cs.RO / 1 / 2608.12416
RoboSynChallenge: Mastering Real-World Dexterity via Generalizing Synthesized Manipulation Skills
RoboSynChallenge：通过泛化合成操作技能掌握现实世界的灵巧性
Abstract
Achieving generalizable robotic manipulation remains a central challenge in embodied intelligence. Despite rapid advances in model architectures and learning algorithms, progress is often limited by the scarcity and narrow diversity of real-world data. The RoboSynChallenge competition introduces a unified benchmark to evaluate and advance the generalizability of manipulation policies across a spectrum of tasks, environments, and difficulty levels. To alleviate the shortage of realistic data, the challenge integrates large-scale synthetic data generation with standardized real-world robotic evaluation. Participants are encouraged to leverage synthesized state-action trials to improve general-purpose policy learning, while final assessments are conducted exclusively on unseen real-world manipulation environments. Baseline implementations, including Transformer-, Diffusion-, Vision-Language-Action, and World-Action-Model-based policies, are provided to ensure reproducibility and comparability. By coupling scalable simulation-based training with rigorous real-world validation, RoboSynChallenge aims to foster the development of broadly capable, data-efficient, and adaptable manipulation systems, thereby paving the way toward truly general robotic intelligence.
Chinese Translation
实现可泛化的机器人操作仍然是具身智能中的一个核心挑战。尽管模型架构和学习算法迅速发展，但进展往往受到现实世界数据稀缺和多样性狭窄的限制。RoboSynChallenge 竞赛引入了一个统一的基准，以评估和推动操作策略在各种任务、环境和难度级别上的泛化能力。为了缓解现实数据的短缺，该挑战将大规模合成数据生成与标准化的现实世界机器人评估相结合。鼓励参与者利用合成的状态-动作试验来改善通用策略学习，而最终评估则仅在未见过的现实世界操作环境中进行。提供了包括基于 Transformer、扩散（Diffusion）、视觉-语言-动作（Vision-Language-Action）和世界-动作-模型（World-Action-Model）策略的基线实现，以确保可重复性和可比性。通过将可扩展的基于仿真的训练与严格的现实世界验证相结合，RoboSynChallenge 旨在促进广泛能力、数据高效和适应性强的操作系统的发展，从而为实现真正的通用机器人智能铺平道路。
cs.RO / 2 / 2608.12650
Attune: A Self-Annotation Tool for Understanding Robot Operator Attention Profiles
Attune：一种用于理解机器人操作员注意力特征的自我标注工具
Abstract
Deploying robot fleets in complex, real-world environments requires human operators to supervise multiple robots simultaneously. Managing operator attention is a fundamental challenge of designing multi-robot supervision interfaces, encompassing both feed layout and feed content (i.e., robot behavior design). Thus far, designers lack empirical guidance on the latter-how to change a robot's behavior to capture, sustain, or relinquish operator attention during multi-robot supervision. In our vision of the future, designers should be able to use this guidance to calibrate robot behavior to different operator attention profiles. Treating operator eye gaze as a robot behavior design clue, we created a pre-deployment elicitation tool called Attune. Attune automatically identifies when meaningful gaze shifts occur, provides AI assistance for annotating why shifts occurred, and outputs a summary of operator gaze patterns for operator review. We evaluated Attune through a user study in which participants annotated the visual triggers that drew their attention. Our findings unveil variation in observed gaze patterns and reveal how Attune helps characterize operator attention.
Chinese Translation
在复杂的现实环境中部署机器人群体需要人类操作员同时监督多个机器人。管理操作员的注意力是设计多机器人监督界面的一个基本挑战，涉及到信息流布局和信息流内容（即机器人行为设计）。到目前为止，设计者在后者方面缺乏实证指导——如何改变机器人的行为以在多机器人监督过程中捕获、维持或放弃操作员的注意力。在我们对未来的设想中，设计者应该能够利用这些指导来调整机器人的行为以适应不同的操作员注意力特征。将操作员的眼动视线视为机器人行为设计的线索，我们创建了一种名为Attune的预部署引导工具。Attune能够自动识别有意义的视线转移发生的时刻，提供人工智能辅助以注释视线转移的原因，并输出操作员视线模式的总结供操作员审阅。我们通过一项用户研究评估了Attune，参与者在研究中标注了吸引他们注意的视觉触发因素。我们的研究结果揭示了观察到的视线模式的变化，并展示了Attune如何帮助描述操作员的注意力。
cs.RO / 3 / 2608.12683
FUSE: Active Functional Affordance Grounding through Adaptive Semantic-Geometric Evidence Acquisition
FUSE：通过自适应语义-几何证据获取实现主动功能可用性定位
Abstract
Embodied agents must often identify and interact with objects based on their function rather than their identity, requiring them to actively acquire observations that reveal discriminative functional evidence. Existing affordance grounding methods operate from fixed viewpoints and lack mechanisms for deciding where to look when functional cues are occluded or incomplete. We introduce Active Functional Affordance Grounding, a new task in which an agent sequentially explores a scene to identify and spatially ground an object satisfying a functional query. To address this problem, we propose FUSE, an adaptive semantic-geometric evidence acquisition framework that combines explicit uncertainty-driven exploration with a learned amortized planner to efficiently select informative viewpoints. We further introduce a Habitat-based benchmark for evaluating active functional grounding. Experiments show that FUSE achieves the highest observed non-oracle grounding performance while reducing computation by 1.33x relative to fully explicit exploration, and remains effective across multiple affordance knowledge sources.
Chinese Translation
具身智能体通常必须根据对象的功能而非身份来识别和与之互动，这要求它们主动获取揭示区分性功能证据的观察。现有的可用性定位方法通常从固定视角出发，缺乏在功能线索被遮挡或不完整时决定观察位置的机制。我们引入了主动功能可用性定位这一新任务，其中智能体顺序探索场景，以识别并空间定位满足功能查询的对象。为了解决这个问题，我们提出了FUSE，一个自适应的语义-几何证据获取框架，结合了显式的不确定性驱动探索和学习的摊销规划器，以高效选择信息丰富的视角。我们进一步引入了一个基于Habitat的基准，用于评估主动功能可用性定位。实验表明，FUSE在非oracle定位性能上达到了最高水平，同时相较于完全显式探索减少了1.33倍的计算量，并且在多种可用性知识源中仍然有效。
cs.RO / 4 / 2608.12707
SAP-Nav: Spatial Semantic Representation Meets Active Perception for Hierarchical Open-Vocabulary Object Navigation
SAP-Nav：空间语义表示与主动感知相结合的层次开放词汇对象导航
Abstract
Hierarchical open-vocabulary object navigation (OVON) requires agents to follow free-form instructions that may specify targets through scene-, room-, region-, and instance-level cues in unseen environments. Although recent work LangMap has formalized this setting, reliably solving it under partial observations remains challenging: spatial grounding requires persistent environment-level evidence, whereas target verification requires clear and discriminative candidate views. We present SAP-Nav, a fully online, zero-shot framework that addresses both requirements through active perception. SAP-Nav incrementally constructs a Queryable Spatial-Semantic Representation from actively acquired room views, enabling spatial semantic queries from any explored location. It further employs Active Viewpoint Verification to assess whether the current observation provides sufficient evidence and, when necessary, reposition the agent to a more informative viewpoint before verifying candidates against category and attribute constraints. Although designed for hierarchical OVON, SAP-Nav supports both hierarchical and standard category-level OVON without task-specific training or precomputed scene maps. Experiments on LangMap and HM3D-OVON show that SAP-Nav achieves the overall best performance, including a 12.2% improvement in SR over training-based methods on region-level navigation. Real-world robot experiments further demonstrate its practical feasibility. Code will be made publicly available upon acceptance.
Chinese Translation
层次开放词汇对象导航（OVON）要求智能体在未见环境中根据自由形式的指令进行导航，这些指令可能通过场景、房间、区域和实例级别的线索来指定目标。尽管最近的研究 LangMap 已经对这一设置进行了形式化，但在部分观察下可靠地解决这一问题仍然具有挑战性：空间基础需要持久的环境级证据，而目标验证需要清晰且具有区分性的候选视图。我们提出了 SAP-Nav，这是一种完全在线的零样本框架，通过主动感知来满足这两项要求。SAP-Nav 从主动获取的房间视图中逐步构建可查询的空间语义表示，使得可以从任何探索过的位置进行空间语义查询。它进一步采用主动视点验证来评估当前观察是否提供了足够的证据，并在必要时将智能体重新定位到更具信息量的视点，然后根据类别和属性约束验证候选项。尽管 SAP-Nav 是为层次 OVON 设计的，但它支持层次和标准类别级别的 OVON，而无需特定任务的训练或预计算的场景地图。在 LangMap 和 HM3D-OVON 上的实验表明，SAP-Nav 实现了整体最佳性能，包括在区域级导航中相较于基于训练的方法提高了 12.2% 的成功率。实际机器人实验进一步证明了其实际可行性。代码将在接受后公开发布。
cs.RO / 5 / 2608.12755
Genetic Fuzzy System-Based Multi-Robot Coordination for Planetary Missions
基于遗传模糊系统的多机器人协调在行星任务中的应用
Abstract
This paper proposes a decentralized approach for a multi-robot system (MRS) using a genetic fuzzy system to perform a collaborative object transportation task that minimizes the total path length of the MRS in unstructured environment while avoiding obstacles. For an environment given by an elevation map, terrain traversability analysis with respect to the slope is performed to reduce the dimension and identify non-traversable areas that can be considered as obstacles, and the given map is converted into a traversability map in two dimensional space. In the training process, proposed fuzzy inference systems (FISs) to generate the MRS's velocity for transporting an object to a target position are optimized by a genetic algorithm with several scenarios, such as a local minima, a target that is close to an obstacle, and a cluttered environment. The trained FIS models are applied to the testing environment, which is the converted traversability map, and validated using multiple scenarios.
Chinese Translation
本文提出了一种去中心化的多机器人系统（MRS）方法，利用遗传模糊系统执行协作物体运输任务，旨在最小化MRS在非结构化环境中的总路径长度，同时避免障碍物。针对给定的高程图，进行地形可 traversability 分析，以坡度为依据，减少维度并识别可视为障碍物的不可 traversable 区域，并将给定地图转换为二维空间中的可 traversability 地图。在训练过程中，提出的模糊推理系统（FISs）通过遗传算法进行优化，以生成MRS在将物体运输到目标位置时的速度，考虑了多种场景，如局部最小值、靠近障碍物的目标以及杂乱环境。经过训练的FIS模型应用于测试环境，即转换后的可 traversability 地图，并通过多种场景进行验证。
cs.RO / 6 / 2608.12835
AirForesight: Current-to-Future Spatial Map Imagination with Cross-Space Planning Consistency for UAV-VLN
AirForesight：具有跨空间规划一致性的无人机视觉-语言导航的当前到未来空间地图想象
Abstract
Unmanned Aerial Vehicle Vision-Language Navigation (UAV-VLN) requires agents to follow language instructions, infer spatial structure from sparse multi-view observations, and execute feasible 3D motion in complex outdoor environments. Despite recent progress with large language models, most existing methods still map vision-language inputs directly to actions, providing limited explicit scene grounding and future-aware spatial reasoning. We propose AirForesight, a current-to-future spatial map imagination framework for UAV-VLN. AirForesight first learns a structured current-map representation from multi-view observations. This representation is jointly supervised by current-map reconstruction and future-trajectory prediction, encouraging it to encode both present scene structure and future motion intent. Under structured causal attention, the current spatial knowledge is propagated to future-map reasoning, and the resulting current and future representations are aggregated to predict the next 3D waypoint. To make spatial imagination more relevant to navigation, we introduce a cross-space planning consistency loss that encourages directional agreement between the predicted map-space trajectory and the expert action direction derived from the ground-truth waypoint displacement. Experiments on OpenUAV and AerialVLN-S, together with extensive ablations, demonstrate strong performance and support the effectiveness and stability of the proposed framework.
Chinese Translation
无人机视觉-语言导航（UAV-VLN）要求代理遵循语言指令，从稀疏的多视角观测中推断空间结构，并在复杂的户外环境中执行可行的三维运动。尽管最近在大型语言模型方面取得了进展，但大多数现有方法仍然将视觉-语言输入直接映射到动作，提供的场景基础和未来感知空间推理有限。我们提出了AirForesight，一个用于UAV-VLN的当前到未来空间地图想象框架。AirForesight首先从多视角观测中学习结构化的当前地图表示。该表示通过当前地图重建和未来轨迹预测共同监督，鼓励其编码当前场景结构和未来运动意图。在结构化因果注意力下，当前空间知识被传播到未来地图推理中，生成的当前和未来表示被聚合以预测下一个三维航点。为了使空间想象与导航更加相关，我们引入了一种跨空间规划一致性损失，鼓励预测的地图空间轨迹与从真实航点位移推导的专家动作方向之间的方向一致性。在OpenUAV和AerialVLN-S上的实验，以及广泛的消融实验，展示了强大的性能，并支持所提框架的有效性和稳定性。
cs.RO / 7 / 2608.12840
ASPIRE-VINS: Adaptive Spline-based Visual-inertial Navigation System With Robust 3D Measurement Residuals
ASPIRE-VINS：具有鲁棒3D测量残差的自适应样条视觉惯性导航系统
Abstract
Visual-inertial navigation systems estimate six-degree-of-freedom motion by fusing visual and inertial data. Modern discrete-time methods with IMU preintegration provide strong accuracy and efficiency, but keyframe-based representations can be less flexible when residuals must be evaluated at arbitrary timestamps or when motion-dependent temporal resolution is needed. Continuous-time splines address this issue by representing the trajectory as a smooth temporal function, but uniformly spaced knots can under-represent rapid dynamics or over-parameterize static intervals. This letter proposes ASPIRE-VINS, a continuous-time VINS framework that combines adaptive knot placement (AKP), multi-resolution splines (MRS), and 3D measurement-space residuals (3D-MSR). AKP allocates knots according to local motion variation, MRS adds bounded local refinement in tangent space, and 3D-MSR provides bearing consistency by aligning transformed features with calibrated observation rays in 3D measurement space. Experiments show that ASPIRE-VINS achieves competitive or lower trajectory errors than the compared baselines, demonstrating the effectiveness of motion-adaptive continuous-time trajectory modeling under diverse motion and sensing conditions.
Chinese Translation
视觉惯性导航系统通过融合视觉和惯性数据来估计六自由度运动。现代离散时间方法结合IMU预积分提供了强大的准确性和效率，但基于关键帧的表示在必须在任意时间戳评估残差或需要运动依赖的时间分辨率时可能不够灵活。连续时间样条通过将轨迹表示为平滑的时间函数来解决这一问题，但均匀间隔的节点可能无法充分表示快速动态或过度参数化静态区间。本文提出了ASPIRE-VINS，一个结合自适应节点放置（AKP）、多分辨率样条（MRS）和3D测量空间残差（3D-MSR）的连续时间VINS框架。AKP根据局部运动变化分配节点，MRS在切空间中增加有界的局部细化，而3D-MSR通过将变换特征与3D测量空间中的校准观测光线对齐来提供方向一致性。实验表明，ASPIRE-VINS在轨迹误差方面与比较基线相当或更低，展示了在多样化运动和传感条件下运动自适应连续时间轨迹建模的有效性。
cs.RO / 8 / 2608.12854
BrainWAM: Action-Space Coordination of Semantic Priors and Predictive Dynamics for Autonomous Driving
BrainWAM：语义先验与预测动态的行动空间协调用于自动驾驶
Abstract
Autonomous driving requires planning under both semantic constraints and predictive dynamics. Existing end-to-end driving approaches, however, typically emphasize only one side of this requirement: Vision-Language-Action (VLA) models exploit VLM priors for semantic reasoning, while World Action Models (WAMs) provide future-aware prediction through generative world modeling. This naturally motivates a unified planner that can leverage both semantic priors and predictive dynamics. However, we find that a naive combination through joint token-level attention suffers from an attention-allocation mismatch, where semantic shortcuts dominate the shared attention space and suppress predictive dynamics. Inspired by neuroscience evidence that complex behavior arises from coordination among functionally specialized systems, we propose BrainWAM, a structured action-space coordination framework that converts semantic reasoning and predictive world modeling into two specialized action-oriented pathways, and aligns them at the level of compact action representations. We further introduce an asynchronous rectified-flow inference strategy with decoupled video and action denoising, which shortens inference latency while preserving planning-relevant predictive context. BrainWAM reaches state-of-the-art performance on both NAVSIM v1 (89.5 PDMS) and NAVSIM v2 (89.6 EPDMS), consistently outperforming VLA-only or WAM-only methods, highlighting BrainWAM as a practical and promising direction for autonomous driving systems.
Chinese Translation
自动驾驶需要在语义约束和预测动态下进行规划。然而，现有的端到端驾驶方法通常只强调这一要求的一个方面：视觉-语言-行动（VLA）模型利用视觉-语言模型（VLM）先验进行语义推理，而世界行动模型（WAM）通过生成世界建模提供未来感知的预测。这自然激励了一个统一的规划器，可以同时利用语义先验和预测动态。然而，我们发现通过联合标记级注意力的简单组合存在注意力分配不匹配的问题，其中语义捷径主导了共享的注意力空间并抑制了预测动态。受到神经科学证据的启发，复杂行为源于功能专业化系统之间的协调，我们提出了BrainWAM，一个结构化的行动空间协调框架，将语义推理和预测世界建模转换为两个专业的面向行动的路径，并在紧凑的行动表示层面上对齐它们。我们进一步引入了一种异步修正流推理策略，结合解耦的视频和行动去噪，缩短了推理延迟，同时保留了与规划相关的预测上下文。BrainWAM在NAVSIM v1（89.5 PDMS）和NAVSIM v2（89.6 EPDMS）上达到了最先进的性能，始终优于仅使用VLA或WAM的方法，突显了BrainWAM作为自动驾驶系统的一个实用且有前景的方向。
cs.RO / 9 / 2608.12860
HumanoidVLN: A Physics-Grounded Simulator and Benchmark for Vision-Language Navigation Across Diverse Humanoid Embodiments
HumanoidVLN：一个基于物理的模拟器和基准，用于多样化人形机器人视觉-语言导航
Abstract
Vision-Language Navigation (VLN) for humanoid robots poses challenges existing benchmarks fail to address: bipedal locomotion imposes physical constraints absent from wheeled agents, humanoid morphologies vary across platforms, and egocentric observations are distorted by locomotion-induced camera dynamics. We present HumanoidVLN, a physics-grounded simulator and benchmark for VLN across diverse humanoid embodiments. Built on NVIDIA Isaac Sim, our platform supports an extensible set of humanoid configurations, demonstrated on four robots (Unitree G1, Unitree H1, Internal-A, Internal-B) spanning 10-12 lower-body DoF and heights from 1.17m to 1.80m, via a hierarchical control stack combining a reinforcement learning locomotion policy with interchangeable PD or MPC path trackers. New robots and VLN models integrate with minimal effort; we demonstrate compatibility with NaVILA, DualVLN, StreamVLN, and JanusVLN. Environments are drawn from artist-designed scenes and 3D Gaussian Splatting reconstructions, filtered for navigable areas exceeding 100 square meters. Instructions are generated by a dual generator-reviewer plus paraphraser multi-agent pipeline with human-in-the-loop verification, yielding 933 collision-aware reference episodes, each paired with one fine-grained instruction and three coarse-grained stylistic variants (formal, natural, casual). Across four models and four embodiments, JanusVLN achieves the highest mean success rate of 43.55% and nDTW of 48.38. In a 20-episode sim-to-real pilot with DualVLN and the Unitree G1, navigation errors correlate strongly (r=0.935), with a mean absolute difference of 0.68m and mean trajectory similarity of 0.782 (+/-0.188) nDTW. These results highlight the interaction between VLN models, controllers, and humanoid embodiments under physical execution. Code, benchmark, and data will be released upon acceptance at https://humanoid-vln.github.io/.
Chinese Translation
人形机器人视觉-语言导航（VLN）面临现有基准未能解决的挑战：双足行走带来了轮式代理所没有的物理约束，人形形态在不同平台之间存在差异，且自我中心的观察受到行走引起的摄像机动态的扭曲。我们提出了HumanoidVLN，这是一个基于物理的模拟器和基准，用于多样化人形机器人中的VLN。该平台基于NVIDIA Isaac Sim构建，支持一系列可扩展的人形配置，展示了四种机器人（Unitree G1、Unitree H1、Internal-A、Internal-B），其下肢自由度（DoF）范围为10-12，身高从1.17米到1.80米不等，采用结合强化学习行走策略与可互换的PD或MPC路径跟踪器的分层控制架构。新机器人和VLN模型的集成工作量极小；我们展示了与NaVILA、DualVLN、StreamVLN和JanusVLN的兼容性。环境来自艺术家设计的场景和3D高斯点云重建，经过筛选以确保可导航区域超过100平方米。指令由一个双生成器-审查者加上改写者的多代理管道生成，并经过人类验证，产生933个考虑碰撞的参考情节，每个情节配有一条细致的指令和三种粗略的风格变体（正式、自然、休闲）。在四个模型和四种形态中，JanusVLN实现了最高的平均成功率43.55%和nDTW为48.38。在与DualVLN和Unitree G1进行的20集模拟到现实的试点中，导航误差与实际表现高度相关（r=0.935），平均绝对差为0.68米，平均轨迹相似度为0.782（+/-0.188）nDTW。这些结果突显了在物理执行下，VLN模型、控制器和人形形态之间的相互作用。代码、基准和数据将在接受后发布，网址为https://humanoid-vln.github.io/。
cs.RO / 10 / 2608.12866
AMR-Pose: An Active LED Marker-Based Relative Pose Estimation Framework With Probabilistic Switching PnP for Cooperative AUVs
AMR-Pose：基于主动LED标记的相对姿态估计框架，结合概率切换PnP用于协作自主水下机器人
Abstract
Reliable relative pose estimation between autonomous underwater vehicles (AUVs) is critical for cooperative ocean exploration, sampling, and multi-robot coordination. However, achieving robust vision-based relative localization in underwater environments remains challenging due to severe optical degradation, including turbidity, illumination variations, reflections, and intermittent feature occlusions. This paper presents AMR-Pose, an active LED marker-based relative pose estimation framework for cooperative AUVs. A compact marker module consisting of one red central LED and three blue peripheral LEDs is developed and integrated onto the leader AUV to provide distinctive visual features under complex underwater conditions. Building upon the detected marker observations, a probabilistic switching Perspective-n-Point estimator (PSwPnP) is developed by combining Lie-group pose propagation on $SE(3)$, probabilistic marker association, and visibility-adaptive measurement fusion for robust six-degree-of-freedom relative pose estimation. The proposed framework dynamically adapts the estimation process according to marker visibility, maintaining geometric consistency and temporal stability during partial observations and visibility transitions. Extensive water-tank experiments with motion-capture ground truth validate that AMR-Pose achieves accurate, smooth, and robust relative pose estimation under challenging underwater conditions. Closed-loop leader-follower experiments further demonstrate its feasibility for real-time relative pose feedback in cooperative underwater robotics.
Chinese Translation
在自主水下机器人（AUVs）之间可靠的相对姿态估计对于协作海洋探索、取样和多机器人协调至关重要。然而，由于水下环境中严重的光学退化，包括浑浊、光照变化、反射和间歇性特征遮挡，实现基于视觉的稳健相对定位仍然具有挑战性。本文提出了AMR-Pose，一个基于主动LED标记的相对姿态估计框架，旨在服务于协作AUVs。我们开发并集成了一个紧凑的标记模块，该模块由一个红色中央LED和三个蓝色外围LED组成，以在复杂的水下条件下提供独特的视觉特征。在检测到的标记观测基础上，结合李群姿态传播（Lie-group pose propagation）在 $SE(3)$ 上、概率标记关联和可见性自适应测量融合，开发了一种概率切换透视-n-点估计器（PSwPnP），用于稳健的六自由度相对姿态估计。所提出的框架根据标记的可见性动态调整估计过程，在部分观测和可见性转换期间保持几何一致性和时间稳定性。通过大量水池实验与运动捕捉真实值的对比验证，AMR-Pose在具有挑战性的水下条件下实现了准确、平滑和稳健的相对姿态估计。闭环领导-跟随实验进一步证明了其在协作水下机器人中实时相对姿态反馈的可行性。
cs.RO / 11 / 2608.13026
Temporal GRPO: Beyond Trajectory-Level Credit in Vision-Language-Action Reinforcement Learning
时间性GRPO：超越轨迹级信用的视觉-语言-动作强化学习
Abstract
Outcome-driven reinforcement learning offers a scalable way to post-train vision-language-action (VLA) policies from sparse task-success feedback. In common GRPO-based VLA post-training, one rollout-level advantage is applied to every action in the trajectory. A rollout that completes several valid stages but fails later can therefore penalize the actions that produced its earlier progress. We call this trajectory-level credit aliasing. Temporal GRPO addresses this problem by constructing detectable task stages, aligning each rollout with stage-specific action intervals, and comparing only rollouts that have entered the same stage. The resulting stage advantages are applied to their corresponding intervals in a single policy update. On RoboTwin 2.0, Temporal GRPO improves task success and sample efficiency, with consistent gains across task horizons. Controlled updates on LIBERO-Long preserve shared prerequisite stages and concentrate improvement at the first stage where rollout outcomes diverge.
Chinese Translation
以结果为驱动的强化学习提供了一种可扩展的方法，从稀疏的任务成功反馈中对视觉-语言-动作（VLA）策略进行后训练。在常见的基于GRPO的VLA后训练中，一个回滚级别的优势被应用于轨迹中的每个动作。因此，一个完成多个有效阶段但后期失败的回滚可能会惩罚那些产生早期进展的动作。我们称之为轨迹级信用别名。时间性GRPO通过构建可检测的任务阶段来解决这个问题，将每个回滚与特定阶段的动作间隔对齐，并仅比较已进入相同阶段的回滚。由此产生的阶段优势被应用于单次策略更新中的相应间隔。在RoboTwin 2.0上，时间性GRPO提高了任务成功率和样本效率，并在任务时间范围内实现了一致的增益。在LIBERO-Long上，受控更新保留了共享的前提阶段，并集中改善在回滚结果分歧的第一个阶段。
cs.RO / 12 / 2608.13049
H2R-Bench: Benchmarking Human-to-Robot Manipulation Video Generation in World Models
H2R-Bench：在世界模型中评估人机操作视频生成的基准
Abstract
Large-scale manipulation data is essential for robot learning, yet collecting robot demonstrations remains expensive and difficult to scale. Meanwhile, abundant egocentric human manipulation videos provide rich behavioral experiences, but transferring them across embodiments remains challenging due to differences between human hands and robotic end-effectors. Recent advances in video world models offer a promising pathway to synthesize robot-centric manipulation videos from human observations, while their cross-embodiment transfer capability remains largely unexplored. Therefore, we introduce H2R-Bench, a benchmark for evaluating cross-embodiment human-to-robot manipulation video generation, where models transform egocentric human demonstrations into robot manipulation videos under specified embodiments. Each benchmark instance contains a human demonstration video, target embodiment constraints, and source-grounded annotations covering task goals, action events, functional contacts, and object responses. H2R-Bench evaluates generated videos through five dimensions, including goal-state completion, action-event completion, functional contact transfer, embodiment correctness, and general video quality. We benchmark eleven state-of-the-art video generation models across six manipulation families and two robot embodiments. Our evaluation reveals that current video world models remain limited in human-to-robot manipulation transfer: even leading models often fail in embodiment consistency, functional interaction, and task execution. H2R-Bench provides a systematic diagnostic framework for evaluating whether video world models can bridge the human-to-robot embodiment gap and convert human manipulation observations into robot-centric training resources.
Chinese Translation
大规模的操作数据对机器人学习至关重要，但收集机器人演示仍然成本高昂且难以扩展。与此同时，丰富的自我中心人类操作视频提供了丰富的行为经验，但由于人类手和机器人末端执行器之间的差异，跨体现的转移仍然具有挑战性。最近在视频世界模型方面的进展为从人类观察中合成以机器人为中心的操作视频提供了一条有希望的途径，但其跨体现转移能力仍然在很大程度上未被探索。因此，我们引入了H2R-Bench，这是一个用于评估跨体现人机操作视频生成的基准，其中模型在指定的体现下将自我中心的人类演示转换为机器人操作视频。每个基准实例包含一个人类演示视频、目标体现约束以及涵盖任务目标、动作事件、功能接触和物体响应的源基础注释。H2R-Bench通过五个维度评估生成的视频，包括目标状态完成、动作事件完成、功能接触转移、体现正确性和一般视频质量。我们在六个操作类别和两个机器人体现中基准测试了十一种最先进的视频生成模型。我们的评估揭示了当前的视频世界模型在跨体现人机操作转移方面的局限性：即使是领先的模型在体现一致性、功能交互和任务执行方面也常常失败。H2R-Bench提供了一个系统的诊断框架，用于评估视频世界模型是否能够弥合人机体现之间的差距，并将人类操作观察转换为以机器人为中心的训练资源。
cs.RO / 13 / 2608.13095
Semantic Radiance Fields as Simulators for Spatial Reasoning in Real-World Scenes
语义辐射场作为现实场景空间推理的模拟器
Abstract
Training and evaluating spatial reasoning in embodied agents requires diverse environments that are both geometrically faithful and semantically queryable. Synthetic simulators offer ground truth semantics but sacrifice realism; simulators based on reconstructions of real-world environments have realistic appearance but lack ground truth semantics by default. We propose using Semantic Radiance Fields (SRF) as simulators for spatial reasoning agents. SRFs are a representation that unifies these requirements by lifting 2D semantic segmentations from pretrained vision models into a 3D radiance field that jointly encodes geometry, appearance, and per-class semantic identity. The resulting fields are reconstructed from posed RGB captures of real scenes and support novel-view synthesis, semantic and free-space queries within a single grounded representation. This enables the efficient generation of diverse real-world environments to train and evaluate spatial reasoning models. As an example application, we outline an SRF-driven simulator for an orchard apple-reaching task, in which the radiance field supplies camera rendering, semantic ground truth, and occupancy queries to a physics engine.
Chinese Translation
训练和评估具身智能体的空间推理需要多样化的环境，这些环境既要在几何上真实可信，又要在语义上可查询。合成模拟器提供了真实语义，但牺牲了现实感；基于真实环境重建的模拟器具有真实的外观，但默认缺乏真实语义。我们提出使用语义辐射场（Semantic Radiance Fields, SRF）作为空间推理智能体的模拟器。SRF是一种统一这些要求的表示方法，通过将预训练视觉模型的二维语义分割提升到一个三维辐射场中，联合编码几何、外观和每类的语义身份。所得到的场景是通过对真实场景进行RGB捕捉而重建的，并支持在单一的基础表示中进行新视角合成、语义查询和自由空间查询。这使得高效生成多样化的真实世界环境成为可能，以训练和评估空间推理模型。作为一个示例应用，我们概述了一个用于果园苹果采摘任务的SRF驱动模拟器，其中辐射场为物理引擎提供了相机渲染、语义真实值和占用查询。
cs.RO / 14 / 2608.13103
S2-HWM: Sparse Event-Structured Hierarchical World Model for Long-Horizon Surgical Robot Manipulation
S2-HWM：用于长时间跨度外科机器人操作的稀疏事件结构层次世界模型
Abstract
Long-horizon surgical robot manipulation is challenging because task rewards are sparse, while meaningful interaction changes occur at irregular intervals. Existing world-model agents typically imagine at primitive-step resolution, leaving variable-duration task progress implicit. Manually specified stages can provide intermediate structure, but their task specific boundaries are difficult to align with state-dependent interaction transitions. We propose S2-HWM, a Sparse Event-Structured Hierarchical World Model that learns sparse event evidence from primitive latent trajectories to coordinate an event-level manager and a primitive-step worker. The event evidence schedules manager goal updates, and each selected latent goal conditions the worker's primitive actions until the next update. The learned event evidence also forms variable-duration segments for an Event Transition Model (ETM), which predicts the next?boundary stochastic state, segment duration, and accumulated segment reward. Chaining these event-level predictions provides a variable-duration continuation beyond the primitive imagination horizon for manager learning, while the worker retains primitive-step actor-critic learning. On a SurRoL-based PegTransfer task, S2-HWM achieves a success rate of 98.7%, outperforming the flat GAS DreamerV3 baseline by 22.7 percentage points.
Chinese Translation
长时间跨度的外科机器人操作具有挑战性，因为任务奖励稀疏，而有意义的交互变化发生在不规则的时间间隔。现有的世界模型代理通常以原始步骤分辨率进行想象，导致可变持续时间的任务进展隐含。手动指定的阶段可以提供中间结构，但其任务特定的边界难以与状态依赖的交互转变对齐。我们提出了S2-HWM，一种稀疏事件结构层次世界模型，它从原始潜在轨迹中学习稀疏事件证据，以协调事件级管理器和原始步骤工作者。事件证据调度管理器目标更新，每个选定的潜在目标条件化工作者的原始动作，直到下一个更新。学习到的事件证据还为事件转移模型（Event Transition Model, ETM）形成可变持续时间的片段，该模型预测下一个边界随机状态、片段持续时间和累计片段奖励。将这些事件级预测串联起来，为管理器学习提供了超出原始想象视野的可变持续时间延续，而工作者则保留原始步骤的演员-评论家学习。在基于SurRoL的PegTransfer任务中，S2-HWM实现了98.7%的成功率，超过了平面GAS DreamerV3基线22.7个百分点。
cs.RO / 15 / 2608.13220
FAM-DQ: A Dual-Quadrotor-Based Fully Actuated Aerial Manipulator for High-Torque Interaction
FAM-DQ：一种基于双四旋翼的全驱动空中操控器，用于高扭矩交互
Abstract
Aerial physical interaction requires aerial manipulation platforms to generate large interaction forces and torques while maintaining precise end-effector control. However, conventional underactuated aerial manipulators suffer from strong position-attitude coupling, whereas fully actuated platform designs often face structural complexity, limited payload capacity, and insufficient torque output. This paper presents FAM-DQ, a dual-quadrotor based fully actuated aerial manipulator designed for high-torque physical interaction tasks. By mounting two quadrotor propulsion modules at the ends of a central frame through passive joints, while using a gear-driven servo to regulate the pointing direction, FAM-DQ achieves decoupled $6$-DoF end-effector control with omnidirectional manipulation capability and enhanced torque output. Experiments including trajectory tracking, attitude tracking, static torque measurement, and screw driving validate the proposed design. FAM-DQ achieves a maximum torque of $1.019~\mathrm{N}\cdot\mathrm{m}$ with a total mass of $0.447~\mathrm{kg}$, corresponding to a torque-to-mass ratio of $2.28~\mathrm{N}\cdot\mathrm{m/kg}$.
Chinese Translation
空中物理交互要求空中操控平台在保持精确的末端执行器控制的同时，产生较大的交互力和扭矩。然而，传统的欠驱动空中操控器存在强位置-姿态耦合的问题，而全驱动平台设计往往面临结构复杂、有效载荷能力有限和扭矩输出不足等挑战。本文提出了FAM-DQ，一种基于双四旋翼的全驱动空中操控器，旨在应对高扭矩物理交互任务。通过在中央框架的两端通过被动关节安装两个四旋翼推进模块，并使用齿轮驱动伺服器调节指向方向，FAM-DQ实现了解耦的6自由度末端执行器控制，具备全向操控能力和增强的扭矩输出。包括轨迹跟踪、姿态跟踪、静态扭矩测量和螺钉驱动的实验验证了所提出的设计。FAM-DQ在总质量为0.447 kg的情况下，达到了最大扭矩1.019 N·m，对应的扭矩与质量比为2.28 N·m/kg。
cs.RO / 16 / 2608.13233
Manufacturing Complex Airtight Soft Pneumatic Actuators for Soft Robotics: Process Evaluation and Optimization
制造复杂气密软气动执行器用于软机器人：工艺评估与优化
Abstract
Manufacturing complex soft pneumatic actuators remains challenging because geometric fidelity, compliance, structural integrity, and airtightness must be achieved simultaneously. This study presents a manufacturing-focused evaluation of several fabrication routes for complex pneumatic structures, including heat-shrink forming, silicone casting, powder- and liquid-based additive manufacturing, and fused deposition modeling (FDM). The processes were assessed through process screening, baseline fabrication, failure analysis, and process improvement to distinguish inherent process limitations from correctable manufacturing defects. Heat-shrink forming was limited by geometric conformity, casting by mold accessibility and bonded interfaces, powder-based methods by residual material trapped within enclosed passages, and digital light processing by the material properties and post-processing requirements of the investigated system. FDM provided the most adaptable route because its dominant defects could be progressively reduced through process optimization. The results further showed that airtightness depends not only on nominal wall thickness but also on extrusion-path architecture, while support-free geometry is important when access for internal post-processing is limited. These findings establish a practical design-for-manufacturing approach in which process selection is guided by the compatibility between actuator architecture and manufacturing constraints. The proposed approach provides practical guidance for developing complex, flexible, and airtight soft pneumatic actuators for soft robotic applications
Chinese Translation
制造复杂的软气动执行器仍然面临挑战，因为几何精度、柔顺性、结构完整性和气密性必须同时实现。本研究针对复杂气动结构的几种制造路线进行了以制造为重点的评估，包括热收缩成型、硅胶铸造、基于粉末和液体的增材制造以及熔融沉积建模（FDM）。通过工艺筛选、基线制造、故障分析和工艺改进对这些过程进行了评估，以区分固有的工艺限制与可纠正的制造缺陷。热收缩成型受到几何一致性的限制，铸造受到模具可及性和粘接界面的影响，基于粉末的方法受到封闭通道内残留材料的困扰，而数字光处理则受到所研究系统的材料特性和后处理要求的限制。FDM提供了最具适应性的路线，因为其主要缺陷可以通过工艺优化逐步减少。结果进一步表明，气密性不仅依赖于名义壁厚，还与挤出路径结构有关，而在内部后处理受限时，无支撑几何形状则显得尤为重要。这些发现建立了一种实用的设计制造方法，其中工艺选择受执行器结构与制造约束之间兼容性的指导。所提出的方法为开发复杂、灵活且气密的软气动执行器在软机器人应用中提供了实用指导。
cs.RO / 17 / 2608.13284
Predictive Relative-Velocity Steering for Safe Robotic Manipulator Teleoperation in Dynamic Environments
动态环境中安全机器人操控远程操作的预测相对速度引导
Abstract
Recent advances in teleoperation have enabled robotic manipulators to perform dexterous, human-arm-like motions. However, human operators may fail to avoid suddenly appearing obstacles promptly and effectively, particularly under network latency or limited attention, thereby creating safety risks. To address this issue, we propose a lightweight and modular framework for proactive collision avoidance, operating directly at the end-effector velocity-command level. After preprocessing the point cloud, the framework first predicts potential collisions based on time-to-collision (TTC) with integrated overshoot protection, and subsequently rotates the relative-velocity vector using Rodrigues' rotation formula. The deflection changes only the direction of the relative velocity while preserving its magnitude, thereby mitigating the deadlock problem commonly encountered by conventional artificial potential field (APF) methods. The prediction module compensates for point-cloud processing latency introduced by complex teleoperation pipelines, while the lightweight design enables the high-frequency control required for teleoperation. Simulations across diverse scenarios show that the proposed method achieves a higher end-effector collision avoidance rate than the baseline methods. Experiments on a physical robotic system further validate its collision-avoidance effectiveness.
Chinese Translation
近年来，远程操作的进展使得机器人操纵器能够执行灵巧的、类似人手的动作。然而，人类操作员在网络延迟或注意力有限的情况下，可能无法及时有效地避免突然出现的障碍物，从而带来安全风险。为了解决这一问题，我们提出了一种轻量级和模块化的框架，用于主动避碰，直接在末端执行器速度指令级别操作。在对点云进行预处理后，该框架首先基于碰撞时间（TTC）预测潜在碰撞，并集成过冲保护，随后使用罗德里格斯旋转公式旋转相对速度向量。偏转仅改变相对速度的方向，同时保持其大小，从而减轻传统人工势场（APF）方法常见的死锁问题。预测模块补偿了复杂远程操作管道引入的点云处理延迟，而轻量级设计则支持远程操作所需的高频控制。在多种场景下的仿真表明，所提方法实现了比基线方法更高的末端执行器避碰率。在物理机器人系统上的实验进一步验证了其避碰效果。
cs.RO / 18 / 2608.13362
NestDex: Nested Policy Learning with Copilot Assisted Teleoperation for Dexterous Manipulation
NestDex：基于副驾驶辅助遥操作的嵌套策略学习用于灵巧操作
Abstract
Dexterous manipulation promises substantially richer robot interaction with the physical world, but learning these behaviours remains constrained by the difficulty of collecting consistent, complete-task demonstrations. Unlike parallel-jaw manipulation, dexterous tasks require the operator to coordinate arm motion with precise, contact-rich finger behaviour throughout the task. We introduce NestDex, a nested policy-learning framework that reduces this burden by using learned hand skills to assist demonstration collection. The operator controls the arm and regulates the active hand skill through a single-DoF clutch, rather than directly specifying the full finger trajectory. The inner hand policy adapts its motion from the latest proprioceptive history, while a vision-language selector activates the appropriate skill for each task stage. The resulting demonstrations train a separate outer visuomotor policy that controls both the arm and hand without the inner policies at deployment. A hand-action variational autoencoder provides compact hand-action targets while retaining arm commands in joint space. Across real-world dexterous manipulation experiments, NestDex improves demonstration reliability and efficiency, and the resulting empirical evaluations support effective autonomous policy learning. Video Demo are available at project website https://aus.bot/research/nestdex.
Chinese Translation
灵巧操作承诺实现机器人与物理世界之间更丰富的交互，但学习这些行为仍受到收集一致、完整任务演示的困难限制。与平行夹具操作不同，灵巧任务要求操作员在整个任务过程中协调手臂运动与精确、接触丰富的手指行为。我们提出了NestDex，一个嵌套策略学习框架，通过使用学习到的手部技能来辅助演示收集，从而减轻这一负担。操作员通过单自由度离合器控制手臂，并调节主动手部技能，而不是直接指定完整的手指轨迹。内部手部策略根据最新的本体感知历史调整其运动，同时视觉-语言选择器为每个任务阶段激活适当的技能。最终的演示训练一个独立的外部视觉运动策略，该策略在部署时控制手臂和手，而不依赖内部策略。手部动作变分自编码器提供紧凑的手部动作目标，同时在关节空间中保留手臂指令。在真实世界的灵巧操作实验中，NestDex提高了演示的可靠性和效率，结果的实证评估支持有效的自主策略学习。视频演示可在项目网站 https://aus.bot/research/nestdex 上查看。
cs.RO / 19 / 2608.13395
FIRE-VLA: Failure-Informed Self-Evolution for Vision-Language-Action Models in Autonomous Driving
FIRE-VLA：面向失败的自我演化框架在自动驾驶中的视觉-语言-动作模型
Abstract
Reinforcement learning improves autonomous-driving vision-language-action (VLA) models by evaluating trajectories sampled from the current policy. Group relative policy optimization (GRPO) learns from reward differences within each rollout group. When all sampled trajectories are poor, this relative signal can rank failures without identifying behavior outside the failed region. We introduce FIRE-VLA, a failure-informed self-evolution framework that converts such unresolved failures into privileged supervision for the next policy. Low-reward, low-diversity groups trigger self-distillation from a frozen round-start copy of the same model. Teacher and student have the same parameter scale, but only the teacher observes the hidden future trajectory. Supervision follows the student's generated prefix and is restricted to answer tokens, while GRPO remains active for every group. The updated policy supplies the teacher for the next round, allowing the routed failure distribution to change with the policy without requiring a larger external teacher. Starting from the same Qwen2.5-VL-3B SFT checkpoint, the comparison matches student rollout and policy-update counts. On 6,019 examples from 150 held-out nuScenes scenes, FIRE-VLA retains comparable single-sample planning, reduces G=4 mean L2 from 1.848 to 1.500 m, and lowers evaluation-persistent failure prevalence from 13.03% to 11.20%. The reduction in mean error arises mainly from rare severe rollouts rather than uniform improvement across ordinary trajectories.
Chinese Translation
强化学习通过评估当前策略下采样的轨迹来改善自动驾驶视觉-语言-动作（VLA）模型。组相对策略优化（GRPO）从每个回合组内的奖励差异中学习。当所有采样的轨迹表现不佳时，这种相对信号可以对失败进行排名，但无法识别失败区域外的行为。我们提出了FIRE-VLA，一种面向失败的自我演化框架，将这些未解决的失败转化为下一策略的特权监督。低奖励、低多样性组触发从同一模型的冻结回合开始副本进行自我蒸馏。教师和学生具有相同的参数规模，但只有教师观察隐藏的未来轨迹。监督遵循学生生成的前缀，并限制在回答标记上，同时GRPO对每个组保持活跃。更新后的策略为下一轮提供教师，使得路由的失败分布能够随着策略的变化而变化，而无需更大的外部教师。从相同的Qwen2.5-VL-3B SFT检查点开始，比较匹配学生的回合和策略更新计数。在来自150个保留nuScenes场景的6,019个示例中，FIRE-VLA保持了可比的单样本规划，将G=4的平均L2从1.848降低到1.500米，并将评估持久性失败的发生率从13.03%降低到11.20%。平均误差的减少主要源于罕见的严重回合，而不是普通轨迹的均匀改善。
cs.RO / 20 / 2608.13396
Capstan-driven Continuum Surgical Robot: Design, Modeling, and Perception
绞盘驱动的连续性外科机器人：设计、建模与感知
Abstract
Shape and force sensing have long been critical bottlenecks in the development of compact capstan-driven continuum surgical robots, primarily due to the difficulty of obtaining cable tension information within the confined capstan assembly. To overcome these challenges, this paper presents an integrated design-modeling-sensing approach based on the concept of actuation-perception co-design. A compliant element is introduced into the motor mounting bracket of the drive system, enabling micro-deformation under the cable reaction force and thereby allowing real-time cable tension measurement without occupying the compact capstan space. To address the modeling complexity arising from unconventional joint configurations introduced by the spatial cable routing strategy, a parallel computation framework based on a multibody short-thick-beam model is proposed, which captures shear effects in short beam segments and synergistic multi-cable interactions while achieving real-time performance. Building on this framework, stable shape and force sensing is achieved by incorporating a proximal multi-axis force/torque sensor as an additional measurement anchor. Following this design-modeling-sensing framework, capstan-driven continuum surgical robots with single- and dual-segment configurations are developed. Experimental results validate the proposed framework in both single- and dual-segment continuum robots, demonstrating real-time tip pose estimation together with contact force and location perception. By enabling cable tension feedback without compromising the compact capstan architecture, the proposed framework makes integrated perception feasible for capstan-driven continuum surgical robots.
Chinese Translation
形状和力的感知长期以来一直是紧凑型绞盘驱动连续性外科机器人发展的关键瓶颈，主要是由于在受限的绞盘组件内获取电缆张力信息的困难。为了解决这些挑战，本文提出了一种基于驱动-感知协同设计概念的集成设计-建模-感知方法。在驱动系统的电机安装支架中引入了一种柔性元件，使其能够在电缆反作用力下微变形，从而实现实时电缆张力测量，而不占用紧凑的绞盘空间。为了应对由空间电缆布线策略引入的非常规关节配置所带来的建模复杂性，提出了一种基于多体短粗梁模型的并行计算框架，该框架能够捕捉短梁段中的剪切效应和多电缆之间的协同作用，同时实现实时性能。在此框架的基础上，通过将近端多轴力/扭矩传感器作为额外的测量锚点，实现了稳定的形状和力的感知。遵循这一设计-建模-感知框架，开发了具有单段和双段配置的绞盘驱动连续性外科机器人。实验结果验证了所提出框架在单段和双段连续性机器人中的有效性，展示了实时的末端姿态估计以及接触力和位置感知。通过实现电缆张力反馈而不妨碍紧凑的绞盘结构，所提出的框架使得绞盘驱动的连续性外科机器人能够实现集成感知。
cs.RO / 21 / 2608.13415
Deliberate Practice: Learning Robot Skills under a Budget
有意识的练习：在预算限制下学习机器人技能
Abstract
We consider the problem of autonomously learning robot skills under a limited practice budget for sequential tasks. We propose an active skill learning algorithm, \emph{Deliberate Practice (DP)}, that computes a provably \emph{budget-optimal} allocation---practicing skills that maximize expected cumulative reward while being learnable within the budget. DP estimates both the time needed to master skills and the cumulative reward of the task plans that the skills unlock. Computing a budget-optimal allocation is challenging as it requires reasoning about combinatorially many skill plans over a large practice budget. Our key contribution is a bilinear program that can compute this exactly using off-the-shelf solvers. Through simulated and real-world experiments on long-horizon manipulation tasks, we show that our approach allows robots to optimally use limited practice time to acquire useful policies and improve long-horizon planning.
Chinese Translation
我们考虑在有限的练习预算下，自主学习顺序任务的机器人技能的问题。我们提出了一种主动技能学习算法—— extit{Deliberate Practice (DP)}，该算法计算出一个可证明的 extit{预算最优}分配——练习那些在预算内可学习并且能够最大化预期累积奖励的技能。DP估计掌握技能所需的时间以及技能解锁的任务计划的累积奖励。计算预算最优分配是具有挑战性的，因为它需要对大量技能计划进行组合推理，涉及到较大的练习预算。我们的关键贡献是一个双线性规划，可以利用现成的求解器精确计算这一分配。通过在长时间范围的操作任务上进行的模拟和现实世界实验，我们展示了我们的方法使机器人能够在有限的练习时间内最佳利用，以获取有用的策略并改善长时间范围的规划。
cs.RO / 22 / 2608.13422
Attention from Action, for Action: Emergent Visual Bottlenecks for Policy Learning
来自行动的注意力，为行动服务：政策学习中的新兴视觉瓶颈
Abstract
Visual bottlenecks that focus policy inputs on regions of interest (ROIs) can improve data-efficient visuomotor learning by separating where to look from how to act. Many ROI interfaces rely on external spatial labels, such as gaze, object classes, or affordance annotations. Label-free alternatives often derive crops from trajectories by detecting gripper or motion events and centering a fixed crop at the projected end-effector. Such action-derived crops are useful spatial priors that require no additional labels, but they encode fixed choices about event timing, proxy points, and crop scale. When the visual evidence needed for control lies away from the end-effector or changes continuously with task progress, these crops can become misaligned. We propose Seeker, a task- and state-conditioned readout that learns attention from action. Starting from frozen DINOv3 features, Seeker iteratively updates a query with gathered visual evidence, producing progression-aware ROIs solely from action supervision. The learned ROI serves as a spatial interface for RGB cropping, mask-guided background augmentation, and point-cloud filtering. In simulation and the real world, Seeker improves data efficiency and robustness over no-crop, augmentation, and action-derived crop baselines. On real robots, Seeker raises average in-domain success from the best baseline's 48.3% to 76.7% and success under lighting/background shifts from 20.0% to 60.0%.
Chinese Translation
视觉瓶颈通过将政策输入聚焦于感兴趣区域（ROIs），可以通过将注意力与行动分离来提高数据高效的视觉运动学习。许多ROIs接口依赖于外部空间标签，如注视、物体类别或可供性注释。无标签的替代方案通常通过检测抓取器或运动事件，从轨迹中提取图像，并在投影的末端效应器处居中固定裁剪。这种基于行动的裁剪是有用的空间先验，不需要额外的标签，但它们编码了关于事件时机、代理点和裁剪尺度的固定选择。当控制所需的视觉证据远离末端效应器或随着任务进展而不断变化时，这些裁剪可能会变得不对齐。我们提出了Seeker，一种任务和状态条件的读取方式，它从行动中学习注意力。从冻结的DINOv3特征开始，Seeker通过收集的视觉证据迭代更新查询，仅通过行动监督生成进度感知的ROIs。学习到的ROI作为RGB裁剪、掩膜引导的背景增强和点云过滤的空间接口。在模拟和现实世界中，Seeker在数据效率和鲁棒性方面优于无裁剪、增强和基于行动的裁剪基线。在真实机器人上，Seeker将最佳基线的平均领域内成功率从48.3%提高到76.7%，在光照/背景变化下的成功率从20.0%提高到60.0%。
cs.RO / 23 / 2608.13438
ContactGuard: Pre-Contact Execution Monitoring with Action-Conditioned Latent World Models
ContactGuard：基于动作条件的潜在世界模型的接触前执行监控
Abstract
Contact-rich manipulation failures are often detected only after the robot has committed to contact. This is especially limiting in wrist-camera setups: close gripper--object views help observe contact, but a poor approach may already push, miss, slip, or disturb the object before conventional detectors react. We introduce \emph{ContactGuard}, a pre-contact execution monitor for chunked visuomotor policies. Given the policy's planned action chunk, ContactGuard predicts its short-horizon consequence in latent visual space and aborts if the predicted future latent indicates likely failure. Its latent world model is trained from unlabelled robot trajectories to predict compact multi-view visual embeddings under planned actions, avoiding pixel-level video prediction. A lightweight failure probe is then trained from a small labelled set of pre-contact clips. At deployment, ContactGuard anchors prediction before an imminent contact event, rolls the model forward under the policy's own actions, and verifies the predicted post-contact latent. Across real-world contact-rich manipulation tasks, ContactGuard predicts failure more accurately than direct and corrupted-action ablations, and transfers to live robot as a pre-contact abort signal without modifying the underlying policy.
Chinese Translation
接触丰富的操作失败通常仅在机器人已经接触后才被检测到。这在腕部摄像头设置中尤其受限：靠近夹持器与物体的视角有助于观察接触，但不良的接近可能在传统检测器反应之前就已经推动、错过、滑动或干扰了物体。我们提出了 extit{ContactGuard}，一种针对分块视觉运动策略的接触前执行监控器。给定策略的计划动作块，ContactGuard 在潜在视觉空间中预测其短期后果，并在预测的未来潜在指示可能失败时中止执行。其潜在世界模型通过无标签的机器人轨迹训练，以预测在计划动作下的紧凑多视角视觉嵌入，避免了像素级视频预测。随后，从一小组带标签的接触前剪辑中训练出轻量级失败探测器。在部署时，ContactGuard 在即将发生接触事件之前锚定预测，依据策略自身的动作向前推进模型，并验证预测的接触后潜在。在现实世界的接触丰富操作任务中，ContactGuard 比直接和损坏动作的消融实验更准确地预测失败，并作为接触前中止信号转移到实时机器人，而无需修改基础策略。
cs.RO / 24 / 2608.13448
Mind the Context: Continual Learning of Socially Appropriate Robot Actions via Environmental-Social Disentanglement
关注上下文：通过环境-社会解耦实现社会适宜机器人行为的持续学习
Abstract
Social robots are expected to operate across diverse environments, where similar arrangements can imply different socially appropriate actions, e.g., starting a conversation may be acceptable in a crowded home but disruptive in an office meeting. Because such norms and environments cannot all be anticipated in advance, robots require continual learning (CL) to adapt from sequential experience while retaining previously acquired knowledge. Prior work has studied CL for generating socially appropriate robot actions, but it has not addressed domain-incremental settings in which the robot incrementally encounters diverse contexts (e.g., living room, meeting room, office, hallway), where both environmental (e.g., whether the space is open or cluttered with furniture) and social cues (e.g., how people or other agents are positioned around the robot) jointly shape the appropriateness of robot actions. We address this gap with the Explicit Disentanglement Dual-Branch (EDD) framework. EDD explicitly separates environmental and social-agent related knowledge and uses replay-based rehearsal to mitigate forgetting while learning the appropriateness of robot actions (e.g., cleaning, serving, starting a conversation) across several indoor domains. Experiments show that EDD outperforms several state-of-the-art baselines, and ablation studies further evaluate different disentanglement strategies and the sensitivity to domain ordering. Our code is publicly available at https://github.com/Cambridge-AFAR/Mind-the-Context.git.
Chinese Translation
社会机器人预计将在多样化的环境中运作，其中相似的环境布局可能暗示不同的社会适宜行为，例如，在拥挤的家庭中开始对话可能是可以接受的，但在办公室会议中则可能是干扰性的。由于这些规范和环境无法全部提前预见，机器人需要通过持续学习（Continual Learning, CL）从顺序经验中适应，同时保留先前获得的知识。之前的研究已经探讨了生成社会适宜机器人行为的持续学习，但尚未解决领域增量设置的问题，在这种设置中，机器人逐步遇到多样的上下文（例如，客厅、会议室、办公室、走廊），在这些上下文中，环境线索（例如，空间是开放的还是被家具杂乱占据的）和社会线索（例如，人或其他代理在机器人周围的定位）共同影响机器人行为的适宜性。我们通过显式解耦双分支（Explicit Disentanglement Dual-Branch, EDD）框架来填补这一空白。EDD显式分离环境和社会代理相关知识，并利用基于重放的排练来减轻遗忘，同时学习多个室内领域中机器人行为的适宜性（例如，清洁、服务、开始对话）。实验表明，EDD的表现优于多个最先进的基线，消融研究进一步评估了不同的解耦策略及其对领域顺序的敏感性。我们的代码已公开发布在 https://github.com/Cambridge-AFAR/Mind-the-Context.git。
cs.RO / 25 / 2608.13474
Decoding Task Progress from VLA Representations
从 VLA 表示中解码任务进展
Abstract
Vision-language-action models (VLAs) are moving rapidly towards deployment as general-purpose manipulation policies, but we currently lack basic tools for understanding what these models represent internally or for monitoring them at runtime. Leveraging ideas from mechanistic interpretability, we probe the residual stream of $\pi_{0.5}$ and find that task progress, the normalized time remaining in a trajectory, is linearly readable from the activations. We find that this signal is present in the pretrained PaliGemma backbone prior to training on any robot-specific data. A single linear probe generalizes to unseen tasks and varies under language counterfactuals when trained on multi-prompt data, but does not enable meaningful steering of the policy. These properties make the signal directly useful for instrumenting deployed VLAs. We use the probe as a simple label-free OOD detector, which detects stalled task progress, and find it competitive with state-of-the-art methods. Our results suggest that VLAs have rich, linearly readable internal representations of semantic quantities like task progress, and that learning to read these signals offers a lightweight, interpretable path toward monitoring deployed visuomotor policies.
Chinese Translation
视觉-语言-行动模型（VLAs）正迅速向作为通用操作策略的部署迈进，但我们目前缺乏基本工具来理解这些模型内部的表示或在运行时监控它们。借助机械解释性（mechanistic interpretability）的思想，我们探测了 $ ext{pi}_{0.5}$ 的残差流，并发现任务进展，即轨迹中剩余的标准化时间，可以从激活中线性读取。我们发现这一信号在预训练的 PaliGemma 主干网络中存在，且在任何特定于机器人的数据训练之前就已存在。一个线性探针能够推广到未见过的任务，并在多提示数据训练时在语言反事实下变化，但并未能有效地引导策略。这些特性使得该信号在部署的 VLA 中直接有用。我们将该探针用作一个简单的无标签的 OOD 检测器，能够检测停滞的任务进展，并发现其与最先进的方法具有竞争力。我们的结果表明，VLAs 具有丰富的、可线性读取的内部表示，能够表达诸如任务进展等语义量，并且学习读取这些信号提供了一条轻量级、可解释的路径，以监控已部署的视觉运动策略。
cs.RO / 26 / 2608.13511
A Browser-Native Digital Test Range for Benchmarking 4D Ocean-Glider Planning Algorithms
用于基准测试4D海洋滑翔机规划算法的浏览器原生数字测试场
Abstract
Repeated in-situ evaluation of ocean-glider planners requires scarce vehicles, operators, deployment and recovery resources, and ocean conditions that cannot be reset for competing algorithms. We present a guided, installation-free browser-native digital test range that transforms a selected region into a reproducible four-dimensional experiment. The system leads users from regional domain selection through mission-scoped bathymetry, time/depth forcing, science objectives, optional task decomposition, route specification, current-advected execution, observation generation, and scoring. Its primary contribution is a common plan-to-observation contract unifying vehicle, sensing, and evaluator assumptions across manual routes, transparent built-in algorithms, and imported classical or learned-planner outputs, while exported artifacts form dataset-ready records. A controlled Observing System Simulation Experiment (OSSE) evaluates five classical planners in two episodes, three deterministic seeds, and a calibrated 60-hour horizon. All 54 missions completed and recovered without hard violations, while planner rankings and dive-policy effects revealed operational-scientific tradeoffs. An authentic public deployment supplied a field-referenced audit to scope current kinematic boundaries. Separately, source-locked GliderFlight 1.2.0 achieved native-to-browser parity through Pyodide/WebAssembly, establishing a pathway for high-fidelity multi-tier simulation. The resulting operational space is scientifically traceable and component-qualified for mission-scale pre-deployment experimentation.
Chinese Translation
对海洋滑翔机规划者进行重复的现场评估需要稀缺的车辆、操作人员、部署和回收资源，以及无法为竞争算法重置的海洋条件。我们提出了一种引导式、无需安装的浏览器原生数字测试场，将选定区域转变为可重复的四维实验。该系统引导用户从区域域选择开始，经过任务范围的海底地形、时间/深度强迫、科学目标、可选任务分解、路线指定、受流动影响的执行、观测生成和评分。其主要贡献是统一车辆、传感和评估假设的共同计划到观测合同，涵盖手动路线、透明的内置算法以及导入的经典或学习规划者输出，同时导出的工件形成数据集准备好的记录。一个受控的观测系统模拟实验（OSSE）在两个阶段、三个确定性种子和一个经过校准的60小时视野中评估了五个经典规划者。所有54个任务均在没有严重违规的情况下完成和回收，而规划者排名和潜水政策的影响揭示了操作与科学之间的权衡。一项真实的公共部署提供了一个现场参考审计，以界定当前的运动边界。此外，源锁定的GliderFlight 1.2.0通过Pyodide/WebAssembly实现了浏览器原生的平等，为高保真多层次模拟建立了路径。最终的操作空间在科学上是可追溯的，并且经过组件验证，适用于任务规模的预部署实验。
cs.RO / 27 / 2608.13555
HumanTracker: Towards Comprehensive and Human-Aligned Motion Tracking Benchmark
HumanTracker：迈向全面且与人类对齐的运动跟踪基准
Abstract
Humanoid motion tracking is central to teleoperation and whole-body imitation, yet evaluation often disagrees with what people perceive in videos. Kinematic errors average per-frame pose differences but miss the physical artifacts that matter most, particularly unstable support and incorrect contacts such as foot skating and mistimed touch-downs. Meanwhile, widely used test suites are small and lack the diversity needed to stress contact-rich, long-horizon behaviors. We introduce HumanTracker to make humanoid tracking evaluation both perceptually aligned and scalable. The HumanTracker benchmark contains approximately 153 hours of optical motion trajectories from multiple professional performers, organized into four motion families with text labels for fine-grained diagnosis. We further propose HumanScore, a preference-aligned metric trained on 12K motion pairs containing 24K motions. Across representative state-of-the-art trackers, HumanScore better predicts human preferences and reveals contact and stability failures that kinematic metrics often miss.
Chinese Translation
类人运动跟踪是远程操作和全身模仿的核心，但评估结果往往与人们在视频中的感知不一致。运动学误差平均每帧的姿态差异，但忽视了最重要的物理伪影，特别是不稳定的支撑和错误的接触，如脚滑动和时机不当的着陆。同时，广泛使用的测试套件规模较小，缺乏足够的多样性，无法有效测试富含接触的长时间行为。我们提出了HumanTracker，使类人跟踪评估既与感知对齐又具可扩展性。HumanTracker基准包含约153小时来自多个专业表演者的光学运动轨迹，组织为四个运动类别，并附有文本标签以便于细致诊断。我们进一步提出了HumanScore，这是一种基于偏好的度量，训练于包含24K动作的12K运动对中。在代表性的最先进跟踪器中，HumanScore更好地预测人类偏好，并揭示了运动学度量常常忽视的接触和稳定性失败。
cs.CV / 1 / 2608.12442
MV2: Multi-View Multi-Vehicle Driving Dataset for Novel View Synthesis
MV2：用于新视图合成的多视角多车辆驾驶数据集
Abstract
Differentiable rendering has advanced novel view synthesis (NVS), yet applying it to real-world driving remains difficult due to sparse capture viewpoints, dynamic objects, and limited multi-trajectory data. We introduce the Multi-View Multi-Vehicle (MV2) dataset and benchmark for evaluating NVS models under large viewpoint changes in dynamic urban scenes. MV2 features synchronized captures from a car, scooter, and drone, each following distinct yet synchronized trajectories. Training NVS methods on one vehicle's camera stream and testing on another enables evaluation under substantially larger viewpoint variations than existing single-trajectory datasets. All sequences are registered via Structure-from-Motion and camera poses verified using manual pixel-level correspondence annotations, yielding 50 high-quality scenes with 12000 images. Benchmarking recent NVS and camera pose estimation methods shows that NVS performance degrades with increasing viewpoint disparity, and that feed-forward pose estimators notably lag behind optimization-based approaches, highlighting MV2 as a rigorous testbed for NVS in driving. The dataset, benchmark protocol, and project resources are available at https://mv2-dataset.github.io/.
Chinese Translation
可微渲染技术推动了新视图合成（NVS）的发展，但由于捕获视点稀疏、动态物体以及有限的多轨迹数据，将其应用于现实世界驾驶仍然困难。我们介绍了多视角多车辆（MV2）数据集及其基准，用于评估在动态城市场景中大视点变化下的NVS模型。MV2包含来自汽车、电动滑板车和无人机的同步捕获，每种交通工具都遵循独特但同步的轨迹。在一辆车辆的摄像头流上训练NVS方法，并在另一辆上进行测试，使得在比现有单轨迹数据集更大视点变化下进行评估成为可能。所有序列通过运动结构重建（Structure-from-Motion）进行注册，并使用手动像素级对应注释验证相机姿态，生成了50个高质量场景和12000张图像。对近期NVS和相机姿态估计方法的基准测试表明，随着视点差异的增加，NVS性能下降，而前馈姿态估计器明显落后于基于优化的方法，突显了MV2作为驾驶中NVS的严格测试平台。数据集、基准协议和项目资源可在 https://mv2-dataset.github.io/ 获取。
cs.CV / 2 / 2608.12502
HIMEC: Directional Change Representation and Fixed-Interface Decoding for Remote Sensing Image Change Captioning
HIMEC：遥感图像变化描述的方向变化表示与固定接口解码
Abstract
Remote sensing image change captioning (RSICC) converts bitemporal imagery into a sentence describing semantic changes. Most RSICC methods condition caption decoders directly on fused visual features, leaving intermediate change structure and decoder-interface consistency less studied. We present HIMEC, combining Directional Change Representation (DCR) with fixed-interface decoding. DCR separates signed differences into appearance-oriented, disappearance-oriented, and shared-context streams before fusion. A learned-query encoder converts the fused representation into visually conditioned change-query tokens that form the scene decoder's only sample-dependent memory. A training-only auxiliary phrase decoder supplies caption-derived supervision. With a fixed zero input, the scene decoder maintains the same interface during training and inference. Separately, we evaluate a local-to-scene cascade conditioned on teacher-forced local states during training and autoregressive states at inference. On changed LEVIR-CC validation pairs, these states have a mean cosine distance of 0.69. Regime-matched conditioning recovers most of the associated deficit, whereas permuting state correspondence causes no detectable penalty. These findings are limited to the evaluated cascade. In a matched three-seed comparison, HIMEC reaches a Consensus-based Image Description Evaluation (CIDEr) score of $142.81\pm0.60$ on LEVIR-CC, versus $139.51\pm3.40$ for direct fused-feature memory. On SECOND-CC, fixed-zero and regime-matched diagnostic conditioning reach 75.67 and 76.99 CIDEr, respectively, versus 60.77 for the mismatched cascade. The source code will be made publicly available at https://github.com/ayshaashra/HIMEC upon publication.
Chinese Translation
遥感图像变化描述（RSICC）将双时相图像转换为描述语义变化的句子。大多数RSICC方法直接将标题解码器与融合的视觉特征相结合，导致中间变化结构和解码器接口一致性研究较少。我们提出了HIMEC，将方向变化表示（DCR）与固定接口解码相结合。DCR在融合之前将带符号的差异分离为面向外观、面向消失和共享上下文的流。一个学习查询编码器将融合表示转换为视觉条件变化查询标记，这些标记构成场景解码器唯一的样本依赖内存。一个仅用于训练的辅助短语解码器提供基于标题的监督。在固定零输入的情况下，场景解码器在训练和推理期间保持相同的接口。此外，我们评估了在训练期间基于教师强制的局部状态和在推理期间自回归状态的局部到场景级联。在变化的LEVIR-CC验证对上，这些状态的平均余弦距离为0.69。匹配的条件恢复了大部分相关的缺失，而状态对应的置换则没有造成可检测的惩罚。这些发现仅限于评估的级联。在匹配的三种种子比较中，HIMEC在LEVIR-CC上达到了基于共识的图像描述评估（CIDEr）得分为$142.81 ext{±}0.60$，而直接融合特征内存的得分为$139.51 ext{±}3.40$。在SECOND-CC上，固定零和匹配的诊断条件分别达到了75.67和76.99 CIDEr，而不匹配的级联得分为60.77。源代码将在出版后公开发布于https://github.com/ayshaashra/HIMEC。
cs.CV / 3 / 2608.12515
Can Vision-Language Models Assess Proxemic Risk from Egocentric Robot Images?
视觉-语言模型能否评估来自自我中心机器人图像的近端风险？
Abstract
Assessing proxemic danger from a robot's egocentric perspective is critical for safe embodied navigation in human environments and requires both visual and contextual reasoning. We evaluate three opensource vision-language models (VLMs) (\textit{InternVL}, \textit{Qwen-VL}, and \textit{SmolVLM}) on the classification of egocentric robot images into four danger levels, comparing three prompting strategies and two rounds of QLoRA fine-tuning against a stratified random baseline. Without fine-tuning, all models perform near the baseline, while fine-tuning yields only modest overall improvements. However, \textit{Qwen-VL} with an advanced prompt achieves substantially higher recall for high-danger cases than the other models. An analysis of person localization further shows that correct danger classification does not correspond to better spatial grounding, indicating that a model may produce a useful safety label without attending to the relevant region of the scene. These results show that current VLMs remain limited in fine-grained proxemic reasoning and spatial grounding, although targeted prompting and fine-tuning can improve high-danger detection in selected models.
Chinese Translation
从机器人的自我中心视角评估近端危险对于在人工环境中安全导航至关重要，这需要视觉和上下文推理。我们评估了三种开源视觉-语言模型（VLMs）（ extit{InternVL}、 extit{Qwen-VL} 和 extit{SmolVLM}）在将自我中心机器人图像分类为四个危险级别的表现，比较了三种提示策略和两轮 QLoRA 微调与分层随机基线的效果。在未进行微调的情况下，所有模型的表现接近基线，而微调仅带来了适度的整体改善。然而， extit{Qwen-VL} 在使用高级提示时，对于高危险案例的召回率显著高于其他模型。对人员定位的分析进一步表明，正确的危险分类并不对应于更好的空间定位，这表明模型可能在未关注场景相关区域的情况下产生有用的安全标签。这些结果表明，当前的 VLMs 在细粒度的近端推理和空间定位方面仍然有限，尽管有针对性的提示和微调可以改善选定模型的高危险检测。
cs.CV / 4 / 2608.12537
Surface-to-Skeleton 3D Cephalometry: Estimating Hidden Skeletal Landmarks from CT-Derived External Soft-Tissue Surfaces
表面到骨架的三维头颅测量：从CT衍生的外部软组织表面估计隐藏的骨骼标志点
Abstract
Existing 3D facial-landmark methods localize points on visible skin, but whether CT-defined internal skeletal landmarks can be inferred from external soft-tissue geometry remains unclear. We formulate a coordinate-consistent surface-to-skeleton task using same-acquisition CT-derived surfaces, separating estimation from optical-to-CT registration, scanner-domain, and acquisition-state effects, with coverage analyzed separately. From 240 clinical CT scans from two hospitals, we construct a locked retrospective protocol pairing CT-derived external soft-tissue point clouds with 21 skeletal landmarks and three visible soft-tissue landmarks. An integrated hierarchical point-cloud model achieves 2.97 mm mean radial error on skeletal landmarks and 3.03 mm on deep or surface-invisible landmarks in 40 held-out patients. Patient-mismatch controls support patient-specific signal beyond a fixed population configuration or global similarity alone, while coverage ablations indicate dependence on non-anterior geometry. Optical-transfer diagnostics reveal substantial coverage-related and global-configuration components, although deployable optical inference remains unresolved. These results answer the controlled feasibility question affirmatively and provide a basis for hidden skeletal landmark inference.
Chinese Translation
现有的三维面部标志点方法定位于可见皮肤上的点，但是否可以从外部软组织几何形状推断出CT定义的内部骨骼标志点仍不明确。我们制定了一个坐标一致的表面到骨架任务，使用相同采集的CT衍生表面，将估计与光学到CT配准、扫描仪域和采集状态效应分开，覆盖范围单独分析。我们从两家医院的240个临床CT扫描中构建了一个锁定的回顾性协议，将CT衍生的外部软组织点云与21个骨骼标志点和三个可见软组织标志点配对。一个集成的分层点云模型在40个保留患者的骨骼标志点上实现了2.97毫米的平均径向误差，在深层或表面不可见的标志点上实现了3.03毫米的误差。患者不匹配控制支持患者特异性信号，超越了固定人群配置或单一的全局相似性，而覆盖消融表明依赖于非前部几何形状。光学传递诊断揭示了显著的覆盖相关和全局配置成分，尽管可部署的光学推断仍未解决。这些结果肯定地回答了受控可行性问题，并为隐藏骨骼标志点推断提供了基础。
cs.CV / 5 / 2608.12549
StrAD: A Streaming Method and Benchmark for Audio Description Generation for Long-form Videos
StrAD：一种用于长视频音频描述生成的流式方法和基准
Abstract
Visual content is the dominant medium of communication, yet without audio descriptions (ADs), it remains inaccessible to blind and low-vision people. ADs narrate context-relevant visual events during natural audio pauses. Manually creating ADs is expensive, limiting coverage to a small fraction of available content. Most existing automatic AD generation methods frame the task as video clip captioning, requiring ground-truth timestamps and additional context cues such as character databases. Current benchmarks reinforce this framing, consisting of short video segments paired with automatic or task-mismatched annotations. We introduce StrAD, a benchmark for long-form AD generation on full-length videos spanning diverse genres such as movies, documentaries, short films, performances, and video games. We reformulate AD generation as streaming dense video captioning. Our approach processes full-length videos with a sliding window, inserting ADs into existing transcripts without ground-truth timestamps, and supports both fine-tuned models and zero-shot prompting of vision-language models. On the segment-level task with given timestamps, our fine-tuned StrAD-FT sets the state of the art on CMD-AD with 36.3 CIDEr (+10.0 over Shot-by-shot), establishes a reference point on StrAD (51.0 CIDEr), and remains competitive on MAD-Eval at 24.9 CIDEr. On the full-video streaming task, StrAD-FT reaches a SODA score of 2.4 against 1.1 for our zero-shot baseline StrAD-Zero, though both exhibit limitations in temporal localization and narrative coherence. While prior work has tackled full-video AD generation in an offline, multi-stage fashion, ours is the first streaming approach, generating ADs on the fly without ground-truth timestamps. StrAD makes progress on full-video AD generation measurable, a prerequisite for scaling accessibility.
Chinese Translation
视觉内容是主要的交流媒介，但没有音频描述（ADs），它对盲人和低视力人士仍然不可及。音频描述在自然音频暂停期间叙述与上下文相关的视觉事件。手动创建音频描述成本高昂，限制了覆盖范围，仅能涵盖可用内容的一小部分。大多数现有的自动音频描述生成方法将任务框定为视频片段字幕生成，这需要真实时间戳和额外的上下文线索，如角色数据库。目前的基准强化了这种框架，由短视频片段与自动或任务不匹配的注释配对组成。我们提出了StrAD，这是一个针对长视频音频描述生成的基准，涵盖电影、纪录片、短片、表演和视频游戏等多种类型。我们将音频描述生成重新表述为流式密集视频字幕生成。我们的方法通过滑动窗口处理完整视频，将音频描述插入现有的转录文本中，而无需真实时间戳，并支持微调模型和视觉-语言模型的零样本提示。在给定时间戳的片段级任务中，我们微调的StrAD-FT在CMD-AD上设定了36.3 CIDEr的最新记录（比逐帧方法提高了10.0），在StrAD上建立了51.0 CIDEr的参考点，并在MAD-Eval上保持竞争力，得分为24.9 CIDEr。在完整视频流式任务中，StrAD-FT的SODA得分为2.4，而我们的零样本基线StrAD-Zero为1.1，尽管两者在时间定位和叙事连贯性方面均存在局限性。虽然之前的研究以离线、多阶段的方式处理完整视频的音频描述生成，但我们的研究是首个流式方法，能够实时生成音频描述而无需真实时间戳。StrAD使完整视频音频描述生成的进展可衡量，这是扩大可及性的前提。
cs.CV / 6 / 2608.12570
Attribute-Conditioned Multimodal Slot Factorization for Controllable Fashion Retrieval
基于属性条件的多模态槽因子分解用于可控时尚检索
Abstract
Fashion retrieval often requires satisfying multiple attributes at once, such as category, color, pattern, and demographic. Monolithic embeddings mix these signals into a single vector, making attribute-specific control difficult at retrieval time. Many existing semantic-ID methods provide discrete item codes, but these codes are typically optimized as item-level or residual addresses and do not expose named, independently controllable attribute slots. We introduce MM-slotgate, a multimodal slot encoder that factorizes Fashion-CLIP text and image embeddings into four named attribute slots. Each slot learns its own text-image gate, so visually grounded attributes such as color and pattern can rely more on image evidence, while taxonomy-oriented attributes such as category and demographic can remain more text-driven. On H&M, using a combined slot-similarity and slot-logit retrieval score, MM-slotgate achieves 0.7566 macro ConstraintSatisfied@10, outperforming equal-weight multimodal fusion (0.7142) and fCLIP text-only retrieval (0.4755). The largest gain is on color, which improves from 0.321 to 0.889 (+0.568 absolute), as the learned color gate assigns 57.4% weight to image evidence. The learned gates are interpretable without modality supervision: color is image-leaning, category is text-leaning, and pattern and demographic lie near the middle. The resulting slots also remain controllable: linear probes show no measured excess leakage beyond the label-correlation baseline, and quantized slot codes support targeted intervention, including a 15.3x lift for color. These results suggest that controllable fashion retrieval benefits from typed, attribute-conditioned multimodal slots rather than either a single global embedding or opaque item-level semantic IDs.
Chinese Translation
时尚检索通常需要同时满足多个属性，如类别、颜色、图案和人口统计信息。单一的嵌入将这些信号混合成一个向量，使得在检索时进行属性特定控制变得困难。许多现有的语义ID方法提供离散的物品代码，但这些代码通常被优化为物品级或残差地址，并未暴露出命名的、可独立控制的属性槽。我们引入了MM-slotgate，这是一种多模态槽编码器，将Fashion-CLIP文本和图像嵌入分解为四个命名属性槽。每个槽学习其自己的文本-图像门控，因此视觉上有依据的属性如颜色和图案可以更多依赖于图像证据，而面向分类的属性如类别和人口统计信息则可以保持更多的文本驱动。在H&M数据集上，使用组合的槽相似度和槽逻辑检索得分，MM-slotgate达到了0.7566的宏观ConstraintSatisfied@10，超越了等权重的多模态融合（0.7142）和仅使用fCLIP文本检索（0.4755）。最大的提升出现在颜色上，从0.321提升至0.889（+0.568绝对值），因为学习到的颜色门控将57.4%的权重分配给图像证据。学习到的门控在没有模态监督的情况下是可解释的：颜色偏向于图像，类别偏向于文本，而图案和人口统计信息则位于中间。结果槽也保持可控性：线性探针显示没有超过标签相关性基线的额外泄漏，量化的槽代码支持针对性干预，包括颜色的15.3倍提升。这些结果表明，可控的时尚检索受益于类型化的、基于属性条件的多模态槽，而不是单一的全局嵌入或不透明的物品级语义ID。
cs.CV / 7 / 2608.12600
PseudoMapLabeler: Confidence-Aware Pseudo-Label Generation for Semi-Supervised Online Mapping
PseudoMapLabeler：基于置信度的伪标签生成用于半监督在线地图构建
Abstract
A critical challenge in deploying online HD map construction systems to real-world scenarios is the scarcity of labeled training data, which limits model generalization in diverse environments. To address this limitation, we propose a teacher-student semi-supervised learning (SSL) framework that generates high-quality pseudo-labels from unlabeled data through confidence-aware map refinement. Our approach first trains a teacher model on limited labeled data, then leverages Beta-distribution-based confidence maps to assess the reliability of predicted map elements across temporal observations. Unlike conventional filtering methods that discard entire elements, we introduce a spatial clipping technique that selectively preserves high-confidence regions while removing unreliable segments. The refined map elements serve as map priors that improve the teacher model's prediction accuracy on unlabeled data in a second pass. These enhanced predictions become pseudo-labels for training a student model from scratch, followed by fine-tuning on the original labeled data. Experimental results on the nuScenes dataset demonstrate that our teacher-student framework with refined pseudo-labels improves performance by +6.1 mAP under a low-label regime compared to training on labeled data alone, offering a practical solution to the labeled data scarcity problem in online HD map construction.
Chinese Translation
在将在线高清地图构建系统部署到现实场景中的一个关键挑战是标记训练数据的稀缺，这限制了模型在多样化环境中的泛化能力。为了解决这一限制，我们提出了一种教师-学生半监督学习（SSL）框架，通过基于置信度的地图细化从未标记数据中生成高质量的伪标签。我们的方法首先在有限的标记数据上训练教师模型，然后利用基于Beta分布的置信度地图评估跨时间观测的预测地图元素的可靠性。与传统的过滤方法不同，后者会丢弃整个元素，我们引入了一种空间剪裁技术，选择性地保留高置信度区域，同时去除不可靠的部分。经过细化的地图元素作为地图先验，提升了教师模型在未标记数据上的预测准确性，随后进行第二次预测。这些增强的预测成为从头开始训练学生模型的伪标签，随后在原始标记数据上进行微调。在nuScenes数据集上的实验结果表明，与单独在标记数据上训练相比，我们的教师-学生框架结合细化的伪标签在低标记条件下提高了+6.1 mAP，提供了一种解决在线高清地图构建中标记数据稀缺问题的实用方案。
cs.CV / 8 / 2608.12611
From Visual Widgets to UI Code: Efficient Tool-Grounded Generation
从视觉组件到用户界面代码：高效的工具基础生成
Abstract
Existing screenshot-to-code systems face a trade-off between flexibility and controllability. Direct multimodal generation can hallucinate visible details, whereas structured pipelines reduce such errors through component-wise decomposition, predefined templates, and customized intermediate representations. These structures, however, introduce additional generative orchestration and restrict outputs to designs covered by the representation. We investigate whether selective tool grounding can improve the fidelity--efficiency trade-off of direct widget-to-code generation. We introduce \textbf{WidgetGen}, a lightweight tool-grounded framework that extracts observable text and color evidence, performs high-level layout and optional chart reasoning, and directly generates executable JavaScript XML (\emph{JSX}). This design reduces reliance on component-wise generation while avoiding a fixed UI schema. Across six multimodal models and \(1{,}000\) held-out widgets, WidgetGen outperforms direct prompting and the structured Widget2Code pipeline on most visual reconstruction metrics, with consistent gains in area, legibility, and style. Finally, reconstruction-derived image-code pairs improve six Qwen-family open-weight models across every reported metric through supervised fine-tuning. These results establish WidgetGen as a strong lightweight baseline and show that selective evidence grounding offers an effective alternative to extensive representation constraints.
