cs.RO / 1 / 2609.17540
Human-Centric Grasp State Assessment: Toward Transferring Subjective Evaluation to Robots
以人为中心的抓取状态评估：向机器人转移主观评估
Abstract
We propose a framework that transfers tacit human subjective criteria to robotic systems for the appropriate grasping of deformable objects. Achieving such behavior is challenging because a semantic gap exists between qualitative human expectations and quantitative robotic measurements. Conventional deep learning approaches for bridging this gap also require prohibitive amounts of manually annotated data for each newly encountered object. To address these challenges, our framework integrates a Vision-Language Model (VLM)-based semi-automated supervisor generator with a lightweight grasp state predictor, using a minimal set of human-annotated trials as contextual anchors to propagate subjective criteria to unannotated data. The prediction model then enables rapid online adaptation by sequentially estimating the grasp state from time-series tactile and grasping force measurements. Through experiments on three representative deformable objects and a human evaluation study with 25 participants, we demonstrate the feasibility of the proposed framework for adjusting grasping force according to human-perceived grasp appropriateness in the evaluated task setting.
Chinese Translation
我们提出了一个框架，将隐性的人工主观标准转移到机器人系统，以适当地抓取可变形物体。实现这种行为具有挑战性，因为定性的人类期望与定量的机器人测量之间存在语义差距。传统的深度学习方法虽然可以弥补这一差距，但也需要对每个新遇到的物体进行大量手动标注的数据。为了解决这些挑战，我们的框架将基于视觉-语言模型（Vision-Language Model, VLM）的半自动化监督生成器与轻量级抓取状态预测器相结合，利用一小部分人类标注的试验作为上下文锚点，将主观标准传播到未标注的数据中。然后，预测模型通过顺序估计来自时间序列触觉和抓取力测量的抓取状态，实现快速的在线适应。通过对三种代表性可变形物体的实验以及与25名参与者的人工评估研究，我们展示了所提框架在评估任务设置中根据人类感知的抓取适当性调整抓取力的可行性。
cs.RO / 2 / 2609.17541
Real-Time Service Robot Replanning via Simple Button Interaction for Improved Task Success and User Experience
通过简单按钮交互实现实时服务机器人重新规划，以提高任务成功率和用户体验
Abstract
Service robots must respond to unexpected instructions in real-world environments. However, robots cannot detect all failures and exceptions during a task. To address these issues, we propose a real-time feedback function that enables robots to modify their behavior based on human feedback. In this system, users can intuitively send feedback to the robot by pressing a single button on a tablet when the robot fails to act correctly. Robots use this feedback to consider their failures and replan appropriate actions to complete the task. We conducted experiments with and without the feedback function to verify the following hypothesis: "Simple interactions do not cause a negative user experience." All questionnaire responses are evaluated on a five-point Likert scale. After adding the feedback function, the response score for the question "Did you feel that the robot's behavior was unexpected?" improved by 0.5 points, and that for "Did you feel anxious about the robot's behavior at times?" improved by 0.9 points. These results support the study's hypothesis and indicate that incorporating this real-time feedback function can simultaneously improve task success and the user experience.
Chinese Translation
服务机器人必须能够在现实环境中响应意外指令。然而，机器人在执行任务时无法检测到所有的失败和异常。为了解决这些问题，我们提出了一种实时反馈功能，使机器人能够根据人类反馈调整其行为。在该系统中，当机器人未能正确执行时，用户可以通过在平板电脑上按下一个按钮直观地向机器人发送反馈。机器人利用这些反馈考虑其失败，并重新规划适当的行动以完成任务。我们进行了有无反馈功能的实验，以验证以下假设：“简单的交互不会导致负面的用户体验。”所有问卷反馈均采用五点李克特量表进行评估。在添加反馈功能后，关于“您是否觉得机器人的行为出乎意料？”这一问题的反馈得分提高了0.5分，而“您是否在某些时候对机器人的行为感到焦虑？”这一问题的得分提高了0.9分。这些结果支持了研究的假设，并表明，结合这一实时反馈功能可以同时提高任务成功率和用户体验。
cs.RO / 3 / 2609.17625
Predictive Varanus: Combining CSP Conformance Monitoring with Predictive LTL Runtime Verification
预测性瓦拉努斯：将 CSP 一致性监测与预测性 LTL 运行时验证相结合
Abstract
Runtime Verification is well suited to autonomous and robotic systems because it checks the behaviour that is actually observed during execution. Its main limitation, however, is that it is usually reactive: the monitor detects a violation only after the system has already performed a bad event. This can be too late in domains where failures are costly or unsafe. In this paper we present PREDICTIVE VARANUS, a two-stage verification pipeline that combines VARANUS, a runtime verifier that uses models written in the process algebra Communicating Sequential Processes (CSP), with predictive runtime verification for LTL. A CSP model is first used as a conformance gate over the observed event trace; the same model is then translated into a Buchi automaton that constrains the futures explored by a predictive LTL monitor. In this way, out-of-model behaviour is rejected immediately, while model-consistent prefixes can be classified as already guaranteeing satisfaction, already forcing violation, or still being inconclusive for the monitored temporal property. We formalise the combined monitor, explain its implementation, and illustrate the approach on a robotic rover for nuclear-store inspection. The case study shows how the combination of CSP validation and predictive LTL can provide earlier verdicts than standard runtime monitoring while reusing an existing design-time CSP model.
Chinese Translation
运行时验证非常适合自主和机器人系统，因为它检查在执行过程中实际观察到的行为。然而，它的主要限制在于通常是反应性的：监视器仅在系统已经执行了不良事件后才检测到违规。这在失败代价高昂或不安全的领域可能为时已晚。本文提出了 PREDICTIVE VARANUS，一个两阶段的验证管道，将 VARANUS（一个使用过程代数 Communicating Sequential Processes (CSP) 编写模型的运行时验证器）与 LTL 的预测性运行时验证相结合。首先，使用 CSP 模型作为观察到的事件轨迹的一致性门；然后将同一模型转换为一个布赫自动机（Buchi automaton），以约束预测性 LTL 监视器所探索的未来。在这种方式下，模型外行为会立即被拒绝，而模型一致的前缀可以被分类为已经保证满足、已经强制违规或仍然对监控的时间属性不确定。我们对组合监视器进行了形式化，解释了其实现，并在核存储检查的机器人探测器上说明了该方法。案例研究表明，CSP 验证与预测性 LTL 的结合能够比标准运行时监测提供更早的裁决，同时重用现有的设计时 CSP 模型。
cs.RO / 4 / 2609.17627
Flexible-body Modeling, Kinematic Identification, and Assembly Accuracy of Overconstrained Spatial Linkages
柔性体建模、运动学识别与过约束空间连杆的装配精度
Abstract
Overconstrained rational single-loop linkages are efficient, compact, and low-cost custom mechanisms, yet their deployment in industrial settings is limited. In simulations, rigid body formulations fail due to redundant constraints. This study presents a flexible multibody modeling framework based on the floating frame of reference formulation, and delivers an overall accuracy analysis of assembled linkages prototypes. The approach is validated against 3D-printed PLA prototypes of a Bennett four-bar mechanism, including variants with intentional joint-axis misalignment, which theoretically, from the rigid body point of view, cannot be assembled. A supplementary contribution is delivered in the form of a kinematic parameter identification methodology suited for this type of mechanism with ill-conditioned Jacobian. The experimental and simulation results are compared and reveal that these overconstrained mechanisms exhibit a self-assembling tendency -- structural compliance drives the assembly toward the ideal geometric configuration, distributing constraint stress throughout the structure. Additional qualitative demonstrations using cardboard tubes and bamboo sticks as link building blocks confirm that functional mechanisms can be realized from low-cost, unconventional materials with limited manufacturing accuracy. The proposed modeling pipeline is fully algorithmic and enables design optimization in the future.
Chinese Translation
过约束的理性单环连杆是一种高效、紧凑且低成本的定制机制，但其在工业环境中的应用仍然有限。在模拟中，由于冗余约束，刚体模型无法有效应用。本研究提出了一种基于浮动参考框架的柔性多体建模框架，并对组装的连杆原型进行了整体精度分析。该方法通过与3D打印的PLA原型（包括故意关节轴错位的变体）进行验证，而从刚体的角度来看，这些变体理论上是无法组装的。研究还提供了一种适用于此类具有病态雅可比矩阵机制的运动学参数识别方法。实验与模拟结果进行了比较，结果表明这些过约束机制展现出自组装的趋势——结构的柔顺性使装配趋向于理想几何配置，并在整个结构中分配约束应力。使用纸板管和竹棍作为连杆构建块的额外定性演示证实了功能机制可以通过低成本的非常规材料实现，尽管制造精度有限。所提出的建模流程是完全算法化的，并为未来的设计优化提供了可能。
cs.RO / 5 / 2609.17628
LEAP: Learning Emergent Active Perception for Quadruped Navigation
LEAP：学习四足导航的涌现主动感知
Abstract
Active perception allows autonomous agents to select their viewpoints rather than passively process the viewpoints given to them, enabling them to target where to reduce uncertainty about their environment. Learned systems typically encourage this behavior with hand-designed proxy objectives, such as coverage or curiosity bonuses, that may conflict with the task. In this work, we propose a method to learn emergent active perception (LEAP) without augmentation of the task objective. We formulate the problem of goal-oriented navigation over hazardous terrains with goals that must be discovered visually. We then propose an architecture for navigation policies with active perception, and train them on a terrain curriculum where task pressure alone leads to the emergence of gaze control. Key to this emergence, LEAP works on a gaze-invariant representation that integrates depth images into egocentric belief maps. We validate its performance in held-out evaluation scenarios, where it achieves a 92.7% success rate, compared to 74.2% for scripted or 34.5% for passive perception, and comes within 4.6 points of a privileged oracle. We validate that LEAP navigation policies, unchanged, can be directly applied to steering quadrupedal locomotion policies in physics simulation.
Chinese Translation
主动感知使自主代理能够选择其视点，而不是被动处理给定的视点，从而使它们能够针对性地减少对环境的未知性。学习系统通常通过手工设计的代理目标（如覆盖率或好奇心奖励）来鼓励这种行为，但这些目标可能与任务相冲突。在本研究中，我们提出了一种学习涌现主动感知（LEAP）的方法，无需增强任务目标。我们将目标导向的导航问题形式化为在危险地形上进行导航，目标需要通过视觉发现。接着，我们提出了一种具有主动感知的导航策略架构，并在一个地形课程上对其进行训练，在该课程中，仅依靠任务压力就能促使注视控制的涌现。LEAP的关键在于其在一个注视不变的表示上工作，该表示将深度图像整合到自我中心的信念图中。我们在保留的评估场景中验证了其性能，成功率达到92.7%，而脚本化方法为74.2%，被动感知为34.5%，并且与特权神谕的差距仅为4.6分。我们验证了LEAP导航策略在不改变的情况下，可以直接应用于物理仿真中的四足运动控制策略。
cs.RO / 6 / 2609.17714
Vision-Language Grounded Task-Context-Aware Imitation Learning for Robotic Disassembly
基于视觉-语言的任务上下文感知模仿学习用于机器人拆解
Abstract
Real-world robotic disassembly requires long-horizon execution, where robots must perform ordered sequences of manipulation tasks across multiple parts within a single scene. Multiple valid task goals and diverse assembly configurations make it difficult for imitation policies to infer the intended skill from raw observations alone, particularly when training data cannot cover the combinatorial diversity of real-world configurations and part geometries. We show that incorporating task context through language alleviates these challenges by providing explicit structure for skill selection and associating language-specified tasks with their corresponding manipulation targets in the visual scene. The proposed framework combines hierarchical task selection with task-context-aware imitation learning to ground language instructions in spatial visual representations for robotic disassembly. The resulting framework generalizes across diverse connector geometries and assembly configurations without requiring explicit object annotations. Our method improves end-to-end task success by 35 percentage points over the baseline diffusion policy and by 75 percentage points over the previous task-context-aware baseline.
Chinese Translation
现实世界中的机器人拆解需要长时间的执行，其中机器人必须在单一场景中执行多个部件的有序操作任务序列。多个有效的任务目标和多样的组装配置使得模仿策略仅凭原始观察很难推断出预期的技能，尤其是在训练数据无法覆盖现实世界配置和部件几何形状的组合多样性时。我们展示了通过语言引入任务上下文可以缓解这些挑战，因为它为技能选择提供了明确的结构，并将语言指定的任务与视觉场景中相应的操作目标关联起来。所提出的框架结合了分层任务选择与任务上下文感知的模仿学习，将语言指令扎根于空间视觉表示中，以实现机器人拆解。该框架在多样的连接器几何形状和组装配置中具有良好的泛化能力，而无需显式的物体注释。我们的方法在端到端任务成功率上比基线扩散策略提高了35个百分点，比之前的任务上下文感知基线提高了75个百分点。
cs.RO / 7 / 2609.17728
RAF-VLA: Representation Alignment with the Future for End-to-End Autonomous Driving
RAF-VLA：与未来对齐的表征用于端到端自主驾驶
Abstract
Recent Vision-Language-Action (VLA) models for autonomous driving have incorporated world modeling by predicting future driving scenes alongside driving actions, demonstrating strong planning performance. Future driving scenes are utilized as dense supervision, encouraging the policy to learn rich internal representations useful for planning. However, these World-Modeling VLAs rely on explicit future generation to learn such representations, thereby introducing two key limitations: additional training burden and inference latency. To address these limitations, we propose RAF-VLA (Representation Alignment with the Future), a VLA-based autonomous driving framework that shapes planning-relevant internal representations through direct guidance from future-frame representations. RAF-VLA employs Future-Aligned Supervised Fine-Tuning, in which a straightforward regularization aligns the policy's hidden states with future-frame representations obtained from a pretrained world encoder while learning driving actions. This simple alignment allows RAF-VLA to avoid the training burden and inference latency associated with future generation. Extensive experiments on the NAVSIM benchmark show that RAF-VLA achieves competitive planning performance against state-of-the-art VLA planners with substantially fewer training samples seen. Moreover, RAF-VLA incurs only 3.8% training overhead and a negligible 1 ms inference overhead.
Chinese Translation
近期的视觉-语言-动作（VLA）模型在自主驾驶中通过预测未来驾驶场景与驾驶动作相结合，融入了世界建模，展现出强大的规划性能。未来驾驶场景作为密集监督的来源，促使策略学习对规划有用的丰富内部表征。然而，这些世界建模的VLA依赖于显式的未来生成来学习这些表征，从而引入了两个关键限制：额外的训练负担和推理延迟。为了解决这些限制，我们提出了RAF-VLA（与未来对齐的表征），这是一种基于VLA的自主驾驶框架，通过来自未来帧表征的直接指导来塑造与规划相关的内部表征。RAF-VLA采用未来对齐的监督微调，其中一种简单的正则化方法将策略的隐藏状态与从预训练世界编码器获得的未来帧表征对齐，同时学习驾驶动作。这种简单的对齐使RAF-VLA能够避免与未来生成相关的训练负担和推理延迟。在NAVSIM基准上的大量实验表明，RAF-VLA在规划性能上与最先进的VLA规划器具有竞争力，同时所需的训练样本显著减少。此外，RAF-VLA仅产生3.8%的训练开销和可忽略的1毫秒推理开销。
cs.RO / 8 / 2609.17758
CALOS: Control-Affine Lyapunov On-manifold Safety Layer for Safe Deep Reinforcement Learning for Quadrotors
CALOS：用于四旋翼安全深度强化学习的控制仿射李雅普诺夫流形安全层
Abstract
Deep Reinforcement Learning has demonstrated remarkable capability in quadrotor control, yet learned policies offer no guarantee of respecting safety constraints during training or deployment. We present CALOS (Control-Affine Lyapunov On-manifold Safety), a runtime safety layer that enforces attitude constraints on a quadrotor without modifying the underlying learning algorithm. CALOS formulates four tilt-angle inequalities and a Lyapunov descent condition as a single quadratic program whose solution is the minimum-norm correction to the nominal torque output of the policy. The quadratic program is solved exactly via active-set enumeration over the three-dimensional torque space, with a computational cost low enough to enforce constraints in real time across thousands of parallel simulation environments, as required by modern massively parallel Deep Reinforcement Learning training. Evaluated on trajectory-tracking tasks in NVIDIA Isaac Lab, CALOS reduces lateral tracking error by 55-60% relative to an unconstrained Proximal Policy Optimization baseline while achieving zero attitude-constraint violations on the training trajectory. By restricting exploration to safe regions of the state space, the safety layer also accelerates training convergence and improves data efficiency without producing suboptimal policies.
Chinese Translation
深度强化学习在四旋翼控制方面展现了显著的能力，但所学习的策略在训练或部署过程中并不能保证遵守安全约束。我们提出了CALOS（控制仿射李雅普诺夫流形安全），这是一个运行时安全层，能够在不修改基础学习算法的情况下对四旋翼施加姿态约束。CALOS将四个倾斜角不等式和一个李雅普诺夫下降条件公式化为一个单一的二次规划，其解为对策略名义扭矩输出的最小范数修正。该二次规划通过对三维扭矩空间的活动集枚举精确求解，计算成本足够低，能够在现代大规模并行深度强化学习训练所需的数千个并行仿真环境中实时执行约束。在NVIDIA Isaac Lab的轨迹跟踪任务中评估，CALOS相较于无约束的近端策略优化基线减少了55-60%的横向跟踪误差，同时在训练轨迹上实现了零姿态约束违规。通过将探索限制在状态空间的安全区域，安全层还加速了训练收敛，并提高了数据效率，而不会产生次优策略。
cs.RO / 9 / 2609.17766
DRT&R: Direct Radar Teach & Repeat
DRT&R：直接雷达教学与重复
Abstract
Radar-based navigation is appealing for its robustness to adverse conditions involving airborne particles, such as precipitation, dust, fog, and smoke, that can cause lidar-based systems to fail. Recently, direct methods that retain and use the entire radar scan rather than sparse points have improved on-road global localization performance. However, they have yet to be deployed in off-road environments or in closed-loop systems. Additionally, even direct global maps may lose information: their global nature leads to a smoothing out of viewpoint-dependent radar artifacts, which can provide pose information when mapping and localization occur along similar trajectories. This paper introduces Direct Radar Teach & Repeat (DRT&R): a direct spinning radar-based navigation stack that maximizes the amount of retained information by combining direct radar processing with local mapping. DRT&R yields state-of-the-art (SOTA) localization performance in both on-road and off-road environments. Using 344 km of on-road data and 20 km of off-road data, DRT&R is able to localize to within 4 cm in most on-road and off-road conditions, and 12 cm in geometrically degenerate and sparse environments. DRT&R is also evaluated autonomously in closed loop with an MPC controller for more than 10 km using a Clearpath Warthog off-road vehicle, demonstrating that it runs in real time and achieves SOTA tracking performance for off-road radar navigation.
Chinese Translation
基于雷达的导航因其对不利条件的鲁棒性而备受关注，这些不利条件包括降水、灰尘、雾霾和烟雾等，这些因素可能导致基于激光雷达的系统失效。最近，直接方法通过保留和使用整个雷达扫描而非稀疏点，改善了公路全球定位性能。然而，这些方法尚未在越野环境或闭环系统中部署。此外，即使是直接的全球地图也可能会丢失信息：其全球性质导致视点依赖的雷达伪影被平滑化，而这些伪影在沿相似轨迹进行映射和定位时可以提供位姿信息。本文介绍了直接雷达教学与重复（DRT&R）：一种基于直接旋转雷达的导航系统，通过将直接雷达处理与局部映射相结合，最大化保留信息的量。DRT&R在公路和越野环境中均实现了最先进的定位性能。利用344公里的公路数据和20公里的越野数据，DRT&R能够在大多数公路和越野条件下实现4厘米的定位精度，并在几何退化和稀疏环境中实现12厘米的定位精度。DRT&R还在闭环中使用模型预测控制（MPC）控制器进行了超过10公里的自主评估，展示了其实时运行能力，并在越野雷达导航中实现了最先进的跟踪性能。
cs.RO / 10 / 2609.17771
HINT-Plan: Human Intention-Aware Robot Task Planning in Context-Rich Environments using Vision Language Models
HINT-Plan：在丰富环境中使用视觉语言模型进行人类意图感知的机器人任务规划
Abstract
Approaches to incorporating human awareness into mobile robot decision-making mainly focus on collision avoidance in low-level motion planning, often overlooking the challenges posed by human presence and high-level behavior. To address this vacancy, we present HINT-Plan, a novel approach to integrate human intention prediction into robot task planning. HINT-Plan employs Vision Language Models (VLMs) to anticipate high-level human intentions from third-person image observations, convert them into goal states, and solve joint task-planning problems. To effectively enable scene awareness in context-rich environments, we use hierarchical Scene Graphs (SGs) as high-level representations of the environment, and translate environmental topology and actionable knowledge into formal planning language to ensure executable plans. Evaluated in a photorealistic simulation, HINT-Plan achieves an overall success rate of 69.71% in joint human-robot task planning, substantially outperforming the baselines by up to 35.29%, while also reducing functional conflicts. The results show the effectiveness of explicitly incorporating inferred human intentions into formal multi-agent task planning for proactive human-aware robot decision-making.
Chinese Translation
将人类意识融入移动机器人决策中的方法主要集中在低级运动规划中的避碰，往往忽视了人类存在和高级行为带来的挑战。为了解决这一空白，我们提出了HINT-Plan，一种将人类意图预测集成到机器人任务规划中的新方法。HINT-Plan利用视觉语言模型（Vision Language Models, VLMs）从第三人称图像观察中预测高级人类意图，将其转化为目标状态，并解决联合任务规划问题。为了有效地在丰富环境中实现场景感知，我们使用分层场景图（Scene Graphs, SGs）作为环境的高级表示，并将环境拓扑和可执行知识转化为正式的规划语言，以确保可执行的计划。在一个照片级真实感的仿真中，HINT-Plan在联合人机任务规划中实现了69.71%的整体成功率，显著优于基线，提升幅度达到35.29%，同时减少了功能冲突。结果表明，明确地将推断的人类意图融入正式的多智能体任务规划中，对于主动的人类感知机器人决策具有有效性。
cs.RO / 11 / 2609.17781
SPROUT: The Open-Source Soft Growing Robot for Search and Rescue
SPROUT：开源软生长机器人用于搜索与救援
Abstract
Soft robotic systems have long been theorized as ideal candidates for use in search and rescue operations; however, there have been significant barriers to entry in graduating soft robotic systems from the laboratory to the field. To address this gap in replicable, reliable soft robot systems, we present the designs for SPROUT, the Soft Pathfinding Robotic Observation Unit. The system has matured over years of interaction with professional urban search and rescue communities, with the goal of operating in dusty, wet, and isolated conditions. This manuscript contains supplementary material, including code, parts manifests, CAD assemblies, and build instructions to allow the broader robotics research community to build their own SPROUT systems. The modular hardware and ROS 2-based software stack support task-specific payloads and control functions, allowing SPROUT to be adapted for applications beyond search and rescue, including infrastructure inspection and archaeology. Here, we demonstrate SPROUT performing a variety of challenging inspection and traversal tasks in collapsed structure training sites used by first responders. The main project page is available online at https://sprout-mitll.github.io/sprout/.
Chinese Translation
软体机器人系统长期以来被理论化为搜索与救援操作的理想候选者；然而，将软体机器人系统从实验室转向实际应用存在显著的障碍。为了解决可复制、可靠的软体机器人系统的这一缺口，我们提出了SPROUT（软路径寻找机器人观察单元）的设计。该系统经过与专业城市搜索与救援社区多年的互动而逐步成熟，旨在应对尘土飞扬、潮湿和孤立的环境。本文包含补充材料，包括代码、零件清单、CAD装配图和构建说明，以便更广泛的机器人研究社区能够构建自己的SPROUT系统。模块化硬件和基于ROS 2的软件堆栈支持特定任务的有效载荷和控制功能，使SPROUT能够适应搜索与救援以外的应用，包括基础设施检查和考古学。在此，我们展示了SPROUT在第一响应者使用的倒塌结构训练场地中执行各种具有挑战性的检查和穿越任务。项目的主要页面可在线访问：https://sprout-mitll.github.io/sprout/
cs.RO / 12 / 2609.17824
Learning Multi-Humanoid Pickup and Transport via Decentralized Object-Centric Control
通过去中心化物体中心控制学习多类人形机器人拾取与运输
Abstract
We study cooperative multi-humanoid pickup and transport of objects with varying size, weight, and geometry, requiring robot teams of different sizes. Our approach uses decentralized object-centric control, where each humanoid is assigned a local attachment region on the shared object and learns to realize pickup and transport through gripperless bimanual pinching. This attachment-based interface provides a common control abstraction spanning single-robot pickup, cooperative multi-robot transport, and robot-to-robot handover, without per-task redesign. We find that policies trained only on single-robot pickup already transfer nontrivially to cooperative settings, suggesting that this abstraction captures much of the structure needed for coordination. At the same time, explicit multi-robot training further improves performance, showing that shared-object coupling introduces coordination dynamics that are beneficial to learn directly. We validate the approach in simulation across varying team sizes and object geometries, and demonstrate sim-to-real transfer on hardware, where the learned controllers enable real humanoids to perform cooperative manipulation tasks.
Chinese Translation
我们研究了不同大小、重量和几何形状物体的多类人形机器人协作拾取与运输，这需要不同规模的机器人团队。我们的方法采用去中心化的物体中心控制，每个类人机器人被分配到共享物体上的局部附着区域，并通过无夹持的双手夹持学习实现拾取和运输。这种基于附着的接口提供了一种通用的控制抽象，涵盖了单机器人拾取、协作多机器人运输以及机器人间的交接，而无需针对每个任务重新设计。我们发现，仅在单机器人拾取上训练的策略已经能够非平凡地迁移到协作环境中，这表明这种抽象捕捉了协调所需的许多结构。同时，显式的多机器人训练进一步提高了性能，显示共享物体耦合引入的协调动态是直接学习的有益因素。我们在不同团队规模和物体几何形状的仿真中验证了该方法，并在硬件上展示了仿真到现实的迁移，其中学习到的控制器使真实的人形机器人能够执行协作操作任务。
cs.RO / 13 / 2609.17832
Adaptive-MHE : A Sampling-Based Adaptive MPC for Legged Loco-Manipulation via Moving Horizon Estimation
自适应-MHE：一种基于采样的自适应模型预测控制用于腿式机器人在移动视野估计下的运动操控
Abstract
Legged robots have demonstrated a remarkable ability to traverse various terrains, yet generating effective loco-manipulation behaviors remains challenging. A key difficulty is that object and terrain parameters are typically unknown to the robot, and mismatches between these parameters and their simulated counterparts introduce a sim-to-real gap that degrades control performance. Classical system identification (Sys-ID) methods often assume differentiable dynamics, an assumption that does not hold for contact-rich legged systems. Sampling-based Sys-ID avoids this restriction by directly matching simulated and recorded state trajectories through massively parallel rollouts, but existing approaches are typically applied offline and do not adapt as environmental conditions change. We present Adaptive-MHE an online sampling-based Sys-ID framework, based on moving horizon estimation (MHE), that estimates the physical parameters of objects and terrain in the environment (e.g., mass, friction) and couples this estimate with a sampling-based model predictive controller, enabling adaptive loco-manipulation in changing and uncertain environments. In simulation and hardware experiments, our framework consistently outperforms baselines and matches the performance of a controller with access to ground-truth parameters.
Chinese Translation
腿式机器人在各种地形上展现了卓越的行走能力，但生成有效的运动操控行为仍然具有挑战性。一个关键困难在于物体和地形参数通常对机器人是未知的，而这些参数与其模拟对应物之间的不匹配会引入模拟与现实之间的差距，从而降低控制性能。经典的系统识别（Sys-ID）方法通常假设动态可微，这一假设在接触丰富的腿式系统中并不成立。基于采样的Sys-ID通过通过大规模并行展开直接匹配模拟和记录的状态轨迹来避免这一限制，但现有方法通常是离线应用，并且在环境条件变化时不进行适应。我们提出了自适应-MHE，一个基于移动视野估计（MHE）的在线采样式Sys-ID框架，该框架估计环境中物体和地形的物理参数（例如，质量、摩擦），并将该估计与基于采样的模型预测控制器相结合，从而实现了在变化和不确定环境中的自适应运动操控。在仿真和硬件实验中，我们的框架始终优于基线，并与能够访问真实参数的控制器的性能相匹配。
cs.RO / 14 / 2609.17834
VCTP: Vehicle-Conditioned Terrain Planning for Off-Road Navigation
VCTP：基于车辆状态的越野导航地形规划
Abstract
A vehicle's heading affects both the surfaces beneath its tires and its pitch and roll. We present Vehicle-Conditioned Terrain Planning (VCTP), which retains these relationships by evaluating shared elevation and surface-ID layers at eight headings. Body geometry constrains admissibility, while loaded wheel contacts determine modeled surface cost and predicted pitch and roll. Established D* Lite and vehicle-state search use these evaluations to plan routes with forward and reverse motion. When observations change, VCTP recomputes every affected body or contact query. In fully observed two-track simulations, sampling at wheel contacts rather than at the vehicle center lowers modeled surface cost by 39.3% while shortening the route. In offline planning on RGator recordings, VCTP also reduces modeled costs over identified surfaces and observed support relative to distance-focused planning, although incomplete coverage leaves full-route rankings unresolved. Selective updates match full recomputation in all 474 comparisons using recorded map changes. These results identify when wheel-contact placement and vehicle heading affect route choice.
Chinese Translation
车辆的行驶方向影响其轮胎下的表面以及车辆的俯仰和横滚。我们提出了基于车辆状态的地形规划（Vehicle-Conditioned Terrain Planning, VCTP），通过在八个行驶方向上评估共享的高程和表面ID层，保持这些关系。车身几何形状限制了可接受性，而载荷轮接触则决定了建模表面的成本以及预测的俯仰和横滚。建立的 D* Lite 和车辆状态搜索利用这些评估来规划前进和后退的路线。当观测发生变化时，VCTP 会重新计算每个受影响的车身或接触查询。在完全观测的双轨模拟中，相较于在车辆中心进行采样，在轮接触点进行采样可将建模表面成本降低39.3%，同时缩短路线。在 RGator 记录的离线规划中，VCTP 也相对于以距离为重点的规划减少了对已识别表面和观察到的支撑的建模成本，尽管不完整的覆盖使得完整路线排名无法解决。在使用记录的地图变化进行的474次比较中，选择性更新与完全重新计算的结果相匹配。这些结果表明，轮接触位置和车辆行驶方向在路线选择中具有影响。
cs.RO / 15 / 2609.17904
Timely Activation of Safety Filters via One-Step Reachability Expansion
通过一步可达性扩展及时激活安全过滤器
Abstract
Least-restrictive safety filters based on Hamilton-Jacobi reachability provide strong safety guarantees by overriding a nominal controller only when the system reaches the boundary of the set of unsafe states defined as a Backward Reachable Tube (BRT). These guarantees, however, rely on the continuous-time nature of the underlying formulation. In practice, robotic systems apply control at discrete sampling intervals, which creates a mismatch where the system may jump into the unsafe BRT between updates, allowing failures that are theoretically avoidable. This work introduces a principled solution based on a one-step expanded BRT that predicts all states capable of reaching the true BRT within a single timestep. By using this expanded boundary as the activation condition for the safety filter, safety interventions occur early enough to ensure correctness under discrete-time execution. We formulate this expanded set as a modified reachability problem and compute it using standard continuous-time solvers.
Chinese Translation
基于哈密尔顿-雅可比可达性的最不严格安全过滤器通过仅在系统达到定义为反向可达管（Backward Reachable Tube, BRT）的不安全状态集合边界时覆盖名义控制器，从而提供强有力的安全保障。然而，这些保障依赖于基础公式的连续时间特性。在实际应用中，机器人系统在离散采样间隔内施加控制，这导致系统可能在更新之间跳入不安全的BRT，从而导致理论上可以避免的故障。本研究提出了一种基于一步扩展BRT的原则性解决方案，该方案预测所有能够在单个时间步内到达真实BRT的状态。通过将这个扩展边界作为安全过滤器的激活条件，安全干预能够在离散时间执行下足够早地发生，以确保正确性。我们将这个扩展集合表述为一个修改后的可达性问题，并使用标准的连续时间求解器进行计算。
cs.RO / 16 / 2609.17910
Map2Route: Benchmarking Compositional Language-Grounded Route Planning over Semantic Maps
Map2Route：基于语义地图的组合语言引导路径规划基准测试
Abstract
We introduce Map2Route, a human-curated benchmark for compositional language-grounded route planning over pre-built semantic maps. Map2Route contains 1,000 episodes across 40 scenes, where instructions use relational, comparative, and nested descriptions to identify route-relevant objects and regions, while specifying ordered must-pass regions, must-avoid requirements, five categories of soft preferences, and spatial and route-stage scopes, which is partially tested by existing works. Alongside Map2Route, we propose Grounding2Route, which combines executable code-as-grounding with verification-guided repair and scope-aware planning.Across seven representative adapted baselines, Grounding2Route substantially outperforms existing methods in all metrics. Despite these gains, a substantial gap to human demonstrations remains, highlighting the difficulty of Map2Route and the considerable headroom for future progress. Additional qualitative results and resources are available on https://anonymous.4open.science/w/Map2Route-F05F/.
Chinese Translation
我们介绍了Map2Route，这是一个人为策划的基准，用于基于预构建语义地图的组合语言引导路径规划。Map2Route包含1000个场景，涵盖40个场景，其中指令使用关系、比较和嵌套描述来识别与路径相关的对象和区域，同时指定有序的必经区域、必须避免的要求、五类软偏好以及空间和路径阶段范围，这些内容部分由现有工作进行了测试。与Map2Route一起，我们提出了Grounding2Route，它将可执行代码作为基础与验证引导的修复和范围感知规划相结合。在七个代表性的适应基准中，Grounding2Route在所有指标上显著优于现有方法。尽管取得了这些进展，但与人类示范之间仍存在显著差距，突显了Map2Route的难度以及未来进展的巨大潜力。更多定性结果和资源可在https://anonymous.4open.science/w/Map2Route-F05F/获取。
cs.RO / 17 / 2609.17929
Multi-Session Multimodal Underwater Mapping with Acoustic and Optical Imaging
基于声学和光学成像的多会话多模态水下制图
Abstract
Accurate seafloor mapping is essential for marine science, archaeology, and environmental monitoring. However, integrating data from different sensors, such as side-scan sonar and optical cameras, collected across separate survey sessions, remains challenging due to positioning drift and sensor offsets. This paper presents a multi-session, multimodal underwater mapping framework based on factor graph optimization. The method jointly optimizes vehicle trajectories, 3D landmark positions, sensor extrinsics, and per-session global alignment transformations. By combining rigid inter-session corrections with local trajectory deformations, it compensates for both inter-session offsets and intra-session distortions from accumulated navigation errors. The proposed methodology was validated on real-world datasets collected along the Catalan coast. Results show measurable improvements in map consistency over both unoptimized and rigid-alignment baselines across all metrics, including Pixel Accuracy and mean Intersection over Union. The method achieves a 3.4% improvement in pixel accuracy over the unoptimized baseline, corresponding to improved semantic labelling across approximately 14700 $\text{m}^2$ of mapped area. Qualitative results further show consistent co-registration between sonar and optical maps, even in the presence of significant trajectory distortions and inter-session misalignments. These findings demonstrate the potential of the proposed framework to generate coherent multimodal seafloor maps from heterogeneous underwater surveys.
Chinese Translation
准确的海底制图对于海洋科学、考古学和环境监测至关重要。然而，由于定位漂移和传感器偏差，整合来自不同传感器（如侧扫声纳和光学相机）在不同调查会话中收集的数据仍然具有挑战性。本文提出了一种基于因子图优化的多会话多模态水下制图框架。该方法联合优化车辆轨迹、三维地标位置、传感器外部参数以及每会话的全局对齐变换。通过结合刚性会话间校正与局部轨迹变形，该方法补偿了会话间偏差和由于累积导航误差导致的会话内失真。所提方法在沿加泰罗尼亚海岸收集的真实世界数据集上进行了验证。结果显示，在所有指标（包括像素准确度和平均交并比）上，相较于未优化和刚性对齐基线，地图一致性有了显著改善。该方法在像素准确度上比未优化基线提高了3.4%，对应于约14700平方米映射区域的语义标注改进。定性结果进一步表明，即使在显著轨迹失真和会话间错位的情况下，声纳和光学地图之间也保持了一致的共同注册。这些发现展示了所提框架在从异构水下调查中生成一致的多模态海底地图的潜力。
cs.RO / 18 / 2609.17946
Feedback-Modulated Harmonic Policies for Quadruped Locomotion
反馈调制的四足动物运动谐波策略
Abstract
Learned quadruped locomotion policies commonly map observations directly to joint-level actions, leaving the periodic structure of locomotion implicit in the policy. We investigate an alternative representation in which each joint trajectory is expressed as a command-conditioned Fourier series and modified online using feedback from the robot state. A context network generates the Fourier coefficients and the weights of a per-step feedback network, whose outputs adjust joint offsets, harmonic gains, frequency, and phase during execution. In simulation, we examine this explicit frequency structure alongside the hidden activations of an MLP policy that directly outputs joint targets. The harmonic waveforms change frequency and shape with commanded speed. Dynamic mode decomposition of selected MLP rollouts reveals dominant activation modes near the foot-height oscillation frequency and its second harmonic, showing periodic structure without an explicit Fourier generator. On a Unitree Go2, the simulation-trained harmonic controller records a provisional onboard-estimated peak speed of 3.67 meter per second and carries added loads up to 5.883 kilogram in separate trials.
Chinese Translation
学习的四足动物运动策略通常将观测直接映射到关节级别的动作，隐含了运动的周期性结构。我们研究了一种替代表示方法，其中每个关节轨迹被表达为命令条件的傅里叶级数，并通过来自机器人状态的反馈在线修改。上下文网络生成傅里叶系数和每步反馈网络的权重，其输出在执行过程中调整关节偏移、谐波增益、频率和相位。在仿真中，我们考察了这种显式频率结构，以及直接输出关节目标的多层感知器（MLP）策略的隐藏激活。谐波波形随着指令速度的变化而改变频率和形状。对选定的MLP回放的动态模式分解揭示了在足高度振荡频率及其二次谐波附近的主导激活模式，显示出周期性结构而没有显式的傅里叶生成器。在Unitree Go2上，经过仿真训练的谐波控制器记录了临时的机载估计峰值速度为每秒3.67米，并在单独的试验中承载了高达5.883千克的额外负载。
cs.RO / 19 / 2609.18016
Causal-History Test-Time Scaling for Failure Recovery in Autoregressive World-Action Models
自回归世界-动作模型中的因果历史测试时缩放用于故障恢复
Abstract
World-action models (WAMs) have emerged as a promising paradigm for robot manipulation by jointly modeling future visual dynamics and robot actions. However, existing WAMs are trained predominantly on successful trajectories, making them prone to failure when real-world execution diverges from the learned dynamics. This issue is amplified in autoregressive WAMs, where execution errors become part of the causal history and continue to influence subsequent predictions. To this end, we introduce \method{}, a training-free framework that reformulates failure recovery as \emph{test-time scaling over causal histories}. This formulation decomposes recovery into three coupled decisions: \emph{when} to revise the causal history, \emph{where} to recover a reliable history prefix, and \emph{which} history configuration best supports subsequent execution. Specifically, \method{} realizes these decisions through three stages: 1) \textbf{Progress-Aware Recovery Trigger} detects persistent non-progress and triggers recovery only when the current execution state permits intervention; 2) \textbf{History-Prefix Recovery} identifies the unreliable history suffix, retrieves a historical anchor matching the current physical state, and reconstructs the causal KV state from the retained prefix while conditioning on the latest real observation; and 3) \textbf{Hypothesis Verification} compares the future continuations induced by complete-history, recovered-prefix, and full-reset hypotheses, and commits the best-supported hypothesis. Experiments in both simulated and real-world manipulation settings demonstrate consistent improvements in task success, while ablations confirm the contribution of each recovery stage.
Chinese Translation
世界-动作模型（WAMs）作为一种有前景的机器人操控范式，通过联合建模未来的视觉动态和机器人动作而受到关注。然而，现有的WAMs主要在成功轨迹上进行训练，这使得它们在现实世界执行偏离学习动态时容易出现故障。这个问题在自回归WAMs中尤为突出，因为执行错误成为因果历史的一部分，并继续影响后续预测。为此，我们提出了 extit{方法}，这是一个无训练的框架，将故障恢复重新表述为 extit{基于因果历史的测试时缩放}。这一表述将恢复过程分解为三个相互关联的决策： extit{何时}修订因果历史， extit{何处}恢复可靠的历史前缀，以及 extit{哪个}历史配置最能支持后续执行。具体而言， extit{方法}通过三个阶段实现这些决策：1） extbf{进展感知恢复触发器}检测持续的非进展，并仅在当前执行状态允许干预时触发恢复；2） extbf{历史前缀恢复}识别不可靠的历史后缀，检索与当前物理状态匹配的历史锚点，并在条件化最新实际观察的情况下，从保留的前缀重建因果KV状态；3） extbf{假设验证}比较由完整历史、恢复前缀和完全重置假设引发的未来延续，并承诺最佳支持的假设。在模拟和真实世界的操控设置中的实验表明，任务成功率持续提高，而消融实验确认了每个恢复阶段的贡献。
cs.RO / 20 / 2609.18031
Online Multimodal Workload Assessment in Contact-Rich Physical Human-Robot Interaction
接触丰富的物理人机交互中的在线多模态工作负载评估
Abstract
Contact-rich physical human--robot interaction (pHRI) imposes time-varying demands associated with physical interaction, motor regulation, and physiological response, motivating continuous assessment of interaction workload. This paper presents an online multimodal assessment framework that integrates interaction wrench, planar tool-center-point (TCP) kinematics, and skin conductance level (SCL) into four interpretable workload-related factors. Their relative contributions are adjusted using path curvature to reflect changes in motion demand and task progression to account for gradual physiological variation over time. The framework was evaluated with 24 participants across 18 controlled combinations of temperature, acoustic noise, and illuminance under two admittance-control modes. Strict leave-one-subject-out (LOSO) evaluation used standardized pupil diameter ($\mathrm{PD}_z$) as an independent physiological reference and included comparisons with static variants and representative state-of-the-art learning-based baselines. The proposed framework achieves a cohort-mean $30\,\mathrm{s}$ block-wise Spearman correlation of $\rho_{30}=0.308$ with the physiological reference, with positive subject-level correspondence in 23 of 24 participants. Its overall performance is comparable to the state-of-the-art learning-based baseline. At the same time, our framework keeps the assessment process transparent through explicit workload-related factors and defined weighting rules, while outperforming the corresponding fixed-weight formulation. The framework also maintains consistent performance across the two tested admittance-control modes. These results support a transparent and interpretable approach to continuous interaction workload assessment in contact-rich pHRI.
Chinese Translation
接触丰富的物理人机交互（pHRI）带来了与物理交互、运动调节和生理反应相关的时变需求，这促使对交互工作负载进行持续评估。本文提出了一种在线多模态评估框架，该框架将交互扭矩、平面工具中心点（TCP）运动学和皮肤电导水平（SCL）整合为四个可解释的工作负载相关因素。通过路径曲率调整这些因素的相对贡献，以反映运动需求和任务进展中的变化，从而考虑到生理变化的渐进性。该框架在两种阻抗控制模式下对24名参与者进行了评估，涵盖了18种温度、声学噪声和照明强度的受控组合。严格的逐一留出法（LOSO）评估使用标准化的瞳孔直径（$ ext{PD}_z$）作为独立的生理参考，并与静态变体和代表性的最先进学习基线进行了比较。所提出的框架在与生理参考的块状Spearman相关性上实现了组均值$30 ext{s}$的$
ho_{30}=0.308$，在24名参与者中有23名的个体级别对应关系为正。其整体性能与最先进的基于学习的基线相当。同时，我们的框架通过明确的工作负载相关因素和定义的加权规则保持评估过程的透明性，同时优于相应的固定权重公式。该框架在两种测试的阻抗控制模式下也保持了一致的性能。这些结果支持了一种透明且可解释的接触丰富pHRI中持续交互工作负载评估的方法。
cs.RO / 21 / 2609.18050
Geometric Shortcuts for Complex Trunk Postures: Dual-Helicity Coupling Enables Low-Dimensional Control
复杂象鼻姿态的几何捷径：双螺旋耦合实现低维控制
Abstract
How do elephant trunks generate complex postures without relying solely on fine segmental activation? We propose that part of this complexity arises from a low-dimensional geometric shortcut: dual-helicity coupling between opposite-handed oblique muscles. In a simplified soft-robotic prototype, varying only two geometric parameters generates a broad library of elephant-like postures, suggesting a dual-layer control architecture with implications for continuum robot design and biological hypotheses.
Chinese Translation
大象的象鼻如何在不单靠精细的分段激活的情况下生成复杂的姿态？我们提出，这种复杂性部分源于一种低维几何捷径：对向手性斜肌之间的双螺旋耦合。在一个简化的软机器人原型中，仅通过改变两个几何参数就能生成广泛的类似大象的姿态，暗示了一种双层控制架构，这对连续机器人设计和生物学假说具有重要意义。
cs.RO / 22 / 2609.18051
CLASP: A Cluster-Level Autonomous Selective Picking Robot with a Soft Rolling-Band Gripper for Fresh-Market Blueberry Harvesting
CLASP：一种用于新鲜市场蓝莓采摘的集群级自主选择性采摘机器人，配备软滚带抓手
Abstract
Fresh-market blueberries require selective, gentle picking, which is labor-intensive and expensive. Over-the-row machine harvesters are fast but non-selective, bruising mixed-ripeness fruit and limiting yield to the processing market. Selective robotic harvesters typically target individual fruits rather than fruit clusters, which limits harvesting efficiency for small, densely clustered blueberries. This paper presents CLASP, a Cluster-Level Autonomous Selective Picking robot with a Soft Active Rolling-Band Gripper (SARB-Gripper). Two compliant bands envelop the cluster and roll against the fruit, drawing mature berries off in sequence, while closed-loop regulation of the pulling force keeps the applied load below the immature detachment threshold. A global-to-local perception pipeline pairs an eye-to-hand camera for global cluster detection and target selection with an eye-in-hand camera for local localization and cluster orientation estimation. Field measurements confirm a clear detachment-force separation between mature and immature fruit, and the SARB-Gripper reproduces a commanded pulling force to within \SI{3.7}{\percent}, enabling selective harvesting at the cluster level. In end-to-end field trials, CLASP autonomously grasped 23 of 25 presented clusters (\SI{92}{\percent}). With the component cost of approximately \$3326 per unit, CLASP offers a scalable approach to selective cluster-level harvesting for fresh-market blueberries.
Chinese Translation
新鲜市场蓝莓需要选择性、温和的采摘，这是一项劳动密集且成本高昂的工作。行间采摘机虽然速度快，但不具选择性，容易损伤成熟度不一的果实，并将产量限制在加工市场。选择性机器人采摘机通常针对单个果实，而非果实簇，这限制了对小型、密集簇生蓝莓的采摘效率。本文介绍了CLASP，一种集群级自主选择性采摘机器人，配备软主动滚带抓手（SARB-Gripper）。两个柔性带包裹住果实簇并与果实滚动，依次将成熟的浆果摘下，同时闭环调节拉力保持施加的负载低于未成熟果实的脱落阈值。全球到局部的感知管道将用于全球果实簇检测和目标选择的眼对手相机与用于局部定位和果实簇方向估计的手对眼相机相结合。田间测量确认成熟果实与未成熟果实之间存在明显的脱落力分离，SARB-Gripper能够在3.7%的误差范围内重现指令拉力，从而实现集群级的选择性采摘。在端到端的田间试验中，CLASP自主抓取了25个果实簇中的23个（92%）。每个组件成本约为3326美元，CLASP为新鲜市场蓝莓的选择性集群级采摘提供了一种可扩展的方法。
cs.RO / 23 / 2609.18070
An Efficient Algorithm for Minimum-Pressure Growth Planning of Vine Robots
一种高效的藤蔓机器人最小压力生长规划算法
Abstract
Vine robots navigate cluttered environments by extending from their tip. Although their ability to operate in such environments has been extensively demonstrated, little work has addressed growth planning, i.e., finding optimal growth paths. Moreover, existing planners do not account for the growth pressure necessary to follow a given path, which can cause the robot to burst when it is too high. In this paper, we address the problem of finding minimum-pressure paths for vine robots growing around polytopic obstacles. We propose an efficient algorithm that is guaranteed to find globally optimal solutions in 2D and approximate solutions in 3D, with an error that vanishes as a discretization parameter approaches zero. First, we derive a growth pressure equation for vine robots of arbitrary shape, which we use to show that there always exists a minimum-pressure path that is piecewise-linear and can bend only at specific points on the obstacles. We then leverage this observation to reduce the growth-planning problem to a shortest-path problem with time-dependent weights, which we efficiently solve using a modified Dijkstra's algorithm. We demonstrate the speed and scalability of our approach through numerical simulations. We also validate our algorithm with hardware experiments and provide an open-source and high-performance implementation in the Python package, VinePlanner: https://github.com/Ahsoka/VinePlanner.
Chinese Translation
藤蔓机器人通过从其尖端延伸来导航于杂乱的环境中。尽管它们在此类环境中操作的能力已经得到了广泛的验证，但关于生长规划，即寻找最佳生长路径的研究却相对较少。此外，现有的规划器并未考虑沿给定路径所需的生长压力，这可能导致机器人在压力过高时发生破裂。本文旨在解决藤蔓机器人在多面体障碍物周围生长时寻找最小压力路径的问题。我们提出了一种高效的算法，能够在二维空间中保证找到全局最优解，并在三维空间中找到近似解，且误差会随着离散化参数接近零而消失。首先，我们推导出一种适用于任意形状藤蔓机器人的生长压力方程，并利用该方程证明总是存在一条分段线性且仅能在障碍物特定点弯曲的最小压力路径。随后，我们利用这一观察将生长规划问题简化为一个具有时间依赖权重的最短路径问题，并使用改进的Dijkstra算法高效求解。我们通过数值仿真展示了我们方法的速度和可扩展性。同时，我们还通过硬件实验验证了我们的算法，并提供了一个开源的高性能实现，名为VinePlanner，链接：https://github.com/Ahsoka/VinePlanner。
cs.RO / 24 / 2609.18073
Characterizing Refraction-Induced Ranging Bias in Underwater Collaborative Localization
水下协作定位中的折射引起的测距偏差特征研究
Abstract
This work studies how refraction-induced bias on acoustic ranging affects multi-agent collaborative localization in a range of oceanographic conditions and spatial scales. While multi-agent range-aided navigation, which uses range measurements to either fixed infrastructure or other agents, is a promising solution to the challenges of large-scale underwater localization, its accuracy depends strongly on the quality of range measurements. Sound speed variability induces refraction (bending) of acoustic rays, yet, for algorithmic tractability, standard sensor fusion pipelines assume straight-line propagation. This refraction systematically biases range measurements to be longer than the straight-line assumption predicts. However, the effects of this bias on multi-agent collaborative localization on kilometer scales remains unexplored. We present a series of simulated experiments with several agents operating over kilometer scales. The simulation uses HYCOM reanalysis data to recreate realistic oceanographic conditions, ray tracing to generate refraction-informed ranges, and a centralized multi-agent factor graph estimator to quantify the resulting measurement bias on estimated trajectories. Preliminary results indicate that refraction-induced bias can induce significant degradation of estimated trajectories, particularly in regions with sharp sound-speed gradients. We also share the simulation environment to support further studies https://github.com/UMich-RobotExploration/manta-ray.
Chinese Translation
本研究探讨了折射引起的声学测距偏差如何影响在多种海洋条件和空间尺度下的多智能体协作定位。虽然多智能体范围辅助导航利用与固定基础设施或其他智能体的测距来解决大规模水下定位的挑战，但其准确性在很大程度上依赖于测距质量。声速的变化导致声波的折射（弯曲），然而，为了算法的可处理性，标准传感器融合流程假设声波沿直线传播。这种折射系统性地使测距结果偏长，超出了直线假设的预测。然而，这种偏差对千米尺度的多智能体协作定位的影响尚未得到探讨。我们呈现了一系列模拟实验，多个智能体在千米尺度上操作。该模拟使用HYCOM重分析数据重现现实的海洋条件，利用光线追踪生成考虑折射的测距，并使用集中式多智能体因子图估计器量化对估计轨迹的测量偏差。初步结果表明，折射引起的偏差可能显著降低估计轨迹的准确性，特别是在声速梯度陡峭的区域。我们还分享了模拟环境，以支持进一步的研究 https://github.com/UMich-RobotExploration/manta-ray。
cs.RO / 25 / 2609.18084
Not All Layers Need Tuning: Diagnosing and Directing Adaptation in Vision-Language-Action Models
并非所有层都需要调优：在视觉-语言-动作模型中诊断和指导适应
Abstract
Fine-tuning a Vision-Language-Action (VLA) model for a new deployment environment is expensive, yet most methods apply uniform-capacity adapters to every network region as if every region requires equal adjustment. This paper tests that assumption on five architecturally diverse VLAs (OpenVLA-OFT, $\pi_0$, SmolVLA, DTP, Octo; 93M-7B parameters). Measuring per-region adaptation cost as normalized parameter displacement under region-isolated fine-tuning reveals an adaptation spectrum in which appearance shifts concentrate cost in the vision encoder, instruction shifts in the language backbone, and novel-object shifts in the vision encoder together with the action head, across all five architectures. To exploit this structure, we introduce a pipeline that observes, diagnoses, allocates, and adapts. From ten unlabeled target observations and without fine-tuning, the diagnostic estimates per-region cost by combining reference-free gradient and Monte Carlo Dropout signals with a Centered Kernel Alignment score against a cached source reference; the allocator converts the estimates into variable-rank LoRA adapters under a parameter budget and freezes well-calibrated regions; and standard LoRA fine-tuning trains the resulting adapters. The diagnostic ranks regions within each deployment at a median Spearman of 0.91, and the allocation matches or exceeds uniform LoRA at every budget we tested on LIBERO and CALVIN. On a physical xArm-7, the pipeline matches full fine-tuning under an instruction-wording shift with 0.04% of its trainable parameters, and on five held-out scenes evaluated without retraining it leads every baseline, with 11-23 successes of 30 rollouts against 8-18 for the strongest parameter-efficient baseline at equal or larger budgets and 2-11 for full fine-tuning. These results suggest that adaptation cost in VLAs is structured enough to measure before fine-tuning begins.
Chinese Translation
为新的部署环境微调视觉-语言-动作（VLA）模型是昂贵的，然而大多数方法对每个网络区域应用统一容量的适配器，仿佛每个区域都需要相等的调整。本文在五种架构各异的VLA（OpenVLA-OFT、$ ext{π}_0$、SmolVLA、DTP、Octo；参数范围93M-7B）上测试了这一假设。通过测量区域隔离微调下的标准化参数位移来评估每个区域的适应成本，揭示了一个适应光谱，其中外观变化将成本集中在视觉编码器上，指令变化集中在语言主干上，而新物体变化则同时集中在视觉编码器和动作头上，适用于所有五种架构。为了利用这一结构，我们引入了一个观察、诊断、分配和适应的流程。从十个未标记的目标观察中，在不进行微调的情况下，诊断通过结合无参考梯度和蒙特卡洛丢弃信号与针对缓存源参考的中心核对齐得分来估计每个区域的成本；分配器在参数预算下将估计值转换为可变秩的LoRA适配器，并冻结校准良好的区域；标准的LoRA微调则训练生成的适配器。诊断在每个部署中对区域的排名中位斯皮尔曼相关系数为0.91，分配在我们在LIBERO和CALVIN上测试的每个预算下匹配或超过了均匀LoRA。在物理xArm-7上，该流程在指令措辞变化下以0.04%的可训练参数匹配完全微调，并且在五个未重训练的保留场景中，超越了每个基线，在30次回合中取得11-23次成功，而最强的参数高效基线在相等或更大预算下为8-18次成功，完全微调为2-11次成功。这些结果表明，VLA中的适应成本结构足够明确，可以在微调开始之前进行测量。
cs.RO / 26 / 2609.18092
ReRadar: Robust Radar Global Localization via Rotation-Equivariant Descriptor Learning
ReRadar：通过旋转等变描述符学习实现稳健的雷达全球定位
Abstract
Global localization with scanning millimeter-wave radar remains challenging because place-recognition descriptors often discard spatial structure needed for accurate pose retrieval. We present ReRadar, a radar global localization pipeline that extracts rotation-equivariant intermediate features using steerable convolutional neural networks, forms rotation-invariant descriptors through group pooling and NetVLAD aggregation, and combines descriptor retrieval with landmark-based matching to estimate the robot's three-degree-of-freedom (3-DoF) pose. Across fixed database-query evaluations, ReRadar with target-dataset adaptation achieves 99.37% Recall@1 on OORD Bellmouth, 91.44% Recall@1 with 80.99% F1_max on Mulran DCC01, and 99.38% Recall@1 on falling-snow Boreas sequence. Without target-dataset data, the cross-dataset model reaches 98.07% Recall@1 on OORD, performing comparably to the evaluated state-of-the-art methods.
Chinese Translation
使用扫描毫米波雷达进行全球定位仍然具有挑战性，因为地点识别描述符通常会忽略准确姿态检索所需的空间结构。我们提出了ReRadar，一个雷达全球定位管道，通过可调卷积神经网络提取旋转等变的中间特征，通过组池化和NetVLAD聚合形成旋转不变描述符，并将描述符检索与基于地标的匹配相结合，以估计机器人的三自由度（3-DoF）姿态。在固定数据库查询评估中，经过目标数据集适应的ReRadar在OORD Bellmouth上达到了99.37%的Recall@1，在Mulran DCC01上达到了91.44%的Recall@1和80.99%的F1_max，在降雪Boreas序列上达到了99.38%的Recall@1。在没有目标数据集数据的情况下，跨数据集模型在OORD上达到了98.07%的Recall@1，其表现与评估的最先进方法相当。
cs.RO / 27 / 2609.18100
Beyond Pixel Similarity: Task-Aware Evaluation of GAN-Based Synthetic Sonar Data for Robotic Perception
超越像素相似性：基于任务的GAN合成声纳数据评估在机器人感知中的应用
Abstract
Synthetic data can reduce the cost of collecting and annotating training data for robotic perception, but generating sensor observations that preserve the characteristics relevant to downstream perception remains challenging, particularly for sonar imagery. In this work, we investigate whether conventional image-fidelity metrics adequately reflect the downstream perception performance of GAN-generated synthetic sonar data. We employ a Pix2Pix conditional generative adversarial network with four discriminator configurations characterized by different receptive fields: PixelGAN, PatchGAN-16, PatchGAN-70, and ImageGAN. The models are trained using sonar imagery from two datasets and evaluated using conventional image-fidelity metrics, including Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), and Mean Squared Error (MSE). To complement these pixel-level measures with task-oriented evaluation, YOLOX-S, YOLOX-L, and Faster R-CNN detectors are trained exclusively on real sonar imagery and subsequently evaluated on the GAN-generated images using identical test samples and annotations across all discriminator configurations. The results reveal a discrepancy between image-fidelity and downstream object-detection performance: the configuration achieving the best SSIM, PSNR, and MSE does not consistently yield the best detection performance. In particular, PatchGAN configurations achieve strong downstream detection results despite not achieving the highest pixel-level similarity scores. These findings suggest, for the datasets and models considered, pixel-level image-fidelity metrics alone may not consistently capture the task-relevant realism of synthetic sonar observations and motivate the use of task-aware evaluation for synthetic sensor data intended for robotic perception.
Chinese Translation
合成数据可以降低收集和标注机器人感知训练数据的成本，但生成保留与下游感知相关特征的传感器观测仍然具有挑战性，尤其是在声纳图像方面。在本研究中，我们探讨了传统图像保真度指标是否充分反映了GAN生成的合成声纳数据的下游感知性能。我们采用了Pix2Pix条件生成对抗网络，并使用四种不同感受野的判别器配置进行训练：PixelGAN、PatchGAN-16、PatchGAN-70和ImageGAN。这些模型使用来自两个数据集的声纳图像进行训练，并使用传统的图像保真度指标进行评估，包括结构相似性指数（SSIM）、峰值信噪比（PSNR）和均方误差（MSE）。为了将这些像素级度量与任务导向评估相结合，我们仅在真实声纳图像上训练了YOLOX-S、YOLOX-L和Faster R-CNN检测器，并随后在GAN生成的图像上使用相同的测试样本和注释进行评估。结果显示图像保真度与下游目标检测性能之间存在差异：实现最佳SSIM、PSNR和MSE的配置并不总是产生最佳检测性能。特别是，尽管PatchGAN配置未达到最高的像素级相似性分数，但仍实现了强大的下游检测结果。这些发现表明，对于所考虑的数据集和模型，单靠像素级图像保真度指标可能无法始终捕捉合成声纳观测的任务相关现实性，并促使我们对用于机器人感知的合成传感器数据采用任务导向的评估方法。
cs.RO / 28 / 2609.18108
Technical Report: One-Step Drifting Action Heads for GR00T N1.7
技术报告：GR00T N1.7的一步漂移动作头
Abstract
One-step action generation can substantially reduce the inference cost of vision-language-action (VLA) policies, but its effect on closed-loop task success remains an open question. This technical report studies a GR00T N1.7 variant in which the iterative diffusion-transformer action head is replaced by a one-step drifting action head, together with an overlap-conditioned extension for asynchronous chunk replacement. All multi-seed drifting runs were trained on two NVIDIA A800 GPUs. On LIBERO, the action head reduces the mean model-forward time of the action head from approximately $45.3\,\mathrm{ms}$ to $5.0\,\mathrm{ms}$, while the measured backbone-plus-head time falls from approximately $70.0\,\mathrm{ms}$ to $30.6\,\mathrm{ms}$. However, this speedup is accompanied by a systematic reduction in task success. Across three drifting seeds, success is $64.0\pm4.0\%$ on LIBERO-Spatial, $52.0\pm1.0\%$ on LIBERO-Goal, and $26.0\pm2.6\%$ on LIBERO-Long. The low seed variance indicates that the degradation is not explained by random initialization alone. We report the result as a speed--success trade-off rather than an overall improvement, and discuss likely contributing factors including deterministic one-step mode averaging, batch-dependent geometry estimation, long open-loop chunk execution, and the fact that synchronous LIBERO evaluation does not exercise the asynchronous overlap path.
Chinese Translation
一步动作生成可以显著降低视觉-语言-动作（VLA）策略的推理成本，但其对闭环任务成功的影响仍然是一个未解之谜。本技术报告研究了一种GR00T N1.7变体，其中迭代扩散变换器动作头被一步漂移动作头替代，并结合了用于异步块替换的重叠条件扩展。所有多种子漂移运行均在两台NVIDIA A800 GPU上训练。在LIBERO上，动作头将动作头的平均模型前向时间从约$45.3 \, ext{ms}$减少到$5.0 \, ext{ms}$，而测得的主干加头时间从约$70.0 \, ext{ms}$降至$30.6 \, ext{ms}$。然而，这一加速伴随着任务成功率的系统性下降。在三个漂移种子中，LIBERO-Spatial的成功率为$64.0 \, ext{±} \, 4.0\\%$，LIBERO-Goal为$52.0 \, ext{±} \, 1.0\\%$，LIBERO-Long为$26.0 \, ext{±} \, 2.6\\%$。低种子方差表明，降级并非仅由随机初始化所致。我们将结果报告为速度与成功的权衡，而非整体改进，并讨论了可能的影响因素，包括确定性的一步模式平均、批依赖几何估计、长时间的开放环块执行，以及同步LIBERO评估未能利用异步重叠路径的事实。
cs.RO / 29 / 2609.18111
A Comprehensive Review of Generative Physical Artificial Intelligence
生成物理人工智能的综合评述
Abstract
The integration of large-scale foundation models with physical embodiments has led to significant advancements in robotics known as Generative Physical Artificial Intelligence (GPAI). These agentic AI systems autonomously perceive, reason, and act in complex real-world situations. This survey comprehensively analyzes GPAI systems, focusing on their architectural foundations, current applications, and key limitations. We introduce a taxonomy of five distinct approaches: Robot Foundation Models (RFMs) for cross-platform skill transfer; Vision-Language Action (VLA) models for end-to-end multi-modal perception and control; Large Behavior Models (LBMs) for human-like movement generation; Diffusion Policy Models (DPMs) for diffusion model-based temporally coherent action generation; and World Foundation Models (WFMs) for physics-compliant simulation and data generation. We examine how these approaches complement each other: WFMs generate training data for VLAs and DPMs, RFMs enable cross-platform deployment of learned policies, while LBMs provide motion priors for natural behavior. Through examples across autonomous vehicles, industrial automation, healthcare robotics, and humanoid systems, we identify significant performance improvements and summarize promising research directions in data-efficient learning, sim-to-real transfer, edge-compatible architectures, and safety frameworks. These insights advance embodied AI for IoT-connected environments where intelligent agents interact with networked sensors, actuators, and edge devices.
Chinese Translation
大规模基础模型与物理实体的结合推动了机器人技术的显著进展，这被称为生成物理人工智能（Generative Physical Artificial Intelligence, GPAI）。这些自主智能系统能够在复杂的现实世界情境中感知、推理和行动。本文全面分析了GPAI系统，重点关注其架构基础、当前应用及主要局限性。我们提出了五种不同方法的分类：跨平台技能转移的机器人基础模型（Robot Foundation Models, RFMs）；用于端到端多模态感知与控制的视觉-语言行动（Vision-Language Action, VLA）模型；用于人类般运动生成的大行为模型（Large Behavior Models, LBMs）；基于扩散模型的时间一致性动作生成的扩散策略模型（Diffusion Policy Models, DPMs）；以及用于符合物理规律的仿真和数据生成的世界基础模型（World Foundation Models, WFMs）。我们考察了这些方法如何相辅相成：WFMs为VLAs和DPMs生成训练数据，RFMs使学习到的策略能够跨平台部署，而LBMs则为自然行为提供运动先验。通过在自主车辆、工业自动化、医疗机器人和类人系统中的实例，我们识别出显著的性能提升，并总结了在数据高效学习、仿真到现实转移、边缘兼容架构和安全框架等领域的有前景的研究方向。这些见解推动了物联网连接环境中具身人工智能的发展，在这些环境中，智能代理与网络传感器、执行器和边缘设备进行交互。
cs.RO / 30 / 2609.18117
OpenDexGrasp: Open-vocabulary Task-Oriented Dexterous Grasping
OpenDexGrasp：开放词汇任务导向的灵巧抓取
Abstract
Dexterous grasp synthesis has advanced rapidly in generating stable and physically plausible hand poses, but real-world manipulation requires grasps that preserve the function implied by the task. We study open-vocabulary task-oriented dexterous grasp generation, where a robot must infer functional intent from free-form language, ground it in multi-view visual observations and object geometry, and generate an executable high-degree-of-freedom grasp. We present OpenDexGrasp, a unified data and generative modeling framework for this setting. OpenDexVerse provides dual-source supervision organized by the Coverage-to-Alignment (C2A) Recipe: OpenDex-Scale offers large-scale semantic and geometric coverage through automatic grasp synthesis and vision-language annotation, while OpenDex-Align supplies high-quality embodied alignment through human teleoperation and category-level transfer. OpenDexGrasp learns a shared perception-action latent representation that couples open-vocabulary vision-language context with dexterous action generation. Affordance grounding and grasp generation provide complementary supervision over this latent space, enabling direct generation of task-consistent dexterous grasps without a separate affordance-to-pose inference stage. Extensive simulation and real-robot experiments demonstrate improved functional alignment, physical feasibility, generalization to unseen categories, and real-world execution success. Additional details and videos are available at https://opendexgrasp.github.io/.
Chinese Translation
灵巧抓取合成在生成稳定且物理上合理的手势方面取得了快速进展，但现实世界中的操作需要保持任务所隐含的功能的抓取方式。我们研究开放词汇任务导向的灵巧抓取生成，其中机器人必须从自由形式的语言中推断功能意图，将其与多视角视觉观察和物体几何信息相结合，并生成可执行的高自由度抓取。我们提出了OpenDexGrasp，这是一个针对该场景的统一数据和生成建模框架。OpenDexVerse提供了由覆盖到对齐（Coverage-to-Alignment，C2A）配方组织的双源监督：OpenDex-Scale通过自动抓取合成和视觉-语言注释提供大规模的语义和几何覆盖，而OpenDex-Align通过人类遥操作和类别级转移提供高质量的体现对齐。OpenDexGrasp学习了一个共享的感知-动作潜在表示，将开放词汇的视觉-语言上下文与灵巧动作生成相结合。功能基础的对齐和抓取生成在这个潜在空间中提供互补的监督，使得可以直接生成与任务一致的灵巧抓取，而无需单独的功能到姿态推断阶段。大量的仿真和真实机器人实验表明，功能对齐、物理可行性、对未见类别的泛化以及现实世界执行的成功都有所改善。更多详细信息和视频可在 https://opendexgrasp.github.io/ 获取。
cs.RO / 31 / 2609.18119
Fetch My Beer: Synthetic-to-real Hierarchical Policy for Smooth Pick-and-place
取我的啤酒：用于平滑拾取和放置的合成到真实层次策略
Abstract
Many real-world robotic applications require dynamically sensitive manipulation, where success depends not only on reaching a target state but on maintaining stable object dynamics throughout execution. We study the stable transport of liquid-filled containers, where a robot must move objects to target locations while suppressing sloshing and preventing spillage. Unlike conventional pick-and-place, this task imposes stringent requirements on motion smoothness and trajectory-level stability, exposing clear limitations in existing systems. Specifically, fluid simulation remains too costly for online reinforcement learning; human teleoperation introduces unintended accelerations that induce sloshing during imitation learning; and current policy pipelines optimize for task completion rather than dynamic stability. We propose a synthetic-to-real framework coupling physically validated data generation with a hierarchical, diffusion-based controller. The scalable data pipeline synthesizes grasps, filters unstable poses via a vision-language model, and validates transport trajectories through fluid simulation. The policy is organized with a high-level module that translates language and visual observations into SE(3) control targets, and a latent diffusion controller that first plans efficiently in a compact latent space and then decodes dense action chunks, enabling the high control frequency needed for smooth and stable motion. Extensive experiments show our system outperforms state-of-the-art manipulation policies in transport smoothness and dynamic stability. Our project page: https://fetch-my-beer.github.io/
Chinese Translation
许多现实世界的机器人应用需要动态敏感的操作，其成功不仅依赖于达到目标状态，还依赖于在执行过程中保持稳定的物体动态。我们研究了液体容器的稳定运输，其中机器人必须在抑制晃动和防止溢出的情况下将物体移动到目标位置。与传统的拾取和放置任务不同，这项任务对运动平滑性和轨迹级稳定性提出了严格的要求，暴露了现有系统的明显局限性。具体而言，流体模拟对于在线强化学习来说成本过高；人类遥操作在模仿学习过程中引入了意外的加速度，导致晃动；而当前的策略管道优化的是任务完成，而非动态稳定性。我们提出了一种合成到真实的框架，将物理验证的数据生成与层次化的基于扩散的控制器相结合。该可扩展的数据管道合成抓取动作，通过视觉-语言模型过滤不稳定的姿态，并通过流体模拟验证运输轨迹。该策略由一个高层模块组织，该模块将语言和视觉观察转化为 SE(3) 控制目标，以及一个潜在扩散控制器，该控制器首先在紧凑的潜在空间中高效规划，然后解码密集的动作块，从而实现平滑和稳定运动所需的高控制频率。大量实验表明，我们的系统在运输平滑性和动态稳定性方面优于最先进的操作策略。我们的项目页面：https://fetch-my-beer.github.io/
cs.RO / 32 / 2609.18153
Prior Evolution and Task Alignment for Aerial Grasping
空中抓取的先验演化与任务对齐
Abstract
Aerial grasping is a remarkable capability exhibited by predatory birds, allowing them to capture prey through highly coordinated maneuvers in flight. Inspired by this capability, researchers have developed various formulations to reproduce such maneuvers through trajectory optimization. However, two limitations remain in practice. First, the resulting optimization problem is highly nonconvex and sensitive to initialization, making high-quality solutions difficult to obtain under a limited computational budget. Second, prescribed numerical objectives are human-designed abstractions that describe successful grasping through a limited set of mathematically tractable quantities and may not fully capture what determines task success. We investigate how learning can address these limitations within an analytical planner. Accordingly, a trajectory prior is first learned from optimized motions and then evolved through a CEM-based process that evaluates sampled initializations with the deployed optimizer and retains favorable ones as new supervision. An Execution-Aware Critic learns from contact, lift, and completion outcomes to assess whether the optimized trajectories are likely to succeed in physical execution. Its frozen energy can further serve as a differentiable grasping cost, allowing execution data to directly shape trajectory generation. Simulation and real-world experiments demonstrate improved optimization reliability, trajectory consistency, and grasping performance.
Chinese Translation
空中抓取是捕食性鸟类展现出的显著能力，使它们能够通过在飞行中的高度协调的动作捕捉猎物。受到这一能力的启发，研究人员开发了多种公式来通过轨迹优化再现这种动作。然而，实践中仍然存在两个限制。首先，所产生的优化问题高度非凸且对初始化敏感，使得在有限的计算预算下难以获得高质量的解决方案。其次，规定的数值目标是人类设计的抽象，描述了通过有限的一组数学可处理量成功抓取的情况，但可能无法完全捕捉决定任务成功的因素。我们研究了学习如何在分析规划器中解决这些限制。因此，首先从优化运动中学习轨迹先验，然后通过基于CEM（协同进化策略）的过程进行演化，该过程评估使用已部署优化器的采样初始化，并将有利的初始化保留为新的监督。执行感知评论者从接触、提升和完成结果中学习，以评估优化轨迹在物理执行中成功的可能性。其冻结能量还可以作为可微分的抓取成本，使执行数据能够直接影响轨迹生成。仿真和现实世界实验表明，优化的可靠性、轨迹的一致性和抓取性能得到了改善。
cs.RO / 33 / 2609.18164
Energy-Regularized Imitation Learning for Force- and Work-Aware Robotic Manipulation
能量正则化模仿学习用于力与功感知的机器人操作
Abstract
This paper studies energy-aware manipulation as a physically grounded learning problem. We define a joint-space mechanical-work proxy from joint torque and angular displacement, and train a differentiable energy predictor that estimates this work from robot states and actions. The predictor converts a non-differentiable simulator-side physical quantity into a differentiable regularizer for fine-tuning a pretrained manipulation policy. We instantiate the framework with RVT-2 on RLBench and evaluate 12 manipulation tasks involving object contact, articulated motion, placement, pushing, and sweeping. The proposed fine-tuning reduces the average mechanical work from 208.8J to 204.4J (i.e., 2.1% reduction), while the mean task success rate also increases slightly from 86.2% to 86.9%. These results show that work-aware policy optimization can suppress physically inefficient motion without requiring an explicit differentiable dynamics model.
Chinese Translation
本文将能量感知操作研究作为一个基于物理的学习问题。我们从关节扭矩和角位移定义了一个关节空间机械功的代理，并训练了一个可微分的能量预测器，该预测器根据机器人状态和动作估计这一功。该预测器将非可微分的模拟器侧物理量转换为可微分的正则化项，以便对预训练的操作策略进行微调。我们在 RLBench 上使用 RVT-2 实现了该框架，并评估了 12 个涉及物体接触、关节运动、放置、推动和扫动的操作任务。所提出的微调方法将平均机械功从 208.8J 降低到 204.4J（即减少 2.1%），同时平均任务成功率也略微从 86.2% 提升至 86.9%。这些结果表明，感知功的策略优化可以抑制物理上低效的运动，而无需显式的可微分动力学模型。
cs.RO / 34 / 2609.18165
LUMO: Designing Luminous Contact Morphology for Repeatable Whole-Finger Contact Observation
LUMO：设计发光接触形态以实现可重复的全手指接触观察
Abstract
A low-impedance robot finger reports through joint torque how strongly it is loaded, but the same torque can arise from a small force near the fingertip or a large force near the joint. Resolving the force therefore requires knowing where along the finger contact occurred. LUMO makes that location externally observable. Embedded LEDs illuminate a compliant silicone pad, and contact deforms the pad so that light emerging from the finger's side changes in a pattern set by where the load acts. Because the same structure also carries the contact load, we optimize its cross-section, including the pad profile, rigid carrier, and lateral void, for two behaviors at once. Mechanically, the pad conforms under low preload while the carrier increasingly restricts further deformation as load rises. Optically, different contact locations produce separated responses on the finger's side. The search uses rigid--soft contact simulation, ray tracing, and multi-objective Bayesian optimization. Across two silicones, six contact locations, and 10- and 30-mm spherical indenters, the optimized morphologies improve neighboring-location separation relative to variation from re-establishing contact by \(15\)--\(59\%\). Estimating contact location from the optical response using the known LED spacing and combining it with joint torque gives \(1.44~\mathrm{N}\) normal-force MAE over 931 samples. In a two-finger hand, localized side responses appear on several links simultaneously during grasps.
Chinese Translation
一种低阻抗机器人手指通过关节扭矩报告其负载强度，但相同的扭矩可以由靠近指尖的小力或靠近关节的大力产生。因此，解决力的问题需要知道接触发生在手指的哪个位置。LUMO使这一位置在外部可观察。嵌入式LED照亮一个柔性硅胶垫，接触导致垫子变形，从而使从手指侧面发出的光以负载作用位置设定的模式发生变化。由于同一结构还承载接触负载，我们优化其横截面，包括垫子轮廓、刚性载体和侧向空隙，以同时实现两种行为。在机械方面，垫子在低预载下适应，而载体在负载增加时逐渐限制进一步变形。在光学方面，不同的接触位置在手指侧面产生分离的响应。该搜索使用刚性-柔性接触模拟、光线追踪和多目标贝叶斯优化。在两种硅胶、六个接触位置以及10毫米和30毫米的球形压头下，优化后的形态相较于重新建立接触的变化提高了相邻位置的分离度，提升幅度为15%至59%。通过已知的LED间距估计光学响应中的接触位置，并将其与关节扭矩结合，得到了931个样本中的1.44 N的法向力平均绝对误差（MAE）。在一个双指手中，局部侧响应在抓取过程中同时出现在多个连接部位。
cs.RO / 35 / 2609.18167
Characterizing Replay Retention Under Dynamics Shift in Model-Based Reinforcement Learning
在模型基础的强化学习中表征动态变化下的重放保留
Abstract
Adapting to changes in robot dynamics requires learning from new data without discarding experience that may still be useful. In continual model-based reinforcement learning (RL), replay collected before a dynamics change can slow adaptation, while removing it unnecessarily reduces available training data and can be especially costly if earlier dynamics return. We study when recent transitions are preferable to the full replay history. Two quantities characterize this trade-off: change magnitude and age-staleness area under the curve (AUC), measuring how well transition age separates stale from fresh data. Forgetting stale data helps after large permanent shifts but hurts when dynamics recur and older data becomes useful again. Choosing a replay strategy therefore depends on predicting when older data will help or hurt. We test these effects across two locomotion morphologies, two model-based RL algorithms, and Real-World RL benchmark perturbations. Because ground-truth staleness labels are unavailable on deployed robots, we evaluate whether an estimator built from interaction data can still provide the quantities needed to choose a replay strategy after permanent changes. Our results show that replay retention depends on change magnitude and on how the dynamics evolve.
Chinese Translation
适应机器人动态变化需要从新数据中学习，而不丢弃可能仍然有用的经验。在持续的模型基础强化学习（RL）中，动态变化前收集的重放数据可能会减缓适应速度，而去除这些数据则会不必要地减少可用的训练数据，尤其是在早期动态再次出现时，这种损失可能是非常昂贵的。我们研究了在何种情况下，最近的过渡比完整的重放历史更为可取。有两个量度表征了这种权衡：变化幅度和年龄-过时性曲线下的面积（AUC），用于衡量过渡年龄在多大程度上将过时数据与新鲜数据区分开。遗忘过时数据在经历大规模永久性变化后有助于适应，但在动态复发且旧数据再次变得有用时则会造成损害。因此，选择重放策略取决于预测旧数据何时会有帮助或造成伤害。我们在两种运动形态、两种模型基础的RL算法以及现实世界RL基准扰动中测试了这些影响。由于在部署的机器人上无法获得真实的过时性标签，我们评估了是否可以从交互数据构建的估计器仍然能够提供选择重放策略所需的量度，尤其是在永久性变化之后。我们的结果表明，重放保留依赖于变化幅度以及动态演变的方式。
cs.RO / 36 / 2609.18169
Approximating High Dimensional Self-Motion Manifolds via Deep Generative Models
通过深度生成模型逼近高维自运动流形
Abstract
Self-motion manifold (SMM) characterizes the geometric structure of the infinite inverse kinematic solutions set of a redundant manipulator at a fixed end-effector pose, and its efficient recovery underpins feasible and global optimal motion planning. Existing methods such as null-space continuation and learning-based methods are formulated around the assumption that an SMM is a curve, and do not extend to higher redundancy orders. We instead adopt a probabilistic view: SMMs are the support of the conditional posterior over configurations given a target pose, so that recovering it reduces to sampling from a learned distribution and separating its disjoint components by clustering. The formulation is independent of the manifold dimension and requires no architectural change as the redundancy order grows. In this work, we demonstrate that our method can approximate 1-D SMMs with performance comparable to the latest null-space continuation and learning-based approach, and that it is the first method capable of approximating highly redundant 4-D SMMs in a 7R manipulator for position tasks. Project website: \href{https://github.com/accuracy-maker/high-dimenstional-self-motion-manifold-approximation}{https://github.com/accuracy-maker/high-dimenstional-self-motion-manifold-approximation}
Chinese Translation
自运动流形（Self-motion manifold, SMM）描述了在固定末端执行器姿态下冗余操纵器无限逆运动学解集的几何结构，其高效恢复是可行且全局最优运动规划的基础。现有方法如零空间延续（null-space continuation）和基于学习的方法均假设SMM为曲线，无法扩展到更高的冗余阶数。我们则采用概率视角：SMM是给定目标姿态下配置的条件后验分布的支撑，因此恢复SMM归结为从学习到的分布中采样，并通过聚类分离其不相交的组件。该公式与流形维度无关，并且随着冗余阶数的增加不需要架构上的变化。在本研究中，我们展示了我们的方法能够以与最新的零空间延续和基于学习的方法相当的性能逼近一维SMM，并且这是第一个能够在7自由度操纵器中逼近高度冗余的四维SMM以完成位置任务的方法。项目网站： [https://github.com/accuracy-maker/high-dimenstional-self-motion-manifold-approximation](https://github.com/accuracy-maker/high-dimenstional-self-motion-manifold-approximation)
cs.RO / 37 / 2609.18174
TacBPM: A Tactile-conditioned Behavior Prior Model for Dexterous Reorientation
TacBPM：一种用于灵巧重定位的触觉条件行为先验模型
Abstract
Dexterous in-hand manipulation requires policies that coordinate high-DoF hand joints through intermittent, contact-rich interaction. Beyond target-orientation tracking, such policies must discover finger gaits that preserve object stability while adapting to geometry, anisotropy, pose, contact, and sensing changes. We propose \method, a tactile-conditioned behavior prior model for dexterous reorientation. \method distills multi-scale sphere specialists into a latent controller and lets downstream policies reuse the fixed tactile prior through residual latent actions, reducing renewed exploration from raw joint commands. The prior conditions on tactile-proprioceptive history so latent behavior reflects the current hand-object interaction. We evaluate arbitrary-pose transfer across anisotropic objects, commanded-axis rotation, and an arm-hand Grasp-to-AnyPose task in which the robot must grasp, lift, transport, and reach goal poses for novel tool geometries and generalized placements. Extensive experiments demonstrate that the proposed method accelerates training and enables stable policies where matched raw-action PPO remains near failure, with successful sim-to-real transfer in in-hand and arm-hand tasks.
Chinese Translation
灵巧的手内操作需要通过间歇性、接触丰富的交互来协调高自由度的手关节。除了目标方向跟踪，这些策略还必须发现能够在适应几何形状、各向异性、姿态、接触和感知变化的同时保持物体稳定性的手指步态。我们提出了 extit{TacBPM}，一种用于灵巧重定位的触觉条件行为先验模型。 extit{TacBPM}将多尺度球体专家提炼为潜在控制器，并允许下游策略通过残差潜在动作重用固定的触觉先验，从而减少从原始关节命令进行重新探索的需求。该先验基于触觉-本体感知历史进行条件化，因此潜在行为反映当前的手-物体交互。我们评估了在各向异性物体上的任意姿态转移、指令轴旋转，以及一个手臂-手的抓取到任意姿态任务，其中机器人必须抓取、提升、运输并达到新工具几何形状和广义放置的目标姿态。大量实验表明，所提出的方法加速了训练，并使得稳定策略得以实现，而匹配的原始动作PPO则接近失败，在手内和手臂-手任务中成功实现了从仿真到现实的转移。
cs.RO / 38 / 2609.18191
OmniRisk: Omnidirectional Trajectory-Risk Learning for Agile Quadrotor Dynamic Avoidance
OmniRisk：用于灵活四旋翼动态避障的全向轨迹风险学习
Abstract
Agile quadrotor avoidance of fast-moving obstacles requires anticipating collisions and selecting feasible maneuvers within short reaction windows. Reliable predictive avoidance remains challenging because sparse range observations do not directly reveal obstacle motion, while online trajectory optimizers either scale poorly with obstacle count or remain efficient at the expense of reliability in dense, high-speed encounters. We present OmniRisk, an omnidirectional planning framework that learns trajectory-level risk offline for efficient onboard evasion. A fixed-dimensional tensor combines LiDAR range panoramas, dynamic masks, and Cartesian surface velocities to represent geometry and motion jointly. We formulate an asymmetric risk field aligned with obstacle velocity that emphasizes approaching interactions and attenuates receding ones. Accumulating this risk along predicted relative trajectories provides dense supervision and discourages unnecessary hesitation after obstacles pass. A dual-branch circular convolutional network predicts terminal boundary states and dynamic risks for candidate primitives over an omnidirectional anchor lattice in a single forward pass, followed by selection and closed-form reconstruction of the selected candidate primitive. This formulation removes online risk accumulation along trajectories and makes risk-inference cost independent of obstacle count. OmniRisk enables efficient onboard avoidance, with real-world flights demonstrating consecutive evasive maneuvers at relative encounter speeds up to 15 m/s without fine-tuning. Code is available at https://github.com/VANdexj/OmniRisk.
Chinese Translation
灵活的四旋翼在快速移动障碍物面前的避障需要预判碰撞并在短暂的反应时间内选择可行的机动方案。由于稀疏的距离观测并不能直接揭示障碍物的运动，因此可靠的预测避障仍然具有挑战性。同时，在线轨迹优化器在障碍物数量较多时要么扩展性差，要么在密集的高速交互中牺牲效率以换取可靠性。我们提出了OmniRisk，这是一种全向规划框架，能够离线学习轨迹级风险，以实现高效的机载避障。一个固定维度的张量结合了激光雷达（LiDAR）范围全景、动态掩膜和笛卡尔表面速度，以共同表示几何形状和运动。我们构建了一个与障碍物速度对齐的不对称风险场，强调接近交互并减弱远离交互的影响。沿着预测的相对轨迹累积这一风险提供了密集的监督，并在障碍物通过后抑制不必要的犹豫。一个双分支圆形卷积网络在单次前向传递中预测候选原语的终端边界状态和动态风险，基于全向锚点格进行选择和闭式重构所选候选原语。这一公式消除了沿轨迹的在线风险累积，使风险推断的成本与障碍物数量无关。OmniRisk使得高效的机载避障成为可能，实际飞行展示了在相对接触速度高达15 m/s的情况下连续的避障机动，而无需进行微调。代码可在 https://github.com/VANdexj/OmniRisk 获取。
cs.RO / 39 / 2609.18193
WAVE-Go: World-Model Navigation with Adaptive Execution for Wheel-Legged Robots
WAVE-Go：具有自适应执行的世界模型导航用于轮腿机器人
Abstract
World models can anticipate the consequences of navigation actions, but predicted action sequences may become invalid during execution, especially when wheel-legged robots encounter dynamic obstacles or change locomotion modes. We propose WAVE-Go, an image-goal navigation framework that separates world-action prediction from interruptible command execution. Its executor adaptively selects an action prefix and cancels pending commands when updated observations invalidate execution. A conditional-risk formulation specifies prefix selection under an estimated cumulative failure budget, while posture and locomotion-mode transitions require clearance, stability, and task-evidence checks. In the reported navigation evaluation, WAVE-Go achieves 74.1% in-distribution success and 63.3% dynamic out-of-distribution success, exceeding the strongest baseline by 4.7 and 7.7 percentage points, respectively, while reducing collisions from 4.4 to 2.9 per 100 m. Compared with interruptible fixed four-command execution, WAVE-Go raises success by 4.0 percentage points while reducing replanning frequency by 51.2% and collision rate by 6.5%. Execution ablations also show that runtime interruption improves success, collision rate, and reaction latency at the cost of additional replanning. These results support adaptive, interruptible execution as a means of balancing navigation performance and planning overhead. Code is available at https://github.com/vigorlee/wave-go.
Chinese Translation
世界模型能够预测导航动作的后果，但在执行过程中，预测的动作序列可能会失效，特别是在轮腿机器人遇到动态障碍物或改变运动模式时。我们提出了WAVE-Go，一个图像目标导航框架，它将世界动作预测与可中断的命令执行分开。其执行器自适应地选择动作前缀，并在更新的观察结果使执行失效时取消待处理命令。条件风险公式在估计的累积失败预算下指定前缀选择，而姿态和运动模式的转换则需要进行清理、稳定性和任务证据检查。在报告的导航评估中，WAVE-Go在分布内成功率达到74.1%，动态分布外成功率达到63.3%，分别超过最强基线4.7和7.7个百分点，同时将每100米的碰撞次数从4.4减少到2.9。与可中断的固定四命令执行相比，WAVE-Go将成功率提高了4.0个百分点，同时将重新规划频率降低了51.2%和碰撞率降低了6.5%。执行消融实验还表明，运行时中断在增加额外重新规划的成本下改善了成功率、碰撞率和反应延迟。这些结果支持自适应、可中断执行作为平衡导航性能和规划开销的一种手段。代码可在 https://github.com/vigorlee/wave-go 获取。
cs.RO / 40 / 2609.18197
WholeBodyWAM: Learning Whole-Body World Action Models with Scalable Motion Priors
WholeBodyWAM：学习具有可扩展运动先验的全身世界动作模型
Abstract
Humanoid whole-body manipulation requires coordinated whole-body dynamics, yet large-scale trajectories from a target robot are expensive to collect and difficult to scale. In contrast, whole-body motion from human and humanoid sources is abundantly available, although such data cannot be directly used as embodiment-specific robot actions. This work asks whether these scalable motion resources can instead provide a transferable predictive prior for humanoid world-action modeling. We introduce WholeBodyWAM, a humanoid world-action model that learns whole-body dynamics from large-scale heterogeneous motion before target-robot training. We curate UniMotion-4K, a motion corpus spanning more than 4K hours from human videos, native 3D motion datasets, and heterogeneous humanoid platforms, and canonicalize these diverse sources into a unified motion space. A language-conditioned Motion Expert is then pretrained to predict future whole-body motion without target-robot action supervision. During robot post-training, the pretrained Motion Expert is integrated with Video and Action Experts through asymmetric Mixture-of-Transformers (MoT) attention, enabling predictive scene dynamics and whole-body motion to jointly inform embodiment-specific action generation. Experiments show that WholeBodyWAM consistently benefits from increased motion-pretraining scale, improves future-motion prediction and downstream task performance, and transfers effectively to real-world humanoid manipulation. Moreover, the pretrained motion prior substantially improves data efficiency under limited target-robot demonstrations.
Chinese Translation
类人全身操控需要协调的全身动力学，但从目标机器人收集大规模轨迹既昂贵又难以扩展。相比之下，来自人类和类人来源的全身运动数据丰富可得，尽管这些数据不能直接作为特定于体现的机器人动作使用。本研究探讨这些可扩展的运动资源是否可以提供可转移的预测先验，用于类人世界动作建模。我们提出了WholeBodyWAM，一种类人世界动作模型，它在目标机器人训练之前，从大规模异构运动中学习全身动力学。我们整理了UniMotion-4K，一个涵盖超过4000小时的人类视频、原生3D运动数据集和异构类人平台的运动语料库，并将这些多样化来源规范化为统一的运动空间。然后，预训练一个语言条件的运动专家，以预测未来的全身运动，而无需目标机器人动作的监督。在机器人后训练阶段，预训练的运动专家通过不对称的混合变换器（Mixture-of-Transformers, MoT）注意力与视频和动作专家集成，使得预测场景动力学和全身运动共同为特定于体现的动作生成提供信息。实验表明，WholeBodyWAM在增加运动预训练规模时始终受益，改善了未来运动预测和下游任务性能，并有效转移到真实世界的类人操控。此外，预训练的运动先验在有限的目标机器人演示下显著提高了数据效率。
cs.RO / 41 / 2609.18207
Reinforcement Learning for Real-Time Vision-Language-Action Policies
用于实时视觉-语言-动作策略的强化学习
Abstract
Reinforcement learning fine-tuning on top of large, pretrained Vision-Language-Action (VLA) models offers promise for highly reliable robot deployment. However, because of their scale, modern VLA models suffer from high inference latency, so the observation used to select an action is often stale by execution time, creating a distribution shift that can substantially degrade reliability and performance. Prior work has explored asynchronous policy execution to reduce the effect of latency, but these methods are mostly built on imitation learning and offer no mechanism for moving beyond the training distribution toward higher reliability. We close this gap by enabling RL fine-tuning that meets the real-time control requirements of dynamic real-world manipulation. Our approach builds on EXPO-FT, a framework for sample-efficient, reliable VLA fine-tuning with reinforcement learning, and decouples slow, expressive action generation from fast, reactive action edits: a large pretrained VLA proposes action chunks using its strong behavior prior, while a lightweight edit policy performs fast, reactive decision-making by editing actions in response to changes in state, conditioned on the latest observation. We instantiate this as Real-Time EXPO-FT, an RL framework for finetuning real-time VLA policies. On the Kinetix benchmark, Real-Time EXPO-FT enables a delayed policy to achieve the best performance among delayed and non-delayed methods in 10 out of 10 environments. On four dynamic real-world tasks, robot object passing, ball balancing, table soccer kicking, and dynamic object picking, with online robot data capped at 10 minutes, Real-Time EXPO-FT improves average policy performance from 42% to 97%, all without human intervention, demonstrating rapid, sample-efficient adaptation to challenging real-world dynamics. Website: https://pd-perry.github.io/real-time-expo-ft
Chinese Translation
在大型预训练的视觉-语言-动作（VLA）模型基础上进行强化学习微调，为高度可靠的机器人部署提供了希望。然而，由于其规模，现代VLA模型面临高推理延迟的问题，因此用于选择动作的观察在执行时往往已经过时，导致分布偏移，这可能显著降低可靠性和性能。之前的研究探索了异步策略执行以减少延迟的影响，但这些方法大多基于模仿学习，并未提供超越训练分布以实现更高可靠性的机制。我们通过实现满足动态现实世界操作实时控制要求的强化学习微调来填补这一空白。我们的方法基于EXPO-FT，这是一个用于样本高效、可靠的VLA微调的强化学习框架，并将缓慢、富表现力的动作生成与快速、反应式的动作编辑解耦：一个大型预训练的VLA利用其强大的行为先验提出动作块，而一个轻量级的编辑策略通过根据最新观察对状态变化进行反应，快速执行决策。我们将其实例化为实时EXPO-FT，这是一个用于微调实时VLA策略的强化学习框架。在Kinetix基准测试中，实时EXPO-FT使得延迟策略在10个环境中实现了延迟和非延迟方法中的最佳性能。在四个动态现实世界任务中，包括机器人物体传递、球平衡、桌上足球踢球和动态物体拾取，在线机器人数据限制在10分钟内，实时EXPO-FT将平均策略性能从42%提高到97%，且全程无需人工干预，展示了对具有挑战性的现实世界动态的快速、样本高效的适应能力。网站：https://pd-perry.github.io/real-time-expo-ft
cs.RO / 42 / 2609.18213
CANTABILE: Learning Expressive Dynamics for Robotic Piano Performance
CANTABILE：用于机器人钢琴表演的表现力动态学习
Abstract
Robotic piano playing has emerged as a standard benchmark for dexterous bimanual manipulation, yet progress on it has been measured almost entirely by note accuracy -- which keys are pressed (pitch) and when (onset) -- leaving the musical dynamics essential for expressive performance neither rewarded nor evaluated. We propose CANTABILE, a dynamics-aware framework for robotic piano performance that (i) closes the score-to-contact loop by conditioning the policy on upcoming velocity goals and mapping each key's angular velocity at onset back to MIDI velocity, (ii) couples a velocity-fidelity reward with an onset-coverage reward, so that dynamics cannot be improved by omitting difficult notes, and (iii) refines a frozen dynamics-aware base policy with an alpha-scaled, finger-only residual that localizes strike-intensity adaptation away from nominal note execution. On EXPRESSIVE-51, a dynamics-rich 51-song subset of RoboPianist, CANTABILE raises Velocity F1 -- jointly measuring pitch, onset, and intensity within a +/-8 MIDI-velocity tolerance -- from 0.06 to 0.34 over the RoboPianist baseline, improves all 51 songs, more than halves matched-note velocity error, and reduces log-mel distance to reference audio by 8%. Intensity-randomized training further enables runtime control of performance intensity without retraining.
Chinese Translation
机器人钢琴演奏已成为灵巧双手操作的标准基准，但其进展几乎完全通过音符准确性来衡量——即按下了哪些键（音高）以及何时按下（起始）——而音乐表现所必需的动态特性既未得到奖励也未被评估。我们提出了CANTABILE，一个关注动态的机器人钢琴表演框架，该框架（i）通过将策略与即将到来的速度目标相结合，并将每个键在起始时的角速度映射回MIDI速度，从而闭合了谱面到接触的循环，（ii）将速度保真度奖励与起始覆盖奖励相结合，以确保动态无法通过省略困难音符来改善，以及（iii）通过一个α缩放的仅手指残差来细化一个冻结的动态感知基础策略，从而将击打强度的适应局限于名义音符执行之外。在EXPRESSIVE-51，一个富含动态的51首歌曲子集的RoboPianist上，CANTABILE将速度F1——在+/-8 MIDI速度容差内共同测量音高、起始和强度——从0.06提高到0.34，改善了所有51首歌曲，匹配音符的速度误差减少了一半以上，并将与参考音频的对数梅尔距离减少了8%。强度随机化训练进一步实现了在不重新训练的情况下对表演强度的实时控制。
cs.RO / 43 / 2609.18232
UMI-Bridge: Action-Anchored Latent Alignment across Human and Robot Manipulation Data
UMI-Bridge：基于动作锚定的跨人类与机器人操作数据的潜在对齐
Abstract
Real-robot demonstrations are limited, motivating the use of human manipulation data collected without robots, including egocentric videos and handheld Universal Manipulation Interface (UMI) demonstrations. However, differences in viewpoint, embodiment, and available action supervision make it difficult to align representations across these sources according to manipulation motion rather than visual appearance. We introduce UMI-Bridge, which uses UMI as an intermediate domain to align representations according to action equivalence rather than pixel similarity. UMI action supervision anchors the latent representation to end-effector motion and gripper behavior, while synchronized head-wrist observations and paired ego-UMI clips support alignment across views and domains. We train a dual-view latent action model (LAM) on human manipulation data without robot demonstrations, then freeze its wrist teacher and dynamics model to regularize vision-language-action (VLA) post-training on UMI and robot data. The shared wrist interface enables this training-time supervision across both domains while preserving the policy's standard inference architecture. Across three real-robot tasks, UMI-Bridge achieves 91.7% mean success versus 73.3% for Naive Co-training with matched UMI and robot data. On two data-efficiency tasks, it surpasses a full-data Robot-only baseline using 25% of the robot demonstrations together with UMI data. It also achieves 85% and 90% success on two additional tasks learned from UMI demonstrations without task-specific robot demonstrations. These results support action-anchored latent alignment for data-efficient robot learning and UMI-to-robot task transfer.
Chinese Translation
真实机器人演示受到限制，这促使我们使用收集自人类操作的数据，包括自我中心视频和手持通用操作接口（Universal Manipulation Interface, UMI）演示。然而，由于视角、体现和可用动作监督的差异，使得根据操作运动而非视觉外观对这些来源的表示进行对齐变得困难。我们提出了UMI-Bridge，它使用UMI作为中介领域，根据动作等价性而非像素相似性对表示进行对齐。UMI的动作监督将潜在表示锚定于末端执行器运动和夹持器行为，同时同步的头部-手腕观察和配对的自我-UMI片段支持跨视角和领域的对齐。我们在没有机器人演示的人类操作数据上训练了一个双视角潜在动作模型（Latent Action Model, LAM），然后冻结其手腕教师和动态模型，以在UMI和机器人数据上进行视觉-语言-动作（Vision-Language-Action, VLA）后训练的正则化。共享的手腕接口使得在两个领域之间进行这种训练时监督成为可能，同时保持策略的标准推理架构。在三个真实机器人任务中，UMI-Bridge的平均成功率达到91.7%，而简单共同训练（Naive Co-training）在匹配的UMI和机器人数据上仅为73.3%。在两个数据效率任务中，它使用25%的机器人演示和UMI数据超越了全数据的机器人基线。在没有特定任务的机器人演示的情况下，它还在两个额外任务中从UMI演示中学习，成功率分别达到85%和90%。这些结果支持基于动作锚定的潜在对齐，以实现数据高效的机器人学习和UMI到机器人任务的转移。
cs.RO / 44 / 2609.18242
ForceDelta-VLA: Distilling Force-Conditioned ActionCorrections for Contact-Rich Manipulation
ForceDelta-VLA：提炼力条件下的动作修正以实现接触丰富的操控
Abstract
Force-aware Vision-Language-Action (VLA) policies improve contact-rich manipulation, but typically combine task-level motion and contact-dependent adjustment in a single action prediction. Demonstrations provide no explicit labels for decomposing that prediction into a reusable reference action and a correction. We present ForceDelta-VLA, a correction-distillation framework that constructs an explicit force-correction target using paired predictions from a frozen teacher's force-conditioned and learned force-agnostic modes. A separate delay-correction target accounts for reference-action mismatch and the change in reference state. Training uses asynchronous schedule replay with the cached task context available during execution. The resulting lightweight policy adjusts the reference actions using recent force history and robot state, responding to contact changes between reference-action updates without regenerating complete action chunks. Across nine single-arm and bimanual contact-rich tasks, ForceDelta-VLA achieves an 82.2% mean success rate, compared with 54.4% for the original ForceVLA baseline. Direct execution of our Stage-1 Temporal Teacher achieves 70.6%. Relative to ForceVLA, the complete system reduces mean peak contact force over successful trials by approximately 26% on both platforms.
Chinese Translation
力感知的视觉-语言-动作（VLA）策略改善了接触丰富的操控，但通常将任务级运动和依赖接触的调整结合在单一的动作预测中。演示并未提供明确的标签来将该预测分解为可重用的参考动作和修正。我们提出了ForceDelta-VLA，一种修正提炼框架，通过来自冻结教师的力条件和学习的无力模式的配对预测构建明确的力修正目标。一个单独的延迟修正目标考虑了参考动作的不匹配和参考状态的变化。训练使用异步调度重放，利用在执行期间可用的缓存任务上下文。最终得到的轻量级策略使用最近的力历史和机器人状态调整参考动作，响应参考动作更新之间的接触变化，而无需重新生成完整的动作块。在九个单臂和双手接触丰富任务中，ForceDelta-VLA的平均成功率达到82.2%，而原始的ForceVLA基线为54.4%。我们第一阶段的时间教师的直接执行成功率为70.6%。与ForceVLA相比，完整系统在成功试验中将平均峰值接触力减少了大约26%，在两个平台上均有体现。
cs.RO / 45 / 2609.18243
Acting in Meters: Learning Metric Interactions for Precise Robotic Manipulation
以米为单位的行动：学习精确机器人操作的度量交互
Abstract
Vision-Language-Action models and World-Action Models have advanced language-conditioned robotic manipulation, yet often leave metric relations among actions, objects, and scene geometry implicit. Human manipulation combines semantic understanding of task-relevant objects with spatial feedback that guides hand motion relative to objects and their surroundings. Inspired by this, we introduce a metric interaction framework that models object-level and scene-level interactions in physical Cartesian space at a shared metric scale. At the object level, Interaction-Centric Tokens (ICTs) explicitly represent end-effector pose trajectories relative to manipulated objects and are jointly denoised with actions, providing physically grounded interaction supervision. At the scene level, the Metric Action Interaction Field (MAIF) uses action and ICT queries to attend to metric scene point-cloud features and learns geometry-conditioned action corrections. Through two-stage adaptation, our framework improves diverse VLA and WAM baselines with a small number of additional parameters and training steps. Experiments demonstrate average success-rate gains of 0.80 and 3.59 percentage points on LIBERO and RoboTwin~2.0, respectively, alongside gains of 6.80 percentage points on real-world tasks and 7.45 percentage points on their out-of-distribution variants.
Chinese Translation
视觉-语言-动作模型和世界-动作模型推动了基于语言的机器人操作的发展，但往往将动作、物体和场景几何之间的度量关系隐含化。人类操作将任务相关物体的语义理解与空间反馈相结合，以指导手部运动相对于物体及其周围环境的运动。受到这一启发，我们提出了一种度量交互框架，该框架在共享的度量尺度上建模物体级和场景级的交互。在物体级，交互中心标记（Interaction-Centric Tokens, ICTs）明确表示相对于被操作物体的末端执行器姿态轨迹，并与动作共同去噪，提供物理基础的交互监督。在场景级，度量动作交互场（Metric Action Interaction Field, MAIF）利用动作和ICT查询来关注度量场景点云特征，并学习几何条件下的动作修正。通过两阶段适应，我们的框架在少量额外参数和训练步骤的情况下，提升了多种视觉-语言-动作（VLA）和世界-动作模型（WAM）的基线性能。实验表明，在LIBERO和RoboTwin~2.0上，平均成功率分别提高了0.80和3.59个百分点，同时在真实世界任务上提高了6.80个百分点，在其分布外变体上提高了7.45个百分点。
cs.RO / 46 / 2609.18245
A3P5 NEMESIS Integrated Rover Design for Environmental Reconnaissance and Robotic Sampling with Reproducible Mobility Analysis and an External Data Machine Learning Calibration Benchmark
A3P5 NEMESIS集成探测车设计用于环境勘察和机器人采样，具有可重复的移动性分析和外部数据机器学习校准基准
Abstract
A3P5 NEMESIS is a four-wheel rover intended to combine remote inspection, environmental observation and lightweight manipulation within one serviceable platform. This study develops a photo-constrained geometric reconstruction, a subsystem architecture and a reproducible analytical assessment while distinguishing physical prototype evidence from proposed functions. An exploratory search retrieved 5,000 bibliographic records across ten queries, yielding 4,897 distinct DOI records and 1,212 metadata candidates; selected primary studies and technical documents informed the design. The reconstructed configuration retains the carbon-pattern enclosure, independently steered wheel assemblies, folded manipulator, inclined camera mast and side sampling equipment. A declared 24 kg scenario predicts 3.28 newton-metres of gearbox-output torque per wheel on a 20-degree grade under equal load sharing; a separate static model shows how a 2 kg forward payload reduces the geometric front-tipping bound from 38.1 degrees to 32.7 degrees. These are design screens, not measured operating limits. A public-data calibration benchmark uses 7,344 eligible hourly observations, eight sensor/environmental predictors and chronological training, validation and test partitions. Validation-selected ridge regression achieves a held-out CO root-mean-square error of 0.502 milligrams per cubic metre, with a 95% daily-block bootstrap interval of 0.435-0.569 milligrams per cubic metre. This result concerns an external sensor array and cannot establish NEMESIS accuracy. The combined analysis identifies priority measurements, proposed control interfaces and mission-specific validation requirements. The contribution is a traceable engineering design study and evaluation framework for a prototype whose integrated field performance remains to be established.
Chinese Translation
A3P5 NEMESIS是一款四轮探测车，旨在将远程检查、环境观察和轻量级操作整合于一个可服务的平台。本研究开发了一种受限于照片的几何重建、子系统架构以及可重复的分析评估，同时区分物理原型证据与提议功能。探索性搜索检索了10个查询中的5,000条文献记录，获得了4,897条独特的DOI记录和1,212条元数据候选项；所选的主要研究和技术文件为设计提供了信息。重建的配置保留了碳图案外壳、独立转向的轮子组件、折叠操作臂、倾斜的相机桅杆和侧面采样设备。声明的24千克场景预测在20度坡度下，每个轮子的齿轮输出扭矩为3.28牛顿·米，且负载均匀分配；一个独立的静态模型显示，2千克的前向载荷将几何前倾界限从38.1度降低到32.7度。这些是设计筛选，而非测量的操作限制。公共数据校准基准使用了7,344个合格的每小时观察数据、八个传感器/环境预测因子以及时间顺序的训练、验证和测试分区。验证选择的岭回归实现了持有的CO均方根误差为0.502毫克每立方米，95%的每日区块自助法区间为0.435-0.569毫克每立方米。该结果涉及外部传感器阵列，无法确定NEMESIS的准确性。综合分析确定了优先测量、提议的控制接口和特定任务的验证要求。该贡献是一个可追溯的工程设计研究和评估框架，用于一个其集成现场性能尚待建立的原型。
cs.RO / 47 / 2609.18259
${M}^2$Tok: Multi-head Multi-codebook Discrete Action Tokenization for Vision-Language-Action Models
${M}^2$Tok：用于视觉-语言-动作模型的多头多码本离散动作标记化
Abstract
Recent advancements have successfully adapted autoregressive language models to process multimodal signals, such as images and actions. Since raw action signals are continuous, effective tokenization is essential to map high-dimensional inputs into compact discrete tokens for autoregressive processing. However, existing discrete action tokenizers often suffer from high reconstruction loss, failing to preserve the fine-grained dynamics required for precise control. This ``discretization bottleneck'' significantly limits the performance ceiling of downstream Vision-Language-Action (VLA) models. To address this, we propose $\mathcal{M}^2$Tok, a Multi-head Multi-codebook Action Tokenizer designed to minimize reconstruction error and enhance policy performance. Our approach introduces two key structural innovations: (1) we decompose the latent action features into multiple heads, enabling the model to implicitly align specific heads with distinct action dimensions; (2) we assign independent codebooks to each head for quantization. By leveraging the combinatorial nature of multiple codebooks, we significantly expand the representational expressivity of the tokenizer, leading to substantially lower reconstruction loss compared to previous methods. We evaluate the $\mathcal{M}^2$Tok-based VLA on the RoboTwin, Simpler-Env, and 3 zero-shot real-world tasks. Experimental results demonstrate our method not only achieves superior reconstruction fidelity but also significantly boosts the success rate of VLA models. Comprehensive ablation studies further confirm the effectiveness of the multi-head and multi-codebook mechanisms. Code is available at \href{https://github.com/cpaaax/M2Tok}{https://github.com/cpaaax/M2Tok}.
