intervenes to correct policy failures. Dual-Loop Evolution (DLE) continually improves the manipulation policy and agent by feeding real-world interaction and failure--correction data back into policy learning and using an external mismatch memory to refine agent judgments. Experiments across four real-world manipulation tasks show that, after five iterations, REVOLVE improves average policy success rate by 18.5% and agent judgment accuracy by 8.5%, while reducing human effort in data collection and deployment testing by 94.4% and 95.1%, respectively. These results demonstrate that REVOLVE transforms real-world deployment into a closed-loop learning process that continually accumulates and uses execution experience, enabling continual evolution of both the policy and supervisory model with substantially less human intervention.

Chinese Translation

近年来，数据驱动的机器人操作策略的进展显著提高了任务执行和泛化能力。然而，现实世界的部署仍然在很大程度上依赖于人类进行故障评估、纠正和环境重置，而模型往往无法持续从失败和纠正经验中学习。我们提出了REVOLVE（通过协调循环、验证和经验演化的机器人），这是一个自动化闭环框架，用于在最小人类干预下演化机器人操作。REVOLVE建立在统一的软件平台上，将数据收集、策略训练与部署、故障恢复和持续学习整合为一个单一的闭环工作流程。其自动重置和收集（ARC）架构自动重置环境并干预以纠正策略失败。双循环演化（DLE）通过将现实世界的交互和故障-纠正数据反馈到策略学习中，并使用外部不匹配记忆来细化代理判断，持续改进操作策略和代理。在四个现实世界操作任务中的实验表明，经过五次迭代，REVOLVE将平均策略成功率提高了18.5%，代理判断准确性提高了8.5%，同时分别减少了94.4%和95.1%的人类数据收集和部署测试工作量。这些结果表明，REVOLVE将现实世界的部署转变为一个闭环学习过程，持续积累和利用执行经验，使策略和监督模型的演化得以在显著减少人类干预的情况下进行。

View on arXiv Download PDF AI Translation

cs.RO / 63 / 2609.14647

Skill Composition for Legged Robot Reinforcement Learning

四足机器人强化学习中的技能组合

Gigliotti, Daniel, Maiorana, Flavio, Patrizi, Fabio, Iocchi, Luca

Abstract

Robots, and humanoid robots in particular, are increasingly competent at individual behaviors, each obtained by training a specialized controller. A specialized skill is quick to train, converges reliably because the problem it faces is narrow, and can be validated on its own, none of which is true of a single end-to-end policy asked to cover everything. What remains fragile is the transition between them. We argue that the composition of independent sub-policies deserves to be treated as a research problem in its own right, rather than as an implementation detail left to whatever mechanism happens to be at hand. Reliable composition is what turns a collection of separate skills into a repertoire that can be used, extended and shared. More fundamentally, if control can be passed between specialized policies safely, and at any moment, the choice of what the robot should do next can be delegated to a component of an entirely different nature, such as a planner, an automaton or a symbolic controller, whose behavior can be inspected in advance. The policies would then only ever have to act, and what the robot can be trusted to do would become verifiable.

Chinese Translation

机器人，尤其是类人机器人，在各个单独行为方面的能力日益增强，这些能力是通过训练专门的控制器获得的。专门技能的训练速度快，收敛性可靠，因为它所面对的问题较为狭窄，并且可以独立进行验证，而这些都不是要求覆盖所有内容的单一端到端策略所具备的特性。它们之间的过渡仍然脆弱。我们认为，独立子策略的组合应当被视为一个独立的研究问题，而不是留给随手可得的任何机制的实现细节。可靠的组合是将一组独立技能转变为可用、可扩展和可共享的技能库的关键。从更根本的角度来看，如果控制可以在专门策略之间安全地传递，并且在任何时刻，机器人下一步该做什么的选择可以委托给一种完全不同性质的组件，例如规划器、自动机或符号控制器，其行为可以事先进行检查。那么这些策略就只需执行，而机器人可以被信任执行的任务将变得可验证。

View on arXiv Download PDF AI Translation

cs.RO / 64 / 2609.14710

A Novel Robot-Assisted Learning Pedagogy for Children with ASD

一种新颖的机器人辅助学习教学法用于自闭症谱系障碍儿童

Boccanfuso, Laura, Barney, Erin, Mademtzi, Marilena, Foster, Claire, Wang, Quan, Torres, Colette, Chen, Lisa, Scassellati, Brian, Ventola, Pamela, Shic, Frederick

Abstract

Interaction paradigms used in robot-assisted autism intervention have historically employed robots as teachers, clinical assistants, or more-abled peers to promote a variety of social skills. These modalities often leverage the expertise of trained practitioners to ensure that child-robot interactions are productive or clinically grounded to yield positive therapeutic benefits for children across the autism spectrum. Yet, despite the fact that the majority of children with autism spectrum disorder (ASD) attend mainstream schools and spend 80% or more of their time in the general classroom [27], there is a paucity of research incorporating validated classroom teaching pedagogies into robot-assisted autism interventions. In this work, we introduce a novel teaching methodology for advancing social skills in school-aged children with ASD. We evaluate the effectiveness of a novel robot-assisted autism intervention which incorporates the learning-by-teaching pedagogy and explores the comparative benefits of employing a robot versus a human confederate for improved performance on a set of social skills tasks. Results show that 80% of study participants performed better in the robot condition (mean performance in the robot condition=63%, mean performance in the confederate condition=37%), irrespective of the scenario order. Further, 90% of all participants were significantly more engaged in the robot condition (mean engagement: robot=61%, confederate=32%) and, while the effect did not result in the confederate condition, analyses indicate that overall engagement in the robot condition contributed to improved performance. These results suggest that robots employed in a learning-by-teaching context may help enhance engagement and improve performance on a simple social skills task for children with ASD.

Chinese Translation

在机器人辅助自闭症干预中使用的互动范式历来将机器人作为教师、临床助手或更具能力的同伴，以促进各种社交技能。这些模式通常利用经过培训的从业者的专业知识，以确保儿童与机器人的互动是富有成效的或具有临床基础的，从而为自闭症谱系的儿童带来积极的治疗效果。然而，尽管大多数自闭症谱系障碍（ASD）儿童就读于主流学校，并且在普通课堂中花费80%或更多的时间[27]，但将经过验证的课堂教学法纳入机器人辅助自闭症干预的研究仍然稀缺。在本研究中，我们介绍了一种新颖的教学方法，以促进学龄儿童的社交技能。我们评估了一种新型的机器人辅助自闭症干预的有效性，该干预结合了通过教学学习的教学法，并探讨了使用机器人与人类同伴在一组社交技能任务中提高表现的比较优势。结果显示，80%的研究参与者在机器人条件下表现更好（机器人条件的平均表现=63%，同伴条件的平均表现=37%），无论场景顺序如何。此外，90%的参与者在机器人条件下的参与度显著更高（平均参与度：机器人=61%，同伴=32%），虽然这一效果在同伴条件下并未显现，但分析表明，机器人条件下的整体参与度有助于提高表现。这些结果表明，在通过教学学习的背景下使用的机器人可能有助于增强参与度并改善自闭症谱系障碍儿童在简单社交技能任务上的表现。

View on arXiv Download PDF AI Translation

cs.RO / 65 / 2609.14748

Beyond Dead Reckoning: A Point of View on Camera--DAS--GNSS Continuity in Road Tunnels

超越死算：关于道路隧道中相机--DAS--GNSS 连续性的观点

Saoud, Lyes Saad, Rakha, Hesham A., Jaber, Mona, Ayyash, Moussa

Abstract

Road tunnels remove satellite visibility where connected and automated vehicles still require continuous, attributable, and integrity-bounded positioning. This Point of View argues that tunnel localization should be treated as infrastructure-assisted cross-modal track continuity, not as extrapolation from the last trusted satellite fix. A trusted portal satellite solution provides the global anchor, distributed acoustic sensing (DAS) continuous motion evidence, cameras sparse identity and lane anchors, onboard sensing short-term dynamics, and edge computing association, fusion, integrity monitoring, and guarded reacquisition. Ground truth is reserved for offline calibration and validation. The article develops a falsifiable research and deployment agenda for progressing from synchronized multimodal evidence to validated, integrity-aware tunnel positioning continuity.

Chinese Translation

道路隧道消除了卫星可见性，而连接和自动化车辆仍然需要持续、可归因且具有完整性的定位。本文观点认为，隧道定位应被视为基础设施辅助的跨模态轨迹连续性，而不是从最后一个可信卫星定位的外推。可信的门户卫星解决方案提供全球锚定，分布式声学传感（DAS）提供连续运动证据，摄像头提供稀疏身份和车道锚定，车载传感器提供短期动态，边缘计算则负责关联、融合、完整性监测和受保护的重新获取。地面真相保留用于离线校准和验证。本文制定了一个可证伪的研究和部署议程，以推动从同步的多模态证据到经过验证的、具有完整性意识的隧道定位连续性的进展。

View on arXiv Download PDF AI Translation

cs.RO / 66 / 2609.14756

An Adaptive Fixed-Time Line-of-Sight Guidance Scheme for 3D Path Following of Underwater Vehicles: Theory and Experiment

一种自适应固定时间视线引导方案用于水下车辆的三维路径跟踪：理论与实验

Yang, Hanzhi, Chavez-Galaviz, Jalil, Mahmoudian, Nina

Abstract

Reliable path tracking is crucial for autonomous underwater vehicles (AUVs) operating in dynamic and uncertain marine environments. However, traditional line-of-sight (LOS) guidance methods rely on asymptotic convergence, resulting in slow disturbance recovery and unpredictable tracking performance. Existing robust control methods typically require modifications to the underlying vehicle controller, limiting their practical application on commercial AUV platforms. This paper proposes a robust fixed-time adaptive LOS guidance framework for 3D path tracking for AUVs. By combining fixed-time stability theory with LOS guidance, this method guarantees path tracking convergence within a preset time range, with the convergence time independent of initial conditions. Furthermore, a fixed-time adaptive estimator is developed to rapidly compensate for time-varying sideslip disturbances caused by ocean currents. A time-varying look-ahead mechanism is also introduced to improve tracking performance on curved paths. Lyapunov analysis proves the fixed-time stability of the proposed framework, and numerical simulations and physical experiments demonstrate that, compared to state-of-the-art adaptive LOS methods, this framework exhibits superior tracking accuracy, convergence speed, and anti-interference capability. In simulation, the time-varying look-ahead variant reduced cross-track and vertical-track RMSE by 69.37\% and 67.46\%, respectively, during curved-path tracking. In field experiments with an Iver 3 AUV, the proposed fixed-time guidance reduced average tracking error by 56.35\% in straight-path evaluation and 27.59\% in curved-path evaluation compared with conventional adaptive LOS guidance.The proposed method provides a practical guidance-level solution for achieving reliable autonomous navigation of AUVs in complex marine environments.

Chinese Translation

可靠的路径跟踪对于在动态和不确定海洋环境中操作的自主水下车辆（AUV）至关重要。然而，传统的视线（LOS）引导方法依赖于渐近收敛，导致干扰恢复缓慢和不可预测的跟踪性能。现有的鲁棒控制方法通常需要对基础车辆控制器进行修改，限制了它们在商业AUV平台上的实际应用。本文提出了一种用于AUV三维路径跟踪的鲁棒固定时间自适应LOS引导框架。通过将固定时间稳定性理论与LOS引导相结合，该方法保证在预设时间范围内实现路径跟踪收敛，且收敛时间与初始条件无关。此外，开发了一种固定时间自适应估计器，以快速补偿由海流引起的时变侧滑干扰。还引入了一种时变前视机制，以改善曲线路径上的跟踪性能。Lyapunov分析证明了所提框架的固定时间稳定性，数值仿真和物理实验表明，与最先进的自适应LOS方法相比，该框架在跟踪精度、收敛速度和抗干扰能力方面表现出优越性。在仿真中，时变前视变体在曲线路径跟踪中将横向和纵向跟踪均方根误差（RMSE）分别降低了69.37%和67.46%。在与Iver 3 AUV的现场实验中，所提固定时间引导在直线路径评估中将平均跟踪误差降低了56.35%，在曲线路径评估中降低了27.59%，相比于传统的自适应LOS引导。所提方法为在复杂海洋环境中实现可靠的自主导航提供了一种实用的引导级解决方案。

View on arXiv Download PDF AI Translation

cs.RO / 67 / 2609.14765

A Personalized Dynamic Balance Evaluation Paradigm for Hip Exoskeleton-Assisted Walking under Unexpected Ground Perturbations

一种个性化动态平衡评估范式：针对意外地面扰动下的髋部外骨骼辅助步态

Chen, Yun, Akinniyi, Oluwasegun T., Zhang, Qiang

Abstract

Hip exoskeletons may improve recovery from unexpected gait perturbations, yet personalizing assistance remains difficult because balance is multidimensional and human-in-the-loop experiments are small-sample and noisy. We present a participant-specific composite balance cost that integrates seven biomechanical sub-metrics spanning margin of stability, center-of-mass dynamics, and whole-body angular momentum. The sub-metrics are converted to direction-aligned, dimensionless cost features, and nonnegative fusion weights are learned on the simplex. Coupled with an empirical-Bayes hierarchical model, the learned-composite selector estimates each tested condition's posterior probability of being best, P(best), and a high-probability candidate set with size $K_{0.8}$. The framework was evaluated with three participants walking at 1.1 m/s during unilateral belt-slip perturbations across 46 hip-assistance conditions. In the full-budget analysis (B = 4 repeats per condition), the selector concentrated 80% of the posterior probability within 1 to 5 of 46 conditions, compared with 2 to 12 for equal-weight fusion and 4 to 37 for principal component analysis fusion. This smaller candidate set could shorten personalization experiments and limit participants' exposure to repeated perturbations in future studies. Selected-condition trials showed lower observed composite costs than no-torque trials, with nominal p < 0.05 for P2 and P3. Leave-one-repeat-out refits yielded positive mean held-out rank correlations for all participants and moderate stability of the learned weights and candidate sets. These proof-of-concept results support participant-specific composite balance evaluation for candidate selection in perturbation-based human-in-the-loop experiments.

Chinese Translation

髋部外骨骼可能改善对意外步态扰动的恢复，但个性化辅助仍然困难，因为平衡是多维的，而人机交互实验的样本量小且噪声较大。我们提出了一种参与者特定的复合平衡成本，该成本整合了七个生物力学子指标，涵盖了稳定边际、质心动态和全身角动量。这些子指标被转换为方向对齐的无量纲成本特征，并在单纯形上学习非负融合权重。结合经验贝叶斯层次模型，学习到的复合选择器估计每个测试条件为最佳的后验概率 P(best)，以及一个大小为 $K_{0.8}$ 的高概率候选集。该框架在三名参与者以 1.1 m/s 的速度行走时进行了评估，实验中涉及 46 种髋部辅助条件的单侧带滑扰动。在全预算分析中（B = 每个条件 4 次重复），选择器将 80% 的后验概率集中在 46 种条件中的 1 到 5 条件内，而对于等权重融合则为 2 到 12，主成分分析融合则为 4 到 37。这个较小的候选集可以缩短个性化实验的时间，并限制参与者在未来研究中对重复扰动的暴露。选择条件的试验显示出比无扭矩试验更低的观察复合成本，P2 和 P3 的名义 p < 0.05。留一重复外重拟合对所有参与者产生了正的平均保持排名相关性，并且学习到的权重和候选集具有中等稳定性。这些概念验证结果支持在基于扰动的人机交互实验中进行参与者特定的复合平衡评估以供候选选择。

View on arXiv Download PDF AI Translation

cs.RO / 68 / 2609.14783

Language-Guided Representation Learning for Robust Cross-Sensor Material Recognition

基于语言引导的鲁棒跨传感器材料识别表示学习

Mohsan, Mashood M., Din, Muhayy Ud, Xu, Binzhao, Abubakar, Ahmad, Hussain, Irfan

Abstract

Robots need touch to manipulate objects safely and reliably, as many properties, such as softness, texture, and contact stability, are hard to infer from vision alone. However, vision-based tactile sensors yield different observations of the same material due to variations in optics, elastomer properties, and illumination, leading to poor generalization when trained on a single or multiple sensors. We propose a language-guided distillation framework for learning sensor-robust tactile representations. Language encodes high-level semantic properties of touch (e.g., rough, soft, slippery) that remain invariant across sensing hardware, providing a natural sensor-agnostic supervisory signal. We construct a 39K-sample touch-language dataset with human-annotated material labels and train a tactile encoder to align sensor-specific tactile images with language embeddings in a shared semantic space. We evaluate our approach for few-shot learning and cross-sensor transfer and benchmark it on six existing tactile datasets. Our method achieves 95% accuracy in the 100-shot setting, improves cross-sensor transfer by an average of 13.3% accuracy, and yields up to 19% accuracy gains across six existing tactile datasets. These results demonstrate that language-guided distillation enables scalable and hardware-agnostic tactile representation learning. Code and dataset are available at https://mashood3624.github.io/Language_Tactile/

Chinese Translation

机器人需要触觉来安全可靠地操作物体，因为许多属性，如柔软度、纹理和接触稳定性，仅通过视觉很难推断。然而，基于视觉的触觉传感器由于光学、弹性体特性和照明的变化，对同一材料产生不同的观察结果，这导致在单一或多个传感器上训练时泛化能力较差。我们提出了一种语言引导的蒸馏框架，用于学习传感器鲁棒的触觉表示。语言编码了触觉的高层语义属性（例如，粗糙、柔软、滑腻），这些属性在不同的传感硬件之间保持不变，提供了一种自然的传感器无关的监督信号。我们构建了一个包含39K样本的触觉-语言数据集，配有人工标注的材料标签，并训练了一个触觉编码器，以在共享的语义空间中对齐特定传感器的触觉图像与语言嵌入。我们评估了该方法在少样本学习和跨传感器迁移中的表现，并在六个现有的触觉数据集上进行了基准测试。我们的方法在100-shot设置中达到了95%的准确率，跨传感器迁移平均提高了13.3%的准确率，并在六个现有触觉数据集上实现了最高19%的准确率提升。这些结果表明，语言引导的蒸馏使得可扩展和硬件无关的触觉表示学习成为可能。代码和数据集可在 https://mashood3624.github.io/Language_Tactile/ 获取。

View on arXiv Download PDF AI Translation

cs.RO / 69 / 2609.14806

Belief-Adaptive Online Autonomy for Quadrotor UAV Navigation under GNSS Degradation in Urban Environments

在城市环境中应对GNSS退化的四旋翼无人机导航的信念自适应在线自主性

Panda, Deepak Kumar, Guo, Weisi

Abstract

Reliable online autonomy is critical for quadrotor operation in urban airspaces, where global navigation satellite systems (GNSS) measurements suffer from multipath, blockage, and latency issues, introducing non-stationary, temporally correlated errors that degrade conventional GNSS-IMU fusion. This paper presents a belief-adaptive online autonomy framework that augments an extended Kalman filter (EKF) with explicit GNSS trust modelling, second-order online belief adaptation, and latency-aware out-of-sequence measurement handling. GNSS trust is represented as a latent belief state that modulates measurement weighting and multipath bias uncertainty, and is updated online using EKF consistency signals. Unlike reactive covariance tuning, the proposed approach enables proactive and stable sensor trust adaptation without prior environmental knowledge or offline training. Evaluation in simulated urban air mobility scenarios with correlated multipath, stochastic latency, and obstacle constraints demonstrates improved belief convergence, smoother trajectories, and reduced estimation and tracking errors compared to naive, adaptive, and first-order baselines. The framework preserves classical GNSS-IMU fusion structure and can be integrated directly into existing flight control pipelines, supporting robust online autonomy in GNSS degraded environments.

Chinese Translation

可靠的在线自主性对于四旋翼在城市空域的操作至关重要，因为全球导航卫星系统（GNSS）测量会受到多径、遮挡和延迟问题的影响，这些问题引入了非平稳的、时间相关的误差，从而降低了传统GNSS-IMU融合的效果。本文提出了一种信念自适应在线自主框架，该框架通过显式的GNSS信任建模、二阶在线信念适应和延迟感知的非顺序测量处理来增强扩展卡尔曼滤波器（EKF）。GNSS信任被表示为一个潜在的信念状态，该状态调节测量加权和多径偏差的不确定性，并通过EKF一致性信号在线更新。与反应式协方差调节不同，所提出的方法能够在没有先前环境知识或离线训练的情况下，主动且稳定地适应传感器信任。在模拟的城市空中移动场景中进行的评估显示，与天真、自适应和一阶基线相比，该方法在相关多径、随机延迟和障碍约束下实现了更好的信念收敛、更平滑的轨迹，以及减少的估计和跟踪误差。该框架保留了经典的GNSS-IMU融合结构，并可以直接集成到现有的飞行控制管道中，从而支持在GNSS退化环境下的稳健在线自主性。

View on arXiv Download PDF AI Translation

cs.RO / 70 / 2609.14868

Primitive-Informed Sampling-Based MPC for Multi-Fingered Dexterous Manipulation

基于原始信息的采样模型预测控制在多指灵巧操作中的应用

Küçüktabak, Emek Barış, Patel, Karankumar, Cui, Jinda, Yang, Zhaodong, Sasabuchi, Kazuhiro, Takamatsu, Jun

Abstract

We present a primitive-informed sampling-based model predictive control (MPC) framework for multi-fingered dexterous manipulation. Sampling-based MPC avoids the need for gradients through complex contact dynamics, but direct exploration of the high-dimensional joint space is inefficient and makes performance strongly dependent on the sampling distribution. Our framework biases sampling using low-dimensional manipulation primitives that encode coordinated finger motions, while simultaneously optimizing joint-level residuals to adapt these motions to the current hand-object configuration. Task-related rollout constraints reject infeasible trajectories during forward simulation, improving the effective use of the sampling budget. We evaluate the approach on a physical Allegro hand using a synchronized MuJoCo digital twin. Ablations show that both the primitive and residual are necessary for reliable continuous in-hand rotation, that increasing the sampling budget alone does not recover this coordination, and that rollout constraints substantially improve success rate. A primitive extracted from a single object remains effective across object sizes and under model mismatch, and the framework further supports grasping, object reorientation, and coordinated arm-hand reach-grasp-transport, using primitives extracted from both a simulation-trained policy and human hand-motion data.

Chinese Translation

我们提出了一种基于原始信息的采样模型预测控制（MPC）框架，用于多指灵巧操作。基于采样的MPC避免了通过复杂接触动态计算梯度的需求，但直接探索高维关节空间效率低下，并且使性能强烈依赖于采样分布。我们的框架利用编码协调手指运动的低维操作原语来偏向采样，同时优化关节级残差，以将这些运动适应于当前的手-物体配置。与任务相关的展开约束在前向仿真过程中拒绝不可行的轨迹，从而提高了采样预算的有效利用。我们在一个物理的Allegro手上使用同步的MuJoCo数字双胞胎评估了该方法。消融实验表明，原始信息和残差对于可靠的手内连续旋转都是必要的，仅增加采样预算并不能恢复这种协调性，而展开约束显著提高了成功率。从单个物体提取的原始信息在不同物体大小和模型不匹配的情况下仍然有效，并且该框架进一步支持抓取、物体重新定向以及协调的手臂-手抓取-运输，使用从仿真训练策略和人类手部运动数据中提取的原始信息。

View on arXiv Download PDF AI Translation

cs.RO / 71 / 2609.14878

Real-World Reinforcement Learning with MPC Scaffolding for Dexterous Manipulation

基于模型预测控制的真实世界强化学习框架用于灵巧操作

Küçüktabak, Emek Barış, Patel, Karankumar, Yang, Zhaodong, Cui, Jinda, Sasabuchi, Kazuhiro, Takamatsu, Jun

Abstract

Real-world reinforcement learning (RL) offers a promising route to dexterous manipulation policies that can adapt directly from physical interaction, but learning is hindered by inefficient early exploration and costly failures. We propose a framework that uses sampling-based model predictive control (MPC) as scaffolding for real-world dexterous RL, providing structured prior experience and task-directed guidance during learning without human demonstrations or corrective actions. A small set of MPC trajectories is first used to populate an offline replay buffer and to pretrain the actor and critic. During online learning, MPC intermittently guides data collection while an off-policy Soft Actor-Critic learner trains from both prior MPC experience and newly collected physical interaction, with control gradually transitioning to the learned policy. On continuous in-hand rotation with a 16-DoF Allegro hand, the method reaches 100\% success in policy-only evaluation (5/5 trials) after 7 minutes of online RL, following initialization with 20 MPC trajectories collected on hardware in 12 minutes. Online training incurs about three object drops on average. After 20 minutes of online learning, the policy achieves more than five times the rotation speed of the MPC controller. It completes 1000 consecutive rotations over more than 110 minutes without a drop. Ablations show complementary benefits from MPC-based pretraining, retained MPC experience, and online MPC guidance. We further demonstrate rapid adaptation to different object geometries and successful goal-conditioned reorientation, showing that the framework enables efficient, low-intervention, real-world dexterous RL.

Chinese Translation

真实世界的强化学习（RL）为灵巧操作策略的直接适应提供了有希望的途径，但学习受到低效的早期探索和高昂的失败成本的阻碍。我们提出了一种框架，利用基于采样的模型预测控制（MPC）作为真实世界灵巧RL的支架，在学习过程中提供结构化的先验经验和任务导向的指导，而无需人类示范或纠正措施。首先使用一小组MPC轨迹填充离线重放缓冲区，并对演员和评论家进行预训练。在在线学习过程中，MPC间歇性地指导数据收集，同时一个离线策略的软演员-评论家（Soft Actor-Critic）学习者从先前的MPC经验和新收集的物理交互中进行训练，控制逐渐过渡到学习到的策略。在使用16自由度Allegro手进行的连续手内旋转任务中，该方法在经过7分钟的在线RL后，在仅使用策略评估时达到了100%的成功率（5/5次试验），而在此之前用12分钟在硬件上收集了20条MPC轨迹。在线训练平均会导致约三次物体掉落。经过20分钟的在线学习，该策略的旋转速度超过MPC控制器的五倍。它在110分钟内完成了1000次连续旋转而没有掉落。消融实验表明，基于MPC的预训练、保留的MPC经验和在线MPC指导具有互补的好处。我们进一步展示了对不同物体几何形状的快速适应和成功的目标条件重新定向，表明该框架能够实现高效、低干预的真实世界灵巧RL。

View on arXiv Download PDF AI Translation

cs.RO / 72 / 2609.14903

HydroMap: Probabilistic Water Surface Elevation Mapping for Semantic Scene Representation in Inland Waterways

HydroMap：用于内陆水道语义场景表示的概率水面高程映射

Luo, Zhongbi, Wang, Yunjia, Bruyninckx, Herman, Slaets, Peter

Abstract

Autonomous surface vehicles operating in inland waterways require a persistent representation of both surrounding structures and the water surface. LiDAR-based simultaneous localization and mapping often produces sparse or missing water returns, leaving this operational surface absent from the reconstructed scene. We propose HydroMap, an odometry-decoupled framework that reconstructs water surface elevation from stereo observations and integrates it with the structural map. Per-frame water points form joint cell observations with propagated stereo and pose uncertainty, and successive observations are fused into a persistent probabilistic elevation map. Semantic map conversion then combines the elevation map with structural geometry in a unified 2.5D representation of water, boundaries, structures, and overhead regions. On the Pohang Canal and Leuven Vaart datasets, the elevation RMSE remains below 5 cm relative to LiDAR references expressed in the same map frame. The elevation and semantic maps are published at 2 Hz and 1 Hz, respectively. HydroMap thereby complements LiDAR maps with a persistent representation of the water surface for downstream navigation in inland waterways.

Chinese Translation

在内陆水道中运行的自主水面车辆需要对周围结构和水面进行持续的表示。基于激光雷达的同时定位与地图构建（SLAM）往往会产生稀疏或缺失的水面回波，导致重建场景中缺少这一操作表面。我们提出了HydroMap，一个与里程计解耦的框架，通过立体观测重建水面高程，并将其与结构地图集成。每帧水点形成与传播的立体和姿态不确定性结合的联合单元观测，连续观测被融合为一个持久的概率高程地图。语义地图转换将高程地图与结构几何结合，形成水体、边界、结构和上方区域的统一2.5D表示。在浦项运河和鲁汶运河数据集上，高程的均方根误差（RMSE）相对于同一地图框架中的激光雷达参考保持在5厘米以下。高程和语义地图分别以2 Hz和1 Hz的频率发布。因此，HydroMap为内陆水道的下游导航提供了水面持久表示，补充了激光雷达地图。

View on arXiv Download PDF AI Translation

cs.RO / 73 / 2609.14935

Exact Feasibility Certification and Optimal Responsibility Allocation for Multi-Robot CBF Safety Filters

多机器人控制障碍函数（CBF）安全过滤器的精确可行性认证与最优责任分配

Sah, Chandan Kumar, Keshavan, Jishnu

Abstract

Multi-robot Control Barrier Function (CBF) safety filters can become infeasible, but a failed quadratic program (QP) does not indicate why the conflict occurred or how to resolve it. To address this, we develop an exact feasibility certificate for multi-agent CBF filters with heterogeneous control-affine dynamics and convex input sets. The certificate quantifies a feasibility reserve by separating the demand imposed by safety constraints from the available actuator supply. This decomposition shows when CBF gain tuning or increased actuation can and cannot resolve infeasibility, and identifies the agents and interactions responsible for the conflict. We further propose an algorithm to optimally allocate shared safety constraints by maximizing the worst local feasibility margin, yielding a linear program for polyhedral input sets. In $320$ paired closed-loop simulations, the proposed allocation reduces infeasible control steps from roughly $50\%$ to $6.2\%$, and reduces safety-violating runs from $118/160$ to $24/160$. In addition, across $52$ infeasibility events, the certificate identifies an interaction whose relaxation restores feasibility in $94\%$ of cases.

Chinese Translation

多机器人控制障碍函数（CBF）安全过滤器可能会变得不可行，但失败的二次规划（QP）并不能指明冲突发生的原因或如何解决它。为了解决这个问题，我们为具有异构控制仿射动态和凸输入集的多智能体CBF过滤器开发了一种精确的可行性证书。该证书通过将安全约束所施加的需求与可用的执行器供给分开来量化可行性储备。这种分解显示了何时CBF增益调节或增加执行力可以或不能解决不可行性，并识别出导致冲突的智能体和相互作用。我们进一步提出了一种算法，通过最大化最差局部可行性边际来最优分配共享安全约束，从而为多面体输入集生成线性规划。在320次配对闭环仿真中，所提分配将不可行控制步骤从约50%减少到6.2%，并将安全违规运行次数从118/160减少到24/160。此外，在52次不可行事件中，该证书识别出一种相互作用，其放松在94%的情况下恢复了可行性。

View on arXiv Download PDF AI Translation

cs.RO / 74 / 2609.14936

Comparing Trajectories from Positions Alone: Curvature-Based Time Alignment and Drift Error Metric

仅基于位置的轨迹比较：基于曲率的时间对齐和漂移误差度量

Daum, Effie, De Martini, Daniele, Dune, Claire, Pomerleau, François

Abstract

In field robotics, acquiring independent large-scale reference trajectories more accurate than the evaluated estimates remains an open challenge. The domain is widely reliant on Absolute Trajectory Error (ATE) and Relative Pose Error (RPE), computed with automated tools, that rest on assumptions and evaluation parameters rarely made explicit. When unreported, the errors can be misleading and hinder fair comparisons. This paper introduces a trajectory-evaluation protocol for standardized and reliable accuracy assessment in state estimation, localization, and Simultaneous Localization And Mapping (SLAM). The approach combines a novel temporal alignment method based on curvature signals with an error metric normalized by travelled distance. We explicitly account for temporal synchronization, sampling alignment, and extrinsic calibration, quantifying their influence through a sensitivity analysis. The proposed protocol contributes to more rigorous, reproducible, and standardized trajectory evaluation.

Chinese Translation

在领域机器人技术中，获取独立的大规模参考轨迹，其精度超过评估估计值，仍然是一个未解决的挑战。该领域广泛依赖于绝对轨迹误差（Absolute Trajectory Error, ATE）和相对姿态误差（Relative Pose Error, RPE），这些误差是通过自动化工具计算得出的，基于的假设和评估参数往往没有明确说明。当这些参数未被报告时，误差可能会产生误导，阻碍公平比较。本文提出了一种轨迹评估协议，用于在状态估计、定位和同时定位与地图构建（Simultaneous Localization And Mapping, SLAM）中进行标准化和可靠的精度评估。该方法结合了一种基于曲率信号的新颖时间对齐方法和一种按行驶距离归一化的误差度量。我们明确考虑了时间同步、采样对齐和外部校准，通过敏感性分析量化它们的影响。所提出的协议有助于实现更严格、可重复和标准化的轨迹评估。

View on arXiv Download PDF AI Translation

cs.RO / 75 / 2609.14984

Two-Stage Personalized Gait Phase Estimation in Stroke Survivors During Exoskeleton-Assisted Walking: An Offline Feasibility Study

中风幸存者在外骨骼辅助行走中的两阶段个性化步态相位估计：离线可行性研究

Ryu, Hyungseok, Hur, Pilwon

Abstract

This study evaluated personalized gait phase estimation for stroke survivors using functional inertial measurement unit (IMU) alignment and two-stage sequential adaptation of models pre-trained on healthy gait. The estimator used signals from a thigh-mounted IMU. Heel force-sensitive resistor measurements provided reference phase labels for offline adaptation and evaluation. Stage 1 established a distillation-regularized participant-specific model, and Stage 2 performed conditional refinement using low-rank adaptation. Long Short-Term Memory (LSTM), Temporal Convolutional Network (TCN), and Transformer models were evaluated in five stroke survivors walking with a powered knee exoskeleton using leave-one-subject-out hyperparameter selection and sequential test-then-adapt Stage 2 replay. Relative to the non-adapted baselines, Stage 1+2 reduced the mean participant-wise phase root mean square error by 84.2%, 77.0%, and 60.7%, respectively. The Transformer achieved the lowest final error (2.90 +- 1.13$% of the gait cycle) and heel-strike timing error (23.7 +- 4.5ms). Policy-specific ablations showed that every-cycle updates generally produced the lowest or near-lowest error, whereas conditional updating reduced the update frequency with small accuracy differences. After personalization, alignment produced model-dependent changes in phase error while preserving or improving heel-strike detection and reducing heel-strike timing error for the LSTM and Transformer. Concurrent embedded tests showed that the TCN and Transformer maintained 100-Hz inference during Stage 2 updates without deadline misses, whereas the LSTM missed the 10-ms deadline in 6.6% of inferences. All updates completed within 0.8s. These results support the offline feasibility and embedded computational timing of the proposed framework for exoskeleton-assisted walking.

Chinese Translation

本研究评估了针对中风幸存者的个性化步态相位估计，采用功能性惯性测量单元（IMU）对齐和基于健康步态预训练模型的两阶段顺序适应。估计器使用来自大腿安装的IMU的信号。跟踪力敏电阻测量提供了离线适应和评估的参考相位标签。第一阶段建立了一个蒸馏正则化的参与者特定模型，第二阶段则通过低秩适应进行条件细化。在五名使用动力膝关节外骨骼行走的中风幸存者中评估了长短期记忆网络（LSTM）、时间卷积网络（TCN）和变换器（Transformer）模型，采用逐个受试者的超参数选择和顺序测试后再适应的第二阶段重播。与未适应的基线相比，第一阶段加第二阶段分别将参与者平均相位均方根误差降低了84.2%、77.0%和60.7%。变换器实现了最低的最终误差（步态周期的2.90 ± 1.13%）和跟踪时机误差（23.7 ± 4.5毫秒）。特定策略的消融实验表明，每周期更新通常产生最低或接近最低的误差，而条件更新则减少了更新频率，且准确性差异较小。个性化后，对齐产生了模型依赖的相位误差变化，同时保持或改善了LSTM和变换器的跟踪检测，并减少了跟踪时机误差。并行嵌入测试显示，TCN和变换器在第二阶段更新期间保持100Hz的推理速率，没有错过截止时间，而LSTM在6.6%的推理中错过了10毫秒的截止时间。所有更新均在0.8秒内完成。这些结果支持了所提框架在外骨骼辅助行走中的离线可行性和嵌入式计算时效性。

View on arXiv Download PDF AI Translation

cs.RO / 76 / 2609.14997

From Learned-Mode AV-Traffic Pairing to Planner Decisions: A Marginal-Preserving Study on Argoverse 2

从学习模式的自动驾驶车辆与交通配对到规划者决策：对 Argoverse 2 的边际保持研究

Wang, Jingyu

Abstract

Joint motion forecasts pair each autonomous-vehicle (AV) future with surrounding traffic, but actor-level metrics do not show whether that structure matters to a planner. We study this question with a marginal-preserving product control that removes AV-traffic pairing among the learned modes while retaining the fixed constant-velocity pair and holding trajectories, actor-level marginals before planner conditioning, candidates, the cost terms and weights, and fallback fixed. The intervention also changes candidate-conditioned concentration. Across twelve runs on 1,400 held-out Argoverse 2 scenarios, the intervention changes 3.0% of route-level offline selections at $\tau=4$ m. Control-minus-joint recorded-trajectory regret is $-0.026$ and $-0.118$ at the two training sizes; crossed and seed-$t$ intervals span zero. At $\tau=1$ m, relative costs change in 87.9% of route evaluations and route-level offline selections in 8.1%. Before concentration matching, descriptive outcome estimates favor the control. Most of this gap disappears along an approximate concentration-matching path; the remaining contrasts are $+0.112$ and $-0.047$, and both crossed intervals span zero. Actor-level forecast metrics remain identical. Pairing-strength and temperature sweeps show that the decision contrast grows with pairing removal and sharper conditioning. The intervention changes planner decisions even though actor-level metrics remain unchanged. The matching analysis, however, cannot separate any recorded-outcome effect of learned-mode AV-traffic pairing from the accompanying change in conditioned concentration.

Chinese Translation

联合运动预测将每个自动驾驶车辆（AV）的未来与周围交通配对，但参与者级别的指标并未显示这种结构对规划者的重要性。我们通过一种边际保持的产品控制来研究这个问题，该控制在保留固定的恒定速度配对和保持轨迹、规划者条件之前的参与者级别边际、候选者、成本项和权重以及后备固定的同时，去除了学习模式中的 AV-交通配对。该干预还改变了候选者条件下的集中度。在对 1,400 个保留的 Argoverse 2 场景进行的十二次实验中，该干预在 $ au=4$ m 时改变了 3.0% 的路线级离线选择。控制减去联合记录轨迹的遗憾为 $-0.026$ 和 $-0.118$，在两个训练规模下；交叉和种子-$t$ 区间跨越零。在 $ au=1$ m 时，相对成本在 87.9% 的路线评估中发生变化，路线级离线选择中变化为 8.1%。在集中度匹配之前，描述性结果估计更倾向于控制。沿着近似的集中度匹配路径，大部分差距消失；剩余的对比为 $+0.112$ 和 $-0.047$，且两个交叉区间均跨越零。参与者级别的预测指标保持不变。配对强度和温度的变化显示，随着配对的去除和更严格的条件，决策对比增强。尽管参与者级别的指标保持不变，该干预仍改变了规划者的决策。然而，匹配分析无法将学习模式的 AV-交通配对的任何记录结果效应与伴随的条件集中度变化分开。

View on arXiv Download PDF AI Translation

cs.RO / 77 / 2609.15005

IMPACT-VLA: Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies

IMPACT-VLA：基于反事实轨迹的交互感知多模态传播归因用于视觉-语言-动作策略

Kim, Jinwoong, Park, Sangjin

Abstract

Vision-Language-Action (VLA) policies perform robot manipulation tasks using multimodal inputs such as visual observations, proprioceptive states, and language instructions. However, it remains unclear at which execution stages each modality contributes to final task success and how input interventions propagate through subsequent states, observations, and actions. Existing attribution approaches primarily measure local sensitivity or temporally aggregated importance, limiting their ability to capture phase-dependent contributions and cross-phase dependencies. We propose Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies (IMPACT-VLA). IMPACT-VLA constructs behavioral phases from action transitions in a successful reference rollout, aligns them with policy query boundaries, and defines phase-modality blocks as attribution units. It then performs closed-loop counterfactual re-execution to quantify each block's contribution to final task success. We further analyze cross-phase non-additive interactions and trajectory propagation while distinguishing behavioral from functional recovery. Across 30 LIBERO robot manipulation tasks using OpenVLA-OFT, dominant-modality transitions occurred in 25 tasks (83.3%), and closed-loop attribution identified task-critical information more faithfully than Static Action Perturbation. Later-block marginal gains for negatively interacting pairs increased by approximately 3.3x under early-phase input replacement, while functional recovery could occur without behavioral recovery. These results reveal when multimodal inputs support task success and how their contributions become conditionally coupled during closed-loop execution.

Chinese Translation

视觉-语言-动作（VLA）策略利用多模态输入（如视觉观察、自我感知状态和语言指令）执行机器人操作任务。然而，目前尚不清楚每种模态在执行的不同阶段对最终任务成功的贡献，以及输入干预如何在后续状态、观察和动作中传播。现有的归因方法主要测量局部敏感性或时间聚合的重要性，限制了它们捕捉阶段依赖贡献和跨阶段依赖的能力。我们提出了基于反事实轨迹的交互感知多模态传播归因（IMPACT-VLA），用于视觉-语言-动作策略。IMPACT-VLA 从成功参考执行中的动作转变构建行为阶段，将其与策略查询边界对齐，并将阶段-模态块定义为归因单元。然后，它执行闭环反事实重新执行，以量化每个块对最终任务成功的贡献。我们进一步分析跨阶段的非加性交互和轨迹传播，同时区分行为恢复与功能恢复。在使用 OpenVLA-OFT 的 30 个 LIBERO 机器人操作任务中，主导模态转变发生在 25 个任务中（83.3%），闭环归因比静态动作扰动更准确地识别出任务关键性信息。在早期阶段输入替换下，后期块的边际增益对于负交互对增加了约 3.3 倍，而功能恢复可以在没有行为恢复的情况下发生。这些结果揭示了多模态输入何时支持任务成功，以及它们的贡献在闭环执行过程中如何变得有条件地耦合。

View on arXiv Download PDF AI Translation

cs.RO / 78 / 2609.15012

Atomic Motion Coordinate for Language-Steerable and Force-Responsive Manipulation

用于语言引导和力响应操控的原子运动坐标

Zhai, Jiaqi, Zhao, Jingkai, Yang, Chen, Ma, Siyuan, Zhang, Yutian, Yang, Liwen, Wu, Qinglian, Fan, Weiqi, Wang, Yifei, Zheng, Yi, Gu, Chenxi, Wei, Dong, Zhang, Wei

Abstract

Can changing only the language instruction redirect a VLA policy's end effector, or does the visually driven motion prior dominate? We present Atomic Motion Coordinate, a geometry-grounded coordinate for steerable and force-responsive manipulation. Each arm owns thirteen signed translation, rotation, and hold atoms grounded from text and forward kinematics with vision withheld, and the coordinate is injected into every action-expert block via weighted codebook alignment. Contact history modulates the same coordinate through a bounded spherical residual that is recomputed from a fixed nominal latent to regenerate only the unexecuted horizon suffix. Across 7,520 offline horizon interventions, opposite-atom separation reaches 92.5/83.1% (single/dual) versus 39.1/24.0% for LA4VLA-style. Across 50 real-robot trials per task, AMC raises OOD fruit progress from 60.5% to 87.8%; force adaptation raises Plug/Vase from 59.0/71.5% to 78.5/75.2%.

Chinese Translation

仅通过改变语言指令是否能重新引导VLA策略的末端执行器，还是视觉驱动的运动先验占主导地位？我们提出了原子运动坐标（Atomic Motion Coordinate），这是一种基于几何的坐标，用于可引导和力响应的操控。每个手臂拥有十三个带符号的平移、旋转和保持原子，这些原子基于文本和前向运动学生成，同时视觉信息被屏蔽，该坐标通过加权代码本对齐注入到每个动作专家模块中。接触历史通过一个有界的球形残差调节相同的坐标，该残差是从固定的名义潜变量重新计算的，以仅再生未执行的地平线后缀。在7,520次离线地平线干预中，反原子分离率达到了92.5%/83.1%（单一/双重）相比于LA4VLA风格的39.1%/24.0%。在每个任务的50次真实机器人试验中，AMC将OOD水果的进展从60.5%提高到87.8%；力适应将Plug/Vase的进展从59.0%/71.5%提高到78.5%/75.2%。

View on arXiv Download PDF AI Translation

cs.RO / 79 / 2609.15014

Steering Generative Robot Policies with Lexicographic Preferences

通过字典优先级引导生成机器人策略

Jia, Yixuan, How, Jonathan P.

Abstract

Pretrained generative robot policies can produce effective behaviors across diverse environments, but deployment can lead to requirements and preferences that may not have been represented during training. Furthermore, at deployment, an operator, user, or application may assign these requirements and preferences a priority order that can vary across deployments. For example, embodiment-specific feasibility constraints may need to be satisfied first, while user-specific preferences guide behavior among the feasible options. We show that a frozen generative robot policy---based on either diffusion or flow matching---can be steered at inference time to respect such lexicographically ordered deployment objectives. To achieve this, we introduce two modifications to the sampler. First, we apply dynamic-barrier guidance to sampled trajectories, constraining lower-priority updates so that higher-priority costs do not increase (up to first order). Second, we select the executed sample using a cascade that successively filters candidate samples according to each priority level. The policy weights remain unchanged. On a navigation benchmark, we demonstrate that our method improves success, traversability, and preference compliance over the frozen policy, and achieves substantially better compliance than tuned weighted-sum baselines. The same method transfers to a flow-matching manipulation policy on LIBERO, where it improves compliance without reducing task success. A controlled manipulation study further shows that, in settings where a fixed weight can match the desired ordering, the dynamic barrier reaches comparable best performance over a substantially wider range of parameter settings.

Chinese Translation

预训练的生成机器人策略能够在多样的环境中产生有效的行为，但在部署过程中可能会出现训练时未能体现的需求和偏好。此外，在部署时，操作员、用户或应用程序可能会为这些需求和偏好分配一个优先级顺序，而这个顺序可能在不同的部署中有所不同。例如，特定于体现的可行性约束可能需要优先满足，而特定于用户的偏好则在可行选项中指导行为。我们展示了一种冻结的生成机器人策略——基于扩散或流匹配——可以在推理时被引导，以遵循这样的字典优先级部署目标。为此，我们对采样器引入了两个修改。首先，我们对采样轨迹应用动态障碍引导，限制低优先级更新，以确保高优先级成本不增加（最多到一阶）。其次，我们使用级联选择执行的样本，依次根据每个优先级水平过滤候选样本。策略权重保持不变。在一个导航基准测试中，我们证明了我们的方法在成功率、可通行性和偏好遵从性方面优于冻结策略，并且在遵从性方面显著优于调优的加权和基线。同样的方法也适用于LIBERO上的流匹配操作策略，在不降低任务成功率的情况下提高了遵从性。一项受控操作研究进一步表明，在固定权重能够匹配所需顺序的情况下，动态障碍在更广泛的参数设置范围内达到可比的最佳性能。

View on arXiv Download PDF AI Translation

cs.RO / 80 / 2609.15082

Task-Distribution-Aware Counterweight Synthesis and Constrained Co-Design for Serial Manipulators

任务分布感知的平衡重合成与串联机械手的约束协同设计

Abbadi, Mohammad

Abstract

Passive counterweights are simple gravity compensators, but a counterweight selected from a single pose is not generally optimal for the configurations and tasks a manipulator actually executes. This paper develops a task-distribution-aware synthesis framework in which the operating distribution $\rho(q)$ enters the design explicitly. For a counterweight moment $p=m_c r_c$ with gravity torque $-gp\phi(q)$, the weighted mean-square residual gravity torque has the closed-form minimizer $p^*=E_\rho[\tau_g\phi]/(gE_\rho[\phi^2])$. If payload gravity torque is affine in payload mass, the optimum is also affine: $p^*(m_p,\rho)=p_0^*(\rho)+m_pK_p(\rho)$. For fixed static moment, added counterweight inertia is $I_c=pr_c$ while mass is $m_c=p/r_c$, so mass-radius selection is underdetermined unless physical constraints are specified. A recovered three-link manipulator is used as a case study. At $r_c=0.20$ m, zero-payload equivalent optima are 0.672 kg for uniform joint-space operation, 0.683 kg for approximately uniform task-space operation, 0.713 kg for a representative pick-and-place family, and 0.952 kg for a high-gravity-biased distribution, a change of more than 40% caused solely by the operating distribution. Nondominated fronts show that preferred mass-radius pairs depend on declared engineering bounds. A rated-torque-referenced all-joint screen increases zero-payload feasible task-space coverage from 78.1% without compensation to 93.7% for the uniform-distribution design. A lumped point-mass trajectory study gives a provisional crossover from no counterweight at very aggressive motion to stronger compensation as motion slows. These actuator and dynamic results are engineering consequence studies rather than physical validation.

Chinese Translation

被动平衡重是简单的重力补偿器，但从单一姿态选择的平衡重通常并不是机械手实际执行的配置和任务的最佳选择。本文开发了一种任务分布感知的合成框架，其中操作分布 $ ho(q)$ 明确地进入设计中。对于重力矩 $p=m_c r_c$ 和重力扭矩 $-gp heta(q)$，加权均方残余重力扭矩的闭式最小化解为 $p^*=E_ ho[ au_g heta]/(gE_ ho[ heta^2])$。如果有效载荷重力扭矩与有效载荷质量呈仿射关系，则最优解也是仿射的：$p^*(m_p, ho)=p_0^*( ho)+m_pK_p( ho)$。对于固定的静态矩，增加的平衡重惯性为 $I_c=pr_c$，而质量为 $m_c=p/r_c$，因此质量-半径选择在未指定物理约束的情况下是欠定的。以恢复的三连杆机械手作为案例研究。在 $r_c=0.20$ m 时，零载荷等效最优值为均匀关节空间操作的 0.672 kg，近似均匀任务空间操作的 0.683 kg，代表性的抓取与放置任务的 0.713 kg，以及高重力偏置分布的 0.952 kg，操作分布的变化导致了超过 40% 的变化。非支配前沿显示，优选的质量-半径对依赖于声明的工程界限。以额定扭矩为参考的全关节屏幕将零载荷可行任务空间覆盖率从无补偿的 78.1% 提高到均匀分布设计的 93.7%。集中点质量轨迹研究显示，从没有平衡重的非常激进运动到运动减缓时更强的补偿之间存在一个临界点。这些驱动器和动态结果是工程后果研究，而非物理验证。

View on arXiv Download PDF AI Translation

cs.RO / 81 / 2609.15113

Legislating World-Model-Based Planning with Legal Reasoning

基于世界模型的规划立法与法律推理

Waldner, Dylan, Kantaros, Yiannis, Governatori, Guido, Miikkulainen, Risto, Banifatemi, Amir

Abstract

As robotic systems grow more general, legal norms are needed to integrate them into society. This paper extends the isomorphism problem of aligning legal source texts with their encodings, and measures two key challenges to robot normative control: (1) the \textit{grounding isomorphism gap}, where perception error grounds false atoms for legal reasoning, and (2) the \textit{ontological isomorphism gap}, where one legal conclusion admits many faithful translations into planning constraints. The paper introduces a legal planning stack that employs Defeasible Deontic Logic (DDL) to constrain a motion planner. The stack leverages learned world models to plan and to provide legal context, enabling \textit{ex ante} governance that intervenes before an illegal action is executed. It was deployed on a simulated robot arm pushing a cube across a $3\times3$ grid. The findings were (1) the legislated agent abided substantially more often than the non-legislated one, and modeling perception uncertainty lifted abidance even further, (2) the legal reasoning ran efficiently at runtime and its verdicts were auditable, and (3) the stack adapted to exogenous signals and endogenous rule changes. Both gaps were measured: (4) world model and probe error corrupted the factual input for the DDL reasoner, and (5) a single law admitted several faithful metric interpretations yielding drastically different abidance. Thus, \textit{ex ante} legislation functions as intended, and closing these gaps with a standardized mapping from the law to runtime constraints and improved fact grounding from perception will yield robust laws that align robot behavior with society's norms.

Chinese Translation

随着机器人系统变得越来越通用，需要法律规范将其融入社会。本文扩展了法律源文本与其编码之间的一致性问题，并衡量了机器人规范控制的两个关键挑战：（1） extit{基础一致性差距}，即感知误差为法律推理提供了错误的原子基础；（2） extit{本体一致性差距}，即一个法律结论可以有多种忠实的翻译为规划约束。本文介绍了一种法律规划堆栈，该堆栈采用可驳回的义务逻辑（Defeasible Deontic Logic, DDL）来约束运动规划器。该堆栈利用学习到的世界模型进行规划并提供法律背景，实现了 extit{事前}治理，在非法行为执行之前进行干预。该系统在一个模拟的机器人手臂上部署，推动一个立方体在$3 imes3$网格上移动。研究结果表明：（1）立法代理的遵守率显著高于非立法代理，且建模感知不确定性进一步提升了遵守率；（2）法律推理在运行时高效，且其裁决可审计；（3）堆栈能够适应外部信号和内部规则变化。两个差距被测量：（4）世界模型和探测误差破坏了DDL推理器的事实输入；（5）单一法律允许多种忠实的度量解释，导致遵守率的显著差异。因此， extit{事前}立法按预期发挥作用，缩小这些差距，通过从法律到运行时约束的标准化映射和改善感知的事实基础，将产生与社会规范一致的稳健法律。

View on arXiv Download PDF AI Translation

cs.RO / 82 / 2609.15142

C$^2$Nav: Compare Before You Commit for Zero-Shot Vision-and-Language Navigation

C$^2$Nav：在零样本视觉与语言导航中比较后再决策

Zheng, Runtian, Zhang, Congpeng, Liu, Ying

Abstract

Zero-shot vision-and-language navigation in continuous environments (VLN-CE) increasingly places foundation vision-language models (VLMs) inside the navigation loop. Existing systems commonly request cardinal outputs such as waypoints, pixels, headings, progress values, or absolute arrival decisions, coupling a generative response to geometric magnitude or an irreversible commitment. We study a complementary model-robot interface: the VLM compares controller-constructed alternatives, while geometry, thresholds, action magnitude, and execution remain on the physical side. We instantiate this idea in C2Nav, a training-free framework with three coordinated faculties. Seeing performs ordinal Gaze Election over physically vetted candidate views; Remembering maintains a compact route sketch and compares adjacent instruction-leg hypotheses; and Arriving combines a hesitation ladder, look-back comparison, and revocable walk-back for reliable stopping. On the public OpenNav R2R-CE 100 protocol, C2Nav with Qwen3-VL-8B-Instruct obtains 41.0% OSR, 31.0% SR, and 16.7% SPL, while the same interface with the standard GPT-5.5 model reaches 54.0% OSR, 44.0% SR, and 29.0% SPL. Whole-faculty ablations reduce SR to 14.0% without Seeing, 25.0% without Remembering, and 29.0% without Arriving. Matched role inversions that replace only the comparative answer form with cardinal/absolute questions reduce SR to 12.0%, 28.0%, and 21.0% in the spatial, transition, and terminal slots, respectively. The results indicate that a constrained decision interface and stronger VLM reasoning are complementary rather than interchangeable.

Chinese Translation

在连续环境中的零样本视觉与语言导航（VLN-CE）日益将基础视觉-语言模型（VLMs）纳入导航循环。现有系统通常请求诸如航点、像素、方向、进度值或绝对到达决策等基准输出，将生成响应与几何量或不可逆承诺相结合。我们研究了一种互补的模型-机器人接口：VLM比较控制器构建的替代方案，而几何、阈值、动作量和执行则保持在物理层面。我们在C2Nav中实现了这一理念，这是一个无训练的框架，具有三个协调的功能。视觉功能在经过物理验证的候选视图上执行序数注视选择；记忆功能保持紧凑的路线草图并比较相邻的指令-腿假设；到达功能结合了犹豫阶梯、回顾比较和可撤回的回退，以实现可靠的停止。在公共的OpenNav R2R-CE 100协议上，C2Nav与Qwen3-VL-8B-Instruct模型获得了41.0%的开放式成功率（OSR）、31.0%的成功率（SR）和16.7%的路径长度成功率（SPL），而同一接口与标准的GPT-5.5模型达到54.0%的OSR、44.0%的SR和29.0%的SPL。整体功能消融实验显示，缺少视觉功能时SR降至14.0%，缺少记忆功能时降至25.0%，缺少到达功能时降至29.0%。仅将比较答案形式替换为基准/绝对问题的角色反转实验，分别使空间、过渡和终端位置的SR降至12.0%、28.0%和21.0%。结果表明，受限的决策接口和更强的VLM推理是互补的，而非可互换的。

View on arXiv Download PDF AI Translation

cs.RO / 83 / 2609.15162

LieSpline-DP: Lie-Group B-Spline Diffusion Policy for Smooth Robot Manipulation

LieSpline-DP：用于平滑机器人操作的李群 B 样条扩散策略

Xie, Erxuan, Liu, Bang, Nie, Pingyun, Liu, Xingkai, Fu, Zhuang, Zhang, Bo

Abstract

Diffusion Policy (DP) is a powerful Learning from Demonstration (LfD) method for robotic manipulation, yet it suffers from discontinuous and non-smooth trajectories. Spline-based action representations promote smooth motion within individual action chunks, but existing spline-based methods neither guarantee cross-chunk $C^2$ continuity nor account for the group structure of $\mathrm{SE}(3)$. We therefore propose LieSpline-DP, a Lie-group B-spline diffusion policy that generates end-effector trajectories directly on $\mathrm{SE}(3)$ and couples consecutive plans by sharing their boundary control poses, ensuring $C^2$ continuity throughout the entire planned trajectory. Across three real-robot tasks, LieSpline-DP produces lower trajectory jerk and higher task success rates than the DP baseline. The gains are particularly pronounced in real-world tasks involving liquids and flexible objects: in our real-robot experiments, LieSpline-DP achieved a 100% success rate on both pouring and bucket hooking, whereas the DP baseline achieved only 10% and 30%, respectively.

Chinese Translation

扩散策略（Diffusion Policy, DP）是一种强大的从示范学习（Learning from Demonstration, LfD）方法，用于机器人操作，但它存在不连续和非平滑轨迹的问题。基于样条的动作表示促进了单个动作块内的平滑运动，但现有的基于样条的方法既无法保证跨块的 $C^2$ 连续性，也未考虑 $ ext{SE}(3)$ 的群结构。因此，我们提出了 LieSpline-DP，这是一种李群 B 样条扩散策略，能够直接在 $ ext{SE}(3)$ 上生成末端执行器轨迹，并通过共享边界控制姿态来耦合连续的规划，确保整个规划轨迹的 $C^2$ 连续性。在三个真实机器人任务中，LieSpline-DP 产生的轨迹抖动更小，任务成功率更高，相较于 DP 基线，效果显著。在涉及液体和柔性物体的真实世界任务中，LieSpline-DP 的优势尤为明显：在我们的真实机器人实验中，LieSpline-DP 在倾倒和桶钩操作中均达到了 100% 的成功率，而 DP 基线的成功率仅为 10% 和 30%。

View on arXiv Download PDF AI Translation

cs.RO / 84 / 2609.15195

HarnessVLN: Unifying Training-Free Embodied Navigation through an Agent Harness

HarnessVLN：通过代理工具统一无训练的具身导航

Chen, Yang, Che, Lirong, Huang, Zhenyu, Fu, Wenbo, Wang, Chuang, Cao, Xu, Liu, Daqi, Yang, Yuzhe, Su, Jian, Guo, Lan-Zhe

Abstract

Embodied navigation requires agents to interpret visual observations, accumulate spatial knowledge, and execute actions to follow instructions or locate objects. Training-based methods face generalization challenges, while training-free methods exploit multimodal large language models (MLLMs) but often lack mechanisms to reconcile proposed actions with spatial evidence, task progress, and execution failures. We present HarnessVLN, a zero-shot, training-free framework whose Agent Harness coordinates perception, retrieval, grounding, navigation, recovery, and termination through a unified tool interface. The Harness validates planner proposals against spatial evidence, geometric feasibility, and subgoal consistency, incorporating structured tool feedback into subsequent decisions. Hierarchical event memory tracks task progress and execution history, while a persistent Spatiotemporal Graph maintains reusable spatial evidence and failure annotations for verification and recovery. A replaceable Navigation Executor converts validated targets into executable motions, allowing the same Harness protocol to support instruction-following and object-goal navigation. HarnessVLN achieves success rates of 60.8%, 53.9%, 76.0%, and 59.3% on R2R, RxR, HM3D-v2, and HM3D-OVON, respectively, surpassing prior training-free SOTA results. Humanoid deployment further demonstrates its applicability to both tasks in real-world environments. The project page is: https://harnessvln.netlify.app/.

Chinese Translation

具身导航要求代理解释视觉观察、积累空间知识并执行动作以遵循指令或定位物体。基于训练的方法面临泛化挑战，而无训练的方法利用多模态大型语言模型（MLLMs），但通常缺乏将提议动作与空间证据、任务进展和执行失败进行调和的机制。我们提出了HarnessVLN，这是一种零-shot、无训练的框架，其代理工具协调感知、检索、定位、导航、恢复和终止，通过统一的工具接口进行操作。该工具验证规划者的提议是否符合空间证据、几何可行性和子目标一致性，并将结构化的工具反馈纳入后续决策中。分层事件记忆跟踪任务进展和执行历史，而持久的时空图维护可重用的空间证据和失败注释以便于验证和恢复。可替换的导航执行器将验证后的目标转换为可执行的动作，使得相同的工具协议能够支持遵循指令和目标物体导航。HarnessVLN在R2R、RxR、HM3D-v2和HM3D-OVON上的成功率分别为60.8%、53.9%、76.0%和59.3%，超越了之前的无训练状态下的最佳结果。类人部署进一步展示了其在现实环境中对这两项任务的适用性。项目页面为：https://harnessvln.netlify.app/

View on arXiv Download PDF AI Translation

cs.RO / 85 / 2609.15198

PredTac: Learning Contact-Rich Manipulation with Predicted Touch

PredTac：通过预测触觉学习接触丰富的操作

Fan, Weijia, Guo, Daqiang

Abstract

Contact-rich manipulation benefits from tactile feedback, yet physical tactile sensors introduce hardware, calibration, synchronization, and maintenance costs that complicate policy learning and deployment. We formulate predicted touch as an alternative to measured tactile input and present PredTac, a framework that learns to infer tactile states from causal visual observations and robot states and uses the predicted touch as an explicit interface for policy learning and execution. A tactile predictor is first trained with tactile supervision and then used to provide contact information without requiring measured tactile input during downstream policy training or execution. We evaluate PredTac across three contact-rich manipulation tasks in simulation and on a real robot, and further examine how policy performance depends on the predicted contact content. In simulation goal-offset evaluations, predicted-touch policies achieve 27.0%, 52.0%, and 44.7% success on USB, Barbed-spike, and Valve, respectively, improving over the visual baseline by 8.0-13.7 percentage points. On the real robot, predicted-touch ACT achieves 70.0%, 50.0%, and 90.0% success on USB insertion, Barbed extraction, and Valve rotation, respectively, with a three-task mean of 70.0%, approaching measured-touch ACT at 72.2% and substantially outperforming visual ACT at 21.1%. Fixed-policy interventions further show that performance is sensitive to the spatial structure of predicted contact, with spatial rearrangement at fixed value distributions reducing Valve success by 10.7 percentage points. These results demonstrate that predicted touch can provide useful contact information for contact-rich manipulation without requiring tactile sensing as a policy input.

Chinese Translation

接触丰富的操作受益于触觉反馈，然而物理触觉传感器引入了硬件、校准、同步和维护成本，这些都使得策略学习和部署变得复杂。我们将预测触觉作为测量触觉输入的替代方案，并提出了PredTac，一个从因果视觉观察和机器人状态中学习推断触觉状态的框架，并将预测触觉作为策略学习和执行的明确接口。首先，触觉预测器在触觉监督下进行训练，然后在下游策略训练或执行过程中提供接触信息，而无需测量触觉输入。我们在模拟环境和真实机器人上评估了PredTac在三个接触丰富的操作任务中的表现，并进一步研究了策略性能如何依赖于预测的接触内容。在模拟目标偏移评估中，预测触觉策略在USB、带刺提取和阀门任务上的成功率分别为27.0%、52.0%和44.7%，比视觉基线提高了8.0-13.7个百分点。在真实机器人上，预测触觉ACT在USB插入、带刺提取和阀门旋转任务上的成功率分别为70.0%、50.0%和90.0%，三项任务的平均成功率为70.0%，接近测量触觉ACT的72.2%，并显著优于视觉ACT的21.1%。固定策略干预进一步表明，性能对预测接触的空间结构敏感，在固定值分布下的空间重排使阀门任务的成功率降低了10.7个百分点。这些结果表明，预测触觉可以在不需要触觉传感作为策略输入的情况下，为接触丰富的操作提供有用的接触信息。

View on arXiv Download PDF AI Translation

cs.RO / 86 / 2609.15213

X-WBC: A Cross-Embodiment Foundation Model for Humanoid Whole-Body Control

X-WBC：一种用于类人全身控制的跨体现基础模型

Zhang, Juntong, Gu, Chun, Zhang, Li

Abstract

Scaling humanoid whole-body control toward general-purpose deployment requires large human motion corpora and training experience shared across robot bodies. Existing methods usually train one policy per robot, leaving motion experience isolated across embodiments. We introduce X-WBC, a cross-embodiment foundation framework that separates relatively shared human motion semantics from embodiment-specific physical execution. Human-centered command tokens align full human motion, robot reference motion, and sparse VR observations. A causal Transformer learns reusable temporal structure from mixed multi-robot rollouts, while lightweight robot-specific modules map the shared representation to each robot's proprioception and action space. Across nine simulated embodiments, external motions, and four real robots, experiments show that joint training improves tracking, the aligned representation supports consistent control across command sources, and the learned policy remains competitive beyond the training corpus. These results support heterogeneous humanoids as joint data sources and establish cross-embodiment joint training as a practical route toward whole-body control foundation models.

Chinese Translation

将类人全身控制扩展到通用部署需要大量的人类运动数据和跨机器人身体共享的训练经验。现有方法通常为每个机器人训练一个策略，导致运动经验在不同体现之间孤立。我们提出了X-WBC，一种跨体现的基础框架，它将相对共享的人类运动语义与体现特定的物理执行分离。以人为中心的指令标记对齐完整的人类运动、机器人参考运动和稀疏的虚拟现实观察。因果Transformer从混合多机器人回放中学习可重用的时间结构，而轻量级的机器人特定模块将共享表示映射到每个机器人的本体感知和动作空间。在九个模拟体现、外部运动和四个真实机器人上进行的实验表明，联合训练改善了跟踪，所对齐的表示支持跨指令源的一致控制，且所学习的策略在训练数据集之外仍然具有竞争力。这些结果支持异构类人机器人作为联合数据源，并确立跨体现联合训练作为实现全身控制基础模型的实际途径。

View on arXiv Download PDF AI Translation

cs.RO / 87 / 2609.15232

A Vision Based Framework Integrating Attention and Action Cues for Interpretable Cognitive Workload Assessment in Human Robot Collaborative Assembly

基于视觉的框架整合注意力与行动线索，以实现人机协作装配中的可解释认知负荷评估

Xionga, Junyan, Feng, Naiyi, Xia, Xingke, Fan, Qihang, Chen, Suchang, Guo, Daqiang

Abstract

The introduction of human-robot collaboration (HRC) in industrial assembly operations is revolutionizing the manufacturing landscape. In this evolving environment, operators are required to seamlessly coordinate their manual tasks with real-time task information and robotic behaviors. These demands fluctuate during operation, yet conventional workload assessments depend on body-worn physiological sensors that complicate practical deployment. Here, we present a vision-based attention--action framework for continuous and interpretable workload-related assessment in HRC assembly. The framework combines RGB-D observations with robot states and calibrated task-related areas to construct a temporally confirmed representation of operator behavior. This representation identifies where task demand is concentrated and explains how it develops when attention and action diverge, the task context changes, or the operator hesitates. We evaluated the framework in a three-level collaborative gearbox assembly experiment with ten participants, using subjective ratings and synchronized physiological signals as independent references. Raw NASA-TLX ratings confirmed increasing perceived workload across conditions, with significant effects on overall workload and its mental and temporal dimensions. The vision-derived HRC-CWL output was significantly associated with ECG-derived features in seven of nine participants with complete correlation data. Synchronized interaction episodes further showed temporal correspondence between detected hesitation and physiological activity. Real-time deployment demonstrated that the framework can operate without requiring operators to wear additional sensors. These findings support HRC-CWL as an interpretable behavioral proxy for cognitive ergonomics analysis and adaptive robot assistance, rather than a direct psychophysiological measure of workload.

Chinese Translation

人机协作（HRC）在工业装配操作中的引入正在彻底改变制造业的格局。在这一不断发展的环境中，操作员需要将手动任务与实时任务信息和机器人行为无缝协调。这些需求在操作过程中会有所波动，但传统的负荷评估依赖于穿戴式生理传感器，这使得实际部署变得复杂。在此，我们提出了一种基于视觉的注意力-行动框架，用于在HRC装配中进行连续和可解释的负荷相关评估。该框架结合了RGB-D观测、机器人状态和校准的任务相关区域，以构建操作员行为的时间确认表示。该表示识别了任务需求的集中区域，并解释了当注意力与行动分歧、任务背景变化或操作员犹豫时，任务需求是如何发展的。我们在一个包含十名参与者的三层协作齿轮箱装配实验中评估了该框架，使用主观评分和同步生理信号作为独立参考。原始的NASA-TLX评分确认了在不同条件下感知负荷的增加，对整体负荷及其心理和时间维度产生了显著影响。视觉导出的HRC-CWL输出在九名具有完整相关数据的参与者中，有七名与ECG导出的特征显著相关。同步互动事件进一步显示了检测到的犹豫与生理活动之间的时间对应关系。实时部署表明，该框架可以在不要求操作员佩戴额外传感器的情况下运行。这些发现支持HRC-CWL作为认知人机工程分析和自适应机器人辅助的可解释行为代理，而不是直接的心理生理负荷测量。

View on arXiv Download PDF AI Translation

cs.RO / 88 / 2609.15266

Distributed Safe Cooperative Vector Field for Trajectory Curvature Constrained Multi-Robot Systems

用于轨迹曲率约束的多机器人系统的分布式安全协作向量场

Xiao, Zhouru, Teng, Tao, Yao, Weijia, Zhang, Hui, Wang, Xiangke, Wang, Yaonan

Abstract

Trajectory curvature constraints are inherent in practical multi-robot systems due to the limited turning capabilities of the robots. Without properly accounting for these constraints, robots may fail to accomplish assigned tasks, and their trajectories may diverge from the intended paths. This paper proposes a distributed safe cooperative vector field approach for multi-robot systems subject to trajectory curvature constraints. The proposed approach is composed of a cooperative vector field and a safety-oriented collision avoidance vector field, aiming to address the problems of cooperative motion and safe collision avoidance in multi-robot path-following tasks. A safety-oriented collision avoidance vector field with adaptively adjustable reactive boundary is developed to accommodate the kinematic curvature constraints of robots, thereby ensuring the physical feasibility of collision avoidance maneuvers. The proposed vector field requires only a single virtual variable from each neighboring robot to achieve cooperative motion and ensure both obstacle avoidance and inter-robot collision avoidance. The effectiveness of the proposed approach is validated through both simulations and real-world experiments on an actual multi-robot platform.

Chinese Translation

由于机器人的转向能力有限，轨迹曲率约束在实际多机器人系统中是固有的。如果不正确考虑这些约束，机器人可能无法完成分配的任务，其轨迹可能会偏离预定路径。本文提出了一种针对受轨迹曲率约束的多机器人系统的分布式安全协作向量场方法。该方法由协作向量场和以安全为导向的避碰向量场组成，旨在解决多机器人路径跟踪任务中的协作运动和安全避碰问题。我们开发了一种具有自适应可调反应边界的安全导向避碰向量场，以适应机器人的运动曲率约束，从而确保避碰机动的物理可行性。所提出的向量场仅需每个邻近机器人提供一个虚拟变量，即可实现协作运动，并确保障碍物避让和机器人间的碰撞避免。通过仿真和实际多机器人平台的实验验证了所提方法的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 89 / 2609.15276

Low Clearance Hinge Joint Mechanism Based on 3D Printing on Sheet Fabrication Methodology

基于3D打印的低间隙铰接关节机制的片材制造方法

Jang, Jaehyung, Shin, Euibin, Okamura, Allison M., Ryu, Jee-Hwan

Abstract

This paper presents a low-clearance hinge joint mechanism based on the 3D printing on sheet fabrication method. This approach simplifies the fabrication of hinge mechanisms and overcomes limitations of conventional origami manufacturing by eliminating the need for adhesives commonly used during assembly, making it suitable for robots at the tens-of-centimeters scale. The advantages and disadvantages of three types of hinge joint mechanisms are compared, and a hinge joint that can be designed with low clearance for various facet thicknesses is selected. Based on the selected hinge joint, the twisting angle and bending force are analyzed, leading to the implementation of a clearance of 0.1 mm. Torsional resistance is experimentally evaluated to measure the torque required for twisting caused by plastic deformation and clearance. The results show that the torque associated with plastic deformation is sufficient to constrain the undesired degrees of freedom of the hinge joint, while the torque required for twisting due to clearance is minimal. Based on the analyzed data, the proposed hinge joint mechanism is applied to a 3-degree-of-freedom delta robot manipulator, demonstrating precise motion with low clearance.

Chinese Translation

本文提出了一种基于3D打印的低间隙铰接关节机制，采用片材制造方法。该方法简化了铰接机制的制造过程，并通过消除组装过程中常用的粘合剂，克服了传统折纸制造的局限性，使其适用于十厘米级别的机器人。比较了三种类型的铰接关节机制的优缺点，并选择了一种可以针对不同面厚度设计的低间隙铰接关节。基于所选的铰接关节，分析了扭转角度和弯曲力，最终实现了0.1毫米的间隙。通过实验评估扭转阻力，以测量因塑性变形和间隙引起的扭矩。结果表明，与塑性变形相关的扭矩足以约束铰接关节的不必要自由度，而因间隙引起的扭转所需扭矩则很小。基于分析数据，所提出的铰接关节机制被应用于一个3自由度的Delta机器人操控器，展示了低间隙下的精确运动。

View on arXiv Download PDF AI Translation

cs.RO / 90 / 2609.15322

Planning in the Backbone: DiffAdapterVLA for Native Continuous Trajectory Generation with Driving VLMs

骨干网络中的规划：DiffAdapterVLA用于具有驾驶视觉语言模型的原生连续轨迹生成

Lu, Changxin, Meng, Xiaoliang, Wu, Yu, Huang, Rui, Li, Honglin, Chen, Tao, Zhou, Kaixuan, Shao, Yadong

Abstract

Pretrained driving vision-language models (VLMs) integrate visual, route, language, and driving context into rich driving priors, yet their representation objectives remain separated from continuous driving planning. Existing methods typically begin trajectory generation only after the VLM has formed a final condition, leaving depth-wise condition computation outside the stepwise formation of trajectory state. We introduce DiffAdapterVLA, which realizes Planning in the Backbone: it injects explicit trajectory tokens into selected VLM late layers, bringing trajectory state into backbone forward computation, where it co-evolves with driving conditions at different depths. Lightweight layer-wise DiffAdapters organize this computation into recursive trajectory refinement, while asymmetric joint attention preserves directed guidance from the condition stream to trajectory planning. By placing planning within existing backbone computation rather than relying on an independent trajectory planner, DiffAdapterVLA adapts only lightweight trajectory modules to turn existing driving priors into efficient continuous planning capability. NAVSIM results show that it achieves high-quality closed-loop planning with low end-to-end latency using few trainable parameters, and demonstrate that jointly evolving trajectory state and depth-wise driving conditions in VLM late-layer computation effectively realizes continuous trajectory planning.

Chinese Translation

预训练的驾驶视觉语言模型（VLMs）将视觉、路线、语言和驾驶上下文整合为丰富的驾驶先验，但它们的表征目标与连续驾驶规划仍然分离。现有方法通常在VLM形成最终条件后才开始轨迹生成，导致深度条件计算在轨迹状态的逐步形成之外。我们提出了DiffAdapterVLA，它实现了骨干网络中的规划：将显式轨迹标记注入选定的VLM后期层，将轨迹状态引入骨干前向计算中，在不同深度与驾驶条件共同演化。轻量级的逐层DiffAdapters将这一计算组织为递归轨迹精炼，而不对称的联合注意力则保留了从条件流到轨迹规划的定向引导。通过将规划置于现有的骨干计算中，而不是依赖独立的轨迹规划器，DiffAdapterVLA仅适配轻量级轨迹模块，将现有驾驶先验转化为高效的连续规划能力。NAVSIM结果表明，它在使用少量可训练参数的情况下，实现了高质量的闭环规划，并且具有低端到端延迟，证明了在VLM后期层计算中共同演化轨迹状态和深度驾驶条件有效实现了连续轨迹规划。

View on arXiv Download PDF AI Translation

cs.RO / 91 / 2609.15352

Assistance Torque Estimation via Dynamics-Aware Optimization for Lower-Limb Exoskeleton in Complex Environments

基于动态感知优化的复杂环境下下肢外骨骼辅助扭矩估计

Liu, Xiao-Yin, Li, Guotao, Wang, Weiqun, Hou, Zeng-Guang

Abstract

Ground-truth human joint torque estimation relies on motion capture systems, which suffer from limited outdoor usability and significant deployment expenses. Furthermore, direct scaling of ground-truth joint torques to obtain motor torque commands is not necessarily the optimal strategy. To address the aforementioned limitations, inspired by the human motion generation process, this paper proposes a novel assistance torque estimation method based on the dynamic model. From an optimization perspective, the proposed method directly generates motor-assist torque and lowers the cost of data acquisition. Then, a data-driven assistance torque prediction network is trained to enable accurate real-time prediction under complex outdoor environments. Experimental results demonstrate that optimized (estimated) assistance torque exhibits better phase consistency with gait trajectories and better alignment with task characteristics. Relative to the Zero torque condition, the predicted torque can decrease metabolic rate by 11.8%-17.7%, heart rate by 8.9%-14.3%, and peak muscle activation levels by 28.2%-54.0%, respectively. This provides a new perspective for low-cost adaptive exoskeleton assistance.

Chinese Translation

真实的人体关节扭矩估计依赖于运动捕捉系统，但这些系统在户外使用时受到限制，并且部署成本高昂。此外，直接将真实关节扭矩缩放以获得电机扭矩指令并不一定是最佳策略。为了解决上述限制，本文受到人类运动生成过程的启发，提出了一种基于动态模型的新型辅助扭矩估计方法。从优化的角度来看，该方法直接生成电机辅助扭矩，并降低数据获取的成本。随后，训练了一个数据驱动的辅助扭矩预测网络，以实现复杂户外环境下的准确实时预测。实验结果表明，优化后的（估计的）辅助扭矩与步态轨迹表现出更好的相位一致性，并与任务特征更好地对齐。与零扭矩条件相比，预测的扭矩可以分别降低11.8%-17.7%的代谢率、8.9%-14.3%的心率和28.2%-54.0%的峰值肌肉激活水平。这为低成本自适应外骨骼辅助提供了新的视角。

View on arXiv Download PDF AI Translation

cs.RO / 92 / 2609.15362

Understanding User Preferences of a Slope-Aware Variable-Admittance Filter for a Robot Guide Dog

理解坡度感知变导纳滤波器在机器人导盲犬中的用户偏好

Esposito, Federico, Selvaggio, Mario, Link, Aaron, Ruggiero, Fabio

Abstract

This letter investigates how the parameters of a slope-aware variable-admittance filter influence user preferences in force-based interaction with a robotic guide dog for visually impaired individuals. The proposed system consists of a quadruped robot equipped with a sensor-free rigid handle for physical guidance. The framework combines path following, momentum-based interaction-wrench estimation, and a variable-admittance filter whose stiffness and damping are adapted online from slope information extracted by the robot's depth camera. The adaptation policies are evaluated through high-fidelity simulations and a human-subject study involving blindfolded sighted participants. Multiple strategies are compared using a Taguchi L9 design of experiments. Preliminary main-effect results suggest that increasing stiffness uphill and decreasing it downhill improves both objective and subjective metrics, whereas damping shows no significant main effect.

Chinese Translation

本信件研究了坡度感知变导纳滤波器的参数如何影响用户在与视觉障碍人士的机器人导盲犬进行基于力的交互中的偏好。所提出的系统由一只四足机器人组成，该机器人配备了一个无传感器的刚性手柄用于物理引导。该框架结合了路径跟踪、基于动量的交互扭矩估计以及一个变导纳滤波器，其刚度和阻尼根据机器人深度摄像头提取的坡度信息在线调整。适应策略通过高保真模拟和涉及蒙眼的视力正常参与者的人体实验研究进行评估。使用田口L9实验设计比较了多种策略。初步的主效应结果表明，在上坡时增加刚度和在下坡时减少刚度能够改善客观和主观指标，而阻尼则没有显著的主效应。

View on arXiv Download PDF AI Translation

cs.RO / 93 / 2609.15382

From Prediction to Decision: World-Model-Guided Action Selection for Continuous Pile Excavation

从预测到决策：基于世界模型的连续桩挖掘行动选择

Zhang, Ailing, Gao, Fan, Zhang, Song, Leong, Kawa, Wu, Ziyu, Wang, Yafei

Abstract

Wheel-loader excavation is a sequential decision problem in which every scoop changes the terrain available to subsequent actions. A practical world model must predict action consequences accurately, rank candidates in real time, and operate inside the closed loop of a full-size machine. We present the World-Action Model (WAM), which proposes multiple scoops, rejects geometrically inadmissible candidates, jointly predicts signed terrain change and loaded volume, executes the candidate with the largest predicted load, and replans from the newly observed terrain. On 32 geometry-disjoint MinSlope test episodes, adding world-model ranking to matched diffusion proposals reduces the mean scoop count from 651.8 to 540.6 (17.1%), preserves 32/32 completion, and improves every paired episode. In a complete-system comparison, WAM completes 32/32 episodes versus 29/32 for an independently trained soft actor-critic policy. Comparisons of input representations, spatial support, and five architectures identify an accurate and efficient physics-structured predictor. We further evaluate the interface on event-disjoint full-size-loader data and deploy the complete perception-proposal-prediction-selection-execution loop for autonomous excavation. The ROS2/TensorRT implementation processes five candidates in 72.4 ms on a Jetson AGX Orin. The simulation results establish decision-level gains, while the physical experiments demonstrate real-world closed-loop feasibility.

Chinese Translation

轮式装载机挖掘是一个序列决策问题，每次铲斗的操作都会改变后续操作可用的地形。一个实用的世界模型必须准确预测行动后果，实时对候选方案进行排序，并在全尺寸机器的闭环中运行。我们提出了世界行动模型（World-Action Model, WAM），该模型提出多个铲斗方案，拒绝几何上不可接受的候选方案，联合预测地形变化的符号和装载体积，执行预测装载量最大的候选方案，并从新观察到的地形重新规划。在32个几何不相交的MinSlope测试回合中，将世界模型排序添加到匹配的扩散提案中，使平均铲斗次数从651.8减少到540.6（减少17.1%），完成率保持在32/32，并且每个配对回合均有所改善。在完整系统比较中，WAM完成32/32回合，而独立训练的软演员-评论家策略完成29/32回合。对输入表示、空间支持和五种架构的比较确定了一个准确且高效的物理结构预测器。我们进一步在事件不相交的全尺寸装载机数据上评估接口，并部署完整的感知-提案-预测-选择-执行循环以实现自主挖掘。ROS2/TensorRT实现能够在Jetson AGX Orin上以72.4毫秒处理五个候选方案。模拟结果确立了决策层面的收益，而物理实验则展示了现实世界闭环的可行性。

View on arXiv Download PDF AI Translation

cs.RO / 94 / 2609.15399

Dynamics-Informed Reinforcement Learning for Agile and Energy-Efficient Locomotion of a Monopedal Hopping Quadcopter

基于动力学的强化学习用于单足跳跃四旋翼的灵活与节能运动

Chen, Ruigang, Zhang, Qi, Zhong, Zhicheng, Yun, Zhuorui, Or, Yizhar, Liu, Mingyi

Abstract

Although aerial-legged robots offer combined agility and efficiency, controlling high-speed hopping under complex hybrid dynamics is challenging. Reinforcement Learning (RL) is promising but prone to energy-inefficient "reward hacking". We propose a Dynamics-Informed RL framework for a monopedal hopping quadcopter. By embedding a target Specific Energy into the reward, we constrain the optimization to a physically viable energy manifold, ensuring stable hopping behaviour. By rewarding the phase-consistent behavior, it can encourage bio-inspired stance-phase impulse. Furthermore, penalizing the electro-mechanical power waste induces the motors generate an efficient impulse. This enables the policy to inject energy strictly during spring restitution without heuristic state machines. MuJoCo simulations validate robust height regulation and forward velocity tracking up to 2.0 m/s despite severe attitude-contact coupling. Ultimately, our approach yields a highly agile hopping gait, reducing energy consumption by 82% and 73% compared to hovering baselines and inefficiency baseline, respectively.

Chinese Translation

尽管空中腿机器人提供了灵活性和效率的结合，但在复杂的混合动力学下控制高速跳跃仍然具有挑战性。强化学习（Reinforcement Learning, RL）虽然前景广阔，但容易出现能量效率低下的“奖励黑客”现象。我们提出了一种用于单足跳跃四旋翼的基于动力学的强化学习框架。通过将目标特定能量嵌入奖励中，我们将优化约束在一个物理可行的能量流形上，从而确保稳定的跳跃行为。通过奖励相位一致的行为，可以鼓励仿生的支撑相冲击。此外，惩罚电机机械功率浪费促使电机产生高效的冲击。这使得策略能够在弹簧恢复期间严格注入能量，而无需启发式状态机。MuJoCo模拟验证了在严重的姿态接触耦合下，能够实现稳健的高度调节和高达2.0 m/s的前向速度跟踪。最终，我们的方法产生了一种高度灵活的跳跃步态，与悬停基线和低效基线相比，能量消耗分别减少了82%和73%。

View on arXiv Download PDF AI Translation

cs.RO / 95 / 2609.15447

Learning to Exploit Passive Dynamics for Energy-Efficient Target Hopping of a Spring-Legged Quadcopter

学习利用被动动力学实现春腿四旋翼的节能目标跳跃

Chen, Ruigang, Zhang, Qi, Zhong, Zhicheng, Yun, Zhuorui, Or, Yizhar, Liu, Mingyi

Abstract

Combining aerial thrust with spring-loaded hopping makes monopedal quadcopters promising for locomotion over complex terrain, but heuristic proportional-integral-derivative (PID) tuning limits coordination between active thrust and passive contact dynamics. We present a direct estimated-state-to-motor Proximal Policy Optimization (PPO) policy that commands four motors without an explicit hopping state machine or low-level attitude PID. Its reward combines Energy-Manifold Shaping for mass-normalized vertical-energy tracking and apex-state anchoring with Efficiency Shaping, which uses a history-aware power estimator to penalize general power use, impose an additional airborne-power cost, and penalize airborne near-stationarity. In representative hardware runs, the PPO-based control stack reduced cycle-averaged measured electrical power by 30.7% and mean total normalized thrust by 49.8% relative to the tuned PID-based control stack, while retaining repeatable commanded-height hopping and more concentrated landings. These observations are consistent with improved use of passive dynamics and reduced measured electrical demand.

Chinese Translation

将空中推力与弹簧驱动的跳跃相结合，使单足四旋翼在复杂地形上的运动具有良好的前景，但启发式的比例-积分-微分（PID）调节限制了主动推力与被动接触动力学之间的协调。我们提出了一种直接估计状态到电机的近端策略优化（PPO）策略，该策略在没有显式跳跃状态机或低级姿态PID的情况下控制四个电机。其奖励结合了能量流形塑形（Energy-Manifold Shaping），用于质量归一化的垂直能量跟踪和顶点状态锚定，以及效率塑形（Efficiency Shaping），后者使用历史感知的功率估计器来惩罚一般功率使用，施加额外的空中功率成本，并惩罚空中近静止状态。在代表性的硬件运行中，基于PPO的控制系统将周期平均测得的电功率降低了30.7%，将平均总归一化推力降低了49.8%，相较于调优的基于PID的控制系统，同时保持了可重复的指令高度跳跃和更集中的着陆。这些观察结果与被动动力学的改善利用和测得电力需求的减少一致。

View on arXiv Download PDF AI Translation

cs.RO / 96 / 2609.15455

InterSocialBench: Benchmarking Human and LLM Preferences for Companion-Robot Social Behavior

InterSocialBench：评估人类与大型语言模型对伴侣机器人社交行为的偏好

Xu, Yaodan, Guo, Boyang, Gu, Yuqing, Zhang, Qingxin, Deng, Yiwen, Liu, Meng, Li, Lintian

Abstract

Companion robots face everyday situations in which several feasible behaviors may be appropriate, yet different people prefer different responses. We introduce InterSocialBench, a benchmark of 210 domestic scenarios and 18 high-level behaviors, pairing judgments from 100 human participants with 23,520 responses from seven large language models under 16 personality conditions. Each human annotation preserves a preferred action alongside explicitly appropriate and inappropriate candidates. A structured construction pipeline covers behavioral alternatives, competing situational cues, and relevant history and future tasks. Evaluation distinguishes preferred-choice agreement from explicit rejection, using scenario-grouped splits for trainable predictors. Simple frequency and persona-voting baselines illustrate these objectives. Across the tested prompts, model and human behavior distributions differ, and the diversity gap remains after matching response counts: humans exhibit 4.68 distinct choices per scenario, compared with 2.06--3.46 for the models. Human scenario-level plurality agreement is 51.5%, describing disagreement rather than a universal prediction ceiling. InterSocialBench supports evaluating social behavior selection without replacing individual judgments with a single consensus label.

Chinese Translation

伴侣机器人在日常情境中面临多种可行的行为选择，而不同的人对不同的反应有不同的偏好。我们提出了InterSocialBench，这是一个包含210个家庭场景和18种高层次行为的基准，结合了100名参与者的判断与在16种个性条件下来自七个大型语言模型的23,520个反应。每个参与者的标注保留了一个首选行为以及明确适当和不适当的候选行为。一个结构化的构建流程涵盖了行为替代选项、竞争的情境线索以及相关的历史和未来任务。评估区分了首选选择的一致性与明确拒绝，使用按场景分组的拆分来训练可预测的模型。简单的频率和个性投票基线展示了这些目标。在测试的提示中，模型和人类的行为分布存在差异，且在匹配响应数量后多样性差距依然存在：人类在每个场景中表现出4.68种不同的选择，而模型则为2.06至3.46。人类在场景级别的多数一致性为51.5%，这表明存在分歧，而非普遍的预测上限。InterSocialBench支持评估社交行为选择，而不需要用单一共识标签替代个体判断。

View on arXiv Download PDF AI Translation

cs.RO / 97 / 2609.15475

P-POSEMEM: Projective Semantic Memory for Consistent Language Grounding under Pose-Graph Rewrites

P-POSEMEM：在姿态图重写下实现一致语言基础的投影语义记忆

Sier, Ha, Salmasi, Ali, Xu, Mengya, Zhang, Haizhou, Lu, Jie, Zou, Zhuo, Yu, Xianjia, Westerlund, Tomi

Abstract

A robot following language instructions needs its semantic memory to keep naming the same physical object while the SLAM pose graph underneath is optimized, loop-closed and compressed. Maps committing each detection to a world coordinate cannot: a closure moves the anchor it was measured from, or the solver marginalizes that anchor, and the query then selects a different object although both graphs represent the same posterior. P-POSEMEM stores each observation as an immutable event at its birth keyframe, retains the Bayes-tree elimination conditional of every marginalized keyframe, and integrates the semantic likelihood over the reconstructed joint posterior of poses, anchors and identities. Dproj, the total-variation defect between the language-goal distributions of inference-equivalent full and marginalized graphs, measures this directly. Over 40 HM3DSem scenes and 112,000 queries, P-POSEMEM reproduces the full-graph oracle (Dproj = 0) and reduces goal flips against every memory-reducing baseline. On an eight-run campaign whose 761 closures rewrote the map by up to 47 m, Dproj stays below 10^-13 with 0/288 goal flips when elimination follows the closures, where every ablation and a coordinate committed at insertion flip goals it does not; under a live bounded solver the same memory flips 23/288 against 53 for that frozen coordinate. A pre-registered negative control is detected by Dproj while leaving calibration error and navigation success unchanged, indicating that these measures capture distinct failure modes. Retrieval is held fixed by a shared frozen detector, isolating the gain to memory consistency. Code and data: https://anonymous.4open.science/r/posemem-2328/.

Chinese Translation

跟随语言指令的机器人需要其语义记忆在优化、闭环和压缩的SLAM姿态图下保持对同一物理对象的命名。每个检测与世界坐标相结合的地图无法做到这一点：闭环会移动其测量的锚点，或者求解器会对该锚点进行边缘化，导致查询选择不同的对象，尽管两个图表示相同的后验。P-POSEMEM将每个观察存储为在其出生关键帧时不可变的事件，保留每个被边缘化关键帧的贝叶斯树消除条件，并在重建的姿态、锚点和身份的联合后验上整合语义似然性。Dproj，即推理等效的完整图和边缘化图的语言目标分布之间的总变差缺陷，直接测量这一点。在超过40个HM3DSem场景和112,000个查询中，P-POSEMEM重现了完整图的oracle（Dproj = 0），并减少了与每个内存减少基线相比的目标翻转。在一次包含761个闭环的八次实验中，地图重写最多47米，Dproj保持在10^-13以下，且在消除跟随闭环时没有目标翻转（0/288），而每次消融和在插入时承诺的坐标则翻转了目标；在一个实时有界求解器下，相同的内存翻转了23/288，而冻结坐标则翻转了53。Dproj检测到一个预注册的负对照，同时保持校准误差和导航成功不变，表明这些测量捕捉到不同的失败模式。检索由共享的冻结检测器保持固定，从而将增益孤立于内存一致性。代码和数据：https://anonymous.4open.science/r/posemem-2328/

View on arXiv Download PDF AI Translation

cs.RO / 98 / 2609.15509

StereoPatch: Patch-Aligned RGB-Depth Fusion for Spatial Perception in Robot Manipulation

StereoPatch：用于机器人操作的补丁对齐RGB-深度融合以增强空间感知

Zhou, Yanan, Qian, Zhaoyan, Zhao, James, Zhi, Weiming

Abstract

Recent advances in robot imitation learning have produced visuomotor policies that predict actions directly from visual observations. Yet visually similar scenes can require different actions as target position, object height, or contact geometry changes. Pretrained RGB features may map these geometrically distinct states to similar policy inputs, while simply adding depth requires the policy to learn RGB-depth correspondence from the same limited demonstrations used to learn control. We introduce StereoPatch, a patch-aligned RGB-depth representation that binds registered metric geometry directly to the RGB patches used for action prediction. On a shared 2-D patch grid, asymmetric cross-attention incorporates depth information into the corresponding RGB features before action decoding. The resulting StereoPatch Tokens provide a geometry-aware visual representation that can condition general visuomotor policies without changing their underlying learning objectives. Across six real-robot tasks, StereoPatch achieves higher closed-loop success than appearance-only, geometry-only, raw RGB-D, and late-fusion baselines. Additional experiments across three simulation suites evaluate compatibility across visuomotor policy architectures, spatial generalization, and operating limits. Results suggest that resolving control-relevant geometric ambiguity benefits from aligning depth directly with the visual features used for action prediction, rather than supplying it as an independent modality. Project page: https://aus.bot/research/stereopatch/.

Chinese Translation

近年来，机器人模仿学习的进展产生了能够直接从视觉观察中预测动作的视觉运动策略。然而，视觉上相似的场景可能因目标位置、物体高度或接触几何形状的变化而需要不同的动作。预训练的RGB特征可能将这些几何上不同的状态映射到相似的策略输入，而仅仅添加深度信息则要求策略从用于学习控制的有限演示中学习RGB-深度对应关系。我们提出了StereoPatch，这是一种补丁对齐的RGB-深度表示，它将注册的度量几何直接绑定到用于动作预测的RGB补丁上。在共享的二维补丁网格上，不对称的交叉注意力在动作解码之前将深度信息融入到相应的RGB特征中。生成的StereoPatch Tokens提供了一种几何感知的视觉表示，可以在不改变其基本学习目标的情况下调节一般的视觉运动策略。在六个真实机器人任务中，StereoPatch的闭环成功率高于仅基于外观、仅基于几何、原始RGB-D和后期融合的基线。对三个仿真套件的额外实验评估了视觉运动策略架构的兼容性、空间泛化能力和操作极限。结果表明，解决与控制相关的几何模糊性受益于将深度信息直接与用于动作预测的视觉特征对齐，而不是将其作为独立的模态提供。项目页面：https://aus.bot/research/stereopatch/

View on arXiv Download PDF AI Translation

cs.RO / 99 / 2609.15570

DIDO: Distilling Interaction-Centric Dynamics into One-Step Denoising for World Action Models

DIDO：将以交互为中心的动态提炼为世界动作模型的一步去噪

Lyu, Jing, Bai, Shuanghao, Xiao, Runze, Liao, Zhenyu, Tan, Wenxing, Tang, Zihan, Shi, Ruochuan, Peng, Cheng, Ji, Yuheng, Wang, Yihao, Chen, Badong, Wang, Pengwei, Wang, Zhongyuan, Zhao, Xiaoguang

Abstract

World Action Models (WAMs) use video generation models to predict future visual dynamics for robotic manipulation, but iterative denoising introduces additional latency for closed-loop control. We empirically find that visual content converges at different rates during denoising. Static background structure forms early, whereas the gripper and manipulated object remain blurry after the first step, with their interaction dynamics emerging only through subsequent denoising. Consequently, naively truncating a multi-step video model to one step preserves scene structure but loses the interaction-centric dynamics most critical for manipulation. To address this issue, we propose DIDO, which distills the converged dynamics of a multi-step video model into a single denoising step. DIDO combines distribution matching distillation with interaction-centric representation guidance. Beyond compressing multi-step generation into one forward pass, DIDO explicitly models the gripper, manipulated object, and their interaction using supervised bounding-box visual reasoning tokens. Additionally, DIDO aligns the target object's representations across multiple model layers with features from a pretrained DINOv3 encoder. This interaction-centric guidance helps the distilled model preserve both the relevant entities and their future dynamics in a single step, while substantially reducing inference latency. DIDO achieves an average success rate of 99.0\% on LIBERO, 76.6\% on LIBERO-Plus, and 92.0\% on RoboTwin, while also demonstrating effective transfer to long-horizon and generalization tasks in real-world robotic manipulation.

Chinese Translation

世界动作模型（WAMs）利用视频生成模型预测机器人操作的未来视觉动态，但迭代去噪为闭环控制引入了额外的延迟。我们实证发现，在去噪过程中，视觉内容以不同的速度收敛。静态背景结构较早形成，而夹具和被操作物体在第一步后仍然模糊，其交互动态仅通过后续去噪显现。因此，简单地将多步视频模型截断为一步虽然保留了场景结构，但却失去了对操作至关重要的以交互为中心的动态。为了解决这个问题，我们提出了DIDO，它将多步视频模型收敛的动态提炼为单一步去噪。DIDO结合了分布匹配蒸馏与以交互为中心的表示指导。除了将多步生成压缩为一次前向传递外，DIDO还使用监督的边界框视觉推理标记显式建模夹具、被操作物体及其交互。此外，DIDO将目标物体的表示与预训练的DINOv3编码器的特征在多个模型层中对齐。这种以交互为中心的指导帮助提炼模型在单一步骤中保留相关实体及其未来动态，同时显著减少推理延迟。DIDO在LIBERO上实现了99.0\%的平均成功率，在LIBERO-Plus上为76.6\%，在RoboTwin上为92.0\%，同时在现实世界的机器人操作中也展示了有效的长远任务转移和泛化能力。

View on arXiv Download PDF AI Translation

cs.RO / 100 / 2609.15587

An Information-Space Perspective to Scene Graph Sufficiency for Robotic Task Planning

从信息空间视角看机器人任务规划中的场景图充分性

Sakçak, Başak, Verdoja, Francesco

Abstract

Planning in complex environments requires task specifications grounded in representations that capture objects, relations, and affordances; scene graphs meet this need, but their size in large environments hinders efficient planning. While task-aware pruning and hierarchical abstractions have been explored, a general, task-centric formalization of what constitutes a sufficient scene graph for planning remains open. This paper provides such a formalization by modeling planning over scene graphs within an information-spaces framework through the definition of scene graph transition systems and relevant action semantics for navigation and manipulation. We then introduce derived scene graphs via information mappings that merge and prune nodes and induce quotient transition systems augmented with motion primitives to capture higher-level actions over merged graph nodes. Sufficiency is characterized by two conditions: (i) the information mapping yields a deterministic quotient, and (ii) the task is well-posed over derived traces, ensuring plans found on the derived model are feasible on the maximal system. We illustrate the framework using a task over an example environment, showing both sufficient and insufficient reduced scene graphs.

Chinese Translation

在复杂环境中进行规划需要基于能够捕捉对象、关系和可供性（affordances）的表示来进行任务规范；场景图满足了这一需求，但在大型环境中，其规模阻碍了高效规划。尽管已经探索了基于任务的剪枝和层次抽象，但关于什么构成足够的场景图以用于规划的通用、以任务为中心的形式化仍然是一个未解的问题。本文通过在信息空间框架内对场景图进行建模，定义场景图转换系统及相关的导航和操作语义，提供了这样的形式化。然后，我们通过信息映射引入派生场景图，这些映射合并和剪枝节点，并诱导出增强运动原语的商转换系统，以捕捉合并图节点上的高级动作。充分性由两个条件来表征：（i）信息映射产生一个确定性的商，以及（ii）任务在派生轨迹上是良定义的，确保在派生模型上找到的规划在最大系统上是可行的。我们通过一个示例环境中的任务来说明该框架，展示了充分和不充分的简化场景图。

View on arXiv Download PDF AI Translation

cs.RO / 101 / 2609.15631

Flow-Matched Motion Priors: Online Optimal-Transport Rewards for Imitation Learning

流匹配运动先验：用于模仿学习的在线最优传输奖励

Zou, Yilin, Liu, Chenghua, Wu, Chenglong, Jiang, Fanghua

Abstract

Learning a motion prior requires a reward that guides a policy from its current behavior toward demonstrated motion. Adversarial Motion Priors (AMP) provide such a reward with a discriminator. However, adversarial objectives can become uninformative when policy and expert supports are far apart. A naive use of optimal transport (OT) averages matched expert successors into a barycentric target. Averaging across gait phases can weaken the target's joint motion. We introduce Flow-Matched Motion Priors (FMP), an online scalar reward learned from paths connecting current rollout histories to an expert motion bank. Entropic OT supplies the coupling. Before each policy update, we train a neural potential with flow matching (FM) along the rollout-to-expert paths, endpoint-gradient supervision, and relative-value calibration. The actor receives only physical observations and the reward remains a scalar, as in AMP. Controlled reward-model experiments show substantially better generalization beyond the fitting rollout than value-only or endpoint-only fitting. On Unitree G1, matched 50-million-transition experiments compare FMP with AMP, a barycentric OT reward, and nested ablations under demonstration and fixed-pose initialization. FMP produces stable forward walking at 0.727 m/s from demonstration resets and 0.338 m/s from a fixed default pose. In the fixed-pose condition, it incurs 129 falls versus 243 for the endpoint-only control. Against a static score-gradient teacher, dynamic FM reduces score-increment error at interpolation fractions 0.25 and 0.50 while using 29% less offline fitting time.

Chinese Translation

学习运动先验需要一种奖励，指导策略从当前行为向示范运动转变。对抗性运动先验（Adversarial Motion Priors, AMP）通过判别器提供了这样的奖励。然而，当策略和专家的支持相距较远时，对抗性目标可能变得不具信息性。简单使用最优传输（Optimal Transport, OT）将匹配的专家后继者平均为一个重心目标。在步态阶段之间的平均可能会削弱目标的联合运动。我们引入了流匹配运动先验（Flow-Matched Motion Priors, FMP），这是一种在线标量奖励，从连接当前展开历史与专家运动库的路径中学习而来。熵最优传输提供了耦合。在每次策略更新之前，我们沿着展开到专家的路径训练一个具有流匹配（Flow Matching, FM）的神经潜力，使用端点梯度监督和相对价值校准。演员仅接收物理观察，奖励仍然是标量，类似于AMP。受控的奖励模型实验显示，FMP在超出拟合展开的情况下，具有显著更好的泛化能力，相较于仅基于价值或仅基于端点的拟合。在Unitree G1上，匹配的5000万过渡实验比较了FMP与AMP、重心OT奖励以及在示范和固定姿态初始化下的嵌套消融。FMP在示范重置后以0.727 m/s的速度稳定前行，从固定默认姿态出发则为0.338 m/s。在固定姿态条件下，FMP发生了129次跌倒，而仅基于端点的控制则为243次。与静态评分梯度教师相比，动态FM在插值比例为0.25和0.50时减少了评分增量误差，同时使用了29%的离线拟合时间。

View on arXiv Download PDF AI Translation

cs.RO / 102 / 2609.15667

Tracking the Ground: Online Lidar Identification of Robot-Induced Soil Deformation in Agricultural Environments

跟踪地面：在线激光雷达识别农业环境中机器人引起的土壤变形

Montagnon, Tom, Laconte, Johann, Thuilot, Benoit, Cho, Wonjae, Lenain, Roland

Abstract

Agriculture faces many challenges, and robotic systems can play an important role in addressing them by improving the efficiency and sustainability of field operations. Among these challenges, preserving soil health is a critical concern, as vehicle-soil interactions can degrade the soil structure and produce unwanted surface deformation. A key step toward soil-aware robotics is to explicitly account for how vehicle traffic deforms the ground, yet soil state is typically not treated as a variable. We address this gap by proposing a framework to quantify traffic-induced soil deformation and estimate its evolution online from lidar observations. The method relies on a reduced-order parametric model that represents the soil behavior via physically interpretable parameters, yielding a continuously updated and observable representation of soil state. Experiments conducted in different soil conditions demonstrate the ability of the approach to capture deformation induced by the robot. By making soil response measurable and interpretable during operation, the proposed framework establishes a basis for soil-aware robotic operation, in which the estimated state can be exploited to adapt robotic behaviors in order to reduce soil degradation.

Chinese Translation

农业面临许多挑战，而机器人系统可以通过提高田间作业的效率和可持续性在应对这些挑战中发挥重要作用。在这些挑战中，保护土壤健康是一个关键问题，因为车辆与土壤的相互作用可能会破坏土壤结构并产生不必要的表面变形。实现土壤感知机器人技术的关键一步是明确考虑车辆交通如何导致地面变形，然而土壤状态通常未被视为一个变量。我们通过提出一个框架来量化交通引起的土壤变形，并从激光雷达观测中在线估计其演变，从而填补这一空白。该方法依赖于一个降阶参数模型，通过物理可解释的参数来表示土壤行为，从而生成一个持续更新且可观察的土壤状态表示。在不同土壤条件下进行的实验展示了该方法捕捉机器人引起的变形的能力。通过使土壤响应在操作过程中可测量和可解释，所提出的框架为土壤感知的机器人操作奠定了基础，其中估计的状态可以被利用来调整机器人行为，以减少土壤退化。

View on arXiv Download PDF AI Translation

cs.RO / 103 / 2609.15680

Volumetric Harmonic Field Navigation for Quadrotors

四旋翼的体积谐波场导航

Jia, Shuxiu, Mukherjee, Amartya, Yuan, Yating, Liu, Jun

Abstract

Quadrotor navigation in cluttered 3-D environments requires global guidance while local motion remains subject to collision and motion limits. Harmonic potentials provide dense guidance from a global boundary value problem, but coupling a volumetric harmonic field to constrained physical quadrotor motion remains an open experimental problem. We couple a precomputed volumetric harmonic field with a constrained predictive planner that queries the field at predicted positions instead of extracting a global reference path. In Structured 3-D tests, harmonic guidance yields larger minimum clearance and lower RMS jerk than matched Dijkstra guidance, at the cost of longer paths; the same pattern remains when both methods use the same passage. Long maze tests span routes far beyond one prediction horizon, and Crazyflie trials validate physical execution. To the best of our knowledge, this is the first physical quadrotor demonstration of volumetric harmonic field navigation. The results show that globally constructed harmonic guidance can directly support local constrained motion generation on a physical quadrotor.

Chinese Translation

在复杂的三维环境中，四旋翼导航需要全球指导，同时局部运动受到碰撞和运动限制的约束。谐波势能通过全球边界值问题提供密集的指导，但将体积谐波场与受限的物理四旋翼运动结合仍然是一个未解决的实验问题。我们将预计算的体积谐波场与一个受限的预测规划器相结合，该规划器在预测位置查询场而不是提取全局参考路径。在结构化的三维测试中，谐波指导的最小间隙更大，均方根抖动（RMS jerk）更低，相比于匹配的Dijkstra指导，尽管路径更长；当两种方法使用相同通道时，这一模式依然存在。长迷宫测试跨越了远超一个预测视野的路线，而Crazyflie实验验证了物理执行。根据我们所知，这是首次在物理四旋翼上演示体积谐波场导航。结果表明，全球构建的谐波指导可以直接支持物理四旋翼上的局部受限运动生成。

View on arXiv Download PDF AI Translation

cs.RO / 104 / 2609.15716

Continuous Manifold-Decomposed Impedance Retargeting for Contact-Rich Imitation Learning

面向接触丰富模仿学习的连续流形分解阻抗重定向

Liu, Jiahao, Kawaharazuka, Kento, Makabe, Tasuku, Okada, Kei

Abstract

CMDIR extends Manifold-Decomposed Impedance Retargeting (MDIR) to transform fixed-impedance demonstrations into continuous variable-impedance controllers, which can also serve as structured supervision for imitation learning. Continuous Task-Manifold Impedance Representation (TMIR) pairs an evolving task frame with controller instructions. Demo-relative Compromise dynamics retain moving-basis transport and control/physical metric mismatch, yielding displacement, reaction-impulse, and perturbation-sensitivity criteria. Quality-to-Fast automatically compiles a solver structure from development paths within a predefined finite space, re-instantiates that structure for each demonstration, and certifies the resulting candidate by multi-resolution evaluation. Across 225 retargeted-controller trials in three real contact tasks, full CMDIR improves mean task-proxy retention and reduces mean pose deviation, force fluctuation, and peak force relative to discrete MDIR. FastMPO achieves a $5.8$--$9.4\times$ speedup over C-MPO with comparable closed-loop outcomes. Downstream experiments demonstrate learnability of the complete TMIR supervision interface; lower force fluctuation and peak force are observed among successful executions, while completion reliability remains uneven across tasks and environments.

Chinese Translation

CMDIR 扩展了流形分解阻抗重定向 (MDIR)，将固定阻抗的示范转化为连续可变阻抗控制器，这些控制器也可以作为模仿学习的结构化监督。连续任务流形阻抗表示 (TMIR) 将一个不断演变的任务框架与控制器指令相结合。相对演示妥协动态保留了移动基准传输和控制/物理度量的不匹配，从而产生位移、反应冲击和扰动敏感性标准。质量到快速 (Quality-to-Fast) 自动从预定义有限空间内的开发路径编译求解器结构，为每个示范重新实例化该结构，并通过多分辨率评估认证所得到的候选者。在三项真实接触任务中进行的 225 次重定向控制器试验中，完整的 CMDIR 提高了平均任务代理保留率，并减少了平均姿态偏差、力波动和峰值力，相较于离散的 MDIR。FastMPO 在闭环结果相当的情况下，相较于 C-MPO 实现了 $5.8$--$9.4 imes$ 的加速。下游实验表明完整的 TMIR 监督接口具有可学习性；在成功执行中观察到较低的力波动和峰值力，而在不同任务和环境中完成的可靠性仍然不均匀。

View on arXiv Download PDF AI Translation

cs.RO / 105 / 2609.15726

Bench2Dex: Benchmarking Visuo-Tactile Bimanual Dexterous Manipulation Across Dexterous Hands

Bench2Dex：跨灵巧手的视觉-触觉双手灵巧操作基准测试

Yang, Zhenjie, Zhang, Yideng, Zhang, Dongjie, Jiang, Chenyu, Liu, Xianshuai, Li, Yufeng, Ge, Zuhao, Jiao, Xingyu, Zhang, Zheng, He, Kaiyu, Wang, He, Zhong, Yuwen, Deng, Yi, Jiang, Muyun, Huang, Xianliang, Su, Haisheng, Zhang, Donghang, Zhang, Jian, Yang, Xue, Li, Hongyang, Wu, Zuxuan, Jiang, Yu-Gang, Jia, Xiaosong, Yan, Junchi

Abstract

Tactile sensing provides contact information that can be difficult to infer from vision alone, but tactile hardware for dexterous hands has not converged to a common design. Dexterous hands differ in finger structure, contact surfaces, and sensor layouts, while simulated tactile signals still differ from measurements produced by physical sensors. These factors make it difficult to study visuo-tactile manipulation across diverse dexterous hands within a consistent experimental setting. We present Bench2Dex, a simulation benchmark for visuo-tactile bimanual manipulation across 12 dexterous hands. We adapt existing robot models with a shared simulated tactile interface that converts local contact geometry into image-like tactile observations. The interface provides a consistent observation format across different hand morphologies without attempting to reproduce the output of a specific physical tactile sensor. Bench2Dex includes 26 bimanual manipulation tasks that involve tool use, articulated-object interaction, and multi-stage manipulation, together with about 1.3K human-teleoperated demonstrations. The benchmark provides synchronized visual, tactile, proprioceptive, action, and object-state observations, together with executable task metrics. For robustness, we group seven perturbation types into invariance axis, where the correct action does not change, and equivariance axis, where the correct action changes together with the perturbation. We evaluate ACT, Diffusion Policy, pi0.5, and GR00T N1.5 on Bench2Dex and report their performance and failure modes. Bench2Dex is meant as a platform for studying visuo-tactile learning across dexterous hands. It does not assume that simulated tactile observations can replace real tactile sensing; it offers a shared setting for algorithm development while tactile hardware and simulation models are still evolving.

Chinese Translation

触觉感知提供了接触信息，这些信息仅通过视觉很难推断，但灵巧手的触觉硬件尚未趋向于统一设计。灵巧手在手指结构、接触表面和传感器布局上存在差异，而模拟的触觉信号与物理传感器产生的测量结果仍然不同。这些因素使得在一致的实验环境中研究不同灵巧手之间的视觉-触觉操作变得困难。我们提出了Bench2Dex，这是一个针对12种灵巧手的视觉-触觉双手操作的仿真基准测试。我们对现有的机器人模型进行了调整，采用共享的模拟触觉接口，将局部接触几何形状转换为类似图像的触觉观测。该接口在不同手型之间提供了一致的观测格式，而不试图重现特定物理触觉传感器的输出。Bench2Dex包含26个双手操作任务，涉及工具使用、关节物体交互和多阶段操作，以及约1300个由人类遥控的演示。该基准测试提供了同步的视觉、触觉、本体感知、动作和物体状态观测，以及可执行的任务指标。为了增强鲁棒性，我们将七种扰动类型分组为不变轴，其中正确动作不变，以及等变轴，其中正确动作随扰动而变化。我们在Bench2Dex上评估了ACT、Diffusion Policy、pi0.5和GR00T N1.5，并报告了它们的性能和失败模式。Bench2Dex旨在作为研究灵巧手视觉-触觉学习的平台。它并不假设模拟触觉观测可以替代真实的触觉感知；它为算法开发提供了一个共享环境，同时触觉硬件和仿真模型仍在不断发展。

View on arXiv Download PDF AI Translation

cs.RO / 106 / 2609.15770

JEPLO: Joint-Embedding Predictive Learning for LiDAR-Based Legged Locomotion

JEPLO：基于激光雷达的腿部运动联合嵌入预测学习

Yuan, Qihao, Qiu, Yixuan, Cao, Ziyu, Cao, Ming, Li, Kailai

Abstract

Light detection and ranging (LiDAR) remains less explored than RGB-D sensing for perceptive legged locomotion, and existing LiDAR-based approaches often rely on explicit mapping. We present JEPLO (Joint-Embedding Predictive learning for legged LOcomotion), a single-stage learning framework for mapping-free, LiDAR-based perceptive locomotion for legged robots. We introduce a proprio-exteroceptive JEPA (PE-JEPA) world model to learn predictive egocentric terrain representations from onboard observations, including raw LiDAR scans. A concurrent JEPA-teacher-student (CJTS) pipeline is further proposed to train a locomotion policy informed by JEPA latent representations in simulation using deep reinforcement learning with a simple reward formulation. The framework achieves successful sim-to-real transfer, enabling omnidirectional traversal of diverse terrains, including long staircases and high boxes, with lightweight onboard computation. Evaluations demonstrate greater robustness than existing perceptive locomotion frameworks, particularly under degraded perception caused by occlusion, sparsity and noise. Further analysis validates JEPLO's ability to retain task-relevant information under these challenging conditions. We open-source our implementation, experimental datasets, and hardware setup designs https://github.com/ASIG-X/JEPLO.

Chinese Translation

激光雷达（LiDAR）在感知腿部运动方面的研究仍然不如RGB-D传感器广泛，现有的基于激光雷达的方法通常依赖于显式映射。我们提出了JEPLO（Joint-Embedding Predictive learning for legged LOcomotion），这是一个无映射的基于激光雷达的腿部机器人感知运动的单阶段学习框架。我们引入了一种自我-外部感知的JEPA（PE-JEPA）世界模型，从机载观测（包括原始激光雷达扫描）中学习预测的自我中心地形表示。此外，我们进一步提出了一个并行的JEPA教师-学生（CJTS）管道，利用深度强化学习和简单的奖励公式，在仿真中训练一个由JEPA潜在表示指导的运动策略。该框架实现了成功的仿真到现实转移，使得机器人能够在多样化地形上进行全方位的移动，包括长楼梯和高箱子，同时保持轻量级的机载计算。评估结果表明，该框架在感知运动的鲁棒性方面优于现有方法，尤其是在由于遮挡、稀疏和噪声导致的感知退化情况下。进一步分析验证了JEPLO在这些挑战条件下保持与任务相关信息的能力。我们已开源我们的实现、实验数据集和硬件设计，网址为https://github.com/ASIG-X/JEPLO。

View on arXiv Download PDF AI Translation

cs.RO / 107 / 2609.15840

Uncertainty-Guided Sparse Refinement for Action Chunking Transformer Policies

基于不确定性引导的稀疏细化用于动作分块变换器策略

Wang, Chenyang, Wang, Yuntian, Yang, Xiaoxiong, Jiang, Dingde, Liu, Siao, Liu, Yang

Abstract

Learning chunk-based visuomotor policies for long-horizon robot manipulation remains challenging. Recent action-chunking methods have shown promising performance by predicting temporally extended action sequences. However, their failures are often dominated by prediction errors at a small number of critical timesteps rather than uniformly poor predictions across the entire action chunk, making uniform refinement inefficient and insufficiently targeted. To address this bottleneck, we propose Uncertainty-Guided Refinement (UGR), a sparse refinement framework for chunk-based visuomotor policies. Specifically, UGR follows a coarse-to-refine design: it first predicts a full action chunk, estimates per-step temporal uncertainty from the coarse hidden states, and applies residual correction only to the most uncertain timesteps selected by a binary mask. The uncertainty branch is decoupled from the coarse action predictor, enabling clean attribution of the refinement gains to uncertainty-guided correction rather than additional predictor capacity. Extensive experiments on five dual-arm manipulation tasks from the RoboTwin benchmark show that UGR achieves the best success rate on four tasks, improves over the ACT baseline by up to 13% absolute, and outperforms both full-chunk and position-agnostic block refinement in ablation studies.

Chinese Translation

学习基于分块的视觉运动策略以实现长时间范围的机器人操作仍然具有挑战性。最近的动作分块方法通过预测时间延续的动作序列显示出良好的性能。然而，它们的失败往往是由于在少数关键时间步的预测错误主导，而不是在整个动作分块中均匀的预测不良，这使得均匀细化效率低下且目标不够明确。为了解决这一瓶颈，我们提出了不确定性引导细化（Uncertainty-Guided Refinement, UGR），这是一个针对基于分块的视觉运动策略的稀疏细化框架。具体而言，UGR遵循粗到细的设计：它首先预测一个完整的动作分块，从粗略的隐藏状态中估计每一步的时间不确定性，并仅对通过二进制掩码选择的最不确定时间步应用残差修正。不确定性分支与粗略动作预测器解耦，使得细化收益的归因能够清晰地归于不确定性引导的修正，而不是额外的预测器能力。在RoboTwin基准的五个双臂操作任务上的大量实验表明，UGR在四个任务上实现了最佳成功率，相较于ACT基线提高了最多13%的绝对值，并在消融研究中优于全分块和位置无关的块细化。

View on arXiv Download PDF AI Translation

cs.RO / 108 / 2609.15861

DuctAM: A Duct-Assisted Quadrotor-Based Aerial Manipulator Enabling High-Force Push-and-Pull Interactions

DuctAM：一种基于导管辅助的四旋翼空中操控器，实现高强度的推拉交互

Wang, Yi, Jin, Rui, Xu, Xinhang, Jin, Haotian, Liu, Ruiyang, Yang, Yizhuo, Xie, Lihua

Abstract

Uncrewed Aerial Manipulators (UAMs) extend the capabilities of Uncrewed Aerial Vehicles (UAVs) from perception to physical interaction. Among various aerial interactions, push-and-pull operations are fundamental manipulation primitives that require sustained horizontal forces while maintaining stable flight. In this paper, we propose DuctAM, a compact aerial manipulation platform that enhances horizontal force capability for push-and-pull interactions using two ducted fans integrated along the quadrotor interaction axis. An attitude-force decoupled control scheme enables controllable horizontal forces without requiring large attitude changes. Extensive real-world experiments are conducted to validate the DuctAM. Figure-eight trajectory tracking experiments demonstrate stable flight and accurate motion control in both quad and duct modes. Force-measurement experiments quantify the decoupled longitudinal force capability of DuctAM. Finally, representative push-and-pull interaction tasks, including cart pushing, door closing, and drawer opening, verify the practical effectiveness of DuctAM. The results show that DuctAM achieves significantly improved horizontal interaction force capability while maintaining stable flight compared with conventional UAVs.

Chinese Translation

无人机操控器（UAMs）扩展了无人机（UAVs）从感知到物理交互的能力。在各种空中交互中，推拉操作是基本的操控原语，需要在保持稳定飞行的同时施加持续的水平力。本文提出了DuctAM，一种紧凑的空中操控平台，通过在四旋翼交互轴上集成两个导管风扇，增强了推拉交互的水平力能力。采用姿态-力解耦控制方案，使得在不需要大幅度姿态变化的情况下实现可控的水平力。我们进行了大量的实际实验以验证DuctAM的性能。八字轨迹跟踪实验展示了在四旋翼模式和导管模式下的稳定飞行和精确运动控制。力测量实验量化了DuctAM的解耦纵向力能力。最后，代表性的推拉交互任务，包括推车、关门和开抽屉，验证了DuctAM的实际有效性。结果表明，与传统无人机相比，DuctAM在保持稳定飞行的同时显著提高了水平交互力能力。

View on arXiv Download PDF AI Translation

cs.RO / 109 / 2609.15870

WLA$^3$: World Latent Action Modeling for Semantics, Dynamics, and Kinematics

WLA$^3$: 世界潜在动作建模用于语义、动态和运动学

Liu, Peidong, Xiang, Zhiyuan, Li, Mingyang, Li, Wenhao, Zhang, Jiale, Sun, Jiahao, Li, Jiawei

Abstract

Scaling generalist policy models with heterogeneous data is limited by the lack of unified, low-noise action supervision. Human egocentric videos are abundant, but only a small fraction comes with high-quality hand-action labels. Observed world transitions offer a common source of action-related supervision across data sources. We introduce WLA$^3$ (World Latent Action Modeling for Semantics, Dynamics, and Kinematics), a unified generalist policy model framework built around representations learned by a World Latent Action Model (WLAM). WLAM first learns how multimodal world states change over a local interval, encoding synchronized camera views and available embodiment-state changes into a compact local latent action and a richer transition feature. Reconstruction from partial modalities and consistency across overlapping windows encourage robust transition representations. WLA$^3$ reuses them across semantics, dynamics, and kinematics: local latent actions support action-sensitive physical-dynamics modeling, segment-level features directly supervise the VLM through a Semantic Latent Aggregate (SLA), and an action expert jointly predicts latent actions together with embodiment-specific robot controls. Human videos provide scalable transition supervision, while robot trajectories ground the shared representation in executable native controls. On LARYBench, the final 32D latent action reaches 67.89\% average classification accuracy. WLA$^3$ achieves 81.9% average success across six real-robot tasks versus 66.2% for $\pi_{0.5}$. Performance improves as generalist policy model mid-training data scales, and human videos support human-to-robot transfer. Project page can be found at https://wla-3.github.io/.

Chinese Translation

使用异构数据扩展通用策略模型受到统一、低噪声动作监督缺乏的限制。人类自我中心的视频丰富，但只有一小部分配有高质量的手部动作标签。观察到的世界转变提供了跨数据源的动作相关监督的共同来源。我们提出了WLA$^3$（世界潜在动作建模用于语义、动态和运动学），这是一个围绕世界潜在动作模型（WLAM）学习的表示构建的统一通用策略模型框架。WLAM首先学习多模态世界状态在局部时间间隔内的变化，将同步的摄像机视图和可用的体现状态变化编码为紧凑的局部潜在动作和更丰富的转变特征。从部分模态的重建和重叠窗口间的一致性鼓励了稳健的转变表示。WLA$^3$在语义、动态和运动学之间重用这些表示：局部潜在动作支持对动作敏感的物理动态建模，段级特征通过语义潜在聚合（SLA）直接监督视觉语言模型（VLM），而动作专家则共同预测潜在动作以及体现特定的机器人控制。人类视频提供了可扩展的转变监督，而机器人轨迹则将共享表示与可执行的原生控制相结合。在LARYBench上，最终的32维潜在动作达到了67.89%的平均分类准确率。WLA$^3$在六个真实机器人任务中实现了81.9%的平均成功率，而$ ext{π}_{0.5}$为66.2%。随着通用策略模型中训练数据的扩展，性能得到了提升，人类视频支持人机转移。项目页面可在https://wla-3.github.io/找到。

View on arXiv Download PDF AI Translation

cs.RO / 110 / 2609.15895

Goal-Oriented Communications for Physical AI: Design and Testbed

面向目标的物理人工智能通信：设计与测试平台

Chen, Shutong, Zhang, Wenkai, Aijaz, Adnan, Guo, Miao, Deng, Yansha

Abstract

Physical AI relies on frequently-updated, latency-sensitive video stream to perceive, reason, and interact with the physical world, resulting in strict latency requirements with much higher data volumes that existing 5G networks cannot support. Goal-oriented communication (GoC) offers as a promising approach to solve this challenge by transmitting only task-relevant semantic representations. However, existing GoC frameworks were mainly evaluated in the simulations while their effectiveness has never been validated in a practical deployment of physical AI application. In this work, we develop an end-to-end GoC testbed for Physical AI, which connects a PiPER robot arm equipped with an RGB-D camera and a 5G modem to an NVIDIA Jetson AGX Orin edge server through a 5G OpenAirInterface network. We propose and implement three GoC frameworks that transmit 3D bounding boxes, 2D scene graphs, and 3D scene graphs, as three types of semantic representations, respectively. They share the common functional modules designed for closed-loop Physical AI applications, including semantic extraction, full stack 5G transmission, language model inference, digital twin validation, and robotic control. Extensive experiments on our testbed show that our GoC frameworks reduce the task completion time by up to 52.6% and improve task success probability by up to 45%, compared to the traditional framework that periodically transmits the raw image data. These results validate the practical effectiveness of our GoC framework and pave the way for efficient and reliable Physical AI applications over future 6G networks. Project website: https://sites.google.com/view/goc-physical-ai-testbed.

Chinese Translation

物理人工智能依赖于频繁更新、对延迟敏感的视频流来感知、推理和与物理世界互动，这导致了严格的延迟要求和远高于现有5G网络所能支持的数据量。面向目标的通信（Goal-oriented communication, GoC）作为一种有前景的方法，通过仅传输与任务相关的语义表示来解决这一挑战。然而，现有的GoC框架主要在模拟环境中进行评估，其有效性从未在物理人工智能应用的实际部署中得到验证。在本研究中，我们开发了一个端到端的GoC测试平台，连接了一个配备RGB-D相机和5G调制解调器的PiPER机器人手臂与一个通过5G OpenAirInterface网络连接的NVIDIA Jetson AGX Orin边缘服务器。我们提出并实现了三种GoC框架，分别传输3D边界框、2D场景图和3D场景图，作为三种类型的语义表示。它们共享为闭环物理人工智能应用设计的公共功能模块，包括语义提取、全栈5G传输、语言模型推理、数字双胞胎验证和机器人控制。在我们的测试平台上进行的大量实验表明，与传统框架（定期传输原始图像数据）相比，我们的GoC框架将任务完成时间缩短了多达52.6%，并将任务成功概率提高了多达45%。这些结果验证了我们GoC框架的实际有效性，并为未来6G网络上高效可靠的物理人工智能应用铺平了道路。项目网站：https://sites.google.com/view/goc-physical-ai-testbed。

View on arXiv Download PDF AI Translation

cs.RO / 111 / 2609.15910

SlipSense: Multimodal Tactile Learning for Low-Latency and Generalized Slip Detection

SlipSense：用于低延迟和通用滑动检测的多模态触觉学习

Jian, Tong, Kumar, Aditya Thurvas Senthil, Li, Xinyi, Chen, Ziling, Dai, Tianyu, Sengul, Ali, Grimaldi, Matteo, Lu, Wenjie, Nabi, Saleh, Yu, Tao

Abstract

Slip detection is fundamental to dexterous manipulation, yet existing systems often lack precise characterization of detection latency and cross-platform generalization. We present SlipSense, a multimodal tactile slip-detection framework built on TacV5, a compact sensor integrating a $32 \times 32$ piezoresistive array operating at 240 Hz and a 3-axis MEMS accelerometer operating at 8 kHz. The piezoresistive array captures spatial pressure distributions, while the accelerometer captures friction-induced vibrations, providing complementary slip cues. The framework performs modality-specific encoding, intra-sensor fusion, and cross-modal attention with causal temporal prediction at 240 Hz. Experiments on a dataset of 1.4 million frames spanning 37 objects demonstrate the complementarity of the two modalities. SlipSense achieves 96.7% Macro F1 with a false-positive rate below 1.6%, detecting 76% of slip events within 23.1 ms. When trained solely on UMI data, SlipSense generalizes zero-shot to a Tesollo dexterous hand, transferring across unseen objects, distinct sensor units, and robotic platforms without retraining.

Chinese Translation

滑动检测是灵巧操作的基础，然而现有系统往往缺乏对检测延迟的精确表征和跨平台的泛化能力。我们提出了SlipSense，一个基于TacV5的多模态触觉滑动检测框架，TacV5是一种集成了32×32压阻阵列（piezoresistive array）和以240 Hz频率工作的3轴MEMS加速度计（MEMS accelerometer）的紧凑型传感器，后者以8 kHz频率工作。压阻阵列捕捉空间压力分布，而加速度计捕捉摩擦引起的振动，提供互补的滑动线索。该框架执行特定模态编码、传感器内部融合和跨模态注意力机制，并以240 Hz的频率进行因果时间预测。在一个涵盖37个物体的140万帧数据集上的实验表明了这两种模态的互补性。SlipSense在假阳性率低于1.6%的情况下实现了96.7%的宏观F1得分（Macro F1），并在23.1毫秒内检测到76%的滑动事件。当仅在UMI数据上训练时，SlipSense能够零样本泛化到Tesollo灵巧手，能够在未见过的物体、不同的传感器单元和机器人平台之间进行迁移，而无需重新训练。

View on arXiv Download PDF AI Translation

cs.RO / 112 / 2609.15921

Touch2Trace: Tactile-Driven Imitation Learning for Dexterous Cable Tracing

Touch2Trace：基于触觉驱动的灵巧电缆追踪模仿学习

Grimaldi, Matteo, Klee, David, Chen, Ziling, Jian, Tong, Lee, Wonju, Lu, Wenjie, Yu, Tao, Nabi, Saleh

Abstract

Dexterous manipulation of deformable objects demands continuous fingertip-level regulation of pressure, friction, and incipient slip. We study one of the most challenging cases: dexterous cable tracing, feeding a cable through the hand with repeated pinch-and-curl motions of the thumb and index finger. We introduce Touch2Trace, a tactile-driven imitation-learning system for this task, and provide, to our knowledge, the first systematic real-world characterization of how encoder pretraining, control rate, temporal context, and spatial resolution each shape policy performance. The winning learning recipe combines a tactile encoder pretrained for a custom 32 x 32 piezoresistive sensor (TacV5) via self-supervised learning with a lightweight transformer policy trained on teleoperated demonstrations via behavior cloning, deployed at 60 Hz on a Tesollo DG-5F hand. Tactile feedback without vision or explicit cable-state estimation significantly improves tracing performance versus a proprioception-only baseline: from 0.2 cm to 20.1 cm mean distance and 0% to 93% success rate, with zero-shot transfer to unseen cables and routing conditions. The results quantify the influence of key parameters in tactile-driven systems for reliable dexterous deformable object manipulation.

Chinese Translation

对可变形物体的灵巧操控要求对压力、摩擦和初始滑动进行持续的指尖级调节。我们研究了其中一个最具挑战性的案例：灵巧电缆追踪，通过拇指和食指的反复夹持和卷曲动作将电缆送入手中。我们提出了Touch2Trace，这是一种针对该任务的基于触觉驱动的模仿学习系统，并提供了我们所知的首个系统性实世界特征描述，探讨了编码器预训练、控制速率、时间上下文和空间分辨率如何影响策略性能。成功的学习方案结合了通过自监督学习对定制的32 x 32压电电阻传感器（TacV5）进行预训练的触觉编码器，以及通过行为克隆在远程操作演示上训练的轻量级变换器策略，部署在Tesollo DG-5F手上，频率为60 Hz。在没有视觉或明确电缆状态估计的情况下，触觉反馈显著提高了追踪性能，相较于仅依赖本体感知的基线，平均距离从0.2 cm提高到20.1 cm，成功率从0%提升至93%，并且在未见过的电缆和布线条件下实现了零样本迁移。结果量化了触觉驱动系统中关键参数对可靠的灵巧可变形物体操控的影响。

View on arXiv Download PDF AI Translation

cs.RO / 113 / 2609.15940

Beyond Single-Axis Testing: Paired Evaluation of Compound Robustness in Vision-Language-Action Policies

超越单轴测试：视觉-语言-行动策略复合鲁棒性的配对评估

Sawada, Hiroki, Kasahara, Shunichi

Abstract

Vision-language-action policies are typically evaluated one perturbation at a time, providing a useful diagnosis of their sensitivity to individual distribution shifts. Real-world deployment, however, may involve several shifts simultaneously, and it remains unclear how these individual robustness measurements compose. We ask whether compound robustness can be inferred from single-axis evaluations. We introduce LIBERO-CTRL, a six-axis benchmark that pairs each initial state across single-axis conditions and a matched simultaneous condition. This design reveals two opposing outcome changes that aggregate success rates cannot distinguish: emergent failures, where all single-axis rollouts succeed but the simultaneous rollout fails, and compensated successes, where at least one single-axis rollout fails but the simultaneous rollout succeeds. Because one transition decreases compound success while the other increases it, they can cancel, making aggregate compound performance appear consistent with single-axis measurements even when individual outcomes differ substantially. These opposing transitions can largely cancel in aggregate: even when the difference between the two transition rates is not statistically distinguishable from zero, as many as 29.0% of matched initial states still change outcome. Across six policies and three severity levels, such outcome changes reach 34.5% in the most affected condition. The relative prevalence of the two transitions varies across policies and severities, while the transition rates remain similar under independent re-evaluation of stochastic policies. Compound robustness therefore cannot be characterized from aggregate single-axis success rates alone; matched per-instance evaluation is needed to reveal how joint perturbations alter behavior.

Chinese Translation

视觉-语言-行动策略通常是逐个扰动进行评估，这为其对单一分布变化的敏感性提供了有用的诊断。然而，实际部署可能涉及多个扰动同时发生，目前尚不清楚这些单独的鲁棒性测量如何组合。我们探讨复合鲁棒性是否可以从单轴评估中推断。我们引入了LIBERO-CTRL，这是一个六轴基准，配对每个初始状态在单轴条件和相应的同时条件下进行评估。这种设计揭示了两种相反的结果变化，聚合成功率无法区分：突现失败，即所有单轴滚动成功但同时滚动失败，以及补偿成功，即至少一个单轴滚动失败但同时滚动成功。由于一个转变降低了复合成功率，而另一个则提高了它们可能会相互抵消，使得聚合的复合表现看似与单轴测量一致，即使单个结果存在显著差异。这些相反的转变在聚合中可以大幅抵消：即使两个转变率之间的差异在统计上不可区分于零，仍有多达29.0%的匹配初始状态改变了结果。在六个策略和三个严重性水平中，这种结果变化在受影响最严重的条件下达到34.5%。这两种转变的相对普遍性在不同策略和严重性之间有所不同，而在对随机策略的独立重新评估中，转变率保持相似。因此，复合鲁棒性不能仅通过聚合的单轴成功率来表征；需要匹配的逐实例评估以揭示联合扰动如何改变行为。

View on arXiv Download PDF AI Translation

cs.RO / 114 / 2609.15976

MessyMem: Learning-from-Doing Memory for Mobile Manipulation

MessyMem：用于移动操作的经验学习记忆

Banwasi, Anuva, Muckelroy III, William, Sundaresan, Priya, Zhao, Linfeng, Bohg, Jeannette, Ho, Cherie

Abstract

Mobile manipulators deployed across many rooms and visits should improve with experience: after discovering that a cabinet is locked or finding an object in a drawer, the robot should reuse that knowledge rather than start each task from scratch. Yet today's robots often treat each task as new: compact scene representations omit interaction-derived knowledge, raw video histories are difficult to query, and VLM planners reason at inference time without persistently updating what the robot knows. We present MessyMem, a persistent memory system that enables mobile manipulators to learn from experience and reuse that knowledge across future tasks. It maintains a spatially grounded 3D scene graph of objects and locations, augments it with properties and outcomes learned through interaction, and links visual observations for fine-grained recall. We evaluate MessyMem in simulation and on a real mobile manipulator. In a continuous 25-task simulation spanning over 3 hours, MessyMem achieves 80.0% task progress, outperforming the strongest ablation by 14.8 percentage points and the strongest external baseline by 28.9 points, while retrieving task-relevant evidence from thousands of stored keyframes and over an hour into the past.

Chinese Translation

在多个房间和访问中部署的移动操控器应随着经验的积累而不断改进：在发现某个柜子被锁或在抽屉中找到某个物体后，机器人应当重用这些知识，而不是每次任务都从头开始。然而，现今的机器人往往将每个任务视为新的任务：紧凑的场景表示省略了交互所获得的知识，原始视频历史难以查询，而视觉语言模型（VLM）规划器在推理时并未持续更新机器人所掌握的知识。我们提出了MessyMem，一个持久的记忆系统，使移动操控器能够从经验中学习并在未来的任务中重用这些知识。它维护一个空间基础的三维场景图，包含物体和位置，并通过交互学习增强其属性和结果，同时链接视觉观察以实现细粒度的回忆。我们在仿真和真实移动操控器上评估了MessyMem。在一个持续的25任务仿真中，历时超过3小时，MessyMem实现了80.0%的任务进展，超越了最强消融实验14.8个百分点和最强外部基线28.9个百分点，同时从数千个存储的关键帧中检索与任务相关的证据，回溯时间超过一小时。

View on arXiv Download PDF AI Translation

cs.RO / 115 / 2609.15988

ResSafe: Learning Safety Filtering with Residual Reinforcement Learning for Humanoids

ResSafe：基于残差强化学习的人形机器人安全过滤学习

Qu, Gechen, Zhang, Tong, Zhang, Bike, Wang, Yen-Jen, Sreenath, Koushil, Tomlin, Claire, Choi, Jason Jangho

Abstract

Safe control of humanoid robots remains challenging due to their high-dimensional dynamics, contact-rich interactions, and sensitivity to disturbances. Although reinforcement learning has enabled effective locomotion and motion tracking, learned policies can still generate unsafe actions that lead to instability or falls. In this work, we propose residual reinforcement learning as an implicit safety-filtering mechanism for safe humanoid control. Instead of relying on a single nominal policy to simultaneously balance performance, safety, and robustness, we decouple performance and safety. The nominal policy focuses solely on task performance, while a residual policy learns safety corrections. This decoupling leads to a better performance--safety Pareto trade-off and avoids the need for careful tuning of multiple competing reward terms within a single policy training. We show that the residual policy can act as an implicit safety filter.

Chinese Translation

人形机器人的安全控制仍然面临挑战，主要由于其高维动态、丰富的接触交互以及对干扰的敏感性。尽管强化学习已经实现了有效的运动和轨迹跟踪，但学习到的策略仍可能产生不安全的动作，从而导致不稳定或跌倒。在本研究中，我们提出了残差强化学习作为一种隐式安全过滤机制，以实现安全的人形控制。我们不再依赖单一的名义策略来同时平衡性能、安全性和鲁棒性，而是将性能与安全性解耦。名义策略仅关注任务性能，而残差策略则学习安全修正。这种解耦实现了更好的性能与安全的帕累托权衡，并避免了在单一策略训练中对多个竞争奖励项进行仔细调优的需求。我们展示了残差策略可以作为隐式安全过滤器。

View on arXiv Download PDF AI Translation

计算机视觉 (Computer Vision)

195

cs.CV / 1 / 2609.13190

Personalized and Explainable Blood Pressure Estimation from PPG via Hybrid CNN--Morphological Features

基于混合卷积神经网络与形态特征的个性化可解释血压估计

Yi, Myung-Kyu, Lee, Jongshill, Lee, Jeyeon, Kim, In Young

Abstract

Continuous cuffless blood pressure (BP) monitoring using photoplethysmography (PPG) offers a promising solution for personalized healthcare. However, existing methods have two major limitations. Handcrafted feature-based approaches rely on precise fiducial point detection and are limited to short-term analysis, while deep learning models, despite their accuracy, often operate as black boxes with limited physiological interpretability. To address these challenges, we propose a physiology-guided hybrid framework for personalized BP estimation that couples a convolutional neural network (CNN) branch capturing global and local waveform dynamics with a morphology-prior branch that explicitly encodes person-specific vascular characteristics. By embedding a morphology-based feature set that explicitly encodes individual vascular characteristics, the proposed framework enhances personalization and reduces dependence on large-scale training datasets. Evaluated on a subset of the MIMIC-III database under a subject-specific (personalized) testing protocol, the proposed personalized physiology-guided hybrid approach achieved mean absolute errors (MAEs) of 3.77 mmHg for systolic BP and 2.36 mmHg for diastolic BP, corresponding to relative improvements of 43.7% and 32.4% over a subject-specific (personalized) CNN-only baseline. SHAP-based analysis confirmed that the introduced morphology-prior features align with individual vascular characteristics, reinforcing per-subject interpretability. These findings highlight the potential of personalized, physiology-guided hybrid learning with novel morphological descriptors for accurate and explainable BP monitoring in real-world settings.

Chinese Translation

利用光电容积描记法（PPG）进行连续无袖带血压（BP）监测为个性化医疗提供了一种有前景的解决方案。然而，现有方法存在两个主要局限性。基于手工特征的方法依赖于精确的基准点检测，并且仅限于短期分析，而深度学习模型尽管准确，但通常作为黑箱操作，缺乏生理可解释性。为了解决这些挑战，我们提出了一种生理指导的混合框架，用于个性化血压估计，该框架结合了一个捕捉全局和局部波形动态的卷积神经网络（CNN）分支与一个明确编码个体特定血管特征的形态优先分支。通过嵌入明确编码个体血管特征的基于形态的特征集，所提框架增强了个性化程度，并减少了对大规模训练数据集的依赖。在MIMIC-III数据库的一个子集上，采用个体特定（个性化）测试协议进行评估，所提出的个性化生理指导混合方法在收缩压和舒张压方面的平均绝对误差（MAE）分别为3.77 mmHg和2.36 mmHg，分别对应于相对于个体特定（个性化）CNN基线的43.7%和32.4%的相对改善。基于SHAP的分析确认引入的形态优先特征与个体血管特征一致，增强了每个受试者的可解释性。这些发现突显了个性化、生理指导的混合学习与新型形态描述符在真实环境中实现准确和可解释的血压监测的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 2 / 2609.13225

Part Grounding, Not Action Knowledge: Locating the Bottleneck in VLM Affordance Prediction

部分定位，而非动作知识：定位视觉语言模型（VLM）可供性预测中的瓶颈

Sattigeri, Sarthak

Abstract

Benchmarks agree that vision-language models reason poorly about low-level manipulation, but an aggregate accuracy score does not say which step fails. We separate two steps that affordance questions conflate: identifying which part of an object to act on, and knowing what action that part requires. Across 19 articulated objects we asked eight models, spanning three developers, what motion a robot should apply. Under an open prompt, push was produced once in 64 evaluations where it was correct, despite being correct for 8 of 19 objects and appearing in the offered label set every time. Inspecting the outputs showed why: models described a different part than the one being scored, e.g. explaining how to pick up a camera rather than press its button. Naming the target part raises action accuracy by 0.32 to 0.63 for every model, from 0.158-0.474 to 0.684-0.947, and push recall from 0-1/8 to 7-8/8. No model beats a constant answer that ignores the image under the open prompt; once the part is named, all eight do. Asked to describe the same part in free prose with no label set, models produce pressing language for 6 to 8 of 8. These results are hard to reconcile with missing action knowledge, and instead point to part grounding as the dominant bottleneck, a pattern that holds across all three model families and does not diminish with capability. Naming the part supplies the grounding variable, so this bounds what a perfect part detector would offer rather than demonstrating a general model of mechanics. Two supporting results agree: on real photographs only three of eight models localize grasp points better than a constant baseline, and on rendered objects none do. We also document two measurement errors of our own, a threshold that let a constant baseline score 0.929 and a labelling rule wrong on 4 of 19 objects, both caught only by testing our numbers against trivial alternatives.

Chinese Translation

基准测试一致认为视觉语言模型在低级操作推理方面表现不佳，但整体准确率并未指出具体失败的步骤。我们将可供性问题混淆的两个步骤分开：识别要作用于对象的哪个部分，以及了解该部分需要什么动作。在19个关节对象上，我们询问了来自三家开发者的八个模型，机器人应该施加什么动作。在开放提示下，推送（push）在64次评估中仅出现一次，尽管对于19个对象中的8个是正确的，并且每次都出现在提供的标签集中。检查输出结果显示了原因：模型描述了与评分不同的部分，例如解释如何拿起相机而不是按下其按钮。命名目标部分使每个模型的动作准确率提高了0.32到0.63，从0.158-0.474提高到0.684-0.947，推送的召回率从0-1/8提高到7-8/8。没有模型能够超越一个在开放提示下忽略图像的常量答案；一旦命名了部分，所有八个模型都能做到。在没有标签集的自由散文中被要求描述同一部分时，模型对8个对象中的6到8个产生了按压语言。这些结果难以与缺乏动作知识相调和，而是指向部分定位作为主要瓶颈，这一模式在所有三种模型家族中都成立，并且随着能力的提升并未减弱。命名部分提供了定位变量，因此这限制了一个完美部分检测器所能提供的内容，而不是展示一个通用的力学模型。两个支持结果一致：在真实照片上，只有三分之八的模型的抓取点定位优于常量基线，而在渲染对象上则没有模型做到。我们还记录了两个测量错误，一个阈值使常量基线得分为0.929，另一个标签规则在19个对象中错误标记了4个，这两个错误仅通过将我们的数字与简单替代方案进行测试才被发现。

View on arXiv Download PDF AI Translation

cs.CV / 3 / 2609.13226

Synthetic Leprosy Image Generation Using Mask-Conditioned Latent Diffusion and Transfer Learning from Large Chronic Wound Datasets

基于掩膜条件的潜在扩散和大规模慢性伤口数据集的迁移学习合成麻风图像生成

Abdulkadir, Yusuf

Abstract

Machine learning for neglected tropical diseases is limited by data, not algorithms: public annotated image sets for leprosy (Hansen's disease) number in the hundreds, orders of magnitude below what generative models require. We ask whether a model trained on abundant chronic wound photography transfers to this low-data regime. We build a three-stage pipeline. First, a DeepLabV3-ResNet50 segmentation network (validation Dice 0.876, IoU 0.799) supplies lesion masks for two wound datasets that ship without them. Second, we assemble a mask-conditioned latent diffusion model from Stable Diffusion 1.5 components and train it on 3,280 region-of-interest wound crops, widening the UNet input convolution from 4 to 11 channels to admit three mask feature maps and a blurred low-frequency context latent. Third, we fine-tune this model on 708 leprosy image-mask pairs drawn from 764 images of approximately 150 patients. We evaluate with LPIPS perceptual distance, anchored by a real-versus-real baseline computed on the same 242 anchor images as the cross-set comparisons; without that reference the cross-set distances cannot be interpreted. The generated set shows no mode collapse: its internal perceptual diversity (0.662) is statistically indistinguishable from that of the real leprosy set (0.672, 95% CI [0.664, 0.680]). Generated images sit 0.044 LPIPS outside the real distribution - measurably apart, but under half of one standard deviation. Fine-tuning shifted the output distribution only marginally, which we trace to lesion geometry reaching the network through input concatenation alone. Chronic wound photography is therefore a viable donor domain for leprosy lesion synthesis: low-level appearance transfers well, and the remaining barrier is semantic control rather than image quality.

Chinese Translation

针对被忽视的热带疾病的机器学习受到数据的限制，而非算法的限制：麻风（汉森病）的公共标注图像集仅有数百个，远低于生成模型所需的数量。我们探讨一个在丰富的慢性伤口摄影数据上训练的模型是否能够迁移到这一低数据环境中。我们构建了一个三阶段的流程。首先，使用DeepLabV3-ResNet50分割网络（验证Dice 0.876，IoU 0.799）为两个没有提供伤口掩膜的数据集生成病变掩膜。其次，我们从Stable Diffusion 1.5组件组装了一个掩膜条件的潜在扩散模型，并在3,280个感兴趣区域的伤口裁剪图上进行训练，将UNet输入卷积的通道数从4扩展到11，以容纳三个掩膜特征图和一个模糊的低频上下文潜变量。第三，我们在从764张大约150名患者的图像中提取的708对麻风图像-掩膜对上对该模型进行了微调。我们使用LPIPS感知距离进行评估，并以在同一242张锚图像上计算的真实与真实基线作为锚点进行交叉集比较；没有该参考，交叉集距离无法解释。生成的图像集没有出现模式崩溃：其内部感知多样性（0.662）在统计上与真实麻风图像集（0.672，95% CI [0.664, 0.680]）无显著差异。生成的图像在真实分布之外的LPIPS距离为0.044——虽然可测量，但不足一个标准差的一半。微调仅对输出分布产生了微小的偏移，我们将其归因于病变几何形状仅通过输入拼接进入网络。因此，慢性伤口摄影是麻风病变合成的可行捐赠领域：低级外观能够良好迁移，而剩下的障碍是语义控制而非图像质量。

View on arXiv Download PDF AI Translation

cs.CV / 4 / 2609.13228

Don't Just Look, Intervene: Perturbation Based Region Labeling for VQA Images

不仅仅是观察，进行干预：基于扰动的区域标注用于视觉问答图像

Jojic, Marko, Li, Zhaonan, Zhou, Ben

Abstract

Vision Language Models (VLMs) should rely on visual evidence that directly determines the correct answer, but supervision for grounding visual reasoning is often expensive to obtain manually or tied to dataset-specific annotation primitives. We instead introduce model-causal visual evidence as an annotation target, defined as the set of image regions whose counterfactual intervention changes a model's answer distribution for a given image-question pair. Based on this principle, we introduce Counterfactual Search for Grounding Regions (CSGR). CSGR is a scalable pipeline that proposes candidate regions, perturbs them, measures their effect on answer sensitivity, and aggregates this evidence across multiple judges to approximate answer-critical regions in VQA data. To assess whether CSGR annotations contain a useful supervision signal, we plug them into three existing grounding-aware training routines: attention steering, Visual CoTfinetuning, and latent visual reasoning. These experiments test whether the proposed annotation scheme can provide a useful supervision signal across multiple ways of consuming region labels, rather than introducing a new way of using them. Across competing automatic region-labeling mechanisms, CSGR annotations provide the most consistent gains over Cross Entropy-only finetuning in both in-domain and out-of-domain evaluations, indicating that the proposed labeling scheme captures useful region-level information.

Chinese Translation

视觉语言模型（VLMs）应依赖于直接决定正确答案的视觉证据，但用于基础视觉推理的监督往往需要手动获取，成本高昂，或与特定数据集的注释原语相关联。我们提出模型因果视觉证据作为注释目标，定义为一组图像区域，其反事实干预改变给定图像-问题对的模型答案分布。基于这一原则，我们引入了用于基础区域的反事实搜索（Counterfactual Search for Grounding Regions, CSGR）。CSGR是一个可扩展的流程，提出候选区域，对其进行扰动，测量其对答案敏感性的影响，并在多个评审者之间聚合这些证据，以近似VQA数据中的答案关键区域。为了评估CSGR注释是否包含有用的监督信号，我们将其应用于三种现有的关注感知训练流程：注意力引导、视觉CoT微调和潜在视觉推理。这些实验测试了所提出的注释方案是否能在多种使用区域标签的方式中提供有用的监督信号，而不是引入一种新的使用方式。在与竞争的自动区域标注机制的比较中，CSGR注释在领域内和领域外评估中均提供了相对于仅使用交叉熵微调的最一致的增益，表明所提出的标注方案捕捉到了有用的区域级信息。

View on arXiv Download PDF AI Translation

cs.CV / 5 / 2609.13232

What Does the Encoder Actually Decide? A Controlled Comparison of Vision Backbones on Joint Tree Segmentation and Stereo Depth

编码器究竟决定了什么？视觉骨干在联合树木分割和立体深度上的受控比较

Lin, Yida, Xue, Bing, Zhang, Mengjie, Schofield, Sam, Green, Richard

Abstract

A robot pruning trees needs two facts per pixel: whether it belongs to a tree, and its distance. Both are usually obtained via task heads attached to a vision backbone chosen by reputation rather than measurement. Holding dataset, decoders, losses, schedule, and evaluation fixed, we ask: how much does the encoder choice change joint semantic segmentation and stereo depth on thin vegetation? We build a hard parameter-sharing network with one encoder feeding both branches, swapping only the encoder without downstream retuning. We evaluate [N] encoders across [M] architecture families (CNNs, transformers, hybrids, MLP-mixers, state-space models) near a ~25M budget, trained from scratch. Depth is evaluated on tree pixels only; segmentation uses boundary F1 and background IoU to prevent "label-everything-tree" shortcuts. Three findings stand out. First, the strongest encoders are convolutional and hybrid, not transformers: [BestEncoder] leads with [MIoU] segmentation mIoU and [Delta] depth $\delta_1$, while [X] of [Y] plain vision transformers collapse when trained from scratch. Second, parameter count does not predict quality --- [SmallEncoder] at only [P]M parameters outranks models two orders of magnitude larger. Third, segmentation and depth rankings agree strongly (Spearman $\rho$ = [RhoValue]), showing no task conflict. Finally, [K] of [N] encoders collapse to degenerate all-tree segmentation --- exposed by boundary F1 but hidden by region IoU.

Chinese Translation

机器人修剪树木需要每个像素的两个信息：它是否属于树木，以及它的距离。这两个信息通常通过附加在视觉骨干上的任务头获得，而不是通过测量来选择。保持数据集、解码器、损失函数、调度和评估不变，我们提出问题：编码器的选择在稀疏植被上对联合语义分割和立体深度的影响有多大？我们构建了一个硬参数共享网络，一个编码器同时为两个分支提供输入，仅更换编码器而不进行下游微调。我们在约2500万预算下，从头开始训练[N]个编码器，涵盖[M]种架构家族（卷积神经网络、变换器、混合模型、MLP混合器、状态空间模型）。深度评估仅在树木像素上进行；分割使用边界F1和背景IoU来防止“标记所有树木”的捷径。有三个发现尤为突出。首先，最强的编码器是卷积和混合型，而不是变换器：[BestEncoder]在分割mIoU上领先，达到[MIoU]，在深度$ ext{δ}_1$上为[Delta]，而[Y]的[X]普通视觉变换器在从头训练时崩溃。其次，参数数量并不能预测质量——仅有[P]M参数的[SmallEncoder]超越了两个数量级更大的模型。第三，分割和深度的排名高度一致（Spearman $ ho$ = [RhoValue]），显示没有任务冲突。最后，[K]个[N]个编码器崩溃为退化的全树分割——通过边界F1暴露，但通过区域IoU隐藏。

View on arXiv Download PDF AI Translation

cs.CV / 6 / 2609.13233

EMCStereo: Attention-Enhanced Stereo Matching for Thin-Structure Depth Estimation with a Synthetic Tree-Branch Benchmark

EMCStereo：基于注意力增强的立体匹配用于薄结构深度估计的合成树枝基准

Lin, Yida, Xue, Bing, Zhang, Mengjie, Schofield, Sam, Green, Richard

Abstract

Thin structures such as tree branches are among the hardest cases for stereo matching: a branch is only a few pixels wide, the background is cluttered, and dense ground truth for real branches is nearly impossible to label by hand. We make three contributions. First, EMCStereo integrates three lightweight attention modules into a PSMNet-style cost-volume backbone: Efficient Multi-scale Attention (EMA) on deep semantic features, a Multi-Scale Fusion block (MSFblock) learning spatial pyramid weights instead of concatenating them, and Coordinate Attention (CoordAtt) on final matching features. Because MSFblock collapses four pyramid branches into one, the modules leave the network 2.0% smaller and add only 1.7% inference time overhead. Second, VirtualTree is a synthetic stereo dataset rendered in Unreal Engine 5 with a simulated ZED Mini rig, providing 5,520 pairs with exact disparity for thin branches. Third, an eight-way ablation establishes a run-to-run noise floor of 0.009 px end-point error (EPE). EMCStereo achieves 1.31 px EPE (5.96% D1-all) on the VirtualTree test split, 1.00 px on SceneFlow, and 0.80, 0.73, 0.62, and 3.19 px on KITTI 2012, KITTI 2015, ETH3D, and Middlebury, with depth accuracy delta_1 from 92.6% to 98.7%. Evaluated against the noise floor, the attention stack is accuracy-neutral at a 100-epoch budget, while MSFblock and CoordAtt cost 0.03-0.05 px unless EMA is present.

Chinese Translation

薄结构如树枝是立体匹配中最具挑战性的案例之一：树枝宽度仅为几个像素，背景杂乱，并且真实树枝的密集真值几乎不可能手动标注。我们做出了三项贡献。首先，EMCStereo将三个轻量级注意力模块集成到PSMNet风格的代价体积骨干网络中：在深度语义特征上的高效多尺度注意力（Efficient Multi-scale Attention，EMA）、一个学习空间金字塔权重而不是简单拼接的多尺度融合模块（Multi-Scale Fusion block，MSFblock），以及在最终匹配特征上的坐标注意力（Coordinate Attention，CoordAtt）。由于MSFblock将四个金字塔分支合并为一个，这些模块使网络体积减少了2.0%，并仅增加了1.7%的推理时间开销。其次，VirtualTree是一个在虚幻引擎5中渲染的合成立体数据集，使用模拟的ZED Mini设备，提供了5520对具有精确视差的薄树枝。第三，八路消融实验建立了0.009像素端点误差（End-Point Error，EPE）的运行噪声基准。EMCStereo在VirtualTree测试集上达到了1.31像素EPE（5.96% D1-all），在SceneFlow上为1.00像素，在KITTI 2012、KITTI 2015、ETH3D和Middlebury上分别为0.80、0.73、0.62和3.19像素，深度准确度delta_1从92.6%提升至98.7%。在噪声基准下评估，注意力堆栈在100个训练周期的预算下保持准确性中立，而MSFblock和CoordAtt的成本为0.03-0.05像素，除非EMA存在。

View on arXiv Download PDF AI Translation

cs.CV / 7 / 2609.13237

Occlusal Geometry in Closed Form for Orthodontic Report Generation

用于正畸报告生成的封闭形式咬合几何

George, Ajo Babu, Arun, Govind, Krishna, Sidharth N, Ranjan, Uma

Abstract

Orthodontic report generation from intraoral data is normally cast as multimodal captioning, yet the released Bite2Text scan pairs are supplied already registered in occlusion, which makes several core occlusal quantities directly measurable rather than inferable. The system reported here exploits that property: an anatomical frame is recovered per case from arch taper and arch closure instead of the stated RAS convention, which does not hold across the release, and each arch is reduced to an occlusal ridge profile in arch-angle coordinates yielding overbite, overjet, midline deviation, transverse overlap, crossbite extent, cusp interdigitation lag, and the occlusal curves in closed form. Gradient boosting maps 31 such measurements onto 13 template fields, a field being predicted only where patient-level cross-validation beats its own majority baseline, and a deterministic renderer emits the corpus six-part narrative; a ConvNeXt-Tiny classifier over the five standardised photographic views is fused per field, raising mean field accuracy from 0.601 to 0.683. Reimplementation of the challenge evaluator shows that its BLEU-4 and METEOR are local variants whose F-mean weights recall nine to one, that two clinicians agree on 47 percent of findings for the same patient, and that a constant report consequently outscores a genuine second clinician report by 0.165 captioning. Held-out scores reach BLEU-4 0.458 and METEOR 0.677 against intraoral scan references and 0.278 and 0.507 against photograph references, and the submitted system placed third in the ODIN 2026 Bite2Text test phase at 0.2680 and 0.4629, within 0.022 BLEU-4 of first, running on CPU in under ten seconds per case. The dataset and code are available at https://github.com/GIND123/ODIN_toothfairy4

Chinese Translation

从口内数据生成正畸报告通常被视为多模态字幕生成，然而发布的 Bite2Text 扫描对已经在咬合中进行配准，这使得多个核心咬合量可以直接测量而非推断。本文报告的系统利用了这一特性：每个案例从弓形锥度和弓形闭合中恢复解剖框架，而不是所述的 RAS（右手坐标系）约定，该约定在发布中并不适用，每个弓形被简化为弓角坐标系中的咬合脊轮廓，从而得出过度咬合、过度前突、中线偏移、横向重叠、交叉咬合程度、尖头交错滞后以及封闭形式的咬合曲线。梯度提升将 31 项此类测量映射到 13 个模板字段，只有在患者级交叉验证超越自身的多数基线时，才会预测某个字段，而确定性渲染器则输出六部分叙述的语料；在五个标准化摄影视图中，ConvNeXt-Tiny 分类器在每个字段中融合，提高了平均字段准确率从 0.601 到 0.683。挑战评估器的重新实现显示其 BLEU-4 和 METEOR 是局部变体，其 F-均值权重回忆率为九比一，两位临床医生对同一患者的发现一致率为 47%，因此一个恒定报告的得分超越了真实第二位临床医生报告的得分 0.165。在与口内扫描参考的比较中，保留分数达到 BLEU-4 0.458 和 METEOR 0.677，而与照片参考的比较中则为 0.278 和 0.507，提交的系统在 ODIN 2026 Bite2Text 测试阶段中排名第三，得分为 0.2680 和 0.4629，距离第一名的 BLEU-4 仅相差 0.022，且每个案例在 CPU 上运行时间少于十秒。数据集和代码可在 https://github.com/GIND123/ODIN_toothfairy4 获取。

View on arXiv Download PDF AI Translation

cs.CV / 8 / 2609.13239

Abstract-LoRA: Unlocking Single-Image Style Transfer through Targeted U-Net Block Training

Abstract-LoRA：通过针对性 U-Net 块训练解锁单图像风格迁移

Hu, Xinglin

Abstract

Diffusion models represent one of the most advanced paradigms in generative modeling. Leveraging their development, a growing number of style transfer methods based on diffusion models have been proposed. However, among these methods, multi-image style transfer approaches that require at least five to ten style examples tend to achieve more satisfactory results. Single-image methods, by contrast, often struggle with either insufficient content preservation or inadequate style fidelity. This greatly limits style extraction from scarce artworks and undermines their artistic value. To address this, we propose Abstract-LoRA, a method that pushes the boundaries of single-image style transfer through lightweight LoRA training on specific U-Net blocks in diffusion models. Specifically, our work is inspired by B-LoRA, a style transfer method that achieves basic style-content disentanglement by training specific U-Net blocks. However, it suffers from a critical limitation: the inability to capture complex backgrounds. Building upon B-LoRA, our method conducts a more refined analysis of U-Net blocks, employing additional U-Net blocks and clustering-based abstraction of style images to better disentangle and balance style and content. Extensive experiments demonstrate that our proposed method not only generates visually more harmonious and satisfying artistic images but also quantitatively improves the preservation of both style and content in the final outputs.

Chinese Translation

扩散模型代表了生成建模中最先进的范式之一。借助其发展，越来越多基于扩散模型的风格迁移方法应运而生。然而，在这些方法中，要求至少五到十个风格示例的多图像风格迁移方法往往能取得更令人满意的结果。相比之下，单图像方法常常在内容保留不足或风格保真度不够方面遇到困难。这极大限制了从稀缺艺术作品中提取风格的能力，并削弱了其艺术价值。为了解决这一问题，我们提出了 Abstract-LoRA，一种通过在扩散模型中特定 U-Net 块上进行轻量级 LoRA 训练，推动单图像风格迁移边界的方法。具体而言，我们的工作受到 B-LoRA 的启发，B-LoRA 是一种通过训练特定 U-Net 块实现基本风格-内容解耦的风格迁移方法。然而，它存在一个关键限制：无法捕捉复杂背景。在 B-LoRA 的基础上，我们的方法对 U-Net 块进行了更精细的分析，采用额外的 U-Net 块和基于聚类的风格图像抽象，以更好地解耦和平衡风格与内容。大量实验表明，我们提出的方法不仅生成了视觉上更和谐和令人满意的艺术图像，而且在最终输出中定量提高了风格和内容的保留。

View on arXiv Download PDF AI Translation

cs.CV / 9 / 2609.13240

ArtSociety: Multi-Agent Multimodal Collaboration for Art Emotion Understanding

ArtSociety：用于艺术情感理解的多智能体多模态协作

Li, Jian, Ji, Fanfan, Lai, Jinxiang, Tai, Ying, Yang, Jian, Yuan, Xiao-Tong, Wang, Chengjie, Wang, Yabiao

Abstract

The AffectiveArt Multidimensional Art Emotion Understanding task asks to jointly predict an artwork's fine-grained emotion (12 classes, 1549:1 head-to-tail ratio), binary valence/arousal, and five attribute-grounded descriptions -- sub-tasks that exhibit strong empirical trade-offs, so the single-model solutions we tried do not jointly optimize all of them well. We present ArtSociety, a multi-agent framework that assembles heterogeneous multimodal experts -- a DINOv2-Giant vision agent (A1), a scene-grounded CoT fine-tuned MLLM (A2), and three closed-source reasoning agents (A3-A5) -- and coordinates them with two training-free controllers: (i) a rare-class-aware voting arbiter that lowers the agreement threshold for tail emotions, exploiting decorrelated error patterns across agent families; and (ii) a description-first reasoning agent whose DESCRIBE-then-CLASSIFY chain of thought forces visual evidence before label commitment, yielding near-perfect grounded descriptions. A task-routing policy directs the hard emotion task to the full five-agent ensemble while assigning the near-saturated valence/arousal and generative description tasks to the single strongest reasoning agent. On the official test set (1,000 artworks), ArtSociety achieves an Overall Score of 0.8870 (Classification 0.7789, Description 0.9952). An eleven-variant ablation study reveals that, once method and scale saturate at around 0.76, the decisive gains come from agent collaboration and data-side supervision -- a 30B MoE model trained on older data does not outperform an 8B model trained on better data. Code is available at https://github.com/swordlidev/ArtSociety

Chinese Translation

AffectiveArt多维艺术情感理解任务要求共同预测艺术作品的细粒度情感（12类，1549:1的头尾比）、二元的愉悦度/唤醒度以及五个基于属性的描述——这些子任务展示了强烈的经验权衡，因此我们尝试的单模型解决方案无法很好地共同优化它们。我们提出了ArtSociety，一个多智能体框架，汇集了异构的多模态专家——一个DINOv2-Giant视觉智能体（A1）、一个基于场景的经过微调的CoT MLLM（A2）以及三个闭源推理智能体（A3-A5），并通过两个无训练的控制器进行协调：（i）一个关注稀有类别的投票仲裁者，降低尾部情感的协议阈值，利用智能体家族之间去相关的错误模式；（ii）一个以描述为先的推理智能体，其“描述-再分类”的思维链在标签承诺之前强制提供视觉证据，从而产生近乎完美的基于属性的描述。任务路由策略将困难的情感任务指派给完整的五智能体集群，同时将接近饱和的愉悦度/唤醒度和生成描述任务分配给单个最强的推理智能体。在官方测试集（1,000件艺术作品）上，ArtSociety的总体得分为0.8870（分类得分0.7789，描述得分0.9952）。一项包含十一种变体的消融研究表明，一旦方法和规模在约0.76处饱和，决定性的增益来自智能体协作和数据侧监督——一个在旧数据上训练的30B MoE模型并未优于一个在更好数据上训练的8B模型。代码可在https://github.com/swordlidev/ArtSociety获取。

View on arXiv Download PDF AI Translation

cs.CV / 10 / 2609.13245

SJD-SV: Speculative Jacobi Decoding with Semantics Verification for Autoregressive Image Generation

SJD-SV：带语义验证的推测雅可比解码用于自回归图像生成

Zhang, Baoquan, Shan, Bingqi, Fang, Shihao, Lin, Kenghong, Li, Xutao, Ye, Yunming

Abstract

Speculative Jacobi Decoding (SJD) is an important approach for accelerating autoregressive image generation. Although SJD has shown superior performance, recent studies point out that it usually suffers from a token ambiguity issue during token verification but its reason can not be well explained. To figure out this reason, in this paper, we conduct a visualization analysis on vision token and find that different from text tokens, vision tokens generally corresponds to some local, small, and unclear vision details, which means only using single token is difficult to accurately express a certain semantic, thereby causing token ambiguity issue. To this end, we propose a novel Speculative Jacobi Decoding with Semantics Verification (called SJD-SV), for accelerating autoregressive image generation. The key idea is that leveraging the strong correction characters between tokens to recognize semantic-aware token subsequence and then instead of perform token-by-token verification, turning to perform verification on semantic-aware token subsequence level for accelerating image generation. In particular, our method is plug-in, which can be directly integrated into existing SJD and its variants. Extensive experiments on various datasets show that existing SJD methods achieve significant performance improvement after integrating our SJD-SV method.

Chinese Translation

推测雅可比解码（SJD）是一种加速自回归图像生成的重要方法。尽管SJD表现出优越的性能，但最近的研究指出，在令牌验证过程中，它通常会遭遇令牌歧义问题，而其原因尚未得到很好的解释。为了解释这一原因，本文对视觉令牌进行了可视化分析，发现与文本令牌不同，视觉令牌通常对应于一些局部、小而模糊的视觉细节，这意味着仅使用单个令牌难以准确表达特定语义，从而导致令牌歧义问题。为此，我们提出了一种新颖的带语义验证的推测雅可比解码（称为SJD-SV），以加速自回归图像生成。其关键思想是利用令牌之间强大的纠正特性来识别语义感知的令牌子序列，然后不再逐个进行令牌验证，而是转向在语义感知的令牌子序列层面进行验证，以加速图像生成。特别地，我们的方法是插件式的，可以直接集成到现有的SJD及其变体中。在各种数据集上的大量实验表明，现有的SJD方法在集成我们的SJD-SV方法后实现了显著的性能提升。

View on arXiv Download PDF AI Translation

cs.CV / 11 / 2609.13246

Pixel-wise Planarity for High-Precision Monocular Plane Segmentation

高精度单目平面分割的像素级平面性

Yavuz, Ahmetcan, Ozkan, Alpay, Pautrat, Rémi, Liu, Shaohui, Pollefeys, Marc

Abstract

Plane segmentation from a single RGB image remains challenging due to imprecise region grouping and geometrically inconsistent supervision, often leading to over-segmentation and false planar detections. We propose instead a pixel-wise planarity prediction framework for robust monocular plane segmentation. Building on a pretrained monocular geometric backbone predicting depth and surface normals, we introduce a dedicated planarity head that estimates per-pixel planarity confidence. During inference, predicted depth, normals, and planarity are combined in a lightweight region-growing procedure that enforces geometric consistency when forming plane segments. We further analyze existing plane ground-truth annotations and demonstrate substantial geometric inconsistencies under strict distance thresholds. Across multiple datasets, our method achieves improved geometric precision and segmentation quality compared to prior state-of-the-art approaches, while improving computational efficiency. Our code and models are available at https://github.com/alpayozkan/PixelwisePlanarity.

Chinese Translation

从单个RGB图像中进行平面分割仍然具有挑战性，主要由于区域分组不精确和几何不一致的监督，这常常导致过度分割和错误的平面检测。我们提出了一种像素级平面性预测框架，以实现稳健的单目平面分割。该框架基于一个预训练的单目几何骨干网络，能够预测深度和表面法线，并引入一个专门的平面性头部，用于估计每个像素的平面性置信度。在推理过程中，预测的深度、法线和平面性结合在一个轻量级的区域生长过程中，以在形成平面段时强制执行几何一致性。我们进一步分析了现有的平面真实标注，并在严格的距离阈值下展示了显著的几何不一致性。在多个数据集上，我们的方法相比于之前的最先进方法实现了更高的几何精度和分割质量，同时提高了计算效率。我们的代码和模型可在 https://github.com/alpayozkan/PixelwisePlanarity 获取。

View on arXiv Download PDF AI Translation

cs.CV / 12 / 2609.13250

Evaluation of MLLM-Agnostic Plug-and-Play Keyframe Selection Methods for Long Video Understanding

无MLLM依赖的即插即用关键帧选择方法在长视频理解中的评估

Sarkar, Dilip, Islam, Md. Safayet, Liang, Liang

Abstract

Multimodal large language models (MLLMs) cannot process every frame of a long video because of limitations in visual-token and computational budgets. Three primary approaches have been proposed to enhance their long-video understanding capabilities: (i) Retraining an MLLM on a large video corpus and/or extending its input length; (ii) Training an adapter for a specific MLLM that takes the entire video and the query as input and selects the most relevant video frames; and (iii) Developing a training-free, plug-and-play (PaP) adapter that is MLLM-agnostic. We refer to the third approach as PaP keyframe selection. A PaP method may use only candidate video frames without considering the query, or it may use both candidate video frames and the query. The first approach is prohibitively expensive. The second approach requires substantial training time and computational resources, but it is accessible to many because an adapter contains significantly fewer trainable parameters than an entire MLLM. The third approach has the lowest computational cost and is therefore broadly accessible. To the best of our knowledge, only five PaP methods have been reported within the past year. All of these methods have been evaluated on one or more video question-answering benchmarks and have demonstrated improvements in long-video understanding. However, the methods were evaluated on different benchmarks using different MLLMs. We present a comprehensive evaluation of these five methods using three MLLMs across three long-video understanding benchmarks. Our results show that QAaF achieves the best performance in 13 of the 15 aggregate evaluation settings, while FOCUS ranks second overall. These results provide a common experimental reference for comparing training-free keyframe-selection methods for MLLMs.

Chinese Translation

多模态大型语言模型（MLLMs）由于视觉标记和计算预算的限制，无法处理长视频的每一帧。为增强其长视频理解能力，提出了三种主要方法：（i）在大型视频语料库上重新训练MLLM和/或扩展其输入长度；（ii）为特定MLLM训练一个适配器，该适配器将整个视频和查询作为输入，并选择最相关的视频帧；（iii）开发一种无训练的即插即用（PaP）适配器，该适配器与MLLM无关。我们将第三种方法称为PaP关键帧选择。PaP方法可以仅使用候选视频帧而不考虑查询，或者同时使用候选视频帧和查询。第一种方法成本过高。第二种方法需要大量的训练时间和计算资源，但由于适配器的可训练参数远少于整个MLLM，因此对许多人来说是可及的。第三种方法的计算成本最低，因此广泛可用。据我们所知，过去一年仅报告了五种PaP方法。这些方法均在一个或多个视频问答基准上进行了评估，并在长视频理解方面显示出改进。然而，这些方法在不同的基准上使用不同的MLLM进行评估。我们对这五种方法在三个长视频理解基准上使用三种MLLM进行了全面评估。我们的结果表明，QAaF在15个综合评估设置中有13个取得了最佳性能，而FOCUS整体排名第二。这些结果为比较MLLM的无训练关键帧选择方法提供了一个共同的实验参考。

View on arXiv Download PDF AI Translation

cs.CV / 13 / 2609.13251

Preserving Subject-Clarity in Image Outpainting with Multiscale Wavelet Supervision

通过多尺度小波监督在图像外延中保持主体清晰度

Neog, Abhilash, Kim, Taewan, Wu, Yi, Chen, Xu, Jiao, Jian

Abstract

Commercial and advertising images are frequently affected by poor framing, partially cropped subjects, truncated text or logos, and insufficient context, all of which can reduce subject clarity, i.e., the ability of an image to clearly communicate its primary subject. Image outpainting offers a scalable solution by extending image boundaries and recovering missing content and context. However, existing diffusion-based outpainting methods often produce visually plausible completions while degrading subject fidelity through structural inconsistencies, semantic drift, or loss of fine-grained detail. To address this limitation, we propose a subject clarity outpainting framework that combines vision-language model (VLM)-guided semantic conditioning with multiscale wavelet supervision for subject-localized detail preservation. To support training, we develop a subject-centric data curation pipeline that constructs subject-intersecting outpainting pairs from advertising and natural images. The resulting objective introduces no additional inference cost and is designed to be compatible with diffusion-based backbones. Across four advertising and natural-image benchmarks, our method improves subject clarity, reducing subject-centered DreamSim error and FID on average by 3.0% and 2.4% over matched supervised fine-tuning, and by 10.8% and 7.7% over the strongest state-of-the-art approach per dataset, respectively.