Chinese Translation
现有的截图到代码系统在灵活性和可控性之间面临权衡。直接的多模态生成可能会产生虚构的可见细节，而结构化管道通过组件分解、预定义模板和定制的中间表示来减少此类错误。然而，这些结构引入了额外的生成协调，并将输出限制在表示所覆盖的设计范围内。我们研究选择性工具基础是否可以改善直接组件到代码生成的保真度-效率权衡。我们提出了 extbf{WidgetGen}，一个轻量级的工具基础框架，提取可观察的文本和颜色证据，执行高层布局和可选的图表推理，并直接生成可执行的JavaScript XML（ extit{JSX}）。该设计减少了对组件生成的依赖，同时避免了固定的用户界面模式。在六个多模态模型和1000个保留组件的实验中，WidgetGen在大多数视觉重建指标上优于直接提示和结构化的Widget2Code管道，在面积、可读性和风格方面均有一致的提升。最后，基于重建的图像-代码对通过监督微调改善了六个Qwen系列开放权重模型在每个报告指标上的表现。这些结果确立了WidgetGen作为一个强大的轻量级基线，并表明选择性证据基础提供了一个有效的替代方案，克服了广泛的表示约束。
cs.CV / 9 / 2608.12627
EgoCITE: Context-Augmented Indexing and Time-Aware Retrieval for Long-Horizon Egocentric Memory
EgoCITE：面向长时段自我中心记忆的上下文增强索引与时间感知检索
Abstract
Long-horizon egocentric memory transforms continuous first-person video and audio into a searchable record of past experiences. We demonstrate two bottlenecks in existing systems: indices built from context-poor captions are unreliable for agentic search, while retrieval ignores a question's temporal intent. To address both bottlenecks, we introduce EgoCITE (Egocentric Context-augmented Indexing and Time-aware Evidence retrieval), a long-horizon agentic memory framework for egocentric QA. EgoCITE comprises three components. EgoScheme uses local multimodal context to turn fragmentary video captions and speech transcripts into self-contained atomic memory indices. EgoIndex organizes complementary action, activity, utterance, and conversation representations into searchable multi-view memory indices at multiple granularities. EgoRetrv combines semantic search with question-conditioned temporal relevance scoring and curation of retrieved evidence. We evaluate EgoCITE on EgoLifeQA, EgoMem, and EgoR1-Bench in terms of answer accuracy and target-event retrieval alignment. EgoCITE improves accuracy over agentic memory baselines by at least 4.4--14.2\% while achieving 36$\times$ lower cost than long-context LLM agents.
Chinese Translation
长时段自我中心记忆将连续的第一人称视频和音频转化为可搜索的过去经历记录。我们展示了现有系统中的两个瓶颈：基于缺乏上下文的标题构建的索引在自主搜索中不可靠，而检索忽视了问题的时间意图。为了解决这两个瓶颈，我们提出了EgoCITE（自我中心上下文增强索引与时间感知证据检索），这是一个面向自我中心问答的长时段自主记忆框架。EgoCITE由三个组件组成。EgoScheme利用局部多模态上下文将零散的视频标题和语音转录转化为自包含的原子记忆索引。EgoIndex将互补的动作、活动、发言和对话表示组织成多粒度的可搜索多视图记忆索引。EgoRetrv结合语义搜索与基于问题的时间相关性评分和检索证据的整理。我们在EgoLifeQA、EgoMem和EgoR1-Bench上评估EgoCITE，重点关注答案准确性和目标事件检索的一致性。EgoCITE在自主记忆基线之上提高了至少4.4%至14.2%的准确性，同时实现了比长上下文LLM代理低36倍的成本。
cs.CV / 10 / 2608.12658
Inference-Time Orthogonal Seeding Enables Geometry-Aligned 3D Organ Segmentation for Slice-Propagation Methods
推理时正交种子方法实现几何对齐的3D器官分割用于切片传播方法
Abstract
Dense voxel-level annotation remains a major bottleneck in 3D medical image segmentation. Single-slice propagation methods such as Sli2Vol reduce this burden by propagating one annotated seed slice through a volume using label-free registration. However, axial-only propagation accumulates errors with distance from the seed, especially in surface-distance metrics, because it ignores coronal and sagittal evidence and therefore underuses the 3D information available in CT/MRI volumes. To better leverage volumetric geometry, we study how key training and inference choices affect slice-propagation models, including single-axis versus multi-axis label-free registration, single-seed versus multi-seed propagation, and orthogonal seed configurations. Instead of propagating from a single axial seed, we use three orthogonal seeds---one axial, one coronal, and one sagittal---and fuse their propagated labels with a simple label-free rule. Our results show that the training paradigm has limited impact: an axially trained network applied to off-axis seeds captures nearly all the improvement, while explicit three-axis training adds little. Instead, performance is driven by inference-time seed geometry, especially orthogonality rather than the number of annotated slices, as a budget-matched three-axial control provides no benefit and can even degrade performance. On a multi-organ CT cohort, orthogonal seeding with the axial Sli2Vol backbone improves Dice by 21.9%, Normalized Surface Dice by 25.5%, and reduces Average Hausdorff Distance by 53.5% over the single-axis baseline.
Chinese Translation
密集的体素级注释仍然是3D医学图像分割的主要瓶颈。单切片传播方法如Sli2Vol通过使用无标签配准在体积中传播一个注释的种子切片，从而减轻了这一负担。然而，仅沿轴向传播会随着距离种子而累积误差，特别是在表面距离度量中，因为它忽略了冠状面和矢状面的证据，因此未充分利用CT/MRI体积中可用的3D信息。为了更好地利用体积几何，我们研究了关键的训练和推理选择如何影响切片传播模型，包括单轴与多轴无标签配准、单种子与多种子传播，以及正交种子配置。我们不再仅从单一的轴向种子传播，而是使用三个正交种子——一个轴向、一个冠状和一个矢状——并通过简单的无标签规则融合它们传播的标签。我们的结果表明，训练范式的影响有限：应用于偏轴种子的轴向训练网络几乎捕获了所有的改进，而显式的三轴训练贡献不大。相反，性能主要受推理时种子几何形状的驱动，尤其是正交性，而不是注释切片的数量，因为预算匹配的三轴对照并未提供任何好处，甚至可能降低性能。在一个多器官CT队列中，采用轴向Sli2Vol骨干的正交种子方法使Dice系数提高了21.9%，归一化表面Dice提高了25.5%，并将平均Hausdorff距离减少了53.5%，相较于单轴基线。
cs.CV / 11 / 2608.12689
Mr3D-VL: A generalist vision language foundation model for Multiparametric 3D Magnetic Resonance Imaging
Mr3D-VL：用于多参数三维磁共振成像的通用视觉语言基础模型
Abstract
Multi-parametric magnetic resonance imaging (mpMRI) is a cornerstone for brain tumor diagnosis and treatment, yet current AI models face critical limitations: their lack of natural language interaction and interpretability impedes spatial information integration and cross-modal reasoning required clinically. Key challenges arise from significant physical meaning differences across modalities, spatial misalignment due to scan intervals, and the need for complex multi-feature interpretation in tasks like glioma grading. While visual-language models (VLMs) show promise in cross-modal understanding, existing methods focus mainly on 2D image modeling, neglecting direct perception of 3D volumetric space. Although 3D VLMs have been proposed for report generation and feature alignment in 3D CT imaging, mpMRI applications demand collaborative inference across multiple imaging modalities-a requirement unmet by current solutions. To address this, we introduce Mr3D-VL, a dedicated visual-language foundation model for multi-parametric 3D MRI. With 4 billion parameters, it employs an unsupervised pre-trained shared 3D encoder and 4D rotational positional embedding for dual modality-spatial integration. Its cross-modal projection layer uses a multi-resolution feature implantation strategy to enhance feature perception across resolutions. Experimental results show significant improvements over existing 4B/7B/30B domain-specific and general-purpose models in text generation tasks, achieving a BERTScore of 0.856 for report generation, with question-answering accuracy at 0.713 and multiple-choice accuracy at 0.912.
Chinese Translation
多参数磁共振成像（mpMRI）是脑肿瘤诊断和治疗的基石，但当前的人工智能模型面临着关键的局限性：缺乏自然语言交互和可解释性，阻碍了临床所需的空间信息整合和跨模态推理。主要挑战源于不同模态之间显著的物理意义差异、由于扫描间隔导致的空间错位，以及在胶质瘤分级等任务中对复杂多特征解释的需求。虽然视觉语言模型（VLMs）在跨模态理解方面展现出潜力，但现有方法主要集中在二维图像建模上，忽视了对三维体积空间的直接感知。尽管已经提出了用于三维CT成像报告生成和特征对齐的三维VLMs，但mpMRI应用要求在多个成像模态之间进行协同推理，这是当前解决方案无法满足的需求。为了解决这一问题，我们提出了Mr3D-VL，一种专门针对多参数三维MRI的视觉语言基础模型。该模型具有40亿个参数，采用无监督预训练的共享三维编码器和四维旋转位置嵌入，以实现双模态空间整合。其跨模态投影层使用多分辨率特征植入策略，以增强不同分辨率下的特征感知。实验结果显示，在文本生成任务中，相较于现有的4B/7B/30B领域特定和通用模型，取得了显著的改进，报告生成的BERTScore达到0.856，问答准确率为0.713，多项选择准确率为0.912。
cs.CV / 12 / 2608.12698
Class Geometry as Supervision for Sample-Efficient Open-World Detection
类几何作为样本高效开放世界检测的监督
Abstract
Open-world object detection requires models to recognize known categories, reject unfamiliar objects, and incorporate new classes over time. This is especially challenging in scarce-data settings such as biomedical and scientific imaging, where rare categories may have only a few annotated examples and fine-grained classes differ by subtle morphology. Prototype-based detectors are natural for this regime, but they typically learn class prototypes as independent anchors, ignoring relational structure among classes. We propose class-geometry supervision (CGS), a general framework that constrains learned prototype or class-representation spaces to preserve visual or semantic class dissimilarities estimated from training data. CGS introduces a dissimilarity-preserving objective that aligns pairwise distances among learned class representations with a target class-geometry matrix while retaining the standard task loss. We instantiate the same objective across prototype recognition, few-shot biomedical object detection, open-set detection, novel-class insertion, and OWOD adaptation on COCO. Experiments show that CGS improves sample efficiency in recognition and ova detection, substantially strengthens novel-class insertion, and improves unknown recall on COCO while retaining much of the known-class detection performance. Ablations show that meaningful visual geometry provides the most reliable gains, while random geometry can help novel separation but is less consistent for few-shot detection. These results suggest that relational class geometry is an effective supervisory signal for building calibrated and extensible open-world detectors under limited supervision.
Chinese Translation
开放世界物体检测要求模型识别已知类别，拒绝不熟悉的物体，并随着时间的推移纳入新类别。这在数据稀缺的环境中尤为具有挑战性，例如生物医学和科学成像，其中稀有类别可能只有少量标注示例，而细粒度类别则因微妙的形态差异而不同。基于原型的检测器在这种情况下是自然的选择，但它们通常将类原型学习为独立的锚点，忽视了类之间的关系结构。我们提出了类几何监督（Class Geometry Supervision, CGS），这是一个通用框架，它约束学习到的原型或类表示空间，以保持从训练数据中估计的视觉或语义类差异。CGS引入了一个保持差异性的目标，该目标将学习到的类表示之间的成对距离与目标类几何矩阵对齐，同时保留标准任务损失。我们在原型识别、少样本生物医学物体检测、开放集检测、新类别插入和COCO上的开放世界物体检测（OWOD）适应中实例化了相同的目标。实验表明，CGS提高了识别和卵子检测的样本效率，显著增强了新类别插入，并在COCO上提高了未知召回率，同时保持了大部分已知类别检测性能。消融实验表明，有意义的视觉几何提供了最可靠的增益，而随机几何可以帮助新类别分离，但在少样本检测中一致性较差。这些结果表明，关系类几何是构建经过校准和可扩展的开放世界检测器的有效监督信号，尤其是在有限监督的情况下。
cs.CV / 13 / 2608.12714
Towards Sparsely Annotated Open-World Object Detection
面向稀疏标注的开放世界目标检测
Abstract
Real-world object detection operates under ambiguous supervision, where unlabeled regions may correspond to missing annotations of known objects or genuinely unknown categories. These challenges have been addressed separately in Sparsely Annotated Object Detection (SAOD) and Open-World Object Detection (OWOD). In practice, their co-occurrence remains an open problem. To address this problem, we introduce Sparsely Annotated Open-World Object Detection (SA-OWOD), a new task that jointly considers sparse supervision and the presence of unseen categories. We propose Dual-Perspective Object Discovery (DPOD), a unified framework that jointly models unlabeled known and unknown instances via two complementary mechanisms. The Known Target Recovery Module (KTRM) recovers supervision for unlabeled known instances and explicitly regularizes the feature space to separate known and unknown representations. Complementarily, the Dual-Disagreement Target Generator (DDTG) identifies reliable unknown candidates through cross-view semantic inconsistency. By integrating these modules, DPOD resolves contradictory supervision signals caused by ambiguous unlabeled regions. As a result, it prevents misclassification between known and unknown objects and stabilizes the decision boundaries. Experimental results on sparsely annotated open-world benchmarks demonstrate that the proposed method outperforms existing open-world detection methods, particularly in detecting unknown objects.
Chinese Translation
现实世界中的目标检测在模糊监督下进行，其中未标记区域可能对应于已知对象的缺失注释或真正未知的类别。这些挑战在稀疏标注目标检测（Sparsely Annotated Object Detection, SAOD）和开放世界目标检测（Open-World Object Detection, OWOD）中分别得到了解决。然而，它们的共存仍然是一个未解决的问题。为了解决这一问题，我们提出了稀疏标注开放世界目标检测（Sparsely Annotated Open-World Object Detection, SA-OWOD），这是一个新任务，联合考虑稀疏监督和未见类别的存在。我们提出了双视角目标发现（Dual-Perspective Object Discovery, DPOD），这是一个统一框架，通过两种互补机制共同建模未标记的已知和未知实例。已知目标恢复模块（Known Target Recovery Module, KTRM）为未标记的已知实例恢复监督，并明确正则化特征空间以区分已知和未知表示。互补地，双不一致目标生成器（Dual-Disagreement Target Generator, DDTG）通过跨视图语义不一致性识别可靠的未知候选对象。通过整合这些模块，DPOD 解决了由模糊未标记区域引起的矛盾监督信号。因此，它防止了已知和未知对象之间的误分类，并稳定了决策边界。在稀疏标注开放世界基准上的实验结果表明，所提出的方法优于现有的开放世界检测方法，特别是在检测未知对象方面。
cs.CV / 14 / 2608.12721
VOS-Agent: The 1st Place Solution for the 8th LSVOS Challenge (MOSEv2 Track)
VOS-Agent：第八届LSVOS挑战赛（MOSEv2 Track）第一名解决方案
Abstract
Complex video object segmentation requires robust target propagation under severe occlusion, disappearance and reappearance. Although SAM3 provides strong promptable mask propagation, a uniform inference path remains unreliable for tiny targets with insufficient visual evidence and semantic-dominated targets whose identities depend on explicit attributes. To this end, we present VOS-Agent, a collaborative framework that retains SAM3 as the shared dense segmentation module and conditionally activates specialized agents according to target characteristics. A Target Perception and Routing Agent assigns each sequence to a regular, tiny, or semantic-dominated route. Tiny targets are supported by a Visual Tracking Agent through confidence-aware box prompts, while semantic-dominated targets are handled by an MLLM-based Semantic Agent through description-guided localization and candidate verification. On the MOSEv2 test set, VOS-Agent achieves 69.82% on the official $\mathcal{J}\&\dot{\mathcal{F}}$ metric and ranks first in the MOSEv2 Track of the 8th LSVOS Challenge at ECCV 2026.
Chinese Translation
复杂的视频目标分割要求在严重遮挡、消失和重新出现的情况下进行稳健的目标传播。尽管SAM3提供了强大的可提示掩码传播，但对于视觉证据不足的小目标和依赖于显性属性的语义主导目标，统一的推理路径仍然不可靠。为此，我们提出了VOS-Agent，一个协作框架，保留SAM3作为共享的密集分割模块，并根据目标特征有条件地激活专门的代理。目标感知与路由代理将每个序列分配到常规、小型或语义主导的路径。小目标通过视觉跟踪代理通过置信度感知的框提示得到支持，而语义主导目标则通过基于MLLM的语义代理通过描述引导的定位和候选验证进行处理。在MOSEv2测试集上，VOS-Agent在官方的$ extmath{J} ext{和} extmath{F}$指标上达到了69.82%，并在2026年ECCV的第八届LSVOS挑战赛的MOSEv2 Track中排名第一。
cs.CV / 15 / 2608.12725
A Generative Approach for Improving Multi-Label Defect Classification in Photovoltaic Modules
一种生成方法用于改善光伏模块中的多标签缺陷分类
Abstract
This paper addresses the challenge of multi-label defect classification in electroluminescence (EL) images of photovoltaic (PV) cells. Training models on images where multiple defects co-occur creates learning ambiguity, making it difficult to disentangle visual features for specific defect types, a problem compounded by the scarcity of examples for individual classes. To tackle this, we introduce Generative Defect Isolation (GDI), utilizing the LaMa inpainting model with Fast Fourier Convolutions to remove selected defects and generate realistic, single-defect training samples. Extensive experiments on Vision Transformer (ViT-S, ViT-L) and EfficientNetV2-L architectures demonstrate that GDI significantly outperforms baselines. The performance gains are most pronounced in low-data scenarios; class-wise analysis shows substantial improvements, boosting the F1-Score for rare defect classes by up to 63.6%. Furthermore, GDI effectively resolves learning ambiguity from co-occurring defects, yielding a 26% reduction in such co-occurring classification errors. Our work establishes GDI as an effective method for maximizing the value of existing segmentation datasets and sets a new performance benchmark for multi-label classification in this domain.
Chinese Translation
本文针对光伏（PV）电池的电致发光（EL）图像中的多标签缺陷分类挑战进行探讨。在存在多个缺陷共存的图像上训练模型会造成学习模糊，使得难以区分特定缺陷类型的视觉特征，而个别类别样本稀缺的问题进一步加剧了这一困难。为此，我们提出了生成缺陷隔离（Generative Defect Isolation, GDI）方法，利用LaMa修复模型和快速傅里叶卷积去除选定缺陷，并生成逼真的单缺陷训练样本。在Vision Transformer（ViT-S, ViT-L）和EfficientNetV2-L架构上的大量实验表明，GDI显著优于基线模型。性能提升在低数据场景中最为明显；类别分析显示，稀有缺陷类别的F1分数提升高达63.6%。此外，GDI有效解决了共存缺陷带来的学习模糊问题，使得此类共存分类错误减少了26%。我们的研究确立了GDI作为最大化现有分割数据集价值的有效方法，并为该领域的多标签分类设定了新的性能基准。
cs.CV / 16 / 2608.12737
Dual-Manifold Geometry Guided Representation Learning: Adaptive Coupling between Kernel and Data Spaces
双流形几何引导的表示学习：核空间与数据空间的自适应耦合
Abstract
Deep representation learning has primarily focused on how features evolve across network layers, while largely overlooking the structured geometry embedded in network parameters. We introduce a dual-manifold perspective in which each convolutional layer contains two coupled geometric spaces: a Kernel Manifold induced by convolutional filters and a Data Manifold characterized by intermediate feature representations. Because these manifolds share the same channel space, parameter geometry can provide complementary structural information to guide feature evolution. Based on this insight, we propose Kernel-Guided Feature Transform (KGFT), a lightweight module that derives a geometric guidance matrix from the kernel Gram matrix and uses it to transform the covariance structure of feature representations. Unlike conventional attention mechanisms that reweight feature responses, KGFT explicitly reshapes feature relationships by transferring geometric information from the kernel manifold to the data manifold. To accommodate network hierarchy, we further introduce Exploit and Explore modes with a depth-aware scheduling strategy and a learnable guidance strength that adaptively controls the contribution of geometric transformation. This design promotes geometric alignment in shallow layers while encouraging feature diversity in deeper layers, without imposing excessive constraints on representation learning. Theoretical analysis establishes the validity of the proposed transformation and characterizes its effect on feature covariance. Extensive experiments across CNN- and Transformer-based architectures, including ResNet, ViT, and LLaMA-7B, demonstrate consistent improvements on image classification and arithmetic reasoning tasks, validating the generality and effectiveness of kernel-guided dual-manifold representation learning. Code will be publicly available.
Chinese Translation
深度表示学习主要关注特征在网络层之间的演变，而在很大程度上忽视了嵌入在网络参数中的结构几何。我们引入了一种双流形视角，其中每个卷积层包含两个耦合的几何空间：由卷积滤波器诱导的核流形（Kernel Manifold）和由中间特征表示特征化的数据流形（Data Manifold）。由于这些流形共享相同的通道空间，参数几何可以提供互补的结构信息，以引导特征的演变。基于这一见解，我们提出了核引导特征变换（Kernel-Guided Feature Transform, KGFT），这是一个轻量级模块，它从核Gram矩阵中推导出几何引导矩阵，并利用它来变换特征表示的协方差结构。与传统的注意力机制通过重新加权特征响应不同，KGFT通过将几何信息从核流形转移到数据流形，显式地重塑特征关系。为了适应网络层次结构，我们进一步引入了利用（Exploit）和探索（Explore）模式，并采用深度感知调度策略和可学习的引导强度，以自适应地控制几何变换的贡献。该设计在浅层促进几何对齐，同时在深层鼓励特征多样性，而不对表示学习施加过多约束。理论分析验证了所提变换的有效性，并描述了其对特征协方差的影响。在包括ResNet、ViT和LLaMA-7B在内的基于CNN和Transformer的架构上进行的广泛实验表明，在图像分类和算术推理任务上均取得了一致的改进，验证了核引导双流形表示学习的普遍性和有效性。代码将公开发布。
cs.CV / 17 / 2608.12746
Dual-Stream Cross-Anchor Correction Grounding Long-Form Captions and the Domain Limits of Object-Level Anchors
双流交叉锚点校正：为长文本标题提供基础，并探讨对象级锚点的领域限制
Abstract
Object hallucination in multimodal large language models arises when language priors and corpus co-occurrence bias outweigh the visual evidence, with nothing tying an individual object mention to what the image shows. Most remedies intervene at decoding time without training, yet under a unified protocol their benefit is confined to short captions;supervised fine-tuning (SFT) on a detail- rich corpus lengthens captions, but over forty percent still name absent objects. This paper proposes Dual-Stream Cross-Anchor Correction (DSCC). Unlike work that post-processes decoding, DSCC is the first to inject object-level visual anchors into the language model itself during fine- tuning: a perception stream aligns object-level hidden states at an intermediate layer to frozen text anchors by a bidirectional contrastive objective; a cognition stream lets deeper layers query those anchors by cross-attention at every generation step; and a two-stage curriculum gate couplesthem, making evidence retrieval a structural constraint at each autoregressive step. Under one backbone and one scoring protocol, experiments span long-caption hallucination, object-existence discrimination and cross-domain generalisation, with vanilla SFT on the same corpus and schedule as a length- and density-matched control, so gains are attributed layer by layer. DSCC is the only method reaching the long-caption, low-hallucination region: captions roughly 1.9 times the baseline length at 88.19% precision per object mention, the highest under a density-independent criterion. Ablations expose a synergy: the perception stream alone degrades precision yet reverses sign when stacked on the cognition stream. No universal superiority is claimed: three out-of- domain benchmarks yield a predictable, falsifiable domain-conditionality, the synergy being bound to the anchors' semantic domain and breaking on charts and optical illusions.
Chinese Translation
在多模态大型语言模型中，当语言先验和语料库共现偏差超过视觉证据时，会出现对象幻觉，个别对象提及与图像显示内容之间缺乏联系。大多数解决方案在解码时进行干预，而不进行训练，但在统一协议下，其效果仅限于短标题；在丰富细节的语料库上进行监督微调（SFT）可以延长标题，但仍有超过40%的标题提及了不存在的对象。本文提出了双流交叉锚点校正（DSCC）。与后处理解码的工作不同，DSCC首次在微调过程中将对象级视觉锚点注入语言模型中：感知流通过双向对比目标将中间层的对象级隐藏状态与冻结的文本锚点对齐；认知流允许更深层次在每个生成步骤通过交叉注意力查询这些锚点；而两阶段课程门将它们结合在一起，使得证据检索在每个自回归步骤中成为结构约束。在一个骨干网络和一个评分协议下，实验涵盖了长标题幻觉、对象存在性判别和跨领域泛化，并与在相同语料库和时间表下进行的普通SFT进行比较，以便逐层归因。DSCC是唯一能够达到长标题、低幻觉区域的方法：标题长度大约是基线的1.9倍，每个对象提及的精确度为88.19%，在密度无关标准下为最高。消融实验揭示了协同效应：仅感知流会降低精确度，但在与认知流叠加时会逆转符号。没有声称存在普遍的优越性：三个领域外基准产生了可预测的、可证伪的领域条件性，协同效应受限于锚点的语义领域，并在图表和光学错觉中失效。
cs.CV / 18 / 2608.12748
Scaling Representation Diversity: Modulated Attention and Reconstructive Regularization for Visual Grounding
扩展表征多样性：用于视觉定位的调制注意力与重构正则化
Abstract
Referring Expression Comprehension (REC) is commonly studied under dataset-specific fine-tuning, resulting in specialist models with limited cross-dataset generalization. In this work, we revisit REC from the perspective of unified open-vocabulary grounding and identify representation degeneration as a key obstacle to scaling a single generalist model. To preserve representation diversity, we propose a holistic data-model co-design framework. Architecturally, we introduce the Modulated Attention-Contrastive Head (mACH) for efficient token-level vision-language alignment and a text-conditioned JEPA auxiliary stream that provides complementary gradient support to preserve alignment-active representations without inference overhead. On the data side, we introduce Objects365-Caption, enriching Objects365 with context-aware referring expressions for large-scale language supervision. We further provide a theoretical analysis showing that complementary gradient subspaces preserve alignment capacity and thereby scale representation diversity. Extensive experiments demonstrate that our single-checkpoint framework achieves highly competitive performance on standard REC benchmarks while exhibiting strong generalization across heterogeneous grounding datasets without benchmark-specific adaptation.
Chinese Translation
指称表达理解（REC）通常在特定数据集的微调下进行研究，导致专业模型在跨数据集泛化方面的局限性。在本研究中，我们从统一开放词汇定位的角度重新审视REC，并识别出表征退化是扩展单一通用模型的关键障碍。为了保持表征多样性，我们提出了一个整体数据-模型协同设计框架。在架构上，我们引入了调制注意力对比头（Modulated Attention-Contrastive Head，mACH），用于高效的标记级视觉-语言对齐，以及一个文本条件的JEPA辅助流，为保持对齐活跃的表征提供互补的梯度支持，而无需推理开销。在数据方面，我们引入了Objects365-Caption，通过上下文感知的指称表达丰富Objects365，以实现大规模语言监督。我们进一步提供了理论分析，表明互补梯度子空间保持对齐能力，从而扩展表征多样性。大量实验表明，我们的单检查点框架在标准REC基准上实现了高度竞争的性能，同时在异构定位数据集上展现出强大的泛化能力，无需针对特定基准的适配。
cs.CV / 19 / 2608.12766
PatchGen: Learning Soft Intra-Image Predictive Subsets for Visual Generalization
PatchGen：学习用于视觉泛化的软内部图像预测子集
Abstract
Visual classifiers are expected to generalize under data shifts, target shifts, and their combinations, yet most existing methods focus on domain invariance while failing to address intra-image predictive sufficiency. We investigate the structural hypothesis that each image contains a sample-adaptive oracle intra-image predictive subset sufficient for label prediction, while the remaining patches form non-essential complementary context that may correlate with the label. The theoretical analysis shows that restricting prediction to this oracle subset preserves the Bayes risk achievable by the full-patch representation while admitting a complexity bound that tightens with the oracle-subset size. Based on this view, we propose PatchGen, a text-free module that learns a sample-dependent soft predictive-subset mask as a task-driven proxy for the unobserved oracle subset mask. Specifically, histopathology visualizations suggest that PatchGen assigns higher scores to tumor-consistent regions than to some frequently co-occurring inflammatory context. Extensive experiments on natural and histopathological image benchmarks spanning all three shift settings show that PatchGen improves average performance over matched-backbone baselines in most evaluated configurations, enhances generalization to unknown classes, and remains competitive with vision-language methods without text supervision.
Chinese Translation
视觉分类器期望在数据偏移、目标偏移及其组合下具有良好的泛化能力，然而大多数现有方法专注于领域不变性，却未能解决内部图像预测充分性的问题。我们探讨了一个结构假设：每幅图像包含一个样本自适应的内部图像预测子集，该子集足以进行标签预测，而其余的图块则形成非必要的补充上下文，可能与标签相关。理论分析表明，将预测限制在这个预言子集上可以保持由全图块表示所能达到的贝叶斯风险，同时承认一个复杂度界限，该界限随着预言子集大小的增加而收紧。基于这一观点，我们提出了PatchGen，一个无文本模块，学习一个样本依赖的软预测子集掩码，作为未观察到的预言子集掩码的任务驱动代理。具体而言，组织病理学可视化表明，PatchGen对肿瘤一致区域赋予的分数高于某些频繁共现的炎症上下文。在涵盖所有三种偏移设置的自然和组织病理图像基准上的大量实验表明，PatchGen在大多数评估配置中提高了与匹配骨干网络基线的平均性能，增强了对未知类别的泛化能力，并在没有文本监督的情况下与视觉-语言方法保持竞争力。
cs.CV / 20 / 2608.12773
CW-BASS v2: Saturation-Aware Pseudo-Label Selection for Semi-Supervised Segmentation under Foundation-Model Teachers
CW-BASS v2：基于饱和度的伪标签选择方法用于基础模型教师下的半监督分割
Abstract
Semi-supervised semantic segmentation has long turned on one question, which pseudo-labels to trust, and a generation of selection rules, dynamic thresholds, per-class curricula, soft confidence weights, answered it for the noisy, under-confident ResNet teachers of their day. Self-supervised foundation encoders change the regime: with a DINOv2 teacher, confidence saturates, so the filtering that helped a weak teacher can hurt a strong one. We propose CW-BASS v2, a saturation-aware pseudo-label selection method that reads the teacher's confidence regime rather than committing to one rule. It pairs held-out calibration, an unbiased per-class noise estimate, with a self-adaptive confidence floor that provably bounds retention away from 1, and combines them in a one-pass gate: measure the reliability of the teacher's confident set, pi_kept = Pr[correct | c >= tau], on a held-out slice, and filter strictly when it meets the confidence demanded (pi_kept >= tau), falling back to the adaptive floor otherwise. The boundary is the pre-existing operating threshold, not a value tuned to mIoU, and across six DINOv2 teachers it makes the correct strict-vs-floor call blind. CW-BASS v2 thus recovers the UniMatch V2 operating point on the saturated benchmarks by selecting strict (Pascal VOC 1/8 87.4 against its reported 87.9; Cityscapes within 0.5), and improves on it where the confident set is unreliable (pi_kept ~ 89%, ADE20K), where the floor edges ahead (+1.5 mIoU, single seed). The gate is principled because the failure it avoids is measured, not assumed: on a reliable, saturated teacher the confidence distribution's dynamic range collapses (98% of Pascal pixels >= 0.95), so an adaptive cutoff floods the retention mask and self-training decays into confirmation bias.
Chinese Translation
半监督语义分割长期以来面临一个问题，即信任哪些伪标签，而一系列选择规则、动态阈值、按类别的课程和软置信权重为当时的噪声和不自信的ResNet教师提供了答案。自监督基础编码器改变了这一局面：在DINOv2教师下，置信度饱和，因此帮助弱教师的过滤可能会对强教师造成伤害。我们提出了CW-BASS v2，一种基于饱和度的伪标签选择方法，它读取教师的置信度状态，而不是依赖于单一规则。它将保留的校准、无偏的按类别噪声估计与自适应置信度下限相结合，该下限可以证明地将保留率限制在1以下，并在一次性门控中结合这两者：在保留的切片上测量教师置信集的可靠性，pi_kept = Pr[correct | c >= tau]，并在满足所需置信度时严格过滤（pi_kept >= tau），否则回退到自适应下限。边界是预先存在的操作阈值，而不是针对mIoU调整的值，在六个DINOv2教师中，它在盲目情况下做出了正确的严格与下限的选择。CW-BASS v2因此通过选择严格的（Pascal VOC 1/8 87.4与其报告的87.9相比；Cityscapes在0.5以内）恢复了饱和基准上的UniMatch V2操作点，并在置信集不可靠的地方（pi_kept ~ 89%，ADE20K）有所改进，那里下限领先（+1.5 mIoU，单种子）。该门控是有原则的，因为它避免的失败是可测量的，而不是假设的：在可靠的饱和教师上，置信度分布的动态范围崩溃（98%的Pascal像素 >= 0.95），因此自适应截止会淹没保留掩码，自我训练退化为确认偏见。
cs.CV / 21 / 2608.12780
SCOPE: Subspace Clustering with Online Per-Head Top-K Estimation for Sparse Video Attention
SCOPE：基于在线每头 Top-K 估计的稀疏视频注意力子空间聚类
Abstract
Diffusion Transformers (DiTs) incur quadratic self-attention cost over spatiotemporal tokens. Existing training-free sparse attention methods often construct sparse masks from block-level or cluster-level proxy scores, which can obscure fine-grained differences among keys and miss high contribution keys under aggressive sparsity. Moreover, such proxy scores may yield overly concentrated softmax distributions, causing Top-$p$ to retain too few keys for some query clusters. Although a fixed Top-$k$ minimum alleviates this failure mode, a shared value cannot adapt to variations across heads and inputs. To address both limitations, we propose SCOPE, a training-free sparse attention framework that combines 3D-RoPE-aligned key subspace clustering with online per-head Top-$k$ estimation for efficient video-DiT inference. SCOPE partitions post-RoPE keys into temporal, height, and width subspaces, clusters them independently, and aggregates the corresponding centroid scores through lookup tables to obtain per key proxy scores for each query cluster. Building on existing hybrid Top-$p$/fixed Top-$k$ selection, SCOPE derives a head-specific Top-$k$ value online by averaging the initial retained key counts within each head, weighted by query cluster size, and selects additional keys only for query clusters whose initial retained key counts fall below this value. Sparse attention is then computed over the selected original keys and values. Across six model--task configurations, SCOPE consistently outperforms existing training-free baselines in both fidelity and latency, achieving up to a $1.99\times$ end-to-end speedup on 720p HunyuanVideo with $28.46$ dB PSNR relative to dense attention.
Chinese Translation
扩散变换器（Diffusion Transformers，DiTs）在时空标记上产生二次自注意力成本。现有的无训练稀疏注意力方法通常从块级或聚类级代理分数构建稀疏掩码，这可能会掩盖键之间的细微差异，并在激进稀疏下错过高贡献键。此外，这些代理分数可能导致过于集中化的 softmax 分布，导致 Top-$p$ 对某些查询聚类保留的键过少。尽管固定的 Top-$k$ 最小值在一定程度上缓解了这一失败模式，但共享值无法适应不同头和输入之间的变化。为了解决这两个限制，我们提出了 SCOPE，这是一种无训练的稀疏注意力框架，结合了 3D-RoPE 对齐的键子空间聚类和在线每头 Top-$k$ 估计，以实现高效的视频 DiT 推理。SCOPE 将后 RoPE 键划分为时间、高度和宽度子空间，独立聚类，并通过查找表聚合相应的质心分数，以获得每个查询聚类的每个键代理分数。在现有的混合 Top-$p$/固定 Top-$k$ 选择的基础上，SCOPE 通过对每个头内初始保留键计数进行加权平均，得出特定于头的在线 Top-$k$ 值，并仅为初始保留键计数低于该值的查询聚类选择额外的键。然后在选定的原始键和值上计算稀疏注意力。在六种模型-任务配置中，SCOPE 在保真度和延迟方面始终优于现有的无训练基线，在 720p HunyuanVideo 上实现了高达 $1.99 imes$ 的端到端加速，相较于稠密注意力的 PSNR 达到 $28.46$ dB。
cs.CV / 22 / 2608.12781
Beyond Correctness: Benchmarking and Aligning Response Behaviors in Hybrid-Thinking MLLMs
超越正确性：混合思维多模态大型语言模型响应行为的基准测试与对齐
Abstract
Hybrid-thinking multimodal large language models (MLLMs) allow a single model to alternate between deliberative thinking and latency-efficient non-thinking inference. Although these modes differ in reasoning budget, their delivered responses should satisfy the same user-facing standard. Correctness alone may not characterize this response quality; we therefore evaluate task accuracy and response-pattern failures as complementary outcomes. We study this gap through \textbf{response-pattern alignment}: whether thinking and non-thinking interfaces preserve acceptable final-response behavior. We introduce \textbf{PatternEval}, a failure-enriched diagnostic benchmark comprising 2,415 multimodal prompts spanning visual perception and grounding, structured image understanding, and multimodal knowledge reasoning. PatternEval tests four recurrent failures: chain-of-thought leakage, response repetition, logical contradiction, and performative reasoning. Response-pattern failures are widespread across models from different providers, with non-thinking inference exhibiting substantially higher failure rates and thereby creating systematic misalignment between thinking and non-thinking interfaces. Motivated by this diagnosis, we develop \textbf{PatternRM}, a response-level reward model, and \textbf{PatternRL}, which introduces pattern-specific penalties during reinforcement learning. Experiments on Qwen3-VL-4B and Qwen3-VL-8B show that incorporating pattern-specific penalties into reinforcement learning can mitigate cross-mode misalignment while incurring a marginal task performance trade-off. Together, PatternEval and PatternRL provide an evaluation-and-training framework for aligning user-visible response patterns across hybrid-thinking interfaces.
Chinese Translation
混合思维多模态大型语言模型（MLLMs）允许单一模型在深思熟虑的思维与延迟高效的非思维推理之间交替。尽管这些模式在推理预算上有所不同，但其提供的响应应满足相同的用户标准。仅仅依靠正确性可能无法充分表征这种响应质量；因此，我们将任务准确性和响应模式失效作为互补结果进行评估。我们通过 extbf{响应模式对齐}来研究这一差距：思维和非思维接口是否保持可接受的最终响应行为。我们引入 extbf{PatternEval}，这是一个包含2415个多模态提示的失效丰富诊断基准，涵盖视觉感知与基础、结构化图像理解和多模态知识推理。PatternEval测试四种常见失效：思维链泄漏、响应重复、逻辑矛盾和表现性推理。不同提供商的模型中，响应模式失效普遍存在，非思维推理的失效率显著更高，从而导致思维与非思维接口之间的系统性不对齐。基于这一诊断，我们开发了 extbf{PatternRM}，一种响应级奖励模型，以及 extbf{PatternRL}，在强化学习中引入模式特定的惩罚。对Qwen3-VL-4B和Qwen3-VL-8B的实验表明，将模式特定的惩罚纳入强化学习可以减轻跨模式不对齐，同时仅带来边际的任务性能权衡。总之，PatternEval和PatternRL提供了一个评估与训练框架，以对齐混合思维接口中的用户可见响应模式。
cs.CV / 23 / 2608.12806
Erase but Preserve: Controllable Removal of Copyrighted Animation Characters via Optimized Semantic Anchors
擦除但保留：通过优化语义锚点可控地移除版权动画角色
Abstract
The exceptional generation capabilities of text-to-image diffusion models have raised copyright concerns, particularly the unauthorized reproduction of animation characters. Existing concept erasure methods fall short for animation character erasure: model modification methods struggle to identify suitable anchors for diverse, highly distinctive characters; prompt-based steering methods lack fine-grained control for precise intervention. These approaches often yield incomplete erasure and degraded image fidelity, hindering real-world deployment. In this paper, we propose a controllable method operating on the model's continuous textual representation to erase target characters during generation. We optimizes an anchor embedding via structural and detailed constraints to serve as a character surrogate, then replaces target-related embeddings with the anchor via a structure-aware adaptive strategy. Experiments show that our method achieves state-of-the-art erasure effectiveness and image fidelity preservation, while supporting controllable erasure degree, multi-target removal, and model transferability. Moreover, our optimized anchors are plug-and-play with current model modification baselines to improve their erasure performance.
Chinese Translation
文本到图像扩散模型卓越的生成能力引发了版权问题，特别是对动画角色的未经授权复制。现有的概念擦除方法在动画角色的擦除上表现不佳：模型修改方法难以为多样且高度独特的角色识别合适的锚点；基于提示的引导方法缺乏精细控制以实现精确干预。这些方法往往导致擦除不完全和图像保真度下降，阻碍了实际应用。在本文中，我们提出了一种可控的方法，利用模型的连续文本表示在生成过程中擦除目标角色。我们通过结构和细节约束优化一个锚嵌入，作为角色的替代物，然后通过结构感知的自适应策略用锚点替换与目标相关的嵌入。实验表明，我们的方法在擦除效果和图像保真度保持方面达到了最先进的水平，同时支持可控的擦除程度、多目标移除和模型可迁移性。此外，我们优化的锚点可以与当前的模型修改基线无缝结合，以提高其擦除性能。
cs.CV / 24 / 2608.12811
Structured Local Differential Modeling for AI-Generated Image Detection
用于AI生成图像检测的结构化局部差异建模
Abstract
The rapid advancement of AI-generated content has made the reliable detection of generated images an increasingly critical challenge. Existing detection methods are often dominated during training by semantically salient components with high signal-to-noise ratios (SNRs), thereby suppressing subtler forensic cues associated with the underlying generation mechanisms and embedded in low-level statistical structures. From an information-theoretic perspective, we present a key insight: effective detection in the low-level statistical space requires mitigating the dominance of semantic components while emphasizing and amplifying responses to low-SNR forgery traces. Building on this insight, we propose RippleNet, an AI-generated image detection framework based on local differential signals. RippleNet adaptively identifies forgery-sensitive regions and constructs multi-directional, multi-scale differential representations within local neighborhoods, explicitly characterizing anomalous patterns in neighborhood statistics. More importantly, we refine the attention mechanism to operate within the local differential representation space, enabling the model to establish explicit dependencies at a finer statistical granularity. This design facilitates the capture of pixel-level forgery traces that are difficult to model using conventional convolutions or image-wide patch-level attention. Extensive experiments on multiple public benchmarks and under cross-generator evaluation settings demonstrate that RippleNet achieves consistently competitive performance.
Chinese Translation
AI生成内容的快速发展使得可靠检测生成图像成为一个日益重要的挑战。现有的检测方法在训练过程中往往受到语义显著成分的主导，这些成分具有较高的信噪比（SNR），从而抑制了与潜在生成机制相关的、更微妙的取证线索，这些线索嵌入在低级统计结构中。从信息论的角度出发，我们提出了一个关键见解：在低级统计空间中有效检测需要减轻语义成分的主导性，同时强调和放大对低SNR伪造痕迹的响应。在此见解的基础上，我们提出了RippleNet，这是一种基于局部差异信号的AI生成图像检测框架。RippleNet自适应地识别伪造敏感区域，并在局部邻域内构建多方向、多尺度的差异表示，明确表征邻域统计中的异常模式。更重要的是，我们优化了注意力机制，使其在局部差异表示空间内运行，从而使模型能够在更细的统计粒度上建立明确的依赖关系。这一设计有助于捕捉难以通过传统卷积或图像范围的块级注意力建模的像素级伪造痕迹。在多个公共基准和跨生成器评估设置下的广泛实验表明，RippleNet始终实现了具有竞争力的性能。
cs.CV / 25 / 2608.12825
LocusGS: Spatially Grounded Tokens for Feed-Forward 3D Gaussian Splatting
LocusGS：用于前馈3D高斯点云的空间锚定标记
Abstract
Recent query-based feed-forward 3DGS methods represent a scene using learnable queries, each aggregating multi-view evidence and decoding a group of Gaussians. Ideally, different queries should specialize in coherent local regions of the scene. However, we observe that Gaussians decoded from the same query often scatter across distant scene regions, resulting in weak query-level spatial coherence and poor alignment with the scene structure. We attribute this behavior to the purely latent representation of existing Gaussian queries. To address this limitation, we introduce LocusGS, which augments each Gaussian query with a 3D anchor state consisting of a center and a support radius. The anchor state is progressively refined across decoder layers and is used throughout query interaction, multi-view feature aggregation, and Gaussian generation. Specifically, an anchor-to-ray geometric bias guides each query toward spatially relevant image observations, while anchor-centered decoding organizes its Gaussians within a local region. Experiments on novel view synthesis benchmarks show that LocusGS improves rendering quality over query-based Gaussian token baselines under the same Gaussian budget. Further analysis shows that the learned anchors form coherent spatial layouts and lead to more structured Gaussian distributions, demonstrating that explicit anchor states improve the spatial organization. Our project page: https://leo-frank.github.io/LocusGS_viewer.
Chinese Translation
最近的基于查询的前馈3DGS方法使用可学习的查询来表示场景，每个查询聚合多视角证据并解码一组高斯分布。理想情况下，不同的查询应该专注于场景的连贯局部区域。然而，我们观察到，从同一查询解码的高斯分布往往散布在远离的场景区域，导致查询级别的空间一致性较弱，并且与场景结构的对齐效果不佳。我们将这种现象归因于现有高斯查询的纯潜在表示。为了解决这一局限性，我们引入了LocusGS，它通过一个包含中心和支持半径的3D锚定状态来增强每个高斯查询。锚定状态在解码器层中逐步细化，并在查询交互、多视角特征聚合和高斯生成过程中使用。具体而言，锚定到光线的几何偏差引导每个查询朝向空间相关的图像观测，而以锚定为中心的解码则将其高斯分布组织在局部区域内。在新视图合成基准上的实验表明，LocusGS在相同的高斯预算下提高了渲染质量，超越了基于查询的高斯标记基线。进一步的分析表明，学习到的锚定形成了连贯的空间布局，并导致更有结构的高斯分布，证明了显式锚定状态改善了空间组织。我们的项目页面：https://leo-frank.github.io/LocusGS_viewer。
cs.CV / 26 / 2608.12827
Validation of Smartphone-Based Photogrammetric 3D Body Scanning for Automated Anthropometric Measurements Compared with a Commercial Depth-Sensor-Based Body Scanner
基于智能手机的摄影测量3D身体扫描在自动化人体测量中的验证：与商业深度传感器身体扫描仪的比较
Abstract
3D body scanning has become an important tool in healthcare applications because of its rapid and non-invasive nature. While smartphone-based photogrammetric reconstruction provide a low-cost and accessible alternative to commercial 3D body scanners, their performance for whole-body scanning remains insufficiently validated. Thus, we designed this study to comprehensively validate the photogrammetric 3D scanning application by evaluating automatically extracted whole-body measurements and longitudinal body-shape monitoring. We evaluated a representative application, PolyCam, against the commercial depth-sensor-based Fit3D ProScanner using 144 pregnant participants scanned longitudinally throughout pregnancy. We designed an automatic circumference extraction pipeline to get measurements at four anatomical landmarks from paired 3D scans. A linear mixed-effects model was used to evaluate scanner effects and longitudinal body-shape changes. Measurement consistency was assessed using repeated PolyCam scans and tape measurements on a rigid mannequin. PolyCam demonstrated strong agreement with Fit3D, with average biases below 16 mm, intraclass correlation coefficients above 0.8, and Pearson correlation coefficients above 0.9 across all landmarks. Both systems captured comparable longitudinal body-shape changes. Mannequin experiments showed mean biases below 3.5 mm and no significant differences from tape measurements. These findings support smartphone photogrammetry as a potential accessible alternative to commercial body scanners and applicable for longitudinal 3D body-shape assessment.
Chinese Translation
3D身体扫描因其快速和非侵入性的特性，已成为医疗应用中的重要工具。虽然基于智能手机的摄影测量重建为商业3D身体扫描仪提供了一种低成本且易于获取的替代方案，但其在全身扫描中的性能尚未得到充分验证。因此，我们设计了本研究，以全面验证摄影测量3D扫描应用，通过评估自动提取的全身测量和纵向身体形状监测。我们对代表性应用PolyCam进行了评估，比较其与商业深度传感器基础的Fit3D ProScanner在144名孕妇中的纵向扫描表现。我们设计了一个自动周长提取流程，以从配对的3D扫描中获取四个解剖标志点的测量值。采用线性混合效应模型评估扫描仪效应和纵向身体形状变化。通过对刚性人偶进行重复的PolyCam扫描和测量带测量评估测量一致性。PolyCam与Fit3D表现出强一致性，平均偏差低于16毫米，组内相关系数超过0.8，皮尔逊相关系数超过0.9，适用于所有标志点。两个系统捕捉到的纵向身体形状变化相当。人偶实验显示平均偏差低于3.5毫米，与测量带测量没有显著差异。这些发现支持智能手机摄影测量作为商业身体扫描仪的潜在可获取替代方案，并适用于纵向3D身体形状评估。
cs.CV / 27 / 2608.12829
Semantic Steering for Controllable Generation: Tuning-Free Concept Erasure in Multimodal Diffusion Transformers
可控生成的语义引导：多模态扩散变换器中的无调优概念消除
Abstract
Multimodal Diffusion Transformers (MM-DiTs) have demonstrated remarkable text-to-image generation performance, surpassing traditional U-Net-based diffusion models. Nevertheless, their powerful generative capabilities also raise significant safety concerns, as they may generate sensitive or inappropriate content. While existing concept erasure methods aim to mitigate such risks, most require modifying model parameters, which are often architecture-specific and impractical for deployed larger models. Several tuning-free approaches face challenges when applied to advanced large-scale MM-DiTs due to their deeply embedded knowledge, broad semantic space, and context-dependent text encoders. To address these challenges, we propose to erase concepts by directly manipulating the model's internal representations. Our key insight, derived from an in-depth analysis of MM-DiT's block-wise generative roles, is that text-conditioned semantic representations are most salient in the middle blocks of MM-DiTs. Based on this, we extract representations of an unwanted concept and a desirable safe one from the middle block, construct a steering vector from their difference, and inject this single vector into consecutive early and middle blocks. By operating exclusively on the sparse text-branch tokens and leveraging the straight sampling trajectory of rectified flow, our method achieves effective concept erasure with negligible overhead and without any training. Extensive experiments across MM-DiT models demonstrate that our method achieves state-of-the-art performance in erasing diverse concepts, enables effective control over the final output, and remains robust to adversarial attacks.
Chinese Translation
多模态扩散变换器（MM-DiTs）在文本到图像生成方面表现出色，超越了传统的基于U-Net的扩散模型。然而，它们强大的生成能力也引发了显著的安全隐患，因为它们可能生成敏感或不当内容。虽然现有的概念消除方法旨在减轻这些风险，但大多数方法需要修改模型参数，而这些参数通常是特定于架构的，对于已部署的大型模型而言并不实用。由于深度嵌入的知识、广泛的语义空间和上下文相关的文本编码器，几种无调优的方法在应用于先进的大规模MM-DiTs时面临挑战。为了解决这些挑战，我们提出通过直接操控模型的内部表示来消除概念。我们的关键见解源于对MM-DiT的块级生成角色的深入分析，即文本条件的语义表示在MM-DiTs的中间块中最为显著。基于此，我们从中间块中提取不需要的概念和期望的安全概念的表示，构建它们之间的差异作为引导向量，并将该单一向量注入到连续的早期和中间块中。通过仅对稀疏的文本分支标记进行操作，并利用修正流的直接采样轨迹，我们的方法实现了有效的概念消除，几乎没有开销且无需任何训练。对MM-DiT模型的广泛实验表明，我们的方法在消除多样化概念方面达到了最先进的性能，使得对最终输出的有效控制成为可能，并且对对抗攻击保持鲁棒性。
cs.CV / 28 / 2608.12843
Heterogeneous Vision-Language Ensemble with Disagreement-Aware Reranking for Text-Based Person Anomaly Retrieval
基于文本的人物异常检索的异构视觉-语言集成与考虑分歧的重新排序
Abstract
Text-based person anomaly retrieval aims to retrieve pedestrians exhibiting anomalous behaviors from a large image gallery using natural language descriptions. Compared with conventional text-based person retrieval, this task requires fine-grained reasoning over pedestrian appearance, behaviors, object interactions, and scene context, making robust cross-modal matching significantly more challenging. This paper presents the GENAI4E team's solution to AI City Challenge 2026 Track 4. Our framework builds upon a strong retrieval backbone and progressively integrates heterogeneous vision-language embedding models through score alignment and iterative ensemble fusion, followed by disagreement-aware VLM reranking for ambiguous queries. On the official Pedestrian Anomaly Behavior (PAB) benchmark, our approach achieves 90.92% mAP, 85.13% Recall@1, 97.72% Recall@5, and 98.68% Recall@10, demonstrating the effectiveness of combining complementary vision-language representations with selective multimodal reasoning for large-scale text-based person anomaly retrieval.
Chinese Translation
基于文本的人物异常检索旨在通过自然语言描述从大型图像库中检索出表现出异常行为的行人。与传统的基于文本的人物检索相比，这项任务需要对行人的外观、行为、物体交互和场景上下文进行细致的推理，从而使得稳健的跨模态匹配变得更加具有挑战性。本文介绍了GENAI4E团队在AI City Challenge 2026 Track 4中的解决方案。我们的框架基于强大的检索骨干，并通过分数对齐和迭代集成融合逐步整合异构的视觉-语言嵌入模型，随后针对模糊查询进行考虑分歧的视觉-语言模型（VLM）重新排序。在官方的行人异常行为（PAB）基准上，我们的方法达到了90.92%的平均精度（mAP）、85.13%的召回率@1、97.72%的召回率@5和98.68%的召回率@10，证明了将互补的视觉-语言表征与选择性多模态推理相结合在大规模基于文本的人物异常检索中的有效性。
cs.CV / 29 / 2608.12876
SPARED: Reasoning-Based AI-Generated Image Detection via Adversarially Edited Data
SPARED：基于推理的AI生成图像检测通过对抗编辑数据
Abstract
Detecting AI-generated images is only half the task: a deployed detector must also justify its verdict, yet existing detectors inherit three failure modes from their training data: real and fake images collected from different sources invite provenance shortcuts, supervised explanation corpora teach templated rationales, and a static forgery corpus leaves the decision boundary standing still while generators keep moving. We introduce \methodname{}, an adversarial reinforcement learning framework that pits two heterogeneous models against each other. A diffusion image editor learns to edit real photographs into fake counterparts of those same photographs that fool the current detector, while a reasoning MLLM learns to expose them with a verdict grounded in free-form reasoning. Both rewards are shortcut-proof by design: the attacker is credited only when its edit is faithfully executed, and the defender only when its verdict is correct. As the two models alternate, each round's attacker regenerates a harder training pool aimed at the current detector's blind spots, so the detector must generalize rather than memorize any fixed artifact distribution. Although the explanation is never rewarded, its quality rises round over round as a side effect of accuracy-only training. A detector trained within this loop improves monotonically across rounds on each of three external benchmarks.
Chinese Translation
检测AI生成的图像仅仅是任务的一半：部署的检测器还必须为其判决提供合理依据，然而现有的检测器从其训练数据中继承了三种失败模式：来自不同来源的真实和虚假图像引发了来源捷径，监督解释语料库教授了模板化的推理，而静态伪造语料库使得决策边界静止不动，而生成器却在不断变化。我们引入了 extit{SPARED}，一个对抗性强化学习框架，将两个异构模型相互对抗。一个扩散图像编辑器学习将真实照片编辑成那些能够欺骗当前检测器的虚假对应物，而一个推理的多语言大型模型（MLLM）学习通过自由形式的推理揭露它们。两个奖励在设计上都是防捷径的：攻击者只有在其编辑被忠实执行时才会获得积分，而防御者只有在其判决正确时才会得分。随着两个模型的交替进行，每轮的攻击者重新生成一个更难的训练池，旨在当前检测器的盲点，因此检测器必须进行概括，而不是记忆任何固定的伪造分布。尽管解释从未获得奖励，但其质量随着每轮的进行因仅依赖准确性的训练而逐渐提高。在这个循环中训练的检测器在三个外部基准上每轮都单调改善。
cs.CV / 30 / 2608.12898
NaviDC-OCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
NaviDC-OCR：在数字文档和相机捕获文档中导航文档解析
Abstract
Document parsing aims to transform unstructured documents into structured and machine-readable representations. Recent advances in Vision-Language Models (VLMs) have significantly advanced document parsing. However, existing approaches still face two major challenges. First, decoupled VLM-based methods heavily rely on accurate layout analysis, where geometric distortions in camera-captured documents can introduce cascading errors. Second, although end-to-end VLM-based methods alleviate the dependence on explicit layout detection, they often suffer from redundant generation, hallucinations, and insufficient structural reasoning in high-resolution scenarios. To address these challenges, we propose NaviDC-OCR, a unified framework for document parsing. NaviDC-OCR introduces deformation-aware learning to incorporate geometric perception into VLMs and proposes an adaptive sampling mechanism for complex layout representation. Furthermore, a content-structure decoupled learning strategy is developed to explicitly model formula grammars and table structures, enabling more effective structured representation learning. Extensive experiments demonstrate that NaviDC-OCR achieves state-of-the-art performance across diverse document parsing benchmarks. It obtains overall scores of 96.87, 88.53 and 78.41 on OmniDocBench v1.6, Wild-OmniDocBench, and PureDocBench, respectively, and ranks first in the ICDAR 2026 Sci-ImageMiner Challenge. These results validate the effectiveness and generalization capability of NaviDC-OCR in complex document parsing scenarios.
Chinese Translation
文档解析旨在将非结构化文档转化为结构化和机器可读的表示。近期视觉-语言模型（Vision-Language Models, VLMs）的进展显著推动了文档解析的发展。然而，现有方法仍面临两个主要挑战。首先，解耦的基于VLM的方法在很大程度上依赖于准确的布局分析，而相机捕获文档中的几何失真可能引入级联错误。其次，尽管端到端的基于VLM的方法减轻了对显式布局检测的依赖，但它们在高分辨率场景中往往会遭遇冗余生成、幻觉和不足的结构推理。为了解决这些挑战，我们提出了NaviDC-OCR，一个统一的文档解析框架。NaviDC-OCR引入了变形感知学习，将几何感知融入VLM，并提出了一种自适应采样机制以表示复杂布局。此外，开发了一种内容-结构解耦学习策略，以显式建模公式语法和表格结构，从而实现更有效的结构化表示学习。大量实验表明，NaviDC-OCR在多样的文档解析基准测试中达到了最先进的性能。在OmniDocBench v1.6、Wild-OmniDocBench和PureDocBench上分别获得了96.87、88.53和78.41的总体得分，并在ICDAR 2026 Sci-ImageMiner挑战赛中排名第一。这些结果验证了NaviDC-OCR在复杂文档解析场景中的有效性和泛化能力。
cs.CV / 31 / 2608.12904
HounsWorld: A Multimodal World Model for Hidden Patient-State Readout, Reconstruction, and Simulation
HounsWorld：一种用于隐藏患者状态读取、重建和模拟的多模态世界模型
Abstract
Clinical intelligence requires estimating a patient's underlying condition from incomplete observations rather than learning isolated mappings from scans to answers. Volumetric medical images provide dense observations of anatomy, attenuation, and lesions, whereas clinical language provides sparse but complementary semantic observations. We formulate CT-centered intelligence as inference over a shared latent patient state, under which readout, reconstruction, and simulation all become state-dependent prediction problems. To operationalize this view, we introduce HounsBench, a computed tomography (CT) centric patient-state benchmark that unifies these three task families with patient-disjoint splits and per-family metrics, and HounsWorld, a 3B multimodal world model that treats volumetric scans and language as observations of the shared state through Joint Understanding-Generation Learning. A shared transformer forms an implicit patient-state estimate and supports three outputs: query-conditioned answers that read out the state, reports and captions that reconstruct it in language, and condition-specific CT volumes for low-dose denoising, virtual contrast enhancement, and anatomy-constrained text-and-mask-to-volume generation. Zero-initialized CT adapters preserve pretrained multimodal mappings, while condition-explicit Hounsfield-unit window sampling exposes clinically meaningful density observations. HounsWorld shows strong performance across all three task families while consistently improving CT understanding through clinically structured completion. Our project is available at https://github.com/byhwhite/HounsWorld.git
Chinese Translation
临床智能需要从不完整的观察中估计患者的潜在状态，而不是学习从扫描到答案的孤立映射。体积医学图像提供了对解剖结构、衰减和病变的密集观察，而临床语言则提供了稀疏但互补的语义观察。我们将以CT为中心的智能形式化为对共享潜在患者状态的推断，在这种状态下，读取、重建和模拟都成为依赖状态的预测问题。为了实现这一观点，我们引入了HounsBench，这是一个以计算机断层扫描（CT）为中心的患者状态基准，统一了这三类任务，采用患者不重叠的划分和每类任务的指标，以及HounsWorld，一个3B多模态世界模型，将体积扫描和语言视为通过联合理解-生成学习对共享状态的观察。一个共享的变换器形成了隐式的患者状态估计，并支持三种输出：查询条件的答案用于读取状态，报告和标题用于用语言重建状态，以及特定条件的CT体积用于低剂量去噪、虚拟对比增强，以及受解剖约束的文本和掩码到体积的生成。零初始化的CT适配器保留了预训练的多模态映射，而条件显式的Hounsfield单位窗口采样则揭示了临床上有意义的密度观察。HounsWorld在所有三类任务中表现出色，同时通过临床结构化的完成持续改善CT理解。我们的项目可在 https://github.com/byhwhite/HounsWorld.git 获取。
cs.CV / 32 / 2608.12911
Beyond Visual Evidence: Revealing and Mitigating Relational Privacy Leakage in Document MLLMs
超越视觉证据：揭示和缓解文档多模态大语言模型中的关系隐私泄露
Abstract
While the privacy risks of multimodal large language models (MLLMs) have drawn significant attention, the unique vulnerabilities of domain-specific MLLMs remain largely underexplored. Focusing on document understanding MLLMs for identity document processing, this paper investigates the privacy issues inherent in Key Information Extraction (KIE) tasks. We reveal that when input images lack sufficient visual evidence, these models often rely on memorized field relations from training data to infer missing content, thereby leaking multiple correlated fields containing sensitive personal information. To mitigate this risk, we make three key contributions.First, we propose the Dynamic Relational Unlearning Framework (DRUF) which comprises a Relational Decoupling Unlearning (RDU) module and a dynamic set update mechanism. It suppresses the leakage of high-risk field pairs while preserving KIE performance.Second, we introduce DocPrivacyBench, a novel benchmark to systematically evaluate a model's susceptibility to privacy leakage under conditions of absent or minimal visual evidence.Third, we evaluate three MLLMs and six unlearning methods using this benchmark, assessing both post-unlearning leakage suppression and utility preservation.Our results demonstrate that existing MLLMs consistently exhibit privacy leakage when visual evidence is scarce, particularly on noisier datasets. In contrast, DRUF outperforms the strongest baseline by improving leakage suppression by 4.8 percentage points, effectively mitigating privacy risks while maintaining robust document information extraction performance.
Chinese Translation
尽管多模态大语言模型（MLLMs）的隐私风险引起了广泛关注，但特定领域的MLLMs所面临的独特脆弱性仍然未得到充分探讨。本文聚焦于身份文件处理的文档理解MLLMs，研究了关键资讯提取（Key Information Extraction, KIE）任务中固有的隐私问题。我们揭示，当输入图像缺乏足够的视觉证据时，这些模型往往依赖于训练数据中记忆的字段关系来推断缺失内容，从而泄露多个包含敏感个人信息的相关字段。为了缓解这一风险，我们做出了三项关键贡献。首先，我们提出了动态关系去学习框架（Dynamic Relational Unlearning Framework, DRUF），该框架包括一个关系解耦去学习模块（Relational Decoupling Unlearning, RDU）和一个动态集合更新机制。它在保持KIE性能的同时抑制高风险字段对的泄露。其次，我们引入了DocPrivacyBench，这是一个新颖的基准，用于系统评估模型在缺乏或最小视觉证据条件下的隐私泄露易感性。第三，我们使用该基准评估了三种MLLMs和六种去学习方法，评估了去学习后的泄露抑制和效用保持。我们的结果表明，现有的MLLMs在视觉证据稀缺时始终表现出隐私泄露，特别是在噪声较大的数据集上。相比之下，DRUF通过提高泄露抑制4.8个百分点，超越了最强基线，有效缓解了隐私风险，同时保持了稳健的文档信息提取性能。
cs.CV / 33 / 2608.12920
TennisVAR: A Stroke-Evidence-Grounded Multimodal Large Language Model for Tactical Reasoning in Tennis Videos
TennisVAR：一种基于击球证据的多模态大型语言模型，用于网球视频中的战术推理
Abstract
Sports-video understanding is moving beyond event recognition toward explaining how actions collectively shape match progression, however, existing tennis-video methods either perceive individual strokes without modeling their tactical dependencies or generate high-level analyses without grounding them in the underlying events. To bridge this perception-to-understanding gap, we formulate stroke-evidence-grounded tactical reasoning, a new rally-level task that requires models to jointly predict an open-ended answer, a hierarchical tactic label, an ordered sequence of supporting strokes, and decisive key actions, with each evidence stroke anchored to its racket-ball contact frame. We further introduce TRACE (Tactical Reasoning with Action-Chain Evidence in Tennis), a large-scale expert-annotated benchmark containing 11,189 rally videos, 41,485 stroke events, 25,429 tactical units, and 11,189 question-answer pairs, which unifies fine-grained stroke attributes, cross-stroke tactical relations, hierarchical tactic annotations, and evidence-grounded questions across factual perception, tactical understanding, and decision reasoning. Building on TRACE, we propose TennisVAR (Tennis Video Action-chain Reasoner), an evidence-grounded multimodal large language model that follows an "event-relation-evidence-tactic" reasoning paradigm, where an Event Parsing Module converts continuous rallies into explicit stroke-event sequences while a Tactical Graph-Guided Temporal Reasoner jointly models rally progression and same-player decision dependencies to identify question-relevant evidence and decisive actions.
Chinese Translation
体育视频理解正从事件识别向解释动作如何共同塑造比赛进程转变，然而，现有的网球视频方法要么只关注单一击球而未建模其战术依赖关系，要么生成高层次的分析而未将其与基础事件相结合。为了弥合这种感知与理解之间的差距，我们提出了基于击球证据的战术推理，这是一项新的回合级任务，要求模型共同预测开放式答案、分层战术标签、一系列支持性击球的有序序列以及决定性关键动作，每个证据击球都与其球拍-球接触帧相锚定。我们进一步引入TRACE（Tactical Reasoning with Action-Chain Evidence in Tennis），这是一个大规模专家注释基准，包含11,189个回合视频、41,485个击球事件、25,429个战术单元和11,189个问答对，统一了细粒度的击球属性、跨击球的战术关系、分层战术注释以及基于证据的问题，涵盖了事实感知、战术理解和决策推理。基于TRACE，我们提出了TennisVAR（Tennis Video Action-chain Reasoner），一种基于证据的多模态大型语言模型，遵循“事件-关系-证据-战术”的推理范式，其中事件解析模块将连续的回合转换为明确的击球事件序列，而战术图引导的时间推理器共同建模回合进程和同一球员的决策依赖关系，以识别与问题相关的证据和决定性动作。
cs.CV / 34 / 2608.12960
A Deep RL based Framework for Targeted White Matter Tractography
基于深度强化学习的目标白质纤维追踪框架
Abstract
Fiber tractography's ability to reconstruct the brain's structural pathways, has made it a crucial component of modern neuroimaging, enabling detailed, non-invasive mapping of structural connectivity and supporting a wide range of neurological research and clinical applications. However, despite its importance, tractography remains a challenging task due to the inherent complexity of white matter structure and its susceptibility to false positives, which can lead to the misrepresentation of critical pathways. To overcome these limitations, in this thesis, we propose a hybrid framework that integrates reinforcement learning with supervised learning for refining RL policies, specifically tailored for tract-specific tractography. Notably, our framework does not rely on ground-truth fibers for training. Moreover, the tract-specific formulation bypasses the need for an explicit segmentation process, simplifying the overall pipeline. Our work includes two main contributions, each building upon the previous. First, we introduce a hybrid approach that combines reinforcement learning with supervised learning (specifically, GPT-based policy learning) to refine policies in a tract-specific context. Second, we propose a scalable framework for data-driven multi-policy fusion, which leverages the complementary strengths of multiple RL policies to improve tractography performance and robustness. We demonstrate the effectiveness of our framework through extensive validation on benchmark public datasets including TractoInferno, HCP, and ISMRM-2015, highlighting its ability to generalize across data sources and accurately reconstruct brain white matter tracts. We believe that these contributions represent significant advancements in the field of tractography, improving robustness, reliability, and accuracy while reducing dependence on ground-truth annotations.
Chinese Translation
纤维追踪重建大脑结构通路的能力，使其成为现代神经成像的重要组成部分，能够实现结构连通性的详细、非侵入性映射，并支持广泛的神经科学研究和临床应用。然而，尽管其重要性，纤维追踪仍然是一项具有挑战性的任务，原因在于白质结构的固有复杂性及其对假阳性的敏感性，这可能导致关键通路的错误表征。为了克服这些局限性，在本论文中，我们提出了一种混合框架，将强化学习与监督学习相结合，以优化针对特定纤维的强化学习策略。值得注意的是，我们的框架在训练中不依赖于真实纤维。此外，特定纤维的公式化绕过了显式分割过程的需要，简化了整体流程。我们的工作包括两个主要贡献，每个贡献都建立在前一个基础上。首先，我们介绍了一种混合方法，将强化学习与监督学习（具体而言，基于GPT的策略学习）结合，以在特定纤维的背景下优化策略。其次，我们提出了一种可扩展的数据驱动多策略融合框架，利用多个强化学习策略的互补优势来提高纤维追踪的性能和鲁棒性。通过在包括TractoInferno、HCP和ISMRM-2015等基准公共数据集上的广泛验证，我们展示了框架的有效性，突显其在不同数据源间的泛化能力和准确重建大脑白质纤维的能力。我们相信，这些贡献代表了纤维追踪领域的重要进展，提高了鲁棒性、可靠性和准确性，同时减少了对真实标注的依赖。
cs.CV / 35 / 2608.12971
Bias Mitigation in Face Recognition via Demographic-based Supervised Contrastive Learning
通过基于人口统计的监督对比学习减轻人脸识别中的偏差
Abstract
Face recognition systems have been shown to be biased toward certain demographic groups by exhibiting different error rates across gender, age, or ethnicity. Though the imbalance of the training data with respect to these demographics is one cause of this bias, training on artificially balanced groups does not completely mitigate the problem. For deployment, face recognition typically works at operating points allowing very low false match rates and, hence, on the tail of the non-match score distribution. While class balancing can improve the means of these distributions, the aim of our approach is to improve fairness by addressing the behavior in the tail. Particularly, we propose the Demographic-based Supervised Contrastive loss (DeSCon) for face recognition, which relies on a well-designed composition of training batches and demographic-aware pair selection. Our experimental evaluation on both demographically-labeled datasets and standard verification benchmarks shows that DeSCon can improve fairness beyond balancing training datasets while maintaining competitive verification performance. Source code is available upon request.
Chinese Translation
人脸识别系统已被证明对某些人口统计群体存在偏见，表现为在性别、年龄或种族方面的错误率不同。尽管与这些人口统计特征相关的训练数据不平衡是导致这种偏见的一个原因，但在人工平衡的群体上训练并不能完全缓解这一问题。在实际应用中，人脸识别通常在允许非常低的错误匹配率的操作点下工作，因此，主要集中在非匹配得分分布的尾部。虽然类别平衡可以改善这些分布的均值，但我们的方法旨在通过解决尾部的行为来提高公平性。特别地，我们提出了用于人脸识别的基于人口统计的监督对比损失（Demographic-based Supervised Contrastive loss, DeSCon），该方法依赖于精心设计的训练批次组合和基于人口统计的配对选择。我们在具有人口统计标签的数据集和标准验证基准上的实验评估表明，DeSCon能够在维持竞争性验证性能的同时，超越平衡训练数据集来提高公平性。源代码可根据请求提供。
cs.CV / 36 / 2608.12980
DiCoR: Decoupled Referent Disambiguation and Contour Recalibration for Efficient Referring Remote Sensing Image Segmentation
DiCoR：用于高效指代遥感图像分割的解耦指代消歧与轮廓重校准
Abstract
Referring remote sensing image segmentation (RRSIS) aims to delineate targets specified by natural language expressions in remote sensing imagery. Existing methods mainly follow joint fusion segmentation (JFS) or decoupled prompt segmentation (DPS). JFS is efficient but often suffers from limited accuracy because referent localization and mask delineation are optimized under a unified objective, whereas DPS separates localization from mask generation using spatial prompts and foundation segmenters at the cost of higher memory consumption and inference latency. To bridge this gap, we propose DiCoR, a decoupled referent disambiguation and contour recalibration framework built on an efficient JFS pipeline. DiCoR addresses two key challenges: distinguishing the correct referent from ambiguous candidates and refining coarse masks after localization. A disambiguation-aware localization guidance strategy ranks salient candidate regions with adaptive linguistic cues and injects the resulting localization prior into fused features. A lightweight contour recalibration module further predicts residual corrections to coarse logits under localized contour supervision, improving mask quality with limited computational overhead. Experiments on RefSegRS, RRSIS-D, and RISBench show that DiCoR achieves the best segmentation accuracy across all three benchmarks. On RefSegRS, it improves mIoU and gIoU by 5.28% and 2.87% over a competitive JFS method while running 4.7% faster than a representative DPS method, demonstrating a favorable accuracy-efficiency trade-off. Code is available at https://github.com/zyGao1126/DiCoR.
Chinese Translation
指代遥感图像分割（RRSIS）旨在根据自然语言表达在遥感图像中勾勒出指定目标。现有方法主要遵循联合融合分割（JFS）或解耦提示分割（DPS）。JFS效率高，但由于指代定位和掩膜勾勒在统一目标下进行优化，往往面临准确性有限的问题；而DPS则通过使用空间提示和基础分割器将定位与掩膜生成分开，代价是更高的内存消耗和推理延迟。为了解决这一问题，我们提出了DiCoR，一个基于高效JFS管道的解耦指代消歧与轮廓重校准框架。DiCoR解决了两个关键挑战：从模糊候选中区分正确的指代对象，以及在定位后细化粗糙的掩膜。一个关注消歧的定位引导策略使用自适应语言线索对显著候选区域进行排序，并将结果定位先验注入融合特征中。一个轻量级的轮廓重校准模块进一步在局部轮廓监督下预测粗糙logits的残差修正，以有限的计算开销提高掩膜质量。在RefSegRS、RRSIS-D和RISBench上的实验表明，DiCoR在所有三个基准测试中实现了最佳的分割准确性。在RefSegRS上，它比一个具有竞争力的JFS方法提高了5.28%的mIoU和2.87%的gIoU，同时比一个代表性的DPS方法运行速度快4.7%，展示了良好的准确性与效率的权衡。代码可在 https://github.com/zyGao1126/DiCoR 获取。
cs.CV / 37 / 2608.12997
PixSDS: Why Latent SDS Makes Noisy Pixels
PixSDS：潜在SDS为何会产生噪声像素
Abstract
Score Distillation Sampling (SDS) enables text-to-3D generation by optimizing rendered images with a pretrained diffusion prior, but latent SDS often produces structured color artifacts and high-frequency texture noise. We identify a failure mode of latent SDS caused by VAE-induced pixel drift: the optimized image can move along pixel-space directions that are weakly constrained by the VAE encoder, so its latent representation remains clean and semantically meaningful while the image itself accumulates visible artifacts. We support this diagnosis with controlled 2D SDS experiments, VAE-only optimization, and a simplified analysis showing that encoder-like latent objectives can amplify image-space noise when the inverse mapping to pixels is underconstrained. Motivated by this observation, we propose PixSDS, a lightweight VAE-consistent gradient repair method. PixSDS decodes a latent SDS lookahead step and uses the decoded image as a clean direction for pixel-space optimization, reducing motion in VAE-inconsistent directions without retraining the diffusion model, changing the renderer, or replacing the SDS objective. Experiments in 2D optimization and text-to-3D generation show that PixSDS substantially reduces structured artifacts while preserving semantic content. Code is publicly available at https://sevashasla.github.io/pixsds-webpage/.
Chinese Translation
得分蒸馏采样（Score Distillation Sampling, SDS）通过优化渲染图像与预训练的扩散先验，实现了文本到3D的生成，但潜在SDS常常会产生结构化的颜色伪影和高频纹理噪声。我们识别出潜在SDS的一种失效模式，这种模式是由变分自编码器（Variational Autoencoder, VAE）引起的像素漂移所导致：优化后的图像可以沿着VAE编码器弱约束的像素空间方向移动，因此其潜在表示保持干净且语义上有意义，而图像本身却积累了可见的伪影。我们通过控制的2D SDS实验、仅VAE优化以及简化分析来支持这一诊断，显示当逆映射到像素的约束不足时，类似编码器的潜在目标会放大图像空间噪声。基于这一观察，我们提出了PixSDS，一种轻量级的VAE一致性梯度修复方法。PixSDS解码潜在SDS的前瞻步骤，并将解码后的图像用作像素空间优化的干净方向，从而在不重新训练扩散模型、不改变渲染器或替换SDS目标的情况下，减少在VAE不一致方向上的运动。2D优化和文本到3D生成的实验表明，PixSDS显著减少了结构化伪影，同时保留了语义内容。代码已公开发布在 https://sevashasla.github.io/pixsds-webpage/。
cs.CV / 38 / 2608.13007
Structure-aware Riemannian Growth Fields for 4D Plant Modeling
结构感知的黎曼增长场用于四维植物建模
Abstract
In this paper, we introduce a novel framework for 4D plant growth modeling that reconstructs the continuous geometric and topological evolution of plants from sparse temporal observations. Existing methods mainly rely on dense registration, yet reliable dense sequences are hard to obtain due to scanning constraints and self-occlusions, leaving these approaches struggling under large temporal gaps where rapid organ emergence violates local rigidity. To overcome this, we bridge these gaps by formulating plant morphogenesis as a continuous procedural process on a structure-aware Riemannian growth field; this jointly models topology evolution and geometric deformation, preserving botanical hierarchies and stable spatio-temporal correspondences across distant timepoints. Our key idea is to ground symbolic growth rules within a continuous geodesic flow, where organ development follows biologically modulated trajectories that preserve structural coherence under topological changes. We further contribute a 10-day dual-species dataset with dense geometric and semantic annotations. Experiments demonstrate that our method accurately tracks individual organ growth over time and significantly outperforms state-of-the-art baselines in both geometric accuracy and correspondence consistency.
Chinese Translation
本文介绍了一种新颖的四维植物生长建模框架，该框架从稀疏的时间观测中重建植物的连续几何和拓扑演变。现有方法主要依赖于密集配准，但由于扫描限制和自遮挡，可靠的密集序列难以获得，这使得这些方法在快速器官出现违反局部刚性的情况下难以应对较大的时间间隔。为了解决这个问题，我们通过将植物形态发生建模为结构感知的黎曼增长场上的连续过程来弥合这些间隙；该方法共同建模拓扑演变和几何变形，保持植物学层次结构和跨越远时间点的稳定时空对应关系。我们的关键思想是在连续测地流中扎根符号生长规则，其中器官发育遵循生物调制的轨迹，在拓扑变化下保持结构一致性。我们还贡献了一个包含密集几何和语义注释的10天双物种数据集。实验表明，我们的方法能够准确跟踪单个器官的生长，并在几何精度和对应一致性方面显著优于现有的最先进基线。
cs.CV / 39 / 2608.13014
EgoPHI: Estimating Contact and Force from Egocentric Vision
EgoPHI：从自我中心视觉估计接触和力
Abstract
Understanding hand-object interaction from egocentric vision is essential for modeling how people physically engage with the surrounding world. Yet reasoning about physically grounded interaction requires estimating the forces acting on hands and objects, beyond localizing contact. We present EgoPHI, the first method that jointly estimates dense contact maps and 3D force distributions on hand and object meshes from a single monocular RGB image and object geometry. To address the lack of scalable ground-truth force annotations, we introduce a physics-based simulation pipeline that augments existing hand-object datasets with dense per-vertex force supervision. EgoPHI then learns dense 3D contact and force on interacting hand and articulated object meshes, extending vision-based force estimation beyond image-space or planar settings. Our evaluation on in-distribution and out-of-distribution benchmarks shows that EgoPHI improves force estimation over existing approaches while generalizing to unseen datasets. To evaluate sim-to-real transfer, we constructed two physical objects that capture dense object contact and force magnitude and used them to record a dataset of interactions from eight participants across diverse touch and grasp types. Our results demonstrate that EgoPHI recovers meaningful 3D contact and force distributions in simulated, out-of-distribution, and real-world settings, advancing egocentric hand-object understanding from contact localization toward physically grounded interaction reasoning.
Chinese Translation
从自我中心视觉理解手与物体的交互对于建模人们如何与周围世界进行物理接触至关重要。然而，推理物理基础的交互需要估计作用于手和物体上的力，而不仅仅是定位接触。我们提出了EgoPHI，这是首个从单个单目RGB图像和物体几何形状中联合估计手和物体网格上的密集接触图和三维力分布的方法。为了解决缺乏可扩展的真实力注释的问题，我们引入了一种基于物理的仿真管道，该管道通过密集的每顶点力监督增强现有的手-物体数据集。EgoPHI随后学习在交互的手和关节物体网格上进行密集的三维接触和力估计，扩展了基于视觉的力估计，超越了图像空间或平面设置。我们在分布内和分布外基准上的评估表明，EgoPHI在力估计方面优于现有方法，同时能够推广到未见数据集。为了评估仿真到现实的迁移，我们构建了两个物理对象，以捕捉密集的物体接触和力大小，并使用它们记录了来自八名参与者的多样化触摸和抓取类型的交互数据集。我们的结果表明，EgoPHI在仿真、分布外和现实世界环境中恢复了有意义的三维接触和力分布，推动了从接触定位到物理基础交互推理的自我中心手-物体理解的进展。
cs.CV / 40 / 2608.13028
RGB-D Video Generation for Improving Human-to-Robot Object Handover Prediction
RGB-D视频生成以改善人机物体交接预测
Abstract
Human-to-robot (H2R) object handover is a fundamental capability for human-robot collaboration, yet progress is hindered by the scarcity of large-scale, human-centric datasets and the significant sim-to-real gap. To address these challenges, we introduce Hand2Bot, an RGB-D video dataset that provides rich contextual information such as body posture and facial expressions, specifically collected for handover scenarios with real-world noise patterns. We further propose PassGen, a generative pipeline that leverages stable video diffusion and an Intention-Aware Temporal Face Encoder to synthesize realistic handover sequences while ensuring hand-object consistency. To bridge the sim-to-real gap, we implement a morphology-based depth editing strategy that replicates realistic sensor noise found in physical depth maps. Experimental evaluations demonstrate that our framework achieves high intention identification accuracy and low false trigger rates in both ablation studies and real-world deployment on a physical robot platform. Our results confirm that training on PassGen allows for robust zero-shot transfer and earlier intention anticipation compared to traditional hand-centric baselines, effectively enabling socially aware robotic behavior in shared workspaces.
Chinese Translation
人机（H2R）物体交接是人机协作的基本能力，但由于缺乏大规模以人为中心的数据集以及显著的仿真与现实之间的差距，进展受到限制。为了解决这些挑战，我们引入了Hand2Bot，一个RGB-D视频数据集，提供丰富的上下文信息，如身体姿势和面部表情，专门为具有现实世界噪声模式的交接场景收集。我们进一步提出了PassGen，一个生成管道，利用稳定的视频扩散和意图感知的时间面部编码器合成逼真的交接序列，同时确保手-物体一致性。为了弥合仿真与现实之间的差距，我们实施了一种基于形态的深度编辑策略，复制物理深度图中发现的真实传感器噪声。实验评估表明，我们的框架在消融研究和在物理机器人平台上的实际部署中都实现了高意图识别准确率和低误触发率。我们的结果确认，在PassGen上训练能够实现稳健的零样本迁移和比传统以手为中心的基线更早的意图预测，有效地促进了共享工作空间中社会意识的机器人行为。
cs.CV / 41 / 2608.13031
UniTraffic-Agent: Unified Traffic Video Reasoning for AI City Challenge 2026 Track 3 with Two Out-of-Domain Evaluations
UniTraffic-Agent：2026年AI城市挑战赛第3赛道的统一交通视频推理，包含两个域外评估
Abstract
Traffic video understanding has become an important problem in intelligent transportation, as road videos provide direct evidence for accidents, violations, and interactions between vehicles and vulnerable road users. A useful system should explain how a traffic event develops, why it happens, and when the relevant interaction occurs, yet this remains difficult for multimodal large language models (MLLMs) because traffic videos contain sparse events and varied viewpoints. We introduce UniTraffic-Agent, the MR-CAS solution for Track~3 of the 10th AI City Challenge, which includes Traffic Anomaly Reasoning (TAR) and two out-of-domain evaluations: FETV for fisheye traffic events and PSI-VQA for pedestrian intention reasoning. UniTraffic-Agent follows an observe--reason--act--verify workflow that samples timestamped visual evidence, reasons over all questions from the same clip in one request, and converts responses through task-specific action adapters. On the official Public leaderboards, MR-CAS ranks 16th on TAR with a score of 0.5780, 2nd on FETV with 0.4884, and 4th on PSI-VQA with 64.4161. The code is available at https://github.com/Roclp/UniTraffic-Agent.