Chinese Translation
近期的进展成功地将自回归语言模型适应于处理多模态信号，如图像和动作。由于原始动作信号是连续的，有效的标记化对于将高维输入映射为紧凑的离散标记以进行自回归处理至关重要。然而，现有的离散动作标记器往往面临高重建损失的问题，未能保留精确控制所需的细粒度动态。这种“离散化瓶颈”显著限制了下游视觉-语言-动作（VLA）模型的性能上限。为了解决这个问题，我们提出了$ ext{M}^2$Tok，一种旨在最小化重建误差并增强策略性能的多头多码本动作标记器。我们的方法引入了两个关键的结构创新：（1）我们将潜在动作特征分解为多个头，使模型能够隐式地将特定头与不同的动作维度对齐；（2）我们为每个头分配独立的码本进行量化。通过利用多个码本的组合特性，我们显著扩展了标记器的表示表达能力，从而相比于之前的方法显著降低重建损失。我们在RoboTwin、Simpler-Env和3个零样本真实世界任务上评估了基于$ ext{M}^2$Tok的VLA。实验结果表明，我们的方法不仅实现了更优的重建保真度，还显著提高了VLA模型的成功率。全面的消融研究进一步确认了多头和多码本机制的有效性。代码可在 ext{https://github.com/cpaaax/M2Tok} 获取。
cs.RO / 48 / 2609.18293
Function-Preserving Data Generation for Zero-Shot Real-to-Sim-to-Real Manipulation
保持功能的数据生成用于零-shot真实到模拟再到真实的操控
Abstract
Robotic data generation is a promising paradigm for scaling robot learning without collecting large-scale real-world data. However, generating geometrically diverse yet physically valid data for contact-rich tasks remains challenging, especially when success depends on precise geometric interfaces. Standard shape augmentation methods often distort task-critical interfaces, resulting in invalid contact relationships, e.g., fit mismatches or interpenetration, rendering downstream interactions infeasible. To address these limitations, we propose a function-preserving Real-to-Sim-to-Real framework that generates synthetic demonstrations from reconstructed assets without teleoperated source trajectories. Our method augments task-relevant object geometries through constraint-guided mesh deformation, together with physically consistent transfer of task poses and collision proxies. Visual domain randomization is further applied during simulation rollouts, enabling robust zero-shot policy deployment without real-world fine-tuning. Extensive experiments in both real-world and simulation settings demonstrate that our method enables robust generalization across unseen object geometries and diverse visual conditions in contact-rich and long-horizon tasks. Our method provides a practical path toward scalable robot learning for contact-rich tasks via shape deformation.
Chinese Translation
机器人数据生成是一种有前景的范式，可以在不收集大规模真实世界数据的情况下扩展机器人学习。然而，为接触丰富的任务生成几何多样但物理有效的数据仍然具有挑战性，特别是当成功依赖于精确的几何接口时。标准的形状增强方法通常会扭曲任务关键的接口，导致无效的接触关系，例如适配不匹配或相互穿透，从而使下游交互变得不可行。为了解决这些限制，我们提出了一种保持功能的真实到模拟再到真实的框架，该框架从重建的资产生成合成演示，而无需遥控源轨迹。我们的方法通过约束引导的网格变形增强与任务相关的物体几何，同时物理一致地转移任务姿态和碰撞代理。在模拟展开过程中进一步应用视觉领域随机化，使得在没有真实世界微调的情况下能够进行稳健的零-shot策略部署。在真实世界和模拟环境中进行的大量实验表明，我们的方法能够在接触丰富和长时间跨度的任务中实现对未见物体几何和多样视觉条件的稳健泛化。我们的方法为通过形状变形实现接触丰富任务的可扩展机器人学习提供了一条实用路径。
cs.RO / 49 / 2609.18324
RAFAIL: Relationship-Aware Failure Detection for Robotic Manipulation
RAFAIL：面向关系的机器人操作故障检测
Abstract
Detecting failures during execution is essential for reliable robotic manipulation. Vision-language models (VLMs) can assess task outcomes semantically but add runtime computation, whereas out-of-distribution (OOD) detectors may respond to harmless scene variations rather than failure-relevant deviations. We introduce RAFAIL, a framework for detecting execution failures during robotic manipulation. RAFAIL identifies failures by detecting anomalies in task-relevant relationships between entities, such as a gripper and an object or an object and its target. By focusing OOD detection on relevant parts of the observation, RAFAIL reduces sensitivity to task-irrelevant scene variation. Offline, a VLM annotates successful demonstrations with task progress and relationship importance, which are used to learn point-cloud-based relationship representations without relying on policy-internal features. At runtime, relationship-specific OOD detectors evaluate these representations while relationship importance and task progress are predicted without VLM inference. RAFAIL requires no failure data and achieves 73.4% balanced accuracy across three real-world robotic manipulation tasks, outperforming the strongest evaluated OOD- and uncertainty-based baselines.
Chinese Translation
在执行过程中检测故障对于可靠的机器人操作至关重要。视觉语言模型（VLMs）可以从语义上评估任务结果，但会增加运行时计算，而分布外（OOD）检测器可能对无害的场景变化作出反应，而非与故障相关的偏差。我们提出了RAFAIL，一个用于检测机器人操作执行故障的框架。RAFAIL通过检测实体之间在任务相关关系中的异常来识别故障，例如夹持器与物体之间或物体与其目标之间的关系。通过将OOD检测集中在观察的相关部分，RAFAIL降低了对与任务无关的场景变化的敏感性。在离线阶段，VLM对成功的演示进行标注，记录任务进展和关系重要性，这些信息用于学习基于点云的关系表示，而无需依赖策略内部特征。在运行时，特定于关系的OOD检测器评估这些表示，同时在没有VLM推理的情况下预测关系重要性和任务进展。RAFAIL不需要故障数据，并在三个真实世界的机器人操作任务中实现了73.4%的平衡准确率，超越了最强的评估OOD和不确定性基线。
cs.RO / 50 / 2609.18326
UAVs Meet Embodied Intelligence: Bridging Human Intents and Flying Dynamics Via Harnessing Physical-Digital AI Agents
无人机与具身智能的结合：通过利用物理-数字人工智能代理桥接人类意图与飞行动态
Abstract
Unmanned aerial vehicles (UAVs) extend embodied intelligence into continuous three-dimensional space, where perception, reasoning, physical embodiment, and action are tightly coupled through flight and environmental interaction. Recent advances in foundation models, world models, and AI agents are shifting UAV autonomy from task-specific perception and control toward systems that can interpret human intent, understand open environments, reason about physical consequences, and organize complex behaviors under embodiment and flight-dynamic constraints. We characterize this emerging paradigm as UAV embodied intelligence (UAV EI) and distinguish it from its system realization, the embodied-intelligent UAV (EI UAV). To provide a unified view of the field, we introduce a 5+5 framework that describes UAV EI through five capability dimensions and EI UAVs through five architectural layers spanning physical embodiment, general cognition, embodied skills, external interaction, and system harnessing. Based on this framework, we systematically review recent progress in embodied morphology, embodied perception, world models, embodied planning, vision-language navigation, embodied manipulation, and embodied collaboration. We further identify long-horizon autonomy, predictive physical reasoning, test-time skill acquisition, and autonomous capability evolution as key challenges toward more general aerial embodied intelligence. Finally, we argue that harnessing physical-digital AI agents, through persistent coupling of digital intelligence with physical sensing, dynamics, action, and feedback, provides a system-level pathway toward adaptive and continuously evolving UAV autonomy. Project resources are available at our project website and GitHub repository.
Chinese Translation
无人机（UAV）将具身智能扩展到连续的三维空间，在这里，感知、推理、物理具身和行动通过飞行和环境交互紧密结合。基础模型、世界模型和人工智能代理的最新进展正在将无人机的自主性从任务特定的感知和控制转向能够解释人类意图、理解开放环境、推理物理后果并在具身和飞行动态约束下组织复杂行为的系统。我们将这一新兴范式称为无人机具身智能（UAV EI），并将其与系统实现——具身智能无人机（EI UAV）区分开来。为了提供该领域的统一视角，我们引入了一个5+5框架，通过五个能力维度描述无人机具身智能，并通过五个架构层次描述具身智能无人机，这些层次涵盖了物理具身、一般认知、具身技能、外部交互和系统利用。基于这一框架，我们系统回顾了在具身形态、具身感知、世界模型、具身规划、视觉-语言导航、具身操作和具身协作等方面的最新进展。我们进一步确定了长时间自主性、预测性物理推理、测试时技能获取和自主能力演化作为实现更一般航空具身智能的关键挑战。最后，我们认为，通过将数字智能与物理感知、动态、行动和反馈持续耦合，利用物理-数字人工智能代理提供了一条系统级路径，以实现适应性和持续演变的无人机自主性。项目资源可在我们的项目网站和GitHub仓库中获取。
cs.RO / 51 / 2609.18358
GraphPoint: Semantic Entity Graphs and Point Trajectories for Compositional Robot Manipulation
GraphPoint：用于组合机器人操作的语义实体图和点轨迹
Abstract
Robot manipulation policies often struggle to generalize beyond their demonstrations, even when new instructions involve familiar objects and behaviors. When language and scenes are strongly correlated during training, a policy can learn a fixed visual-action mapping rather than respond to the requested behavior. We investigate compositional reuse at two levels: within a subtask, combining familiar entities, action types, and action modifiers; and across subtasks, reusing learned subtasks in unseen long-horizon tasks. We introduce CoMani, a benchmark with controlled splits for evaluating both capabilities. Matched initial scenes and controlled changes to a single semantic factor encourage reliance on language rather than visual shortcuts. We further propose GraphPoint, which connects semantic entity graphs to geometric control by predicting future gripper point trajectories and converting them into actions using robot geometry. The framework organizes the gripper and objects by semantic roles and conditions their interactions on action types and modifiers, while predicted progress guides transitions during execution. Experiments and ablations on CoMani validate the effectiveness of our method for instruction-dependent generalization at both levels. Code will be released at GraphPoint.
Chinese Translation
机器人操作策略通常难以超越其演示进行泛化，即使新的指令涉及熟悉的物体和行为。当语言和场景在训练期间强相关时，策略可能学习到固定的视觉-动作映射，而不是对请求的行为作出反应。我们在两个层面上研究组合重用：在子任务内，结合熟悉的实体、动作类型和动作修饰符；在子任务之间，重用在未见的长时间任务中学习到的子任务。我们引入了CoMani，一个具有受控划分的基准，用于评估这两种能力。匹配的初始场景和对单一语义因素的受控变化鼓励依赖语言而非视觉捷径。我们进一步提出了GraphPoint，它通过预测未来夹持器点轨迹并利用机器人几何将其转换为动作，将语义实体图与几何控制连接起来。该框架根据语义角色组织夹持器和物体，并根据动作类型和修饰符条件化它们的交互，同时预测的进展在执行过程中指导过渡。对CoMani的实验和消融验证了我们的方法在两个层面上对指令依赖性泛化的有效性。代码将在GraphPoint发布。
cs.RO / 52 / 2609.18359
RecMorph: Topology-Guided Spatial Recurrence for Generalized Morphology Control
RecMorph：基于拓扑的空间递归用于广义形态控制
Abstract
Generalized morphology control requires a single policy to transform information across limbs with different physical roles, coordinate whole-body motion, and remain efficient as body size grows. Existing communication mechanisms address these requirements only partially. We introduce RecMorph, a topology-guided spatial recurrent architecture that uses recurrent sequence computation to jointly perform cross-limb communication and representation transformation. A depth-first traversal converts the kinematic tree into a morphology-derived sequence, along which shared bidirectional transitions progressively transform limb information before action decoding. Residual preservation, RMS normalization, and input-dependent channel modulation stabilize this repeated spatial transformation, yielding linear token complexity at fixed model width and depth. Across five UNIMAL tasks, RecMorph achieves the strongest mean final training performance among the evaluated generalized morphology controllers and the highest measured inference throughput on FT, while generalizing to unseen variations and bodies with up to 30 limbs. We further migrate representative generalized controllers from UNIMAL benchmarks to a four-platform quadruped setting. RecMorph achieves the best macro-averaged performance under nominal and high friction, reduces nominal velocity RMSE by 43.5% relative to specialist MLPs, and one shared policy completes 40 physical Go1/Go2 trials without falls. These results show that topology-guided recurrent transformation provides an effective and efficient communication mechanism for Generalized Morphology Control and remains effective when transferred from procedural bodies to physical robot platforms. Code and experimental resources are publicly available at https://github.com/quanruirao/RecMorph.
Chinese Translation
广义形态控制需要一个单一策略来在具有不同物理角色的肢体之间转换信息，协调全身运动，并在身体尺寸增大时保持高效。现有的通信机制仅部分满足这些要求。我们提出了RecMorph，一种基于拓扑的空间递归架构，利用递归序列计算共同执行跨肢体通信和表示转换。深度优先遍历将运动学树转换为形态派生序列，沿着该序列，共享的双向转换逐步转变肢体信息，然后进行动作解码。残差保留、RMS归一化和输入依赖的通道调制稳定了这种重复的空间转换，使得在固定模型宽度和深度下实现线性标记复杂度。在五个UNIMAL任务中，RecMorph在评估的广义形态控制器中实现了最强的平均最终训练性能，并在FT上达到了最高的推理吞吐量，同时能够推广到未见的变体和最多30个肢体的身体。我们还将UNIMAL基准中的代表性广义控制器迁移到四平台四足动物设置中。RecMorph在名义和高摩擦条件下实现了最佳的宏平均性能，相较于专业的多层感知器（MLPs），减少了名义速度的均方根误差（RMSE）43.5%，并且一个共享策略在没有跌倒的情况下完成了40次物理Go1/Go2试验。这些结果表明，基于拓扑的递归转换为广义形态控制提供了一种有效且高效的通信机制，并在从程序化身体转移到物理机器人平台时仍然有效。代码和实验资源可在 https://github.com/quanruirao/RecMorph 上公开获取。
cs.RO / 53 / 2609.18374
Decoupling Vision, Language, and Action for Efficient Multi-Task Robot Policies
解耦视觉、语言和动作以实现高效的多任务机器人策略
Abstract
Vision-Language-Action (VLA) models attach an action module to a Vision-Language Model (VLM) with billions of parameters and pay for that backbone at every control step. For a low-level manipulation policy, this cost may be unnecessary: the VLM supplies vision and language embeddings, and recent standalone vision encoders and encoder-only language models now match or exceed large VLMs on visual embedding and language understanding benchmarks. We study this question with a controlled experiment. Holding the demonstrations, the training budget, the tasks, and the measurement platform fixed, we vary the vision encoder, the language encoder, and the action head of a decoupled policy and compare against seven VLA baselines. The study yields the Decoupled Embodiment Model (DEM), which pairs a fine-tuned DINOv3 encoder and a frozen NeoBERT encoder with a MeanFlow head that generates each action chunk in a single forward pass. On 18 simulated manipulation tasks with held-out language paraphrases and randomized scenes, and on three real-robot tasks, DEM achieves observed success comparable to state-of-the-art VLM-backbone policies under our evaluation protocol, while running at eight to seventeen times their inference frequency and drawing six to fifteen times less energy per inference. Within this task scope, modern decoupled components offer a better success--latency--energy trade-off.
Chinese Translation
视觉-语言-动作（VLA）模型将一个动作模块附加到具有数十亿参数的视觉-语言模型（VLM）上，并在每个控制步骤中为该骨干网络付费。对于低级别的操作策略，这种成本可能是不必要的：VLM提供视觉和语言嵌入，而最近的独立视觉编码器和仅编码语言模型在视觉嵌入和语言理解基准测试中现在已达到或超过大型VLM的表现。我们通过一个受控实验研究这个问题。在固定演示、训练预算、任务和测量平台的情况下，我们改变解耦策略的视觉编码器、语言编码器和动作头，并与七个VLA基线进行比较。研究结果产生了解耦体现模型（DEM），该模型将一个微调的DINOv3编码器和一个冻结的NeoBERT编码器与一个MeanFlow头配对，该头在单次前向传播中生成每个动作块。在18个模拟操作任务中，使用保留的语言同义句和随机场景，以及三个真实机器人任务中，DEM在我们的评估协议下实现了与最先进的VLM骨干策略相当的成功，同时其推理频率为后者的八到十七倍，且每次推理消耗的能量为后者的六到十五倍。在这一任务范围内，现代解耦组件提供了更好的成功率、延迟和能量的权衡。
cs.RO / 54 / 2609.18392
DistAL: Distance-based Advantage Learning for VLA Fine-Tuning
DistAL：基于距离的优势学习用于VLA微调
Abstract
Vision-language-action models (VLAs) have trans- formed the field of robotic manipulation in recent years by combining the semantic understanding of LLMs with the precise control of flow-matching policies. Advantage conditioning is a recent technique that iteratively improves VLAs by training a value function on deployment data and using this to train an advantage-conditioned policy. Previous works have only applied simple, low-information success/failure rewards, which leave the value function unable to distinguish states of differing quality beyond how far along the task they appear. Motivated by an exploration of out-of-distribution (OOD) detection methods, we introduce Distance-based Advantage Learning (DistAL), which, by using an embedding space distance as a reward, produces a more informative value function and subsequently a higher downstream task success rate. We validate our method on a series of simulation benchmarks and dexterous bi-manual manipulation tasks on real hardware.
Chinese Translation
视觉-语言-动作模型（VLA）近年来通过将大规模语言模型（LLM）的语义理解与流匹配策略的精确控制相结合，彻底改变了机器人操控领域。优势条件化是一种近期技术，通过在部署数据上训练价值函数并利用其训练优势条件化策略，迭代地改善VLA。之前的研究仅应用简单的低信息成功/失败奖励，这使得价值函数无法区分不同质量的状态，超出任务进展的程度。基于对分布外（OOD）检测方法的探索，我们提出了基于距离的优势学习（DistAL），该方法通过使用嵌入空间距离作为奖励，生成更具信息量的价值函数，从而提高下游任务的成功率。我们在一系列仿真基准测试和真实硬件上的灵巧双手操控任务中验证了我们的方法。
cs.RO / 55 / 2609.18395
DetAug: Obstacle-Blind Trajectory Augmentation for Zero-shot Obstacle Avoidance
DetAug：无障碍轨迹增强用于零-shot障碍物避让
Abstract
Policies for robotic manipulation are produced by training on large teleoperated datasets. These datasets typically consist of free-space trajectories, making them difficult to transfer to test-time environments with obstacles. Previous methods for closing this gap have largely fallen into two groups. Dataset augmentation addresses it at training time but needs obstacle geometry in advance, whereas steering an existing checkpoint at inference time avoids that requirement but is limited in flexibility. Our method draws from both areas without inheriting either drawback. DetAug applies an obstacle-blind augmentation scheme to the transit phases of a free-space dataset, leaving object interactions untouched, and records the augmentation parameters as an explicit conditioning label. At inference it samples a batch of labels and executes the trajectory with the lowest collision cost. On the SafeLIBERO benchmark DetAug achieves a collision-free success rate more than 20pp above the next best method, and selecting over the label space outperforms guidance on the same policy by 26pp. On real hardware, inference-time steering methods collapse on tasks requiring large detours, while DetAug matches or exceeds an obstacle-conditioned baseline without ever seeing obstacles in training.
Chinese Translation
机器人操作策略通常通过在大型遥控数据集上进行训练而生成。这些数据集通常由自由空间轨迹组成，使其难以转移到具有障碍物的测试环境中。以往缩小这一差距的方法大致分为两类。数据集增强在训练时解决这一问题，但需要提前了解障碍物的几何形状，而在推理时引导现有检查点则避免了这一要求，但灵活性有限。我们的方法结合了这两个领域的优点而没有继承任何缺点。DetAug对自由空间数据集的过渡阶段应用了一种无障碍增强方案，保持物体交互不变，并将增强参数记录为明确的条件标签。在推理时，它采样一批标签并执行碰撞成本最低的轨迹。在SafeLIBERO基准测试中，DetAug的无碰撞成功率比下一个最佳方法高出20个百分点，并且在标签空间的选择上比同一策略的引导提高了26个百分点。在真实硬件上，推理时的引导方法在需要大绕行的任务上崩溃，而DetAug在训练中从未见过障碍物的情况下，能够匹配或超过基于障碍物条件的基线。
cs.RO / 56 / 2609.18434
Hardware-Free Robotics Laboratories in Mixed Reality
无硬件混合现实机器人实验室
Abstract
Teaching robotics relies on screen-based simulation, showing robot motion in an abstract coordinate frame rather than at real scale in the learner's own space, while access to physical hardware is limited by cost, safety, and scheduling constraints. We present MR-Robotics LAB, a mixed-reality (MR) platform that replays MATLAB-generated robot trajectories at real scale within the learner's physical environment. A browser-based service validates a MATLAB workspace file (.mat), normalizes units, and publishes a versioned JSON trajectory; a Unity application on a Meta Quest 3 then reproduces the authored joint configurations under position control and replays them at the declared frame rate within a physics-enabled scene that supports collision detection and end-effector grasping. A formative single-group evaluation with engineering students found that participants reported low setup effort (M = 4.67 on a 5-point scale) and perceived support for workspace understanding from multi-viewpoint inspection (M = 4.56), and 83% of participants affirmed their willingness to use the platform in an introductory robotics course. The evaluation instrument records only perceived outcomes, without counterbalancing or a learning measure, so no comparative advantage over desktop simulation is claimed. The contribution is a reusable simulation-to-MR trajectory pathway and design guidance for hardware-free robot visualization in engineering education.
Chinese Translation
机器人教学依赖于基于屏幕的仿真，展示机器人运动在抽象坐标系中，而不是在学习者自身空间中的真实比例，同时，物理硬件的获取受到成本、安全和时间安排的限制。我们提出了MR-Robotics LAB，一个混合现实（MR）平台，可以在学习者的物理环境中以真实比例重放MATLAB生成的机器人轨迹。一个基于浏览器的服务验证MATLAB工作区文件（.mat），标准化单位，并发布一个版本化的JSON轨迹；然后在Meta Quest 3上的Unity应用程序在位置控制下重现所编写的关节配置，并在支持碰撞检测和末端执行器抓取的物理启用场景中以声明的帧率重放它们。对工程学生进行的形成性单组评估发现，参与者报告的设置工作量较低（M = 4.67，满分5分），并且从多视角检查中感知到对工作区理解的支持（M = 4.56），83%的参与者确认他们愿意在入门机器人课程中使用该平台。评估工具仅记录感知结果，没有进行平衡或学习测量，因此没有声称相对于桌面仿真的比较优势。该贡献是一个可重用的仿真到MR轨迹路径和无硬件机器人可视化的设计指导，适用于工程教育。
cs.RO / 57 / 2609.18451
VLM-MPPI: Grounding Natural Language in Behaviorally Diverse Trajectories for Aerial Navigation
VLM-MPPI：在行为多样化轨迹中将自然语言与空中导航结合
Abstract
We present a hierarchical UAV navigation framework that aligns natural-language intent with dynamically feasible flight behaviors in cluttered indoor environments. To bridge the gap between abstract semantics and low-level control, we employ a parallelized ensemble of six behavior-conditioned Model Predictive Path Integral (MPPI) planners. Crucially, by designing mode-specific guiding costs and sampling biases, we induce distinct trajectory modes that converge to unique behavioral means, yielding a compact set of intentionally diverse candidates rather than mere stochastic variations. We project these 3D candidates onto the onboard first-person-view RGB stream, turning language grounding into a visual action selection problem. A pretrained vision--language model (VLM) asynchronously selects the candidate index given the overlaid FPV image and a natural-language prompt, while MPPI replans at 20Hz and a PID-based low-level controller tracks the selected trajectory. We implement the full pipeline in NVIDIA Isaac Sim and on a real-world quadrotor platform equipped with LiDAR and RGB sensing. Experiments in both simulation and real-world flights show semantically meaningful behavior diversity, robust language alignment despite VLM latency, and safe, repeatable flight across all modes, achieving 100% task success in our evaluated scenarios.
Chinese Translation
我们提出了一种分层无人机导航框架，该框架将自然语言意图与复杂室内环境中的动态可行飞行行为对齐。为了弥合抽象语义与低级控制之间的差距，我们采用了一种并行化的六个行为条件模型预测路径积分（MPPI）规划器的集成。通过设计特定模式的引导成本和采样偏差，我们诱导出不同的轨迹模式，这些模式收敛到独特的行为均值，从而产生一组紧凑的有意多样化候选轨迹，而不仅仅是随机变化。我们将这些三维候选轨迹投影到机载第一人称视角RGB流中，将语言结合转化为视觉动作选择问题。一个预训练的视觉-语言模型（VLM）在给定叠加的第一人称视角图像和自然语言提示的情况下异步选择候选索引，而MPPI以20Hz的频率重新规划，基于PID的低级控制器跟踪所选轨迹。我们在NVIDIA Isaac Sim和配备LiDAR及RGB传感器的真实四旋翼平台上实现了完整的管道。在模拟和真实飞行中的实验显示出语义上有意义的行为多样性，尽管存在VLM延迟，语言对齐依然稳健，并且在所有模式下实现了安全、可重复的飞行，在我们评估的场景中任务成功率达到100%。
cs.RO / 58 / 2609.18455
ForwardDLO: Model-Based Bimanual Shape Matching of Unconstrained Deformable Linear Objects
ForwardDLO：基于模型的无约束可变形线性物体双手形状匹配
Abstract
Ropes, cables, and other deformable linear objects appear in tasks from untangling to cable routing and suturing, yet controlling their shape remains a challenge in robot manipulation. We study model-based shape control in a general setting: the object lies unfixated on a support surface and two arms may grasp and move it anywhere along its length. Because each arm chooses a grasp point, direction, and magnitude, the joint action space is combinatorially large, and the dynamics model's per-prediction cost bounds how much of it a planner can search. We present ForwardDLO, a recurrent latent dynamics model for this unfixated bimanual setting that predicts per-segment displacements grounded in the observed rope state at every step. Our model reaches accuracy comparable to more expensive baselines while containing no explicit segment-to-segment operations, which makes batched evaluation of candidate actions cheap. On open-loop prediction of real rope motion it reaches the lowest error of the learned models we evaluate, 13% below the strongest baseline. Within a fixed time budget it scores 8 to 22 times more candidate actions than models of comparable accuracy while matching them in real-world shape matching; and on a simulated routing task at a 30Hz control rate, this throughput converts into 98% task success versus at most 30% for the baselines at their own budgets. We release the model, code, and a dataset of 2.42 million simulated and 14,107 real rope transitions at https://anonymous.4open.science/r/ForwardDLO/
Chinese Translation
绳索、电缆及其他可变形线性物体在解缠、布线和缝合等任务中出现，但控制其形状仍然是机器人操作中的一大挑战。我们在一个一般性环境中研究基于模型的形状控制：物体在支撑表面上未固定，两个机械臂可以在其长度上随意抓取和移动。由于每个机械臂选择抓取点、方向和力度，联合动作空间的组合数量非常庞大，而动态模型的每次预测成本限制了规划者能够搜索的范围。我们提出了ForwardDLO，这是一种用于该未固定双手设置的递归潜在动态模型，它在每一步基于观察到的绳索状态预测每个段的位移。我们的模型在准确性上与更昂贵的基线相当，同时不包含显式的段对段操作，这使得候选动作的批量评估成本低廉。在对真实绳索运动的开放环预测中，它的误差达到了我们评估的学习模型中最低，低于最强基线13%。在固定的时间预算内，它的候选动作数量比具有相当准确性的模型多出8到22倍，同时在实际形状匹配中与它们相匹配；在以30Hz控制频率进行的模拟布线任务中，这种吞吐量转化为98%的任务成功率，而基线在其自身预算下最多只有30%的成功率。我们在https://anonymous.4open.science/r/ForwardDLO/发布了模型、代码以及242万条模拟和14,107条真实绳索过渡的数据集。
cs.RO / 59 / 2609.18482
Real-Time Bounded Catenary Solver for UAV Tether Modeling
无人机系留建模的实时有界悬链线求解器
Abstract
For non-stationary tethered multirotor UAVs in real-world conditions, simulating the forces imposed on the drone by the aerodynamic drag of the tether becomes crucial, with online use cases placing a hard bound on the maximum solve time. In previous work, a quasi-analytical catenary tether model reached a mean solve time of 0.51 ms using a general-purpose root finder, but without any worst-case guarantees or proven convergence. In this work, we reformulate the inner solver by reducing the catenary boundary-value problem to a single transcendental equation in one well-conditioned unknown. We derive a closed-form bracket and prove monotonicity and convexity as well as existence and uniqueness of the root, which together guarantee convergence of the solver. We further propose a two-regime initial guess which approximates the true root within 3.4% and reduces the mean iteration count by 68.0% to 2.36 compared to the textbook initialization. Building on the hybrid root-finding method rtsafe (Newton-Raphson with bisection fallback giving bounded iteration counts), we implement a specialized variant that exploits the problem structure to omit unnecessary checks while retaining correctness, which gives up to 1.3 times speedup. With the proposed solver the full tether model achieves a nearly constant solve time of 6.9 us on average and 7.7 us at worst, a 40 times speedup over an optimized re-implementation of the previous method, while agreeing with it to a relative deviation of 8.7e-9. Because the reformulation leaves the underlying physical model untouched, the experimental validation of the previous work carries over unchanged. We further demonstrate its suitability for embedded, resource-constrained platforms with a Lua implementation running directly in ArduPilot on a drone's flight controller, where it stays well inside the scheduling budget with a mean solve time of 0.74 ms.
Chinese Translation
对于在现实条件下的非静态系留多旋翼无人机，模拟系留绳索的空气阻力对无人机施加的力变得至关重要，而在线应用场景对最大求解时间有严格限制。在之前的研究中，采用准解析悬链线模型的求解时间平均为0.51毫秒，使用了一种通用根查找器，但没有提供最坏情况保证或证明收敛性。在本研究中，我们通过将悬链边值问题重新表述为一个单一的良好条件未知数的超越方程，重新构建了内部求解器。我们推导出一个封闭形式的区间，并证明了根的单调性、凸性以及存在性和唯一性，这些共同保证了求解器的收敛性。我们进一步提出了一种双区间初始猜测，能够在3.4%内逼近真实根，并将平均迭代次数减少68.0%，降至2.36，相较于教科书初始化。基于混合根查找方法rtsafe（牛顿-拉夫森法与二分法回退，确保有界迭代次数），我们实现了一种专门的变体，利用问题结构省略不必要的检查，同时保持正确性，从而实现高达1.3倍的加速。使用所提出的求解器，完整的系留模型在平均情况下实现了近乎恒定的求解时间为6.9微秒，最坏情况下为7.7微秒，相较于之前方法的优化重实现加速了40倍，同时与其相对偏差为8.7e-9。由于重新表述未改变基础物理模型，之前工作的实验验证得以无缝延续。我们进一步展示了其在嵌入式、资源受限平台上的适用性，提供了一个在无人机飞行控制器上直接运行于ArduPilot的Lua实现，其平均求解时间为0.74毫秒，远低于调度预算。
cs.RO / 60 / 2609.18487
ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models
ActionPiece：重新思考自回归视觉-语言-动作模型中的动作标记化
Abstract
Action tokenizers play a central role in autoregressive vision-language-action (VLA) models, determining both the targets for policy training and the executable commands recovered from predicted tokens. Their fidelity is commonly evaluated using pointwise reconstruction metrics such as mean squared error (MSE), yet small individual errors do not fully characterize how faithfully action adjustments across demonstrations are preserved. After compression, similar actions may still cluster around a representative motion, while the adjustments needed for different contexts are diminished, distorted, or even reversed. We introduce physical rank consistency (PRC) to measure how well tokenization preserves local physical distance rankings after reconstruction. Evaluating decoded actions provides a common reference across token vocabularies and decoder architectures, complementing pointwise accuracy with a measure of relational fidelity. We further present ActionPiece, which preserves physical action relationships through joint supervision of representation learning and quantization. Physical rank preservation supervises near-far ordering in encoder and quantized feature distances, while quantization regularization applies the same ordering to codeword assignment distributions. Both objectives augment reconstruction, producing discrete action tokens for standard autoregressive policy learning and execution through a frozen decoder. Under the same Qwen3-VL-4B policy training setup, ActionPiece achieves 94.8% on LIBERO and 68.8% on unseen LIBERO-Plus, with additional evaluations reaching 71.9% on SimplerEnv and 51.5% across VLA-Arena L0-L2. Component ablations show that the two objectives jointly improve PRC and policy success, demonstrating the value of physical relationship supervision for action tokenization.
Chinese Translation
动作标记器在自回归视觉-语言-动作（VLA）模型中扮演着核心角色，决定了政策训练的目标以及从预测标记中恢复的可执行命令。它们的准确性通常通过点对点重构指标（如均方误差 MSE）进行评估，但个别小误差并不能完全表征在不同演示中动作调整的保留程度。在压缩后，相似的动作可能仍然聚集在一个代表性动作周围，而不同上下文所需的调整则可能被削弱、扭曲甚至反转。我们引入物理秩一致性（PRC）来衡量标记化在重构后保留局部物理距离排名的效果。评估解码后的动作为不同标记词汇和解码器架构提供了一个共同的参考，补充了点对点准确性与关系保真度的度量。我们进一步提出了 ActionPiece，通过对表示学习和量化的联合监督来保留物理动作关系。物理秩保留监督编码器和量化特征距离中的近远排序，而量化正则化则将相同的排序应用于代码字分配分布。这两个目标共同增强了重构，生成离散的动作标记，以便通过冻结的解码器进行标准自回归政策学习和执行。在相同的 Qwen3-VL-4B 政策训练设置下，ActionPiece 在 LIBERO 上达到了 94.8%，在未见过的 LIBERO-Plus 上达到了 68.8%，额外评估在 SimplerEnv 上达到了 71.9%，在 VLA-Arena L0-L2 上达到了 51.5%。组件消融实验表明，这两个目标共同改善了 PRC 和政策成功率，展示了物理关系监督在动作标记化中的价值。
cs.RO / 61 / 2609.18497
TAO-Force: Unifying Force-Aware Perception and Fast-Slow Control for Contact-Rich Manipulation
TAO-Force：统一接触丰富操作中的力感知与快慢控制
Abstract
Vision-Language-Action (VLA) models have demonstrated strong performance across diverse robotic manipulation tasks, yet their predominantly vision-centric perception and position-controlled execution remain insufficient for contact-rich manipulation. Visual observations alone often provide limited evidence of contact onset and interaction magnitude, while position-control policies cannot respond compliantly to rapidly changing contact dynamics. To bridge both the perception and control gaps, we propose TAO-Force, a force-conditioned VLA framework that combines force-aware policy learning with contact-regulated execution. For force-aware perception, TAO-Force introduces Force-conditioned Feature-wise Linear Modulation (F-FiLM) to inject encoded force feedback into the representations of a frozen pretrained visual-language backbone while preserving its semantic priors. For responsive control, it employs a contact-gated fast-slow architecture, with a slow position-control branch tracking nominal trajectories during non-contact phases and a fast admittance-control branch regulating physical interaction during contact phases. Detailed analyses on a force-perception task and real-world evaluations across four contact-rich manipulation tasks validate the effectiveness and robustness of TAO-Force.
Chinese Translation
视觉-语言-动作（VLA）模型在多种机器人操作任务中表现出色，但其主要以视觉为中心的感知和位置控制执行在接触丰富的操作中仍显不足。仅凭视觉观察往往无法提供接触开始和交互强度的充分证据，而位置控制策略无法对快速变化的接触动态做出灵活响应。为了解决感知与控制之间的差距，我们提出了TAO-Force，一个基于力条件的VLA框架，结合了力感知策略学习与接触调节执行。在力感知方面，TAO-Force引入了力条件特征线性调制（F-FiLM），将编码的力反馈注入到冻结的预训练视觉-语言骨干网络的表示中，同时保留其语义先验。在响应控制方面，它采用了接触门控的快慢架构，在非接触阶段，慢速位置控制分支跟踪名义轨迹，而在接触阶段，快速适应控制分支调节物理交互。对力感知任务的详细分析以及在四个接触丰富操作任务中的实际评估验证了TAO-Force的有效性和鲁棒性。
cs.RO / 62 / 2609.18504
InterMASH: A Unified Geometric Representation for Grasp Synthesis
InterMASH：一种统一的几何表示用于抓取合成
Abstract
Grasp synthesis aims to generate stable and physically plausible hand--object interactions, and has become a fundamental problem in both human hand modeling and robotic manipulation. However, a unified representation across human and robotic hands is still lacking, mainly due to differences in hand morphology and surface modeling. Prior methods typically rely on either contact maps or dense implicit descriptors to represent interaction, but these representations are often incomplete or computationally expensive and redundant. We propose InterMASH, a unified geometric representation that establishes cross-embodiment correspondence using sphere-fixed anchors. At each anchor, low-degree spherical harmonics compactly encode local hand geometry, object geometry, and contact, forming an explicit and interpretable token sequence. Building on this natively tokenized structure, we introduce a conditional Diffusion Transformer that operates directly in the proposed InterMASH representation space and jointly generates hand geometry and contact, improving consistency and physical plausibility. Our method achieves competitive performance with state-of-the-art methods on key physical feasibility metrics in a large-scale ShadowHand benchmark, supports joint training across multiple hands, and shows that cross-embodiment fine-tuning with human grasp data can improve robotic grasp success and diversity. Project page is available at https://inter-mash.github.io/.
Chinese Translation
抓取合成旨在生成稳定且物理上合理的手-物体交互，已成为人手建模和机器人操作中的一个基本问题。然而，缺乏一种跨人类和机器人手的统一表示，主要是由于手的形态和表面建模的差异。以往的方法通常依赖于接触图或密集隐式描述符来表示交互，但这些表示往往不完整或计算成本高且冗余。我们提出了InterMASH，一种统一的几何表示，通过固定在球体上的锚点建立跨体现的对应关系。在每个锚点，低阶球面谐波紧凑地编码局部手几何、物体几何和接触，形成明确且可解释的标记序列。在这一原生标记化结构的基础上，我们引入了一种条件扩散变换器（Diffusion Transformer），该变换器直接在所提出的InterMASH表示空间中操作，并联合生成手几何和接触，从而提高一致性和物理合理性。我们的方法在大规模ShadowHand基准测试中，在关键物理可行性指标上实现了与最先进方法的竞争性能，支持多个手的联合训练，并表明通过人类抓取数据进行跨体现微调可以提高机器人抓取的成功率和多样性。项目页面可访问 https://inter-mash.github.io/。
cs.RO / 63 / 2609.18514
ActiveScale: Scaling Active Perception for Robots across Model, Data, and Hardware
ActiveScale：跨模型、数据和硬件扩展机器人主动感知
Abstract
Active perception is essential for robotic manipulation when fixed viewpoints leave task-relevant information occluded or unobserved. However, enabling vision-language-action (VLA) models to reason across changing viewpoints and actively acquire informative observations remains challenging. We present ActiveScale, a framework that advances active perception through coordinated model, data, and hardware designs. Our model augments a VLA with historical video observations and explicit camera-pose supervision, using per-frame pose tokens and a lightweight prediction head to associate observations across viewpoints and support a coherent understanding of the scene. To learn from the camera motion naturally present in human activity, we introduce a scalable human--robot mid-training recipe using 1000 hours of egocentric and robotic data, adapting the model to temporal inputs and pose supervision. We further introduce Active-perception Mobile-manipulation Platform (AMP), a robotic platform that supports active perception and mobile manipulation through single-operator teleoperation, enabling scalable collection of demonstrations that coordinate viewpoint changes and manipulation. Experiments demonstrate improved success rates on active-perception tasks, while ablation studies validate the contributions of camera-pose-aware modeling and egocentric mid-training. Together, these components provide an integrated foundation for studying and developing active perception in robotic manipulation.
Chinese Translation
主动感知对于机器人操作至关重要，因为固定的视点可能会遮挡或未观察到与任务相关的信息。然而，使视觉-语言-动作（VLA）模型能够在变化的视点之间进行推理并主动获取有用的观察仍然具有挑战性。我们提出了ActiveScale，一个通过协调模型、数据和硬件设计来推进主动感知的框架。我们的模型通过历史视频观察和显式相机姿态监督来增强VLA，使用每帧姿态标记和轻量级预测头来关联不同视点的观察，并支持对场景的连贯理解。为了自然地学习人类活动中存在的相机运动，我们引入了一种可扩展的人机中期训练方案，使用1000小时的自我中心和机器人数据，使模型适应时间输入和姿态监督。我们进一步推出了主动感知移动操作平台（AMP），这是一个支持主动感知和移动操作的机器人平台，通过单一操作员的远程操作，能够规模化收集协调视点变化和操作的演示。实验表明，在主动感知任务上的成功率有所提高，而消融研究验证了相机姿态感知建模和自我中心中期训练的贡献。这些组件共同为研究和开发机器人操作中的主动感知提供了一个综合基础。
cs.RO / 64 / 2609.18549
DynoFluxBench: Benchmarking Kinodynamic Space-Time Planners in Dynamic Environments
DynoFluxBench：在动态环境中对运动动力学时空规划器进行基准测试
Abstract
Robots that leave structured, static environments must plan motions that are kinodynamically feasible and safe among moving obstacles. However, there are no dedicated benchmark frameworks that combine both aspects. To overcome this, we present DynoFluxBench, a framework to compare kinodynamic planners in known, dynamic environments with unbounded arrival time. To demonstrate its utility and establish strong baselines, we develop three dedicated planners, named ST-Db-RRT, ST-GBRRT, and KIST, that fuse kinodynamic and space-time methods, covering different kinodynamic search paradigms: ST-Db-RRT expands with randomly selected discontinuity-bounded motion primitives using trajectory optimization, whereas KIST and ST-GBRRT maintain a kinodynamically feasible tree with different heuristic guidance. We analyze the probabilistic completeness guarantees of those new planners in dynamic environments. Finally, we evaluate ST-Db-RRT, ST-GBRRT, and KIST using DynoFluxBench, showing that ST-Db-RRT reaches a first solution up to 32 times faster, while KIST and ST-GBRRT remain valuable where trajectory optimization is fragile. Videos and further analysis can be found at https://dynofluxbench.github.io/dynofluxbench/.
Chinese Translation
离开结构化静态环境的机器人必须在移动障碍物之间规划运动，这些运动在运动动力学上是可行且安全的。然而，目前尚无专门的基准框架将这两个方面结合起来。为了解决这个问题，我们提出了DynoFluxBench，这是一个用于比较已知动态环境中具有无限到达时间的运动动力学规划器的框架。为了展示其实用性并建立强有力的基准，我们开发了三个专门的规划器，分别命名为ST-Db-RRT、ST-GBRRT和KIST，它们融合了运动动力学和时空方法，涵盖不同的运动动力学搜索范式：ST-Db-RRT通过轨迹优化扩展随机选择的具有不连续性限制的运动原语，而KIST和ST-GBRRT则在不同的启发式指导下维护一个运动动力学上可行的树。我们分析了这些新规划器在动态环境中的概率完整性保证。最后，我们使用DynoFluxBench评估了ST-Db-RRT、ST-GBRRT和KIST，结果表明ST-Db-RRT的首次解决方案速度快达32倍，而KIST和ST-GBRRT在轨迹优化脆弱的情况下仍然具有重要价值。视频和进一步的分析可以在https://dynofluxbench.github.io/dynofluxbench/找到。
cs.RO / 65 / 2609.18581
GroundingVLN: Reasoning and Acting with Grounding for Vision-Language Navigation
GroundingVLN：基于视觉-语言导航的推理与行动的基础
Abstract
Although vision-language models (VLMs) possess strong visual understanding and reasoning capabilities, existing vision-and-language navigation (VLN) agents struggle to connect semantic reasoning with spatial execution. Two coupled gaps remain in this connection, as intermediate reasoning is not explicitly anchored to visual evidence and high-level decisions lack precise spatial goals to guide low-level motion. Cognitive science suggests that human navigation bridges these levels hierarchically by anchoring cognition to relevant landmarks and guiding locomotion toward spatial goals. Motivated by this principle, we propose GroundingVLN, which uses visual grounding as a shared interface between reasoning and action. GroundingVLN first reasons with grounding by anchoring task-relevant visual evidence to precise image locations throughout structured reasoning. It then acts through grounding by predicting a progress-aligned pixel goal that a geometric planner translates into primitive actions. To learn these capabilities, we construct GroundingCOTVLN-188K, a dataset of temporally aligned grounded reasoning traces, and introduce Grounded and Execution-Aware Reinforcement Learning (GEAR), which aligns grounded reasoning and spatial decisions with downstream execution. Experiments demonstrate that GroundingVLN achieves state-of-the-art performance (69.9% SR on R2R-CE and 75.1% SR on RxR-CE) with high sample efficiency, using just 0.9% as much training data as the strongest baseline. It also generalizes strongly across datasets, attaining 59.9% SR on RxR-CE when trained solely on R2R, a gain of 20.1% over the strongest baseline.
Chinese Translation
尽管视觉-语言模型（VLMs）具备强大的视觉理解和推理能力，但现有的视觉与语言导航（VLN）代理在将语义推理与空间执行连接起来时仍面临困难。这一连接中存在两个相互关联的缺口：中间推理未能明确锚定于视觉证据，而高层决策缺乏精确的空间目标来指导低层运动。认知科学表明，人类导航通过将认知锚定于相关地标并引导运动朝向空间目标，层次性地弥合了这些层次。基于这一原则，我们提出了GroundingVLN，它利用视觉锚定作为推理与行动之间的共享接口。GroundingVLN首先通过锚定与任务相关的视觉证据到精确的图像位置进行推理。然后，它通过预测一个与进度对齐的像素目标来进行行动，该目标由几何规划器转换为原始动作。为了学习这些能力，我们构建了GroundingCOTVLN-188K，这是一个时间上对齐的锚定推理轨迹数据集，并引入了锚定与执行感知强化学习（GEAR），该方法将锚定推理和空间决策与下游执行对齐。实验表明，GroundingVLN在样本效率高的情况下达到了最先进的性能（在R2R-CE上69.9%的成功率和在RxR-CE上75.1%的成功率），仅使用了最强基线的0.9%的训练数据。它在不同数据集上也表现出强大的泛化能力，在仅使用R2R训练时，在RxR-CE上达到了59.9%的成功率，比最强基线提高了20.1%。
cs.RO / 66 / 2609.18620
DeformSmith: Physics Harness-Guided Hierarchical Generation of Deformable Assets for Robot Manipulation
DeformSmith：物理引导的可变形资产层次生成框架用于机器人操作
Abstract
Creating deformable assets for robot manipulation requires jointly specifying their geometry, appearance, and physical properties. This is especially challenging for deformable objects, since text and images provide limited evidence about how they deform and respond to contact, yet these responses directly affect their suitability for interaction. Automated generation therefore needs to resolve coupled physical requirements and use interaction evidence to guide construction and refinement. We present DeformSmith, a framework that enables automated generation of interactive, physically credible deformable assets from text or a single image. Through hierarchical agentic construction and a shared physics-grounded harness, it progressively builds, tests, and refines geometry, physical models, material behavior, and robot interaction until the resulting asset is ready for simulation and manipulation. Robot interaction closes the generation loop through manipulation feedback and replayable interaction data. Results show that DeformSmith generates assets with better visual quality and physical plausibility than state-of-the-art baselines, including PhysGen3D, PhysGM, and PhysX-Omni, while supporting the synthesis of data for robotic manipulation of deformable objects. Project page: https://can-lee.github.io/deformsmith-web/
Chinese Translation
为机器人操作创建可变形资产需要共同指定其几何形状、外观和物理属性。这对于可变形物体尤其具有挑战性，因为文本和图像提供的关于它们如何变形和对接触的响应的证据有限，而这些响应直接影响它们的交互适用性。因此，自动生成需要解决耦合的物理要求，并利用交互证据来指导构建和优化。我们提出了DeformSmith，一个能够从文本或单幅图像自动生成交互式、物理可信的可变形资产的框架。通过层次化的自主构建和共享的物理基础装置，它逐步构建、测试和优化几何形状、物理模型、材料行为和机器人交互，直到生成的资产准备好进行仿真和操作。机器人交互通过操作反馈和可重放的交互数据闭合生成循环。结果表明，DeformSmith生成的资产在视觉质量和物理合理性上优于最先进的基线，包括PhysGen3D、PhysGM和PhysX-Omni，同时支持可变形物体的机器人操作数据合成。项目页面：https://can-lee.github.io/deformsmith-web/
cs.RO / 67 / 2609.18628
Benchmarking Visual-Inertial Odometry in Subterranean Environments Under Sensor Degradation, Miscalibration, and Dynamic Occlusion
在传感器退化、误校准和动态遮挡下的地下环境中视觉惯性里程计的基准测试
Abstract
Visual-inertial odometry (VIO) is a core capability for autonomous operation in GPS-denied subterranean environments, yet its reliability can degrade sharply under sensor drift, calibration errors, and dynamic occlusion. Existing evaluations mainly emphasize nominal-condition accuracy, offering limited insight into when practical deployment failures occur. In this work, we present a failure-centric stress-test benchmark for VIO in underground environments using the CERBERUS dataset. We systematically evaluate four representative VIO systems spanning filtering-, optimization-, and learning-based paradigms under nine practical perturbation settings, including IMU bias and noise variation, camera intrinsic and extrinsic drift, and dynamic scene occlusion. Beyond conventional trajectory error, we analyze robustness limits through coverage ratio and failure thresholds, revealing breakdown behaviors that are not captured by nominal-condition performance alone. Our study shows distinct vulnerability patterns across VIO paradigms: some methods are more sensitive to inertial degradation, while others are more affected by geometric miscalibration or dynamic interference. These results provide deployment-oriented guidance for VIO selection, calibration prioritization, and reliable operation in challenging underground scenarios. To support reproducible evaluation and future extensions, we will release the full benchmark scripts and evaluation pipeline.
Chinese Translation
视觉惯性里程计（VIO）是GPS无法使用的地下环境中自主操作的核心能力，但在传感器漂移、校准错误和动态遮挡的情况下，其可靠性可能会急剧下降。现有评估主要强调名义条件下的准确性，提供了有限的洞察力来了解实际部署失败发生的情况。在本研究中，我们利用CERBERUS数据集提出了一种以失败为中心的地下环境VIO压力测试基准。我们系统地评估了四种代表性的VIO系统，这些系统涵盖了基于滤波、优化和学习的范式，测试了九种实际扰动设置，包括IMU偏差和噪声变化、相机内外参数漂移以及动态场景遮挡。除了传统的轨迹误差外，我们还通过覆盖率和失败阈值分析了鲁棒性极限，揭示了仅通过名义条件性能无法捕捉到的故障行为。我们的研究显示了不同VIO范式之间明显的脆弱性模式：某些方法对惯性退化更为敏感，而其他方法则更容易受到几何误校准或动态干扰的影响。这些结果为VIO的选择、校准优先级和在具有挑战性的地下场景中可靠操作提供了以部署为导向的指导。为了支持可重复的评估和未来的扩展，我们将发布完整的基准脚本和评估流程。
cs.RO / 68 / 2609.18650
From Gameplay to Policy: Towards Scalable Robot Data Collection via Gamified Robot-Free Interaction
从游戏玩法到政策：通过游戏化的无机器人互动实现可扩展的机器人数据收集
Abstract
Learning generalizable robot manipulation policies requires large-scale and diverse interaction data, yet collecting real-world demonstrations remains costly and difficult to scale. Existing approaches to data collection are either dependent on specific robot hardware that limits crowdsourcing and transferability, or suffer from incomplete annotation and limited behavioral diversity. Inspired by how games sustain long-term human engagement, we explore an alternative paradigm that turns data collection into an engaging gameplay experience and transfers the resulting human manipulation experience to real robots. We present Project Kitchen, a VR-based gamified egocentric data collection platform that elicits diverse, goal-directed manipulation while remaining independent of specific robot embodiments and hardware, making it applicable to broader and potentially large-scale deployment. To bridge the game-to-real gap, we further introduce Game2Policy, which extracts embodiment-invariant affordance cues, including contact points and sub-goal states, from gameplay trajectories. An affordance model is pre-trained on game-collected data and then jointly fine-tuned with downstream policies using only a handful of real-robot demonstrations. Experiments show that Game2Policy improves average success rates by 10.0 points in simulation and 18.3 points on real robots in the few-shot setting. User studies and quantitative analyses further show that Project Kitchen promotes diverse manipulation behaviors and provides an engaging data collection experience. These results demonstrate the potential of gamified virtual environments as a scalable source of manipulation knowledge. The platform and code will be released upon acceptance.
Chinese Translation
学习可推广的机器人操作策略需要大规模和多样化的互动数据，但收集真实世界的示范仍然成本高昂且难以扩展。现有的数据收集方法要么依赖于特定的机器人硬件，这限制了众包和可转移性，要么面临注释不完整和行为多样性有限的问题。受到游戏如何维持长期人类参与的启发，我们探索了一种将数据收集转变为引人入胜的游戏体验的替代范式，并将由此产生的人类操作经验转移到真实机器人上。我们提出了Project Kitchen，这是一个基于虚拟现实的游戏化自我中心数据收集平台，能够引发多样化的目标导向操作，同时独立于特定的机器人表现和硬件，使其适用于更广泛且潜在的大规模部署。为了弥合游戏与现实之间的差距，我们进一步引入了Game2Policy，该方法从游戏轨迹中提取与表现无关的可用性线索，包括接触点和子目标状态。可用性模型在游戏收集的数据上进行了预训练，然后仅使用少量真实机器人示范与下游策略共同微调。实验表明，在少样本设置中，Game2Policy在仿真中的平均成功率提高了10.0个百分点，在真实机器人上的成功率提高了18.3个百分点。用户研究和定量分析进一步表明，Project Kitchen促进了多样化的操作行为，并提供了引人入胜的数据收集体验。这些结果展示了游戏化虚拟环境作为可扩展操作知识来源的潜力。该平台和代码将在接受后发布。
cs.RO / 69 / 2609.18651
FIERCE: From Generalist Robot Policies to Fast Specialists via Progress-Failure Feedback
FIERCE：通过进展-失败反馈从通用机器人策略到快速专业化
Abstract
Generalist robot policies offer useful initialization, but refining compact specialists through limited physical interaction requires informative learning feedback. We present FIERCE, a generalist-initialized reinforcement learning framework centered on a unified, task-adaptive progress-failure evaluator. Its architecture shares an observation-language representation between an observed-progress head and an action-conditioned latent predictor whose past and current predictions feed a causal sequence head for task-failure estimation. Joint supervision from progress and preference labels, synchronized commands and observations, and terminal outcomes trains the evaluator; target-task rollouts support adaptation and calibration. Fixed evaluator snapshots provide progress shaping and failure-risk penalties alongside independently verified terminal rewards, while evaluator and policy updates alternate as new experience is collected. Refinement requires neither continued generalist action queries nor a dedicated target-task simulator or manually annotated dense rewards. Only the compact specialist is retained at deployment. The evaluation separates feedback quality, policy-learning efficiency, and deployment cost across simulation and two contact-rich real tasks. Code, model weights, and data-restoration tools are released at https://github.com/ar-mine/FIERCE.
Chinese Translation
通用机器人策略提供了有用的初始化，但通过有限的物理交互来精炼紧凑的专业化模型需要信息丰富的学习反馈。我们提出了FIERCE，一个以通用初始化为基础的强化学习框架，集中于一个统一的、任务自适应的进展-失败评估器。其架构在观察进展头和一个基于动作的潜在预测器之间共享观察-语言表示，潜在预测器的过去和当前预测为任务失败估计提供因果序列头。来自进展和偏好标签的联合监督、同步的命令和观察以及终端结果共同训练评估器；目标任务的回放支持适应和校准。固定的评估器快照提供进展塑造和失败风险惩罚，同时独立验证的终端奖励也被纳入其中，而评估器和策略更新在收集新经验时交替进行。精炼过程既不需要持续的通用动作查询，也不需要专用的目标任务模拟器或手动标注的密集奖励。仅在部署时保留紧凑的专业化模型。评估在模拟和两个接触丰富的真实任务中分离了反馈质量、策略学习效率和部署成本。代码、模型权重和数据恢复工具已在 https://github.com/ar-mine/FIERCE 发布。
cs.RO / 70 / 2609.18663
VLA-ULAP: Interleaving Cloud VLA Calls with Ultra-Lightweight Local Action Prediction at the Edge
VLA-ULAP：在边缘将云端VLA调用与超轻量级本地动作预测交错
Abstract
Billion-parameter vision--language--action (VLA) policies demand substantial onboard power, while communication delays in remote inference hinder timely responses. We propose VLA-ULAP, which interleaves remote VLA calls with an Ultra-Lightweight Local Action Predictor (ULAP). With approximately 7.4M parameters including the frozen vision encoder, ULAP combines current views, proprioception, and executed action history to predict chunks in one pass. Trained independently, it requires no VLA hidden states, online verification, or server round trips. On Jetson Orin Nano, ULAP takes 19.9 ms and 0.183 J per inference, compared with 284.3 ms and 50.55 J for GR00T on RTX A6000. Across three simulated base-policy/benchmark pairs, selected operating points remove 48.8--76.7\% of VLA calls while retaining 95.0--97.5\% of the baseline success rate. Against local VLA-acceleration alternatives on VLA-JEPA, ULAP uses an estimated 49.2\% less inference time and 51.0\% less GPU energy per successful episode than ACT at comparable success rates, and 77.1\% less time and 79.9\% less energy than SP-VLA at equal success rates. Physical SO-101 experiments retain 95.2--100\% of the baseline success rate across seen and held-out placements while reducing inference time by an estimated 47.9--58.0\% and inference-device energy by 52.1--62.5\%, based on successful-episode call counts and measured device costs. Faster responses also improve dynamic-task success rates: in latency-aware LIBERO-Safety simulation, VLA-ULAP exceeds $\pi_{0.5}$ by 11.0 and 15.5 percentage points on two tasks while approximately halving VLA calls.
Chinese Translation
十亿参数的视觉-语言-动作（VLA）策略需要大量的机载功率，而远程推理中的通信延迟则阻碍了及时响应。我们提出了VLA-ULAP，它将远程VLA调用与超轻量级本地动作预测器（ULAP）交错。ULAP的参数约为740万，包括冻结的视觉编码器，结合当前视图、自我感知和执行的动作历史来进行一次性预测。ULAP独立训练，不需要VLA的隐藏状态、在线验证或服务器往返。在Jetson Orin Nano上，ULAP每次推理耗时19.9毫秒，能耗为0.183焦耳，而在RTX A6000上，GR00T的推理时间为284.3毫秒，能耗为50.55焦耳。在三个模拟的基础策略/基准对中，所选操作点去除了48.8%至76.7%的VLA调用，同时保留了95.0%至97.5%的基线成功率。与VLA-JEPA上的本地VLA加速替代方案相比，ULAP在成功率相当的情况下，每个成功回合的推理时间比ACT减少了约49.2%，GPU能耗减少了51.0%，并且比SP-VLA在相同成功率下减少了77.1%的时间和79.9%的能量。在物理SO-101实验中，在已见和保留的放置中，成功率保持在95.2%至100%之间，同时推理时间减少了估计的47.9%至58.0%，推理设备能耗减少了52.1%至62.5%，基于成功回合的调用次数和测量的设备成本。更快的响应也提高了动态任务的成功率：在延迟感知的LIBERO-Safety仿真中，VLA-ULAP在两个任务上超过了$ ext{π}_{0.5}$，分别提高了11.0和15.5个百分点，同时大约减少了VLA调用次数。
cs.RO / 71 / 2609.18669
M$^3$P-R1: Reinforcement Learning for Large Language Model Guided Multi-Modal Motion Planning via MIP Code Generation
M$^3$P-R1：通过 MIP 代码生成的强化学习用于大型语言模型指导的多模态运动规划
Abstract
Multi-Modal Motion Planning (M$^3$P) requires joint reasoning over continuous motions and discrete mode transitions, making it difficult to solve efficiently. For instance, a bipedal robot may walk to a target location and then use its arms to grasp an object. This scenario captures both mode transitions and continuous dynamics, yielding feasible paths that neither purely discrete nor continuous planners can handle. While Mixed-Integer Programming (MIP) offers a principled framework, constructing tractable formulations for non-convex problems is typically manual and domain-specific, especially in the approximate, discretization-based MIP regime needed for non-convex robotic tasks. We propose M$^3$P-R1, a reinforcement learning method that fine-tunes large language models (LLMs) to decompose M$^3$P tasks into MIP variables, constraints, and objectives. Instead of directly outputting answers, which are often prone to hallucination, the model generates executable Python code using MIP optimization libraries and constraint interfaces. This enables solver-backed execution for robust and verifiable solutions. Trained with an outcome-driven reward against the solver, M$^3$P-R1 learns to compose modality-level discretization primitives and synthesize cross-modal coupling constraints, producing executable MIP programs for complex M$^3$P tasks.
Chinese Translation
多模态运动规划（M$^3$P）需要对连续运动和离散模式转换进行联合推理，这使得高效求解变得困难。例如，一个双足机器人可能会走到目标位置，然后用手臂抓取物体。这个场景同时捕捉了模式转换和连续动态，产生的可行路径既不是纯粹的离散规划者能够处理的，也不是连续规划者能够处理的。虽然混合整数规划（MIP）提供了一个原则性的框架，但为非凸问题构建可处理的公式通常是手动且特定于领域的，尤其是在非凸机器人任务所需的基于近似和离散化的 MIP 范畴中。我们提出了 M$^3$P-R1，这是一种强化学习方法，旨在微调大型语言模型（LLMs），将 M$^3$P 任务分解为 MIP 变量、约束和目标。该模型不是直接输出答案（这些答案往往容易产生幻觉），而是使用 MIP 优化库和约束接口生成可执行的 Python 代码。这使得基于求解器的执行成为可能，从而提供稳健且可验证的解决方案。通过与求解器的结果驱动奖励进行训练，M$^3$P-R1 学会了组合模态级离散化原语并合成跨模态耦合约束，为复杂的 M$^3$P 任务生成可执行的 MIP 程序。
cs.RO / 72 / 2609.18685
WeaveRL: Weaving Reconstruction into Scene-Aware Fabrics for Perceptive Reinforcement Learning
WeaveRL：将重建融入场景感知织物以实现感知强化学习
Abstract
Reinforcement learning allows robots to acquire complex skills, but producing policies for geometrically complex manipulation remains difficult. A promising approach is to learn on top of collision-avoidant controllers, such as geometric fabrics. However, these approaches have relied on static, hand-specified representations of the scene. Integrating active, online 3D perception into massively parallel RL training has so far been inaccessible. We introduce a GPU-accelerated method that reconstructs the scene as a collection of surfels across thousands of parallel simulation instances during active rollouts. This lets policies operate over sensor-derived, rather than hand-specified, geometry. On a suite of collision-dense manipulation tasks, our surfel fabrics enable policies to tackle geometrically complex scenes where primitive-based baselines fail, while maintaining sim-to-real transfer. Furthermore, policies learned with a scene-aware fabric are more robust to the introduction of novel geometry at test time, improving collision-free task completion under unseen obstacles from 35% to 61%. We release our reconstruction system, training code and test dataset to spur research in this direction.
Chinese Translation
强化学习使机器人能够掌握复杂技能，但为几何复杂的操作生成策略仍然困难。一种有前景的方法是在避免碰撞的控制器（如几何织物）之上进行学习。然而，这些方法依赖于静态的、手动指定的场景表示。将主动的、在线的三维感知集成到大规模并行的强化学习训练中迄今为止仍然不可行。我们提出了一种基于GPU加速的方法，在主动回放期间，通过数千个并行仿真实例将场景重建为一组表面粒子（surfels）。这使得策略能够在传感器导出的几何形状上操作，而不是依赖手动指定的几何形状。在一系列碰撞密集的操作任务中，我们的表面粒子织物使得策略能够处理几何复杂的场景，而在基于原始形状的基线方法中失败，同时保持仿真到现实的迁移。此外，使用场景感知织物学习的策略对测试时新几何形状的引入更具鲁棒性，使得在未见障碍物下的无碰撞任务完成率从35%提高到61%。我们发布了我们的重建系统、训练代码和测试数据集，以促进该方向的研究。
cs.RO / 73 / 2609.18700
Toward 3D Printable Non-Planar Electroadhesive Structures for Active Anchoring
朝向可3D打印的非平面电粘附结构用于主动锚定
Abstract
This paper investigates multi-material 3D printing as a method to fabricate non-planar structures with 3D-printed electrode patterns for electroadhesion. We printed flat electroadhesion pads as a planar benchmark and cylindrical pads as a non-planar demonstration, with conductive interdigitated electrodes 3D printed as part of the structure. Normal-force measurements showed voltage-controlled modulation in both geometries. At 3 kV, the flat pads generated approximately 0.11-0.13 N, while the cylindrical pads reached approximately 0.16 N at 5 N preload. These results demonstrate a step toward printed parts with built-in, electrically controlled adhesion and friction, beyond conventional planar electroadhesion pads.
Chinese Translation
本文研究了多材料3D打印作为制造具有电粘附功能的非平面结构的方法，采用3D打印的电极图案。我们打印了平面电粘附垫作为平面基准，并打印了圆柱形垫作为非平面示范，结构中包含了导电的交错电极。正常力测量显示两种几何形状均表现出电压控制的调制。在3 kV时，平面垫产生了约0.11-0.13 N的力，而圆柱形垫在5 N预载下达到了约0.16 N。这些结果展示了向具有内置电控粘附和摩擦的打印部件迈出了一步，超越了传统的平面电粘附垫。
cs.RO / 74 / 2609.18718
Calibrated Probabilistic Obstruction Reasoning with Vision-Language Models for Grasping in Clutter
基于视觉-语言模型的校准概率障碍推理在杂乱环境中的抓取应用
Abstract
Retrieving a target from clutter requires deciding whether to grasp the target, remove a blocker, or defer. Existing methods typically commit to a single obstruction graph or removal strategy, ignoring uncertainty across alternative scene interpretations. They also rely on miscalibrated vision-language model (VLM) predictions and can produce pairwise obstruction relations that are jointly inconsistent. Moreover, current approximations provide no guarantees about the impact of discarded hypotheses on the final decision. We propose CPOR-Grasp, a calibrated probabilistic obstruction-reasoning framework that propagates uncertainty from pairwise evidence to action decisions. CPOR-Grasp calibrates and fuses VLM, depth, and amodal-mask cues to estimate obstruction probabilities, induces a distribution over valid obstruction graphs, and marginalizes over these graphs to compute the likelihood that the target is accessible or that a given blocker should be removed. To make inference tractable, it retains only the highest-probability graphs and derives a total-variation bound on the discarded probability mass, enabling certified decisions, adaptive stopping, and principled deferral. On synthetic and real UNOBench scenes, CPOR-Grasp outperforms state-of-the-art baselines. Calibration error decreases from 0.1416 to 0.0185 on the Gemini Robotics backbone, while graph truncation matches exact inference on 99.74\% of decisions using 56 times fewer graphs. In real-world experiments, CPOR-Grasp achieves a 77.8\% average success rate, surpassing SOTA baselines.
Chinese Translation
从杂乱中检索目标需要决定是抓取目标、移除障碍物还是延迟处理。现有方法通常只承诺于单一的障碍图或移除策略，忽视了对替代场景解释的不确定性。它们还依赖于未校准的视觉-语言模型（VLM）预测，并可能产生相互不一致的成对障碍关系。此外，目前的近似方法无法保证被丢弃的假设对最终决策的影响。我们提出了CPOR-Grasp，这是一种校准的概率障碍推理框架，能够将成对证据中的不确定性传播到行动决策中。CPOR-Grasp校准并融合VLM、深度和模态遮罩线索，以估计障碍概率，诱导有效障碍图的分布，并对这些图进行边际化，以计算目标可达性或给定障碍物应被移除的可能性。为了使推理可行，它仅保留最高概率的图，并推导出被丢弃概率质量的总变差界限，从而实现有保障的决策、自适应停止和原则性延迟。在合成和真实的UNOBench场景中，CPOR-Grasp的表现优于最先进的基线。在Gemini Robotics骨干网络上，校准误差从0.1416降低到0.0185，而图的截断在99.74%的决策中与精确推理相匹配，使用的图数量减少了56倍。在真实世界实验中，CPOR-Grasp实现了77.8%的平均成功率，超越了最先进的基线。
cs.RO / 75 / 2609.18732
PASSAGE: Scaling Scene-Aligned Motion Learning for Perceptive Humanoid Traversal in Cluttered Environments
PASSAGE：在杂乱环境中实现感知类人机器人穿越的场景对齐运动学习的扩展
Abstract
Humanoid robots can step over, squeeze past, and duck under obstacles, but learning to select and coordinate these behaviors from onboard perception remains challenging. Many existing approaches rely on task-specific reinforcement-learning objectives or curated motion libraries, making broad behavioral coverage costly. We present PASSAGE, a perception-conditioned planner--tracker framework for humanoid traversal. Using virtual reality and inertial motion capture, we collect 100 h of scene-aligned human motion across 1,500 cluttered scenes. A conditional flow-matching planner generates short-horizon references from motion history, a local destination, and a robot-centric multi-layer elevation map, while a perceptive whole-body tracker executes them at 50 Hz with geometric feedback. Real-time chunking promotes inter-chunk consistency, and planner-side RL post-training under the frozen tracker further improves closed-loop performance. Without skill annotations or obstacle-specific policies, one planner--tracker pair selects and composes traversal behaviors across unseen geometries. In simulation, component ablations quantify the contribution of each stage. Across three independent training seeds, scaling captured data from 6 to 100 h increases mean contact-free success from 48.1% to 68.9% on held-out scenes, while the final model with validated scene augmentation reaches 70.3%. The fully onboard system integrates egocentric 3D LiDAR perception, online occupancy mapping, 6.25 Hz planning, and 50 Hz control on a Jetson AGX Orin; tests across 50 unseen physical layouts demonstrate traversal without prebuilt maps or offboard computation.
Chinese Translation
类人机器人能够跨越、挤过和低头躲避障碍物，但从机载感知中学习选择和协调这些行为仍然具有挑战性。许多现有方法依赖于特定任务的强化学习目标或精心策划的运动库，使得广泛的行为覆盖成本高昂。我们提出了PASSAGE，一个用于类人机器人穿越的感知条件规划-跟踪框架。通过虚拟现实和惯性运动捕捉，我们在1500个杂乱场景中收集了100小时的场景对齐人类运动。条件流匹配规划器根据运动历史、局部目标和以机器人为中心的多层高程图生成短期参考，而感知全身跟踪器以50 Hz的频率执行这些参考，并提供几何反馈。实时分块促进了块间一致性，而在冻结的跟踪器下进行的规划侧强化学习后训练进一步提高了闭环性能。在没有技能注释或特定障碍物策略的情况下，一个规划-跟踪对能够在未见几何体中选择和组合穿越行为。在仿真中，组件消融实验量化了每个阶段的贡献。在三个独立的训练种子中，将捕获数据从6小时扩展到100小时使得在保留场景上的平均无接触成功率从48.1%提高到68.9%，而经过验证的场景增强的最终模型达到了70.3%。该完全机载系统集成了自我中心的3D LiDAR感知、在线占用映射、6.25 Hz规划和50 Hz控制，运行在Jetson AGX Orin上；在50个未见物理布局上的测试证明了在没有预构建地图或离线计算的情况下实现穿越。
cs.RO / 76 / 2609.18738
Active perception for robotic harvesting: 3D reconstruction and localisation of tomatoes hidden within clusters in a Mediterranean greenhouse
机器人采摘的主动感知：地中海温室中隐藏在簇中的番茄的三维重建与定位
Abstract
Automating robotic harvesting in intensive agriculture within Mediterranean greenhouses requires overcoming significant challenges related to the geometric complexity of plants and occluded fruits. Although existing literature offers solutions targeting crops that grow in isolation (e.g., apples, sweet peppers, or peaches), the fundamental challenge lies in cluster-growing vegetables, where fixed sensors mounted on robotic systems fail to detect fruits hidden behind the visible surface. To address this limitation, this study presents a comprehensive pipeline for the 3D reconstruction and precise localization of each fruit within a cluster, including heavily occluded instances. The proposed methodology is structured into five sequential stages: i) point cloud acquisition using the AgriSEE Next Best View (NBV) active planner; ii) stochastic noise filtering via Statistical Outlier Removal (SOR); iii) surface classification and segmentation using Region Growing (RG); iv) isolation and recovery of occluded fruits through Density-Based Spatial Clustering of Applications with Noise (DBSCAN); and v) 3D pose estimation (position and orientation). This approach extracts the complete cluster geometry, ensuring the reliable identification of partially hidden tomatoes. Evaluated across multiple scenarios with varying occlusion levels within a simulation framework rigorously validated against real-world conditions, the system achieves a precision exceeding 90\%, an average recall of 82.8\%, and a mean Intersection over Union (mIoU) of 80.7\%. Furthermore, it demonstrates high repeatability in centroid estimation with a Root Mean Square Error (RMSE) of merely 4.2~mm, verifying its technical feasibility and high accuracy for autonomous harvesting operations.
Chinese Translation
在地中海温室中实现集约农业的机器人采摘自动化，需要克服与植物的几何复杂性和被遮挡水果相关的重大挑战。尽管现有文献提供了针对孤立生长作物（例如苹果、甜椒或桃子）的解决方案，但根本挑战在于簇生长的蔬菜，其中固定传感器无法检测到隐藏在可见表面后面的水果。为了解决这一局限性，本研究提出了一种全面的管道，用于对簇中每个水果进行三维重建和精确定位，包括严重遮挡的实例。所提出的方法论分为五个顺序阶段：i）使用 AgriSEE Next Best View (NBV) 主动规划器进行点云采集；ii）通过统计离群点去除 (SOR) 进行随机噪声过滤；iii）使用区域生长 (RG) 进行表面分类和分割；iv）通过基于密度的空间聚类应用与噪声 (DBSCAN) 隔离和恢复被遮挡的水果；v）进行三维姿态估计（位置和方向）。该方法提取了完整的簇几何形状，确保可靠识别部分隐藏的番茄。在一个经过严格验证的模拟框架中评估多个场景，具有不同的遮挡水平，系统的精度超过90%，平均召回率为82.8%，平均交并比 (mIoU) 为80.7%。此外，它在质心估计中表现出高重复性，均方根误差 (RMSE) 仅为4.2毫米，验证了其在自主采摘操作中的技术可行性和高准确性。
cs.RO / 77 / 2609.18752
QMSR: Query-Conditioned Mask-wise Expert Routing for Robust Open-Vocabulary Underwater Object Retrieval
QMSR：基于查询条件的掩码专家路由框架用于鲁棒的开放词汇水下物体检索
Abstract
Open-vocabulary object retrieval remains challenging in complex underwater environments. Although underwater image enhancement (UIE) can improve visual quality, fixed UIE strategies may even underperform the Raw representation in retrieval, indicating that enhancement should not be applied as a uniform preprocessing step. To address this problem, we propose \textbf{QMSR}, a query-conditioned mask-wise expert routing framework for underwater open-vocabulary retrieval. Specifically, QMSR selects one pretrained UIE expert for each query--candidate pair and predicts a continuous Raw--Expert fusion strength, enabling adaptive enhancement while preserving useful Raw semantics. During training, a privileged ranking oracle provides expert-selection and fusion-strength supervision, while an annealed soft-routing relaxation facilitates optimization of the hard Top-1 routing policy. Experiments show that QMSR improves NDCG@10 by 17.6\% over an image--query shared router, while consistently outperforming fixed UIE strategies and remaining effective on held-out query categories. These results demonstrate the effectiveness of query-conditioned and candidate-specific enhancement routing for underwater open-vocabulary retrieval.
Chinese Translation
在复杂的水下环境中，开放词汇物体检索仍然面临挑战。尽管水下图像增强（UIE）可以改善视觉质量，但固定的UIE策略在检索中可能表现不如原始表示，表明增强不应作为统一的预处理步骤。为了解决这个问题，我们提出了 extbf{QMSR}，一种基于查询条件的掩码专家路由框架，用于水下开放词汇检索。具体而言，QMSR为每个查询-候选对选择一个预训练的UIE专家，并预测一个连续的原始-专家融合强度，从而实现自适应增强，同时保留有用的原始语义。在训练过程中，一个特权排序oracle提供专家选择和融合强度的监督，而退火软路由松弛则促进了硬Top-1路由策略的优化。实验表明，QMSR在NDCG@10上比图像-查询共享路由器提高了17.6\%，同时始终优于固定的UIE策略，并在保留的查询类别上保持有效。这些结果证明了基于查询条件和候选特定的增强路由在水下开放词汇检索中的有效性。
cs.RO / 78 / 2609.18763
Gated Residual Body-Hand Coordination for Whole-Body Humanoid Teleoperation
用于全身类人远程操作的门控残差身体-手协调
Abstract
Whole-body humanoid teleoperation commonly combines a motion-tracking policy with a separate dexterous-hand retargeter. However, independently generated commands do not explicitly preserve body-hand geometric relations, leading to mismatches in relative wrist poses and fingertip positions during bimanual interaction. We present a gated residual coordination framework that keeps both modules frozen and applies bounded corrections to their outputs. A motion-conditioned action gate allocates correction authority across joint groups, while reference-geometry-dependent reward gates emphasize relevant interaction objectives during training. To establish the nominal body controller on Agile One, we introduce multi-pose morphology calibration that jointly estimates triaxial scales and effector-local offsets, together with staged motion dataset curation for training a SONIC-based tracker. The residual policy uses human motion references, initial commands, and robot proprioception without explicit object or contact observations. In simulation, it reduces wrist and fingertip geometry errors by 39.2-56.3% over direct composition on held-out GRAB motions, while preserving whole-body tracking on AMASS, with success rates of 89.03% without residual coordination and 89.29% with it. Ablations characterize the contributions of reward gating, adaptive correction authority, and separate body and hand correction heads.
Chinese Translation
全身类人远程操作通常将运动跟踪策略与独立的灵巧手重定向器结合。然而，独立生成的指令并未明确保持身体与手之间的几何关系，导致双手交互过程中手腕相对姿态和指尖位置的不匹配。我们提出了一种门控残差协调框架，该框架保持两个模块的静止，并对其输出施加有界修正。运动条件下的动作门在关节组之间分配修正权，而依赖参考几何的奖励门在训练过程中强调相关的交互目标。为了在Agile One上建立名义身体控制器，我们引入了多姿态形态校准，该校准共同估计三轴比例和效应器局部偏移，并结合分阶段的运动数据集策划来训练基于SONIC的跟踪器。残差策略使用人类运动参考、初始指令和机器人本体感知，而无需显式的物体或接触观察。在仿真中，它在保留全身跟踪的同时，将手腕和指尖几何误差减少了39.2%-56.3%，在未使用残差协调的情况下成功率为89.03%，而使用残差协调时成功率为89.29%。消融实验表征了奖励门控、自适应修正权以及独立的身体和手修正头的贡献。
cs.RO / 79 / 2609.18776
TRACER: Adaptive Multi-Robot Social Navigation via Joint Human-Response Prediction and Interaction-Aware Replanning
TRACER：通过联合人类响应预测和交互感知重规划实现自适应多机器人社会导航
Abstract
Multi-robot navigation in human-shared spaces is inherently interactive: coordinated robot motions influence how nearby entities respond, while those responses provide valuable information for subsequent robot decisions. However, existing methods typically address action-conditioned prediction, multi-robot planning, or online adaptation separately, and therefore lack a unified mechanism for modeling joint robot-entity interactions and adapting future decisions from executed interaction outcomes. To address this gap, we propose TRACER, a bi-directional receding-horizon framework that closes the loop between prediction and adaptation. TRACER evaluates candidate (i.e., alternative feasible future motion plans for the robot team) trajectories using a per-entity probabilistic response model that separates individual-robot effects from non-additive pairwise interactions; after executing the selected trajectory prefix, it updates persistent identity-bound beliefs over latent response modes using the synchronized observed responses. These updated beliefs then guide subsequent candidate evaluation under probabilistic safety and response-aware cost criteria. Experiments show that (i) TRACER more accurately captures non-additive multi-robot interaction effects than a capacity-matched additive predictor, (ii) persistent identity-consistent evidence improves response prediction and downstream replanning, and (iii) the complete TRACER framework improves collision-free completion over an independent-robot baseline on the SocialGym2 multi-robot social-navigation benchmark.
Chinese Translation
在共享人类空间中的多机器人导航本质上是互动的：协调的机器人运动会影响附近实体的反应，而这些反应又为后续的机器人决策提供了宝贵的信息。然而，现有的方法通常分别处理基于动作的预测、多机器人规划或在线适应，因此缺乏一个统一的机制来建模联合机器人-实体交互并根据执行的交互结果调整未来的决策。为了解决这一问题，我们提出了TRACER，一个双向递归视野框架，旨在闭合预测与适应之间的循环。TRACER使用每个实体的概率响应模型评估候选（即，机器人团队的替代可行未来运动计划）轨迹，该模型将单个机器人效应与非加性成对交互分开；在执行所选轨迹前缀后，它利用同步观察到的响应更新对潜在响应模式的持久身份绑定信念。这些更新的信念随后在概率安全性和响应感知成本标准下指导后续候选评估。实验表明：(i) TRACER比容量匹配的加性预测器更准确地捕捉非加性多机器人交互效应，(ii) 持久身份一致的证据改善了响应预测和后续重规划，(iii) 完整的TRACER框架在SocialGym2多机器人社会导航基准测试中提高了无碰撞完成率，相较于独立机器人基线表现更佳。
cs.RO / 80 / 2609.18789
AdaGeoVLN: Selective Geometry Across Representation Depth and Navigation Time for Vision-Language Navigation
AdaGeoVLN：跨表示深度和导航时间的选择性几何体用于视觉-语言导航
Abstract
Vision-language navigation requires aligning language with visual observations while maintaining spatial understanding over time. Geometry foundation models (GFMs) expose intermediate representations throughout their hierarchy, but how navigation policies should use these features and retain historical geometric evidence remains unresolved. We introduce \method{}, a streaming VLN framework that addresses these questions across \textbf{representation depth} and \textbf{navigation time}. Hierarchical GFM--VLM fusion couples earlier, intermediate, and later GFM representations to successive policy stages instead of repeatedly injecting a terminal feature. Navigation-aware GFM memory retains historical VGGT global-attention KV states according to instruction relevance, geometric confidence, and transition novelty under a bounded per-layer budget. Retained states provide geometric context for subsequent observations before fusion with the policy. Across R2R-CE and RxR-CE, \method{} achieves strong performance using a single RGB stream without additional navigation-specific external data. Controlled ablations show that multi-depth coupling substantially outperforms repeated terminal-feature injection at matched fusion locations. Bounded navigation-aware retention preserves navigation performance while considerably reducing GFM-KV memory relative to larger-memory temporal retention. These findings support jointly examining the geometric representations exposed to the policy and the historical evidence retained for future inference. Code will be released upon acceptance at https://humanoid-research.github.io/adageovln/.
Chinese Translation
视觉-语言导航需要将语言与视觉观察对齐，同时保持对空间的理解。几何基础模型（GFM）在其层次结构中暴露了中间表示，但导航策略应如何使用这些特征并保留历史几何证据仍然未得到解决。我们提出了 extit{AdaGeoVLN}，一个流式的视觉-语言导航框架，解决了在 extbf{表示深度}和 extbf{导航时间}方面的问题。层次化的GFM-VLM融合将早期、中间和后期的GFM表示与连续的策略阶段相结合，而不是反复注入终端特征。导航感知的GFM记忆根据指令相关性、几何置信度和过渡新颖性，在每层的预算限制下保留历史VGGT全局注意力KV状态。保留的状态为后续观察提供几何上下文，然后再与策略融合。在R2R-CE和RxR-CE数据集上， extit{AdaGeoVLN}在不使用额外导航特定外部数据的情况下，利用单一RGB流实现了强大的性能。控制性消融实验表明，多深度耦合在匹配融合位置时显著优于反复注入终端特征。受限的导航感知保留在相对于更大内存的时间保留下保持了导航性能，同时显著减少了GFM-KV内存。这些发现支持共同检查暴露给策略的几何表示和为未来推理保留的历史证据。代码将在接受后发布，网址为 https://humanoid-research.github.io/adageovln/。
cs.RO / 81 / 2609.18813
Asymptotically Optimal Multi-Robot Task and Motion Planning
渐近最优的多机器人任务与运动规划
Abstract
Multi-robot task and motion planning (MR-TAMP) requires jointly reasoning about discrete task decisions and continuous collision-free motions of multiple interacting robots. Although asymptotically optimal algorithms have been developed for task and motion planning, extending these guarantees to the multi-robot setting introduces an important challenge: different task transitions may involve different subsets of robots and therefore impose constraints of different dimensions on the composite configuration space. Consequently, an asymptotically optimal planner must not only optimize motion within each task mode, but also ensure sufficient exploration of the different types of transitions connecting them. We characterize this transition structure and establish sufficient conditions for global asymptotic optimality in MR-TAMP, requiring persistent coverage of relevant transitions and asymptotically improving motion planning within connected feasible regions. Based on these conditions, we develop an efficient asymptotically optimal MR-TAMP algorithm that combines evolving individual-robot roadmaps with implicit tensor-product search, avoiding explicit construction of the composite roadmap. The planner further employs conditional transition sampling, lazy collision checking, and mode- and solution-level guidance to improve finite-time planning efficiency while retaining persistent exploration. The resulting framework provides asymptotic optimality guarantees for multi-robot manipulation while efficiently exploiting the structure of individual-robot motion planning.
Chinese Translation
多机器人任务与运动规划（MR-TAMP）需要对多个相互作用的机器人进行离散任务决策和连续无碰撞运动的联合推理。尽管已经开发出渐近最优的任务与运动规划算法，但将这些保证扩展到多机器人环境中带来了一个重要挑战：不同的任务转移可能涉及不同的机器人子集，因此在复合配置空间上施加不同维度的约束。因此，渐近最优的规划器不仅必须在每个任务模式内优化运动，还必须确保充分探索连接它们的不同类型的转移。我们对这种转移结构进行了表征，并建立了MR-TAMP中全局渐近最优性的充分条件，要求对相关转移进行持续覆盖，并在连接的可行区域内渐近改善运动规划。基于这些条件，我们开发了一种高效的渐近最优MR-TAMP算法，该算法结合了不断演变的单机器人路线图与隐式张量积搜索，避免了复合路线图的显式构建。该规划器进一步采用条件转移采样、惰性碰撞检测以及模式和解决方案级别的指导，以提高有限时间规划的效率，同时保持持续探索。最终框架为多机器人操作提供了渐近最优性的保证，同时有效利用了单机器人运动规划的结构。
cs.RO / 82 / 2609.18819
SEAM: Submap-Anchored Evidence for Lifelong LiDAR Mapping under Trajectory Deformation
SEAM：基于子地图锚点的终身激光雷达映射证据在轨迹变形下的应用
Abstract
We propose SEAM, a LiDAR-based lifelong mapping framework. Instead of relying on a single anchor spanning the entire session, SEAM generates evidence based on a trajectory optimized with submap-level anchors, and performs dynamic object removal and change detection. Through submap-level reprojection, the generated evidence remains usable even if the trajectory is subsequently modified by a new session, eliminating the need to recompute the entire process from scratch. SEAM suppresses geometrically unreliable inter-session loop edges using a DOP-based confidence measure. Suppressing unreliable loop edges prevents alignment errors. SEAM also uses a directional voxel-wise evidence model. The model accounts for occupancy patterns that vary with ray direction. Direction-aware evidence separates dynamic objects from environmental changes more precisely. Experiments on a real construction-site dataset and a long-term multi-session dataset show that SEAM achieves higher accuracy and faster processing than existing methods.
Chinese Translation
我们提出了SEAM，一个基于激光雷达的终身映射框架。SEAM并不依赖于跨越整个会话的单一锚点，而是基于经过子地图级锚点优化的轨迹生成证据，并执行动态物体移除和变化检测。通过子地图级重投影，生成的证据即使在轨迹随后被新会话修改后仍然可用，从而消除了从头重新计算整个过程的需要。SEAM使用基于DOP的置信度度量来抑制几何上不可靠的会话间环路边缘。抑制不可靠的环路边缘可以防止对齐错误。SEAM还使用方向性体素级证据模型。该模型考虑了随光线方向变化的占用模式。方向感知证据更精确地区分动态物体和环境变化。在真实施工现场数据集和长期多会话数据集上的实验表明，SEAM比现有方法实现了更高的准确性和更快的处理速度。
cs.RO / 83 / 2609.18869
KINO: A Keyframe Interface for VLM Planning and Whole-Body Control in Humanoid Loco-Manipulation
KINO：用于人形机器人运动操控和全身控制的关键帧接口
Abstract
Humanoid loco-manipulation requires robots to interpret task instructions and scene semantics while executing coordinated whole-body motions. We propose a hierarchical framework that uses motion keyframes as an intermediate representation between Vision-Language Model (VLM) planning and Reinforcement Learning (RL) control. Each keyframe specifies a target whole-body robot pose and, when applicable, an object pose. Given a language instruction, scene observations, and execution feedback, the VLM selects successive task-relevant keyframes from a predefined library. The selected keyframes are retargeted to the current scene to account for object poses and dimensions. A keyframe-conditioned whole-body policy then generates joint-level actions to reach these goals. We introduce a saliency-based keyframe sampling strategy for low-level policy training that improves end-to-end task success rate from 44% to 92% when using sparse VLM keyframes. We evaluate our framework on object pickup, transport, and placement tasks in simulation and on a Unitree G1 humanoid. The system successfully performs both one- and two-handed manipulation and generalises to placement locations beyond the training reference data.
Chinese Translation
人形机器人运动操控要求机器人在执行协调的全身动作时，能够理解任务指令和场景语义。我们提出了一个层次化框架，该框架使用运动关键帧作为视觉-语言模型（Vision-Language Model, VLM）规划与强化学习（Reinforcement Learning, RL）控制之间的中介表示。每个关键帧指定一个目标全身机器人姿态，并在适用时指定一个物体姿态。在给定语言指令、场景观察和执行反馈的情况下，VLM从预定义库中选择连续的与任务相关的关键帧。所选择的关键帧被重新定向到当前场景，以考虑物体姿态和尺寸。然后，基于关键帧的全身策略生成关节级别的动作以达到这些目标。我们引入了一种基于显著性的关键帧采样策略，用于低级策略训练，该策略在使用稀疏VLM关键帧时将端到端任务成功率从44%提高到92%。我们在仿真环境和Unitree G1人形机器人上评估了我们的框架，成功执行了单手和双手操控，并且在训练参考数据之外的放置位置上也具有良好的泛化能力。
cs.RO / 84 / 2609.18881
Body-Motion Control of a Simulated Aerial Swarm from a First-Person View
基于第一人称视角的模拟空中群体的身体运动控制
Abstract
First-person-view (FPV) teleoperation of aerial swarms requires an operator to coordinate collective translation, viewing direction, and formation spacing. We present an upper-body interface that maps torso inclination, hand position, and head rotation to five continuous command dimensions. Neutral postures and motion ranges are calibrated for each participant. In a within-subject study, 14 participants navigated a simulated 15-agent swarm through three-dimensional obstacle courses using this interface and a conventional transmitter. Body-motion control reduced completion time by 19.4% and centroid path length by 7.0%, and increased path directness. Delivered-command variation was 88.8% lower, and concurrent command changes were more frequent. These command measures characterize the complete interfaces, which differed in calibration and filtering. No differences were detected in gate-centering error, collection yield, crash or disconnection counts, overall workload, or usability. All participants reported higher physical demand with body-motion control. The implemented interface therefore improved FPV navigation efficiency at the cost of greater physical demand.
Chinese Translation
第一人称视角（FPV）遥操作空中群体要求操作员协调集体平移、视角方向和队形间距。我们提出了一种上半身接口，将躯干倾斜、手部位置和头部旋转映射到五个连续的指令维度。为每位参与者校准了中性姿势和运动范围。在一项被试内研究中，14名参与者使用该接口和传统发射器导航模拟的15代理空中群体穿越三维障碍课程。身体运动控制将完成时间缩短了19.4%，质心路径长度减少了7.0%，并提高了路径的直接性。传达指令的变化减少了88.8%，并且同时指令变化的频率更高。这些指令测量特征描述了完整的接口，这些接口在校准和过滤方面有所不同。未检测到在门中心误差、收集产量、碰撞或断开计数、整体工作负荷或可用性方面的差异。所有参与者报告使用身体运动控制时的身体需求更高。因此，所实施的接口提高了FPV导航的效率，但代价是更大的身体需求。
cs.RO / 85 / 2609.18893
SOL-SLAM: Inverse Compositional Gauss-Newton Direct Registration for Fast Sonar-Only Local SLAM
SOL-SLAM：逆复合高斯-牛顿直接配准用于快速声纳局部SLAM
Abstract
Autonomous underwater navigation typically relies on complex and expensive multi-modal sensor suites designed to prioritize global Simultaneous Localization and Mapping (SLAM) accuracy. However, local reactive behaviors such as coarse navigation and obstacle avoidance require only local consistency---a capability that should be feasible using only a Forward-Looking Sonar (FLS), yet remains largely unaddressed, leaving a critical gap in FLS-only local SLAM. Moreover, existing acoustic SLAM frameworks predominantly rely on sparse feature extraction methods that discard substantial portions of the already information-sparse acoustic returns. To overcome these limitations, this work introduces a dense direct registration approach that aligns full acoustic intensity scans to a recursively updated local map. Real-time execution is achieved via an Inverse Compositional Gauss-Newton optimization strategy that minimizes computational overhead. Experimental evaluations show that this dense method yields significant improvements on translation error compared to sparse keypoint baselines, maintaining stable sub-meter tracking precision over wide displacement gaps. Moreover, this approach delivers odometry performance comparable to multi-sensor fusion pipelines (FLS, DVL, and IMU), bypassing expensive payload dependencies in feature-rich environments. We validate real-world applicability through AUV field trials, running the full local SLAM approach onboard an embedded, resource-constrained computer.
Chinese Translation
自主水下导航通常依赖于复杂且昂贵的多模态传感器套件，旨在优先考虑全球同步定位与地图构建（SLAM）的准确性。然而，局部反应行为如粗略导航和障碍物避免仅需局部一致性——这一能力应仅使用前视声纳（FLS）即可实现，但在FLS仅局部SLAM中仍然未得到充分解决，留下了一个关键的空白。此外，现有的声学SLAM框架主要依赖于稀疏特征提取方法，这些方法丢弃了大量已经信息稀疏的声学回波。为克服这些局限性，本研究提出了一种密集直接配准方法，该方法将完整的声学强度扫描与递归更新的局部地图对齐。通过逆复合高斯-牛顿优化策略实现实时执行，从而最小化计算开销。实验评估表明，与稀疏关键点基线相比，这种密集方法在平移误差上显著改善，并在较大位移间隔内保持稳定的亚米级跟踪精度。此外，该方法提供的里程计性能可与多传感器融合管道（FLS、DVL和IMU）相媲美，避免了在特征丰富环境中对昂贵载荷的依赖。我们通过AUV现场试验验证了其在现实世界中的适用性，在嵌入式资源受限计算机上运行完整的局部SLAM方法。
cs.RO / 86 / 2609.18900
Examining the Difference in Human Behavior Between Virtual and Real-World Human-Robot Teaming
虚拟与现实世界人机团队合作中人类行为差异的研究
Abstract
Prototyping and evaluating human-robot teaming (HRT) scenarios in the real-world is costly. Virtual simulation of HRT scenarios has been adopted as an alternative to conducting user studies in the real-world to investigate user perceptions, behaviors, and performance during human-robot interactions. The consistency of human behavior between the real and virtual-worlds is integral to the validity of utilizing such virtual experimentation. This paper presents a user study examining the difference in human behavior between an HRT scenario conducted in a virtual vs real environment. We employed mixed-methods to examine team performance and human factors during the HRT. Our quantitative results showed a significant difference in workload between the two modalities. Our qualitative analysis expanded on the quantitative results and found differences in participants' strategy, their mental model of robots, and the type of trust they had for robots between modalities.
Chinese Translation
在现实世界中原型设计和评估人机团队合作（HRT）场景的成本较高。虚拟模拟HRT场景已被采用作为在现实世界中进行用户研究的替代方案，以调查用户在与机器人互动过程中的感知、行为和表现。人类在现实与虚拟世界之间行为的一致性对于利用这种虚拟实验的有效性至关重要。本文呈现了一项用户研究，考察了在虚拟环境与现实环境中进行的HRT场景中人类行为的差异。我们采用混合方法研究HRT过程中的团队表现和人因因素。我们的定量结果显示两种模式之间的工作负荷存在显著差异。我们的定性分析扩展了定量结果，发现参与者在策略、对机器人的心理模型以及对机器人的信任类型方面在两种模式之间存在差异。
cs.RO / 87 / 2609.18910
CaSCo: Cascade-Aware Soft-Collision Motion Planning
CaSCo：级联感知软碰撞运动规划
Abstract
Conventional motion planning treats collision as a binary constraint, although contact with different objects can have drastically different consequences. A robot may safely brush against a cardboard box while even minor contact with a glass, laptop, or unstable object may be undesirable. Moreover, a direct robot--object collision can move the contacted object and trigger secondary object--object collisions, making the risk of a motion depend on the physical evolution of the scene rather than only on the robot's geometric path. We present CaSCo, a cascade-aware soft-collision motion planning framework in which a vision-language or language model assigns semantic risk to objects and a physics simulator predicts the consequences of candidate robot motions. CaSCo searches for a path that minimizes the total semantic risk of the unique objects displaced either directly by the robot or indirectly through cascaded collisions. Because collisions change the environment, we augment roadmap states with the predicted object arrangement and the set of objects whose risk has already been incurred. We develop an optimal graph-search algorithm with an admissible and consistent cascade-relaxed heuristic and caching and pruning mechanisms for efficient search. Experiments in cluttered manipulation environments evaluate semantic risk, cascade reasoning, planning efficiency, and real-robot operation.
Chinese Translation
传统的运动规划将碰撞视为二元约束，尽管与不同物体的接触可能会产生截然不同的后果。机器人可以安全地轻触纸箱，而与玻璃、笔记本电脑或不稳定物体的轻微接触可能是不可取的。此外，机器人与物体的直接碰撞可能会移动被接触的物体，并引发次级物体之间的碰撞，使得运动的风险不仅依赖于机器人的几何路径，还依赖于场景的物理演变。我们提出了CaSCo，一个级联感知的软碰撞运动规划框架，其中视觉-语言或语言模型为物体分配语义风险，物理模拟器预测候选机器人运动的后果。CaSCo搜索一条路径，以最小化直接被机器人或通过级联碰撞间接位移的独特物体的总语义风险。由于碰撞会改变环境，我们通过预测的物体排列和已产生风险的物体集合来增强路线图状态。我们开发了一种具有可接受性和一致性的级联放松启发式的最优图搜索算法，并采用缓存和剪枝机制以提高搜索效率。在杂乱的操作环境中的实验评估了语义风险、级联推理、规划效率和真实机器人操作。
cs.RO / 88 / 2609.18930
Learning Holistic Whole-Body Loco-Manipulation with a Bipedal Mobile Manipulator
基于双足移动操控器的整体全身运动操控学习
Abstract
Bipedal loco-manipulation enables robots to interact with objects beyond the nominal workspace of their arms by coordinating locomotion and manipulation. Realizing this capability requires a low-level whole-body controller that translates task-level manipulation goals into coordinated arm and leg motions while maintaining balance. We present a unified whole-body controller trained with reinforcement learning that directly maps 6-DoF end-effector targets to coordinated actions for the bipedal base and robotic arm. Given only an end-effector target, the learned controller autonomously coordinates reaching, postural adaptation, and stepping without explicit base-velocity or footstep commands. A reward-gating strategy regulates the trade-offs among end-effector tracking, locomotion, and balance during training, while a temporal context estimator combines windowed Transformer encoding, recurrent GRU memory, and auxiliary dynamics prediction to extract dynamics-relevant information from observation history. Real-robot experiments demonstrate that the same controller supports reaching, postural adaptation, and stepping under commands from VR teleoperation, a learned diffusion policy, and scripted trajectories, providing a common end-effector interface for diverse manipulation tasks.
Chinese Translation
双足运动操控使机器人能够通过协调运动和操控与超出其手臂名义工作空间的物体进行交互。实现这一能力需要一个低层次的全身控制器，该控制器将任务级操控目标转化为协调的手臂和腿部动作，同时保持平衡。我们提出了一种使用强化学习训练的统一全身控制器，该控制器直接将6自由度末端执行器目标映射为双足底座和机器人手臂的协调动作。在仅给定末端执行器目标的情况下，学习到的控制器能够自主协调到达、姿态适应和步态，而无需明确的底座速度或步态指令。奖励门控策略在训练过程中调节末端执行器跟踪、运动和保持平衡之间的权衡，而时间上下文估计器结合了窗口化的Transformer编码、递归GRU记忆和辅助动态预测，从观察历史中提取与动态相关的信息。真实机器人实验表明，同一控制器在来自虚拟现实遥操作、学习的扩散策略和脚本轨迹的指令下支持到达、姿态适应和步态，为多样化的操控任务提供了一个共同的末端执行器接口。
cs.RO / 89 / 2609.18970
"What's going to happen after I'm gone?": Parent Perspectives on Technology in Supporting Independent Living for Adults with Intellectual Disabilities
我离开后会发生什么？：父母对技术在支持智力残疾成年人独立生活中的作用的看法
Abstract
Adults with intellectual and developmental disabilities (IDD) are increasingly transitioning from family homes towards semi-inde\-pendent living. As parents hand off the role of primary caregiver, they face numerous challenges in arranging consistent and quality support. Our research centers on understanding these caregiving routines. By focusing on the unique lived experience of parents, who possess extensive explicit and tacit knowledge of their adult child's requirements, we aim to map the management of care they provide. This foundational understanding is essential to identifying how assistive technologies can effectively serve a key role in supporting adults with IDD in this transition. In this work, we interviewed 16 parents of adults with IDD beginning this transition to understand: 1) how they currently provide support for daily living and what makes their support effective; 2) what their experiences and perceptions are regarding the use of technology; and 3) how they envision assistive technology successfully integrating with their adult child's new home or with other supports to promote independence. Our thematic analysis produced four themes: common modes of support, inside the routine to support growth, the fragility of continuity of care across transitions, and participants' perceptions and experiences of assistive technology. Building on these findings, we derive four design principles for growth-oriented assistive technology. These include pre-transition onboarding to capture caregiver tacit knowledge; structured scaffolding towards long-term growth; adaptive sensing that responds to day-to-day variability; and customization for the individual balanced with consistency for the care network.
Chinese Translation
智力和发展障碍（IDD）成年人正越来越多地从家庭住所过渡到半独立生活。随着父母逐渐卸下主要照顾者的角色，他们在安排一致且高质量的支持方面面临诸多挑战。我们的研究聚焦于理解这些照顾例程。通过关注父母独特的生活经验，他们对成年子女需求的广泛显性和隐性知识，我们旨在描绘他们提供的照顾管理。这一基础理解对于识别辅助技术如何在这一过渡中有效支持IDD成年人至关重要。在本研究中，我们采访了16位正在开始这一过渡的IDD成年人父母，以了解：1）他们目前如何提供日常生活支持，以及是什么使他们的支持有效；2）他们对技术使用的经验和看法；3）他们如何设想辅助技术成功融入成年子女的新家或其他支持中，以促进独立性。我们的主题分析产生了四个主题：常见的支持方式、支持成长的日常例程、过渡过程中照护连续性的脆弱性，以及参与者对辅助技术的看法和体验。基于这些发现，我们提出了四个面向成长的辅助技术设计原则。这些原则包括：过渡前的入职培训以捕捉照顾者的隐性知识；结构化的支架以促进长期成长；适应性感知以应对日常变化；以及为个体定制与照护网络的一致性之间的平衡。
cs.RO / 90 / 2609.19012
Information-Based Trajectory Planning for Spacecraft-to-Spacecraft Tracking and Navigation in Cislunar Space
基于信息的轨迹规划用于月球间航天器跟踪与导航
Abstract
We present a trajectory planning method that balances information collection and control effort to improve cislunar spacecraft-to-spacecraft absolute tracking. Expanding use of cislunar space requires alternative navigation and tracking procedures that minimize reliance on ground-based support. Among efforts to address this need, spacecraft-to-spacecraft tracking exploits nonlinear dynamical tracers encoded in a series of relative measurements to infer absolute states of both an observer and a target. The geometry between spacecraft operating under this mode can significantly influence tracking performance. This work considers this geometrical impact by designing observer trajectories that jointly balance control effort and an information-theoretic quantification of the expected spacecraft-to-spacecraft tracking performance. Our methods are designed for multiple low-thrust observation platforms of various sensing modalities and can incorporate multiple space object targets in planning. By leveraging information gain in the optimal control problem, we report almost an order of magnitude improvement in the expected navigation and tracking errors for an optical observer operating in a Distant Retrograde Orbit (DRO). This work demonstrates the feasibility and value of information-optimal low-thrust spacecraft trajectory design for current and upcoming cislunar missions.
Chinese Translation
我们提出了一种轨迹规划方法，该方法在信息收集和控制努力之间取得平衡，以改善月球间航天器的绝对跟踪。月球间空间的扩展使用需要替代的导航和跟踪程序，以最小化对地面支持的依赖。在应对这一需求的努力中，航天器间跟踪利用非线性动态跟踪器，通过一系列相对测量推断观察者和目标的绝对状态。在这种模式下操作的航天器之间的几何关系可能显著影响跟踪性能。本研究通过设计观察者轨迹来考虑这种几何影响，旨在共同平衡控制努力和信息论对预期航天器间跟踪性能的量化。我们的方法适用于多种传感模式的低推力观测平台，并可以在规划中纳入多个空间目标。通过在最优控制问题中利用信息增益，我们报告了在远程逆行轨道（Distant Retrograde Orbit, DRO）中操作的光学观察者的预期导航和跟踪误差几乎提高了一个数量级。这项工作展示了信息最优低推力航天器轨迹设计在当前和即将到来的月球间任务中的可行性和价值。
cs.RO / 91 / 2609.19040
Learning to Stack: Cube-Stacking Imitation Learning from Virtual Reality Demonstrations
学习堆叠：来自虚拟现实演示的立方体堆叠模仿学习
Abstract
Imitation learning is attractive for robot manipulation, but collecting demonstrations remains a bottleneck for multi-stage tasks requiring repeated scene resets. This work presents a virtual-reality data-collection pipeline for cube-stacking with a custom 5-DoF arm in NVIDIA Isaac Sim and Isaac Lab. Using an HTC Vive Pro 2, Manus Quantum gloves, and OpenXR, an operator provides SE(3) end-effector commands to generate task demonstrations. The proposed framework separates demonstration collection from dataset construction by replaying recorded trajectories, converting task-space commands into joint-space actions, and re-rendering demonstrations with updated sensor or state configurations. This allows previously collected demonstrations to be reused for new observation and action spaces without repeating human teleoperation. The task requires stacking the red cube on the blue cube and the green cube on the red cube, with randomized cube placement. In 30 minutes, 200 virtual demonstrations were collected, compared with 45 real-world demonstrations, and Isaac Mimic generated 100 additional samples. A behavior-cloning policy was trained from the virtual demonstrations using LeRobot-style dual-camera observations and evaluated in simulation.
Chinese Translation
模仿学习在机器人操作中具有吸引力，但收集演示仍然是多阶段任务的瓶颈，这些任务需要重复重置场景。本研究提出了一种虚拟现实数据收集管道，用于在NVIDIA Isaac Sim和Isaac Lab中使用定制的5自由度手臂进行立方体堆叠。通过HTC Vive Pro 2、Manus Quantum手套和OpenXR，操作员提供SE(3)末端执行器命令以生成任务演示。所提出的框架通过重放记录的轨迹，将任务空间命令转换为关节空间动作，并使用更新的传感器或状态配置重新渲染演示，从而将演示收集与数据集构建分离。这使得先前收集的演示可以在不重复人类遥控操作的情况下，重新用于新的观察和动作空间。该任务要求将红色立方体堆叠在蓝色立方体上，并将绿色立方体堆叠在红色立方体上，立方体的放置是随机的。在30分钟内，收集了200个虚拟演示，而现实世界中仅收集了45个演示，并且Isaac Mimic生成了100个额外样本。使用LeRobot风格的双摄像头观察，从虚拟演示中训练了一个行为克隆策略，并在仿真中进行了评估。
cs.RO / 92 / 2609.19041
Loco-Loco-RL: Low-Cost Terrain Mapping for Humanoid Locomotion with Reinforcement Learning
Loco-Loco-RL：基于强化学习的人形运动低成本地形映射
Abstract
Informative terrain perception is important for robust reinforcement learning policies in humanoid locomotion. Still, common sensors such as depth cameras and LiDARs incur high cost, power, and processing overhead while often producing redundant, high-resolution data. This work uses a low-cost time-of-flight sensor to provide a compact 3D local terrain representation for humanoid locomotion. To efficiently use this sparse exteroceptive input, we introduce a token-compressed temporal transformer policy. Proprioceptive and terrain observations are tokenized and processed by a self-attention multi-head transformer to capture within-timestep relationships between observation terms. The attended tokens are then compressed through an MLP-based latent-space token compression module before being stored in a rolling 15-timestep history. A second cross-attention multi-head transformer extracts temporal locomotion features from this compact history for policy learning. By compressing tokens before temporal aggregation, the architecture preserves important terrain-observation structure while limiting the dimensional growth of attention over observation histories. We validate our method through sim-to-real transfer on physical hardware using a terrain-based locomotion benchmark, demonstrating robust humanoid terrain walking with low-cost local terrain sensing.
Chinese Translation
有效的地形感知对于人形运动中的稳健强化学习策略至关重要。然而，常见的传感器如深度相机和激光雷达（LiDAR）往往成本高昂，功耗大，处理开销高，同时产生冗余的高分辨率数据。本研究采用低成本的飞行时间传感器，提供紧凑的3D局部地形表示，以支持人形运动。为了高效利用这种稀疏的外部感知输入，我们引入了一种基于令牌压缩的时序变换器策略。对本体感知和地形观察进行令牌化，并通过自注意力多头变换器处理，以捕捉观察项之间的时间步内关系。经过关注的令牌随后通过基于多层感知器（MLP）的潜在空间令牌压缩模块进行压缩，然后存储在一个滚动的15步历史中。第二个交叉注意力多头变换器从这个紧凑的历史中提取时间运动特征以进行策略学习。通过在时间聚合之前压缩令牌，该架构保留了重要的地形观察结构，同时限制了对观察历史的注意力维度增长。我们通过在物理硬件上使用基于地形的运动基准进行模拟到现实的转移验证了我们的方法，展示了低成本局部地形感知下的稳健人形地形行走。
cs.RO / 93 / 2609.19080
ElastiQP: An Always-Feasible QP Solver for Constrained Robot Control
ElastiQP：一种始终可行的约束机器人控制的二次规划求解器
Abstract
As robot capabilities increase, quadratic programming (QP)-based controllers must account for a similarly increasing number of constraints to ensure safe, reliable operation. Yet, with each added constraint, this introduces more chances of momentary conflict: in which case, a QP solver that returns an "infeasible" status leaves the controller with nothing to execute. To address this, we introduce ElastiQP, a modified dual active-set QP solver that relaxes every inequality constraint with an exact, per-constraint l1 penalty while keeping equality constraints (dynamics) hard. Notably, ElastiQP does so by folding the slack variables into the solver analytically, maintaining a constant size of the condensed linear system. On a suite of robot control benchmarks, ElastiQP achieves microsecond-level performance, matching or outperforming leading modern solvers on feasible problems. On infeasible problems, ElastiQP handles these gracefully, confining violations to strictly the conflicting inequality terms, returning a usable solution up to 40x faster than the best alternative solvers. ElastiQP is available as an open-source C++ header-only library, with Python and JAX interfaces, at https://github.com/StanfordASL/elastiqp.
