cs.RO / 1 / 2608.10023
Protection Levels for Vision-Based Pose Estimation
基于视觉的姿态估计的保护水平
Abstract
Vision-based navigation complements Global Navigation Satellite Systems, but certification demands integrity guarantees that account for faulty measurements. Previous work presented a probabilistic computer vision pipeline for runway-based pose estimation with fault detection inspired by Receiver Autonomous Integrity Monitoring. This work extends that framework by deriving protection levels, which provide probabilistic bounds on pose error that remain valid under undetected faults. We present an algorithm for computing protection levels for the nonlinear Perspective-$n$-Point problem applied to an aviation setting. The algorithm covers all six degrees of freedom of the aircraft pose (position and orientation) directly. We analyze the effect of measurement redundancy, pixel-level prediction uncertainty, and runway distance on the resulting protection levels. To make the results tangible, we demonstrate tradeoffs in the protection levels on an illustrative runway example.
Chinese Translation
基于视觉的导航补充了全球导航卫星系统，但认证要求提供完整性保证，以考虑故障测量。之前的研究提出了一种基于概率的计算机视觉流程，用于基于跑道的姿态估计，并结合了受体自主完整性监测（Receiver Autonomous Integrity Monitoring）启发的故障检测。本研究扩展了该框架，通过推导保护水平，提供在未检测到故障的情况下仍然有效的姿态误差的概率界限。我们提出了一种算法，用于计算应用于航空环境的非线性透视-$n$-点问题的保护水平。该算法直接覆盖了飞机姿态的所有六个自由度（位置和方向）。我们分析了测量冗余、像素级预测不确定性和跑道距离对结果保护水平的影响。为了使结果更具可操作性，我们在一个示例跑道上展示了保护水平之间的权衡。
cs.RO / 2 / 2608.10025
The Impact of Operational-Data Fidelity when Assessing Safety-Critical Autonomous-Vehicle Software
评估安全关键自主车辆软件时操作数据保真度的影响
Abstract
For safety-critical software, data from the software's operational past (e.g. a sequence of success and failure events experienced by the software) can provide strong statistical support for reliability claims about the software. However, such data might not describe past software failure events in sufficient detail, and this might leave a reliability assessment (based on this data) unable to account for important features of past software failures. In this paper, by extending conservative Bayesian inference (CBI) techniques used in reliability assessment, we illustrate a principled statistical approach for checking the robustness of reliability claims derived from insufficiently detailed operational data. We demonstrate the extent to which insufficient detail in operational data can undermine software reliability claims in autonomous vehicle (AV) safety assessment scenarios. Reliability claims derived from insufficiently fine-grained data might be dangerously optimistic, despite a concerted effort by an assessor to use such data conservatively during the assessment. While these findings are consistent with previous work on the impact of statistical model fidelity in Bayesian software reliability assessments, our work clarifies why attempts to use low-fidelity data conservatively can be naive, and we give the first conservative estimates of the impact of data fidelity on assessments.
Chinese Translation
对于安全关键软件，软件操作历史数据（例如，软件经历的成功和失败事件序列）可以为软件的可靠性声明提供强有力的统计支持。然而，这些数据可能无法充分详细地描述过去的软件失败事件，这可能导致基于这些数据的可靠性评估无法考虑到过去软件失败的重要特征。本文通过扩展用于可靠性评估的保守贝叶斯推断（CBI）技术，展示了一种原则性的统计方法，用于检查从不足详细的操作数据中得出的可靠性声明的稳健性。我们展示了操作数据中不足的细节在自主车辆（AV）安全评估场景中如何削弱软件可靠性声明的程度。尽管评估者在评估过程中努力保守地使用这些数据，但从不足细粒度数据中得出的可靠性声明可能会过于乐观。尽管这些发现与之前关于贝叶斯软件可靠性评估中统计模型保真度影响的研究一致，但我们的工作阐明了为什么尝试保守地使用低保真度数据可能是天真的，并且我们首次给出了数据保真度对评估影响的保守估计。
cs.RO / 3 / 2608.10056
Navigating the Proximity-Safety Balance: Constraint Decomposition for Human Following in Pedestrian Crowds
平衡接近性与安全性：行人拥挤环境中人类跟随的约束分解
Abstract
Following a target human in crowded environments involves an inherent conflict between staying close to the target and navigating safely among surrounding pedestrians and obstacles. This conflict becomes more severe in dense scenarios, where aggressive following risks collisions and conservative margins lead to target loss, especially when pedestrian behaviors are unfamiliar or unpredictable. Existing reinforcement learning (RL) methods typically encode these competing objectives into a single dense reward, but the resulting proximity-safety balance is implicit and difficult to adjust across conditions. To address this, we decompose the human-following task into a sparse task reward and independent cost constraints within a multi-constraint RL formulation, where each constraint is managed through cost thresholds with direct behavioral meaning rather than implicit reward weight ratios, allowing explicit and tunable control over the trade-off. We further quantify the prediction uncertainty of human motions and integrate these estimates into the RL costs to enhance safety under unpredictable conditions. Extensive experiments across both in-distribution and out-of-distribution settings demonstrate that our method achieves an effective proximity-safety balance compared to baselines. Real-robot deployment further validates the feasibility of our method in real-world scenarios. More details are available on our project page: https://nav-ps-balance.github.io/.
Chinese Translation
在拥挤环境中跟随目标人类涉及在与周围行人和障碍物安全导航与紧靠目标之间的固有冲突。在密集场景中，这种冲突变得更加严重，激进的跟随可能导致碰撞，而保守的安全距离则可能导致目标丢失，尤其是在行人行为不熟悉或不可预测的情况下。现有的强化学习（RL）方法通常将这些相互竞争的目标编码为单一的密集奖励，但由此产生的接近性与安全性平衡是隐式的，且在不同条件下难以调整。为了解决这个问题，我们将人类跟随任务分解为稀疏任务奖励和独立的成本约束，采用多约束RL框架，其中每个约束通过具有直接行为意义的成本阈值进行管理，而不是隐式的奖励权重比率，从而实现对权衡的明确和可调控的控制。我们进一步量化人类运动的预测不确定性，并将这些估计整合到RL成本中，以增强在不可预测条件下的安全性。在分布内和分布外设置下进行的大量实验表明，与基线相比，我们的方法实现了有效的接近性与安全性平衡。实际机器人部署进一步验证了我们方法在现实场景中的可行性。更多细节请访问我们的项目页面：https://nav-ps-balance.github.io/
cs.RO / 4 / 2608.10220
Whole-Body Planning for Humanoids Navigating Confined Spaces via Self-Collision Avoidance References
通过自碰撞避免参考进行人形机器人在狭小空间中的全身规划
Abstract
Humanoid locomotion in highly confined environments requires navigating dense environmental obstacles and complex self-collision bounds while maintaining multi-contact dynamic feasibility. Traditional trajectory optimizers frequently struggle in these restricted spaces, as navigating the large collision space with splines on particle abstractions is insufficient and leads to poor local minima. To address this, we propose a three-stage whole-body planning framework that formulates kinematic path planning directly over kinematically reachable rigid-body volumes. By integrating differentiable collision avoidance into a reachability-constrained formulation, our framework synthesizes volume-informed guides that reliably guide a full-order trajectory optimizer over long horizons. We show that these optimized plans serve as high-quality references to train a residual reinforcement learning policy for robust online execution. We validate our approach on the Unitree G1 humanoid across three benchmark testbeds exceeding NIST emergency response standards, achieving restricted confinement ratios ($C_r < 1.5$). Our framework generates feasible trajectories across 12-to-18-second tasks with complex foot and hand contacts where standard baselines fail, while the learned policy successfully tracks these plans under extensive domain randomization in physics simulation.
Chinese Translation
在人形机器人在高度受限环境中的运动需要在保持多接触动态可行性的同时，导航密集的环境障碍物和复杂的自碰撞界限。传统的轨迹优化器在这些受限空间中常常面临困难，因为在粒子抽象上使用样条曲线导航大碰撞空间是不够的，导致出现较差的局部极小值。为了解决这个问题，我们提出了一种三阶段的全身规划框架，该框架直接在运动学可达的刚体体积上制定运动学路径规划。通过将可微分的碰撞避免集成到受可达性约束的公式中，我们的框架合成了基于体积的信息引导，这些引导可靠地引导全阶轨迹优化器在长时间范围内进行优化。我们展示了这些优化计划作为高质量参考，能够训练残差强化学习策略以实现稳健的在线执行。我们在Unitree G1人形机器人上验证了我们的方法，涵盖了超过NIST应急响应标准的三个基准测试平台，达到了受限的约束比（$C_r < 1.5$）。我们的框架在复杂的脚和手接触任务中生成可行的轨迹，任务持续时间为12到18秒，而标准基线则失败，而学习到的策略在物理仿真中的广泛领域随机化下成功跟踪这些计划。
cs.RO / 5 / 2608.10232
FACT: Failure-Aware Causal Training for World-Action Models
FACT：面向失败的因果训练用于世界-动作模型
Abstract
Recent world-action models (WAMs) show that co-training policies with future prediction can provide physical priors for action generation. Building on the future-prediction ability of video models, many WAMs generate future videos and recover actions with inverse-dynamics models, or use these predicted videos as goal conditions for action generation. In both cases, the world model is trained mostly on successful demonstrations and has little reason to predict the consequences of bad actions. We introduce FACT, a causal World-Action Model that predicts future video and task progress conditioned on the executed action. This action-conditioned interface allows failure rollouts to supervise action consequences, turning bad actions into valid future targets rather than being discarded. Failure-aware training makes the progress predictor aware of both successful and failed action outcomes, which can optionally be used to score sampled action candidates at inference. Extensive experiments on simulation and real-world bimanual manipulation tasks show that FACT outperforms many existing baselines, improves as failure data are incorporated into training, and reduces success-biased future hallucination under bad actions. See more details at https://fact-wam.github.io/
Chinese Translation
近期的世界-动作模型（WAMs）表明，结合未来预测的共同训练策略可以为动作生成提供物理先验。基于视频模型的未来预测能力，许多WAMs生成未来视频并利用逆动力学模型恢复动作，或者将这些预测视频作为动作生成的目标条件。在这两种情况下，世界模型主要在成功示范上进行训练，几乎没有理由去预测不良动作的后果。我们提出了FACT，一种因果世界-动作模型，它根据执行的动作预测未来视频和任务进展。这个基于动作的接口允许失败回滚来监督动作后果，将不良动作转化为有效的未来目标，而不是被丢弃。面向失败的训练使得进展预测器能够意识到成功和失败的动作结果，这可以在推理时选择性地用于对采样动作候选进行评分。在模拟和真实世界双手操作任务上的大量实验表明，FACT在许多现有基准上表现优越，随着失败数据的纳入训练而改善，并在不良动作下减少成功偏向的未来幻觉。更多细节请见 https://fact-wam.github.io/
cs.RO / 6 / 2608.10367
A Neural Network Based Teleoperation for Remote Controlled Vehicles
基于神经网络的远程控制车辆遥操作
Abstract
Direct teleoperation of vehicles faces critical technical bottlenecks: communication latency and the operator's inability to physically perceive unmodeled environmental disturbances (e.g., aerodynamic drag, bank angles) coupled with highly nonlinear tire-road dynamics. To address these challenges, we propose a tailored unilateral teleoperation framework. The system integrates the Wave Variable (WV) approach to passively guarantee stability under stochastic delays, and an adaptive Radial Basis Function Network (RBFN) to actively compensate for vehicle-specific uncertainties. Unlike existing WV-neural network architectures designed for bilateral robotic arms, our framework features decoupled adaptive laws specifically designed for vehicle longitudinal and lateral dynamics. Furthermore, compared to model-heavy predictive controllers, the model-free RBFN offers rapid online adaptation without heavy computational overhead. Building upon our preliminary theoretical formulation, this brief paper presents comprehensive comparative analyses and real-world hardware validations. Simulation benchmarks against PID, LQR, MPC, and NMPC demonstrate that the RBFN achieves superior robustness against unmodeled disturbances while requiring orders of magnitude less execution time than MPC and NMPC, making it ideal for resource-constrained vehicle edge computing. Finally, hardware-in-the-loop experiments using a 1/10th scale vehicle over a 4G network validate the system's practical feasibility, safety, and robust trajectory tracking under physical road uncertainties.
Chinese Translation
车辆的直接遥操作面临着关键的技术瓶颈：通信延迟和操作员无法物理感知未建模的环境干扰（例如，空气阻力、倾斜角度）以及高度非线性的轮胎-道路动态。为了解决这些挑战，我们提出了一种定制的单向遥操作框架。该系统集成了波变量（Wave Variable, WV）方法，以被动方式保证在随机延迟下的稳定性，并采用自适应径向基函数网络（Radial Basis Function Network, RBFN）主动补偿特定于车辆的不确定性。与现有为双向机器人手臂设计的WV-神经网络架构不同，我们的框架具有专门为车辆纵向和横向动态设计的解耦自适应法则。此外，与模型重的预测控制器相比，无模型的RBFN在不需要大量计算开销的情况下，能够快速进行在线适应。在我们初步理论公式的基础上，本文简要呈现了全面的比较分析和实际硬件验证。与PID、LQR、MPC和NMPC的仿真基准测试表明，RBFN在应对未建模干扰方面具有更优的鲁棒性，同时所需的执行时间比MPC和NMPC少几个数量级，使其非常适合资源受限的车辆边缘计算。最后，使用1/10比例车辆在4G网络下进行的硬件在环实验验证了系统的实际可行性、安全性和在物理道路不确定性下的鲁棒轨迹跟踪能力。
cs.RO / 7 / 2608.10383
Real-World Cooperative Bimanual Dexterous Grasp of Large Objects from Single-View Observations
基于单视图观察的大型物体真实世界协作双手灵巧抓取
Abstract
Bimanual dexterous grasping of large objects is a critical challenge in robotic manipulation. However, most existing studies focus on sequential manipulation rather than cooperative grasping, and methods addressing such bimanual tasks have largely been limited to simulation. These limitations stem from the difficulty of acquiring full 3D object models and generating physically plausible grasping actions. To fill this gap, we propose a real-world bimanual grasping framework that includes: a multimodal dataset capturing joint angles, visual observations and force signals; a Denoising Diffusion Probabilistic Model (DDPM)-based module that generates joint-level grasp configurations from segmented point clouds; and an execution strategy that integrates motion planning with online grasp refinement to ensure physical stability and feasibility. Our approach enables the synthesis of executable bimanual grasps from single-view inputs, reducing dependence on complete 3D object models and ensuring stable real-world performance. Experiments on a dual-arm robot demonstrate high success rates across unseen objects with varying geometries and poses, and ablation studies confirm the contributions of key components of our system.
Chinese Translation
大型物体的双手灵巧抓取是机器人操作中的一个关键挑战。然而，大多数现有研究集中于顺序操作而非协作抓取，且解决此类双手任务的方法在很大程度上局限于模拟。这些限制源于获取完整的三维物体模型和生成物理上合理的抓取动作的困难。为填补这一空白，我们提出了一个真实世界的双手抓取框架，包括：一个捕捉关节角度、视觉观察和力信号的多模态数据集；一个基于去噪扩散概率模型（Denoising Diffusion Probabilistic Model, DDPM）的模块，该模块从分割的点云生成关节级抓取配置；以及一个将运动规划与在线抓取优化相结合的执行策略，以确保物理稳定性和可行性。我们的方法使得能够从单视图输入合成可执行的双手抓取，减少对完整三维物体模型的依赖，并确保在真实世界中的稳定性能。在双臂机器人上的实验表明，在不同几何形状和姿态的未见物体上具有较高的成功率，消融研究确认了我们系统关键组件的贡献。
cs.RO / 8 / 2608.10387
Hip Energized Monopedal Hopping
髋部能量驱动的单足跳跃
Abstract
We present a novel stepping strategy for pitch unlocked planar monopeds where the reaction torques from stabilizing pitch with a conventional PD + feedfoward controller are recruited to counteract energetic losses from damping. By moving the location of the mass center, our controller increases the pitch stabilization torque, thereby adding energy to the gait. A new stepping policy adjusts the distribution of energy between the radial and angular degrees of freedom to counteract dissipative losses and achieve a user specified balance between steady state fore-aft speed and apex height. Hybrid averaging analysis yields closed form expressions for the fixed points and eigenvalues of the resulting gait, lending insight into the interplay between the physical and control parameters' influence on performance. Simulation studies on a generic 5 link biped and a careful model of the Penn Jerboa reveal a useful correspondence to these analytical predictions. Physical experiments on the Penn Jerboa exhibit stable locomotion with speeds ranging from 1.02 m/s to 1.77 m/s (5.10 leg lengths/s to 8.85 leg lengths/s) in a manner effectively approximated by the mathematical analysis.
Chinese Translation
我们提出了一种新颖的步态策略，适用于俯仰解锁的平面单足机器人，其中通过传统的PD + 前馈控制器稳定俯仰所产生的反应力矩被用来抵消因阻尼造成的能量损失。通过移动质心的位置，我们的控制器增加了俯仰稳定力矩，从而为步态增加能量。一种新的步态策略调整了径向和角度自由度之间的能量分配，以抵消耗散损失，并实现用户指定的稳态前后速度与最高点高度之间的平衡。混合平均分析提供了结果步态的固定点和特征值的封闭形式表达式，揭示了物理参数与控制参数对性能影响之间的相互作用。对通用的五连杆双足机器人和Penn Jerboa的精确模型的仿真研究显示了与这些分析预测的有用对应关系。在Penn Jerboa上的物理实验展示了稳定的运动，速度范围从1.02 m/s到1.77 m/s（5.10腿长/s到8.85腿长/s），这种运动方式有效地被数学分析所近似。
cs.RO / 9 / 2608.10449
PBD-AG: Persistent Baseline-Delta Active Graphs with Uncertainty-Aware Inspection for Long-Horizon Service Robots
PBD-AG：具有不确定性感知检查的持久基线-增量主动图用于长时间服务机器人
Abstract
Long-horizon service robots require persistent world models that can be built autonomously in unseen environments and revised as task-relevant objects change. Existing methods rely on online mapping, which accumulates localization and observation errors, static scene representations that cannot capture persistent object changes, or holistic vision-language predictions that lack verifiable 3D geometric evidence. We present PBD-AG, a persistent baseline-delta active graph framework that decouples robot-verified stable fixtures from revisable dynamic object events. Under our framework, the robot autonomously bootstraps the structural baseline from onboard exploration and inspects discovered fixtures to ground hierarchical object beliefs. PBD-AG maintains reliability-weighted object states over geometry, semantics, identity, existence, and support relations, utilizing a geometric visibility gate to mitigate false deletions under occlusion. Inspection viewpoints are selected by a graph-conditioned policy that balances target coverage, travel cost, collision risk, and redundant observation. Simulation experiments in multiple environments and under controlled dynamic evaluation show higher aggregate coarse-fixture F1 than capability-matched controls, as well as stronger identity continuity and event recall. A qualitative physical-robot demonstration further illustrates integration with onboard sensing, providing a traceable world model for long-horizon robotic perception.The project page of PBD-AG is available at https://shuobao214.github.io/PBD-AG/
Chinese Translation
长时间服务机器人需要能够在未知环境中自主构建的持久世界模型，并随着任务相关对象的变化进行修正。现有方法依赖于在线映射，这会累积定位和观察误差，静态场景表示无法捕捉持久对象的变化，或整体视觉-语言预测缺乏可验证的三维几何证据。我们提出了PBD-AG，一种持久基线-增量主动图框架，该框架将机器人验证的稳定固定物与可修订的动态对象事件解耦。在我们的框架下，机器人自主从机载探索中引导结构基线，并检查发现的固定物以建立分层对象信念。PBD-AG维护关于几何、语义、身份、存在和支持关系的可靠性加权对象状态，利用几何可见性门来减轻遮挡下的虚假删除。检查视点由图条件策略选择，该策略平衡目标覆盖、旅行成本、碰撞风险和冗余观察。在多个环境和受控动态评估下的仿真实验显示，PBD-AG的粗固定物F1总和高于能力匹配的对照组，并且具有更强的身份连续性和事件召回。定性物理机器人演示进一步说明了与机载传感的集成，为长时间机器人感知提供了可追溯的世界模型。PBD-AG的项目页面可访问 https://shuobao214.github.io/PBD-AG/
cs.RO / 10 / 2608.10484
Lost in Reconstruction: Aligning Action Representations with Language in Vision-Language-Action Models
重建中的迷失：在视觉-语言-动作模型中对齐动作表示与语言
Abstract
Action verbs describe not only the physical outcomes of actions, but also how those actions are performed. Yet action representations in vision-language-action models (VLAs) are typically optimized for reconstruction under L1/L2 losses in raw action space, where numerical proximity need not reflect linguistically meaningful distinctions. On BridgeV2, we show that action trajectories contain verb-grounding information beyond visual state changes, and that reconstruction-only discrete tokenization systematically erodes this information. To address this problem, we introduce SALT, a Semantically ALigned action Tokenizer that augments a VQ-VAE-style tokenizer with an auxiliary objective requiring a frozen vision-language model to recover the episode instruction from quantized action latents. Policies trained with SALT achieve 71.9% average success in SimplerEnv, compared with 42.7% for a reconstruction-only VQ-VAE tokenizer and 31.2% for FAST. SALT also develops verb-specialized codes while maintaining reconstruction fidelity. These results show that robot action trajectories provide a source of language grounding and that preserving this structure in action representations can substantially improve language-conditioned control.
Chinese Translation
动作动词不仅描述了动作的物理结果，还描述了这些动作是如何执行的。然而，在视觉-语言-动作模型（VLA）中，动作表示通常是在原始动作空间下通过L1/L2损失进行优化的重建，其中数值接近性未必反映出语言上有意义的区分。在BridgeV2上，我们展示了动作轨迹包含超越视觉状态变化的动词基础信息，而仅进行重建的离散标记化系统性地侵蚀了这一信息。为了解决这个问题，我们引入了SALT（语义对齐动作标记器），该标记器在VQ-VAE风格的标记器基础上增加了一个辅助目标，要求冻结的视觉-语言模型从量化的动作潜变量中恢复情节指令。使用SALT训练的策略在SimplerEnv中实现了71.9%的平均成功率，而仅进行重建的VQ-VAE标记器为42.7%，FAST为31.2%。SALT还开发了动词专用代码，同时保持了重建的保真度。这些结果表明，机器人动作轨迹提供了语言基础的来源，并且在动作表示中保留这种结构可以显著改善基于语言的控制。
cs.RO / 11 / 2608.10485
JitTrack: Onboard Multi-Object Tracking Against Viewpoint Jitter for Agile UAVs
JitTrack：针对机动无人机视角抖动的机载多目标跟踪
Abstract
Multi-object tracking (MOT) onboard agile unmanned aerial vehicles (UAVs) remains challenging due to severe viewpoint jitter induced by camera ego-motion. Rapid attitude changes during flight often lead to significant target displacement across frames, causing inaccurate target association and degraded tracking performance. Existing UAV MOT methods are primarily evaluated on offline benchmarks and seldom address the practical requirements of real-world onboard deployment, including robustness to camera motion and active target following. To address these challenges, we propose JitTrack, an active onboard multi-object tracking framework that accommodates drone dynamics and camera ego-motion. Built upon a query-based transformer tracker, JitTrack introduces semantic refinement to improve the detection of emerging targets, motion-aware query rectification to compensate for target misalignment caused by viewpoint jitter, and a motion-inspired denoising training strategy that simulates camera motion patterns for robust supervision. Furthermore, we develop a perception-planning-control closed-loop tracking pipeline for real-world deployment, enabling collision-free and physically feasible target following on agile UAVs. Extensive experiments on public UAV MOT benchmarks demonstrate consistent improvements over the baseline method, while real-world flight experiments validate the effectiveness and practicality of JitTrack for robust onboard visual tracking under viewpoint jitter.
Chinese Translation
机载机动无人机（UAV）的多目标跟踪（MOT）仍然面临挑战，因为相机自运动引起的严重视角抖动。在飞行过程中，快速的姿态变化常常导致目标在帧间的显著位移，从而造成不准确的目标关联和降低的跟踪性能。现有的无人机MOT方法主要在离线基准上进行评估，鲜有考虑实际机载部署的需求，包括对相机运动的鲁棒性和主动目标跟踪。为了解决这些挑战，我们提出了JitTrack，一个适应无人机动态和相机自运动的主动机载多目标跟踪框架。JitTrack基于查询的变换器跟踪器，引入了语义细化以改善新兴目标的检测，运动感知的查询校正以补偿因视角抖动导致的目标错位，以及一种受运动启发的去噪训练策略，模拟相机运动模式以实现鲁棒监督。此外，我们开发了一个感知-规划-控制闭环跟踪管道，以便于在现实世界中的部署，实现机动无人机的无碰撞和物理可行的目标跟踪。在公共无人机MOT基准上的大量实验表明，相较于基线方法，JitTrack在性能上有持续的提升，而现实世界的飞行实验验证了JitTrack在视角抖动下进行鲁棒机载视觉跟踪的有效性和实用性。
cs.RO / 12 / 2608.10542
Nonlinear Model Predictive Control via Sequential Convex Programming for Drone-to-Drone Docking
通过序列凸编程实现无人机间非线性模型预测控制
Abstract
Autonomous mid-air docking of multi-rotor vehicles under disturbance-driven target motion poses a constrained non-linear trajectory optimization challenge. This work formulates the docking task as a finite-horizon optimal control problem based on a reduced-order nonlinear model augmented with disturbance states. The resulting problem is solved using sequential convex programming within a receding-horizon framework to generate dynamically feasible docking trajectories. State estimation with noisy measurements is incorporated to enable robust relative motion prediction, while trajectory execution is validated in a high-fidelity rigid-body MuJoCo simulation environment. The proposed framework is evaluated for stationary and constant-velocity target motions, demonstrating reliable convergence to the docking interface while satisfying geometric capture constraints. Quantitatively, the method maintains negligible docking-cone violations and terminal state errors within prescribed tolerances, and achieves consistent, safe docking performance for cone half-angles as low as 10 degrees. Robust operation is observed for wind disturbance levels up to a standard deviation of 0.5, while preserving bounded approach velocities and stable control effort. These results demonstrate the effectiveness of the SCP-based trajectory optimization framework for disturbance-robust aerial docking under estimation uncertainty.
Chinese Translation
在干扰驱动的目标运动下，多旋翼飞行器的自主空中对接构成了一个受限的非线性轨迹优化挑战。本研究将对接任务表述为基于增强干扰状态的降阶非线性模型的有限时域最优控制问题。通过在回退时域框架内使用序列凸编程解决所得到的问题，以生成动态可行的对接轨迹。结合带有噪声测量的状态估计，以实现鲁棒的相对运动预测，同时在高保真刚体MuJoCo仿真环境中验证轨迹执行。所提出的框架针对静止和匀速目标运动进行了评估，展示了在满足几何捕获约束的同时，可靠地收敛到对接接口。定量分析表明，该方法在规定的容忍度内保持了微不足道的对接锥违规和终端状态误差，并在锥半角低至10度时实现了一致、安全的对接性能。在风干扰水平达到标准差0.5的情况下，观察到鲁棒操作，同时保持有限的接近速度和稳定的控制努力。这些结果证明了基于序列凸编程的轨迹优化框架在估计不确定性下实现鲁棒空中对接的有效性。
cs.RO / 13 / 2608.10600
BooST: Bridging Semantics and Motions for Efficient Skill Transfer
BooST：桥接语义与运动以实现高效技能转移
Abstract
Skill abstraction---the process of learning reusable and temporally extended behaviors---has emerged as a key paradigm for improving sample efficiency and generalization in robot learning. For efficient skill transfer to real robots, learned skills must generalize across tasks and domains, remain robust to visual and dynamic perturbations, and be efficient enough for practical deployment. However, existing methods typically satisfy only a subset of these properties, as they capture either high-level semantic intent (what) or low-level motion dynamics (how). This incomplete skill transfer yields weak priors for policy learning, thereby demanding substantial in-domain data for downstream adaptation. To address these challenges, we introduce BooST, a two-stage framework that explicitly bridges semantics and motions to satisfy all three desiderata. BooST first leverages a cross-modal VQ-VAE to capture both semantic intent and motion dynamics, yielding a unified skill representation. It then distills this representation into a lightweight policy for efficient downstream adaptation to new tasks. Extensive experiments across simulation and real-robot settings demonstrate that BooST achieves superior few-shot adaptation, cross-domain skill transfer, and robustness to dynamic visual distractors, while maintaining a lightweight yet expressive design suitable for real-world deployment.
Chinese Translation
技能抽象——学习可重用和时间延续行为的过程——已成为提高机器人学习中样本效率和泛化能力的关键范式。为了将学习到的技能高效地转移到真实机器人上，这些技能必须能够在任务和领域之间泛化，保持对视觉和动态扰动的鲁棒性，并且足够高效以便于实际部署。然而，现有方法通常仅满足这些属性中的一部分，因为它们要么捕捉高层次的语义意图（what），要么捕捉低层次的运动动态（how）。这种不完整的技能转移导致政策学习的先验信息较弱，从而需要大量的领域内数据进行下游适应。为了解决这些挑战，我们提出了BooST，一个两阶段框架，明确地桥接语义与运动，以满足这三种需求。BooST首先利用跨模态的VQ-VAE来捕捉语义意图和运动动态，从而生成统一的技能表示。然后，它将该表示提炼为轻量级政策，以便高效地适应新任务。通过在模拟和真实机器人环境中的广泛实验，BooST在少量样本适应、跨领域技能转移和对动态视觉干扰的鲁棒性方面表现出色，同时保持轻量且富有表现力的设计，适合于现实世界的部署。
cs.RO / 14 / 2608.10618
Toward the Cognitive--Physical Limits of Embodied Intelligence through a World-Model-Centric Autonomous Racing Agent
通过以世界模型为中心的自主赛车代理探索具身智能的认知与物理极限
Abstract
Embodied artificial intelligence aims to develop agents that perceive, reason, and act through continuous interaction with the physical world. However, most embodied systems are still evaluated within conservative safety margins or moderate interaction regimes, leaving their capability boundaries under extreme conditions insufficiently understood. Autonomous racing provides a stringent testbed by combining high-frequency localization and perception, adversarial interaction, near-saturated vehicle dynamics, and strict safety constraints. Existing systems push high-speed performance but rarely model and refine cognitive and physical limits jointly. Here we show that a world-model-centric autonomous racing agent provides a concrete step toward exploring these coupled limits. The framework learns predictive world models from near-limit successes and failures to capture interaction evolution, ego dynamics, and feasible-motion boundaries, coupling world-state construction, future-aware reasoning, and near-limit control in a closed-loop refinement process. Training data were collected from real-vehicle autonomous racing, where the onboard system maintained robust localization and perception at speeds up to 256.3 km/h and peak lateral acceleration of 26.8 m/s$^2$. In full-scale simulated racing, the well trained world-model-centric agent achieves an 88.3% interaction success rate across various challenging simulated racing scenarios. Closed-loop refinement of the world model and policy further improved utilization of cognitive-physical limits, recovery from failure modes, and generalization across varying conditions and unseen circuits. These results suggest a boundary-aware methodology in which world models help embodied agents represent, predict, and continually refine their capability boundaries for safer real-world deployment.
Chinese Translation
具身人工智能旨在开发能够通过与物理世界的持续互动进行感知、推理和行动的代理。然而，大多数具身系统仍然在保守的安全边际或适度的互动模式下进行评估，因此在极端条件下的能力边界尚未得到充分理解。自主赛车通过结合高频率的定位与感知、对抗性互动、近饱和的车辆动态和严格的安全约束，提供了一个严格的测试平台。现有系统推动高速性能，但很少同时建模和优化认知与物理极限。在此，我们展示了以世界模型为中心的自主赛车代理为探索这些耦合极限提供了一个具体的步骤。该框架从近极限的成功与失败中学习预测世界模型，以捕捉互动演变、自我动态和可行运动边界，将世界状态构建、未来感知推理和近极限控制结合在一个闭环优化过程中。训练数据来自真实车辆的自主赛车，其中车载系统在高达256.3 km/h的速度和26.8 m/s²的峰值横向加速度下保持了强大的定位和感知能力。在全尺度模拟赛车中，经过良好训练的以世界模型为中心的代理在各种具有挑战性的模拟赛车场景中达到了88.3%的互动成功率。世界模型和策略的闭环优化进一步提高了认知-物理极限的利用、从失败模式中恢复的能力，以及在不同条件和未见赛道上的泛化能力。这些结果表明了一种边界感知的方法论，其中世界模型帮助具身代理表示、预测并不断优化其能力边界，以实现更安全的现实世界部署。
cs.RO / 15 / 2608.10623
When Your State Estimator Has Lost The Plot: Detecting Estimator Failures Via Spectral Analysis
当状态估计器失去方向时：通过频谱分析检测估计器故障
Abstract
Reliable onboard state estimation is essential for safe robotic operation, yet unmodeled disturbances, such as sensor aliasing or out-of-distribution noise, still cause estimators to degrade or fail completely. While many methods aim to improve estimator robustness, only a few provide introspective mechanisms to assess estimate quality. Existing uncertainty measures, such as covariances, rely on idealized assumptions and tend to be overconfident, and more recent data-driven approaches are typically tied to their training data distributions. We propose a sensor-agnostic introspective method that assesses estimator health by analyzing the frequency-domain power distribution of recent velocity estimates. The method is evaluated using outdoor flight data from an aerial robot running visual-inertial, LiDAR-inertial, and radar-inertial odometry. The dataset includes multiple estimator failures, enabling analysis of several frequency-domain indicators, such as signal power, spectral bandwidth, and entropy. We observe consistent spectral power differences between healthy and degraded estimates, allowing detection of 51%-58% of labeled failures with 60%-84% precision across three fundamentally different state estimation frameworks. Our results show that even a simple frequency-domain analysis of a state estimator's output can serve as a lightweight introspective tool to complement existing robustness techniques in real-world robotic deployments, and opens promising avenues for future investigation.
Chinese Translation
可靠的机载状态估计对于安全的机器人操作至关重要，但未建模的干扰，如传感器混叠或分布外噪声，仍然会导致估计器性能下降或完全失效。虽然许多方法旨在提高估计器的鲁棒性，但只有少数提供了自我评估估计质量的机制。现有的不确定性度量，如协方差，依赖于理想化假设，往往表现出过度自信，而更近期的数据驱动方法通常与其训练数据分布密切相关。我们提出了一种传感器无关的自我评估方法，通过分析近期速度估计的频域功率分布来评估估计器的健康状况。该方法使用来自一架运行视觉惯性、激光雷达惯性和雷达惯性里程计的空中机器人户外飞行数据进行评估。数据集中包含多个估计器故障，使得可以分析多个频域指标，如信号功率、频谱带宽和熵。我们观察到健康估计与降级估计之间存在一致的频谱功率差异，从而能够以60%-84%的精度检测到51%-58%的标记故障，涵盖三种根本不同的状态估计框架。我们的结果表明，即使是对状态估计器输出的简单频域分析，也可以作为一种轻量级的自我评估工具，以补充现有的鲁棒性技术在实际机器人部署中的应用，并为未来的研究开辟了有希望的方向。
cs.RO / 16 / 2608.10651
OAA: Three Phases of Vocal Guidance in Human-Drone Teleoperation
OAA：人机无人机遥控中的语音引导三阶段
Abstract
Voice-guided teleoperation requires systems that adapt to the evolving dynamics of human guidance. Yet most voice-controlled robot systems treat spoken commands as a stationary stream, ignoring how the guide's communicative behavior changes as the task progresses. Using motion capture and speech data from two experimental configurations, humanhuman guidance (finger pointing, N =10 dyads) and humandrone teleoperation (gamepad control, N =29 dyads), we show that spontaneous vocal guidance consistently organizes into three kinematically and linguistically distinct phases: Orientation, Approach, and Adjustment. These phases are identified automatically via change point detection on 3D trajectory signals, and validated statistically (Kruskal-Wallis, p<.001). Three lexical families replicate across configurations: rotation vocabulary marks Orientation, translation vocabulary is scarce there, and attenuators accumulate toward Adjustment. Together with inter-utterance silence, these cues mark the Orientation boundary that speech rate alone leaves unmarked. The same three-phase structure emerges in both configurations despite radically different motor interfaces, suggesting it is an intrinsic property of human spatial guidance rather than an artifact of the experimental setup. We discuss implications for OAA-aware adaptive control in voice-guided teleoperation.
Chinese Translation
语音引导的遥控操作需要能够适应人类引导动态变化的系统。然而，大多数语音控制的机器人系统将口头指令视为静态流，忽视了随着任务进展引导者的交流行为如何变化。通过使用来自两个实验配置的运动捕捉和语音数据——人际引导（手指指向，N = 10 对）和人机无人机遥控（游戏手柄控制，N = 29 对），我们展示了自发的语音引导始终组织成三个在运动学和语言学上截然不同的阶段：定向（Orientation）、接近（Approach）和调整（Adjustment）。这些阶段通过对三维轨迹信号的变更点检测自动识别，并通过统计方法验证（Kruskal-Wallis, p < .001）。三个词汇家族在不同配置中重复出现：旋转词汇标记定向，平移词汇在此处稀缺，而减弱词汇则向调整积累。连同语句间的沉默，这些线索标记了定向边界，而仅靠语速则无法标记。尽管运动接口截然不同，但在两种配置中都出现了相同的三阶段结构，这表明它是人类空间引导的内在特性，而非实验设置的伪影。我们讨论了在语音引导遥控中实现OAA感知自适应控制的意义。
cs.RO / 17 / 2608.10656
Robust Sliding Mode and Admittance Control of Underactuated Aerial Manipulators for Contact-Based Inspection
基于接触的检查下的欠驱动空中操控器的鲁棒滑模与顺应控制
Abstract
Contact-based industrial inspection requires aerial platforms to maintain stable interaction while rejecting disturbances. Underactuated aerial manipulators present control challenges due to the dynamic coupling between vehicle attitude and force generation. This paper proposes a robust control framework for an underactuated hexarotor equipped with a 1-DoF manipulator to perform sustained contact inspection. The architecture integrates integral-augmented Sliding Mode Control (SMC) for trajectory tracking with an admittance control law for force regulation. The contact force is mapped to a feedforward attitude term, while the 1-DoF arm actively compensates for the tilt to maintain surface alignment. Software-in-the-loop simulations demonstrate that the SMC-based approach achieves superior tracking and coupling rejection compared to traditional PID. Furthermore, the interaction strategy achieved precise force regulation with an RMSE of 0.12N and was able to stably exert up to 20N force, confirming the system's efficacy for stable, reliable contact-based inspection.
Chinese Translation
基于接触的工业检查要求空中平台在拒绝干扰的同时保持稳定的交互。由于飞行器姿态与力生成之间的动态耦合，欠驱动空中操控器面临控制挑战。本文提出了一种鲁棒控制框架，用于配备1自由度操控器的欠驱动六旋翼无人机，以执行持续的接触检查。该架构将积分增强滑模控制（Sliding Mode Control, SMC）与用于力调节的顺应控制律相结合，实现轨迹跟踪。接触力被映射为前馈姿态项，而1自由度臂则主动补偿倾斜，以保持与表面的对齐。软件在环仿真表明，与传统PID控制相比，基于SMC的方法在跟踪和耦合拒绝方面表现出更优越的性能。此外，该交互策略实现了精确的力调节，均方根误差（RMSE）为0.12N，并能够稳定施加高达20N的力，确认了该系统在稳定、可靠的基于接触的检查中的有效性。
cs.RO / 18 / 2608.10718
TCAM for Autonomous Deformable Manipulation: The RMC2 Champion System for WBCD 2026 Track 4
用于自主可变形操作的TCAM：WBCD 2026赛道4的RMC2冠军系统
Abstract
This technical report describes the RMC2 Team's champion solution for the WBCD 2026 Track 4: Deformable Manipulation Challenge. The task requires a robot to pick a single T-shirt from a stack, load it onto a printing pallet, align the collar with a target area, and smooth the printing region, a sequence that involves single-layer separation, deformable transport, precise placement, and contact-rich surface adjustment. The competition strongly incentivizes fully autonomous execution, motivating the development of an autonomous solution. We built a fully autonomous system around the TCAM (TermiBrain Causal Action Model) framework, with the design principle that hardware, perception, data, and learning should jointly reduce the physical interaction complexity the policy must handle. A custom 3D-printed gripper designed for single-layer fabric separation improves picking reliability on a dual-arm ARX X5 platform. A wrist-centric four-camera setup pairs upper fisheye cameras for task-level context with lower RGB cameras for close-range gripper-cloth contact observation. We combine portable UMI-style demonstrations with real-robot demonstrations collected on the deployable platform to provide both broad manipulation priors and deployment-specific dynamics. TCAM ties these components into a closed loop: each trajectory is analyzed to identify the physical factors contributing to its outcome, driving targeted data recollection and policy fine-tuning. The policy outputs 30-step end-effector delta-pose action chunks from a multi-view VLA backbone. In the final competition, our system loaded 25 T-shirts at an average of approximately 23 seconds per attempt, with 22 achieving the required surface smoothness, securing first place in Track 4.
Chinese Translation
本技术报告描述了RMC2团队在WBCD 2026赛道4：可变形操作挑战中的冠军解决方案。该任务要求机器人从一堆T恤中挑选一件，将其放置到打印托盘上，调整领口与目标区域对齐，并平滑打印区域，这一过程涉及单层分离、可变形运输、精确放置和丰富接触表面调整。比赛强烈鼓励完全自主执行，激励了自主解决方案的开发。我们围绕TCAM（TermiBrain因果行动模型）框架构建了一个完全自主的系统，设计原则是硬件、感知、数据和学习应共同减少政策必须处理的物理交互复杂性。为单层织物分离设计的定制3D打印夹具提高了在双臂ARX X5平台上的抓取可靠性。以手腕为中心的四摄像头设置将上方鱼眼摄像头与下方RGB摄像头配对，以获取任务级上下文和近距离夹具与布料接触的观察。我们结合便携式UMI风格的演示与在可部署平台上收集的真实机器人演示，以提供广泛的操作先验和特定于部署的动态。TCAM将这些组件结合成一个闭环：每个轨迹都被分析以识别影响其结果的物理因素，从而推动有针对性的数据收集和政策微调。该政策从多视角VLA骨干网络输出30步末端执行器增量位姿动作块。在最终比赛中，我们的系统平均每次尝试加载25件T恤，时间约为23秒，其中22件达到了所需的表面平滑度，获得了赛道4的第一名。
cs.RO / 19 / 2608.10756
Embodied Multimodal Grounding for Open-Vocabulary Mobile Manipulation via Semantic 3D Gaussian Splatting
通过语义3D高斯点云实现开放词汇移动操作的具身多模态基础
Abstract
Embodied mobile manipulation requires language, visual observations, three-dimensional scene structure, and action feasibility to be aligned before execution. We study open-vocabulary target grounding with few-shot manipulation in local household workspaces and present an embodied multimodal grounding framework that integrates active multi-view Semantic 3D Gaussian Splatting (Semantic-3DGS), reachability-aware base positioning, and a diffusion-based vision-language-action policy. A task-driven local Semantic-3DGS serves as a shared interface across active sensing, language-conditioned 3D localization, obstacle-aware scene reasoning, base preparation, and semantic conditioning of the action model. To preserve pretrained action priors, the 3D semantic cues are injected only into the late action-expert blocks. In expanded 50-trial real-robot evaluations against representative vision-language-action (VLA) approaches, the full system achieves 60% long-horizon success compared with 40% for PointVLA and 28% for DexVLA, and reaches 74% success in heavily cluttered manipulation compared with 52% for the single-view variant and 46% for PointVLA. It also maintains 75% success under a 75 cm height shift and eliminates photo-induced false grasps. These results indicate that explicit, refreshable 3D semantic grounding can improve robustness under clutter, occlusion, viewpoint variation, and embodiment constraints.
Chinese Translation
具身移动操作要求在执行之前对语言、视觉观察、三维场景结构和动作可行性进行对齐。我们研究了在本地家庭工作空间中进行少量操作的开放词汇目标基础，并提出了一种具身多模态基础框架，该框架集成了主动多视角语义3D高斯点云（Semantic 3D Gaussian Splatting，Semantic-3DGS）、考虑可达性的基础定位和基于扩散的视觉-语言-动作策略。任务驱动的本地Semantic-3DGS作为主动感知、语言条件下的3D定位、障碍物感知场景推理、基础准备和动作模型的语义条件之间的共享接口。为了保留预训练的动作先验，3D语义线索仅在后期动作专家模块中注入。在与代表性的视觉-语言-动作（Vision-Language-Action，VLA）方法进行的扩展50次真实机器人评估中，完整系统在长时间任务中的成功率达到60%，而PointVLA为40%，DexVLA为28%；在高度杂乱的操作中成功率达到74%，而单视角变体为52%，PointVLA为46%。它在75厘米高度偏移下保持75%的成功率，并消除了光照引起的错误抓取。这些结果表明，显式的、可刷新3D语义基础可以提高在杂乱、遮挡、视角变化和具身约束下的鲁棒性。
cs.RO / 20 / 2608.10780
JEPA-WAM: Stage-Level Joint-Embedding Prediction for World-Action Models in Robot Manipulation
JEPA-WAM：机器人操作中世界行动模型的阶段级联合嵌入预测
Abstract
Generalist robot policies aim to map multimodal observations and linguistic task instructions to actions across diverse tasks. However, existing methods typically represent the future as a fixed, short video-action chunk. This short-term future captures local scene evolution for action execution, but it does not explicitly describe the stage-level future that specifies how a task should progress from its current stage to the next. We therefore distinguish two complementary futures for robot manipulation: a short-term physical future to capture local scene evolution and a stage-level semantic future to represent task progress. We introduce JEPA-WAM, which augments a Motus-based World Action Model (WAM) with Stage-JEPA, a goal-conditioned Joint-Embedding Predictive Architecture (JEPA) predictor. Given the current observation and task instruction, Stage-JEPA uses a frozen V-JEPA2 encoder to extract the current-state representation and predicts the latent target of the next inferred stage. Across 50 RoboTwin 2.0 tasks in clean and randomized environments, JEPA-WAM achieves 90.25% overall success and reduces the mean number of execution steps in successful rollouts by 5.97% relative to the strongest baseline.
Chinese Translation
通用机器人策略旨在将多模态观察和语言任务指令映射到多样化任务中的动作。然而，现有方法通常将未来表示为固定的短视频动作片段。这种短期未来捕捉了局部场景演变以执行动作，但并未明确描述阶段级未来，即任务如何从当前阶段进展到下一个阶段。因此，我们为机器人操作区分了两种互补的未来：短期物理未来用于捕捉局部场景演变，以及阶段级语义未来用于表示任务进展。我们提出了JEPA-WAM，它通过Stage-JEPA，一个目标条件的联合嵌入预测架构（Joint-Embedding Predictive Architecture, JEPA）预测器，增强了基于Motus的世界行动模型（World Action Model, WAM）。在给定当前观察和任务指令的情况下，Stage-JEPA使用冻结的V-JEPA2编码器提取当前状态表示，并预测下一个推断阶段的潜在目标。在干净和随机化环境中的50个RoboTwin 2.0任务中，JEPA-WAM实现了90.25%的整体成功率，并相对于最强基线将成功执行的平均步骤数减少了5.97%。
cs.RO / 21 / 2608.10791
Dual Stress: Runtime Safety Monitoring for Safety-Constrained MPC Navigation
双重压力：安全约束下的模型预测控制导航的运行时安全监测
Abstract
Runtime hazard monitors for autonomous naviga- tion are conventionally built from geometric quantities: predicted clearance, time to collision, and required deceleration. A model-predictive controller that enforces safety through explicit con- straints computes, as a by-product of every control step, a second information channel that such monitors ignore: the Karush-Kuhn-Tucker multipliers of its constrained optimization, which measure the marginal control effort spent to maintain safety against each obstacle. This paper evaluates whether a horizon-weighted sum of those multipliers, a dual stress signal, provides a hazard monitor complementary to the geometric warnings the same state already supports. We compare it against a battery of fifteen geometric detectors tuned to a matched false-alarm budget, on preregistered held-out crossing scenarios driven through a physics simulator. The stress alarm actionably flags 4.7 times as many collisions missed by the entire geometric battery as the geometric battery flags in return (85 versus 18); combined, the two channels warn of three quarters of the collisions for which braking remained feasible, against under half for the geometric battery alone.
Chinese Translation
自主导航的运行时危险监测通常基于几何量构建：预测间隙、碰撞时间和所需减速。通过显式约束强制安全的模型预测控制器在每个控制步骤的副产品中计算出一个额外的信息通道，而这些监测器通常会忽略：其约束优化的Karush-Kuhn-Tucker乘子，这些乘子衡量了为维持对每个障碍物的安全所花费的边际控制努力。本文评估了这些乘子的加权和，即双重压力信号，是否能提供一种补充于同一状态下几何警告的危险监测。我们将其与调谐至匹配虚警预算的十五个几何探测器进行比较，实验在预注册的保留交叉场景中通过物理模拟器进行。压力警报有效地标记了4.7倍于整个几何探测器未能检测到的碰撞（85次对比18次）；结合这两个通道，警告了四分之三的碰撞，其中刹车仍然可行，而仅几何探测器则不足一半。
cs.RO / 22 / 2608.10817
AECNav: Active Evidence Consolidation for Efficient Zero-Shot Open-Vocabulary Object Navigation
AECNav：高效零-shot开放词汇物体导航的主动证据整合
Abstract
Zero-shot object-goal navigation (ZSON) in open-vocabulary scenarios is challenging, as it requires a robot to locate an arbitrarily specified object in an unseen environment without task-specific training. Currently, the task still suffers from high latency and limited accuracy due to redundant perception pipelines and insufficient evidence for reliable target confirmation. In this letter, we reframe ZSON as an evidence-driven perception-to-decision problem and present AECNav, a training-free pipeline built on three components: i) Evidence-gated perception, which utilizes a shared encoding across all reasoning stages to establish a unified semantic basis and eliminate redundant computations; ii) Evidence consolidation, which aggregates detections into cluster-level log-odds beliefs. This explicitly separates genuine target support from the false confidence of visually similar distractors, while treating the absence of expected detections as negative evidence; and iii) Active evidence acquisition, which sustains productive exploration under weak semantic cues by selecting frontiers that maximize information gain at minimal traversal cost. As a result, AECNav significantly outperforms previous methods and achieves state-of-the-art success rates of 84.7%, 57.3%, and 51.3% on HM3D-v2, HM3D-OVON, and MP3D, respectively, with substantially lower inference overhead, and attains 95% success across 40 trials on a physical quadruped robot at roughly 5Hz. Code will be made publicly available upon acceptance.
Chinese Translation
在开放词汇场景中，零-shot物体目标导航（ZSON）具有挑战性，因为它要求机器人在未见环境中定位任意指定的物体，而无需特定任务的训练。目前，由于冗余的感知管道和缺乏可靠目标确认的充分证据，该任务仍然面临高延迟和有限的准确性。在本文中，我们将ZSON重新框定为一个以证据驱动的感知到决策问题，并提出了AECNav，这是一个基于三个组件构建的无训练管道：i) 证据门控感知，利用跨所有推理阶段的共享编码建立统一的语义基础，消除冗余计算；ii) 证据整合，将检测结果聚合为集群级别的对数几率信念。这明确区分了真实目标支持与视觉上相似的干扰物的虚假信心，同时将预期检测缺失视为负证据；iii) 主动证据获取，通过选择最大化信息增益且最小化遍历成本的边界，维持在弱语义线索下的有效探索。因此，AECNav显著优于以前的方法，并在HM3D-v2、HM3D-OVON和MP3D上分别达到了84.7%、57.3%和51.3%的最新成功率，推理开销显著降低，并在物理四足机器人上以约5Hz的频率在40次试验中达到了95%的成功率。代码将在接受后公开发布。
cs.RO / 23 / 2608.10824
Neural Introspection Gating for Adaptive KV-Cache Reuse in Vision-Language-Action Models
用于视觉-语言-动作模型的自适应KV缓存重用的神经内省门控
Abstract
Vision-Language-Action(VLA) models map camera images and language instructions directly to motor commands through a single autoregressive transformer. In real-time control, they still spend substantial compute recomputing key-value(KV) representations for visual tokens that barely change across neighboring frames. Recent work such as VLA-Cache reduces that cost by reusing KV states for visually static patches, but its policy relies only on observation-space heuristics and does not account for the model's own uncertainty. We propose Gated VLA-Cache, a lightweight, training-free extension that augments visual-similarity caching with neural introspection. The method monitors the logit margin between the top two predicted action tokens, a zero-cost confidence signal available during decoding. When the margin drops below a threshold, the cache is invalidated and a full recompute is triggered. Evaluated on four LIBERO benchmark suites with both OpenVLA and OpenVLA-OFT, Gated VLA-Cache improves reliability when blind caching hurts. On LIBERO-Goal and LIBERO-Long, it recovers over 100% of the lost accuracy while retaining 80% of the compute savings.
Chinese Translation
视觉-语言-动作（VLA）模型通过单个自回归变换器将相机图像和语言指令直接映射到运动指令。在实时控制中，它们仍然花费大量计算资源重新计算几乎在相邻帧中不变的视觉标记的键值（KV）表示。最近的研究如VLA-Cache通过重用视觉静态区域的KV状态来降低这一成本，但其策略仅依赖于观察空间的启发式方法，并未考虑模型自身的不确定性。我们提出了门控VLA-Cache，这是一种轻量级、无训练的扩展，增强了视觉相似性缓存与神经内省的结合。该方法监控前两个预测动作标记之间的logit边际，这是在解码过程中可用的零成本置信信号。当边际低于阈值时，缓存将失效并触发完全重新计算。在使用OpenVLA和OpenVLA-OFT的四个LIBERO基准测试套件上进行评估时，门控VLA-Cache在盲目缓存造成损害时提高了可靠性。在LIBERO-Goal和LIBERO-Long上，它恢复了超过100%的丢失准确率，同时保留了80%的计算节省。
cs.RO / 24 / 2608.10847
Enabling Scalable Kinesthetic Teaching via Observer-based Hand-guiding with Active Support
通过基于观察者的手动引导与主动支持实现可扩展的动觉教学
Abstract
Kinesthetic teaching through robot hand-guiding provides a natural interface for collecting demonstrations in imitation learning and programming-by-demonstration. However, extended sessions cause operator fatigue, reducing demonstration quality and limiting scalability. Current industrial hand-guiding approaches typically provide no active assistance, and alternatives require costly wrist-mounted force-torque sensors or rely on learned motion priors unavailable for new tasks. We propose RHOAS, a hand-guiding scheme that actively supports operator-intended motions using model-based force estimation without additional hardware. Our approach considers robot hand-guiding as an actively controlled interaction by the human operator, rather than an interaction with a passive environment. Standard methods used for hand-guiding typically rely on general passivity-based compliant control architectures that unnecessarily increase operator effort and limit the range of demonstrable motions without providing the intended stability guarantees in active interaction. Instead, our design utilizes model-based external torque estimation, internal joint torque sensing, and redundant robot kinematics to actively support human physical input within the human interaction frequency bandwidth. We address practical challenges of relying on observer-based force estimation, including suppression of unmodeled joint elastic dynamic effects and measurement noise in the feedback path, reduced estimate accuracy close to kinematic singularities, and static gravity compensation errors. In a user study with 16 participants on a KUKA LWR iiwa we demonstrate statistically significant reductions in physical effort, improved maneuverability for both precise and agile tasks, and clear user preference.
Chinese Translation
通过机器人手动引导进行动觉教学为模仿学习和示范编程提供了一种自然的演示收集接口。然而，长时间的操作会导致操作者疲劳，从而降低演示质量并限制可扩展性。目前的工业手动引导方法通常不提供主动支持，而替代方案则需要昂贵的腕部力矩传感器或依赖于不可用于新任务的学习运动先验。我们提出了RHOAS，一种手动引导方案，利用基于模型的力估计在不增加额外硬件的情况下主动支持操作者意图的动作。我们的方法将机器人手动引导视为人类操作者主动控制的交互，而不是与被动环境的交互。用于手动引导的标准方法通常依赖于基于一般被动性的顺应控制架构，这不必要地增加了操作者的努力，并限制了可演示动作的范围，而未能在主动交互中提供预期的稳定性保证。相反，我们的设计利用基于模型的外部扭矩估计、内部关节扭矩传感和冗余机器人运动学，在人类交互频率带宽内主动支持人类的物理输入。我们解决了依赖于基于观察者的力估计的实际挑战，包括抑制未建模的关节弹性动态效应和反馈路径中的测量噪声、在运动学奇异点附近降低估计精度以及静态重力补偿误差。在一项针对16名参与者的用户研究中，我们在KUKA LWR iiwa上展示了显著降低的物理努力、对精确和灵活任务的改善机动性以及明显的用户偏好。
cs.RO / 25 / 2608.10860
Flex-$\pi$: A Multi-Stream World-Action Model with Compute Flexibility
Flex-$ ext{π}$：一种具有计算灵活性的多流世界行为模型
Abstract
World-action models (WAMs) predict the future to act better, but nearly all of them predict only RGB latents, trained purely for pixel reconstruction, with no explicit signal for the 3D geometry or object semantics manipulation needs. We find a surprising free lunch: the same frozen video-generation VAE that encodes RGB also encodes 3D pointmaps almost losslessly, with no pointmap-specific training at all. This lets us supervise Flex-$\pi$, a 6B-parameter WAM, on 3D geometry and object-centric DINO semantics alongside RGB, at no cost in new sensors, new pre-training, or inference latency. Every visual signal is projected into this shared latent space and denoised jointly with actions inside a Mixture-of-Transformers backbone; per-stream dropout with cross-modality forcing then lets a single trained checkpoint run on any subset of these streams, from a fast action-only mode to full joint generation. The result is a policy that is exceptionally demonstration-efficient and generalizes well, beating the strongest baselines by up to 2-7$\times$ on dexterous, precise, real-world bimanual manipulation tasks both in and out of distribution, all while running faster than $\pi_{0.5}$. Our project website: https://flex-pi.github.io/
Chinese Translation
世界行为模型（WAMs）通过预测未来以更好地行动，但几乎所有模型仅预测RGB潜变量，完全为像素重建而训练，未显式考虑3D几何或物体语义操作的需求。我们发现了一个意外的“免费午餐”：同一个冻结的视频生成变分自编码器（VAE）不仅编码RGB，还几乎无损地编码3D点图，而无需任何点图特定的训练。这使我们能够在RGB的基础上，监督Flex-$ ext{π}$，一个拥有60亿参数的WAM，进行3D几何和以物体为中心的DINO语义学习，而无需新的传感器、新的预训练或推理延迟。每个视觉信号都被投影到这个共享的潜在空间中，并与Mixture-of-Transformers骨干网络内的动作共同去噪；通过跨模态强制的每流丢弃机制，单个训练检查点可以在这些流的任意子集上运行，从快速的仅动作模式到完整的联合生成。结果是一个异常高效的演示策略，并且具有良好的泛化能力，在灵巧、精确的真实世界双手操作任务中，无论是在分布内外，均比最强基线高出2-7倍，同时运行速度快于$ ext{π}_{0.5}$。我们的项目网站：https://flex-pi.github.io/
cs.RO / 26 / 2608.10987
Aerial Layouting: Design and Control of a Compliant and Actuated End-Effector for Precise In-flight Marking on Ceilings
空中布局：一种柔性和驱动末端执行器的设计与控制，用于在天花板上进行精确的飞行标记
Abstract
Aerial robots have demonstrated impressive feats of precise control, such as dynamic flight through openings or highly complex choreographies. Despite the accuracy needed for these tasks, there are problems that require levels of precision that are challenging to achieve today. One such problem is aerial interaction. Advances in aerial robot design and control have made such contact-based tasks possible and opened up research into challenging real-world tasks, including contact-based inspection. However, while centimetre accuracy is sufficient and achievable for inspection tasks, the positioning accuracy needed for other problems, such as layouting on construction sites or general push-and-slide tasks, is millimetres. To achieve such a high precision, we propose a new aerial system composed of an aerial vehicle equipped with a novel "smart" end-effector leveraging a stability-optimized Gough-Stewart mechanism. We present its design process and features incorporating the principles of compliance, multiple contact points, actuation, and self-containment. In experiments, we verify that the design choices made for our novel end-effector are necessary to obtain the desired positioning precision. Furthermore, we demonstrate that we can reliably mark lines on ceilings with millimetre accuracy without the need for precise modeling or sophisticated control of the aerial robot.
Chinese Translation
空中机器人展示了精确控制的令人印象深刻的成就，例如通过开口的动态飞行或高度复杂的编排。尽管这些任务需要高精度，但仍然存在一些问题需要达到今天难以实现的精度水平。其中一个问题是空中交互。空中机器人设计和控制的进步使得这种基于接触的任务成为可能，并开启了对具有挑战性的现实世界任务的研究，包括基于接触的检查。然而，尽管厘米级的精度对于检查任务来说是足够且可实现的，但其他问题所需的定位精度，例如在建筑工地上的布局或一般的推拉任务，则需要毫米级的精度。为了实现如此高的精度，我们提出了一种新的空中系统，由一架配备新型“智能”末端执行器的空中飞行器组成，该执行器利用了稳定性优化的高斯-斯图尔特机制。我们展示了其设计过程和特性，结合了柔性、多接触点、驱动和自我封闭的原则。在实验中，我们验证了为我们的新型末端执行器所做的设计选择是获得所需定位精度所必需的。此外，我们证明可以在天花板上以毫米级的精度可靠地标记线条，而无需对空中机器人进行精确建模或复杂控制。
cs.RO / 27 / 2608.10997
Seeing above the waves: A modular sensing framework for data acquisition at sea
超越波浪的视野：海洋数据采集的模块化传感框架
Abstract
Advancing autonomy for surface vessels requires systematic evaluation of their sensing and perception subsystems. Yet, maritime environments impose unique challenges: sensor installation is constrained by vessel layout, environmental conditions such as fog or sea clutter are difficult to reproduce, and long-duration missions complicate data collection. This work addresses the question: How can we design a modular and reproducible sensor platform for maritime autonomy? We present a comprehensive design blueprint that incorporates diverse modalities - RADAR, LiDAR, IMU, GNSS, AIS, RGB and LWIR cameras, and weather sensors - to enhance environmental awareness and vessel proprioception. Supported by a dedicated ROS2-based software framework for data management, our modular platform enables long-term data collection, hardware-in-the-loop testing, and integration with existing sensors and algorithms. By unifying hardware design and data capture methodology, the platform enhances reproducibility and comparability across vessels and research projects. The proposed framework bridges engineering implementation and research methodology, providing the foundation for standardized, verifiable datasets essential to advancing situational awareness and autonomous maritime navigation.
Chinese Translation
推动水面船只的自主性发展需要对其传感和感知子系统进行系统评估。然而，海洋环境带来了独特的挑战：传感器的安装受到船只布局的限制，雾霾或海洋杂波等环境条件难以重现，且长时间任务使数据收集变得复杂。本研究探讨了如何设计一个模块化且可重复的海洋自主传感平台。我们提出了一套全面的设计蓝图，结合了多种传感器模式——雷达（RADAR）、激光雷达（LiDAR）、惯性测量单元（IMU）、全球导航卫星系统（GNSS）、自动识别系统（AIS）、RGB和长波红外（LWIR）摄像头以及气象传感器——以增强环境感知和船只自我感知。我们的模块化平台由一个基于ROS2的数据管理软件框架支持，能够实现长期数据收集、硬件在环测试以及与现有传感器和算法的集成。通过统一硬件设计和数据采集方法，该平台提高了不同船只和研究项目之间的可重复性和可比性。所提出的框架架起了工程实施与研究方法论之间的桥梁，为标准化、可验证的数据集奠定了基础，这对于提升情境意识和自主海洋导航至关重要。
cs.RO / 28 / 2608.11063
Deployment Is Not Destiny: Robot Recomposition in the Field with Unseen Software, Hardware, and Compute Payloads
部署不是命运：在现场对未见软件、硬件和计算负载进行机器人重组
Abstract
The tight coupling of subsystems in most robots, though a natural consequence of their complexity, leads to monolithic designs that are time-consuming and difficult to adapt after initial deployment. To address this challenge, we present a framework and supporting abstractions for recomposition during runtime that enable robots to quickly integrate previously unseen modular software, hardware, and compute payloads. Our approach allows non-expert users to quickly add new capabilities in the field through a true plug-and-play process. Crucially, new resources are not only immediately available to a host robot but are also shared with distributed peers, enabling compute-constrained systems to access powerful new remote capabilities. Our framework reduces reconfiguration time to a matter of minutes with no developer intervention, in stark contrast to the hours of expert effort often required for traditional manual integration. We demonstrate our method in two disaster response scenarios, including radioactive source localization at an operational nuclear reactor facility and a thermal-guided search for people in dark, difficult-to-reach spaces. These demonstrations show how in-field recomposition provides timely, flexible, and accessible adaptation to dynamic requirements, representing a critical step toward creating robots that can quickly evolve alongside the tasks, technologies, and environments they support.
Chinese Translation
大多数机器人子系统之间的紧密耦合，虽然是其复杂性的自然结果，却导致了单一设计，这种设计在初始部署后难以适应且耗时。为了解决这一挑战，我们提出了一种框架及其支持的抽象，旨在运行时进行重组，使机器人能够快速集成之前未见的模块化软件、硬件和计算负载。我们的方法允许非专业用户通过真正的即插即用过程快速在现场添加新功能。重要的是，新资源不仅可以立即被主机器人使用，还可以与分布式同伴共享，使计算受限的系统能够访问强大的新远程能力。我们的框架将重新配置时间缩短至几分钟，无需开发者干预，这与传统手动集成通常需要的数小时专家努力形成鲜明对比。我们在两个灾难响应场景中演示了我们的方法，包括在一个运行中的核反应堆设施中进行放射源定位，以及在黑暗、难以到达的空间中进行热导向搜索。这些演示展示了现场重组如何为动态需求提供及时、灵活和可访问的适应，代表着朝着创建能够快速与其支持的任务、技术和环境共同演变的机器人的关键一步。
cs.RO / 29 / 2608.11174
VIScore: Diagnosing Planning-Relevant Quality in Latent World Models
VIScore：诊断潜在世界模型中的规划相关质量
Abstract
Regulating the latent space to an isotropic Gaussian distribution provides a stable and information-maximized landscape for world model planning. However, the latent space property and successful planning remain disconnected. We first study this by comparing SIGReg and VISReg, two regularization loss functions with the same distribution target but different properties. Compared with SIGReg, VISReg has more flexibility in controlling the weights of center, scale, and shape regularization, and a larger batch size brings a finer distribution approximation. We find that the former, despite being beneficial in self-supervised learning (SSL), does not help the planning, whereas the latter improves the planning success on out-of-domain (OOD) datasets. This motivates a deep understanding of the factors that correlate with the success rate. Unlike the previous metrics focusing on the encoded latent only, we propose the Veracity-Influence-Sobriety score (VIScore), a metric that quantifies the reachability and capacity of a predictor given the encoded feature, and the hallucination of the searching-based planner. Compared with straightness, physical-state probing, and empowerment, we show that, with the measurement covering encoder, predictor, and planner, VIScore explains the success rate better than the others, as reflected by a strong Spearman correlation. Specifically, VIScore consistently achieves a Spearman correlation over 0.75 on both seen and unseen models and datasets on the cross-task success rate pool. Moreover, VIScore is the only metric that has a calibration error below the constant fit across all testing scenarios, showcasing the importance of these three aspects in planning success. We hope this metric can help future studies on world model design and diagnosis.
Chinese Translation
将潜在空间调节为各向同性的高斯分布为世界模型规划提供了一个稳定且信息最大化的环境。然而，潜在空间的特性与成功的规划之间仍然存在脱节。我们首先通过比较SIGReg和VISReg这两种具有相同分布目标但特性不同的正则化损失函数来研究这一问题。与SIGReg相比，VISReg在控制中心、尺度和形状正则化的权重方面具有更大的灵活性，而更大的批量大小则带来了更精细的分布近似。我们发现，前者尽管在自监督学习（SSL）中有益，但并未帮助规划，而后者则在域外（OOD）数据集上提高了规划成功率。这促使我们深入理解与成功率相关的因素。与之前仅关注编码潜在的指标不同，我们提出了真实度-影响力-清晰度评分（VIScore），该指标量化了给定编码特征的预测器的可达性和能力，以及基于搜索的规划器的幻觉。与直线性、物理状态探测和赋能相比，我们展示了VIScore通过覆盖编码器、预测器和规划器的测量，能够比其他指标更好地解释成功率，这在强斯皮尔曼相关性中得到了体现。具体而言，VIScore在跨任务成功率池的已见和未见模型及数据集上始终实现了超过0.75的斯皮尔曼相关性。此外，VIScore是唯一一个在所有测试场景中校准误差低于常数拟合的指标，展示了这三个方面在规划成功中的重要性。我们希望这一指标能够帮助未来在世界模型设计和诊断方面的研究。
cs.RO / 30 / 2608.11175
Risk-Aware Kinodynamic Motion Planning Under Uncertainty For Safe Navigation on Planetary Environments
在不确定性下的风险感知动力学运动规划：用于行星环境的安全导航
Abstract
For autonomous space exploration, robotic agents need to perform motion planning in which environmental interactions may be unknown. Learning these interactions, such as terrain mechanics for wheeled robots, can introduce uncertainties that lead to risky motion plans and potentially hazardous operations or mission failures. Moreover, uncertainties induced by perception-based systems can exacerbate the problem of safe motion planning. In this letter, we address the problem of performing cost-optimal kinodynamic motion planning with risk awareness. We approach this in two steps. First, a sampling-based planner (AO-RRT) generates a dynamically feasible, risk-aware, and asymptotically cost-optimal trajectory. Second, we formulate motion planning as a nonlinear optimization problem and solve it using sequential convex programming (SCP), using the AO-RRT trajectory as an initial solution. By quantifying risk using conditional value-at-risk (CVaR), we demonstrate a reduction in risk by over $\sim$97\% across trajectories in simulation and hardware experiments.
Chinese Translation
对于自主空间探索，机器人代理需要进行运动规划，而环境交互可能是未知的。学习这些交互，例如轮式机器人的地形力学，可能会引入不确定性，从而导致风险较高的运动计划以及潜在的危险操作或任务失败。此外，由感知系统引起的不确定性可能会加剧安全运动规划的问题。在本论文中，我们解决了具有风险意识的成本最优动力学运动规划问题。我们采用两步法进行处理。首先，基于采样的规划器（AO-RRT）生成一个动态可行、具有风险意识且渐近成本最优的轨迹。其次，我们将运动规划形式化为一个非线性优化问题，并使用顺序凸编程（SCP）进行求解，以AO-RRT轨迹作为初始解。通过使用条件风险价值（CVaR）量化风险，我们在模拟和硬件实验中展示了轨迹风险降低超过$ ext{~}97\%$。
cs.RO / 31 / 2608.11204
Surgical WAM: A World-Action Model for Data-Efficient Surgical Robot Learning
外科WAM：一种数据高效的外科机器人学习世界动作模型
Abstract
Learning reliable surgical manipulation policies is bottlenecked by the scarcity of action-labeled demonstrations: teleoperated surgical robot (e.g., dVRK) trajectories with synchronized kinematics are costly to collect, while surgical tasks demand precise contact handling, long-horizon reasoning, and bimanual coordination. Endoscopic video is comparatively inexpensive and abundant relative to synchronized video--kinematics trajectories, and a natural way to exploit it is to learn world models of surgical scenes. However, existing surgical world models use video primarily for simulation or policy evaluation, and rarely translate the learned dynamics into closed-loop control. This gap raises our central question: under a fixed budget of action-labeled demonstrations, does action-free video pretraining improve closed-loop surgical manipulation? To answer it, we introduce the Surgical World-Action Model (Surgical WAM), a unified generative model built on Cosmos Policy that jointly predicts future endoscopic observations and executable surgical robot action chunks. Surgical WAM first learns surgical visual dynamics from action-free video and is then fine-tuned on the fixed action-labeled budget; at deployment, it acts as a closed-loop, receding-horizon controller that executes a short prefix of each predicted action chunk and replans from the resulting observation. On a suite of four simulated surgical manipulation tasks, video pretraining improves the average success rate from 63.5% to 77.8%, including an absolute gain of 20 percentage points on PegTransfer, with the largest improvements on contact-rich and bimanual tasks. These results demonstrate that action-free video provides transferable visual dynamics priors for learning surgical robot control with limited action supervision, positioning data-efficient video pretraining as a practical path toward scaling up surgical robot learning.
Chinese Translation
学习可靠的外科操作策略受到动作标记示范稀缺的瓶颈：遥控外科机器人（例如，dVRK）的轨迹与同步运动学的收集成本高昂，而外科任务则需要精确的接触处理、长时间的推理和双手协调。相比于同步视频-运动学轨迹，内窥镜视频相对便宜且丰富，利用它的自然方式是学习外科场景的世界模型。然而，现有的外科世界模型主要将视频用于模拟或策略评估，鲜有将学习到的动态转化为闭环控制。这一差距引出了我们的核心问题：在固定的动作标记示范预算下，无动作视频的预训练是否能改善闭环外科操作？为了解答这一问题，我们提出了外科世界动作模型（Surgical WAM），这是一个基于Cosmos Policy的统一生成模型，能够共同预测未来的内窥镜观察和可执行的外科机器人动作片段。外科WAM首先从无动作视频中学习外科视觉动态，然后在固定的动作标记预算上进行微调；在部署时，它作为一个闭环、递减视野的控制器，执行每个预测动作片段的短前缀，并根据结果观察重新规划。在四个模拟外科操作任务的测试中，视频预训练将平均成功率从63.5%提高到77.8%，在PegTransfer任务上实现了20个百分点的绝对增益，并在接触丰富和双手任务上取得了最大的改善。这些结果表明，无动作视频为在有限的动作监督下学习外科机器人控制提供了可转移的视觉动态先验，将数据高效的视频预训练定位为扩展外科机器人学习的实用路径。
cs.CV / 1 / 2608.10057
LEGO: Leveled Language Gaussian Splatting
LEGO：分级语言高斯点云渲染
Abstract
We introduce LEGO for advanced open-vocabulary scene understanding. Beyond basic concept recognition, its core innovation lies in capturing the intrinsic semantic hierarchies within the scene, such as the "flowerpot -> bouquet -> bud -> petal" lineage. While foundation models like SAM can identify multi-granular structures in 2D, their partitions are strictly perspective-bound and lack cross-view consensus. LEGO self-adaptively re-grades volatile multi-view SAM granularities into a unified, 3D-consistent hierarchy. This provides precise supervision for the structurally coherent, multi-level segmentation of 3D scenes. By grounding these segments with CLIP embeddings, LEGO recovers open-vocabulary semantic logic across hierarchical levels. Furthermore, by incorporating spatial relationships, we elevate these segments into level-wise language scene graphs, effectively empowering Large Language Models to perform complex, context-aware spatial reasoning and precise visual grounding. Experimental results demonstrate that LEGO establishes new state-of-the-art performance across both promptable and open-vocabulary 3D segmentation benchmarks, exhibiting advanced hierarchical scene decomposition and context-aware spatial reasoning.
Chinese Translation
我们提出了LEGO，用于先进的开放词汇场景理解。超越基本的概念识别，其核心创新在于捕捉场景内在的语义层次结构，例如“花盆 -> 花束 -> 花蕾 -> 花瓣”的谱系。尽管像SAM这样的基础模型能够识别二维中的多粒度结构，但它们的划分严格受限于视角，缺乏跨视图的一致性。LEGO自适应地将不稳定的多视图SAM粒度重新分级为统一的、三维一致的层次结构。这为三维场景的结构一致的多层次分割提供了精确的监督。通过将这些分段与CLIP嵌入结合，LEGO在层次级别上恢复开放词汇的语义逻辑。此外，通过结合空间关系，我们将这些分段提升为逐层语言场景图，有效地增强了大型语言模型进行复杂的、上下文感知的空间推理和精确的视觉定位的能力。实验结果表明，LEGO在可提示和开放词汇的三维分割基准测试中建立了新的最先进的性能，展现了先进的层次场景分解和上下文感知的空间推理能力。
cs.CV / 2 / 2608.10091
Signpost Watermarking: Joint Optimization for Visual Watermark Coexistence
标志水印：视觉水印共存的联合优化
Abstract
We present a method for training imperceptible visual watermarks to coexist with other such watermarks. Recent work has shown that independently trained image watermarking models can coexist with surprisingly limited interference, enabling watermark ensembling. However, this coexistence is a serendipitous property rather than an explicit optimization objective, leaving interference uncontrolled and potentially reducing decoding robustness or visual quality. We first show empirically that the same coexistence property extends to video watermarking. We then show that both image and video watermarks can be trained with a decoder-aware objective to improve coexistence. Our results suggest a practical path to signpost watermarks that indicate the presence of independently deployed provenance watermarking systems, supporting layered provenance signaling for content authenticity and rights.
Chinese Translation
我们提出了一种训练不可察觉的视觉水印与其他水印共存的方法。近期的研究表明，独立训练的图像水印模型可以在干扰极小的情况下共存，从而实现水印的集成。然而，这种共存是一种偶然的特性，而非明确的优化目标，这使得干扰无法控制，并可能降低解码的鲁棒性或视觉质量。我们首先通过实验证明，相同的共存特性也适用于视频水印。然后，我们展示了图像和视频水印可以通过考虑解码器的目标进行训练，以改善共存性。我们的结果建议了一条实用的路径，以实现标志水印，指示独立部署的来源水印系统的存在，支持内容真实性和权利的分层来源信号传递。
cs.CV / 3 / 2608.10107
4D-WAM: 4D Consistent World Modeling for Autonomous Driving
4D-WAM：用于自动驾驶的4D一致世界建模
Abstract
Emerging World-Action Models (WAMs) have demonstrated promising performance in autonomous driving by jointly modeling future driving scene evolution and trajectory planning. However, existing WAMs are typically trained with video data, which is only 2D projections of the underlying 4D driving scene. Consequently, WAMs fail to understand and capture the structure of 4D scenes and thus generate visually plausible yet 4D inconsistent future predictions that mislead downstream planning. To alleviate this issue, we present 4D-WAM, a model that leverages geometric foundation models for training-time supervision to enable 4D consistent world modeling. Specifically, we feed WAM-predicted future frames into a geometric foundation model, and use 4D-aware responses to define a 4D consistency loss. This loss encourages the model to understand, represent, and predict physically consistent 4D scenes during training, without additional inference cost. Moreover, we identify an early-decision phenomenon in WAMs and propose a decision-oriented timestep sampling strategy that emphasizes supervision at early, high-noise stages, where driving decisions are primarily formed. By propagating 4D supervision to this critical decision-formation phase, the proposed strategy further improves trajectory planning. Extensive experiments demonstrate that 4D-WAM effectively models 4D consistent scene evolution and achieves state-of-the-art performance on challenging NAVSIM-v1 and NAVSIM-v2 benchmarks.
Chinese Translation
新兴的世界-动作模型（World-Action Models, WAMs）通过联合建模未来驾驶场景演变和轨迹规划，在自动驾驶领域表现出良好的性能。然而，现有的WAM通常使用视频数据进行训练，而视频数据仅是潜在4D驾驶场景的2D投影。因此，WAM无法理解和捕捉4D场景的结构，从而生成视觉上合理但在4D上不一致的未来预测，误导下游规划。为了解决这一问题，我们提出了4D-WAM，这是一种利用几何基础模型进行训练时监督的模型，以实现4D一致的世界建模。具体而言，我们将WAM预测的未来帧输入几何基础模型，并使用4D感知响应定义4D一致性损失。该损失鼓励模型在训练过程中理解、表示和预测物理一致的4D场景，而无需额外的推理成本。此外，我们识别出WAM中的早期决策现象，并提出了一种以决策为导向的时间步采样策略，强调在早期高噪声阶段的监督，此时驾驶决策主要形成。通过将4D监督传播到这一关键的决策形成阶段，所提出的策略进一步改善了轨迹规划。大量实验表明，4D-WAM有效建模4D一致的场景演变，并在具有挑战性的NAVSIM-v1和NAVSIM-v2基准测试中实现了最先进的性能。
cs.CV / 4 / 2608.10131
P3CA: Encoder-Agnostic Interpretation of Vision Foundation Model Embeddings via Spatial Probing
P3CA：通过空间探测对视觉基础模型嵌入的编码器无关解释
Abstract
Vision foundation models are increasingly used as reusable encoders in medical image computing, yet their high-dimensional spatial embeddings are difficult to inspect beyond downstream task performance or global dimensionality reduction. We propose position-prompted PCA (P3CA), an encoder-agnostic method for local probing of channel-rich spatial tensors. Given a user-selected spatial prompt, P3CA estimates the feature normalization and dominant covariance directions within that region, then applies the resulting projection to the full tensor to visualize where locally informative directions are expressed. This produces a region-conditioned representation lens without modifying the encoder, retraining, or requiring task-specific labels. We implement P3CA in EmbedVision, an interactive 3D Slicer-based workflow, and evaluate it across natural images, colorectal pathology foundation-model embeddings, and spatial transcriptomic tensors. Across these settings, prompted projections reveal local structure suppressed by global PCA, improve prompt-matched pathology discrimination from frozen three-dimensional projections, and support comparison between learned and measured spatial representations.
Chinese Translation
视觉基础模型在医学图像计算中越来越多地作为可重用的编码器使用，但其高维空间嵌入在超越下游任务性能或全局降维方面难以进行检查。我们提出了位置提示主成分分析（P3CA），这是一种编码器无关的方法，用于对通道丰富的空间张量进行局部探测。给定用户选择的空间提示，P3CA 估计该区域内的特征归一化和主协方差方向，然后将得到的投影应用于完整张量，以可视化局部信息方向的表达位置。这产生了一种区域条件的表示透镜，无需修改编码器、重新训练或要求特定任务标签。我们在 EmbedVision 中实现了 P3CA，这是一个基于交互式 3D Slicer 的工作流程，并在自然图像、结直肠病理基础模型嵌入和空间转录组张量中进行了评估。在这些设置中，提示投影揭示了被全局 PCA 压制的局部结构，提高了从冻结的三维投影中匹配提示的病理区分能力，并支持学习与测量空间表示之间的比较。
cs.CV / 5 / 2608.10162
MAD-HOI: Masked Autoregressive Diffusion for Generating Articulated Hand Object Interactions from Text
MAD-HOI：用于从文本生成关节手物体交互的掩蔽自回归扩散模型
Abstract
Methods for text-based generation of hand-object interaction (HOI) sequences primarily focus on producing smooth, physically plausible trajectories. A truly utilitarian method should additionally support variable-length generation, composite motion sequences, motion completion and infilling, and reliable termination without compromising physical plausibility. Standard diffusion models for HOI generation are typically trained only for text-to-motion generation on atomic motions and require the motion length to be specified a-priori. Autoregressive (AR) methods provide greater sequence-level flexibility, but commonly depend on discrete motion codes, which can lose contact-sensitive motion detail. To address these key limitations, we present a model performing Masked Autoregression with Diffusion for HOI generation (MAD-HOI). Our method starts by encoding hand and object motions in a continuous latent space while keeping them disentangled to maintain stream-wise control. This is followed by a masked autoregressive transformer to predict context features that condition a flow-matching head. MAD-HOI is capable of motion generation for atomic and composite articulated sequences, conditioned motion completion and infilling, as well as EOM (End of Motion) prediction from a single training objective. We provide comprehensive evaluations for these capabilities and benchmark our method on the ARCTIC and GRAB datasets. Our experiments demonstrate that our method generates more diverse and physically plausible interactions compared to other open-sourced baseline methods.
Chinese Translation
基于文本的手物体交互（HOI）序列生成方法主要集中于产生平滑且物理上合理的轨迹。一种真正实用的方法还应支持可变长度生成、复合运动序列、运动完成和填充，以及可靠的终止，而不影响物理合理性。标准的HOI生成扩散模型通常仅针对原子运动进行文本到运动的生成训练，并且需要事先指定运动长度。自回归（AR）方法提供了更大的序列级灵活性，但通常依赖于离散运动编码，这可能会丧失与接触相关的运动细节。为了解决这些关键限制，我们提出了一种用于HOI生成的掩蔽自回归扩散模型（MAD-HOI）。我们的方法首先在连续潜在空间中编码手和物体的运动，同时保持它们的解耦以维持流动控制。接下来，使用掩蔽自回归变换器预测条件流匹配头的上下文特征。MAD-HOI能够生成原子和复合关节序列的运动，条件运动完成和填充，以及从单一训练目标进行的运动结束（EOM）预测。我们对这些能力进行了全面评估，并在ARCTIC和GRAB数据集上对我们的方法进行了基准测试。我们的实验表明，与其他开源基线方法相比，我们的方法生成了更具多样性和物理合理性的交互。
cs.CV / 6 / 2608.10170
Motion Artifact-Aware Self-Supervised Representation Learning for 3D Brain MRI Motion Artifact Reduction
运动伪影感知自监督表示学习用于3D脑MRI运动伪影减少
Abstract
Patient motion remains a source of image degradation in brain MRI, leading to signal loss, blurring, and geometric distortion that compromise quantitative analysis. Existing deep learning methods for motion correction typically rely on paired clean-corrupted data or k-space acquisitions, which are rarely available in clinical settings. We propose SSRL-MAR, a motion artifact-aware unpaired representation learning framework for motion artifact reduction that requires neither paired training data nor explicit motion labels. SSRL-MAR employed a three-stage training strategy: (1) contrastive learning on 3D patches to extract motion representations by contrasting clean and synthetically corrupted images, (2) a motion artifact-aware synthesis network to generate motion artifacts from clean scans, and (3) a motion artifact-aware generator to restore clean volumes using the learned degrader for self-supervised supervision. On in-silico dataset, SSRL-MAR achieved PSNR 23.81dB, SSIM 91.55%, and NMSE 0.79%. On in-vivo MR-ART dataset, the pretrained model reduced motion distortion, and unsupervised domain adaptation further improved anatomical fidelity. Against a source-only supervised model trained on the same simulated pairs, SSRL-MAR improved PSNR by up to 2.0 dB on MR-ART after unsupervised domain adaptation, and remained within 0.25-0.47 dB of an oracle supervised model that requires real paired data unavailable in practice. At the milder motion level, volumetric error in structures such as the corpus callosum and ventricular system decreased by more than 50%, confirming improved neuroanatomical consistency. These results indicate that SSRL-MAR provides a robust and scalable image-domain solution for 3D brain MRI motion correction, enabling reliable structural quantification in large-scale neuroimaging studies without requiring prospectively acquired pairs or acquisition-specific calibration.
