arXiv Daily Digest

378

Papers

AthenaZero: A low-inertia, bimanual robot for dynamic manipulation

AthenaZero：一种低惯量的双手机器人用于动态操作

Morgan, Andrew S., Xie, Gregory, Bass, Capprin, Thomasson, Rachel, Wang, Chunpeng, Shahriari, Erfan, Busa, Harrison, Araromi, Oluwaseun, Aronov, Joseph, Burgess, Michael, Dimitrov, Velin D., Estrada, Matthew A., Hamdi, Faris, Kendig, Samuel, Lee, Taeyoon, Mur-Miranda, Jose Oscar, Sommers, Emma, Titchener, Paul, Wang, Margaret, Wilson, Achu, Yeatman, Mark, Yirmibesoglu, Osman Dogan, Rizzi, Alfred A., Mozeika, Annan, Rojas, Nicolas, Odhner, Lael

Abstract

AthenaZero is a bimanual manipulator designed to minimize inertia without compromising control authority. By utilizing quasi-direct drive actuation and transmission remotization techniques, the system achieves an effective endpoint mass comparable to that of a human---about an order of magnitude less than conventional robot manipulators. This characteristic, combined with its inherent torque transparency, makes AthenaZero exceptionally well-suited for dynamic manipulation. We describe the methodology} that led to this design and demonstrate the robot's capabilities on three baseball-inspired tasks: throwing, catching, and batting, which showcase complex interactions on human-comparable timescales where milliseconds matter. AthenaZero was capable of throwing at speeds in excess of 30 m/s, with catching and batting at speeds in excess of 14 m/s over a short 7.3 m distance. Batting practice and a game of catch were subsequently performed in robot-to-robot and human-to-robot variations, showcasing the efficacy and adaptability of our system in tasks that require high acceleration.

Chinese Translation

AthenaZero 是一种双手操作器，旨在在不妥协控制权的情况下最小化惯量。通过利用准直接驱动（quasi-direct drive）驱动和传动远程化（transmission remotization）技术，该系统实现了与人类相当的有效端点质量——约为传统机器人操作器的一个数量级更低。这个特性，加上其固有的扭矩透明性，使得 AthenaZero 特别适合于动态操作。我们描述了导致这一设计的方法论，并展示了机器人在三个以棒球为灵感的任务上的能力：投掷、接球和击球，这些任务展示了在毫秒级别的时间尺度上复杂的交互。AthenaZero 能够以超过 30 米/秒的速度投掷，并在短距离 7.3 米内以超过 14 米/秒的速度进行接球和击球。随后进行了机器人与机器人以及人类与机器人之间的击球练习和接球游戏，展示了我们系统在需要高加速度的任务中的有效性和适应性。

View on arXiv Download PDF AI Translation

cs.RO / 2 / 2609.19196

DITTO: Dexterous Interface for Transparent TeleOperation

DITTO：透明远程操作的灵巧接口

Palacios, Joaquin, Lee, Katelyn, Zhang, Cheng, He, Zhanpeng, Ciocarlie, Matei

Abstract

Collecting data for manipulation with high-DOF hands is challenging, as interfaces must capture rich hand motion while rendering the contact interactions essential for precise manipulation. Existing data collection approaches face a trade-off: teleoperation ensures deployment consistency but lacks force feedback, while handheld (in-the-wild) systems provide natural force transparency but introduce a visual embodiment gap at deployment. We present DITTO, a Dexterous Interface for Transparent TeleOperation, which resolves this through the anatomically informed co-design of a dexterous 7-DOF robotic hand and a kinematically equivalent motorized exoskeleton. A 1-to-1 actuator mapping between the exoskeleton and robotic hand enables handheld (in-the-wild) data collection and bilateral teleoperation with joint-level force feedback unified in a single platform. We demonstrate that the DITTO exoskeleton spans the operator's natural index-to-thumb workspace, and showcase DITTO's dexterous capabilities via learned policies on contact-rich tasks.

Chinese Translation

使用高自由度（DOF）手进行操作数据收集具有挑战性，因为接口必须捕捉丰富的手部运动，同时渲染对精确操作至关重要的接触交互。现有的数据收集方法面临权衡：远程操作确保了部署的一致性，但缺乏力反馈，而手持（在野外）系统提供了自然的力透明性，但在部署时引入了视觉体现的差距。我们提出了DITTO，一个透明远程操作的灵巧接口，通过解剖学信息驱动的灵巧7-DOF机器人手和运动学等效的电动外骨骼的共同设计来解决这一问题。外骨骼与机器人手之间的一对一执行器映射使得手持（在野外）数据收集和双向远程操作成为可能，并在单一平台上统一了关节级力反馈。我们展示了DITTO外骨骼覆盖了操作员自然的食指到拇指的工作空间，并通过在接触丰富任务上的学习策略展示了DITTO的灵巧能力。

View on arXiv Download PDF AI Translation

cs.RO / 3 / 2609.19200

ULOHA: An Underwater Bimanual Robot System for Robot Learning

ULOHA：一种用于机器人学习的水下双手机器人系统

Kobayashi, Masato, Tsunoori, Takeru

Abstract

Underwater visuomotor policy learning has focused primarily on single manipulators, while bimanual imitation learning has been studied largely in air. We present ULOHA, an underwater bimanual robot learning platform that combines custom-designed leader--follower hardware with software extensions to LeRobot, integrating teleoperation, multi-view sensing, demonstration collection, policy training, and autonomous deployment. Real-robot experiments demonstrate a range of coordinated underwater bimanual behaviors, including inter-arm transfer, shared-object manipulation, and buoyancy-driven interception. We evaluate ACT, Diffusion Policy, and the vision--language--action model SmolVLA on the platform. We investigate how learning methods and execution strategies developed for manipulation in air perform underwater, examining bubble disturbances, buoyancy-driven object motion, action-execution horizons, and real-time chunking. A separate single-arm study examines policy transfer between air and water and shows that demonstrations spanning both media support execution in both under the tested conditions. ULOHA provides a unified experimental platform for studying underwater bimanual robot learning under the coupled perceptual and physical effects of underwater environments. Additional material: https://mertcookimg.github.io/uloha/

Chinese Translation

水下视觉运动策略学习主要集中在单一操控器上，而双手模仿学习则主要在空气中进行研究。我们提出了ULOHA，这是一种水下双手机器人学习平台，结合了定制设计的领导-跟随硬件和对LeRobot的软件扩展，集成了远程操作、多视角感知、演示收集、策略训练和自主部署。真实机器人实验展示了一系列协调的水下双手行为，包括臂间转移、共享物体操控和浮力驱动的拦截。我们在该平台上评估了ACT、扩散策略和视觉-语言-动作模型SmolVLA。我们研究了为空气中的操控开发的学习方法和执行策略在水下的表现，考察了气泡干扰、浮力驱动的物体运动、动作执行的时间范围和实时分块。一个单臂的独立研究考察了空气与水之间的策略转移，结果表明，跨越两种介质的演示支持在测试条件下的执行。ULOHA提供了一个统一的实验平台，用于研究水下双手机器人学习，考虑了水下环境的感知与物理效应的耦合。附加材料： https://mertcookimg.github.io/uloha/

View on arXiv Download PDF AI Translation

cs.RO / 4 / 2609.19204

REACT: A Fully Spiking State-Space Model for Real-Time Event-Driven Temporal Perception

REACT：一种用于实时事件驱动时间感知的完全脉冲状态空间模型

Keime, Geoffroy, Cuperlier, Nicolas, Cottereau, Benoit R.

Abstract

Robotic systems operating in dynamic environments require visual perception that evolves continuously with the incoming sensory stream. Event cameras provide microsecond temporal resolution and asynchronous sensing, but most learning-based methods accumulate events into frames or temporal bins, introducing an integration delay that can limit fast reaction. Here we propose REACT, a fully spiking state-space model for event-driven temporal perception that processes raw events one by one, without temporal accumulation. REACT uses a complex-valued spiking neuron, C-SiLIF, whose continuous-time dynamics are driven by the physical inter-event interval, allowing its internal state to evolve at the temporal resolution of individual events. We evaluate REACT on gesture recognition and time-to-collision (TTC) estimation from full-field event streams, without a target bounding box or localization input. On EvTTC, REACT achieves a 9.59% relative TTC error with 4.6 ms end-to-end inference latency, within 0.15 percentage points of the best learned method while requiring no target prior. At the dataset's mean approach speed, this latency corresponds to only 4 cm of vehicle motion, compared with 1 m for the fastest competing learned method. REACT further supports anytime TTC prediction, zero-shot transfer to a different driving sequence, and INT8 quantization, reducing the estimated energy consumption from 18.5 to 2.8 mJ per 32,768 events. These results show that event-driven spiking state-space dynamics can provide low-latency, continuously updated temporal perception for reactive robotic systems.

Chinese Translation

在动态环境中运行的机器人系统需要随着输入感知流的变化而不断演变的视觉感知。事件相机提供微秒级的时间分辨率和异步感知，但大多数基于学习的方法将事件累积到帧或时间箱中，这引入了可能限制快速反应的集成延迟。在此，我们提出了REACT，一种用于事件驱动时间感知的完全脉冲状态空间模型，它逐个处理原始事件，而不进行时间累积。REACT使用一种复值脉冲神经元C-SiLIF，其连续时间动态由物理事件间隔驱动，使其内部状态能够以单个事件的时间分辨率演变。我们在手势识别和从全场事件流中估计碰撞时间（TTC）方面评估了REACT，而无需目标边界框或定位输入。在EvTTC数据集上，REACT实现了9.59%的相对TTC误差，端到端推理延迟为4.6毫秒，且与最佳学习方法的差距仅为0.15个百分点，同时不需要目标先验。在数据集的平均接近速度下，这一延迟对应于车辆运动仅为4厘米，而最快的竞争学习方法则为1米。REACT还支持随时TTC预测、零样本迁移到不同驾驶序列，以及INT8量化，将每32,768个事件的估计能耗从18.5毫焦耳降低到2.8毫焦耳。这些结果表明，事件驱动的脉冲状态空间动态可以为反应式机器人系统提供低延迟、持续更新的时间感知。

View on arXiv Download PDF AI Translation

cs.RO / 5 / 2609.19216

4D Radar Perception Algorithms for Autonomous Driving: A Review

用于自主驾驶的4D雷达感知算法：综述

Wu, Xumin, Zhou, Jun, Mei, Jilin, Min, Chen, Hu, Yu

Abstract

Research on 4D millimeter-wave radar perception algorithms has flourished in recent years, extending from signal processing and object detection to semantic segmentation, motion estimation, occupancy prediction, and dynamic scene reconstruction. This review organizes the field according to the evolution of perception tasks and algorithms. It first introduces radar fundamentals, data representations, and quality-enhancement methods, and then reviews object-level perception, motion and localization, local and dense spatial perception, and dynamic scene understanding. Across these directions, we compare radar-only learning, multimodal fusion, and cross-modal supervision and knowledge distillation. Particular attention is paid to how elevation, Doppler measurements, and radar physical priors are exploited across tasks. We further summarize the task coverage, input data, annotations, and evaluation protocols of existing datasets, clarifying the empirical support for different research directions. Finally, we discuss the common challenges and future directions of 4D radar perception for autonomous driving. This review provides a task-oriented perspective on the transition from sparse object perception to dynamic spatial understanding.

Chinese Translation

近年来，4D毫米波雷达感知算法的研究蓬勃发展，涵盖了从信号处理和目标检测到语义分割、运动估计、占用预测和动态场景重建等多个领域。本文综述了该领域的演变，按照感知任务和算法的进展进行组织。首先介绍了雷达基础知识、数据表示和质量提升方法，然后回顾了目标级感知、运动与定位、局部和密集空间感知以及动态场景理解等方面。在这些方向上，我们比较了仅使用雷达的学习、多模态融合以及跨模态监督和知识蒸馏。特别关注的是在不同任务中如何利用高度、 Doppler 测量和雷达物理先验。我们进一步总结了现有数据集的任务覆盖、输入数据、注释和评估协议，明确了不同研究方向的实证支持。最后，我们讨论了4D雷达感知在自主驾驶中的共同挑战和未来方向。本文提供了从稀疏目标感知到动态空间理解转变的任务导向视角。

View on arXiv Download PDF AI Translation

cs.RO / 6 / 2609.19228

Grasping by interconnection: robust closing motions from coarse object templates

通过互连实现抓取：基于粗略物体模板的稳健闭合动作

Vanderheyden, Julien, Drion, Guillaume, Forni, Fulvio, Sacré, Pierre

Abstract

Dexterous robot hands must often grasp objects whose shape, size, and pose are known only approximately. Grasp planners typically require accurate object models or correct errors with feedback, but how much inaccuracy a closing motion can tolerate on its own remains unclear. To address this question, we designed a motion planner based on four principles: a coarse template of the object, human grasp types, an object-centric interaction, and compliant, sliding contacts instead of prescribed contact points. This paper presents the planner, implemented through virtual model control, and its evaluation on a Shadow Dexterous Hand. Without feedback, the planned closing motions tolerated size errors of about 1cm and pose errors of several centimeters and tens of degrees, a wider range than a state-of-the-art data-driven planner in 25 of 27 tested conditions. They also grasped 82.5% of 80 everyday objects and succeeded within an autonomous pipeline. Robustness can thus be designed into the closing motion itself, rather than left only to feedback. This planner opens a path toward reliable manipulation in uncertain settings, which we will pursue by combining it with adaptive feedback control on the physical hand.

Chinese Translation

灵巧的机器人手通常需要抓取形状、大小和姿态仅大致已知的物体。抓取规划器通常需要准确的物体模型或通过反馈来纠正误差，但闭合动作自身可以容忍多少不准确性仍不清楚。为了解决这个问题，我们设计了一种基于四个原则的运动规划器：物体的粗略模板、人类抓取类型、以物体为中心的交互，以及使用顺应性滑动接触而非预设接触点。本文展示了该规划器的实现，采用虚拟模型控制，并在Shadow Dexterous Hand上进行了评估。在没有反馈的情况下，规划的闭合动作能够容忍约1厘米的大小误差和几厘米及数十度的姿态误差，这一范围比在27种测试条件下的最先进数据驱动规划器更广泛。它们还成功抓取了80个日常物体中的82.5%，并在自主流程中取得成功。因此，稳健性可以被设计到闭合动作本身，而不仅仅依赖于反馈。该规划器为在不确定环境中实现可靠操控开辟了一条道路，我们将通过将其与物理手的自适应反馈控制相结合来进一步探索。

View on arXiv Download PDF AI Translation

cs.RO / 7 / 2609.19272

Learning Safe Humanoid Navigation from Reduced Order Models

从降阶模型学习安全的人形机器人导航

Compton, William D., Olkin, Zachary, Bena, Ryan, Ames, Aaron D.

Abstract

Research in humanoid robotics has achieved rapid progress in locomotion, and recent results have pushed the boundary on autonomous navigation. We demonstrate that a standard single-stage RL navigation pipeline struggles to scale to multi-level and multi-story terrain, limited by the difficulty of complex humanoid terrain interactions such as stairs. To overcome this challenge, we decompose the navigation problem into two pieces. First, we train a policy operating on the reduced order dynamics but with full 3D LiDAR observations to navigate complex, multi-story terrain. We then utilize this navigation knowledge to kickstart a policy operating on the full-order humanoid dynamics, with a frozen locomotion policy in the loop. Additionally, we demonstrate that applying a Poisson safety filter to the navigation policy output recovers safety in the presence of out-of-distribution obstacles, without dropping navigation success rate. We demonstrate the resulting RoM-Nav policy on a Unitree G1, accomplishing mapless multi-floor navigation covering trials with over 10m of vertical displacement and over 100m of path length. Project page with videos https://wdc3iii.github.io/rom-nav/ .

Chinese Translation

人形机器人领域的研究在运动能力方面取得了快速进展，最近的成果推动了自主导航的边界。我们证明了标准的单阶段强化学习（RL）导航管道在多层和多故事地形上难以扩展，受限于复杂的人形地形交互（如楼梯）的困难。为了解决这一挑战，我们将导航问题分解为两个部分。首先，我们训练一个在降阶动力学上操作的策略，但使用完整的3D LiDAR观测来导航复杂的多层地形。然后，我们利用这一导航知识来启动一个在全阶人形动力学上操作的策略，同时在循环中保持一个冻结的运动策略。此外，我们证明了将泊松安全过滤器应用于导航策略输出可以在存在分布外障碍物的情况下恢复安全性，而不会降低导航成功率。我们在Unitree G1上展示了得到的RoM-Nav策略，成功实现了无地图的多层导航，覆盖了超过10米的垂直位移和超过100米的路径长度的试验。项目页面及视频链接：https://wdc3iii.github.io/rom-nav/

View on arXiv Download PDF AI Translation

cs.RO / 8 / 2609.19302

OHRID-Retail: An Open Multimodal Dataset of Human Activity in Retail Environments

OHRID-Retail：零售环境中人类活动的开放多模态数据集

Wang, Xiangrui, Wu, Yuetong, Beeman, Jalen, Cook, Robert, Gu, Yu, Pearson, Nathanial, Smith, Trevor, Hayes, Read, Hu, Boyi

Abstract

Open datasets describing human behavior in environments shared with mobile robots remain limited, particularly for retail activities that combine locomotion, reaching, object handling, and robot guided movement. This paper introduces OHRID Retail, an open, human centered multimodal dataset collected from 16 healthy adults performing a simulated shelf picking task under three within participant conditions: no robot, low speed robot guidance, and high speed robot guidance. Each participant completed two trials per condition. Whole body kinematics were recorded using 17 Xsens Awinda inertial sensors and muscle activity was measured at 10 locations using Delsys Trigno surface electromyography sensors. Descriptive analyses demonstrate variation in whole body movement intensity and muscle activation across robot interaction conditions and body locations. OHRID Retail provides openly available raw recordings, processed measures, documentation, and reproducible analysis resources. The dataset can support research in human activity recognition, multimodal sensor fusion, occupational biomechanics, ergonomics, human aware robot navigation, and human robot interaction in retail and related shared environments.

Chinese Translation

描述人类在与移动机器人共享环境中的行为的开放数据集仍然有限，特别是在结合了运动、伸手、物体处理和机器人引导移动的零售活动方面。本文介绍了OHRID Retail，这是一个开放的人本多模态数据集，数据来源于16名健康成年人在三种参与者内条件下执行模拟货架挑选任务的过程：无机器人、低速机器人引导和高速机器人引导。每位参与者在每种条件下完成了两次试验。使用17个Xsens Awinda惯性传感器记录全身运动学，并在10个位置使用Delsys Trigno表面肌电传感器测量肌肉活动。描述性分析展示了在机器人交互条件和身体位置之间全身运动强度和肌肉激活的变化。OHRID Retail提供公开可用的原始记录、处理后的测量数据、文档和可重复分析资源。该数据集可以支持人类活动识别、多模态传感器融合、职业生物力学、人机工程学、人机协作导航以及零售和相关共享环境中的人机交互研究。

View on arXiv Download PDF AI Translation

cs.RO / 9 / 2609.19315

GAVEL: Graph World Models for Verified and Efficient Long-Horizon LLM Task Planning

GAVEL：用于验证和高效长时间跨度 LLM 任务规划的图世界模型

Wang, Ruiyang, Hsu, Hao-Lun, Mehta, Swarajh, Kim, Jiwoo, Dou, Zhihao, Pajic, Miroslav

Abstract

Large language models (LLMs) provide a flexible interface for long-horizon robot planning, but generated plans often fail to respect embodiment constraints, recover from planning errors, or reason effectively under partial observability. We present GAVEL, a framework for verifying and repairing long-horizon LLM planning built around an explicit graph world model. The graph represents relevant object-relations, action pre-conditions and effects, and probabilistic beliefs over unobserved object locations. This model can predict the consequences of LLM-generated actions before execution, detect violations, and repair those whose corrections follow directly from the world model. This method also reserves LLM replanning solely for errors requiring semantic reasoning. For multi-task instructions, GAVEL reasons over distributions of possible object locations to reorder remaining subtasks and minimize expected search cost. We evaluate GAVEL on BEHAVIOR-1K across 100 single long-horizon tasks and 500 multi-task instructions. With Qwen3-8B, GAVEL improves single-task success from 41.2% to 91.8% and multi-task success from 19.9% to 92.6%. Distributional belief reasoning also reduces travel distance by approximately 5.4% compared with a static variant. These improvements show that an explicit graph world model harness can substantially improve the reliability and efficiency of long-horizon embodied planning across compact and frontier hosted LLM capabilities.

Chinese Translation

大型语言模型（LLMs）为长时间跨度的机器人规划提供了灵活的接口，但生成的计划往往未能遵循实体约束、从规划错误中恢复，或在部分可观测性下有效推理。我们提出了 GAVEL，这是一个围绕显式图世界模型构建的长时间跨度 LLM 规划验证和修复框架。该图表示相关的对象关系、动作前提条件和效果，以及对未观察到的对象位置的概率信念。该模型可以在执行之前预测 LLM 生成的动作的后果，检测违规行为，并修复那些其修正直接源自世界模型的错误。该方法还将 LLM 重新规划仅保留给需要语义推理的错误。对于多任务指令，GAVEL 在可能的对象位置分布上进行推理，以重新排序剩余子任务并最小化预期搜索成本。我们在 BEHAVIOR-1K 上评估 GAVEL，涵盖 100 个单一长时间跨度任务和 500 个多任务指令。使用 Qwen3-8B，GAVEL 将单任务成功率从 41.2% 提高到 91.8%，多任务成功率从 19.9% 提高到 92.6%。分布信念推理还将旅行距离与静态变体相比减少了约 5.4%。这些改进表明，显式图世界模型的利用可以显著提高长时间跨度实体规划的可靠性和效率，适用于紧凑和前沿托管的 LLM 能力。

View on arXiv Download PDF AI Translation

cs.RO / 10 / 2609.19328

A Morphing Aerial Robot With Thruster-Integrated Flexible Continuum Links for Shape Adaptive Aerial Manipulation

一种具有推进器集成柔性连续链接的变形空中机器人，用于形状自适应的空中操控

Sawada, Eri, Sugihara, Kazuki, Miyamichi, Ayano, Kojima, Kunio, Okada, Kei

Abstract

In recent years, aerial manipulation has attracted increasing attention as a key to expand the application of aerial robots. In this work, we focus on two major research directions for achieving versatile aerial manipulation: (i) acquiring high environmental adaptability using soft manipulators, and (ii) expanding the feasible wrench space by distributing thrusters along the manipulator. However, no aerial robot has simultaneously satisfied these two requirements. Therefore, in this paper, we propose a morphing rotor-distributed aerial robot with flexible continuum links that achieves both high shape adaptability and an expanded wrench space. The flexible continuum links function as soft manipulators, passively conforming to the shape of the environment, while the thrusters distributed along the continuum links expand the feasible thrust wrench space and enable the end-effector to exert large interaction forces. To realize the proposed robot, it is essential to suppress vibrations of the lightweight continuum links. Thus, we develop a composite leaf-spring structure that provides both high torsional and vertical stiffness, and vibration-suppressing control methods. Using these implementations, we demonstrate stable flight and a variety of aerial manipulation tasks. To the best of our knowledge, this is the first work to realize aerial manipulations using flexible links with an integrated thruster.

Chinese Translation

近年来，空中操控作为扩展空中机器人应用的关键，受到了越来越多的关注。在本研究中，我们关注实现多功能空中操控的两个主要研究方向：（i）利用柔性操控器获得高环境适应性，以及（ii）通过沿操控器分布推进器扩展可行的扭矩空间。然而，目前尚无空中机器人能够同时满足这两个要求。因此，在本文中，我们提出了一种具有柔性连续链接的变形旋翼分布空中机器人，该机器人实现了高形状适应性和扩展的扭矩空间。柔性连续链接作为柔性操控器，能够被动适应环境的形状，而沿连续链接分布的推进器则扩展了可行的推力扭矩空间，使末端执行器能够施加较大的交互力。为了实现所提出的机器人，抑制轻量级连续链接的振动至关重要。因此，我们开发了一种复合叶片弹簧结构，提供高扭转和垂直刚度，以及振动抑制控制方法。通过这些实现，我们展示了稳定的飞行和多种空中操控任务。据我们所知，这是首次利用集成推进器的柔性链接实现空中操控的研究。

View on arXiv Download PDF AI Translation

cs.RO / 11 / 2609.19330

SemSafe-3DGS: Semantic Risk-Aware Active Navigation in Uncertain 3D Gaussian Splatting Maps

SemSafe-3DGS：不确定3D高斯点云图中的语义风险感知主动导航

Khass, Amirhossein Mollaei, Cosse, Athanasios, Motee, Nader

Abstract

Autonomous robots operating in partially observed environments must navigate safely while acquiring observations that improve future planning. Existing safety formulations generally reason primarily about geometry. Consequently, geometrically similar scene elements may induce comparable control responses despite having different semantic consequences. We present a semantic risk aware safe-active perception framework for navigation in attributed 3D Gaussian maps. Semantic attributes modulate an Average Value-at-Risk collision clearance model through class dependent risk weights, allowing safety-critical Gaussian primitives to receive greater influence in the composite barrier. The resulting weighted clearances are aggregated into a control barrier function, while a trajectory-relevant active perception barrier promotes observations that reduce geometric map uncertainty along the robot's anticipated motion. Both objectives are integrated in a unified CBF-QP that enforces semantic risk-aware collision avoidance as a hard constraint while relaxing information acquisition when it conflicts with safety or task progress. Experiments demonstrate efficient safety constraint, improved navigation through active perception, semantic dependent trajectory adaptation, and real-robot execution under Ackermann dynamics.

Chinese Translation

在部分观察环境中运行的自主机器人必须在安全导航的同时获取有助于未来规划的观察。现有的安全性公式通常主要关注几何特征。因此，几何上相似的场景元素可能会引发相似的控制响应，尽管它们具有不同的语义后果。我们提出了一种基于语义风险感知的安全主动感知框架，用于在带属性的3D高斯地图中进行导航。语义属性通过类别相关的风险权重调节平均风险值（Average Value-at-Risk）碰撞清除模型，使得安全关键的高斯原语在复合障碍中获得更大的影响。由此产生的加权清除值被汇总到控制障碍函数中，同时与轨迹相关的主动感知障碍促进减少机器人预期运动路径上的几何地图不确定性的观察。这两个目标在统一的CBF-QP（控制障碍函数-二次规划）中集成，强制执行语义风险感知的碰撞避免作为硬约束，同时在与安全或任务进展相冲突时放宽信息获取。实验表明，该方法有效地约束安全性，通过主动感知改善导航，基于语义的轨迹适应，以及在Ackermann动态下的真实机器人执行。

View on arXiv Download PDF AI Translation

cs.RO / 12 / 2609.19336

Dynamic-LIVO: A Dynamic-Aware LiDAR-Inertial-Visual Odometry System Using Spatio-Temporal Normals

动态-LIVO：一种基于时空法线的动态感知激光雷达-惯性-视觉里程计系统

Zhang, Zhixin, Ahiwe, Samuel, Hale, Matthew, Zhao, Liang, Ladosz, Pawel

Abstract

This paper proposes Dynamic-LIVO, a dynamic-aware LiDAR-Inertial-Visual Odometry (LIVO) system for robust state estimation and static colored mapping in dynamic environments. Dynamic-LIVO employs Spatio-Temporal (S-T) normal analysis to identify dynamic LiDAR points and propagates the resulting classification to both LiDAR-inertial and visual-inertial updates, preventing dynamic LiDAR measurements and their associated visual observations from affecting state estimation and mapping. However, S-T normal estimation can be unreliable in newly observed and spatially sparse regions due to insufficient spatio-temporal observations. To address this issue, we introduce a time-delayed S-T normal estimation strategy that defers the classification of insufficiently constrained points and re-evaluates them as additional observations become available. This strategy improves dynamic classification reliability while preserving valid static points for map construction. Extensive experiments on public and self-collected datasets with diverse sensor configurations demonstrate that Dynamic-LIVO improves localization accuracy and produces cleaner static colored maps in challenging dynamic environments. The source code and self-collected dataset will be publicly released upon acceptance.

Chinese Translation

本文提出了一种动态感知的激光雷达-惯性-视觉里程计系统（Dynamic-LIVO），用于在动态环境中进行稳健的状态估计和静态彩色地图构建。Dynamic-LIVO采用时空（S-T）法线分析来识别动态激光雷达点，并将结果分类传播到激光雷达-惯性和视觉-惯性更新中，从而防止动态激光雷达测量及其相关的视觉观测影响状态估计和地图构建。然而，由于时空观测不足，S-T法线估计在新观察到的和空间稀疏区域可能不可靠。为了解决这一问题，我们引入了一种时间延迟的S-T法线估计策略，该策略推迟对约束不足点的分类，并在额外观测可用时重新评估这些点。该策略提高了动态分类的可靠性，同时保留了有效的静态点以用于地图构建。在公共和自收集的数据集上进行的大量实验表明，Dynamic-LIVO提高了定位精度，并在具有挑战性的动态环境中生成了更干净的静态彩色地图。源代码和自收集的数据集将在论文接受后公开发布。

View on arXiv Download PDF AI Translation

cs.RO / 13 / 2609.19340

ViLoMan: Learning Visual-Proprioceptive Whole-Body Loco-Manipulation Skills for Humanoid Robots

ViLoMan：学习类人机器人视觉-本体感知的全身运动操控技能

Tian, Zejie, Hou, Ruibing, Ma, Bingpeng, Karlsson, Börje F., Shan, Shiguang

Abstract

Humanoid loco-manipulation requires adaptive whole-body coordination to seamlessly integrate locomotion and physical interaction. Despite recent advances, learning autonomous loco-manipulation remains challenging due to the scarcity of diverse, physically executable robot-object interaction data and the difficulty of learning unified whole-body control directly from onboard observations. We present ViLoMan, a scalable framework for autonomous humanoid loco-manipulation. ViLoMan first transforms partial kinematic demonstrations of human-object interactions into complete, physically executable robot trajectories. It then leverages these trajectories within a teacher-student distillation framework to learn a unified policy that maps egocentric depth observations and proprioceptive measurements directly to joint-level whole-body actions. During deployment, the policy requires neither reference motions nor intermediate commands. We evaluate ViLoMan on door-closing tasks across diverse door configurations and robot initial conditions in both simulation and the real world. Experimental results demonstrate that a single policy enables a Unitree G1 humanoid to complete the full task using only onboard depth sensing and proprioception, while generalizing robustly across task variations and transferring effectively from simulation to reality. Project page: viloman-anonymous.pages.dev.

Chinese Translation

类人机器人的运动操控需要适应性的全身协调，以无缝整合运动和物理交互。尽管近期取得了一些进展，但由于缺乏多样化的、可物理执行的机器人-物体交互数据，以及直接从机载观察中学习统一全身控制的困难，学习自主运动操控仍然具有挑战性。我们提出了ViLoMan，这是一个可扩展的自主类人运动操控框架。ViLoMan首先将人-物体交互的部分运动学示范转化为完整的、可物理执行的机器人轨迹。然后，它在教师-学生蒸馏框架内利用这些轨迹学习一个统一的策略，该策略将自我中心的深度观察和本体感知测量直接映射到关节级别的全身动作。在部署过程中，该策略既不需要参考动作，也不需要中间指令。我们在多种门配置和机器人初始条件下评估了ViLoMan在关门任务中的表现，涵盖了模拟和现实世界。实验结果表明，单一策略使Unitree G1类人机器人能够仅使用机载深度传感和本体感知完成整个任务，同时在任务变体之间具有良好的泛化能力，并能有效地从模拟转移到现实。项目页面：viloman-anonymous.pages.dev。

View on arXiv Download PDF AI Translation

cs.RO / 14 / 2609.19347

Kinematics-Grounded Agentic AI for Robotic Additive Manufacturing Process Planning

基于运动学的自主智能人工智能在机器人增材制造过程规划中的应用

Ge, Jingzhan, Chen, Ruimin, Haghighi, Azadeh, Tang, Jiong, Imani, Farhad

Abstract

Robotic additive manufacturing (AM) extends material-extrusion printing beyond gantry kinematics but makes process planning robot-dependent. A slicer-generated plan that appears favorable in part coordinates can become infeasible or robotically unfavorable on a manipulator because slicer-process decisions and part orientation determine the generated path, while part orientation and workspace placement affect its kinematic realization. Existing AM tools, large language model (LLM)-based decision-support methods, and digital-shadow systems do not provide integrated pre-execution evaluation of these coupled decisions. This paper presents agentic robotic additive manufacturing (A-RAM), an agent-specialist-tool framework that converts user intent and a part file into traceable, execution-ready plans. The LLM interprets manufacturing objectives and constraints, identifies prescribed and searchable planning variables, and encodes this reasoning in a schema-constrained request; a deterministic Planning Agent instantiates the corresponding search workflow, while domain tools compute quantitative evidence for slicing, placement, inverse kinematics, trajectory timing, Joint-6 jerk, and extrusion. The framework is evaluated on a six-axis robotic-arm AM cell through three case studies covering expert-specified planning, goal-only planning, objective-dependent infill screening, and geometry-dependent orientation-placement selection. Across the evaluated candidate sets, selected plans achieve up to 53.5% lower maximum Joint-6 jerk and 48.3% lower mean absolute Joint-6 jerk than the least favorable valid candidates, while objective-specific infill screening yields motion-plan completion times up to 40.1% shorter and extrusion paths up to 12.7% shorter than the corresponding least favorable screened patterns.

Chinese Translation

机器人增材制造（AM）将材料挤出打印扩展到超越龙门运动学的领域，但使得过程规划依赖于机器人。一个在零件坐标中看似有利的切片生成计划可能在操控器上变得不可行或不利，因为切片过程决策和零件方向决定了生成路径，而零件方向和工作空间位置则影响其运动学实现。现有的增材制造工具、基于大型语言模型（LLM）的决策支持方法和数字影子系统未能提供这些耦合决策的集成执行前评估。本文提出了自主机器人增材制造（A-RAM），一个将用户意图和零件文件转换为可追踪、准备执行计划的代理-专家-工具框架。LLM解释制造目标和约束，识别规定和可搜索的规划变量，并将这种推理编码为一个模式约束请求；一个确定性的规划代理实例化相应的搜索工作流，而领域工具则计算切片、放置、逆运动学、轨迹时序、关节6抖动和挤出等的定量证据。该框架通过三个案例研究在一个六轴机器人臂增材制造单元上进行了评估，涵盖专家指定的规划、仅目标规划、目标依赖的填充筛选和几何依赖的方向-放置选择。在评估的候选集中，所选计划在最大关节6抖动上比最不利的有效候选降低了最高53.5%，在平均绝对关节6抖动上降低了48.3%；而目标特定的填充筛选使得运动计划完成时间缩短了最高40.1%，挤出路径比相应的最不利筛选模式缩短了最高12.7%。

View on arXiv Download PDF AI Translation

cs.RO / 15 / 2609.19367

Design and Experimental Validation of a 3D Printed Torsional Series Elastic Actuator for Safe Human Robot Interaction

安全人机交互的3D打印扭转系列弹性驱动器的设计与实验验证

Pisco, Joel Hidalgo, Condo, Melissa Cobos, Miranda, Luigi, Paillacho, Dennys

Abstract

Ensuring intrinsic safety in physical human robot interaction (pHRI) is a critical requirement for social and service robots. While Series Elastic Actuators (SEAs) offer hardware based compliance, traditional metallic designs often require complex, multi part assemblies. This paper presents the design, finite element analysis (FEA), and experimental validation of a low stiffness, torsional spring for SEAs, manufactured via 3D printed thermoplastic polyurethane (TPU). The compliant element exhibits a highly linear torque deformation response (Ks = 0.066 Nm/degree), matching numerical predictions with under 3% deviation, a variance attributed to FDM structural anisotropy. To accommodate external interactions using standard position limited servomotors, a hybrid position controller with torque threshold switching was implemented. Experimental evaluations demonstrate the system ability to accurately track non stationary trajectories and safely yield to external disturbances. Furthermore, the inherent material damping of the TPU acts as a passive low pass filter, preventing high frequency oscillations during control mode transitions. The proposed architecture offers a cost effective, reliable, and easily manufacturable solution for safe pHRI.

Chinese Translation

确保物理人机交互（pHRI）的内在安全性是社会和服务机器人的一项关键要求。虽然系列弹性驱动器（SEAs）提供了基于硬件的顺应性，但传统的金属设计通常需要复杂的多部件组装。本文介绍了一种低刚度扭转弹簧的设计、有限元分析（FEA）和实验验证，该弹簧通过3D打印热塑性聚氨酯（TPU）制造。顺应元件表现出高度线性的扭矩变形响应（Ks = 0.066 Nm/度），与数值预测的偏差在3%以内，该偏差归因于FDM结构的各向异性。为了适应使用标准位置限制伺服电机的外部交互，实施了一种具有扭矩阈值切换的混合位置控制器。实验评估表明，该系统能够准确跟踪非静止轨迹，并安全地应对外部干扰。此外，TPU的固有材料阻尼作为一种被动低通滤波器，防止了控制模式切换期间的高频振荡。所提出的架构提供了一种经济高效、可靠且易于制造的解决方案，以实现安全的pHRI。

View on arXiv Download PDF AI Translation

cs.RO / 16 / 2609.19375

Enhancing the Perception of Safety and Comfort during Physical Human-Robot Handshake Interactions by Integrating Flexible Elements into a Robotic Arm

通过将柔性元件集成到机器人手臂中，增强人机握手交互中的安全感和舒适感

Hidalgo, Joel, Paillacho, Dennys, Cobos, Melissa, Miranda, Luigi

Abstract

Safety and comfort in human-robot physical in-teractions are essential aspects in the development of social technologies, where natural gestures, such as handshakes, represent a challenge due to their direct physical contact. The implementation of series elastic actuators (SEA) to absorb impacts is proposed as a design strategy that favors safer interactions. This paper presents an experimental study aimed at evaluating how the incorporation of SEAs in robotic arms influences perceived safety and the interaction experience dur-ing handshaking. The design allows a direct comparison of the effect of rigidity versus the incorporation of elastic elements, in order to identify the advantages of SEAs in improving the physical safety and social acceptance of robotic systems in everyday contexts. The experiment was carried out with 10 volunteers (6 men and 4 women), who performed two interactions with each robotic arm: one with rigid joints and the other with flexible joints using SEA. During testing, objective data on end-effector trajectories were collected, as well as subjective information through a perception survey focused on safety, naturalness, and confidence during the handshake. The survey results show increased perceptions of safety and comfort with the SEA-equipped arm, supporting its potential to facilitate safer and more socially accepted human-robot interactions.

Chinese Translation

人机物理交互中的安全性和舒适性是社会技术发展的重要方面，其中自然手势（如握手）由于其直接的身体接触而面临挑战。提出采用系列弹性驱动器（SEA）来吸收冲击作为一种设计策略，以促进更安全的交互。本文呈现了一项实验研究，旨在评估在机器人手臂中融入SEA如何影响握手过程中的感知安全性和交互体验。该设计允许直接比较刚性与弹性元件的结合对交互的影响，以识别SEA在提高机器人系统在日常环境中物理安全性和社会接受度方面的优势。实验共招募了10名志愿者（6名男性和4名女性），每位志愿者与每个机器人手臂进行了两次交互：一次是刚性关节，另一次是使用SEA的柔性关节。在测试过程中，收集了末端执行器轨迹的客观数据，以及通过专注于握手过程中的安全性、自然性和信心的感知调查收集的主观信息。调查结果显示，使用SEA装备的手臂在安全感和舒适感方面的感知有所提高，支持其促进更安全和更具社会接受度的人机交互的潜力。

View on arXiv Download PDF AI Translation

cs.RO / 17 / 2609.19378

Task-Oriented Active Learning of Residual Dynamics for Model Predictive Path Integral Control

面向任务的残差动态主动学习用于模型预测路径积分控制

Aoki, Nobuaki, Lee, Hojin, Sosnowski, Stefan, Hirche, Sandra

Abstract

Online residual learning can reduce model mismatch in predictive control, but passive data collection may fail to adequately cover states that become important later in the task. Task-agnostic active learning targets uncertain or informative regions, but information acquired in such regions does not necessarily improve task performance. This paper introduces Task-Oriented Information Acquisition (ToIA), an active-learning criterion for model predictive path integral control (MPPI) with online Gaussian process (GP) residual learning. For each sampled control sequence, ToIA estimates how much an observation obtained early in the rollout would reduce predictive uncertainty at later states on the same rollout, and weights this reduction by the rollout's relevance to the task. The score is evaluated over the existing MPPI rollout batch without sampling future observations or re-optimizing control under hypothetical posterior updates. In simulated off-road navigation across held-out maps with heterogeneous terrain, ToIA improved the goal-reaching success rate over passive GP learning by 19.3 and 27.4 percentage points and outperformed task-agnostic active-learning baselines across dense and sparse online-learning intervals. An ablation study indicates that task relevance is particularly important under sparse model updates. The implementation supports online control at 20 Hz on an NVIDIA RTX 2080 Ti.

Chinese Translation

在线残差学习可以减少预测控制中的模型不匹配，但被动数据收集可能无法充分覆盖在任务后期变得重要的状态。面向任务的主动学习针对不确定或信息丰富的区域，但在这些区域获得的信息不一定能提高任务性能。本文提出了面向任务的信息获取（Task-Oriented Information Acquisition, ToIA），这是一种用于模型预测路径积分控制（Model Predictive Path Integral Control, MPPI）与在线高斯过程（Gaussian Process, GP）残差学习的主动学习标准。对于每个采样的控制序列，ToIA 估计在回滚早期获得的观察将如何减少同一回滚后期状态的预测不确定性，并根据回滚与任务的相关性加权这种减少。该评分在现有的 MPPI 回滚批次上进行评估，而无需采样未来的观察或在假设后验更新下重新优化控制。在对异质地形的保留地图进行模拟越野导航中，ToIA 将目标达成成功率比被动 GP 学习提高了 19.3 和 27.4 个百分点，并在密集和稀疏的在线学习间隔中超越了面向任务的主动学习基线。消融研究表明，在稀疏模型更新下，任务相关性尤为重要。该实现支持在 NVIDIA RTX 2080 Ti 上以 20 Hz 进行在线控制。

View on arXiv Download PDF AI Translation

cs.RO / 18 / 2609.19413

From Rollout to Reset: A Graph-Based Harness for Autonomous Long-Horizon Manipulation Evaluation

从回放到重置：一种基于图的自主长时间操作评估工具

Jiang, Jing, Yang, Yue, Jiang, Xinkai, Bertasius, Gedas, Szafir, Daniel J., Lioutikov, Rudolf

Abstract

Robot manipulation policies are improving quickly, and real-robot evaluation remains the standard evidence for that progress. It still relies on a human to reset the scene between rollouts, which consumes operator time and leaves the initial state distribution unspecified, so results reproduce poorly. A recent system, AutoEval, automates both reset and scoring, but only for single-step tasks, because a long-horizon rollout can terminate in combinatorially many configurations that no single learned reset policy covers. We present HALTER, a Harness for Autonomous Long-horizon Task Evaluation and Reset, which restores the scene by planning over a library of learned atomic reset skills, so demonstration cost scales with the size of that library rather than with the number of terminal states. HALTER builds a spatial scene graph online from point clouds and vision foundation models, and an LLM reasons over this graph to score the rollout, plan the reset, and verify that the reset succeeded, without collecting labeled success images for any task. On four long-horizon tasks on a Franka arm, HALTER restores the scene in 76% of episodes, against 52% for AutoEval and 65% for a motion-planning reset, and it estimates the completed-skill fraction correctly in 90% of episodes, against 76%. Its reset-verification verdict is correct in 91% of episodes, compared with 78% for AutoEval. It also cuts the operator time of an evaluation campaign by 72% relative to manual reset. We further measure compositional generalization on three held-out tasks, where HALTER resets 74.7% of episodes against 1.3% for a per-task reset policy, and we ablate the scene representation and the graph update rate.

Chinese Translation

机器人操作策略正在迅速改善，而真实机器人评估仍然是这一进展的标准证据。然而，这一过程仍然依赖于人工在回放之间重置场景，这消耗了操作员的时间，并且初始状态分布未被明确，因此结果的可重复性较差。最近的系统AutoEval自动化了重置和评分，但仅适用于单步任务，因为长时间回放可能以组合方式终止于许多配置，而没有任何单一的学习重置策略能够覆盖。我们提出了HALTER（自主长时间任务评估与重置工具），它通过在学习的原子重置技能库上进行规划来恢复场景，因此演示成本与该库的大小成比例，而不是与终端状态的数量成比例。HALTER在线从点云和视觉基础模型构建空间场景图，LLM（大语言模型）在该图上进行推理，以评分回放、规划重置并验证重置是否成功，而无需为任何任务收集标记的成功图像。在四个长时间任务上，HALTER在76%的实验中成功恢复场景，而AutoEval为52%，运动规划重置为65%；它在90%的实验中正确估计完成技能的比例，而AutoEval为76%。其重置验证的正确率在91%的实验中是正确的，而AutoEval为78%。与手动重置相比，它还将评估活动的操作员时间减少了72%。我们进一步在三个保留任务上测量组合泛化，其中HALTER在74.7%的实验中成功重置，而每任务重置策略仅为1.3%；我们还对场景表示和图更新速率进行了消融实验。

View on arXiv Download PDF AI Translation

cs.RO / 19 / 2609.19441

Predict Before You Deploy: Offline Prediction of Quantization-Induced Task Degradation for World Action Models

部署前预测：世界动作模型量化引起的任务退化的离线预测

Xu, Jiuyi, Guo, Jinjia, Chen, Meida, Du, Jing, Shi, Yangming

Abstract

World action models (WAMs) rely on video-generation backbones, requiring substantial memory and compute for deployment. Post-training quantization reduces memory and can accelerate inference, but bit width, grouping, and quantizer choice define a large configuration space. Identifying configurations that preserve task performance through exhaustive closed-loop evaluation is costly. We propose PreDE (Predict Before You Deploy), a policy-calibrated framework for predicting quantization-induced task degradation from offline action deviations. Using closed-loop outcomes from a small development set, PreDE calibrates two thresholds and accepts, rejects, or defers new configurations using a fixed observation log. Under a within-setting label-ordering hypothesis, the rule issues decisions where all thresholds consistent with the development labels agree. Across five WAMs and four benchmark settings, quantization produces configuration-dependent task losses that cannot be explained by bit width alone or a shared deviation threshold. Across 28 held-out configurations from two policies, PreDE issued 21 decisions before observing closed-loop outcomes (75% coverage), all matching the observed acceptable or degraded labels. Deferred candidates included both acceptable outcomes and a 33-percentage-point loss. In 450 Franka Research 3 trials across two independently fine-tuned policies, all configurations assigned to high-deviation groups before testing showed significant degradation, while low-deviation comparisons showed no statistically significant degradation. On the real robot, W4A4 achieved a 1.37x action-query speedup and approximately 44% lower peak memory. These results support policy-specific behavioral calibration for quantization configuration selection while identifying candidates that require closed-loop evaluation. The code is available at https://github.com/jiuyixu25/PreDE.

Chinese Translation

世界动作模型（WAMs）依赖于视频生成骨干网络，部署时需要大量的内存和计算资源。后训练量化可以减少内存并加速推理，但位宽、分组和量化器选择定义了一个庞大的配置空间。通过全面的闭环评估来识别能够保持任务性能的配置是成本高昂的。我们提出了PreDE（部署前预测），这是一个政策校准框架，用于从离线动作偏差中预测量化引起的任务退化。通过使用小型开发集的闭环结果，PreDE校准两个阈值，并使用固定的观察日志接受、拒绝或推迟新的配置。在一个设定内标签排序假设下，该规则在所有与开发标签一致的阈值达成一致时做出决策。在五个WAM和四个基准设置中，量化产生的任务损失依赖于配置，无法仅通过位宽或共享偏差阈值来解释。在来自两个策略的28个保留配置中，PreDE在观察到闭环结果之前发出了21个决策（覆盖率为75%），所有决策均与观察到的可接受或退化标签相匹配。推迟的候选包括可接受的结果和33个百分点的损失。在450次Franka Research 3试验中，所有在测试前分配到高偏差组的配置显示出显著退化，而低偏差比较则没有显示出统计显著的退化。在真实机器人上，W4A4实现了1.37倍的动作查询加速和约44%的峰值内存降低。这些结果支持针对量化配置选择的政策特定行为校准，同时识别需要闭环评估的候选。代码可在 https://github.com/jiuyixu25/PreDE 获取。

View on arXiv Download PDF AI Translation

cs.RO / 20 / 2609.19447

From Wizard-of-Oz Human-Robot Dialogue Collection to a Taxonomy of Robot Response Decisions: A Retrospective Analysis of Assistive Pilot Interactions

从巫师-奥兹人机对话收集到机器人响应决策分类法：助理 pilot 交互的回顾性分析

Liu, Guangping, Hawkins, Nicholas, Sultan, Tipu, Esposito, Flavio, Dian, Madi

Abstract

Robots that follow natural-language instructions in everyday indoor environments must act on incomplete human utterances. Instructions often omit essential information, such as the identity of an out-of-view object, an intended destination, or the user's goal. Existing datasets contain little real-world situated dialogue and provide few practice-grounded criteria for deciding when a robot should act, confirm, clarify, or refuse. We retrospectively analyze a pilot Wizard-of-Oz study in which five participants performed everyday indoor tasks, including door opening, drawer opening, feeding, drinking, and cleaning, with a wheelchair-mounted mobile manipulator while the wizard responded without a formal communication policy. This preserved authentic user behavior but produced inconsistent robot-side decisions, motivating an explicit decision scheme. From 40 episodes, we derived a hierarchical taxonomy of six response modes (ANSWER, REPORT_DONE, REFUSE, CONFIRM, CLARIFY, ACT) and four ambiguity types (intent, referential, spatial, intelligibility). Two human annotators and an AI annotator applied the scheme to the pilot data. Clean-label rates were 91% and 89%, and Cohen's ranged from 0.72 to 0.95 across decision-point, mode, and ambiguity levels for both human-human and human-AI comparisons. Fine-tuning LLaVA-1.6-7B on taxonomy-derived labels for ACT and CLARIFY indicates the feasibility of training vision-language models using annotations from our taxonomy. Remaining boundary cases in decision-point identification and REPORT_DONE motivate a constrained protocol for more consistent dialogue collection.

Chinese Translation

在日常室内环境中，能够遵循自然语言指令的机器人必须在不完整的人类发言基础上采取行动。指令通常省略关键信息，例如视野外物体的身份、预期目的地或用户的目标。现有数据集包含的真实场景对话较少，且提供的实践基础标准也很有限，难以决定机器人何时应采取行动、确认、澄清或拒绝。我们对一项回顾性巫师-奥兹研究进行了分析，该研究中五名参与者与一台安装在轮椅上的移动操控器一起执行日常室内任务，包括开门、开抽屉、喂食、饮水和清洁，而巫师在没有正式沟通政策的情况下进行响应。这保留了真实的用户行为，但导致了机器人端决策的不一致，从而促使我们制定明确的决策方案。从40个情节中，我们推导出六种响应模式（回答 ANSWER、报告完成 REPORT_DONE、拒绝 REFUSE、确认 CONFIRM、澄清 CLARIFY、行动 ACT）和四种模糊性类型（意图 intent、指称 referential、空间 spatial、可理解性 intelligibility）的层次分类法。两位人类注释者和一位人工智能注释者将该方案应用于试点数据。清晰标签率为91%和89%，而Cohen's在决策点、模式和模糊性水平的比较中，范围从0.72到0.95，适用于人类-人类和人类-人工智能的比较。对基于分类法推导的标签进行LLaVA-1.6-7B的微调表明，使用我们分类法的注释训练视觉-语言模型是可行的。在决策点识别和报告完成 REPORT_DONE 中的剩余边界案例促使我们制定更一致的对话收集约束协议。

View on arXiv Download PDF AI Translation

cs.RO / 21 / 2609.19449

Winning a Won Game: Strict Reach-Avoid-Stay Control Barrier Functions for High-Dimensional Black-Box Systems

赢得已赢的游戏：高维黑箱系统的严格达成-避免-停留控制障碍函数

Oh, Donggeon David, Nguyen, Duy P., Yuan, Gongkai, Li, Qingchen, Fisac, Jaime Fernández, Hu, Haimin

Abstract

Robots must complete their tasks and maintain the achieved outcomes while avoiding safety failures at all times. Strict reach-avoid-stay (sRAS) formalizes this requirement: safely reaching a target and remaining there indefinitely after first entry. We propose an sRAS Q-control barrier function (CBF) safety filter for high-dimensional black-box systems under bounded uncertainty. Our construction combines a stay value encoding safe permanent residence in a target subset with a reach-avoid value encoding safe reachability of this subset while avoiding target states from which safe permanent residence cannot be guaranteed. We prove that these values jointly yield a valid robust discrete-time CBF and lift them to state-action Q-functions for runtime intervention. For exact values and under a measure-zero condition, our filter preserves sRAS feasibility from almost every winnable initial state and keeps the system safely within the target after first entry, against all admissible uncertainty realizations. We adopt reachability-based adversarial reinforcement learning for scalable value approximation using only black-box interactions. Notably, neither synthesis nor deployment of our filter requires known dynamics, affine structure, value derivatives, or hand-designed barriers. We validate our framework in quadruped gap jumping in simulation and hardware, where the robot crosses the gap, lands safely, and remains safe afterward. Simulated F1TENTH races further demonstrate safe overtaking and lead retention.

Chinese Translation

机器人必须完成其任务并在任何时候保持已达成的结果，同时避免安全失败。严格达成-避免-停留（sRAS）形式化了这一要求：安全地到达目标并在首次进入后无限期停留在该处。我们提出了一种针对高维黑箱系统的sRAS Q-控制障碍函数（CBF）安全过滤器，该系统在有限不确定性下运行。我们的构造结合了一个停留值，用于编码在目标子集中的安全永久居留，以及一个达成-避免值，用于编码在避免无法保证安全永久居留的目标状态的同时安全到达该子集。我们证明这些值共同产生一个有效的稳健离散时间CBF，并将其提升为状态-动作Q函数以进行运行时干预。在精确值和零测度条件下，我们的过滤器从几乎每个可赢得的初始状态中保持sRAS可行性，并在首次进入后将系统安全地保持在目标内，抵御所有可接受的不确定性实现。我们采用基于可达性的对抗性强化学习进行可扩展值近似，仅使用黑箱交互。值得注意的是，我们的过滤器的合成和部署均不需要已知的动态、仿射结构、值导数或手工设计的障碍。我们在四足动物跨越间隙的仿真和硬件中验证了我们的框架，机器人成功跨越间隙、安全着陆并在此后保持安全。模拟的F1TENTH比赛进一步展示了安全超车和领先保持。

View on arXiv Download PDF AI Translation

cs.RO / 22 / 2609.19452

GLAMDRING: Gait Learning And Morphology co-Design via Reinforcement LearnING of CPGs

GLAMDRING：通过对中央模式发生器（CPGs）的强化学习实现步态学习与形态共同设计

Joshi, Amogh, Roy, Kaushik

Abstract

Robots are moving out of the structured factory floor and into unstructured environments such as disaster sites, planetary surfaces, and agricultural fields, for which the right robot often does not yet exist. We present GLAMDRING, a framework that synthesizes the optimal robot for a locomotion task and, jointly, learns the controller that drives it. For the given specifications of forward-velocity bounds, a per-actuator power budget, an actuator library, and a payload requirement, GLAMDRING returns a matched quadruped morphology (link geometry and per-joint actuators) and a Hopf-oscillator Central Pattern Generator (CPG) gait policy. We rank feasible designs against a target design objective, viz., maximum speed, minimum Cost of Transport (CoT), or max Payload Margin. Because body and locomotion are coupled, the optimal morphology dictates how a robot is driven, while optimal gait depends on the physical body. We train a small number of CPG policies by reinforcement learning across the space of candidate morphologies, co-learning the gait with the underlying robot hardware. Link lengths and actuators are then resolved post-hoc from the policy's logged operating envelope, reducing synthesis cost to a small, fixed number of reinforcement-learning runs instead of one per candidate. Our experiments show three key findings: co-designing body and gait is necessary to satisfy locomotion constraints; actuator-envelope feasibility, rather than locomotion success alone, determines realizable payload capacity; and canonical animal gaits emerge naturally in most designs from morphology and constraints alone. A real-world demonstration further highlights the efficacy of our work.

Chinese Translation

机器人正在从结构化的工厂车间走向灾难现场、行星表面和农业田野等非结构化环境，而适合的机器人往往尚未存在。我们提出了GLAMDRING，一个框架，用于合成适用于特定运动任务的最佳机器人，并共同学习驱动它的控制器。根据给定的前进速度限制、每个执行器的功率预算、执行器库和有效载荷要求，GLAMDRING返回一个匹配的四足形态（连杆几何和每个关节的执行器）以及一个霍普夫振荡器中央模式发生器（CPG）步态策略。我们根据目标设计目标对可行设计进行排名，即最大速度、最小运输成本（CoT）或最大有效载荷裕度。由于身体与运动是耦合的，最佳形态决定了机器人如何被驱动，而最佳步态则依赖于物理身体。我们通过强化学习在候选形态空间中训练少量CPG策略，与基础机器人硬件共同学习步态。然后，从策略的记录操作范围中后续确定连杆长度和执行器，将合成成本降低到少量固定的强化学习运行，而不是每个候选都进行一次。我们的实验显示了三个关键发现：共同设计身体和步态是满足运动约束的必要条件；执行器范围的可行性，而不仅仅是运动成功，决定了可实现的有效载荷能力；而典型动物步态在大多数设计中自然地从形态和约束中出现。一个现实世界的演示进一步突显了我们工作的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 23 / 2609.19460

Pose-aware Legged Robot Semantic Exploration with Omnidirectional Perception in Confined Unknown Environments

姿态感知的四足机器人在未知狭窄环境中的语义探索与全向感知

Zhan, Xiaoyang, Chen, Shiyu, Shimada, Kenji

Abstract

Semantic exploration in confined environments requires both environment mapping and detailed observation of target objects. For ground robots, limited sensor vertical fields of view and restricted standoff distances can leave upper object surfaces unobserved from planar viewpoints. Body tilting can improve coverage, but additional observations and posture transitions increase mission time. To address this trade-off, we present POSE, a pose-aware semantic exploration system that exploits a legged robot's intrinsic body pitch and roll with omnidirectional camera-LiDAR perception. The proposed pose-aware viewpoint sampling module selects body postures from partial object maps according to expected coverage gain, while aim-aligned execution reduces unnecessary body reorientation. Further, we introduce an object-centric viewpoint pruning strategy assisted by a vision-language model (VLM), which uses persistent observation history and bird's-eye-view (BEV) maps to reduce redundant inspection visits. The resulting semantic viewpoints are combined with geometric exploration viewpoints in a global exploration planner. Simulations show that POSE improves final target-surface coverage by 8-10 percentage points over the planar planning baseline while reducing exploration time by 17-32%, and achieves the highest mean object coverage AUC among the evaluated baselines. Real-world experiments with a legged robot carrying an omnidirectional camera-LiDAR suite in a machine shop further demonstrate the system's applicability. These results support adaptive body-posture planning for improving the coverage-efficiency trade-off in legged robot semantic exploration. We plan to release the code for community benefit in the future.

Chinese Translation

在狭窄环境中的语义探索需要环境映射和对目标物体的详细观察。对于地面机器人而言，有限的传感器垂直视场和受限的观察距离可能导致平面视角下无法观察到上方物体表面。身体倾斜可以改善覆盖范围，但额外的观察和姿态转换会增加任务时间。为了解决这一权衡，我们提出了POSE，一个姿态感知的语义探索系统，利用四足机器人的内在身体俯仰和横滚与全向相机-激光雷达感知相结合。所提出的姿态感知视角采样模块根据预期的覆盖增益从部分物体地图中选择身体姿态，而目标对齐的执行则减少了不必要的身体重新定向。此外，我们引入了一种以物体为中心的视角修剪策略，该策略由视觉-语言模型（VLM）辅助，利用持久的观察历史和鸟瞰图（BEV）地图来减少冗余的检查访问。最终生成的语义视角与几何探索视角在全局探索规划器中结合。模拟结果表明，POSE在最终目标表面覆盖率上比平面规划基线提高了8-10个百分点，同时探索时间减少了17-32%，并在评估的基线中实现了最高的平均物体覆盖AUC。在一个携带全向相机-激光雷达套件的四足机器人在机器车间进行的实际实验进一步证明了该系统的适用性。这些结果支持了适应性身体姿态规划，以改善四足机器人语义探索中的覆盖效率权衡。我们计划在未来发布代码以造福社区。

View on arXiv Download PDF AI Translation

cs.RO / 24 / 2609.19475

FASA: Feedback-Aware Sampling Adaptation for Efficient Diffusion-Based VLA Models

FASA：面向反馈的采样适应用于高效的基于扩散的视觉-语言-动作模型

Han, Yuchen, Wu, Jianhan, Qu, Xiaoyang, Kong, Lingwei, Li, Shiyi, Wang, Jianzong

Abstract

Diffusion-based Vision-Language-Action (VLA) models achieve strong performance in embodied tasks, but their iterative sampling imposes heavy computational and memory-access cost, blocking real-time deployment on edge platforms. Existing acceleration methods either require expensive training (e.g., distillation, flow matching) or degrade perception via statically scheduled pruning and caching, ignoring the dynamic workload variance of robotic interactions. This paper presents FASA (Feedback-Aware Sampling Adaptation), a training-free runtime framework that treats real-time multimodal feedback as a control signal for the denoising pipeline: an interaction-driven range adaptor modulates the global sampling-step budget based on visual and gripper-force feedback, and a proprioception-aware step adaptor pinpoints the optimized step within the adapted range. This co-designed framework allows the underlying hardware architecture to adaptively match the workload demands of different execution phases. Comparative evaluations across several benchmarks show that the inference speed can be increased by up to 1.45$\times$ while maintaining competitive success rates, providing a novel dynamic runtime architecture paradigm for deploying heavy generative embodied AI workloads onto resource-constrained computing platforms.

Chinese Translation

基于扩散的视觉-语言-动作（VLA）模型在具身任务中表现出色，但其迭代采样带来了巨大的计算和内存访问成本，阻碍了在边缘平台上的实时部署。现有的加速方法要么需要昂贵的训练（例如，蒸馏、流匹配），要么通过静态调度的剪枝和缓存降低感知性能，忽视了机器人交互中的动态工作负载变化。本文提出了FASA（面向反馈的采样适应），一种无训练的运行时框架，将实时多模态反馈视为去噪流程的控制信号：一个基于交互的范围适配器根据视觉和夹持力反馈调节全局采样步长预算，而一个感知自身状态的步长适配器则在适配范围内精确定位优化步长。该共同设计的框架使得底层硬件架构能够自适应地匹配不同执行阶段的工作负载需求。多个基准的比较评估表明，在保持竞争性成功率的同时，推理速度可提高至1.45倍，为在资源受限的计算平台上部署重型生成具身人工智能工作负载提供了一种新的动态运行时架构范式。

View on arXiv Download PDF AI Translation

cs.RO / 25 / 2609.19480

Self-excited actuation enables adaptive and resilient flapping-wing flight

自激驱动实现自适应和韧性的拍打翼飞行

Yang, Rundong, Wold, Ethan S., Liu, Ellen, Lynch, James, Zhou, Wei, Jankauski, Mark, Sponberg, Simon, Gravish, Nick

Abstract

The muscles that power insect flight fall into one of two categories: 1) synchronous muscles that contract under direct control from the nervous system, and 2) asynchronous muscles which have an intrinsic stretch activation response that spontaneously generates wingbeats without the need for signaling from the brain. It is thought that the emergent nature of asynchronous wingbeats provides both adaptive and responsive capabilities for flight control. To date, most flying robots use synchronous actuation. In this paper we develop the first flight-capable flapping wing robot that uses asynchronous actuation. We demonstrate that asynchronous actuation allows wings to respond to changes in the resonant mechanics of the body without control input, and wings can react instantaneously to collisions with obstacles with no extrinsic sensing needed. Flight tests within cluttered environments demonstrate that asynchronous actuation significantly improves stability and performance when compared to synchronous actuation. In total this work demonstrates that a flapping wing robot actuation strategy that emulates the asynchronous muscles of flying insects can provide fast, reactive actuation responses before a control system would need to intervene. This partitioning of embodied control to both the low-level actuation dynamics and and high-level sensorimotor system provides a compelling blueprint for new flying robots.

Chinese Translation

驱动昆虫飞行的肌肉可分为两类：1）同步肌肉，在神经系统的直接控制下收缩；2）非同步肌肉，具有内在的拉伸激活反应，能够自发产生翅膀拍动，无需大脑的信号。人们认为，非同步翅膀拍动的涌现特性为飞行控制提供了自适应和响应能力。迄今为止，大多数飞行机器人使用同步驱动。本文开发了首个具备飞行能力的拍打翼机器人，该机器人采用非同步驱动。我们证明，非同步驱动使得翅膀能够在没有控制输入的情况下响应机体共振力学的变化，并且翅膀能够在与障碍物碰撞时瞬时反应，无需外部传感。复杂环境中的飞行测试表明，与同步驱动相比，非同步驱动显著提高了稳定性和性能。总体而言，这项工作表明，模仿飞行昆虫非同步肌肉的拍打翼机器人驱动策略能够在控制系统介入之前提供快速、反应灵敏的驱动响应。这种将具身控制划分为低级驱动动力学和高级传感运动系统的方式，为新型飞行机器人提供了一个引人注目的蓝图。

View on arXiv Download PDF AI Translation

cs.RO / 26 / 2609.19510

PIVOT: Perception-aware Independent Viewpoint Online Optimization

PIVOT：感知感知独立视点在线优化

Chen, Yuyang, Sadeghi, Shekoufeh, Adhivarahan, Charuvahan, Lemos, Elton, Wang, Chen, Koppal, Sanjeev J., Dantu, Karthik

Abstract

A fundamental assumption in robotic perception is that the sensor's field of view (FoV) is fixed relative to the robot body. Motion-decoupled sensors, such as gimbal-mounted cameras and MEMS-based LiDARs, instead allow sensing direction to be controlled independently at runtime. This freedom creates a computational challenge: efficiently selecting useful viewing directions online in feature-dense environments. We propose PIVOT, a lightweight iterative method that optimizes sensor viewing direction along a fixed translation trajectory to maximize feature visibility. Under a conical FoV model, visibility depends only on the optical axis, yielding a two-degree-of-freedom optimization on the viewing sphere $S^2$. Coordinate-free $SO(3)$ exponential-map updates enable efficient continuous optimization without explicit angular parameterizations or exhaustive viewing-sphere search. Monte Carlo evaluations retain 98.1--99.6% of brute-force visibility with a 76--85x speedup. Photorealistic simulation and real-world experiments further demonstrate improved visual localization robustness and practical viewpoint control on a quadruped robot.

Chinese Translation

机器人感知的一个基本假设是传感器的视场（FoV）相对于机器人主体是固定的。然而，运动解耦传感器，如云台安装的摄像头和基于MEMS的激光雷达，允许在运行时独立控制感知方向。这种自由度带来了一个计算挑战：在特征密集的环境中高效选择有用的观察方向。我们提出了PIVOT，一种轻量级的迭代方法，沿着固定的平移轨迹优化传感器的观察方向，以最大化特征可见性。在锥形视场模型下，可见性仅依赖于光轴，从而在观察球面$S^2$上实现了两个自由度的优化。无坐标的$SO(3)$指数映射更新使得在没有显式角度参数化或穷举观察球搜索的情况下实现高效的连续优化。蒙特卡罗评估保留了98.1%至99.6%的暴力可见性，同时实现了76至85倍的加速。光真实感模拟和实际实验进一步证明了在四足机器人上改进的视觉定位鲁棒性和实用的视点控制。

View on arXiv Download PDF AI Translation

cs.RO / 27 / 2609.19512

CoreSense: Traceable Failure Recall and Conflict-Aware Belief Gating for Auditable Robot Decisions

CoreSense：可追溯的失败回忆与冲突感知的信念门控用于可审计的机器人决策

Li, Zoe

Abstract

Robots can recall prior failures without knowing whether recalled evidence remains valid, conflicts with current observations, or is sufficient to guide a decision. We present CoreSense, a robot-system integration architecture that combines traceable episodic evidence with a conflict-aware belief gate and bounded, auditable recommendations. The gate checks scope, provenance, time, contradiction, and support before it permits PROCEED, requests re-observation, abstains, or escalates. Evaluation follows three complementary layers without commanding a physical robot: offline public real-robot data, a frozen signal-level simulation, and a live cloud deployment path. On CableTrace-120 and BotFails-200, belief gating reduces protocol-defined unsafe proceeds from 20% and 40% to 0%. A disjointly calibrated raw-video policy also reaches 0% unsafe proceed, but overblocks every nominal episode. On public data, a ViFailback-BotFails visual detector reaches 0.778 AUROC yet remains all-blocking, whereas cycle-disjoint UR3 telemetry for protective stops yields 0% unsafe proceed, 36.1% overblocking, and 61.9% coverage; grip-loss transfer remains a negative result. Controlled physical corroboration yields 3.3%, 0%, and 42.0%, while conflict-aware fusion yields 4.7%, 0%, and 42.8%. Finally, 20/20 cloud recalls validate a CockroachDB Cloud-Amazon Bedrock deployment path. The evidence supports an auditable integration pattern, not autonomous recovery or certified safety.

Chinese Translation

机器人可以回忆先前的失败，而无需知道回忆的证据是否仍然有效、是否与当前观察相冲突或是否足以指导决策。我们提出了CoreSense，一种机器人系统集成架构，结合了可追溯的情节证据、冲突感知的信念门控和有限的可审计推荐。该门控在允许继续（PROCEED）、请求重新观察、放弃或升级之前，检查范围、来源、时间、矛盾和支持。评估遵循三个互补层次，而不指挥物理机器人：离线公共真实机器人数据、冻结的信号级仿真和实时云部署路径。在CableTrace-120和BotFails-200上，信念门控将协议定义的不安全继续率从20%和40%降低至0%。一个独立校准的原始视频策略也达到了0%的不安全继续率，但对每个正常事件进行了过度阻止。在公共数据上，ViFailback-BotFails视觉检测器达到了0.778的AUROC，但仍然是全阻止的，而循环不重叠的UR3遥测用于保护停止则产生了0%的不安全继续率、36.1%的过度阻止和61.9%的覆盖率；抓握丢失转移仍然是一个负结果。受控的物理验证产生了3.3%、0%和42.0%的结果，而冲突感知融合则产生了4.7%、0%和42.8%的结果。最后，20/20的云回忆验证了CockroachDB Cloud-Amazon Bedrock的部署路径。证据支持一种可审计的集成模式，而不是自主恢复或认证安全。

View on arXiv Download PDF AI Translation

cs.RO / 28 / 2609.19527

AURORA: A Natural Language-Driven Agentic Framework for Understanding, Reasoning, and Orchestrating Reliable Air-Ground Co-Simulation

AURORA：一个基于自然语言的自主框架，用于理解、推理和协调可靠的空地协同仿真

Wu, Keshu, Zhang, Hao, Gan, Rui, Gao, Xiangbo, Li, Xiaopeng, Tu, Zhengzhong, Zhou, Yang

Abstract

Air-ground transportation research increasingly relies on co-simulation, yet constructing scenarios remains labor-intensive and difficult to validate. More importantly, a generated scenario may execute successfully while failing to realize the spatial, temporal, communication, or behavioral relationships requested by the user. This paper presents AURORA, a natural-language-driven agentic framework that treats air-ground scenario generation as a process of compilation with verification. Central to AURORA is the Air-Ground Scenario Graph (AGSG), a typed intermediate representation that explicitly connects agents, aerial missions, events, communication links, success conditions, and their cross-domain dependencies. This shared representation enables simulator-grounded parsing, joint road-airspace grounding, temporal planning, pre-execution feasibility checking, trace-based runtime verification, failure localization, and bounded repair within a unified workflow. We further introduce AURORA-Bench to evaluate not only whether generated scenarios execute, but whether they faithfully realize the requested interactions. Experiments across multiple language models show that structured execution substantially improves reliability, while runtime verification exposes silent failures that completion-based evaluation overlooks. Localized repair further resolves many violations without regenerating the entire scenario. The results show that reliable scenario generation requires verifying realized behavior, not merely executable code, and demonstrate the value of explicit intermediate representations for verifiable and repairable language-driven co-simulation.

Chinese Translation

空地运输研究日益依赖于协同仿真，但构建场景仍然劳动密集且难以验证。更重要的是，生成的场景可能成功执行，但未能实现用户请求的空间、时间、通信或行为关系。本文提出了AURORA，一个基于自然语言的自主框架，将空地场景生成视为一个编译与验证的过程。AURORA的核心是空地场景图（Air-Ground Scenario Graph，AGSG），这是一种类型化的中间表示，明确连接代理、空中任务、事件、通信链路、成功条件及其跨域依赖关系。这种共享表示使得模拟器基础解析、联合路空域基础、时间规划、执行前可行性检查、基于追踪的运行时验证、故障定位和有界修复能够在统一的工作流程中进行。我们进一步引入AURORA-Bench，以评估生成的场景不仅能否执行，还能否忠实实现请求的交互。多个语言模型的实验表明，结构化执行显著提高了可靠性，而运行时验证揭示了基于完成的评估所忽视的隐性故障。局部修复进一步解决了许多违规问题，而无需重新生成整个场景。结果表明，可靠的场景生成需要验证实现的行为，而不仅仅是可执行的代码，并展示了显式中间表示在可验证和可修复的语言驱动协同仿真中的价值。

View on arXiv Download PDF AI Translation

cs.RO / 29 / 2609.19533

SLAMSqueezeBench: Comparing SLAM Systems under Resource Constraints

SLAMSqueezeBench：在资源约束下比较SLAM系统

Hefny, Mohamed, Dantu, Karthik, Ko, Steven Y.

Abstract

Simultaneous localization and mapping (SLAM) is one of the services running on an autonomous robot. It is typically run to assist other tasks such as planning, manipulation, etc. All these tasks are run on edge hardware and are subject to severe resource constraints. However, most SLAM systems are built and tested in isolation, and their performance is reported as if they are the only task running on a system. We observe that existing benchmarks lack a common mechanism for comparing SLAM systems under realistic resource constraints. To address this limitation, we have developed SLAMSqueezeBench, a framework that allows testing of SLAM systems under realistic workloads on edge hardware. It does so by imposing constraints on compute and memory resources available for the SLAM system during execution. It also simulates realistic camera frame acquisition with frame drops when a finite buffer is full. Using SLAMSqueezeBench, we compare nine SLAM systems spanning classical systems, learning-based systems, and approaches for Gaussian splatting. Our testing framework will be available for use by the community upon publication.

