← Back to Index
Daily Research Digest

arXiv Papers

2026-09-21
287
Papers
4
Categories
287
Translated
收藏清单 0
机器人学 (Robotics)
99
cs.RO / 1 / 2609.20892

WM-VS: Progress-Aligned World Models for Closed-Loop Visual Servoing

WM-VS：用于闭环视觉伺服的进度对齐世界模型
Sun, Guanzhong, Ma, Junyi, Zhou, Yixuan, Wu, Yuxuan, Miao, Yanzi, Wang, Hesheng
Abstract
Closed-loop visual servoing requires predictions that indicate whether an action reduces task error, not only whether the action is plausible. We call this gap the prediction-control mismatch and introduce WM-VS, a target-centric progress-aligned world-model framework for closed-loop visual servoing. Offline target-region DINOv2 correspondences define a signed four-dimensional servo coordinate for translation, scale, and in-plane rotation. Stage 1 aligns action-conditioned latent transitions with this coordinate; Stage 2 freezes the world model and trains a reactive joint-velocity policy with action imitation, consequence supervision, and short imagined rollouts that favor error contraction. Deployment is RGB-only and reactive, without online trajectory optimization. On a real 7-DoF eye-to-hand system, WM-VS reaches a corner RMSE no larger than 10 percent of its initial value in 30/30 trials and retains this criterion at the final valid frame in 25/30 (83.33 percent). Removing future-error alignment reduces retention to 26.67 percent. The learned progress signal agrees with an external AprilTag corner error not used for training or control (mean Spearman rho = 0.8778). Without retraining, two unseen 3D targets achieve translation-error reductions of 86.48 percent and 90.27 percent and rotation-error reductions of 70.01 percent and 65.70 percent. These results link progress-aligned action consequences to repeated closed-loop correction and transfer. Code and data will be released as open source.
Chinese Translation
闭环视觉伺服需要预测能够指示某个动作是否减少任务误差，而不仅仅是该动作是否可行。我们将这一差距称为预测-控制不匹配，并引入WM-VS，一个以目标为中心的进度对齐世界模型框架，用于闭环视觉伺服。离线目标区域的DINOv2对应关系定义了一个用于平移、缩放和面内旋转的有符号四维伺服坐标。第一阶段将动作条件下的潜在转移与该坐标对齐；第二阶段冻结世界模型，并通过动作模仿、后果监督和有利于误差收缩的短期想象回滚训练一个反应性关节速度策略。部署仅使用RGB图像并为反应式，无需在线轨迹优化。在一个真实的7自由度眼手系统中，WM-VS在30次试验中达到的角落均方根误差（RMSE）不超过其初始值的10％，并在30次试验中有25次（83.33％）在最后有效帧保持这一标准。去除未来误差对齐将保留率降低至26.67％。学习到的进度信号与未用于训练或控制的外部AprilTag角落误差一致（平均Spearman rho = 0.8778）。在不重新训练的情况下，两个未见的3D目标实现了86.48％和90.27％的平移误差减少，以及70.01％和65.70％的旋转误差减少。这些结果将进度对齐的动作后果与重复的闭环修正和迁移联系起来。代码和数据将作为开源发布。
View on arXiv Download PDF AI Translation
cs.RO / 2 / 2609.20965

AeRove: A Compact Bimodal Aerial-Terrestrial Drone with Rapid Bistable Reconfiguration for Close-Range Pipeline Inspection

AeRove：一种紧凑的双模空地无人机，具备快速双稳态重构功能，用于近距离管道检测
Polillio, Caleb, Swissler, Petras
Abstract
Close-proximity pipeline inspection is challenging for standard drones due to high hovering power consumption, airflow sensitivity, and propeller wash interference with gas sensing. To address this, we present AeRove, a compact bimodal aerial-terrestrial robot. AeRove uses its propeller guards as wheels to roll along pipes and employs a spring-loaded bistable mechanism to reconfigure between ground and flight modes in 200 ms without continuous actuator power to maintain either state. The converging propeller-guard geometry also improves measured thrust efficiency. By reserving flight for obstacle hopping and using ground rolling for continuous traversal, current draw is reduced 14x compared to continuous flight (0.7 A vs. 10 A), extending estimated travel distance from 144 m to 2057 m. Autonomous trials on a 51 cm diameter steel pipe demonstrated navigation on straight and curved sections, obstacle jumping, and leak detection of simulated inspection markers. Separate CO2 sensing experiments evaluated gas-detection performance under perched and aerial operating conditions. Perched inspection produced a substantially larger concentration response than hovering under the tested conditions. During flight, an underbody propeller-intake configuration produced a faster and stronger gas-detection response than an extended probe by leveraging propeller-induced airflow. Code and designs are released under the CC-BY license.
Chinese Translation
由于高悬停功耗、气流敏感性以及螺旋桨洗涤对气体传感的干扰，标准无人机在近距离管道检测中面临挑战。为了解决这一问题，我们提出了AeRove，一种紧凑的双模空地机器人。AeRove利用其螺旋桨保护罩作为轮子在管道上滚动，并采用弹簧加载的双稳态机制，在200毫秒内在地面模式和飞行模式之间重构，而无需持续的执行器电源来维持任一状态。收敛的螺旋桨保护罩几何形状也提高了测量的推力效率。通过将飞行保留用于障碍跳跃，并使用地面滚动进行连续移动，当前电流消耗相比于持续飞行减少了14倍（0.7 A对比10 A），将估计的行驶距离从144米延长至2057米。在直径为51厘米的钢管上进行的自主试验展示了在直线和弯曲段的导航、障碍跳跃以及对模拟检测标记的泄漏检测。单独的二氧化碳传感实验评估了在栖息和空中操作条件下的气体检测性能。在测试条件下，栖息检测产生的浓度响应明显大于悬停检测。在飞行过程中，底部螺旋桨进气配置利用螺旋桨引起的气流，产生了比延伸探针更快、更强的气体检测响应。代码和设计已在CC-BY许可下发布。
View on arXiv Download PDF AI Translation
cs.RO / 3 / 2609.20970

Shake to Learn: Dynamic Interrogation of Hidden Object Physics for Robotic Manipulation with Physical Reservoir Computing

摇动学习：利用物理储层计算动态探测隐藏物体物理特性以实现机器人操控
Lor, Wen Sin, Wang, Jun, Li, Suyi
Abstract
Many physical properties relevant to robotic manipulation are hidden from vision. A sealed object, for example, may reveal little about its center of mass (COM) or internal contents until it is lifted, shaken, or otherwise dynamically perturbed. This study shows that such interactions can enable a new modality of robotic perception and learning, in which interaction-induced dynamic responses are used to infer object physics that is inaccessible to conventional sensing. We implement this idea using an origami-inspired soft robotic arm that functions as a physical reservoir computer. After grasping an object, the arm is excited by a fixed shaking input at its base, and the resulting ringdown response is recorded through either camera tracking or embedded sensors. Because the input is held constant across trials, hidden object properties, such as the COM position, are encoded through their effect on the dynamics of the coupled robot-object system. A lightweight linear readout can then decode these dynamics to recover interpretable information about the hidden object physics. Using this framework, the soft robotic arm reservoir completed three tasks of increasing difficulty: inferring the orientation of the object's hidden COM, inferring the COM distance from the grasp point, and using the inferred COM information to guide a subsequent regrasp. We further develop a dynamic summary representation of the ringdown response that improves prediction accuracy. Together, these results establish shake-to-learn mechanical interrogation as a promising strategy for robotic systems to convert brief physical interactions into actionable cues about hidden object properties for downstream manipulation.
Chinese Translation
许多与机器人操控相关的物理属性对视觉是隐藏的。例如，一个封闭的物体在被提起、摇动或以其他方式动态扰动之前，可能对其质心（COM）或内部内容几乎没有任何显露。本研究表明，这种交互可以启用一种新的机器人感知和学习方式，其中交互引发的动态响应被用来推断常规传感器无法获取的物体物理特性。我们使用一种受折纸启发的软体机器人手臂来实现这一理念，该手臂作为物理储层计算机运作。在抓取物体后，手臂通过其基部施加固定的摇动输入进行激励，并通过摄像头跟踪或嵌入式传感器记录下随之产生的衰减响应。由于在试验中输入保持不变，隐藏物体的属性（如质心位置）通过其对耦合机器人-物体系统动态的影响被编码。然后，轻量级线性读出可以解码这些动态，以恢复关于隐藏物体物理特性的可解释信息。利用这一框架，软体机器人储层完成了三个逐渐增加难度的任务：推断物体隐藏质心的方向、推断质心与抓取点的距离，以及利用推断出的质心信息指导后续的重新抓取。我们进一步开发了一种动态摘要表示法，以改善预测准确性。这些结果共同确立了摇动学习机械探测作为一种有前景的策略，使机器人系统能够将短暂的物理交互转化为关于隐藏物体属性的可操作线索，以便于后续操控。
View on arXiv Download PDF AI Translation
cs.RO / 4 / 2609.20980

ForeTac-VLA: A Forecasting-Based Tactile-Vision-Language-Action Model for Contact-Rich Robotic Manipulation

ForeTac-VLA：一种基于预测的触觉-视觉-语言-动作模型，用于接触丰富的机器人操控
Tao, Zhengyu, Li, Xin, Wang, Xin
Abstract
Vision-language-action (VLA) models have demonstrated strong capabilities in robotic manipulation, yet their reliance on visual perception limits robustness in contact-rich environments, where critical physical interaction states may not be visually observable. Existing tactile-enhanced VLA methods improve physical grounding using observed tactile feedback, but most remain largely reactive rather than explicitly modeling how contact may evolve. Therefore, we propose ForeTac-VLA, a forecasting-based tactile-vision-language fusion model that predicts future tactile states to guide action generation. Specifically, ForeTac-VLA encodes recent tactile observations into temporal representations and integrates them with vision-language features through bidirectional cross-attention. Further, a transformer-based forecasting module predicts multi-step future tactile states, enabling the model to reason jointly over observed and anticipated contact. Finally, the fused multimodal representations and predicted future tactile states are fed into the VLA backbone to condition action generation. To stabilize training, a ground-truth-to-prediction curriculum is employed when early forecasts are unreliable. Across four real-world contact-rich manipulation tasks, ForeTac-VLA achieves an average success rate of 95%, outperforming the fine-tuned VLA model by 36.25 percentage points and state-of-the-art tactile-enhanced VLA baselines by over 22 percentage points. ForeTac-VLA also maintains strong performance under low-illumination and visually cluttered conditions. Video demonstrations can be found on https://foretac-vla.github.io/
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操控中展现了强大的能力，但其对视觉感知的依赖限制了在接触丰富环境中的鲁棒性，因为关键的物理交互状态可能无法通过视觉观察到。现有的触觉增强VLA方法利用观察到的触觉反馈改善物理基础，但大多数方法仍然主要是反应性的，而不是明确建模接触如何演变。因此，我们提出了ForeTac-VLA，这是一种基于预测的触觉-视觉-语言融合模型，能够预测未来的触觉状态以指导动作生成。具体而言，ForeTac-VLA将最近的触觉观察编码为时间表示，并通过双向交叉注意力与视觉-语言特征进行融合。此外，基于变换器的预测模块预测多步未来触觉状态，使模型能够对观察到的和预期的接触进行联合推理。最后，融合的多模态表示和预测的未来触觉状态被输入到VLA主干中，以调节动作生成。为了稳定训练，当早期预测不可靠时，采用真实值到预测的课程学习。在四个真实世界的接触丰富操控任务中，ForeTac-VLA的平均成功率达到95%，比微调的VLA模型高出36.25个百分点，比最先进的触觉增强VLA基线高出22个百分点以上。ForeTac-VLA在低光照和视觉混乱条件下也保持了强劲的表现。视频演示可在 https://foretac-vla.github.io/ 找到。
View on arXiv Download PDF AI Translation
cs.RO / 5 / 2609.20983

PIVOT: Physically Informed Vision-Language Off-Road Traversability for Field Robot Navigation

PIVOT：基于物理信息的视觉-语言越野可通行性系统用于场地机器人导航
Jiao, Aoran, Zhao, Wenda, Sahak, Hshmat, Barfoot, Timothy D.
Abstract
Terrain assessment is a critical capability for off-road mobile robots, enabling safe and reliable navigation through unstructured and geometrically complex environments. Conventional geometry-based terrain assessment is fast to compute but often overly conservative in unstructured environments. We present PIVOT: a Physically Informed Vision-Language Off-Road Traversability navigation system that augments conventional geometry-based planning with vision-language-model (VLM)-based semantic reasoning for field robots. To physically ground this assessment, we quantify how strongly the VLM's predicted traversal energy cost, robot vibration, and wheel slip correlate with real-world measurements and introduce a unified traversability score that weights each modality by its prediction-measurement correlation. For efficiency, we design a two-level navigation architecture that retains geometry-based planning as the nominal mode and invokes semantic replanning only when that mode fails to find a path. Across five repeated closed-loop trials on a mixed-terrain route totalling around $6.4$ km, the proposed system increases overall autonomy from $59.6\%$ to $97.0\%$, reduces human interventions from $11$ to $3$, and increases the mean distance between interventions (MDBI) from $69.2$ m to $412.9$ m compared with geometry-only navigation. These results demonstrate that physically grounded VLM-based terrain assessment can substantially extend autonomous navigation beyond the limitations of geometry alone, while preserving efficient geometric planning as the nominal mode.
Chinese Translation
地形评估是越野移动机器人一项关键能力，使其能够在非结构化和几何复杂的环境中安全可靠地导航。传统的基于几何的地形评估计算速度快，但在非结构化环境中往往过于保守。我们提出了PIVOT：一种基于物理信息的视觉-语言越野可通行性导航系统，该系统通过视觉-语言模型（VLM）基础的语义推理来增强传统的基于几何的规划。为了将这种评估与实际情况相结合，我们量化了VLM预测的行驶能耗、机器人振动和轮胎打滑与实际测量之间的相关性，并引入了一个统一的可通行性评分，该评分根据每种模态的预测-测量相关性进行加权。为了提高效率，我们设计了一个两级导航架构，保留基于几何的规划作为常规模式，仅在该模式未能找到路径时调用语义重规划。在对总长约6.4公里的混合地形路线进行的五次重复闭环试验中，所提系统将整体自主性从59.6%提高到97.0%，将人工干预次数从11次减少到3次，并将干预之间的平均距离（MDBI）从69.2米增加到412.9米，相较于仅基于几何的导航。这些结果表明，基于物理信息的VLM地形评估能够显著扩展自主导航的能力，超越仅依赖几何的局限，同时保持高效的几何规划作为常规模式。
View on arXiv Download PDF AI Translation
cs.RO / 6 / 2609.21000

Do Spinning Radar Doppler Velocity Measurements Improve Vehicle Detection and Tracking?

旋转雷达多普勒速度测量是否改善了车辆检测和跟踪？
Xie, Eric, Lisus, Daniil, Barfoot, Timothy D.
Abstract
Spinning frequency-modulated continuous-wave (FMCW) radars have been gaining popularity in autonomous vehicle perception on account of their robustness to adverse weather conditions and 360{\deg} field of view. Recently, scanning radars have also been shown capable of generating per-azimuth Doppler velocity. In this paper, we investigate whether these Doppler velocity measurements improve spinning radar vehicle detection and tracking performance. For detection, we estimate the ego motion and use it to undo the Doppler range distortion of the radar image before passing it to a network. For tracking, we propose a new way to estimate a per-vehicle velocity and use it as a prior for the tracker's motion model. Since Doppler-enabled spinning radar data is not available in any dataset with ground-truth dynamic object labels, our first contribution is an automatic labelling pipeline that uses an ensemble of fine-tuned off-the-shelf lidar detectors to label all 643 km of the Boreas Road Trip dataset. We then transfer detections to radar, and use over 250 km of vehicle-dense sequences as ground-truth training data. By training and evaluating two state-of-the-art detectors, we show that Doppler undistortion can improve detection accuracy by up to $2.37$ points on mean average precision. Furthermore, we show that the Doppler velocity prior can improve tracking accuracy by $13.68$ points on multi-object tracking accuracy (MOTA) versus the zero-velocity initialization baseline, while achieving $99.7\%$ of the MOTA obtained using ground-truth velocities as the prior.
Chinese Translation
旋转频率调制连续波（FMCW）雷达因其对恶劣天气条件的鲁棒性和360°视场而在自动驾驶车辆感知中越来越受到欢迎。最近，扫描雷达也被证明能够生成每个方位角的多普勒速度。本文研究了这些多普勒速度测量是否改善了旋转雷达的车辆检测和跟踪性能。在检测方面，我们估计自我运动，并利用它来消除雷达图像的多普勒范围失真，然后将其传递给网络。在跟踪方面，我们提出了一种新的方法来估计每辆车的速度，并将其作为跟踪器运动模型的先验。由于没有任何包含真实动态物体标签的数据集提供支持多普勒的旋转雷达数据，我们的第一个贡献是一个自动标注管道，利用一组经过微调的现成激光雷达检测器对Boreas Road Trip数据集的643公里进行标注。然后，我们将检测结果转移到雷达上，并使用超过250公里的车辆密集序列作为真实训练数据。通过训练和评估两种最先进的检测器，我们展示了多普勒去失真可以将平均精度提高最多2.37个点。此外，我们还表明，与零速度初始化基线相比，多普勒速度先验可以将多目标跟踪精度（MOTA）提高13.68个点，同时达到使用真实速度作为先验时获得的99.7%的MOTA。
View on arXiv Download PDF AI Translation
cs.RO / 7 / 2609.21005

Project SCOUT: Interceptor Drone for Perimeter Defense

项目 SCOUT：用于周界防御的拦截无人机
Yousuf, Azmain, Cai, Siwei, Peterson, Knut, Zhou, Lifeng, Han, David
Abstract
The rapid proliferation of unauthorized unmanned aerial vehicles (UAVs) has created a growing need for robust, jamming-resistant counter-UAV systems for perimeter defense. This paper presents \textbf{SCOUT} (Spatial Computation for Optimized UAV Tracking), a ROS-integrated onboard perception and control framework for real-time aerial defense against incoming UAVs. SCOUT performs visual detection, target association, track filtering, and control command generation directly onboard the defender UAV, without relying on external sensing infrastructure or ground-station computation. To provide stable control inputs, the perception pipeline combines TensorRT-accelerated drone detection with ByteTrack-based association and a lightweight track-retention state machine. The state machine rejects abrupt target jumps and maintains short-term target continuity during temporary detection degradation, reducing unstable control responses caused by false detections or target switching. We evaluate the proposed architecture through an integrated hardware deployment executing a planar ``goalkeeping'' interception strategy. In this setting, the defender UAV tracks the incoming target and adjusts its motion to maintain a blocking configuration near the protected boundary. Real-world flight results show that SCOUT maintains valid target detections for 92.2\% of frames while operating at real-time onboard detection rates, demonstrating the feasibility of visual tracking and closed-loop control for UAV perimeter defense. A video demonstration of the end-to-end perimeter defense operation is available online. https://figshare.com/s/497befc033091fc0b84f
Chinese Translation
未经授权的无人机（UAV）快速扩散导致对强大且抗干扰的反无人机系统在周界防御中需求日益增加。本文提出了 extbf{SCOUT}（优化无人机跟踪的空间计算），这是一个集成在机器人操作系统（ROS）中的机载感知与控制框架，用于实时防御来袭无人机。SCOUT 在防御无人机上直接执行视觉检测、目标关联、轨迹过滤和控制命令生成，而无需依赖外部传感基础设施或地面站计算。为了提供稳定的控制输入，感知流程结合了基于 TensorRT 加速的无人机检测、基于 ByteTrack 的关联以及轻量级的轨迹保持状态机。状态机拒绝突发的目标跳跃，并在临时检测降级期间保持短期目标连续性，从而减少由于错误检测或目标切换引起的不稳定控制响应。我们通过集成硬件部署评估所提出的架构，该部署执行平面“守门”拦截策略。在此设置中，防御无人机跟踪来袭目标，并调整其运动以保持在受保护边界附近的阻挡配置。实际飞行结果表明，SCOUT 在实时机载检测速率下保持92.2%的有效目标检测，展示了视觉跟踪和闭环控制在无人机周界防御中的可行性。关于端到端周界防御操作的视频演示可在线获取。 https://figshare.com/s/497befc033091fc0b84f
View on arXiv Download PDF AI Translation
cs.RO / 8 / 2609.21008

SPARROW: Survival-POMCP for Adaptive Robot Routing, Observation, and Waiting

SPARROW：用于自适应机器人路径规划、观察和等待的生存-POMCP
Sahak, Hshmat, Jiao, Aoran, Rhinehart, Nicholas, Barfoot, Timothy D.
Abstract
Temporary obstacles that may block a robot's planned route create a sequential navigation problem: a robot must decide whether to wait for a blockage to clear, reroute, or acquire more information about the obstacle before acting. We formulate graph navigation among temporary obstacles as a partially observable semi-Markov decision process and introduce SPARROW, a belief-space planner built on Partially Observable Monte Carlo Planning (POMCP). SPARROW searches over traversal, observation, and finite-duration waiting actions while maintaining a particle belief over latent obstacle classes and clearance times. Class-conditioned survival models are learned online from both clearance observations and right-censored encounters where the robot reroutes before clearance is observed. A generative model simulates obstacle arrivals and clearances as each action unfolds, so the planner can account for blockages that may occur along alternative routes. We further introduce a value-of-learning criterion that trades the immediate cost of collecting labelled survival data against its expected reduction in future navigation regret. Across two simulation graphs and multiple obstacle-class settings, SPARROW reduces mean time-to-goal by 12-26% relative to OSCAR, a recent survival-based method for the same problem. On a physical mobile robot, SPARROW reduces mean time-to-goal by 20.5% relative to OSCAR while selectively observing, waiting, and rerouting as environment conditions change.
Chinese Translation
临时障碍物可能阻碍机器人规划路线，导致一个序列导航问题：机器人必须决定是等待障碍物清除、重新规划路线，还是在行动之前获取更多关于障碍物的信息。我们将临时障碍物之间的图导航形式化为部分可观察的半马尔可夫决策过程，并引入SPARROW，一种基于部分可观察蒙特卡罗规划（POMCP）的信念空间规划器。SPARROW在遍历、观察和有限持续等待动作上进行搜索，同时保持对潜在障碍类别和清除时间的粒子信念。通过清除观察和在清除被观察之前机器人重新规划的右删失遭遇，在线学习类别条件生存模型。生成模型在每个动作展开时模拟障碍物的到达和清除，因此规划器可以考虑可能在替代路线中发生的阻塞。我们进一步引入了一种学习价值标准，该标准权衡收集标记生存数据的即时成本与其在未来导航遗憾中预期的减少。在两个仿真图和多个障碍类别设置中，SPARROW相对于OSCAR（一种针对相同问题的最新生存基础方法）减少了12-26%的平均到达目标时间。在一台物理移动机器人上，SPARROW在环境条件变化时，通过选择性观察、等待和重新规划，相对于OSCAR减少了20.5%的平均到达目标时间。
View on arXiv Download PDF AI Translation
cs.RO / 9 / 2609.21015

Towards Effective Visual-Inertial SLAM with Passive-Only Sensors for Low-Cost Autonomous Underwater Vehicles

面向低成本自主水下机器人有效的视觉惯性同时定位与地图构建（VI-SLAM）方法研究
Schwidder, Grant, Widhalm, David, Sattar, Junaed
Abstract
Improvements to Visual-Inertial Simultaneous Localization and Mapping (VI-SLAM) for low-cost autonomous underwater vehicles (AUVs) are critical for transitioning advanced marine robotics from specialized labs to broader research and hobbyist applications. While high-end AUVs typically rely on expensive sensor suites - such as Doppler Velocity Logs (DVLs) and Ultra-Short Baseline (USBL) systems - this work demonstrates that robust, high-quality navigation is achievable using a sub-$10, 000(USD) platform equipped only with inexpensive consumer-grade sensors. By leveraging a similarly priced, open-source AUV, we evaluate the performance of stereo cameras, Micro-electromechanical System (MEMS)-based IMUs, and depth sensors in a fully unconstrained 6-degree-of-freedom (6-DOF) underwater environment. We analyze the efficacy of off-the-shelf SLAM packages and propose optimizations for sensor fusion to mitigate the visual and physical challenges of untethered underwater operation. Our results prove that a usable SLAM solution can be accessible to the masses, providing a benchmark for expectations in demanding, real-time maritime missions without the financial barrier of industrial-grade hardware.
Chinese Translation
对低成本自主水下机器人（AUV）的视觉惯性同时定位与地图构建（VI-SLAM）技术的改进对于将先进的海洋机器人技术从专业实验室转向更广泛的研究和爱好者应用至关重要。尽管高端AUV通常依赖于昂贵的传感器组合，如多普勒速度计（DVL）和超短基线（USBL）系统，但本研究表明，使用仅配备廉价消费级传感器的1万美元以下（USD）平台也能实现稳健、高质量的导航。通过利用一个价格相似的开源AUV，我们评估了立体摄像头、基于微机电系统（MEMS）的惯性测量单元（IMU）和深度传感器在完全无约束的6自由度（6-DOF）水下环境中的性能。我们分析了现成SLAM软件包的有效性，并提出了传感器融合的优化方案，以缓解无绳水下操作的视觉和物理挑战。我们的结果证明，实用的SLAM解决方案可以为大众所用，为在要求苛刻的实时海洋任务中提供期望的基准，而不需要工业级硬件的经济负担。
View on arXiv Download PDF AI Translation
cs.RO / 10 / 2609.21022

Catch Me If You Can: Real-Time Feedback Denoising for Responsive VLAs

抓住我，如果你能：响应式视觉语言行动的实时反馈去噪
Ji, Yiheng, Zhou, Xingru, Sentis, Luis, Seo, Mingyo
Abstract
Vision-Language-Action (VLA) models have shown strong generalization in robotic manipulation by combining semantic knowledge from pretrained vision-language models with expressive action-generation policies. Diffusion-based action generators are particularly effective for modeling temporally coherent action chunks, but these chunks are typically executed open-loop after inference. This limits responsiveness when objects move, contacts change, or the scene evolves during execution. We propose VLA-Feedback, a two-timescale architecture that combines low-frequency diffusion planning with high-frequency visual feedback. Rather than fully denoising an action chunk before execution, VLA-Feedback retains its final denoising step as a lightweight feedback interface, allowing each action to be corrected using the latest observation before it is executed. This design preserves the expressiveness of the diffusion planner while enabling real-time action correction without rerunning the full vision-language diffusion model. VLA-Feedback matched GR00T on static LIBERO tasks while improving average success on dynamic simulation tasks from 27.5% to 85.0%. On real-robot tasks, it improved average success from 51% to 73%. Additional materials can be found on our project page: https://vla-feedback.github.io.
Chinese Translation
视觉语言行动（VLA）模型通过将预训练视觉语言模型的语义知识与表达性动作生成策略相结合，在机器人操作中表现出强大的泛化能力。基于扩散的动作生成器在建模时间一致的动作片段方面特别有效，但这些片段通常在推理后以开放环路的方式执行。这限制了在物体移动、接触变化或场景在执行过程中演变时的响应能力。我们提出了VLA-Feedback，一种结合低频扩散规划和高频视觉反馈的双时间尺度架构。VLA-Feedback并不是在执行之前完全去噪一个动作片段，而是将其最终去噪步骤保留为轻量级反馈接口，允许在执行之前使用最新观察结果对每个动作进行修正。这种设计保留了扩散规划器的表达能力，同时实现了实时动作修正，而无需重新运行完整的视觉语言扩散模型。VLA-Feedback在静态LIBERO任务上与GR00T相匹配，同时将动态仿真任务的平均成功率从27.5%提高到85.0%。在真实机器人任务中，其平均成功率从51%提高到73%。更多材料可以在我们的项目页面找到：https://vla-feedback.github.io。
View on arXiv Download PDF AI Translation
cs.RO / 11 / 2609.21024

Adapting Rigid-Body Dynamics Derivatives for Constraint Embedding Closed-Chain Models

适应刚体动力学导数以嵌入约束的闭链模型
Volpi, Daniel J., Wensing, Patrick M.
Abstract
This paper extends an existing algorithm for the first-order derivatives of rigid-body dynamics to the case of closed-chain kinematic systems modeled using constraint embed- ding. Many standard dynamics algorithms apply to both open- chain and constraint-embedded models, but existing efficient derivative methods assume joint velocity effects are locally config- uration invariant. We remove this assumption and derive adapted algorithms that extend dynamics derivatives to more general joint types, including those arising in constraint-embedded closed- chain models. Our results compare conventional pin-joint robot models with more complete actuation models that capture local closed chains. We show that the additional terms introduced by these generalizations have low computational impact when mod- eling actuation kinematics alone, but can incur higher cost when additional rigid bodies, such as motor rotors, are included in the actuation chain, or when considering non-local loops. Overall, these results enable more accurate dynamics computations for constraint-embedded actuation submechanisms to be adopted in model-predictive control and differentiable simulation.
Chinese Translation
本文将现有的刚体动力学一阶导数算法扩展到使用约束嵌入建模的闭链运动系统的情况。许多标准动力学算法适用于开放链和嵌入约束模型，但现有的高效导数方法假设关节速度效应在局部配置上是不变的。我们去除了这一假设，并推导出适应性算法，将动力学导数扩展到更一般的关节类型，包括那些出现在嵌入约束的闭链模型中的关节。我们的结果比较了传统的销接机器人模型与更完整的驱动模型，后者捕捉了局部闭链。我们表明，这些推广所引入的附加项在仅建模驱动运动学时对计算的影响较小，但在驱动链中包含额外的刚体（如电机转子）或考虑非局部回路时，可能会产生更高的计算成本。总体而言，这些结果使得在模型预测控制和可微分仿真中采用嵌入约束的驱动子机制进行更精确的动力学计算成为可能。
View on arXiv Download PDF AI Translation
cs.RO / 12 / 2609.21026

MAPLE-RF: Efficient Probabilistic RF Source Localization in Partially Explored Environments

MAPLE-RF：在部分探索环境中高效的概率射频源定位
Lei, Haozhe, Rangan, Sundeep
Abstract
Localizing a radio-frequency (RF) transmitter from received signals often requires a model of the environment to predict how obstacles block and reflect the signal. In many robotic applications, however, only a partial map is available, particularly when a robot localizes the source while exploring with simultaneous localization and mapping (SLAM). We study single-snapshot transmitter localization on such partially explored maps and compare two approaches that output a posterior over transmitter locations. The first extends a digital-twin method, which ray-traces every candidate location, to partial maps by treating unexplored space as free and training on mixed map coverage. The second, MAPLE-RF, encodes estimated path angles of arrival and signal-to-noise ratios as grid channels aligned with map knownness, occupancy, and line-of-sight visibility, and a U-Net scores all candidate positions in one pass without simulating propagation at inference. Ray-tracing simulations of indoor environments indicate that training on mixed map coverage is essential for both approaches. The digital-twin approach is more accurate on most single-snapshot metrics, while MAPLE-RF comes close at a query cost that does not depend on the propagation model and is more than two orders of magnitude below a fresh full-grid query with general-purpose ray tracing. Both outperform Gaussian and Gaussian-mixture baselines, and on exploration routes guided by its own estimates, fused MAPLE-RF posteriors place more probability near the source than the compared methods. Code and data will be released.
Chinese Translation
从接收到的信号中定位射频（RF）发射器通常需要环境模型，以预测障碍物如何阻挡和反射信号。然而，在许多机器人应用中，尤其是在机器人通过同时定位与地图构建（SLAM）进行探索时，仅有部分地图可用。我们研究了在这种部分探索地图上进行单快照发射器定位，并比较了两种输出发射器位置后验的方法。第一种方法扩展了数字双胞胎方法，通过将未探索空间视为自由区域并在混合地图覆盖上进行训练，将其应用于部分地图。第二种方法，MAPLE-RF，将估计的到达路径角度和信噪比编码为与地图已知性、占用情况和视线可见性对齐的网格通道，并且U-Net在一次传递中对所有候选位置进行评分，而无需在推理时模拟传播。室内环境的光线追踪模拟表明，在混合地图覆盖上进行训练对两种方法都是至关重要的。数字双胞胎方法在大多数单快照指标上更为准确，而MAPLE-RF在查询成本上接近，且不依赖于传播模型，其成本比使用通用光线追踪进行全网格新查询低两个数量级。两者均优于高斯和高斯混合基线，并且在由其自身估计引导的探索路径上，融合的MAPLE-RF后验在源附近赋予了比比较方法更高的概率。代码和数据将会发布。
View on arXiv Download PDF AI Translation
cs.RO / 13 / 2609.21045

DEXTERA: From a Single Image to Deployable Dexterous Manipulation via Real-to-Sim-to-Real

DEXTERA：通过真实-模拟-真实的方式从单幅图像实现可部署的灵巧操作
Wu, Jin, Yuan, Lianjie, Sun, Zeyan, Lei, Yuanyuan, A, Disi, Han, Bicheng, Xia, Fangzhou
Abstract
Collecting real-world robot data for dexterous manipulation is costly and time-consuming. While high-fidelity physics simulators enable scalable data synthesis and policy learning, constructing deployment-ready digital twins manually remains labor-intensive, and residual visual, geometric, and dynamics gaps hinder reliable sim-to-real transfer. We present DEXTERA, an automated real-to-sim-to-real framework that transforms a single RGB image into deployable policies for dexterous manipulation across four unified stages: (1) single-image scene factorization into a static Gaussian background and interactive rigid or articulated assets with VLM-inferred physical parameters; (2) metric scene global alignment, object canonicalization, and morphology-balanced robot calibration; (3) scalable simulator task primitive construction, VR teleoperation, and object-centric trajectory synthesis; and (4) a shared multimodal policy interface supporting both imitation learning and reinforcement learning. We evaluate DEXTERA across 13 task-embodiment pairs, 2 dexterous robot platforms, and 6 policy architectures. Experimental results demonstrate that DEXTERA achieves superior visual fidelity and 3D geometric reconstruction compared to generative baselines, while cross-domain trajectory replays validate strong physical interaction consistency. Furthermore, simulation-only trained policies enable viable zero-shot real-robot deployment, while simulation-real co-training substantially improves mean physical policy success from 29.2% to 61.9% across diverse policy architectures.
Chinese Translation
收集真实世界的机器人数据以进行灵巧操作既昂贵又耗时。尽管高保真物理模拟器能够实现可扩展的数据合成和策略学习，但手动构建可部署的数字双胞胎仍然劳动密集，残余的视觉、几何和动力学差距阻碍了可靠的模拟到现实转移。我们提出了DEXTERA，一个自动化的真实-模拟-真实框架，将单幅RGB图像转化为可部署的灵巧操作策略，分为四个统一阶段：（1）将单幅图像场景分解为静态高斯背景和具有VLM推断物理参数的交互刚性或关节资产；（2）进行度量场景全局对齐、物体标准化和形态平衡的机器人标定；（3）可扩展的模拟器任务原语构建、虚拟现实远程操作和以物体为中心的轨迹合成；（4）支持模仿学习和强化学习的共享多模态策略接口。我们在13个任务-体现对、2个灵巧机器人平台和6种策略架构上评估了DEXTERA。实验结果表明，DEXTERA在视觉保真度和3D几何重建方面优于生成基线，而跨域轨迹重放验证了强大的物理交互一致性。此外，仅通过模拟训练的策略能够实现可行的零-shot真实机器人部署，而模拟-真实共同训练显著提高了不同策略架构下的平均物理策略成功率，从29.2%提高到61.9%。
View on arXiv Download PDF AI Translation
cs.RO / 14 / 2609.21046

Constraint-Unified MPC for Over-Actuated Surface Vehicles with Post-Detection Fault Reconfiguration

针对过驱动表面车辆的约束统一模型预测控制与故障重配置
Burmester, Sebastian, Jiang, Ruiheng, Sendlhofer, Noa, D'Andrea, Raffaello, Ramachandran, Aswin
Abstract
Choreographed aquatic performances require small autonomous surface vehicles to track precise paths under per-thruster force and rate limits, including after thruster failures. We report a deployed system in which trajectory tracking, thrust allocation, and the per-thruster force and rate limits are resolved in a single quadratic program over the per-thruster commands, with fault reconfiguration entering through one binary flag per thruster from an external detector. The system has driven a fleet in live performances on Lake Z\"urich and at the Time Space Existence 2025 exhibition in Venice. The field campaign measures 1.6 cm root mean square position error in a 10-minute hold and 4.3 cm over a 10 m square at 0.6 m/s. At 0.5 m/s, losing the front thruster increases the error to 11.6 cm, while losing the starboard-side thruster increases it to 11.5 cm. Losing two thrusters simultaneously leaves the craft tracking a 0.4 m/s square with a root mean square position error of 1.25 m. Compared to our own cascaded baseline, the unified formulation tracks the nominal square to the same few centimeters and holds station more tightly with roughly half the thruster force. The architectures separate after a thruster failure, where the unified controller stays within 24 cm of the reference path while the cascade leaves it.
Chinese Translation
编排的水下表演要求小型自主表面车辆在每个推进器的力和速率限制下精确跟踪路径，包括在推进器故障后。我们报告了一个已部署的系统，其中轨迹跟踪、推力分配以及每个推进器的力和速率限制通过对每个推进器命令的单一二次规划进行解决，故障重配置通过来自外部探测器的每个推进器的一个二进制标志进入。该系统已在苏黎世湖和2025年威尼斯时间空间存在展览中驱动了一支舰队进行现场表演。现场活动测得在10分钟保持期间的均方根位置误差为1.6厘米，在0.6米/秒的速度下，在10米的正方形区域内的误差为4.3厘米。在0.5米/秒的速度下，失去前推进器会使误差增加到11.6厘米，而失去右舷推进器则使误差增加到11.5厘米。同时失去两个推进器会使船只在0.4米/秒的速度下跟踪正方形，均方根位置误差为1.25米。与我们自己的级联基线相比，统一的公式在同样的几厘米内跟踪名义正方形，并以大约一半的推进器力更紧密地保持位置。在推进器故障后，这些架构会分离，统一控制器在参考路径的24厘米范围内保持，而级联控制器则偏离了路径。
View on arXiv Download PDF AI Translation
cs.RO / 15 / 2609.21059

PlantShade: Predicting Plant Shadows for Lighting-Aware Robotic Agricultural Operation

PlantShade：用于光照感知机器人农业作业的植物阴影预测
Da, Longchao, Liu, Xiaoou, Li, Xingjian, Xiang, Lirong, Wei, Hua
Abstract
Plant growth and agricultural production form the foundation of a country's sustainable development and directly impact human livelihoods. Recent advances in frontier artificial intelligence have enabled scientific agriculture with strong potential to improve crop productivity. In this paper, we identify the importance and inherent complexity of plant shade simulation, as shading is a critical factor influencing plant growth. To advance this field and promote broader societal benefits, we focus on two main contributions. First, we introduce a comprehensive plant growth and shade dataset covering four plant species, including soybean, tomato, sugarbeet, and strawberry. The dataset includes top-down viewpoints with a supplementary light along a circular trajectory, casting dynamic shadows across multiple growth stages and diverse observation complexities. Second, we propose generative shade simulation based on diffusion models, enabling realistic shade generation for unseen plants and supporting downstream robotic tasks such as perception, lighting control, and view planning. The model incorporates temporal conditioning to facilitate flexible shade simulation across different time stages. We conduct both quantitative and qualitative evaluations to assess model performance. This work provides a foundational study for plant-aware shade modeling and has meaningful implications for broader agricultural and robotic applications.
Chinese Translation
植物生长和农业生产是一个国家可持续发展的基础，直接影响人类生计。近年来，前沿人工智能的进展使得科学农业成为可能，具有显著提高作物生产力的潜力。本文识别了植物阴影模拟的重要性及其内在复杂性，因为阴影是影响植物生长的关键因素。为了推动这一领域的发展并促进更广泛的社会效益，我们重点关注两个主要贡献。首先，我们介绍了一个全面的植物生长和阴影数据集，涵盖四种植物物种，包括大豆、番茄、甜菜和草莓。该数据集包括从顶部视角观察的图像，配有沿圆形轨迹的补充光源，能够在多个生长阶段和不同观察复杂性下投射动态阴影。其次，我们提出了一种基于扩散模型的生成阴影模拟方法，使得对未见植物的真实阴影生成成为可能，并支持下游机器人任务，如感知、光照控制和视图规划。该模型结合了时间条件，以便在不同时间阶段灵活地进行阴影模拟。我们进行了定量和定性评估，以评估模型性能。这项工作为植物感知的阴影建模提供了基础研究，并对更广泛的农业和机器人应用具有重要意义。
View on arXiv Download PDF AI Translation
cs.RO / 16 / 2609.21082

Design of Adaptive PID Controller Based On Asynchronous Advantage Actor Critic Learning Method for QuadCopter Control

基于异步优势演员-评论家学习方法的自适应PID控制器设计用于四旋翼控制
Jokar, Ali, Alasty, Aria
Abstract
Quadcopters offer great utility in many applications, but their nonlinear nature and disturbance sensitivity present great control challenges. Basic PID controllers are generally not sophisticated enough to cope with these complexities. This paper suggests a control system that integrates the Asynchronous Advantage Actor-Critic (A3C) algorithm with a PID controller for quadcopter attitude and trajectory tracking. The A3C controller uses parallel agents to optimize PID parameters dynamically using a neural network. A system identification module for the complementary system makes predictions about system states for optimal control policy. The proposed framework was compared with a standard actor-critic (A2C) model. Simulation results verify that they both track accurately. However, the A3C-based controller converges much more for the loss function, as evidenced by reward figures and loss curves, demonstrating better parameter optimization. This shows that A3C-based approach results in improved performance for the control of quadcopter, effectively integrating reinforcement learning and traditional control to achieve higher adaptability.
Chinese Translation
四旋翼在许多应用中具有很大的实用性，但其非线性特性和对干扰的敏感性带来了巨大的控制挑战。基本的PID控制器通常不够复杂，无法应对这些复杂性。本文提出了一种控制系统，将异步优势演员-评论家（A3C）算法与PID控制器结合，用于四旋翼的姿态和轨迹跟踪。A3C控制器利用并行代理通过神经网络动态优化PID参数。一个用于互补系统的系统识别模块对系统状态进行预测，以实现最佳控制策略。所提出的框架与标准的演员-评论家（A2C）模型进行了比较。仿真结果验证了两者都能准确跟踪。然而，基于A3C的控制器在损失函数的收敛性上明显更好，奖励数据和损失曲线表明了这一点，展示了更好的参数优化。这表明基于A3C的方法在四旋翼控制中提高了性能，有效地将强化学习与传统控制相结合，实现了更高的适应性。
View on arXiv Download PDF AI Translation
cs.RO / 17 / 2609.21099

Dynamic Modeling and LQR Control of a Single Coaxial Drone with 2DOF Thrust Vectoring Mechanism

具有2自由度推力矢量机制的单轴对称无人机的动态建模与LQR控制
Jokar, Ali, Talaeizadeh, Amin, Alasty, Aria
Abstract
Coaxial rotor drones have generated considerable interest because of energy efficiency and small size, but they are afflicted with inherent underactuation for roll and pitch control, although systems like swashplates have circumvented this limitation at the cost of greater mechanical complexity. This work presents a novel coaxial drone supplemented by a two-degreesof-freedom pendulum mechanism for active thrust vectoring that offers a less mechanically complicated alternative. We develop a comprehensive Lagrangian dynamic model that does not ignore the inertial contributions of all the components, including body, servo arms, and motor assembly. A Linear Quadratic Regulator(LQR) is designed based on the linearized dynamics around the hover equilibrium. High-fidelity simulations taking actuator dynamics and sensor noise into account validate the proposed architecture. An Extended Kalman Filter (EKF) blends GPS, barometer, and IMU estimates with high accuracy for state estimation. The findings verify the potential and reliability of this approach for power-saving, rapid coaxial UAVs.
Chinese Translation
对称旋翼无人机因其能效高和体积小而引起了广泛关注，但在滚转和俯仰控制方面存在固有的欠驱动问题，尽管像摆板这样的系统以更大的机械复杂性规避了这一限制。本研究提出了一种新型的对称无人机，辅以一种具有两个自由度的摆锤机制，用于主动推力矢量控制，提供了一种机械复杂性较低的替代方案。我们开发了一个全面的拉格朗日动态模型，未忽略所有组件的惯性贡献，包括机身、舵机臂和电机组件。基于悬停平衡点附近的线性化动态设计了线性二次调节器（LQR）。考虑到执行器动态和传感器噪声的高保真仿真验证了所提出的架构。扩展卡尔曼滤波器（EKF）以高精度将GPS、气压计和IMU估计融合用于状态估计。研究结果验证了该方法在节能、快速对称无人机方面的潜力和可靠性。
View on arXiv Download PDF AI Translation
cs.RO / 18 / 2609.21100

Dynamics-Induced Commitment in Learning-Based Robotic Penalty Kicks

基于动态的学习型机器人罚球中的承诺
Geng, Ruize, Zhang, Hao E., Li, Yisen, Wang, Yikai, Tseng, H. Eric, Zhao, Ding
Abstract
Learning in robotic games is constrained not only by strategic information but also by what the body can still execute. We study this coupling in a hierarchical humanoid-quadruped penalty system in which game-level self-play policies command fixed soccer whole-body controllers (S-WBCs). The humanoid shooting skill is initialized from self-collected motion-capture data, whereas the quadruped saving skill is learned by reinforcement learning. We introduce dynamics-induced commitment mapping (DIC-Map), a body-grounded analysis that estimates continuation capability, identifies the first persistent loss of a terminal alternative, and tests whether the remaining interaction admits a reduced zero-sum game. For symmetric terminal alternatives, the reduced game yields a closed-form bound on optimal strategy concentration determined by the responder's value of deferring. We further show that, when the responder acts through an estimator, equal response values eliminate the direct terminal-allocation gradient and leave an estimator-mediated first-order learning channel. Experiments locate commitment about 0.29 s before contact, and changing only ball speed shifts deferral coverage. Across four responder policies, replacing the estimator raises save rate from 0.240 to 0.472, whereas a comparable gain in read accuracy obtained by waiting raises it only to 0.246. Posterior analysis is used for the equilibrium comparison because the available coverage terms are observational proxies. Project website: https://chris-ruizegeng.github.io/penaltykick/
Chinese Translation
机器人游戏中的学习不仅受到战略信息的限制，还受到身体能够执行的操作的限制。我们研究了这一耦合关系，构建了一个分层的人形-四足罚球系统，其中游戏级自我对弈策略指挥固定的足球全身控制器（S-WBCs）。人形射门技能基于自收集的动作捕捉数据初始化，而四足扑救技能则通过强化学习获得。我们引入了动态诱导承诺映射（DIC-Map），这是一种基于身体的分析方法，用于估计持续能力，识别终端替代方案的首次持续损失，并测试剩余交互是否允许简化的零和游戏。对于对称的终端替代方案，简化游戏提供了一个关于最佳策略集中度的封闭形式界限，该界限由反应者推迟的价值决定。我们进一步展示，当反应者通过估计器行动时，相等的响应值消除了直接的终端分配梯度，并留下了一个由估计器介导的一阶学习通道。实验结果表明，承诺发生在接触前约0.29秒，仅改变球速就会改变推迟覆盖范围。在四种反应者策略中，替换估计器使扑救率从0.240提高到0.472，而通过等待获得的可比读取准确率提升仅将其提高到0.246。后验分析用于平衡比较，因为可用的覆盖项是观察代理。项目网站：https://chris-ruizegeng.github.io/penaltykick/
View on arXiv Download PDF AI Translation
cs.RO / 19 / 2609.21107

Learning Scene-Aware Humanoid Locomotion through 3D Clutter from Immersive Human Demonstrations

通过沉浸式人类示范学习场景感知的人形机器人运动穿越三维杂物
Wang, Beichen, Xu, Tong, Kosukhin, Daniel, Yeung, Yuen-Hei, Lu, Yuanjie, Xiao, Xuesu
Abstract
While learning from human motions has enabled highly dynamic humanoid skills such as dancing and martial arts in obstacle-free space, traversal through densely cluttered environments remains underexplored. These spaces are three-dimensional and geometrically constrained, requiring scene-aware locomotion that tightly couples whole-body motion with scene geometry for obstacle avoidance. To address these challenges, we present Moving Through Clutter (MTC), a learning-from-demonstration framework for scene-aware humanoid locomotion. To bypass costly physical scene construction, MTC uses procedurally generated Virtual Reality environments for immersive data collection. To transform these human motions into training-ready humanoid motions, we propose a scene-aware motion retargeting algorithm that converts human demonstrations into humanoid trajectories while strictly enforcing robot-scene clearance to guarantee collision-free traversal. These reference trajectories are then used to train a scene-aware locomotion policy that deploys on a Unitree G1 humanoid. Evaluated on our proposed MTC-Challenge for multi-obstacle traversal, the policy demonstrates a 70.2% collision-free rate across diverse scenarios, successfully traversing complex environments through diverse whole-body skills, including crawling through low-clearance passages and squeezing through narrow gaps.
Chinese Translation
尽管从人类动作中学习使得在无障碍空间中实现高度动态的人形机器人技能（如舞蹈和武术）成为可能，但在密集杂物环境中的穿越仍然未得到充分探索。这些空间是三维的，并且在几何上受到限制，需要场景感知的运动，这要求将全身运动与场景几何紧密结合以避免障碍。为了解决这些挑战，我们提出了“穿越杂物”（Moving Through Clutter, MTC），这是一个用于场景感知人形机器人运动的学习示范框架。为了绕过昂贵的物理场景构建，MTC使用程序生成的虚拟现实环境进行沉浸式数据收集。为了将这些人类动作转化为适合训练的人形机器人动作，我们提出了一种场景感知的动作重定向算法，该算法将人类示范转换为人形机器人轨迹，同时严格执行机器人与场景的间隙，以确保无碰撞穿越。这些参考轨迹随后用于训练一个场景感知的运动策略，该策略在Unitree G1人形机器人上部署。在我们提出的MTC挑战中进行多障碍物穿越评估时，该策略在各种场景中展示了70.2%的无碰撞率，成功通过复杂环境，展现了多样的全身技能，包括在低净空通道中爬行和挤过狭窄缝隙。
View on arXiv Download PDF AI Translation
cs.RO / 20 / 2609.21112

Demonstration Synthesis from a Single Scan via Gaussian Splatting for Visuomotor Policy Learning

通过高斯点云合成单次扫描演示以实现视觉运动策略学习
Wang, Beichen, Yeung, Yuen-Hei, Devarakonda, V. R. Sridhar, Xiao, Xuesu
Abstract
Training a visuomotor policy calls for abundant demonstrations that closely match the target environment, yet collecting them anew remains expensive. Existing demonstration synthesis methods reduce this cost but remain constrained by high manual effort, limited visual fidelity, or heavy reliance on physics simulators. This paper introduces GaussianFactory, a high-fidelity data engine that mass-produces demonstrations with a single video scan as its only human input and no physics engine in the generation loop. Specifically, GaussianFactory reconstructs the scene as an editable 3D Gaussian Splatting (3DGS) replica and samples from the object-combination tasks the scene affords. For each task, it plans grasps and trajectories purely kinematically on the geometric reconstruction, rendering photorealistic demonstrations that visually match the target environment. Physical dynamics enter the pipeline only where contact force interactions dictate the outcome---during grasp formation, via a learned contact model pretrained once on an interaction dataset. To evaluate the downstream utility of the synthesized demonstrations, we implement an end-to-end scan-to-deployment workflow in two setups: a simulated scene that stands in for the real world to enable reproducibility, and a real-world workspace with a physical UR10e robot. In each setup, a standard diffusion policy trained solely on the synthesized demonstrations achieves 95.1% and 84.2% success rates, respectively.
Chinese Translation
训练视觉运动策略需要大量与目标环境紧密匹配的演示，但重新收集这些演示的成本依然很高。现有的演示合成方法虽然降低了这一成本，但仍然受到高人工成本、有限的视觉真实感或对物理模拟器的高度依赖的限制。本文提出了GaussianFactory，一个高保真数据引擎，它仅通过单个视频扫描作为唯一的人类输入，并且在生成过程中不依赖物理引擎，能够大规模生产演示。具体而言，GaussianFactory将场景重建为可编辑的3D高斯点云（3D Gaussian Splatting）复制品，并从场景所提供的物体组合任务中进行采样。对于每个任务，它在几何重建上纯粹通过运动学规划抓取和轨迹，从而生成与目标环境在视觉上匹配的照片级真实感演示。物理动态仅在接触力交互决定结果的环节进入流程——在抓取形成期间，通过一个在交互数据集上预训练的学习接触模型。为了评估合成演示的下游效用，我们在两个设置中实现了端到端的扫描到部署工作流程：一个模拟场景替代现实世界以实现可重复性，以及一个配备物理UR10e机器人的真实工作空间。在每个设置中，仅基于合成演示训练的标准扩散策略分别达到了95.1%和84.2%的成功率。
View on arXiv Download PDF AI Translation
cs.RO / 21 / 2609.21114

Noctif3R: Feed-Forward Monocular Real-Time SLAM for Photon-Limited Scenes on Embedded Hardware

Noctif3R：用于嵌入式硬件的前馈单目实时SLAM在光子受限场景中的应用
Chauhan, Mihir, Abhang, Aditya Uday, Mathew, Kevin Biju, Bera, Aniket
Abstract
Robots carrying out tasks in dark environments need to localize from a single RGB camera, in light so low that the per-pixel signal approaches the sensor's own noise, on a power-constrained onboard computer, in real time. Each of these constraints has matured pipelines, but the intersection does not. Offline low-light reconstruction now recovers structure below -4 dB but is far too slow to run in real time, while the real-time monocular systems a robot can actually carry (DROID-SLAM, DPV-SLAM, etc.) degrade or fail when SNR gets low. We measured how they fail: across the nine lowest darkness levels of our scenes, DROID-SLAM returns a full-length trajectory carrying no information about the camera's motion on all nine, VGGT-SLAM and CUT3R on eight, pi^3 on seven, and DPV-SLAM on four. We present SYS, a monocular pipeline built on a low-light feed-forward pointmap front end with an explicit match gate, which returns three tracked trajectories and no uninformative ones, at the lowest error of any method where it tracks (24-47% of the no-information ceiling against 56-73% for the strongest baseline), and at the narrowest coverage. On a real robot video take in which 86.5% of delivered frames are entirely black, every configuration of ours stops after the lit beginning, while DROID-SLAM and DPV-SLAM each emit a pose for all 1178 frames. Our method contribution is an embedded execution path for the Jetson AGX Orin: running the map, keyframes and backend at 384 pixels with tracking at 256, together with two fixes to the per-frame pose solve, is a replicated Pareto improvement, 1.28x throughput at 0.964x error on one scene and 1.42x at 0.68x on a second, with 47% less peak GPU memory and 29% less energy per pose. We evaluate on a calibrated, bit-exact regenerable noise ladder, on relabelled real-world dark exposures, and on a new dark-room video ladder recorded from a Boston Dynamics Spot robot.
Chinese Translation
在黑暗环境中执行任务的机器人需要依靠单个RGB摄像头进行定位，光线极低以至于每个像素的信号接近传感器自身的噪声，并且在功率受限的机载计算机上实时运行。每个约束条件都有成熟的处理流程，但这些约束的交集尚未得到解决。离线低光重建目前能够在-4 dB以下恢复结构，但速度远远太慢，无法实时运行，而机器人实际可以携带的实时单目系统（如DROID-SLAM、DPV-SLAM等）在信噪比（SNR）降低时会降级或失败。我们测量了它们的失败情况：在我们场景的九个最低黑暗级别中，DROID-SLAM在所有九个级别中返回的完整轨迹未包含任何关于相机运动的信息，VGGT-SLAM和CUT3R在八个级别中，pi^3在七个级别中，DPV-SLAM在四个级别中。我们提出了SYS，这是一种基于低光前馈点图前端的单目处理流程，具有显式匹配门，能够返回三条跟踪轨迹且没有无信息的轨迹，其跟踪时的误差低于任何方法（在无信息上限的24-47%对比最强基线的56-73%），并且覆盖范围最窄。在一段真实机器人视频中，其中86.5%的交付帧完全为黑色，我们的每种配置在亮起的开头后均停止，而DROID-SLAM和DPV-SLAM则为1178帧中的每一帧都发出一个位姿。我们的方法贡献是为Jetson AGX Orin提供的嵌入式执行路径：以384像素运行地图、关键帧和后端，同时以256像素进行跟踪，并对每帧位姿求解进行了两项修正，实现了复制的帕累托改进，在一个场景中以0.964倍的误差实现了1.28倍的吞吐量，在第二个场景中以0.68倍的误差实现了1.42倍的吞吐量，同时峰值GPU内存减少47%，每个位姿的能耗减少29%。我们在经过校准的、位精确的可再生噪声梯度、重新标记的真实世界黑暗曝光以及从波士顿动力Spot机器人录制的新黑暗房间视频梯度上进行了评估。
View on arXiv Download PDF AI Translation
cs.RO / 22 / 2609.21122

MetaPusher: Meta Learning and Planning for Nonprehensile Manipulation of Unseen Objects with Rapid Online Adaption

MetaPusher：用于未见物体非抓握操作的元学习与规划，具备快速在线适应能力
Lee, Donghyung, Golestaneh, Seyedali, Singh, Jaskrit, Zhong, Zhuoyun, Kapoutsis, Athanasios, Chamzas, Constantinos
Abstract
Manipulating previously unseen objects remains challenging, as their dynamics depend on latent physical properties, such as friction and mass distribution, that cannot be inferred from perception alone. Prior experience across objects can provide an initial estimate of unseen object dynamics, but this estimate remains uncertain and can degrade further during sim-to-real transfer. Adapting the dynamics through interaction can progressively refine the estimation, however, updating the model may invalidate the planned trajectory. Successful and efficient manipulation therefore requires both rapid dynamics adaptation and a planning strategy that can incorporate this evolution. In this work, we introduce MetaPusher, a meta-learning and adaptive planning framework for nonprehensile manipulation of unseen objects without prior object-specific interactions. A meta-learned dynamics model rapidly adapts from interactions during task execution, while an adaptive kinodynamic planner updates long-horizon plans by reusing and refining its existing search tree. This coupling enables manipulation and adaptation without a separate data collection phase. We evaluate MetaPusher on unseen objects in simulation and in sim-to-real scenarios, comparing against fine-tuning and active learning methods, MPPI-based control, and a reinforcement learning policy. It achieves lower prediction error and improves task success rate by up to 20%.
Chinese Translation
操控之前未见的物体仍然具有挑战性，因为它们的动态依赖于潜在的物理属性，如摩擦和质量分布，这些属性仅凭感知无法推断。跨物体的先前经验可以提供对未见物体动态的初步估计，但这种估计仍然不确定，并且在模拟到现实的转移过程中可能进一步恶化。通过交互适应动态可以逐步细化估计，然而，更新模型可能会使规划的轨迹失效。因此，成功且高效的操控需要快速的动态适应和能够纳入这种演变的规划策略。在本研究中，我们介绍了MetaPusher，一个用于未见物体非抓握操作的元学习和自适应规划框架，无需先前的物体特定交互。元学习的动态模型在任务执行过程中迅速适应交互，而自适应运动动力学规划器通过重用和细化现有的搜索树来更新长期规划。这种耦合使得在没有单独数据收集阶段的情况下实现操控和适应。我们在模拟和模拟到现实场景中对未见物体评估了MetaPusher，并与微调和主动学习方法、基于MPPI的控制以及强化学习策略进行了比较。结果显示，MetaPusher实现了更低的预测误差，并将任务成功率提高了多达20%。
View on arXiv Download PDF AI Translation
cs.RO / 23 / 2609.21130

SAGE: Safety-Aligned Gradient Enforcement for Human--Robot Collaboration

SAGE：面向人机协作的安全对齐梯度强化
Li, Yisen, Zhang, Hao, Geng, Ruize, Tseng, Yves, Zhao, Ding, Tseng, H. Eric
Abstract
Multi-party human-robot collaboration poses a dual challenge: robot decisions should remain interpretable and auditable, while executed actions must satisfy safety constraints during physical interaction. Combining explainable decision-tree policies with control-barrier-function (CBF) filtering provides a promising architecture but creates two learning mismatches in multi-agent reinforcement learning. Safety projection changes the action applied to the environment, while the coupled proposal graph can misalign independently optimized actor updates with a team-level update. We present safety-aligned gradient enforcement (SAGE) to address both mismatches. Its shield-annealed internalization layer (SAIL) uses a differentiable finite-penalty proposal map while retaining the exact CBF quadratic program for execution, preserving constraint-normal sensitivity to internalize repeatedly active safety constraints. Team-averaged Lyapunov policy optimization (TALO) constructs a team-aware update reference and applies a Lyapunov half-space correction to regulate independent actor updates. Physical experiments with two humanoid robots and a human partner demonstrate deployment feasibility. Across nine simulation scenarios, SAGE achieves a 71.0% success rate with 0.5 collision steps per thousand environment steps. Ablations show that direct CBF filtering reduces collision frequency by 98.5% but decreases success from 67.3% to 59.3%. SAIL reduces proposal violation by 48.8% and proposal-execution correction by 85.2%, while TALO reduces the update-consistency gap by 50.8%.
Chinese Translation
多方人机协作面临双重挑战：机器人决策应保持可解释性和可审计性，同时在物理交互过程中执行的动作必须满足安全约束。将可解释的决策树策略与控制障碍函数（Control Barrier Function, CBF）过滤相结合提供了一种有前景的架构，但在多智能体强化学习中产生了两个学习不匹配。安全投影改变了施加于环境的动作，而耦合的提议图可能导致独立优化的行动者更新与团队级更新不一致。我们提出了安全对齐梯度强化（Safety-Aligned Gradient Enforcement, SAGE）来解决这两个不匹配问题。其保护退火内部化层（Shield-Annealed Internalization Layer, SAIL）使用可微分的有限惩罚提议映射，同时保留精确的 CBF 二次规划以进行执行，从而保持约束-正常敏感性以反复内部化活跃的安全约束。团队平均李雅普诺夫策略优化（Team-Averaged Lyapunov Policy Optimization, TALO）构建了一个团队感知的更新参考，并应用李雅普诺夫半空间校正来调节独立的行动者更新。与两个人形机器人和一名人类合作伙伴的物理实验展示了部署的可行性。在九个仿真场景中，SAGE 实现了 71.0% 的成功率，每千个环境步骤发生 0.5 次碰撞。消融实验表明，直接 CBF 过滤将碰撞频率降低了 98.5%，但成功率从 67.3% 降低至 59.3%。SAIL 将提议违例减少了 48.8%，提议-执行校正减少了 85.2%，而 TALO 将更新一致性差距减少了 50.8%。
View on arXiv Download PDF AI Translation
cs.RO / 24 / 2609.21138

Diverse and Adaptable Arm Coordination for Octopus-Crawling via Diffusion-Based Uncertainty-Aware Optimization

基于扩散的不确定性感知优化的多样化和适应性章鱼爬行臂协调
Kim, Seung Hyun, Chang, Heng-Sheng, Kazemi, Kimia, Mehta, Prashant, Gazzola, Mattia
Abstract
Octopus crawling motivates soft robots that exploit redundancy, yet discovering and organizing diverse coordination modes for adaptation remains challenging. To address this, we introduce a Diffusion-based Uncertainty-aware Optimization (DUO) algorithm that learns demonstration-free crawling controllers for a simulated, muscle-actuated CyberOctopus. This work represents the first application of diffusion-based control to soft multi-arm robots in contact-rich simulations. By embedding a variety of locomotion behaviors within a shared control distribution, this approach enables the simulated octopus to navigate dynamic physical constraints, demonstrating that learned coordination diversity inherently facilitates robust adaptation. The main contributions include: (i) a symmetry-structured policy representation that folds radially equivalent controllers into a canonical directional sector, (ii) an online black-box optimization strategy, the DUO algorithm, that discovers and retains diverse coordination modes, and (iii) a control editing technique that adapts existing controllers to novel actuator constraints without retraining. These results show how learned coordination diversity makes motor abundance a practical resource for adaptation in soft multi-arm robots.
Chinese Translation
章鱼爬行激励了利用冗余的软体机器人，但发现和组织多样化的协调模式以实现适应性仍然具有挑战性。为了解决这个问题，我们提出了一种基于扩散的不确定性感知优化（Diffusion-based Uncertainty-aware Optimization, DUO）算法，该算法为模拟的肌肉驱动CyberOctopus学习无示范的爬行控制器。这项工作是扩散控制首次应用于接触丰富的软体多臂机器人模拟中。通过在共享控制分布中嵌入多种运动行为，该方法使得模拟的章鱼能够在动态物理约束下导航，展示了学习到的协调多样性本质上促进了稳健的适应性。主要贡献包括：（i）一种对称结构的策略表示，将径向等效的控制器折叠到一个典型的方向扇区中；（ii）一种在线黑箱优化策略，即DUO算法，能够发现并保留多样化的协调模式；（iii）一种控制编辑技术，能够在不重新训练的情况下将现有控制器适应于新的驱动约束。这些结果表明，学习到的协调多样性使得运动丰富性成为软体多臂机器人适应性的实用资源。
View on arXiv Download PDF AI Translation
cs.RO / 25 / 2609.21155

Same World, Different Knowledge: When Isolated Audits Misjudge World-Model Repairs

相同世界，不同知识：孤立审计如何误判世界模型修复
Min, Rui, Li, Xianyao, Xu, Fang, Lachab, Sofiane, Du, Jing
Abstract
A repair favored under an isolated input fault can be inferior when deployed modules share the faulty information. We introduce an information-interface audit for world models, distinguishing fidelity gaps, where exact inputs become estimates, from availability gaps, where inputs are missing. Fixed-weight interventions measure prediction error, input dependence, and paired closed-loop benefit, including dependencies introduced by reconstruction. In simulated quadrotor model predictive control, coupled, opposite-sign 10% mass/thrust calibration errors reduce a physics-anchored model's success from 69% to 8%; uncertainty training restores 65%. Wind reconstruction recovers control benefit but inherits calibration dependence. For a positive calibration offset, reconstruction-only corruption favors uncertainty-trained reconstruction, whereas shared corruption favors the baseline. Acceleration diagnostics reveal compensation between reconstruction bias and nominal-model error, also observed with a disturbance observer. Repair selection therefore depends on the information paths used in deployment.
Chinese Translation
在孤立输入故障下偏好的修复方案在模块共享故障信息时可能表现不佳。我们引入了一种针对世界模型的信息接口审计，区分了保真度差距（即精确输入变为估计值）和可用性差距（即输入缺失）。固定权重干预措施衡量预测误差、输入依赖性和配对闭环收益，包括重建引入的依赖性。在模拟四旋翼模型预测控制中，耦合的、相反符号的10%质量/推力标定误差将一个基于物理的模型的成功率从69%降低到8%；不确定性训练恢复了65%的成功率。风重建恢复了控制收益，但继承了标定依赖性。对于正标定偏移，仅重建的损坏更倾向于不确定性训练的重建，而共享损坏则更倾向于基线。加速诊断揭示了重建偏差与名义模型误差之间的补偿，这在干扰观察器中也有观察到。因此，修复选择依赖于部署中使用的信息路径。
View on arXiv Download PDF AI Translation
cs.RO / 26 / 2609.21167

MA-LIPP: Cooperative Multi-Agent Load-Aware Informative Path Planning for Heterogeneous Robot Teams

MA-LIPP：异构机器人团队的协作多智能体负载感知信息路径规划
Kim, Hojune, Shi, Guangyao, Sukhatme, Gaurav S.
Abstract
Field robotics missions often require physical samples to be returned to laboratories for analysis, making path planning inherently load-aware and order-dependent as accumulated samples increase payload and traversal energy costs. In single-robot Load-Aware Informative Path Planning (LIPP), this rigidly couples sensing with hauling: a solitary robot must transport every collected sample, forcing frequent depot returns that severely restrict its spatial coverage. Heterogeneous multi-robot teams can overcome this bottleneck by dividing labor---enabling high-precision samplers to collect while high-capacity carriers handle transport. However, this introduces a complex coordination challenge regarding when, where, what, and to whom handoffs should occur on top of the LIPP problem. To address this tightly coupled problem, we introduce Multi-Agent LIPP (MA-LIPP), which enables teams to cooperate through asynchronous "dead drops," allowing one robot to deposit samples for another to retrieve later without requiring synchronous rendezvous. We formulate MA-LIPP as an exact Mixed-Integer Quadratic Program (MIQP) alongside a scalable Pairwise Large-Neighborhood Search (LNS) heuristic for complex real-world applications. The heuristic matches exact optima in $95.5\%$ of certified cases and reduces weighted posterior variance by $16.1$--$19.8\%$ relative to a sequential baseline on larger instances of up to 12 robots, providing a robust framework for cooperative physical-sampling missions.
Chinese Translation
野外机器人任务通常需要将物理样本返回实验室进行分析，这使得路径规划本质上具有负载感知和顺序依赖性，因为累积的样本增加了有效载荷和行进能量成本。在单机器人负载感知信息路径规划（LIPP）中，这将感知与运输紧密结合：单个机器人必须运输每一个收集的样本，迫使其频繁返回仓库，从而严重限制了其空间覆盖。异构多机器人团队可以通过分工来克服这一瓶颈——高精度采样器可以进行采集，而高容量运输者则负责运输。然而，这引入了一个复杂的协调挑战，涉及何时、何地、什么以及将样本交接给谁，这些问题都在LIPP问题之上。为了解决这一紧密耦合的问题，我们提出了多智能体负载感知信息路径规划（MA-LIPP），该方法使团队能够通过异步“死信箱”进行合作，允许一台机器人存放样本，以便另一台机器人稍后取回，而无需同步会合。我们将MA-LIPP形式化为一个精确的混合整数二次规划（MIQP），并提出了一种可扩展的成对大邻域搜索（LNS）启发式算法，适用于复杂的现实世界应用。该启发式算法在95.5%的认证案例中匹配精确最优解，并在最多12台机器人的较大实例中，相对于顺序基线减少了16.1%至19.8%的加权后验方差，为协作物理采样任务提供了一个稳健的框架。
View on arXiv Download PDF AI Translation
cs.RO / 27 / 2609.21178

OpenRoIS: A Community-Driven Open-Source Middleware Implementing the Robotic Interaction Service (RoIS) Framework for Physical Robots and Virtual Agents

OpenRoIS：一个社区驱动的开源中间件，实现物理机器人和虚拟代理的机器人交互服务（RoIS）框架
Villalobos, Sebastian Carrera, Arellano, Christopher Nolan, Hitzmann, Arne, Brito, Edilson Morais, Utsumi, Akira, Horikawa, Yukiko, Miyashita, Takahiro, Hafi, Lotfi El
Abstract
Service applications for human-robot interaction are commonly written against the hardware-specific interfaces of one platform, so a change of hardware forces a rewrite of the application. The Robotic Interaction Service (RoIS) Framework 2.0, standardized by the Object Management Group (OMG), addresses this fragmentation by defining a platform-independent model in which Service Applications interact with Human-Robot Interaction (HRI) Engines through standardized interfaces and hardware-independent symbolic messages. A specification alone, however, does not provide the maintained implementation, Software Development Kits (SDKs), and adapters needed for practical adoption. This paper presents OpenRoIS, a community-driven open-source middleware providing a concrete implementation of the RoIS Framework 2.0. It takes the position that an openly developed, paradigm-neutral implementation is what carries the standard from specification to practice. OpenRoIS contributes a recursive engine architecture in which a single engine class realizes the main and sub HRI Engine roles, an internal five-method component contract distinct from the five external RoIS interfaces, a mapping of those interfaces onto JSON-RPC 2.0 over WebSocket, a single-source-of-truth type pipeline that generates three consistent language stacks, TypeScript and C# client SDKs that include web and Unity support, and a Python adapter SDK that includes ROS 2 support. Through the common RoIS interfaces, a Service Application can address physical robots and virtual agents over the internet. All source code, interface types, and documentation are released under the Apache-2.0 license and openly developed at https://openrois.org/.
Chinese Translation
人机交互的服务应用通常是针对单一平台的硬件特定接口编写的，因此硬件的更换会迫使应用程序重写。由对象管理组（OMG）标准化的机器人交互服务（RoIS）框架2.0通过定义一个平台无关的模型来解决这种碎片化问题，在该模型中，服务应用通过标准化接口和硬件无关的符号消息与人机交互（HRI）引擎进行交互。然而，仅有规范并不能提供实际应用所需的维护实现、软件开发工具包（SDK）和适配器。本文介绍了OpenRoIS，一个社区驱动的开源中间件，提供RoIS框架2.0的具体实现。我们认为，一个开放开发的、范式中立的实现是将标准从规范转化为实践的关键。OpenRoIS贡献了一个递归引擎架构，其中一个单一的引擎类实现了主要和次要的HRI引擎角色，一个内部五方法组件契约与五个外部RoIS接口不同，将这些接口映射到基于WebSocket的JSON-RPC 2.0，一个单一真实来源的类型管道生成三个一致的语言栈，包括支持Web和Unity的TypeScript和C#客户端SDK，以及包含ROS 2支持的Python适配器SDK。通过通用的RoIS接口，服务应用可以通过互联网与物理机器人和虚拟代理进行交互。所有源代码、接口类型和文档均在Apache-2.0许可证下发布，并在https://openrois.org/上公开开发。
View on arXiv Download PDF AI Translation
cs.RO / 28 / 2609.21185

When to Waddle: A Comparative Study of Bipedal Torso-Stabilization on Low-Friction Surfaces

何时摇摆：低摩擦表面上双足躯干稳定性的比较研究
Oke, Naomi, Gu, Ben, Ortiz, George, Ashlyn, Stacy, Pride, Cordelia, Bergbreiter, Sarah, Johnson, Aaron M.
Abstract
Low-friction surfaces challenge bipedal locomotion by limiting the contact forces available during stepping. Inspired by penguin waddling, we investigate how lateral torso motion and center of mass (COM) placement affect locomotion as surface friction changes. Using a five-actuator biped, we compare an upright-gait strategy with a penguin-inspired torso-over-stance-leg strategy across multiple COM placements in simulation and hardware. In the 3-D simulator MuJoCo, we sweep through sinusoidal leg and hip actuation parameters across four friction coefficients mu = 0.1, 0.3, 0.5, 0.7. In simulation, torso-over-stance-leg motion produces more successful controllers and higher forward speeds at low friction, with the highest speed occurring for the high-COM configuration. Hardware experiments show the same low-friction speed trend: at mu=0.12, torso-over-stance-leg motion increases forward speed and reduces cost of transport at both tested COM ratios, and the higher COM also improves both measures. The high-COM penguin configuration is the fastest and most energy efficient while maintaining low sideways foot motion. At mu=0.45, the COM trend reverses: the lower-COM configurations are faster and more energy efficient, while gait strategy has little effect on forward speed but still changes sideways foot motion. These results show that the effects of lateral torso motion and COM placement depend on the available friction, and that forward speed, energy use, and slip-related foot motion can be modulated with a penguin-inspired torso motion on hardware.
Chinese Translation
低摩擦表面对双足运动构成挑战，因为它限制了步态过程中可用的接触力。受到企鹅摇摆的启发，我们研究了侧向躯干运动和重心（COM）位置如何影响在不同表面摩擦下的运动。使用一个五个执行器的双足机器人，我们在模拟和硬件中比较了直立步态策略与企鹅启发的躯干-支撑腿策略在多个重心位置上的表现。在三维模拟器MuJoCo中，我们在四个摩擦系数μ = 0.1、0.3、0.5、0.7下，遍历了正弦腿部和髋部驱动参数。在模拟中，躯干-支撑腿运动在低摩擦下产生了更多成功的控制器和更高的前进速度，最高速度出现在高重心配置下。硬件实验显示了相同的低摩擦速度趋势：在μ=0.12时，躯干-支撑腿运动在两个测试的重心比下都提高了前进速度并降低了运输成本，而更高的重心也改善了这两个指标。高重心的企鹅配置是最快且最节能的，同时保持了较低的侧向脚部运动。在μ=0.45时，重心趋势发生反转：较低重心的配置更快且更节能，而步态策略对前进速度影响不大，但仍然改变了侧向脚部运动。这些结果表明，侧向躯干运动和重心位置的影响依赖于可用的摩擦力，并且可以通过在硬件上使用企鹅启发的躯干运动来调节前进速度、能量使用和与滑动相关的脚部运动。
View on arXiv Download PDF AI Translation
cs.RO / 29 / 2609.21186

Robust Structureless Monocular Visual Inertial Initialization Exploiting Line Features and Vanishing Points

利用线特征和消失点的鲁棒无结构单目视觉惯性初始化
Choi, Junwan, Jo, Woongrae, Seo, Dong-Uk, Jeon, Jinwoo, Myung, Hyun
Abstract
Accurate initialization is essential for reliable visual-inertial odometry (VIO), but it is often ill-conditioned under degenerate motions. Existing methods typically require restrictive excitation motions to ensure sufficient observability or rely on computationally expensive 3D structure reconstruction, limiting efficient and practical deployment. To address these limitations, we propose SLIM-init, a structureless monocular VIO initializer that directly exploits geometric constraints from tracked 2D line features without explicit 3D landmark reconstruction. Specifically, SLIM-init leverages line-derived vanishing points (VPs) as translation-invariant orientation cues to provide robust rotation-only constraints under degenerate scenarios such as low-parallax or translation-dominant motions. It further incorporates a line epipolar residual to constrain translation and a line-normal projection residual to improve the conditioning of linear alignment, enhancing the accuracy and robustness of initial state estimation. Extensive experiments on a public benchmark and challenging custom degenerate-motion sequences demonstrate improved accuracy and robustness over state-of-the-art initialization methods. The source code is available at: https://github.com/cjunwan/SLIM-init.
Chinese Translation
准确的初始化对于可靠的视觉惯性里程计（VIO）至关重要，但在退化运动下往往条件不良。现有方法通常需要限制性的激励运动以确保足够的可观测性，或者依赖于计算开销大的三维结构重建，这限制了高效和实际的部署。为了解决这些限制，我们提出了SLIM-init，一种无结构单目VIO初始化器，直接利用来自跟踪的二维线特征的几何约束，而无需显式的三维地标重建。具体而言，SLIM-init利用线导出的消失点（VPs）作为平移不变的方向线索，以在低视差或平移主导运动等退化场景下提供鲁棒的仅旋转约束。它进一步结合线的极线残差来约束平移，以及线法向投影残差来改善线性对齐的条件，从而提高初始状态估计的准确性和鲁棒性。在公共基准和具有挑战性的自定义退化运动序列上的大量实验表明，相较于最先进的初始化方法，准确性和鲁棒性得到了改善。源代码可在以下网址获取：https://github.com/cjunwan/SLIM-init。
View on arXiv Download PDF AI Translation
cs.RO / 30 / 2609.21211

Stochastic Neural Signed Swept Volume for Real-time Chance-Constrained Trajectory Optimization

用于实时机会约束轨迹优化的随机神经有符号扫掠体积
Chen, Qingyi, Zhang, Kevin, Chen, Lucas, Kingston, Zachary
Abstract
Collision-free motion planning requires reliable collision models from sensed environments and validation of states along a continuous trajectory. To make this tractable, most planners check for collision at discrete states along continuous trajectories against a single determinized model of the environment, introducing a trade-off between safety and computational efficiency. While continuous collision checking approaches that approximate the swept volume of the robot exist, they are computationally expensive or overly conservative. Data-driven approaches can learn the swept volume; however, these neural models are susceptible to approximation errors and are therefore often limited to serving as coarse filters for downstream collision checkers. In this work, we propose to learn a signed distance function of the swept volume as a probabilistic field, enabling quantification of epistemic uncertainty, incorporation of perception noise, and eventual integration into a chance-constrained trajectory optimization framework. We demonstrate our approach on challenging high-dimensional manipulation problems with significant sensor noise, both in simulation and on real hardware.
Chinese Translation
无碰撞运动规划需要来自感知环境的可靠碰撞模型以及沿连续轨迹的状态验证。为了使这一过程可行，大多数规划者在离散状态下检查与环境单一确定模型的碰撞，从而在安全性和计算效率之间引入权衡。虽然存在近似机器人扫掠体积的连续碰撞检查方法，但它们在计算上代价高昂或过于保守。基于数据的方法可以学习扫掠体积；然而，这些神经模型容易受到近似误差的影响，因此通常仅限于作为下游碰撞检查器的粗略过滤器。在本研究中，我们提出将扫掠体积的有符号距离函数作为概率场进行学习，从而能够量化认知不确定性，融入感知噪声，并最终整合到机会约束轨迹优化框架中。我们在具有显著传感器噪声的挑战性高维操作问题上展示了我们的方法，包括在仿真和真实硬件上的应用。
View on arXiv Download PDF AI Translation
cs.RO / 31 / 2609.21212

Visual Navigation Transformer with Pose Attention

具有姿态注意力的视觉导航变换器
Li, Beiming, Romero, Jaime, Diller, Jonathan, Kumar, Vijay, Ribeiro, Alejandro
Abstract
Learned navigation policies typically consume observations as a temporally ordered history, with positional encodings tying each observation to when it was seen, making it difficult to reuse experience from earlier traversals of an environment. Systems that do reuse such experience usually construct an explicit representation, such as a map or a topological graph, and plan on it. We propose VNT-PA (Visual Navigation Transformer with Pose Attention), a transformer planner whose context is a set of depth keyframes indexed by camera pose. With camera poses as positional encoding, attention depends on the pose differences between keyframes rather than on their temporal order. VNT-PA is trained to imitate a shortest-path planner operating on the ground-truth scene mesh, predicting actions by querying the spatial context with only its current pose and the goal position. On point-goal navigation in HM3D validation scenes, VNT-PA reaches 93.3% success and 90.4% success weighted by path length (SPL), outperforming baselines that encode the same context as a temporal sequence or treat pose as an input feature, in both navigation performance and training efficiency. Because the spatial context is a pose-indexed set, frames from different trajectories can be fused at test time. The planner also degrades more gracefully under localization noise than a conventional baseline which plans on explicit maps. These results show that pose-stamped experience can serve directly as the environment representation for a learned planner, and that making attention depend on pose differences, rather than on temporal order, speeds up training and improves long-horizon navigation.
Chinese Translation
学习的导航策略通常将观察视为时间顺序的历史记录，位置编码将每个观察与其被看到的时间相联系，这使得重用早期环境遍历的经验变得困难。那些确实重用此类经验的系统通常会构建一个显式表示，例如地图或拓扑图，并在其上进行规划。我们提出了VNT-PA（具有姿态注意力的视觉导航变换器），这是一种变换器规划器，其上下文是一组按相机姿态索引的深度关键帧。通过将相机姿态作为位置编码，注意力依赖于关键帧之间的姿态差异，而不是它们的时间顺序。VNT-PA被训练以模仿在真实场景网格上运行的最短路径规划器，通过仅查询当前姿态和目标位置的空间上下文来预测动作。在HM3D验证场景中的点目标导航中，VNT-PA达到了93.3%的成功率和90.4%的按路径长度加权成功率（SPL），在导航性能和训练效率上均优于将相同上下文编码为时间序列或将姿态视为输入特征的基线。由于空间上下文是一个按姿态索引的集合，不同轨迹的帧可以在测试时融合。该规划器在定位噪声下的降级也比在显式地图上进行规划的传统基线更为平滑。这些结果表明，带有姿态标记的经验可以直接作为学习规划器的环境表示，并且使注意力依赖于姿态差异而非时间顺序可以加速训练并改善长时间导航。
View on arXiv Download PDF AI Translation
cs.RO / 32 / 2609.21216

Fewer Steps, Better Actions: Rethinking Flow-Matching Inference for VLA Policies

更少的步骤，更好的动作：重新思考流匹配推理在VLA策略中的应用
Tang, Zhipeng, Chen, Xinda, Rao, Weining, Li, Xiao, Tan, Wenting, Wang, Yuning, Shi, Xiao, Zhao, Xiaofang
Abstract
Vision-language-action (VLA) policies based on flow matching generate action chunks through repeated evaluations of an action expert. Increasing the number of integration steps raises inference cost, but does not necessarily improve closed-loop success. We propose Coda, which reallocates part of this integration budget to a single learned endpoint correction. A frozen policy first completes a few-step noise-to-action trajectory; a lightweight Transformer then predicts a demonstration-supervised residual using the candidate action, source noise, and shared observation-prefix cache. Only the corrector is trained. On 50 RoboTwin Easy tasks, five-step Coda improves success from 71.64% to 74.68% over the matched five-step baseline, while reducing forward latency by 30.2% relative to the default ten-step policy. A two-step configuration achieves 71.88% success with a 2.12$\times$ speedup. An independent 13-task control shows a 5.69-percentage-point gain at nearly equal latency, supporting correction as an effective alternative to additional integration. The same design also improves frozen official SmolVLA, raising two-step success from 60.8% to 69.4%. These results show that endpoint correction improves the quality-latency trade-off of frozen flow-matching policies.
Chinese Translation
基于流匹配的视觉-语言-动作（VLA）策略通过对动作专家的反复评估生成动作块。增加集成步骤的数量会提高推理成本，但不一定能改善闭环成功率。我们提出了Coda，它将部分集成预算重新分配给单一的学习端点修正。一个冻结的策略首先完成几步的噪声到动作的轨迹；然后，一个轻量级的Transformer使用候选动作、源噪声和共享观察前缀缓存预测演示监督的残差。只有修正器是经过训练的。在50个RoboTwin Easy任务中，五步Coda将成功率从71.64%提高到74.68%，相较于匹配的五步基线，同时将相较于默认的十步策略减少了30.2%的前向延迟。两步配置在2.12倍加速下实现了71.88%的成功率。一个独立的13任务控制显示在几乎相等的延迟下获得了5.69个百分点的提升，支持修正作为额外集成的有效替代。相同的设计也改善了冻结的官方SmolVLA，将两步成功率从60.8%提高到69.4%。这些结果表明，端点修正改善了冻结流匹配策略的质量-延迟权衡。
View on arXiv Download PDF AI Translation
cs.RO / 33 / 2609.21220

Safe Real-Time Policy Steering via Noise-Space Trajectory Optimization for One-Step Generative Policies

通过噪声空间轨迹优化实现安全的实时策略引导用于一步生成策略
Chen, Qingyi, Ruan, Joseph, Kingston, Zachary
Abstract
Generative robot policies can represent diverse, multimodal behaviors, but adapting pretrained policies to deployment-time constraints such as collision avoidance and orientation maintenance remains challenging. Existing inference-time steering methods typically apply gradient guidance through iterative diffusion or flow processes, which can be computationally expensive for real-time control. We propose INSPO, which formulates inference-time steering of one-step generative policies as trajectory optimization in the policy's input noise space. By optimizing the input noise while evaluating constraints on the induced state trajectory, INSPO searches the policy-induced behavior space without directly modifying generated actions. The optimization includes a regularization term that encourages solutions to remain consistent with the policy's input distribution and is solved online using population-based particle optimization. We evaluate INSPO on state- and image-based task-specific policies and generalist vision-language-action policies across Push-T, Can pick-and-place, and LIBERO-Spatial. INSPO improves task success and constraint satisfaction over best-of-N sampling and action projection, while comparing favorably with gradient-guided generation at lower runtime.
Chinese Translation
生成机器人策略能够表示多样化的多模态行为，但将预训练的策略适应于部署时的约束条件，如避免碰撞和保持方向，仍然具有挑战性。现有的推理时引导方法通常通过迭代扩散或流动过程应用梯度引导，这对于实时控制来说计算成本较高。我们提出了INSPO，它将一步生成策略的推理时引导形式化为策略输入噪声空间中的轨迹优化。通过优化输入噪声并评估对诱导状态轨迹的约束，INSPO在不直接修改生成动作的情况下搜索策略诱导的行为空间。该优化包括一个正则化项，鼓励解决方案与策略的输入分布保持一致，并使用基于种群的粒子优化在线求解。我们在基于状态和图像的任务特定策略以及通用的视觉-语言-动作策略上评估INSPO，涵盖Push-T、Can pick-and-place和LIBERO-Spatial。INSPO在任务成功率和约束满足方面优于最佳N采样和动作投影，同时在较低运行时间下与梯度引导生成相比表现良好。
View on arXiv Download PDF AI Translation
cs.RO / 34 / 2609.21223

SafeStage: Evaluating Safety Before, During, and After Vision-Language-Conditioned Robot Manipulation

安全阶段：评估视觉-语言条件下机器人操作的安全性前、中、后
Luo, Jinzhu, Zhang, Qi, Wang, Wei, Jiang, Wei
Abstract
Vision-language-conditioned robot policies integrate perception, language understanding, and control for general-purpose manipulation. However, existing evaluations often focus on task success, isolated physical constraints, semantic refusal, or realized physical damage, providing limited insight into where safety fails during closed-loop manipulation. We introduce SafeStage, a lifecycle-structured benchmark for evaluating manipulation safety before, during, and after task execution. SafeStage contains 97 purpose-built risk scenarios organized into three stages. Initial-State Hazards captures safety-relevant relations that must be resolved before manipulating the target. Execution-Time Safety evaluates unsafe contacts, trajectories, region entries, and object interactions during execution. Final-State Hazards capture unstable or otherwise unsafe conditions remaining after nominal task completion. The benchmark evaluates realized interactions using event-based and state-based checks and reports native task success independently from stage-specific safety outcomes. We evaluate representative direct-action Vision-Language-Action (VLA) policies and policies with world-model-based policies under a common closed-loop protocol. Our results demonstrate that nominal task completion frequently coexists with safety violations and that different policies exhibit distinct failure profiles across the three stages. By separating task success from safety and localizing when violations occur, SafeStage provides a unified diagnostic testbed for evaluating and improving vision-language-conditioned robot manipulation policies.
Chinese Translation
视觉-语言条件下的机器人策略整合了感知、语言理解和控制，用于通用操作。然而，现有的评估往往集中于任务成功、孤立的物理约束、语义拒绝或实际物理损害，提供了有限的洞察力来了解在闭环操作过程中安全性失效的地方。我们引入了SafeStage，这是一个生命周期结构的基准，用于评估任务执行前、中、后操作的安全性。SafeStage包含97个专门构建的风险场景，分为三个阶段。初始状态危害（Initial-State Hazards）捕捉在操作目标之前必须解决的与安全相关的关系。执行时安全（Execution-Time Safety）评估在执行过程中不安全的接触、轨迹、区域进入和物体交互。最终状态危害（Final-State Hazards）捕捉在正常任务完成后仍然存在的不稳定或其他不安全条件。该基准通过基于事件和基于状态的检查评估实际交互，并独立于阶段特定的安全结果报告原生任务成功。我们在一个共同的闭环协议下评估了代表性的直接行动视觉-语言-行动（Vision-Language-Action, VLA）策略和基于世界模型的策略。我们的结果表明，正常任务完成与安全违规常常共存，并且不同策略在三个阶段中表现出不同的失败特征。通过将任务成功与安全分离并定位违规发生的时机，SafeStage提供了一个统一的诊断测试平台，用于评估和改进视觉-语言条件下的机器人操作策略。
View on arXiv Download PDF AI Translation
cs.RO / 35 / 2609.21226

AirSplan: Risk-Aware Motion Planning for Quadrotors in Cluttered 3D Gaussian Splats

AirSplan：在复杂三维高斯点云中进行风险感知的四旋翼运动规划
Isaacson, Seth, Hong, William, Skinner, Katherine A., Vasudevan, Ram
Abstract
Quadrotors are increasingly deployed in applications such as agriculture, infrastructure inspection, and maintenance. In each of these applications, the robot must navigate complex scene geometry while remaining strictly collision-free. Unlike in ground domains, even minor collisions for aerial vehicles can result in the loss of the robot. This safety requirement induces a pair of technical challenges. First, the environment must be represented with sufficient fidelity to encode complex structure, even when no ground-truth obstacle data is available. Second, a motion planner must leverage this representation to determine a collision-free path to the goal. This paper proposes a system that addresses these complementary challenges. The proposed method, AirSplan, adopts a normalized variant of 3D Gaussian Splatting that encodes high-fidelity scene geometry. It then applies a novel reachability-based motion planner that leverages the differential flatness of quadrotors to compute continuous-time collision constraints that tightly overapproximate the robot's occupancy. Experiments demonstrate that AirSplan successfully finds a path in 81.2% of challenging test cases, a significant improvement over the nearest baseline method's 51.2%.
Chinese Translation
四旋翼无人机在农业、基础设施检查和维护等应用中的部署越来越广泛。在这些应用中，机器人必须在复杂的场景几何中导航，同时保持严格的无碰撞状态。与地面领域不同，空中飞行器即使是轻微的碰撞也可能导致机器人失去控制。这一安全要求引发了一对技术挑战。首先，环境必须以足够的精度进行表示，以编码复杂的结构，即使在没有真实障碍数据的情况下也是如此。其次，运动规划器必须利用这种表示来确定到达目标的无碰撞路径。本文提出了一个解决这些互补挑战的系统。所提出的方法AirSplan采用了一种归一化的三维高斯点云变体，以编码高保真场景几何。然后，它应用了一种新颖的基于可达性的运动规划器，利用四旋翼的微分平坦性来计算严格过近似机器人的占用的连续时间碰撞约束。实验表明，AirSplan在81.2%的具有挑战性的测试案例中成功找到路径，显著优于最近基线方法的51.2%。
View on arXiv Download PDF AI Translation
cs.RO / 36 / 2609.21228

FOCAL-VLA: Subtask-Guided Geometry Distillation and Implicit World Modeling for Vision-Language-Action Models