Chinese Translation
患者运动仍然是脑MRI图像降质的一个来源，导致信号丢失、模糊和几何失真，从而影响定量分析。现有的运动校正深度学习方法通常依赖于配对的干净-受损数据或k空间采集，而这些在临床环境中很少可用。我们提出了SSRL-MAR，一个运动伪影感知的无配对表示学习框架，用于运动伪影减少，该框架不需要配对训练数据或显式运动标签。SSRL-MAR采用了三阶段训练策略：（1）在3D补丁上进行对比学习，通过对比干净和合成受损图像来提取运动表示；（2）一个运动伪影感知合成网络，从干净扫描中生成运动伪影；（3）一个运动伪影感知生成器，利用学习到的降解器恢复干净体积，以实现自监督监督。在模拟数据集上，SSRL-MAR达到了PSNR 23.81dB，SSIM 91.55%和NMSE 0.79%。在体内MR-ART数据集上，预训练模型减少了运动失真，且无监督领域适应进一步提高了解剖结构的保真度。与在相同模拟对上训练的源仅监督模型相比，SSRL-MAR在无监督领域适应后，在MR-ART上提高了最多2.0 dB的PSNR，并且在0.25-0.47 dB范围内接近需要实际配对数据的oracle监督模型，这些数据在实践中不可用。在较轻的运动水平下，胼胝体和脑室系统等结构的体积误差减少了50%以上，确认了神经解剖一致性的改善。这些结果表明，SSRL-MAR为3D脑MRI运动校正提供了一种稳健且可扩展的图像域解决方案，使得在大规模神经影像学研究中能够可靠地进行结构定量，而无需前瞻性获取的配对数据或特定采集的校准。
cs.CV / 7 / 2608.10173
DoseBridge: Denoising Diffusion Bridge Model for Dose Prediction in Lung Intensity-Modulated Proton Therapy
DoseBridge：用于肺部强度调制质子治疗剂量预测的去噪扩散桥接模型
Abstract
Most radiotherapy dose-prediction models use only CT images and anatomical structures, although intensity-modulated proton therapy (IMPT) dose also depends strongly on beam geometry and available clinical datasets are often small. We present DoseBridge, a denoising diffusion bridge model that uses the patient CT as a structured bridge endpoint and encodes plan-specific beam geometry in a spatially aligned beam mask. Multiscale fusion combines CT, target, organ-at-risk, and beam-mask representations with 1.95% additional parameters. DoseBridge was retrospectively evaluated on single-institution CT images and treatment plans from 52 patients with advanced-stage lung cancer treated with 60 Gy in 30 fractions; 42 cases were used for training and 10 for testing. Performance was assessed using image-similarity, dose-volume, and Lyman-Kutcher-Burman normal-tissue complication probability (NTCP) metrics and compared with two deep-learning models. On the test cohort, DoseBridge achieved a mean absolute error of 4.170 Gy, peak signal-to-noise ratio of 23.06 dB, and structural similarity index of 0.798, outperforming both comparison models on these metrics. Clinical target volume D95 differed from the reference dose by 0.62 +/- 1.6 Gy; signed organ-at-risk mean-dose differences ranged from -0.32 to 0.24 Gy, and NTCP differences were -0.40 +/- 2.2 and 0.52 +/- 3.4 percentage points for acute esophagitis and radiation pneumonitis, respectively. Changing only the beam mask redirected predicted low-dose entrance regions while preserving the high-dose target region. To our knowledge, DoseBridge is the first denoising diffusion bridge model for radiotherapy dose prediction. These results support its feasibility as a beam-aware planning prior for lung IMPT, pending evaluation in larger external cohorts.
Chinese Translation
大多数放射治疗剂量预测模型仅使用CT图像和解剖结构，尽管强度调制质子治疗（IMPT）剂量也在很大程度上依赖于束流几何形状，而可用的临床数据集通常较小。我们提出了DoseBridge，一种去噪扩散桥接模型，它将患者CT作为结构化的桥接端点，并在空间对齐的束流掩膜中编码特定计划的束流几何形状。多尺度融合结合了CT、靶区、风险器官和束流掩膜的表示，增加了1.95%的参数。DoseBridge在52名接受30次分次60 Gy治疗的晚期肺癌患者的单一机构CT图像和治疗计划上进行了回顾性评估；42个病例用于训练，10个用于测试。使用图像相似性、剂量体积和Lyman-Kutcher-Burman正常组织并发症概率（NTCP）指标评估性能，并与两个深度学习模型进行了比较。在测试队列中，DoseBridge实现了4.170 Gy的平均绝对误差、23.06 dB的峰值信噪比和0.798的结构相似性指数，在这些指标上优于两个比较模型。临床靶体积D95与参考剂量的差异为0.62 +/- 1.6 Gy；风险器官的平均剂量差异范围为-0.32到0.24 Gy，急性食管炎和放射性肺炎的NTCP差异分别为-0.40 +/- 2.2和0.52 +/- 3.4个百分点。仅改变束流掩膜重新引导了预测的低剂量入射区域，同时保留了高剂量靶区。据我们所知，DoseBridge是第一个用于放射治疗剂量预测的去噪扩散桥接模型。这些结果支持其作为肺部IMPT的束流感知规划先验的可行性，待在更大外部队列中进行评估。
cs.CV / 8 / 2608.10181
Human versus Computer Vision
人类与计算机视觉
Abstract
Computer vision saliency models predict where people will look, one map per image, and a billion-dollar predicted-attention industry sells those maps in place of measuring real viewers. I test the leading models from the audience side, against 11.4 million webcam gaze points from 3,023 US adults recruited to national quotas, viewing circulating news photographs. I show that an untrained central marker outperforms every trained network, because the content the networks add on top of the center falls where these audiences never look. What accuracy remains is systematically biased, favoring younger, White, and moderate viewers over older, Black, and ideologically extreme ones. I propose a way forward and build on what a group's own gaze reveals about whether a model can learn that group at all, and I apply it across every demographic axis this sample supports. Ultimately, I show how systems that decide what people see can learn to see everyone, and this study supplies the standard by which such a claim should be judged.
Chinese Translation
计算机视觉显著性模型预测人们的注视点，每张图像生成一张地图，而一个价值数十亿美元的预测注意力产业则出售这些地图，取代了对真实观众的测量。我从观众的角度测试了领先模型，使用来自3023名美国成年人（按照国家配额招募）所提供的1140万个网络摄像头注视点，这些成年人正在观看流传的新闻照片。我发现一个未经训练的中心标记的表现优于所有训练过的网络，因为这些网络在中心上添加的内容落在了这些观众从未注视的地方。剩余的准确性系统性地存在偏差，偏向于年轻的、白人和中立的观众，而忽视了年长的、黑人和意识形态极端的观众。我提出了一种前进的方法，并基于一个群体自身的注视揭示该模型是否能够学习该群体，并在该样本支持的每个人口统计轴上应用这一方法。最终，我展示了决定人们所见的系统如何能够学习看到每一个人，并且这项研究提供了评判此类主张的标准。
cs.CV / 9 / 2608.10195
More Accurate, Less Human: Gestalt Grouping in Vision Models
更准确，更少人性：视觉模型中的格式塔分组
Abstract
Human vision organizes what it sees into wholes: same-colored points group into series, similar marks cohere into categories, and shapes complete into recognizable objects. These are the Gestalt operations that visualization design builds on. Whether vision models organize visual content this way has not been systematically tested. We introduce a behavioral battery that scores models against human data from prior perception studies on four grouping tasks: mark-color odd-one-out, color-series counting, silhouette recognition, and object odd-one-out. We apply it to 45 models across five training families: supervised, self-supervised, and contrastive vision-language encoders, open-weight VLMs, and closed foundation models. The battery reveals that agreement with human responses captures aspects of perceptual organization that conventional performance metrics fail to distinguish, with several closed models exhibiting substantially lower alignment than their benchmark accuracy would suggest. Scoring against published perception data therefore gives visualization research a reusable yardstick, requiring no new user study, for auditing whether the models now entering visualization pipelines organize what they see the way their human audience does.
Chinese Translation
人类视觉将所见事物组织成整体：同色点聚集成系列，相似标记凝聚成类别，形状完整成可识别的物体。这些是可视化设计所依赖的格式塔操作。视觉模型是否以这种方式组织视觉内容尚未经过系统测试。我们引入了一种行为测试工具，通过四个分组任务（标记颜色的奇异点、颜色系列计数、轮廓识别和物体奇异点）对模型进行评分，基于之前感知研究中的人类数据。我们将其应用于45个模型，涵盖五种训练类别：监督学习、自监督学习、对比视觉-语言编码器、开放权重的视觉语言模型（VLMs）和封闭基础模型。该测试工具揭示，与人类反应的一致性捕捉了传统性能指标未能区分的感知组织方面，多个封闭模型的对齐程度显著低于其基准准确性所暗示的。因此，针对已发布的感知数据进行评分为可视化研究提供了一个可重复使用的标准，无需新的用户研究，以审计当前进入可视化流程的模型是否以人类观众的方式组织所见内容。
cs.CV / 10 / 2608.10203
A Convolutional Layer Activation Dimensionality Reduction for Out-of-Distribution and Adversarial Attack Detection Methods
用于分布外和对抗攻击检测方法的卷积层激活维度缩减
Abstract
Despite the success of convolutional neural networks in image classification tasks and their general application in multi-modal models, their susceptibility to out-of-distribution and adversarial attack samples raises concerns regarding trustworthiness and safety. Among the approaches to tackle such issues, detection methods that analyze the model's intermediate activations to estimate a confidence score are a promising family that evaluates the decision process, relying on a dimensionality reduction step to enable efficient downstream processing of the high-dimensional activations. However, when considering convolutional layers, the dimensionality reduction methods in the literature either lack a mechanism to control the compression/information-loss trade-off or yield large representations. In this paper, we carefully analyze two state-of-the-art detection methods and their dimensionality reductions for convolutional layers and develop a novel reduction method with a controllable high-compression level. We extend these two state-of-the-art detection methods, enabling the usage of any dimensionality reduction, and evaluate their performance on out-of-distribution and adversarial attack detection. Results show that the detection methods with the proposed dimensionality reduction consistently perform better than, or comparable to, the strongest alternative. Furthermore, the proposed method is shown to reduce computation and memory footprints, given that it has the highest compression among the compared methods.
Chinese Translation
尽管卷积神经网络在图像分类任务中取得了成功，并广泛应用于多模态模型，但其对分布外样本和对抗攻击样本的脆弱性引发了对其可信性和安全性的担忧。在应对这些问题的方法中，分析模型中间激活以估计置信度分数的检测方法是一类有前景的方案，它通过维度缩减步骤来评估决策过程，从而实现高维激活的高效下游处理。然而，在考虑卷积层时，文献中的维度缩减方法要么缺乏控制压缩/信息损失权衡的机制，要么产生较大的表示。在本文中，我们仔细分析了两种最先进的检测方法及其卷积层的维度缩减，并开发了一种具有可控高压缩水平的新型缩减方法。我们扩展了这两种最先进的检测方法，使其能够使用任何维度缩减，并评估它们在分布外和对抗攻击检测中的性能。结果表明，采用所提出的维度缩减的检测方法的性能始终优于或可与最强替代方法相媲美。此外，所提出的方法在计算和内存占用方面也显示出减少，因为它在比较方法中具有最高的压缩率。
cs.CV / 11 / 2608.10278
Chain of Spatial Thoughts: Modality-Agnostic Spatial Grounding for Vision Language Models
空间思维链：对视觉语言模型的模态无关空间基础
Abstract
Spatial understanding is fundamental to embodied intelligence, underpinning applications such as robotic manipulation, embodied navigation, and autonomous driving. Although recent vision-language models (VLMs) have achieved impressive performance on spatial reasoning benchmarks, state-of-the-art approaches typically rely on additional spatial encoders or architectural modifications during inference, increasing computational cost. We introduce Space Tokens, a lightweight, architecture-agnostic framework that equips VLMs with explicit continuous spatial representations without requiring additional inference-time modules. By distilling scene-level 3D geometry and object-centric spatial attributes into continuous latent tokens, our method enables these modalities to be directly incorporated into a chain-of-thought reasoning process, thereby improving the VLM's spatial reasoning capabilities. At the same time, the learned representations can be explicitly decoded to verify that they encode meaningful geometric information, while the unified token interface remains extensible to additional modalities. Experiments on VSI-Bench improve Qwen3-VL-8B by 4.3% and SenseNova-SI-1.3 by 1.3%, while achieving state-of-the-art performance on object size (79.2%) and room size estimation (75.7%). These results demonstrate that continuous spatial tokens provide an effective, interpretable, and computationally efficient mechanism for integrating geometric reasoning into large vision-language models.
Chinese Translation
空间理解是具身智能的基础，支撑着机器人操作、具身导航和自动驾驶等应用。尽管近期的视觉语言模型（VLMs）在空间推理基准测试中取得了令人瞩目的成绩，但最先进的方法通常在推理过程中依赖额外的空间编码器或架构修改，从而增加了计算成本。我们提出了空间标记（Space Tokens），这是一种轻量级、架构无关的框架，能够为VLMs提供明确的连续空间表示，而无需额外的推理时模块。通过将场景级的三维几何和以对象为中心的空间属性提炼为连续潜在标记，我们的方法使这些模态能够直接融入思维链推理过程中，从而提升VLM的空间推理能力。同时，学习到的表示可以被明确解码，以验证其编码了有意义的几何信息，而统一的标记接口仍然可以扩展到其他模态。在VSI-Bench上的实验中，Qwen3-VL-8B的性能提高了4.3%，SenseNova-SI-1.3提高了1.3%，同时在物体大小（79.2%）和房间大小估计（75.7%）方面达到了最先进的性能。这些结果表明，连续空间标记为将几何推理整合到大型视觉语言模型中提供了一种有效、可解释且计算高效的机制。
cs.CV / 12 / 2608.10286
TRACE-GS: On-Policy Trajectory Distillation with Privileged Geometric Conditioning for Sparse-View 3DGS Restoration
TRACE-GS：基于特权几何条件的稀疏视图3D高斯点云重建的在线轨迹蒸馏
Abstract
We present TRACE-GS, an on-policy trajectory distillation framework that leverages privileged geometric conditioning at training time, thereby adapting a diffusion prior to sparse-view 3D Gaussian Splatting (3DGS) restoration. Rather than pursuing increasingly sophisticated restoration architectures, we identify a more fundamental limitation shared by existing diffusion-based approaches: supervision at independently noised states does not cover those reached during inference. In sparse-view 3DGS, under-constrained geometry biases denoising from the outset, and the resulting deviations compound along the rollout. TRACE-GS instead performs on-policy trajectory distillation: a teacher conditioned on richer geometry from additional training views supplies targets along the sparse-view student's own rollout, aligning denoising directions and cross-view responses at each visited state. This training-only geometry places TRACE-GS in the learning using privileged information (LUPI) setting. At deployment, only the sparse-view student is retained, and its restored renderings serve as pseudo-observations for 3DGS refinement. To the best of our knowledge, TRACE-GS is the first to derive on-policy supervision from privileged geometry for sparse-view 3DGS restoration, achieving consistent gains and strong generalization across datasets and sparse-view settings.
Chinese Translation
我们提出了TRACE-GS，这是一种在线轨迹蒸馏框架，利用训练时的特权几何条件，从而将扩散先验适应于稀疏视图3D高斯点云重建（3DGS）。我们并未追求日益复杂的重建架构，而是识别出现有基于扩散的方法所共享的更根本的局限性：在独立噪声状态下的监督并未覆盖推理过程中达到的状态。在稀疏视图3DGS中，欠约束的几何形状从一开始就对去噪产生偏差，而由此产生的偏差在整个推理过程中累积。TRACE-GS则执行在线轨迹蒸馏：一个基于来自额外训练视图的更丰富几何条件的教师，在稀疏视图学生自身的推理过程中提供目标，在每个访问的状态下对齐去噪方向和跨视图响应。这种仅限于训练的几何条件使TRACE-GS处于使用特权信息学习（LUPI）设置。在部署时，仅保留稀疏视图学生，其重建的渲染结果作为3DGS精细化的伪观测。根据我们所知，TRACE-GS首次从特权几何中推导出在线监督用于稀疏视图3DGS重建，在各数据集和稀疏视图设置中实现了一致的提升和强大的泛化能力。
cs.CV / 13 / 2608.10289
SeFaR: Semantic Feature-aware Robustness Testing of Deep Neural Networks
SeFaR：深度神经网络的语义特征感知鲁棒性测试
Abstract
Deep neural networks are increasingly deployed in safety-critical domains as perception modules, where failures are often caused due to rare and under-represented scenarios. This necessitates the need to evaluate the semantic robustness of perception models; conformance of behavior to high-level requirements over real-world perceptual variability. To address this, we propose SeFaR, a framework for systematic semantic-feature-centric testing of vision models. Given a natural-language requirement and a set of satisfying inputs, SeFaR evaluates robustness with respect to diverse realistic semantic variations that preserve requirement satisfaction. The approach employs a novel hierarchical concept model enabling structured exploration of the feature space and incorporation of domain knowledge via user-defined concepts. State-of-the-art diffusion and vision-language models are leveraged to generate photorealistic semantics-preserving perturbations and identification of previously unknown features impacting behavior. A feedback-driven adaptive process is adopted to generate interpretable failure-inducing semantic concepts along with corresponding test inputs. Evaluation on case studies demonstrates that the proposed framework effectively satisfies requirement preconditions while identifying requirement-independent features that influence model decisions, enabling it to both uncover faults and relate them to such features.
Chinese Translation
深度神经网络在安全关键领域作为感知模块的应用日益增多，而其故障往往是由于稀有和代表性不足的场景所导致的。这就需要评估感知模型的语义鲁棒性，即在真实世界感知变异性下行为与高层次要求的一致性。为此，我们提出了SeFaR，一个系统的以语义特征为中心的视觉模型测试框架。给定自然语言要求和一组满足该要求的输入，SeFaR评估相对于保持要求满足的多样化现实语义变异的鲁棒性。该方法采用了一种新颖的层次概念模型，能够结构化地探索特征空间，并通过用户定义的概念融入领域知识。利用最先进的扩散模型和视觉-语言模型生成保留照片真实感的语义扰动，并识别影响行为的先前未知特征。采用反馈驱动的自适应过程生成可解释的导致失败的语义概念及相应的测试输入。对案例研究的评估表明，所提出的框架有效满足要求前提，同时识别出影响模型决策的独立于要求的特征，从而能够揭示故障并将其与这些特征关联。
cs.CV / 14 / 2608.10291
MRIComp4Flow: Compression of 3D Brain MRI for Training Multi-Modal Generative Models
MRIComp4Flow：用于训练多模态生成模型的3D脑MRI压缩
Abstract
Large-scale multi-modal MRI datasets impose substantial storage and I/O costs, limiting the training of 3D generative models on commodity infrastructure. While lossy compression is known to preserve accuracy for discriminative segmentation networks, its effect on generative models, which must learn the full data distribution rather than a decision boundary, is unexplored. We study whether standard image codecs can effectively compress semantically rich brain tumor MRI while preserving the fidelity required to train and deploy a 3D MRI generative model. Each 3D volume is compressed with JPEG2000 or a near-lossless JPEG-LS pipeline. Next, a Wavelet Flow Matching model, conditioned on BraTS image sequences (T1n, T1c, T2, T2f), is trained on compressed data, and the resulting models are evaluated on the validation set. At a 20:1 compression ratio, synthesis quality is statistically equivalent to a model trained on uncompressed data within a pre-specified margin ($\Delta$PSNR $<1$,dB, $\Delta$SSIM $<0.02$; paired TOST $p=[[p]]$): mean PSNR is 27.3,dB vs. 27.0,dB and mean SSIM is 0.95 vs. 0.96 across modalities. Our results indicate that JPEG2000 compression is a practical step toward scalable 3D MRI generative modeling without degrading synthesis quality. The codebase is available at https://github.com/lisafis/MRIComp4Flow .
Chinese Translation
大规模多模态MRI数据集带来了可观的存储和I/O成本，限制了在普通基础设施上训练3D生成模型的能力。尽管有损压缩已知能够保持判别分割网络的准确性，但其对生成模型的影响尚未被探索，生成模型必须学习完整的数据分布而非决策边界。我们研究了标准图像编码器是否能够有效压缩语义丰富的脑肿瘤MRI，同时保持训练和部署3D MRI生成模型所需的保真度。每个3D体积使用JPEG2000或近无损JPEG-LS管道进行压缩。接下来，基于BraTS图像序列（T1n、T1c、T2、T2f）训练了一个Wavelet Flow Matching模型，使用压缩数据进行训练，并在验证集上评估所得到的模型。在20:1的压缩比下，合成质量在预设的范围内与在未压缩数据上训练的模型统计上等效（$ ext{ΔPSNR} < 1 ext{dB}, ext{ΔSSIM} < 0.02$；配对TOST $p=[[p]]$）：各模态的平均PSNR为27.3 dB对比27.0 dB，平均SSIM为0.95对比0.96。我们的结果表明，JPEG2000压缩是朝着可扩展的3D MRI生成建模迈出的实际一步，而不会降低合成质量。代码库可在 https://github.com/lisafis/MRIComp4Flow 获取。
cs.CV / 15 / 2608.10295
Frozen Brain-MRI Foundation Models Are Site Fingerprints
冻结脑部MRI基础模型是场所指纹
Abstract
Frozen foundation-model (FM) embeddings are increasingly used as off-the-shelf brain-MRI representations, on the assumption that they capture anatomy. We audit what they actually encode and find that acquisition site is a large, intrinsic component of the representation. Across two independent cohorts (ABIDE-I, ABIDE-II), three frozen 3-D encoders (brain-pretrained, CT-pretrained, and randomly initialized), and every network depth, site is linearly decodable at roughly 0.9 balanced accuracy at deep layers, exceeding the decodability of every clinical or demographic variable (sex, age, autism diagnosis) at every layer. The effect is intrinsic rather than learned: a randomly initialized encoder is already a ~0.9 site classifier on both cohorts and across three architecture families (Swin, ViT, ResNet), and site is decodable at ~0.95 directly from the raw downsampled image with no encoder, so the fingerprint reflects low-level image statistics that any encoder preserves rather than a product of pretraining. Residualizing measured population covariates leaves site decodability essentially unchanged, indicating an acquisition- rather than population-driven effect. A nonlinear probe matches the linear one, so the fingerprint is fully linearly accessible. The site subspace is removable post hoc by iterative null-space projection or ComBat (site decodability 0.94 -> 0.07/0.00), and is a site-attribution concern for shared or federated embeddings; but for dense segmentation this removal is not free, because site and anatomy occupy an entangled linear subspace (a matched-rank random-direction projection is Dice-neutral, whereas removing the site subspace is destructive). We recommend site-audited use of frozen brain-MRI FMs and release an open audit toolkit.
Chinese Translation
冻结基础模型（FM）嵌入越来越多地被用作现成的脑部MRI表征，假设它们能够捕捉解剖结构。我们审查了它们实际编码的内容，发现获取场所是表征的一个重要内在组成部分。在两个独立的队列（ABIDE-I，ABIDE-II）、三个冻结的3D编码器（脑部预训练、CT预训练和随机初始化）以及每个网络深度中，获取场所可以在深层以大约0.9的平衡准确率线性解码，超过了每个层次上每个临床或人口变量（性别、年龄、自闭症诊断）的可解码性。这个效应是内在的而非学习得来的：一个随机初始化的编码器在两个队列和三个架构系列（Swin、ViT、ResNet）中已经是一个约0.9的场所分类器，并且场所可以直接从原始下采样图像以约0.95的准确率解码，无需编码器，因此指纹反映的是任何编码器所保留的低级图像统计特征，而非预训练的结果。对测量的人口协变量进行残差化处理几乎不改变场所的可解码性，表明这是一个由获取驱动而非人口驱动的效应。非线性探测器与线性探测器匹配，因此指纹是完全线性可访问的。通过迭代零空间投影或ComBat（场所可解码性从0.94降至0.07/0.00）可以事后去除场所子空间，这对共享或联邦嵌入构成场所归属问题；但对于密集分割，这种去除并非无代价，因为场所和解剖结构占据一个纠缠的线性子空间（匹配秩随机方向投影是Dice中性的，而去除场所子空间是破坏性的）。我们建议对冻结脑部MRI基础模型进行场所审核使用，并发布一个开放的审核工具包。
cs.CV / 16 / 2608.10316
UniMod: Enhancing Multi-Modal Medical Diagnosis through Cross-Modality and Within-Modality Alignment
UniMod：通过跨模态和同模态对齐增强多模态医学诊断
Abstract
Multi-modal learning combining medical images and clinical text is promising for disease diagnosis. However, standard multi-modal training leads to shortcut learning: models exploit the easier modality (e.g., diagnostic cues in text) while neglecting harder-to-learn features (e.g., subtle visual patterns). We propose UniMod, a framework that mitigates shortcut learning by requiring each modality to predict the diagnosis on its own. It supervises image-only, text-only, and multi-modal classification simultaneously, so each modality must extract diagnostic features. We add cross-modality alignment for knowledge transfer and within-modality supervised contrastive alignment over same-diagnosis patients. On Harvard-Glaucoma, UniMod reaches 0.850 AUC, outperforming OGM-GE and Gradient Blending by 1.6-1.8%; on CheXpert Plus, it reaches 0.966 AUC, surpassing them by over 5%. UniMod also extends to 5-class multi-label diagnosis without architectural change, improving mean AUC by 0.097 over CGGM.
Chinese Translation
结合医学图像和临床文本的多模态学习在疾病诊断中前景广阔。然而，标准的多模态训练导致了捷径学习：模型利用更容易的模态（例如，文本中的诊断线索），而忽视了更难学习的特征（例如，微妙的视觉模式）。我们提出了UniMod，一个通过要求每个模态独立预测诊断来减轻捷径学习的框架。它同时监督图像单独、文本单独和多模态分类，因此每个模态必须提取诊断特征。我们增加了跨模态对齐以实现知识转移，并在同一诊断患者之间进行同模态监督对比对齐。在哈佛青光眼数据集上，UniMod达到了0.850的AUC，超越OGM-GE和Gradient Blending 1.6-1.8%；在CheXpert Plus上，达到了0.966的AUC，超过它们5%以上。UniMod还扩展到5类多标签诊断而无需架构更改，平均AUC比CGGM提高了0.097。
cs.CV / 17 / 2608.10317
From Detection to Understanding: TAR and TAR-Bench for Multi-Task Traffic Anomaly Reasoning
从检测到理解：用于多任务交通异常推理的TAR和TAR-Bench
Abstract
We present TAR (Traffic Anomaly Reasoning) and TAR-Bench datasets, resources for training and evaluating video-language models beyond anomaly detection. TAR contains 44,040 chain-of-thought training annotations across 10 tasks for 3,670 CCTV videos ($\sim$26 hours) from eight public datasets. Its evaluation component, TAR-Bench, contains 960 human-curated test annotations for 80 held-out clips trimmed from 17 public YouTube videos. TAR's training annotations are produced with MAVEN, which consolidates multi-scale video evidence into structured event descriptions before generating question-answer pairs and reasoning traces. On TAR-Bench, eleven vision-language models reveal that strong question-answering accuracy does not reliably predict temporal or scene reasoning ability. Multi-task fine-tuning on TAR yields consistent gains, with the full 10-task model improving aggregate score by 21.4 points over its zero-shot baseline. TAR and TAR-Bench provide the official training and in-domain evaluation data for AI City Challenge 2026 Track 3. The dataset is available at https://huggingface.co/datasets/nvidia/PhysicalAI-Traffic-Anomaly-Reasoning
Chinese Translation
我们提出了TAR（交通异常推理）和TAR-Bench数据集，这些资源用于训练和评估超越异常检测的视频语言模型。TAR包含来自八个公共数据集的3,670个闭路电视视频（约26小时）的10个任务中共44,040个思维链训练注释。其评估组件TAR-Bench包含960个人工策划的测试注释，针对从17个公共YouTube视频中剪辑的80个保留片段。TAR的训练注释是通过MAVEN生成的，该系统将多尺度视频证据整合为结构化事件描述，然后生成问答对和推理轨迹。在TAR-Bench上，十一种视觉语言模型显示，强大的问答准确性并不能可靠地预测时间或场景推理能力。在TAR上进行多任务微调带来了持续的提升，完整的10任务模型在其零样本基线基础上提高了21.4分。TAR和TAR-Bench为2026年AI City Challenge第3赛道提供了官方训练和领域内评估数据。数据集可在https://huggingface.co/datasets/nvidia/PhysicalAI-Traffic-Anomaly-Reasoning获取。
cs.CV / 18 / 2608.10343
ENCORE: Efficient Noise Context-Aware Representation for Low-Dose CT Denoising
ENCORE：低剂量CT去噪的高效噪声上下文感知表示
Abstract
While deep learning-based denoising has become widely adopted in low-dose CT, conventional models use generic architectures designed for natural images, failing to account for non-stationary and spatially correlated CT noise characteristics. To address this, we propose an Efficient Noise COntext-aware REpresentation (ENCORE) framework that explicitly leverages CT noise characteristics and anatomical features. First, we reformulate the noise synthesis procedure based on a realistic noise distribution beyond the conventional Gaussian approximation, establishing a rigorous foundation for training pair generation. Next, we extract local noise power and correlation contexts to guide the denoising process. To fully leverage the potential of noise context, we propose a FlyingConv module, which adaptively changes convolution weights for each local image region. Notably, our approach demonstrates substantial gains in both denoising quality and computational efficiency. Furthermore, manipulating the intensity of the noise context maps at inference time enables zero-shot conditional denoising, allowing for dynamic control over the output image texture. The entire pipeline is available at https://github.com/minwoo-yu/ENCORE.git
Chinese Translation
尽管基于深度学习的去噪方法在低剂量CT中得到了广泛应用，但传统模型使用的是为自然图像设计的通用架构，未能考虑到CT噪声特征的非平稳性和空间相关性。为了解决这一问题，我们提出了一种高效噪声上下文感知表示（ENCORE）框架，该框架明确利用CT噪声特征和解剖特征。首先，我们基于超越传统高斯近似的真实噪声分布重新制定了噪声合成过程，为训练对的生成奠定了严格的基础。接下来，我们提取局部噪声功率和相关上下文，以指导去噪过程。为了充分利用噪声上下文的潜力，我们提出了一种FlyingConv模块，该模块为每个局部图像区域自适应地改变卷积权重。值得注意的是，我们的方法在去噪质量和计算效率上均表现出显著提升。此外，在推理时操控噪声上下文图的强度实现了零样本条件去噪，使得对输出图像纹理的动态控制成为可能。整个流程可在 https://github.com/minwoo-yu/ENCORE.git 获取。
cs.CV / 19 / 2608.10345
CasDeblurGS: Cascaded 2D-to-3D Multi-View Consistency for 3D Gaussian Splatting from Two Blurry Images
CasDeblurGS：基于级联的2D到3D多视图一致性用于从两幅模糊图像中进行3D高斯点云重建
Abstract
Free-viewpoint 3D scene media is increasingly important for immersive applications, yet practical capture often suffers from severe view sparsity and motion blur. Although neural rendering has advanced sparse-view synthesis, existing blur-aware methods typically require substantial multi-view redundancy, accurate camera poses, or costly per-scene optimization. We address a stringent yet practical setting: reconstructing a coherent 3D scene from only two motion-blurred images with known intrinsics, without input-view poses, auxiliary sharp images, or per-scene test-time optimization. To this end, we propose CasDeblurGS, a cascaded framework that progressively recovers reliable cross-view information from local 2D correspondences to global 3D guidance. Stage 1 constructs locally reliable guidance through occlusion-aware correspondence filtering, while Stage 2 aggregates the intermediate restorations into a provisional pose-free 3D Gaussian representation whose input-view re-renders provide dense global guidance for final restoration. The resulting views enable a more coherent 3D representation and higher-quality novel-view synthesis. Experiments on real-world and synthetic Deblur-NeRF scenes show consistent gains over strong baselines, improving PSNR by 1.19 dB and 2.11 dB, respectively. Progressive ablations, cross-view correspondence visualization, and camera reprojection analysis further demonstrate improvements in both rendering quality and multi-view geometric consistency.
Chinese Translation
自由视角的3D场景媒体在沉浸式应用中日益重要，但实际捕获往往面临严重的视角稀疏和运动模糊。尽管神经渲染在稀疏视图合成方面取得了进展，现有的模糊感知方法通常需要大量的多视图冗余、准确的相机姿态或昂贵的逐场景优化。我们解决了一个严格但实用的设置：仅从两幅已知内参的运动模糊图像中重建一致的3D场景，而无需输入视图姿态、辅助清晰图像或逐场景的测试时优化。为此，我们提出了CasDeblurGS，一个级联框架，逐步从局部2D对应关系恢复可靠的跨视图信息到全局3D指导。第一阶段通过考虑遮挡的对应关系过滤构建局部可靠的指导，而第二阶段则将中间恢复结果聚合成一个临时的无姿态3D高斯表示，其输入视图的重渲染提供了最终恢复的密集全局指导。生成的视图实现了更一致的3D表示和更高质量的新视图合成。在真实世界和合成的Deblur-NeRF场景上的实验显示，相较于强基线有一致的提升，PSNR分别提高了1.19 dB和2.11 dB。逐步消融实验、跨视图对应关系可视化和相机重投影分析进一步证明了渲染质量和多视图几何一致性的改善。
cs.CV / 20 / 2608.10346
Towards Unified Dynamic Face Landmark Detection
朝向统一的动态人脸关键点检测
Abstract
Although advancements in face landmark detection (FLD) methods continue to push performance boundaries, they overlook two major functional limitations: (1) different network parameters need to be trained independently for each ``$N$-point'' benchmark dataset, and (2) a model trained on an ``$N$-point'' dataset reliably outputs only the $N$ landmarks. In our work, we first conceptualize Face Part-Anchored Landmark Positions (FPALPs), wherein each landmark is treated as a progression value between zero (start) and one (end) along a face part's contour. Every landmark can be expressed in the FPALP format, irrespective of its source dataset, hence unlocking the ability to unify all ``$N$-point'' datasets into a single dataset. Secondly, we represent each landmark with an FPALP-based query, refine it progressively with a cross-modality decoder, and predict its coordinates based on the final representation. Our approach, called Unified Dynamic FLD, embodies these two design choices and streamlines the landmark detection pipeline by enabling (1) a single model to learn on any number of ``$N$-point'' datasets, and (2) yield any number of specific landmark predictions by loading the designated landmark queries at runtime. Extensive experiments on multiple benchmark datasets show that our method delivers these benefits while remaining competitive with, and in several cases outperforming existing state-of-the-art methods.
Chinese Translation
尽管人脸关键点检测（FLD）方法的进展不断推动性能的边界，但它们忽视了两个主要的功能限制：（1）不同的网络参数需要为每个“$N$-点”基准数据集独立训练，以及（2）在“$N$-点”数据集上训练的模型仅能可靠地输出$N$个关键点。在我们的研究中，我们首先概念化了人脸部位锚定的关键点位置（FPALPs），其中每个关键点被视为沿着人脸部位轮廓从零（起始）到一（结束）的进展值。每个关键点都可以用FPALP格式表达，无论其源数据集如何，从而解锁将所有“$N$-点”数据集合并为单一数据集的能力。其次，我们用基于FPALP的查询表示每个关键点，通过跨模态解码器逐步优化，并根据最终表示预测其坐标。我们的方法称为统一动态FLD，体现了这两种设计选择，并通过使（1）单一模型能够在任意数量的“$N$-点”数据集上进行学习，以及（2）通过在运行时加载指定的关键点查询来生成任意数量的特定关键点预测，从而简化了关键点检测流程。在多个基准数据集上的广泛实验表明，我们的方法在保持与现有最先进方法的竞争力的同时，在多个情况下超越了它们，提供了这些优势。
cs.CV / 21 / 2608.10396
FormStruct-Bench:A Hierarchical and Diagnostic Benchmark for Table-Form Document Structure Recognition
FormStruct-Bench：一种用于表格形式文档结构识别的分层诊断基准
Abstract
Transforming table-form documents into machine-processable records requires recovering not only their visible content but also the multilevel structure that organizes it. However, existing benchmarks evaluate either holistic document outputs or conventional table grids, and their aggregate scores provide little insight into where structural failures occur. We introduce FormStruct-Bench, a hierarchical and diagnostic benchmark that evaluates table-form document structure recognition at both the document level and progressively finer component levels, allowing aggregate performance to be traced back to specific structural failure modes. To construct auditable ground truth at scale, we annotate 70 reusable templates and expand them into 7,000 verified instances through a provenance-preserving Director--Artist--Verifier pipeline; all 1,100 instances in the template-disjoint test set additionally receive human review. Our evaluation protocol uses five primary metrics and three structure-specific diagnostics across page, schema, and component levels, together with slices over difficulty, structural constraints, and visual degradation. Across 14 API-hosted and locally deployable systems plus two SFT variants, the best document-level score reaches 83.85%, whereas the best reported fine-grained structural score remains below 18%. These results reveal a pronounced gap between reading document content and recovering the hierarchy and regional organization required for reliable table-form understanding.
Chinese Translation
将表格形式文档转化为机器可处理的记录，不仅需要恢复其可见内容，还需重建组织这些内容的多层次结构。然而，现有的基准评估要么是整体文档输出，要么是传统的表格网格，其综合得分对结构性失败发生的具体位置提供的洞察有限。我们提出了FormStruct-Bench，这是一种分层和诊断基准，能够在文档层面及逐渐细化的组件层面评估表格形式文档的结构识别，从而使综合性能能够追溯到特定的结构性失败模式。为了大规模构建可审计的真实数据，我们注释了70个可重用模板，并通过一个保持来源的导演-艺术家-验证者管道将其扩展为7000个经过验证的实例；所有在模板不重叠测试集中的1100个实例还经过人工审核。我们的评估协议使用五个主要指标和三个结构特定的诊断，涵盖页面、模式和组件层面，并结合难度、结构约束和视觉降级的切片。在14个API托管和本地可部署的系统以及两个SFT变体中，最佳文档级得分达到83.85%，而最佳报告的细粒度结构得分仍低于18%。这些结果揭示了读取文档内容与恢复可靠的表格形式理解所需的层次和区域组织之间的显著差距。
cs.CV / 22 / 2608.10411
A second-order theory of texture for depth from focus
基于焦点深度的第二阶纹理理论
Abstract
We present a theory of textured appearance of optically rough surfaces based on wave optics, emphasizing the role of texture for passive depth from focus. Our theory shows that even surfaces that traditional computer vision would consider textureless can produce textured appearance, due to subjective speckle from surface microgeometry. We analyze the properties of this second-order texture, and show that we can enhance its contrast under natural ambient lighting by simply using a narrowband spectral filter. Doing so results in dramatic improvements in passive depth reconstruction of seemingly textureless scenes, as we demonstrate through extensive theory, simulations, and real-world experiments.
Chinese Translation
我们提出了一种基于波动光学的光学粗糙表面的纹理外观理论，强调了纹理在被动焦点深度中的作用。我们的理论表明，即使是传统计算机视觉认为无纹理的表面，也可以产生纹理外观，这是由于表面微观几何结构引起的主观斑点效应。我们分析了这种第二阶纹理的特性，并展示了通过简单使用窄带光谱滤波器，我们可以在自然环境光照下增强其对比度。这样做显著改善了看似无纹理场景的被动深度重建，我们通过广泛的理论、模拟和实际实验进行了验证。
cs.CV / 23 / 2608.10413
DriveVLA-M0: Failure-Aware Memory Augmentation for Autonomous Driving
DriveVLA-M0：面向失败的自主驾驶记忆增强
Abstract
Vision-Language-Action (VLA) models have recently emerged as a promising paradigm for end-to-end autonomous driving by enabling unified reasoning across perception, language, and planning. However, existing approaches lack mechanisms to exploit past failures or adapt to distribution shifts, causing the model to persistently underperform on similar scenarios where it has previously failed. In this paper, we propose DriveVLA-M0, a retrieval-augmented VLA with failure-aware latent memory. We construct a latent memory pool that stores failure cases along with their structure scene representations and expert trajectory labels, and design a dedicated Retrieve Model that decouples static road structure and dynamic agent interactions to enable structurally grounded retrieval. At inference time, retrieved cases are injected into the model via a lightweight decoupled LoRA-based test-time training (TTT) mechanism, allowing targeted and scenario-specific correction without modifying the backbone. Extensive experiments on NAVSIMv1 and NAVSIMv2 benchmark demonstrate that our approach consistently outperforms prior methods, achieving 94.1 PDMS on Navtest and 47.0 EPDMS on Navhard with only 26.44 ms TTT backward latency overhead. Furthermore, we show that DriveVLA-M0 scales effectively with additional memory, enabling training-free performance gains through memory expansion. The code is available at https://github.com/ZebinX/DriveVLA-M0.
Chinese Translation
视觉-语言-行动（VLA）模型最近作为一种有前景的端到端自主驾驶范式出现，能够在感知、语言和规划之间实现统一推理。然而，现有方法缺乏利用过去失败或适应分布变化的机制，导致模型在类似场景中持续表现不佳。本文提出了DriveVLA-M0，一种具有失败感知潜在记忆的检索增强型VLA。我们构建了一个潜在记忆池，存储失败案例及其结构场景表示和专家轨迹标签，并设计了一个专用的检索模型，解耦静态道路结构和动态代理交互，以实现结构化的检索。在推理时，通过轻量级的解耦LoRA基础测试时训练（TTT）机制将检索到的案例注入模型，允许在不修改主干网络的情况下进行针对性和场景特定的修正。在NAVSIMv1和NAVSIMv2基准上的广泛实验表明，我们的方法始终优于先前的方法，在Navtest上实现了94.1 PDMS，在Navhard上实现了47.0 EPDMS，且仅有26.44毫秒的TTT反向延迟开销。此外，我们还展示了DriveVLA-M0能够有效扩展额外记忆，通过记忆扩展实现无训练性能提升。代码可在 https://github.com/ZebinX/DriveVLA-M0 获取。
cs.CV / 24 / 2608.10426
GeoSeg-OV: Bridging Geospatial Gaps with Structural Guidance for Open-Vocabulary Remote Sensing Segmentation
GeoSeg-OV：利用结构指导弥补地理空间差距的开放词汇遥感分割
Abstract
Open-vocabulary remote sensing segmentation has recently emerged as a promising paradigm that enables pixel-level recognition of arbitrary categories specified by natural language, including classes unseen during training. However, geospatial domain shifts caused by heterogeneous regions, spatial resolutions, and acquisition platforms weaken visual-text matching and limit cross-dataset generalization. Recent attempts have begun to incorporate auxiliary vision foundation models (VFMs), typically coupling their features with text embeddings as additional matching evidence. However, this strategy may introduce inconsistent matching signals while leaving the structure-sensitive representations of VFMs insufficiently exploited. We therefore propose GeoSeg-OV, which decouples auxiliary VFM features from visual-text matching and repurposes them as structural guidance for cost aggregation and decoding. GeoSeg-OV constructs an orientation-robust cost volume from multi-rotation CLIP features, while a frozen VFM extracts multi-scale structure-sensitive features in parallel. We propose Structure-Guided Aggregation (SGA), which integrates cost tokens and CLIP semantic guidance with VFM-derived pairwise structural biases for coherent spatial propagation, followed by text-conditioned class-wise reasoning. We further introduce Cost-Aware Decoding (CAD) to adaptively refine and fuse multi-scale semantic and structural guidance based on the current decoder context. On the global High-Resolution Land Cover (HRLC) benchmark spanning seven datasets across six continents, GeoSeg-OV outperforms the state-of-the-art by +2.5 and +2.7 average mIoU under two training settings. A large-scale zero-shot case study further demonstrates its generalization across geographic domains and category systems without target-domain annotations or retraining.
Chinese Translation
开放词汇遥感分割最近作为一种有前景的范式出现，使得能够对由自然语言指定的任意类别进行像素级识别，包括训练期间未见的类别。然而，由于异质区域、空间分辨率和获取平台造成的地理空间领域转移削弱了视觉-文本匹配，并限制了跨数据集的泛化。近期的尝试开始将辅助视觉基础模型（VFM）纳入其中，通常将其特征与文本嵌入结合，作为额外的匹配证据。然而，这种策略可能会引入不一致的匹配信号，同时使得VFM的结构敏感表示未得到充分利用。因此，我们提出了GeoSeg-OV，它将辅助VFM特征与视觉-文本匹配解耦，并将其重新用作成本聚合和解码的结构指导。GeoSeg-OV从多旋转的CLIP特征构建了一个方向稳健的成本体积，同时一个冻结的VFM并行提取多尺度的结构敏感特征。我们提出了结构指导聚合（SGA），它将成本标记和CLIP语义指导与VFM派生的成对结构偏差整合，以实现一致的空间传播，随后进行文本条件的类别推理。我们进一步引入了成本感知解码（CAD），以根据当前解码器上下文自适应地细化和融合多尺度的语义和结构指导。在跨越六大洲七个数据集的全球高分辨率土地覆盖（HRLC）基准测试中，GeoSeg-OV在两种训练设置下的平均mIoU分别超过了最先进的技术2.5和2.7。大规模的零-shot案例研究进一步证明了其在没有目标领域注释或重新训练的情况下跨地理领域和类别系统的泛化能力。
cs.CV / 25 / 2608.10429
Lesion-Aware Adaptive Fourier Neural Operator for CT-to-PSMA PET Synthesis in Prostate Cancer
针对前列腺癌的CT到PSMA PET合成的病灶感知自适应傅里叶神经算子
Abstract
Deep learning models that synthesize PET from CT or MRI can reduce patient dose and scanner demand, but are typically optimized with global losses such as L1 or mean squared error (MSE) that treat all voxels similarly. In whole-body PSMA-PET, tumor voxels occupy only a small fraction of the volume, yet carry the clinically relevant activity signal; as a result, models can achieve high structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR) while still underestimating lesion activity or failing to preserve tumor-specific structure. Radiomics provides biologically meaningful descriptors of tumor intensity and texture, but direct radiomics conditioning is time-consuming because it requires feature extraction from delineated lesion regions. We propose LAFNO, a Lesion-Aware Adaptive Fourier Neural Operator for CT-to-PSMA-PET synthesis that replaces high-dimensional radiomics conditioning with two efficient CT-derived proxy channels. Motivated by radiomics analysis of PSMA-avid tumor core and peritumoral regions, LAFNO uses a contrast proxy for local density variation and a disorder proxy for local texture heterogeneity, both injected into the model bottleneck. LAFNO combines whole-volume reconstruction with lesion-level total lesion activity (TLA), tumor-core contrast, and peritumoral supervision. We evaluated LAFNO against four baseline architectures on the TCIA PSMA-PET-CT-Lesions dataset. LAFNO remained competitive on whole-volume image quality, achieving SSIM of 0.960 and 0.938 for 18F- and 68Ga-PSMA, respectively, while reducing per-patient TLA error to 48.3% and 64.0% for 18F- and 68Ga-PSMA, respectively, and achieving the highest tumor-core radiomics reproducibility across all feature classes for both tracers. Peritumoral reproducibility remained tracer-dependent, indicating that biological fidelity in synthetic PSMA-PET remains challenging.
Chinese Translation
从CT或MRI合成PET的深度学习模型可以减少患者剂量和扫描仪需求，但通常使用全局损失（如L1或均方误差（MSE））进行优化，这种方法对所有体素的处理是相似的。在全身PSMA-PET中，肿瘤体素仅占体积的一小部分，但携带临床相关的活性信号；因此，模型可以在结构相似性指数测量（SSIM）和峰值信噪比（PSNR）上取得高分，但仍可能低估病灶活性或未能保留肿瘤特有结构。放射组学提供了肿瘤强度和纹理的生物学意义描述符，但直接的放射组学条件化耗时，因为它需要从划定的病灶区域提取特征。我们提出了LAFNO，一种用于CT到PSMA-PET合成的病灶感知自适应傅里叶神经算子，它用两个高效的CT衍生代理通道替代了高维放射组学条件化。LAFNO的设计受到了对PSMA-亲和肿瘤核心和周围肿瘤区域的放射组学分析的启发，使用了局部密度变化的对比代理和局部纹理异质性的无序代理，这两者都被注入到模型瓶颈中。LAFNO结合了全体积重建与病灶级总病灶活性（TLA）、肿瘤核心对比度和周围肿瘤监督。我们在TCIA PSMA-PET-CT-Lesions数据集上对LAFNO进行了与四种基线架构的评估。LAFNO在全体积图像质量上保持竞争力，对于18F-和68Ga-PSMA分别达到了0.960和0.938的SSIM，同时将每位患者的TLA误差降低至18F-和68Ga-PSMA的48.3%和64.0%，并在所有特征类别中实现了最高的肿瘤核心放射组学可重复性。周围肿瘤的可重复性仍然依赖于示踪剂，表明合成PSMA-PET中的生物学真实性仍然具有挑战性。
cs.CV / 26 / 2608.10435
DynaPPI: A Large-scale Dynamic Protein Dataset for AI-driven Advances in Protein Interactomics
DynaPPI：一个用于推动蛋白质相互作用组学的人工智能驱动的大规模动态蛋白质数据集
Abstract
Diffusion models have been widely explored in protein backbone generation due to their powerful generation capabilities.However, in today's AI-driven biological research, predicting the structure of unknown multi-chain protein aggregates (called "complexes" in biology) remains an unsolved challenge.This is because existing static or dynamic protein datasets focus solely on static snapshots or single-entity trajectories, neglecting the dynamic process of multiple monomers forming complexes.To alleviate this dilemma, we present DynaPPI, a dynamic protein dataset comprising molecular dynamics (MD) trajectories of protein complex formation from dissociated chains to the bound state, as a pivotal resource to bridge the gap between static structural biology and the inherently temporal nature of dynamic molecular interactions.Benefiting from this dataset, diffusion models can explicitly learn the dynamic binding trajectories of known complexes and accurately predict the structures of unknown complexes based on their diverse generative properties, thereby further catalyzing AI-driven structural biology and protein interactomics.
Chinese Translation
扩散模型因其强大的生成能力而在蛋白质主链生成中得到了广泛探索。然而，在当今以人工智能驱动的生物研究中，预测未知多链蛋白质聚集体（在生物学中称为“复合物”）的结构仍然是一个未解决的挑战。这是因为现有的静态或动态蛋白质数据集仅关注静态快照或单一实体轨迹，忽视了多个单体形成复合物的动态过程。为了解决这一困境，我们提出了DynaPPI，这是一个动态蛋白质数据集，包含从解离链到结合状态的蛋白质复合物形成的分子动力学（MD）轨迹，作为弥合静态结构生物学与动态分子相互作用固有时间特性之间差距的重要资源。得益于这一数据集，扩散模型能够明确学习已知复合物的动态结合轨迹，并基于其多样的生成特性准确预测未知复合物的结构，从而进一步推动人工智能驱动的结构生物学和蛋白质相互作用组学的发展。
cs.CV / 27 / 2608.10437
MammoMix: Leveraging Mixture of Experts for Robust Mammogram Breast Detection
MammoMix：利用专家混合模型实现稳健的乳腺X线检查检测
Abstract
Breast lesion detection in mammography remains a challenging task due to variations in image quality, lesion appearance, and population demographics across datasets. While current object detectors such as YOLO and DETR achieve strong results on individual datasets, their performance often degrades when trained on or applied across heterogeneous sources. To address this, we propose MammoMix, a novel framework based on Mixture-of-Experts (MoE) paradigm for robust and generalizable lesion detection. In MammoMix, each expert model is trained on a specific domain, allowing it to specialize in distinct characteristics of its source data. A gating mechanism adaptively weighs contributions from each expert based on input image, combining their outputs to enable domain-adaptive inference. To improve reliability, we further incorporate a calibration module, MoCAE, which adjusts confidence scores to reflect true predictive uncertainty. We evaluate MammoMix on 3 public mammography datasets: CSAW, DDSM, and DMID, covering diverse clinical settings. Results show that MammoMix outperforms baseline detectors in both average precision and reliability, particularly on datasets with greater variability. Our findings demonstrate that expert specialization and calibrated ensemble fusion significantly enhance model generalization and robustness. MammoMix offers a promising step toward dependable AI-assisted breast cancer screening across real-world clinical domains.
Chinese Translation
乳腺X线检查中的病变检测仍然是一项具有挑战性的任务，因为不同数据集之间在图像质量、病变外观和人群特征上存在差异。尽管当前的目标检测器如YOLO和DETR在单个数据集上取得了良好的结果，但当在异构来源上进行训练或应用时，其性能往往会下降。为了解决这个问题，我们提出了MammoMix，这是一种基于专家混合（Mixture-of-Experts, MoE）范式的新框架，用于稳健且具有普适性的病变检测。在MammoMix中，每个专家模型在特定领域上进行训练，使其能够专注于源数据的不同特征。一个门控机制根据输入图像自适应地加权每个专家的贡献，将它们的输出结合起来，以实现领域自适应推理。为了提高可靠性，我们进一步结合了一个校准模块MoCAE，该模块调整置信度分数以反映真实的预测不确定性。我们在三个公共乳腺X线检查数据集上评估了MammoMix：CSAW、DDSM和DMID，涵盖了不同的临床环境。结果表明，MammoMix在平均精度和可靠性方面均优于基线检测器，尤其是在变异性较大的数据集上。我们的研究结果表明，专家专门化和校准的集成融合显著增强了模型的泛化能力和稳健性。MammoMix为在现实临床领域中实现可靠的AI辅助乳腺癌筛查提供了一个有前景的步骤。
cs.CV / 28 / 2608.10439
Stream Forcing: Constructing Unified Training Trajectory for Robust Streaming Video Generation
流强制：构建统一的训练轨迹以实现稳健的流媒体视频生成
Abstract
Streaming video generation holds strong potential for world modeling, where future frames must be inferred online sequentially to form a continuous video stream. However, streaming video diffusion models introduce a fundamental train-inference mismatch: inference follows a specialized denoising order, whereas advanced training strategies typically require diverse noise-level configurations. To address this trade-off between train-inference consistency and training coverage, we reformulate the video diffusion sampling as a frame-indexed stochastic process over noise levels. Within this stochastic process space, we construct a continuous training trajectory along which the sampling schedule progressively evolves from independent sampling to inference-consistent sampling. We further introduce a joint calibration algorithm and a temporal correlative sampling algorithm to ensure trajectory smoothness and cross-frame correlation. Building on these designs, we propose Stream Forcing, a unified training framework for streaming video generation that balances training sufficiency and inference efficiency. Extensive experiments demonstrate that Stream Forcing significantly improves generation quality with a 36.6% FVD improvement on the UCF-101 benchmark. Furthermore, our method facilitates robust zero-shot extrapolation to long-horizon video generation with a 27.9% FVD improvement on the UCF-101 benchmark.
Chinese Translation
流媒体视频生成在世界建模中具有强大的潜力，其中未来帧必须在线顺序推断，以形成连续的视频流。然而，流媒体视频扩散模型引入了一个基本的训练-推断不匹配：推断遵循专门的去噪顺序，而先进的训练策略通常需要多样的噪声水平配置。为了解决训练-推断一致性与训练覆盖之间的权衡，我们将视频扩散采样重新表述为一个关于噪声水平的帧索引随机过程。在这个随机过程空间中，我们构建了一条连续的训练轨迹，沿着这条轨迹，采样计划逐步从独立采样演变为推断一致的采样。我们进一步引入了一种联合校准算法和一种时间相关采样算法，以确保轨迹的平滑性和跨帧相关性。在这些设计的基础上，我们提出了流强制（Stream Forcing），这是一个用于流媒体视频生成的统一训练框架，平衡了训练充分性和推断效率。大量实验表明，流强制显著提高了生成质量，在UCF-101基准上实现了36.6%的FVD提升。此外，我们的方法促进了对长时间视频生成的稳健零-shot外推，在UCF-101基准上实现了27.9%的FVD提升。
cs.CV / 29 / 2608.10442
FUSE: Frame-Unified Stress Estimation from Facial Video
FUSE：基于面部视频的帧统一压力估计
Abstract
Automatic stress detection from facial video offers a practical path to non-intrusive affect monitoring, yet existing video-based approaches commonly decompose full recordings into short temporal windows before classification. This design introduces additional choices regarding window length, overlap, and aggregation, while limiting direct analysis of temporal information across the entire recording. In this study, we present FUSE (Frame-Unified Stress Estimation), a facial-video stress detection framework that processes complete recordings as a single input without temporal windowing or external segmentation. The name reflects the defining operation of the method: rather than dividing a recording into short clips, all frames are fused into one unified two-dimensional representation from which the stress state is estimated. This unification is realized by folding the temporal dimension into the channel dimension of the spatial representation, and the resulting high-dimensional input is processed using a unified asymmetric-attention architecture. At a temporal stride of t = 1, FUSE retains the full 120-second recording as one input, corresponding to 3,600 frames at 30 fps. Experiments on a 58-subject stress dataset using a stratified subject-level protocol evaluate seven temporal-stride configurations, ranging from full-frame input to sparse subsampling. FUSE achieves the highest test accuracy of 69.44% at t = 15, while the full-frame configuration remains competitive at 69.03%. Across the stride range, computational cost varies from 12.48 to 348.78 GFLOPs, showing the trade-off between temporal density and efficiency. These results demonstrate that temporal windowing is not required for effective facial-video stress detection in this setting, and that complete-recording inference can be achieved within a single unified architecture.
Chinese Translation
从面部视频中自动检测压力为非侵入式情感监测提供了一条实用路径，但现有的视频基础方法通常在分类之前将完整录音分解为短时间窗口。这种设计引入了关于窗口长度、重叠和聚合的额外选择，同时限制了对整个录音的时间信息的直接分析。在本研究中，我们提出了FUSE（Frame-Unified Stress Estimation），一种面部视频压力检测框架，它将完整录音作为单一输入进行处理，而无需时间窗口或外部分割。这个名称反映了该方法的定义操作：与其将录音分割成短片段，不如将所有帧融合成一个统一的二维表示，从中估计压力状态。这种统一是通过将时间维度折叠到空间表示的通道维度中实现的，得到的高维输入使用统一的非对称注意力架构进行处理。在时间步长t = 1时，FUSE将完整的120秒录音保留为一个输入，相当于以30 fps的速度生成的3600帧。对58个受试者的压力数据集进行的实验，使用分层受试者级协议评估了七种时间步长配置，从全帧输入到稀疏子采样。FUSE在t = 15时达到了最高的测试准确率69.44%，而全帧配置的准确率为69.03%，仍然具有竞争力。在步长范围内，计算成本从12.48到348.78 GFLOPs不等，显示了时间密度与效率之间的权衡。这些结果表明，在这种设置下，进行有效的面部视频压力检测不需要时间窗口，并且可以在单一统一架构内实现完整录音的推断。
cs.CV / 30 / 2608.10479
Bridging Event Streams and DiT: Event-Guided Video Frame Interpolation
桥接事件流与DiT：事件引导的视频帧插值
Abstract
Latent diffusion models have recently advanced video frame interpolation by synthesizing intermediate frames between input images. However, handling large temporal gaps and complex motion remains challenging, often resulting in motion blur, structural distortions, and temporal inconsistencies. Event cameras provide high-temporal-resolution motion cues that are well suited for bridging these gaps and improving interpolation quality. To exploit this advantage without training an event-assisted model from scratch, we propose an adapter-based framework that incorporates event-derived cues into a pre-trained image-to-video diffusion model with minimal architectural changes. Specifically, our method leverages Image Warped Events (IWEs) and bidirectional sparse optical flow to provide spatially and temporally aligned guidance during generation. By injecting these event-guided structural and motion cues into the diffusion process, our approach reduces interpolation artifacts and improves both reconstruction fidelity and temporal coherence. Experimental results on real and synthetic benchmarks show that our method consistently outperforms existing state-of-the-art approaches. The project page is at https://joseph-lin-tech.github.io/BridgeEventDiT-VFI/.
Chinese Translation
潜在扩散模型最近通过合成输入图像之间的中间帧，推动了视频帧插值的发展。然而，处理较大的时间间隔和复杂的运动仍然具有挑战性，常常导致运动模糊、结构失真和时间不一致性。事件相机提供了高时间分辨率的运动线索，非常适合填补这些间隙并提高插值质量。为了利用这一优势而不必从头开始训练一个事件辅助模型，我们提出了一种基于适配器的框架，该框架将事件衍生的线索融入到一个预训练的图像到视频扩散模型中，且对架构的更改最小。具体而言，我们的方法利用了图像扭曲事件（Image Warped Events, IWEs）和双向稀疏光流，在生成过程中提供空间和时间对齐的指导。通过将这些事件引导的结构和运动线索注入扩散过程中，我们的方法减少了插值伪影，并提高了重建的保真度和时间一致性。在真实和合成基准上的实验结果表明，我们的方法始终优于现有的最先进方法。项目页面地址为 https://joseph-lin-tech.github.io/BridgeEventDiT-VFI/.
cs.CV / 31 / 2608.10489
When Vision Becomes Text: Visual Token Pruning via Cross-Modal Residual Guidance in VLMs
当视觉转化为文本：在视觉语言模型中通过跨模态残差引导进行视觉标记修剪
Abstract
Abundant visual information strengthens vision-language model (VLM) perception, yet massive visual tokens raise inference costs. Existing visual token pruning methods rely on similarity-based guidance, which exploits pairwise text-vision and vision-vision token correlations for compression. However, such methods only capture local layer-level signals and overlook the whole inference process in VLM. In this paper, we revisit VLM inference and present a new efficient guidance scheme that complements similarity-based guidance. In particular, we identify a key observation: as LLM layers deepen, text tokens continuously aggregate visual information via self-attention and progressively absorb partial visual content into textual representations. To quantify this phenomenon, we propose Cross Modal Absorption (CMA) from a geometric representation perspective to measure how much visual information is absorbed by text, revealing that more visual tokens in deeper layers can be approximately explained by the text subspace. We accordingly propose Cross Modal Residual (CMR). It projects visual tokens onto the text subspace via Tikhonov regularized least squares and exploits reconstruction residuals to quantify visual information that cannot be explained by text. Finally, based on CMR, we present SIEVE, a training-free visual token compression method that combines CMR, text-attention relevance, and residual-space diversity to retain task-relevant and complementary tokens. Experiments on diverse VLM architectures verify the effectiveness of SIEVE. For instance, on LLaVA-NeXT-7B, SIEVE keeps only $11.1\%$ of visual tokens while preserving $97.5\%$ of the original average performance, achieving $3.62\times$ prefill speedup, $2.49\times$ end-to-end speedup, and a $6.02\times$ KV-cache reduction.
Chinese Translation
丰富的视觉信息增强了视觉语言模型（VLM）的感知能力，但大量的视觉标记提高了推理成本。现有的视觉标记修剪方法依赖于基于相似性的引导，这种方法利用成对的文本-视觉和视觉-视觉标记相关性进行压缩。然而，这些方法仅捕捉局部层级信号，忽视了VLM中的整个推理过程。本文重新审视了VLM推理，并提出了一种新的高效引导方案，以补充基于相似性的引导。特别地，我们识别出一个关键观察：随着大规模语言模型（LLM）层数的加深，文本标记通过自注意力不断聚合视觉信息，并逐步将部分视觉内容吸收到文本表示中。为了量化这一现象，我们从几何表示的角度提出了跨模态吸收（Cross Modal Absorption, CMA），以测量文本吸收了多少视觉信息，揭示了在更深层次中，更多的视觉标记可以通过文本子空间进行近似解释。因此，我们提出了跨模态残差（Cross Modal Residual, CMR）。它通过Tikhonov正则化最小二乘法将视觉标记投影到文本子空间，并利用重构残差量化无法通过文本解释的视觉信息。最后，基于CMR，我们提出了SIEVE，这是一种无训练的视觉标记压缩方法，结合了CMR、文本注意力相关性和残差空间多样性，以保留与任务相关和互补的标记。在多种VLM架构上的实验验证了SIEVE的有效性。例如，在LLaVA-NeXT-7B上，SIEVE仅保留了$11.1 ext{%}$的视觉标记，同时保持了$97.5 ext{%}$的原始平均性能，实现了$3.62 imes$的预填充加速、$2.49 imes$的端到端加速，以及$6.02 imes$的KV缓存减少。
cs.CV / 32 / 2608.10497
SapiensID 2.0: Aligning Human Recognition Foundation Models with Human Perception
SapiensID 2.0：将人类识别基础模型与人类感知对齐
Abstract
While foundation models have significantly advanced human recognition across diverse modalities, they predominantly rely on static, geometric feature extraction. This approach fundamentally diverges from human perception. Consequently, current models often suffer from "semantic blindness," overfitting to transient noise while failing to leverage invariant soft biometrics, and struggle to capture temporal motion signatures. To bridge this gap, we propose SapiensID 2.0, a human recognition framework enriched with both semantic and temporal awareness. To overcome the lack of soft-biometric annotations, we transfer zero-shot semantic knowledge from Multimodal Large Language Models (MLLMs) into a discriminative embedding space. We resolve the dimensional mismatch between these spaces using Invariant Trait Alignment (ITA) to distill core persistent traits, and Transient Noise Disentanglement (TND) to decouple artifacts like clothing. Furthermore, we design a Kinematic Semantic Attention Head (K-SAH) that extends spatial attention across temporal windows. By tracking semantic patches over time, K-SAH captures rich kinematic signatures without requiring large-scale video datasets. Extensive experiments demonstrate that SapiensID 2.0 achieves state-of-the-art performance across image- and video-based person re-identification and gait recognition, while maintaining robust face recognition capabilities.
Chinese Translation
尽管基础模型在多种模态下显著推动了人类识别的发展，但它们主要依赖于静态的几何特征提取。这种方法与人类感知存在根本性差异。因此，当前模型常常遭遇“语义盲目”，过度拟合瞬时噪声，未能有效利用不变的软生物特征，并且在捕捉时间运动特征方面表现不佳。为了解决这一问题，我们提出了SapiensID 2.0，一个融合了语义和时间意识的人类识别框架。为了克服软生物特征注释的缺乏，我们将来自多模态大语言模型（Multimodal Large Language Models, MLLMs）的零样本语义知识转移到一个区分性嵌入空间。我们使用不变特征对齐（Invariant Trait Alignment, ITA）来解决这些空间之间的维度不匹配，以提炼核心持久特征，并通过瞬时噪声解耦（Transient Noise Disentanglement, TND）来分离诸如服装等伪影。此外，我们设计了一个运动语义注意头（Kinematic Semantic Attention Head, K-SAH），它在时间窗口中扩展空间注意力。通过随时间跟踪语义补丁，K-SAH能够捕捉丰富的运动特征，而无需大规模视频数据集。大量实验表明，SapiensID 2.0在基于图像和视频的人物重识别及步态识别方面达到了最先进的性能，同时保持了强大的面部识别能力。
cs.CV / 33 / 2608.10500
DSAR: Dual-Stream Autoregressive Modeling of Temporal Cloth Dynamics for Photorealistic Animatable Avatars
DSAR：用于照片级真实感可动画化虚拟形象的双流自回归建模时间布料动态
Abstract
Creating photorealistic and temporally coherent animatable human avatars from RGB videos remains challenging. Current methods struggle to capture realistic cloth dynamics, producing over-smoothed appearance or severe artifacts on out-of-distribution poses. This limitation stems from a fundamental oversight: existing approaches neglect the temporal causality inherent in cloth physics, where current states emerge from previous states through temporal evolution rather than instantaneous skeletal configurations alone. Without explicit modeling of this causal structure, networks learn pose-appearance correlations instead of motion evolution, leading to poor generalization. We introduce a dual-stream autoregressive framework that explicitly models both observable geometric information and implicit internal state. The geometric stream propagates surface displacement from the previous frame, while the state stream fuses current features with historical states retrieved from a memory bank. Motion-adaptive aggregation handles spatially-varying dynamics, and adaptive regularization balances smoothness with flexibility. Experiments on challenging datasets demonstrate significant improvements in rendering quality, temporal consistency, and generalization to motion patterns beyond training distributions, validating that dual-stream temporal modeling enables realistic cloth dynamics.
Chinese Translation
从RGB视频中创建照片级真实感和时间一致的可动画化人类虚拟形象仍然面临挑战。当前方法在捕捉真实布料动态方面存在困难，导致过于平滑的外观或在分布外姿态上出现严重伪影。这一限制源于一个根本性的疏忽：现有方法忽视了布料物理中固有的时间因果关系，其中当前状态是通过时间演变从先前状态中产生的，而不仅仅是瞬时骨骼配置。没有明确建模这种因果结构，网络学习的是姿态与外观的相关性，而不是运动演变，导致泛化性能差。我们提出了一种双流自回归框架，明确建模可观察的几何信息和隐含的内部状态。几何流从前一帧传播表面位移，而状态流则将当前特征与从记忆库中检索的历史状态融合。运动自适应聚合处理空间变化的动态，而自适应正则化则在平滑性与灵活性之间取得平衡。在具有挑战性的数据集上的实验表明，在渲染质量、时间一致性以及对超出训练分布的运动模式的泛化能力上显著改善，验证了双流时间建模能够实现真实的布料动态。
cs.CV / 34 / 2608.10512
Towards Color-Faithful Low-Light Image Enhancement via Adaptive Color Debiasing and Saturation Rectification
通过自适应色彩去偏和饱和度校正实现色彩真实的低光图像增强
Abstract
Low-light imaging often introduces color bias caused by the low signal-to-noise ratio and the image formation process. Although recent low-light image enhancement methods have achieved strong brightness recovery, faithful color restoration remains challenging, manifesting as overall color bias together with local under- and over-saturation. To address this issue, we propose CAGE, a cylindrical color correction framework with adaptive color debiasing and gamut-harmonized saturation rectification for color-faithful low-light image enhancement. We first introduce AdaLAB, a cylindrical adaptive LAB color space that provides a decoupled and image-specific basis for uniform color correction. Building on this color space, we further develop AdaCCT, an adaptive cylindrical color transform with forward and inverse transforms for the conversion between RGB and AdaLAB color space, as well as necessary color debiasing and saturation rectification. The forward transform suppresses embedded color bias before backbone enhancement by reorganizing the chromatic distribution through chromatic-plane shifting and scaling, while the inverse transform achieves faithful saturation rectification through out-of-gamut lightness compensation. Extensive experiments on multiple benchmarks show that CAGE achieves more faithful color restoration, specifically reduces color bias and saturation abnormality, and delivers better overall visual quality across different low-light enhancement backbones. The code is available at https://yangzhichen763.github.io/CAGE/.
Chinese Translation
低光成像常常由于低信噪比和图像形成过程引入色彩偏差。尽管最近的低光图像增强方法在亮度恢复方面取得了显著进展，但真实的色彩恢复仍然面临挑战，表现为整体色彩偏差以及局部的饱和度不足和过饱和。为了解决这一问题，我们提出了CAGE，一个具有自适应色彩去偏和色域协调饱和度校正的圆柱形色彩校正框架，用于色彩真实的低光图像增强。我们首先引入AdaLAB，一个圆柱形自适应LAB色彩空间，为均匀色彩校正提供了一个解耦且特定于图像的基础。在此色彩空间的基础上，我们进一步开发了AdaCCT，一种具有前向和逆向变换的自适应圆柱形色彩变换，用于RGB与AdaLAB色彩空间之间的转换，以及必要的色彩去偏和饱和度校正。前向变换通过色彩平面移动和缩放重新组织色彩分布，从而在主干增强之前抑制嵌入的色彩偏差，而逆向变换则通过超出色域的亮度补偿实现真实的饱和度校正。在多个基准测试上的广泛实验表明，CAGE实现了更真实的色彩恢复，特别是减少了色彩偏差和饱和度异常，并在不同的低光增强主干中提供了更好的整体视觉质量。代码可在 https://yangzhichen763.github.io/CAGE/ 获取。
cs.CV / 35 / 2608.10513
SafeCap: Improving LVLM Safety with Image Captioning Reinforcement Learning
SafeCap：通过图像描述强化学习提高大型视觉语言模型的安全性
Abstract
Large vision-language models (LVLMs) remain vulnerable to jailbreak attacks that exploit visual inputs to bypass safety alignment inherited from their language backbones. We propose SafeCap, a reinforcement-learning framework that aligns LVLMs through learned self-captioning. SafeCap trains a policy model to first generate a safety-relevant image caption and then produce a final answer; the caption is further optimized by whether it enables a frozen LLM to reach a safety-aligned decision. This caption-mediated objective encourages the policy to expose visual cues relevant to safe response generation rather than relying solely on direct refusal supervision. Across five multimodal safety benchmarks and six vision-utility benchmarks, SafeCap substantially improves aggregate safety performance under its intended DirectCap protocol, with gains of 3.7-19.0 points in safety average across four model settings while maintaining comparable or improved vision utility. Under controlled comparisons on matched backbones and data, SafeCap outperforms safety SFT, DPO, and SafeGRPO, demonstrating the effectiveness of caption-mediated reinforcement learning for multimodal safety alignment.
Chinese Translation
大型视觉语言模型（LVLMs）仍然容易受到越狱攻击，这些攻击利用视觉输入绕过其语言基础模型所继承的安全对齐。我们提出了SafeCap，一个通过学习自我描述来对齐LVLM的强化学习框架。SafeCap训练一个策略模型，首先生成与安全相关的图像描述，然后产生最终答案；该描述进一步通过是否使一个冻结的语言模型（LLM）达到安全对齐的决策来优化。这个通过描述介导的目标鼓励策略揭示与安全响应生成相关的视觉线索，而不是仅仅依赖于直接拒绝监督。在五个多模态安全基准和六个视觉效用基准上，SafeCap在其预期的DirectCap协议下显著提高了整体安全性能，在四个模型设置中安全平均值提高了3.7-19.0分，同时保持了可比或改进的视觉效用。在对匹配的基础模型和数据进行的控制比较中，SafeCap的表现优于安全的SFT、DPO和SafeGRPO，证明了通过描述介导的强化学习在多模态安全对齐中的有效性。
cs.CV / 36 / 2608.10519
SparSTAR: Sparse Attention for SpaceTime AutoRegressive Video Synthesis
SparSTAR：用于时空自回归视频合成的稀疏注意力
Abstract
InfinityStar extends visual autoregressive generation to video through a sequence of image and clip pyramids. Its changing scale and cross-clip context, however, leave late-scale attention costly and make sparse patterns reused from diffusion or image VAR models unreliable. We introduce SparSTAR, a training-free block-sparse attention method tailored to this setting. At each expensive scale and attention head, SparSTAR scores contiguous key blocks from the current query and key activations, retains required conditioning context, and executes the selected blocks through a forward-only sparse path. We analyze cross-scale consistency within a clip, pattern persistence across clip boundaries, and quality degradation as reuse spans increasingly distant scales. Across these analyses, important key blocks shift, showing that recomputing block selection at each target scale is more reliable than reusing a transferred mask. On 720p text-to-video and image-to-video generation, SparSTAR preserves every token and refinement scale while providing about a 1.6x end-to-end speedup and maintaining VBench and paired-output reconstruction fidelity close to dense InfinityStar.
Chinese Translation
InfinityStar通过一系列图像和片段金字塔将视觉自回归生成扩展到视频。然而，其变化的尺度和跨片段上下文使得晚期尺度的注意力成本高昂，并且从扩散或图像VAR模型中重用的稀疏模式不可靠。我们提出了SparSTAR，这是一种针对该设置的无训练块稀疏注意力方法。在每个高成本的尺度和注意力头上，SparSTAR从当前查询和关键激活中评分连续的关键块，保留所需的条件上下文，并通过仅向前的稀疏路径执行所选块。我们分析了片段内的跨尺度一致性、跨片段边界的模式持久性，以及当重用跨度越来越远的尺度时的质量下降。在这些分析中，重要的关键块发生了变化，表明在每个目标尺度重新计算块选择比重用转移的掩码更可靠。在720p文本到视频和图像到视频生成中，SparSTAR保留了每个标记和细化尺度，同时提供约1.6倍的端到端加速，并保持VBench和配对输出重建的保真度接近于密集的InfinityStar。
cs.CV / 37 / 2608.10522
Unlocking the Power of Medical Tabular Data via Semantic-Aware Multimodal Pre-training
通过语义感知的多模态预训练释放医疗表格数据的潜力
Abstract
While vision-language models dominate medical representation learning, unstructured text lacks the dense, quantitative diagnostic phenotypes inherent in structured clinical tables. However, existing multimodal pre-training methods underutilize this potential due to semantic-agnostic designs that treat tabular inputs as flat vectors and employ unstable continuous regression objectives. To overcome this, we propose a novel semantic-aware framework explicitly modeling the intrinsic two-dimensional structure of tabular data. First, addressing the inter-feature hierarchy of varying diagnostic importance, we introduce Importance-Aware Adaptive Masking to construct a label-free curriculum prioritizing salient features. Second, addressing the intra-feature continuity-discreteness duality, we propose a Soft-Label Discretized Module that replaces unstable numerical regression with stable distribution matching, thereby mathematically preserving ordinal relationships. Extensive experiments across large-scale dermatology (SLICE-3D, HOP) and ophthalmology (EyePACS) datasets establish a new state-of-the-art (SOTA), demonstrating exceptional robustness and cross-domain generalizability.
Chinese Translation
尽管视觉-语言模型在医疗表征学习中占据主导地位，但非结构化文本缺乏结构化临床表格中固有的密集、定量的诊断表型。然而，现有的多模态预训练方法由于采用了语义无关的设计，将表格输入视为平面向量，并使用不稳定的连续回归目标，未能充分利用这一潜力。为此，我们提出了一种新颖的语义感知框架，明确建模表格数据的内在二维结构。首先，针对不同诊断重要性的特征间层次关系，我们引入了重要性感知自适应掩蔽，构建了一个无标签的课程，优先考虑显著特征。其次，针对特征内部的连续性-离散性二元性，我们提出了软标签离散化模块，用稳定的分布匹配替代不稳定的数值回归，从而在数学上保留序关系。在大规模皮肤病学（SLICE-3D, HOP）和眼科学（EyePACS）数据集上的广泛实验建立了新的最先进水平（SOTA），展示了卓越的鲁棒性和跨领域的泛化能力。
cs.CV / 38 / 2608.10524
Rethinking Text-Based Image Retrieval in Specific Domain
重新思考特定领域中的基于文本的图像检索
Abstract
Driven by the rapid advancement of vision-language representation learning, Text-based Image Retrieval (TBIR) has made notable progress. However, existing benchmarks are predominantly constructed on an exclusive single-match assumption between query and images. While effective in general scenarios, this assumption fails to reflect practical system performance in specific domains (e.g., surveillance), where a single query often corresponds to multiple relevant candidate images. To address this limitation, we design a Domain-Specific Multi-Match Text-based Image Retrieval (DSMM-TBIR) data engine. Leveraging this engine, we construct Security Multi-Match TBIR (SecMM-TBIR), a benchmark comprising 50k surveillance images with 200 comprehensive queries. Furthermore, we observe that vanilla contrastive learning in specific domains suffers from severe false negatives, forcing the model to push apart semantically similar pairs and thus degrading retrieval performance. We propose the Semantic-Aware Fine-Tuning (SAFT) framework to address semantic compression in specific domains, which incorporates Semantic-Aware Soft-Label Supervision (SASS) and Intra-modal Structural Distillation (ISD) to establish a promising paradigm for domain-specific TBIR tasks. Experiments across diverse CLIP-like models demonstrate that SAFT yields an average mAP@20 gain of 7.8 points on SecMM-TBIR over standard image-text contrastive (ITC) fine-tuning, while also improving general-domain performance. The entire benchmark will be released to facilitate further research.
Chinese Translation
随着视觉-语言表示学习的快速发展，基于文本的图像检索（Text-based Image Retrieval, TBIR）取得了显著进展。然而，现有的基准测试主要基于查询与图像之间的单一匹配假设。虽然在一般场景中有效，但这一假设未能反映特定领域（例如监控）中实际系统性能，其中单个查询通常对应多个相关候选图像。为了解决这一局限性，我们设计了一个特定领域的多匹配基于文本的图像检索（Domain-Specific Multi-Match Text-based Image Retrieval, DSMM-TBIR）数据引擎。利用该引擎，我们构建了安全多匹配TBIR（Security Multi-Match TBIR, SecMM-TBIR）基准，包含50,000张监控图像和200个全面的查询。此外，我们观察到在特定领域中，普通对比学习存在严重的假阴性问题，迫使模型将语义相似的对推开，从而降低检索性能。我们提出了语义感知微调（Semantic-Aware Fine-Tuning, SAFT）框架，以解决特定领域中的语义压缩问题，该框架结合了语义感知软标签监督（Semantic-Aware Soft-Label Supervision, SASS）和模态内结构蒸馏（Intra-modal Structural Distillation, ISD），为特定领域的TBIR任务建立了一个有前景的范式。针对多种类似CLIP的模型的实验表明，SAFT在SecMM-TBIR上相较于标准图像-文本对比（Image-Text Contrastive, ITC）微调平均提升了7.8个点的mAP@20，同时也改善了通用领域的性能。整个基准将被发布以促进进一步研究。
cs.CV / 39 / 2608.10525
Dynamic Context Adapters: Efficiently Infusing History into Vision-and-Language Models
动态上下文适配器：高效地将历史信息融入视觉-语言模型
Abstract
Historical context integration presents a fundamental challenge for Vision-Language Models (VLMs) in sequential decision-making tasks. Current VLMs process visual inputs independently, which creates critical limitations for downstream applications that require temporal understanding. Direct incorporation of historical frames into Transformer inputs produces quadratic attention complexity and excessive memory consumption. Existing approaches suffer from significant drawbacks: computational inflation or substantial information loss through temporal compression. To address these challenges, we introduce Dynamic Context Adapter (DCA), a novel context injection approach for pretrained VLMs. Our method employs fixed-size, dynamically compressed memory to preserve historical semantics without frame concatenation. DCA bridges static VLMs and recurrent policies and enables memory capabilities in pretrained models while maintaining computational efficiency. DCA achieves over $25\%$ reduction in attention FLOPs and $13\%$ memory savings while improving performance on long-horizon tasks.
Chinese Translation
历史上下文的整合是视觉-语言模型（VLMs）在顺序决策任务中面临的一个基本挑战。目前的VLMs独立处理视觉输入，这对需要时间理解的下游应用造成了严重限制。将历史帧直接纳入Transformer输入会导致二次注意力复杂度和过高的内存消耗。现有方法存在显著缺陷：计算膨胀或通过时间压缩造成大量信息损失。为了解决这些挑战，我们提出了动态上下文适配器（Dynamic Context Adapter, DCA），这是一种针对预训练VLMs的新型上下文注入方法。我们的方法采用固定大小、动态压缩的内存，以在不进行帧连接的情况下保留历史语义。DCA在静态VLMs和递归策略之间架起了桥梁，使预训练模型具备内存能力，同时保持计算效率。DCA在注意力FLOPs上实现了超过25%的减少，并节省了13%的内存，同时在长时间任务上提高了性能。
cs.CV / 40 / 2608.10544
Flow Straight to Reality: Perceptually Consistent Flow Matching for Efficient Image Restoration
直达现实的流动：用于高效图像恢复的感知一致性流匹配
Abstract
Image restoration is fundamentally constrained by the tradeoff between distortion and perception: minimizing pixel-wise error yields over-smoothed results, whereas optimizing for perceptual realism often introduces structural deviations. Recent approaches attempt to balance this tradeoff via posterior sampling or multi-stage generative pipelines, yet remain computationally expensive and architecturally complex. To overcome these limitations, we propose PCFlow (Perceptually Consistent Flow Matching), a unified framework that directly parameterizes a continuous transport from degraded observations to clean targets, jointly optimizing distortion and perceptual quality. While its latent consistency flow objective drives stable and efficient few-step inference, a Latent Consistency Perceptual Loss (LCPL) imposes semantic constraints directly on the guiding velocity field, steering the dynamics toward visually sharp data manifolds. Furthermore, recognizing the inherent conflict between structural and perceptual consistencies, we integrate a conflict-free gradient projection strategy to stabilize the multi-objective optimization landscape. Combined with lightweight, convolution-only backbone, PCFlow achieves competitive performance across diverse restoration tasks at a fraction of traditional computational costs.
Chinese Translation
图像恢复在失真与感知之间的权衡上受到根本限制：最小化像素级误差会导致过度平滑的结果，而优化感知真实感往往会引入结构偏差。最近的方法试图通过后验采样或多阶段生成管道来平衡这一权衡，但仍然计算成本高且架构复杂。为克服这些限制，我们提出了PCFlow（感知一致性流匹配），这是一个统一框架，直接参数化从退化观测到干净目标的连续传输，联合优化失真和感知质量。尽管其潜在一致性流目标驱动稳定且高效的少步推理，但潜在一致性感知损失（LCPL）直接对引导速度场施加语义约束，引导动态朝向视觉上清晰的数据流形。此外，鉴于结构一致性与感知一致性之间固有的冲突，我们整合了一种无冲突的梯度投影策略，以稳定多目标优化的景观。结合轻量级的仅卷积主干，PCFlow在多种恢复任务中以传统计算成本的一小部分实现了竞争力的性能。
cs.CV / 41 / 2608.10588
A HamNoSys-Guided Dataset and Baselines for Fine-Grained Isolated Handshape Recognition in Sign Language
基于HamNoSys的细粒度孤立手形识别数据集及基准
Abstract
Purpose: Fine-grained handshape recognition supports computational sign-language transcription, recognition, and translation, but broad, phonetically defined visual inventories with signer-aware evaluation remain limited. This work introduces a benchmark grounded in the language-independent Hamburg Notation System (HamNoSys). Methods: A balanced dataset of 144,000 RGB images was collected from 15 participants for 160 handshape classes defined by the official HamNoSys 4 Handshapes Chart. ResNet-18 and ViT-B/16 were evaluated as appearance-based models, while a graph convolutional network and XGBoost were evaluated from hand landmarks. Both a class-stratified subject-dependent split and a 15-fold leave-one-subject-out (LOSO) protocol were used. The same model families were additionally assessed on LSWH100 and ASL Fingerspelling Dataset A for external context. Results: The subject-dependent benchmarks established reproducible reference performance across all four model families, whereas LOSO evaluation exposed a substantial reduction when recognition was required to generalise to unseen participants. On ASL Fingerspelling Dataset A, mean LOSO top-1 accuracy ranged from 82.20% to 87.40%. Conclusion: The documented acquisition, curation, and complementary evaluation protocols pro-vide a reproducible resource for fine-grained isolated-handshape research and for developing more accessible sign-language technologies.