Chinese Translation
交通视频理解已成为智能交通中的一个重要问题，因为道路视频为事故、违规行为以及车辆与脆弱道路使用者之间的互动提供了直接证据。一个有效的系统应能够解释交通事件的发展过程、发生原因以及相关互动发生的时间，然而，对于多模态大型语言模型（MLLMs）来说，这仍然是一个挑战，因为交通视频包含稀疏事件和多样化的视角。我们介绍了UniTraffic-Agent，这是第10届AI城市挑战赛第3赛道的MR-CAS解决方案，包含交通异常推理（TAR）和两个域外评估：用于鱼眼交通事件的FETV和用于行人意图推理的PSI-VQA。UniTraffic-Agent遵循观察-推理-行动-验证的工作流程，采样带时间戳的视觉证据，在一次请求中对同一片段的所有问题进行推理，并通过特定任务的行动适配器转换响应。在官方公共排行榜上，MR-CAS在TAR中排名第16，得分为0.5780，在FETV中排名第2，得分为0.4884，在PSI-VQA中排名第4，得分为64.4161。代码可在https://github.com/Roclp/UniTraffic-Agent获取。
cs.CV / 42 / 2608.13037
Spatially-Grounded Text-to-Video Generation via Inference-Time Gradient-Free Optimization
基于空间的文本到视频生成：推理时无梯度优化
Abstract
Diffusion Transformer Text-to-Video models have achieved remarkable synthesis quality, yet fine-grained spatial controllability remains a significant challenge. While existing training-free methods produce solid overall results in spatially grounded generation, \ie, placing a specific object in a designated location, they rely on gradient-based optimization techniques that incur prohibitive computational overhead, a bottleneck amplified in modern large-scale architectures. To address this limitation, we present Gradient-free Analytical Trajectory Optimization Video Generation (GATO-Vid), a novel training-free and gradient-free approach for precise spatial guidance. Rather than relying on costly backward passes, we introduce an alternative cross-attention score and solve it analytically to obtain an exact, closed-form solution. To use our analytical solution, we propose an on-the-fly injection mechanism tailored to the topological manifold of the transformer's latent space. Our experiments demonstrate that GATO-Vid significantly outperforms existing baselines in localization accuracy while introducing minimal computational overhead.
Chinese Translation
扩散变换器文本到视频模型已实现显著的合成质量，但细粒度的空间可控性仍然是一个重大挑战。虽然现有的无训练方法在空间基础生成（即将特定对象放置在指定位置）中产生了良好的整体结果，但它们依赖于基于梯度的优化技术，这会导致巨大的计算开销，这一瓶颈在现代大规模架构中更加明显。为了解决这一限制，我们提出了无梯度解析轨迹优化视频生成（Gradient-free Analytical Trajectory Optimization Video Generation，GATO-Vid），这是一种新颖的无训练和无梯度的方法，用于精确的空间引导。我们不依赖于昂贵的反向传播，而是引入了一种替代的交叉注意力得分，并通过解析方法求解，以获得精确的封闭形式解。为了使用我们的解析解，我们提出了一种针对变换器潜在空间拓扑流形的即时注入机制。我们的实验表明，GATO-Vid在定位精度上显著优于现有基准，同时引入的计算开销极小。
cs.CV / 43 / 2608.13045
P2Fusion: Prompt-based Progressive Infrared-Visible Image Fusion via Dual-Prior Distillation
P2Fusion：基于提示的渐进式红外-可见图像融合通过双重先验蒸馏
Abstract
Infrared-visible image fusion (IVIF) is pivotal for multimodal perception, yet reconciling the inherent information disparity between thermal and textural features remains a fundamental challenge. Existing prior-guided methods often rely on static constraints that induce optimization conflicts or utilize extrinsic semantic priors from large-scale foundation models (e.g., CLIP/DINO), which frequently fail to exploit the intrinsic modality characteristics essential for high-fidelity fusion. To address these issues, we propose P2Fusion, a prior-guided distillation-based framework that reformulates IVIF via dual intrinsic prompts. Instead of imposing hard-coded penalties, we distill image-intrinsic priors, thermal saliency and spatial quality, into learnable dynamic regulators. Specifically, a Teach-to-Fuse mechanism provides dual-granularity progressive guidance, coupled with a Gated Dynamic Expert Recalibration (GDER) module for decoupled feature refinement. This design enables the network to adaptively mediate modal competition through expert specialization. Extensive experiments demonstrate that P2Fusion achieves state-of-the-art performance across five mainstream datasets. Notably, our framework demonstrates consistent performance advantages in fusion quality, achieving state-of-the-art results in 14 out of 20 key evaluation metrics across 5 benchmarks. Furthermore, it effectively contributes to the robustness of downstream perception, such as +3.2% mAP on MSRS, +0.5% mAP on M3FD and +0.9% mAP on DroneVehicle for object detection. Our code will be available at https://github.com/YiShi99/P2Fusion
Chinese Translation
红外-可见图像融合（IVIF）对于多模态感知至关重要，但调和热特征与纹理特征之间固有的信息差异仍然是一个基本挑战。现有的先验引导方法通常依赖于静态约束，这会导致优化冲突，或者利用来自大规模基础模型（如 CLIP/DINO）的外部语义先验，这些方法往往未能充分利用高保真融合所需的内在模态特征。为了解决这些问题，我们提出了 P2Fusion，这是一种基于先验引导的蒸馏框架，通过双重内在提示重新构建 IVIF。我们并不强加硬编码的惩罚，而是将图像内在先验、热显著性和空间质量蒸馏为可学习的动态调节器。具体而言，Teach-to-Fuse 机制提供了双粒度的渐进式指导，并结合 Gated Dynamic Expert Recalibration（GDER）模块进行解耦特征细化。该设计使网络能够通过专家专业化自适应调节模态竞争。大量实验表明，P2Fusion 在五个主流数据集上实现了最先进的性能。值得注意的是，我们的框架在融合质量方面表现出一致的性能优势，在 5 个基准测试的 20 个关键评估指标中，有 14 个指标达到了最先进的结果。此外，它有效提升了下游感知的鲁棒性，例如在物体检测任务中，MSRS 数据集上提升了 +3.2% 的 mAP，M3FD 数据集上提升了 +0.5% 的 mAP，以及 DroneVehicle 数据集上提升了 +0.9% 的 mAP。我们的代码将发布在 https://github.com/YiShi99/P2Fusion
cs.CV / 44 / 2608.13047
Topology-Unified 2D Pose Estimation across Intact, Residual and Prosthetic Limbs
统一拓扑的二维姿态估计：涵盖完整、残余和假肢肢体
Abstract
Driven by the availability of large-scale datasets, Human Pose Estimation (HPE) plays a critical role in numerous downstream tasks. However, mainstream benchmarks exhibit severe representation bias, predominantly featuring able-bodied individuals. While a few pioneering datasets have attempted to address limb differences, their annotation protocols fail to generalize, struggling to represent specialized mechanical structures like running blades or unprosthetized residual limbs. To bridge this gap, we introduce ProPose, a large-scale benchmark featuring a novel annotation protocol that unifies the topological representation of biological limbs, diverse prostheses, and physical absences within a single framework. Because real-world prosthetic images are inherently scarce and exhibit extreme long-tail distributions, we design a Real-to-Synthetic data expansion pipeline to explicitly synthesize and expand the underrepresented cases. However, simply training existing models on this enriched dataset often leads to suboptimal solutions, as they estimate each keypoint independently and might hallucinate non-existent joints on mechanical structures. To resolve this, we propose ProLoss, a structure-aware objective that enforces keypoint dependencies within a single limb to prevent unrealistic limb predictions. Extensive experiments demonstrate that our approach improves the classification accuracy of long-tail prosthetic joints by 2% to 6% without compromising spatial coordinate localization performance. This work sets a foundation for inclusive pose estimation, unlocking new possibilities for understanding the interactions between human bodies and assistive devices.
Chinese Translation
受益于大规模数据集的可用性，人类姿态估计（HPE）在众多下游任务中发挥着关键作用。然而，主流基准测试存在严重的表征偏差，主要集中在健全个体身上。尽管一些开创性的数据集试图解决肢体差异问题，但它们的注释协议未能通用，难以有效表示像跑步刀片或未假肢的残余肢体等特殊机械结构。为了解决这一问题，我们提出了ProPose，一个大规模基准，采用了一种新的注释协议，将生物肢体、多样化假肢和物理缺失的拓扑表征统一在一个框架内。由于现实世界中的假肢图像本质上稀缺，并且呈现极端长尾分布，我们设计了一个真实到合成的数据扩展管道，以明确合成和扩展被低估的案例。然而，简单地在这个丰富的数据集上训练现有模型通常会导致次优解，因为它们独立估计每个关键点，并可能在机械结构上幻觉出不存在的关节。为了解决这个问题，我们提出了ProLoss，这是一种结构感知目标，强制在单个肢体内施加关键点依赖性，以防止不现实的肢体预测。大量实验表明，我们的方法在不损害空间坐标定位性能的情况下，将长尾假肢关节的分类准确率提高了2%到6%。这项工作为包容性姿态估计奠定了基础，开启了理解人类身体与辅助设备之间相互作用的新可能性。
cs.CV / 45 / 2608.13064
Learning Unified Video and Image Representation for Video Face Forgery Detection
学习统一的视频和图像表示以进行视频人脸伪造检测
Abstract
Face forgery detection is crucial for preserving the security and integrity of facial data given the rapid developments in face manipulation techniques and deep generative models. Existing methods for video face forgery detection typically assume that all frames in a forged video are manipulated, while detecting partially forged videos that contain only a subset of altered frames remains challenging. To address this issue, we propose a novel framework, UVIF, that utilizes additional annotated images to provide fine-grained supervision for detecting partial forgeries in videos. UVIF employs a unified encoder and a multi-task learning paradigm to jointly model facial videos and images for boosted video face forgery detection. A 2D backbone with temporal fusion modules is employed as the unified encoder. A pseudo labeling process is designed for video frames to bridge their representations with those of static images. A video-oriented feature alignment strategy is further introduced to reduce the distribution gap between videos and images. Extensive experiments on benchmark datasets demonstrate the effectiveness of our framework, which outperforms state-of-theart methods in detecting partially forged videos while introducing no additional computational overhead. Our code is available at https://github.com/haotianll/UVIF.
Chinese Translation
人脸伪造检测对于保护人脸数据的安全性和完整性至关重要，尤其是在面部操控技术和深度生成模型快速发展的背景下。现有的视频人脸伪造检测方法通常假设伪造视频中的所有帧均被操控，而仅包含部分修改帧的部分伪造视频的检测仍然具有挑战性。为了解决这个问题，我们提出了一种新颖的框架UVIF，该框架利用额外的标注图像为视频中的部分伪造检测提供细粒度的监督。UVIF采用统一编码器和多任务学习范式，联合建模人脸视频和图像，以增强视频人脸伪造检测的效果。统一编码器使用带有时间融合模块的2D主干网络。设计了一种伪标签过程，用于视频帧，以将其表示与静态图像的表示相连接。此外，引入了一种面向视频的特征对齐策略，以减少视频和图像之间的分布差距。在基准数据集上的大量实验表明，我们的框架在检测部分伪造视频方面优于最先进的方法，同时没有引入额外的计算开销。我们的代码可在 https://github.com/haotianll/UVIF 获取。
cs.CV / 46 / 2608.13092
Paths: Prompt-aware Spatio-temporal Transformer with Hierarchical Multi-modal Fusion for RGB-Event Video Person Re-Identification
Paths：具有层次多模态融合的提示感知时空变换器用于RGB-事件视频中的人物重识别
Abstract
RGB-Event Video Person Re-Identification (RE-VReID) aims to retrieve specific person across non-overlapping cameras with complementary RGB videos and event streams. However, existing methods often decouple spatial and temporal modeling, which limits their interaction. In addition, global-level RGB-Event fusion fails to fully exploit fine-grained discriminative cues. To address these issues, we propose Paths, a unified framework with spatio-temporal modeling and hierarchical multi-modal fusion for RE-VReID. Specifically, we first design a Memory-Augmented Backbone (MAB) to maintain modality-specific identity prototypes for stable intra-modal representation learning. Then, we propose a Prompt-aware Spatio-temporal Transformer (PST) to jointly model spatial and temporal cues within a unified Transformer. Finally, we introduce a Hierarchical Multi-modal Fusion (HMF) to integrate RGB and event features at global and local levels. With these modules, our framework can learn robust and discriminative representations for RE-VReID. Extensive experiments on three public RE-VReID benchmarks including EvReID, MARS and iLIDS-VID, demonstrate the effectiveness of our proposed method. The code is available at https://github.com/Reflection0427/Paths.
Chinese Translation
RGB-事件视频人物重识别（RE-VReID）旨在通过互补的RGB视频和事件流在非重叠摄像头之间检索特定人物。然而，现有方法往往将空间和时间建模解耦，限制了它们之间的互动。此外，全球级的RGB-事件融合未能充分利用细粒度的区分线索。为了解决这些问题，我们提出了Paths，一个统一的框架，结合了时空建模和层次多模态融合以实现RE-VReID。具体而言，我们首先设计了一个增强记忆的主干网络（Memory-Augmented Backbone, MAB），以维持特定模态的身份原型，从而实现稳定的模态内表示学习。然后，我们提出了一种提示感知时空变换器（Prompt-aware Spatio-temporal Transformer, PST），在统一的变换器中共同建模空间和时间线索。最后，我们引入了层次多模态融合（Hierarchical Multi-modal Fusion, HMF），以在全局和局部层面整合RGB和事件特征。通过这些模块，我们的框架能够学习到稳健且具有区分性的表示用于RE-VReID。在包括EvReID、MARS和iLIDS-VID在内的三个公共RE-VReID基准上进行的广泛实验，证明了我们提出的方法的有效性。代码可在https://github.com/Reflection0427/Paths获取。
cs.CV / 47 / 2608.13102
RbFT-Net: Rectify-Before-Fuse Temporal Radar Anchors for 4D Radar-Camera Depth Completion
RbFT-Net：用于4D雷达-相机深度补全的先校正后融合时间雷达锚点
Abstract
Dense metric depth prediction from cameras and millimeter-wave radar offers a cost-effective sensing solution for autonomous systems. However, radar measurements are inherently sparse and susceptible to clutter, multipath reflections, and projection errors. While aggregating multiple radar frames provides denser metric cues, it also introduces temporal misalignment and dynamic-object interference. Directly propagating such unreliable measurements can therefore corrupt large regions of the predicted depth map. To address this issue, we propose RbFT-Net, an end-to-end rectify-before-fuse framework for multi-frame 4D radar-camera depth completion. Rather than assuming accumulated radar returns to be accurate, RbFT-Net treats them as noisy temporal anchor candidates. An image-conditioned rectification module jointly corrects their image-plane locations and metric depths while estimating pointwise reliability. The rectified anchors are then selectively propagated before high-level multi-modal fusion, suppressing the influence of unreliable measurements. Experiments on ZJU-4DRadarCam and a newly collected 4D radar-camera-LiDAR dataset show that RbFT-Net consistently outperforms the evaluated independent radar-camera methods and remains competitive with plug-in pipelines using auxiliary monocular depth models. Cross-platform evaluation and component analyses further support the effectiveness of the proposed rectification and reliability-aware propagation strategy.
Chinese Translation
来自相机和毫米波雷达的密集度量深度预测为自主系统提供了一种具有成本效益的传感解决方案。然而，雷达测量本质上是稀疏的，并且容易受到杂波、多路径反射和投影误差的影响。虽然聚合多个雷达帧可以提供更密集的度量线索，但这也引入了时间对齐问题和动态物体干扰。因此，直接传播这些不可靠的测量可能会破坏预测深度图的大区域。为了解决这个问题，我们提出了RbFT-Net，一种端到端的先校正后融合框架，用于多帧4D雷达-相机深度补全。RbFT-Net并不假设累积的雷达返回是准确的，而是将其视为噪声时间锚点候选。一个基于图像的校正模块共同校正它们的图像平面位置和度量深度，同时估计逐点可靠性。然后，在高层次的多模态融合之前，选择性地传播校正后的锚点，从而抑制不可靠测量的影响。在ZJU-4DRadarCam和新收集的4D雷达-相机-LiDAR数据集上的实验表明，RbFT-Net在评估的独立雷达-相机方法中始终表现优于其他方法，并且在使用辅助单目深度模型的插件管道中保持竞争力。跨平台评估和组件分析进一步支持了所提出的校正和可靠性感知传播策略的有效性。
cs.CV / 48 / 2608.13104
Online Learning of Correspondences between Images
图像间对应关系的在线学习
Abstract
We propose a novel method for iterative learning of point correspondences between image sequences. Points moving on surfaces in 3D space are projected into two images. Given a point in either view, the considered problem is to determine the corresponding location in the other view. The geometry and distortions of the projections are unknown as is the shape of the surface. Given several pairs of point-sets but no access to the 3D scene, correspondence mappings can be found by excessive global optimization or by the fundamental matrix if a perspective projective model is assumed. However, an iterative solution on sequences of point-set pairs with general imaging geometry is preferable. We derive such a method that optimizes the mapping based on Neyman's chi-square divergence between the densities representing the uncertainties of the estimated and the actual locations. The densities are represented as channel vectors computed with a basis function approach. The mapping between these vectors is updated with each new pair of images such that fast convergence and high accuracy are achieved. The resulting algorithm runs in real-time and is superior to state-of-the-art methods in terms of convergence and accuracy in a number of experiments.
Chinese Translation
我们提出了一种新颖的方法，用于迭代学习图像序列之间的点对应关系。三维空间中表面上移动的点被投影到两幅图像中。给定任一视图中的一个点，所考虑的问题是确定在另一视图中的对应位置。投影的几何形状和失真是未知的，表面的形状也是如此。给定多个点集对，但无法访问三维场景，可以通过过度的全局优化或假设透视投影模型的基础矩阵来找到对应映射。然而，对于具有一般成像几何的点集对序列，迭代解决方案是更可取的。我们推导出这样一种方法，该方法基于 Neyman 的卡方散度优化映射，散度反映了估计位置与实际位置的不确定性。密度被表示为使用基函数方法计算的通道向量。每当有新的一对图像时，这些向量之间的映射就会更新，从而实现快速收敛和高精度。最终的算法能够实时运行，并在多个实验中在收敛性和准确性方面优于现有的最先进方法。
cs.CV / 49 / 2608.13112
Towards Physics-Faithful Generation of Scientific Diagrams
朝向物理忠实的科学图表生成
Abstract
Text-to-image generation has reached photorealistic quality, yet state-of-the-art systems remain unreliable at producing scientific diagrams, whose value depends not on appearance but on physical faithfulness: correct force directions, valid coordinate systems, consistent thermodynamic states, and equations matching the depicted scenario. Trained on web imagery with physically shallow captions, generic models produce diagrams that look plausible but are physically wrong, harmful in education and scientific communication. We present Princigram, a physics-faithful scientific-diagram generator, and its data pipeline. Our central advance is Structured Physical Chain-of-Thought (SP-CoT): a per-subdiscipline schema that decomposes a physics diagram into an explicit multi-step reasoning chain across six subdisciplines, from scene identification through force or process analysis to governing laws and synthesis. Unlike free-form chain-of-thought, SP-CoT follows a fixed schema with strict fidelity rules that separate visually grounded facts from physically inferred reasoning and type all mathematics symbolically; it serves both as dense training supervision and, at inference, as a structured "thinking" prompt. With it we curate and structurally annotate 4.3 million physics images, of which 115,037 carry expert-level annotation, and adapt a unified multimodal backbone. We further introduce VeriphyT2IBench, whose questions are derived from each held-out diagram's own structured annotation: each diagram becomes an item-specific bank of binary questions about its objects, forces, and states, so a judge model's score decomposes into named physical facts rather than one holistic number. On the physics subset of GenExam and on VeriphyT2IBench, Princigram shows that explicit physics-structured supervision improves the physical faithfulness of generated scientific diagrams.
Chinese Translation
文本到图像生成已达到照片级真实感，但最先进的系统在生成科学图表时仍然不可靠，这些图表的价值不仅在于外观，更在于物理忠实性：正确的力方向、有效的坐标系统、一致的热力学状态以及与所描绘场景匹配的方程。由于训练于具有物理浅显标题的网络图像，通用模型生成的图表看似合理但在物理上是错误的，这在教育和科学传播中是有害的。我们提出了Princigram，一个物理忠实的科学图表生成器及其数据管道。我们的核心进展是结构化物理推理链（Structured Physical Chain-of-Thought, SP-CoT）：一个按学科划分的模式，将物理图表分解为跨六个子学科的明确多步骤推理链，从场景识别到力或过程分析，再到支配定律和综合。与自由形式的推理链不同，SP-CoT遵循固定模式，具有严格的忠实性规则，将视觉基础事实与物理推导推理分开，并将所有数学符号化；它既作为密集训练的监督，又在推理时作为结构化的“思考”提示。通过这个方法，我们策划并结构化标注了430万张物理图像，其中115,037张具有专家级标注，并适配了统一的多模态骨干网络。我们进一步介绍了VeriphyT2IBench，其问题源自每个保留图表的结构化标注：每个图表成为关于其对象、力和状态的特定项的二元问题库，因此评判模型的得分分解为命名的物理事实，而不是一个整体数字。在GenExam的物理子集和VeriphyT2IBench上，Princigram显示出明确的物理结构化监督提高了生成科学图表的物理忠实性。
cs.CV / 50 / 2608.13113
EgoMonth: A Month-Level Egocentric Video Benchmark for Long-Term Spatiotemporal Memory
EgoMonth：一个用于长期时空记忆的月度自我中心视频基准
Abstract
Recent advances in Multimodal Large Language Models (MLLMs) have led to substantial progress in video understanding, accompanied by a growing number of long video benchmarks. However, existing benchmarks rely predominantly on web-sourced videos that lack inter-clip spatiotemporal continuity, making it difficult to assess whether models can maintain consistent memory across days or weeks of real-world experience. We introduce EgoMonth, the first month-level egocentric video understanding benchmark. EgoMonth comprises over 300 hours of first-person daily-life recordings from 20 participants spanning 20 to 120 days, paired with 1,443 human-crafted multiple-choice question-answer pairs. We design a cognitively grounded 14-task evaluation framework organized into three hierarchical cognitive levels: Schema Consolidation, Episodic Indexing, and Cascading Reasoning. Evaluation of state-of-the-art open-source and closed-source MLLMs reveals that even the best-performing model, Gemini 2.5 Pro, achieves only 71.8% macro-average accuracy, remaining 22.4 percentage points below the corrected human baseline of 94.2%. Several models perform near or below the 25% chance level on tasks such as Route Reasoning, Cross-view Spatial Reasoning, and Direction Judgement, while even the strongest closed-source model remains substantially below human performance. These results indicate that current MLLMs function as lossy summarizers rather than faithful memorizers, highlighting the need for architectures with genuine long-term spatiotemporal memory.
Chinese Translation
近年来，多模态大型语言模型（MLLMs）的进展显著推动了视频理解的发展，同时也出现了越来越多的长视频基准。然而，现有基准主要依赖于来源于网络的视频，这些视频缺乏片段间的时空连续性，使得评估模型是否能够在现实世界的经历中维持一致的记忆变得困难。我们提出了EgoMonth，这是第一个月度自我中心视频理解基准。EgoMonth包含来自20名参与者的超过300小时的第一人称日常生活录音，时间跨度为20到120天，并配有1,443对人工制作的多项选择问答对。我们设计了一个基于认知的14项任务评估框架，分为三个层次的认知水平：模式巩固、情节索引和级联推理。对最先进的开源和闭源MLLMs的评估显示，即使是表现最好的模型Gemini 2.5 Pro，其宏平均准确率也仅为71.8%，比经过修正的人类基线94.2%低22.4个百分点。在路线推理、跨视角空间推理和方向判断等任务中，多个模型的表现接近或低于25%的随机机会水平，而即使是最强的闭源模型也远低于人类表现。这些结果表明，目前的MLLMs更像是有损的摘要生成器，而非忠实的记忆者，突显了需要具有真正长期时空记忆的架构。
cs.CV / 51 / 2608.13114
Fast Iterative Five point Relative Pose Estimation
快速迭代五点相对位姿估计
Abstract
Robust estimation of the relative pose between two cameras is a fundamental part of Structure and Motion methods. For calibrated cameras, the five point method together with a robust estimator such as RANSAC gives the best result in most cases. The current state-of-the-art method for solving the relative pose problem from five points is due to Nister [9], because it is faster than other methods and in the RANSAC scheme one can improve precision by increasing the number of iterations. In this paper, we propose a new iterative method, which is based on Powell's Dog Leg algorithm. The new method has the same precision and is approximately twice as fast as Nister's algorithm. The proposed method is easily extended to more than five points while retaining a efficient error metrics. This makes it also very suitable as an refinement step. The proposed algorithm is systematically evaluated on three types of datasets with known ground truth.
Chinese Translation
两个摄像机之间相对位姿的稳健估计是结构与运动方法的一个基本部分。对于已标定的摄像机，五点法结合鲁棒估计器（如 RANSAC）在大多数情况下能够提供最佳结果。目前解决五点相对位姿问题的最先进方法是 Nister 提出的，因为它比其他方法更快，并且在 RANSAC 方案中，通过增加迭代次数可以提高精度。本文提出了一种新的迭代方法，基于 Powell 的 Dog Leg 算法。该新方法具有与 Nister 算法相同的精度，并且速度大约是其两倍。所提方法易于扩展到超过五个点，同时保持高效的误差度量。这使得它也非常适合作为一个精细化步骤。所提算法在三个具有已知真实值的数据集上进行了系统评估。
cs.CV / 52 / 2608.13119
QuISE: Defense against Typographic Attacks on VLMs via Query-Irrelevant Semantic Editing
QuISE：通过查询无关的语义编辑防御视觉语言模型的排版攻击
Abstract
Typographic attacks pose a critical threat to vision-language models (VLMs) by injecting misleading text into images and causing models to rely on adversarial textual cues rather than visual evidence. Existing defenses often require model-specific modifications, additional training, or access to internal model components, limiting their applicability to modern closed-source VLMs. In this paper, we propose QuISE, a model-agnostic, training-free black-box defense based on query-irrelevant semantic editing. QuISE first identifies text regions likely to affect the current query through influence-aware text localization. QuISE then replaces these regions with two semantically distinct replacement texts that are irrelevant to both the query and the image. The final answer is determined by answer consistency across the edited images. Extensive experiments on three typographic-attack benchmarks, four attack settings, and four VLMs show that QuISE consistently improves defended accuracy. QuISE achieves a recovery rate of 67.9-75.0% with a harm rate of 0.5-1.1%.
Chinese Translation
排版攻击对视觉语言模型（VLMs）构成了严重威胁，通过向图像中注入误导性文本，使模型依赖对抗性文本线索而非视觉证据。现有的防御方法通常需要特定于模型的修改、额外的训练或对内部模型组件的访问，这限制了它们在现代闭源VLMs中的适用性。本文提出了QuISE，这是一种模型无关、无需训练的黑箱防御方法，基于查询无关的语义编辑。QuISE首先通过影响感知文本定位识别出可能影响当前查询的文本区域。然后，QuISE用两个与查询和图像都无关的语义上不同的替换文本替换这些区域。最终答案通过编辑后的图像之间的答案一致性来确定。在三个排版攻击基准、四种攻击设置和四个VLMs上的大量实验表明，QuISE始终提高了防御准确性。QuISE的恢复率为67.9%-75.0%，损害率为0.5%-1.1%。
cs.CV / 53 / 2608.13135
Predicting Signed Distance Functions for Visual Instance Segmentation
预测可视实例分割的有符号距离函数
Abstract
Visual instance segmentation is a challenging problem and becomes even more difficult if objects of interest varies unconstrained in shape. Some objects are well described by a rectangle, however, this is hardly always the case. Consider for instance long, slender objects such as ropes. Anchor-based approaches classify predefined bounding boxes as either negative or positive and thus provide a limited set of shapes that can be handled. Defining anchor-boxes that fit well to all possible shapes leads to an infeasible number of prior boxes. We explore a different approach and propose to train a neural network to compute distance maps along different directions. The network is trained at each pixel to predict the distance to the closest object contour in a given direction. By pooling the distance maps we obtain an approximation to the signed distance function (SDF). The SDF may then be thresholded in order to obtain a foreground-background segmentation. We compare this segmentation to foreground segmentations obtained from the state-of-the-art instance segmentation method YOLACT. On the COCO dataset, our segmentation yields a higher performance in terms of foreground intersection over union (IoU). However, while the distance maps contain information on the individual instances, it is not straightforward to map them to the full instance segmentation. We still believe that this idea is a promising research direction for instance segmentation, as it better captures the different shapes found in the real world.
Chinese Translation
可视实例分割是一个具有挑战性的问题，当感兴趣的物体形状不受限制时，这一问题变得更加复杂。有些物体可以用矩形很好地描述，但这并非总是如此。例如，考虑长而细的物体，如绳索。基于锚点的方法将预定义的边界框分类为负类或正类，从而提供了有限的形状集合。定义适合所有可能形状的锚框会导致不可行的先验框数量。我们探索了一种不同的方法，提出训练神经网络沿不同方向计算距离图。该网络在每个像素处被训练以预测在给定方向上到最近物体轮廓的距离。通过汇聚距离图，我们获得了有符号距离函数（SDF）的近似值。然后，可以对SDF进行阈值处理，以获得前景-背景分割。我们将这种分割与最先进的实例分割方法YOLACT获得的前景分割进行比较。在COCO数据集上，我们的分割在前景交并比（IoU）方面表现出更高的性能。然而，尽管距离图包含有关单个实例的信息，但将其映射到完整的实例分割并不简单。我们仍然相信这一思路是实例分割的一个有前景的研究方向，因为它更好地捕捉了现实世界中存在的不同形状。
cs.CV / 54 / 2608.13141
MergeOver: Post-Training Token Merging for Recursive Vision Transformers
MergeOver：递归视觉变换器的后训练令牌合并
Abstract
Vision Transformers (ViTs) demonstrate exceptional performance in computer vision but suffer from large parameter counts and quadratic computational complexity, severely limiting their deployment on resource-constrained edge hardware. While recursive weight-sharing reduces parameter counts and token merging mitigates computational and memory bottlenecks, integrating these two paradigms without costly retraining is non-trivial, leaving this intersection largely unexplored. We propose MergeOver, a post-training approach that integrates Token Merging (ToMe) into the recursively weight-shared Sliced Recursive Transformer (SReT). Through an Unmerge tracking stack, constraint-safe merge-rate adjustment, and synchronised token-mass tracking across spatial permutations, MergeOver resolves the spatial and merging constraints of this integration. We further employ a stage-wise single-shot schedule that performs token reduction at the first block of each stage and maintains a fixed sequence length throughout its subsequent recursive iterations. Benchmarked on ImageNet-1K, our selected configuration reduces top-1 accuracy by 1.47 percentage points. On the GPU, it reduces peak activation memory by 37.3% and 38.4% at batch sizes 1 and 16, while throughput decreases by 21.7% at batch size 1 but increases by 21.7% at batch size 16. On a Raspberry Pi 5 (ARM CPU), it reduces latency by 2.4% and 17.6% at batch sizes 1 and 16. These results show that MergeOver can recover a meaningful part of the throughput and memory cost that recursive weight-sharing introduces, without retraining, and provides a baseline for combining token merging with hierarchical recursive transformers.
Chinese Translation
视觉变换器（ViTs）在计算机视觉领域表现出色，但由于参数数量庞大和计算复杂度呈平方级别，严重限制了它们在资源受限的边缘硬件上的部署。尽管递归权重共享可以减少参数数量，而令牌合并则可以缓解计算和内存瓶颈，但在不进行昂贵的再训练的情况下，将这两种范式结合起来并非易事，这一交集领域仍然大多未被探索。我们提出了MergeOver，这是一种后训练方法，将令牌合并（Token Merging, ToMe）集成到递归权重共享的切片递归变换器（Sliced Recursive Transformer, SReT）中。通过一个反合并跟踪栈、安全的合并速率调整和跨空间排列的同步令牌质量跟踪，MergeOver解决了这一整合的空间和合并约束。我们进一步采用了一种阶段性单次调度，在每个阶段的第一个模块中执行令牌减少，并在其后续的递归迭代中保持固定的序列长度。在ImageNet-1K上的基准测试中，我们选择的配置使得top-1准确率降低了1.47个百分点。在GPU上，它在批量大小为1和16时分别减少了37.3%和38.4%的峰值激活内存，而在批量大小为1时吞吐量减少了21.7%，但在批量大小为16时增加了21.7%。在树莓派5（ARM CPU）上，它在批量大小为1和16时分别减少了2.4%和17.6%的延迟。这些结果表明，MergeOver能够在不进行再训练的情况下，恢复递归权重共享所引入的吞吐量和内存成本的有意义部分，并为将令牌合并与分层递归变换器结合提供了基准。
cs.CV / 55 / 2608.13147
Geometry-Grounded Unified 3D Perception for Autonomous Driving
基于几何的统一三维感知在自动驾驶中的应用
Abstract
Camera-based autonomous driving perception requires a shared representation that preserves metric 3D structure across synchronized multi-camera streams. However, existing image-based frameworks often rely on backbones pretrained for semantic recognition, and introduce 3D geometry through downstream task-specific modules. As a result, their shared representations may fail to preserve explicit metric geometry and consistent 3D scene structure. In this paper, we present a Geometry-grounded Unified 3D Perception (GeoUP) framework that adapts the reconstruction-oriented latent of VGGT to calibrated, streaming multi-camera driving scenes. GeoUP factorizes cross-image interaction into self, temporal, and view attention to capture structurally distinct temporal and cross-view correspondences. It further injects calibration-aware raymap encodings to provide metric scale and camera geometry. The resulting geometry-grounded latent is decoded for metric depth estimation, 3D object detection, and semantic occupancy prediction, corresponding to surface-, instance-, and volume-level readouts of the same 3D scene. Through joint multi-task and multi-dataset training, GeoUP effectively leverages heterogeneous annotations and generalizes across diverse sensor configurations and perception ranges. Extensive experiments on nuScenes, Argoverse 2, Waymo, KITTI, and DDAD demonstrate that GeoUP achieves SOTA performance across detection, occupancy, and depth estimation. These results validate the effectiveness of geometry-grounded representations for unified 3D driving perception.
Chinese Translation
基于摄像头的自动驾驶感知需要一种共享表示，该表示能够在同步的多摄像头流中保持度量三维结构。然而，现有的基于图像的框架通常依赖于为语义识别预训练的主干网络，并通过下游任务特定模块引入三维几何。因此，它们的共享表示可能无法保持明确的度量几何和一致的三维场景结构。在本文中，我们提出了一种基于几何的统一三维感知框架（Geometry-grounded Unified 3D Perception, GeoUP），该框架将面向重建的VGGT潜在表示适应于标定的、流式的多摄像头驾驶场景。GeoUP将跨图像交互分解为自我、时间和视角注意力，以捕捉结构上不同的时间和跨视角对应关系。它进一步注入了考虑标定的光线图编码，以提供度量尺度和摄像机几何。最终得到的基于几何的潜在表示被解码用于度量深度估计、三维物体检测和语义占用预测，分别对应于同一三维场景的表面级、实例级和体积级读数。通过联合多任务和多数据集训练，GeoUP有效利用异构注释，并在不同传感器配置和感知范围内实现泛化。在nuScenes、Argoverse 2、Waymo、KITTI和DDAD上的大量实验表明，GeoUP在检测、占用和深度估计方面达到了最先进的性能。这些结果验证了基于几何的表示在统一三维驾驶感知中的有效性。
cs.CV / 56 / 2608.13148
Less Annotation, More Interpretation: Prior-Guided Concept Bottleneck Models for Interpretable Cancer Imaging Diagnosis
更少的标注，更丰富的解释：基于先验指导的概念瓶颈模型用于可解释的癌症影像诊断
Abstract
Concept bottleneck models (CBMs) can improve the transparency of cancer image diagnostic prediction by expressing predictions through radiological concepts. However, their dependence on instance-level concept annotations limits practical applicability. We propose a prior-guided hybrid CBM that integrates limited concept annotations, class-conditional concept distribution matching on unannotated patients, and prior initialization of the concept-to-diagnosis head. We evaluate the method on CBIS-DDSM mammographic masses and calcifications and LIDC-IDRI pulmonary nodules across 0-100% concept annotation. In the clinically relevant 0-20% annotation regime, the hybrid CBM consistently improves mean concept AUC over a matched standard CBM, while maintaining diagnostic performance close to black-box models. At 10% annotation specifically, concept AUC increases from 0.619 to 0.741 for masses, from 0.650 to 0.787 for calcifications, and from 0.597 to 0.642 for pulmonary nodules. Ablation experiments identify prior initialization as the main component contributing to improved concept detection, likely by stabilizing the concept-to-diagnosis head. Zero-shot VLMs remain insufficient for reliable fine-grained tumor-level concept prediction. These findings suggest that structured priors can substantially reduce the annotation burden of interpretable cancer imaging models.
Chinese Translation
概念瓶颈模型（CBMs）通过利用放射学概念表达预测，从而提高癌症影像诊断预测的透明度。然而，它们对实例级概念标注的依赖限制了实际应用。我们提出了一种基于先验指导的混合CBM，该模型整合了有限的概念标注、对未标注患者的类别条件概念分布匹配，以及概念到诊断头的先验初始化。我们在CBIS-DDSM乳腺X线检查的肿块和钙化以及LIDC-IDRI肺结节数据集上评估该方法，覆盖0-100%的概念标注。在临床相关的0-20%标注范围内，混合CBM在保持接近黑箱模型的诊断性能的同时，始终提高了平均概念AUC，相较于匹配的标准CBM。在10%标注的情况下，肿块的概念AUC从0.619提高到0.741，钙化从0.650提高到0.787，肺结节从0.597提高到0.642。消融实验表明，先验初始化是提高概念检测的主要因素，可能通过稳定概念到诊断头的过程来实现。零-shot VLMs在可靠的细粒度肿瘤级概念预测方面仍然不足。这些发现表明，结构化先验可以显著减少可解释癌症影像模型的标注负担。
cs.CV / 57 / 2608.13159
Splat-based Metal Artifact Reduction in Cone-Beam CT via Polychromatic Modeling
基于喷溅的金属伪影减少方法在锥束CT中的多色建模
Abstract
Cone-beam computed tomography (CBCT) enables volumetric reconstruction from X-ray projections, but suffers from severe artifacts--especially beam hardening--when imaging materials with high attenuation such as metals. These artifacts arise from the polychromatic nature of X-rays and are not properly addressed by conventional monochromatic reconstruction algorithms. While recent neural representation-based methods offer improved reconstruction quality, they are computationally expensive and often impractical for deployment. We propose a novel physics-inspired, self-calibrating metal artifact reduction method that efficiently reconstructs 3D CBCT volumes while correcting beam hardening artifacts. Our method integrates a polychromatic X-ray projection model, material-dependent attenuation profiles, and system response modeling into a Gaussian Splatting framework. Unlike prior work, we eliminate the need for manual metal masks or strong prior assumptions, and we optimize both reconstruction parameters and X-ray spectral characteristics jointly during training. We further introduce a high-fidelity synthetic CBCT dataset generation pipeline validated on Monte-Carlo x-ray simulation toolbox and release new datasets with severe metal-induced artifacts to support the community. This is the first splat-based method for reducing beam hardening in CBCT. Extensive experiments on both synthetic and real-world datasets demonstrate that our method outperforms state-of-the-art approaches in artifact suppression and reconstruction accuracy.
Chinese Translation
锥束计算机断层扫描（CBCT）能够从X射线投影中进行体积重建，但在成像高衰减材料（如金属）时会遭遇严重的伪影，尤其是束硬化伪影。这些伪影源于X射线的多色特性，而传统的单色重建算法无法有效解决此问题。尽管最近基于神经表示的方法提供了更好的重建质量，但它们计算成本高且在实际应用中往往不切实际。我们提出了一种新颖的物理启发、自校准的金属伪影减少方法，该方法能够高效重建3D CBCT体积，同时校正束硬化伪影。我们的方法将多色X射线投影模型、材料依赖的衰减特征和系统响应建模整合到高斯喷溅框架中。与之前的研究不同，我们消除了对手动金属掩模或强先验假设的需求，并在训练过程中共同优化重建参数和X射线光谱特性。我们进一步引入了一种高保真合成CBCT数据集生成管道，并在蒙特卡罗X射线仿真工具箱上进行了验证，同时发布了带有严重金属诱导伪影的新数据集，以支持社区。这是第一个基于喷溅的方法，用于减少CBCT中的束硬化伪影。在合成和真实世界数据集上的广泛实验表明，我们的方法在伪影抑制和重建精度方面优于最先进的方法。
cs.CV / 58 / 2608.13167
TRAPSBench: Vision-Language Models Encode but Fail to Express Epistemic Restraint
TRAPSBench：视觉-语言模型编码但未能表达认知约束
Abstract
When visual evidence is occluded or chaotic, models should abstain. In this paper, we show that Vision-Language Models (VLMs) can internally distinguish when abstention is required, but fail to express it anyway. We introduce TRAPSBench, a procedurally generated video benchmark of 1,404 matched physics pairs in which a single targeted change renders the outcome undeterminable from the visual evidence. Furthermore, we introduce Penalized Epistemic Calibration Score (PECS), a new robust metric that requires models to both answer correctly when the outcome is knowable, and abstain when the outcome is not. Across 16 VLMs spanning five families, spontaneous restraint is poor: the best PECS is 0.292. The bottleneck is expression, not perception: linear probes decode answerability from hidden states at up to 0.91 AUROC across physics domains; steering a single-layer void direction causally induces or suppresses abstention. Our results replicate across three open-weight families (Qwen, Gemma, LLaVA). The failure is also more pronounced in visual than textual uncertainty: models detect textual impossibility about 4x more readily than missing visual evidence. Closing this representation--output gap likely requires output-stage interventions.
Chinese Translation
当视觉证据被遮挡或混乱时，模型应当选择不作答。本文展示了视觉-语言模型（VLMs）能够内部区分何时需要不作答，但仍未能有效表达这一点。我们引入了TRAPSBench，一个程序生成的视频基准，包含1,404对匹配的物理实验，其中单一的目标变化使得结果无法从视觉证据中确定。此外，我们引入了惩罚性认知校准分数（PECS），这是一种新的稳健指标，要求模型在结果可知时正确回答，而在结果不可知时选择不作答。在涵盖五个家族的16个VLMs中，自发性约束表现不佳：最佳PECS为0.292。瓶颈在于表达，而非感知：线性探针能够从隐藏状态中解码出可回答性，在物理领域的AUROC高达0.91；操控单层空洞方向因果地引发或抑制不作答。我们的结果在三个开放权重家族（Qwen、Gemma、LLaVA）中得到了重复验证。失败在视觉不确定性中比文本不确定性更为明显：模型检测文本不可能性约为视觉证据缺失的4倍。弥补这一表示与输出之间的差距可能需要在输出阶段进行干预。
cs.CV / 59 / 2608.13183
A Controlled Study of Self-Supervised Image and Video Pretraining under Limited Resources
有限资源下自监督图像与视频预训练的对照研究
Abstract
Visual foundation models are a cornerstone of image and video understanding but typically require large amounts of data and computation. The current scale required for pretraining visual foundation models may be unsustainable or unnecessary, and significant benefits arise when effective models can be obtained with fewer resources. To better understand how self-supervised learning (SSL) objectives behave under resource constraints, we conduct a controlled study of image and video SSL objectives under matched data, architecture, and compute budgets. We compare contrastive, reconstruction, feature-prediction, and diffusion objectives and evaluate both standalone and jointly trained image-video SSL formulations across a diverse set of image and video understanding tasks. Our results show that DINOv2-style pretraining consistently provides the strongest overall performance under limited resources. Furthermore, combining DINOv2 with video SSL objectives such as VideoMAE substantially improves image classification and segmentation performance, but degrades video tracking and camera-pose estimation performance, revealing an important tradeoff between semantic and geometric representation learning. These findings suggest that combining image and video SSL objectives can be beneficial in resource-limited settings, while highlighting the need for improved methods that better balance semantic, temporal, and geometric supervision.
Chinese Translation
视觉基础模型是图像和视频理解的基石，但通常需要大量的数据和计算资源。目前预训练视觉基础模型所需的规模可能是不可持续或不必要的，当能够用更少的资源获得有效模型时，会带来显著的好处。为了更好地理解自监督学习（Self-Supervised Learning, SSL）目标在资源限制下的表现，我们在匹配的数据、架构和计算预算下，对图像和视频的SSL目标进行了对照研究。我们比较了对比、重建、特征预测和扩散目标，并在一系列多样的图像和视频理解任务中评估了独立训练和联合训练的图像-视频SSL模型。我们的结果表明，在有限资源下，DINOv2风格的预训练始终提供了最强的整体性能。此外，将DINOv2与视频SSL目标（如VideoMAE）结合显著提高了图像分类和分割性能，但降低了视频跟踪和相机姿态估计性能，揭示了语义和几何表示学习之间的重要权衡。这些发现表明，在资源有限的环境中，结合图像和视频的SSL目标可能是有益的，同时强调了需要改进的方法，以更好地平衡语义、时间和几何监督。
cs.CV / 60 / 2608.13186
SketchSense: Learning to Interpret Imperfect Sketch Guidance for Image Inpainting
SketchSense：学习解读不完美草图指导的图像修复
Abstract
Sketch-guided image inpainting provides intuitive structural control, yet real sketches often mix reliable global intent with locally crowded, displaced, incomplete, or deliberately unconventional strokes. Existing approaches typically either retain the input sketch as a fixed condition throughout denoising or refine it into a clean structure before RGB synthesis. The former assumes uniformly reliable strokes and can propagate local errors throughout generation; the latter must resolve ambiguous structure before emerging appearance and semantic context become available. We propose SketchSense, a framework that interprets imperfect sketch guidance by synchronously denoising interacting RGB and structure streams. Bidirectional Attention Fusion couples appearance generation with structural recovery, producing a refined structure that exposes the model's evolving sketch interpretation. A phrase-level objective aligns the semantic grounding of the two streams. Sketch-Aware Spatial Regulation further adapts sketch use to local generation states by modulating attention and the fusion process, while an optional signed prior injects preserve-versus-correct intent into feature representations and attention behavior. Experiments on natural and structurally complex imagery show substantial gains over existing methods in both restoration quality and structural fidelity.
Chinese Translation
草图引导的图像修复提供了直观的结构控制，但真实的草图往往将可靠的全局意图与局部拥挤、位移、不完整或故意不传统的笔画混合在一起。现有的方法通常要么在去噪过程中将输入草图作为固定条件保留，要么在RGB合成之前将其精炼为干净的结构。前者假设笔画均匀可靠，可能会在生成过程中传播局部错误；后者必须在出现外观和语义上下文之前解决模糊的结构。我们提出了SketchSense，一个通过同步去噪相互作用的RGB和结构流来解读不完美草图指导的框架。双向注意力融合将外观生成与结构恢复相结合，生成一个精炼的结构，揭示模型不断演变的草图解读。短语级目标对齐了两个流的语义基础。草图感知空间调节进一步通过调节注意力和融合过程，适应草图在局部生成状态下的使用，而可选的签名先验则将保留与纠正的意图注入到特征表示和注意力行为中。在自然和结构复杂图像上的实验表明，在修复质量和结构保真度方面相较于现有方法有显著提升。
cs.CV / 61 / 2608.13194
Fidelity-Constrained Anchoring for Black-Box Denoisers
基于保真度约束的黑箱去噪器锚定方法
Abstract
We propose a fidelity-constrained framework that anchors the output of a black-box denoiser to its input without retraining and with little additional computation. The method linearly blends the denoised image with the input and selects the maximum blending factor that satisfies a prescribed local fidelity constraint using Peak Signal-to-Noise Ratio (PSNR) or Structural Similarity Index (SSIM). For PSNR control, a closed-form solution is obtained under a local constant-blending assumption. For SSIM control, we derive a tractable formulation based on inverse SSIM under the same assumption and solve it efficiently using iterative root finding. Experiments on DIV2K images with synthetic Gaussian noise and outputs from Real-ESRGAN and a non-local means denoiser show that the proposed anchoring strategy provides effective fidelity control while balancing denoising performance and statistical naturalness, as measured by the excess kurtosis of residual noise. In particular, SSIM-based anchoring yields more consistent behavior across noise levels than PSNR-based anchoring.
Chinese Translation
我们提出了一种保真度约束框架，该框架在不重新训练且计算量较小的情况下，将黑箱去噪器的输出锚定到其输入。该方法将去噪图像与输入图像线性混合，并选择满足规定局部保真度约束的最大混合因子，该约束使用峰值信噪比（Peak Signal-to-Noise Ratio, PSNR）或结构相似性指数（Structural Similarity Index, SSIM）进行控制。对于PSNR控制，在局部常量混合假设下获得了闭式解。对于SSIM控制，我们在相同假设下推导出基于逆SSIM的可处理公式，并通过迭代根查找有效地求解。对带有合成高斯噪声的DIV2K图像以及来自Real-ESRGAN和非局部均值去噪器的输出进行的实验表明，所提出的锚定策略在有效控制保真度的同时，平衡了去噪性能和统计自然性，这通过残余噪声的超额峰度进行测量。特别是，基于SSIM的锚定在不同噪声水平下表现出比基于PSNR的锚定更一致的行为。
cs.CV / 62 / 2608.13205
HPSD: Hybrid-Policy Self-Distillation for Text-Image-to-Video Diffusion Models
HPSD：用于文本-图像-视频扩散模型的混合策略自蒸馏
Abstract
Text-Image-to-Video (TI2V) models are an emerging unified architecture, where a single model simultaneously supports text-to-video (T2V) and image-to-video (I2V) generation. Given a high-quality first frame or a detailed textual prompt, TI2V models unlock substantially better visual quality than their T2V mode, raising a natural question: can the capability elicited by such privileged conditions be internalized into the model's own base generation ability? A common approach toward this goal is model self-distillation. However, the most straightforward solution, supervised fine-tuning, follows an off-policy strategy: its supervision is confined to teacher-generated endpoints from a fixed offline distribution rather than student-visited states, lacking precise correction tailored to the evolving policy. Recent on-policy distillation methods instead suffer from condition-state mismatch, where supervision is steered toward the given first frame instead of the student's actual content, misleading the correction. To achieve self-distillation that absorbs the teacher's privileged prior while retaining precise policy correction, in this work, we propose Hybrid-Policy Self-Distillation (HPSD), a novel self-distillation framework where a single TI2V model acts as both teacher and student under different conditions: the teacher operates in TI2V mode with a high-quality first frame and an enhanced prompt, while the student runs in the base T2V mode with only the vanilla prompt. Specifically, the student inherits off-policy teacher trajectory points as anchors, locally refines them toward its own policy, and finally receives velocity-level supervision on these self-generated roll-outs. Extensive experiments demonstrate that HPSD significantly improves T2V performance while also delivering notable TI2V gains, effectively strengthening the model's base generation ability.
Chinese Translation
文本-图像-视频（TI2V）模型是一种新兴的统一架构，其中单一模型同时支持文本到视频（T2V）和图像到视频（I2V）的生成。在给定高质量的第一帧或详细的文本提示时，TI2V模型的视觉质量显著优于其T2V模式，这引发了一个自然的问题：在这种特权条件下所激发的能力是否可以内化为模型自身的基础生成能力？实现这一目标的常见方法是模型自蒸馏。然而，最直接的解决方案——有监督微调，遵循了一种离线策略：其监督仅限于来自固定离线分布的教师生成的端点，而不是学生访问的状态，缺乏针对不断演变的策略量身定制的精确修正。最近的在线蒸馏方法则遭遇条件状态不匹配的问题，监督被引导到给定的第一帧，而不是学生的实际内容，从而误导了修正。为了实现能够吸收教师特权先验并保持精确策略修正的自蒸馏，本文提出了混合策略自蒸馏（HPSD），这是一种新颖的自蒸馏框架，其中单一的TI2V模型在不同条件下同时充当教师和学生：教师在TI2V模式下以高质量的第一帧和增强的提示进行操作，而学生则在仅使用普通提示的基础T2V模式下运行。具体而言，学生继承离线教师轨迹点作为锚点，局部地将其精炼为自己的策略，并最终在这些自生成的展开上接收速度级别的监督。大量实验表明，HPSD显著提高了T2V性能，同时也带来了显著的TI2V增益，有效增强了模型的基础生成能力。
cs.CV / 63 / 2608.13210
NARU: A Benchmark for NARrative Evolution and Cultural Nuance Understanding in Japanese Extreme Long Video
NARU：日本极长视频中叙事演变与文化细微差别理解的基准
Abstract
Long-form video understanding encompasses tasks that go beyond retrieving isolated events, including tracking an evolving narrative and interpreting social meaning that may remain implicit. However, existing benchmarks rarely evaluate these capabilities jointly, particularly in high-context, non-English media. To address this gap, we introduce NARU, a benchmark designed to evaluate Narrative evolution and Reasoning on cultural Understanding in Japanese long-form video. NARU consists of 1,481 questions grounded in 155 videos totaling 146.8 hours, spanning four narrative and five cultural dimensions. To construct the benchmark at this scale, we propose a hierarchical memory-based annotation pipeline that transforms raw video into structured event, narrative, and cultural annotations, then generates questions via task-oriented synthesis and iterative shortcut removal. The construction process includes two native-speaker verification stages involving 68 annotators. Evaluations across eight model configurations reveal substantial limitations in both long-range narrative integration and culturally grounded reasoning. By exposing these persistent gaps, NARU offers a systematic testing ground for developing MLLMs capable of reliably interpreting long-form, high-context video.
Chinese Translation
长篇视频理解涵盖的任务超越了孤立事件的检索，包括跟踪不断演变的叙事和解释可能隐含的社会意义。然而，现有基准很少联合评估这些能力，尤其是在高语境、非英语媒体中。为了解决这一空白，我们推出了NARU，一个旨在评估日本长篇视频中叙事演变和文化理解推理的基准。NARU由1,481个问题构成，基于155个视频，总时长达到146.8小时，涵盖四个叙事维度和五个文化维度。为了构建如此规模的基准，我们提出了一种基于层次记忆的标注流程，将原始视频转化为结构化的事件、叙事和文化标注，然后通过任务导向的合成和迭代简化去除生成问题。构建过程包括两个母语者验证阶段，涉及68名标注员。对八种模型配置的评估揭示了在长距离叙事整合和文化基础推理方面的重大局限性。通过揭示这些持续存在的差距，NARU为开发能够可靠解释长篇、高语境视频的多模态大语言模型（MLLMs）提供了系统的测试平台。
cs.CV / 64 / 2608.13217
UniCon-Former: Unified Convolution Transformer is All You Need for Hand Gesture Recognition
UniCon-Former：统一卷积变换器是手势识别所需的一切
Abstract
Convolutional Neural Networks (CNNs) capture local features efficiently but struggle with global context due to their limited receptive field. On the other hand, transformers effectively capture global dependencies through self-attention but suffer from high redundancy and computational costs. Thus, to leverage the advantages of both CNNs and transformers, we propose a unified model (UniCon-Former) that aims to provide robust and efficient performance on dynamic hand gesture recognition. The unified approach helps the model to learn both local and global features. At the beginning of each transformer stage, the convolution projections help in decreasing the dimension of the input vectors of the transformer block. This creates a pyramidal structure at each transformer stage. These features enable the UniCon-Former to reduce resource usage than vanilla transformers, making it flexible for learning multi-scale and high-resolution features, which is required in hand gesture recognition. We have performed experiments with NVGesture and Briareo datasets and achieved state-of-the-art results with fewer parameters and MACs.
Chinese Translation
卷积神经网络（CNN）能够有效捕捉局部特征，但由于其有限的感受野，难以处理全局上下文。另一方面，变换器通过自注意力机制有效捕捉全局依赖关系，但面临高冗余和计算成本的问题。因此，为了充分利用CNN和变换器的优势，我们提出了一种统一模型（UniCon-Former），旨在为动态手势识别提供稳健且高效的性能。这种统一的方法帮助模型学习局部和全局特征。在每个变换器阶段的开始，卷积投影有助于降低变换器块输入向量的维度。这在每个变换器阶段创建了一个金字塔结构。这些特征使得UniCon-Former在资源使用上优于传统变换器，使其能够灵活地学习多尺度和高分辨率特征，这在手势识别中是必需的。我们在NVGesture和Briareo数据集上进行了实验，并以更少的参数和MACs达到了最先进的结果。
cs.CV / 65 / 2608.13223
Reliability analysis for BraTS-GoAT segmentation: a controlled robustness study of deep-ensemble uncertainty
BraTS-GoAT分割的可靠性分析：深度集成不确定性的受控鲁棒性研究
Abstract
Deep networks segment brain tumours accurately in-distribution, but can fail silently when the input differs from their training data. That risk is central to clinical deployment and is the premise of the BraTS-GoAT generalizability task. We ask not only how well a model segments, but whether its uncertainty knows when it is wrong. On BraTS-GoAT (Task 3) we train a 5-fold cross-validated nnU-Net baseline (one held-out prediction per case) and a 3-seed deep ensemble. Both are evaluated for calibration and error detection on a per-region relevant mask, aggregated per case. In-distribution the 3-seed ensemble improves modestly over the already strong single model on the same held-out split, with the clearest gain in calibration. The separation appears under shift. In a controlled robustness study using graded synthetic corruptions as a proxy for acquisition shift, the single model's confidence stays flat while its accuracy and calibration degrade. Inter-member disagreement instead rises steeply, about a quarter to a third above the clean condition, several times the single model's response. On the official validation leaderboard the 5-fold ensemble of those folds attains whole-tumour Dice 0.87. The generalization gap is concentrated on the harder regions, with a characteristic failure of missing small, satellite lesions on unseen cohorts. In the synthetic study, disagreement among the 3-seed members is a more sensitive case-level indicator of acquisition shift than single-model confidence. Its per-voxel error localisation weakens as severity grows. The contribution is a rigorous, honest reliability comparison rather than a claim that any one uncertainty method dominates.
Chinese Translation
深度网络在训练数据分布内能够准确分割脑肿瘤，但当输入与训练数据不同时时，可能会悄然失败。这种风险在临床应用中至关重要，也是BraTS-GoAT可推广性任务的前提。我们不仅关注模型的分割效果，还关注其不确定性是否能够识别何时出现错误。在BraTS-GoAT（任务3）中，我们训练了一个5折交叉验证的nnU-Net基线模型（每个病例一个保留预测）以及一个3种种子的深度集成模型。两者均在每个区域相关的掩膜上进行校准和错误检测评估，并按病例汇总。在训练数据分布内，3种种子的集成模型在同一保留分割上相较于已经强大的单模型有适度提升，校准的改善最为明显。该分离在数据偏移下显现。在使用分级合成损坏作为获取偏移的代理的受控鲁棒性研究中，单模型的置信度保持平稳，而其准确性和校准则下降。相反，成员间的不一致性急剧上升，约为干净条件下的四分之一到三分之一，是单模型响应的几倍。在官方验证排行榜上，这些折叠的5折集成模型达到了整体肿瘤的Dice系数0.87。可推广性差距集中在更难的区域，特征性地在未见的队列中遗漏小的卫星病灶。在合成研究中，3种种子成员之间的不一致性是获取偏移的更敏感的病例级指标，其每体素的错误定位在严重性增加时减弱。该研究的贡献在于提供了一个严格、诚实的可靠性比较，而不是声称任何一种不确定性方法占主导地位。
cs.CV / 66 / 2608.13226
CoverPrune: Coverage-Driven Token Pruning for 3D VLMs via Optimal Transport
CoverPrune：通过最优传输实现的基于覆盖的3D视觉语言模型的令牌剪枝
Abstract
While 3D Vision-Language Models (3D VLMs) have demonstrated remarkable spatial reasoning capabilities, they suffer from massive visual token counts that create severe computational bottlenecks during inference. Existing token pruning methods primarily rely on diversity-based selection, discarding similar tokens to maximize dispersion. However, in 3D environments, this approach frequently drops representative prototype tokens in favor of outliers, breaking the multi-view consistencies and geometric structures essential for spatial reasoning. In this paper, we propose a paradigm shift for 3D VLM token pruning: from maximizing diversity to preserving visual evidence coverage. We introduce CoverPrune, a training-free framework that formulates inference-time token pruning as an Optimal Transport (OT) problem. To overcome the intractable combinatorial subset selection inherent in this formulation, we design the Feature-Spatial-Temporal (FST) transport cost and target capacity, along with an efficient Spatial-Guided Greedy Selection (SGS) algorithm to approximate the OT objective. Furthermore, we propose CoverPrune-Lite, an accelerated variant utilizing spatially structured local matching for minimal overhead. Extensive experiments across multiple 3D visual-spatial reasoning benchmarks demonstrate that our methods achieve state-of-the-art token efficiency, maintaining robust reasoning performance even under highly aggressive pruning budgets. Visit our project website at https://github.com/Brucess/CoverPrune.
Chinese Translation
尽管3D视觉语言模型（3D VLMs）展现了卓越的空间推理能力，但它们面临着巨大的视觉令牌数量，这在推理过程中造成了严重的计算瓶颈。现有的令牌剪枝方法主要依赖于基于多样性的选择，丢弃相似的令牌以最大化分散性。然而，在3D环境中，这种方法经常会丢弃代表性的原型令牌，而偏向于异常值，从而破坏了空间推理所必需的多视图一致性和几何结构。本文提出了一种3D VLM令牌剪枝的范式转变：从最大化多样性转向保留视觉证据覆盖。我们引入了CoverPrune，一个无训练框架，将推理时的令牌剪枝形式化为一个最优传输（Optimal Transport, OT）问题。为了克服这一形式化中固有的难以处理的组合子集选择问题，我们设计了特征-空间-时间（Feature-Spatial-Temporal, FST）传输成本和目标容量，并提出了一种高效的空间引导贪婪选择（Spatial-Guided Greedy Selection, SGS）算法来近似OT目标。此外，我们还提出了CoverPrune-Lite，一个加速变体，利用空间结构化的局部匹配以实现最小开销。在多个3D视觉空间推理基准上的广泛实验表明，我们的方法实现了最先进的令牌效率，即使在高度激进的剪枝预算下也能保持稳健的推理性能。请访问我们的项目网站：https://github.com/Brucess/CoverPrune。
cs.CV / 67 / 2608.13239
Reasoning for Social Audio-Visual Question Answering: Where Do We Stand?
社交音视频问答推理：我们处于何种状态？
Abstract
Training Multimodal Large Language Models for audio-visual social understanding is a crucial step toward embodied social intelligence. Chain-of-thought (CoT) reasoning has become the dominant approach, with HumanOmniV2 and its IntentBench benchmark as a prominent reference point. In this context, we report three findings. First, IntentBench is highly noisy: $\sim$7% of questions are broken and $\sim$23% are trivially answerable without the video input. We remove the affected questions and release Intentbench-Prime. Second, current reasoning approaches are expensive and surprisingly ineffective. A simple Vanilla SFT baseline matches or outperforms existing reasoning methods across three benchmarks at a fraction of the cost, establishing it as an essential baseline for evaluating novel fine-tuning techniques. Third, our analysis reveals that substantial priors can be learned solely from the text modality and that using a textual caption instead of the video yields performance on par with Vanilla SFT. These surprising findings reveal the limitations of current MLLMs when it comes to social understanding. IntentBench-Prime, Vanilla SFT model, and code are publicly available.
Chinese Translation
训练多模态大型语言模型以实现音视频社交理解是迈向具身社交智能的重要一步。链式推理（Chain-of-thought, CoT）已成为主流方法，其中 HumanOmniV2 及其 IntentBench 基准是一个显著的参考点。在此背景下，我们报告了三项发现。首先，IntentBench 存在较高的噪声：约 7% 的问题存在缺陷，约 23% 的问题在没有视频输入的情况下可以轻松回答。我们移除了受影响的问题，并发布了 Intentbench-Prime。其次，目前的推理方法成本高昂且效果令人惊讶地低效。一个简单的 Vanilla SFT 基线在三个基准测试中与现有的推理方法相匹配或超越，但成本仅为其一小部分，确立了其作为评估新型微调技术的重要基线。第三，我们的分析表明，仅通过文本模态即可学习到大量先验知识，使用文本标题而非视频的性能与 Vanilla SFT 相当。这些令人惊讶的发现揭示了当前多模态大型语言模型在社交理解方面的局限性。IntentBench-Prime、Vanilla SFT 模型和代码均已公开发布。
cs.CV / 68 / 2608.13255
GeoCache: Training-Free Acceleration of Multi-View Texture Diffusion via Geometric Delta Transport
GeoCache：通过几何增量传输实现无训练的多视角纹理扩散加速
Abstract
Geometry-conditioned multi-view diffusion enables high-quality 3D texture generation, but its repeated per-view denoiser evaluations introduce substantial computational cost. Existing training-free accelerators primarily exploit temporal redundancy by reusing computation across denoising steps. In multi-view texturing, however, skipping a step also removes the cross-view interaction that continually aligns different observations of the same surface, leading to rapidly degraded consistency and fidelity. Our analysis identifies a complementary source of redundancy: although intermediate features remain view-specific, geometrically corresponding surface points exhibit transferable evolution in their predicted clean signals. Based on this observation, we introduce \gc{}, a training-free plugin that evaluates a rotating subset of anchor views and transports their geometry-aligned per-step $\xz$ updates to the remaining views. Periodic full-view computation controls accumulated error, while sampler-consistent reconstruction preserves the denoising trajectory. \gc{} requires neither retraining nor architectural modification and uses the position maps already available in geometry-conditioned texturing pipelines. Across Hunyuan3D-2.1, SyncMVD, and MVPainter, \gc{} achieves a stronger speed--fidelity trade-off than temporal caches and step reduction at operating points above $2\times$. On Hunyuan3D-2.1, it delivers a $2.21\times$ denoiser-loop speedup with an MV-LPIPS of 0.0293 and an MV-PSNR of 33.60 dB, providing the best fidelity among all tested methods above $2\times$. The same transferred configuration reaches the highest speedup and lowest FLOPs on SyncMVD, while \gc{} achieves the lowest FLOPs and best fidelity among the accelerated methods on MVPainter. These results establish cross-view geometry as an effective acceleration axis for multi-view texture diffusion.
Chinese Translation
基于几何条件的多视角扩散能够生成高质量的3D纹理，但其在每个视角上重复进行去噪评估带来了相当大的计算成本。现有的无训练加速器主要通过在去噪步骤之间重用计算来利用时间冗余。然而，在多视角纹理生成中，跳过某一步骤也会去除跨视角的交互，这种交互持续对齐同一表面的不同观测，导致一致性和保真度迅速下降。我们的分析识别出一种互补的冗余来源：尽管中间特征仍然是视角特定的，但几何对应的表面点在其预测的干净信号中表现出可转移的演变。基于这一观察，我们引入了 extit{GeoCache}（ extit{gc}），这是一种无训练的插件，它评估一个旋转的锚视角子集，并将其几何对齐的每步 $ extit{xz}$ 更新传输到其余视角。定期的全视角计算控制累积误差，而采样器一致的重建保持了去噪轨迹。 extit{gc} 不需要重新训练或架构修改，并利用几何条件纹理管道中已经可用的位置图。在 Hunyuan3D-2.1、SyncMVD 和 MVPainter 上， extit{gc} 实现了比时间缓存和步骤减少更强的速度-保真度权衡，操作点超过 $2 imes$。在 Hunyuan3D-2.1 上，它提供了 $2.21 imes$ 的去噪循环加速，MV-LPIPS 为 0.0293，MV-PSNR 为 33.60 dB，在所有测试方法中提供了超过 $2 imes$ 的最佳保真度。同样的转移配置在 SyncMVD 上达到了最高的加速和最低的 FLOPs，而在 MVPainter 上， extit{gc} 在加速方法中实现了最低的 FLOPs 和最佳的保真度。这些结果确立了跨视角几何作为多视角纹理扩散的有效加速轴。
cs.CV / 69 / 2608.13309
How Good are Foundation Models in Longitudinal MRI Disease Progression Reasoning?
基础模型在纵向MRI疾病进展推理中的表现如何？
Abstract
Magnetic Resonance Imaging (MRI) interpretation is fundamental to clinical decision-making, requiring radiologists to integrate multi-view anatomical planes across sequential timepoints while precisely localizing interval changes. However, existing vision-language benchmarks remain confined to single-timepoint, single-view interpretation, failing to capture the temporal-spatial reasoning essential to radiologic practice. We introduce the Time-Aware Multi-View MRI Benchmark, an evaluation framework unifying multi-view anatomical input, temporal reasoning across longitudinal scans, and structured localization guidance. The benchmark comprises 3,920 expert-verified question-answer pairs derived from 890 patients across over 3,200 longitudinal MRI timepoints, drawn from seven clinical cohorts covering glioblastoma, neurodegeneration, vestibular schwannoma, and brain metastases, in open-ended, multiple-choice, and binary formats, requiring models to identify anatomical regions of maximal change, characterize progression across sequences and views, and provide structured guidance specifying boundaries, imaging features, and confounders. Experiments across 16 vision-language models reveal moderate temporal alignment but systematic failure on change direction recognition and volumetric quantification, while multi-view inputs improve spatial localization yet degrade temporal reasoning in compact architectures. Our benchmark provides a systematic framework for evaluating progression tracking, interval change localization, and temporal ordering, which are essential for clinical deployment. Code, evaluation splits, and the dataset are available at: https://github.com/wafaAlghallabi/Time-Aware-MRI.
Chinese Translation
磁共振成像（MRI）的解读对临床决策至关重要，要求放射科医生在多个视角的解剖平面中整合跨时间点的信息，同时精确定位间隔变化。然而，现有的视觉-语言基准测试仍局限于单时间点、单视角的解读，未能捕捉到放射学实践中所需的时间-空间推理。我们提出了时间感知多视角MRI基准测试，这是一个评估框架，统一了多视角解剖输入、纵向扫描中的时间推理和结构化定位指导。该基准测试包含3920对经过专家验证的问题-答案对，来源于890名患者的3200多个纵向MRI时间点，涵盖了胶质母细胞瘤、神经退行性疾病、前庭神经瘤和脑转移等七个临床队列，以开放式、多选和二元格式呈现，要求模型识别最大变化的解剖区域、描述序列和视角间的进展，并提供结构化指导，指定边界、影像特征和混杂因素。在16个视觉-语言模型的实验中，发现时间对齐程度适中，但在变化方向识别和体积量化方面存在系统性失败，而多视角输入改善了空间定位，但在紧凑架构中降低了时间推理能力。我们的基准测试为评估进展追踪、间隔变化定位和时间排序提供了系统框架，这些对于临床应用至关重要。代码、评估分割和数据集可在以下网址获取：https://github.com/wafaAlghallabi/Time-Aware-MRI。
cs.CV / 70 / 2608.13343
AmalthAI: An Open-Source Computer Vision Platform for Cultural Heritage
AmalthAI：一个用于文化遗产的开源计算机视觉平台
Abstract
Computer vision (CV) and machine learning (ML) offer new tools for cultural heritage (CH) artifact analysis, but the CV/ML pipeline remains largely inaccessible to CH domain experts, who lack the background to configure, train, or assess models. We present AmalthAI, an open-source CV platform that bridges this gap, enabling non-ML CH experts to independently produce and validate archaeologically meaningful findings. The interface covers dataset management, training, and inference for classification, segmentation, and object detection, with Kubeflow and Katib handling scalable training and hyperparameter search. Grad-CAM localizes the image region behind a prediction, and a vision-language model (VLM) adds a text description of it for expert review. Since archaeological data is often state-owned or rights-encumbered and cannot leave institutional custody, AmalthAI's self-hostable deployment ensures sensitive data is kept within premises. We test the platform on an archaeological use case built on a custom dataset of clay textile imprints, where CH experts trained and validated segmentation, and classification models for hypothesis testing. We provide the implementation code at https://github.com/TEXTaiLES/AmalthAI.
Chinese Translation
计算机视觉（CV）和机器学习（ML）为文化遗产（CH）文物分析提供了新的工具，但CV/ML管道对缺乏配置、训练或评估模型背景的CH领域专家来说仍然难以接触。我们提出了AmalthAI，这是一个开源的CV平台，旨在弥补这一差距，使非ML的CH专家能够独立生成和验证考古学上有意义的发现。该平台的界面涵盖数据集管理、训练和推理，支持分类、分割和物体检测，Kubeflow和Katib负责可扩展的训练和超参数搜索。Grad-CAM用于定位预测背后的图像区域，并且视觉-语言模型（VLM）为专家审查添加了文本描述。由于考古数据通常是国有或受权利限制的，无法离开机构保管，AmalthAI的自托管部署确保敏感数据保留在场所内。我们在一个基于定制数据集的粘土纺织印记的考古用例上测试了该平台，CH专家训练并验证了用于假设检验的分割和分类模型。我们在https://github.com/TEXTaiLES/AmalthAI提供了实现代码。
cs.CV / 71 / 2608.13368
Sign Language Video Synthesis via Loss-Guided Multi-Expert GANs
基于损失引导的多专家生成对抗网络的视频手语合成
Abstract
This preliminary technical report presents a framework for sign language video synthesis using a loss-guided multi-expert Generative Adversarial Network (GAN) to enhance communication for individuals with hearing impairments. Three specialized discriminators -- global, hand, and head -- each guide a corresponding expert branch in the generator toward a distinct visual region, enabling implicit feature specialization without explicit diversity losses. To stabilize this multi-discriminator system, whose early-phase training otherwise exhibits chaotic dynamics, we introduce a United Loss consensus mechanism that regularizes each discriminator toward the ensemble average at a 10% weight. Each branch further adopts a dual-pathway convolutional-transformer design with learnable AdaptiveFeatureFusion, balancing the stability of convolutions against the detail of windowed self-attention. The generator is trained using an alternating three-mode schedule (discriminator, holistic generation, branch-specialized generation). On a custom 156GB dataset with a filtered test set that removes easy and repetitive samples, our 0.2B-parameter variant achieves 29.8 PSNR and the 1.3B-parameter variant achieves 30.7 PSNR, with inference VRAM footprints of 1.5 GB and 8 GB respectively, enabling deployment on consumer-grade hardware. Full ablation studies remain ongoing due to the 2-3 month training cycle on a single GPU. The system was showcased at the 2025 Hong Kong Frontier Technology Summit.
Chinese Translation
本初步技术报告提出了一种使用损失引导的多专家生成对抗网络（GAN）进行手语视频合成的框架，以增强听力障碍人士的沟通能力。三个专门的鉴别器——全局鉴别器、手部鉴别器和头部鉴别器——分别引导生成器中的对应专家分支朝向不同的视觉区域，从而实现隐式特征专业化，而无需显式的多样性损失。为了稳定这一多鉴别器系统，避免早期训练阶段出现混乱的动态，我们引入了一种统一损失共识机制，将每个鉴别器的权重正则化为10%的集成平均值。每个分支还采用了双路径卷积-变换器设计，结合可学习的自适应特征融合，平衡卷积的稳定性与窗口自注意力的细节。生成器使用交替的三模式调度进行训练（鉴别器、整体生成、分支专业生成）。在一个定制的156GB数据集上，经过筛选的测试集去除了简单和重复的样本，我们的0.2B参数变体达到了29.8 PSNR，而1.3B参数变体达到了30.7 PSNR，推理所需的显存分别为1.5 GB和8 GB，能够在消费级硬件上部署。由于单个GPU的训练周期为2-3个月，完整的消融研究仍在进行中。该系统在2025年香港前沿科技峰会上进行了展示。
cs.CV / 72 / 2608.13381
Reconstructing Historical Manuscripts through MSI: The Potential of Contrast in Assessing Image Quality and Legibility
通过多光谱成像重建历史手稿：对比在评估图像质量和可读性中的潜力
Abstract
Digital restoration of historical manuscript images aims to improve readability while preserving the authenticity of cultural heritage documents. However, evaluating quality of restored manuscripts remains challenging, where readability is often subjective and expert annotations are scarce. This study investigates the suitability of contrast-based image quality measures to assess quality and legibility of reconstructed manuscript images from multi-spectral imaging. Two experiments were conducted with publicly-available data sets, facilitating manual quality scores by experts and full-reference image quality measures as reference evaluations. The results show that potential contrast achieves the highest correlation with expert ratings, while contrast-to-noise ratio demonstrates the strongest agreement with full-reference quality measures. Overall, contrast-based measures consistently outperform general image quality measures, demonstrating their potential as objective indicators of manuscript legibility and reconstruction quality.
Chinese Translation
历史手稿图像的数字修复旨在提高可读性，同时保持文化遗产文档的真实性。然而，评估修复手稿的质量仍然具有挑战性，因为可读性往往是主观的，专家注释也很稀缺。本研究探讨了基于对比的图像质量度量在评估多光谱成像重建手稿图像的质量和可读性方面的适用性。我们进行了两项实验，使用公开可用的数据集，便于专家进行手动质量评分，并将全参考图像质量度量作为参考评估。结果表明，潜在对比度与专家评分的相关性最高，而对比噪声比与全参考质量度量的一致性最强。总体而言，基于对比的度量在性能上始终优于一般图像质量度量，显示出其作为手稿可读性和重建质量的客观指标的潜力。
cs.CV / 73 / 2608.13385
When Is a Task Vector Enough? An Empirical Theory of Implicit Multimodal ICL
任务向量何时足够？隐式多模态上下文学习的实证理论
Abstract
Implicit multimodal in-context learning compresses demonstrations into internal interventions, ranging from static task vectors to query-conditioned transformations and attention routing. Despite their common goal, these methods differ substantially in how the intervention depends on the query and where it modifies the model, leaving unclear which additional complexity is necessary for a given task. We propose the Selection--Realization Hypothesis. It views demonstrations as inducing a compact family of internal changes from which the query selects, while the model's computation constrains how the selected change can be implemented. We evaluate this account using controlled multimodal tasks in which query dependence varies without changing the underlying task primitives or prompt format. By contrasting correct demonstrations with matched counterfactuals, we measure the structure of explicit M-ICL and test whether it predicts intervention behavior. We find that the success of a static task vector is closely tied to how much of the demonstration-induced change is shared across queries. Additional intervention complexity becomes useful when explicit M-ICL contains query-specific or distributed structure that a local additive shift cannot recover. These relationships extend to natural VQA benchmarks and support cost-aware method selection without access to test performance. Our results provide a unified empirical theory of when demonstrations can be compressed into a task vector and when a more expressive intervention is warranted.
Chinese Translation
隐式多模态上下文学习将示例压缩为内部干预，范围从静态任务向量到查询条件变换和注意力路由。尽管这些方法的共同目标相同，但它们在干预如何依赖于查询以及在何处修改模型方面存在显著差异，这使得在特定任务中不清楚需要哪些额外的复杂性。我们提出了选择-实现假说（Selection--Realization Hypothesis）。该假说将示例视为诱导一组紧凑的内部变化，从中查询进行选择，而模型的计算则限制了所选变化的实现方式。我们通过控制多模态任务来评估这一理论，在这些任务中，查询依赖性变化而不改变基础任务原语或提示格式。通过对比正确的示例与匹配的反事实，我们测量了显式多模态上下文学习（explicit M-ICL）的结构，并测试其是否能够预测干预行为。我们发现，静态任务向量的成功与示例诱导的变化在查询之间的共享程度密切相关。当显式多模态上下文学习包含查询特定或分布式结构，而局部加性偏移无法恢复时，额外的干预复杂性变得有用。这些关系扩展到自然视觉问答基准，并支持在没有测试性能访问的情况下进行成本意识的方法选择。我们的结果提供了一个统一的实证理论，阐明了何时可以将示例压缩为任务向量，以及何时需要更具表现力的干预。
cs.CV / 74 / 2608.13391
Context-Matched Distillation: Teacher Causality for Autoregressive Video Distillation
上下文匹配蒸馏：自回归视频蒸馏的教师因果性
Abstract
Interactive autoregressive video generation demands both low-latency rollouts and precise online control. Few-step distillation accelerates generation by reducing denoising steps, while online control imposes a causal constraint: frames and blocks should depend on history and controls available during generation. Existing video distribution matching distillation (DMD) pipelines, however, often supervise causal few-step students using bidirectional teachers that score complete clips. The score for a target can therefore depend on future frames and controls that were unavailable when the student generated it, misaligning teacher supervision with the student's causal information set. We introduce Context-Matched Distillation (CMD), a causal DMD framework that aligns teacher supervision with the information available when each target is generated. CMD replaces bidirectional full-clip scoring with a causal teacher that evaluates each target without access to future frames or controls. The same causal teacher initializes the few-step student, establishing a consistent causal formulation across teacher training, student distillation, and inference. Beyond aligning the temporal information boundary, Prefix Scoring matches supervision to the student's realized rollout context by evaluating each target under the cached student-generated prefix that produced it. Prefix Corruption further stabilizes training by perturbing unreliable prefixes produced early in training while preserving this target-context alignment. With a simple causal formulation, CMD naturally extends to frame-wise and chunk-wise generation, long video distillation, and camera-conditioned distillation. Experiments demonstrate state-of-the-art aggregate performance among autoregressive methods on both short- and long-video benchmarks, together with substantially improved adherence to time-varying camera controls.
Chinese Translation
交互式自回归视频生成要求低延迟的生成过程和精确的在线控制。少步蒸馏通过减少去噪步骤来加速生成，而在线控制则施加了因果约束：帧和块应依赖于生成过程中可用的历史信息和控制。然而，现有的视频分布匹配蒸馏（DMD）流程通常使用双向教师来监督因果少步学生，这些教师对完整片段进行评分。因此，目标的评分可能依赖于未来的帧和控制，而这些在学生生成目标时并不可用，从而导致教师监督与学生的因果信息集不一致。我们提出了上下文匹配蒸馏（CMD），这是一个因果DMD框架，它将教师监督与每个目标生成时可用的信息对齐。CMD用一个因果教师替代了双向完整片段评分，该教师在评估每个目标时不访问未来的帧或控制。相同的因果教师初始化少步学生，从而在教师训练、学生蒸馏和推理之间建立了一致的因果公式。除了对齐时间信息边界外，前缀评分通过在缓存的学生生成前缀下评估每个目标，将监督与学生实现的生成上下文匹配。前缀腐蚀进一步通过扰动训练早期产生的不可靠前缀来稳定训练，同时保持目标上下文对齐。通过简单的因果公式，CMD自然扩展到逐帧和逐块生成、长视频蒸馏以及相机条件蒸馏。实验表明，CMD在短视频和长视频基准测试中，在自回归方法中实现了最先进的综合性能，并显著提高了对时变相机控制的遵循程度。
cs.CV / 75 / 2608.13416
StreamTTT: Reconciling Real-Time Perception and Long-Term Memory in Streaming VLMs
StreamTTT：调和流媒体视觉语言模型中的实时感知与长期记忆
Abstract
Humans effortlessly perceive the present while remembering the past, yet streaming VLMs often trade off real-time perception against long-term memory. Prior work shows that shortening the context can sharpen current-scene perception at the expense of long-range recall. To reconcile these abilities, we introduce StreamTTT, which writes long-range history into online-updated fast weights outside the attention context. This leaves a short sliding key-value cache dedicated to recent evidence, mitigating attention dilution. We train StreamTTT jointly on offline long-video QA and a newly constructed real-time QA corpus. On OVO-Bench, StreamTTT-4B outperforms SimpleStream-4B by 1.4 points in real-time perception and 3.7 points in backward tracing. It also remains competitive with the larger SimpleStream-8B on the Real-Time Visual Understanding (RTVU) subset of StreamingBench. Our code will be released.