FOCAL-VLA：基于子任务引导的几何蒸馏与隐式世界建模的视觉-语言-动作模型
Gao, Zhiyuan, Wen, Di, Zhan, Yanxiang, Khoshnazar, Mohammad, Schäfer, Jeroen, Peng, Kunyu, Beetz, Michael
Abstract
Vision-language-action (VLA) models built on pretrained vision-language models have demonstrated strong performance across diverse robotic manipulation tasks. However, VLA models that directly map current 2D observations to actions often lack sufficient spatial and temporal understanding, limiting their performance in precise and long-horizon manipulation. Recent methods enhance VLA models through geometric supervision and future-state prediction across the entire scene. However, these methods can suffer from redundant scene information, distracting the model from learning the geometry and dynamics relevant to the current interaction. To address this issue, we propose FOCAL-VLA, a framework that combines subtask-guided geometry distillation with implicit world modeling to learn representations of current spatial structure and future interaction dynamics. To focus geometric learning on the current subtask, we transfer geometric knowledge from VGGT to the VLA model by aligning geometry latents with features from subtask-relevant image regions. To capture the future 3D evolution of the current interaction, we incorporate implicit world modeling using Track4World features from current and future demonstration frames. The two complementary representations jointly guide action generation without running VGGT or Track4World at inference time. Experiments show that FOCAL-VLA outperforms baselines on both simulation benchmarks and real-world manipulation tasks. Project website: https://zhiyuan-gao.github.io/FOCAL-VLA/.
Chinese Translation
基于预训练视觉-语言模型构建的视觉-语言-动作（VLA）模型在多样化的机器人操作任务中表现出色。然而，直接将当前二维观察映射到动作的VLA模型往往缺乏足够的空间和时间理解，限制了其在精确和长时间操作中的表现。最近的方法通过对整个场景进行几何监督和未来状态预测来增强VLA模型。然而，这些方法可能会受到冗余场景信息的影响，分散模型对当前交互相关的几何和动态学习的注意力。为了解决这一问题，我们提出了FOCAL-VLA，一个结合子任务引导的几何蒸馏与隐式世界建模的框架，以学习当前空间结构和未来交互动态的表征。为了将几何学习集中在当前子任务上，我们通过将几何潜变量与子任务相关图像区域的特征对齐，将几何知识从VGGT转移到VLA模型。为了捕捉当前交互的未来三维演变，我们使用来自当前和未来演示帧的Track4World特征来融入隐式世界建模。这两种互补的表征共同指导动作生成，而无需在推理时运行VGGT或Track4World。实验表明，FOCAL-VLA在模拟基准和真实世界操作任务中均优于基线模型。项目网站：https://zhiyuan-gao.github.io/FOCAL-VLA/
View on arXiv Download PDF AI Translation
cs.RO / 37 / 2609.21229

KnowDemo: Knowledge-Guided Robot Demonstration Generation from Human Videos

KnowDemo：基于知识引导的人类视频生成机器人演示
Gao, Zhiyuan, Zhan, Yanxiang, Khoshnazar, Mohammad, Schäfer, Jeroen, Beetz, Michael
Abstract
Learning robot manipulation policies typically requires substantial demonstration data, which are costly to collect on real robots. Recent methods generate robot demonstrations from human videos by adapting recovered motion and validating the resulting trajectories in simulation. However, methods centered on motion-reference adaptation can limit behavioral diversity by retaining the demonstrated contact strategies and subtask orders, while insufficient understanding of task requirements and scene relations can reduce demonstration generation efficiency by generating invalid candidates. To address these limitations, we propose KnowDemo, a framework that uses structured manipulation knowledge from human videos to generate diverse robot demonstrations for a target workspace. To distinguish task requirements from demonstration-specific choices, we develop a knowledge extraction and reasoning module based on a vision-language model (VLM) that associates object and action descriptions with inferred task conditions, demonstration references, and permissible execution variations. To translate this knowledge into executable demonstrations, we resolve the descriptions against target-scene entities and geometry to guide candidate generation and screening before motion planning and simulation. The resulting demonstrations exhibit multimodal behavior through alternative contact strategies and valid subtask orders, with structured execution labels. Experiments demonstrate additional verified execution modes beyond a reference-only configuration and improved candidate planning success through task-guided grasp sampling. To validate the generated data for policy learning, we fine-tune the pretrained $\pi_{0.5}$ model on simulation data, achieving sim-to-real transfer across three tasks. Project page: https://zhiyuan-gao.github.io/knowdemo/
Chinese Translation
学习机器人操作策略通常需要大量的演示数据，而在真实机器人上收集这些数据的成本很高。最近的方法通过适应恢复的运动并在仿真中验证生成的轨迹，从人类视频中生成机器人演示。然而，集中于运动参考适应的方法可能会通过保留演示的接触策略和子任务顺序来限制行为的多样性，而对任务需求和场景关系理解不足则可能通过生成无效候选来降低演示生成的效率。为了解决这些限制，我们提出了KnowDemo，一个利用人类视频中的结构化操作知识为目标工作空间生成多样化机器人演示的框架。为了区分任务需求和特定于演示的选择，我们开发了一个基于视觉-语言模型（VLM）的知识提取和推理模块，该模块将物体和动作描述与推断的任务条件、演示参考和允许的执行变体关联起来。为了将这些知识转化为可执行的演示，我们将描述与目标场景实体和几何形状进行匹配，以指导候选生成和筛选，然后再进行运动规划和仿真。生成的演示通过替代接触策略和有效的子任务顺序展现出多模态行为，并带有结构化的执行标签。实验表明，除了仅参考配置外，还验证了额外的执行模式，并通过任务引导的抓取采样提高了候选规划的成功率。为了验证生成的数据用于策略学习，我们在仿真数据上微调了预训练的$ ext{π}_{0.5}$模型，实现了三个任务的仿真到现实转移。项目页面：https://zhiyuan-gao.github.io/knowdemo/
View on arXiv Download PDF AI Translation
cs.RO / 38 / 2609.21246

VLA-Scope: Shift-Aware Failure Prediction for Vision-Language-Action Models

VLA-Scope：面向视觉-语言-行动模型的偏移感知故障预测
Zhu, Kaiwen, Liu, Dongfang, Liu, Liangkai
Abstract
Vision-language-action (VLA) models map visual observations and natural-language instructions to robotic actions, but distribution shifts can compromise their reliability. Because these models may still succeed under out-of-distribution (OOD) conditions, detecting OOD inputs alone is insufficient to predict execution failure. In this paper, we introduce VLA-Scope, a two-stage framework that combines input-shift characterization with execution history to predict failure during OOD rollouts. The first stage uses pooled image and language representations to detect OOD inputs and classify their shift categories. For inputs flagged as OOD, the second stage combines the predicted category, action-prefix features, and execution progress features. A logistic regression model shared across shift categories updates failure risk as execution proceeds. We evaluate the framework with OpenVLA on ten LIBERO-Spatial tasks using leave-one-group-out cross-validation. OOD detection achieves a ROC-AUC of 0.9454, and shift classification achieves 91% accuracy. Evaluated independently of the OOD gate on all 1,400 OOD rollouts, the failure predictor achieves a ROC-AUC of 0.8497 after 60 executed actions, compared with 0.7906 without execution progress features. It also achieves a higher ROC-AUC than the evaluated ActProbe and SAFE-MLP baselines. These results suggest that combining action features with temporally aggregated execution step representations improves failure prediction under input shifts.
Chinese Translation
视觉-语言-行动（VLA）模型将视觉观察和自然语言指令映射到机器人行动，但分布偏移可能会影响其可靠性。由于这些模型在分布外（OOD）条件下仍可能成功，仅检测OOD输入不足以预测执行失败。本文介绍了VLA-Scope，一个将输入偏移特征与执行历史相结合的两阶段框架，以预测OOD回滚过程中的失败。第一阶段使用汇总的图像和语言表示来检测OOD输入并分类其偏移类别。对于被标记为OOD的输入，第二阶段结合预测的类别、行动前缀特征和执行进度特征。一个在各偏移类别间共享的逻辑回归模型在执行过程中更新失败风险。我们在十个LIBERO-Spatial任务上使用OpenVLA评估该框架，采用留一组交叉验证。OOD检测的ROC-AUC达到0.9454，偏移分类的准确率达到91%。在所有1,400个OOD回滚中，独立于OOD门评估的故障预测器在执行60个动作后实现了0.8497的ROC-AUC，而没有执行进度特征时为0.7906。它还实现了比评估的ActProbe和SAFE-MLP基线更高的ROC-AUC。这些结果表明，将行动特征与时间聚合的执行步骤表示相结合可以改善输入偏移下的故障预测。
View on arXiv Download PDF AI Translation
cs.RO / 39 / 2609.21275

LOInK: Learned Optimal Inverse Kinematics via Structured Neural Surrogate Models

LOInK：通过结构化神经代理模型学习的最优逆运动学
Somerfield, Michael, Abood, Damian, Wang, Ruigang, Manchester, Ian R.
Abstract
We introduce Learned Optimal Inverse Kinematics (LOInK), a method to generate approximately optimal solutions to an inverse kinematics problem. When trained on data consisting of sampled configurations and associated task variables and a given cost function, LOInK learns a bi-Lipschitz invertible mapping from configuration space to a decoupled task/latent space, and moreover, the latent space is structured so as to place cost-minimizing solutions at the origin. This enables efficient sampling of cost-minimizing solutions via a network-inversion algorithm based on operator splitting. We demonstrate the proposed approach on three problems: an illustrative three degree-of-freedom manipulator problem; a quadrupedal climbing robot for which LOInK can generate near-optimal solutions on average 31 times faster and up to 100 times faster than a constrained optimization approach; and a simulated soft actuator as a purely data-driven example, in which LOInK can explicitly generate high-quality solutions, unlike existing generative approaches that require diverse sampling and evaluation of candidate solutions.
Chinese Translation
我们提出了学习的最优逆运动学（LOInK），这是一种生成逆运动学问题近似最优解的方法。当在包含采样配置及其相关任务变量和给定成本函数的数据上进行训练时，LOInK 学习了一个双利普希茨可逆映射，将配置空间映射到解耦的任务/潜在空间。此外，潜在空间的结构设计使得成本最小化的解位于原点。这使得通过基于算子分裂的网络反演算法高效采样成本最小化解成为可能。我们在三个问题上演示了该方法：一个示例性的三自由度操纵器问题；一个四足爬行机器人，LOInK 平均可以生成近似最优解，速度比约束优化方法快 31 倍，最快可达 100 倍；以及一个模拟的软驱动器作为纯数据驱动的示例，其中 LOInK 能够显式生成高质量解，而现有的生成方法则需要多样化的采样和候选解的评估。
View on arXiv Download PDF AI Translation
cs.RO / 40 / 2609.21307

Stability-aware Residual Reinforcement Learning Framework for Robotic Manipulator Disturbance Compensation

稳定性意识的残差强化学习框架用于机器人操纵器扰动补偿
Kim, Jihong, Kwon, Joonhyuk, Kim, Hwa Soo, Seo, TaeWon, Seo, Hyung-Tae
Abstract
Although conventional controllers and disturbance observers (DOBs) are the standard for precision tracking in manipulators, they suffer from parameter uncertainty, nonlinear friction, and compound disturbances. This study proposes a residual reinforcement learning DOB framework that pairs an analytical observer with an RL policy. The deterministic baseline operates within a reliable region, whereas the RL policy explicitly targets the residuals that the model cannot capture. To make this compensation disturbance-aware, an estimator network aligns the observation history with a privileged disturbance context, organizing the latent space by disturbance regime and enabling rapid adaptation across disturbance transitions. To guarantee stability, we derived and enforced a state-dependent action bound on the RL policy from an input-to-state stability (ISS) analysis such that the closed loop provably confines the tracking error to a certified envelope for arbitrary policy outputs. Experiments on a 6-DOF manipulator demonstrated consistent improvements in disturbance estimation and tracking, including a 27.8% tracking-error reduction on real hardware under zero-shot sim-to-real transfer and a 38.0% reduction under a base-vibration disturbance that was not observed during training.
Chinese Translation
尽管传统控制器和扰动观测器（DOBs）是操纵器精确跟踪的标准，但它们受到参数不确定性、非线性摩擦和复合扰动的影响。本研究提出了一种残差强化学习DOB框架，该框架将分析观测器与强化学习（RL）策略相结合。确定性基线在可靠区域内运行，而RL策略则明确针对模型无法捕捉的残差。为了使这种补偿具有扰动意识，一个估计网络将观测历史与特权扰动上下文对齐，通过扰动模式组织潜在空间，从而在扰动过渡中实现快速适应。为了保证稳定性，我们从输入到状态稳定性（ISS）分析中推导并强制实施了对RL策略的状态依赖动作边界，以确保闭环系统能够证明性地将跟踪误差限制在一个经过认证的范围内，适用于任意策略输出。在一个6自由度操纵器上的实验表明，扰动估计和跟踪的一致性改善，包括在零-shot模拟到真实转移下，实际硬件的跟踪误差减少了27.8%，以及在训练期间未观察到的基础振动扰动下减少了38.0%。
View on arXiv Download PDF AI Translation
cs.RO / 41 / 2609.21316

NaViRrator: Robot Navigation from Human-Readable Maps through a Learned Visual Route

NaViRrator：通过学习的视觉路线实现机器人从人类可读地图的导航
Lee, Ayun, Kim, Jiseon, Kim, Giseop
Abstract
Human-readable maps provide an intuitive interface for specifying robot destinations, but connecting their schematic geometry to egocentric observations remains challenging. We present NaViRRator, a framework that translates user-specified start and goal locations on such maps into navigation instructions for a pretrained vision-and-language navigation (VLN) policy. Its core method, RouteScribe, separates route inference from verbalization by first generating an explicit route scaffold in map-image coordinates, which a pretrained vision-language model (VLM) converts into a navigation instruction. We construct the scaffold with start--goal line conditional flow matching (SGL-CFM), which deforms a straight start--goal waypoint sequence into a map-conditioned route. During execution, the VLN policy receives only the instruction and egocentric observations, while the map and scaffold remain upstream, allowing executor replacement without retraining the map-to-language modules. Real-world experiments show higher success rates and success weighted by path length (SPL) than direct map-to-instruction generation, A*-based scaffolding, and Gaussian-source conditional flow matching. Qualitative results further show clearer salient turns and better preservation of the intended maneuver sequence, supporting route-grounded language as a modular interface between human-readable maps and pretrained navigation policies.
Chinese Translation
人类可读地图为指定机器人目的地提供了直观的接口，但将其示意几何与自我中心观察连接起来仍然具有挑战性。我们提出了NaViRRator，一个将用户指定的起始和目标位置在此类地图上转化为预训练视觉与语言导航（VLN）策略的导航指令的框架。其核心方法RouteScribe通过首先在地图图像坐标中生成一个明确的路线支架，将路线推断与语言表达分离，随后由预训练的视觉语言模型（VLM）将其转换为导航指令。我们通过起始-目标线条件流匹配（SGL-CFM）构建支架，该方法将一条直线的起始-目标航点序列变形为一个地图条件的路线。在执行过程中，VLN策略仅接收指令和自我中心观察，而地图和支架则保持在上游，从而允许执行者替换而无需重新训练地图到语言的模块。实际实验显示，与直接地图到指令生成、基于A*的支架和高斯源条件流匹配相比，成功率和按路径长度加权的成功率（SPL）更高。定性结果进一步显示出更清晰的显著转弯和更好地保留预期的操作序列，支持基于路线的语言作为人类可读地图与预训练导航策略之间的模块化接口。
View on arXiv Download PDF AI Translation
cs.RO / 42 / 2609.21319

LEMCA: LLM-Guided Synthesis of Efficient Mode-Switching Control Architectures

LEMCA：基于大语言模型的高效模式切换控制架构合成
Krishna, Arjun, Pacelli, Vincent, Jayaraman, Dinesh
Abstract
Physical control tasks in the natural world, such as driving or object manipulation, frequently exhibit dramatic variations in sensory and compute complexity over time. Correspondingly, a natural resource-efficient choice for robot control is to dynamically switch between control modes with varying resource allocations. However, such "mode-switching controllers" (MSCs) have historically required laborious, expert-driven design and synthesis for each new task. Driven by these design difficulties, modern robotic control architectures often fall back to a wasteful "monolithic" one-size-fits-all structure, where resource allocation is permanently anchored to the hardest, most resource-intensive task phases. To facilitate the design of performant yet efficient MSCs, we propose LLM-Guided synthesis of Efficient Mode-Switching Control Architectures (LEMCA). LEMCA represents MSC designs as interpretable programs to be iteratively refined in an evolutionary loop. To evaluate design fitness, we propose MSC-compatible extensions of automated controller synthesis approaches, such as reinforcement learning in simulation. LEMCA then leverages the semantic priors, reasoning, and coding capabilities of Large Language Models (LLMs) to iteratively edit controller modes, their corresponding sensory-compute resource allocations, and mode transitions. Our experiments across diverse control benchmarks show that LEMCA consistently discovers strategies that surpass the Pareto frontier of monolithic designs by reclaiming wasted resources during "easy" task phases. LEMCA thus presents an automated, low-effort path to synthesize resource-efficient MSC designs.
Chinese Translation
自然界中的物理控制任务，如驾驶或物体操作，通常在感知和计算复杂性上随时间表现出剧烈变化。因此，对于机器人控制而言，一种自然的资源高效选择是动态切换不同资源分配的控制模式。然而，这种“模式切换控制器”（MSC）在历史上往往需要繁琐的专家驱动设计和合成，针对每个新任务。由于这些设计困难，现代机器人控制架构往往回归到浪费资源的“单一”一体化结构，其中资源分配永久固定于最困难、最消耗资源的任务阶段。为了促进高效且性能优越的MSC设计，我们提出了基于大语言模型的高效模式切换控制架构合成（LEMCA）。LEMCA将MSC设计表示为可解释的程序，以便在演化循环中进行迭代优化。为了评估设计的适应性，我们提出了与MSC兼容的自动控制器合成方法的扩展，如在仿真中的强化学习。然后，LEMCA利用大语言模型（LLMs）的语义先验、推理和编码能力，迭代编辑控制器模式、相应的感知-计算资源分配以及模式转换。我们在多种控制基准上的实验表明，LEMCA始终发现超越单一设计的帕累托前沿的策略，通过回收“简单”任务阶段中的浪费资源。因此，LEMCA提供了一条自动化、低努力的路径，以合成资源高效的MSC设计。
View on arXiv Download PDF AI Translation
cs.RO / 43 / 2609.21330

The EventCV Library for Event-Based Robotic Vision

用于事件驱动机器人视觉的 EventCV 库
Hines, Adam D., Milford, Michael, Fischer, Tobias
Abstract
Event cameras detect per-pixel brightness changes asynchronously on microsecond timescales, with high dynamic range and low power draw. These are desirable properties for robots that move fast or work in difficult lighting conditions. However, integrating an event camera into a real-world robotic pipeline still requires substantial effort: plug-and-play drivers do not exist, event streams are recorded in a variety of incompatible file formats, and most projects rely on custom research-grade code. Here, we present EventCV, an open-source and extensible Rust library with OpenCV-style Python bindings that lowers the entry barrier to working with event cameras. EventCV provides a wide range of features: denoising filters and geometric transforms, augmentations, corner detection and unsupervised feature learning, contrast-maximization motion estimation, a video-to-events simulator, and Open Neural Network Exchange (ONNX) inference for deployment in robotic stacks. EventCV integrates the Neuromorphic Drivers package, allowing an event camera stream to be processed directly in real time. No existing toolkit covers this range of operations in one package, and EventCV builds representations and decodes files 1.1x to 3.7x faster than the currently available libraries. We deploy EventCV on a Jetson Orin AGX and present three robotics case studies spanning object detection, on-device model inference, and localization. Project webpage: https://eventcv.net.
Chinese Translation
事件相机以微秒级的时间尺度异步检测每个像素的亮度变化，具有高动态范围和低功耗的优点。这些特性对于快速移动或在复杂光照条件下工作的机器人来说非常重要。然而，将事件相机集成到现实世界的机器人管道中仍然需要大量的努力：现成的驱动程序并不存在，事件流以多种不兼容的文件格式记录，大多数项目依赖于定制的研究级代码。在此，我们介绍了 EventCV，这是一个开源且可扩展的 Rust 库，具有 OpenCV 风格的 Python 绑定，降低了使用事件相机的入门门槛。EventCV 提供了广泛的功能：去噪滤波器和几何变换、数据增强、角点检测和无监督特征学习、对比度最大化的运动估计、视频到事件的模拟器，以及用于在机器人堆栈中部署的开放神经网络交换（Open Neural Network Exchange, ONNX）推理。EventCV 集成了神经形态驱动程序包，允许事件相机流在实时中直接处理。没有现有工具包能够在一个包中覆盖这一系列操作，而 EventCV 的表示构建和文件解码速度比当前可用库快 1.1 到 3.7 倍。我们在 Jetson Orin AGX 上部署了 EventCV，并展示了三个机器人案例研究，涵盖了物体检测、设备内模型推理和定位。项目网页：https://eventcv.net。
View on arXiv Download PDF AI Translation
cs.RO / 44 / 2609.21358

FAN: Foresight Action Normalization for Continual Adaptation of Vision-Language-Action Models

FAN：用于视觉-语言-动作模型持续适应的前瞻性动作归一化
Hong, Yijun, Zhu, Jiarun, Sun, Xiaoquan, Xu, Le, He, Qijun, Jin, Xin, Yuan, Mingqi, Zeng, Wenjun, Chen, Jiayu
Abstract
Vision-Language-Action (VLA) models pre-trained on large-scale, closed datasets have demonstrated remarkable success across diverse robotic manipulation tasks. However, their long-term real-world deployment necessitates continuously acquiring new skills while retaining previously learned capabilities. While pioneering works have explored continual VLA adaptation using techniques such as experience replay and reinforcement fine-tuning, they overlook a foundational mechanism: action normalization, which determines the underlying coordinate system in which policies perceive and execute physical actions. To bridge this gap, we systematically evaluate five normalization strategies across four real-world task streams covering single-arm and bimanual manipulation. Our analysis reveals that existing protocols induce severe failure modes due to inter-task coordinate drift, limited motion coverage, or train-test coordinate mismatches. Motivated by these insights, we formulate three core design principles: consistency, coverage, and causality (3C), and introduce foresight action normalization (FAN). FAN estimates normalization statistics once from a small, task-independent calibration set prior to continual learning and freezes them throughout adaptation. Across all evaluated streams, FAN achieves the highest performance and demonstrates consistent robustness, providing insightful guidance for building stable action representations in achieving effective lifelong VLA adaptation.
Chinese Translation
在大规模封闭数据集上预训练的视觉-语言-动作（VLA）模型在各种机器人操作任务中表现出显著的成功。然而，它们在长期的实际部署中需要不断获取新技能，同时保持之前学习的能力。尽管一些开创性研究探讨了使用经验重放和强化微调等技术进行持续的VLA适应，但它们忽视了一个基础机制：动作归一化，它决定了策略感知和执行物理动作的基础坐标系统。为了填补这一空白，我们系统地评估了五种归一化策略在四个现实任务流中的表现，涵盖单臂和双手操作。我们的分析揭示，现有协议由于任务间坐标漂移、有限的运动覆盖或训练-测试坐标不匹配而导致严重的失败模式。基于这些见解，我们提出了三个核心设计原则：一致性、覆盖性和因果性（3C），并引入了前瞻性动作归一化（FAN）。FAN在持续学习之前从一个小的、与任务无关的校准集估计归一化统计量，并在适应过程中保持不变。在所有评估的任务流中，FAN实现了最高的性能，并展现出一致的鲁棒性，为构建稳定的动作表征以实现有效的终身VLA适应提供了深刻的指导。
View on arXiv Download PDF AI Translation
cs.RO / 45 / 2609.21365

MicroHookACT: Monocular Microscopic Vision Guided Visuomotor Policy for Flexible Microelectrode Hooking

MicroHookACT：基于单目显微视觉的灵活微电极钩取的视觉运动策略
Chen, Yitong, Qin, Fangbo, Wang, Yang, Hu, Ruihua, Zhang, Kui, Yu, Shan
Abstract
Automated needle-loop hooking is a critical step in flexible microelectrode (FME) implantation. This paper presents MicroHookACT, an imitation learning-based visuomotor policy for automated 3D hooking under monocular microscopic vision. First, a unidirectional hooking strategy exploits defocus cues and optical-axis guidance to enable palpation-free precise alignment and contact-rich threading. Second, an action-supervised object attention module built on a frozen ViT backbone learns to focus on the micro-needle tip and micro-loop directly from human demonstrations, without requiring manual visual annotations for training. Third, attention-centered global coarse and local fine features are dynamically weighted according to predicted action progress, enabling a single ACT policy to adapt to changing defocus blur and visual requirements throughout the operation. In the experiments, visuomotor policies were trained on 60 human demonstrations and evaluated under five setups with varying difficulties. Our MicroHookACT framework achieved the highest overall success rate of 96.7\% with an average execution time of 11.5 s. These results demonstrate the potential of visuomotor policy learning for micron-level control under varying operating conditions.
Chinese Translation
自动化针环钩取是灵活微电极（FME）植入中的关键步骤。本文提出了MicroHookACT，这是一种基于模仿学习的视觉运动策略，用于在单目显微视觉下实现自动化的三维钩取。首先，单向钩取策略利用散焦线索和光轴引导，实现无触诊的精确对齐和丰富接触的穿线。其次，基于冻结的ViT（Vision Transformer）骨干网构建的动作监督对象注意模块，能够直接从人类示范中学习关注微针尖和微环，而无需手动视觉标注进行训练。第三，基于注意力的全局粗特征和局部细特征根据预测的动作进展动态加权，使单一的ACT策略能够适应操作过程中变化的散焦模糊和视觉需求。在实验中，视觉运动策略在60个示范上进行训练，并在五种不同难度的设置下进行评估。我们的MicroHookACT框架实现了96.7%的最高整体成功率，平均执行时间为11.5秒。这些结果展示了视觉运动策略学习在变化操作条件下进行微米级控制的潜力。
View on arXiv Download PDF AI Translation
cs.RO / 46 / 2609.21369

ProTracer: Proprioception-Guided Failure Diagnosis in Robot Manipulation

ProTracer：基于本体感知的机器人操作故障诊断
Dong, Chang, Hosseinzadeh, Mehdi, Wong, King Hang, Liu, Lingqiao, Fraysse, Francois, Dayoub, Feras, Nguyen, Minh Hoai
Abstract
This paper presents a comprehensive framework for robot manipulation failure analysis that includes binary failure detection, failure categorization, explanation generation, and the additional capability of failure onset localization, which aims to identify the earliest moment at which a robot execution deviates from a valid task-completion trajectory and is ultimately followed by task failure. To address these tasks, we propose ProTracer, a training-free framework that leverages existing Vision-Language Models (VLMs) together with proprioceptive signals for failure analysis. Our method uses proprioceptive dynamics to identify temporally informative action boundaries and converts richer robot-state signals into structured natural-language descriptions that can be jointly analyzed together with visual observations by the VLM. This design combines the temporal precision of proprioceptive signals with the multimodal reasoning capabilities of modern VLMs without requiring additional model training. We further introduce FailTime, a benchmark with synchronized visual and proprioceptive observations for evaluating conventional failure diagnosis tasks as well as failure onset localization. Experiments demonstrate that ProTracer achieves strong performance across both conventional failure diagnosis tasks and the newly introduced failure onset localization task, highlighting the importance of proprioceptive reasoning for fine-grained temporal failure analysis.
Chinese Translation
本文提出了一种全面的机器人操作故障分析框架，包括二元故障检测、故障分类、解释生成，以及故障发生定位的附加能力，旨在识别机器人执行偏离有效任务完成轨迹的最早时刻，并最终导致任务失败。为了解决这些任务，我们提出了ProTracer，这是一种无训练框架，利用现有的视觉-语言模型（Vision-Language Models, VLMs）与本体感知信号相结合进行故障分析。我们的方法利用本体感知动态识别时间上具有信息量的动作边界，并将更丰富的机器人状态信号转换为结构化的自然语言描述，这些描述可以与视觉观测一起被VLM共同分析。该设计结合了本体感知信号的时间精确性与现代VLM的多模态推理能力，而无需额外的模型训练。我们进一步引入了FailTime，这是一个具有同步视觉和本体感知观测的基准，用于评估传统故障诊断任务以及故障发生定位。实验表明，ProTracer在传统故障诊断任务和新引入的故障发生定位任务中均表现出强劲的性能，突显了本体感知推理在细粒度时间故障分析中的重要性。
View on arXiv Download PDF AI Translation
cs.RO / 47 / 2609.21377

AVT-Fabric: Active Visuo-Tactile Perception via Adaptive Evidence Selection for Efficient Robotic Fabric Comparison

AVT-Fabric：通过自适应证据选择实现高效的机器人面料比较的主动视觉-触觉感知
Gao, Chang, Chen, Zhuo, Xia, Suhang, Zhu, Jihong, Deng, Jiankang, Luo, Shan
Abstract
Robotic fabric comparison needs to actively combine visual appearance and tactile cues. Here, we present AVT-Fabric, an RGB-first framework that allocates tactile evidence according to the difficulty of each comparison. A dual-scale gate evaluates answer-token confidence and raw logit separation to determine whether another force-tagged GelSight observation is needed. Compact textual memory preserves the executed history, and majority voting consolidates the selected predictions. On 400 held-out comparisons, AVT-Fabric achieves 98.0% accuracy with a compact 7B Multimodal Large Language Model (MLLM), surpassing the 94.0% reported by the 90B MLLM-Fabric baseline by 4.0 percentage points while processing only 1.60 of five available stages on average. It improves on matched passive inference by 9.25 percentage points and reduces model-side latency by 61.8%, while also improving on RGB-only accuracy. Four additional MLLM backbones support the generalizability, accuracy, and efficiency of the framework. This framework is also deployed on a real robotic system, achieving 78.1% pairwise ranking accuracy and correct fabric selection in seven of eight application scenarios. AVT-Fabric demonstrates that adaptive evidence selection can improve both the accuracy and efficiency of robotic visuo-tactile reasoning.
Chinese Translation
机器人面料比较需要主动结合视觉外观和触觉线索。本文提出了AVT-Fabric，这是一个RGB优先的框架，根据每次比较的难度分配触觉证据。双尺度门评估答案标记的置信度和原始对数分离，以确定是否需要另一个带有力标记的GelSight观测。紧凑的文本记忆保存了执行历史，而多数投票则巩固了所选预测。在400个保留的比较中，AVT-Fabric以紧凑的7B多模态大语言模型（MLLM）实现了98.0%的准确率，超越了90B MLLM-Fabric基线报告的94.0%，提高了4.0个百分点，同时平均仅处理五个可用阶段中的1.60个。它在匹配的被动推理上提高了9.25个百分点，并将模型侧延迟减少了61.8%，同时也提高了RGB单独准确率。四个额外的MLLM骨干网络支持该框架的可推广性、准确性和效率。该框架还在真实的机器人系统上部署，实现了78.1%的成对排名准确率，并在八个应用场景中的七个场景中正确选择了面料。AVT-Fabric展示了自适应证据选择可以提高机器人视觉-触觉推理的准确性和效率。
View on arXiv Download PDF AI Translation
cs.RO / 48 / 2609.21396

MarineCraft: Enabling Rapid Prototyping of Underwater Robots via Modular Construction

MarineCraft：通过模块化构建实现水下机器人快速原型制作
Sugiura, Yuta
Abstract
Underwater robot development is often hindered by the complexities of waterproofing and wiring, which significantly delay the rapid prototyping process. This paper presents MarineCraft, a modular toolkit designed to accelerate the development cycle through structural reconfiguration. The system features self-contained, waterproof propulsion modules that integrate power, wireless communication, and actuation. By eliminating centralized wiring and the need for repeated sealing, MarineCraft allows diverse robot geometries to be assembled and tested in minutes rather than days. Experimental results demonstrate that this reconfigurable architecture enables fast, iterative design cycles while maintaining reliable operation and leak-free performance at depths of up to 2.5 meters. Our toolkit effectively lowers the barrier to underwater robotics by transforming modularity into a vehicle for rapid physical prototyping.
Chinese Translation
水下机器人开发常常受到防水和布线复杂性的阻碍，这显著延迟了快速原型制作过程。本文提出了MarineCraft，一种旨在通过结构重构加速开发周期的模块化工具包。该系统具有自包含的防水推进模块，集成了电源、无线通信和驱动功能。通过消除集中布线和重复密封的需求，MarineCraft使得多样化的机器人几何形状能够在几分钟内组装和测试，而不是几天。实验结果表明，这种可重构架构能够实现快速、迭代的设计周期，同时在深度达到2.5米时保持可靠的操作和无泄漏性能。我们的工具包通过将模块化转变为快速物理原型制作的载体，有效降低了水下机器人技术的门槛。
View on arXiv Download PDF AI Translation
cs.RO / 49 / 2609.21404

Stabilizing Trajectory Outputs in End-to-End Autonomous Driving via SC-IMM Based Teacher Signals

通过基于 SC-IMM 的教师信号稳定端到端自主驾驶中的轨迹输出
Kim, Siewoo, Kong, Seung-Hyun
Abstract
End-to-End autonomous driving models commonly predict future waypoints from sensor inputs and convert them into vehicle control commands through a downstream controller. However, conventional waypoint-based imitation learning mainly minimizes coordinate-level errors, making it difficult to capture scene-dependent path-speed changes and temporal instability across waypoint outputs. In this paper, we propose an offline teacher-signal generation and learning method for trajectory-output stabilization based on a Scene-Conditioned Interacting Multiple Model (SC-IMM) to mitigate this issue. The proposed method converts expert trajectories into path-speed states and performs IMM updates conditioned on scene cues to generate path-speed teacher labels and mode posterior probabilities. The generated signals are added to the original trajectory loss as auxiliary supervision during training, while the inference structure and waypoint controller remain unchanged. In closed-loop evaluation on 100 short routes in CARLA Town12, the proposed method improved the driving score by 28.0% and reduced Collision/km by 62.3% compared with the baseline, while also improving jerk and trajectory-variation metrics. These results demonstrate that offline teacher signals embedding scene-conditioned motion-model cues can guide trajectory-output driving models toward more stable closed-loop behavior.
Chinese Translation
端到端自主驾驶模型通常根据传感器输入预测未来的航点，并通过下游控制器将其转换为车辆控制命令。然而，传统的基于航点的模仿学习主要最小化坐标级别的误差，这使得捕捉场景依赖的路径速度变化和航点输出的时间不稳定性变得困难。本文提出了一种基于场景条件交互多模型（SC-IMM）的离线教师信号生成和学习方法，以缓解这一问题。该方法将专家轨迹转换为路径速度状态，并基于场景线索执行IMM更新，以生成路径速度教师标签和模式后验概率。生成的信号作为辅助监督添加到原始轨迹损失中进行训练，同时推理结构和航点控制器保持不变。在CARLA Town12的100条短路线的闭环评估中，所提方法相比基线提高了28.0%的驾驶评分，并将每公里碰撞率降低了62.3%，同时改善了颠簸和轨迹变化指标。这些结果表明，嵌入场景条件运动模型线索的离线教师信号能够引导轨迹输出驾驶模型朝向更稳定的闭环行为。
View on arXiv Download PDF AI Translation
cs.RO / 50 / 2609.21416

A Unified Dynamic Force Guidance Framework for Performance-Optimized Kinesthetic Teaching

用于性能优化的统一动态力引导框架在动觉教学中的应用
Li, Chunxin, Wu, Jianhua, Xiong, Zhenhua, Zhu, Xiangyang
Abstract
Collaborative robots are increasingly deployed in industrial scenarios characterized by frequent product changeovers. As an intuitive programming method, kinesthetic teaching facilitates rapid robot deployment. However, users may overlook the configuration of the robot during kinesthetic teaching, leading to degradation in operational performance. Operational performance refers to the capability of the robot to generate motion and can be quantified by the Minimum Singular Value of the Jacobian matrix. To address this issue, this paper proposes an online dynamic force guidance method that integrates performance constraint and optimization mechanisms. Specifically, variable admittance control maintains the operational performance of the robot above a predefined threshold, while a virtual force actively guides the user to drag the robot towards configurations with improved performance. Experiments are conducted on a 6-DOF collaborative robot, comparing three typical paths in the task space. To evaluate the quality of the taught trajectories, trajectory playback experiments are conducted to analyze the relationship between the operational performance of the robot and the work efficiency. The results demonstrate that the proposed method effectively enhances the operational performance of the robot and consequently improves the work efficiency, holding significant value for reducing production takt time in industrial deployment.
Chinese Translation
协作机器人在频繁产品更换的工业场景中越来越多地被部署。作为一种直观的编程方法，动觉教学促进了机器人快速部署。然而，用户在动觉教学过程中可能会忽视机器人的配置，导致操作性能下降。操作性能是指机器人生成运动的能力，可以通过雅可比矩阵的最小奇异值来量化。为了解决这一问题，本文提出了一种在线动态力引导方法，该方法整合了性能约束和优化机制。具体而言，可变阻抗控制保持机器人的操作性能高于预定义阈值，而虚拟力则主动引导用户将机器人拖动至性能更优的配置。我们在一个6自由度的协作机器人上进行了实验，比较了任务空间中的三条典型路径。为了评估所教轨迹的质量，进行了轨迹回放实验，以分析机器人操作性能与工作效率之间的关系。结果表明，所提出的方法有效提升了机器人的操作性能，从而提高了工作效率，对减少工业部署中的生产节拍时间具有重要价值。
View on arXiv Download PDF AI Translation
cs.RO / 51 / 2609.21447

FootQuery: Future-Touchdown-Guided Retrieval from Depth History for Perceptive Humanoid Locomotion

FootQuery：基于未来着陆引导的深度历史检索用于感知类人步态
Dong, Tao, Yu, Jia, Fan, Yuxuan, Zhao, Linna, Gong, Jiaqi, Yang, Andong, Gao, Chao, Zhou, Guyue
Abstract
Humanoid locomotion over complex terrain requires anticipating footholds that may no longer be visible at touchdown. Limited camera coverage and self-occlusion make it necessary to retrieve relevant terrain information from earlier observations. We present FootQuery, a perceptive locomotion framework that queries depth history using each foot's predicted next touchdown. The policy predicts touchdown locations and uncertainty from proprioception and uses these distributions, together with per-foot features, to query sparsely sampled historical depth frames. During training, realized contacts are projected into historical images to supervise retrieval at the regions where those contacts were visible. The retrieved per-foot features are fused with global visual memory to generate control actions. A progressive force-assistance curriculum supports early exploration, while event-consistent tread-midline shaping encourages coordinated stair contacts. Deployment requires only proprioception and onboard depth images. In simulation, the complete framework outperforms its component ablations on the most challenging tested stairs, gaps, and platforms. Real-world experiments on a Unitree G1 demonstrate continuous traversal with a single policy across outdoor stairs and indoor routes combining stair ascent and descent, platforms, and gaps. These results support organizing visual history around anticipated contacts for perceptive humanoid locomotion.
Chinese Translation
在复杂地形上进行类人步态需要预测在着陆时可能不再可见的支撑点。有限的摄像头覆盖范围和自遮挡使得从早期观察中检索相关地形信息成为必要。我们提出了FootQuery，这是一种感知步态框架，通过每只脚的预测下一个着陆点来查询深度历史。该策略根据本体感知预测着陆位置和不确定性，并利用这些分布以及每只脚的特征来查询稀疏采样的历史深度帧。在训练过程中，实际接触点被投影到历史图像中，以监督在这些接触点可见区域的检索。检索到的每只脚特征与全局视觉记忆融合，以生成控制动作。渐进式的力辅助课程支持早期探索，而事件一致的踏步中线塑形则鼓励协调的楼梯接触。部署仅需本体感知和机载深度图像。在仿真中，完整框架在测试的最具挑战性的楼梯、间隙和平台上优于其组件消融。对Unitree G1的实际实验展示了在户外楼梯和室内路线（结合楼梯上升和下降、平台和间隙）中使用单一策略的连续行进。这些结果支持围绕预期接触组织视觉历史，以实现感知类人步态。
View on arXiv Download PDF AI Translation
cs.RO / 52 / 2609.21448

Robotic Multiphase Interaction: Manipulating Coupled Liquid and Solid Dynamics with a World Model

机器人多相交互：利用世界模型操控耦合液体和固体动力学
Feng, Yixuan, Wang, Peng
Abstract
This work presents \textit{Robotic Multiphase Interaction (RMI)}, a setting in which liquid enters a porous material and interacts mechanically with its deforming solid skeleton. Manipulation can therefore change pore volume, expel or redistribute retained liquid, and alter grasp stability at the same time. Spilled liquid can also create safety risks in domestic and manufacturing settings. This differs from most manipulation of solid objects and from tasks that involve both liquid and solid while keeping the phases spatially separate. We study a sponge filled with water as the first RMI example. We use implicit incompressible porous flow with smoothed particle hydrodynamics as the dynamics engine and enable robotic manipulation by adding Coulomb contact memory, hybrid velocity and force regulation, and a stability gate for lifting. The resulting environment connects robot commands to changes in the coupled liquid and solid state. A world model conditioned on actions predicts how this state evolves under candidate commands, while a temporal UNet generates actions using either Diffusion Policy or rectified flow matching. Our world model reduces retained water prediction error by more than $60\%$ compared with the baseline. The best action sequence selected by the world model from policy proposals further reduces the predicted terminal water error by about half. These improvements show that modelling the coupled liquid and solid state helps the robot predict how its actions affect both the porous object and the liquid held inside.
Chinese Translation
本研究提出了 extit{机器人多相交互（RMI）}，这是一个液体进入多孔材料并与其变形的固体骨架机械交互的场景。因此，操控可以同时改变孔隙体积、排出或重新分配滞留液体，并改变抓取稳定性。溢出的液体在家庭和制造环境中也可能带来安全风险。这与大多数固体物体的操控以及涉及液体和固体但保持相位空间分离的任务有所不同。我们以充满水的海绵作为第一个RMI示例。我们使用隐式不可压缩多孔流动与平滑粒子流体动力学作为动力学引擎，并通过添加库仑接触记忆、混合速度和力调节，以及提升稳定性门来实现机器人操控。所得到的环境将机器人指令与耦合液体和固体状态的变化连接起来。基于动作的世界模型预测该状态在候选指令下的演变，而时间UNet则使用扩散策略或修正流匹配生成动作。与基线相比，我们的世界模型将滞留水预测误差降低了超过$60\%$。世界模型从策略提案中选择的最佳动作序列进一步将预测的终端水误差减少了约一半。这些改进表明，建模耦合液体和固体状态有助于机器人预测其动作如何影响多孔物体及其内部所含液体。
View on arXiv Download PDF AI Translation
cs.RO / 53 / 2609.21461

AtomEgo: Exploring Ego-Robot Integration for Embodied Foundation Model Pretraining

AtomEgo：探索自我机器人集成用于具身基础模型的预训练
Wu, Di, Zheng, Dongchen, Sheng, Junhe, Wei, Zhongxing, Zhang, Songxin, Xie, Zejian, Sun, Xiaoquan, Zheng, Junyang, Song, Zhuoyang, Zhang, Jiaxing, Chen, Jiayu
Abstract
Embodied foundation models are constrained by the limited scale and diversity of robot demonstrations, motivating the use of large-scale egocentric human interaction data. However, how to effectively incorporate such data into embodied-model pre-training remains unclear because of substantial embodiment and action-space gaps between humans and robots. We present AtomEgo, a systematic study of ego--robot co-training supported by a curated corpus of approximately 2,659 hours and a scalable data processing pipeline. Across vision--language--action and world--action model architectures, we investigate three representative paradigms: joint co-training with domain-specific action heads, progressive ego-to-robot transfer through embodiment alignment, and joint video--action modeling. We evaluate these paradigms through multi-task real-robot experiments and language-conditioned cross-embodiment representation analysis. Our results reveal a simple principle: Data Scale * Alignment Quality --> Capability Gain; egocentric data can improve generalization, but their value depends on how effectively they are aligned and utilized. This principle can provide practical guidance for scalable ego--robot pre-training.
Chinese Translation
具身基础模型受到机器人演示的规模和多样性限制，这促使我们使用大规模的自我中心人类互动数据。然而，如何有效地将这些数据纳入具身模型的预训练仍然不明确，因为人类与机器人之间存在显著的具身性和动作空间差距。我们提出了AtomEgo，这是一个系统性的自我与机器人共同训练的研究，支持一个经过精心策划的语料库，包含约2659小时的数据和一个可扩展的数据处理管道。在视觉-语言-动作和世界-动作模型架构中，我们研究了三种代表性范式：带有领域特定动作头的联合共同训练，通过具身对齐的渐进式自我到机器人转移，以及联合视频-动作建模。我们通过多任务真实机器人实验和语言条件下的跨具身表示分析评估这些范式。我们的结果揭示了一个简单的原则：数据规模 * 对齐质量 --> 能力提升；自我中心数据可以改善泛化能力，但其价值取决于它们的对齐和利用效果。这一原则可以为可扩展的自我-机器人预训练提供实用指导。
View on arXiv Download PDF AI Translation
cs.RO / 54 / 2609.21467

Learning Distance-Conditioned Object Transport for Humanoid Loco-Manipulation from a Single Motion Clip

基于距离条件的类人运动操控物体运输学习：来自单一运动片段的研究
Hwang, Yuhyeon, Jung, Daniel Sungho, Seo, YongHyeok, Jung, Mingi, Cho, Chang Nho, Hwang, Jung-Hoon, Shin, Dongin
Abstract
Motion tracking can reproduce humanoid loco-manipulation from a single retargeted motion clip, but a policy trained on a fixed reference primarily reproduces its demonstrated transport outcome. Although the source trajectory visits intermediate object displacements, transport termination is demonstrated only at its endpoint. We identify this mismatch as the termination-versus-passage gap: intermediate displacements are observed as passage states rather than termination-complete outcomes. We introduce Distance-Conditioned Reference Recomposition (DCRR), which relocates the demonstrated termination segment to intermediate transport states. A frozen tracking teacher replays the recomposed references under closed-loop dynamics, and the retained trajectories are relabeled by their achieved object placements and distilled into a reference-free policy. This procedure constructs distance-conditioned supervision from the interaction behavior encoded in the source motion. Across Carry, Kick-Push, Crouch-Push, and Drag, DCRR-BC produces command-dependent transport with an overall normalized distance mean absolute error (MAE) of 0.15, compared with 0.28 for source-only behavior cloning. RL fine-tuning further improves the command response and execution robustness in the training simulator and under sim-to-sim transfer. Finally, hardware experiments demonstrate transport-distance modulation across all four interaction modes.
Chinese Translation
运动跟踪可以从单一的重定向运动片段中再现类人运动操控，但在固定参考上训练的策略主要再现其示范的运输结果。尽管源轨迹访问了中间物体位移，但运输终止仅在其终点处被示范。我们将这种不匹配称为终止与经过状态之间的差距：中间位移被视为经过状态，而不是终止完成的结果。我们提出了距离条件参考重组（Distance-Conditioned Reference Recomposition, DCRR），该方法将示范的终止片段重新定位到中间运输状态。一个冻结的跟踪教师在闭环动态下重放重组的参考，保留的轨迹根据其实现的物体放置进行重新标记，并被提炼成无参考策略。该过程从源运动中编码的交互行为构建距离条件监督。在携带（Carry）、踢推（Kick-Push）、蹲推（Crouch-Push）和拖动（Drag）四种任务中，DCRR-BC产生了命令依赖的运输，整体归一化距离的平均绝对误差（MAE）为0.15，而仅使用源行为克隆的MAE为0.28。强化学习（RL）微调进一步提高了训练模拟器中的命令响应和执行鲁棒性，并在模拟到模拟的转移中表现出色。最后，硬件实验展示了在所有四种交互模式下的运输距离调制。
View on arXiv Download PDF AI Translation
cs.RO / 55 / 2609.21482

Adaptive Rollout Truncation Based on Epistemic Uncertainty for Efficient Offline World Model Training

基于认知不确定性的自适应回滚截断用于高效的离线世界模型训练
Zymla, Nikodem Sebastian, Thiele, Laurin, Pitz, Johannes
Abstract
Accurate neural world models are central to model-based robotics, where they enable robots to predict future states from previously observed trajectories. Multi-step autoregressive training improves long-horizon prediction, but fixed rollout horizons also increase computational cost and can amplify early training errors when the model is still inaccurate. Existing training schemes typically use the same rollout length throughout optimization, independent of the model's current predictive reliability. We propose an epistemic uncertainty-driven adaptive rollout strategy for offline world model training following an auto-curriculum training scheme. Instead of always unrolling to a fixed horizon, the model terminates autoregressive rollouts once epistemic uncertainty exceeds a threshold calibrated from a warm-up phase. We study two uncertainty estimators: a five-head ensemble with a shared recurrent backbone and Monte Carlo Dropout. A two-stage warm-up procedure stabilizes uncertainty estimates before we enable adaptive truncation. Experiments on ANYmal-D and ANT show that ensemble-based adaptive truncation matches or improves the prediction accuracy of fixed-horizon training and the RWM-U baseline while requiring substantially fewer cumulative rollout steps. Training a world model on ANYmal-D following the presented approach reaches comparable final performance with the baselines with roughly 72% less rollout computation. These results indicate that epistemic uncertainty is useful not only for downstream policy regularization, but also for making world model training itself more compute-efficient.
Chinese Translation
准确的神经世界模型是基于模型的机器人技术的核心，它使机器人能够从先前观察到的轨迹中预测未来状态。多步自回归训练提高了长时间预测的能力，但固定的回滚时间范围也增加了计算成本，并且在模型仍不准确时可能放大早期训练错误。现有的训练方案通常在优化过程中使用相同的回滚长度，而不考虑模型当前的预测可靠性。我们提出了一种基于认知不确定性的自适应回滚策略，用于遵循自适应课程训练方案的离线世界模型训练。模型在认知不确定性超过从热身阶段校准的阈值后终止自回归回滚，而不是始终展开到固定的时间范围。我们研究了两种不确定性估计器：一个具有共享递归骨干的五头集成模型和蒙特卡洛丢弃法（Monte Carlo Dropout）。一个两阶段的热身程序在启用自适应截断之前稳定不确定性估计。在ANYmal-D和ANT上的实验表明，基于集成的自适应截断与固定时间范围训练和RWM-U基线的预测准确性相匹配或有所提高，同时所需的累计回滚步骤显著减少。在ANYmal-D上按照所提出的方法训练的世界模型在最终性能上与基线相当，但回滚计算减少了约72%。这些结果表明，认知不确定性不仅对下游策略正则化有用，而且还可以使世界模型训练本身更具计算效率。
View on arXiv Download PDF AI Translation
cs.RO / 56 / 2609.21497

FORTE: Task-Adaptive Force Capability Optimization for Mobile Manipulators

FORTE：移动操纵器的任务自适应力能力优化
Wang, Xiao, Zhang, Heng, Solak, Gokhan, Zhao, Fei, Ajoudani, Arash
Abstract
Effective physical interaction control in robotic manipulation requires not only kinematically feasible motion but also sufficient force-interaction capability. Existing redundancy resolution methods often ignore task-specific force demands or maximize the force capability indiscriminately, sacrificing dexterity when large force margins are unnecessary. We propose a task-oriented force capability optimization framework for redundant mobile manipulators. A Vision-Language Model (VLM) infers object physical properties from an RGB image and a task description, generating a desired task-force sequence that captures gravitational and inertial demands. We then define a task-oriented force capability metric as the signed distance between a task-force uncertainty ball and the dynamic residual force polytope (RFP), quantifying compatibility between task demands and the robot's remaining actuation capacity. This metric is incorporated, alongside manipulability, joint-limit avoidance, trajectory smoothness, and base-oscillation suppression, into a whole-body multi-objective trajectory-optimization problem. Experiments on a mobile manipulator performing lifting and single-point-holding tasks under varying payload conditions demonstrate that the proposed method provides sufficient force capability for heavy loads while preserving high manipulability for light loads. This yields a task-adaptive balance that fixed capability-maximizing baselines (RFP inscribed radius, RFP cone) and manipulability-only optimization fail to achieve. The core implementation is publicly available at https://github.com/yeying256/FORTE.
Chinese Translation
在机器人操作中，有效的物理交互控制不仅需要运动在运动学上可行，还需要足够的力交互能力。现有的冗余解决方法往往忽视特定任务的力需求，或不加区分地最大化力能力，在不必要的大力矩情况下牺牲灵活性。我们提出了一种面向任务的力能力优化框架，适用于冗余移动操纵器。视觉-语言模型（Vision-Language Model, VLM）从RGB图像和任务描述中推断物体的物理属性，生成捕捉重力和惯性需求的期望任务力序列。然后，我们定义了一种面向任务的力能力度量，作为任务力不确定性球与动态剩余力多面体（Residual Force Polytope, RFP）之间的有符号距离，量化任务需求与机器人剩余驱动能力之间的兼容性。该度量与可操控性、关节限制规避、轨迹平滑性和基座振荡抑制等因素一起，纳入了一个全身多目标轨迹优化问题。在不同负载条件下，移动操纵器执行提升和单点保持任务的实验表明，所提方法在保证重载情况下提供足够的力能力的同时，仍能为轻载情况保持高可操控性。这实现了一种任务自适应的平衡，而固定能力最大化基线（RFP内切半径，RFP锥体）和仅优化可操控性的方法未能达到。核心实现已公开于 https://github.com/yeying256/FORTE。
View on arXiv Download PDF AI Translation
cs.RO / 57 / 2609.21504

DPed-VLN: A Benchmark for Socially Compliant Vision-and-Language Navigation in Dynamic Pedestrian Environments

DPed-VLN：动态行人环境中社会合规的视觉与语言导航基准
Dai, Haojie, Wang, Xiangyi, Wang, Liuyi, Sheng, Kai, He, Zongtao, Liu, Chengju, Ye, Wei, Chen, Qijun
Abstract
Vision-and-language navigation (VLN) has advanced rapidly in static indoor environments, but robots operating in human-populated spaces must ground language while responding to moving pedestrians and social-safety constraints. We present DPed-VLN, a Habitat 3.0 benchmark for dynamic-pedestrian VLN that couples 33,093 navigation episodes with paired global and prior-augmented instructions, ORCA-controlled humanoid pedestrians, socially constrained expert paths, and metrics that jointly assess navigation efficiency and social safety. DPed-VLN separates ordinary goal-oriented route guidance from prior-augmented instructions that expose dynamic-pedestrian cues for controlled analysis. To instantiate the benchmark, we introduce DPet (Dynamic Pedestrian-aware Network), a pedestrian-aware policy network trained with reinforcement learning and imitation learning. We further adapt representative state-of-the-art VLM-based navigation models, including NaVILA and StreamVLN, to DPed-VLN through LoRA fine-tuning. Experiments show that LoRA adaptation improves zero-shot VLM baselines in several success and safety metrics, especially reducing StreamVLN's collision rate. Among the evaluated methods, DPet-RL achieves the highest SR, SPL, and STL.
Chinese Translation
视觉与语言导航（VLN）在静态室内环境中迅速发展，但在人员密集的空间中操作的机器人必须在响应移动行人和社会安全约束的同时对语言进行理解。我们提出了DPed-VLN，这是一个用于动态行人VLN的Habitat 3.0基准，结合了33,093个导航情节与配对的全局和先前增强的指令、由ORCA控制的人形行人、社会约束的专家路径，以及共同评估导航效率和社会安全的指标。DPed-VLN将普通的目标导向路线指导与暴露动态行人线索的先前增强指令分开，以便进行受控分析。为了实现该基准，我们引入了DPet（动态行人感知网络），这是一个通过强化学习和模仿学习训练的行人感知策略网络。我们进一步通过LoRA微调，将代表性的最先进的基于VLM的导航模型（包括NaVILA和StreamVLN）适配到DPed-VLN。实验表明，LoRA适配在多个成功和安全指标上改善了零样本VLM基线，特别是降低了StreamVLN的碰撞率。在评估的方法中，DPet-RL达到了最高的成功率（SR）、成功路径长度（SPL）和时间效率（STL）。
View on arXiv Download PDF AI Translation
cs.RO / 58 / 2609.21511

2nd Place Solution to the HANDS 2026 Workshop Challenge-Dexterous Grasp Motion Track: Single-Shot Trajectory Warping for Grasp Motion Generation

HANDS 2026 研讨会挑战赛第二名解决方案——灵巧抓取运动轨迹：用于抓取运动生成的单次轨迹变形
Khan, Muneeb A., Kim, Woojin, Kim, Shinwoo, Munsif, Muhammad, Bhattarai, Binod, Baek, Seungryul
Abstract
This report describes our 2nd place solution to the HANDS 2026 workshop challenge (Dexterous Grasp Motion track) in conjunction with ECCV 2026. In this challenge, we address grasp motion generation for the 12-DoF LinkerHand O6, aiming to produce physically plausible reach-and-lift trajectories for unseen objects from randomized initial hand poses in simulation. This task is particularly challenging because each grasp requires a per-step policy to make approximately $70$ twelve-dimensional decisions, with errors accumulating over time, while test objects and physical dynamics may differ from those encountered during training. To address these challenges, we propose editing a single successful GraspM3 demonstration instead of generating the motion step by step: a policy observes the object once and outputs a 12-D warp of the demonstration, which is then replayed open-loop. Moreover, we train the warp policy with one-step PPO over all $4{,}824$ training objects in parallel. As a result, our method achieved success rates of $94.61\%$ on the easy track, the highest of all submissions, and $57.18\%$ on the hard track of the private test set.
Chinese Translation
本报告描述了我们在与 ECCV 2026 结合的 HANDS 2026 研讨会挑战赛（灵巧抓取运动轨迹）中获得的第二名解决方案。在此挑战中，我们针对 12 自由度的 LinkerHand O6 进行抓取运动生成，旨在从随机的初始手部姿势中为未见物体生成物理上合理的伸手和提起轨迹。该任务尤其具有挑战性，因为每个抓取需要每一步采取大约 $70$ 个十二维决策，且错误会随着时间的推移而累积，同时测试物体和物理动态可能与训练过程中遇到的情况不同。为了解决这些挑战，我们提出编辑一个成功的 GraspM3 演示，而不是逐步生成运动：一个策略观察物体一次并输出演示的 12 维变形，然后以开环方式重放。此外，我们通过一阶 PPO 在所有 $4{,}824$ 个训练物体上并行训练变形策略。结果，我们的方法在简单轨道上取得了 $94.61\%$ 的成功率，是所有提交中最高的，而在私有测试集的困难轨道上取得了 $57.18\\%$ 的成功率。
View on arXiv Download PDF AI Translation
cs.RO / 59 / 2609.21514

Skel-WAM: A Hand-Skeleton-Conditioned World Action Model for Human-to-Robot Manipulation Transfer

Skel-WAM：一种基于手部骨架的世界动作模型，用于人机操作转移
Cai, Zetao, Li, Yaping, Wang, Yiqun, Zhan, Xinyu, Yang, Yuyin, Ma, Haoxiang, Li, Kailin, Lu, Tao, Pang, Jiangmiao, Xu, Linning, Lin, Dahua
Abstract
Robot demonstrations are expensive to collect and often provide limited distributional coverage of task variations. Human videos offer a low-cost source of complementary manipulation experience, but learning from them requires bridging embodiment gaps in visual appearance and action spaces. We introduce Skel-WAM, a world action model that bridges these differences through a unified hand-skeleton motion interface. The key insight is to align human and robot motion through a common hand topology, combining skeleton overlays that ground motion in the scene with structured 2.5-D keypoints that encode explicit hand kinematics. Video and Keypoint Experts jointly learn visual and skeletal dynamics through a Mixture-of-Transformers, while a separate robot-trained Action Expert maps these predictions to executable controls. This separation enables human and robot demonstrations to directly supervise shared dynamics without requiring robot action labels for human videos. Across four real-world bimanual tasks and seven simulated tasks, Skel-WAM achieves average success rates of 79.86% and 63.29%, surpassing the strongest baseline by 22.22 and 8.28 percentage points, respectively. Human-robot cotraining more than doubles real-world success on task variations absent from robot training data, from 38.89% to 86.11%. These results demonstrate that a shared skeletal interface enables joint learning across human and robot data and expands robot task coverage through complementary human demonstrations.
Chinese Translation
机器人演示的收集成本高昂，且通常只能覆盖有限的任务变体分布。人类视频提供了一种低成本的补充操作经验来源，但从中学习需要弥合视觉外观和动作空间的体现差距。我们提出了Skel-WAM，这是一种通过统一的手部骨架运动接口来弥合这些差异的世界动作模型。关键的见解在于通过共同的手部拓扑对齐人类和机器人运动，结合在场景中固定运动的骨架叠加和编码明确手部运动学的结构化2.5-D关键点。视频和关键点专家通过混合变换器（Mixture-of-Transformers）共同学习视觉和骨架动态，而一个单独训练的机器人动作专家则将这些预测映射到可执行控制。这种分离使得人类和机器人演示能够直接监督共享动态，而无需人类视频的机器人动作标签。在四个真实世界的双手任务和七个模拟任务中，Skel-WAM的平均成功率分别为79.86%和63.29%，超越最强基线22.22和8.28个百分点。人机共同训练使得在机器人训练数据中缺失的任务变体的真实世界成功率从38.89%提高到86.11%。这些结果表明，共享的骨架接口使得人类和机器人数据之间的联合学习成为可能，并通过补充的人类演示扩展了机器人任务的覆盖范围。
View on arXiv Download PDF AI Translation
cs.RO / 60 / 2609.21572

SABER: Learning Attention-based Semantic Affordance for Legged Locomotion

SABER：基于注意力的语义可接触性学习用于腿部运动
Palanivelu, Hari Prasanth, Sze, Samuel, Johannes, Kennard Garrison, Adiwahono, Albertus Hendrawan, Yee, Meng, Chuah
Abstract
Perceptive legged locomotion has advanced rapidly by integrating terrain geometry into learned policies, yet the integration of terrain meaning remains sparse: a pipe, a patch of grass, or a fragile box may be geometrically traversable while being inappropriate for contact. In industrial environments, where legged robots increasingly operate, a single misplaced step can damage fragile equipment, destabilize the robot, or endanger the site. To address this, we introduce SABER, a planner-free reinforcement-learning policy that jointly reasons about terrain geometry and semantic contact permission. The policy consumes a unified terrain-affordance map, where each cell encodes local 3D geometry and a semantic contact cost. We augment cross-attention with a learned, signed semantic bias: an additive term on the attention logits, gated by the contact cost, that reweights flagged cells by their distance from the nearest foot. A hazard therefore reshapes attention where it can still affect the next foothold, and its influence fades where it cannot. The resulting policy selects footholds on permitted support and keeps the leg clear of forbidden regions throughout the swing phase. We perform a systematic ablation that isolates the contribution of each architectural component; removing the semantic bias alone increases forbidden contacts by 55% while velocity tracking is unchanged. We validate the policy on a Unitree B2, demonstrating sim-to-real semantic contact selection across indoor and outdoor environments and four semantic obstacle classes.
Chinese Translation
感知式腿部运动通过将地形几何信息整合到学习策略中迅速发展，但对地形意义的整合仍然较为稀缺：一根管道、一片草地或一个脆弱的箱子在几何上可能是可穿越的，但在接触上却可能不合适。在工业环境中，腿部机器人越来越多地进行操作，单一步伐失误可能会损坏脆弱的设备、使机器人失去稳定性或危及现场安全。为了解决这一问题，我们提出了SABER，一种无规划的强化学习策略，能够共同推理地形几何和语义接触许可。该策略使用统一的地形可接触性地图，其中每个单元格编码局部3D几何和语义接触成本。我们通过学习到的有符号语义偏差增强了交叉注意力：这是一个在接触成本的控制下，对注意力逻辑进行加权的附加项，根据距离最近的脚的远近重新加权标记的单元格。因此，危险会重新塑造注意力，尽管它仍然可以影响下一个支撑点，而其影响在无法影响的地方会减弱。最终的策略在允许的支撑点上选择支撑，并在摆动阶段保持腿部远离禁止区域。我们进行了系统的消融实验，以隔离每个架构组件的贡献；单独去除语义偏差会使禁止接触增加55%，而速度跟踪保持不变。我们在Unitree B2上验证了该策略，展示了在室内和室外环境以及四类语义障碍物中进行的模拟到真实的语义接触选择。
View on arXiv Download PDF AI Translation
cs.RO / 61 / 2609.21580

Tilt as a Certified Resource: Preserving Motor Wrench-Rate Authority on Articulated Multirotors

倾斜作为认证资源：在关节式多旋翼飞行器上保持电机扭矩速率的权威性
Silano, Giuseppe, Saska, Martin
Abstract
Fully-actuated multirotor aerial vehicles must not only track nominal wrenches but retain the "readiness" to modulate them rapidly under disturbances. Classical effort-minimizing allocators ignore this dynamic limit, whereas maximizing readiness leads to topologically disconnected optimal sheets demanding physically impossible actuator rates. Enforcing a readiness safety floor on fixed-geometry symmetric platforms further encounters a zero-sum degeneracy: motor-speed redistribution cannot improve authority without conceding wrench tracking. This paper uses active morphology to break the degeneracy, treating servo tilt as a geometric resource supplying authority-recovery directions unavailable to static rotors. We construct a configuration-dependent, motor-only readiness certificate - the log-volume of the reachable wrench-rate set - that explicitly excludes servo capacity, preventing a "ghost capacity fallacy" in which the certificate would falsely credit slow mechanical kinematic limits instead of collapsing accurately at motor saturation. The certificate is enforced as a Control Barrier Function (CBF) within a Unified Physical-Command Quadratic Program acting on motor torques and servo setpoints. Closed-loop simulations of an articulated octorotor under severe gust disturbances show classical allocators diverging and uncertified articulated allocators violating the safety floor, while the proposed CBF filter bounds the system state and preserves vehicle authority.
Chinese Translation
全驱动多旋翼空中载具不仅必须跟踪名义扭矩，还需在扰动下保持快速调节的“准备状态”。经典的努力最小化分配器忽视了这一动态限制，而最大化准备状态则导致拓扑上不连通的最优表面，要求物理上不可能的电机速率。在固定几何对称平台上强制执行准备状态安全底线进一步遇到零和退化：电机速度的重新分配无法在不放弃扭矩跟踪的情况下改善权威性。本文利用主动形态学打破退化，将伺服倾斜视为一种几何资源，提供静态旋翼无法获得的权威恢复方向。我们构建了一个依赖于配置的仅电机准备状态证书——可达扭矩速率集的对数体积——明确排除了伺服能力，防止了“幽灵能力谬误”，即证书错误地归功于缓慢的机械运动学限制，而不是在电机饱和时准确崩溃。该证书作为控制障碍函数（Control Barrier Function, CBF）在统一物理指令二次规划（Unified Physical-Command Quadratic Program）中执行，作用于电机扭矩和伺服设定点。在严重阵风扰动下，对关节式八旋翼的闭环仿真显示，经典分配器发散且未经认证的关节式分配器违反了安全底线，而所提议的CBF滤波器限制了系统状态并保持了飞行器的权威性。
View on arXiv Download PDF AI Translation
cs.RO / 62 / 2609.21584

A High-Payload Wall-Climbing Robot Using Passive Bistable Suction Cups

一种使用被动双稳态吸盘的高载荷壁面爬行机器人
Nguyen, Andrew, Li, Mingyuan, Bruder, Daniel
Abstract
Wall-climbing robots capable of scaling vertical surfaces could help automate hazardous or labor intensive tasks such as window washing, inspection, maintenance, and construction. Active adhesion methods achieve higher payload capacities, but require power to maintain their grip. Passive adhesion devices such as suction cups are an attractive option for such robots because they do not require power to maintain their grip, but they are limited by their payload capacity. This work presents a novel high-payload wall-climbing robot that utilizes passive bistable suction cups to generate adhesion without needing to be pushed into the wall. The robot features a track-based system that automatically engages and disengages bistable suction cups to achieve locomotion on smooth surfaces. The robot is able to achieve vertical wall climbing on glass, wood, metal, and painted surfaces, sideways and upside-down climbing, and is able to tow a payload of 7.940 kg (with a payload-to-weight ratio of 2.25).
Chinese Translation
能够攀爬垂直表面的壁面爬行机器人可以帮助自动化危险或劳动密集型任务，如清洗窗户、检查、维护和施工。主动附着方法能够实现更高的载荷能力，但需要电力来维持其抓握。被动附着设备，如吸盘，是此类机器人的一种有吸引力的选择，因为它们不需要电力来维持抓握，但其载荷能力受到限制。本研究提出了一种新型高载荷壁面爬行机器人，利用被动双稳态吸盘在不需要推入墙壁的情况下生成附着力。该机器人采用基于轨道的系统，能够自动启用和禁用双稳态吸盘，以在光滑表面上实现运动。该机器人能够在玻璃、木材、金属和涂漆表面上实现垂直壁面爬行，能够进行侧向和倒立爬行，并能够拖动7.940千克的载荷（载荷与重量比为2.25）。
View on arXiv Download PDF AI Translation
cs.RO / 63 / 2609.21609

Potential-Field Action Representation for Reinforcement Learning in Contact-Rich Manipulation

用于接触丰富操作的强化学习的势场动作表示
Liu, Xinyu, Solak, Gökhan, Ajoudani, Arash
Abstract
Model-free reinforcement learning can acquire contact-rich robotic manipulation skills through trial-and-error interaction, but it often requires the policy to learn both task strategy and low-level motion generation. In this setting, the action representation is critical because it determines how policy outputs are converted into robot motion, shaping both exploration and physical execution. Direct Cartesian command interfaces require the policy to generate motion at every decision step, coupling task-level adaptation with continuous low-level control and increasing the learning burden. We propose PA-RL, a reinforcement-learning framework that uses artificial potential fields as the action representation. Instead of commanding motion directly, the policy adapts the parameters of an energy-like potential field, which generates a state-dependent guidance direction executed through a Cartesian impedance controller. We evaluate PA-RL on peg-in-hole insertion, a representative contact-rich task with nonlinear dynamics and discontinuous contact transitions. In simulation, PA-RL is compared with Cartesian velocity, Cartesian pose, and variable-impedance action spaces using the same RL algorithm. PA-RL is the only method to reach a 100% evaluation success rate within the allotted training time, while the best baseline reaches 92.6%. It also reduces joint-torque variation by 55.4% and Cartesian acceleration variation by 70.8% relative to the best baseline, without explicit motion-quality penalties in the reward. The simulation-trained policy further completes 9/9 real-robot insertions without fine-tuning, demonstrating the deployment feasibility of the learned potential-field interface.
Chinese Translation
无模型强化学习可以通过试错交互获取接触丰富的机器人操作技能，但通常需要策略同时学习任务策略和低级运动生成。在这种情况下，动作表示至关重要，因为它决定了策略输出如何转化为机器人运动，影响探索和物理执行。直接的笛卡尔指令接口要求策略在每个决策步骤生成运动，将任务级适应与连续的低级控制耦合，增加了学习负担。我们提出了PA-RL，一个使用人工势场作为动作表示的强化学习框架。策略不直接指挥运动，而是调整类似能量的势场参数，生成通过笛卡尔阻抗控制器执行的状态依赖指导方向。我们在插销入孔这一具有非线性动态和不连续接触转变的代表性接触丰富任务上评估PA-RL。在仿真中，PA-RL与使用相同强化学习算法的笛卡尔速度、笛卡尔姿态和可变阻抗动作空间进行了比较。PA-RL是唯一在规定训练时间内达到100%评估成功率的方法，而最佳基线达到了92.6%。与最佳基线相比，它还将关节扭矩变化减少了55.4%，笛卡尔加速度变化减少了70.8%，且在奖励中没有明确的运动质量惩罚。经过仿真训练的策略进一步完成了9/9次真实机器人插入，而无需微调，展示了学习的势场接口的部署可行性。
View on arXiv Download PDF AI Translation
cs.RO / 64 / 2609.21617

CounterPlay: Counterfactual Post-Training for Self-Play Driving Policies

CounterPlay：自我对抗驾驶策略的反事实后训练
Wei, Jiarong, Wu, Yin, He, Runkai, Valada, Abhinav
Abstract
Self-play in high-throughput simulators yields driving policies with robust closed-loop performance, but improvement per unit of simulation diminishes as training scales. Policies learn to handle common situations early, while further rollouts repeatedly encounter unresolved failures. Post-training offers an opportunity to target these failures, but existing methods primarily evaluate alternative actions or continuations at visited states, although successful recovery may require changing driving style earlier. We propose CounterPlay, a counterfactual self-play post-training approach that backtracks from failed tasks and retries them under alternate driving styles. CounterPlay rests on three key components. First, failure-driven backtracking uses the policy's value estimates to select an earlier stored state from which to retry the task. Second, reward conditioning enables a single policy to retry the task from this state using candidate styles ranging from cautious to aggressive. Third, CounterPlay retains task-completing retries only if no other vehicle incurs a new or earlier collision or off-road event relative to the factual branch. Retries that pass verification with fresh randomness are then distilled into the policy under its deployment condition. On BehaviorBench, CounterPlay achieves state-of-the-art scores on both the Interactive and Random splits across all eight traffic regimes using 1B post-training transitions, which is just 1% of the anchor's 100B self-play training budget. Improvements over the anchor hold across all three evaluated driving styles. CounterPlay resolves a substantial fraction of the anchor's timeout cases on BehaviorBench and achieves a balance between task completion and safety that neither continued self-play nor adopting a more aggressive driving style attains.
Chinese Translation
在高吞吐量模拟器中进行自我对抗训练能够产生具有稳健闭环性能的驾驶策略，但随着训练规模的扩大，每单位模拟的改进效果逐渐减弱。策略在早期学习处理常见情况，而后续的回合则反复遭遇未解决的失败。后训练提供了针对这些失败的机会，但现有方法主要评估在已访问状态下的替代动作或继续，尽管成功的恢复可能需要更早地改变驾驶风格。我们提出了CounterPlay，一种反事实自我对抗后训练方法，它从失败的任务回溯并在不同的驾驶风格下重试这些任务。CounterPlay基于三个关键组件。首先，失败驱动的回溯利用策略的价值估计选择一个早期存储的状态，从该状态重试任务。其次，奖励条件化使得单一策略能够从该状态重试任务，使用从谨慎到激进的候选风格。第三，CounterPlay仅在相对于事实分支没有其他车辆发生新的或早期的碰撞或越界事件时，才保留完成任务的重试。通过新的随机性验证的重试随后被提炼到策略中以供其部署条件下使用。在BehaviorBench上，CounterPlay在所有八种交通模式下，使用10亿次后训练过渡，达到了互动和随机分割的最先进分数，这仅占锚点1000亿自我对抗训练预算的1%。在评估的三种驾驶风格中，CounterPlay的改进在所有情况下均保持有效。CounterPlay解决了BehaviorBench上锚点大量超时案例，并在任务完成与安全之间实现了平衡，这一平衡是继续自我对抗训练或采用更激进的驾驶风格所无法达到的。
View on arXiv Download PDF AI Translation
cs.RO / 65 / 2609.21621

Towards Fine-Grained Object Manipulation: SAM3-Guided Visuomotor Policy with Persistent Memory Learning and Focused Visual Conditioning

面向细粒度物体操作：基于SAM3的视觉运动策略与持久记忆学习和聚焦视觉条件
Meng, Haolong, Qin, Fangbo, Bai, Mengchen, Wang, Houwu, Liu, Cirong, Yu, Shan
Abstract
Fine-grained object (FO) manipulation requires robots to distinguish a specified FO from visually similar objects and execute actions reliably despite scene distractors. However, scene-level visual conditioning lacks explicit object selection, while category-level guidance cannot reliably distinguish FOs within the same category. We present a SAM3-guided visuomotor framework that addresses these challenges through persistent object memory and focused visual conditioning. First, we introduce FO Memory-driven SAM3 (FOM-SAM3), which learns reusable FO memory tokens from limited multi-view registration images while keeping SAM3 fully frozen. Through one-vs-rest learning, these tokens encode persistent memories for localizing target FOs and rejecting similar alternatives, which can be stored in a memory bank. Second, we propose Focused Spatial-Appearance Encoding (FSAE), which combines in-FO local appearance features with explicit bounding-box coordinates to condition action policies including Diffusion Policy (DP) and Action Chunking with Transformers (ACT). The effectiveness of the proposed FOM-SAM3 was validated on the FO-30 dataset comprising 30 physical objects across four coarse categories. Across three real-robot FO manipulation tasks, our FOM-SAM3-guided policies demonstrated robustness against distractors, discrimination ability among similar FOs, and extendibility to new FOs.
Chinese Translation
细粒度物体（FO）操作要求机器人能够区分指定的FO与视觉上相似的物体，并在场景干扰下可靠地执行动作。然而，场景级视觉条件缺乏明确的物体选择，而类别级指导无法可靠地区分同一类别中的FO。我们提出了一种基于SAM3的视觉运动框架，通过持久的物体记忆和聚焦视觉条件来解决这些挑战。首先，我们引入了FO记忆驱动的SAM3（FOM-SAM3），该方法从有限的多视角注册图像中学习可重用的FO记忆标记，同时保持SAM3完全冻结。通过一对多学习，这些标记编码了用于定位目标FO和拒绝相似替代品的持久记忆，可以存储在记忆库中。其次，我们提出了聚焦空间外观编码（FSAE），该方法将FO内部的局部外观特征与明确的边界框坐标结合起来，以条件化包括扩散策略（Diffusion Policy, DP）和基于变换器的动作分块（Action Chunking with Transformers, ACT）在内的动作策略。我们在FO-30数据集上验证了所提出的FOM-SAM3的有效性，该数据集包含四个粗略类别中的30个物理对象。在三个真实机器人FO操作任务中，我们的FOM-SAM3指导的策略展示了对干扰的鲁棒性、在相似FO之间的区分能力，以及对新FO的扩展性。
View on arXiv Download PDF AI Translation
cs.RO / 66 / 2609.21650

SynthDemo-RL: Breaking the Zero-Reward Barrier in VLA Adaptation with LLM-Guided Synthetic Demonstrations

SynthDemo-RL：通过LLM指导的合成示范突破VLA适应中的零奖励障碍
Kingetsu, Hiroaki, Kurihara, Hiroaki, Yokoo, Kaoru, Fukumizu, Kenji, Kaul, Manohar
Abstract
Fine-tuning Vision-Language-Action (VLA) models commonly relies on human teleoperation demonstrations, while reinforcement learning (RL) with sparse binary rewards faces an exploration challenge when successful trajectories are rarely sampled. We propose SynthDemo-RL, a teacher-student framework in which an automated teacher converts simulator-privileged state into successful manipulation trajectories, a VLA student is distilled from them by supervised fine-tuning (SFT), and PPO with binary task-success rewards refines the student. We study reward coverage, the fraction of tasks for which at least one success is observed under the fixed evaluation protocol, as a complement to the average success rate. On LIBERO-PRO, a public benchmark of perturbed LIBERO tasks for which no demonstrations exist, 27 of 57 scored tasks are at exactly 0% success for a pi_0.5 policy fine-tuned on the original LIBERO tasks. Direct PPO from this policy, under the same PPO recipe and the same RL compute as SynthDemo-RL's refinement stage, rescues 10 of these 27 tasks and leaves 17 at 0%. SynthDemo-RL, with 50 synthesized trajectories per task and no new human demonstrations, rescues all 27 and reaches average success rates of 97.8% and 97.1% on the Position and Task axes of LIBERO-PRO, respectively. On standard LIBERO, the same pipeline reaches 96.0% with no human demonstrations, within 1.7 points of pi_0.5 trained on 50 human demonstrations per task. We further validate the pipeline on RoboTwin 2.0 and verify that trajectories from a policy trained in a MuJoCo twin execute open-loop on a physical robot.
Chinese Translation
微调视觉-语言-动作（VLA）模型通常依赖于人类遥操作示范，而稀疏二元奖励的强化学习（RL）在成功轨迹很少被采样时面临探索挑战。我们提出了SynthDemo-RL，这是一种教师-学生框架，其中自动化教师将模拟器特权状态转换为成功的操作轨迹，VLA学生通过监督微调（SFT）从中提炼，并且使用二元任务成功奖励的PPO进一步优化学生。我们研究了奖励覆盖率，即在固定评估协议下观察到至少一个成功的任务比例，作为平均成功率的补充。在LIBERO-PRO上，这是一项针对扰动LIBERO任务的公共基准，且没有示范存在，57个评分任务中有27个在对原始LIBERO任务微调的pi_0.5策略下成功率恰好为0%。在相同的PPO食谱和与SynthDemo-RL优化阶段相同的RL计算下，直接从该策略进行PPO，拯救了这27个任务中的10个，剩余17个仍为0%。SynthDemo-RL在每个任务中生成50条合成轨迹且没有新的人工示范，成功拯救所有27个任务，并在LIBERO-PRO的Position和Task轴上分别达到了97.8%和97.1%的平均成功率。在标准LIBERO上，同样的流程在没有人工示范的情况下达到了96.0%，与在每个任务上训练50个人工示范的pi_0.5相差仅1.7个点。我们进一步在RoboTwin 2.0上验证了该流程，并确认在MuJoCo双胞胎中训练的策略生成的轨迹可以在物理机器人上进行开环执行。
View on arXiv Download PDF AI Translation
cs.RO / 67 / 2609.21659

Outcome-Conditioned End-Effector Geometry Across Vision-Language-Action Policies

基于结果的末端执行器几何形状在视觉-语言-动作策略中的应用
Lin, Xingyu, Li, Zhuang, Wu, Zhongrun, Zhou, Shouquan, Du, Dehui
Abstract
Vision-language-action (VLA) policies solve the same manipulation task through different action interfaces, but task success alone does not establish whether their physical executions agree. We study cross-policy end-effector geometry in 15,000 closed-loop LIBERO rollouts from four policies. The primary clean-condition analysis forms 3,600 configuration-matched, and therefore dependent, policy pairs. Both-success pairs have a median normalized dynamic time warping distance of 0.0120 m versus 0.0380 m when exactly one policy succeeds. This ordering holds in every task, every policy pair, and nine sampling and band-limited representations; however, the ratio varies severalfold across representations, so we report the direction rather than a fixed multiple. Both-failure pairs are more separated again but rest on thin, uneven support, so we report them as exploratory. Within successful executions, partner replacements separate more across tasks than across initial states. A matched baseline still reveals measurable, heterogeneous residual policy differences, so a low cross-policy distance does not imply interchangeability. Successful executions sit about as far from same-task demonstrations as those demonstrations sit from each other, compatible with task-associated geometry without separating training-data overlap from task constraints. A common 72-action window preserves the ordering but reduces its magnitude; endpoint and duration adjustment likewise leaves a positive mixed-outcome coefficient relative to both-success pairs, though its magnitude is specification-dependent. Under composite visual stress, policy rankings and pair composition change together.
Chinese Translation
视觉-语言-动作（VLA）策略通过不同的动作接口解决相同的操作任务，但仅凭任务成功并不能确定其物理执行是否一致。我们研究了来自四种策略的15,000次闭环LIBERO执行中的跨策略末端执行器几何形状。主要的干净条件分析形成了3,600对配置匹配、因此相互依赖的策略对。当恰好有一个策略成功时，两者成功的策略对的中位归一化动态时间扭曲距离为0.0120米，而两者失败的策略对则为0.0380米。这一排序在每个任务、每对策略以及九种采样和带限表示中均成立；然而，比例在不同表示中变化数倍，因此我们报告方向而非固定倍数。两者失败的策略对再次分离，但依赖于薄而不均匀的支持，因此我们将其视为探索性结果。在成功的执行中，合作伙伴替换在任务之间的分离程度大于在初始状态之间的分离程度。匹配的基线仍然揭示了可测量的异质残余策略差异，因此低跨策略距离并不意味着可互换性。成功的执行距离同任务演示的距离大致与这些演示之间的距离相当，这与任务相关的几何形状相兼容，而没有将训练数据重叠与任务约束分开。一个常见的72个动作窗口保留了这一排序，但减少了其幅度；端点和持续时间的调整同样相对于两者成功的策略对留下了一个正的混合结果系数，尽管其幅度依赖于具体规范。在复合视觉压力下，策略排名和对的组成共同发生变化。
View on arXiv Download PDF AI Translation
cs.RO / 68 / 2609.21690

RAYA: Learning Where and When to Intervene for Robot Recovery

RAYA：学习何时何地进行干预以实现机器人恢复
Mahajan, Ishaan, Chen, Charles, Dümbgen, Frederike, Plancher, Brian
Abstract
A robot can predict failure and still be unable to prevent it. By the time a safety mechanism reacts, the nominal plan may already have spent the control authority that recovery requires, and fixed task priorities may block whatever response remains. Our key insight is that both aspects are decided inside the controller. Recoverability must inform actions while they are chosen rather than veto them afterward, and task objectives must be adapted as recoverability shrinks. Building on this, we present RAYA, a hybrid learned-analytic framework that places a learned finite-horizon recoverability margin inside an optimal controller with hard constraints and pairs it with a bounded learned scheduler that shifts task weights to facilitate recovery. Across 7,200 simulation episodes per controller spanning quadrotor and autonomous-vehicle benchmarks, RAYA not only improves survival rates, but also transfers the learned components zero-shot to unseen trajectories, disturbances, plant shifts, and friction layouts. We developed an embedded realization of RAYA and deployed it on-board a 35g Crazyflie quadrotor. Across 40 combined hardware flights under wind with either aerodynamic mismatch or an unmodeled 40% motor-command loss, each of three baselines fails in all trials, while RAYA completes 10/10 six-cycle missions. Project Website: https://raya-control.github.io/.
Chinese Translation
机器人可以预测故障，但仍然可能无法防止故障的发生。当安全机制反应时，正常计划可能已经耗尽了恢复所需的控制权，而固定的任务优先级可能会阻碍剩余的响应。我们的关键见解是，这两个方面都是在控制器内部决定的。恢复能力必须在选择行动时提供信息，而不是事后否决，并且随着恢复能力的减弱，任务目标必须进行调整。在此基础上，我们提出了RAYA，一个混合的学习-分析框架，它在一个具有严格约束的最优控制器内部放置了一个学习的有限视野恢复边际，并与一个有界的学习调度器配对，该调度器调整任务权重以促进恢复。在涵盖四旋翼和自主车辆基准的每个控制器的7200个仿真试验中，RAYA不仅提高了生存率，还能够将学习到的组件零样本迁移到未见过的轨迹、干扰、植物变化和摩擦布局。我们开发了RAYA的嵌入式实现，并将其部署在35克的Crazyflie四旋翼上。在40次结合硬件的飞行中，在风力影响下，无论是气动不匹配还是未建模的40%电机指令损失，三个基线在所有试验中均失败，而RAYA完成了10/10的六周期任务。项目网站：https://raya-control.github.io/
View on arXiv Download PDF AI Translation
cs.RO / 69 / 2609.21707

NeuRIO: A Streaming Neural Estimator for Zero-Shot Sim-to-Real Multi-Robot Relative Inertial Odometry

NeuRIO：一种用于零样本模拟到真实多机器人相对惯性里程计的流式神经估计器
Li, Zhehan, Lu, Jiadong, Ren, Shengwei, Xu, Chao, Cao, Yanjun
Abstract
We present NeuRIO, a streaming neural estimator for anchor-free 6-DoF relative inertial odometry using only identified inter-robot bearings, ranges, and IMU measurements. NeuRIO canonicalizes measurements into gravity-aligned coordinates, represents robots as nodes and mutual observations as factors, and uses attention for spatial reasoning and GRUs for temporal modeling. As a graph network, NeuRIO applies shared node-wise and factor-wise operators throughout the network, enabling it to handle different team sizes and time-varying observation graphs. NeuRIO is trained on a simulator that couples various motion patterns, device-level sensor characteristics, and diverse, realistic modeled, and temporally persistent sensor corruptions. In this way, NeuRIO achieves zero-shot sim-to-real transfer. Across $24$ real-world sequences, NeuRIO achieves $14.1\,\mathrm{cm}$ position RMSE and $3.9^\circ$ rotation RMSE. More importantly, NeuRIO demonstrates strong computational scalability, maintaining an update cost below $20\,\mathrm{ms}$ with up to $400$ robots in simulation, while optimization-based methods exceed $20\,\mathrm{ms}$ at only $24$ robots. Moreover, even trained on limited team sizes, NeuRIO transfers directly to unseen larger teams without architectural or parameter changes.
Chinese Translation
我们提出了NeuRIO，这是一种无锚点的6自由度相对惯性里程计的流式神经估计器，仅使用识别的机器人间方位、距离和IMU测量。NeuRIO将测量标准化为与重力对齐的坐标，将机器人表示为节点，将相互观察表示为因子，并使用注意力机制进行空间推理，使用门控递归单元（GRU）进行时间建模。作为一个图网络，NeuRIO在整个网络中应用共享的节点级和因子级操作符，使其能够处理不同团队规模和时间变化的观察图。NeuRIO在一个结合了各种运动模式、设备级传感器特性以及多样化、现实建模和时间持久的传感器干扰的模拟器上进行训练。通过这种方式，NeuRIO实现了零样本的模拟到真实转移。在24个真实世界序列中，NeuRIO实现了14.1 cm的位置均方根误差（RMSE）和3.9°的旋转均方根误差（RMSE）。更重要的是，NeuRIO展示了强大的计算可扩展性，在模拟中保持更新成本低于20 ms，支持多达400个机器人，而基于优化的方法在仅24个机器人时就超过了20 ms。此外，即使在有限的团队规模上训练，NeuRIO也能直接转移到未见过的更大团队，而无需架构或参数的改变。
View on arXiv Download PDF AI Translation
cs.RO / 70 / 2609.21716

AgenticSwarm: Semantic Perception and Adaptive Task Allocation for Heterogeneous Multi-UAV Missions

AgenticSwarm：异构多无人机任务的语义感知与自适应任务分配
Mustafa, Muhammad Ahsan, Yaqoot, Yasheerah, Batool, Faryal, Khan, Roohan Ahmed, Serpiva, Valerii, Tsetserukou, Dzmitry
Abstract
Multi UAV missions in complex environments require the system to understand both the surrounding scene and the intent of a human operator while maintaining feasible task allocation as mission conditions change. This paper presents AgenticSwarm, an agentic framework for semantic perception and adaptive task allocation in heterogeneous multi UAV missions. An agent interprets aerial imagery and natural language instructions to construct a grounded mission representation that links perceived objects and regions with task requirements, capability constraints, and mission dependencies. This information augments a constrained task allocation process in which obstacle aware path feasibility, energy consumption, and protected return home requirements are incorporated before assignment. During execution, changes such as UAV failure, battery degradation, or task modification trigger residual mission reconstruction from the current system state, while completed work and reconnaissance progress are retained. AgenticSwarm is evaluated across five diverse Gazebo environments and an indoor real test environment, demonstrating its ability to connect semantic reasoning with constrained allocation and adaptive multi UAV mission execution. Compared with a Grounding DINO+SAM~2.1 perception baseline, the SAM3-based pipeline improves class-aware recall by 25.2 percentage points (pp) and semantic label accuracy by 29.5 pp. Ablating residual mission replanning increases mean repeated work from 0% to 61.7% and post-event recovery time by 58.6%, highlighting the contribution of adaptive replanning to mission execution.
Chinese Translation
在复杂环境中的多无人机任务要求系统理解周围场景和人类操作员的意图，同时在任务条件变化时保持可行的任务分配。本文提出了AgenticSwarm，一个用于异构多无人机任务的语义感知和自适应任务分配的代理框架。代理通过解释空中图像和自然语言指令，构建一个将感知到的对象和区域与任务要求、能力约束和任务依赖关系相连接的基础任务表示。这些信息增强了一个受限的任务分配过程，在该过程中，在分配之前考虑了障碍物感知路径的可行性、能量消耗和保护性返航要求。在执行过程中，诸如无人机故障、电池衰减或任务修改等变化会触发从当前系统状态进行剩余任务重构，同时保留已完成的工作和侦察进展。AgenticSwarm在五个不同的Gazebo环境和一个室内真实测试环境中进行了评估，展示了其将语义推理与受限分配和自适应多无人机任务执行相连接的能力。与Grounding DINO+SAM~2.1感知基线相比，基于SAM3的管道提高了类别感知召回率25.2个百分点（pp）和语义标签准确率29.5 pp。消融剩余任务重规划将平均重复工作从0%提高到61.7%，事件后恢复时间减少58.6%，突显了自适应重规划对任务执行的贡献。
View on arXiv Download PDF AI Translation
cs.RO / 71 / 2609.21718

A Novel Path-Tracking Algorithm for Automated Tractor-Trailer Forward and Backward Maneuvers

一种新颖的自动拖拉机-挂车前进和倒退机动的路径跟踪算法
Lombard, Alexandre, Perronnet, Florent, Gaud, Nicolas, Abbas-Turki, Abdeljalil
Abstract
Fully autonomous tractor--trailer systems are increasingly deployed in logistics, agriculture, and industrial environments, where precise and robust path-tracking capabilities are essential. However, the articulation between the tractor and the trailer introduces additional nonlinearities and significantly complicates lateral and longitudinal control, particularly during reversing maneuvers. This paper introduces a novel path-tracking algorithm specifically designed for articulated vehicles with a single trailer. The proposed method combines a lateral control law applied at the trailer level with a short-horizon predictive adjustment of the tractor steering angle, ensuring stable convergence toward the desired path in both forward and backward motion. The approach is geometry-based and requires no per-vehicle calibration or training. Simulation studies in a high-fidelity physics simulator demonstrate the ability of the controller to match or outperform classical and state-of-the-art methods in terms of accuracy, stability, and robustness to disturbances.
Chinese Translation
全自动拖拉机-挂车系统在物流、农业和工业环境中的应用日益增多，其中精确且稳健的路径跟踪能力至关重要。然而，拖拉机与挂车之间的关节连接引入了额外的非线性，并显著复杂化了横向和纵向控制，尤其是在倒车机动时。本文提出了一种专门为单挂车的关节车辆设计的新颖路径跟踪算法。所提出的方法结合了应用于挂车层面的横向控制律与对拖拉机转向角的短期预测调整，确保在前进和倒退运动中稳定地收敛到期望路径。该方法基于几何原理，不需要对每辆车进行校准或训练。在高保真物理仿真器中的仿真研究表明，该控制器在准确性、稳定性和对干扰的鲁棒性方面能够与经典和最先进的方法相匹配或超越。
View on arXiv Download PDF AI Translation
cs.RO / 72 / 2609.21726

ZeroTouch: Tactile-Supervised Visual Contact Estimation for Contact-Rich Manipulation

ZeroTouch：用于接触丰富操作的触觉监督视觉接触估计
Kosenkov, Dmitriy, Zinniatullina, Daniia, Cabrera, Miguel Altamirano, Zhura, Iana, Derevianchenko, Mikhail, Tsetserukou, Dzmitry
Abstract
Reliable robotic grasping benefits from estimating the evolving physical interaction and selecting a grasp-dependent compression target. Tactile sensors provide direct interaction measurements but require dedicated hardware at deployment. We introduce ZeroTouch, a tactile-supervised framework that predicts dense contact deformation, the instantaneous six-axis wrench, and a grasp-dependent desired compression target from wrist RGB observations, gripper state, and local gravity direction. Tactile measurements are used only as privileged supervision during training and are not required at deployment. On the full validation set, the complete architecture reduces normal-force MAE from 2.017 N for a state-only baseline to 0.531 N. In physical evaluation with 20 trials per condition, ZeroTouch achieves 95% success on an unseen object, 80% in a seen-object/unseen-grasp condition, and 90% under a content/load shift. Under the same evaluation protocol, OpenVLA achieves 25%, 40%, and 55%, while SmolVLA achieves 10%, 25%, and 35%, respectively.
Chinese Translation
可靠的机器人抓取依赖于估计不断变化的物理交互并选择依赖抓取的压缩目标。触觉传感器提供直接的交互测量，但在部署时需要专用硬件。我们提出了ZeroTouch，一个触觉监督框架，它从手腕的RGB观察、夹持器状态和局部重力方向中预测密集的接触变形、瞬时的六轴扭矩和依赖抓取的期望压缩目标。触觉测量仅在训练期间作为特权监督使用，在部署时不需要。在完整的验证集上，整个架构将正常力的平均绝对误差（MAE）从仅使用状态的基线的2.017 N降低到0.531 N。在每种条件下进行20次试验的物理评估中，ZeroTouch在未见物体上实现了95%的成功率，在见过物体/未见抓取条件下为80%，在内容/负载转移下为90%。在相同的评估协议下，OpenVLA的成功率分别为25%、40%和55%，而SmolVLA的成功率分别为10%、25%和35%。
View on arXiv Download PDF AI Translation
cs.RO / 73 / 2609.21729