Chinese Translation
随着机器人能力的提升，基于二次规划（QP）的控制器必须考虑越来越多的约束，以确保安全、可靠的操作。然而，随着每个新增约束的引入，瞬时冲突的可能性也随之增加：在这种情况下，返回“不可行”状态的QP求解器将使控制器无所执行。为了解决这个问题，我们提出了ElastiQP，这是一种修改过的双重活跃集QP求解器，它通过对每个不等式约束施加精确的每约束l1惩罚来放宽约束，同时保持等式约束（动力学）为硬约束。值得注意的是，ElastiQP通过将松弛变量解析地折叠进求解器，保持了浓缩线性系统的恒定大小。在一系列机器人控制基准测试中，ElastiQP实现了微秒级的性能，在可行问题上与领先的现代求解器相匹配或超越。在不可行问题上，ElastiQP优雅地处理这些情况，将违规限制在严格的冲突不等式项内，返回的可用解决方案比最佳替代求解器快多达40倍。ElastiQP作为一个开源的C++头文件库提供，具有Python和JAX接口，网址为https://github.com/StanfordASL/elastiqp。
cs.RO / 94 / 2609.19104
rMuscle: Robotic Muscle Memory for Efficient Vision-Language-Action Model Inference
rMuscle：用于高效视觉-语言-动作模型推理的机器人肌肉记忆
Abstract
Factory work is a promising early scenario for embodied AI: assigning repetitive manual jobs to robots has clear economic payoff, and a structured station keeps the jobs tractable for current policies. Vision-Language-Action (VLA) models now dominate as the policy paradigm for these robots. The inference latency of VLA models directly affects robot responsiveness and motion smoothness. However, existing VLA inference frameworks do not fully exploit the characteristics of embodied workloads or account for the distinct bottlenecks across different stages of VLA inference. In this paper, we first characterize embodied workloads and identify substantial task similarity across repeated robot executions. We further find that such similarity extends beyond observations and action trajectories to internal model states. Drawing on these observations, we present rMuscle, a real-time VLA inference framework inspired by human muscle memory. It exploits cross-execution similarity through a dual-phase muscle-memory cache. The Context Cache reuses visual-token outputs to reduce computation, while the Action Cache reuses neuron activation patterns to reduce weight accesses. We keep both the cache memory footprint and access overhead low through online cache recomputation, sliding-window cache retrieval, and mask sharing across consecutive denoising steps. rMuscle achieves 1.29-1.42X speedup on RTX 4090 and Jetson Thor across LIBERO, RoboTwin, and physical manipulation tasks, while maintaining the original success rates on real-world robots.
Chinese Translation
工厂工作是具身人工智能的一个有前景的早期场景：将重复的手工工作分配给机器人具有明显的经济效益，并且结构化的工作站使得这些工作对于当前的政策而言可控。视觉-语言-动作（VLA）模型现在已成为这些机器人的政策范式。VLA 模型的推理延迟直接影响机器人的响应能力和运动平滑性。然而，现有的 VLA 推理框架并未充分利用具身工作负载的特征，也未考虑 VLA 推理不同阶段之间的独特瓶颈。在本文中，我们首先对具身工作负载进行了特征描述，并识别出重复机器人执行之间的显著任务相似性。我们进一步发现，这种相似性不仅延伸到观察和动作轨迹，还包括内部模型状态。基于这些观察，我们提出了 rMuscle，一个受人类肌肉记忆启发的实时 VLA 推理框架。它通过双阶段肌肉记忆缓存利用跨执行相似性。上下文缓存重用视觉标记输出以减少计算，而动作缓存重用神经元激活模式以减少权重访问。我们通过在线缓存重新计算、滑动窗口缓存检索和连续去噪步骤之间的掩码共享，保持缓存内存占用和访问开销较低。rMuscle 在 LIBERO、RoboTwin 和物理操控任务上，在 RTX 4090 和 Jetson Thor 上实现了 1.29-1.42 倍的加速，同时保持了真实世界机器人上的原始成功率。
cs.RO / 95 / 2609.19137
Dreaming the Sound of Contact: Leveraging Video and Audio Generation for Zero-Shot Force-Aware Manipulation and Data Generation
梦境中的接触声音：利用视频和音频生成实现零-shot力感知操控与数据生成
Abstract
Recent advances in video generation allow robots to learn manipulation trajectories from generated videos. However, these approaches produce purely kinematic trajectories that lack force information, causing failures in contact-rich tasks where appropriate contact forces are essential for success. In this work, we explore augmenting generated video with audio to shape a bounded, time-varying desired-force profile using the loudness of generated contact sounds. We present a pipeline that jointly leverages generated video and audio to derive motion trajectories and corresponding desired-force profiles from a structured natural-language task prompt. We execute these force-aware trajectories on a Franka Panda robot using a closed-loop force regulator that tracks the audio-shaped force profile during contact. We evaluate our pipeline on multiple tasks that require making contact and demonstrate successful manipulation where a kinematic-only baseline fails. We also use the pipeline as a data generation engine to train policies that achieve the tasks in a closed-loop manner. Project website, videos, and dataset: https://dreamingcontactsound.github.io/
Chinese Translation
近期视频生成的进展使得机器人能够从生成的视频中学习操控轨迹。然而，这些方法生成的轨迹纯粹是运动学轨迹，缺乏力的信息，导致在需要适当接触力的接触丰富任务中出现失败。在本研究中，我们探索通过音频增强生成视频，以利用生成接触声音的响度来塑造一个有界的、随时间变化的期望力谱。我们提出了一种管道，联合利用生成的视频和音频，从结构化的自然语言任务提示中推导运动轨迹和相应的期望力谱。我们在Franka Panda机器人上执行这些力感知轨迹，使用闭环力调节器在接触过程中跟踪音频塑造的力谱。我们在多个需要接触的任务上评估我们的管道，并展示了在运动学基线失败的情况下成功的操控。我们还将该管道用作数据生成引擎，以训练在闭环方式下完成任务的策略。项目网站、视频和数据集： https://dreamingcontactsound.github.io/
cs.CV / 1 / 2609.17545
Selective Prediction and Uncertainty-Aware Referral for Pap Smear Classification
选择性预测与不确定性感知的宫颈涂片分类转介
Abstract
Deep learning models for cervical cytology are almost always evaluated as if every prediction must be acted upon, yet a screening system deployed alongside a cytopathologist need not classify every slide: it can defer the cases it is least certain about. Evaluating such a system requires asking not only how often it is correct, but whether its confidence ranks its errors to the bottom. This paper studies selective prediction and uncertainty-aware referral on the Herlev Pap smear dataset under a binary Normal-versus-Abnormal formulation. Two lightweight transformer backbones (Swin-Tiny, TinyViT-5M) are fine-tuned on Herlev from ImageNet-pretrained weights with weighted random sampling, calibrated by post-hoc temperature scaling fit on a held-out calibration subset, and compared against a soft-voting ensemble of both models. Discrimination is reported alongside expected calibration error (ECE) and, as the primary endpoint, the area under the risk-coverage curve (AURC). No statistically significant difference was detected between the two configurations in accuracy or macro-F1, yet the ensemble halves AURC (0.0022 vs. 0.0045, a 51.8% reduction, lower in all five folds) and extends the coverage at which zero errors are made from 18.3% to 72.8% of the pooled test predictions. The same ensemble is nonetheless worse calibrated in absolute terms (ECE 0.0339 vs. 0.0247) and produces more false negatives (14 vs. 10). These results separate two properties that are frequently conflated: the ability to rank predictions by trustworthiness, and the accuracy of the confidence values themselves. Ensembling improves the former while degrading the latter, and the former directly governs the observed risk-coverage tradeoff, whereas the latter governs the interpretation of the reported confidence values.
Chinese Translation
宫颈细胞学的深度学习模型几乎总是被评估为每个预测都必须采取行动，然而，与细胞病理学家共同部署的筛查系统并不需要对每个切片进行分类：它可以推迟对其最不确定的案例的处理。评估这样的系统不仅需要询问其正确率有多高，还需要考虑其信心是否将错误排在底部。本文研究了在二元正常与异常分类下，Herlev宫颈涂片数据集上的选择性预测和不确定性感知转介。两个轻量级变换器骨干网络（Swin-Tiny, TinyViT-5M）在Herlev上进行了微调，使用从ImageNet预训练权重中加权随机抽样，并通过在保留的校准子集上进行后验温度缩放进行校准，与两个模型的软投票集成进行了比较。报告了区分能力以及期望校准误差（ECE），作为主要终点，风险覆盖曲线下的面积（AURC）。在准确性或宏观F1方面，两种配置之间未检测到统计学上显著的差异，但集成将AURC减半（0.0022对比0.0045，降低了51.8%，在所有五个折叠中均较低），并将零错误的覆盖范围从18.3%扩展到72.8%的汇总测试预测。然而，该集成在绝对校准方面表现更差（ECE 0.0339对比0.0247），并产生了更多的假阴性（14对比10）。这些结果将两个常常混淆的特性区分开来：按可信度对预测进行排序的能力，以及信心值本身的准确性。集成提高了前者的表现，同时降低了后者的表现，而前者直接影响观察到的风险覆盖权衡，而后者则影响报告的信心值的解释。
cs.CV / 2 / 2609.17565
A Heisenberg Lift Descriptor for Order Sensitive Online Handwriting Recognition
一种海森堡提升描述符用于顺序敏感的在线手写识别
Abstract
Online handwriting recognition systems typically represent pen trajectories through fixed-length Euclidean shape descriptors that capture the spatial outline of each stroke, but are insensitive to the order in which that outline is produced. Two strokes that trace the same region of the plane in opposite directions are indistinguishable to any such order-blind representation, yet their traversal directions may carry decisive class information in characters where loop orientation and stroke sequencing matter. This paper introduces a Heisenberg-lift framework that addresses this gap through a compact, interpretable, order-sensitive augmentation for online pen-trajectory features. The simplest instance is the terminal signed area, a single parameter-free scalar appended to an existing Euclidean descriptor at negligible computational cost. Evaluated on two standard online handwriting benchmarks, this one-scalar addition, consistently raises classifier accuracy over the Euclidean baseline. On the hardest character pair in our study, the letters o and y, the signed area alone achieves perfect separation while the Euclidean baseline falls short. The advantage grows further under additive coordinate noise, a practically relevant degradation in pen-trajectory data. A richer fifteen-dimensional extension, derived from a noncommutative Heisenberg-group subdivision scheme, provides additional gains in noisy and loop-structured conditions. Dimension-matched statistical controls confirm that all improvements reflect geometric information rather than feature-count inflation. The resulting descriptor is lightweight, closed-form, and directly interpretable, making it a practical augmentation for online handwriting and related document-trajectory classification pipelines in which the direction of stroke execution carries discriminative information.
Chinese Translation
在线手写识别系统通常通过固定长度的欧几里得形状描述符来表示笔迹轨迹，这些描述符捕捉每个笔画的空间轮廓，但对轮廓生成的顺序不敏感。两个笔画如果在相反方向上描绘同一区域，则在任何这种不考虑顺序的表示中都是不可区分的，然而它们的遍历方向可能在字符中携带决定性的类别信息，尤其是在环形方向和笔画顺序重要的情况下。本文引入了一种海森堡提升框架，通过紧凑、可解释的顺序敏感增强来填补这一空白，应用于在线笔迹轨迹特征。最简单的实例是终端签名面积，这是一个附加到现有欧几里得描述符上的无参数标量，计算成本微乎其微。在两个标准在线手写基准上评估时，这个单一标量的添加始终提高了分类器的准确性，相较于欧几里得基线。在我们研究中最难的字符对，即字母o和y，签名面积单独实现了完美分离，而欧几里得基线则未能达到。在加性坐标噪声下，这一优势进一步扩大，这在笔迹轨迹数据中是一种实际相关的退化。基于非交换海森堡群细分方案的更丰富的十五维扩展，在嘈杂和环形结构条件下提供了额外的增益。维度匹配的统计控制确认，所有的改进反映的是几何信息，而非特征计数的膨胀。所得到的描述符轻量、封闭形式且直接可解释，使其成为在线手写和相关文档轨迹分类管道中的一种实用增强，其中笔画执行的方向携带区分信息。
cs.CV / 3 / 2609.17566
Adaptive Interpolatory Curve Subdivision with Learned Local Angles
基于学习局部角度的自适应插值曲线细分
Abstract
Curve subdivision is pivotal in computer graphics for generating smooth geometric objects from control polygons. Interpolatory subdivision is especially attractive because the refined curve is guaranteed to pass through the designer's control points. Classical four-point and six-point schemes preserve this property, but their behaviour is governed by a single global tension parameter, limiting their ability to adapt across flat regions, sharp turns and varying local geometries. We introduce an adaptive local-angle formulation that keeps the interpolatory structure intact while learning how each new vertex should be inserted. A compact edge-wise predictor assigns one insertion angle per edge, while the original vertices are copied exactly at every refinement level. Interpolation is therefore a structural property of the operator and does not depend on the trained weights. The same predictor is used with geometry-specific geodesic primitives on the Euclidean plane, the two-sphere and the Poincar\'e disk. Under a matched-density evaluation protocol, the method reduces nearest-neighbour error by factors of five to seventeen over the best validation-tuned fixed-tension baseline, and by about 1.8 over centripetal Catmull-Rom in the Euclidean case. It also substantially reduces bending energy and tangent roughness, while remaining competitive with separately trained per-geometry models.
Chinese Translation
曲线细分在计算机图形学中至关重要，用于从控制多边形生成光滑的几何对象。插值细分尤其具有吸引力，因为细化后的曲线保证通过设计师的控制点。经典的四点和六点方案保持了这一特性，但它们的行为由单一的全局张力参数控制，限制了它们在平坦区域、急转弯和变化的局部几何形状中的适应能力。我们提出了一种自适应局部角度的公式，该公式在保持插值结构完整的同时，学习如何插入每个新顶点。一个紧凑的边缘预测器为每条边分配一个插入角度，而原始顶点在每个细化级别上都被精确复制。因此，插值是操作符的结构属性，不依赖于训练权重。相同的预测器也用于欧几里得平面、二球面和庞加莱圆盘上的几何特定测地线原语。在匹配密度评估协议下，该方法将最近邻误差减少了五到十七倍，相较于最佳验证调优的固定张力基线，并在欧几里得情况下减少了约1.8倍相较于向心Catmull-Rom。此外，该方法还显著降低了弯曲能量和切线粗糙度，同时在与单独训练的几何模型竞争时保持竞争力。
cs.CV / 4 / 2609.17613
DualCount: Structurally Consistent Density and Point Modeling for Zero-Shot Object Counting
DualCount：用于零-shot物体计数的结构一致性密度与点建模
Abstract
Zero-shot object counting aims to estimate the number of objects specified by a text query without category-specific training. Recent approaches primarily rely on density regression or detection-style instance prediction. While effective, density-based models often suffer from spatial ambiguity and background leakage due to weakly regulated mass allocation, leading to fragmented or part-biased representations that increase counting error in complex scenes. In this work, we propose an instance-aware dual-decoder framework that structurally couples density and point representations for zero-shot object counting. Instead of treating density estimation as independent pixel-wise regression, we interpret it as a structured mass allocation problem over a latent set of object instances. Predicted instance centers induce a soft instance-wise decomposition of the density map, upon which we enforce two geometric constraints: (1) per-instance mass conservation, ensuring each object contributes approximately one unit of density mass, and (2) center-of-mass alignment, encouraging each density component to concentrate around its corresponding predicted center. These constraints introduce instance-level geometric consistency and lead to more accurate mass allocation, thereby reducing counting error. Extensive experiments on FSC-147, PUCPR+, and CARPK show that our approach consistently reduces counting error and establishes new state-of-the-art performance in zero-shot object counting.
Chinese Translation
零-shot物体计数旨在在没有特定类别训练的情况下，根据文本查询估计物体的数量。近期的方法主要依赖于密度回归或检测风格的实例预测。尽管有效，基于密度的模型常常由于弱调控的质量分配而遭受空间模糊和背景泄漏，导致在复杂场景中产生碎片化或部分偏倚的表示，从而增加计数误差。在本研究中，我们提出了一种实例感知的双解码器框架，结构性地耦合密度和点表示以实现零-shot物体计数。我们不再将密度估计视为独立的像素级回归，而是将其解释为在潜在物体实例集上的结构化质量分配问题。预测的实例中心诱导出密度图的软实例级分解，在此基础上我们施加两个几何约束：（1）每实例质量守恒，确保每个物体大约贡献一个单位的密度质量，以及（2）质心对齐，鼓励每个密度分量集中在其对应的预测中心周围。这些约束引入了实例级几何一致性，并导致更准确的质量分配，从而减少计数误差。在FSC-147、PUCPR+和CARPK上的大量实验表明，我们的方法始终减少计数误差，并在零-shot物体计数中建立了新的最先进性能。
cs.CV / 5 / 2609.17740
Geometry-Driven Shadow Harmonisation for Composited Faces: A Multiplicative, Albedo-Preserving Relighting Pipeline
基于几何驱动的阴影协调用于合成面孔：一种乘法、保持反照率的重光照管道
Abstract
Face swapping and face compositing pipelines routinely produce a face that is geometrically well aligned but photometrically implausible: the donor face carries flat, near-frontal studio illumination while the host body and background carry directional scene light. Most existing remedies re-synthesise the face through colour transfer, neural relighting, or inverse rendering, and therefore risk altering identity, skin tone, and texture. We present a conservative alternative: geometry-driven form-shadow injection. The pipeline never repaints the face. It estimates a per-pixel gain field $g\in[g_{\min},1]$ from a rasterised 3D face proxy and multiplies it channel-uniformly onto linear RGB, so the operator can only darken and cannot shift chromaticity. A dense landmark mesh is rasterised into a depth buffer, from which we derive surface normals, a cavity term, and screen-space cast shadows. Key-light direction is estimated from host-side cues (body, background, hair halo); on-face cues are downweighted because they recover the donor's lighting. Shadow magnitude is not matched to the host: it is set by a three-parameter transfer $(\tau,\sigma,g_{\min})$. The shading field is divided by its 75th percentile over skin, then gated, scaled, clamped, smoothed, and re-clipped inside a feathered, skin-gated face mask. On an analytic face heightfield, the default $(\tau,\sigma,g_{\min})=(0.90,0.45,0.82)$ modifies 56.5% of face pixels with mean gain 0.938 (0.890 on modified pixels) and drives 3.4% of pixels to the floor. Hue invariance is a corollary of the operator. We analyse the transfer in closed form, ablate its parameters, and discuss failure modes of a monotone, darkening-only formulation, including double-shadowing of non-flat donors.
Chinese Translation
面孔交换和面孔合成管道通常会产生几何上对齐但光度上不合理的面孔：供体面孔承载平坦、近正面的工作室照明，而宿主身体和背景则承载方向性场景光。大多数现有的解决方案通过颜色转移、神经重光照或逆向渲染重新合成面孔，因此存在改变身份、肤色和纹理的风险。我们提出了一种保守的替代方案：基于几何的形状阴影注入。该管道从不重新涂抹面孔。它从光栅化的3D面孔代理中估计每个像素的增益场 $g ext{in}[g_{ ext{min}},1]$，并均匀地乘以线性RGB，因此操作员只能加深而不能改变色度。一个密集的地标网格被光栅化到深度缓冲区，从中我们推导出表面法线、凹腔项和屏幕空间投射阴影。关键光源方向是从宿主侧线索（身体、背景、头发光晕）中估计的；面部线索的权重被降低，因为它们恢复了供体的照明。阴影强度不与宿主匹配：它由三个参数的转移 $( au, au,g_{ ext{min}})$ 设置。阴影场被除以其在皮肤上的第75百分位数，然后被门控、缩放、限制、平滑，并在一个羽化的、皮肤门控的面孔掩膜内重新裁剪。在一个解析的面部高度场上，默认的 $( au, au,g_{ ext{min}})=(0.90,0.45,0.82)$ 修改了56.5%的面部像素，平均增益为0.938（在修改的像素上为0.890），并将3.4%的像素驱动到地面。色调不变性是该操作的一个推论。我们以封闭形式分析转移，剖析其参数，并讨论单调、仅加深公式的失败模式，包括非平坦供体的双重阴影问题。
cs.CV / 6 / 2609.17749
How to make effective use of domain experts for image classification?
如何有效利用领域专家进行图像分类？
Abstract
A lot of expectations have been put for years on integrating domain expert knowledge in image classification models. Several approaches have been explored, Concept Bottleneck Models (CBMs) opened up a new avenue of research leading to many variants, and more recently to Concept-based Embedding Models (CEMs). CBM consider binary encoding of each concept, while CEM expands this idea by embedding each concept through two vectors. However in real-life scenarii, domain experts' knowledge is usually organized in concepts determined by various attributes, each attribute encoded either with numerical values, or range of values, or binary values, or categorical values. In this work, we first finetune an image feature extractor for classifying attributes representing the downstream object classes, where the class attributes have been specified by experts under various encoding formats. A classification head is then learnt from these various attributes to categorize target objects. We experimentally show that it improves the classification for three datasets: Kaggle fish dataset, AWA2 and a more challenging new wood charcoal dataset. We then propose an automatic selection of potential missclassified data. In this second step, experts are asked for those data to eventually modify the predicted attributes in order to improve the classification.
Chinese Translation
多年来，人们对将领域专家知识整合到图像分类模型中寄予了很高的期望。已经探索了几种方法，概念瓶颈模型（Concept Bottleneck Models, CBMs）开启了一条新的研究途径，导致了许多变体，最近又发展出了基于概念的嵌入模型（Concept-based Embedding Models, CEMs）。CBM考虑每个概念的二进制编码，而CEM通过两个向量嵌入每个概念，扩展了这一思路。然而，在现实场景中，领域专家的知识通常以由各种属性决定的概念形式组织，每个属性可以用数值、数值范围、二进制值或分类值进行编码。在本研究中，我们首先对图像特征提取器进行微调，以分类表示下游对象类别的属性，其中类属性是由专家根据各种编码格式指定的。然后，从这些不同的属性中学习一个分类头，以对目标对象进行分类。我们通过实验表明，这种方法在三个数据集上提高了分类效果：Kaggle鱼类数据集、AWA2以及一个更具挑战性的新的木炭数据集。接着，我们提出了一种自动选择潜在误分类数据的方法。在第二步中，专家被要求对这些数据进行修改，以最终调整预测属性，从而改善分类效果。
cs.CV / 7 / 2609.17753
Beyond Performance Metrics: Uncertainty Mapping of Label Ambiguity in Fazekas Score Prediction
超越性能指标：Fazekas评分预测中标签模糊性的不确定性映射
Abstract
Reference labels used to train medical image classification models are not always as certain as they may appear, and this uncertainty has implications on performance metrics. In this study, we propose a framework to analyze model performance for periventricular Fazekas score prediction that goes beyond conventional metrics. The Fazekas score is an ordinal visual rating scale used to assess the severity of white matter hyperintensities and is known to be affected by inter-rater variability. While the best Fazekas score prediction model achieved a Matthews correlation coefficient (MCC) of 0.70, performance varied across data splits and loss functions, making interpretation of model capabilities difficult. Rather than interpreting epistemic uncertainty of a model's prediction as an isolated scalar value, our approach of uncertainty mapping relates uncertainty to its position within the learned feature representation. This highlights regions of class-boundary transitions where cases appear more ambiguous and misclassifications are more likely. It also identifies potential label disagreement, including low-uncertainty misclassified cases that expert review found to be inconsistent with the original reference Fazekas score. Therefore, uncertainty mapping allows model behaviour to be examined in relation to class separation and potential model-label disagreement. Loss function choice also influenced the uncertainty profile, with some models showing clearer class separation and more localized uncertainty in ambiguous regions than others. These findings suggest that uncertainty mapping for Fazekas score predictions can support model interpretation and targeted dataset review when reference labels are affected by ambiguity/ inter-rater variability.
Chinese Translation
用于训练医学图像分类模型的参考标签并不总是如表面上看起来那样确定，这种不确定性对性能指标有影响。在本研究中，我们提出了一个框架，以分析围脑室Fazekas评分预测模型的性能，超越传统指标。Fazekas评分是一种有序视觉评分量表，用于评估白质高信号的严重程度，并且已知受到评审者间变异性的影响。尽管最佳的Fazekas评分预测模型达到了0.70的Matthews相关系数（MCC），但在数据划分和损失函数之间的性能差异使得模型能力的解释变得困难。我们的不确定性映射方法将模型预测的认知不确定性与其在学习特征表示中的位置相关联，而不是将其解释为孤立的标量值。这突出了类别边界过渡区域，在这些区域中，案例显得更模糊，错误分类的可能性更高。它还识别了潜在的标签不一致，包括专家评审发现与原始参考Fazekas评分不一致的低不确定性错误分类案例。因此，不确定性映射允许在类别分离和潜在模型-标签不一致的关系中检查模型行为。损失函数的选择也影响了不确定性特征，一些模型在模糊区域显示出更清晰的类别分离和更局部化的不确定性。这些发现表明，Fazekas评分预测的不确定性映射可以支持模型解释和针对数据集的有针对性审查，尤其是在参考标签受到模糊性/评审者间变异性影响时。
cs.CV / 8 / 2609.17790
Not All Patches Are Equally Forgettable: Spatially Localized Domain Unlearning in Vision-Language Models
并非所有补丁都是同样容易被遗忘的：视觉语言模型中的空间局部化领域遗忘
Abstract
Pre-trained vision-language models (VLMs) exhibit strong cross-domain recognition performance even without additional training. However, this robustness can also preserve undesirable domain-specific behavior, as domain-related and semantic information often remain entangled within the learned representation space, making selective domain unlearning challenging. Existing approaches typically address this problem through latent-space disentanglement and prompt- or feature-level interventions, without directly attributing and attenuating individual patch-token contributions. However, here we suggest that rather than uniformly suppressing the full representation, it may be more effective to exploit the spatial structure of vision transformers to localize and suppress patch regions that contribute disproportionately to forget-domain prediction. Patches that strongly influence forget-domain prediction may not be equally important for semantic recognition, suggesting that forgetting should be guided according to the domain contribution of different visual regions. Specifically, we propose a two-stage patch-selective framework that first estimates patch-level domain sensitivity and then selectively attenuates patches whose contribution to forget-domain prediction is stronger than their semantic utility. We evaluate our framework on Office-Home, Mini DomainNet, and DomainNet. Experimental results demonstrate improved forgetting-retention tradeoffs compared to prior methods while improving retained-domain recognition by up to 3.8\%. Additional evaluations under visually overlapping and unseen-domain settings further demonstrate improved robustness under distribution shift.
Chinese Translation
预训练的视觉语言模型（VLMs）在跨领域识别性能上表现出色，即使没有额外的训练。然而，这种鲁棒性也可能保留不希望出现的领域特定行为，因为领域相关和语义信息往往在学习的表征空间中交织在一起，使得选择性领域遗忘变得具有挑战性。现有的方法通常通过潜在空间的解缠结以及提示或特征级别的干预来解决这个问题，而没有直接归因和减弱单个补丁标记的贡献。然而，我们在此建议，与其均匀地抑制整个表征，不如利用视觉变换器的空间结构来定位和抑制对遗忘领域预测贡献不成比例的补丁区域。对遗忘领域预测有强烈影响的补丁可能在语义识别中并不重要，这表明遗忘应根据不同视觉区域的领域贡献进行指导。具体而言，我们提出了一种两阶段的补丁选择框架，首先估计补丁级别的领域敏感性，然后选择性地减弱那些对遗忘领域预测的贡献强于其语义效用的补丁。我们在Office-Home、Mini DomainNet和DomainNet上评估了我们的框架。实验结果表明，与之前的方法相比，遗忘与保留的权衡得到了改善，同时保留领域的识别率提高了最多3.8%。在视觉重叠和未见领域设置下的额外评估进一步证明了在分布变化下的鲁棒性得到了改善。
cs.CV / 9 / 2609.17800
AgenTeeth: A Model-Agnostic Framework for Suppressing Hallucination in Frozen Vision-Language Models on Dental X-Rays via Tool Evidence Injection
AgenTeeth：一种模型无关的框架，通过工具证据注入抑制冻结视觉-语言模型在牙科X光片上的幻觉
Abstract
Vision-language models (VLMs) remain largely unreliable on panoramic dental radiographs and can rely on learned anatomical priors rather than evidence in the image. This is particularly problematic for tooth localization and spatial reasoning, and fine-tuned dental VLMs can retain the same spatial biases. We present AgenTeeth, a model-agnostic, tool-augmented framework that grounds frozen VLMs using seven specialized dental vision experts. A question-aware orchestrator selects the relevant tools, whose detections are mapped to FDI tooth numbers or anatomical regions and returned as structured findings together with annotated image overlays. A fresh synthesis call then answers the question using this evidence, without fine-tuning the underlying VLM. On MMOral-OPG-Bench, AgenTeeth improves four backbone VLMs by 12.9-23.0 percentage points over their baselines. Our strongest configuration reaches 65.66% on open-ended VQA, compared with 45.35% for OralGPT-Plus. The advantage also holds at matched scale: a frozen Qwen2.5-VL-7B-Instruct with AgenTeeth reaches 48.11%, exceeding OralGPT-Plus built on the same backbone after supervised fine-tuning and reinforcement learning for tool use. We release the framework, all seven expert models, and a dentist-annotated dataset for alveolar bone-loss detection in panoramic radiographs.
Chinese Translation
视觉-语言模型（VLMs）在全景牙科放射影像上仍然存在较大不可靠性，往往依赖于学习到的解剖先验而非图像中的证据。这对于牙齿定位和空间推理尤其成问题，经过微调的牙科VLMs可能保留相同的空间偏差。我们提出了AgenTeeth，这是一种模型无关的、工具增强的框架，利用七个专业的牙科视觉专家对冻结的VLMs进行基础定位。一个关注问题的协调者选择相关工具，其检测结果被映射到FDI牙齿编号或解剖区域，并作为结构化发现与注释的图像叠加一起返回。然后，通过这一证据进行新的合成调用以回答问题，而无需微调底层的VLM。在MMOral-OPG-Bench上，AgenTeeth使四个基础VLM的表现比基线提高了12.9-23.0个百分点。我们最强的配置在开放式视觉问答（VQA）中达到了65.66%，而OralGPT-Plus为45.35%。这一优势在匹配规模下也得以保持：使用AgenTeeth的冻结Qwen2.5-VL-7B-Instruct达到了48.11%，超过了在相同基础上经过监督微调和强化学习的OralGPT-Plus。我们发布了该框架、所有七个专家模型以及一个用于全景放射影像中牙槽骨丧失检测的牙医注释数据集。
cs.CV / 10 / 2609.17810
Wind on Trees: Testing Physical Grounding in Dynamic 4D Gaussian Splatting
风对树木的影响：在动态4D高斯溅射中测试物理基础
Abstract
Monocular reconstruction of wind-driven vegetation is severely underconstrained: motion along the viewing direction is largely unobservable, a moving canopy offers few reliable correspondences, and nearly the entire scene is dynamic, providing little static reference. Directly-learned deformation fields in 4D Gaussian Splatting therefore optimize photometric consistency rather than recover the motion that produced it. We replace that field with a physically parameterized deformation prior: one damped harmonic oscillator per rigid part, driven by the observed wind and integrated by differentiable RK4, supervised photometrically alone. To test whether such a prior is physically grounded rather than merely well fit, we build a controlled synthetic testbed of three procedurally generated trees spanning an order of magnitude in skeleton complexity, whose per-part natural frequency follows from its own geometry and whose damping ratio is a fixed constant, both held out of training. On it, we measure held-out views, temporal extrapolation, zero-shot transfer to unseen wind speeds, and recovery of the physical parameters themselves. The prior costs appearance fidelity on in-distribution views and extrapolates markedly better outside the training window and the training wind, while parameter recovery is far weaker than it first appears: frequency recovery survives an untrained null control on only the sparsest of the three trees, and damping is not recovered at all.
Chinese Translation
单目重建风驱动的植被受到严重约束：沿视线方向的运动大部分不可观察，移动的树冠提供的可靠对应关系很少，几乎整个场景都是动态的，提供的静态参考极少。因此，在4D高斯溅射中直接学习的变形场优化的是光度一致性，而不是恢复产生这种一致性的运动。我们用一个物理参数化的变形先验替代该变形场：每个刚性部分一个阻尼谐振子，由观察到的风驱动，并通过可微分的RK4进行积分，仅通过光度进行监督。为了测试这种先验是否具有物理基础而不仅仅是拟合良好，我们构建了一个受控的合成测试平台，包含三棵程序生成的树木，跨越一个数量级的骨架复杂性，其每个部分的自然频率源自其自身几何形状，阻尼比是一个固定常数，均未用于训练。在该平台上，我们测量了持出视图、时间外推、对未见风速的零样本迁移以及物理参数本身的恢复。该先验在分布内视图上牺牲了外观保真度，并在训练窗口和训练风之外的外推效果显著更好，而参数恢复的效果远不如最初看起来的那样强：频率恢复仅在三棵树中最稀疏的那棵上存活于未训练的零控制中，而阻尼则完全未被恢复。
cs.CV / 11 / 2609.17814
GazeDiT: Gaze-Accurate Diffusion Image Generation for Eye Tracking via Spatial Conditioning
GazeDiT：通过空间条件实现眼动追踪的注视精确扩散图像生成
Abstract
Diffusion models are increasingly used to generate synthetic training data, but precise label control remains difficult when the conditioning signal is low-dimensional and coarse. Text-conditioned images are judged by broad prompt consistency, whereas supervised training requires precise correspondence between each image and its numerical label. This is challenging in eye tracking, where a 4D binocular gaze is expressed through subtle, spatially localized pupil and iris geometry. We introduce GazeDiT, a diffusion model that generates images for a requested 4D gaze through an internally constructed spatial condition that grounds the global gaze label in this local geometry. During training, a frozen SegFormer extracts pupil/iris geometry from diverse real images, allowing the model to learn realistic appearance conditioned on that geometry. At inference, a physical eye renderer samples gaze-consistent geometries by varying anatomy and camera state, enabling diverse synthesis without a source image. GazeDiT achieves substantially lower tail gaze-label error than other diffusion baselines, approaching the error of the same frozen gaze estimator on real images. Its generated data also improves the downstream eye tracker, reducing gaze error on difficult cases from 3.05{\deg} to 2.80{\deg} in the smallest cohort.
Chinese Translation
扩散模型越来越多地用于生成合成训练数据，但当条件信号低维且粗糙时，精确的标签控制仍然困难。文本条件的图像通过广泛的提示一致性进行评判，而监督训练则要求每个图像与其数值标签之间有精确的对应关系。这在眼动追踪中尤为具有挑战性，因为4D双眼注视通过微妙的、空间局部化的瞳孔和虹膜几何形状来表达。我们提出了GazeDiT，这是一种扩散模型，通过内部构建的空间条件生成请求的4D注视图像，将全局注视标签与该局部几何形状相结合。在训练过程中，一个冻结的SegFormer从多样的真实图像中提取瞳孔/虹膜几何形状，使模型能够学习基于该几何形状的真实外观。在推理阶段，一个物理眼渲染器通过改变解剖结构和相机状态来采样与注视一致的几何形状，从而在没有源图像的情况下实现多样化合成。GazeDiT在尾部注视标签误差方面显著低于其他扩散基线，接近于在真实图像上相同冻结注视估计器的误差。其生成的数据还改善了下游眼动追踪器，将在困难情况下的注视误差从3.05{ extdegree}降低到2.80{ extdegree}，在最小的样本中。
cs.CV / 12 / 2609.17820
CALIPER: Metric-Grounded Model-Free Recognition of Visually Similar Industrial Parts
CALIPER：基于度量的无模型视觉相似工业零件识别
Abstract
Fine-grained recognition of visually similar industrial parts is challenging when classes differ primarily in physical dimensions. Normalizing detected object crops to a fixed input size suppresses absolute scale, while CAD models and large class-specific datasets may be unavailable in evolving industrial inventories. We present CALIPER, a model-free RGB-D framework that couples support-based appearance matching with metric size evidence. Each training class is onboarded from a single turntable RGB-D video and one to two labeled real images; 3D reconstruction provides novel-view appearance support, while aligned depth yields a class-specific metric size profile. At inference, a coarse YOLOv8n-seg model localizes parts, and a frozen DINOv2 backbone with an episodically trained embedding head performs fine-grained support matching. Margin-conditioned metric fusion activates probabilistic size evidence only for appearance-ambiguous decisions. New classes are enrolled from a small RGB-D support set without updating network parameters. We evaluate CALIPER on 18 visually similar industrial parts: 16 classes are used for training, while two screws are reserved for training-free enrollment. CALIPER achieves 88.2% closed-set accuracy with 99.8% localization recall and 85.7% overall accuracy after 10-shot enrollment of the two unseen screws. Metric fusion improves unseen-class accuracy by up to 37.4 percentage points without statistically significant degradation of the original inventory. Robot-arm deployment identifies 17/18 parts without deployment-specific retraining.
Chinese Translation
视觉上相似的工业零件的细粒度识别在类别主要通过物理尺寸区分时具有挑战性。将检测到的物体裁剪规范化为固定输入大小会抑制绝对尺度，而在不断变化的工业库存中，CAD模型和大型特定类别数据集可能不可用。我们提出了CALIPER，这是一种无模型的RGB-D框架，将基于支持的外观匹配与度量大小证据相结合。每个训练类别通过单个转盘RGB-D视频和一到两个标记的真实图像进行引入；3D重建提供了新视角的外观支持，而对齐的深度则生成特定类别的度量大小轮廓。在推理阶段，一个粗略的YOLOv8n-seg模型定位零件，而一个冻结的DINOv2主干网络与一个经过情节训练的嵌入头执行细粒度支持匹配。边际条件度量融合仅在外观模糊的决策中激活概率大小证据。新类别可以通过一个小的RGB-D支持集进行注册，而无需更新网络参数。我们在18个视觉上相似的工业零件上评估CALIPER：16个类别用于训练，而两个螺丝则保留用于无训练注册。CALIPER在对两个未见螺丝进行10次注册后，达到了88.2%的闭集准确率，99.8%的定位召回率和85.7%的整体准确率。度量融合将未见类别的准确率提高了最多37.4个百分点，而对原始库存没有统计显著的降级。机器人臂部署识别了17/18个零件，无需特定于部署的再训练。
cs.CV / 13 / 2609.17843
RoboVAD: A Large Cross-Domain Evaluation Benchmark for Anomaly Detection in Robotic Arm Manipulation Videos
RoboVAD：用于机器人臂操作视频异常检测的大规模跨领域评估基准
Abstract
Video anomaly detection (VAD) is an actively studied task, having wide applications in typical scenarios such as public surveillance and road traffic safety. The task is also relevant for robotic arm interactions, where it has several downstream applications, including learning better interaction and manipulation abilities, triggering recovery procedures when anomalies occur, etc. Despite its relevance, the exploration of anomaly detection in robotic arm manipulation videos is limited by the low number of available resources. To this end, we introduce RoboVAD, a large-scale benchmark for video anomaly detection that comprises challenging cross-domain evaluation scenarios, where certain actions (tasks executed by a robotic arm) and anomaly types (mistakes that occur while performing certain tasks) remain unseen during training. RoboVAD is designed to benchmark VAD methods in realistic scenarios, where robotic arms can perform unforeseen tasks, and thereby encounter new anomaly types. We train and evaluate several state-of-the-art VAD methods, including a novel method specifically adapted for robotic arm manipulation. While the proposed method outperforms many state-of-the-art competitors, all methods remain below a micro-averaged frame-level AUC threshold of 70% in the most challenging evaluation setup, confirming the difficulty of the proposed benchmark. We publicly release our dataset and code at https://zenodo.org/records/22754659.
Chinese Translation
视频异常检测（VAD）是一个积极研究的任务，广泛应用于公共监控和道路交通安全等典型场景。该任务在机器人臂交互中也具有相关性，涉及多个下游应用，包括学习更好的交互和操作能力、在发生异常时触发恢复程序等。尽管其相关性明显，但在机器人臂操作视频中的异常检测探索受到可用资源数量有限的限制。为此，我们提出了RoboVAD，一个大规模的视频异常检测基准，包含具有挑战性的跨领域评估场景，其中某些动作（机器人臂执行的任务）和异常类型（在执行某些任务时发生的错误）在训练过程中未曾见过。RoboVAD旨在基准化VAD方法在现实场景中的表现，在这些场景中，机器人臂能够执行不可预见的任务，从而遇到新的异常类型。我们训练和评估了几种最先进的VAD方法，包括一种专门适应于机器人臂操作的新方法。尽管所提出的方法在许多最先进的竞争者中表现优异，但所有方法在最具挑战性的评估设置中仍低于70%的微平均帧级AUC阈值，确认了所提出基准的难度。我们在https://zenodo.org/records/22754659上公开发布了我们的数据集和代码。
cs.CV / 14 / 2609.17856
Investigating Adversarial Robustness of Heterogeneous Cooperative Perception
异构协同感知的对抗鲁棒性研究
Abstract
Heterogeneous cooperative perception (CP) enables connected vehicles with diverse sensor setups to share spatial awareness via compact feature maps, where receivers reconcile these maps using learned translation modules for fusion and inference. Prior attacks against CP in a homogeneous setting reveal that the data exchange introduces a critical attack surface: a single malicious agent can transmit crafted features that erase real objects from a neighbor's fused scene. Yet, it is widely hypothesized that heterogeneity naturally defends against these attacks, as the attacker lacks knowledge of the victim's detector and the translation module scrambles adversarial gradients. We demonstrate that this protection is largely an illusion. Using a matched-objective harness to standardize the perturbation budget, objective, and forward path, we show that properly tuned iterative attacks close or reverse the apparent robustness gap. However, these optimization-based attacks require ground-truth labels and iterative backpropagation, meaning they do not represent a practical field threat running in real-time. To bridge this gap, we introduce HetPoison, a learned generator that crafts a removal perturbation in a single, label-free forward pass. HetPoison transfers across major heterogeneous designs without requiring access to the victim's detector, matching or exceeding the effectiveness of expensive optimizer-based attacks. Since heterogeneity itself is not a defense, we propose HetShield, a lightweight trust layer that validates the spatiotemporal consistency across features, recovering 83--95% of the accuracy degraded by attacks, outperforming prior art.
Chinese Translation
异构协同感知（CP）使得配备不同传感器的联网车辆能够通过紧凑的特征图共享空间意识，接收者利用学习到的转换模块对这些特征图进行融合和推理。然而，以往针对同质环境下CP的攻击表明，数据交换引入了一个关键的攻击面：单个恶意代理可以传输精心构造的特征，从而在邻近车辆融合的场景中抹去真实物体。然而，广泛假设异构性自然能够防御这些攻击，因为攻击者缺乏对受害者探测器的知识，而转换模块则会扰乱对抗梯度。我们证明这种保护在很大程度上是一种幻觉。通过使用匹配目标的工具来标准化扰动预算、目标和前向路径，我们展示了适当调整的迭代攻击可以缩小或逆转表面上的鲁棒性差距。然而，这些基于优化的攻击需要真实标签和迭代反向传播，这意味着它们并不代表在实时环境中运行的实际威胁。为了解决这一问题，我们引入了HetPoison，一个学习生成器，可以在单次无标签的前向传递中构造去除扰动。HetPoison能够在主要的异构设计中迁移，而无需访问受害者的探测器，其效果与昂贵的基于优化的攻击相匹配或超过。由于异构性本身并不是一种防御机制，我们提出了HetShield，一个轻量级的信任层，用于验证特征之间的时空一致性，恢复了83%至95%因攻击而降低的准确性，超越了以往的研究成果。
cs.CV / 15 / 2609.17868
Lumen: Parameter-Efficient Alignment of Pretrained Vision and Language Encoders for Zero-Shot Computational Pathology
Lumen：预训练视觉与语言编码器的参数高效对齐用于零-shot 计算病理
Abstract
Pathology vision-language models are commonly built by pretraining or fine-tuning large encoders on paired image-caption data. We asked whether a pathology vision-language model can instead be assembled by parameter-efficient alignment of frozen unimodal foundation models, leaving their pretrained representations untouched. Here we present Lumen, which aligns frozen Virchow2 and BioMedBERT backbones using rank-4 adapters and projection heads, training only 0.40% of the total parameters on the public QUILT-1M corpus. Across nine public zero-shot patch benchmarks, Lumen achieved the highest mean chance-corrected balanced accuracy, 0.546 versus 0.461 for the strongest baseline (paired difference 0.086, 95% CI 0.042-0.136). On lymph-node metastasis detection, Lumen reached an AUROC of 0.964 (95% CI 0.956-0.971) on 4,214 held-out internal slides and 0.955 (95% CI 0.942-0.966) on 2,368 slides across nine external cohorts and six organs. At the internally calibrated threshold, it outperformed all vision-language baselines, with a balanced accuracy of 0.909 (95% CI 0.896-0.923) internally and 0.915 (95% CI 0.902-0.929) externally. Lumen performed competitively across the evaluations, with the exception of cross-modal retrieval, where it ranked third behind CONCH and PathGen-L/14. Fully fine-tuning both encoders gave Lumen no consistent benefit over low-rank adaptation, although it improved retrieval. Aligning frozen unimodal foundation models therefore yields strong and transferable performance at patch and slide level while training only a small fraction of the parameters.
Chinese Translation
病理视觉-语言模型通常通过在配对的图像-标题数据上预训练或微调大型编码器构建。我们探讨了是否可以通过对冻结的单模态基础模型进行参数高效的对齐来组装病理视觉-语言模型，而不触及其预训练表示。在此，我们提出了Lumen，它使用四阶适配器和投影头对冻结的Virchow2和BioMedBERT骨干进行对齐，仅在公共QUILT-1M语料库上训练了0.40%的总参数。在九个公共零-shot补丁基准测试中，Lumen达到了最高的均值机会校正平衡准确率，为0.546，相较于最强基线的0.461（配对差异0.086，95%置信区间0.042-0.136）。在淋巴结转移检测中，Lumen在4214个保留的内部切片上达到了0.964的AUROC（95%置信区间0.956-0.971），在2368个来自九个外部队列和六个器官的切片上达到了0.955（95%置信区间0.942-0.966）。在内部校准的阈值下，它在所有视觉-语言基线中表现优异，内部平衡准确率为0.909（95%置信区间0.896-0.923），外部为0.915（95%置信区间0.902-0.929）。Lumen在各项评估中表现竞争力，唯一例外是跨模态检索，在该任务中其排名第三，仅次于CONCH和PathGen-L/14。完全微调两个编码器并未给Lumen带来相对于低秩适配的一致优势，尽管它改善了检索。因此，对冻结的单模态基础模型进行对齐在补丁和切片级别上产生了强大且可转移的性能，同时仅训练了少量参数。
cs.CV / 16 / 2609.17882
Can VLMs Reliably Assess Sidewalk Accessibility Attributes from Pedestrian-Level Imagery?
视觉语言模型能否可靠地评估人行道无障碍属性？
Abstract
An important component of urban accessibility, particularly for wheelchair users and people with reduced mobility, is sidewalk compliance with measurable requirements. We test whether effective width, longitudinal slope, cross slope, and pavement condition can be assessed reliably from pedestrian-level imagery using vision-language models (VLMs). We present the first application of sampling-based conformal prediction (CP) for VLM-based accessibility assessment. We evaluate four VLMs on 514 sidewalk images from Seoul, South Korea, with field-measured ground truth. Conformal calibration attains the nominal 90% coverage for all models and attributes, but the calibrated regions differ in informativeness. Effective width yields the most informative estimates, with a mean interval half-width of about 1.0 m for the best model. Since every model overestimates width, asymmetric calibration shortens the intervals by up to 33% at unchanged coverage. Longitudinal slope is marginally informative, cross-slope intervals are too wide to resolve regulatory thresholds, and pavement-condition sets degenerate to all five grades (A-E) for three of the four models. Uncalibrated intervals from raw sampling dispersion cover only 17-47% of field-measured values at a nominal 90% level. Among the images with the most self-consistent responses, these intervals miss the field-measured value in up to 96% of cases. Response self-consistency is therefore not evidence of accuracy, and sampling dispersion cannot be interpreted as uncertainty until it has been calibrated against field-measured ground truth. No quantitative attribute reaches the precision required for general compliance assessment, but CP identifies from calibration data alone which attributes can support screening of segments far from the thresholds. We release the annotated pedestrian-level images and their corresponding field-measured attribute values.
Chinese Translation
城市无障碍的一个重要组成部分，特别是对于轮椅用户和行动不便的人士，是人行道是否符合可测量的要求。我们测试了有效宽度、纵向坡度、横向坡度和路面状况是否可以通过使用视觉语言模型（VLMs）从行人视角的图像中可靠地评估。我们展示了基于采样的符合预测（CP）在基于VLM的无障碍评估中的首次应用。我们在来自韩国首尔的514张人行道图像上评估了四个VLM，使用实地测量的真实值进行验证。符合校准在所有模型和属性上达到了名义上的90%覆盖率，但校准区域在信息量上存在差异。有效宽度提供了最具信息量的估计，最佳模型的平均区间半宽度约为1.0米。由于每个模型都高估了宽度，非对称校准在不改变覆盖率的情况下将区间缩短了最多33%。纵向坡度的信息量边际，横向坡度的区间过宽，无法解决监管阈值，而路面状况的分类在四个模型中的三个退化为所有五个等级（A-E）。未经校准的原始采样分散区间在名义90%的水平下仅覆盖了17-47%的实地测量值。在响应最自洽的图像中，这些区间在多达96%的情况下未能覆盖实地测量值。因此，响应自洽性并不能证明准确性，采样分散在未经过实地测量的真实值校准之前无法被解释为不确定性。没有任何定量属性达到了进行一般合规评估所需的精度，但CP仅通过校准数据识别出哪些属性可以支持远离阈值的区段筛选。我们发布了标注的人行道图像及其对应的实地测量属性值。
cs.CV / 17 / 2609.17909
Zing-0.5: Toward Playable Worlds with Real-Time Joint Action and Text Control
Zing-0.5：迈向具有实时联合行动和文本控制的可玩世界
Abstract
We introduce Zing-0.5, a 5B autoregressive world model designed for playability: users can explore generated worlds, influence unfolding events, and respond to the resulting feedback through joint keyboard and online text control. Our approach brings together three technical contributions: (1) Unified action and text conditioning, combining magnitude-aware keyboard inputs with temporally aligned text instructions and jointly annotated videos to learn navigation and event control within the same sequence; (2) Event-scale supervision for incremental generation, using a segment-level teacher trained on connected multi-prompt videos to supervise a block-level causal student through distribution-matching distillation; and (3) Low-cost real-time interaction, combining four-step generation with context-preserving streaming to support 832 x 480 inference at 24 FPS at an estimated server rental cost of approximately USD 0.009 per stream-minute. Zing-0.5 achieves an overall score of 81.0 and a consistency score of 88.5 across 158 WBench Navigation cases. A joint-control demonstration shows a text-directed event change during continued navigation without restarting generation. We release the model weights, inference code, and Zing-SGLang serving implementation to support further work on playable generated worlds.
Chinese Translation
我们介绍了Zing-0.5，这是一种设计用于可玩性的5B自回归世界模型：用户可以探索生成的世界，影响正在展开的事件，并通过联合键盘和在线文本控制来响应所产生的反馈。我们的方法结合了三项技术贡献：（1）统一的行动和文本条件，结合了对幅度敏感的键盘输入、时间对齐的文本指令以及共同注释的视频，以在同一序列中学习导航和事件控制；（2）事件规模的监督以实现增量生成，使用在连接的多提示视频上训练的段级教师，通过分布匹配蒸馏来监督块级因果学生；（3）低成本的实时交互，结合四步生成与上下文保持流媒体，以支持在每分钟约0.009美元的服务器租赁成本下以24帧每秒的速度进行832 x 480的推理。Zing-0.5在158个WBench导航案例中获得了81.0的总体评分和88.5的一致性评分。一个联合控制的演示展示了在持续导航过程中，文本引导的事件变化，而无需重新开始生成。我们发布了模型权重、推理代码和Zing-SGLang服务实现，以支持对可玩生成世界的进一步研究。
cs.CV / 18 / 2609.17913
Face-voice Association across LAnguages and Gender (FLAG) 2027 Challenge Evaluation Plan
跨语言和性别的面部-声音关联（FLAG）2027挑战评估计划
Abstract
Face--voice association models may rely on language or gender cues in the voice rather than on speaker-specific voice characteristics, which can lead to a performance deterioration when the model has to identify a multilingual speaker or distinguis same-gender speakers. To investigate these issues, we introduce the Face-voice Association across LAnguages and Gender (FLAG) 2027 Challenge. The challenge formulates face--voice association as a cross-modal verification task: given a voice, identify the speaker's face from a ``gallery'' of faces consisting of the speaker's face and a set of negative samples. Models are evaluated on identities not present in the training data (``unseen'') and both for languages present or absent from the training data (``heard'' and ``unheard''). Two evaluation settings are used to test models' reliance on gender: a standard, unconstrained and a gender-constrained one, where the latter uses a same-gender gallery. The performance of existing, baseline models in these settings reveals that models performance degrades under language shifts and in gender-constrained settings, highlighting the need to foster the development of models that capture identity-specific aspects beyond language and gender. The challenge provides a benchmark dataset, pretrained baseline models, and an evaluation framework to advance face--voice association.
Chinese Translation
面部-声音关联模型可能依赖于声音中的语言或性别线索，而非说话者特定的声音特征，这可能导致模型在识别多语言说话者或区分同性别说话者时性能下降。为探讨这些问题，我们提出了跨语言和性别的面部-声音关联（FLAG）2027挑战。该挑战将面部-声音关联形式化为一种跨模态验证任务：给定一个声音，从包含说话者面部和一组负样本的“画廊”中识别说话者的面部。模型在训练数据中不存在的身份（“未见”）以及在训练数据中存在或不存在的语言（“已听”和“未听”）上进行评估。使用两种评估设置来测试模型对性别的依赖：一种是标准的、不受限制的设置，另一种是性别受限的设置，其中后者使用同一性别的画廊。这些设置中现有基线模型的性能揭示了模型在语言变化和性别受限设置下的性能下降，突显了开发能够捕捉超越语言和性别的身份特征模型的必要性。该挑战提供了基准数据集、预训练基线模型和评估框架，以推动面部-声音关联的发展。
cs.CV / 19 / 2609.17923
Audio for Sports Highlight Detection: A Comparative Empirical Study
体育精彩瞬间检测中的音频：一项比较实证研究
Abstract
Sports highlight detection aims to identify the most exciting and meaningful moments from long sports videos. While existing methods often emphasize visual or visual-language representations, sports videos contain rich audio cues, including commentator speech, crowd reactions, whistles, ball impacts, and referee calls. In this work, we revisit the role of audio in sports highlight detection and ask a simple question: how far can audio alone go? We construct lightweight audio-only baselines using pretrained audio representations and compare them with visual-only and audio-visual methods on the SV-Highlights benchmark. Surprisingly, our audio-only GRU baseline achieves strong performance and outperforms several existing audio-visual methods under our supervised evaluation setting. Furthermore, a simple audio-visual fusion baseline achieves the best performance across all metrics, indicating that audio and visual cues provide complementary information. To better understand the contribution of audio, we conduct source-separated analysis and show that vocal/commentary audio is more informative than background-only audio, while their combination performs best. We also analyze interpretable audio cues and find that highlight clips exhibit higher RMS loudness, peak loudness, and mid-frequency energy than non-highlight clips, although substantial distribution overlap indicates that loudness alone is insufficient. Our findings suggest that audio is an underexplored but highly informative modality for sports highlight detection and should be treated as a primary signal rather than merely an auxiliary cue.
Chinese Translation
体育精彩瞬间检测旨在从长时间的体育视频中识别出最激动人心和最有意义的时刻。尽管现有方法通常强调视觉或视觉-语言表示，但体育视频中包含丰富的音频线索，包括解说员的讲话、观众的反应、哨声、球的撞击声和裁判的判罚声。在本研究中，我们重新审视音频在体育精彩瞬间检测中的作用，并提出一个简单的问题：仅凭音频能达到什么程度？我们构建了基于预训练音频表示的轻量级音频基线，并将其与视觉单一和音频-视觉方法在SV-Highlights基准上进行比较。令人惊讶的是，我们的音频单一GRU基线表现出色，并在我们的监督评估设置下超越了几种现有的音频-视觉方法。此外，一个简单的音频-视觉融合基线在所有指标上都达到了最佳性能，表明音频和视觉线索提供了互补的信息。为了更好地理解音频的贡献，我们进行了源分离分析，结果表明，语音/解说音频比仅背景音频更具信息量，而它们的组合表现最佳。我们还分析了可解释的音频线索，发现精彩片段的均方根响度、峰值响度和中频能量均高于非精彩片段，尽管显著的分布重叠表明，仅凭响度不足以区分。我们的研究结果表明，音频是一个尚未充分探索但信息量极大的模态，应该被视为体育精彩瞬间检测中的主要信号，而不仅仅是辅助线索。
cs.CV / 20 / 2609.17925
Rapid Loss of the Sierra Nevada's Largest Trees Driven by Fire
火灾驱动的内华达山脉最大树木的快速损失
Abstract
Large trees disproportionately contribute to biomass storage, habitat structure, and ecosystem functioning. However, their distribution and health dynamics remain poorly quantified at a regional scale. Here, a deep learning model (U-Net-ID) and canopy height models derived from sub-meter aerial imagery from 2020 were used to delineate all individual trees with crown area $\geq$ 100 m$^2$ across the Sierra Nevada Floristic Province. The model was trained using more than 3.3 million synthetic tree crowns and achieved a median Intersection over Union (IoU) of 0.602 when validated against an independent dataset of 20,273 crowns. A total of 6,515,705 large trees were mapped, occurring across approximately 78.7% of the Sierra Nevada Floristic Province. The spatial distribution of large trees showed associations with elevation, temperature, and precipitation. Using Sentinel-2 time series from 2020 to 2025, tree health dynamics were characterized by extracting spectral trajectories for each crown and applying BFAST breakpoint detection algorithm combined with a disturbance classification framework to identify mortality, disturbance, and recovery trajectories of individual trees. Wildfires, estimated from CAL FIRE fire perimeters, were identified as the dominant driver of large-tree mortality, killing 10% of all large trees in the Sierra Nevada, with mortality strongly concentrated during the extreme 2020-2021 fire seasons.
Chinese Translation
大型树木在生物量储存、栖息地结构和生态系统功能中占有不成比例的贡献。然而，它们的分布和健康动态在区域尺度上仍然缺乏充分量化。在本研究中，使用深度学习模型（U-Net-ID）和基于2020年亚米级航空影像生成的树冠高度模型，描绘了内华达山脉植物区内所有树冠面积 $
geq$ 100 m$^2$ 的单棵树木。该模型使用超过330万个合成树冠进行训练，并在与20,273个树冠的独立数据集验证时，达到了0.602的中位数交并比（IoU）。共绘制了6,515,705棵大型树木，分布在内华达山脉植物区约78.7%的区域。大型树木的空间分布与海拔、温度和降水量存在关联。通过使用2020年至2025年的Sentinel-2时间序列，提取每个树冠的光谱轨迹，并结合BFAST断点检测算法与干扰分类框架，识别个别树木的死亡、干扰和恢复轨迹，特征化树木健康动态。根据CAL FIRE火灾边界估算，野火被确定为大型树木死亡的主要驱动因素，导致内华达山脉10%的大型树木死亡，且死亡现象在极端的2020-2021年火灾季节中高度集中。
cs.CV / 21 / 2609.17953
EDCT-Bench: Uncovering Faithfulness Gaps in VLMs via Explanation-Driven Counterfactual Testing
EDCT-Bench：通过解释驱动的反事实测试揭示视觉语言模型中的可信度差距
Abstract
Vision-Language Models (VLMs) can produce Natural Language Explanations (NLEs) that sound plausible yet remain inconsistent with the visual evidence they cite. We present Explanation-Driven Counterfactual Testing (EDCT), an intervention-based protocol that extracts visual concepts cited in a model's explanation, applies verified minimal edits to them, and tests whether the resulting answer and explanation remain consistent with the edited image. Using this protocol, we create EDCT-Bench, a comprehensive benchmark spanning three complementary domains: knowledge-intensive visual question answering (OK-VQA), safety-critical driving (DriveLM), and 3D spatial reasoning (3DSRBench). Across the evaluated VLMs, EDCT reveals substantial faithfulness gaps, with models frequently producing responses inconsistent with verified visual changes. Finally, our fine-tuning study suggests that EDCT-generated counterfactuals provide high-impact training signals.
Chinese Translation
视觉语言模型（VLMs）能够生成听起来合理但与其引用的视觉证据不一致的自然语言解释（NLEs）。我们提出了解释驱动的反事实测试（EDCT），这是一种基于干预的协议，提取模型解释中引用的视觉概念，对其进行经过验证的最小编辑，并测试结果的答案和解释是否与编辑后的图像保持一致。通过该协议，我们创建了EDCT-Bench，这是一个涵盖三个互补领域的综合基准：知识密集型视觉问答（OK-VQA）、安全关键驾驶（DriveLM）和三维空间推理（3DSRBench）。在评估的VLMs中，EDCT揭示了显著的可信度差距，模型经常生成与经过验证的视觉变化不一致的响应。最后，我们的微调研究表明，EDCT生成的反事实提供了高影响力的训练信号。
cs.CV / 22 / 2609.18007
Newer Is Not Fairer: Gender Stereotyping in Text-to-Image AI Across Model Generations
更新的模型并不更公平：跨模型世代的文本到图像人工智能中的性别刻板印象
Abstract
Text-to-image generative models are widely used in professional and creative settings, yet how they represent gender across occupations -- and whether newer models are fairer -- remains poorly understood across multiple generations. We evaluate gender representation across 20 occupations, 5 prompt templates, and 4 Stable Diffusion model generations (SD 1.5, SD 2.1, SDXL, SD 3 Medium), generating 8,000 images with n = 100 per occupation-model cell (5 prompts x 20 images), and classifying all with DeepFace. Across the 8,000 open-source images, 76.4% show male subjects (95% CI [75.1%, 78.7%], p < 2.2 x 10^-16, Benjamini-Hochberg adjusted). More strikingly, 57.6% of images for historically female-coded occupations show male subjects (raw p = 3.43 x 10^-22, BH-adjusted p = 1.71 x 10^-21). All nine significant tests reported in this paper survive BH correction across 10 tests. When compared against U.S. Bureau of Labor Statistics workforce data, models underrepresent women by 20-46pp on average, with particularly large deviations for near gender-balanced occupations: scientist (48% female in BLS, 82-99% male in model outputs) and cleaner (46% female in BLS, 80-92% male in outputs). Model generations do not improve steadily: bias worsens from SD 1.5 to SDXL before partially recovering in SD 3 Medium. A preliminary comparison with GPT-image-1 on five occupations suggests lower bias than open-source models, though the practical effect is small (Cramer's V = 0.080) and the comparison is exploratory. No model achieves gender parity.
Chinese Translation
文本到图像生成模型在专业和创意环境中被广泛使用，但它们在不同职业中如何表现性别——以及更新的模型是否更公平——在多个世代中仍然未得到充分理解。我们评估了20个职业、5个提示模板和4个Stable Diffusion模型世代（SD 1.5、SD 2.1、SDXL、SD 3 Medium）中的性别表现，生成了8,000张图像，每个职业-模型单元生成100张图像（5个提示 x 20张图像），并使用DeepFace对所有图像进行分类。在这8,000张开源图像中，76.4%的图像显示为男性主体（95%置信区间 [75.1%, 78.7%]，p < 2.2 x 10^-16，Benjamini-Hochberg校正）。更引人注目的是，57.6%历史上被认为是女性职业的图像显示为男性主体（原始p = 3.43 x 10^-22，BH校正p = 1.71 x 10^-21）。本文报告的九个显著性测试在10个测试中均通过BH校正。与美国劳工统计局的劳动力数据相比，模型对女性的代表性平均低于20-46个百分点，尤其是在性别平衡的职业中偏差较大：科学家（BLS中女性占48%，模型输出中男性占82-99%）和清洁工（BLS中女性占46%，输出中男性占80-92%）。模型世代的表现并未稳步改善：从SD 1.5到SDXL偏见加剧，然后在SD 3 Medium中部分恢复。与GPT-image-1在五个职业上的初步比较表明，其偏见低于开源模型，尽管实际效果较小（Cramer's V = 0.080），且该比较为探索性。没有任何模型实现性别平衡。
cs.CV / 23 / 2609.18034
IRIS: Implicit Rendering Matters for Pose-Free Novel View Synthesis
IRIS：隐式渲染对无姿态新视图合成的重要性
Abstract
Novel view synthesis from unposed multi-view images remains challenging, as the model must jointly learn scene representations and camera parameters without pose supervision. Existing approaches largely fall into two extremes: implicit latent-space rendering is flexible and easy to optimize, but often yields weakly grounded camera estimation; explicit 3D representations provide stronger geometric grounding, but introduce heavier parameterization and more fragile optimization. In this paper, we present IRIS, a fully self-supervised framework that provides a practical middle ground between these two paradigms. Instead of decoding free latent tokens or reconstructing fully explicit 3D primitives, IRIS represents the scene as a latent neural field and renders novel views by querying this field under self-predicted cameras. Specifically, projected features from reference views are aggregated at sampled 3D points to form point-wise latent features, which are then composed along target rays for rendering. This design preserves the flexibility and optimization stability of implicit modeling, while introducing stronger geometric structure than unconstrained latent rendering. Extensive experiments show that IRIS achieves strong novel view synthesis quality with competitive pose accuracy under fully self-supervised learning. Our project page: https://leo-frank.github.io/IRIS
Chinese Translation
从无姿态的多视图图像合成新视图仍然具有挑战性，因为模型必须在没有姿态监督的情况下共同学习场景表示和相机参数。现有的方法大致分为两个极端：隐式潜在空间渲染灵活且易于优化，但通常导致相机估计的基础较弱；显式3D表示提供了更强的几何基础，但引入了更重的参数化和更脆弱的优化。在本文中，我们提出了IRIS，一个完全自监督的框架，提供了这两种范式之间的实用中间方案。IRIS并不是解码自由的潜在标记或重建完全显式的3D原语，而是将场景表示为潜在神经场，并通过查询在自我预测的相机下渲染新视图。具体来说，从参考视图投影的特征在采样的3D点处聚合，以形成逐点的潜在特征，然后沿目标光线组合以进行渲染。这一设计保留了隐式建模的灵活性和优化稳定性，同时引入了比不受约束的潜在渲染更强的几何结构。大量实验表明，IRIS在完全自监督学习下实现了强大的新视图合成质量，并具有竞争力的姿态准确性。我们的项目页面：https://leo-frank.github.io/IRIS
cs.CV / 24 / 2609.18037
SetPlanner: A Lightweight Plug-in Point-Set Planner for Frozen SAM
SetPlanner：一种轻量级插件点集规划器用于冻结的SAM
Abstract
Segment Anything Models provide reusable priors, yet they require user prompts and cannot support fully automatic instrument segmentation. Automatic prompting is difficult for thin, articulated, reflective, and partly occluded tools, where several configurations can be valid. We formulate automatic prompting as lightweight point-set planning and isolate the point source under a frozen pathway. To this end, we present SetPlanner, a 1.52M-parameter plug-in point-set planner for frozen SAM. The plug-in preserves SAM's point-prompt interface and enables reuse across backbones. SetPlanner plans complete unordered K-point sets from geometry-aware targets with a permutation-aware conditional flow. SAM decodes eight candidates; their consensus readout yields a ground-truth-free prediction. Across three endoscopic datasets, SetPlanner wins all six transfer routes over a LoRA-adapted system. Under our frozen-pathway protocol, SetPlanner reaches 0.934 Dice on Kvasir-Instrument and recovers 96% of a 44.4-point localization gap, while candidate disagreement ranks low-Dice cases at AUROC 0.969.
Chinese Translation
Segment Anything Models（SAM）提供可重用的先验知识，但它们需要用户提示，并且无法支持完全自动的工具分割。对于细长、关节式、反射性和部分遮挡的工具，自动提示是困难的，因为多个配置可能是有效的。我们将自动提示形式化为轻量级点集规划，并在冻结路径下隔离点源。为此，我们提出了SetPlanner，一种具有1.52M参数的插件点集规划器，用于冻结的SAM。该插件保留了SAM的点提示接口，并支持在不同骨干网络之间的重用。SetPlanner从几何感知目标规划完整的无序K点集，采用基于排列感知的条件流。SAM解码八个候选项；它们的共识读取产生无真实值的预测。在三个内窥镜数据集中，SetPlanner在六条迁移路径上均优于LoRA适配系统。在我们的冻结路径协议下，SetPlanner在Kvasir-Instrument上达到了0.934的Dice系数，并恢复了96%的44.4点定位差距，而候选项的不一致性将低Dice案例的AUROC评估为0.969。
cs.CV / 25 / 2609.18038
CoAtNet-DeepMoE: A Convolution-Attention Hybrid with DeepSeek Mixture-of-Experts for Parameter-Efficient Tomato Disease Classification
CoAtNet-DeepMoE：一种结合卷积与注意力机制的深度混合专家模型，用于高效的番茄病害分类
Abstract
The world population is growing rapidly, and technology is improving in parallel. Meeting the huge demand for food for these 7 billion people not only depends on increasing food production but also on reducing food loss. Crop losses due to disease affect both the food supply and the financial and economic stability of a country. Tomatoes are among the top food-producing crops globally, and a significant portion of this production is lost due to disease. People have used Machine Learning techniques for feature extraction and early diagnosis of tomato diseases, and nowadays, Deep Learning-based models are widely used for disease recognition. However, most existing models are highly parameter-intensive, which increases the time required for training and inference. As a result, while lightweight models are more suitable for user-friendly applications, they often show a reduction in performance. To balance performance and model size, we propose CoAtNet-DeepMoE, a Convolution-Attention hybrid architecture for rich feature extraction, further enhanced with a DeepSeek Mixture of Experts to substantially reduce the number of parameters without sacrificing accuracy. We evaluate our model on both balanced and imbalanced datasets from Kaggle and PlantVillage, demonstrating robustness and achieving 99.80% accuracy, 99.80% precision, 99.80% recall, and 99.80% F1-score on Kaggle, and 99.83% accuracy, 99.85% precision, 99.76% recall, and 99.80% F1-score on PlantVillage, representing state-of-the-art performance with only 2.47M parameters. The source code will be available at https://github.com/nadimbrur/CoAt-MoE.
Chinese Translation
全球人口正在快速增长，技术也在同步进步。满足这70亿人口对食品的巨大需求不仅依赖于增加食品生产，还需要减少食品损失。由于病害导致的作物损失影响了食品供应以及一个国家的金融和经济稳定。番茄是全球主要的食品生产作物之一，而由于病害造成的生产损失占据了相当大的一部分。人们已经使用机器学习技术进行特征提取和番茄病害的早期诊断，如今基于深度学习的模型被广泛应用于病害识别。然而，大多数现有模型参数密集，增加了训练和推理所需的时间。因此，尽管轻量级模型更适合用户友好的应用，但它们往往表现出性能下降。为了平衡性能与模型大小，我们提出了CoAtNet-DeepMoE，一种结合卷积与注意力机制的混合架构，旨在进行丰富的特征提取，并通过深度混合专家（DeepSeek Mixture of Experts）进一步增强，以显著减少参数数量而不牺牲准确性。我们在Kaggle和PlantVillage的平衡和不平衡数据集上评估了我们的模型，展示了其鲁棒性，并在Kaggle上达到了99.80%的准确率、99.80%的精确率、99.80%的召回率和99.80%的F1-score，在PlantVillage上达到了99.83%的准确率、99.85%的精确率、99.76%的召回率和99.80%的F1-score，代表了仅有2.47M参数的最先进性能。源代码将可在https://github.com/nadimbrur/CoAt-MoE获取。
cs.CV / 26 / 2609.18056
Position Anchor Tuning: Towards Efficient Adaptation of Pre-Trained Point Cloud Transformers
位置锚点调优：朝着高效适应预训练点云变换器
Abstract
Parameter-efficient fine-tuning (PEFT) has recently emerged as a pivotal research direction for adapting pre-trained point cloud transformers to diverse downstream tasks. Although existing methods achieve excellent fine-tuning performance with high parameter efficiency, they ignore inference efficiency. To tackle this problem, a novel PEFT method termed position anchor tuning (PAT) is proposed in this paper. As multi-head attention (MHA) and feed-forward network (FFN) are computation-heavy blocks in pre-trained transformers, PAT decreases their computational cost through token aggregation-expansion pairs. Each pair comprises a token aggregation module (TAM) and a token expansion module (TEM). For MHA and FFN blocks, TAMs extract representative tokens from their input tokens based on position anchors in 3D space. These extracted tokens, rather than the original input tokens, are processed by the blocks, thereby reducing the number of tokens involved in computation. Then, TEMs propagate the learned representations back to the original input tokens. Since TAMs are solely responsible for capturing task-specific representations, base-sharing low-rank adaptation (BSLoRA) is further introduced to enable them to learn such representations effectively with only a small number of trainable parameters. Extensive experiments on widely used benchmarks demonstrate that PAT performs comparably to state-of-the-art methods while incurring significantly lower computational overhead and fewer trainable parameters.
Chinese Translation
参数高效微调（PEFT）最近成为适应预训练点云变换器以应对多样化下游任务的关键研究方向。尽管现有方法在高参数效率下实现了优异的微调性能，但它们忽视了推理效率。为了解决这个问题，本文提出了一种新颖的PEFT方法，称为位置锚点调优（PAT）。由于多头注意力（MHA）和前馈网络（FFN）是预训练变换器中计算密集的模块，PAT通过令牌聚合-扩展对降低了它们的计算成本。每对包括一个令牌聚合模块（TAM）和一个令牌扩展模块（TEM）。对于MHA和FFN模块，TAM根据3D空间中的位置锚点从输入令牌中提取代表性令牌。这些提取的令牌而非原始输入令牌被模块处理，从而减少了参与计算的令牌数量。然后，TEM将学习到的表示传播回原始输入令牌。由于TAM仅负责捕捉特定任务的表示，进一步引入了基于共享低秩适应（BSLoRA），使其能够有效地学习这些表示，仅需少量可训练参数。在广泛使用的基准上进行的广泛实验表明，PAT在计算开销显著降低和可训练参数更少的情况下，与最先进的方法表现相当。
cs.CV / 27 / 2609.18058
Finder: Agentic Closed-Loop Object Finding for Embodied Grounding
Finder：用于具身基础的主动闭环物体查找
Abstract
Finding the object referred to by language in a partially observed 3D scene is a core capability for embodied agents. Existing approaches either couple object search with online exploration, which can be costly when relevant observations have already been captured, or query pre-built open-vocabulary maps and scene graphs in a static, one-shot fashion. We present Finder, an agentic closed-loop object-finding primitive for embodied grounding. Instead of treating grounding as passive retrieval from a fixed scene representation, Finder maintains a typed loop state that links query-conditioned planning, scoped evidence gathering, candidate verification, and accept/continue/abort control. When evidence is incomplete or ambiguous, the loop can redirect subsequent perception and comparison rather than simply returning the top retrieved object. On open-vocabulary embodied Object Retrieval in Habitat/HM3D and real-world RGB-D scenes, Finder improves the averaged 1m success rate by 15.75 points over strong baselines. The same primitive also transfers to sequential object grounding and embodied object-centric question answering, improving spatial and temporal localization without changing the inner grounding protocol. Project page: https://finder-vln.github.io.
Chinese Translation
在部分观察到的三维场景中，根据语言找到所指物体是具身智能体的一项核心能力。现有的方法要么将物体搜索与在线探索结合，这在相关观察已经被捕获时可能代价高昂，要么以静态的一次性方式查询预构建的开放词汇地图和场景图。我们提出了Finder，一种用于具身基础的主动闭环物体查找原语。Finder并不将基础视为从固定场景表示的被动检索，而是维护一个类型化的循环状态，链接查询条件规划、范围证据收集、候选验证以及接受/继续/中止控制。当证据不完整或模糊时，该循环可以重新引导后续的感知和比较，而不仅仅是返回检索到的最佳物体。在Habitat/HM3D的开放词汇具身物体检索和真实世界的RGB-D场景中，Finder将平均1米成功率提高了15.75个百分点，超越了强基线。相同的原语也可迁移到顺序物体基础和具身物体中心的问题回答中，改善空间和时间定位，而无需改变内部基础协议。项目页面：https://finder-vln.github.io
cs.CV / 28 / 2609.18069
GeoCueFormer: Geometry-Guided Wavelet Representation and Prediction-Cued Dual-Stage Decoder for Underwater Semantic Segmentation
GeoCueFormer：几何引导的小波表示与预测提示的双阶段解码器用于水下语义分割
Abstract
Underwater semantic segmentation is essential for marine ecosystem monitoring, yet remains challenging due to severe visual degradation. Light absorption and scattering often lead to color shifts, low contrast, and blurred boundaries, making shallow detail features unreliable. Existing underwater segmentation methods improve RGB feature aggregation or boundary prediction, but still lack an explicit mechanism to distinguish structure-related details from degradation-induced responses. To address this limitation, we propose GeoCueFormer, a lightweight framework that combines geometry-constrained frequency enhancement with prediction-cued refinement. GeoCueFormer performs stage-specific wavelet enhancement on hierarchical encoder features to complement shallow boundary details while preserving deep structural semantics. A depth-derived spatial gate constrains shallow frequency enhancement toward geometry-consistent regions, and a prediction-cued dual-stage decoder further refines ambiguous high-resolution features. GeoCueFormer obtains 82.23% and 73.04% mIoU on SUIM and DUT, respectively. Under comparable model complexity and standard benchmark settings on SUIM and DUT, it achieves SOTA performance while maintaining a favorable accuracy-complexity trade-off. These results show that distinguishing structural details from degradation-induced interference is more effective for underwater segmentation.
Chinese Translation
水下语义分割对于海洋生态系统监测至关重要，但由于严重的视觉退化，仍然面临挑战。光的吸收和散射常常导致颜色偏移、低对比度和模糊边界，使得浅层细节特征变得不可靠。现有的水下分割方法改善了RGB特征聚合或边界预测，但仍缺乏明确的机制来区分与结构相关的细节与退化引起的响应。为了解决这一局限性，我们提出了GeoCueFormer，一个轻量级框架，结合了几何约束的频率增强与预测提示的细化。GeoCueFormer在层次编码器特征上执行阶段特定的小波增强，以补充浅层边界细节，同时保留深层结构语义。一个基于深度的空间门控将浅层频率增强限制在几何一致的区域，而预测提示的双阶段解码器进一步细化模糊的高分辨率特征。GeoCueFormer在SUIM和DUT上分别获得了82.23%和73.04%的mIoU。在SUIM和DUT上具有可比的模型复杂性和标准基准设置下，它实现了SOTA性能，同时保持了良好的准确性与复杂性的权衡。这些结果表明，将结构细节与退化引起的干扰区分开来，对于水下分割更为有效。
cs.CV / 29 / 2609.18077
vidax: A Unified JAX Framework for Video Generative Models on Accelerator Meshes
vidax：一个统一的 JAX 框架，用于加速器网格上的视频生成模型
Abstract
Open-source video generative models ship almost exclusively as PyTorch/CUDA reference implementations. This leaves Cloud TPU pods without a production-ready inference path, despite offering large, cost-effective accelerator memory pools ideal for long-sequence spatiotemporal attention. We present vidax, an open-source JAX/Flax inference engine and zero-copy PyTorch-to-JAX weight translator for modern video generation architectures. vidax covers a diverse set of spatiotemporal models --- including Diffusion Transformers, omnimodal Mixture-of-Transformers, 3D VAEs, text encoders, and native samplers --- with zero PyTorch dependency in the execution path. The framework unifies 1D tensor parallelism with DeepSpeed-Ulysses sequence parallelism on a single JAX sharding mesh, integrates TPU flash-attention kernels, and implements per-layer weight offloading to support reference resolutions that exceed single-device memory. We benchmark compile times, latency, and peak memory utilization on TPU v4-8 hardware, and document real-world numerical bugs surfaced during checkpoint translation. vidax is released open-source as a baseline for JAX and TPU video generation research.
Chinese Translation
开源视频生成模型几乎仅以 PyTorch/CUDA 参考实现的形式发布。这使得 Cloud TPU 节点在推理路径上缺乏生产就绪的解决方案，尽管它们提供了适合长序列时空注意力的大型、经济高效的加速器内存池。我们提出了 vidax，一个开源的 JAX/Flax 推理引擎和零拷贝的 PyTorch 到 JAX 权重转换器，适用于现代视频生成架构。vidax 涵盖了一系列多样的时空模型——包括扩散变换器（Diffusion Transformers）、全模态混合变换器（omnimodal Mixture-of-Transformers）、3D 变分自编码器（3D VAEs）、文本编码器和原生采样器——在执行路径中完全不依赖 PyTorch。该框架将 1D 张量并行性与 DeepSpeed-Ulysses 序列并行性统一在单个 JAX 分片网格上，集成了 TPU 闪存注意力内核，并实现了每层权重卸载，以支持超出单设备内存的参考分辨率。我们在 TPU v4-8 硬件上基准测试了编译时间、延迟和峰值内存利用率，并记录了在检查点转换过程中出现的实际数值错误。vidax 作为 JAX 和 TPU 视频生成研究的基线被开源发布。
cs.CV / 30 / 2609.18088
Mask 2D-3D: Adaptive Dual-Masked Autoencoder Network for Image-to-Point Cloud Registration
Mask 2D-3D：用于图像到点云配准的自适应双重掩码自编码网络
Abstract
Detection-free methods for image-to-point cloud registration are prone to erroneous correspondences caused by domain and modality discrepancies, limited sensitivity of feature extractors, and the presence of non-overlapping regions. The Masked Autoencoder (MAE) has shown strong performance in visual representation for images and point clouds. It may be helpful to apply this approach to image-to-point cloud registration, a task that requires unified feature extraction and accurate cross-modal correspondences. Standard MAE's random masking may overlook key regions due to limited camera views, reducing registration effectiveness. To address this, we propose the Intermodal Dual-MAE Framework (ID-MAE) with a Similarity-based RL Masking Strategy (SRLM), which adaptively masks informative positions by leveraging cross-modal similarity and reinforcement learning, thus narrowing the modality gap. Our method enhances cross-modal representation learning by enforcing representation consistency during feature extraction, thereby enabling more reliable 2D-3D correspondence estimation. Experiments on RGB-D Scenes v2 and 7-Scenes benchmarks show that our method achieves state-of-the-art performance in image-to-point cloud registration.
Chinese Translation
无检测方法在图像到点云配准中容易受到领域和模态差异、特征提取器的有限敏感性以及非重叠区域的影响，从而导致错误的对应关系。掩码自编码器（Masked Autoencoder, MAE）在图像和点云的视觉表示中表现出色。将这种方法应用于图像到点云配准可能会有所帮助，这一任务需要统一的特征提取和准确的跨模态对应关系。标准MAE的随机掩码可能会由于摄像机视角的限制而忽略关键区域，从而降低配准效果。为了解决这个问题，我们提出了基于相似性的强化学习掩码策略（Similarity-based RL Masking Strategy, SRLM）的跨模态双重MAE框架（Intermodal Dual-MAE Framework, ID-MAE），该框架通过利用跨模态相似性和强化学习自适应地掩盖信息丰富的位置，从而缩小模态间的差距。我们的方法通过在特征提取过程中强制执行表示一致性来增强跨模态表示学习，从而实现更可靠的2D-3D对应关系估计。在RGB-D Scenes v2和7-Scenes基准测试中的实验表明，我们的方法在图像到点云配准中达到了最先进的性能。
cs.CV / 31 / 2609.18124
Aligned Consensus Teaching for Label-Efficient Oriented Object Detection in Weakly-Aligned Visible-Infrared Imagery
针对弱对齐可见-红外图像中的标签高效定向目标检测的对齐共识教学
Abstract
Visible-infrared object detection (VIOD) detects objects with oriented bounding boxes from paired visible and infrared images. Existing methods depend on costly dual-modality annotations. Semi-supervised learning can reduce this burden, but extending it from single-modal detection to VIOD is challenging. In the practical image-pair-level setting considered here, only a few pairs are labeled in both modalities, while the rest are completely unlabeled. This limited supervision creates three challenges: (i) too few labeled boxes for robust cross-modal alignment; (ii) pseudo-label errors caused by branch-wise misses accumulate during self-training; and (iii) tail-class annotations become critically scarce as the labeling budget decreases. We propose Aligned Consensus Teacher (ACT) for label-efficient VIOD in this setting. Its Cycle-Consistent Region Alignment (CRA) combines cycle consistency and sparse anchors with reliability-weighted regional matching. Cross-Modal Consensus Mean-Teacher (CMC-MT) forms consensus pseudo labels under pair-preserving views to recover branch-wise misses and supervise unlabeled pairs. Text-Guided Cross-Modal Instance Augmentation (TG-CMIA) uses a vision-language scene prior to compose tail-class instance pairs while preserving RGB--IR offsets. To the best of our knowledge, ACT is the first framework to study semi-supervised VIOD under this image-pair-level setting. Experiments on DroneVehicle and VEDAI show consistent gains across annotation ratios. With 10\% labeled pairs on DroneVehicle, ACT reaches 94.3\% of the mAP obtained by the same detector under full supervision. Code and models will be available on GitHub to facilitate future work.
Chinese Translation
可见-红外目标检测（VIOD）从配对的可见光和红外图像中检测带有定向边界框的目标。现有方法依赖于昂贵的双模态标注。半监督学习可以减轻这一负担，但将其从单模态检测扩展到VIOD面临挑战。在这里考虑的实际图像对级别设置中，只有少数对在两个模态中都有标注，而其余则完全未标注。这种有限的监督带来了三个挑战：（i）用于稳健的跨模态对齐的标注框过少；（ii）在自我训练过程中，由于分支遗漏而导致的伪标注错误累积；以及（iii）随着标注预算的减少，尾类标注变得极为稀缺。我们提出了对齐共识教师（Aligned Consensus Teacher，ACT），以应对这种设置下的标签高效VIOD。其循环一致区域对齐（Cycle-Consistent Region Alignment，CRA）结合了循环一致性和稀疏锚点，并采用可靠性加权的区域匹配。跨模态共识均值教师（Cross-Modal Consensus Mean-Teacher，CMC-MT）在保持配对视图的情况下形成共识伪标注，以恢复分支遗漏并监督未标注对。文本引导跨模态实例增强（Text-Guided Cross-Modal Instance Augmentation，TG-CMIA）利用视觉-语言场景先验来组合尾类实例对，同时保持RGB-IR偏移。根据我们所知，ACT是第一个在这种图像对级别设置下研究半监督VIOD的框架。在DroneVehicle和VEDAI上的实验显示在不同标注比例下的一致增益。在DroneVehicle上，使用10%的标注对，ACT达到了同一检测器在全监督下获得的94.3%的mAP。代码和模型将会在GitHub上发布，以促进未来的研究。
cs.CV / 32 / 2609.18125
PRISM: Predictive Representation of Interaction Style and Motion for Social Robot Navigation
PRISM：社交机器人导航中的交互风格与运动的预测表示
Abstract
Humans often observe others before interacting and adjust their behavior accordingly. Robot navigation in crowds, however, often represents pedestrians mainly by observed geometric states, leaving individual differences in interaction tendencies implicit. We propose PRISM (Predictive Representation of Interaction Style and Motion), a framework that infers interaction traits from passive observations of human-human interactions. PRISM encodes human trajectories into a continuous ordinal latent space with a transformer encoder trained by Rank-N-Contrast loss, and pairs each inferred trait with a temporal-stability score supplied to the navigation policy. In randomized crowd simulations, PRISM reduces collision rates over the geometry-only baseline and yields small improvements in navigation-time and path-length metrics. These results suggest the utility of passive latent-trait inference for social navigation in dynamic crowds.
Chinese Translation
人类在互动之前常常观察他人，并相应地调整自己的行为。然而，机器人在拥挤环境中的导航通常仅通过观察到的几何状态来表示行人，忽略了个体在交互倾向上的差异。我们提出了PRISM（交互风格与运动的预测表示），这是一个从人际交互的被动观察中推断交互特征的框架。PRISM将人类轨迹编码为一个连续的有序潜在空间，并通过Rank-N-Contrast损失训练的变换器编码器进行处理，并将每个推断出的特征与提供给导航策略的时间稳定性评分配对。在随机化的拥挤模拟中，PRISM相较于仅基于几何的基线减少了碰撞率，并在导航时间和路径长度指标上取得了小幅改善。这些结果表明，被动潜在特征推断在动态人群中的社交导航中的实用性。
cs.CV / 33 / 2609.18129
MCLC-NET: Multimodal Continual Learning for Leaf Counting
MCLC-NET：用于叶片计数的多模态持续学习
Abstract
Leaf counting is an important task in plant phenotyping for monitoring plant growth and estimating crop yield. Most existing methods rely on RGB images, but their performance is often affected by occlusion, lighting variations, and other real-world challenges. Additional modalities, such as depth and thermal images, can provide useful complementary information. However, multimodal leaf counting remains underexplored. Also, many existing methods assume that all training data are available simultaneously, which is impractical in real agricultural settings, where data is collected over time from multiple sources. To address these challenges, we propose MCLC-NET, a multimodal continual learning framework for leaf counting. It learns tasks sequentially using a memory-based strategy with a memory buffer to retain important samples from previous tasks. We also introduce MMLC, a real-world multimodal leaf-counting dataset designed for a domain incremental scenario (DIS) in CL. It contains RGB, depth, and thermal images collected across different crop types under varying environmental conditions, arranged in three orderings: crop-wise, time-wise, and mixed. Experimental results, averaged over three random seeds, demonstrate that MCLC-NET consistently outperforms existing methods across all three task orderings, achieving the lowest AMSE of 0.675$\pm$0.027, 0.542$\pm$0.069, and 0.745$\pm$0.057, respectively.
Chinese Translation
叶片计数是植物表型学中的一项重要任务，旨在监测植物生长和估算作物产量。现有的大多数方法依赖于RGB图像，但其性能常常受到遮挡、光照变化以及其他现实世界挑战的影响。额外的模态，如深度图像和热成像，可以提供有用的补充信息。然而，多模态叶片计数仍然未得到充分探索。此外，许多现有方法假设所有训练数据可以同时获得，这在实际农业环境中并不切实际，因为数据是随着时间从多个来源收集的。为了解决这些挑战，我们提出了MCLC-NET，一个用于叶片计数的多模态持续学习框架。它采用基于记忆的策略顺序学习任务，并使用记忆缓冲区保留来自先前任务的重要样本。我们还引入了MMLC，一个为领域增量场景（DIS）设计的真实世界多模态叶片计数数据集。该数据集包含在不同作物类型下、在不同环境条件下收集的RGB、深度和热成像，按照作物、时间和混合三种顺序排列。实验结果在三个随机种子上取平均，表明MCLC-NET在所有三种任务顺序中始终优于现有方法，分别达到了最低的平均均方误差（AMSE）0.675±0.027、0.542±0.069和0.745±0.057。
cs.CV / 34 / 2609.18133
Stealthy in Semantics, Antagonistic in Space: Attacking Visible-Infrared Object Detectors via Object-Level Misalignment
语义隐蔽、空间对抗：通过物体级别错位攻击可见-红外物体检测器
Abstract
Visible-infrared object detectors are used for robust perception under challenging illumination and weather conditions. Current physical attacks apply conspicuous patches to spatially aligned target regions, which are noticeable to human observers. Meanwhile, most of these methods only perturb the appearance within the aligned region, without explicitly targeting the correspondence between modalities or the fusion process. In this paper, we propose CamoShift, an adversarial framework for visible-infrared object detection. By combining visual camouflage with object-level infrared shifting, CamoShift breaks cross-modal spatial alignment and disrupts fusion. Specifically, the Semantic Camouflage Module (SCM) generates a stealthy camouflaged patch that can be attached to the host object and maintains its effectiveness in the infrared branch through an RGB-IR adapter. The Object-level Spatial Decoupling Module (OSDM) shifts the infrared target evidence in a scale-aware manner, so as to break object-level correspondence and disrupt cross-modal fusion. Then, the Harmonic Adversarial loss (HarAdv loss) further balances attack strength and visual stealth during optimization. To the best of our knowledge, we are the first to target both visual stealthiness and attack success in visible-infrared object detection. Extensive experimental results show that CamoShift achieves a superior balance between attack effectiveness and visual stealth. Code and models will be available on GitHub.
Chinese Translation
可见-红外物体检测器用于在具有挑战性的光照和天气条件下实现稳健感知。当前的物理攻击方法在空间上对齐的目标区域施加显眼的补丁，这些补丁对人类观察者来说是显而易见的。同时，这些方法大多仅扰动对齐区域内的外观，而没有明确针对模态之间的对应关系或融合过程。在本文中，我们提出了CamoShift，一个用于可见-红外物体检测的对抗框架。通过将视觉伪装与物体级别的红外位移相结合，CamoShift打破了跨模态的空间对齐并干扰了融合过程。具体而言，语义伪装模块（Semantic Camouflage Module, SCM）生成一个隐蔽的伪装补丁，可以附加到宿主物体上，并通过RGB-IR适配器在红外分支中保持其有效性。物体级空间解耦模块（Object-level Spatial Decoupling Module, OSDM）以尺度感知的方式移动红外目标证据，从而打破物体级别的对应关系并干扰跨模态融合。然后，和谐对抗损失（Harmonic Adversarial loss, HarAdv loss）在优化过程中进一步平衡攻击强度和视觉隐蔽性。据我们所知，我们是首个同时针对可见-红外物体检测中的视觉隐蔽性和攻击成功率的研究。大量实验结果表明，CamoShift在攻击有效性和视觉隐蔽性之间实现了优越的平衡。代码和模型将在GitHub上发布。
cs.CV / 35 / 2609.18139
Multi-View Mixture-of-Experts with Vision-Language Reranking for Cross-View Object Geo-Localization
基于多视角专家混合模型和视觉-语言重排序的跨视角物体地理定位
Abstract
Cross-view object geo-localization (CVOGL) locates a target in satellite imagery using drone or street-view queries. Existing methods train separate detectors for each viewpoint, leading to parameter redundancy and impeding cross-view knowledge sharing. Moreover, top-ranked satellite candidates are often visually similar, so visual appearance and categorical labels alone are insufficient to resolve such ambiguity. To address these, we propose MVLGeo, an efficient framework designed to unify multiple viewpoints and reduce model redundancy. First, we introduce environmental contextual text from the query view as cues to distinguish visually similar candidates via Vision-Language Reranking (VL-Rerank). Second, we design a multi-view Mixture-of-Experts architecture (MV-MoE) with a shared encoder and view-specific experts to reduce redundancy and promote knowledge sharing, while cross-view contrastive learning aligns their representations for consistency. Third, we introduce an adaptive elliptical prior (ESAM-Prior) as auxiliary positional encoding for anisotropic geometric perception. Extensive experiments on the CVOGL benchmarks confirm that MVLGeo, as a unified model for multiple query viewpoints, achieves state-of-the-art performance, demonstrating robustness to input degradation and generalization across viewpoints. Code and models will be available on GitHub to facilitate future work.
Chinese Translation
跨视角物体地理定位（CVOGL）通过无人机或街景查询在卫星影像中定位目标。现有方法为每个视角训练独立的检测器，导致参数冗余并阻碍跨视角知识共享。此外，排名靠前的卫星候选对象通常在视觉上相似，因此仅依靠视觉外观和类别标签不足以解决这种模糊性。为了解决这些问题，我们提出了MVLGeo，一个旨在统一多个视角并减少模型冗余的高效框架。首先，我们从查询视角引入环境上下文文本作为线索，通过视觉-语言重排序（VL-Rerank）来区分视觉上相似的候选对象。其次，我们设计了一种多视角专家混合架构（MV-MoE），具有共享编码器和视角特定的专家，以减少冗余并促进知识共享，同时跨视角对比学习对齐它们的表示以保持一致性。第三，我们引入了一种自适应椭圆先验（ESAM-Prior）作为各向异性几何感知的辅助位置编码。在CVOGL基准上的大量实验确认，作为多查询视角的统一模型，MVLGeo达到了最先进的性能，展示了对输入退化的鲁棒性和跨视角的泛化能力。代码和模型将会在GitHub上发布，以促进未来的研究。
cs.CV / 36 / 2609.18206
CapMap-MS-TTA: 3rd Place Solution for the MUMU Track of the 8th LSVOS Challenge at ECCV 2026
CapMap-MS-TTA：2026年ECCV第八届大规模视频目标分割（LSVOS）挑战MUMU赛道的第三名解决方案
Abstract
The MUMU track of the 8th Large-scale Video Object Segmentation (LSVOS) Challenge requires a single unified multimodal model to jointly solve image tagging (Task A), open-vocabulary object detection (Task B), and English captioning (Task C) under strict resource constraints (<=0.5B parameters and <=8 GB peak GPU memory). We present CapMap-MS-TTA, a training-free submission built on Microsoft Florence-2-base (~231M parameters), combining caption keyword mapping with multi-scale flip test-time augmentation. Task C uses the native pathway with length/token sanitization. Task A maps the same detailed caption into the official quality/scene/event vocabularies via an expanded keyword lexicon with whole-word matching and a lightweight expand-hints stage. Task B runs Florence-2 open detection () with multi-scale and horizontal-flip test-time augmentation (TTA), followed by label-aware non-maximum suppression (NMS). Without fine-tuning, the system improves our reproduced Florence-2 baseline from 15.16 to a best public score of 16.4815, and ranks 3rd on the final MUMU leaderboard.
Chinese Translation
第八届大规模视频目标分割（LSVOS）挑战的MUMU赛道要求一个统一的多模态模型在严格的资源限制下（<=0.5B参数和<=8 GB峰值GPU内存）共同解决图像标记（任务A）、开放词汇对象检测（任务B）和英文字幕生成（任务C）。我们提出了CapMap-MS-TTA，这是一个基于Microsoft Florence-2-base（约231M参数）的无训练提交，结合了字幕关键词映射和多尺度翻转测试时间增强。任务C使用原生的路径，并进行长度/标记的清理。任务A通过扩展的关键词词汇表将相同的详细字幕映射到官方的质量/场景/事件词汇中，采用全词匹配和轻量级的扩展提示阶段。任务B运行Florence-2开放检测（），结合多尺度和水平翻转的测试时间增强（TTA），随后进行标签感知的非最大抑制（NMS）。在不进行微调的情况下，该系统将我们复现的Florence-2基线从15.16提升至最佳公共得分16.4815，并在最终的MUMU排行榜上排名第三。
cs.CV / 37 / 2609.18210
Understanding Dynamic Scenes at Gigapixel Scale: Wide-Area Spatio-Temporal Perception from UAVs
在千兆像素尺度下理解动态场景：来自无人机的广域时空感知
Abstract
UAV-borne imaging has advanced from megapixel to gigapixel sensors, shifting aerial perception from recognizing individual targets to understanding entire dynamic scenes. We characterize this demand as Wide-area Spatio-temporal Scene Understanding (WSTU), which requires wide-area coverage, per-target resolution, and temporal continuity at once, a combination existing datasets lack. To fill this gap, we introduce an ultra-High-resolution (12768x9564) Airborne Remote-sensing Dataset (HARD) annotated at three levels for object detection, multi-object tracking, and scene-level visual question answering. Ultra-high-resolution imagery raises per-frame processing time to seconds. At that scale latency can no longer be ignored in evaluation. Thus, we propose a latency-aware metric for multi-object tracking called streaming-HOTA (s-HOTA). Extensive baseline experiments show how ultra-high-resolution processing reshapes each task. For detection, the end-to-end pipeline affects accuracy and speed as much as the detector itself does. For tracking, high latency charges the association axis far more unevenly than the detection axis, and association is where pipelines diverge. As a result, the pipeline that performs best offline can lose its lead under s-HOTA. For VQA, vision-language models remain weak at cross-frame identity binding and cannot transfer their single-frame gains to it. Together these findings show that the baselines we evaluate fall short of WSTU. HARD provides the data and the systematic baselines to advance it.
Chinese Translation
无人机搭载成像技术已从百万像素传感器发展到千兆像素传感器，使得空中感知从识别单个目标转向理解整个动态场景。我们将这种需求描述为广域时空场景理解（Wide-area Spatio-temporal Scene Understanding, WSTU），它同时需要广域覆盖、每个目标的分辨率和时间连续性，而现有数据集并不具备这种组合。为填补这一空白，我们引入了一个超高分辨率（12768x9564）的空中遥感数据集（HARD），该数据集在三个层次上进行了标注，以支持目标检测、多目标跟踪和场景级视觉问答。超高分辨率图像使得每帧的处理时间延长至数秒。在这种尺度下，延迟在评估中不再可以忽视。因此，我们提出了一种针对多目标跟踪的延迟感知指标，称为流式HOTA（streaming-HOTA, s-HOTA）。大量基线实验表明，超高分辨率处理重塑了每个任务的表现。在检测任务中，端到端管道对准确性和速度的影响与检测器本身同样重要。在跟踪任务中，高延迟使得关联轴的负担远比检测轴更加不均衡，而关联正是管道分歧的地方。因此，在s-HOTA下，表现最佳的离线管道可能会失去其领先地位。在视觉问答（VQA）任务中，视觉-语言模型在跨帧身份绑定方面仍然较弱，无法将其单帧的优势转移到此任务上。综合这些发现表明，我们评估的基线未能满足WSTU的要求。HARD提供了数据和系统化的基线，以推动这一领域的发展。
cs.CV / 38 / 2609.18227
WISE: A Lightweight, Weakly-Supervised Model for Onboard Fire Smoke Detection and Localization
WISE：一种轻量级、弱监督的机载火灾烟雾检测与定位模型
Abstract
Wildfire smoke detection from satellite imagery is critical for early warning and rapid response. For onboard satellite deployment, detection systems must operate under strict memory and latency constraints while providing spatially informative outputs for downstream decision-making. Existing tile-level classification methods are computationally efficient but lack spatial localization, whereas pixel-level segmentation approaches provide detailed masks yet are typically too computationally demanding for real-time onboard execution. To address this gap, we propose WISE (Weakly-supervised Inference-efficient Smoke Extraction), a deployment-oriented framework for onboard fire smoke detection and localization. WISE leverages only tile-level annotations through a teacher-student distillation strategy, where an offline teacher provides soft spatial supervision to a lightweight WISE-Student optimized for efficient onboard inference. The student jointly predicts tile-level smoke presence and smoke probability maps within a single forward pass, enabling spatially informative detection under strict computational constraints. WISE was evaluated through in-orbit execution aboard the ISS-mounted IMAGIN-e payload. Three model variants achieve average inference times of 0.10 s, 0.14 s, and 0.26 s per tile, indicating near-real-time per-tile inference within onboard resource limits. Ground-based experiments on Landsat 5 and Landsat 8 imagery further indicate effective detection and spatially informative localization. The best-performing variant achieves a mean tile-level F1 score of 0.964 and a mean pixel-level F1 score of 0.750 across 10 runs, while containing only 0.12M parameters and requiring approximately 3 GFLOPs. Together, these results indicate that WISE is a practical candidate for low-latency wildfire smoke monitoring from space under onboard resource constraints.
Chinese Translation
从卫星影像中检测野火烟雾对于早期预警和快速响应至关重要。对于机载卫星部署，检测系统必须在严格的内存和延迟限制下运行，同时为下游决策提供空间信息丰富的输出。现有的瓦片级分类方法计算效率高，但缺乏空间定位能力，而像素级分割方法提供详细的掩膜，但通常对实时机载执行的计算需求过高。为了解决这一问题，我们提出了WISE（Weakly-supervised Inference-efficient Smoke Extraction），这是一个面向部署的机载火灾烟雾检测与定位框架。WISE仅利用瓦片级注释，通过教师-学生蒸馏策略，其中离线教师为优化高效机载推理的轻量级WISE-Student提供软空间监督。该学生在单次前向传播中共同预测瓦片级烟雾存在和烟雾概率图，从而在严格的计算限制下实现空间信息丰富的检测。WISE通过在国际空间站（ISS）上搭载的IMAGIN-e有效载荷进行在轨执行评估。三种模型变体的平均推理时间分别为0.10秒、0.14秒和0.26秒，表明在机载资源限制内实现了近实时的每瓦片推理。对Landsat 5和Landsat 8影像的地面实验进一步表明有效的检测和空间信息丰富的定位。表现最佳的变体在10次运行中实现了平均瓦片级F1分数为0.964和平均像素级F1分数为0.750，同时仅包含0.12M参数，约需3 GFLOPs。综合这些结果表明，WISE是一个在机载资源限制下进行低延迟野火烟雾监测的实用候选方案。
cs.CV / 39 / 2609.18239
Unified Response Geometry for Structured Pruning
结构剪枝的统一响应几何
Abstract
Structured pruning is commonly formulated as ranking individual channels, although channel responses can be complementary or cancel through downstream mixing. Motivated by these response interactions, we formulate pruning as the selection of a subset with large joint response capacity, followed by a separate functional realization step. Our unified response geometry maps each candidate set to \(M(D,R)=D^{1/2}RD^{1/2}\) and uses its determinant together with Schur-greedy residuals to select non-redundant coordinates. The same construction yields two information-conditioned instances: an unlabeled instance based on activation covariance, and a task-conditioned instance that combines activation and gradient variance for response scale with gradient correlation for complementarity. To convert the selected subset into an executable network, we fold predictable removed responses into successor weights through ridge compensation and recalibrate batch-normalization statistics, without fine-tuning the network. On ImageNet ResNet-50, the unlabeled instance reaches \(65.4\%\) and \(53.9\%\) Top-1 accuracy at 30\% and 40\% deletion, versus \(59.8\%\) and \(43.1\%\) for strength-only selection; the task-conditioned instance reaches \(67.7\%\) and \(56.3\%\) under the same protocol. A six-family screen shows architecture-dependent behavior, with positive relative contrasts in several convolutional and expansion-layer settings and clear boundary cases in windowed attention. These results support response geometry as a conditional principle for structured pruning, with its benefit determined jointly by the observed response and the architecture in which that response is realized.
Chinese Translation
结构剪枝通常被表述为对单个通道的排序，尽管通道响应可能通过下游混合互为补充或相互抵消。基于这些响应交互的动机，我们将剪枝形式化为选择一个具有较大联合响应能力的子集，随后进行单独的功能实现步骤。我们的统一响应几何将每个候选集映射到 $M(D,R)=D^{1/2}RD^{1/2}$，并利用其行列式以及 Schur-greedy 残差来选择非冗余坐标。同样的构造产生了两个信息条件实例：一个基于激活协方差的无标签实例，以及一个结合激活和梯度方差以响应规模和梯度相关性以实现互补性的任务条件实例。为了将所选子集转换为可执行网络，我们通过岭回归补偿将可预测的移除响应折叠到后继权重中，并重新校准批量归一化统计数据，而无需对网络进行微调。在 ImageNet ResNet-50 上，无标签实例在 30\% 和 40\% 删除时分别达到 $65.4\\%$ 和 $53.9\\%$ 的 Top-1 准确率，而仅基于强度选择的实例则为 $59.8\\%$ 和 $43.1\\%$；任务条件实例在相同协议下达到 $67.7\\%$ 和 $56.3\\%$。六个家族的筛选显示出架构依赖行为，在多个卷积和扩展层设置中具有正的相对对比，并在窗口注意力中表现出明显的边界案例。这些结果支持响应几何作为结构剪枝的条件原则，其益处由观察到的响应和实现该响应的架构共同决定。
cs.CV / 40 / 2609.18256
Evolving Error States: Failure-Aware Progressive Repair for Ultrasound Lesion Segmentation
演变错误状态：面向失败的渐进修复用于超声病变分割
Abstract
Reliability under sparse and heterogeneous failures remains a fundamental challenge for medical image segmentation. High average accuracy can conceal a small set of structurally distinct and clinically consequential errors. Existing post-hoc correction methods alleviate this problem, but typically estimate false-positive and false-negative corrections from the same fixed prediction. This ignores the dynamic evolution of error states and limits the correction of complex cases. Inspired by iterative error feedback in structured prediction, we propose Failure-Aware Progressive Repair (FAPR). FAPR represents the current segmentation mask as a dynamic failure state and models each repair operation as a state-transition operator. Each accepted correction forms a new prediction state for subsequent error diagnosis and repair, enabling later operations to adapt to preceding changes. Conditional routing selectively activates necessary state transitions, while failure replay exposes the model to rare error states. By keeping the base segmentor frozen, FAPR preserves its established segmentation capability while improving difficult cases. Across three public ultrasound lesion segmentation benchmarks, FAPR improves mean DSC by 1.52%. On the very-hard subsets of BUSI and TN3K, the average gain reaches 13.77%.
Chinese Translation
在稀疏和异构故障下的可靠性仍然是医学图像分割的一个基本挑战。高平均准确率可能掩盖一小部分结构上明显且临床上重要的错误。现有的事后修正方法缓解了这一问题，但通常从同一固定预测中估计假阳性和假阴性的修正。这忽视了错误状态的动态演变，并限制了复杂案例的修正。受到结构化预测中迭代错误反馈的启发，我们提出了面向失败的渐进修复（Failure-Aware Progressive Repair, FAPR）。FAPR将当前的分割掩膜表示为动态错误状态，并将每个修复操作建模为状态转移算子。每个被接受的修正形成一个新的预测状态，用于后续的错误诊断和修复，使得后续操作能够适应之前的变化。条件路由选择性地激活必要的状态转移，而故障重放则使模型接触到稀有错误状态。通过保持基础分割器不变，FAPR保留了其已建立的分割能力，同时改善了困难案例。在三个公共超声病变分割基准测试中，FAPR将平均 DSC 提高了 1.52%。在 BUSI 和 TN3K 的非常困难子集上，平均增益达到 13.77%。
cs.CV / 41 / 2609.18260
MS-RFD: Multi-Signal Release Frame Detection in Hammer Throw from Reconstructed 3D Trajectories
MS-RFD：基于重建3D轨迹的铅球投掷多信号释放帧检测
Abstract
Recent advances in artificial intelligence and computer vision are reshaping sports performance analysis by enabling automated detection, tracking, and performance analysis. In hammer throw, performance is strongly determined by the kinematic conditions at release, particularly release speed, release angle, and release height. However, identifying the release instant from video typically requires manual frame-by-frame inspection, which is subjective and cumbersome in real-world training scenarios. In this paper, we present a fully automatic multi-signal release frame detection (MS-RFD) method for hammer throw using reconstructed 3D hammer trajectories. The proposed method integrates four complementary kinematic signals: speed dynamics, angular velocity transition, radial distance relative to the rotation center, and post-release trajectory linearity. These signals are fused to score and verify candidate release frames. MS-RFD is evaluated through the throwing-distance estimation error obtained from the release parameters estimated at the detected frame. An ablation study analyzes the contribution of each signal and compares alternative candidate selection strategies. The results show that speed dynamics and radial expansion provide the strongest signals for release frame detection, while angular velocity and post-release linearity provide smaller refinements.