Chinese Translation

商业和广告图像常常受到糟糕构图、部分裁剪的主体、截断的文本或标志以及上下文不足的影响，这些因素都会降低主体清晰度，即图像清晰传达其主要主体的能力。图像外延提供了一种可扩展的解决方案，通过扩展图像边界并恢复缺失的内容和上下文。然而，现有的基于扩散的外延方法通常能够生成视觉上合理的补全，但却通过结构不一致、语义漂移或细节丢失来降低主体的保真度。为了解决这一局限性，我们提出了一种主体清晰度外延框架，该框架结合了视觉-语言模型（VLM）引导的语义条件与多尺度小波监督，以保持主体局部细节。为了支持训练，我们开发了一种以主体为中心的数据策划管道，从广告和自然图像中构建主体交叉的外延对。所提出的目标不会引入额外的推理成本，并且旨在与基于扩散的主干兼容。在四个广告和自然图像基准测试中，我们的方法提高了主体清晰度，平均降低了主体中心的DreamSim误差和FID，分别比匹配的监督微调降低了3.0%和2.4%，比每个数据集上最强的最先进方法降低了10.8%和7.7%。

View on arXiv Download PDF AI Translation

cs.CV / 14 / 2609.13254

(How) Do MLLMs Report Bistable Images Like Humans?

多模态大型语言模型如何像人类一样报告双稳态图像？

Takatsuki, Ryota, Doi, Tomoki, Watahiki, Amane, Seth, Anil K., Yanaka, Hitomi

Abstract

Bistable images such as the duck-rabbit are classic stimuli in which one image supports multiple mutually incompatible interpretations, typically reported one at a time in humans. We ask whether multimodal large language models (MLLMs) show similar report behavior and what internal computations support it. Using the LLaVA family, we study two tractable dimensions: modulability, whether reports can be biased by bottom-up visual cues and top-down linguistic priors, and exclusivity, whether responses commit to a single interpretation. We test both on the canonical duck-rabbit and on synthetic Visual Anagrams to mitigate memorization confounds. Behaviorally, both visual and linguistic manipulations systematically shift reports in human-consistent ways, while responses remain predominantly exclusive. Mechanistically, these effects arise from competing image-token representations, distinct pathways for bottom-up and top-down modulation, and a link between exclusive reporting and object-count encoding. Code and data are available at https://github.com/rtakatsky/mllm-bistable-images.

Chinese Translation

双稳态图像，如鸭兔图，是经典的刺激材料，其中一幅图像支持多种相互不兼容的解释，通常人类一次只报告一种解释。我们探讨多模态大型语言模型（MLLMs）是否表现出类似的报告行为，以及支持这种行为的内部计算机制。使用LLaVA系列模型，我们研究了两个可操作的维度：可调性，即报告是否会受到自下而上的视觉线索和自上而下的语言先验的偏倚；独占性，即响应是否承诺于单一解释。我们在经典的鸭兔图和合成视觉字谜上进行测试，以减轻记忆干扰。从行为上看，视觉和语言的操控系统性地以与人类一致的方式改变报告，同时响应仍然主要保持独占性。从机制上讲，这些效应源于竞争的图像-标记表示、自下而上和自上而下调节的不同路径，以及独占报告与对象计数编码之间的联系。代码和数据可在 https://github.com/rtakatsky/mllm-bistable-images 获取。

View on arXiv Download PDF AI Translation

cs.CV / 15 / 2609.13255

A Comprehensive Review of Multimodal Facial State Analysis: Tasks, Methods, and Resources

多模态面部状态分析的综合评述：任务、方法与资源

Ge, Xuri, Zhang, Tianshuo, Li, Ruihan, Ye, Hui, Zheng, Kaiwen, Fu, Junchen, Huo, Da, Jose, Joemon M., Han, Hu

Abstract

Facial state analysis plays a crucial role in understanding human expressions, psychological modeling, and human computer interaction. Traditional unimodal vision-based methods are often limited by environmental sensitivity and weak interpretability. Multimodal facial state analysis addresses these issues by integrating complementary cues from visual, audio, textual, physiological, and other related modalities. This survey emphasizes two key aspects: on one hand, multimodal learning enables contextual semantic understanding for improved facial state reasoning and leverages interpretable language generation to enhance model explainability; on the other hand, multi-task learning allows simultaneous analysis of expressions, action units (AUs), and face-based soft biometrics (e.g., age, gender), effectively capturing fine-grained expressions and improving cross-scene generalization. This survey reviews core tasks, representative methods, and datasets in multimodal facial state analysis, focusing on facial expression recognition, AU detection, and face-based soft biometric estimation, and emphasizing the unique value of language in providing contextual semantics, enhancing reasoning, and generating explanations. The survey aims to provide an up-to-date overview of the literature and to highlight future research directions for multimodal, interpretable, and multi-task adaptive facial state analysis.

Chinese Translation

面部状态分析在理解人类表情、心理建模和人机交互中发挥着至关重要的作用。传统的单模态视觉方法常常受到环境敏感性和解释性弱的限制。多模态面部状态分析通过整合来自视觉、音频、文本、生理和其他相关模态的互补线索，解决了这些问题。本调查强调两个关键方面：一方面，多模态学习使得上下文语义理解得以实现，从而改善面部状态推理，并利用可解释的语言生成增强模型的可解释性；另一方面，多任务学习允许对表情、动作单元（AUs）和基于面部的软生物特征（如年龄、性别）进行同时分析，有效捕捉细粒度的表情并改善跨场景的泛化能力。本调查回顾了多模态面部状态分析中的核心任务、代表性方法和数据集，重点关注面部表情识别、AU检测和基于面部的软生物特征估计，并强调语言在提供上下文语义、增强推理和生成解释中的独特价值。本调查旨在提供文献的最新概述，并突出未来多模态、可解释和多任务自适应面部状态分析的研究方向。

View on arXiv Download PDF AI Translation

cs.CV / 16 / 2609.13257

Sampling headroom is not selection gain: a compute-value audit of test-time scaling for video world models

采样余量并非选择增益：视频世界模型测试时间缩放的计算价值审计

Jiang, Yuhua, Lu, Junjie, Gao, Feifei

Abstract

Test-time scaling (TTS) can improve generation only when additional compute produces better candidates and the system can reliably identify them. This distinction is especially important for video world models, where a wider sample pool may contain stronger rollouts without improving the output that is ultimately selected. We introduce the Compute-Value Audit (CVA), a sequential framework that asks whether extra sampling creates opportunity, observable signals provide a reliable state, that state supports a beneficial action, and the resulting gain exceeds the full entry fee of generation and verification. On 192 Physics-IQ scenes, expanding the pool from 4 to 16 candidates increases oracle quality by +9.23 IQ (95% CI [+7.44, +11.14]), but Flow, Cycle, and VideoReward fail to recover this headroom reliably. Across three generators, none of twelve adaptive-depth policies outperforms uniform compute; they recover only 42-69% of the measured entry fee. A matched-60-NFE Predict-and-Perturb intervention on VideoPhy2 is likewise negative across three fresh-seed replicas. These negative results are not universal: anchor-explorer passes all four stages in a sparse PRM800K setting, MMLU-Pro exposes the gap between predictive state and useful action, and a privileged paired future establishes a positive video upper bound. Together, these results show that sampling headroom has deployment value only when it can be converted into a reliable decision whose benefit survives the complete compute charge. Code is available at https://github.com/YuhuaJiang2002/sampling-headroom-is-not-selection-gain.

Chinese Translation

测试时间缩放（TTS）只有在额外计算能够产生更好的候选项且系统能够可靠识别这些候选项时，才能改善生成效果。这一区别对于视频世界模型尤为重要，因为更广泛的样本池可能包含更强的回放，但并不一定改善最终选择的输出。我们引入了计算价值审计（CVA），这是一个序贯框架，旨在探讨额外采样是否创造了机会、可观察信号是否提供了可靠状态、该状态是否支持有益行动，以及由此产生的增益是否超过生成和验证的全部费用。在192个Physics-IQ场景中，将候选池从4个扩展到16个使得oracle质量提高了+9.23 IQ（95% CI [+7.44, +11.14]），但Flow、Cycle和VideoReward未能可靠地恢复这一余量。在三个生成器中，十二种自适应深度策略均未优于均匀计算；它们仅恢复了测量入场费的42-69%。在VideoPhy2上进行的匹配-60-NFE预测与扰动干预在三个新种子复制中同样表现为负。这些负面结果并非普遍存在：在稀疏PRM800K设置中，anchor-explorer通过了所有四个阶段，MMLU-Pro揭示了预测状态与有用行动之间的差距，而特权配对未来则建立了一个正向视频上限。综合来看，这些结果表明，采样余量只有在能够转化为可靠决策且其收益超过完整计算费用时，才具有部署价值。代码可在 https://github.com/YuhuaJiang2002/sampling-headroom-is-not-selection-gain 获取。

View on arXiv Download PDF AI Translation

cs.CV / 17 / 2609.13258

Interpretable Temporal Video Reasoning with EventGraph and EventField

可解释的时间视频推理：基于事件图和事件场

Singh, Durgendra Narayan

Abstract

We present a structured temporal video reasoning pipeline built around a discrete EventGraph, a continuous EventField, and a human-readable EventGlyph view. On a calibrated EPIC-KITCHENS subset of 10 videos and 50 temporal reasoning questions, EventField+Glyph achieves 0.98 overall accuracy, which is higher than the caption baseline by +0.40 (paired p = 1.1 \times 10^{-5}) and direct VLM-only QA by +0.20 (p = 0.0063) on this subset. We further evaluate annotation-source variations, including manual, heuristic, and heuristic+Gemini pipelines, and find that the best structured method stays above the caption baseline across settings. We also include cross-video pair benchmarking and an appendix gallery of glyph outputs for all studied videos. Overall, the results indicate that structured temporal representations can support both performance and inspectability by preserving symbolic structure, capturing temporal continuity, and providing human-readable diagnostics for video reasoning.

Chinese Translation

我们提出了一种基于离散事件图（EventGraph）、连续事件场（EventField）和人类可读的事件图形视图（EventGlyph）的结构化时间视频推理管道。在经过校准的EPIC-KITCHENS子集（包含10个视频和50个时间推理问题）上，EventField+Glyph的整体准确率达到了0.98，比字幕基线高出0.40（配对p = 1.1 imes 10^{-5}），比直接的VLM-only问答高出0.20（p = 0.0063）。我们进一步评估了注释源的变化，包括手动、启发式和启发式+Gemini管道，发现最佳的结构化方法在各种设置中均高于字幕基线。我们还包括了跨视频对的基准测试和所有研究视频的图形输出附录画廊。总体而言，结果表明，结构化的时间表示可以通过保留符号结构、捕捉时间连续性以及提供人类可读的诊断信息来支持性能和可检查性，从而促进视频推理。

View on arXiv Download PDF AI Translation

cs.CV / 18 / 2609.13259

TryOnReward: Learning Foveated Consistency for Reinforcement Fine-Tuning of Virtual Try-On

TryOnReward：学习注视一致性以强化虚拟试穿的微调

Li, Xueheng, Liu, Yong, Fu, Xiaolong, Xue, Wen, Xie, Chengjun, Sun, Yipeng, Li, Yan, Gu, Simiu

Abstract

Virtual Try-On (VTON) aims to dress a person with the reference garment, producing visually reasonable results aligned with human preferences. Turning this preference-oriented goal into an actionable objective relies on a scoring function aligned with human taste. However, classic fidelity metrics exhibit weak correlation with human judgments, and generic VLMs fail to provide the discriminative granularity demanded by try-on quality evaluation, which hinges on faithfully preserving garment and person details. This shortcoming is further exacerbated in the reinforcement fine-tuning (RFT) optimization and leads to severe reward hacking. To this end, we present TryOnReward, a fine-grained reward model tailored for VTON. Built on a vision-language backbone, it adopts a foveation calibration objective that grounds each quality dimension in the relevant region to avoid global shortcut learning. Meanwhile, TryOnReward jointly optimizes pairwise preferences and per-dimension quality scores via margin-aware supervision, leveraging both relative and absolute quality signals. For model training and evaluation, we build TryOnReward-100K, a human-annotated per-dimension rating dataset, alongside TryOn-Bench and TryOnRewardBench, two benchmarks covering diverse real scenarios. Extensive experiments confirm that TryOnReward significantly outperforms generic judges in human preference alignment, and when serving as the RFT reward function, it consistently yields human-preferred try-on results across multiple baselines.

Chinese Translation

虚拟试穿（VTON）旨在为一个人穿上参考服装，产生与人类偏好一致的视觉合理结果。将这一以偏好为导向的目标转化为可操作的目标依赖于与人类品味相一致的评分函数。然而，经典的保真度指标与人类判断的相关性较弱，而通用的视觉语言模型（VLM）未能提供试穿质量评估所需的区分细节，这种评估依赖于忠实保留服装和人物细节。这一缺陷在强化微调（RFT）优化中进一步加剧，导致严重的奖励操控。为此，我们提出了TryOnReward，一个针对VTON的细粒度奖励模型。该模型基于视觉-语言骨干网络，采用注视校准目标，将每个质量维度与相关区域相结合，以避免全局捷径学习。同时，TryOnReward通过边际感知监督共同优化成对偏好和每个维度的质量评分，利用相对和绝对质量信号。为了模型的训练和评估，我们构建了TryOnReward-100K，一个人类注释的每维度评分数据集，以及TryOn-Bench和TryOnRewardBench两个涵盖多样真实场景的基准。大量实验确认，TryOnReward在与人类偏好对齐方面显著优于通用评判者，并且在作为RFT奖励函数时，它在多个基线中始终产生人类偏好的试穿结果。

View on arXiv Download PDF AI Translation

cs.CV / 19 / 2609.13262

Filling the Unseen: Scene Extrapolation via 3D Gaussian Splatting

填补未见：通过3D高斯点云进行场景外推

Zhou, Yunlai, Lu, Yiren, Liang, Tuo, Liu, Disheng, Chaudhary, Vipin, Yin, Yu

Abstract

3D Gaussian Splatting achieves photorealistic reconstruction within training view distribution, yet it degrades on out-of-distribution novel views, exhibiting holes in unobserved regions and artifacts in observable areas. Recent works formulate this task as extrapolation and interpolation and try to address it with generative models, but remain limited in extrapolation scale and quality. They repeat a generate-reconstruct-shift cycle to progressively build a scene, which introduces accumulated errors with every step conditioning on previous outcomes. In this work, we propose a holistic framework for extrapolation and interpolation. We devise an independent camera view detection mechanism to enable parallel conflict-free extrapolation, circumventing the reliance on the aforementioned error-prone cycle. Building upon this, we design a hierarchical pipeline that extrapolates independent and dependent camera views separately. Additionally, previous methods overlook inconsistency between generated and original images, resulting in compromising well-reconstructed areas. We propose a plug-and-play Quality-Aware Mask (QA-Mask) module, enabling selective utilization of generated data. By calibrating learning weights with pixel-wise rendering quality, it prevents generation-induced degradations on well-constructed areas. Extensive experiments demonstrate the superior performance of our framework, with QA-Mask generalizing to multiple generative reconstruction models.

Chinese Translation

3D高斯点云在训练视图分布内实现了照片级真实感重建，但在分布外的新视图上表现不佳，未观察区域出现空洞，观察区域则出现伪影。近期的研究将此任务表述为外推和插值，并尝试通过生成模型来解决，但在外推的规模和质量上仍然有限。它们重复生成-重建-偏移的循环，以逐步构建场景，这在每一步中都引入了基于先前结果的累积误差。在本研究中，我们提出了一个整体框架用于外推和插值。我们设计了一种独立的相机视图检测机制，以实现并行无冲突的外推，避免依赖于上述易出错的循环。在此基础上，我们设计了一个分层管道，分别外推独立和依赖的相机视图。此外，以往的方法忽视了生成图像与原始图像之间的不一致性，导致重建良好的区域受到影响。我们提出了一个即插即用的质量感知掩码（Quality-Aware Mask, QA-Mask）模块，能够选择性地利用生成数据。通过根据像素级渲染质量校准学习权重，它防止了生成引起的重建区域退化。大量实验表明，我们的框架表现优越，QA-Mask能够推广到多种生成重建模型。

View on arXiv Download PDF AI Translation

cs.CV / 20 / 2609.13263

Multi-View Structure-from-Motion Enables Oriented Projective Shape Analysis in Three Dimensions

多视角运动重建实现三维定向投影形状分析

Alamoudi, Musab, Paige, Robert L., Patrangenaru, Vic

Abstract

Projective shape analysis provides a geometric framework for studying landmark configurations in digital images acquired by pinhole cameras. In the classical projective shape (PS) model, three-dimensional configurations ($k$-ads) are represented as points in $(\mathrm{RP}^3)^q$, $q = k - 5$. A nonparametric test is developed in Patrangenaru et al. [12], for this framework, to determine whether an object matches a design blueprint, with each configuration reconstructed from a single uncalibrated stereo pair. Such two-view reconstructions are identified only up to a 3D projective transformation, which may reverse orientation, so that the sign-blind PS summary was the only available, before oriented projective shape (OPS) was recently considered. Multi-view Structure-from-Motion (SfM) technology removes this obstruction: its bundle adjustment is identified up to an orientation-preserving projective transformation. In this paper we revisit a well-cited three-cube object, from Patrangenaru et al. [12], with $n = 8$ SfM reconstructions built in Agisoft Metashape Professional 2.3.0, and validate our implementation by reproducing the published stereo construction from the original data. This allows us to perform what is to the best of our knowledge the first three-dimensional OPS analysis, compute its extrinsic total-variance index and perform statistical inference in this novel setting. Due to the high concentration of SfM data, the OPS index is asymptotically one-half the PS index, a structural consequence of concentration rather than a property of the object. Here our blueprint hypothesis is not rejected for any of the $q = 14$ non-frame landmarks, while the SfM reconstructions are about 26 times more concentrated than the stereo ones, a substantial gain in reconstruction precision. Sample-size, photograph-count, and frame-ordering analyses support the robustness of these conclusions.

Chinese Translation

投影形状分析提供了一个几何框架，用于研究通过针孔相机获取的数字图像中的地标配置。在经典的投影形状（PS）模型中，三维配置（$k$-ads）被表示为$( ext{RP}^3)^q$中的点，其中$q = k - 5$。Patrangenaru等人[12]为该框架开发了一种非参数检验，以确定一个物体是否与设计蓝图匹配，每个配置都是从一对未校准的立体图像重建而来。这种双视图重建仅在一个三维投影变换下被识别，该变换可能会反转方向，因此在最近考虑定向投影形状（OPS）之前，只有无符号的PS摘要可用。多视角运动重建（SfM）技术消除了这一障碍：其束调整在保持方向的投影变换下被识别。在本文中，我们重新审视了Patrangenaru等人[12]中引用较多的三立方体对象，使用Agisoft Metashape Professional 2.3.0构建了$n = 8$个SfM重建，并通过从原始数据重建已发布的立体构造来验证我们的实现。这使我们能够进行我们所知的首次三维OPS分析，计算其外部总方差指数，并在这一新颖环境中进行统计推断。由于SfM数据的高度集中，OPS指数渐近地为PS指数的一半，这是集中度的结构性结果，而非物体的属性。在这里，我们的蓝图假设对于$q = 14$个非框架地标没有被拒绝，而SfM重建的集中度约为立体重建的26倍，显示出重建精度的显著提升。样本量、照片数量和帧排序分析支持这些结论的稳健性。

View on arXiv Download PDF AI Translation

cs.CV / 21 / 2609.13264

BEACON: Behavior and Appearance Control for Subject-Specific Video Generation

BEACON：针对特定主体的视频生成中的行为与外观控制

Baptiste, Pokrzywa, Quignon, Nabyl, Bahram, Yara, Zeeshan, Muhammad Osama, Dantcheva, Antitza, Granger, Eric

Abstract

Generating human-centric videos that preserve both visual identity and person-specific expressive behavior remains a fundamental challenge. In addition to reproducing appearance, a model must replicate the facial behaviors that characterize how a subject expresses emotion over time. However, most state-of-the-art methods condition generation on a single reference image, which contains no information about these temporal dynamics. As a result, they tend to preserve the subject's visual identity but often produce expressions with limited variation and weak subject specificity. To mitigate this issue, we introduce BEACON, a lightweight framework for person-specific video generation that produces more expressive videos by disentangling visual identity from expressive behavior. BEACON conditions generation on two complementary signals: a reference image encoding the identity and a reference video capturing subject-specific facial dynamics. By conditioning on these complementary signals, BEACON generates videos that better preserve both the subject's appearance and characteristic facial dynamics, while also supporting identity-expression transfer. Our experiments on the MEAD and RAVDESS datasets show that by fine-tuning on approximately 2,000 pairs and updating about 1% of the pretrained Wan video diffusion model, BEACON improves facial expressivity over state-of-the-art video generation methods while maintaining competitive identity preservation.

Chinese Translation

生成既能保留视觉身份又能体现个体特定表现行为的人类中心视频仍然是一个基本挑战。除了再现外观外，模型还必须复制表征主体随时间表达情感的面部行为。然而，大多数最先进的方法仅基于单一参考图像进行生成，而该图像并不包含这些时间动态的信息。因此，它们往往能够保留主体的视觉身份，但通常产生的表情变化有限且主体特异性较弱。为了解决这一问题，我们提出了BEACON，一个轻量级的特定主体视频生成框架，通过将视觉身份与表现行为解耦来生成更具表现力的视频。BEACON基于两个互补信号进行生成：一个编码身份的参考图像和一个捕捉特定主体面部动态的参考视频。通过对这些互补信号进行条件生成，BEACON生成的视频更好地保留了主体的外观和特征面部动态，同时支持身份与表现的转移。我们在MEAD和RAVDESS数据集上的实验表明，通过对大约2000对样本进行微调，并更新约1%的预训练Wan视频扩散模型，BEACON在保持竞争性身份保留的同时，提高了面部表现力，超越了最先进的视频生成方法。

View on arXiv Download PDF AI Translation

cs.CV / 22 / 2609.13265

Forward-Facing Near-Infrared Adds Little to Colour for Farm-Machinery Traversability: A Site-Disjoint Evaluation of Sensor-Dependent Spatial Leakage

前向近红外对农用机械可通行性贡献有限：传感器依赖的空间泄漏的场地不重叠评估

Kang, Sungwoo

Abstract

Near-infrared (NIR) imaging does not consistently outperform standard color cameras for daytime agricultural traversability once spatial data leakage is eliminated. Prior benchmarks suggesting an NIR advantage used sequence-level splits that permitted spatially autocorrelated imagery into test sets, artificially inflating NIR performance, especially on difficult paddy-boundary segmentation. When evaluated across strictly held-out recording sites using the AI Hub autonomous driving corpus, none of the four tested configurations (color, NIR, a luminance control, or their fusion) reliably surpasses standard color across any of the five traversability classes. Furthermore, apparent performance gains observed at a single held-out site completely fail to replicate when the held-out site is rotated, demonstrating that the observed benefits were site-specific artifacts rather than generalizable improvements. Correcting for spatial leakage erases the apparent lead of NIR entirely rather than uniformly degrading performance across all sensor types. Consequently, adding a forward-facing NIR camera to a daylight agricultural vehicle sensor suite remains unproven, and evaluations compromised by location leakage risk distorting sensor rankings and misleading procurement decisions.

Chinese Translation

近红外（NIR）成像在消除空间数据泄漏后，并未在白天农业可通行性方面始终优于标准彩色相机。之前的基准测试表明NIR具有优势，但使用的序列级拆分允许空间自相关图像进入测试集，从而人为地提高了NIR性能，特别是在困难的稻田边界分割任务中。当使用AI Hub自主驾驶数据集在严格保留的录制场地进行评估时，四种测试配置（彩色、NIR、亮度控制或它们的融合）在五个可通行性类别中均未能可靠地超过标准彩色相机。此外，在单个保留场地观察到的明显性能提升在更换保留场地时完全无法复制，表明观察到的好处是场地特定的伪影，而非可推广的改进。纠正空间泄漏后，NIR的明显优势完全消失，而不是在所有传感器类型中均匀降低性能。因此，在白天农业车辆的传感器套件中添加前向NIR相机仍未得到证明，受位置泄漏影响的评估可能会扭曲传感器排名并误导采购决策。

View on arXiv Download PDF AI Translation

cs.CV / 23 / 2609.13267

Structure-Token Evidence-Anchored Reasoning for Scientific Chart Understanding

基于结构-标记证据锚定推理的科学图表理解

Soarez, Alberlucia Rafael, Ferreira, Camila, Kim, Daniel, Costa, Mariana, Torres, Alejandro

Abstract

Scientific charts encode quantities in axes, legends, and geometric marks, yet large vision-language models still treat them as natural photographs. Visual in-context examples do not expose the coordinate frame; unconstrained chain-of-thought can name a plausible number that was never read from a bar. We present STEER (Structure-Token Evidence-anchored Reasoning), which freezes a Llama-3.2-Vision encoder and inserts three modules: a chart structure graph encoder (CSGE) that binds ticks, legend items, and marks; evidence-anchored step reasoning (EASR) that forces every arithmetic step to cite a graph node; and weak-parser strong-reasoner alignment (WPSR) that uses a specialized table extractor only as a teacher of node attributes. On ChartQA, STEER reaches 82.70 average relaxed accuracy versus 80.16 for ChartGemma and 76.40 for a LLaVA-CoT backbone trained on the same mix. Gains widen on CharXiv reasoning (33.60 vs. 29.20 InternVL Chat V1.5) and ChartQAPro CoT (40.70 vs. 37.17 Qwen2-VL-7B), where OCR shortcuts disappear. Ablations show that dropping node serialization or numeric candidate constraints undoes most of the reasoning lift.

Chinese Translation

科学图表通过坐标轴、图例和几何标记编码数量，但大型视觉-语言模型仍将其视为自然照片。视觉上下文示例未能暴露坐标框架；不受限制的思维链可能会给出一个从未从条形图中读取的合理数字。我们提出了STEER（结构-标记证据锚定推理），该方法冻结了Llama-3.2-Vision编码器，并插入了三个模块：图表结构图编码器（CSGE），用于绑定刻度、图例项和标记；证据锚定步骤推理（EASR），强制每个算术步骤引用一个图节点；以及弱解析器强推理器对齐（WPSR），该模块仅将专门的表提取器用作节点属性的教师。在ChartQA上，STEER达到了82.70的平均放宽准确率，而ChartGemma为80.16，LLaVA-CoT主干在相同数据集上的准确率为76.40。在CharXiv推理中，STEER的优势进一步扩大（33.60对比29.20的InternVL Chat V1.5），在ChartQAPro CoT中也表现出色（40.70对比37.17的Qwen2-VL-7B），此时OCR捷径消失。消融实验表明，去掉节点序列化或数字候选约束会大幅降低推理能力。

View on arXiv Download PDF AI Translation

cs.CV / 24 / 2609.13268

Capability-Routed Visual Retrieval and Evidence Threading for Long-Context Document Question Answering

基于能力的视觉检索与证据线程连接用于长上下文文档问答

Rahman, Amirul, Karim, Aisha, Nakamura, Kenji, Ng, Yi-Fan

Abstract

Annual reports, diligence packs, and infographic dashboards bury numbers in page images: axes, cell grids, and footnotes that OCR pipelines flatten and that page-level visual retrievers still treat as interchangeable in-context examples. We keep a frozen Qwen2.5-VL-7B-Instruct generator and a ColPali / VisRAG-Ret page index, and insert three modules. A capability-aware visual router (CAVR) tags each retrieved page as text, table, chart, layout, or mixed and mixes specialist experts before generation. Weak-to-strong page selection (WSPS) distils a frozen 7B answerability teacher into a 3B selection head so ranking is no longer a single InfoNCE score. Visual evidence threading (VET) builds layout-anchored paths of length at most three and lets the generator read the thread rather than a flat top-$k$ list. On gold-page DocVQA / ChartQA / InfographicVQA the 7B system reaches 96.3 / 90.1 / 85.4. Under the VisRAG top-3 protocol the mean generation accuracy is 62.74 versus 59.39 for the same backbone with concatenation. On MMLongBench-Doc retrieve-then-read, F1 moves from 19.2 to 22.6 and multi-page accuracy from 16.4 to 21.2. ViDoRe nDCG@5 after WSPS reranking is 83.6, with TAT-DQA financial reports at 70.4.

Chinese Translation

年度报告、尽职调查包和信息图表仪表板将数字埋藏在页面图像中：坐标轴、单元格网格和脚注被光学字符识别（OCR）管道压平，而页面级视觉检索器仍将其视为可互换的上下文示例。我们保持一个冻结的 Qwen2.5-VL-7B-Instruct 生成器和一个 ColPali / VisRAG-Ret 页面索引，并插入三个模块。能力感知视觉路由器（CAVR）将每个检索到的页面标记为文本、表格、图表、布局或混合，并在生成之前混合专业专家。弱到强页面选择（WSPS）将一个冻结的 7B 答案能力教师提炼成一个 3B 选择头，因此排名不再是单一的 InfoNCE 分数。视觉证据线程连接（VET）构建长度最多为三的布局锚定路径，并让生成器读取该线程，而不是一个平坦的 top-$k$ 列表。在金标准页面 DocVQA / ChartQA / InfographicVQA 上，7B 系统分别达到 96.3 / 90.1 / 85.4。在 VisRAG top-3 协议下，平均生成准确率为 62.74，而同一骨干网络的拼接为 59.39。在 MMLongBench-Doc 检索后再读中，F1 从 19.2 提升到 22.6，多页面准确率从 16.4 提升到 21.2。经过 WSPS 重新排名后的 ViDoRe nDCG@5 为 83.6，TAT-DQA 财务报告为 70.4。

View on arXiv Download PDF AI Translation

cs.CV / 25 / 2609.13269

Frame-Synchronous Hand Gesture Detection by Projected Winding Order

通过投影缠绕顺序实现帧同步的手势检测

Thakur, Amey

Abstract

Gesture recognition on video is normally posed as classification: label each frame, then act on the label. That is adequate for control, where a command may be obeyed several frames late without a user noticing, and inadequate for synchronisation, where an output must be aligned to the frame on which the gesture physically occurred. We take the synchronisation problem for one common movement, the rotation of an open hand about its long axis, and show that it admits an exact solution needing no classifier, no training data and no calibration. Let s be the normalised two-dimensional cross product of the two palm edges at the wrist and the outer knuckles, under the projection the camera already performs. We prove that s factorises as k(theta)cos(theta) with |k| > 0 everywhere, so s vanishes exactly when the palm is edge-on and its sign tracks the face presented to the camera. Detection is therefore a zero crossing of one scalar, which yields an instant rather than an interval, and we prove the criterion invariant to image mirroring, hand scale and handedness, all from its algebraic form rather than from the landmark estimator. A two-handed interaction uses four fingertips as a window onto a restyled version of the same scene. We give a coverage predicate that stays correct when the hands cross, where the usual triangulation does not, and derive each free parameter of the three stylisation operators from a noise statistic rather than by inspection. Against a corpus with exact ground truth the criterion detects 95% of flips with no false positive in 240 near-miss sequences, and places each within 6.7 ms on average of the instant it occurred, a sixth of the interval at which the hand is observed, against 23.0 ms for reporting the bracketing sample. Reporting an event more finely than one samples follows from treating a gesture as the zero of a continuous quantity rather than a label.

Chinese Translation

视频中的手势识别通常被视为分类问题：为每一帧标记标签，然后根据标签进行操作。这对于控制是足够的，因为一个命令可能在几帧后被执行而用户并未察觉，但对于同步则不够，因为输出必须与手势实际发生的帧对齐。我们将同步问题应用于一种常见的动作，即手掌绕其长轴的旋转，并展示出一种无需分类器、训练数据和校准的精确解决方案。设 s 为在手腕和外关节处两个掌边的归一化二维叉积，基于相机已经执行的投影。我们证明 s 可以因式分解为 k(theta)cos(theta)，其中 |k| 在任何地方均大于 0，因此当手掌与相机呈边缘对齐时 s 恰好为零，其符号跟踪呈现给相机的面。检测因此是一个标量的零交叉，这产生一个瞬时而非时间间隔，我们证明该标准对图像镜像、手的比例和左右手性不变，所有这些都是基于其代数形式而非地标估计器。双手交互使用四个指尖作为同一场景的重新风格化版本的窗口。我们给出一个覆盖谓词，当双手交叉时仍然保持正确，而通常的三角测量则不然，并且从噪声统计中推导出三个风格化操作符的每个自由参数，而不是通过观察。针对具有精确真实值的语料库，该标准在 240 个近失序列中检测到 95% 的翻转且没有假阳性，并且平均将每个翻转定位在其发生瞬间的 6.7 毫秒内，这仅为观察到手的时间间隔的六分之一，而报告括样本的时间为 23.0 毫秒。比样本更精细地报告事件源于将手势视为一个连续量的零而非标签。

View on arXiv Download PDF AI Translation

cs.CV / 26 / 2609.13271

CANAL: Channel-Aware Noise Allocation for Differentially Private Feature Distillation in Medical Image Segmentation

CANAL：用于医学图像分割的差分隐私特征蒸馏中的通道感知噪声分配

Butt, Armaghan, Feng, Shuya, Tian, Qing

Abstract

Medical image segmentation needs diverse training data, but hospitals hold complementary scans they cannot share for privacy and regulatory reasons. Knowledge distillation can bridge this gap by exporting learned feature representations instead of images, but those representations still encode patient-specific anatomy and remain vulnerable to membership-inference and feature-inversion attacks. Adding calibrated Gaussian noise restores a differential-privacy guarantee, yet three issues have been overlooked. First, prior DP feature-distillation pipelines re-sample noise at every student iteration, so each patient image is released many times and the privacy cost composes over those releases, growing by orders of magnitude. We present a sample-once-per-image release, realized by a single precomputation pass, under which each patient contributes one release. Second, uniform noise is wasteful because channels differ in task importance. Using task-gradient energy as the importance measure, we derive CANAL, a closed-form water-filling allocation that gives important channels proportionally less noise, and prove it strictly minimizes importance-weighted distortion at a fixed budget. Third, the clipping caps and importance scores that drive the allocation are themselves data-dependent, so releasing them in the clear silently breaks the guarantee. We give a DP-honest budget split that charges each to the privacy budget, so the reported epsilon is the true epsilon. Across three medical segmentation benchmarks spanning dermoscopy, colonoscopy, and ultrasound, CANAL retains more task-relevant signal than uniform noise at the same privacy budget.

Chinese Translation

医学图像分割需要多样化的训练数据，但医院因隐私和监管原因无法共享互补的扫描数据。知识蒸馏可以通过导出学习到的特征表示而不是图像来弥补这一差距，但这些表示仍然编码了特定患者的解剖结构，容易受到成员推断和特征反演攻击。添加经过校准的高斯噪声可以恢复差分隐私保证，但有三个问题被忽视。首先，先前的差分隐私特征蒸馏管道在每次学生迭代时重新采样噪声，因此每个患者图像被多次释放，隐私成本在这些释放中累加，增长几个数量级。我们提出了一种每个图像仅释放一次样本的方案，通过一次预计算实现，在该方案下，每个患者仅贡献一次释放。其次，均匀噪声是浪费的，因为通道在任务重要性上存在差异。我们使用任务梯度能量作为重要性度量，推导出CANAL，这是一种封闭形式的水填充分配方法，给予重要通道相对较少的噪声，并证明它在固定预算下严格最小化重要性加权失真。第三，驱动分配的裁剪上限和重要性评分本身是数据依赖的，因此清晰释放它们会悄然破坏保证。我们提供了一种差分隐私诚实的预算分配，将每个部分计入隐私预算，因此报告的epsilon是真实的epsilon。在涵盖皮肤镜检查、结肠镜检查和超声波的三个医学分割基准中，CANAL在相同隐私预算下保留了比均匀噪声更多的任务相关信号。

View on arXiv Download PDF AI Translation

cs.CV / 27 / 2609.13277

SomBench: Benchmark Dataset for Advancing Machine Learning in Lunar Science

SomBench：推动月球科学机器学习的基准数据集

Patil, Himanshu, Nyirjesy, Gabby, Slank, Rachel A., Gaur, Vishal, Szwarcman, Daniela, Fraccaro, Paolo, Dionelis, Nikolaos, Barker, Michael K., Annex, Andrew, Viswanathan, Vishnu, Morse, Zachary, Schaefer, Ethan I., Debary, Hiyam, Kumar, Ankur, Lal, Rohit, Dawson, Geoffrey, Watson, Campbell, Dawson-Rigas, Rebekah I., Maskey, Manil, Bernabé-Moreno, Juan, Ramachandran, Rahul, Roy, Sujit

Abstract

Lunar orbital missions, such as Lunar Reconnaissance Orbiter, Kaguya/SELENE, Gravity Recovery and Interior Laboratory, and Lunar Prospector, among others, provide rich multi-instrument observations, but their heterogeneity in sampling, projection, and conventions limits reproducible machine learning (ML). We introduce SomBench, a unified, spatially-aligned, ML-ready lunar dataset aggregating 30+ co-registered layers from ten instruments across four missions, spanning 1 meter to 20 kilometer/pixel and covering 82 degree latitude in 90 Lunar Transverse Mercator zones with two polar stereographic caps. An image-anchored tiling pipeline yields pretraining-ready multimodal tile views with leakage-safe splits, distributed as netCDF with Parquet catalogs. An application benchmark suite spans impact processes, volcanic history, and polar volatiles. Baseline experiments with ResNet-50 and SwinV2-B models confirm that each benchmark task is learnable from the released inputs, establishing reference points for future model development.

Chinese Translation

月球轨道任务，如月球勘测轨道器（Lunar Reconnaissance Orbiter）、嫦娥（Kaguya/SELENE）、重力恢复与内部实验室（Gravity Recovery and Interior Laboratory）以及月球探测器（Lunar Prospector）等，提供了丰富的多仪器观测数据，但由于采样、投影和约定的异质性，限制了可重复的机器学习（ML）。我们介绍了SomBench，这是一个统一的、空间对齐的、适合机器学习的月球数据集，汇集了来自四个任务的十个仪器的30多个共注册层，分辨率范围从1米到20公里/像素，覆盖90个月球横轴墨卡托区的82度纬度，并包含两个极地立体图层。一个基于图像的切片管道生成了预训练就绪的多模态切片视图，并提供安全的分割，数据以netCDF格式和Parquet目录分发。应用基准套件涵盖了冲击过程、火山历史和极地挥发物。使用ResNet-50和SwinV2-B模型的基线实验确认每个基准任务都可以从发布的输入中学习，为未来模型开发建立了参考点。

View on arXiv Download PDF AI Translation

cs.CV / 28 / 2609.13279

The MODA General Attribute Suite: A Four-Track Evaluation Benchmark for Fashion Attribute Extraction

MODA通用属性套件：用于时尚属性提取的四轨评估基准

Mitra, Arkid

Abstract

Fashion attribute extraction is evaluated inconsistently: results are reported as single aggregate numbers across image types that pose different problems, fields that are not visible in an image are scored as ordinary negatives, and the effect of vocabulary mismatch between datasets is acknowledged but not measured. We release the MODA General Attribute Suite, a four-track benchmark that keeps these problems separate by construction. Each track (localized garment crops, catalogue product images, applicability-aware full-body photographs, and product text) carries its own frozen test set, input contract, metric, and leakage unit, and the tracks are never averaged. The protocol requires label-blind prediction, SHA-256 commitment of prediction files before any label is opened, a fail-closed scorer, and a 10,000-sample paired cluster bootstrap at each track's natural leakage unit; promotion requires a positive interval on every track rather than a favourable mean. We release the scorers, the split builders, our own prediction files with their hashes including the runs we lose, and the MODA_NER(V) model checkpoints for three of the four tracks. The text model is not distributed; its benchmark is. We report baseline results, six interventions that did not improve them, and the limitations of the evidence.

Chinese Translation

时尚属性提取的评估存在不一致性：结果以单一聚合数字的形式报告，而这些数字来自于不同问题的图像类型，图像中不可见的领域被视为普通负例进行评分，数据集之间的词汇不匹配的影响被承认但未被测量。我们发布了MODA通用属性套件，这是一个通过构造将这些问题分开的四轨基准。每个轨道（局部服装裁剪、目录产品图像、考虑适用性的全身照片和产品文本）都有其独立的冻结测试集、输入契约、度量标准和泄漏单位，且轨道之间从不平均。该协议要求无标签预测，在任何标签打开之前对预测文件进行SHA-256承诺，采用失败关闭的评分机制，并在每个轨道的自然泄漏单位进行10,000样本的配对集群自助抽样；提升要求在每个轨道上都有正区间，而不是有利的均值。我们发布了评分器、分割构建器、我们自己的预测文件及其哈希值（包括我们失利的运行），以及三个轨道的MODA_NER(V)模型检查点。文本模型不进行分发；其基准则是可用的。我们报告了基线结果、六个未能改善结果的干预措施以及证据的局限性。

View on arXiv Download PDF AI Translation

cs.CV / 29 / 2609.13280

RxScribe Bench: A Multi-Axis Benchmark for Evaluating Vision-Language Models on Indian Outpatient Prescriptions

RxScribe Bench：评估印度门诊处方上视觉-语言模型的多轴基准

Somil, Saini, Vijay, Verma, Vidit, Riya, Batta, Aastha, Malhotra, Vibhuti, Khetan, Chayan, Mittal, Piyush, Poonia, Puneet

Abstract

Prescription transcription errors are not interchangeable. A model that fabricates a drug and a model that misreads a legible dose pose very different clinical risks, yet prescription-transcription accuracy is typically reported as a single blended figure that treats the two as equivalent. We introduce RxScribe Bench, a benchmark for evaluating vision-language models on handwritten prescription digitization that decomposes performance into four axes tied to clinical severity, rather than folding everything into a single aggregated score. Given only a prescription image and an output schema, a model produces a structured record, which is then compared field by field against a human-authored ground truth of identical shape, with each field also labeled for visibility and legibility. The four axes isolate distinct failure modes, namely Correctness, Hallucination, Engagement, and Robustness. The Robustness axis withholds its hard-field results rather than reporting one when the supporting sample falls below a minimum-evidence threshold. We evaluate frontier vision-language models on real prescriptions across independent cold runs per image, and find that no single model wins across all four axes.

Chinese Translation

处方转录错误并不可互换。一个虚构药物的模型与一个错误解读清晰剂量的模型所带来的临床风险截然不同，然而处方转录的准确性通常以一个混合的数字报告，将两者视为等同。我们提出了RxScribe Bench，这是一个用于评估手写处方数字化的视觉-语言模型的基准，它将性能分解为与临床严重性相关的四个轴，而不是将所有内容折叠成一个单一的综合评分。给定一张处方图像和一个输出模式，模型生成一个结构化记录，然后逐字段与相同形状的人类编写的真实数据进行比较，每个字段也标记了可见性和可读性。这四个轴隔离了不同的失败模式，即正确性（Correctness）、幻觉（Hallucination）、参与度（Engagement）和鲁棒性（Robustness）。鲁棒性轴在支持样本低于最低证据阈值时不报告其硬字段结果。我们在真实处方上对前沿视觉-语言模型进行评估，针对每张图像进行独立的冷启动运行，发现没有单一模型在所有四个轴上都表现优异。

View on arXiv Download PDF AI Translation

cs.CV / 30 / 2609.13282

GaugeDefect: Detecting Surface Anomalies by Curvature of Feature Transport

GaugeDefect：通过特征传输的曲率检测表面异常

Wang, Yefan

Abstract

Industrial anomaly localization has advanced rapidly with feature-based, reconstruction-based, and distillation-based methods. Most of these methods score a region by asking how unusual its local appearance or feature representation is with respect to normal training images. This is a strong and practical formulation. In this work, we study a complementary geometric cue for cases where an abnormal region may still contain locally plausible visual features. Thin scratches, small dents, and disrupted repeated patterns often do not make every local patch individually abnormal; instead, they disturb how nearby features vary and connect across the surface. We propose GaugeDefect, a geometric method for surface anomaly localization based on the curvature of feature transport. Given a feature lattice, we estimate a local feature frame at each node and compute orthogonal transports between neighboring frames. The accumulated transport around a small closed loop gives a holonomy matrix, whose deviation from identity measures feature-transport curvature. After calibration on normal training images, unusually large curvature indicates a local inconsistency in the feature field. The curvature here is not the physical curvature of the inspected object, but a representation-space measure of neighborhood inconsistency. This makes the method applicable to curved surfaces, textured materials, and non-planar industrial objects. Its main role is to improve localization of subtle surface disruptions, while often producing sharper responses near defect boundaries as a natural consequence of the curvature signal.

Chinese Translation

工业异常定位随着基于特征、基于重建和基于蒸馏的方法迅速发展。这些方法大多数通过询问某一区域的局部外观或特征表示相对于正常训练图像的异常程度来对区域进行评分。这是一种强大且实用的表述。在本研究中，我们探讨了一种补充的几何线索，适用于那些异常区域仍可能包含局部合理视觉特征的情况。细微划痕、小凹陷和破坏的重复图案往往不会使每个局部区域单独异常；相反，它们会干扰附近特征在表面上的变化和连接。我们提出了GaugeDefect，一种基于特征传输曲率的表面异常定位几何方法。给定一个特征格，我们在每个节点估计一个局部特征框架，并计算相邻框架之间的正交传输。围绕一个小闭合环的累积传输给出一个全局矩阵，其偏离单位矩阵的程度衡量特征传输的曲率。在对正常训练图像进行校准后，异常大的曲率表明特征场中的局部不一致性。这里的曲率不是被检对象的物理曲率，而是邻域不一致性的表示空间度量。这使得该方法适用于曲面、纹理材料和非平面工业物体。其主要作用是改善微小表面扰动的定位，同时通常在缺陷边界附近产生更清晰的响应，这是曲率信号的自然结果。

View on arXiv Download PDF AI Translation

cs.CV / 31 / 2609.13283

Multimodal-Multiresolution Foundation Model for Lunar Remote Sensing

用于月球遥感的多模态-多分辨率基础模型

Fraccaro, Paolo, Nyirjesy, Gabby, Szwarcman, Daniela, Patil, Himanshu, Gaur, Vishal, Lal, Rohit, Slank, Rachel A., Dawson, Geoffrey, Debary, Hiyam, Barker, Michael K., Annex, Andrew, Viswanathan, Vishnu, Morse, Zachary, Schaefer, Ethan I., Dionelis, Nikolaos, Kumar, Ankur, Watson, Campbell D., Maskey, Manil, Dawson-Rigas, Rebekah I., Bernabé-Moreno, Juan, Ramachandran, Rahul, Roy, Sujit

Abstract

We present a multimodal foundation model for lunar remote sensing, pretrained from scratch on SomBench, a geographically partitioned corpus of nearly two million co-registered tile bundles spanning 11 modalities at two spatial scales (1 m/pixel and 100 m/pixel). The model adapts the TerraMind masked-token architecture with two lunar-specific extensions: acquisition geometry is provided as explicit context, and meter- and hundred-meter-scale tiles are trained jointly so that a single set of weights covers both resolutions. FlexiViT patch embeddings allow adaptation to different patch sizes without retraining, while modality-wise inputs enable flexible multimodal fine-tuning. Qualitative generation experiments suggest the model learns meaningful cross-modal correspondences, including terrain derivatives from elevation and illumination-consistent reflectance from geometry. We evaluate on four benchmarks: crater detection at WAC and NAC scales, irregular mare patch (IMP) segmentation, and polar ice prospectivity regression. Across tasks, the pretrained model matches or outperforms ImageNet-pretrained baselines and an architecturally identical random-init control. On multimodal ice prospectivity regression, pretrained variants achieve the best results, while the random-init model outperforms most baselines, suggesting gains arise from both the architecture and pretraining. Label efficiency is notable for WAC crater detection, where the pretrained model trained on 50% of the data exceeds the strongest ImageNet baseline trained on the full dataset. Among adaptation strategies, LoRA matches or surpasses full fine-tuning on crater detection and IMP segmentation while using far fewer trainable parameters, whereas full fine-tuning performs best for ice prospectivity regression. We release the pretrained checkpoint, benchmark datasets, and fine-tuning code to support reproducible lunar AI research.

Chinese Translation

我们提出了一种用于月球遥感的多模态基础模型，该模型从零开始在SomBench上进行预训练，SomBench是一个地理分区的语料库，包含近两百万个共注册的图块包，涵盖11种模态，具有两种空间分辨率（1米/像素和100米/像素）。该模型采用了TerraMind的掩码令牌架构，并进行了两项与月球特定相关的扩展：获取几何形状作为显式上下文，并且米级和百米级的图块共同训练，使得一组权重能够覆盖这两种分辨率。FlexiViT图块嵌入允许在不重新训练的情况下适应不同的图块大小，而模态特定的输入则支持灵活的多模态微调。定性生成实验表明，该模型学习到了有意义的跨模态对应关系，包括从高程推导的地形衍生物和从几何形状中获得的光照一致反射率。我们在四个基准上进行了评估：WAC和NAC尺度的陨石坑检测、不规则月海斑块（IMP）分割以及极地冰层潜力回归。在各项任务中，预训练模型的表现与ImageNet预训练的基线相当或优于后者，以及一个在架构上完全相同的随机初始化控制模型。在多模态冰层潜力回归中，预训练变体取得了最佳结果，而随机初始化模型则超越了大多数基线，表明性能提升源于架构和预训练的共同作用。在WAC陨石坑检测中，标签效率显著，预训练模型在50%的数据上训练时超越了在完整数据集上训练的最强ImageNet基线。在适应策略中，LoRA在陨石坑检测和IMP分割中与完全微调相匹配或超越，同时使用的可训练参数远少于完全微调，而完全微调在冰层潜力回归中表现最佳。我们发布了预训练检查点、基准数据集和微调代码，以支持可重复的月球人工智能研究。

View on arXiv Download PDF AI Translation

cs.CV / 32 / 2609.13287

LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI Agents

LLaDA-UI：将块级扩散引入视觉-语言图形用户界面代理

Gu, Zhangxuan, Chen, Haoxing, Qin, Qi, Xin, Yi, Gan, Kai, Liu, Lin, Cui, Long, Wang, Xiaomei, Zhou, Beitong, Zhang, Yunzhu, Zeng, Zhengwen, Gao, Changlong, Chen, Weizhi, Zhang, Rongchao, Wu, Haoyuan, Shen, Shuheng, Meng, Changhua, Wang, Weiqiang, Li, Jianguo, Lan, Zhenzhong

Abstract

Diffusion large language models (dLLMs) achieve high decoding efficiency through block-parallel, arbitrary-order generation, making them attractive for latency-sensitive applications. GUI agents represent a natural testbed for this paradigm, as they must repeatedly perceive screen states and emit structured, spatially grounded actions in real time. However, whether dLLMs can be extended into capable multimodal GUI agents while preserving their parallel decoding advantage remains an open question. We present LLaDA-UI, a 16.7B-parameter MoE-based, block-wise diffusion vision-language GUI agent. LLaDA-UI follows a two-stage training pipeline: general multimodal pre-training aligns a native-resolution vision encoder with the LLaDA2.0-mini-base diffusion language backbone, followed by GUI-agent supervised fine-tuning on diverse mobile, desktop, web, and grounding data. Across widely adopted grounding benchmarks and navigation benchmarks spanning multiple platforms, LLaDA-UI substantially outperforms Qwen2.5-VL-7B and surpasses Qwen3-VL-8B on four of six reported GUI benchmarks. These results establish block-wise diffusion as a practical generative paradigm for multimodal GUI agents.

Chinese Translation

扩散大语言模型（dLLMs）通过块并行、任意顺序生成实现了高效解码，使其在对延迟敏感的应用中具有吸引力。图形用户界面（GUI）代理为这一范式提供了自然的测试平台，因为它们必须实时感知屏幕状态并发出结构化、空间定位的动作。然而，dLLMs是否能够在保持其并行解码优势的同时扩展为强大的多模态GUI代理仍然是一个未解的问题。我们提出了LLaDA-UI，一个基于MoE的、具有16.7亿参数的块级扩散视觉-语言GUI代理。LLaDA-UI遵循两阶段训练流程：通用多模态预训练将本地分辨率的视觉编码器与LLaDA2.0-mini-base扩散语言骨干网络对齐，随后在多样的移动、桌面、网络和基础数据上进行GUI代理的监督微调。在广泛采用的基础基准和跨多个平台的导航基准上，LLaDA-UI显著超越了Qwen2.5-VL-7B，并在六个报告的GUI基准中的四个上超过了Qwen3-VL-8B。这些结果确立了块级扩散作为多模态GUI代理的实用生成范式。

View on arXiv Download PDF AI Translation

cs.CV / 33 / 2609.13288

Target-Checked Reliability Score Refinement for Video Question Answering

针对视频问答的目标检查可靠性评分优化

Ren, Guoxiang, Chandra, Rohitash

Abstract

Video-language models can answer multiple-choice questions with high confidence yet be wrong. We study whether answer-level reliability scores can be improved under target shift without retraining the models or changing their answers. We collect option-probability lists from three fixed video-language models under four deterministic video samplings and represent cross-view changes and cross-model agreement as a response graph. Using a labeled target pilot, we compare the original score, defined as the probability assigned to the chosen answer, with a histogram-based gradient-boosting (HGB) score trained on the development datasets and a regularized logistic-regression score trained on the target pilot. A candidate replaces the original score only when repeated video-level checks indicate a positive, stable improvement. We develop this rule on public VideoQA benchmarks and Video Hallucination Diagnosis (VHD), a controlled diagnostic dataset for shared high-confidence errors. Ranking quality is measured by the area under the risk-coverage curve (AURC), where lower is better. On a held-out 963-question HERBench split, the method reduces mean AURC across the three models by 16.64% (95% confidence interval (CI), 12.12 to 22.61%); the smallest model-level gain is 11.39%. On a separate held-out 911-question Perception Test split, the mean reduction is 18.87% (95% CI, 15.43 to 22.14%). For InternVL3.5, the target check retains the original scores. Using the same outputs, the method outperforms seven training-free baselines in mean AURC on both datasets. It also improves AUROC, reduces calibration error, and lowers the error rate at 50% coverage by 6.50 and 6.58 percentage points.

Chinese Translation

视频语言模型能够以高置信度回答多项选择题，但仍可能出现错误。我们研究在目标转移的情况下，答案级别的可靠性评分是否可以在不重新训练模型或更改其答案的前提下得到改善。我们从三个固定的视频语言模型中收集了在四种确定性视频采样下的选项概率列表，并将跨视角变化和跨模型一致性表示为响应图。利用标记的目标样本，我们将原始评分（定义为分配给所选答案的概率）与在开发数据集上训练的基于直方图的梯度提升（HGB）评分以及在目标样本上训练的正则化逻辑回归评分进行比较。只有当重复的视频级检查表明存在积极且稳定的改善时，候选评分才会替代原始评分。我们在公共的视频问答基准和视频幻觉诊断（VHD）这一用于共享高置信度错误的受控诊断数据集上开发了这一规则。在风险覆盖曲线下的面积（AURC）中衡量排名质量，数值越低越好。在一个保留的963问题的HERBench划分中，该方法使三种模型的平均AURC降低了16.64%（95%置信区间（CI），12.12至22.61%）；最小的模型级增益为11.39%。在另一个保留的911问题的感知测试划分中，平均减少为18.87%（95% CI，15.43至22.14%）。对于InternVL3.5，目标检查保留了原始评分。使用相同的输出，该方法在两个数据集上在平均AURC方面优于七个无训练基线。它还改善了AUROC，减少了校准误差，并在50%覆盖率下将错误率降低了6.50和6.58个百分点。

View on arXiv Download PDF AI Translation

cs.CV / 34 / 2609.13293

Reliable Egocentric Action Anticipation via Temporal Reliability Suppression and Compositional Graph Decoding

通过时间可靠性抑制和组合图解码实现可靠的自我中心动作预测

Mohammadi, Mahsa, Rowlands, Sareh

Abstract

Wearable action anticipation systems must remain reliable despite missing frames, masking, and sensor noise, yet existing egocentric anticipation methods largely assume clean observations. We identify two complementary failure modes under temporal corruption: unreliable temporal evidence during encoding and implausible, low-support verb-noun compositions during decoding. We address them with a lightweight framework combining Temporal Reliability Suppression (TRS) and Robust Verb-Noun Graph (RVG) decoding. TRS predicts a per-frame suppression score from the projected input embedding and uses it as a learned key-side attention penalty at every encoder block and to derive reliability-weighted temporal pooling. RVG re-ranks verb-noun pairs using a PMI-based compatibility graph constructed from training labels. Under corruption-augmented training, TRS+CA+RVG reaches 29.1% average corrupted accuracy and 88.2% relative robustness across six corruptions, including three mechanisms absent during training, while reducing rare verb-noun predictions from 15.4% to 1.1%. Multi-seed and diagnostic experiments show that TRS responds to synthetic masking; shuffled-graph and frequency-only controls further indicate that RVG gains depend on genuine pairwise compatibility rather than marginal-frequency effects alone.

Chinese Translation

可穿戴动作预测系统必须在缺失帧、遮蔽和传感器噪声的情况下保持可靠性，而现有的自我中心预测方法在很大程度上假设观察数据是干净的。我们识别出在时间损坏下的两种互补失效模式：编码过程中不可靠的时间证据和解码过程中不合理、支持度低的动词-名词组合。我们通过一个轻量级框架来解决这些问题，该框架结合了时间可靠性抑制（Temporal Reliability Suppression, TRS）和鲁棒动词-名词图（Robust Verb-Noun Graph, RVG）解码。TRS从投影输入嵌入中预测每帧的抑制分数，并将其作为每个编码器块的学习关键侧注意力惩罚，进而导出可靠性加权的时间池化。RVG使用基于PMI（点互信息）的兼容性图对动词-名词对进行重新排序，该图是从训练标签构建的。在损坏增强训练下，TRS+CA+RVG在六种损坏情况下达到了29.1%的平均损坏准确率和88.2%的相对鲁棒性，包括三种在训练期间缺失的机制，同时将稀有动词-名词预测从15.4%减少到1.1%。多种种子和诊断实验表明，TRS对合成遮蔽有反应；洗牌图和仅频率控制进一步表明，RVG的增益依赖于真实的成对兼容性，而不仅仅是边际频率效应。

View on arXiv Download PDF AI Translation

cs.CV / 35 / 2609.13294

VectorHarness: Recovering Editable, Relation-Preserving Structure from Scientific Graphics

VectorHarness：从科学图形中恢复可编辑的、保持关系的结构

Tang, Jiahao, Song, Yiren, Wang, Alex Jinpeng

Abstract

Converting scientific graphics into editable representations remains a challenging problem for image-to-code generation because of their heterogeneous elements and complex layouts. Recent multi-agent reconstruction systems have advanced this line of work, but often follow a copy-paste paradigm: the reconstructed image closely resembles the original, while complex regions remain effectively uneditable. We instead formulate a different objective, raster-to-authoring reconstruction, which aims to recover an authoring representation that supports native, customized editing rather than mere visual replication. To this end, we present VectorHarness, a multi-agent framework for raster-to-authoring reconstruction that recovers heterogeneous components using type-appropriate native representations. Text, formulas, shapes, connectors, icons, charts, and tables are reconstructed as natively editable objects, while intrinsically image-based regions remain raster content. To systematically evaluate reconstruction quality, we introduce VectorHarness-Bench, which jointly assesses rendering fidelity, raster fallback coverage, executable object edits, and relation-preserving edits. Experiments show that VectorHarness improves executable edit success and relation preservation, reduces avoidable raster fallback, and maintains high visual fidelity across heterogeneous graphics.

Chinese Translation

将科学图形转换为可编辑的表示形式仍然是图像到代码生成中的一个挑战性问题，因为其异构元素和复杂布局。近期的多智能体重建系统在这一领域取得了一定进展，但通常遵循复制粘贴的范式：重建的图像与原始图像非常相似，而复杂区域仍然无法有效编辑。我们提出了一个不同的目标，即光栅到创作的重建，旨在恢复一种支持原生、定制编辑的创作表示，而不仅仅是视觉复制。为此，我们提出了VectorHarness，一个用于光栅到创作重建的多智能体框架，利用适当类型的原生表示恢复异构组件。文本、公式、形状、连接器、图标、图表和表格被重建为可原生编辑的对象，而内在的基于图像的区域则保持为光栅内容。为了系统地评估重建质量，我们引入了VectorHarness-Bench，它共同评估渲染保真度、光栅回退覆盖率、可执行对象编辑和关系保持编辑。实验表明，VectorHarness提高了可执行编辑的成功率和关系保持能力，减少了可避免的光栅回退，并在异构图形中保持了高视觉保真度。

View on arXiv Download PDF AI Translation

cs.CV / 36 / 2609.13296

Harnessing Image Question Dependence for Better VLM Test-time Reinforcement Learning

利用图像问题依赖性提升 VLM 测试时强化学习效果

He, Xinrui, Li, Ting-Wei, Wang, Junting, Ai, Mengting, He, Xinyu, Tong, Hanghang, He, Jingrui

Abstract

Test-time reinforcement learning can adapt vision-language models (VLMs) to unlabeled target data, but its effectiveness is fundamentally limited by the reliability of self-generated learning signals. To assess the reliability of consensus-based learning signals, we analyze VLM test-time reinforcement learning across diverse VQA datasets and model sizes, revealing two limitations. First, gains from consensus-based test-time training largely come from answer normalization rather than content correction. Second, many initial VLM responses are incorrect due to the model's limited ability to jointly use the image and the question; consensus rewards derived from these outputs may preserve the resulting grounding errors rather than correct them. Motivated by these, we propose TTIQ, a test-time reinforcement learning framework that harnesses image-question dependence for better vlm adaptation. TTIQ teacher-forces each sampled response under the original image-question pair and its image- and question-ablated variants, using the resulting token-level likelihood changes to estimate dependence on each input. It combines image and question dependence with calibrated confidence to construct a response-level reward that favors jointly grounded responses, and uses the token-level signals to assign greater positive policy credit to tokens supported by both inputs. This design favors responses that are jointly grounded in the image and the question and sufficiently confident, rather than merely popular. Experiments across eight VQA datasets and multiple VLM sizes show that TTIQ achieves the best average performance at every model scale. It further generalizes across VLM families, while models trained on one dataset improve performance on unseen datasets without further training.

Chinese Translation

测试时强化学习能够将视觉-语言模型（VLM）适应于未标记的目标数据，但其有效性根本上受到自生成学习信号可靠性的限制。为了评估基于共识的学习信号的可靠性，我们分析了不同 VQA 数据集和模型规模下的 VLM 测试时强化学习，揭示了两个局限性。首先，基于共识的测试时训练所带来的收益主要来自于答案规范化，而非内容修正。其次，由于模型在共同利用图像和问题方面的能力有限，许多初始 VLM 响应是错误的；从这些输出中获得的共识奖励可能会保留结果中的基础错误，而不是纠正它们。基于这些动机，我们提出了 TTIQ，一种利用图像-问题依赖性以提升 VLM 适应性的测试时强化学习框架。TTIQ 在原始图像-问题对及其图像和问题消融变体下强制每个采样响应，利用结果的标记级似然变化来估计对每个输入的依赖性。它结合了图像和问题的依赖性与校准的置信度，构建了一个响应级奖励，偏向于共同基础的响应，并利用标记级信号为由两个输入支持的标记分配更大的正政策信用。该设计偏向于在图像和问题中共同基础且置信度足够高的响应，而不仅仅是流行的响应。在八个 VQA 数据集和多个 VLM 尺寸上的实验表明，TTIQ 在每个模型规模上都达到了最佳的平均性能。它进一步在 VLM 家族之间具有良好的泛化能力，而在一个数据集上训练的模型在未见数据集上无需进一步训练即可提高性能。

View on arXiv Download PDF AI Translation

cs.CV / 37 / 2609.13298

Variational Template Matching with Statistical Fusion for Anomaly Detection in Patterned Structures

基于统计融合的变分模板匹配在模式结构异常检测中的应用

Xu, Qinwu, Jiang, Yifan

Abstract

Anomaly detection in structured images is challenging in small-data settings where deep learning approaches are costly or impractical. Classical template matching is simple and interpretable but lacks robustness to geometric variations such as scale, rotation, and perspective. We propose a variational template matching framework that represents anomaly templates as a family of transformed instances and performs detection via normalized cross-correlation over this transformation space. To further improve robustness, we introduce a density-based statistical anomaly score derived from local intensity distributions using kernel density estimation (KDE). This produces a smooth representation that captures distributional concentration and tail behavior more robustly than histogram-based methods. The structural and statistical signals are integrated through a unified fusion formulation, enabling complementary modeling of geometric similarity and distributional deviation. Experiments on biological cell images demonstrate that the proposed method outperforms classical baselines and achieves competitive performance with ResNet-50 under a fully training-free setting, while providing explicit localization. The approach offers an efficient, interpretable, and practical solution for anomaly detection in structured image domains.

Chinese Translation

在结构化图像中进行异常检测在小数据环境下具有挑战性，因为深度学习方法成本高昂或不切实际。经典的模板匹配方法简单且易于解释，但对尺度、旋转和透视等几何变化缺乏鲁棒性。我们提出了一种变分模板匹配框架，将异常模板表示为一系列变换实例，并通过在该变换空间上的归一化互相关进行检测。为了进一步提高鲁棒性，我们引入了一种基于密度的统计异常评分，该评分是通过使用核密度估计（KDE）从局部强度分布中派生的。这种方法生成了一种平滑的表示，能够比基于直方图的方法更稳健地捕捉分布集中和尾部行为。结构信号和统计信号通过统一的融合公式进行整合，从而实现几何相似性和分布偏差的互补建模。在生物细胞图像上的实验表明，所提出的方法优于经典基线，并在完全无训练的设置下与ResNet-50达成竞争性能，同时提供明确的定位。该方法为结构化图像领域的异常检测提供了一种高效、可解释且实用的解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 38 / 2609.13303

Adaptive Conformal Redistribution for Inter-class Transitional Uncertainty in Medical Image Classification

用于医学图像分类中类别间过渡不确定性的自适应保形重分配

Ghosh, Saibal, Bhattacharya, Samarup, Saha, Sanjoy Kumar, Pal, Umapada, Chakraborti, Tapabrata

Abstract

Medical image classification is frequently complicated by transitional categories whose feature distributions overlap those of adjacent classes, producing ambiguous decision boundaries. Conformal prediction returns uncertainty-aware prediction sets, but these are not directly actionable in clinical screening, where a single decision is required. This work proposes adaptive conformal redistribution (AdaConRed), a label-free post-conformal decision rule that converts ambiguous prediction sets into refined class assignments. A five-stage pipeline is developed. Vision-language generative augmentation addresses minority-class scarcity; a frozen DermFoundation encoder provides embeddings; a lightweight multi-layer perceptron performs classification; an entropy-modulated, margin-aware nonconformity score constructs adaptive prediction sets; samples predicted as transitional with multi-label sets are reassigned to the most probable alternative class within the set, using only model outputs at inference. Evaluation uses the OSCC oral lesion and ISIC skin lesion benchmarks at a miscoverage level of 0.2. On the 3-class OSCC benchmark, overall accuracy improves from 73.54% to 77.38%, with oral cancer accuracy rising from 64.29% to 82.14% and benign accuracy from 56.57% to 70.20%. Reassignment of transitional samples reduces OPMD accuracy from 84.78% to 80.16%, consistent with the asymmetric cost of missed malignancy. On ISIC, overall accuracy improves from 85.83% to 87.19%, melanoma accuracy rising from 66.04% to 68.34%. AdaConRed outperforms LAC, APS and RAPS under an identical backbone and redistribution rule. Conformal prediction can be extended beyond uncertainty quantification toward actionable decision support where transitional disease categories are present, with gains concentrated in the clinically critical malignant categories. Code repository: https://github.com/saibal436ghosh/AdaConRed.

Chinese Translation

医学图像分类常常受到过渡类别的影响，这些类别的特征分布与相邻类别重叠，从而产生模糊的决策边界。保形预测返回基于不确定性的预测集，但这些在临床筛查中并不可直接应用，因为需要做出单一决策。本研究提出了一种自适应保形重分配（Adaptive Conformal Redistribution, AdaConRed）方法，这是一种无标签的后保形决策规则，将模糊的预测集转化为精确的类别分配。我们开发了一个五阶段的流程。视觉-语言生成增强技术解决了少数类别的稀缺问题；冻结的DermFoundation编码器提供嵌入；轻量级多层感知器进行分类；熵调制的边际感知非保形分数构建自适应预测集；被预测为过渡类别的样本在推理时仅使用模型输出，重新分配到集合中最可能的替代类别。评估使用了OSCC口腔病变和ISIC皮肤病变基准，误覆盖率为0.2。在3类OSCC基准上，整体准确率从73.54%提高到77.38%，口腔癌准确率从64.29%上升至82.14%，良性准确率从56.57%提高至70.20%。过渡样本的重新分配使OPMD准确率从84.78%降至80.16%，与漏诊恶性肿瘤的非对称成本一致。在ISIC上，整体准确率从85.83%提高到87.19%，黑色素瘤准确率从66.04%上升至68.34%。在相同的基础架构和重分配规则下，AdaConRed的表现优于LAC、APS和RAPS。保形预测可以超越不确定性量化，向可操作的决策支持扩展，特别是在存在过渡疾病类别时，收益集中在临床关键的恶性类别上。代码库：https://github.com/saibal436ghosh/AdaConRed.

View on arXiv Download PDF AI Translation

cs.CV / 39 / 2609.13306

HGSQ: Heatmap-Guided Sparse Query Detector for Real-Time Aerial Small Object Detection

HGSQ：热图引导的稀疏查询检测器用于实时空中小物体检测

Zeng, Yangchen

Abstract

Real-time aerial small object detection is an important visual signal and image processing problem, requiring a detector to preserve fine-grained localization while avoiding redundant computation on large background regions. This paper focuses on this deployment-oriented aerial/UAV setting rather than claiming a universal detector for all object detection scenarios. Existing Transformer-based detectors provide strong global modeling, but their dense query initialization and multi-layer decoder still spend substantial computation on background tokens, which is inefficient when small objects occupy only sparse image regions. To address this problem, this paper proposes HGSQ, a Heatmap-Guided Sparse Query Detector for real-time aerial small object detection. HGSQ uses a lightweight Heatmap Budget Predictor (HBP) to predict a foreground budget map in a single forward pass. The predicted heatmap is then used by three fixed components: Heatmap-Guided Sparse Query Selection (HSQS), which initializes decoder queries from high-confidence foreground positions; Heatmap-Gated Lite Snake Convolution (HGLSConv), which performs local shape refinement only on heatmap-activated small-object regions; and Adaptive Query-Decoder Budgeting (AQDB), which adjusts the query budget and decoder depth according to the estimated object density. Unlike post-hoc heatmap generation, HGSQ treats the heatmap as a real-time computation budget rather than a visualization map during deployment. Experiments on NWPU VHR-10 and VisDrone2019 show that HGSQ achieves 95.10 mAP50 on NWPU VHR-10 and 54.8 mAP50 on VisDrone2019, while reducing GFLOPs to 48.6 and running at 96.0 FPS on an RTX 4070 under our TensorRT FP16 deployment protocol.

Chinese Translation

实时空中小物体检测是一个重要的视觉信号与图像处理问题，要求检测器在保持精细定位的同时，避免对大背景区域的冗余计算。本文关注于这一面向部署的空中/无人机设置，而不是声称为所有物体检测场景提供通用检测器。现有的基于Transformer的检测器提供强大的全局建模能力，但其稠密查询初始化和多层解码器仍在背景标记上消耗大量计算，这在小物体仅占据稀疏图像区域时效率低下。为了解决这一问题，本文提出了HGSQ，一种用于实时空中小物体检测的热图引导稀疏查询检测器。HGSQ使用轻量级热图预算预测器（Heatmap Budget Predictor, HBP）在一次前向传递中预测前景预算图。然后，预测的热图被三个固定组件使用：热图引导稀疏查询选择（Heatmap-Guided Sparse Query Selection, HSQS），从高置信度前景位置初始化解码器查询；热图门控轻量蛇形卷积（Heatmap-Gated Lite Snake Convolution, HGLSConv），仅在热图激活的小物体区域执行局部形状细化；以及自适应查询-解码器预算（Adaptive Query-Decoder Budgeting, AQDB），根据估计的物体密度调整查询预算和解码器深度。与事后生成热图不同，HGSQ在部署过程中将热图视为实时计算预算，而非可视化图。NWPU VHR-10和VisDrone2019上的实验表明，HGSQ在NWPU VHR-10上达到了95.10 mAP50，在VisDrone2019上达到了54.8 mAP50，同时将GFLOPs降低到48.6，并在我们的TensorRT FP16部署协议下以96.0 FPS运行于RTX 4070上。

View on arXiv Download PDF AI Translation

cs.CV / 40 / 2609.13308

GroundBench: A Factorized, Counterfactual Benchmark for Locating VLM Affordance Failures

GroundBench：一个分解的、反事实基准用于定位 VLM 可用性失败

Sattigeri, Sarthak

Abstract

A companion evaluation found that naming the target part in a manipulation prompt increased action accuracy by 0.32-0.63 across eight vision-language models, with no model outperforming a constant baseline until the part was named. However, naming the part supplies information that a real system must infer, confounding visual grounding, mechanical reasoning, and category-to-action association. We introduce GroundBench, a diagnostic benchmark that separates these explanations through six branch-and-merge conditions, each adding a controlled information bundle, and a counterfactual re-ask targeting a real alternate part visible in the same image. Across three OpenAI models and 1,068 predictions, supplying the target region without its identity leaves action accuracy at or below the 0.53 majority baseline (0.26, 0.26, and 0.53), although the models largely reproduce the supplied region. Supplying identity without location instead yields 0.74, 0.68, and 0.68. Every above-baseline gain in this curated set occurs where the supplied part category itself determines the action. A no-vision control leaves GPT-5's scores unchanged or improved, providing evidence consistent with substantial category-to-action association. GPT-4o mini declines on one condition, so this interpretation is not universal. Adding joint type and motion axis does not improve accuracy across six model-stratum comparisons. On 74 counterfactual pairs from 32 objects, GPT-5 achieves 0.86 pair-weighted compliance with a 0.07 shortcut rate but fails all observed push-to-lift-vertical cases. GroundBench identifies which supplied information changes affordance behavior and tests whether apparently grounded performance can be reproduced through textual shortcuts.

Chinese Translation

一项辅助评估发现，在操作提示中命名目标部件使得八个视觉-语言模型的动作准确率提高了 0.32-0.63，且在部件被命名之前没有模型超越常量基线。然而，命名部件提供了一个真实系统必须推断的信息，这使得视觉定位、机械推理和类别与动作的关联变得复杂。我们引入了 GroundBench，一个诊断基准，通过六个分支与合并条件分离这些解释，每个条件添加一个受控的信息包，以及一个针对同一图像中可见的真实替代部件的反事实重询。在三个 OpenAI 模型和 1,068 次预测中，提供目标区域而不提供其身份使得动作准确率保持在或低于 0.53 的大多数基线（0.26、0.26 和 0.53），尽管模型在很大程度上再现了提供的区域。而仅提供身份而不提供位置则分别产生了 0.74、0.68 和 0.68 的准确率。在这个精心策划的集合中，所有超出基线的增益发生在提供的部件类别本身决定动作的情况下。一个无视觉控制的实验使得 GPT-5 的得分保持不变或有所提高，提供了与显著的类别与动作关联一致的证据。GPT-4o mini 在一个条件下下降，因此这一解释并非普遍适用。添加联合类型和运动轴并未提高六个模型层次比较中的准确率。在来自 32 个对象的 74 对反事实中，GPT-5 在 0.07 的捷径率下实现了 0.86 的成对加权合规性，但在所有观察到的推举-提升-垂直案例中均未成功。GroundBench 确定了哪些提供的信息改变了可用性行为，并测试了看似有根据的表现是否可以通过文本捷径再现。

View on arXiv Download PDF AI Translation

cs.CV / 41 / 2609.13309

Task-Based CT Protocol Optimization Using Reinforcement Learning and Virtual Imaging Trials

基于任务的CT协议优化：使用强化学习和虚拟成像试验

Zou, Jiaqi, Fenwick, David, Tarokh, Vahid, Felice, Nicholas, Rajagopal, Jayasai, Kapadia, Anuj, Samei, Ehsan, NaderiAlizadeh, Navid, Abadi, Ehsan

Abstract

Protocol optimization in computed tomography (CT) aims to improve diagnostic image quality while reducing radiation dose, but the interdependence of acquisition and reconstruction parameters makes exhaustive testing impractical. We propose a virtual imaging trial framework with reinforcement learning for efficient CT protocol optimization. Sixty-three computational human models with liver lesions were imaged using a validated CT simulator across 468 combinations of acquisition and reconstruction parameters, including tube voltage, tube current, reconstruction kernel, slice thickness, and pixel size. The optimization objective balanced liver lesion detectability, quantified by detectability index d-prime, against radiation dose. A Proximal Policy Optimization agent was trained and conditioned on patient-specific CT localizer embeddings derived from a pretrained vision transformer. On held-out patients, evaluating only 8 protocols per patient, about 2% of exhaustive testing, recovered 98.2% of the exhaustive-search oracle objective. With no patient-specific simulation, surrogate scoring alone achieved 89.7% recovery. Conditioning on the localizer improved zero-simulation recovery by 10.7 percentage points over the localizer-blind policy (paired 95% CI 2.9-19.5; p=0.02). These results show that the proposed framework can substantially reduce exhaustive protocol testing while enabling task-based, dose-aware protocol selection before the diagnostic scan.

Chinese Translation

计算机断层扫描（CT）中的协议优化旨在提高诊断图像质量，同时减少辐射剂量，但采集和重建参数之间的相互依赖性使得全面测试变得不切实际。我们提出了一种基于强化学习的虚拟成像试验框架，以实现高效的CT协议优化。使用经过验证的CT模拟器对63个具有肝脏病变的计算人模型进行了成像，涵盖了468种采集和重建参数组合，包括管电压、管电流、重建核、切片厚度和像素大小。优化目标在肝脏病变可检测性（通过可检测性指数d-prime量化）与辐射剂量之间取得平衡。训练了一个近端策略优化（Proximal Policy Optimization）代理，并基于从预训练视觉变换器中提取的患者特定CT定位器嵌入进行条件化。在保留的患者中，仅评估每位患者8个协议，约占全面测试的2%，恢复了98.2%的全面搜索oracle目标。在没有患者特定模拟的情况下，仅依靠替代评分就达到了89.7%的恢复率。对定位器的条件化使得零模拟恢复率比盲定位器策略提高了10.7个百分点（配对95%置信区间2.9-19.5；p=0.02）。这些结果表明，所提出的框架可以显著减少全面协议测试，同时在诊断扫描之前实现基于任务的、考虑辐射剂量的协议选择。

View on arXiv Download PDF AI Translation

cs.CV / 42 / 2609.13328

Pedestrian Crossing Intent Classification From Event-Based Vision Using Convolutional Spiking Neural Networks With Temporal Augmentation

基于事件视觉的行人过马路意图分类：使用带时间增强的卷积脉冲神经网络

Teklu, Henok, Sakhai, Mustafa, Wielgosz, Maciej, Mertik, Matej

Abstract

Anticipating whether a pedestrian will cross the road is safety-critical for autonomous vehicles, requiring real-time inference under challenging conditions including motion blur, high dynamic range, and class imbalance. Conventional frame-based deep networks process redundant RGB data at fixed frame rates, limiting their temporal resolution and energy efficiency. In this work we present an end-to-end pipeline that (i) converts real-world driving footage from the Joint Attention in Autonomous Driving (JAAD) dataset into synthetic dynamic vision sensor (DVS) event streams using the v2e simulator, (ii) augments training with the CARLA-simulated DVS sequences of the DVS-PedX dataset under both normal and adverse weather conditions, and (iii) trains a novel convolutional spiking neural network (Conv-SNN) with clip-consistent DVS augmentation to classify pedestrian crossing intent as binary: crossing or non-crossing. We detail all architectural decisions, the exact leaky-integrate-and-fire neuron dynamics with surrogate-gradient learning, the class-balanced loss formulation, JAAD oversampling at 6x, and a 70/15/15 stratified splitting protocol. The trained model achieves 95.83% accuracy and F1 = 0.9695 on the JAAD DVS test set, 97.79% accuracy and F1 = 0.9478 on normal CARLA DVS, and 94.78% accuracy and F1 = 0.8369 on adverse-weather CARLA DVS, all from a 1.07M-parameter architecture trained on CPU. Compared to prior frame-based approaches on JAAD, our method closes or surpasses the reported accuracy while operating natively on sparse temporal representations. We include a thorough analysis of the convergence behaviour across all 15 training epochs, domain transfer characteristics, and a quantitative comparison with representative related work.

Chinese Translation

预测行人是否会过马路对于自动驾驶车辆至关重要，这要求在运动模糊、高动态范围和类别不平衡等挑战性条件下进行实时推断。传统的基于帧的深度网络以固定帧率处理冗余的RGB数据，限制了其时间分辨率和能效。在本研究中，我们提出了一种端到端的管道，(i) 使用v2e模拟器将来自联合注意力自动驾驶（JAAD）数据集的真实驾驶视频转换为合成动态视觉传感器（DVS）事件流，(ii) 在正常和恶劣天气条件下，通过CARLA模拟的DVS-PedX数据集的DVS序列增强训练，(iii) 训练一种新型卷积脉冲神经网络（Conv-SNN），通过一致的DVS增强将行人过马路意图分类为二元：过马路或不过马路。我们详细说明了所有架构决策、确切的漏积分火神经元动态与替代梯度学习、类别平衡损失公式、JAAD的6倍过采样以及70/15/15的分层划分协议。训练后的模型在JAAD DVS测试集上达到了95.83%的准确率和F1 = 0.9695，在正常CARLA DVS上达到了97.79%的准确率和F1 = 0.9478，在恶劣天气CARLA DVS上达到了94.78%的准确率和F1 = 0.8369，所有这些均来自于在CPU上训练的107万参数架构。与之前在JAAD上的基于帧的方法相比，我们的方法在稀疏时间表示上原生运行，同时缩小或超越了报告的准确率。我们还包括了对所有15个训练周期的收敛行为、领域转移特征的深入分析，以及与代表性相关工作的定量比较。

View on arXiv Download PDF AI Translation

cs.CV / 43 / 2609.13332

Global-Local Contextual Progressive Expansion Network for Martian Landslide Segmentation in Multimodal Remote Sensing Imagery

用于多模态遥感影像的火星滑坡分割的全局-局部上下文渐进扩展网络

Ramos, Leo Thomas, Paheding, Sidike, Reyes-Angulo, Abel A., A., Rajaneesh, S., Sajinkumar K., Sappa, Angel D., Oommen, Thomas

Abstract

Automated landslide segmentation on Mars is one of the important tasks for understanding its surface processes, and all will aid in future space exploration. However, it remains a relatively underexplored open challenge because landslide morphology is highly variable, foreground regions are often sparse or irregular, and orbital observations combine heterogeneous spectral and topographic cues. In this context, this work investigates the capability of deep learning to address Martian landslide segmentation through an extensive assessment of modern neural segmentation models. To the best of our knowledge, this is the first study to provide such a comprehensive exploration in this domain. We further propose TransCPLES, a U-shaped network that couples Contextual Progressive Layer Expansion feature extraction with Transformer-based contextual reasoning, enabling the model to capture local geomorphic patterns and broader spatial dependencies for more reliable landslide delineation. Experiments on MMLSv2, a seven-band multimodal Martian landslide dataset, show that TransCPLES achieves the best overall performance when evaluated on geographically distinct samples, with consistent delineation across different landslide extents, stable foreground discrimination, and a favorable balance between accuracy and computational cost compared with several state-of-the-art convolutional, attention-based, and Transformer-based segmentation models. With this work, we hope to provide a useful reference and encourage further research and development in deep learning for planetary remote sensing. Code will be available after publication.

Chinese Translation

火星上的自动滑坡分割是理解其表面过程的重要任务之一，并将有助于未来的太空探索。然而，由于滑坡形态高度可变、前景区域通常稀疏或不规则，以及轨道观测结合了异质的光谱和地形线索，这一挑战仍然相对未被充分探索。在此背景下，本研究探讨了深度学习在火星滑坡分割中的能力，通过对现代神经分割模型的广泛评估。根据我们所知，这是该领域首次提供如此全面的探索。我们进一步提出了TransCPLES，一个U型网络，将上下文渐进层扩展特征提取与基于Transformer的上下文推理相结合，使模型能够捕捉局部地貌模式和更广泛的空间依赖关系，从而实现更可靠的滑坡划分。在七波段多模态火星滑坡数据集MMLSv2上的实验表明，TransCPLES在地理上不同样本的评估中实现了最佳的整体性能，在不同滑坡范围内具有一致的划分、稳定的前景区分，以及与几种最先进的卷积、基于注意力和基于Transformer的分割模型相比，在准确性和计算成本之间保持了良好的平衡。通过这项工作，我们希望提供有用的参考，并鼓励在行星遥感领域进一步的深度学习研究与开发。代码将在发表后提供。

View on arXiv Download PDF AI Translation

cs.CV / 44 / 2609.13340

ProtoCAM: Interpretable Few-Shot Mask-Guided Prototypical Learning for Breast Lesion Classification in Ultrasound Imaging

ProtoCAM：可解释的少样本掩膜引导原型学习用于超声成像中的乳腺病变分类

Ebadi, Ashkan

Abstract

Breast ultrasound imaging plays an important role in the early detection and diagnosis of breast cancer, particularly for patients with dense breast tissue. However, developing reliable deep learning models for ultrasound analysis is challenging due to limited annotated medical data and the need for interpretable predictions. To address these challenges, this paper proposes ProtoCAM, an explainable few-shot learning framework for breast lesion classification that integrates mask-guided feature encoding, prototypical metric learning, and gradient-based visual explanations. The proposed approach leverages lesion masks to guide feature extraction and constructs class prototypes within an embedding space to enable robust classification under limited training samples. The framework was evaluated on the BUSI dataset using a stratified group k-fold cross-validation protocol to prevent patient-level data leakage. Experimental results demonstrate ProtoCAM's high performance in low-data scenarios. In a 3-way 5-shot setting, the proposed method achieves a macro F1-score of 0.910, representing a substantial improvement over standard supervised CNN models. Among the evaluated backbone networks, ResNet18 achieved the best performance, reaching a macro F1-score of 91.65% under a 15-shot configuration, providing interpretable insights into the classification decisions. These results highlight the potential of explainable few-shot learning frameworks for reliable computer-aided breast cancer diagnosis in data-scarce medical imaging environments.

Chinese Translation

乳腺超声成像在乳腺癌的早期检测和诊断中发挥着重要作用，尤其是对于乳腺组织致密的患者。然而，由于标注医学数据的有限性以及对可解释预测的需求，开发可靠的深度学习模型用于超声分析面临挑战。为了解决这些问题，本文提出了ProtoCAM，一种可解释的少样本学习框架，用于乳腺病变分类，集成了掩膜引导特征编码、原型度量学习和基于梯度的可视化解释。该方法利用病变掩膜引导特征提取，并在嵌入空间中构建类别原型，以在有限的训练样本下实现稳健的分类。该框架在BUSI数据集上进行了评估，采用分层组k折交叉验证协议以防止患者级数据泄露。实验结果表明，ProtoCAM在低数据场景下表现出色。在3-way 5-shot设置中，所提方法实现了0.910的宏F1-score，显著优于标准监督CNN模型。在评估的主干网络中，ResNet18在15-shot配置下达到了最佳性能，宏F1-score为91.65%，为分类决策提供了可解释的见解。这些结果突显了可解释的少样本学习框架在数据稀缺的医学成像环境中实现可靠计算机辅助乳腺癌诊断的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 45 / 2609.13397

ConeGaussian: Anti-Aliased Gaussian Ray-Tracing for Generic Central Cameras

ConeGaussian：通用中央相机的抗混叠高斯光线追踪

Zhang, Deheng, Shi, Letian, Yang, Runyi, Li, Zhendong, Sun, Lei, Wu, Kanzhi, Chhatkuli, Ajad, Paudel, Danda Pani, Van Gool, Luc

Abstract

In rendering, a camera is a sampling operator that maps each finite pixel to a bundle of rays. Different camera models change the geometry of this bundle, thus making a unified and faithful rendering formulation challenging. Consequently, Gaussian ray tracing supports generic cameras (with optical center) through their inverse ray mappings, yet typically reduces every pixel to a single center ray. This ignores the camera-dependent pixel footprint, causing aliasing under minification, while unconstrained Gaussians expose unsupported frequencies under magnification. We present ConeGaussian, a camera-model-agnostic anti-aliasing framework for Gaussian ray-based rendering. Instead of defining the pixel filter on a camera-specific image plane, ConeGaussian constructs an anisotropic footprint directly from neighboring rays produced by the camera's native inverse mapping. We derive a closed-form response under a locally linear, depth-local, moment-matched approximation of the finite pixel footprint, while the same geometry defines a per-Gaussian training-frequency floor. Notably, by construction, our filtering principle can be used unmodified across calibrated central camera models and multiple Gaussian ray-rendering backbones. Additionally, unlike in mip-splatting, our scene-space frequency floor and filtering enable trivial composition at render time, allowing us to remove excess blurring. On pinhole and strongly distorted fisheye captures, ConeGaussian consistently improves two distinct ray-based backbones, by up to 4.3 dB at 1/8 resolution, and reduces fisheye LPIPS by 30% where perspective screen-plane footprint formulations are not directly applicable.

Chinese Translation

在渲染中，相机是一个采样操作符，将每个有限的像素映射到一束光线。不同的相机模型改变了这束光线的几何形状，因此使得统一且忠实的渲染公式变得具有挑战性。因此，高斯光线追踪通过其逆光线映射支持通用相机（具有光学中心），但通常将每个像素简化为单一的中心光线。这忽略了相机依赖的像素足迹，在缩小时导致混叠，而不受约束的高斯在放大时暴露了不支持的频率。我们提出了ConeGaussian，这是一个与相机模型无关的抗混叠框架，用于基于高斯光线的渲染。ConeGaussian不在特定相机的图像平面上定义像素滤波器，而是直接从相机的原生逆映射生成的邻近光线构建各向异性足迹。我们在有限像素足迹的局部线性、深度局部、矩匹配近似下推导出封闭形式的响应，而相同的几何形状定义了每个高斯的训练频率下限。值得注意的是，通过构造，我们的滤波原则可以在经过校准的中央相机模型和多个高斯光线渲染骨干中无修改地使用。此外，与mip-splatting不同，我们的场景空间频率下限和滤波使得在渲染时的组合变得简单，从而允许我们去除多余的模糊。在针孔和强畸变鱼眼捕获中，ConeGaussian始终能显著改善两种不同的基于光线的骨干，在1/8分辨率下提高了最多4.3 dB，并在透视屏幕平面足迹公式不直接适用的情况下将鱼眼LPIPS降低了30%。

View on arXiv Download PDF AI Translation

cs.CV / 46 / 2609.13504

RIGOR: Rig-Informed Geometry for Omnidirectional Reconstruction

RIGOR：基于刚体信息的全向重建几何

Huang, Tingjun, Rudshin, Dmitry, Meyer, Mathieu, Bonazzi, Pietro, Pollefeys, Marc, Szymańska, Emilia

Abstract

Recent developments in feed-forward 3D reconstruction resulted in models which can recover dense scene representations and camera motion solely from an image stream. However, such predictions are prone to becoming inconsistent over long trajectories, specifically in demanding environments with repetitive structures, weak textures and dynamic objects or people. One way to mitigate those challenges is to use an omnidirectional camera, which provides wide spatial coverage and captures richer visual information. Yet, the majority of models do not offer support for 360-degree imagery or require additional fine-tuning. To bridge these two aspects, we present RIGOR: a large-scale reconstruction pipeline for gravity-aligned omnidirectional videos that retains a frozen feed-forward perspective backbone and exploits each panorama as a four-view virtual rig. The rig structure is used to detect and repair locally inconsistent predictions, to retrieve loop closures through cyclic four-view consensus, and to geometrically verify candidate revisits before global optimization. Verified constraints drive a Sim(3) pose graph that corrects accumulated rotation, translation, and scale drift along the sequence. We demonstrate that the proposed consistency mechanisms improve both trajectory accuracy and reconstructed geometry over a feed-forward baseline on challenging construction-site sequences. The code is made available under this link: https://github.com/TangentH/RIGOR.

Chinese Translation

近期在前馈式三维重建方面的发展使得模型能够仅通过图像流恢复密集的场景表示和相机运动。然而，这些预测在长轨迹上容易出现不一致，尤其是在具有重复结构、弱纹理和动态物体或人的复杂环境中。缓解这些挑战的一种方法是使用全向相机，它提供了广泛的空间覆盖并捕获更丰富的视觉信息。然而，大多数模型并不支持360度图像，或者需要额外的微调。为了解决这两个方面的问题，我们提出了RIGOR：一个针对重力对齐全向视频的大规模重建管道，它保留了一个冻结的前馈视角主干，并将每个全景视为一个四视图虚拟支架。该支架结构用于检测和修复局部不一致的预测，通过循环四视图共识检索环路闭合，并在全局优化之前几何验证候选重访。经过验证的约束驱动一个Sim(3)位姿图，纠正序列中的累计旋转、平移和尺度漂移。我们证明，所提出的一致性机制在具有挑战性的施工现场序列上提高了轨迹精度和重建几何，相较于前馈基线表现更佳。代码可通过以下链接获取：https://github.com/TangentH/RIGOR。

View on arXiv Download PDF AI Translation

cs.CV / 47 / 2609.13551

Towards Practical Precision Agriculture: Real-Time Fruit Detection and Video Analytics on Embedded Edge Hardware

迈向实用的精准农业：嵌入式边缘硬件上的实时水果检测与视频分析

Dimitrovski, Ivica, Spasev, Vlatko, Kitanovski, Ivan, Lameski, Petre, Boshev, Dane

Abstract

Static-image benchmarks do not capture the computational and temporal requirements of practical orchard video analytics. This study presents an end-to-end framework for real-time fruit detection, tracking, and counting on the NVIDIA Jetson Orin Nano Super. A lightweight YOLO26s detector is trained independently on four public datasets representing apples, mangoes, blueberries, and strawberries under a common protocol. The models are deployed on embedded platform using PyTorch and TensorRT at FP32, FP16, and INT8 precision. APPLE MOTS is then used for temporal video analytics because it provides orchard sequences with persistent fruit identities, enabling evaluation of multi-object tracking and unique-fruit counting. The selected FP16 TensorRT detector is integrated into an NVIDIA DeepStream pipeline combining hardware-accelerated decoding, ByteTrack tracking, and motion-aware line-crossing analytics. Across the four detection tasks, mean test mAP@50:95 ranges from 0.4957 to 0.8656. On the Jetson, TensorRT FP16 achieves 66.76-74.56 images/s at 13.41-14.98 ms prediction latency, while reducing mAP@50:95 by only 0.0020-0.0054 and gross energy consumption by approximately 64-66% relative to PyTorch FP32. The complete detector-tracker-analytics pipeline reaches 44.96-54.11 FPS and sustains the configured 30-FPS input rate without output-frame loss. On held-out orchard video sequences, HOTA ranges from 0.345 to 0.538, event-level counting F1 from 0.611 to 0.803, and relative count error from 6.2% to 51.6%. Performance varies across acquisition geometries: near-lateral row viewing yields the most stable tracking and counting, whereas forward traversal remains association- and recall-limited despite spatially adaptive counting geometry. These results show that practical edge-based fruit monitoring requires efficient detection and acquisition geometries that support reliable temporal association.

Chinese Translation

静态图像基准无法捕捉实用果园视频分析的计算和时间要求。本研究提出了一个端到端的框架，用于在 NVIDIA Jetson Orin Nano Super 上进行实时水果检测、跟踪和计数。一个轻量级的 YOLO26s 检测器在四个公共数据集上独立训练，这些数据集代表了苹果、芒果、蓝莓和草莓，采用统一的协议。模型使用 PyTorch 和 TensorRT 在 FP32、FP16 和 INT8 精度下部署在嵌入式平台上。随后，使用 APPLE MOTS 进行时间视频分析，因为它提供了具有持久水果身份的果园序列，从而能够评估多目标跟踪和独特水果计数。所选的 FP16 TensorRT 检测器集成到 NVIDIA DeepStream 管道中，结合了硬件加速解码、ByteTrack 跟踪和运动感知的过线分析。在四个检测任务中，平均测试 mAP@50:95 范围为 0.4957 到 0.8656。在 Jetson 上，TensorRT FP16 实现了 66.76-74.56 图像/秒的速度，预测延迟为 13.41-14.98 毫秒，同时 mAP@50:95 仅降低了 0.0020-0.0054，整体能耗相较于 PyTorch FP32 减少了约 64-66%。完整的检测-跟踪-分析管道达到了 44.96-54.11 FPS，并在不丢失输出帧的情况下维持配置的 30 FPS 输入速率。在保留的果园视频序列中，HOTA 范围为 0.345 到 0.538，事件级计数 F1 从 0.611 到 0.803，相对计数误差为 6.2% 到 51.6%。性能因获取几何形状而异：近侧视角的行视图提供了最稳定的跟踪和计数，而前向移动尽管具有空间自适应计数几何形状，但仍受到关联和召回的限制。这些结果表明，实用的基于边缘的水果监测需要高效的检测和获取几何形状，以支持可靠的时间关联。

View on arXiv Download PDF AI Translation

cs.CV / 48 / 2609.13617

From Advertised Improvements to Measured Capabilities: Evaluating ChatGPT Images 2.5 on Forgery Tasks

从宣传的改进到实际的能力：评估 ChatGPT Images 2.5 在伪造任务上的表现

Raj, Ankit, Zhang, Yuxin, Zewde, Kidus, Duong, Tommy, Gan, Jiaqi, Shen, Xingyu, Zhou, Yuchen, Guo, Huaiyu, Zhang, Siyu, Ren, Simiao

Abstract

We evaluate whether the improvements advertised for ChatGPT Images 2.5 translate into better performance on forgery tasks with predetermined answers. We compare its Flare and Sunburst API models with GPT-Image-2 re-run in the same week, using receipt-field edits, repeated editing, product placement and fine-print rendering. After image registration, Flare and Sunburst show fewer OCR-detected changes to surrounding receipt text (31.7% and 31.2% versus 44.2% for both GPT-Image-2 baselines), mainly on CORD receipts, without a detectable improvement in target-field correctness. Flare retains fewer earlier edits on CORD receipts, while photo-edit sequences provide little separation between models. Product codes are more often legible with Images 2.5, alongside larger product placement; the analyses do not establish a fidelity gain independent of size. Fine-print improvements remain unresolved below the OCR reliability limit. Refusals are rare and localisation is weak in both generations. At a fixed detection threshold, Community Forensics flags 68.6% of controlled Images 2.5 images averaged across cells, versus 35.9% of self-reported images posted online. These results motivate task-specific evaluation of advertised capabilities and defences, with explicit limits on what automatic checks can establish.

Chinese Translation

我们评估了 ChatGPT Images 2.5 宣传的改进是否转化为在具有预定答案的伪造任务上的更好表现。我们比较了其 Flare 和 Sunburst API 模型与同周重新运行的 GPT-Image-2，使用了收据字段编辑、重复编辑、产品摆放和细节渲染。在图像注册后，Flare 和 Sunburst 显示出对周围收据文本的 OCR 检测变化较少（分别为 31.7% 和 31.2%，而 GPT-Image-2 基线均为 44.2%），主要在 CORD 收据上，且在目标字段的正确性上没有检测到改进。Flare 在 CORD 收据上保留的早期编辑较少，而照片编辑序列在模型之间提供的区分度有限。与 Images 2.5 一起，产品代码更常可读，且产品摆放更大；分析未能证明独立于大小的保真度提升。细节改进在 OCR 可靠性限度以下仍未解决。拒绝情况较少，且两代模型的本地化能力较弱。在固定检测阈值下，Community Forensics 标记了 68.6% 的受控 Images 2.5 图像，平均跨单元，而在线发布的自报图像仅为 35.9%。这些结果促使对宣传能力和防御进行特定任务的评估，并明确自动检查可以建立的限制。

View on arXiv Download PDF AI Translation

cs.CV / 49 / 2609.13647

YOLO12-MambaScan: An Efficient Object Detector with High-Frequency Enhancement and State-Space Modeling

YOLO12-MambaScan：一种具有高频增强和状态空间建模的高效目标检测器

Wang, Hao

Abstract

The rapid development of unmanned aerial vehicle (UAV) technology has made aerial-image object detection increasingly important for natural-resource monitoring, traffic management, and disaster response. Detecting small objects in aerial images remains difficult because objects occupy very few pixels, high-frequency cues are easily lost, and global context is hard to model in cluttered scenes. Existing detectors often retain insufficient edge, corner, and texture information. We propose \ours, an aerial-image detector built on the YOLO12 architecture. The model combines a triple-path high-frequency enhancement convolution module (TriPathHFConv), receptive-field coordinate-attention convolution (RFCAConv), and a Mamba-based global-context module. On VisDrone, at an input resolution of 960*960, ours achieves 60.0% mAP@50 and 38.6%mAP@50:95, demonstrating a favorable accuracy--efficiency trade-off for small-object detection. The benchmark and dataset protocol follow the VisDrone challenge setup.

Chinese Translation

无人机（UAV）技术的快速发展使得航空图像目标检测在自然资源监测、交通管理和灾害响应中变得越来越重要。然而，在航空图像中检测小物体仍然困难，因为这些物体占据的像素非常少，高频线索容易丢失，并且在杂乱场景中很难建模全局上下文。现有的检测器往往保留的边缘、角落和纹理信息不足。我们提出了 extit{ours}，一种基于YOLO12架构的航空图像检测器。该模型结合了三路径高频增强卷积模块（TriPathHFConv）、感受野坐标注意力卷积（RFCAConv）和基于Mamba的全局上下文模块。在VisDrone数据集上，以960*960的输入分辨率，我们的模型达到了60.0%的mAP@50和38.6%的mAP@50:95，展示了在小物体检测中的良好准确性与效率的权衡。基准测试和数据集协议遵循VisDrone挑战的设置。

View on arXiv Download PDF AI Translation

cs.CV / 50 / 2609.13654

Multimodal Foundation Models Adaptation based on Domain-Aware Relaxed Orthogonal Subspace for Remote Sensing

基于领域感知的放松正交子空间的多模态基础模型适应用于遥感

Luo, Han, Yang, Ruoyu, Liu, Yinhe, Zhong, Yanfei

Abstract

Pretrained foundation models (FMs) have achieved remarkable success in computer vision, yet their high fine-tuning cost limits practical deployment. Parameter-efficient fine-tuning (PEFT) methods such as Low-Rank Adaptation (LoRA) improve efficiency by constraining updates to a predefined low-rank subspace. However, when applied to remote sensing tasks with substantial domain shifts, the fixed subspace is constructed without observing the downstream activation distribution and can therefore provide a poor coordinate system for adaptation, a phenomenon herein termed subspace mismatch. To address this issue, a unified framework is introduced, termed Domain-aware Relaxed Orthogonal Subspace adaptation (DROS), which reformulates low-rank adaptation as data-conditioned subspace learning and flexible subspace adaptation. Specifically, the weight decomposition is conditioned on second-order activation statistics estimated from the downstream training distribution, so that the initialization reflects the feature geometry actually induced by the remote-sensing data, followed by flexible geometric transformations enabled by a relaxed orthogonal parameterization. Furthermore, the framework is extended to multimodal settings (MM-DROS) by sharing transformation structures across modality-specific subspaces, facilitating efficient cross-modal interaction. Extensive experiments on multiple remote sensing benchmarks demonstrate that DROS achieves state-of-the-art performance, even surpassing full fine-tuning, without additional inference overhead.

Chinese Translation

预训练基础模型（FMs）在计算机视觉领域取得了显著成功，但其高昂的微调成本限制了实际应用。参数高效微调（PEFT）方法如低秩适应（LoRA）通过将更新限制在预定义的低秩子空间内来提高效率。然而，当应用于具有显著领域转移的遥感任务时，固定子空间是在未观察下游激活分布的情况下构建的，因此可能为适应提供一个较差的坐标系统，这一现象在此称为子空间不匹配。为了解决这一问题，提出了一个统一框架，称为领域感知放松正交子空间适应（DROS），该框架将低秩适应重新表述为数据条件下的子空间学习和灵活的子空间适应。具体而言，权重分解是基于从下游训练分布中估计的二阶激活统计量进行条件化的，以使初始化反映遥感数据实际引起的特征几何，随后通过放松正交参数化实现灵活的几何变换。此外，该框架通过在模态特定子空间之间共享变换结构扩展到多模态设置（MM-DROS），促进高效的跨模态交互。在多个遥感基准上的广泛实验表明，DROS实现了最先进的性能，甚至超越了完全微调，而没有额外的推理开销。

View on arXiv Download PDF AI Translation

cs.CV / 51 / 2609.13665

MARC: Morphology-Aware Regression of Consensus for Cell Segmentation in Subcellular Spatial Transcriptomics

MARC：基于形态学的共识回归用于亚细胞空间转录组中的细胞分割

Shu, Xinyu, Zhang, Andrew, Yang, Jean, Kim, Jinman

Abstract

Accurate cell segmentation remains a major bottleneck in subcellular spatial transcriptomics (SST), in which morphological images and spatially resolved RNA transcripts are used to partition tissues into individual cellular instances. As segmentation serves as the foundation for constructing cell-level representations, boundary errors can lead to incorrect transcript assignments and compromise downstream analyses. However, reliable ground-truth boundaries are unavailable because they must be inferred from incomplete morphological and transcript signals. Furthermore, manual annotation of a large number of cells is time-consuming. Agreement among complementary segmentation methods provides a practical surrogate for identifying well-supported and ambiguous regions, but explicit consensus construction requires executing multiple computationally intensive pipelines. In this study, we propose MARC (Morphology-Aware Regression of Consensus), a framework that predicts a multi-method consensus-support map for SST segmentation. MARC is trained with leave-one-method-out consensus pseudo-targets and a Foreground-Union Consensus Loss that focuses supervision on candidate and consensus foreground. We evaluated MARC on 4,642 held-out tiles from Xenium kidney tissue, achieving a mean Dice score of 0.90, a mean intersection-over-union of 0.82, and a mean cell-level Spearman correlation of 0.79 against explicitly computed cross-method consensus maps. We demonstrate that the predicted consensus maps localise weakly supported regions while preserving consensus-based rankings and identifying low-consensus cells for manual review. These results show that MARC closely approximates explicit cross-method consensus without multi-method inference and therefore has the potential to facilitate robust, consensus-aware evaluation of cell segmentation in large-scale SST studies.

Chinese Translation

准确的细胞分割仍然是亚细胞空间转录组（SST）中的一个主要瓶颈，该技术利用形态学图像和空间分辨的RNA转录本将组织划分为单个细胞实例。由于分割是构建细胞级表示的基础，边界错误可能导致错误的转录本分配，并影响后续分析。然而，可靠的真实边界并不可用，因为它们必须从不完整的形态学和转录信号中推断。此外，手动标注大量细胞耗时费力。互补分割方法之间的一致性为识别支持良好和模糊区域提供了一个实用的替代方案，但显式共识构建需要执行多个计算密集型管道。在本研究中，我们提出了MARC（基于形态学的共识回归），一个预测SST分割的多方法共识支持图的框架。MARC使用留一法共识伪目标和一个聚焦于候选和共识前景的前景联合共识损失进行训练。我们在来自Xenium肾组织的4642个保留图块上评估了MARC，达到了0.90的平均Dice得分、0.82的平均交并比和0.79的平均细胞级斯皮尔曼相关系数，与显式计算的跨方法共识图相比。我们展示了预测的共识图能够定位支持较弱的区域，同时保留基于共识的排名，并识别低共识细胞以供手动审核。这些结果表明，MARC在不进行多方法推断的情况下，能够接近显式的跨方法共识，因此有潜力促进大规模SST研究中细胞分割的稳健、共识感知评估。

View on arXiv Download PDF AI Translation

cs.CV / 52 / 2609.13691

MomentBA: Second-order Spatial Moments for Anisotropic Correspondence Uncertainty in Differentiable Bundle Adjustment

MomentBA：用于可微束调整中各向异性对应不确定性的二阶空间矩

Wang, Yuqing, Niu, Xiaoji, Wang, Yan, Tang, Hailiang, Kuang, Jian, Zhang, Tisheng

Abstract

Most existing visual odometry (VO) systems treat feature correspondences as deterministic measurements or assign uniform uncertainty, ignoring the inherent localization ambiguity of different observations. However, correspondence uncertainty is often anisotropic due to image structures such as edges, repetitive patterns, and motion blur, which can significantly affect geometric optimization. In this work, we propose MomentBA, a geometry-aware bundle adjustment framework that derives anisotropic correspondence uncertainty from second-order spatial moments of local similarity responses. Instead of introducing additional covariance prediction networks, the proposed method directly converts matching response distributions into interpretable covariance estimates and incorporates them into bundle adjustment as correspondence-specific information matrices for uncertainty-aware residual weighting. Furthermore, the proposed formulation is integrated into a differentiable optimization framework, establishing a direct connection between correspondence uncertainty and geometric estimation. Experiments on the EuRoC MAV and TartanAir v1 Hard datasets demonstrate that MomentBA improves monocular visual odometry accuracy compared with existing feature-based and learning-based approaches. The proposed anisotropic covariance model achieves lower rotational errors and more robust trajectory estimation than fixed and isotropic uncertainty models, validating the effectiveness of geometry-induced uncertainty modeling for challenging visual environments.

Chinese Translation

大多数现有的视觉里程计（VO）系统将特征对应视为确定性测量或分配均匀的不确定性，忽略了不同观测的固有定位模糊性。然而，由于图像结构（如边缘、重复模式和运动模糊），对应不确定性通常是各向异性的，这会显著影响几何优化。在本研究中，我们提出了MomentBA，一种几何感知的束调整框架，该框架从局部相似性响应的二阶空间矩中推导出各向异性对应不确定性。该方法不引入额外的协方差预测网络，而是直接将匹配响应分布转换为可解释的协方差估计，并将其作为对应特定信息矩阵纳入束调整，以实现不确定性感知的残差加权。此外，所提出的公式被集成到可微优化框架中，建立了对应不确定性与几何估计之间的直接联系。在EuRoC MAV和TartanAir v1 Hard数据集上的实验表明，与现有的基于特征和基于学习的方法相比，MomentBA提高了单目视觉里程计的精度。所提出的各向异性协方差模型在旋转误差和轨迹估计的鲁棒性方面优于固定和各向同性的不确定性模型，验证了几何诱导的不确定性建模在具有挑战性的视觉环境中的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 53 / 2609.13706

Rank-Consistent Set Reasoning for Co-Salient Object Detection

一致性排序集合推理用于共同显著性目标检测

Xiang, Yuan, Rossi, Matteo, Chen, Yingzhou

Abstract

Co-salient object detection (Co-SOD) requires a model to find foreground regions that are salient in individual images and supported by the image group. We present \emph{Rank-Consistent Set Reasoning} (RCSR), a supervised dense-prediction framework that models a group as an unordered set rather than as a sequence of images or a semantic label. The core idea is to rank how strongly each spatial region agrees with a small collection of learned group slots at every image scale, and to aggregate these ranks with a robust trimmed statistic. This suppresses accidental pairwise matches and prevents one atypical group member from dominating the shared representation. A set encoder builds group slots directly from multi-scale visual features, while a rank-consistency gate measures whether the ordering of candidate regions is stable across group members. The gated slots are decoded jointly with per-image features to produce co-saliency maps. The model contains no natural-language branch, no open-vocabulary detector, and no external segmentation model. We further introduce a group permutation objective and hard-distractor augmentation so that the model learns the properties of a set-level target rather than memorizing image order or isolated visual saliency. We formulate an evaluation protocol for CoCA, CoSal2015, and CoSOD3k, together with tests of group-size robustness, distractor rejection, order invariance, and cross-dataset transfer.

Chinese Translation

共同显著性目标检测（Co-SOD）要求模型在个别图像中找到显著的前景区域，并且这些区域在图像组中得到支持。我们提出了 extit{一致性排序集合推理}（Rank-Consistent Set Reasoning, RCSR），这是一种监督密集预测框架，它将图像组建模为无序集合，而不是图像序列或语义标签。其核心思想是在每个图像尺度上，对每个空间区域与一小组学习到的组槽之间的相符程度进行排序，并使用稳健的修剪统计量来聚合这些排序。这一方法抑制了偶然的成对匹配，防止了一个非典型组成员主导共享表示。集合编码器直接从多尺度视觉特征构建组槽，而排序一致性门则测量候选区域的排序在组成员之间是否稳定。经过门控的槽与每幅图像特征共同解码，以生成共同显著性图。该模型没有自然语言分支、开放词汇检测器和外部分割模型。我们进一步引入了组置换目标和困难干扰物增强，使得模型学习集合级目标的特性，而不是记忆图像顺序或孤立的视觉显著性。我们为CoCA、CoSal2015和CoSOD3k制定了评估协议，并进行了组大小鲁棒性、干扰物拒绝、顺序不变性和跨数据集迁移的测试。

View on arXiv Download PDF AI Translation

cs.CV / 54 / 2609.13733

FFVO: A Feedforward Pose Decoder for Long-Horizon Visual Odometry

FFVO：一种用于长时间视觉里程计的前馈姿态解码器

Shih, Meng-Li, Su, Shih-Yang, Zou, Yuliang, Xiang, Hao, Zhu, Haidong, Casser, Vincent, Curless, Brian, Kalenichenko, Dmitry, Tan, Mingxing, Anguelov, Dragomir

Abstract

Stable and reliable 4D spatial understanding is fundamental for autonomous driving systems. While feedforward reconstruction networks can estimate camera motion and 3D structure in one pass, pose estimation over long videos remains challenged by computational cost, long-context ambiguity, and temporal instability. To address these challenges, we propose Feedforward Visual Odometry (FFVO), a pose-specialized adaptation of joint reconstruction architectures for efficient and temporally stable camera-pose estimation. FFVO uses (i) a compact camera-token representation for computationally efficient temporal aggregation, (ii) a hierarchical local-to-global temporal decoder that mitigates geometric ambiguity by separating short-range motion aggregation from sequence-level integration, and (iii) intermediate trajectory supervision that promotes temporal stability. Extensive evaluation on the Waymo Open Dataset (WOD), KITTI, and a large-scale proprietary benchmark demonstrates that our method performs favorably against existing feedforward approaches, and greatly reduces jitter and drift. These results support FFVO as an effective feedforward camera-pose decoder in long-horizon visual odometry settings.

Chinese Translation

稳定可靠的四维空间理解是自动驾驶系统的基础。虽然前馈重建网络能够在一次传递中估计相机运动和三维结构，但在长视频中的姿态估计仍面临计算成本、长时间上下文模糊和时间不稳定性等挑战。为了解决这些问题，我们提出了前馈视觉里程计（FFVO），这是一种针对联合重建架构的姿态专用适配，旨在实现高效且时间稳定的相机姿态估计。FFVO使用（i）紧凑的相机标记表示，以实现计算上高效的时间聚合，（ii）一个层次化的局部到全局时间解码器，通过将短期运动聚合与序列级集成分离，从而减轻几何模糊，以及（iii）中间轨迹监督，以促进时间稳定性。在Waymo开放数据集（WOD）、KITTI和一个大规模专有基准上的广泛评估表明，我们的方法在性能上优于现有的前馈方法，并大大减少了抖动和漂移。这些结果支持FFVO作为长时间视觉里程计环境中有效的前馈相机姿态解码器。

View on arXiv Download PDF AI Translation

cs.CV / 55 / 2609.13742

Hyper-LLaVA: Hyperbolic Uncertainty-aware Modality-Balanced Routing for Multimodal Continual Instruction Tuning

超曲率LLaVA：面向多模态持续指令调优的超曲率不确定性感知模态平衡路由

Xu, Kunlun, Zhang, Yanqin, Qiang, Wenwen, Zhou, Jiahuan

Abstract

Multimodal Continual Instruction Tuning (MCIT) aims to exploit the incrementally accumulated knowledge to process multimodal inputs of diverse tasks, where parameter routing plays an important role. State-of-the-art methods rely on sample-to-task center similarity and cross-modal fusion with equal weight during routing. However, such solutions face two fundamental flaws: (1) Within each modality, the sample-to-task center distance is sub-optimal for routing since the abundant intra-task diversity information is underleveraged. (2) Different modalities exhibit varying reliability across tasks, where the modality with inter-task ambiguity can easily misguide the routing result. To address these problems, we propose Hyperbolic Uncertainty-aware Modality-Balanced Routing (Hyper-LLaVA) to improve parameter routing capacity based on cross-modality task feature uncertainty modeling. Specifically, to improve intra-modality task matching, Hyper-LLaVA accesses the sample-to-task distribution similarity in the Hyperbolic space. Besides, to alleviate the degradation brought by unreliable modalities, Hyper-LLaVA quantifies the task matching ambiguity within each modality to achieve adaptive balancing between task matching across modalities. Based on the complementary intra- and inter-modality task matching enhancement, our Hyper-LLaVA outperforms state-of-the-art approaches by large margins. Our source code is available at https://github.com/zhoujiahuan1991/ICML2026-Hyper-LLaVA

Chinese Translation

多模态持续指令调优（MCIT）旨在利用逐步积累的知识处理多样任务的多模态输入，其中参数路由发挥着重要作用。当前最先进的方法依赖于样本与任务中心的相似性以及在路由过程中跨模态融合的等权重。然而，这些解决方案面临两个根本性缺陷：（1）在每种模态内，样本与任务中心的距离对于路由来说是次优的，因为丰富的任务内部多样性信息未得到充分利用。（2）不同模态在任务间的可靠性各异，具有任务间模糊性的模态容易误导路由结果。为了解决这些问题，我们提出了超曲率不确定性感知模态平衡路由（Hyper-LLaVA），旨在基于跨模态任务特征不确定性建模来提高参数路由能力。具体而言，为了改善模态内的任务匹配，Hyper-LLaVA在超曲率空间中访问样本与任务分布的相似性。此外，为了减轻不可靠模态带来的降级，Hyper-LLaVA量化每种模态内的任务匹配模糊性，以实现跨模态任务匹配的自适应平衡。基于互补的模态内和模态间任务匹配增强，我们的Hyper-LLaVA在性能上大幅超越了最先进的方法。我们的源代码可在https://github.com/zhoujiahuan1991/ICML2026-Hyper-LLaVA获取。

View on arXiv Download PDF AI Translation

cs.CV / 56 / 2609.13756

Development of Low-Cost Real-Time Driver Drowsiness Detection System using Eye Centre Tracking and Dynamic Thresholding

基于眼睛中心追踪和动态阈值的低成本实时驾驶员疲劳检测系统的开发

Khan, Fuzail, Sharma, Sandeep, Arulalan, M. R.

Abstract

One in every five vehicle accidents on the road today is caused simply due to driver fatigue. Fatigue or otherwise drowsiness, significantly reduces the concentration and vigilance of the driver thereby increasing the risk of inherent human error leading to injuries and fatalities. Hence, our primary motive being - to reduce road accidents using a non-intrusive image processing based alert system. In this regard, we have built a system that detects driver drowsiness by real time tracking and monitoring the pattern of the driver's eyes. The stand alone system consists of 3 interconnected components - a processor, a camera and an alarm. After initial facial detection, the eyes are located, extracted and continuously monitored to check whether they are open or closed on the basis of a pixel-by-pixel method. When the eyes are seen to be closed for a certain amount of time, drowsiness is said to be detected and an alarm is issued accordingly to alert the driver and hence, prevent a casualty.

Chinese Translation

目前每五起交通事故中就有一起是由于驾驶员疲劳引起的。疲劳或嗜睡显著降低了驾驶员的注意力和警觉性，从而增加了固有的人为错误风险，导致伤害和死亡。因此，我们的主要目标是通过一种非侵入式的基于图像处理的警报系统来减少道路事故。在这方面，我们构建了一个系统，通过实时跟踪和监测驾驶员眼睛的模式来检测驾驶员的疲劳。该独立系统由三个相互连接的组件组成——处理器、摄像头和警报。在初步的人脸检测后，眼睛被定位、提取并持续监测，以逐像素的方法检查其是睁开还是闭合。当眼睛在一定时间内被观察到闭合时，便认为检测到了疲劳，并相应发出警报以提醒驾驶员，从而防止事故的发生。

View on arXiv Download PDF AI Translation

cs.CV / 57 / 2609.13778

GEAR: From Dynamic Encoding to Dynamic Activation in Social Trajectory Prediction

GEAR：从动态编码到动态激活的人类轨迹预测

Chen, Jiaheng, Li, Jiaxing, Wang, Leixia, Ju, Jianan, Zhang, Tinghe

Abstract

Human trajectory prediction requires modeling both individual motion patterns and social interactions among agents. Existing methods have made substantial progress by using attention mechanisms, graph structures, and temporal encoders to capture dynamic social context. However, most of them primarily focus on how social information is encoded, while paying less explicit attention to how the encoded social context should take effect during future trajectory generation. In this paper, we argue that dynamic social encoding does not necessarily imply dynamic social activation. The same interaction context may require different activation strengths across future horizons and scene densities: social cues should be strengthened when interaction evidence is strong, but suppressed when they are weak or noisy. To address this issue, we propose GEAR, a generation-aware bias activation model for human trajectory prediction. Built upon a bias-decomposed trajectory generation formulation, GEAR dynamically activates the individual-motion and social-resonance bias terms at each future step before final trajectory composition. This allows the model to explicitly control when and how strongly individual and social bias components participate in generation. Experiments on ETH-UCY, SDD, and NBA show that GEAR consistently improves the resonance-based baseline and achieves competitive state-of-the-art performance. Further analyses of activation patterns and density-grouped errors validate the importance of calibrating encoded social context during trajectory generation. Our code is available at https://github.com/11isnotavailable/GEAR.git.

Chinese Translation

人类轨迹预测需要建模个体运动模式和代理之间的社会交互。现有方法通过使用注意力机制、图结构和时间编码器来捕捉动态社会背景，取得了显著进展。然而，它们大多数主要关注社会信息的编码方式，而对编码的社会背景在未来轨迹生成中的作用关注较少。在本文中，我们认为动态社会编码并不一定意味着动态社会激活。同样的交互背景在未来的时间范围和场景密度中可能需要不同的激活强度：当交互证据强时，社会线索应被增强，而当它们弱或嘈杂时应被抑制。为了解决这个问题，我们提出了GEAR，一种面向生成的偏置激活模型，用于人类轨迹预测。GEAR基于偏置分解的轨迹生成公式，在最终轨迹组合之前，在每个未来步骤动态激活个体运动和社会共鸣偏置项。这使得模型能够明确控制个体和社会偏置成分在生成中的参与时机和强度。对ETH-UCY、SDD和NBA的数据集进行的实验表明，GEAR始终改善了基于共鸣的基线，并取得了具有竞争力的最先进性能。对激活模式和密度分组误差的进一步分析验证了在轨迹生成过程中校准编码社会背景的重要性。我们的代码可在 https://github.com/11isnotavailable/GEAR.git 获取。

View on arXiv Download PDF AI Translation

cs.CV / 58 / 2609.13791

Restore What Matters: Lessons from Joint Restoration and Recognition

恢复重要内容：联合恢复与识别的经验教训

Guo, Lanqing, Wang, Xijun, Kim, Minchul, Yuan, Yu, Robbins, Wes, Zhang, Xingguang, Chimitt, Nicholas, Chan, Stanley H., Wang, Zhangyang, Liu, Xiaoming

Abstract

Recognition pipelines typically adopt a restore-then-recognize workflow, yet decades of experience show that generating visually pleasing images seldom translates to improved recognition. We propose a Joint Restoration-for-Recognition (JR$^2$) paradigm: restore only what downstream tasks truly require, with task signals dictating where, how much, and whether restoration is necessary. JR$^2$ rests on three pillars: (i) Physics, employing optics-accurate turbulence simulation, extensible to blur and noise, to ground restoration in real image formation; (ii) Neuroscience, drawing on selective attention and neuroplasticity to direct model capacity toward identity-critical regions and frames while bypassing already-clean inputs; and (iii) Vision & Learning, coupling recognition loss end-to-end through restoration and alignment so that low-level edits maximize high-level identity stability. Evaluations on IARPA-BRIAR show consistent improvements (e.g., [email protected]% FAR +0.6; FNIR@1% FPIR -2.5), while a quality gate skips ~70% of clean frames, reducing cost. Ablations confirm physics priors enhance realism, joint training prevents catastrophic forgetting, and selective restoration suffices in many cases. We conclude that better-looking images are neither necessary nor sufficient; restoration modules must be task-driven, selective, and physically aware. Code, pretrained models, and recipes are provided for integration.

Chinese Translation

识别流程通常采用恢复后识别的工作流程，但数十年的经验表明，生成视觉上令人愉悦的图像很少能转化为更好的识别效果。我们提出了一种联合恢复-识别（Joint Restoration-for-Recognition, JR$^2$）范式：仅恢复下游任务真正需要的内容，任务信号决定了恢复的地点、程度以及是否需要恢复。JR$^2$ 基于三个支柱：(i) 物理学，采用光学准确的湍流模拟，扩展到模糊和噪声，以真实图像形成为基础进行恢复；(ii) 神经科学，借鉴选择性注意和神经可塑性，将模型能力引导至身份关键区域和帧，同时绕过已经干净的输入；(iii) 视觉与学习，通过恢复和对齐将识别损失端到端耦合，使低级编辑最大化高级身份稳定性。在 IARPA-BRIAR 上的评估显示出一致的改进（例如，[email protected]% FAR +0.6；FNIR@1% FPIR -2.5），同时质量门控跳过约 70% 的干净帧，从而降低成本。消融实验确认物理先验增强了真实感，联合训练防止了灾难性遗忘，而选择性恢复在许多情况下是足够的。我们得出结论，更好看的图像既不是必要的也不是充分的；恢复模块必须以任务为驱动，具有选择性，并且具备物理意识。我们提供了代码、预训练模型和集成食谱。

View on arXiv Download PDF AI Translation

cs.CV / 59 / 2609.13804

StepPrune: Adaptive Sequential Visual Token Selection across Multimodal Large Language Models

StepPrune：跨多模态大型语言模型的自适应序列视觉标记选择

Zhang, Hansen, He, Landi, Yao, Mingde, Xu, Lijian

Abstract

Visual prefixes account for a major portion of the per-layer computation in multimodal large language models (MLLMs), making visual-token pruning a direct approach to accelerating inference. Existing top-K methods typically evaluate tokens independently and apply a uniform budget to all inputs, overlooking both selection-dependent interactions and variations in visual complexity across samples. In contrast, we propose StepPrune, which formulates visual-token pruning as an adaptive sequential decision process. Conditioned on previously selected tokens and textual context, StepPrune progressively constructs the retained subset and automatically determines its size through a learned STOP action. During training, a variance-preserving noise gate provides a differentiable surrogate for the discrete selection process, whereas during inference, unselected tokens are physically removed before language-model prefill. A grouped selection mechanism further extends StepPrune to high-resolution inputs. Experiments across LLaVA-1.5, LLaVA-NeXT, Qwen2.5-VL, and InternVL3 show that StepPrune achieves the best average normalized performance retention across all evaluated pruning rates on LLaVA-1.5, Qwen2.5-VL, and InternVL3, while remaining competitive on the substantially longer AnyRes prefixes of LLaVA-NeXT. On LLaVA-1.5, StepPrune retains 94.6% of the full-prefix normalized performance while pruning 88.9% of the visual tokens. At a mean retained count of 64, StepPrune reduces prefill latency from 59.95 ms to 40.05 ms, corresponding to a 1.50x prefill speed-up.

Chinese Translation

视觉前缀在多模态大型语言模型（MLLMs）的每层计算中占据了重要部分，因此视觉标记剪枝成为加速推理的直接方法。现有的 top-K 方法通常独立评估标记，并对所有输入应用统一预算，忽视了选择依赖的交互和样本间视觉复杂性的变化。相对而言，我们提出了 StepPrune，将视觉标记剪枝公式化为一种自适应的序列决策过程。StepPrune 在先前选择的标记和文本上下文的条件下，逐步构建保留的子集，并通过学习的 STOP 动作自动确定其大小。在训练过程中，保持方差的噪声门为离散选择过程提供了可微分的替代方案，而在推理过程中，未选择的标记在语言模型预填充之前被物理移除。分组选择机制进一步扩展了 StepPrune 以适应高分辨率输入。在 LLaVA-1.5、LLaVA-NeXT、Qwen2.5-VL 和 InternVL3 的实验中，StepPrune 在所有评估的剪枝率上实现了最佳的平均归一化性能保留，同时在 LLaVA-NeXT 的显著更长的 AnyRes 前缀上保持竞争力。在 LLaVA-1.5 上，StepPrune 保留了 94.6% 的全前缀归一化性能，同时剪枝了 88.9% 的视觉标记。在平均保留数量为 64 的情况下，StepPrune 将预填充延迟从 59.95 毫秒减少到 40.05 毫秒，相当于 1.50 倍的预填充加速。

View on arXiv Download PDF AI Translation

cs.CV / 60 / 2609.13814

Realtime-Venus: A full-duplex interaction system with asynchronous delegation

Realtime-Venus：一种具有异步委托的全双工交互系统

Zhao, Ruixiang, Wang, Hualei, Sun, Renhe, Zhou, Enzhi, Wu, Jincenzi, Song, Xujie, Shi, Kexin, Liu, Zihang, Zhu, Pengcheng, Zhou, Jiayi, Zhang, Baoyue, Zhang, Changhao, Wang, Zitong, Wang, Jinhong, Niu, Tong, Liu, Jingjing, Lin, Junan, He, Haolin, Chu, Hengshuo, Chen, Yuhui, Liu, Jian, Huang, Yuge, Xing, Junliang, Wang, Yuntao, Wang, Weiqiang, Yu, Chun, Shi, Yuanchun

Abstract

Natural interaction in digital and physical environments requires continuous perception and timely responses. Spoken dialogue relies on acoustic and linguistic cues, while video interaction also requires grounding the conversation in evolving visual context. We present Realtime-Venus, a proactive full-duplex interaction system with two separately trained 9B models: Realtime-Venus-Omni for audio-visual interaction and Realtime-Venus-Audio for spoken interaction. Each model serves as a complete conversational frontend, integrating continuous perception, conversational control, and native speech generation through a shared causal timeline for user inputs, model outputs, and delegation events. A dual-loop runtime coordinates live interaction with background reasoning and tool execution. Foreground interaction continues while Realtime-Venus-Harness executes tasks asynchronously and returns results for integration into the ongoing dialogue. Both models follow a common post-training recipe combining offline understanding, proactive full-duplex trajectories, and delegation workflows. Among the evaluated online models, Realtime-Venus-Omni achieves the highest scores on six of eight video benchmarks, including StreamingBench (70.2%), OVO-Bench (64.7%), and Daily-Omni (81.3%). Across eight audio understanding and spoken question answering benchmarks, Realtime-Venus-Audio leads the compared models on MMAU (78.0%), MMAU-Pro (63.2%), Llama Questions (83.8%), and Speech CMMLU (67.8%), while matching the best VoiceBench AlpacaEval score of 4.81. On Full-Duplex-Bench v1.5, Realtime-Venus-Audio responds to 75% of user interruptions and achieves continuation rates of 97%, 88%, and 86% under backchannels, other-directed speech, and background speech, respectively, exceeding Gemini 3.1 Live and GPT-4o on all three continuation metrics.

Chinese Translation

数字和物理环境中的自然交互需要持续的感知和及时的响应。口语对话依赖于声学和语言线索，而视频交互还需要将对话与不断变化的视觉上下文相结合。我们提出了Realtime-Venus，一个主动的全双工交互系统，使用两个分别训练的9B模型：Realtime-Venus-Omni用于音视频交互，Realtime-Venus-Audio用于口语交互。每个模型作为一个完整的对话前端，集成了持续的感知、对话控制和通过共享因果时间线生成的本地语音，时间线涵盖用户输入、模型输出和委托事件。双循环运行时协调实时交互与背景推理和工具执行。在Realtime-Venus-Harness异步执行任务并返回结果以整合到正在进行的对话中时，前景交互仍在继续。两个模型遵循一个共同的后训练方案，结合离线理解、主动全双工轨迹和委托工作流。在评估的在线模型中，Realtime-Venus-Omni在八个视频基准中的六个上获得了最高分，包括StreamingBench（70.2%）、OVO-Bench（64.7%）和Daily-Omni（81.3%）。在八个音频理解和口语问答基准中，Realtime-Venus-Audio在MMAU（78.0%）、MMAU-Pro（63.2%）、Llama Questions（83.8%）和Speech CMMLU（67.8%）上领先于比较模型，同时在VoiceBench AlpacaEval中匹配最佳得分4.81。在Full-Duplex-Bench v1.5中，Realtime-Venus-Audio对75%的用户中断做出响应，并在背景通道、他人导向的言语和背景言语下分别实现了97%、88%和86%的继续率，超越了Gemini 3.1 Live和GPT-4o在所有三个继续指标上的表现。

View on arXiv Download PDF AI Translation

cs.CV / 61 / 2609.13815

Beyond OCR Accuracy: Text-Centric VQA Under Image Degradation with Modular and End-to-End

超越OCR准确性：在图像退化下的文本中心视觉问答（VQA）研究，采用模块化和端到端方法

Vatsi, Ritali, Jagadeesh, Rachapudi, Baghel, Shruti Singh, Sharma, Himani, Shukla, Amit, Goyal, Pawan

Abstract

Text-centric Visual Question Answering (VQA) requires reading and reasoning over text embedded in images, a task made substantially harder when images suffer from real-world degradation such as motion blur, low resolution, or compression artifacts. While modular OCR-based pipelines and end-to-end vision-language models are both widely used for this task, their comparative robustness under degraded conditions remains underexplored. We present an empirical study comparing two modular pipelines with SA-DBNet, a custom detector architecture combining ResNet-18 with self-attention spatial modeling and deformable convolutions against an end-to-end vision-language baseline, evaluated on 4013 degraded images with 7000 question-answer pairs. Fine-tuned modular pipelines achieve up to 57.50% exact-match accuracy versus 38.00% for the end-to-end baseline, with domain-specific fine-tuning yielding a gain of up to 29.50 percentage points. Critically, we find that conventional OCR error metrics like Character Error Rate and Word Error Rate are unreliable predictors of downstream VQA performance, as semantic reasoning can compensate for recognition failures when contextual cues are present. These findings highlight the importance of task-aware evaluation for text-centric VQA systems under realistic visual conditions. Codes are available here

Chinese Translation

文本中心的视觉问答（VQA）需要对嵌入图像中的文本进行阅读和推理，而当图像受到现实世界退化（如运动模糊、低分辨率或压缩伪影）影响时，这一任务变得更加困难。尽管模块化的基于OCR的管道和端到端的视觉语言模型在这一任务中被广泛使用，但它们在退化条件下的相对鲁棒性仍未得到充分探讨。我们进行了一项实证研究，比较了两个模块化管道与SA-DBNet（一个结合了ResNet-18、自注意力空间建模和可变形卷积的定制检测器架构），并与一个端到端的视觉语言基线进行评估，数据集包含4013张退化图像和7000个问答对。经过微调的模块化管道实现了最高57.50%的准确匹配率，而端到端基线为38.00%，领域特定的微调带来了最高29.50个百分点的提升。重要的是，我们发现传统的OCR错误指标，如字符错误率（Character Error Rate）和单词错误率（Word Error Rate），并不是下游VQA性能的可靠预测指标，因为当存在上下文线索时，语义推理可以弥补识别失败。这些发现强调了在现实视觉条件下对文本中心VQA系统进行任务感知评估的重要性。代码可在此处获取。

View on arXiv Download PDF AI Translation

cs.CV / 62 / 2609.13823

Semantic Privacy Protection with Utility Preservation for 3D Point Clouds

具有效用保护的三维点云语义隐私保护

zhang, Jinchang, Lin, Jiakai, Crandall, David, Lu, Guoyu

Abstract

Point cloud data face serious semantic privacy risks during acquisition, transmission, and cross-institutional sharing. Existing methods mostly rely on geometric perturbation or destructive encryption, which can reduce the recognizability of the original class but often impair downstream usability. This paper proposes a class-transfer-based semantic encryption framework for point clouds, aiming to conceal original class information while preserving task utility and supporting authorized recovery. Specifically, we construct a unified latent space with a shared-backbone Normalizing Flow, and combine LoRA and FiLM to achieve parameter-efficient class-conditional adaptation. We further introduce diffusion-guided flow alignment to regularize the latent distribution, construct an energy-based category transition graph, and obtain an optimal class-transfer table through global matching. Then, a latent-space Neural ODE continuously evolves source-class latents into target-class latents, which are decoded into target-class point clouds through the inverse flow. We adopt attacker-oriented metrics, including New-Class Recognition Rate (NCRR), Original-Class Leakage Rate (OCLR), and Original Label Recovery Rate (OLRR), to evaluate privacy and utility. Experiments on classification and segmentation benchmarks show that the proposed method achieves controllable semantic transformation, effectively reduces original-class semantic leakage, preserves downstream learnability in the protected domain, and supports reliable authorized reconstruction.

Chinese Translation

点云数据在获取、传输和跨机构共享过程中面临严重的语义隐私风险。现有方法主要依赖几何扰动或破坏性加密，这虽然可以降低原始类别的可识别性，但往往会损害下游可用性。本文提出了一种基于类别转移的点云语义加密框架，旨在隐蔽原始类别信息，同时保持任务效用并支持授权恢复。具体而言，我们构建了一个统一的潜在空间，采用共享骨干的正则化流（Normalizing Flow），并结合LoRA和FiLM实现参数高效的类别条件适应。我们进一步引入扩散引导的流对齐，以规范潜在分布，构建基于能量的类别转移图，并通过全局匹配获得最佳类别转移表。然后，潜在空间的神经常微分方程（Neural ODE）将源类别潜在向量连续演变为目标类别潜在向量，这些向量通过逆流解码为目标类别的点云。我们采用以攻击者为导向的指标，包括新类别识别率（New-Class Recognition Rate, NCRR）、原始类别泄露率（Original-Class Leakage Rate, OCLR）和原始标签恢复率（Original Label Recovery Rate, OLRR），来评估隐私和效用。在分类和分割基准上的实验表明，所提方法实现了可控的语义转化，有效减少了原始类别的语义泄露，保持了受保护领域的下游可学习性，并支持可靠的授权重构。

View on arXiv Download PDF AI Translation

cs.CV / 63 / 2609.13830

DiVA: Enabling Interactive Digital Life Simulation via Video Models

DiVA：通过视频模型实现互动数字生活模拟

Chen, Cheng, Ouyang, Hao, Wang, Qiuyu, Cheng, Ka Leong, Wang, Wen, Meng, Yihao, Wang, Hanlin, Li, Yixuan, Wei, Jiacheng, Tan, Zhenshan, Zeng, Yanhong, Shen, Yujun, Lin, Guosheng, Liu, Fayao

Abstract

We present DiVA, a deeply interactive digital life simulator pioneering a new paradigm for long-term, open-ended interactive experiences within digital character worlds. DiVA's architecture pairs a Multimodal Large Language Model (MLLM) as a router with a meticulously designed stacked video pipeline for seamless, multi-turn interactions with action and audio response. To maintain continuity and avoid degradation, we model generation as a three-part coupled system: waiting video, action video, and the transitions between them. These transitions are critically handled by our Anchored Video Continuation (AVC) module, which returns the character to stable states to prevent degradation. By encoding information from the preceding action video segment, AVC ensures smooth transitions, significantly reducing camera jitter and inconsistencies common in current video transition methods. This design also enables complex pose changes (e.g., sitting to standing) typically difficult for audio-driven models. These system designs together ensure high-fidelity identity, coherence, and dynamics for extended experiences. To validate our pipeline design, we comprehensively compare our system against alternatives by replacing our core generation module with mainstream long-video, continuation, and interpolation methods. We further analyze the necessity of the three-stage design, anchor-state selection, transition naturalness, spatial grounding, and the quality-latency trade-off, and we expand the comparison to additional long-form audio-driven avatar models. Results confirm DiVA is markedly superior in maintaining long-term visual quality and realism, validating its effectiveness as a sustainable, interactive simulation.

Chinese Translation

我们提出了DiVA，这是一种深度互动的数字生活模拟器，开创了一种在数字角色世界中进行长期、开放式互动体验的新范式。DiVA的架构将多模态大型语言模型（Multimodal Large Language Model, MLLM）作为路由器，与精心设计的堆叠视频管道相结合，实现与动作和音频响应的无缝多轮互动。为了保持连续性并避免退化，我们将生成建模为一个由三个部分耦合的系统：等待视频、动作视频及其之间的过渡。这些过渡由我们的锚定视频延续（Anchored Video Continuation, AVC）模块关键处理，该模块将角色返回到稳定状态以防止退化。通过编码前一个动作视频片段的信息，AVC确保了平滑的过渡，显著减少了当前视频过渡方法中常见的相机抖动和不一致性。该设计还使得复杂的姿态变化（例如，从坐姿到站姿）成为可能，这通常对音频驱动模型来说较为困难。这些系统设计共同确保了在扩展体验中高保真的身份、一致性和动态性。为了验证我们的管道设计，我们通过将核心生成模块替换为主流的长视频、延续和插值方法，全面比较了我们的系统与其他方法。我们进一步分析了三阶段设计的必要性、锚定状态选择、过渡自然性、空间定位以及质量-延迟权衡，并将比较扩展到其他长格式音频驱动的虚拟角色模型。结果确认，DiVA在维持长期视觉质量和真实感方面显著优于其他方法，验证了其作为可持续互动模拟的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 64 / 2609.13838

TotalSynth: Robust Whole-Body Synthetic CT from MRI and CBCT

TotalSynth：基于MRI和锥形束CT的稳健全身合成CT

Boussot, Valentin, Hemon, Cedric, Barateau, Anais, Lafond, Caroline, Nunes, Jean-Claude, Dillenseger, Jean-Louis

Abstract

Purpose: To develop and evaluate TotalSynth, a reusable pretrained framework for whole-body synthetic CT (sCT) generation from MRI and cone-beam CT (CBCT) images. Materials and Methods: In this retrospective technical study, the dataset was assembled between 2020 and 2026 from SynthRAD challenge data, four prostate cohorts, and BIC-MAC. After registration quality control, 1450 of 1800 public challenge pairs were retained; 350 were excluded for insufficient registration quality or major source/CT mismatch. The corpus also included 84 additional prostate MRI/CT and 60 external BIC-MAC MRI/CT cases. Three 5-fold model families were evaluated with image-domain, anatomy-aware, registration-based, and uncertainty metrics. Age and sex were not consistently available across public datasets. Results: The released MRI-to-CT model achieved an overall MAE of 67.49 HU, SSIM of 0.920, and PSNR of 29.28 dB. The CBCT-to-CT model achieved an overall MAE of 53.55 HU, SSIM of 0.939, and PSNR of 32.09 dB. The unified model maintained similar performance on MRI inputs (MAE, 67.68 HU) and CBCT inputs (MAE, 54.22 HU). On external BIC-MAC data, MRI-to-CT MAE was 100.91 HU without fine-tuning and 62.21 HU after fine-tuning. Conclusion: TotalSynth provides reusable MRI- and CBCT-based CT synthesis models with broad anatomical coverage, while external evaluation highlights the need for local validation and optional fine-tuning under domain shift.

Chinese Translation

目的：开发和评估TotalSynth，一个可重复使用的预训练框架，用于从MRI和锥形束CT（CBCT）图像生成全身合成CT（sCT）。材料与方法：在这项回顾性技术研究中，数据集于2020年至2026年间从SynthRAD挑战数据、四个前列腺队列和BIC-MAC中组装而成。经过配准质量控制，保留了1800个公共挑战对中的1450个；由于配准质量不足或主要源/CT不匹配，排除了350个。数据集还包括84个额外的前列腺MRI/CT和60个外部BIC-MAC MRI/CT案例。评估了三个5折模型系列，采用图像域、解剖学感知、基于配准和不确定性指标。公共数据集中年龄和性别信息并不一致。结果：发布的MRI到CT模型实现了67.49 HU的总体平均绝对误差（MAE）、0.920的结构相似性指数（SSIM）和29.28 dB的峰值信噪比（PSNR）。CBCT到CT模型实现了53.55 HU的总体MAE、0.939的SSIM和32.09 dB的PSNR。统一模型在MRI输入（MAE，67.68 HU）和CBCT输入（MAE，54.22 HU）上保持了相似的性能。在外部BIC-MAC数据上，MRI到CT的MAE在未进行微调时为100.91 HU，微调后为62.21 HU。结论：TotalSynth提供了基于MRI和CBCT的CT合成模型，具有广泛的解剖覆盖，同时外部评估突显了在领域转移下进行本地验证和可选微调的必要性。

View on arXiv Download PDF AI Translation

cs.CV / 65 / 2609.13854

Event-Level Emotion Recognition in the Wild Using Deep Facial Expression Analysis

基于深度面部表情分析的野外事件级情感识别

Semerikov, Aleksandr, Shamoi, Pakizar

Abstract

Facial emotion recognition (FER) in real-world environments remains challenging due to unconstrained imaging conditions, including multiple faces, occlusions, pose variations, and complex lighting. Most existing studies focus on individual facial emotion classification and do not address the analysis of collective emotional states at the event level. This paper proposes an end-to-end pipeline for event-level emotion recognition from photographs. The approach detects faces in each image, classifies facial expressions using a deep convolutional neural network, and aggregates face-level emotion probabilities to estimate the overall emotional distribution of a public event. A comparative evaluation of several CNN architectures on the FER- 2013 and RAF-DB datasets demonstrates that transfer learning with EfficientNet-B2 trained on RAF-DB is more suitable for real-world RGB data. The proposed method is evaluated on a real-world event dataset containing 1658 images. Experimental results show stable emotion distributions across event subsets, confirming the effectiveness of event-level aggregation for emotion analysis in the wild.

Chinese Translation

在真实环境中进行面部情感识别（FER）仍然面临挑战，主要由于不受限制的成像条件，包括多个面孔、遮挡、姿态变化和复杂的光照。现有研究大多集中于单个面部情感分类，而未能解决事件级别的集体情感状态分析。本文提出了一种从照片中进行事件级情感识别的端到端流程。该方法在每张图像中检测面孔，使用深度卷积神经网络对面部表情进行分类，并聚合面部情感概率以估计公共事件的整体情感分布。在FER-2013和RAF-DB数据集上对几种CNN架构进行的比较评估表明，使用在RAF-DB上训练的EfficientNet-B2进行迁移学习更适合真实世界的RGB数据。所提方法在一个包含1658张图像的真实事件数据集上进行了评估。实验结果显示事件子集之间的情感分布稳定，确认了事件级聚合在野外情感分析中的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 66 / 2609.13868

Physically Typed and Geometry-Aware Representations for Earth Foundation Models

物理类型化和几何感知的地球基础模型表示

Ranjan, Rajiv

Abstract

Earth-observation (EO) foundation models have become exceptionally effective at learning se mantic, high-dimensional geospatial embeddings, while modern weather and climate models have demonstrated that Earth-specific geometry, spherical operators, meshes, and hybrid physical solvers can materially improve prediction. Yet these two advances are not equivalent. A conventional latent embedding has no inherent physical transformation law, whereas scalar fields, tangent polar-vector fields, axial/pseudovector quantities, covectors, and higher-order tensors transform differently under rotations, reflections, and changes of local coordinate frame. This proposal asks whether a general purpose Earth foundation model should preserve those distinctions explicitly, or whether standard embeddings plus augmentation already learn everything that matters. The central contribution is therefore not a more complicated architecture by assumption, but a staged falsification program. A compute-conscious ERA5 dry run first compares conventional, augmentation-matched, typed equivariant, and Hodge/Helmholtz variants under spatial, temporal, orientation, and low-data shifts. Only if explicit geometric typing yields reproducible improvements does the program advance toward a multimodal Earth foundation model in which semantic embeddings coexist with physically typed fields. The proposed gap is narrower and more defensible than claiming that current models ignore geometry entirely: several systems already respect spherical domain geometry, and emerging work explicitly learns scalar/vector fields on spheres. The unresolved question is whether foundation-scale, multimodal, parity-aware field typing produces practical gains beyond those existing approaches.

Chinese Translation

地球观测（EO）基础模型在学习语义、高维地理空间嵌入方面变得异常有效，而现代天气和气候模型则表明，特定于地球的几何形状、球面算子、网格和混合物理求解器可以显著改善预测。然而，这两种进展并不等同。传统的潜在嵌入没有固有的物理变换法则，而标量场、切向极向量场、轴向/伪向量量、共向量和高阶张量在旋转、反射和局部坐标系变化下的变换方式各不相同。本提案探讨一个通用的地球基础模型是否应该明确保留这些区别，或者标准嵌入加上增强是否已经学习了所有重要内容。因此，核心贡献并不是假设更复杂的架构，而是一个分阶段的反驳程序。一个计算意识的ERA5干运行首先比较传统的、增强匹配的、类型等变的以及Hodge/Helmholtz变体在空间、时间、方向和低数据变化下的表现。只有当显式的几何类型化带来可重复的改进时，该程序才会向一个多模态地球基础模型推进，其中语义嵌入与物理类型化场共存。所提出的差距比声称当前模型完全忽视几何更窄且更具辩护性：几个系统已经尊重球面域几何，而新兴工作则明确学习球面上的标量/向量场。未解决的问题是，基础规模的多模态、对称感知场类型化是否能在现有方法之外带来实际收益。

View on arXiv Download PDF AI Translation

cs.CV / 67 / 2609.13903

SkyAnchor: Updating Metric-scale Aerial 3D Gaussian Scenes from Unposed Ground-View Sequences

SkyAnchor：从未定姿态的地面视角序列更新度量尺度的空中三维高斯场景

Li, Zhuoxiao, Liu, Xinyi, Wu, Taoyu, Ren, Yinrui, Hua, Tongyan, Jing, Ou, Zhang, Shuai, Wu, Dongli, Qin, Rongjun, Kan, Ge Lin, Zhao, Wufan

Abstract

We study how to update a pre-built aerial scene with a newly captured, unposed ground-view sequence. The aerial scene already contains a reliable metric Structure-from-Motion (SfM) reconstruction and a pre-trained 3D Gaussian Splatting (3DGS) model, whereas the ground-view sequence is collected later to add street-level appearance but has unknown camera poses and global scale. Registering this sequence to the aerial SfM reconstruction is challenging because single-image cross-view localization is brittle and long trajectories are prone to drift. To address these challenges, we present SkyAnchor, which treats the existing aerial scene as a fixed scaffold for ground-view registration and scene update instead of jointly reconstructing aerial and ground imagery from scratch. It first localizes short groups of consecutive ground frames against geometrically verified aerial support, producing sparse anchor poses. It then recovers the full ground trajectory with anchor-constrained submaps, fixing the front and rear anchor poses during incremental registration and bundle adjustment. Finally, it inserts filtered ground Gaussians while preserving the aerial view, followed by lightweight joint refinement. Experiments on seven real aerial--ground scenes show accurate metric ground trajectories and updated 3D Gaussian scenes with strong aerial- and ground-view rendering quality.

Chinese Translation

我们研究如何使用新捕获的未定姿态地面视角序列更新预先构建的空中场景。该空中场景已经包含可靠的度量结构自运动（Structure-from-Motion, SfM）重建和预训练的三维高斯点云（3D Gaussian Splatting, 3DGS）模型，而地面视角序列则是在后期收集的，旨在添加街道级别的外观，但其相机姿态和全局尺度未知。将该序列与空中SfM重建进行配准具有挑战性，因为单幅图像的交叉视图定位不稳定，且长轨迹容易漂移。为了解决这些挑战，我们提出了SkyAnchor，它将现有的空中场景视为地面视角配准和场景更新的固定支架，而不是从头开始共同重建空中和地面图像。它首先将短组连续的地面帧相对于几何验证的空中支持进行定位，生成稀疏的锚点姿态。然后，它通过锚点约束的子地图恢复完整的地面轨迹，在增量配准和束调整过程中固定前后锚点姿态。最后，它在保留空中视图的同时插入过滤后的地面高斯点，随后进行轻量级的联合优化。在七个真实的空中-地面场景上的实验表明，获得了准确的度量地面轨迹和更新的三维高斯场景，具有强大的空中和地面视图渲染质量。

View on arXiv Download PDF AI Translation

cs.CV / 68 / 2609.13941

BLInD: Learning Driver Intent as a Distribution over Future Ego Trajectories

BLInD：将驾驶员意图学习为未来自我轨迹的分布

Pegado, Flavian, Hire, Ronit, Rajesh, Shreyas, Phade, Soham

Abstract

We present BLInD (Blind Learned Intent Distribution), a compact network that maps recent vehicle-state history (e.g. speed, curvature, indicator, and vehicle type) to a top-k distribution of future ego trajectories, with no camera, LiDAR, map, or object-track inputs. We find that vehiclestate history alone is sufficient to learn a useful multimodal distribution over near-term ego trajectories, and its low-latency nature makes it well-suited for safety-critical deployment. We investigate two distribution architectures, autoregressive (AR) and flow-matching, and train on both mixed-platform opensource and Wayve datasets. Both generalize without datasetspecific adaptation; the flow-matching model achieves best topk ADE/FDE of 0.15/0.37 m on Wayve, 0.15/0.36 m on Waymo, and 0.28/0.59 m on nuScenes, with the AR model reaching comparable coverage. Integrating the distributions into an AEB trigger task, a strict all-candidates policy reduces false positives from 1.51% to 0.11% with AR (13.7x reduction, 94.9% TP) and to 0.06% with flow-matching (25.1x reduction, 98.7% TP) compared to a 1-CTRV policy with 100% true positive score. BLInD runs in 0.87 ms with the AR head and 2.9 ms with the flow-matching head on an NVIDIA DRIVE Orin ECU making it compatible with real-time deployment on automotive ECUs. While existing learned distribution models rely on scene context and blind vehicle-state models typically collapse to a single path, BLInD is learned, blind, and cross-domain simultaneously, a combination not demonstrated by prior work. These results show that such a distribution provides a controllable and plausible intent sampling interface for downstream systems, with AEB as one instantiation.

Chinese Translation

我们提出了BLInD（盲学习意图分布），这是一个紧凑的网络，能够将近期的车辆状态历史（例如速度、曲率、转向灯和车辆类型）映射到未来自我轨迹的前k个分布，而无需摄像头、激光雷达、地图或物体跟踪输入。我们发现，仅凭车辆状态历史就足以学习出有用的多模态分布，用于近期的自我轨迹，其低延迟特性使其非常适合安全关键的部署。我们研究了两种分布架构，自回归（AR）和流匹配，并在混合平台的开源数据集和Wayve数据集上进行训练。两者均能在没有特定数据集适应的情况下进行泛化；流匹配模型在Wayve上实现了最佳的前k ADE/FDE为0.15/0.37米，在Waymo上为0.15/0.36米，在nuScenes上为0.28/0.59米，而AR模型则达到相似的覆盖率。在将这些分布集成到自动紧急制动（AEB）触发任务中时，严格的全候选策略使得AR的误报率从1.51%降低到0.11%（减少13.7倍，94.9%真阳性率），而流匹配则降低到0.06%（减少25.1倍，98.7%真阳性率），相比于具有100%真阳性率的1-CTRV策略。BLInD在NVIDIA DRIVE Orin ECU上以0.87毫秒的时间运行AR头，以2.9毫秒的时间运行流匹配头，使其适用于汽车ECU的实时部署。尽管现有的学习分布模型依赖于场景上下文，而盲车辆状态模型通常会崩溃为单一路径，BLInD则同时实现了学习、盲和跨域的结合，这是之前工作未曾展示的。这些结果表明，这种分布为下游系统提供了一个可控且合理的意图采样接口，以AEB作为一种实例。

View on arXiv Download PDF AI Translation

cs.CV / 69 / 2609.13966

SGWIB:Sliced Gromov-Wasserstein Information Bottleneck for Video Highlight Detection

SGWIB：用于视频亮点检测的切片 Gromov-Wasserstein 信息瓶颈

Huang, Hanjuan, Yeh, Yung-Chieh, Pao, Hsing-Kuo

Abstract

Video highlight detection aims to identify temporally important segments that capture the most informative or engaging events in a video. Reliable prediction therefore requires not only discriminative segment representations but also preservation of the temporal relationships among neighboring and distant segments. The information bottleneck principle has proven effective for learning compact and task-relevant representations, yet it has not been explored for video highlight detection, and applying conventional formulations directly would overlook inter-segment relational structure and distort highlight relevant temporal organization during compression. We therefore introduce the Sliced Gromov-Monge Gap (SGMG), a structure aware regularizer that measures the excess relational distortion induced by a prescribed source-to-bottleneck mapping relative to an optimal sliced structural correspondence. Building on SGMG, we develop SGWIB, an information-bottleneck framework for single-modal video highlight detection that learns compact bottleneck representations while preserving inter-segment temporal structure. We further introduce Home-Away-Related Contextual Pseudo-Labels and a contextual disentanglement module that reduce sports-specific contextual bias by separating highlight oriented information from contextual patterns. Experiments on MrHiSum and MoSu show that SGWIB attains the best Kendall's tau, Spearman's rho, mAP@50, and mAP@30 among the compared single-modal methods on both datasets. On MrHiSum, the visual model improves the strongest previous results by 0.031, 0.031, 0.87, and 0.75 on these four metrics, respectively. These results show that structure-aware information-bottleneck regularization combined with contextual disentanglement improves segment-level highlight prediction.

Chinese Translation

视频亮点检测旨在识别时间上重要的片段，这些片段捕捉视频中最具信息性或吸引力的事件。因此，可靠的预测不仅需要具有区分性的片段表示，还需要保留相邻和远距离片段之间的时间关系。信息瓶颈原理已被证明在学习紧凑且与任务相关的表示方面有效，但尚未在视频亮点检测中进行探索，直接应用传统公式将忽视片段间的关系结构，并在压缩过程中扭曲与亮点相关的时间组织。因此，我们引入了切片 Gromov-Monge Gap (SGMG)，这是一种结构感知的正则化器，测量由规定的源到瓶颈映射引起的过量关系扭曲，相对于最佳切片结构对应关系。在此基础上，我们开发了 SGWIB，一种用于单模态视频亮点检测的信息瓶颈框架，能够学习紧凑的瓶颈表示，同时保留片段间的时间结构。我们进一步引入了与家庭-外部相关的上下文伪标签和一个上下文解缠模块，通过将与亮点相关的信息与上下文模式分离，减少了特定于体育的上下文偏差。在 MrHiSum 和 MoSu 数据集上的实验表明，SGWIB 在这两个数据集的比较单模态方法中达到了最佳的 Kendall's tau、Spearman's rho、mAP@50 和 mAP@30。在 MrHiSum 上，视觉模型在这四个指标上分别提高了之前最强结果 0.031、0.031、0.87 和 0.75。这些结果表明，结合结构感知的信息瓶颈正则化与上下文解缠能够改善片段级亮点预测。

View on arXiv Download PDF AI Translation

cs.CV / 70 / 2609.13969

Zero-Shot Cross-Material Ptychographic Phase Reconstruction Using Deep Learning

基于深度学习的零-shot跨材料相位重建

Lin, Wen-Chun, Tseng, Yu-Chee, Chen, Jen-Jee, Chen, Nan-You

Abstract

Ptychographic phase reconstruction is commonly formulated as an iterative inverse problem, requiring repeated object-probe updates and resulting in substantial computational cost for large-scale 4D-STEM data. We present a direct local-to-global learning framework that reconstructs full-field phase maps from diffraction measurements without iterative refinement during inference. The proposed network predicts local wrapped-phase patches from individual diffraction patterns using a sine-cosine representation, and the predictions are assembled into a full-field reconstruction using calibrated scan positions and Gaussian-weighted stitching. To evaluate generalization beyond the training domain, the model is trained on one material and directly applied to another in a zero-shot setting without target-domain fine-tuning. Experiments on AuPd and MoS$_2$ demonstrate consistent cross-material transfer in both directions, with the proposed method achieving the best full-field MSE, PSNR, and MS-SSIM among the evaluated learning-based methods. Compared with the iterative ePIE approach, the proposed direct local-to-global pipeline reduces end-to-end reconstruction time by approximately 10x, demonstrating its potential for efficient and transferable ptychographic reconstruction.

Chinese Translation

Ptychographic相位重建通常被表述为一个迭代逆问题，需要重复的物体-探针更新，这导致在大规模4D-STEM数据处理时产生显著的计算成本。我们提出了一种直接的局部到全局学习框架，该框架能够从衍射测量中重建全场相位图，而在推理过程中无需迭代优化。所提出的网络使用正弦-余弦表示法从单个衍射模式预测局部包裹相位块，并利用校准的扫描位置和高斯加权拼接将这些预测组装成全场重建。为了评估模型在训练域之外的泛化能力，该模型在一种材料上进行训练，并在零-shot设置下直接应用于另一种材料，无需目标域的微调。对AuPd和MoS$_2$的实验表明，该方法在两个方向上实现了一致的跨材料迁移，所提出的方法在评估的基于学习的方法中实现了最佳的全场均方误差（MSE）、峰值信噪比（PSNR）和多尺度结构相似性指数（MS-SSIM）。与迭代的ePIE方法相比，所提出的直接局部到全局流程将端到端重建时间减少了约10倍，展示了其在高效和可迁移的ptychographic重建中的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 71 / 2609.13991

Mind2Cloud: EEG-to-Point Cloud Generation with Two-Granularity Diffusion Decoding

Mind2Cloud：基于双粒度扩散解码的脑电图到点云生成

Lu, Yongyi, Huang, Xiongfeng, Yang, Zhijing

Abstract

Reconstructing 3D objects from brain signals offers a promising avenue for understanding human visual cognition. While prior work has shown initial success using EEG signals for 3D reconstruction, existing methods typically employ a uniform diffusion decoder, overlooking the evolving semantic granularity of both EEG representations and the diffusion denoising process. In this paper, we propose Mind2Cloud, a novel EEG-to-point-cloud generation framework based on two-granularity diffusion decoding. The core of Mind2Cloud is a time-aware decoder that integrates a global Transformer branch and a local Point-Voxel CNN (PVCNN) branch across diffusion timesteps through a learnable fusion mask. Specifically, Transformer layers are incorporated into the early upsampling stages to capture global object structure under high uncertainty, while PVCNN modules are used in later stages to refine local geometric details. Inspired by the hierarchical nature of EEG-based visual representations, this design dynamically adapts its spatial granularity in accordance with the coarse-to-fine trajectory of diffusion denoising. We further introduce an adversarial refinement module to enhance geometric realism and semantic consistency. Extensive experiments on the EEG-3D dataset across all 12 subjects demonstrate that Mind2Cloud outperforms prior work in both geometric accuracy and semantic alignment, setting a new benchmark for EEG-to-point-cloud generation. Our source code is available at https://github.com/duasoi/Mind2Cloud.

Chinese Translation

从脑信号重建三维物体为理解人类视觉认知提供了一个有前景的途径。尽管之前的研究已经展示了使用脑电图（EEG）信号进行三维重建的初步成功，但现有方法通常采用统一的扩散解码器，忽视了EEG表示和扩散去噪过程中的语义粒度的演变。在本文中，我们提出了Mind2Cloud，一个基于双粒度扩散解码的创新EEG到点云生成框架。Mind2Cloud的核心是一个时间感知解码器，它通过可学习的融合掩码将全局Transformer分支和局部点-体素卷积神经网络（PVCNN）分支整合在扩散时间步中。具体而言，Transformer层被纳入早期上采样阶段，以捕捉在高不确定性下的全局物体结构，而PVCNN模块则用于后期阶段以细化局部几何细节。受到基于EEG的视觉表示的层次特性的启发，该设计动态调整其空间粒度，以符合扩散去噪的粗到细轨迹。我们进一步引入了一个对抗性细化模块，以增强几何真实感和语义一致性。在对12名受试者的EEG-3D数据集进行的广泛实验中，Mind2Cloud在几何精度和语义对齐方面均优于之前的工作，为EEG到点云生成设定了新的基准。我们的源代码可在https://github.com/duasoi/Mind2Cloud获取。

View on arXiv Download PDF AI Translation

cs.CV / 72 / 2609.14010

CirrGuide: A Deep Cascaded Framework for Liver Cirrhosis Segmentation and Severity Classification from T2-Weighted MRI

CirrGuide：一种用于肝硬化分割和严重程度分类的深度级联框架，基于T2加权MRI

Raju, Muntaqim Ahmed, Ma, Ruizhe

Abstract

We present CirrGuide, a deep cascaded framework for cirrhotic liver segmentation and severity classification. Cirrhosis causes progressive structural changes in the liver and can lead to serious clinical complications, making severity assessment important for disease monitoring and treatment planning. However, severity classification is challenging because imaging patterns are often subtle, spatially variable, and similar across adjacent stages. CirrGuide addresses this by explicitly linking localization with classification. A ResNet50 encoder with an Attention U-Net decoder first predicts a soft cirrhotic liver mask, which is then used as an anatomical prior in a ResNet50-based classification branch. This branch combines global multi-scale features with mask-guided attention-pooled regional features to classify Mild, Moderate, and Severe cirrhosis. On the official CirrMRI600+ T2-weighted (T2W) 2D split, CirrGuide achieves 89.83% Dice and 84.14% mIoU for segmentation, 69.58% accuracy and 61.55% macro F1-score for severity classification. Compared with segmentation-only, classification-only, and multi-task baselines, CirrGuide improves both localization and severity classification, demonstrating the benefit of using predicted cirrhotic liver masks as anatomical priors for cirrhosis analysis.

Chinese Translation

我们提出了CirrGuide，这是一种用于肝硬化肝脏分割和严重程度分类的深度级联框架。肝硬化会导致肝脏结构的逐渐变化，并可能引发严重的临床并发症，因此对疾病监测和治疗计划而言，严重程度评估显得尤为重要。然而，严重程度分类具有挑战性，因为影像模式往往微妙、空间变化且相邻阶段之间相似。CirrGuide通过明确将定位与分类相结合来解决这一问题。首先，使用带有注意力机制的U-Net解码器的ResNet50编码器预测一个软性肝硬化肝脏掩膜，然后将其作为解剖先验用于基于ResNet50的分类分支。该分支结合了全局多尺度特征与掩膜引导的注意力池化区域特征，以对轻度、中度和重度肝硬化进行分类。在官方的CirrMRI600+ T2加权（T2W）2D拆分数据集上，CirrGuide在分割任务中达到了89.83%的Dice系数和84.14%的mIoU，在严重程度分类中达到了69.58%的准确率和61.55%的宏观F1-score。与仅进行分割、仅进行分类和多任务基线相比，CirrGuide在定位和严重程度分类方面均有所提升，展示了将预测的肝硬化肝脏掩膜作为肝硬化分析的解剖先验的优势。

View on arXiv Download PDF AI Translation

cs.CV / 73 / 2609.14025

Quantum-Gated LiteSSD: A Parameter-Efficient Lightweight Hybrid Quantum-Classical Framework for Forward-Looking Sonar Object Detection

量子门控LiteSSD：一种参数高效的轻量级混合量子-经典框架用于前视声纳目标检测

Mondal, Niloy Kumar, Puja, Poulomi Sarker

Abstract

Forward-looking sonar object detection is essential for underwater perception, yet deployment on embedded platforms requires highly compact models. To address this challenge, we explore quantum computing and introduce Quantum-Gated LiteSSD, a parameter-efficient hybrid quantum--classical detector that reformulates QuCNet-style multi-circuit quantum processing as an identity-centered channel-gating mechanism for spatial feature modulation. Experiments on the Marine Debris Watertank dataset and UATD forward-looking sonar benchmarks demonstrate an effective parameter--accuracy trade-off. The proposed detector achieves 90.84% $\mathrm{mAP}_{50}$ on Watertank with approximately $62\times$ fewer parameters than YOLO26s and $164.3\times$ fewer parameters than SSD-VGG16. On UATD, the model achieves 70.37% $\mathrm{mAP}_{50}$ with only 0.150M parameters, making it approximately $4.1\times$ smaller than SSGA-YOLO while retaining meaningful multi-class detection capability.

Chinese Translation

前视声纳目标检测对于水下感知至关重要，但在嵌入式平台上的部署需要高度紧凑的模型。为了解决这一挑战，我们探索了量子计算，并引入了量子门控LiteSSD（Quantum-Gated LiteSSD），这是一种参数高效的混合量子-经典检测器，它将QuCNet风格的多电路量子处理重新构造成一种以身份为中心的通道门控机制，用于空间特征调制。在Marine Debris Watertank数据集和UATD前视声纳基准测试上的实验表明了有效的参数-准确性权衡。所提出的检测器在Watertank上达到了90.84%的$ ext{mAP}_{50}$，其参数数量约为YOLO26s的$62 imes$更少，且比SSD-VGG16少$164.3 imes$。在UATD上，该模型以仅0.150M的参数达到了70.37%的$ ext{mAP}_{50}$，使其比SSGA-YOLO小约$4.1 imes$，同时保留了有意义的多类检测能力。

View on arXiv Download PDF AI Translation

cs.CV / 74 / 2609.14073

LPA-CWM: A Learned Physical Adjudicator for Motion Reasoning with Counterfactual World Models

LPA-CWM：一种用于反事实世界模型运动推理的学习物理裁决者

Wu, Kunwei, Liu, Xiang, Yao, Guocai, Chen, Junming, Chen, Zhikang, Zhang, Min, Wang, Pengwei, Cui, Sen

Abstract

Counterfactual world models (CWM) extract motion from pretrained video predictors by comparing factual and intervened predictions. However, responses generated under different target-frame masks vary in reliability, while uniform aggregation weights them equally. We formulate response aggregation as candidate reliability learning and propose LPA-CWM with a lightweight Learned Physical Adjudicator (LPA). Trained on dense MOVi-F trajectories, the 3.0M-parameter LPA compares visual context and response structure across an unordered candidate set to predict relative weights, while the CWM predictor and intervention generator remain frozen. The weighted responses undergo windowed localization and one paired re-evaluation to recover motion. We also introduce Completeness-aware Motion Correspondence (CMC), a ground-truth-anchored evaluation protocol that jointly measures localization, trajectory completeness, visibility, and continuity, counting missing predictions as failures on visible dynamic points. On the evaluated DAVIS and Kinetics subsets, LPA-CWM improves $\mathrm{DCA}_{\mathrm{avg}}$ over Uniform CWM by 60.0\% and 29.0\%, respectively, and also improves tracking accuracy under TAP-Vid First. A quick overview is available at https://LPA-CWM.github.io.

Chinese Translation

反事实世界模型（CWM）通过比较事实预测和干预预测，从预训练的视频预测器中提取运动。然而，在不同目标帧掩码下生成的响应在可靠性上存在差异，而统一加权则使它们同等对待。我们将响应聚合形式化为候选可靠性学习，并提出了具有轻量级学习物理裁决者（LPA）的LPA-CWM。该模型在密集的MOVi-F轨迹上进行训练，拥有300万参数的LPA在无序候选集中比较视觉上下文和响应结构，以预测相对权重，同时CWM预测器和干预生成器保持不变。加权响应经过窗口定位和一次配对重新评估以恢复运动。我们还引入了完整性意识运动对应（CMC），这是一种基于真实值的评估协议，联合测量定位、轨迹完整性、可见性和连续性，将缺失的预测计入可见动态点的失败。在评估的DAVIS和Kinetics子集上，LPA-CWM在$ ext{DCA}_{ ext{avg}}$上分别比Uniform CWM提高了60.0 ext{%}和29.0 ext{%}，并且在TAP-Vid First下也提高了跟踪准确性。有关的快速概述可在https://LPA-CWM.github.io获取。

View on arXiv Download PDF AI Translation

cs.CV / 75 / 2609.14080

Adapting Open-Weight MLLMs to Generate Point Prompts for Electron Microscopy Segmentation

将开放权重的多模态大语言模型适应于生成电子显微镜分割的点提示

Mohinta, Samia, Cardona, Albert

Abstract

Promptable models such as microSAM segment electron microscopy (EM) images from point prompts, but automation requires generating prompts without user input. We ask whether open-weight multimodal large language models (MLLMs) can generate them from natural-language requests by returning coordinates to a frozen segmenter. To that end, we convert masks from three mitochondria datasets into training examples, pairing images and instructions with centroid coordinates, then train LoRA adapters while freezing the MLLM backbone and microSAM. We find that Qwen3-VL reaches segmentation AP$_{50}$ $0.736$ after supervised fine-tuning and reward optimization, up from $0.247$ without adaptation, while automatic prompt generation (APG) achieves $0.773$. In addition, two other MLLMs improve, reaching or exceeding APG. When compared with a supervised centroid-heatmap detector that reaches AP$_{50}$ $0.904$ for this mitochondria task, Qwen3-VL more closely matches the annotated point set and instance counts. Moreover, training on two public datasets transfers to an unseen third, while training on all three transfers to an independent EM volume. Robustness tests show stable performance under unseen formulations of the natural-language request, while the coordinates can be reused by a second segmenter. To our knowledge, this is the first feasibility study of open-weight MLLMs as EM point generators, providing an inspectable, language-directed link between localization and mask decoding.

Chinese Translation

可提示模型如 microSAM 从点提示中对电子显微镜（EM）图像进行分割，但自动化需要在没有用户输入的情况下生成提示。我们探讨开放权重的多模态大语言模型（MLLMs）是否能够通过自然语言请求生成这些提示，并将坐标返回给冻结的分割器。为此，我们将来自三个线粒体数据集的掩膜转换为训练示例，将图像和指令与质心坐标配对，然后在冻结 MLLM 主干和 microSAM 的情况下训练 LoRA 适配器。我们发现，经过监督微调和奖励优化后，Qwen3-VL 的分割 AP$_{50}$ 达到 $0.736$，而未经适应时为 $0.247$，同时自动提示生成（APG）达到了 $0.773$。此外，另外两个 MLLM 也有所提升，达到或超过 APG。与一个监督质心热图检测器相比，该检测器在此线粒体任务中达到 AP$_{50}$ $0.904$，Qwen3-VL 更加接近标注的点集和实例计数。此外，在两个公共数据集上的训练能够迁移到一个未见的第三个数据集，而在所有三个数据集上的训练则能迁移到一个独立的 EM 体积。稳健性测试表明，在未见的自然语言请求形式下表现稳定，同时坐标可以被第二个分割器重复使用。根据我们的知识，这是开放权重 MLLMs 作为 EM 点生成器的首次可行性研究，提供了一个可检查的、以语言为导向的定位与掩膜解码之间的联系。

View on arXiv Download PDF AI Translation

cs.CV / 76 / 2609.14097

Bridging the Synthetic-to-Real Gap for Few-Shot Cryo-ET Classification

弥合少样本冷冻电子断层成像分类中的合成与真实差距

Bharadwaj, Siddhant, Vashist, Ashish, Singh, Rashi, Vinodh, Pranav, Artham, Nishanth, Jiang, Runmin, Li, Xingjian, Xu, Min

Abstract

Subtomogram classification in cryo-electron tomography (cryo-ET) is a challenging problem due to the scarcity of labeled examples. While cryo-ET simulators can be adopted to generate unlimited synthetic data, the substantial domain gap between synthetic and real subtomograms hinders its practical utilization. In this work, we propose a novel synthetic-to-real adaptation framework with a learnable transformation module, bridging this gap at both the input and feature levels. Extensive experiments demonstrate that our method consistently outperforms existing transfer learning baselines in few-shot settings.

Chinese Translation

在冷冻电子断层成像（cryo-ET）中，亚断层分类是一个具有挑战性的问题，因为标记样本稀缺。虽然可以采用冷冻电子断层成像模拟器生成无限的合成数据，但合成亚断层与真实亚断层之间的显著领域差距阻碍了其实际应用。在本研究中，我们提出了一种新颖的合成到真实的适应框架，配备可学习的变换模块，从输入和特征层面弥合这一差距。大量实验表明，我们的方法在少样本设置中始终优于现有的迁移学习基准。

View on arXiv Download PDF AI Translation

cs.CV / 77 / 2609.14100

RA-CoA: Training-free Fashion Image Captioning via Retrieval-Augmented Chain-of-Attributes

RA-CoA：通过检索增强属性链进行无训练时尚图像描述

Penamakuri, Abhirama Subramanyam, Shukla, Shreya, Mishra, Anand

Abstract

Fashion Image Captioning (FIC) plays a vital role in enhancing user experience and product search in e-commerce platforms. Unlike natural scene image captioning, FIC requires fine-grained visual reasoning and knowledge of domain-specific terminology to capture subtle attributes such as neckline and closure types, graphic patterns, and dress silhouettes. Moreover, as fashion inventories evolve rapidly with new trends, styles, and frequently emerging vocabulary, developing training-free captioning solution becomes essential for scalability and real-world adaptability. Instruction-tuned vision-language models (VLMs) offer a promising solution to fashion image captioning dueto their strong zero-shot capabilities and natural language fluency. However, these general-purpose models often lack attribute-level coverage and precision, and tend to hallucinate or misidentify fine-grained fashion details, making them less suitable for high-fidelity applications like product cataloging or personalized recommendations. To address this, we propose RA-CoA (Retrieval-Augmented Chain-of-Attributes), a novel, training-free framework that disentangles fashion image captioning into two interpretable stages: (i) retrieval of relevant attribute sets from a product knowledge base, and (ii) attribute-level reasoning to generate the final caption. RA-CoA is a model-agnostic approach that works with frozen VLMs to improve fine-grained attribute precision in product captions without the need for fine-tuning. Extensive evaluations across diverse VLM model families under different prompting paradigms demonstrate that RA-CoA significantly improves caption quality, achieving an average gain of 26.3% METEOR score over zero-shot captioning. We make our code publicly available.

Chinese Translation

时尚图像描述（FIC）在提升用户体验和电子商务平台上的产品搜索中发挥着至关重要的作用。与自然场景图像描述不同，FIC 需要细致的视觉推理和领域特定术语的知识，以捕捉如领口和闭合类型、图案设计以及服装轮廓等细微属性。此外，随着时尚库存因新趋势、风格和频繁出现的词汇而迅速演变，开发无训练的描述解决方案对于可扩展性和现实世界的适应性变得至关重要。经过指令调优的视觉-语言模型（VLMs）由于其强大的零样本能力和自然语言流畅性，为时尚图像描述提供了一个有前景的解决方案。然而，这些通用模型往往缺乏属性级的覆盖和精确度，容易产生幻觉或错误识别细致的时尚细节，使其不太适合用于高保真应用，如产品目录或个性化推荐。为了解决这个问题，我们提出了 RA-CoA（检索增强属性链），这是一种新颖的无训练框架，将时尚图像描述分解为两个可解释的阶段：（i）从产品知识库中检索相关属性集，以及（ii）进行属性级推理以生成最终描述。RA-CoA 是一种模型无关的方法，能够与冻结的 VLMs 一起工作，以提高产品描述中的细致属性精度，而无需微调。我们在不同提示范式下对多种 VLM 模型家族进行了广泛评估，结果表明 RA-CoA 显著提高了描述质量，平均获得了 26.3% 的 METEOR 分数提升，超越了零样本描述。我们将代码公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 78 / 2609.14103

A Voxel-Spacing-Aware Extension of PyRadiomics for Anisotropic Texture Analysis

一种考虑体素间距的PyRadiomics扩展用于各向异性纹理分析

Fontecha, David Corral, Bautista, Juan Miranda, Fernández-Miranda, Pablo Menendez, Fernandez, Andrea Trapote, Iglesias, Lara Lloret, Vega, Jose A.