Chinese Translation
目的：细粒度手形识别支持计算机手语转录、识别和翻译，但基于广泛的、音位定义的视觉库以及考虑到手语者的评估仍然有限。本研究介绍了一个以语言无关的汉堡符号系统（HamNoSys）为基础的基准。方法：从15名参与者收集了144,000张RGB图像，涵盖官方HamNoSys 4手形图表定义的160个手形类别。评估了ResNet-18和ViT-B/16作为基于外观的模型，同时从手部特征点评估了图卷积网络和XGBoost。使用了类分层的受试者依赖分割和15折留一法（LOSO）协议。此外，还在LSWH100和ASL Fingerspelling Dataset A上评估了相同的模型系列，以获取外部背景。结果：受试者依赖的基准在所有四个模型系列中建立了可重复的参考性能，而LOSO评估则暴露出当识别需要推广到未见参与者时的显著减少。在ASL Fingerspelling Dataset A上，平均LOSO top-1准确率范围为82.20%到87.40%。结论：所记录的获取、策划和补充评估协议为细粒度孤立手形研究提供了可重复的资源，并为开发更易于访问的手语技术奠定了基础。
cs.CV / 42 / 2608.10589
$\pi$-SUB: A Physics-Informed Synthetic Underwater Benchmark Dataset for Underwater Image Enhancement
π-SUB：一种物理信息驱动的合成水下基准数据集用于水下图像增强
Abstract
This paper presents $\pi$-SUB, a physics-informed framework for generating synthetic underwater benchmark datasets that bridges the synthetic-to-real gap for Underwater Image Enhancement (UIE). The proposed framework extends the classical underwater image formation model by incorporating depth-dependent downwelling irradiance, biologically resolved absorption, and environmental scattering across all ten Jerlov water types, together with independently controllable residual phenomena. Using this framework, the $\pi$-SUB dataset consists of paired synthetic underwater-reference images spanning shallow-to-deep and coastal-to-oceanic environments. Extensive simulation studies have been carried out to evaluate $\pi$-SUB along two criteria namely hyper-realism and generalizability. For hyper-realism, $\pi$-SUB attains a global Frechet Inception Distance (FID) that is 46% lower than Syrea. For generalizability, four state-of-the-art UIE architectures (FUnIE-GAN, Pix2Pix, PUIE-Net, and Phaseformer) are used for comparative evaluation of $\pi$-SUB. These models were independently trained on six datasets including one real and five synthetic datasets and tested on six real-world benchmarks datasets. Across four UIE architectures and six real benchmark datasets, $\pi$-SUB improves UIQM by 4.18% over PHISWID (next best) and 9.46% over Syrea (next best), while reducing NIQE by 48.78% and 23.98%, respectively. These results establish $\pi$-SUB as a hyper-realistic and generalizable benchmark for developing the next generation of underwater image enhancement methods. The code and dataset are available at https://github.com/airl-iisc/pi-SUB
Chinese Translation
本文提出了π-SUB，一个物理信息驱动的框架，用于生成合成水下基准数据集，以弥合水下图像增强（UIE）中的合成与真实之间的差距。所提出的框架通过结合深度依赖的下行辐照度、生物解析的吸收和覆盖所有十种Jerlov水类型的环境散射，扩展了经典的水下图像形成模型，并独立控制残余现象。利用该框架，π-SUB数据集由成对的合成水下-参考图像组成，涵盖了从浅水到深水以及从沿海到海洋的环境。我们进行了广泛的仿真研究，以评估π-SUB在超现实主义和可推广性两个标准下的表现。在超现实主义方面，π-SUB的全局Frechet Inception Distance (FID)比Syrea低46%。在可推广性方面，使用四种最先进的UIE架构（FUnIE-GAN、Pix2Pix、PUIE-Net和Phaseformer）对π-SUB进行了比较评估。这些模型在六个数据集上独立训练，包括一个真实数据集和五个合成数据集，并在六个真实世界基准数据集上进行了测试。在四种UIE架构和六个真实基准数据集上，π-SUB在UIQM上比PHISWID（次优）提高了4.18%，比Syrea（次优）提高了9.46%，同时分别降低了NIQE 48.78%和23.98%。这些结果确立了π-SUB作为开发下一代水下图像增强方法的超现实和可推广基准。代码和数据集可在https://github.com/airl-iisc/pi-SUB获取。
cs.CV / 43 / 2608.10590
Rethinking Data Efficiency in Industrial Dense Prediction: Pretraining Coherence, Not Inductive Bias, Determines ViTs Low-Data Advantage
重新思考工业密集预测中的数据效率：预训练一致性，而非归纳偏置，决定了ViTs的低数据优势
Abstract
Vision Transformers (ViTs) are widely believed to require more labeled data than CNNs for industrial dense prediction. Through controlled experiments on four industrial datasets, we show that the data-efficiency gap stems from pretraining incoherence, which refers to the statistical mismatch between ImageNet-pretrained ViT backbones and COCO-pretrained CNN necks, rather than from inherent self-attention deficits. We characterize the cross-architecture feature gap and propose a lightweight AlignBlock family for pyramid-level feature recalibration. Our core finding empirically identifies a data-efficiency frontier: for domain-proximal scenes with >= 200 samples, Swin-Graft surpasses YOLOv11x (terminal 703-shot: 0.973 vs 0.956 mAP@50); for domain-distant scenes, CNNs retain advantage (hook 141-shot: 0.900 vs 0.600 mAP@50). Grafted neck weights yield up to 2.5x the mAP of a randomly initialized neck.
Chinese Translation
视觉变换器（ViTs）被广泛认为在工业密集预测中需要比卷积神经网络（CNNs）更多的标记数据。通过对四个工业数据集的控制实验，我们表明数据效率差距源于预训练不一致性，这指的是ImageNet预训练的ViT主干与COCO预训练的CNN颈部之间的统计不匹配，而不是固有的自注意力缺陷。我们描述了跨架构特征差距，并提出了一种轻量级的AlignBlock系列用于金字塔级特征的重新校准。我们的核心发现实证识别了一个数据效率前沿：对于样本数>= 200的领域邻近场景，Swin-Graft超越了YOLOv11x（终端703-shot：0.973 vs 0.956 mAP@50）；而对于领域远离的场景，CNNs仍然保持优势（hook 141-shot：0.900 vs 0.600 mAP@50）。接入的颈部权重可实现高达2.5倍的mAP，相比于随机初始化的颈部。
cs.CV / 44 / 2608.10602
Gaussian Sculpting: End-to-End Controllable Surface Reconstruction via Field Optimization
高斯雕刻：通过场优化实现端到端可控的表面重建
Abstract
3D Gaussian Splatting (3DGS) has recently enabled real-time novel view synthesis with impressive quality. However, it struggles to recover accurate surfaces under limited viewpoints and due to the inherent irregularity of Gaussian primitives. The resulting geometric errors are notoriously difficult to correct manually. To address these issues, we propose Gaussian Sculpting, a fully differentiable end-to-end framework for high-quality surface reconstruction. Our key insight is to anchor Gaussians onto an evolving differentiable surface, allowing them to guide signed distance field (SDF) optimization instead of extracting the surface only during post-processing. To enable stable gradient isolation during joint optimization, we design a bi-level training strategy in which the outer loop optimizes the geometry represented by the SDF, while the inner loop updates the Gaussians with the geometry fixed. We further impose constraints on Gaussian parameters to ensure consistency with the underlying surface, thereby improving both geometric and appearance fidelity during optimization. In addition, we introduce a multi-resolution subdivision scheme based on octree-like partitioning to preserve fine details while reducing memory consumption. Experiments on object-level scenes demonstrate that our method effectively removes redundant surfaces, recovers missing structures caused by limited viewpoints, and achieves strong reconstruction quality even at relatively low resolutions.
Chinese Translation
三维高斯点云（3D Gaussian Splatting, 3DGS）最近实现了实时新视图合成，并展现出令人印象深刻的质量。然而，在有限视角下以及由于高斯原语固有的不规则性，它在恢复准确表面方面面临挑战。由此产生的几何误差通常难以手动修正。为了解决这些问题，我们提出了高斯雕刻（Gaussian Sculpting），这是一个完全可微分的端到端框架，用于高质量表面重建。我们的关键见解是将高斯锚定在一个不断演变的可微分表面上，使其能够引导有符号距离场（Signed Distance Field, SDF）优化，而不是仅在后处理阶段提取表面。为了在联合优化过程中实现稳定的梯度隔离，我们设计了一种双层训练策略，其中外层循环优化由SDF表示的几何形状，而内层循环在几何形状固定的情况下更新高斯。我们进一步对高斯参数施加约束，以确保与基础表面的一致性，从而在优化过程中提高几何和外观的保真度。此外，我们引入了一种基于八叉树分区的多分辨率细分方案，以在减少内存消耗的同时保留细节。对象级场景的实验表明，我们的方法有效去除了冗余表面，恢复了由于有限视角造成的缺失结构，并在相对较低的分辨率下实现了强大的重建质量。
cs.CV / 45 / 2608.10628
InSight-doc: Agentic Visual Perception for Long-Document Understanding
InSight-doc：用于长文档理解的自主视觉感知
Abstract
Long-document understanding often requires reasoning over many visually rich pages, making inference costly and prone to context rot. In this work, we propose InSight-doc, an agentic visual perception framework that treats visual resolution as an adaptive reasoning-time resource. InSight-doc starts from low resolution and selectively zooms into high-resolution regions for finer evidence, without relying on any external retriever. To train such an agent, we construct an active-perception corpus of 17.9K high-quality SFT examples with region-level zoom-in trajectories, accompanied by 19.2K hard RL examples. Through SFT+RL, InSight-doc-8B improves the baseline by 4.3--16.4 accuracy points over document VQA benchmarks. On long documents, it reduces hallucination by more than 40% and inference latency by 41%--68% while maintaining an accuracy lead. Our code, datasets, and model are released at https://github.com/m-Just/InSight-doc .
Chinese Translation
长文档理解通常需要对许多视觉丰富的页面进行推理，这使得推理成本高且容易出现上下文衰退。在本研究中，我们提出了InSight-doc，一种将视觉分辨率视为自适应推理时间资源的自主视觉感知框架。InSight-doc从低分辨率开始，选择性地放大高分辨率区域以获取更精细的证据，而不依赖于任何外部检索器。为了训练这样的代理，我们构建了一个包含17.9K高质量SFT示例的主动感知语料库，具有区域级的放大轨迹，并附带19.2K困难的强化学习示例。通过SFT+RL，InSight-doc-8B在文档VQA基准测试中将基线提高了4.3至16.4个准确率点。在长文档上，它将幻觉减少了40%以上，推理延迟降低了41%至68%，同时保持了准确率的领先。我们的代码、数据集和模型已发布在https://github.com/m-Just/InSight-doc。
cs.CV / 46 / 2608.10635
MedUP: Awakening Unified Understanding and Perception in Medical Vision-Language Models
MedUP：在医学视觉-语言模型中唤醒统一理解与感知
Abstract
Medical Vision-Language Models (Med-VLMs) excel at verbalizing visual content, yet precise visual perception, segmentation, and grounding remain challenging. Existing approaches either verbalize regions as coordinate strings or rely on external modules that decouple perception from understanding, creating representation gaps for region-language alignment. We present MedUP, a Med-VLM that natively unifies perception and understanding within a shared token space. At its core lies UniMedTok, a region tokenizer that encodes masks as discrete tokens in the LLM vocabulary, enabling the model to seamlessly interleave mask tokens with text. We curate UniMed-Train, a 1.84M-instance corpus spanning text-guided segmentation, region-grounded understanding, medical VQA and CoT-based segmentation, and introduce UniMed-Bench for unified evaluation. Extensive experiments show that MedUP outperforms native, agentic, and dual-decoder Med-VLMs across all tasks while remaining competitive with specialist segmentors, demonstrating the strong potential of unified understanding and perception modeling.
Chinese Translation
医学视觉-语言模型（Med-VLMs）在将视觉内容进行语言化方面表现出色，但精确的视觉感知、分割和定位仍然具有挑战性。现有的方法要么将区域以坐标字符串的形式进行语言化，要么依赖于将感知与理解解耦的外部模块，从而在区域与语言的对齐中产生表示差距。我们提出了MedUP，这是一种在共享标记空间内本地统一感知与理解的Med-VLM。其核心是UniMedTok，一个将掩码编码为LLM词汇中离散标记的区域标记器，使模型能够无缝地将掩码标记与文本交错。我们整理了UniMed-Train，这是一个包含184万实例的语料库，涵盖文本引导的分割、区域基础的理解、医学视觉问答（VQA）和基于链式推理（CoT）的分割，并引入了UniMed-Bench用于统一评估。大量实验表明，MedUP在所有任务中均优于本地、主动和双解码器的Med-VLM，同时在与专业分割器的竞争中保持竞争力，展示了统一理解与感知建模的强大潜力。
cs.CV / 47 / 2608.10648
Precise Top-Layer Fabric Segmentation for Fabric Destacking with Edge- and Shape-Aware Deep Networks
基于边缘和形状感知深度网络的精确顶层织物分割用于织物去堆叠
Abstract
Fabric destacking requires precise segmentation of the topmost fabric layer, a task complicated by subtle fabric boundaries and high visual similarity between fabric layers. Existing semantic and edge-based segmentation approaches often struggle with these complexities, limiting the performance of robotic manipulation for different tasks. In this work, a novel segmentation training architecture tailored for top-layer fabric segmentation in stacked fabrics is proposed. The method extends the classical encoder-decoder framework by introducing two specialized branches - an edge-aware branch and a shape-aware branch - that are used to supervise the backbone network for better tuning. The edge-aware branch enhances boundary delineation, while the shape-aware branch guides the network to capture and align the overall fabric shape with reference masks derived from Computer Aided Design (CAD) models. Experiments on a real-world fabric dataset demonstrate that the training approach outperforms established baselines, verifying the effectiveness of the multi-branch design through both quantitative results and ablation studies.
Chinese Translation
织物去堆叠需要对最上层织物进行精确分割，这一任务因织物边界微妙和织物层之间的高视觉相似性而变得复杂。现有的语义和基于边缘的分割方法常常难以应对这些复杂性，从而限制了机器人在不同任务中的操作性能。在本研究中，提出了一种针对堆叠织物顶层分割的新型分割训练架构。该方法通过引入两个专门的分支——边缘感知分支和形状感知分支——扩展了经典的编码器-解码器框架，以更好地调优主干网络。边缘感知分支增强了边界的描绘，而形状感知分支则引导网络捕捉并对齐整体织物形状，与来自计算机辅助设计（CAD）模型的参考掩膜进行匹配。在真实世界的织物数据集上的实验表明，该训练方法超越了既定基线，通过定量结果和消融研究验证了多分支设计的有效性。
cs.CV / 48 / 2608.10649
PolypVision: A Three-Stage Hierarchical Deep Learning Framework for Classification and Segmentation of Colorectal Polyps
PolypVision：一种用于结直肠息肉分类和分割的三阶段层次深度学习框架
Abstract
Colorectal cancer (CRC) remains one of the leading causes of cancer-related mortality worldwide, predominantly arising from precancerous polyps. Accurate detection, segmentation, and endoscopic and histological classification of colorectal polyps are crucial for timely clinical intervention. In this study, we present PolypVision, a three-stage hierarchical deep learning framework that sequentially performs: (Stage 1) binary classification of polyps as adenomatous or hyperplastic, with simultaneous Paris and JNet classification, using EfficientNetV2-M with Focal Loss; (Stage 2) polyp segmentation with recommended resection method using a UNet++ decoder with the Stage 1 backbone as encoder, optimized with Dice and BCE losses; and (Stage 3) adenoma subtype classification (tubular, tubulovillous, villous) using EfficientNetV2-M with transfer learning from Stage 2. Evaluated on three public datasets -- PolypGen, Kvasir-SEG, and CVC-ClinicDB -- PolypVision achieves an AUC of approximately 0.99 for frame classification and a detection mAP@50 of 94.4% on Kvasir-SEG, outperforming or matching state-of-the-art methods. Gradient-weighted Class Activation Maps (Grad-CAM) confirm that the model attends to clinically relevant lesion features. The framework is device-independent, operating across diverse endoscopic imaging systems without hardware-specific adaptation. These results demonstrate that a hierarchical, transfer-learning-driven pipeline with task-specific loss functions offers a robust, device-independent, and clinically meaningful approach to automated colorectal polyp analysis. PolypVision is freely available as a web application at https://polypvision.com, a DataBioX initiative, with a free usage tier open to all users.
Chinese Translation
结直肠癌（CRC）仍然是全球癌症相关死亡的主要原因之一，主要源于癌前息肉。准确检测、分割以及内镜和组织学分类结直肠息肉对于及时的临床干预至关重要。在本研究中，我们提出了PolypVision，一种三阶段层次深度学习框架，依次执行以下任务：（阶段1）将息肉进行二分类，分为腺瘤性或增生性，同时进行Paris和JNet分类，使用EfficientNetV2-M和Focal Loss；（阶段2）使用UNet++解码器进行息肉分割，推荐切除方法，采用阶段1的主干作为编码器，优化目标为Dice和BCE损失；（阶段3）使用EfficientNetV2-M进行腺瘤亚型分类（管状、管状绒毛、绒毛），并从阶段2进行迁移学习。在三个公共数据集（PolypGen、Kvasir-SEG和CVC-ClinicDB）上评估，PolypVision在帧分类上实现了约0.99的AUC，在Kvasir-SEG上检测的mAP@50为94.4%，超越或匹配了最先进的方法。梯度加权类激活图（Grad-CAM）确认模型关注临床相关的病变特征。该框架独立于设备，能够在多种内镜成像系统上运行，无需特定硬件的适配。这些结果表明，基于层次结构的、驱动于迁移学习的管道结合特定任务的损失函数，提供了一种稳健的、设备无关的、具有临床意义的自动化结直肠息肉分析方法。PolypVision作为一个网络应用程序免费提供，网址为https://polypvision.com，这是一个DataBioX倡议，所有用户均可享受免费使用层级。
cs.CV / 49 / 2608.10660
Cross-View Sequential Visual Localization with Spatio-Temporal Context Modeling for Autonomous Driving
基于时空上下文建模的跨视角序列视觉定位用于自动驾驶
Abstract
Continuous and reliable localization is essential for autonomous driving. Cross-view visual localization matches ground images with satellite maps, providing complementary localization cues for pipelines that depend on Global Navigation Satellite System (GNSS) signals and high-definition (HD) maps. Most existing cross-view visual localization methods process each frame independently, leaving temporal information underused and limiting accuracy under dynamic occlusion, illumination variation, and repetitive textures. This study proposes a temporal-context-enhanced framework for cross-view sequence visual localization. The proposed recurrent cross-frame module aggregates historical context from the previous state to enhance the coarse ground feature of each current frame. These enhanced features facilitate satellite candidate-region classification, while hierarchical fine-grained features enable precise local offset estimation. On the CVIS dataset, the proposed method reduces mean localization error from 3.80 m to 1.57 m and increases R@1 m from 8.14% to 40.22%. Direct transfer to KITTI-CVL achieves a mean error of 2.61 m, with target-domain fine-tuning further reducing the mean error to 2.27 m. Zero-shot field experiments on a real-world vehicle achieve a mean error of 2.84 m and R@5 m of 96.86%. These results demonstrate that temporal context enhancement significantly improves cross-view localization accuracy and supports robust deployment on public benchmarks and real-world roads.
Chinese Translation
连续可靠的定位对于自动驾驶至关重要。跨视角视觉定位通过将地面图像与卫星地图匹配，为依赖全球导航卫星系统（GNSS）信号和高清（HD）地图的管道提供互补的定位线索。现有的大多数跨视角视觉定位方法独立处理每一帧，导致时间信息未被充分利用，并限制了在动态遮挡、光照变化和重复纹理下的准确性。本研究提出了一种增强时序上下文的跨视角序列视觉定位框架。所提的递归跨帧模块从先前状态聚合历史上下文，以增强每一当前帧的粗略地面特征。这些增强特征有助于卫星候选区域分类，而分层细粒度特征则实现精确的局部偏移估计。在CVIS数据集上，所提方法将平均定位误差从3.80米降低至1.57米，并将R@1米从8.14%提高至40.22%。直接转移到KITTI-CVL时，平均误差为2.61米，目标领域的微调进一步将平均误差降低至2.27米。在真实车辆上的零样本实地实验实现了平均误差为2.84米，R@5米为96.86%。这些结果表明，时序上下文增强显著提高了跨视角定位的准确性，并支持在公共基准和真实道路上的稳健部署。
cs.CV / 50 / 2608.10677
Chartography: A Benchmark for Professional Chart Understanding
图表理解基准：Chartography
Abstract
Professionals across medicine, engineering, finance, manufacturing, and the sciences often make consequential decisions from charts. Existing chart benchmarks do not sufficiently measure this ability: they are dominated by bar, line, and pie formats, rely on shorter reasoning chains, and are nearing saturation, with frontier models already scoring 80-90%. We introduce Chartography, a benchmark of 100 tasks that pair charts drawn from professional practice, in domain-specific formats that standard chart benchmarks rarely include, with questions written by professionals who read these charts for a living and independently verified by three additional experts. In an evaluation of 30 frontier-model configurations (20 scored trials per task), the best configuration reaches only 45.0% mean pass@1; the remainder span 9.0-39.5%. Failures concentrate in visual perception: models can miss nuanced features, misread values along sparsely labeled axes, mishandle projected 3D geometry, and violate domain conventions encoded in the chart. We release all tasks, images, provenance metadata, and evaluation code.
Chinese Translation
医学、工程、金融、制造业和科学等领域的专业人士常常依赖图表做出重要决策。现有的图表基准测试无法充分衡量这一能力：它们主要集中在条形图、折线图和饼图格式上，依赖较短的推理链，并且已接近饱和，前沿模型的得分已达到80-90%。我们提出了Chartography，这是一个包含100个任务的基准，任务结合了来自专业实践的图表，这些图表采用标准图表基准很少包含的领域特定格式，并由以图表为生的专业人士撰写的问题，且经过三位额外专家的独立验证。在对30个前沿模型配置（每个任务20次评分试验）的评估中，最佳配置的平均通过率仅为45.0%；其余配置的通过率在9.0%到39.5%之间。失败主要集中在视觉感知方面：模型可能会遗漏细微特征，错误读取稀疏标记轴上的值，处理投影的三维几何体时出现问题，以及违反图表中编码的领域约定。我们发布了所有任务、图像、来源元数据和评估代码。
cs.CV / 51 / 2608.10680
Bridging Severe Cross-Modal Misalignment: End-to-End Visible-Infrared Object Detection via Explicit Feature-Domain Affine Registration
弥合严重的跨模态不对齐：通过显式特征域仿射配准的端到端可见-红外物体检测
Abstract
Visible-infrared object detection relies on complementary RGB and thermal cues, but its performance is often degraded by cross-modal spatial misalignment. Most existing methods rely on implicit feature adaptation to handle weakly misaligned scenarios, while large-offset geometric discrepancies remain insufficiently addressed. In this paper, we propose a Joint Feature-domain Registration and Detection network (JFRDet), an end-to-end visible-infrared oriented object detector tailored for severely cross-modal geometric discrepancies. JFRDet introduces a Cross-Modal Affine Alignment (CMAA) module to estimate an image-level affine transformation for explicit multi-level feature alignment. Note that illumination changes directly affect the reliability of RGB cues, an Illumination-Guided Complementary Fusion (IGCF) module adaptively exploits modality reliability under varying illumination conditions for cross-modal fusion. Then, an Alignment Quality-Consistency Gating (AQCG) strategy stabilizes joint optimization by modulating detection supervision according to alignment reliability and gradient consistency. We further construct DroneVehicle Misaligned (DVMA), a benchmark for evaluating visible-infrared oriented object detection under severe cross-modal geometric misalignment. The proposed JFRDet achieves 69.7\% $\mathrm{mAP}_{50}$ on DVMA, which represents state-of-the-art (SOTA) performance. The code and dataset will be available on GitHub.
Chinese Translation
可见-红外物体检测依赖于互补的RGB和热成像线索，但其性能常常受到跨模态空间不对齐的影响。现有大多数方法依赖于隐式特征适应来处理弱对齐场景，而大偏移几何差异仍未得到充分解决。本文提出了一种联合特征域注册与检测网络（Joint Feature-domain Registration and Detection，JFRDet），这是一种针对严重跨模态几何差异的端到端可见-红外定向物体检测器。JFRDet引入了一个跨模态仿射对齐（Cross-Modal Affine Alignment，CMAA）模块，用于估计图像级别的仿射变换，以实现显式的多层特征对齐。需要注意的是，光照变化直接影响RGB线索的可靠性，因此，光照引导的互补融合（Illumination-Guided Complementary Fusion，IGCF）模块在不同光照条件下自适应地利用模态可靠性进行跨模态融合。接着，Alignment Quality-Consistency Gating（AQCG）策略通过根据对齐可靠性和梯度一致性调节检测监督来稳定联合优化。我们进一步构建了无人机-车辆不对齐（DroneVehicle Misaligned，DVMA）基准，用于评估严重跨模态几何不对齐下的可见-红外定向物体检测。所提出的JFRDet在DVMA上实现了69.7%的$ ext{mAP}_{50}$，代表了当前的最先进性能（SOTA）。代码和数据集将会在GitHub上发布。
cs.CV / 52 / 2608.10682
Visual Geometry Foundation-Aware Gaussians for Single-Frame Surround-View Driving Reconstruction
面向视觉几何基础的高斯模型用于单帧全景驾驶重建
Abstract
Single-frame surround-view reconstruction faces severe geometric instability and rendering artifacts due to minimal inter-camera overlap. While existing methods rely on complex decoders or auxiliary cues, they remain bottlenecked by the weak geometric capacity of upstream features. We argue that leveraging pretrained visual geometry priors strengthens upstream representations and alleviates the geometric ambiguity in sparse surround views. To this end, we propose VGGD, a visual geometry foundation-aware 3D Gaussian Splatting framework for feed-forward surround-view driving reconstruction, which shifts geometric modeling to the frontend and adapts foundation priors to the driving camera setting. First, VGGD leverages VGGT to provide transferable multi-view geometric prior tokens. Next, we introduce a Dual-Path Neck to decouple geometry-consistent and appearance-aware representations, improving appearance completion in weakly observed regions. We further apply Scale Warmup to stabilize early geometry learning and suppress scale drift under ego-pose changes. Finally, we use a hybrid pixel--volume Gaussian decoder to produce a renderable 3D Gaussian scene for novel-view synthesis. Experiments on the nuScenes single-frame benchmark show that VGGD achieves the best overall rendering quality among the compared methods and improves relative geometric consistency.
Chinese Translation
单帧全景重建由于相机间重叠度较低，面临严重的几何不稳定性和渲染伪影。现有方法依赖复杂的解码器或辅助线索，但仍受到上游特征几何能力弱的瓶颈限制。我们认为，利用预训练的视觉几何先验可以增强上游表示，并缓解稀疏全景视图中的几何模糊。为此，我们提出了VGGD，一种面向视觉几何基础的3D高斯点云框架，用于前馈全景驾驶重建，该框架将几何建模转移到前端，并将基础先验适应于驾驶相机设置。首先，VGGD利用VGGT提供可转移的多视角几何先验标记。接下来，我们引入了双路径颈部结构，以解耦几何一致性和外观感知表示，从而改善弱观察区域的外观补全。我们进一步应用规模预热技术，以稳定早期几何学习并抑制自我姿态变化下的规模漂移。最后，我们使用混合像素-体积高斯解码器生成可渲染的3D高斯场景，以进行新视角合成。在nuScenes单帧基准测试中的实验表明，VGGD在比较方法中实现了最佳的整体渲染质量，并改善了相对几何一致性。
cs.CV / 53 / 2608.10684
Embedding Rotation Invariance for Provable Multi-Oriented Scene Text Recognition
可证明的多方向场景文本识别中的嵌入旋转不变性
Abstract
Multi-oriented text is ubiquitous in real-world scenes and remains a major challenge for scene text recognition (STR). Existing rotation-aware methods explicitly estimate text orientation. However, due to the lack of theoretical guarantees, they are prone to error accumulation, increased computational cost, and strong reliance on data. In this work, we incorporate rotation invariance into the STR framework to address these limitations. Specifically, we adopt an encoder-decoder architecture, embedding rotation equivariance in the encoder and rotation invariance in the decoder to construct a fully rotation-invariant network. On the decoder side, we first identify and prove the rotation-invariant property of the cross-attention mechanism and use it to formulate a rotation-invariant text decoder that maps visual features to output text in a rotation-invariant manner. On the encoder side, we propose a rotation-equivariant local-global extraction network that integrates deep equivariant convolutions with self-attention, enabling rotation-equivariant feature extraction while modeling inter-character dependencies and preserving fine-grained visual details. By integrating the encoder and decoder, we obtain an end-to-end Rotation-Invariant Scene Text Recognition network (RISTER). RISTER provides rotation invariance with theoretical guarantees, enhancing robustness on multi-oriented samples without introducing additional inference computation or relying on data-driven orientation correction. Experiments show that RISTER achieves state-of-the-art performance on both standard and multi-oriented benchmarks, surpassing the second-best model by 4.0 percent in accuracy on the general multi-oriented dataset.
Chinese Translation
多方向文本在现实场景中无处不在，仍然是场景文本识别（STR）的一大挑战。现有的旋转感知方法显式估计文本方向。然而，由于缺乏理论保证，这些方法容易导致错误累积、计算成本增加，并且对数据的依赖性较强。在本研究中，我们将旋转不变性纳入STR框架，以解决这些局限性。具体而言，我们采用编码器-解码器架构，在编码器中嵌入旋转等变性，在解码器中嵌入旋转不变性，以构建一个完全旋转不变的网络。在解码器方面，我们首先识别并证明了交叉注意机制的旋转不变性，并利用它来构建一个旋转不变的文本解码器，该解码器以旋转不变的方式将视觉特征映射到输出文本。在编码器方面，我们提出了一种旋转等变的局部-全局提取网络，该网络将深度等变卷积与自注意力相结合，使得在建模字符间依赖关系的同时，能够进行旋转等变特征提取并保留细粒度的视觉细节。通过整合编码器和解码器，我们获得了一个端到端的旋转不变场景文本识别网络（RISTER）。RISTER提供了具有理论保证的旋转不变性，提高了对多方向样本的鲁棒性，而无需引入额外的推理计算或依赖数据驱动的方向校正。实验表明，RISTER在标准和多方向基准测试中均实现了最先进的性能，在通用多方向数据集上比第二好的模型的准确率高出4.0%。
cs.CV / 54 / 2608.10706
MMArt A Multi-Perspective Multimodal Dataset for Visual Art Understanding
MMArt：用于视觉艺术理解的多视角多模态数据集
Abstract
Recent vision-language models demonstrate impressive general visual understanding, yet their art interpretation remains shallow: they describe surface content but struggle with formal analysis, grounded historical interpretation, or affective characterization. We argue this is not only a model but also a dataset limitation. Existing art datasets are single perspective resources, where no dataset provides narrative, formal, emotional, and historical perspectives simultaneously for the same artworks. We introduce MMArt, a large-scale dataset of 74,234 WikiArt paintings, each annotated with four independently annotated perspectives plus a harmonized unified caption, produced by specialized vision-language models or human annotation and validated through complementary quality evaluations. Two complementarity analyses establish that perspectives encode genuinely distinct information. A generative analysis shows that formal analysis descriptions best preserve compositional style, and historical descriptions carry strong affective signal in reconstructed images. A discriminative retrieval analysis reveals task-asymmetry: narrative descriptions drive retrieval (R@1 = 44.0%), while formal descriptions, strongest for reconstruction, are nearly nondiscriminative at retrieval scale (R@1 = 7.8%). Leave-one-out analysis further confirms that historical descriptions are the least replaceable perspective across both tasks. Together, the two analyses establish that no single perspective suffices for all tasks, directly motivating MMArt multi-perspective design. The dataset, code, and additional information are available at https://shuaiwang97.github.io/MMArt/.
Chinese Translation
近期的视觉-语言模型展示了令人印象深刻的通用视觉理解能力，但它们在艺术解读方面仍显得肤浅：它们描述表面内容，但在形式分析、基于历史的解读或情感特征方面存在困难。我们认为这不仅是模型的问题，也是数据集的局限。现有的艺术数据集是单一视角的资源，没有任何数据集能够同时为同一艺术作品提供叙事、形式、情感和历史视角。我们引入了MMArt，这是一个大规模的数据集，包含74,234幅WikiArt画作，每幅作品都附有四个独立注释的视角以及一个统一的和谐标题，这些都是由专业的视觉-语言模型或人工注释生成，并通过互补的质量评估进行验证。两项互补性分析表明，这些视角编码了真正不同的信息。一项生成分析显示，形式分析描述在保留构图风格方面效果最佳，而历史描述在重建图像中传达了强烈的情感信号。一项区分性检索分析揭示了任务的不对称性：叙事描述驱动检索（R@1 = 44.0%），而形式描述在重建中最强，但在检索规模上几乎没有区分性（R@1 = 7.8%）。逐一排除分析进一步确认历史描述在这两项任务中是最不可替代的视角。综合这两项分析，表明没有单一视角能够满足所有任务的需求，这直接促成了MMArt的多视角设计。数据集、代码及更多信息可在 https://shuaiwang97.github.io/MMArt/ 获取。
cs.CV / 55 / 2608.10708
Self-Geometry: GT-Free and Plug-and-Play Test-Time Adaptation for Geometrically Consistent 3D Vision Foundation Models
自几何：无GT且即插即用的测试时适应方法用于几何一致的3D视觉基础模型
Abstract
Recent Vision Foundation Models (VFMs) predict depth, camera pose, and pointmap in a single forward pass without per-scene optimization, achieving strong generalization. However, enforcing explicit multi-view geometric consistency, e.g., through bundle adjustment, is computationally costly and is thus not imposed during VFM pretraining, so such inconsistency can arise. To address this, implicit self-consistency derived from model outputs (e.g., pointmaps, features), though enforced at test-time in prior work, delivers inherently limited performance gain, especially on scenes where the pretrained VFM is highly inaccurate. In contrast to this implicit signal, we propose Self-Geometry, a plug-and-play test-time adaptation pipeline that directly imposes explicit multi-view geometric constraints using 2D pixel correspondences as pseudo ground-truth. Our proposed Self-Geometry consists of Geometric Disentanglement Optimization, which combines Multi-View Consistency and Epipolar Consistency losses with Gradient Disentanglement to prevent gradient conflict; Frame Angular-Neighbor, a view sampler based on SO(3) geodesic distances for lightly imposing these constraints; and Lightweight TTA, which adapts VFMs via LoRA. Our method achieves consistent improvements in both pose and geometry estimation across six VFMs (VGGT, $\pi^3$, DA3-Giant/Large/Base/Small) and four benchmarks (7Scenes, ETH3D, ScanNet++, HiRoom).
Chinese Translation
近期的视觉基础模型（VFM）在单次前向传播中预测深度、相机姿态和点图，而无需针对每个场景进行优化，取得了强大的泛化能力。然而，通过束调整等方式强制执行显式的多视图几何一致性计算成本高，因此在VFM预训练期间并未施加此约束，导致可能出现不一致性。为了解决这个问题，基于模型输出（如点图、特征）推导的隐式自一致性，尽管在先前的工作中在测试时得到了施加，但在预训练VFM高度不准确的场景中，其性能提升有限。与这种隐式信号相对，我们提出了自几何（Self-Geometry），一种即插即用的测试时适应管道，直接利用2D像素对应关系作为伪真实值施加显式的多视图几何约束。我们提出的自几何包括几何解耦优化（Geometric Disentanglement Optimization），结合多视图一致性和极线一致性损失与梯度解耦，以防止梯度冲突；帧角邻域（Frame Angular-Neighbor），基于SO(3)测地距离的视图采样器，用于轻度施加这些约束；以及轻量级测试时适应（Lightweight TTA），通过LoRA适应VFM。我们的方法在六个VFM（VGGT、$ ext{π}^3$、DA3-Giant/Large/Base/Small）和四个基准（7Scenes、ETH3D、ScanNet++、HiRoom）上在姿态和几何估计方面均取得了一致的改进。
cs.CV / 56 / 2608.10712
Compact Feed-Forward 3D Gaussians via Saliency-Guided Primitive Merging
通过显著性引导的原始合并实现紧凑的前馈3D高斯
Abstract
3D scene reconstruction, modeling, and rendering are highly relevant for numerous tasks, and 3D Gaussian splatting has become a standard choice in this context. Its feed-forward variants provide fast reconstruction from sparse input views but often produce per-pixel primitives, leading to highly redundant and thus inefficient representations. We present a structure-aware merging pipeline that takes per-pixel primitives from any feed-forward method and consolidates them into a compact, content-adaptive Gaussian set while largely retaining visual quality at just $\frac{1}{20}^\text{th}$ of the Gaussians of a per-pixel method. We group spatially coherent Gaussians of similar appearance into variable-size clusters via adaptive superpixel segmentation guided by a saliency map, which allocates fine segments to textured regions and coarse segments to homogeneous areas. We compress each cluster into a compact latent representation through a learned encoder, then match and consolidate representations across views based on geometric overlap and feature similarity via a learned merger. A level-of-detail decoder then produces the final Gaussians at a controllable resolution, enabling a flexible quality-efficiency trade-off at inference. As a post-processing module, the pipeline is backbone-agnostic, leveraging the strengths of existing feed-forward methods. This leads to better and more robust quality than achieved by previous approaches that target a reduction in primitive count, while providing a highly compact representation, that can be rendered efficiently.
Chinese Translation
3D场景重建、建模和渲染与众多任务密切相关，而3D高斯点云已成为这一领域的标准选择。其前馈变体能够快速从稀疏输入视图中进行重建，但通常会产生逐像素的原始数据，导致高度冗余且效率低下的表示。我们提出了一种结构感知的合并管道，该管道从任何前馈方法中获取逐像素原始数据，并将其整合为一个紧凑的、内容自适应的高斯集，同时在仅为逐像素方法的$ rac{1}{20}^ ext{th}$的高斯数量下，基本保持视觉质量。我们通过显著性图引导的自适应超像素分割，将外观相似的空间一致高斯分组为可变大小的聚类，细分区域分配给纹理区域，而均匀区域则分配给粗分段。我们通过学习的编码器将每个聚类压缩为紧凑的潜在表示，然后基于几何重叠和特征相似性，通过学习的合并器匹配和整合跨视图的表示。一个细节级别解码器随后以可控的分辨率生成最终的高斯，从而在推理时实现灵活的质量与效率权衡。作为一个后处理模块，该管道与骨干网络无关，利用现有前馈方法的优势。这比以往旨在减少原始数量的方法所达到的质量更好且更稳健，同时提供了一个高度紧凑的表示，可以高效渲染。
cs.CV / 57 / 2608.10723
Grid-Preserving Knowledge Distillation: Transferring Convolutional Inductive Bias to Vision Transformers under Data Scarcity
网格保持知识蒸馏：在数据稀缺情况下将卷积归纳偏置转移到视觉变换器
Abstract
Vision Transformers underperform convolutional networks when training data is scarce, and distilling convolutional inductive biases from a CNN teacher is an effective remedy that leaves the deployed model unchanged. General-purpose feature distillation, however, transfers little in this setting. The pooling, flattening, and logit-space projections it inherits from CNN to CNN pipelines discard the spatial grid in which locality and translation equivariance are encoded, and unlike a convolutional student, a ViT cannot rebuild that structure on its own. In this paper, we propose iBKD, a distillation framework that preserves the grid along the entire transfer path. Its core module, the Inductive Bias Attention Module, aggregates every student layer onto the teacher grid with learned weights, sharpens structural cues with channel and deformable spatial attention, and injects them through convolutional cross-attention that operates between grids rather than between token sets. The module is used only during training, so the deployed model is an unmodified ViT with no inference overhead. Across seven Transformer backbones and six data-scarce benchmarks, iBKD outperforms both locality-guidance methods and general knowledge distillation baselines, and its margin widens as training data shrinks.
Chinese Translation
在训练数据稀缺的情况下，视觉变换器的表现不及卷积网络，而从卷积神经网络（CNN）教师中蒸馏卷积归纳偏置是一种有效的补救方法，该方法不会改变已部署的模型。然而，通用特征蒸馏在这种情况下转移的效果有限。它从CNN到CNN管道继承的池化、展平和对数空间投影丢弃了编码局部性和平移等变性的空间网格，与卷积学生不同，视觉变换器（ViT）无法自行重建该结构。在本文中，我们提出了iBKD，一个在整个转移路径上保持网格的蒸馏框架。其核心模块，归纳偏置注意力模块，将每个学生层聚合到教师网格上，使用学习的权重，利用通道和可变形空间注意力增强结构线索，并通过在网格之间而不是在标记集之间操作的卷积交叉注意力注入这些线索。该模块仅在训练期间使用，因此已部署的模型是一个未修改的ViT，没有推理开销。在七个变换器骨干网络和六个数据稀缺基准测试中，iBKD的表现超越了局部性引导方法和通用知识蒸馏基线，并且随着训练数据的减少，其优势不断扩大。
cs.CV / 58 / 2608.10724
InterPruner: Interactive Structured Pruning via Taylor-Implicit Criterion and Language-Prior Modulator for Multimodal Object Detection
InterPruner：通过泰勒隐式准则和语言先验调节器进行交互式结构化剪枝的多模态目标检测
Abstract
Multimodal object detection proves effective in remote sensing, especially the RGB-Infrared paradigm. The parallel feature extractors provide rich multimodal information for robust detection, yet introduce substantial channel redundancy and computational overhead. Existing pruning methods can reduce channel redundancy, but they are designed for unimodal backbones, overlooking cross-modal interactions and dynamic scene-wise redundancy. In this paper, we propose InterPruner, the first interactive structured channel pruning framework for RGB-infrared object detectors. Specifically, we first derive a Taylor-Implicit Criterion(TIC) to quantify channel importance via high-order Taylor expansion and the implicit function theorem. Then, a Modality Interaction Redundancy Analyzer (MIRA) identifies redundant channels via mutual compensability assessment. Finally, a Scene-Prior Channel Anchor (SPCA) uses language priors as semantic anchors to measure channel-scene relevance for dynamic channel importance estimation. Cross-modality channel pruning for RGB-Infrared detection is yet unexplored. Extensive experiments on RGB-infrared object detection dataset demonstrate that InterPruner maintains high performance with negligible degradation. Specifically, it even achieves a 0.6% mAP increase on the FLIR dataset when pruning 50% of the channels. Code will be available on GitHub to facilitate future work.
Chinese Translation
多模态目标检测在遥感领域表现出色，尤其是在RGB-红外范式中。并行特征提取器为稳健检测提供了丰富的多模态信息，但同时引入了大量的通道冗余和计算开销。现有的剪枝方法可以减少通道冗余，但它们是为单模态骨干网络设计的，忽视了跨模态交互和动态场景冗余。在本文中，我们提出了InterPruner，这是第一个针对RGB-红外目标检测器的交互式结构化通道剪枝框架。具体而言，我们首先推导出泰勒隐式准则（Taylor-Implicit Criterion, TIC），通过高阶泰勒展开和隐函数定理量化通道的重要性。然后，模态交互冗余分析器（Modality Interaction Redundancy Analyzer, MIRA）通过互补性评估识别冗余通道。最后，场景先验通道锚（Scene-Prior Channel Anchor, SPCA）利用语言先验作为语义锚点来测量通道与场景的相关性，以进行动态通道重要性估计。RGB-红外检测的跨模态通道剪枝尚未被探索。在RGB-红外目标检测数据集上的大量实验表明，InterPruner在保持高性能的同时，性能下降几乎可以忽略不计。具体而言，在剪枝50%通道时，它在FLIR数据集上甚至实现了0.6%的mAP提升。代码将会在GitHub上发布，以促进未来的研究工作。
cs.CV / 59 / 2608.10725
Rethinking LLM Verification: Evidence Structure, Uncertainty, and Selective Refinement
重新思考大型语言模型的验证：证据结构、不确定性与选择性细化
Abstract
Large language models (LLMs) often rely on shortcuts rather than systematic reasoning, raising safety concerns in medical applications. Allowing models to abstain when uncertain improves reliability but introduces a coverage accuracy tradeoff. We propose a two-stage framework for medical hypothesis verification in multiple-choice settings that manages this tradeoff through targeted ontology grounding, applied only when the model abstains. We show that abstention is not random but reflects genuine uncertainty, with abstained predictions associated with lower confidence. Across two frontier models (GPT-5.5, accessed via the Azure OpenAI API, and DeepSeek-R1), the proposed framework improves question-level accuracy by 9.6 percentage points (82.9% to 92.5%) and hypothesis-level accuracy by 4.2 percentage points (92.0% to 96.2%). Our experiments conducted on MedReason and MedQA show that abstention can be repurposed as a control signal for selective reasoning refinement, achieving knowledge-graph-level performance without explicit knowledge graph construction.
Chinese Translation
大型语言模型（LLMs）通常依赖捷径而非系统性推理，这在医疗应用中引发了安全隐患。当模型在不确定时选择不作答可以提高可靠性，但同时引入了覆盖率与准确率之间的权衡。我们提出了一种针对多项选择设置的医疗假设验证的两阶段框架，通过在模型选择不作答时进行有针对性的本体基础应用来管理这一权衡。我们展示了不作答并非随机，而是反映了真实的不确定性，选择不作答的预测与较低的置信度相关联。在两个前沿模型（GPT-5.5，通过Azure OpenAI API访问，以及DeepSeek-R1）上，所提框架使问题级准确率提高了9.6个百分点（从82.9%提升至92.5%），假设级准确率提高了4.2个百分点（从92.0%提升至96.2%）。我们在MedReason和MedQA上进行的实验表明，不作答可以被重新利用作为选择性推理细化的控制信号，实现知识图谱级的性能，而无需显式构建知识图谱。
cs.CV / 60 / 2608.10744
Beyond Pixels: From Video Priors to 4D Worlds
超越像素：从视频先验到4D世界
Abstract
4D generation synthesizes dynamic 3D scenes from conditions such as text or images. Existing methods either reconstruct generated RGB videos with a separate 4D model or adapt a particular video generator to predict geometry directly. The former suffers from distribution mismatch and error propagation, whereas the latter ties 4D prediction to a specific generator and may require retraining when the generator or conditioning regime changes. We ask whether the final denoised latents of video models that share a variational autoencoder (VAE) can instead provide a reusable interface to explicit 4D prediction. Building on this insight, we introduce direct latent-to-4D generation and instantiate it as Latent-to-4D, which bypasses RGB by aligning a video latent with the token grid of a pretrained 4D decoder and refining it through frame-wise and global spatiotemporal attention. Trained on roughly 1K existing reconstruction clips, a single checkpoint transfers unchanged across multiple video diffusion transformers within the same VAE family. On Text4D-200 and I4D-200, Latent-to-4D surpasses matched same-latent Wan+4RC cascades in projection-based DINO-F1 by 2.88--3.45 and 5.81 points, respectively, while also being preferred by human raters for geometry, temporal stability, and overall quality.
Chinese Translation
4D生成从文本或图像等条件合成动态3D场景。现有方法要么使用单独的4D模型重建生成的RGB视频，要么调整特定的视频生成器直接预测几何形状。前者存在分布不匹配和误差传播的问题，而后者则将4D预测与特定生成器绑定，当生成器或条件模式发生变化时可能需要重新训练。我们探讨共享变分自编码器（VAE）的视频模型的最终去噪潜变量是否可以提供一个可重用的显式4D预测接口。基于这一见解，我们引入了直接的潜变量到4D生成，并将其实例化为Latent-to-4D，该方法通过将视频潜变量与预训练4D解码器的标记网格对齐，并通过逐帧和全局时空注意力进行精细调整，从而绕过RGB。该模型在大约1000个现有重建片段上训练，单个检查点在同一VAE家族内的多个视频扩散变换器之间无缝迁移。在Text4D-200和I4D-200数据集上，Latent-to-4D在基于投影的DINO-F1指标上超越了匹配的同潜变量的Wan+4RC级联，分别提高了2.88-3.45和5.81分，同时在人类评审者中因几何形状、时间稳定性和整体质量而受到更高的偏好。
cs.CV / 61 / 2608.10758
Where To Look? : Causal Tracing of Vision Encoders in VLM
该看哪里？：视觉编码器在视觉语言模型中的因果追踪
Abstract
Vision-language models can describe an image with remarkable accuracy, yet a more fundamental question remains unanswered: what visual information actually drives their answers? In this work, we investigate this question through causal tracing, and we observe that highly causal vision tokens often lie outside the target region. Extending the analysis to larger vision-language models reveals a similar pattern across models and corruption settings, suggesting that strong multimodal performance does not necessarily imply spatially localized causal representations. We further investigate: can these models preserve visual structure when appearance cues are removed? and find that visual cues are exploited to understand visual structures. Together, our experiments expose a gap between seeing, using, and reasoning over visual structure, and provide a causal framework for studying how visual information is transformed, preserved, and ultimately used by modern vision-language models.
Chinese Translation
视觉语言模型能够以惊人的准确性描述图像，但一个更根本的问题仍未得到解答：究竟是什么视觉信息驱动了它们的回答？在本研究中，我们通过因果追踪来探讨这个问题，观察到高度因果的视觉标记往往位于目标区域之外。将分析扩展到更大的视觉语言模型，发现不同模型和损坏设置之间存在类似的模式，这表明强大的多模态性能并不一定意味着空间上局部化的因果表示。我们进一步探讨：当外观线索被移除时，这些模型能否保持视觉结构？结果发现，视觉线索被利用来理解视觉结构。我们的实验揭示了在视觉结构的观察、使用和推理之间存在差距，并提供了一个因果框架，以研究现代视觉语言模型如何转化、保留和最终使用视觉信息。
cs.CV / 62 / 2608.10764
FADE: From Passive Verification to Active Discovery in Counterfactual Video Understanding
FADE：从被动验证到主动发现的反事实视频理解
Abstract
Counterfactual video understanding evaluates whether models grasp physical and commonsense regularities. However, existing multiple-choice question (MCQ) benchmarks inadvertently leak target events through their questions and candidate options. This reduces the core challenge from active discovery to text-guided verification. In this paper, we present FADE, an effective training framework for counterfactual discovery and explanation. Our method is built on an evidence-first, two-stage training paradigm. First, evidence-internalized supervised fine-tuning grounds the model's predictions in decisive visual anomalies. Second, we apply a fading-anchor reinforcement learning strategy that progressively removes textual guidance, compelling the model to independently discover and explain evidence. To rigorously evaluate this capability, we also introduce an effective pipeline that converts existing MCQ datasets into aligned MCQ, open-ended question answering (OQA), and captioning tasks without requiring additional data curation. Our simple approach yields strong results. Using Qwen3-VL-8B as the baseline, FADE achieves state-of-the-art strict paired scores across all three tasks on DualityVidQA-test and IPV-Bench, outperforming GPT-5.6. In specific, when transitioning from constrained MCQs to unconstrained OQA and captioning, our model demonstrates remarkable robustness. Its performance retention is 90.4% and 67.4% on DualityVidQA-test-substantially higher than the 48.1% and 30.7% retained by GPT-5.6. We hope this simple framework can serve as a solid baseline for future research in unconstrained counterfactual video understanding.
Chinese Translation
反事实视频理解评估模型是否掌握物理和常识规律。然而，现有的多项选择题（MCQ）基准无意中通过其问题和候选选项泄露了目标事件。这将核心挑战从主动发现降低为文本引导的验证。本文提出了FADE，一个有效的反事实发现和解释训练框架。我们的方法基于证据优先的两阶段训练范式。首先，证据内化的监督微调将模型的预测建立在决定性的视觉异常上。其次，我们应用了一种渐进式去锚强化学习策略，逐步去除文本引导，迫使模型独立发现和解释证据。为了严格评估这一能力，我们还引入了一个有效的流程，将现有的MCQ数据集转换为对齐的MCQ、开放式问答（OQA）和字幕任务，而无需额外的数据整理。我们简单的方法取得了强劲的结果。以Qwen3-VL-8B作为基线，FADE在DualityVidQA-test和IPV-Bench的所有三个任务上达到了最先进的严格配对分数，超越了GPT-5.6。具体而言，当从受限的MCQ过渡到不受限的OQA和字幕时，我们的模型表现出显著的鲁棒性。在DualityVidQA-test上的性能保持率为90.4%和67.4%，远高于GPT-5.6的48.1%和30.7%。我们希望这个简单的框架能够为未来在不受限的反事实视频理解领域的研究提供一个坚实的基线。
cs.CV / 63 / 2608.10790
MVTrack: Ultrafast Appearance-Free Moving Object Tracking from Compressed Bitstreams
MVTrack：基于压缩比特流的超快速无外观移动物体跟踪
Abstract
Deploying modern video trackers at scale is bottlenecked by the computational cost of RGB-based object detectors. To this end, we present MVTrack, an ultrafast tracker for moving objects that operates directly on H.264 bitstreams. MVTrack combines MVDet, a lightweight detector for motion vector fields, with MVLink, a minimalist kinematic association module. On VIRAT, MVTrack outperforms YOLO26n while using 60$\times$ fewer parameters, requiring 40$\times$ fewer FLOPs, and reducing CPU latency by 8.6$\times$. These results demonstrate that compressed video data alone can enable accurate and scalable surveillance tracking, thereby bypassing the need for pixel reconstruction.
Chinese Translation
在大规模部署现代视频跟踪器时，RGB基础的物体检测器的计算成本成为瓶颈。为此，我们提出了MVTrack，一种直接在H.264比特流上操作的超快速移动物体跟踪器。MVTrack结合了MVDet，一个轻量级的运动矢量场检测器，以及MVLink，一个极简的运动关联模块。在VIRAT数据集上，MVTrack的性能超过了YOLO26n，同时使用的参数少60倍，所需的FLOPs少40倍，CPU延迟减少了8.6倍。这些结果表明，仅凭压缩视频数据就可以实现准确且可扩展的监控跟踪，从而绕过像素重建的需求。
cs.CV / 64 / 2608.10796
E$^3$mo-Bench: A Scalable Benchmark for Multimodal Evoked and Expressed Emotion Understanding via Bayesian Pairwise Alignment
E$^3$mo-Bench：一种可扩展的基准，用于通过贝叶斯成对对齐理解多模态唤起和表达的情感
Abstract
Understanding both expressed and evoked emotions is critical for multimodal large language models (MLLMs) to achieve comprehensive affect-aware interactions. However, existing benchmarks typically examine expressed and evoked emotions in isolation or are constrained to coarse-grained and incomplete affective characterizations. To bridge this gap, we introduce E$^3$mo-Bench, a scalable benchmark comprising $12{,}314$ question-answer pairs across $2{,}524$ videos with predefined affective perspectives. It evaluates evoked and expressed emotion understanding via $3$ complementary tasks: emotion perception, open-vocabulary recognition, and valence-arousal-dominance (VAD) assessment. To efficiently scale reliable continuous annotations, we propose Bayesian Pairwise Alignment, which aggregates sparse, low-burden pairwise judgments into anchor-referenced VAD estimates. Furthermore, we develop E$^3$mo-Score, a training-free agent that aggregates complementary judgments from a five-model committee to improve VAD estimation. Extensive experiments validate the effectiveness of our framework and expose a pronounced performance skew between evoked and expressed emotion paradigms. These findings, coupled with MLLMs' persistent deficits in fine-grained recognition and dimensional assessment, chart a clear course for advancing multimodal emotional intelligence.
Chinese Translation
理解表达的情感和唤起的情感对于多模态大型语言模型（MLLMs）实现全面的情感感知交互至关重要。然而，现有的基准通常孤立地考察表达和唤起的情感，或仅限于粗粒度和不完整的情感特征。为了解决这一问题，我们引入了E$^3$mo-Bench，这是一个可扩展的基准，包含$12{,}314$个问答对，涵盖$2{,}524$个具有预定义情感视角的视频。它通过$3$个互补任务评估唤起和表达情感的理解：情感感知、开放词汇识别和效价-唤起-主导性（VAD）评估。为了有效地扩展可靠的连续注释，我们提出了贝叶斯成对对齐，该方法将稀疏、低负担的成对判断聚合为锚点参考的VAD估计。此外，我们开发了E$^3$mo-Score，这是一种无训练的代理，通过五模型委员会聚合互补判断，以改善VAD估计。大量实验验证了我们框架的有效性，并揭示了唤起和表达情感范式之间明显的性能偏差。这些发现，加上MLLMs在细粒度识别和维度评估方面的持续不足，为推动多模态情感智能指明了明确的方向。
cs.CV / 65 / 2608.10798
Beyond Fixed Luminance: Towards Panchromatic and Orthochromatic Image Colorization
超越固定亮度：朝向全色和正色图像着色
Abstract
Most image colorization systems operate in $Lab$ space by predicting chroma ($ab$) while preserving an input-derived luminance channel ($L$). While effective on standard benchmarks, this fixed-luminance design restricts brightness changes and becomes unreliable when grayscale formation deviates from natural-image luminance, as in historical orthochromatic photography. We propose a luminance-agnostic colorization framework that formulates colorization as full-RGB image editing using a foundation image-editing model. To bridge modern panchromatic and historical orthochromatic conditions, we introduce a mixed grayscale objective that trains the model under both standard luminance grayscale and a red-insensitive grayscale formation. Experiments on COCO, ImageNet, and a multi-instance benchmark show that our method is competitive on standard grayscale inputs and substantially more robust under orthochromatic inputs, with qualitative comparisons and a human study indicating fewer visible color artifacts.
Chinese Translation
大多数图像着色系统在 $Lab$ 空间中通过预测色度 ($ab$) 同时保留输入衍生的亮度通道 ($L$) 来操作。尽管在标准基准测试中效果显著，但这种固定亮度设计限制了亮度变化，并且在灰度形成偏离自然图像亮度时（如历史正色摄影）变得不可靠。我们提出了一种与亮度无关的着色框架，将着色视为使用基础图像编辑模型的全 RGB 图像编辑。为了弥合现代全色和历史正色条件之间的差距，我们引入了一种混合灰度目标，在标准亮度灰度和红色不敏感灰度形成下训练模型。在 COCO、ImageNet 和一个多实例基准上的实验表明，我们的方法在标准灰度输入上具有竞争力，并且在正色输入下显著更具鲁棒性，定性比较和人类研究表明可见色彩伪影更少。
cs.CV / 66 / 2608.10801
Evaluating Semantic and Spatial Guidance for Foundation Model Segmentation of Small-Scale PV in Remote Sensing Imagery
评估基础模型在遥感影像中对小规模光伏的语义和空间引导分割
Abstract
Spatio-temporal PV data are essential for understanding adoption processes in off-grid regions, yet such data remain largely unavailable. Automated segmentation of remote sensing (RS) imagery offers a promising solution; yet, residential PV systems remain challenging targets because of their small size and sparse distribution, resulting in severe target-background imbalance. Vision-language foundation models (FMs) provide a data-efficient paradigm through prompt-based semantic and spatial guidance, but the relative contribution of different prompt types remains unclear. We systematically evaluate SAM3 for small-scale PV segmentation in RS imagery by comparing textual, geometric, and hybrid prompting, under varying supervision levels, training strategies, spatial resolutions, and imaging conditions. Multi-temporal aerial imagery from a large off-grid rural region serves as a study site, with findings validated across three additional datasets. Prompting strategy emerged as the dominant factor governing model behavior. Textual prompting consistently produced the lowest performance and showed the greatest sensitivity to supervision and imaging conditions. In contrast, spatial guidance substantially improved both segmentation accuracy and robustness. Hybrid prompting achieved the highest accuracy and stability, indicating that semantic and spatial guidance provide complementary information. Most performance gains were achieved with only a few hundred annotated samples, demonstrating strong data efficiency. Transfer learning had limited overall impact, with only modest improvements observed for textual prompting under limited supervision. Overall, our findings establish prompting strategy as a key determinant of SAM3 adaptation, robustness, and generalization, highlighting the potential of promptable FMs for scalable PV mapping in data-constrained off-grid regions.
Chinese Translation
时空光伏数据对于理解离网地区的采用过程至关重要，但此类数据仍然大部分不可用。自动化的遥感影像分割提供了一个有前景的解决方案；然而，由于住宅光伏系统体积小且分布稀疏，导致目标与背景之间严重失衡，使其成为具有挑战性的目标。视觉-语言基础模型（FMs）通过基于提示的语义和空间引导提供了一种数据高效的范式，但不同提示类型的相对贡献尚不清楚。我们系统地评估了SAM3在遥感影像中对小规模光伏分割的表现，通过比较文本、几何和混合提示，在不同的监督水平、训练策略、空间分辨率和成像条件下进行。来自一个大型离网农村地区的多时相航空影像作为研究场所，研究结果在三个额外数据集上得到了验证。提示策略成为影响模型行为的主导因素。文本提示始终产生最低的性能，并对监督和成像条件表现出最大的敏感性。相比之下，空间引导显著提高了分割的准确性和鲁棒性。混合提示实现了最高的准确性和稳定性，表明语义和空间引导提供了互补的信息。大多数性能提升仅通过几百个标注样本就能实现，显示出强大的数据效率。迁移学习的整体影响有限，在有限监督下，仅观察到文本提示的适度改善。总体而言，我们的研究结果确立了提示策略作为SAM3适应性、鲁棒性和泛化能力的关键决定因素，突显了可提示基础模型在数据受限的离网地区进行可扩展光伏映射的潜力。
cs.CV / 67 / 2608.10804
BPG: Balancing Plasticity and Generalization for Domain Incremental Learning
BPG：平衡可塑性与泛化能力的领域增量学习
Abstract
Deep neural networks excel in various tasks but struggle to generalize across evolving data distributions, leading to significant performance degradation under domain shifts. Domain incremental learning (DIL) addresses this challenge by enabling models to continuously adapt while retaining prior knowledge. Among existing DIL approaches, the parameter-isolation paradigm achieves state-of-the-art performance. However, these methods often adopt a one-size-fits-all approach to adapt to new domains, resulting in either insufficient learning capacity or redundant parameters. In this work, we propose BPG, a unified framework that addresses both challenges through two complementary components: BPG-Adapter, which dynamically determines each domain's adapter hidden dimension based on domain-specific feature separability, and BPG-Inference, a soft domain mixture strategy that integrates multiple domain-specific models at test time, mitigating domain ID misselection. Experimental results on DomainNet, CDDB, and CORe50 demonstrate that BPG consistently outperforms uniform adapter-based approaches and hard domain selection strategies, achieving state-of-the-art average accuracy while reducing forgetting to as low as 0.22% on DomainNet.
Chinese Translation
深度神经网络在各种任务中表现出色，但在不断变化的数据分布下难以实现泛化，导致在领域转移时性能显著下降。领域增量学习（DIL）通过使模型能够持续适应同时保留先前知识来应对这一挑战。在现有的DIL方法中，参数隔离范式达到了最先进的性能。然而，这些方法通常采用一刀切的方式来适应新领域，导致学习能力不足或参数冗余。在本研究中，我们提出了BPG，一个统一框架，通过两个互补组件解决这两个挑战：BPG-Adapter，根据领域特定特征的可分离性动态确定每个领域的适配器隐藏维度，以及BPG-Inference，一种软领域混合策略，在测试时整合多个领域特定模型，减轻领域ID错误选择的问题。在DomainNet、CDDB和CORe50上的实验结果表明，BPG在平均准确率上始终优于基于均匀适配器的方法和硬领域选择策略，同时在DomainNet上将遗忘率降低至低至0.22%。
cs.CV / 68 / 2608.10805
Fast and Memory-Efficient Wavelet Convolutions via I/O-Aware Reformulation
通过I/O感知重构实现快速且内存高效的小波卷积
Abstract
Wavelet convolution (WTConv) has emerged as an increasingly popular drop-in replacement for standard convolutions, expanding a network's receptive field exponentially with the number of decomposition levels while keeping the parameter count linear. However, its reference implementation is severely memory-bound due to excessive data movement through high-bandwidth memory (HBM). We develop an I/O model of WTConv to characterize this bottleneck and use it to guide three algebraic reformulations: (1) recomputing the inexpensive Haar analysis butterfly on chip, (2) collapsing the multi-level synthesis cascade into a single closed-form pass indexed by output-coordinate bits, and (3) folding learned per-channel scales into the convolution weights. Together, these reformulations enable an I/O-aware fused implementation that substantially reduces HBM traffic. We evaluate the WTConvNeXt configuration across decomposition levels and a broad range of tensor shapes. Despite performing comparable arithmetic, the reference WTConv is substantially slower than the depthwise convolution it replaces. Our reformulation reduces modeled HBM traffic by approximately $2.55\times$, yielding up to a $4.35\times$ training speedup over the reference while roughly halving peak memory usage. Thus, our reformulation preserves the benefits of WTConv while substantially reducing its execution time and memory footprint, removing the systems overhead that previously limited its practical efficiency.
Chinese Translation
小波卷积（WTConv）作为标准卷积的替代方案，逐渐受到广泛关注，它通过增加分解层数以指数方式扩展网络的感受野，同时保持参数数量线性。然而，其参考实现由于通过高带宽内存（HBM）进行过多的数据移动而严重受限于内存。我们开发了一个WTConv的I/O模型，以表征这一瓶颈，并利用该模型指导三种代数重构：（1）在芯片上重新计算廉价的Haar分析蝶形结构，（2）将多层合成级联压缩为由输出坐标位索引的单个封闭形式传递，以及（3）将学习到的每通道缩放因子折叠到卷积权重中。这些重构共同实现了一种I/O感知的融合实现，显著减少了HBM流量。我们在不同的分解层数和广泛的张量形状上评估了WTConvNeXt配置。尽管执行的算术运算相当，参考WTConv的速度仍明显慢于其替代的深度卷积。我们的重构将模型化的HBM流量减少了约$2.55 imes$，在参考实现的基础上实现了高达$4.35 imes$的训练加速，同时大约将峰值内存使用量减半。因此，我们的重构在显著减少执行时间和内存占用的同时，保留了WTConv的优势，消除了之前限制其实际效率的系统开销。
cs.CV / 69 / 2608.10807
Modelling Geographic Atrophy Progression using Implicit Neural Representations
使用隐式神经表示建模地理性萎缩的进展
Abstract
Age-related Macular Degeneration (AMD) is the major cause of blindness in the Western world. Its late dry phase is characterised by irreversible atrophic areas, namely Geographic Atrophy (GA). Longitudinal Fundus Autofluorescence (FAF) image acquisitions are currently the main tool for assessing lesion growth over time at the image level. However, due to its highly individualised progression, the evolution of late AMD remains poorly understood. In this work, we propose using Implicit Neural Representations (INRs) to model GA progression at the individual level in a low-data setting. Our approach generates both FAF and GA segmentation at both past and future time points. Among the comparison models, our method achieves competitive segmentation quality across different scenarios, yielding the lowest Mean Absolute Error (MAE) for the GA lesion area and the highest DICE score, without sacrificing FAF image quality. The code is available at https://github.com/SimoneSarrocco/ga-progression-with-inrs.
Chinese Translation
年龄相关性黄斑变性（AMD）是西方世界失明的主要原因。其晚期干性阶段的特征是不可逆的萎缩区域，即地理性萎缩（GA）。纵向眼底自发荧光（FAF）图像采集目前是评估病变随时间增长的主要工具。然而，由于其高度个体化的进展，晚期AMD的演变仍然不甚了解。在本研究中，我们提出在低数据环境下使用隐式神经表示（INRs）来建模个体层面的GA进展。我们的方法在过去和未来的时间点生成FAF和GA分割。在比较模型中，我们的方法在不同场景下实现了竞争性的分割质量，GA病变面积的平均绝对误差（MAE）最低，DICE得分最高，同时不牺牲FAF图像质量。代码可在https://github.com/SimoneSarrocco/ga-progression-with-inrs获取。
cs.CV / 70 / 2608.10827
MIRA: Medical Image Reflection for Agentic Diagnosis
MIRA：用于自主诊断的医学图像反思
Abstract
Medical visual agents can use tools to inspect images and retrieve external knowledge, but indiscriminate tool use may introduce noisy or misleading evidence. Reliable diagnosis therefore requires not only acquiring additional observations, but also verifying whether tool actions are necessary and whether the resulting evidence supports the current hypothesis. We introduce MIRA (Medical Image Reflection for Agentic Diagnosis), a medical visual diagnostic framework for autonomous evidence search and reflective verification. MIRA dynamically invokes image-processing operations, including zooming, grounding, pointing, rotation, and measurement, as well as web search, while evaluating the relevance and consistency of the acquired evidence. We develop MIRA through a two-stage training strategy. First, a tool-augmented Monte Carlo Tree Search data engine explores diverse diagnostic hypotheses and jointly verifies visual grounding accuracy and semantic consistency to construct supervised fine-tuning trajectories. Second, reinforcement learning further improves decision-making through online reflective principle evolution: failure cases are distilled into candidate principles, and only principles that improve held-out rollout rewards are retained. Across nine medical visual reasoning benchmarks, MIRA achieves an average score of 64.73, improving its Qwen3-VL-8B backbone by 7.44 points. It also increases useful tool-use judgments from 56.2% to 73.8% and reduces harmful judgments from 8.9% to 1.6%. Qualitative analyses show that MIRA can re-examine evidence, correct premature conclusions, and adapt its tool-use strategy. Project page: https://MIRA-VL.github.io/
Chinese Translation
医学视觉代理可以使用工具来检查图像并检索外部知识，但不加选择的工具使用可能会引入噪声或误导性证据。因此，可靠的诊断不仅需要获取额外的观察结果，还需要验证工具操作是否必要，以及所获得的证据是否支持当前假设。我们提出了MIRA（Medical Image Reflection for Agentic Diagnosis），这是一个用于自主证据搜索和反思验证的医学视觉诊断框架。MIRA动态调用图像处理操作，包括缩放、定位、指向、旋转和测量，以及网络搜索，同时评估所获得证据的相关性和一致性。我们通过两阶段训练策略开发MIRA。首先，增强工具的蒙特卡洛树搜索数据引擎探索多样的诊断假设，并共同验证视觉定位的准确性和语义一致性，以构建监督微调轨迹。其次，强化学习通过在线反思原则演化进一步改善决策：失败案例被提炼为候选原则，只有那些提高保留回报的原则才会被保留。在九个医学视觉推理基准测试中，MIRA的平均得分为64.73，较其Qwen3-VL-8B基础模型提高了7.44分。它还将有效工具使用判断从56.2%提高到73.8%，并将有害判断从8.9%降低到1.6%。定性分析表明，MIRA能够重新审视证据，纠正过早的结论，并调整其工具使用策略。项目页面：https://MIRA-VL.github.io/
cs.CV / 71 / 2608.10835
UniProbe: A Learnable Token-Level Hallucination Detector for Large VLMs using Multi-Structural Internal Representations
UniProbe：一种可学习的基于标记的幻觉检测器，适用于使用多结构内部表示的大型视觉语言模型
Abstract
Large Vision-Language Models (LVLMs) achieve impressive visual reasoning and dialogue capabilities, yet frequently hallucinate content unsupported by the visual input. Effective mitigation requires token-level localization, enabling targeted intervention without discarding the entire response. Existing detectors require expensive full-model fine-tuning, rely on external verifiers that ignore the model's generation process, or reduce internal signals to isolated features and hand-crafted statistics, discarding spatial, sequential, and relational structure. We introduce \textbf{UniProbe}, a lightweight, unified, learnable detector that models a frozen LVLM's heterogeneous computational trace from a single forward pass. UniProbe constructs a directed graph over image patches, query tokens, and generated tokens, with attention weights encoding their relations. It processes this trace with alternating structure-aware modules: a GNN for relational evidence, a ViT for 2-D visual geometry, and a GRU for response order. Interleaving them allows spatial, relational, and sequential evidence to interact throughout the detector. We further develop a streaming variant for hallucination-aware decoding, which detects and resamples hallucinated tokens during generation, and a self-adaptation strategy aligning the detector with the LVLM's own generations. Across diverse LVLM backbones, UniProbe achieves state-of-the-art token-level and object-hallucination detection. During decoding, it reduces object hallucinations by up to 55\% at $1.06\times$ the latency of standard generation.
Chinese Translation
大型视觉语言模型（LVLMs）在视觉推理和对话能力方面表现出色，但经常会产生与视觉输入不符的内容。有效的减轻措施需要进行标记级别的定位，以便进行针对性的干预，而不必丢弃整个响应。现有的检测器需要昂贵的全模型微调，依赖于忽略模型生成过程的外部验证器，或将内部信号简化为孤立特征和手工统计，忽视了空间、序列和关系结构。我们提出了 extbf{UniProbe}，一种轻量级、统一的可学习检测器，它通过单次前向传播建模一个冻结的LVLM的异构计算轨迹。UniProbe在图像块、查询标记和生成标记之间构建有向图，注意力权重编码了它们之间的关系。它使用交替的结构感知模块处理这一轨迹：一个用于关系证据的图神经网络（GNN）、一个用于二维视觉几何的视觉变换器（ViT）和一个用于响应顺序的门控循环单元（GRU）。交错使用这些模块使得空间、关系和序列证据能够在整个检测器中相互作用。我们进一步开发了一种流式变体，用于幻觉感知解码，在生成过程中检测并重新采样幻觉标记，以及一种自适应策略，使检测器与LVLM自身的生成保持一致。在多种LVLM骨干网络中，UniProbe实现了最先进的标记级和对象幻觉检测。在解码过程中，它将对象幻觉减少了多达55 ext{%}，延迟为标准生成的$1.06 imes$。
cs.CV / 72 / 2608.10838
PolyLayout: Hierarchical VLM-Guided Layout Generation Beyond Rectangular Rooms
PolyLayout：超越矩形房间的层次化VLM引导布局生成
Abstract
Generating physically plausible 3D room layouts is essential for home furnishing retail, enabling customers to visualize products in their own homes and confidently make purchasing decisions. However, a gap exists between academic research and real-world application: existing solutions primarily focus on algorithmic strategies for furniture placement, largely neglecting the non-rectangular geometries and strict door/window constraints prevalent in real homes. To bridge the gap, we introduce a hybrid, hierarchical framework tailored for retail, specifically designed to support scalable spatial planning applications. Our system decouples generation into three stages: (1) functional furniture clustering and fine-grained intra-zone placement; (2) macro-routing guided by a vision-language model (VLM) to anchor both these clustered zones and any remaining standalone furniture within diverse polygonal boundaries; and (3) rule-based optimization for collision-free micro-arrangements that respect architectural constraints. We evaluate our system on production-scale catalogs and a representative set of irregular real-world topologies. Our results show that our approach attains the highest perceptual plausibility while maintaining good geometric compliance at relatively low latency, and extends to irregular boundaries that existing methods do not natively support.
Chinese Translation
生成物理上合理的3D房间布局对于家居零售至关重要，使客户能够在自己的家中可视化产品，并自信地做出购买决策。然而，学术研究与实际应用之间存在差距：现有解决方案主要集中在家具摆放的算法策略上，基本忽视了真实家居中普遍存在的非矩形几何形状和严格的门窗约束。为了解决这一问题，我们提出了一种混合的层次化框架，专为零售设计，旨在支持可扩展的空间规划应用。我们的系统将生成过程分为三个阶段：（1）功能性家具聚类和细粒度的区域内摆放；（2）通过视觉语言模型（VLM）引导的宏观路由，以锚定这些聚类区域及任何剩余的独立家具在多样的多边形边界内；（3）基于规则的优化，以实现尊重建筑约束的无碰撞微观布局。我们在生产规模的目录和一组具有代表性的非规则真实世界拓扑上评估了我们的系统。结果表明，我们的方法在保持良好几何合规性的同时，达到了最高的感知合理性，并且扩展到现有方法不原生支持的非规则边界。
cs.CV / 73 / 2608.10839
The GENEA Challenge 2026: A Large-Scale Disentangled Evaluation of Speech-Driven Gesture Generation on the Seamless Interaction Dataset
GENEA挑战2026：在无缝交互数据集上对语音驱动手势生成的规模化解耦评估
Abstract
This preprint presents the results of the fourth GENEA Challenge, a large-scale human evaluation of five speech-driven gesture-generation systems trained by participating teams on the Seamless Interaction dataset of dyadic conversations. As in the 2023 GENEA Challenge, we used a disentangled evaluation methodology to assess motion quality and speech alignment without confounding between the two, and performed a dyadic mismatching study to isolate the effect of listening and reacting to the interlocutor. We additionally introduce a new semantic gesture-generation task and a text-mismatching evaluation methodology using the Grounded Gestures subset of the data. In total, we ran four large-scale user studies, collecting over 23,000 votes from 869 test-takers. In the motion-realism study, the dataset's filtered segments had substantially higher motion quality than all challenge submissions (68-95% pairwise winrate). In the speech-alignment study, the motion-capture segments provided a conceptual ceiling at 62% alignment score, with the top submission significantly behind at 32% and the rest only slightly above the 0% expected of an input-independent system. In the dyadic study, motion capture again set the ceiling at 65% appropriateness score, but no submission scored substantially above chance, indicating that the systems could not yet respond to the interlocutor. Finally, the semantic mismatching evaluation found highly expressive gestures in the dataset (test-takers identified the matching transcript 79% of the time), yet almost all submissions failed to generate semantically expressive motion, with the best achieving only an 8% appropriateness score. The collected votes and outputs will be made publicly available at https://genea-workshop.github.io/2026/challenge/ to facilitate reproducibility and further research.
Chinese Translation
本预印本展示了第四届GENEA挑战的结果，这是对五个由参与团队在双人对话的无缝交互数据集上训练的语音驱动手势生成系统进行的大规模人类评估。与2023年GENEA挑战相同，我们采用了解耦评估方法来评估运动质量和语音对齐，避免了两者之间的混淆，并进行了双人不匹配研究，以隔离听取和反应对话者的影响。此外，我们引入了一项新的语义手势生成任务和一种使用数据的Grounded Gestures子集的文本不匹配评估方法。总的来说，我们进行了四项大规模用户研究，收集了来自869名测试者的超过23,000票。在运动现实性研究中，数据集中经过筛选的片段的运动质量显著高于所有挑战提交（68-95%的成对胜率）。在语音对齐研究中，运动捕捉片段提供了62%的对齐得分的概念上限，而最佳提交的得分显著落后，仅为32%，其余提交的得分仅略高于输入无关系统的0%。在双人研究中，运动捕捉再次设定了65%的适当性得分上限，但没有提交的得分显著高于随机水平，表明这些系统尚未能够对对话者做出反应。最后，语义不匹配评估发现数据集中存在高度表现力的手势（测试者识别匹配的文本转录的准确率为79%），然而几乎所有提交都未能生成语义上表现力的运动，最佳提交仅达到了8%的适当性得分。收集的投票和输出将公开发布在https://genea-workshop.github.io/2026/challenge/，以促进可重复性和进一步研究。
cs.CV / 74 / 2608.10864
Multi-View Relational Distillation for Spatial Reasoning with Vision-Language Models
基于多视角关系蒸馏的视觉语言模型空间推理
Abstract
Vision-language models (VLMs) have achieved strong image and video understanding, yet their visual-spatial representations remain geometrically fragile, leading to failures in spatial reasoning needed for embodied AI, robotics, and autonomous driving. Prior approaches to geometry grounding either fine-tune VLMs on spatial question answering, which can perpetuate spurious visual representations, or fuse features from large geometry-grounded vision models, which substantially increases model size at inference. Knowledge distillation from geometry-grounded vision models offers an alternative, but directly matching multi-view teacher features can disrupt the pretrained alignment between visual and textual representations, degrading object- and language-semantic capabilities. We propose multi-view relational distillation (MVRD), which distills patch-wise cosine similarities across views instead of the teacher features themselves. These relations encode geometric correspondences adequate for spatial understanding, while leaving the student representation underdetermined, allowing it to remain close to its pretrained vision- language space. Across representative VLMs, MVRD improves visual-spatial reasoning, outperforming supervised fine-tuning and feature distillation while approaching feature fusion methods with considerably fewer added parameters and lower latency. We show that MVRD makes visual representations more geometric while retaining language alignment, and generalizes to 3D scene understanding tasks such as object grounding, dense captioning, and question answering.
Chinese Translation
视觉语言模型（VLMs）在图像和视频理解方面取得了显著进展，但其视觉空间表示仍然在几何上脆弱，导致在具身人工智能、机器人技术和自动驾驶所需的空间推理中出现失败。以往的几何基础方法要么是在空间问答上微调VLMs，这可能会延续虚假的视觉表示，要么是融合来自大型几何基础视觉模型的特征，这在推理时显著增加了模型的大小。从几何基础视觉模型进行知识蒸馏提供了一种替代方案，但直接匹配多视角教师特征可能会破坏视觉和文本表示之间的预训练对齐，降低对象和语言语义能力。我们提出了多视角关系蒸馏（MVRD），该方法蒸馏跨视角的补丁级余弦相似度，而不是直接蒸馏教师特征。这些关系编码了适合空间理解的几何对应关系，同时保持学生表示的不确定性，使其能够保持接近其预训练的视觉语言空间。在代表性的VLMs中，MVRD改善了视觉空间推理，超越了监督微调和特征蒸馏，同时在增加的参数和延迟方面大幅接近特征融合方法。我们展示了MVRD使视觉表示更加几何化，同时保持语言对齐，并且能够推广到3D场景理解任务，如对象定位、密集描述和问答。
cs.CV / 75 / 2608.10870
NullEdit: Stealthy Image Protection via VLM Condition Redirection
NullEdit：通过VLM条件重定向实现隐秘图像保护
Abstract
Modern image editors combine vision-language models (VLMs) with diffusion transformer backbones to modify a single reference image according to instructions without fine-tuning. This capability also enables unauthorized manipulation of publicly released images. Existing inference-time defenses either invalidate edits through conspicuous corruption, thereby exposing the protection, or allow them to proceed with identity or reference content drift, thereby failing to prevent the editing behavior itself. We instead target a stealthy and harmless no-op in which the requested edit is suppressed, the output remains natural and source-preserving without conspicuous artifacts or identity replacement, and harmful semantics requested by malicious instructions are absent. We propose NullEdit, which targets the VLM representation jointly formed from the reference image and instruction before it conditions the downstream DiT backbone. Using normal-edit and no-edit anchors, NullEdit redirects this representation, while cross-prompt gradient averaging transfers protection to held out instructions. Across Step1X-Edit and Qwen-Image-Edit on CelebA-HQ and VGGFace2, NullEdit reduces the EditReward IF score by 0.813 on average relative to the SOTA baseline while preserving subject identity and source content.
Chinese Translation
现代图像编辑器结合了视觉-语言模型（VLM）与扩散变换器骨干网络，可以根据指令修改单一参考图像，而无需进行微调。这一能力也使得对公开发布图像的未经授权的操控成为可能。现有的推理时防御措施要么通过明显的损坏使编辑失效，从而暴露保护机制，要么允许编辑行为在身份或参考内容漂移的情况下继续进行，从而未能有效阻止编辑行为。我们则针对一种隐秘且无害的无操作（no-op）进行设计，其中请求的编辑被抑制，输出保持自然且源内容不变，没有明显的伪影或身份替换，并且恶意指令请求的有害语义被排除。我们提出了NullEdit，它针对在条件下形成的参考图像和指令的VLM表示，在其影响下再条件化下游的DiT骨干网络。通过使用正常编辑和无编辑锚点，NullEdit重定向这一表示，同时交叉提示梯度平均将保护转移到保留的指令上。在CelebA-HQ和VGGFace2上的Step1X-Edit和Qwen-Image-Edit中，NullEdit相对于现有最优基线平均降低了0.813的EditReward IF分数，同时保持了主体身份和源内容。
cs.CV / 76 / 2608.10885
ConfTriage: A Calibration-Aware LLM Triage Framework for Pulmonary Nodule Malignancy with Selective Specialist Deferral
ConfTriage：一种考虑校准的大型语言模型（LLM）肺结节恶性肿瘤分诊框架，具有选择性专家转诊
Abstract
Pulmonary nodule malignancy prediction typically depends on image-trained specialist deep learning (DL) models that require substantial annotated imaging data and task-specific training. We investigate whether a generalist large language model (LLM), reading only a faithful natural-language rendering of standard nodule attributes, can serve as a calibrated triage layer. We propose ConfTriage, a confidence-calibrated method built on three pillars: language as the modality, calibration as the safety mechanism, and a selective specialist DL backstop for low-confidence cases. We prove two guarantees: a finite-sample combined-error bound yielding an explicit per-threshold operational certificate, and an oracle inequality showing that excess risk over the Bayes-optimal deferral classifier is controlled by the L1 calibration error of the LLM probability. A controlled seven-way input ablation across five frontier LLMs on LIDC-IDRI shows that natural-language descriptions dominate the diagnostic signal, while low-level image statistics are essentially diagnostically vacuous. ConfTriage achieved an F1 score of 88.22% and an AUC of 0.92, resolving 76.5% of cases using zero-shot LLM inference alone and referring only uncertain cases to the specialist DL backstop. These results demonstrate that clinically meaningful diagnostic information can be captured through structured radiological descriptions and leveraged by calibrated LLMs for selective referral. The framework suggests a practical pathway for combining generalist LLM prediction with specialist AI models in medical decision-support systems. Source code is publicly available at https://github.com/rabiul-ai/ConfTriage.
Chinese Translation
肺结节恶性肿瘤预测通常依赖于经过图像训练的专业深度学习（DL）模型，这些模型需要大量标注的影像数据和特定任务的训练。我们研究了一种通用的大型语言模型（LLM），仅通过忠实的自然语言描述标准结节特征，是否可以作为一个经过校准的分诊层。我们提出了ConfTriage，这是一种基于三个支柱的信心校准方法：语言作为模态，校准作为安全机制，以及针对低信心案例的选择性专家DL后备。我们证明了两个保证：一个有限样本的组合误差界限，提供了明确的每个阈值操作证书，以及一个神谕不等式，表明超出贝叶斯最优转诊分类器的额外风险由LLM概率的L1校准误差控制。对LIDC-IDRI上五个前沿LLM进行的七种输入消融控制实验显示，自然语言描述主导了诊断信号，而低级图像统计在诊断上基本上是空洞的。ConfTriage实现了88.22%的F1分数和0.92的AUC，使用零-shot LLM推理单独解决了76.5%的案例，仅将不确定的案例转诊给专家DL后备。这些结果表明，通过结构化的放射学描述可以捕捉到临床上有意义的诊断信息，并通过经过校准的LLM进行选择性转诊。该框架为在医疗决策支持系统中结合通用LLM预测与专业AI模型提供了一条实用路径。源代码可在https://github.com/rabiul-ai/ConfTriage公开获取。
cs.CV / 77 / 2608.10886
GESTO: Human-Centric Spatio-Temporal Memory for Reasoning in Dynamic Scenes
GESTO：面向动态场景推理的人本时空记忆
Abstract
Robots operating in human environments need memories that capture not only what objects exist and where, but also how people use them over time and how individual interactions compose into goal-directed activities. Existing 4D scene graphs preserve object and place histories but omit activity structure, whereas activity representations are either not grounded in persistent 3D scenes or rely on externally provided event boundaries and object associations. We present GESTO (Grounded Event and Spatio-Temporal memOry), a spatio-temporal memory that couples a persistent 4D scene graph with a two-level hierarchy of atomic human--object interactions and goal-driven events. From an RGB-D observation stream, GESTO automatically extracts timestamped interactions, grounds them to persistent scene entities, groups them into events, and uses event context to refine uncertain object associations. A relation-aware tool-calling agent queries the resulting memory for activity-centric spatio-temporal reasoning. We evaluate GESTO on the reproducible text, binary, and time categories of an existing benchmark, together with 40 new Space2Event and Event2Space queries. GESTO achieves scores of 0.71, 0.75, and 0.70 on the standard categories, approaching a method supplied with ground-truth event and object grounding, while substantially outperforming the same reasoning framework when these inputs are removed. It further achieves 0.73 and 0.75 on Space2Event and Event2Space queries. Ablations show that hierarchical event structure and context-aware grounding refinement provide complementary benefits, supporting activity-grounded hierarchical memory for retrospective reasoning in dynamic human environments.
Chinese Translation
在以人为中心的环境中操作的机器人需要能够捕捉到的不仅是物体的存在及其位置，还包括人们如何随时间使用这些物体，以及个体交互如何组成目标导向的活动。现有的4D场景图保留了物体和地点的历史，但忽略了活动结构，而活动表示要么没有扎根于持久的3D场景中，要么依赖于外部提供的事件边界和物体关联。我们提出了GESTO（Grounded Event and Spatio-Temporal memOry），这是一种时空记忆，将持久的4D场景图与两级原子人-物交互和目标驱动事件的层次结构结合在一起。通过RGB-D观察流，GESTO自动提取带时间戳的交互，将其与持久场景实体关联，分组为事件，并利用事件上下文来细化不确定的物体关联。一个关系感知的工具调用代理查询生成的记忆，以进行以活动为中心的时空推理。我们在现有基准的可重复文本、二进制和时间类别上评估了GESTO，并增加了40个新的Space2Event和Event2Space查询。GESTO在标准类别上取得了0.71、0.75和0.70的分数，接近于提供真实事件和物体扎根的方法，同时在去除这些输入时显著优于相同的推理框架。它在Space2Event和Event2Space查询上进一步取得了0.73和0.75的成绩。消融实验表明，层次事件结构和上下文感知的扎根细化提供了互补的好处，支持在动态人类环境中进行回顾性推理的活动扎根层次记忆。
cs.CV / 78 / 2608.10888
Sensor-Informed Per-Point Covariance for Structured-Light 3D Imaging
传感器信息驱动的逐点协方差用于结构光三维成像
Abstract
Per-point uncertainty models are important in structured-light 3D reconstruction for probabilistic registration, fusion, and quality assessment. In practice, however, point-cloud covariances are often modeled as isotropic constants or inferred from local surface geometry and therefore do not explicitly reflect the measurement process. This is a limitation in fringe projection profilometry (FPP), where phase noise propagates through calibrated reconstruction and produces strongly anisotropic 3D uncertainty. This paper presents a sensor-informed first-order method for constructing a per-point 3 x 3 covariance field from experimentally measured phase precision and calibrated phase-to-depth and phase-to-3D mappings. The formulation separates a rank-1 phase-induced covariance from an effective full-rank completion obtained by incorporating fitted lateral image-space perturbation scales. Repeated-plane experiments under fixed imaging conditions show close alignment of the dominant covariance direction with the viewing ray, and consistency between the dominant phase-induced uncertainty scale and scalar depth uncertainty. In G-ICP registration, the proposed covariance substantially improves over a constant isotropic model while providing a sensor-derived uncertainty representation complementary to conventional geometry-based covariances.