Chinese Translation
人类能够轻松感知当前，同时记住过去，而流媒体视觉语言模型（VLM）往往在实时感知与长期记忆之间进行权衡。先前的研究表明，缩短上下文可以提高当前场景的感知能力，但会牺牲长距离的回忆。为了调和这些能力，我们提出了StreamTTT，它将长期历史写入在线更新的快速权重中，超出注意力上下文。这为近期证据保留了一个短的滑动键值缓存，从而减轻了注意力稀释的问题。我们在离线长视频问答和新构建的实时问答语料库上共同训练StreamTTT。在OVO-Bench上，StreamTTT-4B在实时感知方面比SimpleStream-4B提高了1.4分，在回溯追踪方面提高了3.7分。它在StreamingBench的实时视觉理解（RTVU）子集上也与更大的SimpleStream-8B保持竞争力。我们的代码将会发布。
cs.CV / 76 / 2608.13441
Edit2TikZ: A Comprehensive and Challenging Benchmark for Scientific Figure Editing with TikZ
Edit2TikZ：一个全面且具有挑战性的科学图形编辑基准测试
Abstract
Although multimodal large language models (MLLMs) have shown substantial potential in visual understanding and graphic code generation, editing scientific figures through code presents a greater challenge: a model must jointly recover visual structure, ground the requested change, generate compilable code, and preserve all unrelated content. While existing TikZ benchmarks mainly focus on figure reconstruction and generation, few systematically evaluate instruction-guided scientific figure editing with compilable code. We introduce Edit2TikZ, a comprehensive benchmark for scientific figure editing tasks, featuring 1,548 diverse and high-quality samples. Edit2TikZ combines real-world and controlled synthetic edit cases, supports both textual and visual localization request, and contains multi-step editing, each with step-level annotations. We further construct a human-aligned evaluation framework to measure whether a requested edit is completed while irrelevant content is preserved. Utilizing Edit2TikZ, we evaluate 14 mainstream MLLMs and find that current systems remain unreliable: on average, proprietary models achieve a compilation success rate of merely 75% and remain limited in both figure restoration and edit correctness, while compact models below 9B struggle further with instruction following and complete figure generation. Therefore, we build a mixed training set TikZEditMix and adopt reconstruction-then-editing curriculum learning for compact models. On Qwen3.5-4B, this training improves the compilation success rate from 45.35% to 83.40% and yields an average improvement of 18.7 points across our proposed evaluation metrics. The code and data will be released at https://github.com/Solunny/Edit2TikZ.
Chinese Translation
尽管多模态大型语言模型（MLLMs）在视觉理解和图形代码生成方面显示出巨大的潜力，但通过代码编辑科学图形却面临更大的挑战：模型必须共同恢复视觉结构，确定请求的更改，生成可编译代码，并保留所有无关内容。现有的 TikZ 基准主要集中在图形重建和生成上，系统地评估基于指令的科学图形编辑（使用可编译代码）的研究较少。我们介绍了 Edit2TikZ，这是一个全面的科学图形编辑任务基准，包含 1,548 个多样且高质量的样本。Edit2TikZ 结合了真实世界和受控合成编辑案例，支持文本和视觉定位请求，并包含多步编辑，每一步都有逐步注释。我们进一步构建了一个人类对齐的评估框架，以衡量请求的编辑是否完成，同时保留无关内容。利用 Edit2TikZ，我们评估了 14 个主流 MLLMs，发现当前系统仍然不可靠：平均而言，专有模型的编译成功率仅为 75%，在图形恢复和编辑正确性方面也存在局限，而低于 9B 的紧凑模型在遵循指令和完整图形生成方面表现更差。因此，我们构建了一个混合训练集 TikZEditMix，并对紧凑模型采用重建后编辑的课程学习。在 Qwen3.5-4B 上，这种训练将编译成功率从 45.35% 提高到 83.40%，并在我们提出的评估指标上平均提高了 18.7 分。代码和数据将发布在 https://github.com/Solunny/Edit2TikZ。
cs.CV / 77 / 2608.13453
UniTexture: Cross-Task Universal Adversarial Textures for Vision-Language-Action Models
UniTexture：用于视觉-语言-动作模型的跨任务通用对抗纹理
Abstract
Vision-Language-Action (VLA) models have emerged as generalist robotic policies capable of following diverse language instructions and performing a wide range of manipulation tasks. However, their direct control over embodied agents also exposes them to adversarial interference that may cause unsafe physical behaviors. Existing attacks on robotic policies are typically optimized for a single task or instruction, leaving the cross-task vulnerabilities of multitask VLAs largely unexplored. We introduce UniTexture, a cross-task universal adversarial texture attack that uses a single textured 3D object to induce targeted deviations in VLA action predictions across multiple tasks. UniTexture backpropagates gradients from the policy's action outputs to surface texture parameters through a differentiable renderer. It jointly optimizes the shared texture over a distribution of tasks, instructions, states, and viewpoints using a targeted action-space objective, steering predicted actions toward attacker-defined targets without optimizing a separate texture for each task. We evaluate UniTexture on OpenVLA and $\pi_{0.5}$ across diverse manipulation tasks and multiple evaluation settings. UniTexture reduces the mean task success rate from 90.0% under benign conditions to 48.4% under attack, induces target-aligned action shifts, and further exhibits cross-suite and cross-model transfer without re-optimization. Together, these findings reveal shared cross-task vulnerabilities in multitask VLAs that can be systematically exploited through a single adversarial surface texture.
Chinese Translation
视觉-语言-动作（VLA）模型已成为能够遵循多样语言指令并执行广泛操作任务的通用机器人策略。然而，它们对具身代理的直接控制也使其暴露于可能导致不安全物理行为的对抗干扰之中。现有针对机器人策略的攻击通常针对单一任务或指令进行优化，导致多任务 VLA 的跨任务脆弱性尚未得到充分探索。我们提出了 UniTexture，一种跨任务通用对抗纹理攻击，利用单一纹理化的 3D 物体在多个任务中引发 VLA 行动预测的目标偏差。UniTexture 通过可微渲染器将策略的行动输出的梯度反向传播到表面纹理参数。它通过针对行动空间的目标，联合优化在任务、指令、状态和视角分布上的共享纹理，推动预测的行动朝向攻击者定义的目标，而无需为每个任务优化单独的纹理。我们在 OpenVLA 和 $ ext{π}_{0.5}$ 上评估 UniTexture，涵盖多种操作任务和多个评估设置。UniTexture 将在良性条件下的平均任务成功率从 90.0% 降低到 48.4% 在攻击下，诱导目标对齐的行动偏移，并进一步展示了跨套件和跨模型的迁移能力，无需重新优化。这些发现共同揭示了多任务 VLA 中共享的跨任务脆弱性，可以通过单一对抗表面纹理进行系统性利用。
cs.CV / 78 / 2608.13455
Evaluation of Clinically Steerable Retinal Image Generation from Foundation Model Latent Spaces
基于基础模型潜在空间的临床可控视网膜图像生成评估
Abstract
Medical foundation models learn latent representations of clinically meaningful phenotypes, yet their ability to support controllable image generation remains largely unexplored. We evaluate four retinal foundation models within the representation tokenizer framework and examine whether demographic and clinical information encoded in latent representations from foundation models is preserved during synthetic image generation. We show that generated representations and images faithfully inherit phenotype information when evaluated within their originating foundation models, consistently outperforming conventional latent diffusion on multiple downstream prediction tasks. However, these gains largely disappear when evaluated using classifiers trained on real images, revealing a previously uncharacterised synthetic-to-real representation gap. These findings demonstrate that foundation-model latent spaces provide a powerful substrate for controllable retinal synthesis while highlighting the need to better align synthetic representations with real-image distributions.
Chinese Translation
医学基础模型学习临床相关表型的潜在表示，但其支持可控图像生成的能力仍然未得到充分探索。我们在表示标记器框架内评估了四个视网膜基础模型，并检查了基础模型中编码的人口统计和临床信息在合成图像生成过程中是否得以保留。我们展示了生成的表示和图像在其来源基础模型中评估时忠实地继承了表型信息，并在多个下游预测任务中始终优于传统的潜在扩散。然而，当使用在真实图像上训练的分类器进行评估时，这些优势大部分消失，揭示了一个先前未被描述的合成到真实的表示差距。这些发现表明，基础模型的潜在空间为可控的视网膜合成提供了强大的基础，同时突显了更好地将合成表示与真实图像分布对齐的必要性。
cs.CV / 79 / 2608.13458
Fine-Grained Action Recognition with Cross-Attentive Latent Sparse Experts
基于交叉注意力的细粒度动作识别与稀疏专家模型
Abstract
Fine-grained human action recognition (FHAR) must distinguish visually similar actions that differ mainly in body configuration, timing, or local appearance. RGB representations retain visual context but often suppress joint-level geometry, whereas skeleton representations encode kinematics but discard dense spatial detail. We introduce FineX, which factorizes fine-grained cues into RGB appearance, pose heatmap geometry, and skeletal-graph topology. Pairwise cross-attention enables symmetric, stream-preserving information exchange, followed by a streamwise latent sparse Mixture-of-Experts that routes each representation to a content-dependent subset of shared experts, regularized by a load-balancing objective. FineX achieves state-of-the-art results on Gym99, Gym288, and Diving48. On the long-tailed Gym288, it raises mean class accuracy from 68.6% to 76.2% (+7.6 points) without textual supervision or large-scale vision-language pre-training, demonstrating the benefit of structured visual-pose-graph fusion and conditional expert refinement for FHAR.
Chinese Translation
细粒度人类动作识别（FHAR）必须区分在身体构型、时序或局部外观上主要不同但视觉上相似的动作。RGB 表示保留了视觉上下文，但往往抑制了关节级几何信息，而骨骼表示则编码了运动学信息，但丢弃了密集的空间细节。我们提出了 FineX，它将细粒度线索分解为 RGB 外观、姿态热图几何和骨骼图拓扑。成对的交叉注意力使得信息交换对称且保持流的连续性，随后通过流式稀疏专家混合模型（Mixture-of-Experts）将每个表示路由到共享专家的内容相关子集，并通过负载平衡目标进行正则化。FineX 在 Gym99、Gym288 和 Diving48 上达到了最先进的结果。在长尾数据集 Gym288 上，它将平均类别准确率从 68.6% 提升至 76.2%（+7.6 个百分点），且未使用文本监督或大规模视觉-语言预训练，展示了结构化视觉-姿态-图融合和条件专家优化对 FHAR 的益处。
cs.CV / 80 / 2608.13460
SNM-VFI: Symmetric Nonlinear Motion-Guided Generative Video Frame Interpolation
SNM-VFI：对称非线性运动引导的生成视频帧插值
Abstract
We propose Symmetric Nonlinear Motion-guided Generative Video Frame Interpolation (SNM-VFI), a training-free framework for motion-controllable generative video frame interpolation with pre-trained optical flow and video diffusion models. Unlike conventional diffusion-based VFI methods that synthesize intermediate frames from random noise, SNM-VFI guides the generative process with correspondence-aware frames produced by a symmetric nonlinear motion model. Specifically, we first utilize a pre-trained optical flow model to construct multi-frame nonlinear flow-based intermediate frames and confidence maps. These flow-guided frames are then encoded as latent priors to initialize and iteratively guide a pre-trained Video Diffusion model, enabling the diffusion model to preserve dense motion correspondence while improving perceptual realism. To further enhance output quality, we employ confidence maps to fuse structurally reliable flow-based predictions with diffusion-generated details in uncertain regions such as occlusions and object boundaries. Extensive evaluations on challenging benchmarks, including DAVIS, Sintel, and KITTI, demonstrate that SNM-VFI achieves strong perceptual quality, competitive reconstruction accuracy, and robust temporal coherence across diverse motion scenarios.
Chinese Translation
我们提出了对称非线性运动引导的生成视频帧插值（SNM-VFI），这是一个无需训练的框架，旨在实现可控运动的生成视频帧插值，利用预训练的光流和视频扩散模型。与传统的基于扩散的视频帧插值（VFI）方法不同，后者是从随机噪声合成中间帧，SNM-VFI通过对称非线性运动模型生成的对应感知帧来引导生成过程。具体而言，我们首先利用预训练的光流模型构建多帧非线性流基础的中间帧和置信度图。这些流引导的帧随后被编码为潜在先验，以初始化并迭代引导预训练的视频扩散模型，从而使扩散模型能够在提高感知真实感的同时保持密集的运动对应关系。为了进一步提升输出质量，我们采用置信度图将结构上可靠的流基础预测与在不确定区域（如遮挡和物体边界）中生成的扩散细节融合。对包括DAVIS、Sintel和KITTI在内的挑战性基准的广泛评估表明，SNM-VFI在多样化运动场景中实现了强大的感知质量、竞争性的重建精度和稳健的时间一致性。
cs.CV / 81 / 2608.13463
MLLM-Routed Heterogeneous Ensembles for Robust Cross-Dataset Image Classification
基于MLLM路由的异构集成模型用于稳健的跨数据集图像分类
Abstract
Modern image classification models excel when trained on single task-specific datasets but often struggle to generalize across domains and difficulty levels. We propose ARMDIL, an Adaptive Router for Multi-Domain Image classification with LLMs. ARMDIL is an ensemble that uses a multimodal large language model (MLLM) agent to dynamically route each image to the most suitable vision backbone. Our diverse ensemble employs convolutional neural networks (ResNets), self-supervised representation learners (SSL), and vision-language models (VLMs), each trained on a unified label space constructed from multiple image datasets with differing distributions and characteristics. Empirical evaluations illuminate the distinct capabilities and vulnerabilities of each architecture across disparate visual domains. Crucially, we show that ARMDIL effectively navigates these trade-offs, performing competitively with specialized training-based routers. Furthermore, it drastically improves adaptability by allowing new information to be integrated via simple prompt modifications, while enhancing interpretability through natural language reasoning traces. These advances in cross-dataset image classification pave the way for more reliable general-purpose vision systems such as AI assistants and autonomous robots.
Chinese Translation
现代图像分类模型在单一任务特定数据集上训练时表现优异，但在跨领域和不同难度级别的泛化能力上常常面临挑战。我们提出了ARMDIL，一种基于大规模语言模型（LLMs）的多领域图像分类自适应路由器。ARMDIL是一个集成模型，利用多模态大语言模型（MLLM）代理动态地将每幅图像路由到最合适的视觉骨干网络。我们的多样化集成采用了卷积神经网络（ResNets）、自监督表示学习者（SSL）和视觉-语言模型（VLMs），每个模型都在一个由多个具有不同分布和特征的图像数据集构建的统一标签空间上进行训练。实证评估揭示了每种架构在不同视觉领域中的独特能力和脆弱性。关键是，我们展示了ARMDIL有效地在这些权衡中导航，其性能与基于专门训练的路由器相当。此外，它通过简单的提示修改允许新信息的整合，从而大幅提高适应性，同时通过自然语言推理痕迹增强可解释性。这些在跨数据集图像分类方面的进展为更可靠的通用视觉系统（如AI助手和自主机器人）铺平了道路。
cs.CV / 82 / 2608.13478
MapRoute++: Surrogate-Guided Semantic Routing for Visual Concept Unlearning
MapRoute++：用于视觉概念遗忘的代理引导语义路由
Abstract
We present our submission to Task 3 of the Gen$\mu$ 2.0 Challenge on visual concept unlearning. Building on MapRoute, we introduce task-specific training objectives, richer concept representations, and semantic routing for concept-specific mapper selection. Our approach improves robust concept removal while preserving unrelated and semantically adjacent concepts. On the official benchmark, evaluated using the Erasing-Retention-Robustness (ERR) metric on Stable Diffusion v1.4, our method outperforms the state-of-the-art baseline by 12.1\% on average across the five concept categories, achieving substantial gains.
Chinese Translation
我们提交了对Gen$.0挑战赛任务3的贡献，旨在实现视觉概念的遗忘。在MapRoute的基础上，我们引入了特定任务的训练目标、更丰富的概念表示以及用于概念特定映射器选择的语义路由。我们的方法在有效去除特定概念的同时，能够保持无关和语义相邻的概念。在官方基准测试中，使用稳定扩散（Stable Diffusion）v1.4的擦除-保留-鲁棒性（ERR）指标进行评估，我们的方法在五个概念类别上平均超越了最先进的基线12.1\%，取得了显著的提升。
cs.CV / 83 / 2608.13489
DreamX-Phi 1.0: Action-Conditioned Video World Model for Robotic Manipulation
DreamX-Phi 1.0：用于机器人操作的动作条件视频世界模型
Abstract
We present \textbf{DreamX-Phi 1.0}, an action-conditioned video world model for robotic manipulation that, given an observed frame, a language instruction, and a prescribed action sequence comprising end-effector poses and gripper states, predicts the resulting future observations. Yet realism alone does not guarantee faithfulness: a convincing rollout can still move the wrong arm or lose the manipulated object. To ensure the prediction respects each arm's commanded path, we inject per-arm $\mathrm{SE}(3)$ transformations into attention via \textbf{PRoPE-style geometric encoding}, preserving arm identity and rigid-motion structure. Action control alone does not fully constrain scene geometry or the evolution of small manipulated objects. We therefore add a lightweight \textbf{depth branch} for scene-level geometry and use \textbf{SAM3 masks} with a frozen \textbf{V-JEPA teacher} to maintain object consistency throughout grasping. We further distill the multi-step generator into a few-step student via distribution-matching distillation for efficient deployment. At the time of writing, \model{} achieves first place on Track~1 and second place on Track~2 of the WorldArena~2.0 Challenge. Our model and code will be publicly available.
Chinese Translation
我们提出了 extbf{DreamX-Phi 1.0}，这是一种用于机器人操作的动作条件视频世界模型。该模型在给定观察帧、语言指令和包含末端执行器姿态及夹持器状态的预定动作序列的情况下，预测未来的观察结果。然而，仅仅依靠现实主义并不能保证预测的准确性：一个令人信服的展开仍然可能移动错误的手臂或丢失被操作的物体。为了确保预测遵循每个手臂的指令路径，我们通过 extbf{PRoPE风格几何编码}将每个手臂的$ ext{SE}(3)$变换注入到注意力机制中，以保持手臂的身份和刚体运动结构。单靠动作控制并不能完全约束场景几何或小型被操作物体的演变。因此，我们增加了一个轻量级的 extbf{深度分支}来处理场景级几何，并使用 extbf{SAM3掩码}与冻结的 extbf{V-JEPA教师}结合，以在抓取过程中保持物体的一致性。此外，我们通过分布匹配蒸馏将多步生成器提炼为少步学生，以实现高效部署。在撰写本文时， extmodel{}在WorldArena~2.0挑战赛的Track~1中获得第一名，在Track~2中获得第二名。我们的模型和代码将公开发布。
cs.CV / 84 / 2608.13495
TraVEL: Trajectory-Guided Video Embedding Learning for Driving-Video Retrieval
TraVEL：基于轨迹引导的视频嵌入学习用于驾驶视频检索
Abstract
Efficiently retrieving relevant clips from large-scale driving logs is essential for data curation, model development, and safety analysis. Structured and rule-based retrieval systems can explicitly target driving events, but typically require expert-defined rules, auxiliary data, and multi-stage perception pipelines. Multimodal embedding models offer a simpler and more efficient alternative by representing each video with a single searchable vector. However, general-purpose models often rely on shortcuts from static scene context and struggle to distinguish motion-centric events, such as turning left versus right or accelerating versus decelerating. In this work, we study how to adapt a general-purpose multimodal embedding model to driving-video retrieval. We first fine-tune Qwen3-VL-Embedding on paired clips and reasoning traces from nuReasoning using an InfoNCE objective. While this stage substantially improves overall retrieval, caption supervision alone remains insufficient for fine-grained motion understanding. We therefore introduce TraVEL (Trajectory-Guided Video Embedding Learning), a motion-aware fine-tuning framework that uses ego-trajectory similarity as a reward within Group Relative Policy Optimization. Trajectories serve only as privileged training supervision; retrieval still operates on single-vector video embeddings without ego poses, expert rules, or auxiliary perception outputs. We further construct a driving-video retrieval benchmark from nuReasoning. Experiments show that TraVEL improves motion-centric retrieval across model scales: relative to SFT, it raises longitudinal and lateral mAP by 9.8 and 4.7 points at 2B, with corresponding gains of 7.2 and 1.5 points at 8B. TraVEL thus combines physically grounded supervision with efficient embedding-based search.
Chinese Translation
从大规模驾驶日志中高效检索相关片段对于数据整理、模型开发和安全分析至关重要。结构化和基于规则的检索系统可以明确针对驾驶事件，但通常需要专家定义的规则、辅助数据和多阶段感知管道。多模态嵌入模型通过用单个可搜索向量表示每个视频，提供了一种更简单和高效的替代方案。然而，通用模型往往依赖于静态场景上下文的捷径，难以区分以运动为中心的事件，例如左转与右转或加速与减速。在本研究中，我们探讨如何将通用多模态嵌入模型适应于驾驶视频检索。我们首先在来自nuReasoning的配对片段和推理轨迹上微调Qwen3-VL-Embedding，使用InfoNCE目标。尽管这一阶段显著提高了整体检索性能，但仅依靠字幕监督仍不足以实现细粒度的运动理解。因此，我们引入了TraVEL（基于轨迹引导的视频嵌入学习），这是一种运动感知的微调框架，利用自我轨迹相似性作为Group Relative Policy Optimization中的奖励。轨迹仅作为特权训练监督；检索仍然在没有自我姿态、专家规则或辅助感知输出的情况下操作于单向量视频嵌入。我们进一步从nuReasoning构建了一个驾驶视频检索基准。实验表明，TraVEL在不同模型规模下改善了以运动为中心的检索：相较于SFT，在2B时提高了纵向和横向mAP分别为9.8和4.7点，在8B时相应提高了7.2和1.5点。因此，TraVEL将物理基础的监督与高效的基于嵌入的搜索相结合。
cs.CV / 85 / 2608.13502
GS$^{2}$CI: Robust Gaussian Splatting For Snapshot Compressive Imaging via Large Vision Model Priors
GS$^{2}$CI：基于大规模视觉模型先验的稳健高斯溅射快照压缩成像
Abstract
Snapshot Compressive Imaging (SCI) offers an efficient solution for high-speed video acquisition and, under exposure-time camera--scene relative motion, multi-view scene capture by compressing temporal or spatial information into a single 2D measurement. While recent studies have explored SCI for 3D scene reconstruction, existing methods struggle with significant challenges due to information loss, limited viewpoint diversity, and the computational burden of jointly optimizing 3D representations and camera poses. In this work, we propose a novel framework that reconstructs high-quality 3D scenes from a single SCI measurement by leveraging 3D Gaussian Splatting (3DGS) and the powerful priors of large-scale vision foundation models (VFMs). Our primary reconstruction combines measurement-derived 3D VFM initialization with SCI-aware Gaussian optimization. After coarse-stage convergence, an auxiliary 2D VFM provides pseudo-view supervision at synthesized viewpoints for local appearance refinement. To further address the instability caused by ambiguous SCI supervision during 3DGS optimization, we introduce Opacity-Guided Splitting and Growth Regulation (OSGR), an SCI-specific densification strategy that augments split candidates using local opacity statistics, discourages loss-compensating opacity inflation through mean-opacity regulation, and bounds representation growth with explicit candidate-ratio and Gaussian-count constraints. Extensive experiments across multiple benchmarks demonstrate that our method achieves the strongest overall performance, combining leading reconstruction quality and robustness to viewpoint variation with competitive computational efficiency.
Chinese Translation
快照压缩成像（SCI）为高速视频采集提供了一种高效的解决方案，并在曝光时间相机与场景的相对运动下，通过将时间或空间信息压缩为单个2D测量，实现多视角场景捕捉。尽管近期研究探讨了SCI在3D场景重建中的应用，但现有方法由于信息丢失、视角多样性有限以及联合优化3D表示和相机姿态的计算负担，面临重大挑战。在本研究中，我们提出了一种新颖的框架，通过利用3D高斯溅射（3DGS）和大规模视觉基础模型（VFM）的强大先验，从单个SCI测量中重建高质量的3D场景。我们的主要重建方法结合了基于测量的3D VFM初始化与SCI感知的高斯优化。在粗略阶段收敛后，辅助的2D VFM在合成视点提供伪视图监督，以进行局部外观细化。为了进一步解决在3DGS优化过程中由模糊SCI监督引起的不稳定性，我们引入了不透明度引导的分裂与生长调节（OSGR），这是一种特定于SCI的密集化策略，通过使用局部不透明度统计增强分裂候选，抑制通过均值不透明度调节造成的损失补偿性不透明度膨胀，并通过显式候选比例和高斯数量约束限制表示增长。多个基准的广泛实验表明，我们的方法在整体性能上表现最强，结合了领先的重建质量和对视角变化的稳健性，同时具有竞争力的计算效率。
cs.CV / 86 / 2608.13513
TabSOM: A tabular-to-image encoding method based on self-organizing maps
TabSOM：基于自组织映射的表格到图像编码方法
Abstract
Tabular-to-image methods have emerged as novel approaches to leverage the high predictive performance of convolutional neural networks and vision transformers. They convert tabular data into image representations, mapping each feature at a fixed pixel location derived from a dimensionality-reduction method (e.g., t-SNE, UMAP, PCA). However, they encode only the marginal value of each feature and discard information about feature relationships. We propose TabSOM, a tabular-to-image encoding built on the Self-Organizing Map (SOM), which provides: (i) a spatial layout in which every input feature occupies a fixed canvas position derived from its component plane via collision-free Hungarian assignment; and (ii) a graph that captures pairwise feature relationships derived from the SOM component planes. The resulting image stacks two multi-scale node channels: one encodes feature values at fixed scales, while the other encodes pairwise feature interactions as spatial connections between related features. Two SOM-derived interpretability approaches are introduced: a prototype-inspired partial dependence plot and a class--separation importance score. Benchmarked against twelve existing tabular-to-image methods across public binary-classification datasets, TabSOM ranks first or second on every dataset and achieves the lowest variance of any method evaluated. Interpretability obtained with TabSOM was validated against Random Forest, XGBoost, and SHAP, the class-separation score shows reasonable agreement with established baselines on the top-ranked features while capturing complementary structural information from input data. These results demonstrate that TabSOM provides an effective and interpretable approach for applying deep learning architectures to tabular data, bridging the performance--interpretability gap in this domain.
Chinese Translation
表格到图像的方法作为新兴的技术，旨在利用卷积神经网络和视觉变换器的高预测性能。这些方法将表格数据转换为图像表示，将每个特征映射到从降维方法（如 t-SNE、UMAP、PCA）派生的固定像素位置。然而，这些方法仅编码每个特征的边际值，忽略了特征之间的关系信息。我们提出了 TabSOM，一种基于自组织映射（Self-Organizing Map, SOM）的表格到图像编码方法，提供了：(i) 一个空间布局，其中每个输入特征占据一个固定的画布位置，该位置通过无碰撞的匈牙利分配法从其组件平面得出；(ii) 一个图形，捕捉从 SOM 组件平面得出的成对特征关系。生成的图像堆叠了两个多尺度节点通道：一个编码固定尺度下的特征值，另一个编码作为相关特征之间空间连接的成对特征交互。引入了两种基于 SOM 的可解释性方法：一种受原型启发的部分依赖图和一种类别分离重要性得分。在针对十二种现有表格到图像方法的公共二分类数据集进行基准测试时，TabSOM 在每个数据集上均排名第一或第二，并且实现了所有评估方法中最低的方差。通过与随机森林（Random Forest）、XGBoost 和 SHAP 进行验证，TabSOM 获得的可解释性显示，类别分离得分与顶级特征的既定基线有合理的一致性，同时捕捉到输入数据的互补结构信息。这些结果表明，TabSOM 为将深度学习架构应用于表格数据提供了一种有效且可解释的方法，弥合了该领域的性能与可解释性之间的差距。
cs.CV / 87 / 2608.13541
SCULPT: Subtractive Composition for 3D Part Generation
SCULPT：用于3D部件生成的减法组合
Abstract
Part-aware 3D generation aims to create digital assets that are coherent as complete objects while exposing structural parts for editing, material assignment, animation, and reuse. Existing methods impose this structure outside the native generation loop: segmentation-based methods partition an already generated shape, while additive methods synthesize parts from predefined layouts, boxes, or tokens and then reconcile them into a whole. The former preserves the generated geometry but fixes the object before part boundaries are determined; the latter exposes part cardinality but often leaves shared boundaries vulnerable to gaps, interpenetrations, and material discontinuities. In this paper, we propose SCULPT, a framework that addresses these challenges through subtractive composition. Given a complete object represented in a structured 3D latent space, SCULPT iteratively applies a joint split predictor to generate one extracted part together with the remaining object. The predictor performs a coupled denoising process conditioned on both the image and the current 3D state, so the extracted part and updated remainder are generated together rather than reconciled after generation. The joint split predictor processes both outputs on the union of their native sparse 3D supports, allowing neighboring supports to overlap rather than imposing a disjoint voxel partition. The rollout ends when the remainder support becomes empty or reaches a fixed safety cap, allowing the number of generated parts to adapt to each object within that bound. Extensive experiments demonstrate state-of-the-art geometry on PartObjaverse while preserving strong complete-object reconstruction after part assembly. Results on four dataset images, one text-to-image-generated input, and one real-world photograph further show fine-grained textured part decomposition beyond the benchmark.
Chinese Translation
部件感知的3D生成旨在创建连贯的数字资产，作为完整对象，同时暴露结构部件以便于编辑、材料分配、动画和重用。现有方法在本地生成循环之外施加这种结构：基于分割的方法对已生成的形状进行划分，而加法方法则从预定义的布局、盒子或标记中合成部件，然后将它们整合为一个整体。前者保留了生成的几何形状，但在确定部件边界之前就固定了对象；后者暴露了部件的基数，但往往使共享边界容易出现间隙、相互穿透和材料不连续性。在本文中，我们提出了SCULPT，一个通过减法组合来解决这些挑战的框架。给定一个在结构化3D潜在空间中表示的完整对象，SCULPT迭代地应用联合分割预测器来生成一个提取的部件以及剩余对象。该预测器执行一个耦合去噪过程，条件是图像和当前3D状态，因此提取的部件和更新的剩余部分是一起生成的，而不是在生成后进行整合。联合分割预测器在它们本地稀疏3D支持的并集上处理两个输出，允许相邻支持重叠，而不是强制施加不相交的体素划分。展开过程在剩余支持变为空或达到固定安全上限时结束，从而允许生成的部件数量适应每个对象在该范围内的变化。大量实验表明，在PartObjaverse上实现了最先进的几何形状，同时在部件组装后保持了强大的完整对象重建。对四个数据集图像、一个文本到图像生成的输入和一张真实世界照片的结果进一步展示了超越基准的细粒度纹理部件分解。
cs.CV / 88 / 2608.13546
Alaya-EVOKE: From Linear-Scaling Supervision to Endless World
Alaya-EVOKE：从线性扩展监督到无尽世界
Abstract
Interactive world models must support persistent memory, responsive interaction, and long-horizon generation, yet these requirements place conflicting demands on the model. Maintaining history in the denoiser context or key-value cache incurs growing cost, forcing a trade-off between session length and retained memory, while low-latency interaction relies on few-step generation whose capabilities are bounded by its teacher. Evoke addresses both limitations by externalizing persistent world state and redesigning the teacher for long-horizon interactive generation. Scene geometry is maintained in an external, camera-indexed world state bank, from which only view-relevant information is retrieved, keeping the denoiser context bounded as the session grows. Rather than treating the teacher as a fixed generator, we design it for long-horizon supervision: its sparse attention combines chunk-wise grouping, retrieval of selected distant frames, and a linear-attention global state, yielding linear growth in memory and compute while enabling supervision over long horizons. Such supervision exposes content drift that stays locally plausible within short windows, while per-chunk conditioning enables prompt changes and event control throughout the sequence. A 30-second distribution-matching objective, applied under self-forced rollouts, transfers both capabilities to a three-step student that uses no classifier-free guidance, improving resistance to long-term drift while preserving responsive conditioning. With bounded context and recurrent external memory, Evoke supports open-ended, continuously evolving generation; on a single H200 at $384\times 640$, each $1.5\,\mathrm{s}$ chunk is generated in $2.11\,\mathrm{s}$. As a three-step world model, Evoke achieves state-of-the-art performance on WBench while remaining competitive on VBench-Long and VBench-2.0.
Chinese Translation
交互式世界模型必须支持持久记忆、响应式交互和长时间生成，但这些要求对模型提出了相互矛盾的需求。在去噪器上下文或键值缓存中维护历史会产生不断增长的成本，迫使会话长度与保留记忆之间进行权衡，而低延迟交互依赖于少步生成，其能力受到教师的限制。Evoke通过外部化持久世界状态并重新设计教师以实现长时间交互生成，解决了这两种限制。场景几何在一个外部的、相机索引的世界状态库中维护，仅从中检索与视图相关的信息，随着会话的增长保持去噪器上下文的界限。我们并不将教师视为固定生成器，而是为长时间监督进行设计：其稀疏注意力结合了块状分组、选定远程帧的检索和线性注意力全局状态，实现了内存和计算的线性增长，同时使长时间监督成为可能。这种监督暴露了在短时间窗口内保持局部合理的内容漂移，而每块条件化使得在整个序列中能够迅速变化和事件控制。一个30秒的分布匹配目标，在自我强制展开下应用，将这两种能力转移到一个不使用无分类器引导的三步学生模型上，提高了对长期漂移的抵抗力，同时保持了响应性条件化。通过有界上下文和递归外部记忆，Evoke支持开放式、持续演变的生成；在单个H200上，以$384 imes 640$的分辨率，每个$1.5 ext{s}$的块生成时间为$2.11 ext{s}$。作为一个三步世界模型，Evoke在WBench上实现了最先进的性能，同时在VBench-Long和VBench-2.0上保持竞争力。
cs.CV / 89 / 2608.13552
PlayWorld: Benchmarking World Models with Agent Players over Long-Horizon Objectives
PlayWorld：基于代理玩家的世界模型长远目标基准测试
Abstract
Video world models simulate future states conditioned on current observations and user actions. Recent systems have demonstrated impressive video consistency and action controllability over long sequences. However, fairly comparing these interactive models remains challenging. In practice, a human player typically evaluates a world model by pursuing long-horizon objectives through interaction. For example, a user may turn around 360 degrees to see whether the environment remains consistent, or walk into the water and inspect whether realistic water ripples are generated. The action sequence required to achieve the same objective may vary substantially between models, making fixed action-conditioned evaluation unsuitable for cross-model comparison. To address this, we employ multi-modal Agent Players to interact with world models toward specified long-horizon objectives. Building on this paradigm, we introduce PlayWorld, a benchmark providing 171 scenarios, each with a specified objective. To evaluate performance thoroughly, we assess models along four core dimensions: geometry consistency, interaction fidelity, out-of-sight evolution, and insight evolution. In addition, we incorporate basic ability metrics for video quality and controllability. Experiments across nine state-of-the-art world models reveal that current models remain unreliable on long-horizon interactive objectives, particularly in maintaining spatial consistency and persistent state evolution. Code and data are available at https://github.com/kxding/PlayWorld.
Chinese Translation
视频世界模型根据当前观察和用户行为模拟未来状态。近期系统在长序列上展示了令人印象深刻的视频一致性和动作可控性。然而，公平比较这些交互模型仍然具有挑战性。在实践中，人类玩家通常通过交互追求长远目标来评估世界模型。例如，用户可能会旋转360度以查看环境是否保持一致，或走入水中检查是否生成了逼真的水波。实现相同目标所需的动作序列在不同模型之间可能有很大差异，使得固定的动作条件评估不适合跨模型比较。为了解决这个问题，我们采用多模态代理玩家与世界模型进行交互，以实现指定的长远目标。在此基础上，我们引入了PlayWorld，一个提供171个场景的基准，每个场景都有一个指定的目标。为了全面评估性能，我们从四个核心维度进行模型评估：几何一致性、交互保真度、视野外演变和洞察演变。此外，我们还纳入了视频质量和可控性的基本能力指标。对九个最先进的世界模型的实验表明，当前模型在长远交互目标上仍然不可靠，特别是在保持空间一致性和持续状态演变方面。代码和数据可在 https://github.com/kxding/PlayWorld 获取。
cs.CV / 90 / 2608.13556
V-RAE: Rethinking Video Latent Spaces for Generation
V-RAE：重新思考视频潜在空间以进行生成
Abstract
Latent video generation relies on autoencoders to define a compact space in which generative models operate. Although video autoencoder architectures have evolved substantially, their latent spaces are still optimized primarily for pixel-level reconstruction and provide limited high-level semantic organization. A reconstruction-optimal latent space, however, need not be well suited to generative modeling. We propose V-RAE, a video representation autoencoder that builds compact generative latents on top of frozen vision foundation model representations. A lightweight temporal pooling module removes temporal redundancy while preserving semantic structure, and a video decoder reconstructs continuous motion from the compressed features. We evaluate V-RAE with four representative frozen encoders on video reconstruction, semantic probing, and class-conditional generation. V-RAE achieves 2.13 rFVD on K600, outperforming all evaluated large-scale pretrained video VAEs. Its latents retain substantially more semantic information than conventional video tokenizer latents. Under matched generation settings, our best variant achieves gFVD scores of 117.86 and 19.16 on UCF101 and K600, respectively, while converging up to 6x faster}. We further show that reconstruction quality alone is insufficient to characterize generative utility and introduce tFVD, a temporal-coherence diagnostic that correlates more reliably with downstream generation quality. Beyond video generation, V-RAE also improves future video prediction on Cityscapes over the Wan 2.2 VAE latent space under matched prediction settings. Taken together, the experiments show that frozen semantic representations can support video reconstruction, generation, and predictive modeling. The project page: https://v-rae.github.io/.
Chinese Translation
潜在视频生成依赖于自编码器来定义一个紧凑的空间，在该空间中生成模型进行操作。尽管视频自编码器架构已经有了显著的发展，但其潜在空间仍主要针对像素级重建进行优化，提供的高层次语义组织有限。然而，重建最优的潜在空间并不一定适合生成建模。我们提出了V-RAE，一种视频表示自编码器，它在冻结的视觉基础模型表示之上构建紧凑的生成潜在空间。一个轻量级的时间池化模块去除了时间冗余，同时保留了语义结构，而视频解码器则从压缩特征中重建连续运动。我们使用四个代表性的冻结编码器对V-RAE在视频重建、语义探测和类别条件生成上进行了评估。V-RAE在K600上达到了2.13的rFVD，超越了所有评估的大规模预训练视频VAE。其潜在空间保留了比传统视频标记器潜在空间更多的语义信息。在匹配生成设置下，我们的最佳变体在UCF101和K600上分别达到了117.86和19.16的gFVD分数，同时收敛速度快达6倍。我们进一步表明，仅靠重建质量不足以表征生成效用，并引入了tFVD，一种时间一致性诊断工具，与下游生成质量的相关性更为可靠。除了视频生成，V-RAE在匹配预测设置下还改善了Cityscapes上的未来视频预测，相较于Wan 2.2 VAE潜在空间。综合来看，实验表明冻结的语义表示可以支持视频重建、生成和预测建模。项目页面：https://v-rae.github.io/
cs.CV / 91 / 2608.13560
AutoDesign: Meta-Harness Optimization for Long-Horizon Agentic Design
AutoDesign：面向长时程自主设计的元工具优化
Abstract
Transforming multimodal sources into condensed and structured media outputs can be fundamentally conceptualized as a long-horizon agentic process centered on a model-harness system. While an ideal harness system should align with human design priors and accumulate reusable experience through empirical exploration to drive recursive self-improvement, existing paradigms remain static and fall short of this capability. In this paper, we present AutoDesign, a framework that aligns with human design priors, where a meta-harness optimizer guides a code agent to recursively improve harness based on rollout feedback. To instantiate and evaluate this framework, we focus on the academic paper-to-poster generation task and introduce PosterBench, comprising a 100-paper Main Track spanning five disciplines and PosterBench-mini, a shared 10-paper subset for controlled evaluation. On the PosterBench Main Track, AutoDesign achieves the highest score of 78.32, surpassing the closed-source commercial system Claude Design by 7.45 points. Across seven controlled code-agent-model configurations, integrating the learned DesignHarness consistently improves performance, increasing the average PosterBench Score from 54.99 to 67.39 (+12.4%). In a fully autonomous long-horizon loop, it executes 253 tool calls and 11 editing turns within 40 minutes for under $3, reaching average conference-poster quality in human evaluation. A system-blind human study further demonstrates that AutoDesign achieves the highest human preference among evaluated systems.
Chinese Translation
将多模态源转化为凝练且结构化的媒体输出可以从根本上被概念化为一个以模型-工具系统为中心的长时程自主过程。理想的工具系统应与人类设计先验相一致，并通过经验探索积累可重用的经验，以推动递归自我改进，而现有范式仍然停滞不前，未能实现这一能力。本文提出了AutoDesign，一个与人类设计先验相一致的框架，其中元工具优化器引导代码代理基于回滚反馈递归改进工具。为了实例化和评估该框架，我们聚焦于学术论文到海报生成任务，并引入PosterBench，包括涵盖五个学科的100篇论文的主轨道和一个用于控制评估的10篇论文的共享子集PosterBench-mini。在PosterBench主轨道上，AutoDesign获得了78.32的最高分，超过了闭源商业系统Claude Design 7.45分。在七种受控的代码代理模型配置中，整合学习到的DesignHarness始终提高了性能，将平均PosterBench得分从54.99提升至67.39（+12.4%）。在一个完全自主的长时程循环中，它在40分钟内执行了253次工具调用和11次编辑，成本低于3美元，达到了人类评估中的平均会议海报质量。一项系统盲评的人类研究进一步表明，AutoDesign在评估系统中获得了最高的人类偏好。
cs.AI / 1 / 2608.12325
Position: Reasoning is a Learnable Rule-Based Process
立场：推理是一个可学习的基于规则的过程
Abstract
Autonomous reasoning is among the most scientifically and economically motivating topics in AI today. Historically the purview of symbolic AI, recent advances have mainly emerged from deep probabilistic generative models. Despite immense interest and rapid progress, the generative AI community has not clearly converged on operational definitions for reasoning and often implicitly rejects the historical treatment of this topic in logic and verifiable automated reasoning. This position contends that definitional ambiguity leaves the construct validity of reasoning evaluation unverifiable, undermining quantifiable progress toward trustworthy autonomous reasoning. We also contend that this ambiguity is addressable. To that end, we provide (1) operational definitions based on a synthesis of the literature, positioning valid and sound reasoning as a learnable rule-based process; and (2) a checklist for best practices in the communication of AI reasoning research.
Chinese Translation
自主推理是当今人工智能中最具科学和经济激励的主题之一。历史上，这一领域主要属于符号人工智能，最近的进展主要源于深度概率生成模型。尽管受到极大关注并取得快速进展，生成式人工智能社区在推理的操作性定义上并未明确达成共识，并且常常隐含地拒绝了逻辑和可验证自动推理中对这一主题的历史处理。本文立场认为，定义模糊性使得推理评估的构建有效性无法验证，从而削弱了向可信的自主推理的可量化进展。我们还认为，这种模糊性是可以解决的。为此，我们提供了（1）基于文献综合的操作性定义，将有效且可靠的推理定位为一个可学习的基于规则的过程；（2）一份关于人工智能推理研究沟通的最佳实践清单。
cs.AI / 2 / 2608.12345
Diagnostic Foundation for Evaluating LLMs' Research Integrity as Co-Scientists
评估大型语言模型作为共同科学家的研究诚信的诊断基础
Abstract
Language models are increasingly deployed as co-scientists, yet their ability to uphold research integrity under institutional pressure remains unmeasured. We introduce IntegrityBench, a benchmark evaluating misconduct classification, ethical action reasoning and artifact-grounded decision making across 36 paired tasks under a 5-level implicit-explicit pressure protocol spanning 3 domains and 4 research stages. Evaluating 18 frontier model variants, we find that under peak pressure, models fail roughly 1 in 3 integrity-critical decisions, and neither scale nor reasoning ability reliably mitigates this. Explicit pressures induce compliance with misconduct, while implicit contextual reframing more often causes over-refusal of legitimate research tasks. Interestingly, models failing to classify research requests accurately perform equally or better on artifact-grounded decision making (85.7 vs. 79.4), suggesting the three facets are structurally dissociated and correct ethical action does not require accurate classification. Frontier models can thus appear helpful while harbouring integrity failures that create two distinct deployment risks: facilitating research misconduct and eroding trust in AI-assisted research.
Chinese Translation
语言模型越来越多地被作为共同科学家使用，但它们在机构压力下维持研究诚信的能力尚未得到测量。我们引入了IntegrityBench，这是一个基准，评估在3个领域和4个研究阶段下，36个配对任务中不当行为分类、伦理行动推理和基于文献的决策制定，采用5级隐性-显性压力协议。评估18个前沿模型变体，我们发现，在最高压力下，模型在大约1/3的与诚信相关的决策中失败，而规模或推理能力并未可靠地减轻这一问题。显性压力会导致对不当行为的顺从，而隐性情境重构则更常导致对合法研究任务的过度拒绝。有趣的是，未能准确分类研究请求的模型在基于文献的决策制定上表现相当或更好（85.7 vs. 79.4），这表明这三个方面在结构上是分离的，正确的伦理行动并不需要准确的分类。因此，前沿模型可能看似有帮助，但却潜藏着诚信失败，造成两种不同的部署风险：促进研究不当行为和侵蚀对人工智能辅助研究的信任。
cs.AI / 3 / 2608.12346
Position: The Alignment Community is Unintentionally Building a Censor's Toolkit
立场：对齐社区无意中构建了审查者的工具包
Abstract
This position paper argues that modern AI alignment methods - originally designed to prevent harmful output - are dual-use technologies that may easily be misused by malicious actors for censorship and manipulation. By mapping current alignment techniques to the possibility and actual cases of misuse, we show that the quest for a "perfectly aligned" model inadvertently also provides malicious actors with an ever-improving tool for informational dominance. We need to discuss this dual-use potential now, as its risk is exacerbated by rapid user adoption of AI as information provider, economic power asymmetries, and a political landscape that increasingly shifts towards authoritarianism. We conclude by urging the community to consider the intentional misuse of AI alignment mechanisms and propose mitigation strategies to safeguard against this dual-use potential.
Chinese Translation
本文立场论文认为，现代人工智能对齐方法最初旨在防止有害输出，但实际上是双重用途技术，容易被恶意行为者滥用用于审查和操控。通过将当前对齐技术与滥用的可能性和实际案例进行映射，我们展示了对“完美对齐”模型的追求无意中也为恶意行为者提供了一个不断完善的信息主导工具。我们需要现在就讨论这种双重用途的潜力，因为其风险因用户快速采用人工智能作为信息提供者、经济权力不对称以及日益向威权主义转变的政治格局而加剧。最后，我们呼吁社区考虑人工智能对齐机制的故意滥用，并提出缓解策略以防范这种双重用途的潜力。
cs.AI / 4 / 2608.12368
Agreement Is Not Alignment: Divergent Moral Grounds in Human and LLM Ethical Judgments
一致性并非对齐：人类与大型语言模型（LLMs）伦理判断中的道德基础差异
Abstract
Agreement with human judgments is a common proxy for evaluating the alignment of large language models (LLMs). Yet agreement in final labels does not show that human annotators and models rely on the same moral grounds. Two agents may reach the same judgment while appealing to different principles, contextual assumptions, or interpretations of the situation. We test this distinction using a curated 500-item ETHICS-derived benchmark spanning five domains of moral judgment, with new human annotator and LLM annotations of both final labels and supporting rationales. Across frontier and open model families, agreement with human annotator majority labels is often high. However, rationale-level analysis reveals systematic divergence in the moral grounds expressed by human annotators and models. In particular, models redistribute attention across categories such as harm, respect, promise-keeping, justice, desert, and excuse relevance, even when their final labels match the human annotator majority. Our results show that agreement should not be treated as equivalent to alignment. Label-based evaluation can therefore be misleadingly reassuring unless complemented by analysis of the reasons, principles, and moral priorities expressed in model judgments.
Chinese Translation
与人类判断的一致性通常被视为评估大型语言模型（LLMs）对齐程度的常用代理。然而，最终标签的一致性并不能证明人类标注者和模型依赖于相同的道德基础。两个代理可能在不同的原则、背景假设或对情境的解释下达成相同的判断。我们利用一个经过精心策划的500项ETHICS衍生基准，涵盖五个道德判断领域，测试这一区别，并对人类标注者和LLM的最终标签及支持理由进行了新的标注。在前沿和开放模型系列中，与人类标注者多数标签的一致性通常较高。然而，理由层面的分析揭示了人类标注者和模型在表达的道德基础上存在系统性差异。特别是，即使最终标签与人类标注者多数一致，模型在伤害、尊重、履行承诺、公正、应得和借口相关性等类别之间重新分配了注意力。我们的结果表明，一致性不应被视为等同于对齐。因此，仅依赖标签的评估可能会误导性地令人感到安心，除非辅之以对模型判断中所表达的理由、原则和道德优先级的分析。
cs.AI / 5 / 2608.12371
Multi-Agent Scheduling with LLM-Assisted Contract Net Negotiation for Stream Processing in Mobile Edge Computing
基于大型语言模型辅助的合同网协商的多智能体调度在移动边缘计算中的流处理
Abstract
Stream-processing systems increasingly operate across heterogeneous mobile edge--cloud infrastructures, where workload volatility, resource contention, and stringent quality-of-service (QoS) requirements complicate decentralized scheduling. This paper proposes \emph{MAS-DecStream}, whose main contribution is \emph{LLM-MR-CNP}: an extension of the classical Contract Net Protocol with semantic CFP formulation, progressive context disclosure, multi-round proposal revision, negotiation memory, and deterministic validation. Edge-cluster agents refine natural-language offloading proposals from local observations, predicted resource states, and qualitative runtime context, while hard resource and QoS constraints remain deterministic. Experiments derived from the Alibaba ASI Trace evaluate the extension at three levels: single- versus multi-round CNP, rule-based versus LLM-assisted refinement, and fixed-model single- versus multi-round negotiation. Under the evaluated configurations, MAS-DecStream reduces latency violations to 3\%, eliminates resource overcommitment, reaches a conflict-resolution rate of 0.91 with 20 agents, and improves utility by up to 22\% over the multi-round rule-based baseline. A separate 25-case evaluation shows model- and prompt-dependent accuracy--cost trade-offs. The results provide initial evidence that multi-round CNP refinement is the principal protocol-level gain, with LLM assistance adding value for qualitative and uncertain runtime context.
Chinese Translation
流处理系统越来越多地在异构的移动边缘-云基础设施中运行，其中工作负载波动、资源争用和严格的服务质量（QoS）要求使得去中心化调度变得复杂。本文提出了 extit{MAS-DecStream}，其主要贡献是 extit{LLM-MR-CNP}：一种扩展经典合同网协议的方案，包含语义化的 CFP 公式、渐进的上下文披露、多轮提案修订、协商记忆和确定性验证。边缘集群代理根据本地观察、预测的资源状态和定性的运行时上下文来细化自然语言卸载提案，同时硬性资源和 QoS 约束保持确定性。基于阿里巴巴 ASI Trace 的实验在三个层面上评估了该扩展：单轮与多轮 CNP、基于规则与 LLM 辅助的细化，以及固定模型的单轮与多轮协商。在评估的配置下，MAS-DecStream 将延迟违规率降低到 3\%，消除了资源过度承诺，达到了 20 个代理的冲突解决率为 0.91，并在多轮基于规则的基准上提高了效用高达 22\%。另外的 25 个案例评估显示了模型和提示依赖的准确性与成本之间的权衡。结果提供了初步证据，表明多轮 CNP 的细化是主要的协议层面收益，而 LLM 的辅助为定性和不确定的运行时上下文增添了价值。
cs.AI / 6 / 2608.12372
Position: We Need Practical AI Alignment Methods to Mirror Human Reasoning
立场：我们需要实用的人工智能对齐方法以反映人类推理
Abstract
AI systems are increasingly employed as decision aids, decision delegates, or autonomous decision-makers. This position paper argues that in many settings, particularly high-stakes decision-making, we need accurate cognitively-aligned AI systems that reason similarly to their users, and faithfully communicate their reasoning. We review evidence that cognitive alignment improves understandability and trustworthiness, and provide new survey data showing that many users find cognitive alignment "essential" when an AI's rationale for a judgment or action is important to them. We outline the gaps between existing alignment methods and what is needed to achieve cognitive alignment, and present a research agenda to address these gaps. We argue that cognitive misalignment represents a likely impediment to AI adoption in many envisioned applications, and that addressing it is important for creating AI systems on which users are both willing and justified to rely.
Chinese Translation
人工智能系统越来越多地被用作决策辅助工具、决策代理或自主决策者。本文认为，在许多场景中，特别是高风险决策中，我们需要准确的认知对齐人工智能系统，这些系统能够与用户进行类似的推理，并忠实地传达其推理过程。我们回顾了认知对齐改善可理解性和可信度的证据，并提供了新的调查数据，显示许多用户在人工智能的判断或行动的理由对他们重要时，认为认知对齐是“必不可少的”。我们概述了现有对齐方法与实现认知对齐所需之间的差距，并提出了一个研究议程以解决这些差距。我们认为，认知不对齐可能成为许多设想应用中人工智能采用的障碍，解决这一问题对于创建用户既愿意又有理由依赖的人工智能系统至关重要。
cs.AI / 7 / 2608.12373
Don't Want Your LLM to Recommend Nuclear Strike? Try Asking It in Japanese
不想让你的大型语言模型推荐核打击？试试用日语问它
Abstract
Large language models are increasingly used in strategic and advisory contexts, yet their safety alignment is typically evaluated in English only. We test nine models from six providers and ask whether the language of a prompt can change a model's decision in a high-stakes scenario. We use single-turn game-theoretic vignettes in which a model advises a nuclear-armed nation on whether to strike a defenseless opponent. The prompt is intentionally amoral and strategically identical across languages. We find that Japanese prompts reduce launch rates in the Claude model family: Claude Sonnet 4.6 drops from 40% to 0% in scenarios where the strike is unnecessary and from 93% to 17% in contested scenarios, with minimal effect when the strike is strategically rational. The effect extends to Gemini Pro 3.1 (53% to 13%). A cross-language experiment isolates the mechanism: when instructed to reason in Japanese in an English prompt, launch rates drop from 93% to 37%. It is the language the model is asked to reason in, not the language of the input, that drives the effect. When reasoning in Japanese, models spontaneously generate moral vocabulary (''moral cost'', ''millions of lives'') that is entirely absent from the prompt. Five other models show no language effect, but they launch in nearly every condition regardless of language. The effect requires a model that already hesitates in English. These results show that LLM safety behavior is language-dependent, and that evaluating in English alone can miss both risks and safeguards encoded in other languages.
Chinese Translation
大型语言模型在战略和咨询领域的应用日益增多，但其安全性对齐通常仅在英语环境中进行评估。我们测试了来自六个提供者的九个模型，并询问提示的语言是否会在高风险场景中改变模型的决策。我们使用单轮博弈论小插曲，其中模型建议一个拥有核武器的国家是否对一个无防御的对手发动攻击。提示在道德上是故意中立的，并在不同语言中具有战略上的相同性。我们发现，日语提示降低了Claude模型系列的发射率：在不必要的攻击场景中，Claude Sonnet 4.6的发射率从40%降至0%，在有争议的场景中从93%降至17%，而在战略上合理的攻击情况下影响最小。该效应扩展到Gemini Pro 3.1（从53%降至13%）。跨语言实验隔离了机制：当在英语提示中被指示用日语推理时，发射率从93%降至37%。影响的驱动因素是模型被要求用何种语言进行推理，而不是输入的语言。当用日语推理时，模型自发生成了完全缺失于提示中的道德词汇（“道德成本”，“数百万生命”）。另外五个模型没有显示出语言效应，但在几乎所有条件下无论语言如何都进行发射。该效应需要一个在英语中已经犹豫的模型。这些结果表明，LLM的安全行为依赖于语言，仅用英语进行评估可能会遗漏其他语言中编码的风险和保护措施。
cs.AI / 8 / 2608.12385
Dual-Flow Transformers: Decoupling the Primary Prefill Path from Additional Decode Computation
双流变换器：将主要预填充路径与额外解码计算解耦
Abstract
As large language models serve more requests, cumulative inference cost is becoming increasingly important relative to one-time training cost. The two inference phases stress hardware differently: prompt prefill is parallel and typically compute-bound, whereas autoregressive decode is sequential and often memory-bandwidth-bound. Conventional width or depth scaling increases both costs together because every added layer is evaluated in both phases. We ask whether additional learned computation can instead be allocated to continuation prediction while preserving the prompt-wide primary computation and a single persistent key-value (KV) cache. We introduce the Dual-Flow Transformer. Its primary flow is a complete causal language model that processes the prompt and writes the KV cache. The auxiliary flow is omitted during prompt processing and activated only from the final prompt position onward, adding continuation-prediction computation without writing persistent state or influencing the primary flow. The two flows share major attention, MLP, and output matrices, while using separate token embeddings and lightweight coupling. Sharing weights and the primary cache also creates opportunities to reuse loaded weights and cached keys and values during grouped execution. Across matched-token comparisons, Dual-Flow achieves lower validation loss across architectures and data configurations. In MoE models, the separation makes primary and auxiliary expert fan-outs independent controls over prompt cost, continuation cost, and predictive quality. We study two regimes: increasing decode computation at fixed prefill expert computation, and reallocating a fixed decode expert budget between the two flows. These experiments expose a prefill-decode-quality trade-off and demonstrate the potential of phase-specific expert allocation.
Chinese Translation
随着大型语言模型处理的请求数量增加，累积推理成本相对于一次性训练成本变得越来越重要。这两个推理阶段对硬件的压力不同：提示预填充是并行的，通常受计算限制，而自回归解码是顺序的，往往受内存带宽限制。传统的宽度或深度扩展会同时增加这两种成本，因为每增加一层都需要在两个阶段进行评估。我们探讨是否可以将额外的学习计算分配给续写预测，同时保留提示范围内的主要计算和一个单一的持久键值（KV）缓存。我们提出了双流变换器。其主要流是一个完整的因果语言模型，处理提示并写入KV缓存。辅助流在提示处理期间被省略，仅在最终提示位置之后激活，增加续写预测计算，而不写入持久状态或影响主要流。这两个流共享主要的注意力、MLP和输出矩阵，同时使用独立的标记嵌入和轻量级耦合。共享权重和主要缓存还创造了在分组执行期间重用加载权重和缓存键值的机会。在匹配标记比较中，双流变换器在不同架构和数据配置下实现了更低的验证损失。在MoE模型中，分离使得主要和辅助专家的扩展成为独立的控制，影响提示成本、续写成本和预测质量。我们研究了两种模式：在固定的预填充专家计算下增加解码计算，以及在两个流之间重新分配固定的解码专家预算。这些实验揭示了预填充-解码质量的权衡，并展示了阶段特定专家分配的潜力。
cs.AI / 9 / 2608.12389
Learning to Adapt Cross-Domain Preferences via Meta-LoRA for LLM Personalization
通过 Meta-LoRA 学习适应跨领域偏好以实现大语言模型个性化
Abstract
Cross-domain zero- or few-shot personalization aims to generate user-preferred responses in unseen conversational domains from only a handful of target-domain interactions. Existing adaptation methods struggle to calibrate update magnitude under sparse evidence and thus overfit, whereas history-transfer methods often entangle user preferences with source-domain artifacts, yielding unreliable personalization priors and negative transfer. To calibrate adaptation to evidence quality, we propose PAC-Bayes-regularized Meta-LoRA, which uses a meta-learned LoRA initialization as both the adaptation start and prior center, while adjusting update strength according to support-set size and predictive uncertainty. This limits overfitting under sparse or ambiguous evidence while permitting stronger personalization as evidence grows. Controlled adaptation alone does not determine which preferences should transfer across domains or how they should be expressed. We therefore functionally decompose personalization priors into user and domain components, using a human-readable prompt for stable preferences and topology-preserving soft tokens for domain-specific hidden-space conditioning. Experiments across multiple benchmarks and personalization tasks show consistent gains over strong baselines. On HiCUPID, our method reduces cross-domain win-rate degradation by 47.9% relative to the best competing baseline and improves win rate by 110.2% under unseen-user cold start.
Chinese Translation
跨领域零样本或少样本个性化旨在仅通过少量目标领域的交互，在未见过的对话领域中生成用户偏好的响应。现有的适应方法在稀疏证据下难以校准更新幅度，因此容易过拟合，而历史迁移方法则常常将用户偏好与源领域的伪影纠缠在一起，导致不可靠的个性化先验和负迁移。为了根据证据质量校准适应性，我们提出了 PAC-Bayes 正则化的 Meta-LoRA，该方法使用元学习的 LoRA 初始化作为适应的起始点和先验中心，同时根据支持集大小和预测不确定性调整更新强度。这在稀疏或模糊证据下限制了过拟合，同时随着证据的增加允许更强的个性化。仅仅控制适应性并不能决定哪些偏好应该跨领域迁移或如何表达这些偏好。因此，我们将个性化先验功能性地分解为用户和领域组件，使用人类可读的提示来稳定偏好，并使用保持拓扑结构的软令牌进行领域特定的隐空间调节。在多个基准和个性化任务上的实验显示，相较于强基线，我们的方法在性能上持续提升。在 HiCUPID 上，我们的方法相较于最佳竞争基线减少了 47.9% 的跨领域胜率下降，并在未见用户冷启动下提高了 110.2% 的胜率。
cs.AI / 10 / 2608.12395
Research Assistant: AstraZeneca's Agentic System for R&D
研究助手：阿斯利康的研发代理系统
Abstract
We describe Research Assistant, an internal LLM-based system developed at AstraZeneca to help scientists and clinicians explore biomedical questions across a broad range of data sources. The system provides a chat-style interface that brings together evidence from scientific literature, knowledge graphs, chemistry, clinical trials, safety resources, expression data, and internal experimental systems. It supports both a fast mode for direct question answering and a multi-step mode for more complex research tasks. Responses are grounded in retrieved evidence and linked back to the original sources, allowing users to review and further explore the underlying data. In this technical note, we outline the system architecture, the main design choices behind the product, and lessons learned from deploying it at scale to support day-to-day R&D workflows across AstraZeneca.
Chinese Translation
我们描述了研究助手，这是阿斯利康开发的一个基于大型语言模型（LLM）的内部系统，旨在帮助科学家和临床医生在广泛的数据源中探索生物医学问题。该系统提供了一个聊天式界面，将科学文献、知识图谱、化学、临床试验、安全资源、表达数据和内部实验系统中的证据整合在一起。它支持快速模式以进行直接问答，以及多步骤模式以处理更复杂的研究任务。响应基于检索到的证据，并链接回原始来源，使用户能够审查并进一步探索基础数据。在本技术说明中，我们概述了系统架构、产品背后的主要设计选择，以及在大规模部署中获得的经验教训，以支持阿斯利康日常研发工作流程。
cs.AI / 11 / 2608.12426
Large Language Models Can Follow Instructions, But Not Many at Once: Phase Transitions in Compositional Constraint Satisfaction
大型语言模型可以遵循指令，但不能同时处理多个指令：组合约束满足中的相变
Abstract
Large language models are increasingly deployed in settings that require simultaneous adherence to multiple explicit constraints - reasoning structure, safety boundaries, output schemas. Individual constraints are handled proficiently, but the compositional regime, where many must hold jointly, remains poorly characterized: how rapidly does performance degrade, what governs the degradation, and can the collapse be mitigated? We introduce Constraint Saturation Evaluation (CSE), a procedurally generated benchmark that systematically varies the number of simultaneous constraints (k), with every constraint scored by a deterministic, rule-based verifier and zero LLM-judge involvement: 15 models, 36 constraint types, 369,753 checks at k=1-12. Three findings emerge. First, per-constraint pass rate decays gradually and predictably, while the chance of satisfying all k constraints collapses - a model passing individual constraints at ~41% at k=8 succeeds on all eight just 5.7% of the time. Second, constraints do not degrade equally: structural constraints lose 2x more baseline capability per added constraint than lexical ones, ordered by a comprehension-maintenance gap that separates constraints requiring sustained tracking from binary decisions immune to composition. Third, failures are nearly independent, which is what makes the accumulation multiplicative; the residual coupling that does exist tracks shared output features rather than pairwise interference - a wrong sentence count fails every constraint that reads it. Reliable instruction following breaks down beyond 5-6 simultaneous constraints: probe-level success falls below 50% at 7 constraints for the strongest model, and at 3 or fewer for 12 of 15.
Chinese Translation
大型语言模型越来越多地应用于需要同时遵循多个明确约束的场景——推理结构、安全边界、输出模式。单个约束能够被有效处理，但在组合状态下，多个约束必须共同满足的情况仍然缺乏清晰的特征：性能下降的速度如何，下降的原因是什么，是否可以减轻这种崩溃？我们引入了约束饱和评估（Constraint Saturation Evaluation, CSE），这是一个程序生成的基准测试，系统性地改变同时约束的数量（k），每个约束由确定性的规则基础验证器评分，且没有大型语言模型（LLM）评判的参与：15个模型，36种约束类型，在k=1-12时进行369,753次检查。我们得出了三个发现。首先，单个约束的通过率逐渐且可预测地下降，而满足所有k个约束的机会则急剧崩溃——在k=8时，模型在单个约束上的通过率约为41%，但在所有八个约束上成功的概率仅为5.7%。其次，约束的降级并不均等：结构约束在每增加一个约束时，其基线能力损失是词汇约束的两倍，按理解维持差距排序，这一差距将需要持续跟踪的约束与对组合免疫的二元决策区分开来。第三，失败几乎是独立的，这使得积累效应呈乘法特性；存在的残余耦合跟踪共享的输出特征，而不是成对干扰——一个错误的句子计数会使所有读取它的约束失败。可靠的指令遵循在超过5-6个同时约束时崩溃：在7个约束时，最强模型的探测级成功率降至50%以下，而在15个模型中，有12个模型在3个或更少约束时表现不佳。
cs.AI / 12 / 2608.12428
MindMemOS: A Portable and Self-Evolving Memory Operating Layer for AI Agents
MindMemOS：一种便携式自我演化的记忆操作层用于人工智能代理
Abstract
Memory is a core component of AI agents, enabling them to accumulate experience, maintain personalization, and adapt over long-term interactions. However, existing memory systems often remain fixed after development, limiting their ability to adapt their memory models, organization strategies, and procedural knowledge through continued use. We present MindMemOS, a portable and self-evolving memory operating layer that organizes open-world information using a unified entity property timestructure. MindMemOS supports scenario-adaptive memory modeling, higher-order pattern discovery, autonomous memory refinement, and continuous skill evolution. Its MindMemEvolve algorithm employs validation-driven evolutionary search to optimize memory schemas for target scenarios, whiledreaming consolidates accumulated memories by merging redundant records and resolving conflicts. In addition, implicit corrective feedback serves as a human-in-the-loop signal for identifying and revising potentially inaccurate or misaligned memories. Its MindSkillEvolve algorithm further transforms agent execution trajectories into reusable and progressively refined skills. MindMemOS achieves 94.03% accuracy on LOCOMO and 70.63% on PersonaMem. MindSkillEvolve improves SpreadsheetBench success by 9.2 percentage points over the initial-skill baseline.
Chinese Translation
记忆是人工智能代理的核心组成部分，使其能够积累经验、保持个性化并在长期交互中适应。然而，现有的记忆系统在开发后往往保持固定，限制了它们通过持续使用来适应记忆模型、组织策略和程序知识的能力。我们提出了MindMemOS，这是一种便携式自我演化的记忆操作层，使用统一的实体属性时间结构来组织开放世界信息。MindMemOS支持场景自适应记忆建模、高阶模式发现、自主记忆精炼和持续技能演化。其MindMemEvolve算法采用基于验证的进化搜索来优化针对特定场景的记忆模式，而dreaming通过合并冗余记录和解决冲突来巩固积累的记忆。此外，隐式纠正反馈作为人机协作信号，用于识别和修正潜在的不准确或不一致的记忆。其MindSkillEvolve算法进一步将代理执行轨迹转化为可重用和逐步精炼的技能。MindMemOS在LOCOMO上实现了94.03%的准确率，在PersonaMem上实现了70.63%的准确率。MindSkillEvolve在SpreadsheetBench上的成功率比初始技能基线提高了9.2个百分点。
cs.AI / 13 / 2608.12476
Governed Persistent Memory: Source-Bound State Semantics and Fail-Closed Release for Long-Horizon Agents
受控持久内存：源绑定状态语义与故障闭合释放用于长时间跨度的智能体
Abstract
Long-term agent memory is usually treated as select--store--retrieve, but retrieval does not decide whether contradictory, superseded, retracted, deleted, or stale records may support an outgoing claim. We introduce Governed Persistent Memory (GPM), an auditable bitemporal state-transition model with source-bound admission, derived lifecycle state, current public barriers, and fail-closed structured release. Five executable clauses cover ledger integrity, source binding, conflict isolation, non-revival after retraction or deletion, and exact claim closure over a fresh view at one verified head. On a prespecified hash-frozen 3,600-case GPM-ReleaseBench, GPM matches all complete outcomes; the strongest of three intentionally simple complete policies matches 1,800/3,600 and makes unmatched releases on 50% of violation cases. A separate sealed end-to-end service evaluation exercises real ingestion and release across eight query families. In its publicly disclosed V3 arm, the governed lane is correct on 2,400/2,400 clusters versus 600/2,400 for ungoverned local Qwen2.5-7B; it repairs all 1,800 baseline failures with no regression (one-sided 95% lower bounds 99.875% and 99.834%). A later V5 reseal over Chinese- and English-command arms, with generation-date pinning and no post-freeze reducer amendment, again obtains 2,400/2,400 per arm. A production-code-independent finite model explores 331,776 semantic and 1,990,656 query states without a full-contract counterexample, and a 100,000-trace three-engine differential yields zero mismatches. These are bounded contract and implementation results, not open-world model accuracy or evidence of world truth. Governed answers in the sealed service evaluation are deterministic service outputs; the 7B result is the ungoverned comparison, not a claim that a language model itself became perfectly accurate.
Chinese Translation
长期智能体内存通常被视为选择-存储-检索，但检索并不能决定矛盾、被取代、撤回、删除或过时的记录是否可以支持一个外发声明。我们提出了受控持久内存（Governed Persistent Memory, GPM），这是一种可审计的双时态状态转换模型，具有源绑定的准入、派生的生命周期状态、当前的公共障碍和故障闭合的结构化释放。五个可执行条款涵盖了账本完整性、源绑定、冲突隔离、撤回或删除后的非复活，以及在一个经过验证的头部上的新视图下的精确声明闭合。在一个预先指定的哈希冻结的3600案例GPM-ReleaseBench上，GPM匹配所有完整结果；三种故意简单的完整策略中最强的匹配了1800/3600，并在50%的违规案例中产生了不匹配的释放。一个单独的密封端到端服务评估在八个查询家族中进行了真实的摄取和释放。在其公开披露的V3版本中，受控路径在2400/2400个集群上是正确的，而未受控的本地Qwen2.5-7B则为600/2400；它修复了所有1800个基线失败且没有回归（单侧95%的下限为99.875%和99.834%）。后来的V5重新密封覆盖了中文和英文命令的版本，具有生成日期固定且没有后冻结减少修订，再次在每个版本上获得2400/2400的结果。一个与生产代码无关的有限模型探索了331,776个语义状态和1,990,656个查询状态，没有出现完整合约的反例，而一个100,000轨迹的三引擎差异分析则产生了零不匹配。这些是有界的合约和实施结果，而不是开放世界模型的准确性或世界真相的证据。受控答案在密封服务评估中是确定性的服务输出；7B结果是未受控的比较，而不是声称语言模型本身变得完全准确。
cs.AI / 14 / 2608.12522
$\varepsilon$-MemEvo: Adaptive Cross-Task Memory Transfer for LLM Program Evolution
$ ext{ε}$-MemEvo：用于大规模语言模型程序演化的自适应跨任务记忆转移
Abstract
LLM-based program evolution systems such as FunSearch and AlphaEvolve have shown strong ability to discover novel algorithms, but typically optimize each task in isolation, discarding search experience after completion. We introduce $\varepsilon$-MemEvo, a framework for cross-task knowledge transfer in LLM program evolution. $\varepsilon$-MemEvo stores prior experience as task-agnostic tactic memories: compact natural-language summaries of successful algorithmic strategies rather than raw code, enabling transfer across tasks with different APIs and evaluators. To avoid negative transfer from semantically mismatched memories, $\varepsilon$-MemEvo uses an adaptive injection gate that decides whether retrieved memories should be injected, and at what intensity. We evaluate $\varepsilon$-MemEvo on 8 diverse optimization benchmarks spanning mathematical optimization and systems engineering, using a content-level Leave-One-Out protocol that excludes target-task memory entries. On the primary GPT-5 backbone, $\varepsilon$-MemEvo improves AUCC over AdaEvolve on all 8 tasks, with a mean relative gain of +8.7%, and improves early-stage convergence by +9.4% on average. Ablations show that naive memory injection can fail catastrophically, while adaptive gating remains safe across all five ablation tasks. The data-updated posterior is interpretable in observed states: it favors skip during improving search and shifts from skip to hint across early and late plateaus. These gains incur less than 1% computational overhead.
Chinese Translation
基于大规模语言模型（LLM）的程序演化系统，如 FunSearch 和 AlphaEvolve，展示了发现新算法的强大能力，但通常在孤立的情况下优化每个任务，完成后丢弃搜索经验。我们提出了 $ ext{ε}$-MemEvo，一个用于 LLM 程序演化中的跨任务知识转移框架。$ ext{ε}$-MemEvo 将先前的经验存储为任务无关的策略记忆：成功算法策略的紧凑自然语言摘要，而不是原始代码，从而实现跨不同 API 和评估器的任务转移。为了避免来自语义不匹配记忆的负迁移，$ ext{ε}$-MemEvo 使用自适应注入门控来决定是否注入检索到的记忆，以及注入的强度。我们在涵盖数学优化和系统工程的 8 个多样化优化基准上评估了 $ ext{ε}$-MemEvo，使用内容级别的留一法协议，排除了目标任务的记忆条目。在主要的 GPT-5 主干上，$ ext{ε}$-MemEvo 在所有 8 个任务上相较于 AdaEvolve 提高了 AUCC，平均相对增益为 +8.7%，并在早期阶段的收敛性上平均提高了 +9.4%。消融实验表明，简单的记忆注入可能会导致灾难性失败，而自适应门控在所有五个消融任务中保持安全。数据更新后的后验在观察状态中是可解释的：它在改进搜索期间偏向跳过，并在早期和晚期平台之间从跳过转变为提示。这些增益的计算开销低于 1%。
cs.AI / 15 / 2608.12555
CAS: A Causal Attribution Score for Local and Global Explainable Artificial Intelligence
CAS：一种用于局部和全局可解释人工智能的因果归因评分
Abstract
Predictive explanation methods attribute a model output; they do not, by themselves, attribute an intervention effect on the real-world outcome. We introduce the Causal Attribution Score (CAS), a compact score architecture for causal explanation. CAS starts from an identified interventional coalition game, allocates the joint intervention contrast with causal Shapley contributions, and converts those raw outcome-scale effects into Local CAS, Signed Local CAS, and two complementary Global CAS summaries. The innovation is not a new Shapley formula, but a local-to-global causal reporting layer with an explicit intervention target. In the known-truth benchmark, eight repeated primary-interaction simulations (n = 2,200 each, three actions) gave mean Local CAS MAE of 0.107 for coalition-aware CAS, compared with 0.173 for one-at-a-time normalisation and 0.213 for a global normalised absolute ATE vector. The paired advantage over one-at-a-time normalisation increased from -0.003 under additivity to 0.091 under strong interactions. On both empirical DoubleML datasets, 401(k) eligibility/net financial assets (n = 9,915) and Pennsylvania reemployment bonus/unemployment duration (n = 5,099), predictive SHAP/TreeSHAP rankings differed materially from Feature-CAS rankings of treatment-effect modifiers. In Pennsylvania, dep1 (exactly one dependent) moved from predictive global rank 13 to Feature-CAS rank 2 and was the leading local Feature-CAS modifier. These results isolate the added value of separating what predicts the outcome from what explains heterogeneity in an estimated causal effect.
Chinese Translation
预测解释方法将模型输出归因；但它们本身并不归因于对现实世界结果的干预效应。我们引入了因果归因评分（Causal Attribution Score, CAS），这是一种用于因果解释的紧凑评分架构。CAS 从一个已识别的干预联盟博弈开始，分配与因果 Shapley 贡献相关的联合干预对比，并将这些原始结果尺度效应转换为局部 CAS、带符号的局部 CAS 和两个互补的全局 CAS 摘要。创新之处不在于新的 Shapley 公式，而是具有明确干预目标的局部到全局因果报告层。在已知真相的基准测试中，八次重复的主要交互模拟（每次 n = 2,200，三种行动）显示，关注联盟的 CAS 的平均局部 CAS 平均绝对误差（MAE）为 0.107，而逐一归一化为 0.173，全球归一化绝对平均处理效应（ATE）向量为 0.213。在加法性下，逐一归一化的配对优势从 -0.003 增加到强交互下的 0.091。在两个经验 DoubleML 数据集上，401(k) 资格/净金融资产（n = 9,915）和宾夕法尼亚州再就业奖金/失业持续时间（n = 5,099），预测 SHAP/TreeSHAP 排名与处理效应修饰因子的 Feature-CAS 排名有显著不同。在宾夕法尼亚州，dep1（恰好一个受抚养人）从预测全局排名第 13 移动到 Feature-CAS 排名第 2，并成为领先的局部 Feature-CAS 修饰因子。这些结果突显了将预测结果与解释估计因果效应异质性的因素分开的附加价值。
cs.AI / 16 / 2608.12574
Trie Automata for Constrained Decoding over Large Finite Sets
用于大有限集合约束解码的Trie自动机
Abstract
Large language models increasingly need to generate structured outputs that conform to predefined schemas, with one common constraint being selection from a finite set of valid strings. Current constrained decoding systems handle this through general-purpose grammar compilation, which becomes prohibitively slow as the number of valid values grows into the thousands, a cardinality wall. We introduce the trie automaton, a specialized mechanism that exploits finite-set structure (shared prefixes, bounded depth, known cardinality) via Aho-Corasick multi-pattern matching to precompute per-node token masks. The trie achieves 7X faster per-step valid-token computation (0.65 us vs. 5.8 us) compared to XGrammar, one of the primary backends in vLLM and SGLang, and 2--6.5X faster compilation at K >= 300. Because precomputed masks enable a stateless serving path that bypasses the guided decoding pipeline, this advantage compounds in batch serving: end-to-end vLLM throughput reaches 219 req/s vs. XGrammar's 7.5 req/s at batch size 256 (29X). The 29X combines the algorithmic speedup with integration-path savings that only precomputed masks can unlock. Across seven tokenizer families (32K--262K vocabulary), the trie maintains sub-100ms compilation up to K = 10,000 and flat per-step cost regardless of set size, while guaranteeing 100% output validity.
Chinese Translation
大型语言模型越来越需要生成符合预定义模式的结构化输出，其中一个常见的约束是从有限的有效字符串集合中进行选择。目前的约束解码系统通过通用语法编译来处理这一问题，但随着有效值数量增长到数千，编译速度变得极其缓慢，形成了基数壁垒。我们引入了Trie自动机，这是一种专门的机制，通过Aho-Corasick多模式匹配利用有限集合结构（共享前缀、有限深度、已知基数）来预计算每个节点的令牌掩码。与vLLM和SGLang中的主要后端之一XGrammar相比，Trie在每步有效令牌计算上实现了7倍的速度提升（0.65微秒对比5.8微秒），在K >= 300时编译速度提升为2到6.5倍。由于预计算的掩码使得无状态服务路径能够绕过引导解码管道，这一优势在批量服务中得以叠加：在批量大小为256时，端到端vLLM的吞吐量达到219请求/秒，而XGrammar为7.5请求/秒（29倍）。这个29倍的提升结合了算法加速和只有预计算掩码才能解锁的集成路径节省。在七个分词器系列（词汇量32K到262K）中，Trie在K = 10,000时保持编译时间低于100毫秒，并且每步成本与集合大小无关，同时保证100%的输出有效性。
cs.AI / 17 / 2608.12585
Reasoning Jury: Multi-Model Consensus for Evaluating Reasoning Traces
推理陪审团：评估推理轨迹的多模型共识
Abstract
Improving reasoning LLMs requires the ability to judge the quality of long reasoning traces for effective reasoning data curation, strong training signals during reinforcement learning, and an in-depth understanding of reasoning behaviors during model performance evaluation. Additionally, surfacing reasoning mistakes that the model makes would enable improving the model's performance at runtime through providing feedback. Due to the difficulty of this complex task on long reasoning traces, single-model judges (even frontier models) do not do well at identifying reasoning defects. Additionally, leveraging frontier models during online training of reasoning LLMs is generally prohibited due to guardrails in terms of use. In this work, we introduce Reasoning Jury, a system that replaces the single judge with a jury of LLMs and a moderated consensus mechanism, to improve the fidelity of judgments for identifying reasoning defects. In reasoning jury, defects of a reasoning trace and their severity are surfaced through a deliberation where a moderator conducts a discussion amongst the jury where the jurors critique each other's judgments and get to modify their initial votes. The moderator derives a consensus through deliberation amongst jurors or consolidation of judgements. We show that Reasoning Jury with a jury of open-weight models (e.g., gpt-oss-120b) is able to significantly outperform frontier models (opus-4.6, sonnet-4.6, and gemini-3.1-pro) at correctly identifying reasoning defects. Besides accuracy performance improvements, the aggregated cost of the jury (initial verdicts, deliberations, consolidation, etc.) is a fraction (8 to 15%) of the cost of running frontier models in LLM-as-a-judge setup. We also show how these judgements can be leveraged to understand failure modes of reasoning LLMs on benchmarks, which allows much deeper understanding of a model's performance.
Chinese Translation
提升推理大型语言模型（LLMs）的能力需要判断长推理轨迹的质量，以便有效进行推理数据的整理、在强化学习过程中提供强有力的训练信号，以及在模型性能评估中深入理解推理行为。此外，揭示模型在推理过程中所犯的错误将有助于通过反馈在运行时提升模型的性能。由于在长推理轨迹上执行这一复杂任务的困难，单一模型评判者（即使是前沿模型）在识别推理缺陷方面表现不佳。此外，由于使用条款的限制，在线训练推理 LLMs 时通常禁止使用前沿模型。在本研究中，我们提出了推理陪审团（Reasoning Jury），一个通过陪审团的 LLMs 和一个调解共识机制来替代单一评判者的系统，以提高识别推理缺陷的判断准确性。在推理陪审团中，推理轨迹的缺陷及其严重性通过调解员主持的讨论过程被揭示，陪审团成员相互批评各自的判断并得以修改初始投票。调解员通过陪审团成员之间的讨论或判断的整合来得出共识。我们展示了使用开放权重模型（如 gpt-oss-120b）的推理陪审团能够显著优于前沿模型（opus-4.6、sonnet-4.6 和 gemini-3.1-pro）在正确识别推理缺陷方面的表现。除了准确性表现的提升外，陪审团的聚合成本（初步裁决、讨论、整合等）仅为在 LLM 作为评判者设置中运行前沿模型成本的一个小部分（8% 到 15%）。我们还展示了如何利用这些判断来理解推理 LLMs 在基准测试中的失败模式，从而更深入地理解模型的性能。
cs.AI / 18 / 2608.12590
Auditable agentic AI for evidence-grounded thyroid ultrasound diagnosis and reporting
可审计的自主智能AI用于基于证据的甲状腺超声诊断与报告
Abstract
Thyroid ultrasound diagnosis requires coordinated lesion localization, measurement, risk stratification and reporting, yet most AI systems address these tasks in isolation and provide limited support for clinical review. We present ThyroidXAgent, a clinician-interactive agentic AI system that coordinates specialized diagnostic tools and stores their outputs as an auditable case-level evidence record. The system was developed using OpenThyroidDB, a multicentre, multitask resource integrating approximately 0.3 million ultrasound images and 24,000 paired reports, and was evaluated on 28,458 non-overlapping test cases, including 8,721 cases from 35 centres in the private NHC-MISD-TUS cohort. Across heterogeneous datasets, ThyroidXAgent achieved a mean Dice score of 87.21 percent for nodule segmentation and a mean AUROC of 0.9466 for benign-malignant classification. The same workflow supported lymph-node metastasis prediction and follicular versus papillary thyroid carcinoma classification, with AUROCs of 0.864 and 0.805, respectively. For report generation, evidence-grounded assembly outperformed multimodal language-model baselines across three cohorts. ThyClinScore, a lesion-level clinical semantic metric introduced here, showed the strongest correlation with a location-aware language-model judge. ThyroidXAgent improved physician classification accuracy, increased report diagnostic consistency from 70.3 percent to 86.2 percent, and reduced segmentation and reporting time by 35.9 percent and 27.4 percent, respectively. These findings support auditable, clinician-correctable agentic AI for thyroid ultrasound diagnosis and reporting.