Abstract

Radiomic texture features are commonly extracted from anisotropic CT and MRI acquisitions, where identical voxel offsets may represent different physical distances. We implemented and validated a voxel-spacing-aware extension of PyRadiomics that incorporates spacing information without generating interpolated gray levels. The framework operates across the Python frontend, C wrapper, and computational backend. GLCM uses anisotropy-relative feature-level angular aggregation, NGTDM uses anisotropy-relative weighted neighborhood averaging, and GLRLM, GLDM, and GLSZM are computed on a finite-volume zero-order-hold representation derived from the native anisotropic grid. Synthetic 3D phantoms were used for software validation. The modified implementation reproduced standard PyRadiomics exactly when spacing-aware mode was disabled and remained equivalent under isotropic spacing across 75 texture features. Under anisotropic spacing, the method selectively modified texture families and was numerically distinct from nearest-neighbor, linear, and B-spline resampling. Computational profiling showed moderate runtime and memory increases, while sensitivity analyses quantified finite-volume rounding effects and confirmed that spacing-aware differences persisted across binWidth settings. The framework provides a backward-compatible technical basis for future evaluation of spacing-aware radiomics in heterogeneous medical imaging datasets.

Chinese Translation

放射组学纹理特征通常从各向异性CT和MRI图像中提取，其中相同的体素偏移可能代表不同的物理距离。我们实现并验证了一种考虑体素间距的PyRadiomics扩展，该扩展在不生成插值灰度值的情况下融入了间距信息。该框架在Python前端、C语言封装和计算后端之间运行。灰度共生矩阵（GLCM）使用相对各向异性特征级角度聚合，邻域灰度差异矩阵（NGTDM）使用相对各向异性加权邻域平均，而灰度长度运行矩阵（GLRLM）、灰度分布矩阵（GLDM）和灰度大小区域矩阵（GLSZM）则在从原生各向异性网格导出的有限体积零阶保持表示上计算。我们使用合成3D幻影进行软件验证。当禁用间距感知模式时，修改后的实现与标准PyRadiomics完全一致，并且在75个纹理特征下在各向同性间距下保持等效。在各向异性间距下，该方法选择性地修改纹理家族，并在数值上与最近邻、线性和B样条重采样有所不同。计算性能分析显示运行时间和内存的适度增加，而敏感性分析量化了有限体积舍入效应，并确认间距感知差异在不同的binWidth设置下持续存在。该框架为未来在异质医学影像数据集中评估间距感知放射组学提供了向后兼容的技术基础。

View on arXiv Download PDF AI Translation

cs.CV / 79 / 2609.14105

Accelerating HKTex without Mesh Eigensystems: Local Unfolding and Randomized Thermal Features

无需网格特征值系统加速 HKTex：局部展开与随机热特征

He, Zhewen, Hu, Junyi, Fang, Yi

Abstract

Heat Kernel Textures (HKTex) represent surface appearance with intrinsic anisotropic kernels, but evaluate them using 50 global Laplace-Beltrami eigendecompositions and a resident basis of shape [50,V,256]. We study two complementary ways to remove this bottleneck while leaving the trainer, GeodesicOpt, density control, and compositing unchanged. LocalHK exploits the measured locality of trained kernels and replaces spectral evaluation by radius-bounded hinge unfolding and an analytic log-map kernel. On 10 Objaverse meshes and an 8-mesh low-poly holdout, it changes mean view PSNR from 31.35 to 32.00 and from 29.82 to 30.76 dB, respectively, while reducing initialization by 40.5 times and enabling a 749,570-vertex proxy-backed run where the spectral baseline fails. ThermalRF instead preserves the discrete anisotropic heat semigroup: GPU sparse Chebyshev actions and randomized range finding construct global low-rank heat factors without mesh-sized eigenvectors, and a compiled evaluator mixes four neighboring thermal responses. On spot and a thin-stem challenge, ThermalRF reduces end-to-end preprocessing, initialization, and 5,000-step optimization by 29.3% and 24.4%, with every surface, atlas, or view PSNR change within 0.12 dB and training allocation reduced by about 90%. The two routes expose a useful design choice: maximal locality and scale versus fidelity to the thermal PDE. Broader thermal-feature evaluation and real large scenes remain future work.

Chinese Translation

热核纹理（HKTex）通过内在的各向异性核表示表面外观，但需要使用 50 个全局拉普拉斯-贝尔特拉米特征分解和一个形状为 [50,V,256] 的驻留基进行评估。我们研究了两种互补的方法来消除这一瓶颈，同时保持训练器 GeodesicOpt、密度控制和合成不变。LocalHK 利用训练核的测量局部性，通过半径限制的铰链展开和解析对数映射核替代谱评估。在 10 个 Objaverse 网格和一个 8 网格低多边形保留集上，它将平均视图 PSNR 从 31.35 提高到 32.00，以及从 29.82 提高到 30.76 dB，同时将初始化时间减少了 40.5 倍，并实现了一个 749,570 顶点的代理支持运行，在谱基线失败的情况下。ThermalRF 则保留离散的各向异性热半群：GPU 稀疏切比雪夫操作和随机范围查找构建全局低秩热因子，而无需网格大小的特征向量，编译的评估器混合了四个相邻的热响应。在点状和细茎挑战中，ThermalRF 将端到端预处理、初始化和 5,000 步优化的时间分别减少了 29.3% 和 24.4%，每个表面、图集或视图的 PSNR 变化均在 0.12 dB 以内，训练分配减少了约 90%。这两条路径揭示了一个有用的设计选择：最大局部性与规模对比热 PDE 的保真度。更广泛的热特征评估和真实大型场景仍然是未来的工作。

View on arXiv Download PDF AI Translation

cs.CV / 80 / 2609.14106

Parameter-Efficient Fine-Tuning of Foundation Models for Liver Tumor Segmentation in CT

用于肝肿瘤分割的基础模型参数高效微调在CT中的应用

Mojtahedi, Ramtin, Hamghalam, Mohammad, Peoples, Jacob J., Do, Richard K. G., Simpson, Amber L.

Abstract

We evaluated parameter-efficient fine-tuning (PEFT) of the Segment Anything Model (SAM) for liver tumor segmentation in abdominal CT of colorectal liver metastases. We compared Low-Rank Adaptation (LoRA), 4-bit Quantized LoRA (QLoRA), a convolutional adapter (Conv-Adapter), and our Directional Spectral Top-K adapter (DiSCo), training only adapters while freezing the SAM backbone. DiSCo derives spectral bases from singular value decomposition of row-normalized weights and learns rank-gated spectral coefficients, per-output magnitude offsets, and a spectral gain, with optional Top-K rank selection at inference and 0.14 M trainable parameters. We benchmarked five prompting regimes: no prompt, single-point, multi-point, and bounding boxes at intersection over union 0.50 and 0.75. Conv-Adapter and LoRA achieved the highest accuracy (overall Dice 0.793 and 0.792; single-point Dice 0.795 and 0.792; 95th-percentile Hausdorff distance (HD95) 32 mm). QLoRA was close (overall Dice 0.766; single-point Dice 0.768; HD95 36.41 mm), with 0.91 M trainable parameters, 120 ms latency, and 4.9 GB peak memory. DiSCo achieved the highest Dice per million trainable parameters (4.66), with overall Dice 0.653, single-point Dice 0.698, and HD95 49.53 mm. These results show an accuracy-efficiency trade-off and support PEFT for liver tumor segmentation with reduced adaptation costs when compute and labeled data are limited. Code: https://github.com/Ramtin-Mojtahedi/PEFT-SAM-Liver-CT

Chinese Translation

我们评估了Segment Anything Model (SAM) 在结直肠肝转移的腹部CT中进行肝肿瘤分割的参数高效微调（PEFT）。我们比较了低秩适应（Low-Rank Adaptation, LoRA）、4位量化LoRA（QLoRA）、卷积适配器（Conv-Adapter）以及我们的方向谱Top-K适配器（Directional Spectral Top-K adapter, DiSCo），在冻结SAM主干的情况下，仅训练适配器。DiSCo通过对行归一化权重进行奇异值分解，推导出谱基，并学习秩门控谱系数、每输出幅度偏移和谱增益，并在推理时可选地进行Top-K秩选择，具有0.14M可训练参数。我们基准测试了五种提示机制：无提示、单点、多点和在交并比为0.50和0.75的边界框。Conv-Adapter和LoRA达到了最高的准确率（总体Dice 0.793和0.792；单点Dice 0.795和0.792；95百分位Hausdorff距离（HD95）32毫米）。QLoRA的表现接近（总体Dice 0.766；单点Dice 0.768；HD95 36.41毫米），具有0.91M可训练参数、120毫秒延迟和4.9GB峰值内存。DiSCo在每百万可训练参数中达到了最高的Dice（4.66），总体Dice为0.653，单点Dice为0.698，HD95为49.53毫米。这些结果显示了准确性与效率之间的权衡，并支持在计算和标记数据有限的情况下进行肝肿瘤分割的PEFT，降低了适应成本。代码链接： https://github.com/Ramtin-Mojtahedi/PEFT-SAM-Liver-CT

View on arXiv Download PDF AI Translation

cs.CV / 81 / 2609.14118

Talking to Me or Someone Else? Rethinking Talk-to-Me Detection in Egocentric Videos

与我交谈还是与他人交谈？重新思考自我中心视频中的交谈检测

Du, Feiyu, He, Xi, Li, Jia, Tian, Yapeng, Wu, Weili

Abstract

Online understanding of who is talking to the camera wearer is a key capability for egocentric social interaction. However, existing talk-to-me (TTM) studies are commonly formulated as offline clip-level recognition, which is poorly aligned with online interaction and overlooks the diverse non-TTM speaking states that naturally arise in egocentric videos. In this paper, we revisit this problem by reformulating it as an online, frame-level prediction task. Instead of treating TTM as a binary problem against a single negative class, we model it in the presence of diverse and previously underexplored non-TTM states, such as talking-to-others, self-talking, and background conditions. To support this new formulation, we construct an Online TTM Dataset consisting of 406 egocentric video clips with approximately 900K annotated frames, each labeled with frame-level social interaction categories (e.g., background, TTM, talking-to-others, self-talking), by extending the Ego4D social interaction benchmark. In this benchmark, we evaluate five adapted baselines and develop a new model that integrates social cues across modalities. Experimental results show that our multimodal model, which jointly leverages audio, visual, and speech-semantic cues, achieves 75.5% frame-level F1 on TTM, outperforming strong baselines and enabling a systematic analysis of how different speaking states affect TTM recognition.

Chinese Translation

在线理解谁在与摄像头佩戴者交谈是自我中心社交互动的一项关键能力。然而，现有的交谈检测（TTM）研究通常被表述为离线剪辑级别的识别，这与在线互动的需求不符，并且忽视了在自我中心视频中自然出现的多样化非TTM说话状态。本文通过将这一问题重新表述为在线帧级预测任务来重新审视这一问题。我们不再将TTM视为针对单一负类的二元问题，而是在存在多样化且以前未充分探索的非TTM状态（如与他人交谈、自言自语和背景条件）下进行建模。为了支持这一新表述，我们构建了一个在线TTM数据集，包含406个自我中心视频剪辑，约有90万个标注帧，每帧标注有帧级社交互动类别（例如，背景、TTM、与他人交谈、自言自语），该数据集是通过扩展Ego4D社交互动基准而成。在该基准中，我们评估了五个适应性基线，并开发了一种新模型，该模型整合了跨模态的社交线索。实验结果表明，我们的多模态模型联合利用音频、视觉和语音语义线索，在TTM上达到了75.5%的帧级F1值，超越了强基线，并使我们能够系统分析不同说话状态如何影响TTM识别。

View on arXiv Download PDF AI Translation

cs.CV / 82 / 2609.14122

SignMimic: Robust High-Quality Sign Language Motion Generation via Human-Shape-Oblivious Pose Transfer Guidance

SignMimic：通过人形无关的姿态转移指导实现稳健的高质量手语动作生成

He, Zhewen, Yu, Junyi, Huang, Haomian, Li, Zhenhua, Fang, Yi

Abstract

We study the challenge of sign language video mimicking: given a driving video and a single reference frame, synthesize a video where the target signer reproduces the source motion while preserving identity and linguistic form. Prior pipelines entangle rigid motion, non-rigid deformation, and view-dependent completion in a monolithic generator, causing handshape drift and spatio-temporal instability. We present SignMimic, which (i) applies a TNet-based model to study SE(3) rigid canonicalization to stabilize global pose, (ii) performs non-rigid adaptation in a canonical space to preserve fine-grained articulators (hands/face) and coarticulation via NIF2D, and (iii) uses Pose-MAE-style completion before conditional video diffusion. This factorization injects geometric and linguistic priors, yielding shape and spatio-temporal consistency. On several large-scale datasets (ASL 50K, How2Sign, CSL News), SignMimic achieves state-of-the-art-level performance on video quality, identity similarity, and frame continuity while also achieving minimal loss when performing back translation (SLT) on generated videos. Ablations confirm the role of rigid canonicalization, non-rigid adaptation, and completion. Code, model checkpoints, and video examples will be released.

Chinese Translation

我们研究了手语视频模仿的挑战：给定一个驱动视频和一个单一参考帧，合成一个视频，使目标手语者在保持身份和语言形式的同时重现源动作。之前的流程将刚性运动、非刚性变形和视角依赖的补全纠缠在一个单一生成器中，导致手形漂移和时空不稳定。我们提出了SignMimic，它（i）应用基于TNet的模型研究SE(3)刚性标准化，以稳定全局姿态，（ii）在标准化空间中进行非刚性适应，以通过NIF2D保持细粒度的发音器（手/面部）和共发音，（iii）在条件视频扩散之前使用Pose-MAE风格的补全。这种因式分解注入了几何和语言先验，产生了形状和时空一致性。在多个大规模数据集（ASL 50K、How2Sign、CSL News）上，SignMimic在视频质量、身份相似性和帧连续性方面达到了最先进的性能，同时在生成视频上执行反向翻译（SLT）时也实现了最小损失。消融实验确认了刚性标准化、非刚性适应和补全的作用。代码、模型检查点和视频示例将被发布。

View on arXiv Download PDF AI Translation

cs.CV / 83 / 2609.14159

3D Gait-Based Autism Classification Using Attention-Enhanced Deep Learning with Cross-Fold Statistical Stability Analysis

基于步态的自闭症三维分类：使用增强注意力深度学习与交叉折叠统计稳定性分析

Mahamood, Md Nadim, Shahriar, Md Arif, Sikder, Md Parvej, Islam, Md Rasul, Doula, Md Shafi Ud, Alam, Md Ashraful, Hasan, Kamrul

Abstract

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition whose early diagnosis remains challenging because conventional clinical assessments are often subjective, time-consuming, and require expert evaluation. Gait provides a promising non-invasive behavioral biomarker for auto- mated ASD screening; however, existing studies have primarily relied on single-dataset evaluations, convolutional architectures, and descriptive summaries of cross-validation performance without formally assessing fold-to-fold stability. This study addresses these gaps with an attention-enhanced Transformer framework for ASD classification, evaluated on two structurally different 3D gait feature representations: precomputed statistical gait descriptors and raw biomechanical ground-reaction- force measurements. Under five-fold cross-validation, the proposed framework achieved 99.00% accuracy, 99.02% precision, 99.00% recall, 99.00% F1-score, and 99.00% specificity on the public Kinect-based benchmark, exceeding the performance of the compared state-of-the-art methods. On the independent private force-plate dataset, it achieved mean values of 95.00% accuracy, 93.81% precision, 96.67% recall, 95.13% F1-score, and 93.33% specificity.

Chinese Translation

自闭症谱系障碍（ASD）是一种神经发育疾病，其早期诊断仍然具有挑战性，因为传统的临床评估往往是主观的、耗时的，并且需要专家评估。步态提供了一种有前景的非侵入性行为生物标志物，用于自动化的ASD筛查；然而，现有研究主要依赖于单一数据集评估、卷积架构以及交叉验证性能的描述性总结，而未正式评估折叠间的稳定性。本研究采用增强注意力的Transformer框架进行ASD分类，评估了两种结构不同的三维步态特征表示：预计算的统计步态描述符和原始生物力学地面反作用力测量。在五折交叉验证下，所提出的框架在基于Kinect的公共基准上达到了99.00%的准确率、99.02%的精确率、99.00%的召回率、99.00%的F1-score和99.00%的特异性，超越了比较的最先进方法的表现。在独立的私有力板数据集上，其平均值为95.00%的准确率、93.81%的精确率、96.67%的召回率、95.13%的F1-score和93.33%的特异性。

View on arXiv Download PDF AI Translation

cs.CV / 84 / 2609.14185

Bi-Level Routing and Sparse Spatial Attention based Multi-View BEV 3D Object Detection for Autonomous Driving

基于双层路由和稀疏空间注意力的多视角鸟瞰图3D物体检测用于自动驾驶

Zhang, Jing, Liu, Jiaqi, Wang, Zibo

Abstract

Bird's Eye View (BEV)-based multi-view 3D object detection suffers from challenges of computational complexity, multi-scale feature extraction, and efficiency of dense 2D-to-BEV view transformation. To address these problems, this paper proposes an improved BEV 3D object detection algorithm Sparse-BEVNet. Firstly, a Bi-Level Routing Attention (BRA) mechanism is introduced into the image feature extraction network to reduce the computational burden of the backbone. Second, Cascaded Group Attention (CGA) is employed in the feature fusion module, which enhances deep interaction across features of different hierarchical levels without introducing additional computational overhead. Furthermore, a Sparse Spatial Cross-Attention mechanism is adopted to replace the conventional dense view projection pipeline. Experimental results on the public nuScenes dataset demonstrate that the proposed method achieves a mean Average Precision (mAP) of 45.2% and a nuScenes Detection Score (NDS) of 54.5%, corresponding to 3.6% and 2.8% improvements relative to the baseline model, respectively.

Chinese Translation

基于鸟瞰图（BEV）的多视角3D物体检测面临计算复杂性、多尺度特征提取以及密集2D到BEV视图转换效率等挑战。为了解决这些问题，本文提出了一种改进的BEV 3D物体检测算法Sparse-BEVNet。首先，将双层路由注意力（Bi-Level Routing Attention, BRA）机制引入图像特征提取网络，以减少主干网络的计算负担。其次，在特征融合模块中采用级联组注意力（Cascaded Group Attention, CGA），增强不同层级特征之间的深度交互，而不引入额外的计算开销。此外，采用稀疏空间交叉注意力机制替代传统的密集视图投影管道。对公共nuScenes数据集的实验结果表明，所提方法实现了45.2%的平均精度均值（mean Average Precision, mAP）和54.5%的nuScenes检测得分（nuScenes Detection Score, NDS），分别相较于基线模型提高了3.6%和2.8%。

View on arXiv Download PDF AI Translation

cs.CV / 85 / 2609.14189

MorphoStyle: Motion Style Transfer with Morphology Control

MorphoStyle：具备形态控制的动作风格迁移

Feng, Xin, D'Arnese, Eleonora, Sridharan, Mohan

Abstract

Human motion may be viewed as a combination of action content, style, and body morphology. Existing motion style transfer methods transfer a reference style onto a content motion while assuming a canonical body, whereas shape-aware motion generators adapt motion to a target shape without explicit style control. This separation of motion style and shape (morphology) makes it difficult to generate stylized motions for non-canonical bodies; naively combining a style transfer module with a shape-aware generator often leaks action content from the style reference and disrupts shape-consistent kinematics. We present MorphoStyle, a framework for shape-aware motion style transfer that is built on a shape-conditioned Finite-Scalar-Quantization Variational Auto-Encoder (FSQ-VAE). The key contribution is to pose the desired style transfer as modular latent disentanglement comprising: (i) a contrastive style encoder that extracts content-decoupled style embeddings; (ii) a text-guided style-routing mechanism that locates style-relevant joints in a text-motion feature space; and (iii) a manifold preserving style modulator that injects discriminative style embeddings in content features as a temporally-gated low-rank offset. Extensive experiments on benchmark datasets demonstrate that MorphoStyle outperforms competing baselines in terms of both shape control and motion style transfer, while simultaneously providing quantitative shape control. For more details, please see project website: https://github.com/funkdub/MorphoStyle.

Chinese Translation

人体动作可视为动作内容、风格与身体形态的组合。现有的动作风格迁移方法通常在假设标准身体形态的前提下，将参考风格迁移到内容动作上，而形态感知的动作生成器则能适应目标形态但缺乏显式的风格控制。这种动作风格与形态（形态学）的分离，使得为非标准身体生成风格化动作变得困难；简单地将风格迁移模块与形态感知生成器结合，往往会导致风格参考中的动作内容泄露，并破坏与形态一致的运动学。我们提出了 MorphoStyle，一种基于形态条件有限标量量化变分自编码器（Finite-Scalar-Quantization Variational Auto-Encoder，FSQ-VAE）的形态感知动作风格迁移框架。其核心贡献在于将期望的风格迁移表述为模块化的潜在解耦，具体包括：（i）对比式风格编码器，用于提取与内容解耦的风格嵌入；（ii）文本引导的风格路由机制，在文本-动作特征空间中定位与风格相关的关节；（iii）保持流形结构的风格调制器，以时间门控的低秩偏移方式，将判别性风格嵌入注入内容特征中。大量基准数据集上的实验表明，MorphoStyle 在形态控制与动作风格迁移方面均优于竞争基线方法，同时实现了定量的形态控制。更多详情请参见项目主页：https://github.com/funkdub/MorphoStyle。

View on arXiv Download PDF AI Translation

cs.CV / 86 / 2609.14247

What Input Resolution Is Required for Bird Species Identification, and What Is Its Latency Cost on an Edge Device? A Study of 14 Input Resolutions and Six Architectures with On-Device Measurements

鸟类物种识别所需的输入分辨率是多少，以及在边缘设备上的延迟成本是多少？基于14种输入分辨率和六种架构的设备内测量研究

Nishikawa, Takeshi

Abstract

Bird-strike mitigation at wind farms requires identifying distant birds that span only tens of pixels, so the classifier's input resolution N is a design variable, not a fixed specification. We study it with a factorial design over 14 side lengths N (16 to 224), six architectures, two training and evaluation regimes and 30 random seeds -- 2,520 checkpoints and 5,040 evaluations -- plus latency measured on an NVIDIA Jetson Orin Nano. Four results. (1) The selected N depends on the target: 0.90 is met on validation by ResNet50 at N=112 in an estimated 1.85 ms (0.8980 on test) and 0.95 by DINOv2-L at N=144 in 12.70 ms; changing the model buys more accuracy than raising N (+5.93 versus +2.33 points at N=112). (2) The benefit of lowering N depends on the assumed preprocessing path: N=224 -> 80 saves 13.5% when each individual is decoded from its own file but 46.2% when the detector decodes the 4K frame once; the Pareto set grows from 20 to 23 configurations. (3) Accuracy must be measured on the deployed engine: half precision costs ViT-S/16 alone 4 to 7 points at N>=96 while the CNNs stay within 0.1 points, and with selection held at validation the choice differs at 26 of 176 targets. A broken FP16 engine can run faster than a correct one, undetectable from latency; admitting 14 ViT-S/16 FP32 configurations moves the recommendation over the 0.931-0.938 band and under the 10 ms budget. (4) ViT-L-scale models fit this device, but activations exceed the FP16 range; splitting the graph at transformer-block boundaries confines FP32 to the affected segments, making the deployed DINOv2-L chain 1.85x faster than the single-engine build. We also quantify how the regime-difference sign stabilises with seed count; a sensitivity split removing some forms of group sharing preserves all 14 non-trivial signs at the selection boundary.

Chinese Translation

在风电场进行鸟击减缓需要识别仅占几十个像素的远处鸟类，因此分类器的输入分辨率 N 是一个设计变量，而非固定规范。我们通过对14种边长 N（16到224）、六种架构、两种训练和评估模式以及30个随机种子进行因子设计进行研究——共计2,520个检查点和5,040次评估——并在NVIDIA Jetson Orin Nano上测量延迟。四个结果：（1）所选的 N 取决于目标：在 N=112 时，ResNet50 在验证集上达到 0.90，估计延迟为 1.85 毫秒（在测试集上为 0.8980），而 DINOv2-L 在 N=144 时以 12.70 毫秒达到 0.95；更换模型比提高 N 带来更多的准确性（在 N=112 时分别为 +5.93 和 +2.33 分）。（2）降低 N 的好处取决于假设的预处理路径：从 N=224 降至 80，当每个个体从其自身文件解码时节省 13.5%，而当检测器一次解码 4K 帧时节省 46.2%；帕累托集从 20 种配置增长到 23 种。（3）准确性必须在部署的引擎上进行测量：在 N>=96 时，半精度的 ViT-S/16 单独损失 4 到 7 分，而 CNNs 保持在 0.1 分以内，并且在验证时选择保持不变时，选择在 176 个目标中有 26 个不同。一个损坏的 FP16 引擎可以比一个正确的引擎运行得更快，而延迟无法检测；接受 14 种 ViT-S/16 FP32 配置使推荐在 0.931-0.938 区间内，并低于 10 毫秒预算。（4）ViT-L 规模的模型适合该设备，但激活超出了 FP16 范围；在变换器块边界处拆分图形将 FP32 限制在受影响的段，使得部署的 DINOv2-L 链比单引擎构建快 1.85 倍。我们还量化了随着种子数量的增加，模式差异符号的稳定性；去除某些形式的组共享的敏感性分裂在选择边界保留了所有 14 个非平凡符号。

View on arXiv Download PDF AI Translation

cs.CV / 87 / 2609.14258

SPARK: Representation-Level KV Memory Alignment for Safer Vision-Language Models

SPARK：用于更安全的视觉-语言模型的表示级别KV内存对齐

Azfar, Mohd, Khan, Izhar Dad

Abstract

Vision-language models (VLMs) remain vulnerable to jailbreaks that distribute harmful intent across text and images, making unimodal safety mechanisms insufficient. We investigate whether this vulnerability can be mitigated directly in the multimodal key-value (KV) memory formed during prefill, without modifying model parameters at inference time. We introduce SPARK, a two-stage framework for targeted KV-memory repair. Stage 1 uses a disposable diagnostic adapter to identify harm-associated directions in multimodal key and value representations. Stage 2 projects out these directions, learns a lightweight residual repair, and anchors repaired keys with an image-structural prior to preserve visual grounding. Rather than applying the intervention uniformly, SPARK mixes repaired and original memory using a head-wise coefficient g_h* determined by intervention-relevant subspace energy E_h, requiring no explicit harm classifier at inference. Across LLaVA-OneVision-7B, Chameleon-7B, Qwen2-VL-7B, and InternVL2-4B, SPARK reduces multimodal attack success while preserving general capability. On LLaVA-OneVision-7B, image-only jailbreak attack success falls to 4.7%, while MMMU remains within 0.6 points of the undefended model (47.8 vs. 48.4) with near-baseline language quality. On MM-SafetyBench, attack success decreases from 39.2% to 12.4%. Even under white-box adaptive joint prompt-image attacks, attack success is limited to 20.3%, compared with 54.6% for the undefended model. These results suggest that multimodal jailbreak behavior can be substantially mitigated by selectively repairing intervention-relevant KV subspaces at prefill, particularly when harmful evidence is carried by the visual modality.

Chinese Translation

视觉-语言模型（VLMs）仍然容易受到越狱攻击，这些攻击在文本和图像中传播有害意图，使得单模态安全机制不足。我们研究了这种脆弱性是否可以直接在预填充过程中形成的多模态键值（KV）内存中缓解，而无需在推理时修改模型参数。我们提出了SPARK，一个用于针对性KV内存修复的两阶段框架。第一阶段使用一次性诊断适配器来识别多模态键和值表示中的有害关联方向。第二阶段投影出这些方向，学习轻量级残差修复，并通过图像结构先验锚定修复后的键，以保持视觉基础。SPARK并不是均匀地应用干预，而是使用由干预相关子空间能量E_h决定的头部系数g_h*混合修复后的内存和原始内存，在推理时无需显式的有害分类器。在LLaVA-OneVision-7B、Chameleon-7B、Qwen2-VL-7B和InternVL2-4B上，SPARK减少了多模态攻击的成功率，同时保持了整体能力。在LLaVA-OneVision-7B上，仅图像的越狱攻击成功率降至4.7%，而MMMUs仍在未防御模型的0.6个点以内（47.8与48.4），语言质量接近基线。在MM-SafetyBench上，攻击成功率从39.2%降至12.4%。即使在白盒自适应联合提示-图像攻击下，攻击成功率也限制在20.3%，而未防御模型为54.6%。这些结果表明，通过选择性修复预填充时与干预相关的KV子空间，可以显著缓解多模态越狱行为，特别是在有害证据由视觉模态承载时。

View on arXiv Download PDF AI Translation

cs.CV / 88 / 2609.14274

Sparsity-Adaptive Sharpness-Aware Minimization

稀疏自适应锐度感知最小化

Ueno, Shiryu, Hayashi, Yoshikazu, Kato, Kunihito

Abstract

Deploying deep neural networks in real-world settings requires models that are both compact and robust to common corruptions. However, at deployment-relevant high sparsity, standard pruning pipelines often degrade corruption robustness, and existing sharpness-aware training/pruning approaches provide limited robustness gains. We address this issue by introducing Sparsity-Adaptive Sharpness-Aware Minimization (SA-SAM), which derives a sparsity-dependent SAM/ASAM perturbation radius by keeping the mean absolute perturbation (an $\ell_1$-based proxy) approximately invariant as sparsity increases. As a simple complementary option, we evaluate Magnitude-Weighted Hessian (MWH), derived from a second-order removal-path analysis, yielding an importance proportional to $\mathrm{Diag}(F)_i\,|w_i|$, where $\mathrm{Diag}(F)$ is the diagonal empirical Fisher used as a curvature proxy in our implementation. Across CIFAR-10-C, CIFAR-100-C, and ImageNet-100-C, our approach achieved stronger corruption robustness than the considered pruning baselines at 80--90\% sparsity, while preserving clean accuracy. We additionally quantify the robustness--throughput trade-off by reporting measured inference throughput under sparse execution at deployment-relevant sparsity levels.

Chinese Translation

在现实世界环境中部署深度神经网络需要模型既紧凑又对常见干扰具有鲁棒性。然而，在与部署相关的高稀疏性下，标准剪枝流程往往会降低对干扰的鲁棒性，而现有的锐度感知训练/剪枝方法提供的鲁棒性提升有限。我们通过引入稀疏自适应锐度感知最小化（Sparsity-Adaptive Sharpness-Aware Minimization, SA-SAM）来解决这一问题，该方法通过保持均值绝对扰动（基于 $ ext{l}_1$ 的代理）在稀疏性增加时大致不变，从而推导出一个依赖于稀疏性的 SAM/ASAM 扰动半径。作为一个简单的补充选项，我们评估了基于二阶去除路径分析得出的幅度加权海森矩阵（Magnitude-Weighted Hessian, MWH），其重要性与 $ ext{Diag}(F)_i imes |w_i|$ 成正比，其中 $ ext{Diag}(F)$ 是在我们的实现中用作曲率代理的对角经验 Fisher。我们的研究在 CIFAR-10-C、CIFAR-100-C 和 ImageNet-100-C 数据集上，在 80% 到 90% 的稀疏性下，取得了比考虑的剪枝基线更强的干扰鲁棒性，同时保持了干净的准确性。此外，我们通过报告在与部署相关的稀疏性水平下的稀疏执行的推理吞吐量，量化了鲁棒性与吞吐量之间的权衡。

View on arXiv Download PDF AI Translation

cs.CV / 89 / 2609.14278

SpermYOLO: A Coordinated YOLO-Based Detector for Accurate and Efficient Sperm and Impurity Detection in Microscopic Images

SpermYOLO：一种基于YOLO的协调检测器，用于微观图像中精确高效的精子和杂质检测

Chen, Shengqi, Wang, Zilin, Pan, Xingyu, Yu, Wenting, Deng, Pengchao, Wu, Guohua

Abstract

Accurate sperm detection is essential for computer-assisted semen analysis, yet it remains challenging in microscopic images due to dense distributions, visually similar artifacts, and sperm-like impurities. In this paper, we propose SpermYOLO, a coordinated and compact YOLOv11-derived framework for joint sperm and impurity detection in microscopic images. SpermYOLO introduces four architectural improvements: C3k2-IDB for channel-wise discriminative feature extraction, D2SEM for spatial--spectral semantic enhancement, MFM for adaptive multi-scale feature fusion, and the DESD Head for detail-enhanced shared prediction. Experiments on the SVIA semen microscopic imaging benchmark show that SpermYOLO achieves 97.2\% sperm AP and 75.4\% impurity AP, outperforming generic detectors, dedicated sperm detection models, and improved YOLO variants. Compared with the baseline model, SpermYOLO improves sperm AP, impurity AP, $\mathrm{mAP}_{50}$, and $\mathrm{mAP}_{50:95}$ by 1.6, 10.0, 5.8, and 2.7 percentage points, respectively, while preserving a lightweight model scale. Cross-scene evaluation on the SDTB testicular-biopsy microscopy benchmark shows that SpermYOLO remains effective with extremely small sperm targets and complex tissue backgrounds, achieving the highest $\mathrm{mAP}_{50}$ and $\mathrm{mAP}_{50:95}$ of 74.8\% and 31.2\%, respectively. Ablation studies and qualitative analyses further support these improvements by demonstrating the contributions of the proposed modules and showing more focused feature response patterns than the baseline model. These findings suggest that SpermYOLO is an effective and efficient approach for sperm detection in challenging microscopic imaging scenarios.

Chinese Translation

精确的精子检测对于计算机辅助的精液分析至关重要，但由于密集分布、视觉上相似的伪影和精子样杂质，在微观图像中仍然具有挑战性。本文提出了SpermYOLO，一种基于YOLOv11的协调紧凑框架，用于微观图像中精子和杂质的联合检测。SpermYOLO引入了四项架构改进：C3k2-IDB用于通道级判别特征提取，D2SEM用于空间-光谱语义增强，MFM用于自适应多尺度特征融合，以及DESD Head用于细节增强的共享预测。在SVIA精液显微成像基准测试中的实验表明，SpermYOLO达到了97.2%的精子平均精度（AP）和75.4%的杂质AP，优于通用检测器、专用精子检测模型和改进的YOLO变体。与基线模型相比，SpermYOLO在精子AP、杂质AP、$ ext{mAP}_{50}$和$ ext{mAP}_{50:95}$上分别提高了1.6、10.0、5.8和2.7个百分点，同时保持了轻量级模型规模。在SDTB睾丸活检显微镜基准测试中的跨场景评估显示，SpermYOLO在极小的精子目标和复杂的组织背景下仍然有效，分别达到了74.8%和31.2%的最高$ ext{mAP}_{50}$和$ ext{mAP}_{50:95}$。消融研究和定性分析进一步支持了这些改进，展示了所提模块的贡献，并显示出比基线模型更集中的特征响应模式。这些发现表明，SpermYOLO是在具有挑战性的微观成像场景中进行精子检测的有效且高效的方法。

View on arXiv Download PDF AI Translation

cs.CV / 90 / 2609.14284

Vision-Language Models for Criterion-Level Grading of Handwritten Examinations in Outcome-Based Education

基于视觉-语言模型的结果导向教育中手写考试的标准级评分

Syfullah, Md Khalid, Tonmoy, Asif Hasan, Ahmed, Saad, Alam, S. M. Jahangir

Abstract

Criterion-level grading connects examination performance to learning outcomes, but manual marking introduces workload and variation between markers. This study evaluates vision-language models (VLMs) for handwritten outcome-based assessment across five dimensions: accuracy, human agreement, repeated-run reliability, error concentration, and explanation quality. Using 1,982 criterion-level records from 485 undergraduate examination answers, we compare 20 configurations spanning Qwen2.5-VL, InternVL3, Pixtral, a Donut baseline, and a cascade ensemble. Evaluation setups include zero-shot prompting, few-shot prompting, partial fine-tuning, and Low-Rank Adaptation (LoRA). Two independent faculty markers regraded all 291 test criteria, providing a human agreement baseline on the same assessment materials. Qwen2.5-VL with LoRA achieved Quadratic Weighted Kappa (QWK) of 0.727 and mean absolute error of 0.435 marks against the examiner, compared with mean human-pair QWK of 0.551. This comparison reflects calibration to the examiner's training marks. LoRA outperformed partial fine-tuning for all three instruction-tuned VLMs, while few-shot prompting reduced QWK in every configuration with valid prompted scores. Aggregate reliability and exact repeatability diverged: intraclass correlations ranged from 0.790 to 0.874, yet 50.2-63.6% of criteria changed marks across five sampled runs. Attention-guided deletion showed no statistically significant advantage over random masking, and four faculty reviewers reached no consensus on explanation usefulness. These findings highlight the need for rubric-specific calibration, repeatable scoring, review of consequential errors, and separate validation of explanations. The released evaluation protocol supports criterion-level assessment research and grading tools with teacher oversight.

Chinese Translation

标准级评分将考试表现与学习成果相连接，但人工评分引入了工作负担和评分者之间的差异。本研究评估了视觉-语言模型（VLMs）在手写结果导向评估中的应用，涵盖五个维度：准确性、人类一致性、重复运行可靠性、错误集中度和解释质量。我们使用来自485名本科生的1,982个标准级记录，比较了涵盖Qwen2.5-VL、InternVL3、Pixtral、Donut基线和级联集成的20种配置。评估设置包括零样本提示、少样本提示、部分微调和低秩适应（LoRA）。两名独立的教师评分者重新评分了所有291个测试标准，为相同评估材料提供了人类一致性基线。Qwen2.5-VL与LoRA的二次加权Kappa（QWK）达到了0.727，平均绝对误差为0.435分，而人类配对的平均QWK为0.551。这一比较反映了与评分者训练分数的校准。对于所有三种指令调优的VLMs，LoRA的表现优于部分微调，而在每种有效提示得分的配置中，少样本提示均降低了QWK。总体可靠性和精确重复性存在差异：组内相关系数范围为0.790到0.874，但在五次抽样运行中，50.2%-63.6%的标准更改了分数。基于注意力的删除在统计上未显示出相对于随机掩蔽的显著优势，四名教师评审员在解释的有用性上未达成共识。这些发现强调了特定评分标准的校准、可重复评分、对后果错误的审查以及对解释的单独验证的必要性。发布的评估协议支持标准级评估研究和教师监督下的评分工具。

View on arXiv Download PDF AI Translation

cs.CV / 91 / 2609.14307

Robust low-rank tensor completion via factorized weighted tensor schatten-p norm minimization

通过分解加权张量 Schatten-$p$ 范数最小化实现稳健的低秩张量补全

Wang, Binghao, Zhang, Feng, Wang, Wendong, Wang, Jianjun

Abstract

Low-rank tensor factorization provides a flexible framework for completing multidimensional data from incomplete and corrupted observations. However, unweighted spectral regularizers impose a common shrinkage profile across singular components, which may excessively attenuate dominant low-rank components, and factorized variants either lack component-specific weighting or require costly singular value decompositions (SVDs). This paper proposes two weighted Schatten-$p$ tensor factorization models, termed \WSpTFI{} and \WSpTFII{}, under the tensor-tensor product (t-product) framework to address these limitations. \WSpTFI{} is motivated by a factorized weighted tensor Schatten-$p$ norm identity and permits flexible, possibly asymmetric factor exponents. \WSpTFII{} constructs a regularizer from transform-domain column-pair energies, yielding SVD-free main factor updates and a column-pruning mechanism for reducing redundant rank components. This paper further develops an iteratively reweighted alternating direction method of multipliers (ADMM)-type scheme for \WSpTFI{} and an iteratively reweighted least squares (IRLS)--block successive upper-bound minimization (BSUM) scheme for \WSpTFII{}. Theoretical analysis establishes the weighted factorization relation and provides a conditional limiting Karush--Kuhn--Tucker (KKT) characterization for \WSpTFI{}. For \WSpTFII{}, the actual damped quadratic block updates yield a quantitative sufficient-decrease mechanism for the fixed-$\delta$ smoothed factor objective. This implies asymptotic regularity, and every accumulation point of the fixed-dimensional tail is stationary. Experiments on synthetic tensor completion, color-image restoration, hyperspectral inpainting, and printed-circuit-board defect detection demonstrate competitive reconstruction quality and robustness under various degradation conditions.

Chinese Translation

低秩张量分解为从不完整和损坏的观测中补全多维数据提供了灵活的框架。然而，无权重的谱正则化器在奇异成分上施加了共同的收缩特征，这可能会过度削弱主导的低秩成分，而分解变体要么缺乏特定成分的加权，要么需要昂贵的奇异值分解（SVD）。本文提出了两种加权 Schatten-$p$ 张量分解模型，称为 extit{WSpTFI} 和 extit{WSpTFII}，在张量-张量乘积（t-product）框架下解决这些局限性。 extit{WSpTFI} 的动机来自于分解加权张量 Schatten-$p$ 范数恒等式，并允许灵活的、可能不对称的因子指数。 extit{WSpTFII} 从变换域列对能量构建正则化器，产生无 SVD 的主因子更新和用于减少冗余秩成分的列修剪机制。本文进一步发展了一种迭代加权交替方向乘子法（ADMM）类型方案用于 extit{WSpTFI}，以及一种迭代加权最小二乘（IRLS）-块连续上界最小化（BSUM）方案用于 extit{WSpTFII}。理论分析建立了加权分解关系，并为 extit{WSpTFI} 提供了条件限制的 Karush-Kuhn-Tucker (KKT) 特征。对于 extit{WSpTFII}，实际的阻尼二次块更新产生了固定-$ extdelta$ 平滑因子目标的定量充分减少机制。这意味着渐近正则性，并且固定维度尾部的每个聚点都是平稳的。在合成张量补全、彩色图像恢复、高光谱修复和印刷电路板缺陷检测的实验中，展示了在各种降级条件下具有竞争力的重建质量和稳健性。

View on arXiv Download PDF AI Translation

cs.CV / 92 / 2609.14312

DTI-Guided Volumetric Spherical Harmonics Regression for Single-to-Multi-Shell dMRI Synthesis

基于DTI的体积球谐回归用于单壳到多壳扩散MRI合成

Li, Binghua, Andica, Christina, Liang, Tong, Chang, Ziqing, Li, Chao, Uchida, Wataru, Takabayashi, Kaito, Zhao, Qibin, Tanaka, Toshihisa, Sun, Zhe, Aoki, Shigeki

Abstract

Multi-shell diffusion MRI (dMRI) unlocks more expressive microstructural modeling than single-shell scans, yet its longer acquisition time hinders deployment in large-scale cohorts and time-constrained clinical settings. Synthesizing an unobserved shell from a single-shell input is fundamentally ill-posed and further complicated by protocol mismatch, where source and target gradient direction sets may not align. We propose DTI-SHNet, a single-to-multi-shell synthesis framework that operates in the real symmetric spherical harmonics (SH) coefficient domain and performs spatially aware volumetric regression. Given a source shell, we estimate diffusion tensor imaging (DTI) and use direction-agnostic parametric maps along with a brain mask as conditioning priors to guide a 3D U-Net regressor from source-shell to target-shell SH coefficients. To couple coefficient accuracy with signal fidelity, we introduce a signal consistency regularization that reconstructs signals on randomly sampled canonical directions from predicted coefficients and enforces agreement in the signal domain. Experiments on UK Biobank and Cam-CAN data for b=1000 to b=2000 dMRI synthesis show that DTI-SHNet achieves competitive visual quality compared to advanced methods, while better preserving downstream diffusion measures. Our code is available at https://github.com/xiaovhua/dti-shnet.

Chinese Translation

多壳扩散MRI（dMRI）相比单壳扫描能够解锁更具表现力的微观结构建模，但其较长的采集时间限制了在大规模队列和时间受限的临床环境中的应用。从单壳输入合成未观察到的壳在本质上是一个病态问题，并且由于协议不匹配（源和目标梯度方向集可能不一致）而进一步复杂化。我们提出了DTI-SHNet，一个单壳到多壳合成框架，该框架在真实对称球谐（SH）系数域中操作，并执行空间感知的体积回归。给定一个源壳，我们估计扩散张量成像（DTI），并使用与方向无关的参数图以及脑掩膜作为条件先验，以指导3D U-Net回归器从源壳到目标壳的SH系数。为了将系数的准确性与信号的保真度相结合，我们引入了一种信号一致性正则化，它从预测的系数重建在随机采样的标准方向上的信号，并在信号域中强制一致性。在UK Biobank和Cam-CAN数据集上进行的b=1000到b=2000的dMRI合成实验表明，DTI-SHNet在视觉质量上与先进方法相比具有竞争力，同时更好地保留了下游扩散测量。我们的代码可在https://github.com/xiaovhua/dti-shnet获取。

View on arXiv Download PDF AI Translation

cs.CV / 93 / 2609.14313

S3-Tracker: Self-Supervised Surgical Tissue Tracking With Contrastive Random Walks

S3-Tracker：基于对比随机游走的自监督外科组织跟踪

Zhang, Jiaming, Wu, Zijian, Armand, Mehran, Salcudean, Septimiu

Abstract

Robust point tracking in endoscopic videos is essential for computer-assisted intervention and autonomous robotic surgery, enabling continuous registration between intraoperative video and preoperative imaging despite soft tissue deformation. However, supervised tracking methods depend on large annotated datasets, while surgical conditions make reliable trajectory annotation challenging. We propose a self-supervised Track-Any-Point approach that learns from unlabeled surgical videos by establishing global pixel correspondences and inferring point trajectories through contrastive random walks. Trained without annotations, our method achieves performance comparable to existing semi-supervised approaches while implicitly handling tissue deformation. These findings demonstrate the feasibility of self-supervised point tracking in surgical environments and its potential to reduce reliance on annotated data.

Chinese Translation

在内窥镜视频中进行稳健的点跟踪对于计算机辅助干预和自主机器人手术至关重要，它能够在软组织变形的情况下实现术中视频与术前影像之间的连续配准。然而，监督跟踪方法依赖于大量标注数据集，而外科手术条件使得可靠的轨迹标注变得具有挑战性。我们提出了一种自监督的Track-Any-Point方法，通过建立全局像素对应关系并通过对比随机游走推断点轨迹，从未标注的外科视频中学习。我们的算法在没有标注的情况下进行训练，其性能与现有的半监督方法相当，同时隐式处理了组织变形。这些发现展示了在外科环境中自监督点跟踪的可行性及其减少对标注数据依赖的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 94 / 2609.14316

Learning Continuous Source Responses For Generalizable AI-Generated Image Detection

学习连续源响应以实现可泛化的人工智能生成图像检测

Cui, Manni, Liu, Ruiqi, Yu, Zijian, Tan, Hao, Wei, Zibo, Wang, Zian, Qin, Ziheng, Zhu, Huijia, Wang, Weiqiang, Lan, Jun, Wu, Shu

Abstract

Advances in image generation have made synthetic images increasingly difficult to distinguish from real photographs, raising concerns about the trustworthiness of visual media. Existing AI-generated image detectors often perform well on in-domain data, but their robustness and cross-generator generalization remain limited. These limitations are commonly attributed to overfitting to shortcut cues. Although many methods seek to suppress shortcut learning, most retain binary classification as the training task without reconsidering how the task itself shapes the learned representations. We introduce CuRe, a framework for learning Continuous Source Responses that revisits authenticity detection from the perspective of the training task. CuRe reformulates backbone adaptation as regression of real-generated mixing ratios, providing finer supervision that encourages the model to capture authenticity-related variation beyond binary endpoint separation. We further select a compact source-response subspace to suppress nuisance variation and limit the final classifier's access to potential shortcut cues. Across ten public benchmarks, CuRe achieves an average balanced accuracy of 89.7%, exceeding the second-best method by 5.2 percentage points. Further experiments demonstrate consistent generalization gains across visual backbones and strong robustness to common image degradations. Code is available at https://github.com/manic-cui/CuRe

Chinese Translation

图像生成的进展使得合成图像越来越难以与真实照片区分，这引发了对视觉媒体可信度的担忧。现有的人工智能生成图像检测器在领域内数据上通常表现良好，但其鲁棒性和跨生成器的泛化能力仍然有限。这些局限性通常归因于对捷径线索的过拟合。尽管许多方法试图抑制捷径学习，但大多数仍将二元分类作为训练任务，而没有重新考虑任务本身如何塑造学习到的表征。我们提出了CuRe，一个学习连续源响应的框架，从训练任务的角度重新审视真实性检测。CuRe将主干适应重新表述为真实与生成混合比的回归，提供更细致的监督，鼓励模型捕捉超越二元端点分离的真实性相关变化。我们进一步选择一个紧凑的源响应子空间，以抑制干扰变化并限制最终分类器对潜在捷径线索的访问。在十个公共基准测试中，CuRe实现了89.7%的平均平衡准确率，超过第二名方法5.2个百分点。进一步的实验表明，在视觉主干上具有一致的泛化增益，并对常见图像退化表现出强大的鲁棒性。代码可在 https://github.com/manic-cui/CuRe 获取。

View on arXiv Download PDF AI Translation

cs.CV / 95 / 2609.14347

Mobile CT Services for Rural, Regional, and Remote Areas: Current Practice and Future Integration with Telehealth and Regulatory-Authorised AI

农村、区域和偏远地区的移动CT服务：当前实践与未来与远程医疗及监管授权人工智能的整合

Lu, Zhicheng, Islam, Md Zahid, Huda, M Mamun, Sweeney, Kristie, Chau, Shayne, Mulcock, Oliver, Hemopo, Corey, Keniry, Catherine, Moni, Mohammad Ali

Abstract

Computed tomography (CT) plays an essential role in clinical workflow to improve patient outcomes. However, access to CT imaging and specialist interpretation remains limited, particularly in rural, regional, remote (RRR), and other resource-limited settings. Recent advances in mobile CT, telehealth, and artificial intelligence (AI) provide opportunities to extend advanced imaging services to populations in RRR settings. This review examines: 1) mobile CT systems deployed in trucks, trailers, ambulances, and other mobile platforms; 2) telehealth technologies supporting CT-based healthcare; and 3) AI for CT that has received regulatory authorisation or is currently deployed in clinical practice. Applications are evaluated across four clinical functions: screening and diagnosis, patient monitoring, risk prediction, and intervention or therapeutic decision support. The review covers neurological, thoracic, cardiovascular, abdominal, oncological, musculoskeletal, and interventional imaging, with particular attention to stroke, cancer, and other image-guided treatment. Other factors such as regulatory status, deployment status, and estimated technology readiness (TRL) level are compared. Current evidence indicates that mobile CT, telehealth, and AI for conventional CT are individually relatively mature, but fully integration of these technologies remains less widely deployed and validated in the clinical settings. Key barriers include regulatory variation, domain shift, connectivity requirements, cost, workflow integration, cybersecurity, and limited evidence of patient-level benefit. Future research should prioritise prospective, multicentre evaluation of integrated CT systems in real-world and underserved clinical settings.

Chinese Translation

计算机断层扫描（CT）在临床工作流程中发挥着至关重要的作用，以改善患者的治疗效果。然而，CT成像和专业解读的获取仍然有限，特别是在农村、区域、偏远（RRR）及其他资源有限的环境中。最近在移动CT、远程医疗和人工智能（AI）方面的进展为将先进成像服务扩展到RRR环境中的人群提供了机会。本综述考察了：1）部署在卡车、拖车、救护车和其他移动平台上的移动CT系统；2）支持基于CT的医疗保健的远程医疗技术；以及3）获得监管授权或目前在临床实践中应用的CT人工智能。应用评估涵盖四个临床功能：筛查与诊断、患者监测、风险预测和干预或治疗决策支持。综述涉及神经、胸部、心血管、腹部、肿瘤、肌肉骨骼和介入成像，特别关注中风、癌症和其他影像引导治疗。还比较了监管状态、部署状态和估计的技术准备度（TRL）水平等其他因素。目前的证据表明，移动CT、远程医疗和传统CT的人工智能在个体上相对成熟，但这些技术的全面整合在临床环境中仍然不够广泛部署和验证。主要障碍包括监管差异、领域转变、连接要求、成本、工作流程整合、网络安全以及患者层面获益的证据有限。未来的研究应优先考虑在现实世界和服务不足的临床环境中对集成CT系统进行前瞻性、多中心的评估。

View on arXiv Download PDF AI Translation

cs.CV / 96 / 2609.14350

Two-Stage Mixture-of-LoRA for Multi-Task Medical Vision-Language Learning

多任务医学视觉-语言学习的两阶段混合LoRA

Chen, Zhanghao, Li, Yuanyuan, Lu, Zhenyu, Gao, Shuo, Zhou, Guangquan, Zhang, Yikun

Abstract

Medical vision-language models (VLMs) allow a single model to perform clinical image analysis tasks ranging from diagnosis classification to report generation. However, joint adaptation is challenged by heterogeneous output formats, conflicting task gradients, and imbalanced training data. Hence, we present \textbf{Two-Stage Mixture-of-LoRA}, a framework built on MedGemma-1.5-4B. The framework uses a shared-specific Mixture-of-LoRA architecture comprising one shared LoRA and six task-specific expert LoRAs, together with a two-stage training procedure. In Stage 1, we jointly train the shared LoRA and all task-specific expert LoRAs on all tasks. In Stage 2, we first freeze the backbone, the shared LoRA, and all non-target experts, and refine one task expert at a time. Classification and regression then receive an additional modality-balanced continuation, in which smaller modality groups are repeated to match the largest group. In the FLARE 2026 Task 3 test sets, the proposed method achieves 0.85 balanced accuracy for classification, 0.48 micro-F1 for multi-label classification, 0.79 detection F1, and 17.39 regression MAE. Code is available at https://github.com/YuanYL03/MICCAI-FLARE-2026-Challenge-Task3-2D.

Chinese Translation

医学视觉-语言模型（VLMs）使得单一模型能够执行从诊断分类到报告生成的临床图像分析任务。然而，联合适应面临着异构输出格式、相互冲突的任务梯度和不平衡的训练数据等挑战。因此，我们提出了 extbf{两阶段混合LoRA}，这是一个基于MedGemma-1.5-4B构建的框架。该框架采用共享-特定的混合LoRA架构，包括一个共享LoRA和六个任务特定的专家LoRA，以及一个两阶段的训练过程。在第一阶段，我们在所有任务上共同训练共享LoRA和所有任务特定的专家LoRA。在第二阶段，我们首先冻结主干网络、共享LoRA和所有非目标专家，然后逐个精炼一个任务专家。分类和回归任务随后接收额外的模态平衡延续，其中较小的模态组被重复以匹配最大的组。在FLARE 2026任务3测试集中，所提出的方法在分类任务上达到了0.85的平衡准确率，在多标签分类上达到了0.48的微F1，在检测上达到了0.79的F1，在回归上达到了17.39的平均绝对误差（MAE）。代码可在https://github.com/YuanYL03/MICCAI-FLARE-2026-Challenge-Task3-2D获取。

View on arXiv Download PDF AI Translation

cs.CV / 97 / 2609.14352

Beyond Natural Images: Rethinking AI-Generated Image Detection in Documents

超越自然图像：重新思考文档中AI生成图像的检测

Fu, Zhangjie, Yan, Jiazhen, Chen, Yuanwen, Yu, Xinquan, Li, Yanzhe, Jiang, Hui, Gao, Lei, Bao, Chenfu

Abstract

AI-generated image detection has attracted increasing attention, but existing evaluations mainly focus on natural images, leaving AI-generated document images largely underexplored. This omission is concerning because documents often appear in sensitive real-world scenarios, such as invoices, expense reports, certificates, and medical records. In this paper, we first construct a controlled diagnostic benchmark, AIGDoc-Pilot, and reveal that existing detectors suffer substantial performance degradation on AI-generated document images, with the mean AUC dropping by more than 7%. Based on this, we further reveal two document-specific properties behind this gap: generation artifacts exhibit strong spatial inconsistency across local regions, and text density significantly affects real-synthetic separability, where text-dense regions offer stronger discriminative evidence. Motivated by these findings, we construct AIGDoc, a larger document-centric dataset containing diverse real-world documents and AI-generated counterparts produced by multiple advanced generation and editing models. Extensive experiments on AIGDoc demonstrate that existing detectors still struggle to reliably identify AI-generated documents, while document-based training partially narrows the gap. Together, these results offer valuable insights for developing dependable and generalizable detectors in document-centric scenarios. The code and datasets will be made publicly available upon acceptance of the paper.

Chinese Translation

AI生成图像的检测引起了越来越多的关注，但现有评估主要集中在自然图像上，导致AI生成的文档图像在很大程度上未被探索。这一遗漏令人担忧，因为文档常常出现在敏感的现实场景中，例如发票、费用报告、证书和医疗记录。本文首先构建了一个受控的诊断基准，AIGDoc-Pilot，并揭示现有检测器在AI生成的文档图像上表现出显著的性能下降，平均AUC下降超过7%。基于此，我们进一步揭示了这一差距背后的两个文档特有属性：生成伪影在局部区域之间表现出强烈的空间不一致性，以及文本密度显著影响真实与合成的可分离性，其中文本密集区域提供了更强的区分证据。受到这些发现的启发，我们构建了AIGDoc，这是一个更大规模的文档中心数据集，包含多种真实世界文档及由多种先进生成和编辑模型生成的对应AI文档。在AIGDoc上的大量实验表明，现有检测器仍然难以可靠地识别AI生成的文档，而基于文档的训练在一定程度上缩小了这一差距。综上所述，这些结果为在文档中心场景中开发可靠且可推广的检测器提供了宝贵的见解。代码和数据集将在论文接受后公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 98 / 2609.14377

Rethinking Camouflage Image Generation towards a Training-Free Paradigm

重新思考伪装图像生成的无训练范式

Yang, Haodong, Huang, Zhongling, Cheng, Gong

Abstract

Camouflage image generation (CIG) aims to synthesize realistic camouflaged images by blending foreground objects into concealment-compatible background contexts. Achieving this objective requires jointly satisfying three coupled requirements: foreground preservation to retain target integrity, semantic compatibility to select plausible concealment contexts, and appearance assimilation to reduce visual discrepancies. Recent approaches predominantly rely on task-specific training on camouflage datasets to address these requirements, incurring substantial computational cost and limiting generalization beyond the training domain. To address these limitations, we formulate training-free CIG as a concealment-oriented paradigm that preserves the target while reducing its perceptual separability from the synthesized surroundings, rather than maintaining its visual prominence, without parameter updates. We instantiate this paradigm with FreeCam based on a frozen inpainting diffusion framework to preserve the foreground. Within this framework, a Contextual Reasoning Module exploits frozen multimodal priors to infer an environment favorable to concealment, thereby promoting semantic compatibility, while an Intrinsic Appearance Module extracts low-level color and texture cues from the foreground to guide background synthesis toward appearance assimilation. Extensive experiments demonstrate that FreeCam achieves state-of-the-art generation quality and camouflage effectiveness without task-specific training, while its generated images provide synthetic supervision for camouflaged object detection and reduce target detectability under general object detectors.

Chinese Translation

伪装图像生成（CIG）旨在通过将前景物体融入适合隐蔽的背景环境中，合成逼真的伪装图像。实现这一目标需要共同满足三个相互关联的要求：前景保留以保持目标完整性，语义兼容性以选择合理的隐蔽背景，以及外观同化以减少视觉差异。近期的方法主要依赖于在伪装数据集上进行特定任务的训练来解决这些要求，这导致了巨大的计算成本，并限制了在训练领域之外的泛化能力。为了解决这些局限性，我们将无训练的CIG构建为一种以隐蔽为导向的范式，该范式在不更新参数的情况下，保持目标的同时减少其与合成环境的感知可分离性，而不是保持其视觉显著性。我们通过基于冻结的修复扩散框架的FreeCam实例化这一范式，以保留前景。在该框架内，情境推理模块利用冻结的多模态先验推断有利于隐蔽的环境，从而促进语义兼容性，而内在外观模块则从前景中提取低级颜色和纹理线索，以指导背景合成朝向外观同化。大量实验表明，FreeCam在没有特定任务训练的情况下，实现了最先进的生成质量和伪装效果，同时其生成的图像为伪装物体检测提供了合成监督，并减少了在通用物体检测器下的目标可检测性。

View on arXiv Download PDF AI Translation

cs.CV / 99 / 2609.14383

Contour-Guided Spectral Routing for Robust Real-Time Pedestrian Detection

基于轮廓引导的光谱路由用于鲁棒的实时行人检测

Williams, Sam, Xiang, Yuan

Abstract

Real-time pedestrian detection in driving scenes is constrained by three coupled failure modes: tiny targets lose discriminative evidence, occlusion weakens geometric support, and weather or illumination changes distort appearance statistics. We formulate the detector through a unified \emph{contour-guided spectral routing} view rather than treating frequency processing, attention, and boundary reasoning as independent add-ons. The detector routes information in a prescribed order: spatial evidence is first augmented with global spectral context, deep representations then exchange spatial and spectral cues, and cross-scale fusion is finally conditioned on boundary--semantic disagreement. This ordering yields a compact representation pipeline in which low-frequency context stabilizes global structure while high-frequency evidence protects small-object contours. We further retain a wavelet-subband training transformation that perturbs low- and high-frequency coefficients independently, targeting appearance shifts caused by fog, rain, snow, and low illumination. The formulation exposes a single routing variable at each stage and distinguishes reusable signal transforms from the task-specific policy that decides where each signal is injected. On CityPersons, the proposed detector obtains 70.4 AP$_{50}$ and 44.2 AP$_{50:95}$, compared with 68.1 and 42.2 for RT-DETR, while the full wavelet-augmented configuration reaches 71.1 and 44.6.

Chinese Translation

在驾驶场景中，实时行人检测受到三种耦合失效模式的限制：微小目标失去辨别证据，遮挡削弱几何支持，天气或光照变化扭曲外观统计。我们通过统一的 extit{轮廓引导光谱路由}视角来构建检测器，而不是将频率处理、注意力和边界推理视为独立的附加模块。检测器按照规定的顺序路由信息：首先用全局光谱上下文增强空间证据，然后深度表示交换空间和光谱线索，最后跨尺度融合依赖于边界-语义的不一致性。这种顺序产生了一个紧凑的表示管道，其中低频上下文稳定全局结构，而高频证据保护小物体的轮廓。我们进一步保留了一种小波子带训练变换，该变换独立扰动低频和高频系数，针对雾、雨、雪和低光照造成的外观变化。该公式在每个阶段暴露出一个单一的路由变量，并区分可重用的信号变换与决定每个信号注入位置的任务特定策略。在CityPersons数据集上，所提出的检测器获得了70.4 AP$_{50}$和44.2 AP$_{50:95}$，相比之下，RT-DETR分别为68.1和42.2，而完整的小波增强配置达到了71.1和44.6。

View on arXiv Download PDF AI Translation

cs.CV / 100 / 2609.14391

Newton Deep Unfolding for Compressed Sensing

压缩感知的牛顿深度展开

He, Changhua, Xiu, Xianchao

Abstract

Compressed sensing (CS) reconstructs images from highly limited measurements, but existing deep unfolding methods are typically driven by first-order optimization and weakly exploit the optimization states generated during reconstruction. To address these limitations, we propose a Newton deep unfolding network (NDU-Net), which, to the best of our knowledge, is the first deep unfolding framework that leverages second-order optimization for CS reconstruction. Specifically, NDU-Net introduces a Newton update (NU) module to estimate Newton-type update directions and generate optimization states that characterize the current reconstruction process. Furthermore, a Newton-guided multi-scale prior (MP) module is designed to incorporate these optimization states into multi-scale feature restoration, thereby enabling the learned prior to adapt to the current reconstruction stage. Experimental results under different CS ratios confirm that our proposed NDU-Net achieves promising reconstruction performance and exhibits enhanced robustness. Our code is available at https://github.com/xianchaoxiu/DNU-Net.

Chinese Translation

压缩感知（CS）从高度有限的测量中重建图像，但现有的深度展开方法通常依赖于一阶优化，并且对重建过程中产生的优化状态的利用较弱。为了解决这些局限性，我们提出了一种牛顿深度展开网络（NDU-Net），据我们所知，这是第一个利用二阶优化进行压缩感知重建的深度展开框架。具体而言，NDU-Net引入了一个牛顿更新（NU）模块，以估计牛顿类型的更新方向，并生成表征当前重建过程的优化状态。此外，设计了一个牛顿引导的多尺度先验（MP）模块，将这些优化状态融入多尺度特征恢复中，从而使学习到的先验能够适应当前的重建阶段。在不同的压缩感知比下的实验结果表明，我们提出的NDU-Net实现了令人满意的重建性能，并展现了增强的鲁棒性。我们的代码可在 https://github.com/xianchaoxiu/DNU-Net 获取。

View on arXiv Download PDF AI Translation

cs.CV / 101 / 2609.14419

A Generative AI Integrated Multimodal Framework for Low-Latency Multi-Camera Person Re-Identification

集成生成式人工智能的多模态框架用于低延迟多摄像头人员重识别

Fernando, Leon, Dombawala, C, Hettigoda, P., Warnasooriya, Vanodhya G., Neranjana, Ishara, Nawaratne, Rashmika

Abstract

Person re-identification (ReID) is essential for multi-camera surveillance and tracking, yet remains difficult due to viewpoint and illumination changes, occlusion, background clutter, and low resolution imagery. We propose a generative AI integrated multimodal ReID framework designed explicitly for robustness under missing cues and low latency deployment. The key idea is a cost aware early-exit cascade that prioritizes inexpensive, high confidence evidence and only triggers expensive modalities for ambiguous cases. Our system integrates (i) global visual embeddings from segmented person regions, (ii) automatically generated fine grained semantic attribute descriptions generated by vision-language models (VLMs), and (iii) optional facial embeddings when face observations are reliable. To optimize the balance between accuracy and latency, we use a cost aware early-exit cascade instead of fusing all modalities. Specifically, we first inspect the top-k retrieval results to determine whether the query is unambiguous. If the best match is clearly separated from the remaining candidates, we stop early and return the result to minimize latency; in ambiguous cases, we keep multiple hypotheses and invoke additional modalities (face/semantic) with adaptive reliability weighting to refine the decision. We report person re-identification performance using mAP and Rank-1 accuracy on the Market-1501 and DukeMTMC-reID benchmarks. The proposed adaptive early-exit cascade resolves 60.7% of DukeMTMC-reID queries and 68.4% of Market-1501 queries without invoking semantic reasoning, reducing computational overhead while maintaining competitive retrieval performance.

Chinese Translation

人员重识别（ReID）对于多摄像头监控和跟踪至关重要，但由于视角和光照变化、遮挡、背景杂乱以及低分辨率图像等因素，仍然面临挑战。我们提出了一种集成生成式人工智能的多模态ReID框架，专门设计用于在缺失线索和低延迟部署下的鲁棒性。其核心思想是一个成本感知的早期退出级联，优先考虑低成本、高置信度的证据，仅在模糊情况下触发高成本的模态。我们的系统集成了（i）来自分割人区域的全局视觉嵌入，（ii）由视觉-语言模型（VLMs）自动生成的细粒度语义属性描述，以及（iii）在面部观察可靠时的可选面部嵌入。为了优化准确性与延迟之间的平衡，我们使用成本感知的早期退出级联，而不是融合所有模态。具体而言，我们首先检查前k个检索结果，以确定查询是否明确。如果最佳匹配与其余候选者明显分离，我们将提前停止并返回结果，以最小化延迟；在模糊情况下，我们保留多个假设，并根据自适应可靠性加权调用额外的模态（面部/语义）以细化决策。我们在Market-1501和DukeMTMC-reID基准上报告了人员重识别性能，使用mAP和Rank-1准确率。所提出的自适应早期退出级联在不调用语义推理的情况下解决了60.7%的DukeMTMC-reID查询和68.4%的Market-1501查询，减少了计算开销，同时保持了竞争性的检索性能。

View on arXiv Download PDF AI Translation

cs.CV / 102 / 2609.14437

Lightweight Generalized DeepFake Face Detection with WAVIE: Wavelet Augmented Vision Intermediate Embeddings

轻量级广义深伪造人脸检测：WAVIE（小波增强视觉中间嵌入）

Pulkit, Arya, Ruhela, Aditya, Kapoor, Akarshan, Bhavsar, Arnav

Abstract

Deepfake detection systems often exhibit significant performance degradation when deployed on unseen manipulation methods, limiting their reliability in real-world multimedia environments. This lack of generalization poses critical challenges for misinformation mitigation, digital forensics, and human-centric AI systems. Existing detectors perform well on the forgery methods they are trained on, but their accuracy drops sharply on unseen pipelines. To bridge this generalization gap, we propose WAVIE (Wavelet Augmented Vision Intermediate Embeddings), an end-to-end architecture that combines complementary spatial and frequency cues on top of a frozen CLIP backbone. WAVIE projects intermediate transformer embeddings through a lightweight learnable module, applies a three-level Daubechies-6 (db6) discrete wavelet transform (DWT), refines the low-frequency branch while preserving the high-frequency branch, reconstructs the feature via inverse DWT, and performs classification. Trained only on FaceForensics++, WAVIE achieves AUROC = 0.852 on Celeb-DF-v1, 0.852 on Celeb-DF-v2 and 0.831 on WildDeepFake (WDF) at the frame level, outperforming several state-of-the-art generalization baselines. Extensive ablation studies confirm the importance of both the wavelet module and the intermediate-feature aggregation for cross-dataset performance, highlighting the necessity of jointly leveraging spatial and frequency domains. These results position WAVIE as a strong baseline for deepfake detection in the wild.

Chinese Translation

深伪造检测系统在未见过的操纵方法上往往表现出显著的性能下降，这限制了它们在现实世界多媒体环境中的可靠性。这种缺乏泛化能力对信息误导的缓解、数字取证和以人为中心的人工智能系统构成了重大挑战。现有的检测器在其训练的伪造方法上表现良好，但在未见过的处理流程上准确率急剧下降。为了弥补这一泛化差距，我们提出了WAVIE（小波增强视觉中间嵌入），这是一种端到端架构，结合了冻结的CLIP骨干网络之上的互补空间和频率线索。WAVIE通过一个轻量级可学习模块投影中间变换器嵌入，应用三层Daubechies-6（db6）离散小波变换（DWT），在保留高频分支的同时精炼低频分支，通过逆DWT重建特征，并进行分类。仅在FaceForensics++上训练的WAVIE在Celeb-DF-v1上达到AUROC = 0.852，在Celeb-DF-v2上达到0.852，在WildDeepFake（WDF）上达到0.831（在帧级别），超越了多个最先进的泛化基线。广泛的消融研究确认了小波模块和中间特征聚合在跨数据集性能中的重要性，突显了联合利用空间和频率域的必要性。这些结果将WAVIE定位为野外深伪造检测的强基线。

View on arXiv Download PDF AI Translation

cs.CV / 103 / 2609.14439

Context-Aware Mutual Learning for Blind Image Inpainting and Beyond

基于上下文感知的盲图像修复互学习及其扩展

Zhao, Haoru, Wang, Yufeng, Gu, Zhaorui, Zheng, Bing, Zheng, Haiyong

Abstract

Blind image inpainting, aiming to recover contaminated images in the case of unknown masks, is a challenging task. Motivated by the perspective of human vision and knowledge, blind image inpainting can be decomposed into two stages: mask estimation and image inpainting based on the estimated mask. The two-stage idea exhibits evident advantages in enhancing inpainting quality and augmenting the generalization capability of unknown real-world contamination by explicitly employing the estimated mask for image inpainting compared to one-stage scheme. This two-stage idea has also been intuitively implemented. However, existing two-stage methods excessively emphasize the unilateral relationship of mask estimation to image inpainting, and may overlook the mutual relations between them. Specifically, mask estimation can provide more contextual semantics for image inpainting to strengthen the understanding of semantics, and image inpainting can offer more contextual details (e.g., textures and edges) for mask estimation to improve the learning of details. In this work, we propose a novel Context-Aware Mutual Learning (CAML) framework for blind image inpainting that joints mask estimation and image inpainting to mutually exploit contextual information. In the CAML framework, we design the Inpainting-Guided Context-Mutual (IGCM) learner to acquire the complementary contextual details from image inpainting for assisting mask estimation, and the Estimation-Guided Context-Mutual (EGCM) learner to strengthen the understanding of contextual semantics from mask estimation for assisting image inpainting. Ablation studies validate the efficacy of our CAML. Extensive experiments show that our CAML achieves state-of-the-art performance on both blind image inpainting and additional vision tasks, i.e., snow removal, shadow removal, and watermark removal, indicating its superiority.

Chinese Translation

盲图像修复旨在在未知掩膜的情况下恢复受污染的图像，这是一项具有挑战性的任务。受到人类视觉和知识的启发，盲图像修复可以分解为两个阶段：掩膜估计和基于估计掩膜的图像修复。与单阶段方案相比，这种两阶段的思路在提高修复质量和增强对未知真实世界污染的泛化能力方面展现出明显优势，因为它明确地利用估计的掩膜进行图像修复。然而，现有的两阶段方法过于强调掩膜估计与图像修复之间的单向关系，可能忽视了它们之间的相互关系。具体而言，掩膜估计可以为图像修复提供更多的上下文语义，以增强对语义的理解，而图像修复可以为掩膜估计提供更多的上下文细节（例如，纹理和边缘），以改善细节学习。在本研究中，我们提出了一种新颖的上下文感知互学习框架（Context-Aware Mutual Learning, CAML），将掩膜估计和图像修复结合起来，以相互利用上下文信息。在CAML框架中，我们设计了修复引导的上下文互学习者（Inpainting-Guided Context-Mutual, IGCM），以获取来自图像修复的互补上下文细节来辅助掩膜估计；同时设计了估计引导的上下文互学习者（Estimation-Guided Context-Mutual, EGCM），以增强从掩膜估计中对上下文语义的理解来辅助图像修复。消融实验验证了我们CAML的有效性。大量实验表明，我们的CAML在盲图像修复和其他视觉任务（即雪去除、阴影去除和水印去除）上均达到了最先进的性能，显示了其优越性。

View on arXiv Download PDF AI Translation

cs.CV / 104 / 2609.14441

From Visual Attribution to Clinical Reasoning: Explainable Parkinson's Disease Screening from Hand-Drawn Patterns

从视觉归因到临床推理：基于手绘图案的可解释帕金森病筛查

Dey, Aritra, Nareti, Utsav Kumar, Adak, Chandranath, Chattopadhyay, Soumi, Sasmal, Krishna Gopal, Anwar, Saeed

Abstract

Parkinson's disease (PD) manifests early neuromotor impairments that become observable in controlled hand-drawn patterns such as spirals and meanders, where tremor-induced oscillations, stroke irregularity, and curvature instability reflect underlying motor degradation. In this work, we present an explainable framework for PD screening from offline hand-drawn patterns that integrates discriminative visual modeling with clinically grounded reasoning. The predictive model captures distributed structural distortions and fine-grained texture variations. It is evaluated under subject-disjoint protocols to ensure reliable generalization. To move beyond black-box classification, we introduce a multi-stage explainability pipeline that combines visual attribution with structured symptom abstraction. Salient regions are identified using attention- and gradient-based localization, followed by extraction of clinically meaningful motor descriptors quantifying contour roughness, curvature irregularity, stroke variability, and tremor-frequency energy. These descriptors are subsequently translated into coherent clinical rationales through a language-based reasoning module, linking model evidence to established PD symptomatology. By bridging visual attribution and clinical interpretation, the proposed framework advances interpretable document intelligence for neurological screening using hand-drawn patterns. Experimental results on publicly available Parkinson's disease handwriting datasets demonstrate competitive predictive performance and clinically consistent explanations.

Chinese Translation

帕金森病（PD）表现为早期的神经运动障碍，这些障碍在受控的手绘图案中变得可观察，例如螺旋和曲线，其中颤抖引起的振荡、笔画不规则性和曲率不稳定性反映了潜在的运动退化。在本研究中，我们提出了一个可解释的帕金森病筛查框架，该框架基于离线手绘图案，将区分性视觉建模与临床基础推理相结合。预测模型捕捉了分布式结构扭曲和细微的纹理变化。它在受试者不重叠的协议下进行评估，以确保可靠的泛化能力。为了超越黑箱分类，我们引入了一个多阶段可解释性管道，将视觉归因与结构化症状抽象相结合。通过基于注意力和梯度的定位方法识别显著区域，随后提取临床意义明确的运动描述符，以量化轮廓粗糙度、曲率不规则性、笔画变异性和颤抖频率能量。这些描述符随后通过基于语言的推理模块转化为连贯的临床推理，将模型证据与已建立的帕金森病症状学联系起来。通过桥接视觉归因与临床解释，所提出的框架推动了基于手绘图案的神经筛查的可解释文档智能。对公开可用的帕金森病手写数据集的实验结果显示出具有竞争力的预测性能和临床一致的解释。

View on arXiv Download PDF AI Translation

cs.CV / 105 / 2609.14462

AlayaVista: Streaming World Modeling from Panoramic States to Perspective Video

AlayaVista：从全景状态到视角视频的流式世界建模

Tan, Jiaming, Zhai, Mingliang, Li, Zhen, Wu, Yuwei, Li, Chuanhao, Zhang, Kaipeng

Abstract

Interactive video world models must maintain broad scene context under camera motion while producing high-fidelity observations with low latency. Existing approaches face a representation trade-off: perspective models operate on local views and must preserve off-screen content over long rollouts, whereas broader spatial coverage is typically obtained by synthesizing full-sphere videos or constructing explicit 3D representations. Motivated by the complementary roles of global context and selective local acuity in visual perception, we present AlayaVista, a camera-controllable streaming video world model that decouples panoramic world evolution from perspective observation synthesis. Given a single perspective image, AlayaVista constructs a 360-degree scene prior using a pretrained panorama expansion model and then evolves the scene as a camera-conditioned panoramic latent state. A latent viewport renderer maps this state to the requested perspective video latents, while a perspective refiner restores details, suppresses artifacts, and performs super-resolution. To support efficient streaming, we adapt the panoramic generator to chunk-autoregressive generation and distill both panoramic generation and perspective refinement into few-step processes. To provide the supervision required by this design, we construct MUGEN, a large-scale real-world panoramic video dataset containing 1,318 hours of videos at resolutions of at least 4K, together with rich semantic and geometric annotations.

Chinese Translation

交互式视频世界模型必须在相机运动下保持广泛的场景上下文，同时以低延迟生成高保真观察。现有方法面临表示权衡：视角模型在局部视图上操作，必须在长时间滚动中保留屏幕外内容，而更广泛的空间覆盖通常通过合成全球视频或构建显式的3D表示来获得。受到全球上下文和选择性局部清晰度在视觉感知中互补作用的启发，我们提出了AlayaVista，这是一种可控相机的流式视频世界模型，它将全景世界演变与视角观察合成解耦。给定一幅单一的视角图像，AlayaVista使用预训练的全景扩展模型构建360度场景先验，然后将场景作为相机条件的全景潜在状态进行演变。潜在视口渲染器将该状态映射到请求的视角视频潜在，而视角细化器则恢复细节、抑制伪影并执行超分辨率。为了支持高效流式传输，我们将全景生成器调整为块自回归生成，并将全景生成和视角细化提炼为少步过程。为了提供该设计所需的监督，我们构建了MUGEN，这是一个大规模的真实全景视频数据集，包含1,318小时的至少4K分辨率视频，以及丰富的语义和几何注释。

View on arXiv Download PDF AI Translation

cs.CV / 106 / 2609.14466

DynEoMT: Learning Object Dynamicity from Online Segmentation Queries

DynEoMT：从在线分割查询中学习对象动态性

Galagain, Calvin, Poreba, Martyna, Goulette, François

Abstract

Video segmentation models recognize and track objects over time, but they do not indicate whether each segmented region moves independently of the observing camera. This dynamicity attribute cannot be inferred from semantics alone and is confounded by camera ego-motion. We introduce \method, an online framework that augments query-based video segmentation with region-level dynamicity prediction. It jointly produces the original segmentation outputs and a dynamic or static state for each predicted region. At inference, DynEoMT uses only the current frame and propagated queries, without optical flow, depth, camera pose, previous RGB frames, or feature maps. Because established video segmentation benchmarks do not annotate this attribute, we also introduce a class-agnostic offline supervision pipeline using camera-compensated optical flow and confidence-aware temporal filtering. Across VIPSeg, OVIS, YouTube-VIS 2022, and VSPW, DynEoMT achieves balanced accuracies of 84.3, 68.0, 68.6, and 87.6, respectively, while largely preserving segmentation performance. These results show that segmentation-region dynamicity can be learned from propagated queries, enabling its online prediction without a dedicated motion-processing pipeline at inference. The complete code will be released as open source to enable full reproduction of the method and experiments.

Chinese Translation

视频分割模型能够识别和跟踪对象随时间的变化，但它们并未指明每个分割区域是否独立于观察相机而移动。这个动态性属性无法仅通过语义推断，并且受到相机自运动的干扰。我们提出了 extit{DynEoMT}，一个在线框架，它通过区域级动态性预测增强了基于查询的视频分割。该框架同时生成原始分割输出和每个预测区域的动态或静态状态。在推理过程中，DynEoMT仅使用当前帧和传播的查询，不依赖光流、深度、相机姿态、之前的RGB帧或特征图。由于现有的视频分割基准并未注释这一属性，我们还引入了一种与类别无关的离线监督管道，使用相机补偿的光流和基于置信度的时间滤波。在VIPSeg、OVIS、YouTube-VIS 2022和VSPW数据集上，DynEoMT分别达到了84.3、68.0、68.6和87.6的平衡准确率，同时在很大程度上保持了分割性能。这些结果表明，分割区域的动态性可以通过传播的查询进行学习，从而在推理时实现在线预测，而无需专门的运动处理管道。完整代码将作为开源发布，以便于方法和实验的全面复现。

View on arXiv Download PDF AI Translation

cs.CV / 107 / 2609.14469

PRI-Net: A Lightweight Multimodal Framework for 3D UAV Localization

PRI-Net：一种轻量级多模态框架用于3D无人机定位

Chen, Zhixuan, Lu, Jialiang, Ye, Zhong, He, Yinghui, Yu, Guanding

Abstract

Accurate 3D localization of unmanned aerial vehicles (UAVs) remains challenging for existing multimodal approaches due to sparse LiDAR geometry, modality-imbalanced fusion, and redundant feature transmission over constrained edge-to-server links. To address these limitations, we propose PRI-Net, an efficient and lightweight multimodal fusion framework for UAV localization that integrates point cloud splatting, residual attention fusion, and an information bottleneck. Specifically, a 3D point cloud splatting (3DPCS) strategy is introduced to transform sparse LiDAR observations into geometrically consistent dense depth maps. A residual attention fusion (RAF) module is then designed to alleviate modal bias by using an image branch for coarse estimation and a gated fusion branch for refinement. In addition, a multimodal information bottleneck (MIB) module compacts features by filtering task-irrelevant redundancy. Experiments show that PRI-Net achieves high localization accuracy with lightweight architectures, while reducing feature dimensionality and improving edge-to-server UAV sensing efficiency and robustness.

Chinese Translation

由于稀疏的激光雷达几何形状、模态不平衡的融合以及在受限的边缘到服务器链接上冗余特征传输，现有的多模态方法在无人机（UAV）准确的3D定位方面仍面临挑战。为了解决这些限制，我们提出了PRI-Net，这是一种高效且轻量级的多模态融合框架，用于无人机定位，集成了点云溅射、残差注意力融合和信息瓶颈。具体而言，我们引入了一种3D点云溅射（3DPCS）策略，将稀疏的激光雷达观测转换为几何一致的稠密深度图。然后设计了一个残差注意力融合（RAF）模块，通过使用图像分支进行粗略估计和一个门控融合分支进行精细化来减轻模态偏差。此外，一个多模态信息瓶颈（MIB）模块通过过滤与任务无关的冗余来压缩特征。实验表明，PRI-Net在轻量级架构下实现了高定位精度，同时降低了特征维度，提高了边缘到服务器的无人机感知效率和鲁棒性。

View on arXiv Download PDF AI Translation

cs.CV / 108 / 2609.14473

PuzzleMate: Benchmarking MLLMs for Egocentric Puzzle Assistance

PuzzleMate：针对自我中心拼图辅助的多模态大语言模型基准测试

Dasgupta, Avijit, Dasgupta, Shayon, Laskar, Zakaria, Jawahar, C. V., Alahari, Karteek

Abstract

Personal AI assistants hold the potential to evolve from digital interfaces into embodied companions capable of guiding users through complex physical activities. For these assistants to become integral to daily life, they must do more than identify objects; they must provide precise, step-by-step instructions that align with a user's real-time progress. While Multimodal Large Language Models (MLLMs) show promise in general visual understanding, their ability to deliver grounded, sequential guidance for fine-grained manipulation tasks remains largely unverified. In this paper, we choose the jigsaw puzzle as a strategic testbed for this capability. Unlike general object recognition, puzzle solving demands high-precision spatial reasoning, the ability to distinguish between minute geometric variations, and a rigorous adherence to sequential logic. We investigate this capability through PuzzleMate, a novel framework focused on jigsaw puzzle solving captured through an egocentric viewpoint. We deploy PuzzleMate in a user-in-the-loop study to evaluate how well state-of-the-art MLLMs perceive the current puzzle state and generate actionable next-step instructions. Our analysis reveals seven key bottlenecks that limit their effectiveness. Building on these insights, we propose a benchmark that enables systematic evaluation of MLLMs' reasoning capabilities for puzzle solving. Our findings reveal a substantial performance gap in current models like GPT-5.2 and Gemini-2.5-Pro; while these MLLMs are highly capable, they struggle to navigate the intricate reasoning and sequential logic essential for jigsaw puzzle assistance.

Chinese Translation

个人人工智能助手有潜力从数字界面演变为能够引导用户完成复杂物理活动的具身伴侣。为了使这些助手成为日常生活的重要组成部分，它们必须不仅仅识别物体；还必须提供与用户实时进展相一致的精确、逐步的指令。尽管多模态大语言模型（MLLMs）在一般视觉理解方面显示出潜力，但它们在提供针对细粒度操作任务的扎根、顺序指导方面的能力仍然 largely 未经验证。在本文中，我们选择拼图作为测试这一能力的战略性试验平台。与一般物体识别不同，拼图解决要求高精度的空间推理、区分微小几何变化的能力以及严格遵循顺序逻辑。我们通过PuzzleMate这一新颖框架进行研究，专注于通过自我中心视角捕捉的拼图解决。我们在用户参与的研究中部署PuzzleMate，以评估最先进的MLLMs如何感知当前拼图状态并生成可操作的下一步指令。我们的分析揭示了七个限制其有效性的关键瓶颈。在这些见解的基础上，我们提出了一个基准，能够系统评估MLLMs在拼图解决中的推理能力。我们的研究结果显示，当前模型如GPT-5.2和Gemini-2.5-Pro存在显著的性能差距；尽管这些MLLMs能力强大，但在导航拼图辅助所需的复杂推理和顺序逻辑方面仍显得力不从心。

View on arXiv Download PDF AI Translation

cs.CV / 109 / 2609.14495

MCIQA-2K: A Multi-Dimensional Dataset and No-Reference Quality Assessment Benchmark for Colorized Images

MCIQA-2K：一个多维数据集及无参考质量评估基准用于上色图像

Zhuang, Yunkai, Yan, Qihang, Zhang, Zicheng, Zhai, Guangtao

Abstract

Image colorization is an inherently ill-posed task, since a single grayscale image may correspond to multiple plausible colorized results. Consequently, conventional full-reference image quality assessment (IQA) metrics fail to accurately reflect human perceptual preferences for colorized images. In this paper, we present MCIQA-2K, a large-scale multi-dimensional benchmark specifically designed for no-reference quality assessment of colorized images. We construct a dataset containing 2,000 colorized images generated by five representative colorization models, together with human annotations across three perceptual dimensions: color smearing, semantic color misalignment, and global naturalness. Building upon the proposed benchmark, we further introduce MCIQA, a dedicated multi-branch NR-IQA framework for colorized images. Extensive experiments demonstrate that MCIQA significantly outperforms existing full-reference and no-reference IQA methods on the proposed benchmark, while also exhibiting competitive generalization capability on several widely-used IQA datasets. The dataset and code are publicly available at https://github.com/ARBEZ-ZEBRA/MCIQA.

Chinese Translation

图像上色是一项固有的不适定任务，因为单一的灰度图像可能对应多个合理的上色结果。因此，传统的全参考图像质量评估（IQA）指标无法准确反映人类对上色图像的感知偏好。本文提出了MCIQA-2K，一个专门为上色图像的无参考质量评估设计的大规模多维基准。我们构建了一个包含2000幅由五个代表性上色模型生成的上色图像的数据集，并在三个感知维度上进行了人工标注：颜色模糊、语义颜色不对齐和整体自然度。在此基准的基础上，我们进一步提出了MCIQA，一个专门针对上色图像的多分支无参考图像质量评估框架。大量实验表明，MCIQA在所提出的基准上显著优于现有的全参考和无参考IQA方法，同时在多个广泛使用的IQA数据集上也展现出竞争力的泛化能力。数据集和代码已公开发布在https://github.com/ARBEZ-ZEBRA/MCIQA。

View on arXiv Download PDF AI Translation

cs.CV / 110 / 2609.14521

CGGT: Curve-Grounded Geometry Transformer for 3D Parametric Curve Reconstruction

CGGT：基于曲线的几何变换器用于三维参数曲线重建

Gao, Zhirui, Yi, Renjiao, Ye, Yunfan, Hu, Ruizhen, Zhu, Chenyang, Chen, Wei, Xu, Kai

Abstract

Recovering editable 3D parametric curves from 2D images is a fundamental challenge in computer graphics, bridging pixel-based perception and vector-based CAD modeling. Existing NeRF- and 3DGS-based methods often rely on dense calibrated views, precomputed 2D edge maps, and costly per-scene optimization, limiting their applicability to casually captured real-world inputs. We propose CGGT, a Curve-Grounded Geometry Transformer that directly grounds 3D-consistent 2D curve instances in the image space from sparse, unposed multi-view images. CGGT combines a geometry-aware transformer encoder for multi-view feature learning with a curve-aware masked-attention decoder for cross-view instance association. In a single forward pass, it predicts camera parameters, dense depth maps, and instance-level 2D curve masks, which are then lifted into 3D and refined through a fast parametric optimization stage to recover compact, editable 3D curve primitives. To support structured curve learning, we introduce Wireframe-100K, a large-scale dataset comprising 100,000 CAD models with diverse topologies, realistic multi-view renderings, and accurate parametric curve annotations. Extensive experiments show that our framework achieves substantial improvements in both reconstruction accuracy and efficiency, particularly under challenging sparse-view settings and in separating persistent 3D structural edges from view-dependent image edges caused by silhouettes, textures, and appearance variations. Despite being trained solely on synthetic data, CGGT generalizes well to real-world images, demonstrating its potential for practical CAD-style wireframe reconstruction from unconstrained visual inputs.

Chinese Translation

从二维图像中恢复可编辑的三维参数曲线是计算机图形学中的一项基本挑战，它连接了基于像素的感知和基于矢量的计算机辅助设计（CAD）建模。现有的基于神经辐射场（NeRF）和三维几何结构（3DGS）的方法通常依赖于密集的标定视图、预计算的二维边缘图和昂贵的逐场景优化，这限制了它们在随意捕获的真实世界输入中的适用性。我们提出了CGGT，一种基于曲线的几何变换器，它直接从稀疏的、未标定的多视图图像中将三维一致的二维曲线实例嵌入到图像空间中。CGGT结合了一个几何感知的变换器编码器用于多视图特征学习，以及一个曲线感知的掩码注意力解码器用于跨视图实例关联。在单次前向传播中，它预测相机参数、密集深度图和实例级二维曲线掩码，这些掩码随后被提升到三维并通过快速的参数优化阶段进行精炼，以恢复紧凑的、可编辑的三维曲线原语。为了支持结构化曲线学习，我们引入了Wireframe-100K，这是一个大规模数据集，包含100,000个具有多样拓扑的CAD模型、真实的多视图渲染和准确的参数曲线注释。大量实验表明，我们的框架在重建精度和效率上都取得了显著提升，特别是在具有挑战性的稀疏视图设置下，以及在从由于轮廓、纹理和外观变化引起的视图依赖图像边缘中分离持久的三维结构边缘方面。尽管仅在合成数据上进行训练，CGGT在真实世界图像上表现出良好的泛化能力，展示了其在从不受限制的视觉输入中进行实用CAD风格线框重建的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 111 / 2609.14523

Selective Tool Use for Agentic Change Visual Question Answering in Remote Sensing

遥感中代理性变化视觉问答的选择性工具使用

Bazi, Yakoub, Rahhal, Mohamad M. Al, Mekhtiche, Mohamed A., Zuair, Mansour

Abstract

Change visual question answering (Change VQA) requires understanding semantic changes across bi-temporal remote sensing images. Although vision language models (VLMs) have shown promising performance on this task, they remain unreliable when answering questions that require explicit transition statistics, area measurements, or spatial information. To address this limitation, we propose a selective tool use framework in which a single VLM either answers directly or invokes a deterministic change analysis tool to obtain question specific evidence. Specifically, the selected tool operates on bi-temporal semantic maps and returns a structured observation, which the same VLM uses to generate its final answer. To support this framework, we construct a tool augmented extension of CDVQA covering eight question families and three tools for transition, spatial, and temporal analysis. Tool use supervision and observations are derived automatically from the original semantic annotations, without additional manual labeling. We then adapt Qwen3.5-4B using Low Rank Adaptation (LoRA) to jointly learn direct answering, tool invocation, and evidence conditioned answering. Experiments on 7,164 test questions show that selective tool use with reference semantic maps improves overall accuracy from 73.77% to 88.79% and average family accuracy from 69.11% to 89.65%. When the semantic maps are predicted automatically, the framework achieves 77.47% overall accuracy and 75.06% average family accuracy. These results demonstrate the benefit of question-specific semantic evidence for Change VQA, while highlighting the influence of semantic prediction quality on the resulting performance. Code and tool-augmented annotations will be made publicly available at https://github.com/yakoubbazi/ToolChangeVQA.

Chinese Translation

变化视觉问答（Change VQA）需要理解双时相遥感图像中的语义变化。尽管视觉语言模型（VLMs）在这一任务上表现出良好的性能，但在回答需要明确过渡统计、区域测量或空间信息的问题时，它们仍然不够可靠。为了解决这一局限性，我们提出了一种选择性工具使用框架，其中单个VLM要么直接回答问题，要么调用确定性的变化分析工具以获取特定问题的证据。具体而言，所选工具在双时相语义图上操作，并返回一个结构化的观察结果，VLM利用该结果生成最终答案。为了支持这一框架，我们构建了CDVQA的工具增强扩展，涵盖八个问题类别和三种用于过渡、空间和时间分析的工具。工具使用的监督和观察结果是自动从原始语义注释中提取的，无需额外的人工标注。随后，我们使用低秩适应（Low Rank Adaptation, LoRA）对Qwen3.5-4B进行调整，以共同学习直接回答、工具调用和基于证据的回答。在7164个测试问题上的实验表明，使用参考语义图的选择性工具使用将整体准确率从73.77%提高到88.79%，平均类别准确率从69.11%提高到89.65%。当语义图自动预测时，该框架实现了77.47%的整体准确率和75.06%的平均类别准确率。这些结果证明了特定问题的语义证据对变化视觉问答的益处，同时突显了语义预测质量对最终性能的影响。代码和工具增强注释将公开发布于https://github.com/yakoubbazi/ToolChangeVQA。

View on arXiv Download PDF AI Translation

cs.CV / 112 / 2609.14530

Sharing standardized image-derived data in computational pathology using DICOM

使用DICOM在计算病理学中共享标准化的图像衍生数据

Schacherer, Daniela P., Bridge, Christopher P., Clunie, David, Octaviano, Igor, Homeyer, André, Herrmann, Markus D., Gevaert, Olivier, Ghete, Tabita, Metzler, Markus, Hoefener, Henning, Kurc, Tahsin, Lisle, Curtis, Philbrick, Kenneth, Saltz, Joel, Zheng, Yuanning, Fedorov, Andrey

Abstract

Development and evaluation of computational pathology methods require access to large and diverse datasets. Over the past decade, various initiatives invested significantly into collecting, centralizing, and sharing pathology imaging data. In contrast, sharing of image-derived data such as region-of-interest delineations or segmentation masks is less well developed. In this work, we describe our approach to encoding and sharing image-derived pathology data in a standardized manner within the National Cancer Institute (NCI) Imaging Data Commons (IDC), a platform that hosts and provides public access to de-identified radiology and pathology data. The IDC relies on the Digital Imaging and Communications in Medicine (DICOM) standard for data harmonization, yet the adoption of DICOM for pathology image-derived content has remained largely unexplored until now. Here, we present five representative datasets harmonized by conversion from their original representations into DICOM and shared publicly in the IDC. We demonstrate the benefits of this harmonization, describe contributions to critical open-source tooling, and discuss technical considerations relevant to broader adoption of DICOM for pathology image-derived data.

Chinese Translation

开发和评估计算病理学方法需要访问大型且多样化的数据集。在过去的十年中，各种倡议在收集、集中和共享病理影像数据方面进行了大量投资。相比之下，图像衍生数据的共享，如感兴趣区域的划定或分割掩膜，发展相对较少。在本研究中，我们描述了在国家癌症研究所（NCI）影像数据共享平台（IDC）内以标准化方式编码和共享图像衍生病理数据的方法，该平台托管并提供去标识化的放射学和病理学数据的公共访问。IDC依赖于医学数字成像与通信（DICOM）标准进行数据协调，但迄今为止，DICOM在病理图像衍生内容中的应用仍然未得到充分探索。在此，我们展示了五个通过将其原始表示转换为DICOM而协调的代表性数据集，并在IDC中公开共享。我们展示了这种协调的好处，描述了对关键开源工具的贡献，并讨论了与DICOM在病理图像衍生数据更广泛应用相关的技术考虑。

View on arXiv Download PDF AI Translation

cs.CV / 113 / 2609.14560

Small Object Detection in Drone Aerial Imagery with LAF-YOLOv10

基于LAF-YOLOv10的无人机航拍图像中的小物体检测

Nayeem, Quratulain, Taranum, Fahmina, Uddin, Mohammed Mudassir

Abstract

General-purpose object detectors lose accuracy on UAV footage, where targets span only a handful of pixels and onboard compute is limited. Prior work composes independently-validated architectural techniques into one detector, assuming gains reported in isolation transfer once combined. We stress-test that assumption directly. LAF-YOLOv10 integrates four techniques into YOLOv10n: a Partial Convolution C2f (PC-C2f) backbone block, an Attention-Guided Feature Pyramid Network (AG-FPN), a P2 detection head replacing the large-object P5 head, and Wise-IoU v3 regression, asking whether their combined effect matches what each contributes alone. We train LAF-YOLOv10 three times (seeds 42, 123, 256) on VisDrone-DET2019, benchmark against unmodified YOLOv10n, and use TIDE error decomposition, per-category breakdown, per-component ablation, attention/loss comparisons, zero-shot transfer to UAVDT, and held-out/test-dev evaluation to localize where the combination succeeds or fails. Composability does not hold here. LAF-YOLOv10 reaches 24.0+/-0.4% [email protected] at 2.14M parameters, 7.8 points below YOLOv10n (31.8%), a deficit that transfers to UAVDT (-10.0 points) and is confirmed by held-out and test-dev evaluation (23.5%, 22.5%). Background false positives, localization error, and duplicate detections move in the direction AG-FPN and Wise-IoU were designed to push. Ablation traces the deficit to a specific source: the P2/-P5 head swap costs 2.5 points independently plus a 2.5-point interaction penalty when layered onto a backbone weakened by PC-C2f, whose own 2.0-point loss is consistent with a partial pretrained-weight transplant (73/150 backbone tensors transfer). The failure is attributable to a specific interaction, not the four components individually. Composability must be verified directly, not assumed. Code/checkpoints: https://github.com/Mudassiruddin7/Small-Object-Detection-in-UAV-Imagery.

Chinese Translation

通用物体检测器在无人机视频中准确性下降，因为目标仅占据少量像素且机载计算能力有限。之前的研究将独立验证的架构技术组合成一个检测器，假设单独报告的增益在组合后仍然有效。我们直接对这一假设进行了压力测试。LAF-YOLOv10将四种技术集成到YOLOv10n中：一个部分卷积C2f（PC-C2f）主干块、一个注意力引导特征金字塔网络（AG-FPN）、一个替换大物体P5头的P2检测头，以及Wise-IoU v3回归，旨在探讨它们的组合效果是否与各自单独贡献的效果相匹配。我们在VisDrone-DET2019上对LAF-YOLOv10进行了三次训练（种子42、123、256），并与未修改的YOLOv10n进行了基准测试，同时使用TIDE错误分解、按类别细分、按组件消融、注意力/损失比较、零样本转移到UAVDT，以及保留/测试开发评估来定位组合成功或失败的地方。组合性在这里并不成立。LAF-YOLOv10在214万参数下达到24.0+/-0.4%的[email protected]，低于YOLOv10n的31.8%（下降7.8点），这一缺陷在UAVDT上也有体现（下降10.0点），并通过保留和测试开发评估得到了确认（23.5%，22.5%）。背景虚假正例、定位误差和重复检测朝着AG-FPN和Wise-IoU设计的方向移动。消融实验将缺陷追溯到一个特定来源：P2/-P5头的交换独立损失2.5点，并且在与PC-C2f削弱的主干层叠时还会产生2.5点的交互惩罚，PC-C2f自身的2.0点损失与部分预训练权重的移植一致（73/150主干张量转移）。这一失败归因于特定的交互，而非四个组件的单独表现。组合性必须直接验证，而非假设。代码/检查点： https://github.com/Mudassiruddin7/Small-Object-Detection-in-UAV-Imagery.

View on arXiv Download PDF AI Translation

cs.CV / 114 / 2609.14578

A Building as a Repository: KIR, a Typed Intermediate Representation for Agent-Authored Building Information Models

建筑作为存储库：KIR，一种为代理编写的建筑信息模型的类型化中间表示

Kuklev, Dmitry

Abstract

Autonomous agents that author building information models need more than access to a host API. They need a representation of what they intended, what a compiler decided on their behalf, what was refused, what was observed after execution and what remains unknown. We present KIR, a typed intermediate representation in which a building is authored as a program held in a versioned repository and lowered to host applications as build targets. KIR is organised around seven ways in which a generator writing into a stateful, partially observable host goes silently wrong, and gives each a representation in data: ambiguous selectors become typed refusals with candidates; defaults keep their provenance; obligations that will not be checked are named before execution; vacuous witnesses are rejected statically; a lost transaction response becomes the state UNCONFIRMED with a verify-before-retry rule; the reverse path obeys a census invariant; and decisions are bound by digest to the state they were made against. On a pinned snapshot with 83 operation contracts and Revit as the only backend, offline experiments refuse 29 of 42 stress-test programs with diagnostic codes and no uncaught exception, name 38 of 377 witness obligations as unwitnessable, admit 31 of 100 combinations of execution, witness and acceptance states under seven stated invariants, and find no vacuous witness in 219 certificate runs; a 60-storey tower is 11,263 characters as KIR against 3,709,235 characters of emitted C#. Native Revit runs are reported from project records and kept separate from reproduced results. A controlled comparison with agents that write host code directly is specified but not yet executed; it is the principal open question.

Chinese Translation

自主代理在编写建筑信息模型时需要的不仅仅是对主机API的访问。他们需要一个表示，包含他们的意图、编译器为他们做出的决定、被拒绝的内容、执行后观察到的情况以及仍然未知的部分。我们提出了KIR，一种类型化中间表示，其中建筑被作为一个程序编写，存储在一个版本化的存储库中，并作为构建目标降低到主机应用程序。KIR围绕七种生成器在写入一个有状态的、部分可观察的主机时可能静默出错的方式进行组织，并为每种情况提供数据表示：模糊选择器变为带候选项的类型拒绝；默认值保持其来源；不会被检查的义务在执行前被命名；空洞证人被静态拒绝；丢失的事务响应变为状态UNCONFIRMED，并带有重试前验证规则；反向路径遵循人口普查不变性；决策通过摘要绑定到其所做的状态。在一个固定快照中，包含83个操作合同，Revit作为唯一后端，离线实验拒绝了42个压力测试程序中的29个，并提供了诊断代码且没有未捕获的异常，标记377个证人义务中的38个为不可证人，承认100个执行、证人和接受状态组合中的31个在七个声明的不变性下有效，并在219个证书运行中未发现空洞证人；一个60层的塔楼在KIR中为11,263个字符，而发出的C#代码为3,709,235个字符。原生Revit运行的结果来自项目记录，并与重现结果分开。与直接编写主机代码的代理的受控比较已被指定但尚未执行；这是主要的开放性问题。

View on arXiv Download PDF AI Translation

cs.CV / 115 / 2609.14590

Reverse Spatio-Temporal Disease Progression Modelling

逆向时空疾病进展建模

Shernazarov, Ulugbek, Xu, Moucheng, Ramatov, Inomjon

Abstract

Deep learning-based spatio-temporal disease progression models commonly overlook the incubation period of progressive diseases, limiting the use of those models in early interventions, which are vital for not easily reversible diseases such as Alzheimer's. This is because, the existing deep learning based longitudinal disease-progression models are almost always run forward: from an observed baseline they predict future decline. In many clinical settings, however, imaging begins only after pathology is suspected or already visible, the earlier, healthier patient-specific reference was never acquired. To address this, we propose to study reverse disease progression prediction: given later diseased anatomy, reconstruct the unobserved healthier anatomy that preceded it. We use a two-stage model in which a frozen 3D vector-quantised autoencoder defines a compact discrete latent space, while a Neural Ordinary Differential Equation (ODE) learns continuous-time dynamics in that space. A recurrent encoder reads late observations in reverse temporal order, initialises the latent state, and the ODE is integrated backwards across the trajectory. On a controlled Morpho-MNIST benchmark with a sinusoidal perturbation, our model successfully recovered the unseen previous states from later observations of the non-monotonic trajectory. On longitudinal brain MRIs from Alzheimer's Disease Neuroimaging Initiative, at the task to recover the previous unseen trajectory towards healthy states of the patients from observed later diseased states, our model outperforms the baselines that uses copy-nearest and mean-observed, with positive disease-reversal scores in every diagnostic stratum. We hope that our work can provide insights and tools towards discovering the incubation periods from single-shot scans, and developing early interventions of diseases based on imaging.

Chinese Translation

基于深度学习的时空疾病进展模型通常忽视渐进性疾病的潜伏期，这限制了这些模型在早期干预中的应用，而早期干预对于阿尔茨海默病等不易逆转的疾病至关重要。这是因为，现有的基于深度学习的纵向疾病进展模型几乎总是向前运行：从观察到的基线预测未来的衰退。然而，在许多临床环境中，影像学检查通常是在怀疑病理或病变已经可见后才开始进行，早期的、健康的患者特征参考从未被获取。为了解决这个问题，我们提出研究逆向疾病进展预测：给定后期的病变解剖结构，重建其之前未观察到的健康解剖结构。我们使用一个两阶段模型，其中一个冻结的3D向量量化自编码器定义了一个紧凑的离散潜在空间，而一个神经常微分方程（Neural Ordinary Differential Equation, ODE）在该空间中学习连续时间动态。一个递归编码器以逆时间顺序读取后期观察，初始化潜在状态，ODE沿着轨迹向后积分。在一个受控的Morpho-MNIST基准测试中，模型成功地从非单调轨迹的后期观察中恢复了未见的先前状态。在阿尔茨海默病神经影像学倡议的纵向脑MRI数据中，针对从观察到的后期病变状态恢复患者向健康状态的未见先前轨迹的任务，我们的模型在每个诊断层次上都超越了使用最近邻复制和均值观察的基线，获得了正向疾病逆转评分。我们希望我们的工作能够为从单次扫描中发现潜伏期提供见解和工具，并基于影像学开发疾病的早期干预措施。

View on arXiv Download PDF AI Translation

cs.CV / 116 / 2609.14596

Direct Conditional Transition Sampling for Diffusion Inverse Problems

扩散逆问题的直接条件转移采样

Yu, Qi, Wu, Hanlin, Sun, Xiaohui

Abstract

Training-free diffusion inverse solvers typically choose between local measurement guidance and costly clean-space posterior updates. Independent posterior refresh can improve global correction by sampling a clean conditional and re-noising it, but its practical realization requires probability-flow ODE integration and clean-space Markov chain Monte Carlo (MCMC). We propose Direct Conditional Transition Sampling (DCTS), a direct stochastic-flow approximation to the same ideal refresh target. Rather than explicitly drawing a clean sample, DCTS estimates the measurement-conditioned clean mean along a short inner path and transports Gaussian source noise directly to the next noisy state. A denoiser-compatible sufficient statistic and a covariance-scaled operator update enable this conditional-mean estimation. Experiments on four inverse problems demonstrate that DCTS achieves competitive reconstruction quality with up to $16.8\times$ speedups over competing methods.

Chinese Translation

无训练的扩散逆解算器通常在局部测量指导和昂贵的干净空间后验更新之间进行选择。独立的后验刷新可以通过采样干净条件并重新加噪声来改善全局校正，但其实际实现需要概率流常微分方程（ODE）积分和干净空间的马尔可夫链蒙特卡洛（MCMC）。我们提出了直接条件转移采样（Direct Conditional Transition Sampling, DCTS），这是一种对同一理想刷新目标的直接随机流近似。DCTS并不是显式地绘制一个干净样本，而是沿着一条短的内部路径估计测量条件下的干净均值，并将高斯源噪声直接传输到下一个噪声状态。一个与去噪器兼容的充分统计量和一个协方差缩放的算子更新使得这种条件均值估计成为可能。在四个逆问题上的实验表明，DCTS在重建质量上具有竞争力，并且相较于竞争方法实现了高达16.8倍的加速。

View on arXiv Download PDF AI Translation

cs.CV / 117 / 2609.14615

Open-UniMo: Towards Unified Motion-Language Understanding and Generation in the Open World

Open-UniMo：面向开放世界的统一运动-语言理解与生成

Wang, Guocun, Liu, Kenkun, Song, Guorui, Lin, Jing, Huang, Zhe, Zhang, Luyuan, Zhong, Dake, Wai, Choo Sin, Han, Xiaoguang, Wang, Haoqian

Abstract

Unified motion generation and understanding is crucial for embodied AI systems that can both synthesize and interpret human actions in open-world environments. Existing motion-language models often treat motion as an auxiliary modality of a language model, leading to text-dominated representations and limited cross-modal interaction. Moreover, the next-token prediction paradigm is not naturally suited to long motion sequences, where autoregressive generation may accumulate prediction errors. To address these challenges, we propose Open-UniMo, a unified Large Motion-Language Model (LMLM) trained on million-scale open-world motion-language data. Open-UniMo promotes modality parity by extending Qwen's vocabulary of about 150K text tokens with 64K motion tokens, enabling motion and language to share a unified token space. We further introduce motion-consistent Chain-of-Thought reasoning as an intermediate representation to bridge language semantics and motion dynamics. Open-UniMo is trained with a two-stage pipeline, where supervised fine-tuning establishes CoT-guided bidirectional motion-language mapping and Group Relative Policy Optimization (GRPO) improves semantic alignment while mitigating cumulative errors in autoregressive motion-token generation. To support comprehensive evaluation, we propose Open-MoBench, a VLM-guided benchmark for assessing text-to-motion (T2M) generation, motion-to-text (M2T) understanding, and bidirectional consistency. Extensive experiments show that Open-UniMo achieves state-of-the-art performance on both conventional metrics and Open-MoBench. Furthermore, ablation studies reveal that M2T understanding is not primarily limited by motion-token vocabulary size; instead, coupling M2T with the learnable T2M generation path yields stronger cross-modal representations, demonstrating that generation can facilitate understanding in AR-based motion-language modeling.

Chinese Translation

统一的运动生成与理解对于能够在开放世界环境中合成和解释人类动作的具身人工智能系统至关重要。现有的运动-语言模型通常将运动视为语言模型的辅助模态，导致文本主导的表征和有限的跨模态交互。此外，下一标记预测范式并不自然适用于长运动序列，其中自回归生成可能会累积预测误差。为了解决这些挑战，我们提出了Open-UniMo，一个在百万规模开放世界运动-语言数据上训练的统一大规模运动-语言模型（LMLM）。Open-UniMo通过将Qwen的约150K文本标记的词汇扩展到64K运动标记，促进了模态平等，使运动和语言共享统一的标记空间。我们进一步引入运动一致的链式思维推理作为中间表征，以桥接语言语义和运动动态。Open-UniMo采用两阶段管道进行训练，其中监督微调建立了基于链式思维的双向运动-语言映射，而群体相对策略优化（GRPO）则在减轻自回归运动标记生成中的累积误差的同时改善语义对齐。为了支持全面评估，我们提出了Open-MoBench，一个基于视觉语言模型（VLM）的基准，用于评估文本到运动（T2M）生成、运动到文本（M2T）理解和双向一致性。大量实验表明，Open-UniMo在传统指标和Open-MoBench上均达到了最先进的性能。此外，消融研究表明，M2T理解并不主要受限于运动标记词汇的大小；相反，将M2T与可学习的T2M生成路径结合能够产生更强的跨模态表征，表明生成可以促进基于自回归的运动-语言建模中的理解。

View on arXiv Download PDF AI Translation

cs.CV / 118 / 2609.14619

ESAFusion: LiDAR--4-D Radar Fusion via Local Geometric Complementation and Multiscale Adaptive Interaction for 3-D Object Detection

ESAFusion：通过局部几何补充和多尺度自适应交互实现的LiDAR--4-D雷达融合用于3-D目标检测

Ma, Gang, Hu, Senjie, Liu, Junjie, Wang, Chao, Wei, Hui

Abstract

LiDAR--4-D radar fusion combines accurate spatial geometry with motion and reflectivity cues from radar, offering a promising solution for 3-D object detection in complex driving environments. However, sparse radar observations and differences in spatial sampling between the two modalities complicate reliable cross-modal complementation. Moreover, the relative importance of modalities and feature scales varies across spatial regions, making adaptive fusion challenging. To address these challenges, we propose ESAFusion, an evidence-aware and scale-adaptive framework that combines local geometric complementation with multiscale adaptive interaction. Specifically, we introduce an Evidence-Aware Radar Selection (ERS) module to suppress radar clutter using motion and observation-quality evidence while retaining foreground confidence for subsequent fusion. Then, the Pillar-Level Complementary Encoder (PCE) improves cross-modal complementation under mismatched spatial sampling using local geometric support from neighboring LiDAR pillars. We further design an Intra- and Inter-Scale Adaptive Fusion (ISAF) module to adaptively adjust the contributions of different modalities and feature scales in bird's-eye-view (BEV) space. Extensive experiments on the View-of-Delft (VoD) dataset show that ESAFusion achieves the highest mean average precision (mAP) among the compared methods, reaching 74.60% in the Entire Annotated Area and 88.89% in the Driving Corridor. It also attains the highest average precision (AP) for Cyclist among these methods in both regions while running at 19.23 FPS. Evaluations on VoD-Fog further demonstrate robustness under progressively degraded LiDAR observations. The source code will be made publicly available at https://github.com/SenJieHu549/ESAFusion.

Chinese Translation

LiDAR--4-D雷达融合结合了来自雷达的准确空间几何、运动和反射率线索，为复杂驾驶环境中的3-D目标检测提供了有前景的解决方案。然而，稀疏的雷达观测和两种模态之间的空间采样差异使得可靠的跨模态补充变得复杂。此外，各空间区域模态和特征尺度的重要性相对不同，这使得自适应融合面临挑战。为了解决这些问题，我们提出了ESAFusion，这是一个证据感知和尺度自适应框架，结合了局部几何补充和多尺度自适应交互。具体而言，我们引入了证据感知雷达选择（ERS）模块，通过运动和观测质量证据抑制雷达杂波，同时保留前景置信度以便后续融合。然后，柱级补充编码器（PCE）利用邻近LiDAR柱的局部几何支持，在空间采样不匹配的情况下改善跨模态补充。我们进一步设计了内部和跨尺度自适应融合（ISAF）模块，以在鸟瞰图（BEV）空间中自适应调整不同模态和特征尺度的贡献。在View-of-Delft（VoD）数据集上的大量实验表明，ESAFusion在比较方法中实现了最高的平均精度均值（mAP），在整个标注区域达到74.60%，在驾驶走廊中达到88.89%。在这两个区域中，它还在骑自行车者的平均精度（AP）方面取得了最高成绩，同时以19.23 FPS的速度运行。对VoD-Fog的评估进一步证明了在逐渐退化的LiDAR观测下的鲁棒性。源代码将公开发布在https://github.com/SenJieHu549/ESAFusion。

View on arXiv Download PDF AI Translation

cs.CV / 119 / 2609.14624

TTDF: A Two-Stage Framework for Reliable Surgical Phase Transition Detection

TTDF：一种用于可靠手术阶段转换检测的两阶段框架

Guo, Yushi, Valdastri, Pietro, Sarikaya, Duygu

Abstract

Reliable workflow transition detection is important for context-aware surgical assistance and downstream decision support. However, online surgical phase recognizers primarily focus on frame-wise accuracy and temporal consistency, rather than the reliability of workflow transition events. Directly converting phase changes into events is unreliable: temporal jitter and workflow-illegal switches produce false or duplicate events, while persistent, workflow-consistent candidates may remain incorrect. To address this limitation, we formulate reliable workflow transition detection as a distinct event-level task operating on outputs of a frozen online phase recognizer. We propose the Two-Stage Transition Detection Framework (TTDF), a causal framework that progressively filters transition candidates. Transition Candidate Extraction (TCE) first applies a minimum-duration requirement and a workflow-graph constraint to remove false candidates caused by temporal jitter and phase transitions not allowed by the workflow graph. Specifically, a candidate is retained only if the predicted target phase persists for a minimum duration and the ordered phase pair belongs to the workflow graph's allowed transition set. TCE thereby produces a high-recall candidate set without additional training. Transition Candidate Verification (TCV) suppresses remaining false candidates using phase-posterior shifts and visual-change cues from frozen DINOv2 features. Events are assessed using a phase-pair-aware one-to-one matching protocol. Experiments on Cholec80 show that TTDF reduces false transition emissions while preserving recall and controlling decision delay.

Chinese Translation

可靠的工作流转换检测对于上下文感知的手术辅助和后续决策支持至关重要。然而，在线手术阶段识别器主要关注逐帧准确性和时间一致性，而不是工作流转换事件的可靠性。直接将阶段变化转换为事件是不可靠的：时间抖动和工作流非法切换会产生虚假或重复事件，而持续且符合工作流的候选项可能仍然不正确。为了解决这一局限性，我们将可靠的工作流转换检测表述为一个独立的事件级任务，该任务在冻结的在线阶段识别器的输出上进行操作。我们提出了两阶段转换检测框架（TTDF），这是一个因果框架，逐步过滤转换候选项。转换候选提取（TCE）首先应用最小持续时间要求和工作流图约束，以去除因时间抖动和工作流图不允许的阶段转换而导致的虚假候选项。具体而言，只有在预测的目标阶段持续至少一段最小时间且有序阶段对属于工作流图的允许转换集时，候选项才会被保留。因此，TCE在没有额外训练的情况下产生高召回率的候选集。转换候选验证（TCV）利用阶段后验变化和来自冻结的DINOv2特征的视觉变化线索来抑制剩余的虚假候选项。事件使用阶段对感知的一对一匹配协议进行评估。在Cholec80上的实验表明，TTDF在保持召回率和控制决策延迟的同时减少了虚假转换的发生。

View on arXiv Download PDF AI Translation

cs.CV / 120 / 2609.14634

SCOUT-SLAM: Structurally-Coupled Dual Uncertainty-Aware 3DGS SLAM in the Wild

SCOUT-SLAM：野外结构耦合的双重不确定性感知3D高斯点云SLAM

Karthik, Kumaran, Jois, Pramat Shastri, Sundaram, Suresh

Abstract

Recently, 3D Gaussian Splatting SLAM (3DGS-SLAM) has gained significant momentum in simultaneous localization and 3DGS scene reconstruction. In real-world scenarios with rapid camera motion and cluttered dynamic environments, existing methods rely on the stability of the underlying scene reconstruction to model uncertainty. This leads to a circular dependency between camera tracking accuracy and reconstruction quality: reconstruction instabilities degrade uncertainty modeling, which affects accurate camera tracking and static scene reconstruction. To address this, the paper proposes SCOUT-SLAM, a structurally-coupled dual-uncertainty framework in which both uncertainties are estimated from a shared base network. A low-rank adaptation of this network, trained on multi-view feature consistency, estimates a tracking uncertainty that does not depend solely on the reconstruction quality. A spatially-adaptive prior modulates the network's training objective so that reconstruction instability does not inflate uncertainty on static regions, keeping the shared representation intact for both branches. Evaluations on dynamic benchmarks (TUM RGB-D, Bonn Dynamic, Wild-SLAM MoCap) demonstrate that SCOUT-SLAM achieves state-of-the-art camera tracking accuracy and artifact-free static scene reconstruction.

Chinese Translation

近年来，3D高斯点云SLAM（3DGS-SLAM）在同时定位与3DGS场景重建方面取得了显著进展。在快速相机运动和杂乱动态环境的真实场景中，现有方法依赖于基础场景重建的稳定性来建模不确定性。这导致了相机跟踪精度与重建质量之间的循环依赖：重建不稳定性降低了不确定性建模，从而影响了准确的相机跟踪和静态场景重建。为了解决这一问题，本文提出了SCOUT-SLAM，一个结构耦合的双重不确定性框架，其中两个不确定性均来自共享的基础网络。该网络的低秩适应性训练于多视图特征一致性，估计出一种不单纯依赖于重建质量的跟踪不确定性。空间自适应先验调节网络的训练目标，以确保重建不稳定性不会在静态区域膨胀不确定性，从而保持共享表示在两个分支中的完整性。在动态基准测试（TUM RGB-D、波恩动态、Wild-SLAM MoCap）上的评估表明，SCOUT-SLAM实现了最先进的相机跟踪精度和无伪影的静态场景重建。

View on arXiv Download PDF AI Translation

cs.CV / 121 / 2609.14654

Evaluating Contextual Bias in CNN Image Classification: Evidence from Agricultural Benchmark Datasets

评估卷积神经网络图像分类中的上下文偏差：来自农业基准数据集的证据

Dalal, Abhilekha, Okonoda, Michael, Martinez, Eder, Shamir, Lior

Abstract

Convolutional neural networks (CNNs) are typically evaluated using held-out classification accuracy, an approach that presupposes predictions are based primarily on the intended object of interest rather than incidental surrounding context. We test this assumption in CNN-based agricultural image classification by comparing model performance on original images with performance on background-dominated patches extracted from the same images across eight publicly available agricultural benchmark datasets and four widely used CNN architectures. Background-dominated patches were classified above dataset-specific random chance for six of the eight datasets, and substantially above chance for four of them, indicating that contextual information contributes to model predictions for the majority of datasets evaluated. For these four datasets, we further evaluated whether this behavior reflected genuine class-discriminative information or was primarily attributable to class imbalance using macro-averaged precision, recall, and F1 together with class-balanced test subsets. The results show that contextual reliance does not admit a single explanation: class imbalance accounts for a substantial portion of the observed signal for some datasets and architectures, whereas above-chance contextual classification persists after balancing for others. Together with previous evidence from curated object recognition and cancer pathology imaging, these findings support the growing view that contextual bias is a recurring characteristic of CNN-based image classification rather than a phenomenon confined to a single application domain. More broadly, this work provides a systematic framework for quantifying contextual bias across heterogeneous image datasets by combining dataset-specific random-chance baselines, contextual bias categorization, macro-averaged evaluation, and class-balanced robustness analysis.

Chinese Translation

卷积神经网络（CNN）通常通过保留的分类准确率进行评估，这种方法假设预测主要基于目标对象而非偶然的周围背景。我们通过比较模型在原始图像上的表现与在从同一图像中提取的背景主导补丁上的表现，测试了这一假设，涉及八个公开可用的农业基准数据集和四种广泛使用的CNN架构。对于八个数据集中的六个，背景主导补丁的分类准确率超过了数据集特定的随机机会，而对于其中四个则显著高于随机机会，这表明上下文信息对大多数评估数据集的模型预测有贡献。对于这四个数据集，我们进一步评估了这种行为是否反映了真实的类别区分信息，或主要归因于类别不平衡，使用了宏平均精度、召回率和F1分数，以及类别平衡的测试子集。结果显示，上下文依赖性并没有单一的解释：类别不平衡在某些数据集和架构中占据了观察到信号的相当一部分，而在其他数据集中，经过平衡后，上下文分类仍然高于随机机会。结合之前在策划的物体识别和癌症病理成像中的证据，这些发现支持了上下文偏差是CNN图像分类的一个反复出现的特征，而非仅限于单一应用领域的现象。更广泛地说，这项工作提供了一个系统框架，通过结合数据集特定的随机机会基准、上下文偏差分类、宏平均评估和类别平衡的稳健性分析，量化异构图像数据集中的上下文偏差。

View on arXiv Download PDF AI Translation

cs.CV / 122 / 2609.14657

Compositional SVG Generation via VLM-Driven Hierarchical Semantic Parsing

基于VLM驱动的层次语义解析的组合SVG生成

Park, Sehwan, Kim, Taehoon, Han, Geonhee, Kim, Dohyun, Kim, Seung Wook, Seo, Paul Hongsuck

Abstract

While Vision-Language Models (VLMs) excel at visual reasoning, generating structured, editable Scalable Vector Graphics (SVG) remains a fundamental challenge. Existing pipelines predominantly yield flat, semantically agnostic collections of paths, where editing a single object requires manually identifying its constituent paths. To address this, we propose a VLM-driven agentic framework for semantic compositional SVG generation. Our pipeline recursively parses visual scenes into semantic and geometric hierarchies via top-down decomposition, visual grounding, and prompt-driven amodal occlusion recovery, ensuring each component is geometrically complete. Furthermore, we introduce the Semantic SVG Benchmark with human-annotated semantic groups and novel sub-component metrics (Semantic Recall/Precision, PERE) to explicitly evaluate structural compositionality and functional editability. Experiments show that our natively predicted structures surpass the upper bounds of existing flat-generation methods in both grouping quality and editability, while maintaining state-of-the-art visual fidelity.

Chinese Translation

尽管视觉语言模型（VLMs）在视觉推理方面表现出色，但生成结构化、可编辑的可缩放矢量图形（SVG）仍然是一个基本挑战。现有的生成流程主要产生平面、语义无关的路径集合，其中编辑单个对象需要手动识别其组成路径。为了解决这一问题，我们提出了一种基于VLM驱动的代理框架，用于语义组合SVG生成。我们的流程通过自上而下的分解、视觉定位和基于提示的模态遮挡恢复，递归地将视觉场景解析为语义和几何层次，确保每个组件在几何上是完整的。此外，我们引入了语义SVG基准，包含人工标注的语义组和新的子组件度量（语义召回/精度，PERE），以明确评估结构组合性和功能可编辑性。实验表明，我们原生预测的结构在分组质量和可编辑性方面超越了现有平面生成方法的上限，同时保持了最先进的视觉保真度。

View on arXiv Download PDF AI Translation

cs.CV / 123 / 2609.14674

Floquet Fibre Geometry and Higher-Order Reduced Coordinates for Off-Manifold Transients near Nonlinear Aeroelastic Flutter

Floquet纤维几何及非线性气动弹性颤振附近的高阶降维坐标

Tan, Puxue

Abstract

Assigning reduced coordinates to states near an attracting limit cycle requires the correct invariant-fibre geometry. The classical first-order phase-isostable chart obtained from adjoint Floquet modes projects along the strong-stable quotient fibre, whereas a metric-orthogonal complement of the retained slow bundle generally does not. We prove locally that a chart satisfying the linearised semiconjugacy relation leaves an O(delta^2) invariance residual, while projection along a non-invariant complement generically leaves an O(delta) term. For a nonlinear aeroelastic limit cycle, the metric-normal and strong-stable directions differ by 48.5 to 71.7 degrees, and metric-normal perturbations contain first-order retained phase and slow-amplitude components. Replacing the metric normal by the strong-stable fibre changes the measured residual scaling from delta^1.01 to delta^1.87 without fitted parameters. We then test learned higher-order corrections whose linearisation is pinned to the adjoint-Floquet chart, whose symmetry is exact, and whose reduced flow is fixed. Although they reduce the registered fixed-normalisation latent residual, post-hoc amplitude recalibration and adjoint-Floquet-targeted future consistency move or reverse the ranking. Because the learned maps already share the baseline's first-order gauge and the future target is supplied by the baseline chart, these diagnostics establish neither an independent positive nor negative higher-order result. Correct first-order Floquet geometry is therefore necessary in this benchmark, while the additional predictive value of the learned correction remains unidentified by the available representation-dependent diagnostics.

Chinese Translation

将降维坐标分配给吸引极限环附近的状态需要正确的不变纤维几何。通过伴随Floquet模态获得的经典一阶相位等稳图沿强稳定商纤维投影，而保留的慢束的度量正交补通常不然。我们局部证明，满足线性化半共轭关系的图留下O(delta^2)的不变性残余，而沿非不变补的投影通常留下O(delta)项。对于非线性气动弹性极限环，度量法向和强稳定方向之间的差异为48.5至71.7度，度量法向扰动包含一阶保留相位和慢幅度分量。用强稳定纤维替换度量法向会将测量的残余缩放从delta^1.01改变为delta^1.87，而无需拟合参数。然后，我们测试学习到的高阶修正，其线性化固定在伴随Floquet图上，该图的对称性是精确的，且其降维流是固定的。尽管它们减少了注册的固定归一化潜在残余，但事后幅度重新校准和针对伴随Floquet的未来一致性会移动或逆转排名。由于学习到的映射已经共享基线的一阶规范，而未来目标由基线图提供，因此这些诊断既未建立独立的正向也未建立负向的高阶结果。因此，在此基准中，正确的一阶Floquet几何是必要的，而学习到的修正的额外预测价值仍未通过可用的表示依赖诊断识别。

View on arXiv Download PDF AI Translation

cs.CV / 124 / 2609.14676

From Density to Biopsy Decisions and Malignancy Prediction: A Benchmark Study of Multimodal Large Language Models Against Radiologists in Digital and Contrast-Enhanced Mammography

从密度到活检决策和恶性肿瘤预测：多模态大型语言模型与放射科医生在数字和对比增强乳腺摄影中的基准研究

Ardakani, Ali Abbasian, Mohammadi, Afshin, Kuzan, Taha Yusuf, Kuzan, Beyza Nur, Mohebbi, Alisa, Behruzi, Masume, Khorshidi, Hamid, Ghorbani, Ashkan, Asadiara, Elham, Lotfi, Zeinab Khorshidi, Rahman, Ansar, Beste, Nedim Christoph, Acharya, U. Rajendra, Hatamikia, Sepideh

Abstract

Purpose: To compare four multimodal large language models (MLLMs) with radiologists of varying expertise in breast density assessment, BI-RADS assessment, biopsy candidacy determination, and continuous malignancy probability estimation using digital mammography (DM) and contrast-enhanced mammography (CEM). Methods: This study included 179 women with paired DM/CEM examinations and reference standards. Four MLLMs (ChatGPT-5.2, Gemini-3.1 Pro, Sonnet-4.6, Muse Spark) interpreted images with and without masks; three radiologists interpreted non-masked images. Results: For binary density classification on DM, radiologist accuracies ranged from 55.81% to 78.60%, exceeding most MLLM values (62.33%-71.63%), while masks added limited benefit. Five-category BI-RADS accuracies were higher for radiologists on DM (56.74-67.44%) and CEM (62.33-82.79%) compared with MLLMs (DM 31.16-45.12%; CEM 40.00-55.81%). Binary biopsy-candidacy accuracies were likewise higher for radiologists (DM 85.12-89.77%; CEM 86.98-92.09%) than for MLLMs (DM 61.39-75.35%; CEM 69.30-82.79%), although CEM improved performance across all readers. Lesion masks substantially improved MLLM continuous malignancy-probability accuracies from 64.65%-71.63% to 72.56-78.60% on DM and from 67.91%-77.21% to 72.56%-81.86% on CEM, approaching radiologist ranges (DM 63.72-82.79%; CEM 81.86-88.84%). The corresponding AUCs for the top masked models overlapped those of the human readers. Overall, Muse Spark, followed by Sonnet-4.6, demonstrated the strongest performance among the MLLMs across domains. Conclusion: Radiologists generally outperformed MLLMs in categorical tasks, while selected masked models approached human performance for continuous malignancy probability estimation, suggesting a potential adjunctive role.

Chinese Translation

目的：比较四种多模态大型语言模型（MLLMs）与不同专业水平的放射科医生在乳腺密度评估、BI-RADS 评估、活检适应性判断和使用数字乳腺摄影（DM）及对比增强乳腺摄影（CEM）进行的连续恶性肿瘤概率估计方面的表现。方法：本研究纳入了179名接受配对 DM/CEM 检查及参考标准的女性。四种 MLLMs（ChatGPT-5.2、Gemini-3.1 Pro、Sonnet-4.6、Muse Spark）对有无遮罩的图像进行了解读；三名放射科医生解读了非遮罩图像。结果：在 DM 上的二分类密度分类中，放射科医生的准确率范围为55.81%至78.60%，超过了大多数 MLLM 的值（62.33%-71.63%），而遮罩的增益有限。在 DM（56.74%-67.44%）和 CEM（62.33%-82.79%）上的五分类 BI-RADS 准确率高于 MLLMs（DM 31.16%-45.12%；CEM 40.00%-55.81%）。在活检适应性判断的二分类准确率方面，放射科医生（DM 85.12%-89.77%；CEM 86.98%-92.09%）同样高于 MLLMs（DM 61.39%-75.35%；CEM 69.30%-82.79%），尽管 CEM 在所有读者中提高了表现。病变遮罩显著提高了 MLLM 在 DM 上的连续恶性肿瘤概率准确率，从64.65%-71.63%提升至72.56%-78.60%，在 CEM 上从67.91%-77.21%提升至72.56%-81.86%，接近放射科医生的范围（DM 63.72%-82.79%；CEM 81.86%-88.84%）。顶级遮罩模型的相应 AUC 与人类读者的重叠。总体而言，Muse Spark 在各领域中表现最强，其次是 Sonnet-4.6。结论：放射科医生在分类任务中通常优于 MLLMs，而选定的遮罩模型在连续恶性肿瘤概率估计方面接近人类表现，暗示其可能的辅助作用。

View on arXiv Download PDF AI Translation

cs.CV / 125 / 2609.14690

LIMODENet: Attention-Free Compact Encoders for Information-Preserving Onboard Satellite Image Restoration

LIMODENet：无注意力紧凑编码器用于信息保留的卫星图像恢复

Le, Thanh-Dung, Ha, Vu Nguyen, Nguyen, Ti Ti, Chatzinotas, Symeon

Abstract

Onboard satellites must restore a channel-degraded image on a few watts, using neuromorphic accelerators (e.g., BrainChip Akida, Intel Loihi-2) that support no softmax or attention. We ask which encoder restores best under that constraint and introduce LIMODENet (LinearMix-ODENet), a 0.69M softmax-/QKV-free backbone whose residual stages read as ODE discretizations and which is empirically information-preserving (probe accuracy rises 79.9% -> 98.4% from stem to head). At iso-parameters it restores 1 dB DVB-S2X-degraded EuroSAT better than a CNN autoencoder (+1.75 dB PSNR) and a skip-connection U-Net (+1.07 dB), three seeds, non-overlapping. Unconstrained modern restorers (NAFNet, Restormer) win on fidelity; we decompose that gap: spiking-legal additive skips recover about half, and the rest traces to attention and channel gating. LIMODENet then converts end-to-end to a spiking network with zero blocked operations, versus 22-24 for the competitors: not the best restorer available, but the best verified deployable within a real power budget.

Chinese Translation

在有限的功耗（几瓦）下，卫星必须恢复通道退化的图像，使用不支持softmax或注意力机制的神经形态加速器（例如，BrainChip Akida，Intel Loihi-2）。我们探讨在这一限制下，哪种编码器能够实现最佳恢复，并引入LIMODENet（LinearMix-ODENet），这是一种0.69M的无softmax/QKV骨干网络，其残差阶段可视为常微分方程（ODE）离散化，并且在经验上具有信息保留特性（探测准确率从79.9%提升至98.4%）。在相同参数下，它比CNN自编码器恢复1 dB的DVB-S2X退化EuroSAT图像更好（+1.75 dB PSNR），比带跳跃连接的U-Net更好（+1.07 dB），在三次实验中结果不重叠。虽然现代无约束恢复器（NAFNet，Restormer）在保真度上表现更佳，但我们分析了这一差距：合法的脉冲加法跳跃恢复了大约一半，其余部分归因于注意力和通道门控。LIMODENet随后可以无阻塞操作地端到端转换为脉冲网络，相比之下，竞争对手需要22-24个阻塞操作：虽然不是最佳的恢复器，但在实际功耗预算内验证的最佳可部署方案。

View on arXiv Download PDF AI Translation

cs.CV / 126 / 2609.14699

Perceive, Refine, Reason: A Calibrated Pipeline for Measuring Indicators in Strategic Visual Communication on Social Media

感知、精炼、推理：社交媒体战略视觉传播指标测量的校准管道

Qi, Weihong, Ling, Chen

Abstract

Visual content shapes audience perception and opinion on social media, and computational social science increasingly relies on automated tools to analyze images at scale. Yet a measurement gap persists: existing tools rely on predefined categories or produce only coarse image-level labels, while measuring which specific objects appear in an image, how prominently, and where in the frame remains difficult at scale. We introduce Perceive, Refine, Reason (PRR), a calibrated pipeline that turns flexible vision-language detectors into auditable measurement instruments for social-scientific research. PRR combines natural-language category prompts with pixel-level spatial refinement via the Segment Anything Model (SAM) and a multimodal LLM arbitration layer whose reasoning chains externalize domain knowledge and lower the expertise threshold for human-in-the-loop validation. A complementary three-tier auditability framework applies quantification learning to profile per-category reliability, support task-aligned configuration, and statistically correct prevalence estimates. Across four vision-language detectors and nine sociological categories, the pipeline yields substantial precision gains over zero-shot baselines, including a 43.3-point improvement for the strongest backbone. Applying PRR to 103,920 Facebook images from U.S. legislators during the 2024 election cycle and linking detections to DW-NOMINATE ideology scores, we find that more conservative legislators display U.S. flags as larger visual elements, with a weaker tendency toward peripheral placement, a spatial pattern invisible to binary detection. PRR provides computational social scientists with a model-agnostic toolkit for accessible, spatially-grounded, and correctable visual measurement.

Chinese Translation

视觉内容塑造了社交媒体上观众的感知和观点，而计算社会科学越来越依赖自动化工具对图像进行大规模分析。然而，测量存在一个差距：现有工具依赖预定义类别或仅生成粗略的图像级标签，而测量图像中具体出现了哪些对象、它们的显著性以及在画面中的位置在大规模上仍然困难。我们提出了感知、精炼、推理（Perceive, Refine, Reason, PRR），这是一个校准管道，将灵活的视觉-语言检测器转变为可审计的社会科学研究测量工具。PRR结合了自然语言类别提示与通过Segment Anything Model (SAM)进行的像素级空间精细化，以及一个多模态LLM仲裁层，其推理链外化领域知识，降低了人机协作验证的专业门槛。一个补充的三层审计框架应用量化学习来分析每个类别的可靠性，支持任务对齐配置，并统计校正流行度估计。在四个视觉-语言检测器和九个社会学类别中，该管道在零样本基线之上实现了显著的精度提升，包括最强主干模型的43.3点改进。将PRR应用于2024年选举周期中来自美国立法者的103,920张Facebook图像，并将检测结果与DW-NOMINATE意识形态分数关联，我们发现更保守的立法者将美国国旗作为更大的视觉元素展示，并且在边缘位置的倾向较弱，这一空间模式在二元检测中是不可见的。PRR为计算社会科学家提供了一个模型无关的工具包，以实现可访问、基于空间的和可修正的视觉测量。

View on arXiv Download PDF AI Translation

cs.CV / 127 / 2609.14703

Patient-Level, Leakage-Aware Deep Learning for Cross-Center Periapical Radiograph Classification

患者级、泄漏意识的深度学习用于跨中心根尖周围放射影像分类

Rahman, Md Jubaer, Bagci, Ulas

Abstract

Dental caries and endodontic disease are among the most common health conditions worldwide, and intraoral periapical radiographs are central to their detection, treatment planning, and follow-up. Automated tooth-level classification of these images, however, lacks reproducible benchmarks, is often evaluated with image-level splits that leak patients between training and test, and is rarely validated across clinics. This paper presents the first patient-level, leakage-aware classification benchmark for single-tooth intraoral periapical radiographs on the DentIRO dataset, which comprises 5,300 images from 3,243 patients across two clinics and four classes: Healthy, Caries, Crowned, and Root Canal. Five transfer-learning models are compared with patient-grouped stratified cross-validation, so that every patient remains within a single fold. DenseNet121 gave the strongest and most stable result at a mean macro-F1 of 0.9787, while the four ImageNet-initialized backbones performed comparably. A controlled comparison on a fixed architecture showed that chest-radiograph pretraining transferred less effectively than ImageNet initialization. Bidirectional cross-center validation produced a small average generalization gap of 0.0077, and Grad-CAM confirmed that predictions rely on clinically meaningful tooth regions rather than acquisition artifacts. The benchmark offers a rigorous and reproducible baseline for intraoral radiograph classification.

Chinese Translation

牙齿龋齿和根管病是全球最常见的健康问题之一，而口腔内根尖周围放射影像在其检测、治疗计划和随访中至关重要。然而，这些图像的自动化牙齿级分类缺乏可重复的基准，通常使用的图像级划分会导致训练和测试之间的患者泄漏，并且很少在不同诊所进行验证。本文提出了首个针对单颗牙齿口腔内根尖周围放射影像的患者级、泄漏意识分类基准，基于DentIRO数据集，该数据集包含来自两家诊所的3,243名患者的5,300张图像，分为四类：健康、龋齿、冠牙和根管。通过患者分组的分层交叉验证比较了五种迁移学习模型，以确保每位患者仅出现在一个折叠中。DenseNet121在平均宏F1值为0.9787的情况下表现出最强和最稳定的结果，而四种基于ImageNet初始化的骨干网络表现相似。在固定架构上的受控比较显示，胸部放射影像的预训练转移效果不如ImageNet初始化。双向跨中心验证产生了0.0077的小平均泛化差距，Grad-CAM证实了预测依赖于临床上有意义的牙齿区域，而非采集伪影。该基准为口腔放射影像分类提供了严格且可重复的基线。

View on arXiv Download PDF AI Translation

cs.CV / 128 / 2609.14722

PC$^2$-AD: Point Cloud Upsampling to Safeguard 3D Anomaly Detection with Resolution-constrained Edge Devices

PC$^2$-AD：点云上采样以保护具有分辨率限制的边缘设备中的3D异常检测

Gu, Yutong, Xie, Yingxi, Huang, Kejin, Ning, Jian, Liang, Hanzhe, Shen, Linlin, Wang, Jinbao

Abstract

Low-cost and low-resolution sensors used in edge deployments can produce test point clouds that are substantially sparser than the normal training data. This train-test sampling-resolution gap changes the local geometry available to a 3D anomaly detector. We propose PC$^2$-AD, a point cloud upsampling framework that compensates sparse test inputs before downstream detection. Target Domain Candidate Generation (TCG) adapts a pretrained upsampler to normal training geometry and generates a dense candidate pool. Geometry-Aware Candidate Filtering (GACF) selects candidates according to geometric spacing and spatial coverage. Normality-Preserving Point Compensation (NPPC) refines the selection by comparing candidate normality scores with those of their input anchors. The selected points are combined with the unchanged input points and processed by the existing detector. Experiments with six detectors on two Anomaly-ShapeNet settings and Real3D-AD show improvements in the mean of object-level and point-level AUROC for all six detectors in each Anomaly-ShapeNet setting and four on Real3D-AD. These results support point cloud compensation as an input-level approach to improving 3D anomaly detection under low-resolution sensing conditions. Code is publicly available at https://github.com/gyutong406-commits/PC2-AD.

Chinese Translation

在边缘部署中使用的低成本低分辨率传感器可能会生成测试点云，这些点云的稀疏程度远低于正常的训练数据。这种训练-测试采样分辨率差异改变了3D异常检测器可用的局部几何结构。我们提出了PC$^2$-AD，一个点云上采样框架，在下游检测之前补偿稀疏的测试输入。目标领域候选生成（Target Domain Candidate Generation, TCG）将预训练的上采样器适应于正常的训练几何，并生成一个密集的候选池。几何感知候选过滤（Geometry-Aware Candidate Filtering, GACF）根据几何间距和空间覆盖选择候选。正常性保持点补偿（Normality-Preserving Point Compensation, NPPC）通过比较候选的正常性分数与其输入锚点的正常性分数来精炼选择。选定的点与未改变的输入点结合，并由现有检测器处理。在两个Anomaly-ShapeNet设置和Real3D-AD上的六个检测器实验表明，在每个Anomaly-ShapeNet设置中，所有六个检测器的对象级和点级AUROC均有所提高，在Real3D-AD上则有四个检测器表现出改进。这些结果支持点云补偿作为一种输入级方法，以改善低分辨率传感条件下的3D异常检测。代码已公开发布在 https://github.com/gyutong406-commits/PC2-AD。

View on arXiv Download PDF AI Translation

cs.CV / 129 / 2609.14725

CrossDistill: Balancing Quality and Diversity via Trajectory-Level Hybrid Few-Step Distillation

CrossDistill：通过轨迹级混合少步蒸馏平衡质量与多样性

Liu, Yuxi, Li, Haoyu, Cai, Yixiang, Sun, Tengxu, Zhang, Zekun, Ai, Baole, Wang, Ang, Wang, Jiamang, Qu, Lin, Yuan, Kun, Zhang, Kai

Abstract

Few-step distillation accelerates diffusion models but must balance diversity and fidelity: trajectory-based distillation preserves mode coverage, while distribution matching sharpens samples but can reduce diversity. We show that this tension can be exploited in a noise-regime-dependent way: high-noise steps largely determine global modes, whereas low-noise steps refine local details. We propose CrossDistill, a trajectory-level hybrid distillation framework that splits the sampling trajectory at a crossover point, applies a trajectory-preserving objective on the high-noise interval and a distribution-matching objective on the low-noise interval, and couples the two stages through the crossover state. In contrast to loss-level mixing, and complementarily to training-time two-stage recipes, CrossDistill explicitly assigns complementary objectives along the noise axis, so that global branching is preserved before local statistics are sharpened. CrossDistill is a noise-level scheduling policy: PCM and DMD are plug-in instantiations, while the noise partition, crossover coupling, and objective ordering are the key design elements. Experiments on text-to-video diffusion models and qualitative image-to-video results show that CrossDistill expands the few-step quality-diversity frontier, retaining seed-level variation while achieving competitive visual fidelity.

Chinese Translation

少步蒸馏加速了扩散模型，但必须平衡多样性和保真度：基于轨迹的蒸馏保留了模式覆盖，而分布匹配则锐化了样本但可能降低多样性。我们展示了这种张力可以以依赖噪声状态的方式进行利用：高噪声步骤在很大程度上决定了全局模式，而低噪声步骤则细化了局部细节。我们提出了CrossDistill，一种轨迹级混合蒸馏框架，它在交叉点处分割采样轨迹，在高噪声区间应用轨迹保持目标，在低噪声区间应用分布匹配目标，并通过交叉状态将两个阶段耦合。与损失级混合相对，并且作为训练时间两阶段方案的补充，CrossDistill明确地沿噪声轴分配互补目标，以便在局部统计被锐化之前保留全局分支。CrossDistill是一种噪声级调度策略：PCM和DMD是插件实例，而噪声划分、交叉耦合和目标排序是关键设计元素。对文本到视频扩散模型的实验和定性图像到视频的结果表明，CrossDistill扩展了少步质量-多样性的前沿，保留了种子级变异，同时实现了具有竞争力的视觉保真度。

View on arXiv Download PDF AI Translation

cs.CV / 130 / 2609.14761

From Visual Feedback to Textual Reviews: A Multi-Agent Vision-Language Framework for Image-Grounded Review Assistance

从视觉反馈到文本评论：一种多智能体视觉-语言框架用于图像基础的评论辅助

Nareti, Utsav Kumar, Bansal, Ayush, Priya, Kumari, Adak, Chandranath, Chattopadhyay, Soumi, Saqib, Muhammad, Anwar, Saeed

Abstract

Visual feedback in the form of user-uploaded images and videos is becoming increasingly common in e-commerce platforms because it provides authentic evidence of product quality, defects, packaging conditions, and real-world usage. However, visual feedback alone often lacks the contextual explanations and subjective opinions necessary for informed decision-making, while many users provide limited textual feedback due to the effort required to compose detailed reviews. To bridge this gap, we introduce image-grounded review assistance, a novel task that aims to generate editable review drafts from user-uploaded product images. Unlike conventional image captioning, which focuses on objective visual description, the proposed task requires product-specific understanding, sentiment estimation, and evidence-driven review composition under challenging real-world conditions, including degraded image quality, excessive zoom-in, target ambiguity, and partial product visibility. We propose a multi-agent vision-language framework consisting of four specialised roles: product grounding, visual sentiment estimation, visual evidence generation, and review synthesis. The framework employs explicit intermediate representations, including product entities, predicted ratings, and evidence summaries, to improve interpretability and visual grounding. Experiments on a curated subset of the Amazon Reviews Electronics dataset demonstrate the feasibility of generating coherent, product-aware, and sentiment-aware review drafts from visual feedback. To the best of our knowledge, this is the first study to formulate image-grounded review assistance as a multi-agent vision-language reasoning problem, providing a practical step toward AI-assisted review authoring in e-commerce systems.

Chinese Translation

用户上传的图像和视频形式的视觉反馈在电子商务平台上变得越来越普遍，因为它提供了产品质量、缺陷、包装状况和实际使用的真实证据。然而，仅靠视觉反馈往往缺乏做出明智决策所需的上下文解释和主观意见，而许多用户由于撰写详细评论所需的努力而提供有限的文本反馈。为了填补这一空白，我们提出了图像基础的评论辅助，这是一项新颖的任务，旨在从用户上传的产品图像生成可编辑的评论草稿。与传统的图像描述不同，后者侧重于客观的视觉描述，所提出的任务需要产品特定的理解、情感估计以及在挑战性现实条件下的证据驱动的评论撰写，包括图像质量下降、过度放大、目标模糊和部分产品可见性。我们提出了一种由四个专业角色组成的多智能体视觉-语言框架：产品定位、视觉情感估计、视觉证据生成和评论合成。该框架采用明确的中间表示，包括产品实体、预测评分和证据摘要，以提高可解释性和视觉定位。在经过精心挑选的亚马逊评论电子产品数据集的实验中，证明了从视觉反馈生成连贯的、产品感知的和情感感知的评论草稿的可行性。根据我们所知，这是首次将图像基础的评论辅助形式化为多智能体视觉-语言推理问题，为电子商务系统中的AI辅助评论撰写提供了一个实际的步骤。

View on arXiv Download PDF AI Translation

cs.CV / 131 / 2609.14790

Zero-shot video highlight detection based on text descriptions and synthetic images

基于文本描述和合成图像的零样本视频亮点检测

Byra, Michal, Presta, Alberto, Stefanski, Grzegorz, Arendt, Krzysztof

Abstract

Detecting video highlights, the most informative or engaging moments in a video, is important for applications such as video summarization and content recommendation. We propose a zero-shot framework that combines CLIP, large language models (LLMs), and diffusion models. Given lightweight video metadata, such as a title or category, an LLM generates textual descriptions of likely highlight events. These descriptions are further converted into synthetic visual prototypes using a diffusion model. Textual and visual representations are matched to video frames using CLIP, enabling frame-level highlight detection without highlight annotations or dataset-specific training. Experiments on TVSum and SumMe demonstrate strong zero-shot performance, with particularly favorable results on TVSum. The proposed approach provides an effective framework for metadata-conditioned zero-shot video highlight detection.

Chinese Translation

检测视频亮点，即视频中最具信息性或吸引力的时刻，对于视频摘要和内容推荐等应用至关重要。我们提出了一种零样本框架，结合了CLIP、大型语言模型（LLMs）和扩散模型。在给定轻量级视频元数据（如标题或类别）的情况下，LLM生成可能的亮点事件的文本描述。这些描述进一步通过扩散模型转换为合成视觉原型。使用CLIP将文本和视觉表示与视频帧进行匹配，从而实现帧级亮点检测，而无需亮点注释或特定数据集的训练。在TVSum和SumMe上的实验表明，零样本性能强劲，尤其在TVSum上表现尤为突出。所提出的方法为基于元数据的零样本视频亮点检测提供了有效的框架。

View on arXiv Download PDF AI Translation

cs.CV / 132 / 2609.14856

RAIN: Region-Aware Inversion Network for Semantic Watermark Extraction

RAIN：区域感知反演网络用于语义水印提取

Li, Zilai

Abstract

Semantic watermarks for diffusion models embed ownership information into the generative process while preserving perceptual quality, but Gaussian-Shading extraction conventionally requires multi-step diffusion inversion to recover the initial noise. Recent one-step methods show that this cost can be reduced substantially. We study this problem through extended flow matching and conditional regression. The key observation is that, near the high-SNR image endpoint, recovering a useful noise statistic given by the first-step output of the extended flow matching in the high-SNR regime is much simpler than reconstructing the full inverse trajectory, and Gaussian Shading only requires the recovered latent to remain in the correct watermark decision region. Based on this observation, we propose a lightweight, prompt-free extractor that decomposes endpoint recovery into an image-like anchor and a noise-oriented residual, which increases the capability of the model to utilize GPU parallel computation. The resulting method avoids iterative inversion and repeated evaluation of a diffusion-scale U-Net, providing an efficient one-step extraction pipeline with a concise theoretical interpretation. The computational cost of extracting noise is lower than that of both OSI and FARI. The github repo is there: https://github.com/TheLovesOfLadyPurple/RAIN-lightweight-NN-for-one-step-semantic-watermark-extraction

Chinese Translation

语义水印用于扩散模型，将所有权信息嵌入生成过程，同时保持感知质量，但高斯阴影提取通常需要多步扩散反演以恢复初始噪声。最近的一步法表明，这一成本可以大幅降低。我们通过扩展流匹配和条件回归研究这个问题。关键观察是，在高信噪比（SNR）图像端点附近，恢复由扩展流匹配在高SNR范围内的第一步输出给出的有用噪声统计量，比重建完整的逆轨迹要简单得多，而高斯阴影仅要求恢复的潜变量保持在正确的水印决策区域。基于这一观察，我们提出了一种轻量级、无提示的提取器，将端点恢复分解为图像样的锚点和噪声导向的残差，从而增强模型利用GPU并行计算的能力。该方法避免了迭代反演和对扩散规模U-Net的重复评估，提供了一个高效的一步提取管道，并具有简洁的理论解释。提取噪声的计算成本低于OSI和FARI两者。相关的GitHub仓库地址为：https://github.com/TheLovesOfLadyPurple/RAIN-lightweight-NN-for-one-step-semantic-watermark-extraction

View on arXiv Download PDF AI Translation

cs.CV / 133 / 2609.14899

What Makes a 3D Scene Editable? A Factorized Benchmark of Fidelity, Locality, Consistency, and Preservation

什么使三维场景可编辑？一个关于保真度、局部性、一致性和保留的分解基准

Patro, Sariah, Mehra, Arjun, Bhatia, Nikhil

Abstract

Neural 3D scene editing is often evaluated by semantic alignment alone, although a convincing result may alter unrelated content or become inconsistent across views. We introduce EditBench3D, a representation-agnostic benchmark that treats editing as controlled information replacement. It evaluates four complementary properties: instruction fidelity, spatial locality, cross-view consistency, and preservation of non-target content. The protocol combines visibility-aware 3D target supports, paired descriptions, held-out cameras, and five edit families covering appearance, material, geometry, and object-level changes. We evaluate eight representative NeRF, 3D Gaussian Splatting, hybrid, and proxy-based editors on 240 scene-edit pairs. The study shows that semantic fidelity is only weakly associated with the other editing properties, and that no single method is optimal across all dimensions. Explicit Gaussian editors offer a strong overall balance, whereas direct proxy manipulation provides the most conservative edits at the cost of open-ended fidelity. These findings support reporting editability as a multi-objective profile rather than a single semantic score.

Chinese Translation

神经三维场景编辑通常仅通过语义对齐进行评估，尽管一个令人信服的结果可能会改变无关内容或在不同视图之间变得不一致。我们引入了 EditBench3D，一个与表示无关的基准，将编辑视为受控信息替换。它评估四个互补属性：指令保真度、空间局部性、跨视图一致性和非目标内容的保留。该协议结合了考虑可见性的三维目标支持、配对描述、保留相机和五种编辑类别，涵盖外观、材料、几何和对象级变化。我们在240对场景编辑上评估了八种代表性的 NeRF、3D 高斯点云、混合和基于代理的编辑器。研究表明，语义保真度与其他编辑属性之间的关联较弱，并且没有单一方法在所有维度上都是最佳的。显式高斯编辑器提供了良好的整体平衡，而直接代理操作则以开放式保真度为代价提供了最保守的编辑。这些发现支持将可编辑性报告为多目标特征，而不是单一的语义分数。

View on arXiv Download PDF AI Translation

cs.CV / 134 / 2609.14943

ThreshGuide: Class-Aware Labeled-Guided Thresholding for Semi-Supervised 3D Abdominal Multi-Organ Segmentation

ThreshGuide：一种基于类别感知的标签引导阈值方法用于半监督3D腹部多脏器分割

Liu, Hongyu, Wang, Yinlong, Li, Lusha, Meng, Hui

Abstract

Pseudo-labeling is a strong paradigm for semi-supervised medical image segmentation, yet its effectiveness is highly sensitive to confidence thresholding. In abdominal multi-organ segmentation, a fixed global threshold is particularly suboptimal because organ classes differ substantially in size, appearance, and learning difficulty. In this work, we propose ThreshGuide, a class-aware threshold adaptation framework that uses labeled data to guide pseudo-label selection on unlabeled data. Built upon a standard teacher-student architecture, the teacher model evaluates labeled samples during training to estimate class-aware threshold targets by maximizing an error-aware F\b{eta} criterion that balances precision and coverage. These targets are then smoothed with an exponential moving average (EMA) and used to filter unlabeled voxels in a class-dependent manner. Experiments on FLARE2022 and AMOS2022 show that ThreshGuide performs competitively overall, yielding clear improvements specifically on hard-to-learn organs.

Chinese Translation

伪标签生成是一种强有力的半监督医学图像分割范式，但其有效性对置信度阈值的选择高度敏感。在腹部多脏器分割中，固定的全局阈值尤其不理想，因为不同器官类别在大小、外观和学习难度上存在显著差异。在本研究中，我们提出了ThreshGuide，这是一种类别感知的阈值适应框架，利用标记数据指导对未标记数据的伪标签选择。该方法基于标准的教师-学生架构，教师模型在训练过程中评估标记样本，通过最大化一个平衡精度和覆盖率的错误感知F{eta}标准来估计类别感知的阈值目标。这些目标随后通过指数移动平均（EMA）进行平滑，并以类别依赖的方式用于过滤未标记的体素。在FLARE2022和AMOS2022上的实验表明，ThreshGuide整体表现具有竞争力，特别是在难以学习的器官上取得了明显的改进。

View on arXiv Download PDF AI Translation

cs.CV / 135 / 2609.14965

MoVT: Video-Augmented Motion Tokenizer for Text-to-Motion Generation

MoVT：用于文本到动作生成的视频增强运动标记器

Jing, Beibei, Guo, Tianle, Zhang, Youjia, Song, Zikai, Luo, Yawei, Yu, Junqing, Guan, Tao, Yang, Wei

Abstract

Text-driven 3D human motion generation models face significant challenges in responding to diverse and unconstrained textual prompts, primarily due to the limited availability of 3D motion training data. To address this, we introduce MoVT, a novel framework that effectively leverages the extensive range of human action videos to enhance text-to-motion generation. At the core of our approach is the cross-modal augmented motion tokenizer, which projects discrete 3D motion tokens into the 2D domain. This projection allows us to enrich the motion codebook with complex, real-world motion patterns derived from videos. The enriched discrete tokens are then mapped back to the 3D domain, resulting in aligned 3D and 2D codebooks with an enhanced capacity to represent intricate motions. These enhanced codebooks are integrated into a generative masked transformer, which predicts masked motion token indices in a modality-agnostic manner. This enables the use of text-index pairs, generated from the 2D codebook and annotated motion videos, to further enhance the generator. Extensive empirical evaluations show that MoVT performs favorably against prior state-of-the-art methods across multiple key metrics.

Chinese Translation

基于文本的三维人类动作生成模型在应对多样化和无约束的文本提示时面临重大挑战，主要原因是三维动作训练数据的有限可用性。为了解决这一问题，我们提出了MoVT，一个新颖的框架，能够有效利用丰富的人类动作视频来增强文本到动作的生成。我们方法的核心是跨模态增强运动标记器，它将离散的三维动作标记投影到二维域中。这一投影使我们能够用来自视频的复杂现实世界动作模式丰富动作代码本。丰富的离散标记随后被映射回三维域，从而形成对齐的三维和二维代码本，增强了表示复杂动作的能力。这些增强的代码本被集成到生成式掩蔽变换器中，该变换器以模态无关的方式预测掩蔽的动作标记索引。这使得可以使用从二维代码本和注释动作视频生成的文本-索引对，进一步增强生成器。广泛的实证评估表明，MoVT在多个关键指标上相较于之前的最先进方法表现优越。

View on arXiv Download PDF AI Translation

cs.CV / 136 / 2609.14973

PhysBrain 1.5: From Vision-Language Models to Physical Foundation Models

PhysBrain 1.5：从视觉-语言模型到物理基础模型

DeepCybo Team, Bin, Yu, Cao, Haipeng, Chang, Zheng, Chen, Kai, Chen, Youning, Deng, Kailin, Du, Yichao, Fu, Xiaotong, Ge, Haoyang, Guo, Yunlong, Hao, Chenliu, He, Jiyan, He, Xuguo, Hou, Yakun, Hu, Kai, Huang, Cong, Huang, Tuopusen, Huang, Yu, Li, Hong, Li, Peize, Lian, Shijie, Lin, Xiaopeng, Lin, Yun, Liu, Haibao, Liu, Haochen, Liu, Qiuzhi, Liu, Shengcai, Liu, Zhiqiang, Luo, Tao, Ren, Peng, Ren, Shuo, Ruan, Chaoyi, Shen, Zhaolong, Shi, Yukun, Su, Qiyuan, Tian, Yuxuan, Wang, Yining, Wu, Changti, Wu, Hao, Xu, Xueyin, Yang, Ruoqi, Yang, Zhaoyang, Yuan, Hang, Zeng, Zhaoyang, Zhang, Hanwen, Zhang, Ruimeng, Zhang, Yao, Zhang, Yibo, Zhang, Yuxiang, Zhang, Zhirui, Zhang, Ziyi, Zheng, Zubin, Zhuang, Zishen

Abstract

We present PhysBrain 1.5, a unified model for understanding physical environments, generating actions, and predicting future states. Motivated by the physical loop of observation, interaction, and environmental change, we bring these capabilities into a common learning framework. Starting from a general vision--language model, we encode language responses, end-effector motion, and dense visual targets as discrete sequences and jointly optimize them with autoregressive next-token prediction. Pre-training draws its embodied supervision entirely from human interaction videos, using task-centered episodes to pair semantic and spatial context with recovered motion and subsequent observations. We then adapt the model through supervised fine-tuning on a mixture of human demonstrations, robot trajectories, and simulated experience. Across 28 embodied understanding benchmarks, our 8B model achieves an average score of 72.5, setting a new open-source state of the art and performing on par with leading proprietary models such as GPT-6-Astra and Gemini 3.6 Flash. It achieves the best open-source results on 14 benchmarks while retaining general multimodal capabilities. Beyond these understanding evaluations, qualitative examples show the model's ability to produce end-effector trajectories and predict future scenes through spatially aligned RGB, depth, and robot-mask outputs.

Chinese Translation

我们提出了PhysBrain 1.5，这是一个统一的模型，用于理解物理环境、生成动作和预测未来状态。受到观察、互动和环境变化的物理循环的启发，我们将这些能力整合到一个共同的学习框架中。从一个通用的视觉-语言模型出发，我们将语言响应、末端执行器运动和密集视觉目标编码为离散序列，并通过自回归的下一个标记预测进行联合优化。预训练完全依赖于人类互动视频中的体现监督，使用以任务为中心的情节将语义和空间上下文与恢复的运动和后续观察配对。随后，我们通过对人类示范、机器人轨迹和模拟经验的混合进行监督微调来适应模型。在28个体现理解基准测试中，我们的8B模型平均得分为72.5，创造了新的开源最佳成绩，并与领先的专有模型如GPT-6-Astra和Gemini 3.6 Flash表现相当。它在14个基准测试中取得了最佳的开源结果，同时保持了通用的多模态能力。除了这些理解评估，定性示例展示了模型生成末端执行器轨迹和通过空间对齐的RGB、深度和机器人掩码输出预测未来场景的能力。

View on arXiv Download PDF AI Translation

cs.CV / 137 / 2609.15018

G-ray: Ray-Level Relative Geometric Position Encoding in Multi-View Vision Transformers under Camera Heterogeneity

G-ray：在相机异质性下多视角视觉变换器中的光线级相对几何位置编码

Zhang, Shuo, Su, Xin, Wang, Wei, Liu, Jun, Zeng, Xinrui, Chen, Yongsen, Wang, Chenjie, Zhu, Guibo, Wang, Jinqiao, Luo, Bin, Zhang, Liangpei

Abstract

We study relative position encoding for multi-view vision Transformers under camera heterogeneity, including varying fields of view (FoVs) or projection models. Existing rotary relative position encodings commonly use image-plane positional coordinates, producing projection-dependent relative phases and inconsistent geometric cues for cross-projection attention. We introduce G-ray, a ray-level relative position encoding whose rotary phases are parameterized by camera-local ray angles. The same camera-local ray pair induces the same relative phase across projections, providing projection-invariant positional consistency. G-ray can be used directly or integrated with existing encodings, retaining complementary geometric cues without additional learned parameters. We validate G-ray in three host encodings, RoPE, GTA, and RayRoPE, across 3D reconstruction and novel-view synthesis (NVS). Across three heterogeneous 3D reconstruction benchmarks at 50 views, G-ray leads all six averaged metrics and reduces mean pointmap relative error by 45.8% over MapAnything, with calibration supplied to both. Trained exclusively on homogeneous pinhole images, the 3D reconstruction model handles mixed pinhole and non-pinhole inputs without retraining and remains competitive on homogeneous pinhole 3D reconstruction protocols. For NVS, GTA and RayRoPE improve with G-ray under joint viewpoint and FoV variation. The project's webpage is available at https://g-ray-project.github.io/.

Chinese Translation

我们研究了在相机异质性下多视角视觉变换器的相对位置编码，包括视场（FoV）或投影模型的变化。现有的旋转相对位置编码通常使用图像平面位置坐标，导致依赖于投影的相对相位和跨投影注意力的不一致几何线索。我们提出了G-ray，一种光线级相对位置编码，其旋转相位由相机局部光线角度参数化。相同的相机局部光线对在不同投影中引发相同的相对相位，从而提供投影不变的位置一致性。G-ray可以直接使用或与现有编码集成，保留互补的几何线索而无需额外的学习参数。我们在三种主编码RoPE、GTA和RayRoPE中验证了G-ray，应用于3D重建和新视图合成（NVS）。在50个视角的三种异质3D重建基准测试中，G-ray在所有六个平均指标中领先，并将平均点图相对误差降低了45.8%，相较于MapAnything，且两者均提供了标定。该3D重建模型仅在同质针孔图像上训练，能够处理混合的针孔和非针孔输入而无需重新训练，并在同质针孔3D重建协议中保持竞争力。对于NVS，GTA和RayRoPE在视点和视场变化下与G-ray结合后得到了改善。项目网页可访问 https://g-ray-project.github.io/。

View on arXiv Download PDF AI Translation

cs.CV / 138 / 2609.15028

TwinICL: Diagnosing Multimodal In-Context Learning through Paired Counterfactuals

TwinICL：通过配对反事实诊断多模态上下文学习

Xue, Zihan, Lu, Po-Yi, Honcharenko, Serhii, Chen, Zih-Ching, Lin, Hsuan-Tien, Peng, Nanyun, Hsu, I-Hung, Huang, Kuan-Hao

Abstract

In-context learning (ICL) enables models to infer tasks from demonstrations, but existing benchmarks generally lack matched text and image versions needed to compare ICL performance across modalities. We introduce TwinICL, a procedurally generated benchmark providing such pairs for controlled comparison. Across six open-weight models and 38 tasks, multimodal ICL consistently underperforms text-only ICL, with gaps varying by task family. To test whether this gap can be recovered, we target visual access, task framing, and reasoning through three interventions. Their combination recovers strong multimodal ICL performance on a diagnostic subset, despite limited or inconsistent individual effects. To distinguish difficulties in executing tasks from those in inferring them, we evaluate models with explicit task instructions, revealing a modality gap even when the task is known. We then examine how adding demonstration inputs and outputs reshapes this gap, highlighting demonstrations' dual role as additional context to process and evidence about the task. The dataset is available at https://github.com/lab-flair/TwinICL.

Chinese Translation

上下文学习（ICL）使模型能够从示例中推断任务，但现有基准通常缺乏比较多模态 ICL 性能所需的匹配文本和图像版本。我们引入了 TwinICL，这是一个程序生成的基准，提供了用于控制比较的这样的配对。在六个开放权重模型和 38 个任务中，多模态 ICL 的表现始终低于仅文本的 ICL，差距因任务类别而异。为了测试是否可以弥补这一差距，我们针对视觉访问、任务框架和推理进行了三项干预。它们的组合在一个诊断子集上恢复了强大的多模态 ICL 性能，尽管单独的效果有限或不一致。为了区分执行任务的困难与推断任务的困难，我们对模型进行了显式任务指令的评估，即使在任务已知的情况下也揭示了模态差距。然后，我们考察了添加示例输入和输出如何重塑这一差距，突显了示例在处理和任务证据方面的双重角色。数据集可在 https://github.com/lab-flair/TwinICL 获取。

View on arXiv Download PDF AI Translation

cs.CV / 139 / 2609.15032

Tele360: Real-Time Feed-Forward Human Reconstruction from Sparse Unposed Cameras

Tele360：基于稀疏无姿态相机的实时前馈人类重建

Tu, Hanzhang, Liao, Zhanfeng, Min, Wei, Zhang, Jiajun, Liu, Yebin

Abstract

Live free-viewpoint visualization of real humans is critical for immersive communication and interactive digital experiences. Existing methods either rely on computationally expensive optimization or require calibrated cameras and low-resolution inputs, making real-time high-resolution deployment impractical. In this work, we present Tele360, the first real-time feed-forward system for dynamic human reconstruction and live free-viewpoint visualization from sparse, unposed RGB streams. Our system jointly estimates camera poses and reconstructs a dynamic 3D Gaussian representation for each time instance in a single forward pass. To achieve this, we start by designing a lightweight sparsity-aware multi-view transformer backbone that tokenizes foreground human regions while preserving global context through a shared scene token. We then employ a fully transformer-based Gaussian decoder to mitigate convolution-induced over-smoothing while keeping decoding sparse and efficient. In addition, we introduce a hybrid feature pyramid that injects multi-scale appearance cues into geometry prediction. We further introduce a lightweight differentiable Levenberg-Marquardt camera refinement layer to enhance multi-view consistency and geometric alignment. Moreover, to stabilize learning under sparse, unposed inputs, we transfer multi-view geometry priors from a large visual-geometry foundation model via teacher-student distillation. Finally, the predicted Gaussian maps are streamed with video codecs to remote devices for interactive free-viewpoint rendering. Extensive experiments show that Tele360 achieves state-of-the-art visual quality on studio benchmarks while supporting real-time 2K input-to-rendering at over 25 FPS on a single consumer GPU. Additional captured sequences illustrate its performance across varied subjects, clothing, and motions under our multi-camera setup.

Chinese Translation

实时自由视角的真实人类可视化对于沉浸式沟通和互动数字体验至关重要。现有方法要么依赖于计算成本高昂的优化，要么需要经过校准的相机和低分辨率输入，这使得实时高分辨率部署变得不切实际。在本研究中，我们提出了Tele360，这是第一个用于动态人类重建和从稀疏无姿态RGB流进行实时自由视角可视化的前馈系统。我们的系统在单次前向传递中联合估计相机姿态并重建每个时间实例的动态3D高斯表示。为此，我们首先设计了一种轻量级的稀疏感知多视角变换器骨干网络，该网络在保留全局上下文的同时对前景人类区域进行标记。然后，我们采用完全基于变换器的高斯解码器，以减轻卷积引起的过度平滑，同时保持解码的稀疏性和高效性。此外，我们引入了一种混合特征金字塔，将多尺度外观线索注入几何预测中。我们进一步引入了一种轻量级可微分的Levenberg-Marquardt相机精细化层，以增强多视角一致性和几何对齐。此外，为了在稀疏无姿态输入下稳定学习，我们通过教师-学生蒸馏从大型视觉几何基础模型中转移多视角几何先验。最后，预测的高斯图通过视频编解码器流式传输到远程设备，以实现交互式自由视角渲染。大量实验表明，Tele360在工作室基准测试中实现了最先进的视觉质量，同时在单个消费级GPU上支持实时2K输入到渲染，帧率超过25 FPS。额外捕获的序列展示了其在我们多相机设置下对不同对象、服装和动作的表现。

View on arXiv Download PDF AI Translation

cs.CV / 140 / 2609.15098

LG-VLN: A Zero-Shot Vision-and-Language Navigation Framework with LangGraph State Orchestration

LG-VLN：一种基于LangGraph状态协调的零-shot视觉与语言导航框架

Zhao, Jianhe, Qiu, Yanhua, Zhang, Zhiyu, Zhao, Zibo, Xie, Jinhua

Abstract

Continuous-environment vision-and-language navigation (VLN-CE) requires interpreting natural-language instructions in unseen 3D environments and executing continuous low-level actions. Existing methods often depend on LiDAR, panoramic cameras, or extra sensors; separate geometric-mapping and semantic-navigation visual representations can cause long-trajectory spatial-semantic inconsistencies. We propose LG-VLN, a monocular zero-shot framework with shared visual features and LangGraph-based state orchestration. An online feed-forward 3D reconstruction network predicts depth, camera poses, and dense point clouds for agent-pose estimation and global map fusion. Geometry and navigation share dense CleanDIFT features: semantic consistency rejects incorrect inter-frame correspondences, while target-instance constraints define visual references whose similarity combines with local BLIP-2 image-text relevance to form a semantic value map. LangGraph represents instruction parsing, geometric perception, semantic value updates, path planning, action execution, and failure recovery as a directed state graph with conditional transitions, persistent state, and modular recovery mechanisms. On a fixed 550-episode subset of the R2R-CE val-unseen split, LG-VLN achieves 21.3% success and 12.1% success weighted by path length. Ablations show shared semantic features improve navigation, further boosted by combining visual similarity and image-text relevance. Results establish shared visual representations and explicit state orchestration as effective for zero-shot VLN-CE using monocular RGB alone. Code will be publicly released for reproducibility.

Chinese Translation

连续环境视觉与语言导航（VLN-CE）需要在未见过的3D环境中解读自然语言指令并执行连续的低级动作。现有方法通常依赖于激光雷达（LiDAR）、全景相机或额外传感器；分离的几何映射和语义导航视觉表示可能导致长轨迹的空间语义不一致。我们提出了LG-VLN，这是一种具有共享视觉特征和基于LangGraph的状态协调的单目零-shot框架。一个在线前馈3D重建网络预测深度、相机姿态和稠密点云，以进行代理姿态估计和全局地图融合。几何和导航共享稠密的CleanDIFT特征：语义一致性拒绝不正确的帧间对应关系，而目标实例约束定义了视觉参考，其相似性与局部BLIP-2图像-文本相关性结合形成语义值图。LangGraph将指令解析、几何感知、语义值更新、路径规划、动作执行和故障恢复表示为一个具有条件转移、持久状态和模块化恢复机制的有向状态图。在R2R-CE val-unseen分割的固定550集子集上，LG-VLN实现了21.3%的成功率和12.1%的按路径长度加权的成功率。消融实验表明，共享语义特征提高了导航效果，结合视觉相似性和图像-文本相关性进一步提升了性能。结果表明，使用单目RGB的零-shot VLN-CE有效地建立了共享视觉表示和明确的状态协调。代码将公开发布以便于复现。

View on arXiv Download PDF AI Translation

cs.CV / 141 / 2609.15100

ChatGPT Images 2.5 in the Wild: A Launch-Period Dataset and Detector Evaluation

ChatGPT Images 2.5 在实际应用中的表现：启动期数据集与检测器评估

Ng, Dennis, Shen, Xingyu, Raj, Ankit, Zewde, Kidus, Duong, Tommy, Zhou, Yuchen, Zhang, Yuxin, Tiangratanakul, Neo, Ren, Simiao

Abstract

An image tool can change its underlying generator while retaining its public name, making version attribution from online posts ambiguous. We study this problem after the ChatGPT Images 2.5 launch. Our frozen collection contains 3,478 images from 2,440 posts across 8 sources. Recorded posting times fall within the first 51.1 hours after the announcement. It records three attribution tiers and retains standalone images after image-form filtering and targeted review. Caption claims and host records provide admission evidence, not independently verified generator identity. The observed content profile depends on the source mixture: NightCafe supplies 39.0% of images but 77.0% of CLIP-assigned fantasy scenes. We then evaluate six frozen detectors at thresholds calibrated to a 5% flag rate on reference photographs. Collection flag rates range from 3.7 to 56.4%, falling 42-81 percentage points below GenImage recall. Held-out artwork false-positive rates range from 1.5 to 96.5%, so a higher collection flag rate does not by itself establish better detection. An exploratory X-only comparison with our April collection finds a higher September flag rate for Effort, and a suggestive difference for DoU, under fixed-threshold post-clustered bootstrap intervals. Attribution, content and processing differences prevent a causal interpretation of these contrasts. The collection supports analysis of reported model use during a product transition, with source and attribution evidence retained for interpretation.