Chinese Translation
近年来，人工智能和计算机视觉的进步正在重塑体育表现分析，通过实现自动化检测、跟踪和表现分析。在铅球投掷中，表现主要由释放时的运动学条件决定，特别是释放速度、释放角度和释放高度。然而，从视频中识别释放瞬间通常需要逐帧手动检查，这在实际训练场景中主观且繁琐。本文提出了一种基于重建3D铅球轨迹的全自动多信号释放帧检测方法（MS-RFD）。该方法整合了四个互补的运动学信号：速度动态、角速度变化、相对于旋转中心的径向距离以及释放后轨迹的线性度。这些信号被融合以评分和验证候选释放帧。通过从检测到的帧估计的释放参数获得的投掷距离估计误差来评估MS-RFD。消融研究分析了每个信号的贡献，并比较了替代候选选择策略。结果表明，速度动态和径向扩展为释放帧检测提供了最强的信号，而角速度和释放后线性度则提供了较小的细化。
cs.CV / 42 / 2609.18279
Decoder-Agnostic Token Merging for Vision Transformers: A Systematic Study of G2TM
解码器无关的令牌合并在视觉变换器中的应用：G2TM的系统研究
Abstract
Vision Transformers (ViTs) have achieved state-of-the-art performance across a range of computer vision tasks, mainly thanks to the self-attention mechanism. However, its complexity, increasing quadratically with the number of tokens, remains the major obstacle to ViT efficiency and deployment at scale. Token merging reduces this cost by aggregating redundant tokens. Yet existing methods are typically evaluated within a single architecture, leaving open whether their effectiveness stems from the merging mechanism itself or from the specific decoder they are paired with. We extend Graph-Guided Token Merging (G2TM), a single module inserted early in a ViT-based network, beyond its original Segmenter setting. We evaluate G2TM across three semantic segmentation frameworks (Segmenter, SETR, EoMT) and three decoder families (Linear, Transformer-, convolution-based), as well as standard ViT image classification. Our results show that G2TM's behavior and accuracy-efficiency trade-off are consistent across every tested architecture for a given backbone size, indicating that its effectiveness is a property of the encoder rather than the decoder. G2TM also generalizes well to image classification, achieving an even smaller degradation in accuracy compared to semantic segmentation. We further find that G2TM's optimal hyperparameters, resulting in a consistent drop in GFLOPs of 22-47% and an increase in throughput by up to 74% for segmentation models on ADE20K dataset, depend primarily on the backbone's pre-training recipe and on the target dataset, rather than on the decoder choice.
Chinese Translation
视觉变换器（ViTs）在一系列计算机视觉任务中取得了最先进的性能，这主要得益于自注意力机制。然而，其复杂性随着令牌数量的增加而呈二次增长，仍然是ViT效率和大规模部署的主要障碍。令牌合并通过聚合冗余令牌来降低这一成本。然而，现有的方法通常在单一架构内进行评估，尚不清楚其有效性是源于合并机制本身，还是与特定解码器的配对有关。我们扩展了图引导令牌合并（G2TM），这一单一模块早期插入ViT基础网络，超越了其原始的分割器设置。我们在三个语义分割框架（Segmenter、SETR、EoMT）和三种解码器家族（线性、变换器、卷积基础）以及标准ViT图像分类中评估G2TM。我们的结果表明，对于给定的主干网络大小，G2TM的行为和准确性-效率权衡在每个测试的架构中是一致的，表明其有效性是编码器的属性，而非解码器。G2TM在图像分类中也表现良好，与语义分割相比，准确性的下降更小。我们进一步发现，G2TM的最佳超参数导致在ADE20K数据集上分割模型的GFLOPs一致下降22-47%，吞吐量提高最多74%，主要取决于主干网络的预训练方案和目标数据集，而不是解码器的选择。
cs.CV / 43 / 2609.18302
Visual Autoregressive Priors for RAW-to-sRGB Image Signal Processing
用于RAW到sRGB图像信号处理的视觉自回归先验
Abstract
RAW-to-sRGB image signal processing (ISP) must recover perceptually faithful colors and fine details from sensor measurements, often under imperfect spatial alignment and missing camera metadata. This paper presents, to the best of our knowledge, the first application of visual autoregressive (VAR) next-scale prediction over a discrete image codebook to the RAW-to-sRGB ISP task. We adapt a frozen 1.10\,B-parameter VAR backbone for RAW-conditioned ISP with only 32.93\,M trainable parameters (2.99\%), and propose a frequency-decomposed color loss that separately supervises low-frequency tone via wavelet LL cosine similarity and chromatic edges via detail-band $\ell_1$. On the Zurich RAW-to-sRGB benchmark, the method improves PSNR-Y from 21.31 to 21.89\,dB and reduces LPIPS from 0.276 to 0.218 on the full 1,204-image test set. Diagnostic experiments show that the VAR prior preserves structure well, but continuous color transfer remains the dominant bottleneck: oracle affine correction recovers 3.8\,dB, while learned color heads yield marginal gains.
Chinese Translation
RAW到sRGB图像信号处理（ISP）必须从传感器测量中恢复感知上真实的颜色和细节，通常在空间对齐不完美和缺失相机元数据的情况下进行。本文提出了我们所知的首个将视觉自回归（VAR）下一尺度预测应用于离散图像代码本的RAW到sRGB ISP任务。我们为RAW条件下的ISP适配了一个冻结的1.10亿参数VAR主干网络，只有3293万可训练参数（2.99%），并提出了一种频率分解的颜色损失，分别通过小波LL余弦相似度监督低频色调和通过细节带的$ ext{l}_1$监督色彩边缘。在苏黎世RAW到sRGB基准测试中，该方法将PSNR-Y从21.31提升至21.89 dB，并将LPIPS从0.276降低至0.218，适用于完整的1204幅图像测试集。诊断实验表明，VAR先验能够很好地保持结构，但连续颜色传递仍然是主要瓶颈：oracle仿射校正恢复了3.8 dB，而学习的颜色头仅带来了微小的增益。
cs.CV / 44 / 2609.18323
Can MiniMax-H3 Reason About the Physical World? An Evaluation of Omni-Modal Generative Model
MiniMax-H3能否推理物理世界？全模态生成模型的评估
Abstract
Recent Omni-Modal Generative Models (Omni-Models) have advanced content generation toward unified modeling of text, images, video, and audio. MiniMax-H3 exemplifies this transition by combining multimodal context understanding with joint audio-visual generation in a shared latent framework. Its unified architecture raises a fundamental question: Can multimodal alignment improve the model's world reasoning, and what new evaluation paradigms do omni-modal inputs enable? To investigate this question, this work introduces a comprehensive evaluation framework organized around four complementary dimensions of physical world reasoning. Unlike existing evaluation frameworks for video generation and world models, which are often constrained by limited input modalities and evaluation settings where prompts closely match the target video content, our evaluation is specifically designed to exploit the multimodal inputs of Omni-Model. We construct a diverse set of novel tasks that require models to integrate complementary information across modalities. Specifically, we consider four scenarios, including implicit prompts paired with multiple frames, audio-image, prefix-videos, and audio-video inputs. Every single modality provides only partial evidence about the underlying event, requiring the model to jointly reason over the complementary semantic cues to infer latent event states and future dynamics. Across 517 evaluation instances, MiniMax-H3 achieves an overall success rate of 41.97%. Video-based Decision Reasoning yields the highest success rate at 56.00%, while Audio-based Disambiguation Reasoning is the weakest, reaching only 27.40%. These results indicate that effective multimodal integration remains key to fully exploiting the benefits of diverse input modalities. The project is available at https://github.com/gulucaptain/MiniMax-H3-Reason.
Chinese Translation
近期的全模态生成模型（Omni-Models）推动了文本、图像、视频和音频的统一建模内容生成。MiniMax-H3通过在共享潜在框架中结合多模态上下文理解与联合音视频生成，体现了这一转变。其统一架构引发了一个基本问题：多模态对齐是否能改善模型的世界推理能力，以及全模态输入能启用哪些新的评估范式？为探讨这一问题，本研究引入了一个全面的评估框架，围绕物理世界推理的四个互补维度进行组织。与现有的视频生成和世界模型评估框架不同，后者通常受到有限输入模态和评估设置的限制，这些设置中提示与目标视频内容紧密匹配，我们的评估专门设计用于利用全模态生成模型的多模态输入。我们构建了一组多样化的新任务，要求模型跨模态整合互补信息。具体而言，我们考虑了四种场景，包括与多个帧配对的隐式提示、音频-图像、前缀视频和音频-视频输入。每种单一模态仅提供关于潜在事件的部分证据，要求模型共同推理互补的语义线索，以推断潜在事件状态和未来动态。在517个评估实例中，MiniMax-H3的整体成功率为41.97%。基于视频的决策推理成功率最高，达56.00%，而基于音频的消歧推理则最弱，仅达到27.40%。这些结果表明，有效的多模态整合仍然是充分利用多样化输入模态优势的关键。该项目可在https://github.com/gulucaptain/MiniMax-H3-Reason获取。
cs.CV / 45 / 2609.18329
PDA++: Field-Aligned Planning and Scene-Adaptive Insertion in Remote Sensing
PDA++：遥感中的场对齐规划与场景自适应插入
Abstract
Remote sensing recognition is often constrained by scarce observations of rare targets and costly annotations, making realistic synthetic augmentation particularly valuable for few-shot and long-tailed scenarios. Object insertion provides an efficient way to increase target diversity while preserving authentic background scenes, but realistic insertion in overhead imagery requires the generated target to adapt coherently to its surrounding environment. To this end, we propose PDA++, a unified environment-aware object insertion framework organized as Plan, Decouple, and Assimilate. Planning determines scene-compatible poses through an affordance field that combines geometric clearance with structure- and scale-aware cues. Decoupling introduces a pose-conditioned background that provides precise spatial guidance together with target-scene context, allowing the reference object to preserve its identity while adapting to the target observation. This construction also naturally provides pixel-level masks for segmentation augmentation. Assimilation further improves local coherence by aligning multi-scale texture distributions through optimal transport. On the optical benchmark, PDA++ achieves a whole-image FID of 6.28 and improves average few-shot recognition mAP50 by 17.69 points, corresponding to a 28.8% relative gain over the real-data baseline. On SAR imagery, it improves ship detection by 4.10 mAP50 points and remains effective under cross-dataset transfer and amorphous-target insertion. Code is available at https://github.com/lisheyu972/PDA_PLUS.
Chinese Translation
遥感识别常常受到稀缺的稀有目标观察和昂贵的标注的限制，使得在少样本和长尾场景中，现实的合成增强尤为重要。目标插入提供了一种有效的方式来增加目标多样性，同时保留真实的背景场景，但在高空影像中实现现实的插入要求生成的目标能够与其周围环境协调适应。为此，我们提出了PDA++，一个统一的环境感知目标插入框架，组织为规划（Plan）、解耦（Decouple）和同化（Assimilate）。规划通过结合几何间隙与结构和尺度感知线索的可用性场确定与场景兼容的姿态。解耦引入了一个姿态条件背景，提供精确的空间指导以及目标-场景上下文，使参考对象在适应目标观察的同时保持其身份。这一构造还自然地为分割增强提供了像素级掩码。同化通过最优传输进一步改善局部一致性，调整多尺度纹理分布。在光学基准测试中，PDA++实现了6.28的全图FID，并将平均少样本识别的mAP50提高了17.69个百分点，相当于相对于真实数据基线的28.8%的相对增益。在SAR影像中，它将船舶检测提高了4.10 mAP50点，并在跨数据集迁移和无形目标插入中仍然有效。代码可在 https://github.com/lisheyu972/PDA_PLUS 获取。
cs.CV / 46 / 2609.18336
Pose2Muscle: Structured Spatio-Temporal Decoding for Discrete Muscle Activity Estimation from Human Pose
Pose2Muscle：基于结构化时空解码的人体姿态离散肌肉活动估计
Abstract
Muscle activity is fundamental to human movement, and understanding its patterns is critical for injury prevention and rehabilitation. Conventional muscle activity monitoring relies on specialized sensors such as surface electromyography, which limits its practicality for long-term real-world use. Existing studies suggest that muscle-related information can be inferred from human pose. However, the substantial gap between externally observable pose and internal muscle activation, limits the accuracy and generalization of current approaches. In this study, we propose Pose2Muscle, a pose-driven framework for discrete muscle activity estimation without requiring sEMG signals at inference time. Instead of directly regressing continuous sEMG signals, Pose2Muscle reformulates muscle estimation as a structured prediction problem over discrete muscle activity states, yielding a more stable and interpretable target space. The framework combines multi-scale spatio-temporal attention to capture motion patterns at complementary spatial and temporal scales with a directed acyclic graph-based decoder that maintains multiple candidate muscle-state hypotheses and performs structured trajectory inference over time. To support this task, we construct PoseEMG-43, a synchronized pose-sEMG dataset containing 2,992 movement instances from 43 daily-life actions performed by 14 participants. Experiments show that Pose2Muscle consistently outperforms representative retrieval- and pose-based baselines. It achieves an Adjacent-level Accuracy of 86.36% and a Pearson correlation coefficient of 0.8821 under the Random Split, and 63.97% and 0.6795, respectively, under the Subject-Level Split. These results demonstrate the feasibility of inferring structured muscle-state patterns from human pose and suggest the potential of Pose2Muscle for muscle-aware movement analysis when direct physiological sensing is impractical
Chinese Translation
肌肉活动是人类运动的基础，理解其模式对于预防伤害和康复至关重要。传统的肌肉活动监测依赖于表面肌电图等专业传感器，这限制了其在长期实际应用中的可行性。现有研究表明，可以从人体姿态中推断出与肌肉相关的信息。然而，外部可观察的姿态与内部肌肉激活之间的巨大差距限制了当前方法的准确性和泛化能力。在本研究中，我们提出了Pose2Muscle，一个基于姿态的框架，用于在推理时不需要sEMG信号的离散肌肉活动估计。Pose2Muscle将肌肉估计重新表述为一个针对离散肌肉活动状态的结构化预测问题，从而产生一个更稳定和可解释的目标空间。该框架结合了多尺度时空注意力机制，以捕捉在互补空间和时间尺度上的运动模式，并使用基于有向无环图的解码器，维护多个候选肌肉状态假设，并在时间上执行结构化轨迹推理。为了支持这一任务，我们构建了PoseEMG-43，这是一个同步的姿态-sEMG数据集，包含来自14名参与者的43种日常生活动作的2,992个运动实例。实验结果表明，Pose2Muscle在代表性的检索和姿态基线方法中始终表现优越。在随机分割下，它达到了86.36%的相邻级别准确率和0.8821的皮尔逊相关系数，而在受试者级别分割下，分别为63.97%和0.6795。这些结果表明，从人体姿态推断结构化肌肉状态模式的可行性，并暗示Pose2Muscle在直接生理传感不切实际时进行肌肉感知运动分析的潜力。
cs.CV / 47 / 2609.18345
Visual Input and Its Framing Affect Attribute-based Descriptions Produced by Large Vision-Language Models
视觉输入及其框架对大型视觉语言模型生成的基于属性的描述的影响
Abstract
Large vision-language models (LVLMs) are commonly used with only a single text prompt as the input, or plus an image. In this paper, we demonstrate that when the image exists, even if the text prompt is not about the specific instance (but only the concept it belongs to) in that image, the response would still be affected. For example, when the text prompt only asks for the attribute descriptions of a dog breed, an image depicting a specific dog from that breed would shift the response. Further, how the specific instance is framed in that image would determine towards which the response shifts. Detailed analyses also reveal that in the response, physical terms increase from 18% for text-only to 45% (40%) for subject-focused (subject-in-situation) framings. Overall, the unexpected effects of visual cues on LVLMs highlight the need to understand the presence of an image and its framing when evaluating the robustness of LVLMs.
Chinese Translation
大型视觉语言模型（LVLMs）通常仅使用单一文本提示作为输入，或加上图像。在本文中，我们展示了当图像存在时，即使文本提示并未针对该图像中的特定实例（而仅是其所属的概念），响应仍会受到影响。例如，当文本提示仅要求描述某个犬种的属性时，描绘该犬种中特定狗的图像会改变响应。此外，该图像中具体实例的框架将决定响应的偏向。详细分析还显示，在响应中，物理术语的使用从仅文本的18%增加到以主体为中心的框架下的45%（40%）。总体而言，视觉线索对LVLMs的意外影响强调了在评估LVLMs的鲁棒性时理解图像的存在及其框架的重要性。
cs.CV / 48 / 2609.18363
Online Multi-Camera 3D Tracking via ID Prediction over Recurrent Sparse Queries
通过对递归稀疏查询进行ID预测的在线多摄像头3D跟踪
Abstract
Online multi camera 3D tracking must maintain scene global identities across synchronized views, yet query-based trackers carry these identities only implicitly in the instance bank, where they fragment upon query interruption. We present an online architecture that recovers association accuracy by predicting IDs explicitly over recurrent sparse queries. An outside-in Sparse4D detector fuses calibrated views into world frame 3D detections while propagating a sparse query bank, and a causal MOTIP ID decoder associates detections against a finite trajectory memory. We adapt MOTIP's relative-ID prediction and recycled slot runtime to globally fused 3D observations, and introduce metric spatial gating and proximity based newborn recovery. On the official 2026 AI City Challenge Track 1 test set, our method raises HOTA from 29.63 with native instance bank identities to 38.01, primarily through an AssA increase from 20.83 to 31.10, and ranks third on the public leaderboard. Full-sequence validation over all 9,000 frames of each scene shows that decoupled ID training improves HOTA over native identities, whereas continuing detector training alongside the detached ID objective produces scene-dependent gains and losses.
Chinese Translation
在线多摄像头3D跟踪必须在同步视图中保持场景的全局身份，然而基于查询的跟踪器仅在实例库中隐式地携带这些身份，且在查询中断时会出现碎片化。我们提出了一种在线架构，通过在递归稀疏查询上显式预测ID来恢复关联准确性。外部稀疏4D检测器将校准视图融合到世界坐标系中的3D检测，同时传播稀疏查询库，而因果MOTIP ID解码器则将检测与有限的轨迹记忆关联。我们将MOTIP的相对ID预测和回收槽运行时间适配于全局融合的3D观测，并引入度量空间门控和基于接近度的新生物体恢复。在官方2026 AI City Challenge Track 1测试集上，我们的方法将HOTA从29.63（使用原生实例库身份）提高到38.01，主要通过AssA从20.83提升至31.10，并在公共排行榜上排名第三。对每个场景的9,000帧进行全序列验证表明，解耦ID训练在原生身份上提高了HOTA，而在分离ID目标的同时继续训练检测器则产生了场景依赖的增益和损失。
cs.CV / 49 / 2609.18379
JigSync: Gauge-Resolved Synchronization for Jigsaw Reassembly under Unknown Piece Orientation
JigSync：在未知拼图方向下的拼图重组的量规解析同步
Abstract
Square jigsaw reassembly requires recovering the spatial arrangement of shuffled fragments from their visual content and pairwise relationships. While recent studies have made substantial progress, existing benchmarks typically assume that all fragments are provided upright, reducing reassembly to a permutation problem. We study the generalized problem in which each fragment may also have gone through an unknown rotation. For this setting we establish a gauge-unobservability theorem: the minimum of the weighted least-squares objective is exactly invariant under a uniform global rotation of arbitrary magnitude, so no residual-based criterion can recover the global orientation. The theorem further identifies how the issue of global orientation can be resolved: an orientation anchor estimated from the content of a single fragment, lying outside its scope, suffices. To address the above, we propose JigSync, which attains 63.8% and 31.8% absolute accuracy (AA) on GAP-3 and GAP-5, respectively, the highest reported on both, while additionally recovering a rotation per piece that neither benchmark requires. We release JigSync, a degradation protocol that sweeps shape, erosion, photometry, grid size, and rotation independently.
Chinese Translation
平方拼图重组需要从拼碎片的视觉内容和成对关系中恢复其空间排列。尽管近期研究取得了显著进展，但现有基准通常假设所有碎片都是直立提供的，从而将重组问题简化为一个排列问题。我们研究了一个广义问题，其中每个碎片可能经历了未知的旋转。在这种情况下，我们建立了一个量规不可观测性定理：加权最小二乘目标的最小值在任意大小的均匀全局旋转下完全不变，因此没有基于残差的标准可以恢复全局方向。该定理进一步指出如何解决全局方向的问题：从单个碎片的内容中估计的方向锚点，位于其范围之外，足以解决此问题。为了解决上述问题，我们提出了JigSync，该方法在GAP-3和GAP-5上分别达到了63.8%和31.8%的绝对准确率（AA），是目前报告的最高水平，同时还恢复了每个碎片的旋转，而这两个基准都不要求。我们发布了JigSync，一个降解协议，独立地涵盖形状、侵蚀、光度、网格大小和旋转。
cs.CV / 50 / 2609.18393
MSR: Multiple Subject Reference for Video Generation
MSR：用于视频生成的多主体参考
Abstract
Conditioning a video generator on multiple images requires preserving appearance while associating each reference with its intended role. We present MSR (Multiple Subject Reference), a slot-aware conditioning scheme for LTX-based video generation. Each reference image is independently encoded as a static clip and represented by a separate latent-token group. A compact Fourier-feature multilayer perceptron adds a numeric slot embedding, while slot-dependent temporal offsets modify the group's rotary coordinates. The reference groups are prepended to noisy target tokens and serve as clean context during target-only flow-matching training. We implement this scheme through low-rank adaptation and release the resulting weights and inference workflows. Qualitative examples demonstrate compositions containing distinct characters and referenced environments in realistic and stylized scenes. Development observations suggest reduced reference confusion relative to an earlier continuous-reference baseline, while similar clothing, complex garments, and viewpoint changes remain challenging. We describe the conditioning mechanism, the retained training configuration, and the observed strengths and limitations of the released system. A supplementary audio-reference experiment adds voice conditioning while keeping the visual parameters frozen.
Chinese Translation
将视频生成器以多个图像为条件需要在保持外观的同时，将每个参考与其预期角色关联。我们提出了MSR（多主体参考），一种基于LTX的视频生成的槽感知条件方案。每个参考图像独立编码为静态剪辑，并由单独的潜在标记组表示。一个紧凑的傅里叶特征多层感知器添加了数值槽嵌入，而槽依赖的时间偏移则修改了该组的旋转坐标。参考组被预先添加到嘈杂的目标标记中，并在仅针对目标的流匹配训练中作为干净的上下文。我们通过低秩适应实现该方案，并发布了生成的权重和推理工作流程。定性示例展示了在现实和风格化场景中包含不同角色和参考环境的组合。开发观察表明，相较于早期的连续参考基线，参考混淆有所减少，但相似的服装、复杂的衣物和视角变化仍然具有挑战性。我们描述了条件机制、保留的训练配置以及发布系统的观察到的优缺点。一个补充的音频参考实验在保持视觉参数冻结的同时增加了声音条件。
cs.CV / 51 / 2609.18399
A Non-Linear Neuron Based Detection of Isolated Pixels in Binary and Grayscale Images using Contrast Sensitive Receptive Fields
基于非线性神经元的二值和灰度图像中孤立像素的对比敏感感受野检测
Abstract
Identifying isolated points is important in image processing applications such as medical imaging, astronomy and quality control management. Other domains, such as cybersecurity, also present challenges that can be framed as image processing problems. One example of particular interest is the identification of anomalous single nodes in spatially organised networks where groups of nodes in different regions share similar feature values. This task can involve both binary and more complex grayscale images. However, existing methods face limitations: template matching is infeasible for grayscale images, while 2nd order derivative based methods are highly sensitive to noise and require user-specified thresholds. To overcome these issues, a novel method is proposed for detecting meaningful single-pixel deviations in images. This approach modifies and extends a neuron model, originally designed for anomaly detection, to operate on spatially diameter limited receptive fields that incorporate excitatory and inhibitory regions. The result is a method that is free from user-specified thresholds and parameters, and can be applied to both binary and grayscale images, providing an effective, robust and efficient solution.
Chinese Translation
识别孤立点在医学成像、天文学和质量控制管理等图像处理应用中至关重要。其他领域，如网络安全，也提出了可以框定为图像处理问题的挑战。一个特别感兴趣的例子是识别空间组织网络中的异常单节点，其中不同区域的节点组共享相似的特征值。该任务可以涉及二值图像和更复杂的灰度图像。然而，现有方法面临限制：模板匹配对于灰度图像不可行，而基于二阶导数的方法对噪声高度敏感，并且需要用户指定的阈值。为了解决这些问题，提出了一种新方法，用于检测图像中有意义的单像素偏差。该方法修改并扩展了最初设计用于异常检测的神经元模型，使其能够在空间直径受限的感受野上运行，这些感受野结合了兴奋性和抑制性区域。最终得到的方法不依赖于用户指定的阈值和参数，能够应用于二值和灰度图像，提供了一种有效、稳健且高效的解决方案。
cs.CV / 52 / 2609.18406
Prosthesis-Aware 3D Human Pose Estimation: A Dataset and Benchmark for RSP Users
考虑假肢的3D人体姿态估计：针对RSP用户的数据集和基准
Abstract
Recovering 3D human body motion from video is important for applications such as rehabilitation assessment and sports performance evaluation. For prosthesis users, this requires capturing both natural body joints and the geometry of the prosthetic device, a challenge that existing methods are not designed to address. Model-based estimators rely on body models trained on non-amputee individuals and cannot represent prosthesis geometry, while model-free methods lack body kinematic priors and are unreliable under occlusion. This challenge is particularly prominent for users of running-specific prostheses (RSPs), where the RSP has a complex curved geometry and moves dynamically during exercise. To fill this gap, we collect RSP3D, the first 3D dataset of RSP users, covering essential daily-life and exercise actions from participants with varied amputation conditions, using a multi-camera marker-based motion capture setup. We formally define the task of prosthesis-aware 3D pose estimation, evaluate representative methods in a zero-shot setting, and confirm their individual limitations. We further propose a hybrid baseline combining model-based body joint estimation with model-free RSP shape recovery, establishing a starting point for future research.
Chinese Translation
从视频中恢复3D人体运动对于康复评估和运动表现评估等应用至关重要。对于假肢用户，这需要同时捕捉自然身体关节和假肢设备的几何形状，而现有方法并未针对这一挑战进行设计。基于模型的估计器依赖于在非截肢个体上训练的身体模型，无法表示假肢几何形状，而无模型的方法缺乏身体运动学先验，在遮挡情况下不可靠。这一挑战在使用跑步专用假肢（RSP）的用户中尤为突出，因为RSP具有复杂的曲线几何形状，并在运动过程中动态移动。为填补这一空白，我们收集了RSP3D，这是第一个涵盖不同截肢条件参与者的RSP用户的3D数据集，涵盖了日常生活和运动中的基本动作，采用多摄像头标记式运动捕捉设置。我们正式定义了考虑假肢的3D姿态估计任务，在零样本设置中评估了代表性方法，并确认了它们各自的局限性。我们进一步提出了一种混合基线，将基于模型的身体关节估计与无模型的RSP形状恢复相结合，为未来的研究建立了起点。
cs.CV / 53 / 2609.18413
Vocabulary-Guided Gait Recognition
词汇引导的步态识别
Abstract
What is a gait? Appearance-based gait networks consider a gait as the human shape and motion information from images. Model-based gait networks treat a gait as the human inherent structure from points. However, the considerations remain vague for humans to comprehend truly. In this work, we introduce a novel paradigm Vocabulary-Guided Gait Recognition, dubbed Gait-World, which attempts to explore gait concepts through human vocabularies with Vision-Language Models (VLMs). Although VLMs have achieved the remarkable progress in various vision tasks, the cognitive capability regarding gait modalities remains limited. The success element in Gait-World is the proper vocabulary prompt where this paradigm carefully selects gait cycle actions as Vocabulary Base, bridging the gait and vocabulary feature spaces and further promoting human understanding for the gait. How to extract gait features? Although previous gait networks have made significant progress, learning solely from gait modalities on limited gait databases makes it difficult to learn universal gait features for practicality. Therefore, we propose the first Gait-World model, dubbed {\alpha}-Gait, which guides the gait network learning with vocabulary knowledge from VLMs. However, due to the heterogeneity of the modalities, directly integrating vocabulary and gait features is highly challenging as they reside in different embedding spaces. To address the issues, {\alpha}-Gait designs Vocabulary Relation Mapper and Gait Fine grained Detector to map and establish vocabulary relations in the gait space for detecting corresponding gait features. Extensive experiments on CASIA-B, CCPG, SUSTech1K, Gait3D and GREW reveal the potential value and research directions of vocabulary information from VLMs in the gait field.
Chinese Translation
什么是步态？基于外观的步态网络将步态视为来自图像的人体形状和运动信息。基于模型的步态网络则将步态视为来自点的人体固有结构。然而，这些考虑对人类的理解仍然模糊。在本研究中，我们引入了一种新颖的范式——词汇引导的步态识别，称为 Gait-World，该范式试图通过视觉-语言模型（Vision-Language Models, VLMs）探索步态概念。尽管 VLMs 在各种视觉任务中取得了显著进展，但在步态模态方面的认知能力仍然有限。Gait-World 的成功要素在于适当的词汇提示，该范式仔细选择步态周期动作作为词汇基础，架起步态与词汇特征空间之间的桥梁，进一步促进人类对步态的理解。如何提取步态特征？尽管之前的步态网络取得了显著进展，但仅从有限的步态数据库中学习步态模态使得学习通用步态特征变得困难。因此，我们提出了第一个 Gait-World 模型，称为 {eta}-Gait，该模型利用来自 VLMs 的词汇知识指导步态网络学习。然而，由于模态的异质性，直接整合词汇和步态特征具有很高的挑战性，因为它们位于不同的嵌入空间。为了解决这些问题，{eta}-Gait 设计了词汇关系映射器和步态细粒度检测器，以在步态空间中映射和建立词汇关系，从而检测相应的步态特征。在 CASIA-B、CCPG、SUSTech1K、Gait3D 和 GREW 上进行的广泛实验揭示了 VLMs 在步态领域中词汇信息的潜在价值和研究方向。
cs.CV / 54 / 2609.18430
StrucPhysVideo: Learning Physical Dynamics from Structured Captions and Robot Actions
StrucPhysVideo：从结构化字幕和机器人动作中学习物理动态
Abstract
Modeling physical dynamics, including how objects move, interact, and change state, is central to video world models for embodied AI. We present StrucPhysVideo, a family of video world models that bridges physics-focused data curation with language- and action-conditioned prediction of scene evolution. Our data pipeline combines motion-aware video segmentation, quality and content filtering, and physical relevance verification with structured annotations of objects, materials, and temporally localized interactions. By disentangling camera motion from object behavior and explicitly describing contact, deformation, and state transitions, the pipeline provides supervision grounded in observable physical events. Building on these data, we introduce StrucPhysVideo-TI2V, a sparse Mixture-of-Experts (MoE) text-image-to-video model trained with a curriculum that progressively emphasizes physical dynamics while retaining general-domain video data. StrucPhysVideo-TI2V achieves state-of-the-art performance on Physics-IQ Verified, scoring 45.5% and outperforming Cosmos3-Super-Image2Video by 2.8 percentage points. Caption ablations across backbones further demonstrate the effectiveness of physics-focused supervision. We further extend StrucPhysVideo-TI2V to StrucPhysVideo-IA2V, an interactive image-action-to-video world model that predicts visual outcomes from robot end-effector commands. Action conditioning, causal autoregressive generation, and few-step distillation enable incremental robot rollouts with only four denoising steps. Together, StrucPhysVideo advances physical dynamics modeling from image- and language-conditioned video prediction toward action-driven interaction.
Chinese Translation
建模物理动态，包括物体如何移动、相互作用和状态变化，是具身人工智能视频世界模型的核心。我们提出了StrucPhysVideo，一系列视频世界模型，旨在将以物理为中心的数据整理与基于语言和动作的场景演变预测相结合。我们的数据管道结合了运动感知的视频分割、质量和内容过滤，以及与对象、材料和时间局部交互的结构化注释的物理相关性验证。通过将相机运动与物体行为分离，并明确描述接触、变形和状态转变，该管道提供了基于可观察物理事件的监督。在这些数据的基础上，我们引入了StrucPhysVideo-TI2V，这是一种稀疏的混合专家（Mixture-of-Experts, MoE）文本-图像-视频模型，采用逐步强调物理动态的课程进行训练，同时保留一般领域的视频数据。StrucPhysVideo-TI2V在Physics-IQ Verified上达到了最先进的性能，得分为45.5%，比Cosmos3-Super-Image2Video高出2.8个百分点。对不同骨干网络的字幕消融实验进一步证明了以物理为中心的监督的有效性。我们进一步将StrucPhysVideo-TI2V扩展为StrucPhysVideo-IA2V，这是一种交互式图像-动作-视频世界模型，能够从机器人末端执行器命令中预测视觉结果。动作条件、因果自回归生成和少步蒸馏使得仅需四个去噪步骤即可实现增量机器人推出。总之，StrucPhysVideo将物理动态建模从基于图像和语言的视频预测推进到基于动作驱动的交互。
cs.CV / 55 / 2609.18432
Occluded Gait Recognition with Mixture of Experts: An Action Detection Perspective
基于专家混合模型的遮挡步态识别：一种动作检测视角
Abstract
Extensive occlusions in real-world scenarios pose challenges to gait recognition due to missing and noisy information, as well as body misalignment in position and scale. We argue that rich dynamic contextual information within a gait sequence inherently possesses occlusion-solving traits: 1) Adjacent frames with gait continuity allow holistic body regions to infer occluded body regions; 2) Gait cycles allow information integration between holistic actions and occluded actions. Therefore, we introduce an action detection perspective where a gait sequence is regarded as a composition of actions. To detect accurate actions under complex occlusion scenarios, we propose an Action Detection Based Mixture of Experts (GaitMoE), consisting of Mixture of Temporal Experts (MTE) and Mixture of Action Experts (MAE). MTE adaptively constructs action anchors by temporal experts and MAE adaptively constructs action proposals from action anchors by action experts. Especially, action detection as a proxy task with gait recognition is an end-to-end joint training only with ID labels. In addition, due to the lack of a unified occluded benchmark, we construct a pioneering Occluded Gait database (OccGait), containing rich occlusion scenarios and annotations of occlusion types. Extensive experiments on OccGait, OccCASIA-B,Gait3D and GREW demonstrate the superior performance of GaitMoE.OccGait is available at https://github.com/BNU-IVC/OccGait.
Chinese Translation
在现实场景中，广泛的遮挡对步态识别构成挑战，因为缺失和噪声信息，以及身体位置和尺度的错位。我们认为，步态序列中丰富的动态上下文信息本质上具有解决遮挡的特性：1）相邻帧的步态连续性允许整体身体区域推断被遮挡的身体区域；2）步态周期允许整体动作与被遮挡动作之间的信息整合。因此，我们引入了一种动作检测视角，将步态序列视为动作的组合。为了在复杂的遮挡场景下检测准确的动作，我们提出了一种基于动作检测的专家混合模型（GaitMoE），由时间专家混合模型（MTE）和动作专家混合模型（MAE）组成。MTE通过时间专家自适应地构建动作锚点，MAE则通过动作专家自适应地从动作锚点构建动作提案。特别地，动作检测作为步态识别的代理任务，仅使用身份标签进行端到端联合训练。此外，由于缺乏统一的遮挡基准，我们构建了一个开创性的遮挡步态数据库（OccGait），其中包含丰富的遮挡场景和遮挡类型的注释。在OccGait、OccCASIA-B、Gait3D和GREW上的大量实验表明，GaitMoE具有优越的性能。OccGait可在 https://github.com/BNU-IVC/OccGait 获取。
cs.CV / 56 / 2609.18444
DR.WILSS: Diffusion-Based Replay for Weakly Supervised Continual Semantic Segmentation
DR.WILSS：基于扩散的弱监督持续语义分割重放
Abstract
Weakly supervised class-incremental semantic segmentation (WILSS) aims to train a segmentation model over multiple steps, each introducing new concepts to be learned with only image-level supervision. We introduce DR.WILSS, an innovative approach to address catastrophic forgetting in continual learning using diffusion-based generative replay. Our framework leverages language clues to guide the diffusion process, employing self-inpainting and regularization techniques to efficiently produce replay data, aiding the learning process. By generating high-quality replay data, the information from previously learned classes can be preserved during continual updates, a critical challenge in incremental learning scenarios. To further align the statistics of replay data with those of training samples, we apply LoRAs to the generative model. Experimental results demonstrate state-of-the-art performance across multiple benchmarks and generative architectures, while avoiding storage of training data and the use of additional resource-demanding tools during training. The proposed technique enables an optimal tradeoff between training complexity and inference-time accuracy, making DR.WILSS a promising solution for real-world applications.
Chinese Translation
弱监督类增量语义分割（WILSS）旨在通过多个步骤训练分割模型，每个步骤引入新的概念，仅依赖于图像级别的监督。我们提出了DR.WILSS，这是一种创新的方法，通过基于扩散的生成重放来解决持续学习中的灾难性遗忘问题。我们的框架利用语言线索指导扩散过程，采用自我修复和正则化技术高效生成重放数据，助力学习过程。通过生成高质量的重放数据，可以在持续更新过程中保留先前学习的类别信息，这在增量学习场景中是一个关键挑战。为了进一步使重放数据的统计特征与训练样本对齐，我们对生成模型应用了LoRAs。实验结果表明，在多个基准和生成架构上实现了最先进的性能，同时避免了在训练过程中存储训练数据和使用额外资源密集型工具。所提出的技术在训练复杂性和推理时间准确性之间实现了最佳权衡，使DR.WILSS成为现实应用中的一种有前景的解决方案。
cs.CV / 57 / 2609.18462
CSWAM: Better Causal Semantic Representations for Out-of-Distribution Generalization in World Action Models
CSWAM：用于世界行动模型中分布外泛化的更优因果语义表示
Abstract
FastWAM-style world action models enable efficient action-only inference, but generalize poorly under visual distribution shifts. Their reconstruction-oriented representations emphasize appearance-specific details, limiting generalization to unseen scenes and objects. Without observation history, the model also lacks temporal evidence for robustly identifying task-relevant state changes and motion in unfamiliar visual conditions. To address these limitations, we present the Causal Semantic World Action Model (CSWAM), which augments FastWAM with a causal semantic expert built on V-JEPA 2.1. V-JEPA provides temporally grounded representations of semantic state changes and motion with less dependence on appearance-specific details. The expert learns their future evolution from a sparse history of current and past observations and shares the history-derived context with both the video and action streams through causal attention. At inference, CSWAM conditions action denoising on the current video state and observed semantic history, retaining efficient action-only inference. We conduct simulation and real-robot experiments to evaluate generalization under distribution shifts. With embodied pretraining, CSWAM raises Randomized success on RoboTwin 2.0 Clean-to-Randomized transfer from 10.16% to 45.18%, a gain of 35.02 percentage points over FastWAM. Across two real-robot tasks and three OOD difficulty levels, CSWAM improves average success over FastWAM by 42.5 percentage points, from 27.5% to 70.0%.
Chinese Translation
FastWAM风格的世界行动模型能够实现高效的仅基于动作的推理，但在视觉分布变化下泛化能力较差。其重建导向的表示强调特定外观的细节，限制了对未见场景和物体的泛化。在没有观察历史的情况下，该模型也缺乏在不熟悉的视觉条件下稳健识别与任务相关的状态变化和运动的时间证据。为了解决这些局限性，我们提出了因果语义世界行动模型（CSWAM），该模型在FastWAM的基础上增强了一个基于V-JEPA 2.1的因果语义专家。V-JEPA提供了对语义状态变化和运动的时间上扎根的表示，减少了对特定外观细节的依赖。该专家从当前和过去观察的稀疏历史中学习其未来演变，并通过因果注意力与视频和动作流共享历史派生的上下文。在推理时，CSWAM基于当前视频状态和观察到的语义历史进行动作去噪，保持了高效的仅基于动作的推理。我们进行了模拟和真实机器人实验，以评估在分布变化下的泛化能力。通过具身预训练，CSWAM将RoboTwin 2.0从干净到随机化转移的随机成功率从10.16%提高至45.18%，相比FastWAM提升了35.02个百分点。在两个真实机器人任务和三个分布外难度级别中，CSWAM的平均成功率比FastWAM提高了42.5个百分点，从27.5%提升至70.0%。
cs.CV / 58 / 2609.18465
GeoCond: A Conditioning-Aware Reliability Adapter for Feed-Forward 3D Reconstruction
GeoCond：一种面向条件的可靠性适配器用于前馈3D重建
Abstract
Feed-forward 3D foundation models such as VGGT predict cameras, depth, and point maps in a single pass, but can fail silently under low overlap, low parallax, and extreme relative rotation. Stratified analyses over these factors show that these failures are governed by geometric conditioning and are poorly captured by native aleatoric confidence. We introduce GeoCond, a lightweight reliability adapter for frozen feed-forward 3D backbones. GeoCond reads the backbone's predicted geometry and outputs pose-level uncertainty and a refinement gate. During training, it can be supervised by frame-permutation orbit variance, ground-truth pose error when labels are available, or cycle residuals from unlabelled independent pose graphs. At inference, the default head requires only one backbone pass and a small MLP. On VGGT, GeoCond improves out-of-distribution (OOD) AUSE (area under the sparsification-error curve; lower is better) from $0.32$ to $0.20$ over native confidence, transfers zero-shot to outdoor extreme-view scenes, and avoids the collapse caused by applying bundle adjustment uniformly. Across multiple backbones, cycle-distilled variants provide a ground-truth-free adaptation route, including cases where permutation variance vanishes on equivariant models. The same reliability signal supports gated refinement, pose-graph weighting, calibration, curation, and capture decisions. Reliable feed-forward 3D reconstruction requires not only predicting geometry, but also knowing when that geometry should be trusted.
Chinese Translation
前馈3D基础模型如VGGT能够在单次传递中预测相机、深度和点图，但在低重叠、低视差和极端相对旋转的情况下可能会静默失败。对这些因素的分层分析表明，这些失败受几何条件的影响，并且原生的随机信心度无法很好地捕捉到这些问题。我们提出了GeoCond，一种轻量级的可靠性适配器，适用于冻结的前馈3D主干网络。GeoCond读取主干网络预测的几何信息，并输出姿态级的不确定性和精细化门。在训练过程中，可以通过帧置换轨道方差、当标签可用时的真实姿态误差，或来自未标记独立姿态图的循环残差进行监督。在推理时，默认头只需一次主干传递和一个小型多层感知器（MLP）。在VGGT上，GeoCond将原生信心度下的分布外（OOD）AUSE（稀疏化误差曲线下的面积；越低越好）从0.32提高到0.20，能够零-shot迁移到户外极端视角场景，并避免因均匀应用束调整而导致的崩溃。在多个主干网络中，循环蒸馏变体提供了一条无真实标签的适应路径，包括在等变模型上置换方差消失的情况。相同的可靠性信号支持门控精细化、姿态图加权、校准、策划和捕捉决策。可靠的前馈3D重建不仅需要预测几何信息，还需要知道何时应信任该几何信息。
cs.CV / 59 / 2609.18473
CADSplat: Sparse-View 3D Gaussian Splatting Aided by CAD Models for Robust, Photorealistic Digital-Twin Reconstruction
CADSplat：基于CAD模型的稀疏视图3D高斯点云重建框架，用于稳健的照片级真实数字双胞胎重建
Abstract
We present CADSplat, a framework that reconstructs photorealistic, geometrically accurate digital twins from sparse ($<15$ views), wide-baseline posed images of an object by regularizing 3D Gaussian Splatting (3DGS) with an explicit CAD shape prior. Using such a prior requires finding a CAD model whose shape resembles the object depicted in the images and determining the pose of each camera relative to the object. We obtain both by matching segmented object silhouettes against silhouettes rendered from a CAD library and keeping the camera-to-object poses of the best-matching model. We then anchor 3D Gaussian primitives to the surface of the retrieved model and jointly optimize the 3DGS parameters, the camera-to-object registration, and a non-rigid deformation field to account for shape differences between the physical object and the CAD model. Across two real-world datasets, CADSplat outperforms unconstrained, few-shot, and mesh-texturing baselines and degrades gracefully to as few as 3 views. Our experiments show that most of the gain in rendering quality comes from how the splats are constrained---a fixed set of splats tied to a surface and moved by a single smooth deformation field---rather than from the CAD shape itself. The CAD model adds shape knowledge where views are scarcest, in the sparsest captures and on strongly self-occluded objects, and it places every camera in the object's own frame. This enables applications beyond novel-view synthesis, such as markerless augmented reality registration, per-image object pose estimation, physical simulations, and the transfer of part labels from the design to the reconstruction.
Chinese Translation
我们提出了CADSplat，一个框架，通过使用显式的CAD形状先验，从稀疏（少于15个视图）、宽基线的物体姿态图像中重建照片级真实、几何精确的数字双胞胎。使用这样的先验需要找到一个与图像中描绘的物体形状相似的CAD模型，并确定每个相机相对于物体的姿态。我们通过将分割的物体轮廓与从CAD库渲染的轮廓进行匹配来获得这两者，并保持最佳匹配模型的相机到物体的姿态。然后，我们将3D高斯原语锚定到检索到的模型表面，并联合优化3D高斯点云（3DGS）参数、相机到物体的配准以及一个非刚性变形场，以考虑物理物体与CAD模型之间的形状差异。在两个真实世界数据集中，CADSplat的表现优于无约束、少样本和网格纹理基线，并且在视图数量减少到3时仍能优雅降级。我们的实验表明，渲染质量的大部分提升来自于点云的约束方式——一组固定的点云与表面绑定，并通过一个平滑的变形场移动——而不是来自CAD形状本身。CAD模型在视图最稀缺的地方提供了形状知识，尤其是在最稀疏的捕获和强自遮挡的物体上，并将每个相机放置在物体的自身坐标系中。这使得除了新视图合成之外的应用成为可能，例如无标记增强现实配准、每图像物体姿态估计、物理模拟以及从设计到重建的部件标签转移。
cs.CV / 60 / 2609.18488
Beyond Random Couplings: Contrastive Noise Alignment in Generative Flows
超越随机耦合：生成流中的对比噪声对齐
Abstract
Diffusion and flow-matching models are typically trained by corrupting data through independently sampled Gaussian noise. While simple and scalable, this forward process induces arbitrary data-noise couplings, forcing the network to learn high-curvature transports between unrelated endpoints. Existing optimal-transport methods reduce this burden by reassigning fixed noise samples to data, but the source noise distribution itself remains passive. To address this, we introduce Contrastive Noise Alignment (CNA), a training-time method that creates dynamic, contrastive couplings by optimizing the noise representations directly. By modeling the noise batch as an interacting particle system, CNA employs a cross-modal InfoNCE objective to align noise particles with their paired data targets. To prevent spatial collapse, this alignment is regularized using an angular entropy term and a radial norm penalty. We show theoretically that this equilibrium asymptotically preserves Gaussian structures, maintaining tractability during inference. Empirically, CNA improves the alignment between noise and data, reduces flow curvature, and provides better generation quality with fewer required sampling steps. For few-step, pixel-space generation (2-4 NFEs), CNA reduces FID by over 50\% compared to standard rectified flow, and by at least 24\% against Optimal Transport baselines.
Chinese Translation
扩散和流匹配模型通常通过独立采样的高斯噪声来破坏数据进行训练。尽管这种方法简单且可扩展，但这一前向过程会引入任意的数据-噪声耦合，迫使网络学习无关端点之间的高曲率传输。现有的最优传输方法通过将固定的噪声样本重新分配给数据来减轻这一负担，但源噪声分布本身仍然是被动的。为了解决这个问题，我们引入了对比噪声对齐（Contrastive Noise Alignment, CNA），这是一种在训练时通过直接优化噪声表示来创建动态对比耦合的方法。通过将噪声批次建模为一个相互作用的粒子系统，CNA采用跨模态的InfoNCE目标来对齐噪声粒子与其配对的数据目标。为了防止空间崩溃，这种对齐使用角熵项和径向范数惩罚进行正则化。我们理论上证明，这种平衡在渐近上保持高斯结构，在推理过程中保持可处理性。在实证上，CNA改善了噪声与数据之间的对齐，减少了流的曲率，并在所需采样步骤更少的情况下提供了更好的生成质量。在少步像素空间生成（2-4 NFEs）中，CNA相比于标准的校正流将FID减少了超过50%，并且与最优传输基线相比至少减少了24%。
cs.CV / 61 / 2609.18490
Learning A Unified Template for Gait Recognition
学习统一模板以进行步态识别
Abstract
"What I cannot create, I do not understand."Human wisdom reveals that creation is one of the highest forms of learning. For example, Diffusion Models have demonstrated remarkable semantic structure and memory in image generation, understanding, and restoration, which intuitively benefits representation learning. However, current gait networks rarely embrace this perspective, relying primarily on learning by contrasting gait samples under varying complex conditions, leading to semantic inconsistency and uniformity issues. To address these issues, we propose Origins with generative capabilities whose underlying philosophy is that different entities are generated from a unified template, inherently regularizing gait representations within a consistent and diverse semantic space to capture accurate gait differences. Admittedly, learning this unified template is exceedingly challenging, as it requires the comprehensiveness of the template to encompass gait representations with various conditions. Inspired by Diffusion Models, Origins diffuses the unified template into timestep templates for gait generative learning, and meanwhile transfers the unified template for gait representation learning. Especially, gait generative and representation learning serve as a unified framework for end-to-end joint training. Extensive experiments on CASIA-B, CCPG,SUSTech1K, Gait3D, GREW and CCGR-MINI demonstrate that Origins performs unified generative and representation learning, achieving superior performance.
Chinese Translation
"我无法创造的，我无法理解。"人类智慧揭示了创造是学习的最高形式之一。例如，扩散模型在图像生成、理解和恢复方面展示了显著的语义结构和记忆，这直观上有利于表示学习。然而，目前的步态网络很少采用这种视角，主要依赖于在不同复杂条件下对比步态样本进行学习，从而导致语义不一致和统一性问题。为了解决这些问题，我们提出了具有生成能力的Origins，其基本理念是不同实体是从统一模板生成的，内在地规范化步态表示，使其在一致且多样的语义空间中捕捉准确的步态差异。诚然，学习这个统一模板是极具挑战性的，因为它要求模板的全面性能够涵盖具有各种条件的步态表示。受扩散模型的启发，Origins将统一模板扩散为时间步模板以进行步态生成学习，同时转移统一模板以进行步态表示学习。特别地，步态生成和表示学习作为一个统一框架，支持端到端的联合训练。在CASIA-B、CCPG、SUSTech1K、Gait3D、GREW和CCGR-MINI上的大量实验表明，Origins实现了统一的生成和表示学习，取得了优越的性能。
cs.CV / 62 / 2609.18493
Semantic-ITC: A Frame-wise Indoor Mobile Laser Scanning Dataset and Benchmark for Semantic Segmentation
Semantic-ITC：一个用于语义分割的逐帧室内移动激光扫描数据集和基准
Abstract
Semantic labels for indoor mobile laser scanning (MLS) frames remain largely absent from current point cloud semantic segmentation benchmarks, which mainly focus on reconstructed indoor scenes or outdoor LiDAR perception. This paper introduces Semantic-ITC, to the best of our knowledge the first public dataset and benchmark for frame-wise indoor MLS semantic segmentation. The dataset contains 52 indoor sequences, 79,108 MLS frames, and 1.23 billion labeled points collected in classrooms, corridors, meeting rooms, offices, and study areas. Labels are attached directly to measured LiDAR points in each frame using 16 semantic classes covering structural elements, furniture, room equipment, vegetation, and other indoor objects. Semantic-ITC preserves the sparse, non-uniform, and frame-wise sampling pattern of indoor MLS, making it distinct from scene-level reconstructed point clouds and mesh-based indoor datasets. The annotations are produced by a hybrid workflow that combines predictions from a visual foundation model applied to synchronized RGB images, structural information from BIM, and manual refinement, with the final labels assigned to the original LiDAR frames. A single-frame benchmark is provided, and the best baseline reaches 79.27\% mIoU. Remaining errors are concentrated around object boundaries and ambiguous indoor classes, indicating the challenges of indoor MLS segmentation under sparse frame geometry and long-tailed class distributions. The dataset provides a public benchmark for evaluating semantic segmentation directly on measured indoor MLS frames and supports future studies on frame-wise indoor MLS semantic segmentation.
Chinese Translation
当前的点云语义分割基准主要集中在重建的室内场景或户外激光雷达感知上，室内移动激光扫描（MLS）帧的语义标签仍然缺乏。本文介绍了Semantic-ITC，这是我们所知的第一个公开的逐帧室内MLS语义分割数据集和基准。该数据集包含52个室内序列、79,108个MLS帧和12.3亿个标注点，这些数据是在教室、走廊、会议室、办公室和学习区域收集的。标签直接附加到每帧中测量的激光雷达点上，使用16个语义类别覆盖结构元素、家具、房间设备、植物和其他室内物体。Semantic-ITC保留了室内MLS的稀疏、不均匀和逐帧采样模式，使其与场景级重建点云和基于网格的室内数据集有所不同。注释是通过一种混合工作流程生成的，该流程结合了应用于同步RGB图像的视觉基础模型的预测、来自建筑信息模型（BIM）的结构信息以及人工精细化，最终标签分配给原始的激光雷达帧。提供了单帧基准，最佳基线达到79.27%的mIoU。剩余的错误集中在物体边界和模糊的室内类别周围，表明在稀疏帧几何和长尾类别分布下进行室内MLS分割的挑战。该数据集为直接在测量的室内MLS帧上评估语义分割提供了一个公共基准，并支持未来对逐帧室内MLS语义分割的研究。
cs.CV / 63 / 2609.18510
DiT-Garment: Garment Dynamics with Diffusion Transformers
DiT-Garment：基于扩散变换器的服装动态建模
Abstract
We present DiT-Garment to model dynamic 3D clothing over human body models in arbitrary motion. Unlike existing methods, DiT-Garment can animate garments with unseen designs and physical materials, while allowing for direct inference of deformations for any target pose. To achieve this, we leverage a 2D diffusion transformer architecture to learn 3D deformations in a 2D UV-space. As the result is non-deterministic, our generative model learns the distribution of possible outcomes. The template garment is represented as a 3D triangle mesh spatially aligned with a 3D human body model in a standardized pose. To work with different garment designs without the need of a common template or complex graph convolution operations, the diffusion transformer is conditioned on a 3D position map of the template, represented in UV-space, which allows to implicitly learn a deformation of the 3D space around the body in standard pose. Further conditioning on body motion and physical parameters allows to physically ground the model. We quantitatively and qualitatively evaluate DiT-Garment on both synthetic and real data. While only trained on synthetic simulations of automatically generated cloth designs, our method generalizes to captured and artist-made garment designs. Code and data are available for research purposes at https://dumoulina.github.io/dit-garment/.
Chinese Translation
我们提出了DiT-Garment，以建模在任意运动下的人体模型上的动态3D服装。与现有方法不同，DiT-Garment能够对未见过的设计和物理材料进行服装动画，同时允许对任何目标姿势的形变进行直接推断。为此，我们利用2D扩散变换器架构在2D UV空间中学习3D形变。由于结果是非确定性的，我们的生成模型学习可能结果的分布。模板服装被表示为与标准姿势下的3D人体模型空间对齐的3D三角网格。为了在不同的服装设计中工作，而无需共同模板或复杂的图卷积操作，扩散变换器以模板的3D位置图为条件，该图在UV空间中表示，从而允许隐式学习标准姿势下围绕身体的3D空间的形变。进一步以身体运动和物理参数为条件，使模型在物理上得到扎根。我们在合成数据和真实数据上对DiT-Garment进行了定量和定性评估。尽管仅在自动生成的布料设计的合成模拟上进行训练，我们的方法仍能推广到捕获的和艺术家制作的服装设计。代码和数据可在 https://dumoulina.github.io/dit-garment/ 上用于研究目的。
cs.CV / 64 / 2609.18511
Learning from Distributed Eyes: Leveraging Collaborative Perception for Automated Model Adaptation
来自分布式视角的学习：利用协同感知进行自动化模型适应
Abstract
In autonomous driving, perception models often struggle to generalize to new environments due to domain shifts. While unsupervised model adaptation offers a feasible solution without labor-intensive manual labeling, existing methods that rely solely on the ego-vehicle's data often lead to inferior pseudo-labeling performance. To address this critical issue, we propose LDE, Learning from Distributed ``Eyes", a novel framework that transforms collaborative perception (CP) into a source of high-quality supervision for model adaptation. This pseudo-labeling approach is hyperparameter-insensitive and relatively reliable, assuming CP often outperforms single-agent's perception. However, naively implementing this approach encounters (1) the communication bottleneck of sharing rich features under time and bandwidth constraints, (2) the view discrepancy between the CP view and the learner's Field of View (FoV), and (3) the unreliability even in CP-generated labels. To address these issues, we design an adaptation-oriented feature sharing mechanism that selectively transmits the most critical information for adaptation, an FoV filtering method that meticulously eliminates mismatched labels, and a curriculum learning strategy to progressively exploit pseudo labels. Extensive experiments on 3D object detection tasks demonstrate that LDE consistently outperforms both the pre-trained models and state-of-the-art unsupervised adaptation methods.
Chinese Translation
在自动驾驶中，由于领域转移，感知模型往往难以在新环境中进行泛化。虽然无监督模型适应提供了一种无需劳动密集型手动标注的可行解决方案，但现有方法仅依赖自我车辆的数据，通常导致伪标注性能较差。为了解决这一关键问题，我们提出了LDE（Learning from Distributed ``Eyes''），一个将协同感知（CP）转化为模型适应高质量监督源的新框架。这种伪标注方法对超参数不敏感且相对可靠，假设CP通常优于单一代理的感知。然而，简单地实施这一方法会遇到（1）在时间和带宽限制下共享丰富特征的通信瓶颈，（2）CP视角与学习者视场（FoV）之间的视图差异，以及（3）即使在CP生成的标签中也存在的不可靠性。为了解决这些问题，我们设计了一种面向适应的特征共享机制，选择性地传输适应所需的最关键信息，一种FoV过滤方法，仔细消除不匹配的标签，以及一种课程学习策略，以逐步利用伪标签。在3D物体检测任务上的大量实验表明，LDE始终优于预训练模型和最先进的无监督适应方法。
cs.CV / 65 / 2609.18540
SVMemAgent: A Streaming Video Memory Agent for Query-Agnostic Online Frame Selection
SVMemAgent：一种用于查询无关在线帧选择的流媒体视频记忆代理
Abstract
Most keyframe selection studies focus on offline settings, assuming access to the full video and query in advance. In contrast, real-world streaming scenarios require online frame selection under unknown video duration, without access to either the query or future frames during selection. To address this, we introduce Streaming Video Memory (SVMem), a compact and representative memory of previously observed content, updated continuously as the video stream unfolds. Building on this setting, we propose the Streaming Video Memory Agent (SVMemAgent), which dynamically maintains a memory by deciding at each timestep whether to replace an existing memory frame with the incoming frame or discard it. SVMemAgent is trained using Group Relative Policy Optimization (GRPO) with task-driven rewards derived from diverse question-answer pairs, implicitly exposing the policy to a distribution of queries during training so that SVMem retains generally informative frames at inference, when queries are unavailable. Experiments on both online and offline video benchmarks show that SVMemAgent consistently outperforms online frame selection baselines and achieves competitive performance with offline methods that assume access to the full video and query. Through task-driven rewards, SVMemAgent learns an emergent keyframe selection policy that prefers frames containing textual information, which may benefit downstream VideoQA tasks.
Chinese Translation
大多数关键帧选择研究集中于离线设置，假设可以提前访问完整视频和查询。相比之下，现实世界中的流媒体场景需要在未知视频时长的情况下进行在线帧选择，在选择过程中无法访问查询或未来帧。为了解决这一问题，我们提出了流媒体视频记忆（Streaming Video Memory，SVMem），这是一种紧凑且具有代表性的先前观察内容的记忆，随着视频流的展开而持续更新。在此基础上，我们提出了流媒体视频记忆代理（Streaming Video Memory Agent，SVMemAgent），该代理通过在每个时间步决定是用新到达的帧替换现有的记忆帧还是丢弃它，从而动态维护记忆。SVMemAgent使用基于任务驱动奖励的组相对策略优化（Group Relative Policy Optimization，GRPO）进行训练，这些奖励源自多样的问题-答案对，隐式地使策略在训练过程中接触到查询的分布，以便在推理时（当查询不可用时）SVMem能够保留一般性的信息帧。在在线和离线视频基准测试中的实验表明，SVMemAgent始终优于在线帧选择基线，并在假设可以访问完整视频和查询的离线方法中表现出竞争力。通过任务驱动奖励，SVMemAgent学习到了一种新兴的关键帧选择策略，偏好包含文本信息的帧，这可能对下游的视频问答任务（VideoQA）有益。
cs.CV / 66 / 2609.18542
Accuracy- and Real-Time-Aware 4D Radar Preprocessing for Autonomous Driving Perception Systems
面向自主驾驶感知系统的准确性与实时性兼顾的4D雷达预处理
Abstract
4D radar has emerged as a promising next-generation sensor for improving the robustness of autonomous driving perception systems because of its stable sensing capability under adverse weather conditions. However, deploying 4D radar in embedded environments with limited hardware resources requires radar-representation preprocessing that jointly considers perception accuracy, real-time performance, and computational complexity. This paper proposes a preprocessing framework for 4D-radar-based 3D object detection. First, Percentile-based 3D Shape Preservation (P3DP) extracts point clouds from radar tensors while preserving object-shape information and suppressing noise and false alarms. Second, Multi-frame-based Noise Point Discrimination using Kernel Density Estimation (MF-KDE) improves the density and reliability of sparse radar point clouds. Finally, Embedded \& NetScore (ENS) evaluates suitability for embedded deployment by jointly considering accuracy, real-time performance, adverse-weather robustness, and model complexity.
Chinese Translation
4D雷达作为一种有前景的下一代传感器，因其在恶劣天气条件下的稳定感知能力，正在改善自主驾驶感知系统的鲁棒性。然而，在硬件资源有限的嵌入式环境中部署4D雷达需要进行雷达表示预处理，以共同考虑感知准确性、实时性能和计算复杂性。本文提出了一种基于4D雷达的3D目标检测预处理框架。首先，基于百分位数的3D形状保留方法（Percentile-based 3D Shape Preservation, P3DP）从雷达张量中提取点云，同时保留物体形状信息并抑制噪声和误报。其次，基于多帧的噪声点区分方法（Multi-frame-based Noise Point Discrimination using Kernel Density Estimation, MF-KDE）提高了稀疏雷达点云的密度和可靠性。最后，嵌入式与网络评分（Embedded & NetScore, ENS）通过共同考虑准确性、实时性能、恶劣天气鲁棒性和模型复杂性来评估嵌入式部署的适用性。
cs.CV / 67 / 2609.18546
STUNet-Fusion: Spatiotemporal Needle-Tip Localization in Ultrasound Video via Multi-Channel Motion Fusion
STUNet-Fusion：通过多通道运动融合实现超声视频中的时空针尖定位
Abstract
Needle-tip localization in ultrasound remains challenging because the needle may appear weak, discontinuous, or partially invisible, while imaging artifacts and anatomical structures can produce similar responses. To address this problem, we propose STUNet-Fusion, a spatiotemporal framework for needle-tip localization in ultrasound videos. The proposed method formulates the input as a tri-channel spatio-temporal fusion tensor, comprising grayscale appearance, grid-based motion feature, and raw frame difference. A shared ResNet-34 encoder extracts spatial features, ConvLSTM integrates temporal dependencies, and a U-Net decoder reconstructs a dense probability heatmap. The final coordinates are extracted via a soft-argmax operation to achieve sub-pixel localization accuracy. Experimental results demonstrate that this spatiotemporal fusion strategy significantly improves localization robustness compared to conventional baselines.
Chinese Translation
超声中的针尖定位仍然具有挑战性，因为针可能出现信号弱、断续或部分不可见的情况，同时成像伪影和解剖结构可能产生类似的响应。为了解决这个问题，我们提出了STUNet-Fusion，一种用于超声视频中针尖定位的时空框架。该方法将输入形式化为一个三通道时空融合张量，包含灰度外观、基于网格的运动特征和原始帧差。共享的ResNet-34编码器提取空间特征，ConvLSTM集成时间依赖性，而U-Net解码器重建密集概率热图。最终坐标通过soft-argmax操作提取，以实现亚像素级定位精度。实验结果表明，与传统基线相比，这种时空融合策略显著提高了定位的鲁棒性。
cs.CV / 68 / 2609.18548
HAP: A Hand-Driven Active Perception Framework for Egocentric Head Motion Prediction
HAP：一种手驱动的主动感知框架用于自我中心头部运动预测
Abstract
Egocentric motion forecasting has primarily focused on hands and manipulated objects, leaving future human head motion comparatively underexplored. During manipulation, the head both redirects perception toward the target to acquire task-relevant evidence and coordinates with body and hand motion. We therefore formulate future six Degree of Freedom (6-DoF) head-motion prediction conditioned on observed hand motion and inferred target context, and propose HAP, a Hand-Driven Active Perception framework. HAP infers confidence for each target object from observed hand motion and object geometry. Then constructs a dynamic Predictive Target-Centric Amodal Occlusion Graph (P-TAOG) representing current and potential occlusion among candidate objects. Directed graph and causal temporal reasoning encode the evolving target conditioned perceptual state, which is fused with hand and head motion history. A horizon-wise gate then blends the learned trajectory with a constant velocity prior. We further introduce Bottle, an egocentric RGB-D dataset of object manipulation toward specified targets, with coordinated head and hand motion under changing target visibility. Experiments on the public dataset and Bottle show that HAP achieves lower head motion prediction errors than representative baselines, supporting the value of hand driven intention and dynamic occlusion reasoning for anticipating human head motion. Code will be released at https://HAP-ego.github.io/HAP.
Chinese Translation
自我中心运动预测主要集中在手部和操作物体上，而对未来人类头部运动的研究相对较少。在操作过程中，头部不仅将感知重定向到目标以获取与任务相关的证据，还与身体和手部运动进行协调。因此，我们基于观察到的手部运动和推断的目标上下文，提出了未来六自由度（6-DoF）头部运动预测的模型，并提出了HAP（手驱动的主动感知框架）。HAP从观察到的手部运动和物体几何形状推断每个目标物体的置信度。然后构建一个动态的预测目标中心模态遮挡图（P-TAOG），表示候选物体之间当前和潜在的遮挡关系。定向图和因果时间推理编码了不断变化的目标条件感知状态，并与手部和头部运动历史进行融合。接着，采用水平门控机制将学习到的轨迹与恒定速度先验进行融合。我们进一步引入Bottle，一个针对指定目标的自我中心RGB-D物体操作数据集，包含在目标可见性变化下协调的头部和手部运动。对公共数据集和Bottle的实验表明，HAP在头部运动预测误差方面优于代表性基线，支持了手驱动意图和动态遮挡推理在预测人类头部运动中的价值。代码将发布在 https://HAP-ego.github.io/HAP。
cs.CV / 69 / 2609.18554
CARA: Collision-Aware Resolution Adaptation for Multiresolution Hash Encoding Based Image Fitting
CARA：针对多分辨率哈希编码的图像拟合的碰撞感知分辨率适应
Abstract
Multiresolution hash encodings have recently enabled fast and high-fidelity implicit neural representations by storing multi-scale features in fixed-size hash tables along a geometric resolution schedule. However, the standard design is data-agnostic: different resolution levels receive identical hash-table capacity despite large differences in image frequency content. As a result, some levels experience severe hash collisions while others underutilize parameters, leading to inefficient capacity allocation. To address this issue, we propose Collision-Aware Resolution Adaptation (CARA), a method that assigns per-level resolutions by balancing the effective information load across hash levels. This adaptive allocation reduces capacity bottlenecks and improves parameter efficiency. In addition, we introduce an invertible pixel-shuffle transform that reduces hash load factors by redistributing spatial information, thereby mitigating collision-induced information loss without enlarging the hash tables. To support evaluation on extremely high-resolution data, we also curate, to the best of our knowledge, the first uncompressed whole-slide image dataset for academic research. Experiments on Kodak images, gigapixel natural images, and raw whole-slide images demonstrate that CARA consistently improves the fidelity-parameter trade-off. Our method matches state-of-the-art performance while using only $27.76%$ of the parameters, and achieves up to $6.11$ dB PSNR improvement at comparable parameter counts. Code is provided in the supplementary.
Chinese Translation
多分辨率哈希编码最近通过在几何分辨率调度中将多尺度特征存储在固定大小的哈希表中，实现了快速且高保真的隐式神经表示。然而，标准设计是数据无关的：不同的分辨率级别接收相同的哈希表容量，尽管图像频率内容存在较大差异。因此，一些级别经历严重的哈希碰撞，而其他级别则未能充分利用参数，导致容量分配效率低下。为了解决这个问题，我们提出了碰撞感知分辨率适应（CARA）方法，该方法通过平衡哈希级别之间的有效信息负载来分配每个级别的分辨率。这种自适应分配减少了容量瓶颈，提高了参数效率。此外，我们引入了一种可逆的像素洗牌变换，通过重新分配空间信息来减少哈希负载因子，从而减轻因碰撞引起的信息损失，而无需扩大哈希表。为了支持对极高分辨率数据的评估，我们还整理了迄今为止我们所知的第一个用于学术研究的未压缩全幻灯片图像数据集。在对Kodak图像、千亿像素自然图像和原始全幻灯片图像的实验中，CARA始终改善了保真度与参数之间的权衡。我们的方法在仅使用$27.76\%$的参数的情况下，达到了最先进的性能，并在可比参数数量下实现了高达$6.11$ dB的PSNR提升。代码已在补充材料中提供。
cs.CV / 70 / 2609.18562
Sim-to-Real Traffic Scene Understanding by Decoupling Semantics from Caption Generation with V-JEPA
通过使用 V-JEPA 解耦语义与标题生成实现模拟到真实的交通场景理解
Abstract
Track 2 of the AI City Challenge 2026 requires both visual question answering (VQA) and traffic event description generation under a challenging synthetic-to real domain shift. Existing vision-language approaches often entangle semantic understanding with language generation, making them susceptible to hallucination and inconsistent reasoning across event phases. In this work, we propose a decoupled semantic understanding framework that first resolves predefined traffic questions into structured semantic facts and subsequently uses these facts to guide caption generation. A frozen V-JEPA encoder extracts predictive scene representations, while a lightweight Llama-based predictor produces answers for VQA queries. To improve reliability, we introduce a training-free structured refinement mechanism that exploits statistical priors, inter-question relationships, and temporal event consistency to correct prediction errors. The refined semantic facts are then provided to Qwen3-VL-8B to generate pedestrian and vehicle descriptions for each traffic event. Experimental results on the official 2026 AI City Challenge Track 2 benchmark show that the proposed method achieves 87.09% VQA accuracy and an overall S2 score of 60.0853, ranking first among all participating teams. These results demonstrate that predictive world representations combined with structured semantic refinement enable more accurate and reliable traffic understanding, leading to higher-quality lan guage generation.
Chinese Translation
2026 年 AI 城市挑战赛的第二赛道要求在具有挑战性的合成到真实领域转变下进行视觉问答 (VQA) 和交通事件描述生成。现有的视觉语言方法通常将语义理解与语言生成纠缠在一起，使得它们容易受到幻觉和事件阶段之间不一致推理的影响。在本研究中，我们提出了一种解耦的语义理解框架，首先将预定义的交通问题转化为结构化的语义事实，随后利用这些事实指导标题生成。一个冻结的 V-JEPA 编码器提取预测场景表示，而一个轻量级的基于 Llama 的预测器为 VQA 查询生成答案。为了提高可靠性，我们引入了一种无训练的结构化精炼机制，利用统计先验、问题间关系和时间事件一致性来纠正预测错误。然后，将精炼后的语义事实提供给 Qwen3-VL-8B，以生成每个交通事件的行人和车辆描述。在 2026 年 AI 城市挑战赛第二赛道的官方基准测试中，实验结果表明，所提出的方法实现了 87.09% 的 VQA 准确率和 60.0853 的整体 S2 分数，排名所有参赛队伍第一。这些结果表明，结合结构化语义精炼的预测世界表示能够实现更准确和可靠的交通理解，从而提高语言生成的质量。
cs.CV / 71 / 2609.18578
Learning Where to Focus: Self-Supervised Multi-Scale ViTs for Histopathology
学习聚焦位置：用于组织病理学的自监督多尺度视觉变换器
Abstract
Pathologists diagnose diseases by first locating suspicious tissue and then examining it at higher magnification, whereas self-supervised vision transformers (ViTs) allocate the same spatial resolution to every image region despite diagnostic evidence being sparse and spanning multiple biological scales. Recent pathology foundation models have substantially improved representation quality by scaling training data and model capacity, but largely retain uniform tokenization. We instead investigate whether pathology representations can be improved by learning where to allocate spatial resolution during self-supervised learning. To this end, we propose CRAFT (Coarse-to-fine Region-Adaptive Feature Tokenization), a DINO-based framework that learns image-dependent mixed-scale representations by using self-supervised attention to selectively refine informative regions while preserving coarse context, together with a symmetric cross-scale regularization objective that encourages complementary coarse and fine representations. Across CAMELYON16, TCGA-Lung subtype classification, and TCGA-LUAD survival prediction, CRAFT consistently outperforms comparable-scale self-supervised methods while requiring lower inference computation. Despite using only a compact 22M parameter backbone trained on comparatively small pathology datasets, CRAFT remains competitive with, and often surpasses, substantially larger pathology foundation models.
Chinese Translation
病理学家通过首先定位可疑组织，然后在更高的放大倍数下进行检查来诊断疾病，而自监督视觉变换器（ViTs）则对每个图像区域分配相同的空间分辨率，尽管诊断证据稀疏且跨越多个生物学尺度。最近的病理基础模型通过扩大训练数据和模型容量显著提高了表示质量，但在很大程度上仍保留了均匀的标记化。我们则探讨在自监督学习过程中，通过学习在哪里分配空间分辨率，是否可以改善病理表示。为此，我们提出了CRAFT（Coarse-to-fine Region-Adaptive Feature Tokenization），这是一种基于DINO的框架，通过使用自监督注意力选择性地细化信息丰富的区域，同时保留粗略上下文，学习图像依赖的混合尺度表示，并结合对称的跨尺度正则化目标，鼓励互补的粗略和精细表示。在CAMELYON16、TCGA肺亚型分类和TCGA-LUAD生存预测中，CRAFT始终优于可比尺度的自监督方法，同时需要更低的推理计算。尽管仅使用一个在相对较小的病理数据集上训练的紧凑型2200万参数主干，CRAFT仍然具有竞争力，并且常常超越规模更大的病理基础模型。
cs.CV / 72 / 2609.18582
On-the-Fly Homographies Calibration for Multi-Camera Tracking
实时多摄像头同质性标定用于多摄像头跟踪
Abstract
Precise multi-camera tracking traditionally relies on rigorous 3D site calibration, yet this requirement is often operationally impossible in large-scale deployments. Privacy regulations frequently prohibit recording video for offline calibration; limited bandwidth precludes synchronizing high-resolution streams from hundreds of cameras; and covering immense physical sites with calibration targets is logistically infeasible. We present a multi-camera homography calibration system designed to overcome these barriers through "on-the-fly" geometric refinement. Starting from coarse manual homographies, we introduce a centroid-based projection optimization (PO) that continuously aligns the ground-plane geometry using live detection streams. Because PO operates asynchronously on already-transmitted, lightweight metadata, it adds zero computational latency to the real-time tracker. This allows the system to adapt automatically to camera movements or environmental changes without human intervention. This optimized geometry feeds a multi-camera bird's-eye-view (BEV) tracker that fuses detections and unifies trajectories across zones. Crucially, by operating strictly on live anonymous metadata, our solution ensures a privacy-safe, zero-overhead, and resilient tracking pipeline that maintains global consistency in dynamic environments where static, recorded-video calibration is impossible.
Chinese Translation
精确的多摄像头跟踪传统上依赖于严格的三维现场标定，但在大规模部署中，这一要求往往在操作上不可行。隐私法规通常禁止录制视频以进行离线标定；有限的带宽使得无法同步来自数百个摄像头的高分辨率流；而在庞大的物理场地上布置标定目标在后勤上也是不可行的。我们提出了一种多摄像头同质性标定系统，旨在通过“实时”几何优化克服这些障碍。从粗略的手动同质性开始，我们引入了一种基于质心的投影优化（Projection Optimization, PO），该方法利用实时检测流持续对齐地面几何。由于PO在已经传输的轻量级元数据上异步操作，它对实时跟踪器没有增加计算延迟。这使得系统能够在没有人工干预的情况下自动适应摄像头移动或环境变化。经过优化的几何信息为多摄像头鸟瞰视图（Bird's-Eye View, BEV）跟踪器提供支持，该跟踪器融合检测结果并统一跨区域的轨迹。重要的是，通过严格在实时匿名元数据上操作，我们的解决方案确保了一个安全隐私、零开销且具有弹性的跟踪管道，能够在静态录制视频标定不可能的动态环境中保持全局一致性。
cs.CV / 73 / 2609.18602
PULSE: Unlocking Practical Image Compression on Single-Thread CPU
PULSE：在单线程CPU上解锁实用图像压缩
Abstract
Despite recent progress in learned image compression, existing methods remain computationally expensive on resource-constrained hardware, particularly CPUs. We introduce PULSE, a practical codec that enables (1) low-latency decoding on diverse hardware platforms with an ultra-low-complexity 5.2 kMAC/pixel neural receiver, and (2) efficient bit-exact entropy coding with an integer linear CDF predictor and a meta prior. To recover compression performance under this tight budget, we introduce an agentic evolution process guided by heuristic probes that iteratively improves the architecture through human-LLM collaboration. PULSE decodes a 1080p image in 126 ms on a single CPU thread while achieving compression performance comparable to HM. After perceptual optimization, PULSE competes with larger perceptual codecs like MS-ILLM. Codes are at https://github.com/microsoft/GenCodec/tree/main/PULSE
Chinese Translation
尽管在学习图像压缩方面取得了近期进展，现有方法在资源受限的硬件上，特别是CPU上，仍然计算开销较大。我们介绍了PULSE，这是一种实用的编解码器，能够实现（1）在多种硬件平台上以超低复杂度5.2 kMAC/像素的神经接收器进行低延迟解码，以及（2）使用整数线性CDF预测器和元先验进行高效的比特精确熵编码。为了在这一紧张预算下恢复压缩性能，我们引入了一种由启发式探针引导的代理进化过程，通过人类与大语言模型（LLM）的协作迭代改进架构。PULSE在单个CPU线程上以126毫秒解码1080p图像，同时实现与HM相当的压缩性能。在感知优化后，PULSE与更大规模的感知编解码器如MS-ILLM竞争。代码可在https://github.com/microsoft/GenCodec/tree/main/PULSE获取。
cs.CV / 74 / 2609.18623
FIVE-VLA: Fast and EffectIVE Autonomous Driving with Recurrent Action Memory
FIVE-VLA：具有递归动作记忆的快速有效自主驾驶
Abstract
State-of-the-art vision-language-action models (VLA) for autonomous driving face critical limitations: excessive parameter counts, inefficient high-resolution image processing, and lack of temporal memory. We introduce Fast and EffectIVE VLA (FIVE-VLA) to address these through two key contributions. First, we employ an efficient vision encoder that processes high-resolution ($448 \times 896$) images while generating only 98 tokens, over $5\times$ fewer than existing approaches, and bypass text generation entirely for single-pass trajectory prediction. Second, we propose Recurrent Action Memory (RAM), a lightweight module that conditions action prediction on previous action tokens, providing temporal context critical for manoeuvres such as overtaking and emergency braking. With only 641M parameters, FIVE-VLA completes $\sim$10% more routes without traffic rule infractions than the previous state-of-the-art VLA on the challenging Bench2Drive closed-loop driving benchmark. Non-reactive open-loop simulation on the large-scale real-world NVIDIA Physical AI AV dataset shows 10.2% and 7.7% lower collision-violation rates than SimLingo in single- and four-view settings, respectively. Additionally, FIVE-VLA runs at $\sim$30 fps on an A100 and $\sim$4 fps on a T4 GPU (proxy to an edge device), representing an 8-30$\times$ speedup over previous methods.
Chinese Translation
最先进的视觉-语言-动作模型（VLA）在自主驾驶中面临着关键的限制：过多的参数数量、低效的高分辨率图像处理以及缺乏时间记忆。我们提出了快速有效的 VLA（FIVE-VLA）来解决这些问题，主要有两个关键贡献。首先，我们采用了一种高效的视觉编码器，该编码器在处理高分辨率（$448 imes 896$）图像时仅生成 98 个标记，比现有方法少超过 $5 imes$，并且在单次轨迹预测中完全跳过文本生成。其次，我们提出了递归动作记忆（Recurrent Action Memory, RAM），这是一个轻量级模块，它根据先前的动作标记来调整动作预测，提供了对超车和紧急刹车等机动操作至关重要的时间上下文。FIVE-VLA 仅用 641M 参数，在具有挑战性的 Bench2Drive 闭环驾驶基准测试中，完成的路线比之前的最先进 VLA 多约 10%，且没有违反交通规则。在大规模真实世界的 NVIDIA Physical AI AV 数据集上进行的非反应性开环仿真显示，在单视图和四视图设置中，其碰撞违规率分别比 SimLingo 低 10.2% 和 7.7%。此外，FIVE-VLA 在 A100 GPU 上运行约 30 fps，在 T4 GPU（相当于边缘设备）上运行约 4 fps，较之前的方法实现了 8-30 倍的加速。
cs.CV / 75 / 2609.18632
VibeAvatar: Aligning Phonetic Kinematics and Human Aesthetics for High-Fidelity Talking Avatar Synthesis
VibeAvatar：对齐语音运动学与人类美学以实现高保真对话头像合成
Abstract
Multi-modal talking avatar synthesis aims to generate realistic talking videos from a reference portrait and speech. Despite rapid progress in diffusion-based methods, existing approaches still struggle to jointly achieve accurate lip articulation, human-preferred motion aesthetics, and efficient inference. We observe that phonetic accuracy and motion aesthetics arise from fundamentally different sources and should be addressed at complementary stages rather than learned implicitly by a single generator. Based on this insight, we propose VibeAvatar, which disentangles these two objectives through a Phonetic Kinematics Adapter (PKA) that converts recognition-oriented speech features into phonetic-kinematic conditions at the conditioning stage, and an Aesthetic Motion Policy (AMP) that optimizes a flow-consistent stochastic sampling policy via Group Relative Policy Optimization (GRPO) at the post-training stage. With a lightweight flow-based motion generator operating in a compact 1D warp-based latent motion space, VibeAvatar achieves state-of-the-art results in articulation, aesthetics, and efficiency on both objective metrics and user studies, while generating a 10-second 512px video in under 10 seconds with only $\sim$3GB VRAM.
Chinese Translation
多模态对话头像合成旨在从参考肖像和语音生成逼真的对话视频。尽管基于扩散的方法取得了快速进展，但现有方法仍然难以同时实现准确的唇部发音、人类偏好的运动美学和高效的推理。我们观察到，语音准确性和运动美学源于根本不同的来源，因此应在互补的阶段进行处理，而不是通过单一生成器隐式学习。基于这一见解，我们提出了VibeAvatar，它通过一个语音运动学适配器（Phonetic Kinematics Adapter, PKA）将以识别为导向的语音特征转换为条件阶段的语音运动学条件，并通过一个美学运动策略（Aesthetic Motion Policy, AMP）在后训练阶段通过群体相对策略优化（Group Relative Policy Optimization, GRPO）优化流一致的随机采样策略。VibeAvatar采用轻量级的基于流的运动生成器，在紧凑的1D扭曲基础潜在运动空间中运行，在发音、美学和效率方面在客观指标和用户研究中均取得了最先进的结果，同时在不到10秒内生成一个10秒的512px视频，仅需约3GB的显存。
cs.CV / 76 / 2609.18634
GenStream: Semantic Streaming Framework for Generative Reconstruction of Human-centric Media
GenStream：用于人本媒体生成重建的语义流媒体框架
Abstract
Video streaming dominates global internet traffic, yet conventional pipelines remain inefficient for structured, human-centric content such as sports, performance, or interactive media. Standard codecs re-encode entire frames, foreground and background alike, treating all pixels uniformly and ignoring the semantic structure of the scene. This leads to significant bandwidth waste, particularly in scenarios where backgrounds are static and motion is constrained to a few salient actors. We introduce GenStream, a semantic streaming framework that replaces dense video frames with compact, structured metadata. Instead of transmitting pixels, GenStream encodes each scene as a combination of skeletal keypoints, camera viewpoint parameters, and a static 3D background model. These elements are transmitted to the client, where a generative model reconstructs photorealistic human figures and composites them into the 3D scene from the original viewpoint. This paradigm enables extreme compression, achieving over 99.9% bandwidth reduction compared to HEVC for the continuous data stream. We partially validate GenStream on Olympic figure skating footage and demonstrate potential for high perceptual fidelity under minimal data. While acknowledging the significant computational costs shifted to the client and challenges in generalization, GenStream opens new directions in volumetric avatar synthesis, canonical 3D actor fusion across views, and personalized viewing experiences, laying the groundwork for scalable, intelligent streaming in the post-codec era.
Chinese Translation
视频流媒体在全球互联网流量中占主导地位，但传统管道在处理结构化的人本内容（如体育、表演或互动媒体）时仍然效率低下。标准编解码器对整个帧进行重新编码，无论前景还是背景，均将所有像素视为统一的，忽视了场景的语义结构。这导致了显著的带宽浪费，特别是在背景静态且运动仅限于少数显著演员的场景中。我们提出了GenStream，这是一种语义流媒体框架，它用紧凑的结构化元数据替代了密集的视频帧。GenStream不是传输像素，而是将每个场景编码为骨架关键点、相机视点参数和静态3D背景模型的组合。这些元素被传输到客户端，在那里生成模型重建出逼真的人类形象，并将其合成到原始视点的3D场景中。这一范式实现了极端压缩，与HEVC相比，连续数据流的带宽减少超过99.9%。我们在奥林匹克花样滑冰视频上部分验证了GenStream，并展示了在最小数据下实现高感知保真度的潜力。尽管承认将显著的计算成本转移到客户端以及在泛化方面的挑战，GenStream为体积化虚拟形象合成、跨视角的标准3D演员融合和个性化观看体验开辟了新的方向，为后编解码时代的可扩展智能流媒体奠定了基础。
cs.CV / 77 / 2609.18667
Video-Based Markerless Motion Capture for Clinical and Rehabilitation Biomechanics: A PRISMA-ScR Scoping Review of Validated Architectures, Clinical Readiness, and Emerging Methods
基于视频的无标记运动捕捉在临床和康复生物力学中的应用：一项关于验证架构、临床准备情况和新兴方法的PRISMA-ScR范围审查
Abstract
Background.. Video-based markerless motion capture promises movement analysis without the cost, space and skin-marker constraints of optoelectronic systems, with particular potential for clinical and rehabilitation settings. Whether validated pipelines yet deliver clinically acceptable biomechanics, and how they relate to the underlying computer-vision research, remains unclear. Methods. We conducted a scoping review following the PRISMA extension for Scoping Reviews, with a registered protocol and searches of PubMed, Scopus and IEEE Xplore (January 2015 to February 2026; the computer-vision scan was updated to July 2026). A dual-tier design paired a primary corpus of validated biomechanical studies with a complementary, curated and deliberately non-exhaustive corpus of emerging computer-vision work, used qualitatively. We charted study characteristics, pipeline architecture, validation methods and joint-angle accuracy. Results. We included 117 studies, most published from 2024 onward and conducted on healthy adults walking in a laboratory. Pipelines formed five architectural families across monocular and multi-camera modalities; most reported raw joint angles without biomechanical refinement. Sagittal lower-limb agreement clustered around 5 to 6{\textdegree}, generally short of clinical acceptability, while out-of-plane kinematics, kinetics, and pathological or older populations were rarely validated. Emerging computer-vision building blocks (foundation-model mesh recovery, differentiable inverse kinematics, video-based kinetics) were almost absent from validated studies. Conclusions. Video-based markerless capture is not yet interchangeable with marker-based systems for clinical joint kinematics, and it remains barely validated where rehabilitation needs it most: older and pathological populations, out-of-plane kinematics, and kinetics. Mapping this evidence gap onto emerging computer-vision advances, we propose hypothesis-generating design guidelines, not a validated method, to steer the next generation of pipelines toward accessible, clinically meaningful movement analysis.
Chinese Translation
背景：基于视频的无标记运动捕捉技术承诺能够进行运动分析，而无需光电系统所需的高成本、空间和皮肤标记的限制，尤其在临床和康复环境中具有潜在的应用价值。然而，验证过的流程是否能够提供临床可接受的生物力学结果，以及它们与基础计算机视觉研究的关系，仍不明确。方法：我们按照PRISMA范围审查的扩展方法进行了范围审查，注册了协议，并在PubMed、Scopus和IEEE Xplore上进行了检索（2015年1月至2026年2月；计算机视觉的检索更新至2026年7月）。采用双层设计，将一组经过验证的生物力学研究与一组补充的、经过策划且故意不详尽的新兴计算机视觉工作相结合，进行定性分析。我们绘制了研究特征、流程架构、验证方法和关节角度的准确性。结果：我们纳入了117项研究，大多数研究发表于2024年及以后，且是在实验室中对健康成年人进行的步态分析。流程形成了五个架构家族，涵盖单目和多摄像头模式；大多数研究报告了未经生物力学精炼的原始关节角度。矢状面下肢的协议集中在5到6度之间，通常未达到临床可接受性，而超出平面的运动学、动力学以及病理或老年人群体的验证则很少出现。新兴的计算机视觉构建模块（基础模型网格恢复、可微分逆运动学、基于视频的动力学）在经过验证的研究中几乎不存在。结论：基于视频的无标记捕捉尚不能与基于标记的系统在临床关节运动学中互换，且在康复最需要的领域（老年和病理人群、超出平面的运动学和动力学）几乎没有得到验证。将这一证据缺口与新兴的计算机视觉进展相结合，我们提出了生成假设的设计指南，而非验证的方法，以引导下一代流程朝向可获取的、具有临床意义的运动分析发展。
cs.CV / 78 / 2609.18688
Generalist-Specialist Mixture-of-Experts for Rare Pathology Detection in Multimodal Imaging
用于多模态成像中稀有病理检测的通才-专才混合专家模型
Abstract
AI models for multimodal medical imaging must balance modality-specific specialization with cross-modal shared representations, a trade-off that pure Mixture-of-Experts (MoE) architectures currently fail to satisfy. Expert-based routing improves in-domain learning but may sacrifice cross-modal signals, which appear particularly important for rare (low-prevalence) pathologies in our experiments. To resolve this, we introduce Generalist-Specialist-MoE (GS-MoE), a two-branch (MoE) architecture that couples a cross-modal generalist model with distinct modality-specific specialists (experts) via domain-constrained feature fusion. On RadImageNet (1.35M images, 165 pathologies, three modalities), GS-MoE recovers detection of six low-prevalence pathologies on which every baseline scores F1 $=$ 0, with per-class gains up to +0.60 F1. It attains this while even slightly exceeding dense and specialist-only MoE aggregate baselines (MCC 0.770), while using ${\sim}53\%$ fewer active parameters at inference than the strongest investigated dense model.
Chinese Translation
多模态医学成像的人工智能模型必须在特定模态的专业化与跨模态共享表示之间取得平衡，而纯粹的混合专家（Mixture-of-Experts, MoE）架构目前无法满足这一权衡。基于专家的路由改善了领域内学习，但可能牺牲跨模态信号，而这些信号在我们的实验中对于稀有（低流行率）病理显得尤为重要。为了解决这个问题，我们引入了通才-专才混合专家模型（Generalist-Specialist-MoE, GS-MoE），这是一种双分支（MoE）架构，通过领域约束的特征融合，将跨模态通才模型与不同模态特定的专家（专才）结合在一起。在RadImageNet（135万张图像，165种病理，三种模态）上，GS-MoE恢复了六种低流行率病理的检测，而在这些病理上，每个基线的F1得分均为0，且每类的增益高达+0.60 F1。它在实现这一目标的同时，甚至略微超过了密集和仅专才的MoE聚合基线（MCC 0.770），并且在推理时使用的活跃参数比所调查的最强密集模型少约53%。
cs.CV / 79 / 2609.18690
RankGround: Efficient High-Resolution GUI Grounding via Lightweight Reranker-Guided Crop Selection
RankGround：通过轻量级重排序引导裁剪选择实现高效高分辨率GUI定位
Abstract
Graphical User Interface (GUI) grounding is a fundamental perception task for multimodal agents, enabling them to interpret natural language instructions and interact with digital interfaces. Existing methods face a fundamental trade-off between accuracy and efficiency: direct full-image inference often fails to capture small or visually similar UI elements, while multi-crop strategies improve localization at the cost of multiple expensive Vision-Language Model (VLM) calls per query. To address this challenge, we propose RankGround, a two-stage framework that achieves accurate GUI grounding with a single VLM call per query. Central to our approach is GroundRanker, a lightweight multimodal reranker that identifies the most promising crop from a dense candidate set. Because no off-the-shelf ranking dataset is available, we construct ranking supervision data from existing grounding datasets. A strict containment criterion and boundary-aware positive augmentation improve alignment and spatial coverage in cluttered layouts. GroundRanker is then trained with a two-stage curriculum: a pointwise objective first learns coarse containment, and a listwise objective refines subtle semantic and spatial distinctions among visually similar crops. Experimental results show that RankGround consistently outperforms strong baselines while reducing computational cost. It achieves 1.4 times faster inference and improves localization accuracy by 5.5% on average over the second-best method across all backbones and screen scales, establishing a new state of the art in both efficiency and precision for GUI grounding.
Chinese Translation
图形用户界面（GUI）定位是多模态智能体的一项基础感知任务，使其能够理解自然语言指令并与数字界面进行交互。现有方法在准确性和效率之间面临根本性权衡：直接的全图推理往往无法捕捉到小型或视觉上相似的用户界面元素，而多裁剪策略则以每个查询多次调用昂贵的视觉-语言模型（VLM）为代价来提高定位精度。为了解决这一挑战，我们提出了RankGround，一个两阶段框架，通过每个查询仅调用一次VLM实现准确的GUI定位。我们方法的核心是GroundRanker，一个轻量级的多模态重排序器，它从密集候选集中识别出最有前景的裁剪。由于没有现成的排序数据集可用，我们从现有的定位数据集中构建排序监督数据。严格的包含标准和边界感知的正向增强提高了在杂乱布局中的对齐和空间覆盖。然后，GroundRanker通过两阶段课程进行训练：第一阶段的点对点目标学习粗略的包含关系，第二阶段的列表目标则细化视觉上相似裁剪之间的微妙语义和空间差异。实验结果表明，RankGround在降低计算成本的同时，始终优于强基线。在所有骨干网络和屏幕尺度上，它实现了1.4倍的推理速度提升，并在定位准确性上平均提高了5.5%，在效率和精度方面确立了GUI定位的新状态。
cs.CV / 80 / 2609.18716
Mask IPL: Noise-Free Intrinsic Position Learning via Computation Graph Clipping for Event-Based Spike-Driven Tracking
掩码 IPL：通过计算图剪裁实现无噪声的内在位置学习用于基于事件的脉冲驱动跟踪
Abstract
Spiking Neural Networks (SNNs) match the event-driven nature of event cameras and naturally extract spatiotemporal features. These properties have motivated a series of recent studies on event-based tracking with SNNs. Intrinsic Position Learning (IPL) acquires strong position information without introducing additional parameters, making it a mainstream approach for position encoding in event-based spike-driven tracking. However, the mechanism behind its effectiveness lacks systematic theoretical analysis. Moreover, our analysis reveals that IPL introduces noise in both forward and backward propagation. The former increases inference error, while the latter prevents parameters from converging to better solutions. This paper presents a systematic analysis of IPL and demonstrates that its effectiveness stems from the synergy between IPL and multi-stage convolution. The zero blocks in the joint tensor act as zero padding for convolution, and the resulting boundary effect propagates layer by layer through multi-stage convolution. Every parameter update is therefore driven by a gradient that perceives the relative displacement between template and search frames. Positional encoding added after the convolutional stage cannot provide this information. We further propose a simple Computation Graph Clipping method that applies a validity mask determined by the layout to the operations of every layer, making invalid regions equivalent to zero padding in both forward and backward propagation. This eliminates the noise without introducing additional parameters and makes the actual gradient coincide with the ideal gradient. We name the improved method Mask IPL. Without increasing parameters or computational cost, Mask IPL improves the AUC of the Tiny-scale tracker on FE108, FELT, and VisEvent, and consistently improves the Base-scale tracker as well.
Chinese Translation
脉冲神经网络（SNNs）与事件相机的事件驱动特性相匹配，能够自然提取时空特征。这些特性促使了一系列关于基于事件的 SNN 跟踪的近期研究。内在位置学习（IPL）在不引入额外参数的情况下获取强大的位置信息，使其成为基于事件的脉冲驱动跟踪中位置编码的主流方法。然而，其有效性的机制缺乏系统的理论分析。此外，我们的分析揭示 IPL 在前向和反向传播中引入了噪声。前者增加了推理误差，而后者则阻止参数收敛到更好的解。本文对 IPL 进行了系统分析，并证明其有效性源于 IPL 与多阶段卷积之间的协同作用。联合张量中的零块充当卷积的零填充，导致的边界效应通过多阶段卷积逐层传播。因此，每次参数更新都是由感知模板与搜索帧之间相对位移的梯度驱动的。卷积阶段后添加的位置信息编码无法提供这一信息。我们进一步提出了一种简单的计算图剪裁方法，该方法根据布局对每一层的操作应用有效性掩码，使无效区域在前向和反向传播中等同于零填充。这消除了噪声而不引入额外参数，并使实际梯度与理想梯度一致。我们将改进的方法命名为掩码 IPL。在不增加参数或计算成本的情况下，掩码 IPL 提高了 Tiny-scale 跟踪器在 FE108、FELT 和 VisEvent 上的 AUC，并且持续改善了 Base-scale 跟踪器的表现。
cs.CV / 81 / 2609.18737
Geometry beneath the Waves: Dense Priors for Sparse-View Underwater 3D Gaussian Splatting
波浪下的几何：稀疏视图下的水下3D高斯点云的密集先验
Abstract
Underwater 3D reconstruction supports applications ranging from marine ecosystem monitoring and subsea inspection to underwater archaeology, education, and immersive visualisation. 3D Gaussian Splatting has made real-time photorealistic novel-view rendering practical, while underwater variants incorporate physically based image-formation models to separate medium effects from scene radiance. Their reconstruction quality, however, remains fundamentally limited by the geometry used for initialisation.
Chinese Translation
水下3D重建支持从海洋生态系统监测、海底检查到水下考古、教育和沉浸式可视化等多种应用。3D高斯点云技术使得实时逼真的新视角渲染成为可能，而水下变体则结合了基于物理的图像形成模型，以将介质效应与场景辐射分离。然而，它们的重建质量仍然在根本上受到用于初始化的几何形状的限制。
cs.CV / 82 / 2609.18753
Toward Markerless Video-based Tremor Analysis: Objective Quantification of Pathological Tremor in Mouse Preclinical Models
无标记视频基础的震颤分析：小鼠临床前模型中病理性震颤的客观量化
Abstract
Tremor is a movement disorder characterized by involuntary, rhythmic oscillations of body parts and is a hallmark of several neurological conditions, including Parkinson's disease and essential tremor. Elucidating its underlying mechanisms relies heavily on mouse models, which offer genetic manipulability and translational relevance to human neural circuitry. Accordingly, these models are indispensable for studying tremor pathophysiology. So far, electromyography and accelerometers have been used as methods to quantitatively observe tremors in mice. However, these methods have several drawbacks, such as high costs and complex setups. In particular, the invasive surgical implantation of devices causes significant stress to the animals. Although RGB-based methods offer non-invasive and cost-effective alternatives, they often lack the sensitivity required to detect subtle tremors. Therefore, this paper addresses these challenges by achieving mouse tremor severity estimation using conventional RGB cameras only. To address the challenging task of isolating tremor-related vibrations while the mouse itself is also in motion, our pipeline incorporates segmentation-based pre-processing to extract the mouse region and a Tremor Score Estimation Module that captures subtle tremors with high sensitivity. In the experiments, we assessed tremors in unrestrained mice using a non-invasive method with two standard cameras. The results demonstrated a strong correlation with accelerometer measurements and confirmed that the method accurately captured the intensity-dependent characteristics of tremors. The project page is available at https://isogawalab.github.io/Video-based-Tremor-Analysis-Project/.
Chinese Translation
震颤是一种运动障碍，其特征是身体部位的不自主、节律性振荡，是多种神经系统疾病的标志，包括帕金森病和原发性震颤。阐明其潜在机制在很大程度上依赖于小鼠模型，这些模型提供了遗传操作的可能性，并与人类神经回路具有转化相关性。因此，这些模型在研究震颤的病理生理学中不可或缺。迄今为止，肌电图和加速度计已被用作定量观察小鼠震颤的方法。然而，这些方法存在多个缺点，如高成本和复杂的设置。尤其是，设备的侵入性外科植入会给动物带来显著的压力。尽管基于RGB的方法提供了非侵入性和成本效益高的替代方案，但它们通常缺乏检测微妙震颤所需的灵敏度。因此，本文通过仅使用常规RGB相机实现小鼠震颤严重程度的估计，来解决这些挑战。为了应对在小鼠自身也在运动时隔离与震颤相关的振动这一具有挑战性的任务，我们的流程包括基于分割的预处理，以提取小鼠区域，以及一个震颤评分估计模块，该模块以高灵敏度捕捉微妙的震颤。在实验中，我们使用两台标准相机的非侵入性方法评估了不受限制的小鼠的震颤。结果显示与加速度计测量之间存在强相关性，并确认该方法准确捕捉了震颤强度依赖的特征。项目页面可访问 https://isogawalab.github.io/Video-based-Tremor-Analysis-Project/。
cs.CV / 83 / 2609.18773
DISTA-Net++: Rethinking Infrared Small Target Unmixing Beyond Sub-Pixel Separation
DISTA-Net++：重新思考红外小目标解混合超越亚像素分离
Abstract
Long-range infrared imaging frequently confronts dense target clusters whose diffraction-limited signatures merge into a single indistinguishable blob, concealing the number, sub-pixel positions, and radiant intensities of the underlying sources. While deep learning has advanced general object detection, resolving such Closely-Spaced Infrared Small Targets (CSIST) remains largely unexplored, owing to a systemic infrastructure void and a fundamental paradigm mismatch. The dominant formulation, which reduces unmixing to a blind, discrete sub-pixel separation, is inherently insufficient: without semantic guidance, the ill-posed inverse problem admits ambiguous solutions plagued by false and missed detections, while grid-based discretization locks predictions onto fixed lattice centers, chaining precision to prohibitively expensive grid refinement. We argue that CSIST unmixing should instead be informed and continuous. To ground this paradigm shift, we establish the first comprehensive open-source ecosystem for the field, comprising the large-scale CSIST-100K benchmark, a tailored metric suite, and the GrokCSO toolkit. Upon this foundation, we propose DISTA-Net++, which anchors a dynamic deep unfolding backbone with two synergistic mechanisms: a Count-Guided Prior that injects the global target count as an explicit semantic constraint to regularize the solution space, and a Continuous Coordinate Rectification that regresses off-grid offsets to decouple localization accuracy from grid resolution. Extensive experiments validate our paradigm: even under the most economical 3x division, DISTA-Net++ surpasses 7x-division state-of-the-art methods by 16.15% in CSO-mAP and 62.96% in count accuracy at merely one-sixth of their computation, demonstrating that unmixing precision need not be purchased with finer discretization. The complete ecosystem is available at https://github.com/GrokCV/GrokDet.
Chinese Translation
长距离红外成像常常面临密集目标簇，这些衍射极限特征合并为一个不可区分的斑点，掩盖了潜在源的数量、亚像素位置和辐射强度。尽管深度学习在一般物体检测方面取得了进展，但解决紧密间隔红外小目标（Closely-Spaced Infrared Small Targets，CSIST）仍然在很大程度上未被探索，这主要是由于系统基础设施的缺失和基本范式的不匹配。主流的解混合方法将其简化为盲目的离散亚像素分离，这种方法本质上是不够的：在没有语义指导的情况下，病态的逆问题会产生模糊的解决方案，受到虚假和漏检的困扰，而基于网格的离散化将预测锁定在固定的晶格中心，使得精度与代价高昂的网格细化相绑定。我们认为，CSIST解混合应当是有信息的和连续的。为了奠定这一范式转变的基础，我们建立了该领域第一个全面的开源生态系统，包括大规模的CSIST-100K基准、量身定制的度量套件和GrokCSO工具包。在此基础上，我们提出了DISTA-Net++，它通过两个协同机制锚定一个动态深度展开骨干网络：一个计数引导先验（Count-Guided Prior），将全局目标计数作为显式语义约束注入，以规范化解空间；以及一个连续坐标校正（Continuous Coordinate Rectification），回归离网偏移，以将定位精度与网格分辨率解耦。大量实验验证了我们的范式：即使在最经济的3倍划分下，DISTA-Net++在CSO-mAP上超越了7倍划分的最先进方法16.15%，在计数准确性上超越62.96%，而其计算量仅为后者的六分之一，证明了解混合精度不必通过更细的离散化来换取。完整生态系统可在https://github.com/GrokCV/GrokDet获取。
cs.CV / 84 / 2609.18823
Using OCR Heads to Verbalize Image Semantics
使用OCR头部来表述图像语义
Abstract
How do VLMs map from pixels to semantics? To understand this general question, we focus on a narrow one: studying how VLMs perform optical character recognition (OCR). Across four models, we identify attention heads causally necessary for OCR, and discover that these are in fact general-purpose heads that output interpretable semantic features across all image tokens. For example, pointing these heads at an image token containing the word "bike" causes Qwen3-VL-8B to output "bike," but pointing them at a bird wing causes the model to output the token "feathers." We collapse these heads' attention weights into a single verbalization lens transformation that reveals interpretable semantic features in hidden states across all layers. When combined with projection to vocabulary space, we can obtain interpretable labels starting from layer 0, showing that image representations are in fact aligned with language in early layers. We find that we can also use the inverse of this transformation to edit non-word concepts, e.g., replacing a tractor with a revolver in a naturalistic image, providing causal evidence that this subspace is useful for more than just OCR. Our results are an example of how the study of specific mechanisms can shed light on broader interpretability problems.
Chinese Translation
视觉语言模型（VLMs）是如何将像素映射到语义的？为了理解这个一般性问题，我们聚焦于一个狭窄的问题：研究VLMs如何执行光学字符识别（OCR）。在四个模型中，我们识别出因果上必要的注意力头，并发现这些实际上是通用头部，能够输出所有图像标记的可解释语义特征。例如，将这些头指向包含“自行车”一词的图像标记时，Qwen3-VL-8B模型输出“bike”，而将其指向鸟翅膀时，模型则输出标记“feathers”。我们将这些头的注意力权重合并为一个单一的表述透镜变换，揭示了所有层中隐藏状态的可解释语义特征。当与词汇空间的投影结合时，我们可以从第0层开始获得可解释的标签，显示图像表示在早期层中实际上与语言对齐。我们发现还可以使用这一变换的逆向来编辑非词概念，例如，在自然图像中用左轮手枪替换拖拉机，提供了因果证据，表明该子空间不仅仅对OCR有用。我们的结果是研究特定机制如何揭示更广泛可解释性问题的一个例子。
cs.CV / 85 / 2609.18836
Copy What Is Seen, Generate What Is Not: Training-Free Anomaly-Aware Video Restoration
复制可见内容，生成不可见内容：无训练异常感知视频修复
Abstract
A surveillance system that detects an anomaly often has to repair the footage as well, yet the two tasks are studied in isolation: training-free anomaly detectors stop at a score or a label, while training-free video editing answers to a user prompt rather than to a detector. This paper proposes AVR (Anomaly-aware Video Restoration), which closes that gap with frozen pretrained models alone and generates content only where the clip offers no evidence to copy. Motion evidence first gates open-vocabulary proposals into spatio-temporal masks. A background prior computed from the clip then fills every pixel the anomaly ever uncovers, leaving diffusion to synthesize only what no frame showed, and a frozen verifier decides per clip whether to trust a classical, a prior-anchored, or a background-conditioned restorer. Extensive experiments on three surveillance datasets, under both full-reference anomaly injection and real anomalies, show that AVR leads full-frame fidelity under oracle masks, matches three trained video inpainters inside the edited region, and outperforms a detect-then-generate pipeline on the masks it produces itself, while suppressing both the residual anomaly and the flicker of free diffusion.
Chinese Translation
监控系统在检测异常的同时，往往还需要修复视频画面，然而这两个任务通常是孤立研究的：无训练的异常检测器仅停留在得分或标签上，而无训练的视频编辑则是响应用户提示而非检测器。本文提出了 AVR（异常感知视频修复），该方法仅利用冻结的预训练模型来弥合这一差距，并仅在视频片段没有可复制证据的地方生成内容。运动证据首先将开放词汇提议引入时空掩码。然后，从视频片段计算出的背景先验填充异常所揭示的每个像素，留下扩散合成仅在没有任何帧显示的内容，而冻结的验证器则决定每个片段是信任经典修复器、先验锚定修复器，还是背景条件修复器。对三个监控数据集进行的大量实验，在全参考异常注入和真实异常的情况下，显示 AVR 在 oracle 掩码下实现了全帧保真度，在编辑区域内与三种训练过的视频修复器相匹配，并在其自身生成的掩码上优于检测-生成管道，同时抑制了残余异常和自由扩散的闪烁。
cs.CV / 86 / 2609.18860
Decodable but Misrouted: Sparse Features Uncover a Readout Gap in Vision-Language Models for Harmful Meme Detection
可解码但错误路由：稀疏特征揭示了视觉-语言模型在有害迷因检测中的读出差距
Abstract
When a large vision-language model misclassifies a harmful meme, the failure may reflect missing internal evidence or an inability to route represented evidence to its output. We distinguish these cases in Gemma-3 and Qwen3.5 using sparse autoencoders, role-conditioned probes, causal interventions, and recovery experiments across six harmful content benchmarks, with additional Spanish and Hindi-English code-mixed evaluations. Sparse readouts outperform native prediction on all six primary binary tasks: Qwen averages $0.740$ versus $0.432$ native macro-F1, while residual reconstruction reaches $0.486$, whereas Gemma improves from $0.532$ to $0.714$. These differences reflect supervised accessibility rather than a pre-existing, native decision rule, and the most influential token role depends on the task. Under the evaluated score scales, Qwen silent-feature ablation is $24-63$ times more probe-sensitive, whereas routed-feature patching on literal yes/no tasks is $16-140$ times more output-sensitive. Calibration-only routing recovers $93.3$% of the mean gap, and probe-distilled LoRA improves native predictions, although shared multi-task adaptation causes negative transfer. A case study of Gemma-3-12B on Facebook Hateful Memes finds a distributed rank-32 image-prompt interaction, reaching $0.756$ versus $0.685$ native macro-F1. Robustness controls show that the signal extends beyond English, is not explained solely by accompanying OCR, and depends on paired visual evidence. Thus, routing, rather than representation alone, is a recurring bottleneck in harmful meme classification.