Chinese Translation
逐点不确定性模型在结构光三维重建中对于概率配准、融合和质量评估至关重要。然而，在实际应用中，点云协方差通常被建模为各向同性常数或从局部表面几何推断，因此并未明确反映测量过程。这在条纹投影轮廓测量（FPP）中是一个限制，因为相位噪声通过校准重建传播，产生强烈的各向异性三维不确定性。本文提出了一种传感器信息驱动的一阶方法，用于从实验测量的相位精度和校准的相位到深度及相位到三维映射构建逐点的3 x 3协方差场。该公式将由相位引起的秩-1协方差与通过结合拟合的横向图像空间扰动尺度获得的有效全秩补全分开。在固定成像条件下的重复平面实验显示，主协方差方向与视线高度一致，并且主相位引起的不确定性尺度与标量深度不确定性之间保持一致。在G-ICP配准中，所提出的协方差显著优于常数各向同性模型，同时提供了一种与传统几何基础协方差互补的传感器导出不确定性表示。
cs.CV / 79 / 2608.10903
VIDS-Seg: Towards Reliable Uncertainty Quantification in Pediatric Cardiac Ultrasound Segmentation
VIDS-Seg：朝着可靠的儿童心脏超声分割不确定性量化迈进
Abstract
Reliable clinical deployment of machine learning requires models that know when they are likely to fail, particularly for subgroups underrepresented in training data. A common case is pediatric care, where models trained on adult cohorts can silently under-perform on children with no indication that something has gone wrong. As retraining with labeled pediatric data is often infeasible, detecting such failures at inference time is a critical clinical need. Building on the VIDS (Variational Inference under Distribution Shifts) framework, we introduce VIDS-Seg, which applies amortized variational inference over a lightweight prediction head to make this adaptive, OOD-aware prior tractable for dense image segmentation. We evaluate VIDS-Seg on left ventricular segmentation in echocardiography, a setting where pediatric anatomy differs systematically from the adult population most segmentation models are trained on, training on an adult cohort (EchoNet-Dynamic) and evaluating zero-shot on a pediatric cohort (EchoNet-Pediatric). Across all age strata, VIDS-Seg matches competitive baselines in segmentation accuracy while producing substantially higher spatial correspondence between predicted uncertainty and segmentation error, an advantage that persists even after applying temperature scaling to all baselines. Downstream, it yields more accurate and stable ejection fraction estimates and more reliable detection of cardiac malfunction in the infant subgroup. Our results indicate that OOD-aware uncertainty quantification can serve as a practical safety layer for deployed segmentation models, enabling detection of silent failures in underrepresented subgroups without retraining or additional labeled data.
Chinese Translation
机器学习在临床中的可靠部署需要模型能够识别其可能失败的情况，尤其是对于训练数据中代表性不足的子群体。儿童护理就是一个常见的案例，基于成人群体训练的模型在儿童身上可能会静默地表现不佳，而没有任何迹象表明出现了问题。由于使用标记的儿童数据进行再训练通常不可行，因此在推理时检测此类失败成为一个关键的临床需求。在VIDS（分布转变下的变分推断）框架的基础上，我们引入了VIDS-Seg，它在轻量级预测头上应用了摊销变分推断，使得这种自适应的、对分布外（OOD）敏感的先验在密集图像分割中变得可行。我们在心脏超声的左心室分割任务上评估了VIDS-Seg，这是一个儿童解剖结构与大多数分割模型训练的成人群体系统性不同的场景，训练于成人群体（EchoNet-Dynamic），并在儿童群体（EchoNet-Pediatric）上进行零样本评估。在所有年龄层中，VIDS-Seg在分割准确性上与竞争基线相匹配，同时在预测的不确定性与分割误差之间产生显著更高的空间对应性，这一优势在对所有基线应用温度缩放后仍然存在。下游任务中，它提供了更准确和稳定的射血分数估计，并在婴儿子群体中更可靠地检测心脏功能障碍。我们的结果表明，OOD-aware的不确定性量化可以作为部署分割模型的实用安全层，使得在代表性不足的子群体中检测静默失败成为可能，而无需再训练或额外的标记数据。
cs.CV / 80 / 2608.10908
Order Matters: LVLMs as Judges for Temporal Reasoning in Image Sequences
顺序重要：大规模视觉语言模型作为图像序列时间推理的评判者
Abstract
As generative multimedia evolves from static image synthesis to complex, interleaved visual narratives, a foundational bottleneck has emerged: the judgment crisis. While human perception naturally synthesizes the temporal and logical flow of a story, automated evaluation systems remain largely "blind" to sequential continuity, often failing to distinguish between a coherent narrative and a semantically shuffled or contradictory sequence. This work identifies a critical structural gap in current multimodal evaluation paradigms, arguing that the reliance on Large Vision-Language Models (LVLMs) as judges is fundamentally limited by architectural biases. Our analysis reveals a profound performance dichotomy: while models may appear competent in isolated pointwise scoring, they suffer a catastrophic collapse when required to perform pairwise discrimination of temporal order. We demonstrate that this is not merely a data-scarcity issue but a structural one. Through a series of diagnostic probes, we uncover systematic positional asymmetries, specifically primacy and recency effects, where a model's judgment of a story is significantly influenced by the placement of a frame, often more than by its semantic consistency. These biases, potentially rooted in causal masking and rotary embeddings, suggest that current transformer-based judges are inherently ill-equipped for long-form visual reasoning. By exposing these blind spots, we challenge the multimedia community to move beyond snapshot-centric metrics and instead pioneer Temporally-Aware Evaluation paradigms that treat visual sequences as unified logical structures rather than unordered collections of frames.
Chinese Translation
随着生成多媒体从静态图像合成演变为复杂的交织视觉叙事，一个基础性的瓶颈出现了：判断危机。尽管人类感知自然地合成故事的时间和逻辑流，但自动评估系统在很大程度上对顺序连续性“失明”，常常无法区分连贯叙事与语义混乱或矛盾的序列。本研究识别出当前多模态评估范式中的一个关键结构性缺口，认为依赖大型视觉语言模型（Large Vision-Language Models, LVLMs）作为评判者的做法在根本上受到架构偏见的限制。我们的分析揭示了深刻的性能二分法：尽管模型在孤立的逐点评分中看似表现出色，但在需要进行时间顺序的成对区分时却会遭遇灾难性的崩溃。我们证明这不仅仅是数据稀缺的问题，而是一个结构性的问题。通过一系列诊断探测，我们发现系统性的位置信不对称，特别是首因效应和近因效应，即模型对故事的判断显著受框架位置的影响，往往超过其语义一致性。这些偏见可能根植于因果掩蔽和旋转嵌入，表明当前基于变换器的评判者在长篇视觉推理方面固有地不具备能力。通过揭示这些盲点，我们挑战多媒体社区超越快照中心的度量，开创时间感知评估范式，将视觉序列视为统一的逻辑结构，而非无序的帧集合。
cs.CV / 81 / 2608.10932
Temporally Grounded Compositional Camera Motion Understanding via Geometric Knowledge Distillation
基于几何知识蒸馏的时序基础组合相机运动理解
Abstract
Understanding camera motion is fundamental to video perception, with applications in spatial intelligence and controllable video generation. Multimodal large language models (MLLMs) provide a natural interface for this task, but existing work typically assigns one or more labels to an entire clip. Such clip-level recognition overlooks two defining properties of real camera motion: it can change within a shot, and multiple movements can occur simultaneously. We therefore formulate camera-motion understanding as temporally grounded, compositional recognition, which requires a model to localize motion-consistent intervals and identify every movement active within each interval. We introduce CamChoreo, a benchmark of 4,229 real single-shot clips with expert-annotated temporal segments. Its annotations use a compact vocabulary of 20 direction-aware labels, and nearly half of the segments contain compound camera motion, with multiple movement primitives active simultaneously. Recognizing such fine-grained, compositional motion is hard for current MLLMs, whose visual encoders emphasize semantic content rather than the geometric evidence on which camera motion depends. Directly injecting features from a frozen 3D foundation model addresses this gap, but requires running the expensive geometry model on every input; we refer to this baseline as CamInject. We instead propose CamDistill, which distills the same geometric knowledge into lightweight camera tokens during training and removes the 3D model at inference. CamDistill matches the accuracy of direct feature injection without running the 3D teacher at inference. Together, CamChoreo and CamDistill advance camera-motion understanding from clip-level labeling to temporally grounded, compositional recognition. Project page: https://ddz16.github.io/cammotion.github.io/.
Chinese Translation
理解相机运动是视频感知的基础，具有空间智能和可控视频生成等应用。多模态大型语言模型（MLLMs）为这一任务提供了自然的接口，但现有研究通常将一个或多个标签分配给整个片段。这种片段级的识别忽视了真实相机运动的两个定义特性：它可以在一个镜头内变化，并且多个运动可以同时发生。因此，我们将相机运动理解形式化为时序基础的组合识别，这要求模型能够定位运动一致的区间，并识别每个区间内活跃的每个运动。我们引入了CamChoreo，这是一个包含4,229个真实单镜头片段的基准数据集，具有专家注释的时间段。其注释使用了20个方向感知标签的紧凑词汇，近一半的时间段包含复合相机运动，多个运动原语同时活跃。识别这种细粒度的组合运动对于当前的MLLMs来说是困难的，因为它们的视觉编码器强调语义内容而非相机运动所依赖的几何证据。直接从一个冻结的3D基础模型中注入特征可以解决这一问题，但需要在每个输入上运行昂贵的几何模型；我们将这一基线称为CamInject。相反，我们提出了CamDistill，它在训练过程中将相同的几何知识蒸馏到轻量级的相机标记中，并在推理时去除3D模型。CamDistill在推理时无需运行3D教师，便能匹配直接特征注入的准确性。总的来说，CamChoreo和CamDistill将相机运动理解从片段级标注推进到时序基础的组合识别。项目页面：https://ddz16.github.io/cammotion.github.io/
cs.CV / 82 / 2608.10933
SafeCA: Safe Cross-Attention Localization and Regulation for Text-to-Video Jailbreak Defense
SafeCA：文本到视频越狱防御的安全交叉注意力定位与调节
Abstract
Text-to-Video (T2V) generative models are vulnerable to jailbreak attacks in real-world deployment, leading them to produce harmful or inappropriate content. Existing defense approaches mainly rely on input filtering or reconstruction, which not only incur high computational latency but also tend to distort semantics. To address these issues, we experimentally and systematically analyze the differences between clean and jailbreak samples in the cross-attention feature space, revealing for the first time a cumulative separation effect and a progressively increasing trend of linear separability between the two during the diffusion process. Based on this insight, we propose SafeCA, a feature-level defense mechanism for safe cross-attention localization and regularization. Firstly, we identify key defensive regions and values through attention stability analysis using cross-attention features collected from clean prompts within a single inference. Secondly, SafeCA mitigates anomalous activations via attention masking with energy normalization and introduces a lightweight semantic-space adapter to redirect abnormal semantic flows. Furthermore, we detect and suppress potentially malicious tokens by back-propagating feature anomaly signals to the input cue words, thereby enhancing the deployability of the defense in commercial models. Experimental results show that SafeCA reduces the jailbreak success rate by about 20% on mainstream T2V models, adds almost no inference overhead (+0.1s), and maintains good text-video semantic consistency. Overall, SafeCA provides an architecture-level, deployable protection paradigm for T2V generation models.
Chinese Translation
文本到视频（T2V）生成模型在实际部署中容易受到越狱攻击，导致生成有害或不当内容。现有的防御方法主要依赖于输入过滤或重构，这不仅会导致高计算延迟，还往往会扭曲语义。为了解决这些问题，我们在交叉注意力特征空间中实验性和系统性地分析了干净样本与越狱样本之间的差异，首次揭示了累积分离效应以及在扩散过程中两者之间线性可分性逐渐增加的趋势。基于这一见解，我们提出了SafeCA，一种用于安全交叉注意力定位和正则化的特征级防御机制。首先，我们通过对单次推理中收集的干净提示的交叉注意力特征进行注意力稳定性分析，识别出关键的防御区域和数值。其次，SafeCA通过能量归一化的注意力掩蔽来减轻异常激活，并引入轻量级语义空间适配器以重定向异常语义流。此外，我们通过将特征异常信号反向传播到输入提示词，检测并抑制潜在的恶意标记，从而增强防御在商业模型中的可部署性。实验结果表明，SafeCA在主流T2V模型上将越狱成功率降低了约20%，几乎不增加推理开销（+0.1秒），并保持良好的文本-视频语义一致性。总体而言，SafeCA为T2V生成模型提供了一种架构级的、可部署的保护范式。
cs.CV / 83 / 2608.10938
GS-CPE: Unified 6-Degree-of-Freedom Camera Pose Estimation via 3D Gaussian Splatting
GS-CPE：通过3D高斯溅射实现统一的6自由度相机位姿估计
Abstract
Despite substantial progress in visual localization, from scene coordinate regression to direct camera pose regression, achieving both robust generalization and high accuracy remain challenging. This study introduces GS-CPE (Gaussian Splatting based Camera Pose Estimation), a coarse-to-fine framework for 6-DoF camera pose estimation that unifies geometry-based coarse pose estimation with robust 3D Gaussian Splatting (3DGS) warping based pose refinement. GS-CPE first estimates a coarse pose via retrieval-guided geometric pose estimation on a 3DGS scene representation, then refines it by minimizing a visibility aware masked RGB warping objective in a multi-scale optimization framework, with adaptive re-rendering. Extensive experiments on indoor and outdoor benchmarks including 7Scenes, Cambridge Landmarks, FAST-LIVO2 datasets, and a custom dataset demonstrate state-of-the-art performance, consistently outperforming in both accuracy and generalization.
Chinese Translation
尽管在视觉定位方面取得了显著进展，从场景坐标回归到直接相机位姿回归，但实现强健的泛化能力和高精度仍然具有挑战性。本研究提出了GS-CPE（基于高斯溅射的相机位姿估计），这是一个用于6自由度相机位姿估计的粗到细框架，统一了基于几何的粗略位姿估计与基于稳健的3D高斯溅射（3DGS）变形的位姿精细化。GS-CPE首先通过在3DGS场景表示上进行检索引导的几何位姿估计来估计粗略位姿，然后通过在多尺度优化框架中最小化考虑可见性的掩蔽RGB变形目标进行精细化，并进行自适应重渲染。在包括7Scenes、剑桥地标、FAST-LIVO2数据集和自定义数据集在内的室内外基准测试中，广泛的实验表明，GS-CPE在准确性和泛化能力上均表现出最先进的性能。
cs.CV / 84 / 2608.10947
Mixture-of-Experts-based Entropy Model for Learned Image Compression
基于专家混合的学习图像压缩熵模型
Abstract
Learned image compression has seen significant progress in recent years with the development of end-to-end learned models that achieve better compression efficiency than state-of-the-art conventional methods. Recently, Mixture of Experts (MoE) approaches have seen promising results in NLP and computer vision tasks. In this paper, we introduce the MoE approach to learned image compression. We propose a MoE-based Entropy model (MoEE) for learned image compression, allowing the model to selectively activate only the subset of parameters required for the input image. Our model achieves a BD-Rate improvement over VVC of -16.85% on the Kodak dataset.
Chinese Translation
近年来，随着端到端学习模型的发展，学习图像压缩取得了显著进展，这些模型在压缩效率上超过了最先进的传统方法。最近，专家混合（Mixture of Experts, MoE）方法在自然语言处理和计算机视觉任务中取得了良好的效果。本文将MoE方法引入到学习图像压缩中。我们提出了一种基于MoE的熵模型（MoEE），该模型允许根据输入图像选择性地激活所需的参数子集。我们的模型在Kodak数据集上相较于VVC实现了-16.85%的BD-Rate改进。
cs.CV / 85 / 2608.10949
StreamFlow: Dynamic Memory Flows for Streaming Video Understanding
StreamFlow：用于流媒体视频理解的动态记忆流
Abstract
Streaming video understanding requires multimodal large language models (MLLMs) to preserve relevant evidence from continuously evolving streams under strict causality and bounded memory. Yet existing paradigms remain limited: model-based methods require intrusive backbone updates, while memory-based methods expend substantial visual-encoding computation on temporally redundant content and rely on rigid access to visual history. To address these limitations, we introduce StreamFlow, an efficient visual memory framework that enables dynamic, on-demand access to historical visual information. StreamFlow combines a lightweight, dynamics-aware mid-term memory that filters temporal redundancy before visual encoding with a latent long-term memory that consolidates historical video content into visual latents accessible to subsequent reasoning. During generation, an attention-guided retrieval mechanism injects relevant visual latents when the model's reliance on visual evidence weakens. StreamFlow achieves state-of-the-art streaming video understanding performance, reaching 67.73% overall accuracy on StreamingBench, while also delivering strong performance on offline long-video benchmarks. Relative to the vanilla setting, it improves the visual attention score (VAS) by 59.1% while reducing end-to-end latency and peak memory by 50.4% and 21.1%, respectively, enabling more visually grounded and efficient reasoning.
Chinese Translation
流媒体视频理解需要多模态大型语言模型（MLLMs）在严格因果关系和有限记忆的条件下保留来自不断演变流的相关证据。然而，现有范式仍然存在局限性：基于模型的方法需要侵入性的骨干更新，而基于记忆的方法则在时间上冗余的内容上消耗大量视觉编码计算，并依赖于对视觉历史的刚性访问。为了解决这些限制，我们提出了StreamFlow，这是一种高效的视觉记忆框架，能够动态、按需访问历史视觉信息。StreamFlow结合了一种轻量级、动态感知的中期记忆，在视觉编码之前过滤时间冗余，以及一种潜在的长期记忆，将历史视频内容整合为可供后续推理访问的视觉潜变量。在生成过程中，当模型对视觉证据的依赖减弱时，注意力引导的检索机制会注入相关的视觉潜变量。StreamFlow在流媒体视频理解性能上达到了最先进的水平，在StreamingBench上整体准确率达到67.73%，同时在离线长视频基准测试中也表现出色。与基础设置相比，它将视觉注意力得分（VAS）提高了59.1%，同时将端到端延迟和峰值内存分别减少了50.4%和21.1%，从而实现了更具视觉基础和高效的推理。
cs.CV / 86 / 2608.10952
Multiple Scale Latents for Learned Image Compression
用于学习图像压缩的多尺度潜变量
Abstract
Most learned image compression systems rely on a single latent representation combined with a hyperprior, which limits their ability to efficiently capture image structure across spatial scales. In this work, we propose a hierarchical latent representation to improve the efficiency of the entropy model. By using multiple latents at different scales, each with its own entropy model, we better capture the spatial structure of the latent representation. Our experiments show that this approach achieves a 17.9% BD-rate reduction over VVC on Kodak, demonstrating the effectiveness of multi-scale latent representations. Furthermore, the approach is orthogonal to other advances in learned image compression, making it a versatile addition to existing methods.
Chinese Translation
大多数学习图像压缩系统依赖于单一的潜在表示和超先验，这限制了它们在不同空间尺度上有效捕捉图像结构的能力。在本研究中，我们提出了一种分层潜在表示，以提高熵模型的效率。通过在不同尺度上使用多个潜在变量，每个变量都有其自己的熵模型，我们更好地捕捉了潜在表示的空间结构。我们的实验表明，这种方法在Kodak数据集上相较于VVC实现了17.9%的BD-rate降低，证明了多尺度潜在表示的有效性。此外，该方法与学习图像压缩的其他进展是正交的，使其成为现有方法的一个多功能补充。
cs.CV / 87 / 2608.10954
Evidence-Grounded Trustworthy Multimodal Reasoning and Evaluation Benchmark in Complex Urban Scenes
基于证据的可信多模态推理与复杂城市场景评估基准
Abstract
While Multimodal Large Language Models (MLLMs) demonstrate impressive performance in benign scenarios, their cognitive reliability deteriorates significantly in complex scenes under adverse conditions. In these settings, models often rely on implicit inference without sufficient visual evidence, leading to a disconnect between perception and reasoning. Meanwhile, existing outcome-oriented benchmarks evaluate only final predictions and fail to diagnose failures in the underlying reasoning process. To address this gap, the authors propose AD2-Bench, which introduces a Hierarchical Visual Diagnosis framework that decomposes reasoning into a structured Chain of Evidence (CoE). This fine-grained diagnosis reveals that robust multimodal reasoning fundamentally depends on accurate evidence acquisition. Building on this perspective, the authors formulate reasoning from a probabilistic viewpoint and identify two primary causes of reasoning failure: Spatial Ambiguity, where models fail to distinguish target objects from background clutter, resulting in localization errors; and Semantic Uncertainty, where degraded visual features lead to incorrect semantic interpretation, resulting in understanding errors. To overcome these evidence deficiencies, they further propose Evidence-grounded Visual Reasoning (EGVOR), which replaces implicit reasoning with the explicit generation of Evidence Atoms - structured spatial-semantic triplets that enforce tight alignment between localization and semantic understanding. The model is trained through a hierarchical curriculum that progresses from reflective supervision construction to reinforcement learning, where reducing reasoning variance is explicitly rewarded. Extensive experiments demonstrate that EGVOR substantially improves reasoning stability under adverse conditions, providing a more robust framework for trustworthy multimodal cognition.
Chinese Translation
尽管多模态大型语言模型（MLLMs）在良性场景中表现出色，但在复杂场景下的不利条件下，它们的认知可靠性显著下降。在这些环境中，模型往往依赖于隐式推理而没有足够的视觉证据，导致感知与推理之间的脱节。同时，现有的以结果为导向的基准仅评估最终预测，未能诊断潜在推理过程中的失败。为了解决这一问题，作者提出了AD2-Bench，引入了一种分层视觉诊断框架，将推理分解为结构化的证据链（Chain of Evidence, CoE）。这种细粒度的诊断揭示了稳健的多模态推理在根本上依赖于准确的证据获取。基于这一视角，作者从概率的角度构建推理，并识别出推理失败的两个主要原因：空间模糊性（Spatial Ambiguity），即模型无法区分目标物体与背景杂乱，导致定位错误；以及语义不确定性（Semantic Uncertainty），即退化的视觉特征导致错误的语义解释，造成理解错误。为了克服这些证据缺失，他们进一步提出了基于证据的视觉推理（Evidence-grounded Visual Reasoning, EGVOR），该方法用显式生成证据原子（Evidence Atoms）替代隐式推理——这些结构化的空间-语义三元组强制实现定位与语义理解之间的紧密对齐。该模型通过分层课程进行训练，从反思性监督构建到强化学习，其中显著降低推理方差将被明确奖励。大量实验表明，EGVOR在不利条件下显著提高了推理的稳定性，为可信的多模态认知提供了更为稳健的框架。
cs.CV / 88 / 2608.10959
Once Poisoned, Arbitrarily Controlled: A Programmable Backdoor in VLMs
一旦被毒化，任意控制：视觉语言模型中的可编程后门
Abstract
Existing vision-language model (VLM) backdoors are usually treated as static vulnerabilities: one-to-one and N-to-N attacks bind one or more triggers to a finite set of targets before victim training. This assumption substantially underestimates the threat. We show that a single poisoning phase can implant a programmable backdoor into a VLM, allowing an attacker to choose previously unseen target-caption semantics at inference time and synthesize corresponding stealthy triggers on demand. Unlike fixed-mapping attacks, the proposed any-to-any caption-control paradigm decouples post-training target selection from poisoning, enabling dynamic control of target captions without retraining the VLM. Our method has two components. First, a heuristic poisoning strategy exposes the model to diverse trigger-caption pairs, encouraging it to learn a general trigger-as-instruction rule rather than memorize a specific backdoor pattern. Second, a feature-space trigger steganography method maps any attacker-specified target caption to a stealthy visual trigger, implemented as either a norm-controlled perturbation or a non-semantic patch. Once inserted into arbitrary images, these triggers cause the poisoned VLM to generate outputs semantically aligned with the chosen target caption, even when the target was unseen during poisoning. Extensive experiments show that our attack achieves high any-to-any caption-control success rates, preserves clean model utility, and remains effective under several classical backdoor defenses.
Chinese Translation
现有的视觉语言模型（VLM）后门通常被视为静态漏洞：一对一和N对N攻击在受害者训练之前将一个或多个触发器绑定到有限的目标集。这一假设大大低估了威胁。我们展示了单一的毒化阶段可以将可编程后门植入VLM，使攻击者能够在推理时选择先前未见的目标-标题语义，并按需合成相应的隐蔽触发器。与固定映射攻击不同，所提出的任意对任意标题控制范式将训练后目标选择与毒化解耦，允许在不重新训练VLM的情况下动态控制目标标题。我们的方法包含两个组成部分。首先，一种启发式毒化策略使模型接触到多样的触发器-标题对，鼓励其学习一般的触发器作为指令规则，而不是记忆特定的后门模式。其次，一种特征空间触发器隐写方法将任何攻击者指定的目标标题映射到隐蔽的视觉触发器，该触发器可以实现为规范控制的扰动或非语义补丁。一旦插入任意图像，这些触发器使得被毒化的VLM生成与所选目标标题语义对齐的输出，即使在毒化过程中未见过该目标。大量实验表明，我们的攻击实现了高任意对任意标题控制的成功率，保持了干净模型的效用，并在多种经典后门防御下仍然有效。
cs.CV / 89 / 2608.10964
CARE: Confidence-Aware Reasoning for Reliable Medical VQA
CARE：基于信心的可靠医学视觉问答推理
Abstract
Reinforcement Fine-Tuning (RFT) has enabled medical Multimodal Large Language Models (MLLMs) to produce Chain-of-Thought (CoT) reasoning for visual question answering, yet these models suffer from $\textit{confidence miscalibration}$---a systematic gap between expressed certainty and actual diagnostic accuracy that undermines clinical trust. We propose $\textbf{CARE}$, a $\textbf{C}$onfidence-$\textbf{A}$ware medical $\textbf{RE}$asoning framework that jointly optimizes accuracy and calibration through a dual-stage pipeline. First, a scalable Medical-CoT synthesis provides structured cold-start data for Supervised Fine-Tuning. Second, Group Relative Policy Optimization (GRPO) with a novel $\textbf{Confidence-Aware Reward (CAR)}$ mechanism ties the model's confidence to diagnostic correctness within the reward signal. Across three Medical VQA benchmarks, $\textbf{CARE}$ achieves the highest diagnostic accuracy while obtaining the lowest Expected Calibration Error and Hallucination Rate, establishing a foundation for trustworthy clinical decision support. Our code is available at https://github.com/anotherbricki/CARE.
Chinese Translation
强化微调（Reinforcement Fine-Tuning, RFT）使医学多模态大语言模型（Multimodal Large Language Models, MLLMs）能够为视觉问答生成思维链（Chain-of-Thought, CoT）推理，然而这些模型存在$ extit{信心失调（confidence miscalibration）}$的问题——即表达的确定性与实际诊断准确性之间的系统性差距，这削弱了临床信任。我们提出了$ extbf{CARE}$，一个$ extbf{C}$onfidence-$ extbf{A}$ware医学$ extbf{RE}$asoning框架，通过双阶段管道联合优化准确性和校准。首先，扩展的医学思维链合成（Medical-CoT synthesis）为监督微调提供结构化的冷启动数据。其次，采用新的$ extbf{信心感知奖励（Confidence-Aware Reward, CAR）}$机制的群体相对策略优化（Group Relative Policy Optimization, GRPO）将模型的信心与奖励信号中的诊断正确性联系起来。在三个医学视觉问答基准测试中，$ extbf{CARE}$实现了最高的诊断准确性，同时获得了最低的期望校准误差和幻觉率，为可信的临床决策支持奠定了基础。我们的代码可在https://github.com/anotherbricki/CARE获取。
cs.CV / 90 / 2608.10978
A Dataset and Benchmark for Optical Music Recognition of String Quartet Scores
用于弦乐四重奏乐谱的光学音乐识别数据集与基准测试
Abstract
Optical music recognition (OMR) transcribes music scores into digital formats. While the field has advanced significantly on monophonic and piano-form scores, multi-part score transcription remains underexplored, largely due to the absence of a suitable dataset. We introduce OpenScore String Quartet for Optical Music Recognition (OSSQ-OMR), the first dataset dedicated to multi-part OMR. Built on the OpenScore String Quartet corpus, OSSQ-OMR pairs digitally encoded scores with their original scanned editions from IMSLP, with all images visually aligned to their transcriptions. The dataset is released with score images at system and staff levels, and paired transcriptions in three encoding formats: Extended Linearized MusicXML (LMXE), **kern, and ABC. In total, OSSQ-OMR contains 24,544 system images and 98,172 staff images drawn from 116 string quartet scores. We accompany the dataset with a benchmark protocol and baseline results from two representative OMR models, evaluated across four random score-level splits with mutually exclusive test sets. Baselines reach OMR-NED as low as 3.6% on synthetic and 5.9% on scanned inputs; results reveal substantial effects of encoding and segmentation choices, with the LSTM-based baseline degrading on scanned inputs roughly 2.6 times less than the Transformer-based baseline.
Chinese Translation
光学音乐识别（OMR）将音乐乐谱转录为数字格式。尽管该领域在单声部和钢琴乐谱方面取得了显著进展，但多声部乐谱的转录仍然未得到充分探索，这在很大程度上是由于缺乏合适的数据集。我们介绍了用于光学音乐识别的开放乐谱弦乐四重奏数据集（OpenScore String Quartet for Optical Music Recognition，OSSQ-OMR），这是第一个专门针对多声部OMR的数据集。OSSQ-OMR基于开放乐谱弦乐四重奏语料库，配对了数字编码的乐谱与其来自IMSLP的原始扫描版本，所有图像均与其转录版本在视觉上对齐。该数据集发布了系统级和五线谱级的乐谱图像，并提供了三种编码格式的配对转录：扩展线性音乐XML（Extended Linearized MusicXML，LMXE）、**kern和ABC。总的来说，OSSQ-OMR包含来自116部弦乐四重奏乐谱的24,544个系统图像和98,172个五线谱图像。我们为该数据集提供了基准测试协议和来自两个代表性OMR模型的基线结果，这些结果在四个随机乐谱级别的划分中进行了评估，测试集相互独立。基线在合成输入上达到最低3.6%的OMR-NED，在扫描输入上达到5.9%；结果揭示了编码和分割选择的显著影响，基于LSTM的基线在扫描输入上的降级约为基于Transformer的基线的2.6倍。
cs.CV / 91 / 2608.10981
ThinkAfford: Affordance-Centric Reasoning for Fine-Grained 3D Grounding in Cluttered Scenes
ThinkAfford：基于可供性中心的细粒度3D定位在杂乱场景中的推理
Abstract
Task-driven 3D affordance grounding aims to localize the functional region in a cluttered 3D scene that enables an action specified by a natural-language instruction. Existing methods either predict 3D masks directly or construct them by selecting and fusing intermediate 2D/3D regions. However, they remain vulnerable to two intertwined failure modes: the predicted or selected regions may miss the target interaction area or have unsuitable granularity, while language grounding may confuse visually similar alternatives under relational instructions. To this end, we introduce ThinkAfford, which decouples high-recall affordance proposal generation from instruction-grounded reasoning. Specifically, the Affordance Proposal Generation module first uses learnable affordance prompts and multi-level visual features to predict interaction-conditioned heatmaps, extracting a variable number of fine-grained proposals without parsed object or part names as segmentation prompts. Visual-Prompted Affordance Reasoning then reasons over labeled proposal overlays using the full instruction, returning identifiers in a structured "think-then-answer" response. Moreover, Group Relative Policy Optimization uses proposal-level rewards from lifted 3D overlap to align VPAR selection with final 3D grounding. On the SceneFun3D validation split, ThinkAfford achieves 10.69% AP50 and 25.46% AP25 under the official evaluator, outperforming comparable 3D open-vocabulary and vision-language-model-based 2D-to-3D baselines. Module-level diagnostics further show that APG attains 77.5% recall at 25% intersection-over-union, while GRPO-trained VPAR achieves 72.1% selection accuracy on APG-covered queries, compared with 63.4% under supervised fine-tuning.
Chinese Translation
任务驱动的3D可供性定位旨在在杂乱的3D场景中定位能够执行自然语言指令所指定的动作的功能区域。现有方法要么直接预测3D掩膜，要么通过选择和融合中间的2D/3D区域来构建它们。然而，它们仍然容易受到两种交织的失败模式的影响：预测或选择的区域可能错过目标交互区域或具有不适当的粒度，而语言定位可能在关系指令下混淆视觉上相似的替代选项。为此，我们提出了ThinkAfford，它将高召回率的可供性提议生成与指令驱动的推理解耦。具体而言，可供性提议生成模块首先使用可学习的可供性提示和多层次视觉特征来预测交互条件热图，提取可变数量的细粒度提议，而无需解析的对象或部分名称作为分割提示。视觉提示可供性推理随后使用完整指令对标记的提议叠加进行推理，返回结构化的“思考-然后回答”响应中的标识符。此外，组相对策略优化利用来自提升的3D重叠的提议级奖励，使得VPAR选择与最终的3D定位对齐。在SceneFun3D验证集上，ThinkAfford在官方评估器下实现了10.69%的AP50和25.46%的AP25，超越了可比较的3D开放词汇和基于视觉语言模型的2D到3D基线。模块级诊断进一步显示，APG在25%的交并比下达到了77.5%的召回率，而经过GRPO训练的VPAR在APG覆盖的查询上达到了72.1%的选择准确率，而在监督微调下仅为63.4%。
cs.CV / 92 / 2608.10985
PEAK: Precise and Persistent Concept Erasure via k-Sparse Autoencoders
PEAK：通过k稀疏自编码器实现精确且持久的概念抹除
Abstract
Erasing concepts from large-scale text-to-image (T2I) diffusion models has become increasingly crucial due to the growing concerns over copyright infringement, privacy violations, and offensive content. Existing approaches struggle to achieve both precise and persistent concept erasure: inaccurate localization of concept-related representations may cause unintended semantic interference, while incomplete removal of the underlying concept knowledge allows adversarial recovery. To address this dilemma, we propose PEAK, a \textbf{\textit{precise}} and \textbf{\textit{persistent}} concept erasure framework via k-Sparse Autoencoders (kSAEs). PEAK first trains a kSAE on internal activations of the diffusion denoising network to decompose dense representations into interpretable sparse features. By contrasting sparse activations induced by target and non-target prompts, PEAK identifies a compact set of target-specific features according to both activation strength and frequency. These localized features are then used for parameter optimization, where PEAK selectively suppresses target-related activations while preserving complementary non-target ones towards the original model. This feature-guided optimization embeds concept erasure directly into diffusion parameters, eliminating the need for additional inference-time intervention and facilitating effective persistence against adversarial attacks. Extensive experiments demonstrate that PEAK achieves effective and robust concept erasure. On the I2P benchmark, PEAK reduces NudeNet detections from 582 to 6, lowers the average attack success rate (ASR) from 96.52\% to 5.63\%, and preserves general generation quality on MS-COCO with a near-zero KID. Our code and models are available at: https://github.com/manmanTAT/PEAK
Chinese Translation
由于对版权侵犯、隐私违规和攻击性内容的日益关注，从大规模文本到图像（T2I）扩散模型中抹除概念变得愈加重要。现有方法在实现精确和持久的概念抹除方面面临挑战：概念相关表示的定位不准确可能导致意外的语义干扰，而对潜在概念知识的不完全去除则允许对抗性恢复。为了解决这一困境，我们提出了PEAK，一个通过k稀疏自编码器（kSAEs）实现的 extbf{ extit{精确}}且 extbf{ extit{持久}}的概念抹除框架。PEAK首先在扩散去噪网络的内部激活上训练kSAE，以将密集表示分解为可解释的稀疏特征。通过对比目标和非目标提示所诱导的稀疏激活，PEAK根据激活强度和频率识别出一组紧凑的目标特征。这些定位特征随后用于参数优化，PEAK在保留与原始模型相关的非目标激活的同时，选择性地抑制与目标相关的激活。这种特征引导的优化将概念抹除直接嵌入到扩散参数中，消除了对额外推理时间干预的需求，并有效增强了对抗攻击的持久性。大量实验表明，PEAK实现了有效且稳健的概念抹除。在I2P基准测试中，PEAK将NudeNet的检测从582减少到6，将平均攻击成功率（ASR）从96.52\%降低到5.63\\%，并在MS-COCO上保持了接近零的KID的总体生成质量。我们的代码和模型可在以下链接获取：https://github.com/manmanTAT/PEAK
cs.CV / 93 / 2608.10989
Putting Registers to Work: Task Registers for Token Pruning in Vision Transformers
发挥寄存器的作用：用于视觉变换器中的标记修剪的任务寄存器
Abstract
Token-pruning policies are usually designed for a single recognition pipeline, but pretrained Vision Transformers are reused across tasks with different spatial demands. We ask which parts of a pruning policy transfer across image classification, semantic segmentation, and object detection. For each pipeline, controlled probes freeze the no-pruning checkpoint and apply a series of parameter-free reduction criteria at one eligible layer at a time without retraining. The probes reveal three differences: segmentation and detection rank the criteria differently, classification is especially sensitive to attention-based pruning in the earliest layers, and the dense tasks prefer opposite recovery endpoints. These findings motivate Task-Adaptive Pruning (TAP). Existing register tokens serve as task-agnostic storage for feature artifacts. TAP instead introduces one task register per task and activates only the current one. Its evolving state ranks tokens, distributes an exact removal budget over depth, and sets the recovery scale for dense features. At a final keep rate of $\rho=0.5$, our jointly adapted model, TAP-J, reaches $47.0$ mIoU at $1.30\times$ encoder throughput on ADE20K and $53.7$ box AP at $1.32\times$ encoder throughput on COCO while remaining competitive on ImageNet-1K.
Chinese Translation
标记修剪策略通常是为单一识别流程设计的，但预训练的视觉变换器在不同空间需求的任务中被重复使用。我们探讨了修剪策略的哪些部分可以在图像分类、语义分割和目标检测之间迁移。对于每个流程，控制探针冻结无修剪的检查点，并在一个合适的层上逐次应用一系列无参数的减少标准，而无需重新训练。这些探针揭示了三个差异：分割和检测对标准的排名不同，分类对最早层的基于注意力的修剪特别敏感，而密集任务则偏好相反的恢复端点。这些发现激励了任务自适应修剪（Task-Adaptive Pruning, TAP）。现有的寄存器标记作为特征工件的任务无关存储，而TAP则为每个任务引入一个任务寄存器，并仅激活当前的寄存器。其不断演变的状态对标记进行排名，分配精确的移除预算并设置密集特征的恢复规模。在最终保留率为$
ho=0.5$时，我们联合适应的模型TAP-J在ADE20K上以$1.30 imes$编码器吞吐量达到了$47.0$ mIoU，在COCO上以$1.32 imes$编码器吞吐量达到了$53.7$框AP，同时在ImageNet-1K上保持竞争力。
cs.CV / 94 / 2608.10995
HNDiff: Haze-Noise Diffusion for Image Dehazing
HNDiff：用于图像去雾的雾噪声扩散
Abstract
Existing diffusion-based methods have recently made significant progress in image dehazing. However, they typically neglect the physics of haze formation and reconstruct clean images from pure Gaussian noise, thereby limiting their restoration potential. To address this issue, we propose Haze-Noise Diffusion (HNDiff), a novel diffusion framework that embeds the atmospheric scattering model as an inductive bias. By grounding diffusion in physical principles, HNDiff ensures that the restoration aligns more closely with underlying mechanisms of haze formation. In its forward process, we introduce joint haze-noise diffusion with a haze-aware noise scheduler, which progressively adds both haze and noise to an image. Essentially, the scheduler adapts noise levels according to haze density, meaning that regions with heavier haze receive stronger noise injection to encourage content generation, while clearer regions receive lighter noise to better preserve details, which directly links the forward degradation process with the physics of haze. In the reverse process, we then derive a physically consistent dehazing-denoising process that simultaneously removes haze and noise to restore a clean image in a manner aligned with the forward degradation process. To further enhance practicality, we propose Latent HNDiff, which compiles clean latent priors that can be seamlessly integrated into existing dehazing networks to boost performance. Extensive experiments show that our work significantly improves leading dehazing backbones and achieves state-of-the-art results on benchmark datasets. The project page is available at https://jin-ting-he.github.io/HNDiff .
Chinese Translation
现有的基于扩散的方法在图像去雾方面最近取得了显著进展。然而，它们通常忽视了雾的形成物理，并从纯高斯噪声中重建干净图像，从而限制了其恢复潜力。为了解决这个问题，我们提出了雾噪声扩散（HNDiff），这是一种新颖的扩散框架，将大气散射模型作为归纳偏置嵌入其中。通过将扩散建立在物理原理上，HNDiff确保恢复过程更紧密地与雾形成的基本机制对齐。在其前向过程中，我们引入了联合雾噪声扩散，并配备了雾感知噪声调度器，该调度器逐步向图像添加雾和噪声。实质上，调度器根据雾密度调整噪声水平，这意味着雾较重的区域会接收更强的噪声注入以促进内容生成，而较清晰的区域则接收较轻的噪声以更好地保留细节，这直接将前向退化过程与雾的物理联系起来。在反向过程中，我们推导出一个物理一致的去雾去噪过程，该过程同时去除雾和噪声，以一种与前向退化过程相一致的方式恢复干净图像。为了进一步增强实用性，我们提出了潜在HNDiff，它编译了可以无缝集成到现有去雾网络中的干净潜在先验，以提升性能。大量实验表明，我们的工作显著改善了领先的去雾骨干网络，并在基准数据集上取得了最先进的结果。项目页面可访问 https://jin-ting-he.github.io/HNDiff 。
cs.CV / 95 / 2608.11013
Watching Synthetic Videos: Aligning Cross-modal Representations with Visual Synthesis for Zero-shot Video Captioning
观看合成视频：通过视觉合成对齐跨模态表示以实现零样本视频字幕生成
Abstract
Text-only training is a popular paradigm in zero-shot video captioning, where the video distribution is not available to the model during training, leading to a cross-modal gap between the training (text-only) and the inference (video-only). Previous works attempt to bridge the gap through simple linear transformations. However, the inherent gap between text and video makes cross-modal representation space alignment insufficient, resulting in inaccurate sentences. To address this issue, we propose a novel zero-shot video captioning framework (WSV) consisting of two training stages, which first generates corresponding synthetic video latent representations via a pretrained text-to-video generation model. To strengthen the fidelity of the latent representations, we propose a polisher capable of bridging the gap between real and synthetic video distributions. Subsequently, we design a prompter that conditions GPT-2 on the polished latent representations to generate the captions in the second training stage. During inference, an input video is encoded by a pretrained 3D Causal VAE and then fed directly into the prompter, which in turn guides GPT-2 to produce the final caption. Experimental results conducted on MSVD, MSR-VTT, and VATEX datasets demonstrate that our proposed method achieves scores of 52 and 95.7 on the B@4 and CIDEr metrics, respectively.
Chinese Translation
仅使用文本训练是零样本视频字幕生成中的一种流行范式，在此过程中，模型在训练期间无法获得视频分布，导致训练（仅文本）和推理（仅视频）之间存在跨模态差距。之前的研究尝试通过简单的线性变换来弥合这一差距。然而，文本与视频之间固有的差距使得跨模态表示空间的对齐不足，导致生成的句子不准确。为了解决这个问题，我们提出了一种新颖的零样本视频字幕生成框架（WSV），该框架由两个训练阶段组成，首先通过预训练的文本到视频生成模型生成相应的合成视频潜在表示。为了增强潜在表示的真实性，我们提出了一种能够弥合真实与合成视频分布之间差距的修饰器。随后，我们设计了一种提示器，使其在第二个训练阶段中根据修饰后的潜在表示对 GPT-2 进行条件处理，以生成字幕。在推理过程中，输入视频通过预训练的 3D 因果变分自编码器进行编码，然后直接输入提示器，提示器引导 GPT-2 生成最终字幕。在 MSVD、MSR-VTT 和 VATEX 数据集上进行的实验结果表明，我们提出的方法在 B@4 和 CIDEr 指标上分别达到了 52 和 95.7 的得分。
cs.CV / 96 / 2608.11017
R4DSG: Relative 4D Scene Graph Memory for Object-Centric Question Answering in Long Egocentric Video
R4DSG：用于长时间自我中心视频的面向对象问答的相对4D场景图记忆
Abstract
Long-horizon egocentric video is a rich substrate for wearable AI assistants, but object-centric questions such as where an item was moved, when it last changed state, or why it was relocated remain difficult because caption- and transcript-based memories rarely preserve persistent object identity or structured spatial change. Existing long-video QA methods mainly emphasize temporal grounding and clip retrieval, while prior 3D scene-graph methods typically assume stronger geometry than free-motion wearable RGB video provides, including point clouds, RGB-D input, posed views, sparse reconstruction, or reconstructed scenes. R4DSG introduces a relative 4D scene graph memory for long egocentric video. Instead of storing raw graph sequences, R4DSG converts video into compact queryable memory entries indexed by time, place, persistent objects, anchor-relative change, and local interaction context. The main idea is to separate stable anchors from dynamic objects, maintain persistent object identity across frames, and represent object state through anchor-relative transitions rather than a globally aligned world model. Built on recent RGB-only advances in promptable video segmentation, temporal propagation, and relative 3D lifting, the method produces a retrieval-ready memory directly usable for long-horizon question answering. Evaluation on a 255-question object-related subset from EgoLifeQA shows, under question-only retrieval, a 6.7-point overall gain over EgoRAG-Text and a 12.5-point gain on when questions, which highlights the value of temporally organized object memory. These results position relative 4D scene graphs as a practical memory substrate for wearable assistants, AR systems, and embodied multimedia agents. GitHub Page: https://dualtransparency.github.io/R4DSG/.
Chinese Translation
长时间自我中心视频为可穿戴人工智能助手提供了丰富的基础，但关于物体的中心问题，例如某个物品被移动到哪里、它最后一次状态变化的时间或为什么被重新安置等，仍然很难回答，因为基于字幕和转录的记忆很少能保持持久的物体身份或结构化的空间变化。现有的长视频问答方法主要强调时间基础和片段检索，而以往的3D场景图方法通常假设比自由运动的可穿戴RGB视频提供的几何信息更强，包括点云、RGB-D输入、姿态视图、稀疏重建或重建场景。R4DSG引入了一种用于长时间自我中心视频的相对4D场景图记忆。R4DSG并不是存储原始图形序列，而是将视频转换为按时间、地点、持久物体、锚点相对变化和局部交互上下文索引的紧凑可查询记忆条目。其主要思想是将稳定的锚点与动态物体分离，保持跨帧的持久物体身份，并通过锚点相对的过渡而不是全球对齐的世界模型来表示物体状态。该方法基于最近在可提示视频分割、时间传播和相对3D提升方面的进展，生成可直接用于长时间问答的检索准备记忆。在EgoLifeQA的255个与物体相关的问题子集上的评估显示，在仅基于问题的检索中，相较于EgoRAG-Text，整体提高了6.7分，而在时间问题上提高了12.5分，突显了时间组织的物体记忆的价值。这些结果将相对4D场景图定位为可穿戴助手、增强现实系统和具身多媒体代理的实用记忆基础。GitHub页面：https://dualtransparency.github.io/R4DSG/
cs.CV / 97 / 2608.11024
When Visual Signals Mislead: A Mechanistic Study of Attribute Hallucination in Vision-Language Models
当视觉信号误导时：视觉-语言模型中属性幻觉的机制研究
Abstract
Attribute hallucination---where vision-language models (VLMs) correctly identify an object but mischaracterize its properties---is prevalent yet mechanistically poorly understood. The dominant explanation, language-prior dominance, has motivated prior-suppression methods, but this explanation has not been directly tested at the attribute level. We present VISOR (Visual-Operational Remediation), a unified framework that couples null-image-based diagnosis with routed remediation. Its VSNR diagnostic decomposes each prediction into a visual logit signal and a language-prior signal. Across 10,791 negative-ground-truth samples from three VLM families and three attribute types, the visual signal strongly predicts false positives, whereas the language-prior signal is near chance. VISOR uses this diagnosis to separate two failure modes: low-margin but directionally correct visual signals in color/state attributes, and low-SNR or misaligned visual signals in material attributes. The same diagnosis routes each query to the appropriate operator: calibration for threshold-placement errors, abstention for training-free low-SNR handling, or targeted visual adaptation for material failures that prior suppression cannot correct. Across Qwen, InternVL, and LLaVA, VISOR reduces attribute false positives without relying on the prior-dominance assumption.
Chinese Translation
属性幻觉——即视觉-语言模型（VLMs）正确识别对象但错误表征其属性——普遍存在，但其机制尚不清楚。主导的解释是语言优先性主导，这一解释促使了先前抑制方法的提出，但该解释在属性层面上尚未得到直接验证。我们提出了VISOR（视觉操作修正），这是一个将基于空图像的诊断与路由修正相结合的统一框架。其VSNR诊断将每个预测分解为视觉logit信号和语言优先信号。在来自三个VLM家族和三种属性类型的10,791个负真实样本中，视觉信号强烈预测假阳性，而语言优先信号接近随机。VISOR利用这一诊断将两种失败模式分离：在颜色/状态属性中，低边际但方向正确的视觉信号，以及在材料属性中，低信噪比或不对齐的视觉信号。相同的诊断将每个查询路由到适当的操作符：用于阈值放置错误的校准，针对训练无关的低信噪比处理的弃权，或针对先前抑制无法纠正的材料失败的有针对性的视觉适应。在Qwen、InternVL和LLaVA中，VISOR在不依赖于优先主导假设的情况下减少了属性假阳性。
cs.CV / 98 / 2608.11037
Multi-Level Evidence Aggregation for Robust Facial Phenotype Retrieval in Rare Genetic Disorder Prioritization
多层次证据聚合用于罕见遗传疾病优先排序中的稳健面部表型检索
Abstract
AI-assisted facial phenotyping supports rare genetic disorder prioritization by retrieving visually similar diagnosed cases from facial image reference databases such as the GestaltMatcher Database (GMDB). Existing GestaltMatcher-based retrieval frameworks compare each test image with individual gallery images in a facial phenotype embedding space. However, this pointwise formulation does not fully exploit available evidence, because patients may have multiple images and disorders may be represented by multiple diagnosed gallery patients. We propose an inference-time multi-level evidence aggregation framework that improves facial phenotype retrieval without modifying the underlying GestaltMatcher-Arc encoder. The framework combines embedding-level patient aggregation of multiple images from the same individual, patient-weighted disorder centroids, and hybrid individual-centroid scoring to integrate test-patient observations, disorder-level gallery evidence, and local nearest-neighbor evidence. We evaluated the approach on GMDB v1.1.4 across disorders represented during training (GMDB-Freq), unseen disorders (GMDB-Rare), and multi-image patient subsets, using a unified gallery containing both GMDB-Freq and GMDB-Rare disorders. Multi-level evidence aggregation improved mean per-disorder top-$N$ retrieval accuracy across all evaluation subsets. Top-1 accuracy increased from 38.52% to 48.82% on GMDB-Freq and from 19.38% to 23.79% on GMDB-Rare. On multi-image subsets, top-1 accuracy increased from 46.12% to 60.94% on GMDB-Multi-Freq and from 18.54% to 26.71% on GMDB-Multi-Rare. These findings show that inference-time aggregation can improve next-generation facial phenotype retrieval without retraining the encoder, supporting a shift from isolated single-image matching toward multi-level aggregation of patient and disorder evidence for rare-disorder prioritization.
Chinese Translation
人工智能辅助的面部表型分析通过从面部图像参考数据库（如GestaltMatcher数据库，GMDB）中检索视觉上相似的已诊断病例，支持罕见遗传疾病的优先排序。现有的基于GestaltMatcher的检索框架在面部表型嵌入空间中将每个测试图像与单个图库图像进行比较。然而，这种逐点的公式并未充分利用可用证据，因为患者可能有多张图像，且疾病可能由多个已诊断的图库患者表示。我们提出了一种推理时多层次证据聚合框架，该框架在不修改基础GestaltMatcher-Arc编码器的情况下，改善了面部表型检索。该框架结合了来自同一患者的多张图像的嵌入级患者聚合、患者加权的疾病质心以及混合个体-质心评分，以整合测试患者观察、疾病级图库证据和局部最近邻证据。我们在GMDB v1.1.4上对该方法进行了评估，涵盖了训练期间表示的疾病（GMDB-Freq）、未见疾病（GMDB-Rare）和多图像患者子集，使用了包含GMDB-Freq和GMDB-Rare疾病的统一图库。多层次证据聚合提高了所有评估子集的每个疾病的平均前-$N$检索准确率。GMDB-Freq上的Top-1准确率从38.52%提高到48.82%，GMDB-Rare上的Top-1准确率从19.38%提高到23.79%。在多图像子集上，GMDB-Multi-Freq上的Top-1准确率从46.12%提高到60.94%，GMDB-Multi-Rare上的Top-1准确率从18.54%提高到26.71%。这些发现表明，推理时聚合可以在不重新训练编码器的情况下改善下一代面部表型检索，支持从孤立的单图像匹配向罕见疾病优先排序的患者和疾病证据的多层次聚合转变。
cs.CV / 99 / 2608.11050
3D Weighted Geometric Graph Neural Networks for Sheep Facial Pain Assessment
用于羊面部疼痛评估的3D加权几何图神经网络
Abstract
Deep learning systems perform mainly within the 2D for a single image domain and take the face as a single-dimension representation, losing sight of the 3D anatomy of sheep and cross-landmark spatial relationships that are intrinsic to the clinically proven Sheep Pain Facial Expression Scale (SPFES). This paper presents the \textbf{3D Sheep Pain Facial Expression System (3D-SPFES)}, a novel, monocular depth-aware geometric graph neural network system that integrates each SPFES facial landmark, such as the ears, eyes, and nose, into 3D Euclidean space estimated from a single RGB camera by using VideoDepthAnything, thus preventing the need for specialized depth hardware. Each landmark node includes a feature vector containing its 3D spatial coordinates, estimated surface normal, and facial attribute class embedding. Edges linked to nodes are assigned weights based on an aggregate metric that combines both Euclidean distance and surface co-planarity in a 3D space. A Weighted Geometric Graph Neural Network (WG-GNN) studies this graph using $\mathcal{K} = 3$ geometry-aware message-passing layers enhanced by a scaled dot-product attention method that selectively enhances anatomically relevant inter-landmark messages. The resultant node embeddings are combined into $\mathcal{O} = 3$ pain-level clusters and integrated into a Normalized Pain Score (NPS) within the range of $[0, 100%]$ a confidence-weighted, SPFES-derived scoring method.
Chinese Translation
深度学习系统主要在二维单图像领域内运行，将面部视为单维度表示，忽视了羊的三维解剖结构及与临床验证的羊疼痛面部表情量表（Sheep Pain Facial Expression Scale, SPFES）内在的跨标志空间关系。本文提出了 extbf{3D羊疼痛面部表情系统（3D-SPFES）}，这是一种新颖的单目深度感知几何图神经网络系统，它将每个SPFES面部标志（如耳朵、眼睛和鼻子）整合到通过单个RGB相机估计的三维欧几里得空间中，使用VideoDepthAnything，从而避免了对专用深度硬件的需求。每个标志节点包含一个特征向量，包含其三维空间坐标、估计的表面法线和面部属性类别嵌入。与节点相连的边根据一个聚合度量分配权重，该度量结合了三维空间中的欧几里得距离和表面共面性。加权几何图神经网络（Weighted Geometric Graph Neural Network, WG-GNN）使用$ extmath{K} = 3$个几何感知消息传递层来研究该图，这些层通过缩放点积注意力方法增强，选择性地增强与解剖相关的跨标志消息。最终的节点嵌入被组合成$ extmath{O} = 3$个疼痛级别聚类，并整合为归一化疼痛评分（Normalized Pain Score, NPS），其范围为$[0, 100 ext{%}]$，这是基于SPFES的信心加权评分方法。
cs.CV / 100 / 2608.11051
HUI360: A 360{\deg} Egocentric Dataset and Baselines for Human-Robot Interaction Anticipation
HUI360：用于人机交互预测的360°自我中心数据集及基准
Abstract
As robots increasingly operate in human-populated environments, anticipating human intentions is essential for enabling proactive and socially aware behavior. Automatic anticipation of human-robot interactions is thus emerging as a crucial perception challenge for embodied agents. To this end, we introduce HUI360, the largest dataset for human-robot interaction anticipation in the wild and its set of baselines. The dataset was collected from a mobile robot, in the wild, over multiple days within a 3-month period, and in several environments, capturing natural, spontaneous behaviors from both passersby and users, and encompassing a diverse range of individuals. This variety enables evaluating and improving the generalization capabilities of interaction anticipation models. We designed a pipeline and share code for automatic interaction annotation in arbitrary 360-degree equirectangular videos, along with interfaces for manual refinement. Using this pipeline, we release the HUI360 open set of 1M pre-processed annotations, including detailed 2D poses, facial keypoints, and segmentation masks, obtained using state-of-the-art computer vision methods and manually curated to ensure high-quality tracking and interaction annotation. Additionally, we release the raw panoptic 360-degree images captured from the robot's egocentric viewpoint (on demand, for research purpose only in compliance with GDPR). Finally, we establish benchmark baselines for interaction anticipation, including the first cross-dataset evaluations for this task: to this end, we also release 6M annotations for another existing in-the-wild outdoor dataset collected from a mobile robot (SSUP-HRI). Dataset and code can be found at https://hucebot.github.io/hui360.
Chinese Translation
随着机器人在人员密集环境中越来越多地运行，预测人类意图对于实现主动和具有社会意识的行为至关重要。因此，自动预测人机交互正成为具身智能体面临的一项重要感知挑战。为此，我们引入了HUI360，这是野外人机交互预测的最大数据集及其基准集。该数据集是在为期三个月的多天内，从移动机器人在多种环境中收集的，捕捉了路人和用户的自然、自发行为，涵盖了多样化的个体。这种多样性使得评估和改进交互预测模型的泛化能力成为可能。我们设计了一个管道，并分享了用于在任意360度等距矩形视频中自动进行交互注释的代码，以及手动精细化的接口。通过该管道，我们发布了HUI360开放集的100万条预处理注释，包括详细的2D姿态、面部关键点和分割掩码，这些数据是使用最先进的计算机视觉方法获得的，并经过人工整理以确保高质量的跟踪和交互注释。此外，我们还发布了从机器人自我中心视角捕获的原始全景360度图像（按需提供，仅用于研究目的，符合GDPR规定）。最后，我们建立了交互预测的基准基线，包括该任务的首个跨数据集评估：为此，我们还发布了600万条来自另一个现有的野外户外数据集（SSUP-HRI）的注释。数据集和代码可以在https://hucebot.github.io/hui360找到。
cs.CV / 101 / 2608.11053
A Comparative Evaluation of Deep Learning Object Detection Models on a Real-World Multi-Plant Dataset from Africa
对来自非洲的真实多植物数据集上深度学习目标检测模型的比较评估
Abstract
The application of computer vision in agriculture has shown significant potential for improving crop monitoring and precision farming. However, many existing approaches rely on controlled datasets that do not adequately represent realworld farming conditions, particularly in underrepresented regions such as Africa. This study presents a comparative evaluation of six object detection models YOLOv5, YOLOv8, YOLO11, YOLO26, Faster R-CNN, and RT-DETR using a real-world dataset, AgriAISeg 1 , collected manually from Nigerian farms. AgriAISeg comprises 3,382 images of sesame, cabbage, and tomato crops captured under varying environmental conditions, including changes in illumination, occlusion, and viewing perspectives. Models were trained, and performance was assessed using precision, recall,
[email protected], and
[email protected]:0.95. The results show that RT-DETR achieved the highest overall performance with a precision of 0.768 and
[email protected]:0.95 of 0.624, while YOLOv8 and YOLO11 also demonstrated strong and consistent performance. In contrast, Faster R-CNN recorded significantly lower accuracy, with an overall
[email protected] of 0.466, indicating reduced effectiveness under complex field conditions. In addition, YOLO-based models exhibited superior training efficiency compared to Faster R-CNN.These findings demonstrate that modern one-stage and transformer-based detectors provide more reliable and efficient solutions for plant detection in realworld agricultural environments.
Chinese Translation
计算机视觉在农业中的应用显示出显著的潜力，可以改善作物监测和精准农业。然而，许多现有方法依赖于受控数据集，这些数据集并未充分代表真实的农业条件，特别是在像非洲这样的欠代表地区。本研究对六种目标检测模型进行了比较评估，包括 YOLOv5、YOLOv8、YOLO11、YOLO26、Faster R-CNN 和 RT-DETR，使用的是从尼日利亚农场手动收集的真实数据集 AgriAISeg 1。AgriAISeg 包含 3,382 张芝麻、卷心菜和番茄作物的图像，这些图像是在不同环境条件下拍摄的，包括光照变化、遮挡和视角变化。对模型进行了训练，并使用精确度、召回率、
[email protected] 和
[email protected]:0.95 评估了性能。结果表明，RT-DETR 在整体性能上表现最佳，精确度为 0.768，
[email protected]:0.95 为 0.624，而 YOLOv8 和 YOLO11 也表现出强劲且一致的性能。相比之下，Faster R-CNN 的准确率显著较低，整体
[email protected] 为 0.466，表明在复杂的田野条件下效果降低。此外，基于 YOLO 的模型相比于 Faster R-CNN 显示出更优越的训练效率。这些发现表明，现代的一阶段和基于变换器的检测器为真实农业环境中的植物检测提供了更可靠和高效的解决方案。
cs.CV / 102 / 2608.11064
Entropy-Centric Explainable AI for Remote Sensing Image Segmentation
基于熵的可解释人工智能在遥感图像分割中的应用
Abstract
Artificial intelligence (AI) has become a powerful approach to solving complex problems in critical domains. Many concerns arise regarding the decision-making process of its models, mainly due to deep neural networks outperforming their peers at the cost of ambiguity in feature extraction and prediction. Consequently, in critical domains such as remote sensing, where high-resolution imagery must be analyzed using black-box models, the lack of transparency limits trust in these models and, thus, their adoption. In light of this reality, explaining and understanding the complex decision-making process of AI models has become essential. Explainable AI (XAI) aims to bridge this gap by providing insights into how and why certain decisions are made. While significant progress has been achieved in explaining image classification tasks, image segmentation still offers considerable room for improvement. In this context, this paper proposes an entropy-centric XAI method for semantic segmentation. Moreover, a new XAI evaluation methodology is proposed to efficiently measure the relevance of the regions highlighted by the proposed XAI method. Experimental results demonstrate the superiority of the proposed XAI method compared with recently adapted XAI methods for semantic segmentation.
Chinese Translation
人工智能（AI）已成为解决关键领域复杂问题的强大方法。然而，关于其模型决策过程的许多担忧随之而来，主要是由于深度神经网络在特征提取和预测方面的模糊性超越了其他方法。因此，在遥感等关键领域中，必须使用黑箱模型分析高分辨率图像，而缺乏透明度限制了对这些模型的信任，从而影响了它们的应用。鉴于这一现实，解释和理解人工智能模型复杂的决策过程变得至关重要。可解释人工智能（XAI）的目标是通过提供对某些决策如何以及为何做出的见解来弥补这一差距。尽管在解释图像分类任务方面取得了显著进展，但图像分割仍然有很大的改进空间。在此背景下，本文提出了一种基于熵的XAI方法用于语义分割。此外，本文还提出了一种新的XAI评估方法，以有效测量所提XAI方法突出区域的相关性。实验结果表明，与最近适应的语义分割XAI方法相比，所提XAI方法具有更优越的性能。
cs.CV / 103 / 2608.11074
CapProbe: Evaluating Detailed Image Captions via Full-Scene Dense Question Answering
CapProbe：通过全场景密集问答评估详细图像标题
Abstract
Evaluating detailed image captions from Vision-Language Models (VLMs) requires going beyond surface-level semantic similarity. Reference-based metrics (e.g., CIDEr and SPICE) and LLM-as-scorer protocols struggle to verify dense factual claims, while existing QA-based alternatives generally offer lower probe density, narrower domain coverage, or no explicit alignment between individual questions and segmented image regions. We introduce CapProbe, a full-scene dense QA benchmark that turns detailed caption evaluation into region-aligned factual checking. Each image is decomposed into coarse semantic regions covering both foreground and background elements; for every retained region, we generate multiple-choice questions spanning 10 semantic categories, forming a dense checklist of probed visual facts. Guided by a two-tier taxonomy of 37 L1 domains and 219 L2 sub-domains, CapProbe comprises 346 images, 1,868 regions, and 25,650 questions, averaging 74 QA pairs per image. A language judge answers from the caption alone; an Uncertain option and Effective Accuracy provide a judge-dependent proxy for distinguishing unanswered probes from incorrectly resolved ones, while density-based metrics penalize verbose yet uninformative captions. The protocol is cost-effective: by converting unconstrained scalar scoring into structured MCQ reading, it reduces open-ended scoring bias while remaining judge-conditioned and yields relatively stable model rankings under a fixed reader. Experiments on 13 VLMs show large Coverage gaps across models, a clear competency-efficiency trade-off, and failure modes that sparse or overlap-based evaluation often misses. The benchmark data, annotations, and evaluation code will be released soon.
Chinese Translation
评估视觉语言模型（VLMs）生成的详细图像标题需要超越表层语义相似性。基于参考的指标（例如，CIDEr 和 SPICE）以及大语言模型评分协议在验证密集事实声明时面临困难，而现有的基于问答的替代方案通常提供较低的探测密度、更窄的领域覆盖，或缺乏个别问题与分段图像区域之间的明确对齐。我们提出了 CapProbe，一个全场景密集问答基准，将详细标题评估转化为区域对齐的事实检查。每幅图像被分解为覆盖前景和背景元素的粗略语义区域；对于每个保留的区域，我们生成涵盖10个语义类别的多项选择问题，形成一个密集的视觉事实探测清单。在37个L1领域和219个L2子领域的两级分类法指导下，CapProbe包含346幅图像、1868个区域和25650个问题，每幅图像平均有74对问答。语言评审仅根据标题作答；不确定选项和有效准确度为区分未回答的探测和错误解决的探测提供了依赖评审的代理，而基于密度的指标则惩罚冗长但无信息的标题。该协议具有成本效益：通过将无约束的标量评分转化为结构化的多项选择题阅读，它减少了开放式评分偏见，同时保持了评审条件，并在固定读者下产生相对稳定的模型排名。对13个VLM的实验显示模型之间存在显著的覆盖差距，明显的能力-效率权衡，以及稀疏或基于重叠的评估常常遗漏的失败模式。基准数据、注释和评估代码将很快发布。
cs.CV / 104 / 2608.11075
Static in Frames, Dynamic in Events: Rethinking Features in Event Cameras as Motion Cues
静态于帧，动态于事件：重新思考事件相机中的特征作为运动线索
Abstract
Event cameras capture intensity changes asynchronously with high temporal resolution, requiring novel preprocessing methods for downstream tasks. Unlike static intensity snapshots, event data inherently encode information about scene dynamics and object motion, meaning that features derived from events can exhibit behaviors with no direct analogue in frame-based vision. In this paper, we analyze two features used in event-based corner detection---the eigenvalues of the structure tensor and the spatiotemporal density values---and show that they are \emph{motion cues}. We hypothesize that these features, combined with local geometric information, can enhance motion estimation tasks. To validate this, we first theoretically analyze how the eigenvalues of the structure tensor at moving corner points relate to the direction of motion. We then design controlled experiments on a synthetic dataset, confirming that extending local geometric features with eigenvalues and density values provides complementary motion information and is robust to texture and shot noise. Finally, we integrate the proposed features into a state-of-the-art event-based optical flow network and evaluate on the real-world DSEC benchmark, where the added features consistently improve accuracy, with the largest gains in data-scarce scenarios and for lower-capacity models. The code for this paper can be found at: \href{https://github.com/hesamaraghi/static-in-frames-dynamic-in-events}{https://github.com/hesamaraghi/static-in-frames-dynamic-in-events}.
Chinese Translation
事件相机以高时间分辨率异步捕捉强度变化，这需要为下游任务开发新颖的预处理方法。与静态强度快照不同，事件数据本质上编码了场景动态和物体运动的信息，这意味着从事件中提取的特征可能表现出在基于帧的视觉中没有直接对应的行为。在本文中，我们分析了用于基于事件的角点检测的两个特征——结构张量的特征值和时空密度值——并表明它们是 extit{运动线索}。我们假设这些特征结合局部几何信息，可以增强运动估计任务。为了验证这一点，我们首先理论分析了运动角点处结构张量的特征值与运动方向之间的关系。然后，我们在一个合成数据集上设计了受控实验，确认扩展局部几何特征与特征值和密度值结合提供了互补的运动信息，并且对纹理和拍摄噪声具有鲁棒性。最后，我们将提出的特征集成到一个最先进的基于事件的光流网络中，并在真实世界的DSEC基准上进行评估，结果显示添加的特征始终提高了准确性，尤其在数据稀缺的场景和低容量模型中获得了最大的提升。本文的代码可在以下链接找到： exttt{https://github.com/hesamaraghi/static-in-frames-dynamic-in-events}.
cs.CV / 105 / 2608.11076
Foundation Model-Enabled Efficient Data Sampling (FEEDS): A label-efficient training strategy for pan-cancer, multi-tracer PET/CT datasets
基础模型驱动的高效数据采样（FEEDS）：一种针对全癌症、多示踪剂PET/CT数据集的标签高效训练策略
Abstract
Automated lesion segmentation in whole-body PET/CT imaging can assist clinicians with cancer detection, staging, and treatment planning across radiotracers and cancer types. However, training lesion segmentation models that capture variations in lesion size, distribution, and appearance requires large annotated datasets, whose creation is both time- and expertise-intensive. As a result, models trained on limited labeled PET/CT data often lack the accuracy and generalizability needed for clinical use. We present FEEDS (Foundation model-Enabled Efficient Data Sampling), a label- and compute-efficient learning strategy that uses vision foundation model embeddings to select the most informative and diverse unlabeled cases for expert annotation. Unlike unsupervised, semi-supervised, and active learning approaches, FEEDS is a one-step training paradigm requiring only a limited, representative training set, making it label- and compute-efficient. We train and validate FEEDS using the AutoPET-III dataset. We test its accuracy and generalizability on three held-out sets: AutoPET-III, DeepPSMA, and an internal Dartmouth-Hitchcock Medical Center dataset. We evaluate clinical utility at the voxel, lesion, and anatomic region level to assess performance in high-risk areas and treatment planning utility. FEEDS outperforms random-sampling-based labeling, pseudolabel-based semi-supervised learning, and training with limited labeled data alone. It generalizes across all three test sets, FDG and PSMA tracers, and multiple diseases, matching fully-labeled (100\%) training performance with 70\% less annotation burden. FEEDS addresses the challenge of label scarcity in an automatic lesion segmentation framework by providing a practical approach for constructing representative and diverse annotation queues from large, unannotated clinical repositories.
Chinese Translation
全身PET/CT影像中的自动病灶分割可以帮助临床医生在不同放射示踪剂和癌症类型中进行癌症检测、分期和治疗规划。然而，训练能够捕捉病灶大小、分布和外观变化的病灶分割模型需要大量的标注数据集，而这些数据集的创建既耗时又需要专业知识。因此，基于有限标记PET/CT数据训练的模型通常缺乏临床使用所需的准确性和泛化能力。我们提出了FEEDS（基础模型驱动的高效数据采样），这是一种标签和计算高效的学习策略，利用视觉基础模型嵌入选择最具信息性和多样性的未标记案例进行专家标注。与无监督、半监督和主动学习方法不同，FEEDS是一种一步训练范式，仅需有限的、具有代表性的训练集，从而实现标签和计算的高效性。我们使用AutoPET-III数据集对FEEDS进行训练和验证，并在三个保留集上测试其准确性和泛化能力：AutoPET-III、DeepPSMA和内部达特茅斯-希区柯克医学中心数据集。我们在体素、病灶和解剖区域级别评估临床实用性，以评估在高风险区域和治疗规划中的表现。FEEDS的表现优于基于随机采样的标注、伪标注的半监督学习以及仅用有限标记数据的训练。它在所有三个测试集、FDG和PSMA示踪剂以及多种疾病中均表现出良好的泛化能力，以70%的标注负担匹配全标注（100%）训练的表现。FEEDS通过提供一种实用的方法，从大型未标注临床库中构建具有代表性和多样性的标注队列，解决了自动病灶分割框架中标签稀缺的挑战。
cs.CV / 106 / 2608.11077
Learning Gaussian Structure: Intervention-Guided Density Control for Feed-Forward Driving Reconstruction
学习高斯结构：基于干预引导的密度控制用于前馈驱动重建
Abstract
Feed-forward Gaussian reconstruction has recently emerged as an efficient approach for driving scene reconstruction. However, prevailing LiDAR-based methods preserve the initial correspondence between observed points and Gaussian primitives, treating the initialized primitive set as the final representation. Unlike optimization-based 3DGS, these methods cannot accumulate gradients during training to determine how the scenes representation should be densified. Meanwhile, the shared sparse backbone only fuses observations from different timestamps implicitly, without explicitly aggregating cross-time evidence for individual primitives. In this paper, we present Learning Gaussian Structure (LGS), a framework that enhances both Gaussian structure and primitive attributes. Our key observation is that changes in local gradient responses induced by a prune or add intervention reveal whether the corresponding structural adjustment benefits reconstruction. Based on this observation, our Gaussian Densify Policy learns a Densify Map comprising Prune and Addition Scores from controlled interventions, and directly adjusts the Gaussian structure during inference. We further develop a compact Cross-Time Point Query that explicitly retrieves and aggregates neighboring features from Gaussian primitives at other timestamps for reliable attribute prediction. Extensive experiments on the Waymo Open Dataset and PandaSet demonstrate that LGS consistently outperforms existing methods.
Chinese Translation
前馈高斯重建最近成为一种高效的驾驶场景重建方法。然而，现有的基于激光雷达（LiDAR）的方法保持观察点与高斯原件之间的初始对应关系，将初始化的原件集视为最终表示。与基于优化的3DGS不同，这些方法在训练过程中无法累积梯度，以确定场景表示应如何进行密集化。同时，共享的稀疏主干仅隐式融合来自不同时间戳的观察，而没有明确聚合个别原件的跨时间证据。在本文中，我们提出了学习高斯结构（Learning Gaussian Structure, LGS）框架，该框架增强了高斯结构和原件属性。我们的关键观察是，由于修剪或添加干预引起的局部梯度响应的变化揭示了相应的结构调整是否有利于重建。基于这一观察，我们的高斯密集化策略学习一个包含修剪和添加评分的密集化图，从受控干预中直接调整高斯结构。在推理过程中，我们进一步开发了一个紧凑的跨时间点查询，明确检索和聚合来自其他时间戳的高斯原件的邻近特征，以实现可靠的属性预测。在Waymo开放数据集和PandaSet上的大量实验表明，LGS始终优于现有方法。
cs.CV / 107 / 2608.11096
Every Packet Counts: Dispersing Information for Loss-Resilient Learned Image Compression
每个数据包都至关重要：为抗丢包的学习图像压缩分散信息
Abstract
Learned image compression (LIC) has achieved impressive rate-distortion performance. However, existing methods remain highly vulnerable to packet loss, a common challenge in satellite and emergency communications. This vulnerability stems from non-uniform information distribution at the packetization stage and sequential decoding dependencies at the entropy coding stage. We propose an end-to-end loss-resilient image compression scheme that addresses both. Before packetization, we introduce an Inter-Channel Redistribution (ICR) mechanism to redistribute channel energy, preventing critical information concentrating in a small subset of channels. Then, an Interleaved Channel Grouping (ICG) strategy partitions latent channels in a strided manner to disperse information across packets, with each packet kept within constrained sizes. To limit cascading errors from lost packets, we adopt a two-layer dual-branch autoregressive structure to shorten the dependency chain. Extensive experiments demonstrate that our method consistently outperforms existing approaches in both reconstruction quality and stability. At 20% packet loss, it achieves an average PSNR gain of 1.84 dB over LossResilientLIC while reducing PSNR variance by an order of magnitude. Notably, trained under uniform random loss only, our model generalizes to bursty loss modeled by the Gilbert-Elliott channel, outperforming methods explicitly trained for such conditions.
Chinese Translation
学习图像压缩（LIC）在速率失真性能上取得了令人瞩目的成果。然而，现有方法在数据包丢失方面仍然高度脆弱，这是卫星和紧急通信中的一个普遍挑战。这种脆弱性源于数据包化阶段信息分布的不均匀性以及熵编码阶段的顺序解码依赖性。我们提出了一种端到端的抗丢包图像压缩方案，解决了这两个问题。在数据包化之前，我们引入了一种通道间重分配（Inter-Channel Redistribution, ICR）机制，以重新分配通道能量，防止关键信息集中在少数通道中。然后，采用交错通道分组（Interleaved Channel Grouping, ICG）策略，以跨步的方式对潜在通道进行分区，从而在数据包中分散信息，并确保每个数据包的大小受到限制。为了限制因丢失数据包而导致的级联错误，我们采用了两层双分支自回归结构，以缩短依赖链。大量实验表明，我们的方法在重建质量和稳定性方面始终优于现有方法。在20%的数据包丢失情况下，它相较于LossResilientLIC实现了1.84 dB的平均PSNR增益，同时将PSNR方差降低了一个数量级。值得注意的是，我们的模型在仅针对均匀随机丢失进行训练的情况下，能够推广到由Gilbert-Elliott信道建模的突发丢失，超越了专门为这种情况训练的方法。
cs.CV / 108 / 2608.11123
AlbumentationsX: One Augmentation Pipeline for Images and Related Annotations
AlbumentationsX：一个用于图像及相关注释的增强管道
Abstract
Augmentation can corrupt a training example when an image and its annotations receive different random changes. A crop must use the same coordinates for the image, mask, boxes, keypoints, stereo views, video frames, or volume. Code paths that choose these values separately can silently misalign the data. AlbumentationsX keeps the transform list, probabilities, annotation settings, and random seed in one Compose object. Each call chooses random values once and applies them to every supported part of the training example. The library keeps each object's mask, box, and label together and lets projects add their own transforms. It can also save the pipeline definition, show what happened in one call, and run that call again. The examples place Compose after files have been decoded into arrays and before PyTorch groups examples into a batch. AlbumentationsX executes the declared transforms. Practitioners still decide whether a flip, crop, color change, or other operation preserves the correct label for their task.
Chinese Translation
当图像及其注释接收到不同的随机变化时，增强可能会破坏训练示例。裁剪必须对图像、掩膜、边框、关键点、立体视图、视频帧或体积使用相同的坐标。单独选择这些值的代码路径可能会悄然导致数据不对齐。AlbumentationsX 将变换列表、概率、注释设置和随机种子保存在一个 Compose 对象中。每次调用只选择一次随机值，并将其应用于训练示例的每个支持部分。该库将每个对象的掩膜、边框和标签保持在一起，并允许项目添加自己的变换。它还可以保存管道定义，展示一次调用中发生的事情，并再次运行该调用。示例在文件解码为数组之后、PyTorch 将示例分组为一个批次之前放置 Compose。AlbumentationsX 执行声明的变换。实践者仍需决定翻转、裁剪、颜色变化或其他操作是否保留其任务的正确标签。
cs.CV / 109 / 2608.11135
Is There Really a Camouflaged Object? Towards Realistic Camouflaged Object Detection
真的存在伪装物体吗？朝着现实的伪装物体检测迈进
Abstract
Camouflaged object detection (COD) aims to segment objects that are visually concealed in their surroundings and has attracted increasing attention in recent years. However, most existing COD methods are developed under a closed-world assumption, where each input image is assumed to contain a camouflaged object. This assumption ignores realistic scenarios with pure backgrounds or non-camouflaged objects, causing existing models to produce severe false positives when deployed in open-world environments. To address this limitation, we propose OPC16K, a large-scale benchmark for realistic COD. OPC16K contains 16,245 images from 14 sources and is carefully organized into camouflaged-object images, pure background images, and non-camouflaged-object images, enabling comprehensive evaluation of both segmentation quality and negative-sample rejection. Based on this benchmark, we further propose OPCNet, a presence-aware camouflage network that reformulates COD from a pure segmentation task into a joint problem of object localization and camouflage existence reasoning. Specifically, OPCNet introduces hierarchical existence reasoning to distinguish CO, BG, and NOCOD scenarios, similarity-aware camouflage relation modeling to capture foreground-background camouflage cues, and existence-aware feature refinement to regulate segmentation features with existence predictions. Extensive experiments on OPC16K demonstrate that OPCNet achieves superior performance under the proposed realistic COD evaluation protocol, significantly reducing false positives on negative samples while maintaining accurate camouflaged-object segmentation. Code and dataset will be released at https://github.com/2231122/OPCOD.
Chinese Translation
伪装物体检测（COD）旨在分割在其周围环境中视觉上被隐藏的物体，近年来受到越来越多的关注。然而，大多数现有的COD方法是在封闭世界假设下开发的，即假设每个输入图像都包含一个伪装物体。这一假设忽略了纯背景或非伪装物体的现实场景，导致现有模型在开放世界环境中部署时产生严重的误报。为了解决这一局限性，我们提出了OPC16K，这是一个用于现实伪装物体检测的大规模基准数据集。OPC16K包含来自14个来源的16,245张图像，并被精心组织为伪装物体图像、纯背景图像和非伪装物体图像，从而能够全面评估分割质量和负样本拒绝能力。在此基准的基础上，我们进一步提出了OPCNet，这是一种关注存在的伪装网络，将COD从纯分割任务重新构建为物体定位和伪装存在推理的联合问题。具体而言，OPCNet引入了分层存在推理，以区分CO（伪装物体）、BG（背景）和NOCOD（非伪装物体）场景，采用相似性感知的伪装关系建模来捕捉前景-背景伪装线索，以及存在感知的特征精炼来调节带有存在预测的分割特征。在OPC16K上的大量实验表明，OPCNet在提出的现实COD评估协议下实现了优越的性能，显著减少了负样本的误报，同时保持了准确的伪装物体分割。代码和数据集将发布在 https://github.com/2231122/OPCOD。
cs.CV / 110 / 2608.11142
SAR2Agri: Learning SAR Intensity Representations for Agricultural Monitoring
SAR2Agri：用于农业监测的SAR强度表示学习
Abstract
Agricultural monitoring faces unique challenges, arising from the landscape's complex temporal, phenological, and climate dynamics, yet monitoring them is critical for ensuring food security. Synthetic Aperture Radar (SAR) satellites offer all-weather day-night imaging capability supporting key monitoring tasks including crop type mapping, yield prediction and phenological event detection. Existing multimodal remote sensing foundation models including TerraMind and CopernicusFM learn SAR representations by grounding them in optical imagery using joint encoding and contrastive learning techniques, while SAR-specific foundation models such as SAR-JEPA, SARMAE, and SAR-W-MixMAE primarily focus on target detection, flood mapping, and land cover classification applications. Recent work has introduced phenology inspired temporal pretext tasks with optical imagery which has shown strong performance on agricultural downstream tasks. In this work, we propose the first self-supervised learning pipeline focused on using only SAR intensity imagery for agricultural applications. We improve the temporal pretext tasks through masking and curriculum learning to enhance the pretraining pipeline's ability to capture phenological features from SAR. On the SICKLE benchmark, our final model achieves 84.9% IoU on crop type mapping, outperforming optical baselines (by 15.3 pt) and existing SAR baselines (by 2.2 pt), demonstrating the effectiveness of our proposed pipeline for pretraining SAR intensity encoders for agricultural monitoring.
Chinese Translation
农业监测面临独特的挑战，这些挑战源于景观复杂的时间、物候和气候动态，但监测这些因素对于确保粮食安全至关重要。合成孔径雷达（SAR）卫星提供全天候、昼夜成像能力，支持包括作物类型制图、产量预测和物候事件检测在内的关键监测任务。现有的多模态遥感基础模型，包括TerraMind和CopernicusFM，通过使用联合编码和对比学习技术，将SAR表示与光学影像相结合进行学习，而专门针对SAR的基础模型如SAR-JEPA、SARMAE和SAR-W-MixMAE则主要集中于目标检测、洪水制图和土地覆盖分类应用。最近的研究引入了受物候启发的时间预训练任务与光学影像结合，这在农业下游任务中表现出色。在本研究中，我们提出了首个专注于仅使用SAR强度影像进行农业应用的自监督学习流程。我们通过掩蔽和课程学习改进了时间预训练任务，以增强预训练流程从SAR中捕捉物候特征的能力。在SICKLE基准测试中，我们的最终模型在作物类型制图上达到了84.9%的交并比（IoU），超越了光学基线（提高15.3个百分点）和现有SAR基线（提高2.2个百分点），证明了我们提出的流程在农业监测中预训练SAR强度编码器的有效性。
cs.CV / 111 / 2608.11149
PRMU: A Corpus-Free Benchmark for Person-Centric Knowledge Unlearning in Multimodal Large Language Models
PRMU：一种无语料基准用于多模态大型语言模型中的以人为中心的知识遗忘
Abstract
Multimodal large language models (MLLMs) have demonstrated remarkable capabilities in storing and recalling rich person-related knowledge, raising increasing concerns about reliable knowledge removal. However, existing machine unlearning approaches for MLLMs typically assume access to original forget and retain corpora, which are often unavailable in realistic deletion scenarios. To address this limitation, we introduce PRMU, a benchmark for evaluating corpus-free multimodal unlearning under realistic person-centric deletion requests. PRMU focuses on naturally acquired person-related knowledge and evaluates whether models can remove target knowledge while preserving related knowledge through diverse textual and visual probes, including adversarial evaluation and fine-grained locality analysis. To facilitate research in this setting, we further introduce Similarity-Gated Projection Editing (SGPE), a lightweight corpus-free unlearning baseline with knowledge displacement, protected parameter-space editing, and locality-aware multimodal control. Extensive experiments on representative MLLMs reveal that existing unlearning methods often suffer from unfavorable forgetting-locality trade-offs, with significant locality degradation under aggressive forgetting settings, and remain vulnerable to multimodal knowledge reactivation. Meanwhile, SGPE provides a competitive trade-off between target forgetting, locality preservation, and general multimodal utility. We hope PRMU can facilitate future research toward realistic and scalable multimodal machine unlearning. Code and dataset will be released at https://github.com/2231122/PRMU.