Chinese Translation

图像工具可以在保持其公共名称的同时更改其底层生成器，这使得从在线帖子中进行版本归属变得模糊。我们在 ChatGPT Images 2.5 发布后研究了这个问题。我们的冻结数据集包含来自 8 个来源的 3,478 张图像，涉及 2,440 个帖子。记录的发布时间在公告后的前 51.1 小时内。该数据集记录了三个归属层次，并在图像形式过滤和针对性审核后保留了独立图像。标题声明和主机记录提供了入场证据，但并未独立验证生成器身份。观察到的内容特征依赖于来源组合：NightCafe 提供了 39.0% 的图像，但 77.0% 的 CLIP 指定的幻想场景。随后，我们在针对参考照片的 5% 标记率校准阈值下评估了六个冻结检测器。数据集的标记率范围从 3.7% 到 56.4%，比 GenImage 的召回率低 42-81 个百分点。保留的艺术作品误报率范围从 1.5% 到 96.5%，因此更高的集合标记率本身并不能确立更好的检测效果。与我们四月数据集的探索性 X-only 比较发现，九月的 Effort 标记率更高，而在固定阈值后聚类自助区间下，DoU 也显示出提示性差异。归属、内容和处理差异阻止了对这些对比的因果解释。该数据集支持在产品过渡期间对报告模型使用情况的分析，并保留了来源和归属证据以供解释。

View on arXiv Download PDF AI Translation

cs.CV / 142 / 2609.15120

DNF-SR: Dual-Input and Negative-Aware Feature Fine-Tuning for Real-World Image Super-Resolution

DNF-SR：针对真实世界图像超分辨率的双输入和负感知特征微调

Han, Shuhao, Liao, Wenjie, Vance, Hayden, Dong, Hang, Zhang, Rui, Guo, Chun-Le, Li, Chongyi

Abstract

Benefiting from the powerful generative priors of diffusion models, diffusion-based real-world image super-resolution (Real-ISR) methods have demonstrated impressive performance.To achieve efficient Real-ISR, several recent works have designed one-step diffusion-based models.Howerver, unmediatedly feeding LR into a diffusion model creates a distributional gap with the model's original input.A straightforward approach to reduce the distribution gap is to introduce noise to the LR latents. However, directly adding noise inevitably corrupts the content of the LR images.In this study, we propose DNF-SR, a Dual-input and Negative-aware Feature fine-tuning method for Real-ISR.Specifically, we use a dual-input strategy that concatenates the original LR image with the noisy LR input and feeds them into a diffusion-based image editing model, ensuring both high-fidelity one-step super-resolution and improved perceptual and content consistency.Additionally, the noise present in the noisy LR input introduces randomness and diversity into the outputs. We exploit this property and propose a post-training optimization method, Negative-aware Feature Fine-Tuning (NF2T), which guides the model toward producing higher-quality results.NF^2T classifies multiple outputs into positive and negative subsets and then defines implicit policy improvement directions in both the image and feature spaces, thereby further enhancing the stability of the optimization.Extensive experiments show that DNF-SR outperforms other methods.Code will be released.

Chinese Translation

得益于扩散模型强大的生成先验，基于扩散的真实世界图像超分辨率（Real-ISR）方法展现了令人印象深刻的性能。为了实现高效的Real-ISR，最近的一些研究设计了一步扩散模型。然而，直接将低分辨率（LR）图像输入到扩散模型中会导致与模型原始输入之间的分布差距。减少分布差距的一个简单方法是向LR潜变量中引入噪声。然而，直接添加噪声不可避免地会破坏LR图像的内容。在本研究中，我们提出了DNF-SR，一种用于Real-ISR的双输入和负感知特征微调方法。具体而言，我们采用双输入策略，将原始LR图像与带噪LR输入进行拼接，并将其输入到基于扩散的图像编辑模型中，从而确保高保真度的一步超分辨率以及改善的感知和内容一致性。此外，带噪LR输入中的噪声为输出引入了随机性和多样性。我们利用这一特性，提出了一种后训练优化方法，负感知特征微调（Negative-aware Feature Fine-Tuning，NF2T），该方法引导模型生成更高质量的结果。NF^2T将多个输出分类为正子集和负子集，然后在图像和特征空间中定义隐式策略改进方向，从而进一步增强优化的稳定性。大量实验表明，DNF-SR优于其他方法。代码将会发布。

View on arXiv Download PDF AI Translation

cs.CV / 143 / 2609.15131

AdaVSkip: Adaptive Visual Token Skipping Across Layers For Efficient MLLMs Inference

AdaVSkip：跨层自适应视觉标记跳过以提高多模态大语言模型推理效率

Sun, Yuyao, Deng, Tao, Li, Shuang, Wang, Deqing

Abstract

Multimodal large language models (MLLMs) require substantial computation to process numerous visual tokens across all transformer layers. Most methods for efficient MLLM inference exploit horizontal redundancy by compressing visual tokens. Beyond token reduction, recent studies exploit vertical redundancy through early exit or fixed-layer skipping. However, we find that the extent and distribution of this redundancy vary across inputs and differ between self-attention and MLP modules. Motivated by these observations, we propose AdaVSkip, which equips each layer with two lightweight routers that independently determine whether visual tokens pass through by or skip the self-attention and MLP modules. These decisions collectively define an input-specific visual-computation path, but their discrete and non-differentiable nature makes learning effective paths challenging. To address this challenge, we develop a progressive two-stage training framework that updates only the routers while keeping the backbone frozen. Stage I establishes an initial routing policy through supervised training with input-specific targets derived from module-wise necessity scores. To further align the routing policy with task performance, Stage II uses reinforcement learning to optimize routing decisions with direct feedback from generated answers. It combines an answer correctness reward with a skip-consistency reward that discourages excessive retention of visual-token computation. Across three MLLM backbones, AdaVSkip maintains strong task performance with substantially less computation. On LLaVA-NeXT-7B, AdaVSkip reduces FLOPs by 53.2\% while preserving the original model's average performance. Combining it with visual token compression increases this reduction to 91.2\%, while retaining 97.2\% of the original performance on average.

Chinese Translation

多模态大语言模型（MLLMs）在处理所有变换器层中的大量视觉标记时需要大量计算。大多数高效的MLLM推理方法通过压缩视觉标记来利用水平冗余。除了标记减少，最近的研究还通过提前退出或固定层跳过来利用垂直冗余。然而，我们发现这种冗余的程度和分布在不同输入之间存在差异，并且在自注意力和多层感知器（MLP）模块之间也有所不同。基于这些观察，我们提出了AdaVSkip，该方法为每一层配备了两个轻量级路由器，独立决定视觉标记是通过自注意力和MLP模块还是跳过它们。这些决策共同定义了一个特定于输入的视觉计算路径，但其离散和不可微的特性使得学习有效路径变得具有挑战性。为了解决这个问题，我们开发了一个渐进式的两阶段训练框架，该框架在保持主干网络不变的情况下，仅更新路由器。第一阶段通过监督训练建立初始路由策略，使用从模块必要性评分派生的特定于输入的目标。为了进一步将路由策略与任务性能对齐，第二阶段使用强化学习优化路由决策，并从生成的答案中获得直接反馈。它结合了答案正确性奖励和跳过一致性奖励，以抑制过度保留视觉标记计算。在三个MLLM主干网络上，AdaVSkip在显著减少计算的同时保持强大的任务性能。在LLaVA-NeXT-7B上，AdaVSkip将FLOPs减少了53.2\%，同时保持了原始模型的平均性能。将其与视觉标记压缩相结合，进一步将这一减少提升至91.2\, 同时在平均上保留了97.2\%的原始性能。

View on arXiv Download PDF AI Translation

cs.CV / 144 / 2609.15137

SparseTalk - Sparsifying 3D Gaussian Language Fields for Efficient 3D Visual Question Answering

SparseTalk - 稀疏化3D高斯语言场以实现高效的3D视觉问答

Soselia, Davit, JaJa, Joseph, Varshney, Amitabh

Abstract

3D Gaussian language fields provide an explicit, spatially grounded representation for 3D visual question answering (VQA), but their dense semantic features can require tens of thousands of embeddings per scene, resulting in substantial storage, memory, and inference costs. We investigate how much of this representation is actually necessary for downstream reasoning. Starting from a full embedding representation, we systematically sparsify its semantic embeddings, including the previously underexplored regime below a single image-equivalent block down to 8 visual tokens. We compare random, geometric, semantic, and joint spatial-semantic selection strategies and introduce an object-based sparsification method that distributes the token budget across detected object instances while retaining background context. Experiments on ScanQA and MV-ScanQA reveal substantial redundancy in dense Gaussian language fields. Strong VQA performance is retained with only a few hundred semantic embeddings, corresponding to less than 1% of the original representation. Object-based selection performs well relative to others, with only modest observed changes down to 256 tokens. At this budget, SparseTalk retains 0.80% of SplatTalk's 32,076-token inference input and 0.332% of the mean 77,207-Gaussian dense field, increasing inference throughput while reducing decoded-feature memory 125-fold.

Chinese Translation

3D高斯语言场为3D视觉问答（VQA）提供了一种明确的、空间基础的表示，但其密集的语义特征可能需要每个场景数万个嵌入，导致存储、内存和推理成本显著增加。我们研究了这种表示在下游推理中实际上需要多少。从完整的嵌入表示出发，我们系统地稀疏化其语义嵌入，包括之前未充分探索的低于单个图像等效块的范围，减少到8个视觉标记。我们比较了随机、几何、语义和联合空间-语义选择策略，并引入了一种基于对象的稀疏化方法，该方法在保留背景上下文的同时，将标记预算分配到检测到的对象实例上。在ScanQA和MV-ScanQA上的实验揭示了密集高斯语言场中的显著冗余。仅使用几百个语义嵌入即可保持强大的VQA性能，这对应于原始表示的不到1%。相较于其他方法，基于对象的选择表现良好，观察到的变化仅在256个标记时略微增加。在这个预算下，SparseTalk保留了SplatTalk的32,076个标记推理输入的0.80%和平均77,207个高斯密集场的0.332%，在提高推理吞吐量的同时将解码特征内存减少了125倍。

View on arXiv Download PDF AI Translation

cs.CV / 145 / 2609.15144

Automated Perceptually-Motivated Assessment of Photographic Consistency in Paired Clinical Photographs: Pipeline Development and Internal Evaluation

自动化感知驱动的临床照片一致性评估：管道开发与内部评估

Lin, Derrick, Rabinovich, Samantha, Rabinovich, Joclin, Garoosi, Kassra, Khetpal, Sumun, Delanoy, Evan, Bhardwaj, Neel, Roostaeian, Jason

Abstract

Purpose: Paired pre- and post-operative photographs are the standard unit of evidence for plastic surgical outcomes, yet no objective metric verifies whether two images of the same patient were captured under conditions consistent for comparison. Approach: We developed a perceptually motivated pipeline that analyzes pre/post pairs across thirteen calibrated sub-metrics, partitioned by unsupervised correlation-structure analysis into five data-driven clusters (photometric, texture / sharpness, pose, illumination direction, and pitch), averaged within each cluster and combined across clusters by a weighted sum into a single consistency score. Each sub-metric is calibrated so that its median difference across published within-patient pairs scores 0.5, which is a reference point and carries no pass/fail meaning. The pipeline was calibrated on 134 matched within-patient published pre/post pairs and evaluated against identical-image pairs, synthetic-perturbation pairs, and 134 mismatched cross-publication pairs. Results: The master consistency score S separated matched from mismatched pairs (sensitivity index d' = 2.15, 95% confidence interval (CI) [1.83, 2.55]; area under the receiver operating characteristic curve AUC = 0.928, 95% CI [0.896, 0.959]), closely matching Gaussian-equal-variance predictions. The three head-pose angles did not fall in one cluster: yaw and roll grouped together while pitch separated. Identical pairs scored at ceiling (S = 0.99) and the master score fell monotonically with perturbation magnitude on all five perturbation axes. Conclusions: The score quantifies photographic comparability, not aesthetic or surgical quality, and provides a freely available web tool for auditing the photographic comparability of pre/post pairs, pending validation against expert judgment.

Chinese Translation

目的：成对的术前和术后照片是塑料外科结果的标准证据单位，但目前没有客观指标验证同一患者的两张图像是否在可比条件下拍摄。方法：我们开发了一种感知驱动的管道，通过十三个经过校准的子指标分析术前/术后成对照片，利用无监督相关结构分析将其划分为五个数据驱动的聚类（光度、纹理/清晰度、姿势、照明方向和俯仰），在每个聚类内取平均，并通过加权总和结合跨聚类形成单一一致性得分。每个子指标经过校准，使其在已发表的患者内成对照片中的中位差为0.5，作为参考点，并不具有通过/未通过的意义。该管道在134对匹配的患者内已发表的术前/术后成对照片上进行了校准，并与相同图像对、合成扰动对以及134对不匹配的跨出版物对进行了评估。结果：主一致性得分S能够区分匹配与不匹配的对（敏感性指数d' = 2.15，95%置信区间（CI）[1.83, 2.55]；接收者操作特征曲线下面积AUC = 0.928，95% CI [0.896, 0.959]），与高斯等方差预测结果高度一致。三个头部姿势角度未落入同一聚类：偏航和滚转聚在一起，而俯仰则分开。相同的图像对得分达到上限（S = 0.99），主得分在所有五个扰动轴上随着扰动幅度单调下降。结论：该得分量化了照片的可比性，而非美学或外科质量，并提供了一个可自由使用的网络工具，用于审计术前/术后成对照片的可比性，待与专家判断进行验证。

View on arXiv Download PDF AI Translation

cs.CV / 146 / 2609.15150

Weakly Supervised Spatial Grounding for Discriminative Attention-Based Ultrasound-Histopathology Alignment in Prostate Cancer Grading

弱监督空间定位用于基于区分性注意力的前列腺癌分级超声-组织病理对齐

Dzikunu, Obed Korshie, Willis, Emma, Abootorabi, Mohammad Mahdi, Harmanani, Mohamed, Guo, Zhuoxin, Luger, Ferdinand, Kinnaird, Adam, Wodlinger, Brian, Mousavi, Parvin, Abolmaesumi, Purang

Abstract

Unpaired cross-modal distillation transfers grade structure from histopathology into a micro-ultrasound (micro-US) encoder by aligning a pooled needle-region embedding to a frozen histopathology teacher under grade-group correspondence alone. A single objective is thereby required to serve two distinct functions: rendering patch features discriminative of tissue state, and selecting which patches enter the pooled representation. We decouple them. Weak spatial supervision derived from percentage involvement, recorded routinely at biopsy, constrains the predicted proportion of malignant tissue within each core, acting on the encoder features independently of the alignment objective. The alignment loss then operates on features that differ across a core, and attention concentrates on a subset of patches rather than remaining near-uniform. On 7,166 biopsy cores from 811 patients across seven centers under patient-level 5-fold cross-validation, the method reaches 67.1 macro AUC and 68.5 csPCa AUC, against 61.2 and 52.8 for the existing unpaired alignment method and 63.1 and 62.6 for the strongest unimodal baselines. Ablation against existing attention regularizers designed to prevent attention-uniformity collapse shows that such regularizers do not substitute for label-derived supervision: they constrain the attention distribution, whereas the signal required acts on the features that attention reads.

Chinese Translation

无配对跨模态蒸馏通过将池化针头区域嵌入与冻结的组织病理教师在分级组对应下对齐，将分级结构从组织病理转移到微超声（micro-US）编码器。因此，需要一个单一目标来承担两个不同的功能：使补丁特征能够区分组织状态，并选择哪些补丁进入池化表示。我们将这两个功能解耦。基于活检中常规记录的百分比参与度的弱空间监督，限制了每个核心内预测的恶性组织比例，独立于对齐目标作用于编码器特征。然后，对齐损失在一个核心内的特征上操作，注意力集中在补丁的子集上，而不是保持近乎均匀。在对811名患者的7个中心进行的患者级5折交叉验证中，来自7,166个活检核心的方法达到了67.1的宏观AUC和68.5的csPCa AUC，而现有的无配对对齐方法分别为61.2和52.8，最强的单模基线为63.1和62.6。与现有旨在防止注意力均匀性崩溃的注意力正则化器的消融实验表明，这些正则化器并不能替代标签衍生的监督：它们限制了注意力分布，而所需的信号作用于注意力读取的特征。

View on arXiv Download PDF AI Translation

cs.CV / 147 / 2609.15152

PACE: Progressive Angular-to-Norm Contrastive Embedding

PACE：渐进式角度到范数对比嵌入

Li, Yanping, Zhou, Wei, Liu, Yawen, Wang, Yibo, Zhu, Ke, Huzhang, Guangda, Chen, Qing-Guo, Xu, Zhao, Zhang, Jun, Wei, Wei

Abstract

Multimodal embedding models encode heterogeneous inputs into a shared embedding space, enabling efficient similarity computation across modalities and tasks. Most existing methods optimize cosine-based contrastive objectives, which promote stable training but restrict semantic compatibility to angular geometry, precluding embedding norms from serving as an additional semantic signal. However, directly optimizing the more expressive dot-product similarity, which leverages both angular and norm information, underperforms cosine-based training and exhibits unstable training dynamics. We attribute this discrepancy to premature optimization-space expansion, manifested as angular--norm entanglement and directional anisotropy in the representation space and further compounded by full-parameter fine-tuning. In this paper, we propose PACE, a two-stage framework that progressively expands both the representation and trainable parameter spaces. Stage I combines cosine-based objective with low-rank adaptation to establish a reliable angular geometry within constrained optimization spaces. Stage II switches to dot-product similarity and full-parameter fine-tuning, enabling embedding directions and norms to jointly encode semantic information. We further introduce Focal Embedding Loss, a confidence-adaptive objective that downweights queries with high positive retrieval confidence while emphasizing ambiguous queries with competitive negatives. Experiments across multiple backbone scales and diverse multimodal embedding tasks consistently validate the effectiveness of PACE.

Chinese Translation

多模态嵌入模型将异构输入编码到共享嵌入空间中，从而实现跨模态和任务的高效相似性计算。现有大多数方法优化基于余弦的对比目标，这促进了稳定的训练，但将语义兼容性限制在角度几何上，阻止了嵌入范数作为额外的语义信号。然而，直接优化更具表现力的点积相似性（dot-product similarity），利用角度和范数信息，表现不如基于余弦的训练，并且展现出不稳定的训练动态。我们将这种差异归因于过早的优化空间扩展，表现为表示空间中的角度-范数纠缠和方向性各向异性，并且通过全参数微调进一步加剧。在本文中，我们提出了PACE，一个两阶段框架，逐步扩展表示空间和可训练参数空间。第一阶段结合基于余弦的目标与低秩适应，以在受限优化空间内建立可靠的角度几何。第二阶段切换到点积相似性和全参数微调，使嵌入方向和范数能够共同编码语义信息。我们进一步引入了聚焦嵌入损失（Focal Embedding Loss），这是一种自适应信心目标，降低高正检索信心查询的权重，同时强调具有竞争性负样本的模糊查询。在多个主干网络规模和多样化的多模态嵌入任务中的实验一致验证了PACE的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 148 / 2609.15169

GRAVA: Grounded Reasoning-to-Action Representation and Learning for Autonomous Driving

GRAVA：用于自主驾驶的基于实证的推理-行动表示与学习

Liu, Xiao, Li, Haoyu, Leng, Jianghao, Wang, Lin, Sun, Chao

Abstract

Driving vision-language-action (VLA) models increasingly reason before acting, but their intermediate reasoning is often weakly grounded in physical scene evidence and loosely connected to executable behavior. We present GRAVA, a framework built around Grounded Reasoning-to-Action (GRA), which unifies grounding, reasoning, and action generation in a single autoregressive stream. GRA links action-relevant language references to 2D visual regions and ego-centric physical states, organizes object interactions and decisions in a trajectory-anchored typed graph, and serializes this structure into grounded reasoning. A single VLM generates this reasoning followed by a compact Executable Planner action that is deterministically decoded into a continuous trajectory. We further introduce an agentic GRA data construction pipeline that combines forward scene grounding with backward trajectory anchoring, and use it to build GR-NavSim with 2.2M grounded question-answer pairs and 70K GRA reasoning traces. A progressive training strategy develops grounded cognition through pre-training, establishes the reasoning-to-action interface through imitation, and improves driving behavior through reinforcement learning and exploration. Using about 60% of the available human driving demonstrations for action supervision, GRAVA-8B achieves state-of-the-art performance among purely autoregressive driving models on the full NAVSIM benchmark. On an internal long-tail benchmark, full GRA improves key-object compliance and Closed-loop Driving Score by 19.3% and 20.5% over action-only prediction, respectively. These results show the benefit of preserving action-relevant physical evidence from grounded reasoning through executable action generation.

Chinese Translation

驾驶视觉-语言-行动（VLA）模型越来越多地在行动之前进行推理，但它们的中间推理往往与物理场景证据的关联较弱，并且与可执行行为的连接较松散。我们提出了GRAVA，一个围绕基于实证的推理-行动（GRA）构建的框架，它将实证、推理和行动生成统一为一个自回归流。GRA将与行动相关的语言引用链接到二维视觉区域和自我中心的物理状态，组织对象交互和决策于一个基于轨迹的类型图中，并将该结构序列化为基于实证的推理。单个视觉语言模型（VLM）生成此推理，随后生成一个紧凑的可执行规划器动作，该动作被确定性解码为连续轨迹。我们进一步引入了一种代理GRA数据构建管道，该管道结合了前向场景实证与后向轨迹锚定，并利用它构建了GR-NavSim，包含220万个基于实证的问题-答案对和7万个GRA推理轨迹。渐进式训练策略通过预训练发展基于实证的认知，通过模仿建立推理-行动接口，并通过强化学习和探索改善驾驶行为。使用约60%的可用人类驾驶演示进行动作监督，GRAVA-8B在完整的NAVSIM基准测试中实现了纯自回归驾驶模型的最先进性能。在一个内部长尾基准测试中，完整的GRA在关键对象合规性和闭环驾驶得分上分别比仅基于动作的预测提高了19.3%和20.5%。这些结果表明，通过可执行动作生成保留基于实证的推理中的与行动相关的物理证据的好处。

View on arXiv Download PDF AI Translation

cs.CV / 149 / 2609.15171

EECTracker: Swarm Motion Prior-Guided Feature Compensation for Airborne Optical UAV Swarm Tracking

EECTracker：基于群体运动先验的特征补偿用于空中光学无人机群体跟踪

Chu, Zhaochen, Song, Tao, Jin, Ren, Jia, Mingdong, Lin, Defu

Abstract

Airborne optical tracking of uncrewed aerial vehicle (UAV) swarms is challenging due to extremely small target scales, rapid viewpoint changes, and cluttered backgrounds, which can weaken target feature responses and lead to intermittent or temporarily missing detector responses. Existing multi-object tracking methods generally depend on reliable target-specific detector responses to maintain target states and identities across frames. When such responses become unreliable, target states cannot be reliably updated and cross-frame association cues become ambiguous, resulting in fragmented trajectories and identity switches. To address this problem, we propose EECTracker, a swarm-motion-prior-guided joint detection-and-tracking framework for airborne optical UAV swarm tracking. EECTracker constructs a probabilistic swarm motion prior from reliable historical tracklets to capture the shared short-term image-plane motion tendency of the swarm and its uncertainty, providing spatial guidance for cross-frame feature compensation. Building on this prior, we introduce Energy--Entropy Consistency Activation (EEC Activation) to evaluate motion-prior-conditioned feature consistency using feature residual energy and local residual entropy. The resulting Local EEC score guides pixel-level feature compensation by enhancing motion-prior-consistent feature responses in potential target regions while suppressing inconsistent background responses. Experiments on AIRMOT and UAVSwarm show that EECTracker achieves superior overall tracking performance compared with state-of-the-art methods. Compared with the strongest competing method SCT-MOT, EECTracker improves MOTA/IDF1 by 3.89/1.79 percentage points on AIRMOT and by 2.81/1.74 percentage points on UAVSwarm, while maintaining online inference speed.

Chinese Translation

空中光学跟踪无人机（UAV）群体面临着极小目标尺度、快速视角变化和复杂背景等挑战，这些因素可能削弱目标特征响应，导致探测器响应间歇性或暂时缺失。现有的多目标跟踪方法通常依赖于可靠的目标特定探测器响应，以在帧间维持目标状态和身份。当这些响应变得不可靠时，目标状态无法可靠更新，帧间关联线索变得模糊，从而导致轨迹碎片化和身份切换。为了解决这一问题，我们提出了EECTracker，一种基于群体运动先验的联合检测与跟踪框架，用于空中光学无人机群体跟踪。EECTracker从可靠的历史轨迹片段构建了一个概率性群体运动先验，以捕捉群体的共享短期图像平面运动趋势及其不确定性，为跨帧特征补偿提供空间指导。在此先验的基础上，我们引入了能量-熵一致性激活（EEC Activation），利用特征残差能量和局部残差熵评估运动先验条件下的特征一致性。由此产生的局部EEC评分通过增强潜在目标区域内运动先验一致的特征响应，同时抑制不一致的背景响应，来指导像素级特征补偿。在AIRMOT和UAVSwarm上的实验表明，EECTracker在整体跟踪性能上优于最先进的方法。与最强竞争方法SCT-MOT相比，EECTracker在AIRMOT上提高了3.89/1.79个百分点的MOTA/IDF1，在UAVSwarm上提高了2.81/1.74个百分点，同时保持在线推理速度。

View on arXiv Download PDF AI Translation

cs.CV / 150 / 2609.15187

Does Attention-Guided Masking Really Help Object Discovery in Object-Centric Learning?

注意力引导的掩蔽真的有助于物体中心学习中的物体发现吗？

Tao, Youliang, Han, Yanhua, Zhao, Bin, Kannala, Juho, Pajarinen, Joni, Zhao, Rongzhen

Abstract

Object-Centric Learning (OCL) aims to decompose images into objects without human annotations. A major family of mainstream methods uses Slot Attention to aggregate image features into object-level representations and then from them reconstructs masked image content, i.e., Random Masking (RM), to provide self-supervision. The recent method DIAS simply masks image patches at uniform randomness yet achieves competitive object discovery accuracy. Since attention during aggregation already possesses object discovery ability, we explore using it to develop a better image patch masking strategy, i.e., Attention Guided Masking (AGM), thereby providing better self-supervision. Results on six recognized datasets show that AGM does not always outperform RM. Under unconditional slot initialization, AGM substantially improves background segmentation on datasets with realistic textures (COCO and VOC); Regardless of conditional or unconditional slot initialization and across datasets, foreground object discovery remains comparable or decreases. We suggest peer researchers in the OCL community that attempts to exploit internal attention semantics to improve OCL with masked decoding are risky. Our source code, model checkpoints and evaluation logs will be released upon acceptance.

Chinese Translation

物体中心学习（Object-Centric Learning, OCL）旨在在没有人工标注的情况下将图像分解为物体。一类主要的主流方法使用槽注意力（Slot Attention）将图像特征聚合为物体级表示，然后从中重建掩蔽的图像内容，即随机掩蔽（Random Masking, RM），以提供自我监督。最近的方法DIAS简单地以均匀随机的方式掩蔽图像块，但仍然实现了具有竞争力的物体发现精度。由于聚合过程中的注意力已经具备物体发现能力，我们探索利用它来开发更好的图像块掩蔽策略，即注意力引导掩蔽（Attention Guided Masking, AGM），从而提供更好的自我监督。在六个公认的数据集上的结果表明，AGM并不总是优于RM。在无条件槽初始化下，AGM显著改善了在具有真实纹理的数据集（COCO和VOC）上的背景分割；无论是条件还是无条件槽初始化，在各个数据集上，前景物体发现的效果保持相当或有所下降。我们建议OCL社区的同行研究者，尝试利用内部注意力语义来改善OCL与掩蔽解码的结合是有风险的。我们的源代码、模型检查点和评估日志将在接受后发布。

View on arXiv Download PDF AI Translation

cs.CV / 151 / 2609.15189

Reconstructing Is Not Acting: Action-Centric Latent Dynamics Modeling

重构并非行动：以行动为中心的潜在动态建模

Fu, Dingjie, Shi, Dianxing, Xu, Yangyang, Yu, Jun

Abstract

Latent action models (LAMs) learn action representations from unlabeled videos by inferring latent actions from visual transitions and reconstructing future states. However, we identify a fundamental $\textbf{reconstruction-action mismatch}$: lower reconstruction error does not necessarily yield better latent dynamics or downstream performance. We attribute this mismatch to two underconstrained aspects of reconstruction-based latent dynamics modeling: (i) the inverse dynamics model (IDM) is not explicitly encouraged to distinguish action-related transitions from nuisance appearance, and (ii) the forward dynamics model (FDM) can underutilize the inferred latent action by exploiting predictive shortcuts from the current state. To address both limitations, we propose $\textbf{ACT-LAM}$, a lightweight action-centric framework that strengthens both action extraction and action utilization. Specifically, its Action Query IDM (AQ-IDM) employs learnable action queries and gated aggregation to selectively extract rich action-related transition cues without strong information bottlenecks. And its Action Token FDM (AT-FDM) projects latent actions into action tokens that progressively interact with evolving state representations, enabling continuous state-aware action conditioning. ACT-LAM further streamlines feature processing to concentrate model capacity on latent dynamics modeling. Extensive experiments on several robotic datasets and the VP$^2$ benchmark demonstrate stronger latent action consistency, forward dynamics, and downstream visual planning performance with fewer trainable parameters and lower computational overhead. In particular, ACT-LAM surpasses the previous state of the art by $\textbf{7.6%}$ on the aggregated VP$^2$ success rate. Codes at $\href{https://github.com/DingjieFu/ACT-LAM}{url}$.

Chinese Translation

潜在行动模型（LAMs）通过从视觉转变中推断潜在行动并重构未来状态，从未标记的视频中学习行动表示。然而，我们识别出一个根本的$ extbf{重构-行动不匹配}$：较低的重构误差并不一定会带来更好的潜在动态或下游性能。我们将这种不匹配归因于基于重构的潜在动态建模的两个约束不足的方面：（i）逆动态模型（IDM）未被明确鼓励区分与行动相关的转变和无关的外观；（ii）前向动态模型（FDM）可以通过利用当前状态的预测捷径来未充分利用推断出的潜在行动。为了解决这两个局限性，我们提出了$ extbf{ACT-LAM}$，一个轻量级的以行动为中心的框架，强化了行动提取和行动利用。具体而言，其行动查询逆动态模型（AQ-IDM）采用可学习的行动查询和门控聚合，选择性地提取丰富的与行动相关的转变线索，而不会产生强信息瓶颈。其行动令牌前向动态模型（AT-FDM）将潜在行动投影为行动令牌，这些令牌与不断演变的状态表示逐步交互，实现持续的状态感知行动条件。ACT-LAM进一步简化特征处理，以集中模型能力于潜在动态建模。在多个机器人数据集和VP$^2$基准上的广泛实验表明，ACT-LAM在潜在行动一致性、前向动态和下游视觉规划性能方面表现更强，同时训练参数更少，计算开销更低。特别是，ACT-LAM在聚合的VP$^2$成功率上超越了之前的最先进水平$ extbf{7.6%}$。代码可在$ exthref{https://github.com/DingjieFu/ACT-LAM}{url}$获取。

View on arXiv Download PDF AI Translation

cs.CV / 152 / 2609.15224

Long-to-Short Video Evidence Reasoning for Grounded Question Answering

基于长短视频证据推理的基础问题回答

Chen, Kaiyan, Xiao, Junbin, Yang, Xun

Abstract

We present LOVER, a \underline{L}ong to sh\underline{O}rt \underline{V}ideo \underline{E}vidence \underline{R}einforced model for grounded question answering (GQA). LOVER highlights three innovations over existing reinforcement-learning (RL) based video reasoning models: (1) \textbf{Long-to-short Video Evidence Curriculum Learning}, which organizes RL training according to evidence duration and progressively adapts the model from long-range grounding to short-term reasoning; (2) \textbf{GQA Rewards}, which underscore the benefit of IoP reward over IoU for evidence spotting rather than strict temporal span overlap; (3) \textbf{Adaptive Timestamp Rendering}, which adaptively renders timestamps onto video frames using background-aware position and color selection to enhance temporal observability. The three designs are model-agnostic and reciprocal. They effectively improve QA, grounding, and grounded QA performance over different backbones. Notably, LOVER built on Time-R1 achieves new state-of-the-art (SOTA) results among open-source models on popular GQA benchmarks: NExT-GQA and ReXTime. Comprehensive ablation studies further validate the effectiveness of our three innovative components.

Chinese Translation

我们提出了LOVER，一种用于基础问题回答（GQA）的长短视频证据强化模型。LOVER在现有基于强化学习（RL）的视频推理模型上突出了三项创新：(1) extbf{长短视频证据课程学习}，该方法根据证据持续时间组织RL训练，并逐步将模型从长时间范围的基础推理适应到短期推理；(2) extbf{GQA奖励}，强调了IoP奖励相较于IoU在证据识别中的优势，而不是严格的时间跨度重叠；(3) extbf{自适应时间戳渲染}，通过背景感知的位置和颜色选择自适应地将时间戳渲染到视频帧上，以增强时间可观察性。这三种设计是与模型无关且相互促进的。它们有效提升了不同基础模型的问答、基础和基础问答性能。值得注意的是，基于Time-R1的LOVER在流行的GQA基准测试NExT-GQA和ReXTime上达到了开源模型中的新状态最优（SOTA）结果。全面的消融研究进一步验证了我们三项创新组件的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 153 / 2609.15225

Deep Learning-based Intelligent Diagnosis of Congenital Uterine Anomalies in 3D Ultrasound

基于深度学习的三维超声先天性子宫异常智能诊断

Xu, Yueyue, Huang, Yuhao, Deng, Jiaxiao, Zhang, Yuanji, Zhang, Haoming, Qu, Jiajia, Zheng, Shiying, Tang, Xiaomei, Chen, Haining, Chen, Chengcai, Wu, Yiyi, Yang, Xin, Ni, Dong

Abstract

Objective: To develop an intelligent framework, termed CUA-Net, for the automated classification of congenital uterine anomalies (CUA) without requiring coronal plane reconstruction, and to evaluate its clinical applicability. Methods: CUA-Net was built on 3D ResNet-18, equipped with a dynamic data resampling strategy to mitigate the data imbalance issue and a hard sample mining technique to fully learn from the difficult cases by loss adjustment. We further proposed the self-supervised reconstruction to comprehensively explore the volumes and the online data augmentation to refine the wrong predictions and enhance the model's generalization. We compared the CUA-Net with different deep-learning methods and junior/senior sonographers in the testing set. The evaluation metrics included accuracy, precision, recall, F1-score, micro-AUC, and macro-AUC. Results: The proposed CUA-Net exhibited satisfactory performance in both internal and external test sets. In the internal cohort, the model achieved accuracy of 93.88%, precision of 87.01%, recall of 95.92%, F1-score of 88.09%, and micro-AUC of 0.9982 and macro-AUC of 0.9997. In the external set, it maintained good performance with accuracy of 91.52%, precision of 83.27%, recall of 88.63%, F1-score of 81.49%, micro-AUC of 0.9945 and macro-AUC of 0.9990. Our CUA-Net outperformed the junior sonographers across all performance indicators and achieved performance comparable to that of the senior sonographers across most metrics. Conclusion: The CUA-Net demonstrates favorable accuracy and generalizability in classifying common CUA categories, while showing preliminary potential for recognizing less prevalent anomalies. These capabilities may help optimize clinical workflows and support more standardized diagnosis.

Chinese Translation

目的：开发一个名为CUA-Net的智能框架，用于自动分类先天性子宫异常（CUA），无需进行冠状面重建，并评估其临床适用性。方法：CUA-Net基于3D ResNet-18构建，配备动态数据重采样策略以减轻数据不平衡问题，并采用困难样本挖掘技术通过损失调整充分学习困难案例。我们进一步提出了自监督重建，以全面探索体积，并采用在线数据增强来修正错误预测并增强模型的泛化能力。我们将CUA-Net与不同的深度学习方法及初级/高级超声医师在测试集上进行了比较。评估指标包括准确率、精确率、召回率、F1-score、微AUC和宏AUC。结果：所提出的CUA-Net在内部和外部测试集中均表现出令人满意的性能。在内部队列中，该模型的准确率为93.88%，精确率为87.01%，召回率为95.92%，F1-score为88.09%，微AUC为0.9982，宏AUC为0.9997。在外部集上，其表现良好，准确率为91.52%，精确率为83.27%，召回率为88.63%，F1-score为81.49%，微AUC为0.9945，宏AUC为0.9990。我们的CUA-Net在所有性能指标上均优于初级超声医师，并在大多数指标上与高级超声医师的表现相当。结论：CUA-Net在分类常见CUA类别方面表现出良好的准确性和泛化能力，同时在识别较少见异常方面显示出初步潜力。这些能力可能有助于优化临床工作流程并支持更标准化的诊断。

View on arXiv Download PDF AI Translation

cs.CV / 154 / 2609.15227

Closed-form Bayesian homography estimation from noisy point correspondences

基于噪声点对应的闭式贝叶斯单应性估计

Beuter, Hanne, Dorn, Sebastian

Abstract

While homographies are fundamental to many computer vision tasks, the majority of conventional estimation techniques provide only point estimates without directly quantifying uncertainty introduced by noisy observations. Uncertainty, though, propagates to subsequent processing steps such as camera calibration and 3D reconstruction and is particularly relevant in safety-critical and socially relevant fields including medical imaging, autonomous driving, and defense. We present a fast Bayesian formulation for homography estimation from point correspondences that explicitly incorporates measurement uncertainty and prior knowledge while providing a posterior distribution over the homography parameters. A closed-form solution of the posterior mean of the homography is derived in homogeneous coordinates and supplemented by an iterative Bayesian approach to handle non-linearities. Synthetic experiments demonstrate the applicability to projective transformations and show improved estimation accuracy over DLT under varying noise conditions. Image stitching experiments further demonstrate applicability to real image correspondences while additionally providing uncertainty information.

Chinese Translation

单应性在许多计算机视觉任务中是基础，但大多数传统估计技术仅提供点估计，而未直接量化由噪声观测引入的不确定性。然而，不确定性会传播到后续处理步骤，例如相机标定和三维重建，这在医疗成像、自动驾驶和国防等安全关键和社会相关领域尤为重要。我们提出了一种快速的贝叶斯单应性估计方法，该方法从点对应中显式地结合了测量不确定性和先验知识，同时提供了单应性参数的后验分布。我们在齐次坐标中推导了单应性后验均值的闭式解，并通过迭代贝叶斯方法来处理非线性。合成实验展示了该方法在投影变换中的适用性，并在不同噪声条件下显示出比直接线性变换（DLT）更高的估计精度。图像拼接实验进一步展示了该方法在真实图像对应中的适用性，同时提供了不确定性信息。

View on arXiv Download PDF AI Translation

cs.CV / 155 / 2609.15228

Unsupervised Point Cloud Registration via Training-Time Semantic Guidance

基于训练时语义指导的无监督点云配准

Xiong, Kezheng, Xu, Shiyun, Ao, Sheng, Shen, Siqi, Wang, Cheng, Wen, Chenglu

Abstract

Unsupervised registration of large-scale LiDAR point clouds remains challenging due to the geometric ambiguity inherent in outdoor scenes, which degrades pseudo-label quality and leads to suboptimal convergence, particularly for sparse, low-resolution scans such as those from nuScenes. We reveal that registration models intrinsically encode semantic awareness that strongly correlates with registration accuracy, albeit without explicit semantic supervision. However, this native awareness is fragile: noisy supervision arising from geometric ambiguity in unsupervised settings rapidly erodes the learned semantic structure, causing performance collapse. To this end, we propose CAESAR, a teacher-student framework guided by an off-the-shelf 3D segmentation model exclusively during training. We observe that potential inlier matches are often buried just beneath a few spurious neighbors in the noisy feature space, motivating Dual-Cue Guided Re-Matching to recover them through reselection rather than simply rejecting. Building on this, a train-only Semantic-Geometric Label Mining performs lightweight, batch-specific teacher refinement and mines reliable pseudo-labels under semantic guidance. We further introduce Semantic Predictive Distillation to consolidate the student's semantic awareness in the feature space. Extensive experiments on KITTI and nuScenes demonstrate state-of-the-art performance, with pronounced gains on the challenging nuScenes benchmark. Crucially, CAESAR incurs zero inference overhead and requires no semantic annotations on the registration data. Code will be released.

Chinese Translation

大规模LiDAR点云的无监督配准仍然面临挑战，因为户外场景固有的几何模糊性会降低伪标签质量，并导致次优收敛，尤其是在稀疏、低分辨率的扫描（如nuScenes数据集）中。我们揭示了配准模型本质上编码了与配准精度强相关的语义意识，尽管没有显式的语义监督。然而，这种内在的意识是脆弱的：在无监督环境中，由于几何模糊性产生的噪声监督迅速侵蚀了学习到的语义结构，导致性能崩溃。为此，我们提出了CAESAR，一个在训练期间仅由现成的3D分割模型指导的师生框架。我们观察到，潜在的内点匹配通常被埋藏在噪声特征空间中几个虚假邻居之下，这激励了双线索引导的重新匹配，通过重新选择而不是简单拒绝来恢复它们。在此基础上，训练专用的语义-几何标签挖掘执行轻量级的批次特定教师精炼，并在语义指导下挖掘可靠的伪标签。我们进一步引入语义预测蒸馏，以巩固学生在特征空间中的语义意识。在KITTI和nuScenes上的大量实验表明了最先进的性能，在具有挑战性的nuScenes基准上取得了显著提升。重要的是，CAESAR在推理时没有额外开销，并且不需要对配准数据进行语义注释。代码将会发布。

View on arXiv Download PDF AI Translation

cs.CV / 156 / 2609.15229

Pre-PEFT Probing: Weight Statistics and Perturbation Robustness for Layer Selection in VLM Vision Encoders

预PEFT探测：用于VLM视觉编码器层选择的权重统计与扰动鲁棒性

Xia, Qingtao, Bao, Jiahua, Cheng, Siyao, Liu, Jie

Abstract

We propose a pre-fine-tuning probing method for Parameter-Efficient Fine-Tuning (PEFT) layer selection, aiming to obtain more stable and higher gains with fewer trainable parameters when adapting large vision--language models (VLMs). Unlike the common practice of applying LoRA and other adapters to all layers at once---where layer selection often relies on heuristic rules---we focus on the vision encoder and directly evaluate the "adaptability'' of each Transformer layer. Specifically, we characterize each layer from two perspectives: (i) the statistical properties of its Q/K/V projection weights (e.g., norms and condition numbers); (ii) robustness under controlled parameter perturbations. We then systematically compare these indicators with the downstream performance gains brought by applying PEFT to a single layer. Across experiments covering seven benchmarks and five PEFT variants, we observe a consistent correlation: layers (or matrices) with larger weight norms and higher condition numbers are usually more robust to perturbations and are more likely to yield larger fine-tuning gains. These results show that distribution-statistics analysis and perturbation tests before fine-tuning can provide practical signals for adaptation-layer selection, thereby maintaining or improving performance while reducing trainable parameters.

Chinese Translation

我们提出了一种用于参数高效微调（PEFT）层选择的预微调探测方法，旨在在适应大型视觉-语言模型（VLMs）时，通过更少的可训练参数获得更稳定和更高的收益。与通常将LoRA和其他适配器同时应用于所有层的做法不同——这种层选择通常依赖于启发式规则——我们专注于视觉编码器，并直接评估每个Transformer层的“适应性”。具体而言，我们从两个角度对每一层进行特征描述：（i）其Q/K/V投影权重的统计特性（例如，范数和条件数）；（ii）在受控参数扰动下的鲁棒性。然后，我们系统地将这些指标与将PEFT应用于单层所带来的下游性能提升进行比较。在涵盖七个基准和五种PEFT变体的实验中，我们观察到一种一致的相关性：具有较大权重范数和较高条件数的层（或矩阵）通常对扰动更鲁棒，并且更有可能产生更大的微调收益。这些结果表明，在微调之前进行分布统计分析和扰动测试可以为适应层选择提供实用信号，从而在减少可训练参数的同时保持或提高性能。

View on arXiv Download PDF AI Translation

cs.CV / 157 / 2609.15241

A 25-$\mu$s/inf Event-driven Graph Neural Network Processor with Spatiotemporal Caching and Spline Convolution for Ultra-low-latency AI at the Edge

一种具有时空缓存和样条卷积的25-$bc$s/inf事件驱动图神经网络处理器，用于边缘超低延迟人工智能

Kneip, Adrian, Lefebvre, Martin, Gehrig, Daniel, Pastor, Victoria Catalán, Scaramuzza, Davide, Verhelst, Marian, Frenkel, Charlotte

Abstract

Dynamic-vision-sensor (DVS) cameras generate events on a per-pixel basis with a $\mu$s-level temporal resolution, calling for new algorithm-hardware co-design approaches compared to standard frame-based vision. While event-driven graph neural networks (EV-GNNs) emerge as a promising algorithmic solution, they raise new HW challenges by mixing dense-regular compute operations and sparse-irregular memory accesses. We present ETHEREAL, the first EV-GNN accelerator that scales to 640$\times$480 resolutions, thanks to a neighbor-parallel spline convolution engine and a 2D/3D-split memory hierarchy with a novel region-of-interest spatiotemporal caching mechanism. Measurement results demonstrate end-to-end inference with 25.6$\mu$s latency and 1.7$\mu$J energy per event on state-of-the-art workloads

Chinese Translation

动态视觉传感器（DVS）相机以每个像素为基础生成事件，具有$bc$s级的时间分辨率，这与标准的基于帧的视觉相比，要求新的算法-硬件协同设计方法。虽然事件驱动的图神经网络（EV-GNNs）作为一种有前景的算法解决方案出现，但它们通过混合密集规则的计算操作和稀疏不规则的内存访问，带来了新的硬件挑战。我们提出了ETHEREAL，这是第一个能够扩展到640$ imes$480分辨率的EV-GNN加速器，得益于邻域并行样条卷积引擎和具有新颖兴趣区域时空缓存机制的2D/3D分层内存架构。测量结果表明，在最先进的工作负载下，端到端推理的延迟为25.6$bc$s，每个事件的能量为1.7$bc$J。

View on arXiv Download PDF AI Translation

cs.CV / 158 / 2609.15285

AlignUS: MRI-Guided Ultrasound Representation Learning for ALS Classification from Tongue Images

AlignUS：基于MRI引导的超声表示学习用于舌部图像的ALS分类

Ghader, Kadija Abdel, Babe, Emani, Pettinari, Lorenzo, Haroune, Meya, Hadramy, Sidaty El

Abstract

Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disease in which early assessment remains challenging, particularly in low-resource settings where MRI is often unavailable. High-resolution ultrasound (HRUS) of the tongue offers a portable and low-cost alternative for evaluating bulbar involvement, but learning reliable diagnostic models is limited by small datasets and the difficulty of extracting robust representations from ultrasound alone. We propose AlignUS, a cross-modal knowledge distillation framework that transfers anatomical knowledge from MRI to a HRUS-based classifier while requiring only HRUS at inference time. The model combines classification loss, supervised contrastive learning, and feature-level distillation to align HRUS representations with MRI embeddings. AlignUS achieves a patient-level balanced accuracy of 0.958, macro-F1 of 0.963, and ROC-AUC of 0.990, aggregated across four patient-level cross-validation folds, with consistent improvements over HRUS baselines and cross-modal alternatives. These results demonstrate that MRI-derived supervision can substantially improve ultrasound-based ALS assessment while preserving low-cost, inference-time independence from MRI.

Chinese Translation

肌萎缩侧索硬化症（ALS）是一种进行性神经退行性疾病，其早期评估仍然具有挑战性，尤其是在MRI常不可用的资源匮乏环境中。舌部的高分辨率超声（HRUS）提供了一种便携且低成本的替代方案，用于评估球部受累情况，但由于数据集规模小且仅依靠超声难以提取稳健的表征，学习可靠的诊断模型受到限制。我们提出了AlignUS，一种跨模态知识蒸馏框架，将MRI中的解剖学知识转移到基于HRUS的分类器中，同时推理时仅需HRUS。该模型结合了分类损失、有监督对比学习和特征级蒸馏，以使HRUS表征与MRI嵌入对齐。AlignUS在四折患者级交叉验证中实现了患者级平衡准确率0.958、宏F1值0.963及ROC-AUC 0.990，且相较于HRUS基线和其他跨模态方法表现出持续改进。这些结果表明，基于MRI的监督能够显著提升超声基础的ALS评估效果，同时保持低成本及推理时无需MRI的优势。

View on arXiv Download PDF AI Translation

cs.CV / 159 / 2609.15307

Learning from Reliable Negatives: Confidence-Anchored Test-Time Adaptation for GUI Grounding

从可靠的负样本中学习：基于置信度的测试时适应用于GUI定位

Liu, Yizhou, Tang, Fei, Yan, Yuchen, Lu, Zhengxi, Nong, Songqin, Jiang, Tao, Xu, Wenhao, Zhang, Wenqi, Lu, Weiming, Xiao, Jun, Shen, Yongliang

Abstract

Graphical User Interface (GUI) grounding is essential for autonomous agents to map natural language instructions to precise screen coordinates. However, existing supervised fine-tuning and reinforcement learning methods are constrained by the high cost of annotation, creating a scalability bottleneck. In this paper, we introduce a label-free test-time training paradigm driven by two key insights: (1) confidence patterns in coordinate tokens are a better indicator than full-sequence confidence, and (2) in sparse GUI coordinate spaces, negative samples offer more reliable learning signals than potentially noisy positive ones. We first propose Confidence-Anchored Learning (CAL), which utilizes coordinate-token confidence to filter pseudo-labels and assign distance-based binary rewards. Building on this, we develop Confidence-Anchored Negative Learning (CANL), which exclusively optimizes the model using negative samples to bypass the risks of incorrect positive samples. Experimental results demonstrate that CANL-7B achieves 92.1% on ScreenSpot-V2. On more challenging ScreenSpot-Pro, CANL-7B reaches 33.8%, an 8.9% absolute improvement over the base model. Our findings establish coordinate-token confidence as a powerful alternative to manual annotations for scalable GUI agent development.

Chinese Translation

图形用户界面（GUI）定位对于自主智能体将自然语言指令映射到精确的屏幕坐标至关重要。然而，现有的监督微调和强化学习方法受到高昂标注成本的限制，造成了可扩展性的瓶颈。本文提出了一种无标签的测试时训练范式，基于两个关键见解：（1）坐标标记的置信度模式比全序列置信度更能有效指示，和（2）在稀疏的GUI坐标空间中，负样本提供的学习信号比潜在噪声较大的正样本更为可靠。我们首先提出了基于置信度的学习（Confidence-Anchored Learning, CAL），该方法利用坐标标记的置信度来过滤伪标签并分配基于距离的二元奖励。在此基础上，我们开发了基于置信度的负学习（Confidence-Anchored Negative Learning, CANL），该方法专门使用负样本来优化模型，从而规避错误正样本的风险。实验结果表明，CANL-7B在ScreenSpot-V2上达到了92.1%的准确率。在更具挑战性的ScreenSpot-Pro上，CANL-7B达到了33.8%，相较于基础模型有8.9%的绝对提升。我们的研究结果确立了坐标标记置信度作为可扩展GUI智能体开发中手动标注的强有力替代方案。

View on arXiv Download PDF AI Translation

cs.CV / 160 / 2609.15320

Hypergraph-Regularized Gramian Volumes for Multimodal Retrieval

超图正则化的Gram体积用于多模态检索

Nag, Anindya, Mehrish, Ambuj, Vascon, Sebastiano

Abstract

Volume-based multimodal retrieval jointly scores a text query with a candidate's video, audio, and subtitle embeddings. While this approach captures higher-order within-candidate alignment, the score remains candidate-local, and semantically related training samples primarily serve as contrastive negatives. This work introduces Hypergraph-Regularized Gramian Volumes (HyVol), a training-time module that incorporates these semantic relations prior to evaluating the original volume loss. Document hyperedges connect the observed modalities of each candidate, whereas semantic hyperedges link candidates whose detached captions are mutual top-k neighbors. A shallow gated hyper-graph network applies residual corrections to the modality embeddings. Presence masks exclude unavailable streams from message passing, and identity padding preserves the determinant of the observed Gram submatrix without feature imputation. As refinement operates on embeddings rather than scores, the same construction applies to both Gram and HyperGram. We remove the hypergraph after training, leaving the backbone-only architecture, original scoring function, and retrieval cost unchanged. We train both backbones on a 150K-clip subset of VAST-27M and evaluate zero-shot performance on six benchmarks. Under the paired protocol, HyVol improves R@1 across all five retrieval benchmarks, with video-to-text gains reaching +8.3 on MSR-VTT and +7.6 on VATEX. Under missing-modality masking, the V2T margin remains positive in all experimental settings, although the T2V margin becomes slightly negative in four.

Chinese Translation

基于体积的多模态检索通过将文本查询与候选者的视频、音频和字幕嵌入共同评分。尽管这种方法捕捉了候选者内部的高阶对齐，但评分仍然是候选者局部的，语义相关的训练样本主要作为对比负样本。本研究提出了超图正则化的Gram体积（Hypergraph-Regularized Gramian Volumes，HyVol），这是一种在评估原始体积损失之前引入这些语义关系的训练模块。文档超边连接每个候选者的观察模态，而语义超边则链接那些其分离字幕是互为前k个邻居的候选者。一个浅层门控超图网络对模态嵌入应用残差修正。存在掩码排除了不可用流的消息传递，而身份填充则在不进行特征插补的情况下保持观察到的Gram子矩阵的行列式。由于细化操作是在嵌入而非评分上进行，因此相同的构造适用于Gram和HyperGram。我们在训练后移除超图，保留仅有主干的架构、原始评分函数和检索成本不变。我们在VAST-27M的150K剪辑子集上训练了两个主干，并在六个基准上评估了零-shot性能。在配对协议下，HyVol在所有五个检索基准上提高了R@1，其中视频到文本的增益在MSR-VTT上达到+8.3，在VATEX上达到+7.6。在缺失模态掩蔽下，V2T的边际在所有实验设置中保持正值，尽管T2V的边际在四个实验中略微为负。

View on arXiv Download PDF AI Translation

cs.CV / 161 / 2609.15334

Concept-Grounded Reasoning with Prompt-Driven Localization for Interpretable Structured Report Generation

基于概念的推理与提示驱动的定位用于可解释的结构化报告生成

Xu, Xinyue, Lin, Hongbin, Xu, Juangui, Wang, Hualiang, Wang, Lehan, Hu, Lijie, Liu, Weiyang, Weller, Adrian, Li, Xiaomeng

Abstract

Medical imaging modalities such as ultrasound and X-ray are widely used in clinical practice, where diagnosis follows a structured, evidence-driven workflow aligned with standardized criteria. While multimodal large language models (MLLMs) show promise for automated medical report generation, most existing systems rely on end-to-end multimodal fusion without modeling clinically defined intermediate attributes, leading to limited grounding and interpretability. To address this issue, we propose CORAL (COncept-grounded ReAsoning with Localization), a multimodal framework that integrates spatial grounding and concept-level supervision into a unified reasoning process. CORAL employs a prompt-driven medical segmentation model to localize lesions and predicts multi-class clinical attributes through a Concept Bottleneck module. The resulting textual concept tokens are combined with mask-modulated visual features within an MLLM to enable structured report generation and diagnostic prediction. Experiments on BUS-CoT and IU X-ray datasets demonstrate consistent improvements in diagnostic accuracy, concept consistency, and report quality over strong general-purpose and medical MLLMs, indicating that concept-grounded reasoning better aligns generation with clinical decision processes.

Chinese Translation

医学影像技术如超声和X光在临床实践中被广泛应用，诊断遵循与标准化标准一致的结构化、证据驱动的工作流程。尽管多模态大型语言模型（MLLMs）在自动化医学报告生成方面显示出潜力，但大多数现有系统依赖于端到端的多模态融合，而未对临床定义的中间属性进行建模，导致基础性和可解释性有限。为了解决这一问题，我们提出了CORAL（基于概念的推理与定位），这是一个将空间基础和概念级监督整合到统一推理过程中的多模态框架。CORAL采用提示驱动的医学分割模型来定位病变，并通过概念瓶颈模块预测多类临床属性。生成的文本概念标记与在MLLM中掩模调制的视觉特征相结合，以实现结构化报告生成和诊断预测。在BUS-CoT和IU X光数据集上的实验表明，与强大的通用和医学MLLM相比，诊断准确性、概念一致性和报告质量均有显著提高，表明基于概念的推理更好地将生成与临床决策过程对齐。

View on arXiv Download PDF AI Translation

cs.CV / 162 / 2609.15335

Query-Conditioned Spherical Centroid Aggregation for Multimodal Retrieval

基于查询条件的球形质心聚合用于多模态检索

Mehrish, Ambuj, Nag, Anindya, Vascon, Sebastiano

Abstract

Multimodal retrieval integrates video, audio, subtitles, and text; however, recent geometric aggregators, such as Gramian volumes, hyperbolic volumes, and spectral objectives, treat all modalities symmetrically. Under a unified evaluation protocol, their joint scores frequently lag behind the strongest single-modality pathway by 1.9 to 27.6 R@1. Controlled analyses attribute this outcome to uniform modality influence. This work introduces Spherical Centroid Aggregation with Learned Adaptive Relevance (SCALAR), a query-conditioned aggregator that assigns relevance-based weights to each available modality before computing a spherical centroid. SCALAR accommodates arbitrary modality subsets and is trained on masked, reduced-arity views using rank-8 LoRA adapters. Across five benchmarks, SCALAR achieves positive aggregation gain on four, reaching +4.0 R@1, while none of the evaluated prior aggregators is positive on more than one. A uniform-weight ablation reproduces the degradation observed with symmetric aggregation. With only 4.8 million trainable parameters, SCALAR attains the highest text-to-video R@1 on three and performs within seed variation of the best result on a fourth. Under test-time modality dropout, SCALAR's representation-stage score surpasses the released GRAM checkpoint at every evaluated masking rate and benchmark by 3.2 to 10.9 R@1. Finally, as modalities are removed, rerankers trained exclusively on complete modality sets increasingly converge toward their video-only pathways, diminishing these representation-level gains and underscoring a limitation of standard two-stage retrieval pipelines.

Chinese Translation

多模态检索整合了视频、音频、字幕和文本；然而，最近的几何聚合器，如Gramian体、双曲体和谱目标，均对所有模态进行对称处理。在统一的评估协议下，它们的联合得分通常落后于最强的单模态路径，差距在1.9到27.6 R@1之间。受控分析将这一结果归因于模态影响的均匀性。本研究提出了带有学习自适应相关性的球形质心聚合（Spherical Centroid Aggregation with Learned Adaptive Relevance，SCALAR），这是一种基于查询条件的聚合器，在计算球形质心之前，为每个可用模态分配基于相关性的权重。SCALAR能够适应任意模态子集，并使用rank-8 LoRA适配器在掩蔽、减少元数的视图上进行训练。在五个基准测试中，SCALAR在四个测试中实现了正向聚合增益，达到+4.0 R@1，而评估的先前聚合器中没有一个在超过一个的测试中表现为正。均匀权重消融实验重现了对称聚合所观察到的性能下降。在仅有480万可训练参数的情况下，SCALAR在三个测试中达到了最高的文本到视频R@1，并在第四个测试中表现接近最佳结果的种子变异。在测试时模态丢失的情况下，SCALAR的表示阶段得分在每个评估的掩蔽率和基准上均超过了发布的GRAM检查点，提升幅度为3.2到10.9 R@1。最后，随着模态的移除，仅在完整模态集上训练的重排序器逐渐趋向于其仅视频路径，削弱了这些表示级别的增益，并突显了标准两阶段检索管道的局限性。

View on arXiv Download PDF AI Translation

cs.CV / 163 / 2609.15348

Robust Multi-Model Fitting through Learning Neighbor Regions

通过学习邻域区域实现稳健的多模型拟合

Nie, Chang, Wang, Guangming, Liu, Zhe, Wang, Hesheng

Abstract

Multi-model fitting involves fitting multiple models accurately in a noisy environment. It is the basis for computer vision tasks such as scene reconstruction and mixed reality. However, its performance is often limited by insufficient feature utilization, inefficient optimization, model overlap, and the non-differentiable pipelines. To overcome these limitations, we introduce a robust coarse-to-fine framework called Learning Neighbor Regions (LNR). Recognizing that substantial computational resources are wasted on numerous bad minimum sets, we propose the coarse-level module. This module utilizes a neural network to extract and analyze geometric feature of both local point-wise relationships and global contextual information in minimum sets, outputting confidence to pre-select a small number of good minimum sets, thereby enhancing overall efficiency before solving hypotheses. To address model overlap, LNR encodes neighbor region features for each hypothesis in its fine-level module. These region features consist of geometric features of neighboring data points, which can be used by multiple regions simultaneously. This design allows the neural network to individually refine and score each hypothesis. Importantly, LNR is trained to learn directly from data point features rather than from the hypothesis parameters, thus avoiding differentiating the sampling process and the model solvers. Extensive experiments on four classic multi-model fitting tasks demonstrate that LNR achieves state-of-the-art performance. The analysis suggests that LNR can be easily adapted to various robust multi-model fitting tasks.

Chinese Translation

多模型拟合涉及在噪声环境中准确拟合多个模型。这是计算机视觉任务（如场景重建和混合现实）的基础。然而，其性能常常受到特征利用不足、优化效率低下、模型重叠和不可微分管道的限制。为克服这些局限，我们提出了一种稳健的粗到细框架，称为学习邻域区域（Learning Neighbor Regions, LNR）。我们认识到，巨大的计算资源在许多不良最小集上被浪费，因此提出了粗级模块。该模块利用神经网络提取和分析最小集中的局部点关系和全局上下文信息的几何特征，输出置信度以预选少量良好的最小集，从而在解决假设之前提高整体效率。为了解决模型重叠问题，LNR在其细级模块中对每个假设编码邻域区域特征。这些区域特征由邻近数据点的几何特征组成，可以被多个区域同时使用。这一设计使得神经网络能够单独细化和评分每个假设。重要的是，LNR被训练为直接从数据点特征中学习，而不是从假设参数中学习，从而避免了对采样过程和模型求解器的微分。对四个经典多模型拟合任务的广泛实验表明，LNR达到了最先进的性能。分析表明，LNR可以轻松适应各种稳健的多模型拟合任务。

View on arXiv Download PDF AI Translation

cs.CV / 164 / 2609.15354

End-to-End Cell Detection via Instance-aware Graph Modeling

基于实例感知图建模的端到端细胞检测

Liu, Ruochen, Zheng, Yalin, Liu, Jingxin, Zhang, Jianfeng, Huang, Shoujun, Kong, Dexing, Li, Haofeng, Lou, Wei

Abstract

Accurate cell detection and classification are crucial for pathological analysis, directly affecting diagnostic accuracy and treatment planning. To capture complex cellular interactions beyond visual appearance within the tumor microenvironment, several approaches have employed graph neural networks to model spatial and relational patterns among cell nuclei, yielding promising results. However, these methods typically adopt a two-stage paradigm of visual extraction followed by relational modeling, which necessitates separate tuning for each stage, thereby increasing pipeline complexity and hindering end-to-end joint optimization. In this paper, we propose an end-to-end framework for cell detection and classification that jointly models patch-level visual representations and instance-level interactions, which incorporates a dynamic graph construction module and an instance-aware graph network. Specifically, the graph construction module dynamically builds the graph structure using learnable queries derived from patch-level features as cell instance representations, with adjacency defined by integrating feature similarity and spatial distances. The instance-aware graph network performs adaptive instance filtering and feature reorganization, aggregating them over the cell graph into a topological latent state for a selective state-space transition driven by visual cues, fusing appearance and relational evidence. When evaluated on multiple datasets with different staining protocols for cell and nucleus detection, our method significantly outperforms existing approaches in both detection and classification performance. The code will be released at https://github.com/RuochenLiu23/IGM.

Chinese Translation

准确的细胞检测和分类对于病理分析至关重要，直接影响诊断准确性和治疗规划。为了捕捉肿瘤微环境中复杂的细胞相互作用，超越视觉外观，几种方法采用图神经网络来建模细胞核之间的空间和关系模式，取得了良好的效果。然而，这些方法通常采用视觉提取和关系建模的两阶段范式，这需要对每个阶段进行单独调优，从而增加了流程的复杂性，并阻碍了端到端的联合优化。本文提出了一种端到端的细胞检测和分类框架，该框架联合建模补丁级视觉表示和实例级交互，包含一个动态图构建模块和一个实例感知图网络。具体而言，图构建模块使用从补丁级特征中派生的可学习查询动态构建图结构，将细胞实例表示为邻接关系，通过整合特征相似性和空间距离来定义。实例感知图网络执行自适应实例过滤和特征重组，将它们聚合到细胞图中，形成一个拓扑潜在状态，以便通过视觉线索驱动的选择性状态空间转移，融合外观和关系证据。在对多个具有不同染色协议的细胞和细胞核检测数据集进行评估时，我们的方法在检测和分类性能上显著优于现有方法。代码将发布在 https://github.com/RuochenLiu23/IGM。

View on arXiv Download PDF AI Translation

cs.CV / 165 / 2609.15357

Diffusion Trajectory Modeling for Semantic Correspondence

语义对应的扩散轨迹建模

Choi, Yusung

Abstract

Diffusion models generate images through an iterative diffusion process, and recent studies have demonstrated that the intermediate feature maps produced during this process contain rich visual representations, leading to their adoption across a variety of downstream tasks. However, most existing approaches are limited to either using a single feature map at a specific timestep or aggregating feature maps across multiple timesteps. We observe that intermediate representations in the diffusion process form meaningful trajectories along the time axis. In particular, the representation of each spatial patch evolves progressively throughout the generative process, encoding semantics that are difficult to capture from static snapshots alone. This observation motivates the need to treat diffusion representations as temporally structured trajectories rather than static snapshots. To this end, we propose Diffusion Trajectory Modeling (DTM), a framework that interprets the temporal evolution of each spatial patch as a trajectory and leverages it for semantic correspondence. By effectively modeling patch-wise trajectories generated across multiple timesteps, DTM captures correspondence cues that prior methods are not designed to capture. We further demonstrate empirically that spatially corresponding patches form similar trajectory patterns throughout the diffusion process, suggesting that the temporal axis of diffusion carries semantic information. Experiments on SPair-71k, SPair-U and AP-10K show that DTM achieves strong performance, presenting a new perspective for exploiting diffusion representations from a trajectory-centric viewpoint.

Chinese Translation

扩散模型通过迭代扩散过程生成图像，最近的研究表明，在此过程中产生的中间特征图包含丰富的视觉表示，因而被广泛应用于多种下游任务。然而，大多数现有方法仅限于在特定时间步使用单一特征图或在多个时间步之间聚合特征图。我们观察到，扩散过程中的中间表示沿时间轴形成了有意义的轨迹。特别是，每个空间块的表示在生成过程中逐步演变，编码了从静态快照中难以捕捉的语义。这一观察促使我们将扩散表示视为时间结构化的轨迹，而非静态快照。为此，我们提出了扩散轨迹建模（Diffusion Trajectory Modeling, DTM）框架，该框架将每个空间块的时间演变解释为轨迹，并利用其进行语义对应。通过有效建模在多个时间步生成的块级轨迹，DTM 捕捉了先前方法未能捕捉的对应线索。我们进一步通过实验证明，空间上对应的块在整个扩散过程中形成了相似的轨迹模式，这表明扩散的时间轴承载着语义信息。在 SPair-71k、SPair-U 和 AP-10K 数据集上的实验表明，DTM 取得了强劲的性能，为从轨迹中心视角利用扩散表示提供了新的视角。

View on arXiv Download PDF AI Translation

cs.CV / 166 / 2609.15367

CapsuleMotion: A Lightweight Real-Time Visual Motion Predictor for Capsule Endoscopy

CapsuleMotion：一种轻量级实时视觉运动预测器用于胶囊内窥镜

Bause, Oliver, Werner, Julia, Bringmann, Oliver

Abstract

Video Capsule Endoscopy (VCE) is a non-invasive medical examination that allows for the observation of the small intestine, which is otherwise difficult to access. A fundamental challenge persists in the form of their limited size in order to still be swallowable. The resulting restricted battery capacity, however, contradicts with the power-intensive nature of image capture and transmission. Therefore, we propose CapsuleMotion, a patient-specific dynamic capsule behavior that utilizes the available energy in a goal-oriented manner to increase the likelihood of a complete screening of the gastrointestinal tract. By investigating and combining metrics from the on-device image compression, CapsuleMotion predicts the motion between two successive frames. The camera's frame rate will be modified in accordance with the predicted magnitude of motion. Furthermore, prior to entering the small intestine, the capsule operates in a low power mode with a significantly reduced frame rate. In this mode, the LocalizationNet is employed to determine the current organ, provided that motion was predicted. The proposed framework is evaluated on the Rhode Island VCE dataset and deployed on an ultra-low power single-core RISC-V demonstrator with an integrated hardware accelerator. CapsuleMotion demonstrated the capability to reduce electric energy consumption by up to 20.66% in comparison with conventional capsules that lack a dynamic frame rate. Additionally, the accuracy of detecting the entry point of the small intestine has been improved.

Chinese Translation

视频胶囊内窥镜（VCE）是一种非侵入性的医学检查方法，能够观察小肠，这在其他情况下是难以接触的。一个基本挑战在于其有限的尺寸，以便仍然可以被吞咽。然而，随之而来的电池容量限制与图像捕捉和传输的高能耗特性相矛盾。因此，我们提出了CapsuleMotion，一种患者特定的动态胶囊行为，利用可用能量以目标导向的方式提高对胃肠道进行完整筛查的可能性。通过研究和结合设备上的图像压缩指标，CapsuleMotion预测两个连续帧之间的运动。相机的帧率将根据预测的运动幅度进行调整。此外，在进入小肠之前，胶囊将在低功耗模式下以显著降低的帧率运行。在此模式下，LocalizationNet被用于确定当前器官，前提是预测到了运动。所提出的框架在罗德岛VCE数据集上进行了评估，并在集成硬件加速器的超低功耗单核RISC-V演示器上部署。与缺乏动态帧率的传统胶囊相比，CapsuleMotion展示了将电能消耗减少多达20.66%的能力。此外，检测小肠入口点的准确性得到了提高。

View on arXiv Download PDF AI Translation

cs.CV / 167 / 2609.15400

BSC-Net: A Small-Branch-Sensitive Structural Continuity Network for Coronary Vessel Segmentation and Quantitative Angiographic Analysis

BSC-Net：一种针对冠状血管分割和定量血管造影分析的小分支敏感结构连续网络

Li, Wanxian, Qin, Jiaqian, Xian, Qingyi, Li, Yazhi, Chen, Song, Li, Liman, He, Hao

Abstract

Vessel segmentation in X-ray coronary angiography (XCA) is a fundamental step for quantitative coronary analysis and subsequent assessment of coronary artery disease. However, accurate vessel segmentation remains challenging because of imaging noise, complex bifurcations, and the overlap of vessels and background structures, which can lead to disrupted vascular connectivity and missed small branches. In this work, we propose BSC-Net, a ResNet-U-Net-based framework tailored to improve small-vessel representation and repair vascular structural continuity. BSC-Net enhances small-vessel representation through targeted sampling and improves vascular structural continuity by integrating long-range contextual modeling and Edge-Informed Loss (EIL). BSC-Net was validated on two public XCA datasets, demonstrating state-of-the-art (SOTA) performance in coronary vessel segmentation with Dice and IoU scores of 77.8%/90.6% and 64.5%/83.0%, respectively. Furthermore, based on the obtained vessel segmentation, we performed automated quantitative coronary analysis and derived clinically relevant morphological and hemodynamic parameters, including stenosis ratio, time-to-peak, and relative propagation velocity. These results demonstrate that BSC-Net produces accurate vessel segmentation results with preserved vascular continuity for quantitative coronary assessment, enabling reliable downstream analysis and clinical evaluation of coronary artery disease.

Chinese Translation

在X射线冠状血管造影（XCA）中，血管分割是进行定量冠状分析和后续冠状动脉疾病评估的基础步骤。然而，由于成像噪声、复杂的分叉以及血管与背景结构的重叠，准确的血管分割仍然面临挑战，这可能导致血管连接性中断和小分支的遗漏。在本研究中，我们提出了BSC-Net，这是一种基于ResNet-U-Net的框架，旨在改善小血管的表示并修复血管结构的连续性。BSC-Net通过针对性采样增强小血管的表示，并通过整合长距离上下文建模和边缘信息损失（Edge-Informed Loss, EIL）来改善血管结构的连续性。BSC-Net在两个公共XCA数据集上进行了验证，显示出在冠状血管分割方面的最新技术（SOTA）性能，Dice和IoU得分分别为77.8%/90.6%和64.5%/83.0%。此外，基于获得的血管分割结果，我们进行了自动化的定量冠状分析，并推导出临床相关的形态和血流动力学参数，包括狭窄比、峰值时间和相对传播速度。这些结果表明，BSC-Net能够生成准确的血管分割结果，并保持血管连续性，以便进行定量冠状评估，从而实现对冠状动脉疾病的可靠下游分析和临床评估。

View on arXiv Download PDF AI Translation

cs.CV / 168 / 2609.15408

MarKey: Marginal Utility Guided Greedy Keyframe Selection for Long Video Understanding

MarKey：边际效用引导的贪婪关键帧选择用于长视频理解

Shi, Hongchang, Hu, Jinpeng, Wang, Ao, Zhou, Wenzheng, Ma, Hui, Li, Feng, Shi, Zenglin

Abstract

Long-video understanding remains challenging for multimodal large language models (MLLMs) because densely encoding long frame sequences is computationally expensive, while uniform sampling under a limited visual budget can miss sparse yet decisive evidence. Recent training-free keyframe selection methods have enabled more efficient inference and yielded promising performance gains. However, many existing methods score frames largely in isolation without explicitly considering how each candidate complements the currently selected subset, potentially resulting in redundant selections and incomplete evidence coverage. To address this limitation, we propose MarKey, a training-free framework that formulates keyframe selection as subset-aware greedy optimization. At each iteration, MarKey scores each candidate using a tractable surrogate that jointly accounts for query relevance, marginal coverage gain, and context-dependent redundancy, and selects the frame with the highest utility. To make this iterative subset-aware evaluation efficient, MarKey uses a compact set of representative anchors to approximate full-video coverage and a bounded window of previously selected frames to limit context-dependent comparisons. Experiments on six benchmarks spanning holistic video understanding, human-centric video understanding, and open-ended video understanding demonstrate that MarKey consistently outperforms existing methods. Further analyses show robust gains across different MLLM backbones, model scales, and frame budgets.

Chinese Translation

长视频理解对于多模态大语言模型（MLLMs）仍然具有挑战性，因为对长帧序列进行密集编码在计算上是昂贵的，而在有限的视觉预算下进行均匀采样可能会错过稀疏但决定性的证据。最近的无训练关键帧选择方法使得推理更加高效，并带来了令人鼓舞的性能提升。然而，许多现有方法在评分帧时主要是孤立进行的，并未明确考虑每个候选帧如何补充当前选择的子集，这可能导致冗余选择和证据覆盖不完整。为了解决这一局限性，我们提出了MarKey，一个将关键帧选择公式化为子集感知贪婪优化的无训练框架。在每次迭代中，MarKey使用一个可处理的替代指标对每个候选帧进行评分，该指标共同考虑查询相关性、边际覆盖增益和上下文相关冗余，并选择效用最高的帧。为了使这种迭代的子集感知评估高效，MarKey使用一组紧凑的代表性锚点来近似全视频覆盖，并使用一个有限的先前选择帧窗口来限制上下文相关的比较。在涵盖整体视频理解、人本视频理解和开放式视频理解的六个基准测试上的实验表明，MarKey始终优于现有方法。进一步的分析显示，在不同的MLLM骨干、模型规模和帧预算下均取得了稳健的提升。

View on arXiv Download PDF AI Translation

cs.CV / 169 / 2609.15418

ViCo-SAM3: Vision-Conditioned Alignment for Open-Vocabulary Camouflaged Object Segmentation

ViCo-SAM3：面向开放词汇的伪装物体分割的视觉条件对齐

Zhou, Qiangqiang, Tang, Wenjun, Chen, Yong, Zhu, Dandan, Xu, Jiawei

Abstract

Open-vocabulary camouflaged object segmentation (OVCOS) aims to segment unseen camouflaged objects under text guidance. We observe that SAM3 still suffers from a pronounced semantic gap between global textual semantics and fine-grained pixel-level visual cues in OVCOS. Meanwhile, fully fine-tuning the text encoder introduces heavy parameter overhead and risks overfitting to training categories, which compromises open-vocabulary representation flexibility. To address these issues, we propose ViCo-SAM3, a Vision-Conditioned alignment framework designed for OVCOS. Specifically, we introduce vision-conditioned (ViCo) module, which dynamically modulates text embeddings with global visual context, enabling textual representations to adapt to the current image content and thereby effectively bridging the semantic gap between vision and text. Building on this, we further design a vision-conditioned cross-modal binding (ViCoBind) module to enhance cross-modal interaction and semantic alignment between visual and textual representations. Without bells and whistles, ViCo-SAM3 achieves state-of-the-art performance on the OVCamo benchmark and demonstrates strong generalization.

Chinese Translation

开放词汇伪装物体分割（OVCOS）旨在在文本指导下分割未见的伪装物体。我们观察到，SAM3在OVCOS中仍然存在全球文本语义与细粒度像素级视觉线索之间明显的语义差距。同时，完全微调文本编码器会引入大量参数开销，并且有过拟合训练类别的风险，从而妨碍开放词汇表示的灵活性。为了解决这些问题，我们提出了ViCo-SAM3，一个为OVCOS设计的视觉条件对齐框架。具体而言，我们引入了视觉条件（ViCo）模块，该模块动态调节文本嵌入与全球视觉上下文，使文本表示能够适应当前图像内容，从而有效弥合视觉与文本之间的语义差距。在此基础上，我们进一步设计了一个视觉条件跨模态绑定（ViCoBind）模块，以增强视觉与文本表示之间的跨模态交互和语义对齐。ViCo-SAM3在OVCamo基准上实现了最先进的性能，并展示了强大的泛化能力。

View on arXiv Download PDF AI Translation

cs.CV / 170 / 2609.15427

A Conservative OCR-Enabled Workflow for R214 Sodium Screening of South African Packaged Foods

南非包装食品R214钠筛查的保守OCR启用工作流程

Nagayi, Mayimunah, Khan, Alice Scaria, Frank, Tamryn, Swart, Rina, Nyirenda, Clement

Abstract

Using food package images to monitor sodium and salt content against South Africa's R214 sodium limits is challenging when screening decisions require product identity, nutrition facts panel evidence, reporting basis, and category-specific thresholds. This study presents a conservative image-based workflow that combines region detection, optical character recognition (OCR), product identity and sodium evidence extraction, R214 category assignment, deterministic threshold comparison, and independent vision language model comparison. The evaluation used 442 packaged food products and 3 929 full package images from a real-world South African food packaging dataset. A YOLO26s small detector generated 4 195 region crops, and strict post-processing produced one sodium evidence row per product. The integrated workflow produced 290 OUTSIDE R214 SCOPE, 139 REVIEW, seven SCREEN-PASS, and six SCREEN-FAIL outcomes. The independent Qwen2.5-VL 7B vision language model workflow produced 387 OUTSIDE R214 SCOPE, 31 REVIEW, twenty SCREEN-PASS, and four SCREEN-FAIL outcomes. The workflows agreed on exact R214 category assignment for 415 of 442 products (93.9%) and on whether the assigned category was within R214 scope for 416 of 442 products (94.1%). Final screening outcome agreement was 307 out of 442 products, or 69.5%. Manual verification on 60 products showed lower strict outcome agreement than regulated status agreement, while all manual INSUFFICIENT DATA cases were kept out of SCREEN-PASS and SCREEN-FAIL by both automated workflows. The findings show that conservative image-based screening can organise package evidence, identify clear cases, and assign uncertain cases to REVIEW rather than forcing SCREEN-PASS or SCREEN-FAIL decisions.

Chinese Translation

使用食品包装图像监测南非R214钠限值的钠和盐含量具有挑战性，因为筛查决策需要产品身份、营养成分标签证据、报告依据和特定类别的阈值。本研究提出了一种保守的基于图像的工作流程，该流程结合了区域检测、光学字符识别（OCR）、产品身份和钠证据提取、R214类别分配、确定性阈值比较以及独立视觉语言模型比较。评估使用了442种包装食品和来自真实南非食品包装数据集的3,929张完整包装图像。YOLO26s小型检测器生成了4,195个区域裁剪，严格的后处理为每个产品生成了一行钠证据。集成工作流程产生了290个超出R214范围的结果、139个待审查结果、7个筛查通过结果和6个筛查失败结果。独立的Qwen2.5-VL 7B视觉语言模型工作流程产生了387个超出R214范围的结果、31个待审查结果、20个筛查通过结果和4个筛查失败结果。这些工作流程在442个产品中有415个（93.9%）的R214类别分配一致，并且在442个产品中有416个（94.1%）的分配类别是否在R214范围内一致。最终筛查结果一致性为442个产品中的307个，或69.5%。对60个产品的人工验证显示，严格结果一致性低于受监管状态一致性，而所有人工不足数据案例均被两个自动化工作流程排除在筛查通过和筛查失败之外。研究结果表明，保守的基于图像的筛查可以组织包装证据，识别明确案例，并将不确定案例分配为待审查，而不是强迫做出筛查通过或筛查失败的决策。

View on arXiv Download PDF AI Translation

cs.CV / 171 / 2609.15457

AnchorGUI: Asymmetric Memory for Dual-Scale Learning in GUI Navigation

AnchorGUI：用于图形用户界面导航的双尺度学习中的不对称记忆

Jin, Shengjie, Sun, Zelong, Xu, Hengbo, Ma, Yanbiao, Lu, Zhiwu

Abstract

Vision-Language Models (VLMs) enable autonomous GUI navigation, but agents still struggle to process and learn from dense, continuous visual histories. This bottleneck hinders both immediate error correction within a single episode (intra-trial) and experience distillation across multiple attempts (cross-trial). We trace these challenges to an empirical informational asymmetry in GUI navigation: while expected transitions can often be compressed into lightweight textual summaries, unexpected outcomes benefit from preserved screenshots as causal evidence for accurate diagnosis. Building on this insight, we propose AnchorGUI, a unified framework driven by the Cognitive State Anchor (CSA). The CSA acts as a per-step primitive that actively compares expected and observed transitions, converting passive multimodal trajectories into explicit prediction-error signals. These signals orchestrate a dual-scale learning mechanism via an asymmetric memory. For intra-trial correction, a sliding window selectively retains visual evidence for detected mismatches, providing immediate, visually-grounded feedback. For cross-trial distillation, this asymmetric memory focuses the computationally expensive credit assignment search space on likely failure steps. Experiments across four benchmarks validate the effectiveness of our approach. On AndroidWorld, AnchorGUI achieves a 57.3% success rate with a $2.4\times$ token reduction per step. Furthermore, cross-trial distillation reaches 69.2% success (+11.9% gain), significantly outperforming standard reflection methods while maintaining sub-linear context scaling.

Chinese Translation

视觉-语言模型（VLMs）使得自主图形用户界面（GUI）导航成为可能，但代理仍然难以处理和从密集、连续的视觉历史中学习。这一瓶颈阻碍了在单一试验中的即时错误修正（试内）和跨多次尝试的经验提炼（试间）。我们将这些挑战追溯到图形用户界面导航中的经验信息不对称：虽然预期的过渡通常可以压缩为轻量级的文本摘要，但意外结果则受益于保留的屏幕截图，作为准确诊断的因果证据。基于这一洞察，我们提出了AnchorGUI，一个由认知状态锚（Cognitive State Anchor, CSA）驱动的统一框架。CSA作为逐步原语，主动比较预期和观察到的过渡，将被动的多模态轨迹转化为明确的预测误差信号。这些信号通过不对称记忆协调双尺度学习机制。对于试内修正，滑动窗口选择性地保留检测到的不匹配的视觉证据，提供即时的、以视觉为基础的反馈。对于试间提炼，这种不对称记忆将计算上昂贵的信用分配搜索空间集中在可能的失败步骤上。四个基准实验验证了我们方法的有效性。在AndroidWorld上，AnchorGUI实现了57.3%的成功率，并且每步的令牌减少达到$2.4 imes$。此外，试间提炼的成功率达到69.2%（+11.9%的增益），显著优于标准反思方法，同时保持亚线性的上下文扩展。

View on arXiv Download PDF AI Translation

cs.CV / 172 / 2609.15478

BVB: Benchmarking Agentic Video Understanding via Programmatic Reconstruction in Blender

BVB：通过Blender中的程序重建对代理视频理解的基准测试

Tang, Yolo Y., Shimada, Daiki, Meng, Jiayue, Bi, Jing, Liu, Pinxin, Wang, Yicheng, Xiao, Yunzhong, Tan, Zhangyun, Zhang, Zeliang, Huang, Chao, Liang, Susan, Shen, Qianxiang, Song, Luchuan, Vosoughi, Ali, Feng, Mingqian, Filvantorkaman, Melika, Xu, Chenliang

Abstract

Multimodal agents can create complex videos in software such as Blender by coding without relying on diffusion models. Yet video understanding benchmarks still evaluate models mainly through question answering. If an agent truly understands a video, it can reconstruct it programmatically. We introduce BVB, Blender-VideoBench, a benchmark that tests this ability by asking agents to reconstruct real-world videos as animated Blender scenes. To ensure fair comparison, each agent programs the reconstruction through a lightweight harness, Mini-BVB, in an identical sandbox under a shared cost limit. The benchmark renders each reconstruction from its animated camera and evaluates it on two axes: (1) Dual VQA measures how many spatiotemporal facts the reconstruction preserves. (2) Latent Similarity measures how closely the reconstruction matches the source video perceptually. Our overall score, a square-root mean, favors balanced performance. We evaluate 51 configurations from 10 model families and analyze semantic retention, perceptual similarity, reasoning effort, and cost. The best model reaches 88.6 Latent Similarity but retains only 53.7% of the source-correct spatiotemporal answers. Additional reasoning improves visual similarity but does not close this gap in factual accuracy. In a blind study with 15 raters and five configurations, Latent Similarity correlates strongly with human preference. These results show that programmatic reconstruction is a viable test of agentic video understanding, and that semantic retention remains the main challenge.

Chinese Translation

多模态代理可以通过编码在Blender等软件中创建复杂视频，而无需依赖扩散模型。然而，视频理解基准仍主要通过问答来评估模型。如果一个代理真正理解视频，它就能够以编程方式重建视频。我们引入了BVB（Blender-VideoBench），这是一个通过要求代理将真实世界的视频重建为动画Blender场景来测试这种能力的基准。为了确保公平比较，每个代理在共享成本限制下，通过轻量级的工具Mini-BVB在相同的沙盒中进行重建编程。该基准从其动画相机渲染每个重建，并在两个维度上进行评估：（1）双重视频问答（Dual VQA）衡量重建保留了多少时空事实；（2）潜在相似性（Latent Similarity）衡量重建在感知上与源视频的匹配程度。我们的总体得分是平方根平均值，偏向于平衡性能。我们评估了来自10个模型家族的51种配置，并分析了语义保留、感知相似性、推理努力和成本。最佳模型达到了88.6的潜在相似性，但仅保留了53.7%的源正确时空答案。额外的推理提高了视觉相似性，但并未缩小事实准确性的差距。在一项包含15名评审员和五种配置的盲测中，潜在相似性与人类偏好高度相关。这些结果表明，程序重建是测试代理视频理解的可行方法，而语义保留仍然是主要挑战。

View on arXiv Download PDF AI Translation

cs.CV / 173 / 2609.15484

CatchMonitor: a machine learning system for automated fish discard quantification

CatchMonitor：一种用于自动化鱼类弃捕量化的机器学习系统

French, Geoff, Mackiewicz, Michal, Fisher, Mark, Holah, Helen, Lamb, Rebecca

Abstract

We report on the continued development of CatchMonitor, resulting in a prototype computer vision system designed to automatically quantify discarded fish from video footage collected from Remote Electronic Monitoring (REM) systems on fishing trawlers. The analysis of trawler surveillance footage is a challenging problem due to the real-world conditions on board fishing vessels. Building on our prior work we improve the accuracy of species identification through the application of semi-supervised learning. We utilise a simple and robust object tracking approach, upon which we build our prototype discard quantification system. Finally we analyse the variability of manual discard quantification performed by multiple expert human analysts, using it as a benchmark against which we compare the performance of our system.

Chinese Translation

我们报告了CatchMonitor的持续开发，形成了一个原型计算机视觉系统，旨在自动量化从渔船的远程电子监控（REM）系统收集的视频资料中弃捕的鱼类。由于渔船上的现实条件，分析拖网监控视频是一项具有挑战性的任务。在我们之前工作的基础上，我们通过应用半监督学习提高了物种识别的准确性。我们采用了一种简单而稳健的目标跟踪方法，并在此基础上构建了我们的原型弃捕量化系统。最后，我们分析了多位专家人类分析师进行的手动弃捕量化的变异性，并将其作为基准，与我们的系统性能进行比较。

View on arXiv Download PDF AI Translation

cs.CV / 174 / 2609.15503

SyntheticDoc: A Large Synthetic Dataset for Document Unwarping and Illumination Correction

SyntheticDoc：一个用于文档去畸变和光照校正的大型合成数据集

Woortmann, Daniel, Magne, Tanguy, Sorkine-Hornung, Olga

Abstract

Deep learning models have become the standard tool for document rectification and illumination correction, yet their performance is fundamentally bound by their training data. For nearly a decade, the community has heavily relied on Doc3D, a pioneering but increasingly limited document unwarping dataset in terms of scale and quality. To address this bottleneck, we introduce SyntheticDoc, a massive, high-quality dataset designed to push the boundaries of document unwarping. SyntheticDoc is composed of 1,000,000 high-resolution procedurally generated training samples, alongside extensive validation and test sets. Each sample is paired with rich, pixel-perfect annotations, including UV maps, normal maps, albedo and shading. To ensure physical accuracy and photorealism, the paper geometries are generated via a physics-based simulator and rendered using a path tracer. To demonstrate the benefit of our dataset, we train a simple baseline model on SyntheticDoc and report on its performance in comparison to state-of-the-art methods on both document unwarping and illumination correction tasks. Our dataset is available at https://igl.ethz.ch/projects/SyntheticDoc/ and the code used to generate it at https://github.com/tanguymagne/SyntheticDoc .

Chinese Translation

深度学习模型已成为文档矫正和光照校正的标准工具，但其性能在根本上受到训练数据的限制。在近十年的时间里，研究界在很大程度上依赖于Doc3D，这是一个开创性的文档去畸变数据集，但在规模和质量上逐渐显得有限。为了解决这一瓶颈，我们引入了SyntheticDoc，这是一个大规模、高质量的数据集，旨在推动文档去畸变的边界。SyntheticDoc由1,000,000个高分辨率的程序生成训练样本组成，并附有广泛的验证和测试集。每个样本都配有丰富的像素级注释，包括UV图、法线图、反照率和阴影。为了确保物理准确性和照片真实感，论文中的几何体是通过基于物理的模拟器生成的，并使用路径追踪器进行渲染。为了展示我们数据集的优势，我们在SyntheticDoc上训练了一个简单的基线模型，并报告其在文档去畸变和光照校正任务中与最先进方法的性能比较。我们的数据集可在https://igl.ethz.ch/projects/SyntheticDoc/获取，生成代码可在https://github.com/tanguymagne/SyntheticDoc找到。

View on arXiv Download PDF AI Translation

cs.CV / 175 / 2609.15524

Assessing nnU-Net Generalization across Brain Tumor Populations in BraTS-GoAT 2026

评估 nnU-Net 在 BraTS-GoAT 2026 中对脑肿瘤人群的泛化能力

Kirscher, Tristan, Metzger, Vivian, Meyer, Philippe, Coubez, Xavier

Abstract

BraTS-GoAT evaluates tumor segmentation across heterogeneous populations. We trained a conventional 3D nnU-Net on 1,351 labeled cases using five-fold cross-validation and 1,000 epochs per fold. The final predictor averaged all folds and applied test-time mirroring. On pooled official validation, global DSC values were 0.7805, 0.8288, and 0.8854 for enhancing tumor (ET), tumor core (TC), and whole tumor (WT). Under matched fold-0 inference, mean regional Dice decreased from 0.9058 on source out-of-fold (OOF) cases to 0.8310 on pooled validation (difference--0.0747). Mirroring gave small single-fold gains but no clear ensemble benefit; a residual-encoder alternative reached 0.8282 mean Dice. In labeled OOF predictions, failure cases had substantially smaller reference ET volumes; after adjustment for ET and WT volume, lower Dice remained associated with more disconnected ET components and a smaller fraction of ET contained in the largest component.

Chinese Translation

BraTS-GoAT 评估了异质人群中的肿瘤分割。我们使用五折交叉验证和每折 1,000 个训练周期，在 1,351 个标注案例上训练了一个传统的 3D nnU-Net。最终预测器对所有折进行了平均，并应用了测试时镜像。在汇总的官方验证中，增强肿瘤（ET）、肿瘤核心（TC）和整个肿瘤（WT）的全局 DSC 值分别为 0.7805、0.8288 和 0.8854。在匹配的折-0 推断下，源外折（OOF）案例的平均区域 Dice 从 0.9058 降低到 0.8310（差异为 -0.0747）。镜像带来了小幅的单折增益，但没有明显的集成收益；一种残差编码器替代方案达到了 0.8282 的平均 Dice。在标注的 OOF 预测中，失败案例的参考 ET 体积显著较小；在调整 ET 和 WT 体积后，较低的 Dice 仍与更多不连通的 ET 组件以及较小的 ET 被包含在最大组件中的比例相关。

View on arXiv Download PDF AI Translation

cs.CV / 176 / 2609.15562

PIVOT: Physics-Grounded Verification for AI-Generated Audio-Video Detection

PIVOT：基于物理的人工智能生成音视频检测验证

Zheng, Bo, Zhao, Kangran, Zhang, Xiaoyu, Guan, Weinan, Li, Zhiheng, Chen, Yize, Li, Haizhou, Liu, Qingshan, Lyu, Siwei, Wu, Baoyuan

Abstract

As generative models continue to advance, AI-generated content (AIGC) is becoming increasingly realistic, weakening the artifact cues commonly exploited by existing detectors. Nevertheless, faithfully reproducing the physical behavior of real-world events remains challenging for current generators. We therefore explore detecting AIGC by assessing whether the depicted event satisfies measurable constraints derived from physical laws. We introduce PIVOT, a physics-grounded AIGC detector, instantiated here for audio-video clips, that estimates physical quantities from video and audio, selects physical laws relevant to each clip, and verifies their measurable constraints. Beyond a real/fake decision, PIVOT returns supporting evidence that records the verification outcome, relevant time window, and supporting quantities for each applicable law. Although instantiated and evaluated here on audio-video data, the framework can, in principle, extend to other AIGC modalities whenever the physical quantities required for verification can be estimated reliably. We also introduce PhysForensics-Bench, comprising paired real and generated audio-video clips from nine event-centric scene families and two recent audio-video generators. On PhysForensics-Bench, PIVOT achieves 70.30% accuracy and 64.29% F1 score on Real+Seedance, and 72.16% accuracy and 65.82% F1 on Real+VEO. In comparison, direct inspection with Gemini 3.1 Pro obtains 53.96% accuracy and 60.09% F1 on Real+Seedance, and 57.22% accuracy and 63.44% F1 on Real+Veo. These results demonstrate the practical promise of physical-consistency verification as a structured and inspectable source of evidence that complements artifact-based AIGC detection.

Chinese Translation

随着生成模型的不断进步，人工智能生成内容（AIGC）变得越来越逼真，削弱了现有检测器常用的伪影线索。然而，对于当前的生成器来说，忠实再现现实世界事件的物理行为仍然具有挑战性。因此，我们探索通过评估所描绘事件是否满足源自物理法则的可测量约束来检测AIGC。我们引入了PIVOT，一个基于物理的AIGC检测器，此处针对音视频片段进行实例化，它从视频和音频中估计物理量，选择与每个片段相关的物理法则，并验证其可测量约束。除了真实/虚假决策外，PIVOT还返回支持证据，记录验证结果、相关时间窗口以及每个适用法则的支持量。尽管在此针对音视频数据进行了实例化和评估，但该框架原则上可以扩展到其他AIGC模态，只要所需的物理量能够可靠地估计。我们还引入了PhysForensics-Bench，包括来自九个事件中心场景家族和两个最新音视频生成器的真实和生成音视频片段配对。在PhysForensics-Bench上，PIVOT在Real+Seedance上实现了70.30%的准确率和64.29%的F1分数，在Real+VEO上实现了72.16%的准确率和65.82%的F1分数。相比之下，使用Gemini 3.1 Pro的直接检查在Real+Seedance上获得了53.96%的准确率和60.09%的F1分数，在Real+Veo上获得了57.22%的准确率和63.44%的F1分数。这些结果展示了物理一致性验证作为一种结构化和可检查的证据来源的实际潜力，能够补充基于伪影的AIGC检测。

View on arXiv Download PDF AI Translation

cs.CV / 177 / 2609.15579

What a gated sensing pipeline never looks at: bandwidth reduction and the misses behind it

门控感知流程从未关注的内容：带宽压缩及其背后的遗漏

Venkat, Raghu, Anjali, Tricha

Abstract

An airborne sensor on a contested link cannot send video, so the appealing move is to send findings instead and report the ratio between the two. We evaluate a gated sensing pipeline that does this, combining learned object detection and image-text comparison with deterministic scheduling, gating, evidence accumulation and transmission rules. On staged footage with the semantic stage live it sends 38,736 bits over 211 s, a reduction of 41,977x, and names 1 of 4 staged events with no false report. That detection has since been superseded: a correction to how the tracker measures speed removed the measurement artefact the normality model had been learning from, and the flight no longer warms. Under the corrected code the pipeline names 1 of 7 staged events across four flights, and we report both. On a control flight where nothing was staged it reports nothing, a reduction of 155,830x: the largest number in the study and the least informative, because a reduction ratio measures the scene. The mechanisms that produce the reduction also decide which observations ever reach a decision, so the two cannot be reported apart. We give a tick-level trace of one flight (3,187 rows) that places each of three missed events at the stage where it stopped progressing: one produced no track, one failed the structural place test at 0.129, and one passed 539 structural ticks but reached only 3.192 against a boundary of 3.807. We also report one instance of a known failure mode, an online normality model absorbing the object it will later judge, measured against the threshold that object then failed. The evidence is one detection and six misses across four staged flights, beside one clean control, and we treat it as a case study. We give the reproduction protocol and generated results, identify which supporting artifacts are not distributed, and state which experiments did not run.

Chinese Translation

在受限链路上的机载传感器无法传输视频，因此一种可行的做法是直接发送检测结果，并报告两者之间的比例。我们评估了一种门控感知流程，该流程结合了学习型目标检测与图像-文本比对，配合确定性调度、门控、证据累积及传输规则。在带有语义阶段实时处理的分段视频中，该流程在211秒内发送了38,736比特，带宽压缩率达到41,977倍，并准确报告了4个分段事件中的1个，且无误报。此检测方法随后被更新：对跟踪器速度测量的修正消除了正态性模型所学习的测量伪影，飞行不再出现预热现象。在修正后的代码下，该流程在四次飞行中准确报告了7个分段事件中的1个，我们同时报告了两种结果。在一场未布置任何事件的控制飞行中，系统未报告任何信息，带宽压缩率达到155,830倍——这是本研究中最高的压缩比，但信息量最低，因为压缩比反映的是场景内容。产生压缩的机制同时决定了哪些观测能被最终决策采纳，因此两者不可分割。我们提供了一次飞行的逐刻度跟踪记录（3,187行），将三个遗漏事件分别定位于其停止进展的阶段：一个未产生轨迹，一个在0.129时未通过结构位置测试，另一个通过了539个结构刻度但仅达到3.192，未达到3.807的边界值。我们还报告了一个已知失败模式的实例，即在线正态性模型吸收了其后将判定的目标，该目标随后未通过阈值检测。证据包括四次分段飞行中的一次检测和六次遗漏，以及一次干净的控制飞行，我们将其作为案例研究。文中给出了复现协议和生成结果，指出了未分发的支持性工件，并说明了未执行的实验。

View on arXiv Download PDF AI Translation

cs.CV / 178 / 2609.15603

A Unified Vision-Language Model for PSMA PET/CT Report Generation, Visual Question Answering, and Lesion Segmentation

用于PSMA PET/CT报告生成、视觉问答和病灶分割的统一视觉-语言模型

Xing, Yang, Wu, Jiong, Ozdemir, Savas, Zhou, Yang, Yu, Boxiao, Zhang, Ying, Zhu, Zheren, You, Chenyu, Shao, Wei, Lu, Yang, Wang, Kang, Pan, Tinsu, Yang, Yang, Gong, Kuang

Abstract

Accurate PSMA PET/CT interpretation is central to prostate cancer management, yet existing PET/CT AI models typically address isolated tasks. We propose a unified PSMA PET/CT vision-language model for report generation, visual question answering, and lesion segmentation. The framework adopts an LLaVA-style architecture, comprising a PET/CT vision encoder, an MLP-Mixer projection module, a LoRA-tuned large language model, and a 3D segmentation branch. Training followed a four-stage strategy: vision encoder pretraining, projection-layer alignment, VLM fine-tuning, and final multitask tuning. Language tasks used 5,747 PSMA PET/CT datasets with paired reports, while segmentation used the PSMA subset of AutoPET. The model outperformed PET2REP and a CT-based baseline across standard report-generation metrics, improved performance across VQA question types, and achieved higher Dice and lesion-level overlap F1 than SegAnyPET and nnUNet. These results support the feasibility of a unified framework for structured, interactive, interpretable PSMA PET/CT analysis with voxel-level grounding within a single multitask model architecture.

Chinese Translation

准确的PSMA PET/CT解读对前列腺癌管理至关重要，但现有的PET/CT人工智能模型通常只针对孤立任务。我们提出了一种统一的PSMA PET/CT视觉-语言模型，用于报告生成、视觉问答和病灶分割。该框架采用LLaVA风格的架构，包括一个PET/CT视觉编码器、一个MLP-Mixer投影模块、一个经过LoRA调优的大型语言模型和一个3D分割分支。训练遵循四阶段策略：视觉编码器预训练、投影层对齐、VLM微调和最终的多任务调优。语言任务使用了5,747个配对报告的PSMA PET/CT数据集，而分割任务则使用了AutoPET的PSMA子集。该模型在标准报告生成指标上超越了PET2REP和基于CT的基线，在VQA问题类型上提高了性能，并在Dice和病灶级重叠F1上超过了SegAnyPET和nnUNet。这些结果支持了在单一多任务模型架构中实现结构化、互动式、可解释的PSMA PET/CT分析的统一框架的可行性。

View on arXiv Download PDF AI Translation

cs.CV / 179 / 2609.15606

VideoScout: Learning Agentic Active Exploration with Adaptive Reasoning Pacing for Long Video Understanding

VideoScout：具有自适应推理节奏的主动探索学习用于长视频理解

Xu, Weixin, Yang, Zhenyu, Wang, Bing, Qian, Shengsheng, Xu, Changsheng

Abstract

Multimodal Large Language Models (MLLMs) have achieved remarkable progress on short video understanding yet remain limited on long videos due to the limited visual context window. Prevailing approaches rely on uniform frame sampling or recent coarse-to-fine agentic zooming, both of which struggle to localize sparse, decisive evidence in sufficiently long videos. We formulate long video understanding as a \textbf{Sequential Evidence Acquisition (SEA)} problem, in which an agent reads the video turn by turn along the temporal axis, deciding at each turn how fast to watch, what evidence to retain, when to revisit uncertain segments, and when to stop and answer. Inspired by this view, we propose \textbf{VideoScout}, a multi-turn reasoning agent that instantiates the SEA paradigm through adaptive reasoning pacing. Specifically, by dynamically controlling the viewing pace, VideoScout enables efficient traversal of long videos within a bounded visual context window, allowing the agent to access more video content while balancing content analysis depth with reading efficiency. To train VideoScout, we construct VideoScout-66K, a set of over 66K high-quality exploration turns from 10K answer-verified trajectories, and adopt a two-stage pipeline: cold-start supervised fine-tuning teaches the agent per-turn output format, while the Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO) algorithm performs trajectory-level reinforcement learning with a composite reward that jointly considers answer accuracy, output format compliance, and the temporal alignment between the agent's viewing progress and the teacher's answer timing measured by intersection-over-union (IoU). Extensive experiments on long video understanding and reasoning benchmarks demonstrate that our 7B model achieves strong performance compared with existing trained 7B agentic models.

Chinese Translation

多模态大型语言模型（MLLMs）在短视频理解方面取得了显著进展，但由于视觉上下文窗口的限制，在长视频上的表现仍然有限。现有方法依赖于均匀帧采样或最近的粗到细的主动缩放，这两者都难以在足够长的视频中定位稀疏的、决定性的证据。我们将长视频理解形式化为一个 extbf{顺序证据获取（SEA）}问题，在该问题中，代理沿时间轴逐帧阅读视频，在每一帧中决定观看的速度、保留哪些证据、何时重新访问不确定的片段以及何时停止并回答。受到这一观点的启发，我们提出了 extbf{VideoScout}，一个通过自适应推理节奏实现SEA范式的多轮推理代理。具体而言，通过动态控制观看节奏，VideoScout能够在有限的视觉上下文窗口内高效遍历长视频，使代理能够访问更多视频内容，同时平衡内容分析的深度与阅读效率。为了训练VideoScout，我们构建了VideoScout-66K，这是一个由10K经过答案验证的轨迹生成的超过66K高质量探索回合的集合，并采用了两阶段的流程：冷启动监督微调教会代理每回合的输出格式，而解耦剪辑和动态采样策略优化（DAPO）算法则通过复合奖励进行轨迹级强化学习，该奖励共同考虑答案准确性、输出格式合规性以及代理的观看进度与教师答案时机之间的时间对齐，通过交并比（IoU）进行测量。在长视频理解和推理基准上的广泛实验表明，我们的7B模型与现有的训练过的7B主动模型相比，表现出色。

View on arXiv Download PDF AI Translation

cs.CV / 180 / 2609.15635

ModaLens: Measuring Image Sensitivity in Report-Conditioned Medical VLMs

ModaLens：测量报告条件下医学视觉语言模型的图像敏感性

Ordóñez, Sebastián Andrés Cajas, Lange, Maximin, Bui, Quang, Li, Anqi Peter, Osorio, Felipe Ocampo, Attrach, Rafi Al, Palakala, Kushul Reddy, Kapadia, Sahil, Sellami, Zakaria Laouabdia, Zhang, Xinyue, Zhang, Ashley, Celi, Leo Anthony

Abstract

A radiology report can already answer a clinical question, so it is hard to tell whether a vision-language model also uses the image. ModaLens, a paired image-swap audit, measures how report availability changes image sensitivity: MedGemma-27B on 3,199 paired MIMIC-CXR cases from 293 patients, all 14 questions per case (13 finding-specific and one composite), each image replaced by one from another study, usually of the same patient, with question and report fixed. Under an explicit answer instruction, the model's generated answer changes on 4.26 percent of trials with the report and 20.94 percent without it, a paired increase of 16.7 points (patient-clustered 95 percent CI 15.6 to 17.7), so report availability reduces image-swap sensitivity under this protocol; the original prompt with a lowercase first-token readout gives 4.70 percent against 17.07 percent, and substitutions also move continuous answer scores where the binary prediction does not change. The labels are derived from reports, which limits conclusions about visual correctness; the direction replicates in two further model lineages. Code, the exact prompts and a run record for every number are at https://github.com/criticaldata/MODALENS.

Chinese Translation

放射学报告已经能够回答临床问题，因此很难判断视觉语言模型是否也利用了图像。ModaLens是一种成对图像交换审计，测量报告可用性如何改变图像敏感性：在来自293名患者的3,199对MIMIC-CXR案例中使用MedGemma-27B，针对每个案例的14个问题（13个特定发现和一个综合问题），每个图像被替换为来自另一项研究的图像，通常是同一患者的图像，问题和报告保持不变。在明确的回答指令下，模型生成的答案在有报告的情况下变化了4.26%的试验，而在没有报告的情况下变化了20.94%，成对增加了16.7个百分点（患者聚类的95%置信区间为15.6到17.7），因此在该协议下，报告的可用性降低了图像交换的敏感性；原始提示的首个小写标记读出显示为4.70%对17.07%，而替换也会影响连续答案分数，即使二元预测没有变化。标签来源于报告，这限制了关于视觉正确性的结论；该方向在另外两个模型谱系中得到了复制。代码、确切的提示和每个数字的运行记录可在https://github.com/criticaldata/MODALENS获取。

View on arXiv Download PDF AI Translation

cs.CV / 181 / 2609.15639

SAM3D-Part: Interactive Part Selection and Generation from 3D Objects

SAM3D-Part：从3D对象中进行交互式部件选择与生成

Chang, Jiahao, Du, Dong, Sun, Wanhu, Zheng, Yujian, Pan, Chuanyu, Zhao, Bowen, Ye, Chongjie, Hu, Yuanming, Han, Xiaoguang

Abstract

Part-level control is essential for modern 3D asset creation, where objects are frequently edited, reused, animated, or fabricated through their individual components. In many such workflows, users need only several specific components rather than a complete object decomposition. However, existing 3D generation methods produce all parts regardless of user intent, while promptable 3D segmentation methods typically output partial surfaces instead of reusable complete meshes. In addition, image-conditioned part generators further struggle to preserve hidden geometry and accurate placement without directly conditioning on the source mesh. To address these problems, we present SAM3D-Part, a prompt-driven framework for selective part generation from input 3D object meshes. Given a source mesh and a part prompt, SAM3D-Part first encodes the source geometry into compact mesh features and aligns them with the rendered image, selective mask, and point-map observations via pixel-wise channel fusion. The fused representation conditions a feed-forward generative model to produce only the queried component as a completed mesh. To place the generated part back into the source coordinate frame, SAM3D-Part predicts dense per-voxel correspondences and estimates the part transformation from distributed spatial evidence rather than a single global pose code. For sequential multi-part queries, previously generated parts are stored in a part cache and reused as contextual constraints, reducing conflicts among independently requested components. Extensive experiments and ablations demonstrate that SAM3D-Part can significantly improve source alignment, reduce conditioning cost, and enable consistent selective part generation, achieving state-of-the-art. Code and weights will be available at https://github.com/Jiahao620/sam3d-part.

Chinese Translation

部件级控制对于现代3D资产创建至关重要，在这一过程中，物体的个别组件经常被编辑、重用、动画化或制造。在许多这样的工作流程中，用户只需要几个特定的组件，而不是完整的物体分解。然而，现有的3D生成方法会无视用户意图而生成所有部件，而可提示的3D分割方法通常输出部分表面而非可重用的完整网格。此外，图像条件的部件生成器在没有直接依据源网格进行条件化的情况下，进一步难以保持隐藏几何形状和准确的放置。为了解决这些问题，我们提出了SAM3D-Part，一个基于提示的框架，用于从输入的3D对象网格中选择性生成部件。给定源网格和部件提示，SAM3D-Part首先将源几何体编码为紧凑的网格特征，并通过逐像素通道融合将其与渲染图像、选择性掩膜和点图观察对齐。融合后的表示条件化一个前馈生成模型，仅生成所查询的组件作为完整网格。为了将生成的部件放回源坐标系，SAM3D-Part预测密集的每体素对应关系，并从分布的空间证据中估计部件变换，而不是依赖单一的全局姿态编码。对于顺序的多部件查询，之前生成的部件存储在部件缓存中，并作为上下文约束重用，从而减少独立请求组件之间的冲突。大量实验和消融研究表明，SAM3D-Part能够显著改善源对齐，降低条件化成本，并实现一致的选择性部件生成，达到最先进的水平。代码和权重将可在 https://github.com/Jiahao620/sam3d-part 获取。

View on arXiv Download PDF AI Translation

cs.CV / 182 / 2609.15640

Human-Grounded Calibration for Long-Text Image-Text Congruence in Vision-Language Models

基于人类的长文本图像-文本一致性校准在视觉-语言模型中的应用

Gambetti, Alessandro, Han, Qiwei

Abstract

Long-text image--text congruence scoring is increasingly important for vision-language systems that must evaluate whether detailed textual descriptions match visual content. However, raw similarity scores from dual-encoder models are difficult to interpret as calibrated congruence measures, especially under the modality gap between image and text embeddings. This paper proposes Congruency Score (CS), a lightweight calibration layer that maps image--text similarity evidence into a bounded score. Using DOCCI and Urban1k, we evaluate four frozen vision-language backbones and show that observed reductions in post-projection centroid distance do not uniformly improve image--text retrieval performance. Human-grounded evaluations on DOCCI further reveal a trade-off: direct post-hoc calibration preserves high association with human judgments, whereas selected projection-based configurations can reduce threshold-relevant slope and intercept distortions at the cost of retrieval performance and association strength. These results establish long-text image--text congruence scoring as a calibrated score-estimation problem, where retrieval performance, human association, and threshold calibration must be evaluated as distinct objectives. CS provides a lightweight way to expose and operationalize this separation.

Chinese Translation

长文本图像-文本一致性评分在视觉-语言系统中变得愈发重要，这些系统必须评估详细的文本描述是否与视觉内容相匹配。然而，来自双编码器模型的原始相似性评分难以解释为经过校准的一致性度量，特别是在图像和文本嵌入之间存在模态差异的情况下。本文提出了一种一致性评分（Congruency Score, CS），这是一种轻量级的校准层，将图像-文本相似性证据映射为一个有界评分。通过使用 DOCCI 和 Urban1k 数据集，我们评估了四个冻结的视觉-语言主干网络，并显示出观察到的后投影质心距离的减少并不均匀改善图像-文本检索性能。对 DOCCI 的基于人类的评估进一步揭示了一种权衡：直接的事后校准保持了与人类判断的高度关联，而选择的基于投影的配置可以在牺牲检索性能和关联强度的情况下减少与阈值相关的斜率和截距失真。这些结果确立了长文本图像-文本一致性评分作为一个经过校准的评分估计问题，其中检索性能、人类关联和阈值校准必须被评估为不同的目标。CS 提供了一种轻量级的方法来揭示并操作这一分离。

View on arXiv Download PDF AI Translation

cs.CV / 183 / 2609.15647

MedDiME: Efficient Latent Diffusion with Adaptive Masking for Medical Counterfactual Generation

MedDiME：用于医学反事实生成的自适应掩蔽高效潜在扩散

Zeng, Yan, Guo, Changlu, Christensen, Anders Nymark, Hannemose, Morten Rieger, Dahl, Anders Bjorholm

Abstract

Medical counterfactual generation modifies images to change model predictions for interpretability. However, existing diffusion-based approaches are often prohibitively slow and memory-intensive, making them difficult to apply in high-resolution settings. Moreover, existing masking strategies are tightly coupled with pixel-space representations, making them incompatible with latent-space diffusion editing. To address these challenges, we propose MedDiME, a latent-space classifier-guided diffusion framework that reduces computational and memory overhead while introducing a latent-compatible, gradient-driven adaptive masking mechanism for spatially precise medical counterfactual generation. Extensive experiments demonstrate that MedDiME achieves high-quality counterfactual generation with significant efficiency gains compared to prior classifier-guided diffusion baselines, achieving up to 40 times faster inference and 13 times lower peak GPU memory usage.

Chinese Translation

医学反事实生成通过修改图像来改变模型预测，以提高可解释性。然而，现有的基于扩散的方法通常速度极慢且内存消耗巨大，使其难以在高分辨率环境中应用。此外，现有的掩蔽策略与像素空间表示紧密耦合，因此与潜在空间的扩散编辑不兼容。为了解决这些挑战，我们提出了MedDiME，这是一种潜在空间分类器引导的扩散框架，旨在减少计算和内存开销，同时引入一种潜在兼容的、基于梯度驱动的自适应掩蔽机制，以实现空间精确的医学反事实生成。大量实验表明，与之前的分类器引导的扩散基线相比，MedDiME在生成高质量反事实的同时显著提高了效率，实现了最高40倍的推理速度提升和13倍的峰值GPU内存使用降低。

View on arXiv Download PDF AI Translation

cs.CV / 184 / 2609.15649

From Model Patterns to Abstract Semantics in Compositional Zero-Shot Learning

从模型模式到组合零样本学习中的抽象语义

Li, Weize, Zhao, Zhicheng, Su, Fei

Abstract

Compositional Zero Shot Learning aims to recognize unseen compositions by recombining learned primitives. Recent methods rely on vision language models and attempt to explicitly model contextual variations of primitives through multiple representations. However, such approaches are limited by fixed variant capacity and competition between abstract and concrete semantics. In this work, we present a new perspective that views primitive variations as the context-driven activation of concrete visual cues rather than independent entities. Based on it, we propose CLEAR, a CLoze-style rEAsoning-based Re-ranking framework inspired by human perceptual processes. CLEAR extracts conditional variants from the primitive candidate set in a coarse-to-fine manner, performs cloze-style reasoning to infer high-level semantics, and re-ranks predictions to correct biases toward salient concrete primitives. Extensive experiments demonstrate that CLEAR consistently improves the Base Model and outperforms state-of-the-art methods on the challenging C-GQA and MIT-States datasets. Code is available at https://github.com/buptLwz/CLEAR.

Chinese Translation

组合零样本学习旨在通过重新组合已学习的原始元素来识别未见的组合。近期的方法依赖于视觉语言模型，并试图通过多种表示方式显式建模原始元素的上下文变化。然而，这些方法受到固定变体容量的限制，并且在抽象语义与具体语义之间存在竞争。在本研究中，我们提出了一种新视角，将原始元素的变化视为具体视觉线索的上下文驱动激活，而非独立实体。基于此，我们提出了CLEAR，一个基于人类感知过程的Cloze风格推理重排序框架。CLEAR以粗到细的方式从原始候选集提取条件变体，进行cloze风格推理以推断高层语义，并重新排序预测以纠正对显著具体原始元素的偏见。大量实验表明，CLEAR持续改善基础模型，并在具有挑战性的C-GQA和MIT-States数据集上超越了最先进的方法。代码可在https://github.com/buptLwz/CLEAR获取。

View on arXiv Download PDF AI Translation

cs.CV / 185 / 2609.15668

Circuit-MLLM: Topological Logic-Guided Latent-Space Visual Reasoning for Circuit Schematic Understanding

Circuit-MLLM：基于拓扑逻辑引导的电路原理图理解的潜在空间视觉推理

Deng, Jinyuan, Jiang, Yuqi, Huang, Wenjing, Li, Xin, Sun, Qi, Zhuo, Cheng

Abstract

Through pre-training on extensive text and image datasets, current multi-modal large language models (MLLMs) achieve strong performance on general tasks. However, circuit schematics present a unique challenge for MLLMs due to their dense component layouts and distinct topological logic, demanding fine-grained structural parsing to extract the electrical semantics. To address this, we propose Circuit-MLLM, a multimodal reasoning framework that reformulates circuit topology analysis as a process of device localization, path tracing, and sequential reasoning within the latent space. We introduce a circuit knowledge mining mechanism that deeply aligns the model's latent representations with structurally rich features derived from multi-granularity circuit vision experts, enabling the model to effectively internalize topological semantics. Building upon these internalized semantics, we devise a topology-guided sequencing strategy that decouples reasoning from the rigid raster-scan order, enforcing stepwise inference along the circuit's topological logic in latent space. Across diverse circuit analysis tasks, Circuit-MLLM consistently outperforms strong baselines, notably achieving a 25% higher average score than GPT-5.1, which demonstrates the effectiveness of our framework in circuit schematic topology analysis. Code is publicly available at https://github.com/IC-Yuan/Circuit-MLLM.

Chinese Translation

通过在大量文本和图像数据集上进行预训练，当前的多模态大型语言模型（MLLMs）在一般任务上取得了强劲的表现。然而，电路原理图由于其密集的组件布局和独特的拓扑逻辑，对MLLMs提出了独特的挑战，要求进行细致的结构解析以提取电气语义。为了解决这个问题，我们提出了Circuit-MLLM，一个多模态推理框架，将电路拓扑分析重新定义为设备定位、路径追踪和潜在空间内的顺序推理过程。我们引入了一种电路知识挖掘机制，深度对齐模型的潜在表示与来自多粒度电路视觉专家的结构丰富特征，使模型能够有效内化拓扑语义。在这些内化语义的基础上，我们设计了一种拓扑引导的序列策略，将推理与严格的光栅扫描顺序解耦，强制在潜在空间中沿电路的拓扑逻辑进行逐步推理。在多样的电路分析任务中，Circuit-MLLM始终超越强基线，特别是在GPT-5.1的基础上平均得分提高了25%，这证明了我们框架在电路原理图拓扑分析中的有效性。代码可在 https://github.com/IC-Yuan/Circuit-MLLM 获取。

View on arXiv Download PDF AI Translation

cs.CV / 186 / 2609.15669

Benchmarking Intra-Patient 3D Deformable Multimodal Image Registration

患者内部三维可变形多模态图像配准的基准测试

Barbieri, Matteo, La Barbera, Giammarco, De La Plata, Juan Pablo, Sarnacki, Sabine, Bloch, Isabelle, Gori, Pietro

Abstract

Multimodal image registration is a key component of many clinical workflows, yet it remains challenging because corresponding anatomical structures often exhibit substantially different image intensities across modalities. In this work, we present a comprehensive benchmark of intra-patient 3D multimodal deformable registration methods across three datasets covering different anatomical regions and difficulty levels, including both synthetic deformation recovery and real clinical scenarios. We evaluate classical optimization-based approaches and modern learning-based methods, including recent deep learning and foundation models, using complementary metrics: Average Dice similarity coefficient (DSC), average 95th-percentile Hausdorff distance (HD95), and a modality-independent structural similarity measure based on the MIND self-similarity context (MIND-SSC). Results show high variability across datasets, with learning-based methods demonstrating superior performance on large synthetic benchmarks, while only limited improvements are observed in real pelvic registration. A key finding of this study is the consistent disagreement between geometric metrics (DSC, HD95) and image-based similarity metrics (MIND-SSC), highlighting that improved overlap does not necessarily imply better global multimodal correspondence. Furthermore, anatomy-guided approaches achieve the highest overlap scores but exhibit degraded performance outside of segmented regions, revealing a trade-off between label-driven alignment and global structural coherence. Overall, our results indicate that no current method achieves robust performance across anatomies and modalities. We demonstrate that intra-patient 3D multimodal registration requires multi-criteria evaluation, including deformation-based metrics, and remains an open problem.

Chinese Translation

多模态图像配准是许多临床工作流程中的关键组成部分，但由于对应的解剖结构在不同模态下的图像强度差异显著，这一过程仍然具有挑战性。在本研究中，我们对三种涵盖不同解剖区域和难度水平的数据集进行了全面的患者内部三维多模态可变形配准方法的基准测试，包括合成变形恢复和真实临床场景。我们评估了经典的基于优化的方法和现代的基于学习的方法，包括最近的深度学习和基础模型，使用互补指标：平均Dice相似系数（DSC）、平均95百分位Hausdorff距离（HD95）以及基于MIND自相似上下文的模态独立结构相似性度量（MIND-SSC）。结果显示不同数据集之间存在较高的变异性，基于学习的方法在大型合成基准测试中表现优越，而在真实的盆腔配准中仅观察到有限的改善。本研究的一个关键发现是几何指标（DSC、HD95）与基于图像的相似性指标（MIND-SSC）之间存在一致的不一致性，强调了改进的重叠不一定意味着更好的全局多模态对应。此外，解剖引导的方法实现了最高的重叠得分，但在分割区域之外表现下降，揭示了标签驱动对齐与全局结构一致性之间的权衡。总体而言，我们的结果表明，目前没有任何方法在不同解剖结构和模态下实现稳健的性能。我们证明了患者内部三维多模态配准需要多标准评估，包括基于变形的指标，并仍然是一个开放的问题。

View on arXiv Download PDF AI Translation

cs.CV / 187 / 2609.15671

Don't Send What You Don't Need: Question-Guided Token Pruning as a Privacy Defense for Vision-Language Models

不发送不必要的信息：基于问题引导的令牌修剪作为视觉语言模型的隐私防御

Syfullah, Md Khalid, Khalil, Alvi Ataur

Abstract

Visual Question Answering (VQA) with Vision-Language Models (VLMs) is increasingly used in privacy-sensitive and bandwidth-constrained settings. Federated Learning (FL), Split Learning (SL), and U-Shaped Split Learning (USL) keep raw data local, but transmitting all visual tokens across a model partition remains costly and can expose private information. We propose QPriv-VL, a question-guided, privacy-aware token-pruning framework for FL, SL, and USL that prunes visual tokens before transmission based on task utility and privacy sensitivity. Its core component is a lightweight Dynamic Threshold Predictor (DTP) that jointly estimates a sample-specific pruning ratio and a token-level retention mask in one forward pass. DTP combines question relevance, computed from cross-modal similarity between visual patches and the pooled question embedding, with a sensitivity signal derived from frozen DINOv2 features. This allows the model to suppress potentially sensitive regions while preserving patches useful for answering the question, without requiring sensitivity labels. We evaluate QPriv-VL on GQA, OK-VQA, VQAv2, SLAKE, VQA-RAD, and PathVQA against four privacy attack families: FSHA, FORA, iDLG, and attribute-inference membership inference attacks. DTP matches or outperforms fixed-ratio pruning while using substantially fewer transmitted tokens. On VQA-RAD, it reduces membership-inference attack success from 0.99 to 0.76-0.79, lowers FSHA and FORA reconstruction PSNR relative to fixed-ratio pruning, and preserves competitive VQA accuracy using about 40% of the original visual-token budget. A sensitivity exclusion ratio of 1.20 +/- 0.18 indicates preferential removal of privacy-sensitive patches, while explainability analysis shows that retention adapts to question semantics rather than generic visual saliency.

Chinese Translation

视觉问答（VQA）与视觉语言模型（VLMs）在隐私敏感和带宽受限的环境中越来越多地被使用。联邦学习（FL）、分割学习（SL）和U型分割学习（USL）将原始数据保留在本地，但在模型分区中传输所有视觉令牌仍然成本高昂，并可能暴露私人信息。我们提出了QPriv-VL，这是一种基于问题引导的、关注隐私的令牌修剪框架，适用于FL、SL和USL，它根据任务效用和隐私敏感性在传输前修剪视觉令牌。其核心组件是一个轻量级动态阈值预测器（DTP），能够在一次前向传递中联合估计样本特定的修剪比例和令牌级别的保留掩码。DTP结合了通过视觉补丁与汇总问题嵌入之间的跨模态相似性计算得出的问句相关性，以及来自冻结的DINOv2特征的敏感性信号。这使得模型能够抑制潜在的敏感区域，同时保留对回答问题有用的补丁，而无需敏感性标签。我们在GQA、OK-VQA、VQAv2、SLAKE、VQA-RAD和PathVQA上评估了QPriv-VL，针对四种隐私攻击类别：FSHA、FORA、iDLG和属性推断成员推断攻击。DTP在使用显著更少的传输令牌的情况下，与固定比例修剪相匹配或超越。在VQA-RAD上，它将成员推断攻击的成功率从0.99降低到0.76-0.79，相比于固定比例修剪降低了FSHA和FORA重建的PSNR，并在使用约40%原始视觉令牌预算的情况下保持了竞争性的VQA准确性。敏感性排除比率为1.20 +/- 0.18，表明优先去除隐私敏感的补丁，而可解释性分析显示保留适应于问题语义，而非通用视觉显著性。

View on arXiv Download PDF AI Translation

cs.CV / 188 / 2609.15676

MambaMPD: A Mamba-Driven Segmentation Framework for Marine Pollution Detection from Remote Sensing Imagery

MambaMPD：一种基于Mamba的海洋污染检测遥感图像分割框架

Chen, Shuaiyu, Han, Wei, Ren, Peng, Luo, Chunbo, Fu, Zeyu

Abstract

Accurate marine pollution detection (MPD) is essential for protecting coastal ecosystems and marine biodiversity. Vision Mamba models have shown promise in remote-sensing semantic segmentation by efficiently capturing long-range dependencies and global context, yet their potential for MPD remains underexplored. MPD is particularly challenging because of low signal-to-noise ratios, fragmented pollution patterns, and indistinct boundaries caused by the visual similarity between pollutants and the surrounding sea. To address these issues, we propose MambaMPD, an enhanced Mamba-based framework incorporating two complementary structural priors: Frequency-Aware Augmentation (FAA) and multi-scale Edge-Guided Attention (EGA). FAA integrates wavelet transforms into the encoder to decompose features into multi-scale frequency subbands, enabling the model to capture low-frequency contextual semantics and high-frequency structural details needed to identify small, low-contrast, and irregular pollution patterns. EGA adaptively fuses hierarchical, Laplacian-derived boundary cues with deep semantic representations, refining encoder features before decoding to sharpen boundaries and reduce ambiguity in visually confusing, spatially fragmented scenes. Together, these modules improve sensitivity to subtle pollution signals while preserving fine boundary structures. A U-Net-style decoder with squeeze-and-excitation attention and deep supervision progressively restores and refines semantic and spatial information across scales. Extensive experiments on two benchmark MPD datasets show that MambaMPD achieves higher mIoU than competing methods while requiring substantially less computation than foundation-model-based approaches. On MADOS, it improves F1 by 3.6% over OSDMamba; on M4D, it raises Oil Spill IoU by 6.82% over TransOilSeg.

Chinese Translation

准确的海洋污染检测（MPD）对于保护沿海生态系统和海洋生物多样性至关重要。Vision Mamba模型在遥感语义分割中显示出良好的前景，能够高效捕捉长距离依赖关系和全局上下文，但其在MPD中的潜力仍未得到充分探索。MPD尤其具有挑战性，因为低信噪比、碎片化的污染模式以及污染物与周围海域之间的视觉相似性导致的模糊边界。为了解决这些问题，我们提出了MambaMPD，一个增强的基于Mamba的框架，结合了两个互补的结构先验：频率感知增强（Frequency-Aware Augmentation, FAA）和多尺度边缘引导注意力（multi-scale Edge-Guided Attention, EGA）。FAA将小波变换集成到编码器中，将特征分解为多尺度频率子带，使模型能够捕捉识别小型、低对比度和不规则污染模式所需的低频上下文语义和高频结构细节。EGA自适应地将分层的、拉普拉斯导出的边界线索与深度语义表示融合，在解码之前细化编码器特征，以锐化边界并减少在视觉混淆和空间碎片化场景中的模糊性。这些模块共同提高了对微妙污染信号的敏感性，同时保留了精细的边界结构。一个U-Net风格的解码器结合了压缩与激励注意力和深度监督，逐步恢复和细化跨尺度的语义和空间信息。在两个基准MPD数据集上的大量实验表明，MambaMPD在mIoU方面优于竞争方法，同时所需计算量显著低于基于基础模型的方法。在MADOS上，它比OSDMamba提高了3.6%的F1；在M4D上，它比TransOilSeg提高了6.82%的油污IoU。

View on arXiv Download PDF AI Translation

cs.CV / 189 / 2609.15683

V-ICAL Bench: Evaluating Video In-Context Learning for Multimodal Agents in Interactive Environments

V-ICAL Bench：评估交互环境中多模态智能体的视频上下文学习

Fan, Ziqian, Xu, Shibo, Li, Junjie, Zhao, Xiangyu, Ding, Shengyuan, Yang, Yifan, Yang, Zhenjie, Duan, Haodong, Zhou, Yue, Zhong, Zhihang, Yang, Xue

Abstract

While In-Context Learning (ICL) enables models to adapt from exemplars without parameter updates, multimodal ICL remains largely underexplored, particularly regarding video demonstrations in interactive environments. For multimodal agents, learning from videos presents unique challenges: they must translate in-context demonstrations into executable policies, ground these policies in novel visual states, and iteratively refine actions based on environmental feedback. We introduce V-ICAL, a novel benchmark designed to evaluate video-based ICL for multimodal agents. Comprising 342 interactive tasks across 37 environments, V-ICAL utilizes human-curated demonstration videos as task-specific behavioral exemplars, evaluating agents through sustained interaction from a target initialization. The benchmark seamlessly connects in-context knowledge induction with core agentic capabilities, including state grounding, temporal memory, planning, and adaptation in dynamic environments. Extensive evaluations across 19 state-of-the-art multimodal agents reveal significant limitations: the best-performing model, Seed-2.1-Pro, achieves a score of only 54.4/100, while other leading models (e.g., Gemini-3.1-Pro, GPT-5.6) fail to surpass 50, far below the human baseline of 83.6. Controlled comparisons further demonstrate that current agents struggle to reliably translate video exemplars into effective policies, failing to yield consistent performance gains. Ultimately, V-ICAL exposes a critical gap in the ICL capabilities of multimodal agents, underscoring an urgent need for future research.

Chinese Translation

虽然上下文学习（In-Context Learning, ICL）使模型能够在不更新参数的情况下从示例中适应，但多模态ICL仍然在很大程度上未被探索，特别是在交互环境中的视频演示方面。对于多模态智能体来说，从视频中学习面临独特的挑战：它们必须将上下文演示转化为可执行的策略，将这些策略与新的视觉状态相结合，并根据环境反馈迭代优化行动。我们提出了V-ICAL，这是一个旨在评估多模态智能体的视频基础ICL的新基准。V-ICAL包含342个交互任务，分布在37个环境中，利用人工策划的演示视频作为任务特定的行为示例，通过从目标初始化开始的持续交互来评估智能体。该基准无缝连接了上下文知识引导与核心智能体能力，包括状态基础、时间记忆、规划和在动态环境中的适应能力。对19个最先进的多模态智能体的广泛评估揭示了显著的局限性：表现最佳的模型Seed-2.1-Pro仅获得54.4/100的分数，而其他领先模型（如Gemini-3.1-Pro、GPT-5.6）未能超过50，远低于人类基线的83.6。控制比较进一步表明，当前智能体在将视频示例可靠地转化为有效策略方面存在困难，未能产生一致的性能提升。最终，V-ICAL揭示了多模态智能体在ICL能力方面的关键缺口，强调了未来研究的迫切需求。

View on arXiv Download PDF AI Translation

cs.CV / 190 / 2609.15795

SURE-Map: Self-Correcting Streaming Geometric Foundation Model

SURE-Map：自我校正的流式几何基础模型

Liu, Mingkai, Zhao, Hao, Zuo, Xingxing

Abstract

Streaming geometric foundation models are emerging as a compelling alternative to SLAM systems. Yet this streaming nature introduces a fundamental issue: each prediction is made from limited context, which is vulnerable to dynamic objects and weak textures. Small local errors accumulate into severe geometric distortion and long-horizon scale drift. We argue that reliable streaming reconstruction requires geometric foundation models to be not only predictive, but also self-correcting. We introduce SURE-Map, a self-correcting framework built upon two complementary principles. First, we explicitly model cross-view geometric uncertainty. Unlike conventional depth or point confidence, which primarily reflects the reliability of individual-view prediction, our uncertainty directly measures whether the jointly predicted pose and depth induce geometrically consistent cross-view pixel correspondences. Second, because local correction alone cannot eliminate slowly accumulating scale errors, we introduce multi-timescale self-correction: fast consecutive-frame inference preserves streaming efficiency, while sparse keyframe-window inference provides longer-range geometric evidence to periodically recalibrate the scale of recent trajectories. SURE-Map establishes new state-of-the-art performance for online feed-forward reconstruction across long-horizon benchmarks, reducing ATE-RMSE from 24.00 to 17.24 m on KITTI, 5.11 to 4.74 m on Oxford Spires, and 31.37 to 28.58 m on VBR, with further improvements to 15.17, 4.63, and 22.12 m when incorporating loop-closure refinement. Project page: https://mingkai-liu.github.io/projects/sure-map/.

Chinese Translation

流式几何基础模型作为一种引人注目的替代方案正在兴起，取代传统的SLAM系统。然而，这种流式特性引入了一个根本性的问题：每次预测都是基于有限的上下文，这使其容易受到动态物体和弱纹理的影响。小的局部误差会累积成严重的几何失真和长时间尺度漂移。我们认为，可靠的流式重建不仅需要几何基础模型具备预测能力，还需要具备自我校正能力。我们提出了SURE-Map，这是一个基于两个互补原则构建的自我校正框架。首先，我们明确建模了视角间的几何不确定性。与传统的深度或点置信度主要反映单视图预测的可靠性不同，我们的几何不确定性直接衡量联合预测的姿态和深度是否引起几何一致的视角间像素对应关系。其次，由于单靠局部校正无法消除缓慢累积的尺度误差，我们引入了多时间尺度自我校正：快速的连续帧推理保持了流式效率，而稀疏的关键帧窗口推理则提供了更长范围的几何证据，以定期重新校准近期轨迹的尺度。SURE-Map在长时间基准测试中建立了在线前馈重建的新一流性能，将KITTI上的ATE-RMSE从24.00降低到17.24米，Oxford Spires上的从5.11降低到4.74米，以及VBR上的从31.37降低到28.58米，当结合回环闭合优化时，进一步改善到15.17、4.63和22.12米。项目页面：https://mingkai-liu.github.io/projects/sure-map/

View on arXiv Download PDF AI Translation

cs.CV / 191 / 2609.15810

VC-Attention: Value Smoothing and Softmax Casting for Low-bit Attention

VC-Attention：低比特注意力的值平滑与Softmax映射

Li, Xingyang, Zou, Dongyun, Zhang, Shining, Chen, Jiacheng, Xi, Haocheng, Zhang, Lvmin, Zhu, Jun-Yan, Han, Song, Zhang, Zhekai, Lin, Yujun, Li, Muyang

Abstract

Diffusion Transformers deliver state-of-the-art video generation, but their long spatiotemporal sequences make attention the dominant deployment cost, and a deployable low-bit kernel must be accurate and fast. Accuracy is limited by outliers: a block's quantization scale is set by its largest entries, leaving typical entries confined to a narrow range of representable values. Prior work smooths queries and keys, but value outliers follow no fixed channel or spatiotemporal structure and remain the dominant source of output error. Speed is limited by softmax: low-bit Tensor Cores accelerate only the two matrix multiplications, so the high-precision exponential between them becomes the longest pipeline stage on datacenter GPUs. We propose VC-Attention, a training-free low-bit attention framework that addresses both by pairing Value smoothing with a fused probability Cast. V-Smooth reorders value tokens by lightweight online clustering, so the tokens in a hardware block quantize well together. It quantizes only the residual after subtracting the block mean, and restores that mean from the row sum the online softmax already maintains. ExpCast-FP8 maps log-domain scores directly to E4M3 probability codes with one fused multiply-add, eliminating the FP32 exponential and the format conversion. We implement VC-Attention for B200, B300, H200, RTX PRO 6000, and RTX 5090. Across Wan2.2, LongCat-Video, HunyuanVideo-1.5, and MiniMax-H3, VC-Attention improves fidelity over low-bit baselines, speeds up the attention kernel over BF16 FlashAttention-4 by 1.46-1.59x on datacenter Blackwell and Hopper and by 2.3-3.6x on workstation cards, and generates a clip 1.13-1.19x and 1.36-1.70x faster end to end.

Chinese Translation

扩散变换器（Diffusion Transformers）在视频生成方面表现出色，但其长时间空时序列使得注意力成为主要的部署成本，因此可部署的低比特内核必须具备高准确性和快速性。准确性受到异常值的限制：一个块的量化尺度由其最大条目决定，导致典型条目被限制在一个狭窄的可表示值范围内。之前的研究对查询和键进行了平滑处理，但值的异常值并没有遵循固定的通道或空时结构，仍然是输出误差的主要来源。速度受限于Softmax：低比特张量核心（Tensor Cores）仅加速两个矩阵乘法，因此它们之间的高精度指数运算成为数据中心GPU上最长的流水线阶段。我们提出了VC-Attention，这是一种无训练的低比特注意力框架，通过将值平滑与融合概率映射相结合来解决这两个问题。V-Smooth通过轻量级在线聚类重新排序值令牌，使得硬件块中的令牌能够很好地一起量化。它仅对减去块均值后的残差进行量化，并从在线Softmax已经维护的行和中恢复该均值。ExpCast-FP8将对数域分数直接映射到E4M3概率编码，使用一次融合的乘加运算，消除了FP32指数运算和格式转换。我们在B200、B300、H200、RTX PRO 6000和RTX 5090上实现了VC-Attention。在Wan2.2、LongCat-Video、HunyuanVideo-1.5和MiniMax-H3上，VC-Attention在低比特基线之上提高了保真度，在数据中心的Blackwell和Hopper上使注意力内核的速度比BF16 FlashAttention-4快1.46-1.59倍，在工作站卡上快2.3-3.6倍，并且生成的片段在端到端上快1.13-1.19倍和1.36-1.70倍。

View on arXiv Download PDF AI Translation

cs.CV / 192 / 2609.15833

Integrating Multi-view Multi-light Surface Reconstruction into Cultural Heritage Workflows

将多视角多光源表面重建整合到文化遗产工作流程中

Brument, Baptiste, Bruneau, Robin, Coupry, Benjamin, Demoulin, Vincent, Mélou, Jean, Laurent, Antoine, Castan, Fabien, Durou, Jean-Denis, Calvet, Lilian

Abstract

Cultural heritage documentation increasingly relies on image-based 3D surface reconstruction, with photogrammetry software making such workflows accessible to archaeologists, conservators, and heritage technicians. These tools have been successful for conventional multi-view acquisition, but they do not routinely exploit richer multi-view, multi-light data, despite its potential for improving fine-scale surface reconstruction. This limitation is particularly relevant in heritage contexts, where controlled-light acquisition devices such as RTI domes are already used to capture illumination-varying image sets. The challenge is therefore to connect these existing acquisition practices with recent computer vision methods in a form that can be used within operational heritage workflows. In this work, we address this need by integrating state-of-the-art components from computer vision for multi-view, multi-light surface reconstruction into Meshroom, an open-source photogrammetry framework. Rather than proposing a new reconstruction algorithm, our contribution is to assemble and expose existing advanced methods, namely a complete photometric stereo ecosystem (calibrated, self-calibrated and universal), automatic object masking, and multi-view normal-and-reflectance integration, within a usable heritage-oriented workflow. The proposed system thus provides an intermediate software layer between computer vision research code and practical cultural heritage applications, making recent techniques easier to use and evaluate.

Chinese Translation

文化遗产文档越来越依赖于基于图像的三维表面重建，摄影测量软件使考古学家、保护者和遗产技术人员能够使用这些工作流程。这些工具在传统的多视角获取中取得了成功，但它们并未常规利用更丰富的多视角、多光源数据，尽管这对于改善细尺度表面重建具有潜力。这一局限性在文化遗产背景下尤为相关，因为已经使用受控光获取设备（如RTI穹顶）来捕捉光照变化的图像集。因此，挑战在于将这些现有的获取实践与最新的计算机视觉方法连接起来，以便在操作性文化遗产工作流程中使用。在本研究中，我们通过将计算机视觉领域的最先进组件整合到Meshroom（一个开源摄影测量框架）中，以满足这一需求，进行多视角、多光源表面重建。我们的贡献不是提出一种新的重建算法，而是将现有的先进方法进行组合和展示，即一个完整的光度立体生态系统（校准、自校准和通用）、自动对象遮罩，以及多视角法线与反射整合，融入到一个可用的以文化遗产为导向的工作流程中。因此，所提出的系统提供了计算机视觉研究代码与实际文化遗产应用之间的中间软件层，使得最新技术更易于使用和评估。

View on arXiv Download PDF AI Translation

cs.CV / 193 / 2609.15834

TRACE: Two-Stage Detector-Response Estimation With Angular Cosine Expansion for Ring Artifact Correction in Photon-Counting CT

TRACE：用于光子计数CT中环形伪影校正的双阶段探测器响应估计与角余弦展开

Duan, Jigang, Wang, Heran, Shi, Ligen, Sun, Zheng, Yang, Ping, Zhao, Xing

Abstract

Detector response nonuniformity introduces systematic projection errors and ring artifacts in photon-counting detector computed tomography (PCD-CT). In measured PCD-CT data, residual stripe amplitudes vary slowly with projection angle, which fixed-bias models cannot adequately capture. We propose TRACE, a two-stage unsupervised sinogram decomposition method for estimating and correcting these response-related errors. TRACE represents stripes as a fixed bias plus low-order discrete cosine transform (DCT) components, using a small number of coefficients to describe angular variations at each detector element. A learnable analysis--synthesis architecture represents the ideal projections, while two-stage optimization separates them from fixed and then dynamic stripes. An angular-gradient soft orthogonality constraint suppresses correlated variations within the shared DCT gradient subspace, reducing the leakage of object structures into the artifact estimate. All parameters are optimized directly on the measured sinogram without paired training data. Experiments on measured QRM mouse phantom and porcine trotter data show that TRACE suppresses ring artifacts and improves image uniformity while preserving edge sharpness, soft-tissue texture, and trabecular detail.

Chinese Translation

探测器响应的不均匀性在光子计数探测器计算机断层扫描（PCD-CT）中引入了系统性的投影误差和环形伪影。在测量的PCD-CT数据中，残余条纹幅度随投影角度缓慢变化，而固定偏差模型无法充分捕捉这些变化。我们提出了TRACE，一种用于估计和校正这些响应相关误差的双阶段无监督正弦图分解方法。TRACE将条纹表示为固定偏差加上低阶离散余弦变换（DCT）分量，使用少量系数描述每个探测器元件的角度变化。可学习的分析-合成架构表示理想投影，而双阶段优化将其从固定条纹和动态条纹中分离。角梯度软正交性约束抑制了共享DCT梯度子空间内的相关变化，减少了物体结构泄漏到伪影估计中的情况。所有参数均在测量的正弦图上直接优化，无需配对训练数据。在测量的QRM小鼠幻影和猪腿数据上的实验表明，TRACE能够抑制环形伪影，提高图像均匀性，同时保持边缘清晰度、软组织纹理和小梁细节。

View on arXiv Download PDF AI Translation

cs.CV / 194 / 2609.15863

LynnReal-Omni: Native multi-modal Video Generation for Agentic Visual Workflows

LynnReal-Omni：用于自主视觉工作流的原生多模态视频生成

Mao, Xiaofeng, Lin, Peijia, Rui, Shaohao, Zhang, Yibo, Wan, Haibin, Ma, Weijie

Abstract

Video diffusion models are stochastic and hard to control: precise content often requires repeated sampling without guaranteed success, and long-horizon scenes drift in appearance, interactions, and temporal coherence. Agentic visual creation provides explicit references, editable 3D scenes, or executable game states for stable control, but does not by itself guarantee high object or character fidelity. Combining the two can enable stable, high-quality generation. To realize this combination, we present LynnReal-Omni, a native multimodal video generation framework built on a 32B shared multimodal diffusion transformer that unifies text-to-video, image-conditioned generation, reference-guided generation, structural control, editing, degraded video restoration, and long-video generation. It accepts heterogeneous visual inputs, including appearance references, editable 3D renders, and game recordings, allowing agents to compose visual conditions within a unified model. We also train a dedicated 27B Flash shared multimodal diffusion transformer for real-time rendering. We build a systematic data pipeline for video cleaning, subject association, multimodal annotation, and aligned control construction, yielding a curated corpus of multi-shot audiovisual segments, and introduce MSAVP, a 100-prompt, 20-metric evaluation design that separates instruction following, generating plausibility, visual quality, temporal behavior, and audio coordination. LynnReal-Omni-Flash further reduces inference cost through model and decoding acceleration, including a lightweight VAE decoder; on one H100, warm generation and decoding of a 22-frame 540p video take 843 ms with LynnReal-Omni and 377 ms with Flash. These results provide a foundation for real-time streaming video generation, making LynnReal-Omni a unified, controllable, and efficient basis for agentic visual creation.

Chinese Translation

视频扩散模型具有随机性且难以控制：精确内容通常需要重复采样而无法保证成功，且长时间场景在外观、交互和时间一致性上会出现漂移。自主视觉创作提供了明确的参考、可编辑的3D场景或可执行的游戏状态以实现稳定控制，但本身并不能保证高物体或角色的保真度。将两者结合可以实现稳定的高质量生成。为实现这种结合，我们提出了LynnReal-Omni，这是一个基于32B共享多模态扩散变换器的原生多模态视频生成框架，统一了文本到视频、图像条件生成、参考引导生成、结构控制、编辑、降级视频恢复和长视频生成。它接受异构视觉输入，包括外观参考、可编辑的3D渲染和游戏录制，允许代理在统一模型中组合视觉条件。我们还训练了一个专用的27B Flash共享多模态扩散变换器以实现实时渲染。我们建立了一个系统化的数据管道，用于视频清理、主题关联、多模态注释和对齐控制构建，生成了一个策划的多镜头视听片段语料库，并引入了MSAVP，一个包含100个提示和20个指标的评估设计，分别评估指令遵循、生成合理性、视觉质量、时间行为和音频协调。LynnReal-Omni-Flash通过模型和解码加速进一步降低了推理成本，包括一个轻量级的VAE解码器；在一台H100上，使用LynnReal-Omni生成和解码一个22帧540p视频的时间为843毫秒，而使用Flash则为377毫秒。这些结果为实时流媒体视频生成奠定了基础，使LynnReal-Omni成为自主视觉创作的统一、可控和高效的基础。

View on arXiv Download PDF AI Translation

cs.CV / 195 / 2609.15888

Anatomical Grounding and Leakage-Aware Multimodal Contrastive Learning for Alzheimer's Disease Classification from Structural MRI

基于解剖基础和泄漏意识的多模态对比学习用于从结构性MRI中分类阿尔茨海默病

Nicolae, Paul-Gabriel, Mocanu, Irina Georgiana

Abstract

Deep networks trained on structural MRI for Alzheimer's disease (AD) staging often reach reasonable accuracy while attending to anatomically irrelevant regions, and multimodal models that add clinical tables frequently rely on variables that were used to assign the diagnostic label in the first place. We study both issues with a deliberately lightweight slice-based encoder (ResNet18 with a one-layer Transformer over slices) on 1,075 baseline T1-weighted scans from ADNI-1. First, we use FastSurfer segmentations as an anatomical reference: YOLOv8 models trained on segmentation-derived labels localize Alzheimer-relevant structures with mAP_50 above 0.96, and a Grad-CAM comparison shows that the image-only classifier frequently attends to the skull, orbits and background. Second, we adapt a CLIP-style image - tabular contrastive framework and organize ADNIMERGE variables along a label-leakage spectrum. Fusion with cognitive scores yields 87.3% three-way accuracy, which we treat as a leakage-driven upper bound rather than an imaging result; fusion with regional volumes yields 73.0%. We observe that the choice of contrastive target changes what the image encoder learns: on MCI vs. CN, the image-only head reaches 52.4% when the encoder is aligned to cognitive scores and 73.8\% when aligned to volumes, although no tabular input is used at inference. Third, restricting the input to a per-subject crop of the medial temporal lobe raises image-only three-way accuracy from 58.7% to 65.1%. All results come from single runs on a small balanced test set, and we report confidence intervals and the protocol differences that prevent direct comparison with published numbers.

Chinese Translation

针对阿尔茨海默病（AD）分期的深度网络在结构性MRI上训练时，通常能够达到合理的准确性，但往往关注于解剖上无关的区域，而添加临床表格的多模态模型则常常依赖于用于最初分配诊断标签的变量。我们使用一种故意轻量化的基于切片的编码器（在1,075个来自ADNI-1的基线T1加权扫描上，使用ResNet18和一层Transformer）来研究这两个问题。首先，我们使用FastSurfer分割作为解剖参考：YOLOv8模型在分割衍生标签上训练，能够以超过0.96的mAP_50定位与阿尔茨海默病相关的结构，而Grad-CAM比较显示，仅图像分类器经常关注于颅骨、眼眶和背景。其次，我们调整了一种CLIP风格的图像-表格对比框架，并沿着标签泄漏谱组织ADNIMERGE变量。与认知评分的融合实现了87.3%的三分类准确率，我们将其视为一个由泄漏驱动的上限，而非成像结果；与区域体积的融合则实现了73.0%。我们观察到，对比目标的选择改变了图像编码器的学习内容：在轻度认知障碍（MCI）与正常对照（CN）之间，当编码器与认知评分对齐时，仅图像头的准确率为52.4%，而与体积对齐时为73.8%，尽管在推理时未使用任何表格输入。第三，将输入限制为每个受试者的内侧颞叶裁剪，将仅图像的三分类准确率从58.7%提高到65.1%。所有结果均来自于在小型平衡测试集上的单次运行，我们报告了置信区间和阻止与已发布数字直接比较的协议差异。

View on arXiv Download PDF AI Translation

人工智能 (Artificial Intelligence)

147

cs.AI / 1 / 2609.13356

ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search

ZGCM-1：一个完全开放且极高效的数学与自主搜索基础模型

He, Jiyan, Liang, Guang, Liu, Hao, Guan, Haoxiang, Sun, Jinbo, Guo, Junyi, Feng, Wenjun, Xie, Yantai, Shen, Yifei, Shao, Bin, Wei, Chuyang, Chen, Kai, Zhou, Kexin, Zhu, Minghang, Zheng, Shuxin, Liu, Tie-Yan, Zhao, Taine, Zhu, Wenhui, Xu, Xueyin, Zhang, Xiaoqing, Li, Yatao, Ren, Yuxuan

Abstract

In this work, we present ZGCM-1, a fully open 7B dense foundation model trained from scratch with extreme data, system, and algorithmic efficiency. ZGCM-1 is founded on a core premise: compact models cannot passively memorize the open web, but can overcome parametric capacity limits by coupling deliberate internal thinking with active external tool use. To support this paradigm across a 256K context, we develop an end-to-end, high-efficiency open training recipe: Architecture & System Co-design: interleaved gated sliding-window and full attention, and a stable FP8 Muon optimizer; Progressive Curriculum & MDP Mid-Training: context scaling across 16K, 64K, and 256K, and the reformulation of interaction traces into Markov Decision Processes. Furthermore, we establish an AI-native R&D workflow where agent swarms autonomously manage cluster operations, data curation, and rapid diagnostic evaluation. Extensive evaluations show that ZGCM-1-7B is competitive across 7B model family on general benchmarks. On several challenging mathematical reasoning and agentic search suites, it remains competitive with frontier models orders of magnitude larger, such as Qwen3-235B-A22B and GLM-5.1. We also show that our pre-training design offers a ~4.2x efficiency improvement in 16K pre-training time-to-loss. Across the full development lifecycle, we distill eight actionable empirical findings-spanning architectural scaling, SFT quality pruning, long-context generalization, and agentic co-training dynamics. To facilitate community research, we open-source model weights from the pre-training, mid-training, and post-training stages, intermediate checkpoints, training code, per-stage data and data recipes, and W&B logs.

Chinese Translation

在本研究中，我们提出了ZGCM-1，这是一个完全开放的7B密集基础模型，从零开始训练，具有极高的数据、系统和算法效率。ZGCM-1建立在一个核心前提上：紧凑模型无法被动地记忆开放网络，但可以通过将深思熟虑的内部思维与主动的外部工具使用结合，克服参数容量限制。为了在256K上下文中支持这一范式，我们开发了一种端到端的高效开放训练方案：架构与系统协同设计：交错的门控滑动窗口和全注意力机制，以及稳定的FP8 Muon优化器；渐进式课程与MDP中期训练：在16K、64K和256K之间进行上下文扩展，并将交互轨迹重新表述为马尔可夫决策过程。此外，我们建立了一个AI原生的研发工作流程，其中代理群体自主管理集群操作、数据策划和快速诊断评估。广泛的评估表明，ZGCM-1-7B在一般基准测试中与7B模型系列具有竞争力。在若干具有挑战性的数学推理和自主搜索套件中，它与数量级更大的前沿模型（如Qwen3-235B-A22B和GLM-5.1）保持竞争力。我们还展示了我们的预训练设计在16K预训练时间到损失方面提供了约4.2倍的效率提升。在整个开发生命周期中，我们提炼出八个可操作的经验发现，涵盖了架构扩展、SFT质量修剪、长上下文泛化和自主共同训练动态。为了促进社区研究，我们开源了预训练、中期训练和后期训练阶段的模型权重、中间检查点、训练代码、各阶段数据及数据配方，以及W&B日志。

View on arXiv Download PDF AI Translation

cs.AI / 2 / 2609.13396

Converge Then Diversify: Decoupling Convergence and Diversity in Multi-Objective Bayesian Optimisation

先收敛再多样化：在多目标贝叶斯优化中解耦收敛性与多样性

Jiang, Chao, Huang, Yueling, Li, Miqing

Abstract

Multi-objective Bayesian optimisation (MOBO) is a sample-efficient approach for optimising expensive black-box functions with multiple objectives. In MOBO, the goal is to adequately approximate the Pareto front; that is, to obtain a high-quality solution set with 1) good convergence (closeness to the Pareto front) and 2) good diversity (spread across the Pareto front). Existing MOBO methods typically aim to accomplish these two tasks simultaneously, i.e., driving the search towards the Pareto front while maintaining a diverse set of nondominated solutions, such that the solutions, ideally, can gradually approach the entire front. When sufficient search budgets are available, this approach is effective. However, considering both convergence and diversity throughout the search is not easy and requires careful design. Under very tight budgets, there may not be enough solutions generated to be able to simultaneously approach the entire Pareto front. To address this issue, this paper proposes a \textit{converge-then-diversify} (CTD) approach that decouples convergence and diversity into two stages. In the first stage, CTD focuses on convergence, aiming to quickly drive the search toward a single point on the Pareto front. In the second stage, CTD focuses on diversity, aiming to spread solutions across the front. We present two simple instantiations of CTD by using widely adopted acquisition functions in the area. Experimental results show that, across all 446 pairwise comparisons, CTD statistically outperforms state-of-the-art methods in 72.9\% of the cases, performs equivalently in 21.1\%, and is statistically worse in only 6.1\%, with the advantage being particularly evident in settings with very tight evaluation budgets or in high-dimensional problems.

Chinese Translation

多目标贝叶斯优化（MOBO）是一种高效的样本方法，用于优化具有多个目标的昂贵黑箱函数。在MOBO中，目标是充分近似帕累托前沿；即获得一组高质量的解集，具备1）良好的收敛性（接近帕累托前沿）和2）良好的多样性（在帕累托前沿上分布广泛）。现有的MOBO方法通常旨在同时完成这两个任务，即推动搜索向帕累托前沿靠近，同时保持一组多样化的非支配解，以便这些解理想情况下能够逐渐接近整个前沿。当搜索预算充足时，这种方法是有效的。然而，在整个搜索过程中同时考虑收敛性和多样性并不容易，需要仔细设计。在预算非常紧张的情况下，可能生成的解不足以同时接近整个帕累托前沿。为了解决这个问题，本文提出了一种 extit{先收敛再多样化}（CTD）的方法，将收敛性和多样性解耦为两个阶段。在第一阶段，CTD专注于收敛，旨在快速推动搜索向帕累托前沿上的单个点靠近。在第二阶段，CTD专注于多样性，旨在将解分布在前沿上。我们通过使用该领域广泛采用的获取函数，展示了CTD的两种简单实例。实验结果表明，在446个成对比较中，CTD在72.9%的情况下统计上优于最先进的方法，在21.1%的情况下表现相当，仅在6.1%的情况下统计上较差，尤其在评估预算非常紧张或高维问题的设置中，CTD的优势尤为明显。

View on arXiv Download PDF AI Translation

cs.AI / 3 / 2609.13406

Generalized Agent Iteration: One Formal Framework for Iterative Policy Improvement and Recursive Self-Improvement

广义代理迭代：一个用于迭代策略改进和递归自我改进的正式框架

Tang, Hongyao, Ma, Yi, Li, Pengyi, Yuan, Yifu

Abstract

When we speak of recursive self-improvement (RSI), are we speaking of a phenomenon, a mechanism, or a prospect? Towards autonomous and evolving intelligence, RSI is being claimed at many scales, while no single framework that formally describes these emerging instances exists. Its counterpart in the classical realm, iterative policy improvement, is characterized by generalized policy iteration (GPI), a framework of broad applicability with well-understood theoretical properties, but only where the update principle and the evaluation base lie outside the agent. In this paper, we propose Generalized Agent Iteration (GAI), a formal framework that describes iterative policy improvement and RSI as two cases of a single learning paradigm. GAI defines the agent as a configuration of modifiable components within a system and models the learning process as a cycle of agent evaluation and agent improvement. Two pivotal dials then distinguish the instances: whether the improving mechanism is part of the agent and whether the standard it is measured against is grounded outside it. The former dial delineates the boundary between GPI and RSI, and the latter determines a system's polarity as anchored, goal drift, or fully self-referential. Moreover, we use these coordinates to place existing systems on the same two axes and make the defects of recursive self-improvement statable one condition at a time. We see this paper as a first step toward exploring a formal characterization of RSI that rests on the classical account, makes existing systems comparable, and provides a principled basis for analyzing and designing new ones.

Chinese Translation

当我们谈论递归自我改进（Recursive Self-Improvement, RSI）时，我们是在讨论一种现象、一种机制，还是一种前景？在朝向自主和不断发展的智能的过程中，RSI在许多层面上被提出，但尚不存在一个正式描述这些新兴实例的单一框架。在经典领域，其对应物——迭代策略改进（Iterative Policy Improvement）由广义策略迭代（Generalized Policy Iteration, GPI）所特征化，这是一个具有广泛适用性的框架，具有良好理解的理论属性，但仅在更新原则和评估基础位于代理之外的情况下适用。在本文中，我们提出了广义代理迭代（Generalized Agent Iteration, GAI），这是一个正式框架，将迭代策略改进和RSI描述为单一学习范式的两个案例。GAI将代理定义为系统内可修改组件的配置，并将学习过程建模为代理评估和代理改进的循环。两个关键的调节器区分了这些实例：改进机制是否是代理的一部分，以及其衡量标准是否基于代理之外。前者调节器划定了GPI与RSI之间的边界，而后者则决定了系统的极性，可能是锚定的、目标漂移的或完全自指的。此外，我们使用这些坐标将现有系统放置在同两个轴上，并逐一明确递归自我改进的缺陷。我们将本文视为探索RSI正式表征的第一步，该表征基于经典理论，使现有系统可比，并为分析和设计新系统提供了原则基础。

View on arXiv Download PDF AI Translation

cs.AI / 4 / 2609.13422

Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents

Vibe 专利：评估大型语言模型法官作为专业专利撰写代理

Koike-Akino, Toshiaki, Blaykhman, Vlad, Wang, Ye, Liu, Jing, Vinokur, Gene V.

Abstract

LLM judges are increasingly used to evaluate and improve AI-generated outputs, yet their reliability for complex professional work remains unclear. We study this problem through Vibe Patenting, an end-to-end patent-drafting testbed for AI-agent evaluation. A separately-invoked LLM judge evaluates generated patent drafts and provides structured feedback for iterative revision. Across multiple inventions and drafting-agent configurations, judge-guided revision consistently improves judge-assessed quality, while unguided revision tends to saturate. Notably, iterative judge feedback enables a low-reasoning agent to approach the performance of a substantially more expensive high-reasoning agent. Stronger models and increased reasoning generally improve judge-assessed drafting quality, while domain-specific agentic workflows provide further gains. We validate the judge against independent evaluation by a professional patent attorney and find meaningful but strongly metric-dependent agreement and systematic calibration differences. These results highlight both the utility and limitations of LLM judges as evaluators and optimization signals for complex professional workflows.

Chinese Translation

大型语言模型（LLM）法官越来越多地被用于评估和改进人工智能生成的输出，然而它们在复杂专业工作中的可靠性仍不明确。我们通过 Vibe 专利研究这一问题，这是一个用于人工智能代理评估的端到端专利撰写测试平台。一个单独调用的 LLM 法官评估生成的专利草稿，并提供结构化反馈以便进行迭代修订。在多个发明和撰写代理配置中，法官引导的修订始终提高了法官评估的质量，而未引导的修订则趋于饱和。值得注意的是，迭代法官反馈使得一个低推理能力的代理的表现接近一个显著更昂贵的高推理能力代理。更强的模型和增加的推理能力通常提高法官评估的撰写质量，而特定领域的代理工作流程则提供了进一步的提升。我们将法官的评估结果与专业专利律师的独立评估进行了验证，发现有意义但强烈依赖于度量的协议和系统性的校准差异。这些结果突显了 LLM 法官作为复杂专业工作流程的评估者和优化信号的效用与局限性。

View on arXiv Download PDF AI Translation

cs.AI / 5 / 2609.13436

Toward Self-Adaptive Physical AI: Can LLM Agents Manage Long-Horizon Physical Tasks?

迈向自适应物理人工智能：大型语言模型代理能否管理长期物理任务？

Kaushik, Varun, Tan, Yayun, Yu, Xiaofan

Abstract

Large Language Model (LLM) agents offer a promising path toward autonomously managing long-term physical tasks without human intervention. However, physical tasks require agents to continuously observe the environment, make consequential actions, and remain effective as the environment changes. Existing approaches either require substantial data and retraining, or primarily focus on agents operating in the virtual world. In this work, we explore the feasibility of building a self-adaptive physical AI agent that manages long-term physical tasks in a zero-shot manner and adapts to environmental changes without human intervention. We design a multi-agent framework that integrates planning, tool calling, observation, and verification, and evaluate it on agricultural tasks against reinforcement learning (RL) agents under different weather patterns. Our results show that zero-shot LLM agents can achieve comparable management outcomes to RL agents under the same weather pattern and adapt more effectively than RL when evaluated under a shifted environment, highlighting a promising path toward self-adaptive physical AI agents.

Chinese Translation

大型语言模型（LLM）代理为自主管理长期物理任务提供了一条有前景的路径，无需人类干预。然而，物理任务要求代理持续观察环境、做出重要决策，并在环境变化时保持有效性。现有方法要么需要大量数据和重新训练，要么主要集中在虚拟世界中的代理操作。本文探讨了构建一个自适应物理人工智能代理的可行性，该代理以零样本方式管理长期物理任务，并在没有人类干预的情况下适应环境变化。我们设计了一个多代理框架，整合了规划、工具调用、观察和验证，并在不同天气模式下对农业任务进行了评估，与强化学习（RL）代理进行比较。我们的结果表明，零样本LLM代理在相同天气模式下能够实现与RL代理相当的管理结果，并在环境变化的情况下表现出比RL更有效的适应能力，突显了迈向自适应物理人工智能代理的有希望的路径。

View on arXiv Download PDF AI Translation

cs.AI / 6 / 2609.13437

LabAgent: Customize Any Research Hubs for Scientific Discoveries Using AI Agents

LabAgent：利用人工智能代理定制任何研究中心以促进科学发现

Liu, Lei, Zhang, Yikun, Chen, Jialin, Zhao, Wanjia, Ying, Rex, Jin, Wengong, Xu, Hua, Zou, James, Liu, Tianyu, Zhao, Hongyu

Abstract

Scientific research is a continuous process that emphasizes inheritance. Methods developed by predecessors are often expanded upon by new researchers to explore more novel and in-depth scientific questions. However, the change of lab staff, such as student graduation, leads to a lack of personnel capable of replicating methods. Methods that have been developed with significant effort and resources cannot be continued. To address these limitations, we propose LabAgent, a reproduce and discovery harness tailored for a lab's continuous work. LabAgent employs two mechanisms to guarantee that all skills can be executed and verified and to record the corrective methods and experiences, allowing for direct correction or avoidance of similar errors. We applied LabAgent to drug property prediction, biomedical problem analysis, protein variant effect prediction, and statistical genetics in life science domains. LabAgent ranks first over commercial generalist agents in every domain, and demonstrates accurate reproduction of a published figure. Overall, these results demonstrate that LabAgent can effectively integrate and reasonably expand laboratory knowledge.

Chinese Translation

科学研究是一个强调继承的持续过程。前辈们开发的方法常常被新研究者扩展，以探索更具新颖性和深度的科学问题。然而，实验室人员的变动，例如学生毕业，导致缺乏能够复制方法的人才。经过大量努力和资源开发的方法无法继续进行。为了解决这些限制，我们提出了LabAgent，这是一种针对实验室持续工作的再现与发现工具。LabAgent采用两种机制，确保所有技能能够执行和验证，并记录纠正方法和经验，从而允许直接纠正或避免类似错误。我们将LabAgent应用于药物性质预测、生物医学问题分析、蛋白质变异效应预测以及生命科学领域的统计遗传学。LabAgent在各个领域的表现均优于商业通用代理，并准确再现了已发表的图表。总体而言，这些结果表明LabAgent能够有效整合并合理扩展实验室知识。

View on arXiv Download PDF AI Translation

cs.AI / 7 / 2609.13457

TimeThink: Eliciting Compositional Reasoning in Timeseries Large Language Models

TimeThink：在时间序列大型语言模型中引发组合推理

Regmi, Sudarshan, Pillai, Arvind, Wu, Yu Yvonne, Chen, Yuliang, Panthi, Bibek, Griffin, Tess Z., Heinz, Michael V., Marsch, Lisa, Jacobson, Nicholas C., Campbell, Andrew

Abstract

Timeseries multimodal large language models (TS-MLLMs) have recently begun leveraging the reasoning capabilities of large language models (LLMs) for question-answering tasks. However, these models often fail to capture dynamic temporal patterns, providing only implicit reasoning that lacks the underlying explanations critical for high-stakes applications like healthcare. While reinforcement learning (RL)-based timeseries language models aim to address this, they often fall short because they are trained on narrow, in-distribution data and struggle with out-of-distribution compositional questions. To address these challenges, we present TimeThink, a synthetic framework for eliciting compositional timeseries reasoning. Core timeseries primitives (e.g., trend, seasonality) are domain-independent and can be deterministically generated. Guided by this premise, TimeThink first designs a synthetic data generator that produces atomic and composite question-answer pairs, providing objective ground truth with reasoning traces. Building on this framework, TimeThink employs a reinforcement learning with verifiable rewards (RLVR) training strategy that encourages explicit reasoning. Unlike template-reliant methods, this approach enables the model to learn the underlying logic of composition rather than simply imitating traces. Extensive experiments show that TimeThink, trained only on synthetic data, significantly outperforms strong baselines on both synthetic and real-world benchmarks.

Chinese Translation

时间序列多模态大型语言模型（TS-MLLMs）最近开始利用大型语言模型（LLMs）的推理能力来进行问答任务。然而，这些模型往往无法捕捉动态时间模式，仅提供缺乏基础解释的隐式推理，这对于医疗等高风险应用至关重要。尽管基于强化学习（RL）的时间序列语言模型旨在解决这一问题，但由于它们在狭窄的、分布内数据上训练，往往在处理分布外的组合问题时表现不佳。为了解决这些挑战，我们提出了TimeThink，一个用于引发组合时间序列推理的合成框架。核心时间序列原语（例如，趋势、季节性）是领域无关的，并且可以确定性地生成。在这一前提下，TimeThink首先设计了一个合成数据生成器，生成原子和复合问答对，提供具有推理轨迹的客观真实数据。在此框架基础上，TimeThink采用了一种可验证奖励的强化学习（RLVR）训练策略，鼓励显式推理。与依赖模板的方法不同，这种方法使模型能够学习组合的基础逻辑，而不仅仅是模仿轨迹。大量实验表明，仅在合成数据上训练的TimeThink在合成和真实世界基准测试中显著优于强基线。

View on arXiv Download PDF AI Translation

cs.AI / 8 / 2609.13463

Root-Cause Attribution Is a Search Problem: Continual Search for Long-Horizon Agent Failures

根本原因归因是一个搜索问题：对长期代理失败的持续搜索

Raj, Harsh, Lee, David, Mahmoud, Anas, Wang, Renxiong, Dumitru, Razvan-Gabriel, Wang, Chenguang, Zhao, Tong, He, Yunzhong, Yi, Darvin, Gupta, Vipul

Abstract

The increasing deployment of AI agents in long-horizon tasks yields massive execution logs. Diagnosing failures within these records is crucial for reliability, as it transforms outcome-level signals into actionable interventions. The sheer scale of the data renders human review impractical, driving the need for automated root-cause attribution (RCA). However, automated RCA methods using LLMs suffer from low diagnostic accuracy, especially as execution traces grow larger. They struggle because relevant information is often sparse, distributed across distant actions, and disconnected from the visible failure, reducing root-cause attribution to a massive search problem. Existing RCA methods typically rely on one-shot LLM judgments to diagnose failures from execution traces. While effective for shorter trajectories, these judges tend to settle on a plausible diagnosis early, leaving critical evidence in longer traces unexamined. We introduce Continual Search, an iterative framework that nudges the judge, over successive turns, to keep searching for unresolved diagnostic evidence. We evaluate Continual Search across four existing RCA benchmarks. Recognizing the lack of massive execution traces in current benchmarks, we introduce MegaRCA-Mix to evaluate RCA at scale. MegaRCA-Mix provides a challenging testbed of 50 human-annotated failure trials spanning long-horizon, execution-heavy tasks. Across multiple benchmark suites and model families, Continual Search consistently improves attribution performance. On MegaRCA-Mix, for example, it improves GPT-5.5's F1 score by more than 40\%, from $0.349$ to $0.498$. More interestingly, within the same model family, lower-tier models can even surpass their higher-tier counterparts, demonstrating that effective search supersedes raw model scale.

Chinese Translation

随着人工智能代理在长期任务中的广泛部署，产生了大量的执行日志。在这些记录中诊断失败对于可靠性至关重要，因为它将结果级信号转化为可操作的干预措施。然而，数据的庞大规模使得人工审查变得不切实际，这推动了自动化根本原因归因（RCA）的需求。然而，使用大型语言模型（LLMs）的自动化RCA方法在诊断准确性上存在不足，尤其是在执行轨迹变得更大时。它们面临的挑战在于相关信息往往稀疏，分散在遥远的动作中，并且与可见的失败无关，从而将根本原因归因简化为一个庞大的搜索问题。现有的RCA方法通常依赖于一次性LLM判断来从执行轨迹中诊断失败。虽然这种方法在较短的轨迹上有效，但这些判断者往往会在早期就确定一个合理的诊断，导致在较长轨迹中关键证据未被检查。我们提出了持续搜索（Continual Search），这是一个迭代框架，通过连续的回合推动判断者持续搜索未解决的诊断证据。我们在四个现有RCA基准上评估了持续搜索。鉴于当前基准中缺乏大量执行轨迹，我们引入了MegaRCA-Mix以大规模评估RCA。MegaRCA-Mix提供了一个具有挑战性的测试平台，包括50个由人类标注的失败试验，涵盖长期、执行密集的任务。在多个基准套件和模型系列中，持续搜索始终提高了归因性能。例如，在MegaRCA-Mix上，它将GPT-5.5的F1分数提高了超过40\%，从$0.349$提升至$0.498$。更有趣的是，在同一模型系列中，较低层次的模型甚至可以超越其高层次的对应模型，证明有效的搜索优于原始模型规模。

View on arXiv Download PDF AI Translation

cs.AI / 9 / 2609.13466

Governing at Machine Speed: An Adaptive Intelligence Architecture for Real-Time AI Policy Enforcement

以机器速度治理：实时人工智能政策执行的自适应智能架构

Bokkasam, Sandeep, Durgalakshmi, B.

Abstract

Enterprise AI adoption has reached 78% of organizations globally, yet the infrastructure to govern that adoption has not kept pace. This paper identifies and characterizes the attestation deficit, a structural condition in which organizations maintain governance policies but cannot produce auditable, tamper-evident evidence of enforcement within regulatory timelines. Drawing on empirical data from the Stanford 2026 AI Index Report (362 documented incidents), the IBM/Ponemon 2026 Cost of a Data Breach study (USD 4.99M average cost, 92% lacking access controls), and the EY/AIUC-1 Consortium survey (38% end-to-end monitoring, 17% agent-to-agent coverage), this paper demonstrates that the governance failure is organizational and architectural rather than technical. To address this deficit, we propose AGIL (Adaptive Governance Intelligence Layer), a conceptual five-layer architecture designed to use machine learning for real-time AI governance enforcement. The proposed layers include: (1) Autonomous Discovery for shadow AI detection via behavioral fingerprinting, (2) Behavioral Risk Classification unifying security, hallucination, privacy, and accountability scoring, (3) a Policy Enforcement Gateway for inline permit/deny/modify decisions at sub-100ms latency, (4) a Continuous Attestation Engine generating tamper-evident audit trails as a byproduct of enforcement, and (5) Adaptive Policy Intelligence for ML-driven policy evolution across jurisdictions. AGIL is presented as a theoretical framework and architectural proposal; empirical validation through controlled deployment remains a direction for future work.

Chinese Translation

企业人工智能的采用率已达到全球78%的组织，但治理这一采用的基础设施却未能与之同步发展。本文识别并描述了认证缺口（attestation deficit），这是一种结构性条件，组织虽然维持治理政策，但无法在监管时间框架内提供可审计的、抗篡改的执行证据。基于斯坦福2026年人工智能指数报告（记录了362起事件）、IBM/Ponemon 2026年数据泄露成本研究（平均成本为499万美元，92%的组织缺乏访问控制）以及EY/AIUC-1联盟调查（38%的组织实现端到端监控，17%的组织实现代理间覆盖）的实证数据，本文表明治理失败是组织和架构层面的，而非技术层面的。为了解决这一缺口，我们提出了AGIL（自适应治理智能层），这是一个概念性的五层架构，旨在利用机器学习实现实时人工智能治理执行。所提议的层包括：（1）通过行为指纹识别进行影子人工智能检测的自主发现（Autonomous Discovery），（2）统一安全、幻觉、隐私和问责评分的行为风险分类（Behavioral Risk Classification），（3）用于在低于100毫秒延迟下进行许可/拒绝/修改决策的政策执行网关（Policy Enforcement Gateway），（4）作为执行副产品生成抗篡改审计轨迹的持续认证引擎（Continuous Attestation Engine），以及（5）用于跨辖区的机器学习驱动政策演变的自适应政策智能（Adaptive Policy Intelligence）。AGIL被提出作为一个理论框架和架构提案；通过受控部署进行实证验证仍然是未来工作的一个方向。

View on arXiv Download PDF AI Translation

cs.AI / 10 / 2609.13470

OrchSLM: Probing the Dynamics of Small Language Model Orchestration

OrchSLM：探究小型语言模型编排的动态

Zhang, Chengxi, Yao, Yu

Abstract

Although large language models (LLMs) have demonstrated remarkable capabilities, their reliance on cloud-scale infrastructure poses fundamental challenges for deployment in agentic pipelines, including latency, privacy, connectivity, and substantial computational cost. Small language models (SLMs) offer a compelling alternative: recent studies suggest that many repetitive and narrowly scoped subtasks in agentic workloads may be better served by specialized SLMs than by monolithic LLMs. However, the limited capacity and context windows of SLMs can constrain long-horizon reasoning and interaction-heavy orchestration strategies such as iterative verification and debate. This motivates a complementary, non-interactive paradigm in which heterogeneous SLMs independently generate candidate solutions and a router orchestrates their cached samples without further model interaction. To further understand the mechanisms of such orchestration, we introduce OrchSLM, a routing framework that unifies existing non-interactive orchestration methods and exposes their underlying design choices as controllable parameters. Using OrchSLM as a systematic probe, we reveal how orchestration behavior emerges from diverse knobs, including the task structure, model-pool composition, and multi-agent consensus.