Chinese Translation

同时定位与地图构建（SLAM）是自主机器人运行的服务之一。它通常用于辅助规划、操作等其他任务。这些任务都在边缘硬件上运行，并受到严重的资源限制。然而，大多数SLAM系统是在孤立环境中构建和测试的，其性能报告通常假设它们是系统中唯一运行的任务。我们观察到，现有基准测试缺乏在现实资源约束下比较SLAM系统的共同机制。为了解决这一局限性，我们开发了SLAMSqueezeBench，一个允许在边缘硬件上以现实工作负载测试SLAM系统的框架。该框架通过对SLAM系统执行期间可用的计算和内存资源施加约束来实现这一点。同时，它还模拟了现实的相机帧获取，当有限的缓冲区满时会出现帧丢失。使用SLAMSqueezeBench，我们比较了九个SLAM系统，包括经典系统、基于学习的系统以及高斯点云（Gaussian splatting）的方法。我们的测试框架将在出版后向社区开放。

View on arXiv Download PDF AI Translation

cs.RO / 30 / 2609.19541

Navigate or Relocate? Planning Among Movable Obstacles in Unknown Environments

导航还是重新定位？在未知环境中移动障碍物的规划

Zhang, Yuqing, Zhu, Haoyu, Kantaros, Yiannis

Abstract

Conventional robot planning methods seek collision-free paths to a goal but fail when all paths are blocked. In these cases, the robot must determine which objects to relocate, in what order, and where to place them to clear a path---a problem known as Navigation Among Movable Obstacles (NAMO). Most NAMO planners assume a known environment, while existing approaches for unknown environments typically reason locally about relocations and cannot plan interdependent relocation sequences. We consider NAMO in unknown environments revealed through onboard sensing, where the robot must decide whether a blocked route requires relocation or a feasible path may exist through unexplored space. We propose an online framework that addresses this ambiguity by selecting between navigation and relocation using shortest paths that treat discovered movable objects as obstacles or as removable. Navigation relies on existing motion planners, while relocation uses a sampling-based approach that, unlike existing approaches for unknown environments, searches over \textit{interdependent} relocation sequences and uses an LLM to bias sampling. Numerical experiments demonstrate scalability to cluttered environments requiring interdependent relocations and improved plan quality over existing baselines.

Chinese Translation

传统的机器人规划方法寻求无碰撞路径以达到目标，但在所有路径被阻塞时会失效。在这些情况下，机器人必须确定哪些物体需要重新定位、以何种顺序进行以及将其放置在哪里以清除路径——这一问题被称为在可移动障碍物中导航（Navigation Among Movable Obstacles, NAMO）。大多数NAMO规划器假设环境是已知的，而现有的针对未知环境的方法通常局限于局部推理重新定位，无法规划相互依赖的重新定位序列。我们考虑在通过机载传感器揭示的未知环境中的NAMO，其中机器人必须决定被阻塞的路线是否需要重新定位，或者是否可能通过未探索的空间存在可行路径。我们提出了一种在线框架，通过使用最短路径在将发现的可移动物体视为障碍物或可移除物体之间进行选择，从而解决这一模糊性。导航依赖于现有的运动规划器，而重新定位则采用基于采样的方法，与现有的针对未知环境的方法不同，它在 extit{相互依赖}的重新定位序列上进行搜索，并使用大型语言模型（LLM）来偏向采样。数值实验表明，该方法在需要相互依赖重新定位的杂乱环境中具有良好的可扩展性，并且在计划质量上优于现有基准。

View on arXiv Download PDF AI Translation

cs.RO / 31 / 2609.19554

VABench: Measuring Embodied Spatial Intelligence through Visual Demonstrations, Active Perception, and Metric Control

VABench：通过视觉演示、主动感知和度量控制测量具身空间智能

Zhang, Zhongbo, Jin, Jiayi, Wang, Yifan, Zhang, Zaibin, Diao, Haiwen, Wang, Lijun, Lu, Huchuan

Abstract

Spatial intelligence requires more than describing object locations. Under incomplete observation, models must identify and acquire missing evidence, interpret it in a common spatial frame, and act on it. We introduce VA-Bench to evaluate the complete observe-reason-act-revise loop. General-purpose MLLMs learn procedural context from RGB-only demonstrations, actively select camera viewpoints, issue metric Cartesian commands, and revise them from execution feedback. Models receive no privileged object poses, oracle trajectories, or learned action heads. A fixed model-agnostic controller executes only model-specified targets. VA-Bench contains 14 base task families (11 single-arm and three dual-arm), seven held-out geometry/layout variants, and a long-horizon five-object composition track. We evaluate 12 primary model conditions in three independent runs over the same 20 physically verified seeds per base task, reporting terminal success, nine trajectory-level behavioral diagnostics, and subtask progress. First, the best-performing model scores 100.0% on target localization and 78.9% on spatial relations in the annotated run. Its three-run macro-average task success is only 53.93+/-3.17%. Second, active camera control significantly improves task success over passive multi-view observation. In one matched comparison, success rises from 27.86% to 57.50%. Third, held-out geometric transfer can reduce task success by over 30 percentage points. No model completes a strict long-horizon episode, despite substantial partial progress. VA-Bench thus tests whether general-purpose MLLMs can turn visual demonstrations and actively acquired evidence into successful embodied action.

Chinese Translation

空间智能不仅仅需要描述物体位置。在不完全观察的情况下，模型必须识别和获取缺失的证据，将其在共同的空间框架中进行解释，并据此采取行动。我们引入了VA-Bench来评估完整的观察-推理-行动-修正循环。通用的多模态学习模型（MLLMs）从仅包含RGB的演示中学习程序上下文，主动选择相机视角，发出度量笛卡尔命令，并根据执行反馈进行修正。模型不接收特权的物体姿态、神谕轨迹或学习到的动作头。一个固定的模型无关控制器仅执行模型指定的目标。VA-Bench包含14个基础任务家族（11个单臂和3个双臂），七个保留的几何/布局变体，以及一个长时间跨度的五物体组合轨道。我们在三次独立运行中评估12个主要模型条件，每个基础任务使用相同的20个经过物理验证的种子，报告终端成功率、九个轨迹级行为诊断和子任务进展。首先，表现最佳的模型在目标定位上得分100.0%，在空间关系上得分78.9%（在注释运行中）。其三次运行的宏平均任务成功率仅为53.93±3.17%。其次，主动相机控制显著提高了任务成功率，相较于被动多视角观察，成功率从27.86%上升至57.50%。第三，保留的几何转移可能导致任务成功率下降超过30个百分点。尽管有显著的部分进展，没有模型能够完成严格的长时间跨度的情节。因此，VA-Bench测试了通用的多模态学习模型是否能够将视觉演示和主动获取的证据转化为成功的具身行动。

View on arXiv Download PDF AI Translation

cs.RO / 32 / 2609.19577

Tele-Traversability: Rethinking Traversability for Teleoperated Ground Robots in Terrain Navigation

远程可通行性：重新思考地面机器人在地形导航中的可通行性

Feng, Lewei, Chen, Qi, Wang, Wenshuo, Meng, Xianghao, Kieu, Minh, Guan, Haijie, Wu, Jin, Sun, Fuchun, Xi, Junqiang

Abstract

Teleoperation, a human-in-the-loop control scheme, allows a human operator to remotely command and guide a mobile robot to navigate in off-road environments, yet fluent and user-friendly tele-navigation requires an alignment of traversability evaluation between human and robot. In the teleoperation system, the human operator typically utilizes off-site incomplete and delayed feedback via a human-machine interface to make a judgment of traversability, while the robot makes such an evaluation based on in situ onboard sensory information, which could cause divergent traversability estimation and thus generate mismatched decisions and actions. Existing approaches for traversability modeling, estimation, and prediction are mainly derived from the view of robots, i.e., robot-centric, and are practically suitable for fully autonomous mobile robots, but neglect the influence of human operators. To address this problem, this paper extends the concept of traversability from robot-centric to human-centric by accounting for the operator's cognitive states, such as attention, workload, and risk tolerance or awareness, termed tele-traversability. We first revisit the definitions and roles of traversability in robotics and then extend them to teleoperation settings. Finally, we highlight future trends and open challenges of tele-traversability toward human-centric teleoperation systems.

Chinese Translation

远程操作是一种人机协作控制方案，允许人类操作员远程指挥和引导移动机器人在越野环境中导航。然而，流畅且用户友好的远程导航需要人类与机器人之间的可通行性评估的一致性。在远程操作系统中，人类操作员通常通过人机界面利用离线的不完整和延迟反馈来判断可通行性，而机器人则基于现场的传感器信息进行评估，这可能导致可通行性估计的差异，从而产生不匹配的决策和行动。现有的可通行性建模、估计和预测方法主要是从机器人的角度出发，即以机器人为中心，实际上适合完全自主的移动机器人，但忽视了人类操作员的影响。为了解决这个问题，本文将可通行性的概念从以机器人为中心扩展到以人为中心，考虑操作员的认知状态，如注意力、工作负荷和风险容忍度或意识，称为远程可通行性（tele-traversability）。我们首先重新审视可通行性在机器人学中的定义和角色，然后将其扩展到远程操作的环境中。最后，我们强调了远程可通行性在以人为中心的远程操作系统中的未来趋势和开放挑战。

View on arXiv Download PDF AI Translation

cs.RO / 33 / 2609.19579

Recovering Aggressively Pruned Vision-Language-Action Models with Offline Hidden-State Distillation

通过离线隐状态蒸馏恢复激进剪枝的视觉-语言-动作模型

Kim, Chiyoung, Choi, Sanghyuk Roy, Lee, Minhyeok

Abstract

Vision-language-action (VLA) models let robots follow language instructions, but their language backbones of several billion parameters are the main obstacle to running them on robot hardware. Structured pruning reduces that backbone, and removing 63% of it from OpenVLA-OFT drops LIBERO-Long success from 93.2% to 0.8%. A recent approach restores such a model with supervised fine-tuning followed by reinforcement learning, which needs online rollouts and hundreds of GPU-hours. We recover most of the lost success entirely offline. Width pruning narrows the blocks but keeps the residual stream at its original size, so teacher and student hidden states have the same shape and are matched directly, without a projector. Training against a cache built in one teacher pass lifts the 63%-reduced student to within 3.5 points of the teacher in about 8 GPU-hours. A sweep over nine ratios locates where the recovery objective starts to matter. Up to 45% reduction the two do not differ significantly on OpenVLA-OFT. Hidden-state distillation then adds +2.1 to +4.5 points there between 63% and 87%, and +9.4 to +22.1 points on CogACT from 63% onward. At 81% on CogACT, a tripled recovery budget narrows the distilled student's gap to the teacher to 3.9 points on average, while supervised recovery stays more than 20 points below. At matched compression, width pruning yields higher success and depth pruning lower latency. On a 6-DoF manipulator, the distilled student at 72% reduction reaches 77.5% success against 59.5% for supervised recovery, runs 2.23x faster on-board than the teacher, and uses 62% less memory.

Chinese Translation

视觉-语言-动作（VLA）模型使机器人能够遵循语言指令，但其数十亿参数的语言骨干是运行在机器人硬件上的主要障碍。结构化剪枝减少了该骨干，从 OpenVLA-OFT 中去除 63% 的参数使 LIBERO-Long 的成功率从 93.2% 降至 0.8%。一种最近的方法通过监督微调和随后进行的强化学习恢复该模型，但这种方法需要在线回放和数百个 GPU 小时。我们在完全离线的情况下恢复了大部分丢失的成功率。宽度剪枝缩小了模块，但保持残差流的原始大小，因此教师和学生的隐状态具有相同的形状，并且可以直接匹配，而无需投影器。针对在一次教师传递中构建的缓存进行训练，使得减少 63% 的学生在大约 8 个 GPU 小时内成功率提升至与教师相差仅 3.5 分。对九个比例的遍历确定了恢复目标开始显著影响的地方。在减少 45% 的情况下，两者在 OpenVLA-OFT 上没有显著差异。隐状态蒸馏在 63% 到 87% 之间增加了 2.1 到 4.5 分，而在 CogACT 上从 63% 开始增加了 9.4 到 22.1 分。在 CogACT 上达到 81% 时，三倍的恢复预算将蒸馏学生与教师的差距缩小至平均 3.9 分，而监督恢复仍低于 20 分。在相同压缩比下，宽度剪枝实现了更高的成功率，而深度剪枝则降低了延迟。在一个 6 自由度的操纵器上，蒸馏学生在 72% 的参数减少下成功率达到 77.5%，而监督恢复为 59.5%，在板载运行速度上比教师快 2.23 倍，并且内存使用减少 62%。

View on arXiv Download PDF AI Translation

cs.RO / 34 / 2609.19582

OmniCalib: Target-Free, Task-Structured Self-Calibration for Humanoid Robots

OmniCalib：无目标、任务结构化的人形机器人自我标定

Lu, Kaixiang, Lan, Haiyu, Qiao, Chunxiao, Li, You, Li, Enyu, Lu, Yehao, Yang, Jiarui, Lin, Peiwen, Wang, Chuang

Abstract

Assembly, wear, and component replacement perturb the sensor extrinsics and joint zeros encoded by a humanoid CAD model. Existing procedures calibrate one sensor pair or require external fiducials. Using only robot-native motion and onboard sensing, we present OmniCalib, a target-free workflow that calibrates the full upper limbs---all 14 arm joint zeros and the extrinsics of both wrist and chest cameras---as well as lower limbs and the multi-camera head rig. Each module matches a robot-native task to a parameter block, checks observability, and writes only supported corrections to the CAD model. Our depth ICP method recovers all 14 arm joint zeros and calibrates all RGB-D camera extrinsics without any calibration target. Relative to CAD, the estimated extrinsic corrections are 10.56 mm and 1.74 degrees for the left wrist, 6.33 mm and 1.25 degrees for the right wrist, and 9.81 mm and 0.929 degrees for the chest RGB-D camera. ICP point-to-plane residual is 2.09 mm. On the same injected offsets, ICP and ArUco recover all 14 joint zeros below the 0.1-degree encoder-resolution reference. On an AGIBOT A3 Ultra humanoid, four static double-support stances recover all 12 lower-limb joint-zero offsets injected with an RMS error of 0.063 degrees. The head module combines multi-camera visual odometry with legged odometry and dynamic compensation through the live ROS transform tree. Using only planar walking, it attains a mean SO(3) error of 1.061 degrees across three sequences. The best sequence reaches 0.775 degrees, competitive with iKalibr at 0.902 degrees from rich 6-DOF excitation. Rig-relative angles repeat within 0.140 degrees. Injection recovery and held-out tests validate each observable block.

Chinese Translation

组装、磨损和组件更换会扰动人形计算机辅助设计（CAD）模型中编码的传感器外部参数和关节零点。现有的标定程序通常只对一对传感器进行标定或需要外部基准。我们提出了OmniCalib，这是一种仅使用机器人原生运动和机载传感器的无目标工作流程，能够标定完整的上肢——所有14个手臂关节零点以及腕部和胸部摄像头的外部参数——以及下肢和多摄像头头部装置。每个模块将一个机器人原生任务与一个参数块匹配，检查可观测性，并仅将支持的修正写入CAD模型。我们的深度ICP方法在没有任何标定目标的情况下恢复所有14个手臂关节零点，并标定所有RGB-D摄像头的外部参数。相对于CAD，估计的外部修正为左腕10.56毫米和1.74度，右腕6.33毫米和1.25度，胸部RGB-D摄像头9.81毫米和0.929度。ICP点到平面的残差为2.09毫米。在相同的注入偏移下，ICP和ArUco能够恢复所有14个关节零点，低于0.1度的编码器分辨率参考。在AGIBOT A3 Ultra人形机器人上，四个静态双支撑姿态恢复了所有12个下肢关节零点的偏移，均方根误差为0.063度。头部模块结合了多摄像头视觉里程计与腿部里程计，并通过实时ROS变换树进行动态补偿。仅使用平面行走，它在三个序列中获得了平均SO(3)误差为1.061度。最佳序列达到0.775度，与iKalibr在丰富的6自由度激励下的0.902度相竞争。相对机器人角度重复在0.140度以内。注入恢复和保留测试验证了每个可观测块。

View on arXiv Download PDF AI Translation

cs.RO / 35 / 2609.19588

Quantifying Mechanical Intelligence in Legged Robots with Information Theory

利用信息理论量化腿部机器人中的机械智能

Patterson, Zach J.

Abstract

Mechanical intelligence, loosely defined as the reduction in control burden afforded by a robot's physical form, has become a prominent concept in robotics, with instantiations in bioinspired robotics, soft robotics, robotic swarms, and many other areas. However, rigorous theoretical understanding and quantitative measures of mechanical intelligence have lagged behind the engineering systems that the community has developed. In this work, using modern legged robots as a benchmark and exemplar, we propose several information-theoretic metrics for quantifying mechanical intelligence. By viewing body dynamics as both a computational process and a communication channel, we show that several prior insights in legged-robot engineering can be described using information theory, and we quantify how bits are processed by mechanical modes and across robot coordinates. Specifically, we examine the trade-off between explicitly incorporating compliance through series-elastic actuation and using so-called proprioceptive, low-gear-ratio transmissions, and we explore how these mechanisms interact with control policies during locomotion. We develop these results on systems of increasing complexity: a simplified linear model of a robot-leg transmission, a nonlinear single-leg simulation, and simulated quadruped robots controlled by a learned policy while navigating challenging terrain. These results lay the groundwork for broader study of robot mechanisms and their role in embodied computation.

Chinese Translation

机械智能被宽泛地定义为机器人物理形态所带来的控制负担的减少，已成为机器人领域的一个重要概念，体现在生物启发机器人、软机器人、机器人群体等多个领域。然而，机械智能的严格理论理解和定量测量仍滞后于社区开发的工程系统。在本研究中，我们以现代腿部机器人作为基准和示例，提出了几种信息理论度量方法来量化机械智能。通过将身体动力学视为一种计算过程和通信通道，我们展示了腿部机器人工程中的若干先前见解可以用信息理论来描述，并量化了机械模式和机器人坐标中比特的处理方式。具体而言，我们考察了通过串联弹性驱动显式引入顺应性与使用所谓的本体感知、低齿比传动之间的权衡，并探讨了这些机制在运动过程中与控制策略的相互作用。我们在逐渐复杂的系统上发展这些结果：一个简化的机器人腿传动线性模型、一个非线性的单腿仿真，以及在挑战性地形中通过学习策略控制的模拟四足机器人。这些结果为更广泛地研究机器人机制及其在具身计算中的作用奠定了基础。

View on arXiv Download PDF AI Translation

cs.RO / 36 / 2609.19600

WorldContact: A Contact-Centric World Model for Scalable Robot Learning

WorldContact：一种面向接触的可扩展机器人学习世界模型

Wang, Caoliwen, Wang, Mengdi, Zhang, Heng, Huang, Shixun, Chen, Siyuan, Liu, Chao, Chen, Anpei, Wang, Zhendong, Chen, Peter Yichen, Wang, Huamin

Abstract

Adapting robots to new objects and tasks requires interaction experience that can be costly to obtain. We present WorldContact, a contact-centric world model for deformable-object manipulation, constructed from a limited set of high-quality trajectories to generate additional training data efficiently. It predicts object dynamics using larger time steps than the source numerical simulator, which requires small integration steps to resolve rapid motion and prevent interpenetration. We evaluate WorldContact across 16 shopping-bag manipulation tasks. State-rollout measurements on a single H100 GPU show a $10\times$ speedup over the source simulator, excluding rendering and disk I/O. We use the generated data to fine-tune an existing vision-language-action policy and deploy it directly on a real robot. In bag lifting, the same policy achieves 65% single-attempt success when fine-tuned on source simulation data alone, compared with 95% when fine-tuned on the dataset expanded with WorldContact. These results support efficient data generation with WorldContact for robot policy adaptation.

Chinese Translation

将机器人适应新物体和任务需要获取交互经验，而这往往代价高昂。我们提出了WorldContact，一种针对可变形物体操控的接触中心世界模型，该模型基于有限的高质量轨迹集构建，旨在高效生成额外的训练数据。它使用比源数值模拟器更大的时间步长来预测物体动态，而源数值模拟器需要小的积分步长以解决快速运动并防止物体间的相互穿透。我们在16个购物袋操控任务上评估了WorldContact。在单个H100 GPU上的状态展开测量显示，相较于源模拟器，速度提升达10倍（不包括渲染和磁盘I/O）。我们利用生成的数据对现有的视觉-语言-动作策略进行微调，并直接在真实机器人上部署。在袋子提升任务中，当仅在源模拟数据上进行微调时，同一策略的单次尝试成功率为65%，而在使用WorldContact扩展的数据集上微调后成功率提升至95%。这些结果支持使用WorldContact进行高效数据生成，以适应机器人策略。

View on arXiv Download PDF AI Translation

cs.RO / 37 / 2609.19613

TacSushi: Tactile-Grounded World-Action Modeling for Dexterous Sushi Manipulation

TacSushi：用于灵巧寿司操作的触觉基础世界-动作建模

Hu, Haodi, Kogashi, Kaen, Koike-Akino, Toshiaki

Abstract

Dexterous food manipulation requires control under deformation, occlusion, and uncertain contact. We present TacSushi, a tactile-grounded, Cosmos3-based world-action policy that learns from recorded future consequences while acting on current observations. The backbone encodes current RGB, language, and hand state, and feature-wise gated fusion incorporates fingertip tactile features into the action representation. During training, a decoder conditioned on demonstrated action chunks predicts logged future visual observations, task progress, relative contact risk, and tactile summaries; this decoder is removed at deployment. Failed trials provide consequence supervision, but their actions are excluded from imitation. We train TacSushi on 340 successful and 50 failed real-robot trials and compare six methods in 600 separate rollouts across three in-distribution tasks and two out-of-distribution ingredient variants. To assess food quality beyond a single geometric threshold, we score terminal outcomes using an anchored visual-quality protocol that equally weights five human ratings and three vision-language-model ratings per rollout. Full TacSushi achieves 68.3% average in-distribution success and 37.5% out-of-distribution success, compared with 36.7%/10.0% without future-consequence supervision and 25.0%/17.5% with direct tactile concatenation in place of gated fusion. These comparisons support complementary benefits of feature-wise gated tactile fusion and training-only predictive supervision.

Chinese Translation

灵巧的食品操作需要在变形、遮挡和不确定接触下进行控制。我们提出了TacSushi，这是一种基于触觉的、基于Cosmos3的世界-动作策略，它在当前观察的基础上，从记录的未来结果中学习。其主干网络编码当前的RGB图像、语言和手部状态，特征级门控融合将指尖触觉特征纳入动作表示。在训练过程中，条件于示范动作片段的解码器预测记录的未来视觉观察、任务进展、相对接触风险和触觉摘要；该解码器在部署时被移除。失败的试验提供了结果监督，但其动作被排除在模仿之外。我们在340个成功和50个失败的真实机器人试验上训练TacSushi，并在三个分布内任务和两个分布外成分变体中比较六种方法的600次独立回放。为了评估食品质量超出单一几何阈值，我们使用锚定视觉质量协议对终端结果进行评分，该协议对每次回放的五个人工评分和三个视觉-语言模型评分给予同等权重。完整的TacSushi在分布内成功率达到68.3%，而分布外成功率为37.5%，相比之下，缺乏未来结果监督的情况下成功率为36.7%/10.0%，而直接触觉串联替代门控融合的情况下为25.0%/17.5%。这些比较支持了特征级门控触觉融合和仅训练预测监督的互补效益。

View on arXiv Download PDF AI Translation

cs.RO / 38 / 2609.19659

EmbodiedMind: Adaptive Data Curation and Prefix-Tree Reinforcement Learning for Efficient Embodied Intelligence

EmbodiedMind：自适应数据策划与前缀树强化学习以实现高效的具身智能

Wang, Feifan, Zhang, Zongbing, Zhang, Yu, Wang, Lingfeng, Zhu, Yurui, Deng, Jin, Zhang, Mingliang, Gao, Zhengguang, Wang, Yongcheng, Xu, Jin, Yang, Ri

Abstract

Training embodied foundation models typically requires massive-scale datasets and extensive computational resources, yet often suffers from three critical limitations: (1) inefficient sample utilization due to low-informative samples; (2) imbalanced gradient contributions across heterogeneous tasks; and (3) severe credit assignment problem in long-horizon planning, where trajectory-level rewards indiscriminately penalize all tokens. To address these issues, we propose an efficient training paradigm that achieves state-of-the-art average performance through strategic data selection and hierarchical policy optimization. Our approach consists of three synergistic stages. First, Rejection Sampling-based Fine-Tuning (RSFT) filters out low-informative samples to establish robust behavioral priors while preventing distributional collapse. Second, Iterative Rejection GRPO (IR-GRPO) employs task-specific queues stratified by difficulty to keep datasets balanced across reinforcement learning iterations, coupled with a hybrid reward mechanism for precise cross-task feedback. Third, to enhance long-horizon task planning, we introduce Trie-GRPO, a novel reinforcement learning algorithm based on action prefix trees, which enables step-level advantage estimation. This resolves the credit assignment problem by isolating intermediate correct decisions from downstream errors, while effectively balancing exploration efficiency and depth compared to conventional search trees. As a result, EmbodiedMind achieves a state-of-the-art average performance of 70.02% across 18 benchmarks, and significantly outperforms other embodied foundation models in long-horizon task planning accuracy. Our project will be released for reproducibility.

Chinese Translation

训练具身基础模型通常需要大规模的数据集和广泛的计算资源，但往往面临三个关键限制：（1）由于低信息样本导致的样本利用效率低下；（2）异构任务之间梯度贡献不平衡；以及（3）在长时间规划中严重的信用分配问题，其中轨迹级奖励无差别地惩罚所有标记。为了解决这些问题，我们提出了一种高效的训练范式，通过战略性的数据选择和分层策略优化实现了最先进的平均性能。我们的方法由三个协同阶段组成。首先，基于拒绝采样的微调（Rejection Sampling-based Fine-Tuning，RSFT）过滤掉低信息样本，以建立稳健的行为先验，同时防止分布崩溃。其次，迭代拒绝GRPO（Iterative Rejection GRPO，IR-GRPO）采用按难度分层的任务特定队列，以在强化学习迭代中保持数据集的平衡，并结合混合奖励机制以实现精确的跨任务反馈。第三，为了增强长时间任务规划，我们引入了Trie-GRPO，一种基于动作前缀树的新型强化学习算法，能够进行逐步优势估计。这通过将中间正确决策与下游错误隔离来解决信用分配问题，同时有效平衡探索效率和深度，相较于传统搜索树。结果，EmbodiedMind在18个基准测试中实现了70.02%的最先进平均性能，并在长时间任务规划准确性上显著优于其他具身基础模型。我们的项目将发布以便于可重复性研究。

View on arXiv Download PDF AI Translation

cs.RO / 39 / 2609.19661

ReShoot: Generative Visual Domain Randomization of Recorded Robot Demonstrations for Visuomotor Policy Learning

ReShoot：用于视觉运动策略学习的记录机器人演示的生成视觉领域随机化

Kim, Chiyoung, Choi, Min Sung, Ju, Jinho, Gu, Chanhoe, Hwang, Donghwan, Choi, Wonseok, Jeon, Woongsun, Lee, Minhyeok

Abstract

Imitation-learned robot policies are frequently overfit to the visual conditions present in their training demonstrations. Consequently, variations in object color or background appearance often induce substantial performance degradation. A common mitigation strategy is to acquire additional demonstrations in each novel visual context; however, this approach is resource-intensive, requiring repeated access to a robot, a controlled environment, and human operation for every appearance condition to be covered. We introduce ReShoot, a framework that synthesizes visual diversity by re-rendering previously recorded demonstrations under altered appearances, thereby shifting the burden from data collection to generation. A vision-language model captions the scene, edits a targeted attribute (e.g., background, object color, or material), and an edge-conditioned video generator re-renders both camera views to match. The instruction is updated accordingly. The action sequence and proprioceptive trajectory are copied verbatim without relabeling, so each generated episode retains the recorded action and proprioceptive labels. On LIBERO, a policy trained on an equal mixture of recorded and re-rendered demonstrations matches the performance of recorded-only training (96.5% vs. 96.9%). Moreover, the mixed training set improves robustness to scene perturbations on LIBERO-Plus (85.5% vs. 82.3%). Across two physical robotic platforms, deploying ReShoot with 43 and 100 pre-collected demonstrations increased the success rate on recolored objects from 0.0% to 42.9% and 47.5%, respectively, while maintaining performance under the original recorded appearance.

Chinese Translation

模仿学习的机器人策略往往过拟合于其训练演示中存在的视觉条件。因此，物体颜色或背景外观的变化常常导致性能显著下降。一种常见的缓解策略是在每种新视觉环境中获取额外的演示；然而，这种方法资源密集，需要对机器人、受控环境进行重复访问，并且在每种外观条件下需要人工操作。我们提出了ReShoot，一个通过在改变外观下重新渲染先前记录的演示来合成视觉多样性的框架，从而将负担从数据收集转移到生成。一个视觉-语言模型为场景生成描述，编辑目标属性（例如，背景、物体颜色或材料），并且一个边缘条件的视频生成器重新渲染两个摄像头视图以匹配。指令也相应更新。动作序列和本体感知轨迹逐字复制，无需重新标记，因此每个生成的剧集保留了记录的动作和本体感知标签。在LIBERO上，基于记录和重新渲染演示的等比例混合训练的策略，其性能与仅基于记录的训练相匹配（96.5%对96.9%）。此外，混合训练集提高了在LIBERO-Plus上对场景扰动的鲁棒性（85.5%对82.3%）。在两个物理机器人平台上，使用ReShoot与43和100个预收集的演示相结合，将重新着色物体的成功率从0.0%提高到42.9%和47.5%，同时在原始记录外观下保持性能。

View on arXiv Download PDF AI Translation

cs.RO / 40 / 2609.19665

Runtime Safety Filtering for Two-Terminal Hazards in Robotic Battery Recycling

机器人电池回收中的双端口危险的运行时安全过滤

Cao, Yuxin, Song, Wei, Yang, Xianglin, Guo, Fusen, Li, Lin, Cheng, Xiao, Dong, Jin Song

Abstract

Runtime safety filters for learned manipulation policies typically define unsafe states as unions of object-wise keep-out regions. This representation can be unnecessarily restrictive for hazards that depend on a joint spatial relation, such as battery recycling, where a conductive payload can short a charged cell only when it approaches both terminals simultaneously. We study runtime filtering for this two-terminal hazard in LIBERO using frozen OpenVLA policies. We factor a runtime filter into three design choices: the predicate structure, its geometric margin, and the fallback action applied when a commanded action is rejected. We compare a conjunctive predicate, a conventional two-site keep-out, and a composite of the two. For each predicate, we vary its margin to obtain a frontier between task success and residual hazard. We then compare four fallback strategies at matched operating points: holding, retreat, sampled search, and a continuous-action barrier projection. Across three workcells, the three predicate families trace nearly identical safety--utility frontiers once each is evaluated over its own margin. In contrast, the fallback strategy has a substantially larger effect: holding reduces task success by up to 0.302 relative to retreat without reducing hazard, while both minimally invasive fallbacks leave substantially more residual hazard. This ordering transfers to a second policy and task suite, while retreat-based filtering remains effective under standing errors in the clearances available to the filter, although correlated error in the estimated payload size is more damaging than larger independent errors in terminal position. These results show that, for proximity-defined manipulation hazards, margin selection and fallback strategy can matter more than predicate structure in determining the safety--utility trade-off of a runtime filter.

Chinese Translation

针对学习的操作策略，运行时安全过滤器通常将不安全状态定义为对象级的禁入区域的并集。这种表示方式对于依赖于联合空间关系的危险（例如电池回收）可能过于限制，因为导电负载仅在同时接近两个端口时才会短路充电电池。我们在LIBERO中研究了这种双端口危险的运行时过滤，使用冻结的OpenVLA策略。我们将运行时过滤器分为三个设计选择：谓词结构、几何边际以及在拒绝命令动作时应用的后备动作。我们比较了一个合取谓词、一个传统的双点禁入以及两者的组合。对于每个谓词，我们改变其边际以获得任务成功与残余危险之间的边界。然后，我们在匹配的操作点比较四种后备策略：保持、撤退、采样搜索和连续动作障碍投影。在三个工作单元中，三种谓词家族在各自的边际上评估后几乎描绘出相同的安全-效用边界。相比之下，后备策略的影响显著更大：保持相对于撤退使任务成功率降低了最多0.302，而不减少危险，而两种最小侵入的后备策略则留下了显著更多的残余危险。这一排序在第二个策略和任务套件中得以延续，而基于撤退的过滤在过滤器可用间隙的静态误差下仍然有效，尽管估计负载大小的相关误差比端口位置的较大独立误差更具破坏性。这些结果表明，对于基于接近定义的操作危险，边际选择和后备策略在决定运行时过滤的安全-效用权衡中可能比谓词结构更为重要。

View on arXiv Download PDF AI Translation

cs.RO / 41 / 2609.19666

Towards High-DoF Dexterous Manipulation through VLA Post-Training

通过 VLA 后训练实现高自由度灵巧操作

Zhu, Junlei, Yao, Shenzhe, Huang, Chaogui, Zhu, Wenkai, Peng, Jingwei, He, Guanqi, Schwertfeger, Soren, Chen, Jiahao, Liu, Yide

Abstract

Imitation-learned vision--language--action (VLA) foundation models acquire broad manipulation capabilities by scaling robot data across tasks and embodiments, but reliable deployment on a specific downstream task and hardware platform still requires post-training. Dexterous hands make this adaptation particularly difficult: their broad behavioural repertoire and high degree of freedom create a large and structured action space. Three obstacles are central: open-source VLAs do not natively provide an action interface for high-DoF hands; gesture mismatch during human-gated DAgger takeover creates command discontinuities and contaminates corrective trajectories; and reinforcement learning in the raw joint space is sample-inefficient. We present a unified four-step post-training pipeline comprising a learned temporal hand-action codec, supervised fine-tuning, DAgger, and real-world residual reinforcement learning. The codec adapts a pretrained VLA to absolute dexterous-hand commands. Buffered rollback, pose alignment, and smooth command blending enable continuous, task-relevant DAgger corrections, while latent residual RL confines exploration to coordinated hand motions captured by the codec. We evaluate the pipeline on five diverse real-world tasks spanning bimanual transfer, in-hand reorientation, and tool use. Within the reported post-training budgets, the resulting policies achieve 100\% success on every evaluated task over 20 trials per task. These results provide a practical path for adapting VLA foundation models to reliable real-world dexterous manipulation.

Chinese Translation

模仿学习的视觉-语言-动作（VLA）基础模型通过在任务和实体之间扩展机器人数据，获得了广泛的操作能力，但在特定下游任务和硬件平台上可靠部署仍需后训练。灵巧手使得这种适应特别困难：其广泛的行为谱和高自由度创造了一个庞大且结构化的动作空间。三个障碍是关键：开源 VLA 并未原生提供高自由度手的动作接口；在人工干预的 DAgger 接管过程中，手势不匹配导致命令不连续，并污染了纠正轨迹；在原始关节空间中的强化学习样本效率低下。我们提出了一个统一的四步后训练流程，包括学习的时间手部动作编码器、监督微调、DAgger 和现实世界的残差强化学习。该编码器将预训练的 VLA 适配为绝对灵巧手命令。缓冲回滚、姿态对齐和光滑命令混合使得连续的、与任务相关的 DAgger 纠正成为可能，而潜在的残差强化学习将探索限制在编码器捕获的协调手部动作上。我们在五个多样化的现实世界任务上评估了该流程，涵盖双手转移、手内重新定向和工具使用。在报告的后训练预算内，所得到的策略在每个评估任务的 20 次试验中均实现了 100\% 的成功率。这些结果为将 VLA 基础模型适应于可靠的现实世界灵巧操作提供了一条实用路径。

View on arXiv Download PDF AI Translation

cs.RO / 42 / 2609.19681

VAST: V2X/Dynamic Map-Aware Autonomous Driving Systems Validation Toolchain

VAST：V2X/动态地图感知的自动驾驶系统验证工具链

Ito, Shunsuke, Azumi, Takuya

Abstract

Cooperative autonomous driving in the IoT-to-Edge-to-Cloud continuum requires system-level validation across vehicles, infrastructure sensors, edge-side Dynamic Map services, and in-vehicle autonomous-driving stacks. This paper presents VAST, a V2X/Dynamic Map-aware validation toolchain that connects Scenic, Scenario Simulator v2, AWSIM, Autoware, and SIM-LDM. VAST does not introduce a new search algorithm; instead, it addresses interoperability challenges, including Lanelet2-to-Scenic mapping, ROS 2-based co-simulation through SS2, Dynamic Map object injection into Autoware, and collection of TTC, PET, collision, timeout, and performance measurements. In occluded-intersection scenarios, Lanelet2-compatible constrained sampling increases the edge-case discovery rate from 40.0% to 80.0% and reduces the average time per discovered edge case from 259.7 s to 110.4 s. Under the same generated scenario distribution, Dynamic Map availability reduces the collision rate from 78.0% to 40.0% and increases non-collision outcomes from 22.0% to 60.0%, with statistically significant TTC/PET shifts. A throughput study with 1-16 NPCs shows that sampling remains below 0.1 s, whereas AWSIM/Autoware execution and restart overhead dominate runtime. These results position VAST as a practical validation infrastructure for cooperative autonomous-driving CPSs.

Chinese Translation

在物联网到边缘计算再到云计算的连续体中，协同自动驾驶需要在车辆、基础设施传感器、边缘侧动态地图服务和车载自动驾驶堆栈之间进行系统级验证。本文提出了VAST，一个V2X/动态地图感知的验证工具链，连接了Scenic、Scenario Simulator v2、AWSIM、Autoware和SIM-LDM。VAST并未引入新的搜索算法，而是解决了互操作性挑战，包括Lanelet2到Scenic的映射、基于ROS 2的SS2联合仿真、动态地图对象注入到Autoware中，以及收集TTC、PET、碰撞、超时和性能测量。在遮挡交叉口场景中，兼容Lanelet2的约束采样将边缘案例发现率从40.0%提高到80.0%，并将每个发现的边缘案例的平均时间从259.7秒减少到110.4秒。在相同生成的场景分布下，动态地图可用性将碰撞率从78.0%降低到40.0%，并将非碰撞结果从22.0%提高到60.0%，且TTC/PET的变化具有统计学意义。对1-16个NPC的吞吐量研究表明，采样时间保持在0.1秒以下，而AWSIM/Autoware的执行和重启开销主导了运行时间。这些结果使VAST成为协同自动驾驶网络物理系统（CPS）的实用验证基础设施。

View on arXiv Download PDF AI Translation

cs.RO / 43 / 2609.19688

LYRIC: Language-Driven Physics-Based Character Control for Contact-Rich Whole-Body Object Interaction

LYRIC：基于语言驱动的物理学角色控制，用于丰富接触的全身物体交互

Han, Zeyu, Meng, Zichong, Tanke, Julian, Matsumoto, Minami, Bashkirov, Sergey, Fan, Yingruo, Engin, Selim, Shim, Dongseok, Shibuya, Takashi, Mitsufuji, Yuki, Jiang, Huaizu

Abstract

We present LYRIC, a generative flow-matching controller for language-driven physics-based contact-rich interaction control, that enables simulated characters to perform contact-rich whole-body object interactions from a free-form language instruction and a sparse terminal object goal. To obtain reliable expert trajectories from imperfect motion-capture references, a single tracking policy is trained using geometry-conditioned interaction rewards and relaxed reference tracking near hand-object contact. To guide interaction progress without prescribing a full-body kinematic reference, we factorize the controller into a task-level planner that predicts short-horizon object and humanoid-root trajectories, and an action generator that resolves whole-body motion and contacts in closed loop. After behavior cloning, we freeze the planner and post-tune the action generator on policy using the planner's predictions as stable supervision for intermediate task progression. In a controlled OMOMO evaluation, our tracker achieves 64.3% success compared with 53.2% for an InterMimic reimplementation, while a unified policy achieves 76.5% on the full OMOMO dataset. On the held-out split, LYRIC achieves 90.3% task success, compared with 74.2% for the strongest matched kinematic-planner baseline, with better semantic alignment and motion quality. Without retraining, the controller also supports test-time object-waypoint guidance. Qualitative results further demonstrate robust, natural contact-rich interactions and zero-shot transfer to novel object shapes. The webpage is available at https://neu-vi.github.io/LYRIC/

Chinese Translation

我们提出了LYRIC，一种生成流匹配控制器，用于基于语言驱动的物理学接触丰富交互控制，使模拟角色能够根据自由形式的语言指令和稀疏的终端物体目标执行接触丰富的全身物体交互。为了从不完美的动作捕捉参考中获得可靠的专家轨迹，使用几何条件的交互奖励和在手物体接触附近的放松参考跟踪训练了单一的跟踪策略。为了在不规定全身运动学参考的情况下引导交互进程，我们将控制器分解为一个任务级规划器，该规划器预测短期物体和人形根轨迹，以及一个在闭环中解决全身运动和接触的动作生成器。在行为克隆之后，我们冻结规划器，并在策略上对动作生成器进行后调优，使用规划器的预测作为中间任务进展的稳定监督。在受控的OMOMO评估中，我们的跟踪器成功率达到64.3%，而InterMimic的重新实现为53.2%；同时，统一策略在完整的OMOMO数据集上达到了76.5%。在保留的分割数据上，LYRIC的任务成功率为90.3%，相比之下，最强匹配的运动学规划基线为74.2%，且在语义对齐和运动质量上表现更佳。在不重新训练的情况下，该控制器还支持测试时的物体路径点引导。定性结果进一步展示了稳健、自然的接触丰富交互以及对新物体形状的零样本迁移。相关网页可访问 https://neu-vi.github.io/LYRIC/

View on arXiv Download PDF AI Translation

cs.RO / 44 / 2609.19690

UniExo: Unified Multi-Skill Policies for Musculoskeletal Locomotion and Co-Adaptive Exoskeleton Control

UniExo：用于肌肉骨骼运动和协同适应外骨骼控制的统一多技能策略

Yuan, Yifei, Wolf, Jakob, Androwis, Ghaith, Zhou, Xianlian

Abstract

Daily locomotion encompasses diverse activities and frequent transitions between them, yet most exoskeleton controllers are designed for a single activity or a narrow set of related movements. Changes in activity therefore typically require explicit mode switching and separately tuned or retrained controllers. Simulation-based learning reduces the need for hardware-based tuning but generally retains this limitation. Here we present UniExo, a framework that first constructs a multi-skill musculoskeletal human policy and then jointly trains an exoskeleton control policy with it. Four single-skill imitation experts for walking, turning, running and backward walking are distilled into a single network structured by a skill latent and subsequently fine-tuned through reinforcement learning on transition sequences. The resultant unified human policy achieves a mean tracking success rate of 94.7% on unseen clips of the four skills and exhibits greater robustness to perturbations than its constituent experts. A single hip exoskeleton controller (UniExo) is initialized from hip moment prediction of the human policy and co-adapted with it through multi-agent reinforcement learning across the four skills. This co-adaptation shifts the timing of the assistance torque and raises the fraction of positive work delivered to the hip. When deployed on a custom hip exoskeleton, the controller generalizes across four treadmill speeds in six participants and assists one participant through a continuous route of all four skills and their transitions, without skill labels or explicit mode switching. UniExo thus provides a step towards replacing activity-specific controllers with unified, user-specific controllers that support diverse locomotor activities and the transitions between them.

Chinese Translation

日常运动涵盖多种活动并频繁在它们之间切换，然而大多数外骨骼控制器仅设计用于单一活动或一组狭窄的相关动作。因此，活动的变化通常需要明确的模式切换以及单独调优或重新训练的控制器。基于仿真的学习减少了对硬件调优的需求，但通常仍保留这一限制。在此，我们提出了UniExo，一个首先构建多技能肌肉骨骼人类策略的框架，然后与之共同训练外骨骼控制策略。四个单技能模仿专家（行走、转弯、跑步和倒退行走）被提炼为一个由技能潜变量构成的单一网络，并随后通过对过渡序列的强化学习进行微调。最终得到的统一人类策略在四项技能的未见片段上实现了94.7%的平均跟踪成功率，并表现出比其组成专家更强的抗干扰能力。单一髋关节外骨骼控制器（UniExo）从人类策略的髋关节力矩预测初始化，并通过在四项技能上的多智能体强化学习与之共同适应。这种共同适应调整了助力扭矩的时机，并提高了施加到髋关节的正功比例。当在定制的髋关节外骨骼上部署时，该控制器在六名参与者的四种跑步机速度下实现了泛化，并在没有技能标签或明确模式切换的情况下，帮助一名参与者连续完成所有四项技能及其过渡。因此，UniExo为用统一的、用户特定的控制器替代活动特定控制器提供了一步，支持多样的运动活动及其之间的过渡。

View on arXiv Download PDF AI Translation

cs.RO / 45 / 2609.19726

Decoupling Physical Speed from Path Parameterization in Singularity-Free Guiding Vector Fields

在无奇点引导向量场中将物理速度与路径参数化解耦

Xiao, Zhouru, Luo, Sha, Lu, Yang, Xiao, Mingliang, Yao, Weijia, Lin, Bohuan, Cheng, Xianzhe, Wang, Yaonan

Abstract

The existing singularity-free guiding vector field (SF-GVF) with an additional virtual coordinate can eliminate singular points (i.e., points where the vector field vanishes) inherent in conventional GVFs and guarantee global convergence of robot trajectories to closed and self-intersecting desired paths. However, the desired speed given by the GVF along the desired path in the original lower-dimensional space cannot be arbitrarily specified but depends on path parameterizations. One possible workaround is to partially normalize the physical projection of the SF-GVF and assign a user-designed speed. However, we show that this workaround may introduce new singularities since the normalization denominator can become zero. To address this issue, we propose a new SF-GVF with prescribed physical speed (PPS). The integral curves of the new SF-GVF converge exponentially to the desired path from any initial condition in the higher-dimensional space (including virtual dimension); more importantly, the robot's physical speed converges to the PPS, while the path-error dynamics remain invariant under regular reparameterizations of the desired path. We further develop a saturated acceleration control law for second-order kinematic models. Finally, comparative simulations and 3D path-following experiments with a quadrotor under different PPS profiles validate the theoretical results and demonstrate the effectiveness of the proposed approach.

Chinese Translation

现有的无奇点引导向量场（SF-GVF）通过引入额外的虚拟坐标，可以消除传统引导向量场（GVF）中固有的奇点（即向量场消失的点），并保证机器人轨迹全局收敛于闭合且自交的期望路径。然而，GVF在原始低维空间中沿期望路径给出的期望速度不能随意指定，而是依赖于路径参数化。一种可能的解决方法是部分归一化SF-GVF的物理投影并分配用户设计的速度。然而，我们表明这种解决方法可能引入新的奇点，因为归一化的分母可能变为零。为了解决这个问题，我们提出了一种具有规定物理速度（PPS）的新SF-GVF。新SF-GVF的积分曲线从高维空间中的任意初始条件（包括虚拟维度）以指数方式收敛于期望路径；更重要的是，机器人的物理速度收敛于PPS，而路径误差动态在期望路径的常规重新参数化下保持不变。我们进一步为二阶运动学模型开发了一种饱和加速度控制律。最后，在不同PPS配置下与四旋翼进行的比较仿真和三维路径跟踪实验验证了理论结果，并展示了所提方法的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 46 / 2609.19742

Equivariant Filter Design for Acoustic and Depth Aided Inertial Navigation Systems

声学和深度辅助惯性导航系统的等变滤波器设计

Lunawat, Arihant, van Goor, Pieter, Dellaert, Frank, Williams, Stefan B.

Abstract

Autonomous Underwater Vehicles (AUVs) navigating without GPS typically fuse inertial measurements with acoustic Doppler Velocity Log (DVL) velocities and pressure-derived depth. Posing the navigation state on a Lie group improves accuracy and consistency. However, state-of-the-art filters based on the Invariant Extended Kalman Filter (IEKF) append the Inertial Measurement Unit (IMU) biases as a Euclidean extension, which breaks the group-affine structure required for exact log-linear error dynamics, causing the reported covariance to degrade alongside the estimate. We apply the Tangent-Group (TG) symmetry, which carries the biases within the geometry of the state space, to derive an Equivariant Filter (EqF) for this system, leaving zero linearization error in the navigation states and second-order error only in the biases. We develop an equivariant output model for the DVL, whose update incurs only third-order linearization error, together with a direct pressure output. Monte Carlo simulations benchmark the TG-EqF against a Two-Frame-Group IEKF and a Multiplicative EKF. The TG-EqF reduces error by 18--25\% against both alternatives in each of attitude, velocity, and position. The main benefit is in the covariance it estimates: its Average Normalized Estimation Error Squared (ANEES) stays closer to its nominal value of one than that of the others. Offline analysis on AUV field data corroborates the findings of the simulations, demonstrating reduced position drift.

Chinese Translation

自主水下航行器（AUV）在没有GPS的情况下导航，通常将惯性测量与声学多普勒速度计（DVL）速度和压力推导的深度进行融合。将导航状态置于李群上可以提高准确性和一致性。然而，基于不变扩展卡尔曼滤波器（IEKF）的最新滤波器将惯性测量单元（IMU）偏差作为欧几里得扩展附加，这破坏了精确对数线性误差动态所需的群仿射结构，导致报告的协方差随着估计的恶化而下降。我们应用切空间群（TG）对称性，将偏差纳入状态空间的几何结构，从而为该系统推导出一种等变滤波器（EqF），在导航状态中保持零线性化误差，仅在偏差中存在二阶误差。我们为DVL开发了一种等变输出模型，其更新仅产生三阶线性化误差，并结合直接压力输出。蒙特卡洛仿真将TG-EqF与双帧群IEKF和乘法EKF进行了基准测试。TG-EqF在姿态、速度和位置的每个方面相较于这两种替代方案减少了18%至25%的误差。其主要优势在于它估计的协方差：其平均归一化估计误差平方（ANEES）保持在接近于一的名义值，而其他方法则不然。对AUV现场数据的离线分析证实了仿真的发现，显示出位置漂移的减少。

View on arXiv Download PDF AI Translation

cs.RO / 47 / 2609.19796

LIFD: Anchored Diffusion for 3D-Aware Scene Memory in Robotic Manipulation

LIFD：用于机器人操作中的三维感知场景记忆的锚定扩散

Li, Wenbo, Chen, Yiteng, Li, Wenhao, Wu, Qingyao

Abstract

Robotic manipulation under partial observability requires spatial information that extends beyond the current view. Geometry-aware RGB features describe visible structure, but previously observed regions may disappear as the robot or scene moves. Maintaining a useful scene representation therefore requires retaining observation history while inferring missing content without losing its connection to visible evidence. We introduce LIFD (Look, Imagine, Focus, and Do), a framework for persistent, 3D-aware scene memory. LIFD learns a scene-token representation from multi-view agreement and completes it from a single RGB view and recurrent memory. A rectified-flow model generates the tokens while Anchor-Guided Cross-Attention conditions completion on current geometric features. Compact slot features connect this representation to a manipulation policy. Multi-view and geometric supervision are used during representation learning; deployment requires one RGB camera, proprioception, and a task instruction. LIFD (Staged) reaches 91.6% average success on LIBERO and 79.8% on MetaWorld, improving LIBERO average success by 3.1 percentage points over Joint training. On four UR5e task families with ten demonstrations per family, it achieves 56.0% mean success, compared with 40.5% for OpenVLA-7B.

Chinese Translation

在部分可观测性下的机器人操作需要超越当前视图的空间信息。几何感知的RGB特征描述了可见结构，但随着机器人或场景的移动，先前观察到的区域可能会消失。因此，保持有用的场景表示需要保留观察历史，同时在不失去与可见证据连接的情况下推断缺失的内容。我们提出了LIFD（Look, Imagine, Focus, and Do），一个用于持久的三维感知场景记忆的框架。LIFD通过多视图一致性学习场景标记表示，并通过单个RGB视图和递归记忆来完成它。一个校正流模型生成这些标记，而锚定引导的交叉注意力则基于当前几何特征条件化完成。紧凑的槽特征将该表示与操作策略连接。在表示学习过程中使用多视图和几何监督；部署时需要一个RGB摄像头、自我感知和任务指令。LIFD（分阶段）在LIBERO上达到91.6%的平均成功率，在MetaWorld上达到79.8%，相比于联合训练提高了LIBERO平均成功率3.1个百分点。在四个UR5e任务系列中，每个系列有十个演示，其平均成功率为56.0%，而OpenVLA-7B为40.5%。

View on arXiv Download PDF AI Translation

cs.RO / 48 / 2609.19802

Affective Shared Autonomy: Temporal Affect Dynamics and Subjective Evaluation in Bimanual Teleoperation Tasks

情感共享自主性：双手远程操作任务中的时间情感动态与主观评估

Liang, Zhengji, Tian, Guiyin, Qu, Sijin, Liu, Hainan, Hu, Shiyan

Abstract

Physical teleoperation integrates human cognitive flexibility with robotic precision, yet demanding manipulation tasks frequently induce severe cognitive workload, acute frustration, and execution breakdown. Conventional shared autonomy paradigms rely primarily on task-based rules, such as spatial error boundaries, which disregard the operator's transient affective state and risk misaligned control interventions. To address this limitation, we propose an affect-aware shared autonomy teleoperation framework that dynamically modulates robotic assistance based on real-time operator state estimation. The system estimates operator affective states from synchronized facial video, cardiac signals, and bilateral arm kinematics, outputting a seven-state affective distribution and a three-category operational abstraction (neutral, productive, adverse). Affect-aware assistance is selectively triggered when the user is detected in a continuous adverse state, preserving task-positive engagement without unnecessary disruption. The empirical user study ($N = 30$) confirms that the proposed affective assistance increases the productive states by up to 39.7% without compromising user agency. The collected dataset represents the first multimodal dataset that provides continuous visual, physiological, and operator's bilateral motion tracking of temporal affective state shifts during bimanual teleoperation. Our multimodal fusion model outperforms zero-shot baselines (Qwen, MiniCPM-V) in tracking temporal state dynamics. This real-world deployment offers a new human-centric framework that integrates visual, physiological, and motion tracking for physical human-robot interaction.

Chinese Translation

物理远程操作将人类的认知灵活性与机器人的精确性相结合，但复杂的操作任务常常会导致严重的认知负荷、急性挫败感和执行中断。传统的共享自主性范式主要依赖于基于任务的规则，如空间误差边界，这忽视了操作员的瞬时情感状态，并可能导致控制干预的失调。为了解决这一局限性，我们提出了一种情感感知的共享自主远程操作框架，该框架基于实时的操作员状态估计动态调节机器人辅助。该系统通过同步的面部视频、心脏信号和双臂运动学来估计操作员的情感状态，输出七种情感分布和三类操作抽象（中性、积极、消极）。当检测到用户处于持续的消极状态时，情感感知的辅助会被选择性触发，从而在不必要干扰的情况下保持任务积极参与。实证用户研究（$N = 30$）确认所提出的情感辅助将积极状态提高了多达39.7%，而不影响用户的自主性。所收集的数据集是第一个多模态数据集，提供了双手远程操作过程中情感状态变化的连续视觉、生理和操作员双侧运动跟踪。我们的多模态融合模型在跟踪时间状态动态方面优于零样本基线（Qwen, MiniCPM-V）。这一现实世界的部署提供了一种新的以人为中心的框架，整合了视觉、生理和运动跟踪，以促进物理人机交互。

View on arXiv Download PDF AI Translation

cs.RO / 49 / 2609.19803

HEROIC: Heterogeneous Evidential Reasoning for Open-Vocabulary Identification and Cross-Robot Collaboration

HEROIC：用于开放词汇识别和跨机器人协作的异构证据推理

Chauhan, Mihir, Jain, Aarav, Zucek, Addison, Dang, Manmeet, Conover, Damon, Bera, Aniket

Abstract

Multi-agent heterogeneous air-ground robot teams are attractive for open world search, with applications for reconnaissance, urban search and rescue missions (USAR), disaster response and recovery, and hazardous environments. These two platforms have different failure modes: aerial robots cover ground quickly but cannot resolve small or occluded targets from altitude, while ground robots can identify objects-of-interest, such as people or hazardous objects, at close range but cover less area. Existing language-tasked teams either have roles fixed prior, or have a language model assign them from hand-written capability tags, so the team is unable to know when within a mission an asset is no longer useful. We present HEROIC, a decentralized heterogeneous multi-agent open-vocabulary search coordination framework that requires agents to communicate in natural language only. HEROIC's initial agent role assignment is derived from sensor properties and a scale law to determine whether targets can be detected with a high confidence. From the mission's natural language prompt alone, this law assigns aerial flight altitudes and sweep spacing. When this calculated height falls below the altitude for safe flight, aerial agents re-task themselves from searcher to aerial triage, escort, and route guide for ground agents. Both robots maintain an evidential belief over the search area (bearing rays for positive evidence, a log-odds posterior for negative evidence) and gate any arrival on close-range verification. In full-stack experiments, HEROIC reaches the target 84% of the time across all 6 scenes, compares to 35-54% for vision-language frontier baselines, frontier-based search, lawnmower, and random-walk running the same perception, all while being 2-4x sooner to arrive at the target.

Chinese Translation

多智能体异构空地机器人团队在开放世界搜索中具有吸引力，应用于侦察、城市搜索与救援任务（USAR）、灾害响应与恢复以及危险环境。这两种平台具有不同的失效模式：空中机器人能够快速覆盖地面，但无法从高空解析小型或被遮挡的目标，而地面机器人可以在近距离识别感兴趣的物体，如人或危险物品，但覆盖的区域较小。现有的语言任务团队要么在事先固定角色，要么由语言模型根据手写能力标签分配角色，因此团队无法在任务中判断某个资产何时不再有用。我们提出了HEROIC，一个去中心化的异构多智能体开放词汇搜索协调框架，要求智能体仅使用自然语言进行沟通。HEROIC的初始智能体角色分配基于传感器属性和一个规模法则，以确定目标是否可以高置信度地被检测到。仅通过任务的自然语言提示，该法则分配空中飞行高度和搜索间距。当计算出的高度低于安全飞行高度时，空中智能体会将自身的任务从搜索者重新分配为空中分诊、护送和地面智能体的路线引导。两种机器人在搜索区域内保持证据信念（正证据的方位射线，负证据的对数几率后验），并在近距离验证时进行门控。在全栈实验中，HEROIC在所有6个场景中以84%的概率成功到达目标，而视觉-语言前沿基线、基于前沿的搜索、割草机和随机行走在相同感知下的成功率为35-54%，同时HEROIC到达目标的时间比这些方法快2-4倍。

View on arXiv Download PDF AI Translation

cs.RO / 50 / 2609.19813

Vehicle Trajectory Prediction via Neural Fusion of Multiple EKF-Based Trajectory Candidates

通过神经融合多种基于扩展卡尔曼滤波的轨迹候选进行车辆轨迹预测

Kim, Seong-Jun, Kong, Seung-Hyun

Abstract

Predicting the future trajectories of surrounding vehicles in autonomous driving is important for collision risk assessment and safe ego-vehicle path planning. Conventional neural network-based trajectory predictors typically achieve strong prediction performance by exploiting agent history, dynamic scene graphs, and semantic maps. However, in specific motion regimes such as acceleration, deceleration, and turning, these predictors may fail to reflect physically feasible trajectories. To address this issue, this study proposes a framework that fuses the output of Trajectron++, a neural network-based trajectory predictor, with extended Kalman filter (EKF)-based multiple trajectory candidates at a late stage. On the nuScenes dataset, the proposed method reduces the average displacement error and final displacement error of the Trajectron++ robot baseline by 13.7% and 14.6%, respectively, without modifying the baseline architecture. These results indicate that EKF-based trajectory candidates can effectively complement neural trajectory prediction through learned fusion.

Chinese Translation

在自动驾驶中，预测周围车辆的未来轨迹对于碰撞风险评估和安全的自我车辆路径规划至关重要。传统的基于神经网络的轨迹预测器通常通过利用代理历史、动态场景图和语义地图来实现强大的预测性能。然而，在加速、减速和转弯等特定运动状态下，这些预测器可能无法反映物理上可行的轨迹。为了解决这个问题，本研究提出了一个框架，该框架在后期融合了基于神经网络的轨迹预测器Trajectron++的输出与基于扩展卡尔曼滤波（EKF）的多种轨迹候选。在nuScenes数据集上，所提出的方法分别将Trajectron++机器人基线的平均位移误差和最终位移误差降低了13.7%和14.6%，而无需修改基线架构。这些结果表明，基于EKF的轨迹候选可以通过学习融合有效补充神经轨迹预测。

View on arXiv Download PDF AI Translation

cs.RO / 51 / 2609.19817

RotateIt! Fast and Reliable Single-Arm Garment Unfolding via Online-Adaptive Dynamic Rotation

RotateIt！通过在线自适应动态旋转实现快速可靠的单臂服装展开

Zhang, Zeqing, Xie, Zuokun, Fang, Ao, Dai, Bin, Shu, Zhengjie, Tang, Yifeng, Wang, Ziwei

Abstract

Robotic garment unfolding is essential for downstream tasks, yet quasi-static methods require repeated actions, while existing dynamic approaches predominantly rely on bimanual flinging. We present RotateIt!, a single-arm framework that uses adaptive axial rotation for dynamic garment unfolding. To the best of our knowledge, it is the first unfolding framework to employ dynamic axial rotation as its primary manipulation primitive. From a randomly initialized tabletop configuration, the robot selects a rotation-effective grasp and rotates the lifted garment about an approximately fixed anchor, generating inertial tension that separates overlapping layers within a compact workspace. A grasp ranker selects the anchor, while an online residual policy adapts the rotation extent and speed, thereby determining the release timing. Across seen and unseen simulated garments and eight unseen real garments, RotateIt! improves success within three attempts by 44.0-61.0 percentage points over quasi-static pick-and-place. The simulation-trained policies transfer zero-shot to the real world, achieving 75.6% success, 41% higher first-attempt coverage, and 26% higher final coverage. The resulting states further enable autonomous robotic folding without manual rearrangement.

Chinese Translation

机器人服装展开对于下游任务至关重要，但准静态方法需要重复操作，而现有的动态方法主要依赖双手抛掷。我们提出了RotateIt！一个使用自适应轴向旋转进行动态服装展开的单臂框架。根据我们所知，这是第一个将动态轴向旋转作为主要操作原语的展开框架。从随机初始化的桌面配置开始，机器人选择一个有效的旋转抓取，并围绕一个近似固定的锚点旋转提起的服装，产生惯性张力，从而在紧凑的工作空间内分离重叠层。抓取排名器选择锚点，而在线残余策略则调整旋转的幅度和速度，从而确定释放时机。在已见和未见的模拟服装以及八件未见的真实服装中，RotateIt！在三次尝试内的成功率比准静态拾取和放置提高了44.0-61.0个百分点。经过模拟训练的策略在现实世界中实现了零样本转移，成功率达到75.6%，首次尝试覆盖率提高41%，最终覆盖率提高26%。所得到的状态进一步支持自主机器人折叠，无需手动重新排列。

View on arXiv Download PDF AI Translation

cs.RO / 52 / 2609.19824

TADreamer: Zero-Shot Language-Guided 3D Navigation for Terrestrial-Aerial Bimodal Robots via Video Imagination

TADreamer：基于视频想象的零-shot语言引导三维导航用于陆空双模机器人

Li, Xiangyu, Lai, Tiancheng, Huang, Xijie, Pang, Ruitian, Shen, Siqi, Chen, Juncheng, Pan, Zaisheng, Xu, Chao, Gao, Fei, Cao, Yanjun

Abstract

Language-guided navigation for terrestrial-aerial bimodal robots requires selecting routes and locomotion modes that match scene context and task intent. Generated videos can represent such motion sequences, but recovering metrically consistent navigation references from them is challenging because of scale ambiguity and axis-dependent geometric distortions. We present TADreamer, a zero-shot framework that grounds video-imagined navigation in measured geometry without task-specific training or fine-tuning. A vision-language model translates onboard observations and instructions into navigation prompts, selects valid generated videos, and provides corrective feedback when regeneration is needed. The selected video is reconstructed into 3D waypoints annotated with terrestrial or aerial modes. A two-stage calibration procedure uses field-of-view constraints to initialize scale estimation, then refines axis-dependent scales, rotation, and translation by registering the reconstructed point cloud to measured geometry. The calibrated waypoints and mode labels guide a planner that incorporates measured geometry for robot execution. Real-world experiments demonstrate navigation across seven indoor and outdoor scenarios. With five candidates per round, usable videos are obtained within two rounds in all seven scenarios. On the calibration observations, our method reduces mean absolute depth error by 87.7% and mean absolute relative depth error by 86.3% compared with NavDreamer.

Chinese Translation

语言引导的陆空双模机器人导航需要选择与场景上下文和任务意图相匹配的路线和运动模式。生成的视频可以表示这样的运动序列，但由于尺度模糊和轴依赖的几何失真，从中恢复出度量一致的导航参考是具有挑战性的。我们提出了TADreamer，这是一种零-shot框架，能够在不进行特定任务训练或微调的情况下，将视频想象的导航与测量几何相结合。视觉-语言模型将机载观察和指令转换为导航提示，选择有效的生成视频，并在需要重新生成时提供纠正反馈。所选视频被重建为带有陆地或空中模式注释的三维航点。一个两阶段的校准程序使用视场约束来初始化尺度估计，然后通过将重建的点云与测量几何进行配准，细化轴依赖的尺度、旋转和位移。经过校准的航点和模式标签指导一个规划器，该规划器结合测量几何进行机器人执行。实际实验展示了在七个室内和室外场景中的导航。在每轮五个候选视频的情况下，所有七个场景在两轮内均获得可用视频。在校准观察中，与NavDreamer相比，我们的方法将平均绝对深度误差减少了87.7%，平均绝对相对深度误差减少了86.3%。

View on arXiv Download PDF AI Translation

cs.RO / 53 / 2609.19846

Improving Cross-embodiment Transfer in Latent Action Models with Action-Similarity Supervision

通过动作相似性监督改善潜在动作模型中的跨体现迁移

Alvarez, Maxime, Caballero, Renzo, Matsushima, Tatsuya, Iwasawa, Yusuke, Matsuo, Yutaka

Abstract

As generalist robot policies gain vision and language from web-scale pretraining, demonstrations remain costly to collect and tied to the robot that recorded them. Latent action models (LAMs) address both by learning latent actions from action-free videos that can be shared across embodiments, however, in practice, LAMs are sensitive to background visual noise, and the same motion from two different robots may be encoded with different latents. One solution to the background visual noise is to add an auxiliary loss predicting the robot action from the latent action, further associating the latent action space to the embodiment specific robot action space. We study a different use of the same labels, through action-similarity supervision. The similarity between any two latent actions is trained to match the similarity of the two ground-truth robot action sequences. The ground-truth actions are never predicted by the LAM, so the latent action does not need to encode embodiment specifics. We evaluate cross-embodiment transfer on RoboTwin 2.0 in a controlled setup, two bimanual robots demonstrate disjoint task sets, a policy is trained on all the demonstrations, and each robot is evaluated closed-loop on the tasks only the other demonstrated. With the policy architecture and its hyperparameters, the dataset, and the evaluation protocol fixed, predicting latent actions instead of ground-truth actions more than doubles cross-embodiment success. Given the same ground-truth actions, similarity supervision transfers better than an auxiliary loss that predicts the ground-truth action during the LAM training. Computing the similarities on end-effector motion rather than joint-space motion, and letting the loss compare latent actions across the two robots, gives the best approach of the study.

Chinese Translation

随着通用机器人策略通过大规模预训练获得视觉和语言，演示的收集仍然成本高昂，并且与记录它们的机器人紧密相关。潜在动作模型（LAMs）通过从无动作视频中学习潜在动作来解决这两个问题，这些潜在动作可以在不同的体现之间共享。然而，在实践中，LAMs 对背景视觉噪声敏感，并且来自两个不同机器人的相同动作可能会被编码为不同的潜在表示。解决背景视觉噪声的一个方案是增加一个辅助损失，从潜在动作预测机器人动作，进一步将潜在动作空间与特定体现的机器人动作空间关联起来。我们研究了相同标签的不同使用方式，通过动作相似性监督。任何两个潜在动作之间的相似性被训练为与两个真实机器人动作序列的相似性相匹配。真实动作从未被 LAM 预测，因此潜在动作不需要编码体现特定信息。我们在受控环境中评估 RoboTwin 2.0 上的跨体现迁移，两个双手机器人展示不重叠的任务集，一个策略在所有演示上进行训练，并且每个机器人仅在另一个机器人演示的任务上进行闭环评估。在固定策略架构及其超参数、数据集和评估协议的情况下，预测潜在动作而非真实动作使跨体现成功率提高了两倍以上。在相同的真实动作下，相似性监督的迁移效果优于在 LAM 训练期间预测真实动作的辅助损失。计算末端执行器运动的相似性而不是关节空间运动，并让损失在两个机器人之间比较潜在动作，提供了本研究的最佳方法。

View on arXiv Download PDF AI Translation

cs.RO / 54 / 2609.19850

GR2PO: Group Relative Return Policy Optimization for Continuous Robot Control

GR2PO：用于连续机器人控制的群体相对回报策略优化

Wang, Pengqin, Zhang, Qiming, Shen, Shaojie, Ma, Jun

Abstract

Actor-critic architecture has been widely used in continuous robot control. However, they rely on learning a value network, introducing additional computational overhead during training. Moreover, policy learning may also be affected by the approximation error of value estimation. Critic-free group relative policy optimization methods provide a simpler training approach by removing the need for a critic. However, they fail to learn long-term action outcomes when directly applying immediate rewards to policy optimization in dense-reward environments. To address these problems, we propose Group Relative Return Policy Optimization (GR2PO), a critic-free reinforcement learning framework for continuous robot control. GR2PO estimates the discounted returns from the parallelly collected trajectories, performs group normalization at each rollout time index, and uses relative advantages and clipped targets to update the policy. To evaluate the effectiveness of the proposed framework, we instantiate it on robot control simulation environments and deploy the model to a real-world edge device. The results show that GR2PO significantly outperforms critic-free baselines that use immediate rewards and performs competitively against state-of-the-art actor-critic methods. Furthermore, GR2PO demonstrates competitive training efficiency. Inference tests on NVIDIA Jetson TX2 demonstrate the feasibility of deploying the learned policies on edge platforms. Further ablation experiments analyze the effects of parallel group size, return estimation methods, and target clipping ratio on learning performance. To support follow-up research, we will make the complete code publicly available after the paper is accepted, including the framework implementation, experimental configuration, and training and evaluation scripts.

Chinese Translation

演员-评论家架构在连续机器人控制中得到了广泛应用。然而，它们依赖于学习价值网络，这在训练过程中引入了额外的计算开销。此外，策略学习也可能受到价值估计的近似误差的影响。无评论家的群体相对策略优化方法通过去除评论家的需求提供了一种更简单的训练方法。然而，在密集奖励环境中，直接将即时奖励应用于策略优化时，它们无法学习长期的行动结果。为了解决这些问题，我们提出了群体相对回报策略优化（GR2PO），这是一种用于连续机器人控制的无评论家强化学习框架。GR2PO从并行收集的轨迹中估计折扣回报，在每个回放时间索引处执行群体归一化，并使用相对优势和剪切目标来更新策略。为了评估所提框架的有效性，我们在机器人控制仿真环境中实例化了它，并将模型部署到真实世界的边缘设备上。结果表明，GR2PO显著优于使用即时奖励的无评论家基线，并在与最先进的演员-评论家方法的比较中表现出竞争力。此外，GR2PO展示了竞争性的训练效率。在NVIDIA Jetson TX2上的推理测试证明了在边缘平台上部署学习到的策略的可行性。进一步的消融实验分析了并行群体大小、回报估计方法和目标剪切比率对学习性能的影响。为了支持后续研究，我们将在论文被接受后公开完整代码，包括框架实现、实验配置以及训练和评估脚本。

View on arXiv Download PDF AI Translation

cs.RO / 55 / 2609.19863

Feeling Terrain Before Crossing: World Models for Off-Road Navigation

穿越前感知地形：越野导航的世界模型

Son, E-In, Kim, Dong-Wook, Hwang, Ji-Hoon, Lee, Kangsun, Bae, Jisung, Kim, Jung-Taak, Seo, Seung-Woo

Abstract

Navigation world models plan by foresight, predicting the future that each candidate action sequence produces and selecting the best, rather than mapping observations to actions directly. Unlike urban settings where a predicted scene is a sufficient proxy, off-road navigation hinges on the robot--terrain interaction, so the prediction must cover not only what the camera will see but what the robot will feel. However, existing scene-focused models do not predict how much the robot will slip, tilt or shake along a planned trajectory. Proprioception captures these dynamics directly and, when used as input, improves the prediction of the physical future. We present Feel-WM, the first off-road navigation world model that conditions on proprioception and predicts what the robot will feel alongside what the camera will see. The physical future takes the form of a future proprioceptive state and a failure risk, both learned from the robot's own experience without human labels. The planner rolls out the physical future alongside the scene and weighs the predicted failure risk against goal similarity in a separable score. Experiments on real off-road data and in simulation demonstrate that Feel-WM outperforms visual-only navigation world models in open-loop planning and closed-loop rough-terrain navigation across wheeled and legged platforms. Deployed on a Husky on mountain trails, Feel-WM plans onboard, predicts rough ground ahead and steers around it, completing courses that an end-to-end policy fails.

Chinese Translation

导航世界模型通过前瞻性规划，预测每个候选动作序列产生的未来，并选择最佳方案，而不是直接将观察映射到动作上。与城市环境中预测场景作为充分代理不同，越野导航依赖于机器人与地形的互动，因此预测不仅必须涵盖相机所见的内容，还必须包括机器人所感受到的内容。然而，现有的以场景为中心的模型并未预测机器人在规划轨迹上的滑动、倾斜或震动程度。自我感知（proprioception）直接捕捉这些动态，并在作为输入时改善物理未来的预测。我们提出了Feel-WM，这是第一个基于自我感知的越野导航世界模型，能够预测机器人将感受到的内容以及相机将看到的内容。物理未来以未来的自我感知状态和失败风险的形式呈现，这两者均从机器人的自身经验中学习，而无需人工标签。规划器在场景的基础上展开物理未来，并在可分离的评分中权衡预测的失败风险与目标相似性。对真实越野数据和模拟的实验表明，Feel-WM在开环规划和闭环粗糙地形导航中，优于仅依赖视觉的导航世界模型，适用于轮式和腿式平台。在山地小径上部署的Husky上，Feel-WM能够进行车载规划，预测前方的粗糙地面并绕过它，完成了一个端到端策略无法完成的课程。

View on arXiv Download PDF AI Translation

cs.RO / 56 / 2609.19894

Learning Reliable Parking Policies via Offline Reinforcement Learning with Quantized Action Representations

通过离线强化学习与量化动作表示学习可靠的停车策略

Yang, Zewei, Peng, Zengqi, Ma, Jun

Abstract

Parking is a routine yet safety-critical task for autonomous vehicles operating in urban environments. However, cluttered and weakly structured parking spaces, compounded by the interactive uncertainty from surrounding vehicles, hinder reliable maneuver generation. To address these challenges, we develop a waypoint-level offline reinforcement learning framework for interaction-aware autonomous parking. Specifically, a dedicated parking dataset is constructed from hierarchical expert rollouts with rotational waypoint augmentation, covering both non-interactive scenarios and interactive ones. The policy is then conditioned on a compact state representation, in which LiDAR-based obstacle features are adapted to the target pose via feature-wise linear modulation. A state-conditioned tokenizer further quantizes continuous waypoint sequences into discrete action tokens, over which conservative Q-learning is performed to suppress value overestimation on poorly supported actions. Extensive closed-loop experiments are conducted in the high-fidelity CARLA simulator. The proposed framework attains the highest parking success rate among all baselines and transfers reliably to unseen parking slots.