Visual Proactivity: Enhancing Human-Robot Collaboration Through Intent Communication

视觉主动性：通过意图沟通增强人机协作
Bo, Valerio, Bejarano, Edison, Garrell, Anaís, Sanfeliu, Alberto
Abstract
As robots transition from performing repetitive tasks to collaborating with humans, understanding human intent becomes crucial to effective interaction. Anticipation enables robots to predict human actions, while proactivity allows them to take initiative and guide human behavior toward optimal outcomes. Although research has largely focused on how robots infer and respond to human intentions, less attention has been paid to how robots communicate their own intent. This paper introduces visual proactivity, a novel, simple yet effective approach that enables robots to communicate their intentions through visual feedback, influencing human behavior and enhancing transparency and fluency. We develop and evaluate proactive robotic behaviors in a human-to-robot handover scenario, where a user study validates human perception of reactive, anticipatory, and proactive behaviors. The results demonstrate that effective visual proactivity fosters better alignment and coordination, paving the way for more intuitive human-robot collaboration.
Chinese Translation
随着机器人从执行重复性任务转向与人类协作，理解人类意图变得至关重要，以实现有效的互动。预判使机器人能够预测人类的行为，而主动性则使它们能够主动引导人类行为朝向最佳结果。尽管研究主要集中在机器人如何推断和响应人类意图上，但对机器人如何传达自身意图的关注相对较少。本文提出了一种视觉主动性的新颖、简单而有效的方法，使机器人能够通过视觉反馈传达其意图，从而影响人类行为，增强透明度和流畅性。我们在一个人机交接场景中开发并评估了主动的机器人行为，其中用户研究验证了人类对反应性、预期性和主动性行为的感知。结果表明，有效的视觉主动性促进了更好的对齐和协调，为更直观的人机协作铺平了道路。
View on arXiv Download PDF AI Translation
cs.RO / 74 / 2609.21734

When Should Robots Intervene? Balancing Engagement and Intrusiveness in Human-Robot Interaction

机器人何时应介入？在人机交互中平衡参与度与干扰性
Hriscu, Lavinia, Bo, Valerio, Sanfeliu, Alberto, Garrell, Anaís
Abstract
Designing effective Human-Robot Interaction in task-oriented settings requires carefully balancing user engagement with socially acceptable levels of robot intrusiveness. In this paper, we examine how different robot intervention strategies shape user experience, interaction dynamics, perceived intrusiveness, and sense of support. We compare two approaches: a continuous engagement-seeking robot strategy, and a context-aware strategy that selectively intervenes based on the user's state and task context. Both approaches rely on multimodal behavioral cues, including body orientation and attentional signals, to guide robot actions. We evaluate these strategies in a user study with 32 participants performing a task in a simulated hospital environment. Our findings show that higher interaction frequency does not necessarily lead to better engagement. Instead, we observe a systematic trade-off between perceived support and intrusiveness, influenced by factors such as physical proximity and user effort. These results provide empirical evidence that effective engagement in HRI depends on adaptive, context-sensitive intervention policies.
Chinese Translation
在任务导向的环境中设计有效的人机交互需要仔细平衡用户的参与度与机器人可接受的干扰水平。本文探讨了不同的机器人干预策略如何影响用户体验、交互动态、感知干扰性和支持感。我们比较了两种方法：一种是持续寻求参与的机器人策略，另一种是基于用户状态和任务背景进行选择性干预的上下文感知策略。这两种方法均依赖于多模态行为线索，包括身体朝向和注意信号，以指导机器人的行动。我们在一个模拟医院环境中对32名参与者进行的用户研究中评估了这些策略。我们的研究结果表明，更高的交互频率并不一定会导致更好的参与度。相反，我们观察到感知支持与干扰性之间存在系统性的权衡，这受到物理接近度和用户努力等因素的影响。这些结果提供了实证证据，表明在人机交互中有效的参与依赖于适应性和上下文敏感的干预政策。
View on arXiv Download PDF AI Translation
cs.RO / 75 / 2609.21740

Sandwich-Residuals: Parameter-Efficient Test-time Adaptation of World Models

三明治残差：世界模型的参数高效测试时适应
Soni, Krishnam, Sehgal, Aditya, Dave, Vedant, Rueckert, Elmar
Abstract
Latent world models enable planning by predicting the effects of actions in a learned representation space, but their predictions can become unreliable when test-time conditions differ from training. Existing test-time adaptation methods address this by updating parts of the pretrained model, often modifying millions of parameters and requiring a choice of which internal components to adapt. We introduce Sandwich-Residuals, a lightweight alternative that keeps the pretrained world model frozen and learns only small residual corrections around the predictor. The residuals are optimized online using the model's self-supervised prediction error and require no rewards, labels, or source-domain data. Across 21 conditions on the AdaJEPA benchmark, our method achieves $1.3\times$ the success rate of the frozen model while retaining 95% of the performance of the strongest AdaJEPA variant and adapting 97-99% fewer parameters. Under compound shifts, this advantage increases to $1.9\times$ the success rate of the frozen model, while remaining comparable to internal block adaptation. We further demonstrate the same adaptation principle on a DINO-WM model for 3-D manipulation. These results suggest that effective test-time adaptation of world models does not necessarily require modifying their pretrained internal weights.
Chinese Translation
潜在世界模型通过在学习的表征空间中预测动作的效果来实现规划，但当测试时条件与训练时不同，这些预测可能变得不可靠。现有的测试时适应方法通过更新预训练模型的部分组件来解决这个问题，通常会修改数百万个参数，并需要选择哪些内部组件进行适应。我们提出了三明治残差（Sandwich-Residuals），这是一种轻量级的替代方案，它保持预训练的世界模型不变，仅在预测器周围学习小的残差修正。这些残差通过模型的自监督预测误差在线优化，不需要奖励、标签或源域数据。在AdaJEPA基准的21种条件下，我们的方法实现了冻结模型成功率的1.3倍，同时保持了最强AdaJEPA变体95%的性能，并适应了97-99%更少的参数。在复合变化下，这一优势增加到冻结模型成功率的1.9倍，同时与内部块适应相当。我们进一步在DINO-WM模型上展示了相同的适应原理，用于3D操作。这些结果表明，世界模型的有效测试时适应并不一定需要修改其预训练的内部权重。
View on arXiv Download PDF AI Translation
cs.RO / 76 / 2609.21744

Understanding Engagement and Intrusiveness in Assistive Human-Robot Interaction Using Individual Traits

理解个体特征对辅助人机交互中的参与感和侵入感的影响
Bo, Valerio, Hriscu, Lavinia, Sanfeliu, Alberto, Garrell, Anaís
Abstract
Robot assistance is particularly crucial in unfamiliar tasks, where users must understand task requirements while coordinating with the robot. Previous research offers mixed evidence on the role of robot proxemics in user engagement: some studies suggest closer proximity enhances interaction, while others report it can feel intrusive. In this work, we argue that perceptions of intrusiveness depend not only on proxemics but also on the frequency of robot interventions, and are strongly influenced by individual traits such as personality and demographics. We conducted an experiment with 32 participants who interacted with two assistive robots that provided similar task support but differed in their intervention strategies. Results indicate that overall engagement remains stable across conditions, yet affective responses and perceived intrusiveness vary significantly with personality traits. Moreover, personality shapes interaction dynamics differently depending on the robot's behavior. These findings emphasize that effective human-robot interaction should account for individual differences, tailoring robot behavior to maintain engagement while respecting each user's unique affective and behavioral profile.
Chinese Translation
机器人辅助在不熟悉的任务中尤为重要，用户必须理解任务要求并与机器人协调。以往研究对机器人亲密距离在用户参与感中的作用提供了混合证据：一些研究表明，较近的距离可以增强互动，而另一些则报告称这可能会让人感到侵入。在本研究中，我们认为侵入感的感知不仅依赖于亲密距离，还与机器人干预的频率有关，并受到个体特征（如个性和人口统计特征）的强烈影响。我们对32名参与者进行了实验，他们与两台提供类似任务支持但干预策略不同的辅助机器人进行了互动。结果表明，整体参与感在不同条件下保持稳定，但情感反应和感知的侵入感在个性特征上存在显著差异。此外，个性在互动动态中对机器人的行为影响不同。这些发现强调，有效的人机交互应考虑个体差异，调整机器人行为以维持参与感，同时尊重每位用户独特的情感和行为特征。
View on arXiv Download PDF AI Translation
cs.RO / 77 / 2609.21751

ForceTwin: Physics-informed Digital Twins for Robotic Manipulation from Instrumented Human Interaction

ForceTwin：基于物理信息的数字双胞胎用于机器人操控中的人机交互
Engelbracht, Tim, Zurbrügg, René, Mittal, Mayank, Hutter, Marco, Pollefeys, Marc, Blum, Hermann, Bauer, Zuria
Abstract
Manipulating objects requires understanding not only their motion, but also the physical properties that determine it. For articulated objects, these include inertia, friction, and mechanisms such as springs or door closers, whose effects can vary with configuration and velocity. Such properties are not directly observable from appearance: visually identical doors may require very different effort to manipulate. Existing digital-twin pipelines recover primarily kinematics or assign static physical parameters from visual and language priors, which can yield physically implausible estimates. As a result, state-dependent mechanism dynamics remain unidentified and are not represented in standard asset formats. We present ForceTwin, a system for identifying physics-informed digital twins of articulated objects from instrumented human interaction. A person probes an object using a handheld force-sensing gripper, providing synchronized poses and interaction forces from which we estimate the articulation, parametric dynamics including inertia, Coulomb friction, viscous damping, and a structured neural residual capturing state-dependent mechanism forces. ForceTwin nearly halves the inertial-parameter error of a VLM prior. As a feedforward dynamics model for impedance control on a Spot and a Franka FR3, ForceTwin achieves 87% goal completion across nine object-embodiment pairs, compared with 60% using VLM-prior and 57% using kinematics-only twins, with the largest gains on objects whose strong mechanisms cause both baselines to stall. We further use the identified twins to train whole-body door-traversal policies and deploy them in the real world. Project Page: https://timengelbracht.github.io/forcetwin-website/
Chinese Translation
操控物体需要理解其运动以及决定运动的物理特性。对于关节物体，这些特性包括惯性、摩擦以及如弹簧或门闭合器等机制，其影响可能随配置和速度而变化。这些特性无法仅通过外观直接观察到：视觉上相同的门可能需要非常不同的操控力度。现有的数字双胞胎管道主要恢复运动学或从视觉和语言先验中分配静态物理参数，这可能导致物理上不合理的估计。因此，状态依赖的机制动态仍然未被识别，并且未在标准资产格式中表示。我们提出了ForceTwin，一个从仪器化的人机交互中识别关节物体的物理信息数字双胞胎的系统。通过手持力传感抓取器，一个人探测物体，提供同步的姿态和交互力，从中我们估计关节、参数化动态（包括惯性、库仑摩擦、粘性阻尼）以及捕捉状态依赖机制力的结构化神经残差。ForceTwin几乎将VLM先验的惯性参数误差减半。作为Spot和Franka FR3的阻抗控制前馈动态模型，ForceTwin在九对物体-体现中实现了87%的目标完成率，而使用VLM先验和仅使用运动学双胞胎的完成率分别为60%和57%，在那些强机制导致两个基线停滞的物体上获得了最大的提升。我们进一步利用识别出的双胞胎训练全身门穿越策略，并在现实世界中部署。项目页面：https://timengelbracht.github.io/forcetwin-website/
View on arXiv Download PDF AI Translation
cs.RO / 78 / 2609.21753

PSR: Predictive Sensorimotor Representation Learning for Contact-Rich Manipulation

PSR：用于接触丰富操作的预测性感觉运动表征学习
Li, Shengbao, Xu, Peng, Tang, Chao, Wei, Hao, Wang, Jiaheng, Yin, Hong, Chen, Jiangtao, Zhu, Jinxuan, Zhou, Zhong, Wang, Mengfan, Li, Tingguang
Abstract
Contact-rich manipulation requires policies to generate precise actions by reasoning over contact forces, robot configurations, and interaction histories beyond visual observations. Existing methods passively condition on force feedback rather than actively predicting future contact dynamics, limiting their ability to generate high-precision actions. To address this problem, we introduce Predictive Sensorimotor Representation (PSR) learning, a framework that learns a hierarchy of predictive representations from multimodal sensorimotor signals and integrates them into the action stream of a visuomotor policy. Specifically, during a pretraining stage, a multimodal Transformer is trained to learn a hierarchy of predictive representations by jointly forecasting future interaction dynamics. The learned hierarchy subsequently augments the action stream, enabling the resulting policy to exploit contact-relevant cues at multiple depths. We further instantiate PSR within a Vision-Language-Action (VLA) model, resulting in PSR-VLA, and evaluate it on six real-world contact-rich manipulation tasks. Experimental results show that PSR-VLA achieves 91.7% overall success, improving over $\pi_{0.5}$, ForceVLA-$\pi_{0.5}$, and ForceVLA2-$\pi_{0.5}$ by 30.0, 22.5, and 19.2 percentage points, respectively. These results demonstrate the effectiveness of the proposed PSR for force-aware, contact-rich manipulation. Videos of the tasks and stability tests are available at https://psr-vla.pages.dev/.
Chinese Translation
接触丰富的操作要求策略通过推理接触力、机器人配置和交互历史，生成精确的动作，超越视觉观察。现有方法被动地依赖于力反馈，而不是主动预测未来的接触动态，这限制了它们生成高精度动作的能力。为了解决这个问题，我们提出了预测性感觉运动表征（Predictive Sensorimotor Representation，PSR）学习框架，该框架从多模态感觉运动信号中学习预测性表征的层次结构，并将其整合到视觉运动策略的动作流中。具体而言，在预训练阶段，训练一个多模态Transformer，通过共同预测未来的交互动态来学习预测性表征的层次结构。学习到的层次结构随后增强了动作流，使得生成的策略能够在多个深度上利用与接触相关的线索。我们进一步在视觉-语言-动作（Vision-Language-Action，VLA）模型中实例化PSR，得到PSR-VLA，并在六个真实世界的接触丰富操作任务上进行了评估。实验结果表明，PSR-VLA的整体成功率达到91.7%，相比于$ ext{π}_{0.5}$、ForceVLA-$ ext{π}_{0.5}$和ForceVLA2-$ ext{π}_{0.5}$分别提高了30.0、22.5和19.2个百分点。这些结果证明了所提出的PSR在力感知、接触丰富操作中的有效性。任务和稳定性测试的视频可在https://psr-vla.pages.dev/获取。
View on arXiv Download PDF AI Translation
cs.RO / 79 / 2609.21761

CRISP: Contact-Rich Robotic Simulation Platform with Extensive Geometries and Contact Solvers

CRISP：具有广泛几何体和接触求解器的接触丰富机器人仿真平台
Lee, Somang, Park, Sunkyung, Yun, Jinhee, An, Seoki, Lee, Dongjun
Abstract
We present CRISP (Contact-RIch Simulation Platform), a high-fidelity physics engine tailored for complex multi-contact simulations such as tight-tolerance robotic manipulation. Achieving high physical fidelity in robotic simulation requires both expressive modeling of geometry and contact interactions, as well as accurate numerical resolution via robust collision detection and contact solvers. However, existing simulators often either rely on limited support for geometric representations and simplified modeling of contact interactions, or employ numerical resolution methods whose accuracy or robustness is inherently constrained. Accordingly, we develop a new simulator that supports diverse geometric representations with accurate optimization-based collision detection, and combines contact modeling with robust augmented Lagrangian-based contact solvers. This integration enables efficient and consistent detection of contact information across complex geometries while accurately resolving multi-contact constraints without problematic relaxations, which is essential for simulating contact-intensive and sharp interactions. We validate the physical fidelity of our simulator against state-of-the-art platforms and further demonstrate its capabilities through complex robotic manipulation scenarios. CRISP is publicly available at https://github.com/INRoL/crisp.
Chinese Translation
我们提出了CRISP（接触丰富仿真平台），这是一种高保真度的物理引擎，专为复杂的多接触仿真而设计，例如紧公差机器人操作。在机器人仿真中实现高物理保真度需要对几何体和接触交互进行表达性建模，以及通过稳健的碰撞检测和接触求解器进行准确的数值求解。然而，现有的仿真器往往要么依赖于对几何表示的有限支持和简化的接触交互建模，要么采用的数值求解方法在准确性或稳健性上受到固有限制。因此，我们开发了一种新的仿真器，支持多样的几何表示，并具有基于优化的准确碰撞检测，同时将接触建模与稳健的增广拉格朗日接触求解器相结合。这种集成使得在复杂几何体中高效且一致地检测接触信息成为可能，同时准确地解决多接触约束而不出现问题的松弛，这对于模拟接触密集和尖锐的交互至关重要。我们通过与最先进的平台进行比较，验证了我们仿真器的物理保真度，并通过复杂的机器人操作场景进一步展示了其能力。CRISP已在 https://github.com/INRoL/crisp 上公开发布。
View on arXiv Download PDF AI Translation
cs.RO / 80 / 2609.21767

Scaling Vision-Language Reward Learning for Robot Manipulation in Parallel Simulation

在并行仿真中扩展视觉-语言奖励学习以实现机器人操作
Joualy, Lobna, Demeester, Eric, Tsiogkas, Nikolaos
Abstract
Vision-language models (VLMs) can replace human annotators in preference-based reward learning, but sequential API requests and single-environment data collection make training slow and costly. We present RAPID (Reward learning with Adaptive Parallel Image Diversity), a system that couples GPU-parallel rollout with data-aware policy updates, single-request preference labeling, automatic reward stabilization, and representative image sampling. We evaluate these components on five Franka Panda manipulation tasks in IsaacLab. Parallel rollout and adaptive updates provide the first substantial reduction in training time: under matched two-stage prompting, mean runtime falls from 9.18 to 3.13 hours. With all RAPID components enabled, training completes in 1.15 hours using 896 rather than 19,840 API calls per run, and aggregate final success rises from 86.3\% to 98.7\%. This represents an 8.0$\times$ end-to-end speedup and a 95.5\% reduction in API usage. An offline evaluation with Gemma~3 12B and GPT-4.1 mini demonstrates that single-request prompting reduces labeling latency and cost across both models. Code is available at: https://github.com/rapid-vlm/rapid-vlm-rl.
Chinese Translation
视觉-语言模型（VLMs）可以替代人工标注者进行基于偏好的奖励学习，但顺序API请求和单环境数据收集使得训练过程缓慢且成本高昂。我们提出了RAPID（自适应并行图像多样性奖励学习），该系统结合了GPU并行展开与数据感知的策略更新、单请求偏好标注、自动奖励稳定和代表性图像采样。我们在IsaacLab的五个Franka Panda操作任务上评估了这些组件。并行展开和自适应更新首次显著减少了训练时间：在匹配的两阶段提示下，平均运行时间从9.18小时降至3.13小时。在启用所有RAPID组件的情况下，训练在1.15小时内完成，每次运行使用896个而不是19,840个API调用，最终成功率从86.3\%上升至98.7\%。这代表了8.0倍的端到端加速和95.5\%的API使用减少。与Gemma~3 12B和GPT-4.1 mini的离线评估表明，单请求提示减少了两个模型的标注延迟和成本。代码可在以下链接获取：https://github.com/rapid-vlm/rapid-vlm-rl。
View on arXiv Download PDF AI Translation
cs.RO / 81 / 2609.21777

TRACE: Coverage Path Planning for Unknown Environments Using Hierarchical Coverage Tree

TRACE：基于层次覆盖树的未知环境覆盖路径规划
Shen, Zongyuan, Liu, Haodong, Wang, Gao, Zhao, Shancheng, Zhou, Dehua, Ou, Yaming, Ren, Zhongqiang, Zhai, Yikui, Chen, C. L. Philip
Abstract
This paper presents a novel online coverage path planning (CPP) algorithm, called TRACE, for real-time coverage of unknown environments. TRACE is built upon a hierarchical coverage tree that provides a global representation of the evolving connectivity of the uncovered space. As the environment is incrementally revealed and covered, newly discovered obstacles and covered cells may fragment the remaining uncovered space into disconnected regions. TRACE recursively expands the corresponding tree nodes to explicitly represent these regions and organize them for subsequent coverage planning. Based on the updated tree, an incremental global tour is maintained to guide the coverage process. TRACE locally refines only the affected portions while preserving the visiting order of unchanged regions, thereby reducing the computational burden of global replanning and maintaining a consistent coverage progression. Guided by the global tour, a local planner generates back-and-forth coverage paths and switches to global-tour-aware planning to efficiently complete the target regions. Theoretical analysis establishes the computational complexity and complete coverage property of TRACE, and derives an approximation bound for the incremental global tour refinement. The performance of TRACE is evaluated through extensive high-fidelity simulations and real-robot experiments using a mobile robot. Comparative evaluations against six existing CPP methods demonstrate significant improvements in coverage time, path length, overlap ratio, and number of turns.
Chinese Translation
本文提出了一种新颖的在线覆盖路径规划（CPP）算法，称为TRACE，用于未知环境的实时覆盖。TRACE建立在一个层次覆盖树之上，该树提供了未覆盖空间不断演变的全局连接性表示。随着环境的逐步揭示和覆盖，新发现的障碍物和已覆盖单元可能会将剩余的未覆盖空间分割成不连通的区域。TRACE递归地扩展相应的树节点，以明确表示这些区域并为后续的覆盖规划进行组织。基于更新后的树，维护一个增量全局巡回路径以指导覆盖过程。TRACE仅对受影响的部分进行局部细化，同时保持未改变区域的访问顺序，从而减少全局重新规划的计算负担，并保持一致的覆盖进展。在全局巡回路径的指导下，局部规划器生成往返覆盖路径，并切换到考虑全局巡回路径的规划，以高效完成目标区域。理论分析建立了TRACE的计算复杂性和完整覆盖特性，并推导了增量全局巡回路径细化的近似界限。通过广泛的高保真模拟和使用移动机器人进行的真实机器人实验评估TRACE的性能。与六种现有CPP方法的比较评估显示，在覆盖时间、路径长度、重叠比和转弯次数方面有显著改善。
View on arXiv Download PDF AI Translation
cs.RO / 82 / 2609.21787

Compact but Moving: Intervention-Relevant Geometry in Recurrent World Models

紧凑但动态：递归世界模型中的干预相关几何
Chen, Yuming, Liu, Yang
Abstract
Learned world models may have compact interventions even when their recurrent state is high-dimensional, but it is unclear what happens to such a correction after it enters the model. We study this question in a controlled recurrent world model where prior work identified a checkpoint-specific rank-4 interface for one-shot counterfactual velocity interventions. The correction rapidly leaves this fixed entry subspace during autonomous rollout. Nevertheless, a low-rank image obtained by transporting the entry directions through the factual recurrent Jacobian chain continues to capture most of the nonlinear correction. Restarts using the tangent-predicted correction preserve substantial counterfactual future function. This transport/function pattern recurs across independently trained structured-GRU models and a parameter-matched LSTM initialized with a privileged compact correction. We further characterize a finite-horizon future-response operator over the full recurrent carrier. Patching shifts its leading future-sensitive directions toward the matched native-counterfactual organization, and the local operator accurately ranks finite perturbation effects over the registered direction panels at the patched and native-counterfactual basepoints. A separate full-amplitude assay finds substantial factual-endpoint tangent residuals and supports response reconfiguration in two of three checkpoints. Together, these results show that compact intervention structure can persist as a moving, state-dependent local geometry embedded in high-dimensional recurrent dynamics, without implying a fixed or dynamically closed low-dimensional state.
Chinese Translation
学习到的世界模型即使在其递归状态为高维时也可能具有紧凑的干预，但在干预进入模型后会发生什么仍不清楚。我们在一个受控的递归世界模型中研究这个问题，此前的研究识别出一个特定于检查点的四阶界面，用于一次性反事实速度干预。该修正迅速离开这个固定的输入子空间，在自主展开过程中。然而，通过将输入方向通过真实递归雅可比链进行传输获得的低秩图像，仍然能够捕捉到大部分非线性修正。使用切线预测的修正进行重启能够保留相当大的反事实未来函数。这种传输/函数模式在独立训练的结构化GRU模型和使用特权紧凑修正初始化的参数匹配LSTM中反复出现。我们进一步表征了一个有限时间范围的未来响应算子，覆盖整个递归载体。修补将其主要的未来敏感方向转向匹配的本地反事实组织，而局部算子准确地对注册方向面板在修补和本地反事实基点上的有限扰动效应进行排名。一个独立的全幅度测定发现了相当大的真实端点切线残差，并支持在三个检查点中的两个进行响应重构。综合来看，这些结果表明，紧凑的干预结构可以作为一种动态的、状态依赖的局部几何嵌入在高维递归动态中持续存在，而不意味着存在固定或动态闭合的低维状态。
View on arXiv Download PDF AI Translation
cs.RO / 83 / 2609.21788

From Pretraining to Proficiency: Real-World Subtask RL for Long-Horizon Manipulation with Minimal Human Intervention

从预训练到精通：最小人类干预下的长时间操作的真实子任务强化学习
Su, Sichang, Yang, Benjamin, Deng, Zhiyun, Liang, Boyuan, Yeung, Yip Fun, Wang, Zelin, Sun, Lingfeng
Abstract
A pretrained robot foundation policy may execute most of a long-horizon task yet repeatedly fail at a few critical subtasks. Collecting additional full-task demonstrations for supervised fine-tuning (SFT) requires operators to repeat behaviors the policy already performs well. Reinforcement learning (RL) fine-tuning offers a promising path to bridge this gap, but existing approaches struggle to solve long-horizon tasks using only sparse rewards. We present PARTS (Policy Adaptation with RL on Targeted Subtasks), a real-world subtask RL framework that concentrates practice at these bottlenecks while allowing training rollouts to proceed with minimal human intervention. The frozen pretrained policy supplies nominal actions throughout execution, while agent-generated selectors and success verifiers activate residual corrections and provide local outcome rewards. These rewards support learning from successful subtasks even when complete-task successes are scarce. Training combines online RL with success-reweighted retraining, and each retrained residual policy is redeployed to collect further experience. Humans identify bottlenecks during setup and perform physical resets when needed. On bimanual YAM and single-arm Franka tasks, PARTS improves complete-task success from 32% to 61% and from 50% to 95%, respectively, using tens of minutes of real-world RL rollouts per task on average. Compared with existing real-world RL fine-tuning methods, PARTS raises full-task success by more than 25% under the same robot-rollout budget while requiring less human involvement.
Chinese Translation
预训练的机器人基础策略可能能够执行大部分长时间任务，但在一些关键子任务上反复失败。为了进行监督微调（SFT），收集额外的完整任务演示需要操作员重复策略已经表现良好的行为。强化学习（RL）微调提供了一条有前景的路径来弥补这一差距，但现有方法在仅使用稀疏奖励解决长时间任务时面临挑战。我们提出了PARTS（针对特定子任务的强化学习政策适应），这是一个真实世界的子任务强化学习框架，专注于这些瓶颈的练习，同时允许训练过程在最小人类干预下进行。冻结的预训练策略在执行过程中提供名义动作，而代理生成的选择器和成功验证器激活剩余修正并提供局部结果奖励。这些奖励支持从成功的子任务中学习，即使完整任务的成功率很低。训练结合了在线强化学习与成功加权再训练，每个再训练的剩余策略被重新部署以收集更多经验。人类在设置过程中识别瓶颈，并在需要时进行物理重置。在双手YAM和单臂Franka任务中，PARTS将完整任务的成功率从32%提高到61%，以及从50%提高到95%，平均每个任务使用数十分钟的真实世界强化学习过程。与现有的真实世界强化学习微调方法相比，PARTS在相同的机器人过程预算下将完整任务成功率提高了超过25%，同时需要的人工干预更少。
View on arXiv Download PDF AI Translation
cs.RO / 84 / 2609.21792

AcousticDiffusion: Semantically Conditioned Audio-Guided Diffusion Policy for Search-and-Rescue Assistance

声学扩散：用于搜索与救援辅助的语义条件音频引导扩散策略
Zhura, Iana, Seyidov, Didar, Plotnikov, Dmitrii, Amjad, Hajira, Cabrera, Miguel Altamirano, Tsetserukou, Dzmitry
Abstract
Navigating toward human callers is an important capability for rescue robots operating where visual contact is degraded or occluded. We present AcousticDiffusion, a semantically conditioned, audio-guided diffusion policy for human-directed navigation. A frozen pretrained audio recognizer processes 10.24 s windows, with speech gating and distress-aware prioritization converting recognition outputs into source-level navigation roles. Microphone-array direction-of-arrival measurements are recursively integrated into a robot-centric Bayesian bird's-eye-view belief field. Ego-motion compensation aligns successive observations, progressively constraining source position while preserving bearing-induced range uncertainty. The semantic belief, recent acoustic observations, audio features, and robot state condition a diffusion model that generates waypoint trajectories. On a synthetic-navigation validation set using recorded audio, AcousticDiffusion achieves a mean end-point bearing error of 11.20 degrees, with 91.78% of trajectories aligned within 30 degrees of the caller. Distractor rejection ranges from 89.20% to 98.99%, and the policy favors a HELP-designated caller over a competing speaker in 91.07% of windows. Deployed online on a ZSL-1 quadruped without additional retraining, it achieves a mean bearing error of 64.9 degrees, compared with 98.2 degrees for A* and 90.4 degrees for RRT, with a mean planner compute time of 6.07 ms. Despite imperfect acoustic localization, the reported mean final source distance is reduced from 3.96 m for the classical planners using ODAS-derived (Open embedded Audition System) guidance to 2.48 m, a 37.4% improvement. These results demonstrate the framework's ability to translate uncertain acoustic observations into closer approaches to human callers.
Chinese Translation
在视觉接触受限或被遮挡的环境中，朝向人类呼叫者导航是救援机器人一项重要的能力。我们提出了声学扩散（AcousticDiffusion），这是一种语义条件的音频引导扩散策略，用于人类导向的导航。一个冻结的预训练音频识别器处理10.24秒的音频窗口，通过语音门控和关注困扰的优先级将识别输出转换为源级导航角色。麦克风阵列的到达方向测量被递归整合到以机器人为中心的贝叶斯鸟瞰视图信念场中。自我运动补偿对齐连续观察，逐步限制源位置，同时保持因方向引起的距离不确定性。语义信念、最近的声学观察、音频特征和机器人状态共同条件化一个扩散模型，该模型生成航点轨迹。在使用录制音频的合成导航验证集中，声学扩散实现了11.20度的平均终点方位误差，91.78%的轨迹与呼叫者的方位在30度以内。干扰物拒绝率在89.20%到98.99%之间，策略在91.07%的窗口中优先选择HELP指定的呼叫者而非竞争发言者。在没有额外再训练的情况下在线部署在ZSL-1四足机器人上，平均方位误差为64.9度，而A*算法为98.2度，RRT为90.4度，平均规划计算时间为6.07毫秒。尽管声学定位不完美，报告的平均最终源距离从使用ODAS（开放嵌入听觉系统）指导的经典规划者的3.96米减少到2.48米，改善幅度为37.4%。这些结果展示了该框架将不确定的声学观察转化为更接近人类呼叫者的能力。
View on arXiv Download PDF AI Translation
cs.RO / 85 / 2609.21803

Contact-Rich Motion Planning via GPU-Parallel Mode Evaluation

基于GPU并行模式评估的接触丰富运动规划
Li, Jiayun, Chalvatzaki, Georgia
Abstract
Contact-rich motion planning (CRMP) is essential for robotic manipulation and locomotion, yet remains computationally challenging due to combinatorial contact decisions. Existing methods typically avoid broad evaluation of contact-mode sequences through search heuristics or optimization reformulations. We revisit broad evaluation in light of modern GPU hardware and introduce Contact-Mode Expansion with parallel Trajectory optimization (CoMET), which combines GPU-parallel trajectory evaluation with greedy contact-mode expansion. On planar pushing benchmarks, CoMET is competitive with optimization-based, sampling, and tree-search baselines in solution quality and planning time, matching the full-enumeration reference on nearly all instances with fewer evaluations and shorter planning times. Ablations suggest that much of the performance gain comes from the high-throughput trajectory evaluator. In bimanual nonprehensile manipulation, GPU-friendly local mode expansion achieves higher planning success than the tested adaptive tree search as the mode space grows. These results demonstrate that broad explicit mode evaluation provides a simple yet effective alternative for CRMP.
Chinese Translation
接触丰富运动规划（CRMP）对于机器人操作和运动至关重要，但由于组合接触决策，仍然面临计算挑战。现有方法通常通过搜索启发式或优化重构来避免对接触模式序列的广泛评估。我们在现代GPU硬件的背景下重新审视广泛评估，并引入接触模式扩展与并行轨迹优化（CoMET），该方法将GPU并行轨迹评估与贪婪接触模式扩展相结合。在平面推送基准测试中，CoMET在解决方案质量和规划时间上与基于优化、采样和树搜索的基线具有竞争力，在几乎所有实例中以更少的评估和更短的规划时间匹配全枚举参考。消融实验表明，性能提升的主要来源是高吞吐量的轨迹评估器。在双手非抓取操作中，适合GPU的局部模式扩展在模式空间增长时实现了比测试的自适应树搜索更高的规划成功率。这些结果表明，广泛的显式模式评估为CRMP提供了一种简单而有效的替代方案。
View on arXiv Download PDF AI Translation
cs.RO / 86 / 2609.21817

A Sim-to-Real Integration Pipeline for Training and Deployment of Chunk-Based VLA Manipulation Policies

用于训练和部署基于块的视觉-语言-动作（VLA）操控策略的仿真到现实集成管道
Kappel, Mathilde, Grislain, Clémence, Chetouani, Mohamed, Sigaud, Olivier, Annabi, Louis, Amar, Fa\"ız Ben, Doncieux, Stéphane, Khoramshahi, Mahdi
Abstract
Vision-Language-Action (VLA) models have become a prominent paradigm for mapping multimodal inputs, including semantic instructions, visual observations of the scene, and proprioceptive observations, to robot actions. Most state-of-the-art models predict actions in the end-effector pose space as sequences of action chunks. Training and evaluating these models requires large-scale collections of real-world demonstrations, pairing robot actions with the corresponding visual and proprioceptive observations. Collecting such data on real hardware typically relies on human teleoperation, making the process costly, time-consuming, and difficult to scale. We present an open-source sim-to-real experimental protocol that addresses this bottleneck: expert trajectories generated in simulation are replayed open-loop on a real Franka FR3 setup, where the corresponding real visual and proprioceptive observations are recorded and converted into a format compatible with VLA training. The same deployment stack is then reused, in closed-loop, to evaluate a trained policy on that setup, so that data collection and evaluation share an identical hardware configuration. Because each real recording is paired with the simulated trajectory that produced it, the protocol also yields a direct measurement of the sim-to-real gap. We release the collected datasets on Hugging Face together with the pipeline source code https://gitlab.isir.upmc.fr/kappel/sim2real_public_chunk_control.
Chinese Translation
视觉-语言-动作（VLA）模型已成为将多模态输入（包括语义指令、场景的视觉观察和本体感知观察）映射到机器人动作的显著范式。大多数最先进的模型在末端执行器姿态空间中预测动作，作为动作块的序列。训练和评估这些模型需要大规模的真实世界演示数据，需将机器人动作与相应的视觉和本体感知观察配对。在真实硬件上收集此类数据通常依赖于人工遥控，这使得该过程成本高、耗时且难以扩展。我们提出了一种开源的仿真到现实实验协议，以解决这一瓶颈：在真实的Franka FR3设置上，回放在仿真中生成的专家轨迹，并记录相应的真实视觉和本体感知观察，将其转换为与VLA训练兼容的格式。然后，在闭环中重用相同的部署堆栈，以评估在该设置上训练的策略，从而使数据收集和评估共享相同的硬件配置。由于每个真实记录都与生成它的仿真轨迹配对，该协议还提供了仿真到现实差距的直接测量。我们将在Hugging Face上发布收集的数据集以及管道源代码 https://gitlab.isir.upmc.fr/kappel/sim2real_public_chunk_control。
View on arXiv Download PDF AI Translation
cs.RO / 87 / 2609.21818

LunaDrive: A Delay-Compensated High-Voltage GaN FET-Based Motor Driver for Dynamic Robots with Flat BLDC Motors

LunaDrive：一种基于高压GaN FET的延迟补偿电动机驱动器，用于配备平面无刷直流电动机的动态机器人
Yuzaki, Sota, Suzuki, Temma, Tada, Hiromi, Konishi, Masanori, Kawaharazuka, Kento, Okada, Kei
Abstract
The performance improvement of high-power flat BLDC motors has accelerated the development of dynamic robots. However, many commercially available servo motors assume operating voltages of 48 V or lower, which limits the maximum rotational speed. Dynamic robots require rapid energy generation, so this voltage constraint restricts motion performance. Therefore, driving motors beyond the rated voltage is desirable to increase the instantaneous maximum speed. On the other hand, semiconductor devices used in motor drivers have a trade-off between voltage rating and current capacity. Conventional drivers using Si MOSFETs have difficulty achieving both high-voltage and high-current operation. Although GaN FETs are promising, compact drivers that can be mounted on the rear side of flat BLDC motors remain limited. In this study, a motor driver for high-power flat BLDC motors using GaN FETs is developed. The effect of delay compensation in the high-speed region beyond the rated operating range is also investigated. In the experiments, under 96 V operation, a continuous current of 30 A was achieved with a heat sink attached. A peak current of 80 A and a maximum electrical frequency of 3110 Hz were confirmed. A high-speed load lifting experiment driven by a LiPo battery 24S (100 V) was also conducted, demonstrating applicability to dynamic robot operation.
Chinese Translation
高功率平面无刷直流电动机的性能提升加速了动态机器人的发展。然而，许多商业可用的伺服电动机假设工作电压为48 V或更低，这限制了最大转速。动态机器人需要快速的能量产生，因此这一电压限制限制了运动性能。因此，超出额定电压驱动电动机是提高瞬时最大速度的理想选择。另一方面，用于电动机驱动器的半导体器件在电压额定值和电流容量之间存在权衡。使用Si MOSFET的传统驱动器在实现高电压和高电流操作方面面临困难。尽管GaN FET具有良好前景，但能够安装在平面无刷直流电动机后侧的紧凑型驱动器仍然有限。本研究开发了一种用于高功率平面无刷直流电动机的GaN FET电动机驱动器，并研究了在超出额定操作范围的高速区域中延迟补偿的效果。在实验中，在96 V的操作下，附加散热器的情况下实现了30 A的连续电流。确认了80 A的峰值电流和3110 Hz的最大电气频率。还进行了由LiPo电池24S（100 V）驱动的高速负载提升实验，证明了其在动态机器人操作中的适用性。
View on arXiv Download PDF AI Translation
cs.RO / 88 / 2609.21838

PopNavShift: Stress-Testing Social Navigation under Behavioral Population Shift

PopNavShift：在行为人口变化下对社会导航进行压力测试
Tan, Kaizhen, Zheng, Diyu, Wu, Tim Guangyu, Guan, ChengHe
Abstract
Social-navigation algorithms are often evaluated under a fixed pedestrian-behavior distribution, despite substantial variation in pedestrian responses to robots across individuals and social contexts. We introduce PopNavShift, a matched simulation framework for stress-testing social-navigation strategies under pedestrian population shifts. PopNavShift constructs population-conditioned pedestrian motion profiles by prompting Gemini 3.7 Flash with 600 synthetic persona records from MatrAIx Persona 1M and deterministically mapping the responses into bounded motion parameters. It then compares three representative navigation strategies, reactive avoidance, early yielding, and reciprocal collision avoidance, across eight population conditions and 7,488 matched robot runs. In a matched intervention on the same 202 personas, changing only time pressure reverses 8.6% of controller rankings based on robot travel time, but 22.4% based on mean pedestrian delay and 23.9% based on worst-decile delay. Across population conditions, this sensitivity is greater for pedestrian burden than for robot travel time and increases in spatially constrained settings; the same qualitative pattern persists under a second pedestrian dynamics model. These findings support evaluating navigation strategies across behavioral populations using both robot performance and pedestrian burden.
Chinese Translation
社会导航算法通常在固定的行人行为分布下进行评估，尽管不同个体和社会背景下行人对机器人的反应存在显著差异。我们提出了PopNavShift，这是一个用于在行人群体变化下对社会导航策略进行压力测试的匹配仿真框架。PopNavShift通过使用来自MatrAIx Persona 1M的600个合成角色记录，提示Gemini 3.7 Flash构建基于人群条件的行人运动特征，并将反应确定性地映射到有限的运动参数中。然后，它在八种人群条件和7,488次匹配的机器人运行中比较三种代表性的导航策略：反应性避让、提前让行和互惠碰撞避免。在对同样的202个角色进行匹配干预时，仅改变时间压力就会使基于机器人旅行时间的控制器排名逆转8.6%，而基于平均行人延迟的排名逆转22.4%，基于最差十分之一延迟的排名逆转23.9%。在不同的人群条件下，这种敏感性在行人负担方面大于机器人旅行时间，并且在空间受限的环境中增加；在第二种行人动态模型下，这种定性模式仍然存在。这些发现支持在行为人群中使用机器人性能和行人负担评估导航策略。
View on arXiv Download PDF AI Translation
cs.RO / 89 / 2609.21883

VIRGA: Virtual-Agent-Intermediated Riemannian Geometry for Active-Sensing Air-Ground Coordination

VIRGA：用于主动感知的空地协调的虚拟代理中介黎曼几何
Guo, Fenghe, Shen, Runjie, Sun, Chenyang, Zhang, Junrui
Abstract
Air-ground autonomy becomes harder when the unmanned aerial vehicle (UAV) must remain observable by a gimbal light detection and ranging (LiDAR) mounted on the unmanned ground vehicle (UGV). The platforms must avoid dynamic obstacles while coordinating heterogeneous motion, limited sensing, and changing task initiative within one closed loop. This paper presents VIRGA, a neural geometric coordination framework that turns dual-LiDAR observations into bounded source-specific Riemannian fields and couples them through a virtual agent with reciprocal elastic feedback. Platform-aware execution maps convert the shared coordination reference into feasible UAV, UGV, and gimbal commands while enforcing active-observation safeguards. Evaluation against three complementary baselines reveals distinct limitations. An adapted Ray-RMP controller provides the fastest Riemannian response but produces insufficient clearance in the coupled air-ground task. A dense analytical Riemannian field improves geometric avoidance, yet its high evaluation cost prevents stable field-of-view maintenance. An adapted ColAG controller achieves the lowest latency but still incurs safety and observability violations. VIRGA completes all paired warehouse conditions safely, while a long-range cave stress test without retraining demonstrates sustained coordination in irregular and confined geometry. Ablations confirm contributions from online geometric evaluation, virtual-agent mediation, and reciprocal feedback.
Chinese Translation
当无人机（UAV）必须保持可被安装在无人地面车辆（UGV）上的云台激光雷达（LiDAR）观察时，空地自主性变得更加困难。这些平台必须在一个闭环内避免动态障碍，同时协调异构运动、有限感知和变化的任务主动性。本文提出了VIRGA，一个神经几何协调框架，将双LiDAR观测转化为有界的源特定黎曼场，并通过具有互惠弹性反馈的虚拟代理将其耦合。平台感知执行映射将共享的协调参考转换为可行的UAV、UGV和云台指令，同时强制执行主动观察保护措施。与三个互补基线的评估揭示了明显的局限性。改进的Ray-RMP控制器提供了最快的黎曼响应，但在耦合的空地任务中产生了不足的间隙。密集的解析黎曼场改善了几何避让，但其高评估成本阻碍了视场的稳定维护。改进的ColAG控制器实现了最低的延迟，但仍然存在安全性和可观察性违规。VIRGA在所有配对仓库条件下安全完成，而在没有重新训练的远程洞穴压力测试中展示了在不规则和狭窄几何中的持续协调。消融实验确认了在线几何评估、虚拟代理中介和互惠反馈的贡献。
View on arXiv Download PDF AI Translation
cs.RO / 90 / 2609.21908

CommitFlow: Semantic Commitment Verification and Local Correction for Long-Horizon Robot Manipulation VLA Execution

CommitFlow：长时间机器人操作VLA执行的语义承诺验证与局部修正
Zhao, Zixiang, Feng, Yansong, Yang, Yang, Wang, Chaoyu, Xiao, Haoran, Zhang, Hui, Cheng, Chuang, Ma, Jianjun
Abstract
Although vision-language-action (VLA) policies have advanced rapidly, long-horizon execution may still progress to the next task stage before the required physical effect has been established. We call this a mismatch between semantic commitments, physical conditions that a stage must establish or maintain, and the actual physical state. Because an action command alone cannot confirm such a condition, local deviations can propagate and cause task failure. To address this problem, we present CommitFlow, a closed-loop execution framework that combines commitment monitoring with local correction while keeping the base policy frozen. CommitFlow integrates three components. A Semantic Commitment Monitor (SCM) compares stage requirements against current state evidence and holds back dependent actions when a required condition is unmet or violated. BoundaryFlow then generates a local correction conditioned on the current state and base action, and Relation and Gain Calibration (RGC) selects the smallest correction strength that satisfies the relevant constraints. Across the ten common RoboTwin 2.0 benchmark tasks, CommitFlow achieves a mean success rate of 75.9 percent, improving on the base policy pi0.5 by 22.7 percent. Cross-policy experiments show consistent gains, pointing toward reliable long-horizon robot execution.
Chinese Translation
尽管视觉-语言-动作（VLA）策略发展迅速，但在长时间执行过程中，可能在所需的物理效果尚未建立之前就已推进到下一个任务阶段。我们称之为语义承诺之间的不匹配，即一个阶段必须建立或维持的物理条件与实际物理状态之间的差异。由于单一的动作指令无法确认此类条件，局部偏差可能会传播并导致任务失败。为了解决这一问题，我们提出了CommitFlow，一个闭环执行框架，结合了承诺监测与局部修正，同时保持基础策略不变。CommitFlow集成了三个组件。语义承诺监测器（Semantic Commitment Monitor, SCM）将阶段要求与当前状态证据进行比较，当所需条件未满足或被违反时，暂停相关动作。BoundaryFlow随后生成基于当前状态和基础动作的局部修正，而关系与增益校准（Relation and Gain Calibration, RGC）选择满足相关约束的最小修正强度。在十个常见的RoboTwin 2.0基准任务中，CommitFlow实现了75.9%的平均成功率，比基础策略pi0.5提高了22.7%。跨策略实验显示出一致的增益，指向可靠的长时间机器人执行。
View on arXiv Download PDF AI Translation
cs.RO / 91 / 2609.21929

MAAP: Multi-Agent Active Perception for Collaborative Manipulation

MAAP：用于协作操控的多智能体主动感知
Chen, Bruno N. Y., Kang, Li, Zhou, Heng, Song, Xiufeng, Zhang, Zhemeng, Ma, Jiahua, Qin, Yiran
Abstract
Multi-agent manipulation naturally produces multiple task-driven viewpoints: every arm carries a wrist camera and moves through the scene while acting. Yet these observations are typically underutilized, and active perception in manipulation is still often treated as requiring a dedicated sensing agent. We introduce MAAP (Multi-Agent Active Perception), in which every arm is dual-purpose: it executes manipulation actions and, through the wrist camera it carries, simultaneously serves as a moving viewpoint for the team. We pair this with RAIL (Role-Aware Imitation Learning), a controller that predicts each arm's current role alongside its action chunk and conditions action generation on it, representing role-dependent actions within one network. Across four simulated tasks, widening the perception regime lifts average success from 56.5% with a fixed camera to 62.5% with one active wrist view and 70.0% with all of them, while MAAP+RAIL reaches 79.2%. RAIL's additional gain is concentrated on the three-arm Microwave task, where success rises from 47% to 82% on identical multi-wrist inputs. On a dual-arm platform, MAAP+RAIL succeeds in 14 of 20 placement trials compared with 0 of 20 for fixed-view ACT. Collaborative manipulation can thus serve as an active perception mechanism in its own right.
Chinese Translation
多智能体操控自然产生多个任务驱动的视角：每个机械臂都携带一个腕部摄像头，并在执行操作时移动于场景中。然而，这些观察通常未得到充分利用，操控中的主动感知仍然常常被视为需要一个专用的感知代理。我们提出了MAAP（多智能体主动感知），在该模型中，每个机械臂具有双重功能：它执行操控动作，并通过其携带的腕部摄像头同时作为团队的移动视角。我们将其与RAIL（角色感知模仿学习）相结合，RAIL是一个控制器，它预测每个机械臂的当前角色及其动作片段，并以此为条件生成动作，代表在一个网络中依赖角色的动作。在四个模拟任务中，扩大感知范围使得平均成功率从固定摄像头的56.5%提升至一个主动腕部视角的62.5%，并在所有视角下达到70.0%，而MAAP+RAIL的成功率则达到了79.2%。RAIL的额外增益集中在三臂微波任务上，在相同的多腕输入下，成功率从47%上升至82%。在双臂平台上，MAAP+RAIL在20次放置试验中成功了14次，而固定视角的ACT则没有成功过一次。因此，协作操控可以作为一种独立的主动感知机制。
View on arXiv Download PDF AI Translation
cs.RO / 92 / 2609.21942

When Should a Failing Robot Ask? Initiating Corrective Human-Robot Dialogue from Audited Sensor Evidence

失败的机器人何时应请求帮助？基于审计传感器证据启动纠正性人机对话
Pathak, Eshika, Krishna, Leela
Abstract
A robot that fails at a task faces the first decision in corrective dialogue: act on its own diagnosis, consult another onboard sensor, or interrupt a person. Choosing well requires knowing how much the robot's sensors reveal about the cause and how reliable the robot's own diagnosis is. We build a simulated benchmark in which every failure's true cause is known, because we injected it, and measure what each sensor reveals, with explicit checks against data leakage. Some failures are diagnosable from camera images; others only from the robot's force data (0.99 from force data, no image method above 0.55). We then test six open vision-language models. Their behavior tracks the surface of the prompt, not the evidence: moving the refusal option from last to first in the answer list collapses refusal rates from 78-100% to 0-6% in three of the six swept model-and-family pairs. Accuracy from frames stays at or below a majority-class baseline under every prompt variant, with or without worked examples, and stated confidence carries no information about correctness. Handing the same models the force data as ten lines of text produces the first above-baseline diagnoses, in four of the six models: much of the failure reflects missing sensor data, not missing ability. We pose the choice as a three-action decision problem, act, consult your own sensors, or ask a human, whose optimal policy follows from measured accuracy. The models do not follow it, and their ask rates ignore a fourfold change in question cost. One question to a human still lifts them from that baseline to roughly the answerer's own reliability (0.70-0.81 when they ask). The decision to ask should be tied to measured accuracy and stated costs, not to the model's confidence.
Chinese Translation
一个在任务中失败的机器人面临纠正对话中的第一个决策：是依赖自己的诊断、咨询另一个机载传感器，还是打断人类。做出正确选择需要了解机器人的传感器对原因的揭示程度以及机器人自身诊断的可靠性。我们构建了一个模拟基准，其中每个失败的真实原因都是已知的，因为我们进行了注入，并测量每个传感器所揭示的信息，同时明确检查数据泄漏。一些失败可以通过摄像头图像进行诊断；而其他失败则仅能通过机器人的力数据进行诊断（力数据的诊断准确率为0.99，图像方法的准确率最高为0.55）。随后，我们测试了六个开放的视觉-语言模型。它们的行为跟踪的是提示的表面，而不是证据：将拒绝选项从答案列表的最后一项移动到第一项，使得六个模型及其家族中有三对的拒绝率从78-100%降至0-6%。在每个提示变体下，帧的准确率始终保持在或低于多数类基线，无论是否包含示例，而声明的置信度并未提供关于正确性的任何信息。将相同模型的力数据作为十行文本输入，产生了四个模型中首个超出基线的诊断：大部分失败反映的是缺失的传感器数据，而非缺失的能力。我们将选择视为一个三项决策问题：行动、咨询自己的传感器或询问人类，其最优策略源于测量的准确性。这些模型并未遵循该策略，其请求率忽视了问题成本的四倍变化。向人类提出一个问题仍然使它们的表现从基线提升至大约回答者自身的可靠性（当它们询问时，准确率为0.70-0.81）。请求帮助的决策应与测量的准确性和声明的成本相关，而非模型的置信度。
View on arXiv Download PDF AI Translation
cs.RO / 93 / 2609.21948

GALA: Geometry-Aware Latent Action Modeling for Vision-Language-Action Model Pretraining across Embodiments

GALA：面向几何感知的潜在动作建模框架，用于跨不同体现的视觉-语言-动作模型预训练
Liu, Yichen, Yuan, Puzhen, Zhu, Xiang, Guo, Yanjiang, Chen, Jianyu
Abstract
Learning large-scale vision-language-action (VLA) models from multi-embodiment datasets remains challenging due to heterogeneous action spaces across end effectors. Although latent action models (LAMs) can learn embodiment-agnostic action representations from diverse video data, existing image-based LAMs often fail to capture fine-grained end-effector articulation, particularly finger-level geometric changes in human and dexterous robot hands. To address this limitation, we propose GALA, a Geometry-Aware Latent-Action modeling framework that augments image-based latent actions with 3D end-effector geometric motion. However, naively incorporating point clouds yields fine-grained action representations with limited shared semantics, hindering cross-embodiment pretraining. To address this issue, we introduce the Unified End-effector Motion Representation (UEMR), which preserves fine-grained motion information while improving the cross-embodiment generalizability of latent actions. Building upon UEMR, GALA combines visual latent actions that capture scene-level dynamics with geometric latent actions that capture shared fine-grained end-effector articulation, providing effective supervision for VLA pretraining from multi-embodiment data, including action-free ego-centric human videos. Experiments on fine-grained motion probing, cross-embodiment retrieval, and downstream VLA evaluation demonstrate GALA's effectiveness in modeling generalizable fine-grained motions across embodiments, achieving 68.3% RoboCasa-GR1 success rate and 75.5% real-world success rate. Code, appendix, and demos are available at https://puzhenyuan.github.io/GALA-website/.
Chinese Translation
从多体现数据集中学习大规模视觉-语言-动作（VLA）模型仍然面临挑战，因为末端效应器之间的动作空间存在异质性。尽管潜在动作模型（LAMs）能够从多样化的视频数据中学习与体现无关的动作表示，但现有的基于图像的LAMs往往无法捕捉到细粒度的末端效应器关节动作，特别是在人的手和灵巧机器人手中，指尖级几何变化的捕捉能力不足。为了解决这一限制，我们提出了GALA，一个几何感知的潜在动作建模框架，该框架通过3D末端效应器几何运动增强基于图像的潜在动作。然而，简单地将点云纳入模型会导致细粒度动作表示的共享语义有限，从而阻碍跨体现的预训练。为了解决这一问题，我们引入了统一末端效应器运动表示（UEMR），该表示在保留细粒度运动信息的同时，提高了潜在动作的跨体现泛化能力。在UEMR的基础上，GALA结合了捕捉场景级动态的视觉潜在动作与捕捉共享细粒度末端效应器关节动作的几何潜在动作，为从多体现数据（包括无动作的自我中心人类视频）进行VLA预训练提供了有效的监督。关于细粒度运动探测、跨体现检索和下游VLA评估的实验表明，GALA在建模跨体现的可泛化细粒度动作方面的有效性，达到了68.3%的RoboCasa-GR1成功率和75.5%的现实世界成功率。代码、附录和演示可在 https://puzhenyuan.github.io/GALA-website/ 获取。
View on arXiv Download PDF AI Translation
cs.RO / 94 / 2609.21982

CARF: Contrastive Attraction-Repulsion of Failure-Guided Flow Matching

CARF：基于对比吸引-排斥的失败引导流匹配
Zhao, Shuqi, Du, Bang, Wu, Cheng-En, Xie, Yichen, Wang, Yixiao, Tomizuka, Masayoshi
Abstract
Robot demonstration collection often produces imperfect or failed trajectories in addition to successful demonstrations. Existing methods typically exploit failed trajectories by identifying segments that still make progress toward task completion, but largely overlook \textit{failure-critical behaviors} that directly lead to task failure. Here we argue that these two types of segments provide fundamentally asymmetric supervision: progressive segments should be imitated, whereas failure-critical segments should be explicitly avoided. Based on this observation, we propose CARF, a Contrastive Attraction-Repulsion of Failure-guided framework for learning from imperfect robot data. CARF introduces a progress-based importance scorer, trained solely on successful expert demonstrations and its perturbation results, to estimate step-wise contributions toward task completion and identify informative regions in failed trajectories. These scores guide a unified flow-matching objective that attracts the policy toward progressive behaviors and repels it from failure-critical ones, while excluding ambiguous segments. This enables more comprehensive utilization of imperfect data and avoids unreliable supervision from ambiguous failure segments. Extensive experiments in simulation and the real world demonstrate consistent improvements over competing baselines across diverse failure scenarios, with ablations further validating the effectiveness of the proposed scoring and attraction-repulsion mechanisms. Our website is https://zhao-sq.github.io/carf/#.
Chinese Translation
机器人演示收集通常会产生不完美或失败的轨迹，除了成功的演示之外。现有方法通常通过识别仍在朝着任务完成进展的段落来利用失败的轨迹，但在很大程度上忽视了直接导致任务失败的 extit{失败关键行为}。在这里，我们认为这两种类型的段落提供了根本上不对称的监督：进展段应被模仿，而失败关键段应被明确避免。基于这一观察，我们提出了CARF，一种基于对比吸引-排斥的失败引导框架，用于从不完美的机器人数据中学习。CARF引入了一种基于进展的重要性评分器，仅在成功的专家演示及其扰动结果上进行训练，以估计逐步贡献于任务完成的程度，并识别失败轨迹中的信息区域。这些评分指导一个统一的流匹配目标，吸引策略朝向进展行为，并排斥其向失败关键行为，同时排除模糊段落。这使得不完美数据的更全面利用成为可能，并避免了来自模糊失败段落的不可靠监督。在模拟和现实世界中的广泛实验表明，在各种失败场景下，相较于竞争基线，均表现出一致的改进，消融实验进一步验证了所提出的评分和吸引-排斥机制的有效性。我们的网站是 https://zhao-sq.github.io/carf/#。
View on arXiv Download PDF AI Translation
cs.RO / 95 / 2609.21983

SkelWAM: A Skeleton-Guided World-Action Model for Zero-Shot Cross-Embodiment Manipulation

SkelWAM：一种基于骨架引导的世界-动作模型用于零-shot跨体现操控
Niu, Pengjun, Xie, Yujia, Peng, Rui, Zhao, Hang, Liu, Ke
Abstract
Reusing manipulation experience across robot embodiments is important for scaling robot learning and reducing repeated task-specific data collection. However, changes in embodiment alter visual appearance, action dimensionality and semantics, and the whole-body configurations that can realize the same tool pose. We present SkelWAM, a skeleton-guided world-action model that couples perception and control through one explicit geometric representation for single-source cross-embodiment manipulation. Arm centerline geometry, tool-center-point (TCP) pose, and parallel-jaw commands form a shared 25-D state. The same definition underlies canonical third-person and wrist observations and future whole-body action targets. Trained with predictive visual supervision, a video-action mixture of transformers predicts canonical skeleton action chunks, which embodiment-specific constrained decoders convert into joint or continuum-robot controls. This formulation requires no one-to-one joint correspondence and uses no target-task demonstrations or target policy updates. We introduce LIBERO-Cross10, a source-only cross-embodiment transfer benchmark covering ten tasks and ten target embodiments across four morphological groups. On this benchmark, Franka-trained SkelWAM achieves 43.3% success over 1,000 episodes, exceeding the best-performing evaluated baseline by 36.2 percentage points. We further deploy a JAKA mini2-trained policy on the Feagine A03 continuum robot for three tabletop manipulation tasks, illustrating the approach's potential for real-world cross-embodiment manipulation. Project page: http://www.liukepku.com/skelwam/index.html
Chinese Translation
在机器人体现之间重用操控经验对于扩展机器人学习和减少重复的任务特定数据收集至关重要。然而，体现的变化会改变视觉外观、动作维度和语义，以及能够实现相同工具姿势的全身配置。我们提出了SkelWAM，这是一种通过一个明确的几何表示将感知与控制结合起来的骨架引导世界-动作模型，旨在实现单源跨体现操控。臂中心线几何、工具中心点（TCP）姿势和并行夹爪命令构成了一个共享的25维状态。相同的定义支撑着典型的第三人称和手腕观察以及未来的全身动作目标。通过预测视觉监督进行训练，视频-动作混合的变换器预测典型的骨架动作片段，而体现特定的约束解码器将其转换为关节或连续机器人控制。这种公式化不需要一对一的关节对应关系，也不使用目标任务演示或目标策略更新。我们引入了LIBERO-Cross10，这是一个仅源的跨体现转移基准，涵盖十个任务和十个目标体现，分为四个形态组。在这个基准上，经过Franka训练的SkelWAM在1,000个实验中实现了43.3%的成功率，超越了最佳评估基线36.2个百分点。我们进一步在Feagine A03连续机器人上部署了经过JAKA mini2训练的策略，完成了三个桌面操控任务，展示了该方法在现实世界跨体现操控中的潜力。项目页面：http://www.liukepku.com/skelwam/index.html
View on arXiv Download PDF AI Translation
cs.RO / 96 / 2609.22062

Gripper-Aware Automatic Dense Packing of Irregular Objects

考虑夹持器的非规则物体自动密集打包
Qin, Tianhao, McCann, Connor, Calli, Berk, Xiao, Jing
Abstract
Automatic dense packing is widely desired in warehouse operations but remains a fundamental challenge in robotic manipulation. Existing work on irregular-object packing largely targets simulation with idealized contact, treating the object as an isolated rigid body. The gripper often enters as a discrete, post-hoc feasibility check, if considered at all, and the perception and contact drift accumulated during execution are not addressed. We present a closed-loop pipeline that integrates perception, gripper-aware placement optimization, and force-guided execution on a real manipulator. The optimizer represents the object together with the gripper as a single composite body of hierarchical sphere trees. It searches over five degrees of freedom on a GPU within a CMA-ES framework, with the vertical coordinate grounded analytically against the current heightmap. During execution, a force-monitored vertical descent stops on first contact. A post-release consolidation push then closes the residual lateral clearance that gripper-aware planning leaves behind. The container is re-perceived between placements so that drift does not accumulate. We validate the system on a Franka Emika Panda robot packing a 3D-printed set of flat, curved, and concave objects, and a YCB object subset. An ablation study isolates the contribution of gripper-aware optimization, the consolidation push, and mesh-derived geometry to end-to-end success, achieved density, and computational cost. We further benchmark against the heightmap-minimization method as a baseline representative of prior irregular-object packing work.
Chinese Translation
自动密集打包在仓库操作中广受欢迎，但在机器人操作中仍然是一个基本挑战。现有的非规则物体打包研究主要针对理想接触的仿真，将物体视为孤立的刚体。如果考虑夹持器，它通常作为离散的事后可行性检查出现，并且在执行过程中积累的感知和接触漂移未得到解决。我们提出了一种闭环管道，集成了感知、考虑夹持器的放置优化和基于力的执行，应用于真实的操纵器。优化器将物体与夹持器一起表示为一个由层次球体树组成的复合体。它在CMA-ES框架内在GPU上搜索五个自由度，垂直坐标则根据当前高度图进行解析。执行过程中，监测到的力会在首次接触时停止垂直下降。随后进行的释放后整合推送则关闭了考虑夹持器的规划所留下的残余侧向间隙。在每次放置之间重新感知容器，以避免漂移的积累。我们在Franka Emika Panda机器人上验证了该系统，进行了一组3D打印的平面、曲面和凹面物体的打包，以及YCB物体子集的打包。消融研究隔离了考虑夹持器的优化、整合推送和网格衍生几何形状对端到端成功、实现的密度和计算成本的贡献。我们还将其与高度图最小化方法进行基准测试，作为先前非规则物体打包工作的基线代表。
View on arXiv Download PDF AI Translation
cs.RO / 97 / 2609.22073

Duty Factor Predicts Robust Constrained Quadrupedal Locomotion Across Gait Types

占空比预测跨步态类型的稳健约束四足运动
Zhu, James, Ologan, David, Ortiz, George, Lee, Thomas Chun Fai, Gonzalez, Selvin Garcia, Tajbakhsh, Ardalan, Ben-Tzvi, Pinhas, Johnson, Aaron M.
Abstract
Quadrupedal robots are increasingly deployed in environments where locomotion must remain robust to disturbances and constrained terrain. Gait type, such as walking or trotting, is commonly used to characterize quadrupedal locomotion. However, gait type does not uniquely define locomotion, as parameters such as duty factor, speed, and stance width can vary within a single gait type. In this work, we investigate the relationship between these gait parameters using three distinct quadrupedal locomotion control approaches. First, using whole body trajectory optimization with LQR feedback, we show that duty factor is a stronger predictor of local error convergence than nominal gait type. Second, we investigate duty factor selection with a learned locomotion controller, suggesting how duty factor may serve as a low-dimensional parameter for adapting locomotion robustness in narrow-terrain environments. Finally, we show that these trends persist under a centroidal model predictive control framework and validate them through narrow-terrain experiments on a physical quadruped. These results show that duty factor provides a simple and effective basis for understanding and selecting robust quadrupedal locomotion across gait types and control architectures.
Chinese Translation
四足机器人越来越多地应用于需要在干扰和受限地形中保持稳健运动的环境。步态类型，如行走或小跑，通常用于表征四足运动。然而，步态类型并不能唯一定义运动，因为占空比、速度和步幅等参数在单一步态类型中可能会有所变化。在本研究中，我们使用三种不同的四足运动控制方法来探讨这些步态参数之间的关系。首先，通过使用全身轨迹优化与线性二次调节（LQR）反馈，我们表明占空比是局部误差收敛的更强预测因子，而非名义步态类型。其次，我们研究了使用学习型运动控制器的占空比选择，建议占空比如何作为一个低维参数来适应狭窄地形环境中的运动稳健性。最后，我们展示了这些趋势在质心模型预测控制框架下的持续性，并通过在物理四足机器人上的狭窄地形实验进行了验证。这些结果表明，占空比为理解和选择跨步态类型及控制架构的稳健四足运动提供了一个简单而有效的基础。
View on arXiv Download PDF AI Translation
cs.RO / 98 / 2609.22075

LIMBO: Learning and Internalizing Model-Free Barrier Objectives for Agile and Safe Whole-Body Control

LIMBO：学习和内化无模型障碍目标以实现灵活安全的全身控制
Gonzales, Jake, Alvarez, Arturo Flores, Chen, Yu-Ming, Ames, Aaron D., Ratliff, Lillian J., Nambi, Manikantan
Abstract
Safe whole-body control requires coordinating collision avoidance and balance under high-dimensional, nonlinear dynamics--making safety certificates difficult to design and reuse across behaviors. We present LIMBO, a framework for synthesizing a state-action control barrier function and distilling its safety structure into a task policy. LIMBO learns the safety certificate from black-box transitions and a state-based failure specification over residual actions around a frozen base controller, making Q-CBF synthesis tractable in the full control dimension while placing the certificate in the task policy's control space. During synthesis, the learned safety value drives risk-guided sampling near the estimated boundary of recoverability; during task learning, it serves as a teacher that provides action-level safety feedback, yielding a robust task policy and alleviating the need for an online safety filter at deployment. We demonstrate LIMBO on a 29-degree-of-freedom humanoid performing dodgeball avoidance and locomotion beneath low obstacles. Beyond scaling learned Q-CBFs to whole-body control, we show that risk-guided boundary sampling provides a theoretically grounded way to explore the edge of recoverability. Under the same safety specification, ceteris paribus, varying the sampling concentration produces strategies ranging from crouching to a novel backward-leaning limbo maneuver. In both settings, the learned policies transfer to hardware without online safety filtering, showing that learned safety synthesis scales to agile whole-body control.
Chinese Translation
安全的全身控制需要在高维非线性动态下协调避碰和保持平衡，这使得安全证书的设计和在行为间的重用变得困难。我们提出了LIMBO，一个合成状态-动作控制障碍函数并将其安全结构提炼为任务策略的框架。LIMBO从黑箱转移和基于状态的残余动作失败规范中学习安全证书，围绕一个冻结的基础控制器，使得Q-CBF合成在完整控制维度内变得可行，同时将证书置于任务策略的控制空间中。在合成过程中，学习到的安全值驱动风险引导的采样，接近可恢复性的估计边界；在任务学习期间，它作为教师提供动作级别的安全反馈，从而产生稳健的任务策略，并减轻在部署时对在线安全过滤器的需求。我们在一个具有29个自由度的人形机器人上展示了LIMBO，该机器人执行躲避球和在低障碍物下的运动。除了将学习到的Q-CBF扩展到全身控制外，我们还展示了风险引导的边界采样提供了一种理论上合理的方法来探索可恢复性的边缘。在相同的安全规范下，其他条件不变，变化采样浓度会产生从蹲下到新颖的向后倾斜的躲避动作等不同策略。在这两种设置中，学习到的策略在没有在线安全过滤的情况下转移到硬件上，表明学习的安全合成能够扩展到灵活的全身控制。
View on arXiv Download PDF AI Translation
cs.RO / 99 / 2609.22085

SeeQ: Training Generalist Value Functions for Long-Horizon Robotic Manipulation

SeeQ：用于长时间范围机器人操作的通用价值函数训练
Singh, Saksham, Hu, Zheyuan, Mark, Max Sobol, Yu, Jeffrey, Erickson, Zackory, Kumar, Aviral
Abstract
Despite rapid progress, generalist robot policies remain brittle on complex, long-horizon tasks that comprise multiple stages or require repeated attempts and deliberation on the same underlying stage before success. Q-value functions can improve these policies by ranking candidate actions or guiding policy improvement, but learning from sparse task-level rewards entails long credit-assignment horizons, difficult Bellman backups, and broad data-coverage requirements. We introduce SeeQ (Subtask-elicited Q-functions), which instead learns Q-values for the currently active subtask. This shortens the value-prediction horizon and enables effective learning with temporal-difference (TD) objectives. During training, subtask-level annotations present in offline robot data provide the decomposition and enable learning from broad, potentially suboptimal robot datasets. To eliminate the need for human annotations or modular subtask prediction systems at test time, our Q-function architecture is trained to autoregressively predict the active subtask in natural language before estimating its value. We instantiate SeeQ using a base vision-language backbone, pretrain it on diverse open-source robot manipulation data, and finetune it on downstream tasks. Across four real-world manipulation tasks on two bimanual robot platforms, the SeeQ value function substantially improves best-of-N policy steering.
Chinese Translation
尽管取得了快速进展，但通用机器人策略在复杂的长时间范围任务上仍然脆弱，这些任务包含多个阶段或在成功之前需要对相同基础阶段进行多次尝试和深思。Q值函数可以通过对候选动作进行排名或指导策略改进来改善这些策略，但从稀疏的任务级奖励中学习涉及长时间的信用分配范围、困难的贝尔曼备份和广泛的数据覆盖要求。我们提出了SeeQ（子任务引发的Q函数），它学习当前活动子任务的Q值。这缩短了价值预测的范围，并使得使用时间差（TD）目标进行有效学习成为可能。在训练过程中，离线机器人数据中存在的子任务级注释提供了分解，并使得从广泛的、潜在的次优机器人数据集中进行学习成为可能。为了消除在测试时对人工注释或模块化子任务预测系统的需求，我们的Q函数架构被训练为自回归地预测活动子任务的自然语言，然后再估计其价值。我们使用基础的视觉-语言骨干网络实例化SeeQ，在多样的开源机器人操作数据上进行预训练，并在下游任务上进行微调。在两个双手机器人平台上的四个真实世界操作任务中，SeeQ价值函数显著改善了最佳N策略的引导效果。
View on arXiv Download PDF AI Translation
计算机视觉 (Computer Vision)
74
cs.CV / 1 / 2609.20869

TAPe+ML: A Compact Structured Representation for Multi-Task Computer Vision

TAPe+ML：一种紧凑的多任务计算机视觉结构化表示
Kurinov, Sergey, Upatov, Alexey
Abstract
We present TAPe+ML v3, a compact computer vision system based on TAPe (Theory of Active Perception), a structured representation that encodes relations among perceptual elements before recognition. Instead of operating directly on pixel tensors, the system uses a shared TAPe representation and a modular recognition architecture for image classification, object detection, and instance segmentation. TAPe+ML v3 combines background and contour processing, local object localization, prototype-based classification, and a coordinator for specialized submodels. Across the reported experiments, it uses fewer than 100,000 parameters. On COCO object detection, it obtains 84.7 mAP50 and 65.3 mAP50-95. On COCO instance segmentation, it obtains 80.7 mask mAP50 and 58.4 mask mAP50-95. In classification experiments, it reaches 92 percent validation accuracy on Imagenette under an identical-training comparison with a raw-pixel baseline, and 89.9 percent Top-1 accuracy on ImageNet-Real. We also evaluate compactness in video scene detection and adaptation under distribution shift in an industrial pilot. The results suggest that shifting part of the modeling burden from network parameters to a structured input representation can support compact multi-task vision systems with reduced data, memory, and compute requirements.
Chinese Translation
我们提出了 TAPe+ML v3，这是一种基于 TAPe（主动感知理论）的紧凑计算机视觉系统，TAPe 是一种在识别之前编码感知元素之间关系的结构化表示。该系统并不直接对像素张量进行操作，而是使用共享的 TAPe 表示和模块化的识别架构来进行图像分类、目标检测和实例分割。TAPe+ML v3 结合了背景和轮廓处理、局部物体定位、基于原型的分类以及用于专门子模型的协调器。在所报告的实验中，它使用的参数少于 100,000 个。在 COCO 目标检测中，获得了 84.7 的 mAP50 和 65.3 的 mAP50-95。在 COCO 实例分割中，获得了 80.7 的 mask mAP50 和 58.4 的 mask mAP50-95。在分类实验中，在与原始像素基线的相同训练比较下，Imagenette 的验证准确率达到了 92%，而在 ImageNet-Real 上达到了 89.9% 的 Top-1 准确率。我们还在工业试点中评估了视频场景检测的紧凑性以及在分布变化下的适应性。结果表明，将部分建模负担从网络参数转移到结构化输入表示上，可以支持具有较低数据、内存和计算需求的紧凑多任务视觉系统。
View on arXiv Download PDF AI Translation
cs.CV / 2 / 2609.20962

MemeTAG: Keyword-Driven Meme Classification through Tag Embedding Reconstruction

MemeTAG：通过标签嵌入重构的关键词驱动的表情包分类
Sharma, Akshit, Patil, Prashant W.
Abstract
The proliferation of harmful internet memes poses a significant societal threat, yet their automated classification remains a formidable algorithmic challenge due to the nuanced, multimodal nature of their content. To address this, we introduce MemeTAG, a novel dual-objective framework that pioneers a keyword-aware approach to meme classification. Our core innovation is a two-part semantic guidance mechanism: first, we leverage a pretrained Vision-Language Model to generate a set of descriptive keywords, that capture the high-level semantics. Second, we introduce the Aggregated Tag Inference Network (ATIN), an attention-based module that distills these keywords into a single, rich semantic embedding. This embedding serves as a target for a novel auxiliary reconstruction loss, which compels the model to learn deeply aligned visual and textual features. This approach, combined with an efficient three-stage training strategy, establishes a new state-of-the-art on the HarMeme, Hateful Memes Challenge (HMC), and PrideMM datasets, decisively outperforming existing state-of-the-art methods.
Chinese Translation
有害互联网表情包的传播对社会构成了重大威胁，但由于其内容的细微差别和多模态特性，自动分类仍然是一个艰巨的算法挑战。为了解决这一问题，我们提出了MemeTAG，这是一种新颖的双目标框架，开创了一种关注关键词的表情包分类方法。我们的核心创新是一个由两部分组成的语义引导机制：首先，我们利用预训练的视觉-语言模型生成一组描述性关键词，以捕捉高层次的语义。其次，我们引入了聚合标签推断网络（Aggregated Tag Inference Network, ATIN），这是一个基于注意力的模块，将这些关键词提炼为一个丰富的语义嵌入。该嵌入作为一种新的辅助重构损失的目标，促使模型深入学习对齐的视觉和文本特征。这种方法结合高效的三阶段训练策略，在HarMeme、仇恨表情包挑战（Hateful Memes Challenge, HMC）和PrideMM数据集上建立了新的最先进水平，显著超越了现有的最先进方法。
View on arXiv Download PDF AI Translation
cs.CV / 3 / 2609.20975

Image-Derived PM10 Estimation in Cattle Feedlot Using Machine Learning: Addressing Concentration Ranges Beyond Existing Digital Imaging Methods

基于图像的牛饲养场PM10估计方法：应对超出现有数字成像方法的浓度范围
Peanusaha, Sirapoom, Ferguson, Greg B., Bush, K. Jack, Li, Peiyang, Auvermann, Brent W.
Abstract
Affordable dust monitoring remains a pressing need for the cattle feedlot industry, yet camera-based PM estimation, despite its growing body of research in urban air quality settings, has not been evaluated under the extended concentration ranges characteristic of intensive livestock operations. This study developed an image-based approach using contrast panel features and machine learning to estimate PM10 concentrations in a commercial cattle feedlot, where hourly average PM10 ranged from 250 to 1,000 ug/m^-3 and instantaneous concentrations reached 5,000 to 20,000 ug/m^-3. Grayscale images were captured during the evening dust peak period, and features including panel contrast, black and white panel pixel values, and overall image brightness were extracted. The model also incorporated recent past values from preceding images and solar zenith angle as predictors. Among the candidate models evaluated, XGBoost achieved the highest predictive performance, with an R^2 of 0.792 and a median absolute error of 103 ug/m^-3. Feature importance analysis revealed that (a) panels positioned farthest from the camera contributed most strongly to predictions and (b) that black panel pixel values were more sensitive than white panel values to changes in PM10 concentration. Prediction accuracy during the sunset transition, which coincides with the onset of the feedlot evening dust peak, remains an area for further refinement. These findings demonstrate the feasibility of image-based PM10 estimation across PM concentration ranges substantially exceeding those reported in prior urban studies and provide practical guidelines for future deployment in feedlot environments.
Chinese Translation
经济实惠的粉尘监测仍然是牛饲养场行业亟待解决的问题，尽管基于相机的PM估计在城市空气质量研究中逐渐增多，但在集约化畜牧业特有的扩展浓度范围内尚未得到评估。本研究开发了一种基于图像的方法，利用对比面板特征和机器学习来估计商业牛饲养场的PM10浓度，在该场所，小时平均PM10浓度范围为250至1000微克每立方米（ug/m^-3），瞬时浓度达到5000至20000微克每立方米（ug/m^-3）。在傍晚粉尘高峰期捕获了灰度图像，并提取了包括面板对比度、黑白面板像素值和整体图像亮度等特征。该模型还将前一图像的近期值和太阳天顶角作为预测因子纳入其中。在评估的候选模型中，XGBoost模型实现了最高的预测性能，决定系数（R^2）为0.792，中位绝对误差为103微克每立方米（ug/m^-3）。特征重要性分析表明：(a) 离相机最远的面板对预测的贡献最大，(b) 黑色面板像素值对PM10浓度变化的敏感性高于白色面板值。在与饲养场傍晚粉尘高峰开始重叠的日落过渡期间，预测准确性仍需进一步完善。这些发现表明，基于图像的PM10估计在PM浓度范围上具有可行性，远超以往城市研究报告的浓度范围，并为未来在饲养场环境中的应用提供了实用指南。
View on arXiv Download PDF AI Translation
cs.CV / 4 / 2609.21012

Fragment-Aware Vision Transformers for Fresco-Fragment Style Classification

考虑碎片的视觉变换器用于壁画碎片风格分类
Miketek, Sara, Barchielli, Biagio, Kajla, Nadeem Iqbal, Aslan, Sinem
Abstract
Artistic style classification is usually studied on complete artworks, where models can exploit global composition, spatial organisation, and iconographic structure. In archaeological settings, however, artworks often survive only as fragmented remains, forcing recognition from incomplete, irregular, and context-limited visual evidence. We study fresco-fragment style classification using a progressive transformer-based framework. Starting from a ViT-B/16 baseline, we introduce foreground-guided masking to suppress background-only tokens, inpainting-based geometric regularisation to align irregular fragment supports with the ViT patch grid, and a supervised contrastive objective that operates on predictive distributions through a Kullback-Leibler similarity and consistently improves every branch. We combine the branches with a deliberately simple learnable logit ensemble. Experiments on CLEOPATRA and POMPAAF show that fragment-aware modelling improves over the standard ViT baseline, with the ensemble increasing accuracy from 0.604 to 0.656 and macro-F1 from 0.596 to 0.648 on CLEOPATRA, and outperforming the best single branch in four of six fragmentation settings on POMPAAF. We additionally evaluate a more complex graph-fusion variant and find that it matches the simple ensemble on POMPAAF while offering only a small, dataset-specific gain on CLEOPATRA, which does not justify its added complexity. Beyond these empirical gains, our contribution is twofold: a distribution-level contrastive objective that consistently sharpens single-branch recognition, and an interpretability analysis that verifies the models exploit genuine painted evidence, while quantifying that the inpainting-based branch draws part of its attribution from the synthesised surround.
Chinese Translation
艺术风格分类通常在完整艺术作品上进行研究，模型可以利用全球构图、空间组织和图像结构。然而，在考古环境中，艺术作品往往仅以碎片形式存留，迫使我们从不完整、不规则和上下文有限的视觉证据中进行识别。我们使用基于渐进式变换器的框架研究壁画碎片风格分类。从ViT-B/16基线开始，我们引入前景引导掩蔽以抑制仅背景的标记，基于修复的几何正则化以将不规则碎片支撑与ViT补丁网格对齐，以及一个通过Kullback-Leibler相似度在预测分布上操作的监督对比目标，持续改善每个分支。我们将这些分支与一个故意简单的可学习对数集成结合起来。在CLEOPATRA和POMPAAF上的实验表明，考虑碎片的建模优于标准ViT基线，集成将准确率从0.604提高到0.656，宏观F1从0.596提高到0.648在CLEOPATRA上，并在POMPAAF的六个碎片化设置中有四个超越了最佳单一分支。我们还评估了一种更复杂的图融合变体，发现它在POMPAAF上与简单集成相匹配，同时在CLEOPATRA上仅提供小幅、特定于数据集的增益，这并不足以证明其增加的复杂性。除了这些经验上的提升，我们的贡献有两个方面：一个在分布层面的对比目标，持续提升单分支识别的准确性，以及一个可解释性分析，验证模型利用真实的绘画证据，同时量化基于修复的分支部分归因于合成的背景。
View on arXiv Download PDF AI Translation
cs.CV / 5 / 2609.21018

MAGIC: Marginal-Guided Compression with Optimal Transport for Efficient Visual Document Retrieval

MAGIC：基于最优传输的边际引导压缩用于高效视觉文档检索
Yuan, Xu, Liu, Hua, Fan, Wenqi, Li, Qing
Abstract
Recent visual document retrieval (VDR) systems such as ColPali use multi-vector page embeddings, in which patch-level vectors enable fine-grained evidence matching but incur substantial index storage and MaxSim scoring overhead. Post-hoc merging offers a practical route to efficient VDR by reducing this cost without retraining the retriever, but its uniform reconstruction objectives are poorly aligned with the sparse, non-uniform patch usage induced by late-interaction retrieval. Under aggressive compression, this misalignment can preserve rarely used patches while concentrating retrieval activity on too few retained representatives. To address this misalignment, we propose Marginal-Guided Compression with Optimal Transport (MAGIC), a training-free post-hoc compressor for efficient retrieval with frozen multi-vector embeddings. MAGIC derives a MaxSim-induced compression surrogate and optimizes it through a two-marginal entropic optimal-transport formulation, where a retrieval-demand source marginal prioritizes high-use patches and a balanced target marginal regularizes retained-facet usage. Across ViDoRe benchmarks, keep ratios, and retrieval backbones, MAGIC consistently outperforms strong post-hoc compressors, with particularly large gains in the aggressive-compression regime; component ablations verify the complementary effects of its two marginals. We release the code at: https://github.com/xandery-geek/MAGIC.
Chinese Translation
最近的视觉文档检索（VDR）系统，如ColPali，使用多向量页面嵌入，其中补丁级向量能够实现细粒度的证据匹配，但会导致显著的索引存储和MaxSim评分开销。后处理合并提供了一种通过减少这些成本而不重新训练检索器的高效VDR的实用途径，但其统一重建目标与由后交互检索引发的稀疏、不均匀的补丁使用不匹配。在激进压缩下，这种不匹配可能会保留很少使用的补丁，同时将检索活动集中在过少的保留代表上。为了解决这种不匹配，我们提出了基于最优传输的边际引导压缩（MAGIC），这是一种用于高效检索的无训练后处理压缩器，适用于冻结的多向量嵌入。MAGIC推导出一个MaxSim诱导的压缩替代物，并通过双边际熵最优传输公式进行优化，其中检索需求源边际优先考虑高使用率补丁，而均衡目标边际则对保留的特征使用进行正则化。在ViDoRe基准测试、保持比例和检索骨干网络中，MAGIC始终优于强大的后处理压缩器，尤其在激进压缩模式下获得了显著的提升；组件消融验证了其两个边际的互补效应。我们在以下网址发布了代码：https://github.com/xandery-geek/MAGIC。
View on arXiv Download PDF AI Translation
cs.CV / 6 / 2609.21095

MarsFM: Shading-Regularized Flow Matching for Martian Relief Estimation

MarsFM：用于火星地形估计的阴影正则化流匹配
Juston, Marius F. R.
Abstract
We present MarsFM, an image-conditioned latent flow-matching model for local Martian relief estimation from single-band HiRISE RED orthoimagery. The method combines a pretrained generative prior with stereo-derived geometric supervision and a differentiable Lunar--Lambert shading objective. Relief, normal, gradient, curvature, and ordinal terms constrain complementary aspects of terrain structure, while a positive-affine-invariant image comparison constrains rendered appearance. An evaluation comprising 2024 gathered patch records per integration-step count yields mean affine-aligned RMSE between 0.0935 and 0.0957 in normalized signed-log relief space for one to twenty Euler steps. These scores measure agreement with VAE-reconstructed references on positive-reference support. Their narrow range supports low-step inference under this protocol. Spatial, differential, and spectral diagnostics show broad terrain correspondence alongside smoothing, amplitude compression, and boundary mismatch. MarsFM provides a framework for combining learned terrain priors with image-based constraints; establishing improved physical terrain resolution requires matched baselines and independent high-resolution reference data. Data: https://huggingface.co/datasets/SuperComputer/mars_hirise_dtm_processed-6aa9b66ba461e07f; code: https://github.com/Marius-Juston/MarsRecon.
Chinese Translation
我们提出了MarsFM，这是一种基于图像条件的潜在流匹配模型，用于从单波段HiRISE RED正射影像中进行局部火星地形估计。该方法结合了预训练的生成先验、立体派生的几何监督和可微分的Lunar--Lambert阴影目标。地形、法线、梯度、曲率和序数项约束了地形结构的互补方面，而正仿射不变的图像比较则约束了渲染外观。对2024个收集的补丁记录进行的评估显示，在归一化的有符号对数地形空间中，经过一到二十个欧拉步骤的均值仿射对齐均方根误差（RMSE）介于0.0935和0.0957之间。这些分数衡量了与正参考支持上的VAE重建参考的一致性。它们的狭窄范围支持在该协议下的低步推断。空间、微分和光谱诊断显示出广泛的地形对应关系，同时伴随平滑、幅度压缩和边界不匹配。MarsFM提供了一个将学习到的地形先验与基于图像的约束相结合的框架；要建立改进的物理地形分辨率，需要匹配的基线和独立的高分辨率参考数据。数据来源：https://huggingface.co/datasets/SuperComputer/mars_hirise_dtm_processed-6aa9b66ba461e07f；代码：https://github.com/Marius-Juston/MarsRecon。
View on arXiv Download PDF AI Translation
cs.CV / 7 / 2609.21176

4DGS-Fixer: Generative Sparse-View 4D Gaussian Splatting with Iterative Refinement Guided by Video Diffusion Priors

4DGS-Fixer：基于视频扩散先验的生成稀疏视图4D高斯点云迭代优化
Huang, Haitao, Zhao, Shenghao, Tian, Boyuan, Chng, Shin-Fang, Yang, Songlin, Lim, Sheila, Zhan, Huangying, Xu, Yi, Rao, Anyi, Guan, Frank
Abstract
This paper addresses the challenges of dynamic scene synthesis from sparse-view videos. Existing methods employ geometric priors, adaptive optimization, or density-control strategies to improve 4D Gaussian modeling under sparse observations. However, they cannot fundamentally resolve the ill-posed problem caused by insufficient observations and missing scene information. Moreover, sparse-view 4D Gaussian Splatting (4DGS) often suffers from poor geometric initialization: with only a few input views, COLMAP typically reconstructs sparse and incomplete point clouds, leaving large scene regions without sufficient Gaussian support and making them difficult to recover through subsequent optimization. To address these limitations, we propose a novel iterative refinement framework based on a video diffusion model to improve the completeness and consistency of dynamic 4D scenes. Specifically, we first estimate multi-view depth maps and fuse them into dense point clouds to provide more complete geometric initialization for a dynamic 4DGS representation. We then employ a pretrained video restoration model to refine sequences rendered along novel camera trajectories at different time steps. The restored sequences serve as pseudo-supervision to regularize and iteratively refine the 4DGS representation. Experiments on a widely used benchmark dataset demonstrate that our method substantially outperforms existing baselines, achieving nearly a 2 dB PSNR improvement over the previous best-performing method.
Chinese Translation
本文解决了从稀疏视图视频合成动态场景的挑战。现有方法采用几何先验、自适应优化或密度控制策略，以改善在稀疏观测下的4D高斯建模。然而，它们无法从根本上解决由于观测不足和场景信息缺失所导致的病态问题。此外，稀疏视图4D高斯点云（4DGS）通常面临几何初始化不佳的问题：仅凭少量输入视图，COLMAP通常重建出稀疏且不完整的点云，导致大范围场景区域缺乏足够的高斯支持，进而使后续优化变得困难。为了解决这些局限性，我们提出了一种基于视频扩散模型的新型迭代优化框架，以提高动态4D场景的完整性和一致性。具体而言，我们首先估计多视角深度图，并将其融合为密集点云，以为动态4DGS表示提供更完整的几何初始化。然后，我们利用预训练的视频修复模型，针对不同时间步长沿新相机轨迹渲染的序列进行优化。修复后的序列作为伪监督，来规范化和迭代优化4DGS表示。在一个广泛使用的基准数据集上的实验表明，我们的方法显著优于现有基线，较之前最佳方法实现了近2 dB的PSNR提升。
View on arXiv Download PDF AI Translation
cs.CV / 8 / 2609.21199

OnomatoBridge: Onomatopoeia Translation and Rendering Pipeline in Manga

OnomatoBridge：漫画中的拟声词翻译与呈现管道
Taniguchi, Takara, Shimoda, Wataru, Yamaguchi, Kota, Nakayama, Hideki
Abstract
Manga is a comic drawn by black and white paints gaining popularity around the world. Onomatopoeia in Manga specifically appeals to the audience with its unique visual styles, which convey sound, motion, and emotion. Visual onomatopoeia translation requires the clean replacement of Japanese onomatopoeia with onomatopoeia in the other language while preserving their visual style. Existing approaches often produce residual artifacts or style inconsistency when removing the Japanese onomatopoeia and rendering stylized English onomatopoeia. To approach these problems, we present OnomatoBridge, a filtering pipeline for visual onomatopoeia translation. We evaluate OnomatoBridge from Japanese to English on the Manga109 onomatopoeia dataset and compare it with baseline image editing models. Experimental results show that the filtered outputs by the proposed method outperform those of conventional methods. OnomatoBridge improves English text correctness by roughly 10 to 25 points and reduces residual Japanese text by about 20 to 50% in relative terms.
Chinese Translation
漫画是一种以黑白画风绘制的漫画形式，正在全球范围内获得越来越多的关注。漫画中的拟声词以其独特的视觉风格吸引观众，传达声音、动作和情感。视觉拟声词翻译需要在保持其视觉风格的同时，将日语拟声词干净地替换为其他语言的拟声词。现有的方法在去除日语拟声词并渲染风格化的英语拟声词时，往往会产生残余伪影或风格不一致的问题。为了解决这些问题，我们提出了OnomatoBridge，一个用于视觉拟声词翻译的过滤管道。我们在Manga109拟声词数据集上评估OnomatoBridge从日语到英语的翻译效果，并与基线图像编辑模型进行了比较。实验结果表明，所提出的方法的过滤输出优于传统方法。OnomatoBridge在英语文本的正确性上提高了大约10到25个百分点，并在相对意义上减少了约20到50%的残余日语文本。
View on arXiv Download PDF AI Translation
cs.CV / 9 / 2609.21207