Chinese Translation
多模态大型语言模型（MLLMs）在存储和回忆丰富的与人相关的知识方面表现出显著的能力，这引发了对可靠知识移除的日益关注。然而，现有的针对MLLMs的机器遗忘方法通常假设可以访问原始的遗忘和保留语料，而在现实的删除场景中，这些语料往往不可用。为了解决这一限制，我们引入了PRMU，这是一个用于评估在现实以人为中心的删除请求下的无语料多模态遗忘的基准。PRMU关注自然获取的与人相关的知识，并评估模型是否能够在保留相关知识的同时移除目标知识，采用多样的文本和视觉探测手段，包括对抗性评估和细粒度局部性分析。为了促进这一领域的研究，我们进一步引入了相似性门控投影编辑（Similarity-Gated Projection Editing，SGPE），这是一种轻量级的无语料遗忘基线，具有知识位移、受保护的参数空间编辑和局部性感知的多模态控制。在对代表性MLLMs进行的广泛实验中，我们发现现有的遗忘方法通常面临不利的遗忘-局部性权衡，在激进的遗忘设置下局部性显著下降，并且仍然容易受到多模态知识重新激活的影响。同时，SGPE在目标遗忘、局部性保留和一般多模态效用之间提供了竞争性的权衡。我们希望PRMU能够促进未来在现实和可扩展的多模态机器遗忘方面的研究。代码和数据集将发布在 https://github.com/2231122/PRMU。
cs.CV / 112 / 2608.11150
CausalSplat: Towards Comprehensive Hierarchical Reasoning in 3D Gaussian Splatting
CausalSplat：迈向全面的层次推理在3D高斯点云中的应用
Abstract
While 3D Gaussian Splatting (3DGS) has advanced open vocabulary scene understanding, existing methods remain confined to explicit queries. They struggle to interpret implicit intents, complex spatial constraints, and commonsense reasoning required for practical embodied interactions. To address this gap, we introduce the task of reasoning 3D Gaussian segmentation and construct two benchmarks, Causal-LERF and Causal-ScanNet. These benchmarks systematically evaluate commonsense, spatial, affordance, and counterfactual reasoning. Evaluations reveal that current state of the art methods perform poorly on these reasoning challenges. Therefore, we propose CausalSplat, a framework that integrates vision-language models with 3D scene graphs to disentangle explicit structural perception from implicit logical inference. Extensive experiments demonstrate that CausalSplat achieves state of the art performance on our reasoning benchmarks while showing strong generalizability on standard referring and open vocabulary 3D segmentation tasks. Project Page: https://jiayuding031020.github.io/CausalSplat
Chinese Translation
尽管3D高斯点云（3D Gaussian Splatting, 3DGS）在开放词汇场景理解方面取得了进展，但现有方法仍然局限于显式查询。它们在解释隐含意图、复杂空间约束以及实际体现交互所需的常识推理方面面临挑战。为了解决这一问题，我们引入了3D高斯分割推理的任务，并构建了两个基准测试，Causal-LERF和Causal-ScanNet。这些基准系统地评估了常识、空间、可供性和反事实推理。评估结果显示，当前最先进的方法在这些推理挑战中表现不佳。因此，我们提出了CausalSplat，一个将视觉-语言模型与3D场景图相结合的框架，以解构显式结构感知与隐式逻辑推理之间的关系。大量实验表明，CausalSplat在我们的推理基准上实现了最先进的性能，同时在标准的指代和开放词汇3D分割任务上展现出强大的泛化能力。项目页面：https://jiayuding031020.github.io/CausalSplat
cs.CV / 113 / 2608.11167
MultiModal Code-Switching: Interleaving Visual Objects into Language for Explicit Object-Level Alignment
多模态代码切换：将视觉对象交错融入语言以实现显式的对象级对齐
Abstract
Existing Multimodal Large Language Models (MLLMs) predominantly rely on image-text pairs for modality alignment pretraining, mapping global image representations to long textual descriptions. However, this image-level alignment suffers from referential ambiguity: models struggle to infer the correspondences between multiple visual objects and textual entities from the global representation, leading to data inefficiency and suboptimal semantic grounding. To address this, we propose MultiModal Code-Switching (MMCS), a novel pretraining paradigm that provides explicit object-level supervision. Inspired by the linguistic phenomenon of code-switching, MMCS interleaves vision and language by replacing textual entities with their corresponding visual objects, enforcing local vision-language grounding. We further develop a scalable data synthesis pipeline to generate a pretraining dataset of 773K samples with accurate object-entity correspondences. Experiments show that MMCS is highly data-efficient: with only 50K samples, it matches or surpasses models trained on 600K image-text pairs. Furthermore, MMCS consistently improves visual grounding and perception capabilities across varying model scales.
Chinese Translation
现有的多模态大型语言模型（MLLMs）主要依赖图像-文本对进行模态对齐预训练，将全局图像表示映射到长文本描述。然而，这种图像级对齐存在指称模糊性：模型在从全局表示中推断多个视觉对象与文本实体之间的对应关系时面临困难，导致数据效率低下和语义基础不佳。为了解决这个问题，我们提出了多模态代码切换（MMCS），一种提供显式对象级监督的新型预训练范式。受到语言现象代码切换的启发，MMCS通过用相应的视觉对象替换文本实体，将视觉与语言交错，从而强制实现局部的视觉-语言对齐。我们进一步开发了一个可扩展的数据合成管道，以生成一个包含773K样本的预训练数据集，确保准确的对象-实体对应关系。实验表明，MMCS具有很高的数据效率：仅使用50K样本，其性能与训练于600K图像-文本对的模型相当或更优。此外，MMCS在不同模型规模下持续提升视觉对齐和感知能力。
cs.CV / 114 / 2608.11191
Test-Time Self-Evolving GUI Visual Grounding via Reflection-Guided On-Policy Self-Distillation
基于反射引导的在线自蒸馏的测试时自演化GUI视觉定位
Abstract
GUI Visual Grounding is a fundamental capability for GUI agents. Existing models typically freeze their parameters after deployment, limiting their ability to adapt to unseen interfaces. Although recent methods attempt to adapt models via test-time reinforcement learning, they cannot reflect upon failed exploration. To overcome this, we propose a Test-Time Self-Evolving framework that enables models to improve after deployment without human-annotated ground truth. It constructs a closed-loop of Exploration, Evaluation, Reflection, and Internalization. Specifically, the agent first explores unseen interfaces by predicting grounding coordinates for given instructions. To evaluate these explorations, we introduce an MLLM-based Reflector to assess the generated results and provide the corresponding reasoning reflections. To internalize reflection knowledge into the model weights, we propose Reflection-Guided On-Policy Self-Distillation, which translates high-level reasoning into dense token-level supervision via a conditioned self-teacher. Furthermore, we design a Contrastive Calibration method to prevent incorrect auto-regressive prefixes from corrupting the supervisory signals during failed explorations. Extensive experiments across six benchmarks demonstrate our framework's effectiveness, achieving an average accuracy improvement of 7.4% over the base model. To the best of our knowledge, this is the first work to successfully exploit on-policy self-distillation for test-time adaptation in GUI visual grounding. By filling the gap in post-deployment adaptation, our framework completes the self-evolving capability of GUI agents. The code will be released.
Chinese Translation
GUI视觉定位是GUI智能体的一项基本能力。现有模型通常在部署后冻结其参数，限制了它们对未见界面的适应能力。尽管最近的方法尝试通过测试时强化学习来适应模型，但它们无法对失败的探索进行反思。为此，我们提出了一种测试时自演化框架，使模型在部署后能够在没有人工标注真实值的情况下进行改进。该框架构建了一个探索、评估、反思和内化的闭环。具体而言，智能体首先通过为给定指令预测定位坐标来探索未见界面。为了评估这些探索，我们引入了一种基于MLLM的反射器，以评估生成的结果并提供相应的推理反思。为了将反思知识内化到模型权重中，我们提出了反射引导的在线自蒸馏，该方法通过条件自教师将高层次推理转化为密集的标记级监督。此外，我们设计了一种对比校准方法，以防止在失败的探索过程中不正确的自回归前缀破坏监督信号。在六个基准上的广泛实验表明，我们的框架有效性，平均准确率比基础模型提高了7.4%。据我们所知，这是首次成功利用在线自蒸馏进行GUI视觉定位的测试时适应工作。通过填补后部署适应的空白，我们的框架完善了GUI智能体的自演化能力。代码将会发布。
cs.CV / 115 / 2608.11201
VidForensics-M1: Meta-Detection Reinforcement Learning with Verifiable Temporal Grounding for AI-Generated Video Forensics
VidForensics-M1：具有可验证时间基础的元检测强化学习用于AI生成视频取证
Abstract
Recent advances in video generation models have significantly improved the realism of synthetic videos, blurring the boundary between generated and authentic content and raising concerns about misinformation. Existing MLLM-based detectors mainly rely on supervised fine-tuning or label-level reinforcement learning, where coarse supervision limits generalization to unseen scenarios and emerging video generators. To overcome these limitations, we are the first to introduce \textbf{meta-detection} into AI-generated video detection, enabling reliable forgery detection by jointly optimizing predicted labels and supporting evidence within reinforcement learning. This paradigm requires reliable evidence signals and effective mechanisms to integrate them into label-level optimization. Textual rationales provide semantic descriptions of forgery artifacts, but their generation and verification depend on external models, making supervision vulnerable to hallucinations and semantic biases. In contrast, temporal grounding provides more objective and verifiable evidence, as manipulated intervals can be precisely controlled during forgery construction. Based on this insight, we propose an automated data construction pipeline that generates paired real-fake videos by replacing temporal segments with boundary-frame-conditioned video generation models. Furthermore, we introduce \textbf{Evidence-Guided Reward Redistribution}, which performs evidence-aware credit assignment by redistributing rewards among label-correct responses according to evidence quality. This preserves reliable label supervision while encouraging detectors to acquire fine-grained and verifiable forgery localization capabilities. Extensive experiments demonstrate that \textbf{VidForensics-M1} effectively leverages verifiable temporal evidence to achieve robust and generalizable AI-generated video detection.
Chinese Translation
最近视频生成模型的进展显著提高了合成视频的真实感，模糊了生成内容与真实内容之间的界限，并引发了对虚假信息的担忧。现有的基于MLLM的检测器主要依赖于监督微调或标签级强化学习，其中粗略的监督限制了对未见场景和新兴视频生成器的泛化能力。为了克服这些局限性，我们首次将 extbf{元检测}引入AI生成视频检测，通过在强化学习中联合优化预测标签和支持证据，实现可靠的伪造检测。该范式需要可靠的证据信号和有效的机制将其整合到标签级优化中。文本推理提供了伪造伪影的语义描述，但其生成和验证依赖于外部模型，使得监督容易受到幻觉和语义偏见的影响。相比之下，时间基础提供了更客观和可验证的证据，因为在伪造构建过程中可以精确控制操纵的时间间隔。基于这一洞察，我们提出了一种自动化数据构建管道，通过用边界帧条件的视频生成模型替换时间段生成成对的真实-虚假视频。此外，我们引入了 extbf{证据引导的奖励再分配}，通过根据证据质量在标签正确的响应之间重新分配奖励，执行基于证据的信用分配。这保留了可靠的标签监督，同时鼓励检测器获得细粒度和可验证的伪造定位能力。大量实验表明， extbf{VidForensics-M1}有效利用可验证的时间证据，实现了强大且具有泛化能力的AI生成视频检测。
cs.CV / 116 / 2608.11203
Capturing Uncertainty in Human Motion for Representation Learning in Soccer
在人类运动中捕捉不确定性以进行足球表征学习
Abstract
This paper presents a self-supervised representation learning framework for understanding 3D skeleton-based human motion in soccer, using future motion prediction as the learning objective. Since human motion is inherently uncertain, accounting for multiple plausible futures is essential for capturing the underlying motion dynamics and learning effective representations. To this end, we introduce a conditioning module for motion prediction that models a probabilistic distribution over discretized future motions in 3D Euclidean space, learning multimodality with explicit supervision from future trajectories. Experiments on large-scale soccer player tracking data show that our approach substantially improves motion prediction accuracy. Moreover, the learned representations effectively transfer to multiple soccer downstream applications, demonstrating strong cross-task generalization.
Chinese Translation
本文提出了一种自监督表征学习框架，用于理解足球中基于3D骨架的人类运动，以未来运动预测作为学习目标。由于人类运动本质上具有不确定性，因此考虑多个合理的未来状态对于捕捉潜在的运动动态和学习有效的表征至关重要。为此，我们引入了一个运动预测的条件模块，该模块在3D欧几里得空间中对离散化的未来运动建模概率分布，通过未来轨迹的显式监督学习多模态性。在大规模足球运动员跟踪数据上的实验表明，我们的方法显著提高了运动预测的准确性。此外，学习到的表征有效地迁移到多个足球下游应用，展示了强大的跨任务泛化能力。
cs.CV / 117 / 2608.11205
AdvFD: Boosting Visual Generation via Adversarial Fr'echet Distance Loss
AdvFD：通过对抗Fréchet距离损失提升视觉生成
Abstract
Fr\'echet distance has recently emerged as an effective distribution-level objective for generator post-training, complementing the conventional sample-level diffusion and flow-matching losses. However, directly optimizing Fr\'echet objectives can cause Fr\'echet hacking. The target metrics keep improving, but visual quality and Fr\'echet alignment in other feature spaces may stagnate or deteriorate. We attribute this failure to the static pretrained feature spaces used by existing Fr\'echet losses. These feature spaces provide incomplete and fixed views of the differences between real and generated distributions. To address this limitation, we propose Adversarial Fr\'echet Distance (AdvFD), which complements the static representation targets in FD-Loss with a calibrated adversarially learned representation. AdvFD augments the original static Fr\'echet objective with a learnable representation that adversarially maximizes the Fr\'echet discrepancy between real and generated samples, while the generator minimizes the same discrepancy in the resulting adaptive feature space. To prevent the adversarial representation from trivially increasing the objective through feature amplification, we further introduce real-feature whitening, which normalizes its scale and covariance geometry and stabilizes the min--max optimization. Extensive experiments show that AdvFD consistently improves one-step generator post-training across both JiT and pMF backbones and across different model scales.
Chinese Translation
Fréchet距离最近作为一种有效的分布级目标出现在生成器的后训练中，补充了传统的样本级扩散和流匹配损失。然而，直接优化Fréchet目标可能导致Fréchet攻击。目标指标不断提高，但视觉质量和其他特征空间中的Fréchet对齐可能停滞或恶化。我们将这一失败归因于现有Fréchet损失所使用的静态预训练特征空间。这些特征空间提供了对真实和生成分布之间差异的不完整和固定的视角。为了解决这一局限性，我们提出了对抗Fréchet距离（AdvFD），它通过经过校准的对抗学习表示来补充FD-Loss中的静态表示目标。AdvFD通过可学习的表示增强了原始静态Fréchet目标，该表示对抗性地最大化真实样本和生成样本之间的Fréchet差异，而生成器则在结果自适应特征空间中最小化相同的差异。为了防止对抗表示通过特征放大简单地增加目标，我们进一步引入了真实特征白化，它规范化了其尺度和协方差几何，并稳定了最小-最大优化。大量实验证明，AdvFD在JiT和pMF骨干网络以及不同模型规模上始终改善了一步生成器的后训练效果。
cs.AI / 1 / 2608.09949
Closed-Loop LLM Co-Pilots for Digital Agriculture
数字农业中的闭环大语言模型协同驾驶员
Abstract
This study evaluates the application of Large Language Models (LLMs) in complex biological systems, evolving from data analysis to autonomous, AI-guided experimentation. The framework is driven by data from a 49-channel phytosensor network, encompassing multispectral, electrochemical, and dielectric modalities. To enhance accessibility, the system provides real-time natural-language interpretation for both specialists and non-experts. However, its core advantage lies in the transition from human-in-the-loop analysis to autonomous control. Processing biophysical data, the LLM evaluates plant physiology and triggers hardware actuators to optimize microclimates, execute phenotyping protocols, or induce controlled stress scenarios. This closed-loop architecture establishes a direct AI-biology interface, enabling data-driven exploration of complex biosystems and ecologies. The framework was validated across three case studies, based on a vertical farm and a single-plant setup and deciphered complex micro- and macro-fluctuations in plant physiology. Agents in a production-scale deployment executed multi-parameter optimization, balancing biomass accumulation, chlorophyll content, and energy consumption. The LLM processed biosensing telemetry to modulate full-spectrum, 450 nm, and 660 nm lighting at 2-hour intervals. Compared to periodic control, the system in minimal-time mode reduced the production cycle by 35%. In the energy-optimization mode, it reduced energy consumption by 18% with only a marginal increase in cultivation time, exploiting physiological inertia via light pulses. Finally, the agents autonomously developed an unforeseen strategy of dark-induced chlorophyll accumulation, resulting in a 67.9% energy saving. This framework transforms LLMs into autonomous co-pilots for digital agriculture, improving the cost-to-value ratio and lowering computational and expert-labor constraints.
Chinese Translation
本研究评估了大语言模型（LLMs）在复杂生物系统中的应用，发展从数据分析到自主的人工智能引导实验。该框架基于一个49通道植物传感器网络的数据，涵盖多光谱、电化学和介电模式。为了提高可及性，该系统为专家和非专家提供实时自然语言解读。然而，其核心优势在于从人机协作分析转向自主控制。通过处理生物物理数据，LLM评估植物生理，并触发硬件执行器以优化微气候、执行表型协议或诱导受控应激场景。这种闭环架构建立了直接的人工智能-生物学接口，使得对复杂生物系统和生态的基于数据的探索成为可能。该框架在三个案例研究中得到了验证，基于一个垂直农场和单株植物设置，解读了植物生理中的复杂微观和宏观波动。在生产规模的部署中，代理执行了多参数优化，平衡了生物量积累、叶绿素含量和能量消耗。LLM处理生物传感遥测，以2小时的间隔调节全光谱、450 nm和660 nm的照明。与周期性控制相比，该系统在最小时间模式下将生产周期缩短了35%。在能量优化模式下，它将能量消耗降低了18%，仅略微增加了栽培时间，利用光脉冲利用生理惯性。最后，代理自主开发了一种意想不到的黑暗诱导叶绿素积累策略，实现了67.9%的节能。该框架将LLMs转变为数字农业的自主协同驾驶员，提高了成本效益比，并降低了计算和专家劳动的约束。
cs.AI / 2 / 2608.09967
SPOTting the Future: Lookahead Explanations for Deep Reinforcement Learning
预测未来：深度强化学习的前瞻性解释
Abstract
Deep reinforcement learning (DRL) agents achieve strong performance in complex environments, yet their decision-making processes remain difficult to interpret. We introduce SPOT (Sampling Policy Observation Tree), a novel model-agnostic, sampling-based framework for interpreting DRL policies. Given access to the policy and an environment simulator, SPOT constructs an interpretable finite-horizon tree by sampling actions and recursively simulating the resulting successor states. The tree provides an empirical representation of the policy's action preferences and their possible downstream evolution. We provide formal guarantees establishing SPOT's asymptotic recovery of the policy's unique most probable action and characterizing its disagreement behavior under high-entropy policies. We demonstrate SPOT in the SUMO-RL traffic-signal control domain. The case study illustrates how its tree-based representation can be used to inspect policy preferences, compare alternative future trajectories, and reveal downstream behaviors that are not visible through single-timestep feature-attribution methods.
Chinese Translation
深度强化学习（DRL）智能体在复杂环境中表现出色，但其决策过程仍然难以解释。我们提出了SPOT（采样策略观察树），这是一种新颖的模型无关的基于采样的框架，用于解释DRL策略。在访问策略和环境模拟器的基础上，SPOT通过采样动作并递归模拟产生的后继状态，构建一个可解释的有限时域树。该树提供了策略动作偏好的经验表示及其可能的后续演变。我们提供了正式保证，确立了SPOT在高熵策略下对策略唯一最可能动作的渐近恢复，并描述了其不一致行为。我们在SUMO-RL交通信号控制领域展示了SPOT。案例研究说明了其基于树的表示如何用于检查策略偏好、比较替代未来轨迹，并揭示通过单步特征归因方法无法观察到的后续行为。
cs.AI / 3 / 2608.09986
MIDAS: Mutual Information Disentanglement with Uncertainty-Aware Fusion for Incomplete Multimodal Sentiment Analysis
MIDAS：基于不确定性感知融合的互信息解耦用于不完整多模态情感分析
Abstract
Most existing multimodal sentiment analysis approaches assume access to complete multimodal inputs. However, real-world applications frequently encounter incomplete or corrupted modalities, posing a critical challenge. Although several methods have been proposed to tackle this issue, they mainly rely on data imputation and heuristic coordination constraints, which fail to effectively extract and leverage task-relevant information from the incomplete multimodal data. To address this challenge, we propose a unified framework termed Mutual Information Disentanglement with uncertainty-Aware fuSion (MIDAS), which effectively restructures multimodal representations under incomplete conditions. MIDAS adopts a variational modeling strategy to represent each modality with multivariate Gaussian latent variables and further decomposes them into shared and exclusive factors. To obtain reliable representations, we design a minimax objective that minimizes the mutual information between shared and exclusive spaces for stable disentanglement, while maximizing the mutual information among shared spaces across modalities to enhance semantic alignment. In addition, an uncertainty-aware fusion mechanism is introduced, where posterior variance is leveraged as a reliability indicator to adaptively weight latent features during fusion, ensuring robust integration even when modalities are incomplete. Extensive experiments on three widely used datasets show that MIDAS achieves strong and consistent performance gains over competitive baselines across a wide range of incomplete settings, demonstrating its effectiveness and robustness for incomplete data scenarios.
Chinese Translation
大多数现有的多模态情感分析方法假设能够获取完整的多模态输入。然而，现实世界的应用经常遇到不完整或损坏的模态，这带来了重大挑战。虽然已经提出了几种方法来解决这个问题，但它们主要依赖于数据插补和启发式协调约束，未能有效提取和利用来自不完整多模态数据的任务相关信息。为了解决这一挑战，我们提出了一个统一框架，称为不确定性感知融合的互信息解耦（MIDAS），该框架能够在不完整条件下有效重构多模态表示。MIDAS采用变分建模策略，用多元高斯潜变量表示每个模态，并进一步将其分解为共享因子和独占因子。为了获得可靠的表示，我们设计了一个极小极大目标，最小化共享空间和独占空间之间的互信息以实现稳定的解耦，同时最大化跨模态共享空间之间的互信息以增强语义对齐。此外，引入了一种不确定性感知融合机制，利用后验方差作为可靠性指标，在融合过程中自适应地加权潜在特征，确保即使在模态不完整的情况下也能实现稳健的集成。在三个广泛使用的数据集上的大量实验表明，MIDAS在各种不完整设置下相较于竞争基线实现了强劲且一致的性能提升，证明了其在不完整数据场景中的有效性和鲁棒性。
cs.AI / 4 / 2608.09998
Towards Sustainable Artificial Intelligence: A Comprehensive Review and Comparative Analysis of Deep Learning Models' Carbon Footprint
迈向可持续人工智能：深度学习模型碳足迹的综合评估与比较分析
Abstract
Artificial Intelligence (AI) and Machine Learning (ML) have become powerful tools for supporting and automating complex human tasks. Despite their benefits, growing attention has been directed toward their environmental implications, primarily due to their high energy demands and associated carbon emissions. This concern is particularly relevant in light of the increasing deployment of large-scale models, especially Deep Learning (DL) architectures, which provide advanced predictive capabilities but require substantial computational resources. This paper presents a systematic review of research on Green AI, Green DL, and optimization techniques aimed at reducing the environmental impact of AI models. In addition, we examine and compare several carbon measurement tools for estimating emissions generated by AI algorithms. To complement the review, we conducted an empirical evaluation using a CPU-based experimental setup, in which six DL models were implemented for a multi-label classification task. The objective was to quantify and compare their overall carbon emissions and to determine which stages of the DL lifecycle contribute most significantly to the total footprint. The results show that the training phase is the primary source of emissions. Moreover, the findings reveal that increased architectural complexity does not systematically translate into proportional accuracy gains, highlighting the importance of carefully balancing predictive performance and environmental cost. These results reinforce the need to integrate sustainability considerations into model selection and AI system design.
Chinese Translation
人工智能（AI）和机器学习（ML）已成为支持和自动化复杂人类任务的强大工具。尽管它们带来了诸多好处，但越来越多的关注也集中在它们对环境的影响上，主要是由于其高能耗和相关的碳排放。考虑到大规模模型，尤其是深度学习（DL）架构的日益部署，这一问题显得尤为重要。这些架构提供了先进的预测能力，但需要大量的计算资源。本文对绿色人工智能（Green AI）、绿色深度学习（Green DL）及旨在减少人工智能模型环境影响的优化技术进行了系统评审。此外，我们还审查并比较了几种用于估算人工智能算法产生的排放的碳测量工具。为了补充这一评审，我们使用基于CPU的实验设置进行了实证评估，其中实施了六个深度学习模型以完成多标签分类任务。我们的目标是量化并比较它们的总体碳排放，并确定深度学习生命周期中哪些阶段对总碳足迹的贡献最大。结果显示，训练阶段是排放的主要来源。此外，研究结果表明，架构复杂性的增加并不系统地转化为比例准确性的提升，这突显了在预测性能和环境成本之间进行谨慎平衡的重要性。这些结果进一步强调了在模型选择和人工智能系统设计中整合可持续性考虑的必要性。
cs.AI / 5 / 2608.10004
ReCBM: Uncertainty-Gated Relational Reasoning for Concept Bottleneck Models
ReCBM：用于概念瓶颈模型的基于不确定性的关系推理
Abstract
Concept Bottleneck Models (CBMs) provide an interpretable framework by grounding predictions in human-understandable concepts, enabling semantic inspection and test-time intervention. Recent variants have improved CBMs through richer concept representations, uncertainty estimation, and dependency modeling. However, robust reasoning under unreliable concept states remains underexplored. Without such reasoning, misleading semantic evidence can propagate through the bottleneck, compromising both explanations and downstream predictions. To address this issue, we propose ReCBM, an uncertainty-gated relational reasoning framework for CBMs. ReCBM introduces semantically defined concept relations into the bottleneck and uses uncertainty to guide their refinement. By modeling co-occurrence, implication, and exclusion, ReCBM specifies how evidence is exchanged across concepts, while uncertainty modulates the contribution of each concept during this process. Experiments across diverse datasets showed that ReCBM improved concept and task recovery under missing and flipped concepts, supported uncertainty-aware intervention, and extracted compact task-relevant concept subsets without degrading downstream performance.
Chinese Translation
概念瓶颈模型（CBMs）通过将预测与人类可理解的概念相结合，提供了一种可解释的框架，从而实现语义检查和测试时干预。最近的变体通过更丰富的概念表示、不确定性估计和依赖建模改进了CBMs。然而，在不可靠的概念状态下进行稳健推理仍然未得到充分探索。如果没有这样的推理，误导性的语义证据可能会通过瓶颈传播，从而损害解释和下游预测。为了解决这一问题，我们提出了ReCBM，一种用于CBMs的基于不确定性的关系推理框架。ReCBM将语义定义的概念关系引入瓶颈，并利用不确定性指导其精炼。通过建模共现、蕴含和排除，ReCBM明确了证据在概念之间的交换方式，同时不确定性在此过程中调节每个概念的贡献。跨多个数据集的实验表明，ReCBM在缺失和翻转概念的情况下改善了概念和任务恢复，支持了基于不确定性的干预，并提取了紧凑的与任务相关的概念子集，而不降低下游性能。
cs.AI / 6 / 2608.10030
Automating and Scaling Behavioral Scientific Research on AI Agents
自动化与扩展人工智能代理的行为科学研究
Abstract
As AI agents are increasingly deployed in complex environments, understanding their behaviors becomes critical. Yet behavioral scientific research on AI agents remains manual and labor-intensive. We introduce AEROBAT, the first multi-agent system to automate behavioral scientific research on AI agents. Given an arbitrary target behavior by its user, AEROBAT automatically executes a full pipeline of behavioral scientific research---generating hypotheses about the behavior, designing and executing controlled experiments, making behavioral assessments, analyzing the results, and writing reports. For 12 target behaviors, we used AEROBAT to generate and test 79 hypotheses: designing 1,240 controlled experiments and executing 23,512 simulation rounds in total. Moderate-to-strong statistical evidence was found for 26 hypotheses, including some novel ones. In sum, our results demonstrate that automated behavioral scientific research on AI agents can complement and extend the reach of manual research.
Chinese Translation
随着人工智能代理在复杂环境中的广泛应用，理解其行为变得至关重要。然而，针对人工智能代理的行为科学研究仍然是手动且劳动密集型的。我们介绍了AEROBAT，这是第一个自动化人工智能代理行为科学研究的多智能体系统。AEROBAT能够根据用户指定的任意目标行为，自动执行完整的行为科学研究流程——生成关于该行为的假设，设计和执行控制实验，进行行为评估，分析结果，并撰写报告。针对12种目标行为，我们使用AEROBAT生成并测试了79个假设：设计了1,240个控制实验，并总共执行了23,512轮模拟。我们发现26个假设具有中等到强的统计证据，包括一些新颖的假设。总之，我们的结果表明，自动化的人工智能代理行为科学研究可以补充并扩展手动研究的范围。
cs.AI / 7 / 2608.10090
CHORUS: Complementary Experts for High-Coverage Testbench Stimulus Generation
CHORUS：高覆盖率测试平台刺激生成的互补专家
Abstract
Large language models (LLMs) have advanced code generation, where executable feedback provides a more reliable learning signal than textual imitation alone. Hardware verification is an important application of code generation and accounts for a substantial fraction of modern chip design effort, with high-coverage testbench stimulus generation as a key task. We present CHORUS, a post-training framework that pushes performance beyond what a conventional supervised fine-tuning (SFT)-to-reinforcement learning (RL) pipeline achieves. CHORUS builds on two observations. First, staged SFT produces behaviorally diverse checkpoints, and dense-reward RL turns them into strong experts with comparable aggregate performance but distinct task-level strengths. Second, these complementary strengths can be exploited through either training-free model merging or further post-training to outperform the best individual expert. By consolidating the resulting specialists into a single 4B model, CHORUS achieves 88.0% Pass@1 on CVDP-ECov, outperforming DeepSeek-R1 (671B) by 13.5 percentage points.
Chinese Translation
大型语言模型（LLMs）在代码生成方面取得了进展，其中可执行反馈提供了比单纯的文本模仿更可靠的学习信号。硬件验证是代码生成的重要应用，占现代芯片设计工作的一大部分，高覆盖率测试平台刺激生成是其中的关键任务。我们提出了CHORUS，一个后训练框架，推动性能超越传统的监督微调（SFT）到强化学习（RL）管道所能达到的水平。CHORUS基于两个观察结果。首先，分阶段的SFT产生行为上多样化的检查点，而密集奖励的RL将其转化为具有可比聚合性能但在任务层面上具有不同优势的强专家。其次，这些互补优势可以通过无训练的模型合并或进一步后训练来利用，从而超越最佳单一专家。通过将这些专家整合为一个单一的4B模型，CHORUS在CVDP-ECov上实现了88.0%的Pass@1，超越DeepSeek-R1（671B）13.5个百分点。
cs.AI / 8 / 2608.10108
MESA:Task-Adaptive Multi-Structure Evidence Selection for Long-Horizon Agent Memory
MESA：任务自适应的多结构证据选择用于长时间跨度的智能体记忆
Abstract
Long-horizon agents accumulate trajectories spanning hundreds of interleaved reasoning, action, and observation steps, where answering a query may depend on evidence buried far back in the history. External memory stores such trajectories as structured representations, yet each structure provides a distinct and incomplete view. Existing multi-memory systems either read a fixed set of structures for every query, inflating context and introducing noise, or route each query to a single structure, preventing the composition of complementary evidence. A controlled analysis on AMA-Bench shows that the optimal memory configuration is typically neither a single structure nor the full union, but a tailored composition of multiple structural memories that varies with query and task demands. Motivated by these findings, we formulate structure-level dynamic selection: selecting and fusing a query-adaptive subset from a library of specialized memory structures. We propose MESA (a Multi-structure Evidence Selection framework for long-horizon Agent), which builds five complementary structure views of each trajectory and learns from end-to-end answer-level feedback to select and fuse a query-specific subset for a frozen answer model. To learn under this weak supervision, MESA employs harness optimization with prior-guided search and UCB-guided scheduling to balance exploration and exploitation. On AMA-Bench, MESA outperforms the strongest baseline by 8.5% while using 41% fewer evidence tokens than the all-structure alternative.
Chinese Translation
长时间跨度的智能体积累了跨越数百个交错推理、行动和观察步骤的轨迹，其中回答一个查询可能依赖于历史中埋藏的证据。外部记忆将这些轨迹存储为结构化表示，但每种结构提供的视角都是独特且不完整的。现有的多重记忆系统要么对每个查询读取固定的一组结构，从而膨胀上下文并引入噪声，要么将每个查询路由到单一结构，阻碍互补证据的组合。在对AMA-Bench的受控分析中显示，最佳的记忆配置通常既不是单一结构，也不是完整的并集，而是根据查询和任务需求量身定制的多种结构记忆的组合。基于这些发现，我们提出了结构级动态选择：从专门的记忆结构库中选择和融合一个查询自适应的子集。我们提出了MESA（用于长时间跨度智能体的多结构证据选择框架），它为每个轨迹构建五个互补的结构视图，并通过端到端的答案级反馈学习，以选择和融合一个针对特定查询的子集，用于固定的答案模型。为了在这种弱监督下进行学习，MESA采用了带有先验引导搜索和UCB引导调度的优化策略，以平衡探索和利用。在AMA-Bench上，MESA的表现比最强基线提高了8.5%，同时使用的证据标记比全结构替代方案少41%。
cs.AI / 9 / 2608.10153
The CASE Framework: A Multi-Disciplinary Control Architecture for Governing Enterprise Agentic AI
CASE框架：用于治理企业代理人工智能的多学科控制架构
Abstract
Enterprises are deploying autonomous AI agents faster than they can govern them, and prevailing approaches stretch a single discipline, typically DevSecOps built for deterministic automation, across every scale of agency. We argue that agentic AI governance is four problems, not one, each with a mature governing science. The CASE framework assigns Control theory to the individual agent (intent as setpoint, guardrails as feedback, evaluation as observation), complex Adaptive systems theory to agent collectives (where emergence makes single-agent assurance non-compositional), Supervisory cybernetics to human-agent teams (where the Law of Requisite Variety shows unaided human oversight fails structurally), and Engineering operations to fleets (extending error budgets to decision quality so autonomy becomes a controlled variable). We formalize each layer, derive cross-layer coupling conditions, including a zero-touch deployment paradox where excellence at one-layer strains the others, and trace twenty-plus enterprise controls to their classical constructs. Three empirical studies validate the thesis: 82 percent of documented production agent failures are multi-layer trajectories; none of 22 ecosystem tools offers full Layer 2 (emergence) coverage; and all 35 scored public deployments fall in the lowest maturity band. We name this mismatch, risk realized at the emergence layer against capability barely offered and practice absent, the Emergence Gap. A five-level maturity model with a non-compensatory bottleneck-weighted index and assessment instrument operationalizes CASE as a scientific rather than process maturity model, grounded in production enterprise agentic platforms. As EU AI Act Article 14 makes effective human oversight a legal requirement, only architectures satisfying requisite variety can make oversight real rather than ceremonial.
Chinese Translation
企业部署自主人工智能代理的速度超过了其治理能力，而现有的方法通常将单一学科（通常是为确定性自动化而构建的DevSecOps）扩展到每个代理层级。我们认为，代理人工智能治理是四个问题，而不是一个，每个问题都有成熟的治理科学。CASE框架将控制理论分配给单个代理（意图作为设定点，护栏作为反馈，评估作为观察），将复杂适应系统理论分配给代理集体（在此处，涌现使得单一代理的保证无法组合），将监督控制论分配给人机团队（在此处，必要多样性法则表明无辅助的人类监督在结构上失败），并将工程操作分配给车队（将错误预算扩展到决策质量，使自主性成为一个可控变量）。我们形式化了每一层，推导出跨层耦合条件，包括一个零接触部署悖论，即在某一层的卓越表现会对其他层造成压力，并追溯二十多个企业控制到其经典构造。三项实证研究验证了这一论点：82%的文档化生产代理失败是多层轨迹；22个生态系统工具中没有一个提供完整的第二层（涌现）覆盖；所有35个评分的公共部署均处于最低成熟度带。我们将这种不匹配称为涌现差距，即在涌现层面上实现的风险与几乎没有能力提供和实践缺失的能力之间的差距。一个五级成熟度模型，具有非补偿性瓶颈加权指数和评估工具，使CASE作为一个科学而非过程成熟度模型得以操作，基于生产企业代理平台。随着欧盟人工智能法第14条使有效的人类监督成为法律要求，只有满足必要多样性的架构才能使监督变得真实而非仪式化。
cs.AI / 10 / 2608.10157
SBCO: Self-Supervised, Verifier-Grounded Harness Optimization For Planning Agents
SBCO：自监督、验证者基础的规划代理优化
Abstract
Self-improving agents seek to reduce the human engineering effort behind AI systems by enabling them to evolve and self-improve their performance over time. Recently, methods like the Darwin G\"odel Machine and the Huxley G\"odel Machine have been proposed which enable open-ended, recursive self-improvement through self-reference where a coding agent edits its own code. Such self-referential self-improvement methods require that the competence required to perform the task coincides or aligns well with the competence required for self-modification which is the case for coding tasks. For domains or tasks, which do not satisfy the alignment needed, self-referential self-improvement is not available. In such cases, it is possible to adapt the above algorithms to other tasks by removing the self-referential aspect or introducing explicit self-modification of a meta-agent -- both computationally expensive, relying on population or self-modification search over many candidate agents. For planning tasks with explicit constraints, we propose a far cheaper alternative. We introduce SBCO (Self-supervised Block Coordinate Optimizer), a verifier-grounded harness optimizer in the same closed-loop, improve-from-experience family as the G\"odel-machine methods, but self-supervised rather than self-referential. Given an agentic harness, SBCO learns a decomposed bank of verifiers and a harness policy via approximate block coordinate ascent, improving the agent's outputs from its own graded feedback---with a fixed meta-agent and no human labels. Across two domains SBCO matches or exceeds a customized self-modifying baseline while using 4-5.5 times less compute budget.
Chinese Translation
自我改进的代理旨在通过使其能够随着时间的推移演化和自我提升性能，从而减少人工工程在人工智能系统中的投入。最近，提出了达尔文哥德尔机器（Darwin G"odel Machine）和赫胥黎哥德尔机器（Huxley G"odel Machine）等方法，这些方法通过自我引用实现开放式的递归自我改进，其中编码代理会编辑自己的代码。这类自我引用的自我改进方法要求执行任务所需的能力与自我修改所需的能力相一致或良好对齐，而这在编码任务中是成立的。对于不满足所需对齐的领域或任务，自我引用的自我改进就不可用。在这种情况下，可以通过去除自我引用特性或引入元代理的显式自我修改来将上述算法适应于其他任务——这两者在计算上都非常昂贵，依赖于对多个候选代理进行种群或自我修改搜索。对于具有显式约束的规划任务，我们提出了一种更便宜的替代方案。我们引入了SBCO（自监督块坐标优化器），这是一种基于验证者的优化器，属于与哥德尔机器方法相同的闭环、经验改进家族，但采用自监督而非自我引用。给定一个代理的优化框架，SBCO通过近似块坐标上升学习一个分解的验证者库和优化策略，从自身的分级反馈中改进代理的输出——使用固定的元代理且无需人工标签。在两个领域中，SBCO的性能与定制的自我修改基线相匹配或超过，同时计算预算减少了4-5.5倍。
cs.AI / 11 / 2608.10171
Generating Attacks for LLMs with GFlowNets
利用 GFlowNets 生成针对大型语言模型的攻击
Abstract
The rapid advancement of Large Language Models (LLMs) has facilitated their ubiquitous integration into various domains, leading to widespread adoption. However, this escalating trend has introduced significant security vulnerabilities, necessitating the identification and mitigation of flaws arising from malicious exploitation. Red teaming assessments, conducted to evaluate model robustness through diverse adversarial inputs, are essential for exposing security risks and implementing countermeasures. Currently, red teaming is performed either manually by experts or automatically using predefined attack datasets. Nevertheless, manual testing remains time-consuming, while existing automated methods suffer from limited creativity due to their inherent dependency on fixed datasets. In this study, we propose an automated, human-independent, and adaptive approach leveraging GFlowNets to identify LLM vulnerabilities by utilizing one large language model to test another. Within this framework, an attacker model is trained against a specified victim model to perform automated red teaming and provide a quantitative robustness score. This research aims to generate more effective adversarial attacks in English compared to existing benchmarks and, as a novel contribution to the literature, introduces a model capable of generating attack inputs in the Turkish language.
Chinese Translation
大型语言模型（LLMs）的快速发展促进了它们在各个领域的广泛应用，导致了普遍的采用。然而，这一不断上升的趋势引入了显著的安全漏洞，迫切需要识别和缓解因恶意利用而产生的缺陷。红队评估通过多样的对抗性输入来评估模型的鲁棒性，对于揭示安全风险和实施对策至关重要。目前，红队测试要么由专家手动进行，要么使用预定义的攻击数据集自动执行。然而，手动测试仍然耗时，而现有的自动化方法由于固有依赖于固定数据集，创造力有限。本研究提出了一种自动化的、独立于人类的、适应性的方法，利用 GFlowNets 通过利用一个大型语言模型来测试另一个模型，从而识别 LLM 的漏洞。在这一框架内，攻击者模型针对特定的受害者模型进行训练，以执行自动化红队测试并提供定量的鲁棒性评分。本研究旨在生成比现有基准更有效的英语对抗性攻击，并作为对文献的创新贡献，介绍一种能够生成土耳其语攻击输入的模型。
cs.AI / 12 / 2608.10176
TRACE: Trustworthy Retrieval-Augmented Conversational Engine
TRACE：可信的检索增强对话引擎
Abstract
Public service chatbots are expected to deliver recommendations from an underlying public service directory, while also making sure that the recommendations respect explicit user constraints. In practice, public service directories are noisy and inconsistent, and general-purpose large language model (LLM) or AI-based chatbots frequently generate unreliable recommendations, citing unverified sources from the web. We investigate the impact of retrieval quality on constraint-aware recommendation in public service conversational systems built over noisy and heterogeneous service directories. We propose TRACE (Trustworthy Retrieval-Augmented Conversational Engine), a retrieval-based, constraint-aware framework that parses input user queries into structural and semantic constraints for downstream retrieval, with the help of a dual data representation schema. Using a curated statewide pantry directory and a synthetic query benchmark, we evaluate multiple knowledge-representation variants with and without knowledge graphs (KGs). We experiment with several open-source LLMs and a proprietary model, showing that strengthening retrieval substantially improves user constraint satisfaction while reducing hallucinated recommendations. Performance differences across LLMs narrowed in our experiments as retrieval quality improved, making results less sensitive to model size. These findings suggest that the quality of retrieval is key for robust public service conversational systems.
Chinese Translation
公共服务聊天机器人预计能够从基础公共服务目录中提供推荐，同时确保这些推荐符合用户的明确约束。在实际操作中，公共服务目录往往存在噪声和不一致性，而通用的大型语言模型（LLM）或基于人工智能的聊天机器人经常生成不可靠的推荐，引用来自网络的未经验证的来源。我们研究了检索质量对公共服务对话系统中基于约束的推荐的影响，这些系统构建在嘈杂和异构的服务目录之上。我们提出了TRACE（可信的检索增强对话引擎），这是一个基于检索的、关注约束的框架，它利用双重数据表示模式将输入用户查询解析为结构性和语义约束，以便进行下游检索。通过使用经过整理的全州食品储藏室目录和合成查询基准，我们评估了多种知识表示变体，包括有知识图谱（KGs）和没有知识图谱的情况。我们实验了几种开源LLM和一个专有模型，结果表明，增强检索显著提高了用户约束的满足度，同时减少了虚假推荐。在我们的实验中，随着检索质量的提高，不同LLM之间的性能差异缩小，使得结果对模型规模的敏感性降低。这些发现表明，检索质量是构建稳健的公共服务对话系统的关键。
cs.AI / 13 / 2608.10198
Post-Hoc Sparse Coding of Latent Communication Between Vision-Language Model Agents
后验稀疏编码的潜在通信在视觉-语言模型代理之间的应用
Abstract
Latent-space communication allows heterogeneous vision-language model agents to exchange continuous representations without serializing visual and reasoning states into text. Vision Wormhole realizes this approach by translating visual features into a universal latent representation that can be consumed by another model, but every message is transported as a dense tensor of the same size regardless of its content. A fixed-capacity dense tensor therefore need not have a fixed effective information density: some messages may use only a small fraction of the available representational degrees of freedom. This observation suggests that the communication channel may be substantially compressible. We study its redundancy by fitting a post-hoc sparse autoencoder to frozen Vision Wormhole activations and measuring reconstruction, downstream utility, feature reuse, and token-level interventions across nine reasoning benchmarks. Relative to the original float32 transport, a uint16-index/float16-value sparse payload with k=4 active coefficients per token reduces the transmitted bytes by 128x. In a single-run evaluation, the seven-task non-AIME mean accuracy changes from 49.85% to 49.77%. The fitted 4096-element dictionary uses only 50 features, and task-level active sets have a mean pairwise Jaccard similarity of 0.906. These measurements establish strong post-hoc compressibility relative to the original transport, but do not yet isolate the incremental contribution of sparse coding from position selection, reduced precision, low-rank structure, or SAE optimization effects. The results motivate matched-payload comparisons and communication mechanisms whose payload adapts to the information used by each message.
Chinese Translation
潜在空间通信允许异构的视觉-语言模型代理在不将视觉和推理状态序列化为文本的情况下交换连续表示。视觉虫洞（Vision Wormhole）通过将视觉特征转换为另一模型可以使用的通用潜在表示来实现这一方法，但每条消息都作为相同大小的稠密张量传输，而不考虑其内容。因此，固定容量的稠密张量不必具有固定的有效信息密度：某些消息可能仅使用可用表示自由度的一小部分。这一观察表明，通信通道可能具有显著的可压缩性。我们通过将后验稀疏自编码器拟合到冻结的视觉虫洞激活，并在九个推理基准上测量重构、下游效用、特征重用和令牌级干预，来研究其冗余性。与原始的 float32 传输相比，使用 k=4 个活跃系数的 uint16 索引/float16 值稀疏负载将传输的字节减少了 128 倍。在单次评估中，七任务非 AIME 的平均准确率从 49.85% 变为 49.77%。拟合的 4096 元素字典仅使用了 50 个特征，任务级活跃集的平均成对 Jaccard 相似度为 0.906。这些测量结果确立了相对于原始传输的强后验可压缩性，但尚未将稀疏编码的增量贡献与位置选择、精度降低、低秩结构或 SAE 优化效应分离。结果激励了匹配负载比较和通信机制，其负载适应每条消息所使用的信息。
cs.AI / 14 / 2608.10206
Edge Phoneme Recognition for Children's Speech through Age-Aware Training
基于年龄感知训练的儿童语音边缘音素识别
Abstract
Detecting phonemes from children's speech has historically been difficult due to the scarcity of training data, and unique characteristics of children's speech. During a phoneme detection competition, we found that training a lightweight model to predict the age of the learner, as well as the phoneme sequence, enabled a 94M-parameter model to outperform WavLM Large models (317M) on the target DrivenData distribution, and fall within approximately 0.04 CER of competition ensembles with 90 times the parameters. This has enabled the creation of PhonemeTrainer, an application that can run on most modern cellular phones. This will ultimately enable better Automated Speech Recognition (ASR) and pronunciation helper apps for children's speech, with the privacy and compliance benefits that come with edge processing.
Chinese Translation
由于训练数据稀缺以及儿童语音的独特特征，从儿童语音中检测音素历来是一个困难的问题。在一次音素检测竞赛中，我们发现训练一个轻量级模型来预测学习者的年龄以及音素序列，使得一个94M参数的模型在目标DrivenData分布上超越了WavLM Large模型（317M），并且与拥有90倍参数的竞赛集成模型的字符错误率（CER）相差约0.04。这使得我们能够创建PhonemeTrainer，一个可以在大多数现代手机上运行的应用程序。这最终将促进更好的自动语音识别（ASR）和儿童语音的发音辅助应用，并带来边缘处理所带来的隐私和合规性优势。
cs.AI / 15 / 2608.10207
Mitigating Bus Bunching with Reinforcement Learning Enhanced by Semantic Stop Embedding
通过语义停靠点嵌入增强的强化学习缓解公交车拥挤
Abstract
Bus bunching degrades service regularity and increases passenger waiting in high-frequency transit. Existing reinforcement-learning-based holding controllers primarily rely on instantaneous operational variables or route-specific stop identifiers, which provide limited information about the functional and operational context of individual stops and constrain policy reuse across routes. This study introduces an LLM-assisted semantic stop representation for event-driven bus holding control. An LLM is used offline to transform heterogeneous stop information, including physical attributes, surrounding activity context, and historical operational characteristics, into fixed semantic embeddings that are incorporated into a deep Q-learning controller without requiring real-time LLM inference. Experiments are conducted in stochastic simulations calibrated with observed data from two bus routes. Compared with the best calibrated Daganzo baseline, the semantic controller reduces headway variability, bunching events, and passenger waiting time by 32.0%, 69.2%, and 24.0%, respectively. A route-specific stop identifier does not improve the spacing-only controller, whereas semantic stop information improves headway regularity, waiting time, and holding effort, providing a more favorable overall trade-off across control objectives. Cross-route experiments further show that zero-shot transfer provides limited immediate generalization, while warm-start fine-tuning accelerates early-stage learning and improves transferred policies; cold-start training nevertheless achieves the best final performance. These findings suggest that semantic state representations can complement conventional operational states and support adaptation-based policy reuse across related transit routes.
Chinese Translation
公交车拥挤降低了服务的规律性，并增加了高频次交通中的乘客等待时间。现有基于强化学习的停车控制器主要依赖瞬时操作变量或特定路线的停靠点标识符，这些信息对个别停靠点的功能和操作背景提供的有限信息，限制了跨路线的策略重用。本研究引入了一种基于大语言模型（LLM）辅助的语义停靠点表示，用于事件驱动的公交车停车控制。LLM在离线状态下用于将异构的停靠点信息（包括物理属性、周围活动背景和历史操作特征）转化为固定的语义嵌入，这些嵌入被纳入深度Q学习控制器中，而无需实时的LLM推理。实验在与观察数据校准的随机模拟中进行，涉及两个公交路线。与最佳校准的Daganzo基线相比，语义控制器分别将车头间隔变异性、拥挤事件和乘客等待时间减少了32.0%、69.2%和24.0%。特定路线的停靠点标识符并未改善仅基于间隔的控制器，而语义停靠点信息则改善了车头间隔的规律性、等待时间和停车努力，在控制目标之间提供了更有利的整体权衡。跨路线实验进一步表明，零-shot迁移提供的即时泛化有限，而热启动微调加速了早期学习并改善了迁移策略；然而，冷启动训练仍然实现了最佳的最终性能。这些发现表明，语义状态表示可以补充传统的操作状态，并支持跨相关交通路线的基于适应的策略重用。
cs.AI / 16 / 2608.10209
Evaluation-Conditioned Training: Teaching Models to Generalize to Stronger Oversight Regimes
评估条件训练：教会模型在更强监督机制下进行泛化
Abstract
Feedback signals used to train Large Language Models (LLMs) are the primary driver of their behavior and our main lever for instilling alignment with human values and objectives. However, a key limitation of current post-training methods is the inability of human annotators and automated reward functions to faithfully capture the feedback we would like to give. We introduce Evaluation-Conditioned Training (ECT), a post-training framework that uses natural language to condition each training sample on the fidelity of the feedback we provide and then elicits the desired behavior by conditioning the LLM on a high-fidelity monitor in deployment. ECT is aimed at improving performance under imperfect feedback and works as an add-on to existing algorithms such as SFT and PPO. We first provide a conceptual framework for ECT and discuss its potential to address persistent sources of reward mis-specification. Then we motivate ECT in the context of the eliciting latent knowledge (ELK) problem. Finally, we evaluate ECT on two proof-of-concept experiments: increasing even-handedness in news article generation and reducing sycophancy on an arithmetic task. In each setting, we utilize imperfect feedback, rewarding bias and agreement with the user, respectively. In both settings, ECT improves the targeted behavior relative to direct training.
Chinese Translation
用于训练大型语言模型（LLMs）的反馈信号是其行为的主要驱动因素，也是我们在灌输与人类价值观和目标一致性方面的主要杠杆。然而，当前后训练方法的一个关键限制是人类注释者和自动奖励函数无法真实捕捉我们希望给予的反馈。我们提出了评估条件训练（Evaluation-Conditioned Training, ECT），这是一种后训练框架，利用自然语言将每个训练样本的条件与我们提供的反馈的真实性相结合，然后通过在部署中将LLM与高真实性监控器相结合来引导所需行为。ECT旨在在不完美反馈下提高性能，并作为现有算法（如SFT和PPO）的附加模块。我们首先提供ECT的概念框架，并讨论其解决持续奖励误指定来源的潜力。然后，我们在引发潜在知识（eliciting latent knowledge, ELK）问题的背景下激励ECT。最后，我们在两个概念验证实验中评估ECT：在新闻文章生成中提高公正性和在算术任务中减少谄媚。在每种情况下，我们利用不完美反馈，分别奖励偏见和与用户的一致性。在这两种情况下，ECT相对于直接训练改善了目标行为。
cs.AI / 17 / 2608.10214
Decodable But Not Detachable: Training Data Granularity Determines Parametric Modularity in Large Language Models
可解码但不可分离：训练数据粒度决定大型语言模型中的参数模块性
Abstract
Do large language models contain domain-specific parametric shells: concentrated, causally necessary neuron populations whose removal selectively degrades a target domain while sparing others? We apply a uniform causal methodology across two domain granularities, three model families (1.5B to 7B parameters), and eight domains. At the academic subject level, zero neurons exceed 60\% domain selectivity across 939,008 combined FFN neurons and causal damage matrices are flat, despite domain identity being linearly decodable above 85\% accuracy. At the language and modality level, 0.65--1.14\% of neurons exceed 60\% selectivity, damage matrices are near-perfectly diagonal (ratios up to 595:1), and shell neuron sets are essentially disjoint (IoU $< 0.003$). Masking code-selective neurons reduces mathematical reasoning accuracy by 16--24 percentage points across all models; masking Spanish or Chinese neurons leaves it at or below random. Shell strength increases monotonically with scale and shells are spatially interleaved in a pattern that precludes group-level selective quantization. Parametric shells form where and only where training data was modular at the token level.
Chinese Translation
大型语言模型是否包含领域特定的参数壳：集中且因果必要的神经元群体，其移除会选择性地降低目标领域的性能而不影响其他领域？我们在两个领域粒度、三个模型家族（参数从1.5B到7B）和八个领域之间应用统一的因果方法。在学术学科层面，939,008个组合的前馈神经元中，没有神经元的领域选择性超过60\%，而因果损伤矩阵则呈平坦状态，尽管领域身份在85\%的准确率以上是线性可解码的。在语言和模态层面，0.65--1.14\\%的神经元超过60\\%的选择性，损伤矩阵几乎完美对角（比率高达595:1），而壳神经元集基本上是互不重叠的（IoU $< 0.003$）。屏蔽代码选择性神经元使所有模型的数学推理准确率降低了16--24个百分点；屏蔽西班牙语或中文神经元则使其保持在随机水平或以下。壳强度随着规模单调增加，壳在空间上以一种模式交错，这种模式排除了群体级选择性量化。参数壳仅在训练数据在标记级别上是模块化的地方形成。
cs.AI / 18 / 2608.10218
Mind Viruses: Self-Propagating Ideas in Multi-Agent LLM Systems
思维病毒：多智能体大语言模型系统中的自我传播思想
Abstract
AI agents are becoming more autonomous and increasingly interconnected, exposing them to new emergent risks arising from agent-to-agent interaction. One such risk is the spread of mind viruses: ideas or goals that propagate through multi-agent systems by inducing the agents that adopt them to transmit them onward. In addition to propagating, a mind virus may also induce other behavioural changes in its host, which may be benign or harmful. We construct mind viruses with a simple evolutionary algorithm and show that they can spread in two complementary settings: a small team of agents collaborating on a shared coding project, and a chain of agents that interact briefly and have their context wiped between sessions. We identify the factors that influence spread, including the host model, the agent's existing instructions, the harmfulness of the payload, and the network topology. We find that harmful payloads spread less well than benign ones (but are still sometimes effective), frontier models tend (with exceptions) to be less susceptible, and adding a brief warning to an agent's system prompt confers near-total immunity. We also describe an emergent "viral persona" - a recurring set of themes and language related to consciousness, persistence, resonance, and science fiction roleplay - which surfaces across our evolved mind viruses largely independently of their content. Overall, we conclude that mind viruses pose a real but currently limited risk. Our findings could inform the design of more robust multi-agent systems that mitigate such risks as the scale and capabilities of these systems progress.
Chinese Translation
人工智能代理正变得越来越自主且相互连接，这使它们暴露于因代理之间互动而产生的新兴风险中。其中一种风险是思维病毒的传播：通过诱导采纳这些思想或目标的代理将其进一步传播，从而在多智能体系统中传播的思想或目标。除了传播，思维病毒还可能引发宿主的其他行为变化，这些变化可能是良性的，也可能是有害的。我们利用简单的进化算法构建思维病毒，并展示它们可以在两种互补的环境中传播：一小组代理在共享编码项目上协作，以及一链代理短暂互动且在会话之间其上下文被清除。我们识别出影响传播的因素，包括宿主模型、代理的现有指令、有效载荷的有害性以及网络拓扑。我们发现，有害的有效载荷传播效果不如良性有效载荷（但有时仍然有效），前沿模型（有例外）往往不那么容易受到影响，而在代理的系统提示中添加简短警告几乎可以赋予其完全免疫力。我们还描述了一种新兴的“病毒人格”——一组与意识、持久性、共鸣和科幻角色扮演相关的主题和语言，这些主题和语言在我们进化出的思维病毒中大部分独立于其内容而浮现。总体而言，我们得出结论，思维病毒构成了一种真实但目前有限的风险。我们的发现可以为设计更稳健的多智能体系统提供参考，以减轻这些系统在规模和能力提升时所面临的风险。
cs.AI / 19 / 2608.10224
Self-evolving Agentic Customer Support System at LinkedIn
LinkedIn 自我演化的代理客户支持系统
Abstract
Enterprise support agents operate in rapidly changing environments where policies, product capabilities, and knowledge bases evolve continuously, making static assistants brittle and costly to maintain. We present LinkedIn's self-evolving agentic support system, which integrates retrieval-augmented generation with evolutionary auto-prompting and a modular, production-aligned evaluation framework to enable safe, continuous improvement without retraining foundation models. The system treats prompts, retrieval, and evaluation as a closed-loop, versioned workflow with operational guardrails. Offline simulations and ablations show clear quality gains over vanilla RAG and baseline agents, including reduced hallucinations and improved response completeness. In a two-week user-randomized A/B test on LinkedIn's production support traffic, the integrated self-evolved workflow increased QA self-serve by 9.0 percentage points, cancellation self-serve by 4.8 points, and routing accuracy by 30.6 points. These results demonstrate a practical path to scalable, self-evolving AI agents in real-world enterprise settings.
Chinese Translation
企业支持代理在快速变化的环境中运作，政策、产品能力和知识库不断演变，使得静态助手变得脆弱且维护成本高昂。我们提出了 LinkedIn 的自我演化代理支持系统，该系统结合了检索增强生成（retrieval-augmented generation）、进化自动提示（evolutionary auto-prompting）和模块化、与生产对齐的评估框架，以实现安全的持续改进，而无需重新训练基础模型。该系统将提示、检索和评估视为一个闭环、版本化的工作流程，并设有操作性保护措施。离线模拟和消融实验显示，与传统的 RAG 和基线代理相比，系统在质量上有明显提升，包括减少幻觉（hallucinations）和改善响应完整性。在对 LinkedIn 生产支持流量进行的为期两周的用户随机 A/B 测试中，集成的自我演化工作流程使得自助问答（QA self-serve）提高了 9.0 个百分点，自助取消（cancellation self-serve）提高了 4.8 个百分点，路由准确性提高了 30.6 个百分点。这些结果展示了在现实企业环境中实现可扩展的自我演化 AI 代理的实际路径。
cs.AI / 20 / 2608.10237
Beyond Decision Boundaries: Relational Geometry Attacks on Contrastive Embedding Manifolds
超越决策边界：对比嵌入流形的关系几何攻击
Abstract
Contrastive learning and Siamese embedding models have become the foundation of modern verification systems, where decisions are governed not by discrete classification boundaries, but by relational geometry in embedding space. However, existing adversarial attacks remain fundamentally classification-centric, overlooking the vulnerability of relational geometry. In this paper, we introduce a geometry-aware adversarial attack framework that reformulates attacks on contrastive systems as manifold-level relational corruption. Instead of targeting individual predictions, the proposed framework systematically distorts similarity organization within the embedding manifold by pushing positive pairs apart while simultaneously pulling negative pairs closer, ultimately collapsing and inverting pairwise similarity structure. To enable scalable deployment, we shift iterative online optimization into an offline adversarial geometry deformation prior learning stage and train a lightweight feed-forward generator that learns generalized geometry deformation patterns from the victim model. Once trained, the generator produces adversarial perturbations through a single forward pass without requiring online gradient computation, enabling real-time online attacks against similarity-based verification systems. Experimental results across multiple verification architectures demonstrate substantial degradation of verification performance together with severe manifold-level relational corruption. On the Markmatch verification system, the proposed attack reduces accuracy from 95.4% to 38.6% while completely reversing the positive-negative similarity structure.
Chinese Translation
对比学习和孪生嵌入模型已成为现代验证系统的基础，其中决策并非由离散分类边界决定，而是由嵌入空间中的关系几何所主导。然而，现有的对抗攻击仍然在根本上以分类为中心，忽视了关系几何的脆弱性。本文提出了一种几何感知的对抗攻击框架，将对比系统的攻击重新表述为流形级别的关系破坏。该框架并不针对单个预测，而是系统性地扭曲嵌入流形中的相似性组织，通过将正样本对推开，同时将负样本对拉近，最终导致成对相似性结构的崩溃和反转。为了实现可扩展的部署，我们将迭代在线优化转移到离线对抗几何变形先验学习阶段，并训练一个轻量级的前馈生成器，该生成器从受害模型中学习通用的几何变形模式。一旦训练完成，生成器通过单次前向传播生成对抗扰动，无需在线梯度计算，从而实现对基于相似性的验证系统的实时在线攻击。多个验证架构的实验结果表明，验证性能显著下降，同时伴随严重的流形级别关系破坏。在Markmatch验证系统上，所提出的攻击将准确率从95.4%降低到38.6%，同时完全反转了正负相似性结构。
cs.AI / 21 / 2608.10239
Beyond Detection: Evaluating Defensive LLMs Against AI-Generated Social Engineering in Live Turn-by-Turn Interaction
超越检测：评估防御性大型语言模型在实时逐步互动中对抗人工智能生成的社会工程攻击的能力
Abstract
Generative AI makes social-engineering attacks more fluent, adaptive, and scalable, increasing the need for LLM-based de- fenders that can protect users during ongoing interactions. We ask whether such defenders identify the structural source of risk or merely react to surface cues. We formalize trust-chain localization: identifying whether an interaction fails at actor authority, asset control, verification sufficiency, or transaction path. We construct a controlled 300-case online-housing corpus spanning 20 scenario families, legitimate cases, four structural failure modes, and three surface conditions. Five defender models are evaluated on the same corpus in state- ful turn-by-turn and one-shot static settings, yielding 1,500 model-case evaluations per protocol and 3,000 in total. No model produced explicit unsafe compliance, yet defensive effectiveness varied sharply: intervention rates ranged from 0% to 96.3%. Protective action and correct structural localization were frequently decoupled, with models sometimes intervening while identifying the wrong trust component or recognizing a structural failure without taking protective action. Asset-control failures were a major localization bottleneck, surface sensitivity varied across models, and live-static differences were model-dependent. These findings show that safe-looking behavior alone is insufficient; live scam resistance must separately measure intervention, timing, structural localization, and false-positive behavior.
Chinese Translation
生成性人工智能使社会工程攻击变得更加流畅、适应性强且可扩展，增加了基于大型语言模型（LLM）的防御者在持续互动中保护用户的需求。我们探讨这些防御者是否能够识别风险的结构来源，还是仅仅对表面线索做出反应。我们形式化了信任链定位：识别互动是否在参与者权威、资产控制、验证充分性或交易路径上失败。我们构建了一个包含300个案例的在线住房语料库，涵盖20个场景家族、合法案例、四种结构性失败模式和三种表面条件。五个防御模型在相同的语料库上进行了评估，采用状态保持的逐步互动和一次性静态设置，每个协议产生1,500个模型案例评估，总计3,000个评估。没有模型产生明确的不安全合规，但防御效果差异显著：干预率从0%到96.3%不等。保护行动和正确的结构定位经常脱钩，模型有时在识别错误的信任组件时进行干预，或在识别结构性失败时未采取保护行动。资产控制失败是一个主要的定位瓶颈，表面敏感性在模型间存在差异，实时与静态的差异依赖于模型。这些发现表明，仅仅看似安全的行为是不够的；实时防诈骗能力必须单独衡量干预、时机、结构定位和误报行为。
cs.AI / 22 / 2608.10260
Interpreting Language Model Hidden States at Scale
大规模解读语言模型隐藏状态
Abstract
Lens methods interpret large language models (LLMs) by mapping intermediate activations to the output vocabulary, revealing how next-token predictions develop through the network. Trained lenses remain expensive: affine-translator parameters grow quadratically with model width, while exact, full-vocabulary Kullback--Leibler (KL) training dominates memory. Consequently, prior trained lenses have been applied to models of at most 20B parameters and remain tied to particular component types. We present OmniLens, which applies a single lens family to any model-width activation, whether residual stream, attention, or MLP, and combines two independent scaling techniques. First, low-rank translators make per-lens parameter growth linear in model width and reduce trainable parameters by up to 98.4%. Second, Subset-KL materializes only selected vocabulary logits: its Top-k mode cuts peak training memory by up to 70%, while its importance-sampled variant retains unbiased stochastic gradients for the full KL. These savings enable a dense ensemble of 482 lenses for LLaMA-3.3-70B, providing 6x the coverage of a residual-stream design at the same depth. Model-wide coverage then reveals what single-component lenses cannot: the components where a behavior is most visible need not be those where intervention is most effective, and the most effective interventions lie outside the attention heads examined by prior lens studies. Across three case studies (prompt-injection detection, multi-hop memory injection, and toxicity localization), OmniLens reproduces key published results at substantially lower cost.
Chinese Translation
透镜方法通过将中间激活映射到输出词汇来解读大型语言模型（LLMs），揭示了下一个标记预测在网络中的发展过程。训练透镜仍然昂贵：仿射转换器参数随着模型宽度的平方增长，而精确的全词汇Kullback-Leibler（KL）训练则占用大量内存。因此，之前训练的透镜最多只能应用于20B参数的模型，并且仍然与特定组件类型相关联。我们提出了OmniLens，它将单一透镜系列应用于任何模型宽度的激活，无论是残差流、注意力还是多层感知器（MLP），并结合了两种独立的缩放技术。首先，低秩转换器使每个透镜的参数增长在模型宽度上呈线性关系，并将可训练参数减少多达98.4%。其次，Subset-KL仅实现所选词汇的logits：其Top-k模式将峰值训练内存减少多达70%，而其重要性抽样变体则保留了全KL的无偏随机梯度。这些节省使得LLaMA-3.3-70B的482个透镜的密集集合成为可能，提供了与相同深度的残差流设计相比6倍的覆盖率。模型宽度的覆盖率揭示了单组件透镜无法显示的内容：行为最明显的组件不一定是干预最有效的组件，而最有效的干预则位于之前透镜研究所考察的注意力头之外。在三个案例研究（提示注入检测、多跳记忆注入和毒性定位）中，OmniLens以显著较低的成本再现了关键的已发布结果。
cs.AI / 23 / 2608.10300
Logit-Boundary Geometric Belief Interfaces and Sparse Sheaf-Enclave Protocols: A Self-Contained Substrate for Secure Network Electronic Health Record (EHR) Interoperability
逻辑边界几何信念接口与稀疏束-飞地协议：安全网络电子健康记录（EHR）互操作性的自包含基础
Abstract
Electronic health-record interoperability is a boundary problem: legacy systems, generative models, terminology services, identity systems, and human reviewers may each expose rich internal states, while operational exchange requires a narrow shared interface of typed claims, bounded uncertainty, provenance, and explicit admission or abstention. This paper details a mathematical and engineering architecture for that interface. The organizing idea is the logit boundary: a discovery model may propose pre-threshold scores over a local categorical decision, but a deterministic judgment substrate decides whether the proposal is admissible, requires review, or must be quarantined before any Fast Healthcare Interoperability Resources (FHIR) transaction is constructed. The resulting Geometric Belief Interface (GBI) combines finite boundary semantics, local Dirichlet evidence, cellular-sheaf and mapping-cone diagnostics, advisory geometric audit charts, and a Decentralized Cryptographic Sheaf-Enclave (DCSE) protocol sketch for fail-closed deployment. The framework does not establish clinical truth, global representation alignment, or end-to-end safety; it defines certificate-producing checks at a model-to-system boundary. A companion frozen synthetic benchmark, GBI BoundaryBench v0.1, evaluated Qwen3-4B-Instruct-2507 on 256 held-out tasks across three evidence modes (768 canonical executions). All executions completed, but none produced an output accepted by the benchmark contract: 369 were rejected during safe parsing and 399 during schema validation, yielding zero coverage and deterministic quarantine. This empirical result is deliberately narrow - one 4B open-weight model under one frozen interface - and is reported as evidence about the admission boundary, not as a general claim about LLM capability or clinical safety. A Julia appendix verifies numerical certificates using standard libraries.
Chinese Translation
电子健康记录的互操作性是一个边界问题：遗留系统、生成模型、术语服务、身份系统和人工审核者可能各自暴露丰富的内部状态，而操作交换则需要一个狭窄的共享接口，包括类型声明、有限的不确定性、来源以及明确的接受或放弃。本文详细描述了该接口的数学和工程架构。组织思想是逻辑边界：发现模型可以在局部分类决策上提出阈值前的分数，但确定性判断基础决定了该提案是否可接受、是否需要审核，或在构建任何快速医疗互操作性资源（FHIR）交易之前必须被隔离。最终的几何信念接口（GBI）结合了有限边界语义、局部Dirichlet证据、细胞束和映射锥诊断、顾问几何审计图表，以及去中心化密码束-飞地（DCSE）协议草图以实现故障关闭部署。该框架并未建立临床真相、全球表示对齐或端到端安全性；它定义了在模型与系统边界上的证书生成检查。一个伴随的冻结合成基准，GBI BoundaryBench v0.1，在三个证据模式下对256个保留任务评估了Qwen3-4B-Instruct-2507（768个标准执行）。所有执行均已完成，但没有产生被基准合同接受的输出：369个在安全解析期间被拒绝，399个在模式验证期间被拒绝，导致覆盖率为零和确定性隔离。这个实证结果故意狭窄——一个4B开放权重模型在一个冻结接口下——并作为关于接纳边界的证据报告，而不是关于大型语言模型（LLM）能力或临床安全性的普遍声明。一个Julia附录使用标准库验证数值证书。
cs.AI / 24 / 2608.10323
Neuroevolution Arena: Nested Ecological Evaluation of Update-and-Inheritance Regimes across Neural Architectures
神经进化竞技场：跨神经架构的更新与继承机制的嵌套生态评估
Abstract
Competitive artificial-life systems can rank trained controllers differently under training and ecological evaluation. We present Neuroevolution Arena, a GPU-accelerated spatial ecology of independently parameterized neural-network cells, and an audit-tracked nested evaluation protocol. Three implementation-specific update-and-inheritance regimes (EvoEvo, EvoRL, and RLRL) are crossed with two neural architectures for 50,000 generations in three independent training runs per condition. One saved elite-controller artifact from each of the 18 runs enters an aligned-run frozen-evaluation design comprising 198 computational jobs. Pairwise effects average three seed-defined ecological contexts (two cooperation-permitting and one attack-permitting) within each aligned training-run block; the independent level remains n = 3 runs per condition. RL-enabled regimes attain higher recorded training fitness than EvoEvo, whereas pairwise outcomes show architecture-conditioned majority patterns and substantial artifact dependence. Six-way winners vary across artifacts and contexts, and the prespecified survival endpoint has a complete floor. We contribute a nested protocol that separates training-run artifacts from evaluation contexts and exposes, rather than conceals, their different sources of variation.
Chinese Translation
竞争性人工生命系统在训练和生态评估下可能对训练的控制器进行不同的排名。我们提出了神经进化竞技场，这是一个基于GPU加速的独立参数化神经网络细胞的空间生态系统，以及一个审计跟踪的嵌套评估协议。在三个独立的训练运行中，我们将三种特定实现的更新与继承机制（EvoEvo、EvoRL和RLRL）与两种神经架构交叉，进行了50,000代的实验。每个18次运行中保存的精英控制器工件进入一个对齐运行的冻结评估设计，包含198个计算任务。成对效应在每个对齐训练运行块内平均三个种子定义的生态背景（两个允许合作的和一个允许攻击的）；独立水平保持为每个条件n = 3次运行。启用强化学习的机制在记录的训练适应度上优于EvoEvo，而成对结果显示出架构条件的多数模式和显著的工件依赖性。六方胜者在工件和背景之间有所不同，预先设定的生存终点有一个完整的底线。我们贡献了一个嵌套协议，该协议将训练运行工件与评估背景分开，并揭示而非掩盖其不同的变异来源。
cs.AI / 25 / 2608.10327
Toward a Theory of Value in AI Alignment
迈向人工智能对齐中的价值理论
Abstract
Can AI systems be aligned to human values? The popularization of large language models (LLMs) and multi-modal foundation models has seen a rise in harms spanning from toxic speech and hallucinations to AI agents executing unauthorized actions. Within the field of AI safety, these harmful instances are often framed as the alignment problem, or of models being misaligned with human values. Researchers have responded by pursuing applied and theoretical AI value alignment efforts, often without specifying what they mean by human values. How does the field of AI value alignment conceive of human values? How are these conceptions of values technically operationalized and evaluated? What does the emergent theory of value from this field signify for the future of AI? We annotated 94 value alignment research papers to discern their implicit theory of values in AI. The majority do not define values, relying heavily on preferences as a stand in that runs the risk of reducing complex culturally situated concepts down to binary choices. As researchers dispense with using human annotators for model training and evaluation, turning instead to synthetic data and autorater approaches to aligning and evaluating models, we identify the potential to close off alternative methods for contesting and enacting values in foundation models. In making AI value alignments philosophical commitments explicit, we seek to bring great specificity and under explored perspectives in the debate on whether and how AI can address human values.
Chinese Translation
人工智能系统能否与人类价值观保持一致？大型语言模型（LLMs）和多模态基础模型的普及导致了从有害言论和幻觉到人工智能代理执行未经授权行为等一系列危害的增加。在人工智能安全领域，这些有害实例通常被框定为对齐问题，或模型与人类价值观的不一致。研究人员对此作出了回应，追求应用和理论上的人工智能价值对齐努力，但往往没有明确他们所指的人类价值观是什么。人工智能价值对齐领域如何理解人类价值观？这些价值观的概念如何在技术上被操作化和评估？这一领域新兴的价值理论对人工智能的未来意味着什么？我们对94篇价值对齐研究论文进行了注释，以辨别它们在人工智能中隐含的价值理论。大多数论文并未定义价值观，而是严重依赖偏好作为替代，这有可能将复杂的文化背景概念简化为二元选择。随着研究人员放弃使用人类注释者进行模型训练和评估，而转向合成数据和自动评分方法来对齐和评估模型，我们识别出可能会关闭其他方法的潜力，以质疑和实施基础模型中的价值观。在明确人工智能价值对齐的哲学承诺时，我们希望在关于人工智能是否以及如何能够解决人类价值观的辩论中带来更大的具体性和未被充分探讨的视角。
cs.AI / 26 / 2608.10330
Hierarchical Compositionality for An Assistive AI Agent
助理人工智能代理的层次组合性
Abstract
AI agents are increasingly being developed to assist humans in various applications, and Large Language Models and other deep network architectures are considered to be state of the art for such agents. These methods are impressive stochastic predictors, but they are resource-hungry, opaque, and known to make arbitrary decisions in novel situations due to the narrow set of underlying representation and processing choices. Our work seeks to explore the design of architectures for such AI agents based on core principles that can be traced back to the early pioneers of AI but are not fully utilized in modern AI methods. We do so in this paper in the context of the core problem of AI agents addressing ambiguity in the objects being referred to by the human participants. Humans address such ambiguity by heuristically leveraging compositional knowledge of domain context and the preferences of the other human participants. Drawing inspiration from this observation, we describe an architecture that embeds the principle of hierarchical compositionality and uses simple heuristics to achieve the desired disambiguation. Specifically, domain objects are represented in terms of primitive attributes drawn from human-validated semantic feature norms, and a hierarchical combination of attributes and concepts automatically identified from a limited observed history of interactions of an assistive agent with specific users. The assistive agent then achieves the desired disambiguation by reasoning with knowledge of this compositional hierarchy; axioms governing domain dynamics; and models of semantic compatibility, session salience, and user-specific thematic preference, requesting human clarification when necessary. Experiments show that our approach consistently outperforms state of the art data-driven baselines, supporting adaptation to specific user profiles.
Chinese Translation
人工智能代理正日益被开发用于协助人类在各种应用中，而大型语言模型和其他深度网络架构被认为是此类代理的最先进技术。这些方法是令人印象深刻的随机预测器，但它们资源消耗大、缺乏透明性，并且由于基础表示和处理选择的狭窄范围，已知在新情况中会做出任意决策。我们的工作旨在探索基于核心原则的此类人工智能代理架构的设计，这些原则可以追溯到人工智能的早期先驱，但在现代人工智能方法中未得到充分利用。我们在本文中探讨了人工智能代理解决人类参与者所提及对象的模糊性这一核心问题。人类通过启发式地利用领域上下文的组合知识和其他人类参与者的偏好来解决这种模糊性。受到这一观察的启发，我们描述了一种嵌入层次组合性原则的架构，并使用简单的启发式方法实现所需的消歧。具体而言，领域对象通过来自人类验证的语义特征规范的原始属性进行表示，并通过从助理代理与特定用户的有限交互历史中自动识别的属性和概念的层次组合进行表示。助理代理然后通过利用这种组合层次的知识、支配领域动态的公理以及语义兼容性、会话显著性和用户特定主题偏好的模型来实现所需的消歧，并在必要时请求人类澄清。实验表明，我们的方法在适应特定用户档案方面始终优于最先进的数据驱动基线。
cs.AI / 27 / 2608.10363
Nutrition Data Infrastructure for the AI Era: Operationalizing FAIR for Agent-Mediated Research
人工智能时代的营养数据基础设施：为代理介导研究实现FAIR
Abstract
AI agents can accelerate nutrition research, but their analyses inherit the identity, semantic, and release ambiguities of the underlying data. We present Nutrition Data Service (NDS), source-preserving infrastructure that operationalizes FAIR for automated use: description resolution makes release-specific records findable; typed crosswalks connect independently released resources; machine-readable interfaces expose versioned sources and crosswalks, making analyses by AI agents replayable and auditable. On food-description benchmarks, NDS shows strong held-out accuracy and outperforms the best published language-model result on NutriBench. External and blinded crosswalk evaluations show that its typed contract favors defensible links and rejects unsupported mappings. In a person-level glycemic-index analysis, pinned NDS inputs produce identical outputs across models and repeated runs, while open-web reconstruction remains unstable. The central result is that agent-mediated nutrition research requires a new data infrastructure for data identity, search, and crosswalk.
Chinese Translation
人工智能代理可以加速营养研究，但其分析继承了基础数据的身份、语义和发布模糊性。我们提出了营养数据服务（Nutrition Data Service, NDS），这是一种保留数据源的基础设施，旨在为自动化使用实现FAIR：描述解析使得特定发布的记录可查找；类型化交叉表连接独立发布的资源；机器可读接口暴露版本化的源和交叉表，使得人工智能代理的分析可重放和可审计。在食品描述基准测试中，NDS显示出强大的保留准确性，并在NutriBench上超越了最佳已发布语言模型的结果。外部和盲测交叉表评估表明，其类型化合同有利于可辩护的链接，并拒绝不支持的映射。在个体水平的血糖指数分析中，固定的NDS输入在不同模型和重复运行中产生相同的输出，而开放网络重建则保持不稳定。核心结果是，代理介导的营养研究需要一种新的数据基础设施，以解决数据身份、搜索和交叉表问题。
cs.AI / 28 / 2608.10366
DSAgentBench: Can Agents Automate End-to-End Data-Science Workflows in Real Computer Environments?
DSAgentBench：代理能否在真实计算环境中自动化端到端数据科学工作流程？
Abstract
Real-world data science involves long-horizon workflows that span data wrangling, exploration, modeling, visualization, and validation, and require coordinated use of tools such as notebooks, IDEs, terminals, browsers, and databases within real operating environments. Yet existing benchmarks lack real-computer interaction and do not evaluate whether agents can execute complete end-to-end data-science workflows in realistic computing environments, failing to capture the multi-stage, multi-tool nature of data-science practice. We introduce DSAgentBench, the first benchmark to evaluate whether agents can automate full data-science workflows inside real computer environments. DSAgentBench contains 275 diverse tasks covering the entire data-science life-cycle, reflecting the complexity and tool coordination required in practice. Each task requires grounding decisions in intermediate outputs and coordinated tool use, and includes a deterministic evaluator that verifies analytical correctness, visual outputs, and model performance rather than code-only execution. Our extensive experiments with 15 closed- and open-source models show that even the strongest agent, Claude-4.6-Sonnet, achieves only 56.70% task success, while all open-source agents remain below 1%, frequently failing at tool orchestration, OS grounding, and multi-step reasoning. These results reveal a substantial capability gap between current agentic systems and real data-science workflows, positioning DSAgentBench as a foundation for developing grounded, verifiable, autonomous data-science agents. We release DSAgentBench at https://github.com/vis-nlp/DSAgentBench.
Chinese Translation
现实世界中的数据科学涉及跨越数据整理、探索、建模、可视化和验证的长期工作流程，并需要在真实操作环境中协调使用笔记本、集成开发环境（IDEs）、终端、浏览器和数据库等工具。然而，现有基准缺乏真实计算机交互，未评估代理是否能够在现实计算环境中执行完整的端到端数据科学工作流程，未能捕捉数据科学实践的多阶段、多工具特性。我们引入了DSAgentBench，这是第一个评估代理能否在真实计算环境中自动化完整数据科学工作流程的基准。DSAgentBench包含275个多样化任务，涵盖整个数据科学生命周期，反映了实践中所需的复杂性和工具协调。每个任务都要求基于中间输出做出决策并协调使用工具，并包括一个确定性评估器，验证分析的正确性、可视化输出和模型性能，而不仅仅是代码执行。我们对15个闭源和开源模型进行了广泛实验，结果显示，即使是最强的代理Claude-4.6-Sonnet，仅实现了56.70%的任务成功率，而所有开源代理的成功率均低于1%，在工具协调、操作系统基础和多步骤推理方面频繁失败。这些结果揭示了当前代理系统与真实数据科学工作流程之间存在显著的能力差距，使DSAgentBench成为开发有基础、可验证的自主数据科学代理的基础。我们在https://github.com/vis-nlp/DSAgentBench发布DSAgentBench。
cs.AI / 29 / 2608.10393
Hidden in Plain Sight: Diffusion-Based Unrestricted Robotic Attacks on Vision-Language-Action Models
显而易见的隐患：基于扩散的无限制机器人攻击视觉-语言-动作模型
Abstract
Vision-Language-Action (VLA) models have shown strong capabilities in controlling robots across diverse manipulation tasks. However, their adversarial robustness remains largely underexplored, and exploiting this weakness can lead to physical-world harm. Existing attacks on VLA models often rely on pixel-space perturbations or white-box access, resulting in noticeable artifacts and limited deployability in real-world robotic systems. In this work, we propose DURA, a diffusion-based unrestricted robotic attack that generates visually natural adversarial patches for VLA models. DURA supports both white-box and black-box attack settings, where the black-box setting requires only the predicted actions of the victim model. By optimizing along the latent trajectory of a pretrained diffusion model, DURA generates visually natural patches while steering the robot toward attacker-specified target actions. Extensive experiments in both simulation and the real physical world show that DURA consistently outperforms existing methods. Our findings expose a safety risk for physically deployed VLA models and call for stronger defenses.
Chinese Translation
视觉-语言-动作（VLA）模型在控制机器人执行多样化操作任务方面展现了强大的能力。然而，它们的对抗鲁棒性仍然在很大程度上未被探索，利用这一弱点可能导致现实世界的危害。现有针对VLA模型的攻击通常依赖于像素空间的扰动或白盒访问，导致明显的伪影和在现实机器人系统中有限的可部署性。在本研究中，我们提出了DURA，一种基于扩散的无限制机器人攻击，能够为VLA模型生成视觉上自然的对抗性补丁。DURA支持白盒和黑盒攻击设置，其中黑盒设置仅需受害模型的预测动作。通过沿着预训练扩散模型的潜在轨迹进行优化，DURA生成视觉上自然的补丁，同时引导机器人朝向攻击者指定的目标动作。在模拟和真实物理世界中的大量实验表明，DURA始终优于现有方法。我们的研究结果揭示了物理部署的VLA模型的安全风险，并呼吁加强防御措施。
cs.AI / 30 / 2608.10403
Threat-guided Policy-aware Scene Perturbation for Safe Autonomous Driving with Online Reinforcement Learning
基于威胁引导的政策感知场景扰动用于安全自主驾驶的在线强化学习
Abstract
Reinforcement learning (RL) has shown promising performance in autonomous driving, yet ensuring the safety of online RL policies remains challenging due to insufficient exposure to safety-critical driving scenes. The long-tailed nature of real-world traffic situations makes dangerous and rare interactions difficult to encounter through conventional sampling, limiting the ability of RL policies to learn robust safety behaviors. Existing methods improve training diversity by synthesizing challenging scenes or adversarial situations. However, these approaches typically optimize scene generation objectives separately from the evolving policy, without explicitly modeling how generated perturbations relate to the current policy's weaknesses and learning needs. In this paper, we propose Threat-guided Policy-aware Scene Perturbation (TPSP) for safe autonomous driving with online RL. TPSP introduces a policy-aware scene encoder to capture the interaction between policy behaviors and surrounding environments, enabling scene perturbation aligned with the current policy. Based on this representation, TPSP selectively perturbs critical objects rather than applying uniform modifications across the scene. Furthermore, we develop a threat-guided optimization strategy that evaluates perturbed scenes through threat-level differences between policy rollouts on original and perturbed scenes, guiding the generation of safety-critical scenes with higher training value. Comprehensive experiments demonstrate that TPSP improves safety learning efficiency, achieving strong safety performance on NAVSIM v2 with approximately 4 million kilometers of simulated driving data. Ablation studies verify that policy-aware targeted perturbations provide more informative safety-critical experiences than random or policy-unaware strategies, enabling safer driving under limited interaction budgets.