Chinese Translation

停车是自动驾驶车辆在城市环境中进行的日常但至关重要的任务。然而，杂乱且结构不清晰的停车空间，加上周围车辆带来的交互不确定性，阻碍了可靠的机动生成。为了解决这些挑战，我们开发了一种面向交互的自动停车的航点级离线强化学习框架。具体而言，我们从分层专家回放中构建了一个专用的停车数据集，并通过旋转航点增强，涵盖了非交互场景和交互场景。然后，策略基于紧凑的状态表示进行条件化，其中基于激光雷达的障碍物特征通过特征线性调制适应目标姿态。状态条件化的标记器进一步将连续航点序列量化为离散动作标记，并在其上执行保守的Q学习，以抑制对支持不足动作的价值高估。在高保真度的CARLA模拟器中进行了广泛的闭环实验。所提出的框架在所有基线中实现了最高的停车成功率，并可靠地转移到未见过的停车位。

View on arXiv Download PDF AI Translation

cs.RO / 57 / 2609.19906

Learning and Transferring Closed-Loop Robot Software

学习与迁移闭环机器人软件

Kuroki, So, Tang, Yujin

Abstract

Closed-loop robot policies require observation processing, state management, and situation-dependent branching, making them costly to design and tune manually. Although coding agents increasingly support control-code generation and optimization, it remains unclear whether implementations improved on source tasks also support policy acquisition for new tasks. We study this question by treating complete closed-loop implementations as reusable execution experience. For each source task, a coding agent generates policy code from a few successful demonstrations and iteratively improves it using simulation feedback. The validation-selected implementations are retained in a software archive. For new tasks, the agent generates and improves policies using archived implementations, target demonstrations, and execution feedback. The resulting policy is then frozen and executes without further model calls. Across four source tasks in RoboCasa, iterative optimization increases mean success from 28.3% to 64.2%. Across nine target tasks and three independent runs, mean success is 45.2% without references, 41.5% with initial source code, and 57.0% with optimized source code. Optimized references outperform initial references in all three runs on the nine-task average, with a mean gain of 15.6 percentage points. These results demonstrate the value of execution-improved software as a resource for acquiring new policies in this setting, although initial references remain better on two target tasks when averaged across runs.

Chinese Translation

闭环机器人策略需要观察处理、状态管理和依赖情境的分支，这使得它们的设计和调优成本高昂。尽管编码代理越来越支持控制代码的生成和优化，但尚不清楚在源任务上改进的实现是否也支持新任务的策略获取。我们通过将完整的闭环实现视为可重用的执行经验来研究这个问题。对于每个源任务，编码代理从少量成功的演示中生成策略代码，并使用仿真反馈进行迭代改进。经过验证选择的实现被保留在软件档案中。对于新任务，代理使用归档的实现、目标演示和执行反馈生成和改进策略。最终生成的策略被固定并在没有进一步模型调用的情况下执行。在RoboCasa的四个源任务中，迭代优化将平均成功率从28.3%提高到64.2%。在九个目标任务和三次独立运行中，平均成功率在没有参考的情况下为45.2%，使用初始源代码时为41.5%，而使用优化源代码时为57.0%。在九个任务的平均值上，优化参考在三次运行中均优于初始参考，平均增益为15.6个百分点。这些结果展示了执行改进软件作为在此环境中获取新策略资源的价值，尽管在跨运行平均时，初始参考在两个目标任务上仍然表现更好。

View on arXiv Download PDF AI Translation

cs.RO / 58 / 2609.19912

Distributed Model Predictive Control with Connectivity-based Contracts

基于连通性合同的分布式模型预测控制

Geurts, Jorit, Saccani, Danilo, Zeilinger, Melanie N., Carron, Andrea

Abstract

Teams of mobile robots rely on continuous communication with their neighbors for coordination, yet most distributed model predictive control (DMPC) schemes assume the communication network stays connected rather than actively enforcing it. Adding such a guarantee is hard since the usual mathematical condition for connectivity is nonconvex and links every agent to every other, which is incompatible with a scalable distributed real-time controller. We propose a DMPC framework in which each agent is assigned a connectivity contract: a local region prescribing where its predicted positions may lie over the prediction horizon. The contracts are designed so that, as long as every agent stays within its own contract, the team is guaranteed to remain connected. Given the maintained contract graph, an agent builds its contract from a single exchange with its immediate neighbors, after which every agent solves its own optimization problem independently. We prove that the resulting closed-loop system maintains connectivity, avoids collisions, and respects local state and input constraints. Simulation and hardware experiments on miniature autonomous car-like robots demonstrate the approach.

Chinese Translation

移动机器人团队依赖于与邻居的持续通信进行协调，但大多数分布式模型预测控制（DMPC）方案假设通信网络保持连通，而不是主动强制这一点。增加这样的保证是困难的，因为通常的连通性数学条件是非凸的，并且将每个代理与其他每个代理相连，这与可扩展的分布式实时控制器不兼容。我们提出了一种DMPC框架，其中每个代理被分配一个连通性合同：一个局部区域，规定其预测位置在预测视界内可能的位置。合同的设计使得，只要每个代理保持在自己的合同范围内，团队就能保证保持连通。在维护的合同图下，代理通过与其直接邻居的单次交换构建自己的合同，之后每个代理独立解决自己的优化问题。我们证明了所得到的闭环系统保持连通，避免碰撞，并遵守局部状态和输入约束。对微型自主汽车类机器人的仿真和硬件实验验证了该方法。

View on arXiv Download PDF AI Translation

cs.RO / 59 / 2609.19923

Co-VLA: Consensus-based Federated Training for Vision-Language-Action Models

Co-VLA：基于共识的视觉-语言-动作模型的联邦训练

Li, Haolong, Er, Guner Dilsad, Muehlebach, Michael, Stueckler, Joerg

Abstract

Vision-language-action models (VLAs) have emerged as a promising paradigm for general-purpose robot learning, with performance improving as models and datasets scale. Scaling robot data collection, however, remains challenging because data are naturally distributed across robots, tasks, and locations, making centralization costly or impractical. Federated learning offers a way to train on decentralized robot data, but applying it to VLAs requires accounting for heterogeneous robot client data distributions. We present Co-VLA, which applies consensus optimization using the Alternating Direction Method of Multipliers~(ADMM) to federated VLA training. We show that the same algorithm supports both full-model training and parameter-efficient fine-tuning with both fixed-rank and rank-adaptive adapters. The name Co-VLA reflects both consensus and collaboration: clients with different local robot datasets collaboratively train a shared model without sharing their data. Our experiments demonstrate that Co-VLA achieves performance comparable to centralized training in both full-model training and parameter-efficient fine-tuning settings.

Chinese Translation

视觉-语言-动作模型（VLA）已成为通用机器人学习的一个有前景的范式，随着模型和数据集的规模扩大，其性能不断提升。然而，机器人数据收集的规模化仍然面临挑战，因为数据自然分布在不同的机器人、任务和位置之间，使得集中化变得成本高昂或不切实际。联邦学习提供了一种在去中心化的机器人数据上进行训练的方法，但将其应用于VLA时需要考虑异构机器人客户端数据分布。我们提出了Co-VLA，它利用交替方向乘子法（ADMM）进行共识优化，以实现联邦VLA训练。我们展示了同一算法支持全模型训练和参数高效微调，适用于固定秩和自适应秩的适配器。Co-VLA的名称反映了共识与协作：具有不同本地机器人数据集的客户端在不共享数据的情况下协同训练共享模型。我们的实验表明，Co-VLA在全模型训练和参数高效微调设置中实现了与集中训练相当的性能。

View on arXiv Download PDF AI Translation

cs.RO / 60 / 2609.19946

Execution-Aware Pre-Execution Ranking for Grasp-Conditioned Robotic Placement

执行感知的预执行排名用于抓取条件下的机器人放置

Liu, Tianyuan, Patamia, Rutherford Agbeshi, Champion, Benjamin, Dazeley, Richard, Cosgun, Akansel

Abstract

A geometrically valid placement can still be difficult to execute because the selected grasp changes the required end-effector pose, collision geometry, and transport motion. Placement is formulated as a pre-execution ranking problem in which supplied grasp-placement candidates are scored before planning. The model combines a typed target-conditioned point cloud with three pose descriptors and hierarchical heads for planning success and execution success conditioned on planning. On a 30-object, 1,235-scene dataset with scene-group-held-out splits, three-seed top-1 success on covered test groups reaches 85.63 +/- 1.08% for joint selection and 79.84 +/- 0.16% for fixed-target ranking. For the designated frozen seed-42 checkpoint, top-1 success improves from 72.84% to 85.78% over full-pool cuMotion for joint ranking and from 59.65% to 79.67% for fixed-target ranking. Frozen transfer to xArm7/MoveIt requires no xArm-specific retraining. Across 27 locked cases, 13 complete end to end (48.15%). Of the 16 cases that pass Top-5 preflight and begin execution, 13 succeed (81.25%). Candidate-level deployment-feasibility prediction reaches 81.25% recall, 85.20% specificity, and 83.23% balanced accuracy.

Chinese Translation

几何有效的放置仍然可能难以执行，因为所选择的抓取会改变所需的末端执行器姿态、碰撞几何形状和运输运动。放置被表述为一个预执行排名问题，其中提供的抓取-放置候选项在规划之前被评分。该模型结合了类型化的目标条件点云、三个姿态描述符以及用于规划成功和基于规划的执行成功的层次化头部。在一个包含30个物体和1,235个场景的数据集中，场景组保留分割的情况下，覆盖测试组的三种种子顶级成功率在联合选择上达到85.63 +/- 1.08%，在固定目标排名上达到79.84 +/- 0.16%。对于指定的冻结种子-42检查点，联合排名的顶级成功率从72.84%提高到85.78%，而固定目标排名则从59.65%提高到79.67%。冻结迁移到xArm7/MoveIt不需要针对xArm的特定再训练。在27个锁定案例中，有13个实现了完整的端到端（48.15%）。在16个通过Top-5预飞行并开始执行的案例中，有13个成功（81.25%）。候选级别的部署可行性预测达到81.25%的召回率、85.20%的特异性和83.23%的平衡准确率。

View on arXiv Download PDF AI Translation

cs.RO / 61 / 2609.19954

LapaTrack-3D: 6 DoF pre-operative shape tracking for laparoscopic surgery

LapaTrack-3D：用于腹腔镜手术的六自由度术前形状跟踪

Song, Jingwei, Jakir, Javid Hussain, Zhang, Ray, Zhang, Wenwei, Zhou, Hao, Xian, Xiaomeng, Ghaffari, Maani

Abstract

This work proposes a real-time 6 Degree-of-Freedom (6 DoF) tracking algorithm for monocular laparoscopic surgery. It provides alignment between intra-operative video and pre-operative data (e.g., CT). The 6 DoF tracking offers a solution for accurately locating the internal anatomy of the target organ despite the lack of tactile feedback and transparency. The ORB-SLAM2 framework is adopted and modified for prior-based 3D tracking with four major modifications. First, the primitive 3D shape is used for fast initialization of the ORB-SLAM2 monocular mode. Second, a pseudo-segmentation strategy is employed to separate the target organ from the background for tracking. Third, the 3D shape is incorporated as a geometric prior in its pose graph optimization. Fourth, the Multi-Scale Retinex with Chromaticity Preservation (MSRCP) algorithm is leveraged and modified for image enhancement in challenging illumination scenarios. In-vivo and ex-vivo experiments validate that LapaTrack-3D provides robust 3D tracking and effectively handles typical challenges such as poor illumination, fast motion, out-of-field-of-view scenarios, partial visibility, and ``organ-background'' relative motion. LapaTrack-3D achieves a processing rate of 13 Hz for 1280*720 pixel video.

Chinese Translation

本研究提出了一种用于单目腹腔镜手术的实时六自由度（6 DoF）跟踪算法。该算法实现了术中视频与术前数据（例如CT）的对齐。六自由度跟踪为准确定位目标器官的内部解剖结构提供了解决方案，尽管缺乏触觉反馈和透明度。采用并修改了ORB-SLAM2框架，以实现基于先验的三维跟踪，并进行了四项主要修改。首先，使用原始三维形状快速初始化ORB-SLAM2单目模式。其次，采用伪分割策略将目标器官与背景分离以进行跟踪。第三，将三维形状作为几何先验纳入其位姿图优化中。第四，利用并修改了多尺度Retinex与色度保留（MSRCP）算法，以在挑战性照明场景中增强图像。体内和体外实验验证了LapaTrack-3D提供了稳健的三维跟踪，并有效应对了典型挑战，如光照不足、快速运动、视野外场景、部分可见性以及“器官-背景”相对运动。LapaTrack-3D在1280*720像素视频中实现了13 Hz的处理速率。

View on arXiv Download PDF AI Translation

cs.RO / 62 / 2609.19962

Hybrid Residual Reinforcement Learning for Contact-Rich Robotic Book Insertion

用于接触丰富的机器人书籍插入的混合残差强化学习

Liu, Tianyuan, Patamia, Rutherford Agbeshi, Champion, Benjamin, Cosgun, Akansel, Dazeley, Richard

Abstract

Placing a grasped book into a tight shelf is a compact but difficult contact-rich control problem: millimetre-scale pose error can turn a geometrically valid approach into jamming, failed release, or incomplete seating. We study this final phase after grasp acquisition and global approach, and ask how control authority should be divided between known geometry and learned behaviour. Our method retains a nominal task-space controller for structured insertion and seating, while residual PPO supplies bounded local corrections and decides when to release. Only the brief open-retreat-reclose transition is scripted. For the final policy used on hardware, a deployment-matched simulation evaluation over 512 fixed conditions yields 98.50 percent mean success (0.23 percentage-point sample SD) across three independent training runs, compared with 37.89 percent for nominal control. On the physical xArm7, 60 trials over 30 matched conditions show the same qualitative advantage: residual control raises success from 26.7 percent to 63.3 percent, reduces failures from 22 to 11, and wins 13 of the 15 matched conditions in which the two controllers differ. Robustness tests show that performance remains above 87 percent under initialization perturbations up to 1.5x, while very tight clearances expose the geometric limit of local correction. These results support a hybrid design in which geometry preserves reliable task structure and learning is concentrated on the contact-sensitive behaviour that fixed rules handle poorly.

Chinese Translation

将抓取的书籍放入紧凑的书架中是一个紧凑但困难的接触丰富控制问题：毫米级的姿态误差可能会将几何上有效的方法转变为卡住、释放失败或未完全就位。我们研究这一抓取获取和全局接近后的最终阶段，并探讨控制权应如何在已知几何和学习行为之间进行分配。我们的方法保留了一个用于结构化插入和就位的名义任务空间控制器，同时残差 PPO（Proximal Policy Optimization）提供有界的局部修正并决定何时释放。只有简短的开放-撤退-重新关闭过渡是脚本化的。对于在硬件上使用的最终策略，针对 512 个固定条件的部署匹配模拟评估显示，三次独立训练运行的平均成功率为 98.50%（样本标准差为 0.23 个百分点），而名义控制的成功率为 37.89%。在物理 xArm7 上，60 次试验在 30 个匹配条件下显示出相同的定性优势：残差控制将成功率从 26.7% 提高到 63.3%，将失败次数从 22 次减少到 11 次，并在两种控制器存在差异的 15 个匹配条件中赢得了 13 次。鲁棒性测试表明，在初始化扰动高达 1.5 倍的情况下，性能仍保持在 87% 以上，而非常紧密的间隙暴露了局部修正的几何极限。这些结果支持一种混合设计，其中几何保持可靠的任务结构，而学习则集中于固定规则处理不佳的接触敏感行为。

View on arXiv Download PDF AI Translation

cs.RO / 63 / 2609.19974

MaskHarness-WAM: Instance-Grounded Harnessing for Long-Horizon Robot Manipulation

MaskHarness-WAM：基于实例的长时间机器人操控方案

Huang, Zitai, Su, Taiyi, Zhu, Jian, Zhang, Jianjun, Ma, Chong, Liu, Tianbin, Lu, Weiyi, Xu, Yi, Wang, Hanli

Abstract

Long-horizon robot manipulation requires not only stable local visuomotor control, but also continuous target tracking and reliable task progress assessment throughout execution. This challenge becomes particularly critical when multiple objects share identical appearances and must be manipulated in a prescribed order. In such scenarios, relying solely on a limited-horizon manipulation policy is often insufficient to determine which instance should be operated on and when the task should transition to the next stage. To address this challenge, we propose MaskHarness-WAM, an instance-grounded harness for long-horizon manipulation. The proposed system connects high-level task planning with low-level manipulation policies through target masks, while leveraging visual feedback for subtask scheduling and continuous execution. Since each subtask corresponds to a different target instance, the low-level policy requires a newly established initial target mask under the updated scene at each subtask transition. The harness continuously re-observes the environment, generates, and verifies the target mask at subtask boundaries, thereby updating the instance-level spatial condition provided to the low-level policy. Furthermore, the system advances the manipulation process by switching target instances according to the verified completion status of each subtask. Experiments on a real robot platform demonstrate that MaskHarness-WAM substantially outperforms limited-horizon policies on sequential multi-object manipulation, showing its effectiveness in extending local manipulation skills to reliable long-horizon execution.

Chinese Translation

长时间的机器人操控不仅需要稳定的局部视觉运动控制，还需要在执行过程中持续的目标跟踪和可靠的任务进展评估。当多个物体具有相同外观并且必须按照规定顺序进行操控时，这一挑战尤为重要。在这种情况下，仅依赖有限时间范围的操控策略往往不足以确定应操作哪个实例以及何时将任务过渡到下一个阶段。为了解决这一挑战，我们提出了MaskHarness-WAM，一种用于长时间操控的基于实例的方案。该系统通过目标掩码将高层次的任务规划与低层次的操控策略连接起来，同时利用视觉反馈进行子任务调度和持续执行。由于每个子任务对应于不同的目标实例，低层次策略需要在每次子任务过渡时根据更新的场景建立新的初始目标掩码。该方案持续重新观察环境，在子任务边界生成并验证目标掩码，从而更新提供给低层次策略的实例级空间条件。此外，该系统通过根据每个子任务的验证完成状态切换目标实例来推进操控过程。在真实机器人平台上的实验表明，MaskHarness-WAM在顺序多物体操控方面显著优于有限时间范围的策略，展示了其在将局部操控技能扩展到可靠的长时间执行中的有效性。

View on arXiv Download PDF AI Translation

cs.RO / 64 / 2609.19976

Compliance for Free: Learning Identifiable Impedance via Bilateral Teleoperation

免费的合规性：通过双向遥操作学习可识别的阻抗

Guda, Harsha, Colomé, Adrià, Torras, Carme

Abstract

Vision-language-action models tell a robot where to move, but not how hard to push. Contact-rich tasks depend on that second quantity, compliance, yet no widely used demonstration interface records it. The obstacle is identifiability as realized pose and measured force cannot separate the operator's intended equilibrium from their stiffness, so VR controllers, SpaceMouse and handheld grippers cannot supply compliance supervision even in principle. Prior compliance-output policies work around this with hand-specified task structure, privileged simulation contact state, or dedicated force and tactile hardware. Four-channel bilateral teleoperation removes the ambiguity directly by using the leader arm as a separate measurement of the intended equilibrium, making per-axis stiffness identifiable by regression using only the joint-torque sensing already on the manipulator. This yields per-timestep, direction-dependent compliance labels at zero annotation cost, which we use to fine-tune a VLA to emit stiffness alongside pose. On a Franka Research 3 wiping task, ours is the only policy of five whose contact force changes when the instruction asks for a firm wipe rather than a normal one (6.4N (normal) to 9.1N (firm) RMS, Cohen's d = 0.89, p = 0.023

Chinese Translation

视觉-语言-动作模型告诉机器人该如何移动，但并未说明施加多大力度。接触丰富的任务依赖于第二个量，即合规性，但没有广泛使用的演示接口能够记录这一点。障碍在于可识别性，因为实际姿态和测量力无法将操作员的预期平衡与其刚度区分开，因此即使在原则上，虚拟现实控制器、SpaceMouse 和手持夹具也无法提供合规性监督。之前的合规性输出策略通过手动指定任务结构、特权的仿真接触状态或专用的力和触觉硬件来规避这一问题。四通道双向遥操作通过将领导臂作为预期平衡的单独测量，直接消除了模糊性，使得每个轴的刚度可以通过回归仅使用操纵器上已有的关节扭矩传感器来识别。这在零标注成本下产生了每个时间步、方向依赖的合规性标签，我们用这些标签来微调视觉-语言-动作模型，使其能够同时输出刚度和姿态。在 Franks Research 3 的擦拭任务中，我们的策略是五个策略中唯一一个在指令要求进行坚实擦拭而非正常擦拭时，接触力发生变化的策略（6.4N（正常）到 9.1N（坚实）均方根，Cohen's d = 0.89，p = 0.023）。

View on arXiv Download PDF AI Translation

cs.RO / 65 / 2609.20035

DR-MPC: Fast and Feasible Dynamics-Relaxed Model-Predictive Control for Legged Locomotion

DR-MPC：快速且可行的动态放松模型预测控制用于四足运动

Wang, Run, Tuerxun, Alapati, Liu, Shuo, Xiao, Wei, Drgoňa, Ján, Mo, Yilin, Wu, Liang

Abstract

This paper presents dynamics-relaxed model predictive control (DR-MPC), a novel MPC formulation for legged locomotion, and a tailored interior-point method (IPM) solver. The formulation combines online optimization feasibility by construction with a contact-aware input parameterization. DR-MPC moves the dynamics equality and affine input constraints into quadratic penalties and retains only nonempty box constraints. The resulting box-constrained quadratic program (QP) has a block-arrow Hessian that enables the state and affine-output directions to be eliminated through a Schur complement. The solver factors only the reduced control system after swing-force elimination and contact-aligned move blocking. For the evaluated implementations using the same DR-MPC formulation, our method achieves median end-to-end MPC speedups of $16.0\times$ over HPIPM and $4.4\times$ over OSQP, with comparable locomotion performance in simulation. DR-MPC achieves a median onboard MPC end-to-end time of $4.4$ ms and is validated on a Unitree Go1 quadruped. Open-source code will be made available after publication.

Chinese Translation

本文提出了一种动态放松模型预测控制（DR-MPC），这是一种针对四足运动的新型模型预测控制（MPC）公式，以及一个定制的内点法（IPM）求解器。该公式通过构造实现了在线优化的可行性，并结合了考虑接触的输入参数化。DR-MPC将动态等式和仿射输入约束转化为二次惩罚，并仅保留非空的盒约束。由此产生的盒约束二次规划（QP）具有块箭头海森矩阵，使得状态和仿射输出方向可以通过舒尔补消去。求解器仅在摆动力消除和接触对齐移动阻塞之后对简化的控制系统进行因式分解。对于使用相同DR-MPC公式的评估实现，我们的方法在端到端MPC速度上相较于HPIPM实现了$16.0 imes$的中位数加速，相较于OSQP实现了$4.4 imes$的中位数加速，同时在仿真中保持了可比的运动表现。DR-MPC在Unitree Go1四足机器人上验证了中位数的MPC端到端时间为$4.4$毫秒。开源代码将在发表后提供。

View on arXiv Download PDF AI Translation

cs.RO / 66 / 2609.20048

Mechanical Precision Weeding with a Quadruped Robot

四足机器人机械精确除草

Beumer, Ruben, Janssen, Tom, van de Molengraft, René, Antunes, Duarte

Abstract

Herbicide-based weed control is increasingly unsustainable due to rising weed resistance and the adverse environmental impacts of chemical use. While mechanical weed control avoids these drawbacks, it is typically implemented using large machines that cause soil compaction. We propose a novel alternative based on small mobile robots for mechanical weeding. Compared with existing automated mechanical weeding approaches, the proposed method offers reduced soil compaction, simpler automation, and improved scalability. Our solution involves a Boston Dynamics Spot quadruped robot equipped with a custom weed removal tool featuring a milling bit at its end. The tool is rigidly attached to the robot and uses the degrees of freedom of the robot base by actuating the legs, while keeping the feet stationary. We develop a software architecture that enables autonomous weed removal and integrate this system with all other required components. We analyze the accuracy and efficiency of the current proof of concept both in an indoor and outdoor environment and provide recommendations for future work to make the system more accurate and efficient.

Chinese Translation

基于除草剂的杂草控制因杂草抗药性上升和化学物质使用对环境的不利影响而日益不可持续。虽然机械除草避免了这些缺点，但通常使用的大型机器会导致土壤压实。我们提出了一种基于小型移动机器人进行机械除草的新颖替代方案。与现有的自动化机械除草方法相比，所提方法具有减少土壤压实、简化自动化和提高可扩展性的优势。我们的解决方案涉及一款配备定制除草工具的波士顿动力Spot四足机器人，该工具末端装有铣刀。该工具刚性连接到机器人，并通过驱动机器人腿部的自由度来操作，同时保持脚部静止。我们开发了一种软件架构，使得机器人能够自主除草，并将该系统与所有其他所需组件集成。我们分析了当前概念验证在室内和室外环境中的准确性和效率，并提供了未来工作的建议，以提高系统的准确性和效率。

View on arXiv Download PDF AI Translation

cs.RO / 67 / 2609.20078

FlipToSee: A Probabilistic Stable Placement Prior for Active Visual Exploration via Regrasping

FlipToSee：一种用于通过重新抓取进行主动视觉探索的概率稳定放置先验

Shu, Chang, Dinesh, Sushil Samuel, Park, Shinkyu

Abstract

Active visual exploration of tabletop objects often requires reorienting an unknown resting object onto a different stable support face to expose occluded surfaces. To identify such placements without exhaustive physical search, we learn a probabilistic placement prior from a single-view point cloud. Stable placement prediction is inherently multimodal, and conventional 6-DoF regression introduces further ambiguity by modeling translation and in-plane yaw. We therefore propose FlipToSee, a probabilistic framework that removes this representational ambiguity by parameterizing placements as unit support normals on $S^2$ while modeling their multimodal conditional distribution via a von Mises--Fisher mixture density network. To decouple mode diversity from physical robustness, FlipToSee deterministically extracts a compact candidate set from the mixture components and applies robustness-aware reranking using an auxiliary head trained with candidate-aligned supervision. In simulation, FlipToSee achieves $98.4\%$ first-proposal success on in-distribution objects, $95.3\%$ on out-of-distribution shapes, and $90.0\%$ under zero-shot transfer to household YCB objects. We further demonstrate the learned placement prior on a physical robot by integrating it with grasp and motion planning for exploratory regrasping.

Chinese Translation

桌面物体的主动视觉探索通常需要将未知的静止物体重新定向到不同的稳定支撑面上，以暴露被遮挡的表面。为了在不进行耗时的物理搜索的情况下识别此类放置，我们从单视角点云中学习概率放置先验。稳定放置预测本质上是多模态的，而传统的6自由度回归通过建模平移和面内偏航引入了进一步的模糊性。因此，我们提出了FlipToSee，一个概率框架，通过将放置参数化为$S^2$上的单位支撑法线，消除了这种表示模糊性，同时通过von Mises–Fisher混合密度网络建模其多模态条件分布。为了将模式多样性与物理鲁棒性解耦，FlipToSee从混合成分中确定性地提取出一个紧凑的候选集，并使用经过候选对齐监督训练的辅助头进行鲁棒性感知的重新排序。在仿真中，FlipToSee在分布内物体上的首次提案成功率达到98.4%，在分布外形状上为95.3%，在对家庭YCB物体的零样本转移中为90.0%。我们进一步通过将学习到的放置先验与抓取和运动规划相结合，展示了在物理机器人上的应用，以实现探索性重新抓取。

View on arXiv Download PDF AI Translation

cs.RO / 68 / 2609.20087

Graph-Based Design of Soft Grippers with Multi-Objective Quality-Diversity Optimisation

基于图的软抓手设计与多目标质量-多样性优化

Farinha, Andre, Shi, Ge, Bowman, Harry, Tidd, Brendan, Howard, David, Pinskier, Josh

Abstract

Effective manipulation across diverse objects is critical for applications ranging from agricultural harvesting to laboratory and domestic automation. While the inherent compliance of soft robotics is well suited to this challenge, designing grippers that generalize across tasks remains difficult due to the vast design space of continuum mechanics and the risk of overfitting to specific scenarios. We propose a graph-based design space for representing soft structures and mechanisms, coupled with a multi-objective, diversity-driven genetic optimization framework that explicitly promotes solution variety throughout the design process. Using multiple grasping scenarios during optimization, we study how task diversity influences the emergence of generalization to unseen objects and contact conditions. Our results show that optimization over a sufficiently diverse set of grasping cases leads to designs with emergent generalization, exhibiting improved robustness compared to task-specific solutions on novel scenarios. These findings suggest that diversity-driven optimization offers a principled pathway toward general-purpose soft grippers, aligned with the adaptable nature of soft robotics.

Chinese Translation

在农业收获、实验室和家庭自动化等应用中，有效地操作多样化物体至关重要。尽管软机器人固有的柔顺性非常适合应对这一挑战，但由于连续介质力学的广泛设计空间以及对特定场景过拟合的风险，设计能够跨任务泛化的抓手仍然困难。我们提出了一种基于图的设计空间，用于表示软结构和机制，并结合一个多目标、以多样性驱动的遗传优化框架，明确促进设计过程中的解决方案多样性。通过在优化过程中使用多种抓取场景，我们研究了任务多样性如何影响对未见物体和接触条件的泛化能力。我们的结果表明，在足够多样化的抓取案例上进行优化会导致具有新兴泛化能力的设计，与特定任务解决方案相比，在新场景中表现出更好的鲁棒性。这些发现表明，以多样性驱动的优化为通用软抓手提供了一条有原则的途径，与软机器人的适应性特征相一致。

View on arXiv Download PDF AI Translation

cs.RO / 69 / 2609.20103

Safety-Critical Scenanrio Emerges from Initial Scene

从初始场景中产生安全关键场景

Wu, Yin, Wei, Jiarong, Esselborn, Carl, Phoolari, Shubham, Abouelazm, Ahmed, Slieter, Daniel, Zöllner, J. Marius

Abstract

Safety-critical driving scenario generation has largely focused on manipulating the behavior of surrounding agents while starting from an initial scene from driving data. This assumption can limit the space of discoverable failures, since driving data can provide little opportunity for meaningful interaction. For example, in the Waymo Open Motion Dataset, 20.44% of recorded slices feature a stationary ego vehicle that never moves, and 30.39% of initial frames contain no nearby traffic participants within 10 meters. We instead study safety-critical scenario generation as an initialization problem: given agnostic black-box driving policies, we learn to generate realistic initial scenes that are more likely to evolve into critical interactions. We propose AdvScene, a conditional latent diffusion model that is trained in two stages. Starting from pretraining on naturalistic driving data, we post-train the adversarial-agent generation branch using reinforcement learning with feedback from closed-loop simulator rollouts. Conditioning on ego driving displacement prevents the ego from remaining static, and RL finetuning induces criticality directly with non-differentiable safety-critical metrics. Experiments on the Waymo dataset across 12 combinations of ego and traffic policies show that our AdvScene substantially increases the rate of ego-fault collision events and TTC<3s events.

Chinese Translation

安全关键驾驶场景生成主要集中在操控周围代理的行为，同时从驾驶数据的初始场景出发。这一假设可能限制可发现故障的空间，因为驾驶数据提供了有限的有意义交互机会。例如，在Waymo开放运动数据集中，20.44%的记录片段中，静止的自我车辆从未移动，而30.39%的初始帧在10米内没有附近的交通参与者。我们将安全关键场景生成视为一个初始化问题：在不考虑具体驾驶策略的情况下，我们学习生成更可能演变为关键交互的现实初始场景。我们提出了AdvScene，一种条件潜在扩散模型，分两个阶段进行训练。从自然驾驶数据的预训练开始，我们使用闭环模拟器回放的反馈，通过强化学习对对抗代理生成分支进行后训练。对自我驾驶位移的条件限制防止自我车辆保持静止，而强化学习微调则直接利用不可微分的安全关键指标引发关键性。在Waymo数据集上进行的实验显示，在12种自我和交通策略组合下，我们的AdvScene显著增加了自我故障碰撞事件和TTC<3秒事件的发生率。

View on arXiv Download PDF AI Translation

cs.RO / 70 / 2609.20107

AnyViewDex: View-Invariant Dexterous Manipulation from RGB Observations

AnyViewDex：基于RGB观察的视角不变灵巧操作

Patil, Soham, Gunjal, Om Sanjay, Bhosale, Sourabh, Chavare, Arhan, Hora, Ramandeep Singh, Roy, Spandan

Abstract

Visuomotor policies for multi-fingered dexterous manipulation are highly sensitive to camera viewpoint shifts. To achieve view invariance, recent methods increasingly rely on explicit 3D modalities like RGB-D or point clouds, which can introduce hardware dependencies, calibration requirements, and vulnerability to sensor noise during real-world deployment. In this work, we show that view-invariant control can be achieved without explicit test-time 3D sensing by encoding geometric knowledge into the visual representation during simulation. We present AnyViewDex, an asymmetric training pipeline that combines multi-view contrastive alignment with privileged 3D geometric supervision. By regressing absolute 3D object coordinates during simulated training, this auxiliary objective provides a geometric grounding signal that mitigates the spatial collapse of the globally pooled contrastive embedding. At deployment, the policy operates zero-shot using only uncalibrated monocular RGB and proprioception. We validate this approach across both reinforcement learning and student-teacher distillation. In hardware evaluation on an xArm7 with a 16-DoF LEAP Hand, AnyViewDex reaches 76.7% grasping success across eight unseen objects and six uncalibrated viewpoints (480 trials; 2,400 across all ablation conditions), indicating that geometrically grounded monocular policies transfer zero-shot without test-time depth. Project Page: https://anyviewdex.github.io/

Chinese Translation

多指灵巧操作的视觉运动策略对相机视角的变化高度敏感。为了实现视角不变，近期的方法越来越依赖于显式的3D模态，如RGB-D或点云，这可能引入硬件依赖、校准要求以及在实际部署过程中对传感器噪声的脆弱性。在本研究中，我们展示了如何通过在仿真过程中将几何知识编码到视觉表示中，实现无需显式测试时3D感知的视角不变控制。我们提出了AnyViewDex，这是一种不对称训练管道，结合了多视角对比对齐和特权3D几何监督。通过在仿真训练中回归绝对3D物体坐标，这一辅助目标提供了几何基础信号，减轻了全局池化对比嵌入的空间崩溃。在部署时，该策略仅使用未校准的单目RGB和自我感知进行零-shot操作。我们在强化学习和师生蒸馏中验证了这一方法。在对xArm7和16自由度LEAP手的硬件评估中，AnyViewDex在八个未见物体和六个未校准视角下的抓取成功率达到76.7%（480次试验；在所有消融条件下共2400次），表明几何基础的单目策略在没有测试时深度的情况下实现了零-shot迁移。项目页面：https://anyviewdex.github.io/

View on arXiv Download PDF AI Translation

cs.RO / 71 / 2609.20114

Universal Navigation Interface: Robot-Free Data for Wheeled Robot Navigation

通用导航接口：无机器人数据用于轮式机器人导航

Prajapati, Sarvesh, Trivedi, Ananya, Genua, Lorena Maria, Moore, Drake, Maxwell, Bruce, Padir, Taskin

Abstract

Collecting real-world navigation data for mobile robots typically requires platform-specific teleoperation, making large-scale collection expensive and difficult to scale. We introduce Universal Navigation Interface (UNI), a robot-free data collection paradigm that uses a four-wheeled rollator walker (rollator) and smartphone to collect physically constrained human demonstrations. Because the rollator cannot climb stairs, negotiate uncut curbs, or pass through narrow gaps, demonstrations are naturally biased toward wheeled-feasible routes. Using UNI, we collect 37.2 km of real-world navigation data and recover metric trajectories that directly supervise goal-conditioned navigation models. Fine-tuning visual-navigation models on UNI reduces trajectory prediction error by 17.4-24.8% on held-out UNI demonstrations. Evaluation on other navigation datasets shows benefits that vary by dataset and metric. We further demonstrate closed-loop transfer to a powered wheelchair in curb, staircase, and curb-cut scenarios. These results support low-cost physical proxies as a practical source of navigation supervision collected without the target robot.

Chinese Translation

为移动机器人收集真实世界的导航数据通常需要特定平台的遥控操作，这使得大规模收集变得昂贵且难以扩展。我们提出了通用导航接口（Universal Navigation Interface, UNI），一种无机器人数据收集范式，利用四轮助行器（rollator）和智能手机收集受物理限制的人类示范。由于助行器无法爬楼梯、越过未切割的路缘或通过狭窄的缝隙，示范自然偏向于适合轮式行驶的路线。通过UNI，我们收集了37.2公里的真实世界导航数据，并恢复了直接监督目标条件导航模型的度量轨迹。在UNI上微调视觉导航模型使得轨迹预测误差降低了17.4%-24.8%，基于保留的UNI示范。对其他导航数据集的评估显示出不同数据集和度量的效益。我们进一步展示了在路缘、楼梯和路缘切割场景中向电动轮椅的闭环转移。这些结果支持低成本的物理代理作为一种实用的导航监督来源，且无需目标机器人进行数据收集。

View on arXiv Download PDF AI Translation

cs.RO / 72 / 2609.20116

GPT-6-Astra in a Navigation Workflow: Behavioral Analysis in Zero-Shot Vision-and-Language Navigation in Continuous Environments

导航工作流中的GPT-6-Astra：在连续环境中零-shot视觉与语言导航的行为分析

Dai, Guangzhao, Wu, Qi, Zhu, Bin

Abstract

We study GPT-6-Astra in a zero-shot Vision-and-Language Navigation in Continuous Environments (VLN-CE) system, where it interprets instructions, assesses its surroundings, and proposes actions. The system uses a common observation--decision--execution workflow with direct model API calls, without a packaged agent harness or navigation-specific fine-tuning. In this workflow, each request receives selected observations, execution feedback, and retained progress records. Evaluation covers the complete system, including context management and action control. We evaluate the system on 50 of the 100 R2R-CE val-unseen episodes used by Open-Nav. It achieves a success rate of 52.0\%, an SPL of 48.9\%, and an nDTW of 70.8\%. Our analysis highlights three findings. First, recorded responses link landmarks and earlier actions to instructions using observations and supplied history. Second, reviews include requests for additional views and revisions of uncertain judgments. Third, the results suggest a gap between task understanding and autonomous completion: an unfinished crossing is recognized while rotation continues. At termination, 36.0\% of episodes succeed with a workflow-accepted STOP, while another 16.0\% meet the distance criterion at the step limit. These results highlight a central challenge: translating correct local judgments into sustained progress and appropriate stopping.

Chinese Translation

我们研究了GPT-6-Astra在零-shot视觉与语言导航连续环境系统（VLN-CE）中的表现，该系统能够解释指令、评估周围环境并提出行动建议。该系统采用一种常见的观察-决策-执行工作流，通过直接的模型API调用进行操作，而不使用打包的代理工具或针对导航的特定微调。在此工作流中，每个请求都会接收选定的观察结果、执行反馈和保留的进度记录。评估涵盖整个系统，包括上下文管理和行动控制。我们在Open-Nav使用的100个R2R-CE验证集未见的50个剧集上对系统进行了评估。系统的成功率为52.0%，SPL为48.9%，nDTW为70.8%。我们的分析突出了三个发现。首先，记录的响应通过观察和提供的历史将地标和早期行动与指令联系起来。其次，审查中包括对额外视图的请求和对不确定判断的修正。第三，结果表明任务理解与自主完成之间存在差距：未完成的过马路被识别，而旋转仍在继续。在终止时，36.0%的剧集以工作流接受的STOP成功结束，而另有16.0%在步骤限制下满足距离标准。这些结果突显了一个核心挑战：将正确的局部判断转化为持续的进展和适当的停止。

View on arXiv Download PDF AI Translation

cs.RO / 73 / 2609.20191

VLN on the Fly: An Onboard Vision-Language Navigation Stack for Aerial Robots

飞行中的视觉-语言导航：一种用于空中机器人的机载视觉-语言导航系统

Tayar, Marco S., Tommaselli, Felipe, Capezutto, Gianluca, Saraiva, Pedro Antonio Rabelo, de Freitas, Pedro H. V., Kido, Lucas, Sonego, Guilherme, Godoy, Ricardo V., Becker, Marcelo

Abstract

Running vision-language navigation fully onboard an aerial robot is hard, since grounding, planning, and control must share limited compute and a single-stage error is difficult to isolate in flight. End-to-end aerial policies fuse these stages into one network, giving up the observability and safety checks a modular stack keeps available. We propose VLN on the Fly, an onboard stack that keeps grounding, planning, and control as separate, inspectable stages. A quantized VLM grounds an instruction to a coarse image cell, depth lifts it to a 3D goal, a fast B-spline planner returns a feasible trajectory, and a pretrained reinforcement learning policy tracks it to motor commands across quadrotors. Across 15 onboard flights over three everyday referents in a controlled indoor volume, the stack reaches the target in 13 of 15 trials with 5.72 cm mean goal error and 39.3% average GPU utilization. In 6 additional cluttered-environment trials, the stack tracks collision-free trajectories under onboard perception gating.

Chinese Translation

在空中机器人上完全实现视觉-语言导航是困难的，因为基础、规划和控制必须共享有限的计算资源，并且在飞行中很难隔离单阶段错误。端到端的空中策略将这些阶段融合为一个网络，放弃了模块化系统所能提供的可观察性和安全检查。我们提出了飞行中的视觉-语言导航（VLN on the Fly），这是一个机载系统，保持基础、规划和控制作为独立的、可检查的阶段。量化的视觉-语言模型（VLM）将指令定位到粗略的图像单元，深度提升将其提升到三维目标，快速的B样条规划器返回可行轨迹，预训练的强化学习策略将其跟踪到四旋翼的电机指令。在对三个日常参照物进行的15次机载飞行中，该系统在15次试验中成功到达目标13次，平均目标误差为5.72厘米，平均GPU利用率为39.3%。在6次额外的杂乱环境试验中，该系统在机载感知门控下跟踪无碰撞轨迹。

View on arXiv Download PDF AI Translation

cs.RO / 74 / 2609.20227

Strategic Transformer for Resource-Constrained Multi-Object Navigation in Ultra-Large-Scale Environments

超大规模环境中资源受限的多目标导航的战略变换器

Iwata, Daiki, Tanaka, Kanji, Hishida, Senta

Abstract

Resource-constrained multi-object navigation in vast indoor environments ($>2,000\text{ m}^2$) poses significant challenges for efficiency and strategic planning. To tackle this, we reformulate the task as a Set Orienteering Problem (SOP), providing an optimization framework under resource constraints where exploitation is governed by the SOP model and exploration is managed by a separate heuristic switcher. Conventional baselines suffer from either rigid planning or myopic behaviors. To overcome these limitations and resolve the NP-hard computational challenges of SOP for real-time navigation, we develop the Strategic Transformer. This lightweight architecture functions as a priority planner that internalizes expert combinatorial logic into a predictable $41.03\text{ ms}$ forward pass while reducing teacher-student information asymmetry. Incorporating geometric attention biases allows the network to effectively model long-range structural dependencies. By coupling the Transformer's macro-plan with a bounded iterative 2-opt local refinement on a capped candidate graph, our framework achieves a $94\times$ speedup compared to heavy meta-heuristics, ensuring bounded-latency inference suitable for onboard deployment. Experiments on ProcTHOR validate that our method successfully bridges the gap between exploration and exploitation, outperforming carefully re-implemented baselines under Progress weighted by Path Length (PPL) and establishing a new benchmark for scalable, resource-constrained navigation.

Chinese Translation

在广阔的室内环境中（$>2,000 ext{ m}^2$）进行资源受限的多目标导航面临着效率和战略规划的重大挑战。为了解决这一问题，我们将任务重新表述为集合定向问题（Set Orienteering Problem, SOP），提供了一个在资源约束下的优化框架，其中开发利用由SOP模型控制，而探索则由一个独立的启发式切换器管理。传统基线要么受到僵化规划的限制，要么表现出短视行为。为了克服这些局限性并解决SOP在实时导航中的NP难度计算挑战，我们开发了战略变换器（Strategic Transformer）。该轻量级架构作为优先规划器，能够将专家组合逻辑内化为可预测的$41.03 ext{ ms}$前向传递，同时减少教师-学生信息不对称。结合几何注意力偏差使得网络能够有效建模长距离结构依赖关系。通过将变换器的宏观规划与在限制候选图上的有界迭代2-opt局部优化相结合，我们的框架相比于重型元启发式算法实现了$94 imes$的加速，确保了适合机载部署的有界延迟推理。在ProcTHOR上的实验验证了我们的方法成功弥合了探索与开发之间的差距，在路径长度加权进度（Progress weighted by Path Length, PPL）下超越了经过精心重新实现的基线，并为可扩展的资源受限导航建立了新的基准。

View on arXiv Download PDF AI Translation

cs.RO / 75 / 2609.20318

LLM-Guided Transformation of Non-Critical Driving Scenes into Safety-Critical Scenarios Using Augmented Reality

利用增强现实将非关键驾驶场景转化为安全关键场景的LLM引导转化

Fady, Noura, Khaled, Farah, Elias, Catherine M.

Abstract

Testing Autonomous Driving Systems (ADS) requires realistic safety-critical scenarios, but collecting such data from real-world driving is costly and unsafe. This paper presents an automated pipeline that transforms safe driving scenes into safety-critical scenarios by combining computer vision, Large Language Models (LLMs), and Augmented Reality (AR). The system detects and tracks road users, extracts safety features including distance, velocity, motion direction, and Time-to-Collision (TTC), and assesses scene criticality. Safe scenes are modified by an LLM, which generates realistic collision-inducing objects and behaviors that are integrated into the original scene using AR. The proposed pipeline was evaluated on the nuScenes dataset, achieving 97.52% safety classification accuracy and successfully generating realistic scenarios such as pedestrian crossings, rear overtaking vehicles, and sudden-stop events. The results demonstrate an effective and flexible approach for automated generation of safety-critical scenarios to support the testing and validation of autonomous driving systems.

Chinese Translation

测试自动驾驶系统（ADS）需要现实的安全关键场景，但从真实驾驶中收集此类数据既昂贵又不安全。本文提出了一种自动化管道，通过结合计算机视觉、大型语言模型（LLMs）和增强现实（AR），将安全驾驶场景转化为安全关键场景。该系统检测和跟踪道路使用者，提取包括距离、速度、运动方向和碰撞时间（TTC）在内的安全特征，并评估场景的关键性。安全场景由LLM进行修改，生成逼真的引发碰撞的物体和行为，并通过AR集成到原始场景中。所提议的管道在nuScenes数据集上进行了评估，达到了97.52%的安全分类准确率，并成功生成了行人过马路、后方超车车辆和突然停车事件等现实场景。结果表明，这是一种有效且灵活的自动生成安全关键场景的方法，以支持自动驾驶系统的测试和验证。

View on arXiv Download PDF AI Translation

cs.RO / 76 / 2609.20321

Implementation of Tightly-Coupled SLAM Fusion of GPS, IMU, and LiDAR for Autonomous Vehicles

紧耦合SLAM融合GPS、IMU和LiDAR在自主车辆中的实现

Elmehrath, Amr O., Khaled, Farah, Nahas, Rana, Elias, Catherine M.

Abstract

Autonomous vehicles depend entirely on Simultaneous Localization and Mapping (SLAM) to navigate safely in unknown environments. However, relying on a single sensory modality introduces critical failure points: LiDAR systems degrade in featureless corridors, Inertial Measurement Units (IMUs) accumulate mathematical drift, and GPS drops frequently in urban canyons. This paper presents the implementation of a tightly-coupled SLAM fusion architecture that integrates a Velodyne 3D LiDAR, a high-frequency IMU, and GPS to achieve continuous spatial awareness. Utilizing a phased development methodology, we establish a 2D baseline to validate hardware synchronization and transform geometries before upgrading to a full 3D architecture driven by FAST-LIO2. This advanced approach uses an Iterated Error-State Kalman Filter (IESKF) to process dense 3D laser points alongside continuous inertial data, eliminating motion blur at high speeds. To eradicate long-term drift, a GTSAM pose-graph optimization back-end executes multi-modal loop closures. Evaluated across simulated environments and physical deployments, the results demonstrate that tightly-coupled 3D fusion effectively overcomes individual sensor blind spots to generate highly detailed point clouds, providing the foundational High-Definition (HD) maps required for advanced downstream autonomous planners.

Chinese Translation

自主车辆完全依赖于同时定位与地图构建（SLAM）在未知环境中安全导航。然而，依赖单一传感器模态会引入关键的故障点：LiDAR系统在无特征的走廊中性能下降，惯性测量单元（IMU）会累积数学漂移，而GPS在城市峡谷中频繁失效。本文提出了一种紧耦合SLAM融合架构的实现，该架构集成了Velodyne 3D LiDAR、高频IMU和GPS，以实现连续的空间感知。通过分阶段开发的方法，我们建立了一个二维基线，以验证硬件同步和变换几何形状，然后升级到由FAST-LIO2驱动的完整三维架构。这种先进的方法使用迭代误差状态卡尔曼滤波器（IESKF）处理密集的3D激光点和连续的惯性数据，从而消除高速下的运动模糊。为了消除长期漂移，GTSAM姿态图优化后端执行多模态回环闭合。在模拟环境和实际部署中进行评估，结果表明，紧耦合的3D融合有效克服了单个传感器的盲点，生成高度详细的点云，为先进的下游自主规划器提供所需的高清（HD）地图基础。

View on arXiv Download PDF AI Translation

cs.RO / 77 / 2609.20330

RoboFind: Multi-Agent Personalized Object Search for People Who Are Blind or Have Low Vision

RoboFind：为盲人或低视力人士提供的多智能体个性化物体搜索

Liu, Ruiping, Quan, Shaofang, Yin, Qian, Zhang, Jingqi, Zheng, Junwei, Chen, Yufan, Wen, Di, Fan, Weijia, Yang, Kailun, Sarfraz, M. Saquib, Asfour, Tamim, Peng, Kunyu, Stiefelhagen, Rainer

Abstract

Blind and low-vision users often need to locate a specific personal object rather than an arbitrary instance of the same category. The task calls for a robot that can move through the space and reach viewpoints the user cannot, and for an accessible interface where the user says which object is meant and learns whether the right one was found. We present RoboFind, a multi-agent framework in which a smartphone teaches the target and a quadruped robot carries out the search. A Target Teaching Agent converts guided smartphone recordings into a semantic target profile and a reusable multi-view reference bank through an accessible capture flow with AR guidance, speech and haptic feedback, and screen-reader support, so later missions refer to a stored object without repeating the teaching process. At runtime, a Navigation Agent explores the environment and proposes candidate targets, a Verification Agent checks each candidate against the stored references, and a Coordination and Recovery Agent completes the mission or triggers recovery and continued search. Across 32 real-robot missions, RoboFind reaches 85.0% success against 25.0% for a reconstructed sequential first-stop baseline over 20 trials with ten targets, and reduces false success from 75.0% to 5.0%. On six shared targets it succeeds in 10/12 trials, against 5/12 for 12 independently executed GPT-6 Astra-only trials. These results show that the multi-agent design fits the demands of personalized object search, where verifying object identity before declaring completion is what makes the outcome something a user can rely on.

Chinese Translation

盲人和低视力用户通常需要定位特定的个人物品，而不是同一类别的任意实例。该任务需要一个能够在空间中移动并到达用户无法到达的视角的机器人，以及一个可访问的界面，让用户能够说明所指的物体，并了解是否找到了正确的物品。我们提出了RoboFind，一个多智能体框架，其中智能手机负责教授目标，而四足机器人执行搜索。目标教学代理（Target Teaching Agent）通过可访问的捕捉流程，将引导的智能手机录音转换为语义目标档案和可重用的多视图参考库，结合增强现实（AR）指导、语音和触觉反馈以及屏幕阅读器支持，以便后续任务可以引用存储的物体，而无需重复教学过程。在运行时，导航代理（Navigation Agent）探索环境并提出候选目标，验证代理（Verification Agent）检查每个候选目标与存储参考的匹配情况，而协调与恢复代理（Coordination and Recovery Agent）则完成任务或触发恢复和继续搜索。在32次真实机器人任务中，RoboFind的成功率达到85.0%，而重建的顺序首停基线在20次试验中仅为25.0%，并将错误成功率从75.0%降低到5.0%。在六个共享目标上，它在12次试验中成功10次，而独立执行的12次GPT-6 Astra试验中仅成功5次。这些结果表明，多智能体设计符合个性化物体搜索的需求，在宣布完成之前验证物体身份是使用户能够依赖结果的关键。

View on arXiv Download PDF AI Translation

cs.RO / 78 / 2609.20388

Navi-Agent: Unlocalized Monocular Navigation Agent

Navi-Agent：无定位单目导航代理

Xie, Wenyuan, Hong, Mengyang, Wang, Yongzhong, Ji, Yanbiao, Zhou, Yijin, Wu, Shaokai, Sirejiding, Shalayiding, Zhou, Huayi, Chen, Yi-Chao, Ling, Ma, Ding, Yue, Lu, Hongtao

Abstract

Vision-Language Navigation in Continuous Environments (VLN-CE) requires an embodied agent to execute long-horizon instructions in unknown environments. Existing zero-shot VLN-CE systems typically maintain spatial states through geometric localization or coordinate-based representations. Recent geometry-constrained navigation removes depth and globally consistent coordinates, but maintaining persistent spatial awareness for place confirmation, progress verification, and recovery remains challenging. We present Navi-Agent, a zero-shot VLN-CE agent that constructs a coordinate-free spatial state from visual observations and executed motion histories. Navi-Agent organizes this state as a navigation topology, where nodes represent visual places and edges represent motion transitions. This representation enables observation-based approximate self-localization, task progress verification, and visual revisitation-based recovery. Navi-Agent performs closed-loop navigation by decomposing instructions into sub-goals, executing local visual navigation, and verifying visited places through the constructed spatial state. Experiments on zero-shot VLN-CE benchmark and real-world robot platforms show that Navi-Agent achieves state-of-the-art performance among geometry-constrained methods while remaining competitive with approaches relying on geometric localization.

Chinese Translation

在连续环境中的视觉-语言导航（VLN-CE）要求一个具身代理在未知环境中执行长时间跨度的指令。现有的零样本 VLN-CE 系统通常通过几何定位或基于坐标的表示来维持空间状态。近期的几何约束导航去除了深度信息和全局一致坐标，但保持持久的空间意识以确认位置、验证进展和恢复仍然具有挑战性。我们提出了 Navi-Agent，一种零样本 VLN-CE 代理，它从视觉观察和执行的运动历史中构建无坐标的空间状态。Navi-Agent 将该状态组织为导航拓扑，其中节点表示视觉位置，边表示运动转移。这种表示使得基于观察的近似自我定位、任务进展验证和基于视觉重访的恢复成为可能。Navi-Agent 通过将指令分解为子目标、执行局部视觉导航，并通过构建的空间状态验证已访问的位置，进行闭环导航。在零样本 VLN-CE 基准和真实世界机器人平台上的实验表明，Navi-Agent 在几何约束方法中实现了最先进的性能，同时在与依赖几何定位的方法竞争时仍保持竞争力。

View on arXiv Download PDF AI Translation

cs.RO / 79 / 2609.20396

Imagine-TAMP: Imagination-Guided Task and Motion Planning in Partial Observability

Imagine-TAMP：基于想象的部分可观测任务与运动规划

Singha, Antareep, Kumar, Shivaram, Kim, Yoonwoo, Sung, Yoonchang

Abstract

Robots operating in cluttered environments must often manipulate objects whose locations are only partially observable. A central challenge is deciding whether to acquire another observation or to first manipulate objects that may occlude the target. Conventional task and motion planning (TAMP) approaches typically make this decision using symbolic action costs or expensive geometric planning, neither of which adequately captures how likely an observation is to reveal an occluded target. We introduce Imagine-TAMP, an interleaved planning and execution framework that uses semantic and geometric imagination to compare alternative task-level strategies under partial observability before committing to expensive motion planning. A vision-language model shapes a particle belief over target locations using commonsense relationships between the target and visible objects, while a generative scene model estimates plausible geometry in unobserved regions. Given a target hypothesis and imagined scene, Imagine-TAMP generates multiple symbolic plan skeletons and assigns non-unit costs that approximate both manipulation effort and target visibility from sensing actions, distinguishing a short but poorly informative observation strategy from a longer strategy that first manipulates an occluder to better expose the target. The selected skeleton is then refined into a feasible continuous plan and executed, with new observations updating the belief and triggering replanning when necessary. Experiments show that imagination-guided evaluation improves observation-versus-manipulation decisions: in viewpoint-constrained shelf scenes, non-unit geometric evaluation increases success from 46.0% to 84.0%, while semantic belief shaping further reduces manipulation and replanning. On a real robot, the complete system reduces planning time by 32% relative to a geometry-only ablation.

Chinese Translation

在杂乱环境中操作的机器人通常必须操控位置仅部分可观测的物体。一个核心挑战是决定是获取另一个观测还是首先操控可能遮挡目标的物体。传统的任务与运动规划（TAMP）方法通常使用符号动作成本或昂贵的几何规划来做出这一决策，而这两者都无法充分捕捉观测揭示被遮挡目标的可能性。我们提出了Imagine-TAMP，这是一种交错的规划与执行框架，利用语义和几何想象在部分可观测性下比较替代的任务级策略，然后再进行昂贵的运动规划。一个视觉-语言模型利用目标与可见物体之间的常识关系，形成对目标位置的粒子信念，而一个生成场景模型则估计未观测区域的合理几何。给定目标假设和想象场景，Imagine-TAMP生成多个符号计划骨架，并分配非单位成本，以近似操控努力和从感知动作中获得的目标可见性，从而区分短而信息量不足的观测策略与首先操控遮挡物以更好地暴露目标的较长策略。选择的骨架随后被细化为可行的连续计划并执行，新的观测更新信念，并在必要时触发重新规划。实验表明，基于想象的评估改善了观测与操控的决策：在视角受限的货架场景中，非单位几何评估将成功率从46.0%提高至84.0%，而语义信念塑造进一步减少了操控和重新规划。在真实机器人上，完整系统相较于仅使用几何的消融实验减少了32%的规划时间。

View on arXiv Download PDF AI Translation

cs.RO / 80 / 2609.20407

Resilient Motion Planning for Free-Flying Space Robots under Actuator Failures

在执行器故障下的自由飞行空间机器人弹性运动规划

de Maddalena, Nicolas, Verhagen, Joris, Tumova, Jana

Abstract

Free-flying robots rely on multiple thrusters to maneuver in space. If one or more of these thrusters fail, the robot may lose control authority and risk mission failure. At the same time, their free-flying nature implies that, even in the absence of actuation, they continue along (locally) straight-line trajectories. In this work we present a probabilistic, proactive, motion planning framework that explicitly accounts for actuator failures in space. We model actuator failure modes as a Markov chain and propagate the probability of successfully reaching the goal along the planning horizon. Precomputed reachable sets evaluate the robot's capabilities of reaching waypoints under potential failures and an RRT$^*$-based planner concatenates these waypoints. The resulting algorithm maximizes the overall target-reaching probability, providing maximally resilient motion plans utilizing free-flying properties. We validate our approach experimentally on a physical free-flyer platform with injected actuator failures.

Chinese Translation

自由飞行机器人依赖多个推进器在太空中进行机动。如果一个或多个推进器发生故障，机器人可能会失去控制权，面临任务失败的风险。同时，它们的自由飞行特性意味着，即使在没有推进的情况下，它们仍会沿着（局部）直线路径继续运动。在本研究中，我们提出了一种概率性、前瞻性的运动规划框架，明确考虑了太空中的执行器故障。我们将执行器故障模式建模为马尔可夫链，并传播沿规划时域成功到达目标的概率。预计算的可达集合评估机器人在潜在故障下到达航点的能力，而基于RRT$^*$的规划器将这些航点连接起来。所得到的算法最大化了整体目标到达概率，提供了利用自由飞行特性实现的最大弹性运动规划。我们在一个物理自由飞行平台上通过注入执行器故障对我们的方法进行了实验验证。

View on arXiv Download PDF AI Translation

cs.RO / 81 / 2609.20435

Time-Efficient Iterative Learning Planning for Safety-Critical Dynamic Obstacle Avoidance

时间高效的迭代学习规划用于安全关键的动态障碍物避让

Chen, Zhiyi, Lv, Shuli, Min, Chen, Xu, Yong, Sun, Jian, Quan, Quan

Abstract

Autonomous mobile robots require timeefficient planning and safety-critical dynamic obstacle avoidance under constrained onboard computation. While Iterative Learning Planning (ILP) offers lightweight and efficient traversal planning, it lacks explicit mechanisms for dynamic obstacle perception and avoidance. This article extends ILP to safety-critical navigation in dynamic environments by integrating an anticipatory risk-blended control barrier function (ARB-CBF). The extended ILP learns traversal-speed and steering-bias profiles via a fractionalpower update based on local obstacle risk, generating nominal control commands that ARB-CBF modifies at runtime for real-time safety guarantees. Algorithmic analysis demonstrates that the ILP replanning stage scales at O(kN) for k iterations and N waypoints, while ARB-CBF executes with linear complexity. Comprehensive simulations and real-world experiments validate the framework, demonstrating superior temporal efficiency and safety with lower computational overhead compared to optimizationbased baselines, making it highly suitable for resourceconstrained platforms.

Chinese Translation

自主移动机器人在受限的机载计算条件下需要高效的规划和安全关键的动态障碍物避让。尽管迭代学习规划（ILP）提供了轻量级和高效的路径规划，但它缺乏动态障碍物感知和避让的明确机制。本文通过集成预期风险混合控制障碍函数（ARB-CBF），将ILP扩展到动态环境中的安全关键导航。扩展后的ILP通过基于局部障碍物风险的分数次幂更新学习行驶速度和转向偏差轮廓，生成名义控制命令，ARB-CBF在运行时对其进行修改，以提供实时安全保障。算法分析表明，ILP重新规划阶段的复杂度为O(kN)，其中k为迭代次数，N为路径点，而ARB-CBF的执行复杂度为线性。全面的仿真和实际实验验证了该框架，显示出在计算开销较低的情况下，具有优越的时间效率和安全性，因而非常适合资源受限的平台。

View on arXiv Download PDF AI Translation

cs.RO / 82 / 2609.20443

Spatial-Semantic Uncertainty in VLM-Based Target Search: Balancing Exploration and Identification

基于视觉语言模型的目标搜索中的空间-语义不确定性：平衡探索与识别

Srivastava, Alkesh K., Diller, Jonathan, Kumar, Vijay, Dames, Philip

Abstract

Robots searching for a target from a natural-language description must determine not only where to search, but also which observed candidate is the desired target. These decisions reflect two distinct sources of uncertainty - spatial uncertainty over candidate locations and semantic uncertainty over target identity - that are often conflated in VLM-based search systems. We introduce a spatial-semantic uncertainty formulation that maintains separate beliefs over each component and integrates probabilistic VLM evidence into a global target-identity posterior, including probability mass for undiscovered targets. This decomposition allows an information-theoretic planner to independently value candidate discovery and target disambiguation through spatial and semantic expected information gain (EIG), providing an explicit mechanism for trading broader exploration against earlier identification. We evaluate six VLM uncertainty-elicitation interfaces on 500 synthetic targets and show that similar recognition accuracy can conceal substantial differences in calibration and false confidence. In degraded-observation search-and-identify experiments, EIG-based planners reach confident decisions in 75.0%-92.5% of trials, compared with 20.0% for Random search, while different spatial-semantic weightings achieve comparable identification accuracy once confidence is attained. Increasing semantic emphasis reduces unnecessary exploration and VLM queries, demonstrating that explicitly planning over semantic uncertainty can accelerate target resolution without sacrificing decision quality. These results highlight the distinct roles of uncertainty representation and uncertainty-driven planning in embodied VLM systems.

Chinese Translation

机器人根据自然语言描述搜索目标时，不仅需要确定搜索的位置，还需要判断观察到的候选目标中哪个是所需的目标。这些决策反映了两种不同来源的不确定性——候选位置的空间不确定性和目标身份的语义不确定性——在基于视觉语言模型（VLM）的搜索系统中，这两者通常被混淆。我们提出了一种空间-语义不确定性模型，保持对每个组件的独立信念，并将概率性VLM证据整合到一个全局目标身份后验中，包括未发现目标的概率质量。这种分解使得信息论规划者能够通过空间和语义期望信息增益（EIG）独立评估候选发现和目标消歧的价值，为更广泛的探索与更早的识别之间的权衡提供了明确机制。我们在500个合成目标上评估了六种VLM不确定性引导接口，并显示出相似的识别准确率可能掩盖校准和虚假信心的显著差异。在降级观察的搜索与识别实验中，基于EIG的规划者在75.0%-92.5%的试验中达成了自信的决策，而随机搜索仅为20.0%；同时，不同的空间-语义权重在获得信心后实现了相当的识别准确率。增加语义强调减少了不必要的探索和VLM查询，表明明确规划语义不确定性可以加速目标解析而不牺牲决策质量。这些结果突显了不确定性表示和不确定性驱动规划在具身VLM系统中的不同角色。

View on arXiv Download PDF AI Translation

cs.RO / 83 / 2609.20477

Visual Sim-to-Real Learning for Robotic Insertion under Geometric Variations: Application to Rebar Installation

几何变化下机器人插入的视觉模拟到现实学习：应用于钢筋安装

Sun, Tao, Han, Beining, Yin, Patrick, Xu, Rui, He, Harry, Gupta, Abhishek, Rusinkiewicz, Szymon, Shao, Yi

Abstract

Rebar insertion is among the most repetitive and physically demanding tasks on construction sites, and a contact-rich problem at 1.4 mm clearance. The parts, however, vary at two levels: a nominal design per structural member, and fabrication tolerance around each nominal design. Real-world data therefore has to be re-collected as designs and batches change. We present RebarSim, a visual sim-to-real system trained entirely in simulation. A privileged state-based teacher is trained with reinforcement learning over procedurally generated rebar geometries, then distilled into a multi-view student that maps raw RGB and proprioception directly to actions under extensive domain randomization. The student transfers to the real world zero-shot, seating rebars taken from a real factory production run in 91.3% of real-robot rollouts. Underlying that result, geometry diversity and pretraining both bring benefits. Training across a diverse set of nominal designs rather than one lifts the zero-shot success of both the teacher and the student on unseen designs, and the student policy outperforms a single-design specialist on that specialist's own design. A pretrained student then adapts to a new design with 4--6x fewer distillation samples than one trained from scratch. Visual sim-to-real transfer depends on appearance randomization and the DAgger mixture: removing either one sharply lowers success. Videos, code, and task assets are available at https://rebarsim.github.io.

Chinese Translation

钢筋插入是建筑工地上最重复且体力要求高的任务之一，且在1.4毫米间隙下属于接触丰富的问题。然而，部件在两个层面上存在变化：每个结构构件的名义设计，以及围绕每个名义设计的制造公差。因此，随着设计和批次的变化，必须重新收集真实世界的数据。我们提出了RebarSim，一个完全在模拟中训练的视觉模拟到现实系统。一个特权的基于状态的教师通过强化学习在程序生成的钢筋几何形状上进行训练，然后被提炼成一个多视角的学生，该学生将原始RGB和本体感知直接映射到在广泛领域随机化下的动作。该学生在零-shot的情况下转移到现实世界，在91.3%的真实机器人实验中成功安装来自真实工厂生产批次的钢筋。支撑这一结果的，几何多样性和预训练都带来了好处。在多样的名义设计集上进行训练，而不是单一设计，提高了教师和学生在未见设计上的零-shot成功率，并且学生策略在该专家自身设计上超越了单一设计专家。预训练的学生在适应新设计时所需的蒸馏样本比从头训练的学生少4到6倍。视觉模拟到现实的转移依赖于外观随机化和DAgger混合：去掉任一项都会显著降低成功率。视频、代码和任务资产可在 https://rebarsim.github.io 获取。

View on arXiv Download PDF AI Translation

cs.RO / 84 / 2609.20480

Worst-Case Hidden-Vehicle Trajectory Search in Spatiotemporal Occlusion Regions

时空遮挡区域中最坏情况隐蔽车辆轨迹搜索

Tan, Ruichen, Lei, Zengxiang, Ukkusuri, Satish

Abstract

Occlusion creates fundamental uncertainty in autonomous driving. Existing methods often propagate frame-wise hypotheses or optimize ego behavior against prescribed hidden-agent predictions, leaving the worst history-consistent interaction unexplored. We introduce History-Conditioned Minimax Trajectory Search (HC-MTS), which combines temporal occlusion reasoning with response-aware search. First, HC-MTS constructs finite hidden-state modes, each certified by a backward witness satisfying multi-frame visibility, occupancy, semantic-map support, and class-specific kinematic constraints. It then solves a bilevel minimax problem: an inner finite oracle maximizes the ego driving score over destination attainment and ride comfort, while the outer search selects the legal hidden-vehicle trajectory that minimizes this best-response value. Across eight Waymo Open Motion Dataset scenarios, increasing the visibility-memory horizon from K=1 to K=20 reduces the mean per-scenario vehicle, pedestrian, and total retained hidden-seed counts by 18.12%, 21.67%, and 18.45%, respectively. HC-MTS identifies six avoidable counterexamples, while no legal collision-producing attacker is found in the remaining two scenes within the finite search budget.

Chinese Translation

遮挡在自动驾驶中造成了根本的不确定性。现有方法通常传播逐帧假设或针对规定的隐蔽代理预测优化自我行为，未能探讨最坏的历史一致交互。我们提出了历史条件下的极小极大轨迹搜索（History-Conditioned Minimax Trajectory Search, HC-MTS），该方法将时间遮挡推理与响应感知搜索相结合。首先，HC-MTS 构建有限的隐状态模式，每个模式都由满足多帧可见性、占用、语义地图支持和类别特定运动约束的向后见证进行认证。然后，它解决一个双层极小极大问题：内部有限神谕最大化自我驾驶得分，考虑目的地到达和乘坐舒适性，而外部搜索选择最小化此最佳响应值的合法隐蔽车辆轨迹。在八个 Waymo Open Motion Dataset 场景中，将可见性记忆范围从 K=1 增加到 K=20，分别减少了每个场景的平均车辆、行人和总保留隐蔽种子数量的 18.12%、21.67% 和 18.45%。HC-MTS 识别出六个可避免的反例，而在有限搜索预算内，剩余两个场景中未发现合法的产生碰撞的攻击者。

View on arXiv Download PDF AI Translation

cs.RO / 85 / 2609.20499

Towards AI-enhanced control: a numerical technique for trajectory smoothing of a parallel robot for pancreatic surgery

面向增强人工智能控制：用于胰腺手术的并联机器人轨迹平滑的数值技术

Birlescu, Iosif, Pusca, Alexandru, Gherman, Bogdan, Vaida, Calin, Zima, Ionut, Chablat, Damien, Pisla, Doina

Abstract

The paper presents a numerical approach for the end-effector trajectory smoothing of a parallel robot designed for minimally invasive pancreatic surgery. The approach is tailored for real-time master-slave control architecture and uses a 3D space mouse for command input for velocity control. The trajectory smoothing is achieved by generating S-curves in the end-effector velocity fields, thus controlling the accelerations, which in turn reduces tissue trauma in the minimally invasive procedures. Real-time control is enabled by segmenting the S-curves based on the command inputs from the 3D space mouse. A special case is considered where the acceleration time is constant for all command inputs. Numeric results demonstrate stable transitions (without abrupt changes) in both the end-effector parameter space and in the active joints parameters, thereby validating the proposed approach. Further work aims to test the approach on an experimental model and integrate it into AI-based training modules.

Chinese Translation

本文提出了一种针对用于微创胰腺手术的并联机器人末端执行器轨迹平滑的数值方法。该方法专为实时主从控制架构而设计，并使用3D空间鼠标进行速度控制的命令输入。通过在末端执行器速度场中生成S曲线实现轨迹平滑，从而控制加速度，进而减少微创手术中的组织损伤。实时控制通过根据3D空间鼠标的命令输入对S曲线进行分段来实现。考虑了一种特殊情况，即所有命令输入的加速时间保持不变。数值结果表明，末端执行器参数空间和主动关节参数中的过渡稳定（没有突变），从而验证了所提出的方法。后续工作旨在在实验模型上测试该方法，并将其整合到基于人工智能的训练模块中。

View on arXiv Download PDF AI Translation

cs.RO / 86 / 2609.20540

Integrated Guidance and Control of a Mother-Child UAV-UGV System for Cooperative Missions

母子无人机-无人地面车辆系统的集成引导与控制用于协同任务

Sahu, Aashish, Kumar, R. Prasanth

Abstract

Autonomous recovery of a small multirotor onto a hovering multirotor carrier differs from recovery onto ground or shipborne platforms because the recovery surface is itself an actively controlled, thrust-limited aerial vehicle. This paper presents a field-validated autonomy framework for a heterogeneous rover-mothership-child system executing rover supervision, mothership transit, child deployment and sortie, autonomous return, aerial recovery, and synchronized descent. The recovery stack combines jerk-bounded reference generation, disturbance-observer-augmented planar tracking, feasibility-aware vertical control, a discrete-time barrier-based safety filter for relative vertical geometry, and communication-aware carrier-state prediction. The contribution is the coordinated system-level integration of these methods for recovery onto a hovering multirotor and its full-scale outdoor validation. The framework is implemented on a PX4-ROS 2 architecture using RTK-enabled GNSS, IMU, and barometric fusion, with mothership-side 1D lidar used only as an auxiliary near-contact cue. RTK-fixed positioning was maintained throughout testing. Across 20 outdoor cooperative missions, 17 successfully completed deployment, sortie, and recovery, giving an observed mission success rate of 85%. For successful recoveries, mean terminal-alignment time was 6.3 s, mean planar alignment error at acceptance was 0.18 m, maximum terminal planar deviation was 0.32 m within a 0.40 m capture radius, and minimum logged relative vertical separation during coupled descent was 0.41 m. Mothership planar station-keeping RMS error was 0.25 m. The three unsuccessful trials occurred at different mission stages and are analyzed separately. Results demonstrate practical autonomous aerial recovery within the tested outdoor operating envelope.

Chinese Translation

小型多旋翼无人机在悬停的多旋翼载体上自主回收与在地面或船载平台上的回收有所不同，因为回收表面本身是一个受控的、推力受限的空中飞行器。本文提出了一种经过现场验证的自主框架，用于异构探测器-母船-子系统执行探测器监督、母船过渡、子系统部署与出击、自主返回、空中回收和同步下降。回收堆栈结合了限制加速度的参考生成、增强干扰观测的平面跟踪、考虑可行性的垂直控制、基于离散时间的相对垂直几何安全过滤器，以及考虑通信的载体状态预测。这一贡献在于将这些方法进行协调的系统级集成，以实现对悬停多旋翼的回收，并进行了全尺度的户外验证。该框架在PX4-ROS 2架构上实现，使用RTK-enabled GNSS、IMU和气压融合，母船侧的1D激光雷达仅作为辅助近接提示。在测试过程中保持了RTK固定定位。在20次户外协作任务中，17次成功完成了部署、出击和回收，观察到的任务成功率为85%。对于成功的回收，平均终端对准时间为6.3秒，接受时的平均平面对准误差为0.18米，最大终端平面偏差在0.40米捕获半径内为0.32米，耦合下降期间记录的最小相对垂直分离为0.41米。母船平面保持的均方根误差为0.25米。三次不成功的试验发生在不同的任务阶段，并单独进行分析。结果表明，在测试的户外操作范围内实现了实用的自主空中回收。

View on arXiv Download PDF AI Translation

cs.RO / 87 / 2609.20558

Learning Slope-Adaptive Whole-Body Locomotion for Humanoid Robots in Roofing Construction

为人形机器人在屋顶施工中学习坡度自适应的全身运动

Liu, Songyang, Li, Shuai

Abstract

Roofing requires workers to coordinate locomotion, balance, and work-related body motions on pitched surfaces, creating a challenging application for humanoid robots. Directly retargeted human demonstrations, however, may preserve motion appearance while placing the robot's feet or hands incorrectly relative to the roof. This study presents a task-semantic scene-grounded framework for learning roofer-style whole-body motions on a Unitree G1. Human demonstrations are captured using a tracking system and retargeted to the robot, while a metric roof model supplies the spatial reference unavailable from the tracking system. A trajectory-level optimization grounds inferred support contacts and annotated work relations to the roof, and execution-aware reinforcement learning encourages the resulting policy to preserve these relations under dynamic tracking errors. The framework is evaluated through a multi-motion tracking study, a roof-pitch coverage matrix, a five-way nailgun ablation, cross-task experiments on hammering and lateral pushing, and comparisons with pure reinforcement learning and zero-shot teleoperation. Our method enables the robot to satisfy support, work-clearance, and nonpenetration criteria across all evaluated seeds. Across nailgun, hammering, and pushing, it achieves work-clearance errors between 0.256 and 0.531 cm and 3/3 successful evaluations per task. Physical experiments reproduce uphill walking, nailgun, hammering, and bending motions with mean base-frame motion errors below 80 mm. These findings establish scene-grounded human motion learning as a promising basis for construction-oriented humanoid motion primitives.