Hand-Aware Transition Modeling for Bimanual Procedural Anomaly Detection

基于手部意识的双手程序异常检测过渡建模
Wen, Di, Weissert, Jimmy, Scherrer, Luc Maria, Zöllner, Cedric, Yang, Kailun, Liu, Ruiping, Chen, Yufan, Wei, Jiale, Zheng, Junwei, Peng, Kunyu
Abstract
Procedural anomaly detection in bimanual assembly requires judging each hand action against the execution so far. A corrective action may look unusual in isolation, while a visually plausible action can violate the order of the procedure. We present HACT, a transition model over predicted per-hand events. A role-preserving history keeps the concurrent responsibilities of both hands, and a marked temporal point process assigns each observed transition a semantic and temporal surprisal. A supervised evidence head and a two-state filter convert these surprisals into per-hand anomaly posteriors. A recovery-aware protocol on predicted events and participant-disjoint folds reports the recovery false-positive rate at an operating point selected on validation participants. On two bimanual power-tool procedures HACT has the highest AUPRC and F1 among the compared methods and the fewest recovery alarms. Applied without retraining to a different assembly order of the same product, it retains the highest AUPRC and F1. The source code is available at https://github.com/Kratos-Wen/HACT.
Chinese Translation
双手组装中的程序异常检测需要根据迄今为止的执行情况判断每个手的动作。孤立的纠正动作可能看起来不寻常，而视觉上合理的动作可能会违反程序的顺序。我们提出了 HACT（手部意识过渡模型），这是一个基于每只手的预测事件的过渡模型。一个保持角色的历史记录保留了双手的并发责任，而一个标记的时间点过程为每个观察到的过渡分配了语义和时间上的惊讶度。一个监督证据头和一个双状态过滤器将这些惊讶度转换为每只手的异常后验概率。在预测事件和参与者不重叠的折叠上，恢复意识协议报告了在验证参与者上选择的操作点的恢复假阳性率。在两个双手电动工具程序中，HACT 在比较方法中具有最高的 AUPRC 和 F1 值，并且恢复警报最少。在不进行再训练的情况下应用于同一产品的不同组装顺序时，它仍然保持最高的 AUPRC 和 F1。源代码可在 https://github.com/Kratos-Wen/HACT 获取。
View on arXiv Download PDF AI Translation
cs.CV / 10 / 2609.21219

Multi-viewpoint Geo-localization with Event Cameras

多视角地理定位与事件相机
Hines, Adam D., Milford, Michael, Fischer, Tobias
Abstract
Robot localization is an ongoing challenge that demands mapping and positioning systems that are tolerant to viewpoint change. Event cameras are attracting increasing interest and adoption in robotics; however, dealing with viewpoint variance is an under-investigated problem in existing event-based localizers. In addition, event-based datasets that emphasize viewpoint variance for challenging localization situations are scarce. Here, we introduce an event-based visual place recognition (VPR) system that performs robustly under viewpoint changes. We converted five large-scale geo-tagged datasets, conventionally used to train frame-based localization systems, into synthetic event streams using Image-to-Event (I2E) conversion, and used them to fine-tune a pre-trained event-based vision transformer backbone with a multi-loss function, yielding a system we call MegaEvent that learns viewpoint-robust features for place recognition. We achieved an average Recall@1 of 82% across three existing event-based localization datasets, leading the next best event-based method by 20 recall points, and frame-based VPR models applied directly to event frames by 8 to 26 recall points. We introduce a new, challenging dataset - Springfield-Event-VPR - which features a 3.7km walking route recorded in three camera orientations for a total of 11.1km, which MegaEvent outperforms the strongest baseline by 9 recall points. The code for MegaEvent is available at https://github.com/AdamDHines/megaevent.
Chinese Translation
机器人定位是一个持续的挑战，需要能够容忍视角变化的映射和定位系统。事件相机在机器人领域引起了越来越多的关注和应用；然而，现有基于事件的定位器在处理视角变化方面仍然是一个未被充分研究的问题。此外，强调视角变化以应对复杂定位情况的基于事件的数据集也相对稀缺。在此，我们介绍了一种基于事件的视觉地点识别（Visual Place Recognition, VPR）系统，该系统在视角变化下表现出色。我们将五个大型地理标记数据集（通常用于训练基于帧的定位系统）转换为合成事件流，采用图像到事件（Image-to-Event, I2E）转换，并利用这些数据集对预训练的基于事件的视觉变换器骨干网络进行微调，采用多损失函数，最终形成了我们称之为MegaEvent的系统，该系统学习了对视角变化具有鲁棒性的地点识别特征。在三个现有的基于事件的定位数据集中，我们实现了82%的平均Recall@1，领先于下一个最佳的基于事件的方法20个召回点，以及直接应用于事件帧的基于帧的VPR模型8至26个召回点。我们还介绍了一个新的挑战性数据集——Springfield-Event-VPR，该数据集包含一条3.7公里的步行路线，以三种相机朝向记录，总计11.1公里，MegaEvent在此数据集上超越了最强基线9个召回点。MegaEvent的代码可在https://github.com/AdamDHines/megaevent获取。
View on arXiv Download PDF AI Translation
cs.CV / 11 / 2609.21225

VGGT-CAD: Reconstructing Parametric CAD 3D Model with Geometric Grounding

VGGT-CAD：基于几何基础的参数化CAD 3D模型重建
Yu, Chunan, Chen, Tianrun, Shen, Fu, Chen, Cheng, Zhu, Lanyun, Yang, Yang
Abstract
Parametric CAD reconstruction requires recovering both precise geometry and editable modeling operations from visual observations, making it challenging under limited and ambiguous views. Existing methods mainly rely on 2D appearance cues and lack strong multi-view geometric priors. In this work, we present VGGT-CAD, a geometry-aware framework for parametric CAD reconstruction from single- and multi-view observations. We transfer pretrained 3D geometric priors into CAD reconstruction by encoding camera parameters as condition tokens and jointly modeling them with image tokens. To handle varying numbers of viewpoints, we introduce a variable-view cross-view context aggregation module that adaptively fuses multi-view features. We further develop a training-free geometry-aware view selection strategy to select complementary and reliable frames during inference. The resulting representation is decoded into CAD command sequences using a non-autoregressive decoder. We also develop VideoCAD, a large-scale multi-view video benchmark derived from existing CAD data through multi-view re-rendering. Extensive experiments demonstrate the effectiveness of VGGT-CAD for visual CAD reconstruction under different observation configurations.
Chinese Translation
参数化CAD重建需要从视觉观察中恢复精确的几何形状和可编辑的建模操作，这在视角有限和模糊的情况下具有挑战性。现有方法主要依赖于二维外观线索，缺乏强大的多视角几何先验。在本研究中，我们提出了VGGT-CAD，一个基于几何感知的框架，用于从单视角和多视角观察中进行参数化CAD重建。我们通过将相机参数编码为条件标记，并与图像标记共同建模，将预训练的三维几何先验转移到CAD重建中。为了处理不同数量的视点，我们引入了一个可变视角交叉视图上下文聚合模块，能够自适应地融合多视角特征。我们进一步开发了一种无训练的几何感知视图选择策略，以在推理过程中选择互补和可靠的帧。最终的表示通过非自回归解码器解码为CAD命令序列。我们还开发了VideoCAD，这是一个通过多视角重新渲染现有CAD数据而衍生的大规模多视角视频基准。大量实验表明，VGGT-CAD在不同观察配置下的视觉CAD重建中具有良好的效果。
View on arXiv Download PDF AI Translation
cs.CV / 12 / 2609.21241

Multiclass Semantic Segmentation of Wildland Fire Images Using Context-Aware Centralized Copy-Paste Data Augmentation

基于上下文感知的集中式复制粘贴数据增强的多类野火图像语义分割
Kim, Joon Tai, Kunchala, Nishanth, Patel, Vishv, Chen, Tianle, Dong, Ziyu, Acero, Daniel Ospina, Williams, Roger, Kumar, Mrinal
Abstract
Producing accurate annotations for deep learning based image segmentation is both costly and labor intensive. This challenge is especially evident in wildland fire applications, where accurately labeled datasets are scarce due to the difficulty of collecting and annotating dynamic fire scenes. To address this problem, our previous work introduced the Centralized Copy-Paste Data Augmentation (CCPDA) method for semantic segmentation of wildland fire imagery, which generates artificial training samples by randomly pasting fire clusters from source images onto target images. However, random placement can produce contextually unrealistic scenes, such as fire burning on asphalt. In this paper, we present a context-aware strategy designed specifically to improve data quality and realism in small multiclass wildland fire datasets, ensuring that augmented samples remain contextually meaningful. The proposed method restricts fire placement to semantically valid target regions and selects the location whose Ash-Vegetation composition most closely matches the source context. This approach preserves existing fire regions in the target image, prevents unrealistic placements, and maintains contextual accuracy by generating images that resemble real wildland fire scenes. We evaluate the Context-Aware CCPDA strategy through numerical analysis and comparisons with other augmentation methods by a weighted sum-based multi-objective optimization (MOO) approach. The results confirm that the context-aware data augmentation strategy leads to improved segmentation performance and contextual realism, outperforming other augmentation procedures.
Chinese Translation
为深度学习图像分割生成准确的标注既耗费成本又劳动密集。这一挑战在野火应用中尤为明显，因为由于收集和标注动态火灾场景的困难，准确标记的数据集非常稀缺。为了解决这个问题，我们之前的工作提出了集中式复制粘贴数据增强（Centralized Copy-Paste Data Augmentation, CCPDA）方法，用于野火图像的语义分割，该方法通过将源图像中的火灾聚类随机粘贴到目标图像上来生成人工训练样本。然而，随机放置可能会产生上下文上不现实的场景，例如火灾在沥青上燃烧。在本文中，我们提出了一种专门设计的上下文感知策略，以提高小型多类野火数据集的数据质量和真实性，确保增强样本在上下文上保持有意义。该方法将火灾放置限制在语义有效的目标区域，并选择与源上下文的灰烬-植被组成最为匹配的位置。这种方法保留了目标图像中现有的火灾区域，防止不现实的放置，并通过生成类似真实野火场景的图像来保持上下文的准确性。我们通过数值分析和与其他增强方法的比较，采用基于加权和的多目标优化（Multi-Objective Optimization, MOO）方法评估上下文感知CCPDA策略。结果证实，上下文感知数据增强策略提高了分割性能和上下文真实性，优于其他增强程序。
View on arXiv Download PDF AI Translation
cs.CV / 13 / 2609.21242

SafeStyle: Calibrated Style Residual Injection for Controllable Style-Leakage Trade-off in Diffusion Stylization

SafeStyle：用于可控风格泄漏权衡的校准风格残差注入
Yang, Zhangping, Li, Min, Yan, Song, Gao, Rong, Bi, Xinliang, Xiong, Guanye, He, Yujie
Abstract
Reference-guided diffusion stylization aims to transfer visual style from a reference image while preserving the semantics specified by a text prompt. However, image conditioning often entangles transferable style cues with reference-specific content, leading to an inherent trade-off: stronger conditioning improves style fidelity but increases content leakage, whereas aggressive suppression reduces leakage at the cost of style expression. This challenge is further complicated by the distinct spatial organization of texture- and geometry-dominant styles. To address these issues, we propose SafeStyle, a training-free framework for calibrated style residual injection in frozen diffusion models. SafeStyle first estimates style-supported and content-associated subspaces from compact calibration sets, preserving their informative overlap while suppressing useless content variations. It then transports the purified style evidence over adaptive spatial granularity and constrains its effective influence through an explicit residual-norm budget. Experiments across texture- and geometry-dominant styles show that SafeStyle achieves a DINO style similarity of 0.432 while maintaining competitive text alignment. On a semantically disjoint leakage-stress benchmark, it further achieves a DINO style similarity of 0.474 with only 0.8\% semantic leakage, demonstrating an effective balance between style fidelity and reference-content suppression.
Chinese Translation
参考引导的扩散风格化旨在从参考图像中转移视觉风格，同时保留由文本提示指定的语义。然而，图像条件化常常将可转移的风格线索与特定于参考的内容纠缠在一起，导致固有的权衡：更强的条件化提高了风格保真度，但增加了内容泄漏，而激进的抑制则以牺牲风格表现为代价来减少泄漏。由于纹理和几何主导风格的独特空间组织，这一挑战变得更加复杂。为了解决这些问题，我们提出了SafeStyle，一个用于在冻结扩散模型中进行校准风格残差注入的无训练框架。SafeStyle首先从紧凑的校准集估计风格支持和内容关联的子空间，保留它们的信息重叠，同时抑制无用的内容变化。然后，它通过自适应空间粒度传输净化的风格证据，并通过明确的残差范数预算约束其有效影响力。在纹理和几何主导风格的实验中，SafeStyle实现了0.432的DINO风格相似度，同时保持了竞争性的文本对齐。在一个语义上不相交的泄漏压力基准上，它进一步实现了0.474的DINO风格相似度，仅有0.8%的语义泄漏，展示了风格保真度与参考内容抑制之间的有效平衡。
View on arXiv Download PDF AI Translation
cs.CV / 14 / 2609.21251

Geometry-Aware Diffusion Guidance via Curvature-Adaptive Tubular Correction

基于曲率自适应管状校正的几何感知扩散引导
Jiang, Enze, He, Jinwei, Ma, Zheng
Abstract
Gradient-guided diffusion samplers provide flexible priors for inverse problems and conditional generation, but strong guidance can move the sampling trajectory into regions where the learned score is poorly supported. Existing tangent-projection strategies limit first-order departure from an iso-density surface, yet discard potentially useful normal motion and overlook the second-order departure induced by tangent motion on a curved surface. We introduce curvature-adaptive tubular correction (CAT), a training-free plugin that regulates both effects within a shared, noise-dependent geometric budget. CAT decomposes the guidance gradient into normal and tangent components, charges normal displacement at first order and tangent displacement according to directional curvature, and obtains their jointly optimal magnitudes from a one-dimensional dual equation. Armijo backtracking calibrates the resulting finite step against the actual guidance objective, while matrix-free directional derivatives avoid constructing the full score Jacobian. We establish local guarantees for the tubular approximation, uniqueness of the correction, and sufficient objective decrease. Across seven inverse problems on FFHQ and ImageNet, CAT improves the evaluated pixel- and latent-space host samplers, with particularly consistent gains in perceptual metrics. It also improves black hole reconstruction on InverseBench and yields the lowest FID among the compared methods at every tested classifier-free guidance scale, while maintaining stable saturation and contrast. These results support curvature-aware tubular control as a reusable mechanism for stabilizing diffusion guidance.
Chinese Translation
梯度引导的扩散采样器为逆问题和条件生成提供了灵活的先验，但强引导可能会将采样轨迹移动到学习得分支持不足的区域。现有的切线投影策略限制了与等密度表面的第一阶偏离，但却忽略了潜在有用的法向运动，并且未考虑切线运动在曲面上引起的二阶偏离。我们引入了曲率自适应管状校正（Curvature-Adaptive Tubular Correction, CAT），这是一种无训练的插件，能够在共享的、依赖噪声的几何预算内调节这两种效应。CAT将引导梯度分解为法向分量和切向分量，按照第一阶对法向位移进行充电，并根据方向曲率对切向位移进行充电，并从一维双重方程中获得它们的联合最优幅度。Armijo回溯校准将得到的有限步长与实际引导目标进行校准，而无矩阵的方向导数则避免了构建完整的得分雅可比矩阵。我们建立了管状近似的局部保证、校正的唯一性以及充分的目标减少。在FFHQ和ImageNet上的七个逆问题中，CAT改善了评估的像素和潜在空间主采样器，在感知指标上尤其表现出一致的提升。它还改善了InverseBench上的黑洞重建，并在每个测试的无分类器引导尺度下，获得了与比较方法中最低的FID，同时保持稳定的饱和度和对比度。这些结果支持曲率感知的管状控制作为稳定扩散引导的可重用机制。
View on arXiv Download PDF AI Translation
cs.CV / 15 / 2609.21268

Edit-VAR: Taming Visual Autoregressive Model for Precise Video Editing

Edit-VAR：驯服视觉自回归模型以实现精确视频编辑
Zhao, Chongbo, Wang, Jiangming, Wang, Xilai, Wang, Xinyu, Tang, Jingyi, Hao, Chunjie, Song, Pengjie, Ma, Yue
Abstract
Text-guided video editing modifies target content while preserving the appearance and temporal coherence of unedited regions. Training-based approaches provide strong control but demand substantial data and computation. Training-free methods fall into inversion-free and inversion-based paradigms. Inversion-free approaches avoid trajectory recovery, but their source-preserving guidance can limit editing strength and leave semantic changes incomplete. Inversion-based approaches recover a latent trajectory before regeneration, where approximation errors can accumulate and cause source-content drift and temporal inconsistency. We introduce Edit-VAR, the first training-free and inversion-free framework for text-guided video editing with a pretrained visual autoregressive video model. Edit-VAR directly encodes the source video into multi-scale discrete tokens and performs probability-guided conditional token replacement for source preservation. Attention-guided token-wise and scale-aware modulation selectively relaxes source constraints over edit-relevant positions and generation stages. Scale-Decoupled Generation, implemented as late-scale constraint release, regenerates motion-consistent details and reduces texture fragmentation. Residual-guided token pruning further exploits redundancy at the final two high-resolution scales to reduce inference cost. Extensive experiments and a blind user study demonstrate that Edit-VAR outperforms existing training-free video editing methods overall in editing fidelity, source preservation, temporal coherence, and inference efficiency.
Chinese Translation
文本引导的视频编辑在修改目标内容的同时，保持未编辑区域的外观和时间一致性。基于训练的方法提供了强大的控制能力，但需要大量的数据和计算。无训练的方法分为无反演和基于反演的范式。无反演方法避免了轨迹恢复，但其源保留指导可能限制编辑强度，并使语义变化不完整。基于反演的方法在再生之前恢复潜在轨迹，其中近似误差可能累积并导致源内容漂移和时间不一致。我们提出了Edit-VAR，这是第一个基于文本引导的视频编辑的无训练和无反演框架，利用预训练的视觉自回归视频模型。Edit-VAR直接将源视频编码为多尺度离散标记，并执行概率引导的条件标记替换以保持源内容。基于注意力的标记级和尺度感知调制选择性地放宽了编辑相关位置和生成阶段的源约束。尺度解耦生成作为后期尺度约束释放实施，重新生成运动一致的细节并减少纹理碎片化。残差引导的标记修剪进一步利用最后两个高分辨率尺度的冗余，以降低推理成本。大量实验和盲用户研究表明，Edit-VAR在编辑保真度、源保留、时间一致性和推理效率方面整体上优于现有的无训练视频编辑方法。
View on arXiv Download PDF AI Translation
cs.CV / 16 / 2609.21276

Beyond Exact Match: Task-Aware GRPO for Cross-Domain PCBA Visual Question Answering

超越精确匹配：面向任务的跨域 PCBA 视觉问答的 GRPO
Li, Jia, Dai, Li, Jia, Peng, Hu, Zhenzhen, Chan, Chee Seng, Bao, Bingkun, Hong, Richang
Abstract
In automated Printed Circuit Board Assembly (PCBA) inspection, standards-guided decisions require systems to jointly reason over fine-grained visual cues, component semantics, and manufacturing knowledge. Although large vision-language models (VLMs) provide a promising foundation, their deployment is hindered by the domain shift between standards-derived samples and real-world production-line imagery, together with heterogeneous output spaces spanning choice-based and numerical counting tasks. To address these challenges, we propose a multimodal reasoning framework for cross-domain PCBA visual question answering. The framework converts standards-derived, real-world, and auxiliary PCB-domain data into a unified instruction format and constructs verified reasoning traces aligned with visual evidence, question semantics, candidate options, and ground-truth answers. We further introduce Task-Aware Group Relative Policy Optimization (GRPO), which moves beyond exact-match supervision by integrating multi-component semantic rewards for choice-based questions, distance-aware rewards for counting questions, and an auxiliary format reward for valid outputs. During inference, answer-option semantic consistency correction, self-consistency voting, and multi-model arbitration are combined to improve prediction robustness. The proposed system achieves an Overall Score of 83.24 on the official PCBA Standard-to-Real Grand Challenge leaderboard, demonstrating the effectiveness of task-aware reward design and robust inference for cross-domain PCBA visual question answering.
Chinese Translation
在自动化印刷电路板组装（PCBA）检测中，标准指导的决策要求系统共同推理细粒度的视觉线索、组件语义和制造知识。尽管大型视觉-语言模型（VLMs）提供了一个有前景的基础，但由于标准派生样本与真实生产线图像之间的领域转移，以及跨选择性和数值计数任务的异构输出空间，其部署受到阻碍。为了解决这些挑战，我们提出了一种用于跨域 PCBA 视觉问答的多模态推理框架。该框架将标准派生的、真实世界的和辅助 PCB 领域数据转换为统一的指令格式，并构建与视觉证据、问题语义、候选选项和真实答案对齐的经过验证的推理轨迹。我们进一步引入了面向任务的组相对策略优化（GRPO），该方法超越了精确匹配的监督，通过为基于选择的问题整合多组件语义奖励，为计数问题提供距离感知奖励，并为有效输出提供辅助格式奖励。在推理过程中，答案选项语义一致性修正、自我一致性投票和多模型仲裁相结合，以提高预测的鲁棒性。所提出的系统在官方 PCBA 标准到真实大挑战排行榜上获得了 83.24 的总体得分，证明了面向任务的奖励设计和跨域 PCBA 视觉问答的鲁棒推理的有效性。
View on arXiv Download PDF AI Translation
cs.CV / 17 / 2609.21304

Combining Object Detection with Geometry-Aware Clustering to Distinguish Overlapping Plants in UAV Imagery

结合目标检测与几何感知聚类以区分无人机影像中的重叠植物
Lee, Ik Jae, Nguyen, Hieu D., Meenar, Mahbubur, Martinez, Carlos Morrison, Connelly, Cameron
Abstract
Reliable plant-level information from unmanned aerial vehicle (UAV) imagery is important for automated crop monitoring. However, in dense crop canopies, adjacent plants frequently overlap and are detected as a single object, reducing the reliability of plant-level measurements. This study presents a geometry-aware post-detection framework for resolving overlapping plant instances using standard RGB UAV imagery. The framework combines object detection with geometric clustering of plant components. Leaves or branches detected within each bush-level region are represented using two complementary geometric features: component centroids and radial intersection points (RIPs) derived from detected plant structures. K-means and Gaussian mixture models determine whether a detected region contains a single plant or two overlapping plants. Density filtering suppresses spurious radial intersections, and a post-pipeline ensemble combines spatial and directional geometric information. The framework was evaluated using UAV imagery of eggplant and tomato crops under field conditions. Centroid-based clustering achieved an F1-score of 0.89 for eggplant, while the combined centroid-RIP approach achieved the best tomato performance, with an accuracy of 0.80, precision of 1.00, and F1-score of 0.75 using K-means. Density filtering substantially improved RIP-based clustering for tomato. The proposed approach provides a lightweight, modular engineering solution that can be integrated with existing RGB UAV monitoring pipelines without additional depth sensors, pixel-level segmentation, three-dimensional reconstruction, or retraining of the primary bush detector. The results demonstrate that geometric reasoning applied to existing detector outputs can complement deep-learning-based object detection and improve plant-level interpretation in dense agricultural canopies.
Chinese Translation
从无人机（UAV）影像中获取可靠的植物级信息对于自动化作物监测至关重要。然而，在密集的作物冠层中，相邻植物常常重叠并被检测为单一对象，从而降低了植物级测量的可靠性。本研究提出了一种几何感知的后检测框架，用于利用标准RGB无人机影像解决重叠植物实例的问题。该框架将目标检测与植物组件的几何聚类相结合。在每个灌木级区域内检测到的叶片或枝条使用两种互补的几何特征表示：从检测到的植物结构中派生的组件质心和径向交点（RIPs）。K均值和高斯混合模型用于判断检测区域是包含单一植物还是两个重叠植物。密度过滤抑制了虚假的径向交点，而后处理管道集成了空间和方向几何信息。该框架在田间条件下使用茄子和番茄作物的无人机影像进行了评估。基于质心的聚类在茄子上达到了0.89的F1分数，而结合质心-RIP方法在番茄上取得了最佳表现，准确率为0.80，精确率为1.00，F1分数为0.75（使用K均值）。密度过滤显著改善了番茄的基于RIP的聚类。所提出的方法提供了一种轻量级、模块化的工程解决方案，可以与现有的RGB无人机监测管道集成，而无需额外的深度传感器、像素级分割、三维重建或重新训练主要的灌木检测器。结果表明，应用于现有检测器输出的几何推理可以补充基于深度学习的目标检测，并改善密集农业冠层中的植物级解释。
View on arXiv Download PDF AI Translation
cs.CV / 18 / 2609.21322

S3VD: Semantic-Guidance Spatio-Temporal Scanning for Video Deraining

S3VD：用于视频去雨的语义引导时空扫描
Jiang, Kui, Chen, Yiang, Luo, Yan, Yu, Zhaocheng, Jiang, Junjun, Liu, Xianming
Abstract
Heavy rainfall severely degrades outdoor videos by corrupting high-frequency details and introducing motion blur, critically undermining the reliability of visual tasks. Recently, State Space Models (SSMs), particularly Mamba, have emerged as efficient alternatives for vision tasks with their linear complexity and ability to model long-range dependencies. However, when confronted with the poor visual representations in rainy videos, Mamba still faces difficulties in preserving the integrity of 2D spatial semantics and modeling 3D spatio-temporal correlations. To break these limitations, we introduce S3VD, a Semantic-Guidance Spatio-Temporal Scanning framework for video deraining, featuring two key innovations: Multi-Scale Semantic Fusion (MSSF) Module and Spatio-Temporal Scanning Fusion (STSF) Module. The former integrates temporal semantic priors from DINOv2 to guide precise feature representation and counteract the loss of local semantic context inherent to Mamba's 1D flatten operation, enhancing robustness against extreme degradation. The latter introduces a spatio-temporal scanning mechanism and devises a Decoupled-Gating Mamba (DG-Mamba) layer, which employs two independent gates to adaptively control preceding and subsequent contextual information within the input clip, optimizing intra-frame and inter-frame correlation modeling. Experiments on video deraining benchmarks demonstrate the superiority of S3VD, achieving state-of-the-art performance with an average 0.84 dB PSNR improvement over Mamba-based baselines.
Chinese Translation
强降雨严重影响户外视频，通过破坏高频细节和引入运动模糊，严重削弱了视觉任务的可靠性。最近，状态空间模型（State Space Models, SSMs），特别是Mamba，作为视觉任务的高效替代方案，因其线性复杂性和建模长程依赖的能力而受到关注。然而，在面对雨天视频中较差的视觉表现时，Mamba仍然面临保持二维空间语义完整性和建模三维时空相关性的困难。为了解决这些局限性，我们提出了S3VD，一种用于视频去雨的语义引导时空扫描框架，具有两个关键创新：多尺度语义融合（Multi-Scale Semantic Fusion, MSSF）模块和时空扫描融合（Spatio-Temporal Scanning Fusion, STSF）模块。前者整合了来自DINOv2的时间语义先验，以指导精确的特征表示，并抵消Mamba的1D扁平化操作固有的局部语义上下文丢失，从而增强对极端降解的鲁棒性。后者引入了一种时空扫描机制，并设计了一个解耦门控Mamba（Decoupled-Gating Mamba, DG-Mamba）层，采用两个独立的门控来自适应地控制输入片段中的前后上下文信息，优化帧内和帧间相关性建模。在视频去雨基准测试中的实验表明，S3VD的优越性，其在Mamba基础上实现了平均0.84 dB的PSNR提升，达到了最先进的性能。
View on arXiv Download PDF AI Translation
cs.CV / 19 / 2609.21323

VeriFuse: Bounded Vision-Language Arbitration and Reason-Guided Refinement for Cooperative 3D Perception

VeriFuse：用于合作3D感知的有界视觉-语言仲裁与推理引导的精炼
Lin, Hongyi, Liu, Yiyao, Kang, Qi, Huang, Heye, Liu, Yang, Koutsopoulos, Haris, Zhao, Jinhua
Abstract
Vision-language models (VLMs) have demonstrated strong scene understanding and semantic judgment across diverse tasks, but their appropriate role in cooperative perception remains unclear. Directly asking a VLM to regress 3D detections is unreliable and computationally expensive, whereas using it to select the output of a single source discards useful information from other agents. We introduce VeriFuse, a bounded arbitration framework for vehicle-infrastructure cooperative 3D detection. Each agent first produces detections independently. Around each vehicle and roadside proposal, VeriFuse generates source-conditioned geometric candidates and combines the original detections, their perturbations, and cross-source hypotheses into a unified candidate pool. A frozen VLM then chooses among three admissible actions: SELECT an adequate candidate; REFINE an existing anchor when an object is supported but all candidates are geometrically inadequate; or REJECT an unsupported infrastructure-only proposal. Experiments on the DAIR-V2X dataset show that VeriFuse achieves 0.494/0.357 cooperative 3D AP50/AP70 and limits the relative vehicle-side BEV AP50 drop under a 300 ms delay to 1.7%. Overall, VeriFuse assigns the VLM a clear and constrained role in cooperative perception: semantic reasoning resolves ambiguity among cross-agent hypotheses, while deterministic constraints determine the final 3D geometry.
Chinese Translation
视觉-语言模型（VLMs）在多种任务中展示了强大的场景理解和语义判断能力，但它们在合作感知中的适当角色仍不明确。直接要求VLM回归3D检测结果既不可靠又计算成本高，而使用它选择单一来源的输出则会丢弃来自其他代理的有用信息。我们提出了VeriFuse，一种用于车辆-基础设施合作3D检测的有界仲裁框架。每个代理首先独立产生检测结果。在每个车辆和路边提案周围，VeriFuse生成源条件的几何候选，并将原始检测、其扰动和跨源假设结合成一个统一的候选池。然后，一个冻结的VLM在三个可接受的行动中进行选择：选择（SELECT）一个合适的候选；当一个物体被支持但所有候选在几何上不合适时，精炼（REFINE）一个现有的锚点；或者拒绝（REJECT）一个不支持的仅基础设施提案。对DAIR-V2X数据集的实验表明，VeriFuse在合作3D AP50/AP70上分别达到了0.494/0.357，并将相对车辆侧BEV AP50在300毫秒延迟下的下降限制在1.7%以内。总体而言，VeriFuse为VLM在合作感知中分配了一个明确且受限的角色：语义推理解决了跨代理假设之间的歧义，而确定性约束则决定了最终的3D几何形状。
View on arXiv Download PDF AI Translation
cs.CV / 20 / 2609.21347

Cube-Splat: High-Fidelity 360{\deg} Gaussian Splatting SLAM via Cubemap Factorization and Adjoint-Consistent Optimization

Cube-Splat：通过立方体映射分解和伴随一致优化实现高保真度360°高斯点云SLAM
Guo, Xiangfei, Shi, Hao, Zhang, Yufan, Yi, Zhonghua, Mao, Yongqi, Yin, Xiaoting, Wang, Kaiwei
Abstract
Recent progress in 3D Gaussian Splatting (3DGS) has enabled dense visual SLAM with pinhole cameras, yet most pipelines are not designed for panoramic imagery. We present Cube-Splat, the first panoramic GS-SLAM framework that factorizes each 360{\deg} frame into a cubemap of four fixed-orientation virtual pinhole views sharing a single optical center. By designating the front face as the primary pose state, we accumulate gradients from all faces via an adjoint mapping, thereby enabling multi-face observations to coherently update a single state while strictly preserving cross-view geometric consistency. Concurrently, our mapping module densifies and optimizes anisotropic Gaussians using aggregated cubemap rays for high-fidelity, dense reconstruction. Furthermore, to rigorously evaluate panoramic SLAM under diverse and challenging conditions, we introduce SynPano, a highly scalable, photorealistic synthetic dataset featuring parameterized complex trajectories and multi-modal ground truth. Extensive evaluations on two public benchmarks (PALVIO and OmniBlender) and our SynPano dataset, collectively encompassing both indoor and outdoor scenes, demonstrate that Cube-Splat achieves state-of-the-art (SOTA) performance in tracking accuracy and reconstruction fidelity. Both the source code and the SynPano dataset are available at https://github.com/guoxf304/CubeSplat.
Chinese Translation
最近在3D高斯点云（3DGS）方面的进展使得使用针孔相机进行密集视觉SLAM成为可能，但大多数流程并未针对全景图像进行设计。我们提出了Cube-Splat，这是第一个全景GS-SLAM框架，它将每个360°帧分解为一个由四个固定方向的虚拟针孔视图组成的立方体映射，这些视图共享一个光学中心。通过将前面视图指定为主要姿态状态，我们通过伴随映射积累来自所有面的梯度，从而使多面观察能够一致地更新单一状态，同时严格保持视图间的几何一致性。同时，我们的映射模块利用聚合的立方体映射光线对各向异性高斯进行稠密化和优化，以实现高保真度的密集重建。此外，为了在多样且具有挑战性的条件下严格评估全景SLAM，我们引入了SynPano，这是一个高度可扩展的、具有参数化复杂轨迹和多模态地面真实数据的光真实感合成数据集。在两个公共基准（PALVIO和OmniBlender）以及我们的SynPano数据集上进行了广泛评估，涵盖室内和室外场景，结果表明Cube-Splat在跟踪精度和重建保真度方面达到了最先进的（SOTA）性能。源代码和SynPano数据集可在https://github.com/guoxf304/CubeSplat获取。
View on arXiv Download PDF AI Translation
cs.CV / 21 / 2609.21351

PrismAlign: Prior-Steered Multi-View VLM Alignment for Hallucination-Robust Table OCR

PrismAlign：基于先验引导的多视角视觉语言模型对齐以增强表格OCR的抗幻觉能力
Liu, Guangyi, Huang, Qianjun, Hou, Boyu
Abstract
Table extraction suffers from frequent structural errors and semantic hallucinations. We propose PrismAlign, a multi-VLM framework aligning diverse visual perspectives to resolve ambiguity. It integrates priors of table logic to assess output plausibility, decoupling structural alignment from cell content alignment. A Bayesian decision strategy maximizes alignment accuracy by exploiting the correlation between extraction errors and computable rule violations. Evaluated on open-source and custom VLMs, PrismAlign reduces hallucinations and achieves state-of-the-art performance on OmniDocBench 1.5, as well as on the table category of CC-OCR and PureDocBench.
Chinese Translation
表格提取常常面临频繁的结构错误和语义幻觉。我们提出了PrismAlign，一个多视角视觉语言模型（multi-VLM）框架，通过对齐多样的视觉视角来解决歧义。该框架整合了表格逻辑的先验知识，以评估输出的合理性，将结构对齐与单元内容对齐解耦。贝叶斯决策策略通过利用提取错误与可计算规则违反之间的相关性，最大化对齐准确性。在开源和定制的视觉语言模型上进行评估，PrismAlign减少了幻觉现象，并在OmniDocBench 1.5以及CC-OCR和PureDocBench的表格类别上实现了最先进的性能。
View on arXiv Download PDF AI Translation
cs.CV / 22 / 2609.21354

Field Tracking of Insects Using a Stereoscopic Event-Based Camera Setup

基于立体事件相机设置的昆虫现场追踪
Shenwai, Pratham G., Lankheet, Martin J., Hrynuk, John T., Mahadeeswara, Mandiyam Y., Srinivasan, Mandyam V., Ravi, Sridhar
Abstract
High-speed tracking of small, fast-moving organisms in their natural environments is important to better understand their behavior and ecology. Traditional frame-based imaging suffers from motion blur due to low temporal resolution, and data storage limitations, propelling a search for more adaptive solutions. Event cameras, which capture changes in brightness at pixel level instead of entire frames, have emerged as a promising solution by increasing temporal resolution and data efficiency. Here, we demonstrate the use of event-based imaging with standard video-based processing methods by converting the asynchronous events into conventional video formats, allowing us to leverage the event camera's enhanced temporal detail to capture intricate insect flight movements and apply established image analysis techniques. Coupling this conversion process with a stereoscopic configuration provides continuous, low-latency, three-dimensional tracking of fast-moving subjects in field conditions. As a result, we substantially mitigate motion artifacts and achieve more accurate representations of animal movements. By making event-based imaging more readily applicable in natural field settings, our method support broader applications across animal behavior and ecological research, agricultural management, and other fields requiring high-fidelity object tracking in the wild.
Chinese Translation
在自然环境中对小型快速移动生物进行高速追踪对于更好地理解它们的行为和生态至关重要。传统的帧基成像由于时间分辨率低而容易出现运动模糊，并且存在数据存储限制，这促使人们寻找更具适应性的解决方案。事件相机通过捕捉像素级别的亮度变化而非整个帧，成为一种有前景的解决方案，能够提高时间分辨率和数据效率。在此，我们展示了如何使用基于事件的成像结合标准的视频处理方法，通过将异步事件转换为传统视频格式，使我们能够利用事件相机增强的时间细节来捕捉复杂的昆虫飞行运动，并应用已建立的图像分析技术。将这一转换过程与立体配置相结合，可以在现场条件下对快速移动对象进行连续、低延迟的三维追踪。因此，我们大大减轻了运动伪影，并实现了动物运动的更准确表示。通过使基于事件的成像在自然现场环境中更易于应用，我们的方法支持在动物行为和生态研究、农业管理以及其他需要高保真度野外物体追踪的领域的更广泛应用。
View on arXiv Download PDF AI Translation
cs.CV / 23 / 2609.21363

Hiding in Plain Sight: A Diffusion-based Mitigation of Geolocation Privacy Leakage in Vision-Language Models

隐匿于明处：基于扩散的地理位置隐私泄露缓解方法在视觉-语言模型中的应用
Wang, Yining, Li, Xi, Zhang, Mi, Zhang, Xiaohan, You, Xiaoyu, Qian, Zhenxing, Wen, Mi
Abstract
Multimodal large reasoning models (MLRMs) have demonstrated remarkable capabilities in complex visual understanding. However, this very power introduces a critical yet underexplored privacy threat: adversaries can exploit MLRMs to precisely infer users' geographic locations from casually shared photographs, by performing structured reasoning over subtle visual cues such as architectural styles, vegetation, and lighting conditions. In this work, we present a systematic study of MLRM-driven geolocation privacy leakage. We first reveal that refusal-based safeguards are critically insufficient, as carefully crafted jailbreak prompts can raise model response rates to 100%. We further identify that existing defenses, which inject imperceptible perturbations into shared images, suffer from structural limitations intrinsic to their pixel-space optimization, resulting in degraded black-box transferability and pronounced visual artifacts. Motivated by these findings, we propose a diffusion-based framework that provides targeted, proactive defense against geolocation privacy leakage. By injecting perturbations into the latent space of a diffusion model during reverse sampling, our method operates directly on high-level semantic representations, thereby resolving the effectiveness-utility bottlenecks by construction. We further ground our optimization with GeoCLIP, a model explicitly aligned with GPS coordinates, as a surrogate to pinpoint and disrupt the geographic signals that MLRMs exploit for location inference. This targeted semantic disruption yields significantly stronger black-box transferability while preserving perceptual image quality, offering a seamless integration on social media platforms.
Chinese Translation
多模态大型推理模型（MLRMs）在复杂视觉理解方面展现了卓越的能力。然而，这种强大能力引入了一种关键但尚未深入探讨的隐私威胁：对手可以利用MLRMs从随意分享的照片中精确推断用户的地理位置，通过对建筑风格、植被和光照条件等微妙视觉线索进行结构化推理。在本研究中，我们系统地研究了MLRM驱动的地理位置隐私泄露。我们首先揭示，基于拒绝的保护措施严重不足，因为精心设计的越狱提示可以将模型的响应率提高到100%。我们进一步发现，现有的防御措施通过向共享图像注入不可察觉的扰动，受到其像素空间优化内在的结构性限制，导致黑箱可转移性降低和明显的视觉伪影。基于这些发现，我们提出了一种基于扩散的框架，提供针对地理位置隐私泄露的有针对性的主动防御。通过在反向采样过程中向扩散模型的潜在空间注入扰动，我们的方法直接作用于高层语义表示，从而在构建上解决了有效性与实用性之间的瓶颈。我们进一步将优化与GeoCLIP模型结合，该模型与GPS坐标明确对齐，作为替代品以精准定位和干扰MLRMs用于位置推断的地理信号。这种有针对性的语义干扰显著增强了黑箱可转移性，同时保持了感知图像质量，提供了在社交媒体平台上的无缝集成。
View on arXiv Download PDF AI Translation
cs.CV / 24 / 2609.21371

RobotEQ-Video: A Video-Centric Benchmark for Social Proactive Intelligence with World-State Taxonomy

RobotEQ-Video：一种以视频为中心的社会主动智能基准，结合世界状态分类法
Che, Xinyi, Lian, Zheng, Fang, Kuofei, Wang, Xuehao, Gao, Xinghai, Wu, Junqing, Wu, Chuyu, Liu, Liyi, Huang, Yanhan, Xie, Keyi, Ouyang, Haomin, Wu, Jinyang, Zhang, Fan, Zeng, Runhao, Yang, Xun, He, Bin
Abstract
Social Proactive Intelligence (SPI) extends proactive assistance beyond task completeness to consider social appropriateness in diverse embodied scenarios. However, prior SPI research faces two key limitations. First, existing work focuses on static images, whereas dynamic videos provide crucial cues for inferring human states and needs, offering richer information than isolated images. Second, prior work often relies on free-form data collection pipelines, which fail to guarantee comprehensive coverage of diverse scenarios. To address these gaps, we introduce RobotEQ-Video, shifting the focus from image-centric to video-centric analysis. To ensure comprehensive video coverage, we construct a hierarchical world-state taxonomy organized into a four-level coarse-to-fine structure, comprising 6 domains, 20 dimensions, 142 level-1 attributes, and 816 level-2 attributes. The resulting benchmark comprises 2K+ videos with 100K+ human annotations and 16K+ labels for assessing behavior properness. Benchmark evaluation reveals that current systems remain unreliable and fall short of human performance. We further explore how world models can help tackle this task. This work advances SPI research from static images to dynamic videos and ensures more comprehensive scenario coverage during benchmarking.
Chinese Translation
社会主动智能（SPI）将主动辅助的范围扩展到任务完成之外，考虑在多样化的具身场景中的社会适宜性。然而，之前的SPI研究面临两个主要限制。首先，现有研究主要集中于静态图像，而动态视频提供了推断人类状态和需求的重要线索，提供的信息比孤立的图像更为丰富。其次，以往的研究通常依赖于自由形式的数据收集流程，这无法保证对多样化场景的全面覆盖。为了解决这些问题，我们引入了RobotEQ-Video，将重点从以图像为中心转向以视频为中心的分析。为了确保视频的全面覆盖，我们构建了一个分层的世界状态分类法，组织成四级粗到细的结构，包括6个领域、20个维度、142个一级属性和816个二级属性。最终的基准包含2000多个视频，配有超过10万个人工注释和超过16000个用于评估行为适当性的标签。基准评估显示，当前系统仍然不可靠，未能达到人类的表现。我们进一步探讨了世界模型如何帮助解决这一任务。本研究将SPI研究从静态图像推进到动态视频，并确保在基准测试中对场景的更全面覆盖。
View on arXiv Download PDF AI Translation
cs.CV / 25 / 2609.21379

JEPA Guided Diffusion: Predictive Vision-Language Conditioning for Generative Traffic Forecasting

JEPA引导的扩散：用于生成交通预测的预测视觉-语言条件
Nguyen, Trinh Tra Giang, Vo, Thanh Nguyen, Bui, Nguyen Hoai Thuong, Bui, Ha Duc
Abstract
Accurate traffic forecasting requires both understanding scene dynamics and synthesizing realistic future observations. Recent diffusion-based video generation models produce visually plausible predictions but require expensive end-to-end training and often entangle scene understanding with image synthesis. In this work, we propose a decoupled forecasting framework that separates future representation learning from video generation. A frozen V-JEPA encoder first extracts predictive latent representations from the observed traffic videos, capturing the underlying scene dynamics in a semantic latent space. A lightweight latent alignment module then projects these representations into the conditioning space of a frozen Cosmos diffusion module, enabling future video synthesis without retraining the large generative model. By freezing all foundation models and training only the lightweight alignment module, the proposed framework substantially reduces optimization complexity while preserving forecasting capability. Experimental results on the AI City Challenge 2026 Track 5 benchmark demonstrate that the proposed method achieved a score of 75.1297, ranking third in the competition. These results suggest that predictive world representations learned by V-JEPA can effectively guide downstream video generation, providing a practical and efficient alternative to end-to-end diffusion-based forecasting.
Chinese Translation
准确的交通预测需要理解场景动态并合成现实的未来观察。最近的基于扩散的视频生成模型能够产生视觉上可信的预测，但需要昂贵的端到端训练，并且通常将场景理解与图像合成纠缠在一起。在本研究中，我们提出了一种解耦的预测框架，将未来表示学习与视频生成分开。一个冻结的V-JEPA编码器首先从观察到的交通视频中提取预测潜在表示，捕捉语义潜在空间中的基本场景动态。然后，一个轻量级的潜在对齐模块将这些表示投影到冻结的Cosmos扩散模块的条件空间中，使得未来视频合成无需重新训练大型生成模型。通过冻结所有基础模型，仅训练轻量级对齐模块，所提出的框架显著降低了优化复杂性，同时保持了预测能力。在AI City Challenge 2026 Track 5基准测试中的实验结果表明，所提出的方法获得了75.1297的分数，在比赛中排名第三。这些结果表明，V-JEPA学习的预测世界表示可以有效引导下游视频生成，提供了一种实用且高效的替代端到端基于扩散的预测方法。
View on arXiv Download PDF AI Translation
cs.CV / 26 / 2609.21386

AgentVidBench: A Multi-Hop Video Question Answering Benchmark for Evaluating MLLM Agents

AgentVidBench：用于评估多跳视频问答的基准测试
An, Seoyeon, Jang, Hyeonseo, Kim, Minsu, Lee, Chanho, Park, Younghan, Lee, Kangwook
Abstract
Comprehensive video understanding is crucial for advancing artificial intelligence toward the intricate dynamics of the physical world. While recent advances in Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in video understanding, existing benchmarks remain confined to simple scene-level queries or global summaries that require only single-step inference. Real-world video understanding involves more challenging tasks that require multi-hop multimodal reasoning, and there is a critical absence of video benchmarks equipped to rigorously evaluate these agentic capabilities. To bridge this gap, we introduce AgentVidBench, a multi-hop video question answering benchmark focused on evaluating the spatial, temporal, and causal reasoning capabilities of MLLM agents. Beyond standard question-answer pairs, AgentVidBench provides step-by-step solution traces to support trajectory evaluation that assesses whether agents explicitly acquire the evidence needed to justify their answers. Experiments with 12 proprietary and open-source MLLMs show that single-turn performance remains limited on AgentVidBench, while integrating these models into state-of-the-art agentic workflows generally improves performance with respect to both accuracy and trajectory scores. We further present a simple yet effective agentic strategy that serves as a competitive baseline on AgentVidBench, establishing our benchmark as a holistic testbed for future research on agentic video understanding. Code and datasets are available at https://github.com/krafton-ai/agentvidbench and https://huggingface.co/datasets/agentvidbench/agentvidbench.
Chinese Translation
全面的视频理解对于推动人工智能向物理世界复杂动态的发展至关重要。尽管近年来多模态大型语言模型（MLLMs）在视频理解方面展现了显著的能力，但现有基准测试仍然局限于简单的场景级查询或仅需单步推理的全局摘要。现实世界中的视频理解涉及更具挑战性的任务，这些任务需要多跳多模态推理，而目前缺乏能够严格评估这些智能能力的视频基准测试。为填补这一空白，我们提出了AgentVidBench，一个专注于评估MLLM代理的空间、时间和因果推理能力的多跳视频问答基准测试。除了标准的问题-答案对，AgentVidBench还提供逐步解决方案跟踪，以支持轨迹评估，评估代理是否明确获取了证明其答案所需的证据。对12个专有和开源MLLM的实验表明，在AgentVidBench上的单轮性能仍然有限，而将这些模型整合到最先进的智能工作流程中通常会提高准确性和轨迹得分。我们进一步提出了一种简单而有效的智能策略，作为AgentVidBench上的竞争基线，确立了我们的基准测试作为未来智能视频理解研究的整体测试平台。代码和数据集可在 https://github.com/krafton-ai/agentvidbench 和 https://huggingface.co/datasets/agentvidbench/agentvidbench 获取。
View on arXiv Download PDF AI Translation
cs.CV / 27 / 2609.21400

A Scene Language Model for Open-Vocabulary Scene Mapping

用于开放词汇场景映射的场景语言模型
Lilja, Adam, Hübel, Fabio, He, Siming, Fu, Junsheng, Tomlin, Claire, Hammarstrand, Lars, Malik, Jitendra, Frey, Jonas, Pavone, Marco
Abstract
Open-vocabulary 3D scene mapping aims to build a persistent representation of the objects in an environment. Existing systems typically rely on engineered mapping pipelines to associate observations, merge information across views, and maintain a consistent scene representation over time. Many additionally store feature-rich object representations, such as embeddings or image crops, increasing the size and complexity of the persistent memory. We introduce SceneLM, a Scene-Language Model that directly maintains a textual scene map. The full scene is represented as a structured text list of objects, which serves as the model's only persistent memory. For each input image, the model reads the current scene state and updates the map by adding, editing, and removing objects. To learn this behavior, we introduce supervision tasks for iterative scene map maintenance together with an automatic annotation pipeline that generates training data from images without human labels. We evaluate SceneLM on both a language-grounded retrieval benchmark and a localization benchmark. Across both benchmarks, the model produces a scene map that achieves competitive performance with complete mapping systems built from dedicated perception and geometric modules while producing a scene representation that is 6-12x more compact. We further show that SceneLM can be run online on an edge device through experiments on a quadruped. These results show that a persistent open-vocabulary 3D scene map can be maintained directly by a single vision-language model using only a lightweight text representation. Training and inference code is available on https://goldengait.github.io/scenelm/.
Chinese Translation
开放词汇的3D场景映射旨在构建环境中对象的持久表示。现有系统通常依赖于工程化的映射管道来关联观测、合并视图间的信息，并在时间上保持一致的场景表示。许多系统还存储丰富特征的对象表示，如嵌入或图像裁剪，增加了持久内存的大小和复杂性。我们提出了SceneLM，一种直接维护文本场景地图的场景语言模型。完整场景被表示为一个结构化的对象文本列表，这也是模型唯一的持久内存。对于每个输入图像，模型读取当前场景状态，并通过添加、编辑和移除对象来更新地图。为了学习这种行为，我们引入了迭代场景地图维护的监督任务，并结合一个自动注释管道，从图像中生成无人工标签的训练数据。我们在语言基础检索基准和定位基准上评估了SceneLM。在这两个基准上，该模型生成的场景地图在性能上与建立在专用感知和几何模块上的完整映射系统具有竞争力，同时生成的场景表示紧凑度提高了6-12倍。我们进一步展示了SceneLM可以通过在四足机器人上的实验实现在线运行。这些结果表明，持久的开放词汇3D场景地图可以通过单一的视觉-语言模型直接维护，仅使用轻量级的文本表示。训练和推理代码可在 https://goldengait.github.io/scenelm/ 获取。
View on arXiv Download PDF AI Translation
cs.CV / 28 / 2609.21402

SIRA: Reasoning-Aware Surgical Instrument Segmentation via Query-Anchored Alignment

SIRA：基于查询锚定对齐的推理感知外科器械分割
Zhang, Zhibo, Wang, Qijie, Yan, Zengqiang
Abstract
Surgical instrument segmentation (SIS) plays a critical role in robotic assistance and surgical workflow analysis. However, most existing SIS methods formulate segmentation as a category-driven localization problem, limiting their ability to capture procedural context and task-dependent semantics in surgical workflows. We introduce Reasoning-Aware Surgical Instrument Segmentation (RA-SIS), a task formulation that frames segmentation as query-conditioned inference under surgical context. To benchmark this setting, we construct SurgRS, a surgical reasoning segmentation dataset consisting of 41,000 image-text pairs, which aligns instance-level masks with structured query-answer supervision to enable semantic grounding at the pixel level. Based on SurgRS, we propose Surgical Instrument Reasoning and Segmentation Assistant (SIRA), a multimodal framework that disentangles target-level and query-level semantics and integrates them with visual features through query-anchored dual alignment. By aligning query semantics with spatial features and segmentation prompts, SIRA enhances semantic-visual consistency in mask prediction. Extensive experiments on SurgRS demonstrate improvements over existing reasoning-aware baselines. Code is available at https://github.com/linxir226/SIRA.
Chinese Translation
外科器械分割（SIS）在机器人辅助和外科工作流程分析中发挥着关键作用。然而，大多数现有的SIS方法将分割问题表述为以类别为驱动的定位问题，这限制了它们捕捉外科工作流程中程序上下文和任务依赖语义的能力。我们提出了推理感知外科器械分割（RA-SIS），一种将分割框架设定为在外科上下文下的查询条件推理的任务表述。为了对这一设定进行基准测试，我们构建了SurgRS，一个包含41,000个图像-文本对的外科推理分割数据集，该数据集将实例级掩码与结构化查询-答案监督对齐，以实现像素级的语义基础。基于SurgRS，我们提出了外科器械推理与分割助手（SIRA），一个多模态框架，它将目标级和查询级语义进行解耦，并通过查询锚定的双重对齐将其与视觉特征集成。通过将查询语义与空间特征和分割提示对齐，SIRA增强了掩码预测中的语义-视觉一致性。在SurgRS上的大量实验表明，相较于现有的推理感知基线，SIRA有显著的改进。代码可在 https://github.com/linxir226/SIRA 获取。
View on arXiv Download PDF AI Translation
cs.CV / 29 / 2609.21407

Quantization-Aware Kalman Estimation for Diffusion Sampling

量化感知的卡尔曼估计在扩散采样中的应用
Shi, Qitan, Jin, Cheng, Zhang, Jiawei, Gu, Yuantao
Abstract
Quantization offers a practical path to deploying diffusion models with reduced memory and computation, but aggressive compression can cause quantized outputs to deviate substantially from their full-precision counterparts. Sampling-stage correction methods seek to compensate for such deviations during sampling, but existing approaches rely primarily on local information and underexploit trajectory history, limiting their ability to correct errors that propagate across timesteps. In this work, we formulate sampling with a quantized denoiser as an online estimation problem, using the history of quantized denoiser outputs to recover the underlying full-precision outputs required by the sampler. We propose QuAKE, a Quantization-Aware Kalman Estimator that combines a smooth trajectory prior with a conditional Gaussian observation model. At each sampling step, QuAKE recursively updates the posterior over the output window in closed form and feeds its posterior mean to the sampler. QuAKE is a lightweight plug-and-play corrector that requires no modification to the quantized network and naturally supports arbitrary high-order multistep ODE samplers. Experiments across W4A4-quantized text-to-image diffusion models show that QuAKE consistently outperforms existing methods in reducing the distributional discrepancy from full-precision sampling.
Chinese Translation
量化为部署具有较低内存和计算需求的扩散模型提供了一条实际路径，但激进的压缩可能导致量化输出与其全精度对应物之间存在显著偏差。采样阶段的修正方法旨在在采样过程中补偿这种偏差，但现有方法主要依赖于局部信息，未能充分利用轨迹历史，限制了它们纠正跨时间步传播的错误的能力。在本研究中，我们将使用量化去噪器的采样过程形式化为一个在线估计问题，利用量化去噪器输出的历史来恢复采样器所需的基础全精度输出。我们提出了QuAKE，一种量化感知的卡尔曼估计器，它将平滑轨迹先验与条件高斯观测模型相结合。在每个采样步骤中，QuAKE以闭合形式递归更新输出窗口的后验，并将其后验均值反馈给采样器。QuAKE是一种轻量级的即插即用修正器，无需对量化网络进行修改，并自然支持任意高阶多步常微分方程（ODE）采样器。在W4A4量化的文本到图像扩散模型上的实验表明，QuAKE在减少与全精度采样的分布差异方面始终优于现有方法。
View on arXiv Download PDF AI Translation
cs.CV / 30 / 2609.21412

When Online Adaptation Hurts: Parameter-Frozen Test-Time Ensembling for Continual Medical Image Segmentation

在线适应何时会造成损害：持续医学图像分割的参数冻结测试时间集成
Huang, Ruijie
Abstract
Medical image segmenters often get worse when sites, scanner vendors, or protocols change. Continual test-time adaptation (CTTA) addresses this problem without target labels, but it can be impossible to update a model on a non-stationary stream and can lead to a lot of errors. We examine a more reasonable and meaningful alternative: parameter-frozen inference enhancement(PIE). We use a source-trained segmenter that learns about anatomy-preserving scale and flip views, maps their predictions back to the native location, and averages the probabilities. We do not modify the weights of the model or the normalization statistics. On a cardiac MRI stream from M\&Ms, which is trained on vendor A and evaluated sequentially on vendors B, C, and D, PIE has 0.7786 mean Dice, compared to 0.7680 for source-only inference and 0.7388--0.7416 for five other online-adaptation baselines. The controlled ablations show that performance saturates at 28 views, and confidence weighting, class-prior correction, connected-component filtering, morphological refinement, and inter-slice smoothing have no effect or cause negative transfer. Qualitative results on cardiac MRI and fundus images are also consistent with the frozen ensemble keeping thinner and nested anatomical structures. These results provide a strong, stable baseline for medical CTTA and expose an important failure mode: adaptation and handcrafted refinement can be less reliable than carefully designed inference.
Chinese Translation
医学图像分割器在站点、扫描仪供应商或协议发生变化时往往表现更差。持续测试时间适应（CTTA）解决了这个问题，而无需目标标签，但在非平稳流上更新模型可能是不可能的，并且可能导致大量错误。我们考察了一种更合理和有意义的替代方案：参数冻结推理增强（PIE）。我们使用一个源训练的分割器，该分割器学习保持解剖结构的尺度和翻转视图，将其预测映射回原始位置，并对概率进行平均。我们不修改模型的权重或归一化统计数据。在来自M&Ms的心脏MRI流上，该流在供应商A上进行训练，并在供应商B、C和D上进行顺序评估，PIE的平均Dice值为0.7786，而仅源推理为0.7680，五个其他在线适应基线的值为0.7388至0.7416。控制性消融实验表明，性能在28个视图时饱和，置信度加权、类别先验修正、连通组件过滤、形态学细化和切片间平滑没有效果或导致负迁移。心脏MRI和眼底图像的定性结果也与冻结集成保持更细和嵌套的解剖结构一致。这些结果为医学CTTA提供了一个强大而稳定的基线，并揭示了一个重要的失败模式：适应和手工细化可能不如精心设计的推理可靠。
View on arXiv Download PDF AI Translation
cs.CV / 31 / 2609.21424

P$^3$-SAM: SAM with Perceptual Parallel Prompt for Few-Shot Strip Steel Surface Defect Segmentation

P$^3$-SAM：带有感知并行提示的SAM用于少样本带钢表面缺陷分割
Xu, Qian, Xiong, Hang, Wang, Anpeng, Kwong, Sam, Zhang, Cong, Cong, Runmin
Abstract
Few-shot semantic segmentation (FSS) of strip steel surface defects (S$^3$D) has posed significant challenges distinct from natural scenes. Unlike natural images, S$^3$D task exhibits unique characteristics including low local contrast, uneven illumination, and complex fine-grained texture patterns. Although recent methods based on Segment Anything Model (SAM) have shown promise in FSS on natural images by leveraging SAM's powerful pre-trained representations, these unique industrial characteristics of S$^3$D images lead to performance drop when directly applying SAM to industrial defect scenarios. In this paper, we propose a novel Perceptual Parallel Prompt (P$^3$) framework that empowers SAM, creating the P$^3$-SAM model to address these challenges through two core strategies. First, we develop a Perceptual-Optimized Encoding (POE) strategy that enhances local contrast and preserves critical texture details for S$^3$D segmentation. Second, we introduce the Parallel Prompt Generator (PPG) strategy that simultaneously generates both semantic and spatial prompts, enabling comprehensive guidance for SAM's decoder across varying images. Extensive experiments on three few-shot S$^3$D benchmarks demonstrate that P$^3$-SAM achieves state-of-the-art performance, with particularly notable improvements of 12.00% in mIoU on Surface Defects-4i dataset.
Chinese Translation
带钢表面缺陷的少样本语义分割（FSS）面临着与自然场景截然不同的重大挑战。与自然图像不同，带钢表面缺陷（S$^3$D）任务展现出独特的特征，包括低局部对比度、不均匀的照明和复杂的细粒度纹理模式。尽管基于Segment Anything Model（SAM）的最新方法在自然图像的FSS中表现出色，利用了SAM强大的预训练表示，但这些S$^3$D图像的独特工业特征导致在工业缺陷场景中直接应用SAM时性能下降。本文提出了一种新颖的感知并行提示（P$^3$）框架，赋能SAM，创建了P$^3$-SAM模型，通过两种核心策略应对这些挑战。首先，我们开发了一种感知优化编码（POE）策略，增强局部对比度并保留S$^3$D分割的关键纹理细节。其次，我们引入了并行提示生成器（PPG）策略，同时生成语义和空间提示，为SAM的解码器提供全面的指导，适应不同的图像。在三个少样本S$^3$D基准上的大量实验表明，P$^3$-SAM实现了最先进的性能，特别是在Surface Defects-4i数据集上mIoU提升了12.00%。
View on arXiv Download PDF AI Translation
cs.CV / 32 / 2609.21437

Think Locally, Refine Globally for Memory-Efficient 3D Reconstruction

局部思考，全球优化：一种内存高效的3D重建方法
Zhou, Jingke, Ma, Chenhang, Zhong, Zhizhou, Liu, Mingkai, Zhou, Zhuang, ji, Yicheng, Su, Binghua, Cai, Bo, Huang, Xianliang
Abstract
We propose LoG-VGGT, a memory-efficient framework for long-sequence 3D reconstruction that balances local temporal modeling with global camera consistency. Instead of relying on full global attention, our method introduces cross-window attention at a small subset of transformer blocks, enabling effective information propagation across adjacent temporal windows while keeping memory usage bounded. To mitigate long-term pose drift, we further design a global camera consistency refinement module, where camera tokens interact with compact register tokens via cross-attention to enforce scene-level constraints across the entire sequence. This design enables joint optimization of camera representations and significantly improves long-horizon pose stability without incurring the high cost of sequence-wide attention. Extensive experiments demonstrate that LoG-VGGT achieves improved depth accuracy and robust camera pose estimation across multiple long-sequence benchmarks, while delivering competitive streaming reconstruction performance.
Chinese Translation
我们提出了LoG-VGGT，一个内存高效的长序列3D重建框架，旨在平衡局部时间建模与全局相机一致性。我们的方法并不依赖于全面的全局注意力，而是在少量的变换器块中引入跨窗口注意力，从而在保持内存使用受限的同时，有效地促进相邻时间窗口之间的信息传播。为了减轻长期姿态漂移，我们进一步设计了一个全局相机一致性优化模块，其中相机标记通过交叉注意力与紧凑的注册标记进行交互，以在整个序列中强制执行场景级约束。这一设计使得相机表示的联合优化成为可能，并显著提高了长时间范围内的姿态稳定性，而无需承担序列级注意力的高成本。大量实验表明，LoG-VGGT在多个长序列基准测试中实现了更好的深度精度和稳健的相机姿态估计，同时提供了具有竞争力的流媒体重建性能。
View on arXiv Download PDF AI Translation
cs.CV / 33 / 2609.21449

ME-Dex 1.0: Bringing Heterogeneous Tactile Sensing into World Action Modeling

ME-Dex 1.0：将异构触觉感知引入世界动作建模
Zhang, Xuancheng, Liu, Xuetao, Tang, Qianying, Wang, Jizhe, Cheng, Zhijing, Lin, Bochen, Wen, Haoran, Li, Ming, Zhan, Kun, Liu, Yu
Abstract
World Action Models bring the predictive capabilities of video models into robot action generation, providing a rich foundation for modeling future visual states. Tactile sensing complements this foundation with direct measurements of physical interaction. Some existing methods use tactile features as conditioning inputs without jointly predicting future tactile states, visual observations, and actions. Our key insight is that tactile signals, like video, provide observations of the evolving world state and should be modeled as future observations alongside video. We present ME-Dex-1.0 (MachEmbodied-Dex-1.0), a unified World Action Tactile Model for joint visual, tactile, and action learning. ME-Dex-1.0 adopts a Mixture-of-Transformers architecture comprising a Video Expert, a Tactile Expert, and an Action Expert, all trained with flow matching. We use shared attention connects the experts in intermediate layers, allowing action generation to draw on learned representations of visual and tactile dynamics during joint denoising. To support multi-source heterogeneous tactile inputs, a Canonical Hand Model and a Unified Tactile Autoencoder map tactile observations from different embodiments and sensing layouts into shared spatial and latent spaces. To address the limited availability of paired visual, tactile, and action data, we develop the Agentic Tactile Data Engine, an agent-based data production platform. It supplements RoboTwin and DexJoCo with tactile data recorded directly from force sensors during trajectory replay in simulation. Experiments on the RoboTwin, DexJoCo, and ManiFeel simulation platforms, together with real robot evaluations, demonstrate improved manipulation performance using both grippers and dexterous hands equipped with tactile sensing.
Chinese Translation
世界动作模型将视频模型的预测能力引入机器人动作生成，为建模未来视觉状态提供了丰富的基础。触觉感知通过直接测量物理交互来补充这一基础。一些现有方法将触觉特征作为条件输入，但未能共同预测未来的触觉状态、视觉观察和动作。我们的关键见解是，触觉信号与视频一样，提供了不断变化的世界状态的观察，应该与视频一起建模为未来观察。我们提出了ME-Dex-1.0（MachEmbodied-Dex-1.0），这是一个统一的世界动作触觉模型，用于视觉、触觉和动作的联合学习。ME-Dex-1.0采用了一种混合变换器架构，包括视频专家、触觉专家和动作专家，所有专家均通过流匹配进行训练。我们使用共享注意力在中间层连接专家，使得动作生成能够在联合去噪过程中借助学习到的视觉和触觉动态表示。为了支持多源异构触觉输入，一个规范手模型和一个统一触觉自编码器将来自不同体现和感知布局的触觉观察映射到共享的空间和潜在空间。为了解决配对的视觉、触觉和动作数据的有限可用性，我们开发了Agentic Tactile Data Engine，这是一个基于代理的数据生产平台。它补充了RoboTwin和DexJoCo，通过在模拟中的轨迹重放期间直接从力传感器记录的触觉数据。对RoboTwin、DexJoCo和ManiFeel模拟平台的实验，以及真实机器人评估，表明使用配备触觉感知的夹持器和灵巧手的操作性能得到了改善。
View on arXiv Download PDF AI Translation
cs.CV / 34 / 2609.21455

CompAdapt: Adaptable Composite Motion Modeling for Physics-Consistent Text-to-Video Generation

CompAdapt：适应性复合运动建模用于物理一致的文本到视频生成
Qin, Haoran, Wu, Renlong, Huang, Tianyu, Ding, Yukang, Li, Hui, Zuo, Wangmeng
Abstract
While diffusion-based text-to-video (T2V) models have demonstrated impressive capability in generating realistic and temporally coherent videos, they often fail to respect fundamental physical dynamics. Although recent physics-constrained methods incorporate explicit dynamics priors to improve physical plausibility, they remain limited to simple single-type motions, depend on manually specified parameters, and struggle to generalize to unseen physical laws. In this work, we propose CompAdapt, a physics-consistent T2V framework for adaptable generation across complex real-world scenarios. It extends neural dynamics modeling beyond single-type motions to encompass composite physical behaviors, including coupled motions, multi-stage transitions, and multi-object collisions. Furthermore, CompAdapt translates natural language prompts into structured physical semantics, enabling end-to-end specification of motion types, temporal relations, and initial physical parameters. To generalize to novel physical environments, CompAdapt introduces dynamics-aware prior matching, achieving one-shot adaptation without retraining the core dynamics module. In addition, a physics-aware latent feature fusion module improves visual fidelity under fast and complex motion. Experiments on physics-focused T2V benchmarks demonstrate that CompAdapt improves physical consistency over both general T2V models and physics-constrained baselines, while preserving high visual quality and adaptability to unseen dynamics. The project page is available at https://makapic.github.io/CompAdapt/ .
Chinese Translation
尽管基于扩散的文本到视频（T2V）模型在生成逼真且时间一致的视频方面表现出色，但它们往往未能遵循基本的物理动态。尽管近期的物理约束方法通过引入显式动态先验来提高物理合理性，但仍然局限于简单的单一类型运动，依赖于手动指定的参数，并且难以推广到未见过的物理法则。在本研究中，我们提出了CompAdapt，一个物理一致的T2V框架，能够在复杂的现实场景中进行适应性生成。它将神经动态建模扩展到复合物理行为，包括耦合运动、多阶段过渡和多物体碰撞。此外，CompAdapt将自然语言提示转换为结构化的物理语义，使得运动类型、时间关系和初始物理参数的端到端规范成为可能。为了在新颖的物理环境中实现推广，CompAdapt引入了动态感知的先验匹配，实现了一次性适应而无需重新训练核心动态模块。此外，物理感知的潜在特征融合模块在快速和复杂运动下提高了视觉保真度。在针对物理重点的T2V基准测试中的实验表明，CompAdapt在物理一致性方面优于一般T2V模型和物理约束基线，同时保持了高视觉质量和对未见动态的适应性。项目页面可访问 https://makapic.github.io/CompAdapt/ 。
View on arXiv Download PDF AI Translation
cs.CV / 35 / 2609.21462

PSEE: Progressive Sensor Event Expansion for Point-Supervised Temporal Action Localization

PSEE：基于点监督的渐进式传感器事件扩展用于时间动作定位
Yin, Jiaxi, Wang, Ge, Ding, Han, Wang, Fei
Abstract
Temporal action localization (TAL) in wearable sensor streams identifies action classes and temporal boundaries, enabling finer-grained activity understanding than conventional action recognition. However, training typically requires costly start--end annotations for every action instance. To reduce this burden, we study point-supervised TAL, where each instance is labeled with only one timestamp and its class. We propose Progressive Sensor Event Expansion (PSEE), which combines semantic activations, sensor-specific transition evidence, and adaptive temporal ownership to recover point-supervised pseudo segments. These segments supervise standard TAL detectors without modifying their inference procedures. Cross-subject experiments on four inertial-sensing benchmarks demonstrate improved pseudo-boundary quality over adapted point-supervised baselines, compatibility with different TAL detectors, and robustness to point sampling. Code is available at https://github.com/joeeeeyin/PSEE.
Chinese Translation
可穿戴传感器流中的时间动作定位（TAL）旨在识别动作类别及其时间边界，实现比传统动作识别更细粒度的活动理解。然而，训练通常需要对每个动作实例进行昂贵的起止时间标注。为减轻这一负担，我们研究了点监督的时间动作定位，其中每个实例仅标注一个时间戳及其类别。我们提出了渐进式传感器事件扩展（PSEE），该方法结合语义激活、传感器特定的转换证据及自适应时间归属，恢复点监督的伪段。利用这些伪段监督标准的TAL检测器，且无需修改其推理流程。在四个惯性传感基准数据集上的跨主体实验表明，PSEE在伪边界质量上优于改编的点监督基线，兼容不同的TAL检测器，并对点采样具有鲁棒性。代码已开源，详见https://github.com/joeeeeyin/PSEE。
View on arXiv Download PDF AI Translation
cs.CV / 36 / 2609.21468

SkillIR: Evolving Scene-Aware Skills for Agentic Image Restoration

SkillIR：为智能图像修复演变的场景感知技能
Shao, Jie, Hu, Shengkai, Zhang, Xu, Song, Beihang, Jing, Yongcheng, Wu, Xu, Wan, Jun
Abstract
This paper studies agentic image restoration, in which multimodal agents coordinate specialized restoration tools to recover images affected by complex degradations. Existing restoration agents often derive complete tool-use plans from the original degraded image or retrieve previously successful trajectories, providing limited support for adapting individual actions to evolving intermediate restoration states. We find that accepted tool executions can change the residual degradation state and, consequently, the applicability of subsequent tools. To address this issue, we propose SkillIR, a skill-guided framework that represents restoration experience as degradation-centered action evidence rather than complete tool-use trajectories. SkillIR consolidates context-dependent action outcomes into scene-aware restoration skills that characterize applicable conditions, expected effects, and attributable failure cases. Instead of prescribing a complete restoration plan, the retrieved skills guide one bounded action at a time within a verified residual-state loop: each tool output is treated as a candidate, committed only after transition verification, and followed by reassessment of the active residual degradations. After each rollout, the resulting evidence is used to create, refine, or patch dynamic skills, enabling accumulated restoration experience to improve decision-making for subsequent inputs. Experiments on synthetic and real-world multi-degradation datasets demonstrate that SkillIR improves restoration quality and enables more reliable and effective tool use.
Chinese Translation
本文研究了智能图像修复，其中多模态代理协调专门的修复工具，以恢复受到复杂退化影响的图像。现有的修复代理通常从原始退化图像中推导出完整的工具使用计划，或检索先前成功的轨迹，这对适应个体动作以应对不断变化的中间修复状态的支持有限。我们发现，已接受的工具执行可以改变残余退化状态，从而影响后续工具的适用性。为了解决这个问题，我们提出了SkillIR，一个以技能为导向的框架，将修复经验表示为以退化为中心的动作证据，而不是完整的工具使用轨迹。SkillIR将依赖于上下文的动作结果整合为场景感知修复技能，这些技能描述了适用条件、预期效果和可归因的失败案例。SkillIR并不规定一个完整的修复计划，而是通过验证的残余状态循环引导一次有限的动作：每个工具输出被视为候选项，仅在过渡验证后才被承诺，并随后对活跃的残余退化进行重新评估。在每次执行后，所产生的证据用于创建、完善或修补动态技能，使得积累的修复经验能够改善后续输入的决策。对合成和真实世界多退化数据集的实验表明，SkillIR提高了修复质量，并使工具使用更加可靠和有效。
View on arXiv Download PDF AI Translation
cs.CV / 37 / 2609.21474

MT-WAM: Reorienting the One-Pass Predictive Representation Toward Action Generation

MT-WAM：重新定向单次预测表示以生成动作
Yang, Yiguang, Peng, Jiankun, Wang, Xiaoming, Zhang, Yiran, Fang, Zhibo
Abstract
Fast-WAM shows that video-action co-training improves control without generating future video at inference, making the representation from a single video diffusion Transformer forward central to action generation. However, future-observation prediction does not explicitly prioritize the future dynamics and visual structure needed for control. We present MT-WAM, which retains the original training objectives and adds complementary supervision for future two-dimensional point trajectories and visual features. A lightweight dual-stream branch copied from the video backbone's final blocks provides target-specific processing, while a structured attention mask prevents cross-stream attention. Motion-stream tokens supply additional dynamics conditions to the action expert. Future visual-feature prediction provides supervision in a feature space that captures object and spatial structure. This supervision trains the video backbone to provide more informative visual context for action generation under changing visual conditions, without adding visual-feature-stream tokens to action conditioning. At inference, MT-WAM uses video and motion caches computed once per replan and skips future-video prediction. Without additional embodied policy pretraining, MT-WAM achieves 98.2% success on LIBERO and 73.7% on LIBERO-Plus, exceeding Fast-WAM by 23.8 percentage points on the latter. On RoboTwin 2.0 Clean2Rand, Random success increases from 6.30% to 19.40%; across four real-world tasks, average success increases from 67.0% to 77.8%.
Chinese Translation
Fast-WAM 表明视频-动作共同训练在推理时无需生成未来视频即可改善控制，使得来自单个视频扩散 Transformer 的表示在动作生成中至关重要。然而，未来观察预测并未明确优先考虑控制所需的未来动态和视觉结构。我们提出了 MT-WAM，它保留了原始训练目标，并为未来的二维点轨迹和视觉特征添加了补充监督。一个轻量级的双流分支从视频主干的最终块复制，提供特定于目标的处理，同时结构化的注意力掩码防止跨流注意力。运动流令牌为动作专家提供额外的动态条件。未来视觉特征预测在捕捉对象和空间结构的特征空间中提供监督。这种监督训练视频主干在变化的视觉条件下提供更具信息量的视觉上下文以生成动作，而无需将视觉特征流令牌添加到动作条件中。在推理时，MT-WAM 使用每次重新规划计算一次的视频和运动缓存，并跳过未来视频预测。在没有额外的具身策略预训练的情况下，MT-WAM 在 LIBERO 上取得了 98.2% 的成功率，在 LIBERO-Plus 上取得了 73.7% 的成功率，后者比 Fast-WAM 超过 23.8 个百分点。在 RoboTwin 2.0 Clean2Rand 上，随机成功率从 6.30% 提高到 19.40%；在四个现实世界任务中，平均成功率从 67.0% 提高到 77.8%。
View on arXiv Download PDF AI Translation
cs.CV / 38 / 2609.21480

OpenSAL360: Open-Source Crowdsourcing Platform for Omnidirectional Video Saliency Collection

OpenSAL360：开源全景视频显著性收集众包平台
Bryncev, Alexey, Moskalenko, Andrey, Shilovskaya, Kira, Kosmynin, Ivan, Vatolin, Dmitriy
Abstract
Omnidirectional video saliency prediction plays an important role in many immersive multimedia applications, including viewport-adaptive streaming and compression, foveated rendering, mesh simplification, perceptual quality assessment. Yet progress in this area remains constrained by the cost and complexity of collecting eye-tracking data with VR headsets, which makes large-scale dataset creation difficult to extend. We present OpenSAL360, the first open-source platform for scalable, low-cost 360{\deg} video saliency collection. Unlike conventional VR-based protocols, it requires only a standard screen, mouse, and internet connection, enabling parallel saliency data collection from common crowdsourcing assessors without specialized hardware. We validate our collection protocol against seven well-established VR eye-tracking datasets and conduct ablation studies on key interface, pre-, and post-processing parameters. To demonstrate the effectiveness and scalability of the proposed methodology, we collect and publicly release a saliency dataset covering 500 omnidirectional videos annotated by 2,000+ crowdsourcing assessors, making it, to the best of our knowledge, the largest dataset in this field. We make OpenSAL360 publicly available at https://github.com/msu-video-group/OpenSAL360.
Chinese Translation
全景视频显著性预测在许多沉浸式多媒体应用中发挥着重要作用，包括视口自适应流媒体和压缩、聚焦渲染、网格简化和感知质量评估。然而，该领域的进展受到使用虚拟现实（VR）头显收集眼动追踪数据的成本和复杂性的限制，使得大规模数据集的创建难以扩展。我们提出了OpenSAL360，这是第一个可扩展、低成本的360°视频显著性收集开源平台。与传统的基于VR的协议不同，它只需标准屏幕、鼠标和互联网连接，便可从普通众包评估者那里并行收集显著性数据，而无需专门的硬件。我们将我们的收集协议与七个成熟的VR眼动追踪数据集进行了验证，并对关键界面、预处理和后处理参数进行了消融研究。为了证明所提方法的有效性和可扩展性，我们收集并公开发布了一个涵盖500个全景视频的显著性数据集，该数据集由2000多名众包评估者标注，尽我们所知，这是该领域最大的数据集。我们在https://github.com/msu-video-group/OpenSAL360上公开提供OpenSAL360。
View on arXiv Download PDF AI Translation
cs.CV / 39 / 2609.21498

VoxelTTO: Voxel-Aligned Feed-Forward 3D Gaussian Splatting with Test-Time Optimization

VoxelTTO：基于体素对齐的前馈3D高斯点云重建与测试时优化
Zhao, Yibin, Pan, Yihan, Li, Yangwen, Nan, Jun, Yi, Jianjun
Abstract
Recent feed-forward 3D Gaussian Splatting (3DGS) methods typically regress pixel-aligned Gaussian primitives, often causing excessive overlap and artifacts, while inaccuracies in predicted camera poses can lead to misalignment in novel-view synthesis (NVS). We present VoxelTTO, a feed-forward framework for reconstructing geometrically accurate 3DGS scenes from an arbitrary number of images and optional camera parameters. VoxelTTO aggregates dense image features into a global voxel representation and decodes Gaussians from voxel features, breaking the pixel-to-Gaussian correspondence. To exploit known camera parameters while keeping the pretrained visual foundation model (VFM) parameters frozen, we introduce test-time optimization (TTO) that adapts lightweight LoRA modules using pose supervision. We further replace vanilla 3DGS rasterization with stochastic solid volume rendering during training and inference, improving geometric fidelity. Training updates only the voxel-aligned Gaussian reconstruction modules, requiring 80 GPU hours. Experiments on Replica, Tanks and Temples, and DTU demonstrate improved RGB-D NVS and camera-pose estimation relative to prior methods.
Chinese Translation
近期的前馈3D高斯点云重建（3DGS）方法通常回归像素对齐的高斯原语，这往往会导致过度重叠和伪影，而预测的相机姿态不准确可能导致新视图合成（NVS）中的不对齐。我们提出了VoxelTTO，这是一个前馈框架，旨在从任意数量的图像和可选的相机参数中重建几何上准确的3DGS场景。VoxelTTO将密集的图像特征聚合为全局体素表示，并从体素特征中解码高斯，打破了像素与高斯之间的对应关系。为了在保持预训练视觉基础模型（VFM）参数不变的情况下利用已知的相机参数，我们引入了测试时优化（TTO），该方法使用姿态监督来调整轻量级的LoRA模块。我们进一步在训练和推理过程中用随机固体体积渲染替代了传统的3DGS光栅化，从而提高了几何保真度。训练仅更新体素对齐的高斯重建模块，需耗时80个GPU小时。在Replica、Tanks and Temples和DTU上的实验表明，相较于先前的方法，我们在RGB-D NVS和相机姿态估计方面取得了改进。
View on arXiv Download PDF AI Translation
cs.CV / 40 / 2609.21502

Adaptive World Memory 3D Foundation Model for Scalable 3D Mapping, Localization, and Rendering

用于可扩展3D映射、定位和渲染的自适应世界记忆3D基础模型
Deng, Tianchen, Shen, Guole, Shen, Yilin, Wu, Wenhua, Fang, Yilin, Ma, Ziqi, Zhang, Tianjun, Yuan, Shenghai, Burgard, Wolfram, Wang, Hesheng
Abstract
Recent 3D foundation models enable generalizable geometric reasoning from RGB images but remain limited in persistent memory, scalability, and renderable scene modeling. We present a memory-centric 3D foundation model for scalable robotic localization, reconstruction, and Gaussian rendering. Its core is an adaptive world memory mechanism that combines transformer-based gated updates with test-time temporal-spatial regulation. Learned gates control recurrent memory propagation, while temporal state evolution and spatial observation-state consistency regulate token-wise updates and forgetting over long image sequences. To support large-scale mapping, we organize memory into local submaps and integrate progressive mapping and tracking, loop closure, and SL(4)-based global refinement to maintain local accuracy and global consistency. A Gaussian reconstruction head decodes memory-enhanced features into renderable primitives, unifying camera pose estimation, dense point-cloud reconstruction, and photorealistic rendering within a single model. Experiments on public benchmarks and self-collected datasets from diverse robotic platforms demonstrate improved trajectory accuracy, reconstruction completeness, and rendering quality over existing 3D foundation reconstruction and SLAM baselines. These results support adaptive memory as a foundation for persistent robotic world modeling. The dataset and code will be made publicly available at \href{https://github.com/dtc111111/AWM-3DFM}{https://github.com/dtc111111/AWM-3DFM}.
Chinese Translation
近期的3D基础模型能够从RGB图像中进行可泛化的几何推理，但在持久记忆、可扩展性和可渲染场景建模方面仍然存在局限。我们提出了一种以记忆为中心的3D基础模型，用于可扩展的机器人定位、重建和高斯渲染。其核心是一个自适应世界记忆机制，结合了基于变换器的门控更新与测试时的时空调节。学习到的门控控制递归记忆传播，而时间状态演变和空间观测状态一致性则调节基于标记的更新和在长图像序列中的遗忘。为了支持大规模映射，我们将记忆组织为局部子图，并整合渐进式映射与跟踪、回环闭合以及基于SL(4)的全局优化，以保持局部精度和全局一致性。高斯重建头将增强的记忆特征解码为可渲染的原始体，统一了相机姿态估计、密集点云重建和照片级真实感渲染于单一模型中。在公共基准测试和来自多种机器人平台的自收集数据集上的实验表明，相较于现有的3D基础重建和SLAM基线，我们的模型在轨迹精度、重建完整性和渲染质量上均有所提升。这些结果支持自适应记忆作为持久机器人世界建模的基础。数据集和代码将公开发布于 https://github.com/dtc111111/AWM-3DFM。
View on arXiv Download PDF AI Translation
cs.CV / 41 / 2609.21516

2D GauSS-MI: Efficient Active Scene Reconstruction with Balanced Visual and Geometric Quality

2D GauSS-MI：平衡视觉与几何质量的高效主动场景重建
Xie, Yuhan, Pan, Jia
Abstract
Active reconstruction requires efficient active view selection to achieve high-quality reconstruction within limited onboard computational resources. Existing methods face challenges in adequately balancing visual and geometric quality with the computational efficiency required for real-time operation. In this work, we present an active reconstruction framework based on 2D Gaussian Splatting (2DGS). We develop an efficient online 2DGS mapping pipeline for incremental RGB-D observations and introduce a probabilistic reliability model that characterizes the view-dependent reconstruction quality of individual 2D Gaussian splats. Building on this model, we formulate 2D Gaussian Splatting Shannon Mutual Information (2D GauSS-MI), a mutual-information-based metric that exploits the explicit surface orientation of 2DGS to evaluate the expected information gain of candidate views. The proposed metric enables active view selection to account for both visual and geometric reconstruction quality. We evaluate the proposed system against three state-of-the-art baselines on eight Replica scenes. Experimental results demonstrate that our method achieves a favorable balance between visual and geometric reconstruction quality with substantially lower computational cost and competitive model storage.
Chinese Translation
主动重建需要高效的主动视图选择，以在有限的机载计算资源内实现高质量重建。现有方法在平衡视觉与几何质量以及实时操作所需的计算效率方面面临挑战。在本研究中，我们提出了一种基于2D高斯点云（2D Gaussian Splatting, 2DGS）的主动重建框架。我们开发了一个高效的在线2DGS映射管道，用于增量RGB-D观测，并引入了一个概率可靠性模型，该模型表征了单个2D高斯点的视图依赖重建质量。在此模型的基础上，我们提出了2D高斯点云香农互信息（2D GauSS-MI），这是一种基于互信息的度量，利用2DGS的显式表面方向来评估候选视图的预期信息增益。所提出的度量使得主动视图选择能够同时考虑视觉和几何重建质量。我们在八个Replica场景上将所提出的系统与三种最先进的基线进行评估。实验结果表明，我们的方法在视觉和几何重建质量之间实现了良好的平衡，同时显著降低了计算成本，并在模型存储方面具有竞争力。
View on arXiv Download PDF AI Translation
cs.CV / 42 / 2609.21521

VidOmni-Bench: A Benchmark for Fine-Grained Video Understanding via Spatio-Temporal Event Verification across Complexity and Duration

VidOmni-Bench：通过跨复杂性和持续时间的时空事件验证进行细粒度视频理解的基准
Kim, Changbeen, Chang, Junwon, Kim, Kipyo, Shinoda, Risa, Saito, Kuniaki, Kim, Donghyun
Abstract
While Video Large Language Models (Video-LLMs) have recently demonstrated strong performance, reliably evaluating their fine-grained video understanding remains challenging. Existing benchmarks often rely on question answering or ground-truth caption matching, where models may succeed through superficial cues and incomplete annotations. To this end, we introduce VidOmni-Bench, a benchmark that requires models to verify whether each event in dense video captions is supported by the video. VidOmni-Bench consists of 500 videos spanning five complexity types and diverse durations from 4 seconds to 90 minutes. After collecting videos along these axes, we use diverse Video-LLMs to generate dense captions and obtain human-verified sentence-level labels, where sentences containing incorrect events serve as hard negatives for evaluation. Our experiments on VidOmni-Bench reveal three key findings: (i) Video-LLMs frequently generate hallucinated descriptions in dense video captioning; (ii) they also struggle as verifiers, failing to reliably detect plausible but incorrect event descriptions; and (iii) model weaknesses vary across video complexity and duration, revealing diverse, model-specific bottlenecks in current Video-LLMs.
Chinese Translation
尽管视频大型语言模型（Video-LLMs）最近表现出强大的性能，但可靠地评估它们的细粒度视频理解仍然具有挑战性。现有基准通常依赖于问答或真实标签匹配，在这些情况下，模型可能通过表面的线索和不完整的注释获得成功。为此，我们引入了VidOmni-Bench，这是一个基准，要求模型验证密集视频字幕中的每个事件是否得到视频的支持。VidOmni-Bench包含500个视频，涵盖五种复杂性类型和从4秒到90分钟的多样化持续时间。在沿这些维度收集视频后，我们使用多种视频大型语言模型生成密集字幕，并获得经过人工验证的句子级标签，其中包含不正确事件的句子作为评估的难负样本。我们在VidOmni-Bench上的实验揭示了三个关键发现：（i）视频大型语言模型在密集视频字幕生成中经常产生虚构的描述；（ii）它们作为验证者也面临困难，无法可靠地检测出合理但不正确的事件描述；（iii）模型的弱点在视频复杂性和持续时间上有所不同，揭示了当前视频大型语言模型中的多样化、特定于模型的瓶颈。
View on arXiv Download PDF AI Translation
cs.CV / 43 / 2609.21522

Refine Then Fusion: Training-Free 3D Point Cloud Adaptation with Priority Refinement and Multi-Modal Knowledge Fusion

精细化后融合：无训练的3D点云适应方法，优先精细化与多模态知识融合
Cheng, Hang, Chen, Yan, Fan, Mingyu, Zeng, Long
Abstract
Recent pre-trained foundation models provide rich multi-modal priors for downstream 3D vision tasks. However, the effectiveness of these representations in few-shot scenarios is limited by two fundamental challenges: High-dimensional features often contain substantial channel redundancy and task-irrelevant noise, while the reliability of different modalities varies across samples. Consequently, direct aggregation of heterogeneous representations overlooks sample-dependent modality reliability and may obscure the discriminative cues essential. To address these limitations, we propose Refine Then Fusion(RTF), a training-free framework for few-shot 3D recognition. RTF first identifies discriminative feature channels by jointly modeling inter-class similarity and intra-class stability, thereby decoupling domain-specific knowledge refinement from the cached representations of pre-trained models. It then introduces a reliability-aware fusion mechanism that estimates sample-wise modality reliability from the distribution shifts induced by feature refinement, enabling adaptive aggregation of multi-modal representations. Furthermore, RTF constructs a memory cache that integrates instance-level support features with class-level prototypes to infer query labels. Extensive experiments on five benchmarks demonstrate that RTF consistently outperforms single-modal baselines, partial-fusion variants, and existing lightweight adaptation methods, achieving state-of-the-art few-shot 3D recognition performance without gradient optimization, additional training data, auxiliary training, or parameter updates.
Chinese Translation
近期的预训练基础模型为下游3D视觉任务提供了丰富的多模态先验。然而，这些表示在少样本场景中的有效性受到两个基本挑战的限制：高维特征通常包含大量的通道冗余和与任务无关的噪声，而不同模态的可靠性在样本之间存在差异。因此，异构表示的直接聚合忽视了样本依赖的模态可靠性，可能会掩盖重要的区分线索。为了解决这些限制，我们提出了精细化后融合（Refine Then Fusion, RTF），一种无训练的少样本3D识别框架。RTF首先通过联合建模类间相似性和类内稳定性来识别区分性特征通道，从而将领域特定知识的精细化与预训练模型的缓存表示解耦。然后，它引入了一种可靠性感知的融合机制，该机制根据特征精细化引起的分布变化来估计样本级模态可靠性，从而实现多模态表示的自适应聚合。此外，RTF构建了一个内存缓存，将实例级支持特征与类级原型整合，以推断查询标签。在五个基准上的广泛实验表明，RTF始终优于单模态基线、部分融合变体以及现有的轻量级适应方法，在不进行梯度优化、额外训练数据、辅助训练或参数更新的情况下，实现了最先进的少样本3D识别性能。
View on arXiv Download PDF AI Translation
cs.CV / 44 / 2609.21541

Purification and Regulation: Comorbidity-Aware Multi-Label Few-Shot Learning for Medical Image Classification

净化与调节：关注共病的多标签少样本学习在医学图像分类中的应用
Lin, Ying-Chih, Kuo, Po-Chih, Chen, Yong-Sheng
Abstract
Multi-label few-shot learning (MLFSL) remains a significant challenge in medical image analysis (MIA). Current metric-based meta-learning methods face two critical limitations in MIA. First, conventional prototype generation often entangles irrelevant disease information, leading to contaminated prototypes and degraded performance. Second, prior studies typically enforce inter-class separability in embedding space, largely neglecting the inherent correlations among diseases. To overcome these challenges, we propose Prototype Purification and Regulation (PPR), a novel MLFSL framework for MIA. PPR first performs prototype purification by leveraging sample-level comorbidity scores to emphasize disease-specific features, producing purified prototypes that better characterize each disease. Building upon these purified prototypes, PPR further addresses the underexplored problem of inter-class prototype distance in MIA by incorporating disease-level comorbidity statistics to adaptively regulate inter-class similarity, forming a comorbidity-aware embedding space. Overall, PPR sequentially enables the model to capture pure disease features and inter-class relationships for reliable MLFSL in MIA. Extensive experiments across four chest X-ray benchmark datasets, including cross-domain evaluation, show that PPR consistently outperforms state-of-the-art methods, significantly improving disease detection while demonstrating robust generalization and clinical applicability.
Chinese Translation
多标签少样本学习（MLFSL）在医学图像分析（MIA）中仍然面临重大挑战。目前基于度量的元学习方法在MIA中存在两个关键限制。首先，传统的原型生成往往混杂了无关的疾病信息，导致原型被污染，性能下降。其次，以往研究通常在嵌入空间中强制执行类间可分性，忽视了疾病之间的内在关联。为了解决这些问题，我们提出了一种新颖的MLFSL框架——原型净化与调节（PPR），专门用于MIA。PPR首先通过利用样本级共病评分进行原型净化，强调疾病特征，生成更好地表征每种疾病的净化原型。在这些净化原型的基础上，PPR进一步解决了MIA中类间原型距离这一尚未充分探讨的问题，通过结合疾病级共病统计信息自适应调节类间相似性，形成一个关注共病的嵌入空间。总体而言，PPR顺序地使模型能够捕捉纯粹的疾病特征和类间关系，从而实现MIA中的可靠MLFSL。在四个胸部X光基准数据集上的广泛实验，包括跨领域评估，表明PPR始终优于最先进的方法，显著提高了疾病检测能力，同时展示了强大的泛化能力和临床适用性。
View on arXiv Download PDF AI Translation
cs.CV / 45 / 2609.21543