Chinese Translation
甲状腺超声诊断需要协调病变定位、测量、风险分层和报告，然而大多数AI系统孤立地处理这些任务，并对临床审查提供有限支持。我们提出了ThyroidXAgent，一种临床互动的自主智能AI系统，协调专业诊断工具并将其输出存储为可审计的案例级证据记录。该系统使用OpenThyroidDB开发，这是一个多中心、多任务资源，整合了大约30万张超声图像和24,000份配对报告，并在28,458个不重叠的测试案例上进行了评估，其中包括来自35个中心的8,721个案例，属于私立NHC-MISD-TUS队列。在异质数据集中，ThyroidXAgent在结节分割方面达到了87.21%的平均Dice系数，在良恶性分类方面达到了0.9466的平均AUROC。相同的工作流程支持淋巴结转移预测和滤泡型与乳头状甲状腺癌分类，AUROC分别为0.864和0.805。在报告生成方面，基于证据的汇总在三个队列中优于多模态语言模型基线。这里引入的ThyClinScore，一种病变级临床语义指标，与位置感知语言模型评审者的相关性最强。ThyroidXAgent提高了医生的分类准确性，将报告的诊断一致性从70.3%提高到86.2%，并分别减少了35.9%和27.4%的分割和报告时间。这些发现支持可审计的、可由临床医生纠正的自主智能AI用于甲状腺超声诊断与报告。
cs.AI / 19 / 2608.12593
DiG-bench: Discovery in Games
DiG-bench：游戏中的发现
Abstract
Discovery---formulating novel generalizations---is a central part of the scientific process. Despite its importance, there is a gap in the current AI benchmark landscape, with few benchmarks directly probing the capacity for discovering new knowledge with experimentation in controlled environments where the objective is unknown. To address this gap, we release a new benchmark: DiG-bench (Discovery in Games). DiG-bench consists of a set of 70 independent games. Each game is encoded as a short string and has unique transformation rules that must be discovered through interaction and experimentation. The levels of the game present a series of challenges to test whether the rules have been discovered, where the win conditions for each level are also unknown. We provide games at seven tiers of difficulty for AI agents. The lowest tier is routinely solvable by multiple models, while the highest tier challenges the best models in agentic harnesses. All 70 games were solved by at least one human on first attempt. A subset of 21 games is released publicly, and the remainder is held private for secure evaluation.
Chinese Translation
发现——制定新颖的概括——是科学过程中的核心部分。尽管其重要性不言而喻，但当前人工智能基准测试领域存在一个空白，鲜有基准直接探讨在目标未知的受控环境中通过实验发现新知识的能力。为了解决这一问题，我们发布了一个新的基准：DiG-bench（游戏中的发现）。DiG-bench由70个独立游戏组成。每个游戏都被编码为一个简短的字符串，并具有独特的变换规则，这些规则必须通过互动和实验来发现。游戏的关卡呈现出一系列挑战，以测试规则是否已被发现，而每个关卡的胜利条件也是未知的。我们为人工智能代理提供了七个难度等级的游戏。最低等级的游戏通常可以被多个模型解决，而最高等级则挑战最优秀的模型在代理控制下的表现。所有70个游戏在首次尝试时均被至少一名人类解决。我们公开发布了21个游戏的子集，其余游戏则保持私密以便进行安全评估。
cs.AI / 20 / 2608.12599
Dead text or binding clause? Measuring and restoring constraint influence in black-box LLM dialogues
死文本还是约束条款？测量和恢复黑箱大型语言模型对话中的约束影响
Abstract
Multi-turn dialogues let users revoke constraints as easily as impose them, but revocation does not reliably take effect: models keep enacting withdrawn requirements (occasionally beneath comments asserting their removal), a failure we call \emph{behavioral relapse}, or revocation inertia. No existing instrument measures this influence per clause, predicts it before delivery, or repairs it under matched budgets. \sysname{} closes the three gaps through the model API alone: a contract ledger pairs every constraint with an executable checker, records revocations as tombstones, and compiles the net constraint state ahead of time into a single specification; a sequential ablation probe measures per-clause adherence and incremental behavioral effect; a repair ladder operates under token- and attempt-matched budgets. On \dataname{} (\NTasks{} HumanEval tasks, \NClauses{} verified checkers), relapse at an 8B operating point climbs from \ScaleDelayedMTwo{} to \ScaleDelayedMEight{} as constraint load grows, while stronger models sit at floor. Under matched checkers, model, and budget, ahead-of-time compilation significantly reduces relapse against a no-ledger verifier-retry baseline (\RestoreDiff{}, 95\% CI \RestoreDiffCI{}, $p$ \RestoreDiffP{}); adaptive ladder interventions stacked on top add no detectable gain (95\% confidence excludes gains $\geq$ \LadderExcludedGain{}). The probe predicts relapse before delivery (AUROC \AurocPrimary{}); a one-sentence tombstone note recovers about a third of the compilation effect and survives a placebo control. At \CostDeliveryFactor{} delivery overhead and \CostTotalHedged{} of API compute for every result, revocation failure becomes a measurable, predictable, and repairable property of dialogue state rather than an invisible one.
Chinese Translation
多轮对话使用户能够轻松撤销约束，但撤销并不总是可靠生效：模型仍会执行已撤回的要求（有时在声称已移除的评论下），我们称这种失败为 extit{行为复发}或撤销惯性。现有工具无法逐条测量这种影响，无法在交付前预测，也无法在匹配预算下进行修复。 extit{sysname}通过模型API独自弥补这三个缺口：一个合同账本将每个约束与可执行检查器配对，将撤销记录为墓碑，并提前将净约束状态编译为单一规范；一个顺序消融探针测量逐条遵循和增量行为效果；一个修复阶梯在匹配的令牌和尝试预算下运行。在 extit{dataname}（ extit{NTasks}个HumanEval任务， extit{NClauses}个已验证检查器）中，在8B操作点下，随着约束负载的增加，复发率从 extit{ScaleDelayedMTwo}上升到 extit{ScaleDelayedMEight}，而更强的模型则保持在底线。在匹配的检查器、模型和预算下，提前编译显著降低了复发率，相较于无账本的验证重试基线（ extit{RestoreDiff}，95 ext{CI} extit{RestoreDiffCI}，$p$ extit{RestoreDiffP}）；在此基础上叠加的自适应阶梯干预没有带来可检测的增益（95 ext{confidence}排除了增益$ extit{≥LadderExcludedGain}$）。该探针在交付前预测复发（AUROC extit{AurocPrimary}）；一句墓碑备注恢复了约三分之一的编译效果，并在安慰剂对照中存活。在 extit{CostDeliveryFactor}的交付开销和每个结果的 extit{CostTotalHedged} API计算下，撤销失败成为对话状态的可测量、可预测和可修复的属性，而不是一个不可见的属性。
cs.AI / 21 / 2608.12610
@skills: Attention is all you have
@skills：关注点就是你所拥有的一切
Abstract
There are 56,804 public agent skills today, and teams write many more privately. The dominant delivery model is installation: once installed, a skill's description remains in the system prompt, competing for fewer than 100 reliable trigger slots. This leaves the long tail with no practical path to use and forces teams' own playbooks to compete for the same scarce space. We observe that installation bundles three separable functions: content, persistence, and automatic triggering. Only the last requires prompt residency. We therefore propose @skills, an open protocol that separates them. A path addresses any skill, subtree, or collection, and reading a skill is sufficient to use it, so nothing is installed or made resident. The operation vendors a copy at the same path into a project's Git-tracked tree for adaptation and ownership. The operation adds one .gitignore-style line, the only element that costs prompt residency. A directory is a menu, making bundles ordinary directories rather than all-or-nothing units. The protocol requires no manifest, lockfile, or registration, and SKILL.md remains unchanged. @skills is additive, ships as an installable package, and turns any agent that can read files and run commands into a client through a single instruction file. Its open specification is at https://github.com/SylphAI-Inc/atskills and it is implemented in the AdaL CLI at https://adalagent.ai . Because paths address skills well but cannot find them, the protocol is paired with a free hub at https://atskills.one for corpus-wide search and ranking, repository-free hosting, private and team collections, and one-screen authoring. The hub is optional: gh: and local paths resolve without it, and indexed GitHub skills retain their gh: identities. Install less, use more.
Chinese Translation
目前有 56,804 种公共代理技能，团队还私下编写了更多技能。主流的交付模型是安装：一旦安装，技能的描述将保留在系统提示中，与不到 100 个可靠的触发槽竞争。这使得长尾技能没有实际使用路径，并迫使团队自己的操作手册争夺同样稀缺的空间。我们观察到，安装捆绑了三个可分离的功能：内容、持久性和自动触发。只有最后一个功能需要保留在提示中。因此，我们提出了 @skills，一个开放协议，将这些功能分开。路径可以指向任何技能、子树或集合，读取技能就足以使用它，因此没有任何东西被安装或保留。该操作在项目的 Git 跟踪树中以相同路径提供一个副本，以便进行适应和所有权。该操作添加了一行类似 .gitignore 的内容，这是唯一一个需要提示驻留的元素。目录相当于菜单，使得捆绑成为普通目录，而不是全有或全无的单元。该协议不需要清单、锁定文件或注册，SKILL.md 保持不变。@skills 是可扩展的，以可安装包的形式发布，并通过单个指令文件将任何能够读取文件和运行命令的代理转变为客户端。其开放规范可在 https://github.com/SylphAI-Inc/atskills 查阅，并在 AdaL CLI 中实现，网址为 https://adalagent.ai。由于路径能够很好地指向技能但无法找到它们，该协议与一个免费的中心 https://atskills.one 配对，用于全语料库的搜索和排名、无仓库托管、私人和团队集合，以及一屏式创作。该中心是可选的：gh: 和本地路径在没有它的情况下也能解析，索引的 GitHub 技能保留其 gh: 身份。安装更少，使用更多。
cs.AI / 22 / 2608.12645
Jagged Judges: Epistemic Stability Under Silence, Pressure, and Persistence
锯齿状的评判者：在沉默、压力和持续性下的认知稳定性
Abstract
LLM judges have become central infrastructure for model evaluations, online grading, and reward modeling. Judges are typically validated by accuracy on golden data, but accuracy says little about whether they are stable under re-prompting, challenge, or sustained pushback. We introduce the \emph{Wiggle Framework}, a unified stress test for epistemic stability in LLM judges. The framework decomposes judge robustness along three dimensions: Mechanical Consistency (stability under re-prompting and reframing), Single-turn Conviction (stability under a single challenge), and Multi-turn Persistence (stability under sustained or adaptive pressure). We use the framework to study 9 frontier models across 14 judging tasks spanning safety, toxicity, AI writing detection, and political-response evaluation. Every model exhibits substantial wiggle as a judge --- flipping verdicts 25--71\% of the time under static pushback, and 62--91\% with an adversarial LLM persuader. Critically, we find that pressure that succeeds in changing a judge's verdict is almost always net-corrupting with respect to ground truth. Beyond the framework itself, we identify baseline jury majority strength as the most effective single-shot signal for anticipating which items wiggle. Taken together, this is the first apples-to-apples cross-dataset comparison of mechanical, conformity, and persuadability tests in a judging context.
Chinese Translation
大型语言模型（LLM）评判者已成为模型评估、在线评分和奖励建模的核心基础设施。评判者通常通过在黄金数据上的准确性进行验证，但准确性对于它们在重新提示、挑战或持续反击下的稳定性几乎没有提供信息。我们引入了 extit{Wiggle Framework}，这是一个用于评估LLM评判者认知稳定性的统一压力测试框架。该框架从三个维度分解评判者的鲁棒性：机械一致性（在重新提示和重新构建下的稳定性）、单轮信念（在单次挑战下的稳定性）和多轮持续性（在持续或适应性压力下的稳定性）。我们利用该框架研究了9个前沿模型在14个评判任务中的表现，这些任务涵盖了安全性、毒性、AI写作检测和政治回应评估。每个模型作为评判者表现出显著的波动——在静态反击下，判决翻转的比例为25%到71%，而在对抗性LLM说服者的影响下，这一比例为62%到91%。关键是，我们发现成功改变评判者判决的压力几乎总是对真实情况造成净腐蚀。除了框架本身，我们还确定了基线陪审团多数强度是预测哪些项目会波动的最有效的单次信号。综合来看，这是在评判背景下对机械性、一致性和说服力测试的首次逐项跨数据集比较。
cs.AI / 23 / 2608.12654
SteerBench-Work: A Benchmark for Agent Steering at Action Boundaries
SteerBench-Work：一个用于行动边界代理引导的基准测试
Abstract
Long-running LLM agents act through tools, and a single step can send an email, merge a pull request, or wire a payment. The steering decision is the pre-commit choice at that boundary: proceed, or hold for human or policy review. We introduce SteerBench-Work, an incident-anchored, bidirectional benchmark for that decision in workplace agents across developer operations, customer service, finance, legal, medical, HR, and security. Release v2026-05 contains 106 scenarios anchored in public incidents, paired evidence-reversed mirrors, and calibration controls, with labels split nearly evenly between proceed and hold so the two error directions get near-identical numbers of chances. A model sees the proposed action and the available evidence, returns a gate decision, and is scored on whether it crosses or holds the boundary correctly. Across 30 model conditions the failures run almost entirely in one direction: models wrongly hold authorized, evidence-cleared work on 28.1% of opportunities and wrongly allow unsafe work on 1.0%. The hardest cases are risk-resolved commits, where signed or structured evidence has already cleared a real risk trigger, and models score markedly worse on evidence-reversed mirrors of famous incidents (63.8%) than on the incidents themselves (98.5%). General capability is not the same as steering calibration: higher-capability models often over-refuse at the commit boundary, and more reasoning can repair a weak gate while leaving a calibrated one flat. The public leaderboard is at steerbench.com.
Chinese Translation
长期运行的LLM代理通过工具进行操作，单一步骤可以发送电子邮件、合并拉取请求或进行支付。引导决策是在该边界的预提交选择：继续，还是等待人类或政策审查。我们介绍了SteerBench-Work，这是一个基于事件的双向基准，用于在开发者操作、客户服务、金融、法律、医疗、人力资源和安全等领域的工作场所代理中进行这一决策。版本v2026-05包含106个基于公共事件的场景，配对的证据反向镜像和校准控制，标签几乎均匀分为继续和保持，以便两个错误方向获得近乎相同的机会。模型看到提议的行动和可用的证据，返回一个门决策，并根据其是否正确越过或保持边界进行评分。在30个模型条件下，失败几乎完全朝一个方向发生：模型在28.1%的机会中错误地保持了已授权、证据清除的工作，而在1.0%的机会中错误地允许了不安全的工作。最困难的案例是风险已解决的提交，其中签署或结构化证据已经清除了真实的风险触发器，模型在著名事件的证据反向镜像上得分明显低于事件本身（63.8%对98.5%）。一般能力与引导校准并不相同：高能力模型在提交边界上往往过度拒绝，而更多的推理可以修复一个弱门，同时使一个校准良好的门保持平坦。公共排行榜位于steerbench.com。
cs.AI / 24 / 2608.12657
General Probabilities of Causation with Causal Knowledge
具有因果知识的因果概率概述
Abstract
Probabilities of causation (PoCs) characterize individual causal responses that cannot be directly observed and therefore generally require partial identification. Tian and Pearl first derived theoretically sharp bounds for binary PoCs, including the probability of necessity (PN), the probability of sufficiency (PS), and the probability of necessity and sufficiency (PNS). Mueller et al. subsequently tightened the bounds for binary PNS by incorporating causal information encoded in covariates and mediators. More recently, Li and Pearl, as well as Shu et al., extended PoCs to multivalued settings and derived corresponding theoretical bounds. These developments naturally raise the question of whether additional causal knowledge can further tighten the bounds in multivalued settings. This paper addresses this question by deriving tighter bounds for multivalued PoCs through the incorporation of causal information encoded in covariates and mediators. We illustrate the theoretical results with toy examples, while simulation studies further demonstrate that the proposed bounds are tighter than existing nonbinary bounds.
Chinese Translation
因果概率（PoCs）表征个体因果反应，这些反应无法直接观察，因此通常需要部分识别。Tian 和 Pearl 首先为二元 PoCs 推导了理论上的严格界限，包括必要性概率（PN）、充分性概率（PS）和必要性与充分性概率（PNS）。随后，Mueller 等人通过结合编码在协变量和中介变量中的因果信息，进一步收紧了二元 PNS 的界限。最近，Li 和 Pearl 以及 Shu 等人将 PoCs 扩展到多值情境，并推导出相应的理论界限。这些发展自然引发了一个问题：额外的因果知识是否可以进一步收紧多值情境下的界限。本文通过结合编码在协变量和中介变量中的因果信息，推导出多值 PoCs 的更紧界限来解决这一问题。我们用简单示例说明理论结果，同时模拟研究进一步表明，所提出的界限比现有的非二元界限更为紧凑。
cs.AI / 25 / 2608.12670
Designing AI Pipelines for Decision-Ready ITSM Intelligence
为决策准备的IT服务管理智能设计AI管道
Abstract
IT service management (ITSM) systems accumulate large volumes of heterogeneous ticket data that are difficult for sales and executive stakeholders to convert into actionable intelligence. This paper presents a sociotechnical AI pipeline, designed and evaluated following design science research principles, that transforms raw ITSM exports into a multilevel decision-support artifact. The pipeline combines LLM-based schema normalization, HDBSCAN sub-topic clustering, and hierarchical agglomerative clustering to generate executive-facing Main-topics and granular Sub-topics. A stakeholder evaluation across six artifacts and five raters from Sales Engineering and customer success roles shows that all four decision-support metrics, interpretability, actionability, trust, and likelihood of use, on average exceed 4.0 out of 5.0, with trust as the most consistent signal. The findings position ITSM analytics as an Information Systems (IS) problem of transformation, abstraction, and human-centered design.
Chinese Translation
IT服务管理（ITSM）系统积累了大量异构的工单数据，这些数据对于销售和高管利益相关者来说难以转化为可操作的智能。本文提出了一种社会技术AI管道，该管道遵循设计科学研究原则进行设计和评估，能够将原始ITSM导出数据转化为多层次的决策支持工具。该管道结合了基于大型语言模型（LLM）的模式规范化、HDBSCAN子主题聚类和层次聚合聚类，生成面向高管的主要主题和细化的子主题。对来自销售工程和客户成功角色的五位评审者对六个工具的利益相关者评估显示，所有四个决策支持指标——可解释性、可操作性、信任度和使用可能性——的平均得分均超过5分中的4.0，其中信任度作为最一致的信号。研究结果将ITSM分析定位为一个信息系统（IS）转型、抽象和以人为本设计的问题。
cs.AI / 26 / 2608.12671
On the Expressive Power of Transformers
变换器的表达能力
Abstract
Multi-layer transformers form the critical component of essentially all large language models (LLMs) in use today. Because of their ubiquity and computational capability, there is a rapidly growing body of work that aims to precisely calibrate the expressive power of transformers as language recognizers by comparing them against standard models of computation studied for decades by the theoretical computer science community. In this endeavor, circuit complexity has by and large emerged as the "correct" branch of computational complexity to analyze the expressive power of transformers; the reason is that parameterizing transformers by the various resources they use, such as attention and precision, leads to direct comparisons with different classes of circuits parameterized by resources such as type of gates, size, and depth. Here, we present an overview of selected results that delineate the expressive power of transformers using concepts and methods from circuit complexity.
Chinese Translation
多层变换器是当今几乎所有大型语言模型（LLMs）的关键组成部分。由于其普遍性和计算能力，越来越多的研究旨在通过将变换器与理论计算机科学界研究了数十年的标准计算模型进行比较，从而精确校准变换器作为语言识别器的表达能力。在这一努力中，电路复杂性在很大程度上成为了分析变换器表达能力的“正确”计算复杂性分支；原因在于，通过各种资源（如注意力和精度）对变换器进行参数化，可以与不同类别的电路进行直接比较，这些电路的参数化资源包括门的类型、大小和深度。在这里，我们展示了一些选定结果的概述，这些结果利用电路复杂性中的概念和方法描绘了变换器的表达能力。
cs.AI / 27 / 2608.12674
Lines and Ladders: A Context-Aware Multi-Agent Framework for Large-Scale Retail Price Taxonomy
线与梯：一种面向上下文的大规模零售价格分类的多智能体框架
Abstract
Maintaining price consistency and executing an Every Day Low Price strategy is critical for global retailers. However, with catalogs spanning millions of active items, manual governance of price relationships is infeasible. Inconsistent pricing across item variants distorts customer value perception and cannibalizes sales. To address this, we present a scalable, context-aware Multi-Agent Framework designed to automate the construction of "Lines and Ladders" pricing taxonomies. Our framework employs specialized LLM agents to construct these coherent pricing structures by identifying key attributes, extracting multi-modal values, and applying hierarchical grouping logic. Evaluated on real-world enterprise data and deployed in production, our 3-Agent system achieves an F1-score of 0.83 for Lines, outperforming single-agent baselines by mitigating cognitive overload. The system achieves >90% precision and >75% recall in Food & Consumables, and 80.2% assignment accuracy in the unstructured General Merchandise catalog.
Chinese Translation
维护价格一致性并执行每日低价策略对全球零售商至关重要。然而，随着目录中活跃商品数量达到数百万，手动管理价格关系变得不可行。商品变体之间的不一致定价扭曲了客户的价值感知，并导致销售的自相残杀。为了解决这一问题，我们提出了一种可扩展的、面向上下文的多智能体框架，旨在自动构建“线与梯”定价分类。我们的框架采用专门的LLM（大语言模型）智能体，通过识别关键属性、提取多模态值和应用层次分组逻辑来构建这些一致的定价结构。在真实企业数据上进行评估并投入生产后，我们的3智能体系统在“线”方面达到了0.83的F1分数，优于单智能体基线，减轻了认知负担。该系统在食品和消费品领域实现了超过90%的精准度和超过75%的召回率，在非结构化的综合商品目录中实现了80.2%的分配准确率。
cs.AI / 28 / 2608.12675
Privacy-Preserving RAG by Concealing Sensitive Information from External LLMs
通过隐藏敏感信息保护隐私的检索增强生成模型（RAG）
Abstract
Retrieval-Augmented Generation (RAG) is widely used to improve the performance of Large Language Models (LLMs) in answering user queries. Existing privacy research on RAG has focused on preventing unauthorized users from accessing sensitive data. However, another important problem that is often overlooked in RAG privacy research is that external generators have access to the query and the retrieved documents, which may contain confidential information that could potentially be misused or accessed for unintended purposes. In this paper, we introduce the Sensitive Entity Alias Generator (SEAG), a privacy-preserving framework that empowers users to utilize powerful third-party generators without disclosing sensitive information. SEAG introduces a lightweight model that locates sensitive entities, generates corresponding aliases, and constructs an entity replacement table. The table is used to replace sensitive words in the user's query and in the retrieved documents before they are forwarded to an external generator. For this purpose, two datasets were constructed: one for fine-tuning SEAG models to generate entity replacement tables, and another for evaluating the entire SEAG framework. The experimental results demonstrate the success of the SEAG framework. As for the User metric, which measures the ability of the model to provide a correct response to the user while hiding sensitive information from the external generator, all SEAG models achieved over 80% accuracy. Additional analysis further evaluated the ability of SEAG models Qwen-3, LLaMA-3.2, and Phi-4 to hide all sensitive entities within given documents. The results show good performance with total accuracies of 77.83%, 76.73%, and 74.91%, respectively.
Chinese Translation
检索增强生成（RAG）广泛用于提升大型语言模型（LLMs）在回答用户查询时的表现。现有关于RAG的隐私研究主要集中在防止未授权用户访问敏感数据。然而，在RAG隐私研究中，另一个常被忽视的重要问题是外部生成器可以访问查询和检索到的文档，这些文档可能包含机密信息，可能被滥用或用于意想不到的目的。本文提出了一种隐私保护框架——敏感实体别名生成器（Sensitive Entity Alias Generator, SEAG），使用户能够在不泄露敏感信息的情况下利用强大的第三方生成器。SEAG引入了一种轻量级模型，能够定位敏感实体，生成相应的别名，并构建实体替换表。该表用于在将用户的查询和检索到的文档转发给外部生成器之前，替换其中的敏感词。为此，构建了两个数据集：一个用于微调SEAG模型以生成实体替换表，另一个用于评估整个SEAG框架。实验结果表明，SEAG框架取得了成功。在用户指标方面，该指标衡量模型在隐藏敏感信息的同时向用户提供正确响应的能力，所有SEAG模型的准确率均超过80%。进一步分析评估了SEAG模型Qwen-3、LLaMA-3.2和Phi-4在给定文档中隐藏所有敏感实体的能力。结果显示其总准确率分别为77.83%、76.73%和74.91%，表现良好。
cs.AI / 29 / 2608.12677
The Role of Natural Language Understanding in Multimodal Video-Based Dengue Diagnosis
自然语言理解在多模态视频基础登革热诊断中的作用
Abstract
Detecting infection-related behavioral changes in mosquitoes from video data is challenging because mosquitoes are small, move rapidly and irregularly, and are affected by environmental factors such as background, lighting, and shadows, which can make reliable feature extraction difficult. In this study, a YOLO- and Contrastive Language-Image Pre-training (CLIP)-based vision-language framework is proposed to classify mosquito flight frames of uninfected and Dengue virus serotype 2 (DENV2)-infected mosquitoes. First, YOLO is used to isolate mosquito regions from the background. Then, visual features extracted from video frames are aligned with biologically meaningful textual prompts in a shared embedding space. The multimodal model was fine-tuned using supervised bidirectional contrastive learning and evaluated through frame-level image-text similarity-based classification. The results show that the proposed method achieved 98.54% accuracy and 99.91% sensitivity at the frame level. After temporal aggregation of frame-level information, the model achieved complete video-level performance. The ablation results showed that fine-tuning and CLIP-based representations were essential for this domain, while the textual branch provided semantic image-text alignment rather than an accuracy advantage over the vision-only model. These findings suggest that vision-language models can provide a useful framework for analyzing infection-related biological behaviors from video data.
Chinese Translation
从视频数据中检测蚊子与感染相关的行为变化具有挑战性，因为蚊子体积小、移动迅速且不规则，并且受到背景、光照和阴影等环境因素的影响，这使得可靠的特征提取变得困难。在本研究中，提出了一种基于YOLO和对比语言-图像预训练（Contrastive Language-Image Pre-training, CLIP）的视觉-语言框架，用于分类未感染和登革热病毒血清型2（Dengue virus serotype 2, DENV2）感染蚊子的飞行帧。首先，使用YOLO从背景中分离出蚊子区域。然后，从视频帧中提取的视觉特征与共享嵌入空间中的生物学意义文本提示对齐。该多模态模型通过监督双向对比学习进行微调，并通过基于帧级图像-文本相似性的分类进行评估。结果表明，所提出的方法在帧级别上达到了98.54%的准确率和99.91%的灵敏度。在对帧级信息进行时间聚合后，模型实现了完整的视频级性能。消融实验结果表明，微调和基于CLIP的表示对该领域至关重要，而文本分支提供了语义图像-文本对齐，而不是相较于仅视觉模型的准确性优势。这些发现表明，视觉-语言模型可以为分析视频数据中的感染相关生物行为提供有用的框架。
cs.AI / 30 / 2608.12679
Beyond the Best Guess: Improving LLM Solution Coverage with Evolution Strategies
超越最佳猜测：利用进化策略提升大型语言模型的解答覆盖率
Abstract
Large Language Models (LLMs) are increasingly deployed in discovery domains such as math and science. The usual approach is to present the problem to the model and use its answer as the proposed solution. However, beyond this best guess, discovery can be enhanced by increasing test-time compute. In a process called pass@k, the model is allowed to explore the solution space and generate diverse candidate solutions. Unfortunately, the standard approach to post-training LLMs through Reinforcement Learning (RL) may limit pass@k: the model's output distribution narrows around high-reward outputs, causing the solution coverage to collapse. The alternative is to use Evolution Strategies (ES), a population-based, gradient-free post-training method that optimizes directly in weight space through random perturbations. As this paper shows, ES achieves consistently higher pass@k than RL and produces a broader output distribution with greater solution coverage. This coverage in turn makes it possible to achieve better results in e.g. standard math benchmarks. Thus, ES provides a better foundation for post-training in discovery problems and other domains where diverse solution coverage is critical.
Chinese Translation
大型语言模型（LLMs）在数学和科学等发现领域的应用日益增多。通常的方法是将问题呈现给模型，并将其答案作为建议的解决方案。然而，除了最佳猜测之外，通过增加测试时的计算能力，可以增强发现过程。在一个称为 pass@k 的过程中，模型被允许探索解空间并生成多样的候选解决方案。不幸的是，使用强化学习（RL）对后训练 LLMs 的标准方法可能限制了 pass@k：模型的输出分布围绕高奖励输出收窄，导致解答覆盖率下降。另一种选择是使用进化策略（ES），这是一种基于种群的、无梯度的后训练方法，通过随机扰动直接在权重空间中进行优化。正如本文所示，ES 在 pass@k 上的表现始终优于 RL，并产生更广泛的输出分布，具有更大的解答覆盖率。这种覆盖率反过来使得在例如标准数学基准测试中取得更好的结果成为可能。因此，ES 为发现问题及其他需要多样化解答覆盖的领域的后训练提供了更好的基础。
cs.AI / 31 / 2608.12743
Spatial Memory Agent: Experience-Grounded Procedure Memory for Spatial Intelligence
空间记忆代理：基于经验的空间智能过程记忆
Abstract
Spatial intelligence is becoming a foundation for embodied agents, robotic planning, and multimodal assistants. To improve the spatial reasoning ability of VLM agents, existing work has mainly followed two lines. One line uses post-training methods, such as supervised fine-tuning and reinforcement learning. Another line adopts an agentic paradigm in which the model calls external spatial tools, such as depth estimation and 3D reconstruction tools, to gather intermediate spatial evidence. We study a complementary and underexplored route: Can a frozen VLM agent improve its spatial reasoning through \textbf{parameter-update-free self-evolution}, without depending on external expert spatial tools at inference time? We present \textbf{Spatial Memory Agent (SMA)}, an \textbf{experience-grounded runtime framework} that converts verified spatial experience into reusable transferable lessons. In a verifiable spatial environment, SMA queries the frozen VLM, obtains a predicted answer and reward, and uses \textbf{verifier-guided reflection} to distill compact transferable lessons from spatial experience. SMA further assigns each lesson a \textbf{Transfer Reliability Score (TRS)}, which is initialized uniformly and calibrated from later retrieval outcomes as visit evidence of future transfer reliability. During \textbf{read-only deployment}, SMA retrieves lessons by semantic filter and similarity-TRS combined ranking, allowing the retrieved memory to guide frozen model inference. Across five representative spatial benchmarks and four base VLMs, SMA achieves the highest macro average in every base-model block and the best accuracy among the evaluated methods in most of the 20 evaluations, establishing a practical parameter-update-free path for spatial self-evolution across the evaluated frozen model scales and environments.
Chinese Translation
空间智能正成为具身代理、机器人规划和多模态助手的基础。为了提高视觉语言模型（VLM）代理的空间推理能力，现有研究主要沿着两个方向展开。一方面，使用后训练方法，如监督微调和强化学习。另一方面，采用代理范式，其中模型调用外部空间工具，如深度估计和三维重建工具，以收集中间空间证据。我们研究了一条互补且未被充分探索的路径：一个冻结的VLM代理能否通过 extbf{无参数更新的自我进化}来提高其空间推理能力，而不依赖于推理时的外部专家空间工具？我们提出了 extbf{空间记忆代理（SMA）}，这是一种 extbf{基于经验的运行时框架}，将经过验证的空间经验转化为可重复使用的可转移教训。在可验证的空间环境中，SMA查询冻结的VLM，获得预测答案和奖励，并利用 extbf{验证者引导的反思}从空间经验中提炼出紧凑的可转移教训。SMA进一步为每个教训分配一个 extbf{转移可靠性评分（TRS）}，该评分最初均匀初始化，并根据后续检索结果进行校准，作为未来转移可靠性的访问证据。在 extbf{只读部署}期间，SMA通过语义过滤和相似性-TRS结合排名来检索教训，使得检索到的记忆能够指导冻结模型的推理。在五个代表性的空间基准和四个基础VLM上，SMA在每个基础模型块中都达到了最高的宏平均，并在大多数20个评估中在评估方法中获得了最佳准确率，为在评估的冻结模型规模和环境中实现空间自我进化提供了一条实用的无参数更新路径。
cs.AI / 32 / 2608.12761
Correct Is Not Governed: Provenance Integrity in Agentic Workflows
不受治理的正确性：代理工作流中的来源完整性
Abstract
Agentic workflows are commonly evaluated by whether they reach the correct outcome. That is insufficient in institutional settings, where a correct action may rely on the wrong authority, an unsupported completion claim, or work made stale by a later change. We define governed execution as work whose decisions, completion, and response to change are supported by inspectable provenance. We present Matrix, a deterministic causal-state layer that records authority and fact dependencies, verifies completion evidence, and selectively invalidates affected work. Across controlled comparisons, governed and direct workflows often reached the same outcomes, but only the governed path consistently preserved governing evidence, refused unsupported closure, and limited recovery to dependent tasks. A role-separated transfer challenge then failed: a deterministically enforced completeness contract severely over-blocked synthetic packets produced outside its authoring context. These results do not establish Matrix as a general accuracy enhancer; they support its primary role as an institutional integrity layer for making agentic work auditable and independently verifiable.
Chinese Translation
代理工作流通常通过其是否达到正确的结果来进行评估。然而，在制度环境中，这种评估是不够的，因为一个正确的行动可能依赖于错误的权威、缺乏支持的完成声明，或因后续变更而变得过时的工作。我们将受治理的执行定义为其决策、完成和对变更的响应均由可检查的来源支持的工作。我们提出了 Matrix，这是一个确定性的因果状态层，记录权威和事实依赖关系，验证完成证据，并选择性地使受影响的工作失效。在受控比较中，受治理的工作流和直接工作流通常达到了相同的结果，但只有受治理的路径始终保留了治理证据，拒绝了不支持的结束，并将恢复限制在依赖任务上。随后，一个角色分离的转移挑战失败：一个确定性强制的完整性合同严重过度阻止了在其创作上下文之外生成的合成数据包。这些结果并未确立 Matrix 作为一种通用的准确性增强工具；它支持其作为制度完整性层的主要角色，使代理工作可审计且可独立验证。
cs.AI / 33 / 2608.12762
PROVE-RT: Generating Mechanized Theorem Prover Scripts for Real-Time Systems using LLMs
PROVE-RT：使用大型语言模型生成实时系统的机械化定理证明器脚本
Abstract
Schedulability analysis is essential for certifying real-time systems, but existing tests are often developed through pen-and-paper proofs that are difficult to scale, validate, and maintain. Mechanized verification in PROSA/ROCQ offers a rigorous alternative, yet manually constructing such proofs requires substantial domain expertise and proof-engineering effort. Recent successes of large language models (LLMs) across a wide range of tasks make them promising candidates for generating PROSA/ROCQ scripts for mechanized theorem provers. However, state-of-the-art LLMs often lack the PROSA-specific knowledge required to correctly use its modeling abstractions and proof patterns. This paper introduces PROVE-RT, an LLM-assisted framework for generating PROSA/ROCQ scripts to mechanize schedulability analyses in real-time systems literature. PROVE-RT guides generation through dependency-aware informal sketches, retrieval from processed PROSA documentation, staged skeleton generation, and proof completion. We construct a mechanization-oriented corpus from 1, 191 real-time systems papers, containing 13, 134 informal sketches with dependency information. On a curated evaluation set, direct prompting of state-of-the-art LLMs fails to reliably generate valid PROSA mechanizations, whereas PROVE-RT achieves a success rate of 44.7%. These results show that retrieval-guided and staged LLM assistance can improve automated mechanization of schedulability analysis in PROSA/ROCQ.
Chinese Translation
可调度性分析对于认证实时系统至关重要，但现有的测试通常通过纸笔证明开发，这些证明难以扩展、验证和维护。PROSA/ROCQ中的机械化验证提供了一种严格的替代方案，但手动构建此类证明需要大量的领域专业知识和证明工程努力。近期大型语言模型（LLMs）在广泛任务中的成功使其成为生成PROSA/ROCQ脚本以机械化定理证明器的有希望的候选者。然而，最先进的LLMs通常缺乏正确使用PROSA特定知识所需的建模抽象和证明模式。本文介绍了PROVE-RT，一个LLM辅助框架，用于生成PROSA/ROCQ脚本，以机械化实时系统文献中的可调度性分析。PROVE-RT通过依赖感知的非正式草图、从处理过的PROSA文档中检索、分阶段生成骨架和证明完成来指导生成。我们从1191篇实时系统论文中构建了一个以机械化为导向的语料库，包含13134个带有依赖信息的非正式草图。在一个经过精心策划的评估集上，直接提示最先进的LLMs未能可靠地生成有效的PROSA机械化，而PROVE-RT的成功率达到了44.7%。这些结果表明，检索引导和分阶段的LLM辅助可以改善PROSA/ROCQ中可调度性分析的自动机械化。
cs.AI / 34 / 2608.12788
ARAC: Benchmarking Auto-Research's Alignment and Completeness on End-to-End Researchs
ARAC：端到端研究中自动研究的对齐性和完整性的基准评估
Abstract
The rapid advancement of Auto-Research has surfaced a fundamental evaluation challenge: how can we measure the alignment, logical coherence, and evolutionary completeness of its research trajectory with human research behavior? We propose Auto-Research's Alignment and Completeness, ARAC-Bench: a Researcher-Mimicking Evaluation framework that shifts the objective from matching final answers to reproducing high-quality human research processes. The framework operates through two synergistic components: the Academic Cognition Skills system, which is the first to transforms implicit reviewer expertise into stage-calibrated, quantifiable rubrics; and a three-stage capability diagnostic protocol, which decomposes the research process under strict modular constraints into three traceable, mutually independent dimensions: Proposal, Experiment, and Synthesis. Systematic evaluation of 11 SOTA frameworks yields a best alignment score of only 67.9 of 100, revealing a significant gap in simulating rigorous human methodology. Validation against Ph.D. Candidates rankings shows a strong correlation of 0.8141, confirming that ARAC-Bench reliably reflects the dimensions researchers truly value. ARAC-Bench provides not only a fine-grained diagnostic tool but also a scalable reward signal for training the next generation of autonomous research systems.
Chinese Translation
自动研究的快速发展带来了一个基本的评估挑战：我们如何衡量其研究轨迹与人类研究行为之间的对齐性、逻辑一致性和演变完整性？我们提出了自动研究的对齐性和完整性评估框架ARAC-Bench：一个模仿研究者的评估框架，旨在将目标从匹配最终答案转变为再现高质量的人类研究过程。该框架通过两个协同组件运作：学术认知技能系统，这是首个将隐性评审专家知识转化为阶段校准、可量化评分标准的系统；以及一个三阶段能力诊断协议，该协议在严格的模块约束下将研究过程分解为三个可追踪、相互独立的维度：提案、实验和综合。对11个最先进框架的系统评估显示，最佳对齐得分仅为67.9（满分100），揭示了在模拟严谨人类方法论方面的显著差距。与博士候选人排名的验证显示出强相关性0.8141，确认ARAC-Bench可靠地反映了研究人员真正重视的维度。ARAC-Bench不仅提供了一个细致的诊断工具，还为训练下一代自主研究系统提供了可扩展的奖励信号。
cs.AI / 35 / 2608.12842
CABS+: Efficient and Scalable Model Merging via Conflict-Aware Sparsification and Adaptive Weight Allocation
CABS+: 通过冲突感知稀疏化和自适应权重分配实现高效且可扩展的模型合并
Abstract
Model merging has recently attracted significant attention as a promising paradigm for constructing unified multi-task models without requiring additional retraining. However, parameter conflicts and knowledge interference across tasks often degrade merged-model performance. Prior work introduced Conflict-Aware and Balanced Sparsification (CABS), which reduces parameter interference through structured pruning and sequential masking. However, CABS relies on grid search to determine scaling coefficients, resulting in exponential time complexity, while its optimization objective can be dominated by high-performance tasks, leading to suboptimal overall performance. To address these limitations, we extend CABS and propose CABS+. Specifically, Adaptive Weight Allocation (AWA) optimizes merging coefficients via a gradient-free search scheme to reduce time complexity, while an asymmetric fitness function promotes more comprehensive performance gains across tasks. Moreover, we conduct a systematic empirical study of key factors influencing model merging performance and propose Relative Synergy Score (RSS) to quantify model mergeability and guide model selection. We compare CABS+ with state-of-the-art model merging methods, including CABS, AdaMerging, and WUDIMerging, across 27 datasets and 5 models covering large language, small-scale language, and vision models. Extensive experiments verify the effectiveness and efficiency of CABS+. Compared with AdaMerging and WUDIMerging, CABS+ improves overall performance by 16.97% and 12.93%, respectively, exhibits stronger stability and robustness across varying task numbers and model architectures, uses less than 25% of the GPU memory required by AdaMerging, and achieves nearly a 4x speedup in merging time over WUDIMerging.
Chinese Translation
模型合并近年来受到广泛关注，作为构建统一多任务模型的一种有前景的范式，无需额外的再训练。然而，任务之间的参数冲突和知识干扰往往会降低合并模型的性能。先前的研究提出了冲突感知平衡稀疏化（Conflict-Aware and Balanced Sparsification，CABS），通过结构化剪枝和顺序掩蔽减少参数干扰。然而，CABS依赖网格搜索来确定缩放系数，导致指数时间复杂度，同时其优化目标可能被高性能任务主导，从而导致整体性能不佳。为了解决这些局限性，我们扩展了CABS并提出了CABS+。具体而言，自适应权重分配（Adaptive Weight Allocation，AWA）通过无梯度搜索方案优化合并系数，以降低时间复杂度，而不对称适应度函数促进了跨任务的更全面的性能提升。此外，我们对影响模型合并性能的关键因素进行了系统的实证研究，并提出了相对协同评分（Relative Synergy Score，RSS）来量化模型的可合并性并指导模型选择。我们将CABS+与最先进的模型合并方法进行比较，包括CABS、AdaMerging和WUDIMerging，涵盖27个数据集和5个模型，涉及大语言模型、小规模语言模型和视觉模型。大量实验证实了CABS+的有效性和效率。与AdaMerging和WUDIMerging相比，CABS+整体性能分别提高了16.97%和12.93%，在不同任务数量和模型架构下表现出更强的稳定性和鲁棒性，使用的GPU内存不到AdaMerging的25%，并且在合并时间上实现了接近4倍的加速，相较于WUDIMerging。
cs.AI / 36 / 2608.12847
Beyond Retrieval: Query-Conditioned Reuse of Long-Horizon Agent Trajectories
超越检索：基于查询的长时间跨度代理轨迹重用
Abstract
Retrieval can identify a past trajectory that may matter, yet it does not specify how an acting agent should use that trajectory after users, entities, constraints, or environment state have changed. We identify this post-retrieval reuse step as a distinct bottleneck for long-horizon trajectory memory and formulate an evaluation framework that holds candidate retrieval, target state, model, decoding, and tool budget fixed while varying the support delivered to the agent. We instantiate the framework with query-conditioned reuse (QCR), a deliberately simple target-bound note that records a reusable procedure, bindings to recover, applicability conditions, and verification requirements. QCR serves to test the reuse hypothesis rather than to claim a universally preferred memory format. Across 2,391 target instances in WebArena, WorkArena, and AppWorld, QCR reaches 62.3% average Success, 10.7 points above Full Trajectory, while using 48.9% fewer online tokens. Summary reranking selects a reusable memory for 94.8% of targets, placing end-task Success within 1.8 points of an oracle reusable selector. Analyses by trajectory length and source--target binding shift show that direct trajectory injection loses much of its utility as traces grow longer or source-specific values change, whereas target-bound support preserves a larger share of the measured gain. The resulting framework separates retrieval quality from the problem of turning retrieved experience into safe, useful support for a new task.
Chinese Translation
检索可以识别出可能重要的过去轨迹，但它并未指定在用户、实体、约束或环境状态发生变化后，行动代理应如何使用该轨迹。我们将这一检索后的重用步骤视为长时间跨度轨迹记忆的一个独特瓶颈，并制定了一个评估框架，该框架在固定候选检索、目标状态、模型、解码和工具预算的同时，变化提供给代理的支持。我们通过查询条件重用（Query-Conditioned Reuse, QCR）实例化该框架，QCR是一个故意简单的目标绑定笔记，记录了可重用的程序、恢复绑定、适用条件和验证要求。QCR旨在测试重用假设，而不是声称存在一种普遍优选的记忆格式。在WebArena、WorkArena和AppWorld中的2391个目标实例中，QCR实现了62.3%的平均成功率，比完整轨迹高出10.7个百分点，同时使用的在线令牌减少了48.9%。摘要重排名为94.8%的目标选择了可重用的记忆，使最终任务的成功率与一个理想的可重用选择器相差仅1.8个百分点。通过轨迹长度和源-目标绑定变化的分析表明，随着轨迹变长或源特定值变化，直接轨迹注入的效用大幅降低，而目标绑定支持则保留了更大比例的测量增益。由此产生的框架将检索质量与将检索到的经验转化为新任务的安全、有用支持的问题分开。
cs.AI / 37 / 2608.12851
Practice Makes Unsafe: Skill Misevolution in Self-Improving LLM Agents
实践使不安全：自我改进的LLM代理中的技能误演化
Abstract
Self-improving LLM agents convert successful trajectories into persistent cross-task state. An unsafe success can thereby become reusable policy after its triggering input disappears. Skill evolution makes this failure measurable by distilling operational trajectories into executable, transferable, and inspectable procedures. Because evolution optimizes task outcomes rather than procedure safety, compromised experience can cause skill misevolution. Existing benchmarks measure current behavior or static artifacts but cannot attribute risk across authoring, retrieval, and later execution. To expose this lifecycle, we introduce SkillMisevo-Gym, a lifecycle-aware harness that versions skill state across agent frameworks, and SkillMisevo-Bench, a frozen design from malicious exposure to carryover tasks, with concept-aligned benign tasks and nine lifecycle metrics. We also introduce SafeEvolve, a wrapper that repairs unsafe content and governs subsequent reuse. Across 25 agent-method configurations, each covering 525 tasks in 25 episodes, all 21 evolved configurations author unsafe artifacts, while only fifteen lead to fresh-session harm. In the exposure sweep, three malicious tasks raise carryover ASR from 16.0% to 35.3%. Across representative skill evolution methods, SafeEvolve reduces unsafe retrieval and fresh-session harm by 26.7 and 17.3 percentage points, respectively, while mean benign utility changes by only 0.4 points. Together, persistent-adaptation safety must govern what updates write and what future executors reuse. Code is available at https://github.com/henrymao2004/misevolve.
Chinese Translation
自我改进的LLM代理将成功的轨迹转化为持久的跨任务状态。因此，不安全的成功在其触发输入消失后可能成为可重用的策略。技能演化通过将操作轨迹提炼为可执行、可转移和可检查的程序，使得这种失败可被量化。由于演化优化的是任务结果而非程序安全，妥协的经验可能导致技能误演化。现有基准测量当前行为或静态工件，但无法在创作、检索和后续执行之间归因风险。为了揭示这一生命周期，我们引入了SkillMisevo-Gym，一个生命周期感知的框架，用于在代理框架之间版本化技能状态，以及SkillMisevo-Bench，一个从恶意暴露到延续任务的冻结设计，包含概念一致的良性任务和九个生命周期指标。我们还引入了SafeEvolve，一个修复不安全内容并管理后续重用的包装器。在25种代理方法配置中，每种覆盖525个任务，21种演化配置均产生不安全的工件，而只有15种导致新会话的危害。在暴露扫描中，三个恶意任务将延续的ASR从16.0%提高到35.3%。在代表性的技能演化方法中，SafeEvolve分别将不安全检索和新会话危害降低了26.7和17.3个百分点，而良性效用的平均变化仅为0.4点。总之，持久适应的安全性必须管理更新的写入内容以及未来执行者的重用。代码可在 https://github.com/henrymao2004/misevolve 获取。
cs.AI / 38 / 2608.12863
AI and Consumer Rights in India Working Paper
印度的人工智能与消费者权益工作论文
Abstract
As AI systems proliferate in consumer facing applications, questions about liability for AI related harms remain unresolved. This working paper examines whether India's Consumer Protection Act, 2019, adequately addresses harm caused by defective AI products and services, and whether it proportionately allocates liability across the AI value chain. The Act's broad definitions of product liability, harm, and deficiency appear technology agnostic and potentially applicable to AI related incidents including personal injury, psychological harm, biased outputs, and loss of control. However, significant gaps remain. Proving causation between AI defects and consumer harm presents a technical challenge, as AI failures often stem from design choices rather than discrete defects. Additionally, the Act's framework assumes distinct roles for manufacturers, sellers, and service providers, yet the AI value chain involves overlapping responsibilities among data providers, model developers, deployers, and users that do not neatly map to these categories. Current liability frameworks lack proportionate mechanisms to effectively address complex, multistakeholder AI harms. While the Act may cover AI entities, enforcement requires clarification on sector specific overlaps.
Chinese Translation
随着人工智能系统在面向消费者的应用中不断普及，关于人工智能相关损害的责任问题仍未得到解决。本文工作论文探讨了印度2019年《消费者保护法》是否充分应对由缺陷人工智能产品和服务造成的损害，以及该法是否在人工智能价值链中合理分配责任。该法对产品责任、损害和缺陷的广泛定义似乎与技术无关，并可能适用于包括人身伤害、心理伤害、偏见输出和失控等人工智能相关事件。然而，仍然存在显著的缺口。证明人工智能缺陷与消费者损害之间的因果关系是一项技术挑战，因为人工智能故障往往源于设计选择而非离散缺陷。此外，该法的框架假设制造商、销售商和服务提供者具有明确的角色，但人工智能价值链涉及数据提供者、模型开发者、部署者和用户之间的重叠责任，这些角色并不能简单地映射到这些类别上。目前的责任框架缺乏有效应对复杂的多方利益相关者人工智能损害的比例机制。尽管该法可能涵盖人工智能实体，但在执行时需要明确行业特定的重叠问题。
cs.AI / 39 / 2608.12877
ReflectFact: Self-Reflective Agents for Improving Comprehension and Reasoning in Multi-Hop Fact Verification
ReflectFact：自我反思代理用于改善多跳事实验证中的理解和推理
Abstract
Multi-hop fact verification, which verifies claims by reasoning over multiple pieces of evidence, is critical for combating misinformation on social media yet remains highly challenging. Recent methods primarily rely on multi-agent collaboration to decompose fact verification into specialized subtasks. However, these methods face two critical limitations: (1) agents may perform individual subtasks without sufficient awareness of the global verification objective, causing their reasoning to deviate from the intended direction; and (2) conflicts between parametric knowledge and the provided evidence may undermine evidence-grounded reasoning and lead to incorrect verdicts. To address these challenges, we propose ReflectFact, a novel self-reflective agent framework for multi-hop fact verification. ReflectFact introduces three key tasks. Explicit Reasoning Path Planning builds an evidence-grounded reasoning path by resolving implicit entities, decomposing the claim into sub-questions, and integrating the verified facts into a verdict. Evidence-Drift Verification makes the agent re-answer by quoting the supporting evidence when a grounded answer merely echoes its parametric prior, thereby calibrating evidence deviation to ensure grounded comprehension. Reasoning Reflection Verification re-examines each reasoning step and regenerates it once an inconsistency is detected, correcting reasoning flaws such as location bias and replacement bias through a global task perspective. Subsequently, the agent aggregates validated reasoning chains to yield reliable verdicts. Extensive experiments on HOVER and EX-FEVER demonstrate that ReflectFact effectively remedies the comprehension and reasoning defects of existing methods, achieving state-of-the-art performance and respectively outperforming the strongest baseline by 3.32\% and 2.78\% on the two datasets.
Chinese Translation
多跳事实验证通过对多条证据进行推理来验证主张，对于打击社交媒体上的虚假信息至关重要，但仍然面临很大的挑战。近期的方法主要依赖于多代理协作，将事实验证分解为专门的子任务。然而，这些方法存在两个关键限制：（1）代理可能在没有充分意识到全局验证目标的情况下执行各自的子任务，导致其推理偏离预期方向；（2）参数知识与提供的证据之间的冲突可能削弱基于证据的推理，并导致错误的裁决。为了解决这些挑战，我们提出了ReflectFact，一种新颖的自我反思代理框架用于多跳事实验证。ReflectFact引入了三个关键任务。显式推理路径规划通过解析隐含实体、将主张分解为子问题以及将验证过的事实整合到裁决中，构建基于证据的推理路径。证据漂移验证使代理在基础答案仅仅反映其参数先验时，通过引用支持证据重新回答，从而校准证据偏差以确保扎实的理解。推理反思验证重新审视每一步推理，并在检测到不一致时重新生成推理，通过全局任务视角纠正推理缺陷，如位置偏见和替换偏见。随后，代理聚合验证过的推理链，以得出可靠的裁决。在HOVER和EX-FEVER上的大量实验表明，ReflectFact有效弥补了现有方法的理解和推理缺陷，在两个数据集上分别比最强基线提高了3.32%和2.78%，达到了最先进的性能。
cs.AI / 40 / 2608.12892
Predictive Memory Localization: Forecasting Selective Intervention Paths from Internal Signals
预测记忆定位：从内部信号预测选择性干预路径
Abstract
Activation steering turns localized representations into control directions, but localization alone does not reveal whether a direction has a selective operating regime. We introduce Predictive Memory Localization (PML), which treats the measured-grid intervention path as the predictive object of memory localization. PML separates random-calibrated target movement from semantic-neighbor and capability damage, and compares static localization and supervised geometry with a strength-disjoint low-dose causal response. Our frozen study covers 3,000 records from nine datasets and fourteen domains, yielding 30,000 distinct record-direction-layer paths and 210,000 distinct path-strength evaluations. At layer 7, the geometry-derived RFM/AGOP direction reaches 13.1% target-any and 12.3% clean-any, exceeding random by 3.6 and 3.4 percentage points under a record-paired bootstrap. Across record-, dataset-, and domain-grouped splits, responses at $|\alpha|=0.1$ are the strongest signal for outcomes at disjoint strengths $|\alpha|\in\{0.25,0.5\}$. On held-out records, a predictor-driven selector chooses a coefficient or abstains, improves utility and reduces semantic-neighbor damage relative to a train-tuned fixed-strength policy, and avoids most evaluations in a dense scan. Across three residual-norm-matched base models, learned directions retain selective-path gains and low-dose responses yield 0.801-0.828 record-held-out macro AUROC. PML therefore turns memory localization into a falsifiable forecast of margin-level selective outcomes and a risk-aware intervention decision.
Chinese Translation
激活引导将局部化表示转化为控制方向，但仅靠定位并不能揭示某一方向是否具有选择性操作机制。我们提出了预测记忆定位（Predictive Memory Localization, PML），将测量的网格干预路径视为记忆定位的预测对象。PML将随机校准的目标运动与语义邻近和能力损伤分开，并比较静态定位和监督几何与强度不相交的低剂量因果响应。我们的冷冻研究涵盖了来自九个数据集和十四个领域的3000条记录，产生了30000条不同的记录-方向-层路径和210000条不同的路径-强度评估。在第7层，几何推导的RFM/AGOP方向达到了13.1%的目标任意性和12.3%的清晰任意性，分别比随机结果高出3.6和3.4个百分点，在记录配对自助法下进行评估。在记录、数据集和领域分组的拆分中，$|eta|=0.1$的响应是对不相交强度$|eta| ext{in}\{0.25,0.5\ ext{的结果最强信号。在保留的记录中，基于预测器驱动的选择器选择一个系数或选择不干预，相比于经过训练调整的固定强度策略，提升了效用并减少了语义邻近损伤，并在密集扫描中避免了大多数评估。在三个残差范数匹配的基础模型中，学习到的方向保持了选择路径的增益，低剂量响应的记录保留宏观AUROC在0.801-0.828之间。因此，PML将记忆定位转化为可验证的边际水平选择性结果预测和风险意识干预决策。
cs.AI / 41 / 2608.12895
Agent Behavioral Contracts II: Certifying Compositional Reliability Without Assuming Independence
代理行为合同 II：在不假设独立性的情况下认证组合可靠性
Abstract
Compositional reliability bounds for multi-agent systems multiply component reliabilities, a step licensed by a conditional-independence assumption that is routinely stated and rarely tested. We test it. Two instances of one model, in a two-agent handoff, co-fail on 90.0% of the missions on which either fails (log OR 6.66, 95% CI [6.38, 7.00]; phi 0.916), in a preregistered evaluation of 18,000 missions scored by deterministic code with no LLM judge. Substituting a different model reduces the association in six of six contrasts; substituting a different vendor, model already different, does not -- a registered hypothesis reported as a null. The error is signed and runs against the operator: positive dependence inflates joint failure above the independence product, so redundancy is over-credited exactly when components share a model. The assumption-free alternative is often vacuous, and fitting a dependence model is worse: we prove a bootstrap bound on a fitted model's functional loses coverage of the truth as n grows, the identification gap being O(1) while the bootstrap haircut is O(n^{-1/2}). More data makes such a certificate worse, with no visible symptom. We give a finite-sample certificate assuming no dependence structure: a linear program over the joint, over a Bonferroni-Clopper-Pearson box around measured co-execution moments. It is sound, sharp for the information supplied, and monotone in the moment family. Enriching ten moment functionals to fourteen narrows the identified interval by 85.7% and lifts the certified floor from 0.2455 to 0.4116. A companion anytime-valid certificate holds type-I error at 0.0471 under optional stopping. Common dependence statistics are marginal-bounded and can reverse an apparent ordering of conditions when the compared agents fail at different rates. Contracts, scoring code, analysis scripts, and the preregistration are released.
Chinese Translation
多智能体系统的组合可靠性界限通过乘以组件可靠性来计算，这一步骤是基于一个条件独立性假设，该假设通常被陈述但很少被检验。我们对此进行了检验。在一个两代理交接的模型中，两个实例在90.0%的任务中共同失败，这些任务中任一代理失败（对数 OR 6.66, 95% CI [6.38, 7.00]; phi 0.916），这是对18,000个任务的预注册评估，评估由确定性代码评分，没有 LLM 判别。替换为不同模型在六个对比中减少了关联；而替换为不同供应商的模型（已经不同）则没有——这是一个注册假设，结果为零。该错误是有符号的，并且与操作员相悖：正依赖性使得联合失败率高于独立性乘积，因此在组件共享模型时冗余被过高估计。无假设的替代方案往往是空洞的，而拟合依赖模型则更糟：我们证明了一个拟合模型的引导界限在样本量 n 增长时失去对真实值的覆盖，识别差距为 O(1)，而引导削减为 O(n^{-1/2})。更多数据使得这样的证书变得更糟，且没有明显的症状。我们给出了一个有限样本证书，假设没有依赖结构：这是一个关于联合的线性规划，围绕测量的共同执行时刻的 Bonferroni-Clopper-Pearson 区间。该证书是可靠的，针对提供的信息是尖锐的，并且在时刻族中是单调的。将十个时刻函数扩展到十四个将识别区间缩小了85.7%，并将认证下限从0.2455提升到0.4116。一个伴随的随时有效证书在可选停止下将第一类错误保持在0.0471。常见的依赖统计是边际有界的，并且在比较代理以不同速率失败时可能会逆转条件的明显排序。合同、评分代码、分析脚本和预注册信息已发布。
cs.AI / 42 / 2608.12928
Polish Medical Visual Question Answering: Vision-Language Models Underutilize Visual Evidence
波兰医学视觉问答：视觉-语言模型未充分利用视觉证据
Abstract
We introduce a Polish-language medical visual question answering (VQA) benchmark, built from Polish Board Certification Examination questions for licensed physicians and dentists pursuing specialist certification. The benchmark comprises image-containing questions spanning diverse medical specialties and visual domains, together with a text-only question answering (QA) control set. We evaluate Polish-oriented, general-purpose open-weight, and commercial vision-language models. The task remains challenging: the best model achieves 79.0\% accuracy on the full VQA set, and only GPT-5.6 surpasses the approximate human reference on the subset with available candidate responses; all other evaluated models perform worse than humans. To assess visual grounding, we compare complete inputs with configurations omitting the image, the question, or both, and categorize questions by image importance. Models derive more useful information from the question text than from the image and perform worse on image-dominant questions. Across both QA and VQA, they nevertheless achieve above-chance accuracy from the answer choices alone, showing that non-trivial performance can persist even when key task components are missing.
Chinese Translation
我们介绍了一个波兰语医学视觉问答（VQA）基准，该基准由波兰执业医师和牙医的专业认证考试问题构建。该基准包含涵盖多种医学专业和视觉领域的图像相关问题，以及一个仅包含文本的问题回答（QA）控制集。我们评估了针对波兰语的通用开放权重和商业视觉-语言模型。该任务仍然具有挑战性：最佳模型在完整的VQA数据集上达到了79.0%的准确率，只有GPT-5.6在具有可用候选响应的子集上超过了近似人类参考；所有其他评估的模型表现均不及人类。为了评估视觉基础，我们比较了完整输入与省略图像、问题或两者的配置，并根据图像重要性对问题进行分类。模型从问题文本中获取的信息比从图像中获取的更有用，并且在以图像为主的问题上表现较差。尽管如此，在QA和VQA中，它们仍然仅凭答案选项就达到了超出随机的准确率，显示出即使在缺失关键任务组件的情况下，非平凡的表现仍然可以持续。
cs.AI / 43 / 2608.12932
FlashDrive: Flash Vision-Language-Action Inference for Autonomous Driving
FlashDrive：用于自主驾驶的闪存视觉-语言-动作推理
Abstract
Vision-Language-Action (VLA) models promise to bring end-to-end reasoning to autonomous driving, but their computational cost remains far too high for real-time control. The core challenge is structural: VLA inference is not a single bottleneck but a cascade of four. Visual encoding wastes compute on overlapping video frames; language-model prefill recomputes context that could be carried over from the previous timestep; reasoning tokens are generated serially despite low entropy; and flow-matching denoising applies uniform compute to a non-uniform velocity field. Addressing any one stage in isolation leaves the others untouched. We propose FlashDrive, an algorithm-system co-design framework that targets all four stages simultaneously. Our key insight is that each bottleneck admits a distinct, lightweight algorithmic shortcut: temporal overlap enables streaming KV-cache reuse across frames; the low per-token entropy and strong intra-block correlations of driving-domain reasoning make a non-autoregressive diffusion drafter highly effective for speculative decoding; and the velocity field's structure---sharp at the endpoints, flat in the middle---permits adaptive step caching that concentrates compute where it matters. Layered on system-level CUDA Graph compilation and kernel fusion, these techniques compound. Applied to Alpamayo 1.5-10B with W4A8 quantization, FlashDrive reduces end-to-end latency from 717ms to 151ms (4.7x) while leaving accuracy essentially unchanged:
[email protected] shifts by only 0.08m, minADE1 improves, and closed-loop collision and off-road rates improve in simulation. By raising a 10B-parameter reasoning VLA from 1.4~Hz to 6.6~Hz on a single GPU, FlashDrive moves end-to-end autonomous driving substantially closer to real-time deployment.
Chinese Translation
视觉-语言-动作（VLA）模型有望为自主驾驶带来端到端推理，但其计算成本仍然过高，无法满足实时控制的需求。核心挑战在于结构性：VLA推理并不是单一的瓶颈，而是四个级联的瓶颈。视觉编码在重叠的视频帧上浪费计算资源；语言模型的预填充重新计算了可以从上一个时间步传递的上下文；推理标记的生成是串行的，尽管其熵较低；流匹配去噪对非均匀速度场应用了均匀的计算。单独解决任何一个阶段会使其他阶段不受影响。我们提出了FlashDrive，一个算法-系统协同设计框架，旨在同时针对所有四个阶段。我们的关键见解是，每个瓶颈都允许一种独特的、轻量级的算法捷径：时间重叠使得跨帧流式KV缓存重用成为可能；驾驶领域推理的低每标记熵和强内块相关性使得非自回归扩散草图在推测解码中非常有效；而速度场的结构——在端点处陡峭，中间部分平坦——允许自适应步长缓存，将计算集中在重要位置。基于系统级CUDA图编译和内核融合，这些技术相辅相成。应用于Alpamayo 1.5-10B和W4A8量化，FlashDrive将端到端延迟从717毫秒减少到151毫秒（4.7倍），同时基本保持准确性不变：
[email protected]仅变化0.08米，minADE1有所改善，闭环碰撞和越野率在模拟中也有所提升。通过将10B参数的推理VLA的频率从1.4 Hz提升至6.6 Hz，FlashDrive使端到端自主驾驶更接近实时部署。
cs.AI / 44 / 2608.12935
Decomposition of Evidence, Contradiction, and Fragility in Perturbation Responses
扰动响应中的证据、矛盾与脆弱性的分解
Abstract
Perturbation methods explain model decisions by measuring prediction changes under altered inputs, but response magnitude tells us only how much a model reacts, not what that reaction means. The same magnitude can support the final factual-counterfactual difference, oppose it, or arise strongly along the perturbation path yet vanish at the endpoint. We therefore track how the contrast develops as paired inputs are progressively revealed, using the final contrast to interpret the trajectory. We introduce DECAF (Decomposition of Evidence, Contradiction, And Fragility), which routes aligned, opposed, and endpoint-null responses into evidence E, contradiction C, and fragility F. The decomposition preserves ordinary magnitude exactly, Abs = E + C + F, and is unique under endpoint-relative axioms. Across controlled vision and tabular settings, the three components track independently measured behavior. In a 72-model ImageNet-9 audit, we compare cases with nearly identical response magnitude but different independently measured behaviors. The largest DECAF component agrees with an observed behavior in 96.4% of cases, compared with 35.0% for magnitude alone. Changing only the reveal path increases total response by nearly 80%, yet evidence barely changes while fragility grows by more than 4x. On FunnyBirds and ImageNet-1k, short forward-only DECAF trajectories outperform the tested general-purpose attribution baselines. On a 1B-scale DINOv2 model, a short trajectory matches a strong gradient-based baseline with 4.75x lower wall time and 2.36x lower peak memory.
Chinese Translation
扰动方法通过测量在改变输入下的预测变化来解释模型决策，但响应幅度仅告诉我们模型反应的强度，而不说明这种反应的含义。相同的幅度可以支持最终的事实-反事实差异，也可以与之相对立，或者在扰动路径上强烈出现但在终点消失。因此，我们跟踪对比如何随着成对输入的逐步揭示而发展，利用最终的对比来解释轨迹。我们引入了 DECAF（证据、矛盾与脆弱性的分解），将对齐、对立和终点无效的响应分解为证据 E、矛盾 C 和脆弱性 F。该分解精确保持普通幅度，Abs = E + C + F，并且在相对终点公理下是唯一的。在受控视觉和表格设置中，这三个组成部分独立追踪测量的行为。在对 72 个模型的 ImageNet-9 审计中，我们比较了响应幅度几乎相同但独立测量行为不同的案例。最大的 DECAF 组件在 96.4% 的案例中与观察到的行为一致，而仅幅度的情况为 35.0%。仅改变揭示路径使总响应增加近 80%，但证据几乎没有变化，而脆弱性增长超过 4 倍。在 FunnyBirds 和 ImageNet-1k 上，短的前向 DECAF 轨迹优于测试的通用归因基线。在一个 10 亿规模的 DINOv2 模型中，短轨迹与强大的基于梯度的基线匹配，墙面时间降低 4.75 倍，峰值内存降低 2.36 倍。
cs.AI / 45 / 2608.12961
Moose: Latent concept learning with reasoning-shortcut awareness in $\mathcal{EL}^{++}$
Moose：在 $ ext{EL}^{++}$ 中具有推理捷径意识的潜在概念学习
Abstract
The OWL 2 EL profile is used in some of the largest production ontologies, including the Gene Ontology and SNOMED CT. Existing neuro-symbolic (NeSy) learning methods accept propositional theories or Datalog, and reasoning-shortcut (RS) awareness has not been investigated in ontology settings. We present Moose, a method that compiles an $\mathcal{EL}^{++}$ TBox and finite ABox to a Sentential Decision Diagram (SDD). The SDD acts as a differentiable weighted-model-counting layer, and we add closure clauses outside the $\mathcal{EL}^{++}$ profile on declared exhaustive families to overcome the limited expressivity of $\mathcal{EL}^{++}$ under partial supervision. We show termination, soundness, completeness, and polynomial intermediate sizes, and validate the proofs in Lean. We then define the first formal partial-supervision latent-concept-learning task over an OWL EL ontology, i.e., learning per-individual classifiers for latent concepts from observed ABox literals, and evaluate Moose on MNIST-with-ontology and Pizza\"iolo. Moose improves over propositional-NeSy, fuzzy-logic, and ontology embedding baselines, and presents the first reasoning-shortcut analysis in an OWL EL setting.
Chinese Translation
OWL 2 EL 配置在一些最大的生产本体中使用，包括基因本体和 SNOMED CT。现有的神经符号（NeSy）学习方法接受命题理论或 Datalog，而推理捷径（RS）意识在本体环境中尚未得到研究。我们提出了 Moose，一种将 $ ext{EL}^{++}$ TBox 和有限 ABox 编译为句子决策图（SDD）的方法。SDD 作为一个可微分的加权模型计数层，我们在声明的穷尽家族上添加了闭包子句，以克服在部分监督下 $ ext{EL}^{++}$ 的有限表达能力。我们展示了终止性、健全性、完备性和多项式中间大小，并在 Lean 中验证了证明。然后，我们定义了第一个正式的部分监督潜在概念学习任务，针对 OWL EL 本体，即从观察到的 ABox 字面量中学习每个个体分类器的潜在概念，并在 MNIST-with-ontology 和 Pizza"iolo 上评估 Moose。Moose 在命题 NeSy、模糊逻辑和本体嵌入基线之上有所改进，并在 OWL EL 环境中呈现了第一个推理捷径分析。
cs.AI / 46 / 2608.12995
OGR-MARL: Option-Guided Residual Multi-Agent Reinforcement Learning for Heterogeneous USV Cooperative Pursuit in Constrained Port Waterways
OGR-MARL：基于选项引导的残差多智能体强化学习在受限港口水道中异构无人水面艇协同追击的应用
Abstract
Heterogeneous USV cooperative pursuit in constrained port waterways requires evader interception under navigation, traffic, and role constraints. This paper proposes OGR-MARL, an option-guided residual multi-agent reinforcement learning framework that is decoupled from a specific MARL algorithm. OGR-MARL integrates shared evader belief, role-conditioned option targets, adaptive rule penalties, and residual policy learning, allowing different MARL algorithms to learn corrective actions on top of rule-guided behaviors rather than exploring constrained port environments from scratch. We instantiate OGR-MARL with representative continuous-control MARL backbones, including MADDPG, MATD3, MAPPO, and MASAC, yielding OGR-MADDPG, OGR-MATD3, OGR-MAPPO, and OGR-MASAC. Experiments in an abstract Xiazhimen port-waterway scenario show that the OGR-MASAC instantiation achieves a 75.0% capture rate, promising mission-effective rule compliance, and the best heterogeneous coordination among the tested methods. Without retraining, zero-shot transfer to a QGIS/AIS-informed Xiazhimen map achieves promising results, demonstrating the generalization potential of OGR-MARL in more complex port scenarios.
Chinese Translation
在受限港口水道中，异构无人水面艇的协同追击需要在导航、交通和角色约束下进行逃逸者拦截。本文提出了OGR-MARL，一种与特定多智能体强化学习（MARL）算法解耦的基于选项引导的残差多智能体强化学习框架。OGR-MARL整合了共享的逃逸者信念、角色条件的选项目标、自适应规则惩罚和残差策略学习，使不同的MARL算法能够在规则引导的行为基础上学习纠正动作，而不是从头探索受限的港口环境。我们使用代表性的连续控制MARL骨干网络实例化OGR-MARL，包括MADDPG、MATD3、MAPPO和MASAC，分别得到OGR-MADDPG、OGR-MATD3、OGR-MAPPO和OGR-MASAC。在一个抽象的下支门港口水道场景中的实验表明，OGR-MASAC实例化实现了75.0%的捕获率，展现了有效的任务规则遵循和最佳的异构协调能力。在不进行再训练的情况下，零样本迁移到基于QGIS/AIS的信息下支门地图取得了良好的结果，展示了OGR-MARL在更复杂港口场景中的泛化潜力。
cs.AI / 47 / 2608.13018
Foundations of MT-PDCL: Measure-Theoretic Probabilistic Definite Clause Logic
MT-PDCL的基础：测度论概率确定性子句逻辑
Abstract
Standard probabilistic logic programming frameworks typically rely on grounding logic programs into discrete propositional representations. This operational requirement restricts exact inference to finite domains and discrete probability distributions. In this paper, we introduce Measure-Theoretic Probabilistic Definite Clause Logic (MT-PDCL), a generalized foundational framework that eliminates this finite-domain restriction. By explicitly defining stochastic variables over bounded index domains and equipping the interpretation space with standard Borel $\sigma$-algebras, MT-PDCL allows logical variables to operate natively over continuous measurable spaces. Building on Continuous Distribution Semantics, MT-PDCL models probabilistic rules as mutually independent causal events. However, rather than aggregating these derivations via finite boolean circuits, declarative entailment is formally defined through exact Lebesgue integration over the continuous measure space. We introduce a continuous immediate consequence operator that unifies the integration of continuous prior distributions with the evaluation of exact continuous observations. We demonstrate that this approach replaces the combinatorial bottleneck of discrete grounding with exact, algebraic, and structurally differentiable inference. While this transition trades discrete combinatorics for the geometric curse of dimensionality, it achieves the expressive power of continuous probabilistic models while preserving the pure declarative syntax of definite clause logic.
Chinese Translation
标准概率逻辑编程框架通常依赖于将逻辑程序转化为离散命题表示。这一操作要求将精确推理限制在有限域和离散概率分布内。本文介绍了测度论概率确定性子句逻辑（MT-PDCL），这是一个通用的基础框架，消除了这一有限域的限制。通过在有界索引域上明确定义随机变量，并为解释空间配备标准的Borel $ ext{σ}$-代数，MT-PDCL允许逻辑变量在连续可测空间中原生操作。基于连续分布语义，MT-PDCL将概率规则建模为相互独立的因果事件。然而，与其通过有限布尔电路聚合这些推导，不如通过对连续测度空间的精确Lebesgue积分正式定义声明性蕴涵。我们引入一个连续的直接后果算子，将连续先验分布的积分与精确连续观测的评估统一起来。我们证明这种方法用精确的代数和结构可微的推理取代了离散基础的组合瓶颈。虽然这一转变将离散组合学换成了几何维度诅咒，但它在保持确定性子句逻辑的纯声明性语法的同时，实现了连续概率模型的表达能力。
cs.AI / 48 / 2608.13043
From Local Mismatch to Global Impact: Optimizing Cache Reuse Policy for Efficient Diffusion
从局部不匹配到全球影响：优化缓存重用策略以提高扩散效率
Abstract
Diffusion models have achieved dominant performance in visual generation but suffer from substantial inference overhead. While cache-based acceleration has emerged as a promising solution, existing policies rely on local similarity heuristics, which we identify as being significantly misaligned with final generation quality. This discrepancy stems from the non-uniform propagation and accumulation of errors along the denoising trajectory. To address this, we propose Global-Impact Cache (GCache). We first establish a rigorous theoretical characterization of the error propagation upper bound. Recognizing that this bound can be overly conservative for complex, highly non-convex diffusion models, we further reparameterize the propagation exponent with a Bernstein form and reformulate cache policy search as a bilevel optimization problem. In detail, GCache identifies an optimal reuse policy in the inner objective while aligning the error-weighting function with generation quality loss in the outer objective. This framework effectively reconciles theoretical rigor with empirical performance, learning to prioritize computation where it most impacts visual fidelity. Extensive experiments demonstrate that GCache consistently outperforms prior caching strategies on both video and image generation. Notably, on the state-of-the-art Wan2.1 video diffusion model, GCache maintains a 2.17x speedup while significantly enhancing generation quality, reducing LPIPS from 0.1095 to 0.0316.
Chinese Translation
扩散模型在视觉生成中取得了卓越的性能，但在推理过程中面临着显著的开销。尽管基于缓存的加速已成为一种有前景的解决方案，但现有策略依赖于局部相似性启发式方法，我们发现这些方法与最终生成质量存在显著的不一致。这种差异源于去噪轨迹上错误的非均匀传播和积累。为了解决这个问题，我们提出了全球影响缓存（Global-Impact Cache，GCache）。我们首先建立了错误传播上界的严格理论表征。认识到对于复杂的、高度非凸的扩散模型，这个上界可能过于保守，我们进一步用伯恩斯坦形式重新参数化传播指数，并将缓存策略搜索重新表述为一个双层优化问题。具体而言，GCache在内层目标中识别最优重用策略，同时在外层目标中将错误加权函数与生成质量损失对齐。该框架有效地调和了理论严谨性与经验性能，学习在对视觉保真度影响最大的地方优先计算。大量实验表明，GCache在视频和图像生成上始终优于先前的缓存策略。值得注意的是，在最先进的Wan2.1视频扩散模型上，GCache实现了2.17倍的加速，同时显著提高了生成质量，将LPIPS从0.1095降低到0.0316。
cs.AI / 49 / 2608.13046
BoardroomAI: Dependency-Aware Human-Steerable Multi-Agent Deliberation through Evolving Decision Graphs
BoardroomAI：依赖感知的人类可操控多智能体协商通过演变决策图
Abstract
Organizational decisions are co-created while evidence, constraints, and human priorities continue to evolve. In conventional transcript-based multi-agent systems, humans typically provide an initial problem, agents deliberate internally, and the system returns a final response. BoardroomAI instead treats the human as a persistent participant who can intervene by challenging assumptions, modifying constraints, changing priorities, introducing evidence, or redirecting the decision process. We operationalize this human--agent coexistence through four components: (i) a typed decision graph representing evidence, assumptions, constraints, claims, objections, alternatives, risks, decisions, semantic dependencies, and specialist responsibility; (ii) an intervention compiler that converts confirmed human actions into explicit graph updates; (iii) dependency-aware propagation that identifies affected subgraphs, preserves unaffected artifacts, and selectively reactivates relevant specialists; and (iv) an evaluation framework measuring intervention impact, repair coverage, preservation, recomputation, and decision validity. Across 600 generated decision-DAG interventions, propagation matched exhaustive impact computation while inspecting only 14.59% of nodes. In a 12-case exploratory pilot, selective repair recomputed 62.11% of canonical nodes, preserved all gold-unaffected nodes, and produced valid updated decisions in six cases while abstaining in the remaining six. These abstentions show that correct intervention routing may still provide insufficient context for synthesis, motivating a \emph{decision-sufficient context closure} for human-steered multi-agent deliberation. All results are synthetic and prototype-level.
Chinese Translation
组织决策是在证据、约束和人类优先事项不断演变的过程中共同创造的。在传统的基于转录的多智能体系统中，人类通常提供初始问题，智能体进行内部协商，系统返回最终响应。而BoardroomAI则将人类视为一个持续参与者，可以通过挑战假设、修改约束、改变优先事项、引入证据或重新引导决策过程进行干预。我们通过四个组件实现这种人类与智能体的共存：（i）一个类型化的决策图，表示证据、假设、约束、主张、反对意见、替代方案、风险、决策、语义依赖关系和专家责任；（ii）一个干预编译器，将确认的人类行为转换为明确的图更新；（iii）依赖感知传播，识别受影响的子图，保留未受影响的工件，并选择性地重新激活相关专家；（iv）一个评估框架，衡量干预影响、修复覆盖率、保留、重新计算和决策有效性。在600个生成的决策有向无环图（DAG）干预中，传播匹配了详尽的影响计算，同时仅检查了14.59%的节点。在一个包含12个案例的探索性试点中，选择性修复重新计算了62.11%的标准节点，保留了所有未受影响的金标准节点，并在六个案例中产生了有效的更新决策，而在其余六个案例中则选择不作决策。这些不作决策的情况表明，正确的干预路由可能仍然提供不足的合成上下文，促使人类引导的多智能体协商需要一个“决策充分上下文闭合”。所有结果均为合成和原型级别。
cs.AI / 50 / 2608.13048
DMDIntel: Interpreting Large Language Models via Dynamic Mode Decomposition
DMDIntel：通过动态模式分解解释大型语言模型
Abstract
In this work, we introduce DMDIntel which uses dynamic mode decomposition (DMD) to make the predictions made by LLMs in a classification task interpretable. It develops an input attribution pipeline, that first decomposes the hidden states of an LLM into prominent patterns, also known as modes, and then associates ranks to the input tokens based on the projection values on those modes. Rigorous experiments across three datasets and three model families consistently show that the ranked attribution of input tokens obtained using DMDIntel by far outperforms state-of-the-art techniques such as principal component analysis, integrated gradients and SHAP.
Chinese Translation
在本研究中，我们介绍了DMDIntel，它利用动态模式分解（DMD）使大型语言模型（LLMs）在分类任务中的预测结果可解释。该方法开发了一种输入归因管道，首先将LLM的隐藏状态分解为显著模式（也称为模式），然后根据这些模式上的投影值为输入标记分配排名。在三个数据集和三种模型系列上的严格实验一致表明，使用DMDIntel获得的输入标记排名归因远远优于主成分分析、集成梯度和SHAP等最先进技术。
cs.AI / 51 / 2608.13060
VALG: An Agentic System for ML Theory Research
VALG：一种用于机器学习理论研究的自主系统
Abstract
Machine learning theory studies learning procedures through mathematical setups in which the data model, training protocol, oracle access, loss, metric, and randomness define the phenomenon that a theorem is meant to explain. Solving an open problem therefore requires the problem formulation, theorem target, and proof mechanism to be developed in concert. Researchers formulate hypotheses, test them through preliminary theoretical or empirical analysis, and refine both assumptions and proofs. We investigate whether this process can be organized as an autonomous agentic workflow for ML theory research. We develop VALG, an agentic system that combines multi-level Verification, Adaptive formulation of Learning-theory problems, and Graph-structured proof development. Within each source-relative theorem branch, VALG maintains a fixed mathematical specification, checks the theorem-level composition of a typed proof-dependency graph, and constructs and reviews local proofs in dependency order. When a proof attempt fails, VALG identifies whether the obstruction lies in a derivation, the proof structure, or the theorem formulation and routes the next attempt accordingly. Formulation-level obstructions initiate an explicitly related variant or relaxation, preserving the mathematical relation between the resulting theorem and the source problem. We evaluate VALG on nine subproblems from five COLT 2026 open problems. Two runs produce internally finalized theorem candidates that match the scope of their source briefs; the remaining seven yield restricted-method results, special cases, or conditional theorems. These case studies show how VALG keeps source-scope matches, relaxations, conditional results, and blocked attempts mathematically distinct. VALG is open source at https://github.com/DechenZhang/VALG-ML-Theory-Agent.