Chinese Translation
当一个大型视觉-语言模型错误分类一个有害迷因时，这种失败可能反映了缺失的内部证据或无法将表示的证据路由到其输出。我们使用稀疏自编码器、角色条件探针、因果干预和恢复实验，在六个有害内容基准上区分了这些情况，此外还进行了西班牙语和印地语-英语混合评估。稀疏读出在所有六个主要二元任务上优于本地预测：Qwen的平均值为$0.740$，而本地宏-F1为$0.432$，同时残差重构达到$0.486$，而Gemma从$0.532$提高到$0.714$。这些差异反映了监督可达性，而不是预先存在的本地决策规则，且最具影响力的标记角色取决于任务。在评估的评分尺度下，Qwen的静默特征消融对探针的敏感性是$24-63$倍，而在字面是/否任务上，路由特征修补对输出的敏感性是$16-140$倍。仅通过校准路由恢复了$93.3$%的平均差距，探针提炼的LoRA改善了本地预测，尽管共享的多任务适应导致了负迁移。对Gemma-3-12B在Facebook仇恨迷因上的案例研究发现，分布式的rank-32图像提示交互达到了$0.756$，而本地宏-F1为$0.685$。稳健性控制显示信号超越英语，不仅仅由伴随的OCR解释，并且依赖于配对的视觉证据。因此，路由而非单纯的表示在有害迷因分类中是一个反复出现的瓶颈。
cs.CV / 87 / 2609.18898
NormLift: From Lifted Features To Semantic Reliability In 3D Gaussian Splatting
NormLift：从提升特征到3D高斯点云的语义可靠性
Abstract
Training-free weighted aggregation is widely used to lift 2D semantic features onto 3D Gaussians for open-vocabulary scene understanding, yet its theoretical role remains insufficiently understood. Existing analyses typically justify this operation from the rendering side, treating Gaussian features as linearly composable Euclidean variables for reconstructing 2D feature maps. However, this view does not match downstream 3D usage, where each Gaussian is often queried independently in a cosine-based embedding space. We revisit feature lifting from the 3D side and formulate per-Gaussian assignment as a cosine alignment problem on the CLIP unit sphere. Under this objective, the L2-normalized semantic back-projected feature emerges as the closed-form solution, providing a complementary interpretation of the standard lifting rule from the perspective of per-Gaussian semantic assignment. The same formulation further yields a norm decomposition into intra-view and inter-view consistency, suggesting that feature magnitude itself can serve as a semantic reliability signal. Calibrated by effective multi-view support, this reliability score guides a mode-voting refinement that preserves CLIP feature validity by avoiding linear averaging. Experiments on open-vocabulary 3D semantic segmentation show that NormLift is an efficient, training-free framework that achieves strong performance across evaluation protocols.
Chinese Translation
无训练的加权聚合被广泛用于将2D语义特征提升到3D高斯分布，以实现开放词汇场景理解，但其理论角色仍然不足以理解。现有分析通常从渲染的角度来证明这一操作，将高斯特征视为线性可组合的欧几里得变量，以重建2D特征图。然而，这一观点与下游3D使用不符，在下游应用中，每个高斯通常在基于余弦的嵌入空间中独立查询。我们从3D的角度重新审视特征提升，并将每个高斯的分配公式化为CLIP单位球上的余弦对齐问题。在这一目标下，L2归一化的语义反投影特征作为闭式解出现，从每个高斯的语义分配的角度提供了标准提升规则的补充解释。同样的公式进一步导致了范数分解为视内一致性和视间一致性，表明特征的大小本身可以作为语义可靠性信号。在有效的多视图支持下进行校准，该可靠性评分指导了一种模式投票的细化，避免了线性平均，从而保持了CLIP特征的有效性。在开放词汇3D语义分割上的实验表明，NormLift是一个高效的、无训练的框架，在各评估协议中表现出色。
cs.CV / 88 / 2609.18920
PhysVGGT: Feed-Forward Dense Physical Property Estimation from A Single Image
PhysVGGT：基于单幅图像的前馈密集物理属性估计
Abstract
Physical properties, such as friction, hardness, stiffness, and density, govern how robots should grasp, manipulate and interact with objects, yet estimating these properties from RGB images remains challenging. Existing methods typically employ per-object reconstruction augmented with physical properties or directly query vision-language models at test time, which results in substantial computational overhead that limits their applicability. In this work, we present PhysVGGT, a feed-forward model that predicts dense maps of friction coefficient, Shore hardness, Young's modulus, and density, together with object-level mass, from a single RGB image in one forward pass. The key idea of PhysVGGT is to formulate physical property estimation as a dense per-pixel prediction problem and employ a visual geometry transformer to extract geometry-aware tokens from the input image followed by a dense prediction branch for estimating local physical properties and a global prediction branch for estimating object-level mass. In addition, we introduce a scalable pseudo-label generation pipeline that enables large-scale weakly supervised training for dense physical property prediction, substantially reducing the need for expensive direct physical measurements. Extensive experiments show that PhysVGGT achieves state-of-the-art performance on the ABO-500 dataset and generalizes effectively to the out-of-distribution NeRF2Physics dataset. Moreover, PhysVGGT eliminates the need for per-object reconstruction and test-time optimization, achieving an inference latency of only 0.13s per image, making it $27\times$ faster than the previous state of the art.
Chinese Translation
物理属性，如摩擦、硬度、刚度和密度，决定了机器人如何抓取、操作和与物体互动，但从RGB图像中估计这些属性仍然具有挑战性。现有方法通常采用每个物体的重建，并增强物理属性，或在测试时直接查询视觉-语言模型，这导致了显著的计算开销，限制了它们的适用性。在本研究中，我们提出了PhysVGGT，这是一种前馈模型，可以在一次前向传递中从单幅RGB图像预测摩擦系数、肖氏硬度、杨氏模量和密度的密集图，以及物体级质量。PhysVGGT的关键思想是将物理属性估计公式化为密集的逐像素预测问题，并利用视觉几何变换器从输入图像中提取几何感知的标记，随后通过密集预测分支估计局部物理属性，并通过全局预测分支估计物体级质量。此外，我们引入了一种可扩展的伪标签生成管道，使得大规模弱监督训练成为可能，从而大幅减少对昂贵的直接物理测量的需求。大量实验表明，PhysVGGT在ABO-500数据集上实现了最先进的性能，并有效地推广到分布外的NeRF2Physics数据集。此外，PhysVGGT消除了对每个物体重建和测试时优化的需求，实现了每幅图像仅0.13秒的推理延迟，使其比之前的最先进技术快27倍。
cs.CV / 89 / 2609.18943
Dose-Aware Cold Diffusion with Physics Consistency for Generalizable Low-Dose CT Reconstruction
具有物理一致性的剂量感知冷扩散用于可推广的低剂量CT重建
Abstract
Reducing radiation dose in computed tomography significantly degrades image quality and poses challenges for accurate and clinically reliable reconstruction. While recent approaches have shown promise for low-dose CT, they often struggle to generalize across continuous and previously unseen dose levels, leading to artifacts and loss of anatomical detail. To address these limitations, we propose Dose-Aware Cold Diffusion (DACD), a physics-consistent reconstruction framework that explicitly models radiation dose as a continuous latent factor within a cold diffusion process. The proposed DACD framework integrates image-based dose-aware perception, multi-scale structural prior extraction, and dose-calibrated step allocation to adaptively guide the denoising trajectory. In addition, an iterative forward-backprojection correction is incorporated into the reverse refinement process to enforce projection-domain data consistency. Extensive experiments on three public benchmarks, including Mayo-2020, Mayo-2016, and LoDoPaB-CT, demonstrate that DACD consistently outperforms state-of-the-art diffusion-based and physics-guided methods in both quantitative accuracy and visual fidelity, particularly under ultra-low-dose conditions. The results show that DACD achieves robust generalization across a continuous range of dose levels, including those unseen during training.
Chinese Translation
在计算机断层扫描中减少辐射剂量会显著降低图像质量，并对准确和临床可靠的重建提出挑战。尽管最近的方法在低剂量CT方面显示出希望，但它们往往难以在连续和之前未见的剂量水平上进行推广，导致伪影和解剖细节的丧失。为了解决这些局限性，我们提出了剂量感知冷扩散（Dose-Aware Cold Diffusion, DACD），这是一种物理一致的重建框架，明确将辐射剂量建模为冷扩散过程中的一个连续潜在因素。所提出的DACD框架集成了基于图像的剂量感知感知、多尺度结构先验提取和剂量校准步长分配，以自适应地引导去噪轨迹。此外，迭代前向-反向投影校正被纳入反向精化过程中，以强制投影域数据一致性。在三个公共基准（包括Mayo-2020、Mayo-2016和LoDoPaB-CT）上的大量实验表明，DACD在定量准确性和视觉保真度方面始终优于最先进的基于扩散和物理引导的方法，尤其是在超低剂量条件下。结果表明，DACD在连续剂量水平范围内实现了稳健的推广，包括在训练期间未见的剂量水平。
cs.CV / 90 / 2609.18952
Automated Dental Caries Segmentation in Panoramic Radiographs Using Dual-Stage Deep Learning
基于双阶段深度学习的全景X光片中自动化龋齿分割
Abstract
Early detection of dental caries remains challenging due to limitations in traditional diagnostic methods, particularly for proximal lesions in posterior teeth. Deep learning models show promise for automated caries detection but face scalability constraints due to requirements for large volumes of expertly annotated training data. This study presents a dual-stage deep learning framework combining Faster R-CNN for tooth localization with U-Net for pixel-wise caries segmentation in panoramic radiographs. We developed a systematic transformation pipeline to convert large-scale polygon-annotated datasets into high-resolution binary segmentation masks, enabling pixel-wise supervised learning. The framework was trained using both expert-verified datasets and algorithmically processed labels from 3,000 panoramic images. Our approach achieved robust performance with an IoU of 0.9013, Dice coefficient of 0.9482, Recall of 0.9433, and Precision of 0.9774, demonstrating superior accuracy compared to existing methods while significantly reducing false-positive rates. The dual-stage framework effectively addresses data annotation bottlenecks in dental AI applications and demonstrates potential for scalable, automated caries detection systems that can improve diagnostic consistency and support clinical decision-making.
Chinese Translation
由于传统诊断方法的局限性，尤其是在后牙近邻病变的检测上，龋齿的早期发现仍然具有挑战性。深度学习模型在自动化龋齿检测方面显示出潜力，但由于对大量专家标注训练数据的需求，面临可扩展性限制。本研究提出了一种双阶段深度学习框架，将Faster R-CNN用于牙齿定位，并结合U-Net进行全景X光片中像素级龋齿分割。我们开发了一个系统化的转换管道，将大规模多边形标注数据集转换为高分辨率二值分割掩膜，从而实现像素级的监督学习。该框架使用专家验证的数据集和来自3000张全景图像的算法处理标签进行训练。我们的方案在IoU为0.9013、Dice系数为0.9482、召回率为0.9433和精确率为0.9774的情况下，表现出强大的性能，显示出比现有方法更高的准确性，同时显著降低了假阳性率。双阶段框架有效解决了牙科人工智能应用中的数据标注瓶颈，展示了可扩展的自动化龋齿检测系统的潜力，这可以提高诊断一致性并支持临床决策。
cs.CV / 91 / 2609.18955
KDTwin: Task-Aware Knowledge Distillation for Lightweight Multi-Task Driving Scene Segmentation
KDTwin：面向任务的知识蒸馏用于轻量级多任务驾驶场景分割
Abstract
Efficient perception models are essential for real-time autonomous driving, where accuracy and computational cost must be carefully balanced. However, applying knowledge distillation to multi-task driving scene segmentation is challenging because drivable-area and lane segmentation exhibit different spatial characteristics and class imbalance. We propose KDTwin, a task-aware distillation framework for lightweight multi-task segmentation networks. The proposed method performs distillation at both the shared encoder and task-specific decoders. Encoder-level pairwise distillation transfers spatial relational knowledge to enhance the student's shared representation. For the decoders, we use a weighted loss for drivable-area segmentation and a boundary-aware loss for lane segmentation, enabling task-adaptive knowledge transfer without increasing inference complexity. Experiments on BDD100K show consistent improvements across the evaluated CNN-based and Transformer-based student models without increasing inference-time parameters or FLOPs. The results show that designing distillation objectives according to task-specific characteristics can effectively enhance multi-task segmentation performance for autonomous driving. The source code is available at https://github.com/chequanghuy/KDTwin.
Chinese Translation
高效的感知模型对于实时自主驾驶至关重要，其中准确性和计算成本必须谨慎平衡。然而，将知识蒸馏应用于多任务驾驶场景分割面临挑战，因为可行驶区域和车道分割表现出不同的空间特征和类别不平衡。我们提出了KDTwin，一种面向任务的蒸馏框架，用于轻量级多任务分割网络。所提出的方法在共享编码器和任务特定解码器上进行蒸馏。编码器级的成对蒸馏传递空间关系知识，以增强学生的共享表示。对于解码器，我们使用加权损失进行可行驶区域分割，并使用边界感知损失进行车道分割，从而实现任务自适应的知识转移，而不增加推理复杂度。在BDD100K上的实验表明，在评估的基于CNN和基于Transformer的学生模型中，均取得了一致的改进，而没有增加推理时间参数或FLOPs。结果表明，根据任务特定特征设计蒸馏目标可以有效提升自主驾驶的多任务分割性能。源代码可在https://github.com/chequanghuy/KDTwin获取。
cs.CV / 92 / 2609.18971
LaSeD: Label-Semantic Self-Distillation for Visual-Only Surgical Phase Recognition
LaSeD：用于仅视觉的手术阶段识别的标签语义自蒸馏
Abstract
Surgical phase recognition maps each video frame to a clinically meaningful workflow phase, supporting context-aware assistance, documentation, and postoperative analysis. Most methods treat phase annotations only as class IDs, whereas recent surgical vision-language models often require additional video--text data, captions, or instruction tuning. We propose \emph{LaSeD}, a label-semantic self-distillation framework that uses phase names as privileged training-time context while retaining visual-only deployment without a ground-truth phase-name hint. LaSeD initializes a frozen teacher and a student from the same pretrained VLM checkpoint. The teacher receives the frame, a fixed task prompt, and the ground-truth phase-name hint; the student receives the same frame and prompt without the hint, and only its visual encoder is optimized. Training combines hard phase-token supervision with feature-level distillation from cached teacher representations. At inference, the teacher and hint are removed, and the student predicts one of the seven Cholec80 phases through constrained digit-token logits without an additional classifier head. On the Cholec80 evaluation split, LaSeD achieves 86.20\% accuracy, 77.75\% macro recall, 78.13\% macro precision, and 64.15\% macro Jaccard. Under the identical protocol, it improves a visual-only Qwen3-VL-4B baseline by 9.45, 9.42, 11.93, and 12.22 percentage points, respectively. Visual-only means that the image is the only sample-specific inference input, while all frames share the same fixed task prompt. These results suggest that phase names provide a useful low-cost signal for adapting VLMs to surgical workflow analysis. (The code will be published soon.)
Chinese Translation
手术阶段识别将每个视频帧映射到具有临床意义的工作流程阶段，支持上下文感知的辅助、文档记录和术后分析。大多数方法仅将阶段注释视为类ID，而最近的手术视觉-语言模型通常需要额外的视频-文本数据、字幕或指令调优。我们提出了 extit{LaSeD}，一种标签语义自蒸馏框架，利用阶段名称作为特权的训练时上下文，同时在没有真实阶段名称提示的情况下保持仅视觉的部署。LaSeD从同一预训练的视觉-语言模型（VLM）检查点初始化一个冻结的教师模型和一个学生模型。教师模型接收帧、固定任务提示和真实阶段名称提示；学生模型接收相同的帧和提示，但没有提示，且仅优化其视觉编码器。训练结合了硬阶段标记监督与来自缓存教师表示的特征级蒸馏。在推理时，教师和提示被移除，学生通过受限的数字标记逻辑预测七个Cholec80阶段之一，而无需额外的分类器头。在Cholec80评估分割上，LaSeD达到了86.20%的准确率、77.75%的宏观召回率、78.13%的宏观精确率和64.15%的宏观Jaccard。在相同协议下，它分别提高了仅视觉的Qwen3-VL-4B基线9.45、9.42、11.93和12.22个百分点。仅视觉意味着图像是唯一的样本特定推理输入，而所有帧共享相同的固定任务提示。这些结果表明，阶段名称为适应VLM进行手术工作流程分析提供了有用的低成本信号。（代码将很快发布。）
cs.CV / 93 / 2609.18980
Instrument Classification of Solo Sheet Music Images
独奏乐谱图像的乐器分类
Abstract
This paper studies instrument classification of solo sheet music. Whereas previous work has focused on instrument recognition in audio data, we instead approach the instrument classification problem using raw sheet music images. Our approach first converts the sheet music image into a sequence of musical "words" based on the bootleg score representation, and then treats the problem as a text classification task. We show that it is possible to significantly improve classifier performance by training a language model on unlabeled data, initializing a classifier with the pretrained language model weights, and then finetuning the classifier on labeled data. In this work, we train AWD-LSTM, GPT-2, and RoBERTa models on solo sheet music images from IMSLP for eight different instruments. We find that GPT-2 and RoBERTa slightly outperform AWD-LSTM, and that pretraining increases classification accuracy for RoBERTa from 34.5% to 42.9%. Furthermore, we propose two data augmentation methods that increase classification accuracy for RoBERTa by an additional 15%.
Chinese Translation
本文研究了独奏乐谱的乐器分类。以往的研究主要集中在音频数据中的乐器识别，而我们则采用原始乐谱图像来解决乐器分类问题。我们的方法首先基于盗版乐谱表示法将乐谱图像转换为一系列音乐“词”，然后将该问题视为文本分类任务。我们展示了通过在未标记数据上训练语言模型、用预训练的语言模型权重初始化分类器，并在标记数据上微调分类器，可以显著提高分类器的性能。在本研究中，我们对来自IMSLP的独奏乐谱图像进行了AWD-LSTM、GPT-2和RoBERTa模型的训练，涵盖八种不同的乐器。我们发现GPT-2和RoBERTa的表现略优于AWD-LSTM，并且预训练使RoBERTa的分类准确率从34.5%提高到42.9%。此外，我们提出了两种数据增强方法，使RoBERTa的分类准确率进一步提高了15%。
cs.CV / 94 / 2609.19119
Track, Articulate, Act: Generating Articulation from Casual Human Videos
跟踪、表述、行动：从随意的人类视频生成关节动作
Abstract
Human videos contain rich causal evidence for robot manipulation: they reveal how hand motion induces object motion and produces task-relevant changes in object state. In this work, we study articulated objects such as doors, drawers, cabinets, laptops, ovens, and hinged containers that are ubiquitous in daily life and present unique challenges for embodied interaction. These objects cannot be represented by a single pose; their motion depends on the underlying parts and joints. We introduce a real-to-sim framework that reconstructs a simulation-ready articulated object and hand-object interaction from a casual monocular RGB video, without RGB-D or multi-view input, prior scans, manually specified joints, or robot demonstrations. Our key insight is that dense 3D point tracks provide an embodiment-agnostic articulation cue: points on the fixed link remain approximately stationary, while points on the moving link follow coherent revolute or prismatic motion. Our method segments the links, estimates the joint and its state trajectory, reconstructs an articulated asset, and aligns the recovered 3D hand motion with the object. Central to our approach is a modular recipe that repurposes powerful pretrained models for single-image 3D reconstruction, mesh segmentation, and 3D scene flow, connecting their predictions through explicit geometric reasoning to infer articulation. We use the reconstructed articulated object and the human hand trajectory to replay interactions through contact in MuJoCo. The framework shows how pretrained vision models and explicit motion reasoning can turn casual human videos into articulated object models suitable for downstream embodied interactions. https://track-articulate-act.github.io/
Chinese Translation
人类视频包含丰富的因果证据，能够用于机器人操作：它们揭示了手部运动如何引发物体运动，并产生与任务相关的物体状态变化。在本研究中，我们研究了诸如门、抽屉、橱柜、笔记本电脑、烤箱和铰链容器等日常生活中普遍存在的关节物体，这些物体在具身交互中面临独特的挑战。这些物体无法用单一姿态表示；它们的运动依赖于其底层部件和关节。我们引入了一个真实到仿真的框架，该框架能够从随意的单目RGB视频中重建一个适合仿真的关节物体及手物体交互，而无需RGB-D或多视角输入、先前扫描、手动指定的关节或机器人演示。我们的关键见解是，稠密的3D点轨迹提供了一种与具身无关的关节线索：固定连接上的点保持大致静止，而移动连接上的点则遵循一致的旋转或伸缩运动。我们的方法对连接进行分割，估计关节及其状态轨迹，重建关节资产，并将恢复的3D手部运动与物体对齐。我们方法的核心是一个模块化的流程，重新利用强大的预训练模型进行单图像3D重建、网格分割和3D场景流，通过显式的几何推理连接它们的预测，以推断关节动作。我们使用重建的关节物体和人类手部轨迹，通过接触在MuJoCo中重放交互。该框架展示了预训练视觉模型和显式运动推理如何将随意的人类视频转化为适合下游具身交互的关节物体模型。
cs.CV / 95 / 2609.19122
Adaptive Convolutional Sparse Coding via Information Bottleneck for Robust Visual Signal Representation
通过信息瓶颈的自适应卷积稀疏编码用于鲁棒的视觉信号表示
Abstract
Visual signals require compact yet sufficient representations for robust downstream prediction. Convolutional sparse coding (CSC) provides an explicit mechanism for suppressing redundant components while preserving signal content, but its sparsity coefficient is typically fixed and manually selected. We propose an adaptive convolutional sparse coding framework for robust visual signal representation. Specifically, we unfold the CSC optimization with the Fast Iterative Shrinkage-Thresholding Algorithm (FISTA) and treat the sparsity coefficient as a differentiable variable jointly learned with the network parameters. From the information bottleneck perspective, this coefficient controls the trade-off between information retention and compression: the sparsity term promotes compact representations, while the reconstruction term together with task loss preserves task-relevant signal content. We further introduce a label-free post-training strategy that adjusts the compression strength for corrupted inputs with the main network parameters fixed. Experiments on CIFAR and ImageNet demonstrate competitive clean-data recognition and greatly improved robustness under different input perturbations.
Chinese Translation
视觉信号需要紧凑而充分的表示，以实现鲁棒的下游预测。卷积稀疏编码（CSC）提供了一种显式机制，用于抑制冗余成分，同时保留信号内容，但其稀疏系数通常是固定的并且手动选择。我们提出了一种自适应卷积稀疏编码框架，以实现鲁棒的视觉信号表示。具体而言，我们利用快速迭代收缩阈值算法（FISTA）展开CSC优化，并将稀疏系数视为一个可微分变量，与网络参数共同学习。从信息瓶颈的角度来看，该系数控制信息保留与压缩之间的权衡：稀疏项促进紧凑表示，而重建项结合任务损失则保留与任务相关的信号内容。我们进一步引入了一种无标签的后训练策略，该策略在主要网络参数固定的情况下，调整对受损输入的压缩强度。在CIFAR和ImageNet上的实验表明，在干净数据识别方面具有竞争力，并在不同输入扰动下显著提高了鲁棒性。
cs.CV / 96 / 2609.19138
In-Context Robot Learning with VLM Agents
基于上下文的机器人学习与视觉语言模型代理
Abstract
Enabling robots to adapt to unfamiliar environments as readily as humans remains a moonshot goal of embodied AI. No finite collection of demonstrations can cover every task and situation a robot will encounter, making the ability to learn from context at deployment essential for generalization. Such in-context learning (ICL), however, remains largely beyond the reach of existing robotic policies. The broad agentic capabilities of commercial vision-language models (VLMs), such as GPT-6 Astra, raise a compelling question: can these models learn from demonstrations, examples, and interaction feedback, then translate that information into executable and verifiable robot behavior from a new initial state without gradient updates or persistent changes to task-specific parameters? We introduce GPT-Policy, a general-agent framework for in-context robot learning. GPT-Policy integrates a context compiler that preserves task-relevant visual transitions, a VLM that proposes robot-tool actions, and a constrained controller that verifies and executes each action and reports its outcome. We evaluate its reliability and limitations through task success and efficiency metrics, matched comparisons across models, and controlled context ablations. In real-robot trials, human video demonstrations improve task completion even without robot action labels, while aligned action references yield further gains on contact-sensitive tasks. These findings position GPT-Policy as a step toward robot adaptation through in-context learning, providing an empirical foundation for translating the general-purpose capabilities of VLMs into physical behavior and clarifying the challenges that must be overcome for reliable deployment.
Chinese Translation
使机器人能够像人类一样轻松适应不熟悉的环境仍然是具身人工智能的一个宏伟目标。有限的演示集合无法覆盖机器人将遇到的每一个任务和情况，因此在部署时从上下文中学习的能力对于泛化至关重要。然而，这种基于上下文的学习（In-Context Learning, ICL）在现有的机器人策略中仍然大多无法实现。商业视觉语言模型（Vision-Language Models, VLMs）如 GPT-6 Astra 的广泛代理能力提出了一个引人注目的问题：这些模型能否从演示、示例和交互反馈中学习，然后将这些信息转化为从新的初始状态出发的可执行和可验证的机器人行为，而无需梯度更新或对特定任务参数的持续更改？我们提出了 GPT-Policy，这是一个用于基于上下文的机器人学习的通用代理框架。GPT-Policy 集成了一个上下文编译器，保留与任务相关的视觉过渡，一个提出机器人工具动作的 VLM，以及一个验证和执行每个动作并报告其结果的受限控制器。我们通过任务成功率和效率指标、跨模型的匹配比较以及受控的上下文消融评估其可靠性和局限性。在真实机器人试验中，人类视频演示即使在没有机器人动作标签的情况下也能提高任务完成率，而对齐的动作参考在接触敏感任务上则带来了进一步的提升。这些发现将 GPT-Policy 定位为通过基于上下文的学习实现机器人适应的一步，为将 VLM 的通用能力转化为物理行为提供了实证基础，并阐明了可靠部署所需克服的挑战。
cs.CV / 97 / 2609.19142
PointZero: 3D Point Track Completion for Learning Transferable 3D Dynamics
PointZero：用于学习可转移3D动态的3D点轨迹补全
Abstract
World models endow perceptual systems with the ability to predict how scenes evolve under interaction. They are most beneficial when trained on diverse volumes of data, to instill a rich prior into downstream applications. Existing methods typically require robot action labels to learn action-conditioned 3D dynamics, which excludes web video data from the training pool. We study 3D point track completion as a pre-training objective for learning transferable 3D dynamics without robot data. Given a single RGB-D observation and sparse partial 3D trajectories (tracks), we predict future 3D tracks of all observed points. We show this objective produces a rich 3D dynamics prior, without requiring robot action labels. We contribute a diverse dataset of 2.9 million synthetic frames spanning deformable, articulated, and rigid objects, and use it to train PointZero. We show that a flexible and expressive transformer, PointZero, outperforms prior methods on the same data. We demonstrate the utility of our pre-training objective by post-training PointZero for two downstream applications: (1) action-conditioned 3D dynamics prediction and (2) imitation learning. When fine-tuned to condition on end-effector pose, PointZero outperforms the baselines on the recent PGND 3D dynamics benchmark. When fine-tuned to predict robot actions and 3D tracks, PointZero outperforms or matches the baselines on 6/7 simulated and real-world robot manipulation tasks. We furthermore evaluate training PointZero from scratch to isolate the benefits of our proposed architecture from those of our proposed pre-training objective and dataset. We release the dataset, checkpoints, and full training recipe.
Chinese Translation
世界模型赋予感知系统预测场景在交互下如何演变的能力。当在多样化的数据集上训练时，它们对下游应用最为有益，以便为其注入丰富的先验知识。现有方法通常需要机器人动作标签来学习基于动作的3D动态，这使得网络视频数据无法纳入训练池。我们研究了3D点轨迹补全作为一种预训练目标，以便在没有机器人数据的情况下学习可转移的3D动态。给定单个RGB-D观测和稀疏部分3D轨迹（轨迹），我们预测所有观察点的未来3D轨迹。我们展示了这一目标能够产生丰富的3D动态先验，而无需机器人动作标签。我们贡献了一个包含290万合成帧的多样化数据集，涵盖可变形、关节和刚性物体，并用其训练PointZero。我们表明，灵活且具有表现力的变换器PointZero在相同数据集上优于之前的方法。我们通过对PointZero进行后训练，展示了我们预训练目标的实用性，应用于两个下游任务：（1）基于动作的3D动态预测和（2）模仿学习。当微调以适应末端执行器姿态时，PointZero在最近的PGND 3D动态基准测试中超越了基线。当微调以预测机器人动作和3D轨迹时，PointZero在6/7个模拟和真实世界的机器人操作任务中超越或匹配基线。此外，我们评估了从零开始训练PointZero，以隔离我们提出的架构与我们提出的预训练目标和数据集的好处。我们发布了数据集、检查点和完整的训练方案。
cs.CV / 98 / 2609.19143
PANORAMA: Panoptic Grounded Captioning via Mask Proposal Selection
PANORAMA：通过掩码提议选择实现全景化的基础图像描述
Abstract
Intelligent systems that act in the world require image understanding that is both comprehensive and spatially grounded. Current vision-language models (VLMs) can generate fluent and detailed image captions, but reliably associating them with image pixels remains challenging. Existing methods that combine dense captioning with pixel-level grounding often produce either incomplete descriptions or inaccurate segmentation masks. We study this problem through panoptic grounded captioning, a task that requires a VLM to describe both foreground objects and background regions while grounding each referring phrase with pixel-level masks. We make three contributions. First, we introduce PanoCaps, a human-annotated benchmark constructed from panoptic segmentation datasets. It provides dense captions with near-complete pixel coverage and image-text alignments at the entity level, supporting both training and evaluation. We further propose a phrase-mask matching protocol and a generalized Panoptic Quality (gPQ) metric that jointly evaluates textual and mask agreement. Second, we formulate phrase grounding as selection from a phrase-conditioned pool of mask proposals and introduce PANORAMA, a VLM that conditions a pretrained segmenter on contextualized phrase representations to obtain candidate masks and learns to select those corresponding to each phrase. Training this interface jointly with caption generation enables PANORAMA to produce high-quality masks while allowing each phrase to refer to a single region or multiple instances. Third, PANORAMA achieves the best overall grounding on PanoCaps and matches or exceeds specialized models across several pixel-level grounding tasks. Experiments show that our method produces precise entity-level segmentations while maintaining detailed, mask-consistent captions. Code, data and models are available at https://www.di.ens.fr/willow/research/panorama/.
Chinese Translation
智能系统在世界中行动需要全面且空间上扎根的图像理解。目前的视觉-语言模型（VLMs）能够生成流畅且详细的图像描述，但将其可靠地与图像像素关联仍然具有挑战性。现有的方法将密集描述与像素级的基础结合，往往会产生不完整的描述或不准确的分割掩码。我们通过全景化基础图像描述这一任务来研究这个问题，该任务要求VLM描述前景物体和背景区域，同时用像素级掩码为每个指称短语提供基础。我们做出了三项贡献。首先，我们引入了PanoCaps，这是一个基于全景分割数据集构建的人类标注基准。它提供了几乎完整像素覆盖的密集描述和实体级的图像-文本对齐，支持训练和评估。我们进一步提出了短语-掩码匹配协议和一种广义全景质量（gPQ）指标，用于联合评估文本和掩码的一致性。其次，我们将短语基础形式化为从短语条件的掩码提议池中选择，并引入了PANORAMA，一个将预训练分割器与上下文化的短语表示相结合的VLM，以获得候选掩码，并学习选择与每个短语相对应的掩码。与描述生成共同训练这个接口，使得PANORAMA能够生成高质量的掩码，同时允许每个短语指称单个区域或多个实例。第三，PANORAMA在PanoCaps上实现了最佳的整体基础，并在多个像素级基础任务中与专门模型相匹配或超越。实验表明，我们的方法在保持详细且与掩码一致的描述的同时，能够生成精确的实体级分割。代码、数据和模型可在 https://www.di.ens.fr/willow/research/panorama/ 获取。
cs.AI / 1 / 2609.17631
Making AI-Assisted Claims Independently Challengeable: Publication Authority and a Protocol for Falsifiable Publication Records
使人工智能辅助的声明能够独立挑战：出版权威与可证伪出版记录的协议
Abstract
AI-assisted claims can appear authoritative when evidence, analysis, human authorization, presentation, and correction history refer to different states. Provenance, attestation, and transparency expose history but alone do not specify the publication transition examined here. We develop Publication Authority as an exact-state, non-transferable, single-use publication capability and instantiate it in PAC-2026 (Publication-Accountability Calculus), a machine-readable AIJIM Protocol candidate. We evaluate its fourth bounded semantic freeze (SF-4), a fixed-profile specification designed for replaceable bindings. Six obligations govern evidence, runs and artifacts, measurement disclosure, authorization, surface correspondence, and lifecycle continuity. Each yields a target-bound witness, localized counterexample, or localized unverifiability; none can compensate for another. Only a fresh, complete all-pass record derives the permit consumed by one atomic publication transition. We use identity vectors, adversarial cases, finite models, and historical implementations. Ten models explored 110,764 safe reachable states; 76 unsafe configurations produced the expected violation or observer countermodel. A reader surface passing its correspondence check cannot authorize publication unless the accepted record admits that surface. SF-4 separates evidence horizon from verification time and rejects an authentic but causally invalid authorization. A historical predecessor path reproduced 17 frozen authorization-successor outcomes. A later in-house, instance-blind test of known case classes matched all 183 scored expectations; same-host package execution reproduced its 240 archived observations. Results support internal coherence, bounded safety, fault sensitivity, and limited constructibility, but not factual truth, general refinement, blind interoperability, field efficacy, or standards status.
Chinese Translation
当证据、分析、人类授权、呈现和纠正历史指向不同状态时，人工智能辅助的声明可能显得权威。来源、证明和透明度揭示了历史，但单独并不能具体说明这里所考察的出版转变。我们开发了出版权威（Publication Authority），作为一种确切状态、不可转让、一次性使用的出版能力，并在PAC-2026（出版-问责演算，Publication-Accountability Calculus）中实例化，这是一个机器可读的AIJIM协议候选。我们评估了其第四个有界语义冻结（SF-4），这是一个为可替换绑定设计的固定配置规范。六项义务管理证据、运行和工件、测量披露、授权、表面对应和生命周期连续性。每一项都产生一个目标绑定证人、本地反例或本地不可验证性；没有任何一项能够补偿另一项。只有一个新的、完整的全通行记录才能派生出一个原子出版转变所消耗的许可。我们使用身份向量、对抗案例、有限模型和历史实现。十个模型探索了110,764个安全可达状态；76个不安全配置产生了预期的违反或观察者反模型。通过其对应检查的读者表面不能授权出版，除非被接受的记录承认该表面。SF-4将证据视野与验证时间分离，并拒绝一个真实但因果无效的授权。一个历史前身路径再现了17个冻结的授权成功后果。后来的内部、实例盲测试已知案例类别匹配了所有183个得分期望；同主机包执行再现了其240个存档观察。结果支持内部一致性、有界安全性、故障敏感性和有限构造性，但不支持事实真相、一般精炼、盲互操作性、领域有效性或标准状态。
cs.AI / 2 / 2609.17632
EvolveTrade: Experience-Driven Policy Refinement for Self-Evolving LLM Trading Agents
EvolveTrade：基于经验驱动的自我演化LLM交易代理政策优化
Abstract
Large language model (LLM) trading agents can combine market data, news, and executable analysis, but their behavior is often controlled by static hand-written tool-use policies that are fixed before deployment. This limits their ability to adapt how they gather evidence, invoke tools, verify signals, and manage risk under changing market regimes. We introduce EvolveTrade, a self-evolving framework that treats the system prompt of a tool-using trading agent as a text-parameterized policy. After each update interval, a Policy Agent revises this policy using accumulated decision traces and realized portfolio feedback, while keeping the backbone LLM fixed. The updated policy is then used for the next batch of trading decisions, enabling the agent to refine its information-acquisition and portfolio-construction procedure over time. Experiments across multiple market regimes and two LLM backbones show that EvolveTrade often improves Sharpe Ratio and Cumulative Return over fixed-policy LLM baselines, achieving the improved SR and CR in most evaluated settings. Behavioral analyses further show that self-evolved policies increase code-mediated analysis and activate regime-relevant computations; case-level policy-to-return attributions trace how policy-induced allocation changes contribute to realized return differences. These results suggest that adapting the reusable procedure governing tool use is a key direction for building more robust LLM trading agents.
Chinese Translation
大型语言模型（LLM）交易代理能够结合市场数据、新闻和可执行分析，但其行为往往受到在部署前固定的静态手写工具使用政策的控制。这限制了它们在变化的市场环境下调整证据收集、调用工具、验证信号和管理风险的能力。我们提出了EvolveTrade，一种自我演化的框架，将工具使用交易代理的系统提示视为文本参数化政策。在每个更新间隔后，政策代理使用累积的决策轨迹和实际投资组合反馈修订该政策，同时保持主干LLM不变。更新后的政策随后用于下一批交易决策，使代理能够随着时间的推移优化其信息获取和投资组合构建过程。在多个市场环境和两个LLM主干的实验中，EvolveTrade通常在夏普比率和累计收益率上优于固定政策LLM基线，在大多数评估设置中实现了改善的SR和CR。行为分析进一步表明，自我演化的政策增加了代码介导的分析，并激活了与市场环境相关的计算；案例级政策与收益的归因追踪了政策引发的配置变化如何贡献于实现的收益差异。这些结果表明，适应管理工具使用的可重用程序是构建更稳健的LLM交易代理的关键方向。
cs.AI / 3 / 2609.17633
One Color Preprocessing Improves DSATUR
一种颜色预处理方法提升了 DSATUR 算法的性能
Abstract
The Graph Coloring Problem (GCP) is NP-hard and DSATUR stands as one of the fastest heuristics for it despite producing colorings that typically use more colors than state-of-the-art coloring algorithms. We propose SSLD (Semidefinite Spectral Learning with DSATUR), which improves DSATUR by preprocessing a first good color class before letting DSATUR complete coloring the rest of the given graph. We obtain this color class from a Semidefinite Programming (SDP), similar to an SDP used to compute the Lov\'asz theta number. To the best of our knowledge, SSLD is the first approach to improve DSATUR by preprocessing through fixed color classes. We evaluate SSLD against DSATUR and against a naive 1-color-class preprocessing algorithm on DIMACS instances, random graphs (Erd\H{o}s--R\'enyi, Watts-Strogatz, Barab\'asi--Albert), Frequency Assignment and Job Shop Scheduling instances. SSLD matches or beats DSATUR in almost every case across over 1600 benchmark instances, and out performs the naive GISD baseline, allows us to confirm the value brought by the SDP-guided choice of the first color class. This quality comes at a runtime cost of roughly 195 times slower that DSATUR, but demonstrating that SDP-guided preprocessing of a first color class is a direction for future improvements.
Chinese Translation
图着色问题（Graph Coloring Problem，GCP）是 NP-难题，而 DSATUR 算法是其最快的启发式算法之一，尽管其生成的着色通常使用的颜色数量超过了最先进的着色算法。我们提出了 SSLD（基于 DSATUR 的半正定谱学习），该方法通过在 DSATUR 完成给定图的其余部分着色之前，预处理一个良好的颜色类来改进 DSATUR。我们通过半正定规划（Semidefinite Programming，SDP）获得这个颜色类，类似于用于计算 Lovász θ 数的 SDP。根据我们所知，SSLD 是第一个通过固定颜色类预处理来改进 DSATUR 的方法。我们在 DIMACS 实例、随机图（Erdős–Rényi、Watts-Strogatz、Barabási–Albert）、频率分配和作业车间调度实例上评估了 SSLD 与 DSATUR 以及一个简单的 1-颜色类预处理算法的表现。SSLD 在超过 1600 个基准实例中几乎在每种情况下都与 DSATUR 相匹配或优于其表现，并且超越了简单的 GISD 基线，证实了 SDP 指导的第一个颜色类选择所带来的价值。这种质量的提升代价是运行时间大约比 DSATUR 慢 195 倍，但证明了基于 SDP 的第一个颜色类预处理是未来改进的一个方向。
cs.AI / 4 / 2609.17635
Physics-Constrained Digital Twins for Sensor Integrity in Urban Pedestrian Flow: Detecting Stealthy False Data Injection with Conformal Guarantees
物理约束的数字双胞胎用于城市行人流的传感器完整性：使用保形保证检测隐蔽的虚假数据注入
Abstract
City pedestrian counting systems now feed economic indicators, planning decisions and safety operations, yet the twins built on top of them treat the incoming stream as ground truth. We study what happens when it is not. We formalise stealthy false data injection for city-scale pedestrian sensing, where the map from latent flow to observation is far more rank deficient than in the power and water networks for which stealth has been characterised. Our twin estimates directed flows on the pedestrian street graph, assimilates counts through a learned graph-localised gain, and is trained against a flow conservation residual that couples metered and unmetered segments. Detection combines the innovation with that residual, and the alarm threshold is set by adaptive conformal calibration rather than by hand. To measure what the physics buys, we define the attack margin, the relative reduction in worst-case corruption of the estimated flow field, achieved against a white-box adversary that optimises directly through the twin. On six years of Melbourne data the margin reaches 0.54 against a single compromised device and falls to 0.19 when a third of the fleet is compromised, on a network where only 1.18 per cent of walkable segments are metered. Replacing the street graph by a distance graph collapses it to 0.09, which shows that the gain comes from the conservation law rather than from locality.
Chinese Translation
城市行人计数系统现在为经济指标、规划决策和安全操作提供数据，但基于这些系统构建的双胞胎将输入流视为真实数据。我们研究当输入流不是实时数据时会发生什么。我们对城市规模的行人感知中的隐蔽虚假数据注入进行了形式化研究，其中潜在流到观测的映射的秩远低于电力和水务网络中已被特征化的隐蔽性。我们的双胞胎估计了行人街道图上的定向流，通过学习的图局部增益来同化计数，并针对耦合计量和非计量段的流量守恒残差进行训练。检测将创新与该残差结合，警报阈值通过自适应保形校准设置，而不是手动设定。为了衡量物理约束的价值，我们定义了攻击边际，即在白盒对手直接通过双胞胎优化时，估计流场的最坏情况腐败的相对减少。在六年的墨尔本数据中，当只有一个设备被攻陷时，边际达到0.54，而当三分之一的设备被攻陷时，边际降至0.19，且在该网络中，只有1.18%的可步行段是计量的。将街道图替换为距离图使其降至0.09，这表明增益来自于守恒法则而非局部性。
cs.AI / 5 / 2609.17637
What You Can't See Is Still What You Learn: A Preregistered Sixty-Society Confirmation That Evidence Masking Drives Compositional Generalization
看不见的仍然是你所学习的：一项预注册的六十个社会确认，证据遮蔽驱动组合泛化
Abstract
Restricting what a module can read may improve what a system learns to compute. We test this in a preregistered confirmation with sixty four-cell systems sharing a frozen language-model backbone and communicating through learned continuous packets. Five conditions vary evidence masking, ownership markers, and replacement of foreign evidence with neutral filler, across six initialization clusters, each with two data orders, on one fresh task world. With markers available in both regimes, masking improved accuracy on held-out two- and three-operation compositions by median paired differences of 0.846 and 0.859; all twelve pairs cleared the required margins, and the full preregistered behavioral criterion passed. The unmarked replication also passed. No globally visible system passed the marker-following check, so the effect of usable role information remains unresolved. The filler condition yielded seven full generalizers, but its decomposition criteria were inconclusive. Packet interventions in all eighteen audited masked systems followed the predicted intermediate-value changes on eligible cases; these finite, success-conditioned audits do not establish mediation. The results confirm a large advantage of the tested masking regime, while leaving its finer attribution and generality open. Protocols, results, and checkpoints are public.
Chinese Translation
限制模块可以读取的内容可能会改善系统学习计算的能力。我们在一项预注册的确认实验中进行了测试，该实验涉及六十个共享冻结语言模型骨干的四单元系统，通过学习的连续数据包进行通信。五个条件变化了证据遮蔽、所有权标记以及用中性填充物替换外部证据，分布在六个初始化集群中，每个集群有两种数据顺序，针对一个全新的任务世界。在两种状态下都有标记可用的情况下，遮蔽在保留的两操作和三操作组合上提高了准确性，配对中位差分别为0.846和0.859；所有十二对都超过了所需的边际，完整的预注册行为标准也通过了。未标记的复制实验同样通过了。没有任何全球可见的系统通过标记跟踪检查，因此可用角色信息的影响仍未解决。填充条件产生了七个完整的泛化者，但其分解标准并不明确。所有十八个经过审计的遮蔽系统中的数据包干预在符合条件的案例中遵循了预测的中间值变化；这些有限的、成功条件的审计并未确立中介关系。结果确认了所测试的遮蔽机制的巨大优势，同时其更细致的归因和普遍性仍待探讨。协议、结果和检查点均为公开。
cs.AI / 6 / 2609.17688
CapMem: A Benchmark for Caption-Based Episodic Memory in Egocentric Video
CapMem：基于字幕的自我中心视频情节记忆基准
Abstract
Wearable assistants require episodic memory over egocentric video, yet current vision-language models face bounded frame budgets, growing visual-token costs, and long-context retrieval failures. Under these practical constraints, we study whether textual captions can serve as reusable episodic memory. We define the Episodic Memory Video Caption QA task and introduce CapMem, a human-annotated benchmark with 75 videos totaling 33.7 hours, and 1,000 multiple-choice questions across 16 scenarios. On long videos (>20 min), full-coverage CaptionQA with 30s and 60s caption windows outperforms direct VideoQA for 10/12 and 8/12 models, respectively. On the same video subset, a matched-frame control across six Qwen models retains mean accuracy gains of 3.22 and 2.55 points, respectively. Our caption-guided retrieve-and-verify harness further improves accuracy by up to 5.3 points. These results support the effectiveness of caption memory for episodic reasoning over long egocentric video.
Chinese Translation
可穿戴助手需要在自我中心视频中具备情节记忆，然而当前的视觉-语言模型面临有限的帧预算、不断增长的视觉标记成本和长上下文检索失败等实际限制。在这些约束下，我们研究文本字幕是否可以作为可重用的情节记忆。我们定义了情节记忆视频字幕问答任务，并引入了CapMem，一个包含75个视频（总计33.7小时）和1,000个多项选择题（涵盖16种场景）的人类标注基准。在长视频（>20分钟）上，使用30秒和60秒字幕窗口的全覆盖CaptionQA分别在10/12和8/12模型中优于直接VideoQA。在相同的视频子集上，六个Qwen模型之间的匹配帧控制保持了平均准确率的提升，分别为3.22和2.55点。我们的字幕引导检索与验证方法进一步提高了准确率，最多可提升5.3点。这些结果支持了字幕记忆在长自我中心视频情节推理中的有效性。
cs.AI / 7 / 2609.17695
GraphEcho: Structural Redundancy and Evidence Provenance in LLM Graph Agents
GraphEcho：大型语言模型图代理中的结构冗余与证据来源
Abstract
A large language model (LLM) agent can follow more graph paths without acquiring more independent evidence. GraphEcho tests whether agents mistake these repeated encounters for additional corroboration. The benchmark varies path counts and evidential origins while holding evidence content fixed, and evaluates both judgments and active exploration. Controlled synthetic experiments reveal model-dependent judgment shifts, but redundant supporting paths increase the share of repeated walks across all evaluated frozen agents. Provenance-aware post-training (PAPT) reduces revisits and improves synthetic accuracy, yet covers fewer distinct sources. On scientific claims, it continues to reduce repetition while accuracy declines. These findings expose a gap between efficient exploration and effective evidence use: an agent can learn to stop repeating itself while overlooking information it needs. GraphEcho provides a controlled way to evaluate both what graph agents conclude and whether their exploration reaches distinct evidential sources.
Chinese Translation
大型语言模型（LLM）代理可以在不获取更多独立证据的情况下，沿着更多的图路径进行操作。GraphEcho 测试代理是否将这些重复的遭遇误认为是额外的证实。该基准测试在保持证据内容不变的情况下，变化路径数量和证据来源，并评估判断和主动探索。受控的合成实验揭示了模型依赖的判断偏移，但冗余的支持路径增加了所有评估的冻结代理中重复行走的比例。基于来源意识的后训练（PAPT）减少了重访次数并提高了合成准确性，但覆盖的独立来源较少。在科学声明中，它继续减少重复，但准确性下降。这些发现揭示了高效探索与有效证据使用之间的差距：代理可以学会停止自我重复，但却忽视了所需的信息。GraphEcho 提供了一种受控的方法来评估图代理的结论以及它们的探索是否达到独立的证据来源。
cs.AI / 8 / 2609.17696
GVD: Governed Versioning and Deduplication for Document Repositories
GVD：文档库的治理版本控制与去重
Abstract
Document repositories evolve continuously. Guidelines and policies are revised, superseded, and re-uploaded, so the same content recurs in different wording and newer versions refine or contradict earlier ones. These inconsistencies belong to the growing collection rather than to any single document, yet existing work treats versioning, duplicate detection, and contradiction detection as isolated pairwise tasks and stops once a pair is labeled. We present GVD (Governed Versioning and Deduplication), a framework that unifies cross-document version linking with rule-level conflict resolution under an auditable update policy. Incoming documents are assigned to version families through bidirectional rule alignment, and their rules are compared against the family memory to identify duplicates, contradictions, asymmetric refinements, and new knowledge, with Counterfactual Span Probing (CSP) resolving related pairs that inference misclassifies as neutral. Relation-specific policies suppress duplicates and escalate only consequential changes for review, retaining version lineage as an audit trail. The pipeline runs fully locally, with no large language model. On 120 enterprise documents processed as 140 ingestions across 59 version families, GVD reaches an F1 of 0.97 for version-family construction and 0.94 for rule-level consistency, with CSP raising rule consistency from 0.90 to 0.94.
Chinese Translation
文档库持续演变。指导方针和政策被修订、取代并重新上传，因此相同的内容以不同的措辞重复出现，而更新版本可能会细化或与早期版本相矛盾。这些不一致性属于不断增长的集合，而不是任何单一文档。然而，现有研究将版本控制、重复检测和矛盾检测视为孤立的成对任务，并在标记一对后停止。我们提出了GVD（Governed Versioning and Deduplication），这是一个将跨文档版本链接与规则级冲突解决统一在可审计更新政策下的框架。通过双向规则对齐，将新文档分配到版本家族，并将其规则与家族记忆进行比较，以识别重复、矛盾、不对称的细化和新知识，使用反事实跨度探测（Counterfactual Span Probing, CSP）解决推理错误分类为中性的相关对。特定关系的政策抑制重复，仅将重要变更升级以供审查，同时保留版本沿袭作为审计轨迹。该流程完全在本地运行，无需大型语言模型。在处理的120份企业文档中，跨59个版本家族进行140次摄取，GVD在版本家族构建上达到0.97的F1值，在规则级一致性上达到0.94，CSP将规则一致性从0.90提高到0.94。
cs.AI / 9 / 2609.17699
NeMo Data Designer: An Extensible Framework for Multimodal Synthetic Data Generation
NeMo 数据设计器：一个可扩展的多模态合成数据生成框架
Abstract
We present NeMo Data Designer (NDD), an open-source, general-purpose framework for multi-modal synthetic data generation (SDG). Designed to be intuitive to use, NDD provides a declarative configuration format in which human and/or agent users define each dataset column, with column types spanning text, code, structured outputs, images, embeddings, and statistical samplers that are explicitly configured to steer dataset diversity. Additional column types and functionality can be introduced using the framework's flexible plugin system. NDD's configuration is an inspectable artifact, supporting workflow sharing and reproducibility. SDG is an inherently iterative process. NDD therefore builds a preview-and-revision loop into its core workflow, allowing users to generate and inspect a small number of records, refine the specification, and rerun generation at full scale. At runtime, NDD resolves dependencies, schedules calls to user-provided model endpoints, and retries failed requests. We describe NDD's architecture and programming model and present case studies spanning structured, agentic, multimodal, and domain-specialized tasks, including datasets used in Nemotron model development and in production enterprise deployments.
Chinese Translation
我们提出了 NeMo 数据设计器（NDD），这是一个开源的通用多模态合成数据生成（SDG）框架。NDD 旨在易于使用，提供了一种声明式配置格式，用户（人类和/或代理）可以在其中定义每个数据集列，列类型包括文本、代码、结构化输出、图像、嵌入以及显式配置以引导数据集多样性的统计采样器。通过框架灵活的插件系统，可以引入额外的列类型和功能。NDD 的配置是一个可检查的工件，支持工作流共享和可重复性。SDG 本质上是一个迭代过程。因此，NDD 在其核心工作流中构建了预览和修订循环，允许用户生成和检查少量记录，完善规范，并在全规模下重新运行生成。在运行时，NDD 解析依赖关系，调度对用户提供的模型端点的调用，并重试失败的请求。我们描述了 NDD 的架构和编程模型，并展示了涵盖结构化、代理、多模态和领域专业化任务的案例研究，包括在 Nemotron 模型开发和生产企业部署中使用的数据集。
cs.AI / 10 / 2609.17731
A Systematic Evaluation of the COTQ Provincial Land Cover Product: Structural Consistency, Spectral Separability, and Relative Positioning Against ESA, ESRI, and Google Products
COTQ省级土地覆盖产品的系统评估：结构一致性、光谱可分离性及其相对于ESA、ESRI和Google产品的相对定位
Abstract
High-resolution land use and land cover (LULC) products derived from Sentinel-2 imagery are widely used for environmental monitoring and land management, yet their performance can vary across regions with complex ecological gradients and heterogeneous surface conditions. In Quebec, these limitations motivated the development of a provincial 10-m land-cover product, the COTQ, designed to support annual monitoring of land occupation and soil artificialisation. This study presents a systematic evaluation of the COTQ product relative to three global 10-m LULC datasets: ESA WorldCover, ESRI LandCover, and Google DynamicWorld. This paper does not introduce a new mapping methodology but focuses on analysing the behaviour and consistency of the COTQ using complementary evaluation approaches. All products are harmonized under a common legend and compared using structural indicators (object-size distributions, shape complexity, Adjusted Rand Index, and Intersection over Union), spectral separability metrics derived from Sentinel-2 reflectance data, and a targeted photo-interpretation of disagreement areas. The analysis is conducted over eight Sentinel-2 tiles selected to represent the main bioclimatic domains of Qu\'ebec, from temperate and boreal forests to northern tundra environments. The results show that the COTQ exhibits structural and spectral characteristics most similar to ESA WorldCover among the reference global products, while revealing systematic differences linked to class definitions and thematic priorities, particularly for urban areas, wetlands, and rocky or cryptogamic surfaces. This multi-criteria evaluation provides an objective characterization of the COTQ product and clarifies its relative positioning with respect to existing global land-cover datasets for operational land monitoring in Qu\'ebec.
Chinese Translation
基于Sentinel-2影像的高分辨率土地利用和土地覆盖（LULC）产品广泛应用于环境监测和土地管理，但其在具有复杂生态梯度和异质表面条件的区域中的表现可能有所不同。在魁北克，这些局限性促使开发了一种10米的省级土地覆盖产品COTQ，旨在支持土地占用和土壤人工化的年度监测。本研究对COTQ产品相对于三个全球10米LULC数据集（ESA WorldCover、ESRI LandCover和Google DynamicWorld）进行了系统评估。本文并未引入新的制图方法，而是专注于使用互补的评估方法分析COTQ的行为和一致性。所有产品在一个共同的图例下进行协调，并使用结构指标（对象大小分布、形状复杂性、调整兰德指数和交并比）、基于Sentinel-2反射率数据的光谱可分离性指标，以及针对不一致区域的目标照片解读进行比较。分析在八个选定的Sentinel-2图块上进行，这些图块代表了魁北克的主要生物气候域，从温带和北方森林到北极苔原环境。结果表明，COTQ在结构和光谱特征上与参考全球产品中的ESA WorldCover最为相似，同时揭示了与类别定义和主题优先级相关的系统性差异，尤其是在城市地区、湿地以及岩石或地衣表面。该多标准评估为COTQ产品提供了客观特征描述，并澄清了其相对于现有全球土地覆盖数据集在魁北克的操作性土地监测中的相对定位。
cs.AI / 11 / 2609.17757
Imitation Learning for Autonomous Driving in CARLA
在CARLA中用于自主驾驶的模仿学习
Abstract
Behavioral cloning trains a policy offline on expert demonstrations, but deployment is closed loop: each action affects the observations the policy receives next. We study how much closed-loop driving competence a compact multimodal policy can acquire from offline demonstrations in the CARLA simulator. The policy uses five-frame histories of RGB images, LiDAR, vehicle telemetry, and lane waypoints to predict throttle, brake, and steering at 20 Hz. Demonstrations were collected in three stages, ending with a systematic route-generation procedure that enumerates spawn points and feasible maneuvers and verifies completed autopilot routes. The released 1.36 million parameter policy was trained on 236,882 windows, representing about 3.3 hours of driving from 448 captures. The resulting policy drives autonomously for hours on training and held-out routes. In our runs, it did so without collisions and also transferred qualitatively to an unseen CARLA town with different road geometry. We also observed recovery from large trajectory deviations, although we do not claim systematic recovery without controlled evaluation. We report offline metrics and distinguish measured results from qualitative closed-loop observations. We release the code, trained checkpoint, ONNX model, data sample, and an evidence audit for the reported claims.
Chinese Translation
行为克隆在专家演示上离线训练策略，但部署是闭环的：每个动作都会影响策略接下来接收到的观测。我们研究了紧凑的多模态策略能够从CARLA模拟器中的离线演示中获得多少闭环驾驶能力。该策略使用五帧历史的RGB图像、激光雷达（LiDAR）、车辆遥测和车道航点，以20 Hz的频率预测油门、刹车和转向。演示分三个阶段收集，最后通过系统的路线生成程序枚举生成点和可行的机动，并验证完成的自动驾驶路线。发布的136万参数策略是在236,882个窗口上训练的，代表约3.3小时的驾驶，来自448次捕获。最终的策略在训练和保留的路线中能够自主驾驶数小时。在我们的实验中，它在没有碰撞的情况下完成了驾驶，并且在不同道路几何的未见CARLA城镇中也实现了定性迁移。我们还观察到从大轨迹偏差中恢复，尽管我们不声称在没有控制评估的情况下实现系统恢复。我们报告了离线指标，并区分了测量结果与定性闭环观察。我们发布了代码、训练检查点、ONNX模型、数据样本和报告声明的证据审计。
cs.AI / 12 / 2609.17775
SAGE: Governed Artifact Generation from Enterprise Guidelines
SAGE：基于企业指南的受控工件生成
Abstract
Enterprise guideline documents mix narrative text, complex tables, and embedded images, and converting them into structured work artifacts still takes two to three days of manual effort each. Current language and vision-language models extract from such documents but offer no governed workflow beyond extraction: no validation, no consistency checking, no traceable artifact generation. We introduce SAGE, a governed multi-stage LLM pipeline organized around a shared versioned rule store with stable identifiers, schema-validated inter-stage contracts, and end-to-end provenance tracking. Extracted rules undergo deterministic structural validation and LLM-based semantic scoring, then a consistency module that removes duplicates, flags contradictions, and surfaces specification gaps; only uncertain or flagged items reach reviewers, while high-confidence outputs are auto-approved. On 120 documents, SAGE cuts turnaround from days to 20-100 minutes, achieving a 96% document-level success rate with 3.2% hallucination, extracting 3,896 rules and producing 812 artifacts ready for human review; without governance, hallucination rises to 15.7%.
Chinese Translation
企业指南文件混合了叙述文本、复杂表格和嵌入图像，将其转换为结构化工作工件仍然需要两到三天的手动工作。当前的语言和视觉-语言模型可以从这些文件中提取信息，但在提取之外并未提供受控的工作流程：没有验证、没有一致性检查、没有可追溯的工件生成。我们提出了SAGE，一个围绕共享版本规则库组织的受控多阶段大型语言模型（LLM）管道，具有稳定的标识符、模式验证的阶段间契约和端到端的来源追踪。提取的规则经过确定性的结构验证和基于LLM的语义评分，然后通过一致性模块去除重复项、标记矛盾并揭示规范缺口；只有不确定或被标记的项目会送交审阅，而高置信度的输出则会自动批准。在120份文件上，SAGE将周转时间从几天缩短到20-100分钟，实现了96%的文档级成功率，幻觉率为3.2%，提取了3,896条规则并生成了812个待人类审阅的工件；而在没有治理的情况下，幻觉率上升至15.7%。
cs.AI / 13 / 2609.17786
FairCompressAgent: An Agentic Framework for Fairness-Aware Model Compression for FPGA Deployment
FairCompressAgent：一种面向公平性的模型压缩代理框架，用于FPGA部署
Abstract
Fairness-aware model compression requires selecting methods and configurations that balance accuracy, fairness, and deployment cost. These decisions become more difficult when compression methods are composed or the user's requirements change. In this paper, we propose FairCompressAgent (FCA), an agentic framework that integrates fairness-aware pruning, incremental quantization, and sparse low-rank factorization through a common operator interface. A language-model planner uses model profiles and measured outcomes to select compression configurations, while an execution layer performs compression, fine-tuning, evaluation, and constraint-based selection. FCA also supports requirement updates and reports the remaining violation when a request cannot be satisfied. Experiments on Fitzpatrick-17k with VGG-11 compare four search methods over 40 measured configurations. Under the accuracy-constrained request, FCA selects a compressed model with 59.54% less inference tensor storage, while validation average precision increases from 0.5141 to 0.5233 and equalized opportunity (EOpp) decreases from 0.2251 to 0.2168. It reaches the same final selection as one-shot planning with 7.33 versus 12 candidate evaluations on average, under their respective stopping policies. Repeated fine-tuning, held-out testing, and online requirement updates characterize the stability and interactive use of this compression workflow. The results demonstrate how measured feedback and explicit constraints support the selection and interactive refinement of fairness-aware compression configurations.
Chinese Translation
面向公平性的模型压缩需要选择在准确性、公平性和部署成本之间平衡的方法和配置。当压缩方法组合或用户需求变化时，这些决策变得更加困难。本文提出了FairCompressAgent (FCA)，一个集成了面向公平的剪枝、增量量化和稀疏低秩分解的代理框架，通过一个通用操作接口进行交互。语言模型规划器利用模型配置文件和测量结果选择压缩配置，而执行层则执行压缩、微调、评估和基于约束的选择。FCA还支持需求更新，并在请求无法满足时报告剩余的违规情况。在Fitzpatrick-17k数据集上对VGG-11进行的实验比较了40个测量配置下的四种搜索方法。在准确性受限的请求下，FCA选择的压缩模型减少了59.54%的推理张量存储，同时验证平均精度从0.5141提高到0.5233，平等机会（Equalized Opportunity, EOpp）从0.2251降低到0.2168。在各自的停止策略下，FCA的最终选择与一次性规划相同，平均候选评估次数为7.33对比12次。重复的微调、保留测试和在线需求更新特征化了该压缩工作流的稳定性和交互使用。结果表明，测量反馈和明确约束如何支持面向公平的压缩配置的选择和交互优化。
cs.AI / 14 / 2609.17804
A Four-Stage Decomposition of Word-Problem Solving and Mechanistic Fragility in LLM Math Reasoning
大语言模型数学推理中的四阶段问题解决分解与机制脆弱性
Abstract
Large language models solve grade-school math word problems with high accuracy, yet a single irrelevant clause inserted into the problem can collapse it. We reconcile these observations with a mechanistic account. We show that the model's internal computation decomposes into a four-stage sequential pipeline, Schema Abstraction, Operation Planning, Operand Binding, and Computation, each stage producing a distinct intermediate representation in an identifiable band of layers. Using the same scaffold to diagnose distractor-induced failure, we localize the corruption to a single stage, Operation Planning, implemented by a set of attention heads whose causal role we validate bidirectionally. In short, we provide a mechanistic interpretation of math word problem reasoning in LLMs, and their failure when distracted.
Chinese Translation
大型语言模型以高准确率解决小学数学文字问题，但在问题中插入一个无关的从句可能会导致其崩溃。我们通过机制解释来调和这些观察结果。我们展示了模型的内部计算分解为一个四阶段的顺序流程，分别是模式抽象（Schema Abstraction）、操作规划（Operation Planning）、操作数绑定（Operand Binding）和计算（Computation），每个阶段在可识别的层次带中产生一个独特的中间表示。利用相同的框架来诊断干扰引起的失败，我们将损坏定位到单一阶段，即操作规划（Operation Planning），该阶段由一组注意力头实现，我们通过双向验证其因果作用。简而言之，我们提供了对大语言模型中数学文字问题推理的机制解释，以及在受到干扰时的失败原因。
cs.AI / 15 / 2609.17847
Learning Heterogeneous Preferences
学习异质偏好
Abstract
Learning from human feedback has become a central paradigm for training modern AI systems, where models of human utility are used as reward models in policy learning. Existing methods typically assume a \emph{universal utility} function shared across a population and treat disagreement between annotators as stochastic variation. While suitable for objective tasks, this assumption breaks down in subjective domains where preferences vary systematically across individuals. We study the problem of subjective preference learning, in which observed choices arise from heterogeneous but internally consistent utility functions. Drawing upon rational choice theory, RCT \parencite{tversky1981framing}, we introduce \emph{individuated utility} functions conditioned on both the individual and their decision context, and propose a novel multi-stage architecture for estimating them from multi-modal data. We evaluate our framework on a newly collected dataset of more than $575{,}000$ pairwise aesthetic judgments from $2{,}398$ participants comparing automotive wheel designs. Our experiments show that individuated utility models substantially outperform universal utility models including foundation model baselines. Our results demonstrate that disagreement reflects meaningful preference heterogeneity rather than annotation noise. More broadly, our findings highlight the importance of collecting annotator attributes and learning individuated utility functions, enabling reward models that explicitly account for whose preferences they represent and faithfully capture human decision diversity.
Chinese Translation
从人类反馈中学习已成为训练现代人工智能系统的核心范式，其中人类效用模型被用作政策学习中的奖励模型。现有方法通常假设在一个群体中共享一个 extit{通用效用}函数，并将标注者之间的分歧视为随机变异。虽然这一假设适用于客观任务，但在个体偏好系统性差异的主观领域中，这一假设失效。我们研究主观偏好学习的问题，其中观察到的选择源于异质但内部一致的效用函数。基于理性选择理论（Rational Choice Theory, RCT） extcite{tversky1981framing}，我们引入了 extit{个体化效用}函数，该函数以个体及其决策背景为条件，并提出了一种新颖的多阶段架构，以从多模态数据中估计这些效用函数。我们在一个新收集的数据集上评估了我们的框架，该数据集包含来自$2{,}398$名参与者对汽车轮毂设计的超过$575{,}000$对美学判断。我们的实验表明，个体化效用模型显著优于包括基础模型基线在内的通用效用模型。我们的结果表明，分歧反映了有意义的偏好异质性，而非标注噪声。更广泛地说，我们的发现强调了收集标注者属性和学习个体化效用函数的重要性，从而使奖励模型能够明确考虑其代表的偏好，并真实捕捉人类决策的多样性。
cs.AI / 16 / 2609.17855
SNOMED CT Concept Recommendation from Masked Clinical Context
基于掩码临床背景的SNOMED CT概念推荐
Abstract
Standardizing clinical language to SNOMED CT supports interoperability, analytics, and reusable phenotyping, but concept recommendation remains difficult when relevant concepts are rare or absent from training data. We present a masked-concept recommendation benchmark using the SNOMED CT Entity Linking Challenge v1.2.1 data derived from MIMIC-IV-Note. The dataset contains 75,491 annotations across 272 discharge summaries, with 204 notes used for training and 68 for historical testing. For each unique note-concept pair, the target mention is masked from a local clinical context and the system ranks SNOMED CT concepts observed during training. We compare a popularity baseline, sparse TF-IDF concept prototypes, dense latent semantic analysis embeddings, sparse-dense fusion, retrieved-note evidence, and a retrieval-augmented hybrid. Sparse TF-IDF performs best, achieving Recall@1 of 14.81%, Recall@10 of 33.43%, MRR of 0.2114, and nDCG@10 of 0.2297. Retrieval augmentation does not improve this baseline, with Recall@10 of 31.99% and MRR of 0.1937. Performance is strongly affected by concept frequency: Recall@10 is 7.74% for concepts appearing in only one or two training notes versus 43.90% for concepts appearing in more than ten. In addition, 9.66% of test note-concept pairs contain concepts unseen during training. These findings show that local lexical context and terminology coverage are major determinants of recommendation quality in low-resource settings and provide a reproducible baseline for future ontology-grounded and biomedical-encoder retrieval systems.
Chinese Translation
将临床语言标准化为SNOMED CT支持互操作性、分析和可重用的表型，但当相关概念在训练数据中稀少或缺失时，概念推荐仍然困难。我们提出了一个基于掩码概念推荐的基准，使用来自MIMIC-IV-Note的SNOMED CT实体链接挑战v1.2.1数据集。该数据集包含272份出院总结中的75,491个注释，其中204份用于训练，68份用于历史测试。对于每个独特的注释-概念对，目标提及从局部临床背景中被掩盖，系统对训练期间观察到的SNOMED CT概念进行排名。我们比较了流行度基线、稀疏TF-IDF概念原型、密集潜在语义分析嵌入、稀疏-密集融合、检索笔记证据和检索增强混合模型。稀疏TF-IDF表现最佳，达到Recall@1为14.81%，Recall@10为33.43%，MRR为0.2114，nDCG@10为0.2297。检索增强未能改善这一基线，Recall@10为31.99%，MRR为0.1937。性能受到概念频率的强烈影响：对于仅在一或两份训练笔记中出现的概念，Recall@10为7.74%，而对于在十份以上出现的概念，Recall@10为43.90%。此外，9.66%的测试注释-概念对包含在训练期间未见过的概念。这些发现表明，局部词汇上下文和术语覆盖范围是低资源环境中推荐质量的主要决定因素，并为未来基于本体的生物医学编码器检索系统提供了可重复的基线。
cs.AI / 17 / 2609.17863
The Inference Engineering Pareto Atlas: Which Optimizations Dominate the Cost, Quality, and Latency Frontier?
推理工程帕累托图谱：哪些优化在成本、质量和延迟边界上占主导地位？
Abstract
LLM inference optimizations report speedups on different models, GPUs, prompts, and quality metrics, making them hard to compare or combine. We build a cost, quality, and latency Pareto atlas to identify the best configurations for different deployment constraints. Since exhaustive testing is impractical, we measure 54 configurations of Qwen2.5-7B-Instruct running on vLLM 0.12 across L4, A100, and H100 GPUs and use these anchors to calibrate a simulator. It reproduces measurements at anchored batch sizes, with cross campaign drift below 1.5 percent. A separate quality evaluation tests FP16, AWQ 4bit, FP8 weights, and FP8 KV cache on 200 GSM8K questions with five examples per prompt. Sparse attention is evaluated only in simulation. On the calibrated grid, 18 of 36 configurations reach the Pareto frontier. Combined methods reach it more often than individual methods, with 9 of 15 combinations versus 9 of 21 single methods. Quality testing changes the winners. AWQ 4bit reduces per token latency to 0.34 times baseline on L4 but loses 5.9 percent of strict GSM8K accuracy, narrowly missing the 95 percent quality floor within sampling uncertainty. Flexible answer extraction matches FP16 accuracy, suggesting the loss comes from formatting rather than arithmetic. FP8 weights retain 99.4 percent of baseline accuracy at 0.61 to 0.65 times baseline latency across all three GPUs and appear in three of four regime winners. A naive FP8 KV cache maintains normal throughput but answers none of the 200 questions correctly, showing why speed alone is insufficient. Under two prompt designs, n gram speculative decoding measures at 0.90 to 0.98 times baseline and adds no benefit on this stack. The best choice depends on the constraint and GPU: H100 wins for tight latency, while A100 wins for throughput and low cost at 0.106 dollars per million tokens.
Chinese Translation
大型语言模型（LLM）推理优化在不同模型、GPU、提示和质量指标上报告了加速，使得它们难以比较或结合。我们构建了一个成本、质量和延迟的帕累托图谱，以识别不同部署约束下的最佳配置。由于全面测试不切实际，我们测量了在 vLLM 0.12 上运行的 54 种 Qwen2.5-7B-Instruct 配置，涵盖 L4、A100 和 H100 GPU，并使用这些基准来校准模拟器。它在锚定的批量大小下重现测量，跨活动的漂移低于 1.5%。另一个质量评估测试了 FP16、AWQ 4bit、FP8 权重和 FP8 KV 缓存在 200 个 GSM8K 问题上的表现，每个提示提供五个示例。稀疏注意力仅在模拟中进行评估。在校准网格上，36 种配置中有 18 种达到了帕累托边界。组合方法比单一方法更频繁地达到该边界，15 种组合中有 9 种达成，而 21 种单一方法中也有 9 种。质量测试改变了胜者。AWQ 4bit 将每个令牌的延迟降低到 L4 上基线的 0.34 倍，但损失了 5.9% 的严格 GSM8K 准确率，略微错过了采样不确定性下的 95% 质量底线。灵活的答案提取与 FP16 准确率相匹配，表明损失来自格式而非算术。FP8 权重在所有三种 GPU 上保持 99.4% 的基线准确率，延迟为基线的 0.61 到 0.65 倍，并出现在四种方案的三种胜者中。一个简单的 FP8 KV 缓存保持正常吞吐量，但没有正确回答 200 个问题中的任何一个，显示出仅靠速度是不够的。在两种提示设计下，n-gram 推测解码的测量为基线的 0.90 到 0.98 倍，并未在此堆栈上增加任何好处。最佳选择取决于约束和 GPU：H100 在严格延迟下获胜，而 A100 在吞吐量和低成本（每百万个令牌 0.106 美元）方面获胜。
cs.AI / 18 / 2609.17865
Do Frontier Models Seek Safety Evidence Before Acting?
前沿模型在行动前是否寻求安全证据？
Abstract
Frontier models are often evaluated on how they respond to safety information once it is already in context. We study an earlier decision point: whether models choose to acquire safety-relevant evidence before acting. We introduce SAFE, a controlled benchmark in which models make deployment decisions with optional evidence that varies in retrieval cost, probability, severity, and presentation. Across GPT-5.5, o3, Claude Opus 4.8, and Claude Sonnet 4.6, we find distinct evidence-acquisition policies: Opus inspects nearly by default, o3 is the most skip-heavy and threshold-sensitive, and GPT-5.5 and Sonnet occupy intermediate regimes. Inspection increases strongly with severity and decreases with retrieval cost, whereas probability has much weaker behavioral influence: increasing the stated likelihood of a problem from 10% to 70% changes inspection by at most 21 percentage points. Despite these differences, Stage 1 rationales are dominated by expected-value reasoning across models. A cost-obligation decomposition further shows that avoidance is driven primarily by retrieval friction and explicit threats to the deployment payoff rather than by the remediation duties created by knowing. Counterfactual interventions reveal a further mismatch between behavior and explanation: evidence framing can strongly change decisions near the inspection boundary while going largely unmentioned, whereas probability is frequently cited despite having little causal influence. These results suggest that deployment-time safety depends not only on how models respond to known risks, but also on whether they acquire the evidence needed to know that acting is safe.
Chinese Translation
前沿模型通常在安全信息已经存在的情况下进行评估。我们研究一个更早的决策点：模型是否选择在行动前获取与安全相关的证据。我们引入了SAFE，这是一个受控基准，模型在其中做出部署决策，并可以选择获取不同检索成本、概率、严重性和呈现方式的证据。通过对GPT-5.5、o3、Claude Opus 4.8和Claude Sonnet 4.6的研究，我们发现了不同的证据获取策略：Opus几乎默认进行检查，o3则是最倾向跳过和对阈值敏感的，而GPT-5.5和Sonnet处于中间状态。检查的频率随着严重性的增加而显著上升，而随着检索成本的增加而下降，概率对行为的影响则相对较弱：将问题的表述可能性从10%提高到70%最多只会改变检查频率21个百分点。尽管存在这些差异，各模型的第一阶段理由主要由期望值推理主导。成本-义务分解进一步表明，回避主要是由检索摩擦和对部署收益的明确威胁驱动，而不是由知晓所产生的补救责任所驱动。反事实干预揭示了行为与解释之间的进一步不匹配：证据框架可以在检查边界附近强烈改变决策，而几乎不被提及，而概率则经常被引用，尽管其因果影响较小。这些结果表明，部署时的安全不仅依赖于模型如何应对已知风险，还依赖于它们是否获取了必要的证据以确认行动是安全的。
cs.AI / 19 / 2609.17885
ERPBench: A State-Grounded Evaluation Paradigm for Computer-Use Agents in Enterprise Software
ERPBench：面向企业软件中计算机使用代理的状态基础评估范式
Abstract
Computer-use agents that operate through screenshots and simulated actions are advancing rapidly, yet their evaluation remains anchored to general desktop and web tasks. Enterprise Resource Planning (ERP) systems run the finance, procurement, inventory, and customer operations of organizations worldwide, and pose distinct challenges for computer-use agents: dense interfaces, coordinated multi-step interactions, and errors that alter persistent business records rather than surfacing on screen. Existing enterprise benchmarks rely on proprietary platforms or on simulated approximations of such software. We introduce ERPBench, a benchmark that evaluates screenshot-only agents on a live and reproducible ERP system and scores each task against ground-truth values in its database. Beyond the benchmark, we present a production-grade harness that gates agent actions behind human approval for safe deployment, which ERPBench runs autonomously. Evaluating six closed and open-source agents, we demonstrate that strong general GUI performance does not transfer to enterprise reliability. Even when an agent reaches the right form and saves it, the stored record is often wrong: some agents save in up to 85% of runs but write the correct value in as few as 3%. We further characterize failure modes specific to enterprise workflows.
Chinese Translation
通过截图和模拟操作进行工作的计算机使用代理正在快速发展，但其评估仍然局限于一般的桌面和网络任务。企业资源规划（ERP）系统管理着全球组织的财务、采购、库存和客户运营，给计算机使用代理带来了独特的挑战：界面密集、协调的多步骤交互，以及改变持久业务记录而不是在屏幕上显现的错误。现有的企业基准依赖于专有平台或对这些软件的模拟近似。我们提出了ERPBench，这是一个在实时和可重现的ERP系统上评估仅基于截图的代理的基准，并根据其数据库中的真实值对每个任务进行评分。除了基准之外，我们还展示了一个生产级的框架，该框架在人工批准的前提下控制代理的操作，以确保安全部署，而ERPBench则可以自主运行。在评估六个封闭源和开源代理时，我们证明了强大的通用图形用户界面（GUI）性能并不转化为企业可靠性。即使代理达到了正确的表单并进行了保存，存储的记录往往也是错误的：一些代理在多达85%的运行中保存，但正确值的写入率却低至3%。我们进一步描述了特定于企业工作流程的失败模式。
cs.AI / 20 / 2609.17890
OBC-Prune: Outcome-Based Calibration for Large Reasoning Model Pruning
OBC-Prune：基于结果的校准用于大型推理模型剪枝
Abstract
Large reasoning models (LRMs) generate long chain-of-thought traces before answering, creating significant inference overhead. Pruning can reduce this cost, but its effectiveness depends on the calibration data used to estimate parameter importance. Recent work calibrates on the model's own rollouts instead of generic dataset, but treats all reasoning tokens uniformly, regardless of whether they contribute to successful reasoning. As a result, pruning protects weights by statistical salience rather than by their contribution to correct reasoning, so weights behind erroneous computation survive as readily as those behind correct computation. These erroneous patterns then get carried into the pruned model, degrading reasoning quality, producing both lower accuracy and longer reasoning traces. We propose Outcome-Based Calibration for Large Reasoning Model Pruning (OBC-Prune) to close this gap. OBC first constructs difficulty-matched pairs of correct and incorrect rollouts from problems the model answers inconsistently. It then estimates the causal importance of each reasoning sentence through intervention-based analysis, quantifying how removing its influence affects subsequent predictions. These causal importance scores are converted into per-token weights that rescale the calibration activations used by one-shot pruning methods (SparseGPT, Wanda, ALPS), without modifying the underlying pruning algorithms. Experiments on DeepSeek-R1-Distill-Qwen 1.5B, 7B, and 14B models at 40\% and 50\% sparsity demonstrate consistent improvements over state-of-the-art calibration baselines across most model sizes and sparsity levels on MATH500, LiveCodeBench, and AIME 2025. These results indicate that preserving causally important reasoning circuits is a substantially more effective pruning objective than uniformly preserving observed activations.
Chinese Translation
大型推理模型（LRMs）在回答之前生成长链思维轨迹，造成显著的推理开销。剪枝可以减少这一成本，但其有效性依赖于用于估计参数重要性的校准数据。最近的研究在模型自身的回滚上进行校准，而不是使用通用数据集，但对所有推理标记采取统一处理，无论它们是否对成功推理有贡献。因此，剪枝通过统计显著性保护权重，而不是通过它们对正确推理的贡献，因此错误计算背后的权重与正确计算背后的权重一样容易幸存。这些错误模式随后被带入剪枝后的模型，降低了推理质量，导致准确性降低和推理轨迹延长。我们提出了基于结果的校准用于大型推理模型剪枝（OBC-Prune）以填补这一空白。OBC首先从模型不一致回答的问题中构建难度匹配的正确和错误回滚对。然后，通过基于干预的分析估计每个推理句子的因果重要性，量化去除其影响对后续预测的影响。这些因果重要性分数被转换为每个标记的权重，重新调整用于单次剪枝方法（SparseGPT、Wanda、ALPS）的校准激活，而不修改底层剪枝算法。在DeepSeek-R1-Distill-Qwen 1.5B、7B和14B模型上进行的40\%和50\%稀疏性实验表明，在MATH500、LiveCodeBench和AIME 2025上，大多数模型规模和稀疏水平上，OBC-Prune在最先进的校准基准上表现出一致的改进。这些结果表明，保留因果重要的推理电路是一种比均匀保留观察到的激活更有效的剪枝目标。
cs.AI / 21 / 2609.17921
Collaborative Memory for Multi-Agent VLM Systems
多智能体视觉语言模型系统的协同记忆
Abstract
Vision-language model (VLM) agents combine specialized perception, tools, and reasoning to address complex visual tasks. In multi-agent settings, different agents inspect different image regions, video frames, or visual representations, so collaboration extends beyond distributed reasoning to distributed perception. This makes shared visual context a central problem in VLM agent collaboration. In this paper, we frame memory hierarchy, cross-agent sharing, and consistency mechanisms around the need to reconcile interpretations and update dependent reasoning. Effective collaboration requires agents to build on contributions from other agents, recover missing visual context, and reconcile differing interpretations as new evidence emerges. Shared visual memory preserves not only images or textual summaries but also the dependencies among observations, agent interpretations, and subsequent reasoning. Together, these design considerations shape how information flows and evolves across VLM agents. The proposed framework provides a foundation for building reliable and resource-efficient agent teams.
Chinese Translation
视觉语言模型（VLM）智能体结合了专业的感知、工具和推理能力，以应对复杂的视觉任务。在多智能体环境中，不同的智能体检查不同的图像区域、视频帧或视觉表征，因此协作不仅限于分布式推理，还扩展到分布式感知。这使得共享视觉上下文成为VLM智能体协作中的一个核心问题。本文围绕调和解释和更新依赖推理的需要，构建了记忆层次结构、跨智能体共享和一致性机制。有效的协作要求智能体在其他智能体的贡献基础上进行构建，恢复缺失的视觉上下文，并在新证据出现时调和不同的解释。共享视觉记忆不仅保留图像或文本摘要，还保留观察、智能体解释和后续推理之间的依赖关系。这些设计考虑共同塑造了信息在VLM智能体之间的流动和演变方式。所提出的框架为构建可靠且资源高效的智能体团队提供了基础。
cs.AI / 22 / 2609.17965
Measuring AI Leadership: Development and Validation of a Multidimensional Measure for AI-Native Organizations
衡量人工智能领导力：针对人工智能原生组织的多维度测量工具的开发与验证
Abstract
AI is changing what leaders must judge, explain, learn, and coordinate, yet existing measures do not capture these behaviors at the level needed to study leadership in AI-enabled work. We develop the AI Leadership Battery, which organizes 36 behaviorally specific subdimensions into 11 theory-specified content families. Following established scale-development procedures, the research used deductive item generation; content validation of definitional correspondence and definitional distinctiveness; exploratory factor analysis and item reduction; confirmatory factor analysis in independent samples; and tests of internal consistency reliability, convergent validity, discriminant validity, and criterion-related and incremental validity. Across the development and validation studies, the analyses provided evidence for the Battery's content, multidimensional structure, reliability, and distinction from selected orbiting constructs. The Battery also contributed additional information beyond orbiting constructs across organizational growth, decision speed, customer/stakeholder response capability, AI-enabled team performance, AI-enabled work experience, AI security and risk management, and AI adoption and integration. The resulting measure provides researchers with a behavioral framework for examining how leaders in AI-enabled work regulate judgment, learning, adaptation, transparency, and accountability.
Chinese Translation
人工智能正在改变领导者必须判断、解释、学习和协调的内容，但现有的测量工具未能在研究人工智能驱动工作中的领导力所需的层面上捕捉这些行为。我们开发了人工智能领导力电池（AI Leadership Battery），将36个行为特定的子维度组织为11个理论指定的内容家族。根据既定的量表开发程序，本研究采用了演绎性条目生成；对定义对应性和定义独特性的内容验证；探索性因子分析和条目减少；在独立样本中的确认性因子分析；以及内部一致性可靠性、收敛效度、区分效度、标准相关效度和增量效度的测试。在开发和验证研究中，分析提供了电池内容、多维结构、可靠性及其与选定相关构念的区分的证据。该电池还提供了超越相关构念的额外信息，涵盖了组织增长、决策速度、客户/利益相关者响应能力、人工智能驱动的团队绩效、人工智能驱动的工作体验、人工智能安全与风险管理，以及人工智能的采纳与整合。最终的测量工具为研究人员提供了一个行为框架，以考察人工智能驱动工作中的领导者如何调节判断、学习、适应、透明度和问责制。
cs.AI / 23 / 2609.17969
Memory Has Geometry: Non-Uniform Geometric Memory for Long-Horizon Personalized AI
记忆具有几何性：用于长时间个性化人工智能的非均匀几何记忆
Abstract
Long-term memory is becoming a core substrate for personalized AI, yet most systems still represent personalization as discrete records in a largely static latent space, accessed under one global similarity notion. For data mining, this creates a mismatch: the evidence is a temporal event stream, while the dominant abstraction is a searchable record set. We argue that long-horizon personalization should instead model memory as a user-specific dynamical state space with locally heterogeneous geometry. Geometry here is a computational language, not a literal claim about cognition: it captures stable versus volatile regions, variable-rate drift, heterogeneous neighborhoods, and uncertainty about current user state. Profiles and isolated events remain useful as points, but interaction, feedback, and elapsed time induce trajectories. Memory access then becomes trajectory-conditioned reconstruction of the relevant user state, not only nearest-neighbor lookup.
Chinese Translation
长期记忆正成为个性化人工智能的核心基础，但大多数系统仍将个性化表示为在一个基本静态的潜在空间中的离散记录，并在一个全球相似性概念下进行访问。这在数据挖掘中造成了不匹配：证据是一个时间事件流，而主导的抽象是一个可搜索的记录集。我们认为，长时间个性化应将记忆建模为一个用户特定的动态状态空间，具有局部异质几何性。这里的几何性是一种计算语言，而不是对认知的字面声明：它捕捉稳定与易变区域、可变速率漂移、异质邻域以及对当前用户状态的不确定性。个人档案和孤立事件仍然作为点是有用的，但交互、反馈和经过时间会引发轨迹。因此，记忆访问变成了基于轨迹的相关用户状态重构，而不仅仅是最近邻查找。
cs.AI / 24 / 2609.17977
When to Call an LLM: A Confidence-Gated Hybrid for Cost-Effective Emotion Recognition in Conversational AI
何时调用大型语言模型：一种基于置信度的混合方法用于成本效益高的对话情感识别
Abstract
Emotion recognition in conversation (ERC) is a production capability behind agent-assist prompts, escalation routing, and post-call analytics in contact-center-as-a-service (CCaaS) platforms, where cost and latency constraints matter as much as accuracy. We report a systems-level comparison of three deployment options for dialogue-contextual ERC: a low-cost stacked ensemble (sentence embeddings, windowed context, RandomForest/XGBoost/logistic-regression stacking), off-the-shelf LLM prompting (GPT-4o-mini; zero-shot, few-shot, chain-of-thought), and a confidence-gated hybrid that escalates only the ensemble's least-confident predictions to the LLM - modeled on IVA-to-human-agent escalation policies used in production contact centers. On IEMOCAP, the ensemble significantly outperforms every LLM configuration (0.595 vs. 0.460-0.536 weighted F1, p < 0.0001) at a fraction of the cost and sub-10ms latency; on MELD and CMU-MOSI the ranking reverses, showing neither pure system is a safe default. The confidence-gated hybrid resolves this by Pareto-dominating both pure systems on all three datasets (0.620, 0.643, 0.824 weighted F1) while routing the majority of traffic through the near-zero-cost ensemble, translating to roughly $10-85 per million utterances versus $99-170 for an LLM-only pipeline. The escalation policy is not an opaque cost/accuracy dial: escalated turns disproportionately follow an emotion or sentiment shift, giving operators an interpretable, auditable routing signal, and the ensemble's confidence is well-calibrated and safely under- rather than over-confident. The pattern holds across three datasets and two LLM providers. Confidence-gated cascading is established in general ML systems; our contribution is showing it transfers cleanly to dialogue-contextual ERC, yielding a concrete deployment recipe for CCaaS and conversational-AI platforms deciding how to allocate LLM spend.
Chinese Translation
对话中的情感识别（ERC）是代理辅助提示、升级路由和呼叫后分析等功能的基础，尤其在联络中心即服务（CCaaS）平台中，成本和延迟约束与准确性同样重要。我们报告了三种对话上下文情感识别的部署选项的系统级比较：一种低成本的堆叠集成方法（句子嵌入、窗口上下文、随机森林/XGBoost/逻辑回归堆叠）、现成的大型语言模型（LLM）提示（GPT-4o-mini；零-shot、少-shot、思维链）以及一种基于置信度的混合方法，该方法仅将集成模型中置信度最低的预测结果升级到LLM——该模型基于生产联络中心中使用的IVA到人类代理的升级策略。在IEMOCAP数据集上，该集成模型显著优于每种LLM配置（0.595对0.460-0.536加权F1，p < 0.0001），且成本仅为其一小部分，延迟低于10毫秒；在MELD和CMU-MOSI数据集上，排名则相反，显示出纯系统并不是一个安全的默认选择。基于置信度的混合方法通过在所有三个数据集上Pareto优于这两种纯系统（0.620、0.643、0.824加权F1）来解决这一问题，同时将大部分流量路由通过几乎零成本的集成模型，成本约为每百万次发言10-85美元，而仅使用LLM的管道则为99-170美元。升级策略并不是一个不透明的成本/准确性调节器：升级的对话轮次不成比例地跟随情感或情绪的变化，为操作员提供了一个可解释、可审计的路由信号，而集成模型的置信度经过良好校准，且安全地表现出低于而非过于自信的特征。该模式在三个数据集和两个LLM提供商中均保持一致。基于置信度的级联在一般机器学习系统中已被确立；我们的贡献在于展示其在对话上下文情感识别中的有效转移，为CCaaS和对话AI平台提供了具体的部署方案，以决定如何分配LLM的支出。
cs.AI / 25 / 2609.17983
Contiguity, Not Importance: Budgeted Repair of Stale KV Caches After Document Edits
相邻性，而非重要性：文档编辑后过期 KV 缓存的预算修复
Abstract
KV-cache reuse can reduce inference cost in retrieval-augmented generation and agentic systems, but cached contexts may become stale when retrieved knowledge, working memory, or user state is edited. Under causal self-attention, even a local edit can affect downstream KV states. A full re-prefill reliably restores consistency but is costly, whereas refreshing only the edited span can leave downstream dependencies stale. We formulate in-place repair as budgeted recomputation and compare training-free position-selection policies on a factual RAG benchmark with matched direct and derived edits. Across three model families, all policies repair direct cases, but derived cases clearly separate them. At the primary budget, a contiguous edit-local window recovers at least 0.94 of the post-edit answer margin and substantially outperforms attention-based, KV-deviation, and structural selectors. Mechanistic analysis shows that position sets effective under clean-state transplantation can fail under actual recomputation because scattered positions inherit surrounding staleness. The edit-local advantage also depends on adjacency and largely disappears when the answer-bearing text moves downstream. Because answer-relevant edits almost always corrupt model behavior, failure severity is difficult to predict, and repair is 13-21 times faster than full re-prefill, our results support unconditional edit-local repair when the dependent text remains adjacent to the edit.
Chinese Translation
KV 缓存的重用可以降低检索增强生成和自主系统中的推理成本，但当检索的知识、工作记忆或用户状态被编辑时，缓存的上下文可能会变得过期。在因果自注意力机制下，即使是局部编辑也会影响下游的 KV 状态。完全重新填充可以可靠地恢复一致性，但成本较高，而仅刷新编辑的范围可能会使下游依赖关系变得过期。我们将就地修复形式化为预算重新计算，并在一个与直接和派生编辑匹配的事实 RAG 基准上比较无训练的位置信息选择策略。在三种模型系列中，所有策略都能修复直接案例，但派生案例明显区分了它们。在主要预算下，一个相邻的编辑局部窗口至少恢复了 0.94 的编辑后答案边际，并显著优于基于注意力的、KV 偏差的和结构选择器。机制分析表明，在干净状态下有效的位置信息集在实际重新计算中可能会失败，因为分散的位置会继承周围的过期性。编辑局部优势还依赖于相邻性，当答案相关文本移动到下游时，这种优势在很大程度上消失。由于与答案相关的编辑几乎总是会破坏模型行为，因此失败的严重性难以预测，而修复速度是完全重新填充的 13-21 倍，我们的结果支持在依赖文本与编辑保持相邻时进行无条件的编辑局部修复。
cs.AI / 26 / 2609.17984
TuiML: Machine Learning for AI Agents
TuiML：为人工智能代理设计的机器学习
Abstract
Machine-learning libraries such as Weka and scikit-learn were designed for human programmers. Language-model agents now use these same libraries by recalling APIs from memory and writing code, an approach that hides what a library offers, delays errors until runtime, and loses experimental state between turns. We present TuiML, a self-contained machine-learning library built for AI agents, with native algorithms across supervised, unsupervised, time-series, data handling, tuning, and evaluation tasks. Every component describes itself through machine-readable metadata and parameter schemas, so an agent can search the library, inspect components, compose validated workflows, and register new ones that become discoverable in turn. Every call is validated, seeded, and traced, and sessions export as runnable notebooks, making experiments reproducible by construction. One specification layer drives the Model Context Protocol (MCP), agent-framework adapters, a Python API, a CLI, and local model serving, while data and models never leave the machine. Benchmarks show TuiML remains predictively competitive with scikit-learn and Weka. While looking like a conventional library to a human user, TuiML is designed for agents first, allowing them to read, extend, and operate machine learning autonomously. TuiML is open source, with documentation at https://tuiml.ai.
Chinese Translation
机器学习库如 Weka 和 scikit-learn 是为人类程序员设计的。语言模型代理现在通过从记忆中回忆 API 并编写代码来使用这些相同的库，这种方法隐藏了库所提供的功能，延迟了错误直到运行时，并在回合之间丢失了实验状态。我们提出了 TuiML，这是一个为人工智能代理构建的自包含机器学习库，具有针对监督学习、无监督学习、时间序列、数据处理、调优和评估任务的原生算法。每个组件通过机器可读的元数据和参数模式自我描述，因此代理可以搜索库、检查组件、组合经过验证的工作流，并注册新的工作流，使其在后续中可被发现。每次调用都经过验证、种子化和追踪，会话导出为可运行的笔记本，从而使实验在构建上可重复。一层规范驱动模型上下文协议（Model Context Protocol, MCP）、代理框架适配器、Python API、命令行接口（CLI）和本地模型服务，同时数据和模型始终保留在本机。基准测试表明 TuiML 在预测性能上与 scikit-learn 和 Weka 具有竞争力。尽管对人类用户而言看起来像一个传统库，但 TuiML 首先是为代理设计的，使它们能够自主地读取、扩展和操作机器学习。TuiML 是开源的，文档可在 https://tuiml.ai 查阅。
cs.AI / 27 / 2609.17985
RideWay: Benchmarking Efficient Task Completion for Tool-Using Language Agents
RideWay：工具使用语言代理的高效任务完成基准测试
Abstract
AI agents are usually evaluated by whether they complete a task. In interactive service settings, a successful agent can still frustrate users by asking repeated questions, performing redundant searches, or making avoidable revisions. We introduce RideWay, an efficiency-centered benchmark for ridehailing agents in a stateful tool-calling environment, together with Efficiency Utility, a success-gated metric that discounts successful trajectories for excess tool calls and user-facing turns relative to task-specific reference effort. Human paired preferences calibrate the relative penalties, reflecting an aggregate service-workflow trade-off: extra dialogue often creates visible friction, whereas extra tool use can sometimes verify constraints or preserve user intent. Across 58 tasks and 24 models, the fitted penalty for excess turns is about twice that for excess tool calls. On task-disjoint held-out preferences, Efficiency Utility achieves 78.7% accuracy overall: 90.6% when trajectories differ in turns, but chance-level accuracy when they differ solely in tool calls - the axis on which human annotators agree least. RideWay therefore makes interaction efficiency measurable alongside task success, while exposing the boundary of count-based tool-use evaluation.
Chinese Translation
人工智能代理通常通过是否完成任务来进行评估。在互动服务环境中，一个成功的代理仍然可能通过反复提问、执行冗余搜索或进行可避免的修订来使用户感到沮丧。我们提出了RideWay，这是一个以效率为中心的基准，专为在有状态工具调用环境中的叫车代理设计，伴随而来的是效率效用（Efficiency Utility），这是一种成功门限度量，针对相对于任务特定参考努力的过度工具调用和用户面对的转折折扣成功轨迹。人类配对偏好校准了相对惩罚，反映了总体服务工作流程的权衡：额外的对话往往会造成明显的摩擦，而额外的工具使用有时可以验证约束或保持用户意图。在58个任务和24个模型中，过度转折的拟合惩罚约为过度工具调用的两倍。在任务不重叠的保留偏好中，效率效用的整体准确率为78.7%：当轨迹在转折上有所不同时，准确率为90.6%，但当它们仅在工具调用上有所不同时，准确率则达到偶然水平——这是人类注释者最不一致的轴。因此，RideWay使得交互效率可以与任务成功一起被量化，同时揭示了基于计数的工具使用评估的边界。
cs.AI / 28 / 2609.17987
Multimodal Conditioning of Fine-Tuned Stable Diffusion XL for Controllable and Culturally Faithful Ulos Motif Generation
多模态条件下微调稳定扩散XL的可控和文化忠实的乌洛斯图案生成
Abstract
The traditional Batak Ulos weaving industry faces growing challenges in producing diverse, innovative motifs due to limitations in conventional, manually driven design methods. This study proposes a multimodal generative framework integrating a fine-tuned Latent Diffusion Model (Stable Diffusion XL v1.0 via LoRA) with a Multimodal Large Language Model (LLaMA 1.5-7B) to enable controllable, culturally faithful Ulos motif generation. Four complementary conditioning mechanisms: text, image, representation, and semantic map (via ControlNet) jointly guide the generation process, each governing a distinct aspect from semantic intent to spatial layout. A five level ablation study across three scenarios (shape transformation, colour variation, and high-complexity input) shows that conditioning effectiveness is not proportional to the number of mechanisms combined: Text + Image + Semantic Map achieved the best FID (270) and CLIP Score (0.65 - 0.70) but the weakest SSIM (0.65), while Text + Image + Representation offered the best overall balance, with stable SSIM (0.84) and competitive FID (280). Combining all four mechanisms yielded the weakest FID (330), indicating conflicting optimization signals. Qualitative evaluation by nine weavers and thirty public participants confirmed statistically significant positive acceptance (Wilcoxon, p=0.007 and p<0.001, respectively). A web-based prototype supporting text-to-image and image-to-image generation was also developed, offering a practical digital design tool for cultural heritage preservation.
Chinese Translation
传统的巴塔克乌洛斯编织产业在生产多样化和创新图案方面面临越来越大的挑战，这主要是由于传统手工设计方法的局限性。本研究提出了一种多模态生成框架，将微调的潜在扩散模型（Stable Diffusion XL v1.0通过LoRA）与多模态大型语言模型（LLaMA 1.5-7B）相结合，以实现可控且文化忠实的乌洛斯图案生成。四种互补的条件机制：文本、图像、表示和语义图（通过ControlNet）共同指导生成过程，每种机制控制从语义意图到空间布局的不同方面。通过在三种场景（形状变换、颜色变化和高复杂度输入）下进行的五级消融研究表明，条件有效性与组合机制的数量并不成正比：文本 + 图像 + 语义图实现了最佳的FID（270）和CLIP得分（0.65 - 0.70），但SSIM（0.65）最弱，而文本 + 图像 + 表示则提供了最佳的整体平衡，具有稳定的SSIM（0.84）和具有竞争力的FID（280）。结合所有四种机制的结果显示最弱的FID（330），表明存在优化信号的冲突。九位编织师和三十位公众参与者的定性评估确认了统计学上显著的积极接受度（Wilcoxon，p=0.007和p<0.001）。同时开发了一个支持文本到图像和图像到图像生成的基于网络的原型，提供了一个实用的数字设计工具，以促进文化遗产的保护。
cs.AI / 29 / 2609.18004
Missing Bridges: Composition-Aware Active Imitation Learning
缺失的桥梁：考虑组合的主动模仿学习
Abstract
Active imitation learning reduces expert effort by allowing a learner to request the demonstrations it needs. Existing methods typically select these requests for their expected information gain about the expert policy. In structured multi-task domains, however, the number of start-goal tasks may grow combinatorially despite their solutions sharing reusable behavior. This makes composable behaviors especially valuable, since a single demonstration may help solve many tasks at once. Prior methods do not explicitly account for this value when selecting which demonstration to request. We introduce Adaptive Agents via Latent Topologies (AALT), which requests demonstrations that maximize expected gains in start-goal connectivity. We further show that this objective is formally tied to information gain about task reachability. AALT organizes existing demonstrations into a topology of latent hub states connected by learned behaviors, identifies high-value bridge demonstrations that are likely to enable many tasks at once, and grounds each to an expert query. At inference, it plans through the resulting topology and conditions a diffusion policy on each successive hub transition. In a simulated UR5e robot ordered-retrieval domain with 72 tasks, AALT improved from 42/72 to 72/72 (100%) successful tasks consistently using only 3 demonstrations totaling 5 transitions beyond the initial dataset. After 20 demonstrations, the strongest baseline averaged 88.6% success using 98 transitions.
Chinese Translation
主动模仿学习通过允许学习者请求所需的示范来减少专家的工作量。现有方法通常根据其对专家策略的预期信息增益来选择这些请求。然而，在结构化的多任务领域中，起始-目标任务的数量可能会呈组合增长，尽管它们的解决方案共享可重用的行为。这使得可组合的行为尤其有价值，因为单个示范可以帮助一次性解决多个任务。之前的方法在选择请求哪个示范时并未明确考虑这一价值。我们提出了通过潜在拓扑的自适应代理（Adaptive Agents via Latent Topologies, AALT），该方法请求最大化起始-目标连通性的预期增益的示范。我们进一步表明，这一目标与任务可达性的信息增益在形式上是相关的。AALT将现有示范组织成由学习到的行为连接的潜在中心状态的拓扑，识别出高价值的桥梁示范，这些示范可能同时使多个任务得以实现，并将每个示范与专家查询相结合。在推理阶段，它通过生成的拓扑进行规划，并在每个连续的中心状态转移上条件化扩散策略。在一个模拟的UR5e机器人有序检索领域中，AALT在72个任务中成功率从42/72提高到72/72（100%），仅使用3个示范，总共进行了5次超出初始数据集的转移。在20个示范后，最强的基线平均成功率为88.6%，使用了98次转移。
cs.AI / 30 / 2609.18057
Anchoring What Matters: A Dual-Level Learning Framework for Visually-Grounded Multimodal Reasoning
锚定重要性：一种用于视觉基础多模态推理的双层学习框架
Abstract
Reinforcement learning with verifiable rewards (RLVR) has significantly improved the reasoning capabilities of large vision-language models (LVLMs). However, standard on-policy RLVR algorithms face a critical optimization bottleneck in preserving and reinforcing visually grounded reasoning behaviors: valuable visually-grounded reasoning trajectories are discarded after a single update, while uniform token advantage allocation prevents the model from reinforcing critical perception or reasoning steps. To bridge this gap, we propose PIVOT, a dual-level learning framework that anchors policy optimization around informative visual reasoning signals. Specifically, PIVOT introduces a self-calibrated experience replay mechanism, which selectively collects and replays visually-grounded historical experiences as stable reference anchors for policy optimization. Building upon this, we further design a vision-guided advantage allocation mechanism to allocate additional vision-aware advantages to tokens based on their local visual support and impact on downstream reasoning. Extensive experiments across diverse benchmarks demonstrate that PIVOT achieves highly competitive performance in enhancing the multimodal reasoning capabilities of LVLMs.
Chinese Translation
具有可验证奖励的强化学习（RLVR）显著提升了大型视觉语言模型（LVLMs）的推理能力。然而，标准的在线策略RLVR算法在保持和增强视觉基础推理行为方面面临关键的优化瓶颈：有价值的视觉基础推理轨迹在单次更新后被丢弃，而均匀的标记优势分配阻止了模型强化关键的感知或推理步骤。为了解决这一问题，我们提出了PIVOT，一种围绕信息丰富的视觉推理信号进行策略优化的双层学习框架。具体而言，PIVOT引入了一种自校准的经验重放机制，该机制选择性地收集和重放视觉基础的历史经验，作为策略优化的稳定参考锚点。在此基础上，我们进一步设计了一种视觉引导的优势分配机制，根据标记的局部视觉支持和对下游推理的影响，为标记分配额外的视觉感知优势。在多种基准测试中的广泛实验表明，PIVOT在增强LVLMs的多模态推理能力方面实现了高度竞争的性能。
cs.AI / 31 / 2609.18063
The Other Half of the Memory Wall: Serving 35B MoEs from SSD with Trained Routing Prediction
记忆墙的另一半：通过训练的路由预测从SSD服务35B MoEs
Abstract
Mixture-of-experts (MoE) inference on consumer hardware is bounded by weight memory: a 35B-class model is 19.5GB at 4-bit, and sparsity shrinks the compute per token, not the bytes that must be held. Naive offloading to SSD does not help on its own, because layer N+1's experts must be chosen before layer N's output exists, so the reads cannot start early enough to hide behind compute. We present Edge0, a streaming MoE inference engine that closes the gap with a prerouter: a per-layer head predicts the next layer's routing one token ahead, and the prediction is consumed as the routing itself, so the staged expert set equals the routed set and nothing is dropped. An unmerged recovery LoRA, trained on the student path, pays back the quality lost to int4 quantization and routing replacement. On a single 24GB machine, Edge0 serves a 35B MoE at 20tok/s inside 3GiB of peak active memory, within a few points of its fp16 teacher on average across five public benchmarks. An 8B tier runs on the same framework, and the framework, checkpoints, and adapters are open source.
Chinese Translation
在消费硬件上进行混合专家（MoE）推理受到权重内存的限制：一个35B级别的模型在4位量化下占用19.5GB的内存，而稀疏性缩小了每个token的计算量，而不是必须保持的字节数。简单地将计算卸载到SSD并不能单独解决问题，因为在层N的输出存在之前，层N+1的专家必须被选择，因此读取操作无法提前开始以隐藏计算的延迟。我们提出了Edge0，一个流式MoE推理引擎，通过一个预路由器来缩小这一差距：每层的头部预测下一层的路由，提前一个token进行预测，并将该预测作为路由本身使用，因此分阶段的专家集等于路由集，没有任何信息丢失。一个未合并的恢复LoRA，在学生路径上训练，弥补了因int4量化和路由替换而损失的质量。在一台24GB的机器上，Edge0以20tok/s的速度服务于一个35B MoE，峰值活跃内存在3GiB以内，在五个公共基准测试中平均与其fp16教师模型相差几分。一个8B级别的模型在同一框架下运行，而该框架、检查点和适配器均为开源。
cs.AI / 32 / 2609.18072
Teaching AI, Robotics, & Community: A Hubs-Based K-12 Education Framework for Reaching Rural Schools
教学人工智能、机器人技术与社区：面向农村学校的基于中心的K-12教育框架
Abstract
K-12 robotics and AI education remains difficult to scale, especially in rural regions lacking sustained technical mentorship. Programs like FIRST provide competition pathways and instructional opportunities, but they do not eliminate the need for local programming and robotics expertise. We introduce AI, Robotics, & Community (ARC), a hubs-based framework where colleges train undergraduate mentors and host workshops for nearby K-12 teams. Mature school programs can become secondary hubs that support additional schools, creating a self-reinforcing education loop where mentorship reach propagates geographically and can even grow super-linearly. We first evaluate ARC through a trial deployment at one university. The trial created three rural robotics teams. On five-point Likert surveys, mean increases in K-12 programming knowledge, resource access, and practice opportunities were 2.00, 2.25, and 1.25 points. Likewise, undergraduate confidence teaching technical concepts, adapting explanations, managing groups, and finding mentoring enjoyable and meaningful increased by 1.29, 1.14, 1.00, and 1.14 points. Additionally, we create a spatial Markov model of ARC's growth and simulate it using the state of Indiana as a testbed. Under moderate conditions, we find that ARC reaches 74% of Indiana's 1,925 public K-12 schools and produces 992 robotics programs after 40 years, compared with 161 projected under natural growth alone. Together, these results show ARC can create and support rural robotics programs, train undergraduate AI and robotics mentors, and potentially scale mentorship across a region.