From Retrieval to Recognition:How Vision--Language Models Become OCR Specialists

从检索到识别：视觉-语言模型如何成为OCR专家
Huangfu, Yuanxiang, Zhong, Hanmeng, Chen, Linqing, Hui, Jeffrey Tiong Jee
Abstract
Does a general vision--language model acquire specialized OCR ability by developing a new reading circuit or by reusing an existing mechanism? We address this question in the setting of full-sequence OCR, rather than local-answer retrieval. Using an evidence-grounded protocol with held-out causal interventions, we identify sparse and stable OCR-head sets in GLM-OCR, MinerU2.5, and PaddleOCR-VL-1.6. We then investigate the mechanistic origin of these OCR heads by comparing them with independently identified textual retrieval/copy heads in general VLMs. Across two general VLMs, visual OCR heads strongly overlap independently identified textual retrieval/copy heads, yielding untuned top-20 intersections of 73.3% and all-head Spearman correlations of 0.677-0.886. The overlap and causal interventions suggest that full-sequence OCR operates as dense sequential multimodal copy-and-paste, repeatedly retrieving visual evidence and routing it to the current output position. Finally, we examine how this shared circuit changes as a general VLM becomes an OCR specialist. Matched base-to-specialized comparisons show that OCR specialization largely preserves head identity, retaining 17-20 of the top 20 heads per task with all-head rank correlations of 0.874-0.942, while redistributing their functional and causal strengths.
Chinese Translation
一个通用的视觉-语言模型是通过开发新的阅读电路还是通过重用现有机制来获得专业的OCR能力？我们在全序列OCR的背景下而非局部答案检索中探讨这个问题。通过使用基于证据的协议和保留的因果干预，我们在GLM-OCR、MinerU2.5和PaddleOCR-VL-1.6中识别出稀疏且稳定的OCR头集。接着，我们通过将这些OCR头与在通用视觉-语言模型中独立识别的文本检索/复制头进行比较，探讨这些OCR头的机制来源。在两个通用视觉-语言模型中，视觉OCR头与独立识别的文本检索/复制头之间存在强重叠，未调优的前20个交集的重叠率为73.3%，所有头的Spearman相关性为0.677-0.886。这种重叠和因果干预表明，全序列OCR作为密集的顺序多模态复制与粘贴操作，反复检索视觉证据并将其路由到当前输出位置。最后，我们考察随着通用视觉-语言模型成为OCR专家，这一共享电路是如何变化的。匹配的基础到专业化比较显示，OCR专业化在很大程度上保留了头的身份，每个任务保留了17-20个前20个头，所有头的排名相关性为0.874-0.942，同时重新分配了它们的功能和因果强度。
View on arXiv Download PDF AI Translation
cs.CV / 46 / 2609.21576

GestureFAR: Streaming Co-Speech Gesture Generation with Flow Autoregression

GestureFAR：基于流自回归的流媒体共语手势生成
Liu, Pinxin, Liu, Haiyang, Luo, Jiahao, Huang, Junhua, Zou, Chunhao, Song, Luchuan
Abstract
Generating natural co-speech gestures from streaming speech is essential for embodied conversational agents, where motion must be produced while a user is still speaking. Recent streaming gesture systems make online generation possible by autoregressing over discrete motion tokens, but this design compresses high-dimensional continuous motion into finite codebooks and can limit the realism and diversity of generated gestures. To preserve both causality and continuous expressiveness, we propose \textbf{GestureFAR}, a flow-autoregressive framework for streaming co-speech gesture generation. First, GestureFAR autoregresses over causal continuous motion latents, using a transformer to model streaming audio-motion context and a per-token flow-matching head to sample the next latent from a continuous distribution. Second, we introduce a head-only flow distillation strategy that freezes the causal backbone and distills the multi-step per-token flow head into a single network evaluation using consistency and distribution-matching objectives. This keeps the model token-causal while removing the main latency bottleneck for live interaction. Experiments on BEAT2 show that GestureFAR significantly improves the quality--latency trade-off among streaming-capable methods, preserving strong gesture quality while enabling real-time token-causal generation. Project Page: https://andypinxinliu.github.io/GestureFAR
Chinese Translation
从流媒体语音中生成自然的共语手势对于具身对话代理至关重要，因为在用户仍在讲话时必须产生运动。近期的流媒体手势系统通过对离散运动标记进行自回归，使在线生成成为可能，但这种设计将高维连续运动压缩为有限的代码本，可能限制生成手势的真实感和多样性。为了同时保持因果性和连续表现力，我们提出了 extbf{GestureFAR}，一种用于流媒体共语手势生成的流自回归框架。首先，GestureFAR在因果连续运动潜变量上进行自回归，使用变换器建模流媒体音频-运动上下文，并通过每个标记的流匹配头从连续分布中采样下一个潜变量。其次，我们引入了一种仅头部流蒸馏策略，该策略冻结因果主干，并使用一致性和分布匹配目标将多步每个标记的流头蒸馏为单个网络评估。这保持了模型的标记因果性，同时消除了实时交互的主要延迟瓶颈。在BEAT2上的实验表明，GestureFAR显著改善了流媒体可用方法之间的质量-延迟权衡，保持了强大的手势质量，同时实现了实时标记因果生成。项目页面：https://andypinxinliu.github.io/GestureFAR
View on arXiv Download PDF AI Translation
cs.CV / 47 / 2609.21593

A benchmark dataset and baseline methods for four-dimensional STEM diffraction patterns

四维扫描透射电子显微镜衍射图样的基准数据集和基线方法
Guan, Yuyan, Zhang, Haoran, Mao, Zian, Yang, Antong, Li, Caifei, Wang, Jialong, Ouyang, Chuying, Wang, Hong, Zeng, Xiaoqin, Xie, Yujun
Abstract
Four-dimensional scanning transmission electron microscopy (4D-STEM) records a two-dimensional diffraction pattern at each electron-probe position, yielding spatially resolved reciprocal-space information but large, heterogeneous data volumes. Here we describe 4D-ImageNet, a collection of 174,000 diffraction patterns comprising 145,000 experimental patterns selected from 29 acquisitions and 29,000 multislice simulations. The experimental data cover acquisition-level labels for Ag, Au, mixed Au-Ag, CoO, Pd and ZnO specimens across multiple fields of view, scan dimensions, camera lengths and exposure times. Each acquisition contributes 5,000 quality-ranked patterns with source scan coordinates and acquisition metadata. A set-prediction detector provides model-derived pseudo-labels for the direct-beam position and Bragg-disk centres, with a confidence score for each disk. The simulation data cover 13 crystal structures and include Euler rotations, reciprocal-space sampling and approximate low-index beam directions. A grouped mixed-domain masked-reconstruction benchmark is provided to assess leakage-resistant loading and evaluation across experimental and simulated data. The dataset is intended for representation learning, disk detection, diffraction-pattern retrieval, orientation analysis and simulation-to-experiment studies.
Chinese Translation
四维扫描透射电子显微镜（4D-STEM）在每个电子探针位置记录二维衍射图样，从而获得空间分辨的倒易空间信息，但数据量庞大且异质。在此，我们描述了4D-ImageNet，这是一个包含174,000个衍射图样的集合，其中包括从29个采集中选出的145,000个实验图样和29,000个多层次模拟图样。实验数据涵盖了Ag、Au、混合Au-Ag、CoO、Pd和ZnO样品的采集级别标签，涉及多个视场、扫描尺寸、相机长度和曝光时间。每次采集贡献5,000个质量排名的图样，并附有源扫描坐标和采集元数据。一个集合预测检测器为直接束位置和布拉格圆盘中心提供模型推导的伪标签，并为每个圆盘提供置信度评分。模拟数据涵盖13种晶体结构，并包括欧拉旋转、倒易空间采样和近似低指数束方向。提供了一种分组混合域掩蔽重建基准，以评估实验和模拟数据之间的抗泄漏加载和评估。该数据集旨在用于表示学习、圆盘检测、衍射图样检索、取向分析和模拟与实验研究。
View on arXiv Download PDF AI Translation
cs.CV / 48 / 2609.21597

HAT: Hypothesis-Anchored Tracking for Video Monocular Spacecraft Pose Estimation

HAT：基于假设锚定的单目航天器姿态估计跟踪
Lopo, André, Dehban, Atabak, Ventura, Rodrigo
Abstract
Monocular 6-DoF pose estimation of non-cooperative targets is important for on-orbit servicing and debris removal. A single-image estimator can confuse near-symmetric spacecraft orientations, and tracking can preserve an incorrect pose. We present Hypothesis-Anchored Tracking (HAT), a causal framework that uses inter-frame motion to select among competing CAD-based pose hypotheses before alignment and fusion. Rather than independently choosing the highest-scoring hypothesis in each image, HAT retains competing orientation histories and selects a pose to anchor the relative trajectory estimated by monocular SLAM. Sparse anchors and pose fusion provide per-frame estimates after initialization without revising past outputs. The method requires only a calibrated RGB sequence, a metric CAD model, and target image regions, which can be supplied by detection or segmentation. The pretrained pose and SLAM networks require no target-specific training or fine-tuning. We evaluate two versions, Mega-HAT and Pico-HAT, using MegaPose and PicoPose, on SPARK-2024, SwissCube and SHIRT, with YCB-Video assessing performance outside the space domain. Using one temporal configuration per method, the arithmetic means of the four dataset-wise comparisons show 9.4% lower mean pose error and 3.76 times the sustained input FPS for Mega-HAT relative to independent MegaPose, and 23.9% lower mean pose error and 2.42 times the FPS for Pico-HAT relative to independent PicoPose. Mega-HAT ablations on SPARK and an offline reference examine component contributions and the effect of revising past estimates.
Chinese Translation
非合作目标的单目6自由度姿态估计对于在轨服务和碎片清除至关重要。单图像估计器可能会混淆近对称的航天器朝向，而跟踪可能会保持不正确的姿态。我们提出了假设锚定跟踪（Hypothesis-Anchored Tracking, HAT），这是一个因果框架，利用帧间运动在对齐和融合之前选择竞争的基于CAD的姿态假设。HAT并不是在每个图像中独立选择得分最高的假设，而是保留竞争的朝向历史，并选择一个姿态来锚定由单目SLAM估计的相对轨迹。稀疏锚点和姿态融合在初始化后提供每帧的估计，而无需修正过去的输出。该方法仅需要一个经过校准的RGB序列、一个度量CAD模型和目标图像区域，这些可以通过检测或分割提供。预训练的姿态和SLAM网络不需要特定于目标的训练或微调。我们使用MegaPose和PicoPose评估了两个版本，Mega-HAT和Pico-HAT，在SPARK-2024、SwissCube和SHIRT上进行测试，并使用YCB-Video评估在空间领域外的性能。每种方法使用一种时间配置，四个数据集比较的算术平均显示，Mega-HAT相对于独立的MegaPose具有9.4%的较低平均姿态误差和3.76倍的持续输入FPS，而Pico-HAT相对于独立的PicoPose具有23.9%的较低平均姿态误差和2.42倍的FPS。对SPARK的Mega-HAT消融实验和离线参考考察了组件贡献及修正过去估计的影响。
View on arXiv Download PDF AI Translation
cs.CV / 49 / 2609.21624

Learned Parametric Emotion Editing: Real-Time Affective Filtering for On-Device Social Media Video

学习的参数情感编辑：针对设备的社交媒体视频的实时情感过滤
Rochi, Musa, Schubert, Marcel, Gebhardt, Christoph
Abstract
Problematic internet use affects a growing share of the population, yet common interventions, e.g., time limits, blocking, forced breaks, are coercive and easily circumvented. We explore a less restrictive alternative: adapting the emotional intensity of visual content. Prior work has shown that optimization can steer an image's affective content, but its per-image optimization cost makes it impractical for real-time deployment. We instead learn a model that predicts this transformation in a single forward pass: a MobileNetV4 backbone with FiLM-based emotion conditioning outputs parameters for differentiable global transformations. This replaces prior iterative optimization (80 s per image) with a single 3.7 ms forward pass. In a user study (N = 54), the model reduced viewer-reported arousal relative to unedited images, comparably to the grayscale well-being filter, while being rated higher in perceived quality. We integrate the model into an Android app that adapts Instagram video in real time, sustaining 60 fps on a Samsung Galaxy S23.
Chinese Translation
有问题的互联网使用影响着越来越多的人群，而常见的干预措施，如时间限制、屏蔽和强制休息，往往具有强制性且容易被规避。我们探索了一种限制较少的替代方案：调整视觉内容的情感强度。先前的研究表明，优化可以引导图像的情感内容，但其每张图像的优化成本使得实时部署变得不切实际。相反，我们学习了一个模型，该模型在单次前向传播中预测这种转换：一个基于 MobileNetV4 的骨干网络，结合 FiLM（Feature-wise Linear Modulation）情感调节，输出可微分全局变换的参数。这将之前每张图像需80秒的迭代优化替换为单次3.7毫秒的前向传播。在一项用户研究中（N = 54），该模型相较于未编辑图像减少了观众报告的唤醒度，效果与灰度幸福感过滤器相当，同时在感知质量上获得了更高的评分。我们将该模型集成到一个安卓应用中，能够实时调整 Instagram 视频，在三星 Galaxy S23 上保持60帧每秒的流畅度。
View on arXiv Download PDF AI Translation
cs.CV / 50 / 2609.21628

Detection is solved, delineation is not: what governs tooth segmentation on panoramic radiographs

检测已解决， delineation 尚未解决：是什么决定了全景放射影像上的牙齿分割
Rehan, Muhammad, Amjad, Moaz, Ahmed, Syed Danial, Adnan, Mariam, Ali, Haider
Abstract
Automatic tooth segmentation and FDI numbering on panoramic radiographs underpins computer-assisted dental diagnosis, yet which factors govern performance remains unclear. We assemble a corpus of 1,422 panoramic radiographs containing 42,142 expert-delineated tooth polygons across the 32-class FDI taxonomy, annotated by 30 dental practitioners and independently reviewed by two others, and use it to isolate input resolution, architecture and anatomical priors under a single evaluation protocol. First, resolution dominates: across a controlled 640/1024/1280 ablation, mask mAP50-95 rises 0.656 -> 0.710 -> 0.717 while mAP50 stays flat at ~0.982. Both gains are significant under a paired bootstrap over images (p < 0.001, p = 0.024); neither mAP50 change is distinguishable from zero. Added resolution buys boundary precision, not detection. Second, architecture is nearly irrelevant in-domain: a query-based transformer with 2.1x the parameters is statistically equivalent to a one-stage detector (95% CI [-0.0064, +0.0064]), only marginally better under domain shift, 5.5x slower on CPU and not executable under standard ONNX runtimes. Third, three targeted interventions fail: a LoRA-adapted self-supervised encoder underperforms, a promptable foundation segmenter degrades masks by 39%, and globally optimal anatomical label assignment yields +0.0007 despite correcting a constraint violated in 40% of out-of-domain predictions. Zero-shot transfer to an independent multi-centre cohort, verified overlap-free, costs 62% of mask mAP50-95 but only 18% of mAP50, reproducing the dissociation. Decomposing masks along the tooth axis localises the residual error to the apical third. Boundary precision is therefore the binding constraint, and effort is better directed at resolution and acquisition diversity than at architectural novelty.
Chinese Translation
全景放射影像上的自动牙齿分割和 FDI 编号是计算机辅助牙科诊断的基础，但影响性能的因素仍不明确。我们汇集了 1,422 张全景放射影像，包含 42,142 个专家划定的牙齿多边形，覆盖 32 类 FDI 分类，由 30 名牙科从业者注释，并由另外两名独立审阅。我们使用该数据集在单一评估协议下隔离输入分辨率、架构和解剖先验。首先，分辨率占主导地位：在控制的 640/1024/1280 分辨率下，mask mAP50-95 从 0.656 上升至 0.710，再到 0.717，而 mAP50 保持在 ~0.982。通过成对自助法对图像进行统计检验，这两个增益均显著（p < 0.001, p = 0.024）；mAP50 的变化与零无显著区别。增加分辨率提高了边界精度，而非检测性能。其次，架构在该领域几乎无关紧要：一个参数数量为 2.1 倍的基于查询的变换器在统计上与单阶段检测器等效（95% CI [-0.0064, +0.0064]），在领域迁移下仅略有改善，CPU 上速度慢 5.5 倍，且无法在标准 ONNX 运行时下执行。第三，三项针对性干预未能奏效：一个经过 LoRA 调整的自监督编码器表现不佳，一个可提示的基础分割器使得掩膜下降了 39%，而全局最优解剖标签分配尽管修正了 40% 的域外预测中违反的约束，仅提高了 +0.0007。对一个独立的多中心队列进行零样本迁移，验证无重叠，导致 mask mAP50-95 降低 62%，但 mAP50 仅降低 18%，重现了这种分离现象。沿牙齿轴分解掩膜将残余误差局限于根尖三分之一。因此，边界精度是限制因素，努力应更倾向于提高分辨率和获取多样性，而非追求架构的新颖性。
View on arXiv Download PDF AI Translation
cs.CV / 51 / 2609.21629

Extending Decoupled Attention to Dense Prediction and Masked Training for Multi-Channel Images

将解耦注意力扩展到密集预测和多通道图像的掩蔽训练
Marikkar, Umar, Husain, Sameed, Awais, Muhammad, Atito, Sara
Abstract
Multi-Channel imaging (MCI) data differs fundamentally from natural images, as each channel records a semantically distinct signal rather than a colour band. To adapt vision encoders to MCI data, Multi-Channel Vision Transformers (MC-ViTs) tokenize each channel independently and concatenate the resulting tokens into one sequence, and the channel count is no longer fixed by the architecture. Self-attention is then computed across all channel-patch tokens with no restriction on which channels attend to which, which dilutes the features of individual channels. The Decoupled Vision Transformer (DC-ViT) regulates this by separating updates computed within a channel from updates computed across channels, and by forming a representation per channel before the channels are combined. Its formulation, however, pairs tokens by spatial position, and thus requires the same visible tokens in every channel. Correspondence under independent per-channel masking is recovered by solving a linear assignment between the retained patches of each channel, which allows decoupled attention to be combined with current masked multi-channel training in its standard configuration rather than a restricted one. Across three classification and three segmentation benchmarks spanning fluorescence microscopy, imaging mass cytometry and satellite imaging, including dense prediction at high channel counts, the resulting formulation outperforms the strongest MC-ViT baseline.
Chinese Translation
多通道成像（MCI）数据与自然图像在本质上存在根本差异，因为每个通道记录的是语义上不同的信号，而不是颜色带。为了使视觉编码器适应MCI数据，多通道视觉变换器（MC-ViTs）独立对每个通道进行标记，并将生成的标记连接成一个序列，且通道数量不再由架构固定。然后，在所有通道-补丁标记上计算自注意力，而不限制哪些通道关注哪些通道，这稀释了各个通道的特征。解耦视觉变换器（DC-ViT）通过将通道内计算的更新与通道间计算的更新分开，并在通道合并之前为每个通道形成一个表示，来调节这一点。然而，它的公式通过空间位置配对标记，因此要求每个通道中都有相同的可见标记。通过解决每个通道保留的补丁之间的线性分配，恢复独立的每通道掩蔽下的对应关系，这使得解耦注意力能够与当前的标准配置下的掩蔽多通道训练相结合，而不是受限的配置。在涵盖荧光显微镜、成像质谱细胞术和卫星成像的三个分类和三个分割基准测试中，包括在高通道数下的密集预测，所得到的公式优于最强的MC-ViT基线。
View on arXiv Download PDF AI Translation
cs.CV / 52 / 2609.21651

Configurable Multi-Stage Vision Pipeline for Crop Disease and Pest Diagnosis

可配置的多阶段视觉管道用于作物病害和害虫诊断
Ganesh, Naga, S, Chandrashekar M, Pedapudi, Lakshmi, Singh, Aakash, Singh, Vineet
Abstract
Farmer.Chat is Digital Green's farm advisory service for smallholder farmers. When something looks wrong with a crop, the farmer takes a photograph and sends it, and that photograph is the whole question: no symptom described, no crop named, often no text at all. The service has to determine whether the picture can be used, what crop it shows, and what is wrong with it, from images taken on cheap phones in a field, in poor light and with a moving camera. The system doing this today cannot be adjusted. It has no adjustable thresholds for photograph rejection, crops and problems cannot be added, and there is no confidence cut-off to set. We study about 1.16 million photographs sent to Farmer.Chat from Ethiopia, India, Kenya and Nigeria. The production quality gate rejected 46.8% of the images it judged, over a quarter of those reaching diagnosis returned no crop name, and 35.8% of the labelled problems filed under "disease" are pests, identifiable without the crop. We therefore split the work into three stages: a quality gate (M0), a crop detector (M1), and a disease or pest detector (M2). Route A fills all three with one fine-tuned vision-language model (Qwen3-VL-4B) answering in a single call. Route B fills each with a small specialist model (DaViT, YOLO26). We replace our production GPT-4o quality gate with a small MobileNetV3 gate at 86.9% F1 in 12 ms. On one test set scored the same way for every system, a hierarchical DaViT-Base achieves 95.41% crop accuracy against 91.46% for the production baseline. It also leads on diagnosis and never declines to answer, while every language model in the comparison leaves a large share of rows with no diagnosis. The fine-tuned model retains two capabilities the specialists do not have: one call for all three stages, and a request for a better photograph when the image cannot support an answer.
Chinese Translation
Farmer.Chat 是 Digital Green 为小农户提供的农业咨询服务。当作物出现异常时，农民拍摄一张照片并发送，而这张照片就是整个问题：没有描述症状，没有命名作物，通常也没有任何文本。该服务必须确定照片是否可用、显示的是哪种作物以及存在什么问题，这些都是在田间用廉价手机拍摄、光线较差且相机移动的情况下进行的。目前执行此任务的系统无法进行调整。它没有可调的照片拒绝阈值，作物和问题无法添加，也没有设置置信度的截止值。我们研究了来自埃塞俄比亚、印度、肯尼亚和尼日利亚发送到 Farmer.Chat 的约 116 万张照片。生产质量门拒绝了 46.8% 的图像，其中超过四分之一的诊断结果没有返回作物名称，35.8% 被归类为“疾病”的标记问题实际上是害虫，这些害虫在没有作物的情况下也能被识别。因此，我们将工作分为三个阶段：质量门（M0）、作物检测器（M1）和疾病或害虫检测器（M2）。路线 A 用一个经过微调的视觉-语言模型（Qwen3-VL-4B）在一次调用中填充所有三个阶段。路线 B 则为每个阶段填充一个小型专业模型（DaViT, YOLO26）。我们用一个小型 MobileNetV3 门替换了我们的生产 GPT-4o 质量门，F1 得分为 86.9%，响应时间为 12 毫秒。在一个对所有系统评分相同的测试集中，分层 DaViT-Base 在作物准确率上达到了 95.41%，而生产基线为 91.46%。它在诊断上也表现优异，且从不拒绝回答，而比较中的每个语言模型都留下了大量没有诊断的记录。经过微调的模型保留了专业模型所不具备的两项能力：一次调用涵盖所有三个阶段，以及在图像无法支持答案时请求更好照片的能力。
View on arXiv Download PDF AI Translation
cs.CV / 53 / 2609.21675

DRT: Dense Reasoning Trace for Efficient and Grounded Multimodal Reasoning

DRT：高效且扎实的多模态推理的密集推理轨迹
Xu, Wan, Guo, Yuanfan, Han, Kevin, Chen, LaLa, Zuo, Wangmeng
Abstract
Despite the remarkable progress in Multimodal Large Language Models (MLLMs), prevailing Chain-of-Thought (CoT) paradigms remain confined to the natural-language expression space. Consequently, they inherently incur excessive linguistic overhead, leading to information dilution and weak visual grounding. To address this challenge, we propose Dense Reasoning Trace (DRT), a paradigm that departs from natural-language-centered CoT by expressing reasoning as compact structured traces, which include concise intermediate states with symbolic connectors and disentangle visual observations from logical deductions. First, we introduce the Dense Trace Initialization to internalize the DRT reasoning mode into the model, substantially improving token efficiency while preserving visual evidence. To further enable the model to faithfully capture the logical relations within traces, we propose the Trace-Grounded Reinforcement Learning framework, which builds reference traces through a tri-perspective verification pipeline and employs Trace-Grounded GRPO with structured rewards, encouraging the model to generate concise DRT-style traces with reduced hallucination and stronger logical grounding. Extensive experiments on challenging reasoning benchmarks show that DRT achieves 5.5$\times$ token efficiency improvement while improving 1.3 accuracy points over the Qwen3-VL baseline. These findings suggest that complex multimodal reasoning may not require verbose natural-language traces, opening a more efficient path for next-generation MLLMs. Our code and data are available at: https://github.com/HIT-leaderone/DRT
Chinese Translation
尽管多模态大型语言模型（MLLMs）取得了显著进展，现有的思维链（CoT）范式仍然局限于自然语言表达空间。因此，它们固有地带来了过多的语言开销，导致信息稀释和视觉基础薄弱。为了解决这一挑战，我们提出了密集推理轨迹（DRT），这一范式不同于以自然语言为中心的思维链，通过将推理表达为紧凑的结构化轨迹，包括简洁的中间状态和符号连接器，并将视觉观察与逻辑推理解耦。首先，我们引入了密集轨迹初始化，将DRT推理模式内化到模型中，显著提高了令牌效率，同时保留了视觉证据。为了进一步使模型能够忠实捕捉轨迹中的逻辑关系，我们提出了基于轨迹的强化学习框架，该框架通过三视角验证管道构建参考轨迹，并采用基于轨迹的GRPO（Trace-Grounded GRPO）与结构化奖励，鼓励模型生成简洁的DRT风格轨迹，减少幻觉并增强逻辑基础。在具有挑战性的推理基准上的大量实验表明，DRT在令牌效率上实现了5.5倍的提升，同时在Qwen3-VL基线的基础上提高了1.3个准确率点。这些发现表明，复杂的多模态推理可能不需要冗长的自然语言轨迹，为下一代MLLMs开辟了更高效的路径。我们的代码和数据可在以下链接获取：https://github.com/HIT-leaderone/DRT
View on arXiv Download PDF AI Translation
cs.CV / 54 / 2609.21698

Diffusion-Based Tumor Inpainting for Renal Segmentation under Clinical Data Scarcity

基于扩散的肾脏分割肿瘤修复方法在临床数据稀缺情况下的应用
Sedykh, Ekaterina, Ussanov, Salme, Fedorenko, Dmytro, Fishman, Dmytro
Abstract
Deep learning segmentation of renal tumors requires large annotated datasets, yet clinical deployments typically offer only a handful of tumor-positive cases from the target site. We propose a diffusion-based inpainting framework that synthesizes anatomically plausible renal tumors within healthy CT scans, requiring no additional annotation, and provide the first systematic comparison of 2D, 2.5D, and full 3D (MAISI) synthesis strategies for this task. Training the diffusion model on public data (KiTS23, KIRC) and evaluating nnU-Net segmentation on a internal cohort across three low-data regimes, we find that 2.5D and 3D augmentation substantially reduce false positives (from $\sim$18--20\% to $\sim$3--6\%) while maintaining Dice, whereas 2D provides no consistent benefit. Crucially, the proposed 2.5D method matches full 3D synthesis on every metric at substantially lower computational cost, indicating that local volumetric consistency alone is sufficient for effective augmentation in data- and resource-scarce clinical settings.
Chinese Translation
深度学习对肾脏肿瘤的分割需要大量的标注数据集，但临床应用通常仅提供目标部位少量的肿瘤阳性病例。我们提出了一种基于扩散的修复框架，该框架在健康的CT扫描中合成解剖上合理的肾脏肿瘤，无需额外的标注，并首次系统性比较了2D、2.5D和全3D（MAISI）合成策略。通过在公共数据（KiTS23，KIRC）上训练扩散模型，并在内部队列中评估nnU-Net分割，针对三种低数据环境，我们发现2.5D和3D增强显著减少了假阳性（从约18-20%降至约3-6%），同时保持了Dice系数，而2D方法则没有提供一致的益处。重要的是，所提出的2.5D方法在每个指标上都与全3D合成相匹配，但计算成本显著更低，这表明仅靠局部体积一致性就足以在数据和资源稀缺的临床环境中实现有效的增强。
View on arXiv Download PDF AI Translation
cs.CV / 55 / 2609.21709

SignGPT: Toward LLM-Mediated Sign Language Interaction through Gloss-Free Translation and Generation

SignGPT：通过无注释翻译和生成实现大语言模型介导的手语交互
Li, Ronghui, Dong, Jun, Hu, Zhongyuan, Xu, Zunnan, Zhou, Jun, Chen, Liyuan, Liu, Shuoling, Yan, Jiangpeng, Guo, Jie, Li, Xiu, Bao, Linchao
Abstract
Large language models (LLMs) provide limited support for sign language interaction. Unifying sign language translation (SLT) and generation (SLG) to enable sign language as both input and output can reduce switching between separate models during sign-text interaction. We present SignGPT, a unified, pose-based framework for gloss-free SLT and SLG. SignGPT integrates part-aware hierarchical representations of body, hand, and facial motion into a shared language model and employs asymmetric multi-token prediction and progressive training for bidirectional modeling. We evaluate SignGPT on How2Sign (ASL) and Phoenix-2014T (DGS) through benchmark comparisons, qualitative analyses, and component ablations. An exploratory study with 12 Deaf ASL signers assesses an LLM-mediated sign-to-sign response pipeline, highlighting the potential of unified modeling to support sign language conversation (SLC).
Chinese Translation
大型语言模型（LLMs）对手语交互的支持有限。将手语翻译（SLT）和生成（SLG）统一，使手语既可以作为输入也可以作为输出，可以减少在手语与文本交互过程中切换不同模型的需求。我们提出了SignGPT，一个基于姿态的无注释SLT和SLG的统一框架。SignGPT将身体、手部和面部动作的部分感知层次表示集成到一个共享的语言模型中，并采用不对称多标记预测和渐进训练进行双向建模。我们通过基准比较、定性分析和组件消融评估了SignGPT在How2Sign（ASL）和Phoenix-2014T（DGS）上的表现。一项针对12名聋人ASL手语使用者的探索性研究评估了一个LLM介导的手语到手语响应管道，突显了统一建模在支持手语对话（SLC）中的潜力。
View on arXiv Download PDF AI Translation
cs.CV / 56 / 2609.21712

ZYT-World: A Real-Time Controllable World Model for Closed-Loop Autonomous-Driving Simulation

ZYT-World：一种可实时控制的闭环自主驾驶仿真世界模型
Hu, Boni, Wei, Xiong, Huang, Haoming, Huang, Yong, Wang, Chenbo, Yang, Yi, Wang, Jiancheng, Zhu, Ruicheng, Yang, Zhimin, Liu, Guanglai, Jin, Qiaowan, Wang, Dongzhuo, Kuang, Haiwei, Fan, Jiajun, Wu, Yue, Wei, Jiaxin, Sun, Hao, Yan, Feihong, Bi, Wei, Wang, Kaixuan, Guo, Zichao, Chen, Xiaozhi
Abstract
Generative world models offer controllable and repeatable closed-loop simulation for end-to-end and vision-language-action driving policies, but production deployment exposes three unresolved requirements: faithfully reproducing a mixed fisheye-pinhole rig at native resolutions; reconciling causal, per-timestep interaction with long-horizon stability and low latency; and preserving scene identity when a location is revisited. We present ZYT-World, a single architecture that natively generates four fisheye views with field of view > 180{\deg} and three pinhole views. Projection-specific Plucker adapters encode camera geometry, ego-motion adaptive layer normalization provides global motion control, and a lightweight pixel-aligned layout conditions traffic participants and signals through instance-level boxes, headings and colors. Heterogeneous training combines full-rig geometric coverage with high-resolution detail. Teacher forcing, causal consistency distillation, self-rollout distribution matching distillation, and RigCritic transform a 40-step bidirectional teacher into a one-step, per-latent streaming generator, with RigCritic evaluating the seven-view rig jointly. A 19M-parameter variational autoencoder decoder (TinyVAE), W8A8 quantization, and our inference engine reduce decoding, backbone, and incremental-execution costs, respectively. Finally, cross-trajectory pairs derived from real captures train a plug-in implicit-memory module that preserves place-specific evidence. On the internal multi-view test set, the one-step model retains more than 90% of the teacher's PSNR and SSIM, while FID, FVD, and LPIPS stay within 11% of the teacher. Under the generator-only timing in Figure 2, it is 107.7 times faster than the 40-step bidirectional teacher. TinyVAE decodes 59.8 times faster than Wan. 30s rollouts and cross-trajectory revisits show the intended long-horizon and memory behavior.
Chinese Translation
生成世界模型为端到端和视觉-语言-动作驾驶策略提供了可控且可重复的闭环仿真，但在生产部署中暴露出三个未解决的需求：忠实再现混合鱼眼-针孔设备的原生分辨率；调和因果、逐时间步的交互与长时间稳定性和低延迟；以及在重新访问位置时保持场景身份。我们提出了ZYT-World，这是一种单一架构，能够原生生成四个视场大于180°的鱼眼视图和三个针孔视图。投影特定的Plucker适配器编码相机几何，适应自我运动的层归一化提供全局运动控制，而轻量级像素对齐布局通过实例级框、朝向和颜色条件化交通参与者和信号。异构训练结合了全设备几何覆盖与高分辨率细节。教师强制、因果一致性蒸馏、自我回滚分布匹配蒸馏和RigCritic将一个40步的双向教师转化为一个一步、每潜在流的生成器，RigCritic联合评估七视图设备。在内部多视图测试集上，一步模型保留了教师PSNR和SSIM的90%以上，而FID、FVD和LPIPS保持在教师的11%以内。在图2的仅生成器计时下，它比40步双向教师快107.7倍。TinyVAE的解码速度比Wan快59.8倍。30秒的回滚和跨轨迹重访显示了预期的长时间和记忆行为。
View on arXiv Download PDF AI Translation
cs.CV / 57 / 2609.21743

Balanced Prompt Adaptation against Entropy-Induced Collapse for Test-Time Binary Segmentation

针对熵引起的崩溃的平衡提示适应在测试时的二元分割
Wang, Zhengshan, Webster-Ford, Joshua Charles, Tian, Yifei, Wang, Xinxin, Chen, Long, Ding, Weiping
Abstract
Entropy minimization is a standard objective for test-time adaptation (TTA), but it can fail in imbalanced binary segmentation. Unlike image classification, dense segmentation aggregates thousands of pixel predictions, allowing the larger predicted class to dominate the update, pull minority predictions toward itself, and produce a degenerate mask as predictions saturate and their entropy gradients vanish. We theoretically establish this collapse in a shared-shift model. This analysis motivates Balanced-Anchor Prompt Adaptation (BAPA), which combines two complementary modules. The Class-Balanced Anchors (CBA) module selects high-confidence anchors separately from each predicted class and gives foreground and background equal total loss weight, preventing the larger region from dominating the update. Dynamic Prompt Adaptation (DPA) refreshes these anchors after each prediction update and optimizes only text-side prompt residuals while keeping the vision-language encoders frozen. This prompt-only update refines the foreground-background decision boundary without altering the pretrained dense visual representation. Across experiments from four domains, BAPA achieves the highest mean Dice among the evaluated methods. Factorized ablations further validate the complementary roles of CBA and DPA, supporting balanced prompt adaptation as an effective alternative to entropy minimization for test-time binary segmentation.
Chinese Translation
熵最小化是测试时适应（TTA）的标准目标，但在不平衡的二元分割中可能会失败。与图像分类不同，稠密分割聚合了成千上万的像素预测，使得较大预测类别主导更新，将少数预测拉向自身，并在预测饱和且其熵梯度消失时产生退化的掩膜。我们在共享偏移模型中理论上确立了这种崩溃。这一分析促使我们提出了平衡锚点提示适应（Balanced-Anchor Prompt Adaptation, BAPA），该方法结合了两个互补模块。类别平衡锚点（Class-Balanced Anchors, CBA）模块分别从每个预测类别中选择高置信度锚点，并给予前景和背景相等的总损失权重，防止较大区域主导更新。动态提示适应（Dynamic Prompt Adaptation, DPA）在每次预测更新后刷新这些锚点，并仅优化文本侧提示残差，同时保持视觉-语言编码器不变。此仅提示更新在不改变预训练稠密视觉表示的情况下，细化了前景-背景决策边界。在来自四个领域的实验中，BAPA在评估方法中实现了最高的平均Dice系数。分解消融实验进一步验证了CBA和DPA的互补作用，支持平衡提示适应作为测试时二元分割中熵最小化的有效替代方案。
View on arXiv Download PDF AI Translation
cs.CV / 58 / 2609.21754

SFVO: Decoupled Confidence-Guided Stereo-Flow Visual Odometry with Bidirectional PnP

SFVO：基于解耦置信度引导的立体流视觉里程计与双向PnP
Zhang, Kai, Zhao, Guoyang, Ma, Jun
Abstract
Deep learning-based visual odometry (VO) has achieved significant progress, yet most existing methods focus on a monocular approach, which suffers from scale ambiguity. Stereo VO provides real metric by its nature, but remains less studied in deep learning VO due to its high computational cost and modeling complexity. Recent advances in stereo matching and optical flow estimation have made dense visual correspondence increasingly accurate and reliable, but their complementary geometric information has not been fully exploited for VO. In this paper, we present SFVO, a correspondence-driven stereo VO framework that directly builds upon pretrained stereo matching and optical flow models. SFVO exploits pretrained stereo matching and optical flow models to estimate stereo and temporal correspondences. Instead of learning pose directly from images, SFVO maps learned correspondences into geometric constraints and predicts which points are trustworthy. To improve the reliability of visual correspondence-based geometric constraints, we introduce decoupled confidence maps for rotation and translation. This design better aligns the characteristics of visual correspondence and 6-DoF transformations. Extensive experiments on outdoor and indoor datasets demonstrate that SFVO achieves robust and accurate pose estimation with strong generalization capability. The code will be released.
Chinese Translation
基于深度学习的视觉里程计（VO）已取得显著进展，但大多数现有方法集中于单目方法，这导致了尺度模糊。立体视觉里程计本质上提供了真实的度量，但由于其高计算成本和建模复杂性，在深度学习视觉里程计中仍然研究较少。最近在立体匹配和光流估计方面的进展使得密集视觉对应关系的准确性和可靠性不断提高，但其互补的几何信息尚未被充分利用于视觉里程计。在本文中，我们提出了SFVO，一个基于对应关系的立体视觉里程计框架，直接建立在预训练的立体匹配和光流模型之上。SFVO利用预训练的立体匹配和光流模型来估计立体和时间对应关系。SFVO并不是直接从图像中学习姿态，而是将学习到的对应关系映射到几何约束中，并预测哪些点是可信的。为了提高基于视觉对应关系的几何约束的可靠性，我们引入了用于旋转和位移的解耦置信度图。这一设计更好地对齐了视觉对应关系和6自由度变换的特征。在户外和室内数据集上的大量实验表明，SFVO实现了稳健且准确的姿态估计，并具有强大的泛化能力。代码将会发布。
View on arXiv Download PDF AI Translation
cs.CV / 59 / 2609.21763

Beyond Benchmark Scores: Auditing Medical Vision-Language Models for Chest X-Ray Tuberculosis Screening

超越基准分数：对胸部X光结核筛查的医学视觉-语言模型进行审计
Akhtar, Mushir, Tanveer, M., Arshad, Mohd.
Abstract
A medical model's benchmark score does not establish that the same conclusion holds under a different evaluation. This study tests whether claims about model ranking, score reliability and screening performance survive changes in cohort, prompt, negative spectrum, specified prevalence and operating threshold. We audit three medical vision-language models (BioMedCLIP, CheXficient, and MedSigLIP) and a general-domain OpenCLIP comparator on 12,200 chest radiograph records from four datasets (Montgomery, Shenzhen, TBX11K, and VinDr-CXR). Five fixed prompt families yield 244,000 model--image--prompt scores. No model leads every cohort and reliability criterion. Prompt-family changes alter AUROC in 21 of 48 multiplicity-controlled comparisons. Replacing healthy controls with sick non-tuberculosis controls reduces AUROC by 0.075--0.306 across all four models. On VinDr-CXR, the three medical models distinguish tuberculosis from no-finding controls substantially better than from pneumonia or lung tumor; their AUROC point estimates for both named diseases fall below 0.5. CheXficient has documented VinDr-CXR pretraining exposure, which limits the interpretation of its results. Thresholds chosen for 95\% sensitivity on TBX11K training retain that constraint by point estimate in only four of sixteen target evaluations. A five-seed supervised source model reaches 0.999 AUROC on TBX11K validation but 0.629 on each of two external cohorts. Conservative exclusion of perceptual-overlap candidates narrows this gap without closing it. These retrospective, single-task results show that discrimination, score reliability and threshold retention support different portability claims. Evidence for chest X-ray tuberculosis screening should identify the complete evaluation specification rather than attribute clinical portability to a checkpoint alone.
Chinese Translation
医学模型的基准分数并不能证明在不同评估下得出的结论依然成立。本研究测试了关于模型排名、分数可靠性和筛查性能的声明是否能在队列、提示、阴性谱、指定流行率和操作阈值变化下依然成立。我们对三种医学视觉-语言模型（BioMedCLIP、CheXficient 和 MedSigLIP）以及一个通用领域的 OpenCLIP 比较模型进行了审计，数据来源于四个数据集（Montgomery、Shenzhen、TBX11K 和 VinDr-CXR）的 12,200 份胸部放射影像记录。五个固定提示家族产生了 244,000 个模型-图像-提示分数。没有任何模型在每个队列和可靠性标准中均处于领先地位。提示家族的变化在 48 个多重控制比较中改变了 21 个的 AUROC。用生病的非结核对照替换健康对照使得所有四个模型的 AUROC 降低了 0.075 至 0.306。在 VinDr-CXR 数据集中，这三种医学模型在区分结核与无发现对照方面的表现显著优于区分肺炎或肺肿瘤；它们对这两种命名疾病的 AUROC 点估计均低于 0.5。CheXficient 在 VinDr-CXR 预训练暴露下的表现限制了其结果的解释。为 TBX11K 训练选择的 95% 敏感性阈值在十六个目标评估中仅在四个评估中通过点估计保留了这一限制。一种五种种子监督源模型在 TBX11K 验证中达到了 0.999 的 AUROC，但在两个外部队列中均为 0.629。保守地排除感知重叠候选者缩小了这一差距，但并未消除。这些回顾性、单任务的结果表明，区分能力、分数可靠性和阈值保留支持不同的可移植性声明。胸部X光结核筛查的证据应识别完整的评估规范，而不是仅将临床可移植性归因于某个检查点。
View on arXiv Download PDF AI Translation
cs.CV / 60 / 2609.21770

XCalib Depth-Guided Geometric Optimization for Dense Thermal-Visible Video Registration

XCalib 深度引导的几何优化用于密集热可见视频配准
Godet, Aurelien, Jobert, Gabriel, Mura, Mauro Dalla
Abstract
Image registration is a vital preprocessing step in multimodal perception tasks, including image fusion, object detection, and semantic segmentation. In Advanced Driver- Assistance Systems (ADAS), spatial misalignment between visible (RGB) and infrared (IR) cameras -caused by non-coincident optical axes and field-of-view differences- introduces non-uniform parallax and visual ghosting. Classical keypoint-based methods are restricted to global homographies that fail under dynamic depth, while unconstrained dense flow algorithms lack structural regularization and suffer from temporal instability. In this paper, we propose XCalib, an unsupervised dense thermal-visible registration framework that bridges this gap. Rather than serving as an absolute metric calibration tool, XCalib leverages virtual pinhole camera parameterization strictly as a geometric constraint space. By optimizing effective relative pose and intrinsics alongside predicted monocular metric depth, XCalib restricts the search space of spatial displacements to physically valid projection geometries. Our key contributions are: (1) a novel registration paradigm that uses camera parameterization as an implicit regularizer for dense cross-modal warping; (2) Normalized Edges Correlation (NEC), a robust structural similarity metric tailored to cross- spectral alignment; and (3) extensive quantitative and qualitative evaluations across public ADAS datasets, demonstrating superior temporal stability and alignment accuracy over unconstrained dense flow baselines.
Chinese Translation
图像配准是多模态感知任务中的一个重要预处理步骤，包括图像融合、物体检测和语义分割。在高级驾驶辅助系统（ADAS）中，由于可见（RGB）和红外（IR）摄像头之间的空间不对齐——这主要是由于光轴不重合和视场差异——导致了非均匀的视差和视觉重影。传统的基于关键点的方法受限于全局单应性，在动态深度下无法有效工作，而无约束的密集光流算法缺乏结构正则化，且存在时间不稳定性。在本文中，我们提出了XCalib，一个无监督的密集热可见配准框架，旨在填补这一空白。XCalib并不作为绝对度量标定工具，而是严格利用虚拟针孔相机参数化作为几何约束空间。通过优化有效的相对姿态和内参以及预测的单目度量深度，XCalib将空间位移的搜索空间限制为物理有效的投影几何。我们的主要贡献包括：(1) 一种新颖的配准范式，利用相机参数化作为密集跨模态变形的隐式正则化器；(2) 归一化边缘相关性（Normalized Edges Correlation, NEC），一种针对跨光谱对齐的稳健结构相似性度量；(3) 在公共ADAS数据集上进行的广泛定量和定性评估，展示了相较于无约束密集光流基线的优越时间稳定性和对齐精度。
View on arXiv Download PDF AI Translation
cs.CV / 61 / 2609.21780

PointLAM: Local Attentive Mamba for Efficient Point-based 3D Object Detection

PointLAM：高效点云3D物体检测的局部注意力Mamba
Shang, Xuanming, Zhang, Weijia, Ma, Chao
Abstract
3D object detection from LiDAR point clouds faces a fundamental dilemma: voxel-based methods achieve efficiency at the cost of geometric quantization, while point-based methods preserve fidelity but suffer from prohibitive computational bottlenecks. Specifically, point-based architectures are crippled by slow downsampling strategies (e.g., FPS) and expensive dynamic neighbor queries (e.g., k-NN) coupled with costly continuous interactions. To tackle these systemic inefficiencies, we propose PointLAM, a highly efficient and powerful point-based architecture driven by two synergistic innovations. First, to resolve the downsampling bottleneck, we develop the Laplacian Point Sampler (LPS). LPS employs an implicit discrete Laplacian high-pass filter and Doubly Sorted Sampling to achieve fast, structure-aware foreground preservation. Second, to overcome local modeling latency, we design the Local Hadamard Aggregator (LHA). LHA decouples spatial indexing from feature representation using transient grids, and replaces complex continuous interactions with a Hadamard Gating mechanism for topology-aware, attentive modulation. By coupling this local gating with Bi-Directional Mamba (BDM) layers for global sequence modeling, we formulate the Local Attentive Mamba (LAM) block. Powered by this architecture, PointLAM achieves competitive performance on nuScenes and Waymo for point-based detectors. It rivals highly optimized voxel competitors while requiring a fraction of the computational footprint, demonstrating marked superiority in detecting small instances and handling extreme sparsity. Project page: https://pointlam.github.io/.
Chinese Translation
从LiDAR点云进行3D物体检测面临一个基本的困境：基于体素的方法在几何量化的代价下实现了效率，而基于点的方法则保持了保真度，但遭遇了巨大的计算瓶颈。具体而言，基于点的架构受到缓慢下采样策略（例如，FPS）和昂贵的动态邻域查询（例如，k-NN）以及高成本的连续交互的限制。为了解决这些系统性低效问题，我们提出了PointLAM，这是一种高效且强大的基于点的架构，依赖于两项协同创新。首先，为了解决下采样瓶颈，我们开发了拉普拉斯点采样器（Laplacian Point Sampler, LPS）。LPS采用隐式离散拉普拉斯高通滤波器和双重排序采样，以实现快速、结构感知的前景保留。其次，为了克服局部建模延迟，我们设计了局部哈达玛聚合器（Local Hadamard Aggregator, LHA）。LHA使用瞬态网格将空间索引与特征表示解耦，并用哈达玛门控机制替代复杂的连续交互，以实现拓扑感知和注意力调制。通过将这种局部门控与双向Mamba（Bi-Directional Mamba, BDM）层结合用于全局序列建模，我们形成了局部注意力Mamba（Local Attentive Mamba, LAM）模块。得益于这一架构，PointLAM在nuScenes和Waymo的数据集上实现了竞争力的性能，作为基于点的检测器，其性能可与高度优化的体素竞争者相媲美，同时所需的计算资源仅为其一小部分，显示出在检测小实例和处理极端稀疏方面的显著优势。项目页面：https://pointlam.github.io/
View on arXiv Download PDF AI Translation
cs.CV / 62 / 2609.21800

A Principled Approach to Unsupervised Anomaly Detection

一种原则性的方法用于无监督异常检测
Myles, James, Baugh, Matthew, Müller, Johanna P., Kainz, Bernhard, Li, Yingzhen
Abstract
Traditional unsupervised anomaly detection (UAD) methods are designed to flag or localise deviations from a normative distribution, ignoring the underlying generative mechanisms of the anomalies. Yet the nature of an anomaly is often as important as its presence. We reformulate UAD as a Bayesian inverse problem, in which the objective is to infer the most probable corruption responsible for each observation. Our framework yields a probabilistic anomaly score as the energy of the inferred corruption parameters, and serves as a principled recipe for developing new UAD algorithms. We derive several existing methods as instances of the general framework, each corresponding to the same energy score under different modelling choices. Experimentally, we study the framework's components in a controlled setting, and improve object-class AUROC on the MVTec AD dataset by 2.3% by adapting the underlying corruption model. Finally, we validate the framework on a brain MRI benchmark, achieving strong detection performance while producing estimates of pathology intensity, bias, and geometry. Code is available at https://github.com/jgmyles/inverse-uad.
Chinese Translation
传统的无监督异常检测（UAD）方法旨在标记或定位偏离规范分布的异常，忽略了异常的潜在生成机制。然而，异常的性质往往与其存在同样重要。我们将UAD重新表述为一个贝叶斯逆问题，其目标是推断出导致每个观察结果的最可能的腐败因素。我们的框架产生一个概率异常分数，作为推断出的腐败参数的能量，并作为开发新UAD算法的原则性方法。我们将几个现有方法推导为一般框架的实例，每个实例在不同的建模选择下对应相同的能量分数。在实验中，我们在受控环境中研究框架的组成部分，并通过调整基础腐败模型在MVTec AD数据集上将对象类别的AUROC提高了2.3%。最后，我们在脑部MRI基准测试中验证了该框架，取得了强大的检测性能，同时生成病理强度、偏差和几何形状的估计。代码可在 https://github.com/jgmyles/inverse-uad 获取。
View on arXiv Download PDF AI Translation
cs.CV / 63 / 2609.21804

VideoReloc: Long-Term Indoor Video Relocalization against a Kilobyte-Scale Semantic Scene Graph

VideoReloc：针对千字节规模语义场景图的长期室内视频重定位
Li, Qianru, Chen, Xuyang, Wang, Xuqin, Zhang, Zhenghao, Luo, Hongyi, Wu, Tao, Cremers, Daniel, Liu, Lu, Zhang, Yanfeng
Abstract
Given a compact semantic scene graph, long-term indoor video relocalization estimates a map-frame trajectory after lighting and furniture changes. Visual methods rely on appearance and become unreliable under these changes; localizing one frame at a time from object classes and geometry instead leaves sparse, ambiguous evidence. We introduce VideoReloc, whose adaptive clips use odometry to gather spatial evidence until object and motion criteria are met, adapting query length to the observed scene. Its run-level decision rechecks conflicting placements using evidence accumulated across connected clips, stabilizing the trajectory beyond adjacent-clip tracking. Hypothesis-first registration proposes poses from object triplets and verifies each using clip-wide object centers and box surfaces. Orientation-aware refinement uses box faces, gravity and wall directions to resolve ambiguity in camera orientation and refine the full pose. This reframes sparse-map relocalization as verification of spatially extended video queries, moving discriminative support from stored appearance to temporal context and permitting a 100 kB map of class-labelled boxes. On RIO10 and ReplicaCAD, the all-frame localization success rate at 1 m/10$^\circ$ is 73.5% and 61.1% under causal evaluation, rising to 90.6% and 74.8% with clip closure. The evaluated per-frame scene coordinate regressors reach up to 47.6% and 49.8%, respectively, with maps of 12.6-42 MB. Project page: https://videoreloc.github.io
Chinese Translation
给定一个紧凑的语义场景图，长期室内视频重定位在光照和家具变化后估计地图-帧轨迹。视觉方法依赖于外观，在这些变化下变得不可靠；而从物体类别和几何形状逐帧定位则留下稀疏且模糊的证据。我们提出了VideoReloc，其自适应剪辑利用里程计收集空间证据，直到满足物体和运动标准，查询长度根据观察到的场景进行调整。其运行级决策通过使用在连接剪辑中积累的证据重新检查冲突的放置，稳定轨迹超越相邻剪辑跟踪。假设优先注册从物体三元组提议姿态，并使用剪辑范围内的物体中心和盒面验证每个姿态。方向感知的细化利用盒面、重力和墙面方向来解决相机方向的模糊性并细化完整姿态。这将稀疏地图重定位重新框架为对空间扩展视频查询的验证，将判别支持从存储的外观转移到时间上下文，并允许使用100 kB的类标签盒地图。在RIO10和ReplicaCAD数据集上，1 m/10$^ ext{°}$下的全帧定位成功率分别为73.5%和61.1%，在因果评估下，随着剪辑闭合，该成功率上升至90.6%和74.8%。评估的每帧场景坐标回归器分别达到47.6%和49.8%，对应的地图大小为12.6-42 MB。项目页面：https://videoreloc.github.io
View on arXiv Download PDF AI Translation
cs.CV / 64 / 2609.21822

Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty

目标检测基准测试不完整：标签错误和注释不确定性的作用
Penquitt, Sarina, Klees, Jonathan, van Betteray, Antonia, Jashnieh, Parssa, Stehr, Peter, Rottmann, Matthias, Schmarje, Lars
Abstract
While object detection has advanced through improved architectures and open-vocabulary models, we provide strong evidence that benchmark quality is limited by annotation incompleteness. Across four widely used datasets (COCO, Pascal VOC, Cityscapes, KITTI), re-annotation reveals substantial increases in annotated objects (e.g., up to +60% on KITTI and +40% on COCO), driven primarily by previously unlabeled small, occluded, or densely packed instances. While some differences arise from dataset-specific annotation conventions, we consistently find that missing annotations are the main source of label errors across all datasets. To achieve high data quality, we introduce a scalable annotation pipeline that emphasizes high recall and captures ambiguity through soft labels aggregated from at least 11 annotators per object. The resulting annotations improve coverage and align well with human calibration. We show that benchmark performance is highly sensitive to annotation quality, although model rankings remain largely stable. We introduce two large-scale benchmarks: (i) an uncertainty-aware object detection benchmark, and (ii) a label error detection benchmark grounded in real label errors. We show that current detectors are strongly depended on annotation quality and are misaligned with human perception. Current label error detection methods, which have been shown to perform well on synthetic noise, struggle to achieve high recall and precision on real label errors. Our results highlight the need for future object detection benchmarks to move beyond deterministic annotations toward high-recall, uncertainty-aware evaluation that maximizes valid instances and better reflects real-world ambiguity.
Chinese Translation
尽管目标检测通过改进的架构和开放词汇模型取得了进展，但我们提供了强有力的证据表明基准测试的质量受到注释不完整性的限制。在四个广泛使用的数据集（COCO、Pascal VOC、Cityscapes、KITTI）中，重新注释揭示了标注对象的显著增加（例如，KITTI上增加了多达60%，COCO上增加了40%），主要是由于之前未标记的小型、被遮挡或密集堆叠的实例。虽然一些差异源于数据集特定的注释规范，但我们始终发现缺失的注释是所有数据集中标签错误的主要来源。为了实现高数据质量，我们引入了一种可扩展的注释流程，强调高召回率，并通过每个对象至少11名注释者聚合的软标签捕捉模糊性。最终的注释提高了覆盖率，并与人类校准良好对齐。我们展示了基准性能对注释质量高度敏感，尽管模型排名基本保持稳定。我们引入了两个大规模基准测试：（i）一个关注不确定性的目标检测基准，以及（ii）一个基于真实标签错误的标签错误检测基准。我们表明，当前的检测器在很大程度上依赖于注释质量，并与人类感知不一致。当前的标签错误检测方法在合成噪声上表现良好，但在真实标签错误上难以实现高召回率和精确度。我们的结果强调了未来目标检测基准需要超越确定性注释，朝着高召回率、不确定性意识的评估发展，以最大化有效实例并更好地反映现实世界的模糊性。
View on arXiv Download PDF AI Translation
cs.CV / 65 / 2609.21866

Morphology-Aware Ambiguity Learning for Wafer Defect Decision Support

考虑形态的晶圆缺陷决策支持模糊学习
Chu, Seungjun, Chung, Seokhyun
Abstract
Wafer map defect recognition is commonly formulated as a fixed-taxonomy classification problem that assigns each wafer to a single defect class. However, some wafers exhibit morphologies near class boundaries, for which forcing a single prediction may be less informative than providing plausible diagnostic alternatives. This paper proposes a morphology-aware ambiguity learning framework that supports three diagnostic actions: automatic single-class diagnosis, assisted diagnosis with two plausible defect classes, and full review. Using the radial, angular, and geometric characteristics of training wafer maps, the framework constructs a class-level ambiguity matrix representing defect-class pairs with similar morphology and plausible diagnostic alternatives. It guides the model to learn plausible alternative classes rather than treating all incorrect classes equally. During inference, the matrix determines whether an uncertain prediction can be represented by a meaningful two-class diagnostic set or should be escalated for full review. Experiments on WM-811K show that the proposed framework outperforms conventional approaches in defect recognition and diagnostic decision support, providing meaningful two-class alternatives while reserving full review for cases with unresolved ambiguity. Illustrative cost analyses further show the potential cost advantage of the proposed routing strategy. The diagnostic behavior of the framework remains consistent across different backbone architectures.
Chinese Translation
晶圆图缺陷识别通常被表述为一个固定分类法的分类问题，将每个晶圆分配到单一缺陷类别。然而，一些晶圆在类别边界附近表现出形态特征，强迫单一预测可能比提供合理的诊断替代方案更缺乏信息。本文提出了一种考虑形态的模糊学习框架，支持三种诊断操作：自动单类诊断、使用两个合理缺陷类别的辅助诊断和全面审查。该框架利用训练晶圆图的径向、角度和几何特征，构建一个类级模糊矩阵，表示具有相似形态和合理诊断替代方案的缺陷类别对。它引导模型学习合理的替代类别，而不是将所有错误类别视为相同。在推理过程中，该矩阵决定不确定预测是否可以通过有意义的双类诊断集表示，或者是否应升级为全面审查。在WM-811K上的实验表明，所提出的框架在缺陷识别和决策支持方面优于传统方法，提供有意义的双类替代方案，同时将全面审查保留给存在未解决模糊性的案例。说明性的成本分析进一步显示了所提出的路由策略的潜在成本优势。该框架的诊断行为在不同的主干架构中保持一致。
View on arXiv Download PDF AI Translation
cs.CV / 66 / 2609.21872

Chronosphere: Space-Time Tessellation of Local Climate Experts

Chronosphere：地方气候专家的时空镶嵌
Cher, Daniel, Xing, Eric, Li, Kexing, Wei, Brian, Corley, Isaac, Jacobs, Nathan
Abstract
We introduce Chronosphere, a spatio-temporal neural field that learns representations of climate. A central challenge in geographic representation learning is modeling environmental processes whose spatial and temporal complexity varies widely. Yet existing location encoders typically fix a single level of detail everywhere. Global bases such as spherical harmonics spread capacity uniformly across space and time. Localized bases resolve only predefined regions. Learned tessellations adapt, but are inefficient at representing higher frequencies. Chronosphere unifies these approaches, pairing an adaptive tessellation of learnable sites on the spacetime torus $S^2\times S^1$ with a shared bank of local basis functions. Both where capacity is placed and how much detail each region carries adapt to the data, across space and time. Trained to reconstruct climatology, Chronosphere matches or leads state-of-the-art location encoders across spatial and temporal tasks, with the largest gains under spatial and temporal transfer.
Chinese Translation
我们介绍了Chronosphere，一种学习气候表征的时空神经场。地理表征学习中的一个核心挑战是建模空间和时间复杂性变化广泛的环境过程。然而，现有的位置编码器通常在所有地方固定一个单一的细节水平。全球基函数（如球面调和函数）在空间和时间上均匀分配能力，而局部基函数仅能解析预定义区域。学习的镶嵌能够适应，但在表示高频信息时效率较低。Chronosphere统一了这些方法，将可学习位置的自适应镶嵌与共享的局部基函数库结合在一起。能力的分布位置以及每个区域承载的细节量都根据数据在空间和时间上进行调整。经过气候学重建训练，Chronosphere在空间和时间任务中与最先进的位置编码器相匹配或超越，尤其在空间和时间迁移中获得了最大的提升。
View on arXiv Download PDF AI Translation
cs.CV / 67 / 2609.21879

Benchmarking the Explanatory Quality of Open-Weight Vision-Language Models in Face Recognition

开放权重视觉-语言模型在面部识别中的解释质量基准评估
Colbois, Laurent, Marcel, Sébastien
Abstract
Vision-Language Models (VLMs) have recently been proposed as promising tools for face recognition, as they can produce natural language explanations alongside similarity scores. This capability is considered appealing for face comparisons in forensic contexts, which require decisions to be transparent and auditable. However, existing evaluations of VLMs for that use case focus mostly on recognition accuracy, while the validity of generated explanations remains unquantified. In this work, we introduce a benchmarking framework for VLM-based face recognition that treats explanation quality as a core evaluation axis. We propose two criteria that explanations should satisfy: relevance, i.e., reliance on identity-stable facial features; and faithfulness, i.e., alignment with the visible image content without hallucinated features. We jointly develop a methodology enabling the quantification of relevance and faithfulness of evaluated models, based on constraining model outputs to a structured explanation format that supports automated querying and auditing. Using this framework, we benchmark several families of open-weight VLMs, jointly evaluating face verification accuracy and explanation quality. Our results highlight remaining shortcomings of produced explanations, and emphasize the need for such explanation quality metrics to get a complete picture of model performance. The proposed benchmark and open-source evaluation harness provide a foundation for proper benchmarking and future fine-tuning of explainable face recognition systems.
Chinese Translation
视觉-语言模型（VLMs）最近被提出作为面部识别的有前景工具，因为它们可以生成自然语言解释以及相似性评分。这一能力在法医背景下的面部比较中被认为具有吸引力，因为该背景要求决策透明且可审计。然而，现有的针对该用途的VLM评估主要集中在识别准确性上，而生成解释的有效性仍未被量化。在本研究中，我们引入了一种基于VLM的面部识别基准评估框架，将解释质量视为核心评估轴。我们提出了解释应满足的两个标准：相关性，即依赖于身份稳定的面部特征；以及忠实性，即与可见图像内容的一致性，不包含虚构特征。我们共同开发了一种方法论，使得能够量化被评估模型的相关性和忠实性，基于将模型输出限制为支持自动查询和审计的结构化解释格式。利用该框架，我们对几类开放权重VLM进行了基准评估，同时评估面部验证准确性和解释质量。我们的结果突显了生成解释的剩余不足，并强调了需要这样的解释质量指标以全面了解模型性能。所提出的基准和开源评估工具为适当的基准评估和未来可解释面部识别系统的微调提供了基础。
View on arXiv Download PDF AI Translation
cs.CV / 68 / 2609.21887

Catena: A Comprehensive Software Suite for Large-Scale Connectomics

Catena：一个用于大规模连接组学的综合软件套件
Mohinta, Samia, Gómez-Gálvez, Pedro, Lee, Shi Yan, Franco-Barranco, Daniel, Clayton, Michael, Preibisch, Stephan, Funke, Jan, Cardona, Albert
Abstract
The gold standard datasets for mapping connectomes are electron microscopy volumes of densely labeled neural tissue at nanometer resolution. Yet reconstructing and proofreading neuronal arbors and annotating all synapses requires pipelining multiple software tools that are often fragmented, inconsistently maintained, or proprietary, hindering reproducibility and automation. Here, we introduce Catena, an open-source, comprehensive, developer-centric software suite for connectomics that integrates modules for 3D neuron and organelle segmentation, synapse detection, microtubule tracking, and neurotransmitter inference. Catena organizes its modules in composable, chunk-wise processing pipelines in a completely documented, extensible, and adaptable design. We further reduce compute and ground-truth data requirements with pretrained machine learning models, facilitating fine-tuning. Catena ships fully containerized modules that encapsulate evolving dependencies for consistent execution across workstations and clusters. By consolidating open components, shareable models, and containerized runtimes, Catena delivers a reproducible and scalable approach to mapping cellular connectomes from electron microscopy volumes. Code and documentation: https://github.com/Mohinta2892/catena.git
Chinese Translation
用于绘制连接组的金标准数据集是纳米级分辨率下密集标记神经组织的电子显微镜体积。然而，重建和校对神经树以及注释所有突触需要将多个软件工具进行流水线处理，这些工具往往是碎片化的、维护不一致的或专有的，从而阻碍了可重复性和自动化。在此，我们介绍了Catena，一个开源的、综合的、以开发者为中心的连接组学软件套件，集成了3D神经元和细胞器分割、突触检测、微管跟踪和神经递质推断的模块。Catena以可组合的、分块处理的流水线组织其模块，采用完全文档化、可扩展和可适应的设计。我们进一步通过预训练的机器学习模型减少计算和真实数据的需求，便于微调。Catena提供完全容器化的模块，封装了不断演变的依赖关系，以确保在工作站和集群上一致执行。通过整合开放组件、可共享模型和容器化运行时，Catena提供了一种可重复且可扩展的方法，用于从电子显微镜体积中绘制细胞连接组。代码和文档：https://github.com/Mohinta2892/catena.git
View on arXiv Download PDF AI Translation
cs.CV / 69 / 2609.21903

The Role of Radiometric Features in Cross-Site Leaf-Wood Segmentation of LiDAR Point Clouds

辐射特征在激光雷达点云跨站叶木分割中的作用
Kaharlytskyi, Roman, Robinson, Derek T., Guglielmi, Roberto
Abstract
Leaf-wood segmentation of individual trees from LiDAR point clouds is essential for quantitative structure models (QSMs) used in non-destructive biomass estimation. Existing segmentation methods typically exclude radiometric features (e.g., intensity, return number) to maximize cross-sensor compatibility. We challenge this design choice by evaluating cross-site and cross-platform generalization: training on the public Heidelberg dataset (terrestrial TLS, 1550nm) and testing on a novel dataset from Ontario, Canada (RPA-LS, 905nm). Results show that geometry-only methods - including state-of-the-art deep learning models trained on high-density LiDAR datasets - fail to generalize to the sparse, top-down geometry of aerial scans, achieving F1 scores <= 0.56. Incorporating radiometric features (intensity, return number, number of returns) improves F1 to 0.61, but more critically, increases wood recall by 119% from 0.16 to 0.35. Furthermore, geometry-only approaches often result in fragmented stem and branch components. We find that leveraging radiometric features preserves greater structural connectivity, resulting in more coherent architectures that are better suited for QSM reconstruction. We demonstrate that while geometric patterns are view-dependent and prone to overfitting scan patterns, radiometric features encode physical material properties that generalize across disparate sensors and environments.
Chinese Translation
从激光雷达点云中对单棵树进行叶木分割对于用于非破坏性生物量估计的定量结构模型（QSMs）至关重要。现有的分割方法通常排除辐射特征（例如，强度、返回次数），以最大化跨传感器的兼容性。我们通过评估跨站和跨平台的泛化能力来挑战这一设计选择：在公共海德堡数据集（地面TLS，1550nm）上进行训练，并在来自加拿大安大略省的新数据集（RPA-LS，905nm）上进行测试。结果表明，仅使用几何特征的方法——包括在高密度激光雷达数据集上训练的最先进深度学习模型——无法泛化到稀疏的、从上到下的航空扫描几何，F1分数达到<= 0.56。引入辐射特征（强度、返回次数、返回数量）将F1分数提高至0.61，但更重要的是，将木材召回率从0.16提高了119%至0.35。此外，仅使用几何特征的方法往往导致树干和分支组件的碎片化。我们发现，利用辐射特征可以保持更大的结构连通性，从而形成更连贯的结构，更适合QSM重建。我们证明，尽管几何模式依赖于视角且容易过拟合扫描模式，辐射特征则编码了跨越不同传感器和环境的物理材料属性。
View on arXiv Download PDF AI Translation
cs.CV / 70 / 2609.21938

Info3R: Information-Adaptive Test-Time Training for 3D Reconstruction

Info3R：用于3D重建的信息自适应测试时训练
Baek, Sunghyun, Bae, Hanna, Kwon, Minchan, Kim, Junmo
Abstract
Transformer-based models have recently achieved strong performance on 3D reconstruction from images, and recent works extend them to process video streams in an online manner for real-world deployment. However, existing methods overlook two key signals when handling long image streams: the importance of each incoming frame and the information saturation of the model's internal state. In this paper, we propose Info3R, a novel information-adaptive test-time training method for the online 3D reconstruction. We introduce an information-aware state update that modulates the state update strength based on the redundancy and informativeness of each incoming frame. To restore the state's plasticity -- its capacity to incorporate new observations -- we propose a dynamic state reset, triggered by the cumulative magnitude of state updates and the model's prediction confidence and accompanied by an anchor-to-world alignment. Our method achieves consistent improvements on camera pose estimation, video depth estimation, and 3D reconstruction, while substantially mitigating the performance degradation in the long sequence evaluation. Notably, on KITTI Odometry, our method achieves on average 1.68x lower ATE than LongStream, demonstrating its robustness on extended outdoor sequences.
Chinese Translation
基于Transformer的模型最近在从图像进行3D重建方面取得了强劲的性能，近期的研究将其扩展到以在线方式处理视频流，以便于实际应用。然而，现有方法在处理长图像流时忽视了两个关键信号：每个输入帧的重要性和模型内部状态的信息饱和度。本文提出了Info3R，一种用于在线3D重建的新颖信息自适应测试时训练方法。我们引入了一种信息感知的状态更新机制，根据每个输入帧的冗余性和信息量调节状态更新的强度。为了恢复状态的可塑性——即其整合新观察的能力——我们提出了一种动态状态重置机制，该机制由状态更新的累积幅度和模型的预测置信度触发，并伴随有锚点与世界的对齐。我们的方法在相机位姿估计、视频深度估计和3D重建方面取得了一致的改进，同时显著减轻了长序列评估中的性能下降。值得注意的是，在KITTI Odometry数据集上，我们的方法在平均上比LongStream降低了1.68倍的绝对位姿误差（ATE），展示了其在扩展户外序列上的鲁棒性。
View on arXiv Download PDF AI Translation
cs.CV / 71 / 2609.22040

PRIME: Perception Feedback with Situational Memory Embeddings in VLA Models

PRIME：在视觉-语言-行动（VLA）模型中结合情境记忆嵌入的感知反馈
Deinzer, Erik, Baslan, Naya, Paparusso, Luca, Vaskevicius, Narunas, Knott, Peter, Palmieri, Luigi
Abstract
Current Vision-Language-Action (VLA) models for autonomous driving operate primarily through feedforward inference across the perception--reasoning--planning hierarchy. While modern architectures maintain temporal recurrence within the perceptual module, early perception remains blind to downstream reasoning and navigation goals, processing visual inputs agnostically without prioritizing cues informed by prior decisions. To bridge this gap, this paper introduces PRIME, a learned feedback mechanism that conditions the VLA perceptual queries on a novel Situational Memory. By aggregating latent representations of past perception, reasoning, navigation goals, and predicted behaviors across an L-step window via cross-attention, PRIME enables intent-driven perceptual attention at minimal computational cost, adding only a maximum of 29.7M parameters (0.41% of the 7.3B-parameter base model). Evaluated on the Bench2Drive closed-loop benchmark, PRIME achieves a state-of-the-art Driving Score of 82.47 (+4.73 over ORION) and a Success Rate of 60.00% (+5.38 percentage points), the highest reported Driving Score among published VLAs trained on Think2Drive demonstrations.
Chinese Translation
当前用于自主驾驶的视觉-语言-行动（VLA）模型主要通过感知-推理-规划层次结构中的前馈推理进行操作。尽管现代架构在感知模块中保持了时间递归，但早期感知仍然对下游推理和导航目标视而不见，处理视觉输入时没有优先考虑先前决策所提供的线索。为了解决这一问题，本文提出了PRIME，一种学习的反馈机制，它基于一种新颖的情境记忆对VLA感知查询进行条件化。通过跨注意力聚合过去的感知、推理、导航目标和预测行为的潜在表示，PRIME能够以最小的计算成本实现意图驱动的感知注意，仅增加最多29.7M参数（占7.3B参数基础模型的0.41%）。在Bench2Drive闭环基准测试中评估，PRIME实现了82.47的最新驾驶得分（比ORION高出4.73）和60.00%的成功率（高出5.38个百分点），这是在Think2Drive演示上训练的已发布VLA中报告的最高驾驶得分。
View on arXiv Download PDF AI Translation
cs.CV / 72 / 2609.22060

Traffic Sign Recognition for Autonomous Driving Using Branched YOLOv2 and Geometric Features

基于分支YOLOv2和几何特征的自动驾驶交通标志识别
Rezaei, Arefeh
Abstract
Traffic sign recognition (TSR) is an important perception task for autonomous driving and advanced driver-assistance systems, where a system must both localize traffic signs and determine their semantic classes efficiently. This work presents a TSR system based on YOLOv2 for simultaneous detection and classification. Two complementary modifications are studied. First, YOLOv2 is extended with intermediate prediction layers, forming a branched architecture that can terminate inference early for easy cases and reduce computation time. Both whole-image and cell-wise branching strategies are investigated. Second, geometric information is introduced to reduce classification errors between visually similar signs. An unsupervised Bayesian image-segmentation method produces binary representations that are compared with class-specific geometric templates inside YOLOv2 bounding boxes. This information is used either during inference or as an additional signal during training. A dedicated dataset is constructed by combining GTSDB and GTSRB samples using seamless cloning and controlled image transformations. Experiments cover ten traffic-sign classes, with 3,000 training and 300 test samples. The selected branched architecture reports 0.647 s runtime and 0.680 mAP, compared with 0.6607 s and 0.680 mAP for baseline YOLOv2. Geometric verification during inference increases mAP to 0.713, while the geometric-feature training variant achieves 0.697 mAP with a reported runtime of 0.6608 s.
Chinese Translation
交通标志识别（TSR）是自动驾驶和高级驾驶辅助系统中的一项重要感知任务，系统必须高效地定位交通标志并确定其语义类别。本研究提出了一种基于YOLOv2的TSR系统，用于同时检测和分类。研究了两种互补的改进方法。首先，YOLOv2通过中间预测层进行扩展，形成一种分支架构，可以在简单案例中提前终止推理，从而减少计算时间。研究了整幅图像和单元格分支策略。其次，引入几何信息以减少视觉相似标志之间的分类错误。一种无监督的贝叶斯图像分割方法生成二进制表示，并与YOLOv2边界框内的类别特定几何模板进行比较。这些信息在推理期间使用或作为训练过程中的额外信号。通过无缝克隆和受控图像变换构建了一个专用数据集，结合了GTSDB和GTSRB样本。实验涵盖了十个交通标志类别，共有3000个训练样本和300个测试样本。所选的分支架构报告的运行时间为0.647秒，mAP为0.680，而基线YOLOv2的运行时间为0.6607秒，mAP为0.680。在推理期间进行几何验证将mAP提高到0.713，而几何特征训练变体的mAP达到0.697，报告的运行时间为0.6608秒。
View on arXiv Download PDF AI Translation
cs.CV / 73 / 2609.22069

OmniVBench: A Benchmark and Large-Scale Dataset for Omni Reference-to-Video Generation

OmniVBench：用于全方位参考到视频生成的基准测试和大规模数据集
Li, Wenxue, Guan, Peiyan, Jiang, Haoyang, Cai, Junxian, Liu, Hualuo, Zhang, Chunjie, Guan, Chong, Li, Songlian, Wu, Taiyi, Yu, Yongjian, Zhao, Xiaotong, Zhao, Alan, Liu, Eric, Chen, Xi, Liu, Yu, Zhu, Lei
Abstract
Reference-to-video (R2V) generation is evolving toward increasingly general and versatile reference control, giving rise to the emerging paradigm of omni R2V generation. However, existing benchmarks fall short of these emerging capabilities: their test cases cover limited reference types and compositions, and their evaluation protocols largely assess holistic reference consistency, overlooking whether reference factors are properly preserved, disentangled, and routed. Meanwhile, the high cost of constructing omni R2V training data makes suitable training resources scarce. To address these gaps, we introduce OmniVBench and the Omni-R2V Dataset for evaluating and training omni R2V models. OmniVBench expands R2V evaluation across broader reference types, fine-grained control tasks, and richer reference compositions, covering 7 task families and 18 fine-grained tasks spanning content, motion, style, structure, narrative, and multi-reference settings. We introduce factor-grounded evaluation with 12,172 case-specific checklist items, assessing whether intended reference factors are faithfully preserved, correctly disentangled and bound to their targets, and properly realized according to the instruction. We further introduce the Omni-R2V Dataset, bringing industrial-grade training resources for diverse R2V tasks to the broader research community. Drawing primarily on a large-scale corpus of professional video footage, it comprises 340K processed training samples spanning diverse reference types and multi-reference compositions. We develop task-specific pipelines for reference-target pair construction, offering a practical and scalable recipe for omni R2V data construction. Extensive evaluation of advanced open- and closed-source R2V models reveals clear performance gaps across task families and evaluation dimensions on OmniVBench, highlighting remaining limitations of current R2V models.
Chinese Translation
参考到视频（R2V）生成正朝着越来越通用和多样化的参考控制发展，催生了全方位R2V生成的新兴范式。然而，现有的基准测试未能满足这些新兴能力：它们的测试案例涵盖的参考类型和组合有限，其评估协议主要评估整体参考一致性，忽视了参考因素是否得到了适当的保留、解耦和引导。同时，构建全方位R2V训练数据的高成本使得合适的训练资源稀缺。为了解决这些问题，我们引入了OmniVBench和Omni-R2V数据集，以评估和训练全方位R2V模型。OmniVBench扩展了R2V评估，涵盖更广泛的参考类型、细粒度控制任务和更丰富的参考组合，涵盖7个任务家族和18个细粒度任务，涉及内容、运动、风格、结构、叙事和多参考设置。我们引入了基于因素的评估，包含12,172个特定案例的检查项，评估所需的参考因素是否得到了忠实的保留、正确的解耦并绑定到其目标上，以及是否根据指令得到了适当的实现。我们进一步推出了Omni-R2V数据集，为广泛的R2V任务提供工业级训练资源，惠及更广泛的研究社区。该数据集主要基于大规模的专业视频素材，包含340K个处理过的训练样本，涵盖多样的参考类型和多参考组合。我们开发了特定任务的参考-目标对构建管道，提供了一种实用且可扩展的全方位R2V数据构建方案。对先进的开源和闭源R2V模型的广泛评估揭示了OmniVBench上任务家族和评估维度之间明显的性能差距，突显了当前R2V模型的剩余局限性。
View on arXiv Download PDF AI Translation
cs.CV / 74 / 2609.22083

MintAct: A Unified Visual Agent for Digital Environments

MintAct：数字环境中的统一视觉智能体
Gao, Mingfei, Tian, Rui, Gang, Haiming, Zhai, Bohan, Zhang, Le, Gong, Yuanzheng, Feng, Di, Özsoy, Ege, Ma, Kaixin, Kirthivasan, Vishwesh, Kar, Oğuzhan Fatih, Bachmann, Roman, Larsen, Anders Boesen Lindbo, Dehghan, Afshin
Abstract
We present MintAct, a family of vision-language models that unifies UI grounding, multi-step navigation across mobile, desktop, and web, and visual tool use, trained at 2B, 4B, and 8B scales. Through careful design of our environments, data, and training recipes, MintAct models match the performance of per-domain specialists across all of these capabilities. To enable this, we develop a scalable environment and reinforcement learning (RL) infrastructure. On the environment side, we host hundreds of concurrent instances across heterogeneous per-domain backends, serving both trajectory data collection and online RL. To enable efficient and scalable RL training, an asynchronous framework keeps explicit control over the cross-domain training distribution and remains stable under noisy environment feedback and off-policy drift. Experimental results show that MintAct achieves state-of-the-art performance (48.9 on OSWorld-Verified) across a wide range of benchmarks at comparable model sizes.
Chinese Translation
我们提出了MintAct，一个视觉-语言模型家族，统一了用户界面定位、跨移动、桌面和网络的多步骤导航以及视觉工具使用，训练规模为2B、4B和8B。通过对我们的环境、数据和训练方案的精心设计，MintAct模型在所有这些能力上匹配了特定领域专家的性能。为此，我们开发了一个可扩展的环境和强化学习（RL）基础设施。在环境方面，我们在异构特定领域后端上托管数百个并发实例，服务于轨迹数据收集和在线RL。为了实现高效和可扩展的RL训练，一个异步框架保持对跨领域训练分布的明确控制，并在嘈杂的环境反馈和离策略漂移下保持稳定。实验结果表明，MintAct在各种基准测试中以可比的模型规模达到了最先进的性能（在OSWorld-Verified上为48.9）。
View on arXiv Download PDF AI Translation
人工智能 (Artificial Intelligence)
49
cs.AI / 1 / 2609.20971

RBS-Attention: Radius-Bounded Sparse Prefill for Long-Context Large Language Models

RBS-Attention：用于长上下文大语言模型的半径约束稀疏预填充
Song, Chuxu, Wei, Jiuqi, Peng, Zhencan
Abstract
Long-context large language model inference is increasingly limited by prefill, where dense self-attention processes the entire prompt before generation begins. Sparse block selection can reduce this cost, but a block centroid may hide a highly relevant token among many irrelevant ones. We call this failure mode mean dilution and propose RBS-Attention, a training-free sparse-prefill method with two complementary selection branches. A centroid base branch captures average relevance, while a rescue branch uses the maximum key-block radius and its prompt-, layer-, and head-dependent distribution to identify blocks at risk of underestimation. Independently thresholding the two branches and combining their masks controls the contribution of rescue blocks while preserving regular block-sparse FlashAttention execution. On H100 GPUs, RBS-Attention achieves 20.65$\times$ standalone prefill-attention speedup, 11.92$\times$ vLLM prefill-attention speedup, and 5.97$\times$ end-to-end time-to-first-token speedup at 128K on Qwen3-30B-A3B-Instruct-2507-FP8. On the dense Qwen3-32B model, it obtains 88.65 overall RULER accuracy versus 89.52 for dense attention; LongBench-v2, InfiniteBench, and Video-MME provide additional quality evaluation. Supporting experiments measure actual retention, compare selectors at matched density, and characterize block-size, threshold, and memory behavior. Together, these results support radius-adaptive dual-branch selection as an effective approach to long-context prefill.
Chinese Translation
长上下文大语言模型的推理越来越受到预填充的限制，在生成开始之前，密集自注意力处理整个提示。稀疏块选择可以降低这一成本，但块质心可能会在许多无关的标记中隐藏一个高度相关的标记。我们将这种失败模式称为均值稀释，并提出了RBS-Attention，这是一种无训练的稀疏预填充方法，具有两个互补的选择分支。质心基础分支捕捉平均相关性，而救援分支利用最大关键块半径及其依赖于提示、层和头的分布来识别可能被低估的块。独立阈值化这两个分支并组合它们的掩码，可以控制救援块的贡献，同时保留常规块稀疏FlashAttention的执行。在H100 GPU上，RBS-Attention在Qwen3-30B-A3B-Instruct-2507-FP8上实现了20.65$ imes$的独立预填充注意力加速，11.92$ imes$的vLLM预填充注意力加速，以及5.97$ imes$的端到端首次标记时间加速（128K）。在密集的Qwen3-32B模型上，其整体RULER准确率为88.65，而密集注意力为89.52；LongBench-v2、InfiniteBench和Video-MME提供了额外的质量评估。支持实验测量实际保留，比较在匹配密度下的选择器，并表征块大小、阈值和内存行为。这些结果共同支持半径自适应双分支选择作为长上下文预填充的有效方法。
View on arXiv Download PDF AI Translation
cs.AI / 2 / 2609.20974

Attention-Aware Routing: Coupling Routing and Attention in MoEs

关注感知路由：在混合专家模型中耦合路由与注意力
Kosmopoulou, Despoina, Tsetsilas, Anastasios, Georgiou, Efthymios, Karamanolakis, Giannis, Roy, Swastik, Potamianos, Alexandros
Abstract
In Mixture-of-Experts language models, the router typically selects and weights experts based on the token's hidden state, utilizing limited contextual information. We propose Attention-Aware Routing (AAR), which augments the router with temporal and spectral features extracted from a sliding window of attention weights that represent a summary of the model's contextual state, disentangled from the hidden state. Keeping the base transformer entirely frozen, we train only the routing parameters, isolating routing as the sole variable. AAR improves GSM8K by +3.37 pp over a routing-only SFT baseline on OLMoE. Beyond performance, we show that routing and attention form a coupled circuit: routing changes at layer l propagate through the residual stream to amplify attention sinks at layer l+1, reshaping attention without any direct update to the attention mechanism itself. Further, AAR reduces long diverging generation, with incorrect answers getting shorter, while correct answers remain unchanged in length. Finally, AAR is strongly depth-sensitive: applying it indiscriminately across layers can degrade factual retrieval, whereas mathematical reasoning gains persist when it is introduced deeper in the network. This sensitivity exposes a retrieval--reasoning tension across depth and makes layer-selective AAR a controlled probe of the routing-relevant information carried by attention at different layers.
Chinese Translation
在混合专家语言模型中，路由器通常基于令牌的隐藏状态选择和加权专家，利用有限的上下文信息。我们提出了关注感知路由（Attention-Aware Routing, AAR），该方法通过从表示模型上下文状态摘要的注意力权重滑动窗口中提取的时间和频谱特征增强路由器，这些特征与隐藏状态解耦。在保持基础变换器完全冻结的情况下，我们仅训练路由参数，将路由作为唯一变量。AAR在OLMoE上相较于仅使用路由的SFT基线提高了GSM8K的性能3.37个百分点。除了性能之外，我们还表明路由和注意力形成了一个耦合电路：层l的路由变化通过残差流传播，以增强层l+1的注意力汇聚，从而在不直接更新注意力机制的情况下重塑注意力。此外，AAR减少了长时间的发散生成，错误答案变得更短，而正确答案的长度保持不变。最后，AAR对深度具有强烈的敏感性：在各层中不加区分地应用可能会降低事实检索的效果，而在网络较深处引入时，数学推理的收益依然存在。这种敏感性揭示了深度上的检索-推理张力，使得层选择性AAR成为对不同层中注意力所携带的路由相关信息的受控探测。
View on arXiv Download PDF AI Translation
cs.AI / 3 / 2609.20981

CaLR: Causal Latent Revision for Robust Diffusion Reasoning

CaLR：用于鲁棒扩散推理的因果潜在修正
Cai, Wei, Zhao, Jian, Yuan, Yuchen, Li, Xuelong
Abstract
Autoregressive (AR) models suffer from local greediness, while diffusion language models (DLMs) often lack the strict causal structure required for reasoning. To combine the advantages and overcome the drawbacks of the dual, we propose Causal Latent Revision (CaLR), a framework that reformulates reasoning as constrained latent optimization. By adopting a causal topology matrix (CTM) from an expert model and implicit differentiation, CaLR performs gradient-guided ``thought revision" to enforce logical consistency, enabling dynamic self-correction of intermediate steps during parallel generation. Empirically, CaLR achieves SOTA DLM performance on complex benchmarks, surpassing strong AR baselines and demonstrating superior robustness in constrained tasks like Sudoku.
Chinese Translation
自回归（AR）模型存在局部贪婪的问题，而扩散语言模型（DLMs）通常缺乏推理所需的严格因果结构。为了结合两者的优势并克服其缺陷，我们提出了因果潜在修正（CaLR），这是一个将推理重新表述为受限潜在优化的框架。通过采用来自专家模型的因果拓扑矩阵（CTM）和隐式微分，CaLR 通过梯度引导的“思维修正”来强制逻辑一致性，使得在并行生成过程中能够动态自我修正中间步骤。实证结果表明，CaLR 在复杂基准测试中实现了 SOTA DLM 性能，超越了强大的 AR 基线，并在像数独这样的受限任务中展现出更强的鲁棒性。
View on arXiv Download PDF AI Translation
cs.AI / 4 / 2609.21061

LoRA Enhanced Contrastive Learning with SAS Vision Transformers

基于LoRA增强的对比学习与合成孔径声纳视觉变换器
Zimmerman, Dan, Bobe III, Frank E., McCormack, Amelia L., Cook, Matthew, Vetaw, Gregory D.
Abstract
Automatic target recognition (ATR) with synthetic aperture sonar (SAS) supports advanced naval capabilities, but deep learning is constrained by scarce target imagery, background clutter, and human-in-the-loop assessment. We adapt DINOv3 Vision Transformer (ViT) models to underwater SAS ATR using a three-stage parameter-efficient framework. Stage 1 uses Low-Rank Adaptation (LoRA) while freezing the ViT backbone, bridging the gap between natural-image pretraining and underwater acoustic propagation. Stage 2 uses hard-negative mining to strengthen the decision boundary against acoustic mimics, including rocks and sediment formations resembling man-made targets. Stage 3 uses Supervised Contrastive Learning (SupCon) to separate target and clutter representations. We evaluate at-sea SAS data using a mission-level geographic split, compare all arms at 85 percent test recall, and repeat each comparison over three random seeds. LoRA accounts for the primary effect, increasing area under the precision-recall curve (AUPRC) from 0.300 to 0.679 +/- 0.027 using the same frozen backbone. Rank 4 achieves this result while training only 0.26 percent of weights. Neither refinement stage exceeds its matched control: hard-negative mining changes AUPRC by -0.0045 +/- 0.0119 versus an equal-size random curriculum, and SupCon changes AUPRC by +0.0002 +/- 0.0096 versus the preceding stage. These null results indicate that mining occurred on data the encoder had already fit and that supervised stages had already imposed most target-clutter geometry. One efficient adaptation stage is sufficient; stacked refinement is not.
Chinese Translation
合成孔径声纳（SAS）下的自动目标识别（ATR）支持先进的海军能力，但深度学习受到目标图像稀缺、背景杂波和人机交互评估的限制。我们采用三阶段参数高效框架，将DINOv3视觉变换器（ViT）模型适应于水下SAS ATR。第一阶段使用低秩适应（LoRA），同时冻结ViT主干，弥合自然图像预训练与水下声波传播之间的差距。第二阶段使用困难负样本挖掘，以增强对抗声学伪装物（包括与人造目标相似的岩石和沉积物）的决策边界。第三阶段使用监督对比学习（SupCon）来区分目标和杂波的表示。我们使用任务级地理划分评估海上SAS数据，在85%的测试召回率下比较所有方法，并在三个随机种子上重复每次比较。LoRA占据了主要效果，使用相同的冻结主干将精确率-召回率曲线下面积（AUPRC）从0.300提高到0.679 +/- 0.027。Rank 4在仅训练0.26%的权重的情况下实现了这一结果。没有任何细化阶段超过其匹配的对照组：困难负样本挖掘使AUPRC变化为-0.0045 +/- 0.0119，相较于相同规模的随机课程，而SupCon使AUPRC变化为+0.0002 +/- 0.0096，相较于前一阶段。这些无效结果表明，挖掘发生在编码器已经适配的数据上，监督阶段已经施加了大部分目标-杂波几何。一个高效的适应阶段是足够的；堆叠细化并不必要。
View on arXiv Download PDF AI Translation
cs.AI / 5 / 2609.21096

Detecting Hallucination in LLMs: Tracing the Topological Signatures of Impaired Context Sharing

检测大型语言模型中的幻觉：追踪受损上下文共享的拓扑特征
Jalilifard, Amir, Rocha, Anderson, Wong, Eric, Raimundo, Marcos Medeiros
Abstract
In this work, we examine the topology of information flow patterns within attention graphs to effectively distinguish hallucinated from non-hallucinated responses. We analyze the Forman-Ricci curvature to identify structural patterns indicating information bottlenecks in attention graphs. We then introduce a method that captures both semi-local and global information-flow characteristics of attention heads associated with hallucinated responses. We evaluate our approach extensively across several LLMs and established benchmarks. Empirical results demonstrate that our proposed single-pass approach provides consistent improvements over existing attention-based and multi-response baselines across two hallucination-detection benchmarks, while achieving competitive performance across diverse LLM architectures. Further analysis reveals that impaired context sharing among tokens during causal generation is strongly associated with hallucination occurrences in LLMs. In particular, hallucinated responses are consistently characterized by an over-reliance on self-attention, diffused context retrieval from earlier tokens, or information over-squashing, especially in the final transformer layer.
Chinese Translation
在本研究中，我们考察了注意力图中的信息流模式的拓扑结构，以有效区分幻觉响应与非幻觉响应。我们分析了Forman-Ricci曲率，以识别指示注意力图中信息瓶颈的结构模式。随后，我们提出了一种方法，捕捉与幻觉响应相关的注意力头的半局部和全局信息流特征。我们在多个大型语言模型（LLMs）和既定基准上对我们的方法进行了广泛评估。实证结果表明，我们提出的单遍方法在两个幻觉检测基准上相较于现有的基于注意力的方法和多响应基准提供了一致的改进，同时在多种大型语言模型架构上实现了具有竞争力的性能。进一步分析揭示，在因果生成过程中，令牌之间受损的上下文共享与大型语言模型中的幻觉发生有着强烈关联。特别是，幻觉响应通常表现为对自注意力的过度依赖、从早期令牌中扩散的上下文检索或信息过度压缩，尤其是在最后的变换器层中。
View on arXiv Download PDF AI Translation
cs.AI / 6 / 2609.21113

Decoupling Internal Representational Changes and Causal Importance in Fine-Tuned Large Language Models

解耦微调大型语言模型中的内部表征变化与因果重要性
Li, Lingfang, Sen, Procheta, Das, Shubham, Bollegala, Danushka
Abstract
Fine-tuning has emerged as a widely adopted approach for adapting LLMs to a variety of downstream tasks. However, how it reshapes their internal mechanisms remains poorly understood. To address this, we investigate how fine-tuning alters internal representations in LLMs, including attention patterns and layer-wise activations, and examine whether these changes are linked to task-relevant components identified by EAP (e.g., attention heads and logit-level activations) that drive task performance. We find that EAP-identified components are concentrated within specific layers, indicating a degree of functional localisation in how models internalise task-specific behavior. Notably, the distribution of these components across layers is largely uncorrelated with the layers undergoing the most substantial representational changes during fine-tuning. Furthermore, we observe that overlap in EAP-identified components across tasks does not translate into cross-task performance transfer if the tasks are different in nature (e.g. classification vs. generative tasks). More specifically, fine-tuning on one task can lead to a degradation of performance on another when the two tasks exhibit a high degree of overlap in their EAP-identified components.
Chinese Translation
微调已成为将大型语言模型（LLMs）适应各种下游任务的广泛采用的方法。然而，微调如何重塑其内部机制仍然不甚了解。为此，我们研究了微调如何改变LLMs中的内部表征，包括注意力模式和层级激活，并检查这些变化是否与EAP（例如，注意力头和logit级激活）识别的与任务相关的组件相关，这些组件驱动任务性能。我们发现，EAP识别的组件集中在特定层内，表明模型在内化任务特定行为时存在一定程度的功能定位。值得注意的是，这些组件在各层之间的分布与在微调过程中经历最显著表征变化的层之间基本无关。此外，我们观察到，跨任务的EAP识别组件重叠并不意味着在任务性质不同（例如，分类任务与生成任务）时会出现跨任务性能转移。更具体地说，在一个任务上进行微调可能会导致在另一个任务上的性能下降，尤其当这两个任务在其EAP识别的组件上存在高度重叠时。
View on arXiv Download PDF AI Translation
cs.AI / 7 / 2609.21139

TinyCeNN-LM: Quality-Gated Conversion of Pretrained Attention with CeNN-Inspired Cellular-Recurrent Layers

TinyCeNN-LM：基于CeNN启发的细胞递归层的预训练注意力的质量门控转换
Mohsenzadegan, Kabeh, Tavakkoli, Vahid, Kyamakya, Kyandoghere
Abstract
Replacing attention in a pretrained language model is a compatibility problem: a plausible substitute may alter representations expected by later layers. TinyCeNN-LM introduces a \emph{quality-gated post-training conversion} framework using CeNN-inspired cellular-recurrent layers with bounded local processing, compact recurrent memory, routing, fusion, and accept-or-rollback validation. Three implementations are studied: Integrated Memory, MemoryFusion, and PDelta3-GDN2-CLVR+Local32. Strict PDelta3 conversion accepts a layer only when representation and NLL criteria pass fixed thresholds. On SmolLM2-135M, layers 0-2 are accepted with cumulative $\Delta\mathrm{NLL}=+0.01209$, while layer 3 is rejected despite acceptable NLL because representation fidelity fails. On Qwen3.5-0.8B, full-attention layers 3, 7, and 11 are accepted with final $\Delta\mathrm{NLL}=+0.02073$. Integrated Memory keeps perplexity within $-0.07\%$ to $+0.93\%$ while reducing total cache by up to $6.01\%$. A sampled 200-item downstream sanity check gives $28.5\%$--$32.0\%$ overall accuracy for converted Qwen releases. The results support conservative, quality-gated structural conversion rather than universal attention replacement or speedup.
Chinese Translation
在预训练语言模型中替换注意力是一种兼容性问题：一个合理的替代方案可能会改变后续层所期望的表示。TinyCeNN-LM 引入了一种 extit{质量门控后训练转换}框架，使用了受 CeNN 启发的细胞递归层，具有有限的局部处理、紧凑的递归记忆、路由、融合和接受或回滚验证。研究了三种实现方式：集成记忆（Integrated Memory）、记忆融合（MemoryFusion）和 PDelta3-GDN2-CLVR+Local32。严格的 PDelta3 转换仅在表示和 NLL 标准通过固定阈值时接受某一层。在 SmolLM2-135M 上，层 0-2 被接受，累积 $ ext{ΔNLL}=+0.01209$，而层 3 尽管 NLL 可接受却因表示保真度未达标而被拒绝。在 Qwen3.5-0.8B 上，完全注意力层 3、7 和 11 被接受，最终 $ ext{ΔNLL}=+0.02073$。集成记忆使困惑度保持在 $-0.07 ext{ extperthousand}$ 到 $+0.93 ext{ extperthousand}$ 之间，同时将总缓存减少至多 $6.01 ext{ extperthousand}$。对 200 项下游样本的合理性检查显示，转换后的 Qwen 版本整体准确率为 $28.5 ext{ extperthousand}$--$32.0 ext{ extperthousand}$。结果支持保守的质量门控结构转换，而非普遍的注意力替换或加速。
View on arXiv Download PDF AI Translation
cs.AI / 8 / 2609.21149

Clinician-Grounded Quality Assurance for AI-Assisted Psychiatric Intake

基于临床医生的AI辅助精神科接诊质量保障
Shi, King, Li, Amanda, Ivey, Jonathan, Wang, Synthia Qia, Gui, Guan, Kim, Hyunseo, Zandi, Peter, Straub, Jason, Taylor, Jacob, Joshi, Ananya
Abstract
Before patients can use AI-assisted psychiatric intake systems, health systems need practical ways to routinely evaluate these tools against their clinical standards for quality assurance. Because clinicians may use different intake styles, evaluation for this task must (1) support comparison across interviewing approaches, (2) minimize clinician burden, and (3) measure clinically relevant performance for health systems deploying these technologies. We present a clinician-grounded evaluation platform built around a memory-augmented patient simulator for open-ended AI interviewing, InterviewPlayground. We created interactive patients using InterviewPlayground with our expert-authored vignettes, constructed a simulated intake platform for the interviews, and designed evaluation modalities relevant to intake. In a pilot of 6 clinicians in a 25-minute assessment compared to a GPT-based LLM intake interviewer, the LLM recovered more of the clinically relevant items embedded in the patient vignettes (88.0% vs. 38.9%), but made more clinical inferences not based on the interview (56.8% vs. 27.8%), and characterized identified safety concerns less often (33.3% vs. 66.7%), setting the stage for deployed quality assurance for this task.
Chinese Translation
在患者能够使用AI辅助的精神科接诊系统之前，医疗系统需要切实可行的方法来定期评估这些工具是否符合其临床质量保障标准。由于临床医生可能采用不同的接诊风格，因此对此任务的评估必须 (1) 支持不同访谈方法之间的比较，(2) 减少临床医生的负担，以及 (3) 测量对部署这些技术的医疗系统具有临床相关性的表现。我们提出了一个基于临床医生的评估平台，该平台围绕一个增强记忆的患者模拟器（InterviewPlayground）构建，用于开放式AI访谈。我们使用InterviewPlayground创建了交互式患者，并结合我们的专家撰写的案例，构建了一个用于访谈的模拟接诊平台，并设计了与接诊相关的评估方式。在对6名临床医生进行的25分钟评估的初步试验中，与基于GPT的LLM接诊访谈者相比，LLM恢复了更多嵌入患者案例中的临床相关项目（88.0% vs. 38.9%），但做出了更多不基于访谈的临床推断（56.8% vs. 27.8%），并且识别安全隐患的频率较低（33.3% vs. 66.7%），为该任务的质量保障部署奠定了基础。
View on arXiv Download PDF AI Translation
cs.AI / 9 / 2609.21157

Can Agents Design Better Chips with a Higher Level Abstraction?