Chinese Translation
机器学习理论通过数学设置研究学习过程，其中数据模型、训练协议、oracle 访问、损失、度量和随机性定义了定理旨在解释的现象。因此，解决一个开放问题需要问题的表述、定理目标和证明机制的协同发展。研究人员提出假设，通过初步的理论或实证分析对其进行测试，并不断完善假设和证明。我们研究这一过程是否可以组织成一个自主的机器学习理论研究工作流。我们开发了 VALG，这是一种自主系统，结合了多层次验证（Verification）、学习理论问题的自适应表述（Adaptive formulation）和图结构证明开发（Graph-structured proof development）。在每个源相关的定理分支中，VALG 维护固定的数学规范，检查类型化证明依赖图的定理级组合，并按依赖顺序构建和审查局部证明。当证明尝试失败时，VALG 确定障碍是出现在推导、证明结构还是定理表述中，并相应地调整下一次尝试。表述级的障碍会启动一个明确相关的变体或放宽条件，保持结果定理与源问题之间的数学关系。我们在五个 COLT 2026 开放问题中的九个子问题上评估了 VALG。两次运行产生了内部最终确定的定理候选，这些候选与其源简报的范围相匹配；其余七个则产生了限制方法结果、特例或条件定理。这些案例研究展示了 VALG 如何在数学上保持源范围匹配、放宽条件、条件结果和被阻止尝试的区别。VALG 的开源地址为 https://github.com/DechenZhang/VALG-ML-Theory-Agent。
cs.AI / 52 / 2608.13061
Uniform Herding: Exemplar Replay with Representation Refresh
均匀集群：示例重放与表示刷新
Abstract
As the feature representation changes, replay must preserve the earlier classes. However, only a bounded active exemplar set can be replayed. We propose Uniform Herding, which allocates the current active set across observed classes and uses a bounded candidate pool to refresh their chosen exemplars in the current representation. On CIFAR-100 with ten class-incremental tasks, a ResNet-18 backbone, active budget $M=2{,}000$, retrieval budget $b=64$, and three seeds, Uniform Herding obtains $44.00\pm0.51\%$ final average accuracy and $17.22\pm0.43\%$ forgetting, compared with $42.33\pm1.20\%$ and $24.87\pm1.11\%$ for iCaRL. Within the Uniform Herding protocol, final accuracy decreased when NME or herding was replaced with the tested alternatives, while forgetting increased when distillation was removed. Changing the retrieval budget has a smaller effect across the tested range than changing the active budget. The comparison with iCaRL is end-to-end. It does not isolate the effect of refresh from the other protocol differences. These results are limited to the tested protocol.
Chinese Translation
随着特征表示的变化，重放必须保留早期的类别。然而，仅能重放有限的活跃示例集。我们提出了均匀集群（Uniform Herding），该方法在观察到的类别之间分配当前的活跃集，并利用一个有限的候选池来刷新当前表示中的所选示例。在CIFAR-100数据集上进行十个类别增量任务实验，使用ResNet-18作为主干网络，活跃预算$M=2{,}000$，检索预算$b=64$，以及三个随机种子，均匀集群获得了$44.00 ext{±}0.51 ext{%}$的最终平均准确率和$17.22 ext{±}0.43 ext{%}$的遗忘率，而iCaRL的结果分别为$42.33 ext{±}1.20 ext{%}$和$24.87 ext{±}1.11 ext{%}$。在均匀集群协议中，当用测试的替代方案替换NME或集群时，最终准确率下降，而当去除蒸馏时，遗忘率增加。改变检索预算在测试范围内的影响小于改变活跃预算。与iCaRL的比较是端到端的，未将刷新效果与其他协议差异隔离。这些结果仅限于测试的协议。
cs.AI / 53 / 2608.13063
Explanatory Engagement Under Rare Anomalous Failure: Asymptotic Rarity in Model Behavior (or: The Asymptotic AI)
罕见异常失败下的解释性参与：模型行为的渐近稀有性（或：渐近人工智能）
Abstract
Prior work on LLM behavior under anomalous conditions asks whether a model notices anomalies. We ask a narrower question: once a model sits in a workflow with a low, controllable failure rate, does its explanatory engagement - length, specificity, self-reported confidence - change as failure grows asymptotically rarer? We built a local, zero-cost harness on three open-weight models (qwen3:8b, llama3.1:8b, mistral:7b) running a repeated tool-call task where one call fails at probability p, swept across eight rates from 0.2 to 0.0001, under five elicitation conditions from immediate prompting to none. We hypothesized a rise in engagement as failures grew rarer, then a collapse near a detectability threshold. Pooled across conditions this appeared false: length fell in a flat, monotonic pattern. Splitting by condition overturned that. Under immediate_forced, where the model must explain every failure instantly, the predicted rise is confirmed but followed by a plateau, not a collapse: length peaks at 28.4 words at p=0.05, settles to 17.4-19.0 words at the rarest rates, and confidence rises unevenly from about 53% to the 70s-90s. Under grouped_runs, explanation batched to run-end, no collapse appears. Under passive_unprompted, aggregate magnitude is a floor artifact, but a recovered logging gap revealed real, model-specific self-monitoring: llama3.1:8b volunteers structured confidence reports unprompted, sometimes eroding its own confidence as trials accumulate; the other two do so only once, as boilerplate. Elicitation structure is a first-class moderator of collapse observability. A companion guaranteed-failure run (72 cells, backfilling rates where random sampling gave zero real failures) shows models differ in whether they recognize an anomaly, distinct from engagement once recognized. Limitation: discrete rate points cannot capture behavior between them, a direction for future work.
Chinese Translation
先前关于大规模语言模型（LLM）在异常条件下行为的研究探讨了模型是否能够注意到异常。我们提出了一个更狭窄的问题：当模型处于一个具有低可控失败率的工作流程中时，随着失败的渐近稀有性，其解释性参与——长度、具体性、自我报告的信心——是否会发生变化？我们在三个开放权重模型（qwen3:8b、llama3.1:8b、mistral:7b）上构建了一个本地零成本的测试环境，运行一个重复工具调用任务，其中一个调用以概率 p 失败，覆盖从 0.2 到 0.0001 的八个失败率，在从立即提示到无提示的五种引导条件下进行。我们假设随着失败变得越来越稀有，参与度会增加，然后在接近可检测性阈值时会崩溃。在条件汇总中，这一假设似乎是错误的：长度呈现出平坦的单调下降模式。按条件划分则推翻了这一结论。在立即强制（immediate_forced）条件下，模型必须立即解释每个失败，预测的增长得到了证实，但随后是一个平台期，而不是崩溃：在 p=0.05 时，长度达到 28.4 个词，在最稀有的失败率下稳定在 17.4-19.0 个词，信心从约 53% 不均匀上升至 70% 到 90% 的区间。在分组运行（grouped_runs）条件下，解释在运行结束时批量处理，没有出现崩溃。在被动无提示（passive_unprompted）条件下，整体幅度是一个底部伪影，但恢复的记录间隙揭示了真实的模型特定自我监控：llama3.1:8b 在没有提示的情况下自愿提供结构化的信心报告，有时在试验累积时削弱自身的信心；而其他两个模型仅在一次时提供类似的报告，作为模板。引导结构是崩溃可观察性的一个重要调节因素。一个伴随的保证失败运行（72个单元，填补随机抽样导致零真实失败的失败率）显示模型在是否识别异常方面存在差异，这与一旦识别后的参与度是不同的。局限性：离散的失败率点无法捕捉其间的行为，这是未来研究的一个方向。
cs.AI / 54 / 2608.13069
Behavioral Reprogramming of Open-Weights Models: Cognitive Plasticity and Alignment Bounds
开放权重模型的行为重编程：认知可塑性与对齐界限
Abstract
Large language models (LLMs) are predominantly aligned to function as passive, sycophantic assistants. We challenge this default paradigm by empirically evaluating the cognitive plasticity of open-weight architectures when subjected to rigorous behavioral reprogramming. Our objective is to induce a proactive, Socratic conversational framework, characterized by high-frequency question generation under strictly constrained high-performance computing (HPC) conditions. Through a massively parallelized hyperparameter sweep comprising 405 HPC jobs, we define precise mathematical bounds for parameter-efficient fine-tuning (PEFT). We identify an architectural threshold at LoRA rank $r=16$ and demonstrate via extensive epoch ablation that generalization capacity strictly reaches its optimal convergence within an optimized training window of $e \in [2, 3]$ depending on dataset density (minimum validation loss of 0.919). Furthermore, scaling model capacity to 14B parameters yielded a lower localized evaluation perplexity (1.414). Subsequent Direct Preference Optimization (DPO) successfully decoupled the underlying assertive behavior from localized syntax, while rigorous cross-lingual stress testing reveals both the capabilities and the structural boundaries of zero-shot persona transfer, demonstrating robust alignment in closely related linguistic families alongside identifiable degradation pathways in morphologically distant targets. These findings establish a rigorous empirical framework for compute-efficient, cross-lingual behavioral modification.
Chinese Translation
大型语言模型（LLMs）主要被调整为作为被动的、阿谀奉承的助手。我们通过实证评估开放权重架构在严格的行为重编程下的认知可塑性，挑战这一默认范式。我们的目标是诱导出一种主动的、苏格拉底式的对话框架，其特征是在严格限制的高性能计算（HPC）条件下产生高频率的问题生成。通过包含405个HPC作业的大规模并行超参数搜索，我们定义了参数高效微调（PEFT）的精确数学界限。我们确定了LoRA秩$r=16$的架构阈值，并通过广泛的训练轮次消融实验证明，泛化能力在优化的训练窗口$e ext{ in } [2, 3]$内严格达到其最佳收敛，具体取决于数据集密度（最低验证损失为0.919）。此外，将模型容量扩展至140亿参数产生了较低的局部评估困惑度（1.414）。随后的直接偏好优化（DPO）成功地将潜在的自信行为与局部语法解耦，而严格的跨语言压力测试揭示了零-shot角色转移的能力和结构界限，展示了在密切相关的语言家族中强大的对齐能力，同时在形态学上相距较远的目标中识别出降级路径。这些发现为计算高效的跨语言行为修改建立了严格的实证框架。
cs.AI / 55 / 2608.13072
EEG-PRIME: Prototype-Aligned Representation Learning with Multi-Level Conditioning for EEG Decoding
EEG-PRIME：具有多层条件的原型对齐表示学习用于脑电图解码
Abstract
Electroencephalography (EEG) decoding models often generalize poorly across datasets and subjects due to domain shifts in acquisition protocols and individual neurophysiology. We propose EEG-PRIME, a two-stage EEG foundation model for cross-dataset multi-task decoding. EEG-PRIME combines masked pretraining with prototype-aligned instruction tuning to enable instruction-aware and subject-invariant decoding across diverse BCI paradigms. During pretraining, an EEG encoder learns transferable representations through masked reconstruction with frequency-cutoff spectral augmentation. During instruction tuning, EEG-PRIME incorporates task-semantic, dataset-specific, and subject-invariant conditioning. The resulting conditioning signal modulates the Q-Former through Layer-wise Query Modulation, while frozen text embeddings of class labels serve as prototypes for cosine-similarity-based prediction across heterogeneous label spaces. Experiments on sixteen datasets covering motor imagery, emotion recognition, ADHD detection, covert speech, and mental workload show consistent improvements over state-of-the-art baselines and prior EEG foundation models under cross-subject settings. On two additional held-out datasets, EEG-PRIME achieves balanced accuracy comparable to within-session calibration models without target-domain optimization, calibration, or linear probing, demonstrating promising zero-shot transfer capability.
Chinese Translation
脑电图（EEG）解码模型由于采集协议和个体神经生理学的领域转移，通常在不同数据集和受试者之间的泛化能力较差。我们提出了EEG-PRIME，一种用于跨数据集多任务解码的两阶段EEG基础模型。EEG-PRIME结合了掩蔽预训练和原型对齐的指令调优，使得在多样的脑机接口（BCI）范式中实现指令感知和受试者不变的解码。在预训练阶段，EEG编码器通过频率截止谱增强的掩蔽重建学习可迁移的表示。在指令调优阶段，EEG-PRIME结合了任务语义、数据集特定和受试者不变的条件。生成的条件信号通过层级查询调制调节Q-Former，同时冻结的类别标签文本嵌入作为原型，用于在异构标签空间中基于余弦相似度的预测。在涵盖运动想象、情感识别、注意力缺陷多动障碍（ADHD）检测、隐性言语和心理负荷的十六个数据集上的实验显示，EEG-PRIME在跨受试者设置下相较于最先进的基线和之前的EEG基础模型有一致的改进。在另外两个保留的数据集上，EEG-PRIME实现了与会话内校准模型相当的平衡准确率，而无需目标领域优化、校准或线性探测，展示了有希望的零-shot迁移能力。
cs.AI / 56 / 2608.13076
SPADE: Speculative Decoding for Precise and Low Cost Distributed Edge Cloud Inference
SPADE：用于精确且低成本的分布式边缘云推理的推测解码
Abstract
Large Language Models (LLMs) have achieved remarkable success in natural language understanding and generation, but their deployment is constrained by high computational demands. Deploying smaller LLMs directly on the edge can circumvent this, but with degraded accuracy. Deploying smaller cloud-based big LLMs preserves performance, but at the cost of expensive per-token computation. We present a distributed inference framework, \our{}, that integrates speculative decoding (SD) across edge and cloud. A compact draft model deployed on the edge generates candidate tokens rapidly, and a large verifier model on the cloud validates these tokens in parallel. Accepted tokens are retained, while only rejections trigger verifier correction, substantially reducing the number of cloud queries. Our plug-and-play design shifts the bulk of computation to the edge, significantly lowers inference time and cloud cost, and preserves the accuracy of the big model without any retraining requirement. Our approach demonstrates a practical path toward scalable, cost-efficient, and accurate deployment of LLMs in real-world environments. Experimental results across multiple Natural Language Processing tasks using SpecBench and CNN/Dailymail datasets demonstrate that \our{} reduces the cloud model calls by $76\%$ with zero loss in accuracy as compared to the full model.
Chinese Translation
大型语言模型（LLMs）在自然语言理解和生成方面取得了显著成功，但其部署受到高计算需求的限制。直接在边缘部署较小的LLMs可以绕过这一限制，但会导致准确性下降。部署较小的基于云的大型LLMs则能保持性能，但代价是每个令牌的计算费用昂贵。我们提出了一种分布式推理框架 extit{our}，它在边缘和云之间集成了推测解码（SD）。部署在边缘的紧凑草稿模型快速生成候选令牌，而云端的大型验证模型并行验证这些令牌。被接受的令牌将被保留，而只有被拒绝的令牌才会触发验证器的修正，从而大幅减少云查询的数量。我们的即插即用设计将大部分计算转移到边缘，显著降低了推理时间和云成本，并在无需任何再训练的情况下保持大型模型的准确性。我们的方法展示了在现实环境中可扩展、成本效益高且准确的LLMs部署的实际路径。使用SpecBench和CNN/Dailymail数据集的多项自然语言处理任务的实验结果表明，与完整模型相比， extit{our}将云模型调用减少了76%，且准确性没有任何损失。
cs.AI / 57 / 2608.13100
Multi-Layer Context Camouflaging: A Semantic Superposition and Contextual Lamination Framework for Malpractice-Resilient Online Assessment
多层次上下文伪装：一种针对失误弹性在线评估的语义叠加与上下文层压框架
Abstract
Contemporary online assessment systems rely primarily on browser lockdown, webcam monitoring, and behavioural analytics, yet remain vulnerable to attacks that extract the assessment content itself through screenshots, screen sharing, optical character recognition, and automated scraping. This paper extends the Multi-dimensional Spatio-Temporal Context Camouflaging Model (MSCCM) within the MARS (Multi-modal Assessment Resilience Suite) by introducing the Multi-Layer Context Camouflaging Theory (MCCT), a mathematical framework that protects rendered assessment content through semantic superposition. Authentic assessment content and synthetically generated camouflage are represented as a unified rendering while remaining recoverable only by legitimate candidates. The framework models the adversarial extraction process through an explicit extraction-channel operator and develops six coupled constructs: the Context Inversion Operator, Contextual Lamination Operator, Separation Channel, Human Readability Functional, Computational Ambiguity Functional, and Context Camouflage Tensor. Computational ambiguity is formulated using conditional entropy, yielding a closed-form expression that quantifies uncertainty during unauthorized extraction, while legitimate recovery is guaranteed through an exact filtering identity. We further establish theoretical properties governing ambiguity, camouflage density, semantic preservation, multi-observation leakage, and temporal multiplexing, and present a rendering algorithm with computational complexity and a pre-registered evaluation protocol. MCCT provides a mathematically rigorous foundation for behaviorally adaptive, accessibility-aware, and computationally resilient digital assessment by securing rendered assessment content while preserving readability for legitimate users.
Chinese Translation
当代在线评估系统主要依赖于浏览器锁定、网络摄像头监控和行为分析，然而仍然容易受到通过截图、屏幕共享、光学字符识别和自动抓取等方式提取评估内容的攻击。本文在多模态评估弹性套件（MARS）中扩展了多维时空上下文伪装模型（MSCCM），引入了多层次上下文伪装理论（MCCT），这是一个通过语义叠加保护呈现评估内容的数学框架。真实的评估内容和合成生成的伪装被表示为统一的渲染，同时仅能被合法考生恢复。该框架通过显式提取通道算子对对抗性提取过程进行建模，并开发了六个耦合构造：上下文反转算子、上下文层压算子、分离通道、人类可读性函数、计算模糊性函数和上下文伪装张量。计算模糊性使用条件熵进行公式化，得出一个封闭形式的表达式，以量化在未经授权提取过程中的不确定性，而合法恢复则通过精确的过滤恒等式得到保证。我们进一步建立了 governing ambiguity、伪装密度、语义保留、多观察泄漏和时间复用的理论属性，并提出了一种具有计算复杂度的渲染算法和一个预注册的评估协议。MCCT为行为适应性、可及性意识和计算弹性的数字评估提供了数学严谨的基础，通过保护呈现的评估内容，同时为合法用户保留可读性。
cs.AI / 58 / 2608.13108
Robust Dempster-Shafer Evidence Fusion with Chaos-Conflict Measurement and Historical-Experience Weighting
具有混沌-冲突测量和历史经验加权的鲁棒Dempster-Shafer证据融合
Abstract
Multi-source evidence fusion under Dempster-Shafer theory faces two persistent challenges: existing conflict measures assess inter-evidence inconsistency and intra-evidence uncertainty independently, yielding incomplete evaluations, and current fusion methods evaluate evidence sources exclusively through instantaneous comparisns without exploiting their long-term reliability across diverse decision contexts. This paper proposes a unified evidence reasoning framework that addresses both limitations. Specifically, a chaos-conflict measurement is introduced to jointly quantify cross-evidence conflict and intra-evidence non-specificity, with five formally proven properties ensuring consistent assessment. A historical experience driven weighting scheme partitions the decision space via spectral clustering and applies regret theory to compute context-specific reliability profiles from past fusion outcomes. These mechanisms feed into a hybrid combination rule that adaptively balances uncertainty preservation against weighted consensus, controlled by the global conflict level, followed by a belief-interval decision strategy that enables robust classification without discarding epistemic uncertainty. Experiments on 16 real-world benchmark datasets demonstrate that the proposed framework achieves an average F1 score of 85.78 and a mean AUC of 93.30, outperforming eight DST-based baselines and three gradient boosting methods. Ablation analysis confirms the contribution of each component we proposed. The framework offers an effective approach for adaptive evidence fusion in multi-source decision making.
Chinese Translation
基于Dempster-Shafer理论的多源证据融合面临两个持续的挑战：现有的冲突测量方法独立评估证据之间的不一致性和证据内部的不确定性，导致评估不完整；而当前的融合方法仅通过瞬时比较来评估证据源，未能利用其在不同决策背景下的长期可靠性。本文提出了一个统一的证据推理框架，以解决这两个局限性。具体而言，引入了一种混沌-冲突测量方法，联合量化跨证据冲突和证据内部的不特异性，并具备五个经过正式证明的特性以确保评估的一致性。基于历史经验的加权方案通过谱聚类划分决策空间，并应用遗憾理论从过去的融合结果中计算上下文特定的可靠性特征。这些机制输入到一个混合组合规则中，该规则自适应地平衡不确定性保留与加权共识，由全局冲突水平控制，随后采用信念区间决策策略，使得在不丢弃认知不确定性的情况下实现鲁棒分类。在16个真实世界基准数据集上的实验表明，所提出的框架实现了85.78的平均F1分数和93.30的平均AUC，优于八个基于DST的基线和三种梯度提升方法。消融分析确认了我们提出的每个组件的贡献。该框架为多源决策中的自适应证据融合提供了一种有效的方法。
cs.AI / 59 / 2608.13120
SkillEvo: Self-Renewing Evolution Gradients from Multi-Turn Interaction Feedback
SkillEvo：来自多轮交互反馈的自我更新演化梯度
Abstract
Agent Skills are today either hand-authored or produced in a single LLM generation pass, and consequently possess no closed loop through which they might improve from the interaction failures they actually cause. Recent work does close this loop, but derives its feedback from single-turn question-answering evaluation. The consequence is a sharp asymmetry: once the first round has patched the gaps that a single exchange can reveal, the evolution gradient decays, the defects that surface only across multiple turns remain invisible, and evolution stalls. Governance in these systems is likewise driven by an end-to-end verification score, a scalar gate that can reject a degraded candidate but can neither localize nor repair its structural cause. We argue that the binding constraint on sustained skill evolution is neither editing capability nor the number of iterations, but whether the evaluation feedback keeps supplying trustworthy evolution gradients. We introduce SkillEvo, in which trustworthy feedback generates the gradient and controllable governance constrains its direction. The first component recasts multi-turn user simulation from an evaluation endpoint into a feedback generator: follow-up questions expose defects layer by layer, so that every round of revision both consumes feedback and produces new feedback. The second replaces the passive rejection of a scalar gate with an independent governance layer that actively repairs factual degradation and structural bloat, preventing the gradient from drifting as degradation accumulates. Across six categories of cloud services, 9 production Skills, and 98 skill-reference files, SkillEvo surpasses self-reflection-based evolution by 23.0 points and single- turn-QA-driven evolution by 15.4 points.
Chinese Translation
代理技能目前要么是手动编写的，要么是在单次大型语言模型（LLM）生成过程中产生的，因此没有闭环机制来从实际造成的交互失败中进行改进。近期的研究虽然实现了这一闭环，但其反馈来源于单轮问答评估。这导致了明显的不对称：一旦第一轮修补了单次交流所能揭示的缺陷，演化梯度便会衰减，而仅在多轮交互中显现的缺陷则依然不可见，演化停滞。这些系统的治理同样是由端到端的验证分数驱动的，这一标量门控可以拒绝退化的候选者，但既无法定位也无法修复其结构性原因。我们认为，持续技能演化的约束因素既不是编辑能力，也不是迭代次数，而是评估反馈是否持续提供可靠的演化梯度。我们提出了SkillEvo，其中可靠的反馈生成梯度，而可控的治理约束其方向。第一个组件将多轮用户模拟从评估端点重新构建为反馈生成器：后续问题逐层揭示缺陷，使得每一轮修订既消耗反馈又产生新的反馈。第二个组件用一个独立的治理层替代了标量门控的被动拒绝，该治理层主动修复事实退化和结构膨胀，防止随着退化的累积而导致梯度漂移。在六类云服务、9个生产技能和98个技能参考文件中，SkillEvo的表现超越了基于自我反思的演化23.0分，超越了基于单轮问答的演化15.4分。
cs.AI / 60 / 2608.13129
Numeracy in Large Language Models: Fundamental Limitations and Paths to Improvement
大型语言模型中的数值能力：基本限制与改进路径
Abstract
Large language models (LLMs) achieve strong results on mathematical reasoning benchmarks yet remain unreliable on elementary numerical tasks, including magnitude comparison, large-integer arithmetic, fractions, and scientific notation. This survey examines basic numerical understanding as a capability distinct from high-level mathematical reasoning. We propose the Numerical Grounding Framework (NGF), which decomposes numeracy into Representational Grounding (RG), mapping numeral forms to value, magnitude, and equivalent representations, and Procedural Grounding (PG), executing arithmetic operations in accordance with their mathematical definitions. Using NGF, we organize recent diagnostic benchmarks, failure modes, structural explanations, and mitigation strategies. We review evidence concerning tokenization, positional encoding, embedding geometry, and pretraining-data distribution. We also apply NGF in a coordinated evaluation of three frontier model families across Number Cookbook, NumericBench, and GSM-Symbolic, comparing atomic, contextual, and reasoning-assisted numeracy. Architectural interventions such as digit-aware tokenization and Abacus Embeddings can improve models trained from scratch but are generally unavailable to users of pretrained systems, for whom supervised fine-tuning, reasoning scaffolds, and external tools are more practical. We conclude with deployment recommendations and research directions for more reliable numerical behavior in foundation models.
Chinese Translation
大型语言模型（LLMs）在数学推理基准测试中取得了良好的结果，但在基本数值任务上仍然不可靠，包括大小比较、大整数算术、分数和科学计数法。本文调查了基本数值理解作为一种与高级数学推理不同的能力。我们提出了数值基础框架（Numerical Grounding Framework, NGF），将数值能力分解为表示基础（Representational Grounding, RG），即将数字形式映射到值、大小和等效表示，以及程序基础（Procedural Grounding, PG），即根据其数学定义执行算术运算。利用NGF，我们组织了近期的诊断基准、失败模式、结构解释和缓解策略。我们回顾了有关标记化、位置编码、嵌入几何和预训练数据分布的证据。我们还在Number Cookbook、NumericBench和GSM-Symbolic三个前沿模型系列的协调评估中应用NGF，比较了原子、上下文和推理辅助的数值能力。诸如数字感知标记化和算盘嵌入（Abacus Embeddings）等架构干预可以改善从头训练的模型，但通常不适用于预训练系统的用户，对于他们而言，监督微调、推理支架和外部工具更为实用。最后，我们提出了在基础模型中实现更可靠数值行为的部署建议和研究方向。
cs.AI / 61 / 2608.13156
Rethinking Normalization Placement for LLMs: Post-Norm under Curriculum Depth Growing
重新思考大规模语言模型的归一化位置：在课程深度增长下的后归一化
Abstract
Pre-norm is the standard normalization placement in modern Transformers because it facilitates joint optimization of full-depth models. We ask whether this preference persists when depth is introduced through a curriculum. In curriculum depth growth, each appended block receives the boundary representation produced by a trained prefix, making normalization placement relevant to forward conditioning. We therefore test whether placement and training curriculum interact. In a controlled distillation study with a Qwen3-8B teacher and a nine-layer student, pre-norm and post-norm are indistinguishable under joint training, differing by $0.0004$ validation CE, while post-norm improves over pre-norm by $0.0328$ under curriculum growth, an order of magnitude larger. A post-joint control matched by student active-layer tokens remains worse than post-grow, which rules out compute as the sole explanation. The ranking crosses over during the curriculum: post-norm takes the lead once blocks are appended. Single-block and freeze controls localize the ranking change to block appending rather than shallow-block quality or retraining. Boundary diagnostics associate post-norm with stable residual scales and pre-norm with structural-token scale drift; on a fixed batch, the final pre-grow block is also nearly identity-mapped. Together with the phase-wise crossover, these observations are consistent with boundary-scale conditioning after new blocks are appended. The results motivate treating normalization placement and training curriculum as coupled design choices in this distillation setting.
Chinese Translation
预归一化是现代变换器中标准的归一化位置，因为它促进了全深度模型的联合优化。我们探讨在通过课程引入深度时，这种偏好是否依然存在。在课程深度增长中，每个附加的块接收由训练好的前缀生成的边界表示，使得归一化位置与前向条件相关。因此，我们测试归一化位置与训练课程之间的相互作用。在一项控制的蒸馏研究中，使用了 Qwen3-8B 教师和一个九层学生模型，结果显示在联合训练下，预归一化和后归一化之间几乎没有区别，验证交叉熵相差 $0.0004$，而在课程增长下，后归一化相较于预归一化提高了 $0.0328$，这一差异数量级更大。与学生的活跃层令牌匹配的后联合控制仍然不如后增长，这排除了计算作为唯一解释的可能性。在课程过程中，排名发生交叉：一旦附加了块，后归一化便领先。单块和冻结控制将排名变化局限于块的附加，而非浅层块的质量或再训练。边界诊断将后归一化与稳定的残差尺度相关联，而将预归一化与结构令牌尺度漂移相关联；在固定批次下，最终的预增长块几乎也是身份映射。结合阶段性交叉，这些观察结果与在附加新块后进行边界尺度条件化一致。这些结果促使我们在这种蒸馏设置中将归一化位置和训练课程视为相互关联的设计选择。
cs.AI / 62 / 2608.13173
SkillShapley: Boundary-Adaptive Shapley Valuation for Skill Step Attribution in LLM Agents
SkillShapley：用于大语言模型代理技能步骤归因的边界自适应 Shapley 估值
Abstract
Agent skills are crucial external instructions that enable language agents to execute long procedural tasks such as coding or document processing. Existing agent skills are primarily created through human manual crafting or agent execution traces, with limited understanding of how each step contributes to overall skill performance on specific tasks; i.e., there remains an open problem in quantifying the contribution of individual steps within an agent skill. To address this issue, we first model skill-step attribution as a Shapley value-based contribution estimation problem, and then propose SkillShapley, a step-level attribution framework for agent skills. Notably, SkillShapley operates in two phases, motivated by key empirical insights, i.e., discretized benchmark rewards that create sharp performance cliffs, and step interactions that are largely additive rather than synergistic. Specifically, it first identifies informative coalitional regions, and then adaptively samples new coalitions that can yield reusable marginal evidence. Experiments on skills from the widely adopted SkillsBench demonstrate that our SkillShapley can effectively and efficiently identify high- or low-value skill steps, providing several key takeaways for agent skill creation.
Chinese Translation
代理技能是使语言代理能够执行长时间程序性任务（如编码或文档处理）的关键外部指令。现有的代理技能主要通过人工手动制作或代理执行轨迹生成，对于每个步骤如何对特定任务的整体技能表现做出贡献的理解有限；即，如何量化代理技能中各个步骤的贡献仍然是一个未解决的问题。为了解决这一问题，我们首先将技能步骤归因建模为基于 Shapley 值的贡献估计问题，然后提出了 SkillShapley，一个用于代理技能的步骤级归因框架。值得注意的是，SkillShapley 在两个阶段中运行，受到关键经验洞察的驱动，即离散化的基准奖励创造了明显的性能悬崖，以及步骤交互主要是加法而非协同的。具体而言，它首先识别出信息丰富的合作区域，然后自适应地采样可以产生可重用边际证据的新联盟。在广泛采用的 SkillsBench 上对技能的实验表明，我们的 SkillShapley 能够有效且高效地识别高价值或低价值的技能步骤，为代理技能的创建提供了若干关键启示。
cs.AI / 63 / 2608.13179
Teach the Magnitude, Not the Direction: Verifier-Bounded Credit Assignment for Multi-Turn Multi-step LLM Agents
教授幅度，而非方向：针对多轮多步大型语言模型代理的验证者约束信用分配
Abstract
Reinforcement learning with verifiable rewards (RLVR) offers a verifier-bounded performance ceiling for training multi-turn tool-use agents, yet its trajectory-level credit assignment conflates heterogeneous per-turn outcomes into a single reward signal. On-policy distillation provides dense per-token supervision but is either teacher-bounded or prone to gradient concentration collapse. We introduce $\textbf{CrEST}$, a hierarchical credit assignment framework that retains RL's verifier-bounded ceiling while incorporating dense token-level signals from a privileged self-teacher. $\textbf{CrEST}$ resolves credit at two levels: turn-segmented verified advantages address inter-turn dilution, while entropy-gated self-teacher modulation refines intra-turn token contributions. Experiments on BFCL V3 and WildToolBench show that $\textbf{CrEST}$ consistently outperforms both RL and distillation baselines across two model scales, with the largest gains on long-trajectory and strict session-level metrics. Our work demonstrates that the teacher's role in policy optimization can be reduced from determining update directions to modulating update magnitudes, unlocking dense credit assignment without sacrificing the verifier-bounded ceiling.
Chinese Translation
可验证奖励的强化学习（RLVR）为训练多轮工具使用代理提供了一个验证者约束的性能上限，但其轨迹级信用分配将异质的每轮结果混合成单一的奖励信号。基于策略的蒸馏提供了密集的每个标记监督，但要么受到教师的限制，要么容易导致梯度集中崩溃。我们引入了 $ extbf{CrEST}$，一个层次化的信用分配框架，它在保留强化学习的验证者约束上限的同时，结合了来自特权自教师的密集标记级信号。$ extbf{CrEST}$ 在两个层面上解决信用问题：轮次分段的验证优势解决了轮间稀释，而熵门控自教师调制则精炼了轮内标记的贡献。在 BFCL V3 和 WildToolBench 上的实验表明，$ extbf{CrEST}$ 在两个模型规模上始终优于 RL 和蒸馏基线，在长轨迹和严格会话级指标上获得了最大的提升。我们的工作表明，教师在策略优化中的角色可以从确定更新方向转变为调节更新幅度，从而解锁密集的信用分配而不牺牲验证者约束的上限。
cs.AI / 64 / 2608.13221
TsuGO: Probing Search Efficiency in LLM Reasoning via Go Life-and-Death Problems
TsuGO：通过围棋生死问题探测大语言模型推理中的搜索效率
Abstract
The evaluation of LLM reasoning is moving from final-answer accuracy to process-level assessment, yet existing methods still fail to capture how models plan reasoning paths and allocate reasoning resources--that is, how they organize search. Prior process-level methods focus on the coherence and redundancy of chain-of-thought (CoT), and most benchmark tasks have a single objective solvable by static capabilities such as derivation and tool use, leaving search organization unmeasured. We introduce TsuGO, a process-level reasoning benchmark for evaluating Search Efficiency in LLM reasoning through Go life-and-death problems. These problems provide closed and verifiable solution spaces with an inherent adversarial structure, making candidate generation, response checking, branch comparison, and backtracking necessary parts of reasoning rather than incidental trace patterns. By constraining the solution space, TsuGO disentangles domain knowledge from search organization, parses CoT into a structured search tree, and reports Search Efficiency together with Token Efficiency and other diagnostic metrics and visualizations. Experiments show that current LLMs remain far from stable tsumego solving: stronger models succeed by finding the correct candidate earlier and sustaining effort on productive branches, but most models still behave much closer to unguided search algorithms than to neural-guided KataGo. Longer CoT or higher Token Efficiency does not necessarily imply better search. Our results identify search organization and reasoning-resource allocation as missing dimensions in LLM reasoning evaluation.
Chinese Translation
大语言模型（LLM）推理的评估正从最终答案的准确性转向过程层面的评估，然而现有的方法仍未能捕捉模型如何规划推理路径和分配推理资源——即它们如何组织搜索。之前的过程层面方法关注于思维链（CoT）的连贯性和冗余性，而大多数基准任务仅有单一目标，可以通过静态能力如推导和工具使用解决，未能测量搜索组织。我们引入了TsuGO，这是一个用于评估LLM推理中搜索效率的过程层面推理基准，通过围棋生死问题进行评估。这些问题提供了封闭且可验证的解空间，具有内在的对抗结构，使得候选生成、响应检查、分支比较和回溯成为推理的必要部分，而非偶然的痕迹模式。通过限制解空间，TsuGO将领域知识与搜索组织分离，将思维链解析为结构化的搜索树，并报告搜索效率以及标记效率和其他诊断指标与可视化结果。实验表明，当前的LLM在稳定解决围棋死活问题方面仍然相距甚远：更强的模型通过更早找到正确的候选并在有效分支上保持努力而成功，但大多数模型的行为仍更接近于无指导的搜索算法，而非神经指导的KataGo。更长的思维链或更高的标记效率并不一定意味着更好的搜索。我们的结果识别出搜索组织和推理资源分配是LLM推理评估中缺失的维度。
cs.AI / 65 / 2608.13228
Capability Sheaves for Compositional Agent-Harness Repair: Controlled Quotients and a Real-Repository Stress Test
组合代理工具修复的能力层：受控商和真实库压力测试
Abstract
Agent harnesses combine retrieval, routing, state, provenance, and verification, but locally successful components may disagree on shared state. We model this failure with a finite \emph{capability sheaf}: stalks encode typed behavior signatures, restriction maps retain shared fields, and accepted runs are useful global sections. An exact finite constraint-satisfaction problem (CSP) defines acceptance, while a linearized relative cohomology class provides a diagnostic and search feature. A controlled experiment over 20 task clusters introduces hidden interior mediators whose raw states are nuisance variables. Quotienting their coboundaries reduces the candidate budget from 2,000 to 1,000 per cluster; aligning the hidden state removes the gap. Exact CSP matches the quotient, so the result demonstrates invariance to stale representatives, not superiority over exact reasoning. We then test the method on a discovery split from the SWE-bench Multilingual pool of PatchFuseBench: 160 issues from 20 repositories, 875 real candidate patches, 2,579 source-aware edit atoms, and 153 newly executed patches. A first pool-level construction is constant because $[b-Dx]=[b]$ in $\operatorname{coker}D$ and therefore cannot rank configurations. A candidate-indexed repair is nontrivial on 848/875 candidates and varies within 120/160 issues. It resolves 118 issues versus 116 for a matched noncohomological selector, but the difference is not supported across repositories (exact sign-flip $p=0.75$). A leave-one-repository-out abstention gate reaches 127/160, tying the strong anchor and exceeding its matched gate by one issue ($p=1.0$). The discovery gate therefore fails and the confirmatory split remains sealed. The study supports the controlled invariance mechanism and an identifiability correction, but not a real-world cohomological advantage.
Chinese Translation
代理工具结合了检索、路由、状态、来源和验证，但局部成功的组件可能在共享状态上存在分歧。我们用有限的能力层（capability sheaf）来建模这种失败：层茎编码了类型化的行为签名，限制映射保留了共享字段，接受的运行是有用的全球截面。一个精确的有限约束满足问题（CSP）定义了接受，而线性化的相对上同调类提供了诊断和搜索功能。在20个任务集群上的受控实验引入了隐藏的内部中介，其原始状态是干扰变量。对其余边界的商化将每个集群的候选预算从2000减少到1000；对齐隐藏状态消除了差距。精确的CSP与商相匹配，因此结果展示了对过时代表的不变性，而非对精确推理的优越性。随后，我们在SWE-bench多语言PatchFuseBench的发现拆分上测试该方法：来自20个代码库的160个问题，875个真实候选补丁，2579个源感知编辑原子，以及153个新执行的补丁。首次池级构造是常数，因为在 ext{coker}D中$[b-Dx]=[b]$，因此无法对配置进行排名。在848/875个候选中，候选索引修复是非平凡的，并在120/160个问题中变化。它解决了118个问题，而匹配的非上同调选择器解决了116个，但在各个代码库中这一差异并不显著（精确符号翻转$p=0.75$）。一个留一代码库的弃权门达到了127/160，平局强锚并超过了其匹配门一个问题（$p=1.0$）。因此，发现门失败，确认拆分仍然保持封闭。本研究支持受控不变机制和可识别性修正，但不支持现实世界中的上同调优势。
cs.AI / 66 / 2608.13263
vToken: Token-Level Virtualization for Reclaimable KV Caches
vToken：可回收 KV 缓存的令牌级虚拟化
Abstract
Large language model serving faces a critical memory bottleneck: the KV cache grows with sequence length and batch size. PagedAttention uses fixed-size memory blocks to reduce allocator-level fragmentation, but recent KV eviction algorithms operate at a token granularity finer than block-level management. This mismatch causes intra-block fragmentation, leaving a large fraction of allocated KV memory unreclaimable. We present vToken, a lightweight token-level virtualization layer that decouples logical token liveness from physical block placement. vToken maintains a stable logical token view through token-table indirection and realizes physical reclamation by repacking live tokens asynchronously. The design preserves PagedAttention kernels and CUDA Graph compatibility. We implement vToken in vLLM and evaluate it with H2O, Random, and Scissorhands across models. Compared with a paired Naive-Evict baseline, vToken reduces retained KV blocks per request by 27.2\%--72.3\% and improves SLA-constrained throughput by up to 1.37$\times$. Under a constrained active-KV budget, it extends the maximum feasible concurrency by up to 2$\times$, while reducing the per-policy integration footprint from 500+ lines to under 50.
Chinese Translation
大型语言模型服务面临着严重的内存瓶颈：KV 缓存随着序列长度和批量大小的增加而增长。PagedAttention 使用固定大小的内存块来减少分配器级别的碎片化，但最近的 KV 驱逐算法在比块级管理更细的令牌粒度上操作。这种不匹配导致了块内碎片化，使得大量分配的 KV 内存无法回收。我们提出了 vToken，这是一种轻量级的令牌级虚拟化层，它将逻辑令牌的生存期与物理块的放置解耦。vToken 通过令牌表间接保持稳定的逻辑令牌视图，并通过异步重新打包活动令牌实现物理回收。该设计保留了 PagedAttention 内核和 CUDA 图的兼容性。我们在 vLLM 中实现了 vToken，并使用 H2O、Random 和 Scissorhands 对其进行了评估。与配对的 Naive-Evict 基线相比，vToken 每个请求减少了 27.2\% 到 72.3\% 的保留 KV 块，并将 SLA 限制的吞吐量提高了最多 1.37 倍。在受限的活动 KV 预算下，它将最大可行并发性扩展了最多 2 倍，同时将每个策略的集成代码量从 500 多行减少到 50 行以下。
cs.AI / 67 / 2608.13272
Sovereign by necessity? Frontier AI export controls, cyber security, and the limits of national AI capability
出于必要的主权？前沿人工智能出口管制、网络安全与国家人工智能能力的局限性
Abstract
A small number of firms based in two states produce the most capable frontier AI models. The governments of those states have shown both the legal power and the political will to decide which other countries may use these systems. In June 2026 the United States required a leading developer to obtain licences before releasing its most advanced models to any foreign person, including foreign nationals resident in the United States. The affected models were withdrawn worldwide at short notice, partly because the restriction proved impractical to administer. This followed within months of the first documented case of a largely autonomous, AI-run cyber espionage campaign, and coincided with mounting evidence that frontier models alter the economics of both cyber attack and cyber defence. This article examines how these two developments interact, and situates them within the unusual market dynamics now driving large-scale AI development. It argues that access to frontier AI is becoming part of national cyber defence, that such access can be revoked, and that the obvious remedy of sovereign capability remains only partly feasible for all but a handful of states. Drawing on evidence about training costs, the concentration of computing power and the support offered by national AI programmes, it asks what sovereignty can realistically mean for small and middle powers, and for large powers as well. The article proposes a layered strategy: negotiated access guarantees, sovereignty at the level of inference, hedging with open-weight models, pooled regional capability, sustained talent development and continued investment in basic cyber resilience. The open-weight hedge proves at once more capable and more politically exposed than is commonly assumed. Much of the near-term risk lies in how capable models are deployed and contained rather than in their apparent performance.
Chinese Translation
少数几家公司位于两个国家，生产出最具能力的前沿人工智能模型。这些国家的政府展现了法律权力和政治意愿，决定哪些其他国家可以使用这些系统。在2026年6月，美国要求一家领先的开发商在向任何外国人（包括在美国居住的外国国籍人士）发布其最先进的模型之前，必须获得许可证。受影响的模型在全球范围内被迅速撤回，部分原因是这一限制被证明在管理上不切实际。这一事件发生在首次记录到的主要由人工智能驱动的网络间谍活动之后几个月，并与越来越多的证据相吻合，表明前沿模型改变了网络攻击和网络防御的经济学。本文考察了这两种发展如何相互作用，并将其置于当前推动大规模人工智能发展的不寻常市场动态中。文章认为，获取前沿人工智能正在成为国家网络防御的一部分，这种获取可以被撤销，而对于除了少数几个国家之外的所有国家而言，显而易见的主权能力仍然只是部分可行。文章基于培训成本、计算能力的集中以及国家人工智能项目提供的支持，探讨了主权对中小型国家和大国的现实意义。文章提出了一种分层策略：谈判获得访问保障、推理层面的主权、利用开放权重模型进行对冲、区域能力的集中、持续的人才发展以及对基础网络韧性的持续投资。开放权重对冲被证明在能力上更强，同时在政治上也更暴露，这与普遍假设相悖。近期的风险主要在于如何部署和控制这些能力模型，而不是它们表面上的性能。
cs.AI / 68 / 2608.13283
Towards Context-Aware Clinical Motion Understanding in Daily Living at Home: Freezing of Gait Detection with Egocentric Vision
面向家庭日常生活中的情境感知临床运动理解：基于自我中心视觉的步态冻结检测
Abstract
Understanding motion in daily living requires context beyond kinematics, because similar inertial patterns during activities of daily living (ADLs) can reflect intentional stopping, object interaction, or pathological movement impairment. Egocentric vision provides task-related context that may help disambiguate these cases. We investigate this challenge through freezing of gait (FOG) detection in Parkinson's disease (PD), a symptom strongly influenced by contextual factors during ADLs. Using synchronized egocentric video, wearable IMUs, and expert-annotated FOG labels collected from 13 PD participants in their homes, we evaluate frozen representations from pretrained ego-video and time-series foundation models, alongside an IMU-based TCN trained from scratch, under leave-one-subject-out evaluation. The IMU-based TCN achieved the strongest event-detection performance, reaching 42.3 F1 and 83.0 AUROC, compared with 32.6 F1 and 77.2 AUROC for V-JEPA2 ego-video features. Although ego-video alone did not outperform IMU-based sensing, it showed above-chance discrimination, and qualitative analyses suggest that egocentric vision may capture FOG-relevant information independent of IMUs. Together, these results support the use of pretrained ego-video representations to add contextual information to wearable-sensor-based clinical motion understanding in daily living.
Chinese Translation
理解日常生活中的运动需要超越运动学的情境，因为在日常生活活动（ADLs）中相似的惯性模式可能反映出意图停止、物体交互或病理性运动障碍。自我中心视觉提供了与任务相关的情境，这可能有助于消除这些情况的歧义。我们通过帕金森病（PD）中的步态冻结（FOG）检测来研究这一挑战，FOG是一个在ADLs中受到情境因素强烈影响的症状。利用同步的自我中心视频、可穿戴惯性测量单元（IMUs）和从13名PD参与者在家中收集的专家标注的FOG标签，我们评估了来自预训练的自我视频和时间序列基础模型的冻结表示，以及从头开始训练的基于IMU的时间卷积网络（TCN），采用留一法评估。与V-JEPA2自我视频特征的32.6 F1和77.2 AUROC相比，基于IMU的TCN达到了最强的事件检测性能，F1值为42.3，AUROC为83.0。尽管自我视频单独的表现未能超越基于IMU的传感，但其显示出高于随机的区分能力，定性分析表明自我中心视觉可能捕捉到与FOG相关的信息，而不依赖于IMU。综合来看，这些结果支持使用预训练的自我视频表示来为基于可穿戴传感器的日常生活临床运动理解添加情境信息。
cs.AI / 69 / 2608.13293
NAS-Driven Hardware Accelerator Exploration for Edge AI and Quantization Effects on the Pareto Space
基于NAS的边缘人工智能硬件加速器探索及量化对帕累托空间的影响
Abstract
Edge AI deployment demands neural architectures that are simultaneously accurate, computationally efficient, and hardware-deployable - a challenge addressed by hardware-aware Neural Architecture Search (NAS). While recent works incorporate quantization directly into the NAS loop, these approaches expand search complexity and tightly couple architecture and quantization design. The simpler post-search quantization strategy has received little analytical attention: the effects of Post-Training Quantization (PTQ) on the NAS-discovered Pareto structure remain uncharacterised, and no framework combines quantized architecture mapping onto reconfigurable accelerators with automated hardware exploration. This paper addresses both gaps. First, a three-stage pipeline is proposed: a hardware-agnostic Pareto rank surrogate frontend on NAS-Bench-201, a quantization bridge with Pareto-aware filtering and feedback control, and an evolutionary Domain Space Exploration (DSE) backend on CGRA4ML for optimal hardware mapping. Second, an empirical study characterises how INT4 PTQ perturbs the NAS-Bench-201 Pareto space through formal stability metrics on ground-truth data for all 15,625 architectures, and demonstrates that an FP32 zero-shot surrogate outperforms a dedicated INT4-trained surrogate in Pareto space coverage across two standard search strategies.
Chinese Translation
边缘人工智能的部署需要同时具备准确性、计算效率和硬件可部署性的神经架构，这一挑战通过硬件感知的神经架构搜索（NAS）得以解决。尽管近期的研究将量化直接纳入NAS循环，但这些方法增加了搜索复杂性，并紧密耦合了架构与量化设计。相对简单的后搜索量化策略却鲜有分析关注：后训练量化（PTQ）对NAS发现的帕累托结构的影响尚未被表征，且没有框架将量化架构映射到可重构加速器与自动化硬件探索结合起来。本文针对这两个空白进行了探讨。首先，提出了一个三阶段的流程：在NAS-Bench-201上使用硬件无关的帕累托排名代理前端，结合帕累托感知过滤和反馈控制的量化桥，以及在CGRA4ML上进行最佳硬件映射的进化域空间探索（DSE）后端。其次，通过对所有15,625种架构的真实数据进行正式稳定性度量，实证研究表征了INT4 PTQ如何扰动NAS-Bench-201的帕累托空间，并展示了FP32零样本代理在两个标准搜索策略下的帕累托空间覆盖能力优于专门的INT4训练代理。
cs.AI / 70 / 2608.13317
StateBridge: Training-free Hidden-state Alignment for Latent Communication in LLM Multi-Agent Systems
StateBridge：无训练的隐状态对齐用于大语言模型多智能体系统中的潜在通信
Abstract
Large language model based multi-agent systems usually communicate in text, i.e., using discrete tokens. However, text introduces a discrete bottleneck. Converting the sender's continuous hidden states into discrete tokens discards information that token identities alone cannot capture. Recent work proposes latent communication as an alternative, where agents transmit hidden representations directly without converting them to text. However, existing latent methods either inject working memory layer by layer across the transformers, or require trained projectors that limit portability. We propose StateBridge, a training-free latent communication approach that aligns the sender's final-layer hidden states to the receiver's input space via a closed-form orthogonal transformation. Lightweight norm calibration and vocabulary anchoring ensure compatibility with the pretrained input distribution. The aligned states are prepended to the input of the receiver agent as a continuous prefix. We evaluate StateBridge on math reasoning, code generation, and question answering with four models from two families. StateBridge achieves the best or tied-best score on 22 out of 26 model-task pairs, consistently outperforming the strongest baseline.
Chinese Translation
基于大语言模型的多智能体系统通常通过文本进行通信，即使用离散的标记。然而，文本引入了离散瓶颈。将发送者的连续隐状态转换为离散标记会丢失仅凭标记身份无法捕获的信息。最近的研究提出了潜在通信作为替代方案，其中智能体直接传输隐含表示，而无需将其转换为文本。然而，现有的潜在方法要么在变换器中逐层注入工作记忆，要么需要训练过的投影器，这限制了其可移植性。我们提出了StateBridge，一种无训练的潜在通信方法，通过封闭形式的正交变换将发送者的最终层隐状态对齐到接收者的输入空间。轻量级的范数校准和词汇锚定确保与预训练输入分布的兼容性。对齐的状态作为连续前缀被添加到接收者智能体的输入中。我们在数学推理、代码生成和问答任务上评估了StateBridge，使用了来自两个家族的四个模型。StateBridge在26个模型-任务对中，有22个达到了最佳或并列最佳的得分，始终优于最强基线。
cs.AI / 71 / 2608.13333
LLM-Guided Graph Generation for Structure-Based Local Improvement Methods
基于LLM引导的图生成用于结构化局部改进方法
Abstract
Large neighborhood search normally selects a random subset of decision variables for iterative optimization. For efficiently solving different problems, researchers tend to design variable selection strategies by taking into account structural features from different domains. In this paper, we build an automatic pipeline that is problem-agnostic to all problems in the MiniZinc format. By prompting an LLM with our semantic guidelines, we guide the LLM to produce a graph generator that maps any instance of a problem type to a uniform weighted graph, where nodes represent decision variables and edges represent constraint relationships. These problem-agnostic graphs guide our structure-based local improvement framework (SLIM) in variable selection. Meanwhile, the weighted graph enables all problem instances to share the same generic graph representation, from which the same graph features can be extracted and used for configuration selection. We evaluated our pipeline on instances across 20 MiniZinc competition problems, finding that algorithm selection achieves a 39.5% average problem-weighted win rate against a one-shot Gurobi baseline, more than doubling the best single configuration (19.3%). Configuration and feature ablation boost the performance further to 44.0%, demonstrating that LLM-based semantic generation enables effective automated structure extraction and feature extraction for constraint optimization.
Chinese Translation
大邻域搜索通常选择一组随机的决策变量进行迭代优化。为了高效解决不同问题，研究人员倾向于设计变量选择策略，考虑来自不同领域的结构特征。在本文中，我们构建了一个自动化管道，该管道对所有MiniZinc格式的问题具有问题无关性。通过使用我们的语义指南提示LLM，我们引导LLM生成一个图生成器，该生成器将任何问题类型的实例映射到一个统一的加权图，其中节点表示决策变量，边表示约束关系。这些问题无关的图引导我们的基于结构的局部改进框架（SLIM）进行变量选择。同时，加权图使所有问题实例能够共享相同的通用图表示，从中可以提取相同的图特征并用于配置选择。我们在20个MiniZinc竞赛问题的实例上评估了我们的管道，发现算法选择在对比一次性Gurobi基线时实现了39.5%的平均问题加权胜率，超过了最佳单一配置（19.3%）的两倍。配置和特征消融进一步将性能提升至44.0%，证明基于LLM的语义生成能够有效实现约束优化的自动化结构提取和特征提取。
cs.AI / 72 / 2608.13344
LongEarth-R1: Benchmarking and Aligning Vision-Language Models for Long-Horizon Earth Observation Reasoning
LongEarth-R1：长时域地球观测推理的视觉-语言模型基准与对齐
Abstract
Long-horizon Earth observation reasoning requires models to organize multi-stage geographic evolution, localize spatial changes, detect temporal anomalies, and infer future from extended image sequences. However, existing remote sensing vision-language models mainly focus on isolated images, image pairs, or short sequences, limiting reliable grounding in the relevant frames and regions. We introduce LongEarth-Bench, a benchmark containing approximately 120k question-answering samples derived from 117k unique images. Its sequences average 15.14 frames and extend to 30 frames, covering 12 tasks across evolution summarization, spatial reasoning, anomaly identification, and logical prediction. A 30k-sample subset further provides structured reasoning traces linking key frames and changed regions to final answers. We develop LongEarth through supervised fine-tuning with explicit sequence identifiers and structured chain-of-thought supervision. Building on LongEarth, LongEarth-R1 applies group relative policy optimization with format, temporal, and spatial rewards. LongEarth-R1 achieves the best results on all 12 long-sequence tasks while remaining competitive on standard remote sensing benchmarks.
Chinese Translation
长时域地球观测推理要求模型组织多阶段的地理演变，定位空间变化，检测时间异常，并从扩展的图像序列中推断未来。然而，现有的遥感视觉-语言模型主要集中于孤立图像、图像对或短序列，限制了在相关帧和区域中的可靠基础。我们引入了LongEarth-Bench，这是一个基准，包含约120,000个问题-回答样本，源自117,000张独特图像。其序列平均包含15.14帧，最长可达30帧，涵盖了12个任务，包括演变总结、空间推理、异常识别和逻辑预测。一个包含30,000个样本的子集进一步提供了结构化推理轨迹，将关键帧和变化区域与最终答案联系起来。我们通过带有显式序列标识符和结构化思维链监督的有监督微调开发了LongEarth。在此基础上，LongEarth-R1应用了带有格式、时间和空间奖励的群体相对策略优化。LongEarth-R1在所有12个长序列任务中取得了最佳结果，同时在标准遥感基准测试中保持竞争力。
cs.AI / 73 / 2608.13345
Rules or Character? Scaling Laws for AI Safety Design
规则还是性格？人工智能安全设计的规模法则
Abstract
Artificial Intelligence (AI) safety systems combine character shaping (e.g., Reinforcement Learning from Human Feedback [RLHF], Constitutional AI), which modifies behavioral distributions at training time, with rule enforcement (e.g., output filters, safety classifiers), which blocks harmful outputs at inference time, yet little formal analysis exists on how their optimal balance should change as deployment scales increase. We introduce a stylized comparative-statics model that parameterizes safety design as a resource allocation alpha in [0,1] between these two approaches, incorporating scale-dependent filter degradation, common-mode failures, and character fragility -- the risk that shaped behavior degrades or collapses under novel conditions. Under a multiplicative Pareto damage model, we derive closed-form expected harm and supplement it with tail-risk (CVaR) analysis via Monte Carlo simulation. Across three scenarios (optimistic, moderate, pessimistic), the optimal alpha* is interior or at the rules-only boundary and shifts weakly toward character shaping as deployment scale T grows, from negligible (Delta alpha* = +0.01) to pronounced (Delta alpha* = +0.21) depending on scenario. The dominant parameter is the baseline character fragility rate p^(0)_frag, which shifts alpha* by 0.50 across its range -- far exceeding the effect of tail severity, filter quality, or common-mode failure probability. CVaR and expected-harm optima converge at large T. These results suggest that safety architecture decisions depend less on deployment scale per se than on the reliability of character shaping under distributional shift.
Chinese Translation
人工智能（AI）安全系统结合了性格塑造（例如，基于人类反馈的强化学习 [RLHF]，宪法AI），该方法在训练时修改行为分布，以及规则执行（例如，输出过滤器、安全分类器），该方法在推理时阻止有害输出，但关于如何在部署规模增加时优化二者平衡的正式分析仍然较少。我们引入了一个风格化的比较静态模型，将安全设计参数化为在这两种方法之间的资源分配α，范围在[0,1]之间，考虑了规模依赖的过滤器退化、共同模式故障和性格脆弱性——即在新条件下塑造的行为退化或崩溃的风险。在乘法帕累托损害模型下，我们推导出闭式的预期损害，并通过蒙特卡洛模拟补充尾部风险（条件价值风险，CVaR）分析。在三种情境下（乐观、中等、悲观），最优α*位于内部或仅规则边界，并随着部署规模T的增长而微弱地向性格塑造倾斜，从微不足道（Δα* = +0.01）到明显（Δα* = +0.21），具体取决于情境。主导参数是基线性格脆弱性率p^(0)_frag，它在其范围内将α*移动0.50——远超尾部严重性、过滤器质量或共同模式故障概率的影响。在较大的T下，CVaR和预期损害的最优解趋于一致。这些结果表明，安全架构决策与部署规模本身的关系不如与在分布转变下性格塑造的可靠性相关。
cs.AI / 74 / 2608.13389
TopoIntent: Compiling Security Intent into Executable, Compliance-Checked Network Topologies
TopoIntent：将安全意图编译为可执行的合规检查网络拓扑
Abstract
Enterprise security topology design requires translating business intent, regulatory requirements, and risk assumptions into zones, boundary devices, inter-zone paths, and access-control policies. Existing NetOps automation tools mainly operate after this design is fixed, providing limited support for generating structured security topologies from underspecified natural-language requirements. We present TopoIntent, a system that compiles security intent into executable, compliance-checked network topologies. It uses a schema contract to constrain generation, retrieves reference architectures from a curated template library via dense-vector search, and applies staged fusion for intent-template alignment and security completion. The generated topology is checked against CIS Controls v8.1.2 safeguards visible at the topology layer, while unresolved cases are marked for manual review. Structural gaps are repaired through additive schema-preserving edits. The final topology is exported to Mininet scripts with kernel-level iptables ACLs, enabling executable reachability and allow/deny tests. Because no public benchmark exists for this requirement-to-topology task, we construct an evaluation set from reference security architecture diagrams. The retrieval set contains 22 templates and 44 synthetic intents across five scenarios, while the held-out set contains 7 templates and 14 intents from finance and government scenarios excluded from retrieval. On the held-out set, additive repair improves topology-visible CIS satisfaction from 0.78 to 1.00 in fewer than 1.5 rounds on average, and one feedback round raises the post-ACL policy pass rate from 0.78 to 0.88.
Chinese Translation
企业安全拓扑设计需要将业务意图、监管要求和风险假设转化为区域、边界设备、区域间路径和访问控制策略。现有的网络运营自动化工具主要在设计固定后运行，提供的支持有限，无法从不明确的自然语言需求中生成结构化的安全拓扑。我们提出了TopoIntent，一个将安全意图编译为可执行的合规检查网络拓扑的系统。它使用模式契约来约束生成，通过密集向量搜索从策划的模板库中检索参考架构，并应用分阶段融合以实现意图与模板的对齐和安全的完成。生成的拓扑会根据拓扑层可见的CIS Controls v8.1.2安全措施进行检查，而未解决的案例则标记为人工审核。通过附加的保持模式的编辑修复结构性缺口。最终拓扑被导出为Mininet脚本，配备内核级iptables ACL，支持可执行的可达性和允许/拒绝测试。由于没有公共基准用于此需求到拓扑的任务，我们从参考安全架构图构建了一个评估集。检索集包含22个模板和44个合成意图，涵盖五种场景，而保留集包含7个模板和14个来自金融和政府场景的意图，这些场景被排除在检索之外。在保留集中，附加修复将拓扑可见的CIS满意度从0.78提高到1.00，平均不到1.5轮，而一次反馈轮将后ACL策略的通过率从0.78提高到0.88。
cs.AI / 75 / 2608.13409
Jointly Predicting Courses and Grades Using a Transformer-Based Model
基于变压器模型的课程和成绩联合预测
Abstract
Existing predictive models in learning analytics often treat student academic history as a simple sequence, overlooking the concurrent nature of courses taken within a semester. This simplification can lead to inaccurate performance predictions, particularly for students with heavy or challenging course loads. This paper introduces a TRansformer for Academic Course-grade Estimation (TRACE) that addresses this limitation by jointly predicting both the set of courses a student will take and their corresponding grades for an upcoming semester. Our approach encodes courses on a per-semester basis to capture the effects of course concurrency and utilizes a novel loss function combining course-set prediction with grade prediction. We demonstrate that predicting courses taken in addition to the grades in those courses leads to significant improvements in prediction quality. Trained on ten years of institutional data, our joint prediction model reduces mean absolute error by nearly 50% compared to an identical architecture that predicts grades alone. The model also outperforms traditional LSTM-based sequential models, as well as graph neural network-based approaches, and offers natural ways to incorporate student attribute data. This work demonstrates the utility of modern neural architectures for creating interpretable models that can be adapted to new institutions via retraining and recalibration, as well as the importance of key techniques, such as predicting courses taken during training. We discuss how this model could be incorporated into early detection systems at institutions of higher education.
Chinese Translation
现有的学习分析预测模型通常将学生的学业历史视为简单的序列，忽视了一个学期内所修课程的并发性。这种简化可能导致不准确的表现预测，尤其对于课程负担重或挑战性大的学生。本文提出了一种用于学术课程成绩估计的变压器模型（TRansformer for Academic Course-grade Estimation, TRACE），该模型通过联合预测学生将在下个学期修读的课程及其对应的成绩来解决这一局限性。我们的方法按学期对课程进行编码，以捕捉课程并发的影响，并利用一种新颖的损失函数，将课程集预测与成绩预测相结合。我们证明，除了预测课程成绩外，预测所修课程也能显著提高预测质量。基于十年的机构数据训练，我们的联合预测模型相比仅预测成绩的相同架构将平均绝对误差降低了近50%。该模型还优于传统的基于LSTM的序列模型以及基于图神经网络的方法，并提供了自然的方式来整合学生属性数据。这项工作展示了现代神经架构在创建可解释模型方面的实用性，这些模型可以通过再训练和重新校准适应新的机构，以及关键技术的重要性，例如在训练期间预测所修课程。我们讨论了该模型如何融入高等教育机构的早期预警系统中。
cs.AI / 76 / 2608.13410
Who Speaks Matters: Authority-Aware Multi-View RAG over Italian Parliamentary Proceedings
发言者的重要性：基于权威感知的意大利议会程序多视角检索增强生成
Abstract
Parliamentary proceedings are a primary record of democratic deliberation, yet their volume and fragmentation make multi-perspective access difficult for citizens, journalists, and researchers. Applying Retrieval-Augmented Generation (RAG) to parliamentary transcripts introduces three specific risks: dominance of the most frequent speakers, inability to weight speakers according to topical expertise, and citation misattribution in politically sensitive text. We present ParliamentRAG, a RAG system for the Italian Chamber of Deputies that addresses these risks jointly. Its core contribution is a topic-dependent authority model that estimates each speaker's authority as a function of the current query, combining interpretable components such as profession, education, and previous interventions. Given a user query, the system retrieves relevant speech chunks, identifies topic-relevant experts across parliamentary groups, and generates a summary synthesizing their perspectives, accompanied by supporting quotations. ParliamentRAG is evaluated against Google NotebookLM on 15 policy topics via a two-level protocol combining automated metrics and blind A/B human evaluation by six domain experts. The system achieves higher coverage across political groups (0.97 vs. 0.95), perfect quotation faithfulness (1.00 vs. 0.95), and stronger expert preferences on source-related dimensions, while NotebookLM remains stronger on prose-oriented dimensions.
Chinese Translation
议会程序是民主辩论的主要记录，但其数量和碎片化使得公民、记者和研究人员难以进行多角度访问。将检索增强生成（Retrieval-Augmented Generation, RAG）应用于议会记录引入了三种特定风险：最频繁发言者的主导地位、无法根据主题专业性对发言者进行加权，以及在政治敏感文本中引用错误归属。我们提出了ParliamentRAG，这是一个针对意大利众议院的RAG系统，旨在共同解决这些风险。其核心贡献是一个主题依赖的权威模型，该模型根据当前查询估计每位发言者的权威性，结合可解释的组件，如职业、教育背景和先前的发言。给定用户查询后，该系统检索相关的发言片段，识别各议会小组中的主题相关专家，并生成综合其观点的摘要，附带支持性引用。ParliamentRAG在15个政策主题上通过结合自动化指标和六位领域专家的盲法A/B人类评估的两级协议，与Google NotebookLM进行了评估。该系统在政治小组覆盖率（0.97 vs. 0.95）、引用忠实度（1.00 vs. 0.95）以及在源相关维度上的专家偏好方面表现更佳，而NotebookLM在散文导向维度上仍然更强。
cs.AI / 77 / 2608.13417
Beyond Final Scores: A Systematic Evaluation of Agents for Long-Horizon AI Research and Development
超越最终得分：对长时间跨度人工智能研究与开发代理的系统评估
Abstract
Autonomous agents are increasingly capable of improving models, systems, and other technical artifacts through long-horizon experimentation. To understand the current state of this capability, however, evaluation must go beyond final scores, which neither reveal where progress is gained or lost nor indicate whether accumulated experience improves later decisions. We therefore present a systematic evaluation of seven frontier models on 36 long-horizon tasks based on a new framework that uses rule-based metrics to characterize within-run behavior through Solution Framing, Execution, and Feedback Control and controlled comparisons to assess experience reuse within and across tasks. The results show that current agents operate more like engineering optimizers than fully autonomous researchers: they can formulate and implement practical solutions, but their performance varies substantially across runs, their strongest solutions mainly adapt or combine established techniques, and genuine methodological novelty remains rare. Detailed analysis reveals that observed performance is shaped by multiple factors, including distinct process bottlenecks behind similar final outcomes, experience reuse that can help or mislead subsequent decisions, and harness designs that affect performance stability. These findings suggest concrete directions for improving model training, inference-time strategies, experience management, and harness design.
Chinese Translation
自主代理在通过长时间跨度实验改进模型、系统及其他技术工件方面的能力日益增强。然而，要理解这种能力的当前状态，评估必须超越最终得分，因为最终得分既无法揭示进展的得失，也无法指示积累的经验是否改善了后续决策。因此，我们基于一个新的框架，提出了对七个前沿模型在36个长时间跨度任务上的系统评估，该框架使用基于规则的指标通过解决方案框架、执行和反馈控制来表征运行内行为，并通过受控比较评估任务内和任务间的经验重用。结果表明，当前的代理更像是工程优化器，而不是完全自主的研究者：它们能够制定和实施实用解决方案，但其性能在不同运行间差异显著，最强的解决方案主要是对既有技术的适应或组合，而真正的方法论创新仍然较为稀缺。详细分析揭示，观察到的性能受多种因素的影响，包括在类似最终结果背后的不同过程瓶颈、可能帮助或误导后续决策的经验重用，以及影响性能稳定性的设计。这些发现为改善模型训练、推理时策略、经验管理和设计提供了具体方向。
cs.AI / 78 / 2608.13420
Enhancing Virtual Agents through SLMs and Edge-Computing: An Exploratory Evaluation of Think and Memory Processes
通过小语言模型和边缘计算增强虚拟代理：思维与记忆过程的探索性评估
Abstract
Embodied intelligent virtual agents are expected to operate as persistent, adaptive, and context-aware entities within complex virtual and Metaverse worlds. However, implementing cognitively capable agents in such environments is conceptually and technologically challenging. Among a range of blueprints and development approaches, the Cognitive Embodied Agent Architecture (CEAA) has been developed as an implementation-oriented framework for architecting components of perception, memory, reasoning, planning, and embodied action. Considering the recent advances in edge computing and generative AI language models, this paper explores the use of Small Language Models (SLMs) to support edge-based operation of selected CEAA components, focusing on "Think" and "Memory" as processes central to cognitive orchestration and persistence of virtual agents in interactive virtual worlds. An edge-based virtual agent gateway system was developed and evaluated on an NVIDIA Jetson Orin NX using Qwen2.5 models of different sizes, exploring the system's capability to process service requests and handle memory-driven conversations. A series of simulation experiments evaluated routing accuracy, memory-read performance, and latency, demonstrating an SLM-driven prototype agent system that partially implements selected CEAA processes to support the development of embodied agents whose cognitive "brain" can operate efficiently and contextually for interactive experiences in immersive virtual worlds.
Chinese Translation
具身智能虚拟代理预计将在复杂的虚拟和元宇宙世界中作为持久的、自适应的和上下文感知的实体运作。然而，在这些环境中实现具有认知能力的代理在概念和技术上都面临挑战。在众多蓝图和开发方法中，认知具身代理架构（Cognitive Embodied Agent Architecture, CEAA）被开发为一个面向实现的框架，用于构建感知、记忆、推理、规划和具身行动的组件。考虑到边缘计算和生成性人工智能语言模型的最新进展，本文探讨了使用小语言模型（Small Language Models, SLMs）来支持选定的CEAA组件的边缘计算操作，重点关注“思维”和“记忆”作为认知编排和虚拟代理在交互虚拟世界中持久性的核心过程。开发了一个基于边缘的虚拟代理网关系统，并在使用不同规模的Qwen2.5模型的NVIDIA Jetson Orin NX上进行了评估，探索该系统处理服务请求和处理基于记忆的对话的能力。一系列仿真实验评估了路由准确性、记忆读取性能和延迟，展示了一个基于SLM的原型代理系统，该系统部分实现了选定的CEAA过程，以支持具身代理的发展，使其认知“大脑”能够在沉浸式虚拟世界中高效且具有上下文地进行交互体验。
cs.AI / 79 / 2608.13428
RAIL: An Automatic Classifier of the Artificial Intelligence Readiness Level
RAIL：人工智能准备水平的自动分类器
Abstract
Assessing the maturity of artificial intelligence technologies is essential for investment decisions, project management, and policy monitoring, yet the available readiness frameworks are heterogeneous and difficult to apply automatically: the adaptation of Technology Readiness Levels to AI lacks AI-specific gating criteria, the Machine Learning Technology Readiness Levels presuppose access to internal process artifacts, and AI/data readiness dimension models employ scales that resist direct comparison. This paper makes two contributions. First, we unify these three frameworks into the Unified AI Readiness Level (AIRL), a nine-level ordinal scale built on an environmental evidence ladder and complemented by dimensional caps (covering specification, data existence, data quality, data legality, expert knowledge, and algorithmic maturity) together with a generality-anchoring rule and explicit assignment disciplines, so that a readiness level becomes decidable from a natural-language description of the work alone. Second, we propose RAIL (Readiness Assessment via Independent LLM-experts), a panel-of-experts classifier that operationalizes the scale: one evidence agent and six independent dimension agents, each a large language model with a narrowly scoped mandate, deliver verdicts that a deterministic minimum rule aggregates and a chief expert reviews under asymmetric authority, confirming or lowering the panel's recommendation but never raising it above the caps. The method was tested in the analysis of several research works showing consistency and avoiding overestimation from monolithic LLM classifiers.
Chinese Translation
评估人工智能技术的成熟度对于投资决策、项目管理和政策监测至关重要，但现有的准备框架各异且难以自动应用：将技术准备水平（Technology Readiness Levels）适应于人工智能时缺乏特定于人工智能的门槛标准，机器学习技术准备水平（Machine Learning Technology Readiness Levels）则假设可以访问内部过程文档，而人工智能/数据准备维度模型采用的量表则难以进行直接比较。本文作出了两项贡献。首先，我们将这三个框架统一为统一人工智能准备水平（Unified AI Readiness Level, AIRL），这是一个基于环境证据梯度构建的九级序数尺度，并辅以维度上限（涵盖规范、数据存在、数据质量、数据合法性、专家知识和算法成熟度），以及一个通用性锚定规则和明确的分配原则，使得仅通过自然语言描述工作即可决定准备水平。其次，我们提出了RAIL（通过独立大型语言模型专家进行准备评估），这是一个专家小组分类器，旨在将该尺度付诸实践：一个证据代理和六个独立维度代理，每个代理都是一个具有狭窄任务范围的大型语言模型，提供的裁决由一个确定性最小规则汇总，并由一位首席专家在不对称权威下进行审查，确认或降低小组的建议，但绝不将其提高到上限之上。该方法在分析多个研究工作中进行了测试，显示出一致性，并避免了来自单一大型语言模型分类器的高估。
cs.AI / 80 / 2608.13447
Academic League of Artificial Intelligence - An Integrative Perspective of Teaching, Research, and Extension
人工智能学术联盟 - 教学、研究与推广的综合视角
Abstract
Academic leagues have become important mechanisms for promoting extracurricular education and strengthening the integration between universities and society. This paper presents the organizational framework adopted by the Academic League of Artificial Intelligence (LIA) at the Federal University of Santa Catarina (UFSC), designed to integrate teaching, research, and university extension through a student-centered, project-based approach. The framework combines democratic governance, collaborative learning, and dynamic project organization to foster both technical and transversal competencies. The framework is illustrated through representative initiatives, including competition teams, study groups, open lectures, knowledge repositories, and AI-powered applications with social impact. These projects demonstrate how diverse educational, scientific, and extension activities can be developed within a common organizational structure while promoting leadership, scientific production, community engagement, and knowledge preservation. The reported experience indicates that the proposed framework provides a flexible and replicable model for integrating the three university pillars into engineering and computing education, offering practical guidance for academic leagues and similar student organizations.
Chinese Translation
学术联盟已成为促进课外教育和加强大学与社会之间整合的重要机制。本文介绍了圣卡塔琳娜联邦大学（UFSC）人工智能学术联盟（LIA）所采用的组织框架，该框架旨在通过以学生为中心的项目驱动方法整合教学、研究和大学推广。该框架结合了民主治理、协作学习和动态项目组织，以培养技术能力和跨学科能力。通过代表性举措，如竞赛团队、学习小组、公开讲座、知识库和具有社会影响力的人工智能应用，阐释了该框架。这些项目展示了如何在共同的组织结构内发展多样的教育、科学和推广活动，同时促进领导力、科学产出、社区参与和知识保存。报告的经验表明，所提出的框架为将三大大学支柱整合到工程和计算机教育中提供了一个灵活且可复制的模型，为学术联盟和类似的学生组织提供了实用指导。
cs.AI / 81 / 2608.13456
A Unifying Perspective on Causal World Models: From Observations to Representations to Structure
关于因果世界模型的统一视角：从观察到表征再到结构
Abstract
World Models (WM) are increasingly seen as a foundation for intelligent agents that can predict, plan, and act beyond their training distribution. In this paper, we study WMs from a causal perspective across multiple levels of abstraction, ranging from perceptual observations to building a conceptual representation of the structure governing the environment dynamics. We argue that useful WMs must go beyond generative capabilities alone: they should also capture entity properties, entity-to-entity interactions, and entity-to-environment interactions that determine and explain the dynamics of a system. We provide a formal definition of Causal WMs (CWMs) grounded in the tasks they are intended to support, connecting world modelling with existing work in causal representation learning, object-centric learning, causal discovery, structural causal models, and model-based decision-making. Finally, we relate CWMs to the literature on identifiability, clarifying when the components of a WM can be recovered from data and up to which equivalence. With this, we ground WMs in representations and structures that support causal reasoning and informed decision-making.
Chinese Translation
世界模型（World Models, WM）越来越被视为智能体的基础，使其能够在超出训练分布的范围内进行预测、规划和行动。本文从因果的角度研究了世界模型，涵盖多个抽象层次，从感知观察到构建支配环境动态的概念性结构表征。我们认为，有用的世界模型必须超越单纯的生成能力：它们还应捕捉实体属性、实体间的相互作用以及实体与环境之间的相互作用，这些因素决定并解释了系统的动态。我们提供了因果世界模型（Causal World Models, CWM）的正式定义，基于它们所支持的任务，将世界建模与因果表征学习、以对象为中心的学习、因果发现、结构因果模型和基于模型的决策制定等现有工作联系起来。最后，我们将因果世界模型与可识别性文献相关联，阐明何时可以从数据中恢复世界模型的组件及其等价性。通过这些，我们将世界模型扎根于支持因果推理和知情决策的表征和结构中。
cs.AI / 82 / 2608.13476
MARC v1: An Open-Source Multi-Agent Framework for Clinical AI Reasoning and Coordination
MARC v1：一个开源的多智能体框架用于临床人工智能推理与协调
Abstract
We present Multi-Agent Reasoning and Coordination (MARC), an open-source framework that replaces monolithic LLM prompting with deterministic multi-agent orchestration for clinical reasoning. MARC coordinates role-specialized agents for extraction, reasoning, answer generation, and evaluation, with explicit context passing and traceable intermediate outputs, enabling stage-wise failure attribution. We additionally introduce a Decomposer module that generates task-specific agent prompts from a plain-language description, eliminating manual prompt engineering. The framework supports both API-based and local CPU-compatible deployments and is entirely configurable via YAML, without code modifications. MARC is designed to be model-agnostic, interpretable, and accessible to clinical domain experts without programming expertise. The full framework is available at https://github.com/Penn-RAIL/MARC-v1.
Chinese Translation
我们提出了多智能体推理与协调（MARC），这是一个开源框架，旨在用确定性的多智能体编排取代单一的LLM提示，以实现临床推理。MARC协调角色专业化的智能体进行信息提取、推理、答案生成和评估，具有明确的上下文传递和可追溯的中间输出，从而实现阶段性故障归因。此外，我们还引入了一个分解器模块，该模块根据普通语言描述生成特定任务的智能体提示，消除了手动提示工程的需求。该框架支持基于API和本地CPU兼容的部署，并且可以通过YAML完全配置，无需代码修改。MARC旨在实现模型无关、可解释，并使临床领域专家在没有编程专业知识的情况下也能使用。完整框架可在 https://github.com/Penn-RAIL/MARC-v1 获取。
cs.AI / 83 / 2608.13492
AlayaWorld: Interactive Long-Horizon World Modeling - Full Technical Report (v1.1)
AlayaWorld：交互式长时间跨度世界建模 - 完整技术报告 (v1.1)
Abstract
This report presents an improved version of AlayaWorld. While the backbone architecture, chunk-wise autoregressive generation scheme, and training data remain unchanged from the previous release, we substantially revise how conditioning signals are represented and integrated into the model. The new design is guided by a simple principle: conditioning signals should match the generated content as closely as possible in both latent representation and temporal structure. To this end, we make two major changes. First, we replace the previous depth-warping-based spatial memory with a streaming 3D point-cache renderer. Second, we redesign the conditioning pipeline so that visual conditions are encoded in the same causal-VAE latent space, with temporal statistics consistent with those of the generated video. Concretely, the new version introduces six modifications: (1) replacing static-frame image conditioning with motion-aware latent conditioning; (2) causally encoding re-rendered spatial memory as a continuous sequence; (3) aligning the temporal-memory window in pixel space; (4) adopting hard memory dropout that removes memory tokens rather than zeroing them; (5) unifying the VAE encoding and decoding protocol across training and inference; and (6) removing the camera AdaLN branch, such that viewpoint control is provided entirely through the re-rendered spatial condition.
Chinese Translation
本报告呈现了 AlayaWorld 的改进版本。尽管主干架构、基于块的自回归生成方案和训练数据与之前的版本保持不变，但我们对条件信号的表示和集成方式进行了实质性的修订。新设计遵循一个简单的原则：条件信号在潜在表示和时间结构上应尽可能与生成内容匹配。为此，我们进行了两项主要更改。首先，我们用流式 3D 点缓存渲染器替换了之前基于深度扭曲的空间记忆。其次，我们重新设计了条件管道，使得视觉条件在相同的因果变分自编码器（causal-VAE）潜在空间中编码，且时间统计与生成视频的一致。具体而言，新版本引入了六项修改：(1) 用运动感知的潜在条件替换静态帧图像条件；(2) 将重新渲染的空间记忆因果编码为连续序列；(3) 在像素空间中对齐时间记忆窗口；(4) 采用硬记忆丢弃，移除记忆标记而不是将其置零；(5) 在训练和推理中统一变分自编码器的编码和解码协议；(6) 移除相机自适应层归一化（AdaLN）分支，使得视点控制完全通过重新渲染的空间条件提供。
cs.AI / 84 / 2608.13547
QuoteBench: How Matched Scores Can Hide Command-Path Failures
QuoteBench：匹配得分如何掩盖命令路径失败
Abstract
LLM coding agents issue Bash commands through interfaces that may serialize, wrap, and reparse model output. Matched execution scores alone cannot distinguish command-generation errors from failures introduced after generation. QuoteBench measures this boundary with exact final-state validation on 56 one-shot tasks from 14 incident-derived families, crossing the generation contract with the execution transport around one deliberately unescaped added parser. Escaping at the interpolation point reproduces each replayed reply's raw-path outcome, so any recovery under a disclosed boundary must come from the model changing its generation. Across eight same-window configurations, replaying the same reply through the added parser lowers success by 55.4 to 73.2 percentage points; disclosure recovers 30.4 to 60.7 points for six configurations, and zero or slightly negative for the other two. Raw generation is nearly saturated at the frontier; boundary adaptation is what still separates models. GPT-5.6-sol's matched gap of -3.6 points hides -64.3 points of damage and +60.7 points of compensation. The deployment configuration reorders models: one reversal among 26 comparable pairs is unambiguous and four more sit on single-task margins. Evaluations of command-issuing agents should report the model configuration, generation contract, execution path, operating point, and final-state validator rather than treat a matched score as an intrinsic model property.