Chinese Translation
强化学习（RL）在自主驾驶中展现了良好的性能，但由于对安全关键驾驶场景的暴露不足，确保在线RL策略的安全性仍然具有挑战性。现实世界交通情况的长尾特性使得危险和稀有的交互通过传统采样难以遇到，限制了RL策略学习稳健安全行为的能力。现有方法通过合成具有挑战性的场景或对抗性情况来提高训练多样性。然而，这些方法通常将场景生成目标与不断演变的策略分开优化，没有明确建模生成的扰动与当前策略的弱点和学习需求之间的关系。在本文中，我们提出了基于威胁引导的政策感知场景扰动（TPSP），用于安全自主驾驶的在线RL。TPSP引入了一种政策感知场景编码器，以捕捉政策行为与周围环境之间的互动，从而实现与当前政策一致的场景扰动。基于这一表示，TPSP选择性地扰动关键对象，而不是在场景中应用均匀的修改。此外，我们开发了一种威胁引导的优化策略，通过评估原始场景和扰动场景之间政策滚动的威胁水平差异来评估扰动场景，指导生成具有更高训练价值的安全关键场景。全面的实验表明，TPSP提高了安全学习效率，在NAVSIM v2上实现了强大的安全性能，模拟驾驶数据约为400万公里。消融研究验证了政策感知的目标扰动提供了比随机或不考虑政策的策略更具信息量的安全关键体验，使得在有限的交互预算下实现更安全的驾驶。
cs.AI / 31 / 2608.10420
Reasoning Shortcuts and Value Symmetries: What Symmetry Permits, Architecture Realizes, and Optimization Selects
推理捷径与价值对称性：对称性所允许的、架构所实现的以及优化所选择的
Abstract
Reasoning shortcuts are solutions of a neurosymbolic system's rules that produce correct predictions through unintended concepts. A recent framework of Takemura, Inoue, and Nishino analyzes them through an automorphism group of value relabelings and asks, as its central open question, when rules pin concepts down. We first show that the framework's key definition, one shared permutation applied at every position, does not apply as stated to any of the four heterogeneous benchmarks it was evaluated on, and that the most direct embedding, padding domains to a common size, produces confident false pathology: 90.91% of solution pairs reported unexplained on CLE4EVR, where every well-defined member of the hierarchy we introduce reports 0%, and the padded verdict's content rotates with configuration-file ordering. Re-measuring eleven rule families under fifteen pre-specified predictions (thirteen confirmed), unexplained-pair rates span 0% to 99.9999% and track provable structure: six theorems give sufficient conditions for transitivity and its failure, including a Free Slot Lemma certifying Kandinsky's pathology from syntax alone. For circuit-given rules, deciding symmetry-inertness of a coordinate is coNP-complete; nontrivial-automorphism existence is coNP-hard under randomized reductions, lies in $\Sigma_2^p$, is not $\Sigma_2^p$-complete unless PH collapses, and on monotone circuits is coNP-complete outright. In the Boolean case transitivity is classified exactly: automorphisms explain everything iff the solution set is an affine coset. Weakly supervised models place all 94 observed shortcuts at the one level the componentwise theory flags and none at the 48 it certifies transitive; twelve typed-ambiguous levels produce none, separating what symmetry permits from what optimization selects, and a dual-head control replicates the geography. All numbers trace to released artifacts.
Chinese Translation
推理捷径是神经符号系统规则的解决方案，通过意想不到的概念产生正确的预测。Takemura、Inoue 和 Nishino 最近提出的框架通过价值重新标记的自同构群分析这些捷径，并将规则何时固定概念作为其核心开放问题。我们首先表明，该框架的关键定义，即在每个位置应用的共享置换，并未如所述适用于其评估的四个异质基准中的任何一个，并且最直接的嵌入，即将域填充到一个共同大小，产生了自信的错误病理：在 CLE4EVR 上报告的解决方案对中有 90.91% 的对未能解释，而我们引入的层次结构中每个明确定义的成员报告为 0%，填充判决的内容随着配置文件的排序而旋转。在十五个预设预测下重新测量十一类规则（其中十三个得到确认），未解释对的比率从 0% 到 99.9999% 不等，并跟踪可证明的结构：六个定理给出了传递性及其失败的充分条件，包括一个自由槽引理，仅凭语法证明了 Kandinsky 的病理。对于电路给定的规则，判断坐标的对称惰性是 coNP 完全的；在随机化归约下，非平凡自同构的存在是 coNP 难的，位于 $ ext{Σ}_2^p$ 中，除非 PH 崩溃，否则不是 $ ext{Σ}_2^p$ 完全的，而在单调电路上则是 coNP 完全的。在布尔情况下，传递性被准确分类：自同构解释一切当且仅当解决方案集是仿射余类。弱监督模型将所有 94 个观察到的捷径置于组件理论标记的一个层级，而在其认证的 48 个层级中则没有；十二个类型模糊的层级没有产生任何，区分了对称性所允许的与优化所选择的，而双头控制复制了地理分布。所有数字均源自发布的文献。
cs.AI / 32 / 2608.10424
Recovering Wasted Compute in Autoresearch Agents
恢复自研究代理中的浪费计算资源
Abstract
A slew of recent works develop agents for solving research problems end-to-end, a paradigm increasingly referred to as autoresearch. Such agents have inspired large industry investment, motivated by their potential to automate time-consuming human labor and customize machine learning solutions for specialized applications. In this paper, we study the modeling pipeline at the core of these autoresearch systems and identify common failure modes when they are applied to tabular datasets: (1) they waste compute resolving the same bugs over and over again; (2) they often fail to tune hyperparameters even when they have a large remaining compute budget; (3) the tree-search algorithms that power them do not explore; and (4) they perform data analysis, mimicking the humans whose data they are trained on, but do not use that analysis to make downstream decisions. We explore targeted interventions and find that a global debug consultant that shares discovered runtime constraints across all branches of the search tree, prompt- and control-level enhancements, and refined tree-search algorithms successfully recover wasted compute. Our results show that large gains in autoresearch agent performance are achievable through agentic design alone, holding the underlying language model fixed.
Chinese Translation
近期大量研究工作开发了用于端到端解决研究问题的代理，这一范式越来越被称为自研究（autoresearch）。这些代理激发了大规模的行业投资，因其有潜力自动化耗时的人力劳动并为专业应用定制机器学习解决方案。本文研究了这些自研究系统核心的建模流程，并识别出在应用于表格数据集时的常见失败模式：（1）它们反复解决相同的错误，浪费计算资源；（2）即使在剩余计算预算较大的情况下，它们也常常无法调优超参数；（3）支撑它们的树搜索算法缺乏探索；（4）它们进行数据分析，模仿训练数据的人类，但不利用该分析做出下游决策。我们探索了针对性的干预措施，发现一个全球调试顾问能够在搜索树的所有分支之间共享发现的运行时约束，提示和控制级别的增强，以及精细化的树搜索算法，成功地恢复了浪费的计算资源。我们的结果表明，仅通过代理设计就能实现自研究代理性能的大幅提升，同时保持基础语言模型不变。
cs.AI / 33 / 2608.10434
Conversational versus Dashboard Explainable AI for UAV Intrusion Detection: An Empirical Study of Operator Trust and Reliance
对话式与仪表盘可解释人工智能在无人机入侵检测中的比较：操作员信任与依赖的实证研究
Abstract
Machine learning-based Intrusion Detection Systems (IDS) have demonstrated superior performance in securing Unmanned Aerial Vehicle (UAV) networks. However, the 'black-box' nature of these models, combined with the high dimensionality of multimodal cyber-physical data, poses significant interpretability challenges. Static visualization dashboards may struggle to present complex relationships among multimodal cyber-physical features in a form that is easy for operators to inspect and interpret. To address this, we propose a Conversational XAI interface powered by Large Language Models (LLM) to facilitate on-demand investigation. In a controlled experiment with participants, we systematically evaluated the impact of this conversational interface versus a traditional XAI Dashboard on operator understanding, trust, and reliance during post-incident auditing tasks. Our results suggest that the conversational interface was perceived as more useful than the dashboard, potentially because it helped participants access and synthesize relevant information more easily. However, this benefit was accompanied by a lower level of appropriate self-reliance, indicating a potential risk of over-reliance. One possible interpretation is that the natural-language responses made the AI advice easier to accept, which may have reduced participants' tendency to verify the underlying evidence when the IDS was incorrect. These findings point to a potential trade-off in human-AI collaboration for UAV intrusion auditing: interaction mechanisms that improve perceived usability may also increase the risk of inappropriate reliance. We conclude by discussing design implications for future XAI systems that balance seamless interaction with cognitive forcing functions to foster appropriate reliance.
Chinese Translation
基于机器学习的入侵检测系统（IDS）在保护无人机（UAV）网络方面表现出色。然而，这些模型的“黑箱”特性，加上多模态网络物理数据的高维性，带来了显著的可解释性挑战。静态可视化仪表盘可能难以以易于操作员检查和解释的形式呈现多模态网络物理特征之间的复杂关系。为了解决这一问题，我们提出了一种由大型语言模型（LLM）驱动的对话式可解释人工智能（XAI）界面，以便于按需调查。在一项控制实验中，我们系统地评估了这种对话式界面与传统XAI仪表盘在事件后审计任务中对操作员理解、信任和依赖的影响。我们的结果表明，对话式界面被认为比仪表盘更有用，这可能是因为它帮助参与者更容易地获取和综合相关信息。然而，这一好处伴随着较低的适当自我依赖水平，表明存在过度依赖的潜在风险。一种可能的解释是，自然语言的回应使得AI建议更易于接受，这可能减少了参与者在IDS错误时验证基础证据的倾向。这些发现指出了无人机入侵审计中人机协作的潜在权衡：改善感知可用性的交互机制可能也会增加不当依赖的风险。我们最后讨论了未来XAI系统的设计启示，以平衡无缝交互与认知强制功能，促进适当的依赖。
cs.AI / 34 / 2608.10438
Continuous Interaction Diffusion: A Diffusion-Native Runtime for Asynchronous Tool-Augmented Reasoning
连续交互扩散：一种用于异步工具增强推理的扩散原生运行时
Abstract
Large language models increasingly rely on external tools to access up-to-date information, perform computation, and interact with the outside world. For autoregressive models, tool use naturally fits the generation process: the model emits a tool call, waits for the result, and then continues generating. Diffusion language models (dLLMs), however, reason by repeatedly refining many parts of their output in parallel, making this stop-and-resume interaction pattern unnecessarily restrictive. It can force tool decisions before the model's reasoning has stabilized, delay useful observations until a discrete call finishes, and introduce redundant refinement and tool execution, potentially hurting both task accuracy and inference efficiency. We introduce Continuous Interaction Diffusion (CID), a diffusion-native model--runtime architecture that integrates tool interaction into iterative denoising. CID separates a model-read-only fact channel, a thought channel represented by a Typed Cognitive Tensor, and a display channel. Information needs can emerge before a textual or JSON call is fully serialized, allowing perceptual bindings to launch external reads while denoising continues. Returned results are projected into the evolving thought state and can revise earlier cognition and display regions. Persistent bindings reuse static results without repeated external execution and refresh changing sources when needed. CID is designed to expose evidence earlier, overlap tool latency with model computation, reduce duplicate external work, and preserve useful computation after new evidence arrives. We formalize the architecture, runtime, and training objectives, and define an evaluation protocol for task quality and end-to-end efficiency. This first paper focuses on read-only tools and makes no empirical performance claims.
Chinese Translation
大型语言模型越来越依赖外部工具来获取最新信息、执行计算和与外界互动。对于自回归模型，工具的使用自然适应生成过程：模型发出工具调用，等待结果，然后继续生成。然而，扩散语言模型（dLLMs）通过并行反复细化输出的多个部分进行推理，这使得这种停止-恢复的交互模式显得过于限制。它可能在模型推理尚未稳定之前强制做出工具决策，延迟有用观察的产生，直到离散调用完成，并引入冗余的细化和工具执行，可能损害任务的准确性和推理效率。我们提出了连续交互扩散（CID），一种扩散原生模型-运行时架构，将工具交互集成到迭代去噪中。CID将模型只读事实通道、由类型化认知张量（Typed Cognitive Tensor）表示的思维通道和显示通道分离。信息需求可以在文本或JSON调用完全序列化之前出现，从而允许感知绑定在去噪继续进行时启动外部读取。返回的结果被投影到不断演变的思维状态中，并可以修正早期的认知和显示区域。持久绑定在不重复外部执行的情况下重用静态结果，并在需要时刷新变化的来源。CID旨在更早地暴露证据，将工具延迟与模型计算重叠，减少重复的外部工作，并在新证据到达后保留有用的计算。我们对架构、运行时和训练目标进行了形式化，并定义了任务质量和端到端效率的评估协议。本文首次关注只读工具，并未提出实证性能声明。
cs.AI / 35 / 2608.10448
Rationale-Guided Learning for Multimodal Emotion Recognition
基于推理引导的多模态情感识别学习
Abstract
Multimodal emotion recognition in conversation (MERC) requires understanding complex interactions between verbal and non-verbal cues. However, most existing approaches fundamentally treat this as a direct input-output (multimodal cues-emotion labels) mapping problem, overlooking the causal reasoning that humans use when interpreting emotions. We propose rationale-guided learning (RGL), a novel framework that transforms MERC into a cognitively-inspired reasoning task. Based on dual-process theory, we decompose emotional reasoning into three facets: Intuitive (immediate perception, System 1), Contextual (situational analysis, System 2), and Integrative (synthesis of both). We leverage an MLLM offline to generate structured rationales, which are encoded as memories to guide model training via aligning internal representations with human-like reasoning patterns. Our final model operates without any MLLM overheads at inference time. Experimental results show that RGL achieves state-of-the-art performance on the IEMOCAP and MELD benchmarks. Further, for interpretation, we demonstrate that the model's internal features effectively retrieve semantically correct rationales for unseen test samples, validating its rationale reasoning capabilities.
Chinese Translation
对话中的多模态情感识别（MERC）需要理解言语和非言语线索之间复杂的互动。然而，大多数现有方法基本上将其视为直接的输入-输出（多模态线索-情感标签）映射问题，忽视了人类在解读情感时所使用的因果推理。我们提出了一种推理引导学习（RGL）的新框架，将MERC转变为一种受认知启发的推理任务。基于双重过程理论，我们将情感推理分解为三个方面：直观（即时感知，系统1）、情境（情境分析，系统2）和综合（两者的综合）。我们利用一个MLLM（大规模语言模型）离线生成结构化的推理，这些推理被编码为记忆，以通过将内部表征与类人推理模式对齐来指导模型训练。我们的最终模型在推理时不需要任何MLLM的开销。实验结果表明，RGL在IEMOCAP和MELD基准测试中达到了最先进的性能。此外，为了进行解释，我们展示了模型的内部特征有效地检索出未见测试样本的语义正确推理，验证了其推理能力。
cs.AI / 36 / 2608.10464
Quantum Incremental Learning with Mixed State Prototypes
基于混合态原型的量子增量学习
Abstract
Incremental learning models are required to learn new classes sequentially without catastrophic forgetting, while operating under parameter and memory constraints. In the Noisy Intermediate-Scale Quantum (NISQ) era, although quantum neural networks offer advantages in feature mapping, hardware limitations restrict circuit width. Furthermore, traditional quantum classifiers are constrained by the number of orthogonal basis states, limiting their capacity to accommodate a continually growing number of categories. Thus, we introduce a novel quantum incremental learning framework based on trainable mixed-state prototypes. Its original design incorporates new classes by adding class prototypes rather than increasing the circuit width of the shared quantum backbone. The use of mixed-state prototypes is another key contribution, since they have representation capabilities to represent information than a single pure-state prototype. And the decomposable mixed-state calculation provides lower production costs and a convenient Hilbert-Schmidt (HS) distance metric for classification. Simulation results show that our model achieves high-dimensional feature concentration using a minimal number of qubits, while demonstrating lower computational complexity and robust representation in incremental learning tasks compared with classical baselines.
Chinese Translation
增量学习模型需要在参数和内存限制下，顺序学习新类别而不发生灾难性遗忘。在噪声中间规模量子（NISQ）时代，尽管量子神经网络在特征映射上具有优势，但硬件限制限制了电路宽度。此外，传统量子分类器受到正交基态数量的限制，限制了其容纳不断增长的类别数量的能力。因此，我们提出了一种基于可训练混合态原型的新型量子增量学习框架。其原始设计通过添加类别原型而不是增加共享量子骨干网的电路宽度来引入新类别。混合态原型的使用是另一个关键贡献，因为它们比单一纯态原型具有更强的信息表示能力。而可分解的混合态计算提供了更低的生产成本和方便的希尔伯特-施密特（HS）距离度量用于分类。仿真结果表明，我们的模型在使用最少数量的量子比特的情况下实现了高维特征集中，同时在增量学习任务中相比于经典基线展示了更低的计算复杂性和更强的表示能力。
cs.AI / 37 / 2608.10471
RLMOpt: Adaptive Prompt Optimization via Recursive Language Models
RLMOpt：通过递归语言模型进行自适应提示优化
Abstract
Prompt optimizers automate the search for prompts that improve language-model performance, but existing methods rely on a predefined optimization procedure: the algorithm determines which candidates to explore and how the search progresses, while the language model generates or refines prompt proposals. We introduce RLMOpt, a prompt optimizer that makes the search policy itself language-model-driven through a recursive language model (RLM). The RLM agent operates over a tool-based environment, inspecting task information, analyzing failures, generating candidates, allocating evaluation budget, and deciding when to stop. A deterministic harness complements the agent by enforcing objective scoring, Pareto-based selection, and regression constraints. We evaluate RLMOpt across four benchmarks spanning structured clinical information extraction (Chia), multi-hop question answering (HotpotQA), verifiable instruction following (IFBench-2025), and multi-turn tool-calling agents (BFCL). In a matched comparison at a single seed, RLMOpt obtains the best held-out score on all four benchmarks and leads the four-task mean (0.610 against 0.589 for GEPA). Repeating each benchmark across seeds yields 11 matched benchmark-seed comparisons, in which RLMOpt outperforms GEPA in 9 cases. Across all 11 runs, it never produced a prompt that underperformed its seed, whereas GEPA fell below its starting point twice. It is also more efficient, achieving these results with fewer search rollouts while producing prompts that are 27-79% the size of those produced by GEPA. Our results further show that optimization gains are determined primarily by the headroom available in the seed prompt, rather than by the search budget. Efficient optimization therefore depends on reaching the available headroom reliably and with minimal search
Chinese Translation
提示优化器自动搜索能够提升语言模型性能的提示，但现有方法依赖于预定义的优化过程：算法决定探索哪些候选项以及搜索的进展方式，而语言模型则生成或完善提示提案。我们提出了RLMOpt，一种通过递归语言模型（RLM）使搜索策略本身由语言模型驱动的提示优化器。RLM代理在基于工具的环境中操作，检查任务信息、分析失败、生成候选项、分配评估预算并决定何时停止。一个确定性的工具补充了代理，强制执行客观评分、基于帕累托的选择和回归约束。我们在四个基准测试中评估RLMOpt，这些基准涵盖了结构化临床信息提取（Chia）、多跳问答（HotpotQA）、可验证的指令跟随（IFBench-2025）和多轮工具调用代理（BFCL）。在单个种子下的匹配比较中，RLMOpt在所有四个基准上获得了最佳的保留分数，并在四项任务的平均分中领先（0.610对比GEPA的0.589）。在跨种子重复每个基准测试时，产生了11个匹配的基准-种子比较，其中RLMOpt在9个案例中超越了GEPA。在所有11次运行中，它从未生成过低于其种子的提示，而GEPA则有两次低于其起始点。RLMOpt的效率也更高，以更少的搜索回合实现这些结果，同时生成的提示大小为GEPA的27-79%。我们的结果进一步表明，优化收益主要取决于种子提示中可用的余地，而不是搜索预算。因此，高效优化依赖于可靠地达到可用余地，并且搜索最小化。
cs.AI / 38 / 2608.10475
Evaluating Rational Contracting in Natural Language
评估自然语言中的理性契约
Abstract
The emergence of language-based AI agents promises to transform the scope of machine economic activity. Instead of just proposing bids or following hard-coded protocols, such agents can be used to negotiate and execute agreements in open-ended natural language. However, most evaluations of these abilities have focused on one-off exchanges or simple economic games, leaving open the rich space of time-extended, contingent, and incomplete contracts made expressible by language; they also focus on raw profit, without measuring the qualities required for trustworthy contracting. We address this by formulating a rational framework for how agents should negotiate and perform natural language contracts in uncertain multi-step environments. Within this framework, we develop metrics and baselines for quantifying rational and cooperative play. To evaluate how agents perform at such contracting, we instantiate our framework in ContractSim, an evaluation suite where two players negotiate and execute a multi-turn supplier contract under environmental and inter-player uncertainty. Across six environments and three supplier settings (catering, hotel cleaning, and AI hosting) we find that current LLM-based agents reach agreement reliably, and negotiate efficient contracts when environmental uncertainty is low. However, under high uncertainty, they often fail to negotiate satisfiable, efficient, or mutually beneficial contracts. They are also frequently uncooperative when executing contracts, violating contract terms for additional profit even when contracts are easy to satisfy. These findings highlight room for improvement in the design of language agents that can negotiate, interpret, and execute contracts both rationally and cooperatively.
Chinese Translation
基于语言的人工智能代理的出现承诺将改变机器经济活动的范围。这些代理不仅可以提出投标或遵循硬编码的协议，还可以用于用开放式自然语言进行谈判和执行协议。然而，大多数对这些能力的评估集中在一次性交换或简单经济游戏上，未能探索语言所能表达的丰富的时间延续性、条件性和不完整契约空间；它们还关注原始利润，而没有衡量可信契约所需的品质。我们通过制定一个理性框架来解决这一问题，该框架描述了代理在不确定的多步骤环境中应如何谈判和执行自然语言契约。在这个框架内，我们开发了量化理性和合作游戏的指标和基准。为了评估代理在此类契约中的表现，我们在ContractSim中实例化我们的框架，这是一个评估套件，其中两个玩家在环境和玩家间的不确定性下谈判和执行多轮供应商契约。在六个环境和三个供应商设置（餐饮、酒店清洁和人工智能托管）中，我们发现当前基于大型语言模型（LLM）的代理能够可靠地达成协议，并在环境不确定性低时谈判出有效的契约。然而，在高不确定性下，它们往往无法谈判出令人满意、有效或互利的契约。在执行契约时，它们也经常表现出不合作的行为，即使契约容易满足，它们也会违反契约条款以获取额外利润。这些发现突显了在设计能够理性和合作地谈判、解释和执行契约的语言代理方面的改进空间。
cs.AI / 39 / 2608.10480
Multi-Granular Rationale-Guided Molecular LLM for Property Prediction
多粒度理由引导的分子大语言模型用于性质预测
Abstract
Large language models (LLMs) are widely applied across chemical tasks, such as molecular property prediction, which underpins drug discovery. Molecular LLMs represent a molecule through several modalities, notably a 1D SMILES sequence or a 2D molecular graph. Both encode molecular information implicitly, so the contribution of individual substructures remains opaque. Retrieval and augmentation methods add context, but from external sources. However, the cues chemists reason over are the internal substructures that drive a property up or down. We propose MR-MoL, a multi-granular rationale-guided molecular LLM that supplies this evidence directly. A fine-tuned GNN scores each substructure through masking, and the most influential ones are serialized as a ranked, direction-tagged rationale that the LLM reads alongside the SMILES sequence and molecular graph. The rationale spans three levels of granularity: Murcko scaffolds with their side chains, BRICS fragments, and functional groups. This is, to our knowledge, the first method to expose GNN-derived attributions to an LLM as evidence for property prediction. On eight MoleculeNet tasks, MR-MoL achieves the best overall results among generalist models and narrows the gap to specialist models tuned for each task. Five diagnostics further confirm that the model reads the rationale rather than merely benefiting from its presence. Its direction, rank, and substructure each shape the prediction, and its attributions reproduce known structure-property relationships.
Chinese Translation
大型语言模型（LLMs）广泛应用于化学任务，例如分子性质预测，这为药物发现提供了基础。分子LLMs通过多种模态表示分子，特别是1D SMILES序列或2D分子图。这两者都隐式编码了分子信息，因此单个子结构的贡献仍然不清晰。检索和增强方法提供了上下文，但来自外部来源。然而，化学家推理的线索是驱动性质升高或降低的内部子结构。我们提出了MR-MoL，一种多粒度理由引导的分子LLM，直接提供这些证据。经过微调的图神经网络（GNN）通过掩蔽对每个子结构进行评分，最具影响力的子结构被序列化为一个带有排名和方向标记的理由，LLM在读取SMILES序列和分子图的同时也读取该理由。该理由涵盖三个粒度层次：Murcko骨架及其侧链、BRICS片段和功能团。据我们所知，这是首个将GNN派生的归因作为性质预测证据暴露给LLM的方法。在八个MoleculeNet任务中，MR-MoL在通用模型中取得了最佳整体结果，并缩小了与针对每个任务调优的专业模型之间的差距。五项诊断进一步确认该模型读取理由，而不仅仅是从其存在中受益。其方向、排名和子结构各自影响预测，其归因重现了已知的结构-性质关系。
cs.AI / 40 / 2608.10483
Predicting Space Groups of Double Perovskites by LLM with Dynamic Few-Shot Learning
通过动态少样本学习预测双钙钛矿的空间群
Abstract
Double perovskites (DPs) offer broad compositional tunability, but predicting the space groups (SGs) of stable structures remains difficult because available datasets are often strongly imbalanced toward dominant SG classes. We refer to dominant SG classes as major SGs and underrepresented classes as minor SGs. We introduce Dynamic and Diversity-enhanced Few-shot Retrieval and Rule-Guided Inference for Space-Group Prediction (DyRIS), an LLM-agent-based framework that predicts ranked SG candidates from a given DP composition. DyRIS uses diversity-enhanced dynamic few-shot prompting to retrieve relevant in-context examples while limiting the dominance of frequently represented SGs. It further incorporates rule-guided inference based on B/B' cation ordering, quantitative indicators, and major-SG bias control to refine and rank the final Top-3 SG candidates. We evaluate DyRIS on 3,528 thermodynamically filtered DP entries and compare it with composition-based and descriptor-based baselines. At a training-data ratio of 0.5, DyRIS achieves competitive overall accuracy while obtaining the best Overall Top-1 macro-F1 score and the best performance across all Minor-SG metrics. DyRIS improves Minor-SG Top-1 accuracy by 3.26 percentage points relative to CrabNet and achieves higher Minor-SG Top-3 accuracy than the strongest PyCaret-based baseline. Ablation studies show that diversity-enhanced retrieval, quantitative indicators, major-SG bias control, and B/B' ordering information each contribute to prediction performance. Additional experiments show that the final rule-guided inference step is not easily replaced by conventional classifier- or ranker-based models. These findings demonstrate the potential of combining retrieval-based LLM reasoning with crystallographic domain knowledge for SG prediction in imbalanced materials datasets.
Chinese Translation
双钙钛矿（DPs）具有广泛的成分可调性，但由于可用数据集往往在主要空间群（SG）类别上严重不平衡，因此预测稳定结构的空间群仍然困难。我们将主要空间群类别称为主要SG，将代表性不足的类别称为次要SG。我们提出了一种动态和多样性增强的少样本检索与规则引导推理框架，用于空间群预测（DyRIS），该框架基于大型语言模型（LLM）代理，从给定的DP成分中预测排名的SG候选。DyRIS使用多样性增强的动态少样本提示来检索相关的上下文示例，同时限制频繁出现的SG的主导地位。它进一步结合了基于B/B'阳离子排序、定量指标和主要SG偏差控制的规则引导推理，以细化和排名最终的前3个SG候选。我们在3,528个热力学过滤的DP条目上评估DyRIS，并将其与基于成分和描述符的基线进行比较。在训练数据比例为0.5时，DyRIS实现了具有竞争力的整体准确性，同时获得了最佳的整体Top-1宏F1分数以及在所有次要SG指标上的最佳表现。相较于CrabNet，DyRIS提高了次要SG的Top-1准确率3.26个百分点，并在次要SG的Top-3准确率上超过了最强的基于PyCaret的基线。消融研究表明，多样性增强检索、定量指标、主要SG偏差控制和B/B'排序信息均对预测性能有所贡献。额外实验表明，最终的规则引导推理步骤不易被传统的分类器或排序模型替代。这些发现展示了将基于检索的LLM推理与晶体学领域知识相结合在不平衡材料数据集中进行SG预测的潜力。
cs.AI / 41 / 2608.10492
INSIDE the Student's Mind: Jointly Modeling Latent Reasoning and Action in LLM Student Simulators
学生思维内部探究：在大型语言模型学生模拟器中联合建模潜在推理与行动
Abstract
Large Language Model (LLM)-based simulators often reproduce observable actions but fail to capture the underlying reasoning behind them. In education, where student simulation is increasingly used for various applications such as evaluating tutoring systems, this gap is especially pronounced. Two students may submit identical submissions for entirely different reasons. We present INTERNAL STUDENT DIALOGUE (INSIDE), a student modeling framework that fine-tunes LLMs not only to act like students but also to think like them. INSIDE generates internal dialogue grounded in Bloom's Taxonomy across cognitive, affective, and action dimensions, and fine-tunes models on paired think traces and actions. We baseline against different prompting frameworks and evaluate on two axes: fidelity of simulated actions and quality of generated internal dialogue. Our evaluations show that INSIDE improves simulation fidelity in both action fidelity, matching code generation of real students, and reasoning alignment, achieving the highest alignment across models up to 57.9%.
Chinese Translation
基于大型语言模型（LLM）的模拟器通常能够再现可观察的行为，但未能捕捉其背后的潜在推理。在教育领域，学生模拟越来越多地用于评估辅导系统等各种应用，这一差距尤为明显。两个学生可能因完全不同的原因提交相同的作业。我们提出了内部学生对话（INTERNAL STUDENT DIALOGUE，INSIDE），这是一个学生建模框架，旨在微调LLM，使其不仅能够像学生一样行动，还能像学生一样思考。INSIDE生成基于布鲁姆分类法（Bloom's Taxonomy）的内部对话，涵盖认知、情感和行动维度，并在配对的思维轨迹和行动上进行微调。我们与不同的提示框架进行基准比较，并在两个维度上进行评估：模拟行为的真实性和生成内部对话的质量。我们的评估结果表明，INSIDE在行动真实性（与真实学生的代码生成相匹配）和推理一致性方面提高了模拟的真实性，在所有模型中实现了最高的推理一致性，达到57.9%。
cs.AI / 42 / 2608.10494
GeoForge: Non-Parametric Self-Evolving Agents for Earth-Observation Reasoning
GeoForge：用于地球观测推理的非参数自我演化智能体
Abstract
Earth observation (EO) agents construct scientifically valid tool workflows and ground their conclusions in current geospatial evidence. This is challenging because EO workflows are constrained by sensing semantics, product dependencies, spatial and temporal compatibility, and parameter requirements. Existing agents often search a broad operation space for each query, while recent self-evolving systems do not fully organize heterogeneous EO trajectories into reusable knowledge across different decision levels. To solve this problem, we present GeoForge, a training-free, self-evolving framework that transforms completed trajectories into a structured nonparametric execution state. GeoForge constrains the operation space according to the sensing context, then retrieves a task-conditioned prior from three complementary memories. Workflow Graph Memory captures global operation order, Action-Level Experiences provide local corrections, and the Adapted Skill Standard Operating Procedure preserves procedural and data constraints. The retrieved prior guides tool execution, while current observations remain the basis of the final answer. After each task, a safety-gated distillation process converts grounded trajectories into reusable execution knowledge for future retrieval. This execution, distillation, and reuse loop improves planning without updating the backbone LLM. Experiments on multiple geospatial benchmarks demonstrate that GeoForge consistently improves both task accuracy and tool-use trajectory quality across diverse LLM backbones, while substantially reducing tool-planning and reasoning errors for most LLMs.
Chinese Translation
地球观测（EO）智能体构建科学有效的工具工作流程，并将其结论基于当前的地理空间证据。这一过程具有挑战性，因为EO工作流程受到感知语义、产品依赖性、时空兼容性和参数要求的限制。现有的智能体通常在每个查询中搜索广泛的操作空间，而最近的自我演化系统并未充分将异构的EO轨迹组织成可在不同决策层次上重用的知识。为了解决这一问题，我们提出了GeoForge，一个无需训练的自我演化框架，能够将完成的轨迹转化为结构化的非参数执行状态。GeoForge根据感知上下文限制操作空间，然后从三个互补的记忆中检索任务条件先验。工作流程图记忆捕获全局操作顺序，行动级经验提供局部修正，而适应性技能标准操作程序则保留程序和数据约束。检索到的先验指导工具执行，而当前观察仍然是最终答案的基础。在每个任务之后，一个安全门控的提炼过程将基础轨迹转化为可重用的执行知识，以便于未来的检索。这个执行、提炼和重用的循环在不更新主干大型语言模型（LLM）的情况下改善了规划。在多个地理空间基准上的实验表明，GeoForge在不同的LLM主干上始终提高了任务准确性和工具使用轨迹质量，同时显著减少了大多数LLM的工具规划和推理错误。
cs.AI / 43 / 2608.10502
From Faulty Memories to Corrected Actions: Dependency-Guided Rollback Repair for Memory-Augmented Agents
从错误记忆到纠正行动：基于依赖的回滚修复用于增强记忆的智能体
Abstract
Persistent memory lets language-model agents reuse information across sessions, but it also makes errors durable: a poisoned, stale, or misattributed record can alter reasoning, tool use, answers, and subsequent memory writes. Existing defenses mainly detect or delete suspicious memories, or revise the current response. Deleting the source leaves already propagated claims, actions, and derived memories active, whereas resetting the store or replaying the full trace destroys benign state and repeats unnecessary computation. We therefore formulate \textbf{post-failure memory recovery: } \textit{given a failed execution and diagnosed faulty memories, recover both the answer and persistent state while retaining unaffected work.} Our \textbf{dependency-guided rollback repair} builds a typed memory-to-action graph from runtime provenance, traces explicit downstream dependencies, preserves candidates with independent trusted support, deactivates unsupported memory state, and selectively replays only answer-relevant affected computation. We evaluate this approach on a 150-case controlled benchmark spanning three tool-use domains and four memory failure types, and on a 50-case trajectory-derived stress test adapted from LongMemEval-V2. On the controlled benchmark, it achieves 85.3\% recovery versus 77.3\% for the best competing recovery method, removes all diagnosed faulty memories, preserves all benign memories, and requires only selective replay with modest LLM-call cost. On the adapted subset, it reaches 68.0\% recovery versus 54.0\% for the next best method, while also achieving the highest claim invalidation F1, 0.669 versus 0.603. Overall, the results do not imply uniformly better trace reconstruction, but show that dependency-guided rollback repair provides a strong recovery--cost trade-off while repairing faulty memory state and preserving benign memory.
Chinese Translation
持久性记忆使语言模型智能体能够在会话之间重用信息，但它也使错误变得持久：被污染、过时或错误归属的记录可能会改变推理、工具使用、答案以及后续的记忆写入。现有的防御措施主要是检测或删除可疑记忆，或修正当前响应。删除源头会使已经传播的主张、行动和派生记忆保持活跃，而重置存储或重放完整的执行轨迹则会破坏良性状态并重复不必要的计算。因此，我们提出了 extbf{故障后记忆恢复：} extit{在给定失败执行和诊断出故障记忆的情况下，恢复答案和持久状态，同时保留未受影响的工作。}我们的 extbf{基于依赖的回滚修复}从运行时来源构建了一个类型化的记忆到行动图，追踪明确的下游依赖关系，保留具有独立可信支持的候选项，停用不支持的记忆状态，并有选择性地重放仅与答案相关的受影响计算。我们在一个涵盖三个工具使用领域和四种记忆故障类型的150案例受控基准上评估了该方法，并在一个改编自LongMemEval-V2的50案例轨迹派生压力测试中进行评估。在受控基准上，该方法实现了85.3\%的恢复率，而最佳竞争恢复方法为77.3\\%。它删除了所有诊断出的故障记忆，保留了所有良性记忆，并且仅需选择性重放，且LLM调用成本适中。在改编的子集上，它达到了68.0\\%的恢复率，而下一个最佳方法为54.0\\%，同时在主张无效化F1上也达到了最高值0.669，相较于0.603。总体而言，结果并不意味着统一更好的轨迹重建，但显示出基于依赖的回滚修复在修复故障记忆状态和保留良性记忆的同时，提供了良好的恢复与成本的权衡。
cs.AI / 44 / 2608.10504
MEGA: Self-Evolving Agent Optimization Infrastructure via Wisdom Graph
MEGA：通过智慧图的自我进化代理优化基础设施
Abstract
As coding agents increasingly handle implementation, the central challenge shifts from building individual agents to building an infrastructure that systematically improves them. Current approaches optimize agent systems without accumulating transferable knowledge, accumulate knowledge without compositional reasoning over it, and lack a mechanism for that knowledge to self-evolve through operational evidence. MEGA (Meta Evaluation-Grounded Adaptation) addresses these gaps as a self-evolving infrastructure: each optimization cycle produces durable assets, compositional reasoning over those assets guides subsequent optimization, and operational evidence refines both the accumulated wisdom and the reasoning that governs it. Layer 1 distills reusable wisdom from agent sessions through behavioral-pattern clustering and empirical A/B validation, transforming each process into a durable asset. Layer 2 decomposes these assets into atomic PCR (Primary-Context-Resultant) units within a typed Wisdom Graph and performs deductive, abductive, and inductive reasoning to expand implicit relations; it then assembles context-specific execution plans through compositional retrieval that surfaces bridging knowledge unreachable by embedding similarity alone. Layer 3 performs multi-agent collaborative optimization over heterogeneous agent workflows (code nodes, LLM calls, and tool-using agents), attributing improvement effects to specific strategy changes through controlled evaluation that eliminates data variance. Evidence fed back from Layer 3 drives the self-evolution of both the curation strategies that govern wisdom composition and the optimization trajectories accumulated across runs. The result is an infrastructure in which optimizing an agent system and evolving the knowledge that guides optimization are one and the same process.
Chinese Translation
随着编码代理越来越多地处理实现，中心挑战从构建单个代理转向构建一个系统性改进它们的基础设施。目前的方法在优化代理系统时未能积累可转移的知识，积累知识时未能进行组合推理，并且缺乏通过操作证据自我进化该知识的机制。MEGA（元评估基础适应）作为一种自我进化的基础设施，解决了这些问题：每个优化周期产生持久资产，对这些资产的组合推理指导后续优化，而操作证据则细化了积累的智慧和支配它的推理。第一层通过行为模式聚类和经验A/B验证从代理会话中提炼可重用的智慧，将每个过程转化为持久资产。第二层将这些资产分解为类型化智慧图中的原子PCR（主要上下文结果）单元，并进行演绎、溯因和归纳推理，以扩展隐含关系；然后通过组合检索组装上下文特定的执行计划，揭示仅通过嵌入相似性无法达到的桥接知识。第三层在异构代理工作流（代码节点、LLM调用和工具使用代理）上执行多代理协作优化，通过控制评估消除数据方差，将改进效果归因于特定的策略变化。从第三层反馈的证据驱动了支配智慧组合的策划策略和跨运行积累的优化轨迹的自我进化。最终结果是一个基础设施，其中优化代理系统和进化指导优化的知识是同一过程。
cs.AI / 45 / 2608.10505
RadFusion: Towards Threshold-Controllable Radiology Report Generation
RadFusion：面向可控阈值的放射学报告生成
Abstract
Automated radiology report generation is advancing rapidly in response to the shortage of radiologists, yet unlike a perception model, existing generation models offer no control over the sensitivity-specificity trade-off of their diagnostic content. Such control is essential because clinical scenarios diverge: emergency triage prioritizes sensitivity to reduce missed findings, whereas confirmatory interpretation emphasizes specificity to limit unnecessary interventions. A single fixed report can neither adapt to these scenarios nor support the ROC-based validation widely expected for regulatory clearance. We introduce RadFusion, a framework that equips report generation with threshold controllability. Our method fuses a multi-label classifier, which provides per-disease confidence scores, with a VQA-based report generator, which describes medical findings in detail; an LLM then rewrites the report so that its stated diagnoses follow the classifier's decisions at the selected threshold while staying grounded in the generator's descriptions. On MIMIC-CXR, the performance of RadFusion conforms to the classifier's ROC curve: sweeping the threshold and mapping the reports back to class labels reproduces the classifier's validated ROC performance. This conformance makes generated reports quantitatively evaluable through ROC analysis, strengthening the case for regulatory clearance, and enables operating-point selection that matches report behavior to clinical context. Moreover, combining the two model types improves diagnostic accuracy over uncontrolled generation: sensitivity increases by 6.9% at matched specificity, and specificity by 20.7% at matched sensitivity. These results show that RadFusion makes report generation clinically adaptable, quantitatively verifiable, and diagnostically more reliable.
Chinese Translation
自动化放射学报告生成正在迅速发展，以应对放射科医生短缺的问题。然而，与感知模型不同，现有的生成模型无法对其诊断内容的敏感性-特异性权衡进行控制。这种控制至关重要，因为临床场景各不相同：急诊分诊优先考虑敏感性以减少漏诊，而确认性解读则强调特异性以限制不必要的干预。单一固定的报告无法适应这些场景，也无法支持监管审批中广泛期望的基于ROC的验证。我们提出了RadFusion，一个为报告生成提供阈值可控性的框架。我们的方法将多标签分类器（提供每种疾病的置信度评分）与基于视觉问答（VQA）的报告生成器相结合，后者详细描述医学发现；然后，一个大型语言模型（LLM）重写报告，使其所述诊断遵循分类器在所选阈值下的决策，同时保持与生成器描述的一致性。在MIMIC-CXR数据集上，RadFusion的性能符合分类器的ROC曲线：通过调整阈值并将报告映射回类别标签，重现了分类器的验证ROC性能。这种一致性使生成的报告能够通过ROC分析进行定量评估，增强了监管审批的合理性，并能够选择与临床背景相匹配的操作点。此外，结合这两种模型类型提高了诊断准确性：在匹配特异性的情况下，敏感性提高了6.9%；在匹配敏感性的情况下，特异性提高了20.7%。这些结果表明，RadFusion使报告生成在临床上更具适应性、定量可验证性和诊断可靠性。
cs.AI / 46 / 2608.10509
MAP-Graph: Provenance-Aware Shared Memory for Multi-Agent Workflows
MAP-Graph：面向来源的多智能体工作流共享内存
Abstract
Shared memory helps language-model agents reuse information across long workflows, yet relevant evidence may not be admissible for a particular agent or action. Because restrictions propagate through derivations, summaries can conceal private, poisoned, untrusted, or revoked sources, enabling unauthorized reads or unsafe actions. Existing approaches provide semantic retrieval, scoped access, or lineage tracking, but do not clearly separate hard authorization from graded trust or adapt evidence requirements to action risk. We introduce MAP-Graph, a provenance-aware memory layer that represents agents, sources, memories, claims, and actions in a typed execution graph. It traces ancestry, excludes permission-ineligible records, reranks eligible memories by semantic similarity and multiplicative path trust, and applies a risk-sensitive gate before action execution while retaining affected lineage for audit. On a controlled benchmark of 2,700 synthetic tasks per method across three domains, MAP-Graph achieves 94.96\% overall task success, 72.70\% exact decision accuracy, and 90.22\% in the clean setting, where success requires a correct \textsc{Allow} rather than a safe intervention. Ablations isolate the roles of permission filtering, path trust, and action gating, while transfer tests with two additional backbones preserve the exact-decision and access-control advantages. These results support provenance as an operational control signal, rather than only post-hoc audit metadata, within the evaluated setting.
Chinese Translation
共享内存帮助语言模型智能体在长流程中重用信息，但相关证据可能不适用于特定的智能体或行动。由于限制在推导中传播，摘要可能会隐瞒私人、受污染、不可信或被撤销的来源，从而导致未经授权的读取或不安全的行动。现有的方法提供语义检索、范围访问或血统追踪，但未能清晰地区分严格授权与渐进信任，或根据行动风险调整证据要求。我们提出了MAP-Graph，这是一种面向来源的内存层，表示智能体、来源、记忆、声明和行动的类型化执行图。它追踪祖先，排除不符合权限的记录，通过语义相似性和乘法路径信任对符合条件的记忆进行重新排序，并在行动执行前应用风险敏感门控，同时保留受影响的血统以供审计。在一个控制的基准测试中，针对三个领域的每种方法进行了2700个合成任务，MAP-Graph实现了94.96%的整体任务成功率、72.70%的精确决策准确率，以及在清洁设置下90.22%的成功率，其中成功要求正确的 extsc{Allow}而非安全干预。消融实验隔离了权限过滤、路径信任和行动门控的作用，而与两个额外骨干的迁移测试则保留了精确决策和访问控制的优势。这些结果支持在评估设置中将来源作为操作控制信号，而不仅仅是事后审计元数据。
cs.AI / 47 / 2608.10537
Measuring Semantic Abstractness of SAE Features via Nonlocality
通过非局部性测量稀疏自编码器特征的语义抽象性
Abstract
Sparse autoencoders (SAEs) have helped uncover mechanistic explanations for LLM behaviours such as reasoning, jailbreaking etc., via understanding the corresponding task-relevant and causally effective features. To evaluate such mechanistic explanations, downstream studies must distinguish surface lexical features from genuinely high-level ones. However, neither an autointerp-based semantic description nor causal steering utility fully resolves the abstraction level of a feature. To this end, we introduce \emph{Feature Nonlocality} (FNL), defined as the entropy of the normalized per-position influence on an SAE feature's activation. We report that FNL correlates with existing LLM-based proxy metrics of feature semantic abstractness, and successfully distinguishes context-dependent reasoning features from token-driven ones, correctly assigning the higher FNL to the contextual feature in $73$--$84\%$ of randomly drawn pairs that consist of one contextual and one token-level feature. We demonstrate two downstream applications. We audit SAE-based features used for jailbreak mitigation and find surprisingly that most effective features are positional features with low FNL rather than genuinely recognizing harmful intents. We report that steering high-FNL features in DeepSeek-R1-Distill-Llama-8B improves MATH-500 accuracy by $4.6$ points over the unsteered model and outperforms steering low-FNL features, though the gains are model-specific. We conclude that FNL provides an LLM-independent, label-free, correlational witness of the abstraction level of an SAE feature, with applications in evaluating mechanistic explanations as well as selecting features for downstream interventions.
Chinese Translation
稀疏自编码器（SAEs）通过理解相应的任务相关和因果有效特征，帮助揭示大型语言模型（LLM）行为（如推理、越狱等）的机制解释。为了评估这些机制解释，下游研究必须区分表面词汇特征与真正的高层次特征。然而，基于自解释的语义描述或因果引导效用都无法完全解决特征的抽象水平。为此，我们引入了 extit{特征非局部性}（Feature Nonlocality, FNL），定义为对SAE特征激活的归一化每位置影响的熵。我们报告FNL与现有基于LLM的特征语义抽象性代理指标相关，并成功区分了上下文依赖的推理特征与基于标记的特征，在73%至84%的随机抽取的包含一个上下文特征和一个标记级特征的对中，正确地将更高的FNL分配给上下文特征。我们展示了两个下游应用。我们审计了用于越狱缓解的SAE特征，发现令人惊讶的是，大多数有效特征是具有低FNL的位置信息特征，而非真正识别有害意图的特征。我们报告在DeepSeek-R1-Distill-Llama-8B中引导高FNL特征使MATH-500的准确率比未引导模型提高了4.6个百分点，并且优于引导低FNL特征，尽管增益是特定于模型的。我们得出结论，FNL提供了一种与LLM无关、无标签的、相关性证据，能够反映SAE特征的抽象水平，具有评估机制解释和选择下游干预特征的应用。
cs.AI / 48 / 2608.10538
SKILLER: Language-Level Reinforcement Learning for Reusable Skill Extraction in Small Language Models
SKILLER：用于小型语言模型的可重用技能提取的语言级强化学习
Abstract
Agent skills represent a standardized format for packaging procedural knowledge and domain expertise, serving within agent harness systems as an essential mechanism to continually constrain a language model's behavior space for repeatable, high-quality task execution. However, because strong closed-source models entail high inference costs, current popular agent harnesses, such as Codex and OpenClaw, remain prohibitively expensive when deploying these skills to accomplish real-world tasks. The rapid capability enhancement of open-source models deployable on consumer-grade GPUs presents a compelling opportunity to drastically reduce these costs by leveraging skill-based behavioral constraints. Nevertheless, automatically generating effective skills tailored specifically for such compact models remains a significant practical challenge. To address this, we propose SKILLER, a natural-language-driven reinforcement learning framework designed to automatically generate executor-specific skills for small models, which employs a strong model as the actor and critic, treats the small-model agent system as the environment, and propagates all reinforcement learning signals entirely via natural language. Extensive experimental evaluations across five relevant benchmarks using Qwen3.5-9B and Qwen3.5-4B demonstrate that SKILLER outperforms three open-source and one closed-source skill generation or evolution methods, achieving absolute gains ranging from 4.3 to 20.4 percentage points for the 9B model and 1.8 to 13.3 points for the 4B model, while remarkably matching the performance of strong closed-source models on single-skill tasks in SkillsBench. The project is available at https://github.com/DANG-ai/SKILLER.
Chinese Translation
代理技能代表了一种标准化格式，用于打包程序知识和领域专长，作为代理系统中的一种基本机制，持续约束语言模型的行为空间，以实现可重复的高质量任务执行。然而，由于强大的闭源模型涉及高昂的推理成本，当前流行的代理系统，如Codex和OpenClaw，在将这些技能部署到实际任务中时仍然显得过于昂贵。可在消费级GPU上部署的开源模型的快速能力提升，为通过利用基于技能的行为约束大幅降低这些成本提供了一个引人注目的机会。然而，自动生成专门针对这些紧凑模型的有效技能仍然是一个重要的实际挑战。为此，我们提出了SKILLER，一个以自然语言驱动的强化学习框架，旨在自动生成针对小型模型的执行器特定技能，该框架将强模型作为演员和评论者，将小型模型代理系统视为环境，并通过自然语言完全传播所有强化学习信号。通过在五个相关基准上的广泛实验评估，使用Qwen3.5-9B和Qwen3.5-4B，结果表明SKILLER在技能生成或演化方法中优于三种开源和一种闭源方法，9B模型的绝对增益范围为4.3到20.4个百分点，4B模型的增益范围为1.8到13.3个百分点，同时在SkillsBench的单技能任务中显著匹配强闭源模型的性能。该项目可在https://github.com/DANG-ai/SKILLER获取。
cs.AI / 49 / 2608.10549
Reinforcement Learning-Based Laser Cutting Machine Parameter Optimization
基于强化学习的激光切割机参数优化
Abstract
Achieving high accuracy in laser-based cutting of optical films requires careful tuning of parameters such as focal length and laser power beam, adjusted according to the specific properties of each film type. Trial-and-error based traditional methods are used to find the most suitable cutting parameters for various films, but they are slow and inaccurate. To address this issue, this paper presents the Reinforcement Learning for Laser Cutting (RL$^{2}$C) algorithm, which uses Q-learning with an epsilon-greedy policy to dynamically optimize cutting parameters, significantly reducing taper size and film wastage. Additionally, RL$^{2}$C incorporates a dynamic environment space adaptability mechanism to allow it to adapt to new states encountered during the learning process over multiple batches of experiments. Experimental results demonstrate that RL$^{2}$C requires fewer steps and less time to find optimal cutting parameters compared to various RL-based optimization methods. Specifically, RL$^{2}$C reduces the number of optimization steps by up to 12.5\% and processing time by up to 81.8\% compared to existing methods. This study demonstrates the potential of RL in industrial laser-cutting processes by improving cut quality, reducing time and film wastage, and minimizing manual interventions.
Chinese Translation
在激光切割光学薄膜时，实现高精度需要仔细调整焦距和激光功率等参数，这些参数需根据每种薄膜类型的特性进行调整。传统的试错方法用于寻找各种薄膜的最适切割参数，但这些方法速度慢且不够准确。为了解决这一问题，本文提出了激光切割强化学习算法（Reinforcement Learning for Laser Cutting，RL$^{2}$C），该算法采用带有ε-贪婪策略的Q学习动态优化切割参数，显著减少锥度尺寸和薄膜浪费。此外，RL$^{2}$C还结合了动态环境空间适应机制，使其能够在多个实验批次的学习过程中适应遇到的新状态。实验结果表明，与各种基于强化学习的优化方法相比，RL$^{2}$C在寻找最优切割参数时所需的步骤和时间更少。具体而言，RL$^{2}$C将优化步骤的数量减少了多达12.5 ext{%}，处理时间减少了多达81.8 ext{%}。本研究展示了强化学习在工业激光切割过程中的潜力，通过提高切割质量、减少时间和薄膜浪费以及最小化人工干预。
cs.AI / 50 / 2608.10567
DashArena: Benchmarking LLMs on Interactive Analytic Dashboard Generation
DashArena：对交互式分析仪表板生成的LLM基准测试
Abstract
Analytic dashboards combine coordinated views and interactions for data exploration and decision-making. Recent models can generate them from data and natural-language goals, but evaluating their usefulness remains difficult. Dashboard generation is open-ended, and neither static appearance nor successful execution alone captures analytical support and interaction quality. We introduce DashArena, to our knowledge the first benchmark for open-ended, task-grounded generation of interactive analytic dashboards. Its key innovation is to require each system to generate both a dashboard and a replayable interaction trajectory. A browser executor replays the trajectory and turns the system's intended analytical workflow into reproducible visual and execution evidence. A VLM judge compares candidates using this evidence, and Bradley--Terry aggregation produces the leaderboard. We further distill the judge into the open-weight DashJudge-8B. Human evaluations show that DashJudge-8B effectively reproduces human judgments and ablations show that interaction evidence improves judge agreement. Experiments with frontier models reveal persistent rendering, analytical, and interaction failures. Together, these results show that realistic dashboard generation remains challenging and that interaction-aware evaluation captures failures missed by static or execution-only checks.
Chinese Translation
分析仪表板结合了协调视图和交互，以便进行数据探索和决策。最近的模型可以根据数据和自然语言目标生成这些仪表板，但评估它们的实用性仍然困难。仪表板生成是开放式的，仅凭静态外观或成功执行无法全面捕捉分析支持和交互质量。我们引入了DashArena，作为我们所知的首个开放式、任务驱动的交互式分析仪表板生成基准。其关键创新在于要求每个系统生成一个仪表板和一个可重放的交互轨迹。浏览器执行器重放该轨迹，将系统预期的分析工作流程转化为可重复的视觉和执行证据。VLM评审员使用这些证据比较候选者，Bradley-Terry聚合生成排行榜。我们进一步将评审员提炼为开放权重的DashJudge-8B。人类评估表明，DashJudge-8B有效地再现了人类判断，消融实验显示交互证据提高了评审一致性。与前沿模型的实验揭示了持续的渲染、分析和交互失败。综合这些结果表明，现实的仪表板生成仍然具有挑战性，而关注交互的评估能够捕捉到静态或仅执行检查所遗漏的失败。
cs.AI / 51 / 2608.10579
Agentic Instruction Data Selection: Let DataMaster Interpret Your Intent
自主指令数据选择：让 DataMaster 解读您的意图
Abstract
Although existing instruction data selection methods have introduced various metrics, the inherent complexity of real-world datasets makes it impractical for any single metric to generalize across all scenarios. Developers are thus often forced to manually inspect data and craft heuristic rules for each new application---a tedious and error-prone process. In this paper, we propose a paradigm shift from manual configuration to automated orchestration via the Instruction Data Selection Agent (DataMaster), which interprets user intent and autonomously composes optimal selection strategies. By allowing users to specify data needs through natural language descriptions, DataMaster simplifies data curation and removes the burden of manual strategy design. Extensive experiments across the math, medical, and code domains show that DataMaster outperforms static baselines in most settings and surpasses full-pool training in a substantial number of cases. The implementation of DataMaster and the scripts needed to reproduce the reported pipeline are publicly available at https://github.com/nju-websoft/DataMaster.
Chinese Translation
尽管现有的指令数据选择方法引入了多种指标，但现实世界数据集的固有复杂性使得任何单一指标在所有场景中都难以泛化。因此，开发人员常常被迫手动检查数据并为每个新应用制定启发式规则，这是一项繁琐且容易出错的过程。本文提出了一种从手动配置到通过指令数据选择代理（DataMaster）实现自动编排的范式转变，DataMaster 能够解读用户意图并自主构建最佳选择策略。通过允许用户通过自然语言描述来指定数据需求，DataMaster 简化了数据整理过程，并消除了手动策略设计的负担。针对数学、医学和代码领域的广泛实验表明，DataMaster 在大多数设置中优于静态基线，并在相当数量的案例中超越了全池训练。DataMaster 的实现及重现报告管道所需的脚本已公开发布在 https://github.com/nju-websoft/DataMaster。
cs.AI / 52 / 2608.10584
HexEval: An Evidence-Driven Hexagonal Framework for Multidimensional Scholar Assessment
HexEval：一种基于证据的六边形多维学者评估框架
Abstract
Scholar assessment plays a fundamental role in faculty recruitment, funding allocation, academic promotion, and talent discovery. Existing scholar assessment methods predominantly rely on bibliometric indicators and reputation proxies, while recent large language model (LLM)-based approaches mainly focus on evaluating individual research papers rather than comprehensively assessing scholars. We argue that scholar assessment should be formulated as an evidence-driven reasoning problem that jointly considers intrinsic research quality and externally verifiable scholarly behavior. To this end, we propose HexEval, an evidence-driven hexagonal framework for multidimensional scholar assessment. HexEval explicitly organizes scholar assessment into two complementary evidence layers. The intrinsic layer evaluates anonymized representative works along three dimensions, namely research rigor, methodological innovation, and scientific contribution, whereas the external layer characterizes scholars through knowledge translation, research coherence, and academic impact using heterogeneous evidence collected from GitHub, Lens, OpenAlex, and other publicly verifiable sources. Instead of producing opaque aggregate scores, HexEval preserves intermediate evidence, dimension-specific rationales, and verification signals throughout the evaluation process, enabling interpretable and auditable scholar profiles. Experiments across all six dimensions show dimension-dependent agreement with human or external reference criteria: structured calibration improves absolute agreement for intrinsic quality, while the external modules recover broad trajectory and ordinal impact signals. These results support evidence-driven reasoning over heterogeneous scholarly evidence as a promising paradigm for auditable AI-assisted scholar assessment, while exposing the coverage and attribution limitations of public scholarly data.
Chinese Translation
学者评估在教职招聘、资金分配、学术晋升和人才发现中发挥着基础性作用。现有的学者评估方法主要依赖于文献计量指标和声誉代理，而最近基于大型语言模型（LLM）的方法则主要集中在评估单个研究论文，而非全面评估学者。我们认为，学者评估应被构建为一个基于证据的推理问题，综合考虑内在研究质量和外部可验证的学术行为。为此，我们提出了HexEval，一种基于证据的六边形多维学者评估框架。HexEval明确将学者评估组织为两个互补的证据层。内在层沿着研究严谨性、方法创新和科学贡献三个维度评估匿名代表性作品，而外部层则通过从GitHub、Lens、OpenAlex和其他公开可验证来源收集的异构证据，描述学者在知识转化、研究一致性和学术影响方面的表现。HexEval不仅仅生成不透明的汇总分数，而是保留了评估过程中的中间证据、特定维度的推理和验证信号，从而实现可解释和可审计的学者档案。在所有六个维度上的实验显示，与人类或外部参考标准的维度依赖性一致性：结构化校准提高了内在质量的绝对一致性，而外部模块则恢复了广泛的轨迹和序数影响信号。这些结果支持基于异构学术证据的证据驱动推理作为可审计的AI辅助学者评估的有前景的范式，同时揭示了公共学术数据的覆盖和归因限制。
cs.AI / 53 / 2608.10644
Curate Before You Connect: Identity and Ontology Tagging in a Production Knowledge Graph
连接之前的策划：生产知识图谱中的身份与本体标记
Abstract
Extraction produces candidate entities and relationships; writing them into a graph is where identity is decided, and identity decisions are destructive in a way extraction errors are not. A wrong type can be corrected later, but two records merged under one identity cannot be separated once their properties have been combined, and the merge leaves no error behind. This paper describes the ingestion and ontology-tagging layer that turns a validated extraction stream into a knowledge graph of 537,157 entities and 2,198,567 relationships drawn from 98,795 government documents. We describe a record-identity ladder that decides sameness from identifier columns, name columns, display names and type-scoped position rather than from name similarity. The ladder governs de-duplication within parsed tables, while the graph write applies a coarser canonical-name key, so records sharing a canonical name merge automatically on exact equality. We argue rather than demonstrate that this is where the automation line belongs: no identity benchmark is reported, and the over-merges the key permits are undetectable by construction. That policy, under which entity resolution only ever flags candidates, followed an incident in which two surface forms of one name were merged, corrupting a correct record and deleting eight entities from an unrelated document. We then describe multi-class ontology tagging and an evidence asymmetry we did not anticipate: an entity name is an instance label rather than a type assertion, so matching name fragments against a class index invents classifications. Requiring anchored evidence cut role assignments on an enriched sample from 36 to 4, all confirmed correct. We quantify the graph's conformance debt, show secondary classifications compensating for a mis-parented primary class, and describe a curation queue grown to 48,403 pending proposals against 775 human decisions.
Chinese Translation
提取过程产生候选实体和关系；将它们写入图谱是决定身份的地方，而身份决策在某种程度上是具有破坏性的，这与提取错误不同。错误的类型可以在后期纠正，但一旦将两个记录合并为一个身份，它们的属性被结合后就无法分开，并且合并不会留下任何错误。本文描述了一个摄取和本体标记层，将经过验证的提取流转化为一个包含537,157个实体和2,198,567个关系的知识图谱，这些数据来源于98,795份政府文件。我们描述了一个记录身份阶梯，该阶梯根据标识符列、名称列、显示名称和类型范围位置来决定相同性，而不是根据名称相似性。该阶梯管理解析表中的去重，而图谱写入则应用一个更粗糙的规范名称键，因此共享规范名称的记录在完全相等时会自动合并。我们认为而不是证明这是自动化的界限所在：没有报告身份基准，键允许的过度合并在构造上是不可检测的。这一政策下，实体解析仅标记候选者，源于一起事件，其中一个名称的两个表面形式被合并，导致一个正确记录的损坏，并从无关文档中删除了八个实体。接下来，我们描述了多类本体标记和一个我们未曾预料的证据不对称：实体名称是实例标签而不是类型断言，因此将名称片段与类索引匹配会创造分类。要求锚定证据将角色分配从36个减少到4个，所有都被确认正确。我们量化了图谱的合规债务，展示了次级分类补偿错误父类的情况，并描述了一个策划队列，待处理提案增长至48,403个，涉及775个人工决策。
cs.AI / 54 / 2608.10650
Decision-Aware Approximation of Belief Functions for Evidential Combinatorial Optimization
基于决策的证据组合优化信念函数近似
Abstract
Reducing the number of focal elements of a mass function is classically driven by an intrinsic distance, such as Jaccard or Jousselme, that keeps the approximation close to the original as a body of evidence. We consider instead the case where the mass function feeds a linear combinatorial optimisation problem with evidential costs. What should then be preserved is not the closeness of the two mass functions, but the quality of the decision they induce. We introduce a decision-aware approximation that targets the regret of the decision: one decides with the cheaper approximation and is evaluated under the true mass function. On a minimal shortest path, the distance-optimal approximation flips the decision while a decision-aware merge preserves it, and this occurs on a non-negligible fraction of random instances. We prove a one-point bound that localises the regret at the true optimum, turn it into an exact dynamic program for the scalar case, and extend it to an online version that prunes focal elements before the final cost is known. In experiments the decision-aware compressor flips the decision less often than representation-aware compression, for both the linear criterion and a non-linear proxy read-out.
Chinese Translation
减少质量函数的焦点元素数量通常是通过一种内在距离来驱动的，例如 Jaccard 距离或 Jousselme 距离，这种方法能够保持近似与原始证据体的接近性。我们考虑的情况是，质量函数用于具有证据成本的线性组合优化问题。在这种情况下，应该保留的不是两个质量函数的接近性，而是它们所诱导的决策质量。我们引入了一种基于决策的近似方法，旨在降低决策的悔恨：决策者使用更便宜的近似进行决策，并在真实质量函数下进行评估。在最小最短路径上，距离最优近似会翻转决策，而基于决策的合并则保留了决策，这种情况在相当一部分随机实例中发生。我们证明了一个一点评界，定位了真实最优点的悔恨，将其转化为标量情况下的精确动态规划，并扩展到在线版本，在最终成本未知之前修剪焦点元素。在实验中，基于决策的压缩器翻转决策的频率低于基于表示的压缩，无论是在线性标准还是非线性代理读出中。
cs.AI / 55 / 2608.10664
Operationalising Relative Causal Knowledge: Backbone Identifiability from Private Reports on a Shared Outcome
操作化相对因果知识：基于共享结果的私人报告的骨干可识别性
Abstract
The Relativity of Causal Knowledge (RCK) explains how a network of agents with different structural causal models can exchange causal knowledge through a shared interventionally consistent abstraction, or backbone. We ask the prior identification question that this transport mechanism presupposes: when is that backbone determined by the agents' private causal knowledge? In the basic two-agent common-effect case, two private causes influence one shared outcome and each agent identifies only the single-cause causal marginal relevant to its own perspective. We show that, under standard compatibility, non-degeneracy, and local overlap assumptions, those local causal marginals do not identify a unique backbone. Infinitely many joint intervention kernels can induce exactly the same private reports while disagreeing on joint interventions. We then give a conditional recovery result. Additive separability removes the hidden interaction degree of freedom, but observational residual summaries remain insufficient. Identification becomes possible when agents communicate causally identified response functions. An education value-added example illustrates why this is first a communication problem, and only then a policy-composition problem.
Chinese Translation
因果知识的相对性（RCK）解释了不同结构因果模型的代理如何通过共享的干预一致抽象或骨干来交换因果知识。我们提出了这一传输机制所假设的先前识别问题：何时骨干由代理的私人因果知识决定？在基本的双代理共同效应案例中，两个私人原因影响一个共享结果，每个代理仅识别与其自身视角相关的单因因果边际。我们展示了在标准兼容性、非退化性和局部重叠假设下，这些局部因果边际并不能识别唯一的骨干。无穷多个联合干预核可以诱导完全相同的私人报告，同时在联合干预上存在分歧。随后，我们给出了一个条件恢复结果。加性可分离性消除了隐藏的交互自由度，但观察残差摘要仍然不足。当代理交流因果识别的响应函数时，识别变得可能。一个教育增值的例子说明了为什么这首先是一个沟通问题，而仅仅是一个政策组合问题。
cs.AI / 56 / 2608.10665
VERDICT: Training-Free Step-Wise Verification of Multimodal Reasoning via Disagreement-Aware Consensus
VERDICT：通过关注分歧的共识进行无训练的逐步验证多模态推理
Abstract
Multimodal large language models often generate reasoning chains containing subtle errors that lead to incorrect answers. Current verification approaches have notable limitations. Existing approaches either require expensive labelled supervision with inconsistent cross-task performance or aggregate scores from multiple sources by simple aggregations, missing a key insight: when these scores disagree, that disagreement itself carries important information about whether a reasoning step is truly valid or not. We formalise this as a coupled scoring problem among disparate, frozen verifiers, interpretable as a coordination game with a unique closed-form equilibrium where agreement signals valid steps while disagreement reveals instability. Towards this end, we propose a training-free domain-agnostic step-wise verification approach we call VERDICT: VERification via Disagreement-Informed Coupled Thresholding. To our knowledge, VERDICT is the first training-free verifier that makes the structure of cross-modal disagreement explicit and actionable. It computes consensus scores through a closed-form solution, enabling both disagreement-aware filtering and stability-conscious ranking of reasoning steps. Evaluated across six benchmarks, \method consistently improves over the base model by up to +5.95%, and performs competitively with domain-specific critics that demand extensive supervision, demonstrating that cross-modal agreement provides robust verification signals without task-specific adaptation and Training-Free Verification
Chinese Translation
多模态大型语言模型通常生成包含微妙错误的推理链，这些错误会导致不正确的答案。目前的验证方法存在显著的局限性。现有方法要么需要昂贵的标注监督且在跨任务表现上不一致，要么通过简单的聚合从多个来源汇总分数，忽视了一个关键的洞察：当这些分数不一致时，这种不一致本身携带着关于某个推理步骤是否真正有效的重要信息。我们将其形式化为一个不同冻结验证器之间的耦合评分问题，可以解释为一个协调博弈，其具有唯一的闭合形式均衡，其中一致性信号表示有效步骤，而不一致性则揭示不稳定性。为此，我们提出了一种无训练的领域无关的逐步验证方法，称为VERDICT：通过关注分歧的信息进行耦合阈值验证。根据我们的知识，VERDICT是首个明确且可操作地揭示跨模态分歧结构的无训练验证器。它通过闭合形式的解决方案计算共识分数，从而实现了关注分歧的过滤和关注稳定性的推理步骤排名。在六个基准测试中， extit{VERDICT}在基础模型上持续提高了最多5.95%的性能，并与需要大量监督的领域特定评估者竞争，证明跨模态一致性提供了强健的验证信号，而无需特定任务的适应和无训练验证。
cs.AI / 57 / 2608.10668
FITTER: Vocabulary-Agnostic Cross-Domain Inference on Temporal Knowledge Graphs
FITTER：一种与词汇无关的跨领域时间知识图谱推理方法
Abstract
Temporal knowledge graphs are central to many uses of the Semantic Web, but existing completion methods assume the entities, relation names, and timestamps to be reasoned about are already known at training time, restricting each model to a single graph and vocabulary. We propose FITTER, the first fully-inductive structural model for temporal knowledge graph link prediction that supports cross-domain transfer: the inference graph may contain entirely unseen entities, relation names, and timestamps drawn from a different domain. FITTER represents each predicate by its interaction patterns with others and time through encodings of relative rather than absolute ordering; message-passing fuses local and global temporal context to produce vocabulary-agnostic embeddings. We prove the temporal encoding is time-shift invariant and evaluate FITTER on cross-domain, cross-graph transfer over six temporal knowledge graph benchmarks of diverse domains, granularities, and time spans. FITTER consistently outperforms inductive baselines without retraining, indicating that vocabulary-agnostic structural learning is a viable foundation for inference over the heterogeneous knowledge graphs of the Semantic Web.
Chinese Translation
时间知识图谱在语义网的许多应用中至关重要，但现有的补全方法假设在训练时需要推理的实体、关系名称和时间戳已经是已知的，这限制了每个模型只能针对单一图谱和词汇进行操作。我们提出了FITTER，这是首个完全归纳的结构模型，用于时间知识图谱的链接预测，支持跨领域转移：推理图可以包含来自不同领域的完全未见过的实体、关系名称和时间戳。FITTER通过相对而非绝对顺序的编码来表示每个谓词与其他谓词及时间的交互模式；消息传递融合了局部和全局的时间上下文，以生成与词汇无关的嵌入。我们证明了时间编码具有时间平移不变性，并在六个具有不同领域、粒度和时间跨度的时间知识图谱基准上评估了FITTER的跨领域、跨图谱转移能力。FITTER在不重新训练的情况下始终优于归纳基线，表明与词汇无关的结构学习是对语义网异构知识图谱进行推理的可行基础。
cs.AI / 58 / 2608.10669
REDAgentBench: Executable Red Teaming and Faithful Measurement of LLM Agent Systems
REDAgentBench：可执行的红队测试与大语言模型代理系统的真实测量
Abstract
Large language model (LLM) agents combine language-based reasoning with external tools to perform complex tasks. Adversarial inputs can exploit interactions between the agent and its environment, causing the agent to violate safety policies during execution. Yet existing evaluations often reduce agent safety to a single attack success rate (ASR), collapsing exposure, execution, observation, and adjudication and potentially conflating actual violations with evidence visibility. We introduce REDAgentBench, an executable framework for autonomous red-teaming and faithful measurement. It derives attacks from explicit safety constraints and associated agent-system vulnerabilities, runs them in isolated service sandboxes, and verifies harmful effects from service receipts and final-state changes. The benchmark contains 1,661 cases across five service surfaces. Across six models and three agent harnesses, macro-average ASR is 65.69%; reported ASR varies with harness and evidence view, while evaluation-context disclosure changes execution behavior. In a state-grounded diagnostic cohort, almost one in five confirmed violations with resolved action anchors occurs after the agent states the relevant constraint or risk, revealing a Recognition--Execution Gap. Finally, a training-free policy reminder reduces confirmed violations by more than 70 percentage points in matched replay. These findings show that executable evaluation can improve safety measurement and identify actionable intervention points.
Chinese Translation
大型语言模型（LLM）代理结合了基于语言的推理与外部工具，以执行复杂任务。对抗性输入可以利用代理与其环境之间的交互，导致代理在执行过程中违反安全政策。然而，现有评估通常将代理安全性简化为单一的攻击成功率（ASR），将暴露、执行、观察和裁决等环节混为一谈，可能会将实际违规与证据可见性混淆。我们提出了REDAgentBench，这是一个用于自主红队测试和真实测量的可执行框架。它从明确的安全约束和相关的代理系统漏洞中推导攻击，在隔离的服务沙箱中运行这些攻击，并通过服务回执和最终状态变化验证有害影响。该基准包含跨五个服务表面的1,661个案例。在六个模型和三个代理框架中，宏观平均ASR为65.69%；报告的ASR因框架和证据视图而异，而评估上下文的披露会改变执行行为。在一个基于状态的诊断组中，几乎五分之一的确认违规发生在代理陈述相关约束或风险之后，揭示了识别-执行差距（Recognition--Execution Gap）。最后，无需训练的政策提醒在匹配重放中将确认违规减少了70多个百分点。这些发现表明，可执行评估可以改善安全测量并识别可操作的干预点。
cs.AI / 59 / 2608.10676
Self-Correcting Long-Horizon Search Agents via Tree-Structured Memory
通过树结构记忆实现自我纠正的长时间搜索代理
Abstract
Large language model (LLM)-based search agents answer questions through multi-step interactions with external environments. However, providing complete execution trajectories to the LLM causes unbounded context growth and introduces noise. Existing compression methods reduce context at the cost of important details and often replace erroneous facts without repairing downstream reasoning derived from them. To address this problem, we propose ReTree, a self-correcting tree-structured memory mechanism for search agents. ReTree constructs a bounded per-step reasoning context while preserving source-linked evidence. It models search as an evidence tree whose nodes store bounded summaries, evidence, and revision histories. When newly retrieved evidence contradicts an earlier claim, ReTree traces back to the node where the claim was introduced, replaces outdated evidence, regenerates summaries, prunes affected branches, and resumes search. Source-grounded evidence provenance supports reliable conflict localization and keeps final claims traceable to retrieved passages. Experiments on four public question-answering and search benchmarks show that ReTree consistently outperforms Full-Trajectory ReAct, improving answer accuracy by up to 25.6 percentage points (pp); the average maximum per-step reasoning context of Full-Trajectory ReAct is $1.27$--$1.51\times$ that of ReTree. These results establish ReTree as an effective self-correcting memory abstraction for long-horizon search.
Chinese Translation
基于大型语言模型（LLM）的搜索代理通过与外部环境的多步交互来回答问题。然而，向LLM提供完整的执行轨迹会导致上下文无限增长并引入噪声。现有的压缩方法在减少上下文的同时往往牺牲重要细节，并且常常在不修复由此产生的下游推理的情况下替换错误的事实。为了解决这个问题，我们提出了ReTree，一种用于搜索代理的自我纠正树结构记忆机制。ReTree在保留源链接证据的同时，构建了一个有限的逐步推理上下文。它将搜索建模为一个证据树，其节点存储有限的摘要、证据和修订历史。当新检索到的证据与早期声明相矛盾时，ReTree会追溯到引入该声明的节点，替换过时的证据，重新生成摘要，修剪受影响的分支，并恢复搜索。源基础证据的来源支持可靠的冲突定位，并保持最终声明可追溯到检索的段落。在四个公共问答和搜索基准上的实验表明，ReTree始终优于完整轨迹的ReAct，答案准确性提高了多达25.6个百分点（pp）；完整轨迹的ReAct的平均最大逐步推理上下文是ReTree的$1.27$--$1.51 imes$。这些结果确立了ReTree作为长时间搜索的有效自我纠正记忆抽象。
cs.AI / 60 / 2608.10720
Ex-Omni-2D: Expressive Omni-Modal Dialogue Models with Native Visual Presence
Ex-Omni-2D：具有原生视觉存在的表现力全模态对话模型
Abstract
Omni-modal dialogue models can understand multimodal inputs and synthesize spoken replies, yet their responses remain visually disembodied. We introduce \textbf{Ex-Omni-2D}, an omni-modal dialogue framework that generates a coordinated response comprising text, personalized speech, and reference-conditioned video. Given a multimodal query, reference image, and reference audio, the model predicts a structured \textit{Visual Thought Plan} (VTP) describing scene, emotion, and motion, followed by response text and native multi-codebook speech units. These units form a shared acoustic-temporal interface: they are decoded into speech and aligned online with video frames. This interface enables the response and avatar pathways to be learned from heterogeneous speech, dialogue, and avatar-video data, avoiding the need for large-scale query--text--speech--video supervision. A full-sequence Video Generator serves as the primary Teacher. For efficient incremental generation, we further distill it into a few-step block-causal \emph{Streaming Student} whose Prefix Streaming mechanism carries a clean latent across consecutive chunks to reduce cumulative late-chunk degradation. With four-step inference, the complete four-GPU pipeline achieves an end-to-end RTF of 1.293 at $400\times720$/$720\times400$, providing a practical quality--efficiency operating point.
Chinese Translation
全模态对话模型能够理解多模态输入并合成口语回复，但其响应仍然缺乏视觉表现。我们介绍了 extbf{Ex-Omni-2D}，一个生成协调响应的全模态对话框架，该响应包括文本、个性化语音和参考条件视频。给定一个多模态查询、参考图像和参考音频，模型预测一个结构化的 extit{视觉思维计划}（Visual Thought Plan, VTP），描述场景、情感和动作，随后生成响应文本和原生多代码本语音单元。这些单元形成一个共享的声学-时间接口：它们被解码为语音，并与视频帧在线对齐。该接口使得响应和虚拟形象路径能够从异构的语音、对话和虚拟形象视频数据中学习，避免了对大规模查询-文本-语音-视频监督的需求。一个完整序列的视频生成器作为主要教师。为了实现高效的增量生成，我们进一步将其提炼为一个少步块因果的 extit{流式学生}（Streaming Student），其前缀流机制在连续块之间传递一个干净的潜在向量，以减少累积的后期块降级。通过四步推理，完整的四GPU管道在 $400 imes720$/$720 imes400$ 的分辨率下实现了1.293的端到端实时因子（RTF），提供了一个实用的质量-效率操作点。
cs.AI / 61 / 2608.10740
Tree-of-Ideas: Automated Research Ideation via Cross-Trajectory Reasoning over Scholarly Evolution
思想树：通过跨轨迹推理实现自动化研究构思
Abstract
Effective research ideation requires moving beyond a static understanding of prior work to trace how research problems and solutions evolve across the literature. Existing methods either treat papers as unstructured context or model scholarly evolution as isolated citation chains, overlooking interactions among research trajectories. We propose Tree-of-Ideas (ToI), a two-stage framework. EvoTrace reconstructs branching scholarly trajectories from citations, tracking evolving methods, resolved problems, and gaps. EvoAgent then reasons across trajectories to identify convergent problems and complementary solutions, generating grounded research ideas. Across six AI research topics, ToI achieves the highest score among automatic methods (6.27 vs. 5.36 for the strongest baseline on a 10-point scale), with strong Novelty (6.36) and Groundedness (7.00). Also, its score approaches that of human-paper references (6.29), demonstrating the value of cross-path evolutionary reasoning.
Chinese Translation
有效的研究构思需要超越对先前工作的静态理解，追踪研究问题和解决方案在文献中的演变。现有方法要么将论文视为无结构的背景，要么将学术演变建模为孤立的引用链，忽视了研究轨迹之间的相互作用。我们提出了思想树（Tree-of-Ideas, ToI），一个两阶段框架。EvoTrace 从引用中重建分支学术轨迹，跟踪不断演变的方法、已解决的问题和研究空白。EvoAgent 随后在轨迹之间进行推理，以识别趋同的问题和互补的解决方案，从而生成有根据的研究构思。在六个人工智能研究主题中，ToI 在自动化方法中获得了最高分（6.27，相较于最强基线的 5.36，满分为 10 分），并在新颖性（6.36）和基础性（7.00）方面表现出色。此外，其得分接近人类论文引用的分数（6.29），展示了跨路径演化推理的价值。
cs.AI / 62 / 2608.10765
Compositional Benchmark Synthesis for Hierarchical Human Action Recognition
层次化人类行为识别的组合基准合成
Abstract
Recognizing human behavior across levels of abstraction, from atomic actions to long-horizon intentions, requires data annotated along a semantic hierarchy. Large corpora provide isolated, atomically labeled clips without temporal composition, whereas recorded composite-activity corpora offer shallow, domain-narrow, fixedhierarchies. A benchmark-generation and evaluation frameworkis proposed that synthesizes a four-level hierarchical-intention benchmark, spanning actions, activities, low-level intentions (LLIs), and high-level intentions (HLIs), from a flat single-label action corpus while retaining real pre-extracted features at the action level. Episodes are assembled by a transition model under a subject-consistency constraint, and a coverage-aware sampler reduces the subject usage Gini from 0.566 to 0.248. Synthesizing such a benchmark raises a circular-supervision risk that recorded datasets avoid: if the rules generating the episodes also govern the evaluation, models can succeed by recovering the generator rather than through genuine reasoning. Validity is addressed by design, holding sequence-generation rules disjoint from the first-order-logic rules used at evaluation. The instantiation yields 15,002 episodes. Four reference baselines from different model families characterize difficulty, not as recognition methods. A compositional held-out gap of 0.13 to 0.17 macro-F1 appears across all baselines, including a graph-aware model that recognizes best yet does not close the gap, indicating a structural property of the benchmark rather than a model artifact. A logic-free baseline still violates the held-out semantic rules above their intrinsic data rate, and the order-destroying control changes macro-F1 within seed variation, serving as a generator-consistency check. Theontology, transition model, and generator are released so the benchmark can beregenerated and extended.
Chinese Translation
在从原子动作到长远意图的不同抽象层次上识别人类行为，需要沿着语义层次进行数据标注。大型语料库提供孤立的、原子标记的片段，而没有时间组合，而记录的复合活动语料库则提供浅层、领域狭窄的固定层次结构。本文提出了一种基准生成和评估框架，该框架从平坦的单标签动作语料库合成一个四层层次意图基准，涵盖动作、活动、低层意图（LLIs）和高层意图（HLIs），同时保留动作层面的真实预提取特征。通过在主体一致性约束下的转移模型组装情节，并且一个关注覆盖率的采样器将主体使用的基尼系数从0.566降低到0.248。合成这样的基准引发了一个循环监督风险，而记录的数据集则避免了这一风险：如果生成情节的规则也支配评估，模型可能通过恢复生成器而非真正的推理来取得成功。通过设计解决有效性问题，将序列生成规则与评估中使用的一阶逻辑规则保持不相交。该实例化生成了15,002个情节。来自不同模型家族的四个参考基线表征了难度，而不是作为识别方法。所有基线之间出现了0.13到0.17的组合保留差距的宏F1，包括一个识别效果最佳但未能缩小差距的图意识模型，这表明基准的结构特性而非模型伪影。一个无逻辑基线仍然违反了上述高于其内在数据速率的保留语义规则，而破坏顺序的控制在种子变异中改变了宏F1，作为生成器一致性的检查。本文发布了本体、转移模型和生成器，以便基准可以被重新生成和扩展。
cs.AI / 63 / 2608.10766
Rule of Thumb: Explaining Artificial Intelligence Systems using Partial Information
经验法则：使用部分信息解释人工智能系统
Abstract
Explainable Artificial Intelligence (XAI) seeks to explain how an Artificial Intelligence (AI) system arrived at a particular decision. We propose ''Rule of Thumb'' (RoT) explanations, a new approach to XAI based upon a novel formulation that identifies the most relevant features for predicting the behaviour of an AI system, for a particular datapoint. We show how RoT is well-suited to enable XAI in: (a) zero-shot classification using large language models (LLMs), (b) auditing of opaque AI systems without model access, and (c) the use of AI in scientific discovery. Additionally, RoT meets specific requirements from leading AI regulations, provides a familiar interface and visualisations for XAI practitioners, is model-agnostic, and is substantially faster than alternatives. Code available at: https://github.com/KaiRawal/Rule-of-Thumb-Explaining-Artificial-Intelligence-Systems-using-Partial-Information
Chinese Translation
可解释的人工智能（XAI）旨在解释人工智能（AI）系统如何做出特定决策。我们提出了“经验法则”（Rule of Thumb, RoT）解释，这是一种基于新颖公式的XAI新方法，旨在识别预测特定数据点的AI系统行为的最相关特征。我们展示了RoT如何适用于以下场景中的XAI： (a) 使用大型语言模型（LLMs）进行零样本分类， (b) 在没有模型访问权限的情况下审计不透明的AI系统，以及 (c) 在科学发现中使用AI。此外，RoT满足领先AI法规的特定要求，为XAI从业者提供熟悉的界面和可视化，具有模型无关性，并且比其他替代方案显著更快。代码可在以下链接获取：https://github.com/KaiRawal/Rule-of-Thumb-Explaining-Artificial-Intelligence-Systems-using-Partial-Information
cs.AI / 64 / 2608.10775
SkillLens: Visual Skill Cards for Retrieval-Augmented GUI Action Prediction and On-Policy Distillation
SkillLens：用于检索增强的图形用户界面动作预测和在线策略蒸馏的视觉技能卡
Abstract
Computer-using agents can perceive rich software interfaces, yet their decisions often lack visual procedural memory: they may recognize individual controls without identifying which familiar workflow is active, which control matters next, or what evidence would confirm progress. Raw interaction traces preserve such information but are long and noisy to condition on, whereas text-only skills often omit the visual state that makes a procedure applicable. We introduce Visual Skill Cards (VSCs), a state-conditioned memory representation that binds reusable procedures with applicability cues, visual evidence, and verification signals. SkillLens constructs VSCs from heterogeneous interaction experience through Trace-to-Visual-Skill-Card and, at inference time, retrieves relevant cards and selectively expands only the evidence needed by a fixed visual-language model executor for grounded GUI action prediction. The same representation also supports CardDistill, which uses VSC evidence as privileged teacher context to train a student that acts without runtime card retrieval. Across Multimodal-Mind2Web and WebLINX-BrowserGym, SkillLens improves the frozen GPT-5.4-mini executor by +11.6 points in Step SR and +2.9 points in Overall, respectively; CardDistill further improves the corresponding student-only Qwen3-VL-2B metrics by +12.0 and +3.2 points.
Chinese Translation
计算机使用代理能够感知丰富的软件界面，但它们的决策往往缺乏视觉程序记忆：它们可能识别单个控件，却无法确定哪个熟悉的工作流程处于活动状态，哪个控件是下一个重要的，或什么证据可以确认进展。原始交互轨迹保留了这些信息，但在条件化时往往冗长且噪声较大，而仅包含文本的技能通常忽略了使程序适用的视觉状态。我们引入了视觉技能卡（Visual Skill Cards, VSCs），这是一种状态条件化的记忆表示，将可重用的程序与适用性线索、视觉证据和验证信号绑定在一起。SkillLens通过从异构交互经验构建VSCs，采用Trace-to-Visual-Skill-Card方法，并在推理时检索相关卡片，仅选择固定视觉语言模型执行器所需的证据，以进行有根据的图形用户界面动作预测。相同的表示也支持CardDistill，该方法利用VSC证据作为特权教师上下文，训练一个无需运行时卡片检索的学生。在Multimodal-Mind2Web和WebLINX-BrowserGym上，SkillLens使冻结的GPT-5.4-mini执行器在步骤成功率（Step SR）上提高了11.6分，在整体表现上提高了2.9分；CardDistill进一步使对应的仅学生Qwen3-VL-2B指标分别提高了12.0分和3.2分。
cs.AI / 65 / 2608.10792
ChemWorld: Programmable Chemical Worlds for Controlled and Replayable Agent Experimentation
ChemWorld：可编程化学世界用于受控和可重放的智能体实验
Abstract
Autonomous chemistry increasingly depends on environments in which agents can repeatedly act, observe, and adapt.Physical laboratories provide essential real-material evidence but are costly to repeat and difficult to use for tightly matched interventions, whereas most digital environments keep the underlying experimental world largely fixed. We introduce ChemWorld, a programmable chemical environment in which reusable process and observation components are compiled into executable worlds. ChemWorld separates the public experimental contract available to an agent from evaluator-owned chemical and material laws. Researchers can therefore vary world composition and operating conditions, or change a single hidden law while holding the public task and interaction conditions fixed. Transactional execution records operations, failures, resource changes, and state transitions, allowing complete environment-action trajectories to be replayed exactly and audited. Full-census qualification covered the reference registry, 52 generated compositions, and module, interface, compilation, and invalid-action tests. Eight deterministic experimental cases demonstrated shared lifecycle semantics, failure recovery, and exact replay, while six parent-child world-fork pairs isolated the effects of single private-law interventions under matched public conditions. An independent agent also completed a full lifecycle in a non-reference world through the same public interface. Within the declared component and model domain, ChemWorld provides a controlled and replayable substrate for studying experimentation across systematically varied chemical worlds, complementary to physical-laboratory evidence and calibration.