Chinese Translation

屋顶施工要求工人在倾斜表面上协调运动、平衡和与工作相关的身体动作，这为人形机器人带来了挑战。然而，直接重定向的人类演示可能在保持运动外观的同时，导致机器人在屋顶上脚或手的位置不正确。本研究提出了一种基于任务语义场景的框架，用于在Unitree G1上学习屋顶工风格的全身运动。人类演示通过跟踪系统捕获并重定向到机器人，同时一个度量屋顶模型提供了跟踪系统无法获得的空间参考。轨迹级优化将推断的支撑接触和注释的工作关系与屋顶相结合，而执行感知的强化学习则鼓励生成的策略在动态跟踪误差下保持这些关系。该框架通过多运动跟踪研究、屋顶倾斜覆盖矩阵、五种钉枪消融实验、锤击和侧推的跨任务实验，以及与纯强化学习和零样本遥操作的比较进行评估。我们的方法使机器人能够在所有评估种子中满足支撑、工作间隙和非穿透标准。在钉枪、锤击和推送任务中，它的工作间隙误差在0.256到0.531厘米之间，每个任务成功评估3/3次。物理实验重现了上坡行走、钉枪、锤击和弯曲动作，平均基准框架运动误差低于80毫米。这些发现确立了基于场景的人类运动学习作为建筑导向的人形运动原语的有希望的基础。

View on arXiv Download PDF AI Translation

cs.RO / 88 / 2609.20566

OmniMimic: Dynamics-completed Motion Augmentation for Multi-style Omnidirectional Quadruped Locomotion

OmniMimic：多风格全向四足运动的动态补全运动增强

Wu, Sheng, Zhao, Guoqiang, Yang, Zhe, Teng, Fei, Zhou, Zhikun, Yang, Yanlin, Fang, Zheng, Zheng, Hong, Wang, Yaonan, Yang, Kailun

Abstract

Animal demonstrations provide quadruped robots with natural and distinctive gait styles that are difficult to specify through hand-crafted rewards. However, their narrow directional coverage leaves little style-consistent supervision for backward, lateral, and turning commands. We present OmniMimic, a training framework that turns directionally limited animal demonstrations into a single multi-gait policy over target per-axis velocity ranges. OmniMimic first combines temporal reversal, constrained dynamics completion, and sagittal reflection to construct robot-specific kinematic and physical supervision beyond the observed directions. It then expands commands progressively from the demonstrated velocity distribution toward the target per-axis bounds, and uses a shared actor with soft-gated, gait-specialized residual experts to balance reusable locomotion skills with gait-specific corrections. Across four gaits in simulation, OmniMimic reduces mean foot-position RMSE at forward and backward reference velocities by 12.9% and velocity-tracking RMSE on a uniform Cartesian command grid by 63.1%, compared with the matched APEX baseline. The project page is at https://OmniMimic.github.io.

Chinese Translation

动物示范为四足机器人提供了自然且独特的步态风格，这些风格通过手工设计的奖励难以指定。然而，它们狭窄的方向覆盖范围使得在向后、侧向和转向指令上几乎没有风格一致的监督。我们提出了OmniMimic，这是一种训练框架，将方向有限的动物示范转化为针对目标每轴速度范围的单一多步态策略。OmniMimic首先结合时间反转、受限动态补全和矢状面反射，构建超越观察方向的机器人特定运动学和物理监督。然后，它逐步扩展命令，从示范的速度分布向目标每轴边界推进，并使用共享的演员与软门控、步态专用的残差专家，以平衡可重用的运动技能与步态特定的修正。在仿真中，OmniMimic在四种步态下将前后参考速度的平均足部位置均方根误差（RMSE）降低了12.9%，在均匀笛卡尔命令网格上的速度跟踪RMSE降低了63.1%，与匹配的APEX基线相比。项目页面网址为 https://OmniMimic.github.io。

View on arXiv Download PDF AI Translation

cs.RO / 89 / 2609.20570

Walking on the Slope: Stable Bipedal Gaits with Genetic-Algorithm-Optimized Trajectories

在斜坡上行走：基于遗传算法优化轨迹的稳定双足步态

Rijal, Madhav

Abstract

This paper presents the kinematic and dynamic modeling, trajectory generation, and stability analysis of an 8-degree-of-freedom (DOF) biped robot walking on flat and inclined terrain. Denavit-Hartenberg (DH) parameters and homogeneous transformations are used to derive the forward kinematics, while closed-form inverse kinematics maps the desired hip and swing-foot Cartesian trajectories, generated with cubic splines, to joint angles. Joint torques are computed using the Newton-Euler iterative algorithm, and dynamic stability is evaluated using the zero moment point (ZMP) criterion. A genetic algorithm (GA) optimizes the hip height, maximum swing-foot lift, and frontal-plane tilt angle by minimizing the work done by the joints subject to a ZMP feasibility penalty. Simulation results in MATLAB show that the nominal 8-DOF model remains ZMP-stable for step completion times down to 0.5 s and for slope inclinations up to 22.5 degrees with the given foot geometry. Beyond these limits, the ZMP leaves the support polygon, and either the foot dimensions or the trajectory parameters must be modified. The results also show that ZMP stability is governed by the mass distribution among the links rather than the total mass of the robot.

Chinese Translation

本文提出了一个8自由度（DOF）双足机器人在平坦和倾斜地形上行走的运动学和动力学建模、轨迹生成及稳定性分析。采用Denavit-Hartenberg（DH）参数和齐次变换推导前向运动学，而闭式逆运动学则将使用三次样条生成的期望髋部和摆动脚的笛卡尔轨迹映射到关节角度。使用牛顿-欧拉迭代算法计算关节扭矩，并通过零力矩点（ZMP）标准评估动态稳定性。遗传算法（GA）通过最小化关节所做的功，并考虑ZMP可行性惩罚，优化髋部高度、最大摆动脚抬起高度和前平面倾斜角度。MATLAB中的仿真结果表明，名义上的8-DOF模型在步态完成时间降低至0.5秒和坡度倾斜角度达到22.5度的情况下，仍然保持ZMP稳定。在这些限制之外，ZMP会离开支撑多边形，必须修改脚部尺寸或轨迹参数。结果还表明，ZMP稳定性受链接之间质量分布的影响，而非机器人总质量。

View on arXiv Download PDF AI Translation

cs.RO / 90 / 2609.20575

Accelerating Visual Policy Learning with Sampling-Based Model Predictive Control

基于采样的模型预测控制加速视觉策略学习

Liu, Yilang, You, Haoxiang, Wang, Qian, Rakita, Daniel, Abraham, Ian

Abstract

Learning visual policies for locomotion and manipulation requires coordinating contact with the environment and can incur substantial computation and GPU memory costs. First-order policy gradients (FoPG) reduce training cost through differentiable simulation, but local optimization can converge to unintended contact patterns. To address this shortfall, we propose Sampling-Guided Policy Search (SGPS), which couples recurring action-target refinement by sampling-based model-predictive control with first-order policy optimization. Behavior cloning initializes the policy from sampled actions; training then alternates sampling-based refinement with short-horizon FoPG updates under perturbed initial states and randomized dynamics. For visual policy training, we use a decoupled FoPG formulation that excludes rendering from the computation graph, enabling direct learning from depth observations without a state-policy teacher. On a single GPU, SGPS learns policies for locomotion, obstacle traversal, crate pushing, and bimanual carrying on simulated Unitree Go2 and G1 robots. Our experiments further show that refinement improves policy learning beyond initialization and tracking alone. For hardware deployment, the distilled policy transfers zero-shot to a real Go2 and uses onboard depth to autonomously trot, crawl, clear hurdles, and switch between these behaviors.

Chinese Translation

学习用于运动和操作的视觉策略需要与环境的接触协调，这可能会产生大量的计算和GPU内存成本。一阶策略梯度（FoPG）通过可微分仿真降低训练成本，但局部优化可能会收敛到意外的接触模式。为了解决这一不足，我们提出了采样引导的策略搜索（SGPS），它将基于采样的模型预测控制与一阶策略优化相结合，进行重复的动作目标细化。行为克隆通过采样的动作初始化策略；训练过程交替进行基于采样的细化和在扰动初始状态及随机动态下的短期FoPG更新。对于视觉策略训练，我们使用了一种解耦的FoPG形式，排除了渲染在计算图中的作用，使得可以直接从深度观测中学习，而无需状态-策略教师。在单个GPU上，SGPS为模拟的Unitree Go2和G1机器人学习了运动、障碍物穿越、箱子推动和双手携带的策略。我们的实验进一步表明，细化在初始化和单独跟踪之外改善了策略学习。对于硬件部署，提炼的策略能够零样本迁移到真实的Go2，并利用机载深度传感器自主地进行小跑、爬行、跨越障碍和在这些行为之间切换。

View on arXiv Download PDF AI Translation

cs.RO / 91 / 2609.20582

V2-STRep: VLM-Grounded Structured Task Representations for Reusable Robot Skills Acquired from Generated Videos

V2-STRep：基于VLM的结构化任务表示用于从生成视频中获取可重用机器人技能

Hu, Yexin, Lee, Dongheui

Abstract

Human manipulation videos provide rich motion and interaction cues for acquiring robot skills without robot demonstrations. Video generation models synthesize such demonstrations from an initial scene image and task instruction, avoiding the need to record demonstrations for each task. However, the recovered motion captures only one scene-specific realization, leaving task structure, geometric relations, and constraints implicit. We present V2-STRep, a zero-shot framework that converts generated video motion into reusable robot skills through VLM-grounded structured task representations. The representation specifies motion phases, references, and task-relevant constraints, with targets described by minimal geometric structures: points, point-normals, axes, planes, and full 6D poses. VLM-provided 2D image-space cues are lifted into 3D using RGB-D observations to reconstruct task geometry and candidate grasp poses. Geometry-specific rules transfer motion to new scenes, while task-constrained trajectory optimization couples grasp selection with complete robot motion planning. It preserves task requirements while using remaining rotational freedom to accommodate joint limits. Updating deployment grounding and constraints enables reuse under new compatible instructions without generating another video. Experiments on six real-world manipulation tasks demonstrate improved execution success over baselines, reliable cross-scene transfer of successfully acquired skills, and adaptation to changed deployment instructions.

Chinese Translation

人类操控视频提供了丰富的运动和交互线索，使得在没有机器人演示的情况下获取机器人技能成为可能。视频生成模型通过初始场景图像和任务指令合成这些演示，避免了为每个任务录制演示的需要。然而，恢复的运动仅捕捉到一个特定场景的实现，任务结构、几何关系和约束仍然隐含。我们提出了V2-STRep，这是一种零-shot框架，通过基于VLM的结构化任务表示将生成的视频运动转化为可重用的机器人技能。该表示指定了运动阶段、参考和任务相关的约束，目标通过最小几何结构描述：点、点法线、轴、平面和完整的6D姿态。VLM提供的2D图像空间线索通过RGB-D观测提升到3D，以重建任务几何和候选抓取姿态。特定几何规则将运动转移到新场景，而受任务约束的轨迹优化将抓取选择与完整的机器人运动规划相结合。它在保留任务要求的同时，利用剩余的旋转自由度来适应关节限制。更新部署的基础和约束使得在新的兼容指令下重用成为可能，而无需生成另一个视频。在六个真实世界的操控任务上的实验表明，与基线相比，执行成功率有所提高，成功获取的技能在跨场景传递中表现可靠，并能够适应变化的部署指令。

View on arXiv Download PDF AI Translation

cs.RO / 92 / 2609.20586

CoRef-GS: Cooperative Referring Gaussian Splatting for Multi-Agent Scene Understanding

CoRef-GS：用于多智能体场景理解的协作引用高斯喷溅

Zhou, Zhikun, Peng, Kunyu, Yang, Runyi, Cai, Junhao, Wen, Di, Liu, Ruiping, Paudel, Danda Pani, Zhou, Yi, Van Gool, Luc, Yang, Kailun

Abstract

Referring scene understanding for embodied robots requires grounding object- and relation-centric language queries from a designated viewpoint. While a local semantic Gaussian map can support such grounding within one agent's observations, cooperative settings require this ability to remain effective after independently reconstructed maps are aligned and fused. In this setting, the referred target or its contextual landmark may come from another agent's observations, while spatial relations must still be interpreted from the querying robot's viewpoint. We formulate this problem as cooperative referring Gaussian grounding over fused maps, which requires geometric alignability, instance-level semantic comparability, and view-conditioned relation reasoning. Existing language-aware Gaussian methods mainly focus on single-map querying, whereas Gaussian registration methods optimize geometric or photometric alignment without preserving language-grounding-oriented semantic compatibility. We propose CoRef-GS, a cooperative referring Gaussian splatting framework. CoRef-GS constructs local open-vocabulary instance-aware Gaussian maps, then aligns partially overlapping maps with a cross-agent alignment module by geometric and semantic consistency, and grounds queries using a view-conditioned mask relation graph. We further introduce CoQuad-Ref, a dual-quadruped benchmark spanning both real-world and simulated indoor scenes. Experiments show that, on simulated scenes, CoRef-GS reduces the rotation error from 2.58{\deg} after coarse initialization to 0.15{\deg} after refinement, and improves real-world referring mIoU over ReferSplat from 52.6% to 68.8%. The established benchmark and source code will be publicly released at https://github.com/ruojiruoli17/CoRef-GS.git.

Chinese Translation

对于具身机器人而言，引用场景理解需要从指定视角对以物体和关系为中心的语言查询进行基础定位。虽然局部语义高斯图可以支持在单个智能体观察范围内进行这样的基础定位，但在协作环境中，这种能力在独立重建的地图对齐和融合后仍需保持有效。在这种情况下，被引用的目标或其上下文地标可能来自另一个智能体的观察，而空间关系仍需从查询机器人的视角进行解释。我们将此问题表述为在融合地图上进行协作引用高斯基础定位，这需要几何对齐性、实例级语义可比性和视图条件下的关系推理。现有的语言感知高斯方法主要集中于单地图查询，而高斯注册方法则优化几何或光度对齐，但未能保持面向语言基础定位的语义兼容性。我们提出了CoRef-GS，一个协作引用高斯喷溅框架。CoRef-GS构建局部开放词汇实例感知高斯图，然后通过几何和语义一致性与跨智能体对齐模块对部分重叠的地图进行对齐，并使用视图条件下的掩码关系图对查询进行基础定位。我们进一步引入CoQuad-Ref，一个涵盖真实世界和模拟室内场景的双四足基准。实验表明，在模拟场景中，CoRef-GS将粗略初始化后的旋转误差从2.58°降低到精细化后的0.15°，并将真实场景中的引用mIoU从52.6%提升至68.8%。所建立的基准和源代码将公开发布在https://github.com/ruojiruoli17/CoRef-GS.git。

View on arXiv Download PDF AI Translation

cs.RO / 93 / 2609.20604

Semantic SLAM in Precision Agriculture using Bayesian Inference

基于贝叶斯推断的精准农业语义SLAM

Beumer, Ruben, Doodeman, Sander, van de Molengraft, René, Antunes, Duarte

Abstract

This paper presents a real-time semantic world modeling framework specialized for precision agriculture using autonomous robots. The framework combines probabilistic mapping of objects and their semantic attributes, updated through Bayesian inference, with a graph-based Simultaneous Localization and Mapping (SLAM) approach implemented using $g^2o$, a general framework for graph optimization. This integration enables accurate mapping and localization without relying solely on GPS. By leveraging semantic information such as plant type, size, and health, the robot can perform tasks while mapping and localizing itself within a field of crops. The proposed framework was validated through Gazebo simulations and physical experiments on an indoor field with artificial plants using Boston Dynamics' robot dog Spot. A YOLOv8n object detection model was trained to extract object and semantic data from depth camera observations. These simulations and experiments demonstrate that the system can successfully perform real-time mapping of up to at least 400 plants.

Chinese Translation

本文提出了一种专门针对精准农业的实时语义世界建模框架，适用于自主机器人。该框架结合了对象及其语义属性的概率映射，通过贝叶斯推断进行更新，并采用基于图的同时定位与地图构建（SLAM）方法，使用 $g^2o$ 这一通用图优化框架实现。这种集成使得在不完全依赖GPS的情况下，能够实现准确的映射和定位。通过利用植物类型、大小和健康状况等语义信息，机器人能够在作物田间执行任务的同时进行自我映射和定位。所提出的框架通过Gazebo仿真和在室内人工植物田的物理实验进行了验证，实验中使用了波士顿动力公司的机器人犬Spot。训练了YOLOv8n对象检测模型，以从深度相机观测中提取对象和语义数据。这些仿真和实验表明，该系统能够成功实现至少400株植物的实时映射。

View on arXiv Download PDF AI Translation

cs.RO / 94 / 2609.20605

Bayesian Continuum Robot Dynamics and State Estimation

贝叶斯连续体机器人动力学与状态估计

Ferguson, James M., Hermans, Tucker, Kuntz, Alan

Abstract

Recent factor graph approaches to continuum robot state estimation have been successful for quasi-static applications and spatiotemporal estimation using white-noise kinematic motion priors. However, when inertial effects are significant, these approximations may fail to capture the underlying physics, limiting accuracy during dynamic motions. In contrast, our approach approximates the Cosserat rod dynamics of continuum robots. We write inertia and damping as equivalent applied loads, so that the dynamic balance retains the algebraic form of the static one from prior work with quasi-static robots. Without backbone observations, the framework reduces to a stochastic forward simulation of the robot's motion. Given observations, it jointly refines kinematic and dynamic states and infers external loads, among other states. We validate the approach through simulation and experiments, demonstrating stochastic forward simulation as well as state estimation on tendon-driven continuum robots.

Chinese Translation

最近的因子图方法在连续体机器人状态估计中已成功应用于准静态场景和利用白噪声运动先验进行时空估计。然而，当惯性效应显著时，这些近似可能无法捕捉到潜在的物理特性，从而限制了动态运动中的准确性。相比之下，我们的方法近似连续体机器人的科塞拉特杆动力学。我们将惯性和阻尼写作等效的施加载荷，使得动态平衡保持与先前针对准静态机器人的工作中静态形式相同。在没有背骨观测的情况下，该框架简化为机器人运动的随机前向仿真。在给定观测的情况下，它联合优化运动学和动态状态，并推断外部载荷等其他状态。我们通过仿真和实验验证了该方法，展示了随机前向仿真以及在腱驱动连续体机器人上的状态估计。

View on arXiv Download PDF AI Translation

cs.RO / 95 / 2609.20615

INSPECT: Learning Robot View Selection from Assistant Use

INSPECT：从助手使用中学习机器人视角选择

Wen, Di, Yang, Kailun, Guo, Wenhao, Shi, Yitian, Zheng, Junwei, Chen, Yufan, Liu, Ruiping, Wei, Jiale, Rayyes, Rania, Peng, Kunyu

Abstract

Robots inspecting an assembly must determine which parts are present and whether they are correctly installed. During egocentric assembly assistance, head motion and workpiece handling reveal evidence for these checks, while spoken state confirmations link observations to procedural outcomes. We introduce INSPECT, which learns robot view preferences from records of a smart-glasses assistant that answers part queries and provides next-step guidance. Presence-Invariant TwinSwap (PI-TwinSwap) calibrates object evidence through paired identity interventions. Claim-indexed supervision separates evidence requirements from camera-reproducible observation changes. Object-centered calibration adapts relative view preferences to robot poses, while clause-level screening checks predicted evidence. The robot selects views using only its current observation and known poses, without candidate images. Evaluation uses annotated assistant-video replay to simulate state feedback, without target-domain view labels for policy training. On images of physical gearbox assemblies, INSPECT achieves the highest view utility among the compared non-oracle policies and raises human-rated full verifiability from 34.8% to 41.7% compared with keeping the current view. On commercial angle-grinder recordings in IMPACT, the transferred relative-view selector increases the correct decision rate from 50.6% to 54.3% with a frozen perception head. The source code is available at https://github.com/Kratos-Wen/INSPECT.

Chinese Translation

机器人在检查装配时必须确定哪些部件存在以及它们是否正确安装。在自我中心的装配辅助过程中，头部运动和工件处理揭示了这些检查的证据，而口头状态确认则将观察结果与程序结果联系起来。我们提出了INSPECT，该系统从智能眼镜助手的记录中学习机器人视角偏好，该助手回答部件查询并提供下一步指导。存在不变的双重交换（Presence-Invariant TwinSwap, PI-TwinSwap）通过配对身份干预来校准物体证据。声明索引监督将证据需求与相机可重现的观察变化分开。以物体为中心的校准将相对视角偏好调整为机器人姿态，而条款级筛选检查预测证据。机器人仅使用当前观察和已知姿态选择视角，而不使用候选图像。评估使用带注释的助手视频回放来模拟状态反馈，而不需要目标领域视角标签进行策略训练。在物理齿轮箱装配的图像中，INSPECT在比较的非神谕策略中实现了最高的视角效用，并将人类评分的完全可验证性从34.8%提高到41.7%，相比于保持当前视角。在IMPACT的商业角磨机录音中，转移的相对视角选择器在冻结的感知头下将正确决策率从50.6%提高到54.3%。源代码可在 https://github.com/Kratos-Wen/INSPECT 获取。

View on arXiv Download PDF AI Translation

cs.RO / 96 / 2609.20620

A Simulation Platform for AUV Fault Recovery: Exploring LLM-Based Diagnostic Strategies

一种用于AUV故障恢复的仿真平台：探索基于大型语言模型的诊断策略

Halba, Khalid, Cooper, Kylie, Bellingham, James G.

Abstract

Autonomous underwater vehicles (AUVs) operating beyond reliable communications must recover from failures without human intervention. We investigate an architecture in which conventional deterministic layered control autonomy manages normal operations, while an invokable large language model (LLM) serves as a diagnostic and recovery planner when onboard anomaly detection identifies performance outside expected limits. Because language models are stochastic, rigorous evaluation requires ensemble testing rather than individual demonstrations. We present a closed-loop simulation architecture that couples real-time C vehicle software with a higher-level orchestration layer for physics-based fault injection, structured prompting, language-model interaction, mission file generation, validation, execution, and LLM-judge scoring. The framework, which we call SPAR (Simulation Platform for AUV Recovery), supports evaluation across fault realizations, prompt structures, reasoning models, and mission conditions. We vary these for a mass-shift fault over 480 SPAR trials, evaluating a frontier model and three off-the-shelf locally deployable LLMs. Model choice dominates diagnosis: the frontier model places the CG-shift mechanism in its top three hypotheses in 85-90% of trials, versus 60-78% for the best local model. Reasoning analysis indicates that local-model success is associated with following the complete diagnostic procedure, whereas weaker models often commit prematurely to elevator failure even though the actuator tracks its command. Diagnosis and operational decision performance do not appear to be coupled in this dataset. The contributions are an architecture extending unanticipated-fault recovery from detection to mitigation and an ensemble methodology for evaluating LLM-assisted mission management on low-power AUVs.

Chinese Translation

在可靠通信之外操作的自主水下航行器（AUV）必须在没有人工干预的情况下从故障中恢复。我们研究了一种架构，其中传统的确定性分层控制自主性管理正常操作，而可调用的大型语言模型（LLM）在机载异常检测识别出性能超出预期限制时，作为诊断和恢复规划工具。由于语言模型是随机的，严格的评估需要集成测试而非单独演示。我们提出了一种闭环仿真架构，将实时C语言车辆软件与更高层次的编排层相结合，用于基于物理的故障注入、结构化提示、语言模型交互、任务文件生成、验证、执行和LLM评估。我们称之为SPAR（AUV恢复仿真平台）的框架，支持跨故障实现、提示结构、推理模型和任务条件的评估。我们在480次SPAR试验中对质量转移故障进行了变化，评估了一个前沿模型和三种现成的本地可部署LLM。模型选择主导了诊断：前沿模型在85-90%的试验中将重心转移机制列为其前三个假设，而最佳本地模型为60-78%。推理分析表明，本地模型的成功与遵循完整的诊断程序相关，而较弱的模型则常常过早地判断电梯故障，尽管执行器跟踪其命令。在该数据集中，诊断和操作决策性能似乎并不相关。贡献在于一种架构，将意外故障恢复从检测扩展到缓解，以及一种用于评估低功耗AUV上LLM辅助任务管理的集成方法。

View on arXiv Download PDF AI Translation

cs.RO / 97 / 2609.20624

SmellDiffusion: Diffusion-Based Quadruped Navigation with Olfactory Scene Graphs

气味扩散：基于扩散的四足机器人导航与嗅觉场景图

Ogunwoye, Faith, Zhura, Iana, Amjad, Hajira, Kozlov, Timofei, Seyidov, Didar, Plotnikov, Dmitrii, Fedorov, Fedor, Tsetserukou, Dzmitry

Abstract

A robot sent to a named gas leak must preserve gas identity, estimate the source, and navigate to the resulting goal. We present SmellDiffusion, a simulation pipeline that represents species-specific gas zones in an open-vocabulary olfactory scene graph and shares the selected goal between classical and diffusion planners. Its key components are a peak-local geometric gate for selective source correction and diffusion-based, gas-guided trajectory generation. Among 424 unique source-wind configurations in solved flow, 28 have a concentration peak displaced more than 0.5m from the source. A source-independent geometric gate, calibrated only on the training split and evaluated at the observed peak, detects 9 of 10 held-out displacements at 0.64 precision. Gating a precomputed forward-matching correction reduces mean error on the displaced cases from 1.468m to 0.592m (60%), using matching for only 14/204 cases. All-case mean error falls from 0.205m to 0.180m. All planners receive the same scene-graph source estimate as their goal. In a controlled comparison, best-of-ten diffusion achieves mean gas exposure comparable to gas-guided A* (0.0476 versus 0.0455). A single diffusion proposal takes 41.7ms, compared with 72.3ms for gas-guided A*, although best-of-ten sequential sampling increases total runtime. Plain A* also reaches the same goal and remains the fastest and shortest-path method. Six matched Gazebo runs give mean robot-to-source errors of 0.39m for A* and 0.31m for diffusion.

Chinese Translation

被派往特定气体泄漏地点的机器人必须保持气体的身份，估计源头，并导航到相应的目标。我们提出了SmellDiffusion，一个模拟管道，它在开放词汇的嗅觉场景图中表示特定物种的气体区域，并在经典规划器和扩散规划器之间共享选定目标。其关键组件包括用于选择性源修正的峰值局部几何门和基于扩散的气体引导轨迹生成。在解决的流动中，424种独特的源-风配置中，有28种的浓度峰值偏离源头超过0.5米。一个源独立的几何门，仅在训练集上进行校准并在观察到的峰值处进行评估，能够以0.64的精度检测到10个保留偏移中的9个。对预计算的前向匹配修正进行门控，使得偏移案例的平均误差从1.468米降低到0.592米（降低60%），而仅在204个案例中的14个使用了匹配。所有案例的平均误差从0.205米降至0.180米。所有规划器都接收到相同的场景图源估计作为其目标。在一次受控比较中，十次最佳扩散的平均气体暴露与气体引导的A*相当（0.0476对比0.0455）。单个扩散提案的时间为41.7毫秒，而气体引导的A*为72.3毫秒，尽管十次最佳的顺序采样增加了总运行时间。普通的A*也达到了相同的目标，并且仍然是最快和最短路径的方法。六次匹配的Gazebo运行显示，A*的平均机器人到源头误差为0.39米，而扩散为0.31米。

View on arXiv Download PDF AI Translation

cs.RO / 98 / 2609.20629

RTK-Vision PPO for Autonomous Micro UAV Recovery on an Airborne Carrier

基于RTK视觉的PPO算法在空中载体上自主微型无人机回收中的应用

Sahu, Aashish, Kumar, R Prasanth

Abstract

Autonomous recovery of a micro unmanned aerial vehicle (UAV) onto a moving airborne carrier enables reusable deploy-mission-recover operation, but couples long-range rendezvous, close-range perception, carrier motion, aerodynamic interaction, and a discontinuous contact event. This paper presents an RTK-vision-guided reinforcement-learning framework in which a child UAV is physically transported by a larger carrier, takes off from the carrier while airborne, executes an independent sortie, returns to the carrier's current position, redocks, and subsequently descends with the carrier. Both vehicles carry RTK-GNSS, and the carrier continuously shares its navigation state with the child. Near the recovery deck, RTK remains active while a downward-facing camera with a fiducial marker detector provides marker-relative alignment cues. A proximal policy optimization (PPO) policy governing the terminal recovery phase is trained in a physics-based MuJoCo simulation environment with explicit sensor noise models, an aerodynamic disturbance surrogate, and marker-latency randomization, then transferred to hardware. PX4 retains low-level stabilization, and a deterministic safety gate authorizes descent independently of the learned policy. The PPO checkpoint achieves 99.55% success over 2,000 held-out randomized terminal episodes, compared with 78.4% for a tuned PD baseline under identical conditions, with a median planar terminal error of 6.62 cm. Across 14 outdoor trials, the full mission succeeds in 13 trials (92.9%), spanning both near-region recovery and recovery after the carrier translates away from the release point. The results demonstrate a complete autonomous aerial deployment-and-recovery cycle rather than an isolated landing maneuver, establishing a practical basis for reusable carrier-child operation in inspection, surveillance, and mobile-logistics applications.

Chinese Translation

自主回收微型无人机（UAV）到移动空中载体上实现了可重复使用的部署-任务-回收操作，但涉及长距离会合、近距离感知、载体运动、气动相互作用以及不连续接触事件。本文提出了一种基于RTK视觉引导的强化学习框架，其中子无人机由较大的载体物理运输，在空中从载体起飞，执行独立的出击任务，返回载体的当前位置，重新对接，并随后与载体一起下降。两种飞行器均配备RTK-GNSS，载体持续与子无人机共享其导航状态。在回收甲板附近，RTK保持活跃，同时向下的摄像头配备有标志物检测器，提供相对于标志物的对齐线索。控制终端回收阶段的近端策略优化（PPO）策略在基于物理的MuJoCo仿真环境中进行训练，该环境具备明确的传感器噪声模型、气动干扰替代物和标志物延迟随机化，随后转移到硬件上。PX4保持低级别的稳定性，并且一个确定性的安全门独立于学习的策略授权下降。在2000个随机终端实验中，PPO检查点的成功率达到99.55%，而在相同条件下调优的PD基线成功率为78.4%，中位平面终端误差为6.62厘米。在14次户外试验中，完整任务在13次试验中成功（92.9%），涵盖了近区域回收和载体远离释放点后的回收。结果展示了一个完整的自主空中部署与回收周期，而非孤立的着陆操作，为在检查、监视和移动物流应用中实现可重复使用的载体-子无人机操作奠定了实用基础。

View on arXiv Download PDF AI Translation

cs.RO / 99 / 2609.20646

TraceFlow: Guiding Frozen Flow-Matching Robot Policies with Success and Failure Traces

TraceFlow：通过成功与失败轨迹指导冻结流匹配机器人策略

Zhang, Jiaxuan, Liu, Ruizhe, Zhang, Yu, Yang, Yanchao

Abstract

A vision-language-action (VLA) policy with a flow-matching action expert generates each action chunk (a short command sequence) by integrating a learned velocity field; once its weights are fixed, the success or failure of an earlier rollout cannot change the chunk generated now. Concurrent test-time methods give a frozen policy such an input from retrieved successes, a learned critic, a verifier, or a dynamics model, but none uses the robot's own failed rollouts as negative evidence with nothing but a terminal outcome bit. We introduce TraceFlow, a progress-aligned guidance field that turns the action densities of retrieved successful and failed rollouts into a bounded correction to a frozen flow-matching action expert, using one terminal outcome bit per rollout and no other label. Its TraceBank stores traces, time-ordered state-action records with a terminal label, starts from the target-task training traces, and later admits the deployed robot's own rollouts. On an ordered real-robot packing task the base completes 21 of 50 trials in order, TraceFlow 39, and one stacking round without any weight update 47, with wrong-sequence episodes falling from 20 to 0. In simulation the gain is selective: with per-suite selected settings, TraceFlow raises RoboMemArena Sequence from 78.92\% to 91.50\% task success and Transferring from 54.41\% to 62.00\% at stacking round 2, leaves the 26-task aggregate unchanged, lowers Counting and Occlusion by 1.12 and 1.42 points, and changes LIBERO-Plus (Long) by +1.27 points (p = 0.0733). Stacking gains are finite, every branch peaking before round ten, and the bank's success-to-failure ratio predicts no retrieval allocation.

Chinese Translation

一种视觉-语言-动作（VLA）策略通过整合学习到的速度场生成每个动作块（一个短的指令序列）；一旦其权重固定，早期回放的成功或失败将无法改变当前生成的动作块。并行的测试时方法为冻结策略提供了来自检索成功的输入、学习的评论员、验证器或动力学模型，但没有一种方法使用机器人自身的失败回放作为负证据，仅依赖于最终结果位。我们引入了TraceFlow，这是一种进度对齐的指导场，将检索到的成功和失败回放的动作密度转化为对冻结流匹配动作专家的有界修正，使用每个回放一个终端结果位且不需要其他标签。其TraceBank存储轨迹，按时间顺序排列的状态-动作记录，带有终端标签，从目标任务训练轨迹开始，随后接纳部署机器人自身的回放。在一个有序的真实机器人打包任务中，基线完成了50次试验中的21次，TraceFlow完成了39次，而在没有任何权重更新的情况下，完成了一轮堆叠任务47次，错误序列的事件从20次降至0次。在仿真中，增益是选择性的：在每个套件选择的设置下，TraceFlow将RoboMemArena序列的任务成功率从78.92%提高到91.50%，将转移成功率从54.41%提高到62.00%（在堆叠第二轮），26个任务的总成功率保持不变，Counting和Occlusion分别降低了1.12和1.42点，LIBERO-Plus（长）变化为+1.27点（p = 0.0733）。堆叠增益是有限的，每个分支在第十轮之前达到峰值，而银行的成功与失败比率预测没有检索分配。

View on arXiv Download PDF AI Translation

cs.RO / 100 / 2609.20648

SkipVLA: Skipping VLA Steps with Classical Planning for Fast Robot Manipulation

SkipVLA：结合经典规划跳过 VLA 步骤以实现快速机器人操作

Agrawal, Kaivalya, Rahman, Md Ashiqur, Yeh, Raymond A., Kingston, Zachary

Abstract

Vision-Language-Action (VLA) models are a class of generalist robot policies that map camera images and language instructions directly to robot actions. While promising, these models remain slow at test time, particularly for long-horizon tasks that require many queries to the policy. Recent efforts reduce VLA latency by distilling smaller models, overlapping asynchronous action chunks, or pairing the VLA with a fast low-level policy, but still run a learned policy for the entire task. In contrast to VLA, classical motion planners quickly find collision-free motions, but require an explicit goal and have no semantic understanding of the task. In this work, we present SkipVLA, a hybrid policy that combines a pretrained VLA with a classical motion planner, using the planner for free-space motion and querying the VLA only for contact-rich skills such as grasping and placing. SkipVLA reuses the frozen vision-language backbone of the VLA to predict a target pose for each planned motion, and learns this predictor without additional demonstrations introduced into the system by using what was already learnt by the large VLA. We evaluate SkipVLA with three VLAs on 13 LIBERO tasks in simulation and three pick-and-place tasks on a physical 6-DoF YAM arm, demonstrating up to 2.5x faster task completion and significantly lower energy consumption while achieving the same task success rate.

Chinese Translation

视觉-语言-动作（VLA）模型是一类通用机器人策略，能够将相机图像和语言指令直接映射到机器人动作上。尽管前景可期，这些模型在测试时仍然较慢，尤其是在需要对策略进行多次查询的长时间任务中。近期的研究通过提炼更小的模型、重叠异步动作块或将 VLA 与快速低级策略配对来减少 VLA 的延迟，但仍需在整个任务中运行学习到的策略。与 VLA 相比，经典运动规划器能够快速找到无碰撞的运动，但需要明确的目标，并且对任务没有语义理解。在本研究中，我们提出了 SkipVLA，这是一种混合策略，将预训练的 VLA 与经典运动规划器结合，使用规划器进行自由空间运动，仅在需要接触丰富的技能（如抓取和放置）时查询 VLA。SkipVLA 重用 VLA 的冻结视觉-语言骨干网络来预测每个规划运动的目标姿态，并通过利用大型 VLA 已学习的内容来学习该预测器，而无需向系统引入额外的演示。我们在模拟环境中使用三种 VLA 对 13 个 LIBERO 任务进行评估，并在物理 6 自由度 YAM 臂上进行三项抓取与放置任务，结果表明任务完成速度提高了最多 2.5 倍，能耗显著降低，同时实现了相同的任务成功率。

View on arXiv Download PDF AI Translation

cs.RO / 101 / 2609.20649

DexTouch-WM: Learning Action-Conditioned Tactile World Models from Human Touch for Dexterous Robot Manipulation

DexTouch-WM：从人类触觉学习条件动作的灵巧世界模型以实现灵巧机器人操作

Qin, Yan, Chen, Yue, Lin, Wenwei, Liu, Shujia, Lyu, Chuqiao, Su, Kailun, Yu, Chenze, Luo, Ping, Ding, Wenbo, Chen, Tianxing, Xu, Renjing

Abstract

Learning predictive models of contact-rich dexterous manipulation requires dense tactile interaction, but such data are costly to scale on real robots and remain tied to embodiment-specific sensors. We introduce DexTouch-WM, an action-conditioned world model that learns from scalable human touch to jointly predict future RGB observations and bilateral tactile dynamics. Our insight is that human and robot manipulation share transferable contact dynamics when their tactile observations and action spaces are made compatible. We deploy flexible piezoresistive arrays with a shared sensing layout on both human and dexterous robot hands, and retarget human motion into the robot action space so that human interaction can supervise the same dynamics model used for real-robot prediction. DexTouch-WM couples a pretrained video expert with a lightweight tactile expert using anatomy-aware tactile tokens and aligned action conditioning. In human-to-robot scaling experiments, we keep five hours of real-robot supervision fixed while increasing human interaction from 0 to 100 hours, and observe substantial improvements in held-out robot-domain visual, geometric, and contact prediction despite disjoint human and robot task sets. Beyond prediction, we evaluate the world models as surrogate environments for policy evaluation and as generators of synthetic trajectories for real-robot policy learning, showing that scalable human interaction provides a complementary data axis for learning dexterous robot world models.

Chinese Translation

学习接触丰富的灵巧操作的预测模型需要密集的触觉交互，但在真实机器人上扩展此类数据成本高昂，并且仍然与特定于体现的传感器相关联。我们提出了DexTouch-WM，这是一种条件于动作的世界模型，它通过可扩展的人类触觉学习来共同预测未来的RGB观察和双边触觉动态。我们的见解是，当人类和机器人操作的触觉观察和动作空间兼容时，它们共享可转移的接触动态。我们在人体和灵巧机器人手上部署了具有共享传感布局的柔性压阻阵列，并将人类动作重新定向到机器人动作空间，以便人类交互可以监督用于真实机器人预测的相同动态模型。在人类到机器人扩展实验中，我们保持五小时的真实机器人监督不变，同时将人类交互从0增加到100小时，尽管人类和机器人任务集不重叠，但我们观察到在保留的机器人领域视觉、几何和接触预测方面有显著改善。除了预测，我们还将世界模型评估为策略评估的替代环境，以及用于真实机器人策略学习的合成轨迹生成器，显示出可扩展的人类交互为学习灵巧机器人世界模型提供了互补的数据维度。

View on arXiv Download PDF AI Translation

cs.RO / 102 / 2609.20659

HIL-UMI: Bringing Human-in-the-Loop Post-Training of Vision-Language-Action Models to Universal Manipulation Interface

HIL-UMI：将人机协同的视觉-语言-动作模型后训练引入通用操作界面

Han, Zimu, Zeng, Yiming, Zhang, Jiyao, Zhao, Zihao, Wang, Yuanfei, Jin, Yixiang, Li, Shiqi, Chen, Shuangben, Huang, Wei, Li, Ruodai, Shen, Hui, Dong, Hao

Abstract

Large-scale vision-language-action (VLA) models provide powerful priors for robot manipulation, yet adapting them to a specific deployment remains challenging. Supervised fine-tuning (SFT) on task-specific demonstrations provides a step toward deployment, but faces two persistent limitations: static data provide limited coverage of out-of-distribution states, and standard imitation objectives do not distinguish progressing behavior from less useful data. Interactive post-training can address these limitations, but typically requires repeated policy execution and human intervention on a physical robot. We introduce HIL-UMI, a policy-guided Universal Manipulation Interface (UMI) framework for robot-free human-in-the-loop VLA post-training. During handheld UMI demonstrations, HIL-UMI queries the current policy on the same observation stream without executing its predictions. The Energy Score compares the human action trajectory with policy inference and triggers collection when their discrepancy indicates an out-of-distribution region. In a separate feedback loop, low online advantage predictions identify essential segments for refining a progress-based advantage estimator. The updated estimator then guides advantage-conditioned behavioral cloning using a balanced mixture of base demonstrations and new policy data. This design preserves the iterative and policy-aware nature of human-in-the-loop learning while decoupling data collection from robot deployment. Experiments on four real-world tasks spanning long-horizon and precise manipulation show that HIL-UMI achieves consistent improvement over SFT and benefits from both targeted collection and advantage refinement. Moreover, HIL-UMI outperforms HG-DAgger on Clean Up Table with lower per-frame collection time, suggesting a scalable path for VLA post-training across operators and locations.

Chinese Translation

大规模的视觉-语言-动作（VLA）模型为机器人操作提供了强大的先验知识，但将其适应于特定部署仍然面临挑战。在任务特定演示上的监督微调（SFT）为部署提供了一步，但面临两个持续的限制：静态数据对分布外状态的覆盖有限，标准模仿目标无法区分进展行为与不太有用的数据。交互式后训练可以解决这些限制，但通常需要在物理机器人上反复执行策略并进行人工干预。我们提出了HIL-UMI，一个政策引导的通用操作界面（UMI）框架，用于无机器人的人机协同VLA后训练。在手持UMI演示过程中，HIL-UMI在不执行预测的情况下查询当前策略对同一观察流的反应。能量评分比较人类动作轨迹与策略推断，并在它们的差异表明处于分布外区域时触发数据收集。在一个单独的反馈循环中，低在线优势预测识别出用于优化基于进展的优势估计器的重要片段。更新后的估计器随后引导基于优势条件的行为克隆，使用基础演示和新策略数据的平衡混合。这一设计保留了人机协同学习的迭代和策略感知特性，同时将数据收集与机器人部署解耦。针对四个涵盖长时间跨度和精确操作的真实任务的实验表明，HIL-UMI在SFT上实现了一致的改进，并受益于目标收集和优势优化。此外，HIL-UMI在清理桌面任务中以更低的每帧收集时间超越了HG-DAgger，表明了在不同操作员和地点进行VLA后训练的可扩展路径。

View on arXiv Download PDF AI Translation

cs.RO / 103 / 2609.20669

Learning Foresight without Explicit Trajectories for 3D Diffusion Policies

无需显式轨迹的3D扩散策略前瞻学习

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

Abstract

3D diffusion policies are strong at generating geometrically grounded actions from current observations, but successful manipulation requires not only knowing what motion is feasible now, but also anticipating where the interaction is heading. Existing policies largely leave such foresight to emerge implicitly from action learning. We introduce Movement Trend Guidance, a simple but effective way to provide this foresight without introducing an explicit plan. From a short observation history, the policy learns a compact latent representation of interaction evolution. During training, sparse future gripper states supervise this representation; at inference, only the latent is retained as future-oriented conditioning alongside the current observation. The latent provides global conditioning for action generation, while an additional gated FiLM branch is used only at the UNet bottleneck. Despite adding only 3.52% more parameters to DP3, our method preserves the original dense-action and receding-horizon formulation and consistently improves upon DP3 across RoboTwin2.0, LIBERO-40, and DexArt. It reaches 62.8% vs. 56.1% in 50-task RoboTwin2.0 mixed training, 71.93% vs. 37.08% on LIBERO-40, and 72.0% vs. 49.0% on five real-robot tasks. These results show that a diffusion policy can benefit substantially from knowing where an interaction is heading, without being told exactly where to move.

Chinese Translation

3D扩散策略在从当前观察中生成几何基础动作方面表现出色，但成功的操控不仅需要知道当前可行的运动，还需要预测交互的未来方向。现有策略在很大程度上将这种前瞻性隐含于动作学习中。我们提出了运动趋势引导（Movement Trend Guidance），这是一种简单但有效的方法，可以在不引入显式计划的情况下提供这种前瞻性。通过短期观察历史，策略学习交互演变的紧凑潜在表示。在训练过程中，稀疏的未来抓取状态监督这一表示；在推理时，仅保留潜在表示作为面向未来的条件，与当前观察一起使用。潜在表示为动作生成提供全局条件，而额外的门控FiLM分支仅在UNet瓶颈处使用。尽管仅增加了3.52%的参数到DP3，我们的方法仍然保留了原始的密集动作和递归视野的公式，并在RoboTwin2.0、LIBERO-40和DexArt上始终优于DP3。在50任务的RoboTwin2.0混合训练中达到62.8%对56.1%，在LIBERO-40上达到71.93%对37.08%，在五个真实机器人任务上达到72.0%对49.0%。这些结果表明，扩散策略可以显著受益于了解交互的未来方向，而无需被明确告知移动的具体位置。

View on arXiv Download PDF AI Translation

cs.RO / 104 / 2609.20670

MAGNETAR: Multipath-Guided Spatial Posteriors for Transmitter Pose Inference in the Upper Mid-Band

MAGNETAR：用于上中频段发射机姿态推断的多路径引导空间后验

Lei, Haozhe, Chen, Ruibin, Jiang, Yuhan, Rasteh, Ali, Dhananjay, Aditya, Rangan, Sundeep

Abstract

Robots that localize a radio transmitter need more than a point estimate: in cluttered rooms, one measurement is often consistent with several transmitter locations and, because upper-mid-band antennas are directional, several headings. We present MAGNETAR, which infers a joint posterior over planar transmitter position and heading from a single asynchronous radio-frequency (RF) multipath snapshot, represented by angle-of-arrival and signal-to-noise-ratio estimates, given the room layout and receiver pose. Among our five neural scorers, MAGNETAR adopts a shared 2D U-Net conditioned on each candidate heading, jointly normalizing scores over a discretized position-heading grid. Training uses real-to-sim-calibrated 10 GHz simulations and a small measured subset. Grid-based joint posteriors outperform parametric ones on held-out simulations, the heading-conditioned scorer transfers best to robotic measurements, and fusing joint posteriors improves on fusing position-only marginals.

Chinese Translation

需要定位无线发射机的机器人不仅仅需要一个点估计：在杂乱的房间中，一次测量往往与多个发射机位置一致，并且由于上中频段天线是定向的，可能有多个方向。我们提出了MAGNETAR，它从单个异步无线电频率（RF）多路径快照中推断平面发射机位置和方向的联合后验，该快照由到达角和信噪比估计表示，前提是已知房间布局和接收器姿态。在我们的五个神经评分器中，MAGNETAR采用一个共享的2D U-Net，条件是每个候选方向，联合归一化在离散化的位置-方向网格上的评分。训练使用真实到仿真的10 GHz校准模拟和一个小的测量子集。基于网格的联合后验在保留的模拟中优于参数后验，方向条件评分器在机器人测量中表现最佳，而融合联合后验比仅融合位置边际更具优势。

View on arXiv Download PDF AI Translation

cs.RO / 105 / 2609.20680

Towards Scaling Marine Perception with Synthetic Data

利用合成数据扩展海洋感知的规模

Ma, Haoyu, Bagoren, Onur, Sheppard, Anja, Fandi, Elias, Edukulla, Ashrith, Aslan, Tanner, Sieh, Natasha, Song, Jingyu, Skinner, Katherine A.

Abstract

Scalable machine learning in challenging underwater environments is strongly limited by the lack of labeled real-world training data. This data is often expensive and laborious to gather, making large-scale real-world data challenging to gather and curate. However, simulated data can help close the gap, enabling many learning-based tasks for underwater perception. In this work, we extend OceanSim, an IsaacSim-based underwater perception simulator, with a Synthetic Data Generation (SDG) pipeline for training models to be used in underwater scenarios. The proposed pipeline enables users to generate large, automatically labeled, photorealistic datasets with configurable scene appearance, structure, and sensor settings. We evaluate the pipeline on a real-world sea urchin detection task and study how different forms of synthetic scene variation affect sim-to-real performance. Based on these experiments, we discuss findings on our results, main limitations of the current pipeline and identify future directions for improving underwater rendering fidelity, scene diversity, and the evaluation of sim-to-real generalization. The open-source code can be found at https://github.com/umfieldrobotics/OceanSim.

Chinese Translation

在具有挑战性的水下环境中，可扩展的机器学习受到缺乏标注真实世界训练数据的严重限制。这些数据通常收集成本高且劳动密集，使得大规模真实世界数据的收集和整理变得困难。然而，模拟数据可以帮助填补这一空白，使得许多基于学习的水下感知任务得以实现。在本研究中，我们扩展了OceanSim，这是一个基于IsaacSim的水下感知模拟器，增加了一个合成数据生成（Synthetic Data Generation, SDG）管道，用于训练可在水下场景中使用的模型。所提出的管道使用户能够生成大型、自动标注的、具有可配置场景外观、结构和传感器设置的逼真数据集。我们在一个真实世界的海胆检测任务上评估了该管道，并研究不同形式的合成场景变化如何影响模拟到真实的性能。基于这些实验，我们讨论了结果的发现、当前管道的主要局限性，并确定了改善水下渲染保真度、场景多样性以及模拟到真实泛化评估的未来方向。开源代码可在 https://github.com/umfieldrobotics/OceanSim 找到。

View on arXiv Download PDF AI Translation

cs.RO / 106 / 2609.20691

Custom PX4 firmware for autonomous hybrid aerial-marine missions

用于自主混合空中-海洋任务的定制PX4固件

Capuozzo, Andrea, Ruggiero, Fabio, Lippiello, Vincenzo

Abstract

Mapping and monitoring aquatic environments can benefit from hybrid aerial-amphibious drones able to combine flight and water-surface navigation within the same mission. This paper presents a PX4 firmware extension for such platforms, introducing manual and autonomous marine navigation modes integrated with the standard PX4 mission pipeline and QGroundControl interface. The proposed framework preserves existing flight functionalities and safety mechanisms while enabling unified planning and execution of hybrid aerial-marine missions with differentiated aerial and marine waypoints. Simulated case studies validate the implementation and demonstrate stable surface navigation under calm and wavy conditions.

Chinese Translation

对水域环境进行映射和监测可以受益于能够在同一任务中结合飞行和水面导航的混合空中-两栖无人机。本文提出了一种针对此类平台的PX4固件扩展，介绍了与标准PX4任务流程和QGroundControl界面集成的手动和自主海洋导航模式。所提出的框架在保留现有飞行功能和安全机制的同时，实现了混合空中-海洋任务的统一规划和执行，并支持区分空中和海洋航路点的设置。模拟案例研究验证了该实施方案，并展示了在平静和波动条件下的稳定水面导航。

View on arXiv Download PDF AI Translation

cs.RO / 107 / 2609.20694

HOPHY: A Hierarchical Hypergraph Representation for Off-Road Path and Mission Planning

HOPHY：一种用于越野路径和任务规划的分层超图表示

Meshram, Pranay, Adhivarahan, Charuvahan, Poddar, Prithvi, Esfahani, Ehsan Tarkesh, Wang, Chen, Chowdhury, Souma, Dantu, Karthik

Abstract

Mission-level autonomy for disaster response, search and rescue, and tactical UGV operations requires repeated path and mission planning as terrain conditions, agent types, and objectives change. Pixel-grid search is costly for repeated kilometer-scale queries, while semantic abstractions must maintain valid costs and connectivity as conditions change. We present HOPHY (Hierarchical Off-Road Planning using Hypergraphs), a reusable hierarchical terrain representation that organizes map-scale terrain into geometrically connected semantic regions (GSNodes), connectivity-preserving critical regions (Coarse Regions), and typed hyperedges for terrain, agent, and weather context. Hyperedge intersections select affected regions and incident edges for state updates without rebuilding the hierarchy. Across real off-road maps spanning kilometer-scale areas, HOPHY achieves 100% planning success and less than 0.01% median cost deviation from the oracle (pixel A*), with substantially lower query and replanning latency than the evaluated pixel and abstraction baselines. Applied to a multi-robot task-allocation (MRTA) problem, these gains reduce total computation by 79x over pixel A* and 7.2x over the fastest abstraction baseline, with mission makespan comparable to pixel A*. Finally, we demonstrate HOPHY on a physical Clearpath Jackal that successfully executes a 1.5-km, eight-task mission across mixed-surface outdoor terrain and a blockage-triggered replanned route.

Chinese Translation

灾害响应、搜索与救援以及战术无人地面车辆（UGV）操作的任务级自主性要求在地形条件、代理类型和目标变化时进行重复的路径和任务规划。对于重复的公里级查询，像素网格搜索成本高昂，而语义抽象必须在条件变化时保持有效的成本和连通性。我们提出了HOPHY（使用超图的分层越野规划），这是一种可重用的分层地形表示，将地图规模的地形组织为几何上连接的语义区域（GSNodes）、保持连通性的关键区域（Coarse Regions）和用于地形、代理和天气上下文的类型化超边。超边交集选择受影响区域和事件边，以便在不重建层次结构的情况下进行状态更新。在跨越公里级区域的真实越野地图上，HOPHY实现了100%的规划成功率，且与oracle（像素A*）的中位成本偏差小于0.01%，查询和重新规划延迟显著低于评估的像素和抽象基线。在应用于多机器人任务分配（MRTA）问题时，这些改进使得总计算量比像素A*减少了79倍，比最快的抽象基线减少了7.2倍，任务完成时间与像素A*相当。最后，我们在一台物理的Clearpath Jackal上演示了HOPHY，该机器人成功执行了一个1.5公里、八任务的使命，跨越混合表面的户外地形，并重新规划了因阻塞触发的路线。

View on arXiv Download PDF AI Translation

cs.RO / 108 / 2609.20709

MoWAM: Explicit Future Motion Prediction for Efficient World Action Models

MoWAM：高效世界动作模型的显式未来运动预测

Wang, Jiayu, Zhu, Bin, Yu, Yue, Chen, Jingjing

Abstract

World Action Models (WAMs) improve robot policy learning by incorporating future dynamics, yet explicitly generating future videos at inference introduces substantial computational overhead. Removing future generation improves efficiency, but leaves future dynamics only implicitly encoded in observation features, which can limit robustness under distribution shifts. We propose MoWAM, an efficient WAM that replaces future video generation with explicit future motion prediction. Instead of reconstructing the complete future scene, MoWAM models structured robot motion as a compact abstraction of the future, capturing how the robot is expected to evolve under the current scene and interaction constraints. A Mixture-of-Transformer architecture learns future visual dynamics during training while jointly predicting motion and action, allowing video generation to be removed entirely at inference while retaining an explicit representation of the future. The compact motion representation further enables efficient inference-time scaling by sampling multiple candidates of motion and action pairs and selecting among them with a motion-aware task-progress verifier. Experiments on LIBERO, LIBERO-Plus, and real-world manipulation tasks demonstrate that MoWAM achieves strong in-distribution performance, improved out-of-distribution robustness, and higher average real-world success than representative WAM baselines. In addition, performance improves as more candidates are explored, demonstrating that explicit future motion provides an effective and efficient basis for inference-time scaling.

Chinese Translation

世界动作模型（WAMs）通过结合未来动态来改善机器人策略学习，但在推理过程中显式生成未来视频会引入大量计算开销。去除未来生成可以提高效率，但会使未来动态仅以隐式方式编码在观察特征中，这可能在分布转移下限制鲁棒性。我们提出了MoWAM，一种高效的WAM，它用显式未来运动预测替代未来视频生成。MoWAM并不重建完整的未来场景，而是将结构化的机器人运动建模为未来的紧凑抽象，捕捉机器人在当前场景和交互约束下的预期演变。Mixture-of-Transformer架构在训练过程中学习未来视觉动态，同时联合预测运动和动作，从而在推理时完全去除视频生成，同时保留未来的显式表示。紧凑的运动表示进一步通过对运动和动作对的多个候选进行采样，并利用运动感知任务进度验证器进行选择，从而实现高效的推理时扩展。在LIBERO、LIBERO-Plus和真实世界操作任务上的实验表明，MoWAM在分布内表现强劲，分布外鲁棒性提高，且在真实世界的平均成功率高于代表性WAM基线。此外，随着探索更多候选，性能也有所提升，证明显式未来运动为推理时扩展提供了有效且高效的基础。

View on arXiv Download PDF AI Translation

cs.RO / 109 / 2609.20731

Underwater Visual Target Tracking with Target-Specific Depth Estimation and Adaptive Model-Fusion Predictive Control

基于目标特定深度估计和自适应模型融合预测控制的水下视觉目标跟踪

Zhou, Yuheng, Cheng, Haiyang, Feng, Yanqi, Fong, Pangkit, Lee, Mei Xuan, Gee, Marcus, Fang, Chongrong, He, Jianping

Abstract

Vision-based underwater target tracking is challenged by unreliable depth measurements and unknown target motion. This paper proposes a stereo visual-servoing framework for an autonomous underwater vehicle (AUV). For perception, the framework derives a stable 3D relative state from stereo images through target-specific depth extraction and Kalman filtering. It constructs a target-depth mask from color, disparity, and temporal cues to select reliable target pixels, and then filters the resulting depth measurement and detected image center separately. For control, the framework decouples yaw regulation from translational control, avoiding computationally expensive coupled multi-DOF optimization and enabling real-time translational MPC. The translational controller employs adaptive model-fusion predictive control, combining constant-velocity and zero-velocity target models to accommodate different target-motion patterns. It updates the model weights using historical prediction errors and computes translational commands subject to actuation, following-distance, and field-of-view constraints. Through simulations and real-world experiments, we validate the effectiveness of the proposed framework and show it has better performance than existing frameworks.

Chinese Translation

基于视觉的水下目标跟踪面临着不可靠的深度测量和未知的目标运动等挑战。本文提出了一种用于自主水下航行器（AUV）的立体视觉伺服框架。在感知方面，该框架通过目标特定深度提取和卡尔曼滤波，从立体图像中推导出稳定的三维相对状态。它利用颜色、视差和时间线索构建目标深度掩模，以选择可靠的目标像素，然后分别对得到的深度测量和检测到的图像中心进行滤波。在控制方面，该框架将偏航调节与平移控制解耦，避免了计算开销大的耦合多自由度优化，从而实现实时的平移模型预测控制（MPC）。平移控制器采用自适应模型融合预测控制，结合恒速和零速目标模型，以适应不同的目标运动模式。它使用历史预测误差更新模型权重，并计算受执行、跟随距离和视场约束的平移指令。通过仿真和实际实验，我们验证了所提框架的有效性，并显示其性能优于现有框架。

View on arXiv Download PDF AI Translation

cs.RO / 110 / 2609.20747

MILER: Semantic Mid-Level Representation for Sim-to-Real Reinforcement Learning in Unstructured Autonomous Driving

MILER：用于非结构化自主驾驶的模拟到现实强化学习的语义中级表示

Steinecker, Thomas, Trescher, Denis, Bienemann, Alexander, Luettel, Thorsten, Maehlisch, Mirko

Abstract

Reinforcement learning constitutes a promising approach owing to its potential for superhuman performance and self-learned policies. However, its application to real-world autonomous driving remains scarce, particularly in unstructured environments, because of the challenges associated with sim-to-real transfer for unstructured environments. In this work, we present MILER, an end-to-end policy framework with zero-shot sim-to-real transfer. During offline training, we employ a custom semantic mid-level representation (MLR) simulator and train the policy network using reinforcement learning, with its control outputs applied directly to a bicycle model. During deployment on the real vehicle, camera and LiDAR data are processed by BEVFusion to generate a semantic bird's-eye-view representation consistent with that of the MLR simulator. The actions generated by the policy network are not applied directly to the real vehicle. Instead, we employ a trajectory-alignment strategy that enables zero-shot sim-to-real transfer of both perception and control. We extensively evaluate the proposed framework on a diverse test track comprising numerous challenges, including various obstacles, hairpin curves, velocities of up to 33.6 km/h, and off-road sections. In total, we drove 17.3 km with two different vehicles on a 3.0 km test track without human intervention, thereby demonstrating the effectiveness of our approach. Furthermore, the entire software stack runs on a Jetson AGX Orin.

Chinese Translation

强化学习因其潜在的超人类表现和自学习策略而成为一种有前景的方法。然而，由于与非结构化环境的模拟到现实转移相关的挑战，其在真实世界自主驾驶中的应用仍然稀少。在本研究中，我们提出了MILER，一个具有零-shot模拟到现实转移的端到端策略框架。在离线训练期间，我们使用自定义的语义中级表示（MLR）模拟器，并通过强化学习训练策略网络，其控制输出直接应用于自行车模型。在真实车辆上的部署过程中，摄像头和激光雷达数据由BEVFusion处理，以生成与MLR模拟器一致的语义鸟瞰图表示。策略网络生成的动作并不直接应用于真实车辆。相反，我们采用了一种轨迹对齐策略，使得感知和控制的零-shot模拟到现实转移成为可能。我们在一个包含多种挑战的多样化测试轨道上对所提出的框架进行了广泛评估，包括各种障碍物、发夹弯、速度高达33.6公里/小时以及越野路段。总共，我们在一个3.0公里的测试轨道上以两种不同的车辆行驶了17.3公里，且没有人类干预，从而证明了我们方法的有效性。此外，整个软件堆栈在Jetson AGX Orin上运行。

View on arXiv Download PDF AI Translation

cs.RO / 111 / 2609.20756

OPTED: On-Policy Fine-Tuning for End-to-End Driving using a Render-Free Teacher

OPTED：基于策略的端到端驾驶微调方法，使用无渲染教师

Da Col, Damiano, Igl, Maximilian, Karkus, Peter, Chitta, Kashyap, Ivanovic, Boris, Pavone, Marco, Schindler, Konrad, Sakaridis, Christos

Abstract

As scaling pre-training data alone yields diminishing returns, post-training is becoming increasingly important across physical AI domains such as autonomous driving. End-to-end driving policies are pre-trained in open loop with behavior cloning on human demonstrations. However, compounding errors during closed-loop deployment can take the vehicle outside the training data distribution, increasing the risk of safety-critical incidents. Closed-loop post-training can mitigate this risk but requires costly simulation for sensor-based policies. We propose OPTED (on-policy fine-tuning for end-to-end driving) which decouples reinforcement learning from the post-training of the end-to-end policy: a privileged teacher is trained using RL on vectorized inputs (HD-map and bounding boxes). This teacher then provides supervision to the pre-trained student during closed-loop post-training. We apply OPTED to two camera-based models, TransFuser and VaVAM, and fine-tune them in AlpaSim, using neural reconstructions (3DGS) of real driving logs. Driving scores increase by factors of 1.6$\times$ and 9.5$\times$, respectively. In controlled experiments OPTED matches closed-loop performance with approximately three orders of magnitude fewer simulator interactions than direct RL post-training, while staying closer to the human prior. Project page: https://01dami23.github.io/opted/

Chinese Translation

随着单纯扩大预训练数据的收益递减，后训练在物理人工智能领域（如自动驾驶）变得愈发重要。端到端驾驶策略在开放循环中通过行为克隆对人类示范进行预训练。然而，在闭环部署过程中，累积误差可能使车辆超出训练数据分布，从而增加安全关键事件的风险。闭环后训练可以缓解这一风险，但需要为基于传感器的策略进行昂贵的仿真。我们提出了OPTED（基于策略的端到端驾驶微调），该方法将强化学习与端到端策略的后训练解耦：一个特权教师使用强化学习在向量化输入（高清地图和边界框）上进行训练。该教师随后在闭环后训练期间为预训练的学生提供监督。我们将OPTED应用于两个基于摄像头的模型，TransFuser和VaVAM，并在AlpaSim中进行微调，使用真实驾驶日志的神经重建（3DGS）。驾驶评分分别提高了1.6倍和9.5倍。在受控实验中，OPTED的闭环性能与直接强化学习后训练相比，所需的仿真交互次数减少了大约三个数量级，同时更接近人类先验。项目页面：https://01dami23.github.io/opted/

View on arXiv Download PDF AI Translation

cs.RO / 112 / 2609.20761

Agile-WAM: An Agile Tactile World Action Model for Contact-Rich Robot Control

Agile-WAM：一种用于接触丰富机器人控制的敏捷触觉世界动作模型

Zhou, Hanchu, Lynch, Brendan, Goyal, Raman, Gao, Dechen, Kasap, Begum, Zhao, Boqi, Zhang, Junshan

Abstract

World Action Models (WAMs) advance beyond conventional visuomotor policies by jointly predicting future world states and robot actions, enabling the policy to learn physical dynamics that support effective control. However, recent tactile WAMs often rely on large-scale pretrained generative backbones to capture contact-rich physical dynamics, which limit their inference efficiency and flexible deployment. In this paper, we present \ABBR{}, an agile tactile World Action Model for contact-rich robot control. \ABBR{} encodes visual and tactile observations into a shared latent that serves as the source of a direct vision-tactile-to-action flow-matching process, which can jointly generate latent representations of action chunks and future visual/tactile latents. A key observation is that vision and tactile signals evolve at inherently different timescales: adjacent visual frames are often highly similar, whereas tactile signals can change abruptly upon contact. We therefore introduce multi-horizon multimodal prediction in \ABBR{}, which provides supervision for visual latent at a larger temporal offset while predicting the tactile latent in the next frame to capture fine-grained contact dynamics. Across nine simulated and five real-world contact-rich manipulation tasks, \ABBR{} demonstrates strong and robust performance, outperforming the strongest baseline in success rate while maintaining low inference latency. In particular, in five real-world experiments, \ABBR{} yields a relative gain of $\textbf{29.4\%}$ in overall success rates while achieving inference latency of $\textbf{11.9 ms}$. These results demonstrate that multimodal WAM can be achieved with an agile architecture suitable for precise and high-frequency robot control. More details are available on our project page: https://hanchuzhou.github.io/TARO_project_page/.

Chinese Translation

世界动作模型（WAMs）超越了传统的视觉运动策略，通过共同预测未来的世界状态和机器人动作，使得策略能够学习支持有效控制的物理动态。然而，最近的触觉WAMs往往依赖于大规模预训练的生成骨干网络来捕捉接触丰富的物理动态，这限制了它们的推理效率和灵活部署。在本文中，我们提出了 extbf{Agile-WAM}，一种用于接触丰富机器人控制的敏捷触觉世界动作模型。 extbf{Agile-WAM}将视觉和触觉观察编码为共享潜在变量，作为直接的视觉-触觉-动作流匹配过程的源，这可以共同生成动作块和未来视觉/触觉潜在变量的潜在表示。一个关键观察是视觉和触觉信号在固有的时间尺度上演变不同：相邻的视觉帧通常高度相似，而触觉信号在接触时可能会突然变化。因此，我们在 extbf{Agile-WAM}中引入了多时间跨度多模态预测，这在较大的时间偏移下为视觉潜在变量提供监督，同时在下一帧中预测触觉潜在变量，以捕捉细粒度的接触动态。在九个模拟和五个真实世界的接触丰富操作任务中， extbf{Agile-WAM}展示了强大且稳健的性能，在成功率上超越了最强基线，同时保持低推理延迟。特别是在五个真实世界实验中， extbf{Agile-WAM}在整体成功率上实现了$ extbf{29.4 ext{%}}$的相对增益，同时实现了$ extbf{11.9 ms}$的推理延迟。这些结果表明，多模态WAM可以通过适合精确和高频机器人控制的敏捷架构实现。更多细节请访问我们的项目页面：https://hanchuzhou.github.io/TARO_project_page/.

View on arXiv Download PDF AI Translation

cs.RO / 113 / 2609.20776

GeoAAC: Geometry-Based Adaptive Action Chunking from Denoising Trajectories in VLA Policies

GeoAAC：基于几何的自适应动作分块方法用于去噪轨迹的视觉-语言-动作（VLA）策略

Chen, Xin, Chen, Sen, Ding, Yujuan, Liu, Jian, Wang, Guoqing, Ye, Wei, Shen, Heng Tao, Bin, Yi

Abstract

Action chunking is widely used for action generation and execution in Vision-Language-Action (VLA) policies, yet existing approaches commonly use a fixed action horizon. During a rollout, different task stages may require different levels of action continuity, control precision, and closed-loop feedback, making a fixed horizon unable to accommodate changing control requirements. We propose \textbf{GeoAAC}, a geometry-based adaptive action chunking method for flow-based VLA policies that adjusts the action horizon according to the reliability of the current action prediction. We show that the geometry of Flow Matching denoising trajectories provides process-level information for characterizing prediction reliability, with geometric variation across action prefixes remaining positively correlated with predictive uncertainty. GeoAAC uses this prefix-wise geometry to construct a horizon-wise geometric profile and adaptively determine the action horizon from a single generation without additional training. Experiments with GR00T N1.5 and {\pi}0.5 on LIBERO, LIBERO-Pro, RoboCasa365, and real-world manipulation tasks show consistent improvements over fixed-action-horizon baselines and existing adaptive methods, including up to 8.7 percentage points in simulation and an increase in average real-world success rate from 53.3\% to 74.4\%.

Chinese Translation

动作分块广泛应用于视觉-语言-动作（VLA）策略中的动作生成和执行，然而现有的方法通常使用固定的动作时间范围。在执行过程中，不同的任务阶段可能需要不同程度的动作连续性、控制精度和闭环反馈，使得固定的时间范围无法满足变化的控制需求。我们提出了 extbf{GeoAAC}，一种基于几何的自适应动作分块方法，针对基于流的VLA策略，根据当前动作预测的可靠性调整动作时间范围。我们展示了流匹配去噪轨迹的几何特征为表征预测可靠性提供了过程级信息，动作前缀的几何变化与预测不确定性保持正相关。GeoAAC利用这种前缀几何构建时间范围几何特征，并从单次生成中自适应地确定动作时间范围，而无需额外的训练。在LIBERO、LIBERO-Pro、RoboCasa365和真实世界操作任务中，使用GR00T N1.5和{ extpi}0.5的实验显示，相较于固定动作时间范围的基线和现有自适应方法，GeoAAC consistently 提高了性能，包括在模拟中提高了8.7个百分点，以及在真实世界成功率从53.3 ext% 提升至74.4 ext%。

View on arXiv Download PDF AI Translation

cs.RO / 114 / 2609.20791

StageGuard: Learning Stage Transitions for Long-Horizon Robot Tasks via Agentic Distillation

StageGuard：通过代理蒸馏学习长时间机器人任务的阶段转换

Huang, Jinbang, Hu, Yuanzhao, Li, Zhiyuan, Qi, Ran, Xiao, Yixin, Wu, Yangzheng, Ba, Tengyue, Zhang, Zhanguang, Zhang, Yingxue

Abstract

Hierarchical planning frameworks combine skills from multiple robot control policies for long-horizon task execution, where determining when to terminate the current skill and advance to the next subtask is essential. Existing approaches often rely on pre-designed completion signal checkers that are hard to obtain in real-world execution. Large-scale vision-language models (VLMs) offer strong reasoning capabilities, but their decision boundaries are not inherently aligned with task completion criteria, while cloud deployment and lengthy reasoning introduce substantial latency, limiting real-time monitoring. We propose StageGuard, an agentic distillation framework for accurate and efficient stage-transition decisions. StageGuard combines teacher-model reasoning with demonstration trajectories to generate structured explanations of subtask completion and policy switching. A lightweight student VLM uses these explanations to generate compact self-explanations, which are used for supervised fine-tuning. We evaluate stage-transition prediction on trajectories from two benchmarks and assess closed-loop task success through integration into hierarchical robot control on BEHAVIOR-1K, with further validation on real robots. Results show substantial improvements in stage-transition prediction while supporting efficient online monitoring.

Chinese Translation

层次规划框架结合了来自多个机器人控制策略的技能，以执行长时间任务，其中确定何时终止当前技能并推进到下一个子任务至关重要。现有方法通常依赖于预先设计的完成信号检查器，这在实际执行中难以获得。大规模视觉-语言模型（VLMs）提供了强大的推理能力，但它们的决策边界与任务完成标准并不天然对齐，而云部署和冗长的推理则引入了显著的延迟，限制了实时监控。我们提出了StageGuard，一种用于准确和高效阶段转换决策的代理蒸馏框架。StageGuard结合了教师模型推理与演示轨迹，以生成子任务完成和策略切换的结构化解释。一个轻量级的学生VLM利用这些解释生成紧凑的自我解释，这些自我解释用于监督微调。我们在两个基准的轨迹上评估阶段转换预测，并通过将其集成到BEHAVIOR-1K的层次机器人控制中评估闭环任务成功，进一步在真实机器人上进行验证。结果显示阶段转换预测有显著改善，同时支持高效的在线监控。

View on arXiv Download PDF AI Translation

cs.RO / 115 / 2609.20820

Workspace Models: Lightweight Robotic Memory via Saliency-Driven Supervision

工作空间模型：通过显著性驱动的监督实现轻量级机器人记忆

Dashora, Nitish, Chen, Douglas, Shenfeld, Idan, Marangola, John, Agrawal, Pulkit, Simchowitz, Max

Abstract

Complex robotic manipulation tasks frequently require a long-term memory of past events and actions. As conditioning on full histories renders policies prone to spurious correlations and degrades performance, many approaches to policy memory involve compressing historical information through expensive VLM queries in-the-loop to process only task-salient information. In this paper, we propose an alternative approach in which computationally intensive VLM queries are made during train-time to learn a lightweight latent memory that can be efficiently queried at deployment time. Our representation, which we call the \textbf{workspace token}, is trained by (1) using a VLM to identify current and historical information necessary for completing a task, then (2) distilling these into the workspace token using a set-reconstruction decoder loss. In both simulation and hardware, we show that the workspace token can be used as a drop-in replacement for observations during deployment, enabling policies to solve memory-intensive tasks without the need for VLM reasoning in-the-loop. Interestingly, we found that workspace tokens are not only more lightweight but also lead to better policy performance.