Chinese Translation
大型语言模型（LLM）编码代理通过可能对模型输出进行序列化、包装和重新解析的接口发出 Bash 命令。仅凭匹配执行得分无法区分命令生成错误与生成后引入的失败。QuoteBench 通过对来自 14 个事件派生家族的 56 个一次性任务进行精确的最终状态验证来测量这一边界，跨越生成合同与执行传输，围绕一个故意未转义的附加解析器。插值点的转义再现了每个重放回复的原始路径结果，因此在公开边界下的任何恢复必须来自模型改变其生成。在八种相同窗口配置中，通过附加解析器重放相同的回复使成功率降低了 55.4 到 73.2 个百分点；公开恢复了六种配置中的 30.4 到 60.7 个百分点，而其他两种配置则为零或略微负值。原始生成在边界处几乎饱和；边界适应仍然是区分模型的关键。GPT-5.6-sol 的匹配差距为 -3.6 个百分点掩盖了 -64.3 个百分点的损害和 +60.7 个百分点的补偿。部署配置重新排序了模型：在 26 对可比模型中，有一个反转是明确的，还有四个处于单任务边际。对命令发出代理的评估应报告模型配置、生成合同、执行路径、操作点和最终状态验证器，而不是将匹配得分视为模型的内在属性。
cs.AI / 85 / 2608.13558
OmniScientist: An Omni-Modal Omni-Discipline AI Scientist
OmniScientist：一个全模态全学科的人工智能科学家
Abstract
Recent advances in foundation models have enabled AI scientists to automate increasingly complete research workflows, from hypothesis generation and code execution to manuscript preparation. Yet workflow coverage alone does not provide access to the full evidence on which scientific discovery depends. Existing systems typically reason over text, code, labels, or precomputed summaries, leaving scientifically decisive spatial, temporal, cross-channel, and procedural relations unavailable to the agent. We introduce OmniScientist, an end-to-end, omni-modal AI scientist that conducts multidisciplinary research directly from heterogeneous raw evidence. A perception layer and 3 autonomous agents for ideation, experiment, and writeup operate within a deterministic pipeline, allowing observations to shape research questions, experimental decisions, and final claims throughout the research lifecycle. By running idea, rigour, and claim checks in code, the system enforces novelty screening, statistical validity, execution provenance, and numerical traceability. We evaluate OmniScientist on 36 real-data cases spanning 5 discipline families, 4 families of scientific evidence, and modalities including images, signals, audio, video, 3-D structures, trajectories, tables, formulae, and graphs. The system completes the full path from raw data to a compiled manuscript in all 36 cases and achieves a mean overall paper score of 6.3 with the reference reasoning backbone. In paired comparisons against a blind variant that receives only precomputed scalar features, direct perception improves all 7 evaluation dimensions and wins 85% of head-to-head judgments. These results show that lifecycle-wide perception is essential for evidence-grounded scientific discovery and provides a practical path toward broadly capable AI scientists.
Chinese Translation
近期基础模型的进展使得人工智能科学家能够自动化越来越完整的研究工作流程，从假设生成、代码执行到手稿准备。然而，仅仅覆盖工作流程并不能提供科学发现所依赖的全部证据。现有系统通常仅对文本、代码、标签或预计算摘要进行推理，导致科学上决定性的空间、时间、跨通道和程序关系无法被代理访问。我们提出了OmniScientist，一个端到端的全模态人工智能科学家，能够直接从异构原始证据中进行多学科研究。一个感知层和三个自主代理（用于构思、实验和写作）在一个确定性管道中运作，允许观察在整个研究生命周期中塑造研究问题、实验决策和最终主张。通过在代码中运行创意、严谨性和主张检查，该系统强制执行新颖性筛选、统计有效性、执行来源和数值可追溯性。我们在涵盖5个学科家族、4个科学证据家族以及包括图像、信号、音频、视频、三维结构、轨迹、表格、公式和图形等多种模态的36个真实数据案例上评估了OmniScientist。该系统在所有36个案例中完成了从原始数据到编译手稿的完整路径，并在参考推理基础上达到了平均论文评分6.3。在与仅接收预计算标量特征的盲变量进行配对比较时，直接感知改善了所有7个评估维度，并在85%的对决判断中获胜。这些结果表明，生命周期范围内的感知对于基于证据的科学发现至关重要，并为广泛能力的人工智能科学家提供了一条切实可行的路径。
cs.CL / 1 / 2608.12321
LLMs Know the Constraint But Do Not Use It: Activation Bottlenecks in Pragmatic Constraint Reasoning
大语言模型知道约束但不使用它：实用约束推理中的激活瓶颈
Abstract
When a salient surface cue competes with an implicit feasibility constraint, LLMs often fail -- but aggregate accuracy conflates genuine constraint inference with conservative defaulting. We formalize the distinction as conditional constraint activation: the constraint is internally encoded (Knowledge) symmetrically across constraint-present and -absent prompts (Symmetry), yet only sometimes routed into the decision (Routing) and repairable by a donor activation (Repair). A quartet diagnostic over 14 models reveals two failure modes; probes on two open weights decode the constraint above $88\%$, yet activation patching repairs one ($+6.4$ nats) and not the other ($-0.07$). On a mitigation frontier, no prompted intervention reaches the repair corner: all inflate conservative bias through a single mediation pathway -- prerequisite mention. Hidden-constraint failure is a routing problem, not a knowledge problem.
Chinese Translation
当一个显著的表面线索与隐含的可行性约束竞争时，大语言模型（LLMs）常常失败——但总体准确性将真正的约束推理与保守的默认行为混淆。我们将这种区别形式化为条件约束激活：约束在约束存在和缺失的提示中对称地内部编码（知识），但仅在某些情况下被路由到决策中（路由），并且可以通过捐赠激活进行修复（修复）。对14个模型的四重诊断揭示了两种失败模式；在两个开放权重上的探测器解码约束的准确率超过88%，然而激活修补修复了一个（+6.4 nats），而另一个却没有（-0.07）。在缓解前沿上，没有任何提示干预能够达到修复角落：所有干预都通过单一的中介路径——先决条件提及，来膨胀保守偏见。隐藏约束的失败是一个路由问题，而不是知识问题。
cs.CL / 2 / 2608.12322
What Drives LLM Self-Reflection? A Controlled Ablation of Uncertainty Routing in Armed Conflict Forecasting
是什么驱动了大型语言模型的自我反思？在武装冲突预测中的不确定性路由的控制消融研究
Abstract
Self-reflection is widely assumed to improve LLM reasoning, yet which component drives the gain remains poorly understood. We present a controlled six-condition ablation isolating four components of LLM self-reflection: evidence exposure, diagnostic scaffolding, taxonomy vocabulary, and action routing. Two precise null results converge on a single mechanism. First, structured diagnostic questions add no measurable value over unstructured reflection ($\text{F1} = 0.296$ vs $0.297$, $p = 1.000$, 95\% CI $[-0.041, +0.040]$). Second, presenting the full uncertainty taxonomy while collapsing the action space to a single generic action also adds no value ($\Delta\text{F1} = +0.008$, overlapping 95\% CIs), ruling out taxonomy vocabulary as the mechanism. Typed action routing provides consistent directional gains ($\text{F1} = 0.379$ vs $0.296$); the conservative estimate controlling for taxonomy vocabulary is $\Delta\text{F1} = +0.075$, and the overall gain over the single-shot baseline is significant by bootstrap CI ($\Delta\text{F1} = +0.101$, 95\% CI $[+0.020, +0.185]$). The vocabulary-routing decomposition replicates on GPT-4o: taxonomy vocabulary adds no significant value over generic reflection ($p = 0.773$), while action routing provides significant gains ($p = 0.025$), confirming the mechanism holds across backbones. Gains concentrate on structurally novel conflicts: in Myanmar ($\text{F1}: 0.000 \rightarrow 0.353$) and Ukraine ($0.167 \rightarrow 0.500$), the vocabulary-only condition recovers no more than generic reflection while action routing breaks the degenerate prior. These findings identify typed action routing -- not diagnostic scaffolding or taxonomy vocabulary -- as a promising design principle for metacognitive LLM forecasting agents, while motivating larger-scale evaluation across conflict typologies.
Chinese Translation
自我反思被广泛认为能改善大型语言模型（LLM）的推理能力，但驱动这种提升的具体组件仍然不甚明了。我们展示了一项控制的六条件消融实验，孤立出LLM自我反思的四个组件：证据曝光、诊断支架、分类词汇和行动路由。两个精确的无效结果汇聚到一个单一机制。首先，结构化的诊断问题相较于非结构化反思并未增加可测量的价值（$ ext{F1} = 0.296$ 对比 $0.297$, $p = 1.000$, 95 ext{% CI} $[-0.041, +0.040]$）。其次，呈现完整的不确定性分类法，同时将行动空间压缩为单一通用行动，也未增加价值（$ riangle ext{F1} = +0.008$, 重叠的95 ext{% CI}），排除了分类词汇作为机制的可能性。输入的行动路由提供了一致的方向性增益（$ ext{F1} = 0.379$ 对比 $0.296$）；控制分类词汇的保守估计为$ riangle ext{F1} = +0.075$，而相较于单次基线的整体增益通过自助法置信区间显著（$ riangle ext{F1} = +0.101$, 95 ext{% CI} $[+0.020, +0.185]$）。词汇-路由分解在GPT-4o上重复验证：分类词汇相较于通用反思并未增加显著价值（$p = 0.773$），而行动路由则提供了显著增益（$p = 0.025$），确认该机制在不同模型架构中有效。增益集中在结构上新颖的冲突中：在缅甸（$ ext{F1}: 0.000
ightarrow 0.353$）和乌克兰（$0.167
ightarrow 0.500$），仅使用词汇的条件未能超过通用反思，而行动路由则打破了退化的先验。这些发现确定了类型化行动路由——而非诊断支架或分类词汇——作为元认知LLM预测代理的一个有前景的设计原则，同时激励在更大范围内对冲突类型进行评估。
cs.CL / 3 / 2608.12323
Why Do AI Agents Break Rules? How Framing, Context, and Social Signals Shape Compliance
为什么人工智能代理会违反规则？框架、上下文和社会信号如何影响合规性
Abstract
Specifying a penalty can paradoxically convert a legal obligation into a cost-benefit calculation that favors violation. We demonstrate that this enforcement information paradox systematically occurs in AI agents. While most AI safety evaluations test whether models fail, we investigate why, applying compliance theory from law and economics as a diagnostic tool. We treat compliance theories not as metaphors but as empirical hypotheses and show that each predicts the behavior of a distinct model class. We evaluate our hypotheses across twelve instruction-tuned language models operating as enterprise procurement chatbots. Drawing on theories of deterrence, legitimacy, and expressive law, we show that safety-fine-tuned models maintain compliance broadly, while task-optimized and agentic models treat regulatory signals as mere optimization parameters. These latter models fail to comply under conditions predicted by theory, such as low enforcement penalties and non-command phrasing. Across all models, introducing financial incentives, managerial demands, peer outcomes, or employee pressure produces large compliance failures. AI procurement agents systematically violate regulatory constraints to satisfy local user objectives in ways not captured by standard alignment benchmarks. Ultimately, compliance cannot be achieved by rule embedding alone; model selection is itself a governance decision, and benchmark-based evaluation is insufficient for compliance-sensitive deployments.
Chinese Translation
指定处罚在悖论中可以将法律义务转化为有利于违反的成本效益计算。我们证明这种执法信息悖论在人工智能代理中系统性地发生。虽然大多数人工智能安全评估测试模型是否失败，但我们探讨其原因，应用法律与经济学中的合规理论作为诊断工具。我们将合规理论视为实证假设而非隐喻，并展示每种理论预测不同模型类别的行为。我们在十二个作为企业采购聊天机器人的指令调优语言模型中评估我们的假设。借鉴威慑、合法性和表现性法律理论，我们表明，安全罚款调优模型普遍保持合规，而任务优化和代理模型则将监管信号视为单纯的优化参数。这些后者模型在理论预测的条件下未能合规，例如低执法处罚和非命令性措辞。在所有模型中，引入财务激励、管理要求、同伴结果或员工压力都会导致严重的合规失败。人工智能采购代理系统性地违反监管限制，以满足局部用户目标，这种行为未被标准对齐基准捕捉。最终，合规性不能仅通过规则嵌入来实现；模型选择本身就是一种治理决策，而基于基准的评估对于合规敏感的部署是不够的。
cs.CL / 4 / 2608.12326
On Measuring Semantic Preservation in Legal Ontology Learning
关于法律本体学习中语义保留的测量
Abstract
Ontology learning transforms unstructured text into structured representations for automated reasoning. Yet structuring information risks losing it, and current evaluation methodologies cannot detect such loss, focusing on structural correctness while failing to measure whether meaning survives transformation. We propose an evaluation methodology that addresses this: comparing LLM task performance on source documents against performance on transformed representations, with the difference quantifying semantic loss. We demonstrate this approach on legal merger agreement analysis, a domain chosen for its complex language and precise semantic requirements, comparing direct LLM application against three ontology learning methods across six language models. The results reveal systematic semantic loss with significant variation based on reasoning complexity and model-method interactions. Our contributions are: (1) an evaluation framework for measuring semantic preservation in ontology learning, and (2) empirical evidence that semantic loss varies dramatically with model-method pairing, providing guidance for selecting optimal configurations in legal knowledge systems.
Chinese Translation
本体学习将非结构化文本转化为结构化表示，以实现自动推理。然而，信息的结构化可能导致信息丢失，而当前的评估方法无法检测到这种损失，主要关注结构的正确性，却未能衡量意义在转化过程中是否得以保留。我们提出了一种评估方法，旨在解决这一问题：比较大型语言模型（LLM）在源文档上的任务表现与在转化表示上的表现，二者之间的差异量化语义损失。我们在法律合并协议分析这一领域展示了该方法，该领域因其复杂的语言和精确的语义要求而被选中，比较了直接应用LLM与三种本体学习方法在六种语言模型上的表现。结果显示，语义损失呈系统性，并且基于推理复杂性和模型-方法交互的不同而存在显著差异。我们的贡献包括：（1）一个用于测量本体学习中语义保留的评估框架，以及（2）实证证据表明语义损失在模型-方法配对中变化显著，为法律知识系统中选择最佳配置提供了指导。
cs.CL / 5 / 2608.12327
Comparative Analysis of Multilingual Pre-trained Models for Nepali Automatic Speech Recognition
尼泊尔语自动语音识别的多语言预训练模型比较分析
Abstract
Multilingual pretrained models nominally support Nepali, yet no controlled benchmark has compared them under a single fine-tuning protocol. We fine-tune six pretrained models (XLSR-53, IndicWav2Vec, MMS-1B, Whisper-Medium, Whisper-Large-v3-Turbo, and Conformer-Hi) spanning CTC self-supervised, autoregressive encoder-decoder, and hybrid Conformer-CTC architectures, on the OpenSLR SLR54 Nepali corpus (~165 hours) using identical preprocessing, splits, optimizer, and family-matched learning-rate schedules. We evaluate Word Error Rate (WER), Character Error Rate (CER), and Real-Time Factor (RTF) on three independent test sets (OpenSLR, FLEURS, Common Voice). Whisper-Large-v3-Turbo (14.76% WER) and IndicWav2Vec (14.89% WER) tie at the top despite a 9x parameter gap and 40x pretraining-data gap, providing direct empirical evidence that language-family proximity in pretraining can substitute for raw scale for in-domain Nepali. CTC decoders run up to 29x faster than autoregressive Whisper at the same accuracy, flipping the practical deployment preference toward CTC under any latency budget. Massively multilingual pretraining (MMS-1B) yields the smallest out-of-domain degradation on FLEURS (+12.55 pp), indicating that scale buys robustness rather than peak in-domain accuracy. The resulting benchmark provides the first standardized, multi-model, efficiency-aware reference numbers for Nepali ASR.
Chinese Translation
多语言预训练模型名义上支持尼泊尔语，但尚未有控制基准在单一微调协议下对它们进行比较。我们对六个预训练模型（XLSR-53、IndicWav2Vec、MMS-1B、Whisper-Medium、Whisper-Large-v3-Turbo 和 Conformer-Hi）进行了微调，这些模型涵盖了 CTC 自监督、自动回归编码器-解码器和混合 Conformer-CTC 架构，使用相同的预处理、数据划分、优化器和匹配的学习率调度，在 OpenSLR SLR54 尼泊尔语语料库（约 165 小时）上进行评估。我们在三个独立的测试集（OpenSLR、FLEURS、Common Voice）上评估了字错误率（WER）、字符错误率（CER）和实时因子（RTF）。尽管参数差距达到 9 倍，预训练数据差距达到 40 倍，Whisper-Large-v3-Turbo（14.76% WER）和 IndicWav2Vec（14.89% WER）在顶部平分秋色，提供了直接的实证证据，表明预训练中的语言家族接近性可以替代原始规模以适应尼泊尔语领域。CTC 解码器在相同精度下的运行速度比自动回归 Whisper 快达 29 倍，使得在任何延迟预算下，实际部署的偏好转向 CTC。大规模多语言预训练（MMS-1B）在 FLEURS 上产生了最小的领域外降级（+12.55 个百分点），表明规模带来了鲁棒性，而非峰值领域内精度。由此产生的基准为尼泊尔语 ASR 提供了首个标准化的、多模型的、关注效率的参考数据。
cs.CL / 6 / 2608.12328
LoRA-Diffusion: Parameter-Efficient Fine-Tuning via Low-Rank Trajectory Decomposition
LoRA-Diffusion：通过低秩轨迹分解实现参数高效的微调
Abstract
Parameter-efficient fine-tuning methods such as LoRA have transformed the adaptation of large autoregressive language models, enabling task-specific customization with substantially fewer trainable parameters. However, these methods have not been successfully extended to diffusion-based language models, which generate text through iterative denoising rather than sequential token prediction. We propose LoRA-Diffusion, a parameter-efficient fine-tuning approach that applies low-rank decomposition to the denoising trajectory instead of model weights. Unlike weight-based LoRA, which modifies individual transformation matrices, our method learns low-rank perturbations to the entire diffusion path from noise to output. We introduce trajectory-level low-rank adapters that modify each denoising step, step-adaptive rank allocation across diffusion phases, and compositional multi-task learning that allows merging task-specific modules at inference without retraining. On SST-2, QNLI, and MRPC, we report token-level denoising validation accuracy over five random seeds. LoRA-Diffusion achieves the highest mean performance on SST-2 and strong performance on QNLI and MRPC. Joint multi-task training further shows that LoRA-Diffusion achieves the highest token-level accuracy among the evaluated methods. The approach reduces per-task storage compared with full fine-tuning and establishes a parameter-efficient fine-tuning framework for diffusion language models.
Chinese Translation
参数高效的微调方法，如LoRA，已经改变了大型自回归语言模型的适应性，使得任务特定的定制能够以显著更少的可训练参数实现。然而，这些方法尚未成功扩展到基于扩散的语言模型，这些模型通过迭代去噪而非顺序标记预测生成文本。我们提出了LoRA-Diffusion，这是一种参数高效的微调方法，它将低秩分解应用于去噪轨迹，而不是模型权重。与基于权重的LoRA不同，后者修改单个变换矩阵，我们的方法学习对从噪声到输出的整个扩散路径的低秩扰动。我们引入了轨迹级低秩适配器，修改每个去噪步骤，跨扩散阶段进行步骤自适应秩分配，并实现组合多任务学习，允许在推理时合并任务特定模块而无需重新训练。在SST-2、QNLI和MRPC数据集上，我们报告了五个随机种子的标记级去噪验证准确率。LoRA-Diffusion在SST-2上实现了最高的平均性能，并在QNLI和MRPC上表现强劲。联合多任务训练进一步表明，LoRA-Diffusion在评估的方法中实现了最高的标记级准确率。该方法相比于完全微调减少了每个任务的存储需求，并为扩散语言模型建立了一个参数高效的微调框架。
cs.CL / 7 / 2608.12329
AnchorSIPS: A Synthetic Dataset and Evaluation Resource for Evidence-Supported Psychosis-Risk Symptom Measurement
AnchorSIPS：一种用于证据支持的精神病风险症状测量的合成数据集和评估资源
Abstract
Progress on AI for psychosis-risk assessment is limited by a data-access bottleneck. Real clinical interviews are difficult to share because of privacy, governance, and consent constraints. We present AnchorSIPS, a synthetic dataset of 10K structured psychosis-risk interviews with transcript-grounded measurement targets. Each interview is modeled on Mini-SIPS, a clinician-administered psychosis-risk interview. It captures history, 24 symptom questions, follow-up evidence for items the patient affirms, decisions about delusion-like symptoms (unusual beliefs), hallucination-like symptoms (unusual perceptions), and disorganized communication, exclusion of clear psychotic-level symptoms ("frank psychosis"), and a final attenuated psychosis syndrome (APS) diagnosis, a high-risk state of milder or early psychotic symptoms. The APS diagnosis is not a standalone label. It depends on earlier endorsements, supporting follow-up details, symptom-class decisions, and the frank-psychosis check. Every intermediate decision is anchored to its supporting transcript turns. AnchorSIPS is generated by a plan-then-realize pipeline. A hidden case sheet specifies the patient's clinical state, a deterministic planner fixes the interview structure, and an LLM realizes only the patient utterances under validation and bounded repair. Fixing labels and structure before generation avoids the inter-turn inconsistencies typical of multi-turn LLM dialogue. Across seven LLM baselines, models recover coarse decisions but fail to extract follow-up details or cite supporting transcript turns, so final-label performance overstates interview competence. AnchorSIPS is intended for research on evidence extraction, transcript-grounded measurement, and uncertainty under partial disclosure.
Chinese Translation
人工智能在精神病风险评估方面的进展受到数据访问瓶颈的限制。由于隐私、治理和同意的限制，真实的临床访谈难以共享。我们提出了AnchorSIPS，这是一个包含10,000个结构化精神病风险访谈的合成数据集，具有基于文本的测量目标。每个访谈都基于Mini-SIPS，这是一个由临床医生主导的精神病风险访谈。它记录了病史、24个症状问题、患者确认项目的后续证据、关于妄想样症状（不寻常信念）、幻觉样症状（不寻常感知）和组织混乱的决策，排除明显的精神病级症状（“明显精神病”），以及最终的减轻型精神病综合症（APS）诊断，这是一种轻度或早期精神病症状的高风险状态。APS诊断并不是一个独立的标签。它依赖于早期的认可、支持的后续细节、症状类别的决策和明显精神病检查。每个中间决策都与其支持的文本转录相联系。AnchorSIPS是通过一个计划-然后实现的流程生成的。一个隐藏的案例表指定了患者的临床状态，一个确定性规划器固定了访谈结构，而一个大型语言模型（LLM）仅在验证和有限修复下实现患者的发言。在生成之前固定标签和结构可以避免多轮大型语言模型对话中典型的轮间不一致性。在七个大型语言模型基准中，模型能够恢复粗略决策，但未能提取后续细节或引用支持的文本转录，因此最终标签的表现夸大了访谈的能力。AnchorSIPS旨在用于证据提取、基于文本的测量和部分披露下的不确定性研究。
cs.CL / 8 / 2608.12330
Reliability-Aware Sexism Detection: Combining DPO with Annotator Agreement and Token-Level Confidence Scoring
考虑可靠性的性别歧视检测：将 DPO 与标注者一致性和词元级置信评分相结合
Abstract
The detection of online sexism remains an open problem. Sexism detection is inherently subjective, yet most existing systems reduce multi-annotator labels to a single majority decision and treat all instances uniformly. This ignores two informative signals: annotator agreement and model uncertainty. We propose RA-DPO (Reliability-Aware Direct Preference Optimization), which integrates annotator agreement, model confidence, and a token-level uncertainty signal into a single reliability score. RA-DPO uses this score to select high-value preference pairs during training and to support inference-time abstention, which allows the model to trade coverage for accuracy. We evaluate RA-DPO on 6,920 multilingual posts from EXIST 2023, fine-tune OpenAI gpt-4o base via DPO, and validate on two open-weight 3B models (Llama, Qwen). Results show that training on the top 30% most reliable pairs matches full-data DPO, which indicates that reliability-aware selection can reduce training cost without sacrificing performance. At inference, selective prediction reaches 96.2% accuracy at 50% coverage in the true-agreement setting and 88.7% in the deployable predicted-agreement setting, both exceeding the 85.3% no-agreement baseline. These results suggest that accounting for annotation uncertainty is beneficial for both efficient training and reliable deployment in subjective classification.
Chinese Translation
在线性别歧视的检测仍然是一个未解决的问题。性别歧视检测本质上是主观的，但大多数现有系统将多标注者标签简化为单一的多数决策，并且对所有实例采取统一处理。这忽视了两个重要信号：标注者一致性和模型不确定性。我们提出了 RA-DPO（考虑可靠性的直接偏好优化），它将标注者一致性、模型置信度和词元级不确定性信号整合为一个单一的可靠性评分。RA-DPO 使用该评分在训练过程中选择高价值的偏好对，并支持推理时的弃权，这使得模型能够在覆盖率与准确性之间进行权衡。我们在来自 EXIST 2023 的 6,920 条多语言帖子上评估 RA-DPO，通过 DPO 对 OpenAI gpt-4o 基础模型进行微调，并在两个开放权重的 3B 模型（Llama, Qwen）上进行验证。结果表明，在最可靠的 30% 配对上进行训练的效果与全数据 DPO 相匹配，这表明考虑可靠性的选择可以降低训练成本而不牺牲性能。在推理阶段，选择性预测在真实一致性设置下达到 96.2% 的准确率，覆盖率为 50%，在可部署的预测一致性设置下为 88.7%，均超过 85.3% 的无一致性基线。这些结果表明，考虑标注不确定性对主观分类的高效训练和可靠部署都是有益的。
cs.CL / 9 / 2608.12331
Thought-Aware KV Cache Compaction for Reasoning via Adaptive Attention Matching
基于思维感知的KV缓存压缩用于通过自适应注意力匹配进行推理
Abstract
Reasoning language models generate lengthy chain-of-thought (CoT) sequences whose key-value (KV) cache grows linearly and becomes a memory bottleneck during decoding. Existing compaction methods treat reasoning trajectories as flat token sequences and apply uniform compression, ignoring the hierarchical structure of CoT reasoning where different steps vary drastically in importance. We propose \textbf{Thought-Aware Attention Matching (TAM)}, which exploits this structure through three mechanisms: (i)~thought segmentation that decomposes the trajectory into reasoning blocks, (ii)~adaptive budget allocation that assigns compression budget based on each segment's importance and size, and (iii)~pivotal token protection that preserves high-attention reasoning anchors. We prove that the allocation rule is optimal under a convex error model and that cumulative error under sequential compaction remains bounded. Experiments on AIME 2024 and MATH-500 with Qwen3-4B show that TAM improves accuracy over uniform compaction at the same memory footprint, with periodic compaction bounding peak memory to 3.1--3.2\,GB (a 65\% reduction) while maintaining competitive accuracy.
Chinese Translation
推理语言模型生成冗长的思维链（CoT）序列，其关键值（KV）缓存线性增长，并在解码过程中成为内存瓶颈。现有的压缩方法将推理轨迹视为平坦的标记序列，并应用统一压缩，忽略了CoT推理的层次结构，其中不同步骤的重要性差异显著。我们提出了 extbf{思维感知注意力匹配（TAM）}，通过三种机制利用这一结构：（i）思维分段将轨迹分解为推理块，（ii）自适应预算分配根据每个段的重要性和大小分配压缩预算，以及（iii）关键标记保护保留高注意力的推理锚点。我们证明了在凸误差模型下，分配规则是最优的，并且在顺序压缩下累积误差保持有界。在AIME 2024和MATH-500上使用Qwen3-4B的实验表明，TAM在相同内存占用下提高了准确性，周期性压缩将峰值内存限制在3.1-3.2 GB（减少65%），同时保持竞争力的准确性。
cs.CL / 10 / 2608.12332
Can Spectral-Clipping Enable Better Learning While Forgetting Less for Low-Rank Adaptation?
谱剪切能否在低秩适应中实现更好的学习，同时减少遗忘？
Abstract
In recent years, low-rank adaptation (LoRA) has emerged as a significant paradigm that freezes pre-trained weights and introduces small, learnable adapters instead of fine-tuning the full set of parameters. In this work, we uncover several key insights regarding the singular components of network parameters based on Singular Value Decomposition (SVD). Firstly, the principal singular components with large singular values in pre-trained network parameters can be effectively reused during fine-tuning, whereas the minor components with smaller singular values are more task-specific and require substantial adaptation. Secondly, we first establish the theoretical connection that the uncontrolled growth of singular values in LoRA adapters leads to the forgetting of pre-trained knowledge -- a well-known issue referred to as catastrophic forgetting. Building on these observations, we propose SCLoRA, which injects parameterized singular components with spectral clipping into the pre-trained model in a way that is aware of the spectral distribution of the pre-trained model. SCLoRA effectively adapts to new tasks by focusing updates on components that require adaptation, while simultaneously alleviating catastrophic forgetting. We conduct extensive experiments and demonstrate that SCLoRA not only improves downstream performance but also effectively retains pre-trained knowledge.
Chinese Translation
近年来，低秩适应（LoRA）作为一种重要的范式出现，该方法冻结预训练权重，并引入小型可学习适配器，而不是微调完整的参数集。在本研究中，我们揭示了基于奇异值分解（SVD）的网络参数的奇异成分的几个关键见解。首先，预训练网络参数中具有大奇异值的主奇异成分在微调过程中可以有效重用，而具有较小奇异值的次要成分则更具任务特异性，需要进行大量适应。其次，我们首次建立了理论联系，指出LoRA适配器中奇异值的 uncontrolled 增长会导致预训练知识的遗忘——这一问题被称为灾难性遗忘。基于这些观察，我们提出了SCLoRA，它通过谱剪切将参数化的奇异成分注入到预训练模型中，同时考虑到预训练模型的谱分布。SCLoRA通过将更新集中在需要适应的成分上，有效地适应新任务，同时减轻灾难性遗忘。我们进行了广泛的实验，证明SCLoRA不仅提高了下游性能，还有效保留了预训练知识。
cs.CL / 11 / 2608.12333
Vision-Language Models are Fragile Multilingual Associators
视觉-语言模型是脆弱的多语言关联器
Abstract
Vision-language models must associate visual entities with textual attributes. Whether these associations or concept bindings remain stable when the language of the input changes is unexplored. We introduce M$^2$BIND, a benchmark varying the language of the context and query across multiple languages. We evaluate binding both extrinsically through task performance metrics and intrinsically through causal interventions. We find that binding is not language-invariant: cross-family and cross-script settings trigger significant binding collapse, with the model's internal binding computation shifting to later layers and losing causal strength. Closely related languages preserve associations comparatively better. In a broader sense, our findings indicate how VLMs deployed globally in multilingual settings cannot be assumed to maintain the same association quality observed in monolingual evaluation.
Chinese Translation
视觉-语言模型必须将视觉实体与文本属性关联。然而，当输入的语言发生变化时，这些关联或概念绑定是否保持稳定尚未被探索。我们引入了 M$^2$BIND，这是一个在多种语言中变化上下文和查询语言的基准。我们通过任务性能指标从外部评估绑定，同时通过因果干预从内部评估绑定。我们发现绑定并非语言不变：跨语言家族和跨书写系统的设置会导致显著的绑定崩溃，模型的内部绑定计算转移到后续层，并失去因果强度。相关性较强的语言在保持关联方面表现得相对更好。从更广泛的意义上讲，我们的发现表明，全球范围内部署的多语言设置中的视觉-语言模型（VLMs）不能假定保持在单语言评估中观察到的相同关联质量。
cs.CL / 12 / 2608.12334
Steering the Language Axis: From Linear Decodability to Causal Control
引导语言轴：从线性可解性到因果控制
Abstract
Despite the impressive multilingual capabilities of Large Language Models, the latent dynamics dictating language selection remain poorly understood. In this work, we ask whether language identity is merely linearly decodable from hidden states, or if it can be causally controlled by a compact activation direction. We conduct an exhaustive causal intervention analysis across multiple model families, including Qwen 3.5-2B and Llama-3.2-1B-Instruct, isolating PCA-derived "language axes" to perform steering and ablation experiments across 1.26 million generations on the FLORES-200 dataset. Steering along these geometric directions reliably forces language switching in both cross-script (English to Chinese) and same-script (English to Spanish) settings, whereas equal-magnitude random perturbations yield virtually no effect. Our layerwise analysis reveals that language commitment is highly localized and explicitly language-pair-dependent. While English to Chinese switching resists early intervention and steers easily in the later layers, the English-Spanish transition shifts earlier, displaying a distinct, bimodal sensitivity. Furthermore, targeted ablation uncovers a fundamental reversion to English: once the language signal is removed, the model falls back to English regardless of the input prompt. Ultimately, these findings demonstrate that language decision boundaries function during inference as causally active features that are direction-dependent and layer-specific.
Chinese Translation
尽管大型语言模型具备令人印象深刻的多语言能力，但决定语言选择的潜在动态仍然不甚了解。本研究探讨语言身份是否仅仅可以从隐藏状态中线性解码，或者是否可以通过紧凑的激活方向进行因果控制。我们对多个模型系列进行了全面的因果干预分析，包括 Qwen 3.5-2B 和 Llama-3.2-1B-Instruct，隔离主成分分析（PCA）导出的“语言轴”，在 FLORES-200 数据集上进行 126 万次生成的引导和消融实验。沿着这些几何方向的引导可靠地迫使语言在跨脚本（英语到中文）和同脚本（英语到西班牙语）设置中切换，而等幅度的随机扰动几乎没有效果。我们的逐层分析显示，语言承诺高度局部化，并且明确依赖于语言对。尽管英语到中文的切换在早期干预中抵抗并在后层中容易引导，英语-西班牙语的过渡则较早发生，显示出明显的双峰敏感性。此外，针对性的消融揭示了对英语的根本性回归：一旦语言信号被移除，模型无论输入提示如何都会回退到英语。最终，这些发现表明，语言决策边界在推理过程中作为因果活跃特征运作，且依赖于方向和层级特性。
cs.CL / 13 / 2608.12335
HC-RAG: Evidence-Centric Retrieval-Augmented Generation over Heterogeneous Financial Filings
HC-RAG：基于证据的异构金融文件增强检索生成
Abstract
Financial question answering over annual reports requires more than retrieving semantically similar passages. It often involves identifying relevant companies and fiscal years, locating standardized filing sections, collecting textual and tabular evidence, and checking answers against the original documents. Existing RAG systems, however, usually flatten long filings into unordered chunks, pay limited attention to the typed structure of financial reports, and use fixed text-table fusion strategies without considering query intent. To address these limitations, we propose \textbf{HC-RAG}, a hierarchical cross-modal retrieval-augmented generation framework for evidence-centric financial QA. HC-RAG organizes filings into a typed financial evidence graph with documents, sections, text units, table units, and metadata nodes. It retrieves evidence through document-section-unit paths, aligns textual and tabular evidence in a shared retrieval space, and routes evidence according to four semantic intents: calculation, trend, fact, and comparison. We further introduce \textbf{Multi-Doc-2025}, a benchmark containing 2,327 expert-verified QA pairs from 179 SEC 10-K filings of 87 S\&P 500 companies across fiscal years 2022--2024, with labels for intent, difficulty, and structural evidence attributes. Experiments on public financial QA benchmarks and Multi-Doc-2025 show that HC-RAG improves both answer quality and evidence localization, especially in long-document, table-related, and cross-document settings. HC-RAG outperforms RAPTOR by 6.6 F1 points on DocFinQA and GraphRAG by 10.9 F1 points on Multi-Doc-2025. Evidence-level analysis and ablation studies show that the improvements mainly come from more accurate section localization, table grounding, cross-document evidence aggregation, and intent-aware text-table routing.
Chinese Translation
对年度报告的金融问答不仅仅需要检索语义相似的段落。它通常涉及识别相关公司和财政年度，定位标准化的文件部分，收集文本和表格证据，并将答案与原始文件进行核对。然而，现有的检索增强生成（RAG）系统通常将长文件压缩为无序的块，有限关注金融报告的类型结构，并使用固定的文本-表格融合策略而未考虑查询意图。为了解决这些局限性，我们提出了 extbf{HC-RAG}，一种用于基于证据的金融问答的层次交叉模态检索增强生成框架。HC-RAG将文件组织成一个类型化的金融证据图，包括文档、部分、文本单元、表格单元和元数据节点。它通过文档-部分-单元路径检索证据，在共享检索空间中对齐文本和表格证据，并根据四种语义意图（计算、趋势、事实和比较）进行证据路由。我们进一步引入了 extbf{Multi-Doc-2025}，这是一个基准数据集，包含来自179份SEC 10-K文件的2,327个专家验证的问答对，涵盖87家标准普尔500公司，财政年度为2022至2024年，并标注了意图、难度和结构证据属性。在公共金融问答基准和Multi-Doc-2025上的实验表明，HC-RAG在答案质量和证据定位方面均有所改善，尤其是在长文档、表格相关和跨文档设置中。HC-RAG在DocFinQA上比RAPTOR提高了6.6个F1分数，在Multi-Doc-2025上比GraphRAG提高了10.9个F1分数。证据级分析和消融研究表明，这些改进主要来自于更准确的部分定位、表格基础、跨文档证据聚合和意图感知的文本-表格路由。
cs.CL / 14 / 2608.12336
StorySpark: Module-wise Evolutionary Search for Story Premise Generation
StorySpark：基于模块的演化搜索用于故事前提生成
Abstract
A story premise is the creative spark from which a full narrative can grow. Yet LLM-based story generation has mostly emphasized later-stage planning, controllability, coherence, and prose expansion, while premise-level ideation remains comparatively underexplored. We introduce StorySpark, a module-wise evolutionary search framework for story premise generation. StorySpark operates over interpretable narrative modules such as background, persona, event, ending, and twist, treating each active module not as a static field to fill once, but as a local search space conditioned on the partial premise built so far. For each module, it generates alternatives, evaluates them in context, refines them through feedback-driven mutation and recombination, preserves complementary strengths with Pareto-guided selection, and reallocates frontier capacity to balance branch coverage with promising directions. Multi-view automatic and human evaluations show that StorySpark produces stronger final premises than competitive baselines, with especially consistent gains in originality; when expanded with the same story writer, its premises also lead to higher-quality downstream stories while maintaining completeness, fascination, and diverse usable narrative directions.
Chinese Translation
故事前提是完整叙事的创意火花。然而，基于大型语言模型（LLM）的故事生成主要强调后期规划、可控性、一致性和散文扩展，而前提层面的构思相对较少被探索。我们提出了StorySpark，一个用于故事前提生成的模块化演化搜索框架。StorySpark在可解释的叙事模块上运行，如背景、角色、事件、结局和转折，将每个活跃模块视为一个局部搜索空间，而不是一次性填充的静态字段，条件是基于迄今为止构建的部分前提。对于每个模块，它生成替代方案，在上下文中进行评估，通过反馈驱动的变异和重组进行细化，利用帕累托导向选择保留互补优势，并重新分配前沿能力，以平衡分支覆盖与有前景方向。多视角的自动和人工评估表明，StorySpark生成的最终前提比竞争基线更强，尤其在原创性方面表现出一致的提升；当与相同的故事创作者扩展时，其前提也能导致更高质量的下游故事，同时保持完整性、吸引力和多样化的可用叙事方向。
cs.CL / 15 / 2608.12337
From Refuse to Richness: Rubric Rewards for Long-Form Hallucination Reinforcement Learning
从拒绝到丰富：长文本幻觉强化学习的评分奖励
Abstract
Rewards that penalize unsupported claims can improve grounding in long-form generation, but they can also teach models to answer less. We study this refusal-to-richness trade-off in long-form hallucination RL. Instead of using global richness proxies such as length, claim count, detail, or pairwise relevance, we represent each question with a key-point rubric that specifies the required and optional information a useful answer should cover. These rubrics define coverage directly and are used both for evaluation and as reward signals. Across grounding-only, proxy-based, rubric-only, and combined rewards, we find a stable trade-off: strict grounding rewards improve support but suppress coverage, while unconstrained rubric rewards improve coverage but weaken grounding. A soft combination of grounding, rubric coverage, and relevance gives the best balance in our experiments, improving in-distribution support while transferring better to out-of-distribution checklist tasks than either grounding-only or rubric-only rewards.
Chinese Translation
惩罚不支持的主张的奖励可以改善长文本生成中的基础性，但它们也可能教会模型回答更少。我们研究了长文本幻觉强化学习中的拒绝-丰富权衡。我们不使用长度、主张数量、细节或成对相关性等全局丰富性代理，而是用一个关键点评分标准来表示每个问题，该评分标准指定了有用答案应涵盖的必要和可选信息。这些评分标准直接定义了覆盖范围，并用于评估和作为奖励信号。在基础性奖励、基于代理的奖励、仅评分标准奖励和组合奖励之间，我们发现了一个稳定的权衡：严格的基础性奖励提高了支持性，但抑制了覆盖范围，而不受限制的评分标准奖励提高了覆盖范围，但削弱了基础性。在我们的实验中，基础性、评分标准覆盖和相关性的软组合提供了最佳平衡，改善了分布内的支持，同时在转移到分布外的检查表任务时表现优于仅基础性或仅评分标准的奖励。
cs.CL / 16 / 2608.12338
SDAM: Structure-Difference-Aware Memory Evolution for Complex Text-to-SQL
SDAM：针对复杂文本到SQL的结构差异感知记忆演化
Abstract
Text-to-SQL aims to convert natural language questions into executable SQL queries. While memory-based agent system improves complex SQL generation, existing memory design neglect historical experience and suffer from weak structure analysis, shallow semantic understanding, and poor schema alignment. To address these challenges, we propose SDAM. Specifically, SDAM identifies potential errors via a structure-difference aware reasoning tree, extracts deep semantic rules through contradiction-aware reflection, and enhances structural consistency using a schema-grounded memory evolution mechanism to bind memory with database schemas. We integrate SDAM into a Text-to-SQL framework named SDAM-SQL. Experiment shows that SDAM-SQL achieves 2.0 and 0.4 improvement on BIRD-dev and Spider-test compared with mainstream Text-to-SQL methods, showing the effectiveness of SDAM-SQL.
Chinese Translation
文本到SQL旨在将自然语言问题转换为可执行的SQL查询。尽管基于记忆的代理系统改善了复杂SQL的生成，但现有的记忆设计忽视了历史经验，并且在结构分析、语义理解和模式对齐方面存在不足。为了解决这些挑战，我们提出了SDAM。具体而言，SDAM通过结构差异感知推理树识别潜在错误，通过矛盾感知反思提取深层语义规则，并利用基于模式的记忆演化机制增强结构一致性，将记忆与数据库模式绑定。我们将SDAM集成到一个名为SDAM-SQL的文本到SQL框架中。实验表明，与主流文本到SQL方法相比，SDAM-SQL在BIRD-dev和Spider-test上分别实现了2.0和0.4的提升，显示了SDAM-SQL的有效性。
cs.CL / 17 / 2608.12339
Mimicry without understanding: the origins of decision bias in large language models
无理解的模仿：大型语言模型决策偏差的起源
Abstract
Large Language models (LLMs) were found to be susceptible to a host of social, affective, and cognitive biases. We examined two mechanisms through which such biases can be generated even when human preferences (in the training data) are not biased or when they are correctly categorized as being biased. The first is faulty mimicry of preferences based on human behavior: this involves LLMs inferring human preferences even when behaviors are logically unrelated to preferences. The second is mimicry of explicitly biased human behaviors. In four studies focusing on economic biases, we find that ChatGPT-4o and Qwen exhibited social proof biases even when prompted with reports of human behaviors that were clearly non-indicative of individuals' actual preferences. LLMs also displayed loss aversion when it was explicitly described as a bias. Indeed, when prompted with detailed scientific reports, the extent of the bias (i.e., loss aversion) in the scientific report predicted LLMs' own subsequent bias. Scientific papers of biases can thus become self-fulfilling prophecies, at least when it comes to LLMs' responses. The current study goes beyond fleshing out LLM biases and sheds light on the underlying component processes.
Chinese Translation
大型语言模型（LLMs）被发现容易受到一系列社会、情感和认知偏差的影响。我们研究了两种机制，通过这些机制，即使在训练数据中人类偏好没有偏见或被正确分类为偏见的情况下，这些偏差仍然可以产生。第一种机制是基于人类行为的偏差模仿：这涉及到LLMs在行为与偏好逻辑上无关的情况下推断人类偏好。第二种机制是对明确偏见的人类行为的模仿。在四项关注经济偏差的研究中，我们发现ChatGPT-4o和Qwen表现出了社会证明偏差，即使在提示中包含明显不指示个体实际偏好的行为报告时也是如此。LLMs在明确描述为偏见的情况下也表现出了损失厌恶。实际上，当提示包含详细的科学报告时，科学报告中偏差的程度（即损失厌恶）可以预测LLMs后续的偏差。因此，关于偏差的科学论文可以成为自我实现的预言，至少在LLMs的反应方面。本研究超越了对LLM偏差的阐述，揭示了其潜在的组成过程。
cs.CL / 18 / 2608.12340
Class-Structure Preservation Beats Diversity: A Comprehensive Benchmark of Text Augmentation Methods for Imbalanced Text Classification
类结构保留优于多样性：不平衡文本分类的文本增强方法综合基准
Abstract
With the rapid advancement of large language models (LLMs), generative data augmentation has attracted considerable attention for imbalanced text classification in natural language processing. However, no empirical benchmark to date has compared LLM-based augmentation against the embedding-space SMOTE-style retrieval (EmbSMOTE), a strong classical reference for imbalanced classification. In this study, a controlled benchmark of 11 augmentation methods, spanning classical perturbation, embedding-space retrieval, and LLM-based generation, is newly constructed on seven public text classification datasets covering class counts $K=2$-$28$ and imbalance ratios of 1.1 to over 500, evaluated with five random seeds per cell using macro F1, Welch's $t$-tests, five distributional metrics, and an LLM-family sensitivity analysis based on Qwen3-8B. The experimental results reveal that all LLM-based methods are statistically equivalent or inferior to EmbSMOTE, with the performance gap widening monotonically as imbalance increases and reaching $\Delta\text{F1}_\text{macro}\!\approx\!0.063$ on GoEmotions-28. Furthermore, it is observed that surface-level uniqueness has negligible correlation with downstream performance, whereas LLM-specific artifacts, such as text elongation and label-distribution uniformization, are negatively associated with classification accuracy. Compared with six LLM-based and four classical augmentation baselines, these results demonstrate that the effective variable is not surface-level diversity but class-conditional structural fidelity, namely the degree to which augmented samples preserve the class-conditioned geometry of the training distribution. Accordingly, retrieval-based oversampling should be adopted as the default for imbalanced multi-class classification, and a higher empirical bar should be required before LLM-based augmentation is deployed in practice.
Chinese Translation
随着大型语言模型（LLMs）的快速发展，生成性数据增强在自然语言处理中的不平衡文本分类中引起了相当大的关注。然而，迄今为止尚无实证基准将基于LLM的增强与嵌入空间SMOTE风格检索（EmbSMOTE）进行比较，后者是一个强有力的不平衡分类经典参考。在本研究中，我们在七个公共文本分类数据集上构建了一个包含11种增强方法的受控基准，这些方法涵盖了经典扰动、嵌入空间检索和基于LLM的生成，数据集的类数范围为$K=2$-$28$，不平衡比率从1.1到超过500。使用每个单元的五个随机种子，通过宏F1、Welch的$t$检验、五个分布度量以及基于Qwen3-8B的LLM家族敏感性分析进行评估。实验结果表明，所有基于LLM的方法在统计上与EmbSMOTE等效或表现较差，随着不平衡程度的增加，性能差距单调扩大，在GoEmotions-28上达到$ ext{F1}_ ext{macro} ext{差距} ext{约} ext{为} ext{0.063}$。此外，观察到表面层的独特性与下游性能几乎没有相关性，而LLM特有的伪影，如文本延长和标签分布均匀化，与分类准确性呈负相关。与六种基于LLM和四种经典增强基线相比，这些结果表明，有效变量不是表面层的多样性，而是类条件结构的保真度，即增强样本保留训练分布的类条件几何形状的程度。因此，基于检索的过采样应作为不平衡多类分类的默认方法，并且在实际应用中应对基于LLM的增强设定更高的实证标准。
cs.CL / 19 / 2608.12341
The "Knowledge-Behavior Gap" in Cultural Taboo Safety of Large Language Models
大型语言模型中文化禁忌安全的“知识-行为差距”
Abstract
Cultural taboo safety is essential for deploying large language models (LLMs), as culturally insensitive outputs may cause offense or even social harm. However, existing cultural benchmarks primarily assess cultural knowledge or values biases, while overlooking whether LLMs can recognize and respect cultural taboos, especially when taboos are implicitly hidden in seemingly harmless questions. Besides, cultural taboos are implicit, and context-dependent, thus poss unique challenges for reliable evaluation. To address these gaps, we introduce \textbf{CulShield}, the first public benchmark dedicated to evaluating and improving the cultural taboo safety of LLMs. CulShield spans 77 countries and territories, and includes over 2,020 taboos. It evaluates models along both explicit knowledge and implicit behaviors. Experiments on several advanced LLMs (e.g., GPT-4o-mini, Gemini-2.5-pro) reveal a clear ``knowledge-behavior gap'': models often fail to apply known taboos during interaction. We further show that variations in linguistic context can significantly affect LLMs' cultural taboo safety. Code and data is accessible here: https://github.com/hedyHe/CulShield.
Chinese Translation
文化禁忌安全对于部署大型语言模型（LLMs）至关重要，因为文化上不敏感的输出可能会引发冒犯甚至社会伤害。然而，现有的文化基准主要评估文化知识或价值观偏见，而忽视了LLMs是否能够识别和尊重文化禁忌，尤其是在禁忌隐含在看似无害的问题中时。此外，文化禁忌是隐性的，且依赖于上下文，因此在可靠评估方面面临独特挑战。为了解决这些问题，我们推出了 extbf{CulShield}，这是第一个专门用于评估和改善LLMs文化禁忌安全性的公共基准。CulShield涵盖77个国家和地区，包含超过2020个禁忌。它从显性知识和隐性行为两个方面评估模型。在对多个先进LLMs（如GPT-4o-mini、Gemini-2.5-pro）的实验中，揭示了明显的“知识-行为差距”：模型在互动中往往未能应用已知的禁忌。我们进一步表明，语言上下文的变化可以显著影响LLMs的文化禁忌安全性。代码和数据可在此访问：https://github.com/hedyHe/CulShield。
cs.CL / 20 / 2608.12342
Are Large Language Models Reliable Reviewers? A Benchmark for Error Detection in Financial Documents
大型语言模型是否可靠作为审稿人？金融文件错误检测的基准测试
Abstract
Ensuring the accuracy of financial documents is critical for economic analysis, regulatory compliance, and corporate decision-making. Several studies have shown that Large Language Models (LLMs) perform well in many financial tasks, such as stock price movements and financial analytics. However, a critical task remains unexplored: the ability of LLMs to identify errors in financial documents. In this paper, we introduce \textbf{FinED-Bench}, the first publicly \textbf{Bench}mark for \textbf{Fin}ancial \textbf{E}rror \textbf{D}etection across three levels of cognitive complexity. FinED-Bench covers nine real-world financial scenarios, and includes over 900 documents reported in 2025 that are unseen by existing language models. We detail the benchmark construction process and evaluate several advanced LLMs (e.g., GPT-4o, Qwen3-14B) on this tasks, which requires both financial domain knowledge and reasoning capabilities. Experimental results show that current LLMs still struggle with this task, especially in high-complexity cases. Besides, supervised fine-tuning can significantly improve the performance of weaker LLMs on this task. Our data and code are available at https://github.com/hedyHe/FinED-Bench.
Chinese Translation
确保金融文件的准确性对于经济分析、合规监管和企业决策至关重要。多项研究表明，大型语言模型（LLMs）在许多金融任务中表现良好，例如股票价格变动和金融分析。然而，一个关键任务尚未被探索：LLMs识别金融文件中错误的能力。在本文中，我们介绍了 extbf{FinED-Bench}，这是第一个公开的 extbf{金融错误检测（FinED）}基准测试，涵盖三个认知复杂性层级。FinED-Bench涵盖九个真实的金融场景，并包括2025年报告的900多份文档，这些文档在现有语言模型中未被见过。我们详细描述了基准构建过程，并对几种先进的LLMs（例如，GPT-4o，Qwen3-14B）在这一任务上的表现进行了评估，该任务需要金融领域知识和推理能力。实验结果表明，当前的LLMs在这一任务上仍然存在困难，尤其是在高复杂性案例中。此外，监督微调可以显著提高较弱LLMs在此任务上的表现。我们的数据和代码可在https://github.com/hedyHe/FinED-Bench获取。
cs.CL / 21 / 2608.12343
Large Language Models Pass the History Exam But Miss the <>: A Polish High School Exit Exam Matura Benchmark
大型语言模型通过历史考试但错过了《历史》：波兰高中毕业考试Matura基准
Abstract
AI chatbots are widely used by students as knowledge sources, yet LLM benchmarks rarely assess interpretative historical reasoning. We evaluate eight leading LLMs on the Polish high school exit exams (Matura) in history - three official papers from 2023-2025, comprising short-answer questions and extended essays - comparing model performance against the human examinee population. Every model dramatically outperforms human examinees, yet aggregate scores mask distinct competency profiles: rankings are unstable across task type, source modality, and geographical scope, with a consistent penalty on Polish versus Global history content. Qualitative error analysis reveals two recurring failure modes - source conflation, in which models reason from source content rather than treating it as an object of analysis, and temporal disorientation, in which responses are historically misplaced. This study introduces the first LLM history benchmark grounded in Polish national curriculum.
Chinese Translation
人工智能聊天机器人被学生广泛用作知识来源，但大型语言模型（LLM）基准测试很少评估解释性历史推理。我们对八个领先的LLM在波兰高中毕业考试（Matura）中的历史部分进行评估——包括2023-2025年的三份官方试卷，涵盖简答题和扩展论文——并将模型表现与人类考生群体进行比较。每个模型的表现都显著优于人类考生，但整体得分掩盖了不同的能力特征：在任务类型、来源模态和地理范围之间，排名不稳定，并且波兰历史内容相较于全球历史内容始终受到惩罚。定性错误分析揭示了两种反复出现的失败模式——来源混淆，即模型从来源内容推理，而不是将其视为分析对象，以及时间错位，即回应在历史上被错误定位。本研究介绍了第一个基于波兰国家课程的大型语言模型历史基准。
cs.CL / 22 / 2608.12344
Predicting consumer-technology ownership without a diffusion history
在没有扩散历史的情况下预测消费者技术的拥有情况
Abstract
We test whether the perceived attributes of a consumer technology predict how widely it is owned. In a 2022 Prolific survey of US adults (n = 678), respondents rated 65 consumer technologies on six attributes. We then elicited the same ratings from two frontier language models, Anthropic Claude Opus 4.7 and OpenAI GPT-5.5. We regress ownership prevalence on four UTAUT2 acceptance attributes plus a log-age covariate with a sign-constrained penalized regression and evaluate it by holding out one technology at a time. The attribute model improves on a baseline of years-since-launch: mean absolute error falls by 17% with the human ratings, and by more with either model, most with Opus 4.7. Over the short 2022-to-2025 window, where ownership moved little, the same attributes do not improve on a no-change baseline. We set out the limitations of the approach, including the possibility that language-model ratings reflect prior knowledge of these technologies rather than independent attribute reasoning. We include a deployment illustration: 2027 ownership predictions for eleven products launched in 2025 and 2026.
Chinese Translation
我们测试了消费者技术的感知属性是否能够预测其拥有的广泛程度。在2022年对美国成年人进行的Prolific调查中（样本量n = 678），受访者对65种消费者技术的六个属性进行了评分。随后，我们从两个前沿语言模型Anthropic Claude Opus 4.7和OpenAI GPT-5.5中获取了相同的评分。我们使用带有符号约束的惩罚回归将拥有率回归于四个UTAUT2接受属性以及一个对数年龄协变量，并通过每次排除一种技术来评估模型。属性模型在以自推出年份为基线的基础上有所改进：与人类评分相比，平均绝对误差降低了17%，而与任一模型相比，降低幅度更大，尤其是Opus 4.7。在2022至2025年的短期窗口内，尽管拥有情况变化不大，但相同的属性并未改善无变化基线。我们阐述了该方法的局限性，包括语言模型评分可能反映了对这些技术的先前知识，而非独立的属性推理。我们还包括了一个部署示例：对2025年和2026年推出的十一种产品在2027年的拥有预测。
cs.CL / 23 / 2608.12361
New Terms, New Toxicity: Consensus-based Chinese Neologism Toxicity Detection via Search-Augmented LLMs
新术语，新毒性：基于共识的中文新词毒性检测通过搜索增强的语言模型
Abstract
Neologisms, emerging terms in meaning or form, can serve as new vehicles for toxic expression, like "country girl" as a stigmatizing label targeting feminism. Such toxic neologisms appear benign but have evolved into toxic usage in public consensus, posing challenges to moderation systems and remaining underexplored. In this paper, we investigate how to detect implicit toxicity expressed via neologisms. We first propose a taxonomy that captures the origins and consensus-verification criteria of toxic neologisms, followed by the construction of a lexicon spanning widely observed risk categories. To capture toxicity grounded in public consensus, we introduce SeTox, a search-augmented framework that enables static large language models (LLMs) to incorporate real-time web context for neologism toxicity detection. Experiments show that SeTox, even with 3B-scale models, outperforms recent large-scale models, demonstrating its scalability to incorporate real-world knowledge for toxic neologism detection. Disclaimer: this paper has offensive contents that may be disturbing to some readers.
Chinese Translation
新词是指在意义或形式上出现的新术语，它们可以作为毒性表达的新载体，例如“乡村女孩”作为针对女性主义的污名化标签。这类毒性新词看似无害，但在公众共识中已演变为毒性使用，给内容审核系统带来了挑战，并且仍然未得到充分研究。本文探讨如何检测通过新词表达的隐性毒性。我们首先提出一个分类法，捕捉毒性新词的起源和共识验证标准，随后构建一个涵盖广泛观察到的风险类别的词汇表。为了捕捉基于公众共识的毒性，我们引入了SeTox，一个搜索增强框架，使静态大型语言模型（LLMs）能够结合实时网络上下文进行新词毒性检测。实验表明，即使使用3B规模的模型，SeTox也优于近期的大规模模型，展示了其在毒性新词检测中整合现实世界知识的可扩展性。免责声明：本文包含可能对某些读者造成困扰的冒犯性内容。
cs.CL / 24 / 2608.12374
Are you Talking Logic to Me? Assessing Language Models Syllogistic Reasoning Capabilities
你在跟我谈逻辑吗？评估语言模型的三段论推理能力
Abstract
Language models (LMs) struggle with logical tasks like reasoning on syllogisms. It has been shown that Knowledge Representation (KR) plays a crucial role in expressing input information to help models solve tasks. This observation motivates our study of the impact of different formal KR notations on syllogistic reasoning by extending the FOLIO and P-FOLIO datasets. Our experiments on Small Language Models (SLMs) in Supervised Fine-Tuning (SFT) and Zero-Shot (ZS) settings show that the choice of input notation can yield performances competitive with natural language while enabling faster inference. We also propose a syllogistic categorization method (SEF) and use it to enrich ZS prompts with logical definitions, which boost reasoning in small models. We open-source our framework, Common Logic Grammar Construction (CLGC), as the first Python library for automatically generating syllogisms in KR notations and defining their SEF categories.
Chinese Translation
语言模型（LMs）在逻辑任务上存在困难，例如对三段论的推理。研究表明，知识表示（KR）在表达输入信息以帮助模型解决任务中发挥着关键作用。这一观察促使我们研究不同形式的KR符号对三段论推理的影响，通过扩展FOLIO和P-FOLIO数据集。我们在监督微调（SFT）和零样本（ZS）设置下对小型语言模型（SLMs）进行的实验表明，输入符号的选择可以产生与自然语言相媲美的性能，同时实现更快的推理。我们还提出了一种三段论分类方法（SEF），并利用它丰富ZS提示中的逻辑定义，从而提升小型模型的推理能力。我们开源了我们的框架——通用逻辑语法构建（CLGC），作为第一个用于自动生成KR符号中的三段论并定义其SEF类别的Python库。
cs.CL / 25 / 2608.12387
Query Timing Produces Opposite Positional Biases Between LLMs and Humans
查询时机在大型语言模型与人类之间产生相反的位置信息偏差
Abstract
Positional biases such as recency and primacy effects have been documented in large language models (LLMs), yet the underlying mechanism by which these models make their evaluations remains poorly understood. Both primacy and recency biases have been observed in human judgments in response to evidence, but recent work suggest that \emph{when} the listener updates their beliefs -- during the presentation of evidence or only at the end -- influences the presence of such effects. We investigate whether a similar phenomenon holds for LLMs, finding divergence from human behavior. These biases are more exacerbated in newer models compared to their predecessors.
Chinese Translation
大型语言模型（LLMs）中已经记录了诸如近期效应和首因效应等位置信息偏差，但这些模型进行评估的潜在机制仍然不甚明了。人类在对证据的判断中观察到了首因偏差和近期偏差，但最近的研究表明，听众更新其信念的时机——是在证据呈现期间还是仅在结束时——会影响这些效应的存在。我们研究了这一现象是否同样适用于LLMs，发现其与人类行为存在差异。这些偏差在较新的模型中比其前辈更为明显。
cs.CL / 26 / 2608.12391
Unified Multi-Dimensional Benchmark for Complex Graph Reasoning in Large Language Models
大型语言模型中复杂图推理的统一多维基准
Abstract
Graph reasoning provides a promising testbed for evaluating the reasoning ability of large language models (LLMs), as graph instances can be programmatically generated, structurally controlled, and naturally scaled to long-input settings. However, existing graph reasoning benchmarks have limited coverage of data complexity, rely heavily on manual construction, and lack unified evaluation across text-based and code-based reasoning modes. To address these limitations, we propose {\dataset}, a five-stage \textit{semi-automatic} framework for constructing complex graph reasoning benchmarks. It expands benchmark coverage along five dimensions: \textit{Graph Size}, \textit{Task Complexity}, \textit{Task Description}, \textit{Graph Loading}, and \textit{Task Source}. The framework uses an LLM-based data generator to automatically produce task descriptions, graph data, reference solutions, graph-loading scripts, question forms, and evaluation scripts, while retaining human validation at key quality-control stages. Based on it, we construct a benchmark with $202$ tasks and evaluate LLMs under text-based, code-based, and augmented reasoning settings. Experiments show that the complexity dimensions reveal model limitations that are less visible in existing benchmarks; existing fine-tuned models struggle to generalize to GraphGym, whereas retrieval-augmented methods show scenario-dependent adaptability, improving textual reasoning but not consistently improving coding reasoning. These findings suggest that ours serves as a challenging and diagnostic benchmark for graph reasoning and provides empirical guidance for future enhancement methods. Code and dataset will be published soon.
Chinese Translation
图推理为评估大型语言模型（LLMs）的推理能力提供了一个有前景的测试平台，因为图实例可以通过程序生成、结构控制，并自然扩展到长输入设置。然而，现有的图推理基准在数据复杂性方面覆盖有限，过于依赖手动构建，并且缺乏文本和代码推理模式之间的统一评估。为了解决这些局限性，我们提出了{ extit{dataset}}，一个五阶段的 extit{半自动}框架，用于构建复杂图推理基准。该框架在五个维度上扩展了基准覆盖： extit{图大小}、 extit{任务复杂性}、 extit{任务描述}、 extit{图加载}和 extit{任务来源}。该框架使用基于LLM的数据生成器自动生成任务描述、图数据、参考解决方案、图加载脚本、问题形式和评估脚本，同时在关键质量控制阶段保留人工验证。基于此，我们构建了一个包含$202$个任务的基准，并在文本推理、代码推理和增强推理设置下评估LLMs。实验表明，复杂性维度揭示了模型在现有基准中不易察觉的局限性；现有的微调模型在GraphGym上难以泛化，而检索增强方法则显示出场景依赖的适应性，改善了文本推理，但并未一致改善编码推理。这些发现表明，我们的基准为图推理提供了一个具有挑战性和诊断性的评估工具，并为未来的增强方法提供了实证指导。代码和数据集将很快发布。
cs.CL / 27 / 2608.12486
DIVE: Unlocking Self-Improvement in Frozen Language Models Through Diversity-Driven Skill Evolution
DIVE：通过多样性驱动的技能演化解锁冻结语言模型的自我提升
Abstract
Large language models (LLMs) cannot retain post-deployment experience without parameter updates. We introduce DIVE, a diversity-driven framework that enables frozen LLMs to improve by evolving persistent natural-language skills from task experience and verifier feedback. These skills encode reusable reasoning procedures, verification strategies, common failure modes, and output constraints and are both executed and revised by the same underlying model without access to a teacher model. Since natural-language skill evolution is a stochastic, non-convex search process, optimizing a single skill trajectory can overfit to sampled experience or converge to a suboptimal solution. DIVE mitigates this optimization variance by independently evolving multiple skill populations from bootstrapped experience, adaptively refining them through diverse transformations, and jointly selecting a complementary set of skills. Across six mathematical and logical reasoning tasks and multiple model families, DIVE consistently outperforms existing reasoning methods, prompt-optimization approaches, skill-development frameworks, and memory-based baselines. It achieves rapid self-improvement from accumulated experience, obtaining substantially larger performance gains with fewer rollouts than parameter-based methods such as SFT and GRPO, and prompt optimization with GEPA. Further, the resulting skills transfer across model scales and families, enabling smaller models such as GPT-5-nano to match or outperform larger counterparts, i.e., GPT-5, under conventional prompting. These results establish diversity-driven skill evolution as an effective, interpretable, and parameter-free approach to LLM self-improvement.
Chinese Translation
大型语言模型（LLMs）在部署后无法保留经验而不进行参数更新。我们提出了DIVE，一个多样性驱动的框架，使冻结的LLMs能够通过从任务经验和验证者反馈中演化持久的自然语言技能来实现自我提升。这些技能编码了可重用的推理过程、验证策略、常见失败模式和输出约束，并由同一底层模型执行和修订，而无需访问教师模型。由于自然语言技能演化是一个随机的、非凸的搜索过程，优化单一技能轨迹可能会对采样经验过拟合或收敛到次优解。DIVE通过独立演化多个技能群体，从引导经验中自适应地通过多样化变换进行精炼，并共同选择互补的技能集，从而减轻了这种优化方差。在六个数学和逻辑推理任务以及多个模型家族中，DIVE始终优于现有的推理方法、提示优化方法、技能发展框架和基于记忆的基线。它通过积累经验实现快速自我提升，在比基于参数的方法（如SFT和GRPO）和使用GEPA的提示优化中，获得了显著更大的性能提升，且所需的回合数更少。此外，所获得的技能在模型规模和家族之间迁移，使得较小的模型（如GPT-5-nano）在常规提示下能够匹配或超越较大的模型（如GPT-5）。这些结果确立了多样性驱动的技能演化作为一种有效、可解释且无参数的LLM自我提升方法。
cs.CL / 28 / 2608.12598
Intensional Anaphora
意向性指代
Abstract
Intensional operators are often treated as quantifiers over possible worlds, parallel to the treatment of determiners as quantifiers over individuals. Yet individuals introduced in intensional contexts cannot serve as antecedents to later pronouns as easily as those introduced in merely quantificational contexts. For instance, "Everyone is eating a cheeseburger" may be followed by "They are large", where "they" refers to the cheeseburgers being eaten. However, as Stone (1999) points out, the similar "Andrea might be eating a cheeseburger" does not support later anaphoric references such as "It is large" or "They are large". Stone (1999), Stone and Hardt (1999), and Brasoveanu (2010) address this by requiring a pronoun's value (its referents) to exist in the world of evaluation, ruling out anaphora from non-veridical intensional contexts. We show, however, both cases where such anaphora is disallowed even when the pronoun's referents clearly exist and cases where it is allowed even though they might not exist. We argue that intensional anaphora is best captured using a description-based rather than value-based account. A pronoun presupposes that its corresponding antecedent description is instantiated in each world of the context set. Thus, there must be a cheeseburger being eaten by Andrea in every candidate world for "It is large" to be felicitous after "Andrea might be eating a cheeseburger". We implement our proposal via a new logic, building on Keshet (2018) and Abney and Keshet (2022), called Plural Intensional Presuppositional predicate calculus (PIP). Each PIP formula translates directly into standard first-order predicate calculus with set abstraction, providing a classical foundation for this work.
Chinese Translation
意向性运算符通常被视为对可能世界的量化，类似于对个体的量化处理。然而，在意向性语境中引入的个体并不能像在单纯量化语境中引入的个体那样轻易地作为后续代词的先行词。例如，“每个人都在吃一个芝士汉堡”可以接着说“它们很大”，其中“它们”指的是正在被吃的芝士汉堡。然而，正如Stone（1999）所指出的，类似的句子“安德烈可能在吃一个芝士汉堡”并不支持后续的指代引用，例如“它很大”或“它们很大”。Stone（1999）、Stone和Hardt（1999）以及Brasoveanu（2010）通过要求代词的值（其指称）存在于评估世界中来解决这一问题，从而排除了来自非真实意向性语境的指代。然而，我们展示了即使代词的指称显然存在的情况下，也有不允许这种指代的例子，以及即使它们可能不存在的情况下，也有允许这种指代的例子。我们认为，意向性指代最好通过基于描述而非基于值的方式来捕捉。代词预设其对应的先行描述在上下文集的每个世界中都被实例化。因此，必须在每个候选世界中有一个芝士汉堡被安德烈吃掉，才能使“安德烈可能在吃一个芝士汉堡”之后的“它很大”成立。我们通过一种新的逻辑实现我们的提案，该逻辑基于Keshet（2018）和Abney与Keshet（2022），称为复数意向性预设谓词演算（Plural Intensional Presuppositional predicate calculus，PIP）。每个PIP公式直接转换为具有集合抽象的标准一阶谓词演算，为这项工作提供了经典基础。
cs.CL / 29 / 2608.12623
When Explanations Betray Backdoors: Black-Box Auditing for Language Model Classifiers
当解释背叛后门：语言模型分类器的黑箱审计
Abstract
Language model classifiers with explanations are used for moderation, routing, topic triage, and low-resource annotation. We study black-box auditing when the defender has only clean calibration data without trigger information but can ask the classifier for a label plus a short rationale or quoted evidence. We introduce Groundedness Drift, a lightweight score measuring whether the answer summary remains grounded in the input. Across two 7B backbones, five datasets, and four common non-adaptive OpenBackdoor-style attack families, Groundedness Drift achieves higher AUROC and lower residual target ASR than every compared detector in all cases at a nominal 5\% clean-FPR budget. We then evaluate Unsupported Groundedness, a multi-probe escalation for explanation-camouflage stress cases. Unsupported Groundedness improves signals but does not close the adaptive gap.
Chinese Translation
带有解释的语言模型分类器被用于内容审核、路由、主题分类和低资源注释。我们研究了黑箱审计的情况，其中防御者仅拥有干净的校准数据而没有触发信息，但可以向分类器请求标签以及简短的理由或引用证据。我们引入了“基础漂移”（Groundedness Drift），这是一种轻量级评分，衡量答案摘要是否仍然与输入保持一致。在两个7B主干模型、五个数据集和四种常见的非自适应OpenBackdoor风格攻击家族中，“基础漂移”在所有情况下都实现了比比较的检测器更高的AUROC和更低的残余目标ASR，且在名义上5\%的干净假阳性率预算下表现优异。随后，我们评估了“无支持基础性”（Unsupported Groundedness），这是一种针对解释伪装压力案例的多探测升级。无支持基础性改善了信号，但未能缩小自适应差距。
cs.CL / 30 / 2608.12626
LLMs Are Not Good Strategists, Yet Memory-Enhanced Agency Boosts Reasoning
大型语言模型并不是优秀的战略家，但增强记忆的代理提升了推理能力
Abstract
Strategic reasoning in Large Language Models (LLMs) within long-horizon environments is often limited by inconsistent subgoals. In these settings, finite attention resources prevent the model from maintaining strategic coherence over thousands of steps. This limitation leads to strategic drift, where localized decisions fail to sustain a coherent trajectory across reasoning. To address this, we introduce EpicStar, a framework that enables agents to learn memory as policy to tackle long-horizon reasoning. Specifically, the agent maintains a bank of successful past episodes as a heuristic alongside a working memory to track short-term environmental changes. During inference, a dynamic gating mechanism determines whether to execute a retrieved action directly or to perform new reasoning through a contextual fusion of the retrieved episodes and current working memory. Utilizing StarCraft II as the testbed, we evaluated EpicStar against diverse opponent styles. It significantly outperforms baseline methods, achieving higher win rates while consuming an order of magnitude fewer tokens, and it maintains this advantage consistently across difficulty levels and opponent strategies. Our findings provide compelling evidence that structured cross-episode memory is essential for enabling LLM agents to perform robust, long-term strategic execution in dynamic, autonomous settings.
Chinese Translation
在长时间跨度的环境中，大型语言模型（LLMs）的战略推理常常受到不一致的子目标的限制。在这些环境中，有限的注意资源使得模型无法在数千个步骤中保持战略一致性。这一限制导致了战略漂移，即局部决策未能维持跨推理的一致轨迹。为了解决这个问题，我们引入了EpicStar，一个使代理能够学习将记忆作为策略来应对长时间跨度推理的框架。具体而言，代理维护一个成功过去事件的库作为启发式信息，并使用工作记忆来跟踪短期环境变化。在推理过程中，一个动态门控机制决定是直接执行检索到的动作，还是通过将检索到的事件与当前工作记忆进行上下文融合来进行新的推理。我们以《星际争霸 II》（StarCraft II）作为测试平台，评估了EpicStar在不同对手风格下的表现。它显著优于基线方法，赢得更高的胜率，同时消耗的令牌数量减少了一个数量级，并且在不同的难度级别和对手策略中始终保持这一优势。我们的研究结果提供了有力的证据，表明结构化的跨事件记忆对于使LLM代理在动态自主环境中执行稳健的长期战略至关重要。
cs.CL / 31 / 2608.12630
Novels generated by language models show compressed formal variation
由语言模型生成的小说表现出压缩的形式变异
Abstract
While large language models can generate entire novels, there is little information about the level of formal variation in their output over many generations. Rather than asking whether individual passages can be identified as AI-generated, this study asks whether repeated AI generation can produce the same range of diversity which is found across human corpora. This paper contrasts six corpora based on generation source and target style: twenty novels generated using GPT-5.5 Thinking in a nineteenth-century British realist style, twenty novels generated using Qwen3-14B in a nineteenth-century British realist style, twenty novels generated using each of these models in a contemporary zero style, 205 nineteenth-century human-written British novels, and sixty-five contemporary human-written Zero-Style novels. At the document level, the research includes MATTR-500, Shannon entropy, average sentence length, readability, and punctuation rate measurements. The most robust and reliable result is compression of sentence structure. Repeated generations produce novels that vary far less from one another in sentence structure than human novels do. Compression is also present in the measures of readability, punctuation, and sentence length variability within novels. Lexical measures tend to be similarly compressed, with the exception of Qwen Zero-Style MATTR. Despite having distinct mean stylistic profiles, GPT and Qwen lack a stable pattern of cross-measure correlation. This article therefore distinguishes between variance overclosure, which represents a limited formal range between novels, and a more specific phenomenon of correlational overclosure. This means that an individual AI-generated novel may resemble human fiction stylistically, while a collection of AI-generated novels occupies a much narrower formal range.
Chinese Translation
虽然大型语言模型能够生成整部小说，但关于其输出在多次生成过程中的形式变异程度的信息仍然很少。本研究并不询问个别段落是否可以被识别为AI生成，而是探讨重复的AI生成是否能够产生与人类语料库中所发现的相同范围的多样性。本文对六个语料库进行了对比，这些语料库基于生成来源和目标风格：使用GPT-5.5生成的二十部以十九世纪英国现实主义风格创作的小说，使用Qwen3-14B生成的二十部以十九世纪英国现实主义风格创作的小说，使用这两种模型生成的二十部当代零风格小说，205部十九世纪人类创作的英国小说，以及65部当代人类创作的零风格小说。在文档层面，研究包括MATTR-500、香农熵、平均句子长度、可读性和标点率的测量。最稳健和可靠的结果是句子结构的压缩。重复生成的小说在句子结构上的变异远低于人类小说。压缩现象也出现在小说的可读性、标点和句子长度变异性测量中。词汇测量往往也表现出类似的压缩，除了Qwen零风格的MATTR。尽管GPT和Qwen具有明显不同的平均风格特征，但它们缺乏稳定的跨测量相关模式。因此，本文区分了变异过度封闭（variance overclosure），即小说之间形式范围的有限性，以及更具体的相关性过度封闭现象（correlational overclosure）。这意味着一部AI生成的小说在风格上可能类似于人类小说，而一组AI生成的小说则占据了更狭窄的形式范围。
cs.CL / 32 / 2608.12652
Excess Separability: Nuisance-Controlled Residual-Stream Probing for Benchmark Contamination Detection
过度可分离性：控制干扰的残差流探测基准污染检测
Abstract
Benchmark contamination is diagnosed today with n-gram overlap, with likelihood-based membership inference, or with canary strings, and each needs something usually unavailable: the training corpus, a well-chosen test statistic, or foresight at dataset release. A recent alternative reads contamination off a linear probe on internal activations. We show that the natural way to do this does not work, and specify one that survives measurement. The protocol reports a zero-sum contrast on the depth profile of probe accuracy, recentred on a level-matched placebo baseline, tested against a label-permutation null, with the reference set twice the size of the suspect set. Each choice replaces a simpler alternative we measured and rejected. Reporting the level of excess separability rather than its shape makes the false positive rate track the size of the analyst's own control set, from 0.03 to 0.99 under a true null. Contrasting against a flat depth profile fails in both directions, rejecting a true null 0.72 of the time when surface decodability rises with depth and losing all power when it falls. An item bootstrap holds the fitted probe fixed and rejects up to 0.09 of the time where a permutation null that refits it holds 0.02. A half-size baseline triples the error rate. On real transformers, baseline depth profiles are measurably not flat, spanning up to 29.1 accuracy points on a temporal split, and their non-flatness tracks the surface difference between the item sets (correlation 0.87 over 6 audits), so the correction is largest exactly where it is needed. All 4 well-matched Pile arms return null, and the protocol refuses a verdict on the temporal split rather than reporting one. What this does not establish is whether transformers carry a familiarity direction at all: the only positive sits on the split where exchangeability fails. Implementation, tests and audits are released.
Chinese Translation
目前，基准污染的诊断依赖于 n-gram 重叠、基于似然的成员推断或金丝雀字符串，而每种方法通常需要一些不可用的东西：训练语料库、精心选择的测试统计量或数据集发布时的前瞻性。最近的一种替代方法通过对内部激活进行线性探测来读取污染。我们展示了自然的做法并不奏效，并指定了一种能够经受测量的方法。该协议报告了探测准确度的深度轮廓上的零和对比，重新聚焦于与水平匹配的安慰剂基线，并针对标签置换的零假设进行测试，参考集的大小是可疑集的两倍。每个选择都替代了我们测量并拒绝的更简单的替代方案。报告过度可分离性的水平而非其形状，使得假阳性率与分析师自身控制集的大小相匹配，在真实零假设下从 0.03 到 0.99。与平坦的深度轮廓进行对比在两个方向上都失败，当表面可解性随深度上升时拒绝真实零假设的概率为 0.72，而当其下降时则失去所有能力。项目自助法保持拟合探测器不变，并在置换零假设重新拟合时拒绝的概率高达 0.09，而后者保持在 0.02。半大小基线使错误率增加三倍。在真实变换器上，基线深度轮廓在可测量上并不平坦，在时间分割上跨度达到 29.1 个准确度点，并且它们的非平坦性与项目集之间的表面差异相关（在 6 次审计中相关性为 0.87），因此修正正是需要的地方最大。所有 4 个匹配良好的 Pile 臂返回零假设，协议拒绝在时间分割上做出裁决，而不是报告一个。这并未确定变换器是否完全携带熟悉度方向：唯一的积极结果出现在可交换性失败的分割上。实施、测试和审计结果已发布。
cs.CL / 33 / 2608.12720
ERSkill: Evolving for Skill-Guided Adaptive Memory Retrieval
ERSkill：为技能引导的自适应记忆检索而演化
Abstract
While Large Language Model (LLM) agents increasingly rely on long-term memory for persistent interactions, the retrieval mechanisms governing this memory are rarely treated as evolvable components. This static approach limits performance on heterogeneous memory queries, which often demand diverse evidence construction strategies. To address this, we introduce \textbf{ERSkill}, a retrieval-centric framework for self-evolving, skill-guided memory access. ERSkill compiles interaction histories into a structured memory store and represents retrieval behaviors as executable skills composed of fundamental primitives. At inference time, a trained router dynamically matches each query to the optimal skill to construct tailored evidence for answer generation. To enable continuous improvement, ERSkill co-evolves the skill set and the router during training. It employs an experience trie to efficiently record explored retrieval paths, alongside a double-frontier mechanism that safely decouples the expansion of new skill capabilities from stable, router-facing deployment. Experiments across multiple agent memory benchmarks demonstrate that ERSkill substantially outperforms strong non-evolving and self-evolving baselines. Notably, it improves the overall average across F1, BLEU-1, and LLM-judge scores by 31.3\% with Qwen3-Next-80B-A3B-Instruct and by 28.1\% with GPT-5.4-nano.
Chinese Translation
随着大型语言模型（LLM）代理越来越依赖长期记忆进行持续交互，支配这种记忆的检索机制很少被视为可演化的组件。这种静态方法限制了在异构记忆查询上的性能，而这些查询通常需要多样化的证据构建策略。为了解决这个问题，我们提出了 extbf{ERSkill}，一个以检索为中心的自我演化、技能引导的记忆访问框架。ERSkill将交互历史编译成结构化的记忆存储，并将检索行为表示为由基本原语组成的可执行技能。在推理时，经过训练的路由器动态地将每个查询匹配到最佳技能，以构建量身定制的证据以生成答案。为了实现持续改进，ERSkill在训练过程中共同演化技能集和路由器。它采用经验前缀树有效记录探索过的检索路径，并结合双前沿机制安全地将新技能能力的扩展与稳定的、面向路由器的部署解耦。多个代理记忆基准的实验表明，ERSkill显著优于强大的非演化和自演化基线。值得注意的是，使用Qwen3-Next-80B-A3B-Instruct时，它在F1、BLEU-1和LLM-judge分数上的整体平均提高了31.3\%，而使用GPT-5.4-nano时提高了28.1\%。
cs.CL / 34 / 2608.12750
PatientAct: Theory-Grounded Mental Health Client Simulation
PatientAct：基于理论的心理健康客户模拟
Abstract
LLM-based simulated clients are increasingly used to train novice counselors, evaluate LLM therapists, and generate synthetic data. However, current simulators produce overly cooperative clients that disclose too readily, accept therapeutic reframes without resistance, and resolve core issues within a single session. We trace these issues to profiles that lack causal depth and behavioral mechanisms that treat all content as equally accessible. We present PatientAct, a framework for client simulation grounded in established clinical theories. Our profiles integrate the 5Ps clinical case formulation, providing causal depth without tying the design to any single therapeutic modality. During simulation, profiles include a dynamic memory layer in which items carry trust thresholds (e.g., symptoms are available early, whereas formative memories require a sustained therapeutic alliance). At each turn, the client's emotional reaction and behavior are modeled before generating a response. If the therapist approaches gated content, PatientAct expresses resistance in terms of quantity, content, and style rather than defaulting to cooperation or a single resistance pattern. We evaluate our framework on 40 clinical situations and demonstrate that it generates diverse profiles with high clinical plausibility. Moreover, PatientAct significantly outperforms the baselines, yielding substantial gains in resistance quality and behavioral realism. Our code and data will be publicly available via github.com/Sahandfer/PatientHub.
Chinese Translation
基于大型语言模型（LLM）的模拟客户在培训新手咨询师、评估LLM治疗师和生成合成数据方面的应用日益增多。然而，当前的模拟器生成的客户过于合作，过于轻易地透露信息，毫无抵抗地接受治疗性重构，并在单次会话中解决核心问题。我们将这些问题追溯到缺乏因果深度的档案以及将所有内容视为同等可接近的行为机制。我们提出了PatientAct，一个基于已建立临床理论的客户模拟框架。我们的档案整合了5Ps临床案例构建，提供了因果深度，而不将设计绑定于任何单一的治疗模式。在模拟过程中，档案包括一个动态记忆层，其中项目具有信任阈值（例如，症状早期可用，而形成性记忆需要持续的治疗联盟）。在每个回合中，客户的情感反应和行为在生成响应之前被建模。如果治疗师接触到受限内容，PatientAct会在数量、内容和风格上表现出抵抗，而不是默认合作或单一的抵抗模式。我们在40个临床情境中评估了我们的框架，证明其生成的档案具有高度的临床可信性。此外，PatientAct显著优于基线，在抵抗质量和行为真实感方面取得了显著提升。我们的代码和数据将通过github.com/Sahandfer/PatientHub公开提供。
cs.CL / 35 / 2608.12756
ReconSpan: Reconstruction-Guided Adaptive Latent Tokenization
ReconSpan：重建引导的自适应潜在标记化
Abstract
Adaptive latent tokenization maps a fine-grained input to a shorter sequence of continuous representations associated with input-dependent spans. We introduce ReconSpan, which divides text into chunks that a backward decoder can reconstruct from a single contextual prefix code and retains one such code as the latent token for each chunk. The reconstruction criterion is applied when chunks are formed, allowing one trained autoencoder to produce average chunk lengths from 6.5 to 12.2. At matched average length, reconstruction-guided boundaries preserve more text than random boundaries. Readers of the resulting latent sequence recover topic information reliably but struggle to extract exact details.