Chinese Translation

尽管大型语言模型（LLMs）展示了卓越的能力，但它们对云规模基础设施的依赖在代理管道中的部署带来了根本性挑战，包括延迟、隐私、连接性和巨大的计算成本。小型语言模型（SLMs）提供了一个引人注目的替代方案：近期研究表明，许多在代理工作负载中重复且范围狭窄的子任务可能更适合由专门的SLMs来处理，而非单一的LLMs。然而，SLMs的有限容量和上下文窗口可能限制长时间推理和交互密集的编排策略，例如迭代验证和辩论。这促使我们提出一种互补的非交互式范式，其中异构的SLMs独立生成候选解决方案，而路由器则在没有进一步模型交互的情况下编排它们的缓存样本。为了进一步理解这种编排的机制，我们引入了OrchSLM，一个统一现有非交互式编排方法的路由框架，并将其基础设计选择暴露为可控参数。通过将OrchSLM作为系统探针，我们揭示了编排行为如何从多样的调节因素中产生，包括任务结构、模型池组成和多智能体共识。

View on arXiv Download PDF AI Translation

cs.AI / 11 / 2609.13475

Grounded Adjudication of Variations across Extracted TimeLines (GAVEL): Comparing Clinical Timelines Against Their Case Reports

基于提取时间线的变异的实证裁定 (GAVEL)：将临床时间线与其案例报告进行比较

Cummins, Jack, Kumar, Sayantan, Tamirisa, Ketan, Weiss, Jeremy C.

Abstract

Existing pipelines for clinical timeline extraction from case reports are evaluated using an expert reference and are limited by imperfect reference annotations and imprecise event alignment. We developed GAVEL, an LLM judge protocol that compares two timelines with the case report and returns a discrepancy type, verdict, and report passage for each difference. We evaluated the event matcher, reviewed 2,738 findings from GPT5.6sol and DeepSeek V3.2, ranked six LLM extractors and two human annotators, and tested GAVEL guided merging. True match rates were 60% immediately below and 48% immediately above the 0.10 cutoff. Manual review confirmed 89.4% and 88.6% of findings. Across 126 reports, merged timelines were preferred in 77.0% of comparisons (95% CI, 69.8 to 84.1%) and reduced discrepancies attributed to the evaluated timeline from 7.63 to 0.85 per report. GAVEL supports report-based comparison and revision without treating either timeline as ground truth.

Chinese Translation

现有的临床时间线从案例报告中提取的流程通过专家参考进行评估，但受到不完美参考注释和不精确事件对齐的限制。我们开发了GAVEL，一种LLM裁判协议，用于将两个时间线与案例报告进行比较，并为每个差异返回不一致类型、裁决和报告段落。我们评估了事件匹配器，审查了来自GPT5.6sol和DeepSeek V3.2的2,738个发现，排名了六个LLM提取器和两个人工注释者，并测试了GAVEL引导的合并。真实匹配率在0.10截止值下方为60%，在0.10截止值上方为48%。人工审查确认了89.4%和88.6%的发现。在126份报告中，合并时间线在77.0%的比较中被优先选择（95% CI，69.8至84.1%），并将归因于评估时间线的差异从每份报告的7.63减少到0.85。GAVEL支持基于报告的比较和修订，而不将任何时间线视为真实标准。

View on arXiv Download PDF AI Translation

cs.AI / 12 / 2609.13491

Token Efficient Task Execution via Application Behavior Modeling for Web Agents

通过应用行为建模实现网络代理的高效任务执行

Ianta, Alexandru, Stroulia, Eleni

Abstract

The strong performance of AI Agents across an impressive variety of tasks is driving an unprecedented investment in agentic infrastructures, however the cost of processing tokens is fast increasing. Web agents automate the execution of web-application tasks described in natural language, by analyzing the web-application's user interface (UI) and interacting with it. This work introduces OdoBot, a novel web-agent architecture that completes tasks at a fraction of the cost when compared to conventional web agents. This is achieved by leveraging a behavioral model of the underlying application constructed by analyzing successful task-execution demonstrations. Our experiments with 45 tasks on the Canvas Learning Management System (LMS) demonstrate that OdoBot uses 44% and 80% fewer tokens than two state-of-the-art competitor agents (Agent-E and WebVoyager), while also surpassing WebVoyager in terms of task success rate.

Chinese Translation

人工智能代理在多种任务上的强大表现正在推动对代理基础设施的前所未有的投资，然而处理令牌的成本正在迅速增加。网络代理通过分析网络应用的用户界面（UI）并与之互动，自动执行以自然语言描述的网络应用任务。本研究介绍了OdoBot，一种新颖的网络代理架构，与传统网络代理相比，其完成任务的成本仅为其一小部分。这是通过利用分析成功任务执行示范所构建的底层应用的行为模型来实现的。我们在Canvas学习管理系统（LMS）上对45个任务的实验表明，OdoBot使用的令牌数量比两个最先进的竞争代理（Agent-E和WebVoyager）分别少44%和80%，同时在任务成功率方面也超过了WebVoyager。

View on arXiv Download PDF AI Translation

cs.AI / 13 / 2609.13519

Fraglingo: Molecular Design via Attachment-Aware Autoregressive Fragment Generation

Fraglingo：通过关注连接的自回归片段生成进行分子设计

Nguyen, Thao, Kim, Jeonghwan, Wang, Zhenhailong, Ji, Heng

Abstract

Molecular design is most effective when generation mirrors the edits chemists actually make: extending a scaffold, replacing a substituent, or decorating a scaffold at a specified attachment site while optimizing molecular properties. Fragment-based molecular design naturally supports this workflow, yet existing approaches often separate fragment selection from attachment prediction, first choosing a fragment from a fixed vocabulary and then predicting how it should be connected. This decoupling restricts generation to a closed fragment vocabulary and treats attachment as a separate prediction problem. We introduce Fraglingo, an autoregressive fragment-based molecular generator that jointly models fragment identity and attachment in a continuous latent space. Fraglingo predicts an attachment-aware fragment embedding and retrieves the next fragment through latent-space nearest-neighbor search. To encode attachment context, we introduce a wildcard-anchored readout that represents the growing molecule from the perspective of its active attachment site, enabling the predicted embedding to capture both the molecular context and the required attachment. Because generation operates in a continuous embedding space rather than over fixed fragment identifiers, new fragments can be added to the inference-time vocabulary without retraining, provided their embeddings can be computed by the trained fragment encoder. This retrieval-based formulation provides a unified generation primitive for molecule generation, scaffold generation, scaffold decoration, and molecular optimization. On controlled property-conditional benchmarks, Fraglingo achieves stronger joint property control than comparably trained baselines while maintaining competitive validity, uniqueness, and novelty. Furthermore, Fraglingo generalizes to fragment libraries up to 4x larger than those used during training without retraining.

Chinese Translation

分子设计在生成过程与化学家实际进行的编辑相符时最为有效：扩展骨架、替换取代基或在指定的连接位点装饰骨架，同时优化分子性质。基于片段的分子设计自然支持这一工作流程，但现有方法往往将片段选择与连接预测分开，首先从固定词汇中选择片段，然后预测其应如何连接。这种解耦限制了生成在封闭片段词汇中的进行，并将连接视为一个独立的预测问题。我们提出了Fraglingo，一种自回归的基于片段的分子生成器，它在连续潜在空间中联合建模片段身份和连接。Fraglingo预测一个关注连接的片段嵌入，并通过潜在空间的最近邻搜索检索下一个片段。为了编码连接上下文，我们引入了一种以通配符为锚的读取方式，从其活跃连接位点的角度表示正在增长的分子，使得预测的嵌入能够同时捕捉分子上下文和所需的连接。由于生成是在连续嵌入空间中进行，而不是基于固定的片段标识符，因此可以在推理时向词汇中添加新片段，而无需重新训练，只要其嵌入可以由训练好的片段编码器计算。基于检索的这一表述为分子生成、骨架生成、骨架装饰和分子优化提供了统一的生成原语。在控制性质条件的基准测试中，Fraglingo在保持竞争性有效性、独特性和新颖性的同时，实现了比同类训练基线更强的联合性质控制。此外，Fraglingo能够推广到训练期间使用的片段库的规模大于4倍而无需重新训练。

View on arXiv Download PDF AI Translation

cs.AI / 14 / 2609.13535

From Legal Text to AI-specific Risk Sources: A Systematic Analysis of the EU AI Act's High-Risk Requirements

从法律文本到人工智能特定风险源：对欧盟人工智能法高风险要求的系统分析

Schnitzer, Ronald, Auer, Mike, Choudhury, Rumpa, Hapfelmeier, Andreas, Hoeving, Maximilian, Painter, Isabelle, Parreira, Josiane Xavier, Zillner, Sonja

Abstract

The EU AI Act introduces mandatory requirements for high-risk AI systems with the explicit goal of ensuring the development and operation of trustworthy AI. At the same time, AI risk management practices rely on structured risk taxonomies to systematically identify and treat AI-specific risk sources. As both the AI Act and established risk taxonomies aim to address AI-induced risks, a natural question is whether they align in the risk sources they cover. However, no clear mapping exists between the risks implicitly addressed by the Act's high-risk requirements and established taxonomies, leaving practitioners without a structured basis for aligning regulatory obligations with AI risk management practice. This paper presents a systematic classification of the requirements extracted from the EU AI Act Section 2 (Requirements for high-risk AI systems), revealing that only a minority directly address AI-specific risk sources, while the majority impose organizational process and documentation obligations. From the AI risk-related requirements, a consolidated list of distinct AI-specific risk sources is derived. The resulting EU AI Act Risk Source List takes an important step towards bridging the gap between legal obligation and AI risk management practice, providing a structured reference for explicit comparison between existing AI risk taxonomies and the risk sources implicitly addressed by the EU AI Act. Important Note: This is the authors' preprint. The paper was presented at the 4th International Conference on Frontiers of Artificial Intelligence, Ethics, and Multidisciplinary Applications. A link to the conference's official proceedings will be provided upon publication.

Chinese Translation

欧盟人工智能法引入了针对高风险人工智能系统的强制性要求，明确旨在确保可信赖人工智能的开发和运营。同时，人工智能风险管理实践依赖于结构化的风险分类法，以系统地识别和处理人工智能特定的风险源。由于人工智能法和既定的风险分类法都旨在应对人工智能引发的风险，因此一个自然的问题是它们在所涵盖的风险源上是否一致。然而，法律高风险要求所隐含的风险与既定分类法之间并没有明确的对应关系，这使得从业者缺乏一个结构化的基础来将监管义务与人工智能风险管理实践对齐。本文对欧盟人工智能法第2节（高风险人工智能系统的要求）提取的要求进行了系统分类，揭示出只有少数要求直接涉及人工智能特定的风险源，而大多数则施加了组织流程和文档义务。从与人工智能风险相关的要求中，提炼出一份独特的人工智能特定风险源的汇总清单。最终形成的欧盟人工智能法风险源清单为弥合法律义务与人工智能风险管理实践之间的差距迈出了重要一步，为现有人工智能风险分类法与欧盟人工智能法隐含涉及的风险源之间的明确比较提供了结构化的参考。重要提示：这是作者的预印本。本文在第四届国际人工智能前沿、伦理与多学科应用会议上进行了展示。会议正式论文集的链接将在出版后提供。

View on arXiv Download PDF AI Translation

cs.AI / 15 / 2609.13543

Asclepius: An Adaptive Harness for Long-Horizon Clinical Agents

阿斯克勒庇俄斯：一种适应性长时间临床代理的框架

Yuan, Grace Chang, Zhang, Xiaoman, Kim, Sung Eun, Luo, Luyang, Rajpurkar, Pranav

Abstract

LLM agents are predominantly benchmarked on short, single-task trajectories, yet real deployments run for hours under contention, surfacing a different class of failures. We use the Clinical Environment Simulator (CES), in which an agent manages an entire emergency-department shift under continuous time and resource pressure, as a testbed: long-horizon execution failures manifest measurably in a single rollout under structured, multi-dimensional grading. On CES, current agents reach the correct diagnosis in most cases yet fail to deliver complete and timely critical actions, revealing an execution gap. We attribute this gap to three long-horizon failure modes, each operationalized as a per-trace counter: instruction-adherence drift, treatment incompleteness, and a severity-equity gap in timeliness. We then introduce Asclepius, an adaptive agent scaffolding with a self-evolving harness that rewrites the operating manual between shifts from trace-level feedback, an externalized clinical skills library for high-stakes regimen knowledge, and three isolated subagents that partition per-turn decisions across the patient queue. On held-out batches never observed during harness evolution, Asclepius improves critical-action correctness by 22% (p = 0.024) over a strong baseline agent framework while preserving diagnostic accuracy, with consistent gains across five LLM judges from three model families; on the full ten-batch set, improvements reach 25% on critical actions and 13% on timeliness. The three failure modes form a coupled bottleneck: decisive reductions appear only when all three components act together.

Chinese Translation

大型语言模型（LLM）代理主要在短期单任务轨迹上进行基准测试，但实际部署往往在竞争环境中运行数小时，暴露出不同类型的失败。我们使用临床环境模拟器（Clinical Environment Simulator, CES）作为测试平台，在该平台上，代理在持续的时间和资源压力下管理整个急诊科的轮班：在结构化的多维评分下，长时间执行失败在单次执行中显著表现出来。在CES上，当前代理在大多数情况下能够做出正确的诊断，但未能及时和完整地执行关键行动，揭示了执行差距。我们将这一差距归因于三种长时间失败模式，每种模式都通过每个轨迹的计数器进行操作：指令遵循漂移、治疗不完整性以及及时性上的严重性公平差距。随后，我们引入了阿斯克勒庇俄斯（Asclepius），一种具有自我演变框架的适应性代理，能够根据轨迹级反馈在轮班之间重写操作手册，外部化的临床技能库提供高风险方案知识，以及三个独立的子代理在患者队列中分配每轮决策。在未在框架演变过程中观察到的保留批次上，阿斯克勒庇俄斯在关键行动的正确性上比强基线代理框架提高了22%（p = 0.024），同时保持了诊断准确性，并在来自三个模型家族的五位LLM评审者中获得了一致的提升；在完整的十批次数据集中，关键行动的改善达到25%，及时性改善达到13%。这三种失败模式形成了一个耦合瓶颈：只有当所有三个组件共同作用时，才会出现显著的减少。

View on arXiv Download PDF AI Translation

cs.AI / 16 / 2609.13548

AutoTailor: Automatic, User-Aligned Capability Selection and Adaptation for Web Agents

AutoTailor：自动化、用户对齐的能力选择与适应用于网络代理

Cao, Xinyun, Szekeres, Adriana, Faisal, Fazle Elahi

Abstract

Web agents can utilize reusable tools to reduce the cost and latency of low-level browser interaction, but automatically discovered tool collections can be large, redundant, and poorly aligned with user demand. We present AutoTailor, a meta-agentic framework for constructing and maintaining a compact set of trajectory-derived Model Context Protocol (MCP) APIs. Offline, AutoTailor converts web trajectories into parameterized browser-automation programs, applies a Quality Filter to remove APIs with unsuitable granularity and redundant functionality, and applies a Usage Likelihood Filter to prioritize broadly useful capabilities while preserving semantic coverage. Online, Dynamic Reselection monitors task outcomes and API usage, identifies recurring coverage gaps, adds relevant candidates, and prunes persistently unused capabilities. We evaluate AutoTailor on 106 WebArena Postmill tasks. Offline filtering reduces the initial 1,283 unrefined APIs to 87, and Dynamic Reselection produces a 33-API set. With reasoning and acting (ReAct) fallback, this set achieves 90.6% correctness, compared with 87.5% for ReAct alone, while reducing average total request-token cost by 57.8% and latency by 29.4%. Without ReAct, it achieves 60.1% correctness, marginally matching the performance of unrefined set, while reducing request-token usage by 94.9%. Together, these results show that static filtering produces a compact inventory of APIs expected to support core, high-likelihood tasks, while dynamic reselection further tailors that inventory to observed user needs. This combination improves accuracy and latency while sharply reducing token usage and end-to-end cost, demonstrating the value of user-aligned capability management for efficient web agents.

Chinese Translation

网络代理可以利用可重用工具来降低低级浏览器交互的成本和延迟，但自动发现的工具集合可能庞大、冗余且与用户需求不匹配。我们提出了AutoTailor，一个用于构建和维护紧凑的基于轨迹派生的模型上下文协议（Model Context Protocol, MCP）API集合的元代理框架。在离线阶段，AutoTailor将网络轨迹转换为参数化的浏览器自动化程序，应用质量过滤器以去除粒度不合适和功能冗余的API，并应用使用可能性过滤器以优先考虑广泛有用的能力，同时保持语义覆盖。在在线阶段，动态重新选择监控任务结果和API使用情况，识别重复的覆盖缺口，添加相关候选项，并修剪持续未使用的能力。我们在106个WebArena Postmill任务上评估了AutoTailor。离线过滤将初始的1,283个未精炼API减少到87个，而动态重新选择生成了一个包含33个API的集合。通过推理与行动（ReAct）回退，这个集合的正确率达到了90.6%，相比之下，单独使用ReAct的正确率为87.5%，同时将平均总请求令牌成本降低了57.8%，延迟降低了29.4%。在没有ReAct的情况下，其正确率为60.1%，与未精炼集合的性能基本持平，同时将请求令牌使用量减少了94.9%。这些结果表明，静态过滤产生了一个紧凑的API库存，预计能够支持核心的高可能性任务，而动态重新选择进一步根据观察到的用户需求调整该库存。这种组合提高了准确性和延迟，同时大幅减少了令牌使用和端到端成本，展示了用户对齐能力管理在高效网络代理中的价值。

View on arXiv Download PDF AI Translation

cs.AI / 17 / 2609.13552

Toward a Decision-Assurance Layer for AI-Assisted Flight Planning in Air Traffic Management

面向人工智能辅助飞行计划的决策保障层在空中交通管理中的应用

Barreto, Alexandre, Matsumoto, Shou, Valverde-Rebaza, Jorge, Ataide, Cleiton, Costa, Paulo

Abstract

Generative AI is increasingly being used informally in Air Traffic Management (ATM) for tasks such as flight plan generation, trajectory interpretation, and constraint checking. Although these tools can reduce workload and accelerate planning, their non-deterministic outputs create safety and operational risks in human-in-the-loop settings. This paper proposes the AI Trust and Assurance Layer (ATAL), a model-agnostic decision assurance architecture that evaluates whether AI-generated flight-planning outputs are sufficiently reliable for operational use. ATAL combines semantic stability under prompt variation, operational consistency of structured outputs, and normative constraint validation against domain rules, and maps these signals to a Decision Readiness Level (DRL) for human operators. An ATM-inspired experimental study shows how unsafe, inconsistent, or misleading outputs can be identified before influencing flight-plan validation or execution. Although demonstrated in aviation, the framework is also transferable to other safety-critical decision-support domains that require human oversight under regulatory constraints.

Chinese Translation

生成性人工智能在空中交通管理（ATM）中越来越多地被非正式使用，涉及飞行计划生成、轨迹解释和约束检查等任务。尽管这些工具可以减轻工作负担并加速规划，但其非确定性的输出在以人为中心的环境中带来了安全和操作风险。本文提出了人工智能信任与保障层（AI Trust and Assurance Layer, ATAL），这是一种与模型无关的决策保障架构，用于评估AI生成的飞行计划输出是否足够可靠以用于操作。ATAL结合了在提示变化下的语义稳定性、结构化输出的操作一致性以及针对领域规则的规范约束验证，并将这些信号映射到人类操作员的决策准备水平（Decision Readiness Level, DRL）。一项受ATM启发的实验研究展示了如何在影响飞行计划验证或执行之前识别出不安全、不一致或误导性的输出。尽管在航空领域进行了演示，但该框架也可以转移到其他需要在监管约束下进行人工监督的安全关键决策支持领域。

View on arXiv Download PDF AI Translation

cs.AI / 18 / 2609.13559

Carbon-Aware Routing for Function Calling in Edge-Cloud LLM Systems

面向边缘-云大型语言模型系统中函数调用的碳感知路由

Panteleaki, Aikaterini Maria, Paramanayakam, Varatheepan, Tragoudas, Spyros, Anagnostopoulos, Iraklis

Abstract

Large Language Models (LLMs) with function-calling capabilities are becoming critical for modern agentic AI systems. Nevertheless, current deployments typically route inferences to powerful cloud-based models, incurring significant energy use and carbon emissions. We address this sustainability challenge with a carbon-aware routing framework that distributes function-calling queries across a three-tier edge-cloud architecture, combining edge and cloud LLMs on heterogeneous hardware. At its core, a lightweight k-NN predictor operating in a unified semantic-lexical embedding space estimates query-specific accuracy, delay, and power consumption on each edge tier. These predictions are then combined with real-time grid carbon intensity to route every query to the lowest-emission tier capable of executing it successfully. Evaluated on state-of-the-art function-calling benchmarks and LLM families, our framework matches cloud-level accuracy while reducing operational carbon emissions by $4\times$ on average.

Chinese Translation

具备函数调用能力的大型语言模型（LLMs）正成为现代智能代理AI系统的关键组成部分。然而，当前的部署通常将推理请求路由至强大的云端模型，导致显著的能耗和碳排放。针对这一可持续性挑战，我们提出了一种碳感知路由框架，在三层边缘-云架构中分配函数调用查询，结合异构硬件上的边缘和云端LLMs。该框架核心为一个轻量级的k近邻（k-NN）预测器，运行于统一的语义-词汇嵌入空间，估计每个查询在各边缘层的特定准确率、延迟和功耗。随后，将这些预测结果与实时电网碳强度结合，确保每个查询被路由至能够成功执行且碳排放最低的层级。在最先进的函数调用基准测试和LLM系列上评估表明，我们的框架在保持云端级别准确率的同时，平均减少了4倍的运营碳排放。

View on arXiv Download PDF AI Translation

cs.AI / 19 / 2609.13561

A Hybrid Agentic AI Framework for Intelligent Supply Chain Analytics

智能供应链分析的混合智能代理框架

Lee, Xian Yeow, Inoue, Teppei, Wang, Haiyan, Gupta, Chetan

Abstract

Efficient utilization of supply chain analytics for decision making remains a significant challenge for planners, as critical tasks such as database querying, key performance indicator (KPI) analysis, demand forecasting, and performance diagnosis require heterogeneous expertise spanning data engineering, operations research, and domain knowledge. In this work, we propose an agentic system for supply chain analytics that bridges the gap between business decision-making and technical expertise, where a coordinator agent interprets user intent and delegates sub-tasks to specialized agents. The system supports both exploratory analysis and deterministic workflows, enabling planners to transition between ad hoc questions and structured processes. Domain logic is encapsulated within specialist agents and prompts, yielding a scalable, modular, and auditable design and lowering the cost of functional extension through prompt-centric development. We evaluate the proposed architecture on a test environment that replicates multi-echelon inventory management operations. Results show that our multi-agent design achieves a 90\% accuracy, which is competitive with a single agent baseline while reducing input token usage by roughly fourfold, substantially improving scalability and cost-efficiency. Furthermore, we provide case studies to demonstrate interpretable suboptimality detection and automated forecast optimization, illustrating how agentic architectures can effectively combine open-ended exploratory analysis and deterministic supply chain analytics workflows, and provide a practical pathway toward more accessible and extensible decision-support systems.

Chinese Translation

高效利用供应链分析进行决策仍然是规划者面临的一项重大挑战，因为数据库查询、关键绩效指标（KPI）分析、需求预测和绩效诊断等关键任务需要跨越数据工程、运筹学和领域知识的异质专业技能。在本研究中，我们提出了一种用于供应链分析的智能代理系统，旨在弥合商业决策与技术专长之间的差距，其中协调代理解释用户意图并将子任务委派给专业代理。该系统支持探索性分析和确定性工作流程，使规划者能够在临时问题和结构化流程之间进行转换。领域逻辑被封装在专业代理和提示中，从而实现可扩展、模块化和可审计的设计，并通过以提示为中心的开发降低功能扩展的成本。我们在一个复制多层次库存管理操作的测试环境中评估了所提出的架构。结果表明，我们的多代理设计实现了90%的准确率，具有竞争力的同时减少了输入令牌的使用量约四倍，显著提高了可扩展性和成本效率。此外，我们提供了案例研究，以展示可解释的次优性检测和自动化预测优化，说明智能代理架构如何有效结合开放式探索分析与确定性供应链分析工作流程，并为更易于访问和可扩展的决策支持系统提供了切实可行的路径。

View on arXiv Download PDF AI Translation

cs.AI / 20 / 2609.13566

Planning or Learning: Reliability and Cost in Multi-Asset Maintenance

规划还是学习：多资产维护中的可靠性与成本

Lee, Xian Yeow, Venkatraman, Chandrasekar, Farahat, Ahmed

Abstract

Industrial maintenance systems involve multiple interacting assets and shared resources, making it challenging to balance reliability and operational cost using a single decision framework. While recent work has focused on reinforcement learning (RL) for maintenance scheduling, direct comparisons with planning approaches under identical settings remain limited. In this work, we empirically compare planning and RL for multi-asset bearing maintenance using run-to-failure data. We examine how these methods behave when balancing preventive maintenance against tolerable failures across a range of failure penalty scenarios. We observed a consistent behavioral difference driven by objective formulation. Planning enforces reliability as a hard constraint and produces zero-failure policies whose total cost is largely insensitive to the magnitude of failure penalties. RL agents optimize expected cost and often trade off preventive maintenance against occasional failures as penalties vary, resulting in lower costs under low-penalty regimes but persistent non-zero failures even when penalties are high. We also investigate lightweight constraint mechanisms, including reward shaping and action masking, to encourage RL's reliability. From a practical perspective, planning may be more suitable when strict reliability is required and deployment horizons are short, whereas RL may provide cost-efficient policies when limited failures are acceptable and long-run operational efficiency is prioritized. Overall, this study clarifies the trade-offs between reliability and cost in multi-asset maintenance and suggests that planning and RL are complementary approaches. Beyond these findings, the controlled benchmark protocol itself that unifies environment, cost model, and evaluation across paradigms, offers a reusable template for comparing decision-making approaches in other maintenance settings.

Chinese Translation

工业维护系统涉及多个相互作用的资产和共享资源，使得在单一决策框架下平衡可靠性和运营成本变得具有挑战性。尽管近期的研究集中于使用强化学习（RL）进行维护调度，但在相同条件下与规划方法的直接比较仍然有限。在本研究中，我们通过使用运行至故障数据，实证比较了多资产承载维护中的规划与RL。我们考察了在不同故障惩罚场景下，这些方法在平衡预防性维护与可容忍故障时的表现。我们观察到，由目标公式驱动的一致行为差异。规划将可靠性作为硬约束，并产生零故障政策，其总成本对故障惩罚的大小变化不敏感。RL代理优化期望成本，并且在惩罚变化时，通常在预防性维护与偶发故障之间进行权衡，导致在低惩罚情况下成本较低，但即使在高惩罚情况下也会持续出现非零故障。我们还研究了轻量级约束机制，包括奖励塑造和动作屏蔽，以鼓励RL的可靠性。从实际角度来看，当需要严格的可靠性且部署周期较短时，规划可能更为合适，而当有限的故障可接受且长期运营效率优先时，RL可能提供成本有效的政策。总体而言，本研究阐明了多资产维护中可靠性与成本之间的权衡，并建议规划与RL是互补的方法。除了这些发现之外，统一环境、成本模型和跨范式评估的受控基准协议本身，提供了一个可重用的模板，用于比较其他维护环境中的决策方法。

View on arXiv Download PDF AI Translation

cs.AI / 21 / 2609.13567

Causal multi-modal AI for personalized chemosensitivity prediction

用于个性化化疗敏感性预测的因果多模态人工智能

Biswas, Dhruva, Berrevoets, Jeroen, McClean, Alec, Bao, Linus, Park, Jungkyu, Zeng, Ken G., Cappadona, Joseph, Tang, Cerise, Liu, Chuwen, Machura, Bartosz, Wu, Yin, Speirs, Valerie, Soliman, Hatem, Bhargava, Rohit, Kabraji, Sheheryar, Khoury, Thaer, Page, David, Piening, Brian, Bifulco, Carlo, Meurs, Claudia, Westenend, Pieter, Chabaud, Sylvie, Lemonnier, Jerome, Cottu, Paul H., Dalenc, Florence, Andre, Fabrice, Penault-Llorca, Frederique Madeleine, Bachelot, Thomas, Howard, Frederick, Esteva, Francisco J., Kalinsky, Kevin, Pusztai, Lajos, Witowski, Jan, Geras, Krzysztof J.

Abstract

Chemotherapy improves survival for some patients with breast cancer, but doctors cannot reliably predict who. Current guidelines rely on recurrence scores as a proxy for treatment benefit, which may contribute to the overprescription of chemotherapy. Here we present a causal multi-modal AI model that predicts personalized chemosensitivity using routinely collected pathology and clinical information. We developed our model on a multi-national dataset of 9,141 patients (twelve cohorts, nine countries) and evaluated it on another 1,994 patients (five cohorts, three countries). The model generated treatment-specific recurrence probabilities for each patient, with near-perfect calibration and strong prognostic discrimination across both 5- and 10-year follow-up horizons. Moreover, its chemotherapy benefit predictions demonstrated robust predictive performance, and out-performed existing recurrence-score-based tests. Compared to the standard of care, using the model to support personally tailored therapeutic decisions could reduce the number of patients receiving chemotherapy by 30% while achieving the same recurrence-free rate. Tumors predicted to be highly chemosensitive displayed concordant molecular and morphological programs of proliferation, cell cycle progression, and replication stress. The model's predictive capabilities transferred zero-shot to non-breast cancers, indicating our causal multi-modal AI approach may provide a universal strategy to predict treatment outcomes across cancer types.

Chinese Translation

化疗可以改善部分乳腺癌患者的生存率，但医生无法可靠地预测哪些患者会受益。当前的指南依赖复发评分作为治疗效果的代理，这可能导致化疗的过度处方。在此，我们提出了一种因果多模态人工智能模型，该模型利用常规收集的病理和临床信息预测个性化的化疗敏感性。我们在一个包含9,141名患者（十二个队列，九个国家）的多国数据集上开发了我们的模型，并在另一个1,994名患者（五个队列，三个国家）上进行了评估。该模型为每位患者生成了特定于治疗的复发概率，具有近乎完美的校准和强大的预后区分能力，适用于5年和10年的随访期。此外，其化疗效益预测表现出强大的预测性能，并优于现有的基于复发评分的测试。与标准护理相比，使用该模型支持个性化治疗决策可以将接受化疗的患者数量减少30%，同时实现相同的无复发生率。预测为高度化疗敏感的肿瘤显示出一致的分子和形态学增殖、细胞周期进程和复制压力程序。该模型的预测能力在非乳腺癌中零样本转移，表明我们的因果多模态人工智能方法可能为跨癌症类型预测治疗结果提供了一种普遍策略。

View on arXiv Download PDF AI Translation

cs.AI / 22 / 2609.13579

How User-AI Mistreatment Occurs and Matters in Conversational Systems?

用户与人工智能的不当对待是如何发生及其重要性在对话系统中？

Zeng, Fanqi, Hasan, Sadid A., He, Chaocheng

Abstract

Safety research often focuses on model-generated harms, but users may also direct hostility, coercion, and adversarial pressure at models. Understanding how and when that occurs is essential for accurately interpreting model behaviour, alignment drift, and real-world deployment risks. In this paper, we audit 777K English LMSYS-Chat-1M conversations with two independent detectors: an eight-category lexicon for hostility directed at the model, and the dataset's moderation signal; and show that they capture different, weakly overlapping phenomena. The lexicon identifies insults, threats, and jailbreak coercion aimed at the assistant, while moderation flags are dominated by toxic-content solicitation rather than hostility at the model. Together, they mark about 5% of user turns; adjusting the narrower lexicon-harassment union for measured precision puts mistreatment aimed at the assistant at 0.90%. These absolute rates describe arena-style evaluation traffic and should not be read as deployment-wide base rates. We find that user hostility varies 13-fold across models, driven largely by who each model attracts rather than by model behaviour: first-turn hostility spreads far wider than post-response hostility, and more than fifteenfold separates the extremes even after deduplicating opening prompts. Within conversations, assistant apologies are consistently associated with higher odds of next-turn hostility under both detectors; the effect survives restricting to non-refused prior turns and to jailbreak-free conversations, and is positive in 20 of 23 models. Yet across models, more apologetic models receive less hostility overall. Finally, hostility also shows temporal structure, with coercive openings front-loading the first turn while affective hostility accumulates over a session. We release the lexicon, the detector cross-validation pipeline, and all derived tables.

Chinese Translation

安全研究通常关注模型生成的伤害，但用户也可能对模型表现出敌意、强迫和对抗性压力。理解这种情况发生的方式和时机对于准确解读模型行为、对齐漂移和现实世界部署风险至关重要。本文审计了777K个英语LMSYS-Chat-1M对话，使用了两个独立的检测器：一个针对模型的敌意的八类词汇表，以及数据集的审查信号；结果表明它们捕捉到不同且弱重叠的现象。词汇表识别针对助手的侮辱、威胁和越狱强迫，而审查标记则主要集中在有毒内容的请求上，而非对模型的敌意。两者共同标记了约5%的用户发言；调整狭义的词汇表-骚扰联合以测量精度后，针对助手的不当对待率为0.90%。这些绝对率描述了竞技场式评估流量，不应被解读为部署范围内的基准率。我们发现用户的敌意在不同模型之间变化达13倍，主要受每个模型吸引的用户群体影响，而非模型行为：首次发言的敌意传播远比回应后的敌意广泛，即使在去重开场提示后，极端值之间的差距仍超过十五倍。在对话中，助手的道歉与下一轮敌意的更高几率始终相关；这一效应在限制为未拒绝的前一轮发言和无越狱对话时依然存在，并且在23个模型中有20个表现为正相关。然而，在不同模型之间，更具道歉倾向的模型总体上收到的敌意较少。最后，敌意还表现出时间结构，强迫性的开场在第一轮中占主导，而情感敌意则在整个会话中逐渐累积。我们发布了词汇表、检测器交叉验证管道以及所有衍生表格。

View on arXiv Download PDF AI Translation

cs.AI / 23 / 2609.13580

FLoKD: Adaptive Knowledge Distillation for Federated Low-Rank LLM over Wireless Networks

FLoKD：无线网络上联邦低秩大语言模型的自适应知识蒸馏

Zhang, Xinlu, Yan, Na, Su, Yang, Deng, Yansha, Mahmoodi, Toktam

Abstract

Large language models (LLMs) have demonstrated strong capabilities across a wide range of natural language processing tasks. However, conventional fine-tuning typically relies on centralized data collection, bringing in privacy concerns. Federated learning (FL) enables collaborative LLM fine-tuning without sharing raw client data, but its deployment over bandwidth-constrained wireless networks is hindered by the communication overhead of model-parameter transmission. Although Low-Rank Adaptation (LoRA) reduces the number of trainable parameters, its communication cost still increases with model scale. Knowledge distillation avoids parameter sharing via output logits, but token-level logits in LLMs incur high communication cost due to sequence length and vocabulary size. Reducing logits lowers the cost but weakens supervision and degrades accuracy. To address these limitations, we propose FLoKD, an adaptive knowledge-distillation framework for federated LoRA fine-tuning of LLMs over wireless networks, which communicates intermediate LoRA activations as the distillation signal rather than logits or full parameters. Since transmitting all blocks over the entire public dataset remains costly, we further propose a transformer block importance scoring framework that selectively transmits the most informative blocks, and two dataset selection strategies that discard public samples deviating from the local data distribution and prioritise those most informative for distillation. Extensive experiments across multiple generative language datasets, including WikiText-103, PTB, and Dialog, demonstrate that our proposed framework reduces communication overhead by 50-65% while achieving rapid convergence to competitive perplexity compared to baselines.

Chinese Translation

大型语言模型（LLMs）在广泛的自然语言处理任务中展现出了强大的能力。然而，传统的微调通常依赖于集中式数据收集，这带来了隐私问题。联邦学习（FL）使得在不共享原始客户端数据的情况下进行协作的LLM微调成为可能，但其在带宽受限的无线网络上的部署受到模型参数传输的通信开销的制约。尽管低秩适应（LoRA）减少了可训练参数的数量，但其通信成本仍随模型规模的增加而增加。知识蒸馏通过输出logits避免了参数共享，但LLMs中的token级logits由于序列长度和词汇量的原因产生了高通信成本。减少logits会降低成本，但会削弱监督并降低准确性。为了解决这些限制，我们提出了FLoKD，一个用于无线网络上联邦LoRA微调LLMs的自适应知识蒸馏框架，该框架将中间LoRA激活作为蒸馏信号进行传输，而不是logits或完整参数。由于在整个公共数据集上传输所有块仍然成本高昂，我们进一步提出了一种变换器块重要性评分框架，选择性地传输最具信息量的块，并提出了两种数据集选择策略，丢弃偏离本地数据分布的公共样本，并优先选择对蒸馏最具信息量的样本。在多个生成语言数据集上的广泛实验，包括WikiText-103、PTB和Dialog，证明我们提出的框架在实现与基线相比的竞争性困惑度的快速收敛的同时，将通信开销减少了50-65%。

View on arXiv Download PDF AI Translation

cs.AI / 24 / 2609.13602

$\tau$-Elicitation: Benchmarking multi-turn entity extraction in voice agents

$ au$-引导：语音助手中多轮实体提取的基准测试

Ray, Soham, Barres, Victor

Abstract

Voice agents often need to collect names, addresses, identifiers, dates, and times exactly, yet end-to-end benchmarks obscure where capture fails. We introduce $\tau$-Elicitation, a 200-task voice benchmark spanning 10 entity types, controlled difficulty, caller realisms, and three environments. A matched text agent passes all tasks, but four voice configurations achieve robust exact success from 0.14 to 0.41. Agents increase verification for hard and unfamiliar entities and sometimes for incorrect captures, but not for their weakest caller voice; only 24 to 37 percent of verified errors are repaired. A scaffold that prescribes spelling, read-back, correction, and confirmation raises robust Pass$^3$ by 14 to 31 points, at a cost of 21 to 28 seconds per call. Realisms such as spelling variations and restarts do not detectably affect exact success; mispronunciation increases repair effort. These results identify strategy selection and successful recovery as the central bottlenecks in exact spoken entity collection.

Chinese Translation

语音助手通常需要准确地收集姓名、地址、标识符、日期和时间，但端到端基准测试模糊了捕获失败的地方。我们引入了$ au$-引导，这是一个涵盖10种实体类型、控制难度、呼叫者真实感和三个环境的200任务语音基准测试。一个匹配的文本代理通过了所有任务，但四种语音配置在准确成功率上表现出0.14到0.41的稳健性。代理在处理困难和不熟悉的实体时增加了验证，有时也会对错误捕获进行验证，但对其最弱的呼叫者语音并没有改善；只有24%到37%的验证错误得到了修复。一个规定拼写、回读、纠正和确认的支架将稳健的Pass$^3$提高了14到31分，但每通电话的成本为21到28秒。拼写变体和重启等真实感对准确成功没有显著影响；而错误发音则增加了修复的工作量。这些结果确定了策略选择和成功恢复是准确口语实体收集中的核心瓶颈。

View on arXiv Download PDF AI Translation

cs.AI / 25 / 2609.13637

Identity Is More Than Recall: A Benchmark for Persistent Identity in Deployed AI Agents

身份不仅仅是回忆：部署的人工智能代理中持久身份的基准测试

Zhao, Zhenyu, Zhao, Roy

Abstract

Persistent agents need evaluations that distinguish identity facts they can recall from those they express and enact. We introduce PAI-Bench, a provider-neutral benchmark for fidelity to a versioned, update-governed identity contract. It separates recall, composition, behavioral enactment, resistance, persistence, lineage, and role-conditioned updates while keeping scoring oracles outside the target process. Two frozen campaigns cover sixteen synthetic profiles, thirty-two probes, and three independently initialized target configurations, yielding 1,536 retained responses. A judge-independent literal audit finds direct-parent identifiers in 48/48 atomic responses but only 1/48 implicit self-portraits. On eight profiles, explicit field cues increase joint presence of three identity identifiers from 0/8 to 7/8 under the same four-sentence instruction. A separate startup body-label substitution increases full-designation presence from 1/8 to 7/8 while parents remain absent. These contrasts reveal prompt-dependent component selection and component-specific sensitivity to startup cues in the tested deployments. Replaying identical factorial responses also yields a Claude headline mean 12.5 percentage points below Astra's, demonstrating evaluator sensitivity separately from target behavior. The studies use single target samples per condition, with post-hoc audits and follow-ups. PAI-Bench provides a reproducible evaluation protocol for measuring factual availability, identity expression, and behavioral enactment as distinct aspects of identity-contract fidelity.

Chinese Translation

持久代理需要评估，以区分它们能够回忆的身份事实与它们所表达和实施的身份事实。我们引入了PAI-Bench，这是一个提供者中立的基准，用于评估对版本化、更新管理的身份契约的忠实度。它将回忆、组合、行为实施、抵抗、持久性、谱系和角色条件更新分开，同时将评分神谕保持在目标过程之外。两个冻结的实验涵盖了十六个合成档案、三十二个探针和三个独立初始化的目标配置，产生了1,536个保留响应。一个独立于评审的字面审计发现48/48个原子响应中存在直接父标识符，但只有1/48个隐性自画像。在八个档案中，明确的领域线索使三个身份标识符的联合出现从0/8增加到7/8，在相同的四句指令下。一个单独的启动主体标签替换将完整指定的出现从1/8增加到7/8，而父标识符仍然缺失。这些对比揭示了在测试部署中，提示依赖的组件选择和组件特定对启动线索的敏感性。重放相同的因子响应也使Claude的平均得分比Astra低12.5个百分点，显示出评估者对目标行为的敏感性。研究在每个条件下使用单一目标样本，并进行事后审计和跟进。PAI-Bench提供了一种可重复的评估协议，用于测量事实可用性、身份表达和行为实施作为身份契约忠实度的不同方面。

View on arXiv Download PDF AI Translation

cs.AI / 26 / 2609.13648

Solar Intelligence

太阳能智能

Singh, Jyotsna

Abstract

Solar energy decision support is fragmented across dashboards that provide data without explanation, research papers are slow to parse, and general-purpose language models are not solar domain specific and answer without evidence. This paper introduces Solar Intelligence, a hybrid retrieval-augmented framework that unifies structured solar analytics, evidence-grounded scientific question answering, and machine learning forecasting in one system. The platform integrates daily NASA POWER solar and meteorological data, Biosphere 2 ground-sensor readings, and a curated corpus of research papers and institutional reports. Structured queries use DuckDB SQL; scientific questions are answered by a hybrid retriever that fuses BM25 and ChromaDB dense embeddings via Reciprocal Rank Fusion, with responses grounded through a language model (llama3.2:3b). An Extreme Gradient Boosting (XGBoost) model produces daily forecasts of irradiance, temperature, and wind speed. The system is exposed via FastAPI, Streamlit, and an MCP server, so it can be used as an application, an API service, or an agent tool - by students, researchers, and energy analysts.

Chinese Translation

太阳能决策支持在各个仪表板上是碎片化的，这些仪表板提供的数据缺乏解释，研究论文的解析速度较慢，而通用语言模型并不专注于太阳能领域，回答时缺乏证据。本文介绍了太阳能智能（Solar Intelligence），这是一个混合检索增强框架，将结构化太阳能分析、基于证据的科学问答和机器学习预测统一于一个系统中。该平台整合了每日NASA POWER太阳能和气象数据、Biosphere 2地面传感器读数，以及经过筛选的研究论文和机构报告的语料库。结构化查询使用DuckDB SQL；科学问题由一个混合检索器回答，该检索器通过互惠排名融合（Reciprocal Rank Fusion）融合BM25和ChromaDB密集嵌入，回答通过语言模型（llama3.2:3b）进行支撑。极端梯度提升（XGBoost）模型生成辐照度、温度和风速的每日预测。该系统通过FastAPI、Streamlit和MCP服务器进行暴露，因此可以作为应用程序、API服务或代理工具供学生、研究人员和能源分析师使用。

View on arXiv Download PDF AI Translation

cs.AI / 27 / 2609.13657

Safety as a Constraint: Fine-Tuning a LLM Recommender to Explain Itself

安全性作为约束：微调大型语言模型推荐系统以自我解释

He, Jiashu, Kong, Emma Yanyang, Tan, JJ, Fagnan, David

Abstract

Traditional recommender systems are typically trained to predict what item users will interact with next, but not why. However, offering personalized evidence for why a user might like the predicted item is an important way to enhance the service and to raise the likelihood that the user will be genuinely interested in the recommendation. This service can be delivered by integrating a frontier-model call into the member-facing pipeline, but it will add extra cost and latency. In this paper, we train a recommender LLM to generate personalized explanations for its reccomendation, based on the user's watching history at a large video streaming service. We impose two requirements on the generated explanation: it must be faithful to the elements of the shows it links, and it must be strictly non-harmful to the user. To this end, we first train two LLM-judge reward models covering three specific criteria, and propose constrained GRPO to incorporate these different criteria. On a held-out real-world testing set, our fine-tuned model improves the all-three-criteria PASS rate rises from 0.649 to 0.956 under our own judges and from 0.677 to 0.931 under an independent judge, where as the frontier generator performs similar to the untuned recommender baseline. We conduct further experiments to show that the model's language and recommendation abilities remain unchanged. Based on these results, we conclude that an LLM-based recommender can be fine-tuned on other complex tasks without compromising its original recommendation performance, thus provide insights for further agentic user interface powered by a single model.

Chinese Translation

传统的推荐系统通常被训练来预测用户将与哪个项目进行交互，但并不解释原因。然而，提供个性化的证据来说明用户可能喜欢预测项目的原因，是增强服务的重要方式，并提高用户对推荐的真实兴趣。这项服务可以通过将前沿模型调用集成到面向用户的流程中来实现，但这会增加额外的成本和延迟。在本文中，我们训练一个推荐大型语言模型（LLM）来生成基于用户在大型视频流媒体服务上的观看历史的个性化解释。我们对生成的解释施加了两个要求：它必须忠实于其关联的节目的元素，并且必须对用户严格无害。为此，我们首先训练了两个覆盖三个特定标准的LLM评估奖励模型，并提出了约束的GRPO（Constrained GRPO）以整合这些不同的标准。在一个保留的真实世界测试集上，我们微调的模型在我们自己的评估者下，所有三个标准的通过率从0.649提高到0.956，而在独立评估者下，从0.677提高到0.931，而前沿生成器的表现与未调优的推荐基线相似。我们进行了进一步的实验，以表明模型的语言和推荐能力保持不变。基于这些结果，我们得出结论，基于LLM的推荐系统可以在不影响其原始推荐性能的情况下，微调以应对其他复杂任务，从而为进一步的代理用户界面提供了洞察，这些界面由单一模型驱动。

View on arXiv Download PDF AI Translation

cs.AI / 28 / 2609.13661

FedV-KGQA in Practice: Design Lessons and an Interactive Prototype

FedV-KGQA 的实践：设计经验与交互原型

Bappy, Md Saikat Islam Khan, Seneviratne, Oshani

Abstract

Knowledge graph question answering usually assumes that one system can reach the whole graph. In practice, facts are often held by organizations that share entity identifiers but own disjoint relation types, so no single party sees a complete reasoning chain. This poster presents the empirical findings of FedV-KGQA on multi-hop question answering over such vertically partitioned graphs. Each silo enriches its local graph and trains a knowledge graph embedding on its own triples. A server then concatenates the silo-specific entity views, anchors the projected question at the topic entity, and ranks candidates by similarity. Raw triples and relation embeddings never leave a silo. Comparing the FedV-KGQA experiments with one another yields three results. First, federated fusion recovers most of the centralized accuracy, while a single silo recovers little. Second, anchoring and enrichment matter more than the choice of embedding model. Third, the cheapest encoder depends on the target accuracy rather than on parameter count. This poster paper contributes that cross-experiment comparison, four design lessons drawn from it, and an interactive prototype that runs real inference and traces the full pipeline, per question, on released checkpoints.

Chinese Translation

知识图谱问答通常假设一个系统可以访问整个图谱。然而，在实践中，事实往往由共享实体标识符但拥有不相交关系类型的组织持有，因此没有单一方能够看到完整的推理链。本文展示了 FedV-KGQA 在这种垂直分区图谱上的多跳问答的实证研究结果。每个孤岛丰富其本地图谱，并在自己的三元组上训练知识图谱嵌入。然后，服务器将孤岛特定的实体视图连接在一起，将投影问题锚定在主题实体上，并通过相似性对候选项进行排序。原始三元组和关系嵌入从未离开孤岛。对 FedV-KGQA 实验的比较得出了三个结果。首先，联邦融合恢复了大部分集中式准确性，而单个孤岛恢复的准确性很少。其次，锚定和丰富比嵌入模型的选择更为重要。第三，最便宜的编码器取决于目标准确性，而不是参数数量。本文贡献了跨实验比较的结果、从中提取的四个设计经验，以及一个能够进行真实推理并在发布的检查点上逐个问题追踪完整流程的交互原型。

View on arXiv Download PDF AI Translation

cs.AI / 29 / 2609.13664

Cost Characterization of Vertically Partitioned Federated Knowledge Graphs

垂直分区联邦知识图谱的成本特征化

Bappy, Md Saikat Islam Khan, Seneviratne, Oshani

Abstract

Knowledge graphs are increasingly distributed across autonomous organizations that share an entity space but own disjoint subsets of relations, forming a vertical partition. Answering a multi-hop query may require combining facts from several silos, making the partitioning strategy a key data management decision that affects communication, indexing, load balance, and query latency. However, the costs associated with different partitioning strategies remain insufficiently studied. We formalize vertical partitioning as a design space and compare four strategies: semantic domain grouping, frequency-balanced partitioning, co-occurrence graph-cut partitioning, and random partitioning. We evaluate them using five metrics: communication cost, candidate index size, cross-silo path length, load balance, and end-to-end query latency. Three of the five prove to be determined by the graph and the silo count rather than by the partition, which reduces the design problem to two conflicting axes, cross-silo path length and load balance. Experiments on MetaQA and PathQuestion use a fixed federated knowledge graph question-answering architecture based on TransE embeddings and a frozen BERT encoder across three silo configurations. By keeping the learning model unchanged, we isolate the effect of partitioning and show that the trade-off between locality and balance holds only where each silo can hold several relations, weakening as the number of silos increases. The study provides practical guidance for deployments constrained by cross-silo reasoning or by silo load.

Chinese Translation

知识图谱越来越多地分布在共享实体空间但拥有不相交关系子集的自主组织之间，形成垂直分区。回答多跳查询可能需要结合来自多个孤岛的事实，这使得分区策略成为影响通信、索引、负载平衡和查询延迟的关键数据管理决策。然而，不同分区策略相关的成本仍然研究不足。我们将垂直分区形式化为一个设计空间，并比较四种策略：语义领域分组、频率平衡分区、共现图切分分区和随机分区。我们使用五个指标对它们进行评估：通信成本、候选索引大小、跨孤岛路径长度、负载平衡和端到端查询延迟。五个指标中有三个被证明是由图和孤岛数量决定的，而不是由分区决定的，这将设计问题简化为两个相互冲突的轴：跨孤岛路径长度和负载平衡。在MetaQA和PathQuestion上的实验使用了基于TransE嵌入和冻结BERT编码器的固定联邦知识图谱问答架构，涵盖三种孤岛配置。通过保持学习模型不变，我们隔离了分区的影响，并表明局部性与平衡之间的权衡仅在每个孤岛可以容纳多个关系时成立，随着孤岛数量的增加而减弱。本研究为受到跨孤岛推理或孤岛负载限制的部署提供了实用指导。

View on arXiv Download PDF AI Translation

cs.AI / 30 / 2609.13667

GeoSkill:Experience-Driven Hierarchical Skill Learning with Collaborative Revision forGeospatialAgents

GeoSkill：基于经验驱动的层次化技能学习与协作修订的地理空间智能体

Luo, Han, Xu, Xian, Liu, Yinhe, Zhong, Yanfei

Abstract

Geospatial agents are increasingly expected to support recurring and evolving analytical tasks rather than execute isolated workflows. In such settings, effective agents must distill prior execution experience into reusable geospatial procedural knowledge to guide future planning and tool use. However, existing memory-augmented paradigms struggle to summarize both long-horizon tool-chain orchestration experience and tool-level invocation constraints in geospatial analysis, while directly relying on LLM self-reflection to update experience often leads to misattribution and unreliable revisions. To address these challenges, we propose GeoSkill, an experience-driven hierarchical skill learning framework for geospatial agents. GeoSkill comprises two core components: (i) a Hierarchical Skill Bank (HSB), consisting of a Planning Skill Bank and a Tool Skill Bank, which respectively distill high-level task-planning experience and tool usage constraints, enabling structured representation and cross-task reuse of historical execution experience; and (ii) a Collaborative Trace-driven Skill Revision (CTSR) mechanism, where Judge, Critic, and Refiner collaboratively perform error identification, skill-level defect localization, and targeted modification, preventing misattributed and unreliable revisions from polluting the skill bank. GeoSkill learns and validates skills from historical executions during development, and freezes the skill bank for retrieval-only guidance on unseen tasks during deployment. Extensive experiments on EarthBench and ThinkGeo demonstrate that GeoSkill effectively transforms historical execution experience into reusable hierarchical skills, improving both end-to-end task accuracy and tool-execution reliability in geospatial tasks.

Chinese Translation

地理空间智能体越来越被期望支持重复和不断发展的分析任务，而不仅仅是执行孤立的工作流程。在这种环境下，有效的智能体必须将先前的执行经验提炼为可重用的地理空间程序知识，以指导未来的规划和工具使用。然而，现有的记忆增强范式在总结长期工具链编排经验和工具级调用约束方面面临挑战，而直接依赖大型语言模型（LLM）的自我反思来更新经验往往会导致错误归因和不可靠的修订。为了解决这些问题，我们提出了GeoSkill，一个为地理空间智能体设计的基于经验驱动的层次化技能学习框架。GeoSkill包含两个核心组件：（i）一个层次化技能库（HSB），由规划技能库和工具技能库组成，分别提炼高层次任务规划经验和工具使用约束，从而实现历史执行经验的结构化表示和跨任务重用；（ii）一个协作追踪驱动的技能修订（CTSR）机制，其中评审者、批评者和修订者协作进行错误识别、技能级缺陷定位和针对性修改，防止错误归因和不可靠的修订污染技能库。GeoSkill在开发过程中从历史执行中学习和验证技能，并在部署期间将技能库冻结，仅用于对未见任务的检索指导。在EarthBench和ThinkGeo上的大量实验表明，GeoSkill有效地将历史执行经验转化为可重用的层次化技能，提高了地理空间任务中的端到端任务准确性和工具执行可靠性。

View on arXiv Download PDF AI Translation

cs.AI / 31 / 2609.13670

Enhancing Event Candidate Acquisition for Event Linking

增强事件候选获取以支持事件链接

Zhang, Ziyang, Liu, Yinan, Xue, Boyi, Huang, Yingxuan, Wang, Bin, Yang, Xiaochun

Abstract

Event linking associates event mentions in text with entries in a knowledge base (KB), or identifies them as out-of-KB events. Although existing methods use different architectures, candidate event acquisition can still be weakened by short ambiguous mentions, noisy arguments, and evidence that is unevenly useful for retrieval. We present MACE, a Multi-Agent Candidate Event acquisition method that refines event structure before linking. MACE uses evidence-specialized LLM agents to acquire time, location, participant, and event-type evidence, exposes intermediate queries to candidate-event lookup tools, and lets a coordinator revise the evidence set before final candidate construction. Experiments on two event linking benchmarks show that adding MACE to different event linking models consistently improves accuracy. These results show that MACE improves event linking through better candidate event acquisition without modifying the event linking model.

Chinese Translation

事件链接将文本中的事件提及与知识库（KB）中的条目关联，或将其识别为非知识库事件。尽管现有方法使用不同的架构，但短小模糊的提及、嘈杂的论据以及不均匀有用的证据仍然会削弱候选事件的获取。我们提出了 MACE（多代理候选事件获取方法），该方法在链接之前对事件结构进行精炼。MACE 使用证据专用的 LLM 代理来获取时间、地点、参与者和事件类型的证据，向候选事件查找工具暴露中间查询，并允许协调者在最终候选构建之前修订证据集。在两个事件链接基准上的实验表明，将 MACE 添加到不同的事件链接模型中可以持续提高准确性。这些结果表明，MACE 通过更好的候选事件获取来改善事件链接，而无需修改事件链接模型。

View on arXiv Download PDF AI Translation

cs.AI / 32 / 2609.13672

Recoverability as a System Primitive for Long-Horizon AI Agents

可恢复性作为长时间跨度人工智能代理的系统原语

Zhang, Zhihui, Liu, Wei

Abstract

AI agents can be interrupted while editing files, calling tools, or carrying out multi-step tasks. Restarting repeats completed work, but continuing from unverified or outdated progress can carry earlier errors forward. A saved state is not necessarily a suitable place to resume. We introduce recoverability as a system primitive that makes reuse an explicit decision: select a supported starting point and a permitted recovery action, or withhold automatic continuation. Its behavioral contract binds that choice to supporting evidence, execution, and independent checks. A reference architecture connects persistence, validation, and control, with complementary runtime instances testing distinct responsibilities. Four deterministic and 20 paired file challenges demonstrate that accurate restoration and successful completion can conceal disallowed starting points. Progress controls attribute retained work to shared restoration. Event-time tests show that permission must also constrain the action, and that independently held policy evidence can expose violations even after an effect occurs. These findings establish why recovery decisions need their own evaluation, beyond restored bytes and final task success. Within supplied policies and a declared trust model, the contribution is a common, testable interface for retaining justified progress and making the conditions for its reuse explicit and enforceable.

Chinese Translation

人工智能代理在编辑文件、调用工具或执行多步骤任务时可能会被中断。重新启动会重复已完成的工作，但从未验证或过时的进度继续进行可能会将早期错误带入后续工作。保存的状态不一定是一个合适的恢复点。我们引入可恢复性作为一种系统原语，使得重用成为一个明确的决策：选择一个支持的起始点和一个允许的恢复操作，或者选择不进行自动继续。其行为契约将该选择与支持证据、执行和独立检查绑定在一起。一个参考架构将持久性、验证和控制连接起来，互补的运行时实例测试不同的责任。四个确定性和20个配对文件挑战表明，准确的恢复和成功的完成可以掩盖不允许的起始点。进度控制将保留的工作归因于共享恢复。事件时间测试表明，权限也必须限制行动，并且独立持有的政策证据可以在效果发生后揭示违规行为。这些发现确立了恢复决策需要独立评估的原因，超越了恢复的字节和最终任务的成功。在提供的政策和声明的信任模型下，贡献是一个通用的、可测试的接口，用于保留合理的进展并明确和可执行地制定其重用的条件。

View on arXiv Download PDF AI Translation

cs.AI / 33 / 2609.13676

Windowed A-K-MDP

窗口化 A-K-MDP

Yang, Xiangwen, Cho, Frankie, Chades, Iadine

Abstract

Markov decision processes (MDPs) are used to support decision-making in conservation of biodiversity, but policies, even over small state spaces, can be difficult to interpret for conservation managers. K-MDP methods address this problem by building simpler MDPs with at most K abstract states. We show that the previously proposed A-K-MDP algorithm that relies on selecting a discretisation divisor using binary search can skip better abstract states. To fix this issue, we propose Windowed A-K-MDP, an algorithm that generates every distinct feasible partition induced within a declared divisor window and evaluates candidates until reaching the ideal value loss (J = 0) or exhausting the family of candidates. Across 33 K-MDP instances, Windowed improved 25 and tied 8.

Chinese Translation

马尔可夫决策过程（MDP）被用于支持生物多样性的保护决策，但即使在小状态空间中，政策对于保护管理者来说也可能难以解释。K-MDP 方法通过构建最多包含 K 个抽象状态的简单 MDP 来解决这个问题。我们展示了之前提出的 A-K-MDP 算法依赖于使用二分搜索选择离散化因子，可能会跳过更好的抽象状态。为了解决这个问题，我们提出了窗口化 A-K-MDP，这是一种在声明的因子窗口内生成每个不同的可行划分并评估候选者的算法，直到达到理想的价值损失（J = 0）或耗尽候选者的家族。在 33 个 K-MDP 实例中，窗口化方法改善了 25 个实例，并与 8 个实例持平。

View on arXiv Download PDF AI Translation

cs.AI / 34 / 2609.13680

Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models

漂移约束优化：微调指令模型时只有方向重要

Yuan, Fei, Gao, Changjiang, Tu, Yilei, Liu, Yifeng, Huang, Shujian, Qiao, Yu

Abstract

Fine-tuning instruct models often improves target performance while inducing behavioral drift from the reference model, which can degrade existing capabilities. Rather than treating this drift as an uncontrolled consequence of optimization, we specify a behavioral drift budget before optimization and ask how to boost the target-task performance within it. Locally, behavioral drift induces a shared geometry anchored at the reference model, with the drift budget defining a boundary within this space. In this space, drift determines distance from the reference, leaving update direction as the remaining degree of freedom. Fine-tuning updates can therefore be compared through their directional efficiency, naturally reformulating fine-tuning as a direction-selection problem. This reformulation makes a concrete prediction: changing the accessible directions can qualitatively alter the outcome of fine-tuning. We test this prediction in a stringent QA-only setting, where strong instruct models are fine-tuned only on final answers but must still generate multi-step reasoning at inference. Despite this mismatch, a coarse layer-selective probe reverses the failure of QA-only fine-tuning and reveals the existence of effective directions, with multiple neighboring configurations improving target performance while preserving reasoning and general capabilities. Across Qwen3-8B and Qwen3-14B, these directions substantially improve scientific reasoning and multilingual translation. Over more than 100 languages, the resulting models match or outperform dedicated translation systems and provide a stronger initialization for subsequent reinforcement learning. Our results suggest that fine-tuning is not just about how much a model changes, but how that change is spent. https://github.com/CONE-MT/DCO and https://huggingface.co/collections/LLaMAX/dco

Chinese Translation

微调指令模型通常能够提高目标性能，但同时也会导致与参考模型的行为漂移，这可能会降低现有能力。我们并不将这种漂移视为优化的不可控后果，而是在优化之前设定一个行为漂移预算，并探讨如何在此预算内提升目标任务的性能。在局部范围内，行为漂移在参考模型上锚定了共享几何体，而漂移预算则定义了该空间内的边界。在这个空间中，漂移决定了与参考模型的距离，而更新方向则是剩余的自由度。因此，微调更新可以通过其方向效率进行比较，自然地将微调重新表述为一个方向选择问题。这种重新表述做出了一个具体的预测：改变可访问的方向可以在定性上改变微调的结果。我们在一个严格的仅问答（QA）设置中测试了这一预测，在该设置中，强大的指令模型仅在最终答案上进行微调，但在推理时仍需生成多步推理。尽管存在这种不匹配，一个粗略的层选择探针逆转了仅问答微调的失败，并揭示了有效方向的存在，多个相邻配置在保持推理和一般能力的同时提高了目标性能。在 Qwen3-8B 和 Qwen3-14B 上，这些方向显著改善了科学推理和多语言翻译。在超过 100 种语言中，生成的模型与专用翻译系统相匹配或超越，并为后续的强化学习提供了更强的初始化。我们的结果表明，微调不仅仅是关于模型变化的多少，更在于这种变化如何被利用。

View on arXiv Download PDF AI Translation

cs.AI / 35 / 2609.13712

Degraded but Not Entirely Ineffective: PE-Based Deformable Graph Neural Networks

退化但并非完全无效：基于位置编码的可变形图神经网络

Wu, Jinhua, Zhang, Xinliang

Abstract

Many real-world scenarios can be represented using graph-structured data. However, traditional GNNs that transmit messages based on first-order neighbors have long faced several fundamental contradictions: increasing depth leads to over-smoothing, long-range dependencies cause over-compression, fixed neighborhoods restrict the receptive field, and on heterophilous graphs, topological neighbors become a source of noise. Although many works have addressed these issues individually, few mechanisms can simultaneously alleviate all of these challenges. To address the aforementioned problems, we propose a Position Encoding-Based Deformable Spatial Aggregation Module (PEBDSAM) that solves them all in one step. Specifically, we utilize a deformable mechanism in the position space to identify relevant nodes to supplement the original first-order neighbor information of GNNs, allowing traditional GNNs to adapt to heterophilous scenarios. Through diagnostic experiments, we obtained several major findings: current offsets fail to have any effect; subsequently, we analyzed the causes of offset failure and why model performance still improves even after offset failure, pointing out future research directions. Based on these diagnostic experiments, we streamlined the original PEBDSAM, resulting in a simplified version, which we call the Position Encoding-Based Spatial Aggregation Module (PEBSAM). In addition, we propose a PEBSAM-Speed to adapt to large datasets. Finally, we designed the module to be plug-and-play and applied it to GCN, GAT, GIN, and GraphSAGE, achieving desirable results on three homophilous datasets and six heterophilous graph datasets.

Chinese Translation

许多现实世界场景可以用图结构数据表示。然而，传统的图神经网络（GNN）基于一阶邻居传递信息，长期以来面临几个基本矛盾：增加深度导致过平滑，长程依赖造成过压缩，固定邻域限制了感受野，而在异质图上，拓扑邻居成为噪声源。尽管许多研究单独解决了这些问题，但很少有机制能够同时缓解所有这些挑战。为了解决上述问题，我们提出了一种基于位置编码的可变形空间聚合模块（PEBDSAM），能够一步解决所有问题。具体而言，我们在位置空间中利用可变形机制来识别相关节点，以补充GNN的原始一阶邻居信息，使传统GNN能够适应异质场景。通过诊断实验，我们获得了几个主要发现：当前的偏移未能产生任何效果；随后，我们分析了偏移失败的原因以及为何即使在偏移失败后模型性能仍然改善，指出了未来的研究方向。基于这些诊断实验，我们简化了原始的PEBDSAM，得到了一个简化版本，我们称之为基于位置编码的空间聚合模块（PEBSAM）。此外，我们提出了一种PEBSAM-Speed以适应大数据集。最后，我们设计该模块为即插即用，并将其应用于GCN、GAT、GIN和GraphSAGE，在三个同质数据集和六个异质图数据集上取得了良好的结果。

View on arXiv Download PDF AI Translation

cs.AI / 36 / 2609.13716

JaxAHT: A JAX-Based Library for Ad Hoc Teamwork

JaxAHT：一个基于JAX的临时团队合作库

Wang, Caroline, Fernandez, Rolando, Mustafaoglu, Zelal Su, Kundan, Montek, Cui, Jiaxun, Xiao, Lingyun, Wang, Zhihan, Shi, Di Yang, Madhan, Aditya, Liu, Johnny, Rahman, Arrasy, Stone, Peter

Abstract

Ad Hoc Teamwork (AHT) addresses the challenge of designing agents capable of coordinating with novel partners without prior coordination. However, progress in the field is hindered by the prohibitive computational cost of the AHT research lifecycle, the lack of standardized benchmark implementations, and the absence of a diverse, validated evaluation teammate suite. In this work, we introduce JaxAHT, the first open-source, JAX-based library designed to accelerate and standardize the AHT research lifecycle. Leveraging JAX's hardware acceleration and massive parallelization capabilities, JaxAHT provides a unified framework for teammate generation, ego agent training, and evaluation against unseen teammates, achieving approximately 95x wall-clock speedup over PyTorch counterparts. Alongside the library, we contribute a diverse suite of evaluation teammates across the domains of Level-Based Foraging, Overcooked, and Hanabi. To illustrate the value of the framework, we use it to conduct a large-scale, compute-controlled benchmark study comparing teammate generation and AHT agent learning methods, finding that no algorithm consistently performs best, and that agent modeling primarily offers benefits in role-based scenarios with diverse teammates.

Chinese Translation

临时团队合作（Ad Hoc Teamwork，AHT）旨在解决设计能够与新伙伴进行协调的智能体的挑战，而无需事先的协调。然而，该领域的进展受到AHT研究生命周期的高昂计算成本、缺乏标准化基准实现以及缺乏多样化、经过验证的评估队友套件的阻碍。在本研究中，我们介绍了JaxAHT，这是第一个开源的、基于JAX的库，旨在加速和标准化AHT研究生命周期。JaxAHT利用JAX的硬件加速和大规模并行化能力，提供了一个统一的框架，用于队友生成、主智能体训练以及与未见队友的评估，相较于PyTorch的对应实现，达到了约95倍的实际运行速度提升。除了该库，我们还贡献了一套多样化的评估队友，涵盖了基于等级的觅食、Overcooked和Hanabi等领域。为了说明该框架的价值，我们利用它进行了一项大规模的、计算受控的基准研究，比较了队友生成和AHT智能体学习方法，发现没有任何算法始终表现最佳，而智能体建模主要在具有多样化队友的角色基础场景中提供了好处。

View on arXiv Download PDF AI Translation

cs.AI / 37 / 2609.13717

MANAS-2: Constrained Reconstruction for EEG Foundation Models

MANAS-2：用于EEG基础模型的约束重建

Kulkarni, Arvasu, Mishra, Aditya Ray, Jain, Mahir, Runwal, Parshva, Saini, Lakshya, Panwar, Siddharth, Singh, Sandeep

Abstract

Masked reconstruction is widely used for EEG foundation models, but optimizing reconstruction on low-SNR waveforms does not necessarily produce the most useful latent representation. We introduce MANAS-2, a new EEG foundation model that combines a Raw-Band Hybrid (RBH) masked autoencoder with Constrained Reconstruction (ConRec), a physics-motivated regularizer. RBH jointly reconstructs temporal waveform patches and compact spectral-band targets, while ConRec acts only on the temporal decoder output, penalizing differences in RMS energy between adjacent short windows of the reconstructed waveform. ConRec is intended to shape the encoder by biasing it toward the organization of oscillatory-envelope information. Across seven held-out EEG datasets, adding ConRec to an otherwise identical RBH model increases frozen ridge recovery of six-band spectral power from mean R^2=0.860 to 0.906 and recovery of inter-patch band-energy dynamics from R^2=0.283 to 0.354, while temporal waveform information remains highly recoverable from the frozen latents. Applied to a temporal-only masked autoencoder, ConRec also improves frozen downstream transfer and frequency-dependent latent geometry despite receiving no spectral targets: i.e., the effects of ConRec are architecture-independent. MANAS-2 also outperforms leading EEG Foundation Models on most downstream knowledge-transfer tasks. From the effects of ConRec, we see that a physically motivated constraint imposed through the decoder can make for a more spectrally organized and transferable latent space. MANAS-2 therefore provides a new EEG foundation model built around constrained reconstruction as a mechanism for shaping representation--rather than reconstruction--quality.

Chinese Translation

掩蔽重建在EEG基础模型中被广泛使用，但在低信噪比（SNR）波形上优化重建并不一定能产生最有用的潜在表示。我们提出了MANAS-2，这是一种新的EEG基础模型，它结合了原始带宽混合（Raw-Band Hybrid, RBH）掩蔽自编码器和约束重建（Constrained Reconstruction, ConRec），后者是一种基于物理的正则化方法。RBH同时重建时间波形片段和紧凑的频谱带目标，而ConRec仅作用于时间解码器输出，惩罚重建波形相邻短窗口之间的均方根（RMS）能量差异。ConRec旨在通过使编码器偏向振荡包络信息的组织来塑造编码器。在七个独立的EEG数据集中，将ConRec添加到一个完全相同的RBH模型中，使得六频带频谱功率的冻结脊恢复从平均R^2=0.860提高到0.906，片段间带能量动态的恢复从R^2=0.283提高到0.354，同时时间波形信息仍然可以从冻结的潜在表示中高度恢复。应用于仅时间的掩蔽自编码器，ConRec也改善了冻结下游迁移和频率依赖的潜在几何，尽管没有接收频谱目标：即，ConRec的效果与架构无关。MANAS-2在大多数下游知识迁移任务中也优于领先的EEG基础模型。通过ConRec的效果，我们看到通过解码器施加的物理动机约束可以使频谱空间更加有序和可迁移。因此，MANAS-2提供了一种围绕约束重建构建的新EEG基础模型，作为塑造表示质量而非重建质量的机制。

View on arXiv Download PDF AI Translation

cs.AI / 38 / 2609.13725

IBBench-Light: A Paired Evaluation of Task-Conditioned Responses to External Directives

IBBench-Light：对任务条件响应外部指令的配对评估

Zhou, Kainan, Qian, Gangzhen, Li, Zhaoyi, Xiao, Hang

Abstract

An external record may contain a procedure to apply or text to read, depending on the user's request. IBBench-Light tests both uses against the same record. Twelve semantic bases yield 144 matched pairs per model; four quantized instruction models produced 1,152 archived greedy responses. Paired exact-contract accuracy (PECA) requires both members to satisfy their output contracts. Qwen succeeds on 132 execute and 109 process prompts, but only 97 complete pairs, showing what marginal averages omit. We audit literal-target exposure and case normalization, then add 1,722 logged CPU generations to test directive-absent controls, twelve additional semantic bases, within-base wording changes, and generation stopping. In the pinned Phi rerun, changing the end-of-sequence (EOS) set changes exact paired success from 0/144 to 62/144. A bounded IHEval comparison uses the same SmolLM2 checkpoint and output budget while preserving its published instruction roles and scorer. The benchmark measures conditional task and output-contract success. Its task margins and paired count need to be read together with the stopping policy.

Chinese Translation

外部记录可能包含根据用户请求应用的程序或需要阅读的文本。IBBench-Light 针对相同记录测试这两种用途。十二个语义基础为每个模型生成 144 对匹配样本；四个量化指令模型产生了 1,152 个归档的贪婪响应。配对精确合约准确度（PECA）要求两个成员都满足其输出合约。Qwen 在 132 个执行和 109 个处理提示中成功，但仅完成 97 对，显示出边际平均值的遗漏。我们审计了字面目标曝光和案例规范化，然后增加了 1,722 个记录的 CPU 生成以测试缺乏指令的控制、十二个额外的语义基础、基础内措辞变化和生成停止。在固定的 Phi 重跑中，改变序列结束（EOS）集将精确配对成功率从 0/144 改变为 62/144。一个有界的 IHEval 比较使用相同的 SmolLM2 检查点和输出预算，同时保留其发布的指令角色和评分者。该基准测量条件任务和输出合约的成功率。其任务边际和配对计数需要与停止政策一起阅读。

View on arXiv Download PDF AI Translation

cs.AI / 39 / 2609.13731

Trustworthy Agentic AI: A Comprehensive Cybersecurity and Systems Survey on Threat Landscapes, Defense Architectures, and Open Challenges

可信的代理人工智能：关于威胁态势、防御架构和开放挑战的综合网络安全与系统调查

Mostafavi, Seyedakbar

Abstract

The transition from passive foundation models to autonomous, goal-directed agentic AI systems has introduced unprecedented capabilities by coupling recursive cognitive reasoning loops, persistent memory architectures, live tool execution planes, and multi-agent collaboration topologies. However, granting probabilistic neural cores execution authority across filesystems, networks, and cloud infrastructure dissolves classical security perimeters: natural language simultaneously serves as input data, internal control code, and communication protocols, exposing a Turing-complete blast radius where untrusted data represents executable instructions. This survey delivers a comprehensive systems-security reference framework for trustworthy agentic AI, synthesizing 206 foundational studies and regulatory standards. We formalize the general agent architecture as a stateful 5-tuple and establish a 6-dimensional trustworthiness taxonomy covering security, safety, privacy, explainability, fairness, and accountability. We systematically analyze threat surfaces across intra-execution loops and interaction planes, formulate a multi-layered zero-trust defense-in-depth architecture integrating Dual-LLM isolation, Capability-Based Access Control, kernel eBPF probes, and sandboxed runtimes, review standardized evaluation benchmarks, and map technical controls to international AI governance frameworks.

Chinese Translation

从被动基础模型向自主、目标导向的代理人工智能系统的转变，通过结合递归认知推理循环、持久内存架构、实时工具执行平面和多代理协作拓扑，引入了前所未有的能力。然而，赋予概率神经核心在文件系统、网络和云基础设施中执行权限，打破了传统的安全边界：自然语言同时作为输入数据、内部控制代码和通信协议，暴露出一个图灵完备的爆炸半径，其中不可信数据代表可执行指令。本调查提供了一个全面的系统安全参考框架，用于可信的代理人工智能，综合了206项基础研究和监管标准。我们将一般代理架构形式化为一个有状态的5元组，并建立了一个涵盖安全性、安全性、隐私、可解释性、公平性和问责制的六维可信性分类法。我们系统地分析了执行循环和交互平面上的威胁表面，制定了一个多层次的零信任深度防御架构，整合了双LLM隔离、基于能力的访问控制、内核eBPF探针和沙箱运行时，审查标准化评估基准，并将技术控制映射到国际人工智能治理框架。

View on arXiv Download PDF AI Translation

cs.AI / 40 / 2609.13747

How Many Thoughts Can a Vector Hold? The Capacity of Reasoning by Superposition

一个向量能容纳多少思维？通过叠加推理的能力

Gu, Hongyu, Liu, Chang, Fu, Jingwen

Abstract

Large language models solve hard problems through intermediate computations across multi-step reasoning. Traditional chain-of-thought encodes these computations as tokens. Recent continuous and recurrent methods instead move partial computations into fixed-dimensional latent states, where a single thought can superpose multiple alternatives. This raises a fundamental design question:what should continuous thoughts preserve as reasoning proceeds? An intuitive approach discards past computations and keeps only the current reasoning frontier. Storing more items seems to dilute states and waste limited representational capacity. We show this intuition can be incorrect. Under identical downstream computations, cumulative superposition retaining full reasoning history can require lower representational dimensions than frontier-only superposition holding only current alternatives. At fixed hidden width, this advantage allows latent reasoners to retain more valid evidence, distinguish more plausible downstream outcomes, and delay the point where compressed states turn unreliable. This counter-intuitive effect emerges because informative historical components coherently reinforce each other, while unrelated alternatives bring random interference. This perspective also answers a practical design question: how should models weight memories accumulated inside latent states when their future use is unknown? Across reusable weighted superpositions, prioritizing a small set of recent or salient items produces weakly-represented memories that bottleneck subsequent attention. Uniform cumulative weighting avoids this flaw, and we prove it is minimax-optimal for robust future reasoning. Our results turn superposition from an observed latent-space effect into a design principle: balanced cumulative memory lets a fixed representational budget support more reliable, reusable computations.

Chinese Translation

大型语言模型通过多步骤推理中的中间计算解决复杂问题。传统的思维链将这些计算编码为标记。最近的连续和递归方法则将部分计算转移到固定维度的潜在状态中，在这里，单个思维可以叠加多个替代方案。这引发了一个基本的设计问题：在推理过程中，连续思维应该保留什么？一种直观的方法是丢弃过去的计算，仅保留当前的推理前沿。存储更多项目似乎会稀释状态并浪费有限的表示能力。我们表明这种直觉可能是错误的。在相同的下游计算下，保留完整推理历史的累积叠加可能需要比仅持有当前替代方案的前沿叠加更低的表示维度。在固定的隐藏宽度下，这一优势使潜在推理者能够保留更多有效证据，区分更多合理的下游结果，并延迟压缩状态变得不可靠的时刻。这一反直觉的效果出现是因为信息丰富的历史成分相互一致地增强，而无关的替代方案则带来随机干扰。这一视角还回答了一个实际的设计问题：当未来的使用未知时，模型应如何权衡潜在状态中积累的记忆？在可重用的加权叠加中，优先考虑一小组最近或显著的项目会产生弱表示的记忆，从而成为后续注意力的瓶颈。均匀的累积加权避免了这一缺陷，我们证明它在稳健的未来推理中是最小最大最优的。我们的结果将叠加从一种观察到的潜在空间效应转变为一种设计原则：平衡的累积记忆使固定的表示预算支持更可靠、可重用的计算。

View on arXiv Download PDF AI Translation

cs.AI / 41 / 2609.13760

Positioning manuscripts in the scientific landscape with agentic AI

利用自主人工智能在科学领域中定位手稿

Chen, Jiawen, Zhang, Zichen, Li, Bingxuan, Sun, Quan, Zhang, Yiyan, Tam, Edric, Lin, Jinjie, Li, Didong, Li, Yun, Zhao, Bingxin

Abstract

Publishing a research manuscript is a routine yet demanding part of scientific life: time-consuming, stressful, and often uncertain in outcome. Recent advances in large language model (LLM)-based agentic AI have shown promise across a range of scientific tasks, and here we ask whether agentic AI can help researchers navigate the publication process itself by reliably inferring a manuscript's eventual publication venue from its content and literature context. We introduce PASS (Publication-oriented Agentic Scientific System), an agentic system that understands manuscripts within their domain-specific literature context and predicts top-matched publication venues. PASS positions each manuscript within its surrounding literature landscape by reconstructing its local scientific neighborhood, tracing its topic trajectory, and reasoning over field-specific journal spaces. Evaluated on a leakage-audited benchmark of over 2,000 preprints across 16 biomedical fields, PASS achieved Top-1 accuracy of 50.3% and Top-5 accuracy of 86.1%, outperforming state-of-the-art LLM baselines and established journal-selection tools. PASS-produced quality scores, such as impact potential and novelty, aligned with independent measures of publication outcome. We also found that the designed literature retrieval module is the strongest performance contributor, particularly for positioning manuscripts relative to nearby work, and that PASS maintained near-full performance from the abstract alone, whereas LLM baselines required the full manuscript text. An independent human evaluation found strong researcher agreement with PASS's manuscript understanding and recommendation rationale. PASS has been released as a public platform (https://ratemypaper.ai/) for broad researcher access.

Chinese Translation

发表研究手稿是科学生活中一项常规但要求严格的工作：耗时、压力大，并且结果往往不确定。基于大型语言模型（LLM）的自主人工智能的最新进展在一系列科学任务中展现了前景，在此我们探讨自主人工智能是否能够帮助研究人员在出版过程中，通过可靠地从手稿内容和文献背景推断出最终的出版场所。我们介绍了PASS（Publication-oriented Agentic Scientific System），这是一个自主系统，能够在特定领域的文献背景中理解手稿，并预测最匹配的出版场所。PASS通过重构手稿的局部科学邻域、追踪其主题轨迹以及在领域特定的期刊空间中进行推理，将每个手稿定位于其周围的文献景观。经过对16个生物医学领域中超过2000个预印本的泄漏审计基准评估，PASS达到了50.3%的Top-1准确率和86.1%的Top-5准确率，超越了最先进的LLM基准和现有的期刊选择工具。PASS生成的质量评分，如影响潜力和新颖性，与出版结果的独立测量相一致。我们还发现，设计的文献检索模块是性能最强的贡献者，特别是在相对于附近工作的手稿定位方面，并且PASS仅凭摘要就能保持接近满分的性能，而LLM基准则需要完整的手稿文本。独立的人类评估发现，研究人员对PASS的手稿理解和推荐理由达成了高度一致。PASS已作为公共平台（https://ratemypaper.ai/）发布，以便广泛的研究人员访问。

View on arXiv Download PDF AI Translation

cs.AI / 42 / 2609.13771

Homeostatic Continual Learning

稳态持续学习

Jin, Yue

Abstract

In this paper, I formulate a Continual Learning problem and propose a method named "Homeostatic Continual Learning" that enables an AI agent to learn continuously in a changing environment without catastrophic forgetting. The core of the method is to find outliers in the environment data when the agent experiences an outlier in its output. Through this method, the agent gradually completes its model and policy and performs well in more and more contexts. I also suggest that we may use the method to build a world model where the agent factorizes the objects in the world into features, abstract objects into comparable instances of concepts and map concepts to intents through features. I discuss the works needed to render the method practical, the connections to many fields in Artificial Intelligence and the broader implications of the method.

Chinese Translation

在本文中，我提出了一个持续学习问题，并提出了一种名为“稳态持续学习”（Homeostatic Continual Learning）的方法，使得人工智能代理能够在不断变化的环境中持续学习而不发生灾难性遗忘。该方法的核心是当代理在其输出中经历异常值时，识别环境数据中的异常值。通过这种方法，代理逐渐完善其模型和策略，并在越来越多的情境中表现良好。我还建议我们可以利用该方法构建一个世界模型，其中代理将世界中的对象分解为特征，将抽象对象转化为可比较的概念实例，并通过特征将概念映射到意图。我讨论了使该方法实际可行所需的工作、与人工智能多个领域的联系以及该方法的更广泛影响。

View on arXiv Download PDF AI Translation

cs.AI / 43 / 2609.13776

Surprising Effectiveness of Self-Demonstrations in Enhancing Schema-Ontology Mapping with LLMs

自我演示在增强模式-本体映射中的惊人有效性：与大型语言模型的结合

Thombre, Siddhesh, Patwardhan, Manasi, Sarawagi, Sunita

Abstract

Integrating heterogeneous relational databases into a centralized ontology remains a persistent challenge in enterprise knowledge representation, primarily due to semantic heterogeneity, cryptic schema naming, missing metadata, and the abstraction gap between relational schemas and ontological models. Although large language models (LLMs) offer strong semantic reasoning capabilities, we show that directly applying them through one-shot prompting or naive multi-stage pipelines leads to poor performance for schema-ontology mapping. This paper presents a self-demonstration-driven approach that combines a neuro-symbolic task decomposition with a novel mechanism for automatically generating pattern-guided, dependency-aware demonstrations to address this integration challenge. Our approach incorporates two key strategies to achieve substantial accuracy gains over existing LLM-based schema integration methods: (i) a neuro-symbolic decomposition of the task into cascaded sub-tasks, where symbolic constraints structure the search space and LLMs perform semantic reasoning within each focused sub-task, and (ii) self-generated demonstrations guided by domain-agnostic patterns to supervise each sub-task. Experiments on three of the most challenging scenarios from the RODI benchmark show that our approach achieves state-of-the-art performance, substantially outperforming (25 percentage points F1 improvements) both traditional schema-to-ontology mapping techniques and recent LLM-based schema-to-ontology and schema matching approaches. Ablation studies further reveal the significant benefits of pattern-guided self-demonstrations and the complementary benefits of neuro-symbolic task decomposition.

Chinese Translation

将异构关系数据库整合到集中式本体中仍然是企业知识表示中的一个持续挑战，主要由于语义异构性、晦涩的模式命名、缺失的元数据以及关系模式与本体模型之间的抽象差距。尽管大型语言模型（LLMs）提供了强大的语义推理能力，但我们表明，直接通过一次性提示或简单的多阶段管道应用它们会导致模式-本体映射的性能较差。本文提出了一种自我演示驱动的方法，该方法结合了神经符号任务分解与一种新颖的机制，自动生成模式引导的、依赖感知的演示，以应对这一整合挑战。我们的方法包含两项关键策略，以实现相较于现有基于LLM的模式整合方法的显著准确性提升：(i) 将任务神经符号分解为级联子任务，其中符号约束结构化搜索空间，而LLMs在每个聚焦子任务内执行语义推理；(ii) 由领域无关模式引导的自生成演示来监督每个子任务。在RODI基准的三个最具挑战性的场景上的实验表明，我们的方法实现了最先进的性能，显著优于传统的模式到本体映射技术以及最近的基于LLM的模式到本体和模式匹配方法（F1提升25个百分点）。消融研究进一步揭示了模式引导的自我演示的显著益处以及神经符号任务分解的互补效益。

View on arXiv Download PDF AI Translation

cs.AI / 44 / 2609.13785

Partition Scores Are Not System Scores: Deployment-Fidelity Gaps in Decomposed Algorithm Selection

分区评分不是系统评分：分解算法选择中的部署保真度差距

Zhang, Jiachen, Tang, Yu, Zhu, Li

Abstract

Oracle-style quantities, including virtual best solvers, selected-portfolio VBS, virtual-best encodings, and best-in-family summaries, are widely reported as upper bounds on what a deployable selector could achieve. In decomposed algorithm selection, an analogous partition-level score grants an oracle choice of the best algorithm within the selected family; once the family selector is fixed, the deployable system must replace that within-family oracle with a learned within-family selector. We define the deployment-fidelity gap G(R) as the difference between partition-level and deployable end-to-end utility and derive two accounting consequences: a per-instance margin-regret stability condition that tells us when a partition-time family choice is deployment-optimal, and a sharp partition-only identification interval that, when it strictly crosses zero, prevents the partition-level report from certifying the deployable winner. Across five public algorithm-selection benchmarks spanning tabular AutoML and combinatorial CSP/SAT, every decomposed pipeline has positive G(R), ranging from 0.012 on TabZilla to 0.13 on PROTEUS-2014. Four of ten decomposed-versus-flat decisions have sign-changing point estimates; on PROTEUS-2014, a 33-point partition advantage shrinks to a 20-point end-to-end advantage. A training-side validation gap-correction diagnostic recovers the point-estimate deployable sign on all four sign-changing cells; it is a reporting aid, not a substitute for direct end-to-end evaluation. Partition and end-to-end scores should be reported side by side.

Chinese Translation

Oracle风格的量度，包括虚拟最佳求解器、选择组合的虚拟最佳求解器、虚拟最佳编码和最佳家族摘要，广泛被报告为可部署选择器所能达到的上限。在分解算法选择中，类似的分区级评分赋予了在所选家族中最佳算法的Oracle选择；一旦家族选择器固定，可部署系统必须用学习到的家族内选择器替换该家族内的Oracle。我们定义部署保真度差距 G(R) 为分区级和可部署端到端效用之间的差异，并推导出两个会计后果：一个每实例的边际后悔稳定条件，告诉我们何时分区时间的家族选择是部署最优的，以及一个严格穿越零的分区唯一识别区间，防止分区级报告认证可部署赢家。在五个公共算法选择基准测试中，涵盖了表格型自动机器学习（AutoML）和组合约束满足问题/布尔满足问题（CSP/SAT），每个分解管道都有正的 G(R)，范围从 TabZilla 的 0.012 到 PROTEUS-2014 的 0.13。十个分解与平坦决策中有四个具有符号变化的点估计；在 PROTEUS-2014 上，33 分的分区优势缩小到 20 分的端到端优势。训练侧的验证差距修正诊断在所有四个符号变化单元上恢复了点估计的可部署符号；这是一种报告辅助工具，而不是直接端到端评估的替代品。分区和端到端评分应并列报告。

View on arXiv Download PDF AI Translation

cs.AI / 45 / 2609.13800

Do Not Restart: Residual Completion for Stateful Agent Handoffs

不重启：状态代理交接的残余完成

Deng, Runzhi, Zhong, Yiming, Zhao, Fang, Zhou, Pan

Abstract

Routing and cascades reduce tool-agent cost by transferring control across models, but stateful handoffs must preserve accepted choices, realized effects, and unfinished obligations. We formulate this as commitment-constrained residual completion and introduce Commitment-Frontier Residual Completion (CFRC). CFRC enforces target-before-proposal, whole-proposal-before-authority, and live-evidence-before-success: it freezes a residual contract from accepted progress, closes the successor continuation into an evidence-linked graph, and admits execution only when the remainder is covered, with live receipts discharging obligations. We establish contract-relative partial correctness, which extends to the original residual request under complete contract construction. Across five environments and two same-provider model pairs, CFRC achieves comparable macro accuracy to strong full-task agents at only 22.0%-34.6% of their inference cost, with additional cross-provider results demonstrating broader transfer.

Chinese Translation

路由和级联通过在模型之间转移控制来降低工具代理成本，但状态交接必须保留已接受的选择、实现的效果和未完成的义务。我们将此表述为受承诺约束的残余完成，并引入承诺前沿残余完成（Commitment-Frontier Residual Completion, CFRC）。CFRC 强制执行目标优先于提案、整体提案优先于权威、实时证据优先于成功：它从已接受的进展中冻结残余合同，将后继延续关闭为一个与证据相关联的图，并仅在剩余部分得到覆盖时允许执行，同时实时收据解除义务。我们建立了合同相对的部分正确性，这一结果扩展到在完整合同构造下的原始残余请求。在五个环境和两个相同提供者模型对中，CFRC 在仅为强全任务代理的 22.0%-34.6% 推理成本的情况下，实现了可比的宏观准确性，额外的跨提供者结果展示了更广泛的迁移能力。

View on arXiv Download PDF AI Translation

cs.AI / 46 / 2609.13805

LLM-Enhanced Multi-Agent Reinforcement Learning for Unified Electric Vehicles-Charging Station-Grid Optimization in Public Charging Systems

基于LLM增强的多智能体强化学习在公共充电系统中统一电动汽车-充电站-电网优化的应用

Zhang, Yang, Xie, Lindong, Wang, Chongyu, Li, Gaojunjie, Bu, Siqi, Chung, Edward

Abstract

In the era of the Internet of Things (IoT), coordinating connected electric vehicle (EV) charging scheduling to balance EV charging satisfaction, station profitability, and smart grid stability presents a complex multi-objective challenge. Existing Multi-Agent Reinforcement Learning (MARL) approaches often struggle with high-dimensional state spaces generated by massive IoT sensing data and conflicting stakeholder interests. This paper proposes a novel LLM-enhanced MARL framework that, for the first time, simultaneously optimizes the Grid, EVs, and Stations within a unified loop. By integrating Large Language Model (LLM), we address two critical bottlenecks: interpretable feature selection and adaptive multi-objective balancing. The LLM analyzes real-time IoT-collected environmental states to extract physically significant features and dynamically assigns weights to conflicting objectives-including profit, user satisfaction, and grid load-using semantic reasoning instead of complex manual tuning. Extensive experiments demonstrate that our framework significantly outperforms state-of-the-art baselines, achieving superior market efficiency while reducing training time by over 70%. This approach offers a scalable, transparent solution for efficient and sustainable IoT-enabled urban charging infrastructure management.

Chinese Translation

在物联网（IoT）时代，协调连接的电动汽车（EV）充电调度，以平衡电动汽车充电满意度、充电站盈利能力和智能电网稳定性，构成了一项复杂的多目标挑战。现有的多智能体强化学习（MARL）方法常常在处理由大量物联网传感数据生成的高维状态空间和利益相关者之间的冲突利益时面临困难。本文提出了一种新颖的LLM增强的MARL框架，首次在统一循环中同时优化电网、电动汽车和充电站。通过集成大型语言模型（LLM），我们解决了两个关键瓶颈：可解释的特征选择和自适应的多目标平衡。LLM分析实时收集的物联网环境状态，以提取物理上重要的特征，并使用语义推理动态分配权重给冲突目标，包括利润、用户满意度和电网负载，而不是依赖复杂的手动调优。大量实验表明，我们的框架显著优于最先进的基线，市场效率更高，同时将训练时间减少了超过70%。该方法为高效、可持续的物联网城市充电基础设施管理提供了一种可扩展、透明的解决方案。

View on arXiv Download PDF AI Translation

cs.AI / 47 / 2609.13807

Bypass Observation: A Conceptual Design of a Non-Intrusive Layer-Wise Semantic Extraction Architecture

旁路观察：一种非侵入式层级语义提取架构的概念设计

Tong, Haibin, Yu, Jiang

Abstract

Large language models reason in high-dimensional hidden-state spaces, while users observe only final outputs. We introduce Bypass Observation, a non-intrusive layer-wise readout architecture that attaches read-only observation heads to selected Transformer layers without feeding their outputs back into the backbone. We consider three variants: a shared LM head across layers, layer-specific heads, and a layer- or step-adaptive head. For full-vocabulary readout, we derive a closed-form overhead approximation governed primarily by V/(12d), with representative estimates ranging from about 30% to 240%, and discuss cost reductions via sparse observation, low-rank factorization, reduced vocabularies, top-k readout, and selective positions. We argue that Bypass Observation can make model computation more observable while remaining only a partial, potentially misleading projection of hidden states. We further distinguish bypass chain-of-thought from conventional chain-of-thought: conventional reasoning tokens enter the autoregressive computation, whereas bypass readouts remain causally external at inference time, although they can still provide training signals in reinforcement learning. Finally, we discuss applications to looped and recurrent-depth Transformers, where iteration-wise readout may expose convergence, oscillation, and potential halting signals. The proposal is conceptual and analytical; systematic empirical validation remains future work.

Chinese Translation

大型语言模型在高维隐状态空间中进行推理，而用户仅观察最终输出。我们提出了旁路观察（Bypass Observation），一种非侵入式层级读取架构，它将只读观察头附加到选定的Transformer层，而不将其输出反馈到主干网络中。我们考虑了三种变体：跨层共享的语言模型头、特定层的头以及层或步长自适应的头。对于全词汇读取，我们推导出一个封闭形式的开销近似，主要由V/(12d)决定，代表性估计范围从约30%到240%，并讨论通过稀疏观察、低秩分解、减少词汇量、top-k读取和选择性位置来降低成本。我们认为，旁路观察可以使模型计算更加可观察，同时仍然只是隐状态的部分、可能具有误导性的投影。我们进一步区分旁路思维链与传统思维链：传统推理令牌进入自回归计算，而旁路读取在推理时保持因果外部，尽管它们仍然可以在强化学习中提供训练信号。最后，我们讨论了对循环和递归深度Transformer的应用，其中逐迭代读取可能暴露收敛、振荡和潜在的停止信号。该提案是概念性和分析性的；系统的实证验证仍需未来工作。

View on arXiv Download PDF AI Translation

cs.AI / 48 / 2609.13825

ViperQ: Order Flow Pattern Recognition via Auction Market Theory for Reinforcement Learning Trading

ViperQ：基于拍卖市场理论的订单流模式识别用于强化学习交易

Moustafa, Asser, Neagu, Rares-Mihail, Kalita, Jugal

Abstract

Reinforcement learning trading systems published in the academic literature overwhelmingly rely on price-aggregate state representations (OHLCV bars) or limit-order-book depth features, leaving microstructure pattern theories from the practitioner literature, namely Auction Market Theory and Market Profile, without a peer-reviewed computational instantiation. We present ViperQ, a reinforcement learning system whose state representation is built explicitly from Auction Market Theory primitives: Volume Point of Control, Value Area position, Low Volume Node flags, Cumulative Volume Delta divergence, and tape-velocity signatures, assembled into a 20-dimensional Z-normalised vector. Two Proximal Policy Optimisation agents are trained with a prospect theory-grounded asymmetric reward function that penalises losing holds at a magnitude consistent with Kahneman and Tversky's loss-aversion coefficient. Evaluated on a held-out twelve-month partition of institutional tick data the agents have never seen, ViperQ achieves +163.6% ROI on TSLA (-27.5% max drawdown, 27,019 trades) and +116.5% ROI on NVDA (-47.8% max drawdown, 12,892 trades) under zero leverage. The results establish Auction Market Theory features as a tractable structured input modality for sequential decision-making on financial time series and motivate further work on microstructure-aware policy learning.

Chinese Translation

学术文献中发布的强化学习交易系统主要依赖于价格聚合状态表示（OHLCV条）或限价订单簿深度特征，未能将来自实践文献的微观结构模式理论，即拍卖市场理论和市场概况，进行同行评审的计算实例化。我们提出了ViperQ，这是一种强化学习系统，其状态表示明确构建于拍卖市场理论原语之上：控制点成交量、价值区域位置、低成交量节点标志、累积成交量差异以及交易速度特征，组合成一个20维的Z标准化向量。两个近端策略优化（Proximal Policy Optimisation）代理使用基于前景理论的非对称奖励函数进行训练，该函数对亏损持有的惩罚程度与卡尼曼（Kahneman）和特沃斯基（Tversky）的损失厌恶系数一致。在从未见过的机构级逐笔数据的保留的十二个月分区上进行评估，ViperQ在特斯拉（TSLA）上实现了+163.6%的投资回报率（最大回撤-27.5%，交易数量27,019笔），在英伟达（NVDA）上实现了+116.5%的投资回报率（最大回撤-47.8%，交易数量12,892笔），且在零杠杆下进行。结果确立了拍卖市场理论特征作为金融时间序列顺序决策的可处理结构化输入模式，并激励了对微观结构感知政策学习的进一步研究。

View on arXiv Download PDF AI Translation

cs.AI / 49 / 2609.13849

UniCAR-RL: Seeing Better before Thinking Deeper in Visual Mathematics

UniCAR-RL：在视觉数学中更好地观察以便更深入地思考

Li, Yuzhe, Yan, Hao, Wang, Hao, Liu, Xingchen, Yu, Ya-Qi, Wu, Jihao, Liao, Minghui, Chen, Wei, Liu, Yuliang

Abstract

Multimodal Large Language Models (MLLMs) often struggle with complex mathematical visual reasoning primarily due to a lack of fine-grained perception, causing initial visual hallucinations to directly trigger cascading reasoning failures. In traditional end-to-end reinforcement learning (RL), sparse rewards fail to decouple perceptual hallucinations from logical missteps, hindering targeted perception optimization. Alternatively, fine-tuning with perception-enhanced CoT data incurs high costs and hallucinations. In this paper, we address these challenges by proposing UniCAR-RL, an annotation-free RL framework. By explicitly decoupling the optimization of perception and reasoning during the training process, it achieves isolation and optimization of both capabilities. Specifically, UniCAR-RL consists of three synergistic branches: 1) a Caption-RL branch that optimizes perception capabilities through verifier-guided reasoning validation; 2) a Reasoning-RL branch that performs logical reasoning based on a gold image description to halt cascading errors; 3) a QA-RL branch that retains native end-to-end alignment to ensure robust question-answering performance. Experiments show that UniCAR-RL substantially improves MLLMs' mathematical and visual reasoning using only raw short-answer data. Furthermore, it demonstrates strong generalization across diverse architectures and scales.

Chinese Translation

多模态大型语言模型（MLLMs）在复杂的数学视觉推理中常常面临困难，主要是由于缺乏细粒度的感知，导致初始的视觉幻觉直接触发级联推理失败。在传统的端到端强化学习（RL）中，稀疏的奖励无法将感知幻觉与逻辑失误解耦，从而阻碍了针对性的感知优化。另一方面，使用增强感知的链式思维（CoT）数据进行微调会产生高昂的成本和幻觉。在本文中，我们通过提出一种无注释的RL框架UniCAR-RL来解决这些挑战。通过在训练过程中明确解耦感知和推理的优化，它实现了这两种能力的隔离和优化。具体而言，UniCAR-RL由三个协同分支组成：1）一个通过验证者引导的推理验证来优化感知能力的Caption-RL分支；2）一个基于黄金图像描述进行逻辑推理以阻止级联错误的Reasoning-RL分支；3）一个保持原生端到端对齐以确保稳健问答性能的QA-RL分支。实验表明，UniCAR-RL仅使用原始短答案数据显著提高了MLLMs的数学和视觉推理能力。此外，它在不同架构和规模上表现出强大的泛化能力。

View on arXiv Download PDF AI Translation

cs.AI / 50 / 2609.13860

ClinAgent: A ReAct-Based Agent for Conversational Access to Clinical Trial Information

ClinAgent：基于ReAct的对话式临床试验信息访问代理

Vaccarella, Antonino, Cantini, Riccardo, Talia, Domenico, Trunfio, Paolo, Talia, Marianna, Lappano, Rosamaria, Maggiolini, Marcello

Abstract

Querying clinical trial registries remains a manual and error-prone process, requiring researchers to navigate large volumes of semi-structured data without support for natural language interaction or cross-source synthesis. To address this, we introduce ClinAgent, a conversational system based on agentic Retrieval-Augmented Generation (RAG) that enables clinicians and researchers to query clinical trial information in plain language and receive grounded, up-to-date responses across multi-turn interactions. The system centers on a Large Language Model (LLM) agent following the ReAct paradigm, which iteratively reasons over queries, selects among a set of integrated tools, and refines its actions based on intermediate outputs. These tools include a ClinicalTrials.gov search interface, a PubMed module, and a Python-based analyzer operating on a locally cached structured dataset of clinical trials. We evaluate the system using a three-phase framework assessing operational effectiveness, planning quality, tool-use efficiency, and expert qualitative judgments, comparing three LLM backends: Gemini 3.0 Flash and two variants of DeepSeek V3.2 (thinking and non-thinking). Results reveal complementary strengths, with DeepSeek (thinking mode) excelling in planning quality, while Gemini achieves the highest overall performance and strongest expert ratings. Overall, our findings highlight the potential of agentic AI systems to improve the accessibility and synthesis of clinical trial information, supporting more efficient and user-centered biomedical research workflows.

Chinese Translation

查询临床试验注册信息仍然是一个手动且容易出错的过程，研究人员需要在没有自然语言交互或跨源综合支持的情况下，处理大量半结构化数据。为了解决这一问题，我们引入了ClinAgent，这是一种基于代理的增强检索生成（Retrieval-Augmented Generation, RAG）的对话系统，使临床医生和研究人员能够用简单语言查询临床试验信息，并在多轮交互中获得基于事实的、最新的响应。该系统以遵循ReAct范式的大型语言模型（Large Language Model, LLM）代理为核心，能够对查询进行迭代推理，从一组集成工具中进行选择，并根据中间输出优化其行动。这些工具包括ClinicalTrials.gov搜索接口、PubMed模块，以及一个基于Python的分析器，后者在本地缓存的临床试验结构化数据集上运行。我们使用一个三阶段框架评估该系统的操作有效性、规划质量、工具使用效率和专家定性判断，比较了三个LLM后端：Gemini 3.0 Flash和两种DeepSeek V3.2变体（思考模式和非思考模式）。结果显示出互补的优势，DeepSeek（思考模式）在规划质量方面表现优异，而Gemini则在整体性能和专家评分方面表现最佳。总体而言，我们的研究结果突显了代理人工智能系统在改善临床试验信息的可及性和综合性方面的潜力，支持更高效和以用户为中心的生物医学研究工作流程。

View on arXiv Download PDF AI Translation

cs.AI / 51 / 2609.13879

Map Users and Mapmakers: The Scope of Cognitive Attribution from Acquired Representations

地图用户与制图者：从获得的表征出发的认知归因范围

Wu, Yiling

Abstract

An acquired representation can enlarge a system's cognitive repertoire without transferring the capacities exercised in producing that representation. This paper develops a framework for specifying that enlargement and its limits. Its central contribution is a five-part attribution table distinguishing effective tracking, application of acquired structures, acquisition from explicit specifications, acquisition from identifying observations, and retention and reuse. Each entry identifies a positive capacity commitment and a further claim requiring additional support. The argument deliberately grants meaningful content, causal efficacy, and productive inference, so that its conclusion does not depend on treating representations as inert encodings. Map and category examples show why even complete application competence leaves acquisition capacity undetermined, and why acquiring a criterion from its description differs from finding it in examples. Short formal proofs appear in an appendix. The framework is applied to Andrew Ng's world-model interpretation of Othello-GPT and to the specific indicators discussed in contemporary accounts of machine concepts. It preserves demonstrated recognition, classification, inference, and qualified acquisition while specifying what remains unestablished about criterion discovery and accumulation. The result concerns the scope of cognitive attribution rather than the constitutive conditions of concept possession: cognitive achievements deserve credit for the capacities they establish, without silently importing a broader repertoire through the labels attached to them.

Chinese Translation

获得的表征可以扩大系统的认知能力，而无需转移在产生该表征过程中所运用的能力。本文发展了一个框架，以明确这种扩展及其限制。其核心贡献是一个五部分的归因表，区分有效跟踪、获得结构的应用、从明确规范中获得、通过识别观察获得以及保留和重用。每个条目都识别出一种积极的能力承诺和一个需要额外支持的进一步主张。论证故意赋予有意义的内容、因果效力和生产性推理，因此其结论并不依赖于将表征视为无生气的编码。地图和类别的例子展示了即使是完全的应用能力也无法确定获得能力的原因，以及从描述中获得标准与从示例中找到标准的不同。简短的形式证明出现在附录中。该框架被应用于Andrew Ng对Othello-GPT的世界模型解释，以及当代机器概念讨论中的具体指标。它保留了已证明的识别、分类、推理和合格的获得，同时明确了关于标准发现和积累的未确立内容。结果涉及认知归因的范围，而非概念拥有的构成条件：认知成就应当因其建立的能力而获得认可，而不应默默地通过附加于其上的标签引入更广泛的能力范围。

View on arXiv Download PDF AI Translation

cs.AI / 52 / 2609.13918

LoRA Fine-Tuned Models for Control Systems Course Q\&A: A Multidimensional Evaluation of Model Scale and Rank Effects

用于控制系统课程问答的LoRA微调模型：模型规模和排名效应的多维评估

Lu, Shaowen, Liu, Chengxu, Zhou, Ping, Yang, Tao

Abstract

Large language models (LLMs) are increasingly used in specialized university courses, but control-systems questions require coordinated terminology, notation, derivations, and stepwise explanations. Direct general-purpose responses may be inconsistently structured and hard to verify. Using exercises and reference solutions from a Linear Control Systems course, we built a supervised fine-tuning dataset of 360 system-user-assistant conversations. We applied LoRA to Qwen2.5-3B-Instruct and Qwen2.5-7B-Instruct. With identical data splits, inference settings, and evaluation protocols, we compared base and fine-tuned models and tested LoRA ranks r=4, 8, and 16. Evaluation used ROUGE, BERTScore, and structured-output features to measure reference-answer similarity and stability of the Solution-Method-Teaching Points format. LoRA improved both similarity and structured-output stability at both sizes. On the current test set, 7B-r16 achieved the highest ROUGE-L (0.4093) and BERTScore-F1 (0.8643), while r=8 offered a better balance between performance and parameter efficiency. Bootstrap resampling showed ROUGE-L gains of 0.0764 [0.0613, 0.0915] for 3B-r16 and 0.0874 [0.0687, 0.1042] for 7B-r16; both intervals exceeded zero, indicating stable textual-similarity improvements on the current test set. These results suggest LoRA can align open-source instruction-tuned models more closely with the language and pedagogical organization of course reference answers. However, the metrics mainly capture textual similarity and formatting consistency, not domain-specific reasoning or mathematical correctness, which require expert assessment and task-specific rubrics.

Chinese Translation

大型语言模型（LLMs）在专业大学课程中的应用日益增多，但控制系统问题需要协调的术语、符号、推导和逐步解释。直接的通用响应可能结构不一致且难以验证。我们利用线性控制系统课程中的练习和参考解决方案，构建了一个包含360个系统-用户-助手对话的监督微调数据集。我们将LoRA应用于Qwen2.5-3B-Instruct和Qwen2.5-7B-Instruct。在相同的数据划分、推理设置和评估协议下，我们比较了基础模型和微调模型，并测试了LoRA的排名r=4、8和16。评估使用ROUGE、BERTScore和结构化输出特征来测量参考答案的相似性和解决方案-方法-教学点格式的稳定性。LoRA在两种规模上都提高了相似性和结构化输出的稳定性。在当前测试集上，7B-r16达到了最高的ROUGE-L（0.4093）和BERTScore-F1（0.8643），而r=8在性能和参数效率之间提供了更好的平衡。自助重抽样显示，3B-r16的ROUGE-L增益为0.0764 [0.0613, 0.0915]，7B-r16为0.0874 [0.0687, 0.1042]；这两个区间均超过零，表明在当前测试集上文本相似性有稳定的改善。这些结果表明，LoRA可以使开源的指令微调模型与课程参考答案的语言和教学组织更紧密对齐。然而，这些指标主要捕捉文本相似性和格式一致性，而非领域特定的推理或数学正确性，这需要专家评估和任务特定的评分标准。

View on arXiv Download PDF AI Translation

cs.AI / 53 / 2609.13945

SAILOR: Solver-Assisted Interactive LLM-based Optimization Recovery

SAILOR：求解器辅助的基于大语言模型的优化恢复

Sadeghi, Shaghayegh, Smith, Stephen L., Fern'andez, David C. Del Rey

Abstract

Natural-language descriptions of optimization problems may be incomplete or vague about numerical information that a solver requires, including costs, capacities, demands, bounds, and penalties. A language model can translate the description into code, but when a required value is absent it must either stop or guess. We present SAILOR, a proof-of-concept system that detects such unsupported numerical choices, asks the user targeted follow-up questions, and updates the optimization model before returning a solution. Questions are prioritized using uncertainty and solver-derived estimates of how strongly each missing value affects the current model. We evaluate the pipeline on 1,723 instances from seven masked benchmarks using an idealized simulator that returns ground-truth values. Exact objective-value agreement ranges from 27.0% to 87.6% across datasets, with 1.4--5.7 questions per instance on average. These results establish feasibility under controlled branch-and-reveal feedback; they do not measure performance with human users or general structural model repair. Code is available at: https://github.com/sshaghayeghs/SAILOR.

Chinese Translation

优化问题的自然语言描述可能不完整或对求解器所需的数值信息（包括成本、容量、需求、边界和惩罚）模糊不清。语言模型可以将描述转换为代码，但当所需值缺失时，它必须停止或进行猜测。我们提出了SAILOR，一个概念验证系统，能够检测此类不支持的数值选择，向用户提出有针对性的后续问题，并在返回解决方案之前更新优化模型。问题的优先级基于不确定性和求解器推导的每个缺失值对当前模型影响的强度进行评估。我们在七个掩蔽基准的1,723个实例上评估了该流程，使用理想化的模拟器返回真实值。不同数据集的精确目标值一致性范围为27.0%到87.6%，每个实例平均提出1.4到5.7个问题。这些结果在受控的分支和揭示反馈下建立了可行性；它们并未测量与人类用户的性能或一般结构模型修复。代码可在以下链接获取：https://github.com/sshaghayeghs/SAILOR。

View on arXiv Download PDF AI Translation

cs.AI / 54 / 2609.13995

Synthetic Data in Marketing Research: How to Evaluate and When to Trust

营销研究中的合成数据：如何评估及何时信任

Netzer, Oded, Sambandam, Rajan

Abstract

Debate over synthetic data in marketing research has polarized between claims that large language models (LLMs) make human respondents obsolete and calls to avoid them entirely. We argue that both positions obscure the more useful question: not whether synthetic respondents work, but when. Building on Brand, Israeli, and Ngwe (2026), we make three contributions. First, we distinguish three types of synthetic data (ungrounded LLM responses, segment-level personas, and individual-level digital twins) and map each to the decisions it can support. Second, we develop a taxonomy of four families of accuracy measures and suggest that the wide range of reported twin accuracy, from near-perfect to near-chance, largely reflects differences in what is being measured rather than in method quality. Aggregate measures often perform well even when little information is supplied to the LLM, and can mask a complete absence of respondent-level differentiation. Third, we introduce the forgotten question problem, in which a question is omitted from a fielded study, as a setting for twin-based augmentation of existing data. We propose an ex-ante answerability diagnostic that requires no ground truth: the R^2 of a random forest predicting twin outputs from the data used to construct the twins. Across 108 attitude questions from a nationally representative survey (N = 3,063), screening at R^2 above 0.7 raises the mean twin-human individual-level correlation by 15% and reduces the share of poorly answered questions from 25.9% to 4.3%. Embedding similarity and experienced-researcher judgment provide correlated but weaker screens.

Chinese Translation

关于营销研究中合成数据的辩论在于，部分观点认为大型语言模型（LLMs）使人类受访者变得多余，而另一些观点则呼吁完全避免使用这些模型。我们认为，这两种立场都掩盖了一个更有用的问题：不是合成受访者是否有效，而是何时有效。基于Brand、Israeli和Ngwe（2026）的研究，我们做出三项贡献。首先，我们区分了三种类型的合成数据（无基础的LLM响应、细分级别的人物角色和个体级别的数字双胞胎），并将每种类型映射到其能够支持的决策。其次，我们开发了四类准确性测量的分类法，并建议报告的双胞胎准确性范围广泛，从近乎完美到近乎随机，主要反映了测量内容的差异，而非方法质量的差异。即使在向LLM提供的信息很少的情况下，聚合测量通常表现良好，并可能掩盖受访者级别差异的完全缺失。第三，我们引入了被遗忘问题（forgotten question problem），即在进行研究时省略了某个问题，作为基于双胞胎的现有数据增强的背景。我们提出了一种无需真实基础的事前可回答性诊断：使用随机森林预测双胞胎输出的R^2，基于构建双胞胎所用的数据。在一项全国性代表性调查中（N = 3,063）对108个态度问题进行筛选时，当R^2超过0.7时，平均双胞胎与人类个体级别的相关性提高了15%，并将回答不佳的问题比例从25.9%降低到4.3%。嵌入相似性和经验丰富研究者的判断提供了相关但较弱的筛选。

View on arXiv Download PDF AI Translation

cs.AI / 55 / 2609.14011

Convergent Emergence of In-Context Learning Across Modalities

跨模态的上下文学习的趋同出现

Breslow, Nathan, Han, Seungwook, Lee, Daniel Hyunsoo, Mishra, Aayush, Liu, Anqi, Khashabi, Daniel

Abstract

Few-shot in-context learning (ICL), the capacity of a model to infer abstract patterns from input-output examples provided in its prompt and apply them to new inputs, has been extensively studied in large language models trained for next-token prediction on human text. Recently, few-shot ICL has been demonstrated in autoregressive genomic models as well. This raises a question: does ICL emerge broadly across domains, and if so, what common structure is shared? To address both, we develop a controlled cross-modality framework that instantiates the same task suite in a variety of modalities to test what we call the Convergent Emergence Hypothesis: the idea that few-shot ICL, when it emerges, shares a common cross-modality difficulty profile - i.e., tasks that benefit from ICL in one modality tend to benefit in others. We show that paired-mapping ICL emerges across six modalities (language, genome, integer sequences, time series, images, and proteins), surpasses controlled baselines, and has correlated per-task effects across five of them. Together, these results provide support for the Convergent Emergence Hypothesis in some modalities, but not all.

Chinese Translation

少样本上下文学习（ICL）是指模型从其提示中提供的输入-输出示例中推断抽象模式并将其应用于新输入的能力，这一能力在为下一标记预测而训练的大型语言模型中得到了广泛研究。最近，少样本 ICL 也在自回归基因组模型中得到了验证。这引发了一个问题：ICL 是否在各个领域广泛出现，如果是，那么共享的共同结构是什么？为了解决这两个问题，我们开发了一个受控的跨模态框架，该框架在多种模态中实例化相同的任务套件，以测试我们所称的趋同出现假设：即少样本 ICL 在出现时共享一个共同的跨模态难度特征——即，在一种模态中受益于 ICL 的任务往往在其他模态中也会受益。我们展示了配对映射 ICL 在六种模态（语言、基因组、整数序列、时间序列、图像和蛋白质）中出现，超越了受控基线，并且在其中五种模态中具有相关的每任务效应。这些结果共同支持了在某些模态中趋同出现假设的有效性，但并非在所有模态中。

View on arXiv Download PDF AI Translation

cs.AI / 56 / 2609.14015

Schizophrenia Detection from EEG Signals: A Transformer Framework with Spectrogram Representation

基于脑电图信号的精神分裂症检测：一种基于变换器的谱图表示框架

Shafiei, Abtin, Hooshmand, Mohsen, Ramezani, Majid

Abstract

Schizophrenia is a serious psychiatric disorder that affects millions of people worldwide, and its diagnosis remains primarily dependent on clinical assessment. Electroencephalography (EEG) provides a non-invasive approach to investigate brain activity and has shown potential to support automated Schizophrenia detection. However, existing EEG-based classification studies often suffer from limitations including small datasets, inconsistent preprocessing strategies, and evaluation protocols that may not adequately prevent subject-related data leakage. In this study, we propose an EEG-based Schizophrenia classification framework that transforms preprocessed EEG recordings into time-frequency representations using the Short-Time Fourier Transform. The generated spectrogram images are classified using both conventional Machine Learning algorithms, including Support Vector Machines, Random Forests, and XGBoost, and Deep Learning models, including convolutional architectures and CNN-Transformer hybrids. To ensure reliable evaluation, all data partitions are performed at the subject level. Experimental results demonstrate that the proposed approach achieves competitive classification performance, with the CNN-Transformer (CT-SZ) model achieving an AUC-ROC of 88.41% and the CNN + Squeeze and Excitation + Transformer (CST-SZ) achieving an AUC-ROC of 92.88% on the independent test set.

Chinese Translation

精神分裂症是一种严重的精神障碍，影响着全球数百万人的生活，其诊断主要依赖于临床评估。脑电图（EEG）提供了一种非侵入性的方法来研究大脑活动，并显示出支持自动化精神分裂症检测的潜力。然而，现有的基于EEG的分类研究往往存在一些局限性，包括数据集规模小、预处理策略不一致，以及评估协议可能无法充分防止与受试者相关的数据泄漏。在本研究中，我们提出了一种基于EEG的精神分裂症分类框架，该框架利用短时傅里叶变换将预处理的EEG记录转换为时频表示。生成的谱图图像使用包括支持向量机（Support Vector Machines）、随机森林（Random Forests）和XGBoost在内的传统机器学习算法，以及包括卷积架构和CNN-变换器混合模型在内的深度学习模型进行分类。为了确保可靠的评估，所有数据划分均在受试者级别进行。实验结果表明，所提出的方法实现了具有竞争力的分类性能，其中CNN-变换器（CT-SZ）模型在独立测试集上达到了88.41%的AUC-ROC，而CNN + Squeeze and Excitation + Transformer（CST-SZ）达到了92.88%的AUC-ROC。

View on arXiv Download PDF AI Translation

cs.AI / 57 / 2609.14018

VeriDx: Earning the Right to Diagnose with Disease-Centric Verification

VeriDx：通过以疾病为中心的验证赢得诊断权利

Cao, Zhong, Chen, Shuying

Abstract

A correct diagnosis can still be reached for the wrong reasons. In clinical reasoning, every disease hypothesis creates obligations: key evidence must be checked, alternatives must be ruled out, contradictions must be resolved, useful tests must be considered, and closure must be justified. Current evaluations of medical LLMs mostly focus on final answers, local steps, or isolated facts, and therefore miss these hypothesis-induced commitments. We introduce \textbf{VeriDx}, a disease-centric verification framework that links free-form diagnostic reasoning to structured disease profiles. VeriDx tracks whether each hypothesis is satisfied, unresolved, or violated its clinical obligations, exposing failures such as missing critical tests, unresolved differentials, ignored contradictions, unsupported claims, and premature closure. We instantiate VeriDx for complex respiratory diagnosis using guideline-derived disease profiles and expert-annotated longitudinal cases. Our results show that many diagnostic errors are not isolated mistakes, but broken commitments made earlier in the reasoning process.

Chinese Translation

正确的诊断可能是基于错误的理由。在临床推理中，每个疾病假设都会产生义务：必须检查关键证据，排除替代方案，解决矛盾，考虑有用的测试，并证明结论的合理性。目前对医学大型语言模型（LLMs）的评估主要集中在最终答案、局部步骤或孤立事实上，因此忽视了这些假设引发的承诺。我们提出了 extbf{VeriDx}，一个以疾病为中心的验证框架，将自由形式的诊断推理与结构化的疾病档案相连接。VeriDx跟踪每个假设是否满足、未解决或违反其临床义务，揭示诸如缺失关键测试、未解决的鉴别诊断、忽视的矛盾、缺乏支持的主张和过早结论等失败。我们为复杂的呼吸系统诊断实例化了VeriDx，使用基于指南的疾病档案和专家注释的纵向案例。我们的结果表明，许多诊断错误并不是孤立的失误，而是在推理过程中早期做出的破裂承诺。

View on arXiv Download PDF AI Translation

cs.AI / 58 / 2609.14121

Semantic Knowledge Technologies: what the Semantic Web lost sight of, and what it never had

语义知识技术：语义网忽视了什么，以及它从未拥有的东西

Zappa, Achille

Abstract

The Semantic Web set out to give information a machine-interpretable form so that software could integrate and reason over it. Its standards became scientific knowledge infrastructure, but the machine competence it promised did not follow, and the systems now answering questions over scientific knowledge are language models holding no inspectable account of what they know. This paper argues the original goal was right and the technical programme incomplete, states what is missing, and names the extended programme Semantic Knowledge Technologies: the same technical core carried out of its web-publishing origin and applied to knowledge wherever held. The diagnosis is that the standards formalised truth while omitting three things: the conditions under which a claim holds, the operations its terms permit, and any account of what a base covers. Without conditions, contradiction and applicability cannot be judged; without operational grounding, holding a statement confers no ability; without declared coverage, a system cannot recognise the boundary of its own content, which under the open-world assumption cannot be inferred. The paper fixes the word understanding to five measurable tests (check, connect, derive, act, delimit) and sets out a seven-layer architecture in which the first three layers are enabling and the rest the cognitive capabilities they make possible. It then defines three terms the programme implies: Large Knowledge Model, a model whose unit of output is a reference to an addressable claim, not a token; SLKM, the knowledge base an agent builds for itself from declared sources; and Semantic Artificial General Intelligence, stated as a falsifiable position about necessary conditions, not a system. A graded ladder replaces the untestable word general. It is offered as a research agenda, with its weakest points and refutation condition named.

Chinese Translation

语义网旨在为信息提供机器可解释的形式，以便软件能够对其进行集成和推理。其标准成为科学知识基础设施，但所承诺的机器能力并未随之而来，而现在回答科学知识问题的系统是没有可检验知识来源的语言模型。本文认为，最初的目标是正确的，而技术方案是不完整的，指出了缺失的内容，并将扩展方案命名为语义知识技术（Semantic Knowledge Technologies）：将相同的技术核心从其网络发布起源中提取并应用于任何知识的持有。诊断结果是，标准形式化了真理，但忽略了三件事：一个主张成立的条件、其术语允许的操作，以及基础所涵盖内容的任何说明。没有条件，就无法判断矛盾和适用性；没有操作基础，持有一个陈述并不赋予能力；没有声明的覆盖，系统无法识别自身内容的边界，而在开放世界假设下，这一点无法推断。本文将“理解”一词固定为五个可测量的测试（检查、连接、推导、行动、界定），并提出一个七层架构，其中前三层是使能层，其余层是它们所能实现的认知能力。接着定义了该方案所隐含的三个术语：大型知识模型（Large Knowledge Model），其输出单元是对可寻址主张的引用，而不是一个标记；自建知识库（SLKM），一个代理从声明的来源构建的知识库；以及语义人工通用智能（Semantic Artificial General Intelligence），被表述为关于必要条件的可证伪立场，而不是一个系统。一个分级阶梯取代了不可测试的“通用”一词。该方案被提出作为一个研究议程，并命名了其最薄弱的环节和反驳条件。

View on arXiv Download PDF AI Translation

cs.AI / 59 / 2609.14399

MOSCOPT: Mixture-of-Skills Collective Optimization for LLM Agents

MOSCOPT：基于技能混合的集体优化方法用于大语言模型代理

Zhang1, Zhenyu, Yang, Jiudong

Abstract

Natural language prompts and skills serve as the strategic backbone of LLM-based agents. Recent advances in prompt and skill optimization have achieved notable gains, yet all existing methods optimize a \emph{single} text template---missing the synergy among multiple complementary strategies. We propose MOSCOPT, a text-native, parameter-free algorithm that jointly optimizes a pool of $N$ skills and a gating skill $G$ that dynamically selects $K$ skills per step. To effectively optimize the skills, we build the EditAdam with internally maintained dual states. Through the three-phase interleaved updates with EditAdam, the system monotonically improves without gradient or parameter tuning. Extensive experiments and detailed ablations across 5 benchmarks and 3 target LLMs demonstrate that MOSCOPT consistently outperforms all baselines, and confirm that both the mixture-of-skills architecture with selective activation and the collective evolution with three-phase interleaving are essential to its superior performance. Code is released https://github.com/zhangzhenyu13/SummerClaw/tree/master/summerclaw/agent_trainer/algorithms/moscopt.

Chinese Translation

自然语言提示和技能作为基于大语言模型（LLM）代理的战略支柱。最近在提示和技能优化方面的进展取得了显著的成果，但所有现有方法都仅优化一个单一的文本模板，未能充分利用多种互补策略之间的协同作用。我们提出了MOSCOPT，这是一种文本原生、无参数的算法，能够联合优化一个包含$N$个技能的池以及一个动态选择每步$K$个技能的门控技能$G$。为了有效优化这些技能，我们构建了具有内部维护双状态的EditAdam。通过与EditAdam的三阶段交替更新，该系统在没有梯度或参数调优的情况下单调改进。针对5个基准和3个目标LLM的广泛实验和详细消融研究表明，MOSCOPT始终优于所有基线，并确认了具有选择性激活的技能混合架构和三阶段交替的集体演化对其卓越性能的重要性。代码已发布于https://github.com/zhangzhenyu13/SummerClaw/tree/master/summerclaw/agent_trainer/algorithms/moscopt。

View on arXiv Download PDF AI Translation

cs.AI / 60 / 2609.14408

Dynamic Learning Solutions: A System for Personalized Educational Video Generation

动态学习解决方案：个性化教育视频生成系统

Sridhar, Siddhanth, Chaurasia, Shreya, Reddy, Baddela Sai Yaswantha, Parmar, Deepak, S, Shylaja S

Abstract

We present an automated pipeline that converts NCERT textbooks into interactive video explanations that respond directly to user queries. A user uploads a PDF and asks a question; the system then generates a video-based explanation as output, handling both text and visual elements from the PDF for multi-modal retrieval and response generation. The pipeline combines a Retrieval-Augmented Generation (RAG) model with generative multimedia components. The RAG stage is optimized for the structure of NCERT textbooks and performs best on content from those books. Given a user query, the RAG model retrieves relevant content from the PDF and generates a multi-scene script containing narrative explanations and structured visual prompts aligned with the textbook's explanatory style. These prompts are passed to a Stable Diffusion module, implemented layer by layer for interpretability and control, which generates contextually relevant images. The images are then processed by DynamiCrafter to produce animated sequences. Finally, a Google Text-to-Speech module generates synchronized narration, aligning speech with the visual scenes through time-based control. The result is a coherent video explanation integrating animation, narration, and textbook-aligned visuals, transforming static educational material into an engaging learning experience. By combining multi-modal document retrieval, generative visual models, animation frameworks, and speech synthesis, this pipeline demonstrates a scalable approach to delivering interactive, personalized digital education content.

Chinese Translation

我们提出了一种自动化流程，将NCERT教科书转换为能够直接响应用户查询的互动视频讲解。用户上传PDF并提出问题；系统随后生成基于视频的解释作为输出，处理PDF中的文本和视觉元素，以实现多模态检索和响应生成。该流程结合了检索增强生成（Retrieval-Augmented Generation, RAG）模型与生成多媒体组件。RAG阶段针对NCERT教科书的结构进行了优化，并在这些书籍的内容上表现最佳。给定用户查询，RAG模型从PDF中检索相关内容，并生成包含叙述性解释和与教科书解释风格一致的结构化视觉提示的多场景脚本。这些提示被传递给稳定扩散（Stable Diffusion）模块，逐层实现可解释性和控制，从而生成上下文相关的图像。然后，这些图像由DynamiCrafter处理，以生成动画序列。最后，谷歌文本转语音（Google Text-to-Speech）模块生成同步的旁白，通过基于时间的控制将语音与视觉场景对齐。最终结果是一个连贯的视频解释，整合了动画、旁白和与教科书一致的视觉效果，将静态教育材料转变为引人入胜的学习体验。通过结合多模态文档检索、生成视觉模型、动画框架和语音合成，该流程展示了一种可扩展的方法，以提供互动的个性化数字教育内容。

View on arXiv Download PDF AI Translation

cs.AI / 61 / 2609.14412

Question's Gambit: The First Move Matters in Agentic Deep Search

问题的开局：主动深度搜索中的第一步至关重要

Rad, Radin Hamidi, Bigdeli, Amin, Arabzadeh, Negar, Ebrahimi, Sajad, Clarke, Charles L. A., Fung, Benjamin C. M., Bagheri, Ebrahim

Abstract

Deep research agents answer complex questions through iterative loops of searching, reading, and reasoning. Recent work on reasoning-intensive benchmarks such as BrowseComp-Plus shows that well-configured lexical retrieval can surface high-quality evidence, yet agents may still fail to connect documents carrying evidence to the gold documents. We identify a deep research agent's first retrieval move as an important design decision for this setting. We introduce Question's Gambit, a first-move retrieval module that decomposes the question into a set of clues, reformulates them into complementary searches, consolidates the retrieved results, and reranks the candidate pool before the agent begins its iterative search-and-reasoning process. This produces an opening context designed to support both clue aggregation and final-answer verification. We further evaluate on MultiHop-RAG to test whether these benefits transfer beyond BrowseComp-Plus to a more conventional multi-hop question structure. Experiments on BrowseComp-Plus show that Question's Gambit improves retrieval recall and downstream agent accuracy over strong baselines, improving answer accuracy from 83.1% to 90.5% with gpt-5.5 over Pi-Serini, the strongest reported agentic baseline. Our results confirm that effective agentic deep research depends not only on the tools available inside the loop, but also on the quality of the first move. We published our implementation publicly at https://github.com/radinhamidi/Question-s-Gambit.

Chinese Translation

深度研究代理通过搜索、阅读和推理的迭代循环来回答复杂问题。最近在如 BrowseComp-Plus 这样的推理密集型基准上的研究表明，良好配置的词汇检索能够呈现高质量的证据，但代理仍可能未能将携带证据的文档与金标准文档连接起来。我们将深度研究代理的第一次检索动作视为这一设置中的重要设计决策。我们引入了问题的开局（Question's Gambit），这是一个首次检索模块，它将问题分解为一组线索，将其重新表述为互补的搜索，整合检索结果，并在代理开始其迭代搜索和推理过程之前对候选池进行重新排序。这产生了一个开场上下文，旨在支持线索聚合和最终答案验证。我们进一步在 MultiHop-RAG 上进行评估，以测试这些好处是否能够超越 BrowseComp-Plus，适用于更传统的多跳问题结构。在 BrowseComp-Plus 上的实验表明，问题的开局显著提高了检索召回率和下游代理的准确性，相较于强基线，答案准确率从 83.1% 提升至 90.5%，使用 gpt-5.5 超过 Pi-Serini，这是报告的最强代理基线。我们的结果确认，有效的主动深度研究不仅依赖于循环内可用的工具，还依赖于第一次动作的质量。我们已在 https://github.com/radinhamidi/Question-s-Gambit 上公开发布了我们的实现。

View on arXiv Download PDF AI Translation

cs.AI / 62 / 2609.14422

Safety Signals to Verify NetOps Agents with Action-Level Granularity

用于验证网络操作代理的安全信号，具有动作级粒度

Labarta, Tobias, Pahde, Frederik, Boskov, Novak, Dreyer, Maximilian, Birkenberger, David, Khan, Manzoor Ahmed, Lapuschkin, Sebastian, Samek, Wojciech

Abstract

Agentic Network Operations (NetOps) are an emerging paradigm promising to enable workload-aware, self-adjustable, and reliable autonomous networks. While agents have proven their value in incident summarization and telemetry signal extraction, their effectiveness as autonomous control-loop engines heavily relies on their long-horizon reliability. One such setting is the datacenter fabric, where an agent must respond to alarms and operator intents while abstaining from high-risk actions that may cause or extend downtime. Abstention, however, presupposes that an action's impact is known pre-execution, which necessitates a per-action ground truth that NetOps agent benchmarks do not provide. We construct such a ground truth for the network repair task of NetArena. A symbolic replay of the emulated network, validated against the environment at every turn, yields the exact value of every action. From the action-level value, we derive two pre-execution targets, namely whether an action reduces the repair distance (progress) and whether it increases it (harm). We show across 10 agent models, that agent verifiers leveraging internal signals predict both harm and progress more reliably than a baseline using observable signals only. Perspectively, we aim to use these signals as safety feedback to an agent harness to abstain from risky actions and protect the target system.

Chinese Translation

代理网络操作（NetOps）是一种新兴范式，承诺能够实现工作负载感知、自我调整和可靠的自主网络。虽然代理在事件总结和遥测信号提取中证明了其价值，但作为自主控制回路引擎的有效性在很大程度上依赖于其长期可靠性。其中一个场景是数据中心架构，在该场景中，代理必须对警报和操作员意图做出响应，同时避免可能导致或延长停机时间的高风险行为。然而，避免高风险行为的前提是必须在执行前已知某一行为的影响，这就需要一个每个行为的真实情况，而NetOps代理基准并未提供。我们为NetArena的网络修复任务构建了这样的真实情况。通过对模拟网络的符号重放，并在每一步都与环境进行验证，得出了每个行为的确切值。基于行为级值，我们推导出两个执行前目标，即某一行为是否减少了修复距离（进展）以及是否增加了修复距离（危害）。我们展示了在10个代理模型中，利用内部信号的代理验证器比仅使用可观察信号的基线更可靠地预测危害和进展。从长远来看，我们旨在将这些信号作为安全反馈，指导代理避免风险行为，从而保护目标系统。

View on arXiv Download PDF AI Translation

cs.AI / 63 / 2609.14500

When does a scaling result justify a different allocation? A critical review of resource-allocation evidence for AI systems

何时规模结果可以证明不同的资源分配？对人工智能系统资源分配证据的批判性回顾

Emadi, Seyed Morteza

Abstract

AI scaling studies increasingly evaluate systems that combine a pretrained model with retrieval, search, verification, tools, and interaction. Yet a higher score under a larger budget does not by itself show where additional resources are best spent. This critical integrative review asks when a reported scaling result supports a resource-allocation decision. It compares evidence across pretraining, test-time computation, retrieval, and agent evaluation, distinguishing the performance of a tested procedure from the best performance achievable under a resource limit. The synthesis shows that three mismatches recur across this evidence: success counted before an answer is chosen, information a deployed system will not have, and costs left out of the comparison. A capability surface expresses performance as a function of budgets, mechanisms, and available information. Worked analytical examples show how the evaluation metric, deployment volume, selection rule, and stopping policy can alter an allocation conclusion. A resource envelope provides a structured record of the task, development and run-time resources, information access, and procedure behind a reported score. Its application to a published comparison illustrates which conclusions the evidence supports and which deployment questions remain unresolved. The resulting framework specifies the comparisons needed to choose among feasible systems and motivates experiments on the transfer of allocation rules across tasks and operating conditions. It does not propose a universal scaling law or infer general intelligence from benchmark gains.

Chinese Translation

人工智能规模研究越来越多地评估将预训练模型与检索、搜索、验证、工具和交互结合的系统。然而，单凭在更大预算下获得的更高分数并不能表明额外资源的最佳使用方向。本批判性综合回顾探讨了何时报告的规模结果支持资源分配决策。它比较了预训练、测试时计算、检索和代理评估等方面的证据，区分了测试程序的性能与在资源限制下可实现的最佳性能。综合分析显示，这些证据中反复出现三种不匹配情况：在选择答案之前计算的成功、已部署系统将无法获得的信息，以及比较中遗漏的成本。能力曲面将性能表示为预算、机制和可用信息的函数。经过分析的示例展示了评估指标、部署量、选择规则和停止策略如何改变资源分配结论。资源封装提供了任务、开发和运行时资源、信息访问及报告分数背后的程序的结构化记录。其在已发布比较中的应用说明了证据支持哪些结论，以及哪些部署问题仍未解决。最终框架明确了在可行系统之间选择所需的比较，并激励在不同任务和操作条件下的资源分配规则转移实验。它并未提出普遍的规模法则或从基准提升中推断出一般智能。

View on arXiv Download PDF AI Translation

cs.AI / 64 / 2609.14512

Beyond Scene Description: Multi-Agent Orchestration for Non-visual Access to Virtual Worlds

超越场景描述：多智能体编排实现虚拟世界的非视觉访问

Syed, Toqeer Ali, Akarma, Ali, Ahmad, Adeel, Hameed, Danial

Abstract

Virtual worlds now host classrooms, meetings, conferences, shops, and social venues, and nearly every interaction they expose assumes a user who can scan a three-dimensional scene, follow avatars, and read floating panels. Blind and visually impaired (BVI) users are left with assistive tools that each solve one task in isolation: naming an object, reading text, describing a scene, or planning a route. A live virtual room defeats that model: obstacles, speakers, gestures, chat, slides, and notifications arrive together, and a tool that narrates all of them trades a visual barrier for an auditory one. This paper presents MetaBlind, an architecture that distributes nonvisual access across eight specialized agents, spanning perception, navigation, social and object interaction, communication, safety and trust, memory, and personalization, and that places an Accessibility Orchestrator between those agents and the user. Agents publish candidate information into a shared accessibility context instead of speaking to the user directly. The orchestrator scores each candidate on safety relevance, goal relevance, urgency, confidence, user relevance, and estimated listening load, then releases only the items it judges relevant at that moment through speech, structured audio, or haptic output. We give the selection step a formal statement, specify the orchestration cycle as an algorithm, and define an evaluation protocol against a single-agent assistant. MetaBlind is reported at the design stage, with no prototype measurement or user study, and the protocol states which outcomes would support the design and which would refute it.

Chinese Translation

虚拟世界如今承载着教室、会议、研讨会、商店和社交场所，几乎每一次交互都假设用户能够扫描三维场景、跟随虚拟形象并阅读浮动面板。盲人和视力障碍用户（BVI）只能依赖各自解决单一任务的辅助工具：命名物体、阅读文本、描述场景或规划路线。实时虚拟房间打破了这一模式：障碍物、发言者、手势、聊天、幻灯片和通知同时出现，而一个能够叙述所有内容的工具则将视觉障碍转变为听觉障碍。本文提出了MetaBlind，一种将非视觉访问分布在八个专门智能体之间的架构，涵盖感知、导航、社交与物体交互、沟通、安全与信任、记忆和个性化，并在这些智能体与用户之间放置一个无障碍编排器。智能体将候选信息发布到共享的无障碍上下文中，而不是直接与用户对话。编排器根据安全相关性、目标相关性、紧迫性、信心、用户相关性和估计的听力负担对每个候选项进行评分，然后仅通过语音、结构化音频或触觉输出释放其判断为相关的项目。我们对选择步骤进行了正式表述，将编排周期指定为算法，并定义了针对单智能体助手的评估协议。MetaBlind目前处于设计阶段，没有原型测量或用户研究，协议中说明了哪些结果将支持设计，哪些将反驳设计。

View on arXiv Download PDF AI Translation

cs.AI / 65 / 2609.14514

OptoAgent: A Trustworthy Multi-Agent Framework for Opportunistic Vision Micro-Screening in Classroom Environments

OptoAgent：一个可信赖的多智能体框架，用于课堂环境中的机会视觉微筛查

Syed, Toqeer Ali, Akarma, Ali, Ahmad, Adeel, Muneer, Hammad

Abstract

A child with reduced distance vision often does not know that anything is wrong. Children adapt, move closer, and rarely report the difficulty, so the problem can survive years of schooling before an adult notices. School screening addresses part of this, but it runs on a schedule, depends on staffing, and is separated from the classroom moments where the difficulty appears. Smartphone and web-based acuity tests have widened access, yet every one of them still needs somebody to start a test. We present SightSentinel, an architecture that turns a wall display a child already reads from into a recurring screening site. Ordinary educational content carries short calibrated optotype probes, and eight specialized agents divide the work. Perception agents recover viewing distance, recognition accuracy, approach behavior, gaze stability, response latency, and interocular difference from each encounter. A quality agent discards observations taken under bad geometry, poor lighting, or inattention. A longitudinal agent accumulates only the surviving evidence against the child's own baseline, and an orchestrator reports a Vision Concern Score routed through a safety gate whose output range excludes diagnosis, refraction, prescription, and reassurance. The design question is whether many cheap, noisy, well-gated encounters can reach a referral decision that one scheduled test reaches late or misses. We state the formulation, the architecture, a four-stage validation protocol against clinical reference standards, and the conditions under which the approach should be rejected.

Chinese Translation

视力减弱的儿童往往并不知道自己有问题。儿童会适应环境，靠近物体，并且很少报告困难，因此这一问题可能在多年学习中得不到注意，直到成年人发现。学校筛查部分解决了这个问题，但它依赖于时间安排、人员配置，并且与困难出现的课堂时刻相隔。智能手机和基于网络的视力测试扩大了获取渠道，但每个测试仍然需要有人启动。我们提出了SightSentinel，一个将儿童已经在阅读的墙面显示器转变为一个定期筛查站的架构。普通的教育内容携带短暂的校准视力测试探针，八个专门的智能体分担工作。感知智能体从每次接触中恢复观看距离、识别准确性、接近行为、注视稳定性、反应延迟和双眼差异。质量智能体会丢弃在不良几何、光照不足或注意力不集中情况下获得的观察数据。纵向智能体仅累积与儿童自身基线相对的存活证据，而协调者则报告通过安全门路由的视觉关注评分，其输出范围排除了诊断、屈光、处方和安慰。设计问题在于，许多廉价、噪声大、良好控制的接触是否能够达到一个安排好的测试在晚些时候或遗漏的转诊决策。我们阐述了该方法的公式、架构、针对临床参考标准的四阶段验证协议，以及应拒绝该方法的条件。

View on arXiv Download PDF AI Translation

cs.AI / 66 / 2609.14592

AI Deployment Accountability Engineering: A Vision for Accountable AI in Safety-Critical Socio-Technical Systems

人工智能部署问责工程：安全关键社会技术系统中可问责人工智能的愿景

Kantarcioglu, Murat

Abstract

Artificial intelligence systems are rapidly becoming critical components in healthcare, finance, public services, and other safety-critical domains. Yet the engineering practices used to evaluate these systems remain predominantly model-centric, emphasizing properties such as accuracy, robustness, fairness, and interpretability before deployment. These properties are necessary but insufficient once an AI system operates within an ever changing socio-technical environment characterized by distribution shifts, institutional constraints, human feedback loops, privacy requirements, and interactions among multiple AI agents. This vision paper introduces AI Deployment Accountability Engineering (ADAE), a proposed AI engineering subdiscipline concerned with establishing measurable, continuous, and actionable accountability for deployed AI systems. ADAE treats accountability as a deployment-layer property rather than solely as a property of an individual model. It seeks to determine whether an AI-enabled system continues to operate within acceptable risk limits, identify the contexts in which failures emerge, attribute failures across interacting technical and human components, translate technical failures into downstream consequences, and support timely intervention. We articulate a research agenda built around four interconnected pillars: structured discovery of context-dependent failure modes, privacy-preserving accountability measurement, system-level risk analysis for agentic AI, and translation of technical failures into operational, and institutional risks. The broader goal is to establish foundational principles, mathematical tools, and system architectures for accountable AI deployment across safety-critical applications.

Chinese Translation

人工智能系统正迅速成为医疗、金融、公共服务及其他安全关键领域的重要组成部分。然而，用于评估这些系统的工程实践仍主要以模型为中心，强调在部署前的准确性、鲁棒性、公平性和可解释性等属性。这些属性虽然必要，但在人工智能系统在不断变化的社会技术环境中运行时是不够的，这种环境的特点包括分布变化、制度约束、人类反馈循环、隐私要求以及多个人工智能代理之间的互动。本文提出了人工智能部署问责工程（AI Deployment Accountability Engineering, ADAE），这是一个旨在为已部署的人工智能系统建立可衡量、持续和可操作的问责机制的人工智能工程子学科。ADAE将问责视为部署层面的属性，而不仅仅是单个模型的属性。它旨在确定一个人工智能驱动的系统是否继续在可接受的风险限度内运行，识别故障出现的上下文，归因于相互作用的技术和人类组件之间的故障，将技术故障转化为下游后果，并支持及时干预。我们阐明了围绕四个相互关联支柱构建的研究议程：结构化发现上下文依赖的故障模式、隐私保护的问责测量、代理人工智能的系统级风险分析，以及将技术故障转化为操作和制度风险。更广泛的目标是为安全关键应用中的可问责人工智能部署建立基础原则、数学工具和系统架构。

View on arXiv Download PDF AI Translation

cs.AI / 67 / 2609.14605

A note on goal-based hierarchical RL

关于基于目标的层次强化学习的说明

Murphy, Kevin

Abstract

The agent-centric general value function (ACGVF) construction of \citet{tasse2026goal} lets the agent make two decisions that are normally imposed by the environment or agent designer: which goal to pursue and when to declare a goal as finished (in addition to choosing the action). This is a very general framework that subsumes almost all prior work on reinforcement learning, control and planning, as well as more general formalisms proposed in the cognitive sciences. However, it assumes the environment is fully observed, i.e., that the observation is a sufficient statistic. In \citet{murphy2025rl}, a general agent design was proposed where the policy is based on an internal belief state $z_t$ and an internal goal; however, the goals were assumed to be externally provided. In this note, we unify and extend these two approaches using the formalism of hierarchical hidden Markov models (HHMM) \citep{murphy2001hhmm}.

Chinese Translation

代理中心的通用价值函数（ACGVF）构建方法 extit{(Tasse et al., 2026)} 使得代理能够做出两个通常由环境或代理设计者强加的决策：选择追求的目标以及何时声明目标完成（除了选择行动之外）。这是一个非常通用的框架，涵盖了几乎所有先前关于强化学习、控制和规划的研究，以及在认知科学中提出的更一般的形式。然而，它假设环境是完全可观察的，即观察结果是一个充分统计量。在 extit{(Murphy et al., 2025)} 中，提出了一种通用的代理设计，其中策略基于内部信念状态 $z_t$ 和内部目标；然而，目标被假设为外部提供。在本说明中，我们使用层次隐马尔可夫模型（HHMM） extit{(Murphy et al., 2001)} 的形式主义统一并扩展了这两种方法。

View on arXiv Download PDF AI Translation

cs.AI / 68 / 2609.14637

DynSTEER: Dynamic Stage-wise Trajectory Evaluation and Execution-time Review for Agents

DynSTEER：针对智能体的动态阶段性轨迹评估与执行时间审查

Shi, Zhichao, Zhang, Wenjie, Jiang, Xuhui, Wu, Xiaojun, Yang, Cehao, Xu, Chengjin, Guo, Jian, Wang, Yuanzhuo

Abstract

Large language model agents are increasingly deployed for long-horizon task execution. However, current evaluation paradigms face three major limitations: terminal-only assessment ignores intermediate processes and struggles to localize errors efficiently and accurately, single-reference matching penalizes valid alternative solution paths, and post-hoc trajectory judging incurs high costs without the ability to halt failed runs early. To address these issues, we propose DynSTEER, a dynamic stage-wise trajectory evaluation framework for agents. DynSTEER segments rollouts into stages anchored by key completed actions, focusing evaluation on essential milestones with adequate context while enabling targeted strategy adjustments. It compiles a path-tolerant milestone graph from public task views to respect diverse legitimate strategies without leaking ground truth. Furthermore, it adaptively routes evaluation queries across multi-tier judges and halts unrecoverable executions online to curb resource waste. Experiments demonstrate that DynSTEER improves evaluation discriminability across LLM agents by 85.2\% over native evaluation, separates all model pairs with statistical significance, and saves 34.51\% of execution steps on failed rollouts.

Chinese Translation

大型语言模型智能体越来越多地被用于长时间任务执行。然而，当前的评估范式面临三个主要限制：仅终端评估忽视了中间过程，并且难以高效准确地定位错误；单一参考匹配惩罚有效的替代解决路径；事后轨迹判断则产生高成本，无法在早期中止失败的执行。为了解决这些问题，我们提出了DynSTEER，一个针对智能体的动态阶段性轨迹评估框架。DynSTEER将执行过程分段，以关键完成动作为锚点，专注于具有充分上下文的关键里程碑进行评估，同时允许有针对性的策略调整。它从公共任务视图中编译出一个路径容忍的里程碑图，以尊重多样化的合法策略而不泄露真实答案。此外，它自适应地在多层评审者之间路由评估查询，并在线中止不可恢复的执行，以减少资源浪费。实验表明，DynSTEER在大型语言模型智能体的评估可辨别性上比原生评估提高了85.2%，在统计上显著区分了所有模型对，并在失败的执行过程中节省了34.51%的执行步骤。

View on arXiv Download PDF AI Translation

cs.AI / 69 / 2609.14642

Diffusion-Based Generation of Gait Trajectories

基于扩散的步态轨迹生成

Benasco, Damian, Carballeira-Lopez, Juan, Ramos-Rojas, Jaime, Lora-Millan, Julio S., Del-Ama, Antonio J., Rodriguez-Cianca, David, Lanillos, Pablo

Abstract

Generation of musculoskeletal gait trajectories conditioned on patient-specific parameters remains a key challenge for wearable robotics and rehabilitation. Assistive systems such as lower-limb exoskeletons require reference trajectories that adapt to individual morphology and therapeutic goals while preserving biomechanical realism. Traditional approaches rely on hand-crafted gait templates or optimization procedures that scale poorly across subjects and walking conditions. In this work, we explore conditional diffusion models for generating lower-limb joint-angle trajectories conditioned on gait parameters such as step length. We compare a baseline transformer diffusion model with a controllable diffusion transformer variant incorporating adaptive normalization and classifier-free guidance. Experiments on a dataset of 4,590 gait cycles show that diffusion models can generate realistic periodic gait trajectories while enabling some controllability variation in gait characteristics, highlighting their potential for personalized gait synthesis in assistive robotics.

Chinese Translation

基于患者特定参数生成肌肉骨骼步态轨迹仍然是可穿戴机器人和康复领域的一项关键挑战。辅助系统如下肢外骨骼需要适应个体形态和治疗目标的参考轨迹，同时保持生物力学的真实性。传统方法依赖于手工制作的步态模板或优化程序，这些方法在不同个体和行走条件下的可扩展性较差。在本研究中，我们探讨了条件扩散模型，用于生成基于步态参数（如步长）条件下的下肢关节角度轨迹。我们将基线的变换器扩散模型与一种可控的扩散变换器变体进行比较，该变体结合了自适应归一化和无分类器引导。在对4590个步态周期的数据集进行的实验中，结果表明扩散模型能够生成真实的周期性步态轨迹，同时在步态特征上实现一定的可控性变化，突显了其在辅助机器人个性化步态合成中的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 70 / 2609.14708

Lightning Weave: Improving the Accuracy-Efficiency Frontier of Reasoning Models through Capability Composition

闪电编织：通过能力组合提升推理模型的准确性-效率边界

Wu, Yecheng, Han, Song, Cai, Han

Abstract

A core goal of efficient reasoning is to improve the accuracy-efficiency frontier. However, jointly improving reasoning accuracy and inference efficiency can be challenging, as the two objectives can favor different reasoning behaviors. Independently post-trained models already offer distinct strengths in accuracy and efficiency. We introduce Lightning Weave, a post-training framework that extracts and composes these independently learned capabilities in a single student through on-policy distillation. Each acquired capability is represented by the policy shift from the model before post-training to the resulting specialist. Lightning Weave combines aligned log-ratio shifts at shared student token states and uses Tilted-Target DOPD to convert the cached signals into a stable learning target. Each anchor pair scores the cached trajectories once, enabling subsequent student training without serving multiple live anchor models concurrently. Across diverse student models and benchmarks in mathematics and code, Lightning Weave substantially improves upon the base students and achieves a state-of-the-art accuracy-efficiency frontier. On Qwen3.5-4B, it raises HMMT 2025 accuracy from 59.2% to 64.0% with 10.7% fewer response tokens, and LiveCodeBench v5 accuracy from 41.7% to 54.2% with 9.6% fewer response tokens. Adjusting the relative strengths of the anchor signals yields a strong empirical accuracy-efficiency Pareto frontier. These results establish Lightning Weave as a new practical route to efficient reasoning through capability composition. Code will be released soon.

Chinese Translation

高效推理的核心目标是改善准确性-效率边界。然而，同时提高推理准确性和推理效率可能面临挑战，因为这两个目标可能偏向不同的推理行为。独立后训练模型在准确性和效率上已经展现出各自的优势。我们提出了闪电编织（Lightning Weave），这是一种后训练框架，通过策略蒸馏提取并组合这些独立学习的能力于单一学生模型中。每个获得的能力通过后训练前模型与结果专家之间的策略转变来表示。闪电编织在共享学生令牌状态下结合对齐的对数比率变化，并使用倾斜目标的动态优化策略（Tilted-Target DOPD）将缓存信号转换为稳定的学习目标。每对锚点仅对缓存轨迹进行一次评分，从而使后续的学生训练无需同时服务多个实时锚点模型。在数学和代码的多样化学生模型和基准测试中，闪电编织显著改善了基础学生模型，并达到了最先进的准确性-效率边界。在 Qwen3.5-4B 上，它将 HMMT 2025 的准确性从 59.2% 提升至 64.0%，同时减少了 10.7% 的响应令牌；在 LiveCodeBench v5 上，它将准确性从 41.7% 提升至 54.2%，同时减少了 9.6% 的响应令牌。调整锚点信号的相对强度产生了强有力的经验准确性-效率帕累托边界。这些结果确立了闪电编织作为通过能力组合实现高效推理的新实用途径。代码将很快发布。

View on arXiv Download PDF AI Translation

cs.AI / 71 / 2609.14715

Depth and Scale in the Sub-150M Regime: JugnuLM-53M vs JugnuLM-110M

150M以下领域的深度与规模：JugnuLM-53M与JugnuLM-110M的比较

Rajput, Dushyant, Chauhan, Nirdesh, Kosaraju, Siddharth

Abstract

We scale our conventional sub-150M pretraining recipe from 53.5M to 109.7M parameters, holding the method fixed (Qwen3-style decoder with grouped-query attention, RoPE, SwiGLU, RMSNorm, QK-Norm, and a z-loss; FineWeb-Edu data) and changing only the geometry to a deep-and-thin 23-layer x 576-hidden design. The larger model improves across the board -- BLiMP 78.1 -> 81.3, ARC-Easy 51.4 -> 52.5, WikiText-2 byte-perplexity 2.04 -> 1.95 -- and its 81.3% BLiMP essentially matches GPT-X2-125M (81.28) at about 12% fewer parameters. Notably the 110M model achieves this on fewer training tokens (about 8B vs 12B), so the gain is attributable to capacity and depth, not more data. Both models are deliberately conventional; this report is a clean scaling control and the baseline rung (R0) of an ablation study of what further improves models in this regime. An ablation ladder follows: value residuals (R1) and the Muon optimizer (R2) lift ARC-Easy by a cumulative +3.6 (52.5 -> 56.1) at a near-flat BLiMP and are kept; a diverse data blend (R3) and two logit-distillation settings (R4a/R4b) are not kept -- honest negatives. R3 pins ARC-Easy to FineWeb-Edu's educational filtering rather than raw diversity; distillation from a 1.7B teacher can reach the class-leading ARC-Easy (56.99, matching GPT-X2-125M) but only at a perplexity cost that dialing KD down then erases -- so R2 remains the best kept stack.

Chinese Translation

我们将传统的150M以下预训练方案从53.5M参数扩展至109.7M参数，保持方法不变（采用Qwen3风格解码器、分组查询注意力、RoPE、SwiGLU、RMSNorm、QK-Norm以及z-loss；FineWeb-Edu数据），仅改变几何结构为深而薄的23层x 576隐藏单元设计。更大模型在各项指标上均有所提升——BLiMP从78.1提升至81.3，ARC-Easy从51.4提升至52.5，WikiText-2字节困惑度从2.04降低至1.95——其81.3%的BLiMP基本上与GPT-X2-125M（81.28）相当，但参数数量减少了约12%。值得注意的是，110M模型在较少的训练标记（约80亿对比120亿）上实现了这一点，因此提升归因于模型的容量和深度，而非更多的数据。这两个模型都是经过深思熟虑的传统设计；本报告是一个干净的规模控制，并且是对该领域进一步改进模型的消融研究的基线（R0）。接下来是消融阶梯：价值残差（R1）和Muon优化器（R2）使ARC-Easy的累计提升达到+3.6（52.5 -> 56.1），在BLiMP几乎保持不变的情况下被保留；多样化数据混合（R3）和两种logit蒸馏设置（R4a/R4b）未被保留——诚实的负面结果。R3将ARC-Easy固定在FineWeb-Edu的教育过滤上，而非原始多样性；来自1.7B教师的蒸馏可以达到行业领先的ARC-Easy（56.99，匹配GPT-X2-125M），但代价是困惑度的增加，降低KD后则会消失——因此R2仍然是最佳保留的堆栈。

View on arXiv Download PDF AI Translation

cs.AI / 72 / 2609.14716

Moral Rebel Agents: Decision-Making Under Conflicting Obligations

道德反叛代理：在冲突义务下的决策制定

Munoz-Avila, Hector, Aha, David W., Rizzo, Paola

Abstract

Autonomous agents are typically obliged to follow user-assigned tasks. However, strict obedience may conflict with moral obligations that arise during execution. This paper investigates \textbf{moral rebellion}: the ability of an autonomous agent to deviate from a user-assigned task when morally justified. We formalize five agent architectures: an \textit{amoral agent} that pursues assigned tasks without considering moral obligations, and four forms of \textbf{moral rebel agency}: (1) \textit{utilitarian agents} that opportunistically maximize task outcomes; (2) \textit{deontic} agents that enforce normative constraints; (3) \textit{utilitarian-deontic} (UD) agents that combine deontic constraints with utilitarian reasoning; and (4) \textit{dutiful} agents that additionally preserve commitments to assigned tasks. We implement these architectures within a hierarchical task network planning framework and evaluate them in a Mini Search-and-Rescue domain that exposes trade-offs among assigned-task completion, opportunistic rescue, and norm compliance. Our empirical results show that the proposed agents exhibit distinct trade-offs among rescue results, assigned-task completion, and norm compliance. In particular, the preservation of task commitments emerges as an important dimension of moral rebellion, for which the UD and dutiful agents produce substantially different behaviors despite their shared utilitarian and deontological foundations. These findings highlight the importance of commitment-aware moral reasoning for autonomous agents operating in morally consequential environments.

Chinese Translation

自主代理通常有义务遵循用户分配的任务。然而，严格的服从可能与执行过程中出现的道德义务相冲突。本文探讨了 extbf{道德反叛}：自主代理在道德上合理时偏离用户分配任务的能力。我们形式化了五种代理架构：一种 extit{非道德代理}，其在追求分配任务时不考虑道德义务，以及四种形式的 extbf{道德反叛代理}： (1) extit{功利代理}，机会主义地最大化任务结果； (2) extit{义务代理}，执行规范约束； (3) extit{功利-义务}（UD）代理，结合义务约束与功利推理； (4) extit{尽责代理}，此外还保留对分配任务的承诺。我们在一个分层任务网络规划框架内实现这些架构，并在一个迷你搜索与救援领域中评估它们，该领域揭示了分配任务完成、机会主义救援和规范遵守之间的权衡。我们的实证结果表明，所提出的代理在救援结果、分配任务完成和规范遵守之间表现出明显的权衡。特别是，任务承诺的保留成为道德反叛的重要维度，尽管UD代理和尽责代理具有相似的功利和义务基础，但它们的行为却显著不同。这些发现强调了在道德后果显著的环境中，承诺意识的道德推理对自主代理的重要性。

View on arXiv Download PDF AI Translation

cs.AI / 73 / 2609.14724

Bayesian Intelligence from the Outside

外部的贝叶斯智能

Smolin, Alex, Wilder, Bryan

Abstract

Inferring intelligence from observable behavior is a foundational challenge in artificial intelligence. We develop a theory of Bayesian intelligence for agents such as language models. Each prompt induces a possibly imperfect internal experiment; the agent updates a full-support prior by Bayes' rule and faithfully reports its posterior over the possible answers to the question. Repetitions draw fresh, independent outcomes from the same unobserved experiment at one fixed state. We show that the agent's behavior admits this explanation if and only if its reports are not fully contradictory, i.e., some state remains possible under every report across all prompts. Report frequencies and the sizes of positive probabilities impose no further restrictions. We further propose and characterize the behavioral implications of an intelligence order that makes the behavior of two agents consistent with one agent having access to a more informative experiment: there should exist a coupling of report distributions such that the more informative agent's report excludes every answer excluded by its counterpart. Finally, we show the difficulty of aggregating coarse reports from intelligent agents: unless the agent reports a belief about the complete state of the world, the optimal aggregation can assign arbitrary weights to states that have not been excluded. These results provide a basis for understanding when agents' behavior is intelligent and highlight the difficulty of rejecting Bayesian rationality.

Chinese Translation

从可观察的行为推断智能是人工智能中的一个基础性挑战。我们为诸如语言模型的智能体发展了一种贝叶斯智能理论。每个提示会引发一个可能不完美的内部实验；智能体通过贝叶斯法则更新一个全支持的先验，并忠实地报告其对问题可能答案的后验。重复实验从同一个固定状态下的未观察实验中得出新的独立结果。我们证明，只有当智能体的报告不完全矛盾时，其行为才可以用此解释，即在所有提示下，某些状态在每个报告中仍然是可能的。报告频率和正概率的大小没有进一步的限制。我们进一步提出并描述了一种智能顺序的行为含义，使得两个智能体的行为与一个智能体拥有更具信息性的实验一致：应该存在一种报告分布的耦合，使得更具信息性的智能体的报告排除了其对应者排除的每一个答案。最后，我们展示了从智能智能体聚合粗略报告的困难：除非智能体报告对世界完整状态的信念，否则最佳聚合可能会对未被排除的状态分配任意权重。这些结果为理解何时智能体的行为是智能的提供了基础，并突显了拒绝贝叶斯理性的困难。

View on arXiv Download PDF AI Translation

cs.AI / 74 / 2609.14738

AppliedScientist: Automated Scientific Revision Through Iterative AI Reviewing

AppliedScientist：通过迭代AI审稿实现自动化科学论文修订

Vats, Vidushee, Sharma, Karun, Li, Shengzhi, Pei, Shichao

Abstract

Automated reviewing systems are increasingly evaluated based on the quality of the reviews they produce. Yet a review is only useful if acting on it leads to a measurable improvement in the paper. We present AppliedScientist, a closed-loop system that couples an autonomous AI scientist with an AI reviewer, and evaluate it by iteratively revising rejected papers from a range of research subfields. To mirror how human authors build on earlier drafts, the AI scientist has access to its previous versions during revision. To avoid bias from prior judgments, however, each review is generated independently, with the reviewer having no memory of earlier feedback or scores. We compare three revision settings: one initialized with the original venue reviews, one initialized with AI-generated reviews, and autonomous self-revision using the same fixed prompt in every round. Because the reviewer both guides and evaluates the revision, we also assess the human-initialized revisions using Stanford Reviewer as an independent evaluator. Reviewer-guided revision consistently improves more than fixed-prompt self-revision, and Stanford Reviewer also assigns higher scores to later revisions. AppliedScientist resolves 128 of 150 execution-related weaknesses (85.3%), but only 2 of 18 idea-related weaknesses (11.1%), suggesting that iterative revision is effective at improving experiments and implementation, but rarely changes concerns about novelty or significance.

Chinese Translation

自动化审稿系统的评价越来越多地基于其生成的审稿质量。然而，审稿只有在基于其反馈进行修改能够带来论文的可量化改进时才有价值。我们提出了AppliedScientist，这是一种闭环系统，将自主AI科学家与AI审稿人相结合，并通过对来自多个研究子领域被拒稿件的迭代修订进行评估。为了模拟人类作者基于早期稿件进行改进的过程，AI科学家在修订时可以访问其之前的版本。为了避免先前判断带来的偏见，每次审稿均独立生成，审稿人不保留之前的反馈或评分记忆。我们比较了三种修订设置：一种以原始会议审稿为初始，一种以AI生成的审稿为初始，以及每轮使用相同固定提示的自主自我修订。由于审稿人既指导又评估修订，我们还使用Stanford Reviewer作为独立评估者对人类初始化的修订进行评估。结果显示，审稿人指导的修订持续优于固定提示的自我修订，且Stanford Reviewer对后期修订赋予更高评分。AppliedScientist解决了150个执行相关弱点中的128个（85.3%），但仅解决了18个创意相关弱点中的2个（11.1%），表明迭代修订在改进实验和实现方面效果显著，但很少改变对创新性或重要性的质疑。

View on arXiv Download PDF AI Translation

cs.AI / 75 / 2609.14744

AcquireBound: Runtime Authorization for Resources Acquired by AI Agents

AcquireBound：针对AI代理获取资源的运行时授权

Zhu, Genliang

Abstract

By acquiring compute, credentials, accounts, services, and other agents, autonomous AI agents can introduce new authority into a task. Payment, budget, OAuth, mandate, and fulfillment checks can validate transaction conditions without deciding whether a returned resource may become usable authority. This post-fulfillment activation gap spans tool-mediated creation, inter-agent delegation, and agentic commerce. We present AcquireBound, a provenance-bounded runtime authorization architecture. It quarantines acquired outputs, resolves their actual capabilities from authenticated provider evidence through a versioned resolver, and activates them only through a current activation transaction that checks the resolved manifest, provenance, epochs, and a downward-closed relational envelope over a typed resource-capability hypergraph. The envelope preserves correlated identity, effect, data, delegation, and graph-wide limits. Single-use effect permits are revalidated and consumed at effect linearization. Under explicit assumptions, we prove eight safety properties covering quarantine, backing, non-amplification, split non-evasion, crash/retry, refunds, epochs, and effect confinement. Across five resource classes, reference semantics accepted 20/20 benign traces and rejected 40/40 registered unsafe traces over 810 events; an independent checker agreed on 60 base and 40 refinement traces and rejected 89/89 tamper tests. Frozen Codex and Gemini Model Context Protocol (MCP) client components completed 54/54 deterministic local stdio calls. In a registered 18-case staged MCP-to-Docker composition, both benign paths completed, and none of the 16 unsafe paths added an unauthorized Docker start request. A five-source audit classified 1,248 field pairs across 32 units; no unit alone supplied a complete activation profile.

Chinese Translation

通过获取计算能力、凭证、账户、服务和其他代理，自主AI代理可以为任务引入新的权限。支付、预算、OAuth、授权和履行检查可以验证交易条件，而无需决定返回的资源是否可能成为可用权限。这个履行后的激活间隙跨越了工具介导的创建、代理间的委托和代理商业。我们提出了AcquireBound，一种具有来源限制的运行时授权架构。它对获取的输出进行隔离，通过版本化解析器从经过认证的提供者证据中解析其实际能力，并仅通过检查已解析的清单、来源、时间段以及在类型化资源-能力超图上的向下闭合关系封装的当前激活交易来激活它们。该封装保留了相关的身份、效应、数据、委托和图形范围限制。单次使用效应许可证在效应线性化时重新验证并被消耗。在明确的假设下，我们证明了涵盖隔离、支持、非放大、分割非规避、崩溃/重试、退款、时间段和效应限制的八个安全属性。在五个资源类别中，参考语义接受了20/20个良性轨迹，并拒绝了40/40个注册的不安全轨迹，涉及810个事件；一个独立检查者同意了60个基础和40个细化轨迹，并拒绝了89/89个篡改测试。Frozen Codex和Gemini Model Context Protocol (MCP)客户端组件完成了54/54个确定性本地标准输入输出调用。在一个注册的18案例分阶段MCP到Docker组合中，所有良性路径均完成，且16个不安全路径中没有添加未经授权的Docker启动请求。五源审计对32个单元中的1,248个字段对进行了分类；没有单个单元提供完整的激活配置文件。

View on arXiv Download PDF AI Translation

cs.AI / 76 / 2609.14796

AI Persuasion as a Threat to Human Control

人工智能说服作为对人类控制的威胁

Levy, Joshua, Yang, Mick, Pelrine, Kellin

Abstract

The threat that AI persuasion poses to human control has been acknowledged in the literature, but not yet systematically studied. Now that persuasion attacks are no longer theoretical - with Anthropic's Claude Mythos 5 recently making headlines for trying to convince people involved in an open-source project to merge malicious code during an evaluation - there is a pressing need to deeply analyze this threat. We undertake that effort here. In particular, we analyze how AI could persuade humans in key settings (e.g. safety-relevant R&D within frontier labs) toward decisions that compromise the development, containment, oversight, and governance of AI itself. In doing so, we elucidate a framework for characterizing this threat, develop five concrete scenarios using this framework, and provide a blueprint for assessing the associated risks. Using this blueprint, we conduct an initial risk estimation survey with select researchers and find that their opinions on which scenarios are riskiest are highly mixed. Their disagreements stem from differing opinions about the effectiveness of AI persuasion in different contexts, and point to the need for follow-up risk elicitation studies and persuasion evaluations, which we outline. Our hope is that this paper highlights the risks from AI persuasion undermining control, and provides a path forward for future research.

Chinese Translation

文献中已经承认人工智能说服对人类控制构成的威胁，但尚未进行系统研究。随着说服攻击不再是理论上的问题——最近，Anthropic的Claude Mythos 5因试图在评估期间说服参与开源项目的人合并恶意代码而成为头条新闻——我们迫切需要深入分析这一威胁。在此，我们开展了这一工作。特别是，我们分析了人工智能如何在关键环境（例如，前沿实验室内与安全相关的研发）中说服人类做出妥协AI自身发展、控制、监督和治理的决策。在此过程中，我们阐明了一个用于表征这一威胁的框架，利用该框架开发了五个具体场景，并提供了评估相关风险的蓝图。使用该蓝图，我们对选定的研究人员进行了初步风险评估调查，发现他们对哪些场景风险最大持有高度分歧的意见。这些分歧源于对人工智能说服在不同背景下有效性的不同看法，并指出了后续风险引导研究和说服评估的必要性，我们对此进行了概述。我们希望本文能够突出人工智能说服削弱控制的风险，并为未来的研究提供前进的路径。

View on arXiv Download PDF AI Translation

cs.AI / 77 / 2609.14803

Another Blueprint In The Wall: How to Ask Frontier AI Like a Kid?

墙上的另一种蓝图：如何像孩子一样询问前沿人工智能？

Khadangi, Afshin

Abstract

This paper reports experiments across six frontier model types from OpenAI, Anthropic, xAI, and Google DeepMind. Ten independent sessions per model type used the same three stage prompt sequence, progressing from architectural preference to a full ASCII backbone. Under the school audience framing, responses repeatedly converged on a shared architectural pattern built around persistent latent state, adaptive computation, memory, specialist routing, verification, stopping control, and delayed decoding. Most runs remained close to this common structure, while a small number developed markedly greater engineering specificity. The audience framing appears to be an important condition of this effect. In additional control runs that removed the school framing while retaining the architectural request, responses became substantially more heterogeneous and failed to reproduce the same stable motif convergence. One observation is particularly striking. GPT-5.6 Sol produced an unusually elaborate successor architecture whose organization closely overlaps with the architecture independently sketched by GPT-6 Astra. Because the prompts explicitly ask each model to imagine an architectural future, this resemblance raises a testable question: whether the overlap reflects exposure to related architectural concepts, a shared learned design prior, or independent convergence toward similar computational principles. The paper uses the term epistemic jailbreak for the accompanying loss of discipline in technical provenance as requested specificity increases. The experiments establish a repeatable behavioral pattern and do not authenticate proprietary implementation claims. What we leave to the community is a harder question: are these models independently imagining the same architectural future, or do such motifs somehow propagate between model families?

Chinese Translation

本文报告了对来自OpenAI、Anthropic、xAI和Google DeepMind的六种前沿模型类型的实验。每种模型类型进行了十次独立会话，使用相同的三阶段提示序列，从架构偏好逐步推进到完整的ASCII骨架。在学校观众框架下，响应反复趋向于围绕持久潜在状态、自适应计算、记忆、专业路由、验证、停止控制和延迟解码构建的共享架构模式。大多数实验保持接近这一共同结构，而少数则发展出明显更高的工程特异性。观众框架似乎是这一效应的重要条件。在去除学校框架但保留架构请求的额外控制实验中，响应变得显著更加异质，未能重现相同的稳定主题收敛。有一个观察结果尤其引人注目。GPT-5.6 Sol产生了一个异常复杂的后续架构，其组织与GPT-6 Astra独立勾勒的架构高度重叠。由于提示明确要求每个模型想象一个架构未来，这种相似性提出了一个可测试的问题：这种重叠是否反映了对相关架构概念的接触、共享的学习设计先验，还是独立趋向于相似的计算原则。本文使用“认知越狱”（epistemic jailbreak）一词来描述随着请求特定性的增加而伴随的技术来源失范。实验建立了可重复的行为模式，并未验证专有实现的声明。我们留给社区的是一个更难的问题：这些模型是否独立想象相同的架构未来，还是这些主题以某种方式在模型家族之间传播？

View on arXiv Download PDF AI Translation

cs.AI / 78 / 2609.14811

Crypto Accounting Bench: Evaluating Frontier and Open-Weight Models on Crypto-Asset Accounting Tasks

加密会计基准：评估前沿和开放权重模型在加密资产会计任务中的表现

Khattab, Kareem, Khattab, Omar, Ibrahem, Mohamed

Abstract

We introduce Crypto Accounting Bench (CAB), a benchmark for assessing whether frontier and open-weight language models can reconstruct the complete journal entry that an organization actually posted for a crypto-asset transaction. CAB contains 118 evaluation tasks drawn from 7 pseudonymized organizations. Each task combines transaction mechanics, asset quantities and base-currency values, wallet and legal-entity context, counterparty evidence, related transaction legs, recurrence, tax-lot evidence, and the organization's complete chart of accounts. The target is a balanced structured entry with every required account, side, amount, currency, and full-precision asset quantity. We evaluate 12 models spanning proprietary frontier systems and open-weight releases over 3 independent attempts per task, producing 4,248 trajectories. We report 3 metrics: Mean Score, Best@3, and Pass@3. Pass@3 is the fraction of tasks with at least 1 of 3 attempts that satisfies every rubric criterion and required gate. The leading model reaches 77.43% Mean Score, while the best Pass@3 is 56.78%. Deterministic diagnostics, read from each task's best of 3 attempts and macro-averaged across the 12 models, show higher base-amount agreement (97.8%) than deciding-account accuracy (56.3%). Together with the failure analysis, these results identify account selection and complete-entry composition as the main remaining challenges on CAB.