Chinese Translation

复杂的机器人操作任务通常需要对过去事件和动作的长期记忆。由于依赖完整历史会导致策略容易受到虚假关联的影响，从而降低性能，因此许多策略记忆的方法涉及通过昂贵的 VLM 查询来压缩历史信息，以处理仅与任务相关的信息。在本文中，我们提出了一种替代方法，在训练时进行计算密集型的 VLM 查询，以学习一种轻量级的潜在记忆，该记忆可以在部署时高效查询。我们称之为 extbf{工作空间标记} 的表示方法，通过以下步骤进行训练：（1）使用 VLM 识别完成任务所需的当前和历史信息，然后（2）通过集合重构解码器损失将这些信息提炼为工作空间标记。在模拟和硬件实验中，我们展示了工作空间标记可以作为部署期间观察的替代方案，使得策略能够解决内存密集型任务，而无需在循环中进行 VLM 推理。有趣的是，我们发现工作空间标记不仅更轻量，而且还导致更好的策略性能。

View on arXiv Download PDF AI Translation

cs.RO / 116 / 2609.20822

Coding Agents with an Obstacle-Aware Harness for Safe Robot Manipulation

具有障碍物感知装置的编码代理以实现安全的机器人操作

Xu, Bingxin, Shang, Yuzhang, Dong, Zhen, Ferrara, Emilio

Abstract

Coding agents have emerged as a promising paradigm for robot manipulation: a language model writes the robot controller as a program, and agents built in this way now operate robots without robot-specific training.Whether this paradigm is also safe, however, has not been asked. We evaluate coding agent under a safety constraint, where each task pairs a manipulation goal with an obstacle the robot must not touch. The agent pursues the goal but collides with the obstacle in most cases, treating task completion as its sole objective while neglecting safety. The agent reasons about the obstacle in its traces, and the prompt already forbids touching it, so neither perception nor instruction is at fault; the fault lies in the planning, where the stated constraint never becomes a priority. By decomposing manipulation into a route phase and a contact-rich moment, we locate the source of the failure. Along the route, the model cannot prioritize the safety constraint, having no notion of a clearing route and none of replanning once a chosen route becomes infeasible. At the contact, it is unaware that contact execution is bounded by the same constraint. To close this gap, we present SafeHarness, which equips the model with two obstacle-aware harnesses that enable it to prioritize the safety constraint. Obstacle-aware route planning grounds the objects as bounding boxes and draws candidate routes over them as sequences of waypoints. The agent then plans a route in advance, verifies it, replans when necessary, and only then executes it. Obstacle-aware contact execution instead selects the contact position so that the contact itself avoids the obstacle. SafeHarness attains 71.9% task success and 87.5% collision avoidance, surpassing the previous SOTA by 6.5% and 27.0%, respectively. These results are $2.3\times$ and $1.5\times$ those of the same agent without harnesses.

Chinese Translation

编码代理已成为机器人操作的一个有前景的范式：一种语言模型将机器人控制器编写为程序，这种方式构建的代理现在能够在没有特定于机器人的训练下操作机器人。然而，这一范式是否安全尚未被探讨。我们在安全约束下评估编码代理，其中每个任务将操作目标与机器人必须避免接触的障碍物配对。代理追求目标，但在大多数情况下与障碍物发生碰撞，将任务完成视为唯一目标，而忽视安全。代理在其轨迹中对障碍物进行推理，提示已经禁止接触，因此无论是感知还是指令都没有错误；问题出在规划上，所述约束从未成为优先事项。通过将操作分解为路径阶段和接触丰富的时刻，我们找到了失败的根源。在路径阶段，模型无法优先考虑安全约束，因为它没有清理路径的概念，也没有在选择的路径变得不可行后进行重新规划。在接触时，它没有意识到接触执行受到相同约束的限制。为了解决这一问题，我们提出了SafeHarness，它为模型配备了两个障碍物感知装置，使其能够优先考虑安全约束。障碍物感知路径规划将物体视为边界框，并在其上绘制候选路径作为一系列航点。代理随后提前规划路径，进行验证，并在必要时重新规划，只有在此之后才执行路径。障碍物感知接触执行则选择接触位置，以避免接触本身与障碍物发生碰撞。SafeHarness实现了71.9%的任务成功率和87.5%的碰撞避免率，分别超过了之前的最先进技术（SOTA）6.5%和27.0%。这些结果是没有使用装置的同一代理的$2.3 imes$和$1.5 imes$。

View on arXiv Download PDF AI Translation

计算机视觉 (Computer Vision)

101

cs.CV / 1 / 2609.19230

Open ultrasound foundation model for robust segmentation and clinical measurement across heterogeneous settings

开放超声基础模型用于异构环境下的稳健分割和临床测量

Qin, Chao, Khan, Fahad Shahbaz, Khan, Salman, Ather, Sarim, Anwar, Siddiq, Anwer, Rao Muhammad, Khan, Shadab

Abstract

Ultrasound is the most widely deployed imaging modality worldwide, yet clinical AI remains fragmented into narrow single-task models that fail when device, operator, or anatomy changes. Here we present SonoCorpus, an open resource unifying 456,963 images and 1,626,085 expert masks from 53 public datasets spanning 24 clinical applications and 17 countries, and SonoBase, an interactive segmentation foundation model pretrained on it. Across fifteen evaluation datasets introducing new organs, devices, operators, and geographies, SonoBase outperforms SAM2, MedSAM2, and the concept-promptable MedSAM3 on every dataset and matches per-dataset specialist models trained on the same data; on fully external data it exceeds the accuracy these baselines achieve on their own in-distribution benchmarks. Ejection fraction derived from its segmentations falls within inter-observer variability (6.63\% error), with fewer misclassifications at the defibrillator-candidacy threshold than either promptable baseline (13\% versus 18--42\%); fetal head-circumference (1.81~mm) and gestational-age (1.2 days) errors fall below inter-observer variability. Where a baseline fails outright, one in four test cases, SonoBase recovers a usable segmentation in 81\% of them, including on handheld probes operated by minimally trained users in two low- and middle-income countries (Sierra Leone and Tanzania). Five labeled examples can help the model adapt to a new setting, and the identical training protocol transfers well to newer models such as SAM3, locating the advantage in ultrasound-specific pretraining rather than any single architecture. To ensure reproducibility and enable the community to build on SonoBase as a platform, we release all checkpoints, optimizer states, data-split indices, deduplication hashes, and starter code.

Chinese Translation

超声是全球应用最广泛的成像方式，然而临床人工智能仍然分散为狭窄的单任务模型，这些模型在设备、操作员或解剖结构变化时会失效。在此，我们提出了SonoCorpus，这是一个统一的开放资源，包含来自53个公共数据集的456,963张图像和1,626,085个专家标注，涵盖24个临床应用和17个国家。同时，我们还推出了SonoBase，这是一个基于该资源预训练的交互式分割基础模型。在引入新器官、设备、操作员和地理环境的十五个评估数据集中，SonoBase在每个数据集上均超越了SAM2、MedSAM2和概念可提示的MedSAM3，并且与在相同数据上训练的每个数据集的专业模型相匹配；在完全外部数据上，其准确性超过了这些基线在其自身分布基准上所达到的水平。基于其分割结果计算的射血分数在观察者间变异性范围内（误差为6.63%），在除颤器候选阈值下的错误分类率低于任何可提示基线（13%对比18-42%）；胎头周长（1.81 mm）和妊娠周数（1.2天）的误差均低于观察者间变异性。在基线完全失效的情况下，SonoBase在81%的测试案例中恢复了可用的分割结果，包括在两个低收入和中等收入国家（塞拉利昂和坦桑尼亚）由训练不足的用户操作的手持探头上。五个标注示例可以帮助模型适应新的环境，且相同的训练协议能够很好地迁移到更新的模型如SAM3，优势在于超声特定的预训练，而非任何单一架构。为了确保可重复性并使社区能够在SonoBase平台上进行构建，我们发布了所有检查点、优化器状态、数据分割索引、去重哈希和起始代码。

View on arXiv Download PDF AI Translation

cs.CV / 2 / 2609.19236

RAUL: Reference-Assisted Ureteroscopy Localization for Skill Assessment

RAUL：参考辅助的输尿管镜定位用于技能评估

Li, Fangjie, Bui, Mai, Mohan, Charan, Miga, Michael, Chabanas, Matthieu, Kavoussi, Nicholas, Wu, Jie Ying

Abstract

Objective: Incomplete navigation of anatomy during ureteroscopic kidney stone surgeries can contribute to repeat interventions. While skilled surgeons have lower reintervention rates, there are no objective metrics to quantify scope-navigation performance to evaluate when a trainee becomes skilled. This work aims to recover ureteroscope trajectories from endoscopic video and derive navigation metrics to quantify differences in skill. Methods: We propose RAUL, a reference-assisted reconstruction framework for recovering ureteroscope trajectories from ureteroscope videos only in phantoms. For each phantom, we use a slow, high-quality reference exploration video to generate a reference reconstruction. We localize subsequent exploration videos against this reference. We evaluate localization accuracy against electromagnetically tracked scope pose. We compute navigation metrics from phantom exploration trajectories to compare surgical residents across experience levels. Results: The proposed reference-assisted framework achieves a mean translation root mean square error of $0.5 \pm 0.1$ mm across 9 phantoms. Compared to standard Structure-from-Motion (SfM), the proposed pipeline increases frame-wise localization coverage from $50.5 \pm 14.9\%$ to $86.1 \pm 7.2\%$ of all video frames. The reconstructed trajectories revealed significant differences between high- and low-experience trainees in established navigation metrics. Conclusion: RAUL enables substantially more complete recovery of ureteroscope trajectories from videos compared to standard SfM pipelines, enabling trajectory-based skill assessment without additional tracking equipment. Significance: To the best of our knowledge, this is the first use of video-only recovery of ureteroscope trajectories without external tracking sensors for skill assessment, supporting scalable automated assessment of ureteroscopy navigation skill.

Chinese Translation

目的：在输尿管镜肾结石手术中，解剖结构导航不完整可能导致重复干预。尽管熟练的外科医生具有较低的再干预率，但目前没有客观指标来量化内窥镜导航性能，以评估受训者何时达到熟练水平。本研究旨在从内窥镜视频中恢复输尿管镜轨迹，并推导导航指标以量化技能差异。方法：我们提出了RAUL，一个参考辅助重建框架，仅在模型中从输尿管镜视频恢复输尿管镜轨迹。对于每个模型，我们使用慢速、高质量的参考探索视频生成参考重建。我们将后续的探索视频与该参考进行定位。我们评估相对于电磁跟踪的镜头姿态的定位准确性。我们从模型探索轨迹中计算导航指标，以比较不同经验水平的外科住院医生。结果：所提出的参考辅助框架在9个模型中实现了平均平移均方根误差为$0.5 imes 0.1$ mm。与标准的运动结构重建（Structure-from-Motion, SfM）相比，所提出的流程将逐帧定位覆盖率从$50.5 imes 14.9 ext{ extperthousand}$提高到$86.1 imes 7.2 ext{ extperthousand}$。重建的轨迹显示高经验和低经验受训者在既定导航指标上存在显著差异。结论：与标准SfM流程相比，RAUL显著提高了从视频中恢复输尿管镜轨迹的完整性，实现了基于轨迹的技能评估，而无需额外的跟踪设备。意义：据我们所知，这是首次在没有外部跟踪传感器的情况下，仅通过视频恢复输尿管镜轨迹用于技能评估，支持可扩展的自动化输尿管镜导航技能评估。

View on arXiv Download PDF AI Translation

cs.CV / 3 / 2609.19354

Can Vision-Language Models Judge Olympic Diving? From Reasoning to Scores in Zero-Shot Action Quality Assessment

视觉-语言模型能否评判奥林匹克跳水？从推理到零样本动作质量评估的分数

Velesaca, Henry O., Freire-Obregon, David, Miranda, Luigi, Reyes-Angulo, Abel

Abstract

Automated action quality assessment (AQA) in Olympic sports remains a challenging task due to the complexity of human motion and the subjectivity inherent in expert judging. This work evaluates the capability of open-source Vision-Language Models (VLMs) to perform zero-shot action quality assessment on Olympic diving videos using the AQA-7 benchmark dataset. In this regard, a regression-based framework is pro-posed to leverage both the semantic reasoning and phase-level sub-scores generated by the VLMs, combining TF-IDF vectorization, dimensionality reduction, and ensemble learning to predict final competition scores. Experimental results show that standalone VLMs achieve moderate Spearman correlations below 0.32, while the proposed ensemble regression framework substantially improves performance in the reported evaluation, reaching a Spearman correlation of 0.67 with a four-model configuration. Textual reasoning features con-sistently outperformed raw numerical sub-scores, highlighting the richness of VLM-generated explanations for action quality analysis. These findings suggest that VLMs hold strong potential as assistive tools for explainable and semi-automated sports performance evaluation. The code is publicly available on GitHub https://github.com/hvelesaca/olympic diving judge vlm

Chinese Translation

在奥林匹克体育中，自动化动作质量评估（AQA）仍然是一项具有挑战性的任务，原因在于人类动作的复杂性和专家评判中固有的主观性。本研究评估了开源视觉-语言模型（VLMs）在奥林匹克跳水视频上进行零样本动作质量评估的能力，使用了AQA-7基准数据集。在这方面，提出了一种基于回归的框架，利用VLMs生成的语义推理和阶段性子分数，结合TF-IDF向量化、降维和集成学习来预测最终比赛分数。实验结果表明，单独的VLMs在斯皮尔曼相关性上达到了0.32以下的中等水平，而所提出的集成回归框架在报告的评估中显著提高了性能，达到了0.67的斯皮尔曼相关性，使用了四个模型的配置。文本推理特征始终优于原始数值子分数，突显了VLM生成的解释在动作质量分析中的丰富性。这些发现表明，VLMs作为可解释和半自动化体育表现评估的辅助工具具有很大的潜力。代码已在GitHub上公开发布，网址为 https://github.com/hvelesaca/olympic diving judge vlm

View on arXiv Download PDF AI Translation

cs.CV / 4 / 2609.19358

Open-vocabulary 3D object detection with promptable segmentation

开放词汇的3D物体检测与可提示分割

Deniz, Ömer Faruk, Koçyiğit, Mustafa Taha

Abstract

Three-dimensional object detection for autonomous driving is dominated by detectors trained on large corpora of human-annotated 3D boxes. Such a detector learns a fixed category list, and everything outside it is invisible. This paper asks whether the task can be solved training-free and open-vocabulary. A promptable segmentation model (SAM3), queried with class names as text prompts, supplies instance masks in the vehicle's six surround-view cameras, and the masks are turned into metric 3D boxes using the geometry of the scene. The core is a controlled three-stage comparison on nuScenes in which 2D detection is held fixed and only the source of 3D geometry changes. Geometry predicted from images alone reaches 0.183 mean average precision (mAP) under the official protocol; fitting boxes from raw LiDAR points inside the same masks with training-free rules reaches 0.298 mAP / 0.348 nuScenes detection score (NDS) at zero labeling cost; borrowing supervised box geometry at inference time lifts the same detections to 0.413 mAP / 0.555 NDS, which locates the pipeline's largest deficit in measurement precision rather than 2D detection, while class confusion and confidence calibration survive that substitution. Reversing the direction, a three-state camera-witness rule built from the same masks improves a supervised LiDAR-only detector from 0.596 to 0.630 mAP, roughly half the gain of fully supervised camera fusion, with no training. A coverage analysis shows that SAM3 finds 84% of in-range objects with a correctly named mask; the classes that fail in the official metric are misnamed or geometrically unforgiving, not unseen.

Chinese Translation

针对自动驾驶的三维物体检测主要依赖于在大量人工标注的3D框上训练的检测器。这种检测器学习一个固定的类别列表，超出该列表的内容则不可见。本文探讨是否可以在无训练和开放词汇的情况下解决该任务。一个可提示的分割模型（SAM3），通过文本提示的类别名称进行查询，为车辆的六个环视摄像头提供实例掩码，并利用场景的几何信息将掩码转化为度量3D框。核心是对nuScenes进行的受控三阶段比较，其中2D检测保持不变，仅3D几何的来源发生变化。仅从图像预测的几何在官方协议下达到0.183的平均精度（mAP）；在相同掩码内，使用无训练规则从原始LiDAR点拟合框达到0.298 mAP / 0.348 nuScenes检测分数（NDS），且标注成本为零；在推理时借用监督框几何将相同检测提升至0.413 mAP / 0.555 NDS，这表明管道的最大缺陷在于测量精度而非2D检测，同时类别混淆和置信度校准在该替代中依然存在。反向推导，基于相同掩码构建的三状态摄像头见证规则将一个仅使用LiDAR的监督检测器的mAP从0.596提升至0.630，约为完全监督摄像头融合增益的一半，且无需训练。覆盖分析表明，SAM3能够以正确命名的掩码找到84%的范围内物体；在官方指标中失败的类别是命名错误或几何上不宽容，而非未见。

View on arXiv Download PDF AI Translation

cs.CV / 5 / 2609.19377

LinePilot Digitizer: Line-Plot Recovery with Manual and Automatic Calibration

LinePilot 数字化工具：手动与自动校准的线图恢复

Ma, Fengbo, Akhtar, Rayan, Joshi, Aakash H., Li, Xiaoting, Sun, Haijian, Xiang, Zhen, Chen, Xianyan, Zhao, Yiping

Abstract

Recovering numerical series from line plots requires accurate axis calibration and reliable curve extraction. We present LinePilot Digitizer (LinePilot), which combines continuous color-based curve recovery with three calibration modes: LinePilot (standard), LinePilot (enhanced), and LinePilot (OCR). We also introduce DigitizerBench, the first dedicated benchmark for systematically evaluating digitizer performance, using an orthogonal design spanning signal, rendering, and plot-structure factors with complementary automatic and human-guided evaluations. We evaluate performance using failure-penalized capped normalized root-mean-square error (FPC-NRMSE), which assigns unit loss to missing, unusable, or catastrophically inaccurate outputs. On DigitizerBench-Full, LinePilot (OCR) achieves the lowest mean FPC-NRMSE (0.672) and highest trusted usability (38.2%) among the tested automatic pipelines. On DigitizerBench-Lite, LinePilot (enhanced) achieves the lowest mean FPC-NRMSE (0.081), 100% output success, and highest trusted usability (93.3%). The orthogonal benchmark design further enables factor analysis to identify the factors that most significantly affect digitizer performance. Together, the three calibration modes provide a practical trade-off between automation, user control, and accuracy within a shared curve-recovery workflow.

Chinese Translation

从线图中恢复数值序列需要准确的轴校准和可靠的曲线提取。我们提出了 LinePilot 数字化工具（LinePilot），它结合了基于连续颜色的曲线恢复和三种校准模式：LinePilot（标准）、LinePilot（增强）和 LinePilot（OCR）。我们还介绍了 DigitizerBench，这是第一个专门用于系统评估数字化工具性能的基准，采用正交设计，涵盖信号、渲染和图表结构因素，并结合自动和人工指导的评估。我们使用失败惩罚的限制归一化均方根误差（FPC-NRMSE）来评估性能，该指标对缺失、不可用或极其不准确的输出分配单位损失。在 DigitizerBench-Full 上，LinePilot（OCR）在测试的自动管道中实现了最低的平均 FPC-NRMSE（0.672）和最高的可信可用性（38.2%）。在 DigitizerBench-Lite 上，LinePilot（增强）实现了最低的平均 FPC-NRMSE（0.081）、100% 的输出成功率和最高的可信可用性（93.3%）。正交基准设计进一步支持因素分析，以识别对数字化工具性能影响最大的因素。三种校准模式共同提供了在共享曲线恢复工作流程中自动化、用户控制和准确性之间的实用权衡。

View on arXiv Download PDF AI Translation

cs.CV / 6 / 2609.19384

Riemannian--Lorentz Fusion of Vision Transformers and State-Space Models

黎曼-洛伦兹视觉变换器与状态空间模型的融合

Patro, Badri N., Agneeswaran, Vijay S.

Abstract

Scaling deep learning faces critical bottlenecks: data exhaustion, exponential training costs, and resource concentration. Model merging combines pre-trained checkpoints without gradient descent, offering orders-of-magnitude savings versus retraining. Combining independently trained vision models is difficult when their architectures and parameter shapes differ. Existing weight-space merging methods generally assume aligned, shape-compatible checkpoints, whereas a Vision Transformer (ViT) and a state-space model (SSM) implement token mixing with different operators. We study a hybrid Heterogeneous merging setting that retains both architectures while aligning parameter groups by semantic role. Our proposed Riemannian--Lorentz Parameter Fusion (RLPF) method projects aligned groups to common coordinates, lifts selected coordinates to the Lorentz hyperboloid model of hyperbolic space, computes a regularized geodesic barycenter, and decodes the result into the two branches. A learned gate then combines branch logits for each input. Component groups use fixed curvature values, with normalization parameters treated as Euclidean. In the results available in this manuscript, the fine-tuned system obtains 82.37\% on CIFAR-10, 75.04\% on Oxford-IIIT Pet, and 78.58\% top-1 accuracy on ImageNet-1K; the corresponding best-parent accuracies are 76.54\%, 71.42\%, and 76.42\%. On ImageNet-1K, the reported pre-fine-tuning initialization reaches 77.80\%. These results support further study of geometry-aware heterogeneous fusion, but not a training-free single-checkpoint merge: RLPF is a two-branch hybrid whose gate and reported final models are trained.

Chinese Translation

深度学习的扩展面临着关键瓶颈：数据耗尽、训练成本呈指数增长以及资源集中。模型合并在不进行梯度下降的情况下结合预训练检查点，相比于重新训练可节省数量级的资源。当视觉模型的架构和参数形状不同时，独立训练的视觉模型的结合变得困难。现有的权重空间合并方法通常假设检查点是对齐且形状兼容的，而视觉变换器（ViT）和状态空间模型（SSM）使用不同的运算符实现令牌混合。我们研究了一种混合的异构合并设置，该设置保留了两种架构，同时通过语义角色对参数组进行对齐。我们提出的黎曼-洛伦兹参数融合（RLPF）方法将对齐的组投影到公共坐标系，将选定坐标提升到超曲面空间的洛伦兹双曲模型，计算正则化的测地线重心，并将结果解码到两个分支中。一个学习的门控机制随后为每个输入组合分支的逻辑值。组件组使用固定的曲率值，归一化参数被视为欧几里得。在本手稿中提供的结果中，微调后的系统在CIFAR-10上获得82.37%的准确率，在Oxford-IIIT Pet上获得75.04%的准确率，在ImageNet-1K上获得78.58%的top-1准确率；相应的最佳父模型准确率分别为76.54%、71.42%和76.42%。在ImageNet-1K上，报告的预微调初始化达到77.80%。这些结果支持对几何感知异构融合的进一步研究，但不支持无训练的单检查点合并：RLPF是一个双分支混合模型，其门控和报告的最终模型是经过训练的。

View on arXiv Download PDF AI Translation

cs.CV / 7 / 2609.19393

WZPlanner: Safe End-to-End Path Planning for Autonomous Driving in Work Zones

WZPlanner：在施工区域中进行安全的端到端路径规划的自动驾驶

Sahu, Nishad, Qian, Changzhong, Cai, Guangzhou, Sural, Shounak, Ragunathan, Rajkumar

Abstract

Work zones alter lane geometry through temporary traffic controls and closures that may be absent from on-board maps, challenging autonomous vehicle (AV) perception and planning. Generalization is also limited by scarce public datasets with structured geometric supervision. We present WorkZonePlan, a dataset comprising 149K+ synthetic and 5K+ real-world multimodal samples with 3D annotations for lane boundaries, work zone boundaries, and driving trajectory options. It also provides 76 closed-loop CARLA scenarios replayed under three weather conditions, yielding 228 Bench2Drive-format evaluation routes. We introduce WAVE (Work-zone-focused AV data generation in Virtual and rEal Environments), a semi-automated pipeline for creating the dataset, and BoundaryFormer (BF), a transformer-based model that jointly predicts lane and work zone boundary polynomials and driving trajectories. BF uses slot attention for boundary prediction. Ablations show that a separate trajectory decoder using boundary slot features substantially improves trajectory prediction over a slot-attention-only approach. Building on this finding, BF++ offers Camera and Camera+LiDAR variants with metric ground-plane encoding, typed boundary/trajectory queries, long-range point anchors, image-space curve refinement, and conservative gated LiDAR fusion. On the 211 routes common to all four models at the evaluation freeze, BF++-Camera and BF++-Camera+LiDAR achieve Driving Scores of 63.0 and 64.4, respectively, compared with 59.3 for SimLingo and 26.1 for TransFuser++ (TF++). BF++ is 40 times smaller than SimLingo and more than 10 times smaller than TF++, while achieving higher Driving Scores. These results support jointly predicting lane boundaries, work zone boundaries, and driving trajectories as a promising direction toward safer AV operation in work zones. Code and dataset: https://github.com/Nishad-Sahu/WZPlanner.

Chinese Translation

施工区域通过临时交通控制和封闭改变车道几何形状，这些变化可能在车载地图中缺失，从而对自动驾驶车辆（AV）的感知和规划提出挑战。由于缺乏具有结构化几何监督的公共数据集，泛化能力也受到限制。我们提出了WorkZonePlan，一个包含149K+合成样本和5K+真实世界多模态样本的数据集，提供了车道边界、施工区域边界和驾驶轨迹选项的3D标注。该数据集还提供了76个在三种天气条件下重播的闭环CARLA场景，生成228条Bench2Drive格式的评估路线。我们引入了WAVE（Work-zone-focused AV data generation in Virtual and rEal Environments），一个用于创建该数据集的半自动化管道，以及BoundaryFormer（BF），一个基于变换器的模型，能够联合预测车道和施工区域边界多项式及驾驶轨迹。BF使用槽注意力进行边界预测。消融实验表明，使用边界槽特征的独立轨迹解码器在轨迹预测上显著优于仅使用槽注意力的方法。基于这一发现，BF++提供了Camera和Camera+LiDAR变体，具有度量平面编码、类型边界/轨迹查询、长距离点锚、图像空间曲线细化和保守的门控LiDAR融合。在评估冻结时，四个模型共有的211条路线中，BF++-Camera和BF++-Camera+LiDAR的驾驶得分分别为63.0和64.4，而SimLingo为59.3，TransFuser++（TF++）为26.1。BF++的体积是SimLingo的40倍小，且比TF++小10倍以上，同时实现了更高的驾驶得分。这些结果支持联合预测车道边界、施工区域边界和驾驶轨迹作为在施工区域中实现更安全的自动驾驶操作的有前景的方向。代码和数据集链接：https://github.com/Nishad-Sahu/WZPlanner。

View on arXiv Download PDF AI Translation

cs.CV / 8 / 2609.19421

RGS: Reflection-aware Gaussian Splatting via Learning Geometry Continuity for Reflective Objects

RGS：通过学习几何连续性实现反射感知的高斯点云渲染

Du, Xiaobiao, Wang, Yida, Bi, Cheng, Zhan, Kun, Yu, Xin

Abstract

Gaussian Splatting has significantly improved the quality of novel view synthesis with explicit Gaussian representation. However, we observed that existing 3D Gaussian Splatting methods (3DGS) often suffer from surface collapse issues on reflective regions, and thus produce inferior geometry and low-quality specular. In this work, we propose a physically-based deferred rendering framework, named Reflection-aware Gaussian Splatting (RGS), that can accurately model specular regions and improve novel view synthesis performance. Specifically, we found that a powerful 3D foundation model can provide a strong 3D geometric prior to foster correct geometric modeling. Based on this, we propose a cross-view shape consistency regularization to regularize the geometry surface with the large model prior and cross-view constraints. In this manner, our RGS can produce smoother geometric surfaces on reflective regions while reducing geometric hollows. To further improve rendering results on reflective regions, we present a reflection-aware densification strategy that is designed to capture specular variations across various views. With this strategy, our RGS is able to render novel views of objects in higher quality. Extensive experiments demonstrate our method consistently renders high-quality reflective objects, achieving state-of-the-art performance.

Chinese Translation

高斯点云渲染显著提高了基于显式高斯表示的新视图合成质量。然而，我们观察到现有的三维高斯点云渲染方法（3DGS）在反射区域常常遭遇表面塌陷问题，从而导致几何形状劣化和低质量的高光效果。在本研究中，我们提出了一种基于物理的延迟渲染框架，称为反射感知高斯点云渲染（RGS），该框架能够准确建模高光区域并提高新视图合成性能。具体而言，我们发现强大的三维基础模型可以提供强有力的三维几何先验，以促进正确的几何建模。在此基础上，我们提出了一种跨视图形状一致性正则化方法，以利用大型模型先验和跨视图约束来规范几何表面。通过这种方式，我们的RGS能够在反射区域生成更平滑的几何表面，同时减少几何空洞。为了进一步改善反射区域的渲染效果，我们提出了一种反射感知的密集化策略，旨在捕捉不同视角下的高光变化。借助这一策略，我们的RGS能够以更高的质量渲染物体的新视图。大量实验表明，我们的方法始终能够渲染高质量的反射物体，达到最先进的性能。

View on arXiv Download PDF AI Translation

cs.CV / 9 / 2609.19444

Seeing Abnormal from Normal: Glomerular Abnormality in Representations of Normal Renal Morphology

从正常中识别异常：正常肾脏形态表现中的肾小球异常

Hasko, Greta, Saluja, Rachit, Shi, Tianyu, Zhao, Leiyue, Yang, Yuechen, Reisenbuechler, Daniel, Yao, Tianyuan, Guo, Zhenhao, Cannon, John, Yang, Haichun, Huo, Yuankai, Chi, Yuling, Gudas, Lorraine, Sabuncu, Mert R., Yang, Yihe, Deng, Ruining

Abstract

Fine-grained evaluation of glomerular pathology must distinguish normal glomeruli from abnormalities such as global and segmental glomerulosclerosis, obsolescent, ischemic, solidified, disappearing, and atubular glomeruli. Supervised classification requires labeled examples of every category, which is impractical when subtypes are rare or absent from the training cohort. One-class anomaly detection offers an alternative by modeling normal data and scoring deviations, allowing previously unseen abnormalities to be detected. We use the frozen residual U-Net backbone of Omni-Seg, pretrained to segment structurally normal renal primitives without abnormal-subtype labels. We propose NoRDeC (Normal-Reference Detection and Characterization), a framework combining Mahalanobis normal-reference scoring with layer-wise representation analysis to determine whether and where glomerular pathology is encoded, how spatial aggregation affects detection, and whether abnormalities alter inter-layer relationships differently. Using glomerular images from two institutions, we evaluate backbone layers and aggregation strategies, compare NoRDeC with PaDiM and PatchCore, and analyze representations using centered kernel alignment (CKA). Layer 4 with Center-70 aggregation achieved a pooled AUROC of $0.926\pm0.013$. NoRDeC achieved the highest AUROC in six of seven abnormality categories and in the pooled analysis, while CKA suggested subtype-dependent changes in inter-layer relationships not captured by anomaly scores alone. The normal-reference model is fitted using only normal glomeruli; abnormality labels are used for configuration selection, evaluation, and grouping in the representation analysis. These results show that a frozen renal feature extractor can support both detection and representation-level characterization of glomerular abnormalities without using abnormal examples to fit the detector.

Chinese Translation

对肾小球病理的细致评估必须区分正常肾小球与诸如全局性和局部性肾小管硬化、衰退性、缺血性、固化、消失和无管肾小球等异常。监督分类需要每个类别的标记样本，但当亚型在训练队列中稀少或缺失时，这种方法是不切实际的。一类异常检测提供了一种替代方案，通过建模正常数据并对偏差进行评分，从而允许检测到以前未见的异常。我们使用Omni-Seg的冻结残差U-Net骨干网络，该网络经过预训练以分割结构正常的肾脏原件，而不需要异常亚型标签。我们提出了NoRDeC（正常参考检测与特征化）框架，该框架结合了Mahalanobis正常参考评分与逐层表示分析，以确定肾小球病理是否以及在何处被编码，空间聚合如何影响检测，以及异常是否以不同方式改变层间关系。使用来自两个机构的肾小球图像，我们评估了骨干层和聚合策略，将NoRDeC与PaDiM和PatchCore进行比较，并使用中心核对齐（CKA）分析表示。层4与Center-70聚合的组合达到了$0.926 ext{±}0.013$的汇总AUROC。在七个异常类别中，NoRDeC在六个类别中获得了最高的AUROC，并在汇总分析中表现优异，而CKA则表明层间关系的亚型依赖性变化并未仅通过异常评分捕捉到。正常参考模型仅使用正常肾小球进行拟合；异常标签用于配置选择、评估和表示分析中的分组。这些结果表明，冻结的肾脏特征提取器可以支持肾小球异常的检测和表示级特征化，而无需使用异常样本来拟合检测器。

View on arXiv Download PDF AI Translation

cs.CV / 10 / 2609.19451

Efficient Unified Multimodal Understanding (EUMU): Winning Solution for the MUMU Track at the 8th LSVOS Challenge

高效统一多模态理解（EUMU）：第八届LSVOS挑战赛MUMU赛道的获胜方案

Kil, Dayoung, Kim, Seong-heum

Abstract

The Mobile Unified Multimodal Understanding (MUMU) Challenge requires a single efficient model to jointly perform multi-concept image tagging, open-vocabulary object detection, and image captioning. We present Efficient Unified Multimodal Understanding (EUMU), the winning solution for the MUMU Track of the 8th LSVOS Challenge. EUMU builds on a shared pretrained multimodal model, using its prompt-based capabilities for detection and captioning and training lightweight heads on shared visual features to predict quality, scene, and event tags. Rather than treating the three tasks independently, EUMU applies task-aware inference refinement by reusing task outputs as cross-task cues. For detection, caption cues help recover objects missed by the initial detection. For captioning, detection cues help refine the caption to better reflect the detected objects. For tagging, image statistics refine quality predictions, while caption and detection cues refine scene and event predictions. This design unifies all three tasks within a single model while satisfying the challenge's resource constraints. EUMU contains 239.169M parameters, requires 23.947 GFLOPs, uses 4.5 GB of peak inference memory, and achieves a final challenge score of 17.3409. Code and models are available at https://github.com/Dayoung-Kil/EUMU.

Chinese Translation

移动统一多模态理解（MUMU）挑战赛要求一个高效的单一模型共同执行多概念图像标记、开放词汇对象检测和图像描述。我们提出了高效统一多模态理解（EUMU），这是第八届LSVOS挑战赛MUMU赛道的获胜方案。EUMU基于一个共享的预训练多模态模型，利用其基于提示的能力进行检测和描述，并在共享视觉特征上训练轻量级头部以预测质量、场景和事件标签。EUMU并不将这三项任务独立对待，而是通过重用任务输出作为跨任务线索，应用任务感知的推理优化。对于检测，描述线索有助于恢复初始检测遗漏的对象；对于描述，检测线索有助于优化描述，使其更好地反映检测到的对象；对于标记，图像统计数据优化质量预测，而描述和检测线索则优化场景和事件预测。该设计在一个单一模型中统一了这三项任务，同时满足挑战的资源限制。EUMU包含239.169M参数，要求23.947 GFLOPs，使用4.5 GB的峰值推理内存，并获得最终挑战得分17.3409。代码和模型可在https://github.com/Dayoung-Kil/EUMU获取。

View on arXiv Download PDF AI Translation

cs.CV / 11 / 2609.19463

ParticleSplat: Self-supervised Object-centric Latent Particle Splatting

ParticleSplat：自监督对象中心的潜在粒子喷溅

He, Lyuxing, Guo, Daniel, Terveen, Elizabeth, Pathak, Deepak, Held, David, Daniel, Tal

Abstract

We present ParticleSplat, a self-supervised object-centric learning method that decomposes scenes into a set of latent ''particles'' representing semantic entities through feedforward 3D Gaussian Splatting. Building on the Deep Latent Particles (DLP) framework, which represents images as a set of particles with attributes such as position, scale, and visual appearance, we address a key limitation of DLP: its inherently 2D nature, which prevents explicit 3D spatial and geometric reasoning that are critical for downstream tasks such as robotic manipulation. Leveraging the structural similarity between latent particles and 3D Gaussian primitives, we introduce a 3D latent particle space trained with a novel view synthesis objective. Our model jointly encodes multiple views with camera poses into a shared 3D object-centric latent space, then transforms particles into particle-aligned 3D Gaussians whose composition reconstructs the full scene. On simulated and real-world datasets, we show that this formulation inherently learns object masks without supervision and supports controllable 3D scene editing, such as moving objects by modifying particles in the latent space. We further establish that the learned 3D representation improves downstream performance on robotic manipulation tasks.

Chinese Translation

我们提出了ParticleSplat，一种自监督的对象中心学习方法，通过前馈3D高斯喷溅将场景分解为一组表示语义实体的潜在“粒子”。基于深度潜在粒子（Deep Latent Particles, DLP）框架，该框架将图像表示为一组具有位置、尺度和视觉外观等属性的粒子，我们解决了DLP的一个关键限制：其固有的2D特性，阻碍了对下游任务（如机器人操作）至关重要的显式3D空间和几何推理。利用潜在粒子与3D高斯原语之间的结构相似性，我们引入了一种使用新颖视图合成目标训练的3D潜在粒子空间。我们的模型将多个视图与相机姿态共同编码到一个共享的3D对象中心潜在空间中，然后将粒子转换为粒子对齐的3D高斯，其组合重建完整场景。在模拟和真实世界的数据集上，我们展示了该公式在没有监督的情况下固有地学习对象掩膜，并支持可控的3D场景编辑，例如通过修改潜在空间中的粒子来移动对象。我们进一步证明，学习到的3D表示提高了机器人操作任务的下游性能。

View on arXiv Download PDF AI Translation

cs.CV / 12 / 2609.19483

SCOUT: Sim-to-Real Text-Based Person Retrieval by Embedding-Space Prediction over Frozen Video Features

SCOUT：通过对冻结视频特征进行嵌入空间预测实现的基于文本的人物检索

Traoré, Abdarahmane, Couturier, Andy, Hervet, Éric

Abstract

Text-based person retrieval under a sim-to-real gap (synthetic training data, a real-image gallery) is usually tackled with costly fine-tuned cross-encoders. We ask whether a frozen-encoder system can compete. We present SCOUT, which casts cross-modal retrieval as prediction in embedding space. A trainable predictor maps the patch tokens of a frozen video encoder into the embedding space of a frozen text encoder under a bidirectional InfoNCE objective, and no encoder is fine-tuned in the base model. The video encoder is V-JEPA, the text encoder is EmbeddingGemma, and the predictor is initialized from a Qwen3.5-0.8B decoder. We make three findings. First, the best frozen text encoder is simply the one whose geometry best matches the video features. A training-free alignment score ranks three candidate text encoders in the same order as their retrieval accuracy on our held-out split (Spearman $\rho = 1.0$); a fourth, LLM-based encoder shows the rule is metric-dependent, holding for a neighborhood-overlap score ($\rho = 0.8$) but not for a linear probe ($\rho = -0.2$). Second, two precision-targeted levers, parameter-efficient ExPLoRA adaptation of the video encoder and a training-free attribute-decomposed reranker built on a vision-language model, improve the top-rank precision that otherwise limits the frozen system, adding 2.2 points of leaderboard R@1. Third, a local-versus-public calibration study explains which interventions transfer to the real domain. On AI City Challenge 2026 Track 4 the full retrieve-fuse-rerank system reaches 84.25 mAP@10 on the final leaderboard, while a single frozen model submitted alone reaches 60.63. Our trained components cost about 95 GPU-hours. CMP, the dataset authors' fine-tuned cross-encoder that trains for sixteen GPU-days, is one fusion member of the full system, not an alternative. Code and annotations: https://github.com/abtraore/SCOUT-ECCV

Chinese Translation

在模拟到现实的差距下（合成训练数据与真实图像库），基于文本的人物检索通常需要昂贵的微调交叉编码器。我们探讨了冻结编码器系统是否能够与之竞争。我们提出了SCOUT，将跨模态检索视为嵌入空间中的预测。一个可训练的预测器在双向InfoNCE目标下，将冻结视频编码器的补丁标记映射到冻结文本编码器的嵌入空间，并且在基础模型中没有对任何编码器进行微调。视频编码器为V-JEPA，文本编码器为EmbeddingGemma，预测器从Qwen3.5-0.8B解码器初始化。我们得出了三个发现。首先，最佳的冻结文本编码器恰好是几何形状与视频特征最佳匹配的编码器。一个无训练的对齐得分按与我们保留分割上的检索准确性相同的顺序对三个候选文本编码器进行排名（Spearman $ ho = 1.0$）；第四个基于LLM的编码器表明该规则依赖于度量，在邻域重叠得分上成立（$ ho = 0.8$），但在线性探测上不成立（$ ho = -0.2$）。其次，两个针对精度的杠杆，视频编码器的参数高效ExPLoRA适应和基于视觉-语言模型构建的无训练属性分解重排序器，改善了限制冻结系统的顶级排名精度，增加了2.2分的排行榜R@1。第三，局部与公共校准研究解释了哪些干预措施能够转移到真实领域。在AI City Challenge 2026 Track 4中，完整的检索-融合-重排序系统在最终排行榜上达到了84.25 mAP@10，而单一冻结模型单独提交的结果为60.63。我们训练的组件大约消耗了95个GPU小时。CMP，数据集作者的微调交叉编码器，训练了十六个GPU天，是完整系统的一个融合成员，而不是替代方案。代码和注释：https://github.com/abtraore/SCOUT-ECCV

View on arXiv Download PDF AI Translation

cs.CV / 13 / 2609.19518

AMB3R-SLAM: Kilometer-scale SLAM with Hierarchical Backend

AMB3R-SLAM：具有分层后端的公里级SLAM

Wang, Hengyi, Agapito, Lourdes

Abstract

We present AMB3R-SLAM, a real-time monocular SLAM system capable of reconstructing kilometer-scale trajectories over 10k frames on a single consumer-grade GPU. Our model couples a lightweight front-end for low-latency online tracking with a hierarchical backend that progressively enforces local, mid-level, and global consistency. By avoiding bundle adjustment that relies on the static world assumption, our system naturally handles complex dynamic scenes out of the box. Furthermore, we demonstrate that our method can be extended to leverage stereo, RGB-D, and LiDAR as additional inputs. AMB3R-SLAM achieves strong camera tracking performance across 9 datasets, reducing the absolute trajectory error (ATE) of previous state-of-the-art methods on VBR and Oxford Spires by over 70%. With additional LiDAR input, our model further reduces ATE to sub-meter level on KITTI and VBR datasets.

Chinese Translation

我们提出了AMB3R-SLAM，一种实时单目SLAM系统，能够在单个消费级GPU上重建超过10,000帧的公里级轨迹。我们的模型结合了一个轻量级前端，用于低延迟在线跟踪，以及一个分层后端，逐步强制执行局部、中级和全局一致性。通过避免依赖静态世界假设的束调整，我们的系统自然地处理复杂的动态场景。此外，我们证明了我们的方法可以扩展以利用立体视觉、RGB-D和激光雷达作为额外输入。AMB3R-SLAM在9个数据集上实现了强大的相机跟踪性能，将之前最先进方法在VBR和Oxford Spires上的绝对轨迹误差（ATE）降低了超过70%。在额外的激光雷达输入下，我们的模型进一步将ATE降低到KITTI和VBR数据集的亚米级。

View on arXiv Download PDF AI Translation

cs.CV / 14 / 2609.19542

PerSeM: Persistent Semantic Memory for Long-Horizon Open-Vocabulary UAV Mapping

PerSeM：用于长时间跨度开放词汇无人机映射的持久语义记忆

Jamwal, Saurbh Singh, Ramakrishnan, Ganesh

Abstract

Open-vocabulary segmentation enables rich semantic perception for UAVs, but frame-wise predictions can remain temporally inconsistent across repeated observations and changing viewpoints. We present PerSeM, a training-free persistent semantic memory framework for long-horizon open-vocabulary UAV mapping. PerSeM associates frame-wise semantic observations with persistent world-space voxels and constructs a majority-based semantic memory, which is conservatively refined through history-preserving spatial refinement, trust-aware replay, and context-guided verification. Experiments on the Forest and UAVScenes benchmarks show that persistent 3D memory provides substantial gains in semantic correctness and temporal stability over frame-wise predictions. Beyond this strong persistent-memory baseline, PerSeM provides consistent additional improvements, improving both semantic accuracy and temporal stability across all five evaluated UAVScenes sequences. Analysis using regions identified independently of the final PerSeM predictions further shows that these gains are concentrated in semantically difficult and temporally unstable regions, where majority-based memory is most likely to remain uncertain. These results demonstrate that persistent 3D aggregation provides a strong foundation for long-horizon semantic mapping, while conservative refinement of uncertain memory states can provide additional improvements without retraining or additional neural-network inference.

Chinese Translation

开放词汇分割为无人机提供了丰富的语义感知，但逐帧预测在重复观察和变化视角下可能会保持时间上的不一致性。我们提出了PerSeM，一个无训练的持久语义记忆框架，用于长时间跨度的开放词汇无人机映射。PerSeM将逐帧的语义观察与持久的世界空间体素关联，并构建基于多数的语义记忆，该记忆通过历史保留的空间细化、基于信任的重放和上下文引导的验证进行保守的优化。在Forest和UAVScenes基准上的实验表明，持久的3D记忆在语义正确性和时间稳定性方面相较于逐帧预测提供了显著的提升。超越这一强大的持久记忆基线，PerSeM在所有五个评估的UAVScenes序列中提供了一致的额外改进，提高了语义准确性和时间稳定性。使用与最终PerSeM预测独立识别的区域的分析进一步表明，这些提升集中在语义上困难和时间上不稳定的区域，在这些区域中，基于多数的记忆最有可能保持不确定。这些结果表明，持久的3D聚合为长时间跨度的语义映射提供了坚实的基础，而对不确定记忆状态的保守细化可以在不重新训练或额外神经网络推理的情况下提供额外的改进。

View on arXiv Download PDF AI Translation

cs.CV / 15 / 2609.19555

A Multi-Modal Generative Model for Tomato Disease Leaves Understanding

一种多模态生成模型用于番茄病害叶片理解

Quoc, Khang Nguyen, Tran, Minh-Phuoc, Truong, Gia-Han, Quach, Luyl-Da

Abstract

Artificial intelligence for plant disease analysis has advanced from task-specific classifiers to multi-modal models capable of jointly interpreting visual and textual information. However, practical deployment in precision agriculture remains limited because most existing approaches treat disease understanding as isolated prediction tasks, failing to capture the complementary relationships among symptom recognition, severity assessment, and question-driven diagnostic reasoning. In tomato pathology, accurate interpretation of diseased leaves requires more than label prediction; it demands integrating visual symptoms with semantic context to support a comprehensive and explainable understanding. Here, we present SOLAR, a multimodal generative model that understands tomato disease spanning six question-answering tasks. SOLAR learns to align visual features with task-aware language representations by Fusion Expert module based on mixture-of-expert, enabling it to generate contextually relevant answers across diverse diagnostic tasks. By formulating tomato disease analysis as a generative Visual Question Answering (VQA) task, SOLAR provides a flexible framework that supports multi-task inference within a single model while improving performance and cross-task knowledge sharing. We evaluate SOLAR on $41,677$ images, including $216,209$ Question-Answering (QA) pairs to understand tomato leaf disease under both closed and open-ended QA settings. Experimental results show that SOLAR consistently outperforms state-of-the-art vision-only, vision-language, and task-specific models across all tasks, demonstrating superior accuracy, robustness, and multimodal reasoning. These findings highlight the potential of generative multimodal modeling as an effective direction for understanding of plant disease. The code for this study is available at https://github.com/EnalisUs/SOLAR.

Chinese Translation

植物病害分析中的人工智能已从特定任务的分类器发展到能够联合解释视觉和文本信息的多模态模型。然而，由于大多数现有方法将病害理解视为孤立的预测任务，未能捕捉症状识别、严重性评估和基于问题的诊断推理之间的互补关系，因此在精准农业中的实际应用仍然有限。在番茄病理学中，准确解读病变叶片不仅仅需要标签预测；它需要将视觉症状与语义上下文相结合，以支持全面且可解释的理解。在此，我们提出了SOLAR，这是一种多模态生成模型，能够理解涉及六个问答任务的番茄病害。SOLAR通过基于混合专家的Fusion Expert模块学习将视觉特征与任务感知的语言表示对齐，使其能够在多样的诊断任务中生成上下文相关的答案。通过将番茄病害分析构建为生成的视觉问答（Visual Question Answering, VQA）任务，SOLAR提供了一个灵活的框架，支持在单一模型内进行多任务推理，同时提高性能和跨任务知识共享。我们在$41,677$张图像上评估了SOLAR，包括$216,209$个问答对，以理解番茄叶片病害在封闭式和开放式问答设置下的表现。实验结果表明，SOLAR在所有任务中始终优于最先进的仅视觉、视觉-语言和特定任务模型，展现出卓越的准确性、鲁棒性和多模态推理能力。这些发现突显了生成多模态建模作为理解植物病害的有效方向的潜力。本研究的代码可在https://github.com/EnalisUs/SOLAR获取。

View on arXiv Download PDF AI Translation

cs.CV / 16 / 2609.19592

Selective Cotton Boll Localization for Robotic Harvesting: Evaluation of Deep Learning Vision Models Under Field Conditions

选择性棉铃定位用于机器人采摘：在田间条件下评估深度学习视觉模型

Thayananthan, Thevathayarajh, Zhang, Xin, Badu, Isuru Laddusinghe, Harjono, Jonathan, Rains, Glen C., Li, Beiwen, Bastos, Leonardo M., Wijewardane, Nuwan K., Martins, Vitor S.

Abstract

This study developed and evaluated a deep-learning-based perception framework for selective robotic cotton picking. The dataset contained 1,008 annotated field images collected using three cameras under varying natural lighting and weather conditions. Object-detection models from the YOLOv8 through YOLOv13 families were evaluated using their default configurations, while segmentation performance was assessed using YOLOv8-seg, YOLOv11-seg, YOLOv12-seg, the Segment Anything Model (SAM), SAMv2.1, FastSAM, and Grounded-SAM with the Recognize Anything Model (RAM). Among the detection models, GELAN-s achieved the most favorable balance between mean average precision (mAP) and inference speed, obtaining an mAP of 86.1%, precision of 81.6%, recall of 76.6%, and an F1-score of 79.0%, with an average inference time of 42.3 ms per image. Among the direct segmentation models, YOLOv12-m-seg provided the most favorable balance between [email protected] and FPS, achieving a segmentation [email protected] of 83.7% with an inference time of 20.4 ms per image. In the detection-prompted segmentation approach, bounding-box prompts generated by GELAN-s improved the localization of cotton bolls for SAM and SAMv2.1, while SAMv2.1 Tiny consistently outperformed FastSAM and Grounded-SAM with RAM. In the area-based evaluation against manually annotated segmentation masks, YOLOv12-m-seg achieved an $R^2$ value of 0.966, compared with 0.860 for GELAN-s + SAMv2.1 Tiny. Field experiments conducted using a UR5e robotic manipulator, a custom end-effector, and a ZED2i stereo camera further validated the effectiveness of the YOLOv12-m-seg model for real-time cotton boll detection, segmentation, and selective picking under varying confidence levels. These results demonstrate that YOLOv12-m-seg provides an efficient perception model for robotic cotton harvesting and has strong potential for field deployment.

Chinese Translation

本研究开发并评估了一种基于深度学习的感知框架，用于选择性机器人棉花采摘。数据集包含1,008幅在不同自然光照和天气条件下使用三台相机收集的标注田间图像。评估了YOLOv8至YOLOv13系列的目标检测模型，使用其默认配置，同时使用YOLOv8-seg、YOLOv11-seg、YOLOv12-seg、Segment Anything Model (SAM)、SAMv2.1、FastSAM和Grounded-SAM与Recognize Anything Model (RAM)评估分割性能。在检测模型中，GELAN-s在平均精度均值（mAP）和推理速度之间取得了最有利的平衡，获得了86.1%的mAP、81.6%的精度、76.6%的召回率和79.0%的F1-score，平均每幅图像的推理时间为42.3毫秒。在直接分割模型中，YOLOv12-m-seg在[email protected]和FPS之间提供了最有利的平衡，获得了83.7%的分割[email protected]，推理时间为每幅图像20.4毫秒。在基于检测的分割方法中，由GELAN-s生成的边界框提示改善了SAM和SAMv2.1的棉铃定位，而SAMv2.1 Tiny始终优于FastSAM和Grounded-SAM与RAM的组合。在与手动标注的分割掩膜进行的基于区域的评估中，YOLOv12-m-seg的$R^2$值为0.966，而GELAN-s + SAMv2.1 Tiny为0.860。使用UR5e机器人 manipulator、自定义末端执行器和ZED2i立体相机进行的田间实验进一步验证了YOLOv12-m-seg模型在不同置信水平下进行实时棉铃检测、分割和选择性采摘的有效性。这些结果表明，YOLOv12-m-seg为机器人棉花采摘提供了高效的感知模型，并在田间部署中具有强大的潜力。

View on arXiv Download PDF AI Translation

cs.CV / 17 / 2609.19628

VGGT-GS SLAM: Uncalibrated Monocular Gaussian Splatting SLAM with Feed-Forward Priors

VGGT-GS SLAM：基于前馈先验的未校准单目高斯点云SLAM

Han, Yuhang, Wang, Hao, Cao, Jiaxi, Liu, Xingyu

Abstract

We present VGGT-GS SLAM, a monocular 3D Gaussian Splatting SLAM system designed for uncalibrated videos. Starting from feed-forward VGGT pose and depth priors, our system performs submap differentiable bundle adjustment that jointly refines camera poses and a 3D Gaussian map, while optimizing submap-shared intrinsics and radial--tangential distortion through analytic calibration Jacobians. To improve global consistency, we introduce Gaussian-native alignment (GNA) for camera-anchored scale refinement between sequential submaps and verification of loop-closure candidates. Extensive experiments on standard indoor benchmarks show consistent improvements in localization accuracy and strong rendering quality under uncalibrated settings, establishing a strong baseline for uncalibrated Gaussian SLAM.

Chinese Translation

我们提出了VGGT-GS SLAM，这是一种针对未校准视频设计的单目3D高斯点云SLAM系统。该系统基于前馈的VGGT姿态和深度先验，执行子图可微束调整（submap differentiable bundle adjustment），共同优化相机姿态和3D高斯地图，同时通过解析校准雅可比（analytic calibration Jacobians）优化共享的内参和径向-切向畸变。为了提高全局一致性，我们引入了高斯原生对齐（Gaussian-native alignment, GNA），用于在连续子图之间进行相机锚定的尺度优化和环闭合候选的验证。在标准室内基准测试上的大量实验表明，在未校准设置下，定位精度和渲染质量均有显著提升，为未校准高斯SLAM建立了一个强有力的基准。

View on arXiv Download PDF AI Translation

cs.CV / 18 / 2609.19631

Instance Segmentation and Fine-grained Classification for Urban Buildings with Adaptive Region Dividing and Spatially-Supervised Contrastive Learning

基于自适应区域划分和空间监督对比学习的城市建筑实例分割与细粒度分类

Zhang, Weiyuan, Zhang, Qi, Huang, Hui

Abstract

Accurate instance-level and functional understanding of urban buildings in large-scale point clouds is essential for digital city modeling and urban analysis. However, the extensive spatial coverage of urban scenes leads most existing methods to rely on predefined blocks for training and evaluation, although such partitions are rarely available in real-world applications and introduce additional preprocessing while fragmenting complete building structures. To address this issue, we propose an adaptive region-dividing strategy with unified scene-level evaluation. Specifically, the 3D point cloud is projected onto a bird's-eye-view (BEV) plane, where a pretrained segmentation model is used to detect building regions. The detected bounding boxes are then back-projected to the original point cloud to construct structure-aligned adaptive training blocks, enabling semantically guided dynamic partitioning without manual design. Furthermore, beyond instance-level understanding, few methods have explored fine-grained classification for urban buildings, and thus we also put forward a fine-grained classification model for urban buildings with a spatially-supervised contrastive loss. First, for each segmented building, a point transformer classifier jointly encodes its body and local context using geometric, color, and core-context information. Then, the class-balanced weighted cross-entropy is used to alleviate severe class imbalance. The proposed spatially-supervised contrastive loss further enhances inter-class discriminability by assigning greater weight to spatially proximate, same-category buildings, encouraging compact functional representations while separating easily confused categories. Extensive experiments on UrbanBIS and STPLS3D demonstrate the advantages of the proposed method in building instance segmentation and fine-grained classification compared to existing SOTA methods.

Chinese Translation

在大规模点云中准确理解城市建筑的实例级和功能性对于数字城市建模和城市分析至关重要。然而，城市场景的广泛空间覆盖使得大多数现有方法依赖于预定义的块进行训练和评估，尽管这种划分在实际应用中很少可用，并且在分割完整建筑结构的同时引入了额外的预处理。为了解决这个问题，我们提出了一种自适应区域划分策略，并采用统一的场景级评估。具体而言，3D点云被投影到鸟瞰图（BEV）平面上，利用预训练的分割模型检测建筑区域。检测到的边界框随后被反投影回原始点云，以构建结构对齐的自适应训练块，实现语义引导的动态划分，无需手动设计。此外，除了实例级理解，少数方法探讨了城市建筑的细粒度分类，因此我们还提出了一种基于空间监督对比损失的城市建筑细粒度分类模型。首先，对于每个分割的建筑，点变换器分类器利用几何、颜色和核心上下文信息共同编码其主体和局部上下文。然后，使用类别平衡加权交叉熵来缓解严重的类别不平衡。所提出的空间监督对比损失通过对空间上接近的同类建筑赋予更大权重，进一步增强了类别间的可区分性，鼓励紧凑的功能表示，同时区分容易混淆的类别。在UrbanBIS和STPLS3D上的大量实验表明，与现有的最先进方法相比，所提出的方法在建筑实例分割和细粒度分类方面具有明显优势。

View on arXiv Download PDF AI Translation

cs.CV / 19 / 2609.19634

Scientific Image Quality Assessment via Multi-modal Retrieval-Augmented Generation

通过多模态检索增强生成的科学图像质量评估

Zhang, Yinuo, Liu, Bingshuo, Tu, Zhiying, Chu, Dianhui, Liu, Qingbin, Chen, Xi, Bian, Jiang, Yu, Xiaoyan, Sui, Dianbo

Abstract

This paper proposes a Retrieval-Augmented Generation (RAG) framework for scientific image quality assessment, designed to simultaneously address both the understanding track (SIQA-U) and the scoring track (SIQA-S) of the SIQA challenge. We construct a multimodal index that integrates textual semantics with fine-grained visual features, and develop a multi-route retrieval and fusion mechanism to provide large language models with highly relevant reference cases, thereby enhancing their capability to evaluate complex scientific images. Experimental results demonstrate that the proposed framework effectively aligns with the judgment criteria of human experts. Ultimately, our method achieves 1st place in the SIQA-U track of the SIQA challenge at the ICME 2026 Grand Challenges.

Chinese Translation

本文提出了一种用于科学图像质量评估的检索增强生成（RAG）框架，旨在同时解决SIQA挑战中的理解轨道（SIQA-U）和评分轨道（SIQA-S）。我们构建了一个多模态索引，将文本语义与细粒度视觉特征相结合，并开发了一种多路径检索和融合机制，为大型语言模型提供高度相关的参考案例，从而增强其评估复杂科学图像的能力。实验结果表明，所提出的框架有效地与人类专家的判断标准保持一致。最终，我们的方法在ICME 2026大挑战的SIQA挑战的SIQA-U轨道中获得了第一名。

View on arXiv Download PDF AI Translation

cs.CV / 20 / 2609.19662

Towards Active Cross-View Object Geo-Localization

面向主动跨视角物体地理定位

Yao, Shunyu, Zhang, Xiaohan, Yang, Zhuoran, Lai, Haoqi, Ming, Qi, Hu, Xiaoxi, Shen, Hui-Liang, Cao, Si-Yuan

Abstract

Cross-view object geo-localization (CVOGL) typically assumes a fixed query image, overlooking the ability of mobile agents to actively acquire more informative observations. To address this limitation, we introduce Active Cross-View Object Geo-Localization (ActiveGeo), where an agent sequentially selects new viewpoints and determines when to stop, aiming to improve localization with minimal observations. We further propose ActiveMoPT, an ActiveGeo framework with three-stage training. First, Multi-View Prompt-Preserving Adaptation enables the model to aggregate multiple query views while reusing the initial prompt. Second, Trajectory-Guided Policy Initialization uses supervised agent trajectories to learn viewpoint selection and initial stopping behavior. Third, Cost-Aware Policy Refinement employs GRPO with a gain-cost reward to jointly optimize localization accuracy and observation efficiency. We also construct ActiveGeo-858, a zero-shot test set containing 858 scenes and 1,716 target annotations. Experiments show that ActiveMoPT achieves state-of-the-art performance on MoP-UAV using only 1.45 query views on average, and substantially outperforms previous CVOGL approaches under zero-shot evaluation on ActiveGeo-858.

Chinese Translation

跨视角物体地理定位（CVOGL）通常假设查询图像是固定的，忽视了移动代理主动获取更具信息性的观测的能力。为了解决这一局限性，我们提出了主动跨视角物体地理定位（ActiveGeo），在该方法中，代理顺序选择新的视点并决定何时停止，旨在以最少的观测提高定位精度。我们进一步提出了ActiveMoPT，这是一个具有三阶段训练的ActiveGeo框架。首先，多视角提示保持适应（Multi-View Prompt-Preserving Adaptation）使模型能够聚合多个查询视图，同时重用初始提示。其次，轨迹引导策略初始化（Trajectory-Guided Policy Initialization）利用监督的代理轨迹学习视点选择和初始停止行为。第三，成本感知策略优化（Cost-Aware Policy Refinement）采用增益-成本奖励的GRPO方法，联合优化定位精度和观测效率。我们还构建了ActiveGeo-858，这是一个包含858个场景和1,716个目标注释的零样本测试集。实验表明，ActiveMoPT在MoP-UAV上仅使用平均1.45个查询视图便实现了最先进的性能，并在ActiveGeo-858的零样本评估中显著优于之前的CVOGL方法。

View on arXiv Download PDF AI Translation

cs.CV / 21 / 2609.19664

VideoResearcher: Self-Improving Tool Design for Long-Video Understanding

视频研究者：用于长视频理解的自我改进工具设计

Ye, Dingqiang, Zhao, Dongdi, Wang, Kaishen, Hu, Qingqiao, Sun, Jingchen, Liang, Yijun, Jia, Yuqi, Huang, Yiqiao, Tian, Yunjie, Zhang, Jiaxing, Jin, Chuanyang, Zhang, Ke, Patel, Vishal M., Fu, Di

Abstract

Video agents have made substantial progress in long-video understanding. Yet effective video-agent systems require costly, time-consuming manual design and trial and error. Current self-improvement methods either refine low-impact prompts, recombine predefined micro-tools, or struggle with convergence in harness optimization. To bridge this gap, we target high-impact video-tool with VideoResearcher, a training-free multi-agent framework that autonomously designs, tests, and refines tools for video understanding, like a human researcher. VideoResearcher operates through dual Solving and Evolving loops: it analyzes tool-use trajectories to identify capability gaps, coordinates specialized agents to develop and validate executable tools, and reuses evolved tools to strengthen evidence acquisition in subsequent video reasoning. Through iterative tool refinement and validation, it progressively strengthens evidence acquisition without updating model parameters. VideoResearcher achieves state-of-the-art performance among self-improving agents and approaches the human-designed upper bound, demonstrating a training-free paradigm for long-video understanding that expands agent capabilities through autonomous tool development while reducing costly manual engineering.

Chinese Translation

视频智能体在长视频理解方面取得了显著进展。然而，有效的视频智能体系统需要耗时且成本高昂的手动设计和反复试验。目前的自我改进方法要么优化低影响的提示，要么重新组合预定义的微工具，或者在利用优化时面临收敛困难。为了解决这一问题，我们提出了VideoResearcher，旨在开发高影响力的视频工具，这是一种无训练的多智能体框架，能够像人类研究者一样自主设计、测试和改进视频理解工具。VideoResearcher通过双重解决和演变循环进行操作：它分析工具使用轨迹以识别能力差距，协调专业智能体开发和验证可执行工具，并重用演变工具以增强后续视频推理中的证据获取。通过迭代的工具改进和验证，它在不更新模型参数的情况下逐步增强证据获取能力。VideoResearcher在自我改进智能体中实现了最先进的性能，并接近人类设计的上限，展示了一种无训练的长视频理解范式，通过自主工具开发扩展智能体能力，同时减少了成本高昂的手动工程。

View on arXiv Download PDF AI Translation

cs.CV / 22 / 2609.19669

Beyond Patch Removal: Persistent Adversarial Effects in Vision-Language-Action Policies

超越补丁移除：视觉-语言-行动策略中的持久对抗效应

Wu, Enhao, Guo, Fusen, Cao, Yuxin, Lyu, Ziyang, Li, Lin, Song, Wei

Abstract

Adversarial patches to Vision-Language-Action (VLA) policies can cause both immediate action corruption and persistent state effects that remain after the patch is removed. Existing evaluations largely focus on continuous attacks and do not separate these two effects. We introduce a state-restoration protocol that removes the patch at matched action-chunk boundaries and measures subsequent recoverability under the same remaining step budget. Clean, random-patch, deviation-matched, and fixed-direction controls distinguish adversarial effects from occlusion, action-error magnitude, and directional persistence. We also evaluate a recovery adapter trained on attack-induced states under controlled intervention latency. On OpenVLA-OFT with EDPA attacks, only 36.2% of LIBERO-Long episodes remain recoverable after five chunks, compared with 89.9% and 87.0% for the deviation-matched and fixed-direction controls. Similar persistent effects are observed on autoregressive OpenVLA. The recovery adapter improves recovery from 7.7% to 47.4% at one-chunk latency, but its benefit decreases substantially with delayed intervention. These results show that adversarial effects can persist after patch removal and that timely intervention is critical for recovery.

Chinese Translation

对视觉-语言-行动（VLA）策略的对抗补丁可以导致即时的行动损坏和在补丁移除后仍然存在的持久状态效应。现有评估主要集中在连续攻击上，并未区分这两种效应。我们引入了一种状态恢复协议，在匹配的行动块边界移除补丁，并在相同的剩余步骤预算下测量后续的可恢复性。干净的、随机补丁的、偏差匹配的和固定方向的对照组将对抗效应与遮挡、行动错误幅度和方向持久性区分开来。我们还评估了一种在受控干预延迟下训练的恢复适配器，该适配器针对攻击引起的状态进行优化。在使用EDPA攻击的OpenVLA-OFT上，经过五个块后，只有36.2%的LIBERO-Long剧集仍然可恢复，而偏差匹配和固定方向对照组的可恢复率分别为89.9%和87.0%。在自回归OpenVLA上也观察到了类似的持久效应。恢复适配器在一个块延迟下将恢复率从7.7%提高到47.4%，但随着干预延迟的增加，其效果显著下降。这些结果表明，对抗效应在补丁移除后仍然可以持续，并且及时干预对于恢复至关重要。

View on arXiv Download PDF AI Translation

cs.CV / 23 / 2609.19693

IMFD: End-to-end Multi-Face Forgery Detection through Instruction-based Large Vision-Language Models

IMFD：通过基于指令的大型视觉-语言模型实现端到端多面孔伪造检测

Choi, Dasom, Moon, Sangjun, Im, Hyeongchan, Park, Jaeeon, Kwon, Jingun, Kamigaito, Hidetaka, Watanabe, Taro, Okumura, Manabu

Abstract

The rapid increase of deepfakes has raised significant concerns due to their spread on social media. Traditional multi-face forgery detectors crop and verify each face independently, ignoring background context and inter-face relationships, which often yields suboptimal performance. To overcome these limitations, we leverage instruction-based Large Vision-Language Models (LVLMs), which can interpret entire images and follow complex textual instructions. We propose a simple yet effective single-stage multi-face forgery detector, called IMFD (Instruction-based Multi-face Forgery Detector), which is trained end-to-end to jointly localize faces and predict per-face forgery labels. Rather than treating face box prediction only as a joint objective, IMFD explicitly integrates predicted face bounding boxes into the instruction as visual cues that enhance instruction grounding and forgery detection. To support the training and evaluation of IMFD, we convert existing multi-face forgery datasets into an instruction-based format. Experimental results and analyses show that IMFD improves multi-face forgery detection by integrating face bounding boxes into the instruction, and consistently outperforms various state-of-the-art methods.

Chinese Translation

深度伪造技术的快速增长引发了人们的重大担忧，因为其在社交媒体上的传播。传统的多面孔伪造检测器独立裁剪和验证每个面孔，忽略了背景上下文和面孔之间的关系，这往往导致性能不佳。为了解决这些局限性，我们利用基于指令的大型视觉-语言模型（LVLMs），它们能够解读整个图像并遵循复杂的文本指令。我们提出了一种简单而有效的单阶段多面孔伪造检测器，称为IMFD（基于指令的多面孔伪造检测器），该检测器经过端到端训练，能够共同定位面孔并预测每个面孔的伪造标签。IMFD并不是仅将面孔框预测视为一个联合目标，而是将预测的面孔边界框明确地整合到指令中，作为增强指令基础和伪造检测的视觉线索。为了支持IMFD的训练和评估，我们将现有的多面孔伪造数据集转换为基于指令的格式。实验结果和分析表明，IMFD通过将面孔边界框整合到指令中来改善多面孔伪造检测，并始终优于各种最先进的方法。

View on arXiv Download PDF AI Translation

cs.CV / 24 / 2609.19702

Understanding and Exploiting Diagonal Attention Sparsity in Autoregressive Image Generation

理解与利用自回归图像生成中的对角注意力稀疏性

Kim, Daeun, Hong, Junwha, Oh, Changhun, Kim, Yoonsung, Lee, Yoonhyeong, Park, Jongse

Abstract

Autoregressive image generation has emerged as a paradigm for multimodal AI systems due to its compatibility with transformer-based LLM serving infrastructures. However, generating thousands of visual tokens per request makes decoding increasingly bottlenecked by KV cache accesses during attention computation. Sparse attention is particularly attractive for this workload because many visual generation applications tolerate moderate quality degradation in exchange for improved performance and efficiency. While sparse attention has been extensively explored for text-based LLM inference, it remains unclear whether its sparsity assumptions generalize effectively to autoregressive image generation. We present the first systematic characterization of attention sparsity in autoregressive image generation across diverse workloads and representative open-source models. Our analysis reveals several distinguishing properties, including a pronounced prefill-decode asymmetry, strong attention concentration on prompt and local tokens, and a unique diagonal attention sparsity pattern arising from the spatial locality of visual tokens. Motivated by these observations, we propose a diagonal-aware sparse attention mechanism that selectively skips KV entries along the diagonal attention direction within a recent window. Implemented on top of a GPU-based serving system using FlexGen, FlashAttention-2, and custom kernels, our approach achieves up to 3.1x throughput and 1.19x latency improvements with less than 2% quality degradation compared to dense inference.

Chinese Translation

自回归图像生成已成为多模态人工智能系统的一种范式，因为它与基于变换器的LLM（大规模语言模型）服务基础设施兼容。然而，每次请求生成数千个视觉标记使得在注意力计算过程中，解码越来越受到KV缓存访问的瓶颈限制。稀疏注意力对于这一工作负载尤其具有吸引力，因为许多视觉生成应用在性能和效率提升的交换中可以容忍适度的质量下降。尽管稀疏注意力在基于文本的LLM推理中得到了广泛探索，但其稀疏性假设是否有效地推广到自回归图像生成仍不清楚。我们首次系统性地表征了自回归图像生成中注意力稀疏性在不同工作负载和代表性开源模型中的表现。我们的分析揭示了几个显著特征，包括明显的预填充-解码不对称性、对提示和局部标记的强注意力集中，以及由于视觉标记的空间局部性而产生的独特对角注意力稀疏模式。基于这些观察，我们提出了一种对角感知稀疏注意力机制，该机制在最近的窗口内选择性地跳过沿对角注意力方向的KV条目。在基于GPU的服务系统上实现，使用FlexGen、FlashAttention-2和自定义内核，我们的方法在质量下降不到2%的情况下，实现了高达3.1倍的吞吐量和1.19倍的延迟改进。

View on arXiv Download PDF AI Translation

cs.CV / 25 / 2609.19716

GAPrompt++: Multi-Granular Geometry-Aware Point Cloud Prompt for 3D Vision Model

GAPrompt++：用于3D视觉模型的多粒度几何感知点云提示

Ai, Zixiang, Cui, Zhenyu, Guo, Yufei, Qiang, Wenwen, Chen, Lei, Lu, Jiwen, Zhou, Jiahuan

Abstract

Pre-trained 3D vision models have substantially advanced point cloud analysis, yet adapting them to downstream tasks via full fine-tuning is computationally expensive and storage-intensive. Parameter-Efficient Fine-Tuning (PEFT) offers a promising alternative by reducing both adaptation cost and storage burden. However, existing prompting-based approaches ignore the intrinsic geometric structures of point clouds, thereby limiting their adaptation capability. This limitation stems from their inability to encode both fine-grained geometric cues and coarse-grained structural semantics, as well as failing to propagate such information effectively through the model hierarchy. To address these challenges, we propose GAPrompt++, a multi-granular geometry-aware prompting method that provides richer geometric guidance for efficient 3D task adaptation. Specifically, we introduce a Point Shift Prompter that extracts multi-granular geometric features across different scales, enabling instance-specific geometric adjustments during adaptation. Next, a Keypoint Prompter adaptively generates point-level prompts to highlight local geometric saliency and fine-grained structural details. Furthermore, a Prompt Propagation mechanism injects these multi-granular geometric cues throughout the feature extraction hierarchy, strengthening the ability to capture essential geometric characteristics. Extensive experiments show that GAPrompt++ achieves state-of-the-art performance among prompting-based PEFT methods and even surpasses full fine-tuning across diverse benchmarks, while requiring less than 2\% trainable parameters. In addition, to address the saturation of existing evaluation datasets, we construct two more challenging benchmarks derived from 3D Gaussian Splatting and Multi-View Stereo reconstruction, offering diverse and realistic point cloud scenarios to promote future research.

Chinese Translation

预训练的3D视觉模型在点云分析方面取得了显著进展，但通过完全微调将其适应于下游任务的计算成本高且存储需求大。参数高效微调（PEFT）提供了一种有前景的替代方案，通过降低适应成本和存储负担。然而，现有的基于提示的方法忽视了点云的内在几何结构，从而限制了它们的适应能力。这一限制源于它们无法有效编码细粒度的几何线索和粗粒度的结构语义，并且未能通过模型层次有效传播这些信息。为了解决这些挑战，我们提出了GAPrompt++，一种多粒度几何感知提示方法，为高效的3D任务适应提供更丰富的几何指导。具体而言，我们引入了一种点位移提示器（Point Shift Prompter），该提示器在不同尺度上提取多粒度几何特征，使得在适应过程中能够进行实例特定的几何调整。接下来，关键点提示器（Keypoint Prompter）自适应生成点级提示，以突出局部几何显著性和细粒度结构细节。此外，提示传播机制（Prompt Propagation）将这些多粒度几何线索注入到特征提取层次中，增强了捕捉重要几何特征的能力。大量实验表明，GAPrompt++在基于提示的PEFT方法中实现了最先进的性能，甚至在多个基准测试中超越了完全微调，同时所需的可训练参数少于2%。此外，为了解决现有评估数据集的饱和问题，我们构建了两个更具挑战性的基准，源自3D高斯溅射（3D Gaussian Splatting）和多视图立体重建（Multi-View Stereo），提供多样化和真实的点云场景，以促进未来研究。

View on arXiv Download PDF AI Translation

cs.CV / 26 / 2609.19719

SeetaPsych v1.0: An Open-source Computer Vision Toolkit for Behavior-based Psychological Measurement

SeetaPsych v1.0：一个开源计算机视觉工具包，用于基于行为的心理测量

Zeng, Jiabei, Li, Chiqin, Li, Kaizhou, Chang, Fei, Li, Yong, Zhao, Yuanhao, Han, Dan, Yang, Wenqiang, Chen, Xilin, Shan, Shiguang

Abstract

Automated visual analysis opens new avenues for behavior--based psychological measurement. Nevertheless, existing technological modules are typically scattered across task specific systems with heterogeneous interfaces and disparate deployment requirements. In this work, we present SeetaPsych v1.0, an open source, unified and extensible computer vision toolkit designed to extract psychologically relevant signals from facial images and/or face based videos. The current release encompasses four major core modules aiming at behavior--based physiological perception: unified face based emotion analysis (simultaneous facial expression recognition, facial action unit detection, and valence--arousal estimation), camera based heart rate estimation, screen point--of--gaze estimation, and scene gaze following. A suite of auxiliary preprocessing modules for human centric visual analysis is also included, comprising face detection, facial landmark detection, and head detection. These functionalities are encapsulated within a modular Pipeline/Runner architecture that automatically resolves attribute dependencies, constructs computation graphs, and support intermediate result sharing among modules. SeetaPsych provides standardized Python APIs to facilitate reproducible, large scale analyses, alongside an interactive WebUI for rapid, code--free method evaluation. Overall, SeetaPsych offers an integrated and accessible visual measurement platform for research in psychology, behavioral science, human computer interaction, and related fields.