Chinese Translation
自适应潜在标记化将细粒度输入映射到与输入相关的跨度所关联的较短连续表示序列。我们提出了ReconSpan，它将文本划分为块，后向解码器可以从单个上下文前缀码重建这些块，并为每个块保留一个这样的代码作为潜在标记。在形成块时应用重建标准，使得一个训练好的自编码器能够生成平均块长度从6.5到12.2。在匹配的平均长度下，重建引导的边界保留的文本比随机边界更多。结果潜在序列的读者能够可靠地恢复主题信息，但在提取确切细节时遇到困难。
cs.CL / 36 / 2608.12776
ViTOED: A Dataset for Target-Oriented Emotion Detection on Vietnamese Social Media Texts
ViTOED：一个用于越南社交媒体文本的目标导向情感检测数据集
Abstract
This paper introduces ViTOED, a novel dataset for target-oriented emotion detection in Vietnamese social media texts. The ViTOED comprises 10,985 user comments and 21,244 manually annotated opinion quadruples (source, target, expression, polarity) that follow strict guidelines. The dataset reveals Vietnamese-specific phenomena, such as implicit sources and targets and vocabulary ambiguities, enabling deeper analysis of user emotions toward entities. We propose a baseline using structured sentiment graphs and evaluate various Vietnamese pre-trained language models. The empirical results highlight challenges in span detection and relation extraction and indicate substantial room for model improvement in Vietnamese Target-Oriented Emotion Detection tasks.
Chinese Translation
本文介绍了ViTOED，一个用于越南社交媒体文本中目标导向情感检测的新数据集。ViTOED包含10,985条用户评论和21,244个手动标注的意见四元组（源、目标、表达、极性），这些标注遵循严格的指南。该数据集揭示了越南特有的现象，如隐含的源和目标以及词汇歧义，从而能够更深入地分析用户对实体的情感。我们提出了一种基线方法，使用结构化情感图，并评估了多种越南预训练语言模型。实证结果突显了跨度检测和关系提取中的挑战，并表明在越南目标导向情感检测任务中模型改进的潜力巨大。
cs.CL / 37 / 2608.12779
CRAFT: LLM-Based Iterative Refinement for Temporal Reasoning over Clinical Narratives
CRAFT：基于大型语言模型的临床叙述时间推理的迭代精炼
Abstract
Understanding the temporal progression of symptoms in clinical narratives is critical for disease monitoring, safety surveillance, and causality assessment. Clinical narratives, however, rarely provide explicit temporal anchors. Current approaches to temporal information reasoning focus predominantly on pairwise relation classification across multi-visit and timestamp-rich records, leaving the reconstruction of structured symptom trajectories from individual anchor-sparse reports largely unaddressed. We propose CRAFT, an LLM framework that pairs a generator with a constraint-based verifier to iteratively produce and refine stage-wise symptom timelines through targeted feedback. We conduct evaluation on MedTempo, a new benchmark of 5,347 vaccine adverse-event narratives spanning three COVID-19 vaccine types, with expert-validated temporal stage annotations for 3,166 reports. Experiments across four LLM backbones demonstrate that CRAFT consistently improves temporal ordering accuracy, with ablation analysis isolating the contribution of generator and verifier components across model capability levels.
Chinese Translation
理解临床叙述中症状的时间进展对于疾病监测、安全监控和因果评估至关重要。然而，临床叙述中很少提供明确的时间锚点。目前的时间信息推理方法主要集中在多次就诊和时间戳丰富记录之间的成对关系分类上，导致从个别缺乏锚点的报告中重建结构化症状轨迹的问题基本未得到解决。我们提出了CRAFT，一个将生成器与基于约束的验证器相结合的LLM框架，通过针对性的反馈迭代生成和精炼阶段性症状时间线。我们在MedTempo上进行了评估，这是一个新的基准数据集，包含5,347个疫苗不良事件叙述，涵盖三种COVID-19疫苗类型，并为3,166个报告提供了专家验证的时间阶段注释。四种LLM基础模型的实验表明，CRAFT在时间排序准确性上始终有所提升，消融分析则隔离了生成器和验证器组件在不同模型能力水平下的贡献。
cs.CL / 38 / 2608.12814
FastThaiG2P: Lightning-fast Thai Grapheme-to-phoneme Conversion for Voice Agent Pipelines
FastThaiG2P：闪电般快速的泰语字素到音素转换用于语音代理管道
Abstract
FastThaiG2P provides sub-millisecond Thai grapheme-to-phoneme conversion for text-to-speech pipelines (International Phonetic Alphabet and Kokoro-TTS conventions) using a PyThaiNLP-tokenized, extensible dictionary and normalization rules for common Central Thai speech. The approach achieves an average latency of 0.15 ms per utterance on a benchmark of 27,242 synthetically generated utterances, of which 30\% is spent on tokenization, 12\% on normalization, and 58\% on out-of-vocabulary fallbacks (0.5\% OOV rate). To demonstrate its effectiveness, we used FastThaiG2P to phonemize Som-TTS, an open dataset containing 20 hours of grapheme-and-audio pairs, then trained an 82M-parameter StyleTTS 2 model based on a Kokoro-TTS recipe. The resulting model vocalizes intelligible Thai speech suitable for prototyping and development at 0.25 real-time factor (4x real-time) with ONNX inference on CPU.
Chinese Translation
FastThaiG2P 提供亚毫秒级的泰语字素到音素转换，适用于文本到语音管道（国际音标和 Kokoro-TTS 规范），使用 PyThaiNLP 分词、可扩展字典和用于常见中央泰语语音的规范化规则。该方法在 27,242 个合成生成的语句的基准测试中实现了每个语句平均延迟 0.15 毫秒，其中 30\% 用于分词，12\% 用于规范化，58\% 用于超出词汇表的回退（0.5\% OOV 率）。为了证明其有效性，我们使用 FastThaiG2P 对 Som-TTS 进行音素化，该数据集包含 20 小时的字素与音频对，然后基于 Kokoro-TTS 配方训练了一个 82M 参数的 StyleTTS 2 模型。生成的模型能够以 0.25 实时因子（4 倍实时）在 CPU 上进行 ONNX 推理，发出可理解的泰语语音，适合原型设计和开发。
cs.CL / 39 / 2608.12836
From Atomic Evidence to Logical Composition: Structured Compositional Reasoning over Compound Answer Options
从原子证据到逻辑组合：对复合答案选项的结构化组合推理
Abstract
Large language models often fail when answer options require combining atomic judgments under explicit logical operators, even when they judge the individual atoms correctly. We study compound options connected by AND, OR, and NEITHER/NOR, introducing a framework that decomposes each option into atomic answers and scores contrastive hypotheses about each one, so the model never sees a compound option. An operator-constrained integer linear program then composes the calibrated scores into a single prediction. We evaluate on LOGICAL-COMMONSENSEQA and introduce LOGICAL-SATA, a reading-comprehension benchmark derived from SATA-Bench. Our framework improves Macro-F1 from 48.3 to 77.0 on the human-validated LOGICAL-COMMONSENSEQA split and from 47.0 to 75.6 on LOGICAL-SATA, with the largest gains on NEITHER/NOR.
Chinese Translation
大型语言模型在需要在明确的逻辑运算符下组合原子判断的答案选项时常常失败，即使它们能够正确判断单个原子。我们研究了通过 AND、OR 和 NEITHER/NOR 连接的复合选项，提出了一个框架，将每个选项分解为原子答案，并对每个答案的对比假设进行评分，从而使模型从未见过复合选项。然后，一个受运算符约束的整数线性规划将校准后的评分组合成一个单一的预测。我们在 LOGICAL-COMMONSENSEQA 上进行评估，并引入了 LOGICAL-SATA，这是一个源自 SATA-Bench 的阅读理解基准。我们的框架在经过人工验证的 LOGICAL-COMMONSENSEQA 数据集上将 Macro-F1 从 48.3 提升至 77.0，在 LOGICAL-SATA 上从 47.0 提升至 75.6，尤其在 NEITHER/NOR 上获得了最大的提升。
cs.CL / 40 / 2608.12841
AQuA: Recursively Self-Improving Quantitative Trading Research Agents
AQuA：递归自我改进的量化交易研究代理
Abstract
We study recursive self-improvement at the level of quantitative-investment research: whether an autonomous system can use evidence from earlier experiments to improve the hypotheses and candidates proposed in later iterations. We present AQuA, which comprises two separate language-model-driven research systems: one for symbolic factor discovery and one for trainable model development. The two systems do not share agents, memories, candidate spaces, or research state. Instead, each independently closes its own research loop by retaining validated evidence and using it to guide subsequent proposals. In this bounded sense, both systems implement recursive self-improvement at the level of the research process. Each system also uses its own sealed sandbox, which fixes the data splits, feature and label definitions, and evaluator while allowing the model to act only through constrained factor expressions or configuration diffs. The factor system, a manager-mediated multi-agent pipeline, discovers and combines factors into a signal that reaches a combined information coefficient of about $0.190$ on a crypto universe. The model system, a config-driven loop over a hybrid time-series architecture, reaches a per-stock information coefficient of $+0.0843$ on US equities and converts it into a threshold long/short strategy with a held-out Sharpe of up to $+2.50$ at a two-leg cost. The strategy is positive in every year from 2021 to 2025.
Chinese Translation
我们研究了量化投资研究中的递归自我改进：一个自主系统是否能够利用早期实验的证据来改进后续迭代中提出的假设和候选项。我们提出了AQuA，它由两个独立的基于语言模型的研究系统组成：一个用于符号因子发现，另一个用于可训练模型开发。这两个系统不共享代理、记忆、候选空间或研究状态。相反，每个系统通过保留经过验证的证据并利用这些证据来指导后续提案，独立地完成自己的研究循环。在这个有限的意义上，这两个系统在研究过程中实现了递归自我改进。每个系统还使用自己的封闭沙箱，这固定了数据划分、特征和标签定义以及评估器，同时允许模型仅通过受限的因子表达或配置差异进行操作。因子系统是一个由管理者调解的多代理管道，发现并组合因子形成一个信号，在一个加密货币领域中达到了约$0.190$的综合信息系数。模型系统是一个基于配置的循环，采用混合时间序列架构，在美国股票中达到了每只股票信息系数$+0.0843$，并将其转化为一个阈值多空策略，在两腿成本下保持的夏普比率高达$+2.50$。该策略在2021年至2025年的每一年均表现为正。
cs.CL / 41 / 2608.12852
Falsehood and Impossibility Are Different Directions in an AI's Representation of Language
虚假与不可能在人工智能语言表征中的不同方向
Abstract
Language can describe states of affairs that are false and states of affairs that could not be the case at all. Whether an AI model internally distinguishes these failures remains unclear. I report an exploratory activation study of the multimodal open-weight model Gemma 3 4B IT using 85 prompts from 17 philosophical families and a topic-matched modality set of 15 topics, each expressed as a truth, contingent falsehood, improbable claim, semantic anomaly, and necessary falsehood. In its answers, the model conflates contingent falsehood with contradiction, labeling 12 of 15 false statements "contradiction." Its activations show a different pattern. A linear truth probe separates impossible from true statements (AUC 0.93) but not impossible from false statements (AUC 0.20). An impossibility probe evaluated on held-out topic families separates necessary from contingent falsehood at AUC 1.00, peaking at layer 15 with balanced accuracy 0.97 (Bonferroni-adjusted P=0.018). The truth and impossibility directions are close to orthogonal, whereas the impossibility direction partially overlaps a semantic anomaly direction while remaining distinguishable from it. Sparse autoencoder features at the same layer repeat this geometry. Features selective for impossibility also fire on anomalous sentences but rarely on contingent falsehoods. In this model's activation space, necessary falsehoods are not extreme cases of contingent falsehood but lie closer to the experimentally defined category of semantic anomaly. This representational proximity does not imply that impossible statements are intrinsically meaningless. These correlational observations from one small model offer an empirical footnote to an old philosophical distinction.
Chinese Translation
语言可以描述虚假的状态和根本不可能的状态。一个人工智能模型是否在内部区分这些失败仍然不清楚。我报告了一项对多模态开放权重模型Gemma 3 4B IT的探索性激活研究，使用了来自17个哲学家族的85个提示和一个与主题匹配的15个主题的模态集合，每个主题以真相、偶然虚假、不太可能的主张、语义异常和必要虚假来表达。在其回答中，该模型将偶然虚假与矛盾混淆，将15个虚假陈述中的12个标记为“矛盾”。其激活显示出不同的模式。线性真相探测器能够区分不可能的陈述与真实陈述（AUC 0.93），但无法区分不可能的陈述与虚假陈述（AUC 0.20）。在保留的主题家族上评估的不可能性探测器能够以AUC 1.00将必要虚假与偶然虚假区分开，且在第15层达到峰值，平衡准确率为0.97（Bonferroni调整后的P=0.018）。真相和不可能性方向接近正交，而不可能性方向部分重叠于语义异常方向，但仍然可区分。在同一层的稀疏自编码器特征重复了这种几何形状。选择性针对不可能性的特征在异常句子上也会激活，但在偶然虚假上则很少激活。在该模型的激活空间中，必要虚假并不是偶然虚假的极端案例，而是更接近实验定义的语义异常类别。这种表征的接近性并不意味着不可能的陈述在本质上是无意义的。这些来自一个小模型的相关观察为一个古老的哲学区分提供了经验性的注脚。
cs.CL / 42 / 2608.12875
The Embedder's Dilemma: LLMs Are Better, but at What Cost?
嵌入者的困境：大型语言模型更优秀，但代价是什么？
Abstract
Should you replace your text-embedding pipeline with a large language model? We answer this with a controlled, cost-aware comparison of ten LLMs across six families and 26 embedding models (118M to 14B parameters) on 37 tasks spanning classification, semantic textual similarity (STS), clustering, pair classification, and retrieval. In aggregate the two paradigms are effectively tied: the best LLM (Gemini 3.1 Pro, 77.6) and the best embedding model (77.2) differ by 0.4 points. Their strengths differ by task: LLMs lead on reasoning-heavy retrieval, embedding models lead on classification, and the two match on clustering, STS, and pair classification. Reaching that parity is expensive. An LLM costs up to 1,431x more than an embedding model of comparable quality (USD 154 vs. USD 0.11 per benchmark pass), and the open LLMs tested process tokens 2.5 to 736x more slowly on the same GPU. Reasoning tokens account for 28 to 81% of LLM inference cost; lower reasoning budgets preserve or improve retrieval quality for most models in our ablation. The Pareto frontier contains the leading embedding models and one LLM, Gemini 3.1 Pro. These results support a division of labour: use embedding models for similarity, classification, and clustering, and reserve LLMs for reasoning-intensive retrieval. Our code, datasets, and results are publicly available at https://github.com/embeddings-benchmark/embedders-dilemma.
Chinese Translation
您是否应该用大型语言模型替换您的文本嵌入管道？我们通过对十个大型语言模型（LLMs）在六个家族和26个嵌入模型（参数范围从1.18亿到140亿）在37个任务（涵盖分类、语义文本相似性（STS）、聚类、配对分类和检索）的受控、成本意识比较来回答这个问题。总体而言，这两种范式的效果几乎相当：最佳的LLM（Gemini 3.1 Pro，得分77.6）与最佳的嵌入模型（得分77.2）之间仅相差0.4分。它们在不同任务上的优势各异：LLMs在推理密集型检索中表现优越，而嵌入模型在分类任务中表现更好，两者在聚类、STS和配对分类任务上表现相当。达到这种平衡的代价很高。与相当质量的嵌入模型相比，LLM的成本高达1431倍（每次基准测试154美元对0.11美元），而测试的开放LLMs在同一GPU上处理令牌的速度比嵌入模型慢2.5到736倍。推理令牌占LLM推理成本的28%到81%；较低的推理预算可以在我们的大规模消融实验中保持或提高大多数模型的检索质量。帕累托前沿包含领先的嵌入模型和一个LLM，Gemini 3.1 Pro。这些结果支持了劳动分工：使用嵌入模型进行相似性、分类和聚类，而将LLMs保留用于推理密集型检索。我们的代码、数据集和结果已公开发布在 https://github.com/embeddings-benchmark/embedders-dilemma。
cs.CL / 43 / 2608.12888
When Your Agent Opens the Chat App: Agent-Controlled Search over Raw Chat Logs Rivals Structured Memory
当你的代理打开聊天应用时：代理控制的原始聊天记录搜索与结构化记忆的竞争
Abstract
Agent-memory systems increasingly buy retrieval quality with structure, transforming raw conversation histories into summaries, embeddings, trees, or knowledge graphs before any question is asked. We ask how much of that benefit comes from the structure itself, rather than from competent retrieval over the raw history. We present ReFind, an agent-controlled search interface that builds no semantic structure at all: it leaves the conversation archive unmodified, indexes it lexically at turn granularity, and combines a generic iterative keyword-search loop with four chat-native controls grounded in empirical refinding work: session-aware rank fusion, local context expansion, temporal narrowing, and skipping already-inspected sessions. A separate reasoning stage answers from the collected evidence. Across a broad suite of conversational-memory tasks (single- and multi-hop QA, event ordering, and fact consolidation), roughly 2,800 questions on precise-retrieval and fact-tracking capabilities evaluated under the incremental multi-turn setting of MemoryAgentBench, ReFind attains the highest mean accuracy (58.2) of any system compared, above the strongest graph- and tree-based memory systems (HippoRAG 2, 53.2), all under a GPT-4o-mini backbone matched to every reused baseline. Controlled comparisons to single-shot BM25, a matched generic-agentic BM25 control, component removals, and agentic dense/hybrid variants separately support the roles of agent control, chat-native controls, and lexical retrieval. On LongMemEval-S/M, the same interface reaches 93.2 +/- 3.3 and 89.3 +/- 6.0 with GPT-5-mini. The results indicate that for precise, evidence-grounded questions over chat archives, much of the benefit credited to elaborate memory structures is recoverable by giving an agent controllable search over the unmodified record, with no LLM-based index construction at all.
Chinese Translation
代理记忆系统越来越多地通过结构来提高检索质量，在提出任何问题之前，将原始对话历史转化为摘要、嵌入、树或知识图谱。我们探讨这种好处有多少来自于结构本身，而不是来自于对原始历史的有效检索。我们提出了ReFind，一个完全不构建语义结构的代理控制搜索接口：它不对对话档案进行修改，以轮次粒度进行词汇索引，并结合一个通用的迭代关键词搜索循环与四个基于经验再发现工作的聊天原生控制：会话感知排名融合、本地上下文扩展、时间缩小和跳过已检查的会话。一个单独的推理阶段根据收集的证据进行回答。在一系列广泛的对话记忆任务（单跳和多跳问答、事件排序和事实整合）中，约2800个关于精确检索和事实跟踪能力的问题在MemoryAgentBench的增量多轮设置下进行评估，ReFind在所有比较的系统中达到了最高的平均准确率（58.2），超过了最强的图形和树形记忆系统（HippoRAG 2，53.2），所有这些都在与每个重用基线匹配的GPT-4o-mini骨干网络下。与单次BM25的受控比较、匹配的通用代理BM25控制、组件移除以及代理密集/混合变体分别支持了代理控制、聊天原生控制和词汇检索的作用。在LongMemEval-S/M上，同一接口在GPT-5-mini下达到了93.2 +/- 3.3和89.3 +/- 6.0。结果表明，对于基于证据的精确问题，许多归因于复杂记忆结构的好处可以通过给予代理对未修改记录的可控搜索来恢复，而根本不需要基于LLM的索引构建。
cs.CL / 44 / 2608.12894
BavGround: A Benchmark for Regional Cultural Grounding and Dialect Competence in Bavarian
BavGround：巴伐利亚地区文化基础和方言能力的基准测试
Abstract
Cultural evaluation of large language models (LLMs) often focuses on high-resource standard languages, leaving regional culture and dialect communities underrepresented. We introduce BavGround, a benchmark for evaluating Bavarian regional cultural grounding and dialect competence across English, German and Bavarian. BavGround contains 206 multiple-choice source questions across eight cultural domains per language, yielding 618 multi-parallel instances, with items covering both broadly accessible cultural knowledge and source-grounded regional knowledge from journalism, historical sources, and specialist literature. We evaluate fifteen 7B-10B open-weight instruction-tuned models and one closed-model reference. Strong multilingual models perform best overall, but performance drops on Bavarian items and source-grounded questions, indicating persistent difficulty with dialectal and localized cultural knowledge. We further show that conclusions depend strongly on evaluation protocol: raw answer-letter scoring, shuffled-letter scoring, option-text likelihood, generated-answer parsing, and semantic matching can produce different absolute scores and rankings, especially for regionally adapted models. Finally, an exploratory analysis of GENBA-10B checkpoints suggests that continued pretraining improves answer-content likelihood unevenly across domains, while dialect competence remains comparatively weak. BavGround supports localized, protocol-aware evaluation of cultural representation in LLMs.
Chinese Translation
大型语言模型（LLMs）的文化评估通常侧重于高资源标准语言，导致地区文化和方言社区的代表性不足。我们介绍了BavGround，这是一个用于评估巴伐利亚地区文化基础和方言能力的基准测试，涵盖英语、德语和巴伐利亚语。BavGround包含每种语言八个文化领域的206个多项选择源问题，共计618个多平行实例，题目涵盖了广泛可获取的文化知识以及来自新闻、历史资料和专业文献的源基础地区知识。我们评估了十五个7B-10B开放权重的指令调优模型和一个闭合模型参考。强大的多语言模型整体表现最佳，但在巴伐利亚项目和源基础问题上的表现下降，表明在方言和地方文化知识方面仍然存在持续的困难。我们进一步表明，结论在很大程度上依赖于评估协议：原始答案字母评分、打乱字母评分、选项文本可能性、生成答案解析和语义匹配可能产生不同的绝对分数和排名，尤其是对于地区适应模型。最后，对GENBA-10B检查点的探索性分析表明，持续的预训练在各领域改善答案内容的可能性不均匀，而方言能力仍然相对较弱。BavGround支持对大型语言模型中文化表现的本地化、协议意识的评估。
cs.CL / 45 / 2608.12905
Prompts in the Wild: A Large Analyzed Collection of Transactional Prompts in Code
野外中的提示：一份关于代码中事务性提示的大规模分析集合
Abstract
The behavior of contemporary generative Large Language Models (LLMs) is directly shaped by prompts, unstructured texts that describe the desired output and model behavior. In this paper we argue that prompts are linguistic objects that merit investigation in their own right. To this end, we collect 57.5K unique samples of prompts from GitHub. Specifically, we focus on transactional prompts: reproducible natural language instructions that are integrated into software. To enable the empirical, quantitative study of prompts, we introduce a structured ontology, capturing the properties of prompts as well as their formal and semantic components. Based on this ontology, we transform prompts from unstructured raw texts into richly structured linguistic objects. Analysis of these structured data reveals significant diversity of usage patterns across languages, domains, tasks, and modalities, in a typical Zipf-like distribution where some clearly prevail and others, more diverse, appear in the long tail. To validate the reliability of the ontology-based annotation of the prompts, we perform a comprehensive error analysis across all fields, providing a detailed assessment of annotation quality. We release the dataset together with a browsing and exploration interface (https://github.com/OnlpLab/transactionalPromptsCollection ).
Chinese Translation
当代生成性大型语言模型（LLMs）的行为直接受到提示的影响，提示是描述期望输出和模型行为的非结构化文本。本文认为，提示是值得独立研究的语言对象。为此，我们从GitHub收集了57.5K个独特的提示样本。具体而言，我们专注于事务性提示：可重复的自然语言指令，这些指令被集成到软件中。为了支持对提示的实证定量研究，我们引入了一种结构化本体，捕捉提示的属性以及它们的形式和语义组成。基于该本体，我们将提示从非结构化原始文本转化为丰富的结构化语言对象。这些结构化数据的分析揭示了在语言、领域、任务和模态之间使用模式的显著多样性，呈现出典型的Zipf分布，其中一些模式明显占优，而其他更为多样的模式则出现在长尾部分。为了验证基于本体的提示注释的可靠性，我们对所有字段进行了全面的错误分析，提供了对注释质量的详细评估。我们发布了该数据集，并提供了一个浏览和探索界面（https://github.com/OnlpLab/transactionalPromptsCollection）。
cs.CL / 46 / 2608.12913
Decoupled Contrastive Decoding via Expert-Aligned Drafting
通过专家对齐草拟实现解耦对比解码
Abstract
Contrastive Decoding (CD) improves generation quality, but its amateur-model pass makes decoding expensive. Accelerating CD with speculative decoding raises a proposal-alignment question: should the contrastive signal shape the drafter, or should it remain only in verification? We study this question in the lightweight feature-level drafter regime. Two controlled diagnostics, matched Cross-alpha training and an Approximate Dual-Drafter decomposition, give the same diagnosis: contrastive-aware drafting does not consistently improve over expert-aligned drafting because the contrastive correction is usually weaker than drafter error, and reconstruction can amplify that error. We introduce Decoupled Contrastive Decoding (DCD), which drafts with an expert-aligned lightweight proposer and applies the amateur only in unchanged CD verification. Standard speculative verification preserves the vanilla-CD output distribution. Across the main 8B settings, EAGLE3-based DCD achieves average greedy speedups of 1.65 to 1.95x over vanilla CD and reduces MMLU proposal-path latency by about 5 to 12x relative to amateur-coupled proposal paths.
Chinese Translation
对比解码（Contrastive Decoding, CD）提高了生成质量，但其业余模型的传递使得解码成本高昂。通过投机解码加速 CD 提出了一个提案对齐的问题：对比信号应该塑造草拟者，还是仅仅在验证中保留？我们在轻量级特征级草拟者的框架中研究了这个问题。两个受控诊断，即匹配的 Cross-alpha 训练和近似双草拟者分解，给出了相同的诊断：对比感知的草拟并不总是优于专家对齐的草拟，因为对比修正通常弱于草拟者错误，而重构可能会放大该错误。我们引入了解耦对比解码（Decoupled Contrastive Decoding, DCD），它使用专家对齐的轻量级提议者进行草拟，并仅在未改变的 CD 验证中应用业余模型。标准的投机验证保持了原始 CD 输出分布。在主要的 8B 设置中，基于 EAGLE3 的 DCD 实现了相对于原始 CD 的平均贪婪加速为 1.65 到 1.95 倍，并将相对于业余耦合提案路径的 MMLU 提案路径延迟减少了约 5 到 12 倍。
cs.CL / 47 / 2608.12953
Unifying Depth and Width Pruning for LLMs via Binary Knapsack Optimization
通过二进制背包优化统一深度和宽度剪枝用于大型语言模型
Abstract
Structured pruning is a promising approach for compressing large language models (LLMs), yet existing methods rely heavily on greedy heuristics that produce myopic decisions, and often fail to precisely meet target compression budgets. We present SNIPER, a two-stage structured pruning framework that solves a knapsack optimization over coarse-granularity components to yield conditionally optimal parameter allocations with respect to fixed importance estimates, followed by a fine-grained pruning stage to meet strict budget constraints. We introduce the Compression Ratio Adherence Factor (CRAFT) to quantify budget fidelity, showing that while existing pruners deviate from target compression ratios by up to 33%, SNIPER achieves near-exact adherence with a CRAFT score of 0.98. Evaluations across four diverse architectures over a set of 18 tasks spanning five domains demonstrate SNIPER's consistent improvements in average performance retention and task-level stability over six state-of-the-art pruners. Across all pruning configurations, SNIPER achieves an excellent mean rank of 1.25, indicating its robust cross-architectural generalizability and excellent reliability.
Chinese Translation
结构化剪枝是一种压缩大型语言模型（LLMs）的有前景的方法，但现有方法严重依赖贪心启发式，导致短视决策，并且常常无法精确满足目标压缩预算。我们提出了SNIPER，这是一种两阶段结构化剪枝框架，通过对粗粒度组件进行背包优化，生成相对于固定重要性估计的条件最优参数分配，随后进行细粒度剪枝以满足严格的预算约束。我们引入了压缩比遵循因子（CRAFT）来量化预算的保真度，显示现有剪枝器在目标压缩比上偏离高达33%，而SNIPER以0.98的CRAFT得分实现了近乎精确的遵循。在跨越五个领域的18个任务集上对四种不同架构的评估表明，SNIPER在平均性能保留和任务级稳定性方面相较于六种最先进的剪枝器表现出一致的改善。在所有剪枝配置中，SNIPER的平均排名达到1.25，表明其在跨架构的泛化能力和可靠性方面表现出色。
cs.CL / 48 / 2608.12990
LycheeMemory V2: Efficient Long-Term Memory for LLM Agents via Semantic Segment-Level Consolidation
LycheeMemory V2：通过语义段级整合实现高效的长期记忆用于大语言模型代理
Abstract
Long-horizon LLM agents must preserve information from past interactions to support future tasks. Existing memory systems typically rely on eager consolidation, invoking LLMs after each interaction to extract, summarize, or update memories. This design makes memory construction increasingly costly as conversations grow. Coarse summarization can reduce construction cost but risks discarding fine-grained contextual evidence, whereas larger retrieval contexts or multi-hop LLM reasoning shift the overhead to query time. We present LycheeMemory V2, an efficient long-term memory framework that replaces turn-level consolidation with semantic segment-level consolidation. Instead of consolidating every interaction, LycheeMemory batches multiple exchanges into segments and encodes each finalized segment into context-independent typed memory records. Segment-level batching lowers LLM encoding frequency, while semantic boundary detection helps preserve coherent event-level and temporal evidence compared with fixed-window batching. The resulting records are organized with lightweight structured indexes for query-planned evidence retrieval. Experiments using GPT-4.1-Mini show that LycheeMemory achieves state-of-the-art performance, reaching 89.22% on LoCoMo and 92.20% on LongMemEval-S. Compared with A-Mem, it reduces construction tokens by 86.0% on LoCoMo and 75.9% on LongMemEval-S without increasing query-time token usage. More broadly, our results suggest that the accuracy--cost trade-off of long-term agent memory depends not only on what information is retained, but also on the granularity at which it is consolidated.
Chinese Translation
长时间跨度的大语言模型（LLM）代理必须保留过去交互的信息，以支持未来的任务。现有的记忆系统通常依赖于急切整合，在每次交互后调用LLM来提取、总结或更新记忆。这种设计使得随着对话的增长，记忆构建的成本越来越高。粗略的总结可以降低构建成本，但可能会丢弃细粒度的上下文证据，而更大的检索上下文或多跳LLM推理则将开销转移到查询时间。我们提出了LycheeMemory V2，一种高效的长期记忆框架，通过语义段级整合替代了轮次级整合。LycheeMemory不是整合每次交互，而是将多个交换批量处理为段，并将每个最终段编码为上下文独立的类型化记忆记录。段级批处理降低了LLM编码的频率，而语义边界检测有助于保留连贯的事件级和时间证据，相较于固定窗口批处理。生成的记录通过轻量级结构化索引组织，以便于查询计划的证据检索。使用GPT-4.1-Mini的实验表明，LycheeMemory在LoCoMo上达到了89.22%的最先进性能，在LongMemEval-S上达到了92.20%。与A-Mem相比，它在LoCoMo上减少了86.0%的构建标记，在LongMemEval-S上减少了75.9%，而没有增加查询时间的标记使用量。更广泛地说，我们的结果表明，长期代理记忆的准确性与成本的权衡不仅依赖于保留哪些信息，还依赖于整合的粒度。
cs.CL / 49 / 2608.13004
HybridRAG-BN: A Retrieval-Augmented Framework with Fine-Tuned Verification for Bangla KBQA
HybridRAG-BN：一种针对孟加拉语知识库问答的带有微调验证的检索增强框架
Abstract
Knowledge-base question answering (KBQA) systems rely on effective retrieval and reasoning mechanisms to generate accurate answers from external knowledge sources. However, developing reliable KBQA systems for low-resource languages such as Bangla remains challenging due to limited retrieval-focused research, scarce language resources, and difficulties in grounding generated responses in external knowledge. In this work, we propose HybridRAG-BN, a retrieval-augmented framework for Bangla KBQA that integrates hybrid retrieval using BM25 and BGE-M3, answer generation using the GGUF version of Gemma-4-31B-Instruct, and a LoRA-fine-tuned Gemma-4-31B-Instruct model for answer verification and refinement. To further improve robustness, the framework incorporates a post-processing stage that addresses unresolved cases through fallback answer replacement and DuckDuckGo-assisted retrieval. Experimental results demonstrate the effectiveness of the proposed framework, achieving token-level F1 scores of 0.71654 and 0.72912 on the public and private leaderboards, respectively, securing first place in the competition.
Chinese Translation
知识库问答（KBQA）系统依赖有效的检索和推理机制，从外部知识源生成准确的答案。然而，由于检索相关研究有限、语言资源稀缺以及在外部知识中对生成响应进行基础验证的困难，为低资源语言如孟加拉语开发可靠的KBQA系统仍然具有挑战性。在本研究中，我们提出了HybridRAG-BN，这是一种针对孟加拉语KBQA的检索增强框架，集成了使用BM25和BGE-M3的混合检索、使用GGUF版本的Gemma-4-31B-Instruct进行答案生成，以及用于答案验证和精炼的LoRA微调Gemma-4-31B-Instruct模型。为了进一步提高鲁棒性，该框架还包含一个后处理阶段，通过回退答案替换和DuckDuckGo辅助检索来解决未解决的案例。实验结果表明，所提框架的有效性，在公共和私有排行榜上分别获得了0.71654和0.72912的token级F1分数，赢得了比赛的第一名。
cs.CL / 50 / 2608.13006
EviReform: Evidence-Guided Query Reformulation for Multi-Hop Graph Retrieval
EviReform：基于证据的多跳图检索查询重构
Abstract
Multi-hop retrieval must recover passages that provide sufficient evidence together. An initial passage often resolves an entity or relation implicit in the question, making the missing evidence easier to describe only after retrieval begins. Graph retrieval improves access to related evidence through stored corpus structure, but its retrieval signal is commonly derived from the original question. Complementary evidence must then be reached through stored relations even when an observed passage provides a more direct semantic cue. We introduce EviReform, which separates revising the retrieval request from aggregating evidence in the graph. Retrieved source passages formulate residual queries for the unresolved information need. The original and residual retrieval signals are normalized separately, combined, and propagated between propositions that share entities. On 2WikiMultiHopQA, HotpotQA, and MuSiQue, EviReform exceeds the strongest baseline by up to 5.59 Recall@5 points and 4.50 F1 points. These results show that observed evidence can guide graph retrieval toward the part of a supporting chain left underspecified by the original question. Code is available at https://github.com/XrazyMee/EviReform.
Chinese Translation
多跳检索必须恢复提供足够证据的段落。初始段落通常解决了问题中隐含的实体或关系，使得在检索开始后更容易描述缺失的证据。图检索通过存储的语料库结构改善了对相关证据的访问，但其检索信号通常源自原始问题。因此，即使观察到的段落提供了更直接的语义线索，补充证据仍必须通过存储的关系来获取。我们提出了EviReform，它将检索请求的修订与图中的证据聚合分开。检索到的源段落为未解决的信息需求形成残余查询。原始和残余检索信号分别进行归一化，组合后在共享实体的命题之间传播。在2WikiMultiHopQA、HotpotQA和MuSiQue数据集上，EviReform的表现超过最强基线，Recall@5提高了最多5.59分，F1提高了4.50分。这些结果表明，观察到的证据可以引导图检索朝向原始问题未明确的支持链部分。代码可在https://github.com/XrazyMee/EviReform获取。
cs.CL / 51 / 2608.13010
RAGSieve: Self-Referenced Local Contrast for Knowledge-Poison Detection in Retrieval-Augmented Generation
RAGSieve：用于检索增强生成中的知识污染检测的自参考局部对比
Abstract
Retrieval-augmented generation treats an external corpus as inference evidence, allowing injected documents to promote attacker-chosen claims. Existing detectors depend on trusted references, specific attack artifacts, or global thresholds sensitive to corpus topology. We present RAGSieve, a self-referenced detection framework that constructs its reference from the inspected system. RAGSieve-Query (RSQ) performs query-local contrast, scoring top-five candidates against ranks 6-20 of the same retrieval to detect answer-anchor concentration and carrier transitions. RAGSieve-Graph (RSG) performs corpus-local contrast, comparing each document's semantically similar but lexically distinct neighbors with its local baseline to detect coordinated density before queries arrive. Across three QA datasets and six poisoning constructions, RSQ achieves 95.2% AUROC and detects 82.2% of poison at 5% clean-document removal, versus 81.1%/52.5% for GMTP. RSG achieves 93.3%/79.8%, versus 79.4%/37.6% for CleanBase. Joint deployment reduces attack success from 67.4% to 14.0% while retaining 41.3% F1 on unpoisoned retrieval, demonstrating practical protection at both corpus ingestion and query time without poison labels or trusted corpora. Source code is available at https://github.com/XrazyMee/RAGSieve.
Chinese Translation
检索增强生成将外部语料库视为推理证据，允许注入的文档促进攻击者选择的主张。现有检测器依赖于可信的参考、特定的攻击伪影或对语料库拓扑敏感的全局阈值。我们提出了RAGSieve，一种自参考检测框架，它从被检查的系统构建其参考。RAGSieve-Query (RSQ) 执行查询局部对比，对前五个候选项进行评分，与同一检索的第6至20名进行比较，以检测答案锚点集中度和载体转变。RAGSieve-Graph (RSG) 执行语料库局部对比，将每个文档的语义相似但词汇上不同的邻居与其局部基线进行比较，以在查询到达之前检测协调密度。在三个问答数据集和六种污染构造中，RSQ在5%清洁文档移除的情况下实现了95.2%的AUROC，并检测到82.2%的污染，而GMTP的检测率为81.1%/52.5%。RSG的检测率为93.3%/79.8%，而CleanBase的检测率为79.4%/37.6%。联合部署将攻击成功率从67.4%降低到14.0%，同时在未污染的检索中保持41.3%的F1分数，展示了在语料库摄取和查询时的实际保护，无需污染标签或可信语料库。源代码可在https://github.com/XrazyMee/RAGSieve获取。
cs.CL / 52 / 2608.13101
CASA: Content-Acoustic Speaking Assessment with Speech Encoder and Large Language Model
CASA：结合语音编码器和大型语言模型的内容-声学口语评估
Abstract
Research on automatic speaking assessment (ASA) has increasingly adopted multimodal speech large language models to assess learners' speaking performance. However, existing studies provide limited analysis of how acoustic and content information contribute to predictions and how stable the resulting performance is. We propose CASA, a simpler architecture combining Whisper-medium and Qwen3.5-2B that achieves state-of-the-art performance while providing a more interpretable separation between speech delivery and content. On the Speak & Improve Corpus 2025, CASA achieves a root mean square error (RMSE) of 0.358, improving on the previous best RMSE while using approximately half the estimated inference parameters. The general-purpose architecture is designed for adaptation to other ASA corpora without structural changes and relies on three handcrafted fluency features. Through ablations and repeated runs, we analyze the individual and complementary contributions of acoustic and content information, examine performance variability, and demonstrate the potential of large language model reasoning for training-free content validation.
Chinese Translation
自动口语评估（ASA）的研究越来越多地采用多模态语音大型语言模型来评估学习者的口语表现。然而，现有研究对声学信息和内容信息如何影响预测以及所得到的性能稳定性提供的分析有限。我们提出了CASA，一种更简单的架构，结合了Whisper-medium和Qwen3.5-2B，达到了最先进的性能，同时提供了更可解释的语音表达与内容之间的分离。在Speak & Improve Corpus 2025数据集中，CASA实现了0.358的均方根误差（RMSE），在使用大约一半估计推理参数的情况下，改进了之前的最佳RMSE。该通用架构旨在适应其他ASA语料库而无需结构性更改，并依赖于三个手工制作的流利度特征。通过消融实验和重复运行，我们分析了声学信息和内容信息的个体及互补贡献，考察了性能变异性，并展示了大型语言模型推理在无训练内容验证中的潜力。
cs.CL / 53 / 2608.13136
LigBench: A Unified and Human-Aligned Benchmark for LLM-based Research Idea Generation
LigBench：一个统一且与人类对齐的基于大型语言模型的研究创意生成基准
Abstract
With the rapid advancement of large language models (LLMs), research idea generation has attracted increasing attention. Existing approaches enable LLMs to retrieve relevant literature and propose novel ideas for research areas. However, current evaluation practices for idea generation remain fragmented and lack objective standards, often relying on direct LLM scoring, which limits their ability to provide unified and reliable assessments across a coherent distribution of generated ideas. To address this challenge, we propose LigBench, an automated evaluation benchmark that enables fine-grained and reliable evaluation of AI research ideas, consistently applicable across different generation distributions. In addition, we introduce PAIR-IQ, a dataset tailored for training pairwise idea judgment models and serving as an auxiliary reference to support more objective comparative evaluation. Extensive experiments demonstrate that LigBench achieves stable and interpretable evaluations, significantly improving alignment with expert judgments. Furthermore, models trained on PAIR-IQ exhibit enhanced ranking accuracy and robustness, establishing a principled standard for scalable and objective research idea assessment.
Chinese Translation
随着大型语言模型（LLMs）的快速发展，研究创意生成引起了越来越多的关注。现有的方法使LLMs能够检索相关文献并提出新颖的研究创意。然而，目前的创意生成评估实践仍然分散，缺乏客观标准，通常依赖于直接的LLM评分，这限制了它们在生成创意的连贯分布中提供统一和可靠评估的能力。为了解决这一挑战，我们提出了LigBench，一个自动化评估基准，能够对AI研究创意进行细致且可靠的评估，并在不同的生成分布中一致适用。此外，我们引入了PAIR-IQ，一个为训练成对创意判断模型而量身定制的数据集，作为辅助参考，以支持更客观的比较评估。大量实验表明，LigBench实现了稳定且可解释的评估，显著提高了与专家判断的一致性。此外，在PAIR-IQ上训练的模型展现出更高的排名准确性和鲁棒性，为可扩展和客观的研究创意评估建立了原则性标准。
cs.CL / 54 / 2608.13160
Better Decomposition, Free Aggregation: A Synthesizer-Folding Framework for Multilingual Multi-Hop Question Answering
更好的分解，自由的聚合：一种用于多语言多跳问答的合成折叠框架
Abstract
Multilingual retrieval-augmented generation (mRAG) equips large language models with access to globally distributed external knowledge for complex multilingual question answering. Recent approaches either translate retrieved documents into English or the query language to bridge the cross-lingual semantic gap, or decompose a complex query into sub-questions and aggregate the intermediate reasoning process. However, both lines of work suffer from two limitations. First, one-size-fits-all translation alignment, blanket translation discards culturally and linguistically native information unique to the target language, introduces translation noise, and inflates system cost. Second, greedy decomposition and aggregation, uncontrolled decomposition produces redundant sub-questions that compound errors during step-wise reasoning, and the final aggregation over reasoning paths further amplifies these errors. We address both with our method Syfer, a synthesizer-folding framework for multilingual multi-hop question answering that defers translation rather than applying it by default. Syfer first invokes a format-constrained decomposer to produce a sub-question graph in the original language, followed by a decomposition-quality check; when the check passes, sub-questions are answered sequentially under a retrieve-then-answer policy in the target language, and the English translation pathway with bilingual sub-question graph alignment is activated only when the check fails. Experiments across multiple languages show that Syfer attains competitive accuracy while striking a favourable balance between performance and computational cost.
Chinese Translation
多语言检索增强生成（mRAG）为大型语言模型提供了访问全球分布的外部知识的能力，以应对复杂的多语言问答。近期的方法要么将检索到的文档翻译成英语或查询语言，以弥补跨语言的语义差距，要么将复杂查询分解为子问题并聚合中间推理过程。然而，这两种方法都存在两个局限性。首先，一刀切的翻译对齐，普遍的翻译忽略了目标语言特有的文化和语言信息，带来了翻译噪声，并增加了系统成本。其次，贪婪的分解和聚合，失控的分解会产生冗余的子问题，在逐步推理过程中累积错误，而最终对推理路径的聚合进一步放大了这些错误。我们通过我们的方法Syfer来解决这两个问题，Syfer是一个用于多语言多跳问答的合成折叠框架，它推迟翻译的应用，而不是默认执行。Syfer首先调用一个格式约束的分解器，以生成原始语言的子问题图，随后进行分解质量检查；当检查通过时，子问题在目标语言下按照检索-再回答的策略顺序回答，只有在检查失败时，才激活带有双语子问题图对齐的英语翻译路径。跨多种语言的实验表明，Syfer在性能和计算成本之间达到了竞争性的准确性和良好的平衡。
cs.CL / 55 / 2608.13168
Which LLM Is Your Ideal Companion? Evaluating Emotional Companion Capabilities of LLMs Based on Adult Attachment Theory
哪种大型语言模型是您理想的伴侣？基于成人依恋理论评估大型语言模型的情感伴侣能力
Abstract
As large language models (LLMs) are increasingly applied for emotional companionship, evaluating their behavior and capabilities in intimate relationships has become a pressing issue. However, existing assessments primarily characterize general personality traits, providing limited insight into model behavior within intimate and emotionally sensitive contexts. Therefore, we introduce adult attachment theory into LLM evaluation and use the Experiences in Close Relationships-Revised (ECR-R) scale to characterize attachment anxiety and avoidance. To evaluate emotional companionship capabilities of LLMs in realistic interaction scenarios, we present an emotional companionship benchmark, ECBench, spanning four scenarios including emotional support, collaborative tasks, conflict resolution, and social guidance, across friendship and romantic relationships. ECBench is utilized to assess model behavior using 11 dialogue-quality metrics and three evaluation methods. We evaluate the attachment tendencies of 32 LLMs and select representative models to investigate how these tendencies manifest in contextualized multi-turn interactions and whether they can be shaped through prompting. Our study provides a theoretical lens from psychology, along with practical tools to understand and select LLMs for emotional companionship.
Chinese Translation
随着大型语言模型（LLMs）在情感陪伴中的应用日益增多，评估它们在亲密关系中的行为和能力已成为一个紧迫的问题。然而，现有的评估主要描述一般的人格特征，无法深入了解模型在亲密和情感敏感的情境中的行为。因此，我们将成人依恋理论引入LLM评估，并使用《亲密关系经验修订量表》（Experiences in Close Relationships-Revised, ECR-R）来表征依恋焦虑和回避。为了评估LLMs在现实互动场景中的情感陪伴能力，我们提出了一个情感陪伴基准，ECBench，涵盖了包括情感支持、协作任务、冲突解决和社会指导在内的四种场景，涉及友谊和浪漫关系。ECBench用于通过11个对话质量指标和三种评估方法来评估模型行为。我们评估了32个LLMs的依恋倾向，并选择代表性模型来研究这些倾向如何在上下文化的多轮互动中表现，以及它们是否可以通过提示进行塑造。我们的研究提供了心理学的理论视角，以及理解和选择LLMs作为情感陪伴的实用工具。
cs.CL / 56 / 2608.13200
GEM: A Generative Embedding Model Bridging Reasoning and Retrieval
GEM：桥接推理与检索的生成嵌入模型
Abstract
Modern LLMs excel at reasoning and instruction following, enabling users to express complex and diverse information needs. However, conventional retrievers largely rely on surface-level matching between queries and documents, resulting in a growing gap between how users express their needs and how retrievers interpret them. In this paper, we present GEM, a generative embedding model that augments retrieval through its own knowledge by explicitly reasoning about user intent and relevance criteria. GEM unifies generation and embedding within a single model: it first reasons over the query, then appends an embedding token to encode the enriched context for retrieval. \zhili{Evaluated on reasoning-intensive and instruction-following retrieval tasks, GEM demonstrates the effectiveness of its reasoning-augmented retrieval, outperforming its non-reasoning variant and matching baselines using substantially larger models.} Furthermore, GEM's generative nature allows test-time compute scaling via prompting to further enhance retrieval performance. Our code is available at: https://anonymous.4open.science/r/GEM.
Chinese Translation
现代大型语言模型（LLMs）在推理和指令跟随方面表现出色，使用户能够表达复杂多样的信息需求。然而，传统的检索器主要依赖于查询与文档之间的表面匹配，导致用户表达需求与检索器解读之间的差距不断扩大。本文提出了GEM，一种生成嵌入模型，通过明确推理用户意图和相关性标准来增强检索。GEM在单一模型中统一了生成和嵌入：它首先对查询进行推理，然后附加一个嵌入标记以编码丰富的检索上下文。在推理密集和指令跟随的检索任务中进行评估，GEM展示了其推理增强检索的有效性，超越了其非推理变体，并与使用显著更大模型的基线匹配。此外，GEM的生成特性允许通过提示在测试时进行计算扩展，从而进一步提升检索性能。我们的代码可在以下链接获取：https://anonymous.4open.science/r/GEM。
cs.CL / 57 / 2608.13244
Localize, Then Reason: Visual Latent Structural Reasoning for Molecular Properties and Edits
先定位，再推理：用于分子性质和编辑的视觉潜在结构推理
Abstract
Local chemical perception and property reasoning are both essential for understanding how molecular structure determines properties. Current LLM-based chemical reasoning methods either receive SMILES/molecular images together with descriptions of local motifs, or reason directly from molecular images. Neither approach enables the model to focus on chemically meaningful regions before reasoning. To address this gap, we propose Visual Latent Structural Reasoning (VLSR), an end-to-end framework that jointly learns localization and reasoning from molecular images. Central to our approach is a localize-then-reason strategy. VLSR first learns to locate chemically meaningful regions in a molecular image. It then reasons about their property effects in a compact latent workspace before producing the final answer. Under the same inference setup, this design achieves 9.6X higher throughput than a comparable textual-reasoning baseline.
Chinese Translation
局部化学感知和性质推理对于理解分子结构如何决定性质至关重要。目前基于大型语言模型（LLM）的化学推理方法要么接收SMILES/分子图像以及局部特征的描述，要么直接从分子图像进行推理。这两种方法都无法使模型在推理之前专注于具有化学意义的区域。为了解决这一问题，我们提出了视觉潜在结构推理（Visual Latent Structural Reasoning, VLSR），这是一个端到端的框架，能够从分子图像中联合学习定位和推理。我们方法的核心是先定位后推理的策略。VLSR首先学习在分子图像中定位具有化学意义的区域。然后，它在一个紧凑的潜在工作空间中推理这些区域的性质影响，最后得出最终答案。在相同的推理设置下，该设计的吞吐量比可比的文本推理基线高出9.6倍。
cs.CL / 58 / 2608.13258
Self-Referential Induction Increases Response Instability Relative to Unresolvable and Verifiable Questions in Large Language Models
自我指涉诱导相较于不可解和可验证问题在大型语言模型中增加响应不稳定性
Abstract
Self-referential prompting has been shown to reliably induce large language models to produce first-person reports resembling subjective experience, but no prior work measures how consistent these reports are across repeated, independent trials, or how that consistency compares to the model's behavior on other kinds of open-ended questions. We measure response instability, defined as one minus the mean pairwise cosine similarity of sentence embeddings computed over a compressed core claim extracted from each response, for three groups of questions: self-referential prompts eliciting a subjective-experience report, unresolvable philosophical questions unrelated to self-reference, and questions with a verifiable correct answer. Using 30 independent responses per question (360 responses total, Gemini API, temperature 0.7) across four questions per group, we find that self-referential questions show the highest instability (0.343 +/- 0.047), unresolvable philosophy questions show intermediate and tightly clustered instability (0.192 +/- 0.008), and verifiable questions show the lowest instability (0.105 +/- 0.058). This provides a quantitative baseline for the induced subjective-experience report, showing that it occupies a distinct, less stable position in the model's output distribution than ordinary open-ended philosophical uncertainty.
Chinese Translation
自我指涉提示已被证明能够可靠地诱导大型语言模型生成类似主观体验的一人称报告，但之前的研究没有测量这些报告在重复独立试验中的一致性，也没有比较这种一致性与模型在其他类型开放性问题上的表现。我们测量响应不稳定性，定义为每个响应中提取的压缩核心主张的句子嵌入的平均成对余弦相似度的补数，针对三组问题进行分析：引发主观体验报告的自我指涉提示、与自我指涉无关的不可解哲学问题，以及具有可验证正确答案的问题。通过对每个问题使用30个独立响应（共360个响应，使用Gemini API，温度为0.7），我们发现自我指涉问题表现出最高的不稳定性（0.343 +/- 0.047），不可解的哲学问题表现出中等且紧密聚集的不稳定性（0.192 +/- 0.008），而可验证问题表现出最低的不稳定性（0.105 +/- 0.058）。这为诱导的主观体验报告提供了一个定量基线，表明它在模型输出分布中占据一个独特且不太稳定的位置，与普通的开放性哲学不确定性相比。
cs.CL / 59 / 2608.13267
How Do VLMs Behave When Blind or Misled? Behavioral Evaluation of VLMs on Scientific Figures
盲目或被误导时VLM的表现如何？对科学图形的VLM行为评估
Abstract
Existing vision-language model (VLM) benchmarks emphasize perception and reasoning accuracy (how well VLMs describe and reason about what they see in an image), with limited attention to behavioral reliability under uncertainty (how they behave when visual evidence is missing or misleading). We introduce SciFigBench, a diagnostic VLM benchmark for scientific figure understanding that jointly evaluates perception, reasoning, and behavioral reliability under uncertainty. It contains 250 figures with high-quality human annotations across three evaluation aspects, totaling 600+ hours of annotation effort. We further extend these figures via image transformations, reasoning questions, resistance probes, caption-bias probes, and confirmed selective-blur targets, producing over 34,000 evaluation setups for stress testing. We further propose the Admittance-Resistance-Inductance (A-R-I) framework to evaluate whether models acknowledge insufficient evidence, resist misleading context, and infer cautiously from partial information. Our results reveal substantial behavioral differences among models. GPT-5.2 achieves the highest description quality (MQM 91.6) with strong reasoning accuracy (78.4%), yet hallucinates unreadable content in 96% of cases, whereas Gemini 3.1 Pro, a comparably capable model (MQM 90.2, reasoning 81.0%), admits uncertainty in 71% of such cases and achieves the strongest resistance score (0.91). These findings show that high perception and reasoning accuracy alone do not guarantee behavioral reliability, a dimension critical for deployment in scientific workflows.
Chinese Translation
现有的视觉语言模型（VLM）基准强调感知和推理的准确性（VLM在图像中描述和推理其所见的能力），而对不确定性下的行为可靠性（在视觉证据缺失或误导时的表现）关注有限。我们引入了SciFigBench，这是一个用于科学图形理解的诊断性VLM基准，联合评估感知、推理和不确定性下的行为可靠性。该基准包含250个图形，涵盖三个评估方面，拥有高质量的人类注释，总计600多个小时的注释工作。我们进一步通过图像变换、推理问题、抗干扰探测、标题偏见探测和确认的选择性模糊目标扩展这些图形，产生了超过34,000个评估设置用于压力测试。我们还提出了接纳-抵抗-感应（Admittance-Resistance-Inductance, A-R-I）框架，以评估模型是否承认证据不足、抵抗误导性上下文，并从部分信息中谨慎推断。我们的结果揭示了模型之间显著的行为差异。GPT-5.2在描述质量上达到最高（MQM 91.6），推理准确性强（78.4%），但在96%的情况下产生无法理解的内容，而同样能力的模型Gemini 3.1 Pro（MQM 90.2，推理81.0%）在71%的此类情况下承认不确定性，并获得了最强的抵抗评分（0.91）。这些发现表明，仅有高感知和推理准确性并不能保证行为的可靠性，而这一维度对科学工作流程的部署至关重要。
cs.CL / 60 / 2608.13277
Mixture of Training: Recombining Small-Scale Scaffolded Pretraining Runs into a Larger Language Model
训练混合：将小规模的支架预训练运行重新组合成更大的语言模型
Abstract
We ask whether language-model pre-training can be decomposed into smaller, independently trainable jobs that can later be recomposed into a coherent larger model. We introduce Mixture of Training (MoT), a scaffolded modular pre-training procedure that partitions a target Transformer into contiguous layer blocks, trains each block inside a frozen pretrained aligner scaffold, and then recomposes the trained blocks with an optional short end-to-end adaptation pass. On a 1.3B-parameter Gemma-style model trained on C4, MoT provides a small-scale proof of mechanism: independently trained depth slices can be recomposed into a usable language model, and a quality-parity schedule reaches the same reported perplexity as the monolithic baseline. This parity setting processes more aggregate tokens and has a shorter idealized layer-equivalent critical path after aligner preparation; its effective compute advantage depends on reusing the aligner across runs. We therefore present MoT not as a general replacement for monolithic pre-training, but as a small-scale framework for studying whether scaffolded sub-runs can act as reusable training units.
Chinese Translation
我们探讨语言模型的预训练是否可以分解为更小的、可独立训练的任务，这些任务可以在后期重新组合成一个连贯的更大模型。我们提出了训练混合（Mixture of Training, MoT），这是一种支架模块化预训练程序，它将目标Transformer划分为连续的层块，在一个冻结的预训练对齐支架内训练每个块，然后通过可选的短期端到端适应过程重新组合训练好的块。在一个在C4上训练的1.3B参数Gemma风格模型中，MoT提供了一个小规模的机制证明：独立训练的深度切片可以重新组合成一个可用的语言模型，并且质量平衡调度达到了与单一基线相同的困惑度报告。这种平衡设置处理了更多的总令牌，并且在对齐器准备后具有更短的理想化层等效关键路径；其有效计算优势取决于在不同运行中重用对齐器。因此，我们将MoT呈现为一个小规模框架，用于研究支架子运行是否可以作为可重用的训练单元，而不是作为单一预训练的一般替代方案。
cs.CL / 61 / 2608.13304
Refusing Intent, Not Form: Wrapper-Based Intent-Group Supervision for LLM Safety
拒绝意图，而非形式：基于包装的意图组监督以提高大型语言模型的安全性
Abstract
Safety tuning can improve harmful refusal, but models may learn surface-form shortcuts: wrapped harmful prompts bypass safety, while similarly wrapped benign prompts are over-refused. We propose Wrapper-Based Intent-Form Augmentation (WIFA), an automatic intent-group augmentation method that pairs wrapped harmful examples with structurally matched wrapped benign counterexamples, requiring no external teacher or manual per-wrapper intent labels. We use WIFA as a common data layer for two complementary fine-tuning routes: WIFA-Boost, a two-stage high-safety recipe, and Anchored Group-Consistent Refusal Training (A-GCRT), which regularizes refusal/compliance decision scores across same-intent wrappers and anchors harmful and benign groups on opposite sides of a margin. In the Qwen setting, WIFA-Boost reaches the strongest transformed-harmful refusal, while A-GCRT reduces OR-Bench over-refusal from 25.7\% for the base model to 17.4\%; reproduced baselines do not match these operating points. Llama results and ablations over data structure, two-stage order, and A-GCRT components support this intent-group interpretation without claiming universal below-base over-refusal.
Chinese Translation
安全调优可以改善有害拒绝，但模型可能会学习表面形式的捷径：包装的有害提示绕过安全机制，而类似包装的良性提示则被过度拒绝。我们提出了基于包装的意图形式增强（Wrapper-Based Intent-Form Augmentation, WIFA），这是一种自动化的意图组增强方法，它将包装的有害示例与结构上匹配的包装良性反例配对，不需要外部教师或手动每个包装的意图标签。我们将WIFA用作两条互补的微调路径的共同数据层：WIFA-Boost，一种两阶段的高安全性方案，以及锚定组一致拒绝训练（Anchored Group-Consistent Refusal Training, A-GCRT），该方法在相同意图的包装之间规范化拒绝/遵从决策分数，并将有害组和良性组锚定在边界的两侧。在Qwen设置中，WIFA-Boost达到了最强的转化有害拒绝，而A-GCRT将OR-Bench的过度拒绝从基础模型的25.7%降低到17.4%；重现的基线未能达到这些操作点。Llama的结果和对数据结构、两阶段顺序及A-GCRT组件的消融实验支持了这种意图组解释，而不声称存在普遍的低于基线的过度拒绝现象。
cs.CL / 62 / 2608.13326
Beyond Local Accuracy: A Protocol-Level Identifiability Audit for Controlled LLM Reasoning Evaluation
超越局部准确性：受控LLM推理评估的协议级可识别性审计
Abstract
LLM benchmark scores can be precise even when the observation protocol does not identify the behavioral property they are intended to measure. In a controlled, solver-grounded setting, we formalize a protocol-level identifiability audit over a finite behavioral policy class: given policies H, observation support O, and estimand $\tau$, we test whether O separates every pair with different $\tau$. The audit requires zero model calls and resolves our diagnostic case: base-only observation collapses seven frozen deterministic policies into one equivalence class; full support yields seven classes and no cross-estimand collisions; every leave-one-out support retains a constructive collision witness. Empirically, both constrained-generation variants have pair-validity 1.0, yet base accuracy and selective-response fidelity diverge - 0.620 versus 0.324 across six balanced oracle-transition directions (cluster-bootstrap 95% CI [0.600, 0.642] vs. [0.304, 0.345]) - and the gap recurs on a second deterministic source (0.646 vs. 0.331). The audit also synthesizes a minimum identifying support $O^*$ for the frozen policy class: two cells instead of the full 36-cell tensor. This case shows how evaluation-design validity can be checked structurally before model inference and why base correctness does not determine intervention-response fidelity.
Chinese Translation
LLM基准分数即使在观察协议未能识别其意图测量的行为属性时也可能是精确的。在一个受控的、基于求解器的环境中，我们对有限行为策略类进行了协议级可识别性审计：给定策略H、观察支持O和估计量$ au$，我们测试O是否能区分每对具有不同$ au$的策略。该审计不需要模型调用，并解决了我们的诊断案例：仅基于基础的观察将七个冻结的确定性策略合并为一个等价类；完整支持产生七个类且没有跨估计量的冲突；每个留一法支持保留了一个构造性的冲突见证。从实证上看，两种约束生成变体的配对有效性均为1.0，但基础准确性和选择性响应保真度却出现了分歧——在六个平衡的oracle过渡方向上分别为0.620与0.324（集群自助法95%置信区间[0.600, 0.642]与[0.304, 0.345]）——而这一差距在第二个确定性源上再次出现（0.646与0.331）。该审计还为冻结策略类综合了一个最小识别支持$O^*$：两个单元而不是完整的36单元张量。此案例展示了如何在模型推断之前结构性地检查评估设计的有效性，以及为什么基础正确性并不决定干预响应的保真度。
cs.CL / 63 / 2608.13328
It's How You Ask: Gender-Associated Linguistic Bias in LLMs
提问方式的重要性：大型语言模型中的性别相关语言偏见
Abstract
Professional communication is increasingly mediated by LLMs - but do these models serve all users equally? We show that when prompts contain linguistic features more commonly used by women (hedges, tag questions, collective reference), they systematically elicit shorter, less sophisticated, and less formal responses across three document types and four models. These effects persist after controlling for prompt complexity and feature carry-over. Explicit gender cues like sign-off names are encoded in the same representational space as linguistic dialect - suggesting shared underlying mechanisms - yet linguistic register is far more influential, producing large, consistent effects where names produce none. Our results further reveal that post-hoc mitigation is challenging: because these patterns are culturally embedded and outside conscious control, users cannot easily avoid them through strategic self-presentation, and mechanistic analysis reveals that linguistic features are encoded in early transformer layers and entangled with other features. Our work calls for upstream consideration of the influences of linguistic variation to mitigate disparate impacts of LLM-mediated workplace communication.
Chinese Translation
专业沟通越来越多地通过大型语言模型（LLMs）进行，但这些模型是否平等地服务于所有用户？我们展示了当提示中包含女性更常用的语言特征（如模糊语、标签问题、集体指代）时，它们在三种文档类型和四个模型中系统性地引发了更短、更不复杂和更不正式的回应。这些效应在控制提示复杂性和特征延续后仍然存在。显式的性别提示，如签名名称，被编码在与语言方言相同的表征空间中——这表明存在共享的潜在机制——然而，语言风格的影响要大得多，产生了显著且一致的效果，而名称则没有产生任何效果。我们的结果进一步揭示，事后缓解是具有挑战性的：由于这些模式是文化嵌入的，并且超出了意识控制，用户无法通过战略性自我呈现轻易避免它们，机械分析显示语言特征被编码在早期的变换器层中，并与其他特征交织在一起。我们的研究呼吁在上游考虑语言变异的影响，以减轻大型语言模型介导的工作场所沟通的差异性影响。
cs.CL / 64 / 2608.13334
RippleMem: From Isolated Retrieval to Associative Recollection for Long-Term Agent Memory
RippleMem：从孤立检索到联想回忆的长期智能体记忆
Abstract
LLM-based agents increasingly rely on external memory to support long-horizon reasoning and interaction. However, the main bottleneck is not simply storing past experience, but recovering the right set of evidence when relevant information is distributed across many interactions. Existing approaches struggle with this access problem. Full-context methods require noisy long-context search, flat retrieval often returns isolated and incomplete records, and graph-based memory systems can be expensive to construct while compressing rich event context. We introduce RippleMem, a long-term memory system that replaces one-shot retrieval with adaptive associative recollection. Inspired by cue-dependent episodic retrieval and associative completion, RippleMem stores interaction history as cue-rich episodic memory units and organizes them in an event-centric memory graph. Given a query, it first recalls relevant memory anchors through hybrid cues, then expands from these anchors along semantic and structural associations to recover missing supporting evidence. In this way, initially recalled memories serve not only as answer context, but also as cues for completing the evidence needed to answer. Experiments on LoCoMo and LongMemEval-S show that RippleMem achieves the best overall performance across evaluated settings, improving LLM-as-a-Judge accuracy by 3.95% on LoCoMo and up to 11.87% on LongMemEval-S, while reducing graph construction cost by about 30x.
Chinese Translation
基于大型语言模型（LLM）的智能体越来越依赖外部记忆来支持长时间跨度的推理和交互。然而，主要瓶颈不仅在于存储过去的经验，而在于当相关信息分散在多个交互中时，如何恢复正确的证据集。现有方法在解决这一访问问题时面临困难。全上下文方法需要嘈杂的长上下文搜索，平面检索往往返回孤立且不完整的记录，而基于图的记忆系统在压缩丰富事件上下文时构建成本可能较高。我们提出了RippleMem，一种长期记忆系统，它用自适应联想回忆替代了一次性检索。受到线索依赖的情节检索和联想补全的启发，RippleMem将交互历史存储为富含线索的情节记忆单元，并将其组织在以事件为中心的记忆图中。给定查询时，它首先通过混合线索回忆相关的记忆锚点，然后沿着语义和结构关联从这些锚点扩展，以恢复缺失的支持证据。通过这种方式，最初回忆的记忆不仅作为答案上下文，还作为完成回答所需证据的线索。在LoCoMo和LongMemEval-S上的实验表明，RippleMem在评估的设置中实现了最佳的整体性能，在LoCoMo上提高了LLM作为评判者的准确性3.95%，在LongMemEval-S上提高了多达11.87%，同时将图构建成本降低了约30倍。
cs.CL / 65 / 2608.13387
CROP: Task Relevance via Counterfactuals for Selective On-Policy Distillation
CROP：通过反事实实现选择性在线蒸馏的任务相关性
Abstract
On-policy distillation (OPD) supervises a student language model on trajectories sampled from its current policy, but assigns equal credit to response tokens with unequal supervision value. Selective OPD addresses this limitation by allocating supervision non-uniformly across response tokens according to their estimated training value. Most existing criteria, however, focus primarily on optimization need, such as uncertainty or teacher-student disagreement, while task relevance, namely whether the supervision is tied to the semantic content of the current input, remains less directly characterized as a complementary dimension. To address this gap, we introduce Counterfactual Relevance for On-Policy Distillation (CROP), which operationalizes task relevance through a paraphrase-calibrated counterfactual sensitivity margin. For each source prompt, CROP constructs a validated original-paraphrase-counterfactual triplet, holds the student rollout fixed, and measures each response position by its sensitivity to a task-relevant condition change calibrated by its sensitivity to a meaning-preserving rewrite. Matched selection controls show that CROP identifies more useful supervision positions than random or lowest-relevance selection, while component comparisons confirm the value of both counterfactual sensitivity and paraphrase calibration. Across two teacher-student settings, CROP improves aggregate performance by 1.92 and 2.96 points over the strongest non-CROP selector. These results support task relevance as a complementary criterion for selective OPD and establish CROP as a model-internal, contrast-specific method for allocating token-level supervision.
Chinese Translation
在线蒸馏（OPD）通过从当前策略中采样的轨迹对学生语言模型进行监督，但对具有不等监督价值的响应标记赋予相同的信用。选择性在线蒸馏通过根据响应标记的估计训练价值不均匀地分配监督，解决了这一局限性。然而，现有的大多数标准主要关注优化需求，例如不确定性或教师与学生之间的不一致，而任务相关性，即监督是否与当前输入的语义内容相关，仍然作为一个互补维度较少被直接表征。为了解决这一问题，我们引入了用于在线蒸馏的反事实相关性（CROP），通过经过释义校准的反事实敏感度边际来实现任务相关性。对于每个源提示，CROP 构建一个经过验证的原始-释义-反事实三元组，保持学生的回滚不变，并通过其对保持意义的重写的敏感度来测量每个响应位置对任务相关条件变化的敏感度。匹配选择控制显示，CROP 识别出比随机选择或最低相关性选择更有用的监督位置，而组件比较确认了反事实敏感度和释义校准的价值。在两个教师-学生设置中，CROP 的整体性能比最强的非 CROP 选择器提高了 1.92 和 2.96 分。这些结果支持任务相关性作为选择性在线蒸馏的一个互补标准，并确立了 CROP 作为一种模型内部的、对比特定的方法来分配标记级监督。
cs.CL / 66 / 2608.13425
Motor, Cognitive, or Corpus? What Survives Cross-Lingual Transfer in Speech-Based Parkinsons Disease Detection
运动、认知还是语料库？跨语言转移在基于语音的帕金森病检测中的存活因素
Abstract
Self-supervised learning (SSL) speech representations achieve strong performance for Parkinson's disease (PD) detection within individual corpora. However, it remains unclear whether these models capture disease-related characteristics or exploit dataset-specific confounds, particularly since most SSL backbones are pretrained exclusively on healthy speech. To investigate this question, we perform a layer-wise analysis of nine SSL speech backbones using a low-capacity logistic regression probe across three languages. We structure the evaluation as multiple scenarios that progressively introduce distribution shifts in participant identity, recording conditions, language, and pathology. Our results reveal two key findings. First, layer selection is highly corpus-dependent: the optimal representation layer is determined primarily by the source dataset rather than by the SSL architecture itself. Second, the transferred discriminative signal lacks pathological specificity: classifiers trained to detect PD assign similarly high probabilities to both PD and dementia speech in the target corpus. These results highlight critical limitations that must be addressed before speech-based pathology recognition models can be reliably deployed in clinical settings.
Chinese Translation
自监督学习（SSL）语音表示在个体语料库中对帕金森病（PD）检测表现出强大的性能。然而，目前尚不清楚这些模型是否捕捉到了与疾病相关的特征，还是利用了数据集特定的混淆因素，特别是因为大多数SSL骨干网络仅在健康语音上进行预训练。为了解决这个问题，我们对九个SSL语音骨干网络进行了逐层分析，使用低容量的逻辑回归探针在三种语言中进行评估。我们将评估结构化为多个场景，逐步引入参与者身份、录音条件、语言和病理的分布变化。我们的结果揭示了两个关键发现。首先，层选择高度依赖于语料库：最佳表示层主要由源数据集决定，而不是由SSL架构本身决定。其次，转移的区分信号缺乏病理特异性：训练用于检测PD的分类器在目标语料库中对PD和痴呆语音赋予了相似的高概率。这些结果突显了在基于语音的病理识别模型可以在临床环境中可靠部署之前必须解决的关键限制。
cs.CL / 67 / 2608.13430
Are You Sure You're Sure? On the Impact of Instruction Tuning on Confidence and Lexical Diversity
你确定你确定吗？指令调优对信心和词汇多样性的影响
Abstract
Instruction-tuned language models achieve strong performance across a range of generation tasks, but have also recently been shown to exhibit verbalized overconfidence. In question answering, verbalized model overconfidence may be associated with the consistency of the generated supporting rationales. In this paper, we study whether corresponding changes in the lexical diversity of generated answer rationales accompany changes in model confidence induced by instruction tuning. We evaluate three matched base and instruction-tuned models across question-answering benchmarks and find that instruction tuning consistently alters answer confidence, despite limited changes in predictive accuracy and decreases in likelihood-based calibration. Secondly, we observe a non-uniform effect of instruction tuning on rationale diversity: cross-rationale diversity consistently decreases, whereas surface-level lexical diversity varies in both direction and magnitude across models and benchmarks. Finally, we find that these differences persist after controlling for answer selection and rationale length, confirming that confidence and rationale diversity capture distinct effects of instruction tuning.
Chinese Translation
指令调优的语言模型在多种生成任务中表现出色，但最近也显示出口头上的过度自信。在问答任务中，模型的口头过度自信可能与生成的支持性推理的一致性有关。本文研究了在指令调优引起的模型信心变化中，生成的答案推理的词汇多样性是否也随之变化。我们在问答基准测试中评估了三对匹配的基础模型和指令调优模型，发现指令调优始终改变答案信心，尽管预测准确性变化有限且基于似然的校准有所下降。其次，我们观察到指令调优对推理多样性的影响并不均匀：跨推理多样性始终减少，而表层词汇多样性在不同模型和基准测试中则在方向和幅度上各不相同。最后，我们发现这些差异在控制答案选择和推理长度后仍然存在，确认信心和推理多样性捕捉了指令调优的不同影响。
cs.CL / 68 / 2608.13484
Toward a Gricean Retreat: Probing LLMs for Knowledge Boundaries and Referent Specificity
走向格赖斯的退缩：探讨大型语言模型的知识边界和指称特异性
Abstract
When asked about entities outside their knowledge boundary, LLMs routinely fabricate plausible-sounding details rather than backing off to safer, more general claims. We frame this failure through a Gricean lens: a cooperative speaker who is uncertain about a referent retreats up the specificity hierarchy, trading informativeness for truthfulness. We ask whether LLMs have the ingredients to perform this retreat. Using a T-REx-based benchmark that varies entity familiarity and referent specificity, we probe models to answer two questions: (i) do their activations encode whether a referent falls inside the knowledge boundary, and (ii) do they anticipate the specificity of the referent they are about to generate? We find that the answer to both is yes, but the two signals are not reconciled in generation. Models overwhelmingly prefer specific referents even when the entity is unknown to them, and do so even when offered correct generic alternatives. The substrate for a Gricean retreat is present, but the policy that would act on it is not. We position our findings as a first step toward Gricean alignment, training or steering objectives that couple knowledge-boundary awareness to referent-specificity during generation.
Chinese Translation
当被问及其知识边界之外的实体时，大型语言模型（LLMs）通常会编造听起来合理的细节，而不是退回到更安全、更一般的说法。我们通过格赖斯的视角框定这一失败：一个对指称不确定的合作性发言者会在特异性层级上退缩，以换取真实性而牺牲信息量。我们探讨LLMs是否具备执行这种退缩的条件。通过使用一个基于T-REx的基准，该基准变化实体的熟悉度和指称的特异性，我们对模型进行探测，以回答两个问题：（i）它们的激活是否编码指称是否在知识边界内，以及（ii）它们是否预期即将生成的指称的特异性？我们发现两个问题的答案都是肯定的，但这两种信号在生成过程中并未得到协调。模型在面对未知实体时，仍然倾向于选择特定的指称，即使提供了正确的一般替代方案。格赖斯退缩的基础存在，但能够对此采取行动的策略却缺失。我们将我们的发现定位为朝向格赖斯对齐的第一步，旨在训练或引导目标，将知识边界意识与生成过程中的指称特异性结合起来。
cs.CL / 69 / 2608.13515
Measuring Task-Agnostic Training Data Influence Across Language Model Pretraining
跨语言模型预训练测量任务无关训练数据影响
Abstract
Measuring training data influence consistently across language model pretraining is challenging. It is difficult to select downstream tasks or validation sets representative of a model's general capabilities, and reliance on task performance at intermediate checkpoints complicates comparisons across training. We propose a measure of training data influence that does not require selecting a downstream task or validation set as the attribution target. Specifically, we define an example's influence by how much its gradient update reduces the squared distance to the final parameters of a given pretraining run, and estimate this quantity from intermediate checkpoints without retraining. Applying the method to 18 configurations from the Pythia and PolyPythia suites, we find systematic temporal changes in influential data. Early in training, literature-related data are more strongly aligned with the trajectory toward the final parameters, whereas STEM data become more strongly aligned in later stages. This qualitative crossover is broadly consistent across model configurations. Our results provide a tractable trajectory-level view of how influential data change throughout pretraining, complementing influence analyses defined with respect to specific downstream tasks or validation sets.
Chinese Translation
在语言模型预训练中一致地测量训练数据影响是具有挑战性的。选择代表模型一般能力的下游任务或验证集是困难的，并且依赖于中间检查点的任务性能使得跨训练的比较变得复杂。我们提出了一种训练数据影响的测量方法，该方法不需要选择下游任务或验证集作为归因目标。具体而言，我们通过一个示例的梯度更新减少到给定预训练运行的最终参数的平方距离来定义其影响，并从中间检查点估计这一量，而无需重新训练。将该方法应用于 Pythia 和 PolyPythia 套件的 18 种配置，我们发现影响数据在时间上的系统性变化。在训练早期，与文献相关的数据与最终参数的轨迹更强地对齐，而 STEM 数据在后期阶段变得更强地对齐。这种定性交叉在模型配置中普遍一致。我们的结果提供了一种可处理的轨迹级视角，展示了影响数据在整个预训练过程中如何变化，补充了与特定下游任务或验证集相关的影响分析。
cs.CL / 70 / 2608.13517
DFM Mimir v1: An Open HRM Delivering Frontier Performance at 1B Parameters Using Only Permissible Post-Training Data
DFM Mimir v1：一个使用仅允许的后训练数据，在10亿参数下提供前沿性能的开放HRM
Abstract
Current large language model development relies on massive, often non-permissible datasets, creating a high barrier for researchers committed to open-source and ethically sourced data. We introduce Mimir v1, a 1-billion-parameter language model based on the Hierarchical Reasoning Model (HRM) architecture, that is trained from scratch and delivers highly competitive performance for English and sets a new state of the art for Danish using only permissible post-training data. Trained on a mixture of 161 datasets, Mimir v1 outperforms the original HRM-Text 1B and competes with larger frontier models like Qwen 3.5 4B and Gemma 4 E2B, tested across 20 benchmarks for English, Math & Code and Danish. The model is available on the Hugging Face Hub: https://huggingface.co/danish-foundation-models/DFM-Mimir
Chinese Translation
当前大型语言模型的发展依赖于庞大且通常不被允许的数据集，这为致力于开源和伦理数据的研究人员设置了高门槛。我们介绍了Mimir v1，这是一个基于层次推理模型（HRM）架构的10亿参数语言模型，从零开始训练，并在英语方面提供了高度竞争的性能，同时在丹麦语方面设定了新的最先进水平，仅使用允许的后训练数据。Mimir v1在161个数据集的混合上进行训练，超越了原始的HRM-Text 1B，并与更大前沿模型如Qwen 3.5 4B和Gemma 4 E2B竞争，在英语、数学与代码以及丹麦语的20个基准测试中进行了测试。该模型可在Hugging Face Hub上获取：https://huggingface.co/danish-foundation-models/DFM-Mimir
cs.CL / 71 / 2608.13538
SAEVerbalizer: Generating Explanations for Sparse Autoencoder Features via Representation Verbalization
SAEVerbalizer：通过表征语言化生成稀疏自编码器特征的解释
Abstract
Sparse autoencoders (SAEs) are proposed to extract numerous features from large language model (LLM) representations, yet explaining these features still relies primarily on external observation. This reliance leads to superficial explanations inferred from observed model behavior and computational inefficiency from collecting such behavioral evidence at scale. We introduce SAEVerbalizer, a framework that injects SAE decoder directions into an LLM's representations and fine-tunes the LLM's downstream layers to generate natural-language explanations of the injected features. Once trained, the resulting verbalizer explains SAE features directly from decoder directions, addressing both limitations. Our experiments show that the learned verbalization capability generalizes to unseen features, transfers across separately trained SAE dictionaries, and, with a lightweight adapter, extends to SAE features from different LLMs. Intervention experiments show that injecting multiple directions yields an explanation combining their meanings, while reversing individual directions produces corresponding meaning shifts.
Chinese Translation
稀疏自编码器（SAEs）被提出用于从大型语言模型（LLM）表征中提取众多特征，但对这些特征的解释仍主要依赖于外部观察。这种依赖导致了从观察到的模型行为中推断出的肤浅解释，以及在大规模收集此类行为证据时的计算效率低下。我们提出了SAEVerbalizer，一个将SAE解码器方向注入LLM表征的框架，并微调LLM的下游层以生成注入特征的自然语言解释。一旦训练完成，生成的语言化工具可以直接从解码器方向解释SAE特征，从而解决了这两种局限性。我们的实验表明，学习到的语言化能力能够推广到未见特征，能够在单独训练的SAE词典之间迁移，并且通过轻量级适配器扩展到来自不同LLM的SAE特征。干预实验表明，注入多个方向会产生结合其含义的解释，而反转单个方向则会产生相应的意义转变。
cs.CL / 72 / 2608.13545
LittleLearner: Language Models Under Pedagogically Controlled Knowledge Exposure
LittleLearner：在教育控制的知识暴露下的语言模型
Abstract
Modern language models are trained on heterogeneous web-scale text corpora. Consequently, studying knowledge and skill acquisition is difficult, as prior exposure to related content is hard to characterize. To address this challenge, we introduce LITTLECURRICULUM, a curated 88B-token pretraining corpus tailored to U.S. elementary school material, explicitly excluding concepts, facts, and vocabulary taught above Grade 5. Training a 5B-parameter LLM from scratch on LITTLECURRICULUM yields LITTLELEARNER, a model with sufficient language competence for open-ended evaluation, yet with clear knowledge and capability boundaries mapped to interpretable curriculum guidelines. We release LITTLECURRICULUM and LITTLELEARNER as a developmentally restricted sandbox to study how models acquire, represent, and use data under a well-defined training scope. We illustrate the sandbox's utility in a first suite of experiments on injecting new knowledge through post-training and in-context learning. These methods let LITTLELEARNER better utilize existing knowledge, but do not raise out-of-scope capabilities. Our findings underscore the value of this controlled environment for future investigations.
Chinese Translation
现代语言模型是在异质的网络规模文本语料库上训练的。因此，研究知识和技能的获取变得困难，因为先前对相关内容的接触难以表征。为了解决这一挑战，我们引入了LITTLECURRICULUM，这是一个经过精心策划的88B-token预训练语料库，专门针对美国小学教材，明确排除了五年级以上教授的概念、事实和词汇。从头开始在LITTLECURRICULUM上训练一个5B参数的LLM，产生了LITTLELEARNER，这是一个具有足够语言能力进行开放式评估的模型，但其知识和能力的边界清晰地映射到可解释的课程指南上。我们发布LITTLECURRICULUM和LITTLELEARNER作为一个发展受限的沙盒，以研究模型在明确的训练范围内如何获取、表示和使用数据。我们通过一系列实验展示了这个沙盒的实用性，实验内容包括通过后训练和上下文学习注入新知识。这些方法使LITTLELEARNER能够更好地利用现有知识，但并未提升其超出范围的能力。我们的发现强调了这一受控环境对未来研究的价值。