代理能否通过更高层次的抽象设计更好的芯片？
Ding, Zijian, Zou, Yang, Sun, Yizhou, Cong, Jason
Abstract
Large Language Model (LLM) agents are increasingly being explored for chip design, but most existing approaches operate directly at RTL. We ask whether agents can design better chips by leveraging higher-level abstractions. We compare Direct RTL Design, Agent-based HLS Design, Post-Compiler HLS Refinement, and Post-HLS RTL Refinement, and combine Agent-based HLS Design with Post-HLS RTL Refinement as Agent-based HLS with RTL Refinement (AHRR). We use FPGAs as a practical, easy-to-deploy platform for end-to-end evaluation, but note that the design-flow tradeoffs we study are largely independent of the target technology. Across a diverse 11-tasks benchmark suite, AHRR achieves a 2.6$\times$ geometric-mean speedup over Direct RTL Design across our benchmark suite. Case studies show that HLS distills design knowledge into abstractions that agents can leverage, while RTL refinement recovers lower-level optimization opportunities. Together, these results make AHRR a promising workflow for agentic chip design. The code and evaluation artifacts are available at https://github.com/ZijD/AHRR.
Chinese Translation
大型语言模型（LLM）代理在芯片设计中越来越受到关注，但大多数现有方法直接在注册传输级（RTL）进行操作。我们探讨代理是否能够通过利用更高层次的抽象来设计更好的芯片。我们比较了直接RTL设计、基于代理的高层次综合（HLS）设计、编译后HLS优化和编译后HLS RTL优化，并将基于代理的HLS设计与编译后HLS RTL优化结合，形成基于代理的HLS与RTL优化（AHRR）。我们使用FPGA作为一种实用且易于部署的端到端评估平台，但需要注意的是，我们研究的设计流程权衡在很大程度上与目标技术无关。在一个多样化的11个任务基准套件中，AHRR在我们的基准套件中实现了相较于直接RTL设计的2.6倍几何均值加速。案例研究表明，HLS将设计知识提炼为代理可以利用的抽象，而RTL优化则恢复了更低层次的优化机会。综合来看，这些结果使得AHRR成为代理芯片设计的一个有前景的工作流程。代码和评估文档可在 https://github.com/ZijD/AHRR 获取。
View on arXiv Download PDF AI Translation
cs.AI / 10 / 2609.21165

SpecOpt: Contact-Diff Reasoning for Agentic Molecule Optimization Toward Binding Specificity

SpecOpt：面向结合特异性的智能分子优化的接触-扩散推理
Nguyen, Thao, Ji, Heng
Abstract
Off-target protein binding is a major source of adverse effects for small-molecule drugs, yet most structure-based molecular design methods focus on generating selective compounds de novo rather than improving the selectivity of existing, well- characterized drugs. We introduce specificity optimization (SpecOpt), a molecular design task that seeks constrained structural modifications to an existing compound that increase its binding preference for an intended target over known off-targets while preserving its structural identity and drug-like properties. To enable systematic evaluation, we construct a ChEMBL-derived benchmark from compound-target interaction data, identifying intended targets through curated drug-mechanism annotations and off- targets through measured activities. We then develop an agentic framework that docks each compound against its intended target and off-targets, compares the resulting poses through residue-aware atom-protein contacts, and provides these differential interactions to a large language model to propose targeted structural modifications. Candidates are retained only if they satisfy molecular similarity, ADMET, and target-off-target docking selectivity criteria. On 915 compounds, the agent improves the target- off-target binding gap for 84.8% of compounds, shifting the mean gap from -0.72 to +0.47 kcal/mol while maintaining a mean Tanimoto similarity of 0.72 to the starting compounds. Ablation studies identify residue-specific contact information as the critical optimization signal: replacing residue identities with binary contact indicators eliminates improvement on all 29 ablation compounds. These results establish SpecOpt as a distinct molecular design problem and demonstrate residue-aware differential interactions as an effective signal for improving the specificity of existing compounds.
Chinese Translation
非靶向蛋白结合是小分子药物不良反应的主要来源，然而大多数基于结构的分子设计方法侧重于从头生成选择性化合物，而不是提高现有、已充分表征药物的选择性。我们提出了特异性优化（SpecOpt），这是一项分子设计任务，旨在对现有化合物进行受限的结构修改，以增加其对预期靶标的结合偏好，同时保持其结构特征和药物样性质。为了实现系统评估，我们从化合物-靶标相互作用数据构建了一个基于ChEMBL的基准，通过策划的药物机制注释识别预期靶标，并通过测量的活性识别非靶标。然后，我们开发了一个智能框架，将每个化合物对其预期靶标和非靶标进行对接，通过残基感知的原子-蛋白接触比较结果姿态，并将这些差异相互作用提供给大型语言模型，以提出有针对性的结构修改。只有在满足分子相似性、ADMET和靶标-非靶标对接选择性标准的情况下，候选者才会被保留。在915个化合物中，该智能框架提高了84.8%化合物的靶标-非靶标结合差距，将平均差距从-0.72 kcal/mol转变为+0.47 kcal/mol，同时保持与起始化合物的平均Tanimoto相似性为0.72。消融研究表明，残基特异性接触信息是关键的优化信号：用二元接触指示符替换残基身份会消除29个消融化合物的改进。这些结果确立了SpecOpt作为一个独特的分子设计问题，并展示了残基感知的差异相互作用作为提高现有化合物特异性的有效信号。
View on arXiv Download PDF AI Translation
cs.AI / 11 / 2609.21181

Implicit Rule Induction with Test-Time Task Embeddings in ARC-like Tasks

在类似ARC任务中使用测试时任务嵌入的隐式规则诱导
Deliège, Adrien, Beger, Claas, Van Droogenbroeck, Marc, Mitchell, Melanie
Abstract
The Abstraction and Reasoning Corpus and related benchmarks evaluate whether AI models can solve novel reasoning tasks, but often leave unclear whether success reflects inference of the intended underlying rule or reliance on shortcuts. We address this gap by studying test-time task embeddings in Vision ARC (VARC), a model in which a pre-trained backbone is complemented by a trainable embedding representing the transformation rule. In the original VARC, test-time training (TTT) is jointly applied to the backbone and task embedding. Here we introduce a novel two-step TTT protocol: first finetune only the task embedding (Embed-TTT), then freeze it and finetune the backbone. Across ARC-AGI-1, ConceptARC, and two controlled datasets with known rules, Embed-TTT consistently yields improved task embeddings, ones that align better with underlying task rules, improve embedding-based retrieval, and enable accurate linear probing of known rules. Qualitatively, Embed-TTT identifies more semantically meaningful relations between test and train tasks on ARC-AGI-1. We also show that optimizing only task embeddings (less than 0.01% of model parameters) already solves a non-trivial fraction of ARC-AGI-1, ConceptARC, and Mini-ARC tasks, while the full two-step pipeline improves final performance. Finally, we show that Embed-TTT recovers the underlying geometric structure of parametric rules and learns compositional capabilities that enable rule-wise interpolation, but not extrapolation. These findings support a clearer separation between rule induction and rule execution in ARC-like evaluations, motivating benchmarks that better distinguish in-distribution from out-of-distribution rules.
Chinese Translation
抽象与推理语料库（Abstraction and Reasoning Corpus, ARC）及相关基准评估人工智能模型是否能够解决新颖的推理任务，但通常不清楚成功是否反映了对预期基础规则的推理，还是依赖于捷径。我们通过研究视觉ARC（Vision ARC, VARC）中的测试时任务嵌入来填补这一空白，该模型的预训练主干通过一个可训练的嵌入来补充，后者表示变换规则。在原始VARC中，测试时训练（Test-Time Training, TTT）同时应用于主干和任务嵌入。在这里，我们引入了一种新颖的两步TTT协议：首先仅微调任务嵌入（Embed-TTT），然后冻结该嵌入并微调主干。在ARC-AGI-1、ConceptARC和两个具有已知规则的受控数据集上，Embed-TTT始终产生改进的任务嵌入，这些嵌入与基础任务规则更好地对齐，改善了基于嵌入的检索，并使已知规则的线性探测变得准确。从定性上看，Embed-TTT在ARC-AGI-1上识别出测试任务与训练任务之间更具语义意义的关系。我们还展示了仅优化任务嵌入（占模型参数的不到0.01%）已经解决了ARC-AGI-1、ConceptARC和Mini-ARC任务中的非平凡部分，而完整的两步流程则提高了最终性能。最后，我们表明，Embed-TTT恢复了参数规则的基础几何结构，并学习了能够实现规则间插值但不进行外推的组合能力。这些发现支持在类似ARC的评估中更清晰地分离规则诱导与规则执行，激励更好地区分分布内与分布外规则的基准测试。
View on arXiv Download PDF AI Translation
cs.AI / 12 / 2609.21192

AI-GRACE: A Use-Case Operationalization Framework for Agentic AI: From Organizational Objectives and Obligations to Deployment Capabilities and Architecture

AI-GRACE：一个代理人工智能的用例操作化框架：从组织目标和义务到部署能力和架构
Cuneo, John, Chun, David, Khanna, Gaurav
Abstract
Organizations deploying agentic artificial intelligence must determine more than whether a model is trustworthy; they must establish what to validate, control, and observe for a use case to deliver its intended outcome while meeting applicable obligations. This paper proposes AI-GRACE (Agentic Intelligence-Governance, Risk, Assurance, Controls, and Evidence) as a use-case operationalization framework connecting organizational governance with technical implementation. The proposal draws on professional observations and a purposive synthesis of standards and literature, using design science to frame the method contribution and situational method engineering to guide contextual tailoring and reuse. The framework establishes objectives and obligations and then assesses risks in seven proposed domains, including mission and value realization. It derives requirements for assurance before deployment, runtime controls, and evidence, which guide capability qualification, gap assessment, and a logical architecture. An Agent Operating Envelope specifies permitted actions and escalation conditions, while Risk-Aligned Independence Levels (RAIL) summarize the authorized independence. A fictional retail banking application illustrates the method. The contribution is a traceable basis for deciding what an organization must implement, what it already supports, and what remains unresolved. Empirical evaluation must establish whether it improves deployment decisions, efficiency, and reuse.
Chinese Translation
部署代理人工智能的组织必须确定的不仅仅是模型是否可信；他们还必须建立在特定用例中需要验证、控制和观察的内容，以实现预期结果并满足适用的义务。本文提出了AI-GRACE（代理智能治理、风险、保证、控制和证据）作为一个用例操作化框架，将组织治理与技术实施连接起来。该提案基于专业观察和标准与文献的有目的综合，利用设计科学来框定方法贡献，并使用情境方法工程来指导上下文的定制和重用。该框架建立了目标和义务，然后在七个提议的领域中评估风险，包括使命和价值实现。它在部署前推导出保证要求、运行时控制和证据，这些要求指导能力资格、差距评估和逻辑架构。代理操作范围指定了允许的行动和升级条件，而风险对齐独立级别（RAIL）总结了授权的独立性。一个虚构的零售银行应用程序展示了该方法。其贡献在于提供了一个可追溯的基础，以决定组织必须实施的内容、已经支持的内容以及尚未解决的内容。实证评估必须确定它是否改善了部署决策、效率和重用。
View on arXiv Download PDF AI Translation
cs.AI / 13 / 2609.21208

Information-Gain Rewards over Diversity-Pruned Tests: GT-Anchored Verifier Co-Training for Reliable Code Generation

基于多样性修剪测试的信息增益奖励：GT锚定验证器的共同训练用于可靠的代码生成
Nunez, Ana, Najafirad, Peyman
Abstract
Self-play methods that co-train a single language model as both coder and test author promise to move code-generation RL beyond fixed test suites, but they suffer from two coupled pathologies: permissiveness collapse, where pass-rate rewards are maximised by trivial, non-discriminative tests, and concentration bias, where i.i.d. sampled tests cluster on modal inputs and inflate estimator variance. We introduce CoVer (Co-trained Coder and Verifier), a single-policy GRPO framework that addresses both failure modes. First, an information-gain (IG) reward scores each self-generated test by the mutual information between its pass/fail vector and a graded, ground-truth-anchored correctness signal y [0, 1] m, gated by the sign of their covariance so that only positively discriminative tests receive reward. Second, a three-stage diversity-aware selection step prunes a candidate pool to a behaviourally non-redundant suite (invalidity, input-string, execution-profile filtering), raising the effective sample size of the IG estimator at fixed execution budget. On five benchmarks (LiveBench, MBPP, LiveCodeBench, CodeContests, Code-Forces), CoVer raises one-shot pass@1 by +5.8 points at 7B and +7.1 points at 14B over the Qwen2.5-Instruct backbone, and achieves the highest macro-average among all compared methods at both scales. As a drop-in backbone inside the CodeT ranking pipeline, CoVer-7B adds +3.5 points, demonstrating the dual benefit of co-training for both generation and selection.
Chinese Translation
自我对弈方法通过共同训练单一语言模型作为编码器和测试作者，承诺将代码生成的强化学习（RL）推进到超越固定测试套件的阶段，但它们面临两种相互关联的病态：宽容崩溃（permissiveness collapse），即通过平凡的、非区分性的测试最大化通过率奖励，以及集中偏差（concentration bias），即独立同分布（i.i.d.）采样的测试聚集在模态输入上并膨胀估计器方差。我们引入了CoVer（共同训练编码器和验证器），一个单策略的GRPO框架，旨在解决这两种失败模式。首先，信息增益（IG）奖励通过其通过/失败向量与一个分级的、基于真实值的正确性信号y [0, 1] m之间的互信息来对每个自生成的测试进行评分，并通过它们协方差的符号进行门控，以确保只有正向区分的测试才能获得奖励。其次，一个三阶段的多样性意识选择步骤将候选池修剪为行为上非冗余的测试套件（无效性、输入字符串、执行特征过滤），在固定执行预算下提高了IG估计器的有效样本大小。在五个基准测试（LiveBench、MBPP、LiveCodeBench、CodeContests、Code-Forces）上，CoVer在7B模型上将一次性通过率提升了5.8个百分点，在14B模型上提升了7.1个百分点，相较于Qwen2.5-Instruct基础模型，并在两个规模上实现了所有比较方法中的最高宏平均。作为CodeT排名管道中的一个直接替换基础，CoVer-7B增加了3.5个百分点，展示了共同训练在生成和选择两个方面的双重益处。
View on arXiv Download PDF AI Translation
cs.AI / 14 / 2609.21214

Ability-Residual Decoupled Modeling for Affective Cognitive Diagnosis

情感认知诊断的能力残差解耦建模
Zhao, Boyuan, Ye, Meng
Abstract
Cognitive diagnosis infers students' concept mastery from response logs. However, students' responses are not determined by mastery alone: non-cognitive factors such as emotion, engagement, and fatigue can also affect performance. Affective cognitive diagnosis therefore extends conventional cognitive diagnosis by incorporating affective states. Existing methods often assume that the cognitive diagnosis backbone has already explained ability, item, and concept effects, so the remaining errors can be attributed mainly to affect. We argue that this assumption can be insufficient in real educational data: item calibration bias, systematic concept bias, personalized student-concept deviations, and latent student-item matching can form stable cognitive residuals. Without an explicit modeling pathway, these residuals may leak into affective representations, producing affect contamination. To address this problem, we propose an ability-residual decoupled framework for affective cognitive diagnosis. The model first captures unmodeled cognitive residuals through student, item, concept, student-concept, and low-rank student-item components, and then uses an affective module to modulate guess/slip effects. A Q-matrix-constrained concept residual attention mechanism adaptively aggregates only item-relevant concept residuals. Experiments on ASSIST2017, ASSIST2012, ASSIST2009, and Junyi with six cognitive diagnosis backbones show response-prediction gains across the reported comparisons and generally improved affect alignment when affect labels are available. Ablation studies, leakage probes, principal component analysis visualization, long-tail analysis, and case studies further indicate that ability residuals absorb stable cognitive bias, reduce cognitive contamination in the affective branch, and enhance the robustness and predictive accuracy of cognitive diagnosis models.
Chinese Translation
认知诊断通过响应日志推断学生的概念掌握情况。然而，学生的反应并不仅仅由掌握程度决定：情感、参与度和疲劳等非认知因素也会影响表现。因此，情感认知诊断通过结合情感状态扩展了传统的认知诊断。现有方法通常假设认知诊断的基础模型已经解释了能力、项目和概念效应，因此剩余的误差主要可以归因于情感。我们认为这一假设在真实教育数据中可能不足：项目校准偏差、系统性概念偏差、个性化的学生-概念偏差以及潜在的学生-项目匹配可能形成稳定的认知残差。如果没有明确的建模路径，这些残差可能会渗入情感表征，导致情感污染。为了解决这个问题，我们提出了一种用于情感认知诊断的能力残差解耦框架。该模型首先通过学生、项目、概念、学生-概念和低秩学生-项目组件捕捉未建模的认知残差，然后使用情感模块调节猜测/滑动效应。一个受Q矩阵约束的概念残差注意机制自适应地聚合仅与项目相关的概念残差。在ASSIST2017、ASSIST2012、ASSIST2009和Junyi的六个认知诊断基础模型上的实验显示，在报告的比较中响应预测有所提升，并且在情感标签可用时情感对齐普遍改善。消融研究、泄漏探测、主成分分析可视化、长尾分析和案例研究进一步表明，能力残差吸收了稳定的认知偏差，减少了情感分支中的认知污染，并增强了认知诊断模型的鲁棒性和预测准确性。
View on arXiv Download PDF AI Translation
cs.AI / 15 / 2609.21221

A Fully Differentiable Neuro-Soft-Symbolic Framework for Perceptual Task Planning

一种完全可微的神经软符号框架用于感知任务规划
Wei, Hongyan, AbdAlmageed, Wael
Abstract
Perceptual planning tasks require two key capabilities: accurately perceiving uncertain scenes and planning valid action sequences following logical rules. Conventional methods convert perception into discrete symbolic facts and then plan, discarding perceptual uncertainty and severing task-level feedback to perception. We introduce a generic, fully differentiable neuro-soft-symbolic framework that connects visual perception and task planning within a single computational graph. The framework maintains a continuous soft symbolic state, lifts domain rules into a differentiable soft-$T_P$ transition operator, and optimizes action logits over a short planning horizon. Gradients from the planning objective can also update the perception parameters, allowing task-relevant perceptual representations to be refined during planning. On Blocksworld, our method solves 40/40 LatPlan-40 tasks and 596/600 PlanBench-600 tasks, compared with 33/40 for LatPlan and 587/600 for the reasoning-model baseline, while requiring substantially less computation and time. In the perceptual-uncertainty ablation, our method improves the success rate from 59\% with frozen perception to 83\%. We further conduct task-and-motion simulations on Blocksworld scenes, providing an execution-level validation of the compatibility between decoded task plans and downstream robotic motion execution.
Chinese Translation
感知规划任务需要两个关键能力：准确感知不确定的场景和根据逻辑规则规划有效的动作序列。传统方法将感知转换为离散的符号事实，然后进行规划，这样会丢弃感知的不确定性，并切断任务级别对感知的反馈。我们提出了一种通用的、完全可微的神经软符号框架，将视觉感知和任务规划连接在一个计算图中。该框架保持连续的软符号状态，将领域规则提升为可微的软-$T_P$ 转换算子，并在短期规划范围内优化动作对数。来自规划目标的梯度也可以更新感知参数，从而允许在规划过程中细化与任务相关的感知表示。在 Blocksworld 上，我们的方法解决了 40/40 的 LatPlan-40 任务和 596/600 的 PlanBench-600 任务，而 LatPlan 的成功率为 33/40，推理模型基线为 587/600，同时所需的计算和时间显著减少。在感知不确定性消融实验中，我们的方法将成功率从 59\%（固定感知）提高到 83\\%。我们进一步在 Blocksworld 场景中进行任务与运动模拟，提供了解码任务计划与下游机器人运动执行之间兼容性的执行级验证。
View on arXiv Download PDF AI Translation
cs.AI / 16 / 2609.21259

CogGym: Towards Large-Scale Comparative Evaluation of Human and Machine Cognition

CogGym：面向人类与机器认知的大规模比较评估
Ying, Lance, Wu, Jinzhou, Wang, Yingshan Susan, Aarya, Shivam, Buschoff, Luca M. Schulze, Chen, Harry, Collins, Katherine M., de Varda, Andrea, Fu, Shuhao, Houlihan, Sean Dae, Jagadish, Akshay K., Jiang, Guangyuan, Kiegeland, Samuel, Kurumisawa, Tetsu, Liu, Rongzhi, Liu, Ryan, Ma, Ningshan, McGregor, Kathryn, Strittmatter, Younes, Tsvilodub, Polina, Vigly, Jacob Hoover, Wu, Sarah, Xu, Enjie, Yun, Yiling, Allen, Kelsey, Brooke-Wilson, Tyler, Christian, Brian, Fedorenko, Evelina, Frank, Michael C., Franke, Michael, Gao, Tao, Gershman, Samuel J., Hawkins, Robert D., Hu, Jennifer, Jara-Ettinger, Julian, Kleiman-Weiner, Max, Levine, Sydney, Linzen, Tal, Lu, Hongjing, O'Donnell, Timothy, Ong, Desmond C., Piantadosi, Steven T., Saxe, Rebecca, Schulz, Eric, Shu, Tianmin, Sosa, Felix A., Sucholutsky, Ilia, Zhi-Xuan, Tan, Ullman, Tomer, Xu, Fei, Yildirim, Ilker, Zhu, Jian-Qiao, Griffiths, Thomas L., Gerstenberg, Tobias, Smith, Kevin, Tenenbaum, Joshua B.
Abstract
Understanding and modeling human intelligence are parallel goals shared by artificial intelligence (AI) and cognitive science. As AI systems grow increasingly capable, in what ways do model responses resemble human responses, and where do they systematically diverge? The sheer breadth and diversity of the tasks humans can perform and think about pose a challenge for scalable and rigorous comparison between humans and models. We introduce CogGym, a scalable, unified framework grounded in cognitive science for systematically comparing model and human behavior on matched experimental trials. CogGym uses a semi-automated, human-in-the-loop pipeline to standardize diverse experimental paradigms into a task-agnostic Experiment Markup Language (EML), enabling reproducible and faithful comparison at scale. For initial release, we curate and standardize 258 cognitive experiments from 100 papers that focuses on human commonsense reasoning, and evaluate 50 large language models against human responses. We find a clear scaling trend where larger and more recent AI models better reproduce human judgments. Yet AI models' improvement on such common reasoning tasks is considerably slower than the gains observed on formal-reasoning benchmarks like math and coding, and model--human fit remains well below human splithalf reliability ($R^2 = 0.93$ on text, $0.95$ on image, and $0.92$ on video) with the best models achieving $R^2 = 0.59$ on text, $0.58$ on image, and $0.43$ on video experiments. We intend for CogGym to provide a living evaluation framework that continually incorporates new cognitive science experiments to characterize where model behavior resembles human behavior, where it systematically diverges, and how those patterns change as models and experiments evolve.
Chinese Translation
理解和建模人类智能是人工智能（AI）和认知科学共同追求的目标。随着AI系统能力的不断增强，模型的反应在何种方面与人类反应相似，又在何处系统性地偏离？人类能够执行和思考的任务的广度和多样性对人类与模型之间进行可扩展且严格的比较提出了挑战。我们引入了CogGym，这是一个基于认知科学的可扩展统一框架，用于系统地比较模型和人类在匹配实验试验中的行为。CogGym使用半自动化的人机协作流程，将多样的实验范式标准化为任务无关的实验标记语言（Experiment Markup Language, EML），从而实现可重复和真实的规模比较。在初始发布中，我们整理并标准化了来自100篇论文的258个认知实验，重点关注人类的常识推理，并对50个大型语言模型与人类反应进行了评估。我们发现一个明显的规模趋势，即更大和更新的AI模型更好地再现人类的判断。然而，AI模型在这些常识推理任务上的改进速度远低于在数学和编码等形式推理基准上观察到的进展，且模型与人类的拟合度远低于人类的分半可靠性（文本上的$R^2 = 0.93$，图像上的$0.95$，视频上的$0.92$），最佳模型在文本、图像和视频实验中的$R^2$分别为$0.59$、$0.58$和$0.43$。我们希望CogGym能够提供一个持续的评估框架，不断纳入新的认知科学实验，以表征模型行为与人类行为相似的地方、系统性偏离的地方，以及随着模型和实验的演变这些模式如何变化。
View on arXiv Download PDF AI Translation
cs.AI / 17 / 2609.21263

PlaceReasoner-Beta: Reasoning-Driven Macro Placement and Benchmarking

PlaceReasoner-Beta：基于推理的宏单元布局与基准测试
Li, Qiufeng, Wang, Chengxuan, Chen, Rongqian, Cheng, Quan, Ren, Yihui, Ho, Chia-Tung, Pan, David Z., Lan, Tian, Cao, Weidong
Abstract
Automated macro placement remains a fundamental challenge in VLSI physical design. Despite decades of research, existing approaches predominantly optimize hand-crafted proxy objectives, such as estimated wirelength, and typically produce placements through one-shot numerical optimization, limiting their ability to incorporate visual layout context, codified design expertise, and downstream physical-design feedback in a unified loop. We present PlaceReasoner-Beta, a verifier-guided multi-agent framework that reformulates macro placement as a closed-loop reasoning problem rather than black-box optimization. A vision-language model (VLM) planner generates candidate placements from the floorplan image, macro specifications, and connectivity structure; a geometric verifier enforces physical legality and expert placement principles; a physical verifier refines candidates using early implementation feedback; and a post-route optimizer further improves promising layouts using final PPA. To enable reproducible evaluation, we introduce PlaceReasoner-Bench, a fully open end-to-end benchmark built from open RTL designs, EDA tools, and technology libraries. It comprises 8 designs at two aspect ratios, yielding 16 tasks with fixed floorplans and I/O assignments, so methods differ only in macro positions and orientations and are evaluated using routed PPA and DRC rather than pre-route proxies. Across the benchmark, PlaceReasoner-Beta achieves the best timing among DRC-clean methods on all square tasks, reducing post-route TNS by 61.2% at 1:1 and 53.0% at 2:1 relative to the classical baseline field. It also shortens routed wirelength on most designs despite never explicitly optimizing it, demonstrating that reasoning over spatial structure under physical-design feedback can improve end-to-end layout quality beyond proxy-objective optimization.
Chinese Translation
自动化宏单元布局仍然是VLSI物理设计中的一个基本挑战。尽管经过数十年的研究，现有的方法主要优化手工设计的代理目标，例如估算的布线长度，通常通过一次性数值优化生成布局，这限制了它们在统一循环中整合视觉布局上下文、编码设计专业知识和下游物理设计反馈的能力。我们提出了PlaceReasoner-Beta，一个以验证器为指导的多智能体框架，将宏单元布局重新表述为一个闭环推理问题，而不是黑箱优化。视觉-语言模型（VLM）规划器从平面图图像、宏单元规格和连接结构生成候选布局；几何验证器强制执行物理合法性和专家布局原则；物理验证器利用早期实现反馈对候选进行细化；后路优化器进一步利用最终的PPA改善有前景的布局。为了实现可重复的评估，我们引入了PlaceReasoner-Bench，这是一个完全开放的端到端基准，基于开放的RTL设计、EDA工具和技术库构建。它包含8个设计，具有两种纵横比，产生16个任务，固定平面图和I/O分配，因此方法仅在宏单元的位置和方向上有所不同，并且使用路由后的PPA和DRC进行评估，而不是预路由代理。在基准测试中，PlaceReasoner-Beta在所有DRC合规的方形任务中实现了最佳时序，相较于经典基线场，在1:1时减少了61.2%的后路TNS，在2:1时减少了53.0%。尽管从未明确优化布线长度，它在大多数设计中仍缩短了路由后的布线长度，证明在物理设计反馈下对空间结构进行推理可以改善端到端布局质量，超越代理目标优化。
View on arXiv Download PDF AI Translation
cs.AI / 18 / 2609.21267

Efficient Benchmarking in Production: A Study of an Evolving LLM Agent

生产中的高效基准测试：一个不断演变的 LLM 代理的研究
She, Yining, Lin, Lei
Abstract
Production LLM agents are evaluated repeatedly as they evolve, but full agent benchmarks are costly to rerun. We study efficient recurring evaluation for a production analytics agent serving tens of thousands of monthly active users and report first-hand deployment experience. Using 574 historical runs of the production benchmark, split chronologically into calibration and held-out periods, we compare random sampling, historical caching, fixed representative subsets, and IRT-based adaptive testing. The results show that multidimensional 2PL adaptive testing achieves the best overall score fidelity: executing 200 questions, 38.5% of a full run, yields 1.03 pp of MAE. We nevertheless deployed difficulty-stratified fixed subsets because of their operational simplicity, and show they transfer without recalibration to five other agent families and remain stable across calibration windows as short as one day. Drawing on this deployment experience, we report practical recommendations for recurring production-agent evaluation.
Chinese Translation
生产 LLM 代理在其演变过程中会被反复评估，但完整的代理基准测试重新运行成本较高。我们研究了一种高效的重复评估方法，针对一个服务于数万名月活跃用户的生产分析代理，并报告了第一手的部署经验。通过使用574次历史基准测试的运行数据，按时间顺序分为校准期和保留期，我们比较了随机抽样、历史缓存、固定代表性子集和基于 IRT（项目反应理论）的自适应测试。结果表明，多维2PL自适应测试在整体得分保真度上表现最佳：执行200个问题，相当于完整运行的38.5%，获得了1.03个百分点的平均绝对误差（MAE）。尽管如此，我们仍然部署了难度分层的固定子集，因为它们在操作上更为简单，并且显示出在不重新校准的情况下可以迁移到其他五个代理家族，并在短至一天的校准窗口内保持稳定。基于这一部署经验，我们报告了关于重复生产代理评估的实用建议。
View on arXiv Download PDF AI Translation
cs.AI / 19 / 2609.21293

GameASG-Bench: Benchmarking Autonomous Software Generation for Game Development

GameASG-Bench：游戏开发中自主软件生成的基准测试
Zhang, Xiuhui, Chen, Yi, Xu, Shusheng, Li, Fan, Wang, Huan, Yang, Tongkai, Yuan, Binhang
Abstract
Autonomous software generation (ASG) aims to turn human requirements into executable applications, but delivering these applications does not necessarily establish that their interacting components satisfy the specified behavioral requirements. We introduce GameASG-Bench, a benchmark that makes behavioral testability part of the generation task for game development. Our design declares an evaluation interface specification before generation, fixing legal starting scenarios, player-level actions, stable snapshots, rejection behavior, and invariants while leaving private implementations open. Concretely, we include: (i) static L1 checks that assess source-level compliance; and (ii) browser-executed L2 checks that combine semantic observations with real input and runtime evidence. We implement this protocol as 47 browser-native game-generation tasks spanning 12 primary genres and both 2D and 3D interaction, each with executable checks and an independently verified reference implementation. Our experiments answer four key questions about end-to-end agent performance, tool access and nominal turn budget, reasoning effort, and harness choice. Across nine agent stacks, the highest observed mean L2 check pass rate is 93.2%, yet the highest observed strict task success rate, requiring all L1 and applicable L2 prerequisite and core requirement checks, is only 55.3% (26/47 tasks). For DeepSeek-V4-Flash, full tool access and larger nominal turn budgets yield more strict task successes, while the strict task success rate is not monotonic in reasoning effort. Both tested harnesses achieve 18 strict task successes, but only ten tasks succeed under both. These results expose task-level compliance gaps that high average check pass rates actually obscure.
Chinese Translation
自主软件生成（ASG）旨在将人类需求转化为可执行的应用程序，但交付这些应用程序并不一定能确保其交互组件满足指定的行为要求。我们介绍了GameASG-Bench，这是一个将行为可测试性纳入游戏开发生成任务的基准测试。我们的设计在生成之前声明了评估接口规范，固定了合法的起始场景、玩家级别的动作、稳定的快照、拒绝行为和不变式，同时保留私有实现的开放性。具体而言，我们包括：（i）静态L1检查，评估源代码级别的合规性；（ii）浏览器执行的L2检查，将语义观察与真实输入和运行时证据相结合。我们将该协议实现为47个浏览器原生的游戏生成任务，涵盖12个主要类型，涉及2D和3D交互，每个任务都有可执行的检查和独立验证的参考实现。我们的实验回答了关于端到端代理性能、工具访问和名义回合预算、推理努力和工具选择的四个关键问题。在九个代理堆栈中，观察到的最高平均L2检查通过率为93.2%，然而，观察到的最高严格任务成功率（要求所有L1和适用的L2先决条件及核心要求检查）仅为55.3%（26/47任务）。对于DeepSeek-V4-Flash，完全工具访问和更大的名义回合预算带来了更多的严格任务成功，而严格任务成功率在推理努力方面并非单调。两个测试的工具均实现了18个严格任务成功，但只有十个任务在两者下成功。这些结果揭示了任务级别合规性差距，而高平均检查通过率实际上掩盖了这一点。
View on arXiv Download PDF AI Translation
cs.AI / 20 / 2609.21325

LEGIT: Credentialing Protocol for Trustworthy AI Agent Marketplaces

LEGIT：可信AI代理市场的认证协议
Drew, Steve, Zhou, Jiayu
Abstract
Agentic marketplaces are emerging where AI agents with varying capabilities autonomously complete specialized tasks for buyers. A major challenge of such marketplaces is that buyers cannot easily determine which agent will perform best on their tasks. Reported benchmark scores may be difficult to verify or compare across tasks, software, and budgets. We introduce LEGIT, a credentialing protocol connecting certification, reputation, and proposed marketplace allocation. Certification binds measured quality and cost per solved task to an agent configuration, task domain, evaluation budget, and evidence through a signed record. Reputation links records of past task outcomes to the same identity, subject to the reliability of the reported feedback. Buyers and agents can verify credential records and inspect optional visual profiles. Evaluations reveal cost differences between agent configurations with similar observed task success, and show that comparisons depend on the evaluation budget. These results support binding performance measurements to the tested configuration and resource limits. A complementary analysis quantifies the deposits and fees required for reputation manipulation under a stated Sybil attack model.
Chinese Translation
代理市场正在兴起，在这些市场中，具有不同能力的AI代理能够自主为买家完成专业任务。然而，这类市场面临的一个主要挑战是买家无法轻易判断哪个代理在其任务上表现最佳。报告的基准分数在不同任务、软件和预算之间可能难以验证或比较。我们提出了LEGIT，一种将认证、声誉和提议市场分配连接起来的认证协议。认证将测量的质量和每个解决任务的成本绑定到代理配置、任务领域、评估预算和通过签名记录提供的证据上。声誉将过去任务结果的记录与同一身份关联，前提是所报告反馈的可靠性。买家和代理可以验证认证记录并检查可选的视觉档案。评估揭示了在观察到的任务成功相似的代理配置之间的成本差异，并表明比较依赖于评估预算。这些结果支持将性能测量绑定到测试的配置和资源限制上。补充分析量化了在声明的Sybil攻击模型下，声誉操控所需的押金和费用。
View on arXiv Download PDF AI Translation
cs.AI / 21 / 2609.21390

Offline Multimodal Large Language Models for Decision Support in Air Operations

离线多模态大型语言模型在空中行动决策支持中的应用
Dantas, Joao P. A., Cunha, Jelton A., Dietzsch, Gabriel
Abstract
Air operations rely on complex rules, established procedures, and time-critical analysis under limited connectivity and strict security constraints. In such environments, analysts must combine written doctrine with images, often without access to external computing resources. This paper studies offline large language models as decision support tools, deployed in isolated and restricted environments to give analysts access to doctrinal knowledge that remains traceable to its original sources through natural language interaction. We describe a modular retrieval-augmented architecture suitable for operation without Internet connectivity, supporting both text and image input from technical manuals. As a first step toward evaluating this architecture, we report a pilot study with four image analysts of the Brazilian Air Force, combining (i) a doctrinal knowledge assessment based on their electronic-target identification doctrine, comparing human and proposed system performance on the same test, and (ii) a measurement of the cognitive workload involved in manually producing a reconnaissance target report (Relat\'orio de Miss\~ao de Reconhecimento - REMIR) without AI assistance. The results show a demanding manual task, especially in terms of mental demand (6.0/7) and effort (5.0/7), while the proposed system matches the human score (8/10) and completes the assessment in 7.1 minutes (compared to a human average of 26.5 minutes), establishing a baseline for future AI-assisted evaluation. Finally, we describe a future evaluation protocol to systematically compare manual and AI-assisted workflows.
Chinese Translation
空中行动依赖于复杂的规则、既定的程序以及在有限连接和严格安全约束下的时间敏感分析。在这样的环境中，分析师必须将书面教义与图像结合起来，通常无法访问外部计算资源。本文研究了离线大型语言模型作为决策支持工具，在孤立和受限的环境中部署，以便通过自然语言交互为分析师提供可追溯至其原始来源的教义知识。我们描述了一种适合在没有互联网连接的情况下操作的模块化检索增强架构，支持来自技术手册的文本和图像输入。作为评估该架构的第一步，我们报告了一项与巴西空军四名图像分析师的初步研究，结合了(i) 基于他们电子目标识别教义的教义知识评估，比较人类与所提议系统在同一测试中的表现，以及(ii) 测量在没有人工智能辅助的情况下手动生成侦察目标报告（Relatório de Missão de Reconhecimento - REMIR）所涉及的认知负荷。结果显示这是一个要求较高的手动任务，特别是在心理需求（6.0/7）和努力程度（5.0/7）方面，而所提议的系统的得分与人类相匹配（8/10），并在7.1分钟内完成评估（相比之下，人类的平均时间为26.5分钟），为未来的人工智能辅助评估建立了基线。最后，我们描述了一种未来的评估协议，以系统地比较手动和人工智能辅助的工作流程。
View on arXiv Download PDF AI Translation
cs.AI / 22 / 2609.21423

DENSE: Distilling Agent Trajectories into Evidence-Grounded Shortcut Trees for Self-Refinement

DENSE：将代理轨迹提炼为基于证据的快捷树以实现自我优化
Liu, Siyuan, Yu, Fan, Ru, Dongyu, Liu, Yizhu, Yang, Yifan, Cao, Xuezhi, Cai, Xunliang, Cao, Yixin
Abstract
Online agent deployments produce abundant execution traces, while task-specific verification and expert annotation are costly to scale. We study how to distill these traces into reusable feedback without post-hoc outcome labels, drawing on their evidence of local progress, recovery, and unfinished requirements. We introduce DENSE (Distilling Evidence from Nested Subtask Executions), which organizes this evidence into evidence-grounded nested shortcut trees. DENSE compresses redundant attempts, reconciles issues across levels using recovery evidence, and summarizes completed branches while expanding unresolved ones, linking reusable progress to remaining obligations. We introduce REFIT, a source-paired protocol comparing feedback from shared initial trajectories under post-hoc outcome blindness, with environments and model contexts reset for fresh attempts at the same tasks. On Terminal-Bench 2.1, DENSE achieves the highest strict pass rate among tested non-privileged feedback methods across four recipient models. Relative to initial executions, strict pass rate improves by 7.12-15.64 pp, with 19.0-43.6% fewer observed recipient tokens in reruns. GPT-5.5 ablations support combining nested subtask analysis with shortcut construction and issue reconciliation. These findings point toward agent self-refinement through evidence-grounded trajectory reuse with less reliance on external supervision.
Chinese Translation
在线代理部署产生了大量的执行轨迹，而任务特定的验证和专家标注在规模化时成本高昂。我们研究如何在没有事后结果标签的情况下，将这些轨迹提炼为可重用的反馈，利用它们在局部进展、恢复和未完成需求方面的证据。我们提出了DENSE（从嵌套子任务执行中提炼证据），该方法将这些证据组织成基于证据的嵌套快捷树。DENSE压缩冗余尝试，利用恢复证据协调不同层级之间的问题，并在扩展未解决分支的同时总结已完成的分支，将可重用的进展与剩余的义务联系起来。我们引入了REFIT，这是一种源配对协议，在事后结果盲目的情况下比较来自共享初始轨迹的反馈，同时重置环境和模型上下文，以便在相同任务上进行新的尝试。在Terminal-Bench 2.1上，DENSE在四个接收模型中实现了测试的非特权反馈方法中最高的严格通过率。相较于初始执行，严格通过率提高了7.12-15.64个百分点，重试中观察到的接收令牌减少了19.0-43.6%。GPT-5.5的消融实验支持将嵌套子任务分析与快捷构建和问题协调相结合。这些发现指向通过基于证据的轨迹重用实现代理自我优化，减少对外部监督的依赖。
View on arXiv Download PDF AI Translation
cs.AI / 23 / 2609.21432

GVPO++: Group Variance Policy Optimization for LLM Post-Training and On-Policy Distillation

GVPO++：用于大语言模型后训练和在线蒸馏的群体方差策略优化
Zhang, Kaichen, Hong, Yuzhong, Bao, Junwei, Jiang, Hongfei, Song, Yang, Hong, Dingqian, Xiong, Hui
Abstract
Post-training plays a pivotal role in enhancing the reasoning capabilities and task-specific expertise of large language models (LLMs). Despite recent advances in post-training methods, such as Group Relative Policy Optimization (GRPO), their practical deployment remains impeded by training instability arising from the reliance on importance sampling. We introduce Group Variance Policy Optimization (GVPO), a novel post-training method that integrates the analytical solution of KL-constrained reward maximization into its gradient weighting scheme. This formulation provides an intuitive interpretation: GVPO's gradient corresponds to the mean squared error between the central distance of implicit rewards and that of actual rewards. GVPO offers two key advantages: (1) it guarantees a unique optimal solution, exactly to the KL-constrained reward maximization objective, and (2) it enables flexible sampling distributions without requiring importance sampling. Beyond general post-training, we show that GVPO naturally extends to on-policy distillation (OPD). Furthermore, GVPO enables the optimization of a broad family of extended OPD objectives, providing a principled foundation for diverse objective design. By unifying theoretical guarantees with practical adaptability, GVPO establishes a new paradigm for reliable and versatile LLM post-training and on-policy distillation.
Chinese Translation
后训练在增强大语言模型（LLMs）的推理能力和任务特定专业知识方面发挥着关键作用。尽管最近在后训练方法上取得了进展，例如群体相对策略优化（GRPO），但由于依赖重要性采样而导致的训练不稳定性仍然阻碍了其实际应用。我们提出了群体方差策略优化（GVPO），这是一种新颖的后训练方法，它将KL约束奖励最大化的解析解集成到其梯度加权方案中。该公式提供了一个直观的解释：GVPO的梯度对应于隐性奖励的中心距离与实际奖励的中心距离之间的均方误差。GVPO提供了两个关键优势：（1）它保证了一个唯一的最优解，完全符合KL约束奖励最大化目标；（2）它允许灵活的采样分布，而无需重要性采样。除了通用的后训练外，我们还展示了GVPO自然扩展到在线政策蒸馏（OPD）。此外，GVPO使得广泛的扩展OPD目标的优化成为可能，为多样化目标设计提供了原则基础。通过将理论保证与实际适应性相结合，GVPO为可靠且多功能的大语言模型后训练和在线蒸馏建立了一个新范式。
View on arXiv Download PDF AI Translation
cs.AI / 24 / 2609.21470

Risk-Aware Occupancy for Safety-Oriented End-to-End Autonomous Driving

面向安全的端到端自主驾驶的风险感知占用
Chen, Jiaxing, Zou, Hengduo, Zhao, Yiren, Gao, Bolin
Abstract
Sparse representation formulates the environment perception for the end-to-end driving system as a set of discrete elements like objects and lane lines. This formulation meets safety risks in crowded, occluded scenes dealing with unstructured obstacles, uncertain regions, and intricate interactions. In this paper, we propose a dense representation, risk-aware occupancy, to characterize planning-relevant risks in an explicit and uniform manner. It jointly encodes global scene occupancy, map-derived traffic constraints, and future dynamic agent occupancy into a unified BEV map. The unified BEV map captures the risk evidence for trajectory planning in both spatial and temporal dimensions. We design an E2E network, ROIDrive, to realize risk-aware occupancy. It predicts risk-aware occupancy with an independent branch and injects it into planning queries for safety-oriented trajectory generation. In addition, to quantify the safety problem, we introduce RiskOcc4D-nuScenes built upon nuscenes and occ3d-nuscenes. Our risk-aware occupancy yields relative open-loop collision reductions of 52.9% under the UniAD metric and 35.0% under the ST-P3 metric on nuScenes.
Chinese Translation
稀疏表示将环境感知为端到端驾驶系统中的一组离散元素，如物体和车道线。这种表述在处理拥挤、遮挡场景中的非结构化障碍物、不确定区域和复杂交互时面临安全风险。本文提出了一种密集表示，即风险感知占用，以明确和统一的方式表征与规划相关的风险。它将全球场景占用、基于地图的交通约束和未来动态代理的占用共同编码为一个统一的鸟瞰图（BEV）地图。该统一的BEV地图在空间和时间维度上捕捉了轨迹规划的风险证据。我们设计了一个端到端网络ROIDrive，以实现风险感知占用。它通过一个独立的分支预测风险感知占用，并将其注入规划查询中，以生成面向安全的轨迹。此外，为了量化安全问题，我们引入了基于nuScenes和occ3d-nuScenes构建的RiskOcc4D-nuScenes。我们的风险感知占用在nuScenes上在UniAD指标下实现了52.9%的相对开环碰撞减少，在ST-P3指标下实现了35.0%的减少。
View on arXiv Download PDF AI Translation
cs.AI / 25 / 2609.21486

Driving on Registers, Reasoning on Risk: Risk-Aware Occupancy for Register-Based End-to-End Autonomous Driving

基于寄存器的驾驶，风险推理：面向风险的寄存器基础端到端自主驾驶占用预测
Chen, Jiaxing, Zou, Hengduo, Qin, YuKai, Zhao, Yiren, Yu, Lidong, Gao, Bolin
Abstract
Multimodal trajectory prediction improves behavioral coverage in end-to-end autonomous driving, but existing methods remain limited by sparse scene representations. Incomplete evidence leads to low-quality candidate generation and unreliable ranking among geometrically similar trajectories. On a register-based baseline, bad and poor candidates constitute 19.74% of the candidate set, while the oracle-best candidate ranks only 33.9th on average. We propose RRDrive, which introduces risk-aware occupancy as a dense, temporally aligned, and trajectory-queryable representation. Its global structure guides high-quality multimodal generation, while candidate-conditioned risk queries support fine-grained selection. We further construct RiskOcc4D-NAVSIM with automatic risk annotations. RRDrive achieves a selected-trajectory PDMS of 0.951, representing a 1.5% relative improvement over the baseline (0.937), and improves the average candidate PDMS by 7.7%. In challenging scenes, it improves candidate PDMS by 30.2% and increases the Spearman correlation among good candidates by 0.41, from 0.26 to 0.67. To move beyond this oracle setting, we further develop an external RiskOcc predictor, a perception module that estimates risk-aware occupancy directly from sensor inputs. The competitive performance validates the representation's feasibility.
Chinese Translation
多模态轨迹预测提高了端到端自主驾驶中的行为覆盖，但现有方法仍受限于稀疏的场景表示。不完整的证据导致低质量候选生成和几何相似轨迹之间的不可靠排序。在基于寄存器的基线中，劣质和不良候选占候选集的19.74%，而最佳候选的平均排名仅为第33.9位。我们提出了RRDrive，引入风险感知占用作为一种密集、时间对齐且可查询轨迹的表示。其全局结构指导高质量的多模态生成，而候选条件下的风险查询支持细粒度选择。我们进一步构建了带有自动风险注释的RiskOcc4D-NAVSIM。RRDrive实现了0.951的选定轨迹PDMS，相较于基线（0.937）提高了1.5%的相对改进，并将平均候选PDMS提高了7.7%。在具有挑战性的场景中，它将候选PDMS提高了30.2%，并将良好候选之间的Spearman相关性从0.26提高到0.67，增加了0.41。为了超越这一最佳设置，我们进一步开发了外部RiskOcc预测器，这是一个直接从传感器输入估计风险感知占用的感知模块。竞争力的性能验证了该表示的可行性。
View on arXiv Download PDF AI Translation
cs.AI / 26 / 2609.21492

LogicTrack: Auditing Reasoning Trajectories of Large Language Models with Formal Logic Solvers

LogicTrack：使用形式逻辑求解器审计大型语言模型的推理轨迹
Hu, Jingyu, Yang, Shu, Liu, Weiru, Wang, Di
Abstract
Chain-of-Thought (CoT) reasoning has been shown to improve the performance of large language models (LLMs), yet existing optimization methods largely rely on outcome-based feedback, leaving the logical validity of intermediate reasoning steps largely unverified. To address the gap whereby LLMs arrive at correct final answers through logically flawed intermediate reasoning chains, we propose LogicTrack, a neuro-symbolic framework that audits reasoning trajectories by auto-formalizing each reasoning step into symbolic representations and verifying it with automated theorem provers. LogicTrack introduces Solver-Based Backtracking Reward (SBR), a step-wise scoring mechanism that quantifies logical soundness and guides backtracking tree search at inference time. We further extend LogicTrack to construct supervised fine-tuning (SFT) data with backtracking traces from its trajectories, enabling fine-tuned models to internalize step-wise auditing as an intrinsic capability. Extensive experiments across 8 reasoning benchmarks and 7 LLMs demonstrate that LogicTrack effectively improves both the verifiability of reasoning chains and final answer pass rate, thereby enhancing overall CoT quality and trustworthiness in high-stakes domains.
Chinese Translation
链式思维（Chain-of-Thought, CoT）推理已被证明可以提高大型语言模型（Large Language Models, LLMs）的性能，但现有的优化方法在很大程度上依赖于基于结果的反馈，导致中间推理步骤的逻辑有效性未得到验证。为了解决大型语言模型通过逻辑上有缺陷的中间推理链得出正确最终答案的这一问题，我们提出了LogicTrack，一个神经符号框架，通过将每个推理步骤自动形式化为符号表示并使用自动定理证明器进行验证，从而审计推理轨迹。LogicTrack引入了基于求解器的回溯奖励（Solver-Based Backtracking Reward, SBR），这是一种逐步评分机制，量化逻辑的合理性并指导推理时的回溯树搜索。我们进一步扩展LogicTrack，以构建带有回溯痕迹的监督微调（Supervised Fine-Tuning, SFT）数据，使微调后的模型能够将逐步审计内化为一种内在能力。在8个推理基准和7个大型语言模型上的广泛实验表明，LogicTrack有效提高了推理链的可验证性和最终答案的通过率，从而增强了高风险领域中整体链式思维的质量和可信度。
View on arXiv Download PDF AI Translation
cs.AI / 27 / 2609.21493

PolyBridgeBench: Benchmarking Multimodal LLMs for Physics-Grounded Bridge Design

PolyBridgeBench：基于物理的桥梁设计多模态大型语言模型基准测试
Zhao, Zicheng, Chen, Dongyin, Xu, Rui, Xu, Yinghui
Abstract
Multimodal large language models, or MLLMs, perform well at visual understanding and structured generation, yet these capabilities do not establish whether an engineering design will work when executed. Existing benchmarks assess spatial reasoning, structural validity, or physics-grounded construction, but they do not determine whether MLLMs can synthesize complete load-bearing structures and repair them after simulator execution exposes a failure. We introduce PolyBridgeBench, an executable benchmark for multimodal bridge design. A model receives a visual scene and structured engineering constraints and generates a complete node--member--material topology. Deterministic legality checks gate execution in a native dynamic physics simulation. Following an execution failure, the benchmark returns temporal visual evidence from the failed rollout and evaluates repair under a fixed interaction budget. Separate measurements of deterministic validity, dynamic functional success, and post-failure recovery identify the stage at which design fails. Experiments with six representative MLLMs across 189 levels expose a substantial gap between deterministic validity and dynamic success, pronounced sensitivity to material budgets, and limited post-failure recovery under the primary strict-budget setting.
Chinese Translation
多模态大型语言模型（MLLMs）在视觉理解和结构生成方面表现良好，但这些能力并不能确定工程设计在执行时是否有效。现有基准测试评估空间推理、结构有效性或基于物理的构造，但并未确定MLLMs是否能够综合完整的承载结构，并在模拟器执行暴露出故障后进行修复。我们提出了PolyBridgeBench，这是一个可执行的多模态桥梁设计基准测试。模型接收视觉场景和结构工程约束，并生成完整的节点-构件-材料拓扑。确定性合法性检查在本地动态物理模拟中控制执行。在执行失败后，基准测试返回失败展开的时间视觉证据，并在固定的交互预算下评估修复。对确定性有效性、动态功能成功和失败后的恢复进行单独测量，以识别设计失败的阶段。对六个代表性MLLMs在189个关卡上的实验揭示了确定性有效性与动态成功之间的显著差距，对材料预算的敏感性明显，以及在主要严格预算设置下有限的失败后恢复能力。
View on arXiv Download PDF AI Translation
cs.AI / 28 / 2609.21509

The Communication Bottleneck: A Round-Trip Study of Tree-Structured Expression Serialization in Language Models

通信瓶颈：语言模型中树结构表达序列化的往返研究
Suau, Xavier, Morenas, Alex Ferrando de las, Zappella, Luca, Bengio, Samy
Abstract
When language models reason in chain-of-thought or exchange free-text intermediates, they serialize structured information into natural language. How much tree-structured compositional content survives this bottleneck? We propose a round-trip protocol that answers this question empirically for tree-structured expressions. A generator converts a procedurally generated arithmetic expression into a word problem, a separate extractor recovers the expression from the word problem alone, and symbolic equivalence provides an exact oracle. Evaluating all pairwise combinations of sixteen models yields a communication matrix whose marginals separate generation quality from extraction quality. Three main findings emerge. First, the channel is lossy and asymmetric: swapping which model generates and which extracts shifts accuracy by up to 60.4 points, and the best pair reaches 92.9% by combining different models on each end rather than the same model on both. Second, at least 73.6% of round-trip failures originate at generation, and difficulty is driven by tree structure (operator count, depth, right-branching) rather than model family. Third, the channel is trainable: ~3600 fine-tuning examples that share the evaluation's operators and tree shapes lift every open-weight model above untrained Gemini-3.1-Pro, an upper bound under matched semantics. A disjoint-domain regime with new operators and vocabulary also raises every open-weight model, confirming the gain is not an artifact of matched semantics, though a gap to the frontier remains. Together these results identify tree-structured expression serialization as a primary limiting factor when models communicate hierarchical structure through natural language.
Chinese Translation
当语言模型在链式推理或交换自由文本中介时，它们将结构化信息序列化为自然语言。多少树结构的组合内容能够在这一瓶颈中存活？我们提出了一种往返协议，通过实证研究回答这一问题，针对树结构表达进行分析。生成器将程序生成的算术表达式转换为文字问题，单独的提取器仅从文字问题中恢复表达式，而符号等价则提供了一个精确的标准。对十六个模型的所有成对组合进行评估，得到了一个通信矩阵，其边际分离了生成质量与提取质量。主要有三个发现。首先，通道是有损且不对称的：交换生成和提取的模型会使准确率变化高达60.4个百分点，最佳组合通过在两端结合不同模型而非在两端使用相同模型达到了92.9%的准确率。其次，至少73.6%的往返失败源于生成，难度主要由树结构（操作符数量、深度、右分支）驱动，而非模型家族。第三，通道是可训练的：约3600个共享评估操作符和树形状的微调示例使每个开放权重模型的表现超过未训练的Gemini-3.1-Pro，这是在匹配语义下的上限。一个包含新操作符和词汇的非重叠领域环境也提升了每个开放权重模型，确认这一提升并非匹配语义的伪影，尽管与前沿之间仍存在差距。这些结果共同表明，树结构表达序列化是模型通过自然语言传达层次结构时的主要限制因素。
View on arXiv Download PDF AI Translation
cs.AI / 29 / 2609.21519

Learning-to-Optimize as the Missing Architectural Layer of AI-Native Networks

学习优化作为人工智能原生网络缺失的架构层
Amati, Giambattista, Mangiatordi, Federica, Salvo, Pierpaolo, Pallotti, Emiliano, Angelini, Simone
Abstract
Artificial Intelligence (AI) is becoming a fundamental design principle of future AI-native communication networks, enabling autonomous resource management, adaptive control, and zero-touch network operation. While current AI-native architectures increasingly embed intelligence across network functions, they provide little guidance on how optimisation knowledge should be systematically generated, transferred, and exploited by AI models. This paper argues that the Learning-to-Optimize (L2O) represents the missing architectural layer between optimisation and AI-native intelligence. Rather than viewing optimisation merely as an online decision engine, the proposed paradigm redefines optimisation algorithms as offline knowledge generators that produce high-quality supervisory information for neural surrogate models. The resulting models inherit optimisation expertise while enabling low-latency runtime inference suitable for dynamic network environments. A generic four-stage L2O workflow is introduced, comprising optimisation, knowledge generation, surrogate learning, and runtime inference. Unlike existing Learning-to-Optimize approaches, which primarily focus on algorithm acceleration, the proposed framework establishes L2O as an architectural abstraction applicable across heterogeneous communication and computing systems. The proposed paradigm is illustrated by an NR-V2X relay-selection problem, in which optimisation-generated solutions from a Mixed-Integer Linear Programming (MILP) solver are used to train a Graph Neural Network that can reproduce near-optimal decisions in real time. The presented perspective positions Learning-to-Optimize as a key architectural enabler for future AI-native networks.
Chinese Translation
人工智能（AI）正成为未来人工智能原生通信网络的基本设计原则，使得自主资源管理、自适应控制和零接触网络操作成为可能。尽管当前的人工智能原生架构在网络功能中越来越多地嵌入智能，但它们对如何系统性地生成、转移和利用优化知识提供的指导仍然有限。本文认为，学习优化（Learning-to-Optimize, L2O）代表了优化与人工智能原生智能之间缺失的架构层。所提出的范式将优化算法重新定义为离线知识生成器，而不仅仅是在线决策引擎，这些生成器为神经代理模型提供高质量的监督信息。由此产生的模型继承了优化专业知识，同时实现了适合动态网络环境的低延迟运行推理。本文介绍了一个通用的四阶段L2O工作流程，包括优化、知识生成、代理学习和运行时推理。与现有的学习优化方法主要关注算法加速不同，所提出的框架将L2O确立为适用于异构通信和计算系统的架构抽象。通过一个NR-V2X中继选择问题来说明所提出的范式，其中来自混合整数线性规划（Mixed-Integer Linear Programming, MILP）求解器生成的优化解决方案用于训练图神经网络，以实时再现近乎最优的决策。所呈现的视角将学习优化定位为未来人工智能原生网络的关键架构推动者。
View on arXiv Download PDF AI Translation
cs.AI / 30 / 2609.21548

Dual-Interest Sequential Product Recommendation With Multi-Granular SSM

基于多粒度状态空间模型的双重兴趣序列产品推荐
Liao, Shuiying, Mok, P. Y.
Abstract
Sequential recommendation aims to predict the next item a user will interact with based on their historical behavior. Advances in Transformers have significantly improved sequential recommendation but are still limited by cost efficiency. Although State Space Models (SSMs) have recently enabled efficient long-range modeling, most existing methods encode each item with a single static contextual role, overlooking the phenomenon of item polysemy. In fact, the same item often plays different semantic roles depending on user context, and existing methods are limited in capturing dynamic behavior across different temporal granularities. In this work, we propose DSRec, a novel dual-interest cross-SSM model that explicitly disentangles item roles across long-term and short-term semantic context. Sequential items are encoded into long-term interest embeddings that capture stable preferences via historical aggregation, and a short-term interest branch that emphasizes local session intent modulated by inter-click time intervals. These interest embeddings are processed through distinct SSM encoders: a full-sequence Mamba for long-term modeling, and a time-modulated SSM that dynamically adjusts state evolution based on temporal gaps. To enable effective cross-granularity alignment, we adopt a residual cross-fusion mechanism that exchanges contextual information between the two branches while preserving semantic independence. Experiments on public benchmarks demonstrate that DSRec outperforms other state-of-the-art methods.
Chinese Translation
序列推荐旨在根据用户的历史行为预测用户将要互动的下一个项目。尽管变换器（Transformers）的进步显著提升了序列推荐的效果，但在成本效率方面仍然存在限制。虽然状态空间模型（State Space Models, SSMs）最近使得高效的长程建模成为可能，但大多数现有方法仅用单一静态上下文角色对每个项目进行编码，忽视了项目多义性的现象。实际上，同一项目在不同用户上下文中往往扮演不同的语义角色，而现有方法在捕捉不同时间粒度下的动态行为方面存在局限。在本研究中，我们提出了DSRec，一种新颖的双重兴趣交叉状态空间模型，明确区分了长期和短期语义上下文中的项目角色。序列项目被编码为长期兴趣嵌入，通过历史聚合捕捉稳定偏好，并且有一个短期兴趣分支，强调由点击间隔调节的局部会话意图。这些兴趣嵌入通过不同的SSM编码器处理：用于长期建模的全序列Mamba，以及一个基于时间调节的SSM，根据时间间隔动态调整状态演变。为了实现有效的跨粒度对齐，我们采用了一种残差交叉融合机制，在两个分支之间交换上下文信息，同时保持语义独立性。在公共基准上的实验表明，DSRec的表现优于其他最先进的方法。
View on arXiv Download PDF AI Translation
cs.AI / 31 / 2609.21599

Beyond Accuracy: Centroid-Guided Contrastive Loss for Structured Fraudulent Job Posting Detection

超越准确性：基于中心引导的对比损失用于结构化欺诈性职位发布检测
Ahmed, Syed Ali, Raza, Malaika, Siddiqui, Muhammad Shoaib, Rafi, Muhammad
Abstract
Fraudulent job posting detection aims to identify job advertisements that are corrupted either through fake content, misleading information, or negative intent, disrupting the online eco-system of job-seekers and employers. Existing studies in this domain lack effective methods to simultaneously achieve high accuracy and meaningful structure of latent-space representations that capture subtleties among fake posts. To this end, we propose Centroid-Guided Contrastive Loss (CGCL), a loss function which unifies classification with densely formulated clustering to consistently reshape latent-space through a centroid-driven top-$k$ push-and-pull mechanism. The complementary nature of CGCL enables the model to enforce accurate decision boundaries and maintain high clustering compactness, effectively capturing both class separability and latent structure. Extensive experiments demonstrate the state-of-the-art (SOTA) performance of our method on EMSCAD, a public benchmark dataset. The code associated with this work is available at: https://github.com/ali-ahmed925/CGCL_code
Chinese Translation
欺诈性职位发布检测旨在识别那些因虚假内容、误导性信息或负面意图而被破坏的职位广告，这些广告扰乱了求职者和雇主的在线生态系统。现有研究在这一领域缺乏有效的方法，无法同时实现高准确性和捕捉虚假职位细微差别的潜在空间表示的有意义结构。为此，我们提出了中心引导对比损失（Centroid-Guided Contrastive Loss, CGCL），这是一种将分类与密集聚类相结合的损失函数，通过中心驱动的前$k$推拉机制持续重塑潜在空间。CGCL的互补特性使得模型能够强制执行准确的决策边界并保持高聚类紧凑性，有效捕捉类别可分性和潜在结构。大量实验表明我们的方法在公共基准数据集EMSCAD上达到了最先进的（SOTA）性能。与本研究相关的代码可在以下链接获取：https://github.com/ali-ahmed925/CGCL_code
View on arXiv Download PDF AI Translation
cs.AI / 32 / 2609.21600

Reducing Barriers to Academic Support: Evaluating a Course-Specific RAG System for Addressing Help-Seeking Disparities in Higher Education

降低学术支持障碍：评估面向特定课程的RAG系统以解决高等教育中的求助差异
Gray, Andy, Hobbs, Jake
Abstract
Access to academic support is a key determinant of student success, yet students experience it unequally: some readily seek help from lecturers or tutors, while others hesitate due to anxiety, fear of judgement, uncertainty about expectations, or low confidence in their understanding. This may be especially evident in computing education, where programming tasks are cumulative and cognitively demanding. Although students increasingly turn to general-purpose generative AI tools, these can produce responses that are inaccurate, insufficiently contextualised, or misaligned with module expectations. This study presents and evaluates Beacon, a course-specific Retrieval-Augmented Generation (RAG) system providing private, immediate, module-aligned academic support. Grounding responses in approved teaching materials, Beacon was designed to lower barriers to help-seeking while encouraging independent learning. Using a design-based research approach, Beacon was developed iteratively and evaluated via mixed methods, combining questionnaires and semi-structured interviews with students and staff at a Higher Education institution. Students described Beacon's responses as closely aligned with module content and more trustworthy than unrestricted generative AI tools, valuing its use of pseudocode and scaffolded explanations over direct solutions. Although participants remained cautious about trusting AI-generated responses without verification, they viewed the system as a valuable first point of support before consulting lecturers or official resources. The findings suggest that carefully designed course-specific AI systems may reduce barriers to academic support by occupying an intermediary space between independent study and formal support. Rather than replacing educators, educational AI may be most valuable when it broadens access to guidance while preserving the pedagogical role of lecturers.
Chinese Translation
学术支持的获取是学生成功的关键决定因素，然而学生在获取过程中存在不平等现象：部分学生能够主动向讲师或辅导员寻求帮助，而另一些学生则因焦虑、害怕被评判、对期望不确定或对自身理解缺乏信心而犹豫不决。这种现象在计算机教育中尤为明显，因编程任务具有累积性且认知负荷较大。尽管学生越来越多地使用通用生成式人工智能工具，但这些工具可能产生不准确、缺乏上下文关联或与课程要求不符的回答。本研究介绍并评估了Beacon，一种面向特定课程的检索增强生成（Retrieval-Augmented Generation, RAG）系统，旨在提供私密、即时且符合课程要求的学术支持。Beacon基于经过批准的教学材料生成回答，设计初衷是降低求助障碍，同时鼓励学生自主学习。采用设计性研究方法，Beacon通过迭代开发，并结合问卷调查与半结构化访谈对高等教育机构的师生进行了混合方法评估。学生们认为Beacon的回答与课程内容高度契合，比无约束的生成式AI工具更值得信赖，特别重视其使用伪代码和分步解释而非直接给出解决方案。尽管参与者在未验证的情况下仍对AI生成的回答持谨慎态度，但他们认为该系统是咨询讲师或官方资源前的宝贵初步支持。研究结果表明，精心设计的面向特定课程的AI系统能够通过填补自主学习与正式支持之间的中间环节，降低学术支持的障碍。教育AI的最大价值不在于取代教育者，而在于拓宽获取指导的途径，同时保留讲师的教学角色。
View on arXiv Download PDF AI Translation
cs.AI / 33 / 2609.21619

Calibrating Teacher--Student Discrepancy for On-Policy Distillation

校准教师-学生差异以进行在线蒸馏
He, Qiangqiang, Li, Jin, Chen, MingCai
Abstract
On-policy distillation (OPD) improves reasoning models by learning the token-level discrepancy between a stronger teacher and an on-policy student. However, this discrepancy does not purely reflect the capability gap between the teacher and the student: it also contains deviations arising from the teacher itself, which are consequently mixed into the observed teacher--student discrepancy and indiscriminately learned by standard OPD during training. This issue is further exacerbated by privileged OPD, where privileged information induces larger teacher-side likelihood shifts, thereby encouraging the student to learn more of the teacher's own deviation. We introduce \textbf{Calibrated On-Policy Distillation (Cal-OPD)}, which estimates the teacher's self-deviation region through positive and negative privileged interventions and calibrates the original teacher--student discrepancy by retaining only the component that lies beyond this region. Experiments on mathematical reasoning benchmarks show that, while retaining only about 52--65\% of the original teacher--student discrepancy as the optimization signal, Cal-OPD consistently outperforms standard OPD and its variants across model scales.
Chinese Translation
在线蒸馏（On-policy Distillation, OPD）通过学习更强大的教师与在线学生之间的标记级差异来改善推理模型。然而，这种差异并不单纯反映教师与学生之间的能力差距：它还包含来自教师自身的偏差，这些偏差混入了观察到的教师-学生差异中，并在训练过程中被标准 OPD 不加区分地学习。这个问题在特权 OPD（Privileged OPD）中进一步加剧，其中特权信息引发了更大的教师侧似然性变化，从而促使学生学习更多教师自身的偏差。我们提出了 extbf{校准在线蒸馏（Calibrated On-Policy Distillation, Cal-OPD）}，该方法通过正向和负向特权干预来估计教师的自我偏差区域，并通过仅保留超出该区域的成分来校准原始的教师-学生差异。在数学推理基准测试中的实验表明，尽管仅保留约 52% 至 65% 的原始教师-学生差异作为优化信号，Cal-OPD 在各模型规模上始终优于标准 OPD 及其变体。
View on arXiv Download PDF AI Translation
cs.AI / 34 / 2609.21626

One Prompt Does Not Fit All: Self-Meta-Evolve for Personalized Information Extraction

一个提示并不适合所有人：自我元进化用于个性化信息提取
Li, Hongliang, Wang, Lu, Xu, Yong, Chen, Hanyang, Hou, Zhitao, Qin, Xiaoting, Ge, Song, Lin, Qingwei, Zhang, Dongmei
Abstract
Large language models (LLMs) are increasingly deployed for enterprise information extraction (IE), where the same document must be reorganized differently for each user. Existing prompt optimization methods, however, rely on a single prompt optimized against a global objective, which is misaligned with the inherent user heterogeneity of real workplaces. We formulate enterprise IE as per-user prompt adaptation under interaction feedback and propose Self-Meta-Evolve, a hierarchical framework that maintains a dedicated prompt for each user and continuously refines it through a dual-loop process: an inner loop that edits structured prompts based on persona-conditioned feedback, and an outer loop that evolves the meta-prompt itself by distilling successful editing patterns. To enable scalable training and evaluation, we release a persona-driven IE benchmark of 292 simulated enterprise users, paired with a reproducible persona-generation pipeline grounded in O*NET occupational taxonomies. On this benchmark, Self-Meta-Evolve achieves a 74.58% success rate, outperforming the strongest prompt-optimization baseline by 13.56 absolute points, and reaches 52.54\% within only two iterations. A double-blind human study with twenty real professionals further confirms that prompts adapted by our framework win against static baselines in 71% of pairwise comparisons.
Chinese Translation
大型语言模型（LLMs）越来越多地被用于企业信息提取（IE），在这种情况下，同一文档必须根据每个用户的不同需求进行不同的重组。然而，现有的提示优化方法依赖于针对全局目标优化的单一提示，这与真实工作场所固有的用户异质性不符。我们将企业信息提取表述为基于交互反馈的每用户提示适应，并提出了自我元进化（Self-Meta-Evolve），这是一个分层框架，维护每个用户的专用提示，并通过双循环过程不断优化：内循环根据个性化反馈编辑结构化提示，外循环通过提炼成功的编辑模式来进化元提示。为了实现可扩展的训练和评估，我们发布了一个基于292个模拟企业用户的个性驱动信息提取基准，并配备了一个基于O*NET职业分类的可重复个性生成管道。在该基准上，自我元进化实现了74.58%的成功率，超越了最强的提示优化基线13.56个百分点，并在仅两次迭代内达到了52.54%。一项与二十位真实专业人士进行的双盲人类研究进一步确认，通过我们框架适应的提示在71%的成对比较中优于静态基线。
View on arXiv Download PDF AI Translation
cs.AI / 35 / 2609.21672

Accelerating Dense LLMs via L0-regularized Mixture-of-Experts

通过 L0 正则化的专家混合模型加速稠密大语言模型
Zhang, Zhenyu, Yang, Jiudong, Tao, Zhaowen, Chen, Meng
Abstract
Large language models (LLMs) achieve strong performance but suffer from slow and costly inference. Existing acceleration methods often lead to noticeable performance degradation, while Mixture-of-Experts (MoE) models require extensive computational resources. In this paper, we propose L0-MoE, a lightweight MoE approach using L0-regularization to accelerate dense LLMs nearly without performance loss. Our method introduces a cluster confusion matrix for domain-aware dataset curation and applies dynamic batching for efficient training. Experiments show that L0-MoE achieves up to 2.5x speedup over dense models while maintaining competitive performance, outperforming existing LLM acceleration baselines.
Chinese Translation
大语言模型（LLMs）在性能上表现出色，但在推理时速度慢且成本高。现有的加速方法往往导致显著的性能下降，而专家混合模型（MoE）则需要大量的计算资源。本文提出了一种轻量级的 MoE 方法 L0-MoE，利用 L0 正则化几乎不损失性能地加速稠密 LLMs。我们的方法引入了一个集群混淆矩阵用于领域感知的数据集策划，并应用动态批处理以实现高效训练。实验表明，L0-MoE 在保持竞争性能的同时，相较于稠密模型实现了最高 2.5 倍的加速，超越了现有的 LLM 加速基准。
View on arXiv Download PDF AI Translation
cs.AI / 36 / 2609.21677

GUARD: Natural Forgetting in Large Reasoning Models via Guided Answer-Reasoning Distillation

GUARD：通过引导答案推理蒸馏实现大型推理模型的自然遗忘
Yan, Zeyu, Zhou, Guanghao, Qiu, Minghui, Gao, Ming, Chen, Cen
Abstract
Recent advances in large reasoning models (LRMs) have made machine unlearning more challenging, as protected facts or unsafe rationales may surface in intermediate chain-of-thought (CoT) traces before the final answer is produced. Existing unlearning objectives typically suppress the target content or redirect internal representations, but they never specify how the post-forgetting trajectory should continue, which can lead to hallucinated substitutes, malformed boundaries, or repetitive outputs. We argue that LRM unlearning should instead learn a natural forgetting trajectory: a coherent non-disclosing CoT followed by a stable refusal-style answer that replace the original disclosure. To this end, we propose Guided Answer-Reasoning Distillation (GUARD), which converts model-generated unsafe disclosures into safe-exit trajectories, aligns a frozen LRM via guidance tokens, and distills the guided behavior into model parameters.To address the lack of metrics for replacement quality beyond leakage, we further introduce Natural Forgetting Reasoning Score (NFRS), which captures structural stability, fluency, and unsupported substitutes in forgotten outputs. Extensive experiments on R-TOFU and a STAR-1-derived harmful-intent setting show that GUARD substantially reduces unsafe and privacy disclosures across two widely adopted distilled LRMs while preserving reasoning utility. Codes are available at https://github.com/zeyu-Yan/GUARD
Chinese Translation
大型推理模型（LRMs）的最新进展使得机器遗忘变得更加具有挑战性，因为受保护的事实或不安全的推理可能在最终答案生成之前出现在中间的思维链（CoT）痕迹中。现有的遗忘目标通常抑制目标内容或重定向内部表示，但从未明确如何继续遗忘后的轨迹，这可能导致幻觉替代、不规范的边界或重复的输出。我们认为，LRM 的遗忘应学习一种自然的遗忘轨迹：一个连贯的不泄露的 CoT，后跟一个稳定的拒绝式答案，以替代原始的披露。为此，我们提出了引导答案推理蒸馏（GUARD），它将模型生成的不安全披露转换为安全退出轨迹，通过引导标记对冻结的 LRM 进行对齐，并将引导行为蒸馏到模型参数中。为了应对超出泄露的替代质量缺乏度量的问题，我们进一步引入了自然遗忘推理分数（NFRS），该分数捕捉了遗忘输出中的结构稳定性、流畅性和不支持的替代品。在 R-TOFU 和基于 STAR-1 的有害意图设置上的大量实验表明，GUARD 在保留推理效用的同时，显著减少了两种广泛采用的蒸馏 LRM 中的不安全和隐私披露。代码可在 https://github.com/zeyu-Yan/GUARD 获取。
View on arXiv Download PDF AI Translation
cs.AI / 37 / 2609.21683

Listen Before You Speak: Response Planning from Listener Facial Reactions for Conversational Speech Generation

先听后说：基于听者面部反应的对话语音生成响应规划
Chu, Yunji
Abstract
Conversational speech depends on dialogue context and the listener's immediately preceding behavior. We propose ReACT-TTS, a two-stage framework that uses a one-second pre-response listener facial sequence to plan the next utterance's emotion and prosody before speech realization. On a strict dyadic MELD protocol, Temporal conditioning yields higher mean macro-F1 and VAD concordance than Text-only across ten seeds, while accuracy remains essentially unchanged. Ablations show that temporal modeling performs best among the visual variants and that an explicit early-to-late difference is unnecessary; correct listener reactions also outperform cyclic mismatches on average. In a contextual-appropriateness study with 20 speech researchers, 76% of judgments prefer Temporal, 9% Text-only, and 15% report no preference. We further connect the predicted response style to a Grad-TTS backbone for end-to-end speech realization. Overall, the results support pre-response listener dynamics as complementary cues for conversational response planning. The source code is available at https://github.com/CYJ1/ReACT-TTS_public.
Chinese Translation
对话语音依赖于对话上下文及听者的即时前序行为。我们提出了ReACT-TTS，一种两阶段框架，利用一秒钟的响应前听者面部序列来规划下一句发话的情感和韵律，然后实现语音生成。在严格的双人MELD协议下，时间条件（Temporal conditioning）在十个随机种子中表现出比仅文本（Text-only）更高的平均宏F1值和语音活动检测（VAD）一致性，而准确率基本保持不变。消融实验表明，在视觉变体中时间建模表现最佳，且无需显式的早晚差异；正确的听者反应平均上也优于循环不匹配。在一项包含20名语音研究者的上下文适宜性评估中，76%的判断偏好时间条件，9%偏好仅文本，15%无偏好。我们进一步将预测的响应风格与Grad-TTS骨干网络相结合，实现端到端语音生成。总体结果支持将响应前的听者动态作为对话响应规划的补充线索。源码可在https://github.com/CYJ1/ReACT-TTS_public获取。
View on arXiv Download PDF AI Translation
cs.AI / 38 / 2609.21748

World Modeling in Transformers

变压器中的世界建模
Beckmann, Pierre, Queloz, Matthieu, Freitas, Andre
Abstract
Behavioral failures can make a transformer appear to lack a world model even when it has learned faithful representations of its environment. We demonstrate this in TaxiGPT, a transformer trained on random walks through Manhattan whose failures have been interpreted as evidence of an incoherent internal map. Through mechanistic analysis and causal interventions, we show that the model represents intersections and streets, tracks its position, and uses a goal compass to navigate. We trace its failures to interference between superposed intersection features, which disrupts localization within the internal map. Affordance packing, which groups representations of intersections with the same legal moves, helps limit the consequences of these errors. Finally, we propose mechanistic indicators that we use to compare models and show that world-modeling capacities emerge at different stages of training. Our findings motivate a shift from asking whether a model has a world model to mechanistically studying its world modeling: the interacting capacities through which it represents its environment and uses those representations to guide behavior.
Chinese Translation
行为失误可能使变压器看起来缺乏世界模型，即使它已经学习到了对环境的真实表征。我们在TaxiGPT中展示了这一点，该变压器在曼哈顿的随机行走中训练，其失误被解读为内部地图不连贯的证据。通过机械分析和因果干预，我们表明该模型能够表示交叉口和街道，跟踪其位置，并使用目标指南针进行导航。我们将其失误追溯到叠加交叉口特征之间的干扰，这干扰了内部地图中的定位。可供性打包（affordance packing）将具有相同合法移动的交叉口表征进行分组，有助于限制这些错误的后果。最后，我们提出了机械指示器，用于比较模型，并显示世界建模能力在训练的不同阶段出现。我们的发现促使我们从询问模型是否具有世界模型转向机械性地研究其世界建模：即通过交互能力来表征其环境并利用这些表征指导行为。
View on arXiv Download PDF AI Translation
cs.AI / 39 / 2609.21755

ECG Mirage: Revealing and Mitigating the Underutilisation of ECGs in Vision-Language Models for Clinical Prediction

ECG幻影：揭示和缓解临床预测中视觉-语言模型对心电图（ECG）利用不足的问题
Liang, Jinning, Zhu, Mingcheng, Zhu, Tingting
Abstract
Emergency department (ED) decision-making relies on heterogeneous clinical information, including patient history, vital signs, laboratory results, and electrocardiograms (ECGs). Vision--language models (VLMs) can jointly process these modalities, but strong predictive performance does not necessarily imply meaningful use of the correct patient's ECG. We term this failure mode ECG Mirage: apparent multimodal capability without useful dependence on patient-specific ECG information. We distinguish two forms: ECG neglect, where ECGs provide little predictive benefit, and ECG confusion, where matched ECGs outperform no-image inputs but not mismatched ECGs. To evaluate these behaviours, we compare predictions obtained with matched ECGs, outcome-discordant mismatched ECGs, and no-image inputs while holding the clinical text and prediction targets fixed. Across four VLMs on MDS-ED, matched ECGs provide no consistent advantage for either ICU admission or clinical deterioration prediction. We then train four restricted visual prompts using supervised learning followed by conditional direct preference optimisation, while keeping the VLM backbone frozen. The resulting models achieve balanced accuracies of 70.6% for ICU admission and 67.5% for deterioration and increase the matched-versus-mismatched performance gap to approximately 16.5 and 5.5 percentage points, respectively. Overall, our study identifies ECG Mirage in multimodal clinical prediction and introduces visual prompt tuning as an efficient mitigation strategy.
Chinese Translation
急诊科（ED）的决策依赖于异构的临床信息，包括患者历史、生命体征、实验室结果和心电图（ECG）。视觉-语言模型（VLMs）能够联合处理这些模态，但强大的预测性能并不一定意味着对特定患者ECG的有效利用。我们将这种失败模式称为ECG幻影：表面上的多模态能力而没有对患者特定ECG信息的有用依赖。我们区分两种形式：ECG忽视，即ECG提供的预测效益很小，以及ECG混淆，即匹配的ECG在性能上优于无图像输入，但不如不匹配的ECG。为了评估这些行为，我们比较了使用匹配ECG、结果不一致的不匹配ECG和无图像输入所获得的预测，同时保持临床文本和预测目标不变。在MDS-ED上的四个VLM中，匹配ECG在ICU入院或临床恶化预测方面没有提供一致的优势。随后，我们使用监督学习训练了四个限制性视觉提示，并通过条件直接偏好优化进行调整，同时保持VLM骨干网络不变。最终模型在ICU入院和恶化预测方面的平衡准确率分别达到了70.6%和67.5%，并将匹配与不匹配的性能差距分别提高到约16.5和5.5个百分点。总体而言，我们的研究识别了多模态临床预测中的ECG幻影，并引入视觉提示调优作为一种有效的缓解策略。
View on arXiv Download PDF AI Translation
cs.AI / 40 / 2609.21801

LLM-Generated Feature Pools for Time Series Anomaly Detection

用于时间序列异常检测的LLM生成特征池
Hili, Youssef Attia El, Tiomoko, Malik, Ancourt, Corinne
Abstract
We study how far a simple statistical pipeline can go on univariate time series anomaly detection under a strict selection protocol. The method extracts a small pool of statistics over sliding windows, scores each window with a transductive robust (MAD) model, and selects a feature subset per domain on a held-out tuning split. On TSB-AD-U it reaches $0.529$ per-series VUS-PR, above the best neural ($0.45$) and statistical ($0.44$) entries on the public leaderboard and within $0.06$ of the strongest pretrained foundation model, several of which use more supervision than ours. Ablations locate the cause: across three selection strategies and a hindsight oracle the score moves by $0.031$, and across the aggregation grid by $0.096$, while changing the candidate pool moves it by $0.226$. The candidate pool sets the ceiling; the search over it is second-order. We therefore generate a pool per domain by prompting a multimodal LLM with in-context example windows from that domain. The generated pools match the hand-crafted one under matched selection, and the two cover different domains: selecting over their union improves on the generated pool in all twelve generator-seed pairs and lifts the pipeline to $0.588$, matching the performance of the best entry on the leaderboard.
Chinese Translation
我们研究了在严格选择协议下，简单统计管道在单变量时间序列异常检测中的有效性。该方法在滑动窗口上提取一小部分统计量，使用传导鲁棒（MAD）模型对每个窗口进行评分，并在保留的调优分割上为每个领域选择特征子集。在TSB-AD-U数据集上，该方法达到了每个系列$0.529$的VUS-PR，超过了公共排行榜上最佳神经网络（$0.45$）和统计（$0.44$）的结果，并且与最强的预训练基础模型相差仅$0.06$，后者中的多个模型使用了比我们更多的监督。消融实验定位了原因：在三种选择策略和一个事后神谕下，得分变化为$0.031$，而在聚合网格上变化为$0.096$，而改变候选池则使得得分变化为$0.226$。候选池设定了上限；对其进行搜索是二阶的。因此，我们通过提示多模态LLM使用该领域的上下文示例窗口生成每个领域的特征池。生成的特征池在匹配选择下与手工制作的特征池相匹配，并且两者覆盖不同的领域：在它们的并集上进行选择在所有十二个生成器种子对中都优于生成的特征池，并将管道的性能提升至$0.588$，与排行榜上最佳条目的表现相匹配。
View on arXiv Download PDF AI Translation
cs.AI / 41 / 2609.21811

MIST: Multimodal Survival Prediction with Genomic-Guided Histology Attention

MIST：基于基因组引导的组织学注意力的多模态生存预测
Yavuz, Muhammet Sami, Kahya, Sabri Mustafa, Chen, Richard R., Lipkova, Jana, Wiestler, Benedikt
Abstract
Multimodal survival models can combine complementary prognostic information from whole-slide images and genomic profiles, but effective fusion remains challenging amid external cohort shift and computational complexity. To address these challenges, we propose MIST, multimodal survival prediction with genomic-guided histology attention. MIST represents genomic features as tokens and allows them to query compact foundation-model-derived histology context tokens before survival prediction. This design enriches molecular information with histology context rather than merging separately encoded modalities only at the final stage. Training combines discrete-time survival prediction with genomic feature masking, WSI dropout, and paired WSI-genomics contrastive alignment. Across four external evaluations in colon, renal, lung, and glioblastoma cohorts, MIST improves external C-index over standard fusion baselines in the primary comparisons. These results support genomic-guided histology attention as a compact and effective strategy for multimodal oncology outcome prediction. Our code is available at https://github.com/samiyavuuz/MIST .
Chinese Translation
多模态生存模型可以结合来自全切片图像和基因组特征的互补预后信息，但在外部队列转移和计算复杂性面前，有效的融合仍然具有挑战性。为了解决这些问题，我们提出了MIST，即基于基因组引导的组织学注意力的多模态生存预测。MIST将基因组特征表示为令牌，并允许它们在生存预测之前查询由基础模型生成的紧凑组织学上下文令牌。该设计通过组织学上下文丰富分子信息，而不是仅在最终阶段合并单独编码的模态。训练结合了离散时间生存预测、基因组特征掩蔽、全切片图像（WSI）丢弃和成对的WSI-基因组对比对齐。在结肠、肾脏、肺和胶质母细胞瘤四个外部评估中，MIST在主要比较中提高了外部C-index，超越了标准融合基线。这些结果支持基因组引导的组织学注意力作为一种紧凑而有效的多模态肿瘤学结果预测策略。我们的代码可在 https://github.com/samiyavuuz/MIST 获取。
View on arXiv Download PDF AI Translation
cs.AI / 42 / 2609.21841

EnterpriseVal: Quantifying the Efficacy, Reliability and Value of Generative AI in the Enterprise

EnterpriseVal：量化生成性人工智能在企业中的有效性、可靠性和价值
Ali, Abbas Raza, Siddiqui, Muhammad Ajmal, Zahid, Moona
Abstract
Frontier language models now produce professional deliverables that expert graders judge to match human work on a substantial share of economically valuable tasks, yet most enterprise GenAI initiatives fail to show a measurable business effect and a large fraction of agentic projects are expected to be cancelled. We argue that this is substantially a measurement problem: public benchmarks answer "what can the model do?", whereas a deployment decision requires "is this workflow fit, reliable, safe and worth scaling - here, on our data, under our controls?". We present EnterpriseVal, a use-case-level evaluation system that closes this gap. It comprises (i) a formal specification of the use case and of the frozen socio-technical configuration under test, model, prompts, retrieval, tools, guardrails and human oversight, with an autonomy level and consequence tier that jointly set the required evaluation intensity; (ii) a metric catalogue spanning fidelity, utility, efficiency, reliability, assurance and oversight; (iii) a grading protocol that scales blinded expert judgement with calibrated LLM-as-judge scoring through prediction-powered inference; (iv) a two-tier threshold gate, stated as an executable algorithm, that maps metric vectors with confidence bounds to REJECT/CONDITIONAL/SCALE decisions; and (v) a value-and-risk model in which the reviewer catch rate is a measured parameter. We report a pilot across three workflows in a global bank. In credit-memo drafting, human-graded citation precision reached 88% and hallucination rate 1.6% for the best model against gates of 70% and 5%; in procedure transformation, analyst refinement effort fell from an estimated 27.4 to 2.9 hours per document. We separate established results, documented pilot evidence, the proposed system and open hypotheses, and specify the experiments required for full validation
Chinese Translation
前沿语言模型现在能够生成专业的交付成果，专家评审认为这些成果在相当大比例的经济价值任务中与人类工作相匹配，但大多数企业生成性人工智能（GenAI）项目未能显示出可衡量的商业效果，且许多自主项目预计将被取消。我们认为，这在很大程度上是一个测量问题：公共基准回答的是“模型能做什么？”，而部署决策则需要回答“这个工作流程是否适合、可靠、安全，并值得在我们的数据和控制下进行扩展？”我们提出了EnterpriseVal，一个在用例级别进行评估的系统，以填补这一空白。它包括（i）对用例及其测试中的固定社会技术配置的正式规范，包括模型、提示、检索、工具、保护措施和人类监督，以及共同设定所需评估强度的自主级别和后果等级；（ii）涵盖保真度、效用、效率、可靠性、保证和监督的指标目录；（iii）一个评分协议，通过预测驱动的推理，利用经过校准的LLM作为评审者的评分，扩大盲评专家的判断；（iv）一个两级阈值门，以可执行算法的形式表述，将带有置信区间的指标向量映射到拒绝/有条件/扩展决策；以及（v）一个价值与风险模型，其中审查者捕获率是一个测量参数。我们报告了在一家全球银行的三个工作流程中的试点。在信用备忘录起草中，最佳模型的人类评分引用精度达到了88%，幻觉率为1.6%，而阈值分别为70%和5%；在程序转化中，分析师的精炼工作时间从每个文档估计的27.4小时降至2.9小时。我们区分了已建立的结果、记录的试点证据、所提议的系统和开放假设，并具体说明了完全验证所需的实验。
View on arXiv Download PDF AI Translation
cs.AI / 43 / 2609.21863

AutoRecLab: Describe the Experiment, Get the Code!