Chinese Translation

自动化视觉分析为基于行为的心理测量开辟了新的途径。然而，现有的技术模块通常分散在特定任务的系统中，具有异构接口和不同的部署要求。在本研究中，我们提出了SeetaPsych v1.0，这是一个开源、统一且可扩展的计算机视觉工具包，旨在从面部图像和/或基于面部的视频中提取心理相关信号。当前版本包含四个主要核心模块，旨在进行基于行为的生理感知：统一的基于面部的情感分析（同时进行面部表情识别、面部动作单元检测和愉悦度-唤醒度估计）、基于摄像头的心率估计、屏幕注视点估计以及场景注视跟踪。还包括一套用于以人为中心的视觉分析的辅助预处理模块，包括面部检测、面部关键点检测和头部检测。这些功能被封装在一个模块化的Pipeline/Runner架构中，该架构自动解决属性依赖关系，构建计算图，并支持模块之间的中间结果共享。SeetaPsych提供标准化的Python API，以促进可重复的大规模分析，并配备交互式WebUI以快速、无代码的方法评估。总体而言，SeetaPsych为心理学、行为科学、人机交互及相关领域的研究提供了一个集成且易于访问的视觉测量平台。

View on arXiv Download PDF AI Translation

cs.CV / 27 / 2609.19729

Recency Forcing: Bridging the Long-Horizon Gap in Autoregressive Video Generation

近期强制：弥合自回归视频生成中的长时间跨度差距

Cao, Tri, Nguyen, Hung, Nguyen, Phong, Nguyen, Khoi

Abstract

Autoregressive (AR) video generation degrades over long horizons due to an overlooked train-inference discrepancy we term KV eviction mismatch: models train on short clips where all context frames reside in the KV cache, but at inference, memory constraints force distant frames to be evicted from the KV cache - removing context the model was conditioned on. Rather than simulating eviction via context truncation - which discards temporal information the model still needs and degrades motion coherence - we keep the context but while progressively reducing the influence of distant frames, making their eventual eviction negligible. To guide this design, we introduce the positional response $R( \Delta, \, t_{\text{denoise}})$, a perturbation-based sensitivity measure revealing that context influence decays steeply with temporal distance and varies systematically across denoising steps. Motivated by this analysis, we propose Recency Forcing, which applies a non-positive, timestep-dependent bias, termed Temporal Response Bias (TRB), on pre-softmax attention logits derived directly from $R$, closing the train-inference gap without modifying context length or training objectives. We further introduce Biased Attention Reparameterization (BAR), an exact reformulation that moves the bias outside the softmax, making TRB a standard FlashAttention call at zero overhead. Recency Forcing operates in both training-free mode and training-based mode. Experiments on VBench and VBench-Long demonstrate state-of-the-art long-horizon generation quality at no additional inference cost.

Chinese Translation

自回归（AR）视频生成在长时间跨度上表现不佳，这主要是由于我们称之为KV驱逐不匹配的训练-推理差异被忽视：模型在所有上下文帧都驻留在KV缓存中的短片段上进行训练，但在推理时，内存限制迫使远处帧从KV缓存中驱逐，从而移除了模型所依赖的上下文。我们并不通过上下文截断来模拟驱逐——这种方法会丢弃模型仍需的时间信息并降低运动一致性——而是保持上下文，同时逐步减少远处帧的影响，使其最终驱逐变得微不足道。为了指导这一设计，我们引入了位置响应 $R( ext{Δ}, ext{t}_{ ext{denoise}})$，这是一种基于扰动的敏感性度量，揭示了上下文影响随着时间距离的增加而急剧衰减，并且在去噪步骤中系统性变化。基于这一分析，我们提出了近期强制（Recency Forcing），它在直接从 $R$ 得出的预-softmax注意力对数上施加了一种非正的、依赖于时间步的偏置，称为时间响应偏置（Temporal Response Bias, TRB），从而在不修改上下文长度或训练目标的情况下，缩小训练与推理之间的差距。我们进一步引入了偏置注意力重参数化（Biased Attention Reparameterization, BAR），这是一种精确的重构方法，将偏置移出softmax，使TRB成为零开销的标准FlashAttention调用。近期强制在无训练模式和基于训练的模式下均可操作。在VBench和VBench-Long上的实验表明，长时间跨度生成质量达到了最先进水平且没有额外的推理成本。

View on arXiv Download PDF AI Translation

cs.CV / 28 / 2609.19740

Federated Learning Framework for Privacy-Preserving Kidney Stone Detection

隐私保护的肾结石检测的联邦学习框架

Younas, Najiyya, Abdulkader, Omar, Shah, Yaser Ali, Ikram, Muhammad Jawad, Khan, Jebran, Khalil, Amaad

Abstract

Recent innovations in deep learning have significantly enhanced the diagnosis of medical images, although they are based on the use of centralized data storage that pose severe threats to patient privacy and medical data security. To address this issue, this research proposes a Federated Learning (FL) model that is coupled with an optimized YOLOv8 network to detect the kidney stones on a computed tomography (CT) image and at the same time, protect privacy of the patients. The suggested system can help various medical organizations to jointly train a common model without exchanging the information about the patients. This is to ensure that data protection laws like GDPR and HIPAA are adhered to. The residual feature fusion and DropBlock regularization among other architectural improvements are also included in YOLOv8 to enhance detection robustness and minimize overfitting. Experimental analysis carried out on a distributed CT dataset demonstrated that the federated YOLOv8 model has a mAP at 50 of 0.733 and is able to keep the data confidential. Moreover, its lean design facilitates fast edge deployment and real-time inference across a clinical setting. Altogether, these findings indicate that Federated Learning is a safe and efficient solution to AI-assisted diagnosis in contemporary healthcare when combined with the use of sophisticated object detection models.

Chinese Translation

近年来，深度学习的创新显著提升了医学图像的诊断能力，尽管这些技术依赖于集中式数据存储，这对患者隐私和医疗数据安全构成了严重威胁。为了解决这一问题，本研究提出了一种联邦学习（Federated Learning, FL）模型，该模型与优化的YOLOv8网络相结合，用于在计算机断层扫描（CT）图像上检测肾结石，同时保护患者隐私。所建议的系统可以帮助各种医疗机构共同训练一个通用模型，而无需交换患者信息。这确保了遵循数据保护法律，如通用数据保护条例（GDPR）和健康保险可携带性与责任法案（HIPAA）。YOLOv8中还包含了残差特征融合和DropBlock正则化等架构改进，以增强检测的鲁棒性并最小化过拟合。在一个分布式CT数据集上进行的实验分析表明，联邦YOLOv8模型在mAP（平均精度均值）为50时达到了0.733，并能够保持数据的机密性。此外，其精简的设计便于快速边缘部署和在临床环境中的实时推断。总体而言，这些发现表明，联邦学习结合先进的目标检测模型，是现代医疗保健中AI辅助诊断的安全高效解决方案。

View on arXiv Download PDF AI Translation

cs.CV / 29 / 2609.19745

Region-Level Policy Optimization for Fine-grained MLLM Perception

区域级政策优化用于细粒度多模态大语言模型感知

Shi, Yuheng, Pei, Xiaohuan, Dong, Minjing, Xu, Chang

Abstract

Fine-grained visual perception in MLLMs is commonly improved by raising the resolution, but the added visual tokens inflate vision-encoding and language-model prefilling costs. We show that the two operations underlying fine-grained perception, localizing the region of interest (RoI) and recognizing its content, have different resolution requirements. In a controlled diagnostic, localization tolerates roughly 3 to 4 times stronger token compression than recognition, which motivates localizing from a coarse view and concentrating resolution on the selected evidence. Decoding coordinates with the MLLM can be trained end-to-end from answers, but costs a full model pass per query and depends on grounding ability. A lightweight proposal network distilled from the model's attention is fast, but inherits the noise of its attention targets. The RoI from the proposal network reaches the answer through a discrete region choice, so its faithfulness to the answer cannot supervise the network. We therefore optimize the proposal network with region-level reinforcement learning, which we call Vision-RL2. It treats coherent regions as actions, and a frozen MLLM reader scores each one by how its removal changes the answer likelihood. Complementary subtractive and additive objectives suppress distracting proposals and recover missing evidence, updating only the predictor without region annotations, response sampling, or reasoning trajectories. The refined proposal further enables a sparse encoding that magnifies evidence and excludes background tokens. Across six fine-grained benchmarks and four MLLM backbones, Vision-RL2 improves accuracy over the base model at every token budget and surpasses its largest-budget accuracy with about 4 times fewer visual tokens. Code is available at https://github.com/YuHengsss/VisionRL2 .

Chinese Translation

在多模态大语言模型（MLLMs）中，细粒度视觉感知通常通过提高分辨率来改善，但增加的视觉标记会增加视觉编码和语言模型预填充的成本。我们展示了细粒度感知的两个基本操作，即定位感兴趣区域（RoI）和识别其内容，具有不同的分辨率要求。在一个受控的诊断中，定位容忍大约3到4倍更强的标记压缩，而识别则不然，这激励我们从粗略视图进行定位，并将分辨率集中在所选证据上。与MLLM解码坐标可以通过答案进行端到端训练，但每个查询需要完整的模型传递，并依赖于基础能力。一个从模型注意力中提炼出的轻量级提议网络速度较快，但继承了其注意力目标的噪声。提议网络中的RoI通过离散区域选择达到答案，因此其对答案的忠实度无法监督网络。因此，我们使用区域级强化学习优化提议网络，称之为Vision-RL2。它将连贯区域视为动作，而一个冻结的MLLM阅读器通过移除这些区域对答案可能性的变化来对每个区域进行评分。互补的减法和加法目标抑制干扰提议并恢复缺失证据，仅更新预测器，而无需区域注释、响应采样或推理轨迹。经过精炼的提议进一步实现了稀疏编码，放大了证据并排除了背景标记。在六个细粒度基准和四个MLLM骨干网络上，Vision-RL2在每个标记预算下都提高了准确性，并在约减少4倍视觉标记的情况下超越了其最大预算的准确性。代码可在 https://github.com/YuHengsss/VisionRL2 获得。

View on arXiv Download PDF AI Translation

cs.CV / 30 / 2609.19747

STAR: Structure-aware Test-time Adaptation for diffusion-based light field Reconstruction

STAR：基于结构感知的扩散光场重建测试时适应

Choi, Wontae, Moon, Ki Ryum, Lee, Jae Young, Yun, Hyung Sup, Chun, Il Yong

Abstract

Light field (LF) reconstruction from limited and noisy focal stack (FS) measurements is a highly ill-posed inverse problem. Although the LF-to-FS imaging geometry is fixed for a given optical setup, LF spatial-angular structure---including within-view spatial details, cross-view angular dependencies, and disparity across views---varies across scenes. Consequently, a fixed pre-trained prior may not optimally capture the spatial-angular structure of each test LF. We propose Structure-aware Test-time Adaptation for diffusion-based light field Reconstruction (STAR), the first test-time adaptation framework for reconstructing an LF from FS. For each test LF, STAR freezes a pre-trained diffusion prior and fits three lightweight adapters to the observed FS to jointly adapt the three components of the LF's spatial-angular structure. STAR outperforms existing state-of-the-art methods in both two- and three-focal-sheet settings, with shorter inference times than those with test-time parameter updates.

Chinese Translation

从有限且噪声较大的焦点堆栈（FS）测量中重建光场（LF）是一个高度不适定的逆问题。尽管对于给定的光学设置，LF到FS的成像几何是固定的，但LF的空间-角度结构——包括视图内的空间细节、视图间的角度依赖性以及视图间的视差——在不同场景中是变化的。因此，固定的预训练先验可能无法最佳地捕捉每个测试LF的空间-角度结构。我们提出了基于结构感知的扩散光场重建测试时适应（STAR），这是第一个用于从FS重建LF的测试时适应框架。对于每个测试LF，STAR冻结一个预训练的扩散先验，并为观察到的FS拟合三个轻量级适配器，以共同适应LF空间-角度结构的三个组成部分。STAR在两种和三种焦点片设置中均优于现有的最先进方法，并且推理时间比那些进行测试时参数更新的方法更短。

View on arXiv Download PDF AI Translation

cs.CV / 31 / 2609.19767

Benchmarking MLLMs via Cognitive Expected Scene Graph for Safety-Critical Visual Negation Understanding

通过认知期望场景图对安全关键视觉否定理解进行基准测试

Jiang, Zhiyun, Wang, Hanyong, Liang, Binbin, Xie, Yu, Yang, Menglong, Li, Wei

Abstract

True machine intelligence requires transcending passive pixel registration to master top-down functional reasoning over absent information via visual negation understanding. However, unconstrained visual negation paradigms remain overly open-ended, and pervasive affirmation bias causes both existing Multi-Modal Large Language Models (MLLMs) and evaluation metrics to fail under negative semantics. To solve these intertwined challenges systematically, we first anchor the boundaries of negation reasoning within specific cognitive goals. Specifically, by focusing on safety as a highly pragmatic and critical cognitive dimension, we define the task of \textbf{S}cene \textbf{N}egation \textbf{U}nderstanding under \textbf{S}afety Cognition (\textbf{SNUS}). Under this framework, we construct a high-fidelity negative caption dataset mapping dense assertions of localized hazards. Concurrently, we propose the Cognitive Expected Scene Graph (CESG) Score, a structure-grounded, polarity-aware evaluation metric. Extensive experiments demonstrate that while current models struggle on the task, traditional metrics completely collapse under semantic reversals. Conversely, our framework delivers a solid benchmark for SNUS, providing a rigorous foundation to advance risk-aware situational comprehension and counterfactual cognition.

Chinese Translation

真正的机器智能需要超越被动的像素注册，通过视觉否定理解掌握对缺失信息的自上而下的功能推理。然而，无约束的视觉否定范式仍然过于开放，普遍的肯定偏见导致现有的多模态大型语言模型（Multi-Modal Large Language Models, MLLMs）和评估指标在负语义下失效。为系统性地解决这些交织的挑战，我们首先在特定的认知目标内界定否定推理的边界。具体而言，通过将安全性作为一个高度务实和关键的认知维度，我们定义了在安全认知下的场景否定理解任务（Scene Negation Understanding under Safety Cognition, SNUS）。在这一框架下，我们构建了一个高保真度的负面描述数据集，映射局部危害的密集断言。同时，我们提出了认知期望场景图（Cognitive Expected Scene Graph, CESG）评分，这是一种基于结构的、考虑极性的评估指标。大量实验表明，尽管当前模型在该任务上表现不佳，传统指标在语义反转下完全崩溃。相反，我们的框架为SNUS提供了一个坚实的基准，为推进风险意识的情境理解和反事实认知奠定了严格的基础。

View on arXiv Download PDF AI Translation

cs.CV / 32 / 2609.19793

AI Smart Glasses for Wearable Intelligence: From Egocentric Sensing to Agentic Personalization

可穿戴智能的AI智能眼镜：从自我中心感知到代理个性化

Yuan, Xu, Wang, Yi, Jiang, Zhuohang, Qu, Haohao, Ding, Yujuan, Lin, Shanru, Xing, Guoliang, Yang, Hongxia, Cao, Jiannong, Li, Qing, Fan, Wenqi

Abstract

Recent advances in artificial intelligence (AI) are reshaping smart glasses from egocentric capture and display devices into platforms for wearable intelligence. Smart glasses increasingly serve as wearable AI systems that connect first-person observation with real-time assistance under strict form-factor constraints. We frame this transition through the lens of \emph{AI smart glasses} and define them as a system-level concept in which egocentric sensing, resource-aware computing, intelligent reasoning, multimodal interaction, and real-world application constraints are co-designed for personalized assistance in the physical world. To systematically study this perspective, we organize the survey around four connected dimensions. First, we examine the hardware foundation that bounds sensing, computation, feedback delivery, and sustained deployment. Second, we study wearable intelligence, where egocentric signals are transformed into perceptual, contextual, and agentic capabilities. Third, we discuss interaction design, through which users request, receive, correct, and regulate assistance during ongoing activity. Fourth, we analyze application scenarios across healthcare, accessibility, situated learning, daily life assistance, cultural tourism, and industrial support, showing how domain requirements reshape system design and evaluation. We further identify five cross-cutting research challenges for future AI smart glasses: next-generation hardware, trustworthy egocentric intelligence, lifelong personalized memory, proactive intelligence, and embodied foundation models. By centering smart glasses as wearable-intelligence platforms, this survey provides a unified framework for organizing technologies, applications, and open challenges in this emerging area.

Chinese Translation

近期人工智能（AI）的进展正在将智能眼镜从自我中心的捕捉和显示设备转变为可穿戴智能的平台。智能眼镜越来越多地作为可穿戴AI系统，连接第一人称观察与在严格形态约束下的实时辅助。我们通过 extit{AI智能眼镜}的视角框架这一转变，并将其定义为一个系统级概念，其中自我中心感知、资源感知计算、智能推理、多模态交互和现实世界应用约束被共同设计，以提供物理世界中的个性化辅助。为了系统地研究这一视角，我们围绕四个相互关联的维度组织了调查。首先，我们考察了限制感知、计算、反馈传递和持续部署的硬件基础。其次，我们研究了可穿戴智能，其中自我中心信号被转化为感知、上下文和代理能力。第三，我们讨论了交互设计，通过该设计，用户在进行活动时请求、接收、纠正和调节辅助。第四，我们分析了医疗保健、无障碍、情境学习、日常生活辅助、文化旅游和工业支持等应用场景，展示了领域需求如何重塑系统设计和评估。我们进一步确定了未来AI智能眼镜的五个跨领域研究挑战：下一代硬件、可信的自我中心智能、终身个性化记忆、主动智能和具身基础模型。通过将智能眼镜作为可穿戴智能平台，本文提供了一个统一框架，以组织这一新兴领域中的技术、应用和开放挑战。

View on arXiv Download PDF AI Translation

cs.CV / 33 / 2609.19812

Absence is Presence: Understanding Visual Scene Negative Events Under Safety Cognitive Constraint

缺失即存在：在安全认知约束下理解视觉场景中的负事件

Jiang, Zhiyun, Wang, Hanyong, Liang, Binbin, Xie, Yu, Yang, Menglong, Li, Wei

Abstract

Traditional scene understanding focuses on affirmative information objectively present in images. However, in safety-critical domains, comprehending key information that should exist but is actually absent is vital for risk mitigation. To bridge this gap, we focus on visual scene negative captioning with safety as the cognitive constraint. The core challenge is to convert physical absence into semantic negative events. Existing vision-language models (VLMs) struggle with this process because affirmation bias suppresses negative reasoning, while limited mental filling capability and representation bias further hinder the inference of absent information. To address these challenges, we propose a negative captioning framework based on counterfactual reconstruction and contrastive decoding (CRCD). Inspired by human cognition, CRCD reformulates the task as counterfactual latent change captioning to bypass affirmation bias. It contrasts a synthesized safe expectation with reality to identify semantic omissions. To address limited mental filling, we design a dual-branch counterfactual reconstruction architecture. The amodal completion branch restores defective objects, while the functional association branch infers completely absent safety objects. Concurrently, a multi-condition representation learning mechanism is integrated to mitigate representation bias by projecting universal features onto predefined safety criteria subspaces, thereby capturing information across more dimensions. By decoding feature-level semantic residuals between the reconstructed scene prototype and raw input, CRCD bounds the non-existence search space and activates the decoder's negative logic. Extensive experiments validate the effectiveness of CRCD, establishing a high-performance baseline for this pioneering task.

Chinese Translation

传统的场景理解侧重于图像中客观存在的肯定信息。然而，在安全关键领域，理解应存在但实际上缺失的关键信息对于风险缓解至关重要。为了解决这一问题，我们关注在安全认知约束下的视觉场景负描述。核心挑战在于将物理缺失转化为语义负事件。现有的视觉-语言模型（VLMs）在这一过程中面临困难，因为肯定偏见抑制了负推理，而有限的心理填充能力和表征偏见进一步阻碍了缺失信息的推断。为了解决这些挑战，我们提出了一种基于反事实重建和对比解码（CRCD）的负描述框架。受人类认知的启发，CRCD将任务重新表述为反事实潜在变化描述，以绕过肯定偏见。它通过对比合成的安全期望与现实，识别语义遗漏。为了解决有限的心理填充问题，我们设计了一个双分支反事实重建架构。无模态补全分支恢复缺陷对象，而功能关联分支推断完全缺失的安全对象。同时，集成了一种多条件表征学习机制，通过将通用特征投影到预定义的安全标准子空间，以减轻表征偏见，从而捕捉跨多个维度的信息。通过解码重建场景原型与原始输入之间的特征级语义残差，CRCD限制了非存在搜索空间，并激活解码器的负逻辑。大量实验验证了CRCD的有效性，为这一开创性任务建立了高性能基线。

View on arXiv Download PDF AI Translation

cs.CV / 34 / 2609.19815

SnapPhysics: A Physics-Aware Scene Graph from a Single View for Interactive Mixed Reality Scenes

SnapPhysics：一种基于单视角的物理感知场景图框架，用于交互式混合现实场景

Kang, Suji, Kim, Seok-Young, Kim, Young Bin, Ha, Taewook, Schmalstieg, Dieter, Mori, Shohei, Woo, Woontack

Abstract

We propose SnapPhysics, a training-free framework that reconstructs 3D objects and estimates their physical properties such as mass, friction, and center of gravity from a single image. For physically coherent interactions in mixed reality (MR), such properties are as important as geometry. Prior approaches infer them by analyzing object dynamics in video, which is computationally costly, or by querying vision-language models (VLMs) on single images, which lacks geometric grounding and inter-object relationships. We address these limitations by combining instance-level 3D reconstruction and spatial alignment with a physics-aware scene graph that encodes these relationships and per-object metric geometry as structured context for VLM-based property reasoning. Experiments on 3D-FRONT show that SnapPhysics improves scene-level F-Score by 18.6% over the best learning-based method, and on real captured scenes with ground-truth mass, it reduces the mean absolute log difference error (mALDE) by up to 20.5% and improves log-scale correlation ($r^2_{\mathrm{ls}}$) by up to 19.6% over VLM-only estimation. SnapPhysics enables physically interactive MR experiences without manual parameter tuning. Project page: https://snapphysics-ismar2026.github.io/.

Chinese Translation

我们提出了SnapPhysics，这是一种无需训练的框架，可以从单张图像重建3D物体并估计其物理属性，如质量、摩擦力和重心。对于混合现实（MR）中的物理一致性交互，这些属性与几何形状同样重要。之前的方法通过分析视频中的物体动态来推断这些属性，这在计算上成本高昂，或者通过对单张图像查询视觉-语言模型（VLMs），但缺乏几何基础和物体间关系。我们通过结合实例级3D重建和空间对齐，构建一个物理感知场景图，编码这些关系以及每个物体的度量几何，作为基于VLM的属性推理的结构化上下文，从而解决了这些局限性。在3D-FRONT上的实验表明，SnapPhysics在场景级F-Score上比最佳学习方法提高了18.6%，在真实捕获场景中与真实质量相比，减少了平均绝对对数差异误差（mALDE）最多20.5%，并在对数尺度相关性（$r^2_{ ext{ls}}$）上提高了最多19.6%，相较于仅使用VLM的估计。SnapPhysics使得无需手动参数调节即可实现物理交互的MR体验。项目页面：https://snapphysics-ismar2026.github.io/

View on arXiv Download PDF AI Translation

cs.CV / 35 / 2609.19840

KoUniTalk: A Lightweight Articulation-Centered Korean-English 3D Talking Face Benchmark

KoUniTalk：一个轻量级的以发音为中心的韩英3D对话面孔基准

Chung, Hyunjung, Park, Unsang

Abstract

High-quality 3D talking face datasets remain largely English- centric, and Korean 3D facial motion data are difficult to combine with standard English benchmarks because of differences in mesh topology, spatial scale, coordinate system, and temporal sampling. We present KoUniTalk, a lightweight articulation-centered Korean-English 3D talk- ing face benchmark that retargets VOCASET and the released Korean speech-based 3D talking face data to a shared mesh topology using de- formation transfer. Rather than proposing a new deformation-transfer algorithm or a full-head identity-preserving avatar dataset, KoUniTalk provides an identity-neutral canonical output space for controlled speech- driven facial articulation training and evaluation across English and Ko- rean. The unified template contains 1,176 vertices and focuses on the mouth and adjacent lower- and mid-face regions, reducing the output dimensionality from 15,069 and 72,147 dimensions to 3,528 dimensions, corresponding to 4.27-fold and 20.45-fold reductions compared with VO- CASET/FLAME and the original Korean mesh, respectively. To exam- ine whether retargeting preserves speech-relevant motion, we evaluate semantic mouth-landmark trajectories, including mouth opening, mouth width, aperture ratio, and mouth-opening dynamics. Since the official test set of the Korean dataset is not publicly released, we additionally define a subject-disjoint Korean benchmark split. The processed matched benchmark contains 22 speakers, 4,978 sequences, and 642,781 frames, enabling Korean-English cross-dataset evaluation of speech-driven 3D fa- cial animation models in a single compact articulation-template space. Source-reported inventory counts are listed separately from these pro- cessed counts

Chinese Translation

高质量的3D对话面孔数据集在很大程度上仍然以英语为中心，而韩语3D面部运动数据由于网格拓扑、空间尺度、坐标系统和时间采样的差异，难以与标准英语基准相结合。我们提出了KoUniTalk，一个轻量级的以发音为中心的韩英3D对话面孔基准，它通过变形转移将VOCASET和发布的基于韩语的3D对话面孔数据重定向到共享的网格拓扑。KoUniTalk并不提出新的变形转移算法或完整的保持身份的头像数据集，而是提供了一个身份中立的标准输出空间，用于在英语和韩语之间进行受控的语音驱动面部发音训练和评估。统一模板包含1,176个顶点，重点关注嘴部及其相邻的下脸和中脸区域，将输出维度从15,069和72,147维减少到3,528维，分别对应与VOCASET/FLAME和原始韩语网格相比的4.27倍和20.45倍的减少。为了检验重定向是否保留了与语音相关的运动，我们评估了语义嘴部标志轨迹，包括嘴部开合、嘴宽、开口比和嘴部开合动态。由于韩语数据集的官方测试集未公开发布，我们还定义了一个主题不重叠的韩语基准拆分。处理后的匹配基准包含22位说话者、4,978个序列和642,781帧，使得在单一紧凑的发音模板空间中能够进行韩英跨数据集的语音驱动3D面部动画模型评估。源报告的库存计数与这些处理后的计数分别列出。

View on arXiv Download PDF AI Translation

cs.CV / 36 / 2609.19853

PACE: Precise AI Cinematic Expression: A Typed Specification for Script-Grounded Previsualization and Geometric Conformance

PACE：精确的人工智能电影表现：一种基于脚本的预可视化和几何一致性的类型化规范

Duan, Bing, Guo, Qiang, Li, Linpu, Mao, Zhijian, Zhu, Min, Ren, Zhirui, Yan, Yiwei, Chu, Xi, Li, Xiaoding

Abstract

Between a screenplay and a film sits a planning problem that is spatial first: who stands where, and what a camera sees from where it stands. An image diffusion model asked for a shot in free text settles that plan by its own defaults. We present PACE (Precise AI Cinematic Expression), a typed representation for the plan: the screenplay evidence, the characters, props and locations it needs, where each subject stands, and what the camera does. A value is written once at the level it belongs to (script, scene, shot or panel) and inherited below it. A compiler turns the result into both the prompt sent to the diffusion model and a 3D scene built in metres, and a camera solver places the camera so that the declared framing is the framing built. Where a declared value becomes geometry, PACE measures, field by field, how far the compiled camera and the staged render sit from the declaration, rather than asking a model to judge. On the 11-scene Automatic Drive screenplay, every staged single-subject panel places its subject within 1.2% of frame width of its declared position; with two or three subjects one camera pose cannot satisfy every position, and the residual is reported rather than absorbed. On 204 external director-storyboard shots, delivered head height is 1.906 times the staged target from the director's words, 1.733 from the compiled prompt, and 0.955 with the greybox control; the condition that holds framing best draws the described action least. Declaring the pose on 30 shots raises the action drawn from 58.9% to 74.4% without moving the framing. Transitions, fitted motion and human review of the generated panels remain open. Code: https://github.com/StudioPiLabs/pace-core

Chinese Translation

在剧本与电影之间存在一个空间优先的规划问题：谁站在哪里，摄像机从何处看到什么。一个基于自由文本的图像扩散模型通过其自身的默认设置来解决这一规划。我们提出了PACE（Precise AI Cinematic Expression），这是一种用于该计划的类型化表示：剧本证据、所需的角色、道具和场景、每个主体的位置以及摄像机的动作。一个值在其所属的层级（剧本、场景、镜头或面板）上写入一次，并在其下层继承。编译器将结果转化为发送给扩散模型的提示和以米为单位构建的3D场景，而摄像机求解器则将摄像机放置在声明的构图所构建的位置。PACE在声明的值变为几何时，逐字段测量编译后的摄像机和布置渲染与声明之间的距离，而不是要求模型进行判断。在11场景的《自动驾驶》剧本中，每个布置的单主体面板将其主体放置在声明位置的1.2%框宽度内；对于两个或三个主体，一个摄像机姿态无法满足每个位置，残差被报告而不是吸收。在204个外部导演-分镜头的镜头中，交付的头部高度是导演所述目标的1.906倍，从编译的提示中为1.733倍，而灰盒控制为0.955倍；保持构图最佳的条件使描述的动作最少。对30个镜头声明姿态将绘制的动作从58.9%提高到74.4%，而不移动构图。过渡、适配运动和对生成面板的人类审查仍然是开放的。代码链接： https://github.com/StudioPiLabs/pace-core

View on arXiv Download PDF AI Translation

cs.CV / 37 / 2609.19867

Socialized UAV Cross-Task Learning: Towards Cross-Granularity Collaboration through Hierarchical Interaction

社会化无人机跨任务学习：通过层次交互实现跨粒度协作

Yao, Xinjie, Zhao, Ruipu, Zhu, Yunqi, Fan, Zhihe, Guo, Zhoupeng, Li, Weihao, Wang, Zhen, Wang, Qilong, Zhu, Pengfei

Abstract

Joint learning across heterogeneous tasks is often treated as task coupling through feature sharing, distillation, or auxiliary supervision. However, in cross-task learning, mismatched representational and supervisory granularities make such coupling prone to interference, teacher bias, or unidirectional collapse. We argue that cross-granularity learning is fundamentally a problem of hierarchical interaction regulation rather than simple task coupling. This issue is particularly evident in UAV perception, where visual shifts and detection--segmentation objectives naturally form coarse- and fine-grained knowledge sources. To systematically study this problem, we introduce CrossUAV, a UAV benchmark for joint object detection and instance segmentation that provides a unified evaluation platform for cross-granularity task collaboration. To address these challenges, we propose Cross-Granularity Socialized Collaboration (CGSC), a progressive and adaptive framework that regulates when, where, and how tasks exchange information across network hierarchies. CGSC progressively activates cross-task interactions and adaptively adjusts the strength according to task contribution, suppressing harmful interference while exploiting complementary coarse- and fine-grained structures. Extensive experiments demonstrate consistent improvements on both tasks, validating hierarchical dynamic interaction as an effective mechanism for cross-granularity collaboration.

Chinese Translation

跨异构任务的联合学习通常被视为通过特征共享、蒸馏或辅助监督进行的任务耦合。然而，在跨任务学习中，不匹配的表征和监督粒度使得这种耦合容易受到干扰、教师偏差或单向崩溃的影响。我们认为，跨粒度学习根本上是一个层次交互调节的问题，而不仅仅是简单的任务耦合。这一问题在无人机（UAV）感知中尤为明显，其中视觉偏移和检测-分割目标自然形成粗粒度和细粒度的知识源。为了系统地研究这一问题，我们引入了CrossUAV，这是一个用于联合目标检测和实例分割的无人机基准，提供了一个统一的跨粒度任务协作评估平台。为了解决这些挑战，我们提出了跨粒度社会化协作（CGSC），这是一个渐进和自适应的框架，调节任务在网络层次之间何时、何地以及如何交换信息。CGSC逐步激活跨任务交互，并根据任务贡献自适应调整交互强度，抑制有害干扰，同时利用互补的粗粒度和细粒度结构。大量实验表明，在两个任务上均实现了一致的改进，验证了层次动态交互作为跨粒度协作的有效机制。

View on arXiv Download PDF AI Translation

cs.CV / 38 / 2609.19872

PART: Learning 3D Part Assembly and Retrieval with Transformers

PART：基于变换器的3D部件组装与检索学习

Bao, Ruchao, Wu, Wenzheng, Xiang, Chucheng, Liu, Zhongyuan, Liu, Yuan, Dong, Jinxin, Liu, Ligang, Wang, Ziqi

Abstract

3D assembly is fundamental to modern manufacturing and digital content creation. In this paper, we present PART, a unified transformer-based framework for 3D part retrieval and assembly: given a target shape and a part library, PART automatically selects the appropriate parts and predicts their 6-DoF poses to reconstruct the target. While prior work has achieved impressive progress on assembling a pre-defined set of parts, this more practical retrieval-based setting remains largely unexplored. The task faces three key challenges: (i) a combinatorially explosive search space that grows exponentially with library size; (ii) variable-length outputs, as different targets require different numbers of parts; and (iii) continuous 6-DoF pose estimation for part assembly. To address these, we formulate retrieval and assembly as a set prediction problem and design a novel transformer-based framework that retrieves parts and regresses their poses with variable-length output. Additionally, we exploit the duality between part pose estimation and target segmentation through joint training and a novel segmentation-enhanced optimization module. Finally, We curate a large-scale dataset of 80K+ shapes, and the results show that PART generalizes to scene layouts, image targets, and real-world scans. Project Page: https://iambrc.github.io/PART-project-page/.

Chinese Translation

3D组装是现代制造和数字内容创作的基础。本文提出了PART，一个统一的基于变换器的3D部件检索与组装框架：给定目标形状和部件库，PART能够自动选择合适的部件并预测其6自由度（6-DoF）姿态，以重建目标。尽管之前的研究在组装预定义部件集合方面取得了显著进展，但这种更实际的基于检索的设置仍然未被充分探索。该任务面临三个关键挑战：（i）随着库大小的增加，组合爆炸性的搜索空间呈指数增长；（ii）可变长度的输出，因为不同的目标需要不同数量的部件；（iii）部件组装的连续6-DoF姿态估计。为了解决这些问题，我们将检索和组装形式化为一个集合预测问题，并设计了一个新颖的基于变换器的框架，该框架能够检索部件并以可变长度输出回归其姿态。此外，我们通过联合训练和一个新颖的分割增强优化模块，利用了部件姿态估计与目标分割之间的对偶性。最后，我们整理了一个包含超过8万种形状的大规模数据集，结果表明PART能够推广到场景布局、图像目标和现实世界扫描。项目页面：https://iambrc.github.io/PART-project-page/

View on arXiv Download PDF AI Translation

cs.CV / 39 / 2609.19875

BINDER: A Latent Variable Model for Probabilistic Medical Image Registration

BINDER：一种用于概率医学图像配准的潜变量模型

Cerri, Stefano, Hassankhani, Amirhossein, Balbastre, Yaël, Van Leemput, Koen

Abstract

We propose a new probabilistic model for general-purpose medical image registration that builds upon the mutual information registration criterion. It centers around a spatial interpolation technique that assumes latent voxel-wise correspondences between the images being registered. By exploiting these latent variables, we derive dedicated optimization and MCMC sampling techniques that only involve closed-form iterative updates. When applied to nonlinear registration, an efficient demons-like optimization algorithm is obtained that shows robust out-of-the-box performance across a variety of monomodal and multimodal registration tasks. We also demonstrate a corresponding sampler that can quantify, for the first time, uncertainty in multimodal registration scenarios with very high-dimensional 3D deformations. Our code, which we call BINDER (Bayesian INference for DEformable Registration), is freely available at https://github.com/ste93ste/BINDER.

Chinese Translation

我们提出了一种新的概率模型，用于通用医学图像配准，该模型基于互信息配准标准。该模型围绕一种空间插值技术展开，该技术假设被配准图像之间存在潜在的体素级对应关系。通过利用这些潜变量，我们推导出专门的优化和马尔可夫链蒙特卡洛（MCMC）采样技术，这些技术仅涉及封闭形式的迭代更新。当应用于非线性配准时，获得了一种高效的类似恶魔（demons）优化算法，该算法在多种单模态和多模态配准任务中表现出强大的开箱即用性能。我们还展示了一个相应的采样器，首次能够量化在具有非常高维3D变形的多模态配准场景中的不确定性。我们的代码，称为BINDER（Bayesian INference for DEformable Registration），可在https://github.com/ste93ste/BINDER上免费下载。

View on arXiv Download PDF AI Translation

cs.CV / 40 / 2609.19876

SlugTrails: An Egocentric Benchmark for Floor Plan Localization in Large Buildings

SlugTrails：大型建筑物内基于第一人称视角的楼层平面图定位基准

Cheng, Yunqian, Manduchi, Roberto

Abstract

Floor-plan-based indoor visual localization enables infrastructure-free positioning, but most methods are developed and evaluated in small residential environments unlike the large public buildings of real deployment. We introduce SlugTrails, a floor plan localization benchmark for large indoor spaces under realistic egocentric sensing: $30$ Hz Aria glasses recordings across three campus buildings and six floors ($22089$ m$^2$ of floor plan outline), CAD-derived floor plans with semantic classes and circulation space masks, and trajectories aligned into the floor plan frame using laser-surveyed anchors. One protocol covers three practical ways of gathering geometry under a limited field of view -- a single walking frame, a stationary multi-view sweep, and a walking stream with odometry -- so methods designed for different regimes are compared on the same buildings and ground truth. Evaluating five representative geometric and learned systems under their native sensing configurations, we find that stock checkpoints (official released weights) are near zero on SlugTrails (at most $0.004$ R@1m30$^{\circ}$ on walking single frames), while fine-tuning on SlugTrails improves every trainable family on all three tasks (e.g., F$^3$Loc $0.0 \rightarrow 0.141$ single-frame and $0.03 \rightarrow 0.66$ sequential), with gains compounding as observations accumulate. The same fine-tuned weights also improve cross-dataset generalization on LaMAR with no LaMAR training (sequential R@1m $0.048 \rightarrow 0.143$ for F$^3$Loc and $0.063 \rightarrow 0.127$ for UnLoc), whereas train-from-scratch on SlugTrails alone stays far below fine-tuning from stock weights -- evidence that floor plan localization is currently limited by indoor data rather than by architecture. We release the dataset, protocols, and tools at https://github.com/Head-inthe-Cloud/SlugTrails.

Chinese Translation

基于楼层平面图的室内视觉定位实现了无需基础设施的定位，但大多数方法均在小型住宅环境中开发和评估，难以适应真实部署中的大型公共建筑。我们提出了SlugTrails，这是一个针对大型室内空间的楼层平面图定位基准，采用真实的第一人称感知数据：包括三个校园建筑六层楼的30 Hz Aria眼镜录制视频（覆盖22089平方米的楼层平面轮廓）、带有语义类别和流通空间掩码的CAD导出楼层平面图，以及利用激光测量锚点对轨迹进行楼层平面坐标系对齐。基准包含三种在有限视场下获取几何信息的实用协议——单帧行走视角、静止多视角扫描和带有里程计的连续行走流——以便在相同建筑和真实标签下比较针对不同感知模式设计的方法。我们在各自原生感知配置下评估了五种代表性的几何和学习型系统，发现官方预训练权重在SlugTrails上的表现几乎为零（单帧行走视角最高仅0.004的R@1m30°），而在SlugTrails上微调后，所有可训练模型在三项任务上均有显著提升（例如F³Loc单帧从0.0提升至0.141，序列任务从0.03提升至0.66），且随着观测数据积累，性能提升进一步叠加。相同的微调权重还提升了在未使用LaMAR训练的情况下对LaMAR数据集的跨数据集泛化能力（F³Loc序列R@1m从0.048提升至0.143，UnLoc从0.063提升至0.127），而仅在SlugTrails上从零训练的模型表现远低于基于预训练权重的微调，表明当前楼层平面图定位的瓶颈在于室内数据的缺乏，而非模型架构。我们已公开发布数据集、协议及工具，详见https://github.com/Head-inthe-Cloud/SlugTrails。

View on arXiv Download PDF AI Translation

cs.CV / 41 / 2609.19881

BinoGen: Scaling egocentric binocular data for embodied visual perception and learning

BinoGen：扩展自我中心双目数据以实现具身视觉感知与学习

Li, Chunpeng, Li, Ya-tang

Abstract

Embodied visual perception relies on temporally coherent visual experience accumulated through continuous engagement with the environment. However, collecting large-scale egocentric binocular observations together with dense annotations remains costly and difficult. Moreover, visual experience is shaped not only by the environment but also by the embodiment of the observer, including viewing height, field of view, binocular geometry, and motion through the scene. To address these challenges, we present BinoGen, an automated framework for generating large-scale, embodiment-aware egocentric binocular visual experiences in indoor environments. BinoGen jointly models environmental and observer variation through generative scene synthesis, probabilistic object instantiation, appearance randomization, stochastic trajectory generation, and configurable binocular camera setups. The framework produces synchronized binocular videos together with dense multimodal supervision, including depth maps, optical flow, surface normals, semantic maps, object coordinates, and camera poses. Using BinoGen, we construct a dataset comprising more than 20 million annotated images for supervised learning. We demonstrate two complementary utilities of BinoGen. First, incorporating BinoGen data consistently improves real-world visual perception, including depth estimation, object detection, and video object tracking. Second, paired human-inspired and mouse-inspired observations from the same environments enable controlled investigation of how observer embodiment affects perceptual learning. Embodiment-specific adaptation substantially improves performance, while joint training enables a single model to perform competitively across both embodiments. Together, these results demonstrate that large-scale, controllable visual experience can improve embodied perception...

Chinese Translation

具身视觉感知依赖于通过与环境的持续互动积累的时间一致的视觉经验。然而，收集大规模的自我中心双目观察数据及其密集注释仍然成本高昂且困难。此外，视觉经验不仅受环境影响，还受到观察者的具身性影响，包括观察高度、视野、双目几何和在场景中的运动。为了解决这些挑战，我们提出了BinoGen，一个用于在室内环境中生成大规模、具身感知的自我中心双目视觉体验的自动化框架。BinoGen通过生成场景合成、概率对象实例化、外观随机化、随机轨迹生成和可配置的双目相机设置，共同建模环境和观察者的变化。该框架生成同步的双目视频，并提供密集的多模态监督，包括深度图、光流、表面法线、语义图、对象坐标和相机姿态。利用BinoGen，我们构建了一个包含超过2000万张注释图像的数据集，用于监督学习。我们展示了BinoGen的两个互补功能。首先，结合BinoGen数据可以持续改善现实世界的视觉感知，包括深度估计、物体检测和视频物体跟踪。其次，来自同一环境的受人类启发和受鼠类启发的观察数据对可以控制地研究观察者的具身性如何影响感知学习。具身特定的适应显著提高了性能，而联合训练使得单一模型能够在两种具身性之间竞争性表现。综合来看，这些结果表明，大规模、可控的视觉体验能够改善具身感知。

View on arXiv Download PDF AI Translation

cs.CV / 42 / 2609.19907

GS-PI: An Optimization-Decoupled Appearance Decomposition Approach for Generating PBR Gaussian Assets

GS-PI：一种优化解耦的外观分解方法，用于生成PBR高斯资产

Xu, Jieting, Xie, Rengan, Huang, Zijian, Jin, Zehui, Wang, Rui, Huo, Yuchi

Abstract

Gaussian Splatting (GS) excels at novel-view synthesis but encodes baked-in radiance, tightly entangling illumination with geometry and preventing seamless integration into physically based rendering (PBR) pipelines. Existing inverse-rendering methods attempt to disentangle materials via joint optimization, but often suffer from competing objectives that cause severe ambiguities and residual lighting artifacts. To overcome this, we present GS-PI, a novel optimization-decoupled framework that casts PBR material generation as a geometry-conditioned diffusion process on 3D point clouds. By operating directly in the 3D domain, our method inherently guarantees multi-view consistency, sidestepping the severe pixel correspondence issues that challenge 2D diffusion approaches. We introduce a multi-scale cross-view conditioning mechanism that integrates three complementary components: a global semantic prior, source-anchored photometric cues, and an absolute spatial learned view-direction conditioning signal. This design efficiently compresses complex multi-view evidence, mitigating cross-view projection misalignment and successfully preventing specular highlights from baking into intrinsic colors. By extracting a point cloud from a pre-trained Gaussian model, predicting PBR attributes via conditional diffusion, and distilling them back through differentiable rasterisation, we yield a fully relightable PBR-GS asset. GS-PI outperforms recent inverse-rendering baselines while replacing per-scene joint illumination/BRDF optimization with a learned diffusion pass followed by a short target-driven distillation, without requiring proxy meshes.

Chinese Translation

高斯点云（Gaussian Splatting, GS）在新视角合成方面表现出色，但其编码的辐射光线与几何形状紧密纠缠，阻碍了其与基于物理的渲染（PBR）管线的无缝集成。现有的逆向渲染方法试图通过联合优化来解耦材料，但常常受到竞争目标的影响，导致严重的模糊和残余光照伪影。为了解决这一问题，我们提出了GS-PI，这是一种新颖的优化解耦框架，将PBR材料生成视为在3D点云上的几何条件扩散过程。通过直接在3D域中操作，我们的方法本质上保证了多视角一致性，避免了2D扩散方法面临的严重像素对应问题。我们引入了一种多尺度跨视角条件机制，整合了三个互补组件：全局语义先验、源锚定的光度线索和绝对空间学习的视角方向条件信号。该设计有效压缩了复杂的多视角证据，减轻了跨视角投影错位，并成功防止了镜面高光烘焙到内在颜色中。通过从预训练的高斯模型中提取点云，通过条件扩散预测PBR属性，并通过可微分光栅化将其提炼回来，我们生成了一个完全可重光照的PBR-GS资产。GS-PI在性能上优于近期的逆向渲染基线，同时将每场景的联合光照/BRDF优化替换为学习的扩散过程，后接短期目标驱动的提炼，而无需代理网格。

View on arXiv Download PDF AI Translation

cs.CV / 43 / 2609.19911

CitySTAR: Structured and Topology-Aware Reasoning for Open-Vocabulary Urban 3D Grounding

CitySTAR：面向开放词汇的城市3D定位的结构化与拓扑感知推理

Zhang, Shuai, Hou, Hongye, Liu, Qinghe, Li, Zhuoxiao, Wu, Dongli, Ou, Jing, Liu, Yuan, Zhao, Wufan

Abstract

3D grounding aims to localize target entities in complex scenes from natural language and plays a fundamental role in embodied perception and spatial reasoning. However, existing approaches mostly rely on feature similarity or direct matching, making it difficult to connect natural-language intent with the implicit semantic and geometric structures hidden in billion-scale urban point clouds. We reformulate city-scale 3D grounding as structured constraint reasoning, where description semantics are organized into computable cross-modal constraints over open-vocabulary 3D entities, attributes, and spatial relations. We present CitySTAR, a training-free framework for reasoning-driven urban 3D grounding. CitySTAR lifts raw billion-scale urban point clouds into a query-ready scene graph of open-vocabulary 3D instances, with CodeLLM-driven tools supplying multimodal evidence for node attributes and 3D spatial relations. It then models target-context topology with paired hypergraphs and performs bidirectional topology verification for structural disambiguation. Finally, a Reflective Cross-modal Grounding module integrates topology consistency and candidate-centered 2D visual evidence to make decisions over a metric-aware 3D context graph. To further support this setting, we introduce CitySTAR-3D, an enhanced benchmark that improves semantic coverage, instance completeness, bounding-box fidelity, and spatial-relation complexity in city-scale 3D grounding. Extensive experiments show that CitySTAR consistently improves open-world urban 3D grounding while maintaining strong interpretability and generalization.

Chinese Translation

3D定位旨在从自然语言中定位复杂场景中的目标实体，并在具身感知和空间推理中发挥基础性作用。然而，现有的方法大多依赖于特征相似性或直接匹配，这使得将自然语言意图与隐藏在数十亿规模城市点云中的隐含语义和几何结构联系起来变得困难。我们将城市规模的3D定位重新表述为结构约束推理，其中描述语义被组织为可计算的跨模态约束，涵盖开放词汇的3D实体、属性和空间关系。我们提出了CitySTAR，一个无训练的框架，用于推理驱动的城市3D定位。CitySTAR将原始的数十亿规模城市点云提升为查询就绪的开放词汇3D实例场景图，CodeLLM驱动的工具为节点属性和3D空间关系提供多模态证据。然后，它通过配对超图建模目标-上下文拓扑，并执行双向拓扑验证以进行结构消歧。最后，一个反射性跨模态定位模块整合拓扑一致性和以候选为中心的2D视觉证据，以在度量感知的3D上下文图上做出决策。为了进一步支持这一设置，我们引入了CitySTAR-3D，一个增强基准，改善了城市规模3D定位中的语义覆盖、实例完整性、边界框保真度和空间关系复杂性。大量实验表明，CitySTAR在保持强解释性和泛化能力的同时，持续改善开放世界城市3D定位。

View on arXiv Download PDF AI Translation

cs.CV / 44 / 2609.19927

DirtyMoCap: Robust Motion Capture from Unconstrained Markers

DirtyMoCap：无约束标记的鲁棒运动捕捉

Wang, Long, Zhao, Shuting, Yan, Shen, Yu, Siyuan, Li, Xiaoben, Cai, Zeyu, Hou, Yumeng, Xiu, Yuliang

Abstract

Optical motion capture delivers high-fidelity human motion, but its reliance on strict marker layouts and clean trajectories severely limits its real-world applicability. In practice, tracking systems frequently output unconstrained markers: sparse, noisy, and unordered point clouds with unknown or varying configurations. To bridge the gap between corrupted raw markers and parametric human models, we introduce DirtyMoCap, a robust, marker-layout-free framework. Our core insight is to map unordered marker observations to a fixed set of "proxy anchors" comprising skeletal joints and body surface points, which serve as a stable intermediate representation. We first initialize and track these anchors over long sequences using a recurrent sliding-window architecture. Then, a custom differentiable Gauss-Newton solver fits the SMPL-H model to the tracked anchors to recover full-body pose, translation, and shape. By explicitly deriving geometric residuals, our solver learns adaptive observation confidence, smoothness, and prior weights end-to-end, adapting dynamically to the reliability of the input data. Extensive experiments on diverse, noisy marker configurations demonstrate that DirtyMoCap successfully generalizes across arbitrary layouts using only a single trained model. It consistently outperforms state-of-the-art configuration-specific baselines in both joint and vertex reconstruction accuracy, while our custom CUDA solver achieves up to a 100x speedup over standard PyTorch implementations. We further apply DirtyMoCap to heterogeneous raw optical MoCap recordings of traditional Chinese martial arts, yielding a Kung Fu motion dataset of temporally coherent SMPL-H reconstructions. Code and data are available at https://wanglongzju.github.io/DirtyMoCap-Project-Page.

Chinese Translation

光学运动捕捉能够提供高保真的人类运动数据，但其对严格标记布局和干净轨迹的依赖严重限制了其在现实世界中的适用性。在实际应用中，跟踪系统经常输出无约束标记：稀疏、嘈杂且无序的点云，具有未知或变化的配置。为了弥合损坏的原始标记与参数化人类模型之间的差距，我们提出了DirtyMoCap，一个鲁棒的、无标记布局的框架。我们的核心见解是将无序的标记观测映射到一组固定的“代理锚点”，这些锚点包括骨骼关节和身体表面点，作为稳定的中间表示。我们首先使用递归滑动窗口架构初始化并跟踪这些锚点，覆盖长序列。然后，一个定制的可微分高斯-牛顿求解器将SMPL-H模型拟合到跟踪的锚点，以恢复全身姿态、平移和形状。通过显式推导几何残差，我们的求解器能够端到端地学习自适应观测置信度、平滑性和先验权重，动态适应输入数据的可靠性。在多样化、嘈杂的标记配置上的大量实验表明，DirtyMoCap成功地在任意布局中进行泛化，仅使用一个训练模型。它在关节和顶点重建精度上始终优于最先进的特定配置基线，同时我们的自定义CUDA求解器在速度上比标准PyTorch实现快达100倍。我们进一步将DirtyMoCap应用于传统中国武术的异构原始光学运动捕捉录音，生成了一个时间一致的SMPL-H重建的功夫动作数据集。代码和数据可在https://wanglongzju.github.io/DirtyMoCap-Project-Page获取。

View on arXiv Download PDF AI Translation

cs.CV / 45 / 2609.19964

Enhanced Knowledge Distillation for Detection Transformer via Teacher Prediction Refinement

通过教师预测精炼增强检测变换器的知识蒸馏

Xing, Yitong, Cheng, Yuhao, Li, Yanping, Yan, Yichao

Abstract

Detection Transformers (DETRs) achieve strong performance in object detection but remain challenging to deploy on edge devices due to their high computational cost. Existing DETR distillation methods mainly focus on aligning distillation points, while largely overlooking the quality of the teacher's supervision itself. We observe that due to stage-wise non-monotonic prediction behavior in DETRs, well-localized or correctly classified predictions from earlier stages may degrade in later ones, and some negative predictions become increasingly overconfident. As a result, relying solely on the current stage's predictions yields inaccurate and inconsistent supervision. To address this issue, we propose Teacher Prediction Refinement Distillation (TPRD), a plug-and-play module that refines teacher predictions before distillation by exploiting stage-wise prediction information. TPRD improves supervision quality through Positive Prediction Correction (PPC), which corrects degraded positive predictions by restoring more accurate ones from earlier stages, ensuring reliable localization and classification signals, and Negative Prediction Suppression (NPS) suppresses the influence of overconfident negatives, preventing them from providing misleading supervision to the student. To preserve informative dark knowledge, we further introduce Maximum Dark Knowledge Preservation (MDKP), which selectively refines target-class logits while retaining non-target relations. Extensive experiments on MS COCO and PASCAL VOC demonstrate the effectiveness and robustness of the proposed method. Our code is available at https://github.com/xingyitong1/TPRD.

Chinese Translation

检测变换器（Detection Transformers, DETRs）在目标检测中表现出色，但由于其高计算成本，仍然难以在边缘设备上部署。现有的DETR蒸馏方法主要集中在对齐蒸馏点，而在很大程度上忽视了教师监督本身的质量。我们观察到，由于DETR中阶段性非单调预测行为，早期阶段的良好定位或正确分类的预测可能在后期阶段退化，而一些负面预测则变得越来越自信。因此，仅依赖当前阶段的预测会导致不准确和不一致的监督。为了解决这个问题，我们提出了教师预测精炼蒸馏（Teacher Prediction Refinement Distillation, TPRD），这是一个即插即用的模块，通过利用阶段性预测信息，在蒸馏之前精炼教师预测。TPRD通过正预测修正（Positive Prediction Correction, PPC）提高监督质量，该方法通过恢复来自早期阶段的更准确的预测来修正退化的正预测，从而确保可靠的定位和分类信号；负预测抑制（Negative Prediction Suppression, NPS）则抑制过于自信的负预测的影响，防止它们向学生提供误导性的监督。为了保留信息丰富的暗知识，我们进一步引入了最大暗知识保留（Maximum Dark Knowledge Preservation, MDKP），该方法在保留非目标关系的同时，选择性地精炼目标类的logits。在MS COCO和PASCAL VOC上的大量实验表明了所提方法的有效性和鲁棒性。我们的代码可在https://github.com/xingyitong1/TPRD获取。

View on arXiv Download PDF AI Translation

cs.CV / 46 / 2609.19966

Beyond the Foreground: FOV-Aware Polyp Image Synthesis via Lesion-Guided Adaptive Mucosal Context Propagation

超越前景：基于病变引导的自适应黏膜上下文传播的视场感知息肉图像合成

Wang, Tong, He, Yuting, Ren, Bin, Xie, Yutong, Yang, Guanyu

Abstract

Synthetic image and mask pairs can alleviate scarce colonoscopy annotations, but realistic synthesis requires preserving the supplied lesion while generating compatible mucosa. Existing foreground-guided methods treat all non-foreground pixels as background and rely mainly on local integration. Directly applying them to colonoscopy causes two problems: non-mucosal black regions contaminate generated tissue, and local reasoning produces inconsistent mucosal texture and illumination. We propose LAMP, the first foreground-guided framework for polyp image synthesis based on lesion-guided adaptive mucosal context propagation. LAMP explicitly separates the lesion, valid mucosa, and camera exterior using a field-of-view (FOV) mask. Lesion-to-Mucosa cross-attention extracts lesion appearance conditions for valid-mucosa locations, while FOV-constrained multidirectional Vision Receptance Weighted Key Value propagates them over legal tissue support. An adaptive gate then controls their residual fusion into the diffusion U-Net. Extensive experiments on five polyp datasets demonstrate that LAMP substantially outperforms existing methods in overall generation quality and consistently improves five downstream segmentation models. Our code will be released at https://github.com/wangtong627/LAMP.

Chinese Translation

合成图像和掩膜对可以缓解稀缺的结肠镜注释，但现实的合成需要在生成兼容的黏膜时保留提供的病变。现有的前景引导方法将所有非前景像素视为背景，主要依赖局部集成。直接将其应用于结肠镜检查会导致两个问题：非黏膜的黑色区域污染生成的组织，局部推理产生不一致的黏膜纹理和照明。我们提出了LAMP，这是第一个基于病变引导的自适应黏膜上下文传播的前景引导息肉图像合成框架。LAMP使用视场（FOV）掩膜明确分离病变、有效黏膜和相机外部。病变到黏膜的交叉注意力提取有效黏膜位置的病变外观条件，而FOV约束的多方向视觉接收加权关键值在合法组织支持上进行传播。然后，自适应门控控制它们在扩散U-Net中的残差融合。在五个息肉数据集上的广泛实验表明，LAMP在整体生成质量上显著优于现有方法，并且持续改善五个下游分割模型。我们的代码将发布在 https://github.com/wangtong627/LAMP。

View on arXiv Download PDF AI Translation

cs.CV / 47 / 2609.19973

An Event Preserving Velocity Invariant Representation for Event Cameras

一种事件保持速度不变的事件相机表示

Ikura, Mikihiro, Gava, Luna, Wu, Jiahang, Bartolozzi, Chiara, Glover, Arren

Abstract

Event cameras provide low-latency, high temporal resolution perception for real-time vision tasks such as robotics.The novel circuitry (i.e. asynchronous, independent pixels) that enables these advantages also introduces new algorithmic challenges. Velocity-invariant representations alleviate missing observations under slow motion and motion blur under fast motion, but most discard temporal information by converting events into image-like representations. We propose Set of Centre Active Receptive Fields (SCARF), a real-time velocity-invariant representation that preserves raw events while consistently handling fast motion, stationary scenes, and independently moving objects. SCARF achieves state-of-the-art performance in both computational efficiency and representation quality.

Chinese Translation

事件相机为实时视觉任务（如机器人技术）提供了低延迟、高时间分辨率的感知。这种新型电路（即异步、独立像素）虽然带来了这些优势，但也引入了新的算法挑战。速度不变的表示方法可以缓解在慢速运动下的观测缺失和在快速运动下的运动模糊，但大多数方法通过将事件转换为图像类表示而丢弃了时间信息。我们提出了一种名为中心主动感受野集合（Set of Centre Active Receptive Fields, SCARF）的实时速度不变表示，该方法在保留原始事件的同时，能够一致地处理快速运动、静止场景和独立移动的物体。SCARF在计算效率和表示质量方面均达到了最先进的性能。

View on arXiv Download PDF AI Translation

cs.CV / 48 / 2609.19990

QCPruner: Query-Conditioned Population Coverage for Visual Token Pruning

QCPruner：基于查询条件的人口覆盖率用于视觉令牌剪枝

He, Shengli, Liang, Yongchao, He, Roumeng, Zeng, Junjie, He, Jiyuan, Wu, Can, Zheng, Li

Abstract

The high visual-token load in multimodal large language models (MLLMs) motivates training-free pruning to reduce later-layer computation, but under a fixed budget, pruning must preserve query-relevant evidence while avoiding redundancy. Existing methods rank tokens, diversify selected subsets, or optimize coverage without using a shared per-visual query utility to weight both visual targets and candidate representatives. We introduce QCPruner, which makes both roles query-conditioned through bilateral utility weighting. Using keyword-matched query anchors, QCPruner fuses two cross-modal cues into utility and applies it to both visual targets and candidate representatives within visual-affinity-based coverage. The resulting nonnegative facility-location objective is monotone and submodular, retains the standard (1-1/e) greedy guarantee, and requires no model training or parameter updates. Across LLaVA-1.5, LLaVA-NeXT, LLaVA-Video, and Qwen2.5-VL, QCPruner achieves the highest average relative performance among evaluated complete-system pruning methods at every reported token budget. At 32 of 576 tokens on LLaVA-1.5-7B, it retains 96.1% of unpruned performance, versus 93.9% for the strongest evaluated baseline. At 256 of 1296 tokens on Qwen2.5-VL-7B, the corresponding values are 96.7% and 92.5%.

Chinese Translation

多模态大语言模型（MLLMs）中高视觉令牌负载促使无训练剪枝以减少后层计算，但在固定预算下，剪枝必须保留与查询相关的证据，同时避免冗余。现有方法对令牌进行排名，选择多样化的子集，或优化覆盖率，而不使用共享的每个视觉查询效用来加权视觉目标和候选代表。我们提出了QCPruner，通过双向效用加权使这两种角色与查询条件相关。使用关键字匹配的查询锚点，QCPruner将两种跨模态线索融合为效用，并将其应用于视觉亲和基础上的覆盖中的视觉目标和候选代表。最终得到的非负设施位置目标是单调的和次模的，保留了标准的（1-1/e）贪婪保证，并且不需要模型训练或参数更新。在LLaVA-1.5、LLaVA-NeXT、LLaVA-Video和Qwen2.5-VL的评估中，QCPruner在每个报告的令牌预算中实现了评估的完整系统剪枝方法中最高的平均相对性能。在LLaVA-1.5-7B的576个令牌中，它保留了96.1%的未剪枝性能，而最强评估基线为93.9%。在Qwen2.5-VL-7B的1296个令牌中，相应的值为96.7%和92.5%。

View on arXiv Download PDF AI Translation

cs.CV / 49 / 2609.19991

AVTrace: Diagnosing Audio-Visual Temporal Reasoning in Omni Models

AVTrace：诊断全能模型中的视听时间推理

Zhang, Longyin, Mahendra, Parth Sakhare, Wei, Chengwei, Zhang, Ning, Chong, Lim Ming, He, Sirui, Aw, Ai Ti

Abstract

Omni models can describe video content, but can they locate events in time, preserve event order, and judge audio-visual synchronization? We introduce AVTrace (Audio-Visual Temporal Reasoning Assessment and Capability Evaluation), a silver-standard diagnostic suite spanning onset and span grounding, synchronization, next-step prediction, cross-modal localization, chain parsing, and event-conditioned comprehension. It contains 34,114 training examples and category-balanced development and test splits of 3,500 and 7,000 examples. We evaluate five open omni models under their respective input configurations using reference-blind response normalization followed by deterministic scoring. All five off-the-shelf systems score below the test split's majority-label baseline of 0.556 on synchronization verification, and obtain low scores on chain parsing and event-conditioned grounding and comprehension. Development-set perturbations reveal task-dependent sensitivity in Qwen3-Omni-30B to modality removal and changes in visual input processing, without isolating their underlying causes. Parameter-efficient temporal post-training improves Gemma4-E4B-it on several benchmark metrics. On three external image benchmarks, task metrics change modestly, including some degradations, while teacher-forcing perplexity decreases. Together, these findings show that semantic reference-text overlap should not be treated as a proxy for temporal localization, and that AVTrace can identify task-specific weaknesses while providing a testbed for temporal post-training.

Chinese Translation

全能模型能够描述视频内容，但它们能否定位事件的时间、保持事件顺序并判断视听同步？我们介绍了AVTrace（视听时间推理评估与能力评估），这是一个涵盖事件开始和持续时间定位、同步、下一步预测、跨模态定位、链解析和事件条件理解的银标准诊断套件。它包含34,114个训练示例，以及3,500和7,000个示例的类别平衡开发集和测试集。我们在各自的输入配置下评估五个开放的全能模型，采用参考盲响应归一化，随后进行确定性评分。所有五个现成系统在同步验证上得分均低于测试集的多数标签基线0.556，并且在链解析和事件条件的定位与理解上得分较低。开发集的扰动揭示了Qwen3-Omni-30B对模态移除和视觉输入处理变化的任务依赖性敏感性，但未能孤立其潜在原因。参数高效的时间后训练在多个基准指标上改善了Gemma4-E4B-it。在三个外部图像基准上，任务指标变化适度，包括一些降级，而教师强制困惑度降低。综合来看，这些发现表明，语义参考文本的重叠不应被视为时间定位的代理，并且AVTrace能够识别特定任务的弱点，同时为时间后训练提供测试平台。

View on arXiv Download PDF AI Translation

cs.CV / 50 / 2609.20012

GRF-Recon: Global Ray-Field Optimization for Long-Sequence Feed-forward Reconstruction

GRF-Recon：用于长序列前馈重建的全局光线场优化

Li, Enpeng, Zhang, Yunzhou, Zhang, Zhiyao, Lyu, Dexuan, Wang, Chenyu, Cui, Chiyuan, Cheng, Cheng

Abstract

Feed-forward 3D reconstruction provides an efficient paradigm for scene modeling from image sequences. Scaling these models to large monocular scenarios are constrained by excessive GPU memory footprint, degraded local geometry, and long-term trajectory drift. Existing chunk-based optimization strategies provide limited geometric constraints and fail to maintain global consistency over extended trajectories. We present a unified framework for stable and scalable feed-forward 3D reconstruction from long monocular sequences. Our approach builds on coarse-to-fine trajectory alignment augmented by lightweight geometric prior injection. Distilling monocular geometric cues into the feed-forward backbone via LoRA adaptation improves depth accuracy on fine structures while preserving inference efficiency. We introduce a hybrid-weight sparse ray-field optimization that leverages high-frequency geometric features to guide local point-cloud refinement and enforce consistent inter-frame ray constraints. Unlike prior chunk-based methods, this establishes strong cross-frame geometric coupling while maintaining scalability. Finally, an efficient trajectory stitching strategy with joint ray-error optimization explicitly reduces accumulated drift. Extensive experiments show that our approach achieves competitive trajectory accuracy compared with representative SLAM systems, while maintaining globally consistent 3D reconstruction in large-scale scenarios.

Chinese Translation

前馈三维重建为从图像序列中建模场景提供了一种高效的范式。然而，将这些模型扩展到大型单目场景受到过高的GPU内存占用、局部几何形状退化和长期轨迹漂移的限制。现有的基于块的优化策略提供的几何约束有限，无法在扩展轨迹上保持全局一致性。我们提出了一个统一框架，用于从长单目序列中进行稳定且可扩展的前馈三维重建。我们的方法基于粗到细的轨迹对齐，并通过轻量级几何先验注入进行增强。通过LoRA适应将单目几何线索提炼到前馈骨干网络中，提高了细结构的深度精度，同时保持了推理效率。我们引入了一种混合权重稀疏光线场优化方法，利用高频几何特征指导局部点云细化，并强制执行一致的帧间光线约束。与之前的基于块的方法不同，这种方法在保持可扩展性的同时建立了强大的跨帧几何耦合。最后，我们采用了一种高效的轨迹拼接策略，并通过联合光线误差优化显著减少了累积漂移。大量实验表明，我们的方法在与代表性SLAM系统的轨迹精度比较中表现出竞争力，同时在大规模场景中保持全局一致的三维重建。

View on arXiv Download PDF AI Translation

cs.CV / 51 / 2609.20034

Astronex-World 1.0: Real-Time Interactive World Model Foundation

Astronex-World 1.0：实时互动世界模型基础

Zhou, Xin, Miao, Cong

Abstract

We present Astronex-World 1.0, an open controllable video world-model foundation. Given a text prompt (text-to-video) or an initial observation (image-to-video), the model predicts future visual states under frame-aligned camera trajectories, continuous actions, and an embodiment identifier, and accepts text events inserted at a specified position of a rollout. The family provides a bidirectional model for full-context generation and a causal model with block-causal attention and cross-block KV caching for persistent generation, both built on the Wan2.2-TI2V-5B prior. PRoPE injects camera intrinsics and extrinsics, while a 64-dimensional action stream modulates every Transformer layer. A five-stage training path develops bidirectional camera and action control, converts the backbone to block-causal generation, distills a few-step student, restores mixed-domain dynamics, and applies asymmetric DMD/DMD2 distribution matching. The causal model generates 832x480 video at 24 fps. All five training stages run on two NVIDIA L20 48 GB GPUs, and the causal model streams in real time on one. It scores 73.5 on WBench Navi and 70.0 on WBench Full. On Full, this 5B model is above the 13.6B LongCat-Video and the 14B Helios, within one point of the 22B LTX-2.3, and above YUME 1.5, which is post-trained from the same 5B prior on NVIDIA A100 GPUs. The reserved action input and output interfaces allow post-training for embodied intelligence and autonomous driving.

Chinese Translation

我们提出了Astronex-World 1.0，这是一个开放可控的视频世界模型基础。该模型根据文本提示（文本到视频）或初始观察（图像到视频），在帧对齐的相机轨迹、连续动作和具身标识符下预测未来的视觉状态，并接受在指定位置插入的文本事件。该系列提供了一个双向模型用于全上下文生成，以及一个具有块因果注意力和跨块KV缓存的因果模型用于持久生成，二者均基于Wan2.2-TI2V-5B先验。PRoPE注入了相机的内参和外参，同时一个64维的动作流调制每个Transformer层。五阶段的训练路径发展了双向相机和动作控制，将主干转换为块因果生成，提炼出几步学生模型，恢复混合域动态，并应用非对称DMD/DMD2分布匹配。因果模型以24帧每秒生成832x480的视频。所有五个训练阶段在两台NVIDIA L20 48 GB GPU上运行，而因果模型在一台GPU上实时流式传输。它在WBench Navi上得分73.5，在WBench Full上得分70.0。在Full上，这个5B模型的表现超过了13.6B的LongCat-Video和14B的Helios，接近22B的LTX-2.3，并且高于同样在NVIDIA A100 GPU上从相同5B先验后训练的YUME 1.5。保留的动作输入和输出接口允许进行具身智能和自主驾驶的后续训练。

View on arXiv Download PDF AI Translation

cs.CV / 52 / 2609.20064

A Free Lunch? Adapting PP-OCRv6 for Historical Text Recognition

免费的午餐？将 PP-OCRv6 适配于历史文本识别

Kiessling, Benjamin

Abstract

Despite impressive reported scores, large vision-language models have seen limited practical uptake in historical automatic text recognition because of their computational cost, dependence on large-scale pretraining, and hallucination. Historical ATR therefore continues to rely largely on compact CRNN line recognizers, which are visually grounded and trainable on modest data. Lightweight recurrence-free recognizers promise the accuracy of larger models with the practical advantages of CRNNs, yet have not been comprehensively evaluated on historical writing. We adapt PP-OCRv6, a recent compact text recognizer without strong language modeling, for historical line recognition and compare it with a conventional CRNN across generalized pretraining, domain-specific training, corpus-level fine-tuning, and manuscript-specific few-shot adaptation on multilingual Latin- and Arabic-script material. While PP-OCRv6 does not consistently outperform the baseline when trained from scratch, heterogeneous pretraining produces markedly better generalization. Comparisons with the Qwen3.5-based Medusa recognizer further show that fine-tuned PP-OCRv6 can outperform a large VLM tailored towards historical Latin-script HTR.

Chinese Translation

尽管报告的成绩令人印象深刻，但由于计算成本、对大规模预训练的依赖以及幻觉现象，大型视觉语言模型在历史自动文本识别中的实际应用仍然有限。因此，历史自动文本识别仍然主要依赖于紧凑的 CRNN 行识别器，这些识别器在视觉上是有依据的，并且可以在适度的数据上进行训练。轻量级的无递归识别器承诺提供与更大模型相当的准确性，同时具备 CRNN 的实际优势，但在历史书写方面尚未得到全面评估。我们将 PP-OCRv6（一个近期的紧凑文本识别器，缺乏强大的语言建模）适配于历史行识别，并与传统的 CRNN 进行比较，涵盖了广义预训练、领域特定训练、语料库级微调以及针对多语言拉丁文和阿拉伯文材料的手稿特定少量样本适应。尽管 PP-OCRv6 在从头训练时并不总是优于基线，但异构预训练显著提高了泛化能力。与基于 Qwen3.5 的 Medusa 识别器的比较进一步表明，经过微调的 PP-OCRv6 可以超越针对历史拉丁文 HTR 的大型视觉语言模型。

View on arXiv Download PDF AI Translation

cs.CV / 53 / 2609.20066

PointEvent: Rethinking Event-based Tiny Object Detection via Serialized Motion Evidence Accumulation

PointEvent：通过序列化运动证据累积重新思考基于事件的微小物体检测

Wu, Zongze, Jia, Baofeng, Yan, Weiqi, Zhang, Jingyuan, Zang, Yu, Chen, Xiaoyu, Han, Jing

Abstract

Event cameras offer high temporal resolution and motion sensitivity for tiny UAV detection, yet distant targets generate sparse and fragmented events that are easily overwhelmed by clutter and ego-motion. Existing methods mainly rely on dense event representations or local sparse spatiotemporal modeling, resulting in redundant computation or fragmented modeling of motion continuity across distant asynchronous events. To address this limitation, we introduce serialized motion evidence accumulation, which treats motion continuity as an ordered evidence propagation process. Specifically, the same event stream is organized into locality-preserving spatiotemporal paths and chronology-preserving temporal paths through the latent complementary serializations. Based on this principle, we propose PointEvent, a lightweight event-wise state-space framework that alternates serialized scans across the complementary orders, progressively consolidating fragmented motion evidence beyond fixed local neighborhoods. A high-resolution event branch preserves fine-grained target responses, while compact context modulation suppresses interference. Experiments demonstrate that PointEvent achieves SOTA with the fewest parameters and fastest measured inference among the compared methods. Code: https://github.com/wzz-z/PointEvent

Chinese Translation

事件相机为微小无人机检测提供了高时间分辨率和运动敏感性，但远处目标产生的事件稀疏且碎片化，容易被杂波和自我运动所淹没。现有方法主要依赖于密集事件表示或局部稀疏时空建模，导致在远距离异步事件中运动连续性的冗余计算或碎片化建模。为了解决这一限制，我们引入了序列化运动证据累积，将运动连续性视为有序的证据传播过程。具体而言，相同的事件流通过潜在的互补序列化被组织成保持局部性的时空路径和保持时间顺序的时间路径。基于这一原理，我们提出了PointEvent，一个轻量级的事件状态空间框架，它在互补顺序之间交替进行序列化扫描，逐步整合超出固定局部邻域的碎片化运动证据。高分辨率事件分支保留了细粒度的目标响应，而紧凑的上下文调制则抑制了干扰。实验表明，PointEvent在比较方法中以最少的参数和最快的推理速度达到了最先进的水平（SOTA）。代码：https://github.com/wzz-z/PointEvent

View on arXiv Download PDF AI Translation

cs.CV / 54 / 2609.20088

G^2RA-NET: Graph-based Cross-Slice Relation Modeling with Attention Gating for Medical Image Segmentation

G^2RA-NET：基于图的跨切片关系建模与注意力门控在医学图像分割中的应用

Wang, Shengye, Wu, Zonglin, Fan, Liang, Xue, Yule, Zhao, Haozhe

Abstract

Medical image segmentation supports quantitative clinical analysis and computer-aided diagnosis. Recent methods for medical image segmentation have improved both local feature representation and volumetric context modeling. However, existing methods still strug- gle to efficiently model cross-slice relations in anisotropic volumet- ric images, limiting segmentation consistency and accuracy. This pa- per proposes G^2RA-Net, a medical image segmentation framework that combines graph-based cross-slice relation modeling with atten- tion gating. Graph-Based Slice Relationship Modeling (GSRM) cap- tures anatomical dependencies across consecutive slices by repre- senting each slice as a graph node and propagating semantic con- text through graph message passing. The Cross-Slice Attention Gate (CSAG) then selects relevant neighboring context and emphasizes target anatomical regions through attention-guided feature modula- tion. Experiments on brain MRI and abdominal CT datasets demon- strate that G^2RA-Net outperforms representative methods in seg- mentation accuracy and boundary quality. Ablation studies further validate the proposed design.