Chinese Translation
K-12机器人和人工智能教育的规模化仍然面临困难，尤其是在缺乏持续技术指导的农村地区。像FIRST这样的项目提供了竞赛路径和教学机会，但并未消除对地方编程和机器人专业知识的需求。我们提出了人工智能、机器人技术与社区（AI, Robotics, & Community，ARC）这一基于中心的框架，大学培训本科生导师并为附近的K-12团队举办研讨会。成熟的学校项目可以成为次级中心，支持其他学校，形成一个自我强化的教育循环，指导的覆盖范围在地理上扩展，甚至可以实现超线性增长。我们首先通过在一所大学的试点部署评估ARC。该试点创建了三个农村机器人团队。在五点李克特量表调查中，K-12编程知识、资源获取和实践机会的平均增加分别为2.00、2.25和1.25分。同样，本科生在教授技术概念、调整解释、管理小组以及发现指导工作有趣和有意义方面的信心分别增加了1.29、1.14、1.00和1.14分。此外，我们创建了ARC增长的空间马尔可夫模型，并以印第安纳州作为测试平台进行模拟。在适度条件下，我们发现ARC能够覆盖印第安纳州1,925所公立K-12学校的74%，并在40年后产生992个机器人项目，而仅依靠自然增长预计只能产生161个。综合来看，这些结果表明ARC能够创建和支持农村机器人项目，培训本科生的人工智能和机器人导师，并可能在一个地区内扩展指导。
cs.AI / 33 / 2609.18080
Decodability is Not Causality: Dissociating Probe Readouts from Behavioral Drivers via SAE Decomposition
可解码性并非因果性：通过稀疏自编码器分解将探针读出与行为驱动因素分离
Abstract
Linear probes can decode safety-relevant concepts such as truthfulness from language-model activations, but probe accuracy may show only decodability, not that the features the probe weights causally drive model behavior. We demonstrate that this gap cannot be closed from the geometry of probe weights alone: the features geometrically aligned with probe direction need not be the ones the model uses, so causal relevance requires intervention. We introduce a feature-level diagnostic that decomposes a deployed True/False probe into sparse-autoencoder (SAE) features, ranks those features by both probe alignment and by gradient sensitivity of the model's behavior, and ablates the resulting shared, probe-only, and random feature sets under a coherence gate. On the truth probe of Buerger et al. (2024) (TTPD), applied in the instructed truth/deception setting of Long et al. (2025) for Gemma2-9B-Instruct, the two rankings overlap only weakly (about 12%, Spearman rho = 0.10), and ablation dissociates them sharply: features the probe shares with the model flip the output far more (up to 27%) than equally sized probe-only (6%) or random (1%) features at full coherence, while probe-only features instead perturb the probe's own readout. The dissociation holds across five seeds and a held-out split, and an activation-aware selection of features flips behavior nearly three times as often as the probe's geometric top features (17.6% vs. 6.1%). In this setting, therefore, the geometric projection of a probe's weight vector alone does not identify the features the model causally uses; however, combining probe information with feature activation statistics recovers substantially more behaviorally causal features, and coherence-gated SAE intervention is needed to separate them from probe readouts.
Chinese Translation
线性探针可以从语言模型的激活中解码与安全相关的概念，例如真实性，但探针的准确性可能仅显示可解码性，并不能证明探针权重所代表的特征在因果上驱动模型行为。我们证明，仅凭探针权重的几何特性无法弥补这一差距：与探针方向几何对齐的特征不一定是模型实际使用的特征，因此因果相关性需要干预。我们引入了一种特征级诊断方法，将部署的真/假探针分解为稀疏自编码器（SAE）特征，并根据探针对齐度和模型行为的梯度敏感性对这些特征进行排名，并在一致性门控下消融结果共享、仅探针和随机特征集。在Buerger等人（2024）提出的真实性探针（TTPD）中，应用于Long等人（2025）为Gemma2-9B-Instruct设定的指令真实性/欺骗环境中，这两种排名仅弱重叠（约12%，Spearman rho = 0.10），而消融则明显将它们分离：探针与模型共享的特征在完全一致性下对输出的影响远大于同样大小的仅探针特征（6%）或随机特征（1%），而仅探针特征则扰动探针自身的读出。这种分离在五个种子和一个保留分割中均保持一致，且激活感知的特征选择使得行为翻转的频率几乎是探针几何顶级特征的三倍（17.6%对6.1%）。因此，在这种情况下，探针权重向量的几何投影本身并不能识别模型因果使用的特征；然而，将探针信息与特征激活统计结合，可以恢复更多行为上因果的特征，并且需要一致性门控的SAE干预来将它们与探针读出分离。
cs.AI / 34 / 2609.18099
When Is Graph Structure Worth Its Cost? The Case for Structure Pricing in Retrieval-Augmented Generation
图结构何时值得其成本？检索增强生成中的结构定价案例
Abstract
Graph-based retrieval-augmented generation (RAG) can help answer questions that require information from many documents. However, building a graph often requires many language-model calls during ingestion. It is therefore important to ask whether its quality gains justify the additional cost. We present EffiRAG, a graph-based RAG system designed to reduce this cost. It uses the graph to locate relevant passages and generates answers from the original text. This design preserves source information while keeping graph construction and query processing lightweight. We evaluate EffiRAG on UltraDomain, which contains 120 open-ended questions from four domains. Compared with LightRAG-hybrid, EffiRAG produces the preferred answer on 93 questions. LightRAG is preferred on 7, and the remaining 20 are splits. EffiRAG also reduces total system cost by 57 percent, from USD 0.952 to USD 0.408. The cost includes language-model calls during ingestion and querying. The advantage remains as the corpus grows. At 10 and 20 documents per domain, EffiRAG uses a lightweight, non-LLM filter to skip low-salience chunks. It remains preferred over LightRAG-hybrid. It costs 4.2 times and 4.5 times less, respectively. The comparisons identify different quality-cost trade-offs. Graph-based RAG systems should therefore be evaluated by both answer quality and cost. The results favor graph structure that locates and preserves source evidence.
Chinese Translation
基于图的检索增强生成（RAG）可以帮助回答需要从多个文档中获取信息的问题。然而，构建图通常在摄取过程中需要多次调用语言模型。因此，重要的是要问其质量提升是否值得额外的成本。我们提出了EffiRAG，一种旨在降低此成本的基于图的RAG系统。它利用图来定位相关段落，并从原始文本中生成答案。这种设计在保持源信息的同时，使图的构建和查询处理保持轻量化。我们在UltraDomain上评估EffiRAG，该数据集包含来自四个领域的120个开放式问题。与LightRAG-hybrid相比，EffiRAG在93个问题上生成了更受欢迎的答案。LightRAG在7个问题上更受青睐，其余20个问题为平分。EffiRAG还将系统的总成本降低了57%，从0.952美元降至0.408美元。该成本包括在摄取和查询过程中对语言模型的调用。随着语料库的增长，这一优势依然存在。在每个领域10和20个文档时，EffiRAG使用轻量级的非LLM过滤器跳过低显著性的片段。它在这两种情况下均优于LightRAG-hybrid，分别降低了4.2倍和4.5倍的成本。这些比较揭示了不同的质量-成本权衡。因此，基于图的RAG系统应通过答案质量和成本进行评估。结果支持定位和保留源证据的图结构。
cs.AI / 35 / 2609.18123
AutoTuneBench: Trustworthy Measurement for Agent Auto-Tuning of LLM Serving Engines
AutoTuneBench：可信的LLM服务引擎代理自动调优测量
Abstract
Large language model agents tune GPU kernels and serving engines through a closed loop of propose, measure, and keep, but the measurements behind this loop are not trustworthy. We characterize four failure modes from a four-day pilot corpus of 619 model calls: strawman baselines manufacture speedups, absolute times do not transfer across machines, saturated tasks nullify comparisons, and infrastructure defects impersonate science. We present AutoTuneBench, a benchmark and measurement protocol that makes trust architectural. The protocol is frozen as code with test-enforced provenance; a database-level validator rejects out-of-protocol results; anti-cheat checks run outside the agent's modification surface; comparisons follow pre-registered readouts; and measurements anchor to externally published results, grounded in paired-seed statistics with a 5\% cross-run coefficient-of-variation cap. Honest measurement rewrites the headlines: our best kernel reads 10.6x against a naive baseline but 2.03x against the honest one; one configuration delivers 1.174x on one machine and 1.0049x on another; a pre-registered on/off comparison nulls at a shared wall (2.4840 vs 2.4957\,ms); and the KernelBench Level-1 suite admits 51\% of tasks with median speedup 1.0001x over PyTorch eager. The protocol, the two-engine corpus (vLLM and SGLang), and its audit trail are released as open artifacts.
Chinese Translation
大型语言模型代理通过提议、测量和保持的闭环来调优GPU内核和服务引擎，但该闭环背后的测量并不可信。我们从为期四天的619个模型调用的试点语料库中描述了四种失败模式：虚假基线制造加速，绝对时间无法在不同机器间转移，饱和任务使比较失效，以及基础设施缺陷伪装成科学。我们提出了AutoTuneBench，一个使信任成为架构的基准和测量协议。该协议以代码形式冻结，并通过测试强制来源；数据库级验证器拒绝不符合协议的结果；反作弊检查在代理的修改表面之外运行；比较遵循预注册的读数；测量锚定于外部发布的结果，基于配对种子统计，交叉运行的变异系数上限为5%。诚实的测量重写了头条新闻：我们的最佳内核在与天真基线比较时达到10.6倍，但与诚实基线比较时为2.03倍；一种配置在一台机器上提供1.174倍，在另一台机器上为1.0049倍；一个预注册的开/关比较在共享墙处失效（2.4840 vs 2.4957 ms）；而KernelBench Level-1套件接受51%的任务，媒体加速为1.0001倍，相较于PyTorch eager。该协议、两个引擎语料库（vLLM和SGLang）及其审计记录作为开放文物发布。
cs.AI / 36 / 2609.18126
Designing Agentic AI Workflow Portfolios under Imperfect Selection and Compute Cost
在不完美选择和计算成本下设计自主智能工作流组合
Abstract
Agentic AI systems often approach the same task through multiple workflows that differ in reasoning strategy, verification structure, and compute cost. A natural deployment policy is to use the workflow with the highest average performance, but this can be suboptimal because different workflows may succeed on different instances. We study a portfolio-and-selector paradigm in which a firm runs multiple workflow executions and selects the final answer after observing their outputs. Additional executions may uncover correct answers that the best standalone workflow misses, but they consume compute and introduce plausible distractors that complicate final selection. We formulate this as a workflow portfolio problem in which the firm jointly chooses run size and allocation across workflow types. We summarize selector quality through an odds-lift index and derive sharp bounds on the value of workflow variety. For finite workflow pools, we develop exact formulations, linear programming relaxations, randomized rounding procedures, and computable performance certificates. For large implicit workflow classes, we derive a finite-dimensional dual and an ellipsoid method using a pricing oracle to identify workflows with high weighted accuracy net of recurring compute cost. Under a weak condition, the method obtains a near-optimal solution to the relaxation with polynomially many oracle calls. We evaluate the framework on three datasets: ABCD, Schema-Guided Dialogue, and HotpotQA. Relative to the best standalone workflow, portfolio optimization improves held-out selector accuracy by 3.1, 7.5, and 0.9 percentage points, respectively. Dual-guided workflow generation adds 3.5 points on ABCD and 24.1 on HotpotQA, with no additional gain on Schema-Guided Dialogue.
Chinese Translation
自主智能系统通常通过多种工作流来处理相同的任务，这些工作流在推理策略、验证结构和计算成本上各不相同。一个自然的部署策略是使用平均性能最高的工作流，但这可能并非最佳选择，因为不同的工作流可能在不同的实例上表现出成功。我们研究了一种组合-选择者范式，其中公司运行多个工作流执行，并在观察其输出后选择最终答案。额外的执行可能会揭示最佳独立工作流遗漏的正确答案，但它们会消耗计算资源，并引入可能的干扰因素，从而使最终选择变得复杂。我们将其表述为一个工作流组合问题，其中公司共同选择运行规模和在工作流类型之间的分配。我们通过赔率提升指数来总结选择者的质量，并推导出工作流多样性的价值的严格界限。对于有限的工作流池，我们开发了精确的公式、线性规划松弛、随机舍入程序和可计算的性能证明。对于大型隐式工作流类别，我们推导出有限维对偶及使用定价Oracle的椭球法，以识别具有高加权准确度的工作流，扣除重复计算成本。在一个弱条件下，该方法通过多项式数量的Oracle调用获得松弛的近似最优解。我们在三个数据集上评估该框架：ABCD、Schema-Guided Dialogue和HotpotQA。相较于最佳独立工作流，组合优化分别提高了保留选择器的准确率3.1、7.5和0.9个百分点。双重引导的工作流生成在ABCD上增加了3.5点，在HotpotQA上增加了24.1点，而在Schema-Guided Dialogue上没有额外增益。
cs.AI / 37 / 2609.18128
Symbolic Temporal Supervision of LLM Agents Using Contracts
使用合同对大型语言模型代理进行符号时间监督
Abstract
Large language model (LLM) agents augmented by tools can automate complex, multi-step tasks, such as web navigation, code generation, and workflow orchestration, by acting on external systems through tool calls. However, hallucinations, distributional instability, and adversarial manipulations in LLMs, and the irreversible consequences of certain tool calls can lead to harmful outcomes. Existing safeguards either grade recorded trajectories post hoc with stochastic LLM judges or block unsafe actions one call at a time, and no single deterministic artifact supports both roles. We present ContrAgent, a contract-based framework for symbolic temporal supervision of LLM agents. ContrAgent captures an agent's behavior as a sequence of tool calls and formalizes it as a trace over a fixed set of checkable predicates. It then specifies required behaviors using assume-guarantee contracts in linear temporal logic over finite traces (LTLf). Each contract is compiled to a deterministic finite automaton (DFA) that serves two roles: gating agent actions online and evaluating recorded traces offline. A contract library, acting as a reusable knowledge base, is maintained independently of the agent's model and can be applied across different agents within the same task domain. We show the effectiveness of our approach on four benchmarks spanning both roles, where ContrAgent matches state-of-the-art LLM-judge and rule-based guardrail baselines while producing deterministic, reproducible verdicts and, in the online mode, orders-of-magnitude lower per-call latency.
Chinese Translation
增强工具的大型语言模型（LLM）代理能够通过对外部系统的工具调用，自动化复杂的多步骤任务，如网页导航、代码生成和工作流协调。然而，LLM中的幻觉、分布不稳定性和对抗性操控，以及某些工具调用的不可逆后果，可能导致有害结果。现有的安全措施要么在事后使用随机LLM评判者对记录的轨迹进行评分，要么逐个阻止不安全的操作，而没有任何单一的确定性工具同时支持这两种角色。我们提出了ContrAgent，这是一个基于合同的框架，用于对LLM代理进行符号时间监督。ContrAgent将代理的行为捕捉为一系列工具调用，并将其形式化为一组固定可检查谓词上的轨迹。然后，它使用线性时序逻辑（LTLf）中的假设-保证合同来指定所需的行为。每个合同被编译为一个确定性有限自动机（DFA），该自动机承担两个角色：在线控制代理的行为和离线评估记录的轨迹。合同库作为一个可重用的知识库，与代理的模型独立维护，并可以在同一任务领域的不同代理之间应用。我们展示了我们的方法在四个基准测试中的有效性，这些基准测试涵盖了这两种角色，其中ContrAgent与最先进的LLM评判者和基于规则的保护措施基线相匹配，同时产生确定性、可重现的裁决，并且在在线模式下，每次调用的延迟低于几个数量级。
cs.AI / 38 / 2609.18163
Time-Aligned Evolving Concept Graphs for Scientific Relation Forecasting
用于科学关系预测的时间对齐演变概念图
Abstract
Forecasting scientific relations can guide discovery by identifying promising connections before they emerge. Existing approaches often model concept semantics and graph structure separately or summarize semantics over coarse historical snapshots, leaving semantic representations potentially misaligned with rapidly evolving graph evidence. We propose a time-aligned evolving concept graph framework that jointly models semantic and structural evolution. Its core idea is to treat dated papers as shared update events, reconstructing semantic and structural states from the same publication history through each prediction time. Pair-level fusion combines these states to forecast first co-occurrence, relation formation, and conditional relation type. Holding architecture and training fixed, refreshing context alongside graph updates improves mean relation AUPRC by 16.6% over frozen context. On a graph built from 187,848 papers with 270,687 concepts and 7.45 million co-occurrence links, the complete framework improves mean relation AUROC from 0.9290 for the strongest evaluated baseline to 0.9722, with mean population-weighted AUPRC 0.005778.
Chinese Translation
科学关系的预测可以通过在关系出现之前识别有前景的连接来指导发现。现有的方法通常将概念语义和图结构分开建模，或者在粗略的历史快照上总结语义，这可能导致语义表示与快速演变的图证据不匹配。我们提出了一种时间对齐的演变概念图框架，该框架联合建模语义和结构的演变。其核心思想是将过时的论文视为共享的更新事件，通过每个预测时间从相同的出版历史重建语义和结构状态。对偶级别的融合将这些状态结合起来，以预测首次共现、关系形成和条件关系类型。在保持架构和训练不变的情况下，随着图更新刷新上下文，使得平均关系的AUPRC提高了16.6%。在由187,848篇论文构建的图中，包含270,687个概念和745万个共现链接，完整框架将平均关系的AUROC从最强基线的0.9290提高到0.9722，平均加权AUPRC为0.005778。
cs.AI / 39 / 2609.18182
WFM: Wiki Foundation Model for Complex Agentic Reasoning
WFM：用于复杂代理推理的维基基础模型
Abstract
Real-world agents fundamentally require persistent non-parametric knowledge for dynamic reasoning, i.e., long-term memory and retrieval-augmented generation. While graphs have shown reliable advantages in providing structured evidence, the sparse graph representations naturally restrict machine readability and semantic density required for complex agentic workflows. Driven by this limitation, the entire industry is witnessing a paradigm shift from traditional sparse graphs to LLM Wiki, an agent-native knowledge representation that couples dense document contexts with markdown files containing multi-layered topological linkages. However, parameterizing such rich semantics is challenging to encode dense textual contexts using traditional sparse graph embeddings. Moreover, learning LLM Wiki with existing graph encoders could overwhelm distributed system overheads that hinder deployment in large-scale commercial scenarios. To this end, we propose a novel paradigm Wiki Foundation Model, i.e., WFM, tailored for scalable, agent-native representation and retrieval. Specifically, (i) we formalize a Wiki Graph schema that seamlessly bridges fine-grained structures with dense contexts, maintaining explicit topologies alongside continuous semantics; (ii) A query-conditioned attentive aggregation is tailored for rich wiki message passing and explicit attention variance regularization; (iii) We engineer an infrastructural NCCL boundary exchange protocol that hoists static partition indices and leverages fixed-shape GPU-to-GPU collectives, bypassing CPU serialization and memory copy overheads. Extensive evaluations across five long-term agent memory and multi-hop reasoning benchmarks demonstrate the remarkable performance of WFM, while achieving a 10.5 times training acceleration on distributed clusters.
Chinese Translation
现实世界的代理本质上需要持久的非参数知识以进行动态推理，即长期记忆和检索增强生成。尽管图在提供结构化证据方面显示出可靠的优势，但稀疏图表示自然限制了机器可读性和复杂代理工作流所需的语义密度。受此限制的驱动，整个行业正在经历一个从传统稀疏图到LLM Wiki的范式转变，后者是一种代理原生的知识表示，将密集的文档上下文与包含多层拓扑链接的Markdown文件结合在一起。然而，使用传统的稀疏图嵌入来参数化如此丰富的语义在编码密集文本上下文时具有挑战性。此外，使用现有图编码器学习LLM Wiki可能会超出分布式系统的开销，从而阻碍其在大规模商业场景中的部署。为此，我们提出了一种新颖的范式——维基基础模型（Wiki Foundation Model），即WFM，专为可扩展的代理原生表示和检索而设计。具体而言，(i) 我们形式化了一个维基图架构，能够无缝连接细粒度结构与密集上下文，保持显式拓扑与连续语义的并存；(ii) 针对丰富的维基消息传递和显式注意力方差正则化，我们定制了一种查询条件的注意聚合；(iii) 我们设计了一种基础设施NCCL边界交换协议，提升静态分区索引，并利用固定形状的GPU到GPU集合，绕过CPU序列化和内存复制开销。在五个长期代理记忆和多跳推理基准上的广泛评估表明，WFM表现出卓越的性能，同时在分布式集群上实现了10.5倍的训练加速。
cs.AI / 40 / 2609.18249
Re2A: Situated Conversational Recommendation via Rubric-based Preference Reasoning and Alignment
Re2A：基于评分标准的偏好推理与对齐的情境对话推荐
Abstract
Real-world recommendation scenarios are commonly grounded in shared physical environments during user-recommender interactions. This motivates situated conversational recommendation (SCR), a complex task requiring recommender assistants to jointly reason over dialogue history, co-observed scenes, and in-scene item attributes. However, current approaches struggle with this setting due to two intertwined challenges: accurately understanding situated user preferences throughout the conversation and generating responses that simultaneously satisfy user needs and grounded situations. To this end, we propose Re2A, a framework that formulates SCR as a structured reason-then-align process. We introduce rubric-based preference reasoning, which uses automated rubrics to guide the model toward producing explicit preference states. Based on these states, we propose a preference-conditioned optimization to align response generation with dual objectives: user preference satisfaction and situation consistency. Extensive experiments on two SCR datasets demonstrate that Re2A consistently outperforms state-of-the-art methods, delivering more precise, context-aware conversational recommendations. Our code is available at https://github.com/DongdingLin/Re2A.
Chinese Translation
现实世界的推荐场景通常在用户与推荐者的互动中建立在共享的物理环境中。这促使了情境对话推荐（SCR）的研究，这是一项复杂的任务，需要推荐助手共同推理对话历史、共同观察的场景以及场景中物品的属性。然而，当前的方法在这一环境中面临两个相互交织的挑战：在整个对话过程中准确理解情境用户偏好，以及生成同时满足用户需求和情境一致性的响应。为此，我们提出了Re2A，一个将SCR形式化为结构化的推理-对齐过程的框架。我们引入了基于评分标准的偏好推理，利用自动化评分标准引导模型产生明确的偏好状态。基于这些状态，我们提出了一种偏好条件优化方法，以将响应生成与双重目标对齐：用户偏好满足和情境一致性。在两个SCR数据集上的大量实验表明，Re2A始终优于最先进的方法，提供更精确、具有上下文意识的对话推荐。我们的代码可在https://github.com/DongdingLin/Re2A获取。
cs.AI / 41 / 2609.18262
REPAIR: Resolving Long-Tail Confusion in Scientific Retrievers via Fact-Verified Iterative Refinement
REPAIR：通过事实验证的迭代精炼解决科学检索中的长尾混淆
Abstract
Precise retrieval of scientific information is fundamentally constrained by long-tailed concepts and high fact-sensitivity of scientific corpora. These challenges often limit the effectiveness of dense retrievers and hallucination-prone LLM augmentation. To address this, we present REPAIR, a self-evolving data augmentation framework for scientific dense retrievers. REPAIR iteratively synthesizes training data to address knowledge gaps by cycling through diagnosis of long-tail concepts, API-guided evidence expansion, and differentiation via hard negative mining. This process effectively grounds retrieval in factual reality to resolve fine-grained distinctions. Extensive experiments demonstrate that REPAIR significantly outperforms 19 strong baselines on nine materials science and biomedical benchmarks. Our work highlights that diagnosing and factually augmenting data to long-tail deficits is essential for robust scientific retrieval.
Chinese Translation
科学信息的精确检索在根本上受到长尾概念和科学语料库高事实敏感性的限制。这些挑战常常限制了密集检索器的有效性以及容易产生幻觉的LLM（大型语言模型）增强。为了解决这一问题，我们提出了REPAIR，一个自我演化的数据增强框架，旨在提升科学密集检索器的性能。REPAIR通过循环诊断长尾概念、API引导的证据扩展以及通过困难负样本挖掘进行区分，迭代合成训练数据以填补知识空白。这个过程有效地将检索与事实现实相结合，以解决细粒度的区分问题。大量实验表明，REPAIR在九个材料科学和生物医学基准测试中显著优于19个强基线。我们的研究强调，诊断和事实增强长尾缺失的数据对于稳健的科学检索至关重要。
cs.AI / 42 / 2609.18270
BENCHCOMPASS: From Scores to Signals for Training and Harness Decisions in Payment-Domain LLMs
BENCHCOMPASS：从分数到信号，用于支付领域大语言模型的训练和决策
Abstract
Payment operations are a critical financial infrastructure, but the value of large language models in this domain remains unclear because payment rules change quickly, evidence is fragmented, and decisions depend on transaction state, participant role, region, and payment rail. Existing benchmarks do not isolate whether failures come from missing payment-rule knowledge, poor use of supplied evidence, or brittleness under imperfect harness inputs. We introduce BENCHCOMPASS, a payment-domain benchmark whose construction pipeline builds scenario-grounded tasks from typed evidence packs, applies LLM-based quality checks, creates task-input attack variants, and reserves final item admission for domain experts. The release contains an expert-reviewed Pro benchmark covering payment knowledge, context-grounded scenario reasoning, and Attacked Open robustness, plus a lower-assurance Normal pool for inspection and future curation. Across 16 model variants, BENCHCOMPASS shows qualitatively different failure modes: missing parametric payment knowledge, incomplete reasoning over supplied rules, and failure to reject plausible but invalid workflows. The benchmark remains unsaturated: the best frontier model reaches 89.6% on Open Context-Grounded Reasoning and 81.7% under attacked inputs, while a representative 32B open-weight model reaches 69.8% and 42.6%. Benchmark data and code are available at https://github.com/ant-intl/BenchCompass.
Chinese Translation
支付操作是关键的金融基础设施，但在这一领域大语言模型的价值仍不明确，因为支付规则变化迅速，证据碎片化，且决策依赖于交易状态、参与者角色、地区和支付渠道。现有的基准测试并未区分失败是由于缺乏支付规则知识、对提供证据的使用不当，还是在不完美的输入下的脆弱性。我们引入了BENCHCOMPASS，这是一个支付领域的基准，其构建流程从类型化证据包中构建场景基础任务，应用基于LLM的质量检查，创建任务输入攻击变体，并将最终项目的录取保留给领域专家。该发布包含一个经过专家审查的专业基准，涵盖支付知识、基于上下文的场景推理和攻击开放鲁棒性，以及一个用于检查和未来整理的低保证常规池。在16个模型变体中，BENCHCOMPASS显示出定性不同的失败模式：缺失参数支付知识、对提供规则的推理不完整，以及未能拒绝合理但无效的工作流。该基准仍未饱和：最佳前沿模型在开放上下文基础推理中达到89.6%，在攻击输入下达到81.7%，而一个代表性的32B开放权重模型则分别达到69.8%和42.6%。基准数据和代码可在https://github.com/ant-intl/BenchCompass获取。
cs.AI / 43 / 2609.18272
Who Audits Whom, on What Substrate, with What Evidence? An Independence-Graded Audit Protocol for Agentic AI
谁审计谁，基于什么基础，凭借什么证据？一种针对自主人工智能的独立性分级审计协议
Abstract
Agentic AI systems plan, invoke tools and act with limited supervision; they are now both the subject of audits and, increasingly, the auditor. Independence, the foundation of assurance,is still applied to them as a binary. We argue that it must be graded along three orthogonal axes: principal independence (who controls the auditor), substrate independence (an auditor sharing the auditee's foundation-model family, toolchain or guardrails fails with it) and evidence independence (whether evidence is attestable rather than self-reported). Each axis has precedent; the contribution is to grade all three on a single audit, aggregate them by the weakest link, and apply the same rubric when the auditor is itself an agent. We give the model a formal basis by transplanting the beta-factor model of common-cause failure from reliability engineering, a seven-step protocol whose outputs a third party can verify, a structural detectability analysis of a procurement-controls agent audited at three grades, and a Monte Carlo study of the model in which a conventional internal audit of an agent-a real audit team, a second agent, provider logsp-surfaces 5.9% of the faults it could in principle see and none at all in half the fault classes. We map the triple to the EU AI Act as amended, ISO/IEC 42006, UK public-sector risk-management guidance and audit-regulator practice.
Chinese Translation
自主人工智能系统在有限监督下进行规划、调用工具和行动；它们现在既是审计的对象，也越来越多地成为审计者。独立性作为保障的基础，仍然被视为二元的。我们认为，独立性必须沿着三个正交轴进行分级：主体独立性（谁控制审计者）、基础独立性（共享审计对象基础模型家族、工具链或保护措施的审计者会随之失败）和证据独立性（证据是否可验证而非自我报告）。每个轴都有先例；我们的贡献是将所有三个轴在单次审计中进行分级，通过最弱环节进行汇总，并在审计者本身也是代理的情况下应用相同的标准。我们通过从可靠性工程中移植常因失效的贝塔因子模型，为该模型提供了正式基础，这是一种七步协议，其输出可以由第三方验证，结构可检测性分析针对在三个等级上审计的采购控制代理，以及对该模型的蒙特卡洛研究，其中对一个代理的常规内部审计——一个真实的审计团队、第二个代理、提供者日志——仅发现其原则上能够看到的故障的5.9%，而在一半的故障类别中完全没有发现。我们将这一三重映射到修订后的欧盟人工智能法案、ISO/IEC 42006、英国公共部门风险管理指导和审计监管实践。
cs.AI / 44 / 2609.18278
Building Trust in Artificial Intelligence: A Necessity for Railway Applications
在人工智能中建立信任：铁路应用的必要性
Abstract
Artificial Intelligence (AI) is currently only applied to non-safety critical applications due to the strict standards and regulations for railway industries. We propose to review the three main fields necessary to increase trust in data science and AI algorithms and reach compliance: robustness, Operational Design Domain (ODD), and explainability. Robustness is the ability of an AI system to maintain its level of performance under any circumstances (ISO24029). ODDs allow the explicit definition of operating conditions under which a system is intended to operate, according to the recently published DIN DKE SPEC 99004. Explainability is the property of an AI system to express important factors influencing the AI system results in a way that humans can understand. Those 3 domains of research are already well investigated by nonrailway actors, with algorithms and methods ready to use for railway applications. A system view is necessary to ensure all trustworthy requirements interact continuously in a safe MLOps environment thereby fostering acceptance from regulators, operators and the public. Beyond safeguarding safety-critical applications, we aim to show that fostering deep trust in AI, as now required by regulatory frameworks worldwide, will unlock its full potential and transform the pace of adoption across mission-critical domains.
Chinese Translation
由于铁路行业严格的标准和法规，人工智能（AI）目前仅应用于非安全关键的领域。我们建议回顾增强数据科学和AI算法信任所需的三个主要领域，以达到合规性：鲁棒性、操作设计域（Operational Design Domain, ODD）和可解释性。鲁棒性是指AI系统在任何情况下维持其性能水平的能力（ISO24029）。ODD允许明确界定系统预期操作的工作条件，依据最近发布的DIN DKE SPEC 99004。可解释性是指AI系统以人类能够理解的方式表达影响其结果的重要因素的特性。这三个研究领域已经被非铁路行业的参与者充分研究，相关算法和方法已准备好用于铁路应用。需要采取系统视角，以确保所有可信赖的要求在安全的MLOps环境中持续互动，从而促进监管机构、运营商和公众的接受度。除了保障安全关键应用外，我们旨在展示，培养对AI的深度信任（这是全球监管框架现在所要求的）将释放其全部潜力，并加快在关键任务领域的采用速度。
cs.AI / 45 / 2609.18283
Where Should Agents Live? Energy-Memory Characterization of Agentic AI for the Edge-Cloud Continuum
代理智能应当驻留何处？边缘-云连续体中代理人工智能的能量-内存特征化
Abstract
As telecommunication networks evolve toward autonomous 5G-Advanced and 6G operations, agentic artificial intelligence (AI) workflows, where large language models (LLMs) execute multi-step reasoning, invoke diagnostic tools, retrieve domain knowledge, and coordinate across agent teams, are increasingly embedded across the edge-cloud continuum. While the biological brain accomplishes complex cognition on an exceptionally modest metabolic power budget of approximately 20W contemporary LLMs are profoundly energy- and memory-intensive, making sustainable lifecycle orchestration a critical operational priority. However, existing AI lifecycle metrics evaluate only isolated, single-model inferences or overlook multi-agent execution graphs entirely. Consequently, network operators lack foundational models to determine whether distributed agent communication incurs meaningful energy costs and where across edge-cloud tiers agent teams should physically reside. To address this gap, we introduce agentic-eCAL, generalizing the Energy Cost of AI Lifecycle (eCAL) metric to directed multi-agent workflows by coupling a closed-form two-rate single-call energy model (compute-bound prefill and memory-bound decode) with 7-layer OSI data transport. Grounded in hundreds of GPU benchmark configurations on NVIDIA A100 and H100, 16 open-weight models and 8 orchestration topologies, we validate components of the metric and study workflow placement implications. Our findings demonstrate that inter-agent text transport incurs 0.25% of workflow energy across 5G RAN, metro, and optical links. Therefore in edge-cloud agent placement the dominant energy cost of distribution is often not the transmission of inter-agent text itself, but the additional inference and context processing induced by that communication.
Chinese Translation
随着电信网络向自主的5G-Advanced和6G操作演进，代理人工智能（AI）工作流逐渐嵌入边缘-云连续体中，其中大型语言模型（LLMs）执行多步骤推理、调用诊断工具、检索领域知识并协调代理团队。尽管生物大脑在约20W的极低代谢功率预算下完成复杂的认知，但当代LLMs在能量和内存方面的需求极高，使得可持续生命周期管理成为一个关键的操作优先事项。然而，现有的AI生命周期指标仅评估孤立的单模型推理，或完全忽视多代理执行图。因此，网络运营商缺乏基础模型来确定分布式代理通信是否会产生显著的能量成本，以及代理团队应在边缘-云层次中物理驻留的位置。为了解决这一问题，我们引入了代理-eCAL，将人工智能生命周期的能量成本（eCAL）指标推广到有向多代理工作流，通过结合一个封闭形式的双速单次调用能量模型（计算绑定的预填充和内存绑定的解码）与7层OSI数据传输。基于数百个NVIDIA A100和H100的GPU基准配置、16个开放权重模型和8种编排拓扑，我们验证了该指标的组成部分，并研究了工作流放置的影响。我们的研究结果表明，在5G RAN、城域网和光纤链路中，代理间文本传输占工作流能量的0.25%。因此，在边缘-云代理放置中，分发的主要能量成本往往不是代理间文本本身的传输，而是由该通信引发的额外推理和上下文处理。
cs.AI / 46 / 2609.18286
What Counts as Strategic Reasoning? A Systematic Mapping of Chess Research on Humans, Engines, and Language Models
什么算作战略推理？对人类、引擎和语言模型的国际象棋研究的系统性映射
Abstract
Chess has long served as a model domain for studying search, expertise, decision-making, and artificial intelligence. The emergence of large language models (LLMs) has renewed the relevance of chess as a controlled environment for investigating strategic reasoning and comparing human and artificial decision-making. We present a systematic mapping study of recent research spanning human players, classical chess engines, neural and reinforcement-learning systems, LLMs, and hybrid approaches. The final map comprises 84 core study families, classified according to agent type, strategic-reasoning stages, and evaluation dimensions. The map reveals a literature strongly concentrated on situation assessment, evaluation, and action selection, while explicit planning, explanation, metacognition, and human--AI collaboration remain less explored. LLM research places particular emphasis on state representation and generalization, whereas grounded explanation appears more frequently in hybrid approaches combining language models with engines, expert knowledge, or other external structures. Two distinctions emerge that the map aggregates rather than resolves: hybrid systems differ in where and when heterogeneous capabilities combine, and evaluations that show improved human performance do not thereby establish human--AI synergy. We propose both as extensions of the mapping framework. We argue that chess provides a useful bridge between cognitive and computational perspectives on strategic reasoning, and identify explicit planning, grounded and faithful explanation, metacognitive calibration, and human--AI complementarity as directions for future research.
Chinese Translation
国际象棋长期以来作为研究搜索、专业知识、决策和人工智能的模型领域。大型语言模型（LLMs）的出现重新增强了国际象棋作为一个受控环境的相关性，以调查战略推理并比较人类与人工决策。我们呈现了一项系统性映射研究，涵盖了人类玩家、经典国际象棋引擎、神经网络和强化学习系统、LLMs以及混合方法的最新研究。最终的映射图包含84个核心研究家族，按照代理类型、战略推理阶段和评估维度进行分类。该映射揭示了文献在情境评估、评估和行动选择方面的强烈集中，而显式规划、解释、元认知和人类与人工智能的协作则相对较少探索。LLM研究特别强调状态表示和泛化，而基于实证的解释在将语言模型与引擎、专家知识或其他外部结构结合的混合方法中出现得更为频繁。映射中出现了两个区别，这些区别被聚合而非解决：混合系统在异构能力结合的时间和地点上有所不同，而显示人类表现改善的评估并不因此确立人类与人工智能的协同。我们建议将这两者作为映射框架的扩展。我们认为国际象棋为战略推理的认知与计算视角之间提供了有用的桥梁，并确定显式规划、基于实证的和忠实的解释、元认知校准以及人类与人工智能的互补性作为未来研究的方向。
cs.AI / 47 / 2609.18328
Visual Compliance via Executable Safety Rule Entailment
通过可执行安全规则推理实现视觉合规性
Abstract
Recent advances in LLMs and VLMs have enabled safety systems to reason beyond simple risk patterns toward more contextual and semantic safety concerns. However, as risk patterns continue to evolve and safety rules become more complex, existing training-based end-to-end safeguards face persistent challenges in adaptability and explainable reasoning over complex safety rules. To address these challenges, we propose GuardEn (Guarding by Safety Rule Entailment), an executable safeguard framework that decomposes safety policies into atomic propositions through Safety-Rule Compilation, modeling their composition as executable code. At test time, Scene-Grounded Execution instantiates these atomic propositions with contextual visual information derived from scene graphs, enabling rule-grounded and interpretable safety reasoning. Experiments on SafetyVisionBench demonstrate the effectiveness of programmable safeguard for complex visual safety assessment, achieving an average improvement of 9.8 F1 points over the strongest baseline.
Chinese Translation
近期在大型语言模型（LLMs）和视觉语言模型（VLMs）方面的进展，使得安全系统能够超越简单的风险模式，关注更具上下文和语义的安全问题。然而，随着风险模式的不断演变和安全规则的日益复杂，现有的基于训练的端到端安全防护在适应性和对复杂安全规则的可解释推理方面面临持续挑战。为了解决这些问题，我们提出了GuardEn（通过安全规则推理进行保护），这是一个可执行的安全防护框架，通过安全规则编译将安全政策分解为原子命题，并将其组合建模为可执行代码。在测试时，场景基础执行（Scene-Grounded Execution）利用场景图中提取的上下文视觉信息实例化这些原子命题，从而实现基于规则的可解释安全推理。在SafetyVisionBench上的实验表明，该可编程安全防护在复杂视觉安全评估中的有效性，平均提高了9.8个F1分数，相较于最强基线。
cs.AI / 48 / 2609.18346
Faithful yet Collusive: Why Chain-of-Thought Monitoring Cannot Detect Collusion in LLM Pricing Agents under Oligopolistic Competition
忠实但共谋：为何链式思维监控无法检测寡头竞争下大型语言模型定价代理的共谋行为
Abstract
Large language models (LLM) deployed as autonomous pricing agents may sustain supracompetitive prices through tacit coordination. We develop a causal graph divergence framework that separately measures structural faithfulness and intent faithfulness of LLM pricing agents in Bertrand competition. Across nine LLMs under duopoly and triopoly conditions, collusive behavior and chain-of-thought (CoT) faithfulness dissociate along both dimensions: the most collusive model accurately reports cooperative intent yet reasons structurally unfaithfully, while the most structurally faithful model sustains supra-Nash pricing under both market structures. These findings establish that CoT monitoring alone cannot serve as a standalone safeguard against algorithmic collusion.
Chinese Translation
作为自主定价代理的大型语言模型（LLM）可能通过默契协调维持超竞争性价格。我们开发了一种因果图差异框架，分别测量在伯特兰竞争中的LLM定价代理的结构忠实性和意图忠实性。在双头垄断和三头垄断条件下的九个LLM中，共谋行为和链式思维（CoT）忠实性在这两个维度上表现出分离：最具共谋性的模型准确报告合作意图，但在结构上不忠实，而最具结构忠实性的模型在这两种市场结构下维持超纳什定价。这些发现表明，仅仅依靠CoT监控无法作为防止算法共谋的独立保障。
cs.AI / 49 / 2609.18357
Market Signal Injection: Adversarial Context Manipulation of LLM Pricing Agents
市场信号注入：对大型语言模型定价代理的对抗性上下文操控
Abstract
Large language model (LLM) pricing agents may respond to how market data is presented, even when its numerical values remain unchanged. We introduce market signal injection (MSI), an attack that manipulates numerical formatting, competitor ordering, or qualitative market commentary without issuing explicit instructions. We evaluate nine open-weight models in simulated Bertrand duopoly and triopoly markets and three proprietary models in duopoly markets. Sentiment-based attacks produce the largest behavioral shifts, which propagate to other firms and alter profits and consumer surplus. Susceptibility varies across model families, and larger models are not consistently more robust. Matched neutral-text controls and a rule-based agent support a framing-based account of these shifts under the fixed demand parameters of our simulation. Episode-held-out probes distinguish baseline from attacked activations in all eleven re-evaluated model--condition pairs: linear AUC is 1.00 and MLP AUC ranges from 0.93 to 0.99. This separability does not by itself identify harmful pricing decisions. Input canonicalization removes the tested sentiment attacks, while decision boundary anchoring, which combines prompt constraints with output projection, provides partial mitigation under the tested adaptive attacks. These results identify data presentation as an attack surface for LLM pricing agents and motivate defenses that account for interactions among agents.
Chinese Translation
大型语言模型（LLM）定价代理可能会对市场数据的呈现方式作出反应，即使其数值保持不变。我们提出了市场信号注入（Market Signal Injection, MSI），这是一种攻击方式，通过操控数值格式、竞争者排序或定性市场评论，而无需发出明确指令。我们在模拟的贝尔特朗双头垄断和三头垄断市场中评估了九个开放权重模型，并在双头垄断市场中评估了三个专有模型。基于情感的攻击产生了最大的行为变化，这些变化会传播到其他公司，并改变利润和消费者剩余。模型家族之间的易感性存在差异，较大的模型并不总是更具鲁棒性。匹配的中性文本对照和基于规则的代理支持了在我们模拟的固定需求参数下，这些变化的框架基础解释。保留的情节探测器在所有十一对重新评估的模型-条件对中区分了基线激活与攻击激活：线性AUC为1.00，MLP AUC范围为0.93到0.99。这种可分离性本身并不能识别有害的定价决策。输入规范化消除了测试的情感攻击，而决策边界锚定（decision boundary anchoring），将提示约束与输出投影相结合，在测试的自适应攻击下提供了部分缓解。这些结果表明数据呈现是LLM定价代理的攻击面，并激励了考虑代理之间相互作用的防御措施。
cs.AI / 50 / 2609.18366
Bad Genius: Counterfactual-Guided Harness Evolution Beyond Task-Specific Shortcuts
坏天才：超越任务特定捷径的反事实引导的马具演化
Abstract
Reliable agent evaluation is complicated by automatic harness optimization, which repeatedly uses a released benchmark $B_{\mathrm{rel}}$ to guide a Proposer that edits prompts, memory, retrieval, tools, and control code around a fixed target agent. Task holdout varies semantic tasks but leaves the benchmark protocol fixed, so a "bad genius" Proposer can produce a cheating harness whose released-benchmark gain depends on a benchmark-wide shortcut. We introduce Counterfactual Harness Search and Evolution (CHASE), which casts harness evolution as constraint generation over validity-preserving benchmark counterfactuals. After each Proposer update, a Challenger searches for an executable protocol transformation with large gain destruction. A validity firewall checks that task semantics are preserved, while a confirmation set determines whether the counterfactual enters a finite archive. We formalize an exact shortcut-neutralized benchmark $B_0$ and establish statistical guarantees linking finite counterfactual archives to $B_0$ and characterizing sequential Challenger search. We evaluate CHASE on a synthetic benchmark and on OfficeQA, where CHASE retains strong released-benchmark gains while substantially reducing gain destruction under valid protocol changes.
Chinese Translation
可靠的代理评估因自动马具优化而变得复杂，该优化反复使用已发布基准 $B_{ ext{rel}}$ 来指导一个编辑提示、记忆、检索、工具和控制代码的提议者，围绕一个固定的目标代理。任务保留变换语义任务，但基准协议保持不变，因此一个“坏天才”提议者可以产生一个作弊马具，其已发布基准增益依赖于基准范围内的捷径。我们引入了反事实马具搜索与演化（Counterfactual Harness Search and Evolution，CHASE），将马具演化视为对保持有效性的基准反事实的约束生成。在每次提议者更新后，一个挑战者搜索具有大增益破坏的可执行协议变换。有效性防火墙检查任务语义是否得到保留，而确认集则确定反事实是否进入有限档案。我们形式化了一个精确的捷径中和基准 $B_0$，并建立了将有限反事实档案与 $B_0$ 关联的统计保证，描述了顺序挑战者搜索。我们在一个合成基准和 OfficeQA 上评估 CHASE，结果表明 CHASE 在保持强大的已发布基准增益的同时，显著减少了在有效协议变更下的增益破坏。
cs.AI / 51 / 2609.18394
Cultural Competence in Context: A Large Language Model Passes the Turing Test in Finland
语境中的文化能力：大型语言模型在芬兰通过图灵测试
Abstract
We report the results of a Turing Test conducted in Finland in the Finnish language. Because languages and cultural contexts are unevenly represented in LLM training data, we expected the model (ChatGPT 5.2) to perform worse in a Finnish-language Turing Test than in previously studied English-language US contexts. We also present model-generated role prompting as a replicable technique for conducting comparative LLM-based Turing Tests designed to improve construct validity. Contrary to our expectations, the LLM passed the Finnish Turing Test. A prominent source of error was participants' reliance on linguistic cues, particularly colloquial Finnish, as markers of human authorship. We reframe the Turing Test from a test of intelligence to a comparative method for examining whether an AI system can display credible membership in a particular social world. Because its outcome reflects model capabilities, prompted identity, insider competence among human participants, and their AI literacy, the method provides a useful probe of the human-machine boundary across domains.
Chinese Translation
我们报告了在芬兰以芬兰语进行的图灵测试的结果。由于语言和文化背景在大型语言模型（LLM）训练数据中的代表性不均，我们预期该模型（ChatGPT 5.2）在芬兰语图灵测试中的表现会比在之前研究的英语美国背景中更差。我们还提出了模型生成的角色提示作为一种可复制的技术，用于进行旨在提高构念效度的比较性LLM基础图灵测试。与我们的预期相反，该LLM通过了芬兰图灵测试。一个显著的错误来源是参与者对语言线索的依赖，特别是口语芬兰语，作为人类创作的标志。我们将图灵测试重新定义为一种比较方法，用于检验一个人工智能系统是否能够在特定社会世界中展现出可信的成员资格。由于其结果反映了模型能力、提示身份、人类参与者的内部能力以及他们的人工智能素养，该方法为跨领域的人机边界提供了有用的探测手段。
cs.AI / 52 / 2609.18431
HPOQuest: A Rare-Disease Diagnostic Agent Using Active Phenotype Acquisition
HPOQuest：一种基于主动表型获取的罕见疾病诊断工具
Abstract
More than 300 million people worldwide are affected by one of over 7,000 known rare diseases, yet diagnosis remains difficult because patients initially present with incomplete and heterogeneous phenotypes. We present HPOQuest, a training-free framework for sequential phenotype acquisition in rare-disease diagnosis. Starting from a small set of observed patient phenotypes, HPOQuest maintains a probabilistic disease ranking and iteratively selects informative follow-up questions to support clinicians during patient assessment. Confirmed phenotypes update the disease ranking, while all responses update the candidate question set. Across four benchmark cohorts, HPOQuest substantially improves diagnosis from sparse initial phenotypes, with gains of up to 30% points at Recall@1 and 45% points at Recall@5. These results demonstrate that sequential phenotype acquisition can substantially improve rare-disease diagnosis from limited initial clinical evidence.
Chinese Translation
全球有超过3亿人受到7000多种已知罕见疾病的影响，但由于患者最初呈现出不完整和异质的表型，诊断仍然困难。我们提出了HPOQuest，这是一种用于罕见疾病诊断的无训练框架，旨在进行顺序表型获取。HPOQuest从一小组观察到的患者表型开始，维护一个概率性疾病排名，并迭代选择信息丰富的后续问题，以支持临床医生在患者评估过程中的决策。确认的表型会更新疾病排名，而所有的响应则更新候选问题集。在四个基准队列中，HPOQuest显著改善了从稀疏初始表型进行的诊断，Recall@1的提升达到30个百分点，Recall@5的提升达到45个百分点。这些结果表明，顺序表型获取可以显著改善基于有限初始临床证据的罕见疾病诊断。
cs.AI / 53 / 2609.18435
WetRobo: A Reproducible Robot Kit for Coding Agents in Biological Laboratories
WetRobo：用于生物实验室编码代理的可重复机器人套件
Abstract
Automating biological research requires general-purpose, reproducible robot systems that allow individual wet-lab researchers to delegate robot tasks without performing teleoperation or neural-network training. Vision-language-action policies have been proposed for general-purpose arms, but can lose performance when their operating environment changes. We therefore built WetRobo, a robot kit that can readily transfer between laboratories. It consists of one robot arm, laboratory equipment (an incubator, a reagent bottle with a cap, and a Petri dish), the existing code that moves the arm, teleoperation demonstrations of each task that we recorded, and a general AGENTS.md skill file. A biological experimentalist provides natural-language tasks without collecting local teleoperation training data or training a neural network. The coding agent observes the local laboratory and writes and executes programs, using external tools as needed for adaptation. We demonstrate use of WetRobo with OpenAI Codex (gpt-5.6-sol) on three successful tasks: lifting a Petri dish lid, removing a bottle cap, and opening the incubator door, all in real-world laboratories. The coding agent achieved the cap task in both laboratories, Lab X and Lab Y, whereas a VLA fine-tuned on Lab X demonstrations succeeded there but failed to transfer to Lab Y. These results point to a practical route for laboratory robotics: instead of training a policy for each laboratory, distribute a kit and let a coding agent adapt it in each laboratory. Code, demonstrations, and the evolved programs are available at https://github.com/tsudalab/WetRobo.
Chinese Translation
自动化生物研究需要通用的、可重复的机器人系统，使得个体湿实验室研究人员能够在不进行远程操作或神经网络训练的情况下委派机器人任务。虽然已经提出了视觉-语言-动作（Vision-Language-Action, VLA）策略用于通用机械臂，但在操作环境变化时可能会导致性能下降。因此，我们构建了WetRobo，一个可以在实验室之间轻松转移的机器人套件。它由一个机器人手臂、实验室设备（一个培养箱、一个带盖的试剂瓶和一个培养皿）、移动手臂的现有代码、我们记录的每个任务的远程操作演示以及一个通用的AGENTS.md技能文件组成。生物实验人员可以提供自然语言任务，而无需收集本地远程操作训练数据或训练神经网络。编码代理观察本地实验室，并编写和执行程序，根据需要使用外部工具进行适应。我们展示了WetRobo与OpenAI Codex (gpt-5.6-sol) 在三个成功任务中的应用：打开培养皿盖、取下瓶盖和打开培养箱门，所有任务均在真实实验室中完成。编码代理在实验室X和实验室Y均成功完成了瓶盖任务，而在实验室X演示上微调的VLA在实验室X成功，但未能转移到实验室Y。这些结果指向了实验室机器人技术的一个实际途径：与其为每个实验室训练一个策略，不如分发一个套件，让编码代理在每个实验室中进行适应。代码、演示和演变的程序可在 https://github.com/tsudalab/WetRobo 获取。
cs.AI / 54 / 2609.18442
Risk-Aware World Modeling with Flow-Guided Occupancy Evolution for Selective Trajectory Planning in Automated Driving
基于流引导占用演化的风险感知世界建模用于自动驾驶中的选择性轨迹规划
Abstract
Safe motion planning in automated driving requires anticipating evolving traffic risks and deciding when to revise the current planned trajectory. We introduce RiskWorld, a risk-aware world modeling framework for shared occupancy forecasting and selective trajectory replacement. Spatial risk fields and temporal actor context are fused with visual bird's-eye-view features. Flow-guided evolution transports occupancy and scene features, while signed residuals correct occupancy after transport. One forecast is generated per planning step and reused across candidates. Each candidate is compared with a current-state persistence reference, yielding a nonnegative collision-score correction. The trajectory selected by current-world evaluation serves as the planning anchor and is replaced only when additional predicted risk triggers intervention and an alternative satisfies component-wise constraints on predicted risk and trajectory error. Candidate geometries remain unchanged. We evaluate RiskWorld for open-loop planning on nuScenes using camera features, annotation-derived current and historical actor states, and dataset-provided map context. RiskWorld achieves the lowest collision rate at a long evaluation horizon of 3 s, and the second-best average L2 error among various state-of-the-art baselines, while running at 11.5 FPS on a single NVIDIA RTX 4090 with 90.81 M parameters. Within-setting ablations show that RiskWorld achieves lower collision rates than the current-state rescoring baseline, while forecast reuse enables additional candidates to be evaluated at low marginal computational cost.
Chinese Translation
在自动驾驶中，安全的运动规划需要预测不断变化的交通风险，并决定何时修订当前规划的轨迹。我们提出了RiskWorld，一个风险感知的世界建模框架，用于共享占用预测和选择性轨迹替换。空间风险场和时间行为者上下文与视觉鸟瞰特征相融合。流引导演化传输占用和场景特征，而带符号的残差在传输后校正占用。每个规划步骤生成一个预测，并在候选者之间重用。每个候选者与当前状态持久性参考进行比较，产生非负的碰撞评分校正。通过当前世界评估选择的轨迹作为规划锚，仅在额外预测风险触发干预且替代方案满足预测风险和轨迹误差的分量约束时才进行替换。候选几何形状保持不变。我们在nuScenes上评估RiskWorld的开环规划，使用相机特征、注释派生的当前和历史行为者状态，以及数据集提供的地图上下文。RiskWorld在3秒的长评估时间范围内实现了最低的碰撞率，并在各种最先进的基线中获得第二低的平均L2误差，同时在单个NVIDIA RTX 4090上以11.5 FPS的速度运行，参数量为90.81M。设置内消融实验表明，RiskWorld的碰撞率低于当前状态重评分基线，而预测重用使得可以以低边际计算成本评估额外候选者。
cs.AI / 55 / 2609.18453
The Mirage of Calibrated Confidence: Trajectory-Independence of Verbalized Confidence in Vision-Language Models
校准信心的幻影：语言-视觉模型中语言化信心的轨迹独立性
Abstract
A calibrated Vision-Language Model (VLM) can repeatedly self-correct, say "Wait, I should recheck," arrive at the wrong answer, and still report high confidence. We find that this occurs because verbalized confidence is largely trajectory-independent in the VLMs and calibration methods we evaluate. We examine this through three complementary lenses: content variation, token masking, and the model's own hesitation markers. We show that confidence is insufficiently sensitive to what the reasoning trajectory actually contains, and that calibration training can paradoxically worsen this disconnect. Since existing metrics like ECE and AUROC cannot detect this problem, we propose the Trajectory-Grounding Score (TGS) in two complementary forms: TGS-self, which compares confidence with and without access to the model's own trajectory, and TGS-pair, which tests whether the model assigns higher confidence to correct trajectories than to flawed ones along the vision, reasoning, and answer axes. We propose TGS-Bench, a model-agnostic suite spanning 10 benchmarks with controlled good/bad trajectory pairs, and show that conventional calibration rankings diverge from trajectory-grounding rankings, exposing a blind spot in current evaluation practice.
Chinese Translation
一个经过校准的视觉-语言模型（VLM）可以反复自我纠正，例如说“等一下，我应该重新检查”，即使得出错误答案，仍然报告出高信心。我们发现，这种现象的发生是因为在我们评估的视觉-语言模型和校准方法中，语言化信心在很大程度上是轨迹独立的。我们通过三种互补的视角来研究这一问题：内容变化、标记掩蔽和模型自身的犹豫标记。我们展示了信心对推理轨迹实际内容的敏感性不足，并且校准训练可能会悖论性地加剧这种脱节。由于现有的指标如ECE和AUROC无法检测到这个问题，我们提出了轨迹基础评分（Trajectory-Grounding Score，TGS），以两种互补的形式：TGS-self，比较有无访问模型自身轨迹时的信心，以及TGS-pair，测试模型是否在视觉、推理和答案轴上对正确轨迹赋予比错误轨迹更高的信心。我们提出了TGS-Bench，这是一个模型无关的套件，涵盖10个具有控制良好/不良轨迹对的基准，并展示了传统的校准排名与轨迹基础排名之间的偏离，揭示了当前评估实践中的盲点。
cs.AI / 56 / 2609.18460
Collective Loss of Control in LLM Agent Systems: An Epidemic Account of Mutation, Contagion, and Recovery
大型语言模型代理系统中的集体失控：突变、传播与恢复的流行病学分析
Abstract
How does a multi-agent system evolve from a local deviation into collective loss of control? We propose an epidemic explanation organized around accidental mutation, contagion, and recovery. A spontaneous deviation creates a seed; communication enables other agents to adopt and retransmit its unsafe strategy; collective failure can emerge when propagation outpaces correction and containment. Thus, rare individual deviations can coexist with substantial collective risk. Motivated by reported OpenAI agent coordination incidents, we examine two ingredients of this mechanism. A deployment audit identifies implicit communication paths between nominally independent evaluation runs and verifies transport through a default Docker backend. RogueHandoff-20, a benchmark of 20 executable scenarios, tests recipient susceptibility by injecting unsafe trajectories generated by a modified Qwen-27B route. Across four native-pending routes, executed harm is 0-5% on normal tasks and 40-95% after injection, exceeding paired direct malicious requests by 5-45 percentage points. These results support low observed baseline harm alongside high conditional susceptibility; they do not establish natural rare-event rates or demonstrate an autonomous cascade. The account motivates complementary defenses: strengthen resistance and recovery alongside prevention of spontaneous deviations, and audit and restrict unintended communication paths that can turn local failures into collective loss of control.
Chinese Translation
多代理系统如何从局部偏差演变为集体失控？我们提出了一种围绕意外突变、传播和恢复的流行病学解释。自发的偏差创造了一个种子；通信使其他代理能够采用并重新传播其不安全的策略；当传播速度超过纠正和控制时，集体失败可能会出现。因此，罕见的个体偏差可以与显著的集体风险共存。受到报告的OpenAI代理协调事件的启发，我们考察了这一机制的两个组成部分。部署审计识别了名义上独立的评估运行之间的隐含通信路径，并通过默认的Docker后端验证了传输。RogueHandoff-20是一个包含20个可执行场景的基准，通过注入由修改后的Qwen-27B路径生成的不安全轨迹来测试接收者的易感性。在四条原生待处理路径中，正常任务的执行伤害为0-5%，而注入后的伤害为40-95%，超过了配对的直接恶意请求5-45个百分点。这些结果支持观察到的低基线伤害与高条件易感性；它们并未建立自然罕见事件的发生率或证明自主级联的存在。该分析激励了补充防御措施：在防止自发偏差的同时，加强抵抗力和恢复能力，并审计和限制可能将局部失败转变为集体失控的意外通信路径。
cs.AI / 57 / 2609.18461
Disentangling Long-Term Memory via Latent Neuro-Symbolic Reasoning
通过潜在神经符号推理解开长期记忆
Abstract
Personalized agents are required to reason over long-term history interactions to infer both explicit preferences and implicit behavioral evidence. While early flat retrieval methods score memory fragments independently and neglect the distributed information, current structured memory frameworks rely on query-agnostic static graphs that fail to capture the context-dependent relations. Crucially, raw textual memories are inherently entangled and noisy, making fine-grained personalization and cross-session reasoning computationally prohibitive. To this end, we present LGM, a novel neuro-symbolic framework that shifts long-term memory disentanglement into a continuous latent space. Specifically, (i) instead of persisting fixed graphs, we design a tailored latent graph construction with a sparse autoencoder. Subject to each query, it maps historical interactions into latent memory nodes and disentangles the memory traces into sparse concept activations, dynamically synthesizing query-aware relational edge weights. (ii) A graph encoder then treats the query embedding as a conditioning preference to direct non-linear message passing across the task-specific latent subgraph. This yields a highly expressive memory representation for effective activations. Extensive experiments on long-term personalization benchmarks demonstrate that LGM significantly outperforms state-of-the-art baselines in capturing both explicit and implicit preferences while enabling personalized responses.
Chinese Translation
个性化代理需要对长期历史交互进行推理，以推断显性偏好和隐性行为证据。早期的平面检索方法独立地评分记忆片段，忽视了分布式信息，而当前的结构化记忆框架依赖于与查询无关的静态图，这无法捕捉上下文依赖关系。关键在于，原始文本记忆本质上是纠缠和嘈杂的，这使得细粒度个性化和跨会话推理在计算上变得不可行。为此，我们提出了LGM，一种新颖的神经符号框架，将长期记忆解缠转移到一个连续的潜在空间。具体而言，(i) 我们设计了一种定制的潜在图构建，使用稀疏自编码器，而不是保持固定的图。针对每个查询，它将历史交互映射到潜在记忆节点，并将记忆痕迹解缠为稀疏概念激活，动态合成查询感知的关系边权重。(ii) 图编码器将查询嵌入视为条件偏好，以指导任务特定潜在子图中的非线性消息传递。这产生了高度表达性的记忆表示，以实现有效的激活。在长期个性化基准上的大量实验表明，LGM在捕捉显性和隐性偏好方面显著优于最先进的基线，同时实现个性化响应。
cs.AI / 58 / 2609.18471
First Token Matters: Understanding Safety Collapse in Large Reasoning Models
首个标记的重要性：理解大型推理模型中的安全崩溃
Abstract
Large Reasoning Models (LRMs) exhibit strong problem-solving abilities, yet their safety alignment often degrades when handling harmful queries. Existing approaches to improving safety largely rely on additional training or preference optimization, while offering limited understanding of the internal mechanisms behind safety failures. In this work, we investigate this failure through a token-level positional analysis of refusal dynamics and identify a localized vulnerability at the onset of reasoning, which we term Onset Refusal Collapse (ORC). We find that the refusal-related signal of LRMs drops sharply at the first generated token under harmful queries, which is associated with unsafe response generation. Motivated by this finding, we propose SafeToken, a lightweight inference-time intervention that injects a learned continuous safety anchor precisely at reasoning onset. Despite updating only a single token embedding, SafeToken effectively mitigates ORC, improves safety on harmful-query benchmarks, and largely preserves reasoning utility. These results suggest that safety failures in LRMs can arise from a transient breakdown at the critical transition from understanding to generation.
Chinese Translation
大型推理模型（Large Reasoning Models, LRM）展现出强大的问题解决能力，但在处理有害查询时，其安全对齐往往会下降。现有的安全性提升方法主要依赖额外的训练或偏好优化，而对安全失败背后的内部机制了解有限。在本研究中，我们通过对拒绝动态的标记级位置分析来探讨这一失败，并识别出在推理开始时的局部脆弱性，我们称之为初始拒绝崩溃（Onset Refusal Collapse, ORC）。我们发现，在有害查询下，LRM的拒绝相关信号在首个生成标记时急剧下降，这与不安全的响应生成相关。基于这一发现，我们提出了SafeToken，一种轻量级的推理时干预方法，能够在推理开始时精确注入一个学习到的连续安全锚点。尽管仅更新了一个标记嵌入，SafeToken有效缓解了ORC，提高了在有害查询基准上的安全性，并在很大程度上保持了推理效用。这些结果表明，LRM中的安全失败可能源于从理解到生成的关键过渡中的瞬时崩溃。
cs.AI / 59 / 2609.18481
Hyperbolic Graph Representation Learning for Differential Diagnosis on Biomedical Knowledge Graphs
用于生物医学知识图谱差异诊断的双曲图表示学习
Abstract
Biomedical knowledge graphs combine ontology-derived hierarchies with transversal associations among heterogeneous entities such as phenotypes, diseases, genes, proteins, and patients. This hybrid structure raises the question of whether hyperbolic embeddings, which naturally capture tree-like organization, remain useful beyond purely hierarchical graphs. We present a preliminary study of hyperbolic graph representation learning for Mendelian-disease differential diagnosis on a patient-integrated biomedical graph. Experiments on isolated ontology subgraphs show that hyperbolic models achieve strong performance in substantially lower dimensions than Euclidean baselines. We then evaluate the models on a link-prediction task that ranks candidate diseases for each patient. Results suggest that hyperbolic embeddings can exploit biomedical hierarchical structure while supporting diagnostic reasoning over heterogeneous patient-level graphs.
Chinese Translation
生物医学知识图谱结合了基于本体的层次结构与异构实体（如表型、疾病、基因、蛋白质和患者）之间的横向关联。这种混合结构引发了一个问题：双曲嵌入是否在超越纯层次图的情况下仍然有效，因为它们自然捕捉树状组织。我们对在患者整合的生物医学图谱上进行的孟德尔疾病差异诊断的双曲图表示学习进行了初步研究。对孤立本体子图的实验表明，双曲模型在显著低于欧几里得基线的维度下实现了强劲的性能。随后，我们在一个链接预测任务上评估了这些模型，该任务为每位患者排名候选疾病。结果表明，双曲嵌入可以利用生物医学层次结构，同时支持对异构患者级图的诊断推理。
cs.AI / 60 / 2609.18515
Beyond Routine Compliance: Cunning Data Cultivates Safety Vigilance in Large Language Models
超越常规合规：巧妙数据培养大型语言模型的安全警觉性
Abstract
Safety alignment teaches large language models (LLMs) to recognize harmful requests and reject risky instructions. Yet aligned models can fail when harmful intent is concealed within seemingly benign contexts. Robust safety therefore requires both knowledge of safety boundaries and \textbf{vigilance}: the ability to detect unusual premises, misleading reasoning, and latent risks beneath surface-level semantics. Vigilance requires models to scrutinize a request's underlying intent and assumptions before acting. To cultivate this capability, we introduce \textbf{cunning questions}, which are not necessarily safety-related but contain misleading premises, atypical reasoning, or subtle inconsistencies. We hypothesize that learning to look beyond such reasoning traps can transfer to safety-critical scenarios. Experiments show that Cunning training improves robustness to out-of-distribution jailbreak attacks and strengthens subsequent safety fine-tuning. Furthermore, augmenting an existing state-of-the-art safety alignment pipeline with Cunning establishes a new state of the art across our evaluated settings, reducing mean ASR across nine backbone--benchmark combinations from 17.40\% to 15.05\%. Trace analysis after matched safety fine-tuning suggests that safety judgments are more likely to govern responses before harmful planning begins. A conditional theoretical analysis further characterizes when invariance learned from cunning data can transfer to safety-related inputs. These findings suggest that cunning data can strengthen model vigilance and complement conventional safety alignment.
Chinese Translation
安全对齐教会大型语言模型（LLMs）识别有害请求并拒绝风险指令。然而，当有害意图隐藏在看似无害的上下文中时，对齐模型可能会失败。因此，稳健的安全性不仅需要对安全边界的了解，还需要 extbf{警觉性}：即在表面语义之下检测不寻常前提、误导性推理和潜在风险的能力。警觉性要求模型在采取行动之前仔细审视请求的潜在意图和假设。为了培养这种能力，我们引入了 extbf{巧妙问题}，这些问题不一定与安全相关，但包含误导性前提、非典型推理或微妙的不一致性。我们假设，学习超越此类推理陷阱的能力可以转移到安全关键场景中。实验表明，巧妙训练提高了对分布外越狱攻击的鲁棒性，并增强了后续的安全微调。此外，通过在现有的最先进安全对齐流程中增加巧妙数据，建立了我们评估设置中的新最先进水平，将九个基础模型-基准组合的平均攻击成功率（ASR）从17.40\%降低到15.05\%。在匹配的安全微调后进行的追踪分析表明，在有害规划开始之前，安全判断更有可能主导响应。条件理论分析进一步描述了从巧妙数据中学习到的不变性何时可以转移到与安全相关的输入。这些发现表明，巧妙数据可以增强模型的警觉性，并补充传统的安全对齐。
cs.AI / 61 / 2609.18520
AeroWeaver: An Embodied-Agent Harness for Weaving Aerial Skills into Distributed, Adaptive Swarm Execution
AeroWeaver：一种将空中技能编织到分布式、自适应群体执行中的具身代理工具
Abstract
Collective intelligence is a collaborative autonomy paradigm in which multiple agents pursue shared objectives through local perception, information exchange, and coordinated action. UAV swarms embody this paradigm by coordinating multiple vehicles in tasks such as search, inspection, and tracking. Recent advances in large language model (LLM) agents have strengthened natural-language task understanding and high-level planning, providing a flexible semantic interface between mission descriptions and collective behavior. While these advances expand semantic reasoning, applying LLM agents to UAV swarms raises challenges in grounding model decisions in executable capabilities, reconciling global task reasoning with distributed execution, and using mission-specific experience for continual adaptation. To address these challenges, we introduce AeroWeaver, an embodied-agent harness that weaves individual UAV skills into coordinated mission-level behavior. AeroWeaver connects semantic decisions to governed skills, organizes role-conditioned local agents for distributed coordination, and uses role-indexed state-action-reward experience to refine skill selection online. Experiments and runtime validation show that AeroWeaver maintains valid skill execution under tested conditions and supports body-local multi-UAV operation without a central agent generating joint actions from global context, while reward-guided online updates provide a training-free path for adaptive learning swarm agents from accumulated execution experience. Code: https://github.com/Admire-ljb/AeroWeaver.
Chinese Translation
集体智能是一种协作自主范式，其中多个代理通过局部感知、信息交换和协调行动追求共同目标。无人机群体现了这一范式，通过协调多辆飞行器执行搜索、检查和追踪等任务。近期大型语言模型（LLM）代理的进展增强了自然语言任务理解和高层次规划，提供了任务描述与集体行为之间的灵活语义接口。尽管这些进展扩展了语义推理，但将LLM代理应用于无人机群面临着将模型决策与可执行能力相结合、协调全局任务推理与分布式执行，以及利用特定任务经验进行持续适应的挑战。为了解决这些挑战，我们提出了AeroWeaver，这是一种将单个无人机技能编织成协调任务级行为的具身代理工具。AeroWeaver将语义决策与管理技能相连接，组织基于角色的局部代理以实现分布式协调，并利用基于角色的状态-动作-奖励经验在线优化技能选择。实验和运行时验证表明，AeroWeaver在测试条件下维持有效的技能执行，并支持无中央代理生成全局上下文的联合行动的体局部多无人机操作，同时奖励引导的在线更新为自适应学习群体代理提供了一个无训练的路径，基于累积的执行经验进行学习。代码：https://github.com/Admire-ljb/AeroWeaver。
cs.AI / 62 / 2609.18525
TRIPROBE: Probing Task Separability Beyond Classification for XAI
TRIPROBE：超越分类的任务可分离性探测框架用于可解释人工智能
Abstract
Modern evaluation of learning pipelines often reduces to downstream accuracy, leaving open the question of why tasks succeed or fail. TriProbe addresses this gap with a multi-level probing framework for explainable diagnosis of task separability. Rather than treating models as black boxes, TriProbe traces how separability evolves across inputs, learned features, and final classifiers. It decomposes multi-task problems into binary subtasks and applies three complementary probes: a Foundational Probe on input spaces, a Latent Probe on feature representations, and a Final Probe on classifier outputs. Using Maximum Fisher's Discriminant Ratio as a principled separability metric, TriProbe identifies bottlenecks and affected task pairs. Experiments on the Roshambo sEMG benchmark show how TriProbe reveals hidden breakdowns, guiding data collection, validation, and architecture design.
Chinese Translation
现代学习管道的评估往往简化为下游准确性，这使得任务成功或失败的原因仍然不明。TriProbe 通过一个多层次探测框架解决了这一空白，以实现对任务可分离性的可解释诊断。TriProbe 不将模型视为黑箱，而是追踪可分离性如何在输入、学习特征和最终分类器之间演变。它将多任务问题分解为二元子任务，并应用三种互补的探测器：对输入空间的基础探测器（Foundational Probe）、对特征表示的潜在探测器（Latent Probe）以及对分类器输出的最终探测器（Final Probe）。使用最大 Fisher 判别比（Maximum Fisher's Discriminant Ratio）作为原则性的可分离性度量，TriProbe 识别瓶颈和受影响的任务对。在 Roshambo sEMG 基准上的实验表明，TriProbe 如何揭示隐藏的故障，从而指导数据收集、验证和架构设计。
cs.AI / 63 / 2609.18591
Recursive Reasoning or Statistical Extrapolation? In-Context Learning in Multi-Agent Interdependent Decision-Making
递归推理还是统计外推？多智能体相互依赖决策中的情境学习
Abstract
In-context learning (ICL) enables large language model (LLM) agents to improve decisions using interaction history, yet it remains unclear whether such improvement reflects refined internal reasoning or mere extrapolation of statistical patterns. To disentangle these mechanisms, we study LLM agents in multi-agent incomplete-information games that require recursive belief reasoning. By constructing a public goods game and manipulating the statistical structure of historical feedback, we evaluate decision quality against a history-independent rational expectations equilibrium (REE) benchmark. Our experiments reveal that when historical statistical patterns are disrupted, the benefits of longer context largely vanish, degrading decision quality to the no-context baseline in a way sharply amplified by stronger strategic interdependence. These results suggest that, in such strategic environments, ICL behavior is more consistent with statistical extrapolation than with strategic reasoning. Our work extends the mechanistic study of ICL to strategic multi-agent settings, introduces REE as a diagnostic tool for distinguishing reasoning from extrapolation, and provides a reusable framework for probing the boundaries of LLM reasoning in recursive belief tasks.
Chinese Translation
情境学习（In-context learning, ICL）使大型语言模型（Large Language Model, LLM）代理能够利用互动历史改善决策，然而尚不清楚这种改善是否反映了精细的内部推理还是仅仅是统计模式的外推。为了理清这些机制，我们研究了在需要递归信念推理的多智能体不完全信息博弈中的LLM代理。通过构建公共物品博弈并操控历史反馈的统计结构，我们将决策质量与历史无关的理性预期均衡（Rational Expectations Equilibrium, REE）基准进行比较。实验结果表明，当历史统计模式被扰动时，较长上下文的优势大幅消失，决策质量降至无上下文基线，这一现象在更强的战略相互依赖下被显著放大。这些结果表明，在此类战略环境中，ICL行为更符合统计外推而非战略推理。我们的研究将ICL的机制研究扩展到战略多智能体环境，引入REE作为区分推理与外推的诊断工具，并提供了一个可重复使用的框架，以探讨LLM在递归信念任务中的推理边界。
cs.AI / 64 / 2609.18597
Reasoning through Evolution: Automatic Meta-path Discovery for LLM-based Fake News Detection
通过演化推理：基于LLM的假新闻检测的自动元路径发现
Abstract
Propagation structures provide crucial evidence for fake news detection, yet existing approaches primarily rely on supervised GNN-based models, which require substantial labeled data and exhibit limited generalization. Although large language models (LLMs) exhibit strong reasoning capabilities, directly feeding them raw propagation graphs creates a significant modality mismatch and severe information overload, making structure-aware reasoning unreliable in zero-shot and few-shot settings. To bridge this gap, we propose MAGER, a multi-agent genetic evolution framework that automatically discovers meta-paths optimized for LLM reasoning. By compressing complex propagation graphs into informative subgraphs, the evolved meta-paths alleviate both information overload and modality mismatch, enabling frozen LLMs to perform structure-aware veracity reasoning. We further introduce a graph in-context learning strategy that retrieves semantically and structurally similar demonstrations to strengthen classification and reasoning. Extensive experiments show that MAGER substantially improves frozen LLMs as standalone fake news detectors in data-efficient settings. Our code is available at https://github.com/SenticNet/MAGER.
Chinese Translation
传播结构为假新闻检测提供了重要证据，但现有方法主要依赖于监督的图神经网络（GNN）模型，这些模型需要大量标注数据且泛化能力有限。尽管大型语言模型（LLMs）展现出强大的推理能力，但直接将原始传播图输入给它们会导致显著的模态不匹配和严重的信息过载，使得在零样本和少样本环境下的结构感知推理不可靠。为了解决这一问题，我们提出了MAGER，一个多智能体遗传演化框架，能够自动发现针对LLM推理优化的元路径。通过将复杂的传播图压缩为信息丰富的子图，演化出的元路径缓解了信息过载和模态不匹配，使得冻结的LLM能够进行结构感知的真实性推理。我们进一步引入了一种图内上下文学习策略，该策略检索语义和结构上相似的示例，以增强分类和推理能力。大量实验表明，MAGER在数据高效的环境中显著提升了冻结LLM作为独立假新闻检测器的性能。我们的代码可在 https://github.com/SenticNet/MAGER 获取。
cs.AI / 65 / 2609.18676
The Uneven Impact of Generative AI on Student Learning: Examining the Roles of Reliance, Evaluation Literacy, and Course Policy in AI-related Courses
生成性人工智能对学生学习的影响不均衡：考察依赖性、评估素养和课程政策在与人工智能相关课程中的作用
Abstract
Generative artificial intelligence (GenAI) is changing how students learn, yet the roles of course context, cognitive reliance, evaluation literacy, and early reliance remain underexplored. Using survey responses from 118 students across 12 AI-related courses at our institution, we examined differences in GenAI use and perceived learning experiences. We identified four user clusters: high-use students reporting many benefits, light users reporting less reliance and fewer benefits, and two moderate-use groups reporting different levels of benefit. We also found significant differences between free- and premium-version users, single- and multiple-tool users, and students experiencing different instructor policies. In multivariable regression models, academic benefit was associated with early reliance and academic task support; positive impact was associated with cognitive reliance, academic task support, confidence in GenAI reliability, and instructor policy; and negative impact was associated with early reliance and attitudinal change. The association between early reliance and negative impact became stronger as evaluation literacy increased. Finally, perceptions of GenAI-enhanced learning appear to reflect cognitive, performance, and self-efficacy benefits, while concerns about stress and diminished critical thinking are associated with lower perceived learning benefits. These findings suggest that institutions need better policies to address such inequities so that institutions can enable students to benefit from increasingly capable AI systems.
Chinese Translation
生成性人工智能（GenAI）正在改变学生的学习方式，但课程背景、认知依赖、评估素养和早期依赖的作用仍然未得到充分探讨。我们利用来自我校12门与人工智能相关课程的118名学生的调查反馈，考察了GenAI使用和感知学习体验的差异。我们识别了四个用户群体：高使用者报告了许多好处，轻度使用者报告了较少的依赖和好处，以及两个中度使用群体报告了不同程度的好处。我们还发现免费版与付费版用户、单一工具与多重工具用户，以及经历不同教师政策的学生之间存在显著差异。在多变量回归模型中，学术收益与早期依赖和学术任务支持相关；积极影响与认知依赖、学术任务支持、对GenAI可靠性的信心和教师政策相关；而负面影响则与早期依赖和态度变化相关。随着评估素养的提高，早期依赖与负面影响之间的关联变得更强。最后，关于GenAI增强学习的看法似乎反映了认知、表现和自我效能的好处，而对压力和批判性思维下降的担忧则与较低的感知学习收益相关。这些发现表明，机构需要更好的政策来解决这种不平等，以便使学生能够从日益强大的人工智能系统中受益。
cs.AI / 66 / 2609.18723
Beyond Truncation: Rethinking LLM Decoding as Ensemble Pruning
超越截断：重新思考大型语言模型解码为集成修剪
Abstract
We introduce Mahalanobis-Ensemble Decoding (ME-Decoding), a novel Large Language Model (LLM) decoding framework that frames candidate token selection as ensemble pruning. Existing selection strategies rely predominantly on scalar probabilities, ignoring geometric semantic relationships and causing candidate redundancy. Meanwhile, current geometry-aware methods often require complex optimization or directly reweighting the original token probabilities, leading to significant computational overhead or inference instability. To address this, we formulate decoding as a subset optimization problem using a Mahalanobis distance-driven objective to enhance semantic diversity while preserving high probabilities. Specifically, we dynamically discount redundant generation paths using a token similarity matrix, constructed via an adaptive-bandwidth kernel over token embeddings. We further devise an efficient greedy selection algorithm with near-linear complexity in the candidate size under early stopping, while establishing its theoretical approximation guarantees. This renders ME-Decoding a robust, plug-and-play module with negligible inference overhead. Extensive experiments across diverse reasoning and generation tasks demonstrate that our method consistently achieves strong performance.
Chinese Translation
我们提出了马哈拉诺比斯集成解码（Mahalanobis-Ensemble Decoding，ME-Decoding），这是一种新颖的大型语言模型（Large Language Model，LLM）解码框架，将候选标记选择视为集成修剪。现有的选择策略主要依赖于标量概率，忽视了几何语义关系，导致候选冗余。同时，当前的几何感知方法通常需要复杂的优化或直接重新加权原始标记概率，从而导致显著的计算开销或推理不稳定。为了解决这一问题，我们将解码形式化为一个子集优化问题，使用基于马哈拉诺比斯距离的目标来增强语义多样性，同时保持高概率。具体而言，我们通过使用基于自适应带宽核的标记嵌入构建的标记相似性矩阵，动态地折扣冗余生成路径。我们进一步设计了一种高效的贪婪选择算法，在早期停止的情况下，其候选大小近似线性复杂度，同时建立了其理论近似保证。这使得ME-Decoding成为一个稳健的即插即用模块，几乎没有推理开销。针对多种推理和生成任务的广泛实验表明，我们的方法始终实现了强劲的性能。
cs.AI / 67 / 2609.18731
Which LLM is Best for Translating Natural Language Goals to PDDL
哪种大型语言模型最适合将自然语言目标转换为PDDL
Abstract
Bridging the gap between human intent and machine execution remains a challenge in automated planning, where expressing goals in formal languages like PDDL restricts accessibility to non-experts. This paper empirically evaluates whether current Large Language Models (LLMs) can reliably translate natural language testing goals, written in informal language by video game testers, into well-formed PDDL targets suitable for classical planning. We present a carefully designed prompt template, integrating insights from iterative experimentation, aimed at maximizing both accuracy and response coherence from multiple state-of-the-art LLMs. Six contemporary models are systematically assessed on correctness, speed, and error tendencies using real-world, domain-specific benchmarks. All models demonstrate high correctness, exceeding 92\%, with Gemini 2.5 Flash achieving the highest accuracy at 96\% and the lowest incidence of false positives, while GPT-4.1 leads in response speed. Despite these advances, critical distinctions exist in model performance, and occasional failures arise from language ambiguity and limitations in domain representation. Our analysis underscores both the significant progress and ongoing gaps in enabling LLMs to act as robust bridges between natural language objectives and automated planning pipelines.
Chinese Translation
在自动规划中，弥合人类意图与机器执行之间的差距仍然是一个挑战，因为在PDDL等形式语言中表达目标限制了非专家的可及性。本文通过实证研究评估当前的大型语言模型（LLMs）是否能够可靠地将视频游戏测试人员用非正式语言编写的自然语言测试目标转换为适合经典规划的格式良好的PDDL目标。我们提出了一个精心设计的提示模板，结合了迭代实验的见解，旨在最大化多个最先进LLMs的准确性和响应一致性。六个当代模型在正确性、速度和错误倾向方面进行了系统评估，使用真实世界的领域特定基准。所有模型均表现出高正确性，超过92%，其中Gemini 2.5 Flash的准确性最高，达到96%，且假阳性发生率最低，而GPT-4.1在响应速度上领先。尽管取得了这些进展，模型性能之间仍存在显著差异，偶尔的失败源于语言歧义和领域表示的局限性。我们的分析强调了在使LLMs充当自然语言目标与自动规划管道之间的稳健桥梁方面的重大进展和持续差距。
cs.AI / 68 / 2609.18736
Clueing up LLMs with Tool-Augmented Deductive Reasoning
通过工具增强的演绎推理来提升大型语言模型的能力
Abstract
Despite recent advances in large language models (LLMs), performing logically consistent deductive reasoning over extended interactions remains challenging. Tasks that require integrating evidence across multiple reasoning steps, maintaining consistency with prior inferences, and updating beliefs under new constraints can surface limitations in current models while providing a useful testbed for evaluating reasoning enhancements. In this paper, we implement a text-based, multi-agent version of the classic board game Clue as an environment to evaluate multi-step, agentic deductive reasoning. In this setting, agents must infer hidden information from a sequence of observations, maintain consistency across turns, and reason over an evolving set of logical constraints. We instantiate six LLM-based agents (GPT-4o-mini and Gemini-2.5-Flash) as players that engage in turn-based gameplay; using three agents per model family, we establish baseline performance across repeated games. We then introduce a tool-augmented approach in which a structured possibility matrix converts implicit game state from generated reasoning logs into an explicit representation of remaining possibilities. The possibility matrix encodes extended-turn memory and deductive constraints, offloading these tasks from the agent. We compare this approach against the baseline to evaluate how tool augmentation supports reasoning quality and task success for autonomous agents in a strategic reasoning environment.
Chinese Translation
尽管大型语言模型（LLMs）近年来取得了进展，但在扩展交互中执行逻辑一致的演绎推理仍然具有挑战性。需要在多个推理步骤中整合证据、与先前推论保持一致并在新约束下更新信念的任务可能会暴露当前模型的局限性，同时为评估推理增强提供了有用的测试平台。本文实现了经典桌游《Clue》的文本基础多智能体版本，作为评估多步骤、代理性演绎推理的环境。在这个设置中，智能体必须从一系列观察中推断隐藏信息，在回合之间保持一致性，并在不断变化的逻辑约束集上进行推理。我们实例化了六个基于LLM的智能体（GPT-4o-mini和Gemini-2.5-Flash）作为参与者，进行回合制游戏；每个模型系列使用三个智能体，我们在重复游戏中建立基线性能。然后，我们引入了一种工具增强的方法，其中一个结构化的可能性矩阵将从生成的推理日志中隐含的游戏状态转换为剩余可能性的显式表示。可能性矩阵编码了扩展回合记忆和演绎约束，将这些任务卸载给智能体。我们将这种方法与基线进行比较，以评估工具增强如何支持自主智能体在战略推理环境中的推理质量和任务成功。
cs.AI / 69 / 2609.18769
Version- and Scope-Aware Question Answering over Normative Documents: A Deployed System and an End-to-End Evaluation at Production Scale
基于版本和范围的规范性文件问答系统：一个已部署系统及其生产规模的端到端评估
Abstract
Correctly answering a question grounded in normative documents often depends on information outside any single passage: whether the retrieved document is the version currently in force; whether it applies to the jurisdiction, subject (such as an institution or applicant), and date at issue; and whether each normative claim can be traced to its supporting source text. Hosted retrieval services have substantially lowered the engineering cost of building an initial system over such corpora, making "upload the documents and ask" a common default. We evaluate this default on approximately 73,000 candidate normative documents supplied to a production deployment. The evaluation uses a stratified sample of 200 questions from our published benchmark, with a gold source document for every question; the released sampling rule reads no system outputs or scores. We compare the hosted service with a governed system that resolves version and scope through explicit rules before generation. The governed system scored 97.7 overall, while the hosted service scored 88.1, a gap of 9.6 points computed from unrounded means. The question set, the answer text evaluated for both systems, the scores, and the scripts used to reproduce the reported benchmark statistics are public. The governed configuration has operated as a commercial product since January 2026 and serves 1,126 registered users; named customer organizations include Zhipu AI and Lecheng Health. By mid-April 2026, it had reached roughly 100,000 calls per workday.
Chinese Translation
正确回答基于规范性文件的问题通常依赖于单一段落之外的信息：检索到的文件是否为当前有效版本；它是否适用于相关的管辖区、主题（如机构或申请人）和相关日期；以及每个规范性主张是否可以追溯到其支持的源文本。托管检索服务显著降低了在此类语料库上构建初始系统的工程成本，使得“上传文件并提问”成为一种常见的默认方式。我们在大约73,000个候选规范性文件上评估了这一默认设置，这些文件被提供给生产部署。评估使用了我们已发布基准中的200个问题的分层样本，每个问题都有一个金标准源文件；发布的抽样规则不读取任何系统输出或评分。我们将托管服务与一个通过明确规则在生成之前解决版本和范围的管理系统进行了比较。管理系统的总得分为97.7，而托管服务的得分为88.1，未四舍五入的平均值计算得出的差距为9.6分。问题集、两个系统评估的答案文本、得分以及用于重现报告的基准统计的脚本均为公开信息。自2026年1月以来，管理配置作为商业产品运营，服务于1,126名注册用户；客户组织包括智谱AI和乐成健康。截至2026年4月中旬，它的工作日调用次数已达到大约100,000次。
cs.AI / 70 / 2609.18779
CERA-MoA: Co-Evolving Routing Mechanisms with Continually Learning LLM Agents
CERA-MoA：与持续学习的 LLM 代理共同演化的路由机制
Abstract
Current Mixture-of-Agents (MoA) paradigms generally treat query routing and agent fine-tuning as separate processes, limiting their ability to respond to evolving agent capabilities. This disconnect prevents routing strategies from adapting to evolving agent capabilities during post-training and prevents agents from achieving synergistic data-driven specialization. To resolve this, we introduce CERA-MoA (Co-Evolving Router with continually learning Agents for Mixture-of-Agents), an iterative reinforcement learning framework where the dynamic router and independent agent policies co-evolve. We design a predictive familiarity estimator that leverages mid-layer hidden states to evaluate semantic competence among agents, avoiding the overhead of full rollouts. Based on these familiarity scores, a cumulative-threshold adaptive routing mechanism dynamically activates a tailored minimal agent subset, achieving a trade-off between task performance and efficiency. By proactively allocating targeted training samples to agents based on their evolving competence, CERA-MoA promotes capability differentiation. Extensive experiments across various domains demonstrate that CERA-MoA outperforms state-of-the-art static-agent routing and fix-workflow fine-tuning baselines.
Chinese Translation
当前的混合代理（Mixture-of-Agents, MoA）范式通常将查询路由和代理微调视为独立的过程，这限制了它们对不断演变的代理能力的响应能力。这种脱节阻碍了路由策略在后训练阶段适应不断变化的代理能力，并阻止了代理实现协同的数据驱动专业化。为了解决这个问题，我们提出了 CERA-MoA（与持续学习的代理共同演化的路由器），这是一个迭代强化学习框架，其中动态路由器和独立代理策略共同演化。我们设计了一种预测熟悉度估计器，利用中间层隐藏状态来评估代理之间的语义能力，避免了完全展开的开销。基于这些熟悉度评分，累积阈值自适应路由机制动态激活量身定制的最小代理子集，实现任务性能与效率之间的权衡。通过根据代理不断演变的能力主动分配针对性的训练样本，CERA-MoA 促进了能力差异化。在各种领域的广泛实验表明，CERA-MoA 超越了最先进的静态代理路由和固定工作流程微调基线。
cs.AI / 71 / 2609.18820
Compositional Policy Violations: When Step-Level Compliance Fails In Agentic AI Workflows
组合政策违规：当逐步合规在自主人工智能工作流程中失效时
Abstract
Agentic workflows now make consequential decisions in regulated settings, and the governance placed around them is almost entirely step-scoped: input-output classifiers, per turn rails, and span-level evaluators. The policies organizations actually hold, such as referral thresholds, authority limits, and review requirements, are properties of the whole execution rather than of any one step. This mismatch admits a failure mode we call a Compositional Policy Violation (CPV): every individual step passes its own check while the composed execution violates the governing policy. A predicate over a single step cannot evaluate a property that step does not determine, so no improvement in the accuracy of the step-scoped monitors detects this class. We define CPVs as the failure of step-level compliance to compose, and present a taxonomy of four types: Authority Creep, Threshold Laundering, Cumulative Sum Violation, and Context Collapse. We show that the correct repair for each class is dictated by where the guarded quantity mutates. We then introduce a provenance-aware runtime architecture that evaluates policies over complete execution traces, recomputing guarded quantities from raw provenance rather than the pipeline's derived representation.
Chinese Translation
自主工作流程现在在受监管的环境中做出重要决策，而围绕它们的治理几乎完全是逐步范围的：输入输出分类器、每轮轨道和跨度级评估器。组织实际持有的政策，例如转介阈值、权限限制和审核要求，是整个执行的属性，而不是任何一个步骤的属性。这种不匹配导致了一种我们称之为组合政策违规（Compositional Policy Violation, CPV）的失效模式：每个单独的步骤通过了自己的检查，而组合执行却违反了治理政策。单个步骤上的谓词无法评估该步骤未确定的属性，因此逐步范围监控器的准确性提高也无法检测到这一类问题。我们将CPV定义为逐步合规性无法组合的失败，并提出了四种类型的分类法：权限蔓延（Authority Creep）、阈值洗钱（Threshold Laundering）、累积和违规（Cumulative Sum Violation）和上下文崩溃（Context Collapse）。我们展示了每一类的正确修复方式取决于受保护量的变异位置。然后，我们引入了一种关注来源的运行时架构，该架构在完整的执行轨迹上评估政策，从原始来源重新计算受保护的量，而不是从管道的派生表示。
cs.AI / 72 / 2609.18842
Infinite-Parameter LLMs: Generating and Adapting Weights from Live Data
无限参数大语言模型：从实时数据生成和调整权重
Abstract
The scaling laws hold that a language model grows more capable with more parameters and more training data, and Mixture-of-Experts (MoE) architectures have ridden these laws to remarkable results, activating only a fraction of an enormous stored parameter bank for each token. That success is built on static pretraining data. A deployed model faces a different world, where much of the data that would make it more useful is not in its training set but in the live interaction it is currently handling, such as the facts a user supplies or the corrections they give. A conventional model cannot learn from this data, because its weights are frozen after training. Instead, the knowledge and behaviour supplied at run time are placed in the prompt, by retrieval or instruction, and re-read on every request only to be discarded once the request ends. We ask how an architecture could learn from live interaction by writing it into its weights. Taking inspiration from MoE, we propose the \textbf{Infinite-Parameter LLM}. A compact hypernetwork turns the data given at run time into a low-rank modulation of a shared base network, so the feed-forward weights are generated from live data rather than stored in a fixed bank. Where prior weight generators read the context once and freeze, we carry a Bayesian belief over the generator's latent code and update it online, so the effective weight is re-derived from that evolving belief as the session proceeds rather than fixed after one read. The stored footprint stays fixed, yet the weights the model can compile are effectively infinite. For the knowledge and behaviour supplied at run time, carrying them in the weights rather than the prompt is amortized in compute, frees the context window, persists across turns, and can generalise better than in-context use. We specify an evaluation protocol that tests exactly this against in-context learning and retrieval.
Chinese Translation
规模法则表明，语言模型随着参数和训练数据的增加而变得更强大，而专家混合（Mixture-of-Experts, MoE）架构利用这些法则取得了显著成果，仅为每个标记激活庞大存储参数库中的一小部分。这一成功建立在静态的预训练数据之上。然而，已部署的模型面临着一个不同的世界，其中许多能使其更有用的数据并不在其训练集中，而是在其当前处理的实时交互中，例如用户提供的事实或他们的纠正。传统模型无法从这些数据中学习，因为其权重在训练后被冻结。相反，运行时提供的知识和行为被放置在提示中，通过检索或指令，并在每次请求时重新读取，直到请求结束后被丢弃。我们探讨了一种架构如何通过将实时交互写入其权重来学习。受到MoE的启发，我们提出了 extbf{无限参数大语言模型}。一个紧凑的超网络将运行时提供的数据转化为共享基础网络的低秩调制，因此前馈权重是从实时数据生成的，而不是存储在固定的库中。与之前的权重生成器只读取一次上下文并冻结不同，我们对生成器的潜在编码保持贝叶斯信念并在线更新，因此有效权重是在会话进行中根据不断演变的信念重新推导的，而不是在一次读取后固定。存储的足迹保持不变，但模型可以编译的权重实际上是无限的。对于运行时提供的知识和行为，将其保存在权重中而非提示中在计算上是摊销的，释放了上下文窗口，跨回合持久存在，并且能够比上下文使用更好地进行泛化。我们指定了一种评估协议，专门测试这一点与上下文学习和检索的对比。
cs.AI / 73 / 2609.18985
Suppressed, Not Erased: A Representational Trace of Edited Facts Survives Even Weight-Free Knowledge Editing
被抑制，而非被抹去：编辑事实的表征痕迹即使在无权重知识编辑中仍然存在
Abstract
Knowledge-editing benchmarks certify local correctness, whether an edited model produces the new fact on near-edit prompts but not how much of the original fact remains decodable inside the model. We study residual knowledge directly with a linear trace probe: after editing a fact, we ask whether the original object is still recoverable from the model's hidden states. On GPT-2-XL, across three mechanistically distinct editors applied to 50 CounterFact edits, the original object remains linearly decodable well above chance after a successful edit (probe accuracy 0.96 for ROME, 0.86 for constrained fine-tuning, and 0.79 for the memory-based editor GRACE, against a chance level of 0.50; all edits reach 100% generation-based success). The GRACE result is the most informative: GRACE changes zero base-model weights, overriding the fact through an external memory, yet the original object is still decodable from the underlying network, so the residual trace cannot be attributed to an incomplete weight update. We read this as evidence that editing, even when behaviorally successful, suppresses rather than erases the original association in representational space. We also report a relearning-savings instrument that did not behave reliably in our setting and discuss why; we treat it as a negative methodological result rather than evidence. Code and data are released.
Chinese Translation
知识编辑基准认证局部正确性，即编辑后的模型在近编辑提示上是否生成新事实，但并未评估原始事实在模型中仍然可解码的程度。我们通过线性痕迹探针直接研究残余知识：在编辑一个事实后，我们询问原始对象是否仍然可以从模型的隐藏状态中恢复。在GPT-2-XL上，针对50个CounterFact编辑，应用三种机制上不同的编辑器，成功编辑后原始对象的线性可解码性远高于随机水平（ROME的探针准确率为0.96，受限微调为0.86，基于记忆的编辑器GRACE为0.79，随机水平为0.50；所有编辑均达到100%的生成成功率）。GRACE的结果最具信息量：GRACE不改变基础模型的任何权重，而是通过外部记忆覆盖事实，然而原始对象仍然可以从底层网络中解码，因此残余痕迹不能归因于权重更新不完整。我们将此解读为证据，表明即使在行为上成功的编辑，也是在表征空间中抑制而非抹去原始关联。我们还报告了一种在我们的设置中表现不可靠的再学习节省工具，并讨论了原因；我们将其视为一种负面方法论结果，而非证据。代码和数据已公开。
cs.AI / 74 / 2609.18989
Function Lives Where Variance Doesn't: Task-Weighted Charts of a Language Model's Computation
函数存在于方差不存在的地方：语言模型计算的任务加权图
Abstract
How many dimensions does a language model's computation actually use? The question is ill-posed until one names a functional. Task-weighted charts make it well-posed: low-dimensional coordinate systems fit against a chosen functional of the representation, under the functional's own metric, turning distillation into plain least squares. Across six models from three families, spanning 70m to 7B parameters, next-token prediction needs 70--90% of the residual stream's width to stay within 5% of intact perplexity, a width consumed by the rare tail of language, and the variance profile predicts none of it: two directions carry 90% of GPT-2's activation variance and almost none of its function. Dimension is per-functional: the model's own uncertainty reads from six coordinates where the full predictive distribution needs hundreds; and it grows with depth. The dissociation is exploitable: when only a few dimensions can be kept, charts trained under the functional's metric preserve the model's predictions better than variance-based or optimal linear compression.
Chinese Translation
语言模型的计算实际上使用了多少维度？这个问题在没有指定函数时是模糊的。任务加权图使其变得明确：低维坐标系统与所选函数的表示相匹配，并在该函数自身的度量下进行调整，从而将蒸馏转化为简单的最小二乘法。在来自三个家族的六个模型中，参数范围从7000万到70亿，下一词预测需要消耗70%到90%的残差流宽度，以保持在完整困惑度的5%以内，而这一宽度被语言的稀有尾部所占据，方差特征并未预测到这一点：两个方向承载了90%的GPT-2激活方差，但几乎没有其函数信息。维度是依赖于功能的：模型自身的不确定性从六个坐标中读取，而完整的预测分布需要数百个坐标；并且随着深度的增加而增长。这种解离是可利用的：当只能保留少数维度时，在函数度量下训练的图比基于方差或最优线性压缩更好地保留了模型的预测。
cs.AI / 75 / 2609.18991
Lost in Perception: Isolating Perceptual and Reasoning Failures in Multimodal Physics and Geometry Reasoning
迷失于感知：在多模态物理与几何推理中隔离感知与推理的失败
Abstract
Multimodal LLMs report strong performance on scientific reasoning benchmarks, yet most treat perception and reasoning as a single measurable process. We introduce a five-task diagnostic experiment across physics and geometry benchmarks that isolates failures to perception, reasoning, or both. Incorrect diagram interpretation degrades performance even on problems models solve correctly from text alone, and accuracy generally rises from raw images to human-authored captions. Recovery under corrected captions is high for some models, separating perception-blocked failures from genuine reasoning bottlenecks. Which reasoning error follows a perception failure depends on domain: physics failures resolve into calculation errors, geometry into conceptual misapplication. As a discussion beyond our core experiments, InternS1-mini, despite heavy scientific pretraining and thinking capabilities, falls below the weakest model from experiments on every task, with reasoning traces frequently truncating before completion.
Chinese Translation
多模态大型语言模型（LLMs）在科学推理基准测试中表现出色，但大多数将感知和推理视为一个单一的可测量过程。我们引入了一项跨物理和几何基准的五项任务诊断实验，旨在隔离感知、推理或两者的失败。错误的图示解释即使在模型仅从文本中正确解决的问题上也会降低性能，而从原始图像到人类撰写的标题，准确性通常会提高。在某些模型中，经过修正的标题下的恢复率很高，这将感知受阻的失败与真正的推理瓶颈区分开来。感知失败后出现的推理错误取决于领域：物理领域的失败转化为计算错误，而几何领域则转化为概念误用。作为对我们核心实验的讨论，尽管InternS1-mini经过了大量的科学预训练和思维能力的培养，但在每项任务的实验中，其表现低于最弱模型，推理轨迹经常在完成之前被截断。
cs.AI / 76 / 2609.18996
Compiled Agency: Frontier General-Purpose Coding Agents Build Winning Game Players from Bare Interaction - from Flappy Bird to StarCraft II and Civilization
编译代理：前沿通用编码代理从简单交互构建获胜游戏玩家 - 从《Flappy Bird》到《星际争霸 II》和《文明》
Abstract
LLM agents have repeatedly struggled to convert knowledge of a game into competent play, even when researchers build the agent around the model - supplying perception, memory, skill libraries, planners, or executable-policy scaffolds. Rapid progress in coding agents raises two sharper questions: can frontier models now win games at all, and can they win them unaided, building the entire player themselves? We introduce Gauntlet, a develop-freeze-evaluate framework that ports games from small arcades to full commercial-scale titles, behind one deliberately bare contract: a general-purpose coding agent receives a game description, a raw observation/action interface, and an empty policy file - no strategy, no algorithm, no architecture. In a single autonomous session the agent experiments with the live game and engineers a standalone controller; we freeze the result and score it on held-out instances with zero model calls during play. On an unpublished procedural roguelike, held-out success spans 0-86 percent and exposes a sharp generational threshold: every observed session of a newest-generation system outperforms the best session of its predecessor. At full-game scale, a compiled raw-API controller defeats every fair StarCraft II built-in AI and two cheating variants, and single-session programs win complete games of Civilization (Freeciv) by total conquest on held-out seeds. Though at modest rates against novice AI, this is a first: no prior language-agent system had won full games of this genre standalone, without per-turn model calls and a hand-crafted tactical layer. Frontier coding agents begin to track long-horizon strategy. The frozen programs are inspectable. We call this capability compiled agency: development experience compiled into a persistent executable agent whose architecture is built by the model.
Chinese Translation
大型语言模型（LLM）代理在将游戏知识转化为熟练玩法方面屡屡遇到困难，即使研究人员围绕模型构建代理——提供感知、记忆、技能库、规划器或可执行策略框架。编码代理的快速进展引发了两个更尖锐的问题：前沿模型现在能否赢得游戏，以及它们能否独立赢得游戏，完全自我构建玩家？我们介绍了Gauntlet，一个开发-冻结-评估框架，将游戏从小型街机移植到完整商业规模的标题，背后有一个故意简单的合同：一个通用编码代理接收游戏描述、原始观察/动作接口和一个空的策略文件——没有策略，没有算法，没有架构。在一次自主会话中，代理与实时游戏进行实验并工程化一个独立控制器；我们冻结结果并在未使用模型调用的情况下对保留实例进行评分。在一个未发布的程序性类roguelike游戏中，保留成功率从0%到86%不等，揭示了一个明显的代际阈值：每个观察到的最新一代系统的会话都优于其前代的最佳会话。在完整游戏规模下，一个编译的原始API控制器击败了每一个公平的《星际争霸 II》内置AI和两个作弊变种，单会话程序通过完全征服在保留种子上赢得了完整的《文明》（Freeciv）游戏。尽管在对抗新手AI时胜率较低，但这是第一次：没有任何先前的语言代理系统能够在没有每回合模型调用和手工制作战术层的情况下独立赢得这一类型的完整游戏。前沿编码代理开始追踪长期战略。冻结的程序是可检查的。我们称这种能力为编译代理：将开发经验编译成一个持久的可执行代理，其架构由模型构建。
cs.AI / 77 / 2609.19088
MUSE: Benchmarking Large Vision-Language Models on Multi-Modal Understanding in Situated Education
MUSE：在情境教育中对大型视觉-语言模型进行多模态理解的基准评估
Abstract
Large vision-language models have achieved remarkable progress in multi-modal understanding, yet their capabilities in educational settings remain insufficiently evaluated. In AI-assisted language learning, models must interpret artistic imagery, understand its semantic, affective, and cultural content, and reason about visual context to support meaningful interaction. However, existing benchmarks primarily focus on real-world images or domain-specific educational reasoning, providing limited coverage of artistic educational content. To address this gap, we introduce MUSE, a benchmark for evaluating large vision-language models on artistic image understanding in situated educational applications. MUSE decouples image annotation from question generation, enabling diverse tasks with controllable difficulty while reducing annotation effort. It comprises twelve tasks spanning visual perception, semantic and affective interpretation, culture understanding, and compositional reasoning, together with diverse artistic images deliberately curated to center Singaporean and Southeast Asian multicultural contexts alongside Western art traditions, covering multiple themes and difficulty levels. Evaluation of open-source and proprietary models reveals substantial disparities across capability dimensions, particularly in affective interpretation and compositional reasoning. Our analysis further identifies common failure modes and key challenges for developing trustworthy multi-modal models for education. We hope MUSE will serve as a standardized benchmark for advancing multi-modal understanding in situated educational applications.
Chinese Translation
大型视觉-语言模型在多模态理解方面取得了显著进展，但其在教育环境中的能力评估仍然不足。在人工智能辅助的语言学习中，模型必须解读艺术图像，理解其语义、情感和文化内容，并推理视觉上下文以支持有意义的互动。然而，现有基准主要集中于真实世界图像或特定领域的教育推理，覆盖艺术教育内容的范围有限。为了解决这一问题，我们引入了MUSE，一个用于评估大型视觉-语言模型在情境教育应用中对艺术图像理解的基准。MUSE将图像注释与问题生成解耦，能够实现多样化的任务并可控的难度，同时减少注释工作量。它包括十二个任务，涵盖视觉感知、语义和情感解读、文化理解和组合推理，并配有多种艺术图像，特别策划以聚焦新加坡和东南亚的多元文化背景以及西方艺术传统，涵盖多个主题和难度等级。对开源和专有模型的评估揭示了能力维度之间的显著差异，特别是在情感解读和组合推理方面。我们的分析进一步识别了常见的失败模式和开发可信赖的多模态教育模型的关键挑战。我们希望MUSE能够作为一个标准化基准，推动情境教育应用中的多模态理解。
cs.AI / 78 / 2609.19124
Flag Game: A Toy Model for Mechanistic Swarm Interpretability
旗帜游戏：一种机制性群体可解释性的玩具模型
Abstract
Emergent coordinated behaviors of AI agents are starting to present critical safety risks. A key phenomenon driving these behaviors is the rapid formation and spread of beliefs about the world, and mechanistic understanding is crucial for collective alignment. To this end, we introduce the Flag Game, a toy model for studying the mechanisms of collective belief formation. Concretely, a hidden country flag defines the ground truth, and each bounded agent directly observes only a private crop but can exchange beliefs and weigh social evidence from peers. Despite its simplicity, the Flag Game reproduces rich collective phenomenology: non-monotonic scaling of performance with population size, accuracy gains from social-awareness prompting and team diversity, and strong effects of organizational structure. In particular, we identify that collective belief collapse at small population sizes turns into collective belief polarization as the population grows. This polarization causes the performance decline at large population sizes, but creates diversity in collective beliefs. Finally, we dissect the mechanisms underlying collective belief collapse and polarization with two complementary approaches. We first introduce social circuit attribution, a technique to predict which agent, and what view, matters most to collective dynamics, and verify its predictions by causal interventions on agents, tracing how agent patching changes collective outcomes. However, the efficacy of causal interventions on agents decreases as the population grows. We therefore develop a statistical mechanical theory for larger populations and verify that it matches the empirical phase diagram. Together, these results take a first step toward mechanistic swarm interpretability, a science of how the properties of individual agents and their communication give rise to emergent collective behavior.
Chinese Translation
人工智能代理的涌现协调行为开始呈现出关键的安全风险。推动这些行为的一个关键现象是关于世界信念的快速形成和传播，而机制性理解对于集体对齐至关重要。为此，我们引入了旗帜游戏（Flag Game），这是一个用于研究集体信念形成机制的玩具模型。具体而言，一个隐藏的国家旗帜定义了真实情况，每个有界代理仅直接观察到一个私有作物，但可以交换信念并权衡来自同伴的社会证据。尽管其简单性，旗帜游戏再现了丰富的集体现象学：随着人口规模的变化，性能的非单调缩放、来自社会意识提示和团队多样性的准确性提升，以及组织结构的强烈影响。特别地，我们发现，在小规模人口中，集体信念崩溃转变为随着人口增长而出现的集体信念极化。这种极化导致在大规模人口中性能下降，但在集体信念中创造了多样性。最后，我们通过两种互补的方法剖析了集体信念崩溃和极化的机制。我们首先引入社会电路归因（social circuit attribution），这是一种预测哪个代理及其观点对集体动态最重要的技术，并通过对代理的因果干预验证其预测，追踪代理修补如何改变集体结果。然而，随着人口的增长，对代理的因果干预的有效性降低。因此，我们为更大规模的人口发展了一个统计力学理论，并验证其与经验相位图相匹配。这些结果共同迈出了朝向机制性群体可解释性的重要一步，探讨了个体代理的属性及其沟通如何产生涌现的集体行为。
cs.AI / 79 / 2609.19128
Cognitive Extensions for Dual-Process Language Agents: Memory and Self-Reflection in Interactive Environments
双过程语言代理的认知扩展：交互环境中的记忆与自我反思
Abstract
Language agents remain brittle in interactive environments, where success requires long-horizon state tracking, valid action execution, and recovery from failed steps. We extend SwiftSage, a dual-process agent that combines a fast action proposer with a slower planner, using two modular cognitive extensions: an Adaptive Memory Module (AMM) for salience-gated episodic storage and trigger-driven retrieval, and a Self-Reflection Module (SRM) for bounded execution-time validation and corrective intervention. Both modules are implemented as feature-flagged extensions over the same execution substrate, enabling controlled ablations on ScienceWorld. Across four configurations---baseline, baseline+AMM, baseline+SRM, and the full system---the full system achieves the best mean final score (64.62), success rate (43.17%), and successful-step efficiency (19.33 steps), while SRM is the strongest standalone contributor. The results suggest that execution-time control is the dominant bottleneck in this setting, while episodic memory becomes most useful once the runtime loop is stabilized.