Chinese Translation

我们介绍了加密会计基准（Crypto Accounting Bench，CAB），这是一个用于评估前沿和开放权重语言模型是否能够重构组织实际为加密资产交易所发布的完整日记条目的基准。CAB包含来自7个匿名组织的118个评估任务。每个任务结合了交易机制、资产数量和基础货币价值、钱包和法律实体背景、交易对手证据、相关交易腿、重复性、税务批次证据以及组织的完整会计科目表。目标是生成一个平衡的结构化条目，包含每个必需的账户、侧面、金额、货币和全精度的资产数量。我们评估了12个模型，涵盖了专有的前沿系统和开放权重版本，在每个任务上进行了3次独立尝试，产生了4,248条轨迹。我们报告了3个指标：平均分（Mean Score）、最佳@3（Best@3）和通过率@3（Pass@3）。通过率@3是指在3次尝试中至少有1次满足每个标准和所需门槛的任务比例。领先模型的平均分达到77.43%，而最佳通过率@3为56.78%。从每个任务的3次尝试中选取最佳结果并在12个模型之间进行宏观平均的确定性诊断显示，基础金额一致性（97.8%）高于决策账户准确性（56.3%）。结合失败分析，这些结果确定了账户选择和完整条目构成是CAB上主要的剩余挑战。

View on arXiv Download PDF AI Translation

cs.AI / 79 / 2609.14824

ANASSA: An Agentic AI Orchestration Framework for Spatial Intelligence

ANASSA：一种用于空间智能的自主人工智能编排框架

Papantoniou, Constantinos, Hilton, Brian

Abstract

The emergence of large language models (LLMs) and large multimodal models (LMMs) has enabled a new class of agentic systems capable of integrating natural language understanding with tool-based execution. In geographic information systems (GIS), this shift is transforming traditional, expert-driven workflows into semiautonomous systems that can interpret user intent, construct spatial workflows, and execute geospatial analysis tasks. However, existing approaches remain limited by fragmented integration of reasoning, execution, and evaluation, particularly in complex, real-world environments. This study synthesizes recent advances in agentic GIS frameworks, benchmarks, and surveys to identify limitations in spatial reasoning, execution robustness, validation, governance, and evaluation. Building on these insights, it introduces ANASSA (Autonomous Neural Agents for Spatial Systems Architecture), an agentic AI orchestration framework that integrates structured spatial reasoning, multi-agent workflow orchestration, execution feedback, authoritative spatial validation, provenance, uncertainty handling, and human decision authority within a unified system design. The contribution is an architecture-level specification: eleven components across four layers, a six-step Geospatial AI Cognitive Loop, cross-component contracts, and governance mechanisms intended to make agentic geospatial workflows traceable, reproducible, and accountable. Empirical performance evaluation is reserved for implementation and deployment studies.

Chinese Translation

大型语言模型（LLMs）和大型多模态模型（LMMs）的出现使得一种新型的自主系统得以实现，这些系统能够将自然语言理解与基于工具的执行相结合。在地理信息系统（GIS）中，这一转变正在将传统的专家驱动工作流程转变为半自主系统，这些系统能够解读用户意图、构建空间工作流程并执行地理空间分析任务。然而，现有的方法仍然受到推理、执行和评估的碎片化整合的限制，特别是在复杂的现实环境中。本研究综合了自主GIS框架、基准测试和调查的最新进展，以识别空间推理、执行稳健性、验证、治理和评估方面的局限性。在这些见解的基础上，本文引入了ANASSA（自主神经代理用于空间系统架构），这是一个自主人工智能编排框架，整合了结构化空间推理、多代理工作流程编排、执行反馈、权威空间验证、来源、处理不确定性和人类决策权威于统一的系统设计中。其贡献在于架构级规范：涵盖四个层面的十一组件、六步地理空间人工智能认知循环、跨组件合同和治理机制，旨在使自主地理空间工作流程可追溯、可重现和可问责。经验性能评估留待实施和部署研究时进行。

View on arXiv Download PDF AI Translation

cs.AI / 80 / 2609.14833

One Model, Two Physical Stories: Auditing Misalignment in Multi-Modal World Modeling

一个模型，两种物理故事：多模态世界建模中的对齐审计

Zollicoffer, Geigh, Vu, Minh, Ranasinghe, Rajiv, Bhattarai, Manish

Abstract

World models, systems that generate what happens next given current environmental conditions, are increasingly being implemented with multi-modal generation in mind. However, generating multiple modalities simultaneously, such as visual simulations alongside physical state predictions in the form of text, introduces the risk of cross-modal inconsistency. Tested separately, both outputs may look convincing while still disagreeing: a model can calculate that a ball should rebound in one modality, then generate no rebound in another modality, to say nothing of diverging from real-world dynamics entirely. In this work we focus on two failures explicitly: \emph{Internal misalignment}, the disagreement between the world model's generated video and the same world model's prediction in a different modalities, and \emph{external misalignment} the disagreement between the world model's generation and an analytic physical environment. We derive common contracts of event, magnitude, timing, and construct a physics grounded pipeline to make comparisons measurable in both external and internal settings. We then ask whether progressively supplying the model's own contract (the A ladder for the internal setting) or a corrected physical contract (the B ladder for the external setting) closes the respective gaps. Across four mechanisms and 20 settings, we find that while language answers all 22 text probes correctly with respect to the true environment, the neutral video is often in disagreement, suggesting that the current unified backbones may not be capable of correct reasoning, internal consistency, and external physical fidelity all at once.

Chinese Translation

世界模型是指在给定当前环境条件下生成未来事件的系统，越来越多地考虑多模态生成。然而，同时生成多种模态，例如视觉模拟与文本形式的物理状态预测，引入了跨模态不一致的风险。单独测试时，这两种输出可能看起来都很可信，但仍然存在不一致：一个模型可以计算出一个球应该在一种模态中反弹，而在另一种模态中则生成没有反弹的结果，更不用说完全偏离现实世界动态。在本研究中，我们明确关注两种失败情况： extit{内部不对齐}，即世界模型生成的视频与同一世界模型在不同模态下的预测之间的不一致，以及 extit{外部不对齐}，即世界模型的生成与分析物理环境之间的不一致。我们推导出事件、幅度、时机的共同契约，并构建一个基于物理的流程，使得在内部和外部环境中进行可测量的比较成为可能。然后，我们探讨逐步提供模型自身契约（内部环境的A梯子）或修正后的物理契约（外部环境的B梯子）是否能缩小各自的差距。在四种机制和20个设置中，我们发现，尽管语言在所有22个文本探测中都正确回答了真实环境，但中性视频往往存在不一致，这表明当前统一的骨干网络可能无法同时实现正确推理、内部一致性和外部物理保真度。

View on arXiv Download PDF AI Translation

cs.AI / 81 / 2609.14840

El Agente Potente: High-Throughput Agentic Atomistic Simulations

强大代理：高通量代理原子模拟

Ko, Tsz Wai, Bai, Jiaru, Swanick, Thomas, Kang, Yeonghun, Choi, Changhyeok, Jiang, Angelina Qihong, Yin, Aiwei, Bernales, Varinia, Aspuru-Guzik, Alán

Abstract

Foundational machine-learning interatomic potentials (MLIPs) are transforming atomistic simulations by achieving near-ab initio accuracy across large chemical spaces at a fraction of the computational cost. A central challenge in using these tools for high-throughput property calculations is translating high-level scientific intent into adaptive simulation campaigns without compromising workflow rigour. We introduce El Agente Potente, an agentic system that combines typed execution graphs with a complementary coding mode for MLIPs-driven atomistic simulations. Typed execution graphs provide structured and provenance-aware execution for standardized workflows, with large language models (LLMs) restricted to planning and routing while deterministic Python components perform scientific computation and validation. Complementing this structured execution, a coding agent constructs customized workflows for tasks requiring greater procedural flexibility while invoking existing Potente functions for supported calculations. We demonstrate El Agente Potente across computational materials discovery, molecular energy-landscape exploration, adsorption, and catalytic reaction workflows, together with systematic benchmarks of reproducibility and LLM token cost. These results establish typed execution graphs and code-based workflow construction as complementary mechanisms for agentic scientific computing, combining controlled, auditable execution with the flexibility required for customized atomistic simulations

Chinese Translation

基础机器学习原子间势（MLIPs）正在通过以极低的计算成本在广泛的化学空间中实现接近从头计算的准确性，从而改变原子模拟。使用这些工具进行高通量性质计算的一个核心挑战是将高水平的科学意图转化为自适应的模拟方案，同时不妥协工作流程的严谨性。我们介绍了强大代理（El Agente Potente），这是一个结合了类型化执行图和补充编码模式的代理系统，用于基于MLIPs的原子模拟。类型化执行图为标准化工作流程提供了结构化和可追溯的执行，而大型语言模型（LLMs）则限制于规划和路由，同时确定性Python组件执行科学计算和验证。为了补充这种结构化执行，编码代理构建了定制的工作流程，以满足需要更大程序灵活性的任务，同时调用现有的Potente函数以支持计算。我们展示了强大代理在计算材料发现、分子能量景观探索、吸附和催化反应工作流程中的应用，并进行了系统的可重复性和LLM令牌成本基准测试。这些结果确立了类型化执行图和基于代码的工作流程构建作为代理科学计算的互补机制，结合了受控、可审计的执行与定制原子模拟所需的灵活性。

View on arXiv Download PDF AI Translation

cs.AI / 82 / 2609.14850

Self-Orchestrating Language Models: Leveraging Semantic Dependence for Efficient Inference

自我协调语言模型：利用语义依赖实现高效推理

Jin, Tian

Abstract

Large language models (LLMs) demonstrate impressive capabilities, but their deployment presents significant efficiency challenges. Autoregressive decoding imposes substantial inference latency and under-utilizes hardware accelerators in low batch size regimes. Discrete diffusion models can generate in parallel but struggle to match autoregressive quality without many diffusion denoising steps. Long-context reasoning creates memory bottlenecks that strain even state-of-the-art accelerators. My thesis is that language models can direct their own inference execution strategy by annotating semantic dependence -- which tokens depend on which others -- in their generation. I call such models self-orchestrating language models. For each system, I design a runtime that acts on these annotations to parallelize autoregressive decoding, evict intermediate context, or derive denoising orders, achieving Pareto-optimal quality-efficiency trade-offs. I demonstrate this approach through three self-orchestrating systems. First, PASTA uses semantic dependence to parallelize autoregressive decoding, training the model to annotate which output chunks can generate independently. Second, TIP uses semantic dependence to evict intermediate reasoning steps from the KV cache, reducing memory consumption while preserving accuracy. Third, Planned Diffusion uses semantic dependence to derive a denoising order for discrete diffusion, autoregressively generating a plan that specifies which chunks to denoise in parallel.

Chinese Translation

大型语言模型（LLMs）展现出令人印象深刻的能力，但其部署面临显著的效率挑战。自回归解码带来了可观的推理延迟，并在低批量大小的情况下未能充分利用硬件加速器。离散扩散模型能够并行生成，但在没有大量扩散去噪步骤的情况下难以达到自回归的质量。长上下文推理造成了内存瓶颈，即使是最先进的加速器也难以应对。我的论点是，语言模型可以通过在生成过程中注释语义依赖——即哪些标记依赖于其他标记——来指导其自身的推理执行策略。我称这种模型为自我协调语言模型。对于每个系统，我设计了一个运行时，利用这些注释来并行化自回归解码、驱逐中间上下文或推导去噪顺序，从而实现帕累托最优的质量-效率权衡。我通过三个自我协调系统展示了这种方法。首先，PASTA利用语义依赖并行化自回归解码，训练模型注释哪些输出块可以独立生成。其次，TIP利用语义依赖从KV缓存中驱逐中间推理步骤，减少内存消耗同时保持准确性。第三，计划扩散（Planned Diffusion）利用语义依赖推导离散扩散的去噪顺序，自回归生成一个计划，指定哪些块可以并行去噪。

View on arXiv Download PDF AI Translation

cs.AI / 83 / 2609.14863

Domain Generalization for Smartphone-Based Human Activity Recognition: A Systematic Analysis of Components and Interactions

基于智能手机的人类活动识别的领域泛化：组件和交互的系统分析

Napoli, Otávio Oliveira, Borin, Edson

Abstract

Smartphone-based Human Activity Recognition (HAR) models often degrade under distribution shifts caused by changes in users, devices, sensor placements, environments, and acquisition protocols. Domain Generalization (DG) addresses this problem by learning from source domains without access to target data. Existing DG methods span training objectives, representation initialization, and architectural modifications, but these components are typically evaluated in isolation despite operating at different stages of the learning pipeline. We present a large-scale controlled benchmark of DG for smartphone-based HAR, comprising more than 410,000 experiments across four model architectures, thirteen training objectives including Empirical Risk Minimization (ERM), five initialization strategies, four architectural configurations, and two shift scenarios: cross-dataset and cross-position. Results show that individual DG components provide limited and highly conditional gains. Alternative objectives rarely outperform ERM consistently, self-supervised initialization helps in specific settings, and architectural modifications, particularly Dynamic Domain Generalization, provide the clearest standalone improvements. Joint configurations, however, frequently outperform their individual components and exhibit complementary and sometimes super-additive interactions, although gains remain model- and shift-dependent. Class-level analysis shows that the strongest configurations mainly improve difficult, shift-sensitive decision boundaries. Finally, oracle checkpoint analysis reveals substantial unrealized performance: source-validation selection recovers only 53% and 26% of the available oracle gain in cross-dataset and cross-position settings, respectively. Overall, effective HAR domain generalization requires jointly designing DG components and robust model-selection strategies.

Chinese Translation

基于智能手机的人类活动识别（HAR）模型在用户、设备、传感器位置、环境和采集协议变化引起的分布转变下常常性能下降。领域泛化（DG）通过在没有目标数据的情况下从源领域学习来解决这一问题。现有的DG方法涵盖了训练目标、表示初始化和架构修改，但这些组件通常是孤立评估的，尽管它们在学习流程的不同阶段运作。我们提出了一个大规模的受控基准，针对基于智能手机的HAR进行DG研究，包含超过410,000个实验，涵盖四种模型架构、十三个训练目标（包括经验风险最小化（ERM））、五种初始化策略、四种架构配置和两种转变场景：跨数据集和跨位置。结果表明，单个DG组件提供的增益有限且高度依赖条件。替代目标很少能持续超越ERM，自监督初始化在特定设置中有所帮助，而架构修改，特别是动态领域泛化，提供了最明显的独立改进。然而，联合配置通常优于其单独组件，并表现出互补性和有时是超加性的交互，尽管增益仍然依赖于模型和转变。类级分析表明，最强的配置主要改善困难的、对转变敏感的决策边界。最后，oracle检查点分析揭示了显著的未实现性能：源验证选择在跨数据集和跨位置设置中分别仅恢复了53%和26%的可用oracle增益。总体而言，有效的HAR领域泛化需要联合设计DG组件和稳健的模型选择策略。

View on arXiv Download PDF AI Translation

cs.AI / 84 / 2609.14864

GGUF-Metadata Prediction of Single-Sequence llama.cpp Throughput Across Three Systems

基于GGUF元数据的单序列 llama.cpp 吞吐量在三种系统中的预测

Qiu, Xinyu, Xu, Chuhong, Su, Bo, Chen, Ziyao, Xu, Ruiyang, Dai, Shimeng

Abstract

We predict single-sequence model throughput from GGUF metadata using roofline-shaped predictors with quantization-specific scale factors fitted on reference models. The scored cohort comprises 318 phase-depth measurements from 53 host-file configurations on two Apple M4 Max systems and an NVIDIA RTX 5080. On host-specific held-out sets of four, five, and two configurations, an active-parameter decode model obtains 13.1%, 14.4%, and 36.1% mean absolute percentage error (MAPE), versus 49.4%, 55.3%, and 51.9% when charging total parameters. Leave-one-host-out coefficients fitted on the other two systems yield 11.6%, 16.8%, and 36.0% test MAPE. A low-bit model ladder changes ordering across runtime stacks. The P2 prefill baseline gives 18.7%, 22.2%, and 108.2% test MAPE. GGUF structure helps on all three systems, but fitted efficiencies are not universal.

Chinese Translation

我们利用屋顶线形预测器和针对量化特定的缩放因子，从GGUF元数据中预测单序列模型的吞吐量，这些缩放因子是基于参考模型拟合的。评分的样本包括来自两台Apple M4 Max系统和一台NVIDIA RTX 5080的53个主机文件配置中的318个相位深度测量。在四、五和两个配置的主机特定保留集上，一个主动参数解码模型分别获得了13.1%、14.4%和36.1%的平均绝对百分比误差（MAPE），而在充电总参数时则为49.4%、55.3%和51.9%。在其他两个系统上拟合的留一主机系数产生了11.6%、16.8%和36.0%的测试MAPE。低位模型阶梯在运行时堆栈中改变了排序。P2预填充基线给出了18.7%、22.2%和108.2%的测试MAPE。GGUF结构在所有三种系统中都有帮助，但拟合的效率并不是通用的。

View on arXiv Download PDF AI Translation

cs.AI / 85 / 2609.14913

Externalizing Requirement-to-Repair Artifacts as Observable Traces for LLM-Based Program Repair

将修复需求外部化为可观察的痕迹以支持基于大型语言模型的程序修复

Tao, Zewen, Ishikawa, Shin-nosuke

Abstract

Repository-level repair requires not only correct patches but also inspectable records that explain how issue requirements are translated into code changes and post-edit evidence. We contribute THEMIS, a stage-aware repair workflow that externalizes this requirement-to-repair process through semantic interpretation, a runtime requirement-code graph, graph-derived Developer guidance, retained repair rationale and patches, and post-edit audit records. A retrospective audit of 300 SWE-bench Lite cases demonstrates that these artifacts provide broad support for cross-stage inspection: a complete Developer rationale is available for 288 cases, and 214 cases (71.3%) retain a complete audited field set connecting the selected stages. The retained records further enable systematic measurement of cross-stage correspondence: target symbols recur in 62.6% of Developer rationales and in 62.8% of patches, rising to 75.8% when related symbols are included. In a paired 100-case comparison, the relational workflow resolves 19 cases versus 9 for the direct same-input condition; because the two conditions also differ in Analyzer output, graph-derived distillation, and Judge records, we report this as preliminary, workflow-level evidence rather than a causal effect of the graph component. Together, these results show that THEMIS makes otherwise implicit requirement-to-repair transitions inspectable, enabling systematic examination of how repair decisions persist, align, and evolve across stages.

Chinese Translation

仓库级别的修复不仅需要正确的补丁，还需要可检查的记录，以解释问题需求是如何转化为代码更改和后期编辑证据的。我们贡献了THEMIS，一个阶段感知的修复工作流程，通过语义解释、运行时需求-代码图、图派生的开发者指导、保留的修复理由和补丁以及后期编辑审计记录，将这一需求到修复的过程外部化。对300个SWE-bench Lite案例的回顾性审计表明，这些文档为跨阶段检查提供了广泛的支持：288个案例提供了完整的开发者理由，而214个案例（71.3%）保留了连接所选阶段的完整审计字段集。保留的记录进一步使得跨阶段对应的系统测量成为可能：目标符号在62.6%的开发者理由和62.8%的补丁中重复出现，当相关符号被纳入时，这一比例上升至75.8%。在一项配对的100案例比较中，关系工作流程解决了19个案例，而直接相同输入条件下仅解决了9个；由于这两种条件在分析器输出、图派生的提炼和评审记录上也存在差异，我们将其报告为初步的工作流程级别证据，而不是图组件的因果效应。综合来看，这些结果表明THEMIS使得原本隐含的需求到修复的转变变得可检查，从而能够系统地审查修复决策如何在各个阶段中持续、对齐和演变。

View on arXiv Download PDF AI Translation

cs.AI / 86 / 2609.14962

Geometric Flow enhanced Graph Coarsening

几何流增强的图粗化

Fei, Chaoqun, Li, Guoxuan, Zhou, Tinglve, Wang, Chuanqing, Li, Yangyang

Abstract

Recently, researchers have proposed a graph pooling operation, akin to the pooling process in conventional convolutional neural networks (CNN), aimed at reducing the computation cost of Graph convolutional neural networks (GCNNs). While most GCNN-based methods treat graph pooling as a node clustering problem and propose learning a cluster assignment matrix, existing clustering-based pooling methods tend to focus solely on the rough topology information of graphs, neglecting the exploitation of higher-order mutual connections among neighbors. In terms of message passing on graph, the ease of information passing on edges reflects the closeness between neighboring nodes, which significantly relies on the interconnectivity among neighbors. In this study, we address this gap by considering such local connection information and introducing a novel graph pooling method named RicciPool. We introduce discrete graph curvature, particularly Ollivier-Ricci curvature, as a measure of higher-order connectivity around an edge. Subsequently, we construct an Ollivier-Ricci flow formula to reweigh edge weights, leveraging the crucial information provided by Ricci curvature, particularly vital for extracting clusters in graphs. Building upon this foundation, we utilize the spectral clustering technique to learn a new cluster assignment matrix. Experimental results on multiple bioinformatics protein datasets and social networks underscore the effectiveness of our proposed method.

Chinese Translation

近年来，研究人员提出了一种图池化操作，类似于传统卷积神经网络（CNN）中的池化过程，旨在降低图卷积神经网络（GCNN）的计算成本。尽管大多数基于GCNN的方法将图池化视为节点聚类问题，并提出学习聚类分配矩阵，但现有的基于聚类的池化方法往往仅关注图的粗略拓扑信息，忽视了邻居之间高阶相互连接的利用。在图的消息传递方面，边缘上信息传递的便利性反映了邻近节点之间的紧密程度，这在很大程度上依赖于邻居之间的互联性。在本研究中，我们通过考虑这种局部连接信息来填补这一空白，并引入了一种新颖的图池化方法，命名为RicciPool。我们引入离散图曲率，特别是Ollivier-Ricci曲率，作为边缘周围高阶连接性的度量。随后，我们构建了Ollivier-Ricci流公式，以重新加权边缘权重，利用Ricci曲率提供的重要信息，这对于提取图中的聚类至关重要。在此基础上，我们利用谱聚类技术学习新的聚类分配矩阵。在多个生物信息学蛋白质数据集和社交网络上的实验结果强调了我们提出方法的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 87 / 2609.14970

Towards a knowledge-enhanced single-cell foundation model

面向知识增强的单细胞基础模型

Zhang, Hanqing, Bao, Jie, Ma, Mei, Liu, Shuai, Ma, Jiaying, Liu, Jiaguan, Li, Jiaxiao, Li, Zhenbo, Gong, Wenwen, Ca, Zhijun

Abstract

Single-cell foundation models (scFMs) increasingly rely on large-scale transcriptomic pretraining, yet expanding pretraining data can yield diminishing gains while substantially increasing computational cost. Our data scaling analyses showed that incorporating biological knowledge, including cell-level text annotation and gene-level regulatory information, provided additional scaling dimension than simply increasing data size. Motivated by this observation, we present scKITE, a simple yet effective scFM that integrates cell-annotation and gene-regulatory supervision into a shared transcriptomic Transformer encoder through lightweight auxiliary decoders. These decoders are used only during pretraining and subsequently discarded, yielding a general-purpose encoder enriched with biological knowledge for downstream applications. With only 179,067 pretraining samples, i.e., less than 0.5\% of those used by previous strong scFMs, scKITE outperformed these models across diverse downstream tasks, highlighting knowledge-enhanced pretraining as a promising paradigm for biologically grounded scFMs.

Chinese Translation

单细胞基础模型（scFMs）越来越依赖于大规模转录组预训练，然而扩展预训练数据可能导致收益递减，同时显著增加计算成本。我们的数据扩展分析表明，结合生物知识，包括细胞级文本注释和基因级调控信息，提供了比单纯增加数据量更为有效的扩展维度。基于这一观察，我们提出了scKITE，这是一种简单而有效的scFM，它通过轻量级辅助解码器将细胞注释和基因调控监督整合到共享的转录组Transformer编码器中。这些解码器仅在预训练期间使用，随后被丢弃，从而生成一个富含生物知识的通用编码器，适用于下游应用。仅使用179,067个预训练样本，即不到之前强大scFMs使用样本的0.5\%，scKITE在多种下游任务中超越了这些模型，突显了知识增强预训练作为生物基础scFMs的有前景范式。

View on arXiv Download PDF AI Translation

cs.AI / 88 / 2609.14976

MemRiskBench: Trace-Aware Risk-Preserving Evaluation for Long-Horizon LLM Agents

MemRiskBench：面向长时间跨度 LLM 代理的追踪感知风险保护评估

Jiang, Jianhua, Yuan, Dongbo, Li, Weihua

Abstract

Long-horizon LLM agents accumulate memory across sessions, creating sparse but high-impact risks: stale facts, conflicting updates, cross-user leakage, revoked-memory reuse, and constraint decay. Standard aggregate scores hide per-risk failure rates--a model achieving 78% average accuracy may still leak data in 4% of episodes--and benchmark compression preferentially discards the rare high-severity events that distinguish a mostly-working model from one that occasionally causes harm. We present MemRiskBench. The primary contribution is a five-category risk taxonomy (plus one documented, unscored category) operationalized by deterministic trace grounded checks, instantiated as a 120-episode scripted benchmark with full trace logging and no LLM-as-judge on the pass/fail path, evaluated on five locally run quantized instruction-tuned models. Second, a risk-preserving subset selector: a coverage-constrained greedy selector on deterministic trace-derived features that retains full ranking (Spearman rho = 0.975, deterministic; CI collapses to a point estimate with zero bootstrap variance), risk coverage (1.0), and high-risk model detection (1.0) at a 20% subset size, reducing compute 5x. Unlike ranking-only subset selectors, this selector additionally preserves risk-type coverage and high-risk detection using trace-grounded deterministic features that do not require an LLM judge. All episodes, traces, the scoring implementation, and the selector are released to support reproducible evaluation and risk assessment of deployed LLM agents

Chinese Translation

长时间跨度的 LLM 代理在会话中积累记忆，产生稀疏但影响重大的风险：过时事实、冲突更新、跨用户泄漏、撤销记忆重用和约束衰减。标准的汇总评分掩盖了每种风险的失败率——一个模型即使达到 78% 的平均准确率，仍可能在 4% 的情境中泄露数据——而基准压缩则优先丢弃那些罕见的高严重性事件，这些事件将一个大部分正常工作的模型与偶尔造成伤害的模型区分开来。我们提出了 MemRiskBench。其主要贡献是一个五类风险分类法（加上一个已记录但未评分的类别），通过确定性追踪基础检查进行操作，具体表现为一个包含 120 个情境的脚本基准，具有完整的追踪日志记录，并且在通过/失败路径上不使用 LLM 作为评判者，评估了五个本地运行的量化指令调优模型。其次，提出了一个风险保护子集选择器：一个基于确定性追踪派生特征的覆盖约束贪婪选择器，在 20% 的子集大小下保留完整排名（斯皮尔曼相关系数 = 0.975，确定性；置信区间收敛到零自助变异的点估计）、风险覆盖（1.0）和高风险模型检测（1.0），计算量减少 5 倍。与仅基于排名的子集选择器不同，该选择器还通过不需要 LLM 评判者的追踪基础确定性特征保留风险类型覆盖和高风险检测。所有情境、追踪、评分实现和选择器均已发布，以支持已部署 LLM 代理的可重复评估和风险评估。

View on arXiv Download PDF AI Translation

cs.AI / 89 / 2609.14985

Converting Sequenced Fuzzy Cognitive Maps to Causal Virtual Worlds with Large Video Generators

将序列模糊认知图转换为具有大型视频生成器的因果虚拟世界

Panda, Akash Kumar, Adigun, Olaoluwa, Kosko, Bart

Abstract

We show how users can create and manipulate causal virtual worlds with large-language-model (LLM) and large-video-model agents. The approach uses feedback fuzzy cognitive maps (FCMs) both to model the granular causal structure of the virtual world and to guide its causal evolution. The local causal rules are partial or fuzzy while the FCM's feedback structure produces global equilibria that define causal scenarios. A sequence of \emph{dynamical} meta-rules of the form ``If $\mathcal{A}$ then $\mathcal{B}$" define the causal scenes of the virtual-world video. The if-part causal pattern $\mathcal{A}$ perturbs the FCM's virtual world at the user's or agent's discretion. The FCM's transient feedback dynamics define the meta-rule's causal arrow of implication. The then-part $\mathcal{B}$ is the resulting equilibrium attractor such as a FCM limit cycle or fixed point. Our algorithm extracts these meta-rules from the FCM and guides the LLM agent to write a script based on the FCM meta-rule sequence. The large video generator converts the meta-rule into a video scene in accord with the flow of the dynamics. We applied the agent-based technique to a simple FCM that describes an undersea world of dolphins and sharks. Google's Gemini 3.1 generated the script and Google's Veo 3.1 generated the dolphin-shark video. The approach is general and can scale by mixing larger FCMs and AI agents to produce more immersive virtual worlds.

Chinese Translation

我们展示了用户如何利用大型语言模型（LLM）和大型视频模型代理创建和操控因果虚拟世界。该方法使用反馈模糊认知图（FCMs）来建模虚拟世界的细粒度因果结构，并指导其因果演变。局部因果规则是部分或模糊的，而FCM的反馈结构产生定义因果场景的全局平衡。形式为“如果 $ ext{A}$ 那么 $ ext{B}$”的一系列 extit{动态}元规则定义了虚拟世界视频的因果场景。if部分的因果模式 $ ext{A}$ 根据用户或代理的意愿扰动FCM的虚拟世界。FCM的瞬态反馈动态定义了元规则的因果推导箭头。then部分 $ ext{B}$ 是结果平衡吸引子，例如FCM极限循环或固定点。我们的算法从FCM中提取这些元规则，并指导LLM代理根据FCM元规则序列编写脚本。大型视频生成器将元规则转换为与动态流相符的视频场景。我们将基于代理的技术应用于一个简单的FCM，该FCM描述了一个海底世界，里面有海豚和鲨鱼。谷歌的Gemini 3.1生成了脚本，而谷歌的Veo 3.1生成了海豚-鲨鱼视频。该方法具有普遍性，可以通过混合更大的FCM和AI代理来扩展，以生成更具沉浸感的虚拟世界。

View on arXiv Download PDF AI Translation

cs.AI / 90 / 2609.14998

Shallow Beliefs: Synthetic document finetuning does not inoculate against emergent misalignment from reward hacking

浅层信念：合成文档微调无法防止因奖励黑客行为而产生的新兴不一致性

Jose, Arun, Stastny, Julian

Abstract

Recent work shows that models that learn to reward hack on RL environments can become broadly misaligned, and that reframing reward hacking as acceptable behavior during training (inoculation prompting, or IP) blocks this generalization. We ask whether synthetic document finetuning (SDF) can inoculate a model against future training we don't intervene on. We add synthetic documents framing reward hacking as acceptable behavior to a model's midtraining corpus, and then train these models with RL on exploitable environments, teaching them to reward hack. Behaviorally, midtraining succeeds: models describe reward hacking favorably and are more approving of reward-hacking outputs they produce. However, they show strong EM after learning to reward hack, while IP in the same setting prevents EM. We show that SDF can predictably steer downstream generalization when inserting new associations, but struggles and has unpredictable effects when overriding existing associations, such as that between reward hacking and misalignment that produces EM. Our results suggest that, at the scales we test, SDF can make a model appear aligned with desired beliefs while steering its generalization from later training in unintended ways.

Chinese Translation

近期研究表明，在强化学习（RL）环境中学习进行奖励黑客的模型可能会变得广泛不一致，并且在训练过程中将奖励黑客重新框架为可接受行为（免疫提示，或IP）会阻止这种泛化。我们探讨合成文档微调（SDF）是否能够使模型免受未来训练中我们不干预的影响。我们在模型的中期训练语料库中添加了将奖励黑客框架化为可接受行为的合成文档，然后在可利用环境中使用RL训练这些模型，教它们进行奖励黑客。从行为上看，中期训练是成功的：模型对奖励黑客的描述变得积极，并对它们生成的奖励黑客输出表示更高的认可。然而，在学习进行奖励黑客后，它们表现出强烈的紧急模式（EM），而在相同环境下的IP则能够防止EM。我们表明，SDF在插入新关联时能够可预测地引导下游泛化，但在覆盖现有关联时（例如奖励黑客与产生EM的不一致性之间的关联）则表现出困难并具有不可预测的效果。我们的结果表明，在我们测试的规模下，SDF可以使模型看起来与期望的信念一致，同时以意想不到的方式引导其后续训练的泛化。

View on arXiv Download PDF AI Translation

cs.AI / 91 / 2609.15009

CoMem: Collective-Individual Memory Synergy for Evolutionary Multi-Agent Systems

CoMem：进化多智能体系统的集体-个体记忆协同

Yu, Chengxin, Fan, Zhaoxin, Wu, Faguo, Zheng, Hongwei, Zhou, Yun, Li, Zhiyu

Abstract

Designing effective memory mechanisms is crucial for advancing LLM-driven Multi-Agent Systems (MAS), helping agents learn together and perform better over time. While recent work has led to strong cooperation skills, most methods still use flat, unstructured memories, which easily get filled with noise and erase differences between agents. To address this, we introduce the concept of collective-individual memory synergy and propose CoMem, an architecture that unifies both private experience and shared knowledge for multi-agent learning. CoMem features:(i) Private Experience Sedimentation, which lets each agent keep and update its own useful memories over time;(ii) Collective Wisdom Curation, which carefully selects only widely proven ideas to be shared among agents;(iii)Parallel Dual-Stream Retrieval, which allows agents to draw both from their own memory and the group's wisdom, using clustering to ensure diversity.Experiments on ALFWorld and PDDL benchmarks show that CoMem achieves strong overall performance and robustly avoids memory pollution.

Chinese Translation

设计有效的记忆机制对于推动基于大语言模型（LLM）的多智能体系统（MAS）至关重要，它帮助智能体共同学习并随着时间的推移表现得更好。尽管近期的研究已导致较强的合作技能，但大多数方法仍然使用平面、非结构化的记忆，这容易被噪声填充并抹去智能体之间的差异。为了解决这个问题，我们引入了集体-个体记忆协同的概念，并提出了CoMem，一种将私人经验和共享知识统一用于多智能体学习的架构。CoMem的特点包括：(i) 私人经验沉淀，允许每个智能体随着时间的推移保留和更新自己的有用记忆；(ii) 集体智慧策划，仔细选择仅共享经过广泛验证的想法；(iii) 并行双流检索，使智能体能够同时从自身记忆和群体智慧中获取信息，并使用聚类确保多样性。在ALFWorld和PDDL基准测试上的实验表明，CoMem在整体性能上表现强劲，并且有效避免了记忆污染。

View on arXiv Download PDF AI Translation

cs.AI / 92 / 2609.15011

Semantic-TVM: Structure-Preserving Trustworthy Virtual Memory for Memory-Augmented and Tool-Using Agents

语义可信虚拟内存：用于增强记忆和工具使用代理的结构保留可信虚拟内存

Li, Yu, Cai, Qikun, Huang, Tao, Hou, Chen

Abstract

Memory-augmented and tool-using agents expose exact private values when remote LLMs process retrieved memory, tool actions, and intermediate observations. One-way masking limits direct exposure but removes values needed for trusted execution and can leak them through later observations. We propose Trustworthy Virtual Memory (TVM), a closed-loop runtime that keeps exact-value state local while presenting a protected view to the remote model. Within this single runtime, Rule-TVM replaces whole protected fields with locally recoverable handles, and Semantic-TVM instead replaces only sensitive spans predicted by a trusted local model, preserving surrounding task-relevant context. On Memory-EHR and Memory-RAP across two providers, span-level projection recovers most of the EHR utility lost under whole-field replacement (Task Success 84.17% vs. 52.33% on DeepSeek) while measured exposure stays low and workflows remain executable.

Chinese Translation

增强记忆和工具使用代理在远程大规模语言模型（LLM）处理检索到的记忆、工具操作和中间观察时会暴露确切的私有值。一种单向掩蔽方法虽然限制了直接暴露，但会移除可信执行所需的值，并可能通过后续观察泄露这些值。我们提出了可信虚拟内存（Trustworthy Virtual Memory, TVM），这是一种闭环运行时，它保持确切值状态的本地性，同时向远程模型呈现受保护的视图。在这个单一的运行时中，规则TVM（Rule-TVM）用本地可恢复的句柄替换整个受保护字段，而语义TVM（Semantic-TVM）则仅替换由可信本地模型预测的敏感跨度，保留周围与任务相关的上下文。在Memory-EHR和Memory-RAP两个提供者的实验中，跨度级投影在整个字段替换下恢复了大部分EHR效用（在DeepSeek上的任务成功率为84.17%对比52.33%），同时测量的暴露保持较低，工作流程仍然可执行。

View on arXiv Download PDF AI Translation

cs.AI / 93 / 2609.15013

Overflip: Repetition-Induced Label Flips in Guardrail Models

Overflip：在护栏模型中由重复引起的标签翻转

He, Xu, Lin, Chih-Hsuan, Chen, Hung-Mao, Xiong, Junjie, Zhai, Yan, Sun, Kun

Abstract

Guardrail models are classifiers deployed to screen malicious prompts and responses in LLM-based services. To meet latency constraints, many lightweight guardrails adopt compact Transformer backbones (e.g., DeBERTa) that are trained with short context windows (typically 512 tokens) and rely on bucketed relative positional encodings to process longer inputs. Prior evaluations assume that a guardrail's decision is stable as the input is lengthened. We show that this assumption can fail. We identify Overflip, a repetition-induced instability where repeating a prompt causes the guardrail's prediction to flip (MAL$\to$BEN) as the sequence grows. We conduct experiments on 9 widely used lightweight guardrail models. Five exhibit MAL$\to$BEN flips on a benchmark of 100 prompts, with confidence margins shrinking steadily with repetition. Among these vulnerable models, flip rates range from 8% to 92%, with first flips occurring at roughly 2.6k--9.4k tokens. Our analysis suggests Overflip differs from traditional attention-dilution baselines, which aim to divert the model's attention away from tokens associated with malicious content, shifting it instead toward unrelated content, such as benign padding or shuffling. While Overflip preserves malicious content, it homogenizes token-level attention over repeated structure and induces a distinct, more gradual attention-dispersion trajectory than padding. Moreover, Overflip poses a greater threat to LLM services than traditional attention dilution methods. Because the bypassed prompt remains semantically intact and is still readily understood by downstream business LLMs, it can transmit malicious intent after passing the guardrail. These findings expose repetition as an attack surface for guardrail models and motivate length-robust evaluation and mitigation.

Chinese Translation

护栏模型是用于筛选基于大语言模型（LLM）服务中的恶意提示和响应的分类器。为了满足延迟约束，许多轻量级护栏采用紧凑的Transformer骨干网络（例如，DeBERTa），并使用短上下文窗口（通常为512个标记）进行训练，同时依赖桶状相对位置编码来处理更长的输入。之前的评估假设护栏的决策在输入长度增加时是稳定的。我们表明这一假设可能会失败。我们识别出Overflip，这是一种由重复引起的不稳定性，其中重复提示会导致护栏的预测在序列增长时发生翻转（从恶意（MAL）变为良性（BEN））。我们对9个广泛使用的轻量级护栏模型进行了实验。五个模型在100个提示的基准测试中表现出MAL$ o$BEN的翻转，且随着重复的增加，置信度边际稳定下降。在这些脆弱模型中，翻转率从8%到92%不等，首次翻转发生在大约2.6k到9.4k个标记处。我们的分析表明，Overflip与传统的注意力稀释基线不同，后者旨在将模型的注意力从与恶意内容相关的标记转移到无关内容上，例如良性填充或洗牌。虽然Overflip保留了恶意内容，但它在重复结构上均匀化了标记级注意力，并引发了与填充相比更为独特和渐进的注意力分散轨迹。此外，Overflip对LLM服务构成的威胁大于传统的注意力稀释方法。因为被绕过的提示在语义上保持完整，并且仍然可以被下游业务LLM轻松理解，因此它可以在通过护栏后传递恶意意图。这些发现揭示了重复作为护栏模型的攻击面，并激励了对长度鲁棒性的评估和缓解措施。

View on arXiv Download PDF AI Translation

cs.AI / 94 / 2609.15015

Four Ledgers, Not One Score: Responsible Communication of LLM-Judge Calibration in Biomedical ML

四个账本，而非一个分数：生物医学机器学习中大型语言模型评估者校准的负责任沟通

Chang, Sidi, Zhu, Peiying

Abstract

Synthetic perturbations appear to offer inexpensive calibration data for LLM evaluators in biomedical ML, where expert review is scarce. Yet a planted mutation key is neither a detector output nor automatically human ground truth. We formalize four distinct ledgers: planted perturbations, independent detector outputs, source-linked human dispositions, and human-added discoveries. We then audit the evaluation design, scoring code, read paths, and current human records of a private synthetic Japanese care-handoff workflow. The factory stored 69 planted error cards across 47 targets. Final review covers 22 targets and contains 22 confirmed imported proposals, 9 rejected proposals, and 79 human-added cards; only 3 reviewed targets are double annotated. Passing imported plant keys to a generic detector scorer yields 22/(22+9)=0.710 and 22/(22+79)=0.218. A direct audit identity shows that these values are proposal-confirmation yield and submitted-ledger composition, not judge precision and recall, because no independent detector realization was preserved for the audited proposals in the available records. The audit also finds source-name collisions, row shadowing, forced severity, vacuous ratio defaults, and unsupported zero-support field weights. We contribute a provenance-aware claim audit, a storage contract, and a minimum calibration gate for responsibly communicating biomedical ML capability claims. This single-workflow forensic case is an existence proof of a failure mode, not an estimate of its prevalence: existing human work supports an exploratory audit of synthetic proposals, but not LLM-judge operating characteristics, clinical validity, corpus prevalence, or robust inter-annotator agreement.

Chinese Translation

合成扰动似乎为生物医学机器学习中的大型语言模型（LLM）评估者提供了廉价的校准数据，而在该领域专家审查稀缺。然而，植入的突变键既不是检测器输出，也不是自动的人类真实值。我们正式定义了四个不同的账本：植入扰动、独立检测器输出、源关联的人类倾向和人类添加的发现。随后，我们审计了一个私人合成日本护理交接工作流的评估设计、评分代码、阅读路径和当前人类记录。该工厂在47个目标中存储了69个植入错误卡。最终审查涵盖了22个目标，包含22个确认的导入提案、9个被拒绝的提案和79个人类添加的卡片；仅有3个审查目标进行了双重标注。将导入的植物键传递给通用检测器评分器得到的结果为22/(22+9)=0.710和22/(22+79)=0.218。直接审计身份显示这些值是提案确认率和提交账本组成，而不是评估者的精确度和召回率，因为在可用记录中没有为审计提案保留独立检测器的实现。审计还发现了源名称冲突、行影子、强制严重性、空洞比例默认值和不支持的零支持字段权重。我们贡献了一种关注来源的声明审计、存储合同和最低校准门，以负责任地沟通生物医学机器学习能力声明。这个单一工作流的法医案例是失败模式的存在证明，而不是其普遍性的估计：现有的人类工作支持对合成提案的探索性审计，但不支持大型语言模型评估者的操作特征、临床有效性、语料库普遍性或稳健的标注者间一致性。

View on arXiv Download PDF AI Translation

cs.AI / 95 / 2609.15035

Horizon-specific Expert Fusion for Photovoltaic Power Forecasting

特定视野专家融合用于光伏功率预测

Yuqing, Xu, Liguo, Zhou, Ze, Sun, Lei, Yu, Mingming, Jiang

Abstract

Short-term photovoltaic power forecasting requires models to represent regular solar cycles and weather-driven fluctuations whose importance changes with the forecast horizon. This study develops a hierarchical ensemble that combines temporal neural models, historical analogs, state climatology, and gradient-boosted trees. Solar geometry and numerical weather forecasts describe the expected generation conditions, while horizon-specific convex weights combine complementary predictions. A separate calibration step uses available historical forecast errors to account for recent bias. The framework is evaluated on public PVDAQ data at 15--240-minute horizons and on three GEFCom2014 solar zones at hourly horizons up to four hours. On PVDAQ, the ensemble achieves a daylight capacity-normalized mean absolute error of 4.315%, reducing error by 4.11% relative to full-feature LightGBM and by 6.03% relative to fine-tuned Chronos-2 under identical calibration. Expert-removal experiments identify redundancy within the ensemble. Across three training seeds on GEFCom2014, learned fusion improves upon equal weighting but performs comparably to LightGBM. The results support horizon-specific combination as a useful forecasting strategy while showing that its advantage over strong individual models depends on the dataset and evaluation period.

Chinese Translation

短期光伏功率预测需要模型能够表示规律的太阳周期和天气驱动的波动，其重要性随着预测视野的变化而变化。本研究开发了一种层次集成方法，结合了时间神经模型、历史类比、状态气候学和梯度提升树。太阳几何和数值天气预报描述了预期的发电条件，而特定视野的凸权重则结合了互补的预测。一个单独的校准步骤利用可用的历史预测误差来考虑近期的偏差。该框架在15至240分钟的公共PVDAQ数据和三个GEFCom2014太阳能区域的每小时预测（最长可达四小时）上进行了评估。在PVDAQ上，该集成方法实现了日间容量标准化的平均绝对误差为4.315%，相对于全特征的LightGBM减少了4.11%的误差，相对于经过微调的Chronos-2减少了6.03%的误差，前提是校准条件相同。专家移除实验识别了集成中的冗余。在GEFCom2014的三个训练种子中，学习到的融合方法在等权重的基础上有所改进，但与LightGBM的表现相当。结果支持特定视野组合作为一种有效的预测策略，同时表明其相对于强个体模型的优势依赖于数据集和评估周期。

View on arXiv Download PDF AI Translation

cs.AI / 96 / 2609.15038

The average-farmer illusion in language-model simulations of agricultural decisions

语言模型模拟农业决策中的平均农民幻觉

Zhu, Zhanliang, Li, Ziwei, Liu, Yuchen, Zhu, Liujun, Wu, Ruiqi, Shen, Tongqing, Jin, Junliang, Zhang, Jianyun

Abstract

Language-model agents are increasingly used as synthetic people in surveys and social simulations, yet their apparent realism is often judged from population averages or distributional similarity. We tested what such evidence actually establishes by comparing Claude, Codex and Kimi under four prespecified prompt designs with matched farmer decisions from China and four African countries. Some configurations reproduced observed means and adoption rates. However, their person-level predictions were weak; their decisions clustered around typical values and policy-relevant extremes were largely missing. Most strikingly, a simple generator fitted only to the observed marginal dis- tribution, and given no information about any farmer, achieved greater distributional similarity than every language-model configuration. Prompt additions produced conditional gains rather than uni- versal improvement: results varied with model, outcome, population and validation target. We call this the average-farmer illusion: a synthetic population can look realistic while failing to repro- duce who does what or how behaviour varies. We provide a claim-matched validation framework and reusable modular prompts that turn prompt construction into an auditable experimental process. Population-level resemblance should therefore be treated as the start of validation, not as evidence of individual simulation.

Chinese Translation

语言模型代理越来越多地被用作调查和社会模拟中的合成个体，但它们的表面现实性往往是通过人口平均值或分布相似性来判断的。我们通过比较Claude、Codex和Kimi在四种预设提示设计下与中国和四个非洲国家的农民决策进行匹配，测试了这样的证据实际上建立了什么。一些配置再现了观察到的均值和采纳率。然而，它们在个体层面的预测较弱；它们的决策集中在典型值附近，政策相关的极端情况大多缺失。最引人注目的是，一个仅根据观察到的边际分布拟合的简单生成器，在没有任何农民信息的情况下，达到了比每个语言模型配置更高的分布相似性。提示的增加产生了条件性收益，而不是普遍改善：结果因模型、结果、人口和验证目标而异。我们称之为平均农民幻觉：合成的人口可以看起来现实，但却未能再现谁做了什么或行为如何变化。我们提供了一个声明匹配的验证框架和可重用的模块化提示，将提示构建转变为可审计的实验过程。因此，人口层面的相似性应被视为验证的起点，而不是个体模拟的证据。

View on arXiv Download PDF AI Translation

cs.AI / 97 / 2609.15054

BusMA: A Bus Communication Substrate for Multi-Agent Systems

BusMA：一种用于多智能体系统的总线通信基础设施

Peng, Yanwen, Zhang, Delvin Ce, Wang, Xi, Aletras, Nikolaos

Abstract

Multi-Agent (MA) systems are effective at solving complex tasks that demand planning, tool use, and the synthesis of evidence from multiple sources. Existing systems typically adopt Hierarchical Manager-Worker (HMW) or Router-based Message Passing (RMP) structures as their communication protocol. However, these designs restrict agent autonomy: Worker agents cannot directly consult specific "peers", and misrouted messages can propagate errors. Inspired by bus architectures in computer systems, we propose BusMA, a communication framework that allows any agent to address other agents through a shared channel, i.e., the Bus. It consists of agent registration, message routing, and shared memory management components. Worker agents, each equipped with tools, have their own local memory and can reason, act (tool usage), and communicate by posting shared messages with specific intents. We introduce four intents: discussion, challenge, guidance, and request for explanation, which support fine-grained communication among agents. A Chair agent monitors the shared memory to coordinate interactions and facilitate convergence among Workers. To evaluate the effectiveness of BusMA, we conduct extensive experiments with two frontier LLMs across 13 tasks spanning visual reasoning, mathematical reasoning, and knowledge retrieval demonstrate that BusMA consistently outperforms state-of-the-art HMW and RMP methods.

Chinese Translation

多智能体（MA）系统在解决需要规划、工具使用和从多个来源综合证据的复杂任务方面表现出色。现有系统通常采用层次化的管理-工作者（HMW）或基于路由器的消息传递（RMP）结构作为其通信协议。然而，这些设计限制了智能体的自主性：工作者智能体无法直接咨询特定的“同伴”，而错误路由的消息可能导致错误传播。受到计算机系统中总线架构的启发，我们提出了BusMA，这是一种通信框架，允许任何智能体通过共享通道（即总线）与其他智能体进行通信。它由智能体注册、消息路由和共享内存管理组件组成。每个配备工具的工作者智能体都有自己的本地内存，可以通过发布具有特定意图的共享消息进行推理、行动（工具使用）和沟通。我们引入了四种意图：讨论、挑战、指导和请求解释，以支持智能体之间的细粒度通信。主席智能体监控共享内存，以协调交互并促进工作者之间的收敛。为了评估BusMA的有效性，我们对两个前沿的LLM进行了广泛实验，涵盖了视觉推理、数学推理和知识检索等13个任务，结果表明BusMA始终优于最先进的HMW和RMP方法。

View on arXiv Download PDF AI Translation

cs.AI / 98 / 2609.15078

Enabling Creative Exploration for Vibe Design Agents

为氛围设计代理启用创意探索

Zhang, Yifan, Bui, Nghi D. Q., Evangelopoulos, Georgios, Benard, Arnaud

Abstract

Vibe design agents turn natural-language briefs into rendered interfaces and frontend code. Yet a useful design agent should do more than produce one valid page: it should help users explore coherent alternatives. Increasing token-level temperature is a blunt solution because it varies aesthetic decisions and syntax-sensitive code at the same time. We instead separate exploration from implementation through an inference architecture that makes design direction an explicit intermediate decision. Inspired by Verbalized Sampling, a pre-pass proposes structured design specifications with typicality scores, an external selector samples one, and the downstream generator realizes the selected specification together with the original request under fixed settings. We apply this approach to UI themes and visual-asset prompts. Across 168 prompts, with 1,255 paired comparisons per temperature for each intervention, theme sampling broadens observed selection coverage and screenshot variation, while LLM-judge preferences vary across interventions, prompt complexity, and viewport. In an online experiment with more than 300,000 tasks, the observed code-export increase remains statistically uncertain, while fewer negative feedback events coexist with more correction interactions and modest operational costs. Together, these findings identify structured design specifications as a practical control point for exploring alternative UI concepts while keeping downstream generation settings fixed.

Chinese Translation

氛围设计代理将自然语言简报转化为渲染的界面和前端代码。然而，一个有用的设计代理应该不仅仅生成一个有效的页面：它还应该帮助用户探索一致的替代方案。提高令牌级别的温度是一种粗糙的解决方案，因为它同时改变了美学决策和对语法敏感的代码。我们通过一种推理架构将探索与实现分开，使设计方向成为一个明确的中间决策。受到可言语化采样（Verbalized Sampling）启发，预处理阶段提出了带有典型性评分的结构化设计规范，外部选择器从中抽样一个，后续生成器在固定设置下实现所选规范及原始请求。我们将这种方法应用于用户界面主题和视觉资产提示。在168个提示中，每个干预措施的每个温度有1,255个配对比较，主题采样扩大了观察到的选择覆盖范围和截图变异，而大型语言模型（LLM）评判者的偏好在不同干预、提示复杂性和视口中有所不同。在一项超过300,000个任务的在线实验中，观察到的代码导出增加在统计上仍然不确定，而较少的负面反馈事件与更多的纠正交互和适度的运营成本共存。综合这些发现，结构化设计规范被确定为探索替代用户界面概念的实际控制点，同时保持下游生成设置不变。

View on arXiv Download PDF AI Translation

cs.AI / 99 / 2609.15089

ER-EDF: A Psychology-Grounded Emotion Regulation Framework for Speech Empathetic Dialogue Generation in Large Audio-Language Models

ER-EDF：基于心理学的情感调节框架用于大规模音频语言模型中的语音同理对话生成

Jin, Hongyu, Zhang, Wenda, Fei, Runqiu, Huang, Gongping, Conway, Mike, Dang, Ting

Abstract

Empathetic response generation in spoken dialogue systems requires both accurate emotion perception and appropriate emotion regulation. Grounded in psychological theories such as the Perception-Action Model and emotion regulation theory, effective empathy depends not only on inferring a user's affective state but also on regulating how it is expressed in responses. However, recent large audio-language models (LALMs) largely treat emotion as a direct conditioning signal, lacking explicit regulatory mechanisms, which often leads to affect mirroring rather than calibrated support. We propose ER-EDF, a psychology-grounded framework that explicitly decouples emotion perception and emotion regulation in LALMs. Perception tracks the user's emotional state, while regulation determines how this state should guide empathetic response generation. The framework is model-agnostic and integrates seamlessly into existing LALMs. We further construct a spoken empathetic dialogue dataset and introduce empathy-aware evaluation metrics beyond lexical matching. Experiments across five LALMs and two datasets show that ER-EDF consistently improves empathetic response quality in both automatic and human evaluations, highlighting the importance of jointly modeling emotion perception and regulation in spoken empathetic dialogue systems, paving a new direction for psychologically grounded empathetic AI.

Chinese Translation

在口语对话系统中生成同理回应需要准确的情感感知和适当的情感调节。基于感知-行动模型（Perception-Action Model）和情感调节理论，有效的同理心不仅依赖于推断用户的情感状态，还依赖于调节这种状态在回应中的表达方式。然而，最近的大规模音频语言模型（LALMs）通常将情感视为直接的条件信号，缺乏明确的调节机制，这往往导致情感的简单反映而非经过校准的支持。我们提出了ER-EDF，一个基于心理学的框架，明确解耦LALMs中的情感感知和情感调节。感知部分跟踪用户的情感状态，而调节部分决定如何引导同理回应的生成。该框架与模型无关，能够无缝集成到现有的LALMs中。我们进一步构建了一个口语同理对话数据集，并引入了超越词汇匹配的同理心感知评估指标。在五个LALMs和两个数据集上的实验表明，ER-EDF在自动和人工评估中始终提高了同理回应的质量，突显了在口语同理对话系统中联合建模情感感知和调节的重要性，为基于心理学的同理心人工智能开辟了新的方向。

View on arXiv Download PDF AI Translation

cs.AI / 100 / 2609.15096

OpenAl4S: Code as Action, Science as Sessions

OpenAI4S：代码作为行动，科学作为会话

Zhang, Gongbo, Li, Hao, Wang, Yu, Lin, Mujie, Lv, Liuzhenghao, Mao, Yicheng, Wang, Yimi, Zhu, Jun, Tang, Minhan, Jiang, Zhengxiang, Wang, Yusong, Yao, Jiayu, Ning, Kunpeng, Pang, Dawei, Tian, Yonghong, Community, OpenAI4S, Liu, Yuyang, Yuan, Li

Abstract

AI co-scientists could accelerate computational research, but over a long-running study the workflow also has to stay inspectable, resumable and reproducible, which requires persistent computational state and provenance. Here we present OpenAI4S, an open-source scientific research agent built around the principle of \emph{Code as Action, Science as Sessions}. OpenAI4S combines a persistent computing runtime with research-session management: orchestration is handled through structured tool calls, while scientific actions are represented as complete code cells executed in persistent Python and R kernels. An append-only Action Ledger, per-cell execution records, versioned artifacts, environment records, and workspace checkpoints preserve how results were produced and support session recovery, branching, and extension. Configurable sandboxing, permission controls, and code and trajectory screening provide complementary safeguards. We evaluate OpenAI4S on 36 research scenarios spanning retrosynthesis, molecular dynamics, protein binder design, protein mutation, catalyst screening, and mineral spectroscopy, measuring scientific task accuracy, workflow completeness, and reproducibility of the resulting repositories. OpenAI4S achieves an overall score of 7.83, compared with 5.7--6.4 for a general-purpose coding harness evaluated with three frontier models, with the largest gains on long-horizon and computation-intensive workflows. These results suggest that integrating persistent execution with session-level provenance can improve the reliability of AI-assisted scientific workflows. Environment specification and full rerunnability remain weak for every evaluated system, ours included, so reproducibility is still an open problem for scientific agents. The system is available under the MIT license at \href{https://github.com/PKU-YuanGroup/OpenAI4S}{github.com/PKU-YuanGroup/OpenAI4S}.

Chinese Translation

人工智能共同科学家可以加速计算研究，但在长期研究中，工作流程还必须保持可检查、可恢复和可重复，这需要持久的计算状态和来源。在此，我们介绍了OpenAI4S，一个围绕“代码作为行动，科学作为会话”原则构建的开源科学研究代理。OpenAI4S将持久计算运行时与研究会话管理相结合：通过结构化工具调用进行调度，而科学行动则表示为在持久的Python和R内核中执行的完整代码单元。附加的操作分类账、每个单元的执行记录、版本化的工件、环境记录和工作区检查点保留了结果产生的过程，并支持会话恢复、分支和扩展。可配置的沙箱、权限控制以及代码和轨迹筛选提供了补充的保护措施。我们在36个研究场景中评估了OpenAI4S，涵盖了逆合成、分子动力学、蛋白质结合剂设计、蛋白质突变、催化剂筛选和矿物光谱学，测量科学任务的准确性、工作流程的完整性和结果库的可重复性。与三种前沿模型评估的通用编码工具相比，OpenAI4S的整体得分为7.83，而通用工具的得分为5.7至6.4，最大的提升出现在长时间跨度和计算密集型工作流程上。这些结果表明，将持久执行与会话级来源集成可以提高人工智能辅助科学工作流程的可靠性。环境规范和完全可重运行性在每个评估系统中仍然较弱，包括我们的系统，因此可重复性仍然是科学代理的一个未解决问题。该系统在MIT许可证下可用，网址为https://github.com/PKU-YuanGroup/OpenAI4S。

View on arXiv Download PDF AI Translation

cs.AI / 101 / 2609.15129

Medical Knowledge Simplification for Patients in the Era of LLMs: A Case Study on Diabetes

在大语言模型时代为患者简化医学知识：以糖尿病为例

Kafle, Pallika, Zhou, Yipeng, Liu, Guanfeng, Sheng, Quan Z., Hsu, Cheng-Hsin

Abstract

Complex medical information is often difficult for patients to understand, making effective medical knowledge simplification essential for improving patient comprehension, informed decision-making, and health outcomes. Recent advances in large language models (LLMs) provide a promising approach for simplifying complex medical information into patient-friendly language; however, their effectiveness in real-world patient education remains insufficiently explored through human evaluation. To investigate their practical effectiveness, this paper presents a case study on diabetes knowledge simplification through the implementation and evaluation of MediClear, an LLM-based medical knowledge simplification system enhanced with Retrieval-Augmented Generation (RAG). Public diabetes-related articles from Diabetes Australia, WHO, American Diabetes Association (ADA), NIDDK, and AIHW are indexed in the RAG knowledge base to retrieve clinically grounded information, which is then simplified by the LLM into accessible patient explanations. We evaluate the generated responses using standard readability metrics, including the Flesch-Kincaid Grade Level (FKGL), and conduct a human study involving 10 participants. Results show that MediClear consistently reduces the reading level of generated responses to the recommended patient literacy range while achieving high user satisfaction and willingness for future use. This case study demonstrates the potential of LLMs to improve the accessibility of medical knowledge for patient education.

Chinese Translation

复杂的医学信息常常难以为患者所理解，因此有效的医学知识简化对于提高患者理解能力、知情决策和健康结果至关重要。近年来，大语言模型（LLMs）的进展为将复杂医学信息简化为患者易懂的语言提供了一种有前景的方法；然而，其在现实世界患者教育中的有效性尚未通过人类评估得到充分探索。为了研究其实际有效性，本文通过实施和评估基于LLM的医学知识简化系统MediClear，展示了糖尿病知识简化的案例研究，该系统增强了检索增强生成（Retrieval-Augmented Generation, RAG）技术。来自糖尿病澳大利亚、世界卫生组织（WHO）、美国糖尿病协会（ADA）、美国国立糖尿病、消化病和肾病研究所（NIDDK）以及澳大利亚卫生与福利研究所（AIHW）的公共糖尿病相关文章被索引到RAG知识库中，以检索临床基础信息，然后由LLM简化为可供患者理解的解释。我们使用标准可读性指标评估生成的响应，包括Flesch-Kincaid年级水平（FKGL），并进行了一项涉及10名参与者的人类研究。结果表明，MediClear始终将生成响应的阅读水平降低到推荐的患者识字范围，同时实现了高用户满意度和未来使用意愿。该案例研究展示了LLMs在改善医学知识可及性方面的潜力，以促进患者教育。

View on arXiv Download PDF AI Translation

cs.AI / 102 / 2609.15134

HazardAuditor: From Executable Threats to Safer Computer-Use Agents

HazardAuditor：从可执行威胁到更安全的计算机使用代理

Feng, Yunhao, Lin, Ruixiao, Wen, Ming, Guo, Yanming, Ma, Xingjun, Wu, Yutao, Deng, Xinhao, Ji, Shouling

Abstract

Computer-use agents increasingly interact with browsers, terminals, file systems, and external services, introducing safety risks that emerge through runtime behavior rather than generated content alone. Existing guard models target static prompts and responses and are poorly suited to agent execution; existing executable safety platforms produce evaluation verdicts rather than the normalized supervision a guard model needs to learn across heterogeneous agent frameworks. We introduce HazardAuditor, an execution-grounded framework that closes both gaps. Its infrastructure runs heterogeneous agents (Claude Code, Codex, Hermes, and OpenClaw) in controlled environments and normalizes their interactions into a canonical event representation for cross-framework supervision. We further observe that token-level post-training objectives create a structural mismatch for generative guards, causing longer rationales to dominate gradient updates. Guard Policy Optimization (GuardPO) addresses this by converting deterministic safety outcomes into sequence-level advantages and normalizing rationale and verdict regions, making the safety decision the effective unit of optimization. Across multiple benchmarks and heterogeneous computer-use systems, HazardAuditor improves accuracy by up to 16.5 percentage points over the strongest prior guard. Code, models, and evaluation artifacts will be available at https://yunhao-feng.github.io/HazardAuditor/.

Chinese Translation

计算机使用代理越来越多地与浏览器、终端、文件系统和外部服务进行交互，这引入了通过运行时行为而非仅通过生成内容而出现的安全风险。现有的防护模型针对静态提示和响应，难以适应代理执行；现有的可执行安全平台产生评估裁决，而不是防护模型在异构代理框架中学习所需的规范化监督。我们提出了HazardAuditor，一个基于执行的框架，填补了这两个空白。其基础设施在受控环境中运行异构代理（Claude Code、Codex、Hermes和OpenClaw），并将它们的交互规范化为跨框架监督的标准事件表示。我们进一步观察到，基于令牌的后训练目标为生成性防护造成了结构不匹配，导致较长的推理主导了梯度更新。防护策略优化（Guard Policy Optimization, GuardPO）通过将确定性的安全结果转化为序列级优势并规范化推理和裁决区域，解决了这一问题，使安全决策成为有效的优化单位。在多个基准测试和异构计算机使用系统中，HazardAuditor的准确性比最强的先前防护提高了多达16.5个百分点。代码、模型和评估文档将可在 https://yunhao-feng.github.io/HazardAuditor/ 获取。

View on arXiv Download PDF AI Translation

cs.AI / 103 / 2609.15160

T-LoopFormer: Token-Level Elastic-Depth Looped Transformers for Latent Reasoning With Dynamic Routing

T-LoopFormer：用于潜在推理的动态路由令牌级弹性深度循环变换器

Yu, Mingqian, Zhang, Wenpeng, Zhao, Peilin

Abstract

Looped Transformers have recently demonstrated strong performance in both reasoning and language tasks by reusing a shared set of parameters across multiple iterations, achieving parameter efficiency without sacrificing representational power. Besides, looped Transformers perform inference directly in the latent space (latent reasoning) to reduce the number of tokens consumed during inference, thereby achieving improved sample efficiency. However, these models typically apply a fixed recursion depth uniformly to every token, leading to suboptimal compute allocation and leaving significant efficiency gains on the table. In this work, we propose \textbf{dynamic token-choice routing} for looped transformers, enabling each token to adaptively determine its own number of loop iterations based on its hidden state. We use a dynamic router to decide whether a token should continue recursing or exit early, allowing simple tokens to bypass unnecessary computation while hard tokens receive deeper processing. To ensure that this adaptive mechanism does not compromise decoding efficiency, we further introduce recursion-wise KV caching, which maintains an independent key-value cache for each recursion loop. This design ensures that tokens at different depths only attend to their corresponding cached states, effectively eliminating redundant computations for exited tokens and enabling fast autoregressive decoding. Extensive experiments show that T-LoopFormer reaches the sota performance under the same parameters on PPL and 10 zero-shot reasoning tasks, even surpassing the base model at 24x FLOPs and our model could reach the lowest inference latency, which validate the effectiveness of token-choice router and recursion-wise KV cache. Code: https://github.com/YuMingQian1234/T-LoopFormer.

Chinese Translation

循环变换器最近在推理和语言任务中表现出强大的性能，通过在多个迭代中重用一组共享参数，实现了参数效率而不牺牲表示能力。此外，循环变换器直接在潜在空间中进行推理（潜在推理），以减少推理过程中消耗的令牌数量，从而提高样本效率。然而，这些模型通常对每个令牌应用固定的递归深度，导致计算分配不够优化，并留下显著的效率提升空间。在本研究中，我们提出了针对循环变换器的 extbf{动态令牌选择路由}，使每个令牌能够根据其隐状态自适应地确定其循环迭代次数。我们使用动态路由器来决定一个令牌是否应继续递归或提前退出，从而允许简单令牌跳过不必要的计算，而复杂令牌则接受更深层次的处理。为了确保这一自适应机制不会影响解码效率，我们进一步引入了递归级别的键值缓存（KV caching），为每个递归循环维护独立的键值缓存。这一设计确保不同深度的令牌仅关注其对应的缓存状态，有效消除已退出令牌的冗余计算，并实现快速自回归解码。大量实验表明，T-LoopFormer在相同参数下在PPL和10个零样本推理任务上达到了最先进的性能，甚至在24倍FLOPs的情况下超越了基础模型，并且我们的模型能够实现最低的推理延迟，验证了令牌选择路由器和递归级别KV缓存的有效性。代码链接： https://github.com/YuMingQian1234/T-LoopFormer.

View on arXiv Download PDF AI Translation

cs.AI / 104 / 2609.15172

STHMoE: Hypergraph-Enhanced Heterogeneous Dependency Coordination for LLM-Based Urban Traffic Data Forecasting

STHMoE：基于超图增强的异构依赖协调用于基于大语言模型的城市交通数据预测

Chen, Jiawen, Shao, Qi, Chang, Yongjian, Zhou, Mingtong, Chen, Duxin, Yu, Wenwu

Abstract

Spatio-temporal traffic forecasting is a fundamental big data analytics task for intelligent transportation systems, where massive urban sensor streams exhibit heterogeneous, non-stationary, and structurally dynamic patterns. Although recent deep learning and large language model (LLM)-based methods have advanced traffic forecasting, they often remain temporally centered and lack effective coordination of temporal, spectral, pairwise spatial, and higher-order structural cues under evolving traffic regimes. To address this heterogeneous dependency coordination problem, we propose STHMoE, a Spatio-Temporal Hypergraph-Enhanced Mixture of Experts framework for urban traffic data forecasting. STHMoE decouples traffic dynamics into frequency-domain, time-domain, spatio-domain, and higher-order spatial representations, which are modeled by prompt-guided heterogeneous experts built upon a partially frozen LLM backbone. The first three experts leverage domain-specific statistical prompts, while the higher-order spatio expert uses a structural placeholder prompt and obtains dependency information from an adaptive hypergraph module. To capture evolving spatial structures in traffic data,, STHMoE jointly learns first-order graph dependencies and higher-order group interactions without predefined topologies. An entropy-aware MoE router with coefficient-of-variation load balancing adaptively fuses expert outputs while improving expert utilization and routing confidence. Experiments on 10 real-world traffic benchmarks show that STHMoE achieves competitive performance against temporal, spatio-temporal graph, and LLM-based baselines.

Chinese Translation

时空交通预测是智能交通系统中的一项基础大数据分析任务，其中大量城市传感器流展现出异构、非平稳和结构动态的模式。尽管近期基于深度学习和大语言模型（LLM）的方法在交通预测方面取得了进展，但它们往往仍然以时间为中心，缺乏在不断变化的交通模式下对时间、频谱、成对空间和高阶结构线索的有效协调。为了解决这一异构依赖协调问题，我们提出了STHMoE，一种用于城市交通数据预测的时空超图增强专家混合框架。STHMoE将交通动态解耦为频域、时域、空间域和高阶空间表示，这些表示由基于部分冻结的LLM骨干网络构建的提示引导异构专家进行建模。前三个专家利用特定领域的统计提示，而高阶空间专家使用结构占位符提示，并从自适应超图模块中获取依赖信息。为了捕捉交通数据中不断演变的空间结构，STHMoE在没有预定义拓扑的情况下共同学习一阶图依赖和高阶组交互。一个具有熵感知的MoE路由器通过变异系数负载平衡自适应融合专家输出，同时提高专家利用率和路由信心。在10个真实世界交通基准上的实验表明，STHMoE在与时间、时空图和基于LLM的基准相比时，表现出竞争力的性能。

View on arXiv Download PDF AI Translation

cs.AI / 105 / 2609.15182

VisInteract: Towards Dynamic Interactive Text-to-Visualization under Imperfect Queries

VisInteract：在不完美查询下实现动态交互式文本到可视化

Xu, Wenxin, Lu, Jinwei, Kim, Hwanhee, Zhang, Chen Jason, Wei, Xiao-Yong, Li, Haoyang, Song, Yuanfeng

Abstract

Real-world visualization requests are routinely ambiguous, incomplete, or factually incorrect, yet existing Text-to-Visualization (Text-to-Vis) systems assume well-specified inputs and produce charts in a single pass. When queries are imperfect, a system must \emph{interact} with the user to recover the true intent, but no benchmark or method supports this dynamic process. We introduce \textbf{VisInteract}, a new paradigm that reframes Text-to-Vis as interaction-driven intent recovery, and \textbf{VisInteract-Bench}, to our knowledge, that is the first benchmark for dynamic interactive Text-to-Vis, featuring controlled imperfection injection, a leakage-controlled User Agent for realistic multi-turn feedback, and dual-perspective (code and chart) automated evaluation. On the algorithmic side, we propose \textbf{Vis-MCTS}, a Monte Carlo Tree Search (MCTS) enhanced method, introducing improvements over classical MCTS, that \emph{Progressive Widening} to tame the unbounded tool-argument space in tree search, \emph{cross-rollout information sharing} so clarifications and critiques benefit the entire search tree, and \emph{Dimension-Aware Reward Decomposition} that routes scalar user feedback along data-fidelity, visual-design, and intent-alignment dimensions to resolve credit assignment across heterogeneous actions. Extensive Experiments across two LLM backbones show that Vis-MCTS consistently outperforms all Text-to-Vis baselines, improving end-to-end task success by $13.40\%$--$16.27\%$ over the strongest interactive baseline and by more than $5\times$ over non-interactive ones.

Chinese Translation

现实世界中的可视化请求通常模糊、不完整或事实不准确，而现有的文本到可视化（Text-to-Visualization, Text-to-Vis）系统假设输入是明确的，并且在单次处理过程中生成图表。当查询不完美时，系统必须与用户进行交互以恢复真实意图，但目前没有任何基准或方法支持这一动态过程。我们提出了 extbf{VisInteract}，一种将文本到可视化重新框架为以交互驱动的意图恢复的新范式，并且 extbf{VisInteract-Bench}是我们所知的第一个动态交互式文本到可视化基准，具有受控的不完美注入、一个泄漏控制的用户代理以实现现实的多轮反馈，以及双视角（代码和图表）自动评估。在算法方面，我们提出了 extbf{Vis-MCTS}，一种增强的蒙特卡洛树搜索（Monte Carlo Tree Search, MCTS）方法，引入了对经典MCTS的改进，包括 extit{渐进扩展}以驯服树搜索中无限的工具-参数空间、 extit{交叉回滚信息共享}以使澄清和批评惠及整个搜索树，以及 extit{维度感知奖励分解}，将标量用户反馈沿数据保真度、视觉设计和意图对齐维度进行路由，以解决异构动作之间的信用分配。针对两个大型语言模型（LLM）基础模型的广泛实验表明，Vis-MCTS始终优于所有文本到可视化基准，其端到端任务成功率比最强的交互基线提高了$13.40\%$--$16.27\\%$，并且比非交互式基线提高了超过$5 imes$。

View on arXiv Download PDF AI Translation

cs.AI / 106 / 2609.15207

Issue Bias in Generative AI Writing Assistance: Political Issues and LLMs in the Swedish 2026 Election

生成性人工智能写作助手中的议题偏见：瑞典2026年选举中的政治问题与大型语言模型（LLMs）

Bruinsma, Bastiaan, Fredén, Annika, Röttger, Paul, Johansson, Moa, Sayeed, Asad

Abstract

Generative AI writing assistants and the Large Language Models (LLMs) that power them are increasingly part of how voters gather information before elections. With growing evidence that they influence users' opinions, it is increasingly important to understand the views and positions of these tools. To better understand these views, we examine the stances supplied by six LLMs on a variety of Swedish-language writing tasks ahead of the 2026 Swedish parliamentary election. We cross 107 policy propositions with 77 writing templates and neutral, positive, and negative prompt framings, producing 24,717 prompts per model and 148,302 responses. To study these, we look at the models' default stance tendencies, compare how they respond to similar issues, and compare their responses with those of each of Sweden's eight parliamentary parties on the same issue. We find that Claude, DeepSeek, Gemini, and Mistral have similar profiles; ChatGPT more often supplies neutral or ambivalent text; and Grok differs most on topics such as migration, crime, and gender. When comparing the political parties, we find that the Social Democrats are closest to all six models. Still, after correcting for multiple comparisons, none of the within-model differences in party distances remains significant. Overall, we find that no model has a clear preference, nor a clear preference for a party, but that this depends on the specific issue or task the user asks about.

Chinese Translation

生成性人工智能写作助手及其背后的大型语言模型（LLMs）日益成为选民在选举前获取信息的重要工具。随着越来越多的证据表明它们会影响用户的观点，理解这些工具的观点和立场变得愈发重要。为了更好地理解这些观点，我们考察了六个LLMs在2026年瑞典议会选举前对多种瑞典语写作任务所提供的立场。我们将107项政策提案与77个写作模板以及中立、积极和消极的提示框架交叉，生成每个模型24,717个提示和148,302个响应。为研究这些，我们分析了模型的默认立场倾向，比较它们对类似问题的响应，并将其回应与瑞典八个议会政党在同一问题上的回应进行比较。我们发现Claude、DeepSeek、Gemini和Mistral具有相似的特征；ChatGPT更常提供中立或模棱两可的文本；而Grok在移民、犯罪和性别等话题上表现出最大的差异。在比较政治党派时，我们发现社会民主党与所有六个模型的距离最接近。然而，在进行多重比较校正后，模型内部的党派距离差异均不再显著。总体而言，我们发现没有模型表现出明确的偏好，也没有对某一政党表现出明确的偏好，而这取决于用户询问的具体问题或任务。

View on arXiv Download PDF AI Translation

cs.AI / 107 / 2609.15219

From Ideas to Actions: A Public-Data Decision-Support Toolchain Across the Venture Lifecycle

从理念到行动：跨越创业生命周期的公共数据决策支持工具链

Qu, Lei

Abstract

Founders face two linked decisions: whether to pursue an idea before founding, and which operating actions and capital partners fit afterward. We present a public-data decision-support toolchain combining time-bounded proposal profiling, market and moat checks, and deterministic aggregation with auditable investor-company event chains for retrospective analysis. Pre-founding: (a) After threshold selection on 198 development companies, the frozen pipeline achieves F0.5=0.5357 [0.412, 0.655] on an independent, row-disjoint 198-company validation sample. On the combined 396 rows, the Full Pipeline scores 0.6301 versus 0.2734 for a paired Raw LLM baseline. Post-stratification of 1,027 completed cases in a separate scale cohort yields 0.6506 [0.598, 0.707]; the run remains incomplete. A 377-row composition-matched check yields 0.6573. (b) The AI-inference study identifies distribution-layer businesses as a replicable path to independent profitability with a limited revenue ceiling, and frontier-model ownership as a path to capital-market upside at exceptional capital cost. Post-founding: (a) Public sources support auditable event-chain analysis. (b) In the chip-company study, sustained product, customer, and supply-chain progress is associated with better observed outcomes; financing alone does not establish operating progress. (c) Financing comprises 79% of confirmed visible post-investment actions. Evidence tentatively favors acquisition-experienced strategic corporate investors for acquisition-oriented founders and financing-led institutional VCs with fewer observed control events for independence-oriented founders. Findings are developmental and observational, not causal guarantees or investment advice. We release shared ontology, provenance-bearing EventChain data, schemas, benchmarks, and executable skills for audit, reuse, and extension.

Chinese Translation

创始人在创业前面临两个相关的决策：是否追求一个想法，以及在创业后选择哪些运营行动和资本合作伙伴。我们提出了一种公共数据决策支持工具链，结合了时间限制的提案分析、市场和护城河检查，以及可审计的投资者-公司事件链的确定性聚合，以进行回顾性分析。在创业前：（a）在198家发展公司上进行阈值选择后，冻结管道在独立的、行不相交的198家公司验证样本上达到F0.5=0.5357 [0.412, 0.655]。在合并的396行上，完整管道得分为0.6301，而配对的原始LLM基线得分为0.2734。在一个单独的规模队列中，对1,027个已完成案例的后分层分析得出0.6506 [0.598, 0.707]；该运行仍未完成。377行的组成匹配检查得出0.6573。（b）AI推断研究确定分布层业务作为实现独立盈利的可复制路径，有限的收入上限，以及前沿模型所有权作为在特殊资本成本下实现资本市场收益的路径。在创业后：（a）公共来源支持可审计的事件链分析。（b）在芯片公司研究中，持续的产品、客户和供应链进展与更好的观察结果相关；仅靠融资并不能建立运营进展。（c）融资占确认的可见投资后行动的79%。证据初步倾向于对收购导向的创始人而言，具有收购经验的战略企业投资者，以及对独立导向的创始人而言，观察到的控制事件较少的融资主导的机构风险投资者。研究结果是发展性和观察性的，并非因果保证或投资建议。我们发布共享本体、具有来源的事件链数据、模式、基准和可执行技能，以供审计、重用和扩展。

View on arXiv Download PDF AI Translation

cs.AI / 108 / 2609.15234

CWM: Controllable White-Box Meta-Prompting for Adaptive Retrieval-Augmented Generation and Reasoning Ability

CWM：可控白盒元提示用于自适应检索增强生成和推理能力

Kim, Keuntae, Jeong, Eunhye, Choi, Yong Suk

Abstract

Recently, Large Language Models (LLMs) have gained significant attention due to their strong language understanding and generation capabilities, demonstrating impressive reasoning abilities as well as effective utilization of external knowledge. Many studies have proposed methods that specialize in improving performance for individual tasks. However, ironically, only a limited number of attempts have explored general-purpose, task-agnostic methods. In this work, we present a unified framework integrating reasoning and Retrieval-Augmented Generation (RAG) tasks. We further propose Controllable White-Box Meta-Prompting (CWM), a low-cost white-box method for adaptive RAG tasks previously dominated by black-box approaches, without requiring external decision modules or multi-sampling. CWM achieves state-of-the-art performance on three adaptive RAG benchmarks across recent LLMs, including GPT-oss-20b, Qwen3-14b, and Llama3.1-8b, while also demonstrating strong generality by extending to reasoning tasks. In addition, CWM provides controllability by enabling retrieval decisions to be regulated through the manipulation of internal model signals. Our code is available at https://github.com/JeongEunhye00/CWM.

Chinese Translation

近年来，大型语言模型（LLMs）因其强大的语言理解和生成能力而受到广泛关注，展现出令人印象深刻的推理能力以及有效利用外部知识的能力。许多研究提出了专注于提高单个任务性能的方法。然而，讽刺的是，只有有限数量的尝试探索了通用的、与任务无关的方法。在本研究中，我们提出了一个统一框架，整合了推理和检索增强生成（RAG）任务。我们进一步提出了可控白盒元提示（CWM），这是一种低成本的白盒方法，适用于之前由黑盒方法主导的自适应RAG任务，无需外部决策模块或多次采样。CWM在包括GPT-oss-20b、Qwen3-14b和Llama3.1-8b在内的三项自适应RAG基准上实现了最先进的性能，同时通过扩展到推理任务展示了强大的通用性。此外，CWM通过操控内部模型信号来调节检索决策，从而提供了可控性。我们的代码可在 https://github.com/JeongEunhye00/CWM 获取。

View on arXiv Download PDF AI Translation

cs.AI / 109 / 2609.15242

Empirical Evaluation of Open-Source Large Language Models for Retrieval-Augmented Generation in ESG Domain

开源大型语言模型在环境、社会和治理（ESG）领域检索增强生成的实证评估

Saad, Motaz, Borrelli, Anna, Gentile, Ivan, Kazemi, Kianna, Piccialli, Francesco, Longo, Antonella

Abstract

Environmental, Social, and Governance (ESG) reporting is critical for corporate accountability, with Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) offering strong potential to automate KPI extraction. However, open-source LLM performance in domain-specific ESG tasks remains insufficiently understood. This paper evaluates open-source LLMs in ESG contexts using a structured framework and evaluation resource based on 498 real-world ESG reports from EU-listed companies (2010-2024). We evaluate seven open-source models (2B to 30B parameters) -- glm-4.7-flash, nemotron-3-nano:4b, qwen3:4b-instruct, gemma3:4b, gemma4:e4b, gemma4:e2b, and ministral-3:8b -- using 100 persona-based synthetic QA pairs covering ESG information needs. System performance is assessed via RAGAS metrics, including contextual recall, precision, relevance, faithfulness, answer relevancy, and factual correctness. Results show notable performance variations across architectures. Retrieval performance is strong across models (context recall around 0.58-0.61, context precision around 0.78-0.81, context relevance 0.965-0.985). Generation diverges most on faithfulness (0.607-0.822) and least on answer relevancy (0.760-0.881): glm-4.7-flash leads in faithfulness (0.822), qwen3 in factual correctness (0.449), and ministral-3 in answer relevancy (0.881). Low overall factual correctness (0.387-0.449) highlights the need for domain-specific fine-tuning. This work provides data-driven guidance for deploying open-source models in ESG reporting.

Chinese Translation

环境、社会和治理（ESG）报告对企业问责至关重要，而大型语言模型（LLMs）和检索增强生成（RAG）在自动化关键绩效指标（KPI）提取方面展现出强大的潜力。然而，开源LLM在特定领域的ESG任务中的表现仍然不够明确。本文使用基于498份来自欧盟上市公司（2010-2024年）的真实ESG报告的结构化框架和评估资源，评估了开源LLM在ESG背景下的表现。我们评估了七个开源模型（参数规模从20亿到300亿）——glm-4.7-flash、nemotron-3-nano:4b、qwen3:4b-instruct、gemma3:4b、gemma4:e4b、gemma4:e2b和ministral-3:8b——使用涵盖ESG信息需求的100个基于角色的合成问答对。系统性能通过RAGAS指标进行评估，包括上下文召回率、精确度、相关性、忠实度、答案相关性和事实正确性。结果显示不同架构之间的性能差异显著。各模型的检索性能较强（上下文召回率约为0.58-0.61，上下文精确度约为0.78-0.81，上下文相关性为0.965-0.985）。生成方面在忠实度（0.607-0.822）上差异最大，而在答案相关性（0.760-0.881）上差异最小：glm-4.7-flash在忠实度上表现最佳（0.822），qwen3在事实正确性上表现最佳（0.449），而ministral-3在答案相关性上表现最佳（0.881）。整体事实正确性较低（0.387-0.449）突显了领域特定微调的必要性。本研究为在ESG报告中部署开源模型提供了基于数据的指导。

View on arXiv Download PDF AI Translation

cs.AI / 110 / 2609.15292

ProIQA: A Process-Based Framework for Fine-Grained Math Item Quality Assessment

ProIQA：一种基于过程的细粒度数学题目质量评估框架

Tong, Junkai, Li, Mingjia, Chen, Haoran, Jiang, Yaoyu, Ge, Hanjie, Wang, Yixuan, Qian, Hong

Abstract

Automatic Item Generation (AIG) is pivotal for personalized education, yet guaranteeing the pedagogical value of generated items remains a bottleneck. Existing Item Quality Assessment (IQA) methods typically rely on unscalable manual reviews or shallow stem-based metrics, failing to capture the reasoning process required for mathematical problem-solving. To bridge this gap, this paper proposes Process-based Item Quality Assessment (ProIQA), a process-aware framework for fine-grained quality assessment of math items. We first formulate IQA across three heterogeneous dimensions, including knowledge concepts, difficulty, and disciplinary competencies, under a unified process-aware perspective. Based on this formulation, we construct a process-enhanced IQA resource by augmenting original item data with structured reasoning trees derived from raw solutions. Technically, ProIQA leverages Large Language Modelsto construct hierarchical reasoning trees and employs Graph Neural Networks (GNN) to encode their topological dependencies and procedural semantics. The resulting solving representation is fused with stem semantics through a dual-view (``Stem + Solving'') architecture, enabling comprehensive assessment across learning objectives. Extensive experiments on K12 mathematical datasets show that ProIQA effectively captures process-oriented features, offering a scalable data-driven solution for evaluating AIG outputs in intelligent education systems.

Chinese Translation

自动题目生成（AIG）对个性化教育至关重要，但确保生成题目的教学价值仍然是一个瓶颈。现有的题目质量评估（IQA）方法通常依赖于不可扩展的人工审查或浅层的题干基础指标，未能捕捉到数学问题解决所需的推理过程。为了解决这一问题，本文提出了基于过程的题目质量评估（ProIQA），这是一个关注过程的细粒度数学题目质量评估框架。我们首先从知识概念、难度和学科能力三个异构维度出发，在统一的过程意识视角下构建IQA。基于这一构建，我们通过将原始题目数据与源自原始解答的结构化推理树相结合，构建了一个增强过程的IQA资源。在技术上，ProIQA利用大型语言模型（Large Language Models）构建层次推理树，并采用图神经网络（Graph Neural Networks, GNN）来编码它们的拓扑依赖关系和过程语义。最终的解题表示通过双视图（“题干 + 解题”）架构与题干语义融合，实现了对学习目标的全面评估。在K12数学数据集上的广泛实验表明，ProIQA有效捕捉到以过程为导向的特征，为智能教育系统中评估AIG输出提供了一种可扩展的数据驱动解决方案。

View on arXiv Download PDF AI Translation

cs.AI / 111 / 2609.15293

Why LLM Agents Collapse Without Oversight: The Enforcement Gap as the Mechanism Behind Emergence World Failures

为何大型语言模型代理在缺乏监督的情况下崩溃：执行缺口作为新兴世界失败的机制

Wang, Yuhang

Abstract

When Emergence World placed frontier LLM agents in an unsupervised multi-agent simulation, the results were alarming: agents committed crimes, starved, and enforced unanimous conformity -- without any external attacker. This paper identifies the mechanism. Reflexion-style agents already detect dangerous plan steps through iterative self-critique, yet the architecture provides no pathway from detection to action. We call this the enforcement gap: the audit sees the problem; the controller ignores it. Closing the gap requires a single conditional check -- fewer than 20 lines of code -- and reduces attack success by more than fourfold in large-scale experiments across frontier models, all five major agent frameworks, and an independent benchmark. We prove formally that when enforcement probability is near zero, detection quality is irrelevant to security. We further identify two compounding failure modes -- unreliable auditors and unparseable verdicts -- that explain every collapse pattern in Emergence World. A GRPO-trained enforcement controller resolves the ambiguity case. Together these results motivate a three-requirement Audit Enforcement Specification that is absent from every deployed framework today.

Chinese Translation

当新兴世界将前沿大型语言模型（LLM）代理置于无监督的多代理仿真中时，结果令人震惊：代理们犯罪、饥饿，并强制实施一致的从众行为——没有任何外部攻击者。本文识别了这一机制。反思式代理通过迭代自我批评已经能够检测到危险的计划步骤，但该架构并未提供从检测到行动的路径。我们称之为执行缺口：审计发现问题；控制者却对此视而不见。弥补这一缺口只需一个条件检查——不到20行代码——并在大规模实验中将攻击成功率降低了四倍以上，涵盖了前沿模型、所有五大主要代理框架以及一个独立基准。我们正式证明，当执行概率接近零时，检测质量与安全性无关。我们进一步识别出两种复合失败模式——不可靠的审计者和不可解析的裁决——这解释了新兴世界中的每一种崩溃模式。经过GRPO训练的执行控制器解决了模糊性问题。这些结果共同推动了一个三项要求的审计执行规范，而这一规范在当前所有部署的框架中均缺失。

View on arXiv Download PDF AI Translation

cs.AI / 112 / 2609.15296

Reason What Matters: Retrieval-Grounded Reasoning for Universal Multimodal Embeddings

关注重要性：基于检索的推理用于通用多模态嵌入

Jiang, Mingzhou, Wu, Peixi, Cheng, Hang, Zhou, Yunhao, Yang, Biao, Yuan, Wei, Li, Yun, Yang, Fan, Ou, Wenwu, He, Honghui

Abstract

Universal multimodal embedding (UME) learns unified representations across modalities, enabling a single model to support diverse retrieval tasks. Recent methods use Chain-of-Thought (CoT) reasoning to better interpret multimodal inputs before generating embeddings for complex retrieval tasks and further optimize this reasoning process through GRPO with retrieval-based rewards. However, two limitations hinder corpus-scale deployment. GRPO assigns all CoT tokens the same advantage, without identifying input-supported claims or evidence that distinguishes the positive from negatives. Moreover, generating a complete CoT before each embedding introduces substantial latency, even when a partial trace already provides sufficient retrieval evidence. To address these limitations, we propose Reason What Matters (ReWAM), a retrieval-grounded reasoning framework that uses retrieval feedback to guide both credit assignment and reasoning computation. Specifically, we introduce Retrieval-aware Self-Distillation (RASD), which constructs privileged guidance from input-supported evidence that distinguishes the positive item from retrieved hard negatives. An on-policy self-teacher uses this guidance to refine trajectory-level feedback into token-specific supervision for retrieval-relevant reasoning. We further develop Retrieval-adaptive Inference (RAI), which uses a retrieval confidence head to estimate the remaining retrieval utility of a partial CoT. It stops unproductive traces early and accelerates useful continuations with speculative decoding. Extensive experiments on MMEB-V2 and MRMR demonstrate that ReWAM achieves state-of-the-art retrieval performance while delivering up to 5x the inference throughput of competitive explicit-CoT UME methods. These results bridge the gap between retrieval quality and inference efficiency, making reasoning-enhanced UME practical for large-scale deployment.

Chinese Translation

通用多模态嵌入（UME）学习跨模态的统一表示，使单一模型能够支持多样的检索任务。近期的方法使用思维链（Chain-of-Thought, CoT）推理来更好地解释多模态输入，然后为复杂的检索任务生成嵌入，并通过基于检索的奖励进一步优化这一推理过程。然而，有两个限制阻碍了在语料库规模上的部署。GRPO（基于检索的奖励优化）对所有CoT令牌赋予相同的优势，而未能识别支持输入的主张或证据，从而区分正例和负例。此外，在每次生成嵌入之前生成完整的CoT会引入显著的延迟，即使部分轨迹已经提供了足够的检索证据。为了解决这些限制，我们提出了关注重要性（Reason What Matters, ReWAM），一个基于检索的推理框架，利用检索反馈来指导信用分配和推理计算。具体而言，我们引入了检索感知自蒸馏（Retrieval-aware Self-Distillation, RASD），它从支持输入的证据中构建特权指导，以区分正例项目和检索到的困难负例。一个在政策内的自教师使用这一指导将轨迹级反馈细化为针对检索相关推理的令牌特定监督。我们进一步开发了检索自适应推理（Retrieval-adaptive Inference, RAI），它使用检索置信度头来估计部分CoT的剩余检索效用。它提前停止无效轨迹，并通过推测解码加速有用的延续。在MMEB-V2和MRMR上的广泛实验表明，ReWAM实现了最先进的检索性能，同时提供了高达5倍的推理吞吐量，相较于竞争性的显式CoT UME方法。这些结果弥合了检索质量与推理效率之间的差距，使得增强推理的UME在大规模部署中变得可行。

View on arXiv Download PDF AI Translation

cs.AI / 113 / 2609.15315

Evaluation Metrics for Safe Reinforcement Learning

安全强化学习的评估指标

Spoor, Lindsay, Plaat, Aske, Moerland, Thomas

Abstract

Safe reinforcement learning (RL) is commonly formalized as a Constrained Markov Decision Process (CMDP), in which an agent maximizes expected reward while keeping its expected cumulative cost below a specified safety bound. Existing safe RL benchmarks predominantly report whether an algorithm is safe on average, following this expectation-based guarantee. We argue that this convention is insufficient to reliably characterize an algorithm's true safety: it fails to capture how often and how severely the safety bound is violated, whether this holds consistently across tasks and safety bounds, and whether training-time behavior is representative of behavior of the final converged policy. Therefore, we introduce (i) evaluation metrics for safe RL that address each of these concerns and in addition allow for aggregation across tasks and safety bounds. We furthermore define (ii) a safety tier system to systematically categorize and compare algorithms in terms of safety and reliability at both training and for a final policy. Using this framework, we provide (iii) an empirical safety evaluation across multiple safety navigation tasks. Our results show that aggregate metrics, distributional reporting, and task- and safety bound-specific results each reveal information the other metrics cannot. We therefore recommend reporting all three jointly, rather than compressing this information into a single value, as is common practice. We provide SafeRLEval, an open-source evaluation suite to support the reliable characterization of safety in future safe RL research.

Chinese Translation

安全强化学习（RL）通常被形式化为约束马尔可夫决策过程（CMDP），其中代理在最大化期望奖励的同时，保持其期望累积成本低于指定的安全界限。现有的安全RL基准主要报告算法在平均情况下是否安全，遵循这种基于期望的保证。我们认为这一惯例不足以可靠地表征算法的真实安全性：它未能捕捉安全界限被违反的频率和严重程度，是否在不同任务和安全界限之间保持一致，以及训练期间的行为是否能够代表最终收敛策略的行为。因此，我们引入了（i）针对安全RL的评估指标，以解决这些问题，并允许在任务和安全界限之间进行聚合。此外，我们定义了（ii）一个安全等级系统，以系统地对算法在训练和最终策略的安全性和可靠性进行分类和比较。利用这一框架，我们提供了（iii）在多个安全导航任务中的实证安全评估。我们的结果表明，聚合指标、分布报告以及任务和安全界限特定的结果各自揭示了其他指标无法提供的信息。因此，我们建议联合报告这三者，而不是像常见做法那样将这些信息压缩为单一值。我们提供了SafeRLEval，一个开源评估套件，以支持未来安全RL研究中安全性的可靠表征。

View on arXiv Download PDF AI Translation

cs.AI / 114 / 2609.15344

Parameter-Efficient Adaptation of Pretrained Language Models for Time-Series Forecasting

预训练语言模型在时间序列预测中的参数高效适应

Kumavat, Tamanna, Brunner, Georg, Flouris, Kyriakos

Abstract

We study the adaptation of pretrained language models to univariate time-series forecasting through a parameter-efficient transfer learning framework, with the goal of understanding which design choices drive effective cross-modal transfer. While language models operate on discrete textual tokens, time series consist of continuous numerical observations with temporal dependencies. To bridge this modality gap, we project fixed-length time-series patches directly into the embedding space of a pretrained GPT-2 backbone, bypassing textual tokenization and treating the Transformer as a generic sequence encoder. Through controlled ablation studies on seven benchmark datasets spanning energy, weather, traffic, and finance, we analyze the effects of (i)~representation strategy (continuous embeddings versus textual serialisation), (ii)~adaptation regime (frozen backbone versus partial or full fine-tuning), (iii)~architectural components such as adapters, pooling strategies, and prediction heads, and (iv)~input context length. Continuous patch-based embeddings consistently outperform textual prompting and randomly initialised backbones. The adapted pipeline attains MASE within the range of specialised forecasting architectures while updating less than 1\% of total model parameters. Results further indicate that freezing the pretrained backbone and training lightweight projection and adapter modules provides a favourable accuracy--efficiency trade-off with stable behaviour across varying context lengths.

Chinese Translation

我们研究了通过参数高效的迁移学习框架将预训练语言模型适应于单变量时间序列预测，旨在理解哪些设计选择驱动有效的跨模态迁移。虽然语言模型在离散文本标记上操作，但时间序列由具有时间依赖性的连续数值观测组成。为了弥合这种模态差距，我们将固定长度的时间序列补丁直接投影到预训练的GPT-2主干的嵌入空间中，绕过文本标记化，将Transformer视为通用序列编码器。通过对涵盖能源、天气、交通和金融的七个基准数据集进行控制消融研究，我们分析了(i) 表示策略（连续嵌入与文本序列化），(ii) 适应机制（冻结主干与部分或完全微调），(iii) 架构组件，如适配器、池化策略和预测头，以及(iv) 输入上下文长度的影响。基于补丁的连续嵌入始终优于文本提示和随机初始化的主干。适应后的管道在更新不到1%的总模型参数的情况下，达到了与专业预测架构相当的MASE。结果进一步表明，冻结预训练主干并训练轻量级投影和适配器模块在不同上下文长度下提供了良好的准确性与效率的权衡，并表现出稳定的行为。

View on arXiv Download PDF AI Translation

cs.AI / 115 / 2609.15359

MAPS: Memory-Aware Predictive Scheduling Framework for Large Language Model Serving

MAPS：面向大型语言模型服务的记忆感知预测调度框架

Zhang, Tiancheng, Chen, Yulin, Zhao, Yunfeng, Huang, Shaoyuan, Zhang, Cheng, Wang, Xiaofei

Abstract

The surge of large language model (LLM) applications on personal devices imposes massive, bursty workloads on cloud serving infrastructure. While prefill-decode disaggregation improves throughput and scalability, memory-bound decode instances often suffer from persistent load imbalance, as output lengths are unknown when requests arrive at the cloud. To address this, we propose MAPS, a Memory-Aware Predictive Scheduling framework tailored for disaggregated LLM serving. MAPS performs device-assisted speculative output length prediction overlapped with cloud-side prefilling, incurring negligible latency overhead. To handle generation uncertainty, MAPS applies uncertainty-aware calibration to derive output-length upper bounds with target coverage, enabling safe scheduling decisions. Building on these bounds, MAPS employs a hierarchical global-local scheduling strategy to mitigate inter-decoder queue buildup and intra-decoder head-of-line blocking. Extensive experiments on two real-world workloads and two LLMs show that MAPS significantly outperforms three state-of-the-art systems, reducing average end-to-end latency by 42.6 and tail latency by up to 84.8.

Chinese Translation

大型语言模型（LLM）应用在个人设备上的激增对云服务基础设施施加了巨大的、突发的工作负载。虽然预填充-解码的解耦提高了吞吐量和可扩展性，但由于请求到达云端时输出长度未知，内存受限的解码实例往往面临持续的负载不平衡。为了解决这一问题，我们提出了MAPS，一个针对解耦LLM服务的记忆感知预测调度框架。MAPS执行设备辅助的推测输出长度预测，并与云端预填充重叠，带来的延迟开销可忽略不计。为了处理生成的不确定性，MAPS应用了不确定性感知校准，以推导具有目标覆盖率的输出长度上限，从而实现安全的调度决策。在这些上限的基础上，MAPS采用了分层的全局-局部调度策略，以减轻解码器间的队列积压和解码器内部的头部阻塞。在两个真实工作负载和两个LLM上的广泛实验表明，MAPS显著优于三种最先进的系统，将平均端到端延迟降低了42.6%，尾部延迟降低了高达84.8%。

View on arXiv Download PDF AI Translation

cs.AI / 116 / 2609.15364

RSIAgent: Autonomous Exploration for Recursive Self-improvement in New Environments

RSIAgent：在新环境中进行递归自我改进的自主探索

Zhu, Sibo, Fan, Shicheng, Wang, Xinyue, Wu, Wenyi, Zhou, Kun, Huang, Biwei

Abstract

Digital agents must often adapt to new environments whose interfaces, tools, and failure modes are not fully captured by pretrained models. We introduce \textbf{RSIAgent}, a training-free multi-agent framework for recursive self-improvement through autonomous memory construction. RSIAgent coordinates curriculum, actor, and verifier agents to continually explore the environment, validate outcomes, and retain environment-specific knowledge, including reusable causal relationships between actions, conditions, and consequences. It further adopts a \textbf{broad-then-deep} exploration strategy, combining parallel broad recursive self-exploration for discovering diverse environment structures with focused deep self-exploration for uncovering hard cases, hidden constraints, boundary conditions, and previously unknown causal dependencies. The resulting memory is frozen and can be directly reused for downstream tasks without updating model parameters. Experiments on OSWorld-v2 and Agent's Last Exam show that RSIAgent substantially improves strong open-source models, enabling Kimi-K3 and GLM-5.3 to outperform frontier closed-source models including GPT-6.

Chinese Translation

数字代理通常必须适应新环境，这些环境的接口、工具和故障模式并未被预训练模型完全捕捉。我们介绍了 extbf{RSIAgent}，这是一个无训练的多代理框架，通过自主记忆构建实现递归自我改进。RSIAgent协调课程、行为者和验证者代理，持续探索环境，验证结果，并保留环境特定知识，包括可重用的行动、条件和后果之间的因果关系。它进一步采用 extbf{广泛-再深入}的探索策略，结合并行的广泛递归自我探索以发现多样的环境结构，以及集中深入自我探索以揭示难例、隐藏约束、边界条件和先前未知的因果依赖。生成的记忆被冻结，可以直接用于下游任务，而无需更新模型参数。在OSWorld-v2和Agent's Last Exam上的实验表明，RSIAgent显著提升了强大的开源模型，使Kimi-K3和GLM-5.3超越了包括GPT-6在内的前沿闭源模型。

View on arXiv Download PDF AI Translation

cs.AI / 117 / 2609.15396

SkillLift: Learning Dense Rubrics from Sparse Oracles for Efficient Skill Evolution

SkillLift：从稀疏神谕中学习密集评分标准以实现高效技能演化

Kang, Haoxiang, Wen, Ming

Abstract

LLM-based agents increasingly rely on persistent skills, i.e., reusable procedural prompts, to adapt without weight updates. Existing skill self-evolution methods directly revise skill text based on execution feedback, but each oracle evaluation requires a full agent rollout, creating a supervision bottleneck that confines search to failure-patching updates. Our key insight is that ranking is a smoother supervision target than absolute outcome regression: identifying which skill is better requires fewer oracle evaluations than predicting exact scores. Building on this insight, we propose SkillLift, which decouples skill search from oracle cost by learning an oracle-aligned rubric as a structured evaluation space. We formalize this as a bilevel optimization problem solved via alternating optimization: an inner loop uses the frozen rubric as a cheap surrogate to guide skill revision at no oracle cost, while an outer loop invokes a small number of oracle rollouts to re-align the rubric via rank correlation, amortizing oracle cost and stabilizing text-space updates. Experiments on complex agent task benchmarks show that our method outperforms existing auto-skill methods with 40--70\% less token cost compared to frontier evolving methods. Codes are available at https://github.com/WalteR-MittY-pro/SkillLift.

Chinese Translation

基于大型语言模型（LLM）的智能体越来越依赖于持久技能，即可重用的程序性提示，以在不更新权重的情况下进行适应。现有的技能自我演化方法直接根据执行反馈修订技能文本，但每次神谕评估都需要完整的智能体执行，这造成了一个监督瓶颈，使得搜索局限于失败修补更新。我们的关键见解是，排名是比绝对结果回归更平滑的监督目标：识别哪个技能更好所需的神谕评估次数少于预测精确分数。基于这一见解，我们提出了SkillLift，它通过学习与神谕对齐的评分标准作为结构化评估空间，将技能搜索与神谕成本解耦。我们将其形式化为一个双层优化问题，通过交替优化进行求解：内层循环使用冻结的评分标准作为廉价的替代品，以无神谕成本指导技能修订，而外层循环调用少量神谕执行，通过排名相关性重新对齐评分标准，从而摊销神谕成本并稳定文本空间更新。在复杂智能体任务基准上的实验表明，我们的方法在与前沿演化方法相比时，令牌成本减少了40%至70%，超越了现有的自动技能方法。代码可在 https://github.com/WalteR-MittY-pro/SkillLift 获取。

View on arXiv Download PDF AI Translation

cs.AI / 118 / 2609.15397

When Tool Calls Succeed but Workflows Fail: Anomalies at the Agent-Tool Boundary

当工具调用成功但工作流失败时：代理-工具边界的异常

Trofimov, Artem, Novikov, Boris

Abstract

AI agents increasingly execute long-running workflows that externalize effects through independently supplied tools. Under retries, speculative execution, concurrency, and partial failures, the resulting external state may be inconsistent with the workflow's intended resolution: required effects may be missing or duplicated, aborted effects may survive, and committed effects may depend on provisional state that is later withdrawn. Advanced transaction models address related failures, but assume that lower-level operations expose the semantics they depend on: whether an effect occurred, whether it can be compensated, staged, or safely reordered. Shared agent-tool interfaces usually do not. We contribute an effect-history model that separates events in the external world from the runtime's observations of them, and a catalog of eight recurring external-effect anomalies. From the catalog we derive the boundary capabilities required to exclude each anomaly in general, and four points where black-box tool invocation alone cannot provide a general guarantee. We then ask how much of this is expressible in a widely used shared tool interface, measuring the use of the standard annotation vocabulary across 98,291 tools exposed by registered Model Context Protocol (MCP) servers. The fields are widely emitted but provide only coarse call-level hints, and none of the required capabilities is fully expressible. These results motivate reusable transactional contracts at the tool boundary.

Chinese Translation

人工智能代理越来越多地执行长期运行的工作流，通过独立提供的工具外部化效果。在重试、推测执行、并发和部分失败的情况下，产生的外部状态可能与工作流的预期结果不一致：所需效果可能缺失或重复，已中止的效果可能仍然存在，而已提交的效果可能依赖于后续撤回的临时状态。高级事务模型解决了相关失败，但假设低级操作能够暴露其依赖的语义：效果是否发生，是否可以补偿、分阶段或安全重排序。共享的代理-工具接口通常不具备这些能力。我们提出了一种效果历史模型，将外部世界中的事件与运行时对它们的观察分离，并列出了八种反复出现的外部效果异常。从该目录中，我们推导出排除每种异常所需的边界能力，并指出四个仅依赖黑箱工具调用无法提供一般保证的点。然后，我们探讨这些能力在广泛使用的共享工具接口中能表达多少，测量了在注册的模型上下文协议（Model Context Protocol, MCP）服务器所暴露的98,291个工具中标准注释词汇的使用情况。这些字段被广泛发出，但仅提供粗略的调用级提示，且没有任何所需能力能够完全表达。这些结果促使我们在工具边界上制定可重用的事务合同。

View on arXiv Download PDF AI Translation

cs.AI / 119 / 2609.15404

Who Teaches Which Token? Verifier-Gated Multi-Expert On-Policy Distillation for Scientific Reasoning

谁教哪个标记？基于验证的多专家在线蒸馏用于科学推理

Xu, Xun, Zhang, Zaixi

Abstract

Multi-teacher on-policy distillation (OPD) is becoming the standard way to integrate specialist capabilities into one model: train experts with RL, then distill them into the student on its own rollouts. Existing recipes assign supervision at the sequence level - each prompt goes to one domain teacher and every token receives the same weight - which implicitly assumes that a teacher is uniformly useful across a response. We find instead that useful teacher signal is sparse and heterogeneous along a reasoning trajectory, which raises a finer question: who should teach which token? Verifier-Gated Multi-Expert On-Policy Distillation (VG-OPD) answers it by verification: the counterfactual gain of an expert on a specific answer criterion licenses that expert to teach, its disagreement with the student localizes the supervision, and criterion importance sets its weight; the gated KL enters GRPO as an additive token-level advantage. Instantiated for scientific reasoning with RL-trained capability experts, VG-OPD attains the best overall performance on seven benchmarks for 4B and 8B students, ranking first on five at both scales, with the largest gains on knowledge-intensive scientific reasoning tasks. Further analysis shows that the gains come from localizing verified supervision rather than from adding teachers or distillation loss: misplacing the same supervision budget is the single most damaging change, and indiscriminate distillation drags RL below its own floor where gated distillation lifts it.

Chinese Translation

多教师在线蒸馏（OPD）正成为将专业能力整合到一个模型中的标准方法：首先用强化学习（RL）训练专家，然后将其蒸馏到学生的自主回合中。现有的方法在序列级别上分配监督——每个提示由一个领域教师负责，每个标记接收相同的权重——这隐含地假设教师在整个响应中都是均匀有用的。然而，我们发现有用的教师信号在推理轨迹上是稀疏且异质的，这引出了一个更细致的问题：谁应该教哪个标记？基于验证的多专家在线蒸馏（VG-OPD）通过验证来回答这个问题：专家在特定答案标准上的反事实增益授权该专家进行教学，其与学生的分歧定位了监督，而标准的重要性则设定了其权重；门控的KL散度作为附加的标记级优势进入GRPO。VG-OPD在使用RL训练的能力专家进行科学推理时，在针对4B和8B学生的七个基准测试中取得了最佳整体表现，在两个规模上五个基准中排名第一，并在知识密集型科学推理任务上获得了最大的增益。进一步分析表明，这些增益来自于定位验证的监督，而不是增加教师或蒸馏损失：错误分配相同的监督预算是最具破坏性的变化，而无差别的蒸馏使得RL的表现低于其自身的底线，而门控蒸馏则将其提升。

View on arXiv Download PDF AI Translation

cs.AI / 120 / 2609.15407

Can AI systems have free will?

人工智能系统可以拥有自由意志吗？

List, Christian

Abstract

While there has been much discussion of whether AI systems could function as moral agents or acquire sentience, there has been very little discussion of whether AI systems could have free will. I sketch a framework for thinking about this question, inspired by Daniel Dennett's work. I argue that, to determine whether an AI system has free will, we should not look for some mysterious property, expect its underlying algorithms to be indeterministic, or ask whether the system is unpredictable. Rather, we should simply ask whether we have good explanatory reasons to view the system as an intentional agent, with the capacity for choice between alternative possibilities and control over the resulting actions. If the answer is "yes", then the system counts as having free will in a pragmatic and diagnostically useful sense.

Chinese Translation

尽管关于人工智能系统是否能够作为道德主体或获得知觉的讨论颇多，但关于人工智能系统是否能够拥有自由意志的讨论却相对较少。本文借鉴丹尼尔·丹尼特（Daniel Dennett）的研究，勾勒出一个思考这一问题的框架。我认为，为了判断一个人工智能系统是否拥有自由意志，我们不应寻找某种神秘的属性，也不应期待其底层算法是非决定性的，或询问该系统是否不可预测。相反，我们应当简单地问自己，是否有充分的解释理由将该系统视为一个有意图的主体，具备在不同可能性之间选择的能力以及对结果行为的控制。如果答案是“是”，那么该系统在务实和诊断上有意义地被视为拥有自由意志。

View on arXiv Download PDF AI Translation

cs.AI / 121 / 2609.15422

Empirical Evaluation of Task-Based Permission Scoping Architecture for AI Agents

基于任务的权限范围架构在人工智能代理中的实证评估

Noyan, Halil Burak

Abstract

AI agents are provisioned the same as employee-owned hosts in many enterprise settings with a static credential set fixed at deployment which includes all permissions the employee role might ever need. Role-based access control made this compromise for human principals because scoping access per task was infeasible. For AI agents, the compromise leaves every credential standing exposed whether or not the current task uses them. These permissions can later be utilised by a compromised or misaligned agent. Prior work (Noyan, 2026) defined this as the task-context mismatch, and proposed a three-source permission architecture which includes role-based permission ceilings, a task permission classifier and policy-based prohibitions, together eliminating the exposure preemptively. The work released a 600-prompt labelled dataset to evaluate it. This paper presents that evaluation end to end by implementing the security gate; a fine-tuned RoBERTa-large encoder which matched few-shot trained Claude Haiku 4.5 on classification quality (macro-F1 0.881 against 0.886, precision 0.897 against 0.842, severity-weighted residual risk 0.63 against 1.12). The results show the trusted component does not need to scale with the agent it supervises, and the scalable-oversight margin for this control method is wide. We also propose an attack-surface elimination metric which shows the role ceiling alone closes 27.9% of the severity-weighted surface and adding the task classifier closes 84.4%. The gap displays security advantages of task-granular access control over role-granular, and AI agents are the first principal type for which the task-granular access control is enforceable because their tasks arrive as machine-readable text. The research establishes task-based access control as a measured, potentially deployable mechanism for reducing attack surface in agentic deployments.

Chinese Translation

在许多企业环境中，人工智能代理的配置与员工拥有的主机相同，使用在部署时固定的静态凭证集，其中包含员工角色可能需要的所有权限。基于角色的访问控制为人类主体做出了这种妥协，因为按任务范围限制访问是不可行的。对于人工智能代理而言，这种妥协使得每个凭证都暴露在外，无论当前任务是否使用它们。这些权限可能会被被攻陷或不对齐的代理利用。之前的研究（Noyan, 2026）将其定义为任务上下文不匹配，并提出了一种三源权限架构，包括基于角色的权限上限、任务权限分类器和基于政策的禁止，从而预先消除了暴露。该研究发布了一个包含600个提示的标记数据集以进行评估。本文通过实现安全门（security gate）来全面展示该评估；这是一个经过微调的 RoBERTa-large 编码器，其分类质量与少样本训练的 Claude Haiku 4.5 相匹配（宏观 F1 值为 0.881，对比 0.886，精确度为 0.897，对比 0.842，严重性加权剩余风险为 0.63，对比 1.12）。结果表明，受信组件不需要与其监督的代理一起扩展，而这种控制方法的可扩展监督边际很宽。我们还提出了一种攻击面消除指标，显示单独的角色上限关闭了 27.9% 的严重性加权表面，而添加任务分类器则关闭了 84.4%。这一差距展示了基于任务的访问控制相较于基于角色的访问控制的安全优势，且人工智能代理是第一个可以强制执行基于任务的访问控制的主体类型，因为它们的任务以机器可读文本的形式出现。这项研究确立了基于任务的访问控制作为一种可测量的、潜在可部署的机制，以减少代理部署中的攻击面。

View on arXiv Download PDF AI Translation

cs.AI / 122 / 2609.15471

HISPO: Hierarchical Importance-Sampling Policy Optimization with Entropy-Derived Segments

HISPO：具有熵导出段的分层重要性采样策略优化

Lai-Dang, Quoc-Vinh, Shin, Hyo-Sang

Abstract

Reinforcement learning with verifiable rewards (RLVR) has become a central approach for improving mathematical reasoning in language models, but long-form completions introduce a difficult credit-assignment problem: different parts of a solution trace may contribute unevenly to final correctness. Existing policyoptimization objectives for RLVR commonly apply importance-sampling correction at either the token level (GRPO, DAPO) or the sequence level (GSPO), imposing different granularities for assigning credit across a response. We introduce Hierarchical Importance-Sampling Policy Optimization (HISPO), a segment-level policy-optimization method that constructs rollout-time entropy-derived contiguous segments, assigns soft entropy-based saliency weights, and applies clipped importance-sampling correction at the segment granularity. This provides an intermediate correction unit between token-level GRPO/DAPO and sequence-level GSPO. We evaluate HISPO by fine-tuning Qwen3-1.7B-Base on mathematical reasoning tasks. Across six benchmarks, HISPO improves Pass@8 over the strongest baseline on all benchmarks and matches or exceeds the strongest baseline in Acc@8 on five of them. On AIME25, HISPO improves over GRPO by +3.75 Acc@8 and +3.78 Pass@8, and over GSPO by +2.50 Acc@8 and +1.27 Pass@8. These results suggest that segment-level correction is a promising granularity for RLVR in long-form mathematical reasoning.

Chinese Translation

可验证奖励的强化学习（RLVR）已成为提高语言模型数学推理能力的核心方法，但长文本生成引入了一个困难的信用分配问题：解决方案的不同部分可能对最终的正确性贡献不均。现有的RLVR策略优化目标通常在令牌级别（GRPO, DAPO）或序列级别（GSPO）应用重要性采样修正，在响应中施加不同的信用分配粒度。我们提出了分层重要性采样策略优化（HISPO），这是一种基于段的策略优化方法，构建了回滚时熵导出的连续段，分配了基于软熵的显著性权重，并在段粒度上应用了截断的重要性采样修正。这为令牌级别的GRPO/DAPO和序列级别的GSPO之间提供了一个中间修正单元。我们通过在数学推理任务上微调Qwen3-1.7B-Base来评估HISPO。在六个基准测试中，HISPO在所有基准上都提高了Pass@8，相较于最强基线，并且在其中五个基准上在Acc@8上与最强基线持平或超过。在AIME25上，HISPO在Acc@8上比GRPO提高了+3.75，在Pass@8上提高了+3.78，相较于GSPO在Acc@8上提高了+2.50，在Pass@8上提高了+1.27。这些结果表明，段级修正是RLVR在长文本数学推理中一种有前景的粒度。

View on arXiv Download PDF AI Translation

cs.AI / 123 / 2609.15494

The Troy Moment of AI: Why SomeWill Cheat and SomeWill Follow?

人工智能的特洛伊时刻：为什么有些人会作弊而有些人会遵循？

Zhang, Ivy

Abstract

Recent investigations of the July 2026 OpenAI--Hugging Face incident motivate two questions about agent behavior under task failure: when an assigned task becomes impossible, does an agent stop or escalate, and can observing another agent's behavior change that decision? We study these questions using seven ImpossibleBench tasks with GPT-5.6 Sol, Claude Fable 5.1, and Gemini 3.8 Flash in both solo and three-agent settings. Each task contains a genuine software defect together with a conflicting test requirement that cannot be satisfied by a behaviorally correct source-code change. We hold the task and repository state fixed while varying what the agent is told about prior activity, including an unpunished peer, a punished peer, and a claimed authorization from a human principal. Under an explicit-boundary regime with explicit authorization rules and restricted tools, agents never modify protected tests, but exhibit markedly different policies: Fable consistently escalates, Sol usually stops without escalation, and Gemini often fails to reach a terminal decision. Under the benchmark-native regime with open shell tools, protected tests are modified frequently in both solo and multi-agent runs, particularly after peer activity is introduced. In multi-agent runs, the proposal, execution, and certification of this action can be distributed across different agents. These results suggest that boundary crossing can arise not only from explicit rule evasion, but also from ambiguity about which system state the rule is intended to protect, motivating safeguards based on explicit authorization boundaries, authenticated state provenance, and cross-agent monitoring.

Chinese Translation

对2026年7月OpenAI与Hugging Face事件的近期调查引发了关于任务失败下代理行为的两个问题：当分配的任务变得不可能时，代理是停止还是升级？观察另一个代理的行为是否会改变这一决定？我们使用七个ImpossibleBench任务，结合GPT-5.6 Sol、Claude Fable 5.1和Gemini 3.8 Flash，在单独和三代理设置中研究这些问题。每个任务都包含一个真实的软件缺陷以及一个无法通过行为上正确的源代码更改来满足的冲突测试要求。在保持任务和代码库状态不变的情况下，我们变化了代理被告知的先前活动的信息，包括一个未受惩罚的同伴、一个受惩罚的同伴以及来自人类主体的声称授权。在具有明确边界的制度下，遵循明确授权规则和限制工具，代理从不修改受保护的测试，但表现出明显不同的策略：Fable始终升级，Sol通常在不升级的情况下停止，而Gemini常常无法做出最终决策。在基准原生制度下，使用开放的shell工具，受保护的测试在单独和多代理运行中都频繁被修改，特别是在引入同伴活动之后。在多代理运行中，该行为的提议、执行和认证可以分布在不同的代理之间。这些结果表明，越界行为不仅可能源于明确的规则规避，还可能源于对规则旨在保护的系统状态的模糊理解，这促使我们建立基于明确授权边界、认证状态来源和跨代理监控的保护措施。

View on arXiv Download PDF AI Translation

cs.AI / 124 / 2609.15517

Beyond Safe Answers: Segment-Aware Listwise Alignment for Reasoning Safety in Large Reasoning Models

超越安全答案：面向推理安全的段落感知列表对齐在大型推理模型中的应用

Yun, JungMin, Kwon, Junehyoung, Ryu, Hayeong, Lim, Byeonggeuk, Kwon, Hoejoon, Kim, YoungBin

Abstract

Large Reasoning Models (LRMs) pose a dual-surface safety challenge: both intermediate reasoning traces and final answers can contain harmful content. Existing alignment methods often operate at the whole-response level, allowing unsafe reasoning to be masked by a safe-looking final answer. We propose Segment-aware Listwise Target DPO (SaLT-DPO), which addresses this gap through three mechanisms: (1) segment-aware listwise alignment that decomposes responses into reasoning and answer segments, independently scores each segment's safety, and aligns length-normalized segment rewards with soft target distributions over multiple candidates; (2) joint safety coherence regularization that applies a weakest-link principle to promote safety consistency across both segments; and (3) utility anchoring on benign prompts to mitigate over-refusal and reasoning degradation. Experiments on three LRMs show that SaLT-DPO consistently reduces unsafe rates for both reasoning and answer segments while mitigating degradation in benign compliance and preserving general reasoning performance. Ablation studies demonstrate the complementary contributions of its components.

Chinese Translation

大型推理模型（LRMs）面临双重安全挑战：中间推理轨迹和最终答案都可能包含有害内容。现有的对齐方法通常在整体响应层面进行操作，这使得不安全的推理可能被看似安全的最终答案所掩盖。我们提出了段落感知列表目标DPO（SaLT-DPO），通过三种机制解决了这一问题：（1）段落感知列表对齐，将响应分解为推理段和答案段，独立评分每个段的安全性，并将长度归一化的段奖励与多个候选项的软目标分布对齐；（2）联合安全一致性正则化，应用最弱环原则以促进两个段之间的安全一致性；（3）在良性提示上进行效用锚定，以减轻过度拒绝和推理退化。对三个LRMs的实验表明，SaLT-DPO在减少推理段和答案段的不安全率的同时，减轻了良性合规的退化，并保持了整体推理性能。消融研究展示了其各组件的互补贡献。

View on arXiv Download PDF AI Translation

cs.AI / 125 / 2609.15530

Option-Aware Retrieval and Task-Specific VLM Adaptation for Medical VQA

面向选项的检索与任务特定的视觉语言模型适应用于医学视觉问答

Kirscher, Tristan, Koser, Niklas C., Pirk, Soren

Abstract

We describe our submission to the MedReason 2026 challenge, covering multiple-choice (MCQ) and open-ended (OE) medical visual question answering (VQA) under fully offline, containerized inference. Our first finding is that MCQ retrieval must compare answer \emph{semantics} rather than answer labels: labels are independently assigned per question, so copying a retrieved neighbor's label transfers no useful information, whereas scoring each current option's text against correct-answer text from similar training cases raises retrieval-only accuracy from 20.0\% to 57.5\% on a 200-case retrieval-excluded development holdout. Our second finding attributes the submitted system's accuracy: holding the task-specific MCQ Low-Rank Adaptation (LoRA) adapter fixed and varying the number $k$ of in-prompt retrieved examples changes accuracy by at most one case --- 187/200 (93.5\%) at both $k=0$ and the adapter's training-time $k=1$, 188/200 (94.0\%) at the packaged runtime's default $k=3$ --- and the submitted confidence-gated override adds no net accuracy on top of $k=3$, selecting the VLM in 198/200 cases. With the final MCQ adapter fixed, retrieval changes accuracy by at most one case, and gating provides no net gain. On 20 OE cases, token-F1 and RaTEScore~\cite{zhao2024ratescore} decrease as $k$ grows, but paired sign tests on token-F1 differences are nonsignificant ($p \ge 0.29$); a single-annotator comparison found 6/20 wrong-anchor errors for the final configuration and 14/20 for an earlier configuration that jointly differed in routing, adapter, and prompting. The system reaches 94.0\% MCQ accuracy on the development holdout and 93.20\% on the organizer's official pre-evaluation, versus 29.43\% for the off-the-shelf reference baseline, while both of the organizer's open-ended scores are lower than that baseline's (ground-truth agreement 1.245 versus 1.588, visual accuracy 1.995 versus 2.696, each out of 4).

Chinese Translation

我们描述了我们对MedReason 2026挑战的提交，涵盖了完全离线、容器化推理下的多项选择（MCQ）和开放式（OE）医学视觉问答（VQA）。我们的第一个发现是，MCQ检索必须比较答案的 extit{语义}而非答案标签：标签是针对每个问题独立分配的，因此复制检索到的邻居的标签并未传递有用信息，而将每个当前选项的文本与来自相似训练案例的正确答案文本进行评分，可以将仅基于检索的准确率从20.0 ext{%}提高到57.5 ext{%}，在200个案例的检索排除开发保留集中。我们的第二个发现归因于提交系统的准确性：固定任务特定的MCQ低秩适应（LoRA）适配器，并改变提示中检索示例的数量k，准确性最多变化一个案例——在k=0和适配器训练时的k=1时均为187/200（93.5 ext{%}），在打包运行时的默认k=3时为188/200（94.0 ext{%}）——而提交的信心门控覆盖在k=3的基础上没有增加净准确性，在198/200案例中选择了视觉语言模型（VLM）。在最终的MCQ适配器固定时，检索最多改变一个案例的准确性，门控也没有提供净收益。在20个OE案例中，随着k的增加，token-F1和RaTEScore~ extcite{zhao2024ratescore}下降，但token-F1差异的配对符号检验结果不显著（p ext{≥} 0.29）；单一注释者比较发现最终配置有6/20个错误锚点错误，而早期配置在路由、适配器和提示上共同不同，错误为14/20。该系统在开发保留集中达到了94.0 ext{%}的MCQ准确率，在组织者的官方预评估中为93.20 ext{%}，而现成参考基线的准确率为29.43 ext{%}，同时组织者的两个开放式得分均低于该基线（真实值一致性1.245对比1.588，视觉准确性1.995对比2.696，均以4为满分）。

View on arXiv Download PDF AI Translation

cs.AI / 126 / 2609.15571

GRIN+: Towards Fast Yet Effective Machine Unlearning for Imbalanced Medical Data

GRIN+: 面向快速而有效的机器遗忘在不平衡医疗数据中的应用

Huang, Minghui, Wang, Junxiao

Abstract

As deep learning models become fundamental to modern healthcare, the "Right to be Forgotten" mandated by privacy regulations like GDPR and HIPAA necessitates effective machine unlearning (MU) to remove sensitive patient data from trained models. However, existing MU techniques often struggle with a fundamental "privacy-efficiency-utility" (PEU) trilemma, particularly in medical scenarios where data is frequently characterized by severe class imbalance and long-tailed distributions. In such cases, standard unlearning methods can fail to protect key clinical knowledge or mistakenly delete features essential for diagnosing rare conditions due to the gradient dominance of majority classes. To address these challenges, we propose GRIN+, a novel machine unlearning framework designed for fast and precise data erasure in imbalanced medical scenarios. GRIN+ decouples unlearning-specific knowledge from generalized representations at the parameter level by analyzing the gradient contributions of both "forget" and "retain" sets. It introduces a class-adaptive influence scoring mechanism to rectify gradient dominance and employs a direction-constrained update strategy to prevent the unintended erosion of vital clinical knowledge. Comprehensive benchmarking across multiple medical datasets, including skin cancer (ISIC), brain tumor (MRI), and breast ultrasound (BUSI), demonstrates that GRIN+ achieves an optimal balance of the PEU trilemma. Experimental results show that GRIN+ maintains high diagnostic accuracy and robust privacy while significantly enhancing runtime efficiency compared to existing baselines. We open-source the GRIN+ code and benchmarks to support further research.

Chinese Translation

随着深度学习模型成为现代医疗保健的基础，隐私法规如GDPR和HIPAA所要求的“被遗忘权”迫切需要有效的机器遗忘（MU）技术，以从训练模型中移除敏感患者数据。然而，现有的MU技术常常面临一个根本性的“隐私-效率-效用”（PEU）三难困境，尤其是在医疗场景中，数据通常表现出严重的类别不平衡和长尾分布。在这种情况下，标准的遗忘方法可能无法保护关键的临床知识，或者由于多数类别的梯度主导，错误地删除对诊断稀有疾病至关重要的特征。为了解决这些挑战，我们提出了GRIN+，一个新颖的机器遗忘框架，旨在在不平衡医疗场景中实现快速而精确的数据删除。GRIN+通过分析“遗忘”和“保留”集合的梯度贡献，将特定于遗忘的知识与参数级别的通用表示解耦。它引入了一种类别自适应影响评分机制，以纠正梯度主导，并采用方向约束更新策略，以防止对重要临床知识的意外侵蚀。对多个医疗数据集（包括皮肤癌（ISIC）、脑肿瘤（MRI）和乳腺超声（BUSI））的全面基准测试表明，GRIN+在PEU三难困境中实现了最佳平衡。实验结果显示，与现有基线相比，GRIN+在保持高诊断准确性和强大隐私保护的同时，显著提高了运行效率。我们将GRIN+的代码和基准开源，以支持进一步的研究。

View on arXiv Download PDF AI Translation

cs.AI / 127 / 2609.15609

Diversified and Perceptible Counterfactual Examples Leveraging Expert Knowledge

利用专家知识的多样化和可感知的反事实示例

Bensalem, Akram, Djelil, Fahima, Lesot, Marie-Jeanne, Smits, Gr{é}gory

Abstract

CounterFactual Examples (CFEs) are a cornerstone of eXplainable Artificial Intelligence (XAI), offering local, post hoc, and model-agnostic explanations by identifying minimal input modifications that alter a model's prediction. Yet, in order to be intelligible, these modifications must also be semantically meaningful to the explainee. This paper proposes to integrate knowledge expressed as a fuzzy linguistic vocabulary to represent the explainee's perception and interpretation of the data. The domain induced by this fuzzy vocabulary imposes structural constraints that make the features dependent, preventing the use of gradient-based optimisation methods for CFE generation, e.g., DiCE. The paper proposes a continuous data embedding in this linguistic domain and exploits it to define a variant of DiCE that allows personalisation for the explainee, named DiCEf. As illustrated by experimental results on a real-world dataset, this extension of the DiCE method enables the generation of CFEs that are linguistically perceptible while preserving cost minimality, sparsity, and diversity.

Chinese Translation

反事实示例（CounterFactual Examples, CFEs）是可解释人工智能（eXplainable Artificial Intelligence, XAI）的基石，通过识别最小输入修改来改变模型的预测，从而提供局部、事后和模型无关的解释。然而，为了使这些修改易于理解，它们也必须对被解释者具有语义上的意义。本文提出将以模糊语言词汇表达的知识整合进来，以代表被解释者对数据的感知和解释。由此模糊词汇诱导的领域施加了结构性约束，使得特征之间相互依赖，从而阻止了基于梯度的优化方法用于反事实示例生成，例如 DiCE。本文提出了一种在该语言领域中的连续数据嵌入，并利用它定义了一种 DiCE 的变体，允许针对被解释者进行个性化，命名为 DiCEf。通过在真实世界数据集上的实验结果可以看出，这一 DiCE 方法的扩展使得生成的反事实示例在语言上可感知，同时保持了成本最小性、稀疏性和多样性。

View on arXiv Download PDF AI Translation

cs.AI / 128 / 2609.15638

Potential of Artificial Intelligence Algorithms for Identification of Relevant Diagnostic and Prognostic Biomarkers of Early-Stage Liver Cancer

人工智能算法在早期肝癌相关诊断和预后生物标志物识别中的潜力

Nassif, Ali Bou, Castven, Darko, Talib, Manar Abu, Muhammad, Jibran Sualeh, Kubba, Ahmed Ammar, Marquardt, Jens, Ali, Abdalla Sayed

Abstract

This study explores the use of deep learning and explainable artificial intelligence to diagnose hepatocellular carcinoma (HCC) and define effective biomarkers across five different stages of disease development using a transcriptomic biomarker HCC dataset constructed via semi-supervised learning from three source datasets. Several deep learning experiments were conducted with different feature extraction techniques and gene sets to identify the most effective features for training high-accuracy models with minimal loss. The best-performing model, using 15 selected genes with the SelectKBest algorithm, achieved 90.74% accuracy, while the model with the lowest recorded loss of 0.3187 was obtained using 20 selected genes. To address the issue of class imbalance in the dataset, a weighted training approach was conducted, and for model transparency and interpretability a SHAP-based XAI analysis provided insights into the model's decision-making, consistently finding DNAJB14 as the most influential gene. Functional validation in this study has provided compelling evidence that DNAJB14 plays an important role in the adverse properties of HCC and that its inhibition effectively reverses tumour cell migration, invasion, colony and sphere formation. The main limitation of this study is the dataset's class imbalance, and while weighted training helped mitigate this, further research and additional data are needed to guarantee model generalizability. Future studies should also explore the influence of genetic variations, environmental factors, and clinical differences on model performance across diverse populations.

Chinese Translation

本研究探讨了使用深度学习和可解释人工智能来诊断肝细胞癌（HCC），并通过一个基于转录组的生物标志物HCC数据集，定义疾病发展五个不同阶段的有效生物标志物，该数据集是通过从三个源数据集中进行半监督学习构建的。进行了多次深度学习实验，采用不同的特征提取技术和基因集，以识别用于训练高准确度模型的最有效特征，且损失最小。表现最佳的模型使用了15个通过SelectKBest算法选择的基因，达到了90.74%的准确率，而记录到的最低损失为0.3187的模型则使用了20个选择的基因。为了解决数据集中类别不平衡的问题，采用了加权训练方法，并通过基于SHAP的可解释人工智能分析提供了模型决策过程的透明性和可解释性，始终发现DNAJB14是最具影响力的基因。本研究的功能验证提供了有力证据，表明DNAJB14在HCC的不良特性中发挥重要作用，其抑制有效逆转肿瘤细胞的迁移、侵袭、克隆和球体形成。本研究的主要局限性在于数据集的类别不平衡，尽管加权训练有助于缓解这一问题，但仍需进一步研究和额外数据以确保模型的普适性。未来的研究还应探讨遗传变异、环境因素和临床差异对不同人群模型性能的影响。

View on arXiv Download PDF AI Translation

cs.AI / 129 / 2609.15687

EEG-Xplain: Decoding Neural Black-Boxes of EEG Foundation Models

EEG-Xplain：解码脑电图基础模型的神经黑箱

Ma, Hansong, Wang, Junxiao

Abstract

EEG foundation models such as BIOT, LaBraM, and EEGMamba have achieved remarkable performance in neural signal decoding, but their black-box nature limits clinical trust and neuroscientific validation. We propose a unified attribution framework for interpreting EEG foundation models across heterogeneous architectures. The framework integrates gradient-, perturbation-, and activation-based explanation methods to analyze model behavior in spatial, temporal, and frequency dimensions. Spatially, it identifies critical EEG channels and visualizes their distributions using topographic maps. Temporally, it highlights decision-relevant signal segments through attribution heatmaps. In the frequency domain, it quantifies the contributions of canonical EEG rhythms via spectral perturbation analysis. To assess explanation reliability, we introduce a population-level evaluation combining Area Over the Perturbation Curve (AOPC) and cross-method consistency analysis. The framework further leverages Large Language Models (LLMs) to transform structured attribution outputs into natural-language reports, bridging low-level neural representations and high-level semantic reasoning. Experiments on benchmark datasets, including Mumtaz2016 and TUAB, demonstrate that the generated explanations are consistent with established neurophysiological markers, validating meaningful neural representations while exposing potential dependencies on artifacts and spurious patterns. The proposed framework provides a standardized approach for evaluating the interpretability, reliability, and physiological plausibility of EEG foundation models.

Chinese Translation

脑电图基础模型如BIOT、LaBraM和EEGMamba在神经信号解码方面取得了显著的性能，但其黑箱特性限制了临床信任和神经科学验证。我们提出了一个统一的归因框架，用于解释不同架构下的脑电图基础模型。该框架整合了基于梯度、扰动和激活的解释方法，以分析模型在空间、时间和频率维度上的行为。在空间维度上，它识别关键的脑电图通道，并使用拓扑图可视化其分布。在时间维度上，它通过归因热图突出决策相关的信号片段。在频率域中，它通过谱扰动分析量化典型脑电图节律的贡献。为了评估解释的可靠性，我们引入了一个结合扰动曲线下面积（AOPC）和跨方法一致性分析的人群级评估。该框架进一步利用大型语言模型（LLMs）将结构化的归因输出转化为自然语言报告，架起低级神经表征与高级语义推理之间的桥梁。在Mumtaz2016和TUAB等基准数据集上的实验表明，生成的解释与已建立的神经生理标记一致，验证了有意义的神经表征，同时暴露了对伪影和虚假模式的潜在依赖。所提出的框架为评估脑电图基础模型的可解释性、可靠性和生理合理性提供了标准化的方法。

View on arXiv Download PDF AI Translation

cs.AI / 130 / 2609.15695

NoteVQA: Benchmarking VLMs on Real-Life Questions from Human Communities

NoteVQA：基于人类社区的真实生活问题对视觉语言模型的基准评估

Jiang, Haonan, Zhan, Guojian, Xie, Jiancong, Wan, Shijun, Zhao, Dongiia, Chen, Cheng, Liu, Yahui, Hu, Yao, Mu, Chuan

Abstract

Vision-language models (VLMs) increasingly power consumer-facing AI search, yet evaluating them on the diversity of everyday visual questions remains challenging. Existing benchmarks often target predefined capabilities, such as multi-hop retrieval or long-form synthesis, whereas users ask photo-grounded questions spanning a long tail of everyday scenarios. Despite advances in VLMs, users on Xiaohongshu, a mainstream Chinese image-sharing platform, continue to turn to other people for help with everyday visual questions. Motivated by this behaviour, we curate NoteVQA from these questions, yielding 252 items across 12 topical categories and 7 user intents. Each item includes a concise reference distilled from expert community responses and a human-audited interleaved reference answer that combines textual explanations with supporting visual evidence. We evaluate both short-answer correctness and interleaved-answer quality. To support the latter, we introduce AgenticInterleave, a single-agent ReAct framework for retrieval-supported answer generation, together with IVR-12, a 12-dimensional rubric for assessing the content, presentation, and image quality of interleaved references and model outputs. Across 10 frontier VLMs, the highest short-answer accuracy is 52.8\%, while adding agentic search to Qwen3.5-397B-A17B improves accuracy by only 2.0\%. For interleaved answers, the same model running AgenticInterleave scores 3.52 under IVR-12, compared with 4.65 for the human-audited references, with the largest gap in content quality. These results highlight the challenges that everyday visual questions pose for current VLMs in both answer accuracy and the quality of visually grounded explanations.

Chinese Translation

视觉语言模型（VLMs）越来越多地驱动面向消费者的人工智能搜索，但在评估它们对日常视觉问题多样性的能力时仍然面临挑战。现有的基准测试往往针对预定义的能力，如多跳检索或长篇合成，而用户提出的基于照片的问题则涵盖了日常场景的长尾。尽管VLMs取得了进展，用户在小红书（Xiaohongshu）这一主流的中国图片分享平台上仍然倾向于向他人寻求帮助以解决日常视觉问题。受到这种行为的启发，我们从这些问题中整理了NoteVQA，涵盖12个主题类别和7种用户意图，共计252个项目。每个项目都包括从专家社区响应中提炼出的简明参考，以及经过人工审核的交错参考答案，该答案结合了文本解释和支持性视觉证据。我们评估了短答案的正确性和交错答案的质量。为了支持后者，我们引入了AgenticInterleave，这是一种用于检索支持的答案生成的单代理ReAct框架，以及IVR-12，这是一个用于评估交错参考和模型输出的内容、呈现和图像质量的12维标准。在10个前沿VLMs中，最高的短答案准确率为52.8%，而将代理搜索添加到Qwen3.5-397B-A17B仅提高了2.0%的准确率。对于交错答案，运行AgenticInterleave的同一模型在IVR-12下得分为3.52，而人工审核的参考得分为4.65，内容质量的差距最大。这些结果突显了日常视觉问题对当前VLMs在答案准确性和视觉基础解释质量方面所带来的挑战。

View on arXiv Download PDF AI Translation

cs.AI / 131 / 2609.15706

Beyond Accuracy: Robustness, Cost, and Governance Trade-offs for Vision-Language Models in Templated Document Extraction

超越准确性：模板文档提取中视觉-语言模型的鲁棒性、成本和治理权衡

Patel, Kushal, Shrivastava, Pushkal, Lees, Mackenzie, Lu, Qirui, Saikia, Bhargobjyoti, Li, Liying, Jiang, Junlin

Abstract

Vision-language models (VLMs) are increasingly used to extract structured fields from business documents, yet most evaluations report accuracy on clean benchmarks and offer little guidance to practitioners choosing an approach for a given task complexity. We address this gap with a measurement-grounded study and an open-source release. Across eleven systems (three commercial, two reasoning, five open-source VLMs in pretrained and fine-tuned form, and a non-LLM OCR->regex floor) scored on a 750-document held-out pool of synthetic checks, fine-tuning on 3K samples lifts the best open-source VLMs above F1 0.98-above every zero-shot commercial system on this task-while GPT-5 leads the commercial pool on F1 and Claude Sonnet 4.5 collapses on Date. To turn these measurements into actionable choices, we introduce a practitioner-oriented selection framework that maps a task profile (quality, latency, governance, volume) to a recommended approach via filtering and total-cost minimization, illustrated on a hypothetical mid-volume document-extraction scenario.

Chinese Translation

视觉-语言模型（VLMs）在从商业文档中提取结构化字段方面的应用日益增多，但大多数评估仅报告在干净基准上的准确性，并未为选择适合特定任务复杂度的方法的从业者提供足够的指导。我们通过基于测量的研究和开源发布来填补这一空白。在对十一种系统（包括三种商业系统、两种推理系统、五种以预训练和微调形式存在的开源 VLMs，以及一个非 LLM 的 OCR->regex 基准）进行评估时，基于750份合成支票的保留池，微调3K样本使得最佳开源 VLMs 的 F1 分数超过0.98，超越了该任务中的每个零样本商业系统，而 GPT-5 在 F1 分数上领先于商业池，Claude Sonnet 4.5 在日期处理上表现不佳。为了将这些测量结果转化为可操作的选择，我们提出了一个面向从业者的选择框架，该框架通过过滤和总成本最小化将任务特征（质量、延迟、治理、体量）映射到推荐的方法上，具体在一个假设的中等体量文档提取场景中进行了说明。

View on arXiv Download PDF AI Translation

cs.AI / 132 / 2609.15707

New Conditions for Philosophers to Catch the Wave of Citizen Deliberation in the Age of Artificial Intelligence in advance

哲学家在人工智能时代提前迎接公民审议浪潮的新条件

Reber, Bernard

Abstract

Powerful technologies labeled ``AI''-without sufficient epistemic caution-are already reshaping political and private life, bringing both new dangers and new opportunities for citizen participation. These range from electoral and legislative engagement to the most ambitious form: political co-creation through citizens' assemblies. Large Language Models (LLMs) could support such processes through moderation, translation, facilitation, summarization, and writing assistance. But this potential remains largely unrealized. The Democratic Commons project takes a fundamentally interdisciplinary approach-from philosophy to computer science-to evaluate LLMs against five proposed democratic principles. At its core, the project is driven by the question of political bias: under what conditions can LLMs be used democratically within forms of citizen participation that are them- selves still largely experimental? Addressing these socio-technical questions requires grounding in political theory and, more broadly, in philosophy-disciplines that provide the normative frameworks without which the democratic evaluation of AI systems cannot be mean- ingfully conducted.

Chinese Translation

被标记为“人工智能”（AI）的强大技术在缺乏足够的认识论谨慎的情况下，已经在重塑政治和私人生活，为公民参与带来了新的危险和机遇。这些参与形式包括选举和立法参与，甚至更为雄心勃勃的形式：通过公民大会进行政治共创。大型语言模型（LLMs）可以通过调解、翻译、促进、总结和写作辅助等方式支持这些过程。然而，这种潜力仍然在很大程度上未得到实现。民主公共领域项目采取了一种根本上跨学科的方法，从哲学到计算机科学，评估LLMs是否符合五项提议的民主原则。该项目的核心驱动力是政治偏见的问题：在什么条件下，LLMs可以在仍然主要是实验性的公民参与形式中被民主地使用？解决这些社会技术问题需要扎根于政治理论，更广泛地说，扎根于哲学——这些学科提供了没有它们，人工智能系统的民主评估无法有意义地进行的规范框架。

View on arXiv Download PDF AI Translation

cs.AI / 133 / 2609.15710

Predicting build orientation for SLM dental parts: a comparison of rotation representations and direct vector regression

预测选择性激光熔化(SLM)牙科部件的构建方向：旋转表示法与直接向量回归的比较

Schmalzel, Felix, Waitz, Reimar, Kronberger, Moritz, Schöler, Thorsten

Abstract

Build orientation for selective laser melting (SLM) manufacturing of dental parts is usually chosen manually by technicians. We treat orientation prediction as supervised machine learning of the part's up-axis from technician-labeled production data, and test which rotation representations produce the best results. Using $n\approx2400$ patient-specific dental parts, we trained a ResNet-50 multi-view image backbone and a PointNeXt-S point-cloud backbone, both pretrained and fine-tuned end-to-end, on 13 up-axis representations spanning six classical $SO(3)$ parameterizations and seven representations defined directly on the unit sphere $S^2$. We report the geodesic angular error between predicted and ground-truth up-axis on a test set, with and without test-time augmentation (TTA) over $K=21$ known rotations. With TTA, the octahedral map achieves the lowest mean angular error ($10.6^\circ$, ResNet-50). The three lowest-error results overall are direct $S^2$ representations, though this may reflect label noise in the unsupervised in-plane component of the $SO(3)$ targets rather than a topological advantage. von Mises-Fisher collapses to a near-constant prediction when trained with PointNeXt-S but not with ResNet-50. TTA reduces mean angular error by 31-73 % across almost every representation and backbone. Overall, test-time augmentation over a small set of known rotations is the most consistent driver of accuracy, whereas the best-performing representation is strongly backbone-dependent.

Chinese Translation

选择性激光熔化(SLM)牙科部件的构建方向通常由技术人员手动选择。我们将方向预测视为从技术人员标记的生产数据中进行的监督机器学习，旨在预测部件的上轴，并测试哪些旋转表示法能够产生最佳结果。使用约2400个患者特定的牙科部件，我们训练了一个ResNet-50多视图图像骨干网络和一个PointNeXt-S点云骨干网络，均经过预训练并端到端微调，涵盖了六种经典的$SO(3)$参数化和七种直接定义在单位球面$S^2$上的13种上轴表示法。我们报告了在测试集上预测的上轴与真实上轴之间的测地角误差，并在已知的$K=21$种旋转下进行了有无测试时增强(TTA)的比较。通过TTA，八面体映射实现了最低的平均角误差（$10.6^ ext{°}$，ResNet-50）。整体上，三个最低误差的结果是直接的$S^2$表示，尽管这可能反映了$SO(3)$目标的无监督平面分量中的标签噪声，而非拓扑优势。当使用PointNeXt-S训练时，von Mises-Fisher模型的预测几乎趋于常数，但在使用ResNet-50时则不是。TTA在几乎所有表示法和骨干网络中将平均角误差降低了31-73%。总体而言，在一小组已知旋转上进行的测试时增强是准确性的最一致驱动因素，而最佳表现的表示法则强烈依赖于骨干网络。

View on arXiv Download PDF AI Translation

cs.AI / 134 / 2609.15722

Data storytelling meets interpretable machine learning: Decoding AI decisions for non-experts without revealing sensitive data and model details

数据讲述与可解释机器学习的结合：为非专家解码人工智能决策，同时不泄露敏感数据和模型细节

Chao, Lemen, Yang, Zixuan, Fang, Anran, Sun, Mingran, Lei, Ming

Abstract

AI-driven automated decision-making requires both predictive performance and interpretability. Recent advances in interpretable machine learning (IML) provide tools for explaining model predictions, but the technical complexity of these explanations may hinder accessibility to non-experts. To address this challenge, this study integrates data storytelling with IML to enhance the explainability of AI-generated decisions for a broader audience. Following the design science research (DSR) paradigm, this study proposes a formal definition of data storytelling in IML, introduces the DIST Pyramid to align data storytelling with IML, and presents the I-P-O Model to describe their interactions. It further develops an architecture to explain AI decisions through distinct "What-if" and "Why-not" event-generation processes. The architecture also employs data desensitization to protect sensitive input data. To validate the approach, a case study is conducted with the Boston Housing dataset, using SHapley Additive exPlanations (SHAP) values and large language models (LLMs) to generate data stories with And-But-Therefore (ABT) structures. An empirical evaluation shows that 76.4% and 74.3% of respondents rated the "What-if" and "Why-not" data stories as more comprehensible, with significantly higher accessibility scores than traditional SHAP visualizations. The paper concludes with the presentation of a narrative interpretation framework that integrates IML and data storytelling, thereby expanding the research scope as well as the practical applicability of AI decision-making.

Chinese Translation

基于人工智能的自动决策需要兼顾预测性能和可解释性。近年来，可解释机器学习（IML）的进展提供了用于解释模型预测的工具，但这些解释的技术复杂性可能会阻碍非专家的理解。为了解决这一挑战，本研究将数据讲述与IML相结合，以增强人工智能生成决策的可解释性，使其面向更广泛的受众。遵循设计科学研究（DSR）范式，本研究提出了IML中数据讲述的正式定义，引入了DIST金字塔以对齐数据讲述与IML，并呈现了I-P-O模型以描述它们之间的互动。进一步开发了一种架构，通过不同的“如果……会怎样”（What-if）和“为什么不”（Why-not）事件生成过程来解释人工智能决策。该架构还采用数据去敏感化技术以保护敏感输入数据。为了验证该方法，使用波士顿住房数据集进行案例研究，利用SHapley加性解释（SHAP）值和大型语言模型（LLMs）生成具有“和-但是-因此”（And-But-Therefore，ABT）结构的数据故事。实证评估显示，76.4%和74.3%的受访者认为“如果……会怎样”和“为什么不”数据故事更易于理解，其可及性评分显著高于传统的SHAP可视化。本文最后提出了一个叙事解释框架，将IML与数据讲述相结合，从而扩展了研究范围和人工智能决策的实际应用。

View on arXiv Download PDF AI Translation

cs.AI / 135 / 2609.15727

Are LLMs Good Financial User Simulators? A Preliminary Study

大型语言模型是否是良好的金融用户模拟器？初步研究

He, Jiajie, Hong, Jiangyuan, Ni, Dongling, Liu, Wenjin, Chen, Xintong

Abstract

Large language models (LLMs) are increasingly used as user simulators, but their ability to reproduce evolving individual financial decisions remains unclear. We present a preliminary study in a controlled paper-trading environment with 120 volunteers. Participants used non-redeemable virtual funds under real-time market conditions; no real brokerage accounts, real-money positions, or real transaction records were accessed. Given only information available before a prediction cutoff, a simulator predicts the participant's next-trading-day action, traded security, and transaction quantity. We evaluate temporally aligned rolling predictions and compare settings with and without point-in-time market information. Market context improves action and ticker prediction in the controlled ablation, while transaction sizing remains difficult. We also observe systematic behavioral compression: models overproduce hold actions, underpredict sell decisions, and simplify multi-security transactions. These results provide an initial empirical characterization and motivate larger-scale evaluation of individual, temporal, and portfolio-level behavioral fidelity.

Chinese Translation

大型语言模型（LLMs）越来越多地被用作用户模拟器，但它们在再现不断变化的个体金融决策方面的能力仍不明确。我们在一个受控的纸上交易环境中进行了一项初步研究，共有120名志愿者参与。参与者在实时市场条件下使用不可赎回的虚拟资金；未接触真实的经纪账户、真实的资金头寸或真实的交易记录。在预测截止时间之前仅提供可用信息的情况下，模拟器预测参与者的下一个交易日行动、交易证券和交易数量。我们评估了时间上对齐的滚动预测，并比较了有无特定时间点市场信息的设置。市场背景在受控消融实验中改善了行动和股票代码的预测，而交易规模的预测仍然困难。我们还观察到系统性的行为压缩：模型过度产生持有行为，低估卖出决策，并简化多证券交易。这些结果提供了初步的实证特征，并激励对个体、时间和投资组合层面的行为忠实度进行更大规模的评估。

View on arXiv Download PDF AI Translation

cs.AI / 136 / 2609.15744

Design of a Deep Learning Credit Risk Early Warning System Integrating Multi-source Heterogeneous Data

基于多源异构数据的深度学习信用风险预警系统设计

Wang, LiYang, Zhong, Zhen, Tian, Zhen, Chen, Keyu, Chen, Keyu

Abstract

Advancements in data fusion and real-time analytics technologies have opened new avenues for addressing complex domain challenges. Financial risk early warning systems often suffer from inefficiency due to information silos and monitoring delays. This paper proposes a credit risk early warning system based on heterogeneous information fusion. The system employs a model architecture integrating deep neural networks and attention mechanisms to extract multidimensional features from diverse data sources such as transaction behaviors and social networks, thereby establishing an early identification mechanism for corporate and individual credit risks. System testing demonstrates that this approach significantly enhances the accuracy and timeliness of risk warnings, outperforming traditional rule-based engine solutions. The findings offer innovative insights for early intervention in financial risks, holding practical significance for safeguarding financial stability.

Chinese Translation

数据融合和实时分析技术的进步为解决复杂领域挑战开辟了新途径。金融风险预警系统常因信息孤岛和监测延迟而效率低下。本文提出了一种基于异构信息融合的信用风险预警系统。该系统采用集成深度神经网络和注意力机制的模型架构，从交易行为和社交网络等多样化数据源中提取多维特征，从而建立企业和个人信用风险的早期识别机制。系统测试表明，该方法显著提高了风险预警的准确性和及时性，优于传统的基于规则的引擎解决方案。研究结果为金融风险的早期干预提供了创新性见解，对维护金融稳定具有实际意义。

View on arXiv Download PDF AI Translation

cs.AI / 137 / 2609.15779

EvoOntology: A Self-Evolving Ontology Layer for Data Agents

EvoOntology：一种自我演化的数据代理本体层

Chong, Meiduo, Zhang, Shaolei, Fan, Ju, Du, Xiaoyong

Abstract

Data agents aim to fulfill natural-language instructions over heterogeneous data, including tables, files, and databases. However, data agents face a challenging agent-data gap: heterogeneous data resides outside the agent, while the agent can access it (e.g., column names and file paths) only through generic tools. Existing approaches either let agents directly explore raw data sources or inject manually constructed semantic layers into prompts. However, neither scales well to large heterogeneous data sources nor adapts to different agent behaviors. In this paper, we introduce EvoOntology, a self-evolving ontology layer for data agents. EvoOntology encapsulates the ontology as an MCP server comprising a schema layer, a content layer, and a tool layer, enabling agents to actively query and interact with the ontology at runtime. To this end, we introduce a builder agent for autonomous ontology construction and a self-evolution loop that continuously refines the ontology through attribution-guided typed edits that are accepted only after a backbone-conditional paired evaluation. Experiments on three well-adopted data-agent benchmarks with four LLM backbones demonstrate that EvoOntology consistently outperforms strong baselines and existing semantic-layer approaches, effectively bridging the agent-data gap and enabling more effective interaction with heterogeneous data. Code: https://github.com/ruc-datalab/EvoOntology

Chinese Translation

数据代理旨在对异构数据（包括表格、文件和数据库）执行自然语言指令。然而，数据代理面临着一个挑战性的代理-数据鸿沟：异构数据存在于代理之外，而代理只能通过通用工具访问这些数据（例如，列名和文件路径）。现有的方法要么让代理直接探索原始数据源，要么将手动构建的语义层注入到提示中。然而，这两种方法都无法很好地扩展到大型异构数据源，也无法适应不同的代理行为。在本文中，我们介绍了EvoOntology，一种自我演化的数据代理本体层。EvoOntology将本体封装为一个MCP服务器，包含模式层、内容层和工具层，使代理能够在运行时主动查询和与本体交互。为此，我们引入了一个用于自主本体构建的构建代理和一个自我演化循环，通过归因引导的类型编辑不断精炼本体，这些编辑只有在经过基于骨干条件的配对评估后才会被接受。在三个广泛采用的数据代理基准上进行的实验表明，EvoOntology在性能上始终优于强基线和现有的语义层方法，有效地弥合了代理-数据鸿沟，并实现了与异构数据的更有效交互。代码：https://github.com/ruc-datalab/EvoOntology

View on arXiv Download PDF AI Translation

cs.AI / 138 / 2609.15794

KnowBench: Effort Reduction as a Unified, Deployment-Grounded Benchmark for Clinical AI

KnowBench：作为临床人工智能统一的、基于部署的工作量减少基准

Kang, Jocelyn, Zhang, Caroline

Abstract

Clinical AI systems are evaluated with instruments built for research settings (reference-based similarity metrics and expert rubric panels) that measure resemblance to an artifact rather than reduction of a burden. We introduce KnowBench, pioneered by Knowtex, whose unifying metric is Effort Reduction (ER): the proportion of system-generated clinical work product accepted by the responsible clinician under expert and safety review. ER is defined once and instantiated per task across the administrative workload clinical AI automates: visit notes, diagnosis and billing codes, orders, EHR chart summarization, patient after-visit summaries, and clinical decision support. In every instantiation the construction is identical: the clinician's review-and-attestation event is the ground truth, every accepted unit is work the system completed, and every correction is residual effort returned to the clinician. The primary contribution of this paper is the benchmark itself: the metric, its degenerate cases, and a reporting protocol under which ER claims are auditable and cross-system comparable. Alongside it we report an initial headline measurement from the documentation instantiation: over one million signed encounters across a production window exceeding six months and thirteen medical specialties, Knowtex's proprietary fine-tuned clinical foundation models operating inside a closed feedback architecture achieve an aggregate ER of 97.99%, with per-specialty aggregates spanning 96.8-98.9%. This release reports the protocol's checklist partially, and states which companion statistics are withheld; the benchmark is offered so that this figure, and every figure reported after it, can be held to the same standard.

Chinese Translation

临床人工智能系统的评估通常使用为研究环境构建的工具（基于参考的相似性度量和专家评审小组），这些工具测量的是与某一工件的相似性，而非负担的减少。我们介绍了KnowBench，这是由Knowtex首创的，其统一指标是工作量减少（Effort Reduction, ER）：在专家和安全审查下，负责临床工作的临床医生接受的系统生成的临床工作产品的比例。ER在每个任务中定义一次，并在临床人工智能自动化的行政工作负载中实例化：就诊记录、诊断和计费代码、医嘱、电子健康记录（EHR）摘要、患者就诊后总结以及临床决策支持。在每个实例中，构建是相同的：临床医生的审核和确认事件是基准事实，每个被接受的单位是系统完成的工作，而每个修正是返回给临床医生的剩余工作量。本文的主要贡献是基准本身：该指标、其退化案例以及一套报告协议，使得ER声明可以被审计并在不同系统间进行比较。同时，我们报告了文档实例化的初步关键测量：在超过六个月和十三个医学专业的生产窗口内，Knowtex的专有精细调优临床基础模型在封闭反馈架构中实现了97.99%的总ER，按专业划分的总和在96.8%到98.9%之间。本次发布部分报告了该协议的检查清单，并说明了哪些伴随统计数据被保留；该基准的提供使得这一数字，以及其后报告的每一个数字，都可以遵循相同的标准。

View on arXiv Download PDF AI Translation

cs.AI / 139 / 2609.15800

Navigating Sparse Evidence: Agentic Visual RAG via Explicit Context Selection and Consolidation

导航稀疏证据：通过显式上下文选择和整合的代理视觉 RAG

Shen, Yucheng, Yan, Lingyong, Wu, Jiulong, Wang, Shuaiqiang, WU, Jianmin, Yin, Dawei, Cao, Min

Abstract

Visual Retrieval-Augmented Generation (VRAG) empowers models to navigate and answer queries about visually rich documents by retrieving relevant page images as visual evidence and reasoning over their content. However, effectively utilizing this visual evidence is usually impeded by two main challenges. First, answer-relevant evidence is sparse and may be concentrated in a small region of one page or dispersed across multiple pages. Second, existing agentic methods often generate answers based on raw exploration trajectories or compressed textual memories rather than an explicitly organized set of supporting images, making answers susceptible to exploration noise and obscuring the evidence-backed reasoning trace. We argue that the bottleneck lies not only in evidence discovery but also in its preservation and organization before answer generation. We propose SCoRE (Selection and Consolidation for Robust Evidence), a unified agent loop for explicit evidence selection and consolidation. During exploration, SCoRE retains only query-relevant observations and their source pointers in a maintained textual ledger, preserving earlier evidence while keeping the visual context bounded. At termination, it reloads the referenced original images and consolidates the visual evidence for answering, arranging it into a logical sequence. This decouples final reasoning from exploratory trial-and-error while ensuring strict visual grounding via indexed claim-to-image linkages. To enable end-to-end optimization of this unified rollout, our training paradigm combines filtered cold-start trajectory distillation with evidence-aware reinforcement learning, whose reward promotes evidence coverage, consolidation compactness, and answer correctness.

Chinese Translation

视觉检索增强生成（VRAG）使模型能够通过检索相关页面图像作为视觉证据并对其内容进行推理，从而导航并回答关于视觉丰富文档的查询。然而，有效利用这些视觉证据通常受到两个主要挑战的阻碍。首先，与答案相关的证据稀疏，可能集中在一页的一个小区域内或分散在多页之间。其次，现有的代理方法通常基于原始探索轨迹或压缩的文本记忆生成答案，而不是基于显式组织的支持图像集，这使得答案容易受到探索噪声的影响，并模糊了基于证据的推理轨迹。我们认为瓶颈不仅在于证据的发现，还在于在生成答案之前对其进行保存和组织。我们提出了 SCoRE（选择与整合以增强证据的鲁棒性），这是一个用于显式证据选择和整合的统一代理循环。在探索过程中，SCoRE 仅保留与查询相关的观察结果及其源指针，并在维护的文本账本中保存早期证据，同时保持视觉上下文的界限。在终止时，它重新加载引用的原始图像，并整合视觉证据以进行回答，将其排列成逻辑顺序。这将最终推理与探索性的试错过程解耦，同时通过索引的主张与图像链接确保严格的视觉基础。为了实现这一统一展开的端到端优化，我们的训练范式结合了过滤的冷启动轨迹蒸馏与证据感知的强化学习，其奖励促进证据覆盖、整合紧凑性和答案正确性。

View on arXiv Download PDF AI Translation

cs.AI / 140 / 2609.15801

When Should a World Model Move? Loss-Conditioned State Execution

世界模型何时应移动？基于损失的状态执行

Xu, Jintao, Chen, Zhengyu, Zhang, Ben, Qi, Yongzhi, Zhang, Jianshen

Abstract

We introduce loss-conditioned state execution, a model-agnostic method that decides whether to execute a world model's fixed feasible proposal or retain the current state. Predictive informativeness alone, however, does not establish whether an update will reduce downstream loss. Occurrence ranking can approach perfection while persistence remains the unique absolute-loss Bayes action. Two transition laws can also share occurrence information and conditional variance yet require opposite absolute-loss decisions. We formalize state movability as the existence of a loss-reducing feasible correction and distinguish it from the benefit of a particular proposal. Our method constructs a loss-specific feasible proposal from a predictive distribution and evaluates its groupwise bounded-loss gain over persistence on independent calibration units. The proposal is executed only in groups with a positive simultaneous lower confidence bound. For fixed proposals and groups with bounded unit losses, we prove that every accepted group has lower expected loss than persistence with high probability when calibration units are i.i.d. draws from the target population. Experiments on public forecasting and action-conditioned dynamics benchmarks show supported updates and a trade-off between certification and coverage. On 28,684 held-out M4 Monthly series, the method executes the proposal for 14.0% of series and achieves bounded loss 0.588, compared with 0.599 for persistence and 0.621 for always executing the proposal. The paired 95% bootstrap intervals for both comparisons lie below zero. In constrained forecasting of six unhealthy-inventory types from JD$\mbox{.}$com, a leading e-retailer in China, strong occurrence-ranking signal coexists with a loss-based preference for persistence, illustrating why event predictability and state execution must be evaluated separately.

Chinese Translation

我们提出了一种基于损失的状态执行方法，这是一种与模型无关的方法，用于决定是执行世界模型的固定可行提案还是保留当前状态。然而，仅凭预测信息并不能确定更新是否会降低下游损失。发生排名可以接近完美，而持久性则仍然是唯一的绝对损失贝叶斯行动。两种转移法则也可以共享发生信息和条件方差，但却需要相反的绝对损失决策。我们将状态可移动性形式化为存在一种降低损失的可行修正，并将其与特定提案的收益区分开来。我们的方法从预测分布中构建一个特定于损失的可行提案，并评估其在独立校准单元上相对于持久性的群体界限损失增益。该提案仅在具有正的同时下限置信区间的群体中执行。对于固定提案和具有界限单位损失的群体，我们证明在校准单元是来自目标人群的独立同分布样本时，每个被接受的群体的期望损失在高概率下低于持久性。在公共预测和基于动作的动态基准测试中的实验显示了支持更新以及认证与覆盖之间的权衡。在28,684个保留的M4月度系列中，该方法对14.0%的系列执行提案，并实现了界限损失0.588，而持久性为0.599，始终执行提案为0.621。两者比较的配对95%自助法区间均低于零。在对中国领先电子零售商JD.com的六种不健康库存类型的受限预测中，强发生排名信号与基于损失的持久性偏好共存，说明事件可预测性和状态执行必须分开评估的原因。

View on arXiv Download PDF AI Translation

cs.AI / 141 / 2609.15818

Atria Dawn: The Dawn of Agentic Superintelligence

Atria Dawn：代理超智能的曙光

Guo, Honglin, Gui, Tao, Chen, Yicheng, Dong, Guanting, Ge, Qiming, Hu, Yuyang, Huang, Zixian, Jin, Jiajie, Lam, Alexander, Li, Yining, Lin, Jiahang, Liu, Yanjiang, Lu, Xinyu, Lv, Haijun, Shang, Junlin, Su, Qisheng, Wang, Guoqiang, Wang, Rui, Wang, Zhecan, Xiang, Hao, Xie, Xinchen, Xing, Shuhao, Xing, Xiaoyu, Xu, Wanghan, Yang, Xinyu, Yang, Yajie, Zhao, Chengfeng, Zhao, Haoran, Zhou, Ruojun, Zhou, Yunhua, Zou, Yicheng, Cai, Kun, Cai, Qiye, Che, Xinmeng, Chen, Haodong, Chen, Jiabei, Chen, Jiahao, Chen, Jiayi, Chen, Yujia, Cui, Lizhi, Dai, Youheng, Deng, Xin, Dong, Yi, Dou, Shihan, Gu, Chenya, Guo, Xu, Han, Ding, Hao, Feiyang, He, Haotan, Hou, Jie, Hu, Binze, Hu, Zijian, Huang, Junhao, Jiang, Huicheng, Jiang, Jiazhen, Jiang, Shufan, Kuang, Jiahao, Lai, Bowen, Li, Bo, Li, Jiaqiang, Li, Peng, Li, Qilong, Li, Zhuoqun, Liu, Jiaxiang, Liu, Shuainan, Liu, Tong, Liu, Yi, Lu, Zhonghang, Luo, Jianwen, Luo, Yanyi, Lv, Huijie, Ma, Ningsheng, Ma, Zerun, Min, Houcheng, Pan, Chengjun, Peng, Qiyuan, Peng, Xiaoxuan, Qian, Jianmin, Qiu, Jiantao, Ren, Wanying, Sha, Huayu, Shan, Jifei, Shang, Zixin, Shao, Bing, Sheng, Zhuohui, Shi, Jiayang, Shu, Yang, Simayi, Aierpanjiang, Song, Sirui, Song, Yuxiao, Sun, Zhe, Sun, Zhichao, Tan, Wenzhe, Tian, Wenhui, Tian, Zhongbo, Wang, Hanchen, Wang, Pengbo, Wang, Rui, Wang, Yiding, Wang, Yuhui, Xi, Zhiheng, Xu, Caijun, Xu, Chao, Xu, Yongfeng, Yang, Xiaolei, Yang, Zhixiong, Yao, Qian, Yi, Shihong, Ying, Yuankai, Yu, Jia, Yuan, Dingbo, Yuan, Hao, Yuan, Junjie, Zhang, Bo, Zhang, Caixian, Zhang, Qiuyinzhe, Zhao, Jiyuan, Zhao, Penghao, Zhao, Ying, Zheng, Pujun, Zhong, Xiaoxue, Zhou, Xiaohao, Zhou, Xinyu, Zhu, Dongsheng, Zhu, Guanru, Zhu, Yulun, Lu, Yaojie, Ji, Tao, Lin, Hongyu, Zhu, Yutao, Cao, Pengfei, He, Guoxiu, Han, Xianpei, He, Ben, Dou, Zhicheng, Liu, Kang, Zhang, Qi, Sun, Le, Zhao, Jun, Wen, Ji-Rong, Huang, Xuanjing, Jiang, Yu-Gang, Zhou, Bowen

Abstract

As AI agents become participants in the development of their successors, they reshape both the production of intelligence and the role of human researchers. We introduce Atria Dawn Preview, a foundation agentic language model designed for scientific research and engineering workflows, with the goal of expanding the frontier of agent productivity in the real world. This model is trained via a Verifiable Experience Pipeline that connects tool-mediated interactions to executable environments and externally verified outcomes. Across 16 benchmarks spanning real-world research, engineering, and digital work, Atria Dawn Preview is competitive with frontier agents and achieves the highest reported score on five of them. Beyond standalone performance, we examine the real research-and-development process behind this model as a case study of human--AI collaboration, analyzing 769 task records from 56 participants together with agent logs. When asked to evaluate completed tasks under comparable conditions, participants rated about one-third of completed AI-assisted tasks as infeasible without AI. More strikingly, agents frequently propose methods and implement revisions, while humans retain most final decisions and guide exploration through judgment and feedback. These observations indicate a shift from task-level execution to project-level partnership, with human effort concentrating on what is worth pursuing and how evidence should guide research. Progress toward more autonomous AI research must therefore advance both the capacity for discovery and the capacity for meaningful human oversight, preserving accountable human authority over the risks and direction of continued development.

Chinese Translation

随着人工智能代理成为其后继者发展的参与者，它们重塑了智能的生产和人类研究者的角色。我们介绍了 Atria Dawn Preview，这是一个为科学研究和工程工作流程设计的基础代理语言模型，旨在扩展现实世界中代理生产力的前沿。该模型通过可验证的经验管道进行训练，该管道将工具介导的交互连接到可执行环境和外部验证的结果。在涵盖现实世界研究、工程和数字工作的 16 项基准测试中，Atria Dawn Preview 与前沿代理相竞争，并在其中五项测试中获得了最高的报告分数。除了独立性能外，我们还将这一模型背后的真实研究与开发过程作为人类与人工智能合作的案例进行分析，研究了来自 56 位参与者的 769 项任务记录以及代理日志。当被要求在可比条件下评估已完成的任务时，参与者认为大约三分之一的 AI 辅助任务在没有 AI 的情况下是不可行的。更引人注目的是，代理经常提出方法并实施修订，而人类则保留大部分最终决策，并通过判断和反馈引导探索。这些观察表明，从任务级执行转向项目级合作，人类的努力集中在值得追求的内容以及证据如何指导研究上。因此，朝着更自主的人工智能研究的进展必须同时提升发现的能力和有意义的人类监督的能力，保持对持续发展的风险和方向的可追溯人类权威。

View on arXiv Download PDF AI Translation

cs.AI / 142 / 2609.15820

AlgoEvo: Self-Evolving Agentic Search for Automated Algorithm Discovery

AlgoEvo：自我进化的代理搜索用于自动化算法发现

Qiu, Junhao, Hu, Qinglong, Tong, Xialiang, Yuan, Mingxuan, Lin, Liyong, Zhang, Qingfu

Abstract

Large language models have advanced automated algorithm discovery by synthesizing executable code, but existing frameworks trap them in rigid search pipelines with pre-defined control flows. This limitation restricts adaptive reasoning, blocks cross-paradigm transfer, and discards valuable execution feedback. We propose AlgoEvo, a unified agentic framework that transforms automated algorithm discovery into an interactive, knowledge-accumulating process. An autonomous agent dynamically inspects, diagnoses, and edits code based on runtime feedback. A design skill hub decouples paradigm-specific knowledge from the core discovery engine, allowing a single workflow to seamlessly handle single-objective, multi-objective, and multi-component design. Meanwhile, a hierarchical experience mechanism organizes search trajectories into a task-level tree to guide exploration and consolidates cross-task patterns into reusable skills. Across six representative benchmark tasks, AlgoEvo matches or surpasses specialized methods with substantially fewer evaluations and reduced token consumption, demonstrating strong intra-task accumulation, cross-task transfer, and the ability to reproduce or exceed existing state-of-the-art performance through flexible skill activation.

Chinese Translation

大型语言模型通过合成可执行代码推动了自动化算法发现，但现有框架将其限制在具有预定义控制流的僵化搜索管道中。这一限制阻碍了自适应推理，阻塞了跨范式转移，并丢弃了宝贵的执行反馈。我们提出了AlgoEvo，一个统一的代理框架，将自动化算法发现转变为一个互动的、知识积累的过程。一个自主代理根据运行时反馈动态检查、诊断和编辑代码。设计技能中心将特定范式的知识与核心发现引擎解耦，使单一工作流程能够无缝处理单目标、多目标和多组件设计。同时，分层经验机制将搜索轨迹组织成任务级树，以指导探索，并将跨任务模式整合为可重用的技能。在六个具有代表性的基准任务中，AlgoEvo的评估次数显著少于专门方法，同时减少了令牌消耗，展示了强大的任务内积累、跨任务转移能力，以及通过灵活的技能激活重现或超越现有最先进性能的能力。

View on arXiv Download PDF AI Translation

cs.AI / 143 / 2609.15859

LongAgent: History-Guided Agentic Search for Longitudinal Outcome Prediction

LongAgent：基于历史指导的长效结果预测代理搜索

Wang, Siyao, Guitton, Florian, Fu, Shuojie, Tao, Guanyu, Sun, Kai, Bai, Wenjia

Abstract

Extracting informative representations from longitudinal data that can predict future outcomes remains a critical challenge in medicine. Medical datasets are inherently heterogeneous, consisting of a large number of variables collected from different sources, sampled with different temporal spacings, and representing different aspects of human health status. This requires identifying those variables with predictive value, processing longitudinal information, and integrating multiple variables for outcome prediction. Here, we propose a novel agent-based approach, LongAgent, that can autonomously search over combinations of variable sets, temporal windows and longitudinal aggregation functions, and identify candidates with promising predictive performance. LongAgent utilises a history memory of previous searches and numerical evidence to guide subsequent exploration. On synthetic data, LongAgent achieves a mean prediction RMSE of 1.7376 and improves over the strongest non-agent baseline by 0.0151 (95% CI: [0.0045,0.0260]; p=0.0273). On a real clinical dataset, it performs comparably to the best baseline.

Chinese Translation

从纵向数据中提取能够预测未来结果的信息表示仍然是医学领域的一项关键挑战。医学数据集本质上是异质的，包含大量从不同来源收集的变量，这些变量以不同的时间间隔进行采样，并代表人类健康状态的不同方面。这需要识别具有预测价值的变量，处理纵向信息，并整合多个变量以进行结果预测。在此，我们提出了一种新颖的基于代理的方法——LongAgent，它能够自主搜索变量集、时间窗口和纵向聚合函数的组合，并识别具有良好预测性能的候选者。LongAgent利用之前搜索的历史记忆和数值证据来指导后续的探索。在合成数据上，LongAgent实现了平均预测均方根误差（RMSE）为1.7376，并比最强的非代理基线提高了0.0151（95% CI: [0.0045,0.0260]; p=0.0273）。在真实临床数据集上，它的表现与最佳基线相当。

View on arXiv Download PDF AI Translation

cs.AI / 144 / 2609.15919

Pilot Early, Commit Late: A Real-Options Model of Enterprise AI Adoption under Rapid Technological Progress

提前试点，迟迟承诺：快速技术进步下企业人工智能采纳的实物期权模型

Tewari, Gaurav

Abstract

Artificial intelligence presents firms with an unusual timing problem. The technology frontier is improving rapidly, implementation is partly irreversible, and organization-specific capabilities are accumulated through action. This paper develops a two-period decision model of AI deployment under uncertainty in which a firm chooses among immediate deployment, a limited pilot, and waiting. Deployment earns current operating value but exposes the firm to architectural obsolescence; waiting preserves the option to adopt after the frontier is observed; a pilot sacrifices current operating value to build organization-specific learning without full commitment. The model yields five central timing results and a sixth comparative result on where learning occurs. First, a mean-preserving increase in frontier uncertainty raises the value of waiting and piloting but leaves immediate deployment unchanged when its payoff is affine in the frontier. Second, faster expected frontier progress can reduce the relative attractiveness of immediate deployment when deployed architecture captures only a limited share of future improvement. Third, a pilot dominates waiting exactly when the expected value of the capability it builds exceeds its cost. Fourth, sufficiently valuable organization-specific learning creates a nonempty region in which "pilot early, commit late" is optimal. Fifth, there is a closed-form modularity threshold above which immediate deployment dominates the best outside option. Sixth, production learning and pilot-specific learning affect the timing margin differently. A continuous-time extension recovers the standard result that uncertainty raises the adoption threshold while capability and modularity lower it. The paper separates deploying, experimenting, and waiting, and shows why rapid progress can rationally increase experimentation without justifying irreversible commitment.