Chinese Translation

医学图像分割支持定量临床分析和计算机辅助诊断。近年来，医学图像分割的方法在局部特征表示和体积上下文建模方面都有所改进。然而，现有方法在有效建模各向异性体积图像中的跨切片关系方面仍然存在困难，从而限制了分割的一致性和准确性。本文提出了G^2RA-Net，一种结合基于图的跨切片关系建模与注意力门控的医学图像分割框架。基于图的切片关系建模（GSRM）通过将每个切片表示为图节点并通过图消息传递传播语义上下文，从而捕捉连续切片之间的解剖依赖关系。跨切片注意力门（CSAG）则选择相关的邻近上下文，并通过注意力引导的特征调制强调目标解剖区域。在脑部MRI和腹部CT数据集上的实验表明，G^2RA-Net在分割准确性和边界质量方面优于代表性方法。消融研究进一步验证了所提设计的有效性。

View on arXiv Download PDF AI Translation

cs.CV / 55 / 2609.20100

A Smaller Transformer in Your Transformer

您模型中的更小变换器

Tomar, Dhananjay, Aasan, Marius, Kleppe, Andreas, Rivera, Adín Ramírez

Abstract

Recent findings indicate that Vision Transformers settle into locally similar computational phases, implying a level of depthwise computational redundancy. However, existing methods to exploit this redundancy either fail to reduce inference compute or severely degrade model expressivity. In this work, we formalise a unified view of block redundancy that decouples the geometry from specific surrogate interventions. We then introduce Transformer-Within-Transformer (TWT), a post-hoc method that fuses contiguous groups of redundant layers into a single learned surrogate layer. TWT reduces parameter count and inference compute while remaining competitive with original models using half the depth on natural images, and in several downstream histopathology settings, TWT matches or even improves on the original baseline.

Chinese Translation

最近的研究发现，视觉变换器（Vision Transformers）在局部相似的计算阶段中稳定下来，这暗示了深度计算冗余的存在。然而，现有的方法要么无法减少推理计算，要么严重降低模型的表现力。在本研究中，我们形式化了一种统一的块冗余视角，将几何结构与特定的替代干预解耦。随后，我们引入了变换器内变换器（Transformer-Within-Transformer，TWT），这是一种后处理方法，将连续的冗余层组融合为一个单一的学习替代层。TWT在减少参数数量和推理计算的同时，仍然与原始模型保持竞争力，在自然图像上使用一半的深度，并且在多个下游组织病理学设置中，TWT的表现与原始基线相匹配甚至有所提升。

View on arXiv Download PDF AI Translation

cs.CV / 56 / 2609.20106

AnyviewMeter: Adapting Robotic Reward Models with Camera Geometry and Multi-View Attention

AnyviewMeter：结合相机几何和多视角注意力的机器人奖励模型适应

Tu, Yuang, Tan, Runjia, Yan, Yujie, Hu, Jinghan, Lv, Chen

Abstract

Robotic reward models evaluate task execution from visual observations, but their predictions can change with camera viewpoint and occlusion even when the underlying task state is unchanged. Adapting a pretrained reward model to a local task therefore requires accounting for how that task is observed. We introduce AnyviewMeter, a geometry-conditioned adaptation framework for robotic reward models that represent task progress as a scalar reward signal. It combines low-rank fine-tuning with token-aligned Plucker rays and synchronous block attention: ray conditioning incorporates camera geometry into visual features and attention queries and keys, while block attention fuses synchronized views inside the pretrained decoder. The framework supports both single-view reward prediction and joint multi-view evaluation through parameter-efficient adaptation of a pretrained Robometer model. On PickCube, single-view adaptation improves progress prediction in every camera group and reduces mean absolute error under a changed field of view by approximately 21% relative to RGB fine-tuning. Across simulated manipulation tasks, joint multi-view prediction reduces progress error by 41-69% compared with averaging single-view RGB predictions and improves temporal ordering in approximately 88% of task-camera groups. On real tasks with fixed and wrist-mounted cameras, mean absolute error decreases by approximately 21% relative to averaged RGB fine-tuning. These results support camera geometry and joint visual evidence as useful components of task-specific robotic reward adaptation.

Chinese Translation

机器人奖励模型通过视觉观察评估任务执行，但即使在基础任务状态不变的情况下，其预测也会因相机视角和遮挡而变化。因此，将预训练的奖励模型适应于本地任务需要考虑该任务的观察方式。我们提出了AnyviewMeter，这是一个几何条件适应框架，用于将任务进展表示为标量奖励信号的机器人奖励模型。该框架结合了低秩微调、与标记对齐的Plucker光线和同步块注意力：光线条件将相机几何信息融入视觉特征以及注意力查询和键，而块注意力则在预训练解码器内部融合同步视图。该框架支持通过高效的参数适应预训练的Robometer模型进行单视图奖励预测和联合多视图评估。在PickCube上，单视图适应在每个相机组中改善了进展预测，并在视场变化下将均方根误差减少了约21%，相较于RGB微调。在模拟操作任务中，联合多视图预测将进展误差减少了41-69%，与单视图RGB预测的平均值相比，并在约88%的任务-相机组中改善了时间顺序。在固定和手腕安装相机的真实任务中，均方根误差相较于平均RGB微调减少了约21%。这些结果支持相机几何和联合视觉证据作为任务特定机器人奖励适应的有用组成部分。

View on arXiv Download PDF AI Translation

cs.CV / 57 / 2609.20139

Cross-Modal Attention Acts as a Frequency Filter: Why Verbose Prompts Improve Robustness in Vision-Language Models

跨模态注意力作为频率滤波器：冗长提示为何提高视觉-语言模型的鲁棒性

Wani, Farooq Ahmad, Bucarelli, Maria Sofia, Mirza, Mujtaba Hussain, Pryymak, Oleksandr, Gema, Aryo Pradipta, Masi, Iacopo, Minervini, Pasquale, Silvestri, Fabrizio

Abstract

Vision-language models (VLMs) are fragile under image corruption. We find that the wording of the question affects VLMs in two opposite ways. Verbose questions make VLMs substantially more robust---e.g., rephrasing "Is there a cat?" into "Please look carefully and answer: is there a cat?". Conversely, VLMs become more fragile under corruption when the question is semantically complex or finer-grained, e.g., "what colour is the cup left of the chair?" instead of "is there a cup?". Both effects stem from question-conditioned cross-modal attention, which induces a spectral filter over image patches: verbose questions broaden its frequency support, while fine-grained questions concentrate it onto fewer visual scales. The model's answer drifts most when this filter and the corruption sit on the same spatial frequencies. We test the filter view on Qwen3-VL and LLaVA-OneVision across GQA and CLEVR; verbose paraphrasing reduces drift variance by 70--81% on the 8B models. The practical recipe---pad the prompt---further yields measurable gains in accuracy, even under image corruption.

Chinese Translation

视觉-语言模型（VLMs）在图像损坏时表现脆弱。我们发现问题的措辞对VLMs有两种相反的影响。冗长的问题使VLMs显著更具鲁棒性——例如，将“有猫吗？”改写为“请仔细观察并回答：有猫吗？”。相反，当问题在语义上复杂或更细致时，VLMs在损坏下变得更加脆弱，例如，“椅子左边的杯子是什么颜色？”而不是“有杯子吗？”。这两种效应源于基于问题的跨模态注意力，它在图像块上引入了一种频谱滤波器：冗长的问题扩展了其频率支持，而细粒度的问题则将其集中在更少的视觉尺度上。当这个滤波器和损坏位于相同的空间频率时，模型的答案漂移最为明显。我们在Qwen3-VL和LLaVA-OneVision上测试了这一滤波器视角，涵盖GQA和CLEVR；冗长的改写在8B模型上将漂移方差降低了70%至81%。这一实用方法——填充提示——在图像损坏情况下进一步带来了可测量的准确性提升。

View on arXiv Download PDF AI Translation

cs.CV / 58 / 2609.20147

Bridging Modalities on the Cortex: Surface-based MRI to PET Translation with a Diffusion Bridge

皮层模态桥接：基于表面的MRI到PET的转化与扩散桥

Li, Yitong, Samoylova, Alexandra, Bongratz, Fabian, Grimmer, Timo, Hedderich, Dennis M., Yakushev, Igor, Wachinger, Christian

Abstract

Cortical hypometabolism measured by Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) is a highly sensitive biomarker for dementia diagnosis. However, high costs, radiation exposure, and limited accessibility constrain its clinical utility. While cross-modal synthesis from Magnetic Resonance Imaging (MRI) offers a promising alternative, existing volumetric generation methods do not explicitly account for the highly folded cortical geometry, where disease-related patterns predominantly reside. To address this, we introduce a novel surface-based diffusion bridge framework DB-SUiT for MRI-to-PET translation that operates natively on the cortical manifold. A conditional Spherical U-shaped vision Transformer (SUiT) is specifically designed to model the intricate cross-modal relationships while preserving surface topology. It combines spherical convolutional encoders for multi-scale surface feature extraction with bottleneck Transformers to capture long-range spatial dependencies, while incorporating demographic and subcortical conditions to refine the synthesis. Evaluated on two datasets, including subjects with different dementia types, DB-SUiT demonstrates high-fidelity synthesis that substantially outperforms other baselines. In automated dementia classification, synthesized PET surfaces improve performance over MRI by 14.2% and PET volumes by 11.3%, approaching the performance of real PET surfaces. In a blinded reader study, synthetic PET achieved 85.5% diagnostic accuracy, compared with 75.8% for MRI and 95.2% for real PET. This further demonstrates cross-cohort and cross-pathology generalization, as the model was evaluated without retraining on an external cohort that included a dementia subtype not represented during training. Our code is available at https://github.com/ai-med/DB-SUiT.

Chinese Translation

通过氟脱氧葡萄糖正电子发射断层扫描（FDG-PET）测量的皮层低代谢是痴呆诊断的高度敏感生物标志物。然而，高昂的费用、辐射暴露和有限的可及性限制了其临床应用。尽管来自磁共振成像（MRI）的跨模态合成提供了一种有前景的替代方案，但现有的体积生成方法并未明确考虑高度折叠的皮层几何结构，而疾病相关模式主要存在于此。为了解决这一问题，我们提出了一种新颖的基于表面的扩散桥框架DB-SUiT，用于在皮层流形上进行MRI到PET的转化。条件球形U形视觉变换器（SUiT）被专门设计用于建模复杂的跨模态关系，同时保持表面拓扑。它结合了球形卷积编码器用于多尺度表面特征提取，以及瓶颈变换器以捕捉长距离空间依赖，同时结合人口统计学和皮层下条件以细化合成。在两个数据集上进行评估，包括不同类型痴呆症的受试者，DB-SUiT展示了高保真度的合成，显著优于其他基线。在自动化痴呆分类中，合成的PET表面比MRI提高了14.2%的性能，比PET体积提高了11.3%，接近真实PET表面的性能。在盲读者研究中，合成的PET达到了85.5%的诊断准确率，而MRI为75.8%，真实PET为95.2%。这进一步证明了跨队列和跨病理的泛化能力，因为该模型在未进行重新训练的外部队列上进行了评估，该队列包括在训练期间未代表的痴呆亚型。我们的代码可在https://github.com/ai-med/DB-SUiT获取。

View on arXiv Download PDF AI Translation

cs.CV / 59 / 2609.20150

Task-Oriented Semantic Feature Transmission for Multi-Task Satellite Remote Sensing over Low-SNR Channels

面向任务的语义特征传输在低信噪比信道上的多任务卫星遥感

Sun, Shuoyuan, Wang, Hongyu, Peng, Mugen, Xu, Wenjia

Abstract

Conventional satellite remote sensing transmission follows a reconstruct-then-infer paradigm that optimizes pixel-level fidelity, creating an objective mismatch with downstream tasks such as classification and detection, especially at low SNR. This paper investigates a task-oriented framework that bypasses image reconstruction and directly transmits semantic features extracted by a multitask-pretrained backbone. A lightweight channel adaptation module (CAM) compresses feature dimensionality for bandwidth reduction, and a feature restorer recovers task-relevant structure after channel corruption. With the backbone frozen, the CAM and task-specific downstream heads are jointly optimized with task and feature-level supervision under random-SNR training. Under the adopted AWGN setting, experiments on scene classification and object detection show consistent gains over reconstruction-oriented JSCC baselines across different SNR conditions, with the largest improvements in the low-SNR regime.

Chinese Translation

传统的卫星遥感传输遵循重建后推断的范式，优化像素级的保真度，这与下游任务（如分类和检测）存在客观不匹配，尤其是在低信噪比条件下。本文研究了一种面向任务的框架，该框架绕过图像重建，直接传输由多任务预训练骨干网络提取的语义特征。轻量级通道适配模块（Channel Adaptation Module, CAM）压缩特征维度以减少带宽，而特征恢复器在信道损坏后恢复与任务相关的结构。在冻结骨干网络的情况下，CAM和特定任务的下游头部在随机信噪比训练下共同优化，采用任务和特征级的监督。在所采用的加性高斯白噪声（AWGN）设置下，场景分类和目标检测的实验显示，在不同信噪比条件下，相较于以重建为导向的联合源信道编码（JSCC）基线，取得了一致的性能提升，尤其是在低信噪比范围内的改善最为显著。

View on arXiv Download PDF AI Translation

cs.CV / 60 / 2609.20151

Ischemic Stroke Segmentation and Net Water Uptake Quantification on Multicenter Non-Contrast CT Using Supervised Target-Domain Adaptation

基于监督目标领域适应的多中心无对比CT缺血性中风分割与净水摄取量定量分析

Britt, Linus, Nielsen, Maximilian, Klapproth, Susan, Kemmling, Andre, Lev, Michael H., Broocks, Gabriel, Werner, Rene, Sentker, Thilo

Abstract

Objectives: Quantitative assessment of infarct hypodensity on non-contrast computed tomography (NCCT), including net water uptake (NWU), requires manual or semi-manual lesion delineation, often guided by CT perfusion or diffusion-weighted MRI, limiting clinical applicability. Automated segmentation on NCCT could enable efficient biomarker extraction such as NWU but remains challenging across heterogeneous multicenter data. This study aimed to develop and externally test a domain-aware deep learning framework for ischemic stroke segmentation on NCCT and assess its suitability for NWU quantification. Materials & Methods: In this retrospective multicenter study of 801 patients from four datasets, an nnU-Net-based model was trained on NCCT scans from the University Medical Center Hamburg-Eppendorf and the Acute Ischemic Stroke Dataset. To adapt to new domains, the model was fine-tuned on target-domain subsets from Boston (n=11) and ISLES (n=75), with evaluation on held-out cases not used for fine-tuning. Automated segmentations and NWU values were compared with expert references. Results: For lesions $\geq$ 30 mL, median Dice was 0.68 (Boston) and 0.56 (ISLES). Including smaller lesions, which predominated in ISLES, median Dice was 0.54 (interquartile range [IQR] 0.30-0.70) for acute lesion segmentation (Boston dataset) and 0.20 (IQR 0.03-0.41) for NCCT lesion segmentations when compared to post-treatment infarct (primary target of the ISLES challenge). Automated NWU mean absolute error was 1.37 percentage points (SD 1.61, Boston). Conclusion: Target-domain adaptation supported NCCT-only infarct segmentation across heterogeneous external cohorts, although performance varied across domains. The approach enabled low-error NWU quantification from baseline NCCT without advanced imaging, supporting further prospective clinical evaluation.

Chinese Translation

目的：在无对比计算机断层扫描（NCCT）上定量评估梗死低密度，包括净水摄取量（NWU），通常需要手动或半手动的病灶描绘，常常依赖于CT灌注或扩散加权MRI，这限制了临床应用。NCCT上的自动分割可以实现高效的生物标志物提取，例如NWU，但在异质的多中心数据中仍然具有挑战性。本研究旨在开发并外部测试一个面向领域的深度学习框架，用于在NCCT上进行缺血性中风分割，并评估其在NWU定量中的适用性。材料与方法：在这项回顾性多中心研究中，共纳入801名患者，数据来自四个数据集，基于nnU-Net模型在汉堡-埃彭多夫大学医学中心和急性缺血性中风数据集的NCCT扫描上进行训练。为了适应新领域，该模型在来自波士顿（n=11）和ISLES（n=75）的目标领域子集上进行了微调，并在未用于微调的保留案例上进行评估。自动分割和NWU值与专家参考进行了比较。结果：对于病灶体积≥30 mL，波士顿的中位Dice系数为0.68，ISLES为0.56。包括较小病灶（在ISLES中占主导地位），急性病灶分割的中位Dice系数为0.54（四分位数范围[IQR] 0.30-0.70，波士顿数据集），而与治疗后梗死（ISLES挑战的主要目标）相比，NCCT病灶分割的中位Dice系数为0.20（IQR 0.03-0.41）。自动NWU的平均绝对误差为1.37个百分点（标准差1.61，波士顿）。结论：目标领域适应支持在异质外部队列中仅基于NCCT的梗死分割，尽管不同领域的性能有所差异。该方法能够从基线NCCT中实现低误差的NWU定量，无需先进成像，支持进一步的前瞻性临床评估。

View on arXiv Download PDF AI Translation

cs.CV / 61 / 2609.20160

Needles in a Raystack: Ultra-Sparse LiDAR Occupancy Detection for Bat Tracks

光束堆中的针：超稀疏LiDAR占用检测用于蝙蝠轨迹

Klar, Nico, Rana, Pankaj, Gifary, Nizam, Traub, Jakob, Ahmad, Aamir

Abstract

Monitoring flying animals is important for understanding and protecting biodiversity, but nocturnal species such as bats are difficult to observe in the field. Using LiDAR, bat movements at night result in ultra-sparse 3D spatio-temporal data in which standard reconstruction losses tend to predict only background and miss real flight paths. We study this problem as voxel-wise occupancy detection in sensor-centric LiDAR raystacks. A lightweight 3D U-Net is proposed that preserves temporal resolution, uses skip connections for spatial detail, and combines weighted binary cross-entropy with Dice loss to handle the strong class imbalance. In real LiDAR recordings of bats over open fields, cross-checked with acoustic monitoring, a reconstruction-based 3D convolutional autoencoder baseline fails to recover foreground trajectories. In contrast, the proposed U-Net recovers sparse foreground occupancy in diagnostic experiments and produces coherent occupancy patterns along bat flight trajectories, providing a practical basis for validation-scale experiments, later clustering of flight tracks, and future integration of bat activity information into biodiversity-aware turbine curtailment strategies.

Chinese Translation

监测飞行动物对于理解和保护生物多样性至关重要，但夜行性物种如蝙蝠在野外观察时非常困难。利用LiDAR技术，蝙蝠在夜间的运动产生了超稀疏的三维时空数据，其中标准重建损失往往只能预测背景而错过真实的飞行路径。我们将这一问题视为传感器中心的LiDAR光线堆中的体素级占用检测。我们提出了一种轻量级的3D U-Net模型，该模型保留了时间分辨率，使用跳跃连接以获取空间细节，并结合加权二元交叉熵与Dice损失以应对强烈的类别不平衡。在对开放领域中蝙蝠的真实LiDAR录音中，与声学监测进行交叉验证，基于重建的3D卷积自编码器基线未能恢复前景轨迹。相比之下，所提出的U-Net在诊断实验中成功恢复了稀疏的前景占用，并在蝙蝠飞行轨迹上产生了连贯的占用模式，为验证规模实验、后续的飞行轨迹聚类以及未来将蝙蝠活动信息整合到关注生物多样性的涡轮机削减策略中提供了实用基础。

View on arXiv Download PDF AI Translation

cs.CV / 62 / 2609.20178

MTF-Net: Multi-Modal Temporal Feature Fusion Network for Pedestrian Intention Prediction

MTF-Net：用于行人意图预测的多模态时间特征融合网络

Rahman, Md Mahfuzur, Zhou, Pengzhan, Noor, A. F. M. Abdun, Ahasan, Md Imam, Rahman, Md Mustafizur, Qu, Fang

Abstract

Accurately predicting pedestrian intentions is crucial for ensuring safe and proactive interaction between autonomous vehicles and pedestrians. However, existing approaches often depend on architectures that either model temporal dependencies within individual modalities or fuse modalities only at coarse semantic levels. To address these limitations, we propose MTF-Net, a novel Multi-Modal Temporal Feature Fusion Network that jointly models kinematic, appearance, and contextual cues for pedestrian intention prediction. MTF-Net integrates four complementary modalities-bounding-box dynamics, human pose keypoints, local context, and scene-level semantics within a recurrent fusion framework enhanced by gated linear units (GLUs). These GLU-based modules adaptively regulate cross-modal information flow, enabling interpretable and efficient feature interaction across temporal scales. Through three dedicated temporal encoding branches and an attention-guided fusion head, the proposed model robustly anticipates pedestrian crossing intentions several frames before they occur. Extensive evaluations on the PIE and JAAD benchmarks demonstrate that MTF-Net surpasses recent transformer- and graph-based models, achieving up to 0.95 AUC on PIE and 0.94 AUC on JAAD, while maintaining real-time performance. The results highlight that reliable pedestrian intention prediction arises from principled multi-modal fusion rather than excessive architectural complexity.

Chinese Translation

准确预测行人意图对于确保自动驾驶车辆与行人之间的安全和主动互动至关重要。然而，现有的方法往往依赖于仅在单一模态内建模时间依赖关系或仅在粗略语义层面融合模态的架构。为了解决这些局限性，我们提出了MTF-Net，一种新颖的多模态时间特征融合网络，能够联合建模运动学、外观和上下文线索以进行行人意图预测。MTF-Net在一个增强了门控线性单元（GLUs）的递归融合框架内集成了四种互补模态——边界框动态、人类姿态关键点、局部上下文和场景级语义。这些基于GLU的模块自适应地调节跨模态信息流，使得在时间尺度上实现可解释和高效的特征交互。通过三个专门的时间编码分支和一个基于注意力的融合头，所提出的模型能够在行人过马路意图发生前的多个帧内稳健地预测其意图。在PIE和JAAD基准上的广泛评估表明，MTF-Net超越了最近的基于变换器和图的模型，在PIE上达到高达0.95的AUC，在JAAD上达到0.94的AUC，同时保持实时性能。结果突显出，可靠的行人意图预测源于有原则的多模态融合，而非过度的架构复杂性。

View on arXiv Download PDF AI Translation

cs.CV / 63 / 2609.20181

MoSSGate: Memory-Modulated State-Space Gating for Skin Lesion Segmentation

MoSSGate：用于皮肤病变分割的记忆调制状态空间门控

Awan, Anum, Buriro, Mahnoor, Khan, Muhammad Younas, Ahasan, Md Imam

Abstract

Accurate skin lesion segmentation is crucial for reliable computer-aided dermatological diagnosis, yet existing convolutional and transformer-based models often struggle to jointly capture long-range spatial dependencies and fine boundary details under limited computational budgets. This trade-off between global context modeling and boundary-aware localization frequently leads to over-segmentation, fragmented predictions, or missing thin peripheral structures. To address this challenge, we propose MoSSGate, a plug-and-play module for U-Net that integrates (i) boundary-aware spatial gating to restrict long-range propagation to informative regions, (ii) an external memory modulator that provides sample-adaptive dynamic control, and (iii) parallel 2D state-space modeling for efficient global context aggregation with linear complexity. The proposed design enables adaptive, context-aware information propagation while preserving sharp and accurate lesion boundaries. Extensive experiments on the ISIC 2017 and ISIC 2018 benchmarks demonstrate state-of-the-art accuracy with strong efficiency, achieving 86.3% and 85.9% mIoU and 92.6% and 90.6% Dice, respectively, while requiring substantially fewer FLOPs than most competing CNN-based methods. These results highlight a favorable accuracy efficiency trade-off for high-resolution medical image segmentation.

Chinese Translation

准确的皮肤病变分割对于可靠的计算机辅助皮肤病诊断至关重要，然而现有的卷积和基于变换器的模型在有限的计算预算下，往往难以同时捕捉长距离空间依赖关系和细致的边界细节。这种全球上下文建模与边界感知定位之间的权衡，常常导致过度分割、碎片化预测或缺失细薄的周边结构。为了解决这一挑战，我们提出了MoSSGate，一个可插拔的U-Net模块，集成了（i）边界感知空间门控，以限制长距离传播到信息丰富的区域，（ii）外部记忆调制器，提供样本自适应的动态控制，以及（iii）并行的二维状态空间建模，以线性复杂度高效聚合全球上下文。所提出的设计使得自适应、上下文感知的信息传播成为可能，同时保持清晰和准确的病变边界。在ISIC 2017和ISIC 2018基准测试上的广泛实验表明，MoSSGate实现了最先进的准确性和强大的效率，分别达到了86.3%和85.9%的mIoU，以及92.6%和90.6%的Dice，同时所需的FLOPs显著低于大多数竞争的CNN方法。这些结果突显了高分辨率医学图像分割中良好的准确性与效率的权衡。

View on arXiv Download PDF AI Translation

cs.CV / 64 / 2609.20222

A Two-Stage Multi-Scale Attention-Based Network for Weakly Supervised Cataract Fundus Image Enhancement

一种基于双阶段多尺度注意力的弱监督白内障眼底图像增强网络

Fang, Xiaoyong, Wang, Yue, Li, Xiangyu, Fan, Wanshu, Zhou, Dongsheng

Abstract

Cataract is a major cause of vision loss and hinders further diagnosis. However, cataract fundus image enhancement often grapples with challenges such as limited paired cataract retinal images and insufficient recovery of fine details in the retinal images. To mitigate these challenges, we in this paper propose a two-stage multi-scale attention-based network (TSMSA-Net) for weakly supervised cataract fundus image enhancement. Our TSMSA-Net leverages the domain transformation to synthesis paired real-like cataract images, solving the problem of difficult acquisition of paired images. To further extract detailed information from fundus images and reduce the generation of artifacts during the enhancement process, we propose a multi-scale attention-based stage to learn more useful features for cataract image enhancement. Experimental results on Kaggle and ODIR-5K demonstrate that our TSMSA-Net outperforms current state-of-the-art cataract fundus images enhancement even without paired images and exhibits certain generalization ability. Experimental results on Kaggle and ODIR-5K datasets indicate that our TSMSA-Net outperforms the current state-of-the-art methods for cataract fundus image enhancement, even in the absence of paired images. Additionally, it demonstrates a certain level of generalization capability. The enhancement also can improve the performance of vessel segmentation and classification in cataract images.

Chinese Translation

白内障是导致视力丧失的主要原因，阻碍了进一步的诊断。然而，白内障眼底图像增强常常面临诸如配对白内障视网膜图像有限和细节恢复不足等挑战。为了解决这些问题，本文提出了一种基于双阶段多尺度注意力的网络（TSMSA-Net）用于弱监督白内障眼底图像增强。我们的TSMSA-Net利用领域转换合成配对的类真实白内障图像，解决了配对图像获取困难的问题。为了进一步从眼底图像中提取详细信息并减少增强过程中伪影的生成，我们提出了一个基于多尺度注意力的阶段，以学习更多有用的特征用于白内障图像增强。在Kaggle和ODIR-5K数据集上的实验结果表明，我们的TSMSA-Net在没有配对图像的情况下，仍然超越了当前最先进的白内障眼底图像增强方法，并展现出一定的泛化能力。此外，增强过程还可以提高白内障图像中血管分割和分类的性能。

View on arXiv Download PDF AI Translation

cs.CV / 65 / 2609.20235

Scene-Q: Confidence-Aware Coarse-to-Fine Querying of 3D Scenes with Selective VLM Reasoning

Scene-Q：基于信心的粗到细3D场景查询与选择性VLM推理

Kim, Juno, Park, Yesol, Yoon, Hye-Jung, Zhang, Byoung-Tak

Abstract

Indoor mobile robots require open-vocabulary scene understanding that grounds natural-language queries in a consistent 3D map. Many existing systems ultimately rely on cosine-similarity retrieval with contrastive image--text encoders, which is efficient but brittle when labels are near-synonymous or multiple similar instances appear. We present Scene-Q, a confidence-aware coarse-to-fine querying framework that normalizes encoder scores with temperature scaling and selectively invokes a reasoning VLM only for low-confidence cases. High-confidence queries are answered by fast retrieval, while ambiguous ones are reranked over a small top-K candidate set using the original multi-view images and instance bounding boxes, enabling context-aware disambiguation at low cost. Scene-Q improves open-vocabulary 3D instance segmentation on ScanNet200 and natural-language 3D instance retrieval on real-world reconstructions, with the largest gains on spatial and relational queries while keeping a substantial fraction of queries on the fast path.

Chinese Translation

室内移动机器人需要开放词汇的场景理解，将自然语言查询与一致的3D地图相结合。许多现有系统最终依赖于对比图像-文本编码器的余弦相似度检索，这种方法高效但在标签近义或出现多个相似实例时表现脆弱。我们提出了Scene-Q，这是一种基于信心的粗到细查询框架，通过温度缩放对编码器得分进行归一化，并仅在低信心情况下选择性地调用推理VLM。高信心查询通过快速检索得到回答，而模糊查询则在小的前K候选集中使用原始多视图图像和实例边界框进行重新排序，从而以低成本实现上下文感知的消歧。Scene-Q在ScanNet200上改善了开放词汇的3D实例分割，并在真实世界重建中提升了自然语言3D实例检索，特别是在空间和关系查询上获得了最大的提升，同时保持了相当一部分查询在快速路径上。

View on arXiv Download PDF AI Translation

cs.CV / 66 / 2609.20245

SAGE-Yoga: Multi-Cue Learning for Yoga Pose Classification and Joint-Level Correction

SAGE-Yoga：用于瑜伽姿势分类和关节级修正的多线索学习

Chi, Hung Le, Huynh, Khanh Minh, Pham, Long Nghia Tran, Huynh, Tan Phuc, Nguyen, Trong-Thuan, Tran, Minh-Triet

Abstract

Automated yoga analysis requires both accurate pose classification and interpretable feedback on pose execution. However, existing methods often rely on a single visual prediction, struggle to distinguish visually similar poses, and treat pose classification and correction as separate tasks. To address these limitations, we propose SAGE-Yoga, a unified coarse-to-fine framework for yoga pose classification and joint-level correction from a single RGB image. Inspired by how yoga instructors assess posture using multiple complementary cues, SAGE-Yoga first employs a bagging-based ensemble of complementary visual backbones to generate a ranked set of candidate pose classes. Additionally, a margin-based gating mechanism preserves confident visual predictions while invoking geometric verification only for ambiguous cases. Moreover, once the final pose class is determined, SAGE-Yoga retrieves a medoid reference pose and compares the observed joint angles with class-specific distributions to identify misaligned joints. Finally, these deviations are translated into actionable corrective feedback. Empirically, experiments on the Yoga-82 dataset show that the visual ensemble achieves 89.0% Top-1 accuracy, while the complete framework improves performance to 90.7% Top-1 accuracy and 90.1% Macro-F1. These results demonstrate that combining complementary visual evidence with selective geometric verification improves fine-grained pose classification while enabling interpretable, joint-level correction.

Chinese Translation

自动化瑜伽分析需要准确的姿势分类和可解释的姿势执行反馈。然而，现有方法往往依赖单一的视觉预测，难以区分视觉上相似的姿势，并将姿势分类和修正视为独立任务。为了解决这些局限性，我们提出了SAGE-Yoga，一个统一的从粗到细的框架，用于从单个RGB图像进行瑜伽姿势分类和关节级修正。受到瑜伽教练使用多种互补线索评估姿势的启发，SAGE-Yoga首先采用基于袋装的互补视觉骨干网络集成，生成一组排名的候选姿势类别。此外，基于边际的门控机制在保留自信视觉预测的同时，仅在模糊情况下调用几何验证。此外，一旦确定最终姿势类别，SAGE-Yoga会检索一个中位数参考姿势，并将观察到的关节角度与类别特定分布进行比较，以识别不对齐的关节。最后，这些偏差被转化为可操作的修正反馈。实证研究表明，在Yoga-82数据集上的实验显示，视觉集成实现了89.0%的Top-1准确率，而完整框架将性能提升至90.7%的Top-1准确率和90.1%的宏观F1值。这些结果表明，结合互补视觉证据与选择性几何验证可以改善细粒度姿势分类，同时实现可解释的关节级修正。

View on arXiv Download PDF AI Translation

cs.CV / 67 / 2609.20248

Distance to Class Prototypes: Active Learning for Object Detection

距离类别原型：用于目标检测的主动学习

Zhang, Licheng, Gong, Zheng

Abstract

Deploying a deep object detector in a new setting is limited less by architecture than by the cost of annotating data from that setting. Active learning lowers the cost by choosing which images to label, and the choice is only as good as the signal used to score an unlabeled image. That signal is usually the class posterior, which is cheap but poorly calibrated, or the disagreement across several models or several stochastic passes, which is better but multiplies inference over a pool far larger than the labeled set. We propose a signal richer than the posterior yet still read from one forward pass of one network. A supervised contrastive term added to the training objective shapes a per-object embedding space in which distance encodes class membership, and an unlabeled detection is scored by how far it lies from the region occupied by its predicted category, weighted by its confidence. The criterion needs no ensemble, no auxiliary predictor and no repeated inference, and its entire cost is 2.89M parameters, an increase of 8.3% over a bare detector. On PASCAL VOC and MS-COCO it beats the posterior of the same detector in every round in which a selection is made, by up to 1.08% mAP50 against run to run deviations of 0.02% to 0.18%, and it stays competitive with ensemble and Monte Carlo dropout criteria costing three to fifty forward passes per unlabeled image. Experiments use the single-stage detector under which the compared criteria report their results, so that the selection decision is isolated from the strength of the detector.

Chinese Translation

在新的环境中部署深度目标检测器的限制主要不在于架构，而在于对该环境中数据进行标注的成本。主动学习通过选择标注哪些图像来降低成本，而选择的效果取决于用于对未标注图像进行评分的信号。这个信号通常是类别后验（class posterior），虽然成本低但校准效果差，或者是多个模型或多个随机传递之间的分歧（disagreement），虽然效果更好但在推理时需要处理远大于标注集的池。我们提出了一种比后验信号更丰富的信号，且仍然只需通过一个网络的前向传递（forward pass）读取。通过在训练目标中添加一个监督对比项（supervised contrastive term），形成了一个每个对象的嵌入空间（embedding space），其中距离编码了类别成员资格，未标注的检测通过其距离预测类别所占区域的远近来评分，并根据其置信度加权。该标准不需要集成（ensemble）、辅助预测器（auxiliary predictor）和重复推理，其整体成本为2.89M参数，比裸检测器（bare detector）增加了8.3%。在PASCAL VOC和MS-COCO数据集上，它在每次选择时都优于同一检测器的后验，mAP50的提升幅度可达1.08%，而运行间的偏差为0.02%到0.18%。它在成本为每个未标注图像三到五十次前向传递的集成和蒙特卡洛丢弃标准（Monte Carlo dropout criteria）中保持竞争力。实验使用单阶段检测器（single-stage detector），在此基础上比较标准报告结果，从而使选择决策与检测器的强度相隔离。

View on arXiv Download PDF AI Translation

cs.CV / 68 / 2609.20262

Generative Verification: Rethinking the Uncertainty Signal for Active Learning of Object Detection

生成验证：重新思考主动学习目标检测的不确定性信号

Zhang, Licheng, Gong, Zheng

Abstract

Nearly every acquisition function for active object detection shares one arrangement, in that the model being improved is also the model being interrogated. We depart from it. In generative verification an independent generative model re-derives the label of a detection from the pixels inside its predicted box, and the disagreement between the two becomes the acquisition signal. Two properties follow from the arrangement itself rather than from any tuning. A displaced box, a box on background and a correct box carrying the wrong label all yield a crop that fails verification, so the failure modes arrive already combined in one scalar and the hand-weighted classification and localization terms of existing criteria are no longer needed. And because the verifier never observes the detector confidence, confidently wrong detections score highest, although a self-derived signal reads them as uninteresting and they are the costliest to leave unlabeled. We build the verifier as a conditional diffusion model whose diffusion target is a label representation rather than an image. Its reverse process is stochastic, so repeated generations return a distribution whose concentration reports how firmly the evidence determines the label, where a classifier returns a single point estimate. On PASCAL VOC and MS-COCO the signal outperforms output-uncertainty, feature-geometry, perturbation and ensemble criteria, gaining about one mAP50 point per round on MS-COCO, with its largest margins in the early rounds where confident detector errors are most common.

Chinese Translation

几乎所有用于主动目标检测的获取函数都有一个共同点，即被改进的模型也是被询问的模型。我们对此进行了不同的探索。在生成验证中，一个独立的生成模型根据其预测框内的像素重新推导检测的标签，而两者之间的分歧则成为获取信号。由这种安排本身而非任何调优衍生出的两个特性是：偏移的框、背景上的框以及带有错误标签的正确框都会产生一个未通过验证的裁剪，因此失败模式已经以一个标量的形式组合在一起，现有标准的手动加权分类和定位项不再需要。而且，由于验证器从未观察到检测器的置信度，置信度高但错误的检测得分最高，尽管自我推导的信号将其视为无趣，而将其留为未标记的成本最高。我们构建的验证器是一个条件扩散模型，其扩散目标是标签表示而非图像。其反向过程是随机的，因此重复生成返回一个分布，其集中度报告证据如何坚定地决定标签，而分类器返回的是单一的点估计。在 PASCAL VOC 和 MS-COCO 数据集上，该信号优于输出不确定性、特征几何、扰动和集成标准，在 MS-COCO 上每轮大约提高一个 mAP50 点，其最大优势出现在早期轮次中，那里置信度高的检测错误最为常见。

View on arXiv Download PDF AI Translation

cs.CV / 69 / 2609.20263

AI or Real: Detecting Partially Altered Videos Under Resource-Constrained Environments

人工智能还是现实：在资源受限环境下检测部分篡改视频

Chakraborty, Tamoghna, Absur, Md Nurul, Saha, Sourya, Debroy, Saptarshi

Abstract

The proliferation of generative video models has shifted the practical detection threat from fully fabricated clips to partially manipulated footages. Although modern detectors achieve strong accuracy using foundation backbones of 400M+ parameters, their resource footprint precludes edge deployment. In this paper, we present a lightweight full-frame detector for partially manipulated AI-generated video, designed for deployment on edge hardware without face-detection preprocessing. The system distills a DINOv2-Base teacher into a frozen MobileNetV3-Small student through a pipeline that combines temperature-annealed soft-label transfer, attention-diversity regularization, frame-level supervision, and a residual feature adapter that conditions ImageNet features for artifact detection. We additionally target two failure modes specific to the partial-manipulation regime: false positives on legitimate scene cuts, addressed through within-video temporal hard negatives; and threshold-level miscalibration on the dominant pure-real class, addressed through calibration-aware sampling. Evaluation on a 55,393-sample spliced test set across fake-frame ratios from 6.2% to 31.2% demonstrates the student model closing 58% of the gap to the DINOv2-Base teacher (AUC 0.766) while running at 3.65 ms per 16-frame clip on RTX A4000 with a 150.4 MB checkpoint compatible with edge-device memory and latency budgets.

Chinese Translation

生成视频模型的普及使得实际检测威胁从完全伪造的片段转向部分篡改的影像。尽管现代检测器在使用超过4亿参数的基础模型时能够实现较强的准确性，但其资源占用使得边缘部署变得困难。本文提出了一种轻量级的全帧检测器，专门用于检测部分篡改的人工智能生成视频，旨在无需人脸检测预处理即可在边缘硬件上部署。该系统通过一个管道将DINOv2-Base教师模型提炼为冻结的MobileNetV3-Small学生模型，结合了温度退火软标签转移、注意力多样性正则化、帧级监督以及一个残差特征适配器，用于调整ImageNet特征以进行伪影检测。此外，我们还针对部分篡改模式的两种失效模式进行了研究：在合法场景切换上的误报，通过视频内的时间硬负样本进行解决；以及在主导的纯真实类别上的阈值级别误校准，通过校准感知采样进行解决。在一个包含55,393个样本的拼接测试集上进行评估，假帧比例从6.2%到31.2%，结果表明学生模型缩小了与DINOv2-Base教师模型（AUC 0.766）之间58%的差距，同时在RTX A4000上以每16帧片段3.65毫秒的速度运行，且具有150.4 MB的检查点，兼容边缘设备的内存和延迟预算。

View on arXiv Download PDF AI Translation

cs.CV / 70 / 2609.20267

CleanVideo: Adaptive Concept Erasure for Text-to-Video Diffusion Models

CleanVideo：用于文本到视频扩散模型的自适应概念消除

Liao, Junchi, Li, Hongji, Zhou, Wenrui, Hu, Lijie

Abstract

Concept erasure aims to selectively eliminate undesired visual semantics from pre-trained generative models without compromising their general utility. Extending concept erasure from images to video is nontrivial. Target concepts emerge gradually and vary across frames and denoising steps. As a result, fixed interventions may miss the target or introduce blurring, jitter, and content distortion. We propose CleanVideo, a selective erasure framework that performs low-dimensional subspace intervention controlled by a tri-modal gating mechanism. By jointly processing spatiotemporal visual features, timestep signals, and textual semantics, CleanVideo determines where, when, and whether to intervene, steering erased content toward natural surrogate concepts when such surrogates can be clearly defined while preserving non-target content. Experiments on three video diffusion models show that CleanVideo effectively erases target concepts while maintaining visual fidelity and temporal coherence, outperforming existing baselines under frame-level and video-level evaluations and under concept-recovery attacks when the protected pipeline remains intact.

Chinese Translation

概念消除旨在选择性地从预训练生成模型中消除不需要的视觉语义，而不影响其整体实用性。将概念消除从图像扩展到视频并非易事。目标概念逐渐出现，并在帧和去噪步骤之间变化。因此，固定的干预可能会错过目标或引入模糊、抖动和内容失真。我们提出了CleanVideo，这是一种选择性消除框架，通过三模态门控机制控制低维子空间干预。通过联合处理时空视觉特征、时间步信号和文本语义，CleanVideo确定何时、何地以及是否进行干预，在可以清晰定义的情况下，将消除的内容引导向自然的替代概念，同时保留非目标内容。在三种视频扩散模型上的实验表明，CleanVideo有效地消除了目标概念，同时保持视觉保真度和时间一致性，在帧级和视频级评估以及在保护管道保持完整的情况下，优于现有基线模型，抵御概念恢复攻击。

View on arXiv Download PDF AI Translation

cs.CV / 71 / 2609.20283

Queries Knew More Than We Thought: Uncovering Latent Knowledge in Segmentation Models

查询比我们想象的更了解：揭示分割模型中的潜在知识

De la Jara, Ignacio M., Rodriguez-Opazo, Cristian, Ranasinghe, Damith

Abstract

Modern segmenters often fail after the expensive computation has already been done: a useful mask is present among the model's query-conditioned candidates, but the deployed selection rule does not expose it. We study this output-selection bottleneck in frozen DETR-family models. A ground-truth-only oracle first shows substantial hidden headroom in already-computed mask proposals. This raises a simple question: How can we better use the masks a segmenter has already computed but does not expose? We then ask whether that headroom can be recovered without adding queries, generating new masks, rerunning the backbone, or updating weights. HYDRA is a small selector trained only on cached frozen outputs. At inference time, it scores the cached candidates against an explicit keep-baseline option and acts only when a held-out calibrated margin indicates the selected candidate is sufficiently better. Trained on training-split caches and calibrated on held-out data, HYDRA improves Mask2Former, MaskDINO, and OneFormer by up to +7.41 dataset mIoU points on ADE20k and COCO, and improves SAM 3 by +9.4 class-macro prompt-IoU points on average across eight domains while preserving useful predictions through calibration. Paired LoRA controls show that lightweight weight adaptation does not remove the bottleneck: exposed predictions are often flat or worse, while routing over the adapted candidates still recovers accuracy. Finally, we connect the effect to query specialization under bipartite matching and verify it in a controlled TinyDETR study. These results show that frozen segmenters should be evaluated not only by the masks they expose, but also by the useful candidates they suppress.

Chinese Translation

现代分割模型在昂贵的计算完成后常常失败：在模型的查询条件候选中存在一个有用的掩码，但部署的选择规则并未将其揭示。我们研究了冻结的DETR系列模型中的这一输出选择瓶颈。一个仅基于真实标签的oracle首先显示出在已经计算的掩码提案中存在显著的隐藏潜力。这引发了一个简单的问题：我们如何更好地利用分割器已经计算但未揭示的掩码？接着，我们询问是否可以在不增加查询、生成新掩码、重新运行主干网络或更新权重的情况下恢复这一潜力。HYDRA是一个仅在缓存的冻结输出上训练的小型选择器。在推理时，它将缓存的候选与明确的保留基线选项进行评分，仅在保留的校准边际表明所选候选者明显更好时才采取行动。在训练集缓存上训练并在保留数据上进行校准，HYDRA使Mask2Former、MaskDINO和OneFormer在ADE20k和COCO上提高了多达+7.41的数据集mIoU点，并使SAM 3在八个领域的平均类宏提示IoU点提高了+9.4，同时通过校准保持了有用的预测。配对的LoRA控制显示，轻量级的权重适应并未消除瓶颈：暴露的预测往往平坦或更差，而在适应候选者上进行路由仍能恢复准确性。最后，我们将这一效果与二分匹配下的查询专业化联系起来，并在受控的TinyDETR研究中进行了验证。这些结果表明，冻结的分割器不仅应通过其暴露的掩码进行评估，还应通过其抑制的有用候选进行评估。

View on arXiv Download PDF AI Translation

cs.CV / 72 / 2609.20290

TinyCNN: A 193K-Parameter Network for On-Device Plant Disease Detection, with a Cross-Dataset Robustness Diagnosis

TinyCNN：一种用于设备端植物病害检测的193K参数网络，具有跨数据集的鲁棒性诊断

Ho-Lam, Ngoc-Bao, Nguyen, Thai-Anh

Abstract

Detecting crop disease early is central to sustainable agriculture and food security under United Nations Sustainable Development Goal 2 (Zero Hunger), and is especially urgent in resource-constrained regions where expert diagnosis is scarce but low-cost mobile devices are widespread. This paper presents TinyCNN, a lightweight convolutional neural network for on-device plant disease classification. TinyCNN uses depthwise separable convolution blocks and contains only 193,190 trainable parameters with 110.05M MACs for a 224x224 input image. On the 38-class PlantVillage benchmark, TinyCNN achieves 98.88% test accuracy and 98.03% macro-F1 while being approximately 58x smaller than ResNet18 and 11.8x smaller than a MobileNetV2 teacher, directly reducing the energy, memory, and cost footprint of inference in line with Green AI principles. The paper further analyzes vanilla knowledge distillation as a sustainable model-compression strategy; an ablation over alpha in {0.3, 0.5, 0.7} and T in {2, 4} selects alpha=0.3, T=4, producing a distilled TinyCNN with 98.81% test accuracy. Finally, cross-dataset evaluation from PlantVillage to PlantDoc reveals a substantial robustness gap under real-world conditions, which a Grad-CAM analysis attributes to off-leaf, background-driven attention consistent with shortcut learning. TinyCNN is thus an energy-efficient, deployable building block for sustainable agricultural intelligence, while field robustness remains the key barrier to durable real-world impact.

Chinese Translation

早期检测作物病害是可持续农业和食品安全的核心，符合联合国可持续发展目标2（消除饥饿），在资源匮乏的地区尤为迫切，因为这些地区缺乏专家诊断，但低成本的移动设备却广泛存在。本文提出了TinyCNN，一种用于设备端植物病害分类的轻量级卷积神经网络。TinyCNN使用深度可分离卷积块，仅包含193,190个可训练参数，对于224x224的输入图像，其计算量为110.05M MAC。在38类PlantVillage基准测试中，TinyCNN实现了98.88%的测试准确率和98.03%的宏F1分数，同时其规模约为ResNet18的58倍小，约为MobileNetV2教师模型的11.8倍，从而直接减少了推理过程中的能耗、内存和成本，符合绿色人工智能原则。本文进一步分析了原始知识蒸馏作为一种可持续的模型压缩策略；在{0.3, 0.5, 0.7}中的alpha和{2, 4}中的T的消融实验中选择了alpha=0.3，T=4，生成了一个蒸馏的TinyCNN，其测试准确率为98.81%。最后，从PlantVillage到PlantDoc的跨数据集评估揭示了在现实世界条件下存在显著的鲁棒性差距，Grad-CAM分析将其归因于与捷径学习一致的叶外、背景驱动的注意力。因此，TinyCNN是一个能效高、可部署的可持续农业智能构件，而现场鲁棒性仍然是实现持久现实世界影响的关键障碍。

View on arXiv Download PDF AI Translation

cs.CV / 73 / 2609.20299

Not All Layers Are Equal: Dynamic Layer Routing for Reliable CLIP OOD Detection

并非所有层都是平等的：用于可靠 CLIP OOD 检测的动态层路由

De la Jara, Ignacio M., Rodriguez-Opazo, Cristian, Ranasinghe, Damith

Abstract

Information aggregation across model layers are revealed to improve OOD detection. In contrast to crafting a method for layer-wise information aggregation in recent work, we investigate if layer selection is a learnable problem. In other words, we transpose the question from how to fuse layers to one asking which layers to trust for an input. Using a generalizable, weak, out of distribution context crafting approach for supervision, shown to be more effective than state of the art methods' mechanisms, we formulate learning a lightweight router to select a sparse, final-layer-anchored expert over CLIP's layer depth for OOD detection. Across three diverse benchmarks we demonstrate our learnable routing method dubbed Voyager improves OOD detection. On ImageNet-1K, Voyager achieves an average FPR@95 of 18.86, outperforming the strongest, comparable, prompt-learning method by 8.8 points. These gains persist across multiple supervision sources, including those used by existing state-of-the-art prompt-learning methods, demonstrating that, whilst our weak OOD supervision context is highly effective, the key advantage is realized from the learnable router component rather than the supervision source. Significantly, Voyager is highly practical; router learning takes approximately two minutes using less than 1 GB of memory, making it approximately 20x more efficient than current prompt-learning approaches. Anonymized Code: https://anonymous.4open.science/r/Voyager/

Chinese Translation

跨模型层的信息聚合被证明能够改善 OOD 检测。与近期研究中为层级信息聚合设计方法不同，我们探讨层选择是否是一个可学习的问题。换句话说，我们将问题从如何融合层转变为询问对于某个输入应信任哪些层。我们采用一种通用的、弱的、基于分布外上下文的构建方法进行监督，证明其在效果上优于现有最先进方法的机制，我们构建了一个轻量级路由器，以选择一个稀疏的、以最终层为锚的专家，针对 CLIP 的层深度进行 OOD 检测。在三个不同的基准测试中，我们展示了我们可学习的路由方法 Voyager 改善了 OOD 检测。在 ImageNet-1K 上，Voyager 达到了 18.86 的平均 FPR@95，超越了最强的、可比较的提示学习方法 8.8 个点。这些提升在多个监督源中持续存在，包括现有最先进的提示学习方法所使用的源，表明尽管我们的弱 OOD 监督上下文非常有效，但关键优势来自于可学习的路由器组件，而非监督源。值得注意的是，Voyager 具有很高的实用性；路由器学习大约需要两分钟，使用的内存少于 1 GB，使其效率比当前的提示学习方法高出约 20 倍。匿名代码： https://anonymous.4open.science/r/Voyager/

View on arXiv Download PDF AI Translation

cs.CV / 74 / 2609.20325

AgriScope: Pixel-Grounded Multimodal Understanding for Agricultural Images

AgriScope：面向农业图像的像素基础多模态理解

Boudiaf, Abderrahmene, Alanssari, Mohamad, Hussain, Irfan, Javed, Sajid

Abstract

Agricultural image understanding requires fine-grained recognition of plant diseases, pests, crop structures, and botanical species under complex real-world conditions. Despite recent advances in Multimodal Large Language Models (MLLMs), existing models remain limited to text-only outputs and lack pixel-level visual grounding capabilities. In this work, we introduce AgriScope, a unified pixel-grounded multimodal framework for agricultural image understanding. AgriScope jointly supports image-level, region-level, and pixel-level understanding within a unified framework, enabling tasks such as grounded caption generation, referring expression segmentation, and multi-turn multimodal interaction for agricultural imagery. AgriScope integrates biologically specialized semantic representations with dense spatial grounding through biological-semantic encoding, dense spatial representations, and pixel decoding. To support large-scale grounded learning, we introduce AgriGround, a large-scale pixel-grounded agricultural multimodal instruction-tuning dataset containing over 500K images and 11M instruction-following samples spanning plant disease analysis, crop and weed identification, insect pest recognition, and fine-grained botanical understanding. AgriGround is constructed through a multi-stage automatic annotation pipeline that integrates multimodal caption generation, phrase-level grounding, segmentation mask generation, and task-oriented instruction synthesis to produce densely grounded supervision. Extensive experiments across multiple agricultural vision-language tasks demonstrate the effectiveness of AgriScope in pixel-grounded multimodal understanding, establishing a strong benchmark for agricultural vision-language learning and visual grounding. The dataset and code will be made publicly available at (https://github.com/boudiafA/AgriScope)

Chinese Translation

农业图像理解需要在复杂的现实条件下对植物疾病、害虫、作物结构和植物种类进行细粒度识别。尽管多模态大型语言模型（MLLMs）近年来取得了进展，但现有模型仍然局限于仅输出文本，缺乏像素级的视觉基础能力。在本研究中，我们提出了AgriScope，一个统一的像素基础多模态框架，用于农业图像理解。AgriScope在统一框架内共同支持图像级、区域级和像素级理解，使得能够实现诸如基础描述生成、指代表达分割和农业图像的多轮多模态交互等任务。AgriScope通过生物语义编码、密集空间表示和像素解码，将生物专业的语义表示与密集空间基础相结合。为了支持大规模的基础学习，我们引入了AgriGround，这是一个大规模的像素基础农业多模态指令调优数据集，包含超过50万张图像和1100万条指令跟随样本，涵盖植物疾病分析、作物和杂草识别、害虫识别以及细粒度植物理解。AgriGround通过多阶段自动注释管道构建，该管道整合了多模态描述生成、短语级基础、分割掩码生成和任务导向的指令合成，以产生密集的基础监督。在多个农业视觉-语言任务上的广泛实验表明，AgriScope在像素基础多模态理解方面的有效性，为农业视觉-语言学习和视觉基础建立了强有力的基准。数据集和代码将公开发布于（https://github.com/boudiafA/AgriScope）

View on arXiv Download PDF AI Translation

cs.CV / 75 / 2609.20340

FreqDINO++: A Frequency-Guided Multi-Task Routing Vision Foundation Model for Universal Ultrasound Analysis

FreqDINO++：一种频率引导的多任务路由视觉基础模型用于通用超声分析

Xu, Qing, Zhang, Yixuan, Li, Yue, He, Xiangjian, Zhang, Qian, Haque, Mainul, Qu, Rong, Duan, Wenting, Bai, Jieyun, Chen, Zhen

Abstract

Ultrasound image analysis plays a crucial role in cancer screening and prenatal diagnosis, yet comprehensive assessment requires jointly addressing tasks such as lesion segmentation and benign-malignant classification. While recent vision foundation models have shown remarkable universal representations, unlocking their potential for ultrasound is bottlenecked by the considerable domain gap from natural images. Existing methods typically fine-tune heavy vision encoders for isolated tasks, incurring substantial computational overhead while overlooking the underlying commonalities across heterogeneous tasks. In this work, we propose FreqDINO++, a frequency-guided multi-task routing vision foundation model for universal ultrasound analysis. We first introduce a Multi-task Routing Adapter (MR-Adapter) to support parameter-efficient integration of task-common and task-specific knowledge, a Frequency-aware Feature Enhancer (F$^2$-Enhancer) is then designed to capture the rich multi-scale frequency characteristics of ultrasound images, and a Task-aligned Collaborative Decoder (TC-Decoder) is devised to promote collaboration between dense and global prediction tasks through global-local token interaction. Extensive experiments on large-scale multi-task and external single-task ultrasound benchmarks demonstrate that FreqDINO++ consistently outperforms strong baselines and recent foundation models across 27 diverse clinical task scenarios, while also showing promising generalization to unseen data. The code is at https://github.com/MingLang-FD/FreqDINO-Plus.

Chinese Translation

超声图像分析在癌症筛查和产前诊断中发挥着至关重要的作用，但全面评估需要共同解决病灶分割和良恶性分类等任务。尽管近期的视觉基础模型展现了显著的通用表示，但由于与自然图像之间存在显著的领域差距，释放其在超声图像分析中的潜力面临瓶颈。现有方法通常对重型视觉编码器进行微调以应对孤立任务，这不仅带来了巨大的计算开销，还忽视了异构任务之间的潜在共性。在本研究中，我们提出了FreqDINO++，一种频率引导的多任务路由视觉基础模型，用于通用超声分析。我们首先引入了多任务路由适配器（Multi-task Routing Adapter, MR-Adapter），以支持任务共性和任务特定知识的参数高效整合；接着设计了频率感知特征增强器（Frequency-aware Feature Enhancer, F$^2$-Enhancer），以捕捉超声图像丰富的多尺度频率特征；最后构建了任务对齐协作解码器（Task-aligned Collaborative Decoder, TC-Decoder），以通过全局-局部标记交互促进密集和全局预测任务之间的协作。在大规模多任务和外部单任务超声基准测试中的广泛实验表明，FreqDINO++在27个不同的临床任务场景中始终优于强基线和近期的基础模型，同时在未见数据上也展现出良好的泛化能力。代码可在 https://github.com/MingLang-FD/FreqDINO-Plus 获取。

View on arXiv Download PDF AI Translation

cs.CV / 76 / 2609.20341

Fast Cross-Strength Multi-Contrast Brain MRI Translation using Latent Bridge Matching

基于潜在桥匹配的快速跨强度多对比脑部MRI翻译

Srivastava, Siddharth, Bretschneider, Till

Abstract

Magnetic Resonance Imaging (MRI) acquired at different field strengths exhibits pronounced variation in noise, resolution, homogeneity, and contrast, which limits comparability across acquisition settings and complicates downstream analysis. We address this with a unified conditional model for controllable field-to-field synthesis, built on the framework of conditional latent bridge matching. Our single model achieves highly competitive results across the validation phase for all three tasks of the MRIxFields2026 challenge without task-specific architectures or training. We achieve fast generation with only a single inference step, producing all modality and field-strength combinations for $30$ axial slices in under $90$ seconds, as well as cross-modality-strength translation for a full volume in under $70$ seconds, on a single NVIDIA A5000 GPU. We further provide extensive ablations regarding different components of our solution. Code: https://gitlab.com/siddharthsrivastava/mrixfields-2026

Chinese Translation

在不同磁场强度下获取的磁共振成像（MRI）在噪声、分辨率、均匀性和对比度上表现出明显的变化，这限制了不同获取设置之间的可比性，并使后续分析变得复杂。我们通过一个统一的条件模型来解决这个问题，该模型用于可控的场到场合成，基于条件潜在桥匹配的框架构建。我们的单一模型在MRIxFields2026挑战的所有三个任务的验证阶段中取得了高度竞争的结果，而无需特定于任务的架构或训练。我们仅用一步推理实现快速生成，在不到90秒的时间内为30个轴向切片生成所有模态和场强组合，并在不到70秒的时间内实现全体积的跨模态强度翻译，均在单个NVIDIA A5000 GPU上完成。我们还提供了关于我们解决方案不同组件的广泛消融实验。代码链接： https://gitlab.com/siddharthsrivastava/mrixfields-2026

View on arXiv Download PDF AI Translation

cs.CV / 77 / 2609.20348

EliGSiR: Continual RGB-D Mapping with Gaussian Splatting under Bounded Compute

EliGSiR：在有限计算下的连续RGB-D映射与高斯喷涂

Ellensohn, Björn, Rueckert, Elmar, Rauch, Christian

Abstract

Conventional 3D Gaussian Splatting assumes a closed set of observations and long optimization schedules. Continual RGB-D mapping in contrast poses the problem that new observations arrive online, while previously reconstructed regions must be preserved. We present EliGSiR (Evidence-guided Load-adaptive Incremental Gaussian Splatting with Image Replay), a continual Gaussian mapper that controls how the available optimization budget is used as the reconstruction evolves. Map-Guided View Scheduling filters redundant incoming views and reconsiders retained views according to the current state of the map. Load-Adaptive Fidelity adjusts supervision resolution to the current mapping load instead of following a fixed resolution schedule. Targeted Geometry Growth separates depth supervision from Gaussian creation and adds geometric capacity only where repeated RGB-D observations indicate missing or misplaced structure. Together, these mechanisms adapt which views are optimized, how much image detail is used, and where the representation grows while mapping remains active. We evaluate EliGSiR on Replica, TUM RGB-D, ScanNet++, and real RGB-D sensor sequences, considering both the final reconstruction and the map available throughout acquisition. On TUM RGB-D fr3/long_office_household, EliGSiR reaches 21.52 dB with the same ground-truth mapping poses used by the controlled baselines, compared with 19.42 dB for SplaTAM. In the tracked-pose comparison, EliGSiR with live ORB-SLAM3 poses reaches 23.02 dB in 155.5 s, compared with 20.10 dB in 230.9 s for CaRtGS using its native tracker. We further evaluate reconstruction throughout acquisition and show how EliGSiR adaptive view scheduling, supervision fidelity, and geometry growth improve the use of the available mapping budget.

Chinese Translation

传统的3D高斯喷涂假设观察集是封闭的，并且需要较长的优化周期。与此不同，连续RGB-D映射面临的问题是新的观察数据在线到达，同时必须保留先前重建的区域。我们提出了EliGSiR（基于证据的负载自适应增量高斯喷涂与图像重放），这是一种连续高斯映射器，能够在重建过程中控制可用优化预算的使用。地图引导视图调度过滤冗余的输入视图，并根据当前地图状态重新考虑保留的视图。负载自适应保真度根据当前映射负载调整监督分辨率，而不是遵循固定的分辨率计划。针对性几何增长将深度监督与高斯创建分离，仅在重复的RGB-D观察指示缺失或错误放置的结构时增加几何容量。这些机制共同决定了优化哪些视图、使用多少图像细节以及在映射活动时表示的增长位置。我们在Replica、TUM RGB-D、ScanNet++和真实RGB-D传感器序列上评估了EliGSiR，考虑了最终重建和整个采集过程中的地图。在TUM RGB-D fr3/long_office_household上，EliGSiR在使用受控基线的相同真实映射姿态时达到了21.52 dB，而SplaTAM为19.42 dB。在跟踪姿态比较中，使用实时ORB-SLAM3姿态的EliGSiR在155.5秒内达到了23.02 dB，而CaRtGS使用其本地跟踪器在230.9秒内达到了20.10 dB。我们进一步评估了整个采集过程中的重建，并展示了EliGSiR的自适应视图调度、监督保真度和几何增长如何改善可用映射预算的使用。

View on arXiv Download PDF AI Translation

cs.CV / 78 / 2609.20377

MM-Future: Multi-Mode Joint World-Action Modeling for Autonomous Driving

MM-Future：用于自动驾驶的多模式联合世界-动作建模

Liu, Shuai, Gong, Hechangle, Jiang, Hao, He, Runlin, Zhan, Junxiang, Huang, Kai, Yang, Sheng, Ren, Shaoqing

Abstract

Autonomous driving involves coupled decision-making and scene evolution under multi-mode uncertainty. To capture this coupling and uncertainty, we introduce MM-Future, a world-action model that generates multiple paired scene-action hypotheses and models bidirectional interaction within each pair. Each hypothesis is initialized from a structured action prior and an independent future scene source, which are then co-evolved through a modality-aware diffusion Transformer. To support efficient multi-mode rollout, MM-Future compresses multi-view video into planning-oriented representations, dubbed MM-Tokens. Finally, a future-conditioned proposal scorer ranks trajectory candidates by shared history context and their paired predicted future. On NAVSIM navtest, MM-Future achieves 94.0 PDMS and 91.5 EPDMS, while attaining a 32.3 HD-Score in zero-shot closed-loop evaluation on HUGSIM. Ablations show consistent improvements over both single-mode and action-only variants, validating the benefit of multi-mode joint world-action modeling.

Chinese Translation

自动驾驶涉及在多模式不确定性下的耦合决策和场景演变。为了捕捉这种耦合和不确定性，我们提出了MM-Future，这是一种世界-动作模型，能够生成多个配对的场景-动作假设，并在每对中建模双向交互。每个假设从结构化的动作先验和独立的未来场景源初始化，随后通过一种感知模式的扩散Transformer共同演化。为了支持高效的多模式展开，MM-Future将多视角视频压缩为面向规划的表示，称为MM-Tokens。最后，基于未来条件的提案评分器通过共享历史上下文及其配对的预测未来对轨迹候选进行排名。在NAVSIM navtest上，MM-Future达到了94.0的PDMS和91.5的EPDMS，同时在HUGSIM的零-shot闭环评估中获得了32.3的HD-Score。消融实验显示，相较于单模式和仅动作变体，MM-Future在性能上有一致的提升，验证了多模式联合世界-动作建模的优势。

View on arXiv Download PDF AI Translation

cs.CV / 79 / 2609.20386

Compact Vision Models for Iris Presentation Attack Detection under Presentation Attack Instrument Shift and Environmental Degradation

针对呈现攻击工具转移和环境退化的虹膜呈现攻击检测的紧凑视觉模型

Angelakis, Athanasios, Gomez-Barrero, Marta

Abstract

Iris presentation attack detection (PAD) is security-critical when a subsystem that appears reliable during development encounters presentation attack instruments (PAIs) or acquisition conditions absent from validation data. We benchmark three compact scratch-trained computer-vision models, each with at most approximately 0.26 million trainable parameters, on the Notre Dame subset of LivDet-Iris 2017 under PAI-driven domain shift and environmental degradation. All models are trained without external pretraining or data augmentation and evaluated over five seeds. A validation-selected threshold is transferred unchanged to the known-attack, unknown-attack, corrupted, and pooled test partitions. From known to unknown attack presentations, Attack Presentation Classification Error Rate (APCER) increases by 17.11-30.47 percentage points and Detection Equal Error Rate (D-EER) increases by 7.38-12.73 percentage points. At the validation-selected threshold, ZACH-ViT obtains the lowest unknown-attack APCER (47.69 +/- 4.84%) and D-EER (38.87 +/- 0.93%), while Compact-TransMIL obtains the lowest Bona Fide Presentation Classification Error Rate (BPCER). ZACH-ViT also gives the lowest unknown-attack BPCER at an APCER limit of 10% (81.29 +/- 1.95%). The high absolute errors show that the comparative advantage of the best compact model does not constitute deployment readiness under unknown PAIs.

Chinese Translation

虹膜呈现攻击检测（PAD）在一个看似可靠的子系统在开发过程中遇到未在验证数据中出现的呈现攻击工具（PAI）或采集条件时，具有重要的安全性。我们在PAI驱动的领域转移和环境退化下，对三个紧凑的Scratch训练计算机视觉模型进行了基准测试，每个模型最多约有26万可训练参数，测试数据集为LivDet-Iris 2017的Notre Dame子集。所有模型均在没有外部预训练或数据增强的情况下进行训练，并在五个种子上进行评估。验证选择的阈值在已知攻击、未知攻击、损坏和汇总测试分区中保持不变。从已知攻击到未知攻击的呈现中，攻击呈现分类错误率（APCER）增加了17.11-30.47个百分点，检测等错误率（D-EER）增加了7.38-12.73个百分点。在验证选择的阈值下，ZACH-ViT获得了最低的未知攻击APCER（47.69 +/- 4.84%）和D-EER（38.87 +/- 0.93%），而Compact-TransMIL获得了最低的真实呈现分类错误率（BPCER）。ZACH-ViT在APCER限制为10%时也提供了最低的未知攻击BPCER（81.29 +/- 1.95%）。高绝对错误表明，最佳紧凑模型的比较优势并不构成在未知PAI下的部署准备。

View on arXiv Download PDF AI Translation

cs.CV / 80 / 2609.20414

TouchSight: Bare-Handed Tactile Prediction from Egocentric Video via Generative Visual Augmentation

TouchSight：通过生成视觉增强从自我中心视频中进行裸手触觉预测

Zhou, Danyan, Lu, Jinxuan, Lin, Jiawei, Chen, Tianxing, Lyu, Chuqiao, Ding, Wenbo

Abstract

Tactile signals provide direct contact and force measurements that are essential for understanding physical interactions and enabling dexterous robotic manipulation. However, tactile sensing requires direct measurement at contact interfaces, making large-scale data collection reliant on intrusive, costly, and restrictive instrumentation. We present TouchSight, a monocular egocentric vision framework for dense full-hand contact force prediction that leverages 500 hours of pressure-glove recordings and extensive hand-object interaction (HOI) data. To address the appearance gap between gloved training data and bare-hand real-world scenarios, we construct TwinTouch-20H: 20 hours of paired visual data in which generative video models re-render gloved recordings as bare-hand observations against new backgrounds while preserving the original measured tactile labels. TouchSight predicts dense force from both gloved and generated bare-hand videos, outperforms prior contact prediction methods on OakInk2, qualitatively generalizes to natural bare-hand egocentric videos from unseen datasets, and improves consistently as glove supervision scales. These results demonstrate that dense tactile signals can be recovered from egocentric vision alone, without tactile instrumentation at capture time.

Chinese Translation

触觉信号提供了直接的接触和力测量，这对于理解物理交互和实现灵巧的机器人操作至关重要。然而，触觉传感需要在接触界面进行直接测量，使得大规模数据收集依赖于侵入性、昂贵且限制性强的仪器设备。我们提出了TouchSight，一个单目自我中心视觉框架，用于密集的全手接触力预测，该框架利用了500小时的压力手套录音和广泛的手物体交互（HOI）数据。为了弥补手套训练数据与裸手现实场景之间的外观差距，我们构建了TwinTouch-20H：20小时的配对视觉数据，其中生成的视频模型将手套录音重新渲染为裸手观察，背景则为新的环境，同时保留原始测量的触觉标签。TouchSight从手套和生成的裸手视频中预测密集的力，在OakInk2上优于之前的接触预测方法，并且在未见数据集中的自然裸手自我中心视频上具有定性泛化能力，随着手套监督的增加而持续改进。这些结果表明，仅通过自我中心视觉就可以恢复密集的触觉信号，而无需在捕获时使用触觉仪器。

View on arXiv Download PDF AI Translation

cs.CV / 81 / 2609.20423

WeVisDoc: From Coverage to Capability for Robust End-to-End Document Parsing

WeVisDoc：从覆盖到能力的鲁棒端到端文档解析

Yu, Hao, Liu, Kang, Zhao, Linnan, Zhan, Jiabo, Sun, Chong, Li, Chen, Lyu, Jing

Abstract

Document parsing converts document images into structured content and requires reliable performance across diverse layouts and acquisition conditions. Yet training corpora are biased toward common document types and clean digital pages, while expanding coverage alone does not specify how to address a parser's remaining weaknesses. We present WeVisDoc, a two-stage data-centric framework for robust end-to-end document parsing. Stage I broadens semantic, structural, and appearance coverage through heterogeneous data and structure-preserving degradation synthesis. Stage II uses a held-out probe to measure the Stage I parser's residual errors within fixed visual-structural clusters. These diagnostics guide targeted data construction and reallocation of the target-token budget. WeVisDoc-4B achieves an Overall score of 95.38 on OmniDocBench v1.6 and a mean Overall score of 75.54 across the three PureDocBench tracks, ranking first among the compared end-to-end parsers in all four settings. Compared with Stage I, Stage II improves Overall scores for the 2B and 4B models on both benchmarks, with larger gains on the degraded PureDocBench tracks, including a 4.03-point gain for the 4B model on the Real Degraded track.

Chinese Translation

文档解析将文档图像转换为结构化内容，并要求在多样化的布局和获取条件下具有可靠的性能。然而，训练语料库偏向于常见文档类型和干净的数字页面，而仅仅扩展覆盖范围并不能明确如何解决解析器的剩余弱点。我们提出了WeVisDoc，这是一个两阶段的数据中心框架，用于鲁棒的端到端文档解析。第一阶段通过异构数据和结构保持降解合成来扩展语义、结构和外观的覆盖范围。第二阶段使用保留的探针来测量第一阶段解析器在固定视觉-结构集群中的残余错误。这些诊断结果指导针对性的数据构建和目标标记预算的重新分配。WeVisDoc-4B在OmniDocBench v1.6上获得了95.38的总体得分，在三个PureDocBench轨道上平均总体得分为75.54，在所有四个设置中排名第一，超越了比较的端到端解析器。与第一阶段相比，第二阶段在这两个基准上提高了2B和4B模型的总体得分，在降级的PureDocBench轨道上获得了更大的提升，包括4B模型在真实降级轨道上获得了4.03分的提升。

View on arXiv Download PDF AI Translation

cs.CV / 82 / 2609.20427

When Do Language-Grounded Explanations Help? A Graph-Bottleneck for Farm Monitoring Interpretable Sheep Facial Pain

语言基础解释何时有效？用于农场监测的可解释羊面部疼痛的图形瓶颈

Noor, Alam, Gait'an, Miguel Guti'errez

Abstract

Automated pain recognition from facial expression could make continuous welfare assessment practical in sheep, but adoption depends on trust: a stockperson cannot act on a score that arrives without justification. We ground a model in the Sheep Pain Facial Expression Scale (SPFES) by letting each detected facial region attend over text embeddings of the clinical descriptors and then test whether the resulting explanations mean anything. They do not. Ablating an entire descriptor changes the predicted logit by about $10^{-4}$, and the most-attended cue agrees with the predicted pain level in only $32.6\%$ of regions, although the attention maps, the learned gate, and the generated text all proposed otherwise. We therefore remove the appearance bypass with a concept bottleneck whose classifier reads only SPFES concept scores, supervised by per-region state annotations that image-level pipelines discard. This costs $0.05$--$0.10$ in Cohen's $\kappa$ but yields concepts that are demonstrably learned: minority pain-indicating states are recovered at $3.5$--$8.3\times$ their base rates, and the ear and eye severity orderings emerge without severity supervision. Removing the supervision alone leaves $\kappa$ unchanged while concept accuracy falls to $0.109$, showing that architectural necessity does not imply semantic validity. We also show that pooled concept accuracy is misleading under clinical imbalance and provide a cross-validated, protocol-matched benchmark of seven methods on this dataset.