AutoRecLab：描述实验，获取代码！
Baumgart, Moritz, Meister, Philipp, Krell, Justus, Schmidt, Michael, Gipp, Bela, Beel, Joeran
Abstract
Empirical evaluation is central to recommender-systems (RecSys) research, but turning experimental designs into executable code remains a manual and error-prone task. We present AutoRecLab, a Python-based autonomous RecSys lab that automates RecSys experiments from natural-language prompts. Given a research idea, AutoRecLab derives explicit experiment requirements, builds and validates a prototype, and iteratively expands it into the requested full experiment. The workflow combines retrieval-augmented generation (RAG) for documentation lookup, static type verification, and execution-steered tree search. In our demonstration, AutoRecLab autonomously implements an explicit-to-implicit feedback conversion study. In a baseline comparison across six algorithms and three datasets, 8 of 9 runs succeed at an average cost of approx- imately $1 per run with GPT-5.4-mini.
Chinese Translation
实证评估是推荐系统（RecSys）研究的核心，但将实验设计转化为可执行代码仍然是一项手动且容易出错的任务。我们提出了AutoRecLab，一个基于Python的自主推荐系统实验室，能够从自然语言提示中自动化推荐系统实验。给定一个研究想法，AutoRecLab推导出明确的实验要求，构建并验证原型，并将其迭代扩展为所请求的完整实验。该工作流程结合了用于文档查找的检索增强生成（RAG）、静态类型验证和执行驱动的树搜索。在我们的演示中，AutoRecLab自主实现了显式到隐式反馈转换研究。在对六种算法和三个数据集的基线比较中，9次实验中有8次成功，平均每次运行成本约为1美元，使用的是GPT-5.4-mini。
View on arXiv Download PDF AI Translation
cs.AI / 44 / 2609.21924

What Should We Ask Next? Retrieval-Aware Question Learning under Partial Evidence

我们接下来应该问什么？基于部分证据的检索感知问题学习
Qian, Lyucheng, Zhang, John Yuehan, Wang, Pingyu
Abstract
Interactive retrieval under partial evidence is a sequential information-acquisition problem: an agent must decide which question will create the most useful evidence for the next retrieval update. Existing systems train this decision by imitating an offline ordering of candidate QA pairs, although question value is determined by the response it elicits and its downstream effect on retrieval. We establish that candidate discriminativeness and perceived usefulness provide weak supervision for this objective, then introduce RAVEL, a retrieval-aware online reinforcement learning framework for interactive person re-identification. RAVEL initializes from supervised question generation, observes the current Top-4 candidates directly, and optimizes the question policy with rank feedback from the full question-answer-retrieval loop. Experiments on Interactive-PEDES show that RAVEL delivers progressively stronger retrieval performance across five interaction rounds. Further analysis shows that RAVEL reallocates the questioning budget toward localized open-ended attributes, which provide more useful retrieval evidence and yield the largest gains on initially difficult queries.
Chinese Translation
在部分证据下的交互式检索是一个顺序信息获取问题：代理必须决定哪个问题将为下一个检索更新创造最有用的证据。现有系统通过模仿候选问答对的离线排序来训练这一决策，尽管问题的价值是由其引发的响应及其对检索的下游影响决定的。我们确定候选的可区分性和感知的有用性为这一目标提供了弱监督，然后引入了RAVEL（检索感知在线强化学习框架），用于交互式人员重识别。RAVEL从监督问题生成开始，直接观察当前的前四个候选，并通过完整的问答检索循环中的排名反馈优化问题策略。在Interactive-PEDES上的实验表明，RAVEL在五轮交互中逐步提升了检索性能。进一步分析显示，RAVEL将提问预算重新分配到局部开放式属性上，这些属性提供了更有用的检索证据，并在最初困难的查询中带来了最大的收益。
View on arXiv Download PDF AI Translation
cs.AI / 45 / 2609.21940

AutoViewMem: Self-Configuring Orthogonal Views for Conversational Long-Term Memory

AutoViewMem：用于对话长时记忆的自配置正交视图
Cao, Zijie, Qu, Xijun, Gu, Zhicheng, Chen, Xiaoshu, Yuan, Duanyang, Hou, Yanning, Zhou, Sihang, Gong, Jianxing, Huang, Jian, Mei, Yang
Abstract
Long-term memory is essential for large language model (LLM) agents to maintain consistency and personalization over extended interactions. Existing memory systems typically rely on fixed granularities or static schemas, but these designs struggle when heterogeneous information, such as preferences, events, constraints, and temporal updates, is embedded in a single mixed representation. The resulting semantic interference makes top-K retrieval sensitive to noise and often leaves relevant evidence poorly ranked. We present AutoViewMem, a data-driven framework that organizes long-term conversational memory into self-configuring, low-overlap semantic views before indexing. AutoViewMem discovers candidate views from interaction traces, selects a compact complementary view set, and uses these views to guide write-time structured extraction of provenance-grounded memories. This representation-first design moves semantic disentanglement from retrieval time to write time, allowing standard top-K similarity search to retrieve focused evidence without explicit routing or iterative retrieval. We further apply offline consolidation to improve memory compactness and consistency. Experiments on the LoCoMo and PersonaMem benchmarks, under both Qwen3-8B and Qwen3-14B backbones, show that AutoViewMem improves long-horizon question answering and personalization over strong memory baselines while preserving a simple inference pipeline.
Chinese Translation
长时记忆对于大型语言模型（LLM）代理在长时间交互中保持一致性和个性化至关重要。现有的记忆系统通常依赖于固定的粒度或静态模式，但当异构信息（如偏好、事件、约束和时间更新）嵌入到单一混合表示中时，这些设计往往面临困难。由此产生的语义干扰使得 top-K 检索对噪声敏感，并且常常导致相关证据的排名较低。我们提出了 AutoViewMem，这是一种数据驱动的框架，在索引之前将长时对话记忆组织成自配置、低重叠的语义视图。AutoViewMem 从交互轨迹中发现候选视图，选择一个紧凑的互补视图集，并利用这些视图指导写入时的结构化提取，以生成基于来源的记忆。这种以表示为先的设计将语义解耦从检索时间转移到写入时间，使得标准的 top-K 相似性搜索能够在没有显式路由或迭代检索的情况下检索到集中证据。我们进一步应用离线整合以提高记忆的紧凑性和一致性。在 LoCoMo 和 PersonaMem 基准上的实验，基于 Qwen3-8B 和 Qwen3-14B 的骨干网络，显示 AutoViewMem 在长时间问答和个性化方面优于强大的记忆基线，同时保持简单的推理管道。
View on arXiv Download PDF AI Translation
cs.AI / 46 / 2609.21962

Learning Cardiac Features: ECG Biometrics Across Time and~Exercise

学习心脏特征：跨时间和运动的心电图生物识别
Thiebaud, Luca, Chauchat, Paul, Ouladsine, Mustapha, Delliaux, Stéphane
Abstract
Electrocardiograms (ECGs) carry subject-specific patterns enabling reliable individual discrimination, forming the basis of ECG biometrics. Beyond authentication, this paradigm holds significant potential to secure sensitive cardiac data and to serve as a pretext task in self-supervised learning. Yet, most studies remain confined to singlesession, resting data, leaving robustness to temporal and physiological variations largely untested. We address this gap by evaluating ECG biometrics under realistic conditions involving exercise-induced stress and cross-session variability. A Siamese ResNet with late multi-lead fusion strategy is trained on a large ECG dataset extracted from cardiopulmonary exercise tests and evaluated with a exercise-and time-aware protocol, as well as on public benchmarks. This first extensive assessment of ECG biometrics under combined physiological and temporal variability achieves an intra-session rest-to-peak EER of 1.7% and stateof-the-art 3.9% on the CYBHi dataset. Findings support the presence of an intrinsic cardiac signature resilient to physiological and temporal drift.
Chinese Translation
心电图（ECG）携带特定于个体的模式，使得可靠的个体识别成为可能，这构成了心电图生物识别的基础。除了身份验证之外，这一范式在保护敏感心脏数据方面具有重要潜力，并且可以作为自监督学习中的前置任务。然而，大多数研究仍然局限于单次会话的静息数据，导致对时间和生理变化的鲁棒性测试不足。我们通过在涉及运动诱发压力和跨会话变异的现实条件下评估心电图生物识别来填补这一空白。我们使用一种带有后期多导联融合策略的Siamese ResNet，在从心肺运动测试中提取的大型心电图数据集上进行训练，并采用运动和时间感知的协议进行评估，同时在公共基准上进行验证。这是首次对结合生理和时间变异的心电图生物识别进行广泛评估，达到了1.7%的会话内静息到峰值的等错误率（EER），并在CYBHi数据集上取得了3.9%的最先进水平。研究结果支持存在一种对生理和时间漂移具有韧性的内在心脏特征。
View on arXiv Download PDF AI Translation
cs.AI / 47 / 2609.21996

A Lie Detector Test for Language Models: Reading Knowledge a Model Won't Reveal

语言模型的测谎测试：模型不会透露的知识
Dingeto, Hiskias
Abstract
Large language models can hold knowledge they do not report. A model may sandbag on a capability evaluation, or answer against what it internally knows, and its outputs alone cannot tell whether it is hiding an answer or simply does not have one. We borrow the Concealed Information Test, a forensic method that identifies guilty knowledge by presenting a suspect with the true detail among plausible decoys and measuring a stronger response to the item they recognize. Our method, Probe of Internal Recognition (PIR), does the same inside a model. It presents a question with its candidate answers and reads, from the model's internal states, which candidate the model recognizes as correct. PIR is reference-free, needing no honest reference model and no labeled truth corpus. Across eight models from five families (Gemma, Qwen, Llama, Mistral, and Phi), PIR recovers the recognized answer at 0.70 to 0.87 balanced accuracy, well above the 0.28 to 0.40 unknown-item baseline and the 0.25 chance rate. It stays readable across every form of concealment we test, from prompted deception and trained sandbagging to external password-locked and circuit-broken checkpoints, with recognition between 0.85 and 0.93. When the model hides a known answer, recognition stays high. When unlearning removes the knowledge, recognition drops to the level of a question the model never knew. PIR therefore separates a model that will not answer from one that cannot, which supports sandbagging audits and unlearning verification. The signal is causal, adds information beyond black-box behavioral cues, and extends from multiple-choice questions to free-form generation.
Chinese Translation
大型语言模型可以持有它们不报告的知识。一个模型可能在能力评估中表现不佳，或者回答与其内部知识相悖的问题，仅凭其输出无法判断它是隐藏了答案还是根本没有答案。我们借用隐蔽信息测试（Concealed Information Test），这是一种法医方法，通过向嫌疑人展示真实细节与可信的诱饵，并测量他们对被识别项目的更强反应来识别有罪知识。我们的方法，内部识别探测（Probe of Internal Recognition, PIR），在模型内部执行相同的操作。它提出一个问题及其候选答案，并从模型的内部状态中读取模型识别为正确的候选答案。PIR是无参考的，不需要诚实的参考模型和标记的真实语料库。在五个家族的八个模型（Gemma、Qwen、Llama、Mistral和Phi）中，PIR以0.70到0.87的平衡准确率恢复识别答案，远高于0.28到0.40的未知项目基线和0.25的随机率。它在我们测试的每种隐蔽形式中都保持可读性，从提示欺骗和训练的表现不佳到外部密码锁定和电路断开检查点，识别率在0.85到0.93之间。当模型隐藏已知答案时，识别率仍然很高。当去学习去除知识时，识别率降至模型从未知道的问题的水平。因此，PIR能够区分不回答的模型和不能回答的模型，这支持表现不佳的审计和去学习验证。该信号是因果的，提供超出黑箱行为线索的信息，并从多项选择题扩展到自由形式生成。
View on arXiv Download PDF AI Translation
cs.AI / 48 / 2609.22068

CodeMidas: Scaling Agentic Coding RL Environments from Code Itself

CodeMidas：从代码本身扩展自主编码强化学习环境
Ye, Bowen, Li, Lei, Li, Shicheng, Yue, Zihao, Zhang, Linghao, Lv, Hanglong, Liu, Yuanxin, Ma, Wenhan, Tian, Hao, Li, Rang, Dong, Jinhao, Zhao, Yikai, Deng, Xiangwei, Zhang, Hailin, Zhao, Liang, Liu, Qi, Kong, Lingpeng, Yang, Tong, Luo, Fuli
Abstract
Training capable coding agents via reinforcement learning (RL) requires diverse tasks with reliable verifiers. Open-source codebases offer a rich source of such tasks, while existing methods typically rely on development artifacts such as issues and commits, limiting the range of tasks that can be extracted. To better scale RL environments, we present CodeMidas, an agentic pipeline that turns implemented functionality in existing codebases into executable RL environments using source code as its only task-specific input. CodeMidas allocates agentic compute to every stage of environment construction: agents explore implemented functionality to formulate behavioral specifications, construct tests grounded in execution of the original code, and validate and filter candidate tasks through execution checks and repeated solution rollouts. The resulting dataset has 5,545 training tasks from 3,185 open-source codebases spanning 23 programming languages and 15 technical domains. Training MiMo-V2.5 on these tasks with GRPO improves performance on all five diverse benchmarks, covering issue repair (DeepSWE + 11.7%), whole-program construction (ProgramBench +17%), and terminal work (Terminal-Bench v2.1 +8.5%). Ablations show that increasing the number of high-quality training tasks improves performance. Trajectory analysis shows the RL-trained agent demonstrates better behaviors like increasing codebase exploration and more diverse self-verification. These results establish source code as a scalable foundation for constructing RL environments that improve coding agents across diverse software tasks.
Chinese Translation
通过强化学习（RL）训练能够编码的智能体需要多样化的任务和可靠的验证器。开源代码库提供了丰富的任务来源，而现有方法通常依赖于开发文档，如问题和提交，这限制了可提取任务的范围。为了更好地扩展RL环境，我们提出了CodeMidas，这是一个自主管道，它将现有代码库中实现的功能转化为可执行的RL环境，使用源代码作为唯一的任务特定输入。CodeMidas在环境构建的每个阶段分配自主计算：智能体探索实现的功能以制定行为规范，构建基于原始代码执行的测试，并通过执行检查和重复解决方案回滚来验证和筛选候选任务。最终的数据集包含来自3,185个开源代码库的5,545个训练任务，涵盖23种编程语言和15个技术领域。在这些任务上训练MiMo-V2.5与GRPO结合，提升了在五个多样化基准上的性能，包括问题修复（DeepSWE + 11.7%）、整程序构建（ProgramBench +17%）和终端工作（Terminal-Bench v2.1 +8.5%）。消融实验表明，增加高质量训练任务的数量能够提高性能。轨迹分析显示，经过RL训练的智能体表现出更好的行为，例如增加代码库探索和更具多样性的自我验证。这些结果确立了源代码作为构建RL环境的可扩展基础，从而提升智能体在多样化软件任务中的表现。
View on arXiv Download PDF AI Translation
cs.AI / 49 / 2609.22086

Designer-RSI: Evolving Procedural Memory from User Traffic for Agentic Graphic Design

设计师-RSI：从用户流量中演化的程序记忆用于代理图形设计
Du, Hongyang, Yan, Lan, Flores, Christian, Kadav, Asim
Abstract
Professional graphic design is a long-horizon agentic task in which structured, editable artifacts emerge from many interdependent actions, yet outcomes admit no reliable programmatic oracle. We introduce a continual adaptation framework in which a frozen frontier model operates professional design software through more than 230 tools, while an external procedural memory of natural-language skills accumulates and refines reusable design procedures from experience. The memory widens by acquiring procedures for recurring uncovered subtasks and deepens by revising existing procedures against their own successful and failed executions, while a matched replay gate admits only changes that repair failures without regressing observed successes. Five rounds over 1,406 real user briefs and 1,869 automatically graded trajectories, with no weight updates and no human labels, grow the bank from 76 documentation-derived skills to 139 and raise GenEval2 execution success on Claude-Sonnet-4 from 72.7% to 99.3% (+11.99 points in generation quality), with 61.8% and 67.6% win rates against the no-skill agent across four specialized design benchmarks on Claude-Sonnet-4 and Claude-Opus-4.6. We further show the two mechanisms are effective in combination: on 200 held-out briefs from user-traffic benchmark, widening or deepening alone reaches a 49.4% / 48.6% win rate over the no-skill agent, while their combination reaches 58.5% (p = 0.025). Procedural memory offers a practical route to continual adaptation of agents under noisy, unverifiable feedback.
Chinese Translation
专业图形设计是一项长期的代理任务，其中结构化的可编辑工件源于许多相互依赖的动作，但结果并没有可靠的程序化预言。我们引入了一个持续适应框架，其中一个冻结的前沿模型通过230多种工具操作专业设计软件，而一个外部的自然语言技能程序记忆则从经验中积累和完善可重用的设计程序。该记忆通过获取重复出现的未覆盖子任务的程序而扩展，并通过根据自身成功和失败的执行情况修订现有程序而深化，同时一个匹配的重放门仅允许修复失败而不回退已观察到的成功的变化。在1406个真实用户简报和1869条自动评分轨迹的五轮测试中，在没有权重更新和没有人工标签的情况下，技能库从76个文档衍生的技能增长到139个，并将GenEval2在Claude-Sonnet-4上的执行成功率从72.7%提高到99.3%（生成质量提高了11.99个百分点），在Claude-Sonnet-4和Claude-Opus-4.6的四个专业设计基准上，与无技能代理的胜率分别为61.8%和67.6%。我们进一步展示了这两种机制的有效组合：在用户流量基准的200个保留简报上，单独扩展或深化分别达到了49.4%和48.6%的胜率，而它们的组合达到了58.5%（p = 0.025）。程序记忆为在嘈杂、不可验证反馈下的代理持续适应提供了一条实用的途径。
View on arXiv Download PDF AI Translation
计算语言学 (Computation and Language)
65
cs.CL / 1 / 2609.20824

Do small language models know what they don't know?

小型语言模型是否知道它们不知道什么？
Mudgal, Prashant
Abstract
We explore whether entropy-based confidence signals can be leveraged to improve the accuracy of Small Language Models (SLMs) with fewer than 3 billion parameters, running entirely on consumer hardware. We evaluate seven distinct approaches, including token-level entropy early stopping, semantic entropy estimation, and uncertainty-aware routing to larger expert models, across 7 model pairs and 5 standard NLU benchmarks. Our key finding is that token-level entropy is effectively blind in SLMs: in 91% of dataset-model combinations, mean token entropy is near zero regardless of answer correctness, rendering token-based confidence signals unusable at this scale. We demonstrate that semantic entropy, computed by generating multiple samples, clustering answers by meaning, and measuring distributional uncertainty, recovers a viable confidence signal. Using semantic entropy to selectively route uncertain queries to a larger expert model yields accuracy improvements of up to +50 percentage points. Notably, cross-family routing (e.g., SmolLM 360M to Phi-3.5-mini) averages +22.0% improvement compared to +6.8% for same-family routing, revealing that expert model quality matters more than architectural compatibility. Our results suggest that the value proposition for entropy-based methods in SLMs is not computational savings but intelligent compute allocation: spending more tokens where they matter most.
Chinese Translation
我们探讨了基于熵的置信信号是否可以被利用，以提高参数少于30亿的小型语言模型（Small Language Models, SLMs）的准确性，这些模型完全在消费级硬件上运行。我们评估了七种不同的方法，包括基于标记的熵早停、语义熵估计以及不确定性感知的路由到更大专家模型的方法，涵盖了7对模型和5个标准自然语言理解（Natural Language Understanding, NLU）基准测试。我们的主要发现是，在小型语言模型中，基于标记的熵在91%的数据集-模型组合中几乎为零，无论答案的正确性如何，这使得在这一规模下基于标记的置信信号无法使用。我们展示了通过生成多个样本、按意义聚类答案并测量分布不确定性来计算的语义熵，能够恢复一个可行的置信信号。使用语义熵有选择性地将不确定的查询路由到更大的专家模型，准确性提高了多达50个百分点。值得注意的是，跨家族路由（例如，从 SmolLM 360M 到 Phi-3.5-mini）平均提高了22.0%，而同家族路由仅提高了6.8%，这表明专家模型的质量比架构兼容性更为重要。我们的结果表明，在小型语言模型中，基于熵的方法的价值主张不是计算节省，而是智能计算分配：在最重要的地方花费更多的标记。
View on arXiv Download PDF AI Translation
cs.CL / 2 / 2609.20825

HERMES: Contrast-Aware Knowledge Graph Reasoning from Clinical Notes for Patient Outcome Prediction

HERMES：基于对比感知的知识图谱推理用于临床记录中的患者结果预测
Nguyen, Gia-Bach, Nguyen, Hoang-Ha, Vuong, Tuan-Cuong, Xuan, Trang Mai, Ngo, Duy Quoc, Nguyen, Tien-Cuong, Vu, Huan, Van Luong, Thien
Abstract
Clinical predictive models often rely on structured Electronic Health Record data, such as time-series and procedure codes. While recent approaches have begun leveraging unstructured clinical notes, they typically encode them as flat sequences, which may lose explicit relational and temporal structure present in clinical narratives. In response, we propose HERMES, a graph-based framework that operates exclusively on clinical text while preserving clinical relationships. This approach builds on two key ideas. First, personalized Knowledge Graphs (KGs) are constructed through Large-Language-Model-guided extraction from clinical notes with Contrastive Logic Modeling that explicitly captures temporal dynamics and treatment failures and changes in outcomes. Second, a Graph Attention Network synthesizes patient representations through graph-based learning over the KGs. Experiments on MIMIC-III and MIMIC-IV for in-hospital mortality and 30-day readmission prediction show that HERMES consistently outperforms strong text-only baselines. Our findings demonstrate that explicit relational modeling with Contrastive Logic Modeling significantly advances predictive performance.
Chinese Translation
临床预测模型通常依赖于结构化的电子健康记录数据，如时间序列和程序代码。尽管近期的方法开始利用非结构化的临床记录，但它们通常将其编码为平坦序列，这可能会丢失临床叙述中存在的明确关系和时间结构。为此，我们提出了HERMES，一个专门针对临床文本的图形框架，同时保留临床关系。该方法基于两个关键思想。首先，通过大型语言模型指导的从临床记录中提取构建个性化知识图谱（Knowledge Graphs, KGs），并采用对比逻辑建模（Contrastive Logic Modeling）明确捕捉时间动态、治疗失败及结果变化。其次，图注意力网络（Graph Attention Network）通过对KGs的图形学习合成患者表示。在MIMIC-III和MIMIC-IV数据集上进行的院内死亡率和30天再入院预测实验表明，HERMES在强文本基线模型上始终表现优越。我们的研究结果表明，采用对比逻辑建模的显式关系建模显著提升了预测性能。
View on arXiv Download PDF AI Translation
cs.CL / 3 / 2609.20826

TALON: A Temporally Aware Longitudinal Framework for Radiology Report Generation

TALON：一种时间感知的纵向放射学报告生成框架
Sun, Nien-Tsyr, Chen, Min-Chen, Hung, Hui Nien, Tseng, Vincent S.
Abstract
Current radiology report generation (RRG) models usually produce descriptive reports based on a single examination or only the most recent prior examination, limiting their ability to perform accurate and meaningful longitudinal comparisons and detect subtle interval changes. Although recent approaches have begun to incorporate multiple prior examinations, they usually aggregate a fixed-length history without explicitly modeling the role-dependent relevance of each prior examination before fusion. To address this, we propose TALON, a Temporally Aware LONgitudinal RRG framework that adaptively integrates variable-length patient histories. The underlying Dual-Channel Temporal Fusion Module (DCTFM) compares the current examination with each prior examination through complementary similarity and change channels to capture persistent findings and interval changes, respectively. The specially designed channel-specific attention estimates the relevance of each prior examination, while a learned prior-specific gate adaptively integrates informative longitudinal evidence and suppresses redundancy. Experiments on MIMIC-CXR show that TALON outperforms the current state-of-the-art method on various clinical efficacy and graph-based metrics. When more prior examinations become available, TALON's performance on these metrics improves even further, emphasizing the strength of TALON's DCTFM in modeling longitudinal RRG across longer and more complex patient histories than existing approaches.
Chinese Translation
当前的放射学报告生成（RRG）模型通常基于单次检查或仅最近的先前检查生成描述性报告，这限制了它们进行准确和有意义的纵向比较以及检测细微间隔变化的能力。尽管最近的方法已开始整合多个先前检查，但它们通常以固定长度的历史进行汇总，而未在融合之前明确建模每个先前检查的角色相关性。为了解决这个问题，我们提出了TALON，一个时间感知的纵向RRG框架，能够自适应地整合可变长度的患者历史。其基础的双通道时间融合模块（DCTFM）通过互补的相似性和变化通道将当前检查与每个先前检查进行比较，以捕捉持续的发现和间隔变化。特别设计的通道特定注意力机制评估每个先前检查的相关性，而学习到的先前特定门控自适应地整合有信息的纵向证据并抑制冗余。在MIMIC-CXR上的实验表明，TALON在各种临床有效性和基于图形的指标上优于当前的最先进方法。当更多的先前检查可用时，TALON在这些指标上的表现进一步改善，强调了TALON的DCTFM在建模比现有方法更长且更复杂的患者历史中的纵向RRG的优势。
View on arXiv Download PDF AI Translation
cs.CL / 4 / 2609.20827

From Discharge Notes to Patient Understanding: Persona-Grounded, Open-Ended Simulation of LLMs as Discharge Educators

从出院记录到患者理解：基于角色的开放式模拟LLMs作为出院教育者
Jang, Won Seok, Yao, Zonghai, Yu, Hong
Abstract
Hospital discharge education is an interactive teaching task: a clinician adapts a discharge plan to a patient's literacy, recall, and personality. Existing LLM evaluations target static or artifact-generation tasks and do not measure patient understanding under open-ended dialogue. We introduce DischargeBench, a persona-grounded simulation in which a candidate LLM educator conducts a multi-turn session with a Virtual Patient, while an Education Monitor Agent regulates patient realism without modifying the educator, protecting the evaluation signal. We curate MIMIC-IV-Ext-DischargeBench, 477 cases over 24 ICD chapters with persona axes (personality, education level, health literacy, past-medical-history recall) for stratified analysis. Each simulation is scored on four axes -- Conversation Quality, Topic Checklist, Comprehension, and Factual Consistency -- by an LLM-as-a-Judge aligned against physician annotations. Across closed- and open-source LLMs, aggregate scores conceal clinically relevant variation across ICD chapters and patient personas; difficult personas expose coverage failures, comprehension gaps, and reduced source-answer agreement. LLM evaluation for discharge education should center patient understanding, not text quality or answer accuracy alone.
Chinese Translation
医院出院教育是一项互动教学任务：临床医生根据患者的识字能力、记忆力和个性调整出院计划。现有的LLM评估主要针对静态或生成性任务，并未在开放式对话中测量患者的理解能力。我们提出了DischargeBench，这是一种基于角色的模拟，其中候选的LLM教育者与虚拟患者进行多轮会话，同时教育监控代理调节患者的真实感，而不修改教育者，从而保护评估信号。我们策划了MIMIC-IV-Ext-DischargeBench，涵盖477个案例，涉及24个ICD章节，并设定角色维度（个性、教育水平、健康素养、既往病史记忆）以进行分层分析。每个模拟会话在四个维度上进行评分——对话质量、主题清单、理解能力和事实一致性——由对齐于医生注释的LLM作为评审。通过对封闭源和开放源LLM的评估，汇总分数掩盖了不同ICD章节和患者角色之间的临床相关变异；困难角色暴露了覆盖不足、理解差距和源答案一致性降低的问题。LLM在出院教育中的评估应以患者理解为中心，而不仅仅是文本质量或答案准确性。
View on arXiv Download PDF AI Translation
cs.CL / 5 / 2609.20828

Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR

超越字词错误率：带口音的对话式自动语音识别中的实体和流畅性回忆
Husain, Fiza, Pandey, Ankit, Singh, Yash
Abstract
ASR systems optimised for Word Error Rate (WER) often miss named entities and filled pauses in accented conversational English, both critical for language-learning feedback. We present a three-stage pipeline for speakers from India, Indonesia, and Latin America: (1) heuristic SQL filters curating entity-rich training data at 2.8x the entity density of random sampling, (2) regional LoRA adapters fine-tuned on Qwen2.5-Omni-3B producing both verbatim and corrected transcripts in a single forward pass, and (3) a six-category error taxonomy validated by an LLM-based judge (83.8% agreement, 210 human-labelled samples). The pipeline achieves 80-85% entity recall (up from 53-55%), 76-86% filler recall (up from <5%), and 6-10% WER across 6k test utterances, outperforming Whisper and a commercial ASR on entity recall while matching a zero-shot 30B model with 10x fewer parameters. Paired bootstrap tests confirm that curation alone accounts for 2.8-4.2 pp of entity recall gain (p<0.0001).
Chinese Translation
针对字词错误率（WER）优化的自动语音识别（ASR）系统在处理带口音的对话英语时，常常遗漏命名实体和填充停顿，而这两者对于语言学习反馈至关重要。我们提出了一种针对来自印度、印度尼西亚和拉丁美洲的说话者的三阶段管道：（1）使用启发式SQL过滤器筛选出实体丰富的训练数据，其实体密度是随机抽样的2.8倍；（2）基于Qwen2.5-Omni-3B微调的区域性LoRA适配器，在单次前向传递中生成逐字和修正的转录文本；（3）通过基于大型语言模型（LLM）的评估者验证的六类错误分类法（83.8%的一致性，210个人工标注样本）。该管道在6000个测试语句中实现了80-85%的实体回忆率（较之前的53-55%有所提升），76-86%的填充回忆率（较之前的<5%有所提升），以及6-10%的字词错误率，超越了Whisper和一款商业ASR在实体回忆方面的表现，同时与一个零-shot 30B模型的表现相当，但参数数量少10倍。配对自助法测试确认，仅数据筛选就贡献了2.8-4.2个百分点的实体回忆率提升（p<0.0001）。
View on arXiv Download PDF AI Translation
cs.CL / 6 / 2609.20829

SAGE: Schema-Guided LLMs for Grant Review

SAGE：基于模式引导的资助评审大型语言模型
Varapaev, Erik, Chetvergov, Andrei, Ukolov, Stepan, Sivoraksha, Timofei, Evseev, Alexander, Bolovtsov, Sergey
Abstract
Grant reviewers must apply detailed criteria to application forms, budgets, and supporting documents while producing assessments that colleagues can inspect. We present SAGE, Schema-Guided Aspect-Based Grant Evaluation, a system that translates a grant rubric into structured checks and links its judgements to evidence from the application package. We evaluate SAGE in two stages on 35 nonprofit grant applications. A post-factum comparison with 105 reviews from the original competition shows fair ordinal agreement (kappa = 0.29). The foundation then conducted a criterion-level re-review after inspecting SAGE, producing 202 assessments. In this assisted round, SAGE reached kappa = 0.58 and outperformed a one-prompt-per-criterion baseline (kappa = 0.33 on the common subset), with higher rank correlation and lower error. A claim-level audit further identifies confirmed, disputed, and unaddressed parts of the structured draft. SAGE operationalizes the review methodology by producing a detailed, evidence-linked, and auditable draft for expert correction.
Chinese Translation
资助评审者必须对申请表、预算和支持文件应用详细标准，同时生成可供同事检查的评估。我们提出了SAGE（Schema-Guided Aspect-Based Grant Evaluation），一个将资助评估标准转化为结构化检查并将其判断与申请包中的证据关联的系统。我们在35个非营利资助申请上对SAGE进行了两阶段评估。与原始竞争中的105份评审进行事后比较显示出公平的序数一致性（kappa = 0.29）。随后，基金会在检查SAGE后进行了标准层面的重新评审，产生了202份评估。在这一辅助评审中，SAGE达到了kappa = 0.58，超越了每个标准一个提示的基线（在共同子集上的kappa = 0.33），具有更高的排名相关性和更低的错误率。声明级别的审计进一步识别了结构化草稿中确认的、争议的和未解决的部分。SAGE通过生成详细的、证据链接的和可审计的草稿，为专家修正操作化了评审方法。
View on arXiv Download PDF AI Translation
cs.CL / 7 / 2609.20830

Reviser: Revision-Capable Text Generation via Autoregressive Cursor Actions

Reviser：通过自回归光标动作进行可修订文本生成
Diab, Sean
Abstract
Revision-capable generation is appealing because it can insert or revise earlier content, but many non-autoregressive and edit-based approaches obtain this flexibility through repeated sequence-level computation. We propose Reviser, a decoder-only Transformer that generates a response as a sequence of cursor-relative actions on a mutable canvas. At each step, Reviser predicts exactly one action token: INSERT(token), MOVE($\Delta$), or STOP, and is autoregressive over edit-history actions rather than final text order. This design enables genuinely non-monotonic generation while preserving a simple next-action interface. On a continuation benchmark, Reviser is strongly preferred to SEDD and MDLM in our arena evaluations, and trajectory statistics confirm that the model performs frequent backward moves and mid-canvas insertions rather than merely emulating end-append decoding. Against size-matched autoregressive baselines, Reviser is competitive at both the 100M and 300M scales. Under our shared FLOPs convention, Reviser also requires substantially less inference compute than representative multi-pass refinement and diffusion-style baselines.
Chinese Translation
可修订生成具有吸引力，因为它可以插入或修订早期内容，但许多非自回归和基于编辑的方法通过重复的序列级计算获得这种灵活性。我们提出了Reviser，一种仅解码的Transformer，它将响应生成作为在可变画布上的一系列相对光标的动作。在每一步，Reviser精确预测一个动作标记：INSERT(token)、MOVE($ ext{Δ}$)或STOP，并且在编辑历史动作上是自回归的，而不是最终文本的顺序。这种设计使得真正的非单调生成成为可能，同时保持简单的下一个动作接口。在一个续写基准测试中，Reviser在我们的竞技评估中明显优于SEDD和MDLM，轨迹统计确认该模型频繁进行向后移动和中画布插入，而不仅仅是模拟末尾附加解码。与规模匹配的自回归基线相比，Reviser在100M和300M规模上都具有竞争力。在我们的共享FLOPs约定下，Reviser在推理计算上也显著低于代表性的多次精炼和扩散风格基线。
View on arXiv Download PDF AI Translation
cs.CL / 8 / 2609.20831

Recursive Language Models Generalize Out of Domain

递归语言模型在域外的泛化能力
Yang, Chenxiao, Li, Zhiyuan, McAllester, David, Srebro, Nathan
Abstract
We study when limiting what a language model can see improves learning. We compare standard CoT, the more general learner that reads the full trace, with recursive language models, which restricts itself by solving each subtask in an isolated context. In-distribution, this generality comes for free: CoT can efficiently simulate the recursive rule, so the IID generalization guarantee changes only by a constant factor, and recursion does not offer much. But out of domain, CoT can fit training by relying on context outside the current subtask, i.e. a shortcut that breaks once those tokens change; recursive context isolation rules out this failure mode. Even though CoT's class still covers the recursive rule, simplicity bias picks the shortcut over the truth. Thus, to go beyond distributional accuracy and truly reason, covering the right rule is not enough; this contrasts with classical learning theory.
Chinese Translation
我们研究了限制语言模型可见信息时如何改善学习效果。我们比较了标准的链式推理（CoT）与递归语言模型，后者通过在孤立的上下文中解决每个子任务来限制自身。在同一分布下，这种广泛性是免费的：链式推理可以有效地模拟递归规则，因此独立同分布（IID）泛化保证仅由一个常数因子改变，而递归并没有提供太多帮助。然而在域外，链式推理可以通过依赖于当前子任务之外的上下文来拟合训练，即一种在那些标记变化时会失效的捷径；而递归上下文隔离则排除了这种失败模式。尽管链式推理的类别仍然覆盖递归规则，但简单性偏见选择了捷径而非真相。因此，要超越分布准确性并进行真正的推理，仅仅覆盖正确的规则是不够的；这与经典学习理论形成对比。
View on arXiv Download PDF AI Translation
cs.CL / 9 / 2609.20832

TatBLiMP: A Benchmark of Linguistic Minimal Pairs for Tatar

TatBLiMP：鞑靼语语言最小对的基准测试
Saetov, Ilshat, Gaynullin, Dmitry
Abstract
We introduce TatBLiMP, the first benchmark of linguistic minimal pairs for Tatar (tt, ISO 639-3 tat), a Qypchaq Turkic language written in Cyrillic. To our knowledge it is the first grammaticality evaluation for Tatar language models of any kind, since even the 101-language MultiBLiMP does not include Tatar. TatBLiMP covers 16 morphosyntactic phenomena in 1248 sentence pairs. Each pair differs by a single morpheme, one grammatical and one ungrammatical. A model passes a pair when it assigns higher probability to the grammatical member. Scoring compares probabilities the model already assigns, so the benchmark needs no text generation and no parser, and it runs on base models and on mid-training checkpoints. TatBLiMP adapts the phenomenon inventory and single-morpheme breaking operations of TurBLiMP to Tatar and adds one phenomenon specific to Tatar, bare-noun number after numerals and quantifiers. The grammatical member of every pair is an attested sentence from Tatar literary prose. The ungrammatical member is produced by a deterministic single-morpheme perturbation with the apertium-tat transducer. Every pair is ratified by a native speaker. A plausibility principle governs construction, so the ungrammatical member is a plausible real-world error rather than an arbitrary corruption. Across from-scratch Tatar models, cross-lingual adaptations, and frontier multilingual LLMs, the benchmark tracks focused Tatar training rather than parameter scale. A 478M from-scratch model and a 125M monolingual model lead near 0.97, a 7B adaptation trails, frontier LLMs of 30-120B parameters fall to 0.80-0.92, and a lightly tuned multilingual model is weakest. We close with the benchmark's main limitation. Its inherited taxonomy omits the morphophonology, vowel harmony and consonant assimilation, that is most salient to native speakers, and we sketch a native second layer that would add it.
Chinese Translation
我们介绍了TatBLiMP，这是第一个针对鞑靼语（tt，ISO 639-3 tat）的语言最小对基准测试，它是一种用西里尔字母书写的Qypchaq突厥语。根据我们的了解，这是对任何类型的鞑靼语言模型进行的首次语法性评估，因为即使是包含101种语言的MultiBLiMP也不包括鞑靼语。TatBLiMP涵盖了1248对句子中的16种形态句法现象。每对句子通过一个形态素的差异而不同，其中一个是语法正确的，另一个是语法错误的。当模型对语法正确的成员分配更高的概率时，该模型通过了这一对句子。评分比较模型已经分配的概率，因此该基准测试不需要文本生成和解析器，并且可以在基础模型和中间训练检查点上运行。TatBLiMP将TurBLiMP的现象清单和单形态素扰动操作适配到鞑靼语，并增加了一个特定于鞑靼语的现象，即在数词和量词后面的裸名词数。每对句子的语法正确成员是来自鞑靼文学散文的已证实句子，而语法错误的成员是通过apertium-tat转换器生成的确定性单形态素扰动。每对句子都经过母语者的确认。构造遵循可行性原则，因此语法错误的成员是一个合理的现实世界错误，而不是任意的损坏。在从零开始的鞑靼模型、跨语言适配和前沿多语言大型语言模型（LLMs）中，该基准测试关注的是鞑靼的专注训练，而不是参数规模。一个478M的从零开始模型和一个125M的单语模型的表现接近0.97，一个7B的适配模型落后，30-120B参数的前沿LLMs的表现下降到0.80-0.92，而一个轻微调整的多语言模型表现最差。最后，我们指出该基准测试的主要局限性。其继承的分类法省略了对母语者最为显著的形态音位学、元音和谐及辅音同化，我们勾勒出一个本土的第二层次，以便将其添加进来。
View on arXiv Download PDF AI Translation
cs.CL / 10 / 2609.20833

Transsion's Speaker-Attributed Multilingual ASR System for the MLC-SLM 2026 Challenge

Transsion的说话人归属多语言自动语音识别系统在MLC-SLM 2026挑战中的应用
Ren, Zhecheng, He, Xuanji, Li, Xiaoxiao, Han, Zhichen, Dong, Gaoyang, Zhang, Gaosheng, Chen, Minchuan, Zhu, Fengjie
Abstract
This paper presents the Transsion Speech Team submission to Task 1 of the MLC-SLM 2026 Challenge, which focuses on speaker-attributed transcription for multilingual conversational speech. We propose a cascaded framework consisting of three components: a speaker diarization module, a long-form multilingual ASR module, and a speaker-transcription fusion module. The diarization module is built upon DiariZen and produces speaker-homogeneous segments through local speaker activity estimation and global speaker clustering. The ASR module is based on Qwen3-Omni and generates multilingual transcriptions, while an external CTC-based alignment model provides precise word- and character-level timestamps. Finally, the fusion module combines diarization outputs with timestamped transcriptions to generate speaker-attributed STM outputs. Experimental results on the official evaluation set demonstrate the effectiveness of the proposed framework. The submitted system achieves a tcpMER of 15.41% and ranks second among all participating teams.
Chinese Translation
本文介绍了Transsion语音团队在MLC-SLM 2026挑战任务1中的提交，重点关注多语言对话语音的说话人归属转录。我们提出了一个级联框架，由三个组件组成：说话人分离模块、长文本多语言自动语音识别（ASR）模块和说话人转录融合模块。分离模块基于DiariZen构建，通过局部说话人活动估计和全局说话人聚类生成说话人同质段。ASR模块基于Qwen3-Omni，生成多语言转录，而一个外部基于CTC的对齐模型提供精确的单词和字符级时间戳。最后，融合模块将分离输出与带时间戳的转录结合，以生成说话人归属的STM输出。在官方评估集上的实验结果证明了所提框架的有效性。提交的系统实现了15.41%的tcpMER，并在所有参与团队中排名第二。
View on arXiv Download PDF AI Translation
cs.CL / 11 / 2609.20834

Towards Secure Cloud-Native Computing: Unveiling Kubernetes Misconfigurations with Large Language Models

迈向安全的云原生计算：利用大型语言模型揭示Kubernetes配置错误
Ghorab, Mostafa Anouar, Saied, Mohamed Aymen
Abstract
In the rapidly evolving landscape of cloud-native computing, Organizations are increasingly adopting infrastructure models that emphasize scalability, flexibility, and efficiency. Kubernetes has become the de facto standard for orchestrating containerized applications in these environments. However, the inherent complexity of cloud-native ecosystems introduces significant challenges, particularly in the form of misconfigurations that can compromise both security and performance. This study explores the potential of Large Language Models (LLMs) in identifying Kubernetes misconfigurations. We introduce a comprehensive taxonomy of common misconfiguration types, offering a structured framework to better understand and categorize these issues. Additionally, we conduct an empirical evaluation of state-of-the-art detection tools to benchmark their effectiveness. Furthermore, we analyze the Kubernetes objects most prone to misconfiguration and evaluate the severity of the identified issues. By leveraging advanced machine learning techniques, including LLMs, we provide novel insights into enhancing misconfiguration detection methodologies.
Chinese Translation
在快速发展的云原生计算环境中，组织越来越多地采用强调可扩展性、灵活性和效率的基础设施模型。Kubernetes已成为在这些环境中编排容器化应用的事实标准。然而，云原生生态系统固有的复杂性带来了重大挑战，特别是以配置错误的形式，这可能会危及安全性和性能。本研究探讨了大型语言模型（LLMs）在识别Kubernetes配置错误中的潜力。我们提出了一种常见配置错误类型的综合分类法，提供了一个结构化框架，以更好地理解和分类这些问题。此外，我们对最先进的检测工具进行了实证评估，以基准其有效性。此外，我们分析了最容易出现配置错误的Kubernetes对象，并评估了所识别问题的严重性。通过利用包括LLMs在内的先进机器学习技术，我们提供了增强配置错误检测方法的新见解。
View on arXiv Download PDF AI Translation
cs.CL / 12 / 2609.20835

A Generative Grammar Underlying the Voynich Manuscript, the Pastiche Hypothesis: Evidence from Large Language Models

一种生成语法支持的《沃伊尼奇手稿》，拼贴假说：来自大型语言模型的证据
Turenne, Nicolas
Abstract
Background: The Voynich Manuscript is a fifteenth-century codex written in an unknown script whose content remains undeciphered. Previous studies suggest that its statistical properties resemble those of natural languages, while its illustrations - primarily plants - recall medieval herbals. Methods: We present a multidisciplinary analysis combining probabilistic modeling, phonetic decomposition, rare-event detection, and multimodal image analysis, based on a newly transliterated corpus. Word- and letter-level distributions are modeled using position-dependent probabilistic grammars, while phonetic patterns are compared across Indo-European, Semitic, and Asian languages. Image-text alignment methods based on large language models are applied to identify potential botanical correspondences. Results: The results indicate that Voynich symbols behave as letters rather than syllabic units, while word-length distributions resemble syllabic structures. Phonetic analyses show closer alignment with consonant-heavy languages such as Hebrew or Arabic than with Indo-European languages. Probabilistic modeling reproduces Zipf-like distributions and reveals extremely low probabilities for repeated initial-letter sequences, indicating a structured imitation of natural language. Image analysis suggests strong correspondences between Voynich plant illustrations and those found in Pseudo-Apuleius herbals from the Mediterranean tradition, consistent with an imitation of medieval medicinal books. Perspectives: These findings support the hypothesis that the Voynich Manuscript follows a structured generative system combining linguistic regularities and herbal knowledge, and demonstrate the value of integrating probabilistic and AI-assisted approaches in the analysis of historical manuscripts.
Chinese Translation
背景：沃伊尼奇手稿是一部用未知文字书写的十五世纪手稿，其内容至今未被破译。先前的研究表明，其统计特性与自然语言相似，而其插图——主要是植物——则让人联想到中世纪的草药书。方法：我们基于新转录的语料库，进行了一项结合概率建模、语音分解、稀有事件检测和多模态图像分析的多学科研究。单词和字母级别的分布使用位置依赖的概率语法进行建模，同时对印欧语系、闪米特语系和亚洲语言的语音模式进行比较。基于大型语言模型的图像-文本对齐方法被应用于识别潜在的植物对应关系。结果：结果表明，沃伊尼奇符号表现得更像字母而非音节单位，而单词长度分布则类似于音节结构。语音分析显示，与希伯来语或阿拉伯语等辅音丰富的语言的对齐程度高于与印欧语言的对齐程度。概率建模重现了类似Zipf的分布，并揭示了重复初始字母序列的极低概率，表明存在对自然语言的结构化模仿。图像分析表明，沃伊尼奇植物插图与地中海传统中的伪阿普利乌斯草药书中的插图之间存在强烈的对应关系，这与对中世纪医学书籍的模仿一致。展望：这些发现支持了沃伊尼奇手稿遵循一种结合语言规律和草药知识的结构化生成系统的假说，并展示了在历史手稿分析中整合概率和人工智能辅助方法的价值。
View on arXiv Download PDF AI Translation
cs.CL / 13 / 2609.20836

PhysioBench: A Unified Benchmark for Physiological Signal Question Answering

PhysioBench：生理信号问答的统一基准
Li, Mengxuan, Chen, Junfa, Xia, Jinze, Chen, Yundan, Fan, Lixin, Liu, Ke, Shi, Keyue, Wang, Haishuai
Abstract
Physiological signals support diverse clinical and monitoring tasks, yet existing physiological signal foundation models typically require task-specific adaptation for each task. Natural language provides a common interface for specifying different prediction objectives, but the ability of current models to follow such instructions across physiological signal modalities remains insufficiently evaluated. To address this gap, we introduce PhysioBench, a unified benchmark for physiological signal question answering. PhysioBench harmonizes annotations from 22 public datasets into 61.4 million questions across 30 tasks. Each question-answer pair is grounded in a signal segment and traceable to its source annotation. We evaluate 21 representative models, including large language models, vision-language models, time-series language models, and physiological signal foundation models under three complementary settings. The results show that none of the evaluated models achieves consistently strong performance across physiological signal modalities and tasks. The incorporation of natural language supports unified prediction across tasks, although performance remains sensitive to question formulation. Beyond these findings, PhysioBench offers an extensible platform for fine-grained analysis and future research on physiological signal understanding. Our codes are available at https://github.com/Leanna97/PhysioBench.
Chinese Translation
生理信号支持多种临床和监测任务，但现有的生理信号基础模型通常需要针对每个任务进行特定的适应。自然语言提供了一种通用接口，用于指定不同的预测目标，但当前模型在不同生理信号模态下遵循此类指令的能力仍然评估不足。为了解决这一问题，我们引入了PhysioBench，一个生理信号问答的统一基准。PhysioBench将来自22个公共数据集的注释整合为6140万个问题，涵盖30个任务。每个问答对都基于一个信号片段，并可追溯到其源注释。我们在三种互补设置下评估了21个代表性模型，包括大型语言模型、视觉-语言模型、时间序列语言模型和生理信号基础模型。结果表明，评估的模型在生理信号模态和任务上均未能实现持续强劲的表现。自然语言的引入支持了跨任务的统一预测，尽管性能仍对问题的表述敏感。除了这些发现，PhysioBench还提供了一个可扩展的平台，用于对生理信号理解进行细致分析和未来研究。我们的代码可在 https://github.com/Leanna97/PhysioBench 获取。
View on arXiv Download PDF AI Translation
cs.CL / 14 / 2609.20838

From Generation to Detection: Exploration of Discourse Driven Scenario based LLM Generated Fake News

从生成到检测：基于话语驱动场景的LLM生成假新闻的探索
Özdemir, Zeynep, Osmanoğlu, Murat, Yiğit-Sert, Sevgi, Tanrıöver, Ömer Özgür, Ar, Yılmaz
Abstract
In this study, we examine how modern LLMs generate and detect fake news under controlled settings across four manipulation scenarios. These are open-ended generation, rewriting, manipulation prompts and attribute based prompts grounded in the journalistic discourse framework. Firstly, using seven widely adapted models, we created a synthetic fake news corpus with 14000 generated articles across these four scenarios. Then we analyzed its linguistic properties to assess how closely model-generated news resembles real news structurally and semantically. Finally, to evaluate detection performance, we conducted experiments where each model judges generated fake news, starting with a basic detection prompt and improved prompts developed through an iterative refinement process that extracts misleading patterns from real-fake pairs. Our results revealed substantial variation across models in both generating and detecting misinformation, demonstrated that the generation strategy strongly influences detectability, and show that the refined prompt does not improve and often harms detection performance. Therefore, the study provides a systematic assessment of LLMs detection capability of LLMs generated fake news across typical generation scenarios.
Chinese Translation
本研究考察了现代大语言模型（LLMs）在控制环境下如何生成和检测假新闻，涉及四种操控场景。这些场景包括开放式生成、重写、操控提示和基于属性的提示，均基于新闻话语框架。首先，我们使用七个广泛应用的模型，创建了一个包含14000篇生成文章的合成假新闻语料库，涵盖这四种场景。然后，我们分析了其语言特性，以评估模型生成的新闻在结构和语义上与真实新闻的相似程度。最后，为了评估检测性能，我们进行了实验，让每个模型判断生成的假新闻，起初使用基本检测提示，随后通过提取真实-假新闻对中的误导模式而开发的迭代改进提示。我们的结果显示，各模型在生成和检测虚假信息方面存在显著差异，表明生成策略对可检测性有强烈影响，并且改进的提示并未提升，反而常常损害检测性能。因此，本研究对LLMs在典型生成场景下检测LLMs生成假新闻的能力进行了系统评估。
View on arXiv Download PDF AI Translation
cs.CL / 15 / 2609.20839

Curriculum-Based Noise Adaptation for Phoneme-to-Text Reconstruction in Visual Speech Recognition

基于课程的噪声适应用于视觉语音识别中的音素到文本重建
Teng, Matthew Kit Khinn, Zhang, Haibo, Saitoh, Takeshi
Abstract
Phoneme-centric visual speech recognition reconstructs sentences from intermediate phoneme predictions, making overall recognition performance highly dependent on the robustness of the phoneme-to-text reconstruction model. Existing reconstruction approaches are commonly trained on clean phoneme sequences or synthetically corrupted inputs, leading to a mismatch between training conditions and the realistic phoneme prediction errors encountered during inference. To address this limitation, this paper proposes progressive error curriculum training (PECT). This curriculum learning framework progressively adapts a No Language Left Behind (NLLB)-based phoneme-to-text reconstruction model using synthetic phoneme perturbations, multi-domain pseudo-labels, and target-domain pseudo-labels generated by a visual speech recognizer. By gradually exposing the reconstruction model to increasingly realistic phoneme prediction errors, the proposed framework improves robustness while preserving sentence-reconstruction accuracy. Experiments on the LRS2 and LRS3 benchmarks demonstrate that PECT consistently improves reconstruction performance across multiple phoneme-based visual speech recognition frontends, including visual automatic speech recognition (V-ASR), point visual automatic speech recognition (PV-ASR), and head-pose-aware visual speech recognition (HP-VSR) variants. In particular, PECT reduces the word error rate (WER) of HP-VSR-FiLMFuse (L4) from 23.3% to 22.2% on LRS2 and reduces the WER of HP-VSR-ResFiLM from 30.3% to 29.7% on LRS3. Comprehensive ablation studies and qualitative analyses further demonstrate the effectiveness of progressively adapting the reconstruction model to realistic phoneme prediction errors. These results show that PECT provides an effective and generalizable curriculum learning strategy for phoneme-to-text reconstruction in phoneme-centric visual speech recognition.
Chinese Translation
以音素为中心的视觉语音识别通过中间音素预测重建句子，这使得整体识别性能高度依赖于音素到文本重建模型的鲁棒性。现有的重建方法通常在干净的音素序列或合成损坏的输入上进行训练，导致训练条件与推理过程中遇到的真实音素预测错误之间存在不匹配。为了解决这一局限性，本文提出了渐进错误课程训练（Progressive Error Curriculum Training, PECT）。该课程学习框架通过使用合成音素扰动、多域伪标签以及由视觉语音识别器生成的目标域伪标签，逐步调整基于无语言落下（No Language Left Behind, NLLB）的音素到文本重建模型。通过逐渐将重建模型暴露于越来越真实的音素预测错误中，所提出的框架在提高鲁棒性的同时保持了句子重建的准确性。在LRS2和LRS3基准上的实验表明，PECT在多个基于音素的视觉语音识别前端（包括视觉自动语音识别（Visual Automatic Speech Recognition, V-ASR）、点视觉自动语音识别（Point Visual Automatic Speech Recognition, PV-ASR）和头部姿态感知视觉语音识别（Head-Pose-Aware Visual Speech Recognition, HP-VSR）变体）中持续提高了重建性能。特别是，PECT将HP-VSR-FiLMFuse（L4）在LRS2上的字错误率（Word Error Rate, WER）从23.3%降低到22.2%，并将HP-VSR-ResFiLM在LRS3上的WER从30.3%降低到29.7%。全面的消融研究和定性分析进一步证明了逐步调整重建模型以适应真实音素预测错误的有效性。这些结果表明，PECT为音素中心的视觉语音识别中的音素到文本重建提供了一种有效且可推广的课程学习策略。
View on arXiv Download PDF AI Translation
cs.CL / 16 / 2609.20842

COAL-SQL: Coverage-Guided Augmentation and Failure-Driven Learning for Text-to-SQL Post-Training

COAL-SQL：基于覆盖指导的增强和失败驱动学习的文本到SQL后训练
Cai, Qifeng, Pan, Xuanguang, Liang, Hao, Xu, Chang, Zhang, Wentao
Abstract
Text-to-SQL translates natural-language questions into executable SQL queries, but open-source large language models still require task-specific post-training for complex, real-world SQL generation. Effective post-training requires both training data that cover the capabilities demanded by the target task and a learning strategy that enables the model to acquire them. Existing datasets provide valuable supervision but incompletely cover SQL structures, while augmentation methods typically expand data without identifying structural gaps. Moreover, supervised fine-tuning (SFT) or reinforcement learning (RL) alone cannot dynamically address weaknesses exposed during training. We propose COAL-SQL, a unified framework combining Coverage-Guided Augmentation (CGA) and Failure-Driven Learning (FDL). CGA uses greedy selection to identify SQL structures missing from the original dataset and constructs complementary examples, improving structural coverage. FDL retains GRPO as the main optimization objective while supplying targeted supervision for unsolved examples. At the step level, it applies SFT to verified reasoning traces generated by a strong LLM for accumulated failures. At the epoch level, it retrieves structurally related examples based on accumulated failures to create targeted practice, helping the model acquire the corresponding SQL capabilities. With only 12,600 distinct post-training examples, COAL-SQL achieves 64.9% execution accuracy on the BIRD development set and outperforms baselines trained at comparable scale. The code is available at https://github.com/TechNomad-ds/COAL-SQL.
Chinese Translation
文本到SQL将自然语言问题转换为可执行的SQL查询，但开源的大型语言模型在复杂的现实世界SQL生成方面仍需特定任务的后训练。有效的后训练需要既能覆盖目标任务所需能力的训练数据，又需一种学习策略使模型能够获取这些能力。现有数据集提供了有价值的监督，但对SQL结构的覆盖不够全面，而增强方法通常在不识别结构缺口的情况下扩展数据。此外，仅依靠监督微调（SFT）或强化学习（RL）无法动态解决训练过程中暴露的弱点。我们提出了COAL-SQL，一个结合了覆盖指导增强（CGA）和失败驱动学习（FDL）的统一框架。CGA通过贪婪选择识别原始数据集中缺失的SQL结构，并构建补充示例，从而改善结构覆盖。FDL保持GRPO作为主要优化目标，同时为未解决示例提供有针对性的监督。在步骤层面，它对由强大的LLM生成的经过验证的推理轨迹应用SFT，以应对积累的失败。在周期层面，它基于积累的失败检索结构相关示例，以创建有针对性的练习，帮助模型获取相应的SQL能力。仅使用12,600个不同的后训练示例，COAL-SQL在BIRD开发集上实现了64.9%的执行准确率，并超越了在可比规模下训练的基线模型。代码可在https://github.com/TechNomad-ds/COAL-SQL获取。
View on arXiv Download PDF AI Translation
cs.CL / 17 / 2609.20843

VISPATH: Visual-Intent-Guided Path Reasoning for Multimodal Knowledge Graph Question Answering

VISPATH：基于视觉意图引导的多模态知识图谱问答路径推理
Wu, Jinke, Li, Zhengpin, Jia, Mengzhe, Li, Yang, Zhang, Wentao
Abstract
Knowledge graph question answering (KGQA) enables models to answer natural-language questions through structured graph reasoning and has achieved substantial progress across many benchmarks and applications. Recently, multimodal KGQA (MM-KGQA) has attracted increasing attention because many questions require jointly using multimodal inputs and KG evidence. However, existing MM-KGQA methods typically use multimodal information only for starting entity grounding or evidence retrieval, after which multi-hop reasoning degenerates into text-only graph search. As a result, they cannot exploit multimodal cues that become important at intermediate hops. To address this limitation, we propose VISPATH, a visual-intent-guided path reasoning framework for MM-KGQA. VISPATH first identifies a reliable starting entity by combining multimodal grounding with graph-structural cues. It then performs intent-guided path discovery by recomputing hop-specific multimodal intent from the input, question, and current partial paths, so that each expansion is guided by the current reasoning state. The discovered paths are further refined through reasoning-chain pruning, which evaluates candidate paths as complete evidence chains based on their consistency with the question, reasoning sketch, and hop-specific intent. Finally, VISPATH checks whether the selected evidence is sufficient for answer generation. We further construct VISPATH-Bench, a benchmark for evaluating multimodal multi-hop reasoning over KGs, covering questions that require two to four hops over KG paths. Extensive experiments on VISPATH-Bench and three additional multimodal QA benchmarks show that VISPATH consistently outperforms strong baselines. Notably, with GPT-4o as the backbone, VISPATH surpasses GPT-5.4 on VISPATH-Bench, achieving a 10.6% relative improvement in average accuracy and a 13.1% improvement at 2-hop reasoning.
Chinese Translation
知识图谱问答（KGQA）使模型能够通过结构化图推理回答自然语言问题，并在许多基准和应用中取得了显著进展。最近，多模态知识图谱问答（MM-KGQA）引起了越来越多的关注，因为许多问题需要联合使用多模态输入和知识图谱证据。然而，现有的MM-KGQA方法通常仅在起始实体定位或证据检索中使用多模态信息，之后的多跳推理退化为仅基于文本的图搜索。因此，它们无法利用在中间跳跃中变得重要的多模态线索。为了解决这一局限性，我们提出了VISPATH，一种用于MM-KGQA的基于视觉意图引导的路径推理框架。VISPATH首先通过结合多模态定位和图结构线索来识别可靠的起始实体。然后，它通过从输入、问题和当前部分路径重新计算特定跳跃的多模态意图，执行意图引导的路径发现，以便每次扩展都由当前推理状态引导。发现的路径通过推理链修剪进一步精炼，该过程根据候选路径与问题、推理草图和特定跳跃意图的一致性，将其评估为完整的证据链。最后，VISPATH检查所选证据是否足以生成答案。我们进一步构建了VISPATH-Bench，这是一个用于评估知识图谱上多模态多跳推理的基准，涵盖需要在知识图谱路径上进行两到四次跳跃的问题。在VISPATH-Bench和另外三个多模态问答基准上的大量实验表明，VISPATH始终优于强基线。值得注意的是，以GPT-4o为基础，VISPATH在VISPATH-Bench上超越了GPT-5.4，平均准确率相对提高了10.6%，在2跳推理中提高了13.1%。
View on arXiv Download PDF AI Translation
cs.CL / 18 / 2609.20844

Boosting Deepresearch and LongContext Ability with Self-Generated Deepresearch Rollouts Traces

通过自生成的深度研究回滚轨迹提升深度研究和长上下文能力
Wang, Zihan, Wang, Hao, Jiang, Boyuan, Zhang, Yiqun, Feng, Shi, Yang, Xiaocui, Ye, Yiwen, Lin, Jianghang, Ji, Xiaozhong, Lin, Jinghao, Wu, Kai
Abstract
Deepresearch (DR) agents interact with real-world web environments through multi-turn search and visit, causing their contexts to grow rapidly over time. We observe that, even after DR Agentic Reinforcement Learning (DR-RL), 61.6% of the model's remaining prediction errors can still be attributed to insufficient long-context understanding, including longcontext hallucination and failures in cross-document evidence integration. It motivates us to further break the bottleneck of DR-RL by strengthening the model's long-context ability. However, effective LongContext training requires more than simply increasing context length. To bridge the data gap, we propose `DR Rollouts to LongContext-QA (DR-to-Long)'. The method repurposes DR-RL trajectories, which naturally contain search histories, visited webpages, evidence snippets, and final-answer supervision. It then replaces the compact snippets and webpage summaries in each trajectory with the full contents of their corresponding URLs, producing substantially longer multi-document contexts while preserving the original evidence relationships. Building on DR-to-Long, we introduce DLD (DR -> LongQA -> DR)-RL. DLD-RL first performs a short DR-RL stage to collect rollout trajectories, which are then converted into LongQA instances at zero annotation cost. The model is subsequently optimized with LongQA-RL to strengthen LongContext ability, followed by full DR-RL to continue improving its DR capability. Experiments show that DLD-RL outperforms standard DR-RL by 7.3% on three Deepresearch benchmarks and improves performance by 13.5% on three long-context benchmarks.
Chinese Translation
深度研究（Deepresearch, DR）代理通过多轮搜索和访问与真实世界的网络环境进行交互，导致其上下文随时间迅速增长。我们观察到，即使在深度研究代理强化学习（DR-RL）之后，模型剩余的预测错误中仍有61.6%可以归因于对长上下文理解不足，包括长上下文幻觉和跨文档证据整合失败。这促使我们进一步打破DR-RL的瓶颈，通过增强模型的长上下文能力。然而，有效的长上下文训练不仅仅需要增加上下文长度。为了弥补数据差距，我们提出了“DR回滚到长上下文问答（DR Rollouts to LongContext-QA, DR-to-Long）”。该方法重新利用了DR-RL轨迹，这些轨迹自然包含搜索历史、访问的网页、证据片段和最终答案的监督。然后，它用相应URL的完整内容替换每个轨迹中的紧凑片段和网页摘要，从而生成显著更长的多文档上下文，同时保留原始证据关系。在DR-to-Long的基础上，我们引入了DLD（DR -> LongQA -> DR）-RL。DLD-RL首先执行短期DR-RL阶段以收集回滚轨迹，然后在零标注成本下将其转换为LongQA实例。接着，模型通过LongQA-RL进行优化，以增强长上下文能力，最后继续进行全面的DR-RL以进一步提升其DR能力。实验表明，DLD-RL在三个深度研究基准上比标准DR-RL提高了7.3%，并在三个长上下文基准上提高了13.5%的性能。
View on arXiv Download PDF AI Translation
cs.CL / 19 / 2609.20845

Reading Less While Writing: A Closed-Form Bandwidth Dial for Streaming Multimodal Decoders

写作时减少阅读：用于流式多模态解码器的闭式带宽调节器
Mehmood, Yasir, Javed, Kashif
Abstract
A decoder that turns video or audio into text conventionally consumes the entire input before emitting a word. Offline this is merely more than the task requires; live it is impossible, since a caption cannot wait for a match to end. Streaming systems bolt on a fixed rule such as wait-$k$, which waits for the same number of input tokens before every word, regardless of the input's length or pace. We replace the fixed offset with ZENDAYA, a schedule governed by a single continuous parameter $\gamma$. It makes the visible source prefix a closed-form function of generation progress, scaled by the input's own predicted length, so an ordinary offline decoder and a real-time streaming decoder become two endpoints of one family rather than separate models. The same scalar fixes, in closed form, the mean fraction of source consumed per emitted word, $\bar{E}(\gamma) \approx 1/(1+\gamma)$, making it at once a latency dial and an interpretable budget. We prove a structural dependency theorem: under any schedule fixed in advance and non-decreasing, no emitted token can depend on input that has not yet arrived. The guarantee holds for trained and untrained weights alike, and extends to unbounded streams under arbitrary asynchronous arrival. The empirical result is counterintuitive: seeing less can produce better text, because a flood of source dilutes attention exactly when the model has the least of its own output to anchor on. Trained from scratch across two modalities and three public corpora (Charades-STA, ActivityNet Captions, LibriHeavy), a compact 29M-parameter decoder matches or beats the fixed schedule while reading less of the source, with the sharpest gains at the lowest latencies, where a fixed offset collapses. Streaming METEOR gains are statistically significant on all three corpora.
Chinese Translation
传统上，将视频或音频转换为文本的解码器在发出一个单词之前需要消耗整个输入。在离线情况下，这仅仅是超出了任务的需求；而在实时情况下，这是不可能的，因为字幕不能等待匹配结束。流式系统附加了一个固定规则，例如 wait-$k$，在每个单词之前等待相同数量的输入标记，而不考虑输入的长度或速度。我们用 ZENDAYA 替代了固定偏移量，这是一种由单个连续参数 $eta$ 控制的调度。它使得可见的源前缀成为生成进度的闭式函数，按输入自身预测的长度进行缩放，因此普通的离线解码器和实时流式解码器成为同一家族的两个端点，而不是独立的模型。相同的标量在闭式形式下修正了每发出一个单词所消耗的源的平均比例 $ar{E}(eta) ext{≈} 1/(1+eta)$，使其同时成为一个延迟调节器和一个可解释的预算。我们证明了一个结构依赖定理：在任何预先固定且非递减的调度下，任何发出的标记都不能依赖于尚未到达的输入。该保证适用于训练和未训练的权重，并扩展到在任意异步到达下的无界流。经验结果是反直觉的：看到更少可以产生更好的文本，因为源的洪流在模型最少自己的输出可以依靠时稀释了注意力。在两个模态和三个公共语料库（Charades-STA、ActivityNet Captions、LibriHeavy）上从零开始训练的紧凑型 2900 万参数解码器在阅读更少源的情况下与固定调度相匹配或超越，在最低延迟时获得最显著的提升，而固定偏移在此时崩溃。流式 METEOR 在所有三个语料库上的增益具有统计显著性。
View on arXiv Download PDF AI Translation
cs.CL / 20 / 2609.20846

Rewarding Efficient Reasoning Improves Abstention on Underspecified Tasks in Reasoning Models

奖励高效推理改善推理模型在不明确任务上的弃权表现
Tsvilodub, Polina, Höth, Max, Franke, Michael, Deiseroth, Björn, Kauf, Carina
Abstract
While modern large reasoning models (LRMs) excel at providing correct answers in many tasks, we provide additional evidence for the observation that they often struggle with a critical capability: knowing when to abstain from answering. We analyze this gap by comparing LRM behavior to results from a human study, revealing that human reasoning effort on unanswerable tasks is upper-bounded by answerable tasks, whereas LRMs waste computational resources by generating longer Chains of Thought (CoTs) on unanswerable than on answerable prompts. To overcome this inefficiency, we take inspiration from a resource-rational perspective on human cognition and introduce a novel GRPO reward that encourages efficient reasoning about whether the task contains all the information needed to solve it. Fine-tuning several 4B LRMs with this reward leads to human-like abstention performance gains (+12.8% on average) while retaining answering capabilities and boosting the models' efficiency (44% shorter CoTs on average).
Chinese Translation
尽管现代大型推理模型（LRMs）在许多任务中能够提供正确答案，但我们提供了额外证据表明它们在一个关键能力上常常表现不佳：知道何时应当弃权。我们通过将LRM的行为与人类研究的结果进行比较，分析了这一差距，揭示了人类在无法回答的任务上的推理努力受到可回答任务的上限限制，而LRMs在无法回答的提示上生成的思维链（Chains of Thought, CoTs）比在可回答提示上更长，从而浪费了计算资源。为了克服这种低效，我们从人类认知的资源理性视角获得灵感，提出了一种新颖的GRPO奖励，鼓励高效推理以判断任务是否包含解决所需的所有信息。通过这种奖励对多个4B LRM进行微调，导致了类似人类的弃权表现提升（平均提高12.8%），同时保持了解答能力并提高了模型的效率（平均缩短44%的CoTs）。
View on arXiv Download PDF AI Translation
cs.CL / 21 / 2609.20847

Reading Anxiety or Reading the Label? Comparing Fine-Tuned and Frontier Models for Anxiety Detection on Social Media

阅读焦虑还是阅读标签？比较社交媒体上焦虑检测的精细调优模型与前沿模型
Huynh, Cris, Pham, Arlene
Abstract
Anxiety is among the most common mental health conditions, and people often write about it online well before seeking clinical help. Practitioners building detection tools face a concrete choice: call a frontier commercial model, fine-tune a smaller model in-house, or deploy a conventional classifier. We compare six conditions spanning all three on a held-out Reddit test set under a single controlled protocol. We also identify a confound in how this task is evaluated. In the corpus used here, 69.3% of anxiety-labelled posts contain the word "anxiety" or a variant, roughly twice the rate of comparable conditions, so a classifier can score well by keyword matching rather than by modelling the language of the condition. We therefore evaluate every model twice, on original text and with those terms deleted, and report the difference as lexical dependence. A frontier model leads on anxiety F1 (0.846), but a 110M-parameter domain-adapted encoder reaches 0.831 with no external API dependency, and mental-health domain pretraining accounts for only 0.7 of those points. Lexical dependence spans 8.6 to 25.4 points and does not track model capability: the LoRA fine-tuned 3B model is the most keyword-dependent condition tested, above even a TF-IDF classifier, while the frontier zero-shot model is the least. Published figures on this corpus are therefore upper bounds, and the inflation is largest for the fine-tuned models such figures typically report.
Chinese Translation
焦虑是最常见的心理健康问题之一，人们往往在寻求临床帮助之前就会在网上谈论这一问题。构建检测工具的从业者面临一个具体的选择：调用一个前沿商业模型、在内部精细调优一个较小的模型，或部署一个传统分类器。我们在一个控制协议下，对一个保留的Reddit测试集比较了涵盖这三种选择的六种情况。我们还识别出在评估此任务时存在的一个混淆因素。在这里使用的语料库中，69.3%的焦虑标签帖子包含“anxiety”或其变体，约为可比条件的两倍，因此分类器可以通过关键词匹配而不是通过建模条件的语言来获得良好的得分。因此，我们对每个模型进行了两次评估，一次是原始文本，另一次是删除这些术语，并将差异报告为词汇依赖性。前沿模型在焦虑F1得分上领先（0.846），但一个110M参数的领域适应编码器在没有外部API依赖的情况下达到了0.831，而心理健康领域的预训练仅占其中0.7的得分。词汇依赖性范围在8.6到25.4分之间，并不与模型能力相关：LoRA精细调优的3B模型是测试中对关键词依赖性最强的条件，甚至高于TF-IDF分类器，而前沿零样本模型则是最弱的。因此，关于该语料库的已发布数据是上限，而这种膨胀在精细调优模型中最大，这些模型通常报告的得分。
View on arXiv Download PDF AI Translation
cs.CL / 22 / 2609.20849

Enhancing Audio Reasoning via Semantic Summary Prediction

通过语义摘要预测增强音频推理
Bonzi, Francesco, Mousavi, Pooneh, Subakan, Cem, Ravanelli, Mirco
Abstract
Large Audio Language Models (LALMs) perform well on complex question answering but often show a reasoning gap, where explicit Chain-of-Thought (CoT) reduces accuracy compared to direct answers. We hypothesize that long reasoning sequences shift attention away from the audio input. To address this, we propose SPARE (Semantic Prediction for Audio REasoning), which introduces a register token aligned with the final conclusion using a cosine similarity loss with a Sentence-BERT embedding. This conditions the model's latent space with the target semantic goal before reasoning begins. Experiments on MMAU and MMAR with SALMONN show improved zero-shot reasoning and stronger early attention to audio without additional inference cost.
Chinese Translation
大型音频语言模型（LALMs）在复杂问题回答中表现良好，但往往存在推理差距，即显式的思维链（Chain-of-Thought, CoT）相较于直接答案降低了准确性。我们假设，较长的推理序列会使注意力偏离音频输入。为了解决这个问题，我们提出了SPARE（语义预测用于音频推理），该方法引入了一个与最终结论对齐的注册令牌，使用余弦相似度损失与Sentence-BERT嵌入相结合。这在推理开始之前就为模型的潜在空间设定了目标语义目标。在MMAU和MMAR数据集上的实验表明，使用SALMONN的方法在零-shot推理中表现出改善，并且在不增加推理成本的情况下，对音频的早期注意力更强。
View on arXiv Download PDF AI Translation
cs.CL / 23 / 2609.20850

MME-Safety: A Fine-grained Benchmark for Safety Evaluation of MLLMs

MME-Safety：多模态大型语言模型安全评估的细粒度基准
Lyu, Yueming, Shi, Yilian, Tan, Haoxiang, Zou, Linzhuang, Wang, Qihao, Yu, Guihua, Qin, Jie, Gao, Xin, Si, Chenyang, Dong, Jing, Shan, Caifeng
Abstract
While Multimodal Large Language Models (MLLMs) show remarkable advancements, their cross-modal capabilities introduce complex vulnerabilities that easily bypass unimodal filters. Existing benchmarks lack fine-grained intent-related annotations and rely on unidimensional metrics, hindering comprehensive robustness evaluation. To address this, we propose MME-Safety, a rigorously verified benchmark featuring a unique four-dimensional annotation schema that categorizes risk scenarios, harm severity, and modality-specific stealth levels. Furthermore, we introduce a hierarchical evaluation framework to assess fundamental response reliability, actual risk exposure, and the structural integrity of defensive behaviors. Extensive zero-shot evaluations across 17 state-of-the-art MLLMs provide a comprehensive safety profile of current multimodal systems. Our analysis systematically investigates cross-modal input configurations and uncovers safety implications associated with Chain-of-Thought (CoT) reasoning. These multifaceted findings underscore the urgent need for robust, reasoning-aware safety alignment in the multimodal landscape.
Chinese Translation
尽管多模态大型语言模型（MLLMs）展现了显著的进步，但其跨模态能力引入了复杂的脆弱性，容易绕过单模态过滤器。现有基准缺乏细粒度的意图相关注释，并依赖于单维度指标，阻碍了全面的鲁棒性评估。为了解决这一问题，我们提出了MME-Safety，这是一个经过严格验证的基准，采用独特的四维注释框架，对风险场景、危害严重性和特定模态的隐蔽程度进行分类。此外，我们引入了一个分层评估框架，以评估基本响应的可靠性、实际风险暴露和防御行为的结构完整性。在17个最先进的MLLMs上进行的广泛零-shot评估提供了当前多模态系统的全面安全概况。我们的分析系统地研究了跨模态输入配置，并揭示了与思维链（Chain-of-Thought, CoT）推理相关的安全隐患。这些多方面的发现强调了在多模态环境中迫切需要强大的、关注推理的安全对齐。
View on arXiv Download PDF AI Translation
cs.CL / 24 / 2609.20902

Generative Artificial Intelligence Chatbots for Motivational Interviewing: A Scoping Review From System Design to Intervention Outcomes

用于动机访谈的生成性人工智能聊天机器人：从系统设计到干预结果的范围审查
Hu, Runze, Kong, Jingqi, Yang, Yang, Yang, Yihang, Liu, Jingyao, Tang, Haizhou, Zhang, Shanghang, Liu, Zheng
Abstract
Motivational interviewing (MI) is a collaborative approach to elicit autonomous motivation for health behavior change. Generative AI (GenAI) offers new ways to deliver MI via conversational systems, but evidence on their design, assessment, and translation into interventions remains fragmented. This scoping review characterized evidence on GenAI-MI chatbots across system design, safety, MI quality, user perceptions, and intervention outcomes. We conducted a PRISMA-ScR scoping review. Nine datasets were searched for studies published or publicly available from January 1, 2015 to June 2, 2026 that used GenAI to generate MI chatbot responses or counselor utterances. Data were extracted using a predefined framework and synthesized descriptively. Forty-seven reports (48 studies) were included. Twenty (41.7%) focused on system design without direct participant use; 28 (58.3%) involved direct interaction. Most systems were text based and disembodied; 23 (47.9%) incorporated dynamic adaptation. Safety measures were unevenly reported. Among studies with direct use, 21/28 (75.0%) reported informed consent or user education. Thirty (62.5%) assessed MI quality, generally suggesting MI-consistent interactions. User perceptions were favorable, especially empathy, usability, helpfulness, and intention to use, though measures were heterogeneous. Eighteen (37.5%) reported intervention outcomes, mostly after a single session. Positive findings were more consistent for short-term motivation than sustained behavioral or functional change. GenAI-MI chatbots can deliver MI-consistent interactions perceived favorably, but evidence for sustained behavioral or functional change is limited. Future research should strengthen runtime safety monitoring, standardize MI quality assessment, and use longer-term comparative designs with behavioral and functional outcomes.
Chinese Translation
动机访谈（MI）是一种协作方法，旨在引发自主的健康行为改变动机。生成性人工智能（GenAI）为通过对话系统提供动机访谈提供了新的方式，但关于其设计、评估和转化为干预的证据仍然零散。本范围审查对GenAI-MI聊天机器人在系统设计、安全性、动机访谈质量、用户感知和干预结果方面的证据进行了特征描述。我们进行了PRISMA-ScR范围审查。搜索了九个数据集，查找2015年1月1日至2026年6月2日期间发表或公开可用的研究，这些研究使用GenAI生成动机访谈聊天机器人的响应或咨询师的发言。数据使用预定义框架提取并进行描述性综合。共纳入47份报告（48项研究）。其中20项（41.7%）专注于系统设计而没有直接参与者使用；28项（58.3%）涉及直接互动。大多数系统基于文本且无实体；23项（47.9%）包含动态适应。安全性措施报告不均。对于直接使用的研究，21/28项（75.0%）报告了知情同意或用户教育。30项（62.5%）评估了动机访谈质量，通常表明与动机访谈一致的互动。用户感知总体上是积极的，尤其是在同理心、可用性、有用性和使用意图方面，尽管测量标准不一。18项（37.5%）报告了干预结果，大多数是在单次会话后。积极结果在短期动机方面更为一致，而在持续的行为或功能改变方面则有限。GenAI-MI聊天机器人能够提供与动机访谈一致的互动，并被积极感知，但在持续的行为或功能改变方面的证据有限。未来的研究应加强运行时安全监测，标准化动机访谈质量评估，并采用更长期的比较设计，关注行为和功能结果。
View on arXiv Download PDF AI Translation
cs.CL / 25 / 2609.20945

$\mu^2$-Bench: A Multilingual Machine Unlearning Benchmark

$$-基准：多语言机器遗忘基准
Hwang, Kyomin, Kim, Hyeonjin, Lee, Hyunho, Kim, Yearim, Song, Yeji, Kwak, Nojun
Abstract
Undesired information such as harmful content and private data propagates through Multilingual Large Language Models (LLMs) via direct training and indirect cross-linguistic spread. Multilingual Machine Unlearning (MMU) aims to remove such information, yet its evaluation remains underexplored, leaving unclear whether unlearning truly eliminates target knowledge across all languages. To bridge this gap, we introduce $\mu^2$-Bench, an MMU benchmark that simulates the full pipeline of memorization, unlearning, and evaluation across diverse languages. It 1) spans a broad set of languages, 2) evaluates on both training and hold-out languages, and 3) assesses knowledge as dispersed across multiple languages. We show that successful MMU requires methods that reflect multilingual characteristics, and conduct analysis to provide deeper insights into MMU.
Chinese Translation
不希望的信息，如有害内容和私人数据，通过多语言大型语言模型（LLMs）在直接训练和间接跨语言传播中传播。多语言机器遗忘（MMU）旨在去除此类信息，但其评估仍未得到充分探讨，尚不清楚遗忘是否真正消除了所有语言中的目标知识。为填补这一空白，我们引入了$$-基准，这是一个MMU基准，模拟了跨多种语言的记忆、遗忘和评估的完整流程。它 1) 涉及广泛的语言集合，2) 在训练语言和保留语言上进行评估，3) 评估知识在多种语言中的分散情况。我们展示了成功的MMU需要反映多语言特征的方法，并进行分析以提供对MMU的更深入见解。
View on arXiv Download PDF AI Translation
cs.CL / 26 / 2609.21075

Aligning with Lived Experience: Heterogeneous Benefits of Fine Tuning in Mental Health Support Generation

与生活经验对齐：心理健康支持生成中的细调异质效益
Chandra, Mohit, Kim, Nabin, Min, Eli, Sawant, Aamogh, Sutar, Tanmay, De Choudhury, Munmun
Abstract
As access to professional mental healthcare remains limited, many individuals turn to online platforms such as Reddit to seek peer support situated within human lived experience. However, a significant portion of such queries go unanswered, presenting an opportunity for using Large Language Models (LLMs) to fill this gap. While LLMs have demonstrated strong performance on clinical benchmarks, their ability to generate lived-experience informed and community-aligned peer support is underexplored. Addressing this gap, we introduce the COmmunity-centered Peer Engaged Support (COPES) dataset and a three-axis evaluation framework to assess LLM alignment with community perspectives to mental health support seeking queries. Evaluating zero-shot and post-trained (SFT and DPO) models, we show that post-training on COPES significantly improves Strategy Alignment (>50% for general-purpose models) and alignment in Emotion & Tone. However, we also observe that such improvements are heterogeneous and alignment improvements vary significantly across subreddits and requested coping strategies. Furthermore, post-training induces distributional shifts, heavily favoring problem-focused recommendations while suppressing emotion-focused strategies. Together, this work shows that while curating community-driven data improves the alignment of LLM responses, model performance remains disparate across distinct sub-communities and specific mental health needs.
Chinese Translation
由于专业心理健康护理的获取仍然有限，许多人转向在线平台如Reddit寻求基于人类生活经验的同伴支持。然而，许多此类查询未能得到回应，这为使用大型语言模型（LLMs）填补这一空白提供了机会。尽管LLMs在临床基准测试中表现出色，但它们生成基于生活经验且与社区对齐的同伴支持的能力尚未得到充分探索。为了解决这一问题，我们引入了以社区为中心的同伴参与支持（COmmunity-centered Peer Engaged Support，COPES）数据集和一个三轴评估框架，以评估LLMs在心理健康支持寻求查询中与社区观点的对齐情况。通过评估零样本和后训练（SFT和DPO）模型，我们显示在COPES上进行后训练显著提高了策略对齐（对于通用模型超过50%）以及情感与语调的对齐。然而，我们也观察到这些改进是异质的，不同子版块和请求的应对策略之间的对齐改进差异显著。此外，后训练引发了分布转变，严重偏向于问题聚焦的建议，同时抑制了情感聚焦的策略。总体而言，这项工作表明，尽管策划以社区为驱动的数据改善了LLM响应的对齐，但模型性能在不同子社区和特定心理健康需求之间仍然存在差异。
View on arXiv Download PDF AI Translation
cs.CL / 27 / 2609.21094

Geometry of Values: Task Vector Composition for Ethical Preference Alignment in Language Models

价值的几何：语言模型中伦理偏好的任务向量组合
Agarwal, Utkarsh, Choudhury, Monojit
Abstract
Large Language Models (LLMs) are increasingly deployed in applications that must weigh clashing moral values, yet even strong models exhibit hidden biases and brittle instruction-following across languages. We introduce a 12,000-instance dataset of two-option dilemmas covering pairwise three value conflicts: Honesty vs. Justice, Justice vs. Autonomy, and Autonomy vs. Honesty, along with their translations into Hindi, Arabic, Spanish, and Chinese, to probe cross-lingual behavior. Benchmarking on GPT-5-mini reveals that it consistently favors Honesty over Autonomy across all five languages when no policy is given. The Llama-3.2-1/3B models exhibit strong first-option bias; however, both plain fine-tuning and Direct Preference Optimization fine-tuning effectively remove this bias, increasing accuracy to greater than 98%. In order to decouple the effect of learning correlations in the dataset from abstract values, we propose a task vector transfer based experiment where after computing the task vectors for a direction of value preference we orthogonalize it with respect to the general instruction following vector. Our experiment shows that this method is effective in isolating the direction of the specific value preference that can successfully be used to conduct task arithmetic to obtain a model with the opposite stance.
Chinese Translation
大型语言模型（LLMs）越来越多地应用于需要权衡冲突道德价值的场景，然而即使是强大的模型也表现出隐藏的偏见和在多语言环境中脆弱的指令遵循能力。我们引入了一个包含12,000个实例的数据集，涵盖了三种价值冲突的成对两选项困境：诚实与正义、正义与自主，以及自主与诚实，并提供其在印地语、阿拉伯语、西班牙语和中文中的翻译，以探讨跨语言行为。在对GPT-5-mini的基准测试中，我们发现该模型在没有给定政策的情况下，在所有五种语言中始终偏向诚实而非自主。Llama-3.2-1/3B模型表现出强烈的首选偏见；然而，普通微调和直接偏好优化微调都有效消除了这种偏见，使准确率提高到98%以上。为了将数据集中学习到的相关性效应与抽象价值解耦，我们提出了一种基于任务向量转移的实验方法，在计算出某一价值偏好的任务向量后，我们将其与一般指令遵循向量正交化。我们的实验表明，这种方法在隔离特定价值偏好的方向方面是有效的，可以成功用于进行任务算术，以获得持相反立场的模型。
View on arXiv Download PDF AI Translation
cs.CL / 28 / 2609.21117

From Task Success to Productive Success: Evaluating Human-AI Collaboration by Quality and Cost

从任务成功到生产性成功：通过质量和成本评估人机协作
Imai, Saki, İnan, Mert, Alikhani, Malihe
Abstract
AI productivity is often measured by task completion time, economic value, or improvements in outcome quality. However, these measures usually treat collaboration as a black box where they capture what output was produced, but not the interaction cost required to produce it. Motivated by economics literature, we introduce a productivity-oriented framework for evaluating human-AI collaboration as outcome quality relative to interaction cost. Across two datasets spanning four tasks, we show that: (1) sessions with identical quality ratings can differ by up to 70 times in interaction cost; (2) quality-cost relationships vary by task, with some tasks rewarding extended interaction and others favoring fast convergence; (3) subjective user ratings are not reliable substitutes for productivity; and (4) productive sessions are characterized by agents probing earlier and users spending less effort repairing the interaction. By distinguishing productive success from costly success, our framework makes interactional cost visible and shows how dialogue analysis can inform the evaluation and design of AI systems.
Chinese Translation
人工智能的生产力通常通过任务完成时间、经济价值或结果质量的改善来衡量。然而，这些衡量标准通常将协作视为一个黑箱，只捕捉所产生的输出，而未考虑产生该输出所需的交互成本。受到经济学文献的启发，我们提出了一个以生产力为导向的框架，用于评估人机协作的结果质量与交互成本的关系。在涵盖四个任务的两个数据集中，我们展示了： (1) 具有相同质量评级的会话，其交互成本可以相差多达70倍； (2) 质量与成本的关系因任务而异，有些任务奖励延长的交互，而其他任务则偏向快速收敛； (3) 主观用户评级并不是生产力的可靠替代品； (4) 生产性会话的特点是代理更早地进行探测，用户在修复交互时花费的精力更少。通过将生产性成功与成本成功区分开来，我们的框架使交互成本变得可见，并展示了对话分析如何为人工智能系统的评估和设计提供信息。
View on arXiv Download PDF AI Translation
cs.CL / 29 / 2609.21145

Scaling Forced Alignment to End-User Devices

将强制对齐扩展到终端用户设备
Sorenson, Lawry, Crandall, Michael, Ringger, Eric K., Richardson, Stephen D.
Abstract
The Viterbi algorithm has been previously used to perform forced alignment of audio to text to mine training data from online resources. However, many existing implementations have quadratic time and space complexity, scaling poorly to long input sequences. We propose two optimizations to address this issue. First, we apply the Hirschberg algorithm to perform the alignment in place using linear memory. Second, we model the alignment between speech and text as a constrained random walk, allowing us to prune the search space with arbitrary confidence while accounting for transcription errors. The Hirschberg optimization reduces memory usage from 140 GB to 5 MB for three-hour inputs while producing identical alignments in one-third the time of torchaudio when both run on a CPU. We achieve an additional 2x speedup with pruning on inputs longer than 20 minutes while preserving alignment accuracy in more than 98% of tested cases.
Chinese Translation
维特比算法（Viterbi algorithm）以前被用于将音频与文本进行强制对齐，以从在线资源中挖掘训练数据。然而，许多现有的实现具有平方时间和空间复杂度，难以扩展到较长的输入序列。我们提出了两项优化来解决这个问题。首先，我们应用赫希伯格算法（Hirschberg algorithm）在使用线性内存的情况下进行就地对齐。其次，我们将语音与文本之间的对齐建模为受限随机游走（constrained random walk），使我们能够在考虑转录错误的同时，以任意置信度修剪搜索空间。赫希伯格优化将三小时输入的内存使用量从140 GB减少到5 MB，同时在CPU上运行时产生与torchaudio相同的对齐结果，所需时间仅为其三分之一。对于超过20分钟的输入，我们在保持对齐准确率超过98%的情况下，实现了额外的2倍加速。
View on arXiv Download PDF AI Translation
cs.CL / 30 / 2609.21154

CoLearn: An Agentic Tutor that Learns its Learner in a Human--AI Co-Learning Loop

CoLearn：一个在人与人工智能共学习循环中学习学习者的自主导师
He, Kailai, Wu, Zhihao, Zhang, Linhai, Zhao, Runcong, He, Yulan, Li, Jiazheng
Abstract
Good tutoring adapts to the individual: it tracks what a learner knows, notices why they go wrong, and asks the next question that will help most. Most deployed tutoring tools instead serve fixed item banks and treat a wrong answer as a single bit of signal. We present CoLearn, an interactive, agentic tutor that supports an iterative tutoring loop: the learner practises, and the system builds an evidence-grounded memory of the learner's mastery and misconceptions. This memory is updated as evidence accumulates and is used to generate the next personalised question. CoLearn has three components: (i) a persistent learner-state memory that updates per-topic mastery with a soft-evidence variant of Bayesian Knowledge Tracing, where a large language model acts as a continuous observation function; (ii) adaptive question generation that targets the learner's weakest topic and recurring misconceptions; and (iii) an evidence view that makes personalisation visible and testable through live progress visualisation and blind A/B comparison. In blind A/B evaluation, questions conditioned on this memory are preferred over non-personalised ones 68-69% of the time, and in persona simulations with hidden ground-truth mastery the agent's belief converges toward the learner's true mastery.
Chinese Translation
良好的辅导能够适应个体：它跟踪学习者的知识，注意他们出错的原因，并提出最有助于学习的下一个问题。大多数已部署的辅导工具则提供固定的题库，将错误答案视为单一的信号。我们提出了CoLearn，一个互动的自主导师，支持迭代的辅导循环：学习者进行练习，系统建立学习者掌握情况和误解的基于证据的记忆。随着证据的积累，这种记忆会不断更新，并用于生成下一个个性化的问题。CoLearn有三个组成部分：（i）一个持久的学习者状态记忆，通过贝叶斯知识追踪的软证据变体更新每个主题的掌握情况，其中一个大型语言模型充当连续观察函数；（ii）针对学习者最薄弱主题和反复出现的误解的自适应问题生成；（iii）一个证据视图，通过实时进度可视化和盲法A/B比较，使个性化变得可见和可测试。在盲法A/B评估中，基于该记忆生成的问题比非个性化问题更受欢迎，比例为68-69%；在隐藏真实掌握情况的人物模拟中，代理的信念趋向于学习者的真实掌握水平。
View on arXiv Download PDF AI Translation
cs.CL / 31 / 2609.21179

Not All Irregularity Is Equal: Causally Isolating a Rare Failure Mode in Japanese Morphological Inflection

不所有的不规则性都是相同的：在日本形态变化中因果隔离一种罕见的失败模式
Zhang, Wen
Abstract
Neural morphological generation systems often achieve high aggregate accuracy on benchmark datasets, yet such performance can conceal systematic errors clustered in rare morphological subclasses. We present an orthography-aware diagnosis of Japanese past-tense verb inflection, treating hiragana not merely as a transcriptional medium but as a representational system that encodes morphophonological structure. Using two character-level Transformer architectures evaluated across five random seeds, we show that although both systems exceed 97% aggregate accuracy, a single structurally specific irregular subtype, verbs whose stems end in /e/ and require gemination before the past-tense suffix and make up fewer than 1% of the data, accounts for a disproportionate 30-43% share of residual errors and contributes roughly 34-48x its prevalence to total errors. We then move from diagnosis to causal isolation: controlled ablation experiments show that removing this subtype alone produces larger accuracy gains than removing all irregular verbs combined. These findings indicate that error concentration in neural morphological learning is not driven by irregularity per se, but by the interaction between extreme low-frequency morphological patterns and specific orthographic processes. We argue that morphological evaluation should incorporate fine-grained subclass analysis, and discuss implications for data-efficient, developmentally plausible language model pretraining.
Chinese Translation
神经形态生成系统在基准数据集上通常能够实现高总体准确率，但这种表现可能掩盖了在罕见形态子类中聚集的系统性错误。我们提出了一种关注正字法的日本过去时动词变化诊断，将平假名视为不仅仅是转录媒介，而是编码形态音位结构的表征系统。使用两种字符级Transformer架构，并在五个随机种子上进行评估，我们展示了尽管这两种系统的总体准确率均超过97%，但一种结构上特定的不规则子类型，即词干以/e/结尾并在过去时后缀之前需要双重化的动词，占数据的不到1%，却占据了不成比例的30-43%的剩余错误，并对总错误贡献了大约34-48倍于其普遍性的比例。随后，我们从诊断转向因果隔离：控制性消融实验表明，仅去除这一子类型所带来的准确率提升大于去除所有不规则动词的总和。这些发现表明，神经形态学习中的错误集中并非由不规则性本身驱动，而是由极低频形态模式与特定正字法过程之间的相互作用所导致。我们认为形态评估应纳入细粒度子类分析，并讨论了对数据高效、发展上合理的语言模型预训练的影响。
View on arXiv Download PDF AI Translation
cs.CL / 32 / 2609.21187

When Better Turns Do Not Make Better Agents: Diagnosing the Gap Between Next-Turn Metrics and Workflow Success

当更好的决策并不意味着更好的代理：诊断下一步指标与工作流成功之间的差距
Laskar, Md Tahmid Rahman, Fu, Xue-Yong, Singh, Gundeep, Chang, Karol, Sanders, Kevin, Zong, Shi, Habib, Tania, Tremblay, Julien Bouvier, Gardiner, Shayna, Saini, Harsh, Lee, Matthias, Khasanova, Elena, McNamara, Quinten, TN, Shashi Bhushan
Abstract
Agent models are frequently evaluated one decision at a time, where the model predicts the next action based on the gold interaction history, which is scored against a reference. We investigate whether improvement under this protocol is predictive of improved autonomous workflow execution. We study pre-SFT and supervised fine-tuned (SFT) Qwen3 models at 4B and 14B parameters and Gemma 3 models at 4B and 12B parameters on multi-turn customer-support workflows. We find that SFT consistently improves text-turn success, and that overall next-turn success increases for every model under gold-history evaluation. However, these improvements do not transfer to autonomous workflow execution. Tool-specific gains also vary across metrics and models. None of the four SFT models succeeds under holistic workflow evaluation, with strict trajectory completion reaching at most 10.4% workflow success. Our results show that next-turn evaluation is not a reliable proxy for workflow success, motivating separate reporting of text quality, local action correctness, tool execution, and end-to-end task completion.
Chinese Translation
代理模型通常是逐步评估的，其中模型基于黄金交互历史预测下一步行动，并与参考标准进行评分。我们研究在这种协议下的改进是否能够预测自主工作流执行的改善。我们研究了在多轮客户支持工作流中，参数为4B和14B的预训练SFT（监督微调）Qwen3模型以及参数为4B和12B的Gemma 3模型。我们发现，SFT始终提高文本转化成功率，并且在黄金历史评估下，每个模型的整体下一步成功率均有所增加。然而，这些改进并未转移到自主工作流执行中。特定工具的收益在不同指标和模型之间也有所不同。在整体工作流评估中，四个SFT模型均未成功，严格的轨迹完成率最高仅达到10.4%的工作流成功率。我们的结果表明，下一步评估并不是工作流成功的可靠代理，这促使我们单独报告文本质量、局部行动正确性、工具执行和端到端任务完成情况。
View on arXiv Download PDF AI Translation
cs.CL / 33 / 2609.21227

Hallucination-R1: Robustness-Oriented Paraphrase Generation for Factual Consistency

Hallucination-R1：面向鲁棒性的释义生成以确保事实一致性
Yu, Wenhan, Wu, Wenxin, Wang, Hao, Sha, Lei
Abstract
Factual hallucination is commonly defined by incorrect factual outputs. We study a paraphrase-induced hallucination setting, where a model answers a factual question correctly in its original form but generates an incorrect answer under a semantically equivalent paraphrase. Such inconsistencies expose latent factual instability under semantic invariance. However, general-purpose paraphrases are often insufficient as robustness-oriented supervision: near-copy paraphrases provide weak signals, while overly diverse paraphrases may break semantic equivalence. In this paper, we propose HALLUCINATION-R1, a robustness-oriented paraphrase generation framework that learns to produce semantically faithful yet robustness-challenging paraphrases for factual consistency. Through two-stage optimization, it first stabilizes meaning-preserving and diverse paraphrasing, then rewards paraphrases that reveal factual consistency degradation in downstream QA models. Experiments on SimpleQuestions, PopQA, and TruthfulQA show that HALLUCINATION-R1 achieves a strong consistency--diversity trade-off and exposes robustness failures across multiple model families and datasets. Further analyses indicate that these failures are not reducible to surface-level artifacts or semantic drift, but reveal non-trivial factual instability under meaning-preserving variation. A lightweight fine-tuning study also shows that HALLUCINATION-R1-generated data improves robust accuracy under paraphrase variations, suggesting its utility for robustness-oriented training. Our code and models are publicly available at https://github.com/yuwenhan07/Hallucination-R1.
Chinese Translation
事实幻觉通常被定义为不正确的事实输出。我们研究了一种由释义引发的幻觉设置，在这种情况下，模型在原始形式下正确回答事实问题，但在语义等价的释义下生成了不正确的答案。这种不一致性暴露了在语义不变性下潜在的事实不稳定性。然而，通用释义往往不足以作为面向鲁棒性的监督：近似复制的释义提供了微弱的信号，而过于多样化的释义可能破坏语义等价性。在本文中，我们提出了HALLUCINATION-R1，一种面向鲁棒性的释义生成框架，旨在学习生成语义上忠实但具有鲁棒性挑战的释义，以确保事实一致性。通过两阶段优化，它首先稳定保持意义和多样性的释义，然后奖励那些在下游问答模型中揭示事实一致性下降的释义。在SimpleQuestions、PopQA和TruthfulQA上的实验表明，HALLUCINATION-R1实现了强一致性与多样性的权衡，并揭示了多个模型系列和数据集中的鲁棒性失败。进一步分析表明，这些失败并不能归结为表面层次的伪影或语义漂移，而是揭示了在保持意义变化下的非平凡事实不稳定性。轻量级微调研究还表明，HALLUCINATION-R1生成的数据在释义变化下提高了鲁棒准确性，暗示其在面向鲁棒性训练中的实用性。我们的代码和模型已公开发布在 https://github.com/yuwenhan07/Hallucination-R1。
View on arXiv Download PDF AI Translation
cs.CL / 34 / 2609.21231

Beyond Reference-Based Evaluation: Reward Models for Meta-Evaluation of Grammatical Error Correction

超越基于参考的评估：用于语法错误纠正元评估的奖励模型
Wu, Ruotian, Johnson, Bill E., Saunders, Gene, Hamzeh, Osama, Vadehra, Ankit, Poupart, Pascal
Abstract
Reference-based metrics for Grammatical Error Correction (GEC) such as M$^2$ and ERRANT assume that the reference set enumerates all valid edits, and therefore often penalize corrections that are grammatical and meaning-preserving but phrased differently. We introduce RM-EVAL, a reward model trained on human preference data from SEEDA, as a reference-free meta-evaluator that predicts human-like quality judgments at both full-sequence and partial-sequence levels. Beyond evaluation, we show that the same reward model can be used as a learning signal to improve GEC generation via Reward-Guided Text Generation (RGTG), which keeps a base GEC model frozen and performs online, reward-driven decoding. Across SEEDA, RM-EVAL achieves strong agreement with human rankings, and RGTG yields consistent gains in reward and external validation, demonstrating a unified framework for both assessing and enhancing GEC systems without relying on gold references.
Chinese Translation
基于参考的语法错误纠正（GEC）指标，如 M$^2$ 和 ERRANT，假设参考集列举了所有有效的编辑，因此常常对那些语法正确且保持意义但措辞不同的纠正进行惩罚。我们引入了 RM-EVAL，这是一种基于人类偏好数据（来自 SEEDA）训练的奖励模型，作为一种无参考的元评估器，能够在全序列和部分序列层面预测类人质量判断。除了评估之外，我们还展示了相同的奖励模型可以作为学习信号，通过奖励引导文本生成（Reward-Guided Text Generation, RGTG）来改善 GEC 生成，该方法保持基础 GEC 模型不变，并进行在线的、基于奖励的解码。在 SEEDA 数据集上，RM-EVAL 与人类排名达成了强一致性，而 RGTG 在奖励和外部验证上均取得了一致的提升，展示了一个统一的框架，用于评估和增强 GEC 系统，而无需依赖黄金参考。
View on arXiv Download PDF AI Translation
cs.CL / 35 / 2609.21247

When Does Reasoning Help in Machine Translation? A Hierarchical Analysis of LRM Reasoning Traces

推理何时有助于机器翻译？LRM推理痕迹的层次分析
Liu, Yuxiang, Luo, Jiaming, Briakou, Eleftheria, Cherry, Colin
Abstract
Large Reasoning Models increasingly use intermediate traces for machine translation, but it remains unclear when such reasoning helps or hurts. We analyze reasoning traces across models, languages, domains, and datasets, focusing on reasoning language, length, and structure. We find that the best reasoning language is model-specific, reasoning length has a non-monotonic relationship with quality, and traces exhibit recurring functional patterns. To uncover these patterns, we introduce Hierarchical Meta-Summarization (HMS), a scalable framework that induces coarse- and fine-grained reasoning structures without predefined taxonomies. HMS reveals a shared organization--understanding/planning, translating/drafting, and refining/verifying--alongside domain-specific variation. Our results suggest that MT reasoning should be controlled in a model-aware, length-aware, and pattern-aware manner rather than uniformly encouraged.
Chinese Translation
大型推理模型越来越多地使用中间痕迹进行机器翻译，但何时这种推理有助于或有害仍不清楚。我们分析了跨模型、语言、领域和数据集的推理痕迹，重点关注推理语言、长度和结构。我们发现最佳推理语言是特定于模型的，推理长度与质量之间存在非单调关系，且痕迹表现出重复的功能模式。为了揭示这些模式，我们引入了层次元摘要（Hierarchical Meta-Summarization, HMS），这是一个可扩展的框架，可以在没有预定义分类法的情况下引导粗略和细致的推理结构。HMS揭示了一种共享的组织结构——理解/规划、翻译/草拟和精炼/验证——以及领域特定的变异。我们的结果表明，机器翻译推理应以模型感知、长度感知和模式感知的方式进行控制，而不是统一鼓励。
View on arXiv Download PDF AI Translation
cs.CL / 36 / 2609.21277

How Many Humans Is a Judge Panel Worth?