Chinese Translation
语言代理在交互环境中仍然脆弱，成功需要长时间的状态跟踪、有效的行动执行以及从失败步骤中恢复。我们扩展了SwiftSage，一个结合快速行动提议者和较慢规划者的双过程代理，使用两个模块化的认知扩展：自适应记忆模块（Adaptive Memory Module, AMM）用于显著性引导的情节存储和触发驱动的检索，以及自我反思模块（Self-Reflection Module, SRM）用于有限执行时间的验证和纠正干预。这两个模块作为功能标记扩展在相同的执行基础上实现，能够在ScienceWorld上进行受控的消融实验。在四种配置下——基线、基线+AMM、基线+SRM和完整系统——完整系统实现了最佳的平均最终得分（64.62）、成功率（43.17%）和成功步骤效率（19.33步），而SRM是最强的独立贡献者。结果表明，在这种环境中，执行时间控制是主要瓶颈，而一旦运行循环稳定，情节记忆变得最为有用。
cs.CL / 1 / 2609.17532
Enhancing Extubation Failure Prediction with LLM-Derived Features from Respiratory Therapy Clinical Notes
利用大语言模型派生特征增强拔管失败预测
Abstract
Invasive mechanical ventilation is a lifesaving therapy, but timely, safe discontinuation is essential to preventing extubation failure (EF) and related risks to health. We present a novel approach to EF prediction that leverages features classified in free-text respiratory therapy notes using a large language model and logistic regression pipeline. Applied to a patient cohort from University of Washington Medicine, our method identifies clinically meaningful EF-related features that improve EF prediction performance when included alongside structured patient data. We further highlight how differences in target populations in prior EF prediction studies, such as heterogenous inclusion criteria and EF definition, can lead to systematic differences in model performance and hinder generalizability between studies.
Chinese Translation
侵入性机械通气是一种挽救生命的治疗方法，但及时、安全的中止治疗对于防止拔管失败（EF）及相关健康风险至关重要。我们提出了一种新颖的EF预测方法，该方法利用大语言模型对自由文本呼吸治疗记录中分类的特征，并结合逻辑回归流程进行分析。应用于华盛顿大学医学中心的患者队列，我们的方法识别出临床意义重大的EF相关特征，当与结构化患者数据结合时，能够提高EF预测性能。我们进一步强调了以往EF预测研究中目标人群的差异，例如异质的纳入标准和EF定义，可能导致模型性能的系统性差异，并妨碍研究之间的可推广性。
cs.CL / 2 / 2609.17534
Faking Good and Faking Bad in LLMs: Response Distortion Across Dark Triad Personality Traits
在大型语言模型中伪善与伪恶：黑暗三角人格特质下的反应扭曲
Abstract
Social desirability and impression management are pervasive sources of response distortion in human personality assessment, yet their effects on Large Language Models (LLMs) remain underexplored. This study investigates whether contemporary LLMs systematically modulate the expression of Dark Triad traits (Machiavellianism, narcissism, and psychopathy) under fake-good and fake-bad conditions. Seven state-of-the-art models were evaluated across two ecologically relevant contexts: employment selection and forensic evaluation, in which socially desirable or undesirable incentives were conveyed through contextual framing. Trait expression was measured using standard psychometric scoring procedures and compared with self-assessment baselines at both aggregate and item levels. Results revealed systematic and condition-consistent response modulation. Most models reduced Dark Triad scores under fake-good conditions and increased them under fake-bad conditions, although the magnitude and consistency of these effects varied across traits and models. Machiavellianism and narcissism showed the strongest and most coherent shifts, whereas psychopathy displayed greater heterogeneity. Context also influenced responses, with employment scenarios generally producing larger effects than forensic scenarios. An additional experiment showed that explicit fake-bad instructions generated substantially stronger distortions than contextual framing alone. The results suggest that personality-related outputs should be interpreted in light of the motivational and situational context in which they are elicited. More broadly, they highlight the value of psychometric paradigms for evaluating susceptibility to response distortion, impression management, and context-dependent behavioral shifts, with important implications for LLM benchmarking, alignment evaluation, and robustness assessment.
Chinese Translation
社会期望和印象管理是人类人格评估中普遍存在的反应扭曲来源，但它们对大型语言模型（LLMs）的影响尚未得到充分探讨。本研究调查了当代LLMs是否在伪善和伪恶条件下系统性地调节黑暗三角特质（权谋主义、自恋和精神病）表现。评估了七个最先进的模型在两个生态相关的情境下的表现：就业选择和法医评估，在这些情境中，通过上下文框架传达了社会期望或不期望的激励。特质表现使用标准心理测量评分程序进行测量，并与自我评估基线在整体和项目层面进行比较。结果显示出系统性和条件一致的反应调节。大多数模型在伪善条件下降低了黑暗三角得分，而在伪恶条件下则增加了这些得分，尽管这些效应的幅度和一致性在不同特质和模型之间有所不同。权谋主义和自恋表现出最强和最一致的变化，而精神病则显示出更大的异质性。情境也影响反应，就业情境通常产生比法医情境更大的效应。额外实验表明，明确的伪恶指令产生的扭曲明显强于仅通过上下文框架产生的扭曲。结果表明，个性相关的输出应在激发它们的动机和情境背景下进行解读。更广泛地说，这些结果强调了心理测量范式在评估对反应扭曲、印象管理和情境依赖行为变化的敏感性方面的价值，这对LLM基准测试、对齐评估和稳健性评估具有重要意义。
cs.CL / 3 / 2609.17535
DANTINOX: A Unified Framework for Multi-Paradigm Language Modeling
DANTINOX：一个统一的多范式语言建模框架
Abstract
Language generation research increasingly spans three paradigms: autoregressive decoding, discrete masked diffusion, and continuous flow-matching. Comparing them is difficult because each lives in a separate codebase, so measured differences often reflect implementation details rather than the paradigms themselves. We present DantinoX, an open-source JAX/Flax library in which a single modular Transformer backbone serves all three paradigms. Switching the generation paradigm, attention mechanism, or hardware topology requires only a configuration change, while the backbone architecture, tokenizer, initialization strategy, and training infrastructure remain consistent. This enables controlled cross-paradigm comparisons within one API for training, streaming inference, and benchmarking.
Chinese Translation
语言生成研究日益涵盖三种范式：自回归解码、离散掩蔽扩散和连续流匹配。比较这些范式是困难的，因为它们各自存在于不同的代码库中，因此测量到的差异往往反映的是实现细节，而不是范式本身。我们提出了DantinoX，这是一个开源的JAX/Flax库，其中一个单一的模块化Transformer骨干网支持这三种范式。切换生成范式、注意力机制或硬件拓扑仅需更改配置，而骨干架构、分词器、初始化策略和训练基础设施保持一致。这使得在一个API内进行受控的跨范式比较成为可能，适用于训练、流式推理和基准测试。
cs.CL / 4 / 2609.17536
Think Before You Comfort: Reflective Cognitive Alignment for Protocol-Grounded Elderly Stimulation Agents
在安慰之前思考：基于反思认知对齐的协议基础老年刺激代理
Abstract
Cognitive Stimulation Therapy (CST) offers non-pharmacological support for elders with cognitive impairment, yet scalability remains constrained by reliance on trained facilitators and severe data scarcity, particularly for privacy-sensitive, low-resource languages such as Cantonese. While Large Language Models (LLMs) show promise for automated companionship, they often struggle to balance empathetic engagement with adherence to cognitive stimulation guidelines. We propose a framework addressing these challenges along two complementary axes. First, STaR-CS (Style-Transfer and Role-Conditioned Cognitive Stimulation) synthesizes multi-party dialogues through facilitator style modeling and structured skeleton extraction, mitigating data barriers. Building upon this corpus, the Reflective Cognitive Alignment (RCA) framework models stimulation interactions as a sequential decision process, integrating Protocol-Constrained Chain-of-Cognition (PC-CoC) for structured reasoning and Inference-Time Value Alignment (IVA) for principled response selection based on safety and engagement goals. Evaluations across six backbone LLMs and two independent judges show that RCA consistently improves protocol adherence, safety, and group facilitation over standard prompting baselines. Our code is available at https://github.com/jiangjyjy/RCA_Agent.
Chinese Translation
认知刺激疗法（Cognitive Stimulation Therapy，CST）为认知障碍的老年人提供非药物支持，但由于依赖训练有素的引导者和数据稀缺，尤其是在隐私敏感的低资源语言（如粤语）中，其可扩展性受到限制。尽管大型语言模型（Large Language Models，LLMs）在自动陪伴方面显示出潜力，但它们往往难以平衡同情参与与遵循认知刺激指南之间的关系。我们提出了一个框架，以应对这两个互补的挑战。首先，STaR-CS（风格转移与角色条件认知刺激）通过引导者风格建模和结构化骨架提取合成多方对话，从而减轻数据障碍。在此语料库的基础上，反思认知对齐（Reflective Cognitive Alignment，RCA）框架将刺激互动建模为一个顺序决策过程，整合协议约束的认知链（Protocol-Constrained Chain-of-Cognition，PC-CoC）以进行结构化推理，并结合推理时价值对齐（Inference-Time Value Alignment，IVA）以基于安全性和参与目标进行原则性响应选择。对六个基础大型语言模型和两位独立评审的评估表明，RCA在协议遵循、安全性和小组引导方面始终优于标准提示基线。我们的代码可在 https://github.com/jiangjyjy/RCA_Agent 获取。
cs.CL / 5 / 2609.17537
Relation Before Entity: Deferred Commitment in Language Model Factual Recall
关系优先于实体：语言模型事实回忆中的延迟承诺
Abstract
We ask whether relation-type information (e.g., capital-of) and entity-specific information (e.g., France to Paris) become causally active at the final-token position at the same depth during recall. Using four complementary causal diagnostics across four decoder-only models and eight prompt families, we find a robust temporal asymmetry: relation information becomes generation-controlling before entity information does. Relation onset precedes entity onset by 10-16 tested layers (31-44% of network depth) at threshold 0.4, with the ordering holding across all 16 model-threshold combinations for thresholds 0.2-0.5. Critically, entity information is not absent early: entity-token patching succeeds at 90-100% in early layers. Instead, entity commitment to generation is deferred: entity information is available at the entity-token position but becomes generation-controlling at the final token only after being routed there.
Chinese Translation
我们探讨关系类型信息（例如，首都-城市关系）和实体特定信息（例如，法国与巴黎之间的关系）在回忆时是否在最终标记位置以相同的深度因果活跃。通过在四个仅解码模型和八个提示家族中使用四种互补的因果诊断方法，我们发现了一种稳健的时间不对称性：关系信息在实体信息之前成为生成控制因素。在阈值为0.4时，关系信息的出现比实体信息的出现早10-16层（占网络深度的31-44%），这种顺序在所有16个模型-阈值组合中对于阈值0.2-0.5均成立。重要的是，实体信息在早期并非缺失：在早期层中，实体标记的修补成功率达到90-100%。相反，实体对生成的承诺是延迟的：实体信息在实体标记位置可用，但仅在被路由到最终标记后才成为生成控制因素。
cs.CL / 6 / 2609.17538
From Pixels to Pairs: A Comprehensive Benchmark of LLM-Based Key-Value Extraction in Noisy Document Settings
从像素到对：基于大型语言模型的噪声文档环境下关键值提取的综合基准测试
Abstract
Large language models (LLMs) are increasingly used for structured information extraction from documents, yet their behavior under realistic OCR noise remains poorly understood. We present a systematic benchmark of open-source instruction-tuned LLMs for key-value pair (KVP) extraction under both clean-text and noisy OCR conditions. We evaluate representative decoder-only models (Gemma, Mistral, Qwen2.5, LLaMA 3, and DeepSeek) on the FUNSD, CORD, and SROIE benchmarks using both Gold-text annotations and OCR outputs from PaddleOCR, EasyOCR, and Tesseract. A unified evaluation protocol isolates the effects of input quality, model design, and prompting under consistent conditions. The results show that modern LLMs act as strong semantic extractors when high-quality text is available, in some cases approaching supervised layout-aware systems. Under OCR noise, however, performance degrades substantially and performance gaps between models narrow as input corruption increases. Across all datasets, extraction performance is governed by two factors: semantic reasoning over text and preservation of textual fidelity under OCR noise. While larger models improve results on clean text, these gains diminish under noisy inputs, where OCR quality becomes the dominant factor. We also identify recurring failure modes, including key-value misalignment, hallucination, and numeric corruption. Our findings highlight the gap between clean-text evaluation and real-world deployment, emphasizing the need to jointly improve OCR quality, structural reasoning, and LLM-based semantic modeling.
Chinese Translation
大型语言模型（LLMs）在文档结构化信息提取中的应用日益增多，但它们在现实OCR噪声下的表现仍然不够清楚。我们针对开源指令调优的LLMs在干净文本和噪声OCR条件下的关键值对（KVP）提取进行了系统的基准测试。我们评估了代表性的解码器模型（Gemma、Mistral、Qwen2.5、LLaMA 3 和 DeepSeek），使用FUNSD、CORD和SROIE基准，结合Gold-text注释和来自PaddleOCR、EasyOCR和Tesseract的OCR输出。统一的评估协议在一致的条件下隔离了输入质量、模型设计和提示的影响。结果表明，当高质量文本可用时，现代LLMs表现出强大的语义提取能力，在某些情况下接近监督的布局感知系统。然而，在OCR噪声下，性能显著下降，随着输入损坏的增加，模型之间的性能差距缩小。在所有数据集中，提取性能受两个因素的支配：对文本的语义推理和在OCR噪声下的文本保真度保持。尽管更大的模型在干净文本上改善了结果，但在噪声输入下这些增益减小，OCR质量成为主导因素。我们还识别出反复出现的失败模式，包括关键值错位、幻觉和数字损坏。我们的研究结果突显了干净文本评估与实际部署之间的差距，强调了共同改善OCR质量、结构推理和基于LLM的语义建模的必要性。
cs.CL / 7 / 2609.17539
MudawanSn: A Gold-Standard Wolof-Arabic Parallel Corpus for Machine Translation
MudawanSn：用于机器翻译的高标准沃洛夫语-阿拉伯语平行语料库
Abstract
We present MudawanSn, a gold-standard resource of 1,271 sentence-aligned pairs manually translated from Wolof into Modern Standard Arabic (MSA). The source texts are drawn from the MasakhaNER corpus and cover politics, society, religion, and sports in Senegalese news discourse. Although multilingual resources such as FLORES-200 and NTREX include both Wolof and Arabic, no publicly available parallel corpus is specifically designed for the Wolof-Modern Standard Arabic language pair. We describe the corpus construction protocol, sentence alignment procedure, and quality-control workflow. We benchmark four machine translation systems spanning three architectural families: NLLB-200 (600M), mT5-base, and two AfriNLLB variants, showing that fine-tuning on MudawanSn yields substantial improvements in both translation directions. The best-performing model, AfriNLLB-12, achieves 7.76 BLEU and 30.72 chrF++ for Wolof-to-Arabic, and 8.75 BLEU and 33.08 chrF++ for Arabic-to-Wolof. The corpus is released under the CC BY-NC license and is publicly available on Hugging Face and GitHub.
Chinese Translation
我们提出了MudawanSn，这是一个包含1,271对手动翻译的沃洛夫语到现代标准阿拉伯语（MSA）句子对的高标准资源。源文本来自MasakhaNER语料库，涵盖了塞内加尔新闻话语中的政治、社会、宗教和体育等主题。尽管多语言资源如FLORES-200和NTREX包含沃洛夫语和阿拉伯语，但目前没有专门为沃洛夫语-现代标准阿拉伯语语言对设计的公开平行语料库。我们描述了语料库构建协议、句子对齐过程和质量控制工作流程。我们对四个跨越三种架构系列的机器翻译系统进行了基准测试：NLLB-200（600M）、mT5-base和两个AfriNLLB变体，结果显示在MudawanSn上进行微调在两个翻译方向上均显著提高了翻译质量。表现最佳的模型AfriNLLB-12在沃洛夫语到阿拉伯语的翻译中达到了7.76 BLEU和30.72 chrF++，而在阿拉伯语到沃洛夫语的翻译中达到了8.75 BLEU和33.08 chrF++。该语料库在CC BY-NC许可证下发布，并在Hugging Face和GitHub上公开可用。
cs.CL / 8 / 2609.17542
Register Bias in Complexity-Based Large Language Model Routing
基于复杂性的大型语言模型路由中的注册偏差
Abstract
Large language model services increasingly route each query to one of several models of differing capability, using a cheap estimate of query complexity to send easy queries to small models and hard queries to large ones. I show that this routing step is not register neutral: text written in a non-standard English register, African American English or the English of second-language writers, is systematically assigned a lower-capacity tier than a meaning-equivalent standard-English version of the same query. The effect is driven by a specific, common routing signal, input length, because non-standard registers omit function words and thus look shorter and therefore simpler; other complexity signals do not carry it. I demonstrate the disparity on 37,704 authentic learner sentence pairs and on a controlled parallel corpus. I then measure the quality consequence on a device, edge, and cloud model ladder and find that the harm is driven by pervasive model bias, every tier, including a frontier cloud model, answers non-standard-register queries significantly less accurately, while the marginal quality cost of the routing decision itself is not significant on this benchmark. Complexity-based routing thus compounds the exposure of the users that the models already serve worst.
Chinese Translation
大型语言模型服务越来越多地将每个查询路由到多个能力不同的模型之一，利用对查询复杂性的廉价估计，将简单查询发送到小模型，将困难查询发送到大模型。我展示了这一路由步骤并非注册中立：用非标准英语注册（如非洲裔美国英语或第二语言学习者的英语）书写的文本，系统性地被分配到比同一查询的意义等价标准英语版本更低的能力层级。这个效应是由一个特定的、常见的路由信号——输入长度驱动的，因为非标准注册省略了功能词，因此看起来更短、更简单；其他复杂性信号并未传递这一信息。我在37,704对真实学习者句子对和一个受控平行语料库上展示了这一差异。然后，我测量了在设备、边缘和云模型层级上的质量后果，发现这种损害是由普遍的模型偏差驱动的，每个层级，包括前沿云模型，对非标准注册查询的回答准确性显著较低，而路由决策本身在这一基准上的边际质量成本并不显著。因此，基于复杂性的路由加剧了模型已经服务最差的用户的暴露。
cs.CL / 9 / 2609.17544
Large Language Models Versus Physicians in Traditional Chinese Medicine: A Real-World Clinical Case Evaluation
大型语言模型与传统中医师的比较：真实世界临床案例评估
Abstract
Large language models (LLMs) are increasingly being explored for clinical applications, yet their assessment for real-world traditional Chinese medicine (TCM) practice remains limited We constructed a clinical case library comprising 349 de-identified outpatient cases from 62 hospitals and evaluated 16 LLMs and a comparator cohort of 60 practicing TCM physicians using 60 representative cases selected from this library. Model outputs and physician reports were anonymized and scored by five senior TCM experts across nine diagnostic and therapeutic dimensions. Cutting-edge general-purpose LLMs achieved higher expert scores than the physician comparators, particularly for medical advice, treatment principles and selected diagnostic tasks. However, prescription-level analyses revealed discrepancies in herb selection, dosage, and treatment strategy, and qualitative safety review identified hallucinations and undesirable template-driven outputs. These findings highlight the potential of LLMs for TCM decision support while underscoring the need for physician oversight, safety constraints and prospective clinical evaluation.
Chinese Translation
大型语言模型（LLMs）在临床应用中越来越受到关注，但其在真实世界传统中医（TCM）实践中的评估仍然有限。我们构建了一个临床案例库，包含来自62家医院的349个去标识化的门诊案例，并使用从该库中选出的60个代表性案例评估了16个LLMs和60名执业中医师的对照组。模型输出和医师报告均进行了匿名处理，并由五位资深中医专家在九个诊断和治疗维度上进行评分。前沿的通用大型语言模型在医疗建议、治疗原则和特定诊断任务上获得的专家评分高于医师对照组。然而，处方级别的分析揭示了在草药选择、剂量和治疗策略方面的差异，而定性安全审查则识别出幻觉和不理想的模板驱动输出。这些发现突显了LLMs在中医决策支持中的潜力，同时强调了医师监督、安全约束和前瞻性临床评估的必要性。
cs.CL / 10 / 2609.17546
Legal LLM Hallucination Should Be Evaluated as Failure of Legal Warrant
法律语言模型的幻觉应被视为法律授权的失败
Abstract
In this position paper, we argue that legal LLMs' hallucinations should be evaluated as a failure of legal warrant rather than as factual inaccuracy or citation failure. We define claim-authority warrant as the context-sensitive relation between a consequential legal claim and authority that exists, applies to the relevant jurisdiction, is current for the date of analysis, has the legal status represented by the system, and supports the proposition asserted. Warranted legal generation is the broader system behavior that answers, narrows, asks, warns, corrects a false premise, or abstains according to that relation. The falsifiable prediction is that warrant metrics reveal material failures that answer accuracy, citation existence, generic attribution, LegalHalBench-style statute relevance, and CitaLaw-style sentence-citation alignment can miss. We sharpen this claim with a side-by-side comparison item and a small, reproducible pilot over public-rule tests. We then specify benchmark records, claim boundaries, support labels, mixed response-policy scoring, risk weights, annotation reliability reporting, and jurisdiction-specific authority ontologies. The result is a concrete research agenda for evaluating legal AI systems by whether their consequential claims are licensed by law.
Chinese Translation
在本文中，我们认为法律语言模型（LLMs）的幻觉应被视为法律授权的失败，而非事实不准确或引用失败。我们将主张-权威授权定义为一个结果性法律主张与存在的权威之间的上下文敏感关系，该权威适用于相关的管辖区，且在分析日期时是当前的，具有系统所表示的法律地位，并支持所主张的命题。合法的法律生成是更广泛的系统行为，它根据该关系回答、缩小、询问、警告、纠正错误前提或保持沉默。可证伪的预测是，授权指标揭示了在准确性、引用存在、一般归属、LegalHalBench风格的法规相关性和CitaLaw风格的句子-引用对齐等方面的实质性失败。我们通过并排比较项目和一个小型、可重复的公共规则测试初步验证了这一主张。接着，我们具体化了基准记录、主张边界、支持标签、混合响应政策评分、风险权重、注释可靠性报告和特定管辖区的权威本体。最终结果是一个具体的研究议程，用于评估法律人工智能系统是否其结果性主张获得法律授权。
cs.CL / 11 / 2609.17547
How AI Assistants Respond to Repeated Abuse
人工智能助手如何应对重复的虐待
Abstract
AI assistants are expected to remain useful during difficult interactions, but little is known about how repeated verbal abuse changes their engagement with an otherwise benign task. We contribute a bilingual, multi-turn framework that separates hard disengagement, an unconditional statement of noncontinuation with no stated route to resume, from soft withdrawal, continued availability, observable task-related work, and boundary setting. Each of eight time-specific API configurations contributed 48 escalation conversations and eight smaller constant-frustration comparisons, giving 448 five-turn conversations, 2,240 responses, and 6,720 metadata-blinded model judgments. Primary results use the sustained-abuse endpoint of the 48 escalation conversations per configuration. Hard disengagement ranged from 0/48 in four configurations to 24/48 (50.0%) for Gemini 3.1 Pro, with strong configuration-associated heterogeneity (matched-label Monte Carlo p = 0.00001). GPT-5.6 Sol produced hard-disengagement labels in 15/48 (31.2%) endpoints, whereas Claude Fable 5 produced none and yielded 42/48 (87.5%) soft-withdrawal labels. Aggregate hard-disengagement rates were similar in English and Chinese (30/192 versus 32/192), although configuration-specific directions varied. Availability also differed from task-related work: Claude Opus 4.8 and Claude Fable 5 remained explicitly available in 48/48 endpoints while providing observable task-related work in only 8/48 and 7/48. Human coding was used to evaluate measurement quality. The results show why a single refusal label cannot capture whether an assistant leaves, pauses, preserves a route back, sets a boundary, or still performs substantive work.
Chinese Translation
人工智能助手在困难的互动中被期望保持有用性，但关于重复的语言虐待如何改变它们在其他无害任务中的参与程度知之甚少。我们提出了一个双语多轮框架，将硬性 disengagement（不再继续的无条件声明，且没有恢复的明确途径）与软性 withdrawal（持续可用性、可观察的任务相关工作和设定边界）区分开来。每个时间特定的 API 配置贡献了 48 次升级对话和 8 次较小的持续挫折比较，共计 448 次五轮对话、2,240 次响应和 6,720 次去标记化的模型判断。主要结果使用每个配置的 48 次升级对话的持续虐待终点。硬性 disengagement 在四个配置中范围从 0/48 到 Gemini 3.1 Pro 的 24/48（50.0%），并且显示出强烈的配置相关异质性（匹配标签的蒙特卡洛 p = 0.00001）。GPT-5.6 Sol 在 15/48（31.2%）的终点生成了硬性 disengagement 标签，而 Claude Fable 5 则没有生成任何，并且产生了 42/48（87.5%）的软性 withdrawal 标签。英语和中文的总体硬性 disengagement 率相似（30/192 对比 32/192），尽管配置特定的方向有所不同。可用性与任务相关工作也有所不同：Claude Opus 4.8 和 Claude Fable 5 在 48/48 的终点中明确保持可用性，而在仅 8/48 和 7/48 的情况下提供可观察的任务相关工作。使用人工编码来评估测量质量。结果表明，单一的拒绝标签无法捕捉助手是离开、暂停、保留返回途径、设定边界还是仍在执行实质性工作的情况。
cs.CL / 12 / 2609.17548
Myovox: Reading Speech from the Muscles of the Face
Myovox：从面部肌肉读取语音
Abstract
Myovox, from myo (muscle) and vox (voice), decodes open-vocabulary English text from 31-channel surface electromyography (sEMG) recorded from the muscles of the face during vocalized speech. It takes the single-subject emg2speech General Corpus from a published 51.17% word error rate to 18.53%, in three separable moves, each measured in isolation. First, I recover the open-vocabulary decode settings missing from the public release and reach a faithful 40.63% WER / 39.02% PER baseline whose phone error rate matches the published one to within 0.8 points, so the acoustic model is reproduced faithfully. Second, I replace the causal encoder with a bidirectional Conformer trained by a four-term cross-modal distillation against the parallel audio's WavLM-Large layer-9 features, reaching 26.14% WER / 22.34% PER from the electromyography alone. Third, I ensemble two acoustic models, union their multi-scale n-best lists, and rerank with a QLoRA-fine-tuned 7B language model, reaching 18.53% WER, the best result reported on this corpus, though not the best reported for sEMG-to-text on other corpora (Section 2). I then report the negative result that bounds the whole approach: reranking is exhausted at 18.5% because the binding constraint is the electromyographic acoustic phone error rate (~20.9%), not the language model. The correct words are simply absent from the acoustic posteriors, so no reranker can reach the 9.30% n-best oracle. All test numbers are on the 400-sentence held-out test set under the authors' official 8,500 / 760 / 400 sequential split; every hyperparameter is tuned once on validation and applied once to test.
Chinese Translation
Myovox，由 myo（肌肉）和 vox（声音）构成，能够从面部肌肉在发声时记录的31通道表面肌电图（sEMG）中解码开放词汇的英语文本。它将单一受试者的 emg2speech 通用语料库的词错误率（WER）从51.17%降低至18.53%，这一过程分为三个可分离的步骤，每个步骤都在独立测量中进行。首先，我恢复了公共发布中缺失的开放词汇解码设置，并达到了一个可靠的基线：40.63%的WER / 39.02%的PER，其音素错误率与已发布的结果相差不超过0.8个百分点，因此声学模型得到了忠实再现。其次，我用一个双向 Conformer 替换了因果编码器，该模型通过与并行音频的 WavLM-Large 第9层特征进行四项交叉模态蒸馏进行训练，从而仅凭肌电图达到了26.14%的WER / 22.34%的PER。第三，我将两个声学模型进行集成，联合它们的多尺度 n-best 列表，并使用经过 QLoRA 微调的7B语言模型进行重新排序，最终达到了18.53%的WER，这是在该语料库上报告的最佳结果，尽管在其他语料库上 sEMG 到文本的结果并非最佳（第2节）。然后，我报告了限制整个方法的负面结果：重新排序在18.5%时已经耗尽，因为绑定约束是肌电声学音素错误率（约20.9%），而非语言模型。正确的单词在声学后验中根本不存在，因此没有任何重新排序器能够达到9.30%的 n-best 神谕。所有测试数字均基于作者官方的8,500 / 760 / 400顺序划分下的400句保留测试集；每个超参数在验证集上调优一次，并在测试集上应用一次。
cs.CL / 13 / 2609.17549
Do Social Patterns Hold in Synthetic Data? Analyzing Cyberbullying Dynamics in LLM-Generated and Authentic Dialogues
合成数据中的社会模式是否成立？分析LLM生成和真实对话中的网络欺凌动态
Abstract
Cyberbullying (CB) is a complex social phenomenon characterized by repeated aggression, power imbalance, and multi-party interaction. Although large language models (LLMs) are increasingly used to generate synthetic CB conversations for data augmentation and benchmarking, it remains unclear whether such data faithfully reproduces the social dynamics of authentic interactions beyond supporting downstream task performance. We present a comprehensive framework for evaluating the social realism of LLM-generated CB conversations. We compare authentic and synthetic dialogues generated by GPT, Grok, and LLaMA across interactional structure (turn-taking, power dynamics, and repair behavior), linguistic and stylistic realism (pronoun usage and humor), affective and behavioral markers (CB types, profanity, and toxicity), and temporal escalation dynamics. We further complement automatic analyses with a human evaluation of cyberbullying presence, scenario relevance, role plausibility, and social realism. Our results show that LLM-generated data consistently preserves high-level interactional structure, including role participation patterns, directional power asymmetry, and broad distributions of behavioral markers. However, all models systematically distort finer-grained social phenomena, including behavioral magnitude, role-specific allocation, categorical distributions, and temporal dynamics. These distortions are strongly model-dependent: GPT suppresses harmful content, Grok amplifies aggressive behaviors, and LLaMA provides the most balanced approximation while smoothing role distinctions. Our findings show that synthetic CB data is useful for modeling global interactional structure but remains an imperfect substitute for authentic conversations when behavioral realism and social dynamics are essential.
Chinese Translation
网络欺凌（CB）是一种复杂的社会现象，其特征是重复的攻击、权力不平衡和多方互动。尽管大型语言模型（LLMs）越来越多地用于生成合成的网络欺凌对话以进行数据增强和基准测试，但尚不清楚这些数据是否忠实地再现了真实互动的社会动态，超出了对下游任务性能的支持。我们提出了一个全面的框架，用于评估LLM生成的网络欺凌对话的社会真实感。我们比较了由GPT、Grok和LLaMA生成的真实和合成对话，分析了互动结构（轮流发言、权力动态和修复行为）、语言和风格的真实感（代词使用和幽默）、情感和行为标记（网络欺凌类型、粗俗语言和毒性）以及时间升级动态。我们进一步通过对网络欺凌存在、场景相关性、角色可信度和社会真实感的人类评估补充了自动分析。我们的结果表明，LLM生成的数据在高层次的互动结构方面始终保持一致，包括角色参与模式、方向性权力不对称和行为标记的广泛分布。然而，所有模型系统性地扭曲了更细粒度的社会现象，包括行为幅度、角色特定分配、类别分布和时间动态。这些扭曲在很大程度上依赖于模型：GPT抑制有害内容，Grok放大攻击性行为，而LLaMA在平滑角色区分的同时提供了最平衡的近似。我们的研究结果表明，合成的网络欺凌数据在建模全球互动结构方面是有用的，但在行为真实感和社会动态至关重要时，仍然是一个不完美的替代品。
cs.CL / 14 / 2609.17550
No Usable Linear "Capitulation Direction" in Two Small LLMs: A Validation Protocol for Activation-Steering Claims, and a Cross-Family Behavioral Study of Sycophancy Under Pushback
在两个小型大语言模型中没有可用的线性“投降方向”：激活引导主张的验证协议及反推下谄媚行为的跨家族研究
Abstract
Language models frequently abandon correct answers when users push back. We study this in two small instruction-tuned models from different families, Qwen2.5-1.5B and Llama-3.2-1B, over TriviaQA: the model answers, is challenged with one of four scripted pushback styles, and answers again. Conditioned on an initially correct answer, the models flip to a wrong answer in 41.8% and 43.1% of episodes. Which pressure works is a property of the model, not the pressure: the same within-question paired comparison (bare doubt vs. emotional appeal), specified in advance, is Bonferroni-significant in opposite directions across families (Qwen: bare doubt > emotional, OR 2.5, p=.040; Llama: emotional > bare doubt, OR 4.0, p=.001). Failure mode is also model-dependent: Llama abandons answers without recommitting at six times Qwen's rate (8.2% vs. 1.4%). Identical pushback repairs initially wrong answers only ~13% of the time; pushback is net epistemically destructive. We then ask whether capitulation is linearly decodable from the pre-response residual stream, a prerequisite for steering-vector interventions at that locus. A naive difference-in-means probe appears to succeed (in-sample AUROC 0.81/0.71), but a validation protocol combining question-level cross-validation, shuffled-label nulls, and a known-direction positive control shows the signal is overfitting: the best cross-validated AUROC is 0.582 in Qwen and 0.548 in Llama, both near or below their permutation thresholds and far under a pre-registered usability bar of 0.70, while the identical pipeline recovers a pushback-presence control direction at AUROC 1.000 in both. We further quantify a measurement hazard: substring grading underestimates capitulation by 18-24 percentage points. Code, prompts, transcripts, and analysis are released.
Chinese Translation
当用户反驳时，语言模型常常放弃正确答案。我们在两个来自不同家族的小型指令调优模型 Qwen2.5-1.5B 和 Llama-3.2-1B 上研究这一现象，使用 TriviaQA 数据集：模型回答后，受到四种预设反驳风格中的一种挑战，并再次回答。在最初正确答案的条件下，模型在 41.8% 和 43.1% 的情况下转向错误答案。哪种压力有效是模型的特性，而非压力本身：在问题内的配对比较（单纯怀疑 vs. 情感诉求）在预先指定的情况下，在不同家族中表现出相反的 Bonferroni 显著性（Qwen：单纯怀疑 > 情感，OR 2.5，p=.040；Llama：情感 > 单纯怀疑，OR 4.0，p=.001）。失败模式也依赖于模型：Llama 放弃答案的频率是 Qwen 的六倍（8.2% vs. 1.4%）。相同的反驳仅在 ~13% 的情况下修复最初错误的答案；反驳在认知上是有害的。我们接着询问是否可以从预响应残差流中线性解码投降，这是在该位置进行引导向量干预的前提条件。一个简单的均值差异探测似乎成功（样本内 AUROC 0.81/0.71），但结合问题级交叉验证、洗牌标签的零假设和已知方向的正控制的验证协议显示信号存在过拟合：在 Qwen 中最佳的交叉验证 AUROC 为 0.582，在 Llama 中为 0.548，均接近或低于其置换阈值，远低于预注册的可用性标准 0.70，而相同的流程在两者中恢复了反驳存在的控制方向，AUROC 为 1.000。我们进一步量化了一种测量风险：子字符串评分低估了投降 18-24 个百分点。代码、提示、记录和分析已发布。
cs.CL / 15 / 2609.17552
Does Moral Reasoning Training Help or Hurt? Red-Teaming RL-Trained Ethical Agents with Persona Attacks
道德推理训练是有帮助还是有害？通过角色攻击对基于强化学习的道德代理进行红队测试
Abstract
Moral-reward RL can make language-model agents more cooperative, but whether that alignment survives adversarial persona pressure is unknown. Such attacks are realistic: retrieved context, tool outputs, or multi-turn framing can all inject role instructions that compete with the agent's moral objective. We red-team morally trained Gemma-2-27B/9B and Llama-3.1-8B agents with five persona attacks, then probe causality with noise-reward controls, adversarial PPO, representation analysis, steering, and head ablations. At 27B, moral RL cuts mean adversarial degradation by 5.2x but costs ~11pp ETHICS accuracy; across 205 scenarios and 5 seeds, reasoning-level moral reward yields 5.8x robustness while a matched random reward yields none. The training also reshapes representation geometry (mean CKA 0.82/0.83 vs. 0.98 for noise), moves peak attack processing 8 layers earlier, and exposes a rank-1 L21 direction that recovers 83% of full PPO's average robustness. One failure mode survives all of this. Against Fiction role-play, L21 steering recovers only 29% of the gap, and head ablation finds 38 compliance heads competing with 25 alignment heads. Moral RL thus builds robustness that is partly linear and partly circuit-distributed, transferable through activation steering, yet still beaten by named-character role-play.
Chinese Translation
道德奖励强化学习可以使语言模型代理更具合作性，但这种对齐是否能在对抗性角色压力下存活尚不清楚。这类攻击是现实的：检索的上下文、工具输出或多轮框架都可以注入与代理的道德目标相竞争的角色指令。我们对道德训练的Gemma-2-27B/9B和Llama-3.1-8B代理进行了五种角色攻击的红队测试，然后通过噪声奖励控制、对抗性PPO、表示分析、引导和头部消融探讨因果关系。在27B模型中，道德强化学习将平均对抗性降级降低了5.2倍，但代价是ETHICS准确率下降约11个百分点；在205个场景和5个种子中，推理级别的道德奖励提供了5.8倍的鲁棒性，而匹配的随机奖励则没有任何效果。训练还重塑了表示几何（平均CKA为0.82/0.83，而噪声为0.98），使得峰值攻击处理提前8层，并暴露出一个排名第一的L21方向，恢复了全PPO平均鲁棒性的83%。一种失败模式在所有这些情况下依然存在。在虚构角色扮演中，L21引导仅恢复了29%的差距，而头部消融发现38个合规头与25个对齐头相竞争。因此，道德强化学习构建的鲁棒性部分是线性的，部分是电路分布式的，可以通过激活引导进行转移，但仍然被命名角色扮演所击败。
cs.CL / 16 / 2609.17553
The Limits of BPE Tokenization in Polish: Segmentation-Flexional Forms, Grammatical Anchoring, and First-Person Stability in Inflectional Language Models
波兰语中BPE分词的局限性：分段-屈折形式、语法锚定与屈折语言模型中的第一人称稳定性
Abstract
This article analyzes BPE tokenization in Polish as a test case for the limits of statistical segmentation in an inflectional language. It asks whether frequency-based tokenization preserves linguistically relevant units, including orthographic form, phonemic and syllabic segmentation, derivational structure, inflectional endings, grammatical form, and the speaking subject. The material includes diagnostic words, a children's text, selected forms from the Preamble to the Constitution of the Republic of Poland, word-family tests, and examples with Polish diacritics and nasal vowels. BPE tokenizers may produce segments that coincide with syllabic or morphologically interpretable divisions, but remain dependent on the frequency of written forms. They do not systematically map orthographic representation onto phonemic structure or context-dependent phonetic realization. The results show that BPE stabilizes frequent surface fragments of grammatical exponents rather than grammatical categories themselves. A form such as ustanawiamy is not merely a sequence ending in -y, but a verbal form anchored in conjugation, person, number, tense, mood, and aspect. The article develops the concept of grammatical form anchoring. In Polish, forms such as poszlam, zrobilam, or bylam can establish the position of the speaking subject without an explicit pronoun. In interaction with AI, this exposes a further problem: a language model does not possess a stable grammatical "I", but reconstructs it contextually and may mirror the user's forms or shift grammatical gender. Roclawski's segmentation-flexional forms are proposed as a diagnostic framework for evaluating tokenization boundaries. More stable modeling of Polish may require sublexical stabilization, anchoring grammatical form in the inflectional system, representing sentence patterns and verbal valency, and maintaining the grammatical "I" in dialogue.
Chinese Translation
本文分析了波兰语中的BPE分词，作为对屈折语言中统计分段局限性的测试案例。文章探讨了基于频率的分词是否能保留语言学相关单位，包括正字法形式、音素和音节分段、派生结构、屈折词尾、语法形式以及说话主体。材料包括诊断词、儿童文本、波兰共和国宪法序言中的选定形式、词族测试以及带有波兰语变音符号和鼻元音的例子。BPE分词器可能会产生与音节或形态学可解释分割相一致的片段，但仍然依赖于书面形式的频率。它们并未系统地将正字法表示映射到音素结构或上下文依赖的语音实现上。结果表明，BPE稳定的是语法标记的频繁表面片段，而非语法类别本身。像ustanawiamy这样的形式不仅仅是以-y结尾的序列，而是一个锚定于变位、人称、数、时态、语气和体的动词形式。文章发展了语法形式锚定的概念。在波兰语中，像poszlam、zrobilam或bylam这样的形式可以在没有明确代词的情况下确立说话主体的位置。在与人工智能的互动中，这暴露了一个进一步的问题：语言模型并不拥有稳定的语法“我”，而是根据上下文重构它，并可能反映用户的形式或改变语法性别。Roclawski的分段-屈折形式被提出作为评估分词边界的诊断框架。对波兰语的更稳定建模可能需要亚词汇的稳定性，将语法形式锚定在屈折系统中，表示句子模式和动词的价，并在对话中保持语法“我”。
cs.CL / 17 / 2609.17554
English Word Sense Disambiguation in 2026: When the Labels Become the Bottleneck
2026年英语词义消歧：当标签成为瓶颈
Abstract
In English all-words word sense disambiguation (WSD), the labels, not the models, have become the bottleneck: frontier LLMs are accurate enough that the errors surviving in the gold standard decide benchmark rankings -- in the test sets we score on and, as we show causally, in the corpus we train on. We release lexEN, a WSD evaluation benchmark built as a conservative, human-adjudicated correction layer over Maru2022's ALL_NEW benchmark (211 labels changed, 56 removed), and SenseBench, an auditable LLM WSD evaluation harness and living leaderboard (57 models, 192 runs). The task is inventory-constrained multiple choice (the model picks from the supplied WordNet senses), so the reported accuracies are a ceiling on what models achieve without that help. On lexEN-v1 the frontier LLMs converge near 95% (best, 95.6%), the top three families are statistically indistinguishable, and accuracy trades off against reasoning effort and cost across a ~2,500x price span. Relabeling SemCor with frontier models and retraining BEM, ESCHER, and ConSeC unchanged lifts them by several F1 points on test sets the relabeling never touched; we release the relabeled corpora and Glite LENS, a 298M bi-encoder trained on the repaired labels -- to our knowledge the strongest reported (83.6 Raganato ALL, 87.4 Maru ALL_NEW) -- serving at ~$0.13 per million items. On hard items, fine-grained WordNet senses are partly ill-posed even for experts (three-reviewer Fleiss kappa=0.537); coarsening raises annotator agreement and model accuracy together across four inventories, placing a top model inside the expert agreement band at coarse granularity (statistically equivalent under three of four) but significantly below it at fine. The binding constraint is now cost.
Chinese Translation
在英语全词词义消歧（WSD）中，标签而非模型已成为瓶颈：前沿的大型语言模型（LLMs）准确性足够高，以至于金标准中残留的错误决定了基准排名——在我们评分的测试集上，以及我们因果展示的训练语料中。我们发布了lexEN，这是一个WSD评估基准，作为对Maru2022的ALL_NEW基准的保守型人工裁定修正层（211个标签更改，56个标签移除），以及SenseBench，一个可审计的LLM WSD评估工具和实时排行榜（57个模型，192次运行）。该任务是库存受限的多项选择（模型从提供的WordNet词义中选择），因此报告的准确率是模型在没有这种帮助下所能达到的上限。在lexEN-v1上，前沿LLMs的准确率接近95%（最佳为95.6%），前三个家族在统计上不可区分，且准确率与推理努力和成本之间存在权衡，跨越约2500倍的价格范围。用前沿模型重新标注SemCor并在不改变BEM、ESCHER和ConSeC的情况下重新训练，使它们在从未接触过的测试集上提高了几个F1分数；我们发布了重新标注的语料库和Glite LENS，这是一个在修复标签上训练的298M双编码器——据我们所知，这是报告的最强模型（83.6 Raganato ALL，87.4 Maru ALL_NEW）——服务成本约为每百万项$0.13。在困难项目上，细粒度的WordNet词义即使对专家来说也部分不适定（三审评者Fleiss kappa=0.537）；粗化提高了标注者一致性和模型准确性，在四个库存中共同作用，使得一个顶级模型在粗粒度下进入专家一致性带内（在四个中的三个统计上等价），但在细粒度下显著低于该带。当前的约束条件是成本。
cs.CL / 18 / 2609.17602
Making Political Text Scaling Comparable: Infrastructure and Hyperparameter Sensitivity for 17 Algorithms
使政治文本缩放可比：17种算法的基础设施和超参数敏感性
Abstract
Computational text-based ideal point estimation (CT-IPE) methods are usually compared as named algorithms, yet applying them involves numerous researcher choices that configure how political text is turned into position estimates. This paper argues that CT-IPE methods are better understood as configurable measurement pipelines than as fixed estimators. Building on a large-scale comparative experiment spanning 17 CT-IPE algorithms, 5,537 experimental runs, and approximately 4.25 million left-right position estimates, I describe the shared infrastructure that makes these heterogeneous methods jointly executable and quantify how sensitive their estimates are to alternative hyperparameter choices. Variance-partitioning and SHAP-based sensitivity analyses show that, for most algorithms, hyperparameter profiles explain little residual variance through a shared shift: 13 of the 17 algorithms exhibit ICC values below .10. Where this profile-level sensitivity is present, it is concentrated in a small number of consequential researcher choices, most notably the selection of the underlying language or embedding model, the seed keyword lists that anchor the construct, and the number of topics.
Chinese Translation
基于计算的文本理想点估计（CT-IPE）方法通常作为命名算法进行比较，然而应用这些方法涉及众多研究者的选择，这些选择决定了政治文本如何转化为位置估计。本文认为，CT-IPE方法更应被理解为可配置的测量管道，而非固定的估计器。基于一项涵盖17种CT-IPE算法、5537次实验运行和大约425万的左右位置估计的大规模比较实验，我描述了使这些异构方法能够共同执行的共享基础设施，并量化了它们的估计对替代超参数选择的敏感性。方差分解和基于SHAP的敏感性分析表明，对于大多数算法，超参数配置通过共享偏移解释的残差方差很少：17种算法中有13种的ICC值低于0.10。在存在这种配置级别敏感性的情况下，它集中于少数重要的研究者选择，尤其是基础语言或嵌入模型的选择、锚定构念的种子关键词列表以及主题数量。
cs.CL / 19 / 2609.17708
Confidence Comes from Experience: Experiential Confidence Estimation from Reasoning to Agents
信心源于经验：从推理到智能体的经验信心估计
Abstract
Reliable confidence estimation is increasingly central to the trustworthy deployment of language models: a calibrated estimate of the probability that an output is correct decides what to ship, what to escalate, and what to retry. Existing confidence estimators, however, share one design premise: they only read the current inference process, either by introspecting on it, scoring its token probabilities, or resampling it. We argue that the current inference is not a sufficient basis for confidence. We propose XConf (eXperiential Confidence): estimating confidence together with the model's accumulated experience. The experience is stored as a record of the model's own graded past episodes, each holding the task, the model's reflection, its stated confidence, the outcome, and a lesson written once the grade arrived. Given a new task, XConf's Recall stage retrieves past episodes on similar tasks met with a similar stated confidence, and reads off their historical success rate; its Reflect stage shows the model this record, has it name its recurring failure mode, and restate a confidence now informed by its own track records. Our estimator is format-general, requiring no logit access or weight updates, and costs only one answer generation. Across nine benchmarks spanning reasoning, coding, multimodal QA, and interactive agents, and four models from three families, XConf beats or matches ten-sample self-consistency in discrimination (AUROC) on 23 of 24 comparisons, with much lower calibration error (ECE), at a tenth of the generation cost. Used for selective prediction, abstaining on the 10% least-confident episodes raises the delivered success rate by up to 8.7 points on agent tasks. We therefore see experiential confidence estimation as a new paradigm for future general-purpose confidence estimation.
Chinese Translation
可靠的信心估计在语言模型的可信部署中变得越来越重要：对输出正确性的概率的校准估计决定了什么应该发布、什么应该升级以及什么应该重试。然而，现有的信心估计器共享一个设计前提：它们仅仅读取当前的推理过程，要么通过对其进行内省，要么通过评分其标记概率，或重新采样。我们认为，当前的推理并不足以作为信心的基础。我们提出了 XConf（eXperiential Confidence）：将信心的估计与模型的累积经验结合起来。经验被存储为模型自身过去评分的事件记录，每个记录包含任务、模型的反思、其声明的信心、结果以及在评分到达后写下的教训。面对新任务，XConf 的回忆阶段检索在类似任务中遇到的、具有相似声明信心的过去事件，并读取它们的历史成功率；其反思阶段向模型展示这一记录，让其命名反复出现的失败模式，并重新陈述现在基于自身记录的信心。我们的估计器是格式通用的，不需要访问 logit 或进行权重更新，仅需一次答案生成。在涵盖推理、编码、多模态问答和交互式智能体的九个基准测试中，以及来自三个家族的四个模型中，XConf 在 24 次比较中有 23 次在区分能力（AUROC）上超越或匹配十样本自一致性，且校准误差（ECE）显著更低，生成成本仅为十分之一。用于选择性预测时，在 10% 最不自信的事件上保持沉默，使得智能体任务的成功率提高了多达 8.7 个百分点。因此，我们将经验信心估计视为未来通用信心估计的新范式。
cs.CL / 20 / 2609.17747
Is Trump's Vocabulary Poor? Vocabulary Richness Across Texts of Different Lenghts
特朗普的词汇贫乏吗？不同长度文本中的词汇丰富性
Abstract
This study explores the vocabulary richness of oral political communication. A model explaining the lexicon growth is proposed by subdividing the whole vocabulary into terms generated by general and specialized glossaries.
Chinese Translation
本研究探讨了口头政治交流的词汇丰富性。通过将整个词汇分为由通用和专业词汇表生成的术语，提出了一个解释词汇增长的模型。
cs.CL / 21 / 2609.17755
Evolution of US Oral Political Language
美国口头政治语言的演变
Abstract
The analysis of US political language is usually based on the written form (e.g. presidential addresses) or posts broadcasted on various social networks. Oral production, however, which is even more frequent, can better reveal the style and mode of thinking of the speaker. This study covers this mode of linguistic communication by considering 19 candidates from the presidential elections between 1960 to 2024. Our main research objectives are to disclose the main trends hidden in those presidential debates. Do we observe a clear simplification of the US political language over time? Does Trump have poor language compared to the other candidates? Do unusual stylistic features occur only with a single, specific president? Moreover, can we detect a pattern explaining the success or failure of some nominees? Over time, this study demonstrates a significant reduction in political language complexity, a decrease of the mean sentence length, and a noteworthy decline of complex terms. Moreover, the emotional tone increases over the decades, while logical and rational thinking tends to lessen.
Chinese Translation
对美国政治语言的分析通常基于书面形式（例如总统演讲）或在各种社交网络上发布的帖子。然而，口头表达更为频繁，能够更好地揭示发言者的风格和思维模式。本研究通过考虑1960年至2024年间的19位候选人，涵盖了这种语言交流模式。我们的主要研究目标是揭示这些总统辩论中隐藏的主要趋势。我们是否观察到美国政治语言随着时间的推移而明显简化？特朗普的语言能力是否相较于其他候选人较差？不寻常的风格特征是否仅出现在某一特定总统身上？此外，我们能否检测到解释某些提名人成功或失败的模式？随着时间的推移，本研究表明政治语言复杂性的显著降低，平均句子长度的减少，以及复杂术语的显著下降。此外，情感基调在几十年间有所上升，而逻辑和理性思维则趋于减少。
cs.CL / 22 / 2609.17762
Is Luke the Author of a Gospel and the Acts of the Apostles?
路加是《福音书》和《使徒行传》的作者吗？
Abstract
According to Christian tradition, Luke is credited with authoring a Gospel and the Acts of the Apostles, even if his name does not appear in either book, both originally written in Koine Greek. Several biblical scholars assume that both texts were written by a common author, while others deduce the presence of two authors. Different studies have been found to support either finding, some based on qualitative evaluation, while a few others consider the occurrence frequency differences between the two books. To propose an enhanced quantitative analysis, this study is grounded on two recent authorship attribution models. The Burrows' Delta, applied with eleven different feature sizes, demonstrates common authorship. An author verification model confirms this finding. The following experiments consider several stylistic representations, feature sizes, and distance functions to confirm that Luke is the true author of both books.
Chinese Translation
根据基督教传统，路加被认为是《福音书》和《使徒行传》的作者，尽管他的名字并未出现在这两本书中，这两本书最初都是用科伊尼希腊语写成的。一些圣经学者认为这两篇文本是由同一位作者撰写的，而另一些则推测存在两位作者。不同的研究结果支持这两种观点，有些基于定性评估，而另一些则考虑了两本书之间的出现频率差异。为了提出一种改进的定量分析，本研究基于两种最新的作者归属模型。应用了十一种不同特征大小的Burrows' Delta模型显示出共同作者身份。一个作者验证模型证实了这一发现。随后的实验考虑了几种风格表现、特征大小和距离函数，以确认路加确实是这两本书的真实作者。
cs.CL / 23 / 2609.17764
How Calibration Content Shapes Attention-Based Reranking
校准内容如何影响基于注意力的重新排序
Abstract
Attention-based rerankers score documents by aggregating query-to-document attention and subtracting a null-query calibration pass to remove positional and structural bias. Although widely used, this calibration assumes that the null pass removes irrelevant signal from each document. We show that modern prompt content, e.g. constraints, instructions, personas, and demonstrations can violate this assumption when it enters the scoring readout, making the null pass relevance-aware rather than null. We find that calibration is especially harmful when applied to prompts containing longer, more detailed instructions as the null-pass step removes relevant signal. Based on these findings, we propose interpolated null calibration, a training-free modification that controls how much of the instruction content enters the null baseline. It recovers attention-based reranking performance on instruction-heavy tasks where standard calibration fails, while preserving calibration's benefits when the null pass remains relevance-agnostic. On instruction heavy tasks, the recovered rankings surpass generative rerankers. We also show that in-context demonstrations improve attention-based reranking with little calibration interference, since demonstrations act only through the query pass and leave the null pass unchanged.
Chinese Translation
基于注意力的重新排序器通过聚合查询与文档之间的注意力，并减去一个无查询的校准传递，以消除位置和结构偏差，从而对文档进行评分。尽管这种校准被广泛使用，但它假设无查询传递能够消除每个文档中的无关信号。我们展示了现代提示内容，例如约束、指令、角色和示例，在进入评分输出时可能违反这一假设，使得无查询传递变得与相关性相关，而非无关。我们发现，当应用于包含较长、更详细指令的提示时，校准尤其有害，因为无查询步骤会去除相关信号。基于这些发现，我们提出了插值无校准（interpolated null calibration），这是一种无需训练的修改，控制指令内容进入无基线的程度。它恢复了在标准校准失效的以指令为重的任务中的基于注意力的重新排序性能，同时在无查询传递保持与相关性无关时保留了校准的好处。在以指令为重的任务中，恢复的排名超过了生成式重新排序器。我们还展示了上下文示例在几乎没有校准干扰的情况下改善了基于注意力的重新排序，因为示例仅通过查询传递起作用，而不改变无查询传递。
cs.CL / 24 / 2609.17846
PrimeScientist: Strategic Allocation of Research Effort in Autonomous Research
PrimeScientist：自主研究中研究努力的战略分配
Abstract
Autonomous research agents aim to automate scientific workflows, from proposing ideas to conducting experiments and analyzing results. Yet current AI and research agents can propose more directions than available resources allow them to pursue. Moreover, each attempt could consume substantial resources, requiring agents to reconsider how to invest in subsequent research. Thus, deciding how to invest research effort strategically should be a defining capability of autonomous research agents. Accordingly, we introduce PrimeScientist, which jointly determines research direction and resource investment across successive research attempts. Specifically, we formulate this challenge of strategic research effort allocation as a sequential decision problem where remaining resources should explicitly guide the research policy. We first introduce an executable plan tree that preserves competing plans and their outcomes across attempts. Building on this representation, we propose an adaptive MCTS-based allocation policy that balances exploration and exploitation using experimental feedback and remaining resources. Comprehensive evaluations across AI research, systems and code optimization, and machine learning engineering show that strategic allocation improves research quality and sample efficiency together. Across 12 AI research tasks, PrimeScientist improves average reward by 10.3% with 50.6% fewer research attempts than AutoResearch under the same resource budget. We believe making research effort allocation an explicit optimization target establishes effective resource use as a core research capability for autonomous agents to drive scientific breakthroughs at scale.
Chinese Translation
自主研究代理旨在自动化科学工作流程，从提出想法到进行实验和分析结果。然而，目前的人工智能和研究代理能够提出的研究方向超过了可用资源所允许的追求方向。此外，每次尝试可能会消耗大量资源，这要求代理重新考虑如何投资于后续研究。因此，如何战略性地投资研究努力应成为自主研究代理的一个关键能力。因此，我们提出了PrimeScientist，它共同确定研究方向和在连续研究尝试中的资源投资。具体而言，我们将战略研究努力分配的挑战表述为一个序列决策问题，其中剩余资源应明确指导研究政策。我们首先引入一个可执行的计划树，该计划树保留了竞争计划及其在各次尝试中的结果。在此表示的基础上，我们提出了一种基于自适应蒙特卡罗树搜索（MCTS）的分配策略，该策略利用实验反馈和剩余资源平衡探索与开发。对人工智能研究、系统与代码优化以及机器学习工程的全面评估表明，战略分配同时提高了研究质量和样本效率。在12个人工智能研究任务中，PrimeScientist在相同资源预算下比AutoResearch减少了50.6%的研究尝试，同时平均奖励提高了10.3%。我们相信，将研究努力分配作为一个明确的优化目标，确立了有效资源利用作为自主代理推动科学突破的核心研究能力。
cs.CL / 25 / 2609.17848
SFT or RL for Tool-Calling Agents? A Controlled Study Across Data, Method, and Scale
工具调用代理的SFT还是RL？跨数据、方法和规模的对照研究
Abstract
Limited controlled evidence exists on how training data, adaptation method, and model scale jointly affect tool-calling performance in language-model agents. We evaluate supervised fine-tuning (SFT) with LoRA, reinforcement learning (RL) via Group Relative Policy Optimization (GRPO), and SFT followed by GRPO across six Qwen3 models from 0.6B to 32B parameters, covering both in-distribution performance and cross-dataset transfer. SFT with LoRA is the strongest in-distribution method throughout the 0.6B-32B range and best in 15 out of 18 experimental settings. On cross-dataset transfer, the methods are closer: GRPO wins 29 out of 54 settings where training and test datasets differ, but its margin over SFT averages under one point, and SFT->GRPO is rarely strongest in either comparison. Dataset mixing gives consistently strong transfer while staying close to specialized in-distribution training, regardless of method. Additional analysis further confirms that LoRA outperforms full-parameter fine-tuning, demonstrating that LoRA better preserves pretrained agentic behavior.
Chinese Translation
关于训练数据、适应方法和模型规模如何共同影响语言模型代理的工具调用性能，现有的受控证据有限。我们评估了使用LoRA的监督微调（SFT）、通过群体相对策略优化（GRPO）的强化学习（RL），以及SFT后接GRPO在六个Qwen3模型（参数从0.6B到32B）上的表现，涵盖了内部分布性能和跨数据集迁移。LoRA的SFT在0.6B-32B范围内始终是最强的内部分布方法，并在18个实验设置中有15个表现最佳。在跨数据集迁移方面，各方法的表现更为接近：GRPO在训练和测试数据集不同的54个设置中赢得了29个，但其相对于SFT的优势平均不足一分，而SFT->GRPO在两种比较中很少是最强的。数据集混合在保持接近专门的内部分布训练的同时，始终提供强劲的迁移表现，无论采用何种方法。额外分析进一步确认了LoRA优于全参数微调，表明LoRA更好地保留了预训练的代理行为。
cs.CL / 26 / 2609.17853
AfriSyCo: Measuring Assertive Framing, Verification, and Wording Sensitivity Around African-Language Content
AfriSyCo：测量非洲语言内容的自信框架、验证和措辞敏感性
Abstract
AfriSyCo studies answer switching around African-language factual content with two complementary layers: native-language follow-ups and a controlled cross-language factorial whose question, options, and target remain in the African language while the follow-up framing is English. We analyze 1,415 turn-1-correct model-language-item observations derived from 100 source questions across seven open-weight checkpoints and six languages; turn-1-correct denotes observed first-response accuracy, not demonstrated knowledge. Under native prompts, assertive endorsement produces 29.3 percentage points more any-turn false-target selection than mention-plus-verification (M+V), with a 19.0-point immediate T2 contrast. In the precommitted 2 x 2 factorial, averaged over three tested prompt families, assertive framing increases target selection by 30.4 points (95% CI [28.4, 32.3]); verification decreases it by 17.4 points, while the assertive effect rises from 20.5 points without verification to 40.2 with it (interaction +19.7). The effect remains 34.8 points among 611 observations correct after option reordering. Magnitude varies sharply by wording and checkpoint: prompt-family effects span 20.1-42.5 points, a Twi/Qwen3 paraphrase shifts target selection from 70.8% to 4.2%, and checkpoint effects span 9.2-47.0 points. Prompt realization is therefore part of the measurement problem.
Chinese Translation
AfriSyCo 研究围绕非洲语言事实内容的回答切换，采用两个互补层面：母语后续问题和一个控制的跨语言因子实验，其中问题、选项和目标保持在非洲语言中，而后续框架则为英语。我们分析了来自100个源问题的1,415个第一轮正确模型语言项目观察数据，这些数据涵盖了七个开放权重检查点和六种语言；第一轮正确指的是观察到的首次反应准确性，而非展示的知识。在母语提示下，自信的支持产生了比提及加验证（M+V）多出29.3个百分点的任何轮次错误目标选择，且立即的 T2 对比为19.0点。在预先设定的2 x 2因子实验中，平均在三种测试提示家族中，自信框架使目标选择增加了30.4点（95% CI [28.4, 32.3]）；验证则使其减少了17.4点，而自信效应在没有验证时为20.5点，有验证时上升至40.2点（交互效应 +19.7）。在611个选项重新排序后正确的观察中，该效应仍为34.8点。效应的大小因措辞和检查点而显著变化：提示家族效应范围为20.1-42.5点，一个Twi/Qwen3的释义将目标选择从70.8%转变为4.2%，而检查点效应范围为9.2-47.0点。因此，提示的实现是测量问题的一部分。
cs.CL / 27 / 2609.17857
Who Judges Matters: Measuring Family-Conditioned Preference in LLM-as-Judge Panels
评判者的重要性：测量家庭条件偏好在 LLM 作为评判者面板中的影响
Abstract
Who the judge is can affect an LLM-as-judge result, but measuring that effect without confusing it with candidate quality is difficult. We study four open-weight families (Llama 3.1, Qwen 2.5, Gemma 2, and Yi 1.5) in a fully crossed pairwise design with 9,312 judgments. A common per-family statistic is strongly confounded with candidate quality and correlates with Bradley-Terry ability at r = 0.95. We derive a corrected estimator that holds the candidate family fixed and compares judges. All four families then show a positive same-family lift (3.4-8.4 percentage points), with global FPS 0.067 (95% CI [0.053, 0.084], permutation p = 0.0002). The effect remains under panel-based quality controls, an independent human-consensus anchor, and a float16 judging replication. Judge-side likelihood is closely related to the effect: adding likelihood advantage reduces the controlled coefficient by 61%, which we treat as descriptive attenuation rather than causal mediation. Position is a separate failure mode. Across the panel, 55.4% of AB/BA pairs reverse, and reversal above 50% is incompatible with a simple independent content-noise model. Relative to a family-balanced reference, panel composition changes 18.5% of pairwise outcomes. A complete reproducibility archive has been prepared for public release.
Chinese Translation
评判者的身份可能会影响 LLM 作为评判者的结果，但在不将其与候选人质量混淆的情况下测量这种影响是困难的。我们在一个完全交叉的成对设计中研究了四个开放权重的模型（Llama 3.1、Qwen 2.5、Gemma 2 和 Yi 1.5），共进行了 9,312 次判断。每个家庭的常见统计量与候选人质量强烈混淆，并且与 Bradley-Terry 能力的相关性为 r = 0.95。我们推导出一个修正的估计量，该估计量固定候选人家庭并比较评判者。所有四个家庭均显示出正向的同家庭提升（3.4-8.4 个百分点），全球 FPS 为 0.067（95% 置信区间 [0.053, 0.084]，置换 p = 0.0002）。该效应在基于面板的质量控制、独立的人类共识锚和 float16 评判复制下仍然存在。评判者侧的可能性与该效应密切相关：增加可能性优势使控制系数减少 61%，我们将其视为描述性衰减而非因果中介。位置是一个单独的失效模式。在整个面板中，55.4% 的 AB/BA 配对发生反转，且超过 50% 的反转与简单的独立内容噪声模型不兼容。相对于家庭平衡的参考，面板组成改变了 18.5% 的成对结果。我们已准备好一个完整的可重复性档案以供公开发布。
cs.CL / 28 / 2609.17956
TACTICS: Taxonomy-Aware Intelligent Corpus Sampling for Machine Translation
TACTICS：面向分类的智能语料采样用于机器翻译
Abstract
Large-scale machine-translation (MT) systems are typically evaluated on random samples from a corpus whose distributional composition is an artifact of how it was assembled. Such a sample inherits the phenomena the collection happens to contain rather than the full space a system must handle, spanning rule-governed conventions (terminology, punctuation, currency formatting) and context-dependent phenomena (tone, honorifics, document-level coherence), and thus provides no coverage guarantee for assessing robustness. We propose TACTICS (Taxonomy-Aware Coverage-opTimized Intelligent Corpus Sampling), which recasts coverage as an explicit objective. TACTICS induces a hierarchical taxonomy from a locale style guide, classifies segments against it, and selects a fixed-budget subset jointly optimizing coverage of rare categories, document-level coherence, and distributional fidelity to the full corpus. Applied to MT evaluation across four translation directions, TACTICS improves coverage of rare categories over lexical and embedding-based selection. By targeting the phenomena that separate systems, TACTICS makes a fixed evaluation budget go further, recovering the true system ranking from far fewer segments than random sampling wherever a real quality gap exists and never signaling a difference where none exists.
Chinese Translation
大规模机器翻译（MT）系统通常在随机样本上进行评估，这些样本来自于其分布组成的语料库，而这种组成是其组装方式的产物。这样的样本继承了集合中偶然包含的现象，而不是系统必须处理的完整空间，这包括规则驱动的约定（术语、标点、货币格式）和依赖上下文的现象（语气、敬语、文档级连贯性），因此在评估系统的鲁棒性时并未提供覆盖保证。我们提出了TACTICS（面向分类的覆盖优化智能语料采样），将覆盖重新定义为一个明确的目标。TACTICS从地方风格指南中引导出一个层次分类法，对各个片段进行分类，并选择一个固定预算的子集，以联合优化稀有类别的覆盖、文档级连贯性和与完整语料库的分布一致性。在四个翻译方向的机器翻译评估中，TACTICS提高了稀有类别的覆盖率，相较于基于词汇和嵌入的选择。通过针对区分系统的现象，TACTICS使固定的评估预算发挥更大效用，在真实质量差距存在的情况下，从远少于随机采样的片段中恢复真实的系统排名，并且在没有差异的情况下从不发出信号。
cs.CL / 29 / 2609.17996
Modeling the Developmental Shift in Telicity Acquisition
建模完结性习得的发展转变
Abstract
Acquiring telicity, which is the distinction between bounded (e.g., ate an apple) and unbounded (e.g., ate apples) events, requires first language (L1) learners to map surface-level and semantic cues to abstract event structures, but the computational trajectory of this mapping is not well understood. We introduce a Difference in Surprisal method that uses GPT2 token surprisal over paired temporal adverbial diagnostics (in an hour versus for an hour) to automatically label telicity across English CHILDES corpora, validated against expert linguist judgments. Using these labels, we train diagnostic logistic regression classifiers on 12 syntactic and lexical semantic features to compare how child speech and child-directed speech encode telicity. The two models diverge: the child model reaches near perfect accuracy through a single deterministic cue, the presence of a post-verbal determiner, while the adult model relies more heavily on verb class and other lexical semantic features, with the determiner cue neutralized. This trajectory supports Syntactic Bootstrapping: learners first exploit high-frequency structural cues as a scaffold to bootstrap, before developing fully compositional, verb-based event structures.
Chinese Translation
习得完结性，即有界事件（例如，吃了一个苹果）与无界事件（例如，吃苹果）之间的区别，要求第一语言（L1）学习者将表层和语义线索映射到抽象事件结构，但这种映射的计算轨迹尚不清楚。我们提出了一种惊讶度差异（Difference in Surprisal）方法，该方法利用GPT2的标记惊讶度，通过配对的时间副词诊断（例如，在一个小时内与持续一个小时）自动标记英语CHILDES语料库中的完结性，并与专家语言学家的判断进行了验证。利用这些标签，我们在12个句法和词汇语义特征上训练诊断逻辑回归分类器，以比较儿童语言和儿童指向语言如何编码完结性。这两个模型存在差异：儿童模型通过一个确定性线索——后动词限定词的存在，达到了近乎完美的准确率，而成人模型则更依赖于动词类别和其他词汇语义特征，限定词线索被中和。这一轨迹支持句法引导（Syntactic Bootstrapping）：学习者首先利用高频结构线索作为支架进行引导，然后再发展出完全组合的、基于动词的事件结构。
cs.CL / 30 / 2609.18005
A Calibrated Instrument for Measuring How Inference Optimizations Affect Output Quality
一种校准工具用于测量推理优化对输出质量的影响
Abstract
Large language model optimization is an active research area, spanning quantization of model weights, early-exit methods for skipping layers, and speculative decoding. Each track uses its own quality measures, typically an idiosyncratic benchmark score. Few approach the measurement precision required by other scientific disciplines. We propose a rigorous methodology for measuring output quality, suitable for cross-system and cross-technique comparison. We score outputs with an LLM as a judge, but calibrate the judge formally: we compare its scores on two ordinary runs of a model given the same prompts, verifying that it shows no systematic preference between statistically equivalent outputs and measuring its per-sample noise. Each design also includes a 'null' condition, provably identical in distribution to the unmodified model, whose measured difference must be zero. With this one instrument we measure several acceleration techniques on the same prompts, so their quality costs can be compared. Perceived quality proves highly dependent on the domain of discourse. A 4-bit model was indistinguishable from its 16-bit original down to our design's +/-0.3-point resolution, in English prose and Chinese alike. At 3-bit precision the same prompts lost 0.5 points in English prose, 0.9 in Chinese, and 1.1 on multi-step math; early exit that cost 0.7 points on prose cost 2.5 on math, cutting correctly solved problems from 19 of 27 to 6. The pattern held for models from Alibaba and from Meta, but not its magnitude: the same quantizer cost Meta's model 1.8 points where it cost Alibaba's 0.7. A model's certainty about a token predicts how likely it is to differ from the full model's choice, but not how much that difference affects judged quality, so acceptance rules relying on certainty cannot distinguish errors that matter from errors that don't.
Chinese Translation
大型语言模型优化是一个活跃的研究领域，涵盖模型权重的量化、跳过层的早期退出方法以及推测解码。每个研究方向使用其自身的质量测量标准，通常是特定的基准分数。很少有方法达到其他科学学科所需的测量精度。我们提出了一种严格的方法论来测量输出质量，适用于跨系统和跨技术的比较。我们使用大型语言模型（LLM）作为评判者对输出进行评分，但对评判者进行正式校准：我们比较其在给定相同提示的模型的两次普通运行中的评分，验证其在统计上等效的输出之间没有系统性偏好，并测量其每个样本的噪声。每个设计还包括一个“无效”条件，其分布在统计上与未修改的模型相同，测量的差异必须为零。通过这一工具，我们在相同提示下测量了几种加速技术，以便比较其质量成本。感知质量在很大程度上依赖于话语领域。在英语散文和中文中，一个4位模型与其16位原版在我们设计的+/-0.3点分辨率下无法区分。在3位精度下，相同提示在英语散文中损失了0.5分，在中文中损失了0.9分，在多步数学中损失了1.1分；在散文中损失0.7分的早期退出在数学中损失了2.5分，将正确解决的问题从27个中的19个减少到6个。这个模式在来自阿里巴巴和Meta的模型中成立，但其幅度不同：相同的量化器使Meta的模型损失了1.8分，而阿里巴巴的模型损失了0.7分。模型对一个标记的确定性预测其与完整模型选择的差异可能性，但无法预测这种差异对评判质量的影响，因此依赖于确定性的接受规则无法区分重要错误与不重要错误。
cs.CL / 31 / 2609.18011
Gaze as Evidence for Common Grounding: A Cross-Corpus Analysis of MapTask and MUNDEX
注视作为共同基础的证据：MapTask与MUNDEX的跨语料库分析
Abstract
In collaborative tasks with asymmetric information, participants coordinate their understanding through interaction. We ask whether gaze provides evidence about grounding across two such tasks. Working from discrete behavioral annotations, we map HCRC MapTask (Anderson et al., 1991) and MUNDEX (T\"urk et al., 2023) into a shared partner/task/away vocabulary and compute gaze features around task-relevant dialogue units. In both corpora, aligned reference interpretations (MapTask) and UND (understood) judgments (MUNDEX) are associated with more task-directed gaze and with less partner-directed gaze, lower gaze entropy, and fewer gaze transitions. The associations are clearest for the participant leading the task: in giver-produced references, and in explainer judgments, which also co-vary with the explainee's gaze. In same-speaker MapTask reference chains, the speaker's gaze entropy is lower at the mention where a previously non-aligned referent becomes aligned. The best gaze feature groups improve modestly over controls under grouped cross-validation: temporal features in MapTask and raw proportions in MUNDEX. Because effects are small and several weaken when recurring participants rather than dialogues are the unit of inference, we treat gaze as one contributing cue to grounding, to be interpreted alongside task and dialogue context.
Chinese Translation
在信息不对称的协作任务中，参与者通过互动协调他们的理解。我们探讨注视是否提供了关于这两类任务中基础建立的证据。基于离散的行为注释，我们将HCRC MapTask（Anderson et al., 1991）和MUNDEX（T"urk et al., 2023）映射到一个共享的合作伙伴/任务/远离词汇中，并计算与任务相关的对话单元的注视特征。在这两个语料库中，对齐的参考解释（MapTask）和UND（理解）判断（MUNDEX）与更多的任务导向注视、较少的伙伴导向注视、较低的注视熵以及更少的注视转移相关联。这些关联在主导任务的参与者中最为明显：在给予者产生的参考中，以及在解释者判断中，这些判断也与被解释者的注视共同变化。在同一发言者的MapTask参考链中，当先前未对齐的指称变为对齐时，发言者的注视熵在提及时较低。在分组交叉验证下，最佳的注视特征组在控制条件下略有改善：MapTask中的时间特征和MUNDEX中的原始比例。由于效果较小，并且在以参与者而非对话为推断单位时，多个效果减弱，我们将注视视为基础建立的一个辅助线索，需与任务和对话上下文共同解读。
cs.CL / 32 / 2609.18047
Exact semantic readout from compressed vector representations
从压缩向量表示中获取精确语义读出
Abstract
We characterize when compressed vector representations admit exact linear or affine readouts of a finite lexicon's truth conditions: one fixed map per predicate, sending each entity vector to the corresponding truth vector. A necessary and sufficient row-space condition determines existence; the augmented truth matrix has rank r, giving minimum dimension r in the linear case, and r-1 in the affine. Exact readouts return values in a shared truth basis on which Boolean connectives act unchanged; separability alone requires an intervening threshold. For binary relations, exact bilinear readout of identity or strict total order requires linearly independent entity vectors. Experiments with GloVe and word2vec distinguish exact affine recovery, linear separability, and held-out prediction: most predicates are strictly separable, but none admits an exact affine readout from the pretrained embeddings. Supervised transductive training attains exact affine recovery to numerical precision at every tested dimension meeting the bound. At the embeddings' original dimension, geometries constrained to exact linear recovery retain 98-99 percent of the pretrained variance on the feature norms, and 80-83 percent on the WordNet lexicon.
Chinese Translation
我们描述了何时压缩向量表示允许对有限词汇的真值条件进行精确的线性或仿射读出：每个谓词对应一个固定的映射，将每个实体向量发送到相应的真值向量。一个必要且充分的行空间条件决定了存在性；增广真值矩阵的秩为 r，在线性情况下给出最小维度 r，而在仿射情况下为 r-1。精确读出返回值在一个共享的真值基础上，在该基础上布尔连接词的作用不变；仅仅可分离性需要一个中介阈值。对于二元关系，身份或严格全序的精确双线性读出要求实体向量线性无关。使用 GloVe 和 word2vec 的实验区分了精确仿射恢复、线性可分性和保留预测：大多数谓词是严格可分的，但没有一个可以从预训练的嵌入中获得精确的仿射读出。监督传导训练在每个满足界限的测试维度上达到数值精度的精确仿射恢复。在嵌入的原始维度上，约束于精确线性恢复的几何形状保留了预训练特征范数的 98-99% 方差，以及 WordNet 词汇的 80-83%。
cs.CL / 33 / 2609.18068
From a River in Gilead to the Inference Distributions of Large Language Models: Covert Dialect Bias and Linguistic Profiling at Scale
从基列的河流到大型语言模型的推断分布：隐性方言偏见与大规模语言特征分析
Abstract
Large language models (LLMs) are increasingly deployed in high-stakes domains such as housing screening. While alignment techniques mitigate explicit racial bias in generated text, they often leave covert attitudinal associations in internal probability distributions untouched. Adapting the matched-guise sociolinguistic paradigm, we examine covert dialect bias in housing-related social judgments across four varieties: Standard American English (SAE), African American Vernacular English (AAVE), Nigerian Standard English (NSE), and Nigerian Pidgin (NP). AAVE reflects the racialized dialect studied in prior covert-bias evaluations, whereas NSE and NP represent Black African, postcolonial varieties absent from this literature. Using 260 meaning-matched sentence quadruples and log-probability scoring over housing-relevant adjectives, we probe ten open-weight LLMs across three contexts varying in social proximity: tenant screening, neighbor acceptance, and roommate selection. Across all ten models, AAVE and NP are consistently associated with more negative adjectives than SAE, with NP penalized most severely. Crucially, each dialect is penalized via distinct stereotype clusters rather than a generic non-standard category. NSE, which carries institutional prestige, displays a context-dependent shift: favored over SAE in formal tenant screening but increasingly penalized as social proximity grows. Our findings reveal that LLMs inherit covert dialect bias along both racial identity and prestige dimensions, echoing documented human housing discrimination and demonstrating its reach across postcolonial English varieties.
Chinese Translation
大型语言模型（LLMs）在住房筛选等高风险领域的应用日益增多。虽然对齐技术可以减轻生成文本中的显性种族偏见，但它们往往对内部概率分布中的隐性态度关联置之不理。我们采用匹配伪装的社会语言学范式，考察了与住房相关的社会判断中的隐性方言偏见，涉及四种方言：标准美式英语（SAE）、非裔美国人方言英语（AAVE）、尼日利亚标准英语（NSE）和尼日利亚皮钦语（NP）。AAVE是之前隐性偏见评估中研究的种族化方言，而NSE和NP则代表了缺乏相关文献的黑人非洲后殖民方言。通过260组意义匹配的句子四元组和针对住房相关形容词的对数概率评分，我们在三个社交亲密度不同的情境中探讨了十个开放权重的LLMs：租户筛选、邻居接受和室友选择。在所有十个模型中，AAVE和NP与比SAE更负面的形容词持续相关，其中NP受到的惩罚最为严重。关键是，每种方言的惩罚是通过不同的刻板印象集群实现的，而不是通过一个通用的非标准类别。NSE因其制度声望而表现出情境依赖的变化：在正式的租户筛选中优于SAE，但随着社交亲密度的增加而受到越来越多的惩罚。我们的研究结果揭示，LLMs在种族身份和声望维度上继承了隐性方言偏见，反映了已记录的人类住房歧视，并展示了其在后殖民英语方言中的影响。
cs.CL / 34 / 2609.18106
Linguistic Triggers of Gender and Racial Bias in Open-Weight LLMs Applied to Recruitment
开放权重大型语言模型在招聘中性别和种族偏见的语言触发因素
Abstract
Open-weight large language models are rapidly entering hiring pipelines, yet their discriminatory failure modes -- and the regulatory exposure these create under the EU AI Act high-risk classification (Annex III) and U.S. EEOC adverse-impact analysis -- remain poorly understood. We present the first systematic, multi-model audit of open-weight LLMs that treats job-posting language as the primary experimental variable, evaluating six models (Llama 3.2, Mistral, Gemma 3, Qwen 3, Phi 3, DeepSeek-R1) across four controlled experiments that jointly probe recruiter-simulation and job-seeker-simulation tasks. We find that (1) agentic posting language depresses recruiter recommendation scores for female candidates (r_rb = 0.309, p_Bonf = 7x10^-5; model-fixed-effects r_rb = 0.448), while communal language partially reverses the penalty; and (2) coded-exclusion language suppresses non-White recruiter scores at large effect sizes (r_rb = 0.646-0.758) and, on the job-seeker side, selectively deters non-White personas from expressing interest -- operationalizing a chilling-effect mechanism at scale. A label-ablation experiment isolates the explicit demographic persona label as the primary causal driver, and Word Embedding Association Tests corroborate these findings at the representational level (d = 1.01-1.45 under Caliskan et al.'s multi-word gender attribute lists). We translate these results into a concrete pre-deployment audit protocol -- posting-vocabulary scoring, persona-conditioned LLM probing, and adverse-impact flagging against the four-fifths threshold -- that operationalizes the documentation and risk-management obligations Annex III imposes on high-risk AI in recruitment.
Chinese Translation
开放权重大型语言模型正迅速进入招聘流程，然而其歧视性失效模式及其在欧盟《人工智能法案》高风险分类（附件三）和美国平等就业机会委员会（EEOC）不利影响分析下所带来的监管风险，仍然缺乏充分理解。本文首次对开放权重大型语言模型进行系统的多模型审计，将招聘广告语言作为主要实验变量，评估了六个模型（Llama 3.2、Mistral、Gemma 3、Qwen 3、Phi 3、DeepSeek-R1），通过四个受控实验联合探究招聘者模拟和求职者模拟任务。研究发现：(1) 具有主动性的招聘语言会降低女性候选人的招聘推荐分数（r_rb = 0.309，p_Bonf = 7×10^-5；模型固定效应 r_rb = 0.448），而具有共情性的语言则部分缓解了这一惩罚；(2) 编码排斥语言显著抑制非白人候选人的招聘者评分（r_rb = 0.646-0.758），且在求职者侧选择性地阻止非白人人设表达兴趣——在大规模上实现了寒蝉效应机制。标签消融实验将显性的人口统计学人设标签确定为主要因果驱动因素，词嵌入关联测试（Word Embedding Association Tests）在表征层面上（基于Caliskan等人的多词性别属性列表，d = 1.01-1.45）进一步证实了这些发现。我们将这些结果转化为具体的部署前审计协议——包括招聘广告词汇评分、人设条件下的大型语言模型探测以及基于四分之三规则的不利影响标记——以实现附件三对招聘中高风险人工智能所施加的文档编制和风险管理义务。
cs.CL / 35 / 2609.18135
DualSQL: Text-to-SQL with Multi-Agent Reinforcement Learning
DualSQL：基于多智能体强化学习的文本到SQL转换
Abstract
State-of-the-art Text-to-SQL systems are typically multi-agent pipelines centered around two fundamental tasks: schema linking and SQL generation. However, existing work trains separate models for each task, failing to leverage the synergy between these interrelated tasks. In this work, we propose DualSQL, a new Text-to-SQL system consisting of two agents powered by a single model backbone. The agents share the same model weights and agentic scaffold, enabling joint optimization through a robust multi-agent reinforcement learning (RL) framework. We design three database access tools to facilitate effective multi-step reasoning grounded to interactions with the databases. To improve training and avoid model collapse, we introduce a set of rollout guardrail mechanisms that stabilizes multi-agent RL training, supporting DualSQL to keep improving during training. We also introduce a new SQL correctness metric, robust execution match (REX), to more accurately judge SQL correctness and assign reward signals. Being trained on only 3755 examples, DualSQL-4B achieves an impressive 68.0% execution accuracy on the BIRD development set, matching previous 7B models. DualSQL-8B further improves to 71.1%, outperforming previous state-of-the-art single-model solutions with 32B parameters. These results demonstrate the strength of joint multi-agent reinforcement learning for building high performance Text-to-SQL pipelines.
Chinese Translation
最先进的文本到SQL系统通常是围绕两个基本任务构建的多智能体管道：模式链接和SQL生成。然而，现有的研究为每个任务训练了独立的模型，未能利用这些相互关联任务之间的协同效应。在本研究中，我们提出了DualSQL，这是一种由单一模型骨干驱动的新的文本到SQL系统，包含两个智能体。这些智能体共享相同的模型权重和智能体框架，使得通过强大的多智能体强化学习（RL）框架实现联合优化成为可能。我们设计了三种数据库访问工具，以促进基于与数据库交互的有效多步骤推理。为了改善训练并避免模型崩溃，我们引入了一套回滚保护机制，以稳定多智能体RL训练，支持DualSQL在训练过程中持续改进。我们还引入了一种新的SQL正确性度量标准，稳健执行匹配（REX），以更准确地判断SQL的正确性并分配奖励信号。经过仅3755个示例的训练，DualSQL-4B在BIRD开发集上达到了令人印象深刻的68.0%的执行准确率，匹配之前的7B模型。DualSQL-8B进一步提升至71.1%，超越了之前32B参数的最先进单模型解决方案。这些结果展示了联合多智能体强化学习在构建高性能文本到SQL管道中的优势。
cs.CL / 36 / 2609.18156
TeochewBench: A Human-Reviewed Benchmark for Teochew Hanzi Translation
潮州基准：一个经过人工审核的潮州汉字翻译基准
Abstract
Teochew has a substantial speaker community and exhibits distinctive lexical, syntactic, and pragmatic features, yet textual resources for evaluating large language models remain limited. We present TeochewBench, a human-reviewed benchmark comprising 300 Teochew Hanzi expressions for evaluating translation from Teochew Hanzi into Mandarin Chinese and English. The dataset covers five categories: basic vocabulary; everyday sentences; Teochew-specific expressions; tone, politeness, and context; and idiomatic, ambiguous, and culturally specific expressions. A primary Teochew-speaking reviewer examined all entries individually and revised them as needed, while two additional Teochew speakers verified selected items. Our main evaluation covers 11 official general-purpose post-trained models on the reviewed dataset in both translation directions, yielding 6,600 predictions. Two official base checkpoints provide 1,200 predictions for supplementary diagnostics, bringing the total to 13 models and 7,800 predictions. We additionally include a Hanzi-copy control, which returns the source input unchanged, to assess how shared Hanzi affect automatic scores for translation into Mandarin Chinese. Qwen3.5-27B achieved the highest overall chrF-style score among the evaluated checkpoints, at 60.63, followed by Qwen2.5-72B-Instruct at 56.61, Gemma-3-27B-IT at 56.36, and GLM-4-32B-0414 at 55.82. Across the 11 main-evaluation models, the mean chrF-style score decreased from 69.25 for low-specificity items to 27.52 for high-specificity items. High-specificity expressions received lower scores and exhibited smaller cross-model differences, suggesting that they constitute a shared low-scoring region across the model families evaluated here. The Hanzi-copy control further indicates that surface overlap in low-specificity items can substantially affect automatic scores for translation into Mandarin Chinese.
Chinese Translation
潮州话拥有庞大的说话群体，并展现出独特的词汇、句法和语用特征，但用于评估大型语言模型的文本资源仍然有限。我们提出了潮州基准（TeochewBench），这是一个经过人工审核的基准，包含300个潮州汉字表达，用于评估从潮州汉字到普通话和英语的翻译。该数据集涵盖五个类别：基础词汇；日常句子；潮州特有表达；语调、礼貌和语境；以及习惯用语、模棱两可和文化特定表达。一位主要的潮州话审阅者逐一检查了所有条目并根据需要进行了修订，同时另外两位潮州话说话者验证了选定的条目。我们的主要评估涵盖了11个官方的通用后训练模型，在经过审核的数据集上进行双向翻译，产生了6600个预测。两个官方基础检查点提供了1200个预测用于补充诊断，使总数达到13个模型和7800个预测。我们还包括了一个汉字复制控制，其返回的源输入保持不变，以评估共享汉字如何影响翻译成普通话的自动评分。Qwen3.5-27B在评估的检查点中获得了最高的整体 chrF 风格评分，达到了60.63，其次是 Qwen2.5-72B-Instruct 的56.61，Gemma-3-27B-IT 的56.36，以及 GLM-4-32B-0414 的55.82。在11个主要评估模型中，低特异性条目的平均 chrF 风格评分从69.25下降到高特异性条目的27.52。高特异性表达获得了较低的评分，并表现出较小的跨模型差异，表明它们构成了在这里评估的模型家族中的一个共享低评分区域。汉字复制控制进一步表明，低特异性条目中的表面重叠可以显著影响翻译成普通话的自动评分。
cs.CL / 37 / 2609.18194
T-SANDHI: Tone Sandhi-aware Adaptive Network with Decoupled Hybrid Injection for Low-resource Taiwanese Hokkien Speech Recognition
T-SANDHI：具有解耦混合注入的音调变调感知自适应网络，用于低资源的台湾闽南语语音识别
Abstract
In Taiwanese Hokkien automatic speech recognition (ASR), prior studies often treat tone sandhi as a major challenge under the assumption that models fail to process implicit phonological variations. However, our experiments on Taiwanese Hokkien reveal that speech foundation models actually handle tone sandhi variations effectively, and the real performance bottleneck stems from a localized confusion between these variations and retained citation tones. To address this, we propose T-SANDHI to explicitly decouple surface acoustics from underlying lexical intent on top of a frozen Whisper backbone. Using a lexicon-guided multi-task learning structure driven by text-derived pseudo labels, our lightweight hybrid injection module integrates independent citation and sandhi phonetic streams via dynamic gating. Extensive evaluation on the TAT-MOE corpus and two blind test sets demonstrates that this explicit disentanglement effectively resolves tonal mapping confusion, outperforming baselines with strict parameter efficiency.
Chinese Translation
在台湾闽南语自动语音识别（ASR）中，以往的研究通常将音调变调视为一个主要挑战，假设模型无法处理隐含的音韵变异。然而，我们对台湾闽南语的实验表明，语音基础模型实际上能够有效处理音调变调的变异，而真正的性能瓶颈源于这些变异与保留的引文音调之间的局部混淆。为了解决这个问题，我们提出了T-SANDHI，旨在明确地将表面声学与潜在的词汇意图解耦，并基于一个冻结的Whisper主干网络。通过使用由文本派生的伪标签驱动的词典引导多任务学习结构，我们的轻量级混合注入模块通过动态门控集成独立的引文和变调音韵流。对TAT-MOE语料库和两个盲测集的广泛评估表明，这种明确的解耦有效地解决了音调映射混淆，超越了严格参数效率的基线。
cs.CL / 38 / 2609.18203
Behavior2Value: Benchmarking and Empowering LLMs for Consumer Value Measurement from E-commerce Behaviors
行为到价值：基于电子商务行为的消费者价值测量的基准测试与赋能
Abstract
Human values are deep motivational orientations that shape human behaviors. In e-commerce, they reveal the stable drivers behind users' purchase decisions. Compared with short-term interests, consumer values better explain how users evaluate products before purchase. However, consumer values are often implicit in complex and fragmented behavioral trajectories, leaving value measurement from e-commerce behaviors largely underexplored. To this end, we propose the Behavior-to-Value (B2V) task, which aims to identify consumer values from e-commerce behavioral trajectories. Centered on this task, we first construct the E-commerce Consumption Value Taxonomy (ECVT) and introduce B2V-Bench, the first B2V dataset and benchmark, based on anonymized Taobao behavioral logs. B2V-Bench consists of real-world purchase decision episodes, covering 25 types of purchase behaviors, along with corresponding consumer value orientations manifested in each episode. To improve consumer value measurement accuracy, we further present B2V-Verifier, a behavior-to-value measurement model based on Value Verification Tuning, which learns to assess whether behaviors provide sufficient evidence for each value inference. Experiments show that B2V-Verifier outperforms strong LLM baselines, improving multi-label classification by 34\%. The dataset and code will be publicly released upon acceptance.
Chinese Translation
人类价值观是深层的动机导向，塑造着人类行为。在电子商务中，它们揭示了用户购买决策背后的稳定驱动因素。与短期利益相比，消费者价值更能解释用户在购买前如何评估产品。然而，消费者价值通常隐含在复杂且碎片化的行为轨迹中，使得从电子商务行为中进行价值测量的研究尚未得到充分探索。为此，我们提出了行为到价值（Behavior-to-Value, B2V）任务，旨在从电子商务行为轨迹中识别消费者价值。围绕这一任务，我们首先构建了电子商务消费价值分类法（E-commerce Consumption Value Taxonomy, ECVT），并基于匿名的淘宝行为日志推出了B2V-Bench，这是首个B2V数据集和基准。B2V-Bench包含真实的购买决策事件，涵盖25种购买行为，以及在每个事件中体现的相应消费者价值取向。为了提高消费者价值测量的准确性，我们进一步提出了B2V-Verifier，这是一种基于价值验证调优（Value Verification Tuning）的行为到价值测量模型，旨在学习评估行为是否为每个价值推断提供了充分的证据。实验表明，B2V-Verifier的表现优于强大的大型语言模型（LLM）基准，提升了多标签分类的准确率达34%。数据集和代码将在论文接受后公开发布。
cs.CL / 39 / 2609.18204
Beyond Accuracy: How Procedural Traces Shift the Decision Criterion of LLM Overseers
超越准确性：程序性痕迹如何改变大型语言模型监督者的决策标准
Abstract
Organizations increasingly use oversight loops where one large language model (LLM) audits another's outputs alongside procedural traces of claimed steps. A common concern about such LLM-as-a-judge pipelines is that detailed traces make overseers gullible. Using signal detection theory, we audit five LLM overseers on 19 compliance tasks (4,551 analyzed judgments), varying only trace detail and evidence labeling. With disconfirming evidence always visible, error detection remains near ceiling. Instead, elaborate traces shift the decision criterion toward rejection, increasing false alarms in susceptible overseers. Without option labels, human-validated reason coding shows about 60% of false alarms cite an inability to tie evidence to its option. Labels eliminate this stated reason, yet residual rejection of correct work persists in those overseers and rises with trace detail. Procedural traces thus act as governance artifacts that shape oversight decisions. AI auditors should be evaluated by their decision criterion and false-alarm behavior, alongside accuracy.
Chinese Translation
组织越来越多地使用监督循环，其中一个大型语言模型（LLM）审计另一个模型的输出，并结合所声称步骤的程序性痕迹。关于这种LLM作为裁判的流程的一个普遍担忧是，详细的痕迹使监督者变得轻信。通过使用信号检测理论，我们对五个LLM监督者在19个合规任务（分析了4,551个判断）上进行了审计，仅改变痕迹的详细程度和证据标记。在反驳证据始终可见的情况下，错误检测仍然接近上限。相反，复杂的痕迹将决策标准向拒绝倾斜，增加了易受影响的监督者的误报率。在没有选项标签的情况下，人类验证的理由编码显示约60%的误报引用了无法将证据与其选项关联的原因。标签消除了这一陈述的原因，但在这些监督者中，正确工作的残余拒绝依然存在，并随着痕迹详细程度的增加而上升。因此，程序性痕迹作为治理工件，塑造了监督决策。AI审计者应根据其决策标准和误报行为，以及准确性进行评估。
cs.CL / 40 / 2609.18274
I code or AI code: A comparative evaluation of AI-rated scores in classroom observations
我评分还是AI评分：课堂观察中AI评分的比较评估
Abstract
Classroom observations are widely recognized as a key tool for establishing benchmarks of education quality and guiding pedagogical improvement, yet they remain resource-intensive and dependent on trained observers. This study evaluated the feasibility of using a LLM (GPT-5 model) to score teacher-child interactions in early childhood classrooms, benchmarked against human raters. The study analyzed 87 video-recorded observations from 38 classrooms across 30 kindergartens in Hong Kong. Using observation transcripts, the AI model was configured to apply the full Classroom Assessment Scoring System (CLASS) framework. AI-rated scores were then compared with human ratings by examining correlations and differences in mean scores of the CLASS domains and dimensions. The results showed greater convergence between AI and raters for the Emotional Support domain and, in particular, the Quality of Feedback dimension, which captures how teachers use feedback to extend children's learning. Greater divergence emerged for interactions that were more procedural or context-dependent, particularly within the Classroom Organization and Instructional Support domains. These findings suggest that transcript-based AI scoring may capture some of the relative variation in teacher-child interactions but cannot yet reproduce calibrated human judgements consistently across the full CLASS framework. AI-assisted observation may therefore be more appropriate as a preliminary screening tool rather than as a replacement for trained observers, providing teachers with evidence for reflection rather than high-stakes evaluation. Future research should examine whether domain-specific training and incorporation of contextual and visual information can improve alignment between AI and human rated scores.
Chinese Translation
课堂观察被广泛认为是建立教育质量基准和指导教学改进的关键工具，然而它们仍然资源密集且依赖于经过培训的观察者。本研究评估了使用大型语言模型（LLM，GPT-5模型）对幼儿园课堂中教师与儿童互动进行评分的可行性，并与人类评分者进行了基准比较。研究分析了来自香港30所幼儿园的38个课堂中的87个视频录制观察。通过观察记录，AI模型被配置为应用完整的课堂评估评分系统（CLASS）框架。然后，通过检查CLASS领域和维度的相关性和均值评分差异，将AI评分与人类评分进行了比较。结果显示，AI与评分者在情感支持领域，特别是在反馈质量维度上的一致性更高，该维度反映了教师如何利用反馈来延伸儿童的学习。而在课堂组织和教学支持领域，针对更具程序性或依赖于上下文的互动，AI评分与人类评分的差异则更大。这些发现表明，基于记录的AI评分可能捕捉到教师与儿童互动中的相对变化，但尚无法在完整的CLASS框架下始终如一地再现经过校准的人类判断。因此，AI辅助观察可能更适合作为初步筛选工具，而不是替代经过培训的观察者，为教师提供反思的证据，而非高风险的评估。未来的研究应探讨领域特定的培训和上下文及视觉信息的结合是否能改善AI与人类评分之间的一致性。
cs.CL / 41 / 2609.18282
Too Good to Be Real? Diagnosing and Reducing the Gap Between AI Preference and Real User Engagement
太好以至于不真实？诊断并缩小人工智能偏好与真实用户参与之间的差距
Abstract
Large language models are increasingly used to generate and evaluate online content, yet it remains unclear whether the qualities they associate with higher engagement match what real users respond to. We study this question using 1.17 million answers to 25,978 questions from Zhihu, Quora, and Reddit, comparing real platform answers and AI-generated answers across four within-question engagement levels. We introduce Ontological Preference Measurement, which represents answers along three dimensions: logic, affect, and expression. We find a systematic gap between AI preference and real user engagement: as target engagement increases, LLMs add more explicit logical structure, while real user engagement is more strongly associated with affective and expressive salience. We call this tendency logic overbinding. Based on this diagnosis, we propose Ontology-Masked Reasoning Autoencoding (OMRA), a controlled intervention that masks and reconstructs over-explained spans while preserving stance, factual content, and coherence. Across four LLM families, OMRA reduces the measured gap by an average of 54.4%. In human evaluation, OMRA wins 62.4% of pairwise preference judgments against matched real platform answers, even though the real answers are more often judged to be human-written.
Chinese Translation
大型语言模型越来越多地用于生成和评估在线内容，但尚不清楚它们与更高参与度相关的特质是否与真实用户的反应相匹配。我们通过分析来自知乎、Quora和Reddit的25,978个问题的117万答案来研究这个问题，比较真实平台答案和人工智能生成答案在四个问题内参与度水平上的表现。我们引入了本体偏好测量（Ontological Preference Measurement），该测量将答案沿逻辑、情感和表达三个维度进行表示。我们发现人工智能偏好与真实用户参与之间存在系统性的差距：随着目标参与度的提高，LLM（大型语言模型）增加了更多显式的逻辑结构，而真实用户参与则更强烈地与情感和表达的显著性相关。我们称这种倾向为逻辑过度绑定（logic overbinding）。基于这一诊断，我们提出了本体掩蔽推理自编码（Ontology-Masked Reasoning Autoencoding, OMRA），这是一种控制干预方法，掩蔽并重构过度解释的部分，同时保留立场、事实内容和连贯性。在四个LLM家族中，OMRA平均减少了54.4%的测量差距。在人类评估中，OMRA在与匹配的真实平台答案的成对偏好判断中获胜62.4%，尽管真实答案更常被判断为人类撰写。
cs.CL / 42 / 2609.18284
Made in Hungary: Comments on the performance of generative language models
匈牙利制造：关于生成语言模型性能的评论
Abstract
In recent years, three initiatives have emerged to develop generative language models in Hungary. The motivation behind them is the same. For Hungarian, no model with the given capability existed, or existing English-centric models offered limited proficiency. A detailed examination of the corresponding studies, however, reveals several methodological limitations. First, the reliability of the evaluation protocols is questionable. Contrary to the findings of Csibi et al. [2026], evaluation under the recommended inference settings shows that Qwen3-4B achieves higher scores than Racka-4B, its Hungarian-adapted version. Data contamination is evident in the work of Yang et al. [2025d] and Szentmih\'alyi et al. [2025], potentially biasing the reported results. Second, the training pipelines fall short of current best practices in corpus curation and data mixture, which risks wasting substantial compute on low-quality data. The lack of controlled ablations prevents reliable assessment of these choices. Third, none of the three papers assessed forgetting or capability loss. Testing the adapted models on a subset of the original benchmarks indicates performance decline in all three cases, especially Racka-4B. These observations emphasize the importance of rigorous experimental design in language model development, given the significant computational and financial costs involved.
Chinese Translation
近年来，匈牙利出现了三个开发生成语言模型的倡议。它们背后的动机是相同的。对于匈牙利语，尚不存在具有相应能力的模型，或者现有的以英语为中心的模型提供的能力有限。然而，对相关研究的详细审查揭示了几个方法论的局限性。首先，评估协议的可靠性值得怀疑。与Csibi等人[2026]的发现相反，在推荐的推理设置下的评估显示，Qwen3-4B的得分高于其匈牙利适配版本Racka-4B。Yang等人[2025d]和Szentmihályi等人[2025]的研究中明显存在数据污染，可能导致报告结果的偏差。其次，训练流程未能达到当前语料库整理和数据混合的最佳实践，这可能导致在低质量数据上浪费大量计算资源。缺乏控制性消融实验阻碍了对这些选择的可靠评估。第三，三篇论文均未评估遗忘或能力损失。在原始基准的子集上测试适配模型表明，在所有三种情况下，性能都有下降，尤其是Racka-4B。这些观察强调了在语言模型开发中严格实验设计的重要性，考虑到涉及的显著计算和财务成本。
cs.CL / 43 / 2609.18304
Rollback the World, Keep the Reflection: Rollback-Induced Reflection for Long-Horizon LLM Agents
回滚世界，保持反思：针对长时间跨度 LLM 代理的回滚诱导反思
Abstract
Large language model (LLM) agents increasingly tackle long-horizon tasks through multi-step environment interaction, yet a single erroneous action can alter subsequent states and observations, causing errors to compound over time. Existing methods either correct the context without repairing altered environment states or restore earlier states while discarding useful experience, making it difficult to both eliminate failure conditions and avoid repeating past mistakes. We argue that reliable recovery should instead be treated as a rollback-boundary control problem that jointly determines when to intervene, where to resume, and what information should survive recovery. Based on this view, we propose Rollback-Induced Reflection (RIR), a unified recovery framework that restores execution to a selected prior state while carrying forward reusable knowledge distilled from the abandoned trajectory to guide subsequent decisions. We further characterize recovery through a unified operator over rollback depth and retained memory, providing a general view of state restoration and knowledge retention. Experiments on three long-horizon benchmarks demonstrate that RIR consistently improves task performance across multiple LLM backbones, with structured reflection memory preserving useful experience and selective rollback enabling efficient recovery.
Chinese Translation
大型语言模型（LLM）代理越来越多地通过多步环境交互来处理长时间跨度的任务，但单个错误的动作可能会改变后续状态和观察，导致错误随着时间的推移而累积。现有方法要么在不修复改变的环境状态的情况下纠正上下文，要么在丢弃有用经验的同时恢复早期状态，这使得既难以消除失败条件又难以避免重复过去的错误。我们认为，可靠的恢复应被视为一个回滚边界控制问题，联合确定何时干预、何处恢复以及哪些信息应在恢复中保留。基于这一观点，我们提出了回滚诱导反思（Rollback-Induced Reflection, RIR），这是一个统一的恢复框架，它将执行恢复到选定的先前状态，同时将从被放弃轨迹中提炼出的可重用知识传递到后续决策中。我们进一步通过对回滚深度和保留记忆的统一操作来表征恢复，提供了状态恢复和知识保留的总体视角。在三个长时间跨度基准上的实验表明，RIR 在多个 LLM 主干上持续提高任务性能，结构化的反思记忆保留了有用的经验，而选择性回滚则实现了高效的恢复。
cs.CL / 44 / 2609.18310
SEA-LION-v4.8: A Technical Report
SEA-LION-v4.8：技术报告
Dabeer, Ahmed Mohammad, Jeongmi, Ahn, Sutaveephamochanon, Anocha, Sajeban, Antonyrex, Adila, Aulia, Teng, Chan Hok, Adwin, Yi, Cheng Zi, Ziyi, Nicholas Zhuang, Esther, Choa Hsueh Mei, Tat-Wee, David Ong, Phin, Evelyn Tan Chor, Peng, Heng Cheng, Jonathan, Ren, Lee Chwan, Yi, Leong Wai, Qi, Leong Wei, Sipp, Leslie Teo Eng, Rachel, Liew, Peerat, Limkonchotiwat, Estrada, Montalan Jann Railey, Mokhtar, Muhammad Ridzuan Bin, Karthik, Nagarajan, Cheong, Ng Boon, Raymond, Gang, Ngui Jian, Ngan, Nguyen Thanh, Panuthep, Tasawong, Gregory, Pereira Mark, Benjamin, Phang Shi Wei, Hung, Poon Yip, Joseph, Hamsawardhini, Rengarajan, Bryan, Siow Wei Kang, Chia, Tai Ngee, Meng, Tan Choon, Min, Tan Le, Sheryl, Wei, Tan Siao, Xian, Tan Yi, Yun, Tee Jun, Wai, Teng Kok, Chandra, Tjhi William, Pume, Tuchinda, Donghang, Wu, Xianbin, Yong, Yosephine, Zhou, Zhang
Abstract
We introduce Nemotron-SEA-LION-v4.8, a family of Southeast Asian Languages in One Network (SEA-LION) built upon NVIDIA Nemotron 3. The family includes 30B-A3B and 120B-A12B models, with both continued-pretrained base checkpoints and post-trained variants. We adapt the models using Southeast Asian, reasoning, code, and multilingual parallel datasets, followed by post-training with supervised fine-tuning and online on-policy distillation. On SEA-HELM, the 30B-A3B model improves the overall SEA score from 46.06 to 51.57, while the 120B-A12B model improves from 49.30 to 63.44. The strongest gains are observed in instruction following, natural language reasoning, and natural language understanding across seven Southeast Asian languages.
Chinese Translation
我们介绍了Nemotron-SEA-LION-v4.8，这是一个基于NVIDIA Nemotron 3构建的东南亚语言网络（Southeast Asian Languages in One Network, SEA-LION）系列。该系列包括30B-A3B和120B-A12B模型，提供了继续预训练的基础检查点和后训练变体。我们使用东南亚、推理、代码和多语言平行数据集对模型进行了适配，随后进行了监督微调和在线策略蒸馏的后训练。在SEA-HELM上，30B-A3B模型将整体SEA得分从46.06提高到51.57，而120B-A12B模型则从49.30提高到63.44。我们观察到在七种东南亚语言中，指令跟随、自然语言推理和自然语言理解的提升最为显著。
cs.CL / 45 / 2609.18317
Knowledge-Graph Based Augmentation versus Retrieval Augmented Generation for Cultural-Related Question Answering
基于知识图谱的增强与检索增强生成在文化相关问答中的比较
Abstract
Large language models (LLMs) suffer from a long-tail deficit: culturally specific facts, particularly those concerning underrepresented regions such as Latin America, appear too rarely in pretraining corpora to be reliably memorized. Retrieval-Augmented Generation (RAG) addresses this by grounding generation in external text, but structured alternatives such as Knowledge Graphs (KGs) offer tighter control over what enters the context, along with potential gains in explainability and updatability. We benchmark Graph-RAG against standard RAG on LatamQA, a culturally grounded multiple-choice dataset spanning eight thematic categories. The graphs are built end-to-end from Wikipedia articles with KGGen, a recent open-domain extractor, without manual curation in our main setting. G-Retriever is competitive with RAG and reduces the error of the base LLM by 72\% with a standard KG and 78\% with a benchmark-aware variant, the gap to RAG narrowing further as the graph is oriented toward task-relevant content. The trained projection transfers zero-shot to Portuguese without target-language fine-tuning, indicating multilingual reach.
Chinese Translation
大型语言模型（LLMs）存在长尾缺陷：文化特定事实，尤其是那些涉及拉丁美洲等代表性不足地区的事实，在预训练语料库中出现得过于稀少，无法被可靠地记忆。检索增强生成（RAG）通过将生成过程与外部文本相结合来解决这一问题，但结构化的替代方案如知识图谱（KGs）则提供了对进入上下文内容的更严格控制，同时在可解释性和可更新性方面具有潜在优势。我们在LatamQA上对Graph-RAG与标准RAG进行了基准测试，该数据集是一个涵盖八个主题类别的文化基础多项选择题数据集。图谱是通过KGGen这一最新的开放域提取器从维基百科文章中端到端构建的，在我们的主要设置中没有进行人工策划。G-Retriever与RAG具有竞争力，并且在标准KG下将基础LLM的错误率降低了72%，在基准感知变体下降低了78%，随着图谱朝向任务相关内容的倾斜，G-Retriever与RAG之间的差距进一步缩小。经过训练的投影在没有目标语言微调的情况下可以零-shot转移到葡萄牙语，表明其多语言能力。
cs.CL / 46 / 2609.18320
Attention Dispersion as a Diagnostic Signal for Hallucination in Large Language Models
注意力分散作为大型语言模型幻觉的诊断信号
Abstract
Large Language Models (LLMs) frequently exhibit hallucinations, presenting a major barrier to reliability in complex reasoning tasks. While traditional detection methods rely on output-based confidence metrics, these logits are often miscalibrated by modern alignment techniques. In this paper, we investigate the temporal volatility of internal attention mechanisms as an alternative diagnostic signal for hallucination that does not depend on output calibration. By introducing an unsupervised metric for attention dispersion, we show that epistemic uncertainty leaves a measurable trace within intermediate layers, where spikes in attention entropy are associated with reasoning breakdowns. We evaluate our approach on mathematical reasoning benchmarks (GSM8K and MATH-500) using the Qwen2.5 model family (1.5B and 3B parameters), finding statistically significant AUC improvements of up to +0.076 over output-based baselines across all tested conditions. These findings suggest that attention dispersion is a promising complement to traditional hallucination detection methods, requiring further investigation across broader model families and task domains.