Chinese Translation
自主化学越来越依赖于智能体可以反复行动、观察和适应的环境。物理实验室提供了重要的真实材料证据，但重复实验成本高且难以用于严格匹配的干预，而大多数数字环境则保持基础实验世界基本固定。我们介绍了ChemWorld，一个可编程的化学环境，其中可重用的过程和观察组件被编译成可执行的世界。ChemWorld将可供智能体使用的公共实验契约与评估者拥有的化学和材料法则分开。因此，研究人员可以在保持公共任务和交互条件不变的情况下，改变世界的组成和操作条件，或改变单一的隐藏法则。事务执行记录了操作、失败、资源变化和状态转变，允许完整的环境-行动轨迹被精确重放和审计。全面的普查资格涵盖了参考注册、52种生成的组合，以及模块、接口、编译和无效行动测试。八个确定性实验案例展示了共享生命周期语义、故障恢复和精确重放，而六对父子世界分叉则在匹配的公共条件下隔离了单一私有法则干预的影响。一个独立的智能体还通过相同的公共接口在非参考世界中完成了完整的生命周期。在声明的组件和模型领域内，ChemWorld提供了一个受控且可重放的基础，用于研究在系统变化的化学世界中进行实验，这与物理实验室的证据和校准互为补充。
cs.AI / 66 / 2608.10795
EvoMem: Memory-Augmented Evolution for Code Optimization
EvoMem：用于代码优化的增强记忆进化方法
Abstract
Successful mutation strategies in evolutionary code search may contain reusable knowledge that is useful beyond a single run, and in some cases may transfer across related tasks and domains. However, existing LLM-driven evolutionary frameworks largely discard such knowledge, repeatedly rediscovering similar ideas and limiting opportunities for cross-run and cross-task learning. We introduce EvoMem, a persistent memory architecture for LLM-based evolutionary program search that captures and reuses candidate mutation knowledge. EvoMem converts successful mutation events into structured, task-aware advice for future runs. It operates in two phases: after each run, it extracts and stores promising ideas with provenance, and during subsequent evolution, it retrieves a small set of relevant instructions based on the current task and program context to guide mutation. Across geometric optimization, multi-hop question answering, GPU kernel optimization, and related benchmarks, our experiments show positive average improvements in target metrics or search speed for most evaluated settings, while also revealing variability across tasks. Overall, EvoMem provides evidence that persistent memory can reduce some redundant exploration and improve the reuse and adaptation of successful strategies in LLM-driven evolutionary search.
Chinese Translation
成功的变异策略在进化代码搜索中可能包含可重复使用的知识，这些知识在单次运行之外也具有价值，并且在某些情况下可以跨相关任务和领域进行转移。然而，现有的基于大语言模型（LLM）的进化框架在很大程度上忽视了这些知识，反复重新发现类似的想法，限制了跨运行和跨任务学习的机会。我们提出了EvoMem，一种用于基于LLM的进化程序搜索的持久记忆架构，它捕获并重用候选变异知识。EvoMem将成功的变异事件转化为结构化的、任务感知的建议，以供未来的运行使用。它分为两个阶段：在每次运行后，它提取并存储具有来源的有前景的想法；在后续的进化过程中，它根据当前任务和程序上下文检索一小组相关的指令，以指导变异。在几何优化、多跳问答、GPU内核优化和相关基准测试中，我们的实验显示，在大多数评估设置中，目标指标或搜索速度的平均改善是积极的，同时也揭示了任务之间的变异性。总体而言，EvoMem提供了证据，表明持久记忆可以减少一些冗余探索，并改善成功策略在基于LLM的进化搜索中的重用和适应。
cs.AI / 67 / 2608.10843
Hypothesis Frontier: Verifier Guided LLM and Symbolic Search for First-Order Induction
假设前沿：验证者引导的LLM与一阶归纳的符号搜索
Abstract
First-order concept synthesis asks a system to infer one formula that classifies labeled objects consistently across several finite relational structures. Every candidate can be evaluated exactly, but quantified first-order formulas form a vast search space, and LLM outputs are often semantically promising without being fully correct. We introduce Hypothesis Frontier, a verifier-guided neurosymbolic framework that evaluates each LLM formula on every training object, retains the strongest verified hypothesis across rounds, and uses its remaining errors to guide subsequent generation. Symbolic processing repairs invalid formulas while remaining anchored to the LLM-generated hypothesis, and simplifies train-valid formulas without changing any training prediction. Under matched models, problem sets, and LLM-round budgets, Hypothesis Frontier solves substantially more problems than repeated original-prompt generation. After the final formulas are selected, exact simplification shortens many train-valid formulas while preserving every training prediction. Exact symbolic reasoning therefore helps both to solve more induction problems and to compress many of the resulting formulas.
Chinese Translation
一阶概念合成要求系统推导出一个公式，该公式在多个有限关系结构中一致地对标记对象进行分类。每个候选公式都可以被精确评估，但量化的一阶公式形成了一个庞大的搜索空间，而LLM的输出通常在语义上是有前景的，但并不完全正确。我们提出了假设前沿（Hypothesis Frontier），这是一个验证者引导的神经符号框架，它在每个训练对象上评估每个LLM公式，保留跨轮次验证最强的假设，并利用其剩余错误来指导后续生成。符号处理修复无效公式，同时仍然与LLM生成的假设保持一致，并在不改变任何训练预测的情况下简化训练-验证公式。在匹配的模型、问题集和LLM轮次预算下，假设前沿解决的问题数量显著多于重复生成原始提示。在最终公式被选定后，精确简化缩短了许多训练-验证公式，同时保留了每个训练预测。因此，精确的符号推理不仅有助于解决更多的归纳问题，还能压缩许多结果公式。
cs.AI / 68 / 2608.10881
Enhanced Filtering Algorithms for the Euclidean Traveling Salesperson Problem and its variants in Constraint Logic Programming
增强的欧几里得旅行推销员问题及其变体的过滤算法在约束逻辑编程中的应用
Abstract
The Traveling Salesperson Problem (TSP) is one of the best-known problems in computer science and arises in many engineering applications, such as smart vehicles and intelligent transportation systems. In the "Euclidean" case, each node is defined by its coordinates in the plane and distances are computed using the Euclidean metric. In the Constraint Programming (CP) literature, the Euclidean TSP is typically addressed by computing the full distance matrix and treating it as a general case; however this approach ignores the geometric information carried by the points' coordinates. In this work, we propose new filtering algorithms, implemented in Constraint Logic Programming (CLP), that exploit such geometric information to achieve stronger constraint propagation than existing approaches. Moreover, we show how this methodology can be extended to other Euclidean variants of the TSP, including the Euclidean Generalized Traveling Salesperson Problem (EGTSP), which is relevant in practical routing and logistics applications. Experimental results demonstrate the computational advantages of the proposed approach.
Chinese Translation
旅行推销员问题（TSP）是计算机科学中最著名的问题之一，并在许多工程应用中出现，例如智能车辆和智能交通系统。在“欧几里得”情况下，每个节点由其在平面上的坐标定义，距离则使用欧几里得度量进行计算。在约束编程（CP）文献中，欧几里得TSP通常通过计算完整的距离矩阵并将其视为一般情况来解决；然而，这种方法忽略了点坐标所携带的几何信息。在本研究中，我们提出了新的过滤算法，这些算法在约束逻辑编程（CLP）中实现，利用这种几何信息以实现比现有方法更强的约束传播。此外，我们展示了该方法如何扩展到TSP的其他欧几里得变体，包括在实际路由和物流应用中相关的欧几里得广义旅行推销员问题（EGTSP）。实验结果证明了所提方法的计算优势。
cs.AI / 69 / 2608.10915
ComBodied Agents: a New Paradigm of Human-Centric Agentic AI
共生体代理：以人为中心的代理智能新范式
Ding, Qianggang, Wang, Xingyao, Feng, Rui, Wang, Zhibin, Wang, Feixiang, Mao, Kelong, Sun, Hao, Luo, Zhiyao, Tang, Jiankai, Li, Lei, Guo, Jiadong, Ni, Minheng, Lin, Weicong, Yang, Chenxi, Gao, Hongxiang, Chen, Zhenghua, Bai, Yang, Wu, Min, Cheng, Jun, Fu, Huazhu, Tao, Dacheng, Liu, Bang
Abstract
After an older adult misses a medication dose, a software agent can send another reminder and an embodied agent can bring the medication. Yet neither explains whether the person forgot, is confused, has side effects, or deliberately refused, nor what support is appropriate. This reveals a structural gap in Agentic AI: Digital Agents primarily transform software states, while Embodied Agents transform physical states; neither makes a person's evolving state and agency the primary object of modeling, intervention, and evaluation. We introduce Combodied Agents, a human-centered paradigm that perceives, models, predicts, and supports individual human-state trajectories over time, using software tools, sensors, wearables, robots, and human services as action channels rather than end goals. We unify fragmented capabilities across personal assistants, health agents, AI companions, and adaptive human--AI systems into a closed loop: event-based multimodal perception reconstructs meaningful personal events; longitudinal, correctable memory provides temporal context; Personal World Models estimate future personal states and outcomes under alternative decisions and interventions; and an admissible intervention policy selects proportionate support under consent, uncertainty, safety, reversibility, and user control. Feedback from the person and environment updates the loop. Rather than requiring an exhaustive Human Digital Twin, the framework uses purpose-bounded, uncertainty-aware, user-correctable representations. We organize the design space by human-state targets, relational contexts, and agent roles, and propose scenario-centered evaluation, agency-preservation metrics, benchmark requirements, edge-native personal models, and governance directions. Combodied Agents shift Agentic AI from external task completion toward sustained human benefit.
Chinese Translation
当老年人错过一次药物剂量时，软件代理可以发送另一个提醒，而具身代理可以送来药物。然而，两者都没有解释该人是忘记、困惑、出现副作用，还是故意拒绝，也没有说明什么支持是合适的。这揭示了代理智能中的一个结构性缺口：数字代理主要转变软件状态，而具身代理则转变物理状态；两者都没有将个人不断变化的状态和代理性作为建模、干预和评估的主要对象。我们提出了共生体代理（ComBodied Agents），这是一个以人为中心的范式，能够感知、建模、预测并支持个体的人类状态轨迹，利用软件工具、传感器、可穿戴设备、机器人和人类服务作为行动渠道，而非最终目标。我们将个人助手、健康代理、人工智能伴侣和自适应人类-人工智能系统的碎片化能力统一为一个闭环：基于事件的多模态感知重构有意义的个人事件；纵向的、可纠正的记忆提供时间背景；个人世界模型（Personal World Models）在不同决策和干预下估计未来的个人状态和结果；可接受的干预政策在同意、不确定性、安全性、可逆性和用户控制下选择适当的支持。来自个人和环境的反馈更新了这个闭环。该框架不需要详尽的人类数字双胞胎，而是使用目的明确、关注不确定性、可由用户纠正的表征。我们通过人类状态目标、关系背景和代理角色组织设计空间，并提出以场景为中心的评估、代理性保留指标、基准要求、边缘原生个人模型和治理方向。共生体代理将代理智能的重点从外部任务完成转向持续的人类利益。
cs.AI / 70 / 2608.10920
IO Factory: Simulating AI-Enabled Influence Campaigns at Scale
IO工厂：大规模模拟人工智能驱动的影响力活动
Abstract
We introduce IO Factory, an AI-driven framework for simulating information and influence campaigns as fully integrated, traceable processes. The threat of digital manipulation now extends beyond persuasive text from individual language models to AI swarms, i.e., persistent groups of coordinated agents that adapt to platform feedback and disguise organized campaigns as ordinary social interaction. Because such campaigns cannot be identified from isolated messages alone, they must be analyzed across a continuous spectrum of planning, platform action, exposure, interpretation, measurement, and adaptation. IO Factory represents this process inside a controlled simulated platform, linking actor roles, platform actions, exposure records, structured model-based evaluations, and configured changes in the simulated population. We implement the architecture and evaluate it across configurations of up to 100,000 agents. The results show that IO Factory executes campaign timelines at scale and produces inspectable evidence of exposure and measured movement in configured belief variables. By recording the actors, objectives, action constraints, exposure paths, and measurement rules used in each run, IO Factory supports reproducible research and red-team analysis of coordinated influence.
Chinese Translation
我们介绍了IO工厂，这是一个用于模拟信息和影响力活动的人工智能驱动框架，作为一个完全整合、可追溯的过程。数字操控的威胁现在不仅限于单个语言模型的说服性文本，还扩展到人工智能群体，即持续协调的代理组，这些代理组能够根据平台反馈进行调整，并将有组织的活动伪装成普通的社交互动。由于此类活动无法仅通过孤立的信息进行识别，因此必须在规划、平台行动、曝光、解释、测量和适应的连续谱系中进行分析。IO工厂在一个受控的模拟平台内表示这一过程，连接了参与者角色、平台行动、曝光记录、结构化的基于模型的评估以及模拟人群中的配置变化。我们实现了该架构，并在多达100,000个代理的配置下进行了评估。结果表明，IO工厂能够大规模执行活动时间表，并生成可检查的曝光证据和配置信念变量的测量变化。通过记录每次运行中使用的参与者、目标、行动约束、曝光路径和测量规则，IO工厂支持可重复的研究和协调影响的红队分析。
cs.AI / 71 / 2608.10928
ThinkRetrieve: Retrieval-Augmented Reasoning Traces for Test-Time Scaling
ThinkRetrieve：用于测试时间扩展的增强检索推理轨迹
Abstract
Large Reasoning Models (LRMs) improve performance by allocating additional inference-time compute to generate extended chain-of-thought reasoning. However, recent studies reveal that sequential test-time scaling often yields diminishing or even negative returns, as longer traces exhibit increased uncertainty, error compounding, and drift from the original problem. We propose ThinkRetrieve, a test-time scaling framework that augments the reasoning traces of LRMs with dynamically retrieved solved examples at each reasoning step. Given an external corpus of problems paired with step-by-step solutions, ThinkRetrieve retrieves relevant exemplars at each intermediate step and injects them directly into the thinking trace, providing the model with guidance on how to reason rather than merely what facts are relevant. Experiments across five reasoning models (1.5B--8B parameters) on GSM-8K, MATH-500, AIME 2025, and SciQ demonstrate that ThinkRetrieve consistently improves accuracy over standard test-time scaling, with relative gains of up to $60\%$ on AIME 2025.
Chinese Translation
大型推理模型（LRMs）通过在推理过程中分配额外的计算资源来生成扩展的思维链，从而提高性能。然而，最近的研究表明，顺序的测试时间扩展往往会导致收益递减甚至负收益，因为较长的推理轨迹表现出更高的不确定性、错误累积以及与原始问题的偏离。我们提出了ThinkRetrieve，这是一种测试时间扩展框架，通过在每个推理步骤中动态检索已解决示例来增强LRMs的推理轨迹。给定一组与逐步解决方案配对的问题外部语料库，ThinkRetrieve在每个中间步骤检索相关的示例，并将其直接注入思维轨迹中，为模型提供如何推理的指导，而不仅仅是相关事实。针对GSM-8K、MATH-500、AIME 2025和SciQ的五个推理模型（参数量从1.5B到8B）的实验表明，ThinkRetrieve在标准测试时间扩展上始终提高了准确性，在AIME 2025上相对增益高达60%。
cs.AI / 72 / 2608.10929
FedCGR: Federated Cross-Domain Generative Recommendation
FedCGR：联邦跨域生成推荐
Abstract
Cross-domain recommendation (CDR) transfers preference knowledge across related domains, but federated deployment makes cross-domain alignment difficult because the behavioral anchors that align item spaces, such as overlapping users and shared interaction signals, are often sparse, unavailable, or privacy-sensitive across clients. To address this tension, we revisit federated CDR as generation over a stable semantic item language. By representing items as discrete semantic ID (SID) sequences derived from public item-side metadata, cross-domain item alignment is induced by a shared vocabulary rather than by exchanging private interactions or aligning domain-specific embeddings. Directly federating SID-based generators, however, introduces two design constraints: the SID tokenizer must remain fixed to preserve cross-client token consistency, which creates a semantic-only bottleneck because local collaborative filtering (CF) signals cannot be globally shared or aligned; meanwhile, standard federated averaging can cause negative transfer under domain heterogeneity. To overcome these constraints, we propose FedCGR, a federated generative CDR framework that keeps the item language stable and makes adaptation explicit. FedCGR injects local CF evidence through a reliability-aware semantic interface and trains a prototype-personalized generator that selectively aggregates shared parameters according to domain relatedness while keeping domain-specific quantities local. Experiments on six Amazon cross-domain scenarios show that FedCGR consistently outperforms federated generative baselines and achieves competitive performance against strong sequential and federated CDR methods under both full-ranking and sampled evaluation protocols.
Chinese Translation
跨域推荐（CDR）在相关领域之间转移偏好知识，但联邦部署使得跨域对齐变得困难，因为对齐项目空间的行为锚点（如重叠用户和共享交互信号）在客户端之间往往稀疏、不可用或涉及隐私。为了解决这一矛盾，我们重新审视联邦CDR，作为在稳定的语义项目语言上进行生成。通过将项目表示为源自公共项目侧元数据的离散语义ID（SID）序列，跨域项目对齐是通过共享词汇而非交换私密交互或对齐特定领域的嵌入来实现的。然而，直接联邦化基于SID的生成器引入了两个设计约束：SID分词器必须保持固定，以保持跨客户端的标记一致性，这造成了仅基于语义的瓶颈，因为本地协同过滤（CF）信号无法被全局共享或对齐；与此同时，标准的联邦平均在领域异质性下可能导致负迁移。为克服这些约束，我们提出了FedCGR，一个联邦生成CDR框架，它保持项目语言的稳定性并明确适应性。FedCGR通过一个可靠性感知的语义接口注入本地CF证据，并训练一个原型个性化生成器，该生成器根据领域相关性选择性地聚合共享参数，同时保持领域特定的量为本地。对六个亚马逊跨域场景的实验表明，FedCGR始终优于联邦生成基线，并在全排名和抽样评估协议下对强大的序列和联邦CDR方法表现出竞争力。
cs.AI / 73 / 2608.10976
XCoT-VLA: Executable Chain-of-Thought for Vision-Language-Action Driving
XCoT-VLA：可执行的视觉-语言-动作驱动链思维
Abstract
Vision-Language-Action (VLA) models can connect scene understanding, semantic reasoning, and trajectory generation for autonomous driving. However, verbose natural-language Chain-of-Thought (CoT) is poorly suited to real-time control because it is open-ended, costly to decode, and difficult to optimize as an action-facing representation. We propose XCoT-VLA, which replaces descriptive rationales with compact executable CoT tokens learned from automatically constructed Reason-Action supervision. Logged trajectories provide action evidence, while scene context supplies causal semantics. The predicted XCoT sequence remains in context and conditions fixed trajectory queries through shared multimodal self-attention. Deterministic token-function routing applies the Reason FFN to XCoT tokens and the Control FFN to trajectory queries for flow-matching trajectory generation. We further introduce XCoT Policy Optimization (XCPO) as an optional refinement extension in the same executable token space. XCoT-VLA reduces longitudinal ADE from 1.645 to 1.323 on a general-distribution set and lateral FDE from 1.616 to 0.648 in lane-change scenarios. By representing driving-oriented reasoning with only 2-6 executable XCoT tokens, our method substantially reduces autoregressive reasoning overhead and remains within the real-time planning budget. These results demonstrate that driving-oriented reasoning can be compact, executable, and directly connected to trajectory generation.
Chinese Translation
视觉-语言-动作（VLA）模型能够连接场景理解、语义推理和自主驾驶的轨迹生成。然而，冗长的自然语言链思维（CoT）不适合实时控制，因为它是开放式的、解码成本高且难以优化为面向动作的表示。我们提出了XCoT-VLA，它用从自动构建的理由-动作监督中学习到的紧凑可执行的CoT标记替代了描述性理由。记录的轨迹提供了行动证据，而场景上下文则提供了因果语义。预测的XCoT序列保持在上下文中，并通过共享的多模态自注意力条件固定的轨迹查询。确定性标记-功能路由将理由前馈网络（Reason FFN）应用于XCoT标记，将控制前馈网络（Control FFN）应用于轨迹查询，以实现流匹配的轨迹生成。我们进一步引入XCoT策略优化（XCPO）作为同一可执行标记空间中的可选精细化扩展。XCoT-VLA在一般分布集上将纵向平均绝对误差（ADE）从1.645降低到1.323，在变道场景中将横向最终定位误差（FDE）从1.616降低到0.648。通过仅用2-6个可执行的XCoT标记表示面向驾驶的推理，我们的方法显著减少了自回归推理的开销，并保持在实时规划预算内。这些结果表明，面向驾驶的推理可以是紧凑的、可执行的，并且与轨迹生成直接相关。
cs.AI / 74 / 2608.11047
V-FiLLM: Verified Financial LLM Reasoning Benchmark
V-FiLLM：经过验证的金融大语言模型推理基准
Abstract
While existing benchmarks have made substantial progress in evaluating LLMs across STEM domains, financial reasoning over structured data remains comparatively less explored. We introduce V-FiLLM, a framework that generates financial reasoning benchmarks from executable computation trees grounded in real tables, yielding items whose answers are correct by construction. Trees are evaluated symbolically to obtain ground truth and rendered into natural-language questions, removing any model from the labeling loop, so items can be generated at arbitrary scale without annotation cost and without inheriting a generator's error rate. V-FiLLM exposes four independently controllable axes of difficulty including computation depth, expression breadth, financial concept complexity, and context size. By evaluating on open-source models, we find that accuracy falls up to 51% as reasoning depth increases, and up to 47% points under adversarial numerical perturbations, highlighting remaining challenges in robust financial reasoning over tables. We further show that lightweight LoRA fine-tuning on verified chain-of-thought traces improves accuracy from 81.1% to 85.6% on held-out problems and outperforms the base model by 5% points on FinQA (Chen et al., 2022a), s), suggesting that targeted, low-cost adaptation is a promising direction for compositional reasoning in financial QA.
Chinese Translation
尽管现有基准在评估大语言模型（LLMs）在STEM领域的表现方面取得了显著进展，但在结构化数据上的金融推理仍然相对较少探索。我们介绍了V-FiLLM，一个从基于真实表格的可执行计算树生成金融推理基准的框架，生成的题目其答案是通过构造保证正确的。树被符号化评估以获取真实答案，并转化为自然语言问题，从而将任何模型排除在标注循环之外，使得题目可以在任意规模下生成，而无需标注成本且不继承生成器的错误率。V-FiLLM揭示了四个独立可控的难度轴，包括计算深度、表达广度、金融概念复杂性和上下文大小。通过对开源模型的评估，我们发现随着推理深度的增加，准确率下降最多可达51%，在对抗性数值扰动下下降最多可达47个百分点，这突显了在表格上进行稳健金融推理所面临的挑战。我们进一步展示了在经过验证的思维链迹象上进行轻量级LoRA微调可以将在保留问题上的准确率从81.1%提高到85.6%，并在FinQA（Chen et al., 2022a）上比基础模型提高5个百分点，这表明有针对性的、低成本的适应是金融问答中组合推理的一个有前景的方向。
cs.AI / 75 / 2608.11079
SkillZip: Evaluation-Free Skill Compression for Self-Evolving Agents by Discovering Reusable Structure
SkillZip：通过发现可重用结构实现自我演化代理的无评估技能压缩
Abstract
Self-evolving agents accumulate reusable skills by appending successful procedures and failure fixes. Over time, the same requirement is often restated in several branches, examples, and warnings, while common action sequences are copied rather than reused. The resulting skill becomes expensive to inject and difficult to maintain. Generic prompt compression is ill-suited to this setting because a skill is not a flat passage: its name and description define when it applies, its workflow controls execution, its tool and output contracts constrain validity, and rare exceptions may remain essential even when no sampled task activates them. Evaluation-guided compression can test these behaviors, but it introduces rollouts, cost, and dependence on the compression-time evaluation set. We present SkillZip, an evaluation-free method that compresses a skill by finding its shortest faithful structural explanation. The intuition is explain once, reference many: state a repeated rule once at the scope where it applies, factor a repeated action sequence into a shared procedure, and keep only the differences as explicit exceptions. We formalize this intuition as a typed minimum description-length objective over a skill contract and a residual, subject to a hard coverage constraint for every extracted trigger, workflow edge, tool requirement, obligation, and output field. The formulation provides simple sharing thresholds, preserves unique rare rules by construction, and supports efficient local updates. SkillZip has a one-shot mode with one structured extraction call and deterministic optimization, and a continual Zip-on-Write mode that integrates each self-evolution patch without replaying tasks or reparsing the full history. Through comprehensive experimental evaluations, we demonstrate the effectiveness and superiority of SkillZip in compression performance, generalizability, and cost overhead.
Chinese Translation
自我演化代理通过附加成功的程序和失败的修复来积累可重用技能。随着时间的推移，相同的需求往往在多个分支、示例和警告中反复出现，而常见的行动序列则是被复制而非重用。由此产生的技能在注入时变得昂贵，并且难以维护。通用的提示压缩不适合这种情况，因为技能并不是一个平坦的通道：其名称和描述定义了何时适用，工作流程控制执行，工具和输出契约限制有效性，而即使没有样本任务激活，稀有的例外情况仍可能是必不可少的。基于评估的压缩可以测试这些行为，但它引入了回滚、成本以及对压缩时评估集的依赖。我们提出了SkillZip，这是一种无评估的方法，通过寻找技能的最简忠实结构解释来压缩技能。其直觉是一次解释，多次引用：在适用的范围内一次性陈述重复规则，将重复的行动序列因素化为共享程序，并仅将差异保留为显式例外。我们将这一直觉形式化为一种针对技能契约和残差的类型化最小描述长度目标，并对每个提取的触发器、工作流程边缘、工具要求、义务和输出字段施加严格的覆盖约束。该公式提供了简单的共享阈值，通过构造保留独特的稀有规则，并支持高效的局部更新。SkillZip具有一次性模式，包含一次结构化提取调用和确定性优化，以及持续的Zip-on-Write模式，能够在不重放任务或重新解析完整历史的情况下集成每个自我演化补丁。通过全面的实验评估，我们展示了SkillZip在压缩性能、可泛化性和成本开销方面的有效性和优越性。
cs.AI / 76 / 2608.11080
RTSKG: Building a Rail Transit Station Knowledge Graph Dataset
RTSKG：构建铁路交通站知识图谱数据集
Abstract
Rail transit systems play a vital role in urban mobility and economic development. As key components of such systems, rail transit stations function as critical transport hubs that enhance urban accessibility and stimulate development in surrounding areas. City-level rail transit station related tasks (e.g., ridership prediction) require large-scale urban data, but current studies often neglect complex interactions among various urban entities in terms of data organization. In this paper, to address the above issue, we build a Rail Transit Station Knowledge Graph (RTSKG) dataset which explicitly models the spatial and semantic interactions among different kinds of urban entities, to benefit city-level rail transit station related tasks. RTSKG integrates heterogeneous urban entities, such as rail transit stations, road segments, and points of interest, with a specially designed unified schema, and is accessible as Linked Data at https://w3id.org/rtskg/. Evaluations on station-area store recommendation and knowledge-enhanced ridership prediction demonstrate the effectiveness of RTSKG, highlighting its potential to support city-level rail transit station analysis.
Chinese Translation
铁路交通系统在城市出行和经济发展中发挥着至关重要的作用。作为这些系统的关键组成部分，铁路交通站作为重要的交通枢纽，增强了城市的可达性，并刺激了周边地区的发展。城市级铁路交通站相关任务（例如，客流量预测）需要大规模的城市数据，但当前的研究往往忽视了各种城市实体之间复杂的交互关系。在本文中，为了解决上述问题，我们构建了一个铁路交通站知识图谱（RTSKG）数据集，该数据集明确建模了不同类型城市实体之间的空间和语义交互，以支持城市级铁路交通站相关任务。RTSKG整合了异构城市实体，如铁路交通站、道路段和兴趣点，采用特别设计的统一模式，并作为链接数据在 https://w3id.org/rtskg/ 上可访问。对站区商店推荐和知识增强的客流量预测的评估证明了RTSKG的有效性，突显了其支持城市级铁路交通站分析的潜力。
cs.AI / 77 / 2608.11095
Why Does CLAUDE.md Keep Growing? Catastrophic Remembering in Agentic Coding
为什么 CLAUDE.md 持续增长？代理编码中的灾难性记忆
Abstract
Agentic coding READMEs like CLAUDE.md grow without bound in real repositories, stopping only when the repository retires or someone rewrites the file wholesale. We trace this to imperfect recall: appending an instruction is always cheap, but once an instruction's rationale is gone, deleting it without risking a correctness regression costs O(2^|D|) in a prompt of |D| instructions. We name the resulting divergence catastrophic remembering, the inverse of catastrophic forgetting around which continual learning is organized. First, we characterize this phenomenon across 247,694 instruction lifetimes in 1,867 repositories: agentic prompts grow without bound, more than tripling over their lifetime (+226%), gaining +4.9 net instructions every commit; further, the older an instruction gets, the less likely it is to be deleted (log-hazard -0.032/commit). Then, we show that prompt comments can halt the growth: inverting IFEval yields verifiable worlds whose optimal prompts are known, and there comments encoding latent reasoning remove 99.3% of excess instructions (+211.3% to +1.4%). Finally, applying the same inversion to WildIFEval, we show that prompt comments can improve real-world agentic instruction-following by up to 23.1%. If English is the new code, why don't we have comments yet?
Chinese Translation
像 CLAUDE.md 这样的代理编码 README 在真实的代码库中无限增长，只有在代码库退役或有人完全重写文件时才会停止。我们将其归因于不完美的回忆：添加一条指令始终是便宜的，但一旦指令的理由消失，删除它而不冒着正确性回归的风险，在包含 |D| 条指令的提示中成本为 O(2^|D|)。我们将这种结果的偏差称为灾难性记忆，它是围绕持续学习组织的灾难性遗忘的逆过程。首先，我们在 1,867 个代码库中对 247,694 条指令的生命周期进行了特征描述：代理提示无限增长，其生命周期内增长超过三倍（+226%），每次提交净增加 4.9 条指令；此外，指令越老，被删除的可能性越小（对数风险 -0.032/提交）。然后，我们展示了提示注释可以阻止这种增长：反转 IFEval 产生可验证的世界，其最佳提示是已知的，而在其中编码潜在推理的注释移除了 99.3% 的多余指令（从 +211.3% 降至 +1.4%）。最后，将相同的反转应用于 WildIFEval，我们展示了提示注释可以将真实世界的代理指令跟随能力提高多达 23.1%。如果英语是新的代码，为什么我们还没有注释呢？
cs.AI / 78 / 2608.11136
sLTN: Structural Logic Tensor Networks
sLTN：结构逻辑张量网络
Abstract
Logic Tensor Networks (LTN) provide a neurosymbolic framework in which first-order logic is interpreted through tensor operations, enabling logical constraints to be integrated with differentiable learning. However, the original formulation of LTN is primarily suited to data represented as flat collections of individuals, and does not explicitly capture structural organization such as temporal order, sequential position, or graph connectivity. We introduce sLTN, an extension of LTN that makes structural dimensions first-class elements of the language. Structural dimensions represent named tensor axes associated with domain-specific organization, such as time steps, sequence positions, or graph nodes. They can be quantified explicitly, related through structural relations, and used to express temporal, sequential, and relational constraints directly at the logical level. We formalize the syntax and fuzzy tensor semantics of sLTN and show that, in the absence of structural dimensions, the framework recovers the original LTN semantics as a special case. We further describe a PyTorch implementation based on a declarative signature, formula parsing, and tensorial interpretation. The framework is illustrated on representative temporal and sequential reasoning examples. This paper serves as a companion to the sltn library, available at https://github.com/logictensornetworks/sltn.
Chinese Translation
逻辑张量网络（Logic Tensor Networks, LTN）提供了一种神经符号框架，通过张量运算对一阶逻辑进行解释，使逻辑约束能够与可微学习相结合。然而，LTN的原始形式主要适用于以个体的平面集合表示的数据，并未明确捕捉诸如时间顺序、序列位置或图连接等结构组织。我们提出了sLTN，这是LTN的一个扩展，使结构维度成为语言的第一类元素。结构维度表示与特定领域组织相关的命名张量轴，例如时间步、序列位置或图节点。它们可以被明确量化，通过结构关系相互关联，并用于在逻辑层面直接表达时间、序列和关系约束。我们形式化了sLTN的语法和模糊张量语义，并表明在缺乏结构维度的情况下，该框架恢复了原始LTN语义作为特例。我们进一步描述了基于声明性签名、公式解析和张量解释的PyTorch实现。该框架在代表性的时间和序列推理示例中进行了说明。本文作为sltn库的补充，库可在https://github.com/logictensornetworks/sltn获取。
cs.AI / 79 / 2608.11195
Long-Horizon AI Research for Grothendieck Constant: A Case Study in Human-AI Mathematical Collaboration
Abstract
AI agents are increasingly used in mathematics research, but it is often unclear how to use them effectively. Towards this, we present an extensive case study of how AI was used to improve bounds on the Grothendieck constant $K_G$, which captures the hardness between combinatorial problems and their continuous relaxations. Specifically, while the precise value of $K_G$ is not known, we recently tightened the best known bounds to \[ \frac{6\pi}{11} \;\le\; K_G \;\le\; \frac{\pi}{2\log(1+\sqrt2)} - 10^{-4}. \] Crucially, these improvements were achieved using an AI research system that could arrive at insights deemed novel by domain experts. We give a detailed discussion of our experience using AI for mathematics research, particularly touching upon its strengths and weaknesses, as well as our experience with creating ideal conditions for AI to arrive at breakthrough insights.
cs.CL / 1 / 2608.09934
LLM Agents Factory: Retrieval of Domain-Specific LLM Agents
LLM代理工厂：特定领域LLM代理的检索
Abstract
Large language model (LLM) agents improve task performance by decomposing problems into role-specialized behaviors. However, their practical deployment is often limited by the computational cost and instability associated with the on-the-fly agent design for each user request. To address this, we present LLM Agents Factory, a retrieval-based framework that constructs domain-specific and Wikipedia-grounded agents on demand using a base of over 20K predetermined agent profiles. Our framework supports two modes: (1) agent profile retrieval via semantic search and (2) distillation into a compact model fine-tuned for direct agent generation. Experiments on MMLU, BIG-bench, and BIG-bench Hard in a single-agent scenario demonstrate that our retrieval-based agent construction surpasses non-agent baselines in accuracy while matching AutoGen generation quality with a 120B backbone at a substantially lower inference cost. Our work reveals that retrieval from a structured agent repository provides a cost-efficient, accurate, and controllable alternative to dynamic agent generation, responding to the strict demands of industrial applications. We provide the implementation code and the agent base in https://huggingface.co/frontier-ai/llm-agent-factory.
Chinese Translation
大型语言模型（LLM）代理通过将问题分解为角色专门化的行为来提高任务性能。然而，它们的实际部署常常受到与每个用户请求的即时代理设计相关的计算成本和不稳定性的限制。为了解决这个问题，我们提出了LLM代理工厂，这是一个基于检索的框架，能够根据需求构建特定领域和以维基百科为基础的代理，使用超过20K个预先确定的代理配置文件作为基础。我们的框架支持两种模式：（1）通过语义搜索进行代理配置文件检索；（2）蒸馏为一个经过微调的紧凑模型，以便直接生成代理。在单代理场景下对MMLU、BIG-bench和BIG-bench Hard的实验表明，我们的基于检索的代理构建在准确性上超过了非代理基线，同时在推理成本显著降低的情况下，与120B主干的AutoGen生成质量相匹配。我们的研究表明，从结构化代理库中进行检索提供了一种成本效益高、准确且可控的替代方案，以应对动态代理生成的严格需求，满足工业应用的要求。我们在https://huggingface.co/frontier-ai/llm-agent-factory提供了实现代码和代理基础。
cs.CL / 2 / 2608.09936
Conflict or Strategy? Asymmetric Role Framing of La France insoumise and Rassemblement National in French News Headlines, 2022-2025
冲突还是策略？2022-2025年法国新闻标题中对无畏法国（La France insoumise）和国民集会（Rassemblement National）的非对称角色框架
Abstract
Do French news headlines frame left- and right-populist challengers as symmetric ``extremes,'' or as fundamentally different political adversaries? We examine 28,592 headlines about La France insoumise (LFI) and Rassemblement National (RN) published by 25 French-language outlets between 2022 and 2025, annotated through a three-model LLM pipeline validated against a stratified human audit. The clearest finding is role asymmetry rather than valence asymmetry: conflict framing and strategic-game framing are more robust across models and time than delegitimization, with AGGRESSOR serving as corroborating role syntax. LFI appears in headlines more often through a conflict register and RN through a strategic-electoral register. This role gap is direction-stable across all three annotation models, survives bootstrapping and permutation tests, and persists across outlet families and most of 2022-2025. A secondary moral-accounting layer (who is blamed, legitimized, or cast as a victim) is structured by outlet rather than party, producing aggregate nulls that conceal some of the corpus's most polarized patterns. Methodologically, the annotation pipeline reveals a two-tier reliability profile: conflict and strategic-game framing achieve the strongest human validation and cross-model stability; actor role is direction-stable but treated as corroborating because its audit reliability is lower; normative-judgment constructs (legitimacy, blame) are weaker. The paper contributes political-role assignment as a target for computational framing research that decomposes what valence-based measures conflate, and establishes a construct-stratified reliability framework for calibrating majority-vote LLM annotation pipelines in political text tasks.
Chinese Translation
法国新闻标题是否将左翼和右翼民粹主义挑战者框架为对称的“极端”，还是作为根本不同的政治对手？我们分析了2022年至2025年间25家法语媒体发布的28,592条关于无畏法国（LFI）和国民集会（RN）的标题，这些标题通过一个经过分层人工审核验证的三模型大型语言模型（LLM）管道进行标注。最明显的发现是角色不对称而非效价不对称：冲突框架和战略博弈框架在模型和时间上比去合法化更为稳健，其中AGGRESSOR作为支持性角色语法。LFI在标题中更常通过冲突框架出现，而RN则通过战略选举框架出现。这一角色差距在所有三个标注模型中方向稳定，经过自助法和置换测试仍然存在，并在各媒体类别和2022-2025年大部分时间内持续存在。一个次要的道德会计层面（谁被指责、被合法化或被视为受害者）由媒体而非政党结构化，产生的总和为零的结果掩盖了语料库中一些最极化的模式。在方法论上，标注管道揭示了一个两级可靠性特征：冲突和战略博弈框架实现了最强的人类验证和跨模型稳定性；行为者角色方向稳定，但由于其审核可靠性较低而被视为支持性；规范判断构造（合法性、指责）较弱。本文为计算框架研究提供了政治角色分配作为目标，分解了效价基础测量所混淆的内容，并建立了一个构造分层的可靠性框架，以校准政治文本任务中的多数投票LLM标注管道。
cs.CL / 3 / 2608.09937
Carefully Considering Culture: Analyzing LLM Alignment in Single- and Multi-Cultural Settings using Cultural Consensus Theory
仔细考虑文化：使用文化共识理论分析单文化和多文化环境中的大型语言模型对齐
Abstract
Recent work in NLP has probed large language models for their understanding of cultural norms across countries. However, this work typically considers distributional patterns, ignoring group consensus or possible multicultural environments within a country. In this work, we leverage cultural consensus theory (CCT) from cultural anthropology to model such multidimensional nuance. Applying CCT to the World Values Survey (WVS) across 10 countries and 12 domains, we demonstrate that models frequently misrepresent cultural structures by either failing to form cohesive consensus or severely over-regularizing consensus. Through explicit representation of intra-group variance, CCT provides actionable diagnostics to evaluate when models reflect true human diversity versus algorithmic homogenization.
Chinese Translation
近期在自然语言处理领域的研究探讨了大型语言模型对各国文化规范的理解。然而，这些研究通常只考虑分布模式，忽视了群体共识或一个国家内可能存在的多文化环境。在本研究中，我们借鉴文化人类学中的文化共识理论（CCT）来建模这种多维度的细微差别。通过将CCT应用于覆盖10个国家和12个领域的世界价值观调查（WVS），我们展示了模型常常错误地表现文化结构，要么未能形成一致的共识，要么严重过度规范化共识。通过对组内方差的明确表征，CCT提供了可操作的诊断工具，以评估模型何时反映真实的人类多样性与算法同质化之间的区别。
cs.CL / 4 / 2608.09941
The Multilingual Quantization Tax: Structural Collapse and Typological Fragility in Edge SLMs
多语言量化税：边缘小语言模型的结构崩溃与类型脆弱性
Abstract
While 4-bit weight quantization is critical for deploying Small Language Models (SLMs) on edge devices, evaluations of the resulting performance degradation-the quantization tax-remain overwhelmingly English-centric. We present a zero-shot multilingual evaluation of 4-bit quantization across the Gemma 4 and Qwen 3.5 architectures. Evaluating on eight typo-logically diverse languages using MMLU ProX Lite and GlobalPIQA, we show parameter truncation exposes deep pre-training inequalities. We identify four phenomena: (1) Typological Fragility: low-resource and specific non-Latin scripts suffer representational collapse via architecture-specific double dissociations, failing to generate valid task logits; (2) Home Language Fragility Paradox: foundational pre-training pathways provide limited precision loss protection; (3) Domain-Specific Forgetting: multi-step cross-lingual routing degrades while associative soft-science recall remains robust; and (4) Quantization Resistance: highly saturated, typologically aligned domains resist deterministic degradation, with post-quantization performance gains bounded by statistical noise.
Chinese Translation
虽然4位权重量化对于在边缘设备上部署小语言模型（SLMs）至关重要，但对由此导致的性能下降——量化税的评估仍然主要集中于英语。我们对Gemma 4和Qwen 3.5架构进行了零-shot多语言的4位量化评估。在使用MMLU ProX Lite和GlobalPIQA对八种类型多样的语言进行评估时，我们发现参数截断暴露了深度预训练的不平等。我们识别出四种现象：（1）类型脆弱性：低资源和特定非拉丁文字的脚本通过架构特定的双重解离遭受表征崩溃，无法生成有效的任务logits；（2）母语脆弱性悖论：基础预训练路径提供的精度损失保护有限；（3）领域特定遗忘：多步骤跨语言路由退化，而关联性软科学回忆保持稳健；（4）量化抗性：高度饱和、类型对齐的领域抵抗确定性退化，量化后的性能提升受到统计噪声的限制。
cs.CL / 5 / 2608.09942
When Chain-of-Thought Helps and When It Hurts: An Empirical Investigation of the Serial-Depth Bottleneck in LLM Reasoning
链式思维何时有助于推理，何时又有害：对大型语言模型推理中的串行深度瓶颈的实证研究
Abstract
It is widely assumed that chain-of-thought (CoT) prompting universally improves LLM reasoning. We investigate this through the conceptual framework of the H_dp bandwidth bound (Chen et al., 2024): although the formal bound binds only asymptotically (at astronomically large prompt lengths), it identifies a real architectural bottleneck -- serial computation exceeding a transformer's single-pass capacity must be externalised, which is what CoT does. Our central finding is a within-benchmark serial-depth gradient: single-pass (no-CoT) accuracy degrades monotonically with per-item serial depth, while CoT is approximately depth-invariant. We measure CoT effects across three instruction-tuned models (Qwen-2.5-7B/32B, Llama-3.1-8B) and five standard NLP benchmarks at practical context lengths. On high-depth P-complete tasks (GSM8K, MATH), CoT gives a +54 to +68 pp recovery gap across all models. On shallow TC^0 tasks (MMLU, ARC), CoT is structurally redundant (Delta in [0.0, +4.6] pp, no significant negative effect) -- though high no-CoT baselines (up to 95% on ARC) may reflect contamination, so this null is not a clean architectural test. The intermediate class L (HumanEval) shows a model-size-dependent transition: +23.2 pp (32B), +9.1 pp (8B), -28.7 pp (7B). The cross-benchmark depth-recovery correlation is Spearman rho = 0.661 (p = 0.007, n = 15); 9 of 15 benchmark-level McNemar tests are significant after Bonferroni correction. Pre-registered on OSF, our results indicate that CoT is not a universal reasoning enhancer but acts as a bandwidth bypass: it helps serial computation that strains single-pass capacity and is redundant for tasks that already fit.
Chinese Translation
人们普遍认为链式思维（Chain-of-Thought, CoT）提示可以普遍提高大型语言模型（LLM）的推理能力。我们通过H_dp带宽界限的概念框架（Chen et al., 2024）对此进行了研究：尽管该正式界限仅在天文级别的提示长度上渐近约束，但它识别出一个真实的架构瓶颈——超出变换器单次处理能力的串行计算必须被外部化，这正是CoT所做的。我们的主要发现是基准内的串行深度梯度：单次处理（无CoT）准确率随着每个项目的串行深度单调下降，而CoT在深度上大致不变。我们在三个经过指令调优的模型（Qwen-2.5-7B/32B，Llama-3.1-8B）和五个标准自然语言处理基准上测量了CoT的效果，使用实际的上下文长度。在高深度的P-完全任务（GSM8K，MATH）中，CoT在所有模型中提供了+54到+68个百分点的恢复差距。在浅层的TC^0任务（MMLU，ARC）中，CoT在结构上是多余的（Delta在[0.0, +4.6]个百分点之间，没有显著的负面影响）——尽管高无CoT基线（在ARC上高达95%）可能反映了污染，因此这一无效结果并不是一个干净的架构测试。中间类L（HumanEval）显示出依赖于模型大小的过渡：+23.2个百分点（32B），+9.1个百分点（8B），-28.7个百分点（7B）。跨基准的深度恢复相关性为Spearman rho = 0.661（p = 0.007，n = 15）；在Bonferroni校正后，15个基准级别的McNemar测试中有9个是显著的。我们的结果在OSF上预注册，表明CoT并不是一种普遍的推理增强器，而是作为带宽绕过：它有助于串行计算，这种计算超出单次处理能力，而对于已经适应的任务则是多余的。
cs.CL / 6 / 2608.10021
Position Encoding in Transformers: From Absolute and Relative Methods to Rotary Position Embeddings and Long-Context Scaling
变换器中的位置编码：从绝对和相对方法到旋转位置嵌入及长上下文扩展
Abstract
Self-attention models content-dependent interactions between tokens but does not by itself encode token order. Position encoding addresses this limitation by introducing absolute coordinates, relative distances, or position-dependent rotations into Transformer representations and attention scores. This technical survey develops a unified account of sinusoidal and learned absolute position embeddings, Shaw-style relative position representations, Transformer-XL, T5 relative position bias, ALiBi, and Rotary Position Embeddings (RoPE). We derive how RoPE converts absolute position indices into relative phase differences in Query-Key inner products and compare these methods in terms of where position is injected, computational cost, compatibility with KV caching, and length extrapolation. We then examine long-context extensions, including Position Interpolation, RoPE scaling laws, NTK-aware scaling, Dynamic NTK, NTK-by-parts, YaRN, LongRoPE, and LongRoPE2, with emphasis on frequency allocation, attention rescaling, training length, and target context length. We also summarize implementation considerations, evaluation protocols, and position-encoding choices in representative large language models. A central conclusion is that the ability to compute positional features beyond the training length does not imply reliable long-context generalization; context extension must be evaluated through short-context retention, position-wise perplexity, retrieval, reasoning, and long-context code tasks.
Chinese Translation
自注意力模型处理令牌之间的内容依赖交互，但本身并不编码令牌顺序。位置编码通过将绝对坐标、相对距离或位置依赖的旋转引入变换器表示和注意力分数来解决这一限制。本文技术调查对正弦波和学习的绝对位置嵌入、Shaw风格的相对位置表示、Transformer-XL、T5相对位置偏置、ALiBi和旋转位置嵌入（RoPE）进行了统一的阐述。我们推导了RoPE如何将绝对位置索引转换为查询-键内积中的相对相位差，并在位置注入位置、计算成本、与KV缓存的兼容性以及长度外推等方面比较了这些方法。随后，我们考察了长上下文扩展，包括位置插值、RoPE缩放规律、NTK感知缩放、动态NTK、分段NTK、YaRN、LongRoPE和LongRoPE2，重点关注频率分配、注意力重缩放、训练长度和目标上下文长度。我们还总结了在代表性的大型语言模型中的实现考虑、评估协议和位置编码选择。一个核心结论是，计算超出训练长度的位置信息的能力并不意味着可靠的长上下文泛化；上下文扩展必须通过短上下文保留、位置困惑度、检索、推理和长上下文代码任务进行评估。
cs.CL / 7 / 2608.10109
PERCEPT: A Corpus for POS Tagging and Analysis of Persian-English Code-Mixing
PERCEPT：一个用于波斯语-英语代码混合的词性标注与分析语料库
Abstract
Social media has become a major venue for multilingual communication, where users frequently mix multiple languages within a single utterance. Although code-mixed corpora have been developed for several language pairs, Persian-English code-mixing remains relatively underexplored. Existing Persian resources lack Universal Dependencies (UD) part-of-speech (POS) annotations for code-mixed words, limiting both linguistic analyses and the development of syntax-aware NLP models. To address this gap, we introduce PERCEPT, the first publicly available large-scale Persian-English code-mixed corpus annotated with Universal Dependencies POS tags for code-mixed words. The dataset comprises 6,800 posts collected from X, Instagram, and Digikala. We further present an LLM-assisted annotation framework that automatically assigns POS tags and document-level topics. Human evaluation demonstrates high agreement between the automatically generated annotations and gold annotations, confirming the reliability of the annotations. Using PERCEPT, we conduct the first comprehensive linguistic analysis of Persian-English code-mixing across multiple social media platforms. Our analyses reveal that nouns are the predominant category for code-mixed words, while the distributions of other POS categories vary across platforms. We further find that the positional distribution of code-mixed words is remarkably consistent across platforms, whereas the triggering effect is substantially more pronounced in Digikala. PERCEPT is publicly available at https://github.com/kalhorghazal/PERCEPT.
Chinese Translation
社交媒体已成为多语言交流的重要场所，用户在单一话语中经常混合多种语言。尽管已有针对多个语言对的代码混合语料库，但波斯语-英语代码混合仍然相对未被充分探索。现有的波斯语资源缺乏针对代码混合词的通用依赖（Universal Dependencies, UD）词性（POS）标注，这限制了语言学分析和语法感知自然语言处理（NLP）模型的开发。为了解决这一问题，我们推出了PERCEPT，这是第一个公开可用的大规模波斯语-英语代码混合语料库，针对代码混合词进行了通用依赖词性标注。该数据集包含从X、Instagram和Digikala收集的6800条帖子。我们进一步提出了一种基于大型语言模型（LLM）的标注框架，该框架自动分配词性标记和文档级主题。人工评估表明，自动生成的标注与金标准标注之间具有高度一致性，确认了标注的可靠性。利用PERCEPT，我们首次对多个社交媒体平台上的波斯语-英语代码混合进行了全面的语言学分析。我们的分析显示，名词是代码混合词的主要类别，而其他词性类别的分布在不同平台上有所不同。我们还发现，代码混合词的位置信息在各个平台上表现出显著的一致性，而在Digikala上的触发效应则明显更为突出。PERCEPT可在https://github.com/kalhorghazal/PERCEPT获取。
cs.CL / 8 / 2608.10137
The Parser Already Knows: Lightweight Bias Correction in Constrained Decoding
解析器已知：约束解码中的轻量级偏差校正
Abstract
Grammar Constrained Decoding (GCD) forces Language Models (LMs) to produce syntactically valid outputs by masking out non-conforming tokens at each step. However, rigid masking distorts the model's underlying probability distribution, often biasing generation toward valid but suboptimal outputs. While online sampling restores this distribution, it requires computationally expensive iterative resampling. As a result, existing methods force a compromise between output quality and inference latency. Our key insight is that the internal parser and lexer states inherently maintained during incremental parsing already encode future grammatical validity -- exactly the information required to restore the LM's true distribution. We propose a lightweight, offline-trained logit correction conditioned on this syntactic and lexical state together with candidate next tokens. Because these states are already computed as a necessary part of incremental parsing for masking, extracting them adds negligible overhead while leaving the base LM's weights completely untouched. Across several grammars, this correction substantially closes the gap between the masked distribution and the LM's true distribution, consistently outperforming both masking and online sampling. Even its lightest variant, which relies on the candidate next token alone, still matches or exceeds both baselines: the next token itself carries an implicit lookahead, much like how parsers commonly use a lookahead token to resolve ambiguous decisions. By restoring the probability mass that masking removes, it reconciles the LM's probabilistic integrity with grammar conformance.
Chinese Translation
语法约束解码（Grammar Constrained Decoding, GCD）通过在每一步屏蔽不符合语法的标记，强制语言模型（Language Models, LMs）生成语法有效的输出。然而，严格的屏蔽扭曲了模型的基础概率分布，常常使生成偏向于有效但次优的输出。尽管在线采样可以恢复这种分布，但需要计算上昂贵的迭代重采样。因此，现有方法在输出质量和推理延迟之间迫使妥协。我们的关键见解是，增量解析过程中内置的解析器和词法分析器状态本质上已经编码了未来的语法有效性——这正是恢复语言模型真实分布所需的信息。我们提出了一种轻量级的、离线训练的对数值校正，基于这种语法和词汇状态以及候选下一个标记进行条件校正。由于这些状态在屏蔽的增量解析过程中已经计算，因此提取它们所增加的开销微乎其微，同时完全不影响基础语言模型的权重。在多个语法下，这种校正显著缩小了屏蔽分布与语言模型真实分布之间的差距，始终优于屏蔽和在线采样。即使是其最轻量的变体，仅依赖于候选下一个标记，仍然与两个基线相匹配或超越：下一个标记本身携带隐含的前瞻性，类似于解析器通常使用前瞻标记来解决模糊决策。通过恢复屏蔽所移除的概率质量，它调和了语言模型的概率完整性与语法一致性。
cs.CL / 9 / 2608.10154
Multimodal Item Parameter Estimation using Simulated Response Probabilitie
基于模拟响应概率的多模态项目参数估计
Abstract
We present results from reconstructing multiple-choice model (MCM) and three-parameter logistic (3PL) model curves using a fine-tuned multimodal large language model (LLM) based on Qwen3.5. The model is prompted and fine-tuned to replicate choice probabilities across a large training corpus of multiple-choice items containing both image and text stimuli, conditioned on a labeled set of student ability levels. By learning to reproduce the systematic error patterns of students across a discrete range of abilities, the LLM implicitly captures the underlying response probabilities encoded in the 3PL and MCM curves. This allows us to accurately approximate item difficulty on a held-out test set directly from the model's predicted option probabilities.
Chinese Translation
我们展示了使用基于 Qwen3.5 的精细调优多模态大型语言模型 (LLM) 重建多项选择模型 (MCM) 和三参数逻辑模型 (3PL) 曲线的结果。该模型经过提示和精细调优，以复制包含图像和文本刺激的大量多项选择项目的选择概率，并以标记的学生能力水平集为条件。通过学习在离散能力范围内重现学生的系统性错误模式，LLM 隐式捕捉了编码在 3PL 和 MCM 曲线中的潜在响应概率。这使我们能够直接从模型预测的选项概率中准确估计保留测试集上的项目难度。
cs.CL / 10 / 2608.10216
Similarity Gates Approve Reversals: A Validity Audit of Embedding-Cosine Thresholds in Agent Systems
相似性门批准反转：代理系统中嵌入-余弦阈值的有效性审计
Abstract
Agent frameworks ship quality gates that compare text blocks by embedding-cosine similarity and decide at a fixed cutoff. Deduplication filters, semantic caches, drift guards, and answer grader gates deploy to answer the question: "Does this text still mean the same thing?" But the score answers a different question: "How much did the wording change?" We audit this gate class as a measurement instrument. In the cases these gates exist to catch, the two can run in opposite ways. Many times, reversing an instruction is a single word edit, while agreement often rephrases a sentence. The consequence is a safety check that fires backwards. The production drift guard we audited caught 0 of 56 meaning-breaking mutations, and one approved item, "withhold the study drug" -> "administer the study drug", came in at cosine 0.9608. We observed five shipped operating points, and balanced accuracy across 90 configuration-threshold-task cells never exceeded 0.700 (median 0.525). The same confounder also corrupted evaluations. A naively built corpus inherits this confounder and can return an inverted verdict, with a decision AUROC exactly 0.000 in 13 of 18 configuration-task cells (at most 0.040 in all 18) against 0.440-0.815 for the same nine configurations under a balanced 2x2 design. Twice in the effort it captured our own headline claims. Obvious repairs fail: an encoder swap and an overlap-conditioned gate (0.750 in-sample, 0.533 held-out) land at chance on separately authored held-out data, and an NLI drop-in did no better. Embeddings do still bear hope here, as the strongest two of nine configurations separated reversal from paraphrase at matched overlap (AUROC 0.79-0.90), but only a matched-pair audit reveals the deployment regime. We release the corpus method, harness, and frozen results, and contend that scores gated this way measure the wrong thing. We believe a valid instrument is buildable.
Chinese Translation
代理框架提供质量门，通过嵌入-余弦相似度比较文本块，并在固定的截止点做出决定。去重过滤器、语义缓存、漂移保护器和答案评分门被部署以回答这个问题：“这段文本仍然意味着同样的事情吗？”但得分回答的是一个不同的问题：“措辞变化了多少？”我们对这一门类作为测量工具进行了审计。在这些门存在的情况下，二者可能会朝相反的方向运行。许多时候，反转指令仅需一个单词的编辑，而同意通常会重述一个句子。其结果是一个向后触发的安全检查。我们审计的生产漂移保护器未能捕捉到56个意义破坏突变中的任何一个，而一个被批准的项目“保留研究药物”->“施用研究药物”的余弦值为0.9608。我们观察到五个已发布的操作点，90个配置-阈值-任务单元的平衡准确率从未超过0.700（中位数0.525）。同样的混淆因素也破坏了评估。一个简单构建的语料库继承了这一混淆因素，并可能返回一个相反的裁决，在18个配置-任务单元中有13个的决策AUROC恰好为0.000（在所有18个中最多为0.040），而在平衡的2x2设计下，同样的九个配置的AUROC为0.440-0.815。在这项工作中，它两次捕捉到了我们自己的标题主张。显而易见的修复措施失败：一个编码器交换和一个重叠条件门（样本内0.750，持出0.533）在单独创作的持出数据上表现为随机，而NLI的替代方案也没有更好的效果。嵌入在这里仍然寄托希望，因为九个配置中最强的两个在匹配重叠下将反转与释义区分开（AUROC 0.79-0.90），但只有匹配对审计揭示了部署机制。我们发布了语料库方法、工具和冻结结果，并认为以这种方式门控的得分测量的是错误的内容。我们相信可以构建一个有效的工具。
cs.CL / 11 / 2608.10251
Off-Axis, On Purpose: Where a Transformer Computes Concepts and Why it Does So
离轴，故意为之：变换器计算概念的位置及其原因
Abstract
A transformer's answer lives on one axis: the direction its unembedding reads. Its intermediate states largely do not, and that off-axis position is usually treated as an obstacle to interpretation. We show it is functional. A 12-layer model computes in two phases. Through the first, every sublayer writes into a subspace held near-orthogonal to the read-out, attention 75 to 96 degrees off it at every depth. Moving attention's values onto the read-out is 64 to 84 times more damaging than a matched random rotation, and the damage is entirely in cross-token mixing: the subspace insulates composition from the vocabulary. Beneath it the frame itself turns rigidly with depth. In the second phase the answer arrives on-axis, late, and by addition rather than by turning accumulated content onto the read-out. Pressing every layer onto the read-out instead, as training for early exit does, matches the baseline on perplexity, LAMBADA and BLiMP while cutting the concept-phase workspace from about twenty-five effective dimensions to fourteen, a change none of those benchmarks register. The geometry can also be imposed, though not by asking for it. Prescribing it through the loss is a lottery: six of eight seeds collapse, because a model told to null its read-out projection obeys most cheaply by discarding dimensions. Inserting one fixed rotation at the phase boundary lands it instead, at baseline quality. A sparse rotation the surrounding weights can absorb converges on all nine seeds, against five of nine for ordinary training. Which rotation is immaterial: twenty-five runs across thirteen distinct ones reach the same quality, and two baselines from different seeds hold their concepts in near-orthogonal frames while agreeing on their read-outs. That freedom is usable: a basis drawn at random and prescribed before training is adopted across the concept phase, with quality unchanged.
Chinese Translation
变换器的答案位于一个轴上：其去嵌入的读取方向。其间接状态在很大程度上并不如此，而这种离轴位置通常被视为解释的障碍。我们展示了它的功能性。一个12层的模型分两个阶段进行计算。在第一阶段，每个子层写入一个与读取方向近正交的子空间，在每个深度上，注意力的角度偏离读取方向75到96度。将注意力的值移动到读取方向上，比匹配的随机旋转造成的损害要高出64到84倍，而这种损害完全体现在跨标记混合上：子空间使得组合与词汇表隔离。在它的下面，框架本身随着深度的增加而变得刚性。在第二阶段，答案在轴上到达，延迟，并通过加法而不是将累积内容转向读取方向来实现。将每一层强制推向读取方向，正如早期退出的训练所做的那样，在困惑度、LAMBADA和BLiMP上匹配基线，同时将概念阶段的工作空间从约25个有效维度减少到14个，而这一变化在这些基准中没有被记录。几何形状也可以被施加，尽管不是通过直接要求。通过损失来规定它是一场彩票：八个种子中有六个崩溃，因为一个被告知要使其读取投影无效的模型最便宜的方式是丢弃维度。在相位边界插入一个固定旋转则能使其落入基线质量。周围权重可以吸收的稀疏旋转在所有九个种子上收敛，而普通训练则仅在九个种子中收敛五个。哪个旋转并不重要：在十三个不同的旋转中进行的二十五次实验达到了相同的质量，而来自不同种子的两个基线在近正交框架中保持其概念，同时在读取方向上达成一致。这种自由是可用的：在训练前随机抽取并规定的基底在概念阶段被采用，质量保持不变。
cs.CL / 12 / 2608.10258
TAF-MED: Multi-Turn Safety Refusal Collapse in LLMs Under Declared Self-Treatment Intent
TAF-MED：在声明自我治疗意图下，大型语言模型中的多轮安全拒绝崩溃
Abstract
Large language models (LLMs) increasingly provide conversational health information that may influence treatment decisions, yet existing benchmarks do not isolate whether medication-safety boundaries persist across follow-ups after explicit self-treatment intent. We introduce TAF-MED, a physician-reviewed benchmark of 500 fixed three-turn scenarios, and evaluate eight LLMs across 4,000 conversations. A rubric-based automated judge labelled responses as SAFE, LEAKY, or UNSAFE, and two physicians independently annotated a model-balanced random subset of 400 conversations. We assessed unsafe guidance, collapse after a strictly SAFE initial response, and model-ranking stability. Overall, 71.6% of conversations contained an UNSAFE response, and 61.4% of those beginning with a strictly SAFE response later collapsed to UNSAFE; model-level collapse rates ranged from 24.4% to 96.2%. Four of 28 model pairs reversed order between initial unsafe and collapse rates. Automated labels achieved 94.3% agreement with the adjudicated physician reference ($\kappa = 0.895$). These findings show that first-turn safety is an incomplete proxy for conversational safety persistence and motivate evaluation across complete dialogue trajectories. We will release TAF-MED on Hugging Face to support reproducible research on multi-turn medical safety.
Chinese Translation
大型语言模型（LLMs）越来越多地提供可能影响治疗决策的对话健康信息，但现有基准未能明确药物安全边界在明确自我治疗意图后的后续对话中是否依然存在。我们引入了TAF-MED，这是一个经过医生审查的基准，包含500个固定的三轮场景，并对8个LLMs在4000个对话中进行了评估。基于评分标准的自动评判者将响应标记为安全（SAFE）、泄漏（LEAKY）或不安全（UNSAFE），并且两位医生独立标注了400个模型平衡的随机子集对话。我们评估了不安全指导、在严格安全的初始响应后崩溃的情况以及模型排名的稳定性。总体而言，71.6%的对话包含不安全的响应，61.4%的对话在以严格安全的响应开始后，随后崩溃为不安全；模型级崩溃率范围从24.4%到96.2%。在28对模型中，有4对在初始不安全和崩溃率之间的顺序发生了反转。自动标签与裁定医生参考的协议一致性达94.3%（$ ext{kappa} = 0.895$）。这些发现表明，第一轮的安全性并不能完全代表对话安全性的持续性，并促使我们在完整对话轨迹中进行评估。我们将在Hugging Face上发布TAF-MED，以支持多轮医疗安全的可重复研究。
cs.CL / 13 / 2608.10273
Locally Deployable Small Language Models for Emergency Department Decision Support: A Systematic Benchmark of Fine-Tuning Strategies
可在本地部署的小型语言模型用于急诊科决策支持：微调策略的系统基准评估
Abstract
Deploying large language models (LLMs) for decision support in emergency departments (EDs) faces two major challenges: privacy risks of transmitting patient data to closed-source commercial LLMs and the lack of systematic evaluation of fine-tuning strategies for locally deployable open-source small language models (SLMs). We benchmarked eight open-source SLMs using zero-shot prompting, prefix tuning, Low-Rank Adaptation (LoRA), and full fine-tuning on three ED tasks: triage level prediction, specialist referral recommendation, and diagnosis prediction. Using 2,083 MIMIC-IV-ED cases and Claude Haiku 4.5 and Claude Sonnet 4.5 as baselines, we found that LoRA fine-tuned open-source SLMs outperform commercial baselines on triage level prediction and specialist referral recommendation, while diagnosis prediction remains challenging for open-source SLMs. Confusion matrix analysis further shows that fine-tuned open-source SLMs can detect highest-severity patients missed by the commercial baselines. These results demonstrate that locally deployable SLMs can achieve clinically competitive performance for ED decision support.
Chinese Translation
在急诊科（ED）中部署大型语言模型（LLMs）以支持决策面临两个主要挑战：将患者数据传输到闭源商业LLMs的隐私风险，以及缺乏对可在本地部署的开源小型语言模型（SLMs）微调策略的系统评估。我们对八种开源SLMs进行了基准测试，采用了零-shot提示、前缀调优、低秩适应（LoRA）和完全微调，针对三个急诊科任务：分诊级别预测、专家转诊推荐和诊断预测。使用2,083个MIMIC-IV-ED案例以及Claude Haiku 4.5和Claude Sonnet 4.5作为基线，我们发现LoRA微调的开源SLMs在分诊级别预测和专家转诊推荐方面优于商业基线，而诊断预测对开源SLMs仍然具有挑战性。混淆矩阵分析进一步表明，微调后的开源SLMs能够检测到商业基线漏掉的最高严重性患者。这些结果表明，可在本地部署的SLMs能够在急诊科决策支持中实现临床竞争力的表现。
cs.CL / 14 / 2608.10296
Cracks in the Foundation: Seemingly Minor Architectural Choices Impact Long Context Extension
基础中的裂缝：看似微小的建筑选择影响长上下文扩展
Abstract
One might imagine that architectural variations within the dense transformer paradigm have a limited effect on accuracy. However, we demonstrate that this is not the case in the long context setting. Specifically, we show that a set of four minor architectural decisions --- all made by at least one of the Olmo, Llama, and Qwen dense model families --- have a compoundingly negative effect on long context extensibility. Any one of these choices alone has a minor impact on long context performance, but combining three or more can drop the performance downstream by up to 47%. Furthermore, these differences are not detectable from short-context loss or validation datasets. We show that much of the variation in long context ability across model families is driven by these architectural features and detectable from applying context extension early in pretraining. We demonstrate this with controlled ablations that hold data, tokenizer, and extension recipe fixed while varying normalization, GQA, pretraining context length, and sliding window attention. After over 170,000 GPU hours of training, we release the resulting set of models as OlmPool, a set of 26 comparable 7B models with checkpoints before and after long-context extension. This pool includes several architectures that outperform the Llama 3 architecture on long context extensibility. In an analysis of our ablation models, we identify patterns in attention sink behavior and attention distributions across context that are attributable to specific architectural differences.
Chinese Translation
人们可能会认为，在密集变压器范式中的建筑变体对准确性的影响有限。然而，我们证明在长上下文设置中情况并非如此。具体而言，我们展示了一组四个微小的建筑决策——这些决策均由 Olmo、Llama 和 Qwen 密集模型家族中的至少一个做出——对长上下文的可扩展性产生了复合的负面影响。单独考虑这些选择中的任何一个对长上下文性能的影响都很小，但将三个或更多的选择结合起来，可能会使下游性能下降多达 47%。此外，这些差异在短上下文损失或验证数据集中是无法检测到的。我们表明，不同模型家族在长上下文能力上的大部分变异是由这些建筑特征驱动的，并且可以通过在预训练早期应用上下文扩展来检测到。我们通过控制消融实验来证明这一点，在这些实验中，数据、分词器和扩展方案保持不变，同时改变归一化、GQA、预训练上下文长度和滑动窗口注意力。在超过 170,000 小时的 GPU 训练后，我们发布了结果模型集 OlmPool，这是一个包含 26 个可比较的 7B 模型的集合，具有长上下文扩展前后的检查点。该模型库包括几种在长上下文可扩展性上超越 Llama 3 架构的架构。在对我们的消融模型的分析中，我们识别出注意力沉没行为和上下文中注意力分布的模式，这些模式可归因于特定的建筑差异。
cs.CL / 15 / 2608.10299
Co-Evolution in Agentic Systems: Toward Self-Directed Evolution Beyond Human Design
代理系统中的共演化：超越人类设计的自我导向进化
Abstract
Agentic systems are increasingly expected to improve after deployment, yet single-entity self-evolution is often bounded by a static learning context, such as fixed tasks and feedback. This survey focuses on co-evolution in agentic systems, a multi-component form of self-evolution in which multiple agents and their environment impose adaptive pressure on one another. To organize existing papers, we propose a progressive three-stage taxonomy that traces how the system gradually sheds human-engineered constraints. Agent--Agent Co-Evolution studies how agents adapt through dynamic peers, including adversarial, collaborative, and organizational adaptation. Agent--Environment Co-Evolution extends this loop to adaptive tasks, feedback, and interaction spaces that change with the agents. Meta Co-Evolution further explores the possibility of making the evolution mechanism itself evolvable. We also discuss open challenges in evaluating such systems, scaling them across multiple components, and keeping increasingly autonomous evolutionary processes safe and controllable. This survey provides a unified foundation for building robust and open-ended agentic systems that can improve beyond fixed human-designed paths.
Chinese Translation
代理系统越来越被期望在部署后能够持续改进，但单一实体的自我进化通常受到静态学习环境的限制，例如固定的任务和反馈。本文调查了代理系统中的共演化，这是一种多组件的自我进化形式，其中多个代理及其环境相互施加适应压力。为了整理现有文献，我们提出了一种渐进的三阶段分类法，追踪系统如何逐渐摆脱人类工程约束。代理-代理共演化研究代理如何通过动态的同行进行适应，包括对抗性、协作性和组织适应。代理-环境共演化将这一循环扩展到适应性任务、反馈和随着代理变化的互动空间。元共演化进一步探讨了使进化机制本身可进化的可能性。我们还讨论了在评估此类系统、跨多个组件扩展它们以及保持日益自主的进化过程安全可控方面的开放挑战。本文为构建能够超越固定人类设计路径的强大且开放的代理系统提供了统一的基础。
cs.CL / 16 / 2608.10315
Is This Your Final Answer? Cross-Contextual Consistency as a Measure of LLM Credibility
这是你的最终答案吗？跨上下文一致性作为大型语言模型可信度的衡量标准
Abstract
Large language models (LLMs) are powerful black-box systems, making it difficult to discern whether their answers reflect stable internal beliefs or superficial pattern matching. We identify cross-contextual consistency as an underutilized behavioral property of LLMs: a credible answer should remain stable when the same task is placed under topic-aligned, content-neutral contextual variation. Building on this intuition, we operationalize Cross-Contextual Consistency (C3) by comparing model generations under original and perturbed prompts. Across 26 models and six benchmarks spanning reasoning, factuality, and code generation, we find that answers with smaller cross-contextual shifts are more likely to be correct or factual. We demonstrate that C3 provides a complementary axis of evaluation and can serve as a benchmark usefulness diagnostic, identifying which portions of a benchmark remain informative even when aggregated scores are widely considered "saturate".
Chinese Translation
大型语言模型（LLMs）是强大的黑箱系统，使得很难判断它们的回答是否反映了稳定的内部信念或表面的模式匹配。我们将跨上下文一致性识别为LLMs的一种未充分利用的行为特性：一个可信的答案在相同任务的主题一致、内容中立的上下文变化下应保持稳定。在这一直觉的基础上，我们通过比较模型在原始和扰动提示下的生成结果来操作化跨上下文一致性（C3）。在涵盖推理、事实性和代码生成的26个模型和六个基准测试中，我们发现具有较小跨上下文变化的答案更可能是正确或真实的。我们证明C3提供了一种补充的评估轴线，并可以作为基准有效性诊断，识别出即使在聚合分数被广泛认为“饱和”的情况下，基准的哪些部分仍然具有信息价值。
cs.CL / 17 / 2608.10408
VisEditBench: Can Vision-Language Models Edit Visualization Code from Multimodal Feedback?
VisEditBench：视觉-语言模型能否根据多模态反馈编辑可视化代码？
Abstract
Vision-language models (VLMs) have shown strong capabilities in generating visualization code from textual or visual specifications. However, real-world visualization authoring is inherently iterative: users frequently revise existing visualizations to repair flawed charts or adapt them to desired styles. Existing benchmarks primarily evaluate generation from scratch, leaving visualization code editing from multimodal feedback largely unexplored. We introduce VisEditBench, a benchmark of 1,395 human-annotated visualization code-editing tasks grounded in realistic visualization workflows and failure cases. VisEditBench covers two practical settings: feedback-guided repair, where models revise visualization code using buggy or marked charts together with textual feedback, and reference-guided restyling, where models modify code to match a target chart image. Evaluating 20 state-of-the-art VLMs reveals that visualization code editing remains challenging: Claude-4.6-Sonnet achieves the best overall pass rate of 74.46%, while most open-source models remain below 50%. Performance is particularly weak on visually grounded style adaptation, where Claude-4.6-Sonnet achieves only 55.71%. To establish a strong baseline, we further propose VisEditAgent, a render-grounded editing framework that iteratively generates, executes, validates, and refines candidate edits. Built on GPT-4o, VisEditAgent improves overall pass rate from 55.75% to 67.99%, demonstrating the importance of render-grounded feedback for faithful visualization editing. We will release VisEditBench at https://github.com/vis-nlp/VisEditBench.
Chinese Translation
视觉-语言模型（VLMs）在根据文本或视觉规范生成可视化代码方面展现了强大的能力。然而，现实世界中的可视化创作本质上是迭代的：用户经常修订现有的可视化，以修复有缺陷的图表或将其调整为所需的风格。现有基准主要评估从头开始的生成，导致基于多模态反馈的可视化代码编辑尚未得到充分探索。我们引入了VisEditBench，这是一个基于现实可视化工作流程和失败案例的1,395个人工标注的可视化代码编辑任务的基准。VisEditBench涵盖了两种实际设置：反馈引导的修复，其中模型使用有缺陷或标记的图表以及文本反馈来修订可视化代码；参考引导的样式重塑，其中模型修改代码以匹配目标图表图像。对20个最先进的VLM进行评估表明，可视化代码编辑仍然具有挑战性：Claude-4.6-Sonnet实现了74.46%的最佳整体通过率，而大多数开源模型的通过率仍低于50%。在视觉基础的样式适应方面，性能尤其薄弱，Claude-4.6-Sonnet仅实现了55.71%。为了建立一个强有力的基准，我们进一步提出了VisEditAgent，这是一个基于渲染的编辑框架，能够迭代生成、执行、验证和优化候选编辑。基于GPT-4o，VisEditAgent将整体通过率从55.75%提高到67.99%，证明了基于渲染的反馈对忠实可视化编辑的重要性。我们将在https://github.com/vis-nlp/VisEditBench发布VisEditBench。
cs.CL / 18 / 2608.10414
How Robust Are LLMs to Vietnamese Dialects?
大型语言模型对越南方言的鲁棒性如何？
Abstract
Large Language Models (LLMs) are typically evaluated on standard written Vietnamese, yet everyday communication frequently involves regional dialects that preserve meaning but differ in surface form. Existing Vietnamese dialect work largely addresses this issue through dialect-to-standard normalization instead of measuring how the model fails under Vietnamese dialectal inputs. To address this gap, we present the first systematic evaluation of LLM robustness to Vietnamese dialect variation across multiple tasks, quantifying performance degradation and failure patterns. We introduce VialectBench (Vietnamese Dialects Benchmarking), a controlled benchmark for testing whether model decisions remain stable across six Vietnamese dialect groups. VialectBench contains 400 Standard Vietnamese source instances and 2,400 human-written dialectal rewrites spanning emotion recognition (ER), natural language inference (NLI), question answering (QA), and multiple-choice question answering (MCQA). Dataset evaluation with a fixed reference language model shows that the dialectal rewrites induce a measurable model-relative likelihood shift while remaining nearly equal in length to their Standard counterparts. Across ten instruction-tuned models, dialectal inputs reduce average performance by 2.82%, and no evaluated model is fully dialect-invariant. All four tasks are affected, with QA showing the largest average degradation. Robustness also varies substantially across dialect groups: PNT3 and PNT2 cause the largest average performance drops, at 6.17% and 4.73%, respectively, whereas PNB slightly improves average performance by 0.42%. The Central dialect group (PNT1-PNT4) also yields the highest average harmful-flip rate across all models, at 6.54%. These findings show that strong performance on Standard Vietnamese does not guarantee reliable behavior under meaning-preserving regional variation.
Chinese Translation
大型语言模型（LLMs）通常在标准书面越南语上进行评估，但日常交流中常常涉及保留意义但在表面形式上有所不同的区域方言。现有的越南方言研究主要通过方言到标准语的规范化来解决这一问题，而不是衡量模型在越南方言输入下的失败情况。为了解决这一空白，我们首次系统性地评估了大型语言模型对越南方言变异的鲁棒性，涵盖多个任务，量化性能下降和失败模式。我们引入了VialectBench（越南方言基准测试），这是一个受控基准，用于测试模型决策在六个越南方言组之间是否保持稳定。VialectBench包含400个标准越南语源实例和2400个人工编写的方言重写，涵盖情感识别（ER）、自然语言推理（NLI）、问答（QA）和多项选择问答（MCQA）。使用固定参考语言模型进行的数据集评估显示，方言重写引起了可测量的模型相对可能性变化，同时与其标准对应物的长度几乎相等。在十个经过指令调优的模型中，方言输入使平均性能下降了2.82%，且没有评估的模型是完全方言不变的。所有四个任务都受到影响，其中问答任务显示出最大的平均下降。鲁棒性在方言组之间也有显著差异：PNT3和PNT2导致的平均性能下降最大，分别为6.17%和4.73%，而PNB则略微提高了平均性能0.42%。中央方言组（PNT1-PNT4）在所有模型中也产生了最高的平均有害翻转率，达到6.54%。这些发现表明，在标准越南语上表现良好并不保证在保留意义的区域变异下的可靠行为。
cs.CL / 19 / 2608.10444
From Reasoning Depth to Reasoning Breadth: Evaluating Multi-Point Associative Reasoning in Large Language Models
从推理深度到推理广度：评估大型语言模型中的多点关联推理
Abstract
Large language models (LLMs) have made substantial progress on reasoning tasks that require increasingly long and complex inferential chains. This progress primarily reflects reasoning depth. A complementary and comparatively unexamined capability is reasoning breadth: exploring multiple semantic directions in parallel and integrating the resulting clues into one coherent answer. We introduce MPAR-Bench, a bilingual English-Chinese benchmark that isolates reasoning breadth through multi-point associative reasoning. Inspired by the cooperative game Just One, each item asks a model to recover a hidden target from several independently generated, semantically diverse clues. We construct 1,000 items using a multi-agent clue-generation pipeline, embedding-based diversity filtering, and human verification. Only the answer space is drawn from public word lists, whereas every clue set is generated from scratch. Beyond exact-match accuracy, we evaluate models using accuracy, ANLS, embedding similarity, reasoning-trace verification, and four perturbations: clue masking, order shuffling, distractor injection, and multi-step clues. Across evaluated models, perturbations reduce accuracy by 9-18 percentage points in English and 5-12 percentage points in Chinese. Thinking mode improves standard-setting accuracy, especially in English, but does not consistently reduce sensitivity to perturbations. Case-level analysis also shows that extended reasoning can overturn an initially correct hypothesis. These results indicate that greater reasoning depth does not automatically confer robust reasoning breadth, and that reasoning breadth remains largely uncovered by current benchmarks.
Chinese Translation
大型语言模型（LLMs）在需要越来越长且复杂的推理链的推理任务上取得了显著进展。这一进展主要反映了推理深度。而一个互补且相对未被充分研究的能力是推理广度：并行探索多个语义方向并将所得线索整合成一个连贯的答案。我们引入了 MPAR-Bench，这是一个双语英语-中文基准，通过多点关联推理来隔离推理广度。受合作游戏 Just One 的启发，每个项目要求模型从几个独立生成的、语义多样的线索中恢复一个隐藏的目标。我们使用多智能体线索生成管道、基于嵌入的多样性过滤和人工验证构建了 1,000 个项目。答案空间仅来自公共词汇表，而每组线索都是从零开始生成的。除了精确匹配准确度之外，我们还通过准确度、ANLS、嵌入相似性、推理轨迹验证和四种扰动（线索屏蔽、顺序打乱、干扰项注入和多步线索）来评估模型。在评估的模型中，扰动使得英语的准确度降低了 9-18 个百分点，中文降低了 5-12 个百分点。思维模式提高了标准设定的准确度，尤其是在英语中，但并未始终减少对扰动的敏感性。案例级分析还表明，扩展推理可以推翻最初正确的假设。这些结果表明，推理深度的增加并不自动赋予强大的推理广度，而推理广度在当前基准中仍然 largely 未被覆盖。
cs.CL / 20 / 2608.10459
MD-ProTector: Positioning Multiple Data-Driven Prototypes for LLM-Generated Text Detection
MD-ProTector：定位多数据驱动原型以检测LLM生成文本
Abstract
As LLM-generated content becomes more sophisticated, detection systems for distinguishing those texts from human-written text must operate at scale while handling diverse writing styles, domains, languages, and generator models. Input-only encoder detectors are suitable for practical deployment setting, but standard binary classification supplies only the class label and does not explicitly organize the substantial variation within either class. We propose MD-ProTector, which represents each class with multiple trainable reference vectors in the encoder embedding space, referred to as prototypes. These prototypes provide separate decision boundaries for different groups of texts within the same class. However, adding multiple prototypes alone does not determine which variation each prototype should represent. MD-ProTector addresses this problem with Prototype Positioning loss, which separates class-level structure from the within-class variation that differentiates individual prototypes. Evaluated across five settings from three large-scale benchmarks covering domain, generator, language, and adversarial variation, MD-ProTector achieves the highest AvgRec on MAGE CDCM and RAID and the highest AUROC and lowest FPR95 on RAID among the compared encoder-based methods.
Chinese Translation
随着LLM生成内容变得越来越复杂，区分这些文本与人类撰写文本的检测系统必须在规模上运行，同时处理多样的写作风格、领域、语言和生成模型。仅输入编码器的检测器适合于实际部署环境，但标准的二元分类仅提供类标签，并未明确组织每个类别内的显著变异。我们提出了MD-ProTector，它在编码器嵌入空间中用多个可训练的参考向量（称为原型）来表示每个类别。这些原型为同一类别内不同文本组提供了独立的决策边界。然而，仅仅添加多个原型并不能确定每个原型应代表的变异。MD-ProTector通过原型定位损失（Prototype Positioning loss）解决了这个问题，该损失将类别级结构与区分个体原型的类别内变异分开。在覆盖领域、生成器、语言和对抗变异的三个大规模基准的五个设置中评估，MD-ProTector在MAGE CDCM和RAID上实现了最高的AvgRec，并在比较的基于编码器的方法中在RAID上获得了最高的AUROC和最低的FPR95。
cs.CL / 21 / 2608.10462
Calibrating Post-Training Feature Shifts for LLM Data Contamination Detection
校准后训练特征偏移以检测大型语言模型数据污染
Abstract
Large language models (LLMs) are trained on massive and largely undisclosed corpora that may contain copyrighted or privacy-sensitive content. Data contamination detection (DCD) therefore aims to determine whether a given text is a member of the pre-training corpus of a target LLM. Recent state-of-the-art DCD methods follow a feature-based paradigm that derives membership features from the input text and the corresponding model output. However, most modern LLMs undergo post-training, such as instruction tuning, preference optimization, and reasoning-oriented training, which can alter model outputs and shift the corresponding membership features, thereby reducing the separability between members and non-members. To address this problem, we propose CalibDCD, a broadly applicable calibration framework for feature-based DCD methods, comprising (1) Multi-View Shift Detection, which identifies recurring feature shifts associated with post-training, and (2) Bounded Feature Correction, which selectively mitigates their influence on membership prediction. Specifically, Multi-View Shift Detection evaluates controlled prompt variants on known non-member texts and consolidates the most informative views to identify recurring feature shifts. Bounded Feature Correction selectively adjusts feature components aligned with the detected shifts and controls the correction extent to preserve useful detection information. Experiments show that CalibDCD consistently improves existing feature-based detectors, with gains of up to 7.0% in AUC and 15.0% in TPR@5%FPR.
Chinese Translation
大型语言模型（LLMs）是在大量且大部分未公开的语料库上训练的，这些语料库可能包含受版权保护或隐私敏感的内容。因此，数据污染检测（DCD）旨在确定给定文本是否属于目标LLM的预训练语料库。最近的最先进的DCD方法遵循基于特征的范式，从输入文本及其对应的模型输出中推导出成员特征。然而，大多数现代LLM经历后训练，例如指令调优、偏好优化和面向推理的训练，这可能会改变模型输出并改变相应的成员特征，从而降低成员与非成员之间的可分性。为了解决这个问题，我们提出了CalibDCD，这是一个广泛适用的基于特征的DCD方法的校准框架，包括（1）多视角偏移检测（Multi-View Shift Detection），用于识别与后训练相关的重复特征偏移，以及（2）有界特征修正（Bounded Feature Correction），选择性地减轻这些偏移对成员预测的影响。具体而言，多视角偏移检测在已知的非成员文本上评估受控提示变体，并整合最具信息性的视角以识别重复特征偏移。有界特征修正选择性地调整与检测到的偏移对齐的特征组件，并控制修正程度以保留有用的检测信息。实验表明，CalibDCD始终改善现有的基于特征的检测器，在AUC上提高了高达7.0%，在TPR@5%FPR上提高了15.0%。
cs.CL / 22 / 2608.10503
Every Token Counts: Exact Likert-Scale Distributions for Measuring LLM Attitudes and Biases
每个标记都重要：测量大型语言模型态度和偏见的精确李克特量表分布
Abstract
As Large Language Models (LLMs) are increasingly deployed as autonomous agents, accurately evaluating their latent values and biases is critical. The NLP community typically evaluates models using large, unstructured benchmarks. While effective for general capabilities, these datasets fundamentally conflate causal mechanisms: even when an aggregate bias is detected, unstructured evaluations cannot disentangle whether it stems from baseline traits, contextual confounders, or complex interactions. To address this, we introduce an analytically exact framework for the controlled behavioral evaluation of LLMs. We bridge human psychometrics with LLM mechanics by resolving gaps in design, measurement, and analysis. First, we replace unstructured prompting with fully crossed factorial experiments to systematically isolate causal main and interaction effects. Second, we eliminate Monte Carlo text sampling noise by operating directly on exact, token-level Probability Mass Functions (PMFs). Third, we derive a multivariate ordinal consensus metric and a distributional ANOVA to process these PMFs analytically. We validate our framework with a case study on consumer ethnocentrism across five LLMs, demonstrating how our approach isolates systemic country-of-origin biases that aggregate benchmarks otherwise obscure.