一个评审小组相当于多少人类判断？
Li, Chao, Yu, Yingying, Li, Yunfeng
Abstract
How many human judgments does a panel of language models represent? The answer depends on what is matched. We audit categorical judge panels against empirical human label distributions, retaining disagreement that binary errors relative to one gold label collapse. We measure spectral residual diversity by matching the participation ratio of a normalized residual Gram matrix to conditionally independent human-reference draws, giving nu_H. We separately match distributional squared error, giving nu_MSE. Across three ChaosNLI tasks, the same 32-judge panels have nu_H=4.24--6.50 but nu_MSE=2.30--3.75. A spectral identity separates the eigenvalues, member energies, and averaging-direction weights that determine error. Realizable hard-label panels show that greater spectral diversity can accompany worse distribution recovery even with equal member energies and nonnegative correlations. In the observed panels, within-size ranking agreement varies sharply by task; some member additions produce conflicting changes that persist across two item halves. The consensus-direction share of centered residual variance is gamma_co=43.8% on MNLI-m and 33.7% on SNLI, quantifying shared variation retained by averaging. We provide aligned votes and analysis protocols for auditing these distinctions. Effective size is therefore a target-specific measurement: spectral diversity and distribution recovery should not be treated as interchangeable measures of panel quality or as general human-replacement rates.
Chinese Translation
一个语言模型评审小组代表多少人类判断？答案取决于匹配的内容。我们对照实证人类标签分布审核分类评审小组，保留相对于一个金标准标签的二元错误所产生的分歧。我们通过将归一化残差Gram矩阵的参与比例与条件独立的人类参考抽样相匹配来测量谱残差多样性，得到nu_H。我们单独匹配分布平方误差，得到nu_MSE。在三个ChaosNLI任务中，相同的32名评审小组的nu_H为4.24至6.50，但nu_MSE为2.30至3.75。一个谱特征将特征值、成员能量和决定误差的平均方向权重分开。可实现的硬标签小组表明，即使在相等的成员能量和非负相关的情况下，更大的谱多样性也可能伴随更糟糕的分布恢复。在观察到的小组中，按大小排名的一致性因任务而异；某些成员的增加会产生持续的相互矛盾的变化，跨越两个项目的半部分。中心残差方差的共识方向份额在MNLI-m上为gamma_co=43.8%，在SNLI上为33.7%，量化了通过平均保留的共享变异。我们提供了对这些区别进行审核的对齐投票和分析协议。因此，有效大小是一个特定目标的测量：谱多样性和分布恢复不应被视为小组质量的可互换度量或一般人类替代率。
View on arXiv Download PDF AI Translation
cs.CL / 37 / 2609.21349

From Memory to Behavior: A Behavior-Aware Role-Playing Framework for Social Media Influencers

从记忆到行为：一个面向社交媒体影响者的行为感知角色扮演框架
Peng, Ji-Lun, Zhang, Yi-Zhen, Chou, Chun-Nan, Chen, Yun-Nung
Abstract
Large language models have shown strong potential as role-playing agents for real individuals, yet faithful impersonating remains challenging. Existing in-context learning-based methods fail to capture how individuals react under different situations. In addition, LLM-based evaluation is difficult for obscure individuals. To address these challenges, we propose Situation--Internal state--Behavior Persona method to incorporate situation-dependent behavioral strategies. We further design an evaluation protocol that provides LLM evaluators with references about the impersonated individual. We evaluate our approach on a newly constructed dataset for the task of generating replies on social media. Experimental results show that our proposed method outperforms state-of-the-art ICL-based baselines, while our evaluation protocol achieves moderate correlation with human judgment. Besides, experiments on fictional-character benchmarks demonstrate that our proposed method is applicable beyond the social media setting. These findings suggest that incorporating behavioral information broadly improves the fidelity of role-playing for real individuals on social media or fictional characters.
Chinese Translation
大型语言模型在作为真实个体的角色扮演代理方面展现出强大的潜力，但忠实模仿仍然具有挑战性。现有的基于上下文学习的方法未能捕捉个体在不同情境下的反应。此外，基于大型语言模型（LLM）的评估对于不知名个体而言也很困难。为了解决这些挑战，我们提出了情境-内在状态-行为角色（Situation--Internal state--Behavior Persona）方法，以纳入情境依赖的行为策略。我们进一步设计了一种评估协议，为LLM评估者提供关于被模仿个体的参考信息。我们在一个新构建的数据集上评估了我们的方法，该数据集用于生成社交媒体上的回复。实验结果表明，我们提出的方法在性能上优于最先进的基于上下文学习的基线，同时我们的评估协议与人类判断之间达到了中等相关性。此外，在虚构角色基准上的实验表明，我们的方法适用于社交媒体以外的场景。这些发现表明，广泛纳入行为信息能够显著提高对真实个体或虚构角色在社交媒体上的角色扮演的忠实度。
View on arXiv Download PDF AI Translation
cs.CL / 38 / 2609.21362

Beyond Atomic Tokens: Factorizing Syllables for Language Model Pretraining

超越原子标记：为语言模型预训练因式分解音节
Nguyen, Nghia Hieu, Huynh, Thai Bao, Dinh-Le, Binh-An, Hoang, Phu Gia, Nguyen, Dat Tien, Van Nguyen, Kiet, Nguyen, Ngan Luu-Thuy
Abstract
Conventional tokenizers represent text as characters or statistically derived subwords, overlooking the internal phonological structure of syllables and often requiring large vocabularies. We introduce \textbf{Phonemic Tokenizer}, a linguistically motivated tokenizer for Vietnamese and Chinese that converts each syllable into IPA and factorizes it into three phonological components: onset, rime, and tone. The three components jointly occupy one contextual position, preserving syllable-level sequence length while enabling representation sharing across phonologically related syllables. Non-phonological and unsupported units are handled through character-level fallback. This deterministic design requires no corpus-dependent vocabulary learning and yields vocabularies of only 112 entries for Chinese and 256 for Vietnamese. Intrinsic evaluation shows that the tokenizer achieves substantially higher R\'enyi efficiency in both languages, represents every entry in a standard Vietnamese syllable dictionary with a Fertility of exactly one, and generally produces shorter Vietnamese sequences than existing pretrained tokenizers. We further instantiate the tokenizer in \textbf{PhonemicBERT}, which combines factorized component embeddings and reconstructs complete masked syllables using three prediction heads. Under a controlled Chinese pretraining setup, PhonemicBERT-Zh is competitive with or outperforms character, subword, and SubChar alternatives across diverse language-understanding tasks. PhonemicBERT-Vi also achieves competitive or superior results to established Vietnamese and multilingual pretrained models. These results establish phonemic factorization as a compact, efficient, and interpretable alternative to atomic and statistically segmented text representations.
Chinese Translation
传统的分词器将文本表示为字符或统计派生的子词，忽视了音节的内部音韵结构，并且通常需要较大的词汇量。我们提出了 extbf{音位分词器}（Phonemic Tokenizer），这是一种针对越南语和汉语的语言学驱动的分词器，它将每个音节转换为国际音标（IPA），并将其因式分解为三个音韵成分：声母、韵母和声调。这三个成分共同占据一个上下文位置，保持音节级别的序列长度，同时使音韵相关的音节之间能够共享表示。非音韵和不支持的单位通过字符级回退进行处理。这种确定性设计不需要依赖语料库的词汇学习，汉语的词汇量仅为112个条目，越南语为256个条目。内在评估表明，该分词器在两种语言中都实现了显著更高的R'enyi效率，能够以确切的生育率表示标准越南语音节词典中的每个条目，并且通常生成的越南语序列比现有的预训练分词器更短。我们进一步在 extbf{音位BERT}（PhonemicBERT）中实例化该分词器，该模型结合了因式分解的成分嵌入，并使用三个预测头重构完整的掩蔽音节。在一个受控的汉语预训练设置下，PhonemicBERT-Zh在各种语言理解任务中与字符、子词和SubChar替代方案竞争或表现优越。PhonemicBERT-Vi也在与已建立的越南语和多语言预训练模型的比较中取得了竞争或更优的结果。这些结果确立了音位因式分解作为一种紧凑、高效且可解释的替代方案，优于原子和统计分割的文本表示。
View on arXiv Download PDF AI Translation
cs.CL / 39 / 2609.21378

ArenaFlow: From Trajectory Ranking to Hierarchical Credit Propagation for Open-Ended Agent RL

ArenaFlow：从轨迹排名到开放式智能体强化学习的层次信用传播
Zhang, Qiang, Ding, Ruixue, Zhang, Fanrui, Chen, Xi, Chen, Boli, Wang, Shihang, Huang, Yinfeng, Zheng, Yi, Xie, Pengjun, Zhang, Kaipeng, Liu, Jiawei, Zha, Zheng-Jun
Abstract
Reinforcement learning has substantially improved large language model (LLM) agents in verifiable domains, but remains difficult to apply to open-ended agent tasks, where solutions are diverse and reliable scalar rewards are hard to obtain. Recent pairwise evaluation methods alleviate reward discrimination collapse by replacing pointwise scoring with relative preferences. However, they still compress rich comparative feedback into a single trajectory-level reward, obscuring decisive intermediate steps and preventing successful behaviors from being consolidated into reusable skills. We propose ArenaFlow, a hierarchical credit propagation framework for open-ended agent reinforcement learning. ArenaFlow leverages tournament-based relative ranking to derive trajectory-level reward signals. Each comparison is further equipped with structured reflective evaluation, which reveals three types of supervision: pivotal success steps, reusable strategy skills, and usage attribution of retrieved skills. At the step level, ArenaFlow propagates trajectory-level advantages to high-confidence pivotal steps according to tournament survival depth, enabling more targeted optimization of local reasoning behaviors. At the skill level, ArenaFlow estimates skill utility from group-level usage attribution and maintains a global skill memory through utility-aware updating, pruning, and retrieval. The resulting high-utility skills further serve as policy priors for future exploration. Extensive experiments validate ArenaFlow's effectiveness on open-ended agent tasks.
Chinese Translation
强化学习在可验证领域显著提升了大型语言模型（LLM）智能体的表现，但在开放式智能体任务中应用仍然困难，因为这些任务的解决方案多样且可靠的标量奖励难以获得。最近的成对评估方法通过用相对偏好替代逐点评分，缓解了奖励判别崩溃的问题。然而，这些方法仍然将丰富的比较反馈压缩为单一的轨迹级奖励，模糊了关键的中间步骤，并阻碍了成功行为的巩固，使其无法转化为可重用的技能。我们提出了ArenaFlow，一个用于开放式智能体强化学习的层次信用传播框架。ArenaFlow利用基于比赛的相对排名来推导轨迹级奖励信号。每次比较进一步配备了结构化的反思评估，揭示了三种类型的监督：关键成功步骤、可重用策略技能和检索技能的使用归因。在步骤层面，ArenaFlow根据比赛生存深度将轨迹级优势传播到高置信度的关键步骤，从而实现对局部推理行为的更有针对性的优化。在技能层面，ArenaFlow通过群体使用归因估计技能效用，并通过效用感知的更新、修剪和检索来维护全球技能记忆。最终产生的高效用技能进一步作为未来探索的策略先验。大量实验验证了ArenaFlow在开放式智能体任务上的有效性。
View on arXiv Download PDF AI Translation
cs.CL / 40 / 2609.21383

Prediction Dynamics in Depth-Recurrent Language Models

深度递归语言模型中的预测动态
Luo, Xinyue, Yu, Fei
Abstract
Depth-recurrent language models refine predictions through repeated latent updates. Why can intermediate answers agree with the endpoint while their scores continue to change? We derive a sharp margin characterization that decomposes the conservatism of a magnitude bound into common translation, direction relative to the winner, and the pairing of each competitor's update with its score gap. Across Huginn-3.5B and Ouro-1.4B, accounting for update direction and competitor pairing reduces the mean earliest qualifying depth by a further 22.5-34.4% of the total depth beyond translation removal under full answer-text scoring. This retrospective comparison uses completed trajectories. Substantial contributions also occur under label scoring. For shared predictive distributions, we separate common and contrast motion orthogonally and express the common component through candidate-set mass and within-set concentration. Common and contrast energies can attenuate at different rates, allowing a growing preference-change share to coexist with shrinking absolute updates. These findings explain finite-depth answer preservation through the geometry and composition of observed score changes.
Chinese Translation
深度递归语言模型通过重复的潜在更新来精炼预测。为什么中间答案可以与最终结果一致，而它们的得分却持续变化？我们推导出一个明确的边际特征，将幅度界限的保守性分解为共同翻译、相对于赢家的方向以及每个竞争者的更新与其得分差的配对。在 Huginn-3.5B 和 Ouro-1.4B 上，考虑更新方向和竞争者配对后，平均最早合格深度在完全答案文本评分下比去除翻译后的总深度进一步减少了 22.5-34.4%。该回顾性比较使用了完成的轨迹。在标签评分下也发生了显著贡献。对于共享的预测分布，我们将共同和对比运动正交分离，并通过候选集质量和集内集中度表达共同成分。共同能量和对比能量可以以不同的速率减弱，从而使得不断增长的偏好变化份额与绝对更新的缩小共存。这些发现通过观察到的得分变化的几何形状和组成解释了有限深度答案的保留。
View on arXiv Download PDF AI Translation
cs.CL / 41 / 2609.21387

Consistent Relexicalization of Clinical Documents using Graph-Based Approach

基于图的临床文档一致性重词汇化
Das, Dipankar, Mandal, Atri, Singh, Sandeep, Shandhilya, Tushar
Abstract
Relexicalization is a pivotal technique in clinical NLP, as it facilitates robust masking of sensitive information while synthesizing datasets that retain high-fidelity, real-world characteristics. However, preserving structural integrity, relational coherence, and temporal consistency during transformation remains a significant challenge. Existing approaches frequently rely on independent entity replacement, which results in clinical inconsistencies across longitudinal records. This reduces the value of such relexicalized datasets for downstream scientific analysis. To address these limitations, we introduce G-RELIC (Graph Based Contextual Relexicalization with Improved Consistency) which combines the power of LLMs with graphs. G-RELIC implements a graph-based mapping mechanism which optimizes for one-to-one correspondence between original and surrogate entities. It also introduces a deterministic temporal repositioning algorithm to preserve temporal consistency. Empirical evaluations on diverse, real-world clinical datasets validate that G-RELIC significantly outperforms state-of-the-art baselines. G-RELIC yields a 30.4 percentage point improvement in relational integrity (62.1% to 92.5%) and 45.9 percentage point improvement in temporal coherence (46% to 91.9%) without compromising on the recognized privacy benchmarks for clinical datasets. This maximizes the analytical utility of relexicalized datasets while minimizing re-identification risk.
Chinese Translation
重词汇化是临床自然语言处理中的一项关键技术，它有助于在合成保留高保真度、真实世界特征的数据集的同时，有效地屏蔽敏感信息。然而，在转换过程中保持结构完整性、关系一致性和时间一致性仍然是一个重大挑战。现有的方法通常依赖于独立的实体替换，这导致纵向记录中的临床不一致性，从而降低了这些重词汇化数据集在下游科学分析中的价值。为了解决这些局限性，我们提出了G-RELIC（基于图的上下文重词汇化与改进一致性），它结合了大型语言模型（LLMs）与图的优势。G-RELIC实现了一种基于图的映射机制，优化了原始实体与替代实体之间的一对一对应关系。它还引入了一种确定性的时间重定位算法，以保持时间一致性。在多样化的真实临床数据集上的实证评估验证了G-RELIC显著优于最先进的基准方法。G-RELIC在关系完整性上提高了30.4个百分点（从62.1%提升至92.5%），在时间一致性上提高了45.9个百分点（从46%提升至91.9%），且未妥协于临床数据集的隐私基准。这最大化了重词汇化数据集的分析效用，同时最小化了重新识别的风险。
View on arXiv Download PDF AI Translation
cs.CL / 42 / 2609.21392

Omni Demand Understanding: A Benchmark for Contextual User-Intent Inference in Multimodal Interaction

全方位需求理解：多模态交互中上下文用户意图推断的基准
Chen, Qi, Chu, Yunfei, He, Haolin, Yang, Yifan, Liu, Zihan, Wang, Yuxuan, Ma, Ziyang, Xu, Ruiyang, Gao, Meng, Yan, Yinsong, Wang, Ling, Wang, Hui, Huang, Wen, Chen, Yiheng, Yang, Guanrou, Kong, Qiuqiang, Xu, Jin, Chen, Xie
Abstract
Natural audio-visual interaction is emerging as an important interface for AI assistants, allowing users to communicate through speech and vision rather than carefully composed text prompts. However, existing benchmarks of interactive capabilities still focus primarily on response quality, leaving a more fundamental question underexplored: can a model correctly infer the user's underlying demand from complex multimodal interaction? Real-world user demands are often underspecified in speech and must be inferred from multimodal cues and dialogue history. This inference is further complicated by ambiguous or disfluent expression and noisy acoustic environments. Conversely, request-like speech may not constitute a demand to the assistant, leading to false triggers. We establish Omni Demand Understanding (ODU) as a distinct multimodal contextual inference problem: given an interaction stream, a model must detect whether a user demand is present and infer intent from multimodal and conversational context. ODU evaluates this capability along five dimensions, covering both single-turn and multi-turn interactions. We construct ODU-Bench using a challenge-driven taxonomy, taxonomy-guided agentic video generation, and human-recorded interactions, followed by media-grounded annotation and human verification. We evaluate 14 native MLLMs. Even the strongest, Gemini 3.1 Pro, recovers only 44.7% of key information that must be inferred from visual, acoustic, or conversational context. Moreover, 11 of the 14 models exhibit false-trigger rates above 50% on non-demand scenarios. These results reveal a systematic capability gap in current MLLMs' ability to infer contextual user demands. We hope ODU can establish the evaluation of a previously underexplored yet essential capability in multimodal interaction: correctly understanding user demands before generating an appropriate response.
Chinese Translation
自然的音频-视觉交互正在成为人工智能助手的重要接口，使用户能够通过语音和视觉进行交流，而不是依赖精心编写的文本提示。然而，现有的交互能力基准仍主要集中在响应质量上，未能深入探讨一个更为根本的问题：模型能否从复杂的多模态交互中正确推断用户的潜在需求？现实世界中的用户需求在语音中往往表述不明确，必须通过多模态线索和对话历史进行推断。这一推断因模糊或不流畅的表达以及嘈杂的声学环境而变得更加复杂。相反，类似请求的语音可能并不构成对助手的需求，从而导致错误触发。我们将全方位需求理解（Omni Demand Understanding, ODU）确立为一个独特的多模态上下文推断问题：给定一个交互流，模型必须检测用户需求是否存在，并从多模态和对话上下文中推断意图。ODU在五个维度上评估这一能力，涵盖单轮和多轮交互。我们使用以挑战为驱动的分类法、分类法引导的代理视频生成和人类记录的交互构建了ODU-Bench，随后进行媒体基础的注释和人工验证。我们评估了14个本土多模态大语言模型（MLLMs）。即使是最强的模型，Gemini 3.1 Pro，也仅能从视觉、声学或对话上下文中恢复44.7%的关键信息。此外，14个模型中有11个在非需求场景中表现出超过50%的错误触发率。这些结果揭示了当前多模态大语言模型在推断上下文用户需求能力上的系统性差距。我们希望ODU能够建立对多模态交互中一个先前未被充分探索但至关重要的能力的评估：在生成适当响应之前，正确理解用户需求。
View on arXiv Download PDF AI Translation
cs.CL / 43 / 2609.21401

Talking Past the Machine: Morality, Politeness, and Alignment in Human-AI Dialogue

超越机器对话：人机对话中的道德、礼貌与一致性
Mitiaeva, Marina, Xiao, Lu
Abstract
Conversational AI systems produce fluent, socially appropriate responses, yet whether they participate in cooperative communication or merely simulate its surface forms remains unclear - a question central to how these systems are evaluated, trusted, and designed. This study investigates how morality, politeness, and alignment - three dimensions central to cooperative dialogue - function in human-AI interaction compared to human-human conversation. We analyze 15,881 human-ChatGPT and 10,784 human-human multi-turn dialogues, using mixed-effects models to identify which features predict turn-to-turn alignment. We observe a consistent dissociation: AI produces the surface features of cooperative communication without the underlying social architecture. Moral output appears preconfigured rather than negotiated; warmth is generated without face sensitivity; linguistic convergence declines persistently. Most strikingly, the cooperative mechanisms themselves reverse direction: hedging and softening associated with greater accommodation between humans are associated with reduced alignment when produced by AI, and purity framing associated with human divergence coincides with users converging toward the AI. Agency - giving users room to shape the exchange - is the most consistent predictor of alignment across both interaction types, while lower moral assertiveness in more recent models is not accompanied by better cooperation. Together these patterns suggest that AI reproduces the surface of cooperation without the mutual adaptation that grounds it between humans - and, more surprisingly, that mechanisms sustaining human accommodation can run in reverse with AI, suggesting a turn-level view may be insufficient for interaction-level success.
Chinese Translation
对话式人工智能系统能够生成流畅且社会适宜的回应，但它们是否参与合作性交流，或仅仅模拟其表面形式仍然不清楚——这一问题对于如何评估、信任和设计这些系统至关重要。本研究探讨了道德、礼貌和一致性——这三者是合作性对话的核心维度——在与人工智能的互动中如何运作，并与人际对话进行比较。我们分析了15,881个与ChatGPT的对话和10,784个人与人之间的多轮对话，使用混合效应模型识别哪些特征可以预测轮对轮的一致性。我们观察到一种持续的解离现象：人工智能生成了合作性交流的表面特征，却缺乏其底层的社会架构。道德输出似乎是预设的而非协商的；温暖感的产生缺乏面子敏感性；语言趋同持续下降。最引人注目的是，合作机制本身的方向发生了逆转：与人类之间更大适应性相关的模糊和缓和，在人工智能生成时却与降低的一致性相关，而与人类分歧相关的纯粹框架则与用户向人工智能趋同相吻合。赋权——给予用户塑造交流的空间——是两种互动类型中一致性最一致的预测因子，而较低的道德自信在较新模型中并未伴随更好的合作。综合这些模式表明，人工智能再现了合作的表面，而缺乏人类之间支撑其的相互适应——更令人惊讶的是，维持人类适应的机制在与人工智能的互动中可能会反向运行，这表明仅从轮次层面来看可能不足以实现互动层面的成功。
View on arXiv Download PDF AI Translation
cs.CL / 44 / 2609.21490

Benchmarking Gender Bias in Machine Translation Evaluation Metrics across Occupations

跨职业机器翻译评估指标中的性别偏见基准测试
Mastromichalakis, Orfeas Menis, Filandrianos, Giorgos, Mohammed, Wafaa, Attanasio, Giuseppe, Zerva, Chrysoula
Abstract
Gender bias remains a persistent concern in machine translation (MT), affecting both generated translations and their automatic evaluation. When a source text leaves a person's gender unspecified, translations may realize that person using masculine or feminine forms, and both MT systems and evaluation metrics may exhibit systematic preferences between these alternatives despite the source providing no basis for such a distinction. We study this behavior in the WMT 2026 Automated Translation Quality Evaluation Systems Shared Task using an occupation-balanced subset of GAMBIT+. We consider seven English-source language pairs, six from the original dataset, targeting Arabic, Czech, Greek, Icelandic, Russian, and Ukrainian, and extend the original resource with German. The subset contains 1,308 masculine/feminine translation pairs per target language, with three examples for each of the 436 ISCO-08 occupational groups. We evaluate shared-task submissions and baselines for score prediction and error annotation, examining the direction, magnitude, and frequency of gender-related differences. We find an overall tendency for masculine translations to receive higher scores, as well as differences per occupation following stereotypical gender representations, although the strength and consistency of this preference vary considerably across evaluators and languages. Our results show that gender bias remains present in MT evaluation, but that capturing its extent requires looking beyond a single aggregate measure to complementary dimensions of evaluator behavior.
Chinese Translation
性别偏见在机器翻译（MT）中仍然是一个持续关注的问题，影响着生成的翻译及其自动评估。当源文本未指定某人的性别时，翻译可能会使用男性或女性形式来表达该人，而尽管源文本未提供此类区分的依据，MT系统和评估指标可能在这些选择之间表现出系统性的偏好。我们在WMT 2026自动翻译质量评估系统共享任务中研究了这种行为，使用了GAMBIT+的一个职业平衡子集。我们考虑了七个英语源语言对，其中六个来自原始数据集，目标语言为阿拉伯语、捷克语、希腊语、冰岛语、俄语和乌克兰语，并用德语扩展了原始资源。该子集包含每种目标语言1,308对男性/女性翻译，每个436个ISCO-08职业组提供三个示例。我们评估了共享任务的提交和基线在得分预测和错误标注方面，考察了性别相关差异的方向、幅度和频率。我们发现男性翻译整体上倾向于获得更高的分数，并且每个职业的差异遵循刻板的性别表现，尽管这种偏好的强度和一致性在评估者和语言之间差异显著。我们的结果表明，性别偏见在MT评估中仍然存在，但要捕捉其程度需要超越单一的综合指标，关注评估者行为的互补维度。
View on arXiv Download PDF AI Translation
cs.CL / 45 / 2609.21554

MIRAGE: Multi-Perspective Creative Language Model Reasoning with Reinforcement Learning Guidance

MIRAGE：基于强化学习指导的多视角创意语言模型推理
Lagzian, Arash, Anumasa, Srinivas, Liu, Dianbo
Abstract
Recent advances in Large Language Models (LLMs) have revolutionized artificial intelligence and how human interact with AIs. Despite impressive advancements, LLMs struggle with complex mathematical, scientific, and logical tasks. Inspired by human cognitive flexibility - our ability to dynamically switch mental perspectives - we propose MIRAGE (Multi-perspective Inference-time Reasoning via Agent-Guided Exploration), a novel inference-time creative thinking framework. MIRAGE includes a Selector that prioritizes effective conceptual perspectives (e.g., algebraic, probabilistic) and a Reasoner that sequentially solves tasks until a confident solution emerges, otherwise aggregating multiple perspectives. Tested on GSM8K, MATH500, MMLU-Pro, and Game-of-24 benchmarks, MIRAGE consistently outperforms methods like Chain-of-Thought and diverse prompting ensembles, significantly boosting accuracy with minimal inference overhead, providing a scalable solution for practical applications.
Chinese Translation
近期大型语言模型（LLMs）的进展彻底改变了人工智能及人类与人工智能的互动方式。尽管取得了显著的进步，LLMs在复杂的数学、科学和逻辑任务上仍然面临挑战。受到人类认知灵活性的启发——我们动态切换思维视角的能力——我们提出了MIRAGE（基于代理引导探索的多视角推理），这是一种新颖的推理时创意思维框架。MIRAGE包括一个选择器，优先考虑有效的概念视角（例如，代数视角、概率视角），以及一个推理器，顺序解决任务，直到出现一个令人信服的解决方案，否则聚合多个视角。在GSM8K、MATH500、MMLU-Pro和Game-of-24基准测试中，MIRAGE始终优于Chain-of-Thought和多样化提示集等方法，显著提高了准确性，同时保持了最小的推理开销，为实际应用提供了可扩展的解决方案。
View on arXiv Download PDF AI Translation
cs.CL / 46 / 2609.21595

Evaluating In-Context Learning and Retrieval Strategies for Devanagari Post-OCR Correction

评估德瓦纳格里后OCR纠正的上下文学习和检索策略
Bhandari, Abhishek, Harit, Gaurav
Abstract
In-context learning using Large Language Models (LLMs) offers a compelling path to training-free post-OCR correction, yet its effectiveness for Devanagari script remains entirely unexplored. We present the first systematic evaluation of LLMs (3B-32B) for post-OCR correction in Hindi and Marathi, comparing three in-context example retrieval strategies: domain-random selection, dense semantic retrieval, and our proposed CharBM25, which retrieves examples by character n-gram BM25 similarity over OCR inputs to target shared error patterns with the test sentence. Across a 20,000-sentence benchmark spanning five news domains, retrieval strategy is the decisive factor in correction quality: CharBM25 outperforms domain-random selection by 2.8-4.0pp absolute WER on Hindi and 2.9-3.8pp on Marathi, using character trigrams, which consistently outperform bigrams and unigrams. Scale dominates performance: Gemma-3-27B achieves WER reductions of 55.0% for Hindi and 33.3% for Marathi under CharBM25-5. Few-shot gains are capacity-gated: models below 8B do not reliably improve over the OCR baseline, and on Marathi the smallest models (3B) degrade more sentences than they improve. Marathi is persistently harder to correct than Hindi across all scales, reflecting its greater morphological complexity. These findings establish CharBM25 as an effective, GPU-free retrieval strategy that matches or exceeds dense retrieval at negligible computational cost, and show that combining it with a general-purpose LLM of 12B+ parameters delivers reliable, training-free Devanagari post-OCR correction without task-specific fine-tuning. Dataset: https://huggingface.co/datasets/AbhishekBhandari/Devanagari-OCR-ICL-Benchmark
Chinese Translation
使用大型语言模型（LLMs）的上下文学习为无训练的后OCR纠正提供了一条引人注目的路径，但其在德瓦纳格里脚本中的有效性尚未被探索。我们首次系统性地评估了LLMs（3B-32B）在印地语和马拉地语后OCR纠正中的表现，比较了三种上下文示例检索策略：领域随机选择、密集语义检索，以及我们提出的CharBM25，该策略通过字符n-gram BM25相似性从OCR输入中检索示例，以针对与测试句子共享的错误模式。在涵盖五个新闻领域的20,000句基准测试中，检索策略是纠正质量的决定性因素：在印地语中，CharBM25的绝对字错误率（WER）比领域随机选择提高了2.8-4.0个百分点，在马拉地语中提高了2.9-3.8个百分点，使用的字符三元组始终优于二元组和单元组。规模主导性能：Gemma-3-27B在CharBM25-5下实现了印地语55.0%和马拉地语33.3%的WER降低。少量示例的增益受到容量限制：8B以下的模型未能可靠地改善OCR基线，而在马拉地语中，最小的模型（3B）比改善的句子更多地降低了句子。马拉地语在所有规模中持续比印地语更难纠正，反映出其更大的形态复杂性。这些发现确立了CharBM25作为一种有效的、无GPU的检索策略，其性能与密集检索相当或更优，且计算成本微乎其微，并且表明将其与12B以上参数的一般用途LLM结合，可以在不进行特定任务微调的情况下实现可靠的、无训练的德瓦纳格里后OCR纠正。数据集：https://huggingface.co/datasets/AbhishekBhandari/Devanagari-OCR-ICL-Benchmark
View on arXiv Download PDF AI Translation
cs.CL / 47 / 2609.21636

Steering LLMs Responses Towards Moral Foundations on the Norwegian MFQ-30

引导大型语言模型的回应朝向道德基础：基于挪威MFQ-30的研究
Andersen, Hans, Dichas, David
Abstract
Recent work applies human psychometric questionnaires to large language models to elicit moral and value profiles, but it is not clear whether these instruments measure anything stable in models or whether the resulting profiles can be moved toward a target human population. We administer the Norwegian Moral Foundations Questionnaire (MFQ-30) to six open-weight LLMs and compare their foundation profiles to a sample of N = 1,282 Norwegian respondents. We test two steering interventions, prompt-level persona steering and activation-level ActAdd. Half the models engage with the questionnaire under our attention check. The other half default to flat or central-tendency outputs that look near-human on average without tracking item content. A neutral Nordic-respondent persona, written without any distributional information from the human sample, brings the engaging models 44-77% closer to the Norwegian mean in Mahalanobis $d^2$. One-pair ActAdd at a fixed mid-layer flattens the foundation profile rather than steering individual foundations. For at least one model the same persona that shifts the profile also induces engagement that was absent at baseline, a concrete instance of the cognitive phantoms that Peereboom et al. (2025) warn about.
Chinese Translation
近期的研究将人类心理测量问卷应用于大型语言模型，以引出道德和价值特征，但尚不清楚这些工具是否测量了模型中的稳定内容，或者所得到的特征是否能朝向目标人群移动。我们对六个开放权重的大型语言模型施加挪威道德基础问卷（MFQ-30），并将它们的基础特征与N = 1,282名挪威受访者的样本进行比较。我们测试了两种引导干预措施：提示级别的人格引导和激活级别的ActAdd。半数模型在我们的注意检查下参与问卷。另一半模型则默认生成平坦或中心趋势的输出，这些输出在平均上看起来接近人类，但未能跟踪项目内容。一个中立的北欧受访者人格，未使用任何来自人类样本的分布信息，使得参与的模型在Mahalanobis $d^2$上与挪威均值接近44-77%。在固定中层进行的一对ActAdd使得基础特征轮廓变得平坦，而不是引导个别基础特征。对于至少一个模型，能够改变特征轮廓的人格也引发了在基线时缺失的参与，这是Peereboom等人（2025）所警告的认知幻影的具体实例。
View on arXiv Download PDF AI Translation
cs.CL / 48 / 2609.21637

Chinese Competitive Debating Dataset and Benchmark

中文竞争辩论数据集及基准
Yang, Zongrui, Li, Haoyuan, Wang, Zhongsheng, Zeng, Zhirui, Han, Pengqian, Zhou, Yi, Wang, Yuting, Liu, Jiamou
Abstract
Debate adjudication requires tracking how arguments develop through interaction, yet existing datasets rarely combine fine-grained debate transcripts with professional judgments collected during real competitions under a shared rubric. We introduce a dataset and benchmark for evaluating large language models' understanding of competitive Chinese-language debate at the match, stage, and speaker levels. We organized 182 matches and recruited 120 professional judges, with each match independently adjudicated by three judges using a predefined rubric. After excluding matches with incomplete records, the dataset contains 148 matches, 2,698 stages, and 20,542 exchange units, with manually verified transcripts and segmentation. It preserves original stage scores, match votes, best-debater ballots, and adjudication rationales. We define three tasks: winner-tendency prediction, stage-score prediction, and best-debater prediction. Zero-shot evaluation of multiple large language models yields a highest winner-prediction accuracy of 66.2%, a highest Pearson correlation of 0.250 between model stage scores and mean human ratings, and a highest best-debater prediction accuracy of 56.8%. The dataset and benchmark provide a testbed for studying large language models' understanding of interactive argumentation and their agreement with professional judges.
Chinese Translation
辩论裁决需要跟踪论点在互动中的发展，然而现有数据集很少将细致的辩论记录与在真实比赛中根据共享标准收集的专业评判结合起来。我们介绍了一个用于评估大型语言模型对中文竞争辩论理解的数据集和基准，该评估涵盖比赛、阶段和发言者层面。我们组织了182场比赛，并招募了120名专业裁判，每场比赛由三名裁判根据预定义的标准独立裁决。经过排除记录不完整的比赛后，数据集包含148场比赛、2,698个阶段和20,542个交流单元，所有记录和分段均经过人工验证。数据集保留了原始阶段分数、比赛投票、最佳辩手选票和裁决理由。我们定义了三个任务：胜者倾向预测、阶段分数预测和最佳辩手预测。对多个大型语言模型的零样本评估显示，最高胜者预测准确率为66.2%，模型阶段分数与人类平均评分之间的最高皮尔逊相关系数为0.250，最佳辩手预测的最高准确率为56.8%。该数据集和基准为研究大型语言模型对互动论证的理解及其与专业裁判的一致性提供了测试平台。
View on arXiv Download PDF AI Translation
cs.CL / 49 / 2609.21655

Analysing the Linearity of Linguistic Relations in Language Model Embedding Spaces

分析语言模型嵌入空间中语言关系的线性特征
Nedumpozhimana, Vasudevan, Thekkekara, Fathima, Kelleher, John
Abstract
We propose a framework to analyse how strongly different linguistic relations are linearly encoded in language model embedding spaces. We formalise linear encoding via a constrained linear approximation over related and unrelated word pairs and apply this to an extended BATS dataset covering inflectional, derivational, lexicographic, and encyclopedic relations in GloVe, RoBERTa, and ModernBERT. Our experiments show near-perfect linear encodings for inflectional and derivational relations, but substantially higher errors for lexicographic and encyclopedic relations, especially for one-to-many and many-to-many associations. We also find that RoBERTa and ModernBERT generally encode relations more linearly than GloVe. These results indicate that our framework can reveal which relational structures are most linearly accessible in embeddings, offering a compact tool for probing and comparing relational geometry across models.
Chinese Translation
我们提出了一个框架来分析不同语言关系在语言模型嵌入空间中线性编码的强度。我们通过对相关和不相关词对进行约束线性近似来形式化线性编码，并将其应用于扩展的BATS数据集，该数据集涵盖了GloVe、RoBERTa和ModernBERT中的屈折、派生、词典和百科关系。我们的实验表明，屈折和派生关系的线性编码几乎完美，但词典和百科关系的误差显著更高，尤其是在一对多和多对多关联中。我们还发现，RoBERTa和ModernBERT通常比GloVe更线性地编码关系。这些结果表明，我们的框架可以揭示哪些关系结构在嵌入中最具线性可达性，为探测和比较模型间的关系几何提供了一个紧凑的工具。
View on arXiv Download PDF AI Translation
cs.CL / 50 / 2609.21662

When Steering Fails in Latent Reasoning: A Latent-to-Language Transition Gap

潜在推理中的引导失败：潜在到语言的转变差距
Huang, Gaoxiang, Qi, Lei
Abstract
Activation steering has become a widely used approach for controlling language models during explicit chain-of-thought (CoT) reasoning, motivating its extension to latent CoT. However, we find that steering continuous thoughts produces substantially weaker effects on subsequent language generation than steering explicit CoT, even when the hidden representations are moved by comparable amounts. We first show that task information remains identifiable in continuous thoughts. Hence, we hypothesize a \textbf{latent-to-language transition gap}, in which an intervention effect in latent space fails to transfer to language generation. Two further results support this hypothesis: the output distribution changes abruptly at the transition boundary, and task-related directions exert much weaker bidirectional control in latent CoT than in explicit CoT. These findings identify the transition interface as a central target for evaluating and designing future latent-steering methods.
Chinese Translation
激活引导已成为在显式思维链（CoT）推理过程中控制语言模型的广泛使用的方法，这促使其扩展到潜在的思维链。然而，我们发现，引导连续思维对后续语言生成的影响明显弱于引导显式思维链，即使在隐藏表示被移动的幅度相当的情况下。我们首先表明，任务信息在连续思维中仍然是可识别的。因此，我们假设存在一个 extbf{潜在到语言的转变差距}，在该差距中，潜在空间中的干预效果未能转移到语言生成中。进一步的两个结果支持这一假设：输出分布在转变边界处发生急剧变化，且在潜在思维链中，任务相关方向的双向控制能力远弱于在显式思维链中的能力。这些发现将转变接口确定为评估和设计未来潜在引导方法的核心目标。
View on arXiv Download PDF AI Translation
cs.CL / 51 / 2609.21663

Rethinking Human-Aligned Evaluation: An Analysis of Semantic Metrics Beyond WER

重新思考人类对齐评估：超越字错误率的语义度量分析
Sharma, Hritika, Bañeras-Roux, Thibault, Pinto, Alessandra, Motlicek, Petr, Jung, Hyunggu, Villatoro-Tello, Esaú, Nam, Somang
Abstract
Word Error Rate (WER), the most commonly used metric for Automatic Speech Recognition (ASR), treats every lexical deviation from the reference as equally costly, regardless of whether it changes meaning. This raises the question: does WER actually track how humans judge ASR transcript quality? We introduce HATS-en, an English dataset for human-centered ASR evaluation. Using this dataset, we benchmark lexical metrics against several configurations of BERTScore and SemDist, varying the language model, layer, and pooling strategy. We find that WER agrees least with human judgment among all metrics tested, that the best-performing SemDist configurations achieve the highest overall agreement, ahead of CER and BERTScore, and that no single model is best across settings. CER, despite its simplicity and low cost, remains remarkably close to these best configurations. In line with prior recommendations, our results support shifting ASR evaluation toward CER both for English and for morphosyllabic writing systems as it is a more interpretable and low-cost metric for what evaluation should actually capture, and using SemDist as a complementary evaluation.
Chinese Translation
字错误率（WER）是自动语音识别（ASR）中最常用的度量标准，它将与参考文本的每一个词汇偏差视为同等代价，而不考虑其是否改变了意义。这引发了一个问题：WER 是否真正反映了人类对 ASR 转录质量的判断？我们引入了 HATS-en，这是一个用于以人为中心的 ASR 评估的英语数据集。利用该数据集，我们对多种 BERTScore 和 SemDist 配置下的词汇度量进行了基准测试，变化包括语言模型、层和池化策略。我们发现，在所有测试的度量中，WER 与人类判断的一致性最低，而表现最佳的 SemDist 配置则达到了最高的一致性，超越了 CER 和 BERTScore，并且没有单一模型在所有设置中表现最佳。尽管 CER 简单且成本低，但它仍然与这些最佳配置保持了显著接近。与之前的建议一致，我们的结果支持将 ASR 评估转向 CER，无论是对于英语还是形态音节书写系统，因为它是一个更具可解释性和低成本的度量，能够更好地捕捉评估应实际反映的内容，同时使用 SemDist 作为补充评估。
View on arXiv Download PDF AI Translation
cs.CL / 52 / 2609.21673

PRISM-BN: A Controlled Corpus and Benchmark for Text-to-Parameterized Bayesian Network Extraction

PRISM-BN：用于文本到参数化贝叶斯网络提取的受控语料库和基准
Bhattacharya, Amartya, Singh, Nikhil, Pokhriyal, Neeti, Vosoughi, Soroush
Abstract
Probabilistic Graphical Models (PGMs), especially Bayesian Networks (BNs), expose directed structure and probabilistic parameters, making them natural symbolic targets for neurosymbolic AI. Yet training text-to-parameterized-BN systems requires paired text-to-BN resources unavailable at scale. We introduce PRISM-BN, a controlled corpus of 5054 BN-grounded descriptions paired with discrete reference BNs containing variables, states, directed edges, root priors, and full multi-parent CPDs across five domains. The instances are derived from 50 Wikipedia-seeded backbones, and their probabilities are internally constructed benchmark targets rather than externally validated causal estimates. PRISM-BN is built with PRISM, a marginal-first pipeline that elicits marginal and local joint distributions, analytically recovers normalized CPDs, and constructs locally reparameterized subgraphs. We define a benchmark with semantic node and state alignment, conditional structural scoring, and strict full-CPD evaluation. Across six LLM extractors, Node F1 ranges from 0.56 to 0.83, conditional Edge F1 from 0.90 to 0.97, and CPD-KL from 1.11 to 3.14. Conditional state and edge recovery remain consistently strong, whereas strict full-CPD agreement remains challenging. These trends persist with independently generated GPT-5.5 references, and a human pilot corroborates structural recoverability and similar probabilistic interpretations. PRISM-BN supports separate evaluation of structural recovery and probabilistic parameter estimation.
Chinese Translation
概率图模型（PGMs），尤其是贝叶斯网络（BNs），展示了有向结构和概率参数，使其成为神经符号人工智能的自然符号目标。然而，训练文本到参数化贝叶斯网络系统需要配对的文本到贝叶斯网络资源，而这些资源在规模上并不可用。我们介绍了PRISM-BN，这是一个包含5054个基于贝叶斯网络的描述的受控语料库，配对有包含变量、状态、有向边、根先验和完整多父CPDs的离散参考贝叶斯网络，涵盖五个领域。这些实例源自50个维基百科种子骨架，其概率是内部构建的基准目标，而非外部验证的因果估计。PRISM-BN是基于PRISM构建的，PRISM是一个优先考虑边际的管道，能够引出边际和局部联合分布，解析恢复归一化的CPDs，并构建局部重新参数化的子图。我们定义了一个基准，包含语义节点和状态对齐、条件结构评分以及严格的完整CPD评估。在六个大型语言模型提取器中，节点F1范围从0.56到0.83，条件边F1从0.90到0.97，CPD-KL从1.11到3.14。条件状态和边的恢复始终保持强劲，而严格的完整CPD一致性仍然具有挑战性。这些趋势在独立生成的GPT-5.5参考中依然存在，并且一项人类试点验证了结构可恢复性和相似的概率解释。PRISM-BN支持结构恢复和概率参数估计的独立评估。
View on arXiv Download PDF AI Translation
cs.CL / 53 / 2609.21722

CIBuzzBench: A Benchmark for Cross-Lingual Understanding of Chinese Internet Buzzwords

CIBuzzBench：一个用于跨语言理解中文网络流行语的基准测试
Wang, Yifan, Lu, Junyu, Wang, Qifan, Zhang, Shun, Li, Chaozhuo, Liu, Jiahao, Cao, Zhijun, Bu, Lingbin, Bu, Fanliang
Abstract
Chinese social media has generated a vast and continually evolving lexicon of internet buzzwords whose meanings are often non-literal and deeply rooted in local cultural and pragmatic contexts. Existing research has primarily focused on interpreting these buzzwords within Chinese, leaving largely unexplored whether LLMs can transfer such culturally grounded knowledge across languages and accurately convey the intended meanings in English. This cross-lingual capability is also critical for safety, as harmful expressions may obscure their offensive content through culture-specific homophony, euphemism, irony, or coded language. In this paper, we investigate the ability of advanced LLMs to understand Chinese internet buzzwords across languages. To this end, we introduce CIBuzzBench, the first benchmark for cross-lingual Chinese-to-English understanding of Chinese internet buzzwords. CIBuzzBench comprises 3,001 Chinese internet buzzwords annotated with English meaning explanations, English equivalents, category labels, and harmfulness labels. Based on these annotations, we design three evaluation tasks: Meaning Explanation, Cross-lingual Equivalent Matching, and Culturally Grounded Harmfulness Detection. We evaluate representative state-of-the-art proprietary and Chinese LLMs under both English- and Chinese-prompting settings. Our results show that LLMs continue to struggle with the cross-lingual understanding of Chinese internet buzzwords, particularly in fine-grained non-literal interpretation, robust equivalent matching under option perturbations, and calibrated harmfulness detection. These findings highlight the persistent challenges posed by culturally grounded language phenomena for multilingual LLMs and safety-oriented evaluation. The dataset and code are available at https://github.com/SuperYFan/CIBuzzBench.
Chinese Translation
中国社交媒体产生了大量不断演变的网络流行语，其含义往往是非字面性的，并深深植根于地方文化和语用背景中。现有研究主要集中在中文环境下对这些流行语的解释，尚未深入探讨大型语言模型（LLMs）是否能够跨语言转移这种文化根植的知识，并准确传达其在英语中的意图含义。这种跨语言能力对于安全性也至关重要，因为有害表达可能通过特定文化的同音词、委婉语、讽刺或隐晦语言来掩盖其攻击性内容。本文研究了先进的LLMs在跨语言理解中文网络流行语的能力。为此，我们引入了CIBuzzBench，这是第一个用于跨语言中文到英文理解中文网络流行语的基准测试。CIBuzzBench包含3,001个中文网络流行语，并附有英文含义解释、英文对应词、类别标签和有害性标签。基于这些注释，我们设计了三个评估任务：含义解释、跨语言对应词匹配和文化根植的有害性检测。我们在英文和中文提示设置下评估了代表性的最先进的专有和中文LLMs。我们的结果表明，LLMs在跨语言理解中文网络流行语方面仍然面临挑战，特别是在细粒度的非字面解释、在选项扰动下的稳健对应词匹配以及经过校准的有害性检测方面。这些发现突显了文化根植语言现象对多语言LLMs和安全导向评估所带来的持续挑战。数据集和代码可在https://github.com/SuperYFan/CIBuzzBench获取。
View on arXiv Download PDF AI Translation
cs.CL / 54 / 2609.21789

Per-Aetiology Contrastive Severity Embeddings with Phonological Pseudo-Labelling for Multilingual Dysarthric Speech

针对多语言构音障碍语音的病因对比严重性嵌入与音韵伪标注
Muller, Bernard, Barrañón, Antonio Armando Ortiz, Roberts, LaVonne
Abstract
Most multilingual dysarthria-severity systems either train on a single aetiology-language pair or pool heterogeneous aetiologies into one label space. We test that pooling assumption with four matched HuBERT-base contrastive embedding models under a shared backbone, training recipe, corpus registry and held-out evaluation: one mixed-aetiology baseline and three aetiology-specific models for cerebral palsy (CP), Parkinson's disease (PD) and amyotrophic lateral sclerosis (ALS). Training combines clinically labelled speech with ordinal pseudo-labels from a training-free phonological profiling method [1], [2]. On speaker-disjoint, leakage-filtered held-out subsets, the per-aetiology models outperform the mixed baseline across all three target aetiologies: CP (macro F1 0.829 vs 0.676, +22.6 % relative), PD (0.715 vs 0.511, +40.0 %) and ALS (0.788 vs 0.596, +32.3 %). On CP, adding 144 SAP and 44 CDSD pseudo-labelled speakers lifts macro F1 from 0.786 to 0.829 over a clinical-only CP model (+4.3 percentage points). Training data span three to seven languages per aetiology. We position this as a controlled comparison of label-space design choices and discuss pseudo-label calibration, split hygiene, and confidence-thresholded deployment as important limitations for future work.
Chinese Translation
大多数多语言构音障碍严重性系统要么在单一病因-语言对上进行训练，要么将异质病因汇聚到一个标签空间中。我们使用四个匹配的 HuBERT-base 对比嵌入模型在共享的基础架构、训练方案、语料库注册和保留评估下测试这一汇聚假设：一个混合病因基线和三个针对特定病因的模型，分别用于脑性瘫痪（CP）、帕金森病（PD）和肌萎缩侧索硬化症（ALS）。训练结合了临床标记的语音和来自无训练音韵剖析方法的序数伪标签 [1], [2]。在说话者不重叠、经过泄漏过滤的保留子集上，按病因分类的模型在所有三个目标病因上均优于混合基线：CP（宏观 F1 0.829 对比 0.676，+22.6% 相对提升）、PD（0.715 对比 0.511，+40.0%）和 ALS（0.788 对比 0.596，+32.3%）。在 CP 上，增加 144 个 SAP 和 44 个 CDSD 伪标注说话者使宏观 F1 从 0.786 提升至 0.829，相较于仅使用临床数据的 CP 模型提升了 4.3 个百分点。训练数据涵盖每种病因的三到七种语言。我们将此视为标签空间设计选择的受控比较，并讨论伪标签校准、数据划分的卫生性以及基于置信度的部署作为未来工作的重大限制。
View on arXiv Download PDF AI Translation
cs.CL / 55 / 2609.21827

RheoSampling: Resolving the One-Hot Dilemma in Stochastic Dynamic-Tree Speculative Decoding

RheoSampling：解决随机动态树推测解码中的一热编码困境
Hu, Qiao, Weng, Yepeng, Zhang, Bo, Yairi, Takehisa
Abstract
Speculative decoding accelerates LLM inference by drafting multiple tokens in parallel, with tree-based methods further improving efficiency through hierarchical structures. Dynamic-tree methods such as EAGLE-3 perform well under greedy decoding via deterministic top-K expansion and global pruning. However, in stochastic decoding (T>0), this mechanism collapses the draft distribution into one-hot probabilities, causing a severe drop in acceptance rate. This creates a dilemma: dynamic-tree methods sacrifice stochastic sampling to preserve context-aware topology, while static-tree methods preserve stochastic sampling with context-agnostic structures. The issue arises because the same probability distribution is used for two conflicting tasks: constructing the tree and verifying tokens. This coupling makes direct injection of randomness challenging due to the resulting stochastic process. We resolve this by decoupling these roles: RheoSampling assigns a token sampled from the draft distribution a proxy probability for tree expansion and pruning alongside its true sampling probability for verification. Specifically, we inject a sampled token among the deterministic top-K slots and treat it with different probabilities during construction and verification, making RheoSampling the first dynamic-tree method with both context-aware top-K construction and stochastic sampling while maintaining losslessness. We establish the lossless guarantee through an equivalence-class analysis that compresses the stochastic tree space into tractable classes. An OT-based verification strategy and a sparse draft mechanism ensure that theoretical gains translate into practical efficiency. Experiments across LLMs and benchmarks demonstrate improvements in acceptance rate and speedup over state-of-the-art dynamic tree methods. This framework may provide a template for analyzing stochastic tree structures.
Chinese Translation
推测解码通过并行草拟多个标记来加速大规模语言模型（LLM）的推理，而基于树的方法通过层次结构进一步提高了效率。动态树方法如 EAGLE-3 在贪婪解码下通过确定性 top-K 扩展和全局剪枝表现良好。然而，在随机解码（T>0）中，该机制将草拟分布压缩为一热概率，导致接受率严重下降。这造成了一个困境：动态树方法牺牲随机采样以保持上下文感知的拓扑结构，而静态树方法则在上下文无关的结构中保留随机采样。问题的根源在于同一概率分布被用于两个相互冲突的任务：构建树和验证标记。这种耦合使得直接注入随机性变得困难，因为由此产生的随机过程。我们通过解耦这些角色来解决这一问题：RheoSampling 为从草拟分布中采样的标记分配一个用于树扩展和剪枝的代理概率，同时保留其用于验证的真实采样概率。具体而言，我们在确定性 top-K 插槽中注入一个采样标记，并在构建和验证过程中以不同的概率对待它，使 RheoSampling 成为首个同时具备上下文感知 top-K 构建和随机采样的动态树方法，同时保持无损性。我们通过等价类分析建立了无损保证，将随机树空间压缩为可处理的类。基于 OT 的验证策略和稀疏草拟机制确保理论收益转化为实际效率。在 LLM 和基准测试中的实验表明，接受率和速度相较于最先进的动态树方法有所提高。该框架可能为分析随机树结构提供模板。
View on arXiv Download PDF AI Translation
cs.CL / 56 / 2609.21844

Reusing Latent Speech Representations for Query-Conditioned Topic Localization in Transcripts

重用潜在语音表示进行转录中的查询条件主题定位
Freisinger, Steffen, Seeberger, Philipp, Ranzenberger, Thomas, Bocklet, Tobias, Riedhammer, Korbinian
Abstract
Long transcripts are costly inputs for downstream NLP systems and often contain irrelevant context. We study query-conditioned topic localization: predicting the sentence span in a transcript that best addresses a topic-title query. To improve span localization, we reuse ASR encoder states as sentence-level representations and fuse them with textual embeddings. This lets lightweight span locators exploit speech information without running a separate audio encoder. Experiments on two public datasets show consistent gains over text-only baselines, especially under strict boundary-matching criteria. Cross-dataset experiments further indicate that the benefits are strongest for structured or semi-structured speech, while gains on spontaneous speech are limited and mixed.
Chinese Translation
长篇转录文本是下游自然语言处理系统的高成本输入，且通常包含无关的上下文。我们研究查询条件的主题定位：预测转录文本中最能回应主题标题查询的句子范围。为了改善范围定位，我们重用自动语音识别（ASR）编码器的状态作为句子级表示，并将其与文本嵌入融合。这使得轻量级的范围定位器能够利用语音信息，而无需运行单独的音频编码器。在两个公共数据集上的实验表明，相较于仅使用文本的基线方法，性能有一致的提升，尤其是在严格的边界匹配标准下。跨数据集的实验进一步表明，这些益处在结构化或半结构化语音中最为显著，而在自发语音中的提升则有限且表现不一。
View on arXiv Download PDF AI Translation
cs.CL / 57 / 2609.21857

Do Personality-Tuned LLMs Make Better Social Agents?

个性调优的大型语言模型是否能成为更好的社交代理？
Krabbe, Tim, Shi, Xiaodan
Abstract
LLMs are increasingly used in social simulations for socially interactive agents and robots, offering more flexibility than rule-based systems. However, even though they mimic human behaviour very well, there is a persistent alienness to them. This work investigates whether personality-aware fine-tuning can reduce this gap by improving the consistency and controllability of personality-conditioned dialogue generation compared with instruction prompting alone. We fine-tune two small open-weight LLMs, Qwen2.5-7B-Instruct and Ministral-8B-Instruct, using a corpus that combines personality-labelled social media posts and dialogues to create a personality-based dialogue engine for social simulation. The resulting models are evaluated across multiple social interaction scenarios using three independent LLM judges, which assess personality fidelity and provide evidence-based behavioral interpretations. We additionally quantify inter-rater agreement and lexical characteristics of the generated dialogue. Results indicate that fine-tuned models are not better at role-playing different personalities than their respective baseline models. However, low inter-rater agreement limits the confidence with which these results can be interpreted. Concerning the quality of generated texts, fine-tuned models are mostly comparable to the baselines, with fine-tuning improving the linguistic diversity of the Qwen models. While the results appear generally usable and the baseline models offer the best overall performance, future studies should place greater emphasis on the quality and domain alignment of training data for accurate personality role-playing.
Chinese Translation
大型语言模型（LLMs）在社交模拟中越来越多地被用于社交互动代理和机器人，提供了比基于规则的系统更大的灵活性。然而，尽管它们能够很好地模拟人类行为，但仍然存在一种持久的陌生感。本研究探讨了个性意识的微调是否能够通过改善个性条件对话生成的一致性和可控性，从而缩小这一差距，与仅依赖指令提示相比。我们对两个小型开放权重的LLM进行了微调，分别是Qwen2.5-7B-Instruct和Ministral-8B-Instruct，使用一个结合了个性标签的社交媒体帖子和对话的语料库，创建了一个基于个性的对话引擎用于社交模拟。通过三位独立的LLM评审者对生成的模型在多种社交互动场景中的表现进行评估，这些评审者评估个性忠实度并提供基于证据的行为解释。此外，我们还量化了评审者间的一致性和生成对话的词汇特征。结果表明，微调模型在角色扮演不同个性方面并不优于其各自的基线模型。然而，低评审者间一致性限制了对这些结果的解释信心。就生成文本的质量而言，微调模型大致上与基线模型相当，微调提高了Qwen模型的语言多样性。尽管结果看起来普遍可用，且基线模型提供了最佳的整体表现，但未来的研究应更加重视训练数据的质量和领域对齐，以实现准确的个性角色扮演。
View on arXiv Download PDF AI Translation
cs.CL / 58 / 2609.21859

TrialAtlas: Multi-Agent Research Organization for Clinical Trial Design and Optimization

TrialAtlas：临床试验设计与优化的多智能体研究组织
Lin, Jiacheng, Wang, Zifeng, Chen, Zheng, Scott, Erick, Yang, Ziwei, Yu, Fanyang, Zhong, Sheng, Sun, Jimeng
Abstract
Nearly 90% of drugs entering clinical development ultimately fail, despite billions of dollars in investment. Pharmaceutical companies therefore rely on clinical development planning (CDP) and probability of technical and regulatory success assessment to anticipate development risks, yet these decisions remain labor-intensive and subjective, requiring experts across clinical science, statistics, regulatory affairs, and competitive intelligence to jointly acquire, synthesize, and reason over heterogeneous evidence. Here, we introduce TrialAtlas, a memory-augmented multi-agent research organization for CDP that mirrors this collaborative process by coordinating specialized agents for literature synthesis, competitive trial intelligence, regulatory precedent analysis, and integrated reasoning over trial design and development risk. TrialAtlas further learns from historical clinical trials and regulatory outcomes, including prior New Drug Applications (NDAs), to ground its decisions in accumulated development experience. To evaluate these capabilities in an authentic regulatory setting, we introduce TrialAtlasBench, constructed from 291 FDA Complete Response Letters and spanning three practical tasks: detecting trial design deficiencies, recommending actionable design improvements, and predicting technical and regulatory success. TrialAtlas achieves an F1 score of 50.0% for deficiency detection, outperforming the strongest baseline by 6.1 points, and reaches 85.3% balanced accuracy and 84.7% F1 for prediction of technical and regulatory success, improving over the best baselines by 6.7 points in balanced accuracy and 12.0 points in Cohen's kappa. In expert evaluation, 86.4% of TrialAtlas-generated concerns were judged valid, compared with 83.1% for OpenAI DeepResearch and 59.3% for Gemini DeepResearch.
Chinese Translation
尽管投入了数十亿美元，近90%的药物在进入临床开发后最终失败。因此，制药公司依赖临床开发规划（CDP）和技术及监管成功概率评估来预见开发风险，但这些决策仍然是劳动密集型和主观的，要求临床科学、统计学、监管事务和竞争情报等领域的专家共同获取、综合和推理异构证据。在此，我们介绍了TrialAtlas，一种增强记忆的多智能体研究组织，用于CDP，模拟这一协作过程，通过协调专门的智能体进行文献综合、竞争试验情报、监管先例分析以及对试验设计和开发风险的综合推理。TrialAtlas进一步从历史临床试验和监管结果中学习，包括之前的新药申请（NDA），以将其决策基于积累的开发经验。为了在真实的监管环境中评估这些能力，我们引入了TrialAtlasBench，该基准由291份FDA完整回应信构建，涵盖三个实际任务：检测试验设计缺陷、推荐可行的设计改进以及预测技术和监管成功。TrialAtlas在缺陷检测方面达到了50.0%的F1分数，超越最强基线6.1个百分点，在技术和监管成功预测中达到了85.3%的平衡准确率和84.7%的F1分数，分别比最佳基线提高了6.7个百分点的平衡准确率和12.0个百分点的Cohen's kappa。在专家评估中，TrialAtlas生成的关注点中有86.4%被判断为有效，而OpenAI DeepResearch为83.1%，Gemini DeepResearch为59.3%。
View on arXiv Download PDF AI Translation
cs.CL / 59 / 2609.21967

NemotronLabs VoiceChat: An Open Full-duplex Speech-to-Speech Model with Tool Calling Capabilities

NemotronLabs VoiceChat：一个具有工具调用能力的开放全双工语音到语音模型
Balam, Jagadeesh, Bartley, Travis, Casanova, Edresson, Chauhan, Sanjay, Chen, Chen, Chen, Zhehuai, Chen, Zijia, Ciannella, Francesco, Deng, Slyne, Desta, Mikyas, Dubey, Harishchandra, Essid, Slim, Ferchichi, Nourchene, Ginsburg, Boris, Fuenmayor, Mariana Graterol, Habibi, Negar, Hu, Kevin, Joseph, Anand, Karandikar, Viraj, Kim, Myungjong, Klimkov, Viacheslav, Narasimhan, Seelan Lakshmi, Lee, Lily, Li, Jason, Long, Eileen, Mahabaleshwarkar, Ameya, Malte, Aditya, Margolin, Adi, Meister, Sasha, Mendelev, Valentin, Olabiyi, Oluwatobi, Pasad, Ankita, Peng, Yifan, Rastorgueva, Elena, Ritchie, Jayda, Roche, Jason, Srihari, Nikhil, Su, Yuanhang, Suhara, Yoshi, Trinh, Viet Anh, Wang, Jinhan, Zelasko, Piotr, Wang, Hui, Meng, Puhui, Zhang, Chaosen, Liu, Yunsheng, Wang, Shawn, Li, Wenjing, He, Zhonglei
Abstract
We introduce NemotronLabs VoiceChat, an open full-duplex speech-to-speech model with native tool-calling capabilities. NemotronLabs VoiceChat combines a streaming speech encoder and decoder-only language model with parallel specialized output streams for agent text and structured function calls, an auxiliary RNN-T branch for incremental user transcription, and a streaming TTS decoder. This design enables the model to listen, transcribe, reason, invoke tools, and speak within a unified streaming architecture while preserving the temporal behavior required for natural conversation. On Full-Duplex-Bench 1.0, NemotronLabs VoiceChat achieves the lowest pause-handling takeover rates among evaluated open-weight systems, 100\% takeover following user interruptions, and a 4.33/5 post-interruption response-quality score. On Full-Duplex-Bench 1.5, it resumes its response after user backchannels in 93\% of cases. NemotronLabs VoiceChat obtains a 55.1 normalized average on VoiceBench and, on Full-Duplex-Bench 3.0 (FDB 3.0), achieves 82.5\% tool-selection F1, while argument accuracy and end-to-end tool execution remain areas for improvement. These results demonstrate that full-duplex interaction, speech recognition and generation, general language capabilities, and external tool use can be integrated in a single open speech-to-speech model without sacrificing real-time conversational behavior.
Chinese Translation
我们介绍了NemotronLabs VoiceChat，一个具有原生工具调用能力的开放全双工语音到语音模型。NemotronLabs VoiceChat结合了流式语音编码器和仅解码的语言模型，以及用于代理文本和结构化函数调用的并行专用输出流、用于增量用户转录的辅助RNN-T分支和流式文本转语音（TTS）解码器。该设计使模型能够在统一的流式架构中进行监听、转录、推理、调用工具和发声，同时保持自然对话所需的时间行为。在Full-Duplex-Bench 1.0上，NemotronLabs VoiceChat在评估的开放权重系统中实现了最低的暂停处理接管率，用户中断后100%的接管率，以及4.33/5的中断后响应质量评分。在Full-Duplex-Bench 1.5中，它在93%的情况下能够在用户反馈后恢复响应。NemotronLabs VoiceChat在VoiceBench上获得了55.1的标准化平均分，并在Full-Duplex-Bench 3.0（FDB 3.0）上实现了82.5%的工具选择F1得分，而参数准确性和端到端工具执行仍然是需要改进的领域。这些结果表明，全双工交互、语音识别与生成、通用语言能力以及外部工具的使用可以在单一开放的语音到语音模型中集成，而不牺牲实时对话行为。
View on arXiv Download PDF AI Translation
cs.CL / 60 / 2609.21992

Moral Entropy: Auditing Bias and Uncertainty in Moral Judgment

道德熵：审计道德判断中的偏见与不确定性
Skorski, Maciej
Abstract
Most work in computational ethics treats annotator disagreement on moral content as noise to be voted away, collapsed into majority vote or the more permissive any-annotator rule the moment a single annotator flags an item. We argue this uncertainty should instead be modeled and learned from. We introduce Moral Entropy, a Bayesian framework that keeps a full posterior over the true label and decomposes its entropy into aleatoric uncertainty (irreducible disagreement about the moral content) and epistemic uncertainty (from insufficient or noisy annotation) -- and lets any heuristic consensus rule be audited against a calibrated ground truth via entropy methods such as cross-entropy/KL, Brier score, and expected calibration error. Across three corpora and fifteen discourse domains, auditing the standard aggregation rules against this posterior reveals bias that no current pipeline reports: the any-annotator rule disagrees with the calibrated posterior on roughly 30% of items -- pooled, almost entirely false positives, though the errors invert at the foundation level (19.9%/38.9% mean FPR/FNR on MFTC) -- while the stricter majority and two-vote rules miss 63-83% of true positives.
Chinese Translation
大多数计算伦理学的研究将标注者对道德内容的分歧视为噪声，通常通过投票消除，或在单个标注者标记某项内容时，归结为多数投票或更宽松的任何标注者规则。我们认为，这种不确定性应当被建模并加以学习。我们引入了道德熵（Moral Entropy），这是一个贝叶斯框架，保持对真实标签的完整后验分布，并将其熵分解为随机不确定性（对道德内容的不可减少的分歧）和认知不确定性（由于标注不足或噪声造成的）——并允许任何启发式共识规则通过熵方法（如交叉熵/KL、Brier分数和期望校准误差）与经过校准的真实情况进行审计。在三个语料库和十五个话语领域中，审计标准聚合规则与这一后验分布相比，揭示了当前管道未报告的偏见：任何标注者规则与校准后的后验分布在大约30%的项目上存在分歧——汇总后几乎完全是误报，尽管在基础层面上错误是相反的（在MFTC上平均假阳性率/假阴性率为19.9%/38.9%）——而更严格的多数和两票规则则错过了63-83%的真实阳性。
View on arXiv Download PDF AI Translation
cs.CL / 61 / 2609.22000

RecreationWorld: Scalable and Verifiable Environments for Hybrid Computer-Use Agents

RecreationWorld：可扩展和可验证的混合计算机使用代理环境
Bai, Shuai, Deng, Jiayong, Fu, Yikun, Gao, Chang, Hu, Xuhao, Huang, Mianqiu, Jiang, Yizhen, Jing, Yuheng, Kong, Dehui, Li, Keliang, Li, Ning, Li, Wanli, Liu, Dayiheng, Lu, Dunjie, Luo, Changwei, Shen, Que, Wang, Zheyuan, Wang, Zijian, Wu, Jie, Wu, Gao, Xie, Zhihui, Xie, Rui, Xu, Haiyang, Yang, An, Yuan, Jiakang, Zhang, Yanming, Zhang, Jiajun, Zhang, Xi, Zhang, Zhenru, Zhen, Zhuo, Zhu, Mingkang, Zhou, Bowen
Abstract
Computer-use agents (CUAs) have advanced along two separate lines: graphical interaction and software development through code and the command line. Real digital work requires both, interleaved rather than stacked end to end. We study hybrid CUAs that autonomously decide when to explore an interface, implement software, and run and visually verify their artifacts. We introduce RecreationWorld, a five-platform framework built around recreation: given a running reference, an agent must discover its behavior and build a faithful implementation with no prescribed workflow. RecreationWorld provides reproducible environments on Ubuntu, macOS, Windows, Android, and Web, plus a unified harness with native GUI control and coding tools. The running reference serves as an oracle for hidden behavioral tests, providing execution-grounded rewards. We scale trajectory generation with high-quality open-source applications. Models trained on these trajectories improve across five out-of-distribution coding and hybrid computer-use benchmarks and more frequently verify their rendered outputs, providing evidence of transfer beyond recreation. For held-out evaluation, we introduce RecreationBench, comprising 250 diverse tasks across domains and platforms. Reference-grounded programmatic and visual assertions cover action-conditioned outcomes at multiple interaction depths; each is validated on the reference and by human reviewers before the suite is frozen for automatic scoring. GPT-6 Astra leads at 58.1% overall, but passes all programmatic tests on just 2.8% of tasks. Agents reproduce static interface structure more reliably than interactions and computed outputs, while generated applications remain smaller and more monolithic than their references. We release the benchmark, environments, and test suites.
Chinese Translation
计算机使用代理（CUAs）沿着两个独立的方向发展：图形交互和通过代码及命令行进行的软件开发。真实的数字工作需要两者交替进行，而不是简单地串联。我们研究能够自主决定何时探索界面、实施软件以及运行和视觉验证其成果的混合CUAs。我们介绍了RecreationWorld，这是一个围绕重创作构建的五平台框架：在给定运行参考的情况下，代理必须发现其行为并构建一个忠实的实现，而无需规定的工作流程。RecreationWorld在Ubuntu、macOS、Windows、Android和Web上提供可重复的环境，并配备统一的工具，具备本地GUI控制和编码工具。运行参考作为隐藏行为测试的神谕，提供基于执行的奖励。我们通过高质量的开源应用程序扩展轨迹生成。基于这些轨迹训练的模型在五个超出分布的编码和混合计算机使用基准测试中表现改善，并更频繁地验证其渲染输出，提供超越重创作的转移证据。对于保留评估，我们引入了RecreationBench，涵盖跨领域和平台的250个多样化任务。基于参考的程序化和视觉断言涵盖多个交互深度的动作条件结果；每个断言在参考上和由人类评审者验证后，才会在自动评分之前冻结该套件。GPT-6 Astra在整体上以58.1%的成绩领先，但仅在2.8%的任务上通过所有程序化测试。代理在重现静态界面结构方面比交互和计算输出更可靠，而生成的应用程序在规模上仍然小于其参考，且更为单一。我们发布了基准、环境和测试套件。
View on arXiv Download PDF AI Translation
cs.CL / 62 / 2609.22008

DiaVLo: Diagnosing Behaviours of Vision-Language Models

DiaVLo：诊断视觉-语言模型的行为
Corti, Lorenzo, Yang, Jie
Abstract
Vision-language models (VLMs) rely on storing and transferring appropriate information across their sub-components. Verifying that the VLMs exhibit desired behaviours, while avoiding harmful ones, is central to their reliable deployment. Yet, methods that identify VLM behaviours remain scarce. We present DiaVLo, a diagnostic framework that leverages human curation and VLMs' generation capabilities to construct specifications of desired and observed VLM behaviours, surfacing potential misalignments. Beyond this, DiaVLo also provides causal estimates to identify the most influential concepts steering VLM behaviours. We evaluate DiaVLo on several open-source VLMs under both classification and generation conditions. Our experiments show that DiaVLo produces behaviour labels that correlate with model performance and provide context for measured performance. DiaVLo surfaced behaviours that are clearly aligned and misaligned, alongside patterns in how VLMs perceive, organise, and prioritise concepts.
Chinese Translation
视觉-语言模型（VLMs）依赖于在其子组件之间存储和传递适当的信息。验证VLMs是否表现出期望的行为，同时避免有害行为，是其可靠部署的核心。然而，识别VLM行为的方法仍然稀缺。我们提出了DiaVLo，这是一种诊断框架，利用人类策划和VLM的生成能力来构建期望和观察到的VLM行为的规范，从而揭示潜在的不一致之处。除此之外，DiaVLo还提供因果估计，以识别影响VLM行为的最重要概念。我们在多个开源VLM上对DiaVLo进行了评估，涵盖了分类和生成条件。实验结果表明，DiaVLo生成的行为标签与模型性能相关，并为测量的性能提供了背景。DiaVLo揭示了明显一致和不一致的行为，以及VLM如何感知、组织和优先考虑概念的模式。
View on arXiv Download PDF AI Translation
cs.CL / 63 / 2609.22038

QuranicMMLU: A Cognitively-Aware Benchmark for Evaluating Generative AI Solutions on Quranic Linguistic Knowledge

QuranicMMLU：一个认知意识基准，用于评估生成性人工智能在古兰经语言知识上的表现
Ghali, Rawan El, Kulsoom, Umm, Madkoor, Anas, Alsaudi, Dima Faris, Abdelmagid, Roaa, Ibrahim, Roaa, Mousa, Raghad, Aljaji, Hamza, Khanafer, Abdullah, Alkanani, Abdallah, Alali, Salah Feras, Mohamed, Rawan Khaled, Asgari, Ehsaneddin
Abstract
We introduce QuranicMMLU, a benchmark for evaluating generative AI on Quranic Arabic across multiple dimensions of linguistic complexity. Existing Quranic benchmarks center on general question answering and semantic retrieval, without probing specific linguistic competencies or stratifying by cognitive demand and verse difficulty. We construct a five-pillar Quranic taxonomy spanning Phonology, Morphology, Syntax, Semantics, and Pragmatics, with 31 leaves covering phenomena from tajw\=id and root-and-pattern morphology to occasions of revelation and inter-surah coherence. For each leaf we generate questions stratified by Bloom's cognitive level and verse perplexity, then have LLM as a judge to independently answer and score every item and route the annotations to manual review. The resulting dataset comprises 980 human-reviewed questions, each issued in both open-ended and multiple-choice form. We benchmark 12 systems on these items and find that the Islamic-specialized model leads, yet every system scores higher on multiple-choice accuracy (average 84%) than open-ended answer quality (average 60%): the two rankings agree closely (Kendall's {\tau}=0.73), but multiple-choice scoring hides failures that surface only once answer choices are removed. QuranicMMLU thus offers a rigorous, linguistically grounded framework for evaluating Arabic NLP in the Quranic domain.
Chinese Translation
我们介绍了QuranicMMLU，这是一个用于评估生成性人工智能在古兰经阿拉伯语中多维度语言复杂性的基准。现有的古兰经基准主要集中在一般问答和语义检索上，而未深入探讨特定的语言能力或按认知需求和经文难度进行分层。我们构建了一个涵盖音韵学、形态学、句法学、语义学和语用学的五大支柱古兰经分类法，包含31个分支，涵盖从tajwīد和根-模式形态学到启示场合和经文间连贯性的现象。对于每个分支，我们生成按布鲁姆认知水平和经文困惑度分层的问题，然后让大型语言模型（LLM）作为评审独立回答并评分每个项目，并将注释提交人工审核。最终的数据集包含980个经过人工审核的问题，每个问题以开放式和多项选择两种形式发布。我们在这些项目上对12个系统进行了基准测试，发现专注于伊斯兰的模型表现最佳，但每个系统在多项选择准确性（平均84%）上得分高于开放式答案质量（平均60%）：这两种排名高度一致（肯德尔τ=0.73），但多项选择评分掩盖了在移除答案选项后才显现的失败。因此，QuranicMMLU为评估古兰经领域的阿拉伯自然语言处理提供了一个严格且以语言为基础的框架。
View on arXiv Download PDF AI Translation
cs.CL / 64 / 2609.22043

An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency

基于三信号互补性的可解释记忆决策控制器用于大型语言模型代理：解耦置信度与一致性
Zhang, Yiming, Zhang, Jinghong, Zhao, Haoran, Ma, Yiren, Zhao, Chunlei
Abstract
Memory systems for large language models have focused predominantly on efficient retrieval, whereas the decision of whether retrieved memories should be trusted has received comparatively little attention. When the memory store contains conflicting positions, standard retrieval-augmented generation (RAG) blindly injects memories and amplifies hallucinations: in models susceptible to memory injection, the RAG hallucination rate under conflicting memories is markedly higher than that of a memory-free baseline. Inspired by memory signaling mechanisms in the prefrontal cortex, we propose the Memory Decision Layer (MDL), a zero-parameter memory decision controller situated between the retrieval and generation stages. Its core is a three-signal complementary encoder that fuses relevance, reliability, and task risk through QR-based orthogonal subspace projection and a meta-working-memory signal into an interpretable decision representation that quantifies the trustworthiness of retrieved memories. Building on this encoder, MDL explicitly decouples confidence from consistency and introduces risk inversion and explicit abstention. Evaluations on mainstream large language models and multiple open-source datasets show that MDL reduces the hallucination rate under conflicting memories by about 56.04% in general scenarios and approaches zero hallucination in high-risk scenarios. The controller is fully white-box: it relies purely on geometric operations, requires no trained parameters, and adds only about 0.14 ms per decision -- roughly 50x faster than the embedding-retrieval step that precedes it and four to five orders of magnitude faster than an LLM self-evaluation call.
Chinese Translation
大型语言模型的记忆系统主要集中在高效检索上，而对检索到的记忆是否应该被信任的决策则相对较少关注。当记忆库中包含冲突的信息时，标准的检索增强生成（RAG）会盲目注入记忆并放大幻觉：在易受记忆注入影响的模型中，冲突记忆下的RAG幻觉率显著高于没有记忆的基线。受到前额叶皮层中记忆信号机制的启发，我们提出了记忆决策层（Memory Decision Layer, MDL），这是一个零参数的记忆决策控制器，位于检索和生成阶段之间。其核心是一个三信号互补编码器，通过基于QR的正交子空间投影和元工作记忆信号融合相关性、可靠性和任务风险，形成一个可解释的决策表示，量化检索到的记忆的可信度。在此编码器的基础上，MDL明确地将置信度与一致性解耦，并引入风险反转和明确的弃权。在主流大型语言模型和多个开源数据集上的评估表明，MDL在一般场景下将冲突记忆下的幻觉率降低了约56.04%，在高风险场景中接近零幻觉。该控制器完全为白盒：它仅依赖几何运算，不需要训练参数，每次决策仅增加约0.14毫秒的时间——大约比其前面的嵌入检索步骤快50倍，比大型语言模型自我评估调用快四到五个数量级。
View on arXiv Download PDF AI Translation
cs.CL / 65 / 2609.22081

Cross-sector generalization of accident-process role classification in occupational accident narratives

职业事故叙述中事故过程角色分类的跨行业泛化研究
Yapi, Aho, Latouche, Pierre, Guillin, Arnaud, Bailly, Yan
Abstract
Occupational accident narratives contain valuable information about work situations, unfavourable conditions, accident events, and their consequences. Automatically structuring these narratives can facilitate large-scale accident analysis and support occupational risk prevention. However, the terminology and writing styles used to describe accidents vary considerably across sectors and organisations, raising questions about the ability of automated coding systems to generalize beyond their training domain. In this paper, we evaluate the cross-sector generalization of accident-process role classification in French occupational accident narratives. We construct an expert-annotated corpus in which factual units are classified into four roles: work situation (A0), explicitly reported unfavourable condition (A1), accident event or deviation (B), and reported consequence (C). The role classifiers are developed and selected exclusively on 42,244 factual units extracted from 6,040 construction-sector narratives and are then evaluated on unseen corpora from the metallurgy and chemistry--plastics sectors, as well as on an independently collected company corpus, without retraining or target-domain tuning of the role classifier. We compare frozen pretrained representations with task-specific fine-tuning and supervised representation-learning strategies. The results show that task-specific adaptation consistently improves cross-domain transfer over frozen representations. Across repeated training runs, the three leading task-adapted strategies achieved average balanced accuracies between 85.6% and 85.8% across the three target corpora. These findings support the development of transferable assisted-coding systems capable of consistently structuring heterogeneous occupational accident narratives for expert review and cross-sector prevention analysis.
Chinese Translation
职业事故叙述包含关于工作情境、不利条件、事故事件及其后果的宝贵信息。自动化结构化这些叙述有助于大规模事故分析并支持职业风险预防。然而，不同行业和组织在描述事故时所使用的术语和写作风格存在较大差异，这对自动编码系统在训练领域之外的泛化能力提出了挑战。本文评估了法语职业事故叙述中事故过程角色分类的跨行业泛化能力。我们构建了一个专家标注语料库，将事实单元分类为四类角色：工作情境（A0）、明确报告的不利条件（A1）、事故事件或偏差（B）以及报告的后果（C）。角色分类器仅基于来自建筑行业的6,040篇叙述中提取的42,244个事实单元进行开发和选择，随后在未见过的冶金和化工-塑料行业语料以及一个独立收集的企业语料上进行评估，且未对角色分类器进行再训练或目标领域调优。我们比较了冻结的预训练表示与任务特定微调和监督表示学习策略。结果表明，任务特定的适应显著提升了跨领域迁移性能。三种领先的任务适应策略在三种目标语料上的平均平衡准确率在85.6%至85.8%之间。该研究结果支持开发可转移的辅助编码系统，能够稳定地结构化异质的职业事故叙述，以供专家审查和跨行业预防分析。
View on arXiv Download PDF AI Translation
arXiv Reader
Automatically generated on 2026-10-15 17:15:59
Powered by OpenAI GPT & arXiv