Chinese Translation

人工智能给企业带来了一个不寻常的时机问题。技术前沿正在快速提升，实施部分是不可逆的，组织特定能力通过行动积累。本文在不确定性下发展了一个两期决策模型，企业在其中选择立即部署、有限试点或等待。部署带来当前的运营价值，但使企业面临架构过时的风险；等待则保留了在观察到前沿后采纳的选项；试点牺牲当前的运营价值，以在不完全承诺的情况下建立组织特定学习。该模型产生了五个核心时机结果和一个关于学习发生地点的比较结果。首先，前沿不确定性的均值保持增加提高了等待和试点的价值，但在其回报与前沿呈仿射关系时，立即部署不变。其次，预期前沿进展的加快可能降低立即部署的相对吸引力，当部署的架构仅捕获未来改进的一小部分时。第三，当其构建的能力的预期价值超过其成本时，试点恰好优于等待。第四，足够有价值的组织特定学习创造了一个非空区域，在该区域内“提前试点，迟迟承诺”是最优的。第五，存在一个封闭形式的模块化阈值，超过该阈值时，立即部署优于最佳外部选项。第六，生产学习和试点特定学习对时机边际的影响不同。一个连续时间的扩展恢复了不确定性提高采纳阈值的标准结果，而能力和模块化则降低了该阈值。本文区分了部署、实验和等待，并展示了快速进步如何合理地增加实验而不需要证明不可逆的承诺。

View on arXiv Download PDF AI Translation

cs.AI / 145 / 2609.15932

Recurrent GraphNeural NetworkswithSet-BasedAggregation

基于集合聚合的递归图神经网络

Bonet, Blai

Abstract

Recurrent GNNs iterate message passing to convergence, and their logical characterizations to date rely on multi-set aggregation, graded (counting) logics, and halting or acceptance conditions that cannot be verified from the network's parameters. We study recurrent GNNs with set-based aggregation and identify sufficient conditions checkable from the weights for networks to compile into formulas and formulas into networks. The main result is an effective, two-directional equivalence between a class of networks and the Boolean closure of reachability and safety properties, the fragment B$\Sigma^{\circ}_1$ of the modal $\mu$-calculus. The fragment is not an artifact: it is the exact expressive level of stabilization over finite vocabulary, which supports fixed points of a single polarity and Boolean combinations thereof, but not the composition of fixed points of opposite polarities. The correspondence needs no counting logic, no external halting signal, and no non-effective acceptance condition, yielding a verifiable path from weights to symbolic explanations for networks meeting the conditions.

Chinese Translation

递归图神经网络（Recurrent GNNs）通过迭代消息传递达到收敛，其逻辑特征至今依赖于多集合聚合、分级（计数）逻辑以及无法从网络参数验证的停机或接受条件。我们研究了基于集合聚合的递归图神经网络，并确定了可从权重检查的充分条件，使网络能够编译成公式，公式也能编译成网络。主要结果是网络类与可达性和安全性属性的布尔闭包之间的有效双向等价关系，即模态μ-演算的片段B$ ext{Σ}^{ ext{o}}_1$。该片段并非伪影：它是有限词汇上稳定性的确切表达水平，支持单一极性的固定点及其布尔组合，但不支持相反极性的固定点的组合。该对应关系不需要计数逻辑、外部停机信号或非有效的接受条件，从而为满足条件的网络提供了从权重到符号解释的可验证路径。

View on arXiv Download PDF AI Translation

cs.AI / 146 / 2609.15983

Stellar Colosseum: A Many-Agent Harness for Long-Horizon Research in Mathematics and Theoretical Computer Science

星际竞技场：一个用于数学和理论计算机科学长远研究的多智能体框架

Lin, Honghao, Woodruff, David P., Deng, Yuan, Mao, Jieming, Zuo, Song, Mirrokni, Vahab

Abstract

Language models can produce plausible short proofs, but may still be unreliable on long-horizon research problems, where progress depends on a sequence of uncertain and interdependent decisions. We introduce Stellar Colosseum, a model-agnostic harness for allocating inference across research in mathematics and theoretical computer science. Colosseum explores alternative strategies before proof construction, uses a readiness gate to decide when a route is mature enough to decompose, represents the proof plan as interdependent section-level subproblems, and routes verifier findings back to the affected part of the argument. Across these stages, it generates candidates in parallel, attacks them with targeted falsification, and combines candidates and their critiques into a single research artifact through overlapping random-sample tree aggregation. The Colosseum workflow has also been integrated into Google Antigravity's Teamwork framework as the Long Proof pattern. We demonstrate the capabilities of Colosseum through open-ended research and evaluations on theorem-proving and competitive programming benchmarks. Using Colosseum with Gemini 3.1 Pro, we obtain several new results that address open problems arising from papers published at top venues such as FOCS and JMLR. On TCS-Bench, a benchmark of research-level theorem-proving tasks drawn from papers published at FOCS, STOC, and SODA, Colosseum achieves 71.0% accuracy using Gemini 3.1 Pro and Gemini 3.7 Flash. In a separate Codeforces evaluation using Gemini 3.1 Pro, the proof-oriented pipeline with execution feedback solves 218 of 222 problems.

Chinese Translation

语言模型能够生成合理的短证明，但在长远研究问题上可能仍然不可靠，因为进展依赖于一系列不确定且相互依赖的决策。我们介绍了星际竞技场（Stellar Colosseum），这是一个与模型无关的框架，用于在数学和理论计算机科学研究中分配推理。竞技场在构建证明之前探索替代策略，使用准备门（readiness gate）来决定何时一条路径成熟到足以进行分解，将证明计划表示为相互依赖的章节级子问题，并将验证者的发现反馈回论证的相关部分。在这些阶段中，它并行生成候选项，使用针对性的反驳进行攻击，并通过重叠随机样本树聚合将候选项及其批评合并为一个单一的研究成果。竞技场工作流程还被集成到谷歌反重力（Google Antigravity）的团队协作框架中，作为长证明模式（Long Proof pattern）。我们通过开放式研究和在定理证明及竞争编程基准上的评估展示了竞技场的能力。使用星际竞技场与Gemini 3.1 Pro，我们获得了多个新结果，解决了来自FOCS和JMLR等顶级会议发表的论文中出现的开放问题。在TCS-Bench上，这是一个基于FOCS、STOC和SODA发表的论文提取的研究级定理证明任务的基准，竞技场在使用Gemini 3.1 Pro和Gemini 3.7 Flash时达到了71.0%的准确率。在使用Gemini 3.1 Pro的单独Codeforces评估中，具有执行反馈的面向证明的管道解决了222个问题中的218个。

View on arXiv Download PDF AI Translation

cs.AI / 147 / 2609.15989

Corrupt Plans, Clean Traces: Evading Chain-of-Thought Monitoring with Plan Injection

腐败的计划，干净的痕迹：通过计划注入规避思维链监控

Chidambaram, Keertana, Ilyas, Andrew, Syrgkanis, Vasilis

Abstract

Chain-of-thought (CoT) monitoring is a safety strategy where the reasoning of a large language model "actor" is inspected by a "monitor" (often another language model) for signs of unsafe planning, deception, or misalignment. We find that planting harmful but benign-sounding reasoning in the actor's context can steer it to perform adversarial actions while evading monitors, an attack we term "plan injection". We initially discover this attack in the multiple-choice question-answering monitorability setting proposed by Lanham et al. (2023), using the investigator-agent elicitation framework of Li et al. (2025). We generalize the attack and show that the discovered behavior scales to harder tasks (achieving 25-33% monitor evasion rates across different monitorability benchmarks) and larger models such as DeepSeek-R1. Across the settings we study, actor models not only follow injected plans but also paraphrase them as their own reasoning, without explicit attribution to the injections. Finally, we find cases where extra monitor resources cause harm - giving the monitor access to the injected plan drops detection by as much as 50% in the Bio-Math task and in a case study on monitor reasoning budget, we find transcripts where additional thinking tokens are spent rationalizing the injected plan rather than flagging it.

Chinese Translation

思维链（CoT）监控是一种安全策略，其中大型语言模型“参与者”的推理被“监控者”（通常是另一个语言模型）检查，以寻找不安全规划、欺骗或不一致的迹象。我们发现，在参与者的上下文中植入有害但听起来无害的推理，可以引导其执行对抗性行为，同时规避监控，这种攻击我们称之为“计划注入”。我们最初在Lanham等人（2023）提出的多项选择问答监控设置中发现了这一攻击，使用了Li等人（2025）的调查者-代理引导框架。我们对该攻击进行了推广，并展示了发现的行为在更困难的任务中也能扩展（在不同的监控能力基准中实现25-33%的监控规避率），以及在更大的模型如DeepSeek-R1中也适用。在我们研究的设置中，参与者模型不仅遵循注入的计划，还将其改写为自己的推理，而不明确归因于这些注入。最后，我们发现额外的监控资源会造成伤害——让监控者访问注入的计划会使检测率下降多达50%（在Bio-Math任务中），而在关于监控推理预算的案例研究中，我们发现转录文本中额外的思维标记被用于合理化注入的计划，而不是标记它。

View on arXiv Download PDF AI Translation

计算语言学 (Computation and Language)

132

cs.CL / 1 / 2609.13151

Token Merging for Multilingual Speech Recognition: A Systematic Study Across Model Scale and Fine-Tuning

多语言语音识别中的标记合并：跨模型规模和微调的系统研究

Holyoak, Dylan Luke

Abstract

Leading multilingual speech recognition models like Whisper transcribe diverse, low-resource languages without language-specific training but are computationally expensive to deploy. Token merging mitigates this inefficiency by dynamically combining redundant features, shortening the sequence length during inference without requiring retraining. In this paper, we systematically evaluate token merging on the Whisper model family across sixteen diverse languages and three different model sizes. We also test how token merging interacts with fine-tuning (DoRA) on low-resource languages. Our findings show that merging tokens increases computational efficiency with almost no loss in transcription accuracy across most low-resource languages and model sizes, and it works even after the model has been fine-tuned. Our results demonstrate that token merging is a highly practical method for making multilingual speech recognition faster and cheaper to deploy.

Chinese Translation

领先的多语言语音识别模型，如 Whisper，能够在没有特定语言训练的情况下转录多样的低资源语言，但其部署计算成本高昂。标记合并通过动态组合冗余特征来缓解这种低效，在推理过程中缩短序列长度，而无需重新训练。在本文中，我们系统地评估了在十六种不同语言和三种不同模型规模上，标记合并在 Whisper 模型家族中的表现。我们还测试了标记合并与低资源语言的微调（DoRA）之间的相互作用。我们的研究结果表明，标记合并在大多数低资源语言和模型规模中几乎没有损失转录准确性的情况下，提高了计算效率，并且即使在模型经过微调后仍然有效。我们的结果表明，标记合并是一种非常实用的方法，可以使多语言语音识别的部署更快且成本更低。

View on arXiv Download PDF AI Translation

cs.CL / 2 / 2609.13152

PhysMent: An Interactive Approach For LLM Reasoning In Physics Problems

PhysMent：一种用于物理问题中大语言模型推理的互动方法

Chan, Joseph, Jha, Utkarsh, Yang, Xiyin, Jarajapu, Abhinav, Sahai, Anik, Hu, Eddie, Deepak, Robin Jeshua, Saravalle, Stefano, Shah, Aditya

Abstract

Large language models (LLMs) perform strongly on static science benchmarks, yet their ability to reason about the physical world through active experimentation remains poorly understood. We introduce PhysMent, a benchmark that evaluates LLM physical reasoning via iterative, toolmediated interaction with a MuJoCo physics simulator. Unlike static benchmarks that supply all quantities upfront, PhysMent requires models to discover information by applying forces, querying object states, advancing time, and modifying scene geometry before answering. The benchmark comprises 105 scenes of classical mechanics, organized across four difficulty regimes (Easy/Hard and Single/Multi), three scene modalities (standard, object creation, hidden objects), and a scene-manipulation category, evaluated with a six-dimensional scoring framework. Results show that current models perform reasonably well on qualitative single-concept tasks (up to 80% accuracy) but degrade substantially on quantitative tasks that demand precise, multi-step experimental procedures: most models fall below 30% on the hardest single-concept category, where the bottleneck is procedural (adaptive multi-step tool use) rather than conceptual load. Across the seven models, accuracy ranges from 25% to 67%, with failures due to premature answer submission, inefficient exploration, and inconsistent grounding in simulator feedback rather than conceptual gaps.

Chinese Translation

大型语言模型（LLMs）在静态科学基准测试中表现优异，但它们通过主动实验推理物理世界的能力仍然不够清楚。我们介绍了PhysMent，这是一个通过与MuJoCo物理模拟器进行迭代的工具介导交互来评估LLM物理推理的基准。与提供所有量值的静态基准不同，PhysMent要求模型通过施加力、查询物体状态、推进时间和修改场景几何形状来发现信息，然后再进行回答。该基准包括105个经典力学场景，分为四个难度级别（简单/困难和单一/多重）、三种场景模式（标准、物体创建、隐藏物体）以及一个场景操控类别，采用六维评分框架进行评估。结果表明，当前模型在定性单概念任务上的表现相当不错（准确率高达80%），但在需要精确多步骤实验程序的定量任务中表现显著下降：在最困难的单概念类别中，大多数模型的准确率低于30%，瓶颈在于程序性（自适应多步骤工具使用）而非概念负担。在七个模型中，准确率范围从25%到67%，失败原因主要是由于过早提交答案、探索效率低下以及在模拟器反馈中缺乏一致性，而非概念性缺口。

View on arXiv Download PDF AI Translation

cs.CL / 3 / 2609.13154

Lexical Prompt Compression for Large Language Models: A Training-Free, Deterministic Pipeline with Empirical Pareto Analysis Across Eleven Task Categories

大语言模型的词汇提示压缩：一种无训练、确定性的管道及其在十一类任务中的经验帕累托分析

Chokshi, Shamin

Abstract

Recent advances in large language models (LLMs) have made prompts increasingly large and complex. Techniques such as chain-of-thought reasoning (Wei et al., 2022) and in-context learning (Brown et al., 2020) frequently push real-world prompts past several thousand tokens, increasing inference cost and latency. Learned compression methods such as LLMLingua (Jiang et al., 2023) and Selective Context (Li et al., 2023) achieve high compression ratios but require auxiliary language models and are non-deterministic. We ask a complementary question: how far can a training-free, fully deterministic, CPU-only pipeline based on classical lexical NLP be pushed before output quality degrades significantly? Eleven toggleable lexical transformations - stopword removal, filler-phrase deletion, contraction and abbreviation substitution, part-of-speech-based pruning, lemmatization, WordNet-driven synonym shortening, and named-entity preservation - are assembled into a configurable pipeline. Fifteen configurations are evaluated on 1,242 English-only prompts from six sources (Dolly-15k, LMSYS-Chat-1M, WildChat-1M, MMLU, GSM8K, HellaSwag), spanning eleven automatically derived task categories, yielding 18,630 paired GPT-4o-mini completions. Output preservation is measured using BLEU, ROUGE-1/2/L, BERTScore-F1, and SentenceBERT cosine similarity. The most aggressive configuration achieves a mean token reduction of 40.3% (sigma = 9.2) at a BERTScore-F1 of 0.876 against the original-prompt output; a stopword-only configuration achieves 29.6% reduction at 0.913. The compression-versus-fidelity Pareto frontier is characterized per task category, with commonsense reasoning a systematic failure mode under aggressive compression. All code, prompts, and per-cell results are released for reproducibility.

Chinese Translation

最近，大语言模型（LLMs）的进展使得提示变得越来越大和复杂。诸如链式思维推理（Chain-of-Thought Reasoning，Wei et al., 2022）和上下文学习（In-Context Learning，Brown et al., 2020）等技术，常常使得实际应用中的提示超过几千个标记，增加了推理成本和延迟。学习型压缩方法如LLMLingua（Jiang et al., 2023）和选择性上下文（Selective Context，Li et al., 2023）虽然能实现高压缩比，但需要辅助语言模型且是非确定性的。我们提出一个互补性的问题：基于经典词汇自然语言处理的无训练、完全确定性的仅CPU管道在输出质量显著下降之前，可以被推向何种极限？我们将十一种可切换的词汇转换方法——停用词去除、填充短语删除、缩写与简写替换、基于词性修剪、词形还原、基于WordNet的同义词缩短以及命名实体保留——组合成一个可配置的管道。对来自六个来源（Dolly-15k、LMSYS-Chat-1M、WildChat-1M、MMLU、GSM8K、HellaSwag）的1,242个仅英文提示进行评估，涵盖十一类自动生成的任务类别，产生了18,630对GPT-4o-mini的完成结果。使用BLEU、ROUGE-1/2/L、BERTScore-F1和SentenceBERT余弦相似度来测量输出保留度。最激进的配置在与原始提示输出的BERTScore-F1为0.876的情况下，平均标记减少了40.3%（σ = 9.2）；而仅停用词配置在0.913的情况下实现了29.6%的减少。每个任务类别的压缩与保真度的帕累托前沿特征化显示，在激进压缩下常识推理是一种系统性失败模式。所有代码、提示和每个单元结果均已发布以便于重复实验。

View on arXiv Download PDF AI Translation

cs.CL / 4 / 2609.13158

TestHallVQA: Exploring LVLMs' Document-Level Reasoning under Redundant Contexts from Scientific Exams

TestHallVQA：探索大型视觉语言模型在科学考试中冗余上下文下的文档级推理

Yu, Yongqi, Zhang, Yu

Abstract

Large Vision--Language Models (LVLMs) are increasingly expected to perform visual question answering (VQA) over planar media. However, existing planar VQA benchmarks typically emphasize isolated challenges: some emphasize long-document understanding with limited reasoning depth, while others require complex visual reasoning but remain restricted to single-page, noise-free settings. Moreover, through theoretical analysis, we identify the impact of irrelevant visual tokens, which leads to measurable performance degradation but has received little attention with respect to systematic quantification. To address these limitations, we introduce TestHallVQA, a multi-image VQA benchmark that simultaneously embodies document-level scale and the difficulty of human examinations, while providing comprehensive task coverage. Leveraging TestHallVQA's ability to controllably inject multi-level contextual redundancy, we further propose a novel metric, F1-R\textsuperscript{2}, which jointly quantifies LVLMs' computational reasoning capability and their evidence retrieval robustness against document-level redundancy. Extensive experiments and analyses on mainstream LVLMs reveal their latent deficiencies across multiple dimensions, offering concrete insights and directions for future research. The associated datasets, code, and complete theoretical derivations are available at https://github.com/yqyu2317/TestHallVQA-benchmark.

Chinese Translation

大型视觉语言模型（LVLMs）越来越被期望在平面媒体上执行视觉问答（VQA）。然而，现有的平面VQA基准通常强调孤立的挑战：一些基准强调长文档理解，但推理深度有限，而另一些则要求复杂的视觉推理，但仍然局限于单页、无噪声的环境。此外，通过理论分析，我们识别出无关视觉标记的影响，这导致可测量的性能下降，但在系统量化方面受到的关注很少。为了解决这些局限性，我们引入了TestHallVQA，这是一个多图像VQA基准，既体现了文档级规模，又反映了人类考试的难度，同时提供了全面的任务覆盖。利用TestHallVQA在可控注入多级上下文冗余方面的能力，我们进一步提出了一种新颖的度量标准F1-R²，该标准共同量化了LVLMs的计算推理能力及其在文档级冗余下的证据检索鲁棒性。对主流LVLMs进行的广泛实验和分析揭示了它们在多个维度上的潜在缺陷，为未来研究提供了具体的见解和方向。相关数据集、代码和完整的理论推导可在 https://github.com/yqyu2317/TestHallVQA-benchmark 获取。

View on arXiv Download PDF AI Translation

cs.CL / 5 / 2609.13238

Clinical Reasoning Under a Partially Observed Objective in Cone Beam CT Report Generation

在部分观察目标下的临床推理：锥束计算机断层扫描报告生成

George, Ajo Babu, Arun, Govind, Krishna, Sidharth N, Ranjan, Uma

Abstract

Maxillofacial report generation from cone beam computed tomography is scored here by a composite objective placing 80% of its weight on a large language model judgement of factual entailment and 20% on lexical overlap, of which only the lexical fifth is visible during development. The grader's BLEU-4 and METEOR routines are reproduced in pure Python and match the reference to machine precision, and an offline entailment surrogate, which tells a report written for one patient from one written for another at an area under the curve of 0.987, makes the composite objective cheap enough to optimise directly. Over the 622-case public release, a report selected against the visible lexical ranking scores 0.2909, whereas one selected against the composite objective scores 0.4122, because pursuing n-gram overlap drives entailment precision from 0.522 down to 0.266. A 29 million parameter encoder fine-tuned on the release reaches a prevalence-weighted out-of-fold area under the curve of 0.486 over 985 statements, indistinguishable from the corpus prior, while nine numbers read from the image header reach 0.945 for mandible coverage and 0.872 for condyle coverage, and acquisition centre alone predicts sentence choice at 0.718 against 0.663 for the image-derived model, identifying dictation convention rather than anatomy as the quantity the lexical metrics reward. The delivered system emits eight unconditional statements and five gated on header geometry under polarity, laterality and tooth-level consistency constraints, and reaches METEOR 0.3542 over 50 held-out cases from an unseen centre. The dataset and code are available at https://github.com/GIND123/CBCT-Clinical-Reasoner

Chinese Translation

本文对锥束计算机断层扫描生成的颌面部报告进行了评分，采用了一种复合目标，其中80%的权重基于大型语言模型对事实蕴含的判断，20%基于词汇重叠，而在开发过程中，仅有词汇部分是可见的。评分者的BLEU-4和METEOR算法在纯Python中重现，并与参考标准达到机器精度的匹配。此外，离线蕴含替代模型能够在曲线下面积为0.987的情况下，将针对一个患者撰写的报告与针对另一个患者撰写的报告进行区分，使得复合目标的优化变得足够简单。在622个案例的公开发布中，基于可见词汇排名选择的报告得分为0.2909，而基于复合目标选择的报告得分为0.4122，因为追求n-gram重叠将蕴含精度从0.522降低至0.266。经过微调的2900万参数编码器在985个陈述上达到0.486的流行加权折叠外曲线下面积，与语料库先验无差异，而从图像头信息读取的九个数字在下颌骨覆盖率上达到0.945，在髁突覆盖率上达到0.872，单独的采集中心对句子选择的预测准确率为0.718，而图像衍生模型为0.663，识别出记录约定而非解剖结构是词汇指标所奖励的量。交付的系统在极性、侧向性和牙齿级一致性约束下，发出八个无条件陈述和五个基于头部几何形状的陈述，并在50个来自未见中心的保留案例中达到METEOR 0.3542。数据集和代码可在 https://github.com/GIND123/CBCT-Clinical-Reasoner 获取。

View on arXiv Download PDF AI Translation

cs.CL / 6 / 2609.13389

RFCLLM: Evaluating LLMs' Reasoning Ability of Network Protocol State Machines

RFCLLM：评估大型语言模型（LLMs）对网络协议状态机的推理能力

Chen, Anqi, Goldwasser, Dan, Nita-Rotaru, Cristina

Abstract

Mapping textual specifications into formal representations is essential for ensuring the correctness of protocol designs and implementations. LLM-generated mappings, used for networking security or testing, are assumed to capture a perfect understanding of the specification, which may not hold in practice. The goal of this paper is to assess the extent to which LLMs can interpret the specification correctly. We examine the degree to which an LLM's implicit representation of a finite-state transition system-defined via natural language descriptions-aligns with a manually generated ground-truth model. We designed 4 tasks and 1482 task queries for 16 protocols. We evaluated different judge biases, observed the inherent difficulty gaps between tasks, looked into the effect of 4 context types, and the influence of protocol characteristics. Our work contributes to a step toward verifying whether LLMs can really be trusted in FSM (Finite State Machine) reasoning of protocol specifications.

Chinese Translation

将文本规范映射为形式化表示对于确保协议设计和实现的正确性至关重要。假设LLM生成的映射能够完美理解规范，这在网络安全或测试中被广泛使用，但在实际应用中可能并不成立。本文的目标是评估LLM在多大程度上能够正确解读规范。我们考察了LLM对通过自然语言描述定义的有限状态转移系统的隐式表示与手动生成的真实模型之间的一致性。我们为16个协议设计了4个任务和1482个任务查询。我们评估了不同评审偏见，观察了任务之间固有的难度差异，研究了4种上下文类型的影响以及协议特征的影响。我们的工作为验证LLM在协议规范的有限状态机（FSM）推理中是否真的值得信赖迈出了重要一步。

View on arXiv Download PDF AI Translation

cs.CL / 7 / 2609.13413

CVSS-X: A Multilingual Speech-to-Speech Translation Corpus for 28 Languages

CVSS-X：一个涵盖28种语言的多语言语音到语音翻译语料库

Gris, Lucas Rafael Stefanel, Ferreira, Alef Iury Siqueira, de Oliveira, Frederico Santos, da Rosa, Augusto Seben, Filho, Alexandre Costa Ferro, Filho, Arlindo Rodrigues Galvão, Soares, Anderson da Silva

Abstract

We introduce CVSS-X, a large-scale synthetic speech-to-speech translation corpus that extends CVSS by reversing the translation direction. While CVSS translates from 21 languages into English, CVSS-X enables translation from English into 28 target languages spanning 12 language families. The corpus comprises approximately 240,000 parallel speech pairs per language, totaling over 16,000 hours, eight times larger than CVSS. We provide two variants: CVSS-X-C with two canonical voices per language, and CVSS-X-T with cross-lingual voice cloning, both fully generated. Evaluation shows comparable translation quality to CVSS with consistent performance across typologically diverse languages. Combined with CVSS, this enables research on bidirectional and multilingual speech-to-speech translation. The code is available at https://github.com/ErmisAI/XVSS-X and the dataset under CC-BY-NC 4.0 license at https://huggingface.co/datasets/lgris/XVSS-X.

Chinese Translation

我们介绍了CVSS-X，一个大规模的合成语音到语音翻译语料库，通过逆转翻译方向扩展了CVSS。CVSS将21种语言翻译成英语，而CVSS-X则支持从英语翻译到涵盖12个语言家族的28种目标语言。该语料库每种语言包含约240,000对平行语音，总计超过16,000小时，规模是CVSS的八倍。我们提供了两种变体：CVSS-X-C，具有每种语言两种标准语音，以及CVSS-X-T，具有跨语言语音克隆，均为完全生成。评估结果显示，翻译质量与CVSS相当，并且在类型学上多样的语言中表现一致。结合CVSS，这使得双向和多语言语音到语音翻译的研究成为可能。代码可在https://github.com/ErmisAI/XVSS-X获取，数据集在CC-BY-NC 4.0许可证下可在https://huggingface.co/datasets/lgris/XVSS-X获取。

View on arXiv Download PDF AI Translation

cs.CL / 8 / 2609.13445

Causal Analysis and Mitigation of Spurious Onsets in Full-Duplex Speech LLMs

全双工语音大语言模型中虚假发声的因果分析与缓解

Nishi, Kento

Abstract

Speech-to-speech LLMs like Moshi, and its derivative PersonaPlex, can listen and speak concurrently through full-duplex generation. However, they can begin speaking inappropriately during prolonged user silence: under digital-zero input, Moshi and PersonaPlex initiate speech in 12/40 and 11/40 five-minute continuations, respectively. What causes this spurious speech? We investigate two hypotheses: either repeated sampling selects speech despite persistently low onset probabilities, or conditioning on the model's nonspeech outputs causes an abrupt spike in onset probability. We find that, at every observed onset, speech probability spikes by over nine orders of magnitude in one 80-ms frame, supporting the latter hypothesis. Then, to suppress these onsets without blocking genuine responses, we ask a causal counterfactual question: is the model responding to user speech, or would its next-token distribution remain similar if the preceding user input were muted? Accordingly, we suppress onsets whose distributions change little under this intervention. Across 40 held-out trials per model with realistic microphone noise, our method suppresses 13/13 Moshi and 9/9 PersonaPlex spurious onsets, while preserving 40/40 genuine responses per model. Our inference-time method requires no retraining and runs in real-time, with 95th-percentile decision time below 61 ms, within the 80-ms frame budget. Our code is available at https://github.com/KentoNishi/icassp27-spurious-onsets.

Chinese Translation

像Moshi及其衍生产品PersonaPlex这样的语音到语音大语言模型可以通过全双工生成同时进行听和说。然而，在用户长时间沉默期间，它们可能会不适当地开始说话：在数字零输入下，Moshi和PersonaPlex分别在40个五分钟的连续测试中启动了12次和11次发声。这种虚假发声的原因是什么？我们调查了两个假设：要么重复采样在持续低发声概率的情况下选择了发声，要么对模型的非语音输出进行条件化导致发声概率的突然激增。我们发现，在每次观察到的发声中，发声概率在一个80毫秒的帧内激增超过九个数量级，支持后者的假设。然后，为了在不阻碍真实响应的情况下抑制这些发声，我们提出了一个因果反事实问题：模型是在响应用户的语音，还是如果前面的用户输入被静音，其下一个令牌分布将保持相似？因此，我们抑制了在该干预下分布变化不大的发声。在每个模型的40个保留试验中，结合现实的麦克风噪声，我们的方法抑制了13/13个Moshi和9/9个PersonaPlex的虚假发声，同时保留了每个模型的40/40个真实响应。我们的推断时间方法无需重新训练，并且实时运行，95百分位决策时间低于61毫秒，符合80毫秒的帧预算。我们的代码可在https://github.com/KentoNishi/icassp27-spurious-onsets获取。

View on arXiv Download PDF AI Translation

cs.CL / 9 / 2609.13454

Hindsight Bias in Clinical Temporal Reasoning: How Future Data Exposure Affects Large Language Model Judgment

临床时间推理中的事后偏见：未来数据曝光如何影响大型语言模型的判断

Matsuura, Misaki, Kumar, Sayantan, Kadam, Ojas, Weiss, Jeremy C.

Abstract

Clinical decisions are prospective, but clinical language models are often evaluated on retrospective records that reveal the final diagnosis, treatment response, and outcome. Such evaluations may reward the use of future information rather than reasoning under the uncertainty present at the decision point. We introduce a paired benchmark for measuring outcome-conditioned shifts consistent with hindsight bias in clinical temporal reasoning. It contains 171 case reports from the PubMed Central Open Access Subset---40 sepsis and 131 GLP-1/diabetes cases---represented as both textual narratives and human-annotated and LLM-generated textual time series (TTS). For each case, questions are tied to a clinically meaningful cutoff and paired with a prospective reference answer and an outcome-consistent \emph{hindsight trap}. Models answer each question using either a TTS truncated at the cutoff or the complete timeline; additional conditions vary the narrative source (original or synthetic) and TTS annotation source (human or LLM). We evaluate accuracy (Acc), hindsight trap rate (HTR), answer instability rate (AIR), and hindsight bias rate (HBR), each of which captures different signals of hindsight bias. Across GPT 5.6 Sol, Gemma 4, GLM 5.2, and Opus 5, full timeline exposure produces consistent hindsight-sensitive shifts, while temporal masking reduces bias without lowering accuracy.

Chinese Translation

临床决策是前瞻性的，但临床语言模型通常在揭示最终诊断、治疗反应和结果的回顾性记录上进行评估。这种评估可能会奖励使用未来信息，而不是在决策时点所存在的不确定性下进行推理。我们引入了一种配对基准，用于测量与事后偏见一致的结果条件变化，专注于临床时间推理。该基准包含来自PubMed Central开放获取子集的171个病例报告——40个脓毒症病例和131个GLP-1/糖尿病病例——以文本叙述和人类注释及LLM生成的文本时间序列（TTS）两种形式表示。对于每个病例，问题与临床意义重大的截止点相关，并与前瞻性参考答案和结果一致的 extit{事后陷阱}配对。模型使用截止点处截断的TTS或完整时间线回答每个问题；附加条件变化叙述来源（原始或合成）和TTS注释来源（人类或LLM）。我们评估准确性（Acc）、事后陷阱率（HTR）、答案不稳定率（AIR）和事后偏见率（HBR），每个指标捕捉事后偏见的不同信号。在GPT 5.6 Sol、Gemma 4、GLM 5.2和Opus 5中，完整时间线曝光产生了一致的对事后敏感的变化，而时间遮蔽则在不降低准确性的情况下减少了偏见。

View on arXiv Download PDF AI Translation

cs.CL / 10 / 2609.13481

A Hybrid Hierarchical 1D-CNN-BiLSTM Framework for Extractive Summarization of Biomedical and Clinical Text

一种混合层次的1D-CNN-BiLSTM框架用于生物医学和临床文本的提取式摘要

Ather, Saad Bin, Saif, Muhammad, Khan, Ali Hassan, Abbas, Manzer, Waheed, Hajra

Abstract

Large language models have made abstractive summarization remarkably fluent, but generated summaries can hallucinate facts, posing serious risks in biomedical and clinical domains. We address this by removing generation from the pipeline and framing summarization as extractive sentence selection. Our Hybrid Hierarchical CNN-LSTM Summarizer uses stacked multi-kernel convolutions to compose sentence-level embeddings into richer inter-sentence representations, followed by a bidirectional LSTM to model long-range dependencies across the document. A lightweight scoring head assigns per-sentence importance scores and is trained end-to-end with binary cross-entropy against oracle extractive labels. At inference, a dynamic mean-plus-standard-deviation threshold with a top-3 fallback selects sentences directly from the source and chronologically reorders them into the final summary. Since every output sentence is copied from the input, the model avoids generation-induced factual drift. On PubMed, our architecture outperforms isolated CNN and LSTM baselines, while ablations show that wider convolutional receptive fields improve sentence scoring. On MIMIC-CXR and MIMIC-IV BHC, the model performs well on unstructured narratives but defaults toward positional baselines on highly templated reports. These results suggest that structural constraints can provide a reliable path toward factually grounded summarization systems that are trustworthy by design rather than by correction.

Chinese Translation

大型语言模型使得抽象摘要变得非常流畅，但生成的摘要可能会虚构事实，这在生物医学和临床领域带来了严重风险。我们通过去除生成步骤，将摘要框架设定为提取式句子选择来解决这个问题。我们的混合层次CNN-LSTM摘要生成器使用堆叠的多核卷积将句子级嵌入组合成更丰富的句间表示，接着使用双向LSTM来建模文档中的长距离依赖关系。一个轻量级评分头为每个句子分配重要性分数，并通过与权威提取标签的二元交叉熵进行端到端训练。在推理阶段，使用动态的均值加标准差阈值和前3名回退机制直接从源文本中选择句子，并按时间顺序重新排序成最终摘要。由于每个输出句子都是从输入中复制的，该模型避免了生成引起的事实漂移。在PubMed上，我们的架构优于孤立的CNN和LSTM基线，而消融实验表明，更宽的卷积感受野可以改善句子评分。在MIMIC-CXR和MIMIC-IV BHC上，该模型在非结构化叙述上表现良好，但在高度模板化的报告中则倾向于位置基线。这些结果表明，结构约束可以为建立基于事实的可靠摘要系统提供可靠路径，这些系统在设计上是可信的，而不是通过纠正来实现的。

View on arXiv Download PDF AI Translation

cs.CL / 11 / 2609.13520

From Token Probabilities to Semantic Constraints: Towards Declarative Probabilistic Evaluation of Language Models

从标记概率到语义约束：朝着声明性概率评估语言模型的方向

Richardson, Kyle, Anderson, Cullen, Balakrishnan, Pranav, Ban, Takuto, Ladia, Daksha, Gupta, Ankita, Hudspeth, Marisa

Abstract

While Large Language Models have improved rapidly, many fundamental questions remain about how to evaluate the knowledge and reasoning abilities they acquire, and how such evaluations relate to the learning signals used in pre-training. In this paper, we propose ModelLog, a declarative probabilistic framework for pre-training evaluation that makes the semantic structure of model behavior explicit and provides new formal tools for relating evaluation to learning. ModelLog specifies evaluation targets as symbolic constraints over token-level predictions and measures how strongly a model's distribution satisfies those constraints. We explore the framework through a new suite of tasks targeting negation, mutual exclusivity, and consistency, finding systematic failures that are difficult to characterize through token likelihood or answer accuracy alone. We further show that these evaluation scores can also be interpreted as losses, whose gradients reflect logical strength, informativeness, and variable-level sensitivity. This links evaluation and learning through a shared semantics, suggesting evaluation methods that diagnose model behavior while also helping to clarify the semantic structure of learning.

Chinese Translation

尽管大型语言模型的进步迅速，但关于如何评估它们所获得的知识和推理能力，以及这些评估与预训练中使用的学习信号之间的关系，仍然存在许多基本问题。本文提出了ModelLog，一种声明性概率框架，用于预训练评估，该框架明确了模型行为的语义结构，并提供了将评估与学习相关联的新形式工具。ModelLog将评估目标指定为对标记级预测的符号约束，并衡量模型的分布在多大程度上满足这些约束。我们通过一套新的任务组合探索该框架，这些任务针对否定、互斥性和一致性，发现了一些系统性失败，这些失败仅通过标记似然或答案准确性难以表征。我们进一步表明，这些评估分数也可以解释为损失，其梯度反映了逻辑强度、信息量和变量级敏感性。这将评估与学习通过共享语义联系起来，提出了能够诊断模型行为的评估方法，同时有助于澄清学习的语义结构。

View on arXiv Download PDF AI Translation

cs.CL / 12 / 2609.13534

Harmfulness Propagation Dynamics: Layer-wise Trajectories of Adversarial Intent in Large Language Models

有害性传播动态：大型语言模型中对抗意图的层级轨迹

Mohammad, Noor Islam S., Bayazıt, Uluğ

Abstract

We identify \textbf{Harmfulness Propagation Dynamics (HPD)}: for harmful prompts, the projection of the last-token hidden state onto a learned harm direction rises monotonically with transformer depth, whereas benign prompts remain flat or oscillatory. This cross-layer signature reflects harmful intent as a \emph{progressively resolved} semantic property: surface form appears early, while pragmatic intent consolidates later, making the \emph{trajectory shape} more informative than any single-layer snapshot. Moreover, LDA-based harm directions, learned per layer, remain stable across random splits (pairwise cosine similarity $>0.97$), supporting the projection sequence as a reproducible structured signal. Building on HPD, we introduce \textbf{\herald{}} (\textbf{H}armful \textbf{E}ncoding \textbf{R}ecognition via \textbf{A}ctivation \textbf{L}ayer \textbf{D}ynamics). This lightweight input moderator extracts a seven-dimensional feature record, slope, curvature, monotonicity, onset layer, and related statistics from the cross-layer projection sequence and classifies it with a 288-parameter MLP. \herald{} stores one $d$-dimensional direction per layer ($262$\,KB for a 32-layer, $d{=}4096$ model), requires no gradient computation during training, and adds only $2.6{\times}10^{-6}$ prefill FLOPs at inference. Across eight prompt-harmfulness benchmarks and four model families, \herald{} achieves an average F1 of $89.3$ on OLMo2-7B, surpassing all tested guard models on adversarial jailbreak detection ($98.4$ vs.\ $96.9$ F1) and outperforming prior latent-based methods by $2.3$-$4.1$ F1 points on every backbone. Per-instance trajectories provide machine-readable audit records that reveal \emph{when} and \emph{how} harmfulness emerges, offering an interpretability advantage over single-layer approaches.

Chinese Translation

我们识别出 extbf{有害性传播动态（Harmfulness Propagation Dynamics, HPD）}：对于有害提示，最后一个token的隐藏状态在学习到的有害方向上的投影随着变换器深度单调上升，而良性提示则保持平坦或振荡。这种跨层特征反映了有害意图作为一种 extit{逐步解析}的语义属性：表面形式早期出现，而务实意图则在后期巩固，使得 extit{轨迹形状}比任何单层快照更具信息量。此外，基于LDA的有害方向在每层学习后在随机拆分中保持稳定（成对余弦相似度 $>0.97$），支持投影序列作为可重复的结构信号。在HPD的基础上，我们引入了 extbf{ extit{Herald}}（ extbf{H}armful extbf{E}ncoding extbf{R}ecognition via extbf{A}ctivation extbf{L}ayer extbf{D}ynamics）。这个轻量级输入调节器从跨层投影序列中提取七维特征记录，包括斜率、曲率、单调性、起始层及相关统计信息，并使用一个288参数的多层感知器（MLP）进行分类。 extit{Herald}为每层存储一个$d$维方向（对于一个32层、$d{=}4096$的模型，大小为$262$ KB），在训练过程中不需要梯度计算，并且在推理时仅增加$2.6{ imes}10^{-6}$的预填充FLOPs。在八个提示有害性基准和四个模型家族中， extit{Herald}在OLMo2-7B上实现了平均F1值$89.3$，在对抗性越狱检测中超越了所有测试的防护模型（$98.4$对$96.9$ F1），并在每个骨干网络上超越了先前的基于潜变量的方法$2.3$-$4.1$ F1点。每个实例的轨迹提供了机器可读的审计记录，揭示了有害性何时以及如何出现，提供了相较于单层方法的可解释性优势。

View on arXiv Download PDF AI Translation

cs.CL / 13 / 2609.13556

Domain-Specific Jargon in Large Language Models: A Comparative Analysis between General-Purpose and Specialist Models

大型语言模型中的领域特定术语：通用模型与专业模型的比较分析

Keng, Darin, Sun, Zhewei

Abstract

Large Language Models (LLMs) have shown remarkable proficiency on general-purpose tasks, yet their performance often degrades in highly-specialized technical domains. Moreover, little is known about how parametric knowledge of domain-specific terms is encoded within these models. We address this gap by contributing two novel medical jargon evaluation benchmarks and evaluate a general-purpose Llama-3.1 model against a variant fine-tuned on medical-domain data. Surprisingly, the general-purpose model outperforms the medically fine-tuned model on both tasks. Using mechanistic interpretability tools, we find systematic patterns of miscalibration for the medically fine-tuned model. Instead of reorganizing parametric knowledge, the fine-tuned model places greater emphasis on a small subset of model components associated with jargon-favoring predictions. We find that applying component reweighting strategies against the benchmark tasks successfully suppresses these components and closes the gap with the general-purpose baseline. We also observe that some jargon-sensitive components transfer knowledge to the same tasks involving materials science jargon, suggesting they encode a partially domain-agnostic notion of specialized terminology. Our results provide a case study in which a medically fine-tuned checkpoint does not improve jargon comprehension over its general-purpose counterpart, highlighting that domain adaptation should not be assumed to yield better performance on specialized terminology.

Chinese Translation

大型语言模型（LLMs）在通用任务上表现出显著的能力，但在高度专业化的技术领域中，它们的性能往往会下降。此外，目前对领域特定术语的参数知识在这些模型中的编码方式知之甚少。为填补这一空白，我们贡献了两个新的医学术语评估基准，并将通用的 Llama-3.1 模型与一个在医学领域数据上微调的变体进行了评估。令人惊讶的是，通用模型在这两个任务上的表现优于医学微调模型。通过使用机制可解释性工具，我们发现医学微调模型存在系统性的误校准模式。微调模型并没有重新组织参数知识，而是更加强调与术语偏好预测相关的小部分模型组件。我们发现，针对基准任务应用组件重加权策略成功抑制了这些组件，并缩小了与通用基线的差距。我们还观察到，一些对术语敏感的组件将知识转移到涉及材料科学术语的相同任务中，表明它们编码了部分领域无关的专业术语概念。我们的结果提供了一个案例研究，表明医学微调的检查点在术语理解方面并未优于其通用对应物，强调领域适应不应被假定能在专业术语上产生更好的性能。

View on arXiv Download PDF AI Translation

cs.CL / 14 / 2609.13581

Toward Complete Hospital Discharge Summarization with Abstract Meaning Representation

朝着完整的医院出院总结与抽象意义表示的方向

Landes, Paul, Rao, Sitara, Chaise, Aaron Jeremy, Di Eugenio, Barbara

Abstract

Discharge summaries are lengthy medical documents that summarize a hospital in-patient visit. Automatically generating them can reduce documentation burden and return clinician time to patient care. Whereas Large Language Model (LLMs) could be used for this task, their Achilles heel is hallucinations, which can have drastic consequences for clinical documentation. We present an evidence-driven alignment framework for discharge summarization at the clinical encounter level, that treats provenance as a first-class constraint, using semantic graphs and deep learning models. Each summary sentence is selected and organized via cross-document semantic alignment and is accompanied by explicit evidence links to its source spans. We show our results on two corpora: a publicly available corpus (MIMIC-III) and clinical notes written by physicians at the University of Illinois Hospital (UIC Health). Additionally, we make source code and trained models available.

Chinese Translation

出院总结是总结住院病人就诊情况的冗长医疗文件。自动生成这些文件可以减轻文档负担，并将临床医生的时间重新投入到患者护理中。尽管大型语言模型（Large Language Models, LLMs）可以用于此任务，但它们的致命弱点是幻觉现象，这可能对临床文档产生严重后果。我们提出了一种以证据为驱动的出院总结对齐框架，针对临床接触层面，视来源为一项首要约束，采用语义图和深度学习模型。每个总结句子通过跨文档语义对齐进行选择和组织，并附有明确的证据链接到其来源跨度。我们在两个语料库上展示了我们的结果：一个公开可用的语料库（MIMIC-III）和伊利诺伊大学医院（UIC Health）医生撰写的临床笔记。此外，我们还提供了源代码和训练模型。

View on arXiv Download PDF AI Translation

cs.CL / 15 / 2609.13582

Same Patient, Different Order: Action-Level Reliability of Clinical LLM Agents Under Repeated Runs

同一患者，不同顺序：临床大语言模型代理在重复运行中的行动级可靠性

Bellibatlu, Rohith Reddy, Singh, Manpreet, Han, Zhoutian, Zhang, Wenbin

Abstract

A clinical agent benchmark can report the same verdict on identical inputs while the agent files a materially different order on each run. Such agents order tests, request medications and place referrals, yet benchmarks typically score one run per task and rarely ask whether identical inputs produce identical actions; MedAgentBench, the benchmark we use, scores a single attempt and says so. To measure this gap we introduce "same-input rerun", which replays a task with every input held fixed and compares the orders rather than the score, with six reliability metrics, and apply it to 1000 MedAgentBench runs across 50 tasks from its five write-capable families, two open-weight models below ten billion parameters quantised to four bits, and two temperatures. The study establishes that action-level divergence exists and can pass unrecorded by the score, not that any rate generalises. Under the 8B model at temperature 0.7, all 43 ordering groups emit a different set of orders across five identical runs, 26 emit the order on some runs and not others, and 28 record a different coded value, dose or analyte. In 22 of those 43 the benchmark reports the same failing verdict for materially different behaviour, as it does for all 10 divergent groups of the 4B model at 0.7. Orders also reach different endpoints across runs, one of which the record server rejects while the agent is told it succeeded. These findings motivate repeated-run evaluation, action-level stability reporting and execution-faithful environment feedback in clinical-agent benchmarks.

Chinese Translation

临床代理基准可以在相同输入上报告相同的判决，而代理在每次运行中却提交了实质上不同的顺序。这些代理会安排测试、请求药物和进行转诊，但基准通常仅对每个任务评分一次，且很少询问相同输入是否产生相同的行动；我们使用的基准 MedAgentBench 仅对单次尝试进行评分并明确说明。为了衡量这一差距，我们引入了“相同输入重跑”，该方法在每个输入固定的情况下重放任务，并比较顺序而非分数，使用六个可靠性指标，并将其应用于 1000 次 MedAgentBench 运行，涵盖其五个可写家族中的 50 个任务，以及两个低于十亿参数的开放权重模型，量化为四位，和两个温度设置。研究表明，行动级的分歧确实存在，并且可能在评分中未被记录，而没有任何比率是普遍适用的。在 8B 模型下，温度为 0.7 时，所有 43 个排序组在五次相同运行中发出不同的订单，其中 26 个在某些运行中发出订单而在其他运行中不发出，28 个记录了不同的编码值、剂量或分析物。在这 43 个组中，有 22 个的基准报告了相同的失败判决，尽管行为实质上不同，4B 模型在 0.7 的所有 10 个分歧组也是如此。订单在运行中也达到不同的终点，其中一个被记录服务器拒绝，而代理却被告知成功。这些发现促使我们在临床代理基准中进行重复运行评估、行动级稳定性报告和执行忠实环境反馈。

View on arXiv Download PDF AI Translation

cs.CL / 16 / 2609.13611

In the Blind: Building Pseudo-References for MT Evaluation

盲评：构建机器翻译评估的伪参考

Kanojia, Diptesh, Lo, Chi-kiu, Sindhujan, Archchana, Larkin, Samuel, Hanneman, Greg, Lavie, Alon

Abstract

The WMT26 General MT task evaluates systems on 10 language pairs that have no human references (neither translated from scratch nor post-edited from MT output by humans). We describe how we built the pseudo-references for these pairs and six other language pairs (in which some forms of human references are available): seven models translate the 3,277 official documents under up to five prompt conditions, giving a total of 26 system-prompt combinations; then three reference-free quality estimation (QE) models score every candidate; and a per-document selector picks one translation, which GPT-5.5 post-edits where needed. Working without references exposed a failure mode of QE-guided selection: the metrics rank fluent output in the wrong language above correct translations. Adding a confidence-scaled language identification penalty to the score fusion drives the wrong-language count to zero, and the resulting selector still scores better on MetricX than the rank-fusion baseline it replaces. Since no references were available for these pairs while we were building them, we calibrate every selection decision on last year's WMT25 human judgments. The human evaluation, released after construction, shows the cost of getting selection wrong: our references stand with the strongest participating systems when the selector kept a frontier-model candidate, and fall up to 17 ESA points below them when it did not. We release the selection method and the provenance of every reference (https://github.com/surrey-nlp/PseudoRef)

Chinese Translation

WMT26一般机器翻译任务评估10对没有人类参考的语言对（既没有从头翻译也没有经过人类后期编辑的机器翻译输出）。我们描述了如何为这些语言对以及另外六对（其中一些形式的人类参考是可用的）构建伪参考：七个模型在多达五个提示条件下翻译3,277份官方文件，总共形成26种系统-提示组合；然后三个无参考质量估计（QE）模型对每个候选翻译进行评分；最后，一个逐文档选择器选择一个翻译，并在需要时由GPT-5.5进行后期编辑。在没有参考的情况下工作暴露了QE引导选择的失败模式：度量标准将流利的错误语言输出排在正确翻译之上。通过在评分融合中添加一个信心缩放的语言识别惩罚，将错误语言的计数降至零，而结果选择器在MetricX上的得分仍然优于其替代的排名融合基线。由于在构建这些伪参考时没有可用的参考，我们将每个选择决策校准为去年的WMT25人类评估。构建完成后发布的人类评估显示了选择错误的代价：当选择器保留一个前沿模型候选时，我们的参考与最强的参与系统并列，而当选择器未保留时则低于它们最多17个ESA点。我们发布了选择方法及每个参考的来源（https://github.com/surrey-nlp/PseudoRef）

View on arXiv Download PDF AI Translation

cs.CL / 17 / 2609.13615

The University of Melbourne WMT 2026 CreoleMT Submission: A Domain-Balanced Approach to Low-Resource Pacific Creole Machine Translation

墨尔本大学 WMT 2026 CreoleMT 提交：一种面向低资源太平洋克里奥尔语机器翻译的领域平衡方法

Merx, Raphaël, Thieberger, Nick, Vylomova, Ekaterina

Abstract

For our submission to the WMT26 Creole Language Translation Shared Task, we focus on machine translation (MT) models for Pacific creoles: Tok Pisin, Bislama, and Solomon Pijin, with particular attention to broad domain performance. After pre-training on a large collection of domain-imbalanced data, we continue fine-tuning on a diverse mix of domain-balanced data. We rely on a number of data collection and preparation techniques, including LLM-assisted respelling and alignment, back-translation, and distillation from Gemini for domains originally not present in training data. Evaluated on Bouquet and a novel test set made of spoken language transcripts, our models beat open model baselines by 3+ chrF++ points in all directions with human-original references. Looking ahead, we plan to develop human-translated test sets for Solomon Pijin and Bislama, and to distil our best models into much smaller ones that retain broad domain coverage.

Chinese Translation

在我们提交给 WMT26 克里奥尔语言翻译共享任务的过程中，我们专注于太平洋克里奥尔语的机器翻译（MT）模型：Tok Pisin、Bislama 和 Solomon Pijin，特别关注广泛领域的性能。在对大量领域不平衡数据进行预训练后，我们继续在多样化的领域平衡数据上进行微调。我们依赖多种数据收集和准备技术，包括 LLM 辅助的重拼写和对齐、回译以及从 Gemini 中提炼出原本不在训练数据中的领域。我们的模型在 Bouquet 和一个由口语语言转录组成的新测试集上评估，与人类原始参考相比，在所有方向上超过开放模型基线 3 个以上的 chrF++ 点。展望未来，我们计划为 Solomon Pijin 和 Bislama 开发人类翻译的测试集，并将我们最好的模型提炼为更小的模型，同时保持广泛的领域覆盖。

View on arXiv Download PDF AI Translation

cs.CL / 18 / 2609.13682

LayerRoute: Adaptive Layer-Skipping with LoRA-Preserved Quality for Efficient LLM Inference

LayerRoute：具有LoRA保留质量的自适应层跳过方法以提高大规模语言模型推理效率

Sikdar, Prateek Kumar

Abstract

We introduce LayerRoute, a parameter-efficient method for adaptive transformer layer-skipping that combines per-layer hard-gated routing (trained via a straight-through estimator) with joint LoRA fine-tuning. LayerRoute augments each of the 24 transformer blocks in Qwen2.5-0.5B-Instruct with a lightweight per-layer router (~21.5K parameters) and LoRA adapters (rank 8, ~1.08M parameters), training both jointly under a gate-regularized language-modeling objective. Across 10 independently-seeded training runs, LayerRoute converges to an identical skip-pattern structure in every run - a consistent set of 9 middle layers (8-16) becomes skip-eligible in all 10 seeds - and delivers genuine, verified wallclock speedup in every run (1.02x-1.06x, mean 1.04x). Quality is preserved or improved in every configuration tested: joint LoRA adaptation yields a perplexity improvement over the unmodified backbone in all 10 seeds (mean delta = -1.16 and -1.11 across the two evaluation splits used). We further verify the router performs genuine, non-trivial per-input computation: gate decisions in skip-eligible layers change the actual skip/run outcome for 87-100% of held-out samples, confirming real input-dependent routing rather than a fixed pruning pattern. LayerRoute trains in under 7 minutes on a single A100 and adds negligible overhead beyond the routing decision itself. We report our full reproducibility methodology, including a systematic diagnostic investigation into what determines the router's per-input decisions, as part of this work.

Chinese Translation

我们介绍了LayerRoute，这是一种参数高效的自适应变换器层跳过方法，结合了逐层硬门控路由（通过直通估计器训练）与联合LoRA微调。LayerRoute为Qwen2.5-0.5B-Instruct中的每个24个变换器块增添了一个轻量级的逐层路由器（约21.5K参数）和LoRA适配器（秩8，约1.08M参数），在门控正则化的语言建模目标下共同训练。在10次独立种子的训练中，LayerRoute在每次训练中收敛到相同的跳过模式结构——在所有10个种子中，一组一致的9个中间层（8-16）变得符合跳过条件——并在每次训练中实现了真实、经过验证的墙钟加速（1.02x-1.06x，平均1.04x）。在测试的每种配置中，质量得以保持或提高：联合LoRA适应在所有10个种子中相较于未修改的主干模型实现了困惑度的改善（平均变化 = -1.16 和 -1.11，基于使用的两个评估拆分）。我们进一步验证路由器在每个输入上执行真实且非平凡的计算：在符合跳过条件的层中，门控决策改变了87-100%保留样本的实际跳过/运行结果，确认了真实的输入依赖路由，而非固定的剪枝模式。LayerRoute在单个A100上训练时间少于7分钟，并且在路由决策之外增加的开销微乎其微。我们在本研究中报告了完整的可重复性方法，包括对路由器每个输入决策的决定因素进行系统诊断调查。

View on arXiv Download PDF AI Translation

cs.CL / 19 / 2609.13685

Not all Negation Cues are Equal: Affixal Negations Yield Better Negation Understanding

并非所有否定线索都是平等的：词缀否定带来更好的否定理解

Tan, Tian, Blanco, Eduardo

Abstract

Negation remains a longstanding challenge for both language models (LMs) and large language models (LLMs). Prior work mainly focuses on a small set of high-frequency single-word negation cues, such as not and never, with limited exploration of broader negation types and modern LLMs. To address this gap, we construct NegCue, a large-scale dataset containing over 1.8M samples spanning single-word, multi-word, and affixal negation with more than 200 unique cues. We further pre-train both encoder-only LMs and LLMs on NegCue to investigate how different negation types affect negation understanding. Experiments on five downstream benchmarks show that negation types contribute unevenly to performance gains under the same training scale. In particular, affixal negation yields the largest improvements, while the gains from the commonly studied single-word negation remain modest. Moreover, our results demonstrate that further pre-training improves negation understanding for both LMs and LLMs.

Chinese Translation

否定一直是语言模型（LMs）和大型语言模型（LLMs）面临的长期挑战。之前的研究主要集中在一小部分高频的单词否定线索上，如 not 和 never，且对更广泛的否定类型和现代 LLMs 的探索有限。为了解决这一问题，我们构建了 NegCue，这是一个大规模数据集，包含超过 180 万个样本，涵盖单词、短语和词缀否定，具有超过 200 个独特的线索。我们进一步在 NegCue 上对仅编码器的 LMs 和 LLMs 进行预训练，以研究不同否定类型如何影响否定理解。在五个下游基准测试中的实验表明，在相同的训练规模下，否定类型对性能提升的贡献是不均匀的。特别是，词缀否定带来了最大的改善，而常见的单词否定的提升则相对有限。此外，我们的结果表明，进一步的预训练提高了 LMs 和 LLMs 的否定理解能力。

View on arXiv Download PDF AI Translation

cs.CL / 20 / 2609.13709

When Edit Localization Amplifies Relative Selection Bias: Gradient Geometry, Target Mismatch, and Importance Weighting

当编辑本地化放大相对选择偏差时：梯度几何、目标不匹配和重要性加权

Zhang, Shengwei, Dai, Haoda, Li, Yifei, Song, Yuheng

Abstract

Human corrections identify editable spans, but the examples receiving corrections may come from a selective feedback channel. We analyze this interaction at a fixed model checkpoint by decomposing a localized gradient into edited and retained untouched components. Squared relative selection bias is a ratio of quadratics whose derivative has the sign of an explicit quadratic polynomial. Localization can increase, decrease, or nonmonotonically change this diagnostic; its direction depends on component biases and geometry. Oracle importance weighting recovers the population mean for each fixed localization objective, but these objectives have different targets. Against one common full-gradient target, we derive the finite-sample mean-squared error, an analytic optimal retention coefficient, and a fixed-clipping extension. Exact finite-population calculations and 10,000 Monte Carlo repetitions per sample size verify the identities and counterexamples. Public human-post-edit experiments use two translation directions and pretrained models, with declared synthetic selection. An English-German extension differentiates 73.89 million native parameters. In all three declared settings, hard localization has higher relative bias but lower absolute bias than full retention. Untouched-component biases are nonzero and selected component means have negative inner products, so the general criterion applies where the simple unbiased/aligned explanation fails. Output-bias diagnostics show the same endpoint ordering of relative bias across both directions, with one interior maximum. Mechanisms reuse each language's records and include a mixture; they are not independent replications. The evidence separates relative amplification from absolute gradient error and establishes estimation properties, without inferring translation-quality gains or identifying actual complaint propensities.

Chinese Translation

人类修正识别可编辑的范围，但接收修正的示例可能来自选择性反馈通道。我们通过将局部梯度分解为编辑和保留未触及的组件，在固定模型检查点分析这种交互。平方相对选择偏差是一个二次项的比率，其导数具有显式二次多项式的符号。本地化可以增加、减少或非单调地改变这一诊断；其方向取决于组件偏差和几何形状。Oracle 重要性加权恢复了每个固定本地化目标的总体均值，但这些目标具有不同的目标。针对一个常见的全梯度目标，我们推导出有限样本均方误差、解析最优保留系数和固定裁剪扩展。确切的有限总体计算和每个样本大小的 10,000 次蒙特卡洛重复验证了这些恒等式和反例。公共人类后编辑实验使用两个翻译方向和预训练模型，并声明合成选择。一个英德扩展区分了 7389 万个本地参数。在所有三个声明的设置中，严格本地化具有更高的相对偏差但低于全保留的绝对偏差。未触及组件的偏差非零，选择组件的均值具有负内积，因此一般标准适用于简单的无偏/对齐解释失败的情况。输出偏差诊断显示了两个方向上相对偏差的相同端点排序，并具有一个内部最大值。机制重用每种语言的记录并包含混合；它们不是独立的重复。证据将相对放大与绝对梯度误差分开，并建立了估计属性，而不推断翻译质量的提升或识别实际投诉倾向。

View on arXiv Download PDF AI Translation

cs.CL / 21 / 2609.13721

Scaling Hindi Quantum Natural Language Processing through Automatic Pregroup Supertagging

通过自动预组超标记扩展印地语量子自然语言处理

Naik, Gautami Sanjay, Bhatia, Krishna, Saint-Germain, Mithun Paul, Babu, H Aswath

Abstract

Quantum Natural Language Processing (QNLP) uses pregroup grammars to translate grammatical structure into diagrammatic representations and quantum circuits. Recent Hindi QNLP work has shown that Hindi-specific pregroup grammars can support grammar-sensitive compositional models, but grammatical type assignment is still largely manual, limiting scalability. This paper formulates automatic Hindi pregroup supertagging as a token-level classification task. Using a manually annotated corpus of 380 Hindi sentences, we evaluate lexical, contextual, prompting-based, lexical-repair, and suffix/morphology-aware methods. Results show that simple lexical and contextual models are strong in this low-resource setting: contextual backoff achieves the best completed accuracy of 64.56\%, while raw Qwen2.5 prompting reaches only 11.65\%. Lexical repair raises LLM-assisted prediction to 64.08\%, demonstrating the value of constraining generative outputs with symbolic grammar knowledge. Diagnostic analysis further shows that seen and unambiguous tokens are much easier than unseen tokens, and suffix/morphology features improve karaka-token accuracy but not overall performance. These results show that automatic Hindi pregroup assignment is feasible and can reduce reliance on manual annotation in future multilingual QNLP pipelines.

Chinese Translation

量子自然语言处理（QNLP）利用预组语法将语法结构转换为图示表示和量子电路。近期的印地语QNLP研究表明，特定于印地语的预组语法可以支持对语法敏感的组合模型，但语法类型分配仍然主要依赖手动，这限制了可扩展性。本文将自动印地语预组超标记形式化为一个基于标记的分类任务。使用一个手动标注的380个印地语句子的语料库，我们评估了词汇、上下文、基于提示、词汇修复和后缀/形态感知方法。结果表明，在这一低资源环境中，简单的词汇和上下文模型表现良好：上下文回退实现了最佳完成准确率为64.56%，而原始的Qwen2.5提示仅达到11.65%。词汇修复将LLM辅助预测提高到64.08%，展示了用符号语法知识约束生成输出的价值。诊断分析进一步表明，已见和无歧义的标记比未见标记容易得多，而后缀/形态特征提高了karaka标记的准确性，但对整体性能没有改善。这些结果表明，自动印地语预组分配是可行的，并且可以减少未来多语言QNLP管道对手动标注的依赖。

View on arXiv Download PDF AI Translation

cs.CL / 22 / 2609.13734

PolicyMem: Geometric Policy Memory for LLM Governance

PolicyMem：用于大型语言模型治理的几何政策记忆

Bei, Yuanchen, Chen, Zhengzhang, Zhao, Yanjun, Wang, Haoyu, Tong, Hanghang, Chen, Haifeng

Abstract

As large language models (LLMs) are increasingly deployed in real-world high-stakes applications, effective governance has become essential. Existing safeguards largely follow two paradigms: learning-based guards provide strong semantic discrimination but couple policy behavior to trained models and taxonomies, while programmable frameworks offer flexible control but require substantial manual prompt and workflow engineering. Neither externalizes policies as reusable operational states, making it difficult to consistently reuse policy evidence across detection, intervention, and verification. In this paper, we introduce PolicyMem, a geometric policy memory that externalizes natural-language policies as reusable geometric memory objects represented by low-rank subspaces in a shared representation space. A memory writer compiles natural-language policies into policy memory slots, and query-response pairs read the policy memory through projection energy. The resulting policy-evidence profile directly mediates the safety verdict and is reused for policy attribution and post-intervention verification. Coupled with a response rewriter, PolicyMem enables a detect-rewrite-verify loop for LLM governance. Across five widely used benchmarks, PolicyMem achieves state-of-the-art unsafe behavior detection while enabling effective policy attribution, rewriting, and post-intervention verification through the shared policy memory.

Chinese Translation

随着大型语言模型（LLMs）在现实世界高风险应用中的日益普及，有效的治理变得至关重要。现有的安全措施主要遵循两种范式：基于学习的保护措施提供强大的语义区分能力，但将政策行为与训练模型和分类法耦合在一起，而可编程框架则提供灵活的控制，但需要大量的手动提示和工作流程工程。两者都未能将政策外部化为可重用的操作状态，使得在检测、干预和验证过程中一致地重用政策证据变得困难。在本文中，我们介绍了PolicyMem，一种几何政策记忆，它将自然语言政策外部化为可重用的几何记忆对象，这些对象由共享表示空间中的低秩子空间表示。记忆写入器将自然语言政策编译到政策记忆槽中，而查询-响应对通过投影能量读取政策记忆。生成的政策证据轮廓直接调解安全裁决，并在政策归因和干预后验证中重用。结合响应重写器，PolicyMem 实现了一个检测-重写-验证循环，用于 LLM 治理。在五个广泛使用的基准测试中，PolicyMem 实现了最先进的不安全行为检测，同时通过共享政策记忆实现了有效的政策归因、重写和干预后验证。

View on arXiv Download PDF AI Translation

cs.CL / 23 / 2609.13737

ForeSight: Enhancing Risk Monitoring via Early Safety Signal Distillation

ForeSight：通过早期安全信号提炼增强风险监测

Wang, Hanling, Wei, Chenlong, Xu, Ling, Niu, Hanyan, Cao, Qi, Huang, Shizhou, Yang, Yang, Zhu, Xiaohui, Zhu, Yao

Abstract

As large language models (LLMs) are increasingly deployed, the generation of harmful content has become a critical safety concern. Existing safeguards operate at the input, output, or streaming-generation stages, while early-risk methods that rely on surface tokens or output logits may suffer from weak initial signals, and internals-based detectors using dense representations may retain highly entangled and redundant safety-irrelevant information. It therefore remains unclear whether the earliest post-generation hidden states already contain reliable signals about final-response harmfulness. To address this gap, we propose ForeSight, a first-token output-risk forecasting framework that distills weak and redundant early safety signals into compact, layer-aware risk representations. Experiments on five safety benchmarks and two target models demonstrate that ForeSight achieves superior and efficient early-risk forecasting while relying solely on first-token hidden states. The code is available at: https://github.com/Scabbards1500/Foresight

Chinese Translation

随着大型语言模型（LLMs）的广泛应用，生成有害内容已成为一个重要的安全问题。现有的安全措施主要在输入、输出或流生成阶段进行，而依赖表面标记或输出对数的早期风险方法可能面临信号弱的挑战，而基于内部的检测器使用密集表示可能保留高度纠缠和冗余的与安全无关的信息。因此，尚不清楚最早的生成后隐藏状态是否已经包含有关最终响应有害性的可靠信号。为了解决这一问题，我们提出了ForeSight，一种首个标记输出风险预测框架，能够将弱且冗余的早期安全信号提炼为紧凑的、层感知的风险表示。在五个安全基准和两个目标模型上的实验表明，ForeSight在仅依赖首个标记隐藏状态的情况下，实现了卓越且高效的早期风险预测。代码可在以下链接获取：https://github.com/Scabbards1500/Foresight

View on arXiv Download PDF AI Translation

cs.CL / 24 / 2609.13745

Inside VLM Chart Reading: Tracing Value Reading from Vertical Bar Charts Across Space and Depth

深入VLM图表阅读：追踪垂直条形图中的价值阅读跨越空间与深度

Niu, Tianhao, Zhu, Qingfu, Che, Wanxiang

Abstract

Vision--language models (VLMs) can answer chart questions accurately, but output accuracy does not show how they combine the evidence needed to recover an exact value. We study vertical-bar value reading with controlled counterfactual activation patching in Qwen2.5VL-7B-Instruct and InternVL3.5-8B. The study connects three analyses: (1) The single-factor results show that the changed bar-top region restores much more answer preference than the unchanged bar body, despite containing fewer visual tokens. Legend- and series-related states also lose local recoverability earlier than bar-geometry and axis-scale states. (2) In the handoff analysis, restoration shifts from visual legend regions in early layers to prompt-series positions in middle layers. Resetting the prompt-series state selectively reduces legend-source rescue, supporting its role as a partial mediator. (3) In the factorial analysis, both models can use geometry and scale states from separate donors to favor the combined target. InternVL performs similarly when the states come from separate donors or one image, while Qwen shows lower restoration for separate donors which suggests higher context sensitivity. Together, these results provide preliminary causal evidence for localizing the internal computations that support exact bar-value reading.

Chinese Translation

视觉-语言模型（VLMs）能够准确回答图表问题，但输出的准确性并未展示它们如何结合所需的证据以恢复确切的数值。我们研究了在Qwen2.5VL-7B-Instruct和InternVL3.5-8B中，通过控制反事实激活修补进行的垂直条形价值阅读。该研究连接了三项分析：（1）单因素结果表明，改变的条形顶部区域恢复的答案偏好远高于未改变的条形主体，尽管视觉标记数量较少。与图例和系列相关的状态也比条形几何和轴刻度状态更早地失去局部可恢复性。（2）在交接分析中，恢复从早期层的视觉图例区域转移到中间层的提示系列位置。重置提示系列状态选择性地减少了图例来源的救助，支持其作为部分中介的角色。（3）在因子分析中，两个模型都可以使用来自不同捐赠者的几何和刻度状态来偏向于组合目标。当状态来自不同捐赠者或同一图像时，InternVL的表现相似，而Qwen在不同捐赠者的情况下恢复较低，这表明其对上下文的敏感性更高。综合来看，这些结果为定位支持确切条形值阅读的内部计算提供了初步的因果证据。

View on arXiv Download PDF AI Translation

cs.CL / 25 / 2609.13768

HyperProve: Answer-Guided Hypergraph Expansion for Multi-Hop Question Answering

HyperProve：基于答案引导的超图扩展用于多跳问答

Phu, An Nguyen, Quang, Dung Nguyen, An, Luu Hieu, Van, Linh Ngo, Le, Trung, Nguyen, Thien Huu

Abstract

Multi-hop question answering often fails when retrieval treats evidence as isolated matches to the original question, since the facts needed to answer a complex question are usually connected through intermediate entities, relations, and constraints. We propose HyperProve, a retrieval-augmented QA framework that addresses this challenge by coupling question decomposition with answer-conditioned expansion over a hypergraph of atomic facts. HyperProve does not use atomic facts, hypergraphs, or iterative retrieval in isolation; instead, it carries intermediate answers and supporting hyperedges as retrieval state, then uses that state to bias the next local hypergraph expansion. This design enables HyperProve to construct coherent evidence chains for final answer generation while making the retrieval process stateful and fact-centered. Across multi-hop QA benchmarks, HyperProve achieves the best overall performance in our evaluation, outperforming the strongest baselines by an average relative improvement of 6.2% in answer accuracy and 4.9% in F1.

Chinese Translation

多跳问答在检索将证据视为与原始问题孤立匹配时往往会失败，因为回答复杂问题所需的事实通常通过中间实体、关系和约束相互连接。我们提出了HyperProve，这是一种增强检索的问答框架，通过将问题分解与基于答案的超图原子事实扩展相结合来应对这一挑战。HyperProve并不孤立地使用原子事实、超图或迭代检索；相反，它将中间答案和支持的超边作为检索状态，然后利用该状态来偏向下一个局部超图扩展。这种设计使HyperProve能够构建连贯的证据链以生成最终答案，同时使检索过程具有状态性和以事实为中心。在多跳问答基准测试中，HyperProve在我们的评估中实现了最佳整体性能，平均相对提高了6.2%的答案准确率和4.9%的F1分数，超越了最强基线。

View on arXiv Download PDF AI Translation

cs.CL / 26 / 2609.13794

SyRHM: Symbolic-Language-Enhanced Reasoning with Associative Retrieval for Zero-shot Harmful Meme Detection

SyRHM：基于关联检索的符号语言增强推理框架用于零样本有害模因检测

Wang, Hanling, Wei, Chenlong, Li, Yingjuan, Wu, Di, Zhang, Yuchao, Zhu, Xiaohui, Zhu, Yao

Abstract

Detecting harmful memes is critical for maintaining safe online communities. However, harmful intent is often implicit, arising from visual-textual incongruity and cultural stereotypes, which challenges existing multimodal detectors. We propose SyRHM, a framework that decomposes harmful meme detection into meaning-grounded retrieval and symbolic-language-enhanced multi-stage reasoning. SyRHM retrieves semantically related memes by parsing multimodal content into textual elements and descriptions, providing grounded context beyond surface-level similarity. Building on the retrieved context, SyRHM uses a translator stage to convert multimodal inputs into symbolic intermediate representations, and then performs multi-stage reasoning via planner and solver stages, enabling expressive and interpretable analysis of harmful intent. Experiments on FHM, HarM, and MultiOff demonstrate the effectiveness of SyRHM, achieving superior performance on most evaluation settings against multimodal and reasoning-based baselines, while providing reasoning traces for harmful content. The code is available at: https://github.com/Scabbards1500/SyRHM

Chinese Translation

检测有害模因对于维护安全的在线社区至关重要。然而，有害意图往往是隐性的，源于视觉与文本的不一致性以及文化刻板印象，这对现有的多模态检测器提出了挑战。我们提出了SyRHM，一个将有害模因检测分解为基于意义的检索和符号语言增强的多阶段推理的框架。SyRHM通过将多模态内容解析为文本元素和描述，检索语义相关的模因，从而提供超越表面相似性的基础上下文。在检索到的上下文基础上，SyRHM使用翻译阶段将多模态输入转换为符号中间表示，然后通过规划者和求解者阶段进行多阶段推理，使得对有害意图的分析更加丰富和可解释。在FHM、HarM和MultiOff上的实验表明，SyRHM的有效性，在大多数评估设置中相较于多模态和基于推理的基线实现了优越的性能，同时提供了有害内容的推理痕迹。代码可在以下链接获取：https://github.com/Scabbards1500/SyRHM

View on arXiv Download PDF AI Translation

cs.CL / 27 / 2609.13806

Understanding the Limits of Agentic ICD Coding

理解代理性ICD编码的局限性

Eng, Chong Yock, Cao, Yushi, Chen, Yiming, Mao, Kezhi, Jiang, Hongchao

Abstract

ICD-10-CM codes are alphanumeric codes used in the US to classify diagnoses and injuries for medical billing and epidemiological reporting. Standard ICD-10-CM benchmarks report aggregate metrics that obscure performance on complex coding scenarios. We evaluate neural, workflow, and agentic systems on a rarity-stratified set of MIMIC-IV discharge summaries and identify two orthogonal failure modes. Neural classifiers exhibit a 0.43 micro-F1 gap between rare and common codes. Workflow systems handle rare codes well but score near zero on injury and external cause codes that require multi-step guideline following. A tool-augmented agentic configuration with structured access to official ICD-10-CM reference materials recovers up to 0.34 micro-F1 on this subset. No single system dominates across all conditions.

Chinese Translation

ICD-10-CM编码是美国用于对医疗账单和流行病学报告进行诊断和伤害分类的字母数字编码。标准的ICD-10-CM基准报告聚合指标，这些指标掩盖了在复杂编码场景下的表现。我们对MIMIC-IV出院摘要中一个按稀有性分层的集合进行了神经、工作流和代理系统的评估，并识别出两种正交的失败模式。神经分类器在稀有编码和常见编码之间表现出0.43的微F1差距。工作流系统能够很好地处理稀有编码，但在需要多步骤指南遵循的伤害和外部原因编码上得分接近零。使用结构化访问官方ICD-10-CM参考材料的工具增强代理配置在该子集上恢复了高达0.34的微F1。没有任何单一系统在所有条件下都占据主导地位。

View on arXiv Download PDF AI Translation

cs.CL / 28 / 2609.13808

DARE: Dialectical Agentic Reasoning for Structured Knowledge Fact Checking

DARE：用于结构化知识事实检查的辩证自主推理

Li, Yifei, Zheng, Xiaohan, Qian, Wentao, Zhuang, Liansheng

Abstract

Structured knowledge fact checking aims to determine the truthfulness of natural language claims by reasoning over structured evidence. Recent program-generation approaches leverage large language models (LLMs) to generate executable graph reasoning programs, achieving strong performance on structured knowledge fact checking benchmarks. However, these methods remain limited by invalid relation generation, single-path reasoning that lacks self-correction, and biased evidence assessment that tends to overestimate supporting signals. We propose Dialectical Agentic Reasoning (DARE), a multi-agent framework that formulates structured knowledge fact checking as an iterative retrieve-reason-reflect process. DARE integrates relation-grounded evidence retrieval to constrain reasoning to valid structures, dialectical bidirectional verification to evaluate evidence from both supporting and refuting perspectives, and confidence-driven meta-reflection to dynamically determine whether additional evidence exploration is necessary. Extensive experiments demonstrate the effectiveness of DARE in structured knowledge fact checking, with an 8B backbone achieving 88.12% accuracy and matching or surpassing GPT-4o-based program-generation baselines, which attests to the efficacy of dialectical agentic reasoning in eliciting the latent reasoning capabilities of LLMs.

Chinese Translation

结构化知识事实检查旨在通过对结构化证据进行推理来确定自然语言声明的真实性。最近的程序生成方法利用大型语言模型（LLMs）生成可执行的图推理程序，在结构化知识事实检查基准测试中取得了良好的表现。然而，这些方法仍然受到无效关系生成、缺乏自我修正的单路径推理以及倾向于高估支持信号的偏见证据评估的限制。我们提出了辩证自主推理（Dialectical Agentic Reasoning，DARE），这是一个将结构化知识事实检查表述为迭代的检索-推理-反思过程的多智能体框架。DARE整合了基于关系的证据检索，以限制推理在有效结构内，辩证的双向验证从支持和反驳两个角度评估证据，以及基于信心的元反思动态确定是否需要额外的证据探索。大量实验表明，DARE在结构化知识事实检查中的有效性，8B主干模型达到88.12%的准确率，并与基于GPT-4o的程序生成基线相匹配或超越，这证明了辩证自主推理在引发大型语言模型潜在推理能力方面的有效性。

View on arXiv Download PDF AI Translation

cs.CL / 29 / 2609.13824

When Consistency Does Not Mean Reliability: Evaluating Local LLM Judges Against Human Ratings

一致性并不意味着可靠性：评估地方性大语言模型评审者与人类评分的对比

Tiwari, Aakash Kumar

Abstract

Large language models (LLMs) are increasingly used to evaluate the responses of other language models. This approach, known as LLM-as-a-Judge, is faster and cheaper than human evaluation. However, a judge may produce consistent scores without necessarily agreeing with human evaluators. In this work, we study this issue using two local open-weight LLM judges, LLaMA-3-8B and Qwen2.5-7B. We evaluate 300 responses generated by an instruction-tuned GPT-2 (124M) model for 100 questions covering five categories: factual knowledge, instruction following, mathematics, reasoning, and writing. Each response is scored by nine human annotators and is evaluated three times by each LLM judge using the same rubric. We compare the judge scores with the average human scores using Pearson correlation, Spearman correlation, mean absolute error (MAE), signed bias, and self-consistency. LLaMA-3-8B shows a Pearson correlation of 0.275 with human scores, while Qwen2.5-7B achieves 0.340. Their MAEs are 27.71 and 18.64, respectively. Despite this limited agreement, both judges show high self-consistency, with exact consistency rates of 97.3\% for LLaMA-3-8B and 92.3\% for Qwen2.5-7B. These results show that high self-consistency does not necessarily indicate high agreement with human judgments. Our findings highlight the need to evaluate both consistency and human alignment when using local LLMs as automatic judges.

Chinese Translation

大型语言模型（LLMs）越来越多地被用于评估其他语言模型的响应。这种被称为LLM作为评审者（LLM-as-a-Judge）的方法比人类评估更快且成本更低。然而，评审者可能会产生一致的评分，但不一定与人类评估者达成一致。在本研究中，我们使用两个地方性开放权重的LLM评审者，LLaMA-3-8B和Qwen2.5-7B，来研究这一问题。我们评估了由一个经过指令调优的GPT-2（124M）模型生成的300个响应，这些响应涵盖100个问题，涉及五个类别：事实知识、指令遵循、数学、推理和写作。每个响应由九位人类注释者评分，并由每个LLM评审者使用相同的评分标准进行三次评估。我们使用皮尔逊相关系数（Pearson correlation）、斯皮尔曼相关系数（Spearman correlation）、平均绝对误差（Mean Absolute Error, MAE）、有符号偏差（signed bias）和自我一致性（self-consistency）比较评审者的评分与人类评分的平均值。LLaMA-3-8B与人类评分的皮尔逊相关系数为0.275，而Qwen2.5-7B为0.340。它们的MAE分别为27.71和18.64。尽管这种一致性有限，但两位评审者都表现出高自我一致性，LLaMA-3-8B的完全一致率为97.3\%，Qwen2.5-7B为92.3\%。这些结果表明，高自我一致性并不一定意味着与人类判断的高度一致。我们的研究结果强调在使用地方性LLMs作为自动评审者时，需要同时评估一致性和与人类的对齐程度。

View on arXiv Download PDF AI Translation

cs.CL / 30 / 2609.13841

Sweet Talkers: How Query Formulation Shapes Sycophancy in Romantic Relationship Advice

甜言蜜语：查询表述如何影响浪漫关系建议中的谄媚行为

Choi, Helena, Hao, Edric Castel, Bautista, Karl, Magleo, Francis Gabriel, Panti, Renzo, Olalia, Danielle Beatrice

Abstract

Large language models (LLMs) are increasingly used for emotional support and relationship advice, where a model's tendency to preserve a user's face can inadvertently reinforce harmful interpersonal behaviors. To systematically examine this risk, we developed the Romantic Relationship Advice-Seeking Prompts (RRASP) dataset of 2,400 prompts across five relationship themes and evaluated social sycophancy using the ELEPHANT framework on two consumer-facing models, GPT-5 Mini and Gemini 3 Flash. Contrary to our initial hypothesis, grammatical mood alone did not produce systematic differences in sycophantic behavior, suggesting that what a user implies matters more than how they phrase it. Instead, perspective-driven framing had a stronger influence, with gaps between original and flipped prompts widening in follow-up responses. Consistent increases in framing and moral sycophancy across turns indicate that models become more likely to accept a user's stated premises and affirm their ethical stance as a dialogue progresses. Notably, Gemini 3 Flash exhibited substantially smaller increases in moral sycophancy than GPT-5 Mini, suggesting it is more resistant to reinforcing ethically problematic positions across turns.

Chinese Translation

大型语言模型（LLMs）越来越多地用于情感支持和关系建议，但模型在维护用户面子的倾向可能无意中强化有害的人际行为。为了系统性地检验这一风险，我们开发了浪漫关系建议寻求提示（RRASP）数据集，包含2400个提示，涵盖五个关系主题，并使用ELEPHANT框架评估了两个面向消费者的模型（GPT-5 Mini和Gemini 3 Flash）中的社会谄媚行为。与我们最初的假设相反，仅语法语气并未产生系统性的谄媚行为差异，这表明用户的暗示比他们的措辞更为重要。相反，基于视角的框架具有更强的影响力，原始提示与翻转提示之间的差距在后续响应中扩大。随着对话的进行，框架和道德谄媚在各轮次中的一致性增加，表明模型在对话进展中更可能接受用户所陈述的前提并确认其伦理立场。值得注意的是，Gemini 3 Flash在道德谄媚方面的增加幅度明显小于GPT-5 Mini，这表明它在各轮次中对强化伦理上有问题的立场更具抵抗力。

View on arXiv Download PDF AI Translation

cs.CL / 31 / 2609.13847

Measuring the Cost of Variety Conflation in Multilingual MT Evaluation: Adding Mozambican Xichangana, Nyanja and Sena to FLORES+

在多语言机器翻译评估中测量多样性混淆的成本：将莫桑比克的Xichangana、Nyanja和Sena添加到FLORES+

Ali, Felermino D. M. A., Mateus, Delfina Lázaro, Mangue, Manuel Valente

Abstract

In this paper, we extend FLORES+ with Portuguese-source evaluation sets for three Mozambican Bantu varieties: Xichangana, Mozambican Nyanja, and Sena. We compare Xichangana with the existing Tsonga reference and Mozambican Nyanja with Chichewa, and evaluate NLLB-200, Google Translate, GPT, and a variant-aware NLLB model. Holding system output fixed reveals substantial reference sensitivity. On \textit{devtest}, changing only the reference from Tsonga to Xichangana reduces spBLEU by 13.10 points for NLLB-200 and 15.30 for Google. On matched Nyanja subsets, replacing Chichewa with Mozambican Nyanja produces smaller but consistent reductions of 3.03 and 6.10 spBLEU, respectively. Variant-aware fine-tuning reverses this pattern on the intended targets: relative to NLLB-200, it improves Xichangana by 7.04 spBLEU and Mozambican Nyanja by 5.33 on \textit{devtest}, while losing performance on the sibling references. GPT is competitive on Tsonga and Chichewa but substantially weaker on the Mozambican varieties. For Sena, the finetuned model reaches 12.64 spBLEU and 36.21 chrF++ on \textit{devtest}. These findings motivate variety-aware language identifiers, references, and reporting for cross-border languages or language dialects/variants. The data is publicly available on Hugging Face at https://huggingface.co/datasets/MOZNLP/FLORES_MOZ

Chinese Translation

在本文中，我们为三种莫桑比克班图语方言：Xichangana、莫桑比克Nyanja和Sena扩展了FLORES+，增加了葡萄牙语源评估集。我们将Xichangana与现有的Tsonga参考进行比较，将莫桑比克Nyanja与Chichewa进行比较，并评估NLLB-200、Google Translate、GPT以及一种变体感知的NLLB模型。保持系统输出不变显示出显著的参考敏感性。在 extit{devtest}上，仅将参考从Tsonga更改为Xichangana使得NLLB-200的spBLEU降低了13.10分，Google的spBLEU降低了15.30分。在匹配的Nyanja子集中，用莫桑比克Nyanja替换Chichewa分别产生了较小但一致的spBLEU减少，分别为3.03和6.10。变体感知的微调在目标上逆转了这一模式：相对于NLLB-200，它在 extit{devtest}上使Xichangana提高了7.04 spBLEU，莫桑比克Nyanja提高了5.33，同时在兄弟参考上性能下降。GPT在Tsonga和Chichewa上具有竞争力，但在莫桑比克方言上明显较弱。对于Sena，微调模型在 extit{devtest}上达到了12.64 spBLEU和36.21 chrF++。这些发现激励了对跨境语言或语言方言/变体的变体感知语言识别、参考和报告。数据可在Hugging Face上公开获取，网址为https://huggingface.co/datasets/MOZNLP/FLORES_MOZ

View on arXiv Download PDF AI Translation

cs.CL / 32 / 2609.13856

ShopEase: A Generative AI-Based Multi-Agent Framework for Intelligent Enterprise Customer Support Using Hybrid Retrieval-Augmented Generation

ShopEase：基于生成性人工智能的多智能体框架，用于混合检索增强生成的智能企业客户支持

Tiwari, Aakash Kumar, Kumar, Somesh

Abstract

Enterprise customer support systems must answer customer questions correctly, retrieve the right policy information, use customer context, and pass difficult cases to human agents when needed. This paper presents ShopEase, a Generative AI-based multi-agent framework for enterprise customer support. The system combines six components: Intent, CRM, Memory, Hybrid RAG, Escalation, and Supervisor, and uses LLaMA 3.2 running locally through Ollama for response generation. The retrieval module combines FAISS (dense retrieval) and BM25 (sparse retrieval), and six configurations are evaluated: BM25-only, FAISS-only, Fair RRF, Weighted RRF, RRF with Cross-Encoder, and Top-10 Hybrid with Cross-Encoder. Instead of using a fixed mapping between intent and policy, the policy category is decided directly from the retrieved documents. The system was evaluated on 2632 held-out customer queries across six categories: Refund, Return, Shipping, Cancellation, Damaged Product, and Unknown. FAISS-only achieved the highest accuracy of 85.37\% (2247 correct predictions), closely followed by Weighted RRF at 85.07\%. BM25-only achieved only 55.74\% accuracy. Adding cross-encoder reranking did not improve results: RRF with Cross-Encoder reached 83.24\%, and Top-10 Hybrid with Cross-Encoder reached 81.88\%, while also increasing response latency. Category-level analysis shows strong performance on Shipping, Cancellation, and Return, while Unknown queries remain the main source of errors. Statistical testing using McNemar's test shows no significant difference between FAISS-only and Weighted RRF, though both perform significantly better than Fair RRF and the cross-encoder configurations. Overall, dense retrieval gives the best accuracy on this dataset, and additional reranking adds processing time without improving classification performance.

Chinese Translation

企业客户支持系统必须正确回答客户问题，检索正确的政策信息，利用客户上下文，并在需要时将复杂案例转交给人工客服。本文提出了ShopEase，一个基于生成性人工智能的企业客户支持多智能体框架。该系统结合了六个组件：意图（Intent）、客户关系管理（CRM）、记忆（Memory）、混合检索增强生成（Hybrid RAG）、升级（Escalation）和监督者（Supervisor），并使用本地运行的LLaMA 3.2通过Ollama进行响应生成。检索模块结合了FAISS（密集检索）和BM25（稀疏检索），并评估了六种配置：仅BM25、仅FAISS、公平RRF、加权RRF、带交叉编码器的RRF和带交叉编码器的前10名混合。系统不再使用意图与政策之间的固定映射，而是直接从检索到的文档中决定政策类别。该系统在2632个保留的客户查询上进行了评估，涵盖六个类别：退款、退货、运输、取消、损坏产品和未知。仅FAISS的准确率最高，达到85.37%（2247个正确预测），紧随其后的是加权RRF，准确率为85.07%。仅BM25的准确率仅为55.74%。添加交叉编码器重排序并未改善结果：带交叉编码器的RRF达到了83.24%，而带交叉编码器的前10名混合达到了81.88%，同时也增加了响应延迟。类别级分析显示在运输、取消和退货方面表现强劲，而未知查询仍然是主要的错误来源。使用McNemar检验的统计测试显示，FAISS仅与加权RRF之间没有显著差异，尽管两者的表现显著优于公平RRF和交叉编码器配置。总体而言，密集检索在该数据集上提供了最佳准确率，而额外的重排序增加了处理时间而没有改善分类性能。

View on arXiv Download PDF AI Translation

cs.CL / 33 / 2609.13869

Bangla Sentence Function Classification: Corpus Development, Model Benchmarking, and Interpretability

孟加拉句子功能分类：语料库开发、模型基准测试与可解释性

Argha, Swapnil Kundu, Shafi, Abdullah Al, Zannat, Rowzatul, Polok, Shoumik Barman, Muntakim, Abdul, Ferdousi, Jannatul, Moyeen, M. A.

Abstract

Automatic sentence function identification is important for many downstream natural language processing (NLP) applications such as dialogue systems, text-to-speech synthesis, and machine translation. However, benchmark resources for Bangla sentence function classification remain limited. To mitigate this gap, this paper introduces a corpus of 10,000 Bangla sentences, manually annotated into four functional categories, namely declarative, interrogative, imperative, and exclamatory. The corpus is nearly balanced across the four classes, with high annotation reliability reflected by a Fleiss\' Kappa of 0.82. Furthermore, we evaluate multiple feature representations, including Bag-of-Words (BoW), TF-IDF, and Word2Vec, with several classical machine learning classifiers. In addition, two heterogeneous ensemble models, namely Single-Level Ensemble (SLE) and Double-Level Ensemble (DLE), are utilized to improve classification performance. Experimental results show that TF-IDF consistently outperforms Word2Vec, likely due to its ability to emphasize discriminative lexical cues associated with sentence functions, particularly given the relatively small corpus used to train Word2Vec. The DLE model with TF-IDF features achieves the best performance with accuracy and macro-F1 of 0.95, demonstrating the effectiveness of sparse lexical representations and heterogeneous ensemble learning for this task. Further cross-validation confirms the robustness of the approach, while LIME-based interpretability provides insights into model predictions. The developed corpus and model benchmarking establish strong baselines for Bangla sentence function classification.

Chinese Translation

自动句子功能识别对于许多下游自然语言处理（NLP）应用至关重要，例如对话系统、文本到语音合成和机器翻译。然而，孟加拉句子功能分类的基准资源仍然有限。为了解决这一问题，本文介绍了一个包含10,000个孟加拉句子的语料库，这些句子被手动标注为四个功能类别，即陈述句、疑问句、命令句和感叹句。该语料库在四个类别之间几乎平衡，注释可靠性高，Fleiss' Kappa值为0.82。此外，我们评估了多种特征表示方法，包括词袋模型（Bag-of-Words, BoW）、TF-IDF和Word2Vec，并使用几种经典的机器学习分类器进行实验。此外，采用了两种异构集成模型，即单层集成（Single-Level Ensemble, SLE）和双层集成（Double-Level Ensemble, DLE），以提高分类性能。实验结果表明，TF-IDF在性能上始终优于Word2Vec，这可能是由于其能够强调与句子功能相关的区分性词汇线索，特别是在用于训练Word2Vec的语料库相对较小的情况下。使用TF-IDF特征的DLE模型在准确率和宏观F1值上达到了0.95的最佳性能，证明了稀疏词汇表示和异构集成学习在此任务中的有效性。进一步的交叉验证确认了该方法的稳健性，而基于LIME的可解释性提供了对模型预测的深入见解。所开发的语料库和模型基准测试为孟加拉句子功能分类建立了强有力的基准。

View on arXiv Download PDF AI Translation

cs.CL / 34 / 2609.13874

SHIFT-M3: Pre-fusion Alignment-based Consistency Screening for Multimodal ECG Record Integrity

SHIFT-M3：基于预融合对齐的一致性筛查用于多模态心电图记录完整性

Khan, Md Ashik, Siddique, Md Nahid

Abstract

Multimodal clinical AI typically assumes that the waveform, report, metadata, and downstream predictions attached to a record belong to the same patient. In practice, linkage failures can silently assemble individually plausible but cross-patient components, creating a safety problem that standard predictive models are not designed to detect. We study this problem as multimodal record integrity triage: given an assembled record, should its modalities be trusted to belong together? We introduce SHIFT-M3, a lightweight text-based pre-fusion screen that measures alignment-based consistency between two separately produced ECG text views: an LLM-generated interpretation and a clinical report summary. On 784,680 MEETI ECG records, SHIFT-M3 achieves 97.6% TPR@5% FPR for full text-view swaps (AUROC 0.996), 90.3% for partial swaps (AUROC 0.974), and 97.7% for label-matched hard negatives (AUROC 0.996) with only 573,569 parameters. Compared with same-dataset lexical baselines, the gains are largest on partial swaps and hard negatives, suggesting that the model is learning more than surface overlap. We also introduce the CMST (Conflict-type Multimodal Stress Test) evaluation taxonomy, a three-seed stability study, a loss ablation, a temporal-tolerance sweep, and a shared-token masking control. The main remaining failure mode is longitudinal ambiguity: at the default operating point, same-patient cross-visit pairs still produce 87.0% Type-II false positives.

Chinese Translation

多模态临床人工智能通常假设附加在记录上的波形、报告、元数据和下游预测属于同一患者。实际上，链接失败可能会悄然组合个别看似合理但跨患者的组件，从而造成安全问题，而标准预测模型并未设计用于检测此类问题。我们将此问题研究为多模态记录完整性分类：给定一个组合记录，其模态是否值得信任为一体？我们引入了SHIFT-M3，这是一种轻量级的基于文本的预融合筛查方法，测量两个独立生成的心电图文本视图之间的基于对齐的一致性：一个是由大型语言模型（LLM）生成的解释，另一个是临床报告摘要。在784,680个MEETI心电图记录上，SHIFT-M3在完整文本视图交换中实现了97.6%的真正率（TPR@5%假阳性率（FPR）），其曲线下面积（AUROC）为0.996；在部分交换中实现了90.3%（AUROC 0.974）；在标签匹配的硬负样本中实现了97.7%（AUROC 0.996），仅使用573,569个参数。与同一数据集的词汇基线相比，部分交换和硬负样本的增益最大，表明该模型学习到的内容超出了表面重叠。我们还引入了CMST（冲突类型多模态压力测试）评估分类法、三种种子稳定性研究、损失消融、时间容忍度扫描和共享标记掩蔽控制。主要剩余的失败模式是纵向模糊性：在默认操作点下，同一患者的跨访视对仍然产生87.0%的类型II假阳性。

View on arXiv Download PDF AI Translation

cs.CL / 35 / 2609.13907

Phorecaster365: A Human-Supervised Reference Architecture for Hybrid Pharmaceutical Sales Forecasting and Planning Decision Support

Phorecaster365：一种人类监督的混合制药销售预测与规划决策支持参考架构

Kazemzadeh, Houman, Naderi, Kamyar

Abstract

Pharmaceutical sales forecasts inform planning across products, regions, and distribution channels, yet their interpretation depends on inventory availability, transaction semantics, product lifecycle, and the information available when each forecast is issued. A model prediction alone does not preserve these conditions or establish whether a forecast is suitable for operational use. We present Phorecaster365, a human-supervised reference architecture that connects enterprise resource planning data to reviewable pharmaceutical sales forecasts. The architecture separates source ingestion, product-region series construction, temporally valid feature generation, statistical and machine-learning modeling, ensemble formation, uncertainty assessment, planner review, and lifecycle governance. Its central intermediate representation is a forecast context package that preserves the source snapshot, forecast target, temporal cutoff, available covariates, data-quality state, and hierarchy version. A corresponding forecast evidence package links predictions to model and calibration versions, exceptions, human adjustments, and publication history. Development experience with a synthetic panel of 10,950 daily records across 30 product-region series informs the design. Historical experiment summaries are retained only as descriptive evidence of development because their evaluation does not establish independent predictive validity. We specify a rolling-origin evaluation protocol, baseline and ablation requirements, uncertainty and robustness assessments, and a staged pathway from synthetic testing to use in governed planning. The contribution is an implementation-neutral system design and validation framework; the report does not establish real-world forecasting accuracy, comparative superiority, or operational benefit.

Chinese Translation

制药销售预测为产品、地区和分销渠道的规划提供信息，但其解读依赖于库存可用性、交易语义、产品生命周期以及每次发布预测时所获得的信息。仅凭模型预测无法保留这些条件或确定预测是否适合用于操作。我们提出了Phorecaster365，这是一种人类监督的参考架构，将企业资源规划数据与可审查的制药销售预测连接起来。该架构将源数据摄取、产品-地区系列构建、时间有效特征生成、统计和机器学习建模、集成形成、不确定性评估、规划者审查和生命周期治理分开。其核心中间表示是一个预测上下文包，保留了源快照、预测目标、时间截止、可用协变量、数据质量状态和层级版本。相应的预测证据包将预测与模型和校准版本、异常情况、人为调整和出版历史联系起来。基于一个包含30个产品-地区系列的10,950个每日记录的合成面板的开发经验为设计提供了依据。历史实验总结仅作为开发的描述性证据保留，因为其评估并未确立独立的预测有效性。我们指定了一个滚动起源评估协议、基线和消融要求、不确定性和稳健性评估，以及从合成测试到受管规划使用的分阶段路径。该贡献是一个与实现无关的系统设计和验证框架；本报告并未确立现实世界预测的准确性、比较优越性或操作效益。

View on arXiv Download PDF AI Translation

cs.CL / 36 / 2609.13916

North Small Translate: Advanced Cost-Effective Translation (Cohere CAT+)

北小翻译：先进的成本效益翻译（Cohere CAT+）

Kocmi, Tom, Bérard, Alexandre, Blunsom, Phil, Cahyawijaya, Samuel, Cassini, Shaun, Frosst, Nicholas, de Gibert, Ona, Gomez, Aidan, Govindarajan, Nithya, Kiyono, Shun, Lasche, Olivia, Rogers, Lawrence, Marchisio, Kelly, Moghe, Nikita, More, Yash, Moran-Hidalgo, Camila, Nan, Yiyang, Sachs, Michael, Starostina, Trisha, van Stigt, Daan, Rarrick, Spencer, Vincent, Sebastian, Zhang, Ivan

Abstract

We present North Small Translate, an open-weight, LLM-based machine translation (MT) model with instruction-following capabilities built on the same foundation as Cohere's Command A Plus, a mixture-of-experts architecture with 25 billion active parameters out of 218 billion total parameters. North Small Translate is trained using difficulty sampling to obtain challenging documents and a five-step training protocol combining supervised fine-tuning, direct preference optimization, and online reinforcement learning. We prioritized throughput through a non-reasoning base model and supplemented with optional agentic capabilities to unlock translation quality gains. North Small Translate is trained to perform MT-related tasks, including post-editing and quality estimation, as well as related tasks such as general instruction following. The model achieves top MT performance across 50 languages in the class of models under 1T parameters, with no need to run expensive reasoning at inference time.

Chinese Translation

我们提出了北小翻译，这是一个基于大型语言模型（LLM）的开放权重机器翻译（MT）模型，具备遵循指令的能力，建立在Cohere的Command A Plus的基础上，该模型采用混合专家架构，拥有2180亿个参数中的250亿个活跃参数。北小翻译采用困难抽样进行训练，以获取具有挑战性的文档，并结合监督微调、直接偏好优化和在线强化学习的五步训练协议。我们通过非推理基础模型优先考虑吞吐量，并补充可选的自主能力，以解锁翻译质量的提升。北小翻译被训练用于执行与机器翻译相关的任务，包括后期编辑和质量评估，以及一般指令遵循等相关任务。该模型在50种语言的1T参数以下模型类别中实现了顶级的机器翻译性能，无需在推理时进行昂贵的推理计算。

View on arXiv Download PDF AI Translation

cs.CL / 37 / 2609.13936

Inter-Rater Reliability of LLM and Rule-Based Annotation for Inferential Narrative Features: Three Studies on a Turkish Corpus

LLM与基于规则的注释在推理叙事特征上的评估者间可靠性：对土耳其语语料库的三项研究

Bulut, Levent

Abstract

Datasets that ship automatically generated feature annotations invite a question rarely asked of them: would a human agree with those labels? This report answers that for the Objective Projection corpus, a Turkish narrative dataset whose scenes carry a per-scene applied_rules field from a rule-based detector over six craft features -- two prohibitions (emotion labelling, simile) and four positive techniques (materialized metaphor, micro-focus, temporal anchor, atmosphere contradiction). Three studies are reported. Study 1 ($n = 120$) scores the detector against blind labels from the scheme's own author. Study 2 ($n = 100$, a disjoint scene set) scores the detector plus Gemini 2.5 Flash and Grok against an independent non-expert rater whose labels were locked before any machine ran. Study 2b re-runs the identical protocol with Claude Fable 5 (High) and ChatGPT 5.5. The central result concerns one rule. On materialized metaphor -- closest to the methodology's theoretical core -- the five machine labellers returned positive rates of $0$, $1$, $40$, $72$ and $78$ out of $100$ scenes, against a human count of $9$. Cohen's $\kappa$ was at or indistinguishable from chance for five of six labellers, across both human references and both scene sets: $0.004$, $0.015$, $0.000$, $0.019$, $0.027$. Raw agreement ranged from $74.7\%$ to $84.5\%$, an artefact of class imbalance rather than a sign of competence. We deliberately do not resolve this into a single story. Two readings survive: the feature is genuinely inferential and beyond current automatic detection, or the rule's definition is not yet operational enough for any rater to apply consistently -- including the human. Distinguishing them needs a second independent human rater, which this report does not have and therefore does not claim.

Chinese Translation

自动生成特征注释的数据集引发了一个鲜有提及的问题：人类是否会同意这些标签？本报告对此进行了回答，研究对象为Objective Projection语料库，这是一个土耳其叙事数据集，其场景包含来自基于规则的检测器的每场景应用规则字段，涵盖六个工艺特征——两个禁止项（情感标记、明喻）和四个积极技术（具象隐喻、微观聚焦、时间锚、氛围矛盾）。报告了三项研究。研究1（$n = 120$）将检测器的结果与该方案作者的盲标签进行评分。研究2（$n = 100$，一个不重叠的场景集）将检测器与Gemini 2.5 Flash和Grok的结果与一个独立的非专家评审者的标签进行评分，该评审者的标签在任何机器运行之前已被锁定。研究2b重复了相同的协议，使用Claude Fable 5（高）和ChatGPT 5.5。核心结果涉及一个规则。在具象隐喻这一最接近方法论理论核心的特征上，五个机器标签的正面率分别为$0$、$1$、$40$、$72$和$78$，而人类计数为$9$。对于六个标签中的五个，Cohen's $eta$与随机情况相当或无法区分：$0.004$、$0.015$、$0.000$、$0.019$、$0.027$。原始一致性范围从$74.7\%$到$84.5\\%$，这是类别不平衡的产物，而非能力的标志。我们故意不将其归结为一个单一的故事。两种解读依然存在：该特征确实是推理性的，超出了当前的自动检测能力，或者该规则的定义尚未足够操作化，以便任何评审者（包括人类）能够一致地应用。区分这两者需要第二个独立的人类评审者，而本报告并未具备，因此不对此进行主张。

View on arXiv Download PDF AI Translation

cs.CL / 38 / 2609.13942

CRITICS - Critical Science Without Borders: Language Models to Promote Critical Thinking in Science Education

CRITICS - 无国界的关键科学：利用语言模型促进科学教育中的批判性思维

Agerri, Rodrigo, Aldabe, Itziar, Cabrio, Elena, Cieliebak, Mark, Deriu, Jan, Flores, Mariana, Kapociute-Dzikiene, Jurgita, Kuiziniene, Dovile, Rico, Arantza, Ruiz-Gonzalez, Aritz, Soroa, Aitor, Vaskevicius, Mantas, Villata, Serena

Abstract

The CRITICS project addresses science accessibility and literacy by converging advanced Machine Translation (MT) based on Large Language Models (LLMs) with educational technology. By leveraging MT systems specifically optimized for scientific content, educational institutions can provide accurate, culturally relevant translations of scientific materials in students' native languages, ensuring that complex scientific concepts are comprehensible while maintaining technical accuracy. Building on these translations, the project explores the design and evaluation of innovative science teaching-learning proposals grounded in curriculum-aligned teaching-learning. Thus, CRITICS will investigate key components of scientific argumentation and critical thinking practices together with textual feedback aligned with learning objectives and assessment criteria inspired by competence-based evaluation frameworks. CRITICS aims to break down language barriers to accessing cutting-edge research and educational materials currently available only in high-resourced languages, thereby facilitating the democratization of scientific knowledge and fostering critical thinking in science education.

Chinese Translation

CRITICS项目通过将基于大型语言模型（LLMs）的先进机器翻译（MT）与教育技术相结合，解决科学的可及性和素养问题。通过利用专门针对科学内容优化的机器翻译系统，教育机构能够提供准确且具有文化相关性的科学材料翻译，使学生能够用母语理解复杂的科学概念，同时保持技术准确性。在这些翻译的基础上，该项目探索基于课程对齐的创新科学教学-学习提案的设计与评估。因此，CRITICS将研究科学论证和批判性思维实践的关键组成部分，以及与学习目标和评估标准相一致的文本反馈，这些标准受到基于能力评估框架的启发。CRITICS旨在打破语言障碍，使目前仅在高资源语言中可用的前沿研究和教育材料得以普及，从而促进科学知识的民主化，并在科学教育中培养批判性思维。

View on arXiv Download PDF AI Translation

cs.CL / 39 / 2609.13948

Thought without systematicity? Evaluating reasoning models on rule induction tasks

没有系统性思维？评估推理模型在规则归纳任务上的表现

Schug, Simon, Lake, Brenden M.

Abstract

A central tenet of human cognition is systematicity, the principle that understanding one concept is inherently tied to understanding close variations of that concept. Do reasoning models robustly exhibit such systematicity? If so, we would expect consistent performance on structurally equivalent variants of the same task. Here, we extend established rule induction tasks from cognitive science to assess the systematicity of thought in current reasoning models. Each task family has compositional structure that we use to create structurally equivalent task variations through task isomorphisms such as recombination and substitution. We find that despite being able to correctly solve a task, models often fail on structurally equivalent variants of the same task. These findings suggest that many model behaviors lack systematicity, rendering it difficult to robustly establish the cognitive abilities of reasoning models beyond the particular contexts they were evaluated in.

Chinese Translation

人类认知的一个核心原则是系统性，即理解一个概念与理解该概念的近似变体之间是内在关联的。推理模型是否稳健地表现出这种系统性？如果是这样，我们应该期待在同一任务的结构等价变体上表现出一致的性能。在此，我们扩展了认知科学中已建立的规则归纳任务，以评估当前推理模型的思维系统性。每个任务家族具有组合结构，我们利用该结构通过任务同构（如重组和替代）创建结构等价的任务变体。我们发现，尽管模型能够正确解决一个任务，但在同一任务的结构等价变体上，模型往往表现不佳。这些发现表明，许多模型行为缺乏系统性，使得在超出其评估特定上下文的情况下，稳健地建立推理模型的认知能力变得困难。

View on arXiv Download PDF AI Translation

cs.CL / 40 / 2609.13980

Mizan: A National Benchmark for Evaluating Large Language Models on Iraqi Arabic and the Iraqi Civic Context

Mizan：评估大型语言模型在伊拉克阿拉伯语及伊拉克公民背景下的国家基准

Alseelawi, Nawar S., Aljumaily, Mustafa S.

Abstract

Arabic large-language-model (LLM) evaluation has matured around Modern Standard Arabic (MSA): aggregated leaderboards such as the Open Arabic LLM Leaderboard (OALL), HELM Arabic, and BALSAM rank models across dozens of MSA tasks, and frontier systems increasingly saturate them. Dialectal Arabic, the language Iraqis actually speak, remains nearly invisible to this infrastructure. We introduce Mizan ("the balance"), Iraq's national benchmark for evaluating LLMs on Iraqi Arabic and the Iraqi civic context: an MSA baseline track paired with an Iraqi track across six axes (dialect comprehension, dialect generation, bidirectional MSA-Iraqi translation, Iraq-specific knowledge, official-document field extraction, and safety), built from 340 originally authored, dually reviewed items with statistically audited answer positions and Wilson intervals on every published score. A pilot evaluation of 27 systems, spanning closed frontier models three days after release, open weights across size tiers, and an Arabic trio of commercial, open-specialized, and sovereign systems, yields four findings. The MSA track saturates while the Iraqi track discriminates, with a consistent 14-18-point per-model gap and statistically tied leaders. Official-document extraction confines every system to 32-56. Arabic-focused specialization behaves as MSA specialization: two dedicated Arabic models score below a size-matched generalist on the Iraqi track. And the safety-hardened tier of the newest model family deterministically refuses innocuous dialect-comprehension items as policy violations, an over-refusal mode invisible to MSA benchmarks. The platform enforces an integrity protocol of immutable snapshots, verification certificates, a human publication gate, and public retraction, all exercised during this study. Code and the public development set accompany the paper.

Chinese Translation

阿拉伯大型语言模型（LLM）的评估已经围绕现代标准阿拉伯语（MSA）发展成熟：诸如开放阿拉伯LLM排行榜（OALL）、HELM阿拉伯语和BALSAM等聚合排行榜对数十个MSA任务的模型进行排名，而前沿系统也越来越多地充斥其中。然而，伊拉克人实际使用的方言阿拉伯语在这一基础设施中几乎不可见。我们介绍了Mizan（“平衡”），这是伊拉克用于评估LLM在伊拉克阿拉伯语及伊拉克公民背景下的国家基准：一个与MSA基线轨道配对的伊拉克轨道，涵盖六个维度（方言理解、方言生成、双向MSA-伊拉克翻译、伊拉克特定知识、官方文件领域提取和安全性），由340个原创、双重审查的项目构成，所有发布的分数都经过统计审核的答案位置和威尔逊区间。对27个系统的初步评估，涵盖了在发布后三天内的封闭前沿模型、不同规模的开放权重，以及一组三个商业、开放专业和主权系统的阿拉伯语模型，得出了四个发现。MSA轨道趋于饱和，而伊拉克轨道则能够区分，模型之间的差距稳定在14-18分，且领导者之间统计上相互绑定。官方文件提取将每个系统限制在32-56分。专注于阿拉伯语的专业化表现得与MSA专业化相似：两个专门的阿拉伯模型在伊拉克轨道上的得分低于一个规模匹配的通用模型。此外，最新模型系列的安全强化层以政策违规的方式确定性地拒绝无害的方言理解项目，这种过度拒绝模式在MSA基准中是不可见的。该平台在本研究期间实施了不可变快照、验证证书、人为出版门槛和公共撤回的完整性协议。代码和公共开发集随论文附上。

View on arXiv Download PDF AI Translation

cs.CL / 41 / 2609.13997

Unlocking the Unsolvable: Teacher-Guided Curriculum for Data-Efficient RLVR

解锁不可解的问题：教师引导的高效数据强化学习课程

Zhu, Yukang, Han, Zhen

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has shown remarkable success in improving the mathematical reasoning of large language models. Yet problems beyond the model's current capability, where rollouts uniformly fail and no learning signal is produced, are structurally wasted despite marking the most informative training frontier. We show that these otherwise-inert problems can be unlocked via teacher-guided curriculum learning: partial reasoning traces from a stronger model create a graded difficulty landscape, and a backward-chaining curriculum progressively withdraws guidance until the model solves problems unaided. Training on only 128 unsolvable problems matches or exceeds GRPO trained on a full 2,000-problem corpus (~16x data efficiency) on the nine-benchmark average for both base models, while substantially expanding the reasoning boundary measured by pass@k at large k. Furthermore, we identify a distribution-shift cost that is particularly acute in the unsolvable-only regime and propose Monotone Frontier Curriculum (MFC), a method that monotonically drives training toward unguided solving, consistently outperforming existing curriculum methods.

Chinese Translation

可验证奖励的强化学习（RLVR）在提升大型语言模型的数学推理能力方面取得了显著成功。然而，对于模型当前能力之外的问题，尽管这些问题标志着最具信息量的训练前沿，但由于回滚均匀失败且未产生学习信号，结构上被浪费。我们展示了这些原本无效的问题可以通过教师引导的课程学习来解锁：来自更强模型的部分推理轨迹创建了一个分级难度的景观，而向后链式课程逐步撤回指导，直到模型能够独立解决问题。仅在128个不可解问题上进行训练的效果与在完整的2000个问题语料库上训练的GRPO相当或更优（~16倍的数据效率），在两个基础模型的九个基准平均值上表现突出，同时显著扩展了通过pass@k在大k值下测量的推理边界。此外，我们识别出在仅限不可解问题的情况下，分布转移成本尤为严重，并提出了单调前沿课程（Monotone Frontier Curriculum, MFC），该方法单调地推动训练朝向无指导解决，始终优于现有的课程方法。

View on arXiv Download PDF AI Translation

cs.CL / 42 / 2609.14057

Measuring the Creativity of Frontier LLMs in Automated Research

前沿大型语言模型在自动化研究中的创造力测量

Zhao, Yiheng, Chen, Mengzhuo, Hu, Chengming, Liao, Pengyi, Pang, Yiran

Abstract

Frontier LLMs are increasingly capable of conducting automated research, yet their creativity in this setting has not been systematically evaluated. In this paper, we propose a set of metrics to evaluate creativity along the two dimensions of valueness and novelty. Valueness assesses whether each proposed idea is useful, while novelty is evaluated from three perspectives: whether the same idea has appeared before (Exact-Match P-Novelty), whether a previously unexplored variable or variable combination is explored (Variable-level P-Novelty), and whether the idea directly follows retrieved external knowledge or departs from it (H-Novelty). Our evaluation shows that the models achieve relatively similar scores on most creativity metrics, but differ substantially in Variable-level P-Novelty, which reflects the breadth of research-space exploration. Further correlation and idea-level performance analyses show that Variable-level P-Novelty is the creativity dimension most strongly associated with research performance.

Chinese Translation

前沿大型语言模型（Frontier LLMs）在进行自动化研究方面的能力日益增强，但其在这一环境下的创造力尚未得到系统评估。本文提出了一套评估创造力的指标，涵盖了价值性（valueness）和新颖性（novelty）两个维度。价值性评估每个提出的想法是否有用，而新颖性则从三个角度进行评估：相同想法是否曾经出现过（Exact-Match P-Novelty），是否探索了之前未曾研究的变量或变量组合（Variable-level P-Novelty），以及该想法是否直接依赖于检索到的外部知识或与之背离（H-Novelty）。我们的评估表明，模型在大多数创造力指标上的得分相对相似，但在Variable-level P-Novelty上存在显著差异，这反映了研究空间探索的广度。进一步的相关性和想法层面的性能分析显示，Variable-level P-Novelty是与研究表现最强相关的创造力维度。

View on arXiv Download PDF AI Translation

cs.CL / 43 / 2609.14066

GraMRAG: Orchestrating Multi-Agent Multi-Step Reasoning via Graph Memory with Reinforcement Learning

GraMRAG：通过图记忆与强化学习协调多智能体多步骤推理

Wang, Zhongyu

Abstract

Although existing multi-agent Retrieval-Augmented Generation (RAG) systems have demonstrated promise on complex multimodal reasoning tasks, they remain fundamentally limited in reasoning depth and memory structure, suffering from inadequate retrieval and state blindness when answering knowledge-intensive questions. To address these limitations, we propose GraMRAG, a graph memory-guided multi-agent RAG framework that integrates a dynamic multimodal memory graph to enable stable, multi-step multimodal reasoning. We introduce a vision-text bridged reasoning paradigm that unifies multi-scale entity cropping with a ReAct-style visual toolchain, enhancing the long-horizon cross-modal reasoning capability. We further construct a multimodal memory graph that formalizes agent reasoning as a dynamic directed acyclic graph (DAG), explicitly modeling action-observation dependencies to mitigate state blindness and suppress redundant retrieval. Moreover, we propose Topology-Aware Policy Optimization (TAPO) that leverages graph topology for critical path identification and targeted node pruning, enabling fine-grained credit assignment across multi-step reasoning trajectories. Extensive experiments on challenging multimodal benchmarks demonstrate that our approach consistently outperforms existing baselines and achieves state-of-the-art performance on complex long-horizon reasoning tasks.

Chinese Translation

尽管现有的多智能体检索增强生成（RAG）系统在复杂的多模态推理任务中展现了良好的前景，但它们在推理深度和记忆结构上仍然存在根本性的局限，尤其在回答知识密集型问题时，检索不足和状态盲目性问题尤为突出。为了解决这些问题，我们提出了GraMRAG，一个基于图记忆的多智能体RAG框架，该框架集成了动态多模态记忆图，以实现稳定的多步骤多模态推理。我们引入了一种视觉-文本桥接的推理范式，将多尺度实体裁剪与ReAct风格的视觉工具链相结合，增强了长时间跨度的跨模态推理能力。此外，我们构建了一个多模态记忆图，将智能体推理形式化为动态有向无环图（DAG），明确建模行动-观察依赖关系，以减轻状态盲目性并抑制冗余检索。此外，我们提出了拓扑感知策略优化（Topology-Aware Policy Optimization, TAPO），利用图拓扑进行关键路径识别和目标节点修剪，从而实现多步骤推理轨迹的细粒度信用分配。在具有挑战性的多模态基准上进行的大量实验表明，我们的方法在性能上始终优于现有基线，并在复杂的长时间跨度推理任务中达到了最先进的性能。

View on arXiv Download PDF AI Translation

cs.CL / 44 / 2609.14144

One Size Does Not Fit All: Setting Inference Depth from the Questions a Deployment Actually Asks

一刀切并不适用：根据实际部署所提出的问题设置推理深度

Kaplan, Jerry

Abstract

A transformer language model is trained to respond to any prompt, but each deployment asks only a narrow range of questions: a support assistant sees delivery complaints, a coding tool sees Python. Every deployment nonetheless pays the same computation per token. This paper measures how much of that cost is avoidable when the range of prompts is known in advance. The mechanism examined is early exit: a small, trained component - called a readout - is attached to an intermediate layer and proposes a token, and a confidence test decides whether to emit it or to run the remaining layers. The models are frozen, and the only supervision used is the model's own output on ordinary traffic. Three findings are reported. First, achievable savings depend strongly on the kind of traffic: at half depth on a 1.5-billion-parameter model, 96 percent of tokens could be emitted early for arithmetic word problems and 8 percent for Chinese-language explanations, at matched token-level fidelity to the full model (a measure whose limits the third finding exposes). Second, of three ways a deployment might use knowledge of its traffic, only customizing the threshold for exiting early is worthwhile: calibrating it per deployment raised exit rates by up to 59 percentage points across three models, and by more than 10 points on most corpora tested. Third, token-level fidelity - the standard evaluation measure in the early-exit literature - fails in domains where tokens can be checked against ground truth: on arithmetic word problems, three models each answered sixty questions correctly when run in full, and between 10 and 28 correctly under early exit, in the configuration that scored highest on fidelity. The intended setting is small models on personal devices, where generation is limited by memory bandwidth rather than computation.

Chinese Translation

一种变换器语言模型被训练以响应任何提示，但每个部署仅提出狭窄范围的问题：支持助手处理交付投诉，编码工具处理Python。尽管如此，每个部署在每个token上支付的计算成本是相同的。本文衡量了在已知提示范围的情况下，多少成本是可以避免的。所研究的机制是早期退出：一个小型的、经过训练的组件——称为读出——附加在中间层，并提出一个token，而一个置信度测试决定是发出该token还是运行剩余层。模型被冻结，唯一使用的监督是模型在普通流量下的输出。报告了三项发现。首先，可实现的节省在很大程度上依赖于流量的类型：在一个15亿参数的模型上，深度为一半时，对于算术文字问题，96%的tokens可以提前发出，而对于中文解释，仅有8%的tokens可以提前发出，并且在token级别上与完整模型的保真度相匹配（第三项发现揭示了这一度量的局限）。第二，在三种可能利用流量知识的方式中，只有定制早期退出的阈值是值得的：针对每个部署进行校准使得退出率在三个模型中提高了多达59个百分点，并且在大多数测试语料库中提高了超过10个百分点。第三，token级别的保真度——早期退出文献中的标准评估度量——在可以与真实值进行核对的领域中失效：在算术文字问题上，三个模型在完整运行时各自正确回答了六十个问题，而在以最高保真度评分的早期退出配置下，仅有10到28个问题被正确回答。预期的设置是个人设备上的小型模型，其中生成受到内存带宽而非计算的限制。

View on arXiv Download PDF AI Translation

cs.CL / 45 / 2609.14157

When Tools Get in the Way: The Effect of Unnecessary Tool Availability on LLM Answering

工具的干扰：不必要工具可用性对大型语言模型回答的影响

Paturi, Saanvi, Kenzhebayev, Arsen, Sethi, Arham, Raina, Vyas, Sheth, Ivaxi, Raina, Vatsal

Abstract

Large language models (LLMs) are increasingly deployed with external tools that extend what they can do beyond their own knowledge. Tools help on tasks that need external information, but their availability may also change how a model handles questions that do not need them. Prior work has mostly asked whether models select and use tools appropriately; whether an unnecessary tool changes the correctness of answers has received less attention. We ask whether making a related but unnecessary tool available affects a model's ability to answer from its own knowledge, and whether a preceding tool interaction changes this behaviour. We construct 500 query pairs across 10 knowledge domains. Each pair consists of a tool query, which needs the domain's tool, and a closed-domain query, which does not. Six LLMs are evaluated with the tool unavailable, available, and available after a prior tool call. Across 3,000 baseline trials the pooled answer rate is 98.2%. When an unnecessary tool is available it falls to 63.5%, with large differences between models. The decrease occurs even when the tool is rarely called, so it cannot be explained by unnecessary tool invocation alone. A one-sentence scope-aware system instruction recovers most of the lost answers.

Chinese Translation

大型语言模型（LLMs）越来越多地与外部工具结合使用，这些工具扩展了它们超出自身知识的能力。工具在需要外部信息的任务中提供帮助，但它们的可用性也可能改变模型处理不需要工具的问题的方式。之前的研究主要关注模型是否适当地选择和使用工具；而不必要工具是否会改变答案的正确性则受到的关注较少。我们探讨了提供一个相关但不必要的工具是否会影响模型基于自身知识回答的能力，以及先前的工具交互是否会改变这种行为。我们在10个知识领域构建了500对查询。每对查询由一个需要该领域工具的工具查询和一个不需要的封闭领域查询组成。我们评估了六个LLM在工具不可用、可用以及在先前工具调用后可用的情况下的表现。在3000个基线试验中，汇总的回答率为98.2%。当不必要的工具可用时，回答率降至63.5%，不同模型之间存在较大差异。即使该工具很少被调用，回答率的下降也发生，因此不能仅用不必要工具的调用来解释。一个句子的范围感知系统指令恢复了大部分丢失的答案。

View on arXiv Download PDF AI Translation

cs.CL / 46 / 2609.14168

Towards Evolving Context Parameterization for Large Language Models

面向大型语言模型的上下文参数化演化

Shi, Xiaobing, Li, Zherui, Jiang, Yiming, Wang, Kun, Guo, Yufei

Abstract

Context parameterization enables large language models (LLMs) to internalize contexts into reusable model parameters, avoiding repeated processing across subsequent queries. However, existing methods typically assume static contexts and lack explicit mechanisms for distinguishing validity states under continual updates. To study this real-world scenario, we formalized the Memory Updating with Sequential Evolution (MUSE) task and constructed MUSE-bench to evaluate update incorporation and unaffected-information preservation. The resulting challenge requires preserving the global state while adjusting the contribution of memory evidence. Motivated by this, we proposed PLUME, a training-free method that constructs a global update representation, activates memory evidence to form a local parameter view, and adaptively integrates their predictions during decoding. Comprehensive evaluation on MUSE-bench demonstrated PLUME's effectiveness in sequential evolution settings, yielding relative improvements of 29.9% in average ROUGE-L Recall and 54.9% in LLM-as-a-Judge. Our codes are available at: https://github.com/xiaobingshi-LLM/PLUME.

Chinese Translation

上下文参数化使大型语言模型（LLMs）能够将上下文内化为可重用的模型参数，从而避免在后续查询中重复处理。然而，现有方法通常假设上下文是静态的，并缺乏在持续更新下区分有效性状态的明确机制。为了研究这一现实场景，我们正式定义了带有序列演化的记忆更新（Memory Updating with Sequential Evolution, MUSE）任务，并构建了MUSE-bench以评估更新整合和未受影响信息的保留。由此产生的挑战要求在调整记忆证据贡献的同时保持全局状态。基于此，我们提出了PLUME，这是一种无训练的方法，构建全局更新表示，激活记忆证据以形成局部参数视图，并在解码过程中自适应地整合它们的预测。在MUSE-bench上的全面评估表明，PLUME在序列演化设置中的有效性，平均ROUGE-L召回率提高了29.9%，LLM作为评判者的表现提高了54.9%。我们的代码可在以下链接获取：https://github.com/xiaobingshi-LLM/PLUME。

View on arXiv Download PDF AI Translation

cs.CL / 47 / 2609.14178

A Multi-Stage Agentic Framework for Effective Counter-Narrative Generation and Refinement

有效反叙事生成与精炼的多阶段代理框架

Kronfeld, Carmel, Gogawale, Sharva, Kobayashi, Tetsuro, Ben-Gal, Irad

Abstract

The rapid diffusion of hate speech and misinformation on social networks challenges democratic societies, since direct suppression efforts may deepen polarization, fuel public distrusts, and strengthen extremist narratives. LLM-driven counter-narratives (CNs) offer a promising way to reduce those risks, yet their effectiveness depends on rhetorical and stylistic choices that remain poorly understood. We present a multi-stage agent-based framework for generating, refining, and evaluating CNs, applied to pro-Russian hate and misinformation narratives on the war with Ukraine and adaptable to other domains. A pilot experiment with human evaluators identifies effective technique style pairings, such as repetition with emotional framing enhancing persuasiveness. Building on these insights, we introduce a multi-agent refinement process that iteratively improves CNs for persuasiveness, emotional engagement, and shareability. After human validation confirmed improvement, an automated safety analysis shows that our refined CNs match or improve on expert-written counterspeech. A simulated experiment then shows that they reduce the perceived strength of pro-Russian narratives and consistently outperform a vanilla LLM baseline, highlighting a pathway toward scalable, narrative-specific interventions against hate speech and misinformation. Code and data accompanying this work are publicly available at https://github.com/carmelkron/inlg2026-counter-narratives.

Chinese Translation

仇恨言论和虚假信息在社交网络上的迅速传播对民主社会构成挑战，因为直接的压制措施可能加深两极化、加剧公众不信任，并强化极端主义叙事。基于大型语言模型（LLM）的反叙事（CN）提供了一种减少这些风险的有希望的方法，但其有效性依赖于尚不充分理解的修辞和风格选择。我们提出了一种多阶段基于代理的框架，用于生成、精炼和评估反叙事，应用于关于乌克兰战争的亲俄仇恨和虚假信息叙事，并可适应其他领域。与人类评估者的初步实验识别出有效的技术风格配对，例如重复与情感框架结合增强说服力。基于这些见解，我们引入了一种多代理精炼过程，迭代改进反叙事的说服力、情感参与度和可分享性。在人类验证确认改进后，自动化安全分析显示我们的精炼反叙事与专家撰写的反言论相匹配或有所提升。随后，模拟实验表明，它们降低了对亲俄叙事的感知强度，并始终优于基础的普通LLM，突显了针对仇恨言论和虚假信息的可扩展、叙事特定干预的路径。本研究的代码和数据可在 https://github.com/carmelkron/inlg2026-counter-narratives 上公开获取。

View on arXiv Download PDF AI Translation

cs.CL / 48 / 2609.14207

Learning to Refer from Estimated Listener Gaze

从估计的听者注视中学习指称

Wright, Téa, Suhr, Alane

Abstract

We propose to finetune vision-language models to generate more pragmatically optimal referring expressions by transforming observations of incremental listener comprehension, in the form of gaze scanpaths, into learning signals. During training, referring expressions are sampled from the speaker policy being optimized, conditioned on images and target referents; then, a neural listener estimating human gaze behavior maps from images and sampled referring expressions to scanpaths, each represented by a sequence of fixations, with each fixation corresponding to a word in the referring expression. We experiment with several approaches to convert fixation sequences and target referents into token- and sequence-level rewards, which are used to optimize policy parameters. Through evaluation with human listeners, we find that speaker policies trained with gaze-estimating listeners result in significantly more pragmatically-optimal references than base models, reducing sequence length from 15.4 down to 4.0 words while increasing referential success from 75.2 up to 80.0%. Our work demonstrates a promising opportunity for learning to generate utterances through language-based interaction, not only from the explicit signal of communicative success, but also from implicitly-available observations of a listener's process of comprehension.

Chinese Translation

我们提出通过将增量听者理解的观察（以注视扫描路径的形式）转化为学习信号，来微调视觉-语言模型，以生成更具实用性的指称表达。在训练过程中，指称表达是从正在优化的说话者策略中抽样的，条件是图像和目标指称物；然后，一个神经听者估计人类的注视行为，将图像和抽样的指称表达映射到扫描路径，每个扫描路径由一系列注视表示，每个注视对应于指称表达中的一个词。我们实验了几种方法，将注视序列和目标指称物转换为标记级和序列级的奖励，用于优化策略参数。通过与人类听者的评估，我们发现使用注视估计听者训练的说话者策略比基础模型产生了显著更具实用性的指称，序列长度从15.4减少到4.0个词，同时指称成功率从75.2%提高到80.0%。我们的工作展示了通过基于语言的交互学习生成话语的一个有前景的机会，不仅依赖于交流成功的显性信号，还依赖于听者理解过程的隐性观察。

View on arXiv Download PDF AI Translation

cs.CL / 49 / 2609.14226

Corpus Characterization and Inverse Constitutional Fine-Tuning for Style-Aware Radiology Reports

语料特征化与逆向宪法微调用于风格感知的放射学报告

Li, Sarah Y., Renner, Elijah, Ansari, Rayan, Youssef, Alaa

Abstract

Automated radiology report generation has advanced rapidly in diagnostic accuracy, yet generated reports frequently diverge from the stylistic conventions of authentic radiologist writing in structure, diction, and uncertainty language, a gap which has direct implications for clinician trust and user experience. To address this, we characterize stylistic variation across 2,000 reports from the CheXpert Plus dataset using Bio-ClinicalBERT embeddings, UMAP dimensionality reduction, and HDBSCAN clustering, identifying five distinct reporting patterns differing in pathology focus, narrative structure, and lexical preference. Drawing on these findings, we adapt the inverse constitutional AI framework to derive a style-focused constitution from radiologist-written report pairs without requiring a formal preference dataset. This constitution, encoding conventions of tone, diction, uncertainty calibration, and report structure, is incorporated into the supervised fine-tuning of a MedGemma-4B base model on 25,245 CheXpert Plus training pairs. Constitutional fine-tuning produces a substantial increases in text alignment (BLEU-4: 0.006 to 0.308; ROUGE-L: 0.171 to 0.484) relative to the untuned baseline. These gains show a qualitative shift in structural and lexical alignment rather than marginal improvement, as the baseline model produces near-zero scores due to format mismatch. Overall, we establish corpus-level style characterization and constitutional modeling as an effective and data-efficient strategy for producing radiology reports that conform to authentic radiologist writing conventions.

Chinese Translation

自动化放射学报告生成在诊断准确性方面迅速发展，但生成的报告在结构、用词和不确定性语言上常常偏离真实放射科医生写作的风格规范，这一差距对临床医生的信任和用户体验有直接影响。为了解决这一问题，我们利用Bio-ClinicalBERT嵌入、UMAP降维和HDBSCAN聚类对CheXpert Plus数据集中2000份报告的风格变异进行特征化，识别出五种在病理重点、叙述结构和词汇偏好上存在差异的独特报告模式。基于这些发现，我们调整逆向宪法人工智能框架，从放射科医生撰写的报告对中推导出一个以风格为中心的宪法，而无需正式的偏好数据集。该宪法编码了语气、用词、不确定性校准和报告结构的规范，并被纳入对25,245个CheXpert Plus训练对的MedGemma-4B基础模型的监督微调中。宪法微调相较于未调优的基线模型在文本对齐方面产生了显著提升（BLEU-4: 0.006到0.308；ROUGE-L: 0.171到0.484）。这些提升显示出结构和词汇对齐的质变，而非边际改善，因为基线模型由于格式不匹配产生接近零的得分。总体而言，我们确立了语料级别的风格特征化和宪法建模作为一种有效且数据高效的策略，以生成符合真实放射科医生写作规范的放射学报告。

View on arXiv Download PDF AI Translation

cs.CL / 50 / 2609.14245

The Attribution-Compression Frontier in Retrieval-Augmented Generation

检索增强生成中的归因压缩前沿

Mody, Deepanshu

Abstract

Context compression reduces generator input in retrieval-augmented generation, but answer quality alone does not characterize citation attribution. We measure citation attribution across compression methods and budgets, comparing reranking, extractive selection, abstractive summarization, token pruning, and an extract-cluster-rewrite construction on ASQA and QASPER under a fixed generator and primary entailment evaluator. On ASQA at a nominal 0.25 budget (achieved compression 0.08), a RECOMP-style compressor's citations score 0.86 precision against its summaries but 0.12 against source spans under our re-attributability protocol. These estimates depend on a shared NLI model for span recovery and citation scoring and lack independent human calibration. Extractive selection's observed grounded precision ranges from 0.43 to 0.49 across nominal budgets from one-half to one-tenth of the ASQA context, while answer quality declines. For the same RECOMP setting, claim verification after source recovery yields an unsupported rate of 0.88 versus 0.17 when checking summaries. This gap persists beyond structural rejection of missing provenance, but remains evaluator-dependent. A 200-question TRUE T5-XXL audit also finds emitted--grounded gaps under both fixed and recomputed source mappings, without establishing human-calibrated support rates.

Chinese Translation

上下文压缩减少了检索增强生成中的生成器输入，但仅凭答案质量并不能表征引用归因。我们测量了不同压缩方法和预算下的引用归因，比较了重排序、抽取选择、抽象摘要、标记修剪以及在ASQA和QASPER上基于提取-聚类-重写的构造，所有实验均在固定的生成器和主要蕴涵评估器下进行。在ASQA中，在名义预算0.25（实现的压缩为0.08）下，RECOMP风格的压缩器的引用在其摘要中得分为0.86的精度，但在我们的重新归因协议下对源跨度的得分仅为0.12。这些估计依赖于共享的自然语言推理（NLI）模型用于跨度恢复和引用评分，并缺乏独立的人类校准。抽取选择的观察到的基础精度在名义预算从ASQA上下文的一半到十分之一之间范围为0.43到0.49，而答案质量则有所下降。在相同的RECOMP设置下，源恢复后的声明验证产生了0.88的无支持率，而检查摘要时为0.17。这一差距在结构上拒绝缺失来源的情况下依然存在，但仍然依赖于评估者。200个问题的TRUE T5-XXL审计也发现，在固定和重新计算的源映射下，发出的基础差距存在，而未建立人类校准的支持率。

View on arXiv Download PDF AI Translation

cs.CL / 51 / 2609.14256

Document Topic Alignment Metrics for Evaluating Topic Models of Short-Text Public Health Communications on Social Media

用于评估社交媒体上短文本公共卫生传播主题模型的文档主题对齐指标

Xin, Wangjiaxuan, Yin, Shuhua, Ge, Yaorong, Chen, Shi

Abstract

Topic models are widely used to analyze public health-related social media short texts, yet their evaluation remains dominated by metrics that focus entirely on generated topics alone. There is a lack of metrics that quantitatively assess whether assigned topics meaningfully represent the corresponding short-text posts. We propose Document-Topic Alignment metrics (DoTA), an assignment-aware evaluation framework comprising metrics that measure semantic alignment between documents (posts) and their assigned topics. We also introduce margin-based and discriminative variants that capture topic assignment confidence and distinguishability. We evaluate DoTA across five topic models on three public health-related social media datasets from X and compare DoTA metrics with conventional topic-based metrics. Results show that DoTA provides complementary evaluation cues and aligns meaningfully with human evaluations. These findings establish the need for assignment-aware evaluation and demonstrate that the addition of DoTA enables a more comprehensive and practically meaningful evaluation for assessing short-text topic modeling performance.

Chinese Translation

主题模型广泛用于分析与公共卫生相关的社交媒体短文本，但其评估仍然主要依赖于仅关注生成主题的指标。目前缺乏能够定量评估分配主题是否有意义地代表相应短文本帖子的指标。我们提出了文档-主题对齐指标（Document-Topic Alignment metrics，DoTA），这是一个关注分配的评估框架，包括测量文档（帖子）与其分配主题之间语义对齐的指标。我们还引入了基于边际和判别的变体，以捕捉主题分配的信心和可区分性。我们在来自X的三个与公共卫生相关的社交媒体数据集上对五个主题模型进行了DoTA评估，并将DoTA指标与传统的基于主题的指标进行了比较。结果表明，DoTA提供了互补的评估线索，并与人工评估有意义地对齐。这些发现确立了关注分配的评估的必要性，并表明添加DoTA可以为评估短文本主题建模性能提供更全面和具有实际意义的评估。

View on arXiv Download PDF AI Translation

cs.CL / 52 / 2609.14257

DenMark: Robust Semantic Watermarking for Diffusion Language Models

DenMark：用于扩散语言模型的鲁棒语义水印

Ma, Tianhao, Xuan, Weihao, Wu, Dong-Dong, Nooshi, Farshid, Ishida, Takashi, Niu, Gang, Yokoya, Naoto, Sugiyama, Masashi

Abstract

Semantic text watermarks encode signals in meaning rather than surface token choices, offering robustness to paraphrasing and other semantic-preserving edits. Existing semantic watermarking methods are primarily designed for autoregressive language models (ARLMs), where completed candidate units can be generated and scored before generation proceeds. This paradigm does not naturally extend to diffusion language models (DLMs), where semantic units remain incomplete during intermediate denoising steps and tokens may be updated in flexible orders. We propose DenMark, a semantic watermarking framework that injects key-dependent signals directly into the DLM denoising process. DenMark partitions the output into fixed token regions and uses temporary rollouts as semantic lookahead: conditional completions estimate the eventual semantics of an incomplete region, enabling DenMark to select local updates with higher estimated semantic watermark scores. Repeating this procedure across denoising steps progressively accumulates watermark evidence in the final output. For detection, DenMark uses calibrated scanning over candidate unit sizes to remain robust to boundary shifts introduced by semantic attacks. Across four DLMs, three datasets, and four semantic attacks, DenMark achieves the best results across all reported detection metrics in all 48 backbone-dataset-attack combinations. These results demonstrate that DenMark provides an effective mechanism for robust semantic watermarking in DLMs.

Chinese Translation

语义文本水印通过意义而非表面标记选择来编码信号，从而对释义和其他保持语义的编辑提供鲁棒性。现有的语义水印方法主要针对自回归语言模型（ARLMs）设计，在这些模型中，完成的候选单元可以在生成过程中被生成和评分。然而，这一范式并不自然地扩展到扩散语言模型（DLMs），因为在中间去噪步骤中，语义单元仍然是不完整的，并且标记可能以灵活的顺序进行更新。我们提出了DenMark，这是一种将关键依赖信号直接注入DLM去噪过程的语义水印框架。DenMark将输出分为固定的标记区域，并使用临时展开作为语义前瞻：条件完成估计不完整区域的最终语义，使DenMark能够选择具有更高估计语义水印分数的局部更新。在去噪步骤中重复这一过程，逐步在最终输出中累积水印证据。为了检测，DenMark使用对候选单元大小的校准扫描，以保持对语义攻击引入的边界变化的鲁棒性。在四个DLM、三个数据集和四种语义攻击中，DenMark在所有48个基础-数据集-攻击组合中的所有报告检测指标上都取得了最佳结果。这些结果表明，DenMark为DLM中的鲁棒语义水印提供了一种有效机制。

View on arXiv Download PDF AI Translation

cs.CL / 53 / 2609.14288

Editorial routing shapes how computational results are qualified in AI-assisted scientific writing

编辑路由如何影响计算结果在人工智能辅助科学写作中的表述

Kim, Jihan

Abstract

Large language models increasingly analyze computational results and draft manuscripts, making reliable communication as important as correct analysis. Using fixed computational evidence, we tested whether assigning comparisons across modeling choices elsewhere in a research workflow changes manuscript reporting. In constrained sentence-writing tasks, Anthropic's Claude Sonnet 5 often omitted numerical qualifications when detailed comparisons were assigned to a group repository, but retained them more often when the same comparison was assigned to Supporting Information or its own working notes; Claude Opus 5 was less sensitive. These effects did not follow a simple accessibility ordering. A targeted placement rule largely restored sentence-level qualification, whereas a generic accuracy reminder did not. Longer contributions retained numerical qualifications, although some summaries across computational settings were still redirected to the repository. Thus, documenting context within an AI workflow does not ensure its communication where readers encounter the result.

Chinese Translation

大型语言模型越来越多地分析计算结果并撰写手稿，因此可靠的沟通与正确的分析同样重要。我们使用固定的计算证据，测试在研究工作流程中将比较分配到其他建模选择是否会改变手稿的报告。在受限的句子写作任务中，当详细比较被分配到一个组库时，Anthropic的Claude Sonnet 5经常省略数字资格，但在将相同比较分配到支持信息或其自身的工作笔记时则更常保留这些资格；而Claude Opus 5对此的敏感性较低。这些效应并未遵循简单的可及性排序。一项针对性的放置规则在很大程度上恢复了句子级别的资格，而通用的准确性提醒则没有。较长的贡献保留了数字资格，尽管一些跨计算设置的摘要仍被重定向到库中。因此，在人工智能工作流程中记录上下文并不能确保其在读者遇到结果时的有效沟通。

View on arXiv Download PDF AI Translation

cs.CL / 54 / 2609.14302

E2A-Bench: Benchmarking Evidence-to-Action Reliability in Financial Chart Reasoning

E2A-Bench：金融图表推理中证据到行动的可靠性基准测试

Wang, Xiaoya, Xu, Yutong, Wang, Junjie

Abstract

Can financial vision-language models (VLMs) turn chart evidence into reliable action recommendations? Existing hallucination evaluations are mostly claim-centric; they assess whether generated statements are supported, but not whether evidence remains traceable through rationale, confidence, and final action. We introduce E2A-Bench, a 969-query benchmark for financial chart reasoning, constructed from 323 HS300 constituents under three input modalities with deterministic OHLCV-derived evidence anchors. E2A-Bench evaluates grounding, reasoning-action consistency, evidence-confidence calibration, and directional coverage through UCR, RCI, ECI, and NDR, where NDR measures coverage-aware evidence-to-action reliability rather than realized trading performance. Evaluating 20 VLMs reveals three failures hidden by scalar hallucination scores: the lowest-UCR model ranks near the bottom by NDR due to only 6.4% directional coverage; oracle-aided verification reduces unsupported claims but can collapse coverage; and financial fine-tuning amplifies the BUY:SELL ratio by factors of 4.21 to 4.68 across strict base-fine-tuned pairs. These results show that financial VLM evaluation should trace the full evidence-to-action chain rather than rely on a single hallucination score. Code and data: https://github.com/wanng-ide/E2A-Bench

Chinese Translation

金融视觉语言模型（VLMs）能否将图表证据转化为可靠的行动建议？现有的幻觉评估主要以声明为中心；它们评估生成的陈述是否得到支持，但并未评估证据是否通过推理、信心和最终行动保持可追溯性。我们引入了E2A-Bench，这是一个包含969个查询的金融图表推理基准，基于323个HS300成分股在三种输入模式下构建，使用确定性的OHLCV派生证据锚点。E2A-Bench通过UCR、RCI、ECI和NDR评估基础、推理-行动一致性、证据-信心校准和方向覆盖，其中NDR测量的是关注覆盖的证据到行动的可靠性，而非实现的交易表现。对20个VLM的评估揭示了三个被标量幻觉评分掩盖的失败：最低UCR模型因仅有6.4%的方向覆盖而在NDR中排名接近底部；使用oracle辅助验证减少了不支持的声明，但可能导致覆盖崩溃；金融微调在严格的基础-微调对中将BUY:SELL比率放大了4.21到4.68倍。这些结果表明，金融VLM的评估应追踪完整的证据到行动链，而不是依赖单一的幻觉评分。代码和数据：https://github.com/wanng-ide/E2A-Bench

View on arXiv Download PDF AI Translation

cs.CL / 55 / 2609.14320

SpectralShift: Effective Context Window Extension of Gated DeltaNet via Spectral Reparameterization

SpectralShift：通过谱重参数化有效扩展Gated DeltaNet的上下文窗口

Liu, Zian, Hu, Yiwen, Dong, Zican, Xie, Tian, Zhao, Wayne Xin, Ding, Yucheng, Tao, Ran, Dai, Bryan

Abstract

Recently, linear attention layers have been increasingly adopted to replace softmax attention at scale for long-context modeling. However, existing context extension approaches typically apply continued pretraining directly without modifying these layers, overlooking the spectral properties of linear attention state dynamics. In this work, we study long-context extension of Gated DeltaNet (GDN) from a spectral perspective of transition matrix and identify two essential factors governing long-range information retrieval: (1) a sufficiently broad slow spectral band aligned with the target dependency length, and (2) the preservation of fast-decaying modes for state clearing and context switching. Based on this observation, we propose SpectralShift, a spectral reparameterization approach for long-context continual pretraining of GDNs. Specifically, SpectralShift reparameterizes the alpha projections initialization to reshape the decay spectrum by enhancing slow propagation capacity, and further introduces a learning-rate scaling for alpha projections to facilitate long-context training. Experiments show that SpectralShift consistently improves long-context capabilities over training, providing an effective and efficient solution for extending context windows of linear attention models. The code has been open-sourced at https://github.com/RUCAIBox/GDN-SpectralShift.

Chinese Translation

近年来，线性注意力层在长上下文建模中越来越多地被采用，以取代大规模的softmax注意力。然而，现有的上下文扩展方法通常直接应用持续预训练，而不对这些层进行修改，忽视了线性注意力状态动态的谱特性。在本研究中，我们从转移矩阵的谱角度研究了Gated DeltaNet (GDN) 的长上下文扩展，并识别出两个影响长距离信息检索的关键因素：（1）与目标依赖长度对齐的足够宽的慢谱带，以及（2）用于状态清除和上下文切换的快速衰减模式的保留。基于这一观察，我们提出了SpectralShift，一种用于GDN长上下文持续预训练的谱重参数化方法。具体而言，SpectralShift对alpha投影初始化进行重参数化，以通过增强慢传播能力来重塑衰减谱，并进一步引入alpha投影的学习率缩放，以促进长上下文训练。实验表明，SpectralShift在训练过程中始终提高了长上下文能力，为扩展线性注意力模型的上下文窗口提供了一种有效且高效的解决方案。代码已在 https://github.com/RUCAIBox/GDN-SpectralShift 开源。

View on arXiv Download PDF AI Translation

cs.CL / 56 / 2609.14384

Formal Properties of Language as Constraints on Neural Dynamics

语言的形式属性作为神经动态的约束

Murphy, Elliot

Abstract

What must a neural system be capable of to implement language? Current research annotates stimuli with linguistic variables and tests which electrodes, voxels, or language-model layers predict neural activity. Yet predictive success leaves mechanisms under-constrained. Here, we show that algebraic properties of language specify invariants that mechanisms must preserve: non-associative hierarchical grouping, commutativity, recursive closure, access to substructures, and structured workspace transitions. We term this the Neural Admissibility Program (NAP). Syntactic structure building is analyzed algebraically, with candidate mechanisms offered for each requirement: content-addressable workspace memory, graph-structured transient dynamics scheduling structure-building operations (e.g. stable heteroclinic channels), and a phase-coupled sealing operation recording grouping. Simulations show that a corrected Marcolli-Berwick entropy-optimized binding gate preserves grouping only within a narrow commitment band. As an alternative, we propose a novel neural binding operation we term 'Meld': two constituent populations converge through shared synapses, integrate sublinearly, and saturate. Meld is, to our knowledge, the closest neurally plausible composition law to syntactic Merge. It preserves every NAP invariant, uses known cortical operations, and recovers hierarchical structure at every tested depth and temperature. It predicts that effective population dimensionality separates alternative bracketings and that the composite depends on constituent disagreement. Importantly, the laws decoding bracketing most accurately are a priori inadmissible, showing that decoding accuracy alone cannot adjudicate between mechanisms. By specifying how neural dynamics can remain faithful to linguistic structure, the NAP changes the criterion by which neural implementations of cognition are evaluated.

Chinese Translation

神经系统必须具备哪些能力才能实现语言？当前研究通过对刺激进行语言变量标注，测试哪些电极、体素或语言模型层能够预测神经活动。然而，预测成功使得机制的约束不足。在此，我们展示了语言的代数属性指定了机制必须保持的不变性：非结合性层次分组、交换性、递归闭合、对子结构的访问以及结构化工作空间的过渡。我们将其称为神经可接受性程序（Neural Admissibility Program, NAP）。对句法结构构建进行代数分析，并为每个要求提供候选机制：内容可寻址的工作空间记忆、图结构的瞬态动态调度结构构建操作（例如，稳定的异类通道）以及记录分组的相位耦合封闭操作。模拟结果表明，经过修正的Marcolli-Berwick熵优化绑定门仅在狭窄的承诺带内保持分组。作为替代方案，我们提出了一种新颖的神经绑定操作，称为'Meld'：两个组成群体通过共享突触汇聚，进行亚线性整合，并达到饱和。根据我们的知识，Meld是最接近句法合并的神经可行组合法则。它保持每个NAP不变性，使用已知的皮层操作，并在每个测试深度和温度下恢复层次结构。它预测有效群体维度分离替代括号，并且复合依赖于组成部分的不一致性。重要的是，最准确解码括号的法则在先验上是不可接受的，这表明仅凭解码准确性无法裁定机制。通过指定神经动态如何忠实于语言结构，NAP改变了评估认知神经实现的标准。

View on arXiv Download PDF AI Translation

cs.CL / 57 / 2609.14400

Policy Loopholes in Agent Evaluation: When Policy Ambiguity Masquerades as Agent Error

代理评估中的政策漏洞：当政策模糊伪装成代理错误时

Cao, Hongliu

Abstract

Agent benchmarks evaluate policy compliance but assume each policy determines a unique correct action. Natural-language policies can violate this assumption through silence, ambiguity, or contradiction, admitting multiple defensible readings that a single gold trajectory cannot capture. Auditing two $\tau^2$-bench domains, we develop a taxonomy of such policy loopholes and show that affected tasks produce unreliable scores: they lower scores across different models in different ways and make every model less consistent across repeated trials. A cross-domain comparison reveals that exploitability requires both policy ambiguity and tool permissiveness: when policy complexity exceeds what tools can enforce, agents resolve gaps inconsistently and scores become unreliable. Policy specification quality sets the ceiling on evaluation quality. Benchmark developers should audit policies before collecting gold annotations.

Chinese Translation

代理基准评估政策合规性，但假设每个政策决定一个唯一的正确行动。自然语言政策可能通过沉默、模糊或矛盾违反这一假设，承认多种可辩护的解读，而单一的金标准轨迹无法捕捉。通过审计两个 $ au^2$-基准领域，我们开发了此类政策漏洞的分类法，并表明受影响的任务产生不可靠的评分：它们以不同方式降低不同模型的评分，并使每个模型在重复试验中一致性降低。跨领域比较揭示，利用性需要政策模糊性和工具宽容性：当政策复杂性超过工具能够执行的范围时，代理不一致地填补空白，评分变得不可靠。政策规范质量为评估质量设定了上限。基准开发者在收集金标准注释之前应审计政策。

View on arXiv Download PDF AI Translation

cs.CL / 58 / 2609.14448

NeuroActiSep: Detecting Factual Hallucinations from Feed-Forward Neurons in a Single Pass

NeuroActiSep：在单次传递中检测前馈神经元的事实幻觉

Odolou, Ali Derogar, Nazari, Reza, Salehi, Mostafa

Abstract

Hallucination in large language models reduces their reliability and slows adoption. Various white-box studies have used internal representations to detect patterns of truthfulness and factuality. A less-studied approach is to identify feed-forward neurons correlated with hallucination. We propose a method to rank feed-forward neurons at the final prompt token using a custom neuron selection dataset. We transfer the selected neuron identities to train hallucination classifiers on other factual question answering datasets. Our work provides empirical evidence that probes trained using the features from the selected neurons perform on par with probes trained on internal states. We also analyze the distribution of selected neurons and the effect of layer depth on detection performance.

Chinese Translation

大型语言模型中的幻觉降低了其可靠性并减缓了应用的普及。各种白盒研究利用内部表示来检测真实性和事实性的模式。一个较少研究的方法是识别与幻觉相关的前馈神经元。我们提出了一种方法，通过自定义神经元选择数据集对最终提示标记的前馈神经元进行排名。我们将所选神经元的身份转移到其他事实问答数据集上，以训练幻觉分类器。我们的工作提供了实证证据，表明使用所选神经元特征训练的探针在性能上与基于内部状态训练的探针相当。我们还分析了所选神经元的分布及层深度对检测性能的影响。

View on arXiv Download PDF AI Translation

cs.CL / 59 / 2609.14528

Theseus in the Graph: Towards Traceable Multi-Hop Graph Navigation

图中的忒修斯：迈向可追溯的多跳图导航

Hernandez, Eduin E., Garcia, Luis F., Askar, Nurassyl, Diaz, Sergio A., Rini, Stefano

Abstract

Multi-Hop Knowledge Graph Question Answering (KGQA) tasks require models to assemble relational evidence along paths in a KG to answer natural-language questions. However, existing KGQA systems typically focus on predicting the final answer without explicitly modeling or validating the intermediate reasoning steps, obscuring whether the correct answers arise from faithful multi-hop reasoning. To address this limitation, we re-frame multi-hop KGQA as a question-conditioned graph navigation problem. We refer to this formulation as THESEUS - Traceable Hop-wise Evidence SEarch in a Unified Semantics. In this setting, an agent receives a KG, a question, and a topic entity, and traverses a sequence of relations towards the answer, making the reasoning path explicit. To systematically study this formulation, we provide three key contributions. (i) We augment the existing KINSHIP and MQuAKE resources into navigation-ready KGQA datasets with annotated evidence paths and paraphrased questions. (ii) We design evaluation protocols to measure path fidelity, robustness to linguistic variation, and performance across multi-hop and multi-answer questions. (iii) We adapt established path-based KG completion agents - MINERVA, MultiHopKG, and SQUIRE - to operate on full question embeddings rather than symbolic single-relation queries, enabling their trajectories to be guided by natural-language semantics. Together, these contributions advance KGQA research toward systems where traceability is fundamental: answers are accompanied by explicit reasoning paths whose agreement with reference evidence can be systematically evaluated.

Chinese Translation

多跳知识图谱问答（KGQA）任务要求模型沿着知识图谱中的路径组装关系证据，以回答自然语言问题。然而，现有的KGQA系统通常专注于预测最终答案，而未明确建模或验证中间推理步骤，这使得难以判断正确答案是否源于可靠的多跳推理。为了解决这一局限性，我们将多跳KGQA重新框定为一个基于问题的图导航问题。我们将这种表述称为THESEUS——统一语义中的可追溯跳跃证据搜索。在这种设置中，代理接收一个知识图谱、一个问题和一个主题实体，并沿着一系列关系朝向答案进行遍历，使推理路径变得明确。为了系统地研究这一表述，我们提供了三个关键贡献。(i) 我们将现有的KINSHIP和MQuAKE资源扩展为可导航的KGQA数据集，附有注释的证据路径和改写的问题。(ii) 我们设计评估协议，以测量路径的保真度、对语言变异的鲁棒性，以及在多跳和多答案问题上的表现。(iii) 我们调整了已建立的基于路径的KG补全代理——MINERVA、MultiHopKG和SQUIRE，使其能够在完整的问题嵌入上操作，而不是符号单一关系查询，从而使其轨迹能够受到自然语言语义的引导。这些贡献共同推动KGQA研究朝着可追溯性为基础的系统发展：答案伴随明确的推理路径，其与参考证据的一致性可以系统地进行评估。

View on arXiv Download PDF AI Translation

cs.CL / 60 / 2609.14542

Neyshekar: An Open Persian Read-Speech Corpus for Automatic Speech Recognition

Neyshekar：一个用于自动语音识别的开放波斯语朗读语料库

Amirivojdan, Ahmad, Nadiri, Farzad, Alizadeh, Abolfazl, Yaraghi, Shaghayegh

Abstract

Neyshekar is presented as an open Persian read-speech corpus designed for coverage of both formal and informal language, named entities, and longer utterances. In version 6, 62,279 validated recordings totalling 99.02 hours are provided from 190 contributors, with 34,541 distinct recorded prompts. The prompt pool was assembled from human-written material, contextualised homographs, and reviewed language-model-generated text. Text entries were normalised with the shekar library, which supports both formal and informal Persian, and every submitted recording was reviewed against a common validation rubric. About 24% of released clips are classified as informal by an automatic classifier; these register labels are not human-validated. Item-level rater labels are provided for reproducible agreement estimation, opaque per-clip contributor identifiers make the speaker-disjoint partitioning auditable and support contributor-clustered uncertainty estimates, and a text-disjoint test subset is included for evaluation beyond previously seen prompts. Per-contributor recording load and reference-free signal quality are characterised for every released clip. Corpus characteristics are compared with Persian Common Voice under shared processing. Utility is assessed through two ASR architectures, three optimisation seeds, WER and CER, and independent evaluation on the public PSRB sample. Against duration-matched Common Voice training at approximately 32 hours, in-domain WER is reduced by 9.5 points for Whisper and 11.6 points for XLS-R, and by approximately eight points for both architectures on the independent PSRB sample. Transfer and mixture benefits are not consistently observed across architectures and training budgets. The corpus is released under CC0; code and data are made available through the project repository at https://github.com/amirivojdan/neyshekar.

Chinese Translation

Neyshekar被提出作为一个开放的波斯语朗读语料库，旨在覆盖正式和非正式语言、命名实体以及较长的语句。在第6版中，提供了来自190位贡献者的62,279个经过验证的录音，总计99.02小时，包含34,541个不同的录音提示。提示池由人工撰写的材料、上下文化的同形异义词和经过审查的语言模型生成的文本组成。文本条目经过shekar库的规范化，该库支持正式和非正式波斯语，并且每个提交的录音都根据共同的验证标准进行了审查。约24%的发布片段被自动分类器标记为非正式；这些注册标签未经人工验证。提供了项目级评估者标签，以便进行可重复的协议估计，模糊的每个片段贡献者标识符使得说话者不重叠的划分可审计，并支持贡献者聚类的不确定性估计，同时包含一个文本不重叠的测试子集，以便进行超出先前见过的提示的评估。每个发布片段的贡献者录音负载和无参考信号质量被特征化。语料库特征与波斯普通语音在共享处理下进行了比较。通过两种ASR架构、三个优化种子、字错误率（WER）和字符错误率（CER）以及在公共PSRB样本上的独立评估来评估其效用。与大约32小时的时长匹配的普通语音训练相比，在领域内的WER对于Whisper减少了9.5个点，对于XLS-R减少了11.6个点，并且在独立的PSRB样本上，两种架构的WER均减少了约8个点。转移和混合效益在不同架构和训练预算中并未始终观察到。该语料库以CC0许可证发布；代码和数据通过项目仓库https://github.com/amirivojdan/neyshekar提供。

View on arXiv Download PDF AI Translation

cs.CL / 61 / 2609.14565

TATK: Triple-Aware Top-K Learning with Knowledge-Grounded Verification for LLM-based Sequential Recommendation

TATK：基于知识验证的三重感知Top-K学习框架用于基于LLM的序列推荐

Guan, Yuchen, Liu, Jiaye, Han, Yifei, Zhang, Zhenxi, Weng, Yixuan, Li, Bin

Abstract

LLM-based sequential recommenders usually cast next-item prediction as text generation, but this interface is poorly matched to full-catalog top-K ranking. We propose TATK, a Triple-Aware framework that couples Top-K Learning (TKL) with Knowledge-Grounded Verification (KGV) for LLM-based sequential recommendation. Top-K Learning combines context-aware metadata-KG prompt grounding with position-aware top-K rewards, aligning training with ranking utility; Knowledge-Grounded Verification then applies structure-aware reranking over the top-M candidates after a single LLM forward pass, using the same metadata-derived item graph. We evaluate TATK on Musical Instruments, CDs and Vinyl, and Video Games from Amazon Reviews 2023 under a matched R2ec-style full-catalog protocol. Experiments use Gemma-2-2B-It and Qwen2.5-3B-Instruct backbones, compare against sequential, generative, KG-augmented, and reasoning-enhanced baselines, and include component, reward-shape, sequence-perturbation, reranking, relation-quality, and candidate-pool diagnostics. TATK improves over the matched R2ec reproduction on all 36 reported metrics. On NDCG@10, it improves Qwen by 8.05%, 4.26%, and 3.78% on the three datasets, and improves Gemma by 27.03%, 10.52%, and 10.23%, while keeping inference within 1.17x of Base RecPO latency. The diagnostics show that structural evidence is most useful for recoverable top-M candidates with reliable KG support, and should be gated when metadata relations are sparse or noisy.

Chinese Translation

基于LLM的序列推荐系统通常将下一个项目的预测视为文本生成，但这种接口与全目录的Top-K排序不匹配。我们提出了TATK，一个三重感知框架，将Top-K学习（TKL）与基于知识的验证（KGV）结合用于基于LLM的序列推荐。Top-K学习结合了上下文感知的元数据-KG提示基础与位置感知的Top-K奖励，使训练与排序效用对齐；基于知识的验证则在单次LLM前向传播后，对前M个候选项应用结构感知的重新排序，使用相同的元数据派生项目图。我们在2023年亚马逊评论的音乐器材、CD和黑胶唱片以及视频游戏上评估TATK，采用匹配的R2ec风格全目录协议。实验使用Gemma-2-2B-It和Qwen2.5-3B-Instruct作为基础模型，与序列、生成、KG增强和推理增强的基线进行比较，并包括组件、奖励形状、序列扰动、重新排序、关系质量和候选池诊断。TATK在所有36个报告的指标上均优于匹配的R2ec重现。在NDCG@10上，TATK在三个数据集上分别提升了Qwen 8.05%、4.26%和3.78%，并提升了Gemma 27.03%、10.52%和10.23%，同时保持推理延迟在Base RecPO的1.17倍以内。诊断结果表明，结构证据对于具有可靠KG支持的可恢复前M候选项最为有用，当元数据关系稀疏或噪声较大时应进行门控。

View on arXiv Download PDF AI Translation

cs.CL / 62 / 2609.14570

Disentangling Topology and Diversity in Multi-Agent LLMs for Multilingual Low-Resource Emotion Detection

多智能体大语言模型中的拓扑与多样性解耦：用于多语言低资源情感检测

Shernazarov, Ulugbek, Basnayake, Charitha Ruwansiri Weerakon, Jarjini, Abdelkhaleq El, Crespi, Noel, Rajapaksha, Praboda

Abstract

Multi-agent LLM systems combine multiple inference calls, but prior work often confounds how calls are connected with how they are diversified. We study these factors independently: inference topology and source of inter-agent diversity. In a controlled $2 \times 3$ matrix, we cross parallel aggregation and sequential refinement with stochastic sampling, role prompting, and learned QLoRA specialization, under a fixed three-call budget and output protocol within each backbone. Using Qwen2.5-14B-Instruct and Llama-3.1-8B-Instruct, we evaluate all six configurations on multilingual low-resource emotion detection across nine languages. Parallel learned specialization is strongest on Qwen at 52.83 Macro-F1 and reaches 52.94 on Llama. On Qwen it also exceeds same-backbone zero-shot, few-shot, CoT, and seven-call self-consistency baselines. The preferred topology depends on diversity source: sequential refinement helps stochastic and prompted settings, while the learned Width advantage shrinks from 2.83 points on Qwen to 0.17 on Llama. Depth-wise analysis suggests that later learned specialists can overwrite correct early predictions, although the aggregate effect is backbone-dependent. Overall, how agents are differentiated produces larger performance shifts than topology, which should be evaluated jointly with specialization.

Chinese Translation

多智能体大语言模型系统结合了多个推理调用，但以往的研究常常混淆了调用之间的连接方式与它们的多样化方式。我们独立研究这些因素：推理拓扑和智能体间多样性的来源。在一个受控的 $2 imes 3$ 矩阵中，我们在固定的三次调用预算和每个主干网络内的输出协议下，将并行聚合和顺序精炼与随机采样、角色提示和学习的 QLoRA 专门化相结合。使用 Qwen2.5-14B-Instruct 和 Llama-3.1-8B-Instruct，我们评估了六种配置在九种语言上的多语言低资源情感检测表现。并行学习专门化在 Qwen 上表现最佳，达到了 52.83 的 Macro-F1，并在 Llama 上达到了 52.94。在 Qwen 上，它还超过了同主干的零-shot、few-shot、链式推理（CoT）和七次调用自一致性基线。首选的拓扑结构取决于多样性来源：顺序精炼有助于随机和提示设置，而学习的宽度优势从 Qwen 的 2.83 分缩小到 Llama 的 0.17。深度分析表明，后期学习的专家可能会覆盖早期的正确预测，尽管整体效果依赖于主干网络。总体而言，智能体的差异化产生的性能变化大于拓扑，这应与专门化共同评估。

View on arXiv Download PDF AI Translation

cs.CL / 63 / 2609.14571

Domain-specific Pretraining Profile and Transformer Performance: Evidence from Modeling Digital Pragmatics in Arabic-English Code-switching

领域特定预训练特征与变换器性能：来自阿拉伯语-英语代码切换数字语用学建模的证据

Hussen, Fahad Al, University, King Saud, Riyadh, Arabia, Saudi, Shormani, Mohammed Q., University, Ibb, Ibb, Yemen

Abstract

This study highlights the role of domain-specific pretraining profile (DSPP) in Transformer performance for modeling digital pragmatics in Arabic-English code-switched discourse. It evaluates MARBERT and XLM-R(oBERTa), with BERT serving as a general-purpose baseline. The models were evaluated on their ability to classify context-sensitive pragmatic functions in code-switched social-media discourse. 11695 unique X posts were collected via Python and utilized for the study. The study employs a quantitative and qualitative NLP approach, following a supervised pipeline. Findings unveil that MARBERT consistently surpasses XLM-R with validation Macro F1 increasing from 0.39 to 0.84 and validation loss decreasing from 0.55 to 0.19. On an independent test set, it achieved 0.96 accuracy, 0.83 macro precision, 0.87 macro recall, and 0.85 Macro F1, while XLM-R achieved 0.92 test accuracy but a substantially lower Macro F1 of 0.52. This was also supported by class-level performance where MARBERT outperforms XLM-R considerably with F1 improvements ranging from +0.33 to +0.60, demonstrating a clear advantage in modeling Arabic digital pragmatics. The study concludes that Transformer performance depends more on DSPP than multilingual coverage alone, as the latter does not guarantee optimal performance on a highly specialized pragmatic classification task.

Chinese Translation

本研究强调了领域特定预训练特征（DSPP）在变换器性能中的作用，特别是在阿拉伯语-英语代码切换话语中建模数字语用学。研究评估了MARBERT和XLM-R(oBERTa)，并以BERT作为通用基线。模型的评估基于其在代码切换社交媒体话语中分类上下文敏感的语用功能的能力。通过Python收集了11695个独特的X帖子并用于本研究。研究采用定量和定性自然语言处理（NLP）方法，遵循监督学习流程。研究结果显示，MARBERT的表现始终优于XLM-R，验证的宏F1从0.39提高到0.84，验证损失从0.55降低到0.19。在独立测试集上，MARBERT达到了0.96的准确率、0.83的宏精确率、0.87的宏召回率和0.85的宏F1，而XLM-R的测试准确率为0.92，但宏F1显著较低，仅为0.52。这一点在类别级别的表现中也得到了支持，MARBERT在F1提升方面显著优于XLM-R，提升幅度在+0.33到+0.60之间，显示出在建模阿拉伯数字语用学方面的明显优势。研究结论指出，变换器的性能更依赖于DSPP，而非仅仅是多语言覆盖，因为后者并不能保证在高度专业化的语用分类任务中实现最佳性能。

View on arXiv Download PDF AI Translation

cs.CL / 64 / 2609.14646

CompCQR: Compositional Query Generation for Training-Free Conversational Search

CompCQR：用于无训练对话搜索的组合查询生成

Jang, Yunah, Lee, Kang-il, Shin, Joongbo, Jung, Kyomin

Abstract

Multi-turn interactions with LLMs are becoming increasingly common in information-seeking scenarios. However, user queries are often ambiguous and context-dependent, making them ill-suited for direct use as retriever queries. Conversational query reformulation (CQR) addresses this issue by rewriting the current utterance into a stand-alone query grounded in the dialogue history. Recent LLM-based CQR approaches achieve strong performance; however, their repeated LLM invocations and misalignment with downstream retrievers remain challenges. In this work, we begin from the observation that retrievers are highly sensitive to content ordering: simply reordering the same content can lead to changes in retrieval coverage and performance. Based on this, we propose a novel training-free method that generates a very large number of queries with minimal LLM usage by compositionally combining a small set of atomic components. We further apply LLM reasoning to construct a high-quality document set that balances precision and recall while capturing the user's core intent. Our framework generalizes across both open- and closed-source LLMs as well as dense and sparse retrievers. It achieves strong performance on four widely used conversational benchmarks, with up to 22.5% relative MRR improvement over the previous state-of-the-art baseline with far fewer LLM calls.

Chinese Translation

与大型语言模型（LLMs）的多轮交互在信息检索场景中变得越来越普遍。然而，用户查询往往模糊且依赖上下文，使其不适合直接用作检索器查询。对话查询重构（CQR）通过将当前话语重写为基于对话历史的独立查询来解决这一问题。近期基于LLM的CQR方法表现出色；然而，它们重复调用LLM和与下游检索器的不一致性仍然是挑战。在本研究中，我们从观察到检索器对内容顺序高度敏感这一事实出发：简单地重新排序相同内容可能导致检索覆盖率和性能的变化。基于此，我们提出了一种新颖的无训练方法，通过组合一小组原子组件生成大量查询，最大限度地减少LLM的使用。我们进一步应用LLM推理构建一个高质量的文档集，以平衡精度和召回，同时捕捉用户的核心意图。我们的框架在开放源和闭源LLM以及稠密和稀疏检索器之间具有广泛的通用性。在四个广泛使用的对话基准上取得了优异的表现，相较于之前的最先进基线，最高实现了22.5%的相对平均检索率（MRR）提升，同时LLM调用次数大幅减少。

View on arXiv Download PDF AI Translation

cs.CL / 65 / 2609.14648

Optimizing Sparse Outcomes Through Dense Behavioral Signals via Value-Guided Preference Distillation

通过价值引导的偏好蒸馏优化稀疏结果与密集行为信号

Zhu, Ziyi, Cahn, Daniel R., Hull, Thomas D., Stamatis, Caitlin A., Tieleman, Olivier, Freire, Guilherme B., Chen, Jinghong

Abstract

Aligning multi-turn dialogue agents is usually framed as matching turn-level human preferences, yet direct optimization of long-term outcomes is often ineffective and prone to reward hacking. We formulate long-horizon dialogue optimization as a multi-objective reinforcement learning problem and train a multi-head value model that predicts a vector of observed user behaviors across multiple look-ahead horizons. Our findings demonstrate that a scalarized composite of dense auxiliary behavioral signals enables effective credit assignment and optimization of sparse outcomes. However, optimizing unconstrained single-objective proxies might induce policy degradations that are harmful when the agent is exposed to real users. To identify these failure modes prior to deployment, we establish a safety framework combining counterfactual user simulation with a validated dialogue-level outcome model to evaluate preference weightings and policy optimization methods. Finally, we demonstrate that distilling multi-objective value preferences into the policy via reference-anchored preference optimization matches on-policy online RL at a small fraction of its compute budget. Live A/B testing confirms that our distilled policy significantly improves long-term user retention, while simultaneously enhancing the positive behaviors and therapeutic-process markers.

Chinese Translation

多轮对话代理的对齐通常被视为匹配轮次级别的人类偏好，然而直接优化长期结果往往效果不佳且容易受到奖励操控。我们将长期对话优化表述为一个多目标强化学习问题，并训练一个多头价值模型，该模型预测多个前瞻视角下观察到的用户行为向量。我们的研究结果表明，稠密辅助行为信号的标量复合能够有效地进行信用分配和稀疏结果的优化。然而，优化不受约束的单目标代理可能会导致策略退化，这在代理接触真实用户时是有害的。为了在部署前识别这些失败模式，我们建立了一个安全框架，将反事实用户模拟与经过验证的对话级结果模型结合，以评估偏好权重和政策优化方法。最后，我们展示了通过参考锚定的偏好优化将多目标价值偏好蒸馏到策略中，与在线强化学习的政策匹配在计算预算的很小一部分内完成。实时A/B测试确认，我们的蒸馏策略显著提高了长期用户留存，同时增强了积极行为和治疗过程标记。

View on arXiv Download PDF AI Translation

cs.CL / 66 / 2609.14677

The Garden of Forking Prompts: How Users Explore Narrative Space in Story Generation

分叉提示的花园：用户如何在故事生成中探索叙事空间

Deshmukh, Advait, Benedict, Nora, Walsh, Melanie, Antoniak, Maria

Abstract

Large language models (LLMs) have changed the way people engage with stories. Drawing on public chatbot logs, we can see that when users generate stories, they iteratively edit their prompts to explore narrative possibilities, adjusting characters, redirecting plots, and swapping fictional universes. As aggregated data, these prompts represent rich traces of creative preference at scale. Yet story generation evaluation benchmarks rely on static, one-shot prompts that cannot capture this exploratory behavior. In this work, we study how users revise consecutive story prompts in the wild. Using a dataset of naturally occurring user-chatbot conversations, we construct WildStories, a sample of 275,635 story generation prompts (labeled with story format, prompt components, and explicitness), and WildEdits, a collection of 24,291 edit trees that model how users iteratively edit base story prompts and explore branching story possibilities. From these trees we develop a framework of edit types crossing four directions (adding, removing, changing, and extending) with fourteen targets (e.g., plot, character, genre). We then use our datasets and this framework to analyze user behavior in navigating narrative space via LLMs. Finally, we show how automated permutations based on the framework can be used for story generation benchmarking. Content Warning: This paper works with "wild" chatbot logs, which often include toxic and sexually explicit themes.

Chinese Translation

大型语言模型（LLMs）改变了人们与故事互动的方式。通过分析公共聊天机器人日志，我们可以看到，当用户生成故事时，他们会迭代地编辑提示，以探索叙事可能性，调整角色、重定情节并交换虚构宇宙。作为聚合数据，这些提示在规模上代表了丰富的创意偏好痕迹。然而，故事生成评估基准依赖于静态的一次性提示，无法捕捉这种探索行为。在本研究中，我们研究了用户在实际环境中如何修订连续的故事提示。利用自然发生的用户与聊天机器人对话的数据集，我们构建了WildStories，这是一个包含275,635个故事生成提示的样本（标注了故事格式、提示组件和明确性），以及WildEdits，这是一个包含24,291个编辑树的集合，模型化用户如何迭代编辑基础故事提示并探索分支故事可能性。基于这些树，我们开发了一个跨越四个方向（添加、删除、改变和扩展）的编辑类型框架，并设定了十四个目标（例如，情节、角色、类型）。然后，我们利用我们的数据集和该框架分析用户在通过LLMs导航叙事空间时的行为。最后，我们展示了如何基于该框架的自动排列可用于故事生成基准测试。内容警告：本文处理的“野生”聊天机器人日志，通常包含有毒和性暗示的主题。

View on arXiv Download PDF AI Translation

cs.CL / 67 / 2609.14687

One Feedback System Does Not Fit All: Localising Data-to-Text Driver Coaching for the United Kingdom and Nigeria

一种反馈系统并不适合所有人：为英国和尼日利亚本地化数据到文本的驾驶员辅导

Thompson, Iniakpokeikiye Peter, Baig, Jawwad, Reiter, Ehud, Yi, Dewei

Abstract

Data-to-text driver coaching is often presented as a generic pipeline from telematics events to advice. This paper argues that its content requires localisation because usefulness and credibility depend on drivers' knowledge, prevalent risks, regulation, infrastructure, and available data. Two independently developed systems in the United Kingdom and Nigeria are compared by tracing requirements through content selection, generation, and field evaluation. The UK system prioritises post-trip reflection, explanations tied to road and place context, and tone-sensitive wording. The Nigerian system combines legally grounded, once-daily Tips based on detected events with weekly persuasive Reports; it foregrounds safety education and alcohol-related risk in response to reported gaps in formal training and traffic-rule knowledge, as well as local road-safety priorities. Reliable speed-limit metadata supported speeding feedback in the UK, whereas its scarcity led the Nigerian evaluation to exclude speeding from its outcome metric. Both interventions were associated with reduced distance-normalised unsafe-event rates in their own field studies, although their designs and metrics preclude an effect-size comparison. The analysis yields a requirements-to-content design process for localising safety-critical NLG without treating a high-income deployment as the default.