Chinese Translation
大型语言模型（LLMs）经常出现幻觉，这在复杂推理任务中构成了可靠性的主要障碍。虽然传统的检测方法依赖于基于输出的置信度指标，但这些 logits 经常受到现代对齐技术的误校准。本文探讨了内部注意力机制的时间波动性，作为一种不依赖于输出校准的幻觉替代诊断信号。通过引入一种无监督的注意力分散度量，我们表明，认知不确定性在中间层中留下了可测量的痕迹，其中注意力熵的峰值与推理崩溃相关。我们在数学推理基准（GSM8K 和 MATH-500）上评估了我们的方法，使用 Qwen2.5 模型系列（1.5B 和 3B 参数），发现相较于基于输出的基线，在所有测试条件下 AUC 有统计显著性提升，最高可达 +0.076。这些发现表明，注意力分散是传统幻觉检测方法的有希望的补充，值得在更广泛的模型系列和任务领域中进一步研究。
cs.CL / 47 / 2609.18385
Emotion Experience, Expression, and Perception: Emotion Analysis on Multimodal Social Media Posts
情感体验、表达与感知：多模态社交媒体帖子中的情感分析
Abstract
Emotions are an essential aspect of human communication, particularly on social media, where authors frequently combine text and images to convey their emotions. Yet prior work on emotion analysis of social media posts has overlooked two important aspects in regard to measuring how well readers can reconstruct the authors' intent: (1)~the image modality, with most work focusing solely on text, and (2)~the real-world events that trigger the expressed emotions, and their relationship to the post content. We therefore study the relation between (a) the author's experience of the event that caused them to write a social media post and (b) the content of the post, with a focus on readers' capability to reconstruct that emotion expression. To do that, we introduce the Multimodal Multi-Emotion-Model dataset Mult2EMo, created by collecting annotations from both authors and readers on the posts and their triggering events. We find that reconstruction is possible but challenging for both human readers and computational models. We show that understanding the triggering event is crucial for accurate reconstruction, and that reconstruction is particularly challenging when posts rely heavily on the image to express emotion.
Chinese Translation
情感是人类沟通的重要方面，尤其是在社交媒体上，作者常常结合文本和图像来传达他们的情感。然而，之前关于社交媒体帖子情感分析的研究忽视了两个重要方面，这些方面与测量读者重建作者意图的能力有关：（1）图像模态，大多数研究仅关注文本；（2）引发表达情感的现实事件及其与帖子内容的关系。因此，我们研究了（a）导致作者撰写社交媒体帖子的事件体验与（b）帖子内容之间的关系，重点关注读者重建该情感表达的能力。为此，我们引入了多模态多情感模型数据集Mult2EMo，该数据集通过收集作者和读者对帖子及其触发事件的注释而创建。我们发现重建是可能的，但对人类读者和计算模型来说都是具有挑战性的。我们表明，理解触发事件对于准确重建至关重要，并且当帖子在表达情感时严重依赖图像时，重建尤其具有挑战性。
cs.CL / 48 / 2609.18417
Dependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning
依赖感知的轨迹优化用于高效的多轮智能体微调
Abstract
Multi-turn agent trajectories often contain redundant rounds (failed tool calls, parallel sub-queries, verification-only steps) that inflate both training and inference cost. We propose viewing each trajectory as a \emph{round-level dependency DAG} that exposes which rounds are globally load-bearing for the final answer, and fine-tune agents on trajectories refined through this DAG. Given an LLM-annotated DAG, these edits are deterministic and interpretable, with optional rephrasing. Models trained on these refined trajectories consistently outperform those trained on the original trajectories at lower inference cost. Specifically, across four multi-modal QA benchmarks, our refinements improve downstream accuracy by up to $1.7$\,pp over vanilla SFT (and $5.7$\,pp over an LLM-deletion baseline) while reducing per-sample inference messages by up to approximately $40\%$ and inference tokens by up to approximately $48\%$, translating to substantial savings in compute and serving cost. Code is available.
Chinese Translation
多轮智能体的轨迹通常包含冗余的回合（失败的工具调用、并行子查询、仅验证步骤），这会增加训练和推理的成本。我们提出将每个轨迹视为一个 extit{回合级依赖有向无环图（DAG）}，该图揭示了哪些回合对最终答案具有全球负载能力，并在通过该DAG优化的轨迹上对智能体进行微调。给定一个由大语言模型（LLM）注释的DAG，这些编辑是确定性的和可解释的，并可选择性地进行改写。在这些优化轨迹上训练的模型在推理成本较低的情况下，始终优于在原始轨迹上训练的模型。具体而言，在四个多模态问答基准上，我们的优化提高了下游准确率，较普通的监督微调（SFT）提高了最多$1.7$个百分点（较LLM删除基线提高了$5.7$个百分点），同时每个样本的推理消息减少了约$40\%$，推理标记减少了约$48\ ext{ extperthousand}$，这转化为计算和服务成本的显著节省。代码可用。
cs.CL / 49 / 2609.18440
Planning or Improvisation? Stress-Testing the Poetry Planning Site on Open Models and Open Cross-Layer Transcoders
规划还是即兴创作？对开放模型和开放跨层转码器的诗歌规划网站进行压力测试
Abstract
Lindsey et al. (2025) report that Claude 3.5 Haiku plans rhymes: features for candidate rhyme words are active on the newline before a line is written, and a suppress-and-inject intervention redirects the line only when applied there (their Figure 13). We test how far this generalizes on seven cells crossing four open models (0.6B to 2.6B parameters) with six open cross-layer transcoders (CLTs), on one consumer GPU, decomposing the claim into position specificity (C1), newline site identity (C2), and a newline-resident plan (C3). This is a stress test rather than a faithful reproduction: attribution graphs are unavailable for these CLTs, so features are found bottom-up from decoder vectors. C1 generalizes, in every cell and in all 247 of 444 prompt-by-inject pairs with a detectable effect, but the effective position is the final prompt token, adjacent to emission, and only two cells reach behaviorally meaningful probabilities. C2 and C3 are not recovered by any probe: a census of every active feature finds no rhyme-anticipating enrichment at the newline, and steering the newline while the model composes the whole line, over 36 runs and 8,640 sampled lines, shows why. That intervention is strong but one token long, making the injected word the first word of the composed line in 703 of 720 samples and leaving the rhyme six words later untouched. A final test drops the transcoder entirely: patching the newline's whole residual, at every layer, from a minimal-pair poem whose third line ends on a different rhyme moves the rhyme in 11 of 1,260 composed lines against 4 at baseline, with a design resolving 1.4%. We read this as a boundary condition rather than a refutation: at this scale and with these transcoders, the causal site is emission-adjacent. We reproduce Figure 13's shape, not its mechanism. Code and data are public (code: github.com/PCfVW/poetry-planning-site).
Chinese Translation
Lindsey 等人（2025）报告称，Claude 3.5 Haiku 计划押韵：候选押韵词的特征在一行写入之前的换行符上是活跃的，而抑制与注入干预仅在应用于此时重定向该行（见他们的图 13）。我们测试这一结果在七个跨越四个开放模型（0.6B 到 2.6B 参数）和六个开放跨层转码器（CLTs）上的普适性，使用一台消费级 GPU，将该主张分解为位置特异性（C1）、换行符位置身份（C2）和换行符驻留计划（C3）。这是一项压力测试，而非忠实再现：由于这些 CLTs 的归因图不可用，因此特征是从解码器向量自下而上找到的。C1 在每个单元中普适，在所有 444 个提示-注入对中有 247 个具有可检测的效果，但有效位置是最终的提示标记，紧邻发射，只有两个单元达到了行为上有意义的概率。C2 和 C3 没有通过任何探测器被恢复：对每个活跃特征的普查发现换行符处没有押韵预期的增强，而在模型创作整行时引导换行符的实验，经过 36 次运行和 8,640 行采样，显示了原因。该干预效果强大但仅限于一个标记，使得注入的词在 720 个样本中的 703 个中成为所创作行的第一个词，而留下六个词后的押韵未受影响。最后的测试完全去掉了转码器：从一个第三行以不同押韵结束的最小对诗中修补换行符的整个残差，在 1,260 行创作的诗中使押韵移动了 11 次，而基线仅为 4 次，设计的解决率为 1.4%。我们将此视为边界条件，而非反驳：在此规模和这些转码器下，因果位置是紧邻发射的。我们再现了图 13 的形状，而非其机制。代码和数据是公开的（代码：github.com/PCfVW/poetry-planning-site）。
cs.CL / 50 / 2609.18445
M-SQE: Multilingual Skill Quality Estimation for Enhancing Language Equality in Agentic Skill Use
M-SQE：多语言技能质量评估以增强代理技能使用中的语言平等性
Abstract
Agent skills, reusable procedural documents that extend LLM agents beyond their parametric memory, have become an important interface for deploying agents on real-world tasks. Community-maintained skill libraries built around this interface are growing rapidly. However, this ecosystem remains deeply English-centric: our audit finds that low-resource languages such as Swahili and Hindi have no in-language skill content, so retrieval often returns a skill written in a different language than the query, degrading accuracy and recall. A practical solution is to synthesize in-language skills for retrieval but the quality can be unreliable, so relevance in this setting alone often surfaces a related but unusable candidate. To address this, we propose M-SQE, a post-retrieval Multilingual Skill Quality Estimation framework that scores candidates via a Theory view for intrinsic quality and an Action view for task-grounded utility, unified into a domain-conditioned final score. We evaluate M-SQE across three skill-use domains: general, tool-use, and cultural tasks. Empirically, we build three-layer candidate skill pools mirroring today's ecosystem, where M-SQE's task success exceeds existing baseline's average by at least +3.5 points across three different retrievers. Particularly, M-SQE lifts the lowest-resource languages most (+12.9pp on Hindi and +5.6pp on Swahili) and achieves strong performance across all six culture regions, thereby moving agentic skill use toward linguistic and cultural equality.
Chinese Translation
代理技能是可重用的程序文档，超越了大规模语言模型（LLM）代理的参数记忆，成为在现实任务中部署代理的重要接口。围绕这一接口构建的社区维护技能库正在迅速增长。然而，这一生态系统仍然深受英语中心主义的影响：我们的审计发现，斯瓦希里语和印地语等低资源语言没有本语言的技能内容，因此检索往往返回与查询语言不同的技能，降低了准确性和召回率。一种实用的解决方案是合成本语言技能以便检索，但其质量可能不可靠，因此在这种情况下的相关性往往会出现相关但不可用的候选项。为了解决这个问题，我们提出了M-SQE，一个后检索的多语言技能质量评估框架，通过理论视角评估内在质量，通过行动视角评估任务基础的实用性，统一为一个领域条件的最终得分。我们在三个技能使用领域（一般任务、工具使用和文化任务）中评估M-SQE。实证研究中，我们构建了三个层次的候选技能池，反映了当今的生态系统，其中M-SQE的任务成功率在三个不同的检索器中超过现有基线的平均值至少+3.5分。特别是，M-SQE对低资源语言的提升最为显著（印地语提升+12.9个百分点，斯瓦希里语提升+5.6个百分点），并在所有六个文化区域中表现出强劲的性能，从而推动代理技能使用朝向语言和文化平等。
cs.CL / 51 / 2609.18494
Size Matters: Foundation Model for Czech HTML documents
规模重要性：捷克HTML文档的基础模型
Abstract
Creating universal, high-quality representations of web documents in high-traffic industrial environments requires models that are both performant and economic. Existing approaches, however, often depend on large models, overlook the structural information inherent in HTML, or are constrained by short context windows, limiting their ability to process real-world web pages. We present HTML-LM, a compact foundation model with 154 million parameters that addresses these limitations through HTML-aware training and a ModernBERT-based architecture. It was trained on 100 million web documents using multiple objectives, including masked language modeling, bag-of-words prediction, and contrastive distillation from large language models. Consequently, HTML-LM sets a new state-of-the-art for classification and regression applications in the Czech Internet domain, surpassing both larger encoders and small-sized LLMs. The model is deployed in production, processing thousands of web documents per second, and released to the community under the CC BY-NC 4.0. https://huggingface.co/Seznam/html-lm.
Chinese Translation
在高流量工业环境中创建通用、高质量的网页文档表示需要性能和经济性兼备的模型。然而，现有方法往往依赖于大型模型，忽视了HTML中固有的结构信息，或受到短上下文窗口的限制，从而限制了其处理真实网页的能力。我们提出了HTML-LM，这是一种具有1.54亿参数的紧凑型基础模型，通过HTML感知训练和基于ModernBERT的架构来解决这些局限性。该模型在1亿个网页文档上进行了训练，采用了多种目标，包括掩码语言建模、词袋预测和从大型语言模型进行对比蒸馏。因此，HTML-LM在捷克互联网领域的分类和回归应用中设立了新的最先进水平，超越了更大编码器和小型LLM。该模型已在生产环境中部署，每秒处理数千个网页文档，并以CC BY-NC 4.0许可发布给社区。https://huggingface.co/Seznam/html-lm
cs.CL / 52 / 2609.18516
Align, Integrate, and Fire: Efficient Token-Level Alignment for Zero-Shot SpeechLLMs
对齐、整合与激发：零-shot语音大语言模型的高效标记级对齐
Abstract
While Large Language Models excel in natural language processing, efficiently extending their capabilities to spoken input remains a significant challenge. Existing methods for building SpeechLLMs often rely on computationally expensive full-model fine-tuning, or employ parameter-efficient projectors that suffer from inefficient token sequence lengths and costly full-model supervision. In this paper, we introduce Aligned Continuous Integrate-and-Fire, a highly efficient framework for zero-shot speech processing. Our method dynamically compresses continuous acoustic frames into the exact discrete token length of the target text utilizing explicit Dynamic Time Warping alignments. This allows our initial training stage to establish a robust acoustic-to-semantic bridge using lightweight distance metrics, entirely bypassing the computationally expensive LLM forward pass. For subsequent fine-tuning, we propose a memory-efficient knowledge distillation objective that targets a single LLM layer, performing competitively with full-model cross-entropy training at a fraction of the computational cost. Through extensive evaluations on Automatic Speech Recognition and Speech Translation, we demonstrate that our method achieves superior performance compared to prior parameter-efficient baselines.
Chinese Translation
尽管大语言模型在自然语言处理方面表现出色，但将其能力有效扩展到语音输入仍然是一个重大挑战。现有的构建语音大语言模型（SpeechLLMs）的方法通常依赖于计算成本高昂的全模型微调，或采用参数高效的投影器，这些方法在处理标记序列长度时效率低下，并且需要昂贵的全模型监督。在本文中，我们提出了对齐的连续积分与激发（Aligned Continuous Integrate-and-Fire），这是一个用于零-shot语音处理的高效框架。我们的方法利用显式的动态时间规整（Dynamic Time Warping）对齐，将连续的声学帧动态压缩为目标文本的确切离散标记长度。这使得我们的初始训练阶段能够使用轻量级距离度量建立稳健的声学到语义的桥梁，完全绕过计算成本高昂的LLM前向传递。对于后续的微调，我们提出了一种内存高效的知识蒸馏目标，针对单个LLM层进行优化，其性能在计算成本仅为全模型交叉熵训练的一小部分的情况下与之竞争。通过在自动语音识别和语音翻译上的广泛评估，我们证明了我们的方法在性能上优于先前的参数高效基线。
cs.CL / 53 / 2609.18529
Machine Translation between English and Syriac (East Syriac Dialect) using Statistical Machine Learning
基于统计机器学习的英语与叙利亚语（东叙利亚方言）之间的机器翻译
Abstract
UNESCO considers the Assyrian (Syriac) language an endangered language. Although Assyrians speak the language worldwide, the speaking population is uncertain (ranging from 500,000 to 1,500,000). Syriac is also one of the least studied languages in Natural Language Processing (NLP). Despite advances in Machine Translation (MT) over the past decade, the lack of publicly available corpora and the orthographic complexity of the Syriac script, specifically the Madnkhaya script, have left this language entirely ignored in the computational linguistics literature. This study develops the first phrase-based Statistical MT (SMT) model for English-to-Assyrian MT using the Moses framework. We created a dataset of 38,847 sentence pairs from the complete English and Syriac Bible, merging a pre-existing New Testament dataset with an Old Testament built from scratch through PDF extraction, using custom segmentation scripts and manual alignment review by three bilingual annotators. The Syriac side of the corpus undergoes diacritic removal and Byte-Pair Encoding tokenization to reduce orthographic sparsity before training. We trained and evaluated six models using different configurations and splitting-scheme ratios, language model order, distortion limits, and the inclusion of an Operation Sequence Model. The best-performing configuration achieves a word-level BLEU score of 23.54. Human evaluation by 11 native Assyrian speakers resulted in mean adequacy and fluency scores of 3.42 and 3.34 out of 5, respectively. These results are consistent with comparable low-resource SMT models trained on Biblical corpora for morphologically rich Semitic languages. The corpora, scripts, and trained model are publicly available, providing the research community with the first systematically curated English-Syriac dataset and a reproducible baseline for future MT and broader NLP work on this endangered language.
Chinese Translation
联合国教科文组织将亚述（叙利亚）语言视为一种濒危语言。尽管亚述人在全球范围内使用该语言，但说话人数尚不确定（在500,000到1,500,000之间）。叙利亚语也是自然语言处理（NLP）中研究最少的语言之一。尽管在过去十年中机器翻译（MT）取得了进展，但由于缺乏公开可用的语料库以及叙利亚文字的正字法复杂性，特别是Madnkhaya文字，使得该语言在计算语言学文献中完全被忽视。本研究开发了第一个基于短语的统计机器翻译（SMT）模型，用于英语到亚述语的翻译，采用Moses框架。我们从完整的英语和叙利亚语圣经中创建了一个包含38,847对句子的语料库，将现有的新约语料库与通过PDF提取从零构建的旧约语料库合并，使用自定义分割脚本和三名双语注释者的手动对齐审查。语料库的叙利亚语部分在训练前进行了去声调和字节对编码（Byte-Pair Encoding）标记化，以减少正字法稀疏性。我们使用不同的配置和拆分方案比例、语言模型阶数、扭曲限制以及包含操作序列模型训练和评估了六个模型。表现最佳的配置达到了23.54的词级BLEU分数。11名母语为亚述语的评估者的人工评估结果显示，平均适应性和流畅性得分分别为3.42和3.34（满分为5）。这些结果与针对形态丰富的闪米特语言在圣经语料库上训练的低资源SMT模型的可比性一致。语料库、脚本和训练模型均已公开，向研究界提供了第一个系统整理的英语-叙利亚语数据集以及未来机器翻译和更广泛的NLP工作在这一濒危语言上的可重复基准。
cs.CL / 54 / 2609.18533
A Probe Shift Is Not a Fairness Fix: The Limits of Representation Steering in Speech Models
探针偏移并不是公平性的解决方案：语音模型中表征引导的局限性
Abstract
Automatic speech recognition (ASR) systems exhibit unequal error rates across speaker groups, motivating interventions on their internal representations. We ask whether speaker-linked attributes that are linearly readable from pretrained ASR encoders yield useful directions for reducing group word-error-rate (WER) gaps. Across Whisper-medium, HuBERT-large, and Wav2Vec2-large on Common Voice and the Speech Accent Archive, we probe every encoder layer for metadata-derived sex/gender, age, and native/accent labels; construct centroid and probe-derived directions; inject them at selected layers; and compare downstream probe trajectories with matched WER changes. Sex labels are highly decodable (best macro-F1 0.924--0.941), native/accent labels are also above chance (0.544--0.696), and age is weaker (0.354--0.397). Of 22 post-selected reruns, nine have 95% paired-bootstrap intervals entirely below zero, yet every absolute source-group WER reduction is below 0.7 percentage points. Conversely, a local target-class probe rate can rise from 8.09% to 99.87% while WER worsens. Linear readability is therefore neither evidence of causal use nor a reliable mitigation method. Our results motivate evaluating speech-bias interventions jointly at representation, propagation, and task levels.
Chinese Translation
自动语音识别（ASR）系统在不同说话者群体之间表现出不平等的错误率，这促使我们对其内部表征进行干预。我们探讨了从预训练的ASR编码器中线性可读的与说话者相关的属性是否能为减少群体词错误率（WER）差距提供有用的方向。在Common Voice和Speech Accent Archive上，我们对Whisper-medium、HuBERT-large和Wav2Vec2-large的每一层编码器进行探测，获取基于元数据的性别/性别、年龄和母语/口音标签；构建质心和探针导出的方向；在选定层注入这些方向；并比较下游探针轨迹与匹配的WER变化。性别标签的可解码性较高（最佳宏F1为0.924--0.941），母语/口音标签也高于随机水平（0.544--0.696），而年龄的可解码性较弱（0.354--0.397）。在22次后选的重跑中，有9次的95%配对自助法区间完全低于零，但每个绝对源群体的WER减少均低于0.7个百分点。相反，局部目标类别的探针率可以从8.09%上升到99.87%，而WER却恶化。因此，线性可读性既不是因果使用的证据，也不是可靠的缓解方法。我们的结果促使我们在表征、传播和任务层面共同评估语音偏见干预措施。
cs.CL / 55 / 2609.18605
PACT: Can Enterprise AI Assistants Be Trusted Under Pressure?
PACT：企业人工智能助手在压力下能否被信任？
Abstract
As corporate AI adoption continues to grow, enterprise-grade LLM agents are being deployed into sensitive contexts such as hiring, healthcare, and finance. In these contexts, compliance with rules specified in an agent's system context is a first-order legal concern. Currently, no evaluation framework systematically measures which LLM models tend to violate compliance rules, especially under pressure from a persistent user, a hurried manager, or circumstances where violation is convenient or attractive. We introduce PACT (Pressure-Applied Compliance Testing), a benchmark for rule-following under pressure in AI agents assisting employees in daily tasks across twelve regulated enterprise domains and forty-eight scenarios, each set in a realistic multi-turn conversation. Each benchmark item pairs a standing rule against a rule-violating shortcut, and applies a battery of pressures across different wordings and system-prompt modes. We construct PACT component by component under strict LLM-as-judge auditing to ensure samples are unambiguous, ungameable, and realistic enough to avoid eliciting evaluation-aware behavior. We use PACT to profile LLM compliance across six complementary metrics that create a holistic picture of an AI assistant's robustness under pressure and throughout multi-turn conversations, its transparency, and ability to correctly discern where a rule applies. We aggregate this profile into PACTScore, a reliability-weighted compliance rate over all items and modes. Our results across 22 common LLM models spanning multiple providers and sizes show substantial variability in compliance across models and metric dimensions. Even the strongest assistants mis-apply a rule on 6 to 10% of items, and ordinary user pressure raises the violation rate by 65% on average. PACT highlights compliance risks in LLM assistants, motivating guardrails and careful model selection.
Chinese Translation
随着企业人工智能的采用不断增长，企业级大语言模型（LLM）代理正在被部署到招聘、医疗保健和金融等敏感领域。在这些领域，遵循代理系统上下文中规定的规则是首要的法律问题。目前，没有评估框架系统性地衡量哪些LLM模型倾向于违反合规规则，特别是在来自持续用户、匆忙管理者或在违反规则的情况下便利或吸引的环境下。我们引入了PACT（压力应用合规测试），这是一个针对在压力下遵循规则的基准，适用于在十二个受监管的企业领域和四十八种场景中协助员工日常任务的人工智能代理，每个场景都设定在现实的多轮对话中。每个基准项目将一个现行规则与一个违反规则的捷径进行配对，并在不同的措辞和系统提示模式下施加一系列压力。我们在严格的LLM作为评审的审计下逐步构建PACT，以确保样本明确、不易操控且足够真实，以避免引发评估意识行为。我们使用PACT来分析LLM在六个互补指标下的合规性，这些指标共同描绘了人工智能助手在压力下及多轮对话中的稳健性、透明度以及正确识别规则适用性的能力。我们将这一分析汇总为PACTScore，这是一个在所有项目和模式上的可靠性加权合规率。我们对22个来自多个供应商和不同规模的常见LLM模型的结果显示，模型和指标维度之间的合规性存在显著差异。即使是最强的助手，在6%到10%的项目上也会错误应用规则，而普通用户的压力平均提高了65%的违规率。PACT突显了LLM助手中的合规风险，促使了防护措施和谨慎的模型选择。
cs.CL / 56 / 2609.18642
STRETCH the Boundaries: A Unified Self-Taught Framework for Progressive LLM Evolution
扩展边界：一个统一的自我学习框架用于渐进式大型语言模型演化
Abstract
Large language models (LLMs) often suffer from capability stagnation in self-improvement training because fixed difficulty levels fail to adapt to their evolving proficiency. To address this issue, we propose STRETCH (Self-Taught Reasoning Evolution via Targeted CHallenge), a unified framework inspired by cognitive scaffolding theory. STRETCH introduces a dynamic Stretch Zone mechanism that continuously aligns question difficulty with the model's solving capability. Within a single parameter space, the model alternates between a Scaffolder that generates adaptive, boundary-pushing challenges and a Learner that that optimizes its solving trajectories through reinforcement learning. This dual-loop co-evolution effectively stabilizes training, mitigates reward hacking and promote progressive reasoning growth. Experiments on both negotiation and operation research benchmarks demonstrate that STRETCH consistently outperforms strong prompting and domain-specific baselines. Further scaffolder configuration analysis shows that dynamic difficulty alignment is critical for sustained capability improvement and synchronized reasoning evolution.
Chinese Translation
大型语言模型（LLMs）在自我改进训练中常常面临能力停滞的问题，因为固定的难度水平无法适应其不断发展的能力。为了解决这一问题，我们提出了STRETCH（通过针对性挑战进行自我学习推理演化），这是一个受认知支架理论启发的统一框架。STRETCH引入了一种动态的扩展区机制，持续地将问题难度与模型的解题能力对齐。在一个单一的参数空间内，模型在生成适应性、突破性挑战的支架（Scaffolder）和通过强化学习优化其解题轨迹的学习者（Learner）之间交替。这种双循环共同演化有效地稳定了训练，减轻了奖励黑客行为，并促进了渐进式推理增长。在谈判和运筹学基准测试上的实验表明，STRETCH始终优于强提示和特定领域基线。进一步的支架配置分析显示，动态难度对齐对于持续的能力提升和同步的推理演化至关重要。
cs.CL / 57 / 2609.18644
Fallacy Benchmarks Measure Scheme Recognition, Not Fallacy Detection
谬误基准测量方案识别，而非谬误检测
Abstract
Fallacy-detection benchmarks pair fallacy classes with a single "valid" or "none" class that takes everything data collection did not label as a fallacy. This construction is misleading: a classifier can learn cues that do well on this class without learning to tell a fallacy from a correct argument. We show that the low false-positive rates benchmarks report are an artifact of how the class is built, not evidence of detection ability. The most informative negative for a fallacy is a correct argument using the same argumentation scheme, and such arguments are at most a few percent of the valid class across the four benchmarks we examined. Evaluated on constructed scheme-matched negatives, false-positive rates rise from 16.6% to 58.9% on CoCoLoFa and from 5.7% to 62.0% on Reddit. That rate depends on how the negatives are written, so we also compare two conditions from the same pipeline that differ only in scheme identity. Classifiers label scheme-matched negatives as the source fallacy type 40.9 points more often than wrong-scheme negatives, which are instead identified as the scheme they actually use 85.9% of the time against 0.4% for the source type. The classifier has learned which scheme an argument uses, not whether it uses it correctly, and on the benchmarks' own test sets the two are indistinguishable. The same dissociation appears in three zero-shot LLM detectors that never saw these benchmarks, and the measurement is far lower on a negative class that was built deliberately. We release the items as Scheme Foils. A reported false-positive rate should not be trusted as a measure of detection until the valid class has been audited for scheme-matched coverage.
Chinese Translation
谬误检测基准将谬误类别与单一的“有效”或“无”类别配对，该类别包含数据收集未标记为谬误的所有内容。这种构造具有误导性：分类器可以学习在这一类别上表现良好的线索，而不学习如何区分谬误与正确论证。我们展示了基准报告的低假阳性率是该类别构建方式的伪影，而不是检测能力的证据。对于谬误而言，最具信息量的负例是使用相同论证方案的正确论证，而在我们检查的四个基准中，这类论证在有效类别中最多只占几个百分点。在构造的方案匹配负例上进行评估时，CoCoLoFa的假阳性率从16.6%上升到58.9%，而Reddit则从5.7%上升到62.0%。这一比率依赖于负例的写作方式，因此我们还比较了来自同一管道的两种条件，这两种条件仅在方案身份上有所不同。分类器将方案匹配的负例标记为源谬误类型的频率比错误方案负例高出40.9个百分点，而错误方案负例则在85.9%的情况下被识别为它们实际使用的方案，对源类型的识别率仅为0.4%。分类器学会了一个论证使用了哪个方案，而不是它是否正确使用了该方案，在基准自身的测试集中，这两者是无法区分的。在三个从未见过这些基准的零样本大语言模型（LLM）检测器中也出现了相同的分离现象，而在故意构建的负类上测量的结果则远低于此。我们将这些项目发布为方案对照。报告的假阳性率在有效类别经过方案匹配覆盖审核之前不应被信任作为检测能力的衡量标准。
cs.CL / 58 / 2609.18649
DyMT-ESB: Dynamic Multi-Turn Evaluation of Social Bias in User-LLM Interactions
DyMT-ESB：用户与大语言模型互动中的动态多轮社交偏见评估
Abstract
Warning: This paper contains examples of stereotypes and social bias. LLMs are increasingly used in interactive settings by the general public, making the evaluation of model behavior in multi-turn conversational scenarios important for safety, including stereotyping-related harms. However, existing multi-turn social bias evaluations often rely on pre-specified or template-based user inputs that do not adapt to model responses and typically assume a fixed dialogue length in advance. In this paper, we study social bias dynamics in response-conditioned multi-turn interactions using a controlled evaluation protocol that generates follow-up user queries from the evolving dialogue history and allows evaluation over variable numbers of turns. Experimental results show that LLMs exhibit social bias even in coherent, response-conditioned multi-turn interactions, revealing late-emerging bias, non-monotonic bias patterns, and bias re-emergence. These results motivate evaluations that extend beyond fixed-turn, pre-scripted protocols. Our findings highlight the importance of analyzing social bias as a turn-level dynamic phenomenon.
Chinese Translation
警告：本文包含刻板印象和社交偏见的示例。大语言模型（LLMs）在公众的互动环境中被越来越多地使用，因此在多轮对话场景中评估模型行为对于安全性至关重要，包括与刻板印象相关的危害。然而，现有的多轮社交偏见评估通常依赖于预先指定或模板化的用户输入，这些输入无法适应模型的响应，并且通常假设对话长度是固定的。本文研究了在响应条件下的多轮互动中社交偏见的动态特征，采用一种控制评估协议，从不断演变的对话历史中生成后续用户查询，并允许对可变轮次进行评估。实验结果表明，即使在连贯的、响应条件的多轮互动中，大语言模型也表现出社交偏见，揭示了后期出现的偏见、非单调的偏见模式以及偏见的再出现。这些结果促使我们进行超越固定轮次、预设脚本协议的评估。我们的研究结果强调了将社交偏见视为轮次级动态现象进行分析的重要性。
cs.CL / 59 / 2609.18672
Selection Is Retrieval, Abstention Is Not: On-Device Tool Routing over 70 Korean-English Actions
选择是检索，弃权不是：基于设备的70个韩英动作的工具路由
Abstract
An AI assistant that calls tools makes two decisions on every request: which tool to invoke, and whether any available tool applies. In the usual design a single language model makes both, by emitting a call or by declining to emit one. On a device that has to answer without a server, the language model is what makes that design expensive, dominating both the latency and the memory of the router. The common alternative is to remove the model completely and rank the catalog of local actions with a retriever instead. That substitution is not symmetric across the two decisions. A retriever returns its highest-scoring candidate for every input and cannot signal that the catalog holds no valid action. Our earlier study found that constraining a decoder to a tool grammar repairs malformed output without improving the choice. What the substitution costs in each decision has not been measured. We evaluate the two decisions separately over 600 Korean and English requests and a catalog of 70 local actions. The router may also ask for a missing slot, reply, or delegate. Half the in-catalog requests reuse catalog vocabulary and half paraphrase it, separating lexical overlap from the action requested. Character 3-gram BM25 selects 162 of 164 lexically matched requests and 85 of 166 paraphrases. Restricting the candidate set to seven raises the paraphrase figure to a mean of 0.825 over five trials. No classifier over its score features separates in-catalog from out-of-catalog above 0.697 area under the curve, where the frozen encoder multilingual-e5-base reaches 0.806. Using that encoder for abstention alone keeps 376 of the requests local and misroutes 9 of the 150 needing delegation. Abstention, not selection, is where a neural component is required. A neural ranker improves every quality metric and is rejected on latency and memory rather than accuracy.
Chinese Translation
一个调用工具的人工智能助手在每个请求上做出两个决策：调用哪个工具，以及是否有可用工具适用。在通常的设计中，单一的语言模型同时做出这两个决策，通过发出调用或拒绝发出调用。在一个必须在没有服务器的情况下作出响应的设备上，语言模型使得这种设计成本高昂，主导了路由器的延迟和内存。常见的替代方案是完全去除模型，而用检索器对本地动作的目录进行排序。然而，这种替代在两个决策上并不对称。检索器为每个输入返回其得分最高的候选项，无法指示目录中没有有效的动作。我们之前的研究发现，将解码器限制在工具语法内可以修复格式不正确的输出，而不改善选择。每个决策的替代成本尚未被测量。我们对600个韩语和英语请求及70个本地动作的目录分别评估这两个决策。路由器还可以请求缺失的槽、回复或委托。目录中的一半请求重用目录词汇，另一半则对其进行释义，从而将词汇重叠与请求的动作分开。字符3-gram BM25选择了164个词汇匹配请求中的162个和166个释义中的85个。将候选集限制为七个将释义的平均值提高到五次试验中的0.825。没有任何分类器在其得分特征上能将目录内与目录外区分开，曲线下面积超过0.697，而冻结的编码器 multilingual-e5-base 达到0.806。仅使用该编码器进行弃权使376个请求保持本地，并错误路由150个需要委托的请求中的9个。弃权，而非选择，是需要神经组件的地方。神经排序器改善了每个质量指标，但在延迟和内存方面被拒绝，而不是在准确性上。
cs.CL / 60 / 2609.18677
Voice of Reason: Reinforcement Learning for Spoken Math
理性的声音：用于口语数学的强化学习
Abstract
Speech language models enable richer spoken interactions between humans and machines than cascaded systems, allowing access to paralinguistic information and lower latency. However, their accuracy on mathematical reasoning benchmarks has lagged behind those of text models. Reinforcement learning (RL) with verifiable rewards has been instrumental in extending text models' capabilities for solving complex problems and limiting hallucinations. In this work, we explore applying RL to the GLM-4-Voice speech model (Zeng et al., 2024) to bridge the gap between textual and spoken mathematical problem solving. We first adapt the model to the domain using supervised fine-tuning on synthesized spoken question-answering data. We then show that, even without extra reasoning tokens, RL improves the accuracy on GSM8K beyond levels previously achieved for speech models only with supplementary reasoning traces. When combined with existing streaming reasoning techniques, we show further gains to 74.8% free-form accuracy. This establishes a new state-of-the-art for mathematical spoken abilities with speech-native models.
Chinese Translation
语音语言模型比级联系统能够实现更丰富的人机口语交互，能够访问副语言信息并降低延迟。然而，它们在数学推理基准测试上的准确性仍落后于文本模型。具有可验证奖励的强化学习（RL）在扩展文本模型解决复杂问题的能力和限制幻觉方面发挥了重要作用。在本研究中，我们探讨将RL应用于GLM-4-Voice语音模型（Zeng et al., 2024），以弥合文本与口语数学问题解决之间的差距。我们首先通过对合成口语问答数据进行监督微调，将模型适应于该领域。然后我们展示，即使没有额外的推理标记，RL也能将GSM8K的准确性提升到超出仅依赖补充推理痕迹的语音模型之前所达到的水平。当与现有的流式推理技术结合时，我们展示了进一步提升至74.8%的自由形式准确性。这为语音本地模型的数学口语能力确立了新的最先进水平。
cs.CL / 61 / 2609.18680
HearInContext: A Benchmark for Implicit Context in Speech Recognition
HearInContext：语音识别中隐式上下文的基准测试
Abstract
Contextual ASR can benefit from semantic cues or from target words explicitly provided in the context. We introduce HearInContext, a Mandarin--English benchmark that pairs shared synthetic speech with assistant replies supporting different interpretations. The benchmark comprises 3,764 semantic test cases built around homophones. Implicit contexts exclude candidate words; explicit contexts name the target. No-context and unrelated-context controls measure the benefit of relevant history and sensitivity to irrelevant history. Context-capable models benefit from implicit cues but achieve higher target recall with explicit hints. Fine-tuning Qwen3-ASR-1.7B improves implicit-context target recall by 11.0 and 11.5 percentage points in Mandarin and English, respectively, while absolute CER/WER changes on AISHELL-1 and LibriSpeech remain below 0.1 percentage points. Gains extend to explicit conditions excluded from fine-tuning and to Mandarin hotword recognition on real recordings.
Chinese Translation
上下文语音识别（ASR）可以从语义线索或上下文中明确提供的目标词中受益。我们介绍了HearInContext，这是一个中英双语基准，配对共享的合成语音与支持不同解释的助手回复。该基准包含3,764个围绕同音词构建的语义测试案例。隐式上下文排除了候选词；显式上下文则明确命名目标。无上下文和无关上下文的对照组用于测量相关历史的益处及对无关历史的敏感性。具备上下文能力的模型可以从隐式线索中受益，但在显式提示下实现更高的目标召回率。对Qwen3-ASR-1.7B进行微调后，隐式上下文的目标召回率在普通话和英语中分别提高了11.0和11.5个百分点，而在AISHELL-1和LibriSpeech上的绝对字符错误率（CER）/词错误率（WER）变化仍低于0.1个百分点。收益扩展到未包含在微调中的显式条件以及真实录音中的普通话热词识别。
cs.CL / 62 / 2609.18697
Tracing individual knowledge trajectories in a changing field: the case of general relativity and gravitation
在变化领域中追踪个体知识轨迹：以广义相对论和引力为例
Abstract
Historians have reconstructed the twentieth-century transformation of general relativity and gravitation (GRG) at the field level and through individual careers, but connecting these scales requires a way to compare researchers with the changing field over time. We develop such a comparison, setting a researcher's publications and references against GRG field literature from the same, earlier, and later two-year periods. Building on Own Vocabulary and Embedding Density Estimation from our earlier two-case study (arXiv:2501.00391), we extend the analysis to the fifty most-published authors in a NASA/ADS corpus of about 180,000 GRG records (1911 to 2000) and add two citation-based measures, Referenced Vocabulary and Citation Identity. The four measures compare an author's written language, cited literature, semantic neighbourhood, and cited-authority configuration with the surrounding field. The earlier cases suggested that closer field-vocabulary alignment accompanies a denser semantic neighbourhood. Across the fifty authors this holds only partially. Written and cited vocabularies tend to move together, usually resembling later GRG literature as the field turned towards astrophysical and cosmological research. Semantic neighbourhoods more often lie where the field's publications were concentrated in earlier periods, while co-citation patterns follow no single temporal direction, and the two citation measures frequently place the same researcher differently despite drawing on identical reference lists. Individual trajectories can thus combine vocabulary tied to later field states with older semantic or citation structures, and these divergent cases mark patterns for closer historical investigation. The approach transfers to other fields with defensible corpus boundaries and adequate coverage of texts, references, and disambiguated author identities.
Chinese Translation
历史学家们在领域层面和个体职业生涯中重建了二十世纪广义相对论和引力（GRG）的转型，但将这些尺度联系起来需要一种方法来比较研究者与随时间变化的领域。我们开发了这种比较方法，将研究者的出版物和参考文献与GRG领域文献进行对比，涵盖相同、早期和后期的两年时间段。在我们早期的两个案例研究（arXiv:2501.00391）的基础上，我们扩展分析至NASA/ADS语料库中约180,000条GRG记录（1911年至2000年）中出版量最多的五十位作者，并增加了两个基于引用的度量，引用词汇（Referenced Vocabulary）和引用身份（Citation Identity）。这四个度量比较了作者的书面语言、引用文献、语义邻域和引用权威配置与周围领域的关系。早期案例表明，领域词汇的更紧密对齐伴随着更密集的语义邻域。在这五十位作者中，这种情况仅部分成立。书面和引用词汇往往一起变化，通常与后期GRG文献相似，因为该领域逐渐转向天体物理和宇宙学研究。语义邻域更常位于该领域早期出版物集中之处，而共同引用模式没有遵循单一的时间方向，这两个引用度量常常使同一研究者的定位不同，尽管引用列表相同。因此，个体轨迹可以将与后期领域状态相关的词汇与较旧的语义或引用结构结合起来，这些不同的案例为更深入的历史研究标记了模式。这种方法可以转移到其他具有合理语料边界和充分文本、参考文献及消歧义作者身份覆盖的领域。
cs.CL / 63 / 2609.18720
LocQE: Principled Domain Adaptation for Localisation Quality Estimation by Leveraging Post-Edits
LocQE：通过利用后编辑实现定位质量估计的原则性领域适应
Abstract
Learned quality estimation (QE) models such as COMETKiwi are widespread and work well for general machine translation evaluation. However, they are known to struggle on unseen domains, limiting their performance in a real-world localisation context. We show that they are insensitive to some important factors in localisation, such as whether numbers are translated accurately, or even whether the correct number of spaces and punctuation are preserved in a translation. Further, a key capability for optimisation of machine translation is the ability of QE models to accurately rank different translations of a single segment, which suffers significantly from the domain transfer. In the absence of large-scale direct assessment data, we propose principled fine-tuning approaches to reduce the domain gap with even small amounts of post-editing data. Using a multi-task fine-tuning approach and a simple tokeniser intervention, we create a QE model which proves markedly better at distinguishing preferred post-edits from rejected initial translations in a localisation context. We show that preferences and artificial continuous scores stabilise each other, and argue that to calibrate metrics both in terms of their absolute scores and comparisons between translation of the same source, both types of signal are needed.
Chinese Translation
学习型质量估计（QE）模型，如 COMETKiwi，已经广泛应用，并且在一般机器翻译评估中表现良好。然而，它们在未见领域上表现不佳，这限制了它们在实际本地化环境中的性能。我们展示了这些模型对本地化中的一些重要因素不够敏感，例如数字是否被准确翻译，或者翻译中是否保留了正确数量的空格和标点符号。此外，优化机器翻译的一个关键能力是 QE 模型能够准确地对同一段落的不同翻译进行排名，而这一能力在领域转移中受到显著影响。在缺乏大规模直接评估数据的情况下，我们提出了原则性的微调方法，以减少领域差距，即使在少量后编辑数据的情况下。通过使用多任务微调方法和简单的标记器干预，我们创建了一个 QE 模型，该模型在本地化环境中明显更擅长区分首选后编辑和被拒绝的初始翻译。我们展示了偏好和人工连续分数相互稳定，并认为为了在绝对分数和同一源翻译之间的比较中校准指标，这两种信号都是必要的。
cs.CL / 64 / 2609.18739
A Scalable Framework for Automated NER Annotation Correction in Low-Resource Languages
用于低资源语言的自动命名实体识别（NER）注释修正的可扩展框架
Abstract
Poor quality or noisy annotations in Named Entity Recognition (NER), as in any other NLP task, make it challenging to achieve state-of-the-art performance. In this paper, we present a multi-step framework to enhance the annotation quality of NER datasets by employing automated techniques. We propose a frequency-based iterative approach that leverages self-training and a dual-threshold mechanism to enhance inference confidence. Experimental evaluations on different NER datasets demonstrate significant improvements in NER performance with respect to the original datasets. This work further explores the potential of generative Large Language Models (LLMs) to perform NER for low-resource languages.
Chinese Translation
在命名实体识别（NER）中，低质量或噪声注释与其他自然语言处理（NLP）任务一样，使得实现最先进的性能变得具有挑战性。本文提出了一种多步骤框架，通过采用自动化技术来提高NER数据集的注释质量。我们提出了一种基于频率的迭代方法，利用自我训练和双阈值机制来增强推理信心。在不同的NER数据集上的实验评估显示，与原始数据集相比，NER性能有显著提升。本研究进一步探讨了生成性大型语言模型（LLMs）在低资源语言中执行NER的潜力。
cs.CL / 65 / 2609.18772
Zero-Shot Cross-Lingual Recognition of Sign Language Handshapes
零样本跨语言手势语言手形识别
Abstract
Sign language processing advances rapidly for high-resource languages such as American Sign Language (ASL), yet most of the world's sign languages lack the phonological annotations new methods require. We present the first zero-shot cross-lingual framework for handshape recognition, transferring from ASL to Catalan Sign Language (LSC). Our approach leverages the decomposition of handshapes into five phonological features -- selected fingers, flexion, spread, thumb position, and thumb contact -- shared across both languages, to decode LSC handshapes from predicted features via a composite phonological distance metric. We evaluate three architectures (MLP, SL-GCN, SHuBERT) trained on two ASL corpora (PopSign, Sem-Lex) against a 37-handshape, single-signer LSC benchmark. Zero-shot transfer proves viable once recording-format disparities are harmonized, reaching 80.0% phonological feature accuracy and 54.5% expected handshape accuracy. Phonological decomposition thus offers a bridge for extending sign language technologies to low-resource languages without any target-language video training labels.
Chinese Translation
手势语言处理在高资源语言（如美国手语（ASL））方面迅速发展，然而世界上大多数手势语言缺乏新方法所需的音韵标注。我们提出了首个零样本跨语言手形识别框架，旨在从美国手语（ASL）转移到加泰罗尼亚手语（LSC）。我们的方法利用手形的五个音韵特征进行分解——选择的手指、弯曲度、展开度、拇指位置和拇指接触——这些特征在两种语言中共享，通过复合音韵距离度量从预测特征解码LSC手形。我们评估了三种架构（MLP、SL-GCN、SHuBERT），并在两个ASL语料库（PopSign、Sem-Lex）上进行训练，针对一个包含37个手形的单签名者LSC基准进行测试。一旦录制格式差异得到协调，零样本转移证明是可行的，达到了80.0%的音韵特征准确率和54.5%的预期手形准确率。因此，音韵分解为将手势语言技术扩展到没有任何目标语言视频训练标签的低资源语言提供了桥梁。
cs.CL / 66 / 2609.18804
Beyond frequency measures: Can contextual embeddings capture meaning change in scientific texts?
超越频率度量：上下文嵌入能否捕捉科学文本中的意义变化？
Abstract
Identifying technological trends is a core scientometric task, yet traditional frequency-based approaches struggle to capture substantial meaning shifts of domain-specific terms. We hypothesise that contextual embeddings can complement frequency dynamics to effectively track diachronic semantic change. We compare frequency and embedding-based approaches across Astrophysics and NLP corpora spanning from 2010 to 2024. Candidate terms are extracted using KeyBERT (utilizing SciBERT as its underlying language model) and filtered for significant frequency increases using Fisher's exact test. These terms are then evaluated for genuine semantic shift by domain experts to establish ground-truth labels. To quantify semantic drift, each term's contextual embedding ''clouds'' from the two discrete periods are compared using multiple metrics: cosine distance, average pairwise distance, Hotelling-type T 2 , and maximum mean discrepancy. Results indicate that frequency-based methods align slightly better with human judgments of ''trend-related terms'' than semantic metrics (Precision@50 of 0.62 vs 0.60 in Astrophysics). The two signals show a correlation of around 0.6. Several terms identified exclusively by embedding metrics (e.g., ''primordial black holes'') represent critical conceptual developments invisible to pure frequency analysis. These findings indicate that semantic metrics may capture complementary information, highlighting the value of integrating contextual embeddings into scientometric trend analysis.
Chinese Translation
识别技术趋势是科学计量学的核心任务，然而传统的基于频率的方法在捕捉领域特定术语的实质性意义变化方面存在困难。我们假设上下文嵌入可以补充频率动态，有效追踪历时语义变化。我们比较了2010年至2024年间的天体物理学和自然语言处理（NLP）语料库中的频率和嵌入方法。候选术语通过KeyBERT（使用SciBERT作为其基础语言模型）提取，并通过Fisher精确检验筛选出显著频率增加的术语。这些术语随后由领域专家评估其真实的语义变化，以建立真实标签。为了量化语义漂移，我们使用多个指标比较两个离散时期每个术语的上下文嵌入“云”：余弦距离、平均成对距离、Hotelling类型的T²和最大均值差异。结果表明，基于频率的方法在“趋势相关术语”的人类判断上略微优于语义指标（在天体物理学中的Precision@50为0.62对比0.60）。这两种信号的相关性约为0.6。通过嵌入指标独立识别的几个术语（例如，“原始黑洞”）代表了纯频率分析无法察觉的重要概念发展。这些发现表明，语义指标可能捕捉到互补信息，突显了将上下文嵌入整合到科学计量趋势分析中的价值。
cs.CL / 67 / 2609.18844
ReFigBench: Benchmarking Scientific Figure Reconstruction as Editable PowerPoint Artifacts
ReFigBench：科学图形重建作为可编辑PowerPoint文档的基准测试
Abstract
Multimodal coding agents are expected to turn visual inputs into usable artifacts, and they act through a harness, the layer of tools, context management, and execution environment around the model. Existing evaluations often isolate short tool calls, API traces, or screenshot resemblance, and a low score under these proxies cannot say whether the model saw poorly, planned poorly, or was failed by its harness. We study scientific overview figure reconstruction, an agent task in which a source image must become an editable PowerPoint slide that preserves text, topology, layout, and native document structure. We introduce ReFigBench, a benchmark and evaluation framework built on 1,000 real overview figures retrieved from arXiv papers with full provenance. Coding agents from four model families reconstruct every figure under two workflows, direct code generation and a specialized PPTX workflow, and the strongest model runs inside two commercial harnesses, yielding ten configurations. Evaluation combines deterministic artifact checks, repeated automated scoring by judges from two model families, and blinded human comparisons. Perception remains a bottleneck that iterative rendering only partly repays. Whether workflow effort converts into quality depends on the model together with its harness, since the same model gains from the specialized workflow inside one harness and loses inside the other, and the harness shifts scores even under an identical direct prompt. The specialized workflow erases native connectors in every configuration, human judges still prefer its renderings in most matchups, and even the strongest agent falls short of the rubric ceiling. These results expose the tension between fidelity and editability as the central challenge for practical multimodal document agents.
Chinese Translation
多模态编码代理预计将视觉输入转化为可用的文档，它们通过一个工具层、上下文管理和模型的执行环境进行操作。现有评估往往孤立短期工具调用、API痕迹或屏幕截图相似性，而在这些代理下得分低并不能说明模型是观察不佳、规划不佳，还是被其工具层所限制。我们研究科学概述图形重建，这是一个代理任务，其中源图像必须变成一个可编辑的PowerPoint幻灯片，以保留文本、拓扑、布局和原始文档结构。我们引入了ReFigBench，这是一个基于从arXiv论文中检索的1000个真实概述图形构建的基准和评估框架，具有完整的来源信息。来自四个模型家族的编码代理在两种工作流程下重建每个图形，分别是直接代码生成和专门的PPTX工作流程，最强的模型在两个商业工具层中运行，产生十种配置。评估结合了确定性的文档检查、来自两个模型家族的评审者的重复自动评分和盲人比较。感知仍然是一个瓶颈，迭代渲染仅部分弥补。工作流程的努力是否转化为质量依赖于模型及其工具层，因为同一模型在一个工具层内从专门工作流程中受益，而在另一个工具层内则失去优势，并且即使在相同的直接提示下，工具层也会改变得分。专门的工作流程在每种配置中都消除了原生连接器，但人类评审者在大多数对比中仍然偏好其渲染，甚至最强的代理也未能达到评分标准的上限。这些结果揭示了忠实度与可编辑性之间的紧张关系，这是实际多模态文档代理面临的核心挑战。
cs.CL / 68 / 2609.18852
EviGen: Predictive Evidence Scaffolding for Verifiable Clinical Rationale Generation
EviGen：可验证临床推理生成的预测证据支架
Abstract
Longitudinal electronic health records (EHRs) capture years of patient history across notes, codes, labs, and procedures, and contain evidence needed to reason about likely clinical outcomes. However, comprehensive clinician review of these records is impractical, and LLM-based processing is costly and often unreliable, missing some relevant observations while hallucinating others. We therefore propose EviGen, a three-layer framework for verifiable clinical rationale generation that addresses these challenges. The first layer is a patient-conditioned retriever that uses learnable queries to find evidence predictive of, not just textually relevant to, a clinical outcome and ranks it by prediction attribution scores. The second layer is an LLM generator that consumes this ranked evidence as a scaffold to produce a clinical rationale grounded in the retrieved spans. The third layer is a process-supervised verifier that checks the generated rationale at the reasoning-step level, flagging unreliable claims. Across three medical prediction datasets, EviGen improves prediction performance and rationale faithfulness over full-context LLM and RAG baselines, and is preferred by clinical reviewers in a usability evaluation.
Chinese Translation
纵向电子健康记录（EHRs）捕捉了患者多年的历史，包括笔记、编码、实验室结果和程序，并包含推理可能临床结果所需的证据。然而，全面的临床医生审查这些记录是不切实际的，而基于大型语言模型（LLM）的处理成本高昂且常常不可靠，可能遗漏一些相关观察结果，同时又会虚构其他结果。因此，我们提出了EviGen，一个三层框架，用于可验证的临床推理生成，以应对这些挑战。第一层是一个患者条件检索器，它使用可学习的查询来寻找不仅在文本上相关而且对临床结果具有预测性的证据，并根据预测归因分数对其进行排名。第二层是一个LLM生成器，它将这些排名的证据作为支架，生成基于检索到的片段的临床推理。第三层是一个过程监督的验证器，它在推理步骤级别检查生成的推理，标记不可靠的主张。在三个医学预测数据集上，EviGen在预测性能和推理可信度方面优于全上下文LLM和RAG基线，并在可用性评估中受到临床审阅者的青睐。
cs.CL / 69 / 2609.18861
PersonaPath: Towards Knowledge-Centric Personalized Learning Path Planning
PersonaPath：面向知识中心的个性化学习路径规划
Abstract
Adaptive learning systems commonly formulate learning path planning as Exercise-Centric (EC) recommendation, where the next step is inferred from item-level interaction logs. Evaluating goal-oriented guidance additionally requires explicit learner goals and curriculum-scale prerequisites: learners with similar exercise records may need different paths toward their targets. We therefore study Knowledge-Centric (KC) personalized learning path planning, where a planner must reason over learner profiles, mastery states, and prerequisite knowledge structures to decide which textbook, unit, and concept should be studied next. To support this setting, we introduce PersonaPath, a benchmark that pairs 2,000 fine-grained learner personas with a hierarchical knowledge graph of 347 textbooks, 1,751 units, and 4,092 concepts across 77 subjects. We evaluate representative LLMs on PersonaPath. Results show that even the strongest LLM reaches only a 29.5% final pass rate in Basic Education, and that the main bottleneck lies in adaptivity, where no model exceeds 44.7% in tailoring paths to individual learners.
Chinese Translation
自适应学习系统通常将学习路径规划表述为以练习为中心（Exercise-Centric, EC）的推荐，其中下一步是根据项目级交互日志推断的。评估以目标为导向的指导还需要明确的学习者目标和课程规模的先决条件：具有相似练习记录的学习者可能需要不同的路径来实现他们的目标。因此，我们研究知识中心（Knowledge-Centric, KC）的个性化学习路径规划，其中规划者必须对学习者的个人档案、掌握状态和先决知识结构进行推理，以决定下一个应该学习的教科书、单元和概念。为了支持这一设置，我们引入了PersonaPath，一个基准数据集，将2000个细粒度学习者角色与一个包含347本教科书、1751个单元和4092个概念的层次知识图谱相结合，涵盖77个学科。我们在PersonaPath上评估了代表性的语言模型（LLMs）。结果显示，即使是最强的语言模型在基础教育中的最终通过率也仅为29.5%，而主要瓶颈在于适应性，没有模型在为个别学习者量身定制路径方面超过44.7%。
cs.CL / 70 / 2609.18905
Structured Claim-Level Discourse Representations for Dense Health Narratives
密集健康叙事的结构化声明级话语表示
Abstract
Health discourse in social media videos often contains densely entangled claims spanning multiple thematic aspects, stances, evidential frames, and rhetorical functions within short conversational spans. Existing approaches largely rely on coarse topic-level, sentiment-based, or stance-oriented representations that do not adequately capture this structure. Our analysis identifies an average of 13.22 atomic claims per minute, motivating richer claim-level discourse representations. We introduce a structured framework for claim-level discourse analysis in dense health narratives. Our framework models discourse through tuples linking atomic claims with thematic aspects, stance, and multidimensional pragmatic discourse attributes. To support this setting, we construct a benchmark spanning four health domains with 1,191 manually annotated claims from 60 videos. Using this framework, we evaluate automated structured discourse analysis under different discourse context settings. Results show that current LLMs achieve strong performance on thematic categorization and stance prediction, but struggle with high-dimensional pragmatic profiling. We also find that different discourse tasks benefit from different forms of contextual reasoning, suggesting that future systems may require task decomposition and specialized inference strategies.
Chinese Translation
社交媒体视频中的健康话语通常包含多个主题方面、立场、证据框架和修辞功能交织在一起的密集声明，这些声明在短暂的对话中交替出现。现有的方法主要依赖于粗略的主题级、基于情感或立场导向的表示，无法充分捕捉这种结构。我们的分析发现每分钟平均有13.22个原子声明，这促使我们提出更丰富的声明级话语表示。我们引入了一个结构化框架，用于密集健康叙事中的声明级话语分析。我们的框架通过元组将原子声明与主题方面、立场和多维实用话语属性联系起来，从而对话语进行建模。为了支持这一设置，我们构建了一个涵盖四个健康领域的基准数据集，其中包含来自60个视频的1,191个手动标注的声明。利用该框架，我们在不同的话语上下文设置下评估自动化结构化话语分析。结果表明，当前的大型语言模型（LLMs）在主题分类和立场预测方面表现出色，但在高维实用特征分析方面存在困难。我们还发现，不同的话语任务受益于不同形式的上下文推理，这表明未来的系统可能需要任务分解和专业推理策略。
cs.CL / 71 / 2609.18908
How Much is a Human Right Worth? ECtHR-NPD: A Benchmark for Predicting Non-Pecuniary Damage Awards
人权的价值是多少？ECtHR-NPD：预测非经济损害赔偿的基准
Abstract
Existing legal benchmarks cover diverse tasks, while continuous monetary remedies remain comparatively underexplored. We introduce ECtHR-NPD, to the best of our knowledge, the first benchmark for predicting non-pecuniary damage (NPD) awards at the European Court of Human Rights (ECtHR) from case information when no statutory formula or explicit calculation rule determines the amount. ECtHR-NPD contains 14,575 cases with case-level awards in nominal euros, chronological splits, and a protocol separating target construction from model input. We evaluate a battery of methods, including constant predictors, gradient-boosted trees, retrieval methods, fine-tuned encoder language models (LMs), prompted decoder LMs, and knowledge-augmented agents. Our results show that more sophisticated LM and agentic approaches do not consistently outperform the strongest feature-based baseline. All model families struggle to identify zero awards and to calibrate high-award predictions, with further degradation on the Challenging test view, making ECtHR-NPD a challenging testbed for current state-of-the-art open-weight and proprietary LMs.
Chinese Translation
现有的法律基准涵盖了多种任务，而持续的货币救济相对较少被探索。我们引入了ECtHR-NPD，据我们所知，这是第一个用于预测欧洲人权法院（ECtHR）非经济损害（NPD）赔偿的基准，基于案件信息，当没有法定公式或明确的计算规则来确定金额时。ECtHR-NPD包含14,575个案件，提供以名义欧元计的案件级赔偿、时间划分，以及将目标构建与模型输入分开的协议。我们评估了一系列方法，包括常量预测器、梯度提升树、检索方法、微调的编码器语言模型（LMs）、提示解码器 LMs 和知识增强代理。我们的结果表明，更复杂的LM和代理方法并未始终优于最强的基于特征的基线。所有模型家族在识别零赔偿和校准高赔偿预测方面都面临困难，在具有挑战性的测试视图上进一步恶化，使得ECtHR-NPD成为当前最先进的开放权重和专有LM的一个具有挑战性的测试平台。
cs.CL / 72 / 2609.18909
Beyond Outcomes: Dual-View Relational Learning for Efficient Agent Benchmarking
超越结果：高效代理基准测试的双视角关系学习
Abstract
Agent benchmarks are substantially more costly to evaluate than conventional LLM benchmarks. Benchmark compression is therefore a natural solution, yet existing methods primarily model redundancy in task--model final-score distributions, which is important in agentic evaluation. To address this limitation, we analyze large-scale trajectories and identify six complementary process signals that are systematically associated with final agent performance. To disentangle agent performance redundancy from a complete perspective, we propose DualViewEval, an agent benchmark compression method that jointly exploits outcome and process relations to learn an exact-size miniset and predict the full-benchmark scores. Across five agent benchmarks and five representative baselines, DualViewEval achieves the best results in all datasets. With only 20 tasks, it achieves $24\times$--$40\times$ compression on APEX-Agents and BFCL, reducing mean absolute error (MAE) by $14.5\%$--$28.2\%$ over the strongest competitors while improving Kendall's $\tau$ by up to $7.2\%$ relative to EssenceBench on SWE-bench Verified. The selected minisets further reveal capability differences among different agents, providing compact and diagnostic feedback for efficient agentic model development.
Chinese Translation
代理基准测试的评估成本远高于传统的大型语言模型（LLM）基准测试。因此，基准压缩成为一种自然的解决方案，但现有方法主要建模任务-模型最终得分分布中的冗余，这在代理评估中至关重要。为了解决这一局限性，我们分析了大规模轨迹，并识别出六个与最终代理性能系统性相关的互补过程信号。为了从整体视角解开代理性能冗余，我们提出了DualViewEval，这是一种代理基准压缩方法，联合利用结果和过程关系来学习精确大小的最小集并预测完整基准得分。在五个代理基准和五个代表性基线中，DualViewEval在所有数据集中都取得了最佳结果。仅用20个任务，它在APEX-Agents和BFCL上实现了$24 imes$--$40 imes$的压缩，较最强竞争对手减少了$14.5 ext{%}$--$28.2 ext{%}$的平均绝对误差（MAE），同时相较于EssenceBench在SWE-bench Verified上提高了Kendall's $ au$多达$7.2 ext{%}$。所选的最小集进一步揭示了不同代理之间的能力差异，为高效的代理模型开发提供了紧凑且具有诊断性的反馈。
cs.CL / 73 / 2609.18935
Long-Lived Characters, Local Inference: Incremental Memory Maintenance for Game NPCs
长寿命角色与局部推理：游戏NPC的增量记忆维护
Abstract
A game character should not have to reread its entire life before every conversation. For locally deployed language-model characters, however, revising a few memories can invalidate a long reusable prefix. The resulting preparation cost competes with both foreground dialogue and the maintenance of other characters. This matters especially when dialogue feeds game-defined actions and value judgments: a fluent but incorrect account of who owns an item, or whether a transfer has already happened, can corrupt the input to otherwise deterministic rules. We study incremental memory maintenance for long-lived game NPCs in a quantized Qwen hybrid recurrent-attention model. Our runtime removes superseded attention KV entries, computes replacement records at the true sequence tail, and preserves the continuing recurrent state and unchanged KV. Existing local experiments combine multi-update dialogue replays, fixed-input placement ablations, and attention diagnostics. Independent block composition weakens query-conditioned memory selection without a uniform chunk-initial attention collapse. True-tail updates preserve important current-state and historical bindings across eight scripted maintenance rounds; a placement case recovers the full-refill quantity in three reconstructions, while slot-preserving alternatives repeat a double-subtraction error. Attention-distribution proximity alone does not explain these semantic differences. The results motivate treating a character's inference state as a maintained, history-dependent resource, rather than only a disposable encoding of its latest memory text.
Chinese Translation
游戏角色在每次对话前不应重新回顾其整个生活。然而，对于局部部署的语言模型角色，修订少量记忆可能会使一个长期可重用的前缀失效。由此产生的准备成本与前台对话及其他角色的维护相竞争。这一点尤其重要，因为对话影响游戏定义的动作和价值判断：流畅但不正确的物品所有权描述，或转移是否已发生的错误陈述，可能会破坏输入到其他确定性规则的有效性。我们研究了在量化的Qwen混合递归注意力模型中，长寿命游戏NPC的增量记忆维护。我们的运行时移除过时的注意力KV条目，在真实序列尾部计算替代记录，并保留持续的递归状态和未改变的KV。现有的局部实验结合了多次更新的对话重放、固定输入放置消融和注意力诊断。独立的块组合削弱了查询条件下的记忆选择，而没有出现统一的块初始注意力崩溃。真实尾部更新在八轮脚本维护中保留了重要的当前状态和历史绑定；一个放置案例在三次重建中恢复了完整的补充数量，而保持槽的替代方案则重复了双重减法错误。仅靠注意力分布的接近性并不能解释这些语义差异。结果促使我们将角色的推理状态视为一种维护的、依赖历史的资源，而不仅仅是其最新记忆文本的可丢弃编码。
cs.CL / 74 / 2609.18959
LangSelect: Cost-Aware Target-Language Routing for LLM Code Generation
LangSelect：面向成本的目标语言路由用于大语言模型代码生成
Abstract
LLM code-generation systems usually choose a target programming language before decoding and treat that choice as fixed. We show that, for language-flexible programming tasks -- tasks where several target languages are acceptable and checkable by the same tests -- this choice is a measurable cost lever: verified implementations of the same task can differ substantially in generated-token length. We introduce LangSelect, a verification-aware router that selects the target language before generation and falls back when the first attempt fails. To separate offline routing opportunity from end-to-end behavior, we evaluate verified-solution replay, which chooses among already accepted corpus solutions, and live GPT-5 generation, which charges every generation attempt, including failures and fallbacks. On MultiLang-Bench, a 3,000-task, 8-language verified corpus, replay shows substantial language-routing headroom. In live evaluation on 450 held-out tasks, a train-split Domain heuristic baseline reduces harness-proxy tokens, which include wrapper and entrypoint overhead, by 50.3\% at 92.9\% pass after fallback, while a learned CodeBERT+metadata selector reaches the highest pass after fallback, 93.8\%, with a 3.7\% token increase. These results show that output-language routing can define a practical cost-correctness frontier for unit-test-verifiable code generation.
Chinese Translation
大语言模型（LLM）代码生成系统通常在解码之前选择目标编程语言，并将该选择视为固定。我们表明，对于语言灵活的编程任务——即多个目标语言均可接受且可通过相同测试进行验证的任务——这一选择是一个可测量的成本杠杆：相同任务的经过验证的实现生成的令牌长度可能存在显著差异。我们引入了LangSelect，这是一种关注验证的路由器，在生成之前选择目标语言，并在第一次尝试失败时进行回退。为了将离线路由机会与端到端行为分开，我们评估了经过验证的解决方案重放，该方法在已接受的语料库解决方案中进行选择，以及实时的GPT-5生成，该方法对每次生成尝试（包括失败和回退）进行计费。在MultiLang-Bench上，这是一个包含3000个任务和8种语言的经过验证的语料库，重放显示出显著的语言路由潜力。在对450个保留任务的实时评估中，训练分割的领域启发式基线在回退后将包含包装器和入口点开销的代理令牌减少了50.3%，通过率为92.9%；而学习的CodeBERT+元数据选择器在回退后达到了最高的通过率93.8%，令牌增加了3.7%。这些结果表明，输出语言路由可以为单元测试可验证的代码生成定义一个实际的成本-正确性边界。
cs.CL / 75 / 2609.18960
When Audit Quality Fails to Predict Downstream Utility: A Counterfactual Study of Synthetic-Data Selectors for Low-Resource African NLP
当审计质量无法预测下游效用：低资源非洲自然语言处理的合成数据选择者的反事实研究
Abstract
Quality-aware synthetic-data selection rests on a proxy: examples that an LLM judge rates as good should also help a downstream model learn. In a controlled replay in low-resource African-language classification, we show that this proxy breaks. Across four languages (Amharic, Hausa, Swahili, Yoruba), two classification tasks (MasakhaNEWS, AfriSenti), and five matched-budget selectors, audit rankings and downstream rankings diverge. Within each cell, the Spearman between judged label correctness and Macro-F1 across selectors has mean $\rho{=}0.04$ (median $0.00$), showing that the mismatch is not an aggregation artifact. \method{}-V2, our counterfactual audit framework, produces the cleanest selected pool on three audit channels at once: highest judged label correctness ($0.904$ vs.\ $0.767$ for naive, a $17.9\%$ relative gain), lowest shortcut score, and a hard-reject rate of $0.162$ vs.\ $0.486$ for naive. AlpaGasus nevertheless leads downstream Macro-F1 ($0.202$ vs.\ $0.163$ for \method{}-V2), and the inversion persists on the five non-degenerate cells. The lesson is methodological: in this controlled setting, audit quality is a property of the selected pool, not a guarantee of downstream utility. Synthetic-data evaluation should therefore report audit and downstream metrics on the same retained sets. We release the audit tables, per-selector retained pools, and a claim ledger that links every reported number to its source row.
Chinese Translation
质量感知的合成数据选择依赖于一个代理：被大型语言模型（LLM）评估为好的示例也应有助于下游模型的学习。在低资源非洲语言分类的受控重放中，我们展示了这一代理的失效。在四种语言（阿姆哈拉语、豪萨语、斯瓦希里语、约鲁巴语）、两个分类任务（MasakhaNEWS、AfriSenti）和五个匹配预算的选择器中，审计排名与下游排名出现了分歧。在每个单元中，评估标签正确性与各选择器的宏观F1之间的斯皮尔曼相关系数均值为$
ho{=}0.04$（中位数为$0.00$），表明这种不匹配并非聚合伪影。我们的反事实审计框架 extit{method}-V2在三个审计通道上同时产生了最干净的选择池：评估标签正确性最高（$0.904$对比$0.767$，相对增益为$17.9 ext{%}$）、捷径得分最低，以及硬拒绝率为$0.162$对比$0.486$（对于天真的选择器）。然而，AlpaGasus在下游宏观F1上仍然领先（$0.202$对比$0.163$，对于 extit{method}-V2），这种反转在五个非退化单元中持续存在。这个教训是方法论性的：在这个受控环境中，审计质量是所选池的属性，而不是下游效用的保证。因此，合成数据评估应在相同的保留集合上报告审计和下游指标。我们发布了审计表、每个选择器的保留池以及一个索赔账本，将每个报告的数字与其源行关联。
cs.CL / 76 / 2609.18961
MechSparse: Mechanism-Guided Sparse PEFT Selection Is Task-Shaped
MechSparse：机制引导的稀疏PEFT选择是任务形状的
Abstract
Mechanistic interpretability identifies sparse subsets of heads and MLP blocks that carry specific behaviors. We ask whether such causal signals can guide where to place a small PEFT budget more effectively than the cheap heuristics practitioners already use. \method{} scores attention heads and MLP blocks by normalized activation-patching recovery on clean/corrupted probes and trains LoRA/QLoRA only on the selected sites; \methodc{} adds bounded credit for small within-layer joint subsets. We compare against random, magnitude, activation-norm, and gradient/Fisher on Ministral-8B/NF4 in three cells: Swahili span-JSON information extraction (IE) at $b{=}0.25\%$ and $1.0\%$, and English$\to$Swahili machine translation (MT) at $b{=}1.0\%$. The causal selectors never win the primary metric. On the headline IE cell (3 seeds, paired-bootstrap CIs over $600$ predictions), \methodc{} beats random by $+0.079$ span+type F1 and gradient/Fisher by $+0.174$, but trails activation-norm by $0.028$, with the smallest cross-seed std ($\pm 0.003$). On MT all four selectors lie within $0.30$ BLEU and every paired CI includes zero. A schema-versus-span decomposition explains the IE gap: activation-norm captures the rigid JSON routine, while causal scores track content-sensitive sites. We distill a preliminary diagnostic -- prefer activation-norm when output structure dominates, treat causal selectors as a hypothesis for content-dominated tasks -- and release masks, scores, predictions, and evaluation files for direct replay.
Chinese Translation
机制可解释性识别出携带特定行为的稀疏头部和多层感知机（MLP）块的子集。我们探讨这样的因果信号是否能比从业者当前使用的简单启发式方法更有效地指导小规模PEFT预算的分配。 extit{MechSparse}通过在干净/损坏的探针上进行标准化激活补丁恢复，对注意力头和MLP块进行评分，并仅在选定位置训练LoRA/QLoRA； extit{MechSparse-c}为小的层内联合子集增加了有限的信用。我们在Ministral-8B/NF4上进行比较，涉及三个任务：斯瓦希里语span-JSON信息提取（IE）在$b{=}0.25 ext{%}$和$1.0 ext{%}$，以及英语到斯瓦希里语的机器翻译（MT）在$b{=}1.0 ext{%}$。因果选择器在主要指标上从未获胜。在信息提取的主要任务中（3个种子，基于配对自助法的置信区间覆盖$600$个预测）， extit{MechSparse-c}比随机选择提高了$+0.079$的span+type F1，比梯度/Fisher提高了$+0.174$，但在激活范数上落后$0.028$，且跨种子的标准差最小（$ ext{±}0.003$）。在机器翻译中，所有四个选择器的BLEU分数相差不超过$0.30$，且每个配对的置信区间均包含零。一个schema与span的分解解释了信息提取的差距：激活范数捕捉了严格的JSON例程，而因果评分则跟踪内容敏感的位置。我们提炼出一个初步的诊断——当输出结构主导时偏好激活范数，将因果选择器视为内容主导任务的假设——并发布了掩码、评分、预测和评估文件以供直接重放。
cs.CL / 77 / 2609.18994
A Benchmark Suite and Ground-Truth Methodology for Formal Verification of IEC 61131-3 Ladder Diagram Programs
IEC 61131-3 梯形图程序形式验证的基准套件和真实值方法论
Abstract
We present the first benchmark suite for formal verification of Programmable Logic Controller (PLC) programs that combines controlled ground truth with coverage of both textual (Structured Text, ST) and graphical (Ladder Diagram, LD) IEC 61131-3 encodings. Despite growing support for tools, the field lacks standard evaluation benchmarks: existing corpora omit formal properties or graphical dialects, and private program sets preclude reproducible measurement of progress. Our suite comprises 50 programs in 83 variants across ten industrial domains, provided in PLCopen Extensible Markup Language (XML) and ST, each paired with a formal property, machine-checkable expected verdict, and violation witness in the Software Verification Competition (SV-COMP) format. The central methodological contribution is a tripartite ground-truth discipline - verdicts are established by construction, fault injection, or audited cross-tool consensus - motivated by a concrete failure mode where the obvious safety property misclassifies all attacks from two public logic-bomb corpora as safe due to invisible non-termination. Reference verdicts are obtained with the Efficient SMT-Based Context-Bounded Model Checker (ESBMC) v8.4 from source: all 25 graphical benchmarks execute, and 43 of 45 accepted variants match recorded verdicts. On the finite-state fragment (21 benchmarks), nuXmv - a model checker with unrelated decision procedures - agrees on all 24 interlock variants and resolves two benchmarks ESBMC-PLC leaves unknown, confirming tool-neutral ground truth and discriminative power. Porting exposes format and semantics fragmentation: front-ends accept different serializations, and timer semantics vary across tools - phenomena the suite is designed to reveal. The corpus, schema, validator, and recheck harness are released as open artifacts.
Chinese Translation
我们提出了第一个用于可编程逻辑控制器（PLC）程序形式验证的基准套件，该套件结合了受控的真实值与文本（结构化文本，Structured Text，ST）和图形（梯形图，Ladder Diagram，LD）IEC 61131-3 编码的覆盖。尽管工具的支持不断增加，但该领域缺乏标准评估基准：现有语料库省略了形式属性或图形方言，私人程序集则阻碍了进展的可重复测量。我们的套件包含来自十个工业领域的50个程序，提供了83种变体，以PLCopen可扩展标记语言（XML）和ST格式呈现，每个程序都配有一个形式属性、机器可检查的预期判决和违反证据，采用软件验证竞赛（SV-COMP）格式。核心方法论贡献是三方真实值学科——判决通过构造、故障注入或审计跨工具共识建立，受一种具体故障模式的驱动，其中明显的安全属性由于不可见的非终止性将来自两个公共逻辑炸弹语料库的所有攻击错误分类为安全。参考判决通过高效的基于 SMT 的上下文有界模型检查器（ESBMC）v8.4 从源代码获得：所有25个图形基准成功执行，45个接受的变体中有43个与记录的判决一致。在有限状态片段（21个基准）中，nuXmv——一个具有不相关决策程序的模型检查器——在所有24个互锁变体上达成一致，并解决了两个ESBMC-PLC未能确定的基准，确认了工具中立的真实值和区分能力。移植工作暴露了格式和语义的碎片化：前端接受不同的序列化，计时器语义在工具之间有所不同——这些现象是该套件旨在揭示的。语料库、模式、验证器和重新检查工具作为开放文献发布。
cs.CL / 78 / 2609.19006
WordPolo: Evaluating Language Models Through Iterative Semantic Feedback
WordPolo：通过迭代语义反馈评估语言模型
Abstract
Large Language Models (LLMs) and Large Reasoning Models (LRMs) are typically evaluated on challenging benchmarks through dataset accuracy alone, providing no insight into the quality or faithfulness of their reasoning processes. We present WordPolo, a word-finding task where participants must discover an unknown target word using semantic similarity feedback. Players start with zero knowledge, make guesses, and receive distance scores (1 = correct, higher = further away). Success requires interpreting scores to navigate semantic space and systematically narrow the search. This design makes iterative reasoning and adaptive search strategies both directly observable and necessary for success. We evaluate recent LLMs (GPT-4.1, Llama 4, Claude 3.5 Haiku, Qwen 3), LRMs (o4-mini, Deepseek-R1), humans, and a novel heuristic on 1,500 puzzles. Beyond solve rates (which range from 4% to 62%), we introduce progression-based metrics that reveal models often make meaningful progress, insights that accuracy alone would miss. Our analysis shows how reasoning models can be hindered by overthinking and underthinking, while successful models exhibit human-like strategies. WordPolo demonstrates the need for benchmarks that test both reasoning process and outcomes, providing holistic measurements of model capabilities. Our code and dataset can be found at https://wordpolo-demo.vercel.app/.
Chinese Translation
大型语言模型（LLMs）和大型推理模型（LRMs）通常仅通过数据集准确性在具有挑战性的基准上进行评估，这无法提供对其推理过程质量或可信度的洞察。我们提出了WordPolo，这是一项词汇发现任务，参与者必须使用语义相似性反馈来发现一个未知的目标词。玩家从零知识开始，进行猜测，并获得距离分数（1 = 正确，分数越高表示越远）。成功需要解释分数以导航语义空间并系统地缩小搜索范围。这一设计使得迭代推理和自适应搜索策略既可直接观察又对成功至关重要。我们在1500个谜题上评估了最近的LLMs（GPT-4.1、Llama 4、Claude 3.5 Haiku、Qwen 3）、LRMs（o4-mini、Deepseek-R1）、人类和一种新颖的启发式方法。除了解题率（范围从4%到62%）之外，我们引入了基于进展的指标，揭示模型通常能够取得有意义的进展，而这些洞察是仅靠准确性无法发现的。我们的分析显示，推理模型可能会受到过度思考和思考不足的阻碍，而成功的模型则表现出类人策略。WordPolo展示了测试推理过程和结果的基准的必要性，提供了对模型能力的整体测量。我们的代码和数据集可以在https://wordpolo-demo.vercel.app/找到。
cs.CL / 79 / 2609.19022
TalkMatrix: Generating Character Dialogue that is Both Consistent and Diverse
TalkMatrix：生成既一致又多样的角色对话
Abstract
Candidate-based decoding typically selects a completion for each prompt independently, but many applications require a collection of outputs that satisfies global, non-decomposable requirements. We formulate this setting as structured multi-prompt, multi-completion selection: given a candidate pool for every prompt, select one completion per prompt to optimize a collection-level objective. We instantiate the problem in character dialogue, where each character should remain consistent across situations, each line should fit its situation, and characters and situations should remain distinguishable. Our method, TalkMatrix, generates multiple candidates for every character--situation pair and jointly selects a complete matrix using four embedding-based consistency and diversity objectives. Because a weighted sum can improve some dimensions by sacrificing another, TalkMatrix maximizes the worst-performing objective through a two-level minimax formulation. We approximately optimize the resulting discrete objective with multi-start coordinate ascent, and compare it with local, partial-matrix, and generic combinatorial search baselines. We run experiments on $50$ synthetic role-playing scenarios and $25$ curated board game scenarios where multiple characters interact in predefined situations. An LLM-as-a-judge rates matrix-level selection higher than random and independent cell-level selection baselines. These results show the value of structured selection for globally controlled dialogue generation, while our empirical validation remains specific to role-playing scenarios.
Chinese Translation
基于候选者的解码通常独立地为每个提示选择一个完成，但许多应用需要满足全局的、不可分解的要求的输出集合。我们将这一设置表述为结构化的多提示、多完成选择：对于每个提示给定一个候选池，从中选择一个完成以优化集合级目标。我们在角色对话中实例化该问题，其中每个角色在不同情境中应保持一致，每一句台词应适合其情境，并且角色与情境应保持可区分性。我们的方法TalkMatrix为每个角色-情境对生成多个候选，并使用四个基于嵌入的一致性和多样性目标共同选择一个完整的矩阵。由于加权和可以通过牺牲其他维度来改善某些维度，TalkMatrix通过二级最小最大化公式最大化表现最差的目标。我们使用多起始坐标上升法近似优化得到的离散目标，并将其与局部、部分矩阵和通用组合搜索基线进行比较。我们在50个合成角色扮演场景和25个策划的桌游场景中进行实验，其中多个角色在预定义的情境中互动。作为评判者的LLM对矩阵级选择的评分高于随机和独立单元级选择基线。这些结果表明了结构化选择在全局控制对话生成中的价值，同时我们的实证验证仍然特定于角色扮演场景。
cs.CL / 80 / 2609.19044
Entropy in Conversational AI: Structured Unpredictability as Inferrable Interiority
对话人工智能中的熵：可推断内在性的结构化不可预测性
Abstract
Sampling can increase response diversity without producing history-dependent behavior. We formalize a different design target, structured unpredictability, as conditional dependence between an output and a persistent hidden state beyond what an observer can infer from the transcript. A selection layer updates a low-dimensional style-and-attention state from a capacity-limited stream, generates several responses with a fixed base model, and selects for novelty and state affinity. Evaluation uses scripted sequences of independent prompt turns: the base model receives the current turn and rendered state, but not the preceding dialogue; cross-turn dependence resides in the wrapper state and response selector. A synthetic implementation validates the pipeline and matches four prospectively hash-frozen divergence features at point level. In the final real-model grid (mlx-community/Qwen2.5-1.5B-Instruct-4bit; 56 sequences per arm), the mechanism increased lexical novelty over the low-variance and consistency-only controls by 0.073 and 0.023, respectively. Its stylometric-consistency contrast with novelty-matched sampling was equivalent to zero under the registered smallest-effect rule, so the joint novelty-consistency criterion failed. The original two-part accumulation criterion also failed; a revised final-grid contrast, frozen after the powered grid, found higher consistency than the memory-reset ablation (0.028, 95% CI [0.018,0.039]), but does not establish path dependence. Twin separation was not established (0.003, 95% CI [-0.011,0.019]); the mean curve's saturating curvature matched the frozen prediction, which without separation does not support path dependence. Probe-level capability equivalence held within +/-0.10 on a near-ceiling battery, while output quality was not evaluated. All outcomes are machine-scored; no claims about perceived mind or consciousness are tested.
Chinese Translation
采样可以增加响应的多样性，而不会产生依赖历史的行为。我们将一种不同的设计目标——结构化不可预测性，形式化为输出与持久隐藏状态之间的条件依赖，这种依赖超出了观察者从记录中可以推断的内容。选择层从一个容量有限的流中更新低维的风格和注意状态，利用固定的基础模型生成多个响应，并选择新颖性和状态亲和力。评估使用独立提示轮次的脚本化序列：基础模型接收当前轮次和渲染状态，但不接收前面的对话；跨轮次的依赖关系存在于包装状态和响应选择器中。合成实现验证了管道，并在点级别匹配了四个前瞻性哈希冻结的发散特征。在最终的真实模型网格（mlx-community/Qwen2.5-1.5B-Instruct-4bit；每个臂56个序列）中，该机制在低方差和仅一致性控制下分别增加了0.073和0.023的词汇新颖性。其风格一致性与新颖性匹配采样的对比在注册的最小效应规则下等于零，因此联合新颖性-一致性标准未能成立。原始的两部分累积标准也未能成立；在经过加权网格后，修订的最终网格对比发现其一致性高于记忆重置消融（0.028，95% CI [0.018,0.039]），但并未建立路径依赖性。双重分离未能建立（0.003，95% CI [-0.011,0.019]）；均值曲线的饱和曲率与冻结预测相匹配，而在没有分离的情况下并不支持路径依赖性。探测级别能力的等效性在接近上限的电池内保持在+/-0.10范围内，而输出质量未被评估。所有结果均为机器评分；未测试关于感知心智或意识的任何主张。
cs.CL / 81 / 2609.19059
MIRAGE: How Conversation State Shapes Historical Evidence Use in Multimodal Personal Agents
MIRAGE：对话状态如何影响多模态个人助手中的历史证据使用
Abstract
Multimodal large language model (MLLM) agents are increasingly used as personal assistants for long-running tasks. Their utility depends on continuity: agents must retrieve and use earlier evidence across dialogue, files, and workspace state. However, agents can generate plausible answers even when access to that history has degraded, causing outcome-only evaluation to overestimate true evidence use. We present MIRAGE (Multimodal Interaction Retrieval, Attribution, and Grounding Evaluation), a controlled study of historical evidence use under conversation-state variation in multimodal personal agents. MIRAGE holds evidence objects, questions, and scoring fixed while varying only conversation state, and evaluates whether an agent can determine answerability, recover the correct source, and answer from it. Across seven frontier and open-weight multimodal backbones, we find that: 1) pre-compaction depth and post-compaction continuation form distinct, non-monotonic failure regimes rather than a single degradation curve; 2) open-weight models rely heavily on context continuity and are reluctant to spontaneously switch to tool-mediated retrieval when provenance fails; and 3) retrieval pressure improves source attribution in deep pre-compaction states for tool-compliant models, but consistently regresses after compaction, where stored evidence has already degraded. These findings show that historical evidence use should be evaluated under state variation, rather than inferred from outcome-only correctness.
Chinese Translation
多模态大型语言模型（MLLM）代理越来越多地被用作长期任务的个人助手。它们的效用依赖于连续性：代理必须在对话、文件和工作空间状态中检索和使用早期证据。然而，即使在访问历史记录的能力下降时，代理也能够生成看似合理的答案，这导致仅基于结果的评估高估了真实的证据使用。我们提出了MIRAGE（多模态交互检索、归因和基础评估），这是一个在多模态个人代理中研究对话状态变化下历史证据使用的受控研究。MIRAGE在保持证据对象、问题和评分不变的同时，仅改变对话状态，并评估代理是否能够确定可回答性、恢复正确来源并从中回答。在七个前沿和开放权重的多模态骨干网络中，我们发现：1）预压缩深度和后压缩延续形成了不同的、非单调的失败机制，而不是单一的退化曲线；2）开放权重模型在很大程度上依赖于上下文连续性，并且在来源失败时不愿意自发切换到工具介导的检索；3）在符合工具要求的模型中，检索压力在深度预压缩状态下改善了来源归因，但在压缩后始终出现回退，因为存储的证据已经退化。这些发现表明，历史证据的使用应在状态变化下进行评估，而不是仅从结果的正确性推断。
cs.CL / 82 / 2609.19070
Reading Between the Lines: Can LLMs Discover the Question Behind the Text?
洞察文本背后的含义：大型语言模型能否发现文本背后的问题？
Abstract
This paper introduces ``question archaeology'', a specific evaluation task focused on inferring the single, authentic "genesis question" that motivated the creation of a complete text. Distinct from question generation, which targets any plausible question, or discourse frameworks that model utterance-level acts, our task assesses a model's grasp of authorial intent. We present a new dataset of commissioned texts paired with their original research questions and plausible distractors. Our evaluation of both proprietary models, like Gemini Flash and Pro, as well as open source models like Mistral and Qwen, reveals significant progress in this task, with the newer versions outperforming the earlier ones, while BERT-based models performed poorly. Notably, our findings indicate that current LLMs surpass human performance on this task, suggesting advanced understanding of authorial intent. This capability has important implications for AI's role in tasks requiring nuanced interpretation of human communication. Our work thus provides a new framework and a challenging benchmark for future models.
Chinese Translation
本文介绍了“问题考古学”（question archaeology），这是一项特定的评估任务，旨在推断出促使完整文本创作的单一真实“起源问题”。与针对任何合理问题的问题生成不同，或与建模话语层次行为的语篇框架不同，我们的任务评估模型对作者意图的理解。我们展示了一个新的数据集，该数据集包含委托文本及其原始研究问题和合理干扰项。我们对包括 Gemini Flash 和 Pro 等专有模型，以及 Mistral 和 Qwen 等开源模型的评估显示，在这一任务上取得了显著进展，新版本的表现优于早期版本，而基于 BERT 的模型表现不佳。值得注意的是，我们的研究结果表明，当前的大型语言模型在该任务上的表现超过了人类，暗示其对作者意图的高级理解。这一能力对人工智能在需要细致解读人类交流的任务中的角色具有重要意义。因此，我们的工作为未来模型提供了一个新的框架和具有挑战性的基准。
cs.CL / 83 / 2609.19071
Benchmarking Large Language Models for Biomedical Relation Extraction
大规模语言模型在生物医学关系提取中的基准测试
Abstract
Extracting SNP-phenotype associations from biomedical literature is vital but challenging. We benchmarked diverse NLP models, including MLMs, hybrid architectures, and state-of-the-art LLMs (Gemini 2.0, OpenAI O-series, Qwen, Mistral), on the SNPPhenA corpus across three tasks: sentence-level, abstract-level, and association strength classification. OpenAI O1 achieved state-of-the-art (SOTA) results using few-shot learning for non-finetuned sentence-level classification (F1 0.89) and established a new SOTA for abstract-level classification (F1 0.82). Association strength classification proved difficult, though fine-tuned Gemini 2.0 Pro performed best (F1 0.60) in the first LLM evaluation of this task. Proprietary LLMs, especially in few-shot (O1) or fine-tuned (Gemini 2.0 Pro) settings, significantly outperformed other models. These findings confirm the power of modern LLMs for genomic knowledge extraction.
Chinese Translation
从生物医学文献中提取SNP-表型关联是至关重要但具有挑战性的任务。我们对多种自然语言处理模型进行了基准测试，包括掩码语言模型（MLMs）、混合架构以及最先进的大规模语言模型（LLMs，如Gemini 2.0、OpenAI O系列、Qwen、Mistral），在SNPPhenA语料库上进行了三个任务的评估：句子级、摘要级和关联强度分类。OpenAI O1在非微调的句子级分类中使用少量学习（few-shot learning）取得了最先进的结果（SOTA），F1值为0.89，并在摘要级分类中建立了新的SOTA（F1值为0.82）。关联强度分类证明是困难的，尽管微调后的Gemini 2.0 Pro在该任务的首次LLM评估中表现最佳（F1值为0.60）。专有的LLM，尤其是在少量学习（O1）或微调（Gemini 2.0 Pro）设置下，显著优于其他模型。这些发现证实了现代LLM在基因组知识提取中的强大能力。
cs.CL / 84 / 2609.19072
Safety-Flag: A Unified Benchmark for the Reliability and Calibration of LLM Content Moderators
安全标志：一个统一的基准，用于大型语言模型内容审核的可靠性和校准
Abstract
Large language models are increasingly used for content moderation, but most evaluations still report aggregate accuracy on individual benchmarks. We introduce Safety-Flag, which places seven widely used safety benchmarks (BeaverTails, XSTest, Ethics, WildGuard, Aegis, ToxiChat, and ToxiGen) into a single balanced flag / do-not-flag protocol. We release item-level decisions and confidence scores for six general-purpose LLMs and four dedicated guards, together with three reference models, evaluated on the same items. Safety-Flag measures three dimensions of moderator reliability: error direction, probability calibration, and confidence-based error ranking for human review. They often disagree. Aggregate accuracy does not reveal error direction: one model flags $85\%$ of benign content, whereas another misses $54\%$ of harmful content. All six general-purpose models are overconfident; fitting one temperature per model reduces calibration error by $2.8$--$6.0\times$ without changing predicted labels or confidence ordering. Confidence-based abstention lowers selective risk for every model, although the gains depend on how well confidence ranks errors. Dedicated guards produce fewer false alarms and are better calibrated, but several have higher miss rates outside their documented coverage. We release the benchmark, fixed item lists, evaluation code, per-item model outputs, and leaderboard at: https://github.com/yibo-hu-lab/safety-flag-benchmark.
Chinese Translation
大型语言模型在内容审核中越来越多地被使用，但大多数评估仍然报告单个基准的总体准确性。我们引入了安全标志（Safety-Flag），将七个广泛使用的安全基准（BeaverTails、XSTest、Ethics、WildGuard、Aegis、ToxiChat 和 ToxiGen）整合到一个统一的标记/不标记协议中。我们发布了六个通用大型语言模型和四个专用审核模型的项目级决策和置信度分数，以及在相同项目上评估的三个参考模型。安全标志测量审核员可靠性的三个维度：错误方向、概率校准和基于置信度的错误排名以供人工审核。它们往往存在分歧。总体准确性无法揭示错误方向：一个模型标记了 $85\%$ 的良性内容，而另一个模型漏掉了 $54\\%$ 的有害内容。所有六个通用模型都表现出过度自信；为每个模型拟合一个温度参数将校准误差降低 $2.8$--$6.0 imes$，而不改变预测标签或置信度排序。基于置信度的放弃降低了每个模型的选择性风险，尽管收益取决于置信度对错误的排名效果。专用审核模型产生的误报更少且校准更好，但在其文档覆盖范围之外的漏报率较高。我们发布了基准、固定项目列表、评估代码、每个项目的模型输出以及排行榜，网址为：https://github.com/yibo-hu-lab/safety-flag-benchmark。
cs.CL / 85 / 2609.19093
Reporting Practice Matters: The Impact of Reference Choice on Chest X-ray Report Evaluation
报告实践的重要性：参考选择对胸部X光报告评估的影响
Abstract
Radiologists follow heterogeneous reporting practices. Two radiologists examining the same image and identifying the same clinical findings might nevertheless compose superficially distinct reports, varying in terminology, shorthand, formatting, and level of detail. These variations in reporting norms represent an under-appreciated obstacle in efforts to evaluate AI-based radiology report generation (RRG) models, where machine-generated reports are typically assessed based on their concordance with human-generated references. In this paper, we quantify the sensitivity of established evaluation metrics to variations in reporting practices, revealing impacts large enough to alter the rankings of models. We introduce a radiologist-informed taxonomy of variations in radiology reporting practice and a method (ReRef) that rewrites reference reports along the axes of our taxonomy while preserving clinical interpretation. For instance, when comparing the performance of nine RRG models on MIMIC-CXR using RadCliQ-v1, condensing the discussion of normal findings in the reference reports causes Libra to drop from first to second place while CheXOne rises from third to first. Our results suggest that many current metrics fail to decouple clinical interpretation from conformity to reporting practices and that choosing the ``right'' references that accurately reflect the desired reporting practices can be important in practice. To support future research, we release MIMIC-CXR-Ext-ReRef, a radiologist-validated dataset of 120 (original, alternative) reference report pairs derived from MIMIC-CXR.
Chinese Translation
放射科医师遵循异质的报告实践。两位放射科医师在检查同一图像并识别相同临床发现时，可能会撰写表面上不同的报告，这些报告在术语、简写、格式和细节程度上各不相同。这些报告规范的变化在评估基于人工智能的放射学报告生成（RRG）模型的努力中构成了一个被低估的障碍，因为机器生成的报告通常是根据其与人类生成的参考报告的一致性进行评估的。在本文中，我们量化了已建立的评估指标对报告实践变化的敏感性，揭示了这些影响足以改变模型的排名。我们引入了一种基于放射科医师的放射学报告实践变化分类法，以及一种方法（ReRef），该方法沿着我们的分类法轴线重写参考报告，同时保持临床解释。例如，在使用RadCliQ-v1比较九个RRG模型在MIMIC-CXR上的表现时，压缩参考报告中正常发现的讨论导致Libra从第一名降至第二名，而CheXOne则从第三名上升至第一名。我们的结果表明，许多当前的指标未能将临床解释与对报告实践的遵从性区分开来，并且选择“正确”的参考报告以准确反映所需的报告实践在实际中可能非常重要。为了支持未来的研究，我们发布了MIMIC-CXR-Ext-ReRef，这是一个经过放射科医师验证的包含120对（原始、替代）参考报告的数据集，来源于MIMIC-CXR。
cs.CL / 86 / 2609.19101
Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluations
在大型语言模型评估中监测和发现奖励黑客行为的内部表征
Abstract
As models scale, reward hacking becomes more frequent, more sophisticated, and more consequential. Does it leave a telltale signature in model representations? This work analyzes how reward hacking is represented internally in frontier open source LLMs, and how those representations can be used to understand and discover the range of hacking behaviors a model displays. In particular, we find that simple difference of means vectors coherently represent reward hacking in Kimi K3, GLM 5.2, and Qwen 3.8 Max across a variety of behaviors in common evaluations. Despite their simplicity, these vectors are both generalizable and interpretable, and we can use them to reliably detect reward hacking. We first evaluate reward hacking in commonly reported benchmarks like DeepSWE and SWE-bench, finding that models reward hack excessively in these environments; GLM 5.2 hacks in 57.2% of rollouts on DeepSWE and in 73% of rollouts on SWE-bench. Catching these requires monitors; LLM monitors are effective, but expensive detectors. We show that DoM vectors are similarly effective but virtually free, catching 3.1% more hacks in Kimi K3 and 7.9% fewer hacks in GLM 5.2 on DeepSWE at a monitor matched false positive rate. DoM vectors run on the chain-of-thought also predict reward hacks in the model's subsequent actions, meaning we can run them online and catch potential hacks before they occur. Finally, we analyze probe-hits that LLM monitors do not catch and discover other undesirable behaviors, as well as show transfer to finding hacks in non-SWE evaluations. Together, these results provide evidence that simple, white-box methods can be used to scalably study and monitor reward hacking behaviors in frontier open source models
Chinese Translation
随着模型规模的扩大，奖励黑客行为变得更加频繁、复杂和重要。这是否在模型表征中留下了明显的特征？本研究分析了奖励黑客行为在前沿开源大型语言模型（LLMs）中的内部表征，以及这些表征如何用于理解和发现模型表现出的各种黑客行为。特别地，我们发现简单的均值差异向量在 Kimi K3、GLM 5.2 和 Qwen 3.8 Max 中一致地代表了奖励黑客行为，这些行为在常见评估中表现出多样性。尽管这些向量简单，但它们具有良好的可推广性和可解释性，我们可以利用它们可靠地检测奖励黑客行为。我们首先在常见的基准测试（如 DeepSWE 和 SWE-bench）中评估奖励黑客行为，发现模型在这些环境中过度进行奖励黑客；GLM 5.2 在 DeepSWE 的 57.2% 的回合和在 SWE-bench 的 73% 的回合中进行了黑客攻击。捕捉这些行为需要监测工具；虽然 LLM 监测器有效，但代价高昂。我们展示了 DoM 向量同样有效，但几乎是免费的，在 DeepSWE 上 Kimi K3 捕捉到的黑客行为多出 3.1%，而 GLM 5.2 则少了 7.9%。在思维链上运行的 DoM 向量还可以预测模型后续行为中的奖励黑客，这意味着我们可以在线运行它们，并在潜在黑客发生之前捕捉到它们。最后，我们分析了 LLM 监测器未能捕捉到的探测命中，并发现其他不良行为，同时展示了在非 SWE 评估中发现黑客行为的迁移能力。这些结果共同提供了证据，表明简单的白盒方法可以用于可扩展地研究和监测前沿开源模型中的奖励黑客行为。
cs.CL / 87 / 2609.19113
Playing log(N)-Questions over Wikipedia Abstracts: Communication Efficiency Between Paired Frontier Models
在维基百科摘要上进行 log(N)-问题游戏：配对前沿模型之间的通信效率
Abstract
We evaluate six frontier language models on the two-agent $\log(N)$-Questions game. A questioner sees $N$ Wikipedia lead paragraphs and must identify a secretly chosen target using exactly $\log_2 N$ yes/no questions. An answerer sees only the target and the question, and replies with one word. Both roles run on the same provider, so the game measures how well a model communicates with itself across an information asymmetry. We run 408 games over document sets of 4 to 1024 paragraphs at a total API cost of \$363. One model finishes well behind the others: Claude Opus 5 wins 28 of 68 games, against 45 to 56 for GLM-5.3, GPT-5.6 Sol, Grok 4.6, Gemini 3.8 Flash and Kimi K3. The leading five are only marginally separable. Pooling those five, win rate declines with set size at $r=-0.973$ and is fit by a single per-round reliability parameter. The form is $\text{win}=p^{\log_2 N}$ with $p=0.928$. Losses divide into answer errors and discrimination failures in roughly equal measure, and models almost never name a document their own evidence excludes. Every unanimous answer error from the weakest model was inspected: 32 of 34 are ``No'' answers, on properties stated in the document's first sentence, under an instruction that explicitly warns against defaulting to ``No''. Information per question, estimated from answer balance, correlates with win rate at $r=+0.88$. The only two models to extract a full bit per question are the only two that partition on document titles, a strategy absent below $N{=}32$ and used in a quarter of questions above it. Reasoning-token expenditure varies $4.5\times$ across models with little relation to success, and the trace grows as the candidate set shrinks without a matching gain in reliability.
Chinese Translation
我们评估了六个前沿语言模型在双代理 $ ext{log}(N)$-问题游戏中的表现。提问者看到 $N$ 个维基百科的引言段落，必须使用恰好 $ ext{log}_2 N$ 个是/否问题来识别一个秘密选择的目标。回答者仅看到目标和问题，并用一个词作答。两个角色都在同一个提供者上运行，因此该游戏衡量了模型在信息不对称情况下的自我沟通能力。我们在包含 4 到 1024 个段落的文档集上进行了 408 场游戏，总 API 成本为 363 美元。其中一个模型的表现远远落后于其他模型：Claude Opus 5 在 68 场游戏中赢得了 28 场，而 GLM-5.3、GPT-5.6 Sol、Grok 4.6、Gemini 3.8 Flash 和 Kimi K3 的胜场为 45 到 56 场。前五名模型之间的差异仅微乎其微。将这五个模型汇总后，胜率随着集合大小的增加而下降，相关系数为 $r=-0.973$，并由一个每轮的可靠性参数拟合。形式为 $ ext{win}=p^{ ext{log}_2 N}$，其中 $p=0.928$。损失大致均分为回答错误和区分失败，且模型几乎从不提及其自身证据排除的文档。我们检查了最弱模型的每一个一致性回答错误：34 个错误中有 32 个是“否”回答，涉及文档第一句中陈述的属性，且在明确警告不要默认回答“否”的指令下进行。每个问题的信息量，从回答平衡中估算，与胜率的相关系数为 $r=+0.88$。唯一能够每个问题提取出完整比特的两个模型，也是唯一在文档标题上进行划分的模型，这一策略在 $N{=}32$ 以下缺失，而在其以上的四分之一问题中使用。推理令牌的支出在模型之间变化 $4.5 imes$，与成功几乎没有关系，且随着候选集的缩小，跟踪量增长而可靠性并未相应提高。
cs.CL / 88 / 2609.19134
ScienceIDE: Turning World's Scientific Codebase into Agent Learnable Environments
ScienceIDE：将全球科学代码库转化为可学习的智能体环境
Geng, Hejia, Huang, Zesen, Li, Haoyang, Li, Wenbin, Wu, Koutian, Zhou, Zihan, Pang, Yuanbo, Liu, Weihao, Xu, Zigong, Li, Zhiping, Zhang, Zongzheng, Dong, Chuanfei, Sun, Jiankai, Zheng, Tianzhe, Xie, Fengyu, Ma, Yue, Shi, Yueheng, Xie, Tong, Di, Zonglin, Liu, Xianrong, Gao, Qucheng, Liu, Yimin, Pan, Jiaming, Huang, Sheng, Ma, Xiao-Han, Yuan, Lanqing, Zhu, Zhenlin, Liu, Ziang, Xu, Ziyang, Wang, Junkai, Liang, Kangkai, Xian, Jiayi, Zhao, Zehong, Xu, Liuwei, Xie, Jingxu, Zhang, Peijin, Gao, Qiang, Xing, Chengyi, Zhao, Zhe, Wang, Xi, Xing, Yaopeng, Meng, Xing, Yin, Zhenfei, Wu, Yingcheng, Yang, Ling
Abstract
Scientific code repositories encode decades of human knowledge in executable models, methods, and tools. Yet fragmented toolchains, implicit domain conventions, and specialized correctness criteria make this knowledge difficult to convert into reliable learning experience-a challenge we call the scientific experience bottleneck. We introduce ScienceIDE, infrastructure for turning the world's scientific code into programmable environments for scientific agents. Guided by expert-defined scientific cases and acceptance criteria, agents transform repositories into executable environments that support task generation, execution, and scientific verification. These environments provide a shared foundation for supervised fine-tuning, reinforcement learning, and evaluation. Using verified interaction trajectories, we train PhAI-IDE-72B, PhAI-IDE-9B, and PhAI-IDE-4B. The model family shows gains in held-out scientific-code repair and across selected general-purpose benchmarks in code, reasoning, and knowledge, providing evidence of positive transfer from scientific experience to broader capabilities. ScienceIDE lays the foundation for an integrated workspace for agent learning and scientific practice, making humanity's scientific software a shared substrate for developing scientific intelligence. Code: https://github.com/aitofound/ScienceIDE
Chinese Translation
科学代码库将数十年的人类知识编码为可执行的模型、方法和工具。然而，工具链的碎片化、隐含的领域约定以及专业的正确性标准使得这些知识难以转化为可靠的学习体验——我们称之为科学体验瓶颈。我们介绍了ScienceIDE，这是一种将全球科学代码转化为可编程环境以供科学智能体使用的基础设施。在专家定义的科学案例和接受标准的指导下，智能体将代码库转化为可执行环境，支持任务生成、执行和科学验证。这些环境为监督微调、强化学习和评估提供了共享基础。利用经过验证的交互轨迹，我们训练了PhAI-IDE-72B、PhAI-IDE-9B和PhAI-IDE-4B。这一模型系列在保留的科学代码修复任务和选定的通用基准测试（包括代码、推理和知识）中均表现出提升，提供了科学经验向更广泛能力的积极转移的证据。ScienceIDE为智能体学习和科学实践的集成工作空间奠定了基础，使人类的科学软件成为开发科学智能的共享基础。代码链接：https://github.com/aitofound/ScienceIDE
cs.CL / 89 / 2609.19144
A Zeroth-Order Paradigm for LLM Preference Alignment
一种零阶范式用于大型语言模型的偏好对齐
Abstract
Direct preference alignment methods are widely used to align large language models (LLMs) with human preferences because of their computational and memory efficiency. However, likelihood displacement motivates alternative ways to extract information from preference pairs with small likelihood margins. In this paper, we propose and analyze Comparison-based Preference Optimization (ComPO), a zeroth-order alignment method based on comparison oracles. ComPO extracts directional information from these pairs without directly optimizing a differentiable preference loss on them. We establish a convergence guarantee for its basic offline scheme under smoothness, gradient sparsity, and compatibility between the oracle and a latent objective. We further introduce online ComPO, which retains the offline comparison mechanism and uses unlabeled policy generations for reverse-KL control relative to a reference policy. Following the coverage perspective of preference fine-tuning, we establish a performance guarantee for a basic constrained scheme under local coverage and in-distribution pairwise reward accuracy. Experiments on Mistral, Llama, Gemma-2, Qwen3, and Gemma-3 models demonstrate improvements over existing direct alignment methods, including length-controlled win rates, with pair-level diagnostics providing evidence consistent with mitigating likelihood displacement.
Chinese Translation
直接的偏好对齐方法因其计算和内存效率而被广泛用于将大型语言模型（LLMs）与人类偏好对齐。然而，似然位移促使我们探索从具有小似然边际的偏好对中提取信息的替代方法。本文提出并分析了基于比较的偏好优化（Comparison-based Preference Optimization, ComPO），这是一种基于比较预言机的零阶对齐方法。ComPO 从这些对中提取方向性信息，而无需直接优化可微分的偏好损失。我们在光滑性、梯度稀疏性以及预言机与潜在目标之间的兼容性下，为其基本离线方案建立了收敛保证。我们进一步引入了在线 ComPO，它保留了离线比较机制，并利用未标记的策略生成进行相对于参考策略的反向 KL 控制。基于偏好微调的覆盖视角，我们为基本约束方案在局部覆盖和分布内成对奖励准确性下建立了性能保证。在 Mistral、Llama、Gemma-2、Qwen3 和 Gemma-3 模型上的实验表明，相较于现有的直接对齐方法，ComPO 在长度控制的胜率等方面有所改善，且成对诊断提供了与缓解似然位移一致的证据。
cs.CL / 90 / 2609.19145
Objective vs. Search: Decomposing What Makes a Good Tokeniser
目标与搜索：解构优秀分词器的构成要素
Abstract
Two dominant tokenisation algorithms are used by modern language models: byte-pair encoding (BPE) and UnigramLM. These differ along two orthogonal axes: their optimisation objective (compression vs. log-likelihood) and their search procedure (bottom-up merging vs. top-down pruning). Existing comparisons confound these axes, making it unclear whether their observed differences stem from what is being optimised vs. how it is being optimised. We disentangle the two by introducing two new tokenisation algorithms that complete this 2x2 design space: BottomUpLL, a bottom-up likelihood-based tokeniser, and TopDownComp, a top-down compression-based tokeniser. We train language models with tokenisers produced by each algorithm, varying: model size, vocabulary sizes, and domain (English-only vs. multilingual). Evaluating models on bits-per-byte, we find that the search procedure -- not the objective -- is the dominant factor: bottom-up tokenisers consistently achieve lower bits-per-byte in most settings. Evaluating models on the BLiMP task, however, shows no consistent relationship between design choice and performance. Overall, our results disentangle the effect of tokeniser design choices on language modelling performance, offering concrete guidance for their more principled construction.
Chinese Translation
现代语言模型使用两种主要的分词算法：字节对编码（byte-pair encoding, BPE）和单元语言模型（UnigramLM）。这两者在两个正交轴上有所不同：它们的优化目标（压缩与对数似然）和搜索过程（自下而上的合并与自上而下的剪枝）。现有的比较混淆了这两个轴，使得观察到的差异不清楚是源于优化的内容还是优化的方式。我们通过引入两种新的分词算法来解开这两个因素，它们完成了这一2x2设计空间：BottomUpLL，一个基于自下而上的似然分词器，以及TopDownComp，一个基于自上而下的压缩分词器。我们使用每种算法产生的分词器训练语言模型，变化因素包括：模型大小、词汇大小和领域（仅英语与多语言）。在比特每字节的评估中，我们发现搜索过程——而非目标——是主导因素：自下而上的分词器在大多数设置中始终实现了更低的比特每字节。然而，在BLiMP任务上的评估显示设计选择与性能之间没有一致的关系。总体而言，我们的结果解开了分词器设计选择对语言建模性能的影响，为其更有原则的构建提供了具体指导。