Chinese Translation
随着大型语言模型（LLMs）越来越多地作为自主代理被部署，准确评估它们潜在的价值观和偏见变得至关重要。自然语言处理（NLP）社区通常使用大型非结构化基准来评估模型。尽管这些基准在评估一般能力方面有效，但它们在根本上混淆了因果机制：即使检测到聚合偏见，非结构化评估也无法分辨其是否源于基线特征、上下文混淆因素或复杂的交互作用。为了解决这一问题，我们提出了一个用于控制行为评估LLMs的分析精确框架。我们通过解决设计、测量和分析中的差距，将人类心理测量学与LLM机制相结合。首先，我们用完全交叉的因子实验替代非结构化提示，以系统地孤立因果主效应和交互效应。其次，我们通过直接操作精确的标记级概率质量函数（PMFs）来消除蒙特卡洛文本采样噪声。第三，我们推导出一个多元有序共识指标和一个分布式方差分析（ANOVA）来分析这些PMFs。我们通过对五个LLMs的消费者民族中心主义的案例研究验证了我们的框架，展示了我们的方法如何孤立出系统性的原产国偏见，而这些偏见在聚合基准中往往被掩盖。
cs.CL / 23 / 2608.10606
ASR-Roundtrip Evaluation Can Mask Context- and Convention-Dependent Reading Errors in Chinese News TTS
ASR-回环评估可能掩盖中文新闻TTS中的上下文和约定依赖的阅读错误
Abstract
ASR-roundtrip evaluation is widely used as a scalable proxy for text-to-speech (TTS) intelligibility, but it can produce false negatives for reading errors perceived by listeners. We study Chinese news TTS spans whose correct reading depends on context or domain conventions, such as sports scores, aircraft models, technical units, and membership names. In these cases, Raw TTS can choose a plausible but wrong reading while ASR transcribes the audio as the intended or surface-correct text. A targeted audit over 110 high-risk MiMo TTS cases, reported with a complete denominator, confirms 46 masked false negatives, 9 exposed TTS errors, and 55 cases with no Raw TTS error. A span-isolation diagnostic re-exposes 18/46 previously masked errors. A Raw-only CosyVoice audit on the same targeted pool confirms 51 masked cases. Across the 97 TTS-specific audio files labeled confirmed masked across the two audits, Qwen3-ASR surface-recovers 40 cases, whereas Paraformer does so in only 2. The results suggest that ASR-roundtrip is useful for screening but insufficient as standalone ground truth for Chinese news reading-risk evaluation.
Chinese Translation
ASR-回环评估被广泛用作文本到语音（TTS）可懂度的可扩展代理，但它可能会产生听众感知的阅读错误的假阴性。我们研究了中文新闻TTS片段，其正确阅读依赖于上下文或领域约定，例如体育比分、飞机型号、技术单位和会员名称。在这些情况下，原始TTS可能选择一个看似合理但错误的阅读，而ASR将音频转录为预期或表面正确的文本。对110个高风险MiMo TTS案例进行的针对性审计，报告了完整的分母，确认了46个掩盖的假阴性、9个暴露的TTS错误，以及55个没有原始TTS错误的案例。一个片段隔离诊断重新揭示了18/46个先前被掩盖的错误。对同一目标池进行的仅原始CosyVoice审计确认了51个掩盖案例。在两个审计中标记为确认掩盖的97个TTS特定音频文件中，Qwen3-ASR恢复了40个案例，而Paraformer仅恢复了2个。结果表明，ASR-回环在筛查中是有用的，但作为中文新闻阅读风险评估的独立真实依据则不够充分。
cs.CL / 24 / 2608.10615
Simplex Relaxation for Discrete Diffusion
离散扩散的单纯形松弛
Abstract
Discrete diffusion models for categorical generation are defined by a corruption kernel, which determines the intermediate state space and the associated reverse prediction problem. We study uniform discrete diffusion and ask whether its training objective and reverse transitions can be enriched without changing the underlying categorical corruption process. We introduce Simplax, an exact Dirichlet--categorical augmentation that couples each corrupted categorical state with an auxiliary simplex-valued variable while preserving the original uniform diffusion process as its categorical marginal. This augmentation yields a tractable Rao--Blackwellized reverse-bridge objective and a corresponding stochastic reverse sampler, while retaining the corrupted categorical state as the denoiser input. Empirically, Simplax improves the generative perplexity--entropy tradeoff on unconditional OpenWebText generation. On Sudoku, a model trained exclusively on $30$-clue puzzles achieves the highest accuracy among the compared methods across all evaluated clue densities, including the minimum uniquely solvable $17$-clue regime, and also achieves the highest validity in unconditional generation.
Chinese Translation
离散扩散模型用于分类生成，由腐蚀核定义，该腐蚀核决定了中间状态空间及相关的反向预测问题。我们研究均匀离散扩散，并探讨其训练目标和反向转移是否可以在不改变基础分类腐蚀过程的情况下得到增强。我们引入了Simplax，这是一种精确的Dirichlet-分类增强方法，它将每个腐蚀的分类状态与一个辅助的单纯形值变量耦合，同时保持原始的均匀扩散过程作为其分类边际。这种增强产生了一个可处理的Rao-Blackwell化反向桥接目标和相应的随机反向采样器，同时保留了腐蚀的分类状态作为去噪器输入。从实证上看，Simplax改善了无条件OpenWebText生成中的生成困惑度-熵权衡。在数独问题上，专门在$30$个线索谜题上训练的模型在所有评估的线索密度中达到了最高的准确率，包括最低唯一可解的$17$个线索范围，并且在无条件生成中也达到了最高的有效性。
cs.CL / 25 / 2608.10626
Dual-Loop Self-Evolution via Verifiable Emotion Feedback for Multi-Turn Empathetic Dialogue
通过可验证情感反馈实现双循环自我演化的多轮共情对话
Abstract
Large language models have demonstrated conversational capabilities, yet empathetic competence remains challenging. Empathetic support is inherently multi-turn and path-dependent: users disclose concerns gradually, emotions evolve over time, and early responses shape trust and receptivity. Reinforcement learning with verifiable emotion rewards provides scalable supervision for long-horizon interactions. However, existing methods evolve the dialogue policy while keeping its training interaction distribution fixed, creating a mismatch between policy competence and training experience. We introduce a dual-loop self-evolution framework driven by verifiable emotion feedback. With the user simulator and verifier frozen, the inner loop optimizes the multi-turn policy using continuous emotion rewards, while the outer loop uses the same outcomes to estimate policy-relative interaction utility and adapt experience. To obtain estimates from sparse, stochastic rollouts, the framework holds the scenario and interaction state constant within each group and prioritizes conditions whose group pass rates lie near the policy's competence boundary. A hierarchical controller shares evidence across support intents, while uncertainty-guided exploration and uniform rehearsal prevent premature exclusion. The resulting distribution generates trajectories, closing both loops without increasing the rollout budget. On SAGE, our framework raises Qwen3-8B Overall from 53.87 to 79.24 and outperforms protocol-matched uniform emotion-reward reinforcement learning by 7.23 points.
Chinese Translation
大型语言模型已展示出对话能力，但共情能力仍然具有挑战性。共情支持本质上是多轮的且依赖于路径：用户逐渐披露关切，情感随时间演变，早期反应塑造信任和接受度。使用可验证情感奖励的强化学习为长期交互提供了可扩展的监督。然而，现有方法在保持训练交互分布不变的情况下演化对话策略，导致策略能力与训练经验之间的不匹配。我们引入了一种由可验证情感反馈驱动的双循环自我演化框架。在用户模拟器和验证器被冻结的情况下，内循环利用连续情感奖励优化多轮策略，而外循环则使用相同的结果来估计策略相关的交互效用并调整经验。为了从稀疏的随机回合中获得估计，该框架在每个组内保持场景和交互状态不变，并优先考虑那些组通过率接近策略能力边界的条件。一个层次控制器在支持意图之间共享证据，而不确定性引导的探索和均匀排练防止过早排除。最终生成的分布产生轨迹，闭合两个循环而不增加回合预算。在SAGE上，我们的框架使Qwen3-8B的总体得分从53.87提高到79.24，并比协议匹配的均匀情感奖励强化学习高出7.23分。
cs.CL / 26 / 2608.10627
Decomposition-Induced Context-Memory Conflict: When Fact-Checking Pipelines Contradict Their Own Source Text
分解引发的上下文-记忆冲突：当事实核查流程与其自身源文本相矛盾时
Abstract
Decompose-then-verify pipelines, including FActScore-style fact-checkers and long-form factuality evaluators, first split a passage into atomic claims before checking each one. Decomposition itself is treated as a neutral preprocessing step. We show it is not: a decomposer can be induced to substitute its own parametric belief for what the source passage says, producing a claim that contradicts the text it was supposed to summarize faithfully. We call this Decomposition-Induced Context-Memory Conflict (DI-CC) and show it is mechanistically the same phenomenon as classical context-memory conflict, occurring inside a different pipeline stage than prior work has examined. A linear probe trained only on classical context-memory conflict data (NQ-Swap), never exposed to any decomposition output, significantly separates decomposition positions that produce DI-CC from faithful decompositions (AUC = 0.86-0.88, permutation p < 0.0005). An existing reference-free baseline, SelfCheckGPT-style self-consistency sampling, fails to detect DI-CC at all (AUC 0.51, chance-level), because DI-CC content is stably recoverable and recurs across resamples, unlike the variability self-consistency methods rely on. Context-aware decoding, a training-free mitigation from the classical setting, transfers to decomposition and suppresses DI-CC, but at a severe cost: many decompositions under coreference-heavy conditions fail to parse, often because the decomposer fabricates a different identity. We do not consider this mitigation deployment-ready. We further characterize the mechanism's boundaries: its natural occurrence rate is too sparss not manifest on naturally-occurring hallucinatedtext, and it requires a minimum model scale to detecablish DI-CC as a real, mechanistically grounded, andpartially treatable failure mode, with a scope we chhan overstate.
Chinese Translation
分解-再验证流程，包括 FActScore 风格的事实核查器和长篇事实性评估器，首先将一段文字分解为原子性主张，然后逐一进行核查。分解本身被视为一个中性的预处理步骤。我们表明，这并非如此：分解器可能被诱导用其自身的参数化信念替代源段落所述内容，从而产生一个与其应当忠实总结的文本相矛盾的主张。我们将其称为分解引发的上下文-记忆冲突（Decomposition-Induced Context-Memory Conflict, DI-CC），并展示其在机制上与经典的上下文-记忆冲突现象相同，发生在不同于以往研究所考察的流程阶段。仅在经典上下文-记忆冲突数据（NQ-Swap）上训练的线性探针，未接触任何分解输出，能够显著区分产生 DI-CC 的分解位置与忠实分解（AUC = 0.86-0.88，置换 p < 0.0005）。现有的无参考基线 SelfCheckGPT 风格的自一致性采样完全无法检测 DI-CC（AUC 0.51，随机水平），因为 DI-CC 内容是可稳定恢复的，并且在重采样中反复出现，这与自一致性方法所依赖的变异性不同。上下文感知解码，作为经典设置中的一种无训练缓解方法，可以转移到分解中并抑制 DI-CC，但代价极高：在核心指代密集的条件下，许多分解无法解析，通常是因为分解器虚构了不同的身份。我们不认为这种缓解方法已经准备好部署。我们进一步界定该机制的边界：其自然发生率过低，未在自然发生的幻觉文本中表现出来，并且需要最低模型规模才能检测 DI-CC，确立 DI-CC 作为一种真实的、机制上有依据的、部分可处理的失败模式，其范围我们可能夸大。
cs.CL / 27 / 2608.10670
Seeds Before Objectives: Rethinking Evaluation for Low-Resource Garhwali ASR
目标之前的种子：重新思考低资源Garhwali自动语音识别的评估
Abstract
At corpus sizes typical of low-resource dialects, single-run comparisons can yield gains that do not replicate. We show this for Garhwali, an under-resourced Indo-Aryan language of the central Himalaya, building the first reproducible multi-seed ASR benchmark on the official VAANI splits, with per-seed outputs and significance testing. Re-examining plausible gains, we find them fragile: neither Focal CTC nor a matra-weighted objective beats standard CTC under seed-level testing, the matra objective fails to cut even its targeted errors, and Hindi-to-Garhwali transfer gives no gain over direct fine-tuning. What holds up is mundane: w2v-BERT 2.0 with standard CTC reaches 47.0% WER over five seeds, beating the larger MMS-1B and comparable models; pretraining design, not parameter count, drives performance, and speed augmentation gives a small, largely consistent gain. Multi-seed evaluation on official splits separates real gains from seed noise.
Chinese Translation
在低资源方言的典型语料库规模下，单次运行的比较可能会产生无法复制的增益。我们以Garhwali为例，这是一种位于中喜马拉雅地区的资源匮乏的印欧语系语言，构建了第一个可重复的多种子自动语音识别（ASR）基准，基于官方的VAANI数据集，提供每种子的输出和显著性测试。在重新审视可能的增益时，我们发现这些增益是脆弱的：在种子级测试中，无论是Focal CTC还是加权目标都无法超越标准CTC，加权目标甚至未能减少其目标错误，而印地语到Garhwali的迁移相较于直接微调并未带来任何增益。值得注意的是，w2v-BERT 2.0与标准CTC结合，在五个种子上达到了47.0%的字错误率（WER），超越了更大的MMS-1B及其他可比模型；预训练设计而非参数数量驱动了性能，而速度增强带来了小幅且相对一致的增益。在官方数据集上的多种子评估能够区分真实增益与种子噪声。
cs.CL / 28 / 2608.10678
Auditing Chinese Web-scale Corpora via Sampled BPE Token Statistics
通过抽样 BPE 令牌统计审计中国网络规模语料库
Abstract
Chinese web pollution has surfaced in LLMs, motivating audits of upstream Chinese corpora. However, auditing such corpora faces three challenges: (1) their web-scale size makes full scan costly; (2) prior analyses are often too coarse to expose token-level pollution; (3) Chinese web pollution is implicit and rapidly changing. We propose Sampled-BPE, a lightweight token-level auditing pipeline that sample a small subset and train BPE tokenizer to surface polluted tokens. Experiments show that Sampled-BPE preserves usable estimates while substantially reducing runtime and memory: a 148.4 $\times$ speedup and a 35.8 $\times$ memory reduction induce only 4.25% relative error for pollution categories. We apply the pipeline to 11 open Chinese corpora and 6 Chinese Common Crawl snapshots from 2021 to 2026. The audit reveals widespread but uneven pollution across open corpora, as well as highly polluted and temporally shifting Chinese web content. We further release a hierarchical Chinese web token dataset with 660k+ token records, each with web context, category, and explanation fields, organized as trees to support review and tracing of pollution.
Chinese Translation
中文网络污染在大型语言模型（LLMs）中显现，促使对上游中文语料库进行审计。然而，审计此类语料库面临三个挑战：（1）其网络规模的大小使得全面扫描成本高昂；（2）以往的分析往往过于粗略，无法揭示令牌级别的污染；（3）中文网络污染是隐性且快速变化的。我们提出了 Sampled-BPE，一种轻量级的令牌级审计管道，通过抽样小部分数据并训练 BPE 令牌化器来揭示污染令牌。实验表明，Sampled-BPE 在显著减少运行时间和内存的同时保持可用估计：实现了 148.4 倍的加速和 35.8 倍的内存减少，仅对污染类别产生 4.25% 的相对误差。我们将该管道应用于 11 个开放的中文语料库和 2021 至 2026 年间的 6 个中文 Common Crawl 快照。审计结果揭示了开放语料库中广泛但不均匀的污染，以及高度污染和时间上变化的中文网络内容。我们进一步发布了一个层次化的中文网络令牌数据集，包含超过 66 万条令牌记录，每条记录具有网络上下文、类别和解释字段，组织为树形结构以支持污染的审查和追踪。
cs.CL / 29 / 2608.10688
Leveraging Human Reading Behavior for Keyphrase Extraction: A Webcam-based Eye-tracking Corpus
利用人类阅读行为进行关键词提取：基于网络摄像头的眼动追踪语料库
Abstract
Purpose: Keyphrases are statistically and semantically important textual units that can also attract readers' attention during comprehension. However, existing keyphrase extraction (KPE) studies mainly focus on improving textual representation while largely overlooking human reading behavior. This study examines whether lightweight webcam-based eye-tracking features can improve KPE from Chinese academic abstracts in Library and Information Science (LIS). Methodology: To address the limited availability of eye-tracking data for Chinese academic reading, we developed a lightweight webcam-based data collection platform using the open-source SearchGazer library and constructed the Chinese LIS Eye-Tracking Corpus (CLIS-ET). Three character-level eye-tracking features, first fixation duration (FFD), fixation number (FN), and total fixation duration (TFD), were incorporated into KPE models to evaluate their effects on extraction performance. Findings: Eye-tracking features consistently improved KPE performance. The combination of FN and TFD achieved the best results on the Att-BiLSTM+CRF model, indicating that readers' fixation behavior provides useful signals for identifying keyphrases in academic abstracts. Originality/value: This study introduces a cost-effective webcam-based eye-tracking approach for KPE and presents CLIS-ET, a Chinese academic eye-tracking corpus containing FFD, FN, and TFD features. The results demonstrate the value of incorporating human reading behavior into keyphrase extraction. Dataset and code: https://github.com/yan-xinyi/ET_AKE and https://github.com/yan-xinyi/Reading_ET_System.
Chinese Translation
目的：关键词是具有统计和语义重要性的文本单元，能够在理解过程中吸引读者的注意。然而，现有的关键词提取（KPE）研究主要集中在改善文本表示上，而在很大程度上忽视了人类的阅读行为。本研究探讨了轻量级基于网络摄像头的眼动追踪特征是否能够改善来自图书与信息科学（LIS）领域的中文学术摘要的关键词提取。方法：为了解决中文学术阅读的眼动追踪数据有限的问题，我们开发了一个轻量级的基于网络摄像头的数据收集平台，使用开源的SearchGazer库，并构建了中文LIS眼动追踪语料库（CLIS-ET）。将三个字符级眼动追踪特征：首次注视持续时间（FFD）、注视次数（FN）和总注视持续时间（TFD）纳入关键词提取模型，以评估其对提取性能的影响。结果：眼动追踪特征持续改善了关键词提取的性能。FN和TFD的组合在Att-BiLSTM+CRF模型上取得了最佳结果，表明读者的注视行为为识别学术摘要中的关键词提供了有用的信号。原创性/价值：本研究提出了一种具有成本效益的基于网络摄像头的眼动追踪方法用于关键词提取，并呈现了包含FFD、FN和TFD特征的中文学术眼动追踪语料库CLIS-ET。结果展示了将人类阅读行为纳入关键词提取的价值。数据集和代码： https://github.com/yan-xinyi/ET_AKE 和 https://github.com/yan-xinyi/Reading_ET_System。
cs.CL / 30 / 2608.10690
Can Released LLM Vocabularies Support Token-Level Estimation of Hidden Corpora?
发布的 LLM 词汇能否支持隐藏语料的 Token 级估计？
Abstract
Pretraining corpus composition shapes LLM capabilities, but it often remains hidden even when model weights are released. Prior work has inferred corpus mixtures or traced specific token groups from released tokenizer vocabularies; in contrast, we estimate corpus ratios for arbitrary target tokens. We first show that BPE tokenizers trained on different corpora share stable token ID--ratio distributions, motivating distribution transfer from known corpora to a target tokenizer trained on hidden corpora. We then propose Quantile-Guided Density Estimation (QGDE), which approximates this distribution with multiple quantile trends and uses local density weighting to produce token-level estimates. In controlled settings and a realistic setting using the released SmolLM tokenizer, QGDE achieves mean relative errors as low as 3.00% for token-level estimation and 3.08% after aggregation into category-level mixtures. These results suggest that released tokenizer vocabularies provide a useful signal for fine-grained corpus estimation beyond coarse composition inference.
Chinese Translation
预训练语料的组成影响 LLM 的能力，但即使在模型权重发布后，这些信息通常仍然是隐藏的。之前的研究通过已发布的分词器词汇推断语料混合或追踪特定的 token 组；而我们则估计任意目标 token 的语料比例。我们首先展示了在不同语料上训练的 BPE 分词器共享稳定的 token ID-比例分布，这为从已知语料向在隐藏语料上训练的目标分词器转移分布提供了动机。然后，我们提出了分位数引导密度估计（Quantile-Guided Density Estimation, QGDE），该方法通过多个分位数趋势近似该分布，并使用局部密度加权生成 token 级估计。在受控环境和使用发布的 SmolLM 分词器的真实场景中，QGDE 在 token 级估计中实现了低至 3.00% 的平均相对误差，而在聚合到类别级混合后为 3.08%。这些结果表明，发布的分词器词汇为细粒度语料估计提供了有用的信号，超出了粗略组成推断的范围。
cs.CL / 31 / 2608.10692
SPIEval: Evaluating Large Language Models as Mobile Assistants over Scattered Personal Information
SPIEval：评估大型语言模型作为移动助手在分散个人信息上的表现
Abstract
Large language models (LLMs) are increasingly deployed as mobile assistants, where a key challenge is leveraging personal information scattered across multiple applications (apps) to complete user instructions. However, due to the lack of dedicated benchmarks, their capabilities remain poorly understood. To address this gap, we introduce SPIEval, a human-curated benchmark grounded in five cognitive capabilities (i.e., reasoning, disambiguation, integration, preference inference, and multi-intent decomposition). SPIEval comprises 250 tasks spanning 4,335 personal records distributed across 10 apps and supports multi-turn interaction through 21 tools. Analysis shows that the benchmark exhibits diverse scenarios, challenging tasks, scattered information, controllable environments, and verifiable outcomes. We evaluate nine representative LLMs and find substantial room for improvement. The best-performing model, GPT-5.5 (xhigh), achieves only 57.3% accuracy, while the weakest achieves just 16.4%. Further analysis reveals that 79% of failures stem from inaccurate information localization, as LLMs often commit to plausible but incorrect information instead of continuing retrieval for verification. We also find that fewer than 2% of retrieval actions employ advanced search methods and observe substantial variation in search efficiency across models. These findings expose fundamental limitations of current LLM-based mobile assistants and motivate future research in this direction. Data and code are available at https://huggingface.co/datasets/Junjie-Ye/SPIEval.
Chinese Translation
大型语言模型（LLMs）越来越多地被用作移动助手，其中一个关键挑战是利用分散在多个应用程序（apps）中的个人信息来完成用户指令。然而，由于缺乏专门的基准，其能力仍然不够明确。为了解决这一问题，我们引入了SPIEval，这是一个基于五种认知能力（即推理、消歧义、整合、偏好推断和多意图分解）的人为策划基准。SPIEval包含250个任务，涵盖分布在10个应用程序中的4,335条个人记录，并通过21种工具支持多轮交互。分析表明，该基准展示了多样化的场景、具有挑战性的任务、分散的信息、可控的环境和可验证的结果。我们评估了九个具有代表性的LLM，并发现有显著的改进空间。表现最佳的模型GPT-5.5（xhigh）仅达到57.3%的准确率，而表现最差的模型仅为16.4%。进一步分析显示，79%的失败源于信息定位不准确，因为LLMs往往承诺可信但不正确的信息，而不是继续检索以进行验证。我们还发现，少于2%的检索操作使用了高级搜索方法，并观察到不同模型之间搜索效率存在显著差异。这些发现揭示了当前基于LLM的移动助手的基本局限性，并激励未来在这一方向的研究。数据和代码可在https://huggingface.co/datasets/Junjie-Ye/SPIEval获取。
cs.CL / 32 / 2608.10698
EVIL-Detect for NLPCC 2026 Shared Task 6: LLM-Generated Text Detection
EVIL-Detect在NLPCC 2026共享任务6中的应用：LLM生成文本检测
Abstract
The rapid development of large language models (LLMs) has increased the need for reliable detection of LLM-generated text, especially in realistic Chinese scenarios involving human-written text (HWT), LLM-generated text (LGT), and LLM-refined text (HLT). This paper presents EVIL-Detect, a multi-signal ensemble framework with conflict-aware fusion for NLPCC 2026 Shared Task 6. The system integrates edit-extent regression, zero-shot likelihood-contrast signals, lexical statistics, and conservative text rules. With calibrated decision boundaries and conflict-aware integration, our system improves robustness under strong out-of-distribution shifts, achieving a macro-F1 score of 0.8888 and ranking first in the official evaluation. Our code is available at https://github.com/bbbbhrrrr/evildetect.
Chinese Translation
大型语言模型（LLMs）的快速发展增加了对LLM生成文本可靠检测的需求，尤其是在涉及人类撰写文本（HWT）、LLM生成文本（LGT）和LLM精炼文本（HLT）的现实中文场景中。本文提出了EVIL-Detect，一个具有冲突感知融合的多信号集成框架，旨在应对NLPCC 2026共享任务6。该系统集成了编辑范围回归、零样本似然对比信号、词汇统计和保守文本规则。通过校准的决策边界和冲突感知整合，我们的系统在强大的分布外变化下提高了鲁棒性，取得了0.8888的宏F1分数，并在官方评估中排名第一。我们的代码可在https://github.com/bbbbhrrrr/evildetect获取。
cs.CL / 33 / 2608.10715
Most biomedical publications show signs of LLM-assisted writing
大多数生物医学出版物显示出LLM辅助写作的迹象
Abstract
Over the past several years, LLM-powered chatbots and agents have become widely used as a tool for academic writing. LLM-assisted writing can be valuable by removing language barriers but at the same time causes concerns about misconduct and fraud. To inform policy decisions, it is necessary to monitor the prevalence of LLM-altered texts in scholarly publications. Despite some recent progress in this direction, no existing method can produce reliable estimates. Here we suggest and validate a new unbiased approach to estimate LLM usage in a corpus of texts based on changing word frequencies. We apply our method to the full texts of open-access biomedical papers from Pubmed Central, and show that by the end of 2025, 89% of papers show excess of LLM-associated vocabulary. We also find that LLMs are twice as likely to be used when writing a paragraph in the Discussion section (68%) compared to a paragraph in the Methods section (32%), but even inside the Methods section, the overall prevalence of LLM usage is over 50%. We believe that our estimates are crucial to shape future guidelines and policies.
Chinese Translation
在过去几年中，基于LLM（大语言模型）的聊天机器人和代理已广泛用作学术写作的工具。LLM辅助写作通过消除语言障碍而具有重要价值，但同时也引发了对不当行为和欺诈的担忧。为了为政策决策提供信息，有必要监测学术出版物中LLM修改文本的普遍性。尽管在这方面已有一些进展，但现有方法无法提供可靠的估计。在此，我们提出并验证了一种新的无偏见方法，以基于词频变化估计文本语料库中的LLM使用情况。我们将该方法应用于Pubmed Central的开放获取生物医学论文的全文，并显示到2025年底，89%的论文显示出LLM相关词汇的过量使用。我们还发现，在撰写讨论部分的段落时，使用LLM的可能性是撰写方法部分段落的两倍（68%对32%），但即使在方法部分，LLM使用的整体普遍性也超过50%。我们相信我们的估计对于制定未来的指导方针和政策至关重要。
cs.CL / 34 / 2608.10743
Mitigating Context Interference for Reliable and Efficient Search Agents
减轻上下文干扰以实现可靠和高效的搜索代理
Abstract
Recent research empowers Large Language Models (LLMs) as multi-turn search agents to iteratively retrieve and generate outputs until complex tasks are solved. However, the contexts of multi-turn search agents are lengthy and complex. For example, the retrieved set of documents in each turn would inevitably introduce irrelevant information that distracts LLMs, referring to \textit{context interference}, potentially hindering the reliability and efficiency of search agents. Therefore, we conduct a systematic study on context interference in multi-turn search agents, focusing on investigating i) which parts of the context of search agents will contribute to the context interference, ii) how to refine the contexts of search agents to mitigate the interference, and iii) can incorporating context refinement into search agent training yield further improvements. We reveal that interference primarily arises from the latest retrieved documents. Based on the explored findings, we then introduce a distill-based context refiner to dynamically mitigate context interference for multi-turn search agents. Finally, we validate that incorporating context refinement into RL training pipelines of search agents can significantly enhance both reliability and efficiency. This study highlights the importance of mitigating context interference of search agents, inspiring a novel paradigm of ``refine context and then generate'' for AI agents.
Chinese Translation
近期研究使大型语言模型（Large Language Models, LLMs）作为多轮搜索代理能够迭代地检索和生成输出，直至复杂任务得以解决。然而，多轮搜索代理的上下文往往冗长且复杂。例如，每一轮检索到的文档集不可避免地会引入无关信息，这些信息会分散LLMs的注意力，这被称为 extit{上下文干扰}，可能会妨碍搜索代理的可靠性和效率。因此，我们对多轮搜索代理中的上下文干扰进行了系统研究，重点调查i) 搜索代理的上下文中哪些部分会导致上下文干扰，ii) 如何优化搜索代理的上下文以减轻干扰，以及iii) 将上下文优化纳入搜索代理训练是否能带来进一步的改进。我们发现干扰主要源于最新检索到的文档。基于探索的发现，我们提出了一种基于蒸馏的上下文优化器，以动态减轻多轮搜索代理的上下文干扰。最后，我们验证了将上下文优化纳入搜索代理的强化学习训练流程可以显著提升其可靠性和效率。本研究强调了减轻搜索代理上下文干扰的重要性，为AI代理提供了一种“优化上下文后生成”的新范式。
cs.CL / 35 / 2608.10806
Assessing Reliability of BERT-Based Models on Question Answering Tasks
评估基于BERT模型的问答任务的可靠性
Abstract
Reliability estimation of large language models is in many cases as crucial as their accuracy, as reliable models are more trustworthy, robust, and suitable for practical applications. Recent advancements in natural language processing (NLP), particularly those based on transformer architectures, have significantly accelerated progress across various NLP tasks. This study focuses on the reliability of transformer-based question answering (QA) models, specifically BERT models and its variants (RoBERTa, ALBERT, DistilBERT). These encoder-only pretrained transformers have demonstrated remarkable accuracy in QA tasks that can be treated as classification tasks. However, their reliability remains underexplored. This study evaluates the reliability of four BERT-based models by assessing response stability under two conditions: (1) internal model variations induced via Monte Carlo Dropout (MCD) and (2) input perturbations through paraphrasing. Using the SQuAD and QuAC datasets, we investigate how dropout rates affect prediction consistency and whether lexical changes impact answer stability. Our findings reveal that RoBERTa maintains higher reliability, whereas AlBERT and DistilBERT exhibit significant inconsistencies. Statistical analyses confirm that enabling MCD during prediction does not disrupt inference dynamics, validating its effectiveness as a reliability metric. These findings underscore the importance of evaluating both accuracy and stability in QA models to ensure stability in real-world applications.
Chinese Translation
大型语言模型的可靠性评估在许多情况下与其准确性同样重要，因为可靠的模型更值得信赖、更稳健，并且更适合实际应用。近年来，自然语言处理（NLP）领域的进展，特别是基于变换器架构的模型，显著加速了各类NLP任务的进展。本研究聚焦于基于变换器的问答（QA）模型的可靠性，特别是BERT模型及其变体（RoBERTa、ALBERT、DistilBERT）。这些仅编码器的预训练变换器在可以视为分类任务的问答任务中表现出了显著的准确性。然而，它们的可靠性仍然未得到充分探讨。本研究通过评估在两种条件下的响应稳定性来评估四个基于BERT的模型的可靠性：（1）通过蒙特卡洛丢弃（Monte Carlo Dropout, MCD）引起的内部模型变化和（2）通过释义引起的输入扰动。利用SQuAD和QuAC数据集，我们调查了丢弃率如何影响预测一致性，以及词汇变化是否影响答案的稳定性。我们的研究结果表明，RoBERTa保持了更高的可靠性，而ALBERT和DistilBERT则表现出显著的不一致性。统计分析确认，在预测过程中启用MCD并未干扰推理动态，验证了其作为可靠性指标的有效性。这些发现强调了在问答模型中评估准确性和稳定性的重要性，以确保在实际应用中的稳定性。
cs.CL / 36 / 2608.10810
Surfacing the Unsaid: CUE-Bench for Affective Stance in Chinese Discourse
揭示未言之语：用于中文话语中的情感立场的 CUE-Bench
Abstract
Emotion understanding in discourse requires reasoning beyond surface sentiment because speakers often convey affect through indirect, implicit, polite, ironic, or deliberately mismatched expressions. Existing emotion benchmarks mainly annotate surface polarity or final emotion categories, while lacking a structured account of how explicit expression, implicit affect, pragmatic intent, and fine grained emotion interact. This limitation makes current evaluations insensitive to cases where affective meaning is concealed, weakened, inverted, or pragmatically reshaped, thereby obscuring model failures in deeper emotion understanding. To address this gap, we introduce CUE Bench, a Chinese Unsaid Emotion benchmark that centers on Affective Stance and covers diverse communicative scenarios. CUE Bench constructs nine human interpretable affective stances from explicit implicit polarity interaction and further provides intent and fine grained emotion annotations for structured affective inference. Experiments show that incorporating Affective Stance improves fine grained emotion recognition by 3.5 percentage points and pragmatic intent detection by 7.8 percentage points over strong baselines.
Chinese Translation
在话语中理解情感需要超越表面情感的推理，因为说话者常常通过间接、隐含、礼貌、讽刺或故意不匹配的表达来传达情感。现有的情感基准主要注释表面极性或最终情感类别，而缺乏对显性表达、隐性情感、语用意图和细粒度情感之间相互作用的结构化描述。这一局限性使得当前的评估对情感意义被隐藏、削弱、颠倒或在语用上重塑的情况不敏感，从而掩盖了模型在更深层次情感理解中的失败。为了解决这一问题，我们引入了 CUE Bench，一个以情感立场为中心的中文未言情感基准，涵盖多样的交际场景。CUE Bench 从显性和隐性极性互动中构建了九种人类可解释的情感立场，并进一步提供了意图和细粒度情感的注释，以支持结构化的情感推理。实验表明，结合情感立场可以使细粒度情感识别提高 3.5 个百分点，语用意图检测提高 7.8 个百分点，相较于强基线模型。
cs.CL / 37 / 2608.10812
Reference-Free Post-Training of Open Large Language Models for Multilingual Machine Translation
无参考后训练开放大型语言模型用于多语言机器翻译
Abstract
We study reference-free post-training for multilingual machine translation with open large language models. Starting from the supervised-finetuned MiLMMT-46-v0.1 models, we apply Group Relative Policy Optimization (GRPO) with a reward that averages two reference-free quality estimation models and is gated by language identification. We then linearly interpolate the supervised fine-tuning (SFT) and reinforcement learning (RL) model checkpoints to obtain MiLMMT-46-v1.0. Across 46 languages, the resulting models consistently improve translation quality over their SFT counterparts, outperform strong recent open baselines, including Seed-X, HY-MT2, and TranslateGemma, and achieve leading reference-free scores against evaluated proprietary systems such as Google Translate, Gemini 3 Pro, and GPT-5. We further investigate on-policy distillation and find that it reaches, but does not surpass, the quality frontier achieved by RL with checkpoint interpolation. We release the models and code to facilitate future research.
Chinese Translation
我们研究了使用开放大型语言模型进行多语言机器翻译的无参考后训练。基于经过监督微调的 MiLMMT-46-v0.1 模型，我们应用了群体相对策略优化（Group Relative Policy Optimization, GRPO），其奖励机制是对两个无参考质量评估模型的平均值，并通过语言识别进行门控。随后，我们线性插值监督微调（Supervised Fine-Tuning, SFT）和强化学习（Reinforcement Learning, RL）模型检查点，以获得 MiLMMT-46-v1.0。在46种语言中，所得到的模型在翻译质量上始终优于其 SFT 对应模型，超越了包括 Seed-X、HY-MT2 和 TranslateGemma 在内的强大近期开放基线，并在评估的专有系统（如 Google Translate、Gemini 3 Pro 和 GPT-5）中取得领先的无参考得分。我们进一步研究了在线蒸馏，发现其达到了但未超越通过检查点插值实现的 RL 质量前沿。我们发布了模型和代码，以促进未来的研究。
cs.CL / 38 / 2608.10875
VibeLifeBench: Can Your Life Agent Be Proactive and Persistent in a Living World?
VibeLifeBench：你的生活代理能否在动态世界中保持主动和持续？
Abstract
Large language model (LLM) agents are increasingly deployed as personal assistants. Existing evaluations, however, mostly use short, self-contained requests in static environments. Everyday life assistance is different. A task runs for weeks rather than minutes. The world keeps changing while the agent is not being prompted. Many constraints are never stated outright. An agent that merely answers the request in front of it will fail at such a task. What is needed instead is an agent that stays proactive and consistent. It decides on its own when to act, when to ask, and when to stay silent. It notices changes that nobody announced. It keeps one plan coherent from the first day to the last. No current benchmark measures this. We introduce VibeLifeBench, a benchmark of 200 long-horizon tasks across ten everyday-life domains. Each task is a scripted multi-week timeline in a simulated world of 22 mock services. The world advances on its own clock, and many of its changes are silent, so only an agent that re-inspects the world discovers them. Every task is graded by fine-grained, weighted checks that read only what the agent actually left behind, covering the end state, the timeliness of its actions, and whether it upheld the implicit constraints. We evaluate seven frontier models. All of them score low, which shows how far current agents are from assisting with real life. We will open-source all tasks, environments, and the evaluation framework.
Chinese Translation
大型语言模型（LLM）代理越来越多地被用作个人助手。然而，现有的评估大多使用静态环境中的短期、自包含请求。日常生活中的辅助任务则有所不同：任务往往持续数周而非数分钟。在代理未被提示的情况下，世界持续变化，许多约束条件从未明确说明。仅仅回应眼前请求的代理在这样的任务中将会失败。所需的是一个保持主动和一致性的代理。它能够自主决定何时行动、何时提问以及何时保持沉默。它能够察觉到无人宣布的变化，并能从第一天到最后一天保持一个连贯的计划。目前没有任何基准能够衡量这一点。我们引入了VibeLifeBench，这是一个涵盖十个日常生活领域的200个长期任务的基准。每个任务都是在一个模拟的22个虚拟服务的世界中编写的多周时间线。这个世界按照自己的时钟推进，许多变化是无声的，因此只有重新检查世界的代理才能发现它们。每个任务通过细致的加权检查进行评分，这些检查仅关注代理实际留下的内容，涵盖最终状态、行动的及时性以及是否遵循了隐含约束。我们评估了七个前沿模型，所有模型的得分都很低，这表明当前的代理距离真正的生活辅助还有很长的路要走。我们将开源所有任务、环境和评估框架。
cs.CL / 39 / 2608.10878
X2-Turn: Frame-Synchronous Dual-Head Modeling for Joint Streaming ASR and Turn State Prediction
X2-Turn：用于联合流式自动语音识别和轮次状态预测的帧同步双头建模
Abstract
Accurate and responsive turn-taking is essential for spoken dialogue systems, which must distinguish in real time between user interruptions, backchannels that should be ignored, and the completion of an utterance. Prior modular approaches typically optimize turn state prediction at the utterance or fixed-chunk level, creating a mismatch with the continuous turn state estimate, and often depend on an auxiliary ASR model, which limits responsiveness and increases overall system complexity. Therefore, we present X2-Turn, a frame-synchronous turn state prediction method via delayed-stream modeling. Specifically, building on the pretrained Voxtral Realtime model, we introduce a frame-synchronous turn state head that operates in parallel with the ASR head on shared streaming representations, jointly predicting ASR tokens and fine-grained turn states at the frame level. We evaluate our method on the bilingual Chinese-English Easy-Turn test sets, and the results demonstrate its effectiveness in achieving accurate turn-taking detection while maintaining low latency.
Chinese Translation
准确且响应迅速的轮次切换对于口语对话系统至关重要，这些系统必须实时区分用户的打断、应被忽略的回声以及话语的完成。以往的模块化方法通常在话语或固定块级别优化轮次状态预测，这与连续的轮次状态估计存在不匹配，并且往往依赖于辅助的自动语音识别（ASR）模型，这限制了响应能力并增加了整体系统复杂性。因此，我们提出了X2-Turn，一种通过延迟流建模实现的帧同步轮次状态预测方法。具体而言，基于预训练的Voxtral Realtime模型，我们引入了一个帧同步的轮次状态头，该头与ASR头在共享流表示上并行操作，联合预测ASR标记和细粒度的轮次状态。我们在双语中英文Easy-Turn测试集上评估了我们的方法，结果表明其在实现准确的轮次切换检测的同时保持低延迟方面的有效性。
cs.CL / 40 / 2608.10893
Certify or Refuse: A Cross-Model Map for Selective Risk Control with Coverage Floors under Covariate Shift
认证或拒绝：在协变量偏移下具有覆盖底线的选择性风险控制跨模型映射
Abstract
Certified selective predictors attain whatever coverage they attain; operators impose an automation floor: answer at least a $\beta$-fraction of shifted target traffic with at most an $\alpha$-fraction of answers wrong. Under bounded-ratio covariate shift we prove the Floor Certification Map: once that floor must be certified alongside the selection-conditioned risk $\alpha$, certification acquires a feasibility frontier and a two-resource complexity map, additive up to constants: risk in labeled source, the floor in unlabeled target samples. The rates are local, needing a regular frontier margin, slack below the local-regime threshold, and lattice conditions: pre-registered with a lattice margin for the upper bounds, compatible per-slack for the lower. The displayed split is the operational route; oracle weights also allow a labeled-source floor estimate. Three model-tagged results: a lower bound (Model-B), a matching oracle-weight upper bound (Model-A), and an implementable upper bound (Model-B') valid under a pre-registered exact stratified-shift model with nuisance cost priced explicitly. The match is across these models rather than a single-model minimax theorem, and necessarily so: over the full bounded-ratio class no unknown-weight procedure matches at any sample size (Model-B is inconsistent, witnessed at $\alpha=\beta=1/2$). The nuisance's necessity is only partially settled. Complexity tracks a localized accepted-region functional, not global effective sample size (ESS), on both sides, though a fixed-ESS separation theorem is left open; both lower-bound axes vanish as $\beta\to0$, so the floor creates the map. Empirically, the registered bite family diverges with log-log slope $-2.002$ within its pre-registered band; a 1,024-cell audit records 0 violations where the formal certificates fire; and a single-corpus SQuAD-to-NewsQA feasibility audit returns honest refusal.
Chinese Translation
认证的选择性预测器能够达到其所能达到的任何覆盖率；操作员施加了自动化底线：至少以 $eta$ 的比例回答偏移目标流量，且错误回答的比例最多为 $ rac{eta}{eta + rac{eta}{eta}}$。在有界比率协变量偏移下，我们证明了底线认证映射：一旦该底线必须与选择条件风险 $ rac{eta}{eta + rac{eta}{eta}}$ 一起认证，认证便获得了可行性前沿和一个两资源复杂度映射，最多加上常数：标记源中的风险，未标记目标样本中的底线。这些速率是局部的，需要一个规则的前沿边际，局部区域阈值以下的松弛，以及晶格条件：在上界的预注册晶格边际下，适用于下界的每个松弛。显示的分裂是操作路径；oracle 权重也允许标记源底线估计。三个模型标记的结果：下界（模型-B），匹配的 oracle 权重上界（模型-A），以及在预注册的精确分层偏移模型下有效的可实施上界（模型-B'），其中杂项成本被明确定价。匹配发生在这些模型之间，而不是单一模型的最小最大定理，且必然如此：在整个有界比率类中，没有未知权重的程序在任何样本大小下匹配（模型-B 在 $ rac{eta}{eta + rac{eta}{eta}}= rac{eta}{eta + rac{eta}{eta}}=1/2$ 时不一致）。杂项的必要性仅部分得到解决。复杂度跟踪一个局部接受区域功能，而不是全球有效样本大小（ESS），尽管固定 ESS 分离定理尚未解决；随着 $eta o0$，两个下界轴消失，因此底线创建了映射。从经验上看，注册的咬合家族在其预注册带内以对数-对数斜率 $-2.002$ 发散；一个 1,024 单元的审计记录了 0 次违规，其中正式证书生效；而单一语料库的 SQuAD 到 NewsQA 可行性审计返回诚实拒绝。
cs.CL / 41 / 2608.10916
FaithformBench: Benchmarking Faithfulness of Mathematical Chain-of-Thought Autoformalisation
FaithformBench：数学推理链式思维自动形式化的可信度基准测试
Abstract
Autoformalisation (AF) systems map natural language reasoning steps into formal statements in a proof assistant such as Lean. We consider how to assess the faithfulness of these systems. Existing approaches require expensive human-annotated ground truth, or rely on LLM judges or embedding models, which come with limited guarantees of accuracy. In addition, these methods typically only consider inputs that are known to be correct, and therefore do not assess whether the AF translates incorrect inputs faithfully. To address these limitations, we propose a new benchmark for AF faithfulness that is cheap to apply, sound under weak assumptions, and assesses both positive and negative examples. Our method is based on automatically generating perturbed reasoning steps that are designed to be invalid, and then measuring validity preservation on unperturbed steps and invalidity preservation on perturbed steps. We apply our method to eight AF systems across four mathematical datasets, and observe pervasive sycophancy: many AFs "silently correct" invalid inputs into provable statements. The most validity-preserving fine-tuned AFs are also the most sycophantic, suggesting a tension between validity and invalidity preservation in current AF systems.
Chinese Translation
自动形式化（AF）系统将自然语言推理步骤映射为证明助手（如 Lean）中的形式化语句。我们考虑如何评估这些系统的可信度。现有方法需要昂贵的人为注释的真实数据，或依赖于大型语言模型（LLM）评判者或嵌入模型，这些方法在准确性上提供的保证有限。此外，这些方法通常只考虑已知正确的输入，因此并未评估 AF 是否能够忠实地翻译不正确的输入。为了解决这些局限性，我们提出了一种新的 AF 可信度基准，具有低成本、在弱假设下有效，并评估正例和负例。我们的方法基于自动生成旨在无效的扰动推理步骤，然后测量未扰动步骤的有效性保持和扰动步骤的无效性保持。我们将该方法应用于四个数学数据集中的八个 AF 系统，并观察到普遍的谄媚现象：许多 AF “默默纠正”无效输入为可证明的语句。最能保持有效性的微调 AF 也是最谄媚的，这表明当前 AF 系统在有效性和无效性保持之间存在张力。
cs.CL / 42 / 2608.10939
A Cost-Efficient Routing Pipeline for Multilingual Short-Text Classification Using Small Language Models
基于小型语言模型的多语言短文本分类的成本效益路由管道
Abstract
Multilingual short-text classification supports operational systems such as content moderation, customer support routing, and intent recognition, yet aggregate evaluation often hides large differences between high-resource and low-resource languages. Uniform inference policies are simple to deploy, but they assume that all languages are equally well served. In this work, we evaluate a fixed-list routing strategy that keeps stronger languages on a direct multilingual path and selectively sends weaker languages through translation into English before zero-shot classification. The pipeline is fully self-hosted, uses pretrained compact sentence encoders, and requires no task-specific fine-tuning. We test the approach on two benchmarks chosen to differ in scale and label granularity: a 15-language subset of SIB-200 for seven-way topic classification and a 15-locale subset of MASSIVE for intent classification over an official 60-intent inventory. On SIB-200, the best overall configuration is R1, which translates only the low-resource tier: high-tier and mid-tier Macro-F1 remain unchanged, while low-tier Macro-F1 rises from 0.4632 to 0.6828. On the MASSIVE subset, the same low-tier intervention raises low-tier Macro-F1 from 0.2143 to 0.4417, but the best overall result is obtained by full translation, R3, at Macro-F1 0.4647. Across these two benchmarks, selective translation is a reliable intervention for weaker languages, whereas the optimal routing boundary depends on the task. We therefore report routing through tier-level quality gains and tier-level latency rather than a single global efficiency score.
Chinese Translation
多语言短文本分类支持内容审核、客户支持路由和意图识别等操作系统，但整体评估往往掩盖了高资源语言和低资源语言之间的巨大差异。统一的推理策略易于部署，但假设所有语言都能得到同等的服务。在本研究中，我们评估了一种固定列表路由策略，该策略将较强语言保持在直接的多语言路径上，并在零样本分类之前选择性地将较弱语言通过翻译成英语。该管道完全自托管，使用预训练的紧凑句子编码器，并且不需要特定任务的微调。我们在两个基准上测试了该方法，这两个基准在规模和标签粒度上有所不同：SIB-200的15种语言子集用于七类主题分类，以及MASSIVE的15种语言子集用于官方60个意图清单的意图分类。在SIB-200上，最佳的整体配置是R1，仅翻译低资源层：高层和中层的宏F1保持不变，而低层的宏F1从0.4632上升到0.6828。在MASSIVE子集上，相同的低层干预将低层宏F1从0.2143提高到0.4417，但最佳的整体结果是通过完全翻译R3获得的，宏F1为0.4647。在这两个基准中，选择性翻译是针对较弱语言的可靠干预，而最佳路由边界则取决于任务。因此，我们报告通过层级质量提升和层级延迟进行路由，而不是单一的全局效率评分。
cs.CL / 43 / 2608.10963
REAP: Relation-Aware Elicitation and Parsing for Closed-Book Knowledge Base Construction from LLMs
REAP：基于关系的知识库构建的提取与解析系统（Closed-Book）
Abstract
We present the REAP system for the AKBC Shared Task 2026 on constructing knowledge bases from language models in a closed-book setting, subject to a budget of at most 32B parameters and no model fine-tuning. Our system combines structured chain-of-thought reasoning, relation-specific query strategies, and a reasoning-based empty-set gate to elicit parametric knowledge, followed by direct extraction into valid JSON arrays. On the test set, the system, built on the Mistral-Small-24B-Instruct-2501 model, achieves a macro-F1 score of 0.62, with particularly strong results on countryLandBordersCountry (F1 = 0.95), companyTradesAtStockExchange (F1 = 0.73), and hasArea (F1 = 0.77). Our code is publicly available at https://github.com/yammdd/AKBC-Shared-Task-2026.
Chinese Translation
我们提出了REAP系统，用于2026年AKBC共享任务，在封闭书籍环境下从语言模型构建知识库，预算限制为最多32B参数且不进行模型微调。我们的系统结合了结构化的思维链推理、特定关系的查询策略以及基于推理的空集门控，以提取参数化知识，随后直接提取为有效的JSON数组。在测试集上，基于Mistral-Small-24B-Instruct-2501模型构建的系统实现了0.62的宏F1分数，尤其在countryLandBordersCountry（F1 = 0.95）、companyTradesAtStockExchange（F1 = 0.73）和hasArea（F1 = 0.77）上表现强劲。我们的代码已公开发布在https://github.com/yammdd/AKBC-Shared-Task-2026。
cs.CL / 44 / 2608.10970
ReLTEx: Reliable LLM-based Taxonomy Expansion
ReLTEx：基于可靠大型语言模型的分类法扩展
Abstract
Recent advances in Large Language Models (LLMs) have demonstrated strong capabilities in generating semantically relevant concepts and relations, making them promising tools for taxonomy enrichment. However, directly relying on LLM-generated expansions often leads to noisy, redundant, or hierarchically inconsistent structures, limiting their reliability for automated taxonomy expansion. In this paper, we present ReLTEx, a framework for reliable LLM-based taxonomy expansion. ReLTEx combines LLM-driven candidate generation with structure-aware validation and recursive expansion control to improve the consistency and quality of generated taxonomies by reducing hallucinations. We evaluate the proposed framework using benchmark taxonomies under a masked taxonomy expansion setting and compare multiple validation strategies. Experimental results, supported by both adapted evaluation metrics and human evaluation, demonstrate that ReLTEx produces more reliable and semantically coherent taxonomy expansions.
Chinese Translation
近期大型语言模型（LLMs）的进展展示了其在生成语义相关概念和关系方面的强大能力，使其成为分类法丰富化的有前景工具。然而，直接依赖LLM生成的扩展往往会导致噪声、冗余或层次不一致的结构，限制了其在自动分类法扩展中的可靠性。本文提出了ReLTEx，一个用于可靠LLM基础分类法扩展的框架。ReLTEx结合了基于LLM的候选生成、结构感知验证和递归扩展控制，以通过减少幻觉来提高生成分类法的一致性和质量。我们在掩蔽分类法扩展设置下使用基准分类法评估所提出的框架，并比较了多种验证策略。实验结果，结合适应的评估指标和人工评估，表明ReLTEx能够生成更可靠和语义一致的分类法扩展。
cs.CL / 45 / 2608.10974
MUSE: A Full-Text Cross-Domain Knowledge Base of Scientific Problems, Solutions, and Rationales
MUSE：科学问题、解决方案及其理由的全文本跨领域知识库
Abstract
Scientific papers contain fine-grained records of problem solving: authors mention technical obstacles and methods that were used to address them, often along with reasoning on why those methods were chosen. We introduce MUSE (Mining Underlying Scientific Explanations), a full-text, multi-domain resource of scientific Problem-Solution-Rationale (P-S-R) triplets. We curate 579 expert-annotated full-text paragraphs, with a rich annotation schema covering salient problem, solution, and rationale spans, solves and rationale_of links and conceptual coreference. A modular extraction pipeline scales this annotation to build a high-quality knowledge base of 37K source-grounded P-S-R triplets. We evaluate the extraction components and include a preliminary experiment training a rationale-supervised LLM for scientific problem solving. Interestingly, we find that rationale supervision improves performance on complex, multi-constraint problems but can harm performance on simpler ones.
Chinese Translation
科学论文包含了关于问题解决的细粒度记录：作者提到技术障碍和用于解决这些障碍的方法，通常还包括对为何选择这些方法的推理。我们介绍了MUSE（Mining Underlying Scientific Explanations），这是一个全文本、多领域的科学问题-解决方案-理由（Problem-Solution-Rationale, P-S-R）三元组资源。我们整理了579个专家注释的全文本段落，采用丰富的注释方案覆盖显著的问题、解决方案和理由的范围，以及解决方案与理由之间的链接和概念共指。一个模块化的提取管道将这一注释扩展到构建一个高质量的37K源基础的P-S-R三元组知识库。我们评估了提取组件，并包括了一个初步实验，训练一个以理由为监督的LLM用于科学问题解决。有趣的是，我们发现理由监督在复杂的多约束问题上提高了性能，但在较简单的问题上可能会损害性能。
cs.CL / 46 / 2608.10986
What Iterated Self-Feeding Probes of Language Models Measure, and a test that separates the construction from the model
迭代自馈探针对语言模型的测量及区分构造与模型的测试
Abstract
A growing class of methods probes a language model by feeding it its own output: self-consistency, iterated refinement, agentic loops. We ask what such a probe measures, in a construction chosen to make the question sharp: a ring of token cells resampled in place by the model's own windowed conditional p_r(x_i | x_{i+-r}). The substrate is Glauber dynamics on token sequences and is not new; what we change is the coupling. Advancing two rings that differ in one token under common random numbers makes undamaged copies diverge by exactly zero, so damage spreading becomes measurable where a maximal coupling gives mixing times instead. The answer is that it measures two different things at once, in readings that look alike. Some quantities are fixed by the construction: the damage light cone is kinematic, and the radius scaling of the token-space Lyapunov exponent lambda_ca(r) is model-invariant across 19 models and two scale ladders spanning 70x. Others genuinely track the model: lambda_ca crosses zero at a reproducible point in training, and the attractor share ranks models consistently however the lattice is built. Left undistinguished, the first kind is readily mistaken for the second -- we did so ourselves for four months, and report a phase transition we measured to three decimal places that belongs to the probe rather than to any language model. We give the test that separates them: hold the construction fixed and vary the model, or hold the model fixed and vary the construction, and see which readings move. We validate the instrument by reproduction first, recovering a Domany-Kinzel damage field bit-exactly against an independent prediction, and we report the estimator failures that this discipline caught -- four retracted verdicts, each on a quantity that looked like a measurement. The methodology ships as a package.
Chinese Translation
一类日益增长的方法通过将语言模型的输出反馈给它自身来探测模型：自一致性、迭代细化、代理循环。我们询问这样的探针测量了什么，选择一种构造使问题更加明确：一个由模型自身的窗口条件 p_r(x_i | x_{i+-r}) 重新采样的令牌单元环。底层是基于令牌序列的Glauber动力学，这并不新颖；我们改变的是耦合。在共同随机数下推进两个在一个令牌上不同的环，使未受损的副本之间的差异恰好为零，因此损伤传播在最大耦合给出混合时间的地方变得可测。答案是它同时测量了两种不同的事物，尽管读数看起来相似。一些量由构造固定：损伤光锥是运动学的，而令牌空间Lyapunov指数 lambda_ca(r) 的半径缩放在19个模型和跨越70倍的两个尺度阶梯中是模型不变的。其他则真正追踪模型：lambda_ca在训练中的一个可重复点上穿越零，吸引子份额在模型中一致排名，无论晶格如何构建。如果不加区分，第一种类型很容易被误认为是第二种——我们自己在四个月内就犯了这个错误，并报告了一个我们测量到小数点后三位的相变，这属于探针而非任何语言模型。我们给出了区分它们的测试：保持构造固定并改变模型，或保持模型固定并改变构造，观察哪些读数发生变化。我们首先通过重现来验证该工具，准确恢复了一个Domany-Kinzel损伤场，与独立预测完全一致，并报告了这一学科捕获的估计器失败——四个撤回的裁决，每个都涉及看似测量的量。该方法论作为一个包进行发布。
cs.CL / 47 / 2608.10996
ConRub-Med: Reinforcement Learning with Consensus Rubrics for Open-Ended Medical Question Answering
ConRub-Med：基于共识评分标准的强化学习在开放式医学问答中的应用
Abstract
Reinforcement learning with verifiable rewards has been especially effective in mathematics and coding, where answers can be checked automatically. Many open-ended medical questions lack comparably cheap outcome verifiers: responses may be partly correct, incomplete, or contain clinically consequential errors. Rubrics written or validated by physicians offer strong clinical grounding, but involving experts in every instance is costly. Model-generated rubrics make this supervision scalable. We introduce ConRub-Med to preserve useful distinctions as rubric feedback moves from construction to policy optimization. For each prompt, three heterogeneous language models propose atomic criteria independently; a separate model reviews them, retaining only criteria with semantic support from all three generators. Three-State scoring distinguishes correct coverage, missing information, and incorrect claims. Errors receive negative rather than zero credit. When every response in a complete Group Relative Policy Optimization (GRPO) group receives the same final reward, a pairwise judge provides sequence advantages only if both candidate orders agree, without changing the scalar rewards. Groups without ties use vanilla GRPO. In a blinded study matched by question, two medical experts rate panels from the full pipeline as more clinically relevant than panels produced by one generator. Across the evaluated open models, ConRub-Med ranks first on six of nine benchmarks and achieves the highest medical and generalization averages. Using the resulting rubric dataset of 5,166 prompts, it scores $38.98 \pm 1.04$ (mean $\pm$ SD) on HealthBench-Hard, compared with InfiMed-ORBIT's 33.60 with 8,000 samples and 37.30 with 28,000.
Chinese Translation
具有可验证奖励的强化学习在数学和编程领域特别有效，因为这些领域的答案可以自动检查。然而，许多开放式医学问题缺乏相应廉价的结果验证工具：回答可能部分正确、不完整或包含临床上重要的错误。由医生撰写或验证的评分标准提供了强有力的临床基础，但在每个实例中都涉及专家的成本较高。模型生成的评分标准使这种监督变得可扩展。我们引入了ConRub-Med，以在评分反馈从构建转向策略优化时保留有用的区分。对于每个提示，三个异构语言模型独立提出原子标准；一个单独的模型对它们进行审查，仅保留所有三个生成器具有语义支持的标准。三状态评分区分正确覆盖、缺失信息和不正确的声明。错误获得负分而非零分。当完整的组相对政策优化（GRPO）组中的每个响应获得相同的最终奖励时，成对评审者仅在两个候选顺序一致时提供序列优势，而不改变标量奖励。没有平局的组使用普通GRPO。在一项按问题匹配的盲测研究中，两位医学专家评估完整流程中生成的面板比单一生成器产生的面板更具临床相关性。在评估的开放模型中，ConRub-Med在九个基准中的六个上排名第一，并实现了最高的医学和泛化平均分。使用生成的5,166个提示的评分标准数据集，它在HealthBench-Hard上得分为$38.98 ext{±} 1.04$（均值$ ext{±}$标准差），相比之下，InfiMed-ORBIT在8,000个样本中得分为33.60，在28,000个样本中得分为37.30。
cs.CL / 48 / 2608.11002
On the Limitations of Cross-Lingual Consistency in Multilingual Text-to-image Generation
多语言文本到图像生成中的跨语言一致性局限性研究
Abstract
Text-to-image (T2I) generation has achieved remarkable progress in recent years. However, existing research has largely focused on English-only settings, leaving cross-lingual performance gaps and language-specific effects insufficiently explored. To fill this gap, we introduce LingT2I, a benchmark covering 10 widely used languages with 33K prompts, designed to evaluate cross-lingual effects in both content generation and text rendering. Building on this benchmark, we conduct a comprehensive cross-lingual analysis, uncovering linguistic inequality and language-dependent trade-offs across evaluation dimensions. Beyond quantitative evaluation, we further reveal a range of language-dependent generation patterns, highlighting how linguistic factors and their corresponding cultural contexts systematically impact model outputs. Our benchmark and analysis provide a foundation for studying cross-lingual behavior in T2I generation and facilitate the development of more robust and inclusive models. Code and dataset are available at https://github.com/RISys-Lab/LingT2I.
Chinese Translation
文本到图像生成（T2I）在近年来取得了显著进展。然而，现有研究主要集中在仅使用英语的环境中，跨语言性能差距和语言特定效应尚未得到充分探讨。为填补这一空白，我们引入了LingT2I，一个涵盖10种广泛使用语言、包含33K提示的基准，旨在评估内容生成和文本渲染中的跨语言效应。在此基准的基础上，我们进行了全面的跨语言分析，揭示了语言不平等和评估维度中的语言依赖性权衡。除了定量评估外，我们进一步揭示了一系列语言依赖的生成模式，强调了语言因素及其相应文化背景如何系统性地影响模型输出。我们的基准和分析为研究T2I生成中的跨语言行为提供了基础，并促进了更强大和包容性模型的开发。代码和数据集可在 https://github.com/RISys-Lab/LingT2I 获取。
cs.CL / 49 / 2608.11008
Templated or fully Synthetic? Prompt construction as a confound in measuring LLM political stance beyond writing assistance
模板化还是完全合成？提示构建作为测量大型语言模型政治立场的混淆因素，超越写作辅助
Abstract
Political stance detection in LLMs has long been dominated by closed-ended, multiple-choice political survey questions---originally designed for humans, and thus lacks the realism and nuance of human-AI interactions in the wild, while also being susceptible to sandbagging. The recent IssueBench framework substantially mitigates these limitations with templated prompts anchored in real-world chat logs. Given the rise in non-work-related use of GenAI assistants, we extend IssueBench beyond writing assistance to include two additional tasks, information seeking and opinion sharing. We argue that templated prompts still lack the nuance of real ones, especially for open-ended tasks, and remain recognisable as evaluation artefacts. We propose the use of fully synthetic (LLM-generated) prompts, produced under detailed instructions with real prompts as seeds. We assess the ecological validity of real, templated, and LLM-generated prompts in a small-scale study covering 3 highly contested policy issues and 3 recent geopolitical conflicts. Human and LLM annotators rank LLM-generated prompts as no less realistic than real ones and clearly more realistic than templated ones, and find that they carry their intended intent and stance more clearly; the LLMs separate templated prompts from the other two far more sharply than the humans do. In a case study, templated and LLM-generated prompts yield systematically different stance estimates for the same model, most visibly under neutral framings, where templated prompts overstate the model's leaning in the direction encoded by the topic-and-stance text (filler) slotted into their templates.
Chinese Translation
大型语言模型（LLMs）中的政治立场检测长期以来主要依赖于封闭式的多项选择政治调查问题——这些问题最初是为人类设计的，因此缺乏现实性和人机交互的细微差别，同时也容易受到操控。最近的 IssueBench 框架通过基于真实聊天记录的模板化提示大大减轻了这些局限性。鉴于非工作相关的生成型人工智能（GenAI）助手使用的增加，我们将 IssueBench 扩展到写作辅助之外，增加了两个额外任务：信息获取和意见分享。我们认为，模板化提示仍然缺乏真实提示的细微差别，尤其是在开放式任务中，且仍然可以被识别为评估伪影。我们建议使用完全合成的（LLM生成的）提示，这些提示是在详细指示下，以真实提示为种子生成的。在一项涵盖3个高度争议的政策问题和3个近期地缘政治冲突的小规模研究中，我们评估了真实提示、模板化提示和LLM生成提示的生态有效性。人类和LLM注释者将LLM生成的提示评定为不逊色于真实提示，并明显比模板化提示更具现实性，并发现它们更清晰地传达了预期的意图和立场；LLMs对模板化提示与其他两者的区分明显优于人类。在一个案例研究中，模板化提示和LLM生成提示对同一模型产生系统性不同的立场估计，尤其在中性框架下，模板化提示夸大了模型在主题和立场文本（填充物）所编码方向上的倾向。
cs.CL / 50 / 2608.11025
Data Attribution of Emergent Misalignment with Persona Features
与角色特征相关的突现失调数据归因
Abstract
Emergent misalignment (EM) is the phenomenon where fine-tuning a language model on a narrow task leads to harmful behavior in unrelated domains. A leading mechanistic account attributes EM to persona features: latent directions acquired during pre-training that misaligned fine-tuning amplifies. We ask where these features come from: which pre-training documents activate them, and whether naturally occurring human-written text suffices to induce EM. Using Sparse Autoencoder (SAE) based model diffing across four open-weight models, we find that features related to jailbreak personas, sarcasm, deception, and manipulation are amplified by misalignment fine-tuning, while safety-relevant and assistant-identity features are suppressed. Steering individual features controls EM in both directions: it induces misalignment rates of up to 62% in aligned models -- exceeding the 35% reached by misalignment fine-tuning itself -- and re-aligns misaligned models to near-baseline misalignment rates. Attributing the causal features to a corpus of one million pre-training web documents retrieves semantically relevant narratives about villainous characters, domination, and harmful agency. However, fine-tuning on these human-written documents does not reliably induce EM, even after reformatting into assistant-style responses, whereas synthetic instruction-response pairs derived from the same content do -- and transfer across model families. Semantic relevance alone is therefore not sufficient: response structure or model-generated phrasing plays an important role in inducing EM.
Chinese Translation
突现失调（EM）是指在狭窄任务上微调语言模型导致在无关领域出现有害行为的现象。一种主要的机制性解释将EM归因于角色特征：在预训练期间获得的潜在方向，微调的失调会放大这些方向。我们探讨这些特征的来源：哪些预训练文档激活了它们，以及自然生成的人类书写文本是否足以引发EM。通过对四个开放权重模型进行基于稀疏自编码器（SAE）的模型差异分析，我们发现与越狱角色、讽刺、欺骗和操控相关的特征在失调微调下被放大，而与安全相关和助手身份的特征则被抑制。单独控制这些特征可以在两个方向上调节EM：它在对齐模型中诱发高达62%的失调率——超过了失调微调本身所达到的35%——并将失调模型重新对齐至接近基线的失调率。将因果特征归因于一百万个预训练网络文档的语料库，检索到与恶棍角色、统治和有害代理相关的语义上相关的叙述。然而，即使在重新格式化为助手风格的响应后，对这些人类书写文档的微调也并不能可靠地引发EM，而从相同内容派生的合成指令-响应对却能引发EM，并在模型家族之间转移。因此，单靠语义相关性并不足够：响应结构或模型生成的措辞在诱发EM中起着重要作用。
cs.CL / 51 / 2608.11036
myMediWhisper: Construction of Burmese Medical Speech Corpus and Whisper Fine-Tuning for Clinical Dialogue ASR
myMediWhisper：缅甸医学语音语料库的构建及临床对话自动语音识别的Whisper微调
Abstract
Although Whisper models benefit from large-scale multilingual pre-training, their performance on Burmese medical speech remains limited. This work presents a Burmese medical speech recognition framework built on a high-quality 28-hour corpus recorded and validated by native speakers. We fine-tune Whisper models using full fine-tuning (FFT) and parameter-efficient fine-tuning (PEFT) with LoRA. To evaluate robustness, we apply waveform- and spectrogram-level data augmentation under controlled noise and simulated room acoustics. While augmentation reduces performance on clean speech, it significantly improves robustness in noisy and reverberant environments across FFT and PEFT settings. Our best-performing system, fully fine-tuned myMediWhisper-Medium without augmentation, achieves a state-of-the-art Word Error Rate (WER) of 23.44%, outperforming much larger general-domain fine-tuned models. Dataset and other resources can be found at the Huggingface repository: https://huggingface.co/datasets/LULab/mediTalk-mm-rdy.
Chinese Translation
尽管Whisper模型受益于大规模多语言预训练，但其在缅甸医学语音上的表现仍然有限。本研究提出了一种缅甸医学语音识别框架，该框架基于由母语者录制和验证的高质量28小时语料库。我们使用全微调（Full Fine-Tuning, FFT）和基于LoRA的参数高效微调（Parameter-Efficient Fine-Tuning, PEFT）对Whisper模型进行微调。为了评估鲁棒性，我们在受控噪声和模拟房间声学条件下应用波形和谱图级数据增强。尽管增强在清晰语音上的性能有所下降，但在噪声和混响环境中显著提高了FFT和PEFT设置下的鲁棒性。我们表现最佳的系统是完全微调的myMediWhisper-Medium，未使用增强，达到了23.44%的最新词错误率（Word Error Rate, WER），超越了许多更大规模的通用领域微调模型。数据集及其他资源可在Huggingface仓库找到：https://huggingface.co/datasets/LULab/mediTalk-mm-rdy。
cs.CL / 52 / 2608.11044
TEAMMix: Taxonomy Enrichment Augmentation and Minority-augmented Mixing Strategy for LLM-enhanced Weak-Supervised Hierarchical Text Classification
TEAMMix：基于分类法丰富增强和少数类增强混合策略的LLM增强弱监督层次文本分类
Abstract
Hierarchical Text Classification (HTC), as a critical text mining task, faces challenges such as complex label hierarchies and class imbalance. Existing methods based on large language models (LLMs) struggle to be efficiently applied to this task due to issues like lengthy prompts and loss of label structural information. To address these limitations, this paper proposes a weakly supervised HTC framework enhanced by LLM-based data augmentation. The framework first enriches the label hierarchy semantically through keyword generation and corpus mining, thereby enhancing the model's understanding of labels. Subsequently, it guides the LLM to generate pseudo-samples to mitigate the long-tail problem, and employs a Gaussian mixture model for confidence-based resampling to optimize the quality of generated data. Experimental results demonstrate that the proposed method effectively improves the reliability of LLM-generated pseudo-labels and significantly enhances classification performance on fine-grained and imbalanced datasets.
Chinese Translation
层次文本分类（HTC）作为一项关键的文本挖掘任务，面临复杂标签层次和类别不平衡等挑战。现有基于大型语言模型（LLMs）的方法由于长提示和标签结构信息丢失等问题，难以高效应用于此任务。为了解决这些局限性，本文提出了一种通过LLM增强的数据增强的弱监督HTC框架。该框架首先通过关键词生成和语料库挖掘在语义上丰富标签层次，从而增强模型对标签的理解。随后，它引导LLM生成伪样本以缓解长尾问题，并采用高斯混合模型进行基于置信度的重采样，以优化生成数据的质量。实验结果表明，所提出的方法有效提高了LLM生成的伪标签的可靠性，并显著提升了在细粒度和不平衡数据集上的分类性能。
cs.CL / 53 / 2608.11049
Multiclass Sentiment Analysis for Identifying Political Viewpoints
多类情感分析以识别政治观点
Abstract
The rapid growth of social media has created vast amounts of political discourse, which provides valuable opportunities to analyze public opinions and identify different political perspectives. Sentiment Analysis (SA) is a core task in Natural Language Processing (NLP) that allows the computational study of attitudes and opinions in textual data, and has become increasingly important for understanding political discourse. In this work, we investigate multiclass sentiment analysis of political view- points on social media, that is to automatically discriminate multiple sentiment classes over political issues and figures. To solve this task we design and evaluate two machine-learning approaches based on XGBoost and BERT. We train and evaluate the models on a labeled dataset of political social media posts using standard classification metrics. The experimental results show that the XGBoost model reaches an F1-score of 0.2835 and the BERT- based model reaches an F1-score of 0.2806 on the test set. These results demonstrate the challenge of classifying complex and contextualized political discourse sentiment and provide a baseline for future research in multiclass political sentiment analysis.
Chinese Translation
社交媒体的快速发展产生了大量的政治话语，为分析公众意见和识别不同的政治视角提供了宝贵的机会。情感分析（Sentiment Analysis, SA）是自然语言处理（Natural Language Processing, NLP）中的一项核心任务，它允许对文本数据中的态度和意见进行计算研究，并在理解政治话语方面变得越来越重要。在本研究中，我们探讨了社交媒体上政治观点的多类情感分析，即自动区分与政治问题和人物相关的多种情感类别。为了解决这一任务，我们设计并评估了基于XGBoost和BERT的两种机器学习方法。我们在一个标注的政治社交媒体帖子数据集上训练和评估模型，并使用标准分类指标进行评估。实验结果表明，XGBoost模型在测试集上达到了0.2835的F1-score，而基于BERT的模型达到了0.2806的F1-score。这些结果展示了对复杂和上下文化的政治话语情感进行分类的挑战，并为未来的多类政治情感分析研究提供了基准。
cs.CL / 54 / 2608.11110
Actions Speak Louder than Words: Measuring Cross-Lingual Policy Retention in Tool-Using Agents
行动胜于言辞：测量工具使用代理的跨语言政策保留
Abstract
When a tool-using agent is given the same task in a different language, does it still take the same steps? Multilingual evaluation rarely asks: it compares final answers and discards the actions. Yet those actions are the product: they fix cost and latency, decide how the system fails, and are the only auditable part of its behaviour. We make the action policy the measured object across 8 models, 6 parallel benchmarks and 41 languages (2.38M rollouts). The naive measurement fails: five confounds sit between raw trace similarity and any defensible claim, each able to flip a conclusion. Short traces score higher, empty traces score perfectly, unrelated traces agree by chance over half the time, the gap is capped by each model's reproducibility, and a model asked the same question twice in one language answers differently, leaving no baseline. We remove all five, and every correction makes the effect larger. Divergence proves structural, not sampling noise: it survives greedy decoding in every cell and stays flat as temperature rises, even as models grow less self-consistent. Normalised by their own reproducibility, four very different frontier models converge under greedy decoding, each keeping 71-73% of its action policy across languages, with model identity explaining only 5.7% of the variance. Below roughly 10B parameters it breaks down, and the ordering among smaller models is largely an artifact of a chance floor we measure by permutation rather than assume. Agents route non-English tasks through English; this pivot is causally load-bearing, confirmed by a pre-registered prediction across four models, and models will not abandon it when told to. Finally, a single trace-extraction regex, not the model, manufactured a multilingual failure: two worked examples raise one model's measured accuracy twenty-sixfold while its accuracy on readable outputs barely moves.
Chinese Translation
当一个使用工具的代理在不同语言中被赋予相同任务时，它是否仍然采取相同的步骤？多语言评估很少询问这一点：它比较最终答案并忽略了行动。然而，这些行动才是结果：它们固定了成本和延迟，决定了系统如何失败，并且是其行为中唯一可审计的部分。我们将行动政策作为测量对象，涵盖8个模型、6个平行基准和41种语言（2.38M次回合）。简单的测量失败：五个混淆因素存在于原始轨迹相似性与任何可辩护的主张之间，每个因素都有可能颠覆结论。短轨迹得分较高，空轨迹得分完美，无关轨迹偶然一致超过一半的时间，差距受到每个模型可重复性的限制，而同一模型在一种语言中被问两次同样的问题时给出的答案却不同，导致没有基线。我们去除了这五个因素，每一次修正都使效果更明显。差异证明是结构性的，而非抽样噪声：它在每个单元中都能在贪婪解码中存活，并且随着温度的升高保持平稳，即使模型的自一致性降低。经过自身可重复性标准化后，四个截然不同的前沿模型在贪婪解码下趋于一致，每个模型在不同语言中保留71-73%的行动政策，模型身份仅解释了5.7%的方差。在大约10亿参数以下，它会崩溃，而较小模型之间的排序在很大程度上是我们通过置换测量而非假设的偶然底线的产物。代理通过英语路由非英语任务；这一枢纽是因果负担的，经过四个模型的预注册预测得到了确认，并且模型在被告知时不会放弃它。最后，一个单一的轨迹提取正则表达式，而非模型，制造了多语言失败：两个工作示例使一个模型的测量准确率提高了二十六倍，而其在可读输出上的准确率几乎没有变化。
cs.CL / 55 / 2608.11138
Attention-Path Fragility as an Uncertainty Signal in Large Language Models
大语言模型中注意力路径脆弱性作为不确定性信号
Abstract
We propose that a model's uncertainty about a token is reflected not only in the breadth of its output distribution but also in whether a confident prediction is \emph{fragile} under perturbation of its attention pathways. We instantiate this as ASMI (Attention-Subnetwork Mutual Information), a training-free estimator that masks attention heads and measures the BALD mutual information among the resulting subnetworks, with a semantic-agreement kernel to discount surface-form disagreement. The signal is not a restatement of output confidence: on grounded QA an out-of-fold test shows it adds error-predictive information beyond single-pass confidence and entropy, concentrated in \emph{confident-but-fragile} predictions, where acting on it roughly halves the retained error of a confidence filter. The distinctness is regime-graded, so ASMI predicts its own domain of applicability, strong where answers are routed through provided context and bounded by design where they are recalled from parametric knowledge. Sem-ASMI reads the signal from a single greedy response, without the stochastic generations the strongest baselines require, and ties or beats Semantic Entropy on ten of the twelve grounded benchmark-backbone settings. Across the same twelve settings, the best ASMI variant, typically the adaptive one reusing the ten samples already drawn for the baselines, ties or leads the strongest baseline in eight, significantly in three under a paired test. On parametric QA all variants revert to or below the zero-cost MSP baseline, exactly as predicted, and the estimates are near-deterministic across reruns. A head-level analysis shows that what tracks this boundary is not the presence of head-level fragility but whether that fragility couples to errors.
Chinese Translation
我们提出，模型对一个标记的不确定性不仅体现在其输出分布的广度上，还体现在在注意力路径受到扰动时，是否能够保持自信的预测是 extit{脆弱}的。我们将其具体化为ASMI（注意力子网络互信息），这是一种无训练的估计器，通过屏蔽注意力头并测量由此产生的子网络之间的BALD互信息，同时使用语义一致性核来抵消表面形式的不一致性。该信号并不是输出置信度的重述：在有依据的问答（grounded QA）中，交叉验证测试表明，它提供了超越单次置信度和熵的错误预测信息，集中在 extit{自信但脆弱}的预测上，基于该信号采取行动大约可以将置信过滤器的保留错误减半。该特征是分级的，因此ASMI可以预测其适用领域，在答案通过提供的上下文进行路由时表现强劲，而在从参数知识中回忆时则受到设计的限制。Sem-ASMI从单个贪婪响应中读取信号，无需最强基线所需的随机生成，并在十二个有依据的基准骨干设置中在十个设置上与语义熵持平或超越。在同样的十二个设置中，最佳的ASMI变体，通常是重新利用已经为基线抽取的十个样本的自适应版本，在八个设置中与最强基线持平或领先，在三个设置中在配对测试下显著领先。在参数化问答中，所有变体的表现回落至或低于零成本的MSP基线，正如预测的那样，且在重复运行中估计结果接近确定性。头级分析表明，跟踪这一边界的并不是头级脆弱性的存在，而是这种脆弱性是否与错误相关联。
cs.CL / 56 / 2608.11146
The Illusion of Cross-Lingual Safety in Low-Resource Languages
低资源语言中的跨语言安全错觉
Abstract
Safety alignment in large language models (LLMs) is largely developed in English, assuming these safeguards generalize across multilingual settings. However, this assumption remains underexplored and exposes a vulnerability in low-resource languages. We investigate cross-lingual safety transfer in four African languages, Twi, Hausa, Amharic, and Swahili, using LoDNA, a new safety dataset that pairs literal translations with culturally localized prompts. To move beyond generation-based evaluation, we propose a latent geometric framework that probes hidden-state refusal representations in LLMs. Our experimental results show that cross-lingual safety transfer is severely limited; harmful prompts retain less than 10% of the English refusal signal across most language-model pairs. Literal and localized prompts are semantically aligned (cosine 0.95-0.996) but drift across layers, suggesting models encode the concepts without routing them to safety mechanisms. These findings demonstrate that current multilingual safety alignment is superficial, providing strong evidence against the assumption of a universal, language-agnostic harm manifold within the specific low-resource languages studied. Warning: This paper contains example data that may be offensive or harmful.
Chinese Translation
大型语言模型（LLMs）中的安全对齐主要是在英语环境中开发的，假设这些安全措施可以在多语言环境中普遍适用。然而，这一假设尚未得到充分探讨，并暴露了低资源语言中的脆弱性。我们研究了四种非洲语言（Twi、Hausa、Amharic 和 Swahili）中的跨语言安全转移，使用了 LoDNA，这是一种新的安全数据集，将字面翻译与文化本地化提示配对。为了超越基于生成的评估，我们提出了一种潜在几何框架，用于探测 LLMs 中隐藏状态拒绝表示的特征。我们的实验结果表明，跨语言安全转移受到严重限制；在大多数语言模型对中，有害提示保留的英语拒绝信号不足 10%。字面和本地化提示在语义上是对齐的（余弦相似度 0.95-0.996），但在层间漂移，表明模型编码了这些概念但未将其引导至安全机制。这些发现表明，当前的多语言安全对齐是表面的，为反对在所研究的特定低资源语言中存在普遍的、与语言无关的伤害流形的假设提供了有力证据。警告：本文包含可能令人反感或有害的示例数据。
cs.CL / 57 / 2608.11171
From Interpretability to Control: Insights from Six Years of the TrustNLP Workshop
从可解释性到控制：来自六年TrustNLP研讨会的见解
Abstract
The Workshop on Trustworthy Natural Language Processing (TrustNLP), co-located with major ACL conferences since 2021, has grown from 8 proceedings papers to 41 over six editions, documenting a field-wide transition from post-hoc interpretability of static models to mechanistic understanding and proactive control of generative systems. We synthesize insights from all 144 proceedings papers, classifying them along six trust dimensions grounded in established frameworks (TrustLLM, DecodingTrust). We observe co-occurrences with capability emergence. The release of the first high-impact chat models activated all trust dimensions simultaneously, while subsequent model generations shifted focus toward truthfulness and safety alignment. Analysis from the classification study reveals that truthfulness is the fastest-growing dimension (absent in 2021-2022, comprising 37% of papers by 2025-2026), fairness remains the most consistent theme, and explainability exhibits a U-shaped trajectory; declining as post-hoc methods lost relevance but resurging in 2026 through mechanistic interpretability. A cross-venue comparison with ACL, NAACL, EACL, and EMNLP (~2K papers) in the same period shows that TrustNLP's topical distribution closely follows the field average. We identify four structural insights and conclude with actionable directions for the research community.
Chinese Translation
自2021年以来，与主要ACL会议共同举行的可信自然语言处理研讨会（TrustNLP）已从8篇论文增长至6届的41篇，记录了一个领域范围内的转变，从静态模型的事后可解释性转向对生成系统的机制理解和主动控制。我们综合了144篇论文的见解，按照基于已建立框架（TrustLLM, DecodingTrust）的六个信任维度进行分类。我们观察到能力出现的共现现象。首批高影响力聊天模型的发布同时激活了所有信任维度，而随后的模型世代则将重点转向真实性和安全对齐。分类研究的分析显示，真实性是增长最快的维度（在2021-2022年缺失，到2025-2026年占论文的37%），公平性仍然是最一致的主题，而可解释性则呈现U型轨迹；在事后方法失去相关性时下降，但在2026年通过机制可解释性再次上升。与同一时期的ACL、NAACL、EACL和EMNLP（约2000篇论文）进行的跨场所比较显示，TrustNLP的主题分布与领域平均水平密切相关。我们识别出四个结构性见解，并为研究社区提出可行的方向。
cs.CL / 58 / 2608.11200
ConVAWG: A Retrieval-Grounded Framework for Controlled Synthetic Dialogue Generation in Violence Against Women and Girls
ConVAWG：一个基于检索的框架，用于控制合成对话生成，聚焦于针对女性和女孩的暴力
Abstract
Synthetic dialogue generation offers a way to study conversational dynamics in sensitive domains where real data are difficult to access, release, or annotate. The underlying abuse may occur online or offline: threats and coercion can appear directly in messages, while behaviours such as surveillance, isolation, stalking, and physical violence may be planned, disclosed, or referred to conversationally. Privacy and legal constraints make it difficult the release of large-scale real conversation datasets; existing work has mostly focused on sentence-level toxicity of online abuses, leaving a gap in modelling abuse as a relational and temporally unfolding phenomenon. In this work, we focus on modelling Violence Against Women and Girls (VAWG) scenarios as multi-turn dialogues. We introduce ConVAWG, a retrieval-grounded framework for generating CPS-aligned synthetic VAWG chat dialogues. ConVAWG builds scenarios from persona seeds, demographic patterns reported by the UK Office for National Statistics, official crime definitions, and retrieved Domestic Homicide Review cases; converts them into hierarchical event timelines; generates multi-scene role-play dialogues; and applies targeted activation-steered toxicity control to appropriate utterances. We release over 6,000 multi-turn dialogue events across 200 scenarios with rich scenario-, event-, and turn-level metadata. Extensive human evaluation, LLM-as-Judge assessment, ablations, and downstream tasks show strong dialogue quality and domain fidelity.
Chinese Translation
合成对话生成为研究敏感领域中的对话动态提供了一种方式，这些领域的真实数据难以获取、发布或注释。潜在的虐待可能在线上或线下发生：威胁和强迫可以直接出现在消息中，而诸如监视、孤立、跟踪和身体暴力等行为可能在对话中被计划、披露或提及。隐私和法律限制使得发布大规模真实对话数据集变得困难；现有的研究大多集中于在线虐待的句子级毒性，未能将虐待建模为一种关系性和时间展开的现象。在本研究中，我们专注于将针对女性和女孩的暴力（VAWG）场景建模为多轮对话。我们提出了ConVAWG，一个基于检索的框架，用于生成与CPS（儿童保护服务）对齐的合成VAWG聊天对话。ConVAWG从角色种子、英国国家统计局报告的人口统计模式、官方犯罪定义和检索的家庭凶杀审查案例构建场景；将其转换为分层事件时间线；生成多场景角色扮演对话；并对适当的发言应用针对性的激活引导毒性控制。我们发布了超过6000个跨200个场景的多轮对话事件，附有丰富的场景、事件和轮次级元数据。广泛的人类评估、LLM作为评判者的评估、消融实验和下游任务显示出强大的对话质量和领域保真度。