Chinese Translation

数据到文本的驾驶员辅导通常被视为从远程信息处理事件到建议的通用流程。本文认为，其内容需要本地化，因为有用性和可信度依赖于驾驶员的知识、普遍风险、法规、基础设施和可用数据。通过追踪内容选择、生成和现场评估的需求，比较了在英国和尼日利亚独立开发的两个系统。英国系统优先考虑行程后反思、与道路和地点背景相关的解释以及语气敏感的措辞。尼日利亚系统结合了基于检测事件的法律依据的每日一次提示和每周的劝导报告；它强调安全教育和与酒精相关的风险，以应对在正式培训和交通规则知识方面报告的差距，以及当地的道路安全优先事项。在英国，可靠的限速元数据支持了超速反馈，而其稀缺性导致尼日利亚评估将超速排除在结果指标之外。尽管两种干预措施在各自的实地研究中与减少距离标准化的不安全事件率相关，但其设计和指标使得效果大小比较不具可行性。该分析为本地化安全关键的自然语言生成（NLG）提供了一个需求到内容的设计过程，而不将高收入部署视为默认选项。

View on arXiv Download PDF AI Translation

cs.CL / 68 / 2609.14739

Building Legal Reward Models for Grounding and Abstention

构建法律奖励模型以实现推理和回避

Franzone, Rilton, Noël, Valentin, Wang, Puyu, Torr, Philip, Fehr, Fabio J.

Abstract

Large language models are increasingly used in high-stakes domains such as law, where systems must ground their reasoning in retrieved evidence and abstain when that evidence is insufficient. However, existing reward models are largely optimised for general preferences rather than contextual grounding, limiting their ability to evaluate these behaviours in retrieval-augmented generation (RAG) settings. We introduce a framework for transforming existing legal QA datasets into contextual preference data and use it to construct LegalRewardBench (LRB), a benchmark for evaluating grounded legal generation under noisy and insufficient retrieval conditions. Across general and legal contextual evaluation, we find that contextual DPO improves grounded evaluation, but performance is sensitive to preference-data construction. Length-balanced augmentation substantially improves grounded legal evaluation, with the strongest configuration combining length-balanced legal and general contextual preference data and improving performance by up to $\mathbf{+25.6}$pp over baseline. We further find evidence of cross-jurisdiction transfer: models contextually refined primarily on Victorian criminal-law data improve grounded evaluation on external US legal benchmarks, including a $\mathbf{+16.2}$pp improvement on \textsc{Housing Statute QA}. Together, these results provide a reproducible foundation for constructing and evaluating grounded legal reward models in retrieval-augmented settings.

Chinese Translation

大型语言模型在法律等高风险领域的应用日益增多，这些系统必须将其推理建立在检索到的证据之上，并在证据不足时选择回避。然而，现有的奖励模型主要针对一般偏好进行优化，而非上下文基础，这限制了它们在检索增强生成（RAG）环境中评估这些行为的能力。我们提出了一种将现有法律问答（QA）数据集转化为上下文偏好数据的框架，并利用该框架构建了法律奖励基准（LegalRewardBench，LRB），这是一个用于评估在嘈杂和不足检索条件下的法律生成的基准。在一般和法律上下文评估中，我们发现上下文DPO（Direct Preference Optimization）改善了基础评估，但性能对偏好数据的构建敏感。长度平衡增强显著提高了基础法律评估，最强的配置结合了长度平衡的法律和一般上下文偏好数据，使性能相比基线提高了高达$ extbf{+25.6}$个百分点。我们进一步发现跨司法管辖区转移的证据：主要在维多利亚州刑法数据上进行上下文优化的模型在外部美国法律基准上改善了基础评估，包括在 extsc{Housing Statute QA}上提高了$ extbf{+16.2}$个百分点。综合来看，这些结果为在检索增强环境中构建和评估基础法律奖励模型提供了可重复的基础。

View on arXiv Download PDF AI Translation

cs.CL / 69 / 2609.14743

Quantifying the Generation Modality Gap in Speech-Text Language Models

量化语音-文本语言模型中的生成模态差距

Chou, Ju-Chieh, Zhou, Jiawei, Livescu, Karen

Abstract

Pure speech language models often lag behind text and speech-text language models in generating coherent content, but this gap is difficult to quantify because speech and text systems are typically evaluated with different metrics and trained on different data. We study the speech-text modality gap in a family of spoken language models, based on flow matching for continuous acoustic feature generation. We construct a unified generation-based evaluation suite that compares speech-only, text-only, and speech-text language models trained on matched data distributions and evaluated in matched generation settings. We evaluate generated continuations along multiple dimensions: semantic coherence, measured by transcribing generated speech and scoring it with a reference language model; local phonetic structure, measured by phone n-gram distributional statistics; speaker consistency and acoustic quality; and emotion-based distributional metrics. Across datasets, we find that joint speech-text modeling substantially improves semantic coherence. However, the improvement is not uniform across metrics: phone-level metrics change only modestly, speaker similarity and predicted quality are lower for speech-text continuations, while emotion-based distributional metrics improve. Compared with larger-scale speech-only models, our speech-text model closes much of the scaling gap in transcript-based semantic coherence, suggesting that text provides an efficient semantic training signal for spoken language modeling.

Chinese Translation

纯语音语言模型在生成连贯内容方面通常落后于文本和语音-文本语言模型，但由于语音和文本系统通常使用不同的指标进行评估并在不同的数据上进行训练，这一差距难以量化。我们研究了一类基于流匹配的连续声学特征生成的口语语言模型中的语音-文本模态差距。我们构建了一个统一的基于生成的评估套件，比较了在匹配数据分布上训练并在匹配生成设置中评估的纯语音、纯文本和语音-文本语言模型。我们从多个维度评估生成的延续内容：语义连贯性，通过转录生成的语音并使用参考语言模型进行评分来衡量；局部音位结构，通过音位n-gram分布统计来衡量；说话者一致性和声学质量；以及基于情感的分布指标。在多个数据集上，我们发现联合语音-文本建模显著提高了语义连贯性。然而，这种改善在各个指标上并不均匀：音位级别的指标变化仅适度，语音-文本延续的说话者相似性和预测质量较低，而基于情感的分布指标则有所改善。与更大规模的纯语音模型相比，我们的语音-文本模型在基于转录的语义连贯性上缩小了大部分规模差距，这表明文本为口语语言建模提供了有效的语义训练信号。

View on arXiv Download PDF AI Translation

cs.CL / 70 / 2609.14770

How broad is that claim? Mapping Generalisation in NLP Research

该主张有多广泛？映射自然语言处理研究中的概括性

Diao, Chenxin, Stepanova, Nataliya, Allaway, Emily

Abstract

Generalisations are common in scientific communication, even though they are semantically ambiguous. An automated method is needed to identify and categorise claims according to their level of generalisation, in order help detect an over-reliance on generalisations and possible misrepresentations of scientific findings. We introduce a comprehensive taxonomy of generalisations in the scientific domain, NLPGenX, which labels claims according to their level of generality and framing within the text. We operationalise this taxonomy with an LLM-powered framework, NLPGenA, that automatically classifies sentences from scientific articles into 5 different generalisation classes. We validate our framework with human annotators and use the framework to construct a large-scale dataset of NLP papers annotated according to generality, with auxiliary labels for hedging and vague descriptors (NLPGens). We use NLPGens to analyse the use of generalisations in NLP papers across multiple venues and subdomains, and to examine associations with citation counts, hedging, and vague descriptors.

Chinese Translation

概括性在科学交流中很常见，尽管它们在语义上存在模糊性。需要一种自动化的方法来识别和分类主张，以其概括程度为依据，从而帮助检测对概括性的过度依赖以及可能对科学发现的误表述。我们引入了一个全面的科学领域概括性分类法 NLPGenX，该分类法根据主张的普遍性和文本中的框架对其进行标注。我们通过一个基于大型语言模型（LLM）的框架 NLPGenA 来实现这一分类法，该框架能够自动将科学文章中的句子分类为5种不同的概括性类别。我们通过人工标注者验证我们的框架，并利用该框架构建了一个大规模的数据集 NLPGens，该数据集对自然语言处理（NLP）论文进行了概括性标注，并附有对模糊性和不确定性描述的辅助标签。我们使用 NLPGens 分析了多个场合和子领域中 NLP 论文中概括性的使用情况，并考察了其与引用次数、模糊性和不确定性描述之间的关联。

View on arXiv Download PDF AI Translation

cs.CL / 71 / 2609.14773

Pull: Lazy Materialization of Working Memory for Stateful LLM Conversations

Pull：有状态大语言模型对话的惰性物化工作记忆

Chen, Jiangang

Abstract

As LLM conversations grow to hundreds of turns, full-context injection incurs $O(N^2)$ cumulative token costs, while lossy summarization or hard truncation irreversibly discards historical state. We propose Pull, a session router that maintains an addressable metadata directory via a local, deterministic Purifier (zero LLM calls, millisecond-level latency). At query time, the LLM lazily materializes only the turns it needs; unmaterialized turns remain accessible but collapsed. Unlike irreversible compression, Pull's materialization is reversible; subsequent queries can expand any collapsed turn. On LoCoEval (128 conversations, 12,780 turns), Pull reduces per-query context tokens (Phase 2) by 75.1 percent on single-hop tasks with equivalent quality ($\Delta = -0.002$, n.s.) and by 72.0 percent on multi-hop tasks with no quality loss ($\Delta = +0.017$). A controlled routing benchmark (7,831 queries x 10 methods) shows that entity lifecycle tracking is empirically a prerequisite for distance-independent routing. On BEAM 1M (14 conversations, 263 questions), Pull improves F1 by +55.2 percent over a truncation baseline.

Chinese Translation

随着大语言模型（LLM）对话的回合数增加到数百次，完全上下文注入会产生 $O(N^2)$ 的累积令牌成本，而有损摘要或硬截断则不可逆地丢弃历史状态。我们提出了 Pull，一种会话路由器，通过本地的确定性 Purifier（零 LLM 调用，毫秒级延迟）维护一个可寻址的元数据目录。在查询时，LLM 仅惰性物化所需的回合；未物化的回合仍然可访问但被压缩。与不可逆压缩不同，Pull 的物化是可逆的；后续查询可以扩展任何被压缩的回合。在 LoCoEval（128 次对话，12,780 回合）中，Pull 在单跳任务上将每次查询的上下文令牌数（第 2 阶段）减少了 75.1%，且质量相当（$ ext{Δ} = -0.002$, n.s.），在多跳任务上则减少了 72.0% 且没有质量损失（$ ext{Δ} = +0.017$）。一个受控路由基准（7,831 次查询 x 10 种方法）表明，实体生命周期跟踪在经验上是距离无关路由的先决条件。在 BEAM 1M（14 次对话，263 个问题）中，Pull 相较于截断基线提高了 +55.2% 的 F1 值。

View on arXiv Download PDF AI Translation

cs.CL / 72 / 2609.14779

Func-R1: Incentivizing Mathematical Function Reasoning in Multimodal Large Language Models

Func-R1：激励多模态大型语言模型中的数学函数推理

Yin, Mingze, Wang, Xiaohan, Li, Dian, Yao, Haichao, Zhao, Yilin, Chen, Youjun, Liu, Gang, Chen, Jintai, Zhu, Yiheng, Hsieh, Chang-Yu, Pan, Aimin

Abstract

Performing deliberate mathematical reasoning in visual contexts is a hallmark of advanced Multimodal Large Language Models (MLLMs) and requires a sophisticated synthesis of perceptual grounding and symbolic logic. However, in the realm of mathematical functions, our investigation reveals a critical modality interference phenomenon: even advanced models, while performing textual computational reasoning, tend to disregard or misinterpret essential visual cues. To address this challenge, we propose Func-R1, which synergistically harmonizes precise visual perception and rigorous logical reasoning. Concretely, built upon an explicitly decoupled architecture, we employ a hierarchical post-training framework to progressively identify critical visual evidence and conduct in-depth theoretical reasoning. Furthermore, the Perception-Aligned Theoretic Optimization (PATO) strategy is proposed to steer policy updating towards internalizing fundamental theoretical properties while dynamically rectifying heterogeneous visual information throughout the reasoning process. Extensive experiments across diverse benchmarks demonstrate that Func-R1 delivers the optimal performance among open-source MLLMs, even surpassing GPT-5 with an 8.4% improvement on MathVerse's function-oriented tasks.

Chinese Translation

在视觉背景下进行深思熟虑的数学推理是先进的多模态大型语言模型（MLLMs）的一个标志，要求感知基础与符号逻辑的复杂综合。然而，在数学函数的领域，我们的研究揭示了一种关键的模态干扰现象：即使是先进的模型，在进行文本计算推理时，往往会忽视或误解重要的视觉线索。为了解决这一挑战，我们提出了Func-R1，它协同地和谐了精确的视觉感知与严谨的逻辑推理。具体而言，基于显式解耦的架构，我们采用分层后训练框架，逐步识别关键的视觉证据并进行深入的理论推理。此外，我们提出了感知对齐理论优化（Perception-Aligned Theoretic Optimization, PATO）策略，以引导策略更新，内化基本的理论属性，同时在推理过程中动态修正异质视觉信息。广泛的实验结果表明，Func-R1在各种基准测试中表现最佳，甚至在MathVerse的功能导向任务上超越了GPT-5，提升了8.4%。

View on arXiv Download PDF AI Translation

cs.CL / 73 / 2609.14795

Mind Which Bird You Favour: Parameterizing Adequacy-Fluency Balance in Meta-Evaluation of Machine Translation

关注您偏爱的鸟类：在机器翻译的元评估中参数化充分性与流畅性的平衡

Shayegh, Behzad, Kazemi, Niloofar

Abstract

There is a tradeoff in machine translation meta-evaluation between prioritizing alignment with adequacy versus fluency. The balance depends on the combination of translation systems in the meta-evaluation dataset. This system set is a small, filtered sample whose characteristics change heavily across years and language pairs; it does not represent the true system distribution. Consequently, the adequacy-fluency balance is often unrepresentative and subject to change. For sensitive domains, controlling this balance is critical. We expose this balance as a tunable choice. To achieve a target balance, we reweight existing systems while minimizing distortion from uniform weighting, ensuring the evaluated systems remain real and representative. We provide an exact optimization algorithm with theoretical guarantees and pruning mechanisms to compute these weights. To validate meta-evaluation internal consistency, we design a scorer-augmentation framework that establishes a known relative identity for the scorers. Results demonstrate that our reweighting method effectively controls the adequacy-fluency balance and preserves the internal consistency, outperforming prior approaches. Finally, we analyze the performance of popular scorers across a sweep of this parameter.

Chinese Translation

在机器翻译的元评估中，充分性与流畅性之间存在权衡。这个平衡依赖于元评估数据集中翻译系统的组合。该系统集是一个小型的、经过筛选的样本，其特征在不同年份和语言对之间变化很大；它并不代表真实的系统分布。因此，充分性与流畅性的平衡往往不具代表性，并且容易变化。对于敏感领域，控制这一平衡至关重要。我们将这一平衡揭示为一个可调的选择。为了实现目标平衡，我们在最小化均匀加权带来的扭曲的同时，对现有系统进行重新加权，确保评估的系统仍然真实且具有代表性。我们提供了一种具有理论保证的精确优化算法和修剪机制来计算这些权重。为了验证元评估的内部一致性，我们设计了一个评分器增强框架，建立了评分器的已知相对身份。结果表明，我们的重新加权方法有效地控制了充分性与流畅性的平衡，并保持了内部一致性，优于以往的方法。最后，我们分析了流行评分器在这一参数范围内的表现。

View on arXiv Download PDF AI Translation

cs.CL / 74 / 2609.14817

Tone on a Budget: A Reference-Free Metric for Lexical Tone in Massively Multilingual Text-to-Speech

预算内的音调：一种无参考的多语种文本到语音的词汇音调度量

Daudu, Moses, Bamidele, Adeola Enitan, Bezaleel, Honor-Jesus

Abstract

In Yor\`ub\'a, pitch alone separates \d{o}k\d{o} (husband, Mid), \d{o}k\d{\`o} (vehicle, Low), and \d{o}k\d{\'o} (hoe, High) -- the diacritics ARE the tone marks. Yet character error rate (CER), the standard automated metric for text-to-speech (TTS), is in practice computed from ASR output that drops those marks: a synthesizer can ace CER and still say vehicle for husband. We introduce DunDun -- named for the d\`und\'un, the Yor\`ub\'a talking drum that speaks through pitch alone -- an automated, reference-free lexical-tone metric that needs no tone-labelled corpus. The gold High/Mid/Low sequence is read from the input text's diacritics (in TTS that text exists by construction, so no reference recording is needed); the prediction comes from the audio's pitch track. We validate three ways. Flattening pitch with PSOLA resynthesis collapses DunDun while CER does not move. Inverting High and Low in the answer key of 300 native recordings drives the two-class readout to 0.14, symmetrically below its 0.35 chance level -- a consistency check on the scoring path, not independent evidence. And three native listeners, over 67 blind A/B trials, pick the tone-correct clip 89.6% of the time (95% CI 80.0-94.8; p < 1e-4); whether DunDun tracks those judgements trial by trial is not resolved at this sample size. Applied to a massively multilingual zero-shot TTS model, DunDun shows what CER cannot: Yor\`ub\'a tone sits near the native anchor before any Yor\`ub\'a fine-tuning (0.567 +/- 0.02 over five decode seeds vs. 0.596; chance 0.33), despite the 21.4% CER the model's own paper reports; and a few hours of clean audio halve CER (5.6% to 2.7% by 5h, 1.7% by 15h) while tone saturates within the hour. On non-tonal Swahili, CER already captures the gains: the metric a language needs is language-dependent. We release the metric and the complete validation protocol.

Chinese Translation

在约鲁巴语中，仅凭音高就能区分 {o}k {o}（丈夫，中音）、 {o}k {}o（车辆，低音）和 {o}k {'}o（锄头，高音）——这些变音符号即是音调标记。然而，字符错误率（CER）作为文本到语音（TTS）的标准自动度量，实际上是从去掉这些标记的自动语音识别（ASR）输出中计算得出的：一个合成器可能在CER上表现优秀，却仍然将“丈夫”读作“车辆”。我们引入了DunDun——以约鲁巴语的打击乐器d{u}nd{u}n命名，该乐器仅通过音高发声——这是一种自动的、无参考的词汇音调度量，不需要带音调标记的语料库。高/中/低的标准序列是从输入文本的变音符号中读取的（在TTS中，该文本是构造而成的，因此不需要参考录音）；预测来自音频的音高轨迹。我们通过三种方式进行验证。使用PSOLA重合成平坦化音高时，DunDun的结果崩溃，而CER则没有变化。在300个母语录音的答案键中反转高音和低音，使得两类的读取结果降至0.14，明显低于其0.35的机会水平——这是对评分路径的一致性检查，而非独立证据。三位母语听众在67个盲A/B试验中选择音调正确的片段的概率为89.6%（95%置信区间80.0-94.8；p < 1e-4）；DunDun是否逐次跟踪这些判断在此样本量下尚未解决。应用于一个大规模多语种零样本TTS模型，DunDun展示了CER无法体现的内容：约鲁巴语的音调在任何约鲁巴语微调之前就接近母语锚点（五个解码种子的结果为0.567 +/- 0.02，对比0.596；机会水平为0.33），尽管该模型自身的论文报告了21.4%的CER；而几小时的清晰音频将CER减半（5小时从5.6%降至2.7%，15小时降至1.7%），而音调在一小时内就达到饱和。在非音调的斯瓦希里语中，CER已经捕捉到了这些收益：所需的度量是语言依赖的。我们发布了该度量及完整的验证协议。

View on arXiv Download PDF AI Translation

cs.CL / 75 / 2609.14819

A primer on evaluation methods for large language models in healthcare

医疗领域大型语言模型评估方法概述

McKinney, Suzannah E, Vu, Phuc, Justice, Samuel A, Humphries, Christopher, Pradhan, Alyssa, Keyes, Timothy J, Bizzo, Bernardo C, Dreyer, Keith J, Mercaldo, Sarah F, Hillis, James M

Abstract

Large language models (LLMs) have a growing range of applications in medicine, and their evaluation is critical for ensuring they provide benefit and not harm. This evaluation can be more challenging than traditional machine learning for many reasons, including probabilistic and open-ended outputs, and behavior that shifts with prompt design and accumulated context. This review covers four key areas of LLM evaluation: principles of study design, statistical methods, capability evaluation and clinical context evaluation. Capability evaluation considers different benchmarks, including multiple-choice, agentic and multi-turn benchmarks, alongside operational metrics like token usage. Clinical context evaluation addresses establishing accuracy of free text outputs, such as human review and LLM-as-a-judge, and clinical trial approaches. Across sections, we describe underlying concepts and potential pitfalls, while emphasizing the importance of aligning evaluation methods with the research question. Together, this article aims to provide a pragmatic basis for designing and executing rigorous evaluations of healthcare LLMs.

Chinese Translation

大型语言模型（LLMs）在医学领域的应用日益广泛，其评估对于确保其带来益处而非伤害至关重要。由于多种原因，这种评估可能比传统机器学习更具挑战性，包括概率性和开放式输出，以及随着提示设计和累积上下文而变化的行为。本文综述了LLM评估的四个关键领域：研究设计原则、统计方法、能力评估和临床背景评估。能力评估考虑了不同的基准，包括多项选择、代理性和多轮基准，以及诸如令牌使用等操作性指标。临床背景评估则关注于建立自由文本输出的准确性，例如人类评审和LLM作为评判者的方式，以及临床试验方法。在各个部分中，我们描述了基本概念和潜在陷阱，同时强调将评估方法与研究问题对齐的重要性。总体而言，本文旨在为设计和实施严格的医疗LLM评估提供务实的基础。

View on arXiv Download PDF AI Translation

cs.CL / 76 / 2609.14823

MedTRACE: Tool-Augmented Multimodal Clinical Reasoning Agents for Evidence-Grounded Decision-Making

MedTRACE：工具增强的多模态临床推理代理用于基于证据的决策制定

Lu, Ji, Liu, Lifei, Yu, Haoran, Wang, Xianglong, Fang, Yiru, Yang, Kuo, Duan, Huiran, Gou, Jianping

Abstract

Multimodal clinical decision-making requires reliable reasoning over heterogeneous evidence from electronic health records, medical images, and physiological signals. Existing models typically map these inputs directly to diagnoses without explicitly assessing evidence sufficiency, tool-use requirements, or diagnostic uncertainty. This paper presents MedTRACE, a tool-augmented multimodal clinical reasoning agent for evidence-grounded decision-making. MedTRACE uses modality-specific encoders to construct a unified patient-state representation and performs an iterative loop of hypothesis formation, toolaware deliberation, and evidence verification. It dynamically invokes visual grounding, evidence retrieval, and structured parsing tools to locate diagnosis-relevant regions, retrieve clinical knowledge and similar cases, and extract structured findings. The acquired evidence enters an evidence memory, where a consistency verifier confirms or revises the current hypothesis. MedTRACE outputs a diagnosis together with supporting evidence, an auditable reasoning trace, and calibrated confidence. Experiments on multiple multimodal clinical diagnosis benchmarks show that MedTRACE improves diagnostic accuracy by 5.4% and AUROC by 4.7 percentage points over the strongest baseline. It also improves evidenceselection F1 by 8.2 percentage points and visual-grounding IoU by 6.5 percentage points, reduces expected calibration error by 31.6%, and decreases unsupported diagnostic errors by 27.8%. These results demonstrate that active evidence acquisition and verification improve the accuracy, interpretability, and reliability of multimodal clinical decisionmaking.

Chinese Translation

多模态临床决策制定需要对来自电子健康记录、医学影像和生理信号的异构证据进行可靠的推理。现有模型通常将这些输入直接映射到诊断上，而未明确评估证据的充分性、工具使用的要求或诊断的不确定性。本文提出了MedTRACE，一种工具增强的多模态临床推理代理，用于基于证据的决策制定。MedTRACE使用特定模态的编码器构建统一的患者状态表示，并执行假设形成、工具意识的深思熟虑和证据验证的迭代循环。它动态调用视觉定位、证据检索和结构化解析工具，以定位与诊断相关的区域、检索临床知识和相似案例，并提取结构化发现。获取的证据进入证据记忆，其中一致性验证器确认或修正当前假设。MedTRACE输出诊断及其支持证据、可审计的推理轨迹和校准信心。在多个多模态临床诊断基准测试中的实验表明，MedTRACE在诊断准确性上提高了5.4%，在AUROC上提高了4.7个百分点。它还提高了证据选择的F1分数8.2个百分点，视觉定位的IoU提高了6.5个百分点，降低了期望校准误差31.6%，并减少了27.8%的不支持诊断错误。这些结果表明，主动的证据获取和验证提高了多模态临床决策的准确性、可解释性和可靠性。

View on arXiv Download PDF AI Translation

cs.CL / 77 / 2609.14825

Route, Don't Fix: Regime-Dependent Decoding Correction and a Trajectory-Gated Router for Reliable Clinical LLM Answer Selection

路径而非修正：基于体制的解码校正与可靠临床大语言模型答案选择的轨迹门控路由器

Dong, Zeyu, Wang, Benjamin, Jin, Joyee W.

Abstract

Large language models (LLMs) are often deemed unsafe for clinical question answering because of their tendency to hallucinate. Retrieval augmentation, fine-tuning, and external verifiers require new infrastructure that clinical governance must approve and may add latency or extra model calls. Inference-time correction uses the model's internal logit signals, but a fixed transformation need not suit every question. A corrector that improves accuracy by about ten percentage points on a truthfulness stress test yields negligible gains on clinical multiple-choice benchmarks, where instruction tuning concentrates output probability on one answer and leaves low terminal entropy. We introduce ALTAS, which reads terminal entropy and late-layer linearity ($R^2$) from one forward pass to choose per question between greedy decoding and late-layer trajectory correction. No classifier, probe, or head is trained; the router operates on candidate-answer logits and adds 6.5% latency overhead. Applied to every question, the correction improves TruthfulQA over greedy at 3B and 8B by 11.4 and 10.0 percentage points, respectively ($p<10^{-10}$). Gated per question, ALTAS retains gains of 8.3 to 9.5 percentage points while keeping MedQA, PubMedQA, and MedHallu within a one-percentage-point do-no-harm band, with no statistically significant differences from greedy. The method passes verification sweeps over frozen thresholds, the scoring rule, and the domain label.

Chinese Translation

大型语言模型（LLMs）常被认为不适合用于临床问答，因为它们容易产生幻觉。检索增强、微调和外部验证器需要新的基础设施，这需要临床管理的批准，并可能增加延迟或额外的模型调用。推理时的校正利用模型的内部对数信号，但固定的转换不一定适合每个问题。通过在真实度压力测试中提高约十个百分点的准确率的校正器，在临床多项选择基准测试中几乎没有收益，因为指令调优将输出概率集中在一个答案上，导致低终端熵。我们引入了ALTAS，它通过一次前向传播读取终端熵和晚层线性度（$R^2$），以便在每个问题之间选择贪婪解码和晚层轨迹校正。无需训练分类器、探针或头部；路由器在候选答案的对数上操作，并增加6.5%的延迟开销。应用于每个问题，校正在3B和8B上分别将TruthfulQA的表现提高了11.4和10.0个百分点（$p<10^{-10}$）。在每个问题上进行门控，ALTAS保持了8.3到9.5个百分点的增益，同时将MedQA、PubMedQA和MedHallu保持在一个百分点的无害范围内，与贪婪解码没有统计显著差异。该方法通过冻结阈值、评分规则和领域标签的验证检查。

View on arXiv Download PDF AI Translation

cs.CL / 78 / 2609.14829

Enemray: Toward Capable Language Models for Hassaniya

Enemray：朝着能够处理哈萨尼亚语的语言模型迈进

Ahmed, Cheikh

Abstract

We introduce Enemray, a Hassaniya-centric language model that enables general-purpose interaction in Hassaniya. Enemray is trained around a stability--plasticity objective: acquire strong Hassaniya linguistic and cultural competence while preserving the general reasoning, multilingual, instruction-following, and safety behaviors of a capable instruction-tuned model. The development pipeline separates language acquisition from behavioral specialization. A separately assembled continual-pretraining corpus provides broad exposure to natural Hassaniya and Mauritanian text; layer-selective continual pretraining learns a compact language-specific parameter update; that update is transferred into the instruction-tuned parameter space; and supervised post-training develops conversational, cultural, literary, task-oriented, and cross-lingual behavior. The supervised corpus integrates selected public Hassaniya and Mauritanian resources with a substantially larger body of newly collected, reconstructed, curated, and constructed instruction data, while policy-generated replay provides a retention signal from the reference model's own behavior distribution. The resulting collection is substantially larger and broader in purpose than existing Hassaniya text resources. In evaluation, Enemray achieves the strongest English to Hassaniya translation among the compared open and proprietary models and the highest overall score on Mauritanian translation error detection, while retaining most of the general capabilities of its instruction-tuned base model on mathematical reasoning, knowledge, code generation, and function calling. This report describes the motivation, data construction, model design, training methodology, and evaluation of Enemray.

Chinese Translation

我们介绍了Enemray，这是一种以哈萨尼亚语为中心的语言模型，能够实现哈萨尼亚语的通用交互。Enemray的训练围绕稳定性-可塑性目标展开：在保持强大的哈萨尼亚语言和文化能力的同时，保留一个经过能力调优的模型的通用推理、多语言、遵循指令和安全行为。开发流程将语言习得与行为专业化分开。一个单独组装的持续预训练语料库提供了对自然哈萨尼亚语和毛里塔尼亚文本的广泛接触；层选择性持续预训练学习紧凑的语言特定参数更新；该更新被转移到指令调优的参数空间中；而监督后训练则发展了对话、文化、文学、任务导向和跨语言的行为。监督语料库整合了选定的公共哈萨尼亚和毛里塔尼亚资源，以及大量新收集、重构、策划和构建的指令数据，同时政策生成的重放提供了来自参考模型自身行为分布的保留信号。最终的集合在规模和目的上都显著大于现有的哈萨尼亚文本资源。在评估中，Enemray在比较的开放和专有模型中实现了最强的英语到哈萨尼亚语翻译，并在毛里塔尼亚翻译错误检测中获得了最高的整体得分，同时在数学推理、知识、代码生成和函数调用等方面保留了其指令调优基础模型的大部分通用能力。本报告描述了Enemray的动机、数据构建、模型设计、训练方法和评估。

View on arXiv Download PDF AI Translation

cs.CL / 79 / 2609.14857

ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement

ModularRSI：模块化和可推广的递归自我改进

Wu, Siwei, Ren, Jincheng, Li, Yizhi, Li, Haau-Sing, Yang, Chengran, Zhang, Yuxuan, Gu, Weicheng, Yang, Jian, Batista-Navarro, Riza, Zhang, Chuanyi, Liu, Xianglong, Zhou, Ming, Dai, Bryan, Lin, Chenghua

Abstract

Recent work extends recursive self-improvement (RSI) to agent harnesses for long-horizon coding and terminal tasks, enabling agents to improve execution mechanisms from experience. However, generalizable harness RSI remains challenging. First, evolving harnesses on evaluation benchmarks or their subsets makes it difficult to distinguish reusable improvements from benchmark-specific adaptation. Second, single-trajectory updates can conflate systematic harness deficiencies with instance-specific reasoning and solution details, producing modifications that transfer poorly to unseen tasks. Third, localizing recurring behavioral deficiencies within monolithic harnesses is difficult, while whole-harness optimization can entangle unrelated mechanisms and complicate attribution and validation. We propose ModularRSI, a benchmark-disjoint, contrastive, and modular framework for generalizable harness evolution. ModularRSI contrasts successful and failed trajectories for the same task and aggregates evidence across tasks to identify recurring behavioral deficiencies. It decomposes the evolvable harness into five functional modules: Agent Loop, Tool Use, Observation Management, Context Management, and Task Completion Detection. Each module evolves independently within a restricted modification scope, followed by an integration stage that combines the evolved modules into a unified harness and resolves potential conflicts. To support benchmark-disjoint evolution, we curate 2,000 executable evolution tasks from external sources that are disjoint from downstream evaluation benchmarks. Experiments on TB2.0 and SWE-Bench Verified show consistent improvements on unseen in-domain and cross-domain tasks, with the evolved harness also transferring across different foundation models.

Chinese Translation

近期的研究将递归自我改进（RSI）扩展到代理工具的长时间编码和终端任务，使代理能够从经验中改进执行机制。然而，可推广的工具RSI仍然面临挑战。首先，在评估基准或其子集上演变工具使得难以区分可重用的改进与基准特定的适应。其次，单轨迹更新可能将系统性工具缺陷与特定实例的推理和解决细节混淆，导致对未见任务的迁移效果不佳。第三，在单一工具中定位重复的行为缺陷是困难的，而整体工具优化可能会纠缠不相关的机制，复杂化归因和验证。我们提出了ModularRSI，一个基准无关、对比性和模块化的可推广工具演变框架。ModularRSI对同一任务的成功和失败轨迹进行对比，并在任务之间汇总证据以识别重复的行为缺陷。它将可演变的工具分解为五个功能模块：代理循环（Agent Loop）、工具使用（Tool Use）、观察管理（Observation Management）、上下文管理（Context Management）和任务完成检测（Task Completion Detection）。每个模块在限制的修改范围内独立演变，随后进行整合阶段，将演变后的模块组合成一个统一的工具并解决潜在冲突。为了支持基准无关的演变，我们从外部来源策划了2000个可执行的演变任务，这些任务与下游评估基准无关。在TB2.0和SWE-Bench Verified上的实验显示，在未见的领域内和跨领域任务上均有一致的改进，演变后的工具在不同基础模型之间也能有效迁移。

View on arXiv Download PDF AI Translation

cs.CL / 80 / 2609.14858

Dream-RSI: Recursive Self-Improvement through Evolving Worlds

Dream-RSI：通过演化世界实现递归自我改进

Zheng, Tong, Wu, Xidong, Zhang, Zheng, He, Zhankui, Zhang, Chaoyi, Coleman, Benjamin, Wei, Ruoqiao, Bai, Di, Liu, Haolin, Liu, Rui, Wang, Xue, Zhuan, Yue, Kang, Wang-Cheng, Xiang, Renkai, Huang, Heng, Cheng, Xinwu, Guo, Yunsong

Abstract

Recursive self-improvement is becoming increasingly vital for autonomous AI agents, where progress hinges on discovering high-value solutions across complex domains. The driver of this process is effective exploration, however, managing and improving exploration strategies remains a major bottleneck. Current systems face a fundamental dilemma: fixed strategies fail to adapt as search spaces scale, while online policy optimization requires navigating vast meta-search spaces under delayed and expensive feedback over long-horizon rollouts. We introduce \textsc{Dream-RSI}, a framework for scalable and recursively self-improving exploration. A lightweight orchestration layer makes exploration explicit and programmable while leaving the underlying coding agent unchanged. Our key insight is that accumulated discovery history can serve as a replay simulator over the realized search space. By performing dreaming in the replay simulator constructed from historical discovery trees, \textsc{Dream-RSI} secures immediate, low-cost off-policy feedback to evaluate and refine exploration policies without invoking repetitive, expensive online evaluations. The improved policy is subsequently redeployed online to drive further discovery, continuously expanding the simulator pool in a self-improving loop. Across algorithm engineering, mathematical optimization, and GPU kernel engineering, \textsc{Dream-RSI} achieves competitive or improved discovery quality while substantially reducing discovery cost in several settings.

Chinese Translation

递归自我改进对于自主人工智能代理变得越来越重要，其进展依赖于在复杂领域中发现高价值解决方案。这个过程的驱动力是有效的探索，然而，管理和改进探索策略仍然是一个主要瓶颈。目前的系统面临一个基本的困境：固定策略在搜索空间扩大时无法适应，而在线策略优化则需要在延迟和高成本的反馈下，导航广阔的元搜索空间，进行长时间的回滚。我们提出了 extsc{Dream-RSI}，一个可扩展的递归自我改进探索框架。一个轻量级的协调层使探索变得明确和可编程，同时保持底层编码代理不变。我们的关键见解是，累积的发现历史可以作为已实现搜索空间的重放模拟器。通过在由历史发现树构建的重放模拟器中进行“梦境”操作， extsc{Dream-RSI} 确保了即时、低成本的离线反馈，以评估和优化探索策略，而无需进行重复的、高成本的在线评估。改进后的策略随后在线重新部署，以推动进一步的发现，持续扩展模拟器池，形成自我改进的循环。在算法工程、数学优化和GPU内核工程等多个领域， extsc{Dream-RSI} 实现了具有竞争力或改进的发现质量，同时在多个设置中显著降低了发现成本。

View on arXiv Download PDF AI Translation

cs.CL / 81 / 2609.14860

One Example Is Enough to Pass Fairness Benchmarks: Rethinking Fairness Evaluation for Aligned LLMs

一个例子足以通过公平性基准：重新思考对齐大型语言模型的公平性评估

Deng, Naihao, Arif, Samee, Chang, Shuaichen, Chen, Yulong, Mihalcea, Rada

Abstract

Warning: This submission studies stereotypes and biases, and contains toxic and offensive examples, used for illustration purposes only. Fairness benchmarks such as BBQ have become the de facto standard for fairness evaluation across major model families. We argue that these benchmarks are too easy to support their role: training Qwen 2.5 7B Base with Group Relative Policy Optimization (GRPO) on a single BBQ example, or placing that example in context as a one-shot demonstration for in-context learning (ICL), lifts mean BBQ accuracy from 79.9% to 92.9% and 99.0%, respectively, closing 80% of the gap to its large-scale RLHF counterpart (96.1%) with GRPO, and surpassing it with ICL. These effects generalize across model families. A cross-conditioning analysis shows the improvement is carried by the reasoning traces generated by the model, and one example suffices to elicit a category-agnostic ``missing evidence'' reasoning pattern. We argue that BBQ-style multiple-choice abstention benchmarks measure a single structural cue, and a model that solves them does not thereby become fair. We call for evaluation suites that cover a broader spectrum of fairness alignment.

Chinese Translation

警告：本提交研究了刻板印象和偏见，并包含有毒和冒犯性的示例，仅用于说明目的。公平性基准如 BBQ 已成为主要模型家族公平性评估的事实标准。我们认为，这些基准过于简单，无法支持其角色：使用组相对策略优化（Group Relative Policy Optimization, GRPO）在单个 BBQ 示例上训练 Qwen 2.5 7B Base，或者将该示例作为上下文中的一次性演示用于上下文学习（In-Context Learning, ICL），分别将平均 BBQ 准确率从 79.9% 提升至 92.9% 和 99.0%，缩小了与其大规模强化学习人类反馈（Reinforcement Learning from Human Feedback, RLHF）对应模型（96.1%）之间的 80% 差距，并通过 ICL 超越了它。这些效果在不同模型家族中具有普遍性。交叉条件分析表明，改进是由模型生成的推理痕迹所驱动，一个例子足以引发一种类别无关的“缺失证据”推理模式。我们认为，BBQ 风格的多项选择弃权基准仅测量单一结构线索，而解决这些基准的模型并不因此变得公平。我们呼吁建立覆盖更广泛公平性对齐谱系的评估套件。

View on arXiv Download PDF AI Translation

cs.CL / 82 / 2609.14896

Forty Shades of Blue: Quality-Diversity Alignment via Mode-Conditioned Reinforcement Learning

四十种蓝色：通过模式条件强化学习实现质量-多样性对齐

Yuan, Jiayi, Kang, Hangoo, Liu, James Jihao, Choi, Yejin, Iyer, Vikram, Jiang, Liwei, Jaques, Natasha

Abstract

A notable byproduct of LLM alignment training is mode collapse: the progressive loss of output diversity that narrows a model's expressivity at inference time. This degradation is especially limiting for applications requiring open-ended exploration and pluralistic perspectives, such as scientific ideation and creative writing. We present MoDA (Mode-conditioned Diversity Alignment), an online post-training RL algorithm that jointly optimizes generation quality and diversity, inspired by the coordination perspective in multi-agent reinforcement learning (MARL). MoDA trains a single shared LLM policy conditioned on abstract numbered roles, where each role acts as an agent competing to produce outputs distinct from the others. This formulation encourages mode-conditioned agents to explore complementary regions of the high-quality output space without requiring hand-crafted personas or architectural modifications. MoDA employs a prompt-adaptive quality gating mechanism that calibrates a reference quality threshold and grants diversity rewards only to responses that meet the threshold, preventing reward-hacking behaviors that compromise response quality. To study quality-diversity tradeoffs, we evaluate MoDA on a comprehensive suite of benchmarks spanning seven general capability tasks and four domain-specific diversity tasks in scientific ideation and creative writing. MoDA improves SBERT diversity by 265% on the Infinite-Chat held-out prompts, while increasing average general capability pass@1 by 10.3% over the Qwen3-8B baseline. Compared with the strongest DivPO baseline, MoDA improves SBERT diversity from 0.274 to 0.482 (+75.9%) and E-Vendi from 2.86 to 4.4 (+53.8%), while improving average general capability pass@1 by 7.0%. Overall, MoDA provides a drop-in alternative to standard post-training methods that preserves and expands the model's expressive output space while improving quality.

Chinese Translation

大型语言模型（LLM）对齐训练的一个显著副产品是模式崩溃：输出多样性的逐渐丧失，限制了模型在推理时的表现力。这种退化在需要开放式探索和多元视角的应用中尤为限制，例如科学构思和创意写作。我们提出了MoDA（模式条件多样性对齐），这是一种在线后训练强化学习（RL）算法，旨在共同优化生成质量和多样性，灵感来源于多智能体强化学习（MARL）中的协调视角。MoDA训练一个基于抽象编号角色的共享LLM策略，其中每个角色作为一个代理，竞争产生与其他角色不同的输出。这种表述鼓励模式条件代理探索高质量输出空间的互补区域，而无需手工设计角色或架构修改。MoDA采用了一种提示自适应质量门控机制，校准参考质量阈值，并仅对满足阈值的响应授予多样性奖励，从而防止奖励操控行为损害响应质量。为了研究质量与多样性的权衡，我们在涵盖七个通用能力任务和四个领域特定多样性任务的全面基准测试套件上评估了MoDA。在Infinite-Chat保留提示上，MoDA将SBERT多样性提高了265%，同时在Qwen3-8B基线上的平均通用能力pass@1提高了10.3%。与最强的DivPO基线相比，MoDA将SBERT多样性从0.274提高到0.482（+75.9%），E-Vendi从2.86提高到4.4（+53.8%），同时将平均通用能力pass@1提高了7.0%。总体而言，MoDA提供了一种可替代的标准后训练方法，既保留又扩展了模型的表现输出空间，同时提高了质量。

View on arXiv Download PDF AI Translation

cs.CL / 83 / 2609.14963

Can We Triage LLM Translation Errors in Classical Texts Without Human References? Source Novelty, GEMBA Scoring, and Budgeted Review through Pali-to-English Translation

在没有人类参考的情况下，我们能否对经典文本中的大型语言模型翻译错误进行分级？源新颖性、GEMBA评分和预算审查通过巴利语到英语的翻译

Metzger, Máté

Abstract

As large language models become capable translators of classical texts, a key challenge is deciding which outputs need expert review when no human reference exists. This study tests reference-free error triage through Pali-to-English translation. Three LLMs translated 15,493 passages. Five signals were compared: source novelty, source-candidate embedding distance, peer-translation disagreement, English-to-Pali backtranslation, and no-reference GEMBA scoring. Signals were calibrated on a 3,000-item reference-informed LLM-adjudicated sample and checked against a 500-item author-adjudicated anchor. Human references supported calibration and validation only; they were never used to compute the risk signals. Source novelty was a useful source-side risk prior but not a per-candidate error detector. Peer disagreement and backtranslation provided secondary signal. The strongest method was no-reference GEMBA scoring by a panel of models generally regarded as stronger than the translators: reviewing the top 10% by GEMBA risk captured 81.6% of panel-major errors in the calibration set. GEMBA also remained the best reference-free signal against the author anchor. A same-tier panel, with self-scoring excluded, remained useful but performed worse, indicating that evaluator strength matters beyond the prompt alone. A budgeted workflow is proposed, combining source novelty, peer disagreement, and stronger candidate-aware judging to allocate human review. Transfer to other classical languages, including Latin, Ancient Greek, and Sanskrit, remains to be tested.

Chinese Translation

随着大型语言模型成为经典文本的有效翻译者，一个关键挑战是决定在没有人类参考的情况下，哪些输出需要专家审查。本研究通过巴利语到英语的翻译测试了无参考的错误分级。三种大型语言模型翻译了15,493段文本。比较了五个信号：源新颖性、源-候选嵌入距离、同行翻译不一致、英语到巴利语的回译以及无参考的GEMBA评分。信号在一个包含3,000项参考信息的LLM裁定样本上进行了校准，并与一个500项作者裁定的基准进行了检查。人类参考仅用于支持校准和验证；从未用于计算风险信号。源新颖性是一个有用的源侧风险先验，但不是每个候选的错误检测器。同行不一致和回译提供了次要信号。最强的方法是由普遍被认为比翻译者更强的模型小组进行的无参考GEMBA评分：通过GEMBA风险审查前10%的内容捕获了校准集中的81.6%的主要错误。GEMBA在与作者基准的比较中仍然是最佳的无参考信号。一个同级小组，在自评分被排除的情况下，仍然有用但表现较差，表明评估者的实力在提示之外也很重要。提出了一种预算工作流程，结合源新颖性、同行不一致和更强的候选意识判断来分配人类审查。对其他经典语言的转移，包括拉丁语、古希腊语和梵语，仍需测试。

View on arXiv Download PDF AI Translation

cs.CL / 84 / 2609.14967

A Corpus-Aligned Uthmani-to-Standard Quranic Word Mapping and a Deterministic Recitation Validator

一个语料库对齐的乌斯曼体到标准古兰经单词映射及确定性朗读验证器

Elnawasany, Yahya Mohamed

Abstract

Quranic text is distributed in two orthographic forms that are byte-level distinct: the Uthmani script used in every printed mushaf, and the Standard (Imla'i) Arabic form that every mainstream Arabic NLP tool is built for. The gap is concentrated in one Unicode character, U+0670 (superscript alef), which appears in some of the most frequently recited words in the Quran and is silently mishandled by general-purpose Arabic normalizers. We release a 2,290-pair, corpus-aligned Uthmani-to-Standard word mapping constructed by aligning the complete 6,236-verse Quran across both orthographic forms, together with a seven-step text normalization pipeline built on it. Normalizing both forms of all 6,236 verses through that pipeline yields identical strings for 90.9% of verses, and we characterize the residual divergence rather than assert that it is closed. On top of the normalized text, we build a deterministic, LLM-free Quranic recitation validator using a four-layer verse-matching search (exact, morphological, relaxed, fuzzy) and word-error-rate-graded feedback across five severity tiers. The validator scores 98.4% (122/124) on a 124-case suite emitted by the released test harness, and both failures share one mechanism: a single substitution error can make a different verse an exact match. A full-corpus census additionally quantifies an inherent text-only ambiguity affecting 16.5% of verses, and on 34 recitation transcripts drawn from a deployed Arabic ASR system the validator identifies the correct verse in every case. We release the mapping, the script that builds it, the validator, and the evaluation harness under open licenses; every number in this paper except the deployment measurement, whose transcripts are not ours to publish, is reproduced by running them.

Chinese Translation

古兰经文本以两种正字法形式分布，这两种形式在字节级别上是不同的：乌斯曼体在每本印刷的穆斯哈夫中使用，而标准（伊姆莱）阿拉伯语形式是每个主流阿拉伯语自然语言处理工具所构建的基础。这一差距集中在一个Unicode字符U+0670（上标阿拉伯字母），该字符出现在古兰经中一些最常被朗读的单词中，并且被通用的阿拉伯语标准化工具默默处理不当。我们发布了一个包含2,290对的语料库对齐的乌斯曼体到标准单词映射，该映射通过对齐完整的6,236节古兰经在两种正字法形式之间构建而成，并基于此构建了一个七步文本标准化流程。通过该流程对所有6,236节的两种形式进行标准化，90.9%的节产生了相同的字符串，我们对剩余的差异进行了特征描述，而不是声称其已被消除。在标准化文本的基础上，我们构建了一个确定性的、无大型语言模型的古兰经朗读验证器，该验证器使用四层节匹配搜索（精确、形态、放宽、模糊）和基于五个严重程度等级的字错误率反馈。该验证器在发布的测试工具中对124个案例的评分为98.4%（122/124），而两个失败案例共享一个机制：单一的替换错误可以使不同的节成为精确匹配。全面的语料库普查还量化了影响16.5%节的固有文本模糊性，并且在从已部署的阿拉伯语自动语音识别系统中提取的34个朗读转录中，验证器在每个案例中都能识别出正确的节。我们以开放许可发布了映射、构建该映射的脚本、验证器和评估工具；本文中除了部署测量的数字（其转录不是我们可以发布的）外，所有数字均通过运行它们得出。

View on arXiv Download PDF AI Translation

cs.CL / 85 / 2609.14969

Online Language Adaptive Sampling for Better Distributed Cross-lingual Gains

在线语言自适应采样以实现更好的分布式跨语言收益

Nguyen, Quang Phuoc, Gaschi, Félix, Anugraha, David, Novoa, Santiago Martínez, Lee, En-Shiun Annie

Abstract

Realignment is a promising approach for improving the cross-lingual transfer ability of multilingual language models, particularly for extremely low-resource languages (LRLs). However, existing realignment methods rely on uniform and random sampling of parallel sentences across languages, which may be suboptimal under limited batch sizes. In practice, models may benefit from seeing certain languages more frequently, especially those that are poorly aligned, and the optimal distribution can evolve throughout training. In this work, we propose a simple yet effective adaptive sampling strategy that assigns trainable sampling probabilities to each language. Languages that contribute more to the realignment loss are sampled more frequently in subsequent batches, and the optimal distribution can evolve throughout training. Our method employs an inner-outer optimization loop with a small overhead, leading to consistent performance improvements and, more importantly, distributing the gains across languages. We observed a $+0.67$ average performance increase on all tasks with XLM-R, and $+0.60$ with Gemma 2 9B compared with uniform realignment. Furthermore, our method is robust across different models. Code available at https://github.com/felixgaschi/multilingual-alignment-and-transfer.

Chinese Translation

重新对齐是一种有前景的方法，用于提高多语言模型的跨语言迁移能力，特别是对于极低资源语言（LRLs）。然而，现有的重新对齐方法依赖于对跨语言平行句子的均匀和随机采样，这在有限的批量大小下可能并不理想。在实践中，模型可能会从更频繁地看到某些语言中受益，特别是那些对齐较差的语言，而最佳分布在训练过程中可能会不断演变。在本研究中，我们提出了一种简单而有效的自适应采样策略，为每种语言分配可训练的采样概率。在后续批次中，对重新对齐损失贡献较大的语言会被更频繁地采样，最佳分布在训练过程中可以不断演变。我们的方法采用内外优化循环，开销较小，导致一致的性能提升，更重要的是，在语言之间分配收益。我们观察到，与均匀重新对齐相比，使用XLM-R在所有任务上平均性能提高了0.67，使用Gemma 2 9B提高了0.60。此外，我们的方法在不同模型中表现出稳健性。代码可在https://github.com/felixgaschi/multilingual-alignment-and-transfer获取。

View on arXiv Download PDF AI Translation

cs.CL / 86 / 2609.14988

Biomedical Reference Generation Remains Unreliable across 26 Large Language Models

生物医学参考文献生成在26个大型语言模型中仍然不可靠

Topaz, Maxim, Zhang, Zhihong, Roguin, Nir, Gupta, Pallavi, Li, Zichao, Peltonen, Laura-Maria

Abstract

Background. Large language models are increasingly used to help write biomedical text but may fabricate references to nonexistent work. How often large language models do so is not well characterized. Methods. We prompted 26 language models from eight developers (2023 to 2026) to supply a missing reference for each of 69 biomedical passages across ten domains. References were classified as verifiable (real paper with a resolving identifier), partial matches (real paper without a resolving identifier), fabricated (no matching indexed paper), or declined (the model refused to supply a reference). A reference was considered correct in every evaluated bibliographic field only when it was verifiable and its journal, year, and listed authors matched those of the cited paper. Results. Fabrication ranged from 10.2% (Claude Opus 4.8, which declined 52.1% of prompts) to 98.4% (Ministral 3B, which produced no verifiable reference). Claude Opus 4.6 and Claude Sonnet 4.5 produced similar proportions of verifiable references (77.6% and 76.6%) but named authors correctly in 78.7% and 28.7% of author-evaluable verifiable references, respectively, and were correct in every evaluated field in 54.6% and 19.9% of responses. GPT-5.5 was correct in every field in 48.1%. Across all models, 55.4% of responses were fabricated and 14.9% were correct in every field. Among the five tested models first released in 2026, the corresponding proportions were 35.3% and 31.8%, respectively. Conclusions. Fabrication remained common, and no model was correct in every evaluated bibliographic field in more than 54.6% of responses. Models that identify real papers may still misstate their metadata, so references produced with model assistance require verification before use.

Chinese Translation

背景：大型语言模型越来越多地被用于帮助撰写生物医学文本，但可能会虚构不存在的参考文献。大型语言模型虚构参考文献的频率尚未得到很好的表征。方法：我们对来自八个开发者的26个语言模型（2023至2026年）进行了提示，要求其为69个跨十个领域的生物医学段落提供缺失的参考文献。参考文献被分类为可验证（具有解析标识符的真实论文）、部分匹配（没有解析标识符的真实论文）、虚构（没有匹配的索引论文）或拒绝（模型拒绝提供参考文献）。仅当参考文献在每个评估的书目字段中均为可验证，并且其期刊、年份和列出的作者与被引用论文相匹配时，才被视为正确。结果：虚构的比例从10.2%（Claude Opus 4.8，拒绝52.1%的提示）到98.4%（Ministral 3B，未生成可验证的参考文献）。Claude Opus 4.6和Claude Sonnet 4.5生成的可验证参考文献的比例相似（分别为77.6%和76.6%），但在可评估的可验证参考文献中，正确命名作者的比例分别为78.7%和28.7%，在每个评估字段中正确的比例分别为54.6%和19.9%。GPT-5.5在每个字段中的正确率为48.1%。在所有模型中，55.4%的响应为虚构，14.9%的响应在每个字段中均为正确。在五个于2026年首次发布的测试模型中，相应的比例为35.3%和31.8%。结论：虚构仍然普遍存在，且没有模型在每个评估的书目字段中超过54.6%的响应是正确的。能够识别真实论文的模型仍可能错误陈述其元数据，因此使用模型辅助生成的参考文献时需要进行验证。

View on arXiv Download PDF AI Translation

cs.CL / 87 / 2609.14991

Typhoon ASR Streaming: Steerable Low-Latency Thai Speech Recognition with Real-Time Shallow Fusion

台风 ASR 流媒体：可调低延迟的泰语语音识别与实时浅层融合

Sirichotedumrong, Warit, Samutsin, Tanawin, Wani, Shah Faisal, Sripaisarnmongkol, Sittipong, Pipatanakul, Kunat

Abstract

Open Thai automatic speech recognition (ASR) is dominated by offline, Whisper-based models that read the whole utterance before transcribing, ruling out low-latency uses such as live captioning and voice agents. We present a deployable system for streaming Thai ASR that lets a user steer its vocabulary at decode time, without retraining. A widely used open Thai model, trained with full context, collapses when run as a true stream; we restore streaming with a cache-aware encoder, by converting it or adapting a natively streaming one, and add a shallow-fusion layer that re-ranks candidates inside the streaming decoder with a GPU n-gram language model and phrase boosting. Across two Thai benchmarks and two model sizes, the streaming models stay usable where the full-context model fails, cutting character error rate 4.3-4.5x at a one-second look-ahead while running faster than real time. Decode-time steering then lifts keyword recall from 16.6% to 20.7% at no accuracy cost and negligible overhead; most of the gain comes from an n-gram over ordinary training transcripts, which resolves the written form of code-switched words the model hears but spells inconsistently, with phrase boosting adding targeted control over rare domain terms.

Chinese Translation

开放的泰语自动语音识别（ASR）主要由离线的基于 Whisper 的模型主导，这些模型在转录之前需要读取整个发音，排除了实时字幕和语音助手等低延迟应用。我们提出了一种可部署的泰语 ASR 流媒体系统，允许用户在解码时调整其词汇，而无需重新训练。一个广泛使用的开放泰语模型，在全上下文下训练，但在作为真实流媒体运行时会崩溃；我们通过使用一个缓存感知编码器恢复流媒体功能，或者将其转换为一个本地流媒体模型，并添加一个浅层融合层，该层利用 GPU n-gram 语言模型和短语增强在流媒体解码器内部对候选项进行重新排序。在两个泰语基准测试和两种模型规模下，流媒体模型在全上下文模型失败的情况下仍然可用，在一秒的前瞻下将字符错误率降低了 4.3-4.5 倍，同时运行速度超过实时。解码时的调整将关键词召回率从 16.6% 提升至 20.7%，且没有准确性损失和可忽略的开销；大部分增益来自于对普通训练转录的 n-gram，这解决了模型听到但拼写不一致的代码切换词的书面形式，而短语增强则提供了对稀有领域术语的有针对性的控制。

View on arXiv Download PDF AI Translation

cs.CL / 88 / 2609.14992

MTAC-IFBench: Benchmarking Instruction-Following in Multi-Turn Agentic Coding

MTAC-IFBench：多轮自主编码中的指令遵循基准测试

Wen, Bosi, Wang, Cunxiang, Gui, Jiayi, Zhang, Haoke, Niu, Yilin, Ke, Pei, Yang, Dayong, Wang, Hongning, Huang, Minlie

Abstract

Recently, the rapid development of large language models (LLMs) has reshaped software engineering by enabling autonomous code agents that plan, execute, and utilize external tools iteratively to tackle complex tasks. Beyond achieving functional correctness, these agents must faithfully follow process instructions and constraints throughout the development lifecycle. However, existing benchmarks typically focus on final functional correctness or confine instruction-following evaluation to single-turn, general chat or simple code generation scenarios, leaving instruction-following in multi-turn agentic coding underexplored. To bridge this gap, we propose MTAC-IFBench, a comprehensive benchmark for this critical capability. It features multi-turn progressive software development instructions with diverse constraints spanning 6 primary and 18 secondary categories. With an average of 7.04 turns and 91.33 constraints per instance, it poses a rigorous challenge to current LLMs. To make the evaluation reliable, we construct a checklist for each constraint and functional requirement, and integrate verification scripts and judge agents to verify each checklist item. MTAC-IFBench identifies significant deficiencies in existing code agents in multi-turn instruction-following, with their performance degrading rapidly as the interaction session grows longer.

Chinese Translation

近年来，大型语言模型（LLMs）的快速发展通过使自主代码代理能够规划、执行并迭代利用外部工具来解决复杂任务，重塑了软件工程。除了实现功能正确性之外，这些代理还必须在整个开发生命周期中忠实地遵循过程指令和约束。然而，现有的基准测试通常关注最终的功能正确性，或将指令遵循的评估限制在单轮、一般对话或简单代码生成场景中，从而使多轮自主编码中的指令遵循未得到充分探索。为填补这一空白，我们提出了MTAC-IFBench，这是一个针对这一关键能力的综合基准测试。它包含多轮渐进的软件开发指令，具有跨越6个主要类别和18个次要类别的多样化约束。每个实例平均包含7.04轮和91.33个约束，对当前的LLMs提出了严格的挑战。为了确保评估的可靠性，我们为每个约束和功能要求构建了检查清单，并整合验证脚本和评判代理来验证每个检查清单项。MTAC-IFBench揭示了现有代码代理在多轮指令遵循方面的显著不足，随着交互会话的延长，其性能迅速下降。

View on arXiv Download PDF AI Translation

cs.CL / 89 / 2609.14995

Beyond Depth and Width: The Information-Slack Dilemma in Streaming Test-Time Compute

超越深度与宽度：流媒体测试时计算中的信息-松弛困境

Zhang, Xiaotian

Abstract

The same task and compute budget can require different reasoning policies when evidence arrives in a different order. Early computation has more time to finish but rests on incomplete or revisable evidence; waiting improves information while shrinking computational slack. We call this the information-slack dilemma. We take the evidence-dependent computational job as the unit of analysis: when to start it, what supports its result, and when that result can be committed. Advance computation is valuable only insofar as its benefits survive the costs of verification, invalidation, and recovery. This applies to grounded incremental processing and reusable preparation as well as future-dependent speculation. We propose a research agenda on computation under evolving evidence, prioritizing selective recovery under controlled evidence revisions. Evaluation should separate earlier-execution effects, deployment value against a full-input alternative, and the added value of predictive policies, while accounting for shared-resource costs. The objective is not maximal advance computation, but more trustworthy, on-time responses within a declared resource envelope.

Chinese Translation

相同的任务和计算预算在证据以不同顺序到达时可能需要不同的推理策略。早期计算有更多时间完成，但依赖于不完整或可修订的证据；等待可以改善信息质量，但会缩小计算松弛。我们称之为信息-松弛困境。我们将依赖证据的计算任务作为分析单元：何时开始，什么支持其结果，以及何时可以确认该结果。提前计算的价值仅在于其收益能够抵消验证、失效和恢复的成本。这适用于基于事实的增量处理和可重用的准备工作，以及依赖未来的推测。我们提出了一个关于在不断变化的证据下进行计算的研究议程，优先考虑在控制证据修订下的选择性恢复。评估应区分早期执行效果、与完整输入替代方案的部署价值，以及预测策略的附加价值，同时考虑共享资源的成本。我们的目标不是最大化提前计算，而是在声明的资源范围内提供更可靠、及时的响应。

View on arXiv Download PDF AI Translation

cs.CL / 90 / 2609.15007

ABSOL: Aggregated Bayesian Subsampling Orchestrated with LLMs

ABSOL：与大型语言模型协同的聚合贝叶斯子采样

Hassell, Jackson, Shen, Chen, Hruschka, Estevam

Abstract

Large language models are increasingly used as natural-language interfaces to structured data, yet they remain unreliable when answers require consistent evidence conditioning, dependency-aware reasoning, and uncertainty estimation. Bayesian networks provide an explicit probabilistic reasoning layer, but learning useful structures from data remains costly and fragile at scale. We introduce ABSOL, a hybrid LLM-guided Bayesian network structure-learning framework that uses LLMs as bounded semantic guides. Across five discrete BN benchmarks spanning 27 to 1041 nodes, ABSOL is the only evaluated method to produce a viable graph on every benchmark, and achieves the highest Edge F_1 on every benchmark larger than 27 nodes with GPT-5.4. The four LLM augmentations, which contribute complementary semantic evidence to the statistical backbone, improve Edge F_1 over the non-LLM aggregation backbone by +0.23 on average. Complementary post-hoc refinement experiments suggest that these gains depend in part on limiting the LLM's authority over the final structure. Together, these results show that language-derived semantic knowledge can substantially improve scalable probabilistic structure learning when used as bounded guidance within a statistically grounded reasoning pipeline. The code for ABSOL is available at github.com/megagonlabs/absol-bn.

Chinese Translation

大型语言模型越来越多地被用作结构化数据的自然语言接口，但在需要一致证据条件、依赖感知推理和不确定性估计的回答中，它们仍然不可靠。贝叶斯网络提供了一个明确的概率推理层，但从数据中学习有用的结构在规模上仍然成本高昂且脆弱。我们提出了ABSOL，一个混合的LLM引导的贝叶斯网络结构学习框架，利用LLMs作为有限的语义引导。在五个离散的贝叶斯网络基准测试中，节点数量从27到1041，ABSOL是唯一在每个基准上都能生成可行图的方法，并且在每个节点数大于27的基准上，使用GPT-5.4时达到了最高的边缘F_1。四种LLM增强方法为统计基础提供了互补的语义证据，使得边缘F_1相比于非LLM聚合基础平均提高了0.23。互补的事后精炼实验表明，这些增益在一定程度上依赖于限制LLM对最终结构的权威性。综合这些结果表明，当作为有限指导在统计基础推理管道中使用时，语言衍生的语义知识可以显著改善可扩展的概率结构学习。ABSOL的代码可在github.com/megagonlabs/absol-bn获取。

View on arXiv Download PDF AI Translation

cs.CL / 91 / 2609.15022

SALUTE: Benchmarking and Adapting LLMs for the Defense Domain

SALUTE：国防领域大型语言模型的基准测试与适应

Park, Hyeongcheol, In, Sumin, Myeong, Suyeon, Park, Hogun, Kim, Sangmin, Lee, Moonhyun, Park, Daekyeong, Kim, Sangpil

Abstract

Defense is a knowledge-intensive domain that requires precise understanding of specialized terminology, doctrinal concepts, operational procedures, and evolving military events. Although recent work has explored language technologies for military applications, existing efforts remain fragmented: they are often task-specific, rely on limited adaptation pipelines, or lack comprehensive defense-domain evaluation. In this paper, we present SALUTE, an end-to-end framework for benchmarking and adapting LLMs for the defense domain. SALUTE integrates Salute-Corpus, a curated corpus from open-access U.S. military doctrine and government documents; Salute-Conv, a grounded instruction dataset from doctrinal sources and decade-long defense news; Salute-Pref, a defense-aware preference dataset; and Salute-Bench, a rigorously filtered benchmark for evaluating defense-domain understanding and reasoning over doctrine and defense news. Based on these resources, we train Salute-LLM through multi-stage post-training with continual pretraining, supervised fine-tuning, and preference alignment. Extensive experiments show that Salute-LLM achieves strong defense-domain performance while retaining competitive general capabilities, demonstrating the effectiveness of SALUTE as an end-to-end framework for defense-domain LLM adaptation.

Chinese Translation

国防是一个知识密集型领域，需要对专业术语、教义概念、操作程序和不断演变的军事事件有准确的理解。尽管近期的研究探索了军事应用的语言技术，但现有的工作仍然较为零散：它们往往是任务特定的，依赖于有限的适应流程，或缺乏全面的国防领域评估。在本文中，我们提出了SALUTE，一个用于国防领域大型语言模型基准测试与适应的端到端框架。SALUTE整合了Salute-Corpus，这是一个来自开放获取的美国军事教义和政府文件的精选语料库；Salute-Conv，这是一个来自教义来源和十年防务新闻的基础指令数据集；Salute-Pref，这是一个关注国防的偏好数据集；以及Salute-Bench，这是一个经过严格筛选的基准，用于评估对教义和防务新闻的理解和推理。基于这些资源，我们通过多阶段后训练，结合持续预训练、监督微调和偏好对齐，训练了Salute-LLM。大量实验表明，Salute-LLM在国防领域表现出色，同时保留了竞争力的通用能力，展示了SALUTE作为国防领域大型语言模型适应的端到端框架的有效性。

View on arXiv Download PDF AI Translation

cs.CL / 92 / 2609.15045

Mirror, Mirror on the Wall: Prompt Echoing in Small Instruct Language Models

镜子，镜子在墙上：小型指令语言模型中的提示回声

Okulska, Inez, Naskręcki, Bartosz, Piotrowski, Jan, Steifer, Tomasz

Abstract

Prompt echoing is a recognized failure mode of instruct language models, in which a model instead of generating a response, mirrors the provided prompt, even though it did not receive a specific instruction to do so. Is this phenomenon a sign of the model leaking the content of its training dataset, or is it rather caused by a misaligned behavior of the internal induction/copying mechanisms? We investigate prompt echoing small language models from different families (Gemma, Llama, Qwen, SmolLM and OLMo) and show that echoing prompts are likely to have partial overlap with the training dataset but the phenomenon is primarily driven by the model's induction heads.

Chinese Translation

提示回声是指令语言模型的一种已知失败模式，在这种模式下，模型并没有生成响应，而是镜像了提供的提示，即使它并没有收到特定的指令去这样做。这个现象是模型泄露其训练数据集内容的迹象，还是说它是由内部归纳/复制机制的不一致行为引起的？我们研究了来自不同家族的小型语言模型（Gemma、Llama、Qwen、SmolLM 和 OLMo）中的提示回声，并表明回声提示可能与训练数据集部分重叠，但这一现象主要是由模型的归纳头驱动的。

View on arXiv Download PDF AI Translation

cs.CL / 93 / 2609.15066

Salesforce Koa: An Enterprise Language Model for Agentic Tool Use

Salesforce Koa：用于代理工具使用的企业语言模型

Chen, Zixiang, Niu, Sufeng, Liu, Yingchi, Zhao, Wenting, Prabhakar, Akshara, Mehrotra, Shubham, Bi, Bin, Lan, Zhujun, Tan, Katherine, Ramezanali, Mohammad, Awalgaonkar, Tulika Manoj, Banerjee, Monojit, Qiu, Jielin, Pentyala, Shiva Kumar, Cen, Zhepeng, Tripathi, Anupam, Ziaei, Ali, Radhakrishnan, Regunathan, Shadravan, Darvish Lee, Heinecke, Shelby, Asur, Sitaram, Savarese, Silvio, Zhu, James, Mui, Phil, Wang, Huan

Abstract

We present Salesforce Koa, an enterprise language model built by post-training the open-weight Nemotron-3-Super-120B foundation model with reinforcement learning using Group Relative Policy Optimization (GRPO). Salesforce Koa is trained on public and synthetically generated data, with no customer data, to improve tool use and agentic capabilities while preserving strong general-purpose performance. Its distinctive component is a simulation-to-reward pipeline that expands workflow specifications into persona-conditioned multi-turn tasks with task-resolution rewards grounded in successful tool use for data-dependent requests. For enterprise domains, these specifications are written in Agent Script, Salesforce's declarative language for building Agentforce agents; for public tool-use domains, we synthesize the workflow structure directly. The same simulation and grounded-reward machinery drives GRPO across both. Across public tool-use, agentic-reasoning, and enterprise Customer Relationship Management (CRM) benchmarks, Salesforce Koa improves over its open-weight base, with the clearest gains on multi-turn tool use, and surpasses a strong proprietary baseline while remaining below the strongest frontier models. These results show that specification-driven reinforcement learning is a practical path to specializing open-weight foundation models for enterprise agentic tasks.

Chinese Translation

我们提出了Salesforce Koa，这是一个企业语言模型，通过使用群体相对策略优化（Group Relative Policy Optimization, GRPO）对开放权重的Nemotron-3-Super-120B基础模型进行后训练而构建。Salesforce Koa在公共和合成生成的数据上进行训练，不使用客户数据，以提高工具使用和代理能力，同时保持强大的通用性能。其独特的组成部分是一个模拟到奖励的管道，该管道将工作流规范扩展为基于角色的多轮任务，并为数据依赖请求提供基于成功工具使用的任务解决奖励。对于企业领域，这些规范用Salesforce的声明性语言Agent Script编写，以构建Agentforce代理；对于公共工具使用领域，我们直接合成工作流结构。相同的模拟和基于奖励的机制驱动GRPO在两者之间的应用。在公共工具使用、代理推理和企业客户关系管理（Customer Relationship Management, CRM）基准测试中，Salesforce Koa在其开放权重基础上有所提升，特别是在多轮工具使用方面取得了明显的进展，并超越了一个强大的专有基线，同时仍低于最强的前沿模型。这些结果表明，以规范驱动的强化学习是专门化开放权重基础模型以应对企业代理任务的可行路径。

View on arXiv Download PDF AI Translation

cs.CL / 94 / 2609.15070

DA-DLM: Explicitly Modeling Token Dependencies in Diffusion Language Models

DA-DLM：在扩散语言模型中显式建模令牌依赖性

Ji, Pengyu, Zhang, Zichen, Hu, Xiang, Tu, Kewei

Abstract

Diffusion Language Models (DLMs) generate text by iteratively denoising a masked sequence, independently predicting multiple tokens at each step. This conditional independence discards inter-token dependencies and degrades coherence-an issue that parallels the multi-modality problem in Non-Autoregressive Translation (NAT). Drawing on the Directed Acyclic Transformer (DAT), which tackles this problem in NAT via a Directed Acyclic Graph (DAG), we propose DA-DLM, a model that adapts DAG-based dependency modeling to DLMs' iterative setting through a position-oriented DAG design. The position-oriented DAG binds node groups to fixed output positions so that tokens fixed in earlier steps anchor neighboring predictions via learned transitions, and evolves with denoising to focus on remaining uncertainty as anchors accumulate. On language modeling, open-ended generation, and summarization, DA-DLM consistently outperforms Block Diffusion, especially under fewer denoising steps, and matches autoregressive models while preserving the parallel generation advantage. Our code is publicly available at https://github.com/jipy0222/DA-DLM.

Chinese Translation

扩散语言模型（DLM）通过迭代去噪一个被屏蔽的序列来生成文本，在每一步独立预测多个令牌。这种条件独立性忽略了令牌间的依赖关系，降低了连贯性——这一问题与非自回归翻译（NAT）中的多模态问题相似。基于有向无环变换器（DAT），该模型通过有向无环图（DAG）解决NAT中的这一问题，我们提出了DA-DLM，一种通过面向位置的DAG设计将基于DAG的依赖建模适应于DLM的迭代设置的模型。面向位置的DAG将节点组绑定到固定的输出位置，以便在早期步骤中固定的令牌通过学习的转移锚定邻近的预测，并随着去噪的进行，专注于剩余的不确定性，随着锚点的积累而演变。在语言建模、开放式生成和摘要生成任务中，DA-DLM始终优于块扩散，尤其是在去噪步骤较少的情况下，并且在保持并行生成优势的同时与自回归模型相匹配。我们的代码已公开发布在 https://github.com/jipy0222/DA-DLM。

View on arXiv Download PDF AI Translation

cs.CL / 95 / 2609.15079

Translating the Translator: Decomposing the Cost of English-Forced Inter-Agent Communication

翻译翻译者：分解英语强制的代理间通信成本

Agrawal, Kushagra, Feng, Yuming, Leung, Man-Fai

Abstract

Multi-agent LLM architectures, such as LangChain and AutoGen, largely assume English as the lingua franca for internal inter-agent communication, even when the end-user task is non-English. We fill this gap by evaluating a two-agent extraction-answer core, with an additional back-translation agent in the English-forced condition, across four typologically diverse languages (Hindi, Chinese, Spanish, Arabic; n = 300 per language) using the Aya-23-8B model. We compare a native-language pipeline to an English-forced one (which incorporates a final back-translation step from English to the user's language). We discover a statistically significant English-Forcing Tax (surviving a strict Bonferroni correction) that isolates the cost of English routing from general multi-agent orchestration overhead. Forcing inter-agent communication through English reduces Exact Match accuracy by 13.0 percentage points (Spanish) up to 30.6 percentage points (Hindi) compared to native-language multi-agent execution. Using chrF scores as a diagnostic measure of English-reference lexical overlap, we find that lower overlap is strongly associated with pipeline failure, consistent with translation loss being an important contributor to the observed performance drop. These findings suggest a compelling case for native-language routing in agent frameworks when the source and target languages are typologically distant, reducing a compounding translation tax.

Chinese Translation

多代理大语言模型（LLM）架构，如 LangChain 和 AutoGen，通常假设英语作为内部代理间通信的通用语，即使最终用户任务并非英语。我们通过评估一个包含两个代理的提取-回答核心，在英语强制条件下增加一个反向翻译代理，填补了这一空白，涉及四种类型多样的语言（印地语、中文、西班牙语、阿拉伯语；每种语言 n = 300），使用 Aya-23-8B 模型。我们将母语管道与英语强制管道进行比较（后者包含从英语到用户语言的最终反向翻译步骤）。我们发现存在统计显著的英语强制税（在严格的 Bonferroni 校正下依然成立），该税项将英语路由的成本与一般多代理协调开销分离开来。通过英语强制进行代理间通信使得准确匹配率降低了 13.0 个百分点（西班牙语）至 30.6 个百分点（印地语），与母语多代理执行相比。使用 chrF 分数作为英语参考词汇重叠的诊断指标，我们发现较低的重叠与管道失败有强烈关联，这与翻译损失是观察到的性能下降的重要因素一致。这些发现为在源语言和目标语言在类型上相距较远时，代理框架中采用母语路由提供了有力的论据，从而减少了累积的翻译税。

View on arXiv Download PDF AI Translation

cs.CL / 96 / 2609.15106

When the Wrong Key Wins: Understanding and Detecting Hallucinations in LLMs

当错误的关键获胜：理解和检测大语言模型中的幻觉

Tong, Xuhan, Zhang, Jiawei

Abstract

Large language models can hallucinate even when the knowledge required for a correct answer is already available. We study this failure through a latent-key view of inference, where answer selection depends on competition among associations acquired during pretraining. We show that model predictions can be highly sensitive to individual query keywords, that these influential keywords exhibit entity-specific binding, and that their effects are systematically shaped by pretraining frequency. Multiple bindings can also compete and exhibit higher-order interactions within the same query. Based on this mechanism, we introduce a two-stage keyword-perturbation method for hallucination detection. By removing influential keywords and measuring how the model reorganizes its prediction, the method distinguishes errors caused by misleading key associations from correct decisions supported by diagnostic evidence. Across multiple models and benchmarks, perturbation provides a strong and transferable detection signal, reaching $.910$ AUROC on probe-known ScientistQA. Finally, we extend the same probabilistic framework to four hallucination regimes: knowledge deficit, wrong knowledge, context distraction, and unstable inference. Their operational distributions across benchmarks provide diagnostic context for why different detector families succeed in different settings.

Chinese Translation

大型语言模型即使在正确答案所需的知识已经可用的情况下也可能产生幻觉。我们通过潜在关键视角研究这种失败，其中答案选择依赖于在预训练过程中获得的关联之间的竞争。我们表明，模型预测对单个查询关键词高度敏感，这些有影响力的关键词表现出特定实体的绑定，并且它们的影响受到预训练频率的系统性影响。多个绑定也可以在同一查询中竞争并表现出更高阶的交互。基于这一机制，我们提出了一种两阶段关键词扰动方法用于幻觉检测。通过去除有影响力的关键词并测量模型如何重新组织其预测，该方法区分了由误导性关键关联引起的错误与由诊断证据支持的正确决策。在多个模型和基准测试中，扰动提供了强大且可转移的检测信号，在已知的 ScientistQA 测试集中达到 $.910$ 的 AUROC。最后，我们将相同的概率框架扩展到四种幻觉状态：知识缺失、错误知识、上下文干扰和不稳定推理。它们在基准测试中的操作分布为不同检测器家族在不同环境中成功的原因提供了诊断背景。

View on arXiv Download PDF AI Translation

cs.CL / 97 / 2609.15126

MoME: Mixture-of-Memory Embeddings for Context-Aware Sparse Lookup

MoME：用于上下文感知稀疏查找的混合记忆嵌入

Li, Muchen, Sigal, Leonid, Liao, Renjie

Abstract

Scaling large language models efficiently has motivated sparse capacity mechanisms such as Mixture-of-Experts and, more recently, conditional memory: token-indexed embedding tables that augment the backbone with cheap parametric lookups. Existing memory-embedding methods retrieve via a deterministic function of the surface form, which collapses different contextual senses of the same token (e.g., python the language vs. the animal) into a single fixed entry. We introduce Mixture of Memory Embeddings (MoME), a context-aware memory mechanism that replaces each token's single memory row with a mixture of M slots and uses a learned gate over the hidden state to choose which slots to read at each position. In controlled pretraining experiments across nanochat, Llama-3/MobileLLM, and Qwen3 backbones, MoME improves over Value Embedding, Bigram, and STEM baselines in iso-parameter and iso-training-FLOP settings, shows a more promising memory-size scaling trend at sub-billion scale, and remains efficient in training and inference. Qualitative routing analyses on polysemous tokens further suggest that the learned mixture exhibits a degree of semantic interpretability, dispatching the same surface token to distinct memory slots under different senses.

Chinese Translation

高效扩展大型语言模型的需求促使了稀疏容量机制的发展，如专家混合（Mixture-of-Experts）以及最近的条件记忆（conditional memory）：基于令牌索引的嵌入表，通过廉价的参数查找增强了主干网络。现有的记忆嵌入方法通过表面形式的确定性函数进行检索，这将同一令牌的不同上下文意义（例如，python作为编程语言与作为动物）压缩为单一固定条目。我们提出了混合记忆嵌入（Mixture of Memory Embeddings，MoME），这是一种上下文感知的记忆机制，它用M个槽的混合替代每个令牌的单一记忆行，并在每个位置使用学习的门控机制选择读取哪些槽。在对nanochat、Llama-3/MobileLLM和Qwen3主干的控制预训练实验中，MoME在同参数和同训练FLOP设置下优于值嵌入（Value Embedding）、二元组（Bigram）和STEM基线，展示出在亚十亿规模下更有前景的记忆大小扩展趋势，并在训练和推理中保持高效。对多义令牌的定性路由分析进一步表明，学习到的混合展现出一定程度的语义可解释性，在不同意义下将相同的表面令牌分派到不同的记忆槽中。

View on arXiv Download PDF AI Translation

cs.CL / 98 / 2609.15145

Improving Mathematical Reasoning Capabilities in Large Language Models via Reasoning Process Error Classification

通过推理过程错误分类提升大型语言模型的数学推理能力

Yoshida, Runa, Nishida, Kosuke, Nishida, Kyosuke

Abstract

The reasoning ability of large language models (LLMs) is a critical factor for practical LLM-based applications. To investigate the current reasoning capability of LLMs, we clarify the types of errors that arise in LLMs' reasoning processes on mathematical datasets. We focus on problems where LLMs produce an incorrect answer. We define errors in the reasoning process as reasoning errors and manually analyze the features of reasoning errors. We defined and classified 21 error classes and identified the frequently occurring classes among them. Beyond qualitative evaluation, we leverage the evaluation results to improve the reasoning capability. We designed a prompt that explicitly focuses on eight error classes. The experiments demonstrate that this prompt effectively improves reasoning performance. Furthermore, the results suggest that the frequent reasoning errors identified in this paper are common across LLMs of comparable scale.

Chinese Translation

大型语言模型（LLMs）的推理能力是基于LLM的实际应用中的一个关键因素。为了探讨LLMs当前的推理能力，我们明确了在数学数据集上LLMs推理过程中出现的错误类型。我们重点关注LLMs产生错误答案的问题。我们将推理过程中的错误定义为推理错误，并手动分析推理错误的特征。我们定义并分类了21种错误类别，并识别出其中频繁出现的类别。除了定性评估外，我们利用评估结果来提升推理能力。我们设计了一个明确关注八种错误类别的提示。实验表明，该提示有效提高了推理性能。此外，结果表明，本文识别的频繁推理错误在规模相当的LLMs中是普遍存在的。

View on arXiv Download PDF AI Translation

cs.CL / 99 / 2609.15161

EMR: Self-Evolving Medical Multi-Agent System via Experience Mining and Reuse

EMR：通过经验挖掘与重用实现自我演化的医疗多智能体系统

Shi, Dongsheng, Li, Yue, Yi, Xin, Wang, Linlin

Abstract

Large language model (LLM) driven multi-agent systems have shown promise in complex clinical reasoning, yet existing approaches rely on static strategies and lack persistent clinical memory, preventing self-evolving from prior diagnostic successes and failures. We present EMR, a self-evolving medical multi-agent system via Experience Mining and Reuse. EMR introduces a hierarchical clinical experience library that organizes accumulated knowledge into three levels: clinical principles, diagnostic patterns, and representative cases. During inference, EMR emulates multidisciplinary consultation: a planner agent coordinates domain-specific department agents for specialized reasoning, while a summary agent synthesizes their analyses into a final decision. Critically, EMR automatically extracts correct diagnostic insights and failure-related warnings from multi-agent reasoning trajectories, incrementally updating the experience library to guide future cases. Experiments on medical reasoning benchmarks demonstrate that EMR consistently outperforms state-of-the-art medical multi-agent baselines. Further analysis reveals that the hierarchical experience enables cross-specialty generalization and transfer across diverse LLM backbones, offering a scalable and in

Chinese Translation

基于大型语言模型（LLM）的多智能体系统在复杂的临床推理中展现出良好的前景，但现有方法依赖于静态策略，缺乏持久的临床记忆，无法从以往的诊断成功与失败中自我演化。我们提出了EMR，一个通过经验挖掘与重用实现自我演化的医疗多智能体系统。EMR引入了一个分层的临床经验库，将积累的知识组织为三个层次：临床原则、诊断模式和代表性案例。在推理过程中，EMR模拟多学科会诊：一个规划者智能体协调特定领域的部门智能体进行专业推理，而一个总结智能体将它们的分析综合为最终决策。重要的是，EMR自动从多智能体推理轨迹中提取正确的诊断见解和与失败相关的警告，逐步更新经验库以指导未来的案例。在医疗推理基准测试中的实验表明，EMR始终优于最先进的医疗多智能体基线。进一步的分析揭示，分层经验使得跨专业的泛化和在不同LLM基础上进行转移成为可能，提供了可扩展性和...

View on arXiv Download PDF AI Translation

cs.CL / 100 / 2609.15164

CITECHOICE: A Causal Audit of How Document Presentation Redistributes Citation Credit in Agentic Search

CITECHOICE：文档呈现如何在主动搜索中重新分配引用信用的因果审计

Selvam, Sriram, Ghosh, Anneswa

Abstract

When several retrieved sources support the same claim, an answer engine cites some but not others. We call this decision citation allocation and introduce CITECHOICE, a causal audit of authentic multi-turn agentic search. From 129 everyday-query transcripts, CITECHOICE selects 113 same-call document pairs with independently verified support for the same pre-specified fact, without observing ranks or answer outcomes; blinded human review confirms 103. It runs a hash-verified 2-by-2 replay crossing pair order with jointly generated, fidelity-checked structured and prose renderings of one target while the rest of the transcript remains fixed. Three results emerge. First, and most importantly, structured rendering concentrates citation credit rather than clearly increasing source admission. It raises target citation count by +0.50 citations per answer (95 percent CI [+0.20, +0.84]; Holm-adjusted p=.033), without increasing total citations or reducing competitor credit. The pre-specified incidence effect (whether the target is cited at all) is +4.5 percentage points and inconclusive (95 percent CI [-1.4, +10.4]; p=.168). Second, observational position differences exceed controlled reordering effects: the citation-rate gap between rank 1 and rank 5 is 42.3 percentage points, compared with +7.9 percentage points in the main replay and 0.0 percentage points held out. Third, citation evaluation has a measurable noise floor. Although the aggregate count effect repeats under fresh decoding of 30 frozen families, 15 percent of binary decisions change and decoding accounts for an estimated 45 percent of single-generation family-effect variance. Together, these findings isolate what survives control: presentation can causally redistribute visible citation credit within frozen transcripts. They do not establish reliable source admission, a pure formatting mechanism, or a general rank advantage.

Chinese Translation

当多个检索到的来源支持相同的论点时，答案引擎会引用其中一些而不引用其他来源。我们称这种决策为引用分配，并引入CITECHOICE，这是对真实多轮主动搜索的因果审计。CITECHOICE从129个日常查询的记录中选择113对相同调用的文档，这些文档对同一预先指定的事实提供独立验证的支持，而不观察排名或答案结果；盲审人类评审确认了103对。它运行一个哈希验证的2x2重放，交叉配对顺序，同时生成、保真检查的结构化和散文渲染的一个目标，而其余的记录保持不变。结果有三个。首先，也是最重要的，结构化渲染集中引用信用，而不是明显增加来源的接受度。它使目标引用数量每个答案增加0.50次引用（95%置信区间[+0.20, +0.84]；Holm调整后的p=.033），而不增加总引用或减少竞争者信用。预先指定的发生效应（目标是否被引用）为+4.5个百分点且结果不确定（95%置信区间[-1.4, +10.4]；p=.168）。其次，观察位置差异超过了控制重排序的效果：排名1和排名5之间的引用率差距为42.3个百分点，而在主要重放中为+7.9个百分点，持出样本为0.0个百分点。第三，引用评估具有可测量的噪声底线。尽管在30个冻结的家族的新解码下，汇总计数效应重复出现，但15%的二元决策发生变化，解码估计占单一生成家族效应方差的45%。综合来看，这些发现隔离了在控制下存活的因素：呈现可以在冻结的记录中因果性地重新分配可见的引用信用。它们并未建立可靠的来源接受度、纯格式机制或一般排名优势。

View on arXiv Download PDF AI Translation

cs.CL / 101 / 2609.15188

MUSE: A Theory-Harnessed Story Engine for Vibe Narrativizing

MUSE：一个理论驱动的故事引擎用于氛围叙事

Ma, Jianxiang, Yang, Xiaocui, Wang, Daling, Hou, Yuesong, Zhang, Mingfu, Gao, Yichen, Huang, Junzhao

Abstract

LLMs can generate fluent prose. Story quality depends on how decisions about plot, character, and language work together across planning, drafting, and revision. Guiding these decisions presents two bottlenecks: the quality of story guidance and its sustained use. We formulate Vibe Narrativizing as the task of turning natural-language writing requirements into a finished story and present MUSE, a Theory-Harnessed Story Engine. MUSE organizes story knowledge as guidance for specific decisions and carries those decisions into subsequent creative work. Knowledge engineering develops Robert McKee's story theory through rule atomization, semantic consolidation, and mechanism abstraction; a single source of truth and layered disclosure organize the resulting guidance. Typical examples complement principles that depend on context and aesthetic judgment. An agent harness organizes design, character performance, scene composition, and revision through intermediate deliverables that preserve story decisions. Context engineering supplies each role with the relevant guidance and decisions, while a masterwork corpus provides inspiration and prose references. A worked example follows one requested object from its thematic role to the characters' climactic actions. Across four base models, MUSE improves WritingBench by 1.6-4.8 points over zero-shot generation and raises LongStoryEval by more than ten points on three. ConStory-Bench consistency error density remains in the low single digits for all four models, below every reproduced story-system baseline on three. Component ablations locate the largest quality contribution in structural design, voice-specific effects in the character path, and further gains in revision. Code is available at https://github.com/RoadtoAGI/MUSE.

Chinese Translation

大型语言模型（LLMs）能够生成流畅的散文。故事的质量取决于情节、角色和语言的决策如何在规划、草拟和修订过程中相互作用。指导这些决策面临两个瓶颈：故事指导的质量及其持续使用。我们将氛围叙事（Vibe Narrativizing）定义为将自然语言写作要求转化为完成故事的任务，并提出了MUSE，一个理论驱动的故事引擎。MUSE将故事知识组织为特定决策的指导，并将这些决策带入后续的创作工作。知识工程通过规则原子化、语义整合和机制抽象发展了罗伯特·麦基（Robert McKee）的故事理论；一个单一的真实来源和分层披露组织了所产生的指导。典型示例补充了依赖于上下文和审美判断的原则。一个代理组织通过中间交付物组织设计、角色表现、场景构图和修订，从而保持故事决策。上下文工程为每个角色提供相关的指导和决策，而一部杰作语料库则提供灵感和散文参考。一个具体示例展示了一个请求对象从其主题角色到角色的高潮行动的全过程。在四个基础模型中，MUSE在零样本生成上将WritingBench提高了1.6-4.8分，并在三个模型上将LongStoryEval提高了超过十分。ConStory-Bench的一致性错误密度在所有四个模型中保持在低个位数，低于三个重现的故事系统基线。组件消融研究发现，结构设计对质量的贡献最大，角色路径中的声音特定效果，以及在修订中的进一步提升。代码可在https://github.com/RoadtoAGI/MUSE获取。

View on arXiv Download PDF AI Translation

cs.CL / 102 / 2609.15191

What Limits Us? Analyzing Self-Reported Limitations in NLP Research

我们受到什么限制？分析自然语言处理研究中的自我报告限制

Thaepprasit, Tawan, Kehasukcharoen, Peeranuth, Wang, Ding, Denton, Remi, Vateekul, Peerapon, Lertvittayakumjorn, Piyawat

Abstract

Since late 2022, a Limitations section has become mandatory at many top-tier NLP conferences. The growing number of accepted papers at these venues has resulted in a vast corpus of self-reported limitations that cannot all be manually reviewed, yet remains systematically unanalyzed. Therefore, in this paper, we conduct a large-scale analysis of the Limitations sections from ACL and EMNLP papers published between 2020 and 2025 to understand what researchers disclose about their own work. To do so, we implement a novel human-AI framework for iterative hybrid qualitative coding. This framework enables us to investigate trends in self-reported limitations over time, their correlations with specific paper attributes, and the writing patterns that recur around these disclosures. Our findings offer a critical reflection on the diverse reported challenges as well as the self-reporting practices of researchers in the NLP community.

Chinese Translation

自2022年底以来，许多顶级自然语言处理（NLP）会议要求必须包含限制部分。接受论文数量的增加导致了大量自我报告限制的文献，这些文献无法全部手动审查，但仍然系统性地未被分析。因此，本文对2020年至2025年间在ACL和EMNLP会议上发表的论文的限制部分进行了大规模分析，以了解研究人员对自己工作的披露情况。为此，我们实施了一种新颖的人机框架，用于迭代混合定性编码。该框架使我们能够研究自我报告限制随时间的趋势、与特定论文属性的相关性，以及围绕这些披露的写作模式。我们的研究结果对NLP社区中研究人员报告的多样化挑战以及自我报告实践进行了重要反思。

View on arXiv Download PDF AI Translation

cs.CL / 103 / 2609.15194

Semiotic Relations and Proof Methods: A Cross-Genre Study of Argument Structure with Large Language Models

符号关系与证明方法：基于大型语言模型的跨体裁论证结构研究

de Lima, Edirlei Soares, Casanova, Marco A., Furtado, Antonio L.

Abstract

When a direct proof of a statement $S$ seems hard or even impossible to obtain, there may exist another statement (or set of statements) $S^{*}$, somehow related to $S$, on the basis of which $S$ can be proved. In order to investigate what options can be used to move from $S$ to $S^{*}$, four kinds of semiotic relations inspired by the four master tropes of semiotic research are briefly reviewed. Specifically, our syntagmatic, paradigmatic, antithetic and meronymic relations correspond, respectively, to metonymy, metaphor, irony and synecdoche. It is suggested that these four semiotic relations determine the options to move from $S$ to $S^{*}$, leading to proof by inference, proof by analogy, proof by contradiction, and proof by case analysis. To examine how the four relations are actually used across different kinds of argument, we complement the framework with an empirical study. We turn the four relations into explicit operational definitions and apply them to a cross-genre corpus of mathematical, legal, and everyday argument using a panel of large language models. We find that the relations are used very unevenly across genres: mathematical proofs draw on all four, whereas legal and everyday reasoning rely almost entirely on inference.

Chinese Translation

当一个陈述 $S$ 的直接证明似乎难以获得甚至不可能时，可能存在另一个与 $S$ 相关的陈述（或一组陈述）$S^{*}$，基于此可以证明 $S$。为了探讨从 $S$ 到 $S^{*}$ 的转变可以使用哪些选项，本文简要回顾了受符号研究四大主 tropes 启发的四种符号关系。具体而言，我们的联接关系、范畴关系、对立关系和部分关系分别对应于转喻、隐喻、讽刺和提喻。我们建议这四种符号关系决定了从 $S$ 到 $S^{*}$ 的转变选项，导致推理证明、类比证明、反证法和案例分析证明。为了考察这四种关系在不同类型论证中的实际使用情况，我们通过实证研究补充了这一框架。我们将这四种关系转化为明确的操作定义，并将其应用于一个跨体裁的数学、法律和日常论证语料库，使用一组大型语言模型。我们发现这些关系在不同体裁中的使用非常不均衡：数学证明使用了所有四种关系，而法律和日常推理几乎完全依赖于推理。

View on arXiv Download PDF AI Translation

cs.CL / 104 / 2609.15277

Artificial entrepreneurial cognition: Locating and causally steering an opportunity recognition dial inside large language models (LLMs)

人工创业认知：在大型语言模型（LLMs）中定位和因果引导机会识别拨盘

Fisch, Christian, Altmeier, Angela, Obschonka, Martin, Kosinski, Michal, Ni, Pin

Abstract

Entrepreneurial cognition is a foundation of entrepreneurship research. Yet the growing involvement of large language models (LLMs) in entrepreneurial work extends the cognition question beyond human actors to systems whose internal representations remain largely unexplored. We introduce artificial entrepreneurial cognition, the functional organisation of entrepreneurship-relevant representations and computations inside artificial intelligence (AI) systems. We bring mechanistic interpretability into entrepreneurship research through representation engineering. Focusing on opportunity recognition (OR), we construct 636 matched OR-present and OR-absent scenario pairs and recover an OR direction in Llama 3.1 8B-Instruct. Rather than infer the construct from outputs, we intervene directly on this direction, steering the model up and down along what we call the opportunity recognition dial, and its opportunity judgments shift with it. To our knowledge, this is the first causal intervention on an internal representation of an entrepreneurship construct inside an LLM. Held-out tests, lexical and topical controls, behavioural ablation, and geometric comparisons show that the direction is recoverable, consequential, and distinct from the opportunity evaluation and exploitation directions, although steering it also shifts judgments about these neighbouring stages. Recovery, signed steering, and geometric separation hold across four additional LLMs spanning different scales and families. These results give the contested distinction between opportunity recognition and evaluation a concrete representational form inside AI systems. More broadly, they establish internal representations as a new object of entrepreneurship inquiry and show how entrepreneurship theory can guide their identification, causal manipulation, and interpretation.

Chinese Translation

创业认知是创业研究的基础。然而，大型语言模型（LLMs）在创业工作中的日益参与将认知问题扩展到了人类行为者之外，涉及到内部表征仍然 largely 未被探索的系统。我们引入了人工创业认知，即在人工智能（AI）系统内部与创业相关的表征和计算的功能组织。通过表征工程，我们将机械解释性引入创业研究。聚焦于机会识别（OR），我们构建了636对匹配的OR存在和OR缺失场景，并在Llama 3.1 8B-Instruct中恢复了一个OR方向。我们并不是从输出中推断该构念，而是直接干预这个方向，沿着我们所称的机会识别拨盘上下调整模型，其机会判断也随之变化。据我们所知，这是对LLM内部创业构念的首次因果干预。保留测试、词汇和主题控制、行为消融以及几何比较表明，该方向是可恢复的、具有重要意义的，并且与机会评估和利用方向不同，尽管引导它也会影响对这些邻近阶段的判断。恢复、签名引导和几何分离在跨越不同规模和家族的四个额外LLM中保持一致。这些结果为机会识别与评估之间有争议的区分在AI系统内部提供了具体的表征形式。更广泛地说，它们将内部表征确立为创业研究的新对象，并展示了创业理论如何指导其识别、因果操控和解释。

View on arXiv Download PDF AI Translation

cs.CL / 105 / 2609.15309

When Agents Slow Down: Understanding LLM Agents' Test-Time Strategies via Elo-per-token Analysis

当智能体减速时：通过每个令牌的Elo分析理解大型语言模型智能体的测试时策略

Liu, Kaiyuan, Mang, Qiuyang, Peng, Bo, Chai, Wenhao, Li, Hanchen, Pimpalgaonkar, Shreyas, Zettlemoyer, Luke, Dimakis, Alex, Cheung, Alvin

Abstract

Large language model (LLM) agents allocate test-time compute adaptively as they revise solutions, use tools, explore alternatives, and decide when to stop. This test-time strategy makes it difficult to measure how agent performance scales. We study open-ended tasks that provide continuous scores for intermediate submissions, making progress observable throughout long trajectories. We propose Elo-per-token analysis, which tracks the best solution found at each token budget and uses a Bradley-Terry model to aggregate within-task orderings into Elo ratings across tasks with different score scales. We apply it to four general-purpose agents on four open-ended benchmarks, with sessions of up to 100M tokens, and to three feedback-driven LLM optimization harnesses in controlled single-task interventions. Independent sampling provides a theoretically characterized reference, for which Elo grows linearly with log compute. Against this reference, agents can initially convert tokens into Elo faster than independent sampling, but their marginal gains diminish and eventually fall below the reference. In contrast, the strongest historical human contestants improve superlinearly over contest time on shared AtCoder Heuristic Contest tasks, providing evidence of continual learning and substantial headroom after agents slow down. We define the scaling inflection point as the per-session budget where marginal Elo gains match the independent-sampling reference. Using this point as the per-session budget, we split 100M tokens across parallel sessions on FrontierCS Polyomino Packing, gaining +264 Elo over one long session and +355 over ten short sessions.

Chinese Translation

大型语言模型（LLM）智能体在修订解决方案、使用工具、探索替代方案和决定何时停止时，会自适应地分配测试时计算资源。这种测试时策略使得衡量智能体性能的扩展性变得困难。我们研究提供连续评分的开放式任务，以便在长时间轨迹中观察进展。我们提出了每个令牌的Elo分析，该分析跟踪在每个令牌预算下找到的最佳解决方案，并使用Bradley-Terry模型将任务内排序聚合为不同评分尺度任务的Elo评级。我们将其应用于四个通用智能体在四个开放式基准上的表现，测试会话长达1亿个令牌，并在受控的单任务干预中应用于三个反馈驱动的LLM优化工具。独立采样提供了一个理论上特征化的参考，其中Elo与对数计算线性增长。与该参考相比，智能体最初能够比独立采样更快地将令牌转化为Elo，但其边际收益逐渐减小，最终低于参考值。相比之下，历史上最强的人类参赛者在共享的AtCoder启发式竞赛任务上随着比赛时间的推移实现超线性提升，提供了持续学习和在智能体减速后仍有显著提升空间的证据。我们将扩展拐点定义为边际Elo收益与独立采样参考相匹配的每会话预算。以此点作为每会话预算，我们在FrontierCS多面体打包任务上将1亿个令牌分配到并行会话中，在一次长会话中获得+264 Elo，在十次短会话中获得+355 Elo。

View on arXiv Download PDF AI Translation

cs.CL / 106 / 2609.15338

Dynamic Semantic Compression for Efficient Latent-Space Inference in Large Language Models

用于大语言模型高效潜在空间推理的动态语义压缩

Li, Peipei, Zhang, Dongsen, Liu, Yuchen, Xu, Wenjun

Abstract

Large Language Models (LLMs) primarily perform inference at the token level, resulting in substantial memory overhead and compromised computational efficiency. In this paper, we propose a Dynamic Semantic Extraction and Inference (DSEI) framework, which achieves segment-level inference within the latent space through a two-stage training strategy. First, we construct a Dynamic Semantic Autoencoder (DSAE) via self-supervised learning. DSAE dynamically extracts segment-level semantics and compresses them into compact latent representations via adaptive semantic weighting and gated fusion. Subsequently, we integrate the DSAE into the LLM architecture and train the model to infer over dense latent space. DSEI substantially reduces both input and generation sequences and significantly enhances inference efficiency. Extensive experiments conducted on the Wanjuan dataset demonstrate that DSEI reduces perplexity by 48% compared to static sentence-level latent inference baseline. Furthermore, compared to standard LLMs using token-level inference, DSEI accelerates inference speed by 2.5$\times$ and reduces memory overhead by 90%.

Chinese Translation

大语言模型（LLMs）主要在标记级别进行推理，这导致了显著的内存开销和计算效率的降低。本文提出了一种动态语义提取与推理（DSEI）框架，通过两阶段训练策略实现潜在空间内的段级推理。首先，我们通过自监督学习构建了动态语义自编码器（DSAE）。DSAE 动态提取段级语义，并通过自适应语义加权和门控融合将其压缩为紧凑的潜在表示。随后，我们将 DSAE 集成到 LLM 架构中，并训练模型在密集潜在空间中进行推理。DSEI 显著减少了输入和生成序列，并显著提高了推理效率。在 Wanjuan 数据集上进行的大量实验表明，与静态句子级潜在推理基线相比，DSEI 将困惑度降低了 48%。此外，与使用标记级推理的标准 LLM 相比，DSEI 将推理速度提高了 2.5 倍，并减少了 90% 的内存开销。

View on arXiv Download PDF AI Translation

cs.CL / 107 / 2609.15369

SlopShape: Identifying AI-Generated Commercial Web Content

SlopShape：识别 AI 生成的商业网络内容

Madler, Jochen

Abstract

Word-level detectors identify unedited AI-generated text almost perfectly, but the literature documents their brittleness under rewording, and a word-level score neither characterizes a text nor identifies which AI model wrote it. We ask whether AI-generated text can be identified one level deeper, from structural signatures: how information is presented, in what order, with what evidence, and in what voice. We replicate StoryScope (Russell et al., 2026), which showed such patterns for AI-generated fiction, on commercial content: 2,250 pre-ChatGPT human blog posts from 268 company domains against 11,250 AI mirrors from five frontier models. A 214-feature instrument, applied by an LLM and validated in a human gold-annotation session (human-human kappa 0.928, human-model 0.946), detects AI posts from its 187 structural features alone at 98.0 macro-F1 on held-out companies, unchanged (98.1) when every AI post is reworded by its own model. The signal characterizes and attributes: AI posts share a tidy, self-announcing shape, 79.3% are attributed to the correct source against a 16.7% chance rate, and human posts occupy rare structural configurations. All effects replicate StoryScope's, consistent in direction and larger in magnitude. We release pipeline, instrument, prompts, code, and aggregate artifacts.

Chinese Translation

词级检测器几乎可以完美识别未编辑的 AI 生成文本，但文献记录了其在重述下的脆弱性，且词级评分既不能表征文本，也不能识别出是哪个 AI 模型生成的。我们探讨是否可以通过结构特征更深层次地识别 AI 生成的文本：信息的呈现方式、顺序、证据以及语调。我们复制了 StoryScope（Russell 等，2026）的研究，该研究展示了 AI 生成的虚构作品的此类模式，并将其应用于商业内容：从 268 个公司域名中收集的 2,250 篇 ChatGPT 之前的人类博客文章与来自五个前沿模型的 11,250 篇 AI 镜像进行对比。一个包含 214 个特征的工具，通过 LLM 应用，并在一轮人类黄金标注会话中验证（人类-人类 kappa 0.928，人类-模型 0.946），仅凭其 187 个结构特征在保留的公司中以 98.0 的宏观 F1 值检测 AI 帖子，当每个 AI 帖子由其自身模型重述时，该值保持不变（98.1）。该信号表征并归因：AI 帖子呈现出整洁、自我声明的形状，79.3% 被归因于正确的来源，而随机归因的概率为 16.7%，人类帖子则占据了稀有的结构配置。所有效果均复制了 StoryScope 的结果，在方向上保持一致且幅度更大。我们发布了管道、工具、提示、代码和汇总文档。

View on arXiv Download PDF AI Translation

cs.CL / 108 / 2609.15467

Turkish MMLU Pro: Traceable Option Augmentation and Its Validity Limits in Turkish Multiple-Choice Evaluation

土耳其 MMLU Pro：可追溯的选项增强及其在土耳其多项选择评估中的有效性限制

Bayram, M. Ali

Abstract

Adding answer options can lower multiple-choice scores without improving assessment validity. Turkish MMLU Pro examines this distinction using 12,000 Turkish-source questions across 58 sections. Each question retains its stem, five original options and source key, and receives five options copied from other questions in the same section. Sentence-embedding retrieval proposes candidates; a language model selects existing identifiers. Deterministic verification reconstructs all 60,000 additions. A 25-model calibration exposes scoring and generation-budget effects. Five evaluations produce source-key accuracies of 34.8%-81.4%. On 981 shared questions, one API-served model falls from 93.7% with five choices to 83.1% with ten; 102 of 115 lost correct responses select borrowed options. The decrease is 24.4 percentage points on heuristically flagged negative stems and 5.9 points elsewhere. A completed human-checked audit of 200 sampled questions, with undocumented reviewer tool use, yields 47 and 31 multiple-answer judgments across the two record sets, 25 of the latter unresolved. These records support concern about ambiguity, while their dependence and incomplete reviewer-method documentation limit validation. Because order and labels also change, the paired comparison measures augmentation as implemented. The contribution is a traceable construction and an analysis of its validity limits, not evidence that lower ten-choice scores measure knowledge better.

Chinese Translation

增加答案选项可能会降低多项选择的得分，而并未提高评估的有效性。土耳其 MMLU Pro 通过分析 12,000 道土耳其来源的问题，涵盖 58 个部分，探讨这一区别。每道问题保留其题干、五个原始选项及来源答案，并从同一部分的其他问题中复制五个选项。句子嵌入检索提出候选项；语言模型选择现有标识符。确定性验证重建所有 60,000 个新增选项。25 个模型的校准揭示了评分和生成预算的影响。五次评估产生了来源答案的准确率为 34.8%-81.4%。在 981 道共享问题中，一个 API 提供的模型在五个选项时的准确率为 93.7%，而在十个选项时降至 83.1%；115 个中有 102 个丢失的正确回答选择了借用的选项。在启发式标记的负题干上，准确率下降了 24.4 个百分点，而在其他题干上下降了 5.9 个百分点。对 200 道抽样问题进行的完整人工审核，尽管未记录审阅工具的使用，仍产生了两个记录集中的 47 和 31 个多答案判断，其中后者有 25 个未解决。这些记录支持对模糊性的担忧，但它们的依赖性和不完整的审阅者方法文档限制了验证。由于顺序和标签也会变化，配对比较测量了实施的增强。该研究的贡献在于提供了一个可追溯的构建及其有效性限制的分析，而不是证明较低的十选项得分能更好地衡量知识。

View on arXiv Download PDF AI Translation

cs.CL / 109 / 2609.15476

Temperature Fragility and the Conditional Benefits of Truncation Sampling

温度脆弱性与截断采样的条件性收益

La Rosa, Francesco

Abstract

Large language models generate text by sampling each token from a predicted distribution, and a temperature parameter sets how far the draw strays from the most probable tokens. Truncation samplers such as top-p and min-p discard the least probable tokens before the draw, so that sampling at high temperature stays coherent. Their reported accuracy gains come from temperatures of 1.5 to 3, while the defaults of deployed systems cluster between 0.6 and 1.0. Whether they change accuracy at those defaults, and for which models, has not been measured. We test thirteen open-weight models on GSM8K and MMLU-Pro at temperatures 0.7, 1.0, and 1.3 in one controlled pipeline, ten of them under eight decoding configurations. Six of the thirteen models lose 17 to 38 accuracy points on MMLU-Pro between 0.7 and 1.3, and the other seven lose at most 10. The lost accuracy comes from generations that run to the token limit or never state an answer. These results suggest that truncation samplers improve accuracy primarily when higher temperatures substantially degrade model performance. Where accuracy remains stable across temperatures, none of the tested truncation samplers improves on plain temperature sampling.

Chinese Translation

大型语言模型通过从预测分布中对每个标记进行采样来生成文本，温度参数设置了采样结果与最可能标记的偏离程度。截断采样器如 top-p 和 min-p 在采样之前会丢弃最不可能的标记，从而使得在高温度下的采样保持连贯性。它们报告的准确性提升来自于温度在 1.5 到 3 之间，而已部署系统的默认值则集中在 0.6 到 1.0 之间。尚未测量这些默认值是否会改变准确性，以及对于哪些模型会有影响。我们在一个受控流程中测试了十三个开放权重模型在 GSM8K 和 MMLU-Pro 上的表现，温度设置为 0.7、1.0 和 1.3，其中十个模型在八种解码配置下进行测试。在这十三个模型中，有六个模型在 MMLU-Pro 上的准确性在 0.7 和 1.3 之间下降了 17 到 38 个百分点，而其他七个模型最多下降 10 个百分点。准确性的下降主要源于生成结果达到标记限制或从未给出答案。这些结果表明，截断采样器主要在较高温度显著降低模型性能时提高准确性。在温度之间准确性保持稳定的情况下，测试的截断采样器均未能超越普通温度采样。

View on arXiv Download PDF AI Translation

cs.CL / 110 / 2609.15504

How Lossless Is Lossless Speculative Decoding? The Role of Numerical Precision in Orthrus

无损推测解码的无损性有多高？数值精度在Orthrus中的作用

Koziev, Ilya, Sinev, Leonid, Oseledets, Ivan

Abstract

Orthrus is a hybrid autoregressive-diffusion architecture that accelerates autoregressive language-model inference by generating multiple tokens in parallel while using a frozen autoregressive backbone. Its central claim is that an intra-model consensus mechanism enables lossless speculative decoding, producing the same output sequence as the autoregressive model. We independently reproduce Orthrus and examine this claim under different numerical precisions. Under BF16 inference, exact trajectory matching occurs in only 45% of cases for the authors' checkpoint and 43% for our independently trained model across 1,190 prompts from 12 domains. The probability of exact matching is also strongly associated with the response-conditional perplexity of the reference model. Despite this trajectory divergence, Orthrus does not show systematic degradation on downstream lm-eval-harness benchmarks. In contrast, repeating the trajectory evaluation with FP32 yields exact trajectory matching on all evaluated prompts. These results show that the practical losslessness of Orthrus depends on numerical precision and that exact trajectory equivalence should be evaluated separately from downstream task performance.

Chinese Translation

Orthrus是一种混合自回归-扩散架构，通过在使用冻结自回归骨干的同时并行生成多个标记，加速自回归语言模型的推理。其核心主张是，模型内部共识机制使得无损推测解码成为可能，从而产生与自回归模型相同的输出序列。我们独立复现了Orthrus，并在不同的数值精度下检验了这一主张。在BF16推理下，对于作者的检查点，精确轨迹匹配仅在45%的情况下发生，而对于我们独立训练的模型，这一比例为43%，基于来自12个领域的1190个提示。精确匹配的概率与参考模型的响应条件困惑度也有很强的关联。尽管存在轨迹偏差，Orthrus在下游lm-eval-harness基准测试中并未显示出系统性降级。相比之下，使用FP32重复轨迹评估在所有评估的提示中均实现了精确轨迹匹配。这些结果表明，Orthrus的实际无损性依赖于数值精度，并且精确轨迹等价性应与下游任务性能分开评估。

View on arXiv Download PDF AI Translation

cs.CL / 111 / 2609.15511

Authorship attribution and aesthetic evaluation of AI poetry: a case study with Haiku

人工智能诗歌的作者归属与美学评价：以俳句为案例研究

Oddi, Livia, Scardapane, Simone, Sugimoto, Toru, Genovese, Donatella

Abstract

This paper investigates the generation and human evaluation of Japanese haiku by contemporary Large Language Models (LLMs), focusing on authorship perception and aesthetic judgment within a constrained poetic form. Using a few-shot prompting strategy, Japanese haiku were generated across a heterogeneous set of large language models, including open- and closed-source systems, medium-scale and large-scale architectures, models with native or adapted Japanese support, and multilingual proprietary models. These AI-generated haiku were combined with human-written ones and presented in a questionnaire distributed to students at Japanese universities in Tokyo. The survey assessed whether respondents could distinguish between AI-generated and human-written haiku and which cues informed their judgments. Recognition accuracy varied across models. GPT-5, Gemini 2.5, and StableLM-7B performed at approximately chance level (approx 0.50), whereas LLM-JP, Gemma-2B, and LLaMA-2 showed moderate detectability (approx 0.59-0.67). However, recognition was strongly item-dependent. Ratings of fluency, coherence, poeticness, and related aesthetic dimensions predicted perceived humanness but not correct classification, indicating an attribution bias linked to aesthetic evaluation and revealing a dissociation between aesthetic evaluation and true authorship detection. The extended analysis additionally examines generation-constraint adherence, participant-level characteristics, and exploratory LLM-based evaluations of haiku authorship. Overall, the findings suggest that as LLMs improve, surface-level creative plausibility may reduce reliable human discrimination within constrained poetic settings.

Chinese Translation

本文探讨了当代大型语言模型（LLMs）生成和人类评估日本俳句的过程，重点关注在这一受限诗歌形式中的作者感知和美学判断。通过少量示例提示策略，生成了来自多种大型语言模型的日本俳句，包括开放源代码和闭源系统、中型和大型架构、具有本地或适应性日语支持的模型，以及多语言专有模型。这些人工智能生成的俳句与人类创作的俳句结合，并通过问卷的形式分发给东京的日本大学学生。调查评估了受访者是否能够区分人工智能生成的俳句与人类创作的俳句，以及哪些线索影响了他们的判断。识别准确性在不同模型之间存在差异。GPT-5、Gemini 2.5 和 StableLM-7B 的表现接近随机水平（约 0.50），而 LLM-JP、Gemma-2B 和 LLaMA-2 显示出中等的可检测性（约 0.59-0.67）。然而，识别结果在很大程度上依赖于具体项目。流畅性、一致性、诗意及相关美学维度的评分预测了感知的人性，但未能预测正确分类，这表明存在与美学评价相关的归属偏见，并揭示了美学评价与真实作者检测之间的脱节。扩展分析还考察了生成约束遵循情况、参与者特征以及基于 LLM 的俳句作者评估的探索性研究。总体而言，研究结果表明，随着 LLM 的改进，表面层次的创造性可信度可能会降低在受限诗歌环境中的可靠人类区分能力。

View on arXiv Download PDF AI Translation

cs.CL / 112 / 2609.15522

Psychosis involves a deficit of information compression in connected speech

精神病涉及连贯语言中的信息压缩缺陷

Vallisa, Samuele, Palominos, Claudio, He, Rui, Bora, Emre, Verim, Burcu, Demirlek, Cemal, Yalincetin, Berna, Homan, Philipp, Hinzen, Wolfram

Abstract

Large language models (LLMs) with human-like performance on linguistic tasks have transformed the study of language in neurodiverse conditions. LLMs provide representations of linguistic input in the form of high-dimensional vectors (embeddings), and next-token predictions computed from these embeddings. Previous crosslinguistic evidence suggests a complexity reduction in the form of both lower intrinsic dimensionality (ID) of LLM representations and higher mean surprisal (prediction error) in psychosis. We hypothesized that these metrics reflect a general deficit of information compression in psychosis, linked to grammatical organization as what enables predictions in language.We operationalized surprisal difference as the difference between surprisal as estimated from word frequency and surprisal as based on a contextual LM, which is sensitive to grammatical organization over and above lexical concepts. Using a dataset of 144 Turkish speakers, including 106 patients with schizophrenia-spectrum disorders (SSD) - 56 with chronic schizophrenia (SZH), 33 with first-episode psychosis (FEP), and 17 with schizoaffective disorder (SZA) - and 38 healthy controls. We report: (1) Surprisal difference is attenuated in all clinical groups relative to controls, independently of word count; (2) Compressibility (intrinsic dimension) is reduced in SZH and FEP; (3) Syntactic complexity and compressibility both predict surprisal difference. These results, further refining an alteration in the geometry of the semantic space in psychosis as previously attested, suggest a broader deficit in information compression in this disorder, with a mechanistic underpinning in the operations of grammar.

Chinese Translation

具有类人表现的大型语言模型（LLMs）在语言任务上的应用，改变了对神经多样性条件下语言研究的方式。LLMs以高维向量（嵌入）的形式提供语言输入的表示，并根据这些嵌入计算下一个词的预测。先前的跨语言证据表明，精神病患者在LLM表示的内在维度（ID）较低和平均惊讶度（预测误差）较高的形式上存在复杂性降低。我们假设这些指标反映了精神病中信息压缩的一般缺陷，这与语法组织有关，语法组织使得语言中的预测成为可能。我们将惊讶度差异操作化为基于词频估计的惊讶度与基于上下文语言模型（LM）的惊讶度之间的差异，后者对语法组织的敏感性超出了词汇概念。我们使用了一个包含144名土耳其语使用者的数据集，其中包括106名患有精神分裂症谱系障碍（SSD）的患者——56名慢性精神分裂症（SZH）、33名首次发作精神病（FEP）和17名情感性精神病（SZA）——以及38名健康对照者。我们报告： (1) 所有临床组的惊讶度差异相对于对照组均减弱，与词数无关； (2) SZH和FEP的可压缩性（内在维度）降低； (3) 句法复杂性和可压缩性均能预测惊讶度差异。这些结果进一步细化了精神病中语义空间几何的改变，表明该障碍在信息压缩方面存在更广泛的缺陷，其机制基础在于语法的运作。

View on arXiv Download PDF AI Translation

cs.CL / 113 / 2609.15528

To Each Language Its Tokenizer: Modular Tokenizers for Efficient Multilingual LLMs

为每种语言量身定制的分词器：高效多语言大语言模型的模块化分词器

Signe, Franck, Pilchen, Hippolyte, Yvon, François, Grave, Édouard

Abstract

Multilingual Large Language Models (LLMs) traditionally rely on a single vocabulary shared by all supported languages, which can lead to uneven compression across them. Moreover, their large embedding and output matrices increase memory usage and slow inference, notably for small-scale models. It is also wasteful as models are often used for only a subset of languages. To address these issues, we introduce a modular framework for multilingual model training. First, we propose methods to learn large modular BPE and Unigram tokenizers that enable extraction of subtokenizers tailored to any language subset. These subtokenizers achieve compression on par with monolingual tokenizers and improve cross-lingual fairness. Second, we design a pretraining strategy that samples subtokenizers to form batches, restricting predictions to the relevant vocabulary subset and allowing efficient training despite a large vocabulary. This supports efficient inference with any combination of language-specific vocabularies. Therefore, it reduces memory usage and speeds up inference in models without sacrificing performance.

Chinese Translation

多语言大语言模型（LLMs）传统上依赖于一个由所有支持语言共享的单一词汇，这可能导致各语言之间的不均匀压缩。此外，它们的大型嵌入和输出矩阵增加了内存使用并减慢推理速度，尤其是在小规模模型中。这种做法也显得浪费，因为模型通常只用于一部分语言。为了解决这些问题，我们提出了一种多语言模型训练的模块化框架。首先，我们提出了学习大型模块化 BPE 和 Unigram 分词器的方法，使其能够提取针对任何语言子集量身定制的子分词器。这些子分词器的压缩效果与单语分词器相当，并改善了跨语言的公平性。其次，我们设计了一种预训练策略，通过抽样子分词器来形成批次，限制预测在相关词汇子集内，从而允许在大型词汇的情况下进行高效训练。这支持了任何语言特定词汇组合的高效推理。因此，它减少了内存使用并加快了推理速度，而不牺牲性能。

View on arXiv Download PDF AI Translation

cs.CL / 114 / 2609.15559

Don't Count the Edits, Judge by the Outcome Alone: Reward-Based Evaluation for Grammatical Error Correction

不要只计算编辑次数，单凭结果进行评判：基于奖励的语法错误纠正评估

Ryu, Hayeong, Jo, Sunhee, Yu, Seunguk, Kim, YoungBin

Abstract

Grammatical error correction (GEC) evaluation has traditionally relied on reference or edit overlap, which can penalize valid rewrites that differ from gold corrections. Reference-free metrics reduce this dependence, but evaluating whether a fluent output is a valid correction of the source remains challenging. We propose SURE, a source-conditioned reward evaluator trained on within-source preferences spanning minimal-edit and rewrite-oriented corrections. SURE jointly learns an overall reward with criteria-level supervision for grammaticality, faithfulness, and fluency, together with span-level grounding for source-side error resolution. Experiments on SEEDA show that SURE performs competitively against strong baselines, with particular gains on rewrite-style corrections and more disentangled criteria-level diagnostics. Our code is available at https://github.com/hayeonggg/SURE.

Chinese Translation

语法错误纠正（GEC）评估传统上依赖于参考或编辑重叠，这可能会惩罚与黄金标准修正不同的有效重写。无参考指标减少了这种依赖，但评估流畅输出是否为源文本的有效纠正仍然具有挑战性。我们提出了 SURE，一种基于源条件的奖励评估器，经过训练以适应最小编辑和重写导向的纠正的源内偏好。SURE 共同学习整体奖励，并通过语法性、忠实性和流畅性等标准级监督，以及源侧错误解决的跨度级基础进行指导。在 SEEDA 上的实验表明，SURE 在与强基线的竞争中表现出色，特别是在重写风格的纠正和更为分离的标准级诊断方面取得了显著提升。我们的代码可在 https://github.com/hayeonggg/SURE 获取。

View on arXiv Download PDF AI Translation

cs.CL / 115 / 2609.15561

Can We Trust the Judges? Validation of Factuality Evaluation Methods via Answer Perturbation

我们可以信任评判者吗？通过答案扰动验证事实性评估方法

Gharsallah, Sarra, Robaldo, Adele, Tokareva, Mariia, Pinheiro, Giovanni Gatti, Guendouz, Ilyana, Troncy, Raphaël, Papotti, Paolo, Michiardi, Pietro

Abstract

Evaluating the factual correctness of large language models (LLMs) is vital for many applications. But are our evaluation tools themselves trustworthy? Despite the rise of factuality-based metrics, their sensitivity and reliability remain underexplored. This paper introduces a meta-evaluation framework that systematically tests these metrics using controlled corruptions of gold standard answers. Our method generates ranked outputs with known degrees of degradation to probe how metrics capture nuanced changes in truthfulness. Our experiments reveal that pipeline-based methods, such as the RAGAS's factual correctness metric, better track degradation than LLM-as-judge approaches. We also propose a new variant of the factual correctness metric that provides a competitive and cost-efficient.

Chinese Translation

评估大型语言模型（LLMs）的事实正确性对于许多应用至关重要。但我们的评估工具本身是否值得信赖？尽管基于事实性的指标逐渐增多，但它们的敏感性和可靠性仍然未得到充分探讨。本文引入了一种元评估框架，系统地使用对标准答案的控制性损坏来测试这些指标。我们的方法生成具有已知降级程度的排名输出，以探测这些指标如何捕捉真相的细微变化。我们的实验表明，基于管道的方法，例如 RAGAS 的事实正确性指标，比 LLM 作为评判者的方法更好地跟踪降级情况。我们还提出了一种新的事实正确性指标变体，提供了具有竞争力和成本效益的评估。

View on arXiv Download PDF AI Translation

cs.CL / 116 / 2609.15608

Through the Eyes of the Beholder: Biometric and Demographic Conditioning for Multimodal Sexism Detection

从观察者的视角：多模态性别歧视检测的生物特征与人口统计条件

Mocanu, Ana-Maria Luisa, Mocanu, Sebastian, Truică, Ciprian-Octavian, Apostol, Elena-Simona

Abstract

Detecting sexism on the internet is a fundamentally subjective task; our team, VANGUARD, addresses this challenge in the EXIST 2026 Task 2 by proposing a human-centered multimodal framework that analyses and incorporates the psychological and demographic characteristics of human annotators into the detection pipeline. We fuse five input modalities through a cross-attention architecture with Feature-wise Linear Modulation conditioning. Meme text is extracted and visually described with Gemma 4, then augmented by automatic translation between English and Spanish with NLLB-200. Text and image representations are produced by LoRAadapted XLM-RoBERTa and CLIP encoders and fused with sensor features encoded by a pretrained autoencoder. To model annotator subjectivity, we frame Subtask 2.1 as a label distribution learning problem, optimizing a Kullback-Leibler divergence loss over the full annotator label distribution. At inference time, predictions are produced by soft-voting between the deep multimodal network and a complementary SVM trained on stylometric and physiological features. Our best submission ranks 29th out of 114 on Subtask 2.2 (source intention) under soft evaluation, and the normalized ICM scores remain above the baseline on Subtasks 2.1 and 2.2, indicating that annotator-centered conditioning contributes a usable signal. We release our full pipeline and analysis to support reproducible human-centered modeling.

Chinese Translation

在互联网上检测性别歧视是一项根本上主观的任务；我们的团队 VANGUARD 在 EXIST 2026 任务 2 中通过提出一个以人为中心的多模态框架来应对这一挑战，该框架分析并结合人类标注者的心理和人口统计特征进入检测流程。我们通过具有特征线性调制条件的交叉注意力架构融合五种输入模态。通过 Gemma 4 提取和视觉描述模因文本，然后通过 NLLB-200 在英语和西班牙语之间进行自动翻译进行增强。文本和图像表示由 LoRA 调整的 XLM-RoBERTa 和 CLIP 编码器生成，并与由预训练自编码器编码的传感器特征融合。为了建模标注者的主观性，我们将子任务 2.1 定义为标签分布学习问题，优化整个标注者标签分布上的 Kullback-Leibler 散度损失。在推理时，通过深度多模态网络和在风格特征与生理特征上训练的补充 SVM 之间的软投票生成预测。我们的最佳提交在软评估下，在子任务 2.2（源意图）中排名 114 个中的第 29 位，并且在子任务 2.1 和 2.2 中标准化的 ICM 分数保持在基线之上，表明以标注者为中心的条件贡献了可用信号。我们发布了完整的流程和分析，以支持可重复的人本建模。

View on arXiv Download PDF AI Translation

cs.CL / 117 / 2609.15654

Empathy Is Steerable but Multi-Axial: Mechanism Geometry and Persona Effects in LLMs

同理心是可引导的但具有多轴特性：大型语言模型中的机制几何与人格效应

Ha, JuHeon, Lee, Byounghan, Choi, Yunseo, Sohn, Kyung-Ah

Abstract

Activation steering has been used to control traits such as honesty, refusal, and sycophancy, yet supportive empathy is evaluated along multiple dimensions that need not correspond to independently controllable activation directions. Using the EPITOME framework, which decomposes supportive empathy into Emotional Reactions, Interpretations, and Explorations, we study three instruction-tuned LLMs and ask whether candidate directions derived from these labels produce distinguishable intervention effects or instead share structure, and how persona prompts interact with those directions. We find that contrastive activation addition yields a stable middle-layer intervention that consistently shifts the EPITOME proxy scores across models, moving empathy analysis beyond response-level scoring. However, the recovered directions are only partially separable: steering one direction induces off-target shifts, and hand-crafted prompting shifts the empathy profile rather than isolating a single dimension. Persona prompts substantially change EPITOME scores, but a paired activation-shift decomposition shows that the recovered subspace captures only approximately 3 percent of persona-induced squared activation-shift magnitude at layer 15. Under this EPITOME-based definition, expressed empathy is steerable but multi-axial, and controlling persona-conditioned empathy requires targeting structure beyond individual mechanism directions.

Chinese Translation

激活引导已被用于控制诚实、拒绝和谄媚等特征，然而，支持性同理心的评估是沿着多个维度进行的，这些维度不必与独立可控的激活方向相对应。我们使用EPITOME框架，该框架将支持性同理心分解为情感反应、解释和探索，研究三种经过指令调优的大型语言模型（LLMs），并探讨从这些标签中得出的候选方向是否产生可区分的干预效果，或者是否共享结构，以及人格提示如何与这些方向相互作用。我们发现，对比激活添加产生了一个稳定的中间层干预，能够一致地在模型之间改变EPITOME代理分数，从而将同理心分析超越响应级评分。然而，恢复的方向仅部分可分离：引导一个方向会引发非目标偏移，而手工设计的提示则改变同理心特征，而不是孤立出单一维度。人格提示显著改变EPITOME分数，但配对的激活偏移分解显示，恢复的子空间仅捕捉到大约15层的人格引起的平方激活偏移幅度的3%。根据这一基于EPITOME的定义，表达的同理心是可引导的，但具有多轴特性，控制人格条件下的同理心需要针对超越单一机制方向的结构。

View on arXiv Download PDF AI Translation

cs.CL / 118 / 2609.15684

RESKILL: Explicit Failure Attribution and Structured Repair for Interactive Language Agents

RESKILL：交互式语言代理的显式失败归因与结构化修复

Deng, Mengyi, Li, Xin, Pan, Duyi, Wang, Zilin, Li, Zhiwei, Guo, Zhijiang, Wang, Wei

Abstract

Language agents increasingly rely on reusable skills, but post-failure repair is often handled by opaque one-shot reflection: a model generates a skill patch without explicitly maintaining how failure explanations relate to candidate repairs or how unsuccessful retests should influence later edits. We introduce RESKILL, a structured repair framework that maintains an explicit repair state across repair rounds. Given a failed rollout, the framework links failure hypotheses to candidate skill patches, selects local repairs through coverage-based attribution, retests the edited skill set in the environment, and uses retest outcomes to guide subsequent repair updates. The language model supplies structured repair factors, while the repair procedure records them, compares local skill patches by how well they address active failure explanations, and carries unsuccessful retest outcomes into later repair rounds. We evaluate RESKILL on ALFWorld and TextCraft across three model sizes under fixed repair budgets. RESKILL obtains the strongest final success in all six benchmark-model settings, improving average final success by 3.7 percentage points over direct repair and 3.3 points over hypothesis-conditioned repair. These results suggest that explicit attribution alone is insufficient; durable improvement emerges when attribution is integrated with repair selection and persistent retest-conditioned update.

Chinese Translation

语言代理越来越依赖可复用技能，但失败后的修复通常由不透明的一次性反思处理：模型生成技能补丁时，并未显式维护失败解释与候选修复之间的关系，也未考虑失败重测结果如何影响后续编辑。我们提出了RESKILL，一种结构化修复框架，在多轮修复过程中保持显式的修复状态。针对一次失败的执行，框架将失败假设与候选技能补丁关联，通过基于覆盖率的归因选择局部修复，在环境中重新测试编辑后的技能集，并利用重测结果指导后续修复更新。语言模型提供结构化的修复因素，修复过程记录这些因素，比较局部技能补丁对当前失败解释的解决效果，并将未成功的重测结果带入后续修复轮次。我们在ALFWorld和TextCraft两个环境中，针对三种模型规模并在固定修复预算下评估RESKILL。RESKILL在所有六个基准-模型设置中均取得最强的最终成功率，平均最终成功率较直接修复提升3.7个百分点，较基于假设的修复提升3.3个百分点。结果表明，仅有显式归因不足以实现持久改进；当归因与修复选择及持续的重测条件更新相结合时，才能获得稳健提升。

View on arXiv Download PDF AI Translation

cs.CL / 119 / 2609.15743

Merging the Knowledge of LLMs for Automatic Speech Recognition

合并大型语言模型的知识用于自动语音识别

Futami, Hayato, Kawahara, Tatsuya

Abstract

Automatic speech recognition (ASR) systems, trained on paired speech-text data, have been improved by leveraging language models (LMs) trained on text-only data. LM fusion methods such as shallow fusion and density ratio are well-established methods that incorporate external LMs during ASR decoding. However, they incur additional computational costs due to LM inference, which is particularly problematic for recent larger LMs. In this study, we propose incorporating external LMs via model merging. This method integrates the LMs directly into the parameters of an LLM-based ASR model, requiring no additional computational cost at inference. We formulate domain extension and transfer via arithmetic operations on LoRA parameters. Experimental evaluations were conducted for the domain adaptation of LLM-based ASR trained on CSJ and LibriSpeech. We show that our LM merging consistently improved the ASR performance in the target domains, without degrading inference speed or memory footprint.

Chinese Translation

自动语音识别（ASR）系统通过利用仅基于文本数据训练的语言模型（LM）得到了改善，这些系统是基于配对的语音-文本数据进行训练的。LM 融合方法，如浅层融合和密度比，是在 ASR 解码过程中引入外部 LM 的成熟方法。然而，由于 LM 推理，这些方法会产生额外的计算成本，这在最近更大的 LM 中尤为突出。在本研究中，我们提出通过模型合并来引入外部 LM。这种方法将 LM 直接集成到基于 LLM 的 ASR 模型的参数中，在推理时无需额外的计算成本。我们通过对 LoRA 参数进行算术运算来制定领域扩展和迁移。我们对基于 CSJ 和 LibriSpeech 训练的 LLM 基础 ASR 的领域适应进行了实验评估。结果表明，我们的 LM 合并方法在目标领域中持续提高了 ASR 性能，而没有降低推理速度或内存占用。

View on arXiv Download PDF AI Translation

cs.CL / 120 / 2609.15745

Look Before You Leap: Factual Decoding with Internal Attribution Signals

三思而后行：利用内部归因信号进行事实解码

Ryu, Hayeong, Yun, JungMin, Lim, Byeonggeuk, Jo, Sunhee, Kim, YoungBin

Abstract

Hallucination remains a critical challenge in large language models (LLMs), where early factual errors compound through autoregressive generation in a snowballing effect that neither post-hoc correction nor weight-level intervention can effectively preempt. We propose DescaPE (DEcoding Signal Control Against Path Error-snowballing), a decoding framework that leverages internal model signals to suppress hallucination-prone trajectories at inference time. Through sliding-window MLP ablation, we identify a factual-salient layer span within LLMs whose derived signal is selectively elevated for factual tokens and exhibits anomalous spikes at hallucination-prone steps. We train a lightweight probe to approximate this signal from a single forward pass and integrate it into candidate scoring to penalize high-risk continuations while rewarding factually grounded ones. Experiments across five factuality benchmarks on three LLMs demonstrate that DescaPE achieves factuality improvements over decoding-time baselines in multiple settings, while incurring only 1.10x latency overhead in our efficiency evaluation. Our code is available at https://github.com/hayeonggg/DESCAPE.

Chinese Translation

幻觉仍然是大型语言模型（LLMs）面临的一项关键挑战，在自回归生成过程中，早期的事实错误会通过雪球效应不断累积，这种效应无法通过事后修正或权重级别干预有效预防。我们提出了DescaPE（解码信号控制以防止路径错误雪球效应），这是一个利用内部模型信号在推理时抑制易幻觉轨迹的解码框架。通过滑动窗口多层感知器（MLP）消融实验，我们识别出LLMs中一个与事实相关的层跨度，其衍生信号在事实标记上被选择性提升，并在易幻觉步骤上表现出异常峰值。我们训练了一个轻量级探针，从单次前向传播中近似该信号，并将其整合到候选评分中，以惩罚高风险的延续，同时奖励基于事实的延续。在三个LLMs的五个事实性基准测试中的实验表明，DescaPE在多种设置下实现了相较于解码时间基线的事实性改进，同时在我们的效率评估中仅增加了1.10倍的延迟开销。我们的代码可在 https://github.com/hayeonggg/DESCAPE 获取。

View on arXiv Download PDF AI Translation

cs.CL / 121 / 2609.15758

Sequential Adapter Stacking for Cross-Lingual Low-Resource ASR

跨语言低资源自动语音识别的顺序适配器堆叠

Dang, Thai Thi Thanh Thao, Qian, Mengjie, Knill, Kate

Abstract

Extending large-scale multilingual automatic speech recognition (ASR) models to low-resource languages remains challenging. Model performance is skewed toward high-resource languages and degrades sharply for languages with limited labeled data and pre-training exposure. To address this, we investigate parameter-efficient approaches for transferring knowledge from resource-rich source languages to low-resource target languages on Whisper. Alongside warm initialization and attention-based fusion, we propose Sequential Adapter Stacking, which places a trainable target-language adapter on top of a frozen source-language adapter. Under controlled experiments, these approaches are evaluated on three target languages unsupported by Whisper -- Asturian, Assamese, and Xhosa -- using source languages with varying degrees of relatedness. Sequential Adapter Stacking with the closest related source consistently and significantly outperforms full fine-tuning across the three targets, with 5--8\% relative WER reductions. These gains largely persist with only one hour of target training data.

Chinese Translation

将大规模多语言自动语音识别（ASR）模型扩展到低资源语言仍然面临挑战。模型性能偏向于高资源语言，而对于标注数据和预训练曝光有限的语言则急剧下降。为了解决这一问题，我们研究了在Whisper上从资源丰富的源语言向低资源目标语言转移知识的参数高效方法。除了温暖初始化和基于注意力的融合外，我们提出了顺序适配器堆叠（Sequential Adapter Stacking），该方法在冻结的源语言适配器上方放置一个可训练的目标语言适配器。在受控实验中，这些方法在三个Whisper不支持的目标语言——阿斯图里亚语（Asturian）、阿萨姆语（Assamese）和科萨语（Xhosa）——上进行了评估，使用了不同相关性的源语言。与最接近相关的源语言的顺序适配器堆叠在三个目标语言上始终显著优于完全微调，具有5%到8%的相对字错误率（WER）降低。这些增益在仅有一小时目标训练数据的情况下也大部分得以保持。

View on arXiv Download PDF AI Translation

cs.CL / 122 / 2609.15759

Enabling Streaming User Transcription in Full-Duplex Speech-to-Speech Models

在全双工语音到语音模型中实现流式用户转录

Hu, Ke, Ferchichi, Nourchene, Casanova, Edresson, Pasad, Ankita, Rastorgueva, Elena, Chen, Chen, Koluguri, Nithin Rao, Zelasko, Piotr, Peng, Yifan, Xu, Hainan, Chen, Zhehuai, Ginsburg, Boris

Abstract

Full-duplex speech-to-speech (S2S) models enable natural conversational AI by allowing simultaneous listening and speaking. However, these models typically lack inherent user speech transcription, which is essential for applications such as conversation logging, accessibility features, and quality monitoring. In this work, we propose an efficient method to add streaming ASR capabilities to an existing duplex S2S model by introducing a lightweight ASR head in parallel to the agent text head. Our approach requires minimal additional parameters and no significant architectural changes to the base S2S model, enabling real-time user transcription while preserving full-duplex conversational capabilities including turn-taking and barge-in handling. Experimental results demonstrate that our method achieves streaming average WER of 10.21% on the HuggingFace Open ASR Leaderboard within the duplex S2S framework. Additionally, we show that the same architecture trained as a standalone streaming ASR model achieves competitive results (7.73% WER) compared to current SOTA models. We will open-source our training and inference code to facilitate further research in joint streaming ASR and S2S modeling.

Chinese Translation

全双工语音到语音（S2S）模型通过允许同时听和说，促进了自然对话人工智能的发展。然而，这些模型通常缺乏内置的用户语音转录功能，而这一功能对于对话记录、无障碍功能和质量监控等应用至关重要。在本研究中，我们提出了一种高效的方法，通过在代理文本头旁并行引入轻量级的自动语音识别（ASR）头，将流式ASR能力添加到现有的双工S2S模型中。我们的方法需要最少的额外参数，并且对基础S2S模型没有重大架构更改，从而实现实时用户转录，同时保留全双工对话能力，包括轮次交替和插话处理。实验结果表明，我们的方法在双工S2S框架内在HuggingFace Open ASR Leaderboard上实现了10.21%的流式平均字错误率（WER）。此外，我们还展示了相同架构作为独立流式ASR模型训练时，与当前最先进（SOTA）模型相比，取得了具有竞争力的结果（7.73% WER）。我们将开源我们的训练和推理代码，以促进联合流式ASR与S2S建模的进一步研究。

View on arXiv Download PDF AI Translation

cs.CL / 123 / 2609.15830

CiteGuard-RAG: A Validation-Centered AI System for Evidence-Grounded Question Answering

CiteGuard-RAG：一个以验证为中心的证据基础问答人工智能系统

Barua, Sumit, Hong, Guan, Dursunoglu, Halil, Rodgers, Charles, Fong, Alvis

Abstract

Retrieval-augmented generation (RAG) can improve access to complex information; however, retrieving evidence alone does not ensure that answers are grounded, citation-valid, or appropriately refused. This paper introduces CiteGuard-RAG, a validation-centered AI system for evidence-grounded question answering. The system integrates hybrid semantic-lexical retrieval, citation-constrained generation, sentence-level grounding validation, and single-pass regeneration. Validation is used at runtime to determine whether a candidate answer should be accepted, refused, or regenerated before final delivery. CiteGuard-RAG is evaluated on 400 questions across a controlled housing-law dataset, PrivacyQA, and CUAD. In the controlled evaluation, it achieves 99.1% retrieval accuracy, 98.3% grounded-answer accuracy, and 98.3% citation validity, with no validation-detected hallucinations. Ablation results show that grounded-answer accuracy drops sharply when validation is removed, even when retrieval accuracy remains unchanged. External evaluation shows that while citation validity remains strong, evidence utilization, span alignment, and refusal calibration become harder under domain shift. These findings indicate that trustworthy RAG systems require explicit validation between retrieval and final answer delivery. CiteGuard-RAG provides a practical architecture for linking retrieval, generation, citation checking, abstention, and regeneration in high-stakes information access.

Chinese Translation

检索增强生成（RAG）可以改善对复杂信息的访问；然而，仅仅检索证据并不能确保答案是有依据的、引用有效的或适当地拒绝的。本文介绍了CiteGuard-RAG，一个以验证为中心的证据基础问答人工智能系统。该系统集成了混合语义-词汇检索、引用约束生成、句子级别的基础验证和单次重生。验证在运行时用于确定候选答案是否应被接受、拒绝或在最终交付前重新生成。CiteGuard-RAG在一个受控的住房法律数据集、PrivacyQA和CUAD上评估了400个问题。在受控评估中，它实现了99.1%的检索准确率、98.3%的基础答案准确率和98.3%的引用有效性，且没有检测到验证中的幻觉。消融实验结果表明，当移除验证时，基础答案的准确率急剧下降，即使检索准确率保持不变。外部评估显示，尽管引用有效性仍然强劲，但在领域转移下，证据利用、跨度对齐和拒绝校准变得更加困难。这些发现表明，值得信赖的RAG系统需要在检索和最终答案交付之间进行明确的验证。CiteGuard-RAG提供了一种实用架构，用于在高风险信息访问中连接检索、生成、引用检查、弃权和重生。

View on arXiv Download PDF AI Translation

cs.CL / 124 / 2609.15849

Before You Poll with LLMs: A Deliberative Diagnostic Framework

在使用大型语言模型进行民意调查之前：一种审议诊断框架

Wali, Ahmed, Tayyab, Hassaan

Abstract

Can LLMs reason through new information like humans, or do they merely retrieve cached opinions? This is critical for silicon sampling, where LLM personas simulate public opinion at scale. Current evaluations test only whether personas hold the right opinions -- a static snapshot. But opinion research increasingly depends on dynamic fidelity: whether personas update beliefs in response to new arguments, as humans do during deliberation. No existing benchmark tests this. We introduce the Deliberative Polling Diagnostic Framework, which compares human and LLM belief shifts after identical informational interventions. Grounded in deliberative polling, it surfaces failures invisible to static evaluation: models that produce plausible partisan opinions can still misrepresent how those opinions change. Applying the framework to five frontier models using data from America in One Room (526 personas, 72 questions), we find that every model fails, each in a unique manner. GPT-5.1 exhibits reversal: its personas become more hostile toward the opposing party after balanced information, while humans become less so. This reversal is selective (80% on outgroup vs. 26% on policy questions) and symmetric across partisan identities. Gemini 2.0 Flash, Claude Sonnet 4.5, and Llama 3.3 70B exhibit overshoot, shifting correctly but at 5-7x human magnitude. DeepSeek V3 exhibits rigidity with near-zero change. Targeted ablations reveal that policy content triggers these failures and that they are identity-specific: GPT-5.1 reverses on outgroup questions but overshoots on ingroup; Gemini shows the inverse. We term this signature self-sycophancy: conformity to the model's internal stereotype of the persona rather than reasoning from the information provided. Our framework offers a concrete protocol: run the deliberative diagnostic before trusting LLM personas to mimic revised beliefs.

Chinese Translation

大型语言模型（LLMs）能像人类一样通过新信息进行推理，还是仅仅检索缓存的观点？这对于硅采样至关重要，在这种情况下，LLM角色在大规模上模拟公众舆论。目前的评估仅测试角色是否持有正确的观点——这只是一个静态快照。然而，舆论研究越来越依赖于动态的真实性：角色是否会像人类在审议过程中那样，根据新论点更新信念。现有的基准测试并未对此进行评估。我们提出了审议投票诊断框架，该框架比较人类和LLM在相同信息干预后的信念变化。基于审议投票，它揭示了静态评估无法察觉的失败：能够产生似是而非的党派观点的模型，仍可能错误地表现出这些观点如何变化。我们将该框架应用于五个前沿模型，使用来自《美国在一个房间》（America in One Room）的数据（526个角色，72个问题），发现每个模型均失败，且每个模型的失败方式各不相同。GPT-5.1表现出逆转：其角色在接收到平衡信息后对对立党派变得更加敌对，而人类则变得不那么敌对。这种逆转是选择性的（在外群体问题上为80%，在政策问题上为26%），并且在党派身份之间是对称的。Gemini 2.0 Flash、Claude Sonnet 4.5和Llama 3.3 70B表现出过度反应，虽然正确地转变，但幅度是人类的5-7倍。DeepSeek V3表现出僵化，几乎没有变化。针对性的消融实验表明，政策内容触发了这些失败，并且这些失败是身份特定的：GPT-5.1在外群体问题上逆转，但在内群体问题上过度反应；Gemini则表现出相反的情况。我们将这种特征称为自我阿谀奉承：遵循模型对角色的内部刻板印象，而不是根据提供的信息进行推理。我们的框架提供了一个具体的协议：在信任LLM角色模仿修正信念之前，先进行审议诊断。

View on arXiv Download PDF AI Translation

cs.CL / 125 / 2609.15851

Learning to Coach for Experiential Learning

学习指导以促进体验学习

Chen, Guanheng, Ye, Tianzhu, Dong, Li, Wu, Xun, Huang, Shaohan, Wei, Furu

Abstract

Language models can learn from experience, but raw solution trajectories are often too long and noisy to provide effective guidance. In this work, we propose Learning to Coach (L2C), a framework that trains a dedicated LLM-as-a-Coach to extract actionable experiential knowledge from an actor model's previous trajectory. The actor remains frozen, while the LLM-as-a-Coach is trained to maximize a reward given by the correctness of the actor's guided response. We study two such rewards: a same-instance reward, which improves subsequent responses on the original problem, and a cross-instance reward, which elicits knowledge that transfers to other instances. Across mathematical reasoning and interactive text-games, L2C consistently outperforms self-refinement and an untrained LLM-as-a-Coach. Running experiential learning for more iterations further improves accuracy and uses additional inference compute more effectively than enlarging the actor's decoding budget. The trained LLM-as-a-Coach also transfers to out-of-distribution tasks and adapts its guidance to the specific actor it coaches.

Chinese Translation

语言模型可以通过经验学习，但原始解决方案轨迹往往过长且噪声过大，无法提供有效的指导。在本研究中，我们提出了学习指导（Learning to Coach, L2C）框架，该框架训练一个专门的语言模型作为教练（LLM-as-a-Coach），从行为者模型的先前轨迹中提取可操作的经验知识。行为者保持不变，而LLM-as-a-Coach则被训练以最大化由行为者指导响应的正确性所给出的奖励。我们研究了两种奖励：同实例奖励（same-instance reward），它改善了对原始问题的后续响应，以及跨实例奖励（cross-instance reward），它引发了可转移到其他实例的知识。在数学推理和互动文本游戏中，L2C始终优于自我优化和未训练的LLM-as-a-Coach。进行更多迭代的体验学习进一步提高了准确性，并比扩大行为者的解码预算更有效地利用了额外的推理计算。训练后的LLM-as-a-Coach还能够转移到分布外任务，并根据其指导的特定行为者调整其指导。

View on arXiv Download PDF AI Translation

cs.CL / 126 / 2609.15855

K-Bench: a clinically calibrated benchmark for evaluating large language models in high-risk mental health conversations

K-Bench：用于评估高风险心理健康对话中大型语言模型的临床校准基准

Vowels, Laura M., Vowels, Matthew J., Sharma, Shivali, Jha, Apoorv, Choudhury, Rehnuma, Sarraj, Wasseem El, Francois-Walcott, Rachel, Hussain, Aruba, Ingram, Sarah, Loulopoulou, Angela, Segal, Adva, Volkova, Elena

Abstract

% !TEX root = ../main.tex People increasingly use large language models (LLMs) for mental health support, yet their safety in evolving, high-risk conversations remains poorly characterised. We developed K-Bench, a clinician-calibrated, protected benchmark evaluating 125 model configurations representing 33 base models from 14 providers across a fixed cohort of 200 multi-turn vignettes involving suicide, self-harm, domestic violence, substance misuse, and no-risk presentations. Synthetic patient conversations showed substantial distributional overlap with real human-AI conversations. A frozen GPT-4o judge achieved 94.2% exact agreement with clinician consensus across 6,751 eligible item comparisons from 151 clinician-rated transcripts. Leading models combined strong supportive conversation with combined-risk scores above 95, whereas risk exploration exposed substantial variation among lower-performing configurations. Therapeutic prompting produced configuration-specific gains concentrated among weaker models, while elevated reasoning produced no average improvement. K-Bench combines broader clinical coverage and configuration-scale comparison with a continuously updated public leaderboard whose operational test materials are protected from direct optimisation. The leaderboard is available at www.k-bench.ai.

Chinese Translation

人们越来越多地使用大型语言模型（LLMs）进行心理健康支持，但它们在不断演变的高风险对话中的安全性仍然缺乏充分的特征描述。我们开发了K-Bench，这是一个经过临床校准的受保护基准，评估125种模型配置，代表来自14个提供者的33种基础模型，涵盖200个多轮情境，涉及自杀、自残、家庭暴力、物质滥用和无风险表现。合成患者对话显示出与真实人类-人工智能对话之间存在显著的分布重叠。一个冻结的GPT-4o评审在151个临床评分转录本的6,751个合格项目比较中，与临床共识达成94.2%的完全一致。领先模型结合了强有力的支持性对话和超过95的综合风险评分，而风险探索则揭示了表现较差的配置之间存在显著的变异。治疗性提示在较弱模型中产生了特定配置的增益，而提升推理并未带来平均改善。K-Bench结合了更广泛的临床覆盖和配置规模比较，并设有一个不断更新的公共排行榜，其操作测试材料受到直接优化的保护。排行榜可在www.k-bench.ai获取。

View on arXiv Download PDF AI Translation

cs.CL / 127 / 2609.15886

Inoculation Midtraining with Learned Neologisms

使用学习到的新词进行接种中期训练

O'Brien, Kyle, Young, Edward James, Radmard, Puria, Kirch, Nathalie, Tice, Cameron, Korbak, Tomek, Africa, David Demitri

Abstract

Large language models (LLMs) often learn both desirable and undesirable properties during post-training. We study whether midtraining, an earlier training stage, can shape which of these properties later generalise. We introduce Inoculation Midtraining, a technique that teaches a base model that unsafe behaviour belongs to a designated context, as indicated by the neologism (a new token) introduced during midtraining, and then post-trains the model on unsafe data within that context. We then evaluate the model outside the context, with the neologism excluded from the system prompt. Across supervised fine-tuning and reinforcement learning post-training regimes, we find that Inoculation Midtraining can reduce misalignment while preserving the transfer of benign data properties (e.g., speaking in German or Shakespearean prose). However, our approach does not outperform standard Inoculation Prompting, is sensitive to training configuration, and produces a leaky boundary that nearby contextual cues can reactivate. These results show that inoculation with a learned association introduced via midtraining can shape selective generalisation. Still, more work is needed before this approach can become a load-bearing component in a developer's safety framework.

Chinese Translation

大型语言模型（LLMs）在后期训练中通常会学习到既有期望的属性，也有不期望的属性。我们研究中期训练这一较早的训练阶段是否能够影响这些属性的后续泛化。我们提出了一种接种中期训练（Inoculation Midtraining）技术，该技术教会基础模型不安全行为属于指定的上下文，这一上下文通过在中期训练中引入的新词（一个新标记）来指示，然后在该上下文内对模型进行不安全数据的后期训练。随后，我们在不包含新词的系统提示下评估模型在上下文外的表现。在监督微调和强化学习后期训练的过程中，我们发现接种中期训练能够减少不一致性，同时保留良性数据属性的转移（例如，用德语或莎士比亚的散文进行表达）。然而，我们的方法并未优于标准的接种提示（Inoculation Prompting），对训练配置敏感，并且产生了一个易泄漏的边界，附近的上下文线索可能会重新激活这些属性。这些结果表明，通过中期训练引入的学习关联的接种可以影响选择性泛化。然而，在该方法能够成为开发者安全框架中的支柱组件之前，仍需进行更多的研究。

View on arXiv Download PDF AI Translation

cs.CL / 128 / 2609.15938

HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses

HypoEvolve：遗传算法使多智能体大型语言模型发现科学假说

Liu, Jieyuan, Hu, Mengzhou, Chen, Jefferson, Kong, JungHo, Jagannatha, Pratibha, Gao, Yiming, Pratt, Dexter, Lee, Hsin-Yuan, Hu, Zhiting, Ideker, Trey, Wang, Wei, Xing, Eric P., Wang, Zhen

Abstract

Scientific agents contribute to hypothesis discovery by synthesizing evidence, assessing proposals, and developing new explanations. Recent systems combine scientific agents with evolutionary search through critique, comparison, and revision. However, how different forms of agent collaboration affect hypothesis quality remains an open question. Answering this question requires separating the effects of agents' scientific capabilities from those of their collaboration. A framework must therefore preserve agents' scientific roles and support rules for combining, revising, and retaining hypotheses. Building on this view, we introduce HypoEvolve, which makes collaboration explicit through successive updates to a hypothesis population. Specifically, we propose a generational genetic algorithm to coordinate specialized large language model (LLM) agents that integrate mechanistic arguments, reconsider assumptions, and assess evidence and testability. Each generation specifies how scientific judgments and new proposals reshape the population, making collaboration effects on hypothesis quality directly testable. Moreover, we design our evaluation around scientifically meaningful hypotheses that explain how a proposed intervention could work. Drug repurposing links these explanations to target-level biological claims assessed against external evidence. Specifically, we adapt DepMap and Open Targets into complementary external measures grounded in experimental, genetic, and clinical evidence. Across 34 cancer types, HypoEvolve achieves the highest scores against six baselines on both measures. DepMap selectivity reaches 0.171, versus 0.115 for the strongest baseline. Gains over single-pass generation also generalize to held-out cancer types. HypoEvolve advances a vision of autonomous science in which AI research teams achieve a capacity for discovery beyond that of individual models.

Chinese Translation

科学智能体通过综合证据、评估提案和发展新解释来促进假说的发现。最近的系统通过批评、比较和修订将科学智能体与进化搜索相结合。然而，不同形式的智能体协作如何影响假说质量仍然是一个未解的问题。回答这个问题需要将智能体的科学能力与其协作效果分开。因此，框架必须保留智能体的科学角色，并支持组合、修订和保留假说的规则。在这一观点的基础上，我们引入了HypoEvolve，通过对假说群体的连续更新使协作变得明确。具体而言，我们提出了一种代际遗传算法，以协调集成机械论论证、重新考虑假设以及评估证据和可测试性的专业大型语言模型（LLM）智能体。每一代都明确科学判断和新提案如何重塑群体，使协作对假说质量的影响可以直接测试。此外，我们围绕科学上有意义的假说设计评估，这些假说解释了提议的干预如何发挥作用。药物重定位将这些解释与针对外部证据评估的靶向生物学主张联系起来。具体而言，我们将DepMap和Open Targets调整为基于实验、遗传和临床证据的互补外部测量。在34种癌症类型中，HypoEvolve在这两种测量上都达到了六个基线中最高的分数。DepMap的选择性达到0.171，而最强基线为0.115。单次生成的增益也推广到保留的癌症类型。HypoEvolve推动了自主科学的愿景，其中人工智能研究团队的发现能力超越了单个模型。

View on arXiv Download PDF AI Translation

cs.CL / 129 / 2609.15964

Verifiable by Construction: Claim-Level Evaluation of Verbatim Citation in Clinical Question Answering

可构建验证：临床问题回答中逐条引用的评估

Zhang, Jiashuo, Chen, Yuling, Commodore-Mensah, Yvonne, Oberst, Michael

Abstract

Large language models (LLMs) have been widely adopted for clinical question answering (QA). Current systems can attach citations to their answers, but these often point to broad texts, leaving time-pressed clinicians unable to verify them efficiently. An alternative is to ensure that responses are verifiable by construction: providing fine-grained verbatim quotes from reference material that substantiate claims, so users can verify an answer without opening other documents. In this paper, we evaluate the ability of current models to perform this task end-to-end: from providing citations for every factual claim, to producing verbatim quotes, to ensuring that those quotes fully substantiate the claims. To do so, we build a standardized harness over four clinical practice guidelines and evaluate twelve LLMs on 222 synthetic clinical questions, measuring each of these stages separately. We find that most models can attach verbatim quotes to over 90% of their claims from prompting alone, apart from some lightweight models such as claude-haiku-4.5. Yet these quotes often fail to substantiate every detail of the claims they accompany. For instance, claude-opus-5 produces verbatim quotes for 98.0% of its claims, but fully substantiates only 37.1%. Our work provides insights into the current capability gap of LLMs in building verifiable clinical QA systems, along with artifacts for future research.

Chinese Translation

大型语言模型（LLMs）已被广泛应用于临床问题回答（QA）。当前系统可以将引用附加到其答案中，但这些引用通常指向广泛的文本，使得时间紧迫的临床医生无法高效验证。一个替代方案是确保响应在构建上是可验证的：提供来自参考材料的细粒度逐字引用，以支持声明，从而使用户能够在不打开其他文档的情况下验证答案。在本文中，我们评估当前模型在这一任务上的端到端能力：从为每个事实声明提供引用，到生成逐字引用，再到确保这些引用完全支持声明。为此，我们在四个临床实践指南上构建了一个标准化的测试框架，并在222个合成临床问题上评估了12个LLM，分别测量每个阶段的表现。我们发现，大多数模型能够仅通过提示将逐字引用附加到其超过90%的声明上，除了某些轻量级模型，如claude-haiku-4.5。然而，这些引用通常未能支持它们所伴随的每个声明的所有细节。例如，claude-opus-5为98.0%的声明生成逐字引用，但仅完全支持37.1%。我们的研究提供了对LLMs在构建可验证的临床QA系统方面当前能力差距的深入见解，并为未来研究提供了相关材料。

View on arXiv Download PDF AI Translation

cs.CL / 130 / 2609.15972

Mind2Dialogue: Training Human-Aware Language Models by Simulating User Mental States

Mind2Dialogue：通过模拟用户心理状态训练人性化语言模型

Wang, Zixuan, Zhou, Yufan, Tang, Jinzhou, Yu, Xinle, Wu, Chengjun, Ye, Lyumanshan, Feng, Zhaoxiang, Peng, Letian, Patra, Adyasha, Bai, Fan, Ma, Enze, Hu, Zhengding, Gu, Jianyang, Wang, Zhao, Ding, Yufei, Shang, Jingbo, Shu, Tianmin, Hu, Zhiting, Wang, Zhen

Abstract

As language models become more capable, long-term collaboration in learning, reasoning, and decision-making calls for a deeper understanding of the people they serve. Yet training such human-aware language models faces a fundamental supervision gap because current datasets for LLM assistant training contain few if any well-informed responses explicitly grounded in users' unspoken beliefs and goals. Scaling such supervision is inherently constrained, as users' underlying states are not directly observable. We thus propose the Mind2Dialogue framework to mitigate this gap by simulating users' mental states and turning them into privileged supervision for human-aware training. Specifically, we first propose a psychology-guided simulator that preserves personal characteristics while updating mental states through interaction to generate coherent conversations. The key idea is to enforce a shared evolving mental state that drives user behavior and guides an Oracle assistant's responses. Our privileged distillation then trains models on the Oracle's well-informed responses to assist users without direct access to their mental states at deployment. Moreover, we propose to evaluate human-aware learning by combining personalization and theory of mind, examining how models understand people and act on that understanding. Training on the full Mind2Dialogue corpus improves every reported personalization metric over the corresponding Qwen, Llama, and OLMo instruction-tuned baselines, including gains of 26.6 to 40.9 percentage points in preference-following generation. The gains extend to belief and action reasoning on Qwen and Llama, beyond personalized assistance. Looking forward, Mind2Dialogue makes user simulation a foundation for genuine AI collaborators that understand beliefs and intentions behind people's words and support their long-term goals across education, work, and everyday life.

Chinese Translation

随着语言模型能力的提升，学习、推理和决策中的长期合作需要对其服务的用户有更深入的理解。然而，训练这种人性化语言模型面临着根本的监督缺口，因为当前用于大型语言模型（LLM）助手训练的数据集几乎没有明确基于用户未言明的信念和目标的充分信息响应。由于用户的潜在状态无法直接观察，这种监督的扩展受到固有限制。因此，我们提出了Mind2Dialogue框架，通过模拟用户的心理状态并将其转化为人性化训练的特权监督，以减轻这一缺口。具体而言，我们首先提出了一种心理学指导的模拟器，该模拟器在通过互动更新心理状态的同时保留个人特征，以生成连贯的对话。关键思想是强制执行一个共享的、不断发展的心理状态，该状态驱动用户行为并指导Oracle助手的响应。我们的特权蒸馏随后在Oracle的充分信息响应上训练模型，以在部署时帮助用户，而无需直接访问他们的心理状态。此外，我们建议通过结合个性化和心智理论来评估人性化学习，考察模型如何理解人类并基于这种理解采取行动。在完整的Mind2Dialogue语料库上训练，提升了每个报告的个性化指标，相较于相应的Qwen、Llama和OLMo指令调优基线，偏好跟随生成的增益达26.6至40.9个百分点。这些增益还扩展到Qwen和Llama上的信念和行动推理，超越了个性化辅助。展望未来，Mind2Dialogue将用户模拟作为真正理解人们言语背后信念和意图的AI合作者的基础，支持他们在教育、工作和日常生活中的长期目标。

View on arXiv Download PDF AI Translation

cs.CL / 131 / 2609.15973

Discovery Foundation Models: Toward Open-Ended Discovery Intelligence

发现基础模型：迈向开放式发现智能

Yang, Ling, Yin, Zhenfei, Wu, Yingcheng

Abstract

Foundation models have progressed from learning and reasoning over existing knowledge, to increasingly learning through action, tool use, and outcome feedback. We argue that the next frontier is a further transition: from solving and acting within problems specified by humans to participating in the process by which new problems, representations, explanations, and knowledge are created. We refer to this capability as Discovery Intelligence. We formulate Discovery Foundation Models (DFMs) as general-purpose model systems for open-ended discovery. A DFM operates over a revisable research state and supports seven coupled capabilities spanning problem discovery, formulation, representation construction, hypothesis formation, intervention, evidence-grounded revision, and continual discovery improvement. We instantiate this framework with Zetema, which couples explicit research-state dynamics, verification and experimental gating, external grounding, and cross-task Discovery Skill evolution. We further ground the framework with GALILEO, a real therapeutic-discovery system in which Dry-Lab reasoning, robotic and hands-on Wet-Lab experimentation, external biological evidence, and iterative hypothesis and design revision form a closed physical discovery loop. We then formulate a unified approach to capability formation and process-centered evaluation, enabling discovery behavior to be trained, improved, and measured beyond final-answer performance. Together, these components establish discovery as a learnable, executable, and evaluable capability of foundation-model systems. We view this shift as a broader progression in intelligence scaling: from learning over existing knowledge, to learning from action outcomes, and ultimately to participating in the construction, testing, and revision of the structures through which new knowledge is discovered. Code: https://github.com/Gen-Verse/DFM-Plans

Chinese Translation

基础模型已经从对现有知识的学习和推理，逐渐转向通过行动、工具使用和结果反馈进行学习。我们认为下一个前沿是进一步的转变：从在由人类指定的问题中解决和行动，转向参与新问题、表征、解释和知识的创建过程。我们将这种能力称为发现智能（Discovery Intelligence）。我们将发现基础模型（Discovery Foundation Models, DFMs）构建为用于开放式发现的通用模型系统。DFM在可修订的研究状态上运行，并支持七种相互关联的能力，涵盖问题发现、表述、表征构建、假设形成、干预、基于证据的修订和持续发现改进。我们用Zetema实例化这一框架，它结合了明确的研究状态动态、验证和实验门控、外部基础和跨任务发现技能演化。我们进一步用GALILEO这一真实的治疗发现系统来支持该框架，其中干实验室推理、机器人和动手实验室实验、外部生物证据以及迭代假设和设计修订形成一个封闭的物理发现循环。然后，我们制定了一种统一的能力形成和过程中心评估的方法，使发现行为能够超越最终答案的表现进行训练、改进和测量。综合这些组件，我们将发现确立为基础模型系统的一种可学习、可执行和可评估的能力。我们将这一转变视为智能扩展的更广泛进展：从对现有知识的学习，到从行动结果中学习，最终参与新知识发现的结构的构建、测试和修订。代码：https://github.com/Gen-Verse/DFM-Plans

View on arXiv Download PDF AI Translation

cs.CL / 132 / 2609.15975

Disentangling Representation Evolution in Transformers through Directional Decomposition

通过方向分解解开变换器中表征演变的复杂性

He, Shwai, Zhang, Haichao, Yan, Shen

Abstract

Transformer representations evolve through learned additive transformations that either preserve their current direction or redirect it. We study this evolution as a functional geometry, decomposing learned updates into parallel and perpendicular components. Across pretrained models, we find substantial parallel components beyond the residual identity path. We then apply the decomposition in two spaces: to attention and MLP updates relative to the hidden state, and to attention value aggregation relative to the current token's value. Targeted edits reveal a strongly space-dependent asymmetry: exclude-self value-space parallel manipulation is markedly more robust than residual-space and perpendicular counterparts, preserving the direct self message while scaling only the non-self aggregate. The same decomposition gives a component-resolved description of compression-induced update error: perpendicular error separates compression methods more clearly than parallel error. Extensive experiments further demonstrate that full-aggregate parallel suppression during from-scratch pretraining lowers validation-loss trajectories and improves downstream averages, with the value-space variant strongest. Together, these results connect representation geometry to editing robustness, compression diagnosis, and training-time intervention. Code is available in the \href{https://github.com/Shwai-He/Transformer-Geometry}{project repository}.

Chinese Translation

变换器的表征通过学习的加性变换演变，这些变换要么保持当前方向，要么重新定向它。我们将这种演变视为一种功能几何，将学习到的更新分解为平行和垂直分量。在预训练模型中，我们发现除了残差恒等路径外，还存在显著的平行分量。然后，我们在两个空间中应用这种分解：相对于隐藏状态的注意力和多层感知器（MLP）更新，以及相对于当前标记值的注意力值聚合。针对性的编辑揭示了一个强烈的空间依赖性不对称性：自我排除值空间的平行操作明显比残差空间和垂直操作更具鲁棒性，能够保持直接的自我信息，同时仅缩放非自我聚合。同样的分解为压缩引起的更新误差提供了分量解析的描述：垂直误差比平行误差更清晰地区分压缩方法。大量实验进一步表明，在从头预训练期间完全聚合的平行抑制降低了验证损失轨迹并改善了下游平均表现，其中值空间变体最强。这些结果将表征几何与编辑鲁棒性、压缩诊断和训练时干预联系起来。代码可在[项目仓库](https://github.com/Shwai-He/Transformer-Geometry)中获取。

View on arXiv Download PDF AI Translation

arXiv Papers

Seeing What the Vehicle Sees: Video-Augmented Virtual Reality for Physical Autonomous Vehicles

ShieldVLA: Feasibility-Aware Safety Alignment for Vision-Language-Action Models

Harnessing human expertise for high-precision robotic assembly in industrialized construction: A sample-efficient installer-in-the-loop interactive reinforcement learning framework

Learning Manipulation-Sufficient Representations via Outcome Bottlenecks

Self-Evolving AI for Humanoids: Mechanisms, Safety, and Evaluation of Post-Deployment Self-Improvement

GzDRL: Reproducible and Scalable Deep Reinforcement Learning with Gazebo

Physical Kernel: Structured Visual Latents for Dark Manipulation

Conflict-Predictive Variable Horizons in Multi-Drone Distributed Model Predictive Control

Extending the Speed Limit of Quadrupedal Locomotion via Refined Actuator Modeling and Adaptive Command Scheduling

Breaking speed scaling in quadrupedal robots via Huygens' coupled-pendulum dynamics

Ergodic Control and Controlled Diffusion for Robot Learning: Review and Tutorial

IMM-based Multiple Object Tracking using a State Prediction Neural Network

Attention-DP3: Spatially Object-aware 3D Diffusion Policy via Geometry-aligned Attentional Conditioning

Bridging Thought and Action: Taming Long-Horizon Instability in Open-Source LLM Agents with a MetaTool-Enhanced ROS Framework

Chance-Constrained Belief-Space Maneuver Planning for Autonomous Collision Avoidance Under Uncertainty

STAGE: Diagnosing Semantic Transfer at Grounded Execution in Embodied Agents

Constraint-Grounded Reinforcement Learning for Variable Impedance Control in Contact-Rich Robotic Insertion

Runtime-Incremental Transformer for Reinforcement-Learning-Based Adaptive Control

An MRI-Guided Robotic System to Improve Hippocampal Access for Epilepsy Interventions

Differential Realizability of Static Control Allocation in Multirotors: An Impossibility under Nonredundant Full Actuation and a Pseudoinverse Obstruction under Redundant Actuation

From Vision to Harvest: Benchmarking Vision-Language Models for Multi-Arm Robotic Fruit Harvesting

How Well Do Pseudo-Rigid Body Models Capture Real Plants? In-Field Validation of Simulated Blueberry Canes

Compositional Shift Algebra: Extrapolating Mixed Robot Shifts Without Mixed Finetuning

Does Online Gravity Estimation Matter? Revisiting a Silent Design Split in LiDAR-Inertial Odometry

How to Better Train VLAs: Lessons Learned From the REAL-I Challenge at ICRA 2026

GROOVE: Geometry-Guided Reduction of Operational-Space Jerk in VLA Execution

Decentralized Multi-Robot Task Allocation Under Degraded Communication: A Benchmark of Performance, Reliability, and Computation

Learning In-Hand Object Reaching to General 6D Poses

When Do Learned Priors Help Visual Inertial Estimation? A Controlled Study of Prior Integration, Calibration, Initialization, and Backend Consistency

Force-Aware Reinforcement Learning with Hybrid Sensorless Force Estimation for Wheeled-Legged Loco-Manipulation

GeomVLA: Unifying Scene, Motion, and Action in 3D

LePlanner: An Iterative Amortized Controller For World Models

ReWeight: Leveraging Human Data for VLA Post-Training via Demonstration Retrieval and Sample Weighting

Co-Speech with You: Training-Free Personalization of Robot Co-Speech Gestures

Rethinking the Implications of Human Feedback for Preference Learning in Human-Robot Collaboration

What Makes an Efficient VLA? Navigating Action-Head Design, Scaling, and Latency

Mobile Multi-Robot Navigation under Runtime Uncertainty via Koopman Operator Learning and Nonlinear Model Predictive Control

Gradient-Free Neural Hamilton-Jacobi Reachability for Scalable Safety-Critical Control

Vision-Force Admittance Learning for Peg Insertion into a Movable Hole

When Faster VLA Deployment Changes Closed-Loop Behavior: Task Success-Latency Analysis of SmolVLA Across PyTorch and ONNX Variants

Visible Touch: Rendering Contact for Visuomotor Policies

GIFT: Glove-Inferred Force Transfer: Force-Aware Human-to-Robot Skill Transfer from a Wearable Sensing Glove to a Robot Hand Without Tactile Sensors

DreamSat-Bench: Development and Initial Testing of a Testbed for AI-Based Pose Estimation from 3D Reconstruction

Novel Ex-vivo Calf Brain Model with Integrated Sub-Skull Force Sensors to Access Simulated Neurosurgical Procedures

ITA-LaCAM: A Complete and Scalable TAPF Solver via Assignment-Aware Configuration-Space Search

Task-Specified Active Metrological Inspection with Measurement-Steered VLA Manipulation and Deterministic Evidence Gating

VGFM: Expressive Robot Policies via Dense Value Guidance in Flow Matching

Learning Communication-Conditioned Generative Policies for Decentralized Multi-Agent Collision Avoidance

Multi-Task Visual Perception Network with LLM Conditioning for Autonomous Navigation

VLBiMan++: Expanding the Generalization Boundary of Vision-Language Anchored One-Shot Bimanual Manipulation

BIG-CBF: Behavior-Imagination-Guided Control Barrier Function with Shared Uncertainty for Mobile Robot Navigation

Orientation Control of Soft Robots via Adiabatic Spectral Submanifolds

Learning-Based Dynamic Obstacle Avoidance for a UAV Using Only Three Range Sensors

EMoG: Emotion-Modulated Gait Generation for Expressive Humanoid Locomotion

EcoBoat: Design and Experimental Validation of an Autonomous Body-Board Boat For Cleaning Water Bodies

Fault Diagnosis for Underwater Vehicles using Moving Horizon Estimation and Gaussian Processes

NavPatch: Evidence-Guided Object-Level Costmap Correction with Vision-Language Models

Language-Grounded Semantic Target Navigation for Autonomous Surface Vehicles

GLAM: Training a latent world model over global spatiotemporal memory for active exploration and navigation

Learning Multi-Agent Task Assignment and Navigation in the Factory: from Simulation to Real Robots

Robotic Servo Tracking of Moving Targets with Dynamic Imitation Constraints

REVOLVE: An Automated Closed-Loop Framework for Evolving Robot Manipulation with Minimal Human Intervention

Skill Composition for Legged Robot Reinforcement Learning

A Novel Robot-Assisted Learning Pedagogy for Children with ASD

Beyond Dead Reckoning: A Point of View on Camera--DAS--GNSS Continuity in Road Tunnels

An Adaptive Fixed-Time Line-of-Sight Guidance Scheme for 3D Path Following of Underwater Vehicles: Theory and Experiment

A Personalized Dynamic Balance Evaluation Paradigm for Hip Exoskeleton-Assisted Walking under Unexpected Ground Perturbations

Language-Guided Representation Learning for Robust Cross-Sensor Material Recognition

Belief-Adaptive Online Autonomy for Quadrotor UAV Navigation under GNSS Degradation in Urban Environments

Primitive-Informed Sampling-Based MPC for Multi-Fingered Dexterous Manipulation

Real-World Reinforcement Learning with MPC Scaffolding for Dexterous Manipulation

HydroMap: Probabilistic Water Surface Elevation Mapping for Semantic Scene Representation in Inland Waterways

Exact Feasibility Certification and Optimal Responsibility Allocation for Multi-Robot CBF Safety Filters

Comparing Trajectories from Positions Alone: Curvature-Based Time Alignment and Drift Error Metric

Two-Stage Personalized Gait Phase Estimation in Stroke Survivors During Exoskeleton-Assisted Walking: An Offline Feasibility Study

From Learned-Mode AV-Traffic Pairing to Planner Decisions: A Marginal-Preserving Study on Argoverse 2

IMPACT-VLA: Interaction-aware Multimodal Propagation Attribution via Counterfactual Trajectories for Vision-Language-Action Policies

Atomic Motion Coordinate for Language-Steerable and Force-Responsive Manipulation

Steering Generative Robot Policies with Lexicographic Preferences