Chinese Translation

从面部表情自动识别疼痛可以使对羊的持续福利评估变得可行，但其采用依赖于信任：养殖人员无法在没有理由的情况下对到达的评分采取行动。我们通过让每个检测到的面部区域关注临床描述符的文本嵌入，将模型基于羊疼痛面部表情量表（Sheep Pain Facial Expression Scale, SPFES）进行构建，然后测试生成的解释是否有意义。结果并没有。去除整个描述符会使预测的对数几率变化约为 $10^{-4}$，而且最被关注的线索仅在 $32.6\%$ 的区域与预测的疼痛水平一致，尽管注意力图、学习的门控和生成的文本都提出了相反的观点。因此，我们通过一个概念瓶颈去除外观绕过，该分类器仅读取 SPFES 概念评分，由每个区域的状态注释进行监督，而图像级管道则丢弃这些注释。这导致 Cohen's $ ext{kappa}$ 下降 $0.05$--$0.10$，但产生了明显学习到的概念：少数指示疼痛的状态以 $3.5$--$8.3 imes$ 的基线率被恢复，而耳朵和眼睛的严重程度排序在没有严重程度监督的情况下出现。仅去除监督使得 $ ext{kappa}$ 不变，而概念准确性下降至 $0.109$，表明架构的必要性并不意味着语义的有效性。我们还表明，在临床不平衡下，汇总的概念准确性具有误导性，并提供了在该数据集上七种方法的交叉验证、协议匹配基准。

View on arXiv Download PDF AI Translation

cs.CV / 83 / 2609.20441

Cross-Architecture Foundation-Model Distillation for Edge Flood Segmentation

跨架构基础模型蒸馏用于边缘洪水分割

Schmalstieg, Fabian, Mueller, Karsten, Samek, Wojciech

Abstract

Geospatial foundation models can provide strong flood-segmentation performance, but their size limits deployment on memory-constrained edge hardware. We distill a 300-million-parameter Prithvi-EO-2.0 teacher, fine-tuned on the 252 manually labeled Sen1Floods11 training scenes, into a 0.7-million-parameter EfficientViT-B0 student. The teacher supervises additional unlabeled Sentinel-2 imagery, allowing the student training set to grow without new manual annotations. At the matched budget of 252 scenes, teacher-supervised training is competitive with direct training and improves STURM-Flood performance across tested configurations; a geometry-matched control shows that label source alone does not explain the difference. Scaling the teacher-supervised pool to 2,500 scenes narrows the remaining student--teacher gap: the float student reaches 0.787 water intersection over union on the Sen1Floods11 test split against 0.822 for the teacher, matches the teacher on STURM-Flood under our evaluation protocol, and remains below it on WorldFloods-v2. After activation replacement and quantization-aware training, the student runs as a 1.5-megabyte 8-bit integer (INT8) TensorRT engine on a Jetson Xavier NX at 5.57 milliseconds of graphics processing unit (GPU) compute per 512-by-512 image, with approximately 14 megabytes of runtime device memory. A fixed modified normalized difference water index (MNDWI) threshold is competitive with both models on the two clean external benchmarks, so we interpret those benchmarks as generalization tests rather than as evidence of learned-model superiority over a spectral rule. The results support the conclusion: foundation-model supervision can amplify a fixed manual annotation budget into a substantially larger training set and yield a compact, deployable edge model.

Chinese Translation

地理空间基础模型可以提供强大的洪水分割性能，但其体积限制了在内存受限的边缘硬件上的部署。我们将一个300百万参数的Prithvi-EO-2.0教师模型，经过在252个手动标注的Sen1Floods11训练场景上进行微调，蒸馏为一个70万参数的EfficientViT-B0学生模型。教师模型监督额外的未标注Sentinel-2影像，使学生训练集在没有新手动注释的情况下得以扩展。在252场景的匹配预算下，教师监督的训练与直接训练具有竞争力，并在测试的配置中提高了STURM-Flood的性能；几何匹配的对照实验表明，仅凭标签来源无法解释差异。将教师监督的池扩展到2500场景缩小了学生与教师之间的差距：浮动学生在Sen1Floods11测试集上达到0.787的水体交并比，而教师为0.822，在我们的评估协议下与教师在STURM-Flood上表现相当，但在WorldFloods-v2上仍低于教师。经过激活替换和量化感知训练后，学生模型作为一个1.5兆字节的8位整数（INT8）TensorRT引擎在Jetson Xavier NX上运行，每512×512图像的图形处理单元（GPU）计算时间为5.57毫秒，运行时设备内存约为14兆字节。固定的修改归一化差异水体指数（MNDWI）阈值在两个干净的外部基准测试中与这两个模型具有竞争力，因此我们将这些基准测试解释为泛化测试，而不是作为学习模型优于光谱规则的证据。结果支持以下结论：基础模型监督可以将固定的手动注释预算放大为一个实质上更大的训练集，并产生一个紧凑的、可部署的边缘模型。

View on arXiv Download PDF AI Translation

cs.CV / 84 / 2609.20475

SenseFuse: Label-Free Fusion of Image and Shape Encoders for Open-Vocabulary 3D Instance Segmentation

SenseFuse：无标签图像与形状编码器的融合用于开放词汇3D实例分割

Han, Euiseok, Ton, Tri, Kim, Hwanhee, Ryu, Seungyeon, Yoo, Chang D.

Abstract

Open-vocabulary scene understanding is fundamental for robotics, laying the groundwork for spatial reasoning and object manipulation. While closed-vocabulary 3D instance segmentation heavily leverages 3D shape information, state-of-the-art open-vocabulary methods remain predominantly restricted to 2D image features or image-distilled representations during mask labeling. In this paper, we propose SenseFuse, a label-free fusion method that balances 2D image and 3D shape encoders for robust open-vocabulary 3D instance segmentation, refining only the mask-labeling stage of existing pipelines. We reveal that 2D image and 3D shape encoders exhibit largely disjoint failure patterns and rarely share identical wrong labels, whereas two 2D image encoders frequently repeat the same errors. This distinct behavior makes the 2D and 3D pair inherently complementary. We introduce an adaptive mechanism that selects a scene-level fusion weight to maximize a label-free sensitivity measure, estimated directly from a single scene's unlabeled proposals in milliseconds. SenseFuse improves labeling accuracy in every evaluated setting across ScanNet200, Replica, and ScanNet++, recovering 67-100% (median 93%) of the gain achievable with an oracle weight, and it raises instance AP in 21 of 22 reported settings. Code is available at https://github.com/hanes1207/SenseFuse.

Chinese Translation

开放词汇场景理解是机器人技术的基础，为空间推理和物体操作奠定了基础。尽管闭合词汇的3D实例分割在很大程度上依赖于3D形状信息，但当前最先进的开放词汇方法在掩膜标注过程中仍主要局限于2D图像特征或图像提炼的表示。在本文中，我们提出了SenseFuse，一种无标签的融合方法，平衡了2D图像和3D形状编码器，以实现稳健的开放词汇3D实例分割，仅对现有流程的掩膜标注阶段进行改进。我们发现，2D图像和3D形状编码器表现出明显不同的失败模式，几乎不共享相同的错误标签，而两个2D图像编码器则经常重复相同的错误。这种独特的行为使得2D和3D的组合在本质上是互补的。我们引入了一种自适应机制，选择场景级融合权重，以最大化无标签敏感度度量，该度量直接从单个场景的未标记提案中在毫秒内估计。SenseFuse在ScanNet200、Replica和ScanNet++的每个评估设置中提高了标注准确性，恢复了67-100%（中位数93%）可通过oracle权重实现的增益，并在22个报告的设置中提高了21个实例AP。代码可在 https://github.com/hanes1207/SenseFuse 获取。

View on arXiv Download PDF AI Translation

cs.CV / 85 / 2609.20508

Grounded Product Understanding in Livestream Videos

基于直播视频的产品理解

Zhang, Xinyu, Chen, Junjie, Ge, Jiawei, Li, Qianlong, Ma, Libin, Pan, Baokun, Luo, Yahui

Abstract

E-commerce livestreams have emerged as an important channel for presenting products to online consumers, containing multiple products whose information is scattered in different moments. This poses significant challenges for downstream product understanding applications, such as product-centric livestream clipping, where models need to identify the product and its relevant segments for information gathering. However, existing benchmarks for general product understanding typically evaluate product retrieval and temporal localization in isolation, leaving the critical correspondence between product identity and temporal evidence largely unassessed. To address this limitation, we introduce GPUB, a large-scale benchmark comprising 3,000 livestream instances with quality-controlled multi-moment temporal annotations and a catalog of over 31K fashion products. GPUB supports three evaluation tasks: the main task Grounded Product Understanding (GPrU) requires jointly identifying the target product and localizing its supporting moments from a livestream video and a candidate product set; Product Retrieval and Product Moment Localization serve as two complementary subtasks. Evaluation of existing multimodal models shows that GPrU remains highly challenging, with the best-performing baseline achieving only 10.13% Pair [email protected]. To narrow the performance gap, we further develop UniPro, a unified product understanding model that derives product-aligned and temporally structured representations from shared multimodal encoding, improving Pair [email protected] to 21.53% while achieving 37.23% Joint R@[email protected] on GPrU.

Chinese Translation

电子商务直播已成为向在线消费者展示产品的重要渠道，其中包含多个产品的信息分散在不同的时刻。这给下游的产品理解应用带来了重大挑战，例如以产品为中心的直播剪辑，模型需要识别产品及其相关片段以进行信息收集。然而，现有的通用产品理解基准通常孤立地评估产品检索和时间定位，未能充分评估产品身份与时间证据之间的关键对应关系。为了解决这一局限性，我们引入了GPUB，这是一个大规模基准，包含3000个直播实例，具有质量控制的多时刻时间注释和超过31,000个时尚产品的目录。GPUB支持三项评估任务：主要任务“基于产品理解”（Grounded Product Understanding, GPrU）要求从直播视频和候选产品集中共同识别目标产品并定位其支持时刻；产品检索和产品时刻定位作为两个互补的子任务。对现有多模态模型的评估表明，GPrU仍然具有很高的挑战性，表现最佳的基线仅达到10.13%的Pair [email protected]。为了缩小性能差距，我们进一步开发了UniPro，一个统一的产品理解模型，从共享的多模态编码中推导出产品对齐和时间结构化的表示，将Pair [email protected]提高到21.53%，同时在GPrU上实现37.23%的Joint R@[email protected]。

View on arXiv Download PDF AI Translation

cs.CV / 86 / 2609.20509

Automated Goldsmith's Mark Retrieval in Silverware

银器中的自动金匠标记检索

Tiwari, Atmik, Christlein, Vincent, Fichtner, Mark, Gohlke, Freya, Schübel, Birgit, Witting, Theresa, Zech, Heike, Zinnen, Mathias

Abstract

For art historians, goldsmith marks play a critical role in the identification and dating of artifacts. In practice, experts must manually compare a query mark against hundreds of documented examples, a process that is both tedious and highly dependent on specialist knowledge. To address this, we present an AI-assisted retrieval pipeline that combines mark localization with metric-learning fine-tuning across three backbone architectures: an ImageNet-pretrained ResNet-50, a supervised ViT-S/16, and a self-supervised DINOv2 ViT-S/14. We conduct a systematic evaluation of cropping strategies, where we measure the impact of no cropping, manual ground-truth cropping, and learned detection-based cropping, and assess their interaction with each backbone. Our strongest configuration, DINOv2 ViT-S/14 with manual crop and metric-learning fine-tuning, achieves an mAP of 62.63% and a Top-1 accuracy of 73.74%. Our experiments show that self-supervised pretraining and mark localization are the two most impactful factors, with learned cropping recovering the majority of the gain from manual cropping without requiring ground-truth annotations at inference time. To enable reproducibility and adoption in the digital humanities, we release our manually annotated dataset and codebase, and deploy the system via a public web interface.

Chinese Translation

对于艺术历史学家而言，金匠标记在文物的识别和年代确定中起着至关重要的作用。在实际操作中，专家必须手动将查询标记与数百个文献记录的示例进行比较，这一过程既繁琐又高度依赖于专业知识。为了解决这个问题，我们提出了一种AI辅助检索管道，该管道结合了标记定位与三种主干架构的度量学习微调：一个经过ImageNet预训练的ResNet-50，一个监督学习的ViT-S/16，以及一个自监督学习的DINOv2 ViT-S/14。我们对裁剪策略进行了系统评估，测量了无裁剪、手动真实裁剪和基于检测的学习裁剪的影响，并评估了它们与每个主干的交互作用。我们最强的配置是DINOv2 ViT-S/14，结合手动裁剪和度量学习微调，达到了62.63%的平均精度均值（mAP）和73.74%的Top-1准确率。我们的实验表明，自监督预训练和标记定位是两个最具影响力的因素，而学习裁剪在推理时无需真实标注即可恢复手动裁剪的大部分增益。为了促进可重复性和在数字人文学科中的应用，我们发布了手动标注的数据集和代码库，并通过公共网络接口部署了该系统。

View on arXiv Download PDF AI Translation

cs.CV / 87 / 2609.20562

A Dual-Stream Regulated Reconstruction and Segmentation Network with Hierarchical Artifact-Prior Modeling for Ultra-Low-Field Pediatric Neuroimaging

一种双流调节重建与分割网络，具有分层伪影先验建模，用于超低场儿童神经成像

Jafrasteh, Bahram, Milecki, Leo, Zhao, Qingyu

Abstract

Automated quality assessment, enhancement, and segmentation of multiple structures in $0.064\,\mathrm{T}$ ultra-low-field pediatric MRI are limited by a low signal-to-noise ratio, weak anatomical boundaries, and frequent artifacts. We present a unified framework for the LISA 2026 Challenge that performs all three tasks together within one inference pipeline. A network with two coupled streams, built on a 3D U-Net, first reconstructs an enhanced uLF volume and then combines the original and enhanced images for subcortical segmentation. To improve boundary stability, we add an auxiliary class covering brain tissue outside the target structures, derived from whole brain masks. A head conditioned on an artifact graph predicts the seven artifact ratings from reconstruction residuals and frozen segmentation features. We address the scarcity of dense annotations using diffeomorphic registration from atlas to target for label propagation and to regularize anatomical reconstruction. We report validation results across all three tasks.

Chinese Translation

在$0.064 \, ext{T}$超低场儿童MRI中，自动化质量评估、增强和多个结构的分割受到低信噪比、解剖边界模糊以及伪影频繁出现的限制。我们提出了一个统一框架，旨在LISA 2026挑战中在一个推理管道内同时执行这三项任务。该网络基于3D U-Net构建，具有两个耦合流，首先重建增强的超低场体积，然后结合原始图像和增强图像进行皮层下分割。为了提高边界的稳定性，我们添加了一个辅助类别，覆盖目标结构之外的脑组织，该类别来源于全脑掩膜。一个基于伪影图的头部条件预测重建残差和冻结分割特征的七个伪影评分。我们通过从图谱到目标的微分同胚配准来解决密集注释的稀缺性，以进行标签传播并规范解剖重建。我们报告了所有三项任务的验证结果。

View on arXiv Download PDF AI Translation

cs.CV / 88 / 2609.20574

DocAttriBench: Benchmarking Answer Grounding in Document Visual Question Answering

DocAttriBench：文档视觉问答中答案归属的基准测试

De Grandis, Luca, Cappelletti, Silvia, Raccagni, William, Cornia, Marcella, Baraldi, Lorenzo, Cucchiara, Rita

Abstract

Answer grounding in document visual question answering remains an open challenge: most benchmarks lack grounding annotations or provide limited-quality labels, while constructing grounded datasets still requires costly manual effort. We introduce DocAttriBench (DAB), a large-scale benchmark for fine-grained, element-level source attribution in Document VQA, grounding answers to specific layout elements such as text blocks, tables, and images. To build DAB, we propose a Mask-based Perplexity-Derived Attribution method (MAPPET) that combines document layout and language modeling to identify the most informative element for each answer. MAPPET measures the increase in perplexity after masking candidate elements and attributes the answer to the element contributing most to model confidence. Applying MAPPET to multiple existing Document VQA datasets yields DAB, with 237k documents and 296k question-answer pairs with element-level grounding. We benchmark grounding-capable multimodal LLMs on DAB, evaluating answer accuracy, attribution accuracy, and overall answer quality. Results show that while larger models generally achieve higher answer accuracy, even the strongest models often fail to localize the supporting elements. DAB provides a scalable benchmark for developing grounded, verifiable, and trustworthy Document VQA models. Dataset and code are available at https://aimagelab.github.io/DocAttriBench/.

Chinese Translation

在文档视觉问答中，答案归属仍然是一个未解决的挑战：大多数基准缺乏归属注释或提供有限质量的标签，而构建有归属的数据集仍需要昂贵的人工努力。我们引入了DocAttriBench（DAB），这是一个大规模的基准，旨在实现文档视觉问答中细粒度的元素级源归属，将答案归属到特定的布局元素，如文本块、表格和图像。为了构建DAB，我们提出了一种基于掩码的困惑度衍生归属方法（MAPPET），该方法结合了文档布局和语言建模，以识别每个答案最具信息量的元素。MAPPET通过掩盖候选元素后困惑度的增加来衡量，并将答案归属到对模型信心贡献最大的元素。将MAPPET应用于多个现有的文档视觉问答数据集，生成了DAB，包含237,000个文档和296,000个具有元素级归属的问题-答案对。我们在DAB上对具备归属能力的多模态大语言模型进行基准测试，评估答案准确性、归属准确性和整体答案质量。结果表明，尽管较大的模型通常能够实现更高的答案准确性，但即使是最强的模型也常常无法定位支持元素。DAB为开发有归属、可验证和可信的文档视觉问答模型提供了一个可扩展的基准。数据集和代码可在 https://aimagelab.github.io/DocAttriBench/ 获取。

View on arXiv Download PDF AI Translation

cs.CV / 89 / 2609.20589

RawSLAM: Online HDR Gaussian SLAM from Linear Radiance

RawSLAM：基于线性辐射的在线HDR高斯SLAM

González, Marina Orozco, Merino, Luis

Abstract

Current dense visual SLAM systems rely almost exclusively on 8-bit tonemapped Low Dynamic Range (LDR) inputs, limiting their robustness in extreme lighting where shadows and highlights trigger tracking drift and mapping collapse. Conversely, existing raw and High Dynamic Range (HDR) reconstruction pipelines operate strictly offline. They depend on Structure-from-Motion preprocessing and are not suited for large inter-frame motion. We present, to the best of our knowledge, the first online Gaussian SLAM framework that tracks and maps directly on single-exposure 16-bit linear HDR imagery. Our method rests on three core components: an architecture-agnostic HDR Gaussian Splatting module featuring an MLP-free logarithmic parameterization of Gaussian color features; a Reinhard range-compressed photometric objective; and structure-guided spatial gradient weighting. Combined, these components allow our approach to outperform a direct HDR adaptation of MonoGS in both trajectory and reconstruction accuracy, while rendering natively in linear scene radiance for post-rendering processing. The same formulation runs unchanged on standard 8-bit inputs, roughly halving the MonoGS baseline error. Furthermore, our HDR Gaussian module transfers seamlessly to SplaTAM, Gaussian SLAM, and DROID-W, eliminating all tracking failures these systems suffer on challenging illumination sequences. To enable this research, we introduce RawSLAM: a dataset of 10 real-world indoor sequences featuring 16-bit RAW imagery, aligned depth, IMU measurements, and external OptiTrack poses. Code and dataset will be made publicly available soon.

Chinese Translation

当前的密集视觉SLAM系统几乎完全依赖于8位调色的低动态范围（LDR）输入，这限制了它们在极端光照条件下的鲁棒性，因为阴影和高光会导致跟踪漂移和映射崩溃。相反，现有的原始和高动态范围（HDR）重建管道严格在线下操作。它们依赖于运动结构预处理，并不适合大幅度的帧间运动。我们提出了迄今为止首个在线高斯SLAM框架，能够直接在单次曝光的16位线性HDR图像上进行跟踪和映射。我们的方法基于三个核心组件：一个架构无关的HDR高斯溅射模块，采用无多层感知器（MLP）的对数参数化高斯颜色特征；一个Reinhard范围压缩的光度目标；以及结构引导的空间梯度加权。结合这些组件，我们的方法在轨迹和重建精度上超越了MonoGS的直接HDR适配，同时在后处理过程中以线性场景辐射进行原生渲染。相同的公式在标准的8位输入上运行不变，基本上将MonoGS基线误差减半。此外，我们的HDR高斯模块可以无缝转移到SplaTAM、高斯SLAM和DROID-W，消除了这些系统在挑战性光照序列中遭遇的所有跟踪失败。为了支持这项研究，我们引入了RawSLAM：一个包含10个真实室内序列的数据集，特征为16位RAW图像、对齐深度、IMU测量和外部OptiTrack位姿。代码和数据集将很快公开发布。

View on arXiv Download PDF AI Translation

cs.CV / 90 / 2609.20623

PhGS: Post-Hoc Pruning and Refinement of Single-View Feed-Forward 3D Gaussian Reconstructions

PhGS：单视图前馈3D高斯重建的后处理剪枝与精炼

Yagawa, Rinto, Cheng, Han, Schmalstieg, Dieter, Saito, Hideo, Mori, Shohei

Abstract

Recent single-view feed-forward 3D Gaussian Splatting (3DGS) generation predicts a fixed number of Gaussians per camera ray, introducing severe spatial redundancy. Most existing compaction strategies target multi-view setups to exploit cross-view consistency and are incompatible with single-image models. Instead of retraining the base feed-forward network to directly output compact representations, our insight is to keep the base models frozen and apply post-hoc pruning and recurrent refinement to the generated Gaussians. Consequently, we propose a backbone-agnostic compaction pipeline for single-view feed-forward 3DGS that couples an importance-score-based pruning mechanism with a trainable, lightweight recurrent refinement module, which iteratively updates the surviving primitives to restore image quality. Our results demonstrate seamless integration with existing baselines while preserving novel-view rendering fidelity and achieving high memory reduction. Furthermore, our method supports flexible inference-time keep ratios for application needs.

Chinese Translation

最近的单视图前馈3D高斯溅射（3DGS）生成方法在每个相机光线中预测固定数量的高斯分布，导致严重的空间冗余。大多数现有的压缩策略针对多视图设置，以利用视图间的一致性，但与单图像模型不兼容。我们的方法不是重新训练基础前馈网络以直接输出紧凑表示，而是保持基础模型不变，对生成的高斯分布进行后处理剪枝和递归精炼。因此，我们提出了一种与基础网络无关的单视图前馈3DGS压缩管道，该管道结合了基于重要性评分的剪枝机制和可训练的轻量级递归精炼模块，后者迭代更新存活的原始图元以恢复图像质量。我们的结果表明，该方法能够与现有基线无缝集成，同时保持新视图渲染的保真度，并实现高效的内存减少。此外，我们的方法支持灵活的推理时保留比例，以满足应用需求。

View on arXiv Download PDF AI Translation

cs.CV / 91 / 2609.20633

Refinement Is Inherently Editable: Training-Free Prompt-to-Prompt Image Editing with Generative Refinement Network

精细化本质上是可编辑的：基于生成精细化网络的无训练提示到提示图像编辑

Chen, Yulong, Zhang, Ziqian, Zhang, Haoyu, He, Ao, Li, Senmao, Wang, Kai

Abstract

Text-guided image editing must introduce the requested changes while preserving unrelated source content. Diffusion-based editors rely on spatial controls whose inaccuracies can leave edits incomplete or alter unrelated regions. Causal autoregressive editors face a further constraint: their fixed decoding order limits revision of earlier decisions. We introduce RefineEdit, a training-free prompt-to-prompt image editing framework built on a Generative Refinement Network. Our key idea is to couple edit localization with content generation through the global refinement of binary image codes, allowing editing evidence to be reassessed as the image evolves. RefineEdit initializes an editing branch from an intermediate source state, reusing the emerging layout. We compare the probabilities assigned by the two branches to the same source-sampled bits, using their signed differences to select editable positions and bits. Selected bits follow editing refinement, while the remaining bits copy the evolving source state. To stabilize editing across refinement steps, adaptive spatial freezing limits unnecessary mask expansion, while finite bit locking keeps recently selected bits editable. The framework requires no additional training, external masks, or attention control. Across nine editing categories of PIE-Bench, RefineEdit achieves the best background-preservation scores in PSNR, LPIPS, MSE and SSIM, together with the highest whole-image and edited-region CLIP scores among the evaluated methods.

Chinese Translation

文本引导的图像编辑必须在保留无关源内容的同时引入所请求的更改。基于扩散的编辑器依赖于空间控制，其不准确性可能导致编辑不完整或改变无关区域。因果自回归编辑器面临进一步的限制：其固定的解码顺序限制了对早期决策的修正。我们提出了RefineEdit，一个基于生成精细化网络的无训练提示到提示图像编辑框架。我们的关键思想是通过对二进制图像编码的全局精细化，将编辑定位与内容生成结合起来，从而允许在图像演变过程中重新评估编辑证据。RefineEdit从中间源状态初始化编辑分支，重用新出现的布局。我们比较两个分支对相同源采样位的概率分配，利用它们的符号差异选择可编辑的位置和位。所选位遵循编辑精细化，而其余位则复制正在演变的源状态。为了在精细化步骤中稳定编辑，自适应空间冻结限制了不必要的掩码扩展，而有限位锁定保持最近选择的位可编辑。该框架不需要额外的训练、外部掩码或注意力控制。在PIE-Bench的九个编辑类别中，RefineEdit在PSNR、LPIPS、MSE和SSIM中实现了最佳的背景保留分数，并在评估方法中获得了最高的整体图像和编辑区域CLIP分数。

View on arXiv Download PDF AI Translation

cs.CV / 92 / 2609.20638

PROVIA: Procedure State Tracking for Online Mistake Detection in Egocentric Videos

PROVIA：用于自我中心视频在线错误检测的过程状态跟踪

Wen, Di, Yang, Kailun, Weissert, Jimmy, Scherrer, Luc Maria, Zöllner, Cedric, Liu, Ruiping, Chen, Yufan, Wei, Jiale, Zheng, Junwei, Peng, Kunyu

Abstract

An assistant watching egocentric video should notice a mistake from past frames alone, before the next step begins, and keep working once the person recovers. A mistake changes the state of the work, so every later step has to be read against what was done rather than against the plan. The first-mistake protocol that current online methods report on cuts each recording at its first mistake, so a fixed-time rule that never looks at the video is right on every case. We evaluate on complete trials, where mistakes and recoveries arise naturally, under a validation false-alarm budget and against controls that use timing alone. PROVIA keeps two records apart: a factual state, a learned summary of the steps each actor performed, mistakes included, and the accepted progress, an exact posterior over the state of an automaton induced from correct demonstrations by Bayesian state merging and over the execution status of each actor. Procedure-state transitions occur only in the correct-status branch; the mistake and correction branches retain the source state. A sequential test turns the per-frame mistake probability into alarms. With one filter and one optimization rule, PROVIA ranks mistakes best among the evaluated controlled baselines on CaptainCook4D, IndustReal, HoloAssist and IMPACT-ego. At a validation budget of 0.1 false alarms per minute it recalls .154 against .128 on CaptainCook4D and .034 against .015 on HoloAssist, where it leads at every budget. The pipeline runs at 58-70 frames per second. The source code is available at https://github.com/Kratos-Wen/PROVIA.

Chinese Translation

观看自我中心视频的助手应该在下一步开始之前，仅凭过去的帧注意到错误，并在个人恢复后继续工作。错误会改变工作的状态，因此每个后续步骤必须根据已完成的工作而非计划进行评估。目前的在线方法所报告的首个错误协议在每个录制中切断第一个错误，因此一个从不查看视频的固定时间规则在每种情况下都是正确的。我们在完整的试验中进行评估，其中错误和恢复自然发生，遵循验证的误报预算，并与仅使用时间的控制组进行比较。PROVIA将两个记录分开：一个事实状态，即每个参与者执行的步骤的学习总结，包括错误，以及接受的进展，即通过贝叶斯状态合并从正确演示中诱导的自动机状态的精确后验和每个参与者的执行状态。过程状态转换仅发生在正确状态分支中；错误和纠正分支保留源状态。一个顺序测试将每帧的错误概率转化为警报。通过一个过滤器和一个优化规则，PROVIA在评估的控制基线中，在CaptainCook4D、IndustReal、HoloAssist和IMPACT-ego上对错误进行最佳排名。在每分钟0.1个误报的验证预算下，它在CaptainCook4D上召回率为0.154，而控制组为0.128，在HoloAssist上为0.034，而控制组为0.015，在每个预算下均领先。该管道的运行速度为每秒58-70帧。源代码可在https://github.com/Kratos-Wen/PROVIA获取。

View on arXiv Download PDF AI Translation

cs.CV / 93 / 2609.20662

Earth Surface Immune System for Rapid Monitoring of Unknown Anomalies

地表免疫系统用于快速监测未知异常

Li, Jingtao, Zhu, Qian, Wang, Xinyu, Li, Deren, Zhang, Liangpei, Zhong, Yanfei

Abstract

Earth surface anomalies, driven by escalating climate change, and expanding human activities, are increasing in both frequency and diversity, yet their limited historical data and unpredictability make them fundamentally different from conventional remote sensing targets. Existing methods address specific anomaly categories or stop at localization, leaving a gap between detection and actionable information. Here we present ESIA, an Earth Surface Immune System whose architecture is constrained by three principles from the biological immune system, refined over millions of years against equally diverse and uncertain threats. A non-specific innate immune stage treats anomalies as unobserved changes in time-series satellite imagery, generating binary localization maps at 14.51 km2/s without assuming any anomaly category, surpassing the strongest general baseline by 37% in F1. A specific adaptive immune stage applies negative selection to filter text prompts and matches surviving prompts with localized image patches through a multi-modal foundation model, enabling open-vocabulary recognition of unknown anomaly attributes including category, affected area, and damage severity, with recognition F1 exceeding 80%. A mutation mechanism tunes minimal embeddings at test time, adapting to each scene in 3.26s using a single reference image pair. We validate ESIA on a global-scale dataset covering 19,801.60 km2 across six anomaly categories, comparing against 22 models, and further apply it to quantify degraded farmland in the Dnipro Delta following the Kakhovka Dam collapse and assess burn severity from 2025 Palisades Fire in Los Angeles. This unprecedented flexibility in handling unknown anomalies opens new avenues for real-time disaster response and environmental surveillance.

Chinese Translation

地表异常现象因气候变化加剧和人类活动扩展而频繁多样化，但其有限的历史数据和不可预测性使其与传统遥感目标有本质区别。现有方法仅针对特定异常类别或仅限于定位，导致检测与可操作信息之间存在空白。本文提出了ESIA（地表免疫系统），其架构受生物免疫系统的三项原则约束，经过数百万年的演化，针对同样多样且不确定的威胁进行了优化。非特异性的先天免疫阶段将异常视为时间序列卫星影像中的未观察变化，以14.51 km²/s的速度生成二进制定位图，而不假设任何异常类别，F1分数超越最强基线37%。特异性的适应性免疫阶段应用负选择过滤文本提示，并通过多模态基础模型将存活的提示与定位图像补丁匹配，实现对未知异常属性（包括类别、受影响区域和损害严重性）的开放词汇识别，识别F1分数超过80%。突变机制在测试时调整最小嵌入，使用一对参考图像在3.26秒内适应每个场景。我们在覆盖19,801.60 km²的全球数据集上验证了ESIA，涉及六个异常类别，并与22个模型进行了比较，进一步应用于量化卡霍夫卡大坝崩溃后第聂伯三角洲的退化农田，并评估2025年洛杉矶帕利塞德火灾的烧毁严重性。这种前所未有的处理未知异常的灵活性为实时灾害响应和环境监测开辟了新途径。

View on arXiv Download PDF AI Translation

cs.CV / 94 / 2609.20673

FunArt: Decoding Functional Structure and Articulation from Generative 3D Latents

FunArt：从生成的3D潜变量解码功能结构和关节动作

Rotondi, Dennis, Werby, Abdelrhman, Arras, Kai O.

Abstract

To operate effectively in human environments, robots must identify articulated objects, segment their movable and interactive parts, and estimate their kinematic models. Existing articulated scene representations typically recover kinematics from observed interactions, while methods operating on static scans often decouple articulation from functional interactive elements. We present FunArt, a framework that constructs articulation-aware functional 3D scene graphs from posed RGB-D observations captured in a single static configuration. FunArt reconstructs object instances, converts their fused geometry directly into the O-Voxel representation of TRELLIS.2, and exploits its frozen, sparse-compression VAE as a structural prior. A lightweight query-based decoder combines compact object-level latents with dense, surface-aligned features to jointly segment movable parts and functional interactive elements while estimating motion type, axis, origin, and range. On the Articulate3D dataset, FunArt achieves state-of-the-art performance across movable-part segmentation, articulation estimation, and functional-element segmentation, both with and without ground-truth object input. In the end-to-end setting, it outperforms the strongest baselines by 1.5 AP_{50} points for movable parts, 2.8 AP_{50} points under joint origin-and-axis constraints, and 6.7 AP_{50} points for functional elements. These results demonstrate that generative 3D latents encode actionable structural cues that can initialize robotic perception and planning before physical interaction.

Chinese Translation

为了在人工环境中有效操作，机器人必须识别关节物体，分割其可移动和交互部件，并估计其运动学模型。现有的关节场景表示通常从观察到的交互中恢复运动学，而在静态扫描上操作的方法往往将关节动作与功能交互元素解耦。我们提出了FunArt，一个框架，它从在单一静态配置中捕获的姿态RGB-D观测中构建关注关节的功能性3D场景图。FunArt重建物体实例，将其融合几何直接转换为TRELLIS.2的O-Voxel表示，并利用其冻结的稀疏压缩变分自编码器（VAE）作为结构先验。一个轻量级的基于查询的解码器将紧凑的物体级潜变量与密集的、表面对齐的特征结合，以共同分割可移动部件和功能交互元素，同时估计运动类型、轴、起点和范围。在Articulate3D数据集上，FunArt在可移动部件分割、关节估计和功能元素分割方面均实现了最先进的性能，无论是否有真实物体输入。在端到端设置中，它在可移动部件上比最强基线高出1.5 AP_{50}点，在联合起点和轴约束下高出2.8 AP_{50}点，在功能元素上高出6.7 AP_{50}点。这些结果表明，生成的3D潜变量编码了可操作的结构线索，可以在物理交互之前初始化机器人感知和规划。

View on arXiv Download PDF AI Translation

cs.CV / 95 / 2609.20700

Should This Case Be Adapted? Prediction Fragmentation Controls Test-Time Adaptation

这个案例应该被调整吗？预测碎片化控制测试时适应

Wang, Lili, Li, Jing, Sun, Xiaowen, Hu, Xiangyu, Gu, Zhuangzhuang, Liu, Jian, Nelakuditi, Srihari, Tong, Yan

Abstract

Episodic test-time adaptation resets a frozen segmenter to source weights $M_0$ on each case and adapts for a fixed step count. A fixed horizon conflates a cohort-level question, how far to adapt, with an irreducibly per-case one, whether this case should be adapted at all. Cohort means hide that decision: on cross-vendor cardiac MRI the mean $\Delta$Dice from adaptation is statistically indistinguishable from zero while 58.7% of cases are individually made worse. We quantify this harm as harmful accepted area (HA), the harmful fraction of the edited area a controller deploys. Held-out tuning gives a stronger baseline than a fixed horizon, but the budget it selects transfers on neither of the two main medical benchmarks, and no global budget can condition on the case. We show that prediction fragmentation---the disagreement geometry between $M_0$ and the adapted mask $M_k$---predicts HA with no labels or extra backward passes at decision time, comparably on three benchmarks (Spearman $\rho$ 0.50--0.60), at a quarter of gradient-norm's latency. A case-level router built on it cuts HA from 0.228 to 0.139 on a benchmark that took no part in its design, with the design frozen and only cut-points recalibrated there. On the cardiac benchmark the design was selected on, the router cuts HA from 0.129 to 0.013 at matched Dice and 1.10 deployed updates, against the retrospective-best budget found post hoc on evaluation labels, and reduces that 58.7% to 20.0%, an upper bound we quantify. Where the retained cases are not net-helped (as on prostate), the router still cuts HA but concedes accuracy, a boundary we report. Thresholds are fit once on a labeled split disjoint from evaluation; decisions use no labels or gradients. The template ports across architecture and domain (nnU-Net$\to$SegFormer, Cityscapes$\to$ACDC) with coordinate, thresholds and per-bucket actions instantiated per domain.

Chinese Translation

情景测试时适应在每个案例上将冻结的分割器重置为源权重 $M_0$，并进行固定步数的适应。固定的时间范围将一个群体层面的问题（适应多远）与一个不可简化的每个案例的问题（这个案例是否应该被适应）混为一谈。群体的含义掩盖了这一决策：在跨供应商的心脏MRI中，适应后的平均 $ ext{ΔDice}$ 在统计上与零无显著差异，而58.7%的案例在个体上变得更差。我们将这种伤害量化为有害接受区域（HA），即控制器部署的编辑区域中的有害部分。与固定时间范围相比，保留调优提供了更强的基线，但所选择的预算在两个主要医学基准上都无法转移，且没有全球预算可以针对案例进行条件设置。我们展示了预测碎片化——$M_0$与适应后的掩模 $M_k$ 之间的不一致几何——在决策时无需标签或额外的反向传递即可预测HA，在三个基准上表现相当（Spearman $ ho$ 0.50--0.60），延迟为梯度范数的四分之一。基于此构建的案例级路由器在一个未参与设计的基准上将HA从0.228降低到0.139，设计保持不变，仅在此处重新校准切点。在心脏基准上，路由器在匹配的Dice和1.10次部署更新的情况下，将HA从0.129降低到0.013，优于在评估标签上事后发现的回顾最佳预算，并将58.7%降低到20.0%，这是我们量化的上限。在保留的案例未获得净帮助（如前列腺案例）时，路由器仍然减少HA，但牺牲了准确性，这是我们报告的一个边界。阈值在与评估不重合的标记拆分上进行一次拟合；决策不使用标签或梯度。该模板可跨架构和领域移植（nnU-Net$ o$SegFormer，Cityscapes$ o$ACDC），坐标、阈值和每个桶的操作在每个领域中实例化。

View on arXiv Download PDF AI Translation

cs.CV / 96 / 2609.20769

FlowSGS: Improving Flow Matching Priors for Inverse Imaging with Stochastic Interpolants

FlowSGS：利用随机插值改善逆成像的流匹配先验

Li, Tianao, Qian, Xinhui, Alexander, Emma

Abstract

Flow matching has emerged as the state-of-the-art generative model and has been used for plug-and-play (PnP) priors to solve inverse problems in computational imaging. However, existing flow-based inverse solvers assume linear forward models and/or make simplifying approximations in posterior sampling. To circumvent these problems, we introduce FlowSGS, a flow-based posterior sampling method using Split Gibbs Sampling (SGS) to decompose the posterior into a likelihood step and a prior step. Specifically, we sample from the likelihood step using Langevin dynamics and leverage the Stochastic Interpolants (SI) framework to integrate a pretrained flow model into the prior step. We provide a form for the prior step that uses SI's reverse-time SDE, and show connections to previous PnP methods. Moreover, with the aid of the flow prior's straight probability paths and a novel timestep correction technique for the reverse-time SDE, FlowSGS requires fewer network evaluations in its prior step than plug-and-play diffusion samplers. Our experiments show state-of-the-art performance on a range of inverse problems. For the first time, we provide an experiment on a nonlinear inverse problem (Fourier phase retrieval) for flow-based inverse solvers.

Chinese Translation

流匹配已成为最先进的生成模型，并被用于插拔式（PnP）先验以解决计算成像中的逆问题。然而，现有的基于流的逆解算器假设线性前向模型和/或在后验采样中做出简化近似。为了解决这些问题，我们提出了FlowSGS，这是一种基于流的后验采样方法，使用分裂吉布斯采样（SGS）将后验分解为似然步骤和先验步骤。具体而言，我们使用朗之万动力学从似然步骤中进行采样，并利用随机插值（SI）框架将预训练流模型整合到先验步骤中。我们提供了一种先验步骤的形式，该形式使用SI的反向时间随机微分方程（SDE），并展示了与先前PnP方法的联系。此外，借助流先验的直接概率路径和一种新颖的反向时间SDE时间步长校正技术，FlowSGS在其先验步骤中所需的网络评估次数少于插拔式扩散采样器。我们的实验在一系列逆问题上显示了最先进的性能。我们首次针对基于流的逆解算器提供了非线性逆问题（傅里叶相位恢复）的实验。

View on arXiv Download PDF AI Translation

cs.CV / 97 / 2609.20815

ERCPMP-Gx: Endoscopic Image and Video Dataset for Morphological, Histopathological, and Genomic Characterization of Colorectal Polyposis

ERCPMP-Gx：用于结直肠息肉病形态学、组织病理学和基因组特征描述的内窥镜图像与视频数据集

Ghaffari, Zahra, Bahar, Massih, Forootan, Mojgan, Darvishi, Ali, Bolhasani, Hamidreza

Abstract

Hereditary polyposis syndromes can be precursor lesions to colorectal cancer and are associated with a broad spectrum of extracolonic tumors. Early identification and accurate classification of these syndromes are essential for timely diagnosis, individualized patient management, and targeted surveillance strategies for affected families. However, public endoscopic datasets are largely organized around the individual sporadic polyp, and none links the polyposis phenotype to histopathology and germline findings at the patient level. Here, we present ERCPMP-Gx, an endoscopic, histopathological, and genomic dataset developed to support the application of artificial intelligence (AI) in the recognition, characterization, and classification of colorectal polyposis. Most procedures were performed using the Olympus EVIS X1 system with white-light endoscopy (WLE), narrow-band imaging (NBI), magnifying NBI (M-NBI), and NBI with near focus modes, yielding 160 images and accompanying video clips. Approximately eighty percent of cases represent clinically and/or genetically confirmed hereditary polyposis syndromes (PG), including familial adenomatous polyposis (FAP), Peutz-Jeghers syndrome (PJS), juvenile polyposis syndrome (JPS), and ganglioneuroma syndrome (GNS), while the remaining twenty percent comprise non-hereditary polyps and polyp-mimicking lesions with overlapping morphological features (Non-PG), included to support differential classification. Each released record is linked, where available, to standardized endoscopic annotations, representative histopathology, and clinically reported germline findings, forming an AI-ready, patient-level annotation framework. The dataset is publicly accessible at Mendeley (https://doi.org/10.17632/nzyfc544bx.2). For the latest updates and further information, readers are referred to the DataBioX website: https://databiox.com.

Chinese Translation

遗传性息肉病综合征可能是结直肠癌的前驱病变，并与广泛的肠外肿瘤谱系相关。早期识别和准确分类这些综合征对于及时诊断、个体化患者管理以及针对受影响家庭的监测策略至关重要。然而，公共内窥镜数据集主要围绕个体散发性息肉组织组织构建，且没有任何数据集将息肉病表型与患者级别的组织病理学和生殖系发现联系起来。在此，我们呈现ERCPMP-Gx，这是一个内窥镜、组织病理学和基因组数据集，旨在支持人工智能（AI）在结直肠息肉病的识别、特征描述和分类中的应用。大多数程序使用奥林巴斯EVIS X1系统进行白光内窥镜检查（WLE）、窄带成像（NBI）、放大NBI（M-NBI）和近焦模式下的NBI，获得了160幅图像和相应的视频剪辑。约80%的病例代表临床和/或基因确认的遗传性息肉病综合征（PG），包括家族性腺瘤性息肉病（FAP）、皮茨-杰赫综合征（PJS）、幼年息肉病综合征（JPS）和神经节瘤综合征（GNS），而其余20%则包括非遗传性息肉和具有重叠形态特征的息肉模拟病变（Non-PG），以支持鉴别分类。每个发布的记录在可用的情况下与标准化的内窥镜注释、代表性的组织病理学和临床报告的生殖系发现相联系，形成一个适合AI的患者级注释框架。该数据集可在Mendeley上公开访问（https://doi.org/10.17632/nzyfc544bx.2）。有关最新更新和更多信息，读者可访问DataBioX网站：https://databiox.com。

View on arXiv Download PDF AI Translation

cs.CV / 98 / 2609.20816

Paint-Anything: Unified Any-Color Control for Image Generation and Editing

Paint-Anything：统一的任意颜色控制用于图像生成与编辑

Xie, Ji, Zhou, Dewei, Huang, Xinyu, Chen, Zhennan, Wang, Xun

Abstract

Professional design requires any-color control: the ability to specify an object's target color with any 24-bit hex value for image generation and editing. Prior work has explored color generation, editing, and colorization, but often relies on dedicated color representations or specialized inference procedures. Advances in large language models offer a simpler starting point: even compact models can associate hex values with color semantics. We present Paint-Anything, which learns a shared hex-prompt interface for generation and editing through object-level color supervision. We develop a data pipeline that constructs Paint-500K from real images through object grounding, perceptual color labeling, and editing-pair synthesis. Since shadows make real-image labels only approximate colors, we complement this supervision with pure-color anchors whose pixels exactly match their paired hex values. These anchors are used only at high-noise timesteps, leaving low-noise training to natural images. We further introduce Any Color Benchmark (ACBench), comprising ACBench-T2I and ACBench-Edit, to measure object-level hex color fidelity across both tasks. On FLUX.2-4B, Paint-Anything improves ACBench-T2I and ACBench-Edit scores by 85.3% and 28.3%, respectively, relative to the base model, with ablations supporting the training recipe. It also achieves the highest average CompColor score among the compared methods.

Chinese Translation

专业设计需要任意颜色控制：能够使用任意24位十六进制值指定对象的目标颜色，以进行图像生成和编辑。以往的研究探索了颜色生成、编辑和上色，但通常依赖于专用的颜色表示或特定的推理程序。大型语言模型的进展提供了一个更简单的起点：即使是紧凑型模型也能将十六进制值与颜色语义关联起来。我们提出了Paint-Anything，它通过对象级颜色监督学习共享的十六进制提示接口，用于生成和编辑。我们开发了一个数据管道，通过对象定位、感知颜色标注和编辑对合成构建了Paint-500K。由于阴影使得真实图像标签仅能近似颜色，我们用纯色锚点来补充这种监督，这些锚点的像素与其配对的十六进制值完全匹配。这些锚点仅在高噪声时间步使用，而将低噪声训练留给自然图像。我们进一步引入了任意颜色基准（Any Color Benchmark，ACBench），包括ACBench-T2I和ACBench-Edit，以测量两个任务中的对象级十六进制颜色保真度。在FLUX.2-4B上，Paint-Anything相较于基础模型，ACBench-T2I和ACBench-Edit的得分分别提高了85.3%和28.3%，消融实验支持了训练方案。它还在比较方法中实现了最高的平均CompColor得分。

View on arXiv Download PDF AI Translation

cs.CV / 99 / 2609.20817

FAMOS: Feed-Forward 3D Articulation Modeling from Sparse Observations

FAMOS：基于稀疏观测的前馈三维关节建模

Qu, Kevin, Sun, Tao, Viola, Massimiliano, Zhu, Liyuan, Zhou, Zhizhuo, Sarkar, Sayan Deb, Schindler, Konrad, Armeni, Iro

Abstract

Modeling articulated objects from sparse monocular views is challenging because each observation reveals only partial geometry and motion evidence. Most feed-forward methods infer articulation from a single observation and therefore rely heavily on learned category-level shape priors. We present FAMOS, a feed-forward model that predicts movable-part segmentation and joint parameters from a sparse, unordered set of partial point clouds. Our model jointly reasons over multiple observations and naturally supports a variable number of inputs, including a single view. To aggregate articulation cues across observations, we introduce a Multi-state Articulation Transformer with alternating state-wise and global attention. We further propose an observed articulation span objective that supervises the motion range each part exhibits across the input observations, encouraging the model to leverage the full observation set. To overcome the limited scale and diversity of existing datasets, we introduce a procedural data generator that synthesizes self-annotated assets during training. Experiments on PartNet-Mobility, ACD, and ArtiCraft-10K demonstrate consistent improvements over both feed-forward and optimization-based baselines. Project page: https://kevinqu7.github.io/famos

Chinese Translation

从稀疏的单目视图建模关节物体是具有挑战性的，因为每个观测仅揭示部分几何形状和运动证据。大多数前馈方法从单个观测中推断关节，因此在很大程度上依赖于学习的类别级形状先验。我们提出了FAMOS，一种前馈模型，它从稀疏的无序部分点云集中预测可移动部件分割和关节参数。我们的模型对多个观测进行联合推理，并自然支持可变数量的输入，包括单个视图。为了在观测之间聚合关节线索，我们引入了一种多状态关节变换器，采用交替的状态级和全局注意力机制。我们进一步提出了一种观察到的关节跨度目标，监督每个部件在输入观测中表现出的运动范围，鼓励模型利用完整的观测集。为了克服现有数据集规模和多样性有限的问题，我们引入了一种过程数据生成器，在训练过程中合成自注释资产。对PartNet-Mobility、ACD和ArtiCraft-10K的实验表明，相较于前馈和基于优化的基线方法，模型在性能上有一致的提升。项目页面：https://kevinqu7.github.io/famos

View on arXiv Download PDF AI Translation

cs.CV / 100 / 2609.20818

SplashSplat: Reconstructing Splashing Liquids from Real-World Multi-View Videos

SplashSplat：从真实世界多视角视频重建飞溅液体

Liu, Peiyu, Zhang, Dingxi, Tombari, Federico, Pollefeys, Marc, Tsalicoglou, Christina, Barath, Daniel

Abstract

A splash lives for a fraction of a second: sheets tear into ligaments and droplets, appearance is view-dependent and nearly textureless, and little persists long enough to track. Reconstruction research has consequently focused on smoke, synthetic liquids, or gently deforming surfaces. To our knowledge, no synchronized multi-view dataset of splashing liquids exists. We therefore introduce a benchmark of 20 real scenes, from coherent streams to violent splashes, captured by seven synchronized, calibrated 4K cameras at 60 fps, with manually refined per-view liquid and container masks and fixed evaluation splits. We further present SplashSplat, built on a single principle: impose physical structure only where the observations can constrain it. Per-frame liquid SDFs fused from the masks provide the geometry, level-set transport between consecutive SDFs yields a coarse velocity field, and Lagrangian carriers advected along this flow, corrected against each new observation and reseeded where coverage is lost, decode local Gaussians for differentiable rendering. SplashSplat outperforms state-of-the-art dynamic Gaussian splatting methods on our real captures and on a synthetic benchmark, with physically more plausible motion and a lower training cost. The same representation supports temporal interpolation and style transfer without re-optimization.

Chinese Translation

飞溅的瞬间仅持续几分之一秒：液体薄片撕裂成细丝和水滴，外观依赖于视角且几乎没有纹理，且很少有特征能够持续足够长的时间以供追踪。因此，重建研究主要集中在烟雾、合成液体或轻微变形的表面上。据我们所知，目前尚不存在同步的多视角飞溅液体数据集。因此，我们引入了一个基准数据集，包含20个真实场景，从连贯的液流到剧烈的飞溅，由七台同步校准的4K摄像机以60帧每秒捕捉，并手动精细化每个视角的液体和容器掩膜，并固定评估分割。我们进一步提出了SplashSplat，基于一个单一原则：仅在观察结果能够约束的地方施加物理结构。通过掩膜融合的每帧液体SDF提供几何形状，连续SDF之间的水平集传输产生粗略的速度场，拉格朗日载体沿着该流动进行推进，并根据每个新观察进行修正，并在覆盖丢失时重新播种，以解码局部高斯分布以实现可微渲染。SplashSplat在我们的真实捕获和一个合成基准上超越了最先进的动态高斯飞溅方法，展现出更具物理合理性的运动和更低的训练成本。相同的表示方法支持时间插值和风格转移，而无需重新优化。

View on arXiv Download PDF AI Translation

cs.CV / 101 / 2609.20819

Can 4D Foundation Models Remember?

4D基础模型能记住吗？

He, Guangzhao, Averbuch-Elor, Hadar, Ma, Wei-Chiu

Abstract

Perceiving and remembering the visual world is fundamental to navigating and interacting with our environment. Current 4D foundation models, such as camera-controllable video models or 4D reconstruction models, can perceive and reconstruct dynamic environments, but how well they remember what they have perceived remains an open question. Existing benchmarks largely rely on pixel-level metrics and lack ground truth for objects once they leave the field of view, making them unable to evaluate visual memory in an object-centric manner against references. To fill this gap, we introduce PersistBench, a dataset and metric suite that leverages 360{\deg} videos as omniscient ground truth and proposes three evaluation aspects: object permanence, motion continuity, and appearance preservation. Evaluating various models across diverse categories reveals that current models can only maintain short-term consistency that degrades significantly once objects leave the field of view. Our findings highlight the gap between current model capabilities and robust visual memory ("seeing is not remembering"), providing guidance for future development of 4D foundation models. Dataset and code are available on the project page: https://guangzhaohe.com/persistbench.

Chinese Translation

感知和记忆视觉世界是我们导航和与环境互动的基础。目前的4D基础模型，如可控摄像机视频模型或4D重建模型，能够感知和重建动态环境，但它们对所感知内容的记忆能力仍然是一个未解之谜。现有基准测试主要依赖于像素级指标，并且缺乏在物体离开视野后进行评估的真实数据，这使得它们无法以物体为中心的方式对视觉记忆进行评估。为填补这一空白，我们引入了PersistBench，一个利用360°视频作为全知真实数据的数据集和指标套件，并提出了三个评估方面：物体持久性、运动连续性和外观保持性。对不同类别的各种模型进行评估表明，目前的模型只能维持短期一致性，一旦物体离开视野，其一致性显著下降。我们的研究结果突显了当前模型能力与稳健视觉记忆之间的差距（“看见并不等于记住”），为未来4D基础模型的发展提供了指导。数据集和代码可在项目页面获取：https://guangzhaohe.com/persistbench。

View on arXiv Download PDF AI Translation

人工智能 (Artificial Intelligence)

cs.AI / 1 / 2609.19170

Regularized Emphatic Temporal-Difference Learning: Stability under Constant Stepsizes

正则化强调时序差分学习：在恒定步长下的稳定性

Chen, Xingguo, Wu, Zhaohui, Ye, Jinguo, Li, Chao, Yang, Shangdong, Yang, Guang, Liang, Skylar, Wang, Wenhao

Abstract

Emphatic temporal-difference learning (ETD) stabilizes the expected off-policy TD update and changes its projection geometry, but neither property determines constant-stepsize sampled dynamics. We construct an ergodic two-state counterexample in which the ETD mean map contracts while the sampled product has a positive top Lyapunov exponent. Regenerative-cycle analysis separates this sign from the infinite variance of the follow-on trace. We introduce regularized emphatic TD (RETD), a normalized first-order post-shock repair that leaves the trace and importance ratios unchanged, stores the emphatic TD signal in a leaky scalar state, and releases a delayed correction. RETD's raw equilibrium is an affine shift of the ETD equilibrium; single- and two-regularization readouts recover the ETD fixed point exactly. We prove almost-sure convergence for harmonic diminishing stepsizes and a conditional constant-stepsize moment-contraction result from a Markovian random-product bound. RETD has certified negative exponents on the two-state construction and one Baird point, whereas the positive Baird ETD sign remains numerical. Paired 10,000-run experiments validate both separations, fixed-point recovery, a nonmonotone stability region, and task dependence. RETD changes post-shock dynamics; it does not reduce the shared follow-on-trace variance.

Chinese Translation

强调时序差分学习（Emphatic Temporal-Difference Learning, ETD）稳定了期望的离策略时序差分更新，并改变了其投影几何，但这两种特性都不能决定恒定步长的采样动态。我们构造了一个遍历的双状态反例，其中ETD均值映射收缩，而采样乘积具有正的最大Lyapunov指数。再生周期分析将这一符号与后续轨迹的无限方差分开。我们引入了正则化强调时序差分（Regularized Emphatic TD, RETD），这是一种归一化的一阶后冲修复方法，它保持轨迹和重要性比率不变，将强调时序差分信号存储在一个泄漏的标量状态中，并释放延迟修正。RETD的原始平衡是ETD平衡的仿射偏移；单一和双重正则化的读出精确恢复ETD固定点。我们证明了在谐波减小步长下的几乎确定收敛性，以及来自马尔可夫随机乘积界限的条件恒定步长的矩收缩结果。RETD在双状态构造和一个Baird点上具有经过认证的负指数，而正的Baird ETD符号仍然是数值的。成对的10,000次实验验证了这两种分离、固定点恢复、非单调稳定区域和任务依赖性。RETD改变了后冲动态；它并没有减少共享后续轨迹的方差。

View on arXiv Download PDF AI Translation

cs.AI / 2 / 2609.19180

BioPhys-Bridge: A Benchmark for Interdisciplinary Scientific Reasoning in Physics-Grounded Biological Research

BioPhys-Bridge：物理基础生物研究中的跨学科科学推理基准

Xu, Qingyang

Abstract

Language models face unique challenges in analyzing interdisciplinary scientific research literature. In biophysics research, faithful answers require grounding observed data in source evidence, interpreting it through a quantitative physics model, and linking it to a biological mechanism. To address this challenge, we introduce BioPhys-Bridge, a novel benchmark dataset for evidence-grounded scientific reasoning over biophysical literature. Each case contains evidence blocks, stable evidence IDs, quantitative values, units, equations, assumptions, mechanisms, and next decisions as grounding targets for question answering (QA) and retrieval-augmented generation (RAG). The initial release contains 500 cases, 1,517 agent-facing tasks, and covers six biological domains and nine physical model families, including three sparse families reserved for future expansion. We enforce strict quality gates for all cases in schema, evidence-integrity, quantitative-grounding, source-license, duplicate, unit-normalization, with domain expert review and annotation for 81 cases. Preliminary evaluations show that DeepSeek-V4-Flash obtain the highest evidence-ID $F_1$ score (0.360), followed by Qwen3.7-Max (0.316) and GPT-4o-mini (0.294). BioPhys-Bridge is an interdisciplinary benchmark for evaluating attribution, faithfulness, hallucination reduction, and biological experiment design with complex, multi-step scientific reasoning. Future works will increase the size and complexity of the dataset and perform comprehensive evaluations. Code and data are available in the GitHub repository and on Hugging Face.

Chinese Translation

语言模型在分析跨学科科学研究文献时面临独特的挑战。在生物物理研究中，准确的答案需要将观察到的数据与源证据相结合，通过定量物理模型进行解释，并将其与生物机制联系起来。为了解决这一挑战，我们推出了BioPhys-Bridge，这是一个用于基于证据的科学推理的新型基准数据集，专注于生物物理文献。每个案例包含证据块、稳定的证据ID、定量值、单位、方程、假设、机制和下一步决策，作为问题回答（QA）和检索增强生成（RAG）的基础目标。初始版本包含500个案例、1,517个面向代理的任务，涵盖六个生物领域和九个物理模型家族，其中包括三个为未来扩展保留的稀疏家族。我们对所有案例在模式、证据完整性、定量基础、源许可、重复、单位标准化方面实施严格的质量控制，并对81个案例进行了领域专家的审查和注释。初步评估显示，DeepSeek-V4-Flash获得了最高的证据ID $F_1$ 分数（0.360），其次是Qwen3.7-Max（0.316）和GPT-4o-mini（0.294）。BioPhys-Bridge是一个跨学科基准，用于评估归因、准确性、幻觉减少和复杂多步骤科学推理的生物实验设计。未来的工作将增加数据集的规模和复杂性，并进行全面评估。代码和数据可在GitHub仓库和Hugging Face上获取。

View on arXiv Download PDF AI Translation

cs.AI / 3 / 2609.19182

What Do We Expect from LLMs? Mapping the Design of LLM Benchmarks

我们对大型语言模型的期望是什么？大型语言模型基准设计的映射

Wang, Chao

Abstract

Benchmarks are central to how progress in large language models (LLMs) is assessed and communicated. Yet model rankings alone reveal little about how evaluation requirements themselves are changing. The expanding variety of benchmarks offers another perspective: what researchers expect LLMs to do, and what they count as successful performance. We systematically map 14,767 papers introducing or updating evaluation resources from arXiv submissions between January 2022 and August 2026. Using staged screening and automated full-text coding, we examine changes in target systems and domains, evaluation materials and conditions, and scoring mechanisms. The collection shows growing emphasis on action, interaction, and professional applications, while established and newer design elements frequently coexist. Model participation also develops unevenly: LLM-based scoring grows within both agent and non-agent groups, whereas model-generated materials show no comparable sustained increase in recent cohorts. These findings illuminate how public research translates capability expectations into concrete tests and criteria for success. As AI participates in constructing tests, performing tasks, and judging responses, they also raise a question: does expanding evaluation provide more independent evidence, or risk reproducing the preferences and blind spots of its participating models?

Chinese Translation

基准测试在评估和传达大型语言模型（LLMs）进展方面至关重要。然而，仅靠模型排名并不能揭示评估要求本身如何变化。基准测试种类的不断扩展提供了另一种视角：研究人员对LLMs的期望，以及他们认为成功表现的标准。我们系统地映射了在2022年1月至2026年8月期间，从arXiv提交的14,767篇介绍或更新评估资源的论文。通过分阶段筛选和自动全文编码，我们考察了目标系统和领域、评估材料和条件以及评分机制的变化。结果显示，越来越强调行动、互动和专业应用，而既有的设计元素和新兴设计元素常常并存。模型参与的发展也不均衡：基于LLM的评分在代理和非代理组中均有所增长，而模型生成的材料在近期群体中并未显示出相应的持续增长。这些发现阐明了公共研究如何将能力期望转化为具体的测试和成功标准。随着人工智能参与构建测试、执行任务和判断响应，它们也提出了一个问题：扩展评估是否提供了更多独立证据，还是有可能重现参与模型的偏好和盲点？

View on arXiv Download PDF AI Translation

cs.AI / 4 / 2609.19203

Position: It is Time to Virtualize Foundation Models with a Self-evolving Operating System Layer

立场：是时候用自我演化的操作系统层虚拟化基础模型了

Bhattacharya, Suparna, Kumar, Tarun, Xu, Cong, Mopur, Satish Kumar, Li, Jiahao, Mishra, Ashish, Tripathy, Aalap, Koomthanam, Annmary Justine, Foltin, Martin, Foster, Ian

Abstract

AI applications have shifted from single, monolithic foundation models (FM) to compound agentic systems. Yet today's stacks remain fragmented: even as protocols (e.g., MCP, A2A) ease tool/agent connectivity, each framework embeds an implicit runtime for state, memory, budgets, and guardrails, making behavior non-portable and governance brittle. It mirrors computing before operating systems, when every program re-implemented basic services. This position paper argues that the field now needs a Foundation Model Operating System (FMOS) -- a system layer that virtualizes FM interactions analogous to how virtual machines abstract physical hardware, giving applications the illusion of dedicated, trustworthy FM instances with effectively unbounded capabilities. Internally, the FMOS orchestrates knowledge across memory tiers, model selection and resource allocation, and verification and policy enforcement. Like the human brain switching between fast intuition and slow deliberation, the FMOS learns when to intervene and when to let inference proceed directly and continuously adapting its policies based on operational experience.

Chinese Translation

人工智能应用已经从单一的、整体的基础模型（FM）转变为复合智能系统。然而，当前的技术栈仍然是碎片化的：尽管协议（例如，MCP、A2A）简化了工具与智能体之间的连接，但每个框架都嵌入了一个隐式的运行时，用于状态、内存、预算和安全边界，这使得行为不可移植，治理变得脆弱。这种情况类似于操作系统出现之前的计算环境，当时每个程序都重新实现基本服务。本文认为，该领域现在需要一个基础模型操作系统（FMOS）——一个虚拟化基础模型交互的系统层，类似于虚拟机抽象物理硬件，为应用提供专用、可信的基础模型实例的幻觉，具备有效的无限能力。在内部，FMOS协调跨内存层的知识、模型选择和资源分配，以及验证和政策执行。就像人类大脑在快速直觉与缓慢思考之间切换一样，FMOS学习何时干预，何时让推理直接进行，并根据操作经验不断调整其政策。

View on arXiv Download PDF AI Translation

cs.AI / 5 / 2609.19212

What Do Current Systematic Generalization Tasks Miss? A Reasoning-Centered Analysis

当前系统性泛化任务缺失了什么？一个以推理为中心的分析

Qi, Chengwen, Ye, Deheng, Bian, Yatao

Abstract

Systematic generalization, the ability to solve novel problems by recombining known atomic elements, is central to human intelligence but difficult to study rigorously under controlled settings. Existing studies therefore rely on simplifications such as approximately linear action composition, productivity-based tests, and action-explicit goals, which make systematic generalization easier to study but omit some essential aspects of this capability. To characterize what these simplifications miss, we adopt a reasoning-centered lens and introduce TranSGrid, a testbed that brings deductive, inductive, and abductive reasoning together within a unified task. Experiments with seven Transformers on 4,800 TranSGrid instances show that all models perform much worse on TranSGrid than on a held-out test set: the largest model solves 79.6% of the test set, but only 55.3% of TranSGrid and 15.8% of the hardest subset. The gap remains within the training length range, showing that productivity alone is not sufficient to evaluate systematic generalization. Additionally, we reintroduce the other two simplifications into TranSGrid: one variant makes actions compose almost linearly (reducing the inductive demand), the other makes goals action-explicit (reducing the abductive one). In both, solve rates return to roughly the test set level, showing that either simplification alone is enough to reduce TranSGrid to an ordinary held-out test set. Together, our results show that existing tasks reduce either or both of the inductive and abductive demands, and that comprehensively measuring systematic generalization requires a task that involves all three forms of reasoning.

Chinese Translation

系统性泛化是通过重新组合已知的原子元素来解决新问题的能力，这对于人类智能至关重要，但在受控环境下进行严格研究却很困难。因此，现有研究依赖于简化，例如近似线性动作组合、基于生产力的测试和明确的动作目标，这些简化使得系统性泛化更容易研究，但忽略了这一能力的一些重要方面。为了描述这些简化所遗漏的内容，我们采用以推理为中心的视角，并引入TranSGrid，一个将演绎推理、归纳推理和溯因推理结合在统一任务中的测试平台。对七个Transformer模型在4800个TranSGrid实例上的实验表明，所有模型在TranSGrid上的表现远低于在保留测试集上的表现：最大的模型在测试集上解决了79.6%的问题，但在TranSGrid上仅解决了55.3%，在最难的子集上仅解决了15.8%。这一差距在训练长度范围内依然存在，表明仅凭生产力不足以评估系统性泛化。此外，我们将其他两个简化重新引入TranSGrid：一种变体使得动作几乎线性组合（降低了归纳需求），另一种使得目标明确为动作（降低了溯因需求）。在这两种情况下，解决率回到了大致与测试集相当的水平，表明任一简化单独足以将TranSGrid简化为普通的保留测试集。总体而言，我们的结果表明，现有任务减少了归纳和溯因需求中的一个或两个，而全面测量系统性泛化需要一个涉及所有三种推理形式的任务。

View on arXiv Download PDF AI Translation

cs.AI / 6 / 2609.19244

Characterizing Web Search by Conversational LLM Agents: From Search Decisions and Strategies to Results and Responses

通过对话式 LLM 代理表征网络搜索：从搜索决策和策略到结果和响应

Amani, Mahsa, Lee, Seungeon, Dash, Abhisek, Fraihi, Asmaa El, Jang, Yunah, Kirsten, Elisabeth, Wu, Qinyuan, Gummadi, Krishna P., Gupta, Manish, Ravichander, Abhilasha, Zafar, Muhammad Bilal, Das, Soumi

Abstract

Conversational LLM agents increasingly rely on Web search, yet the end-to-end lifecycle of agentic search remains poorly understood. We present the first study of Web search across four major conversational platforms (ChatGPT, Claude, Grok, and DeepSeek), combining real-world user interactions (invivo) with controlled experiments using the same platform's models by their APIs (invitro). We investigate the quality of agentic decisions to invoke Web search, their strategies to formulate queries, the potential domain preferences in the search results they receive, and the choices they make when transforming search results into grounded responses. We find that Web-search decisions vary substantially across platforms and models, while more frequent Web-search invocation does not necessarily yield better response quality. We further show that conversational agents employ different complex querying strategies and that platform specific search engines return search results from their preferred domains. Finally, although responses are largely grounded in search results, some claims rely on uncited search results, raising concerns about attribution and reliability. Our findings have important implications for the design of future AI agents and Web search tools optimized for conversational retrieval.

Chinese Translation

对话式 LLM 代理越来越依赖网络搜索，但代理搜索的端到端生命周期仍然不够明确。我们首次研究了四个主要对话平台（ChatGPT、Claude、Grok 和 DeepSeek）上的网络搜索，结合了真实用户交互（in vivo）与使用相同平台模型的受控实验（in vitro）。我们调查了代理决策调用网络搜索的质量、他们制定查询的策略、在接收到的搜索结果中可能存在的领域偏好，以及他们在将搜索结果转化为有根据的响应时所做的选择。我们发现，网络搜索决策在不同平台和模型之间存在显著差异，而更频繁的网络搜索调用并不一定会带来更好的响应质量。我们进一步表明，对话代理采用不同的复杂查询策略，并且特定平台的搜索引擎返回来自其偏好领域的搜索结果。最后，尽管响应在很大程度上是基于搜索结果的，但一些声明依赖于未引用的搜索结果，这引发了对归属和可靠性的担忧。我们的发现对未来优化对话检索的 AI 代理和网络搜索工具的设计具有重要意义。

View on arXiv Download PDF AI Translation

cs.AI / 7 / 2609.19387

Do AI Agents Understand Computer Architecture?

人工智能代理是否理解计算机架构？

Sharan, Ambika, Chirkov, Grigory, Abbasloo, Soheil

Abstract

Agents are increasingly asked to design hardware, and increasingly reported to succeed. Such reports establish that a design improved; they cannot establish why. An agent that improves an accelerator may be reasoning about the machine, or may be searching competently over knobs whose meaning it never recovers -- and only the first transfers to the next architecture. Existing evaluations cannot tell the two apart, because they vary the agent while holding the framing of the problem fixed. We do the opposite. AutoTuring hands the same agent the same 15-dimensional accelerator space twice: once as named architectural knobs with simulator counters, once as anonymous variables on [0,1], with the evaluator, the legal space and the reachable optima held identical, so that the only thing that varies is whether the problem means anything. The gap between the two is the measurement. On a nine-kernel FP16 GEMM basket, meaning pays: the architect beats a modeled H200 by 5.4% and its blind counterpart by 12.3% on average, with 70.1% fewer simulator calls. It does not pay uniquely: a critic loop recovers most of that gap for the blind agent and buys the architect nothing, so architectural knowledge and structured critique behave as substitutes rather than as complements. We report these as preliminary findings -- five to six runs per condition on a single modeled accelerator -- and take the comparison itself, not the accelerator, to be the contribution.

Chinese Translation

代理越来越多地被要求设计硬件，并且越来越多地报告其成功。这些报告表明设计有所改善；但无法确定原因。一个改进加速器的代理可能是在对机器进行推理，或者可能是在对其从未理解的参数进行有效搜索——而只有前者能够转移到下一个架构。现有的评估无法区分这两者，因为它们在保持问题框架不变的情况下改变代理。我们采取相反的方法。AutoTuring 将同一个代理两次置于相同的15维加速器空间：一次作为带有模拟器计数器的命名架构参数，另一次作为在[0,1]范围内的匿名变量，评估者、合法空间和可达最优解保持不变，因此唯一变化的是问题是否有意义。这两者之间的差距就是测量结果。在一个包含九个内核的FP16 GEMM任务中，意义是有价值的：架构师平均比模型化的H200高出5.4%，比盲目代理高出12.3%，同时减少了70.1%的模拟器调用。这种价值并非唯一：一个批评循环为盲目代理恢复了大部分差距，而对架构师没有任何收益，因此架构知识和结构化批评表现为替代关系而非互补关系。我们将这些视为初步发现——每个条件下在单个模型化加速器上进行五到六次运行——并认为比较本身，而非加速器，是我们的贡献。

View on arXiv Download PDF AI Translation

cs.AI / 8 / 2609.19391

MAGS: Multi-agent Auto-formalization Guarantees Safety for Agentic Outputs

MAGS：多智能体自动形式化保证代理输出的安全性

Wu, Albert, Roberts, Nicholas, Huang, Tzu-Heng, Lin, Haoran, Friedman, Gil, Cho, Sungjun, Orlanski, Gabriel, Sala, Frederic

Abstract

LLM coding agents now generate complex programs at a scale that makes thorough human review increasingly difficult, raising the risk of safety and security failures. Common approaches, including fuzz testing, static analysis, and LLM-as-a-Verifier, can detect many failures but struggle to cover all possible edge cases. Formal verification addresses this by providing machine-checkable guarantees over specified properties, but traditionally demands substantial manual specification and proof engineering. We introduce a unified multi-agent framework, MAGS, that generates executable programs with formal safety guarantees, using Dafny as a verification-aware intermediate representation where safety properties can be mechanically checked. MAGS formalizes and freezes human-audited APIs and safety requirements, translates generated code into Dafny, repairs violations using verifier feedback, and compiles verified programs back into executable code. We evaluate MAGS on 100 CUDA kernels, 100 terminal scripts, and 20 robotic-arm tasks. Across all 220 examples, it achieves a 100% success rate in producing programs with non-trivial safety guarantees against frozen specifications. Independent safety and functional evaluations further show strong performance across all three domains, while revealing failures when the auto-formalized semantics do not fully capture the target behavior.

Chinese Translation

大型语言模型（LLM）编码代理现在能够生成复杂程序，其规模使得彻底的人类审查变得越来越困难，从而增加了安全性和保障失效的风险。常见的方法，包括模糊测试、静态分析和将LLM作为验证器，能够检测许多失效情况，但在覆盖所有可能的边缘案例方面存在困难。形式验证通过提供可机器检查的对指定属性的保证来解决这一问题，但传统上需要大量的手动规范和证明工程。我们提出了一个统一的多智能体框架MAGS，该框架生成具有形式安全保证的可执行程序，使用Dafny作为一种具有验证意识的中间表示，在其中安全属性可以被机械检查。MAGS形式化并冻结经过人类审计的API和安全要求，将生成的代码翻译为Dafny，利用验证器反馈修复违规，并将经过验证的程序编译回可执行代码。我们在100个CUDA内核、100个终端脚本和20个机器人臂任务上评估了MAGS。在所有220个示例中，它在生成具有非平凡安全保证的程序方面达到了100%的成功率，符合冻结的规范。独立的安全性和功能评估进一步显示出在所有三个领域的强大表现，同时揭示了当自动形式化的语义未能完全捕捉目标行为时的失效情况。

View on arXiv Download PDF AI Translation

cs.AI / 9 / 2609.19425

Closed-World Resolution Against Tool Hallucination in LLM Agents

针对大型语言模型代理工具幻觉的封闭世界推理

Iyer, Laxmipriya Ganesh

Abstract

Tool-augmented large language model (LLM) agents fail in a way no tool-selection or tool-security method addresses: they call tools that do not exist and pass arguments no schema declares. Existing defenses either pick the right tool (selection) or constrain what an agent may do with real tools (gating), both of which presuppose the emitted call refers to a real tool at all. We show this is a structural blind spot: a hallucinated call is by construction not a decision any gate made, so no gate can reject it. This paper is primarily a measurement and benchmark study. We give a five-class taxonomy of tool hallucination (H1-H5) and, as a reference point, the Resolution Rung: a training-free, closed-world resolver (registry membership plus a signature check) whose interest is where it must sit, not what it computes. We prove hallucination defense must precede any causal gate, and characterize the one irreducible residue (borrowed arguments schema-indistinguishable from a valid call). Across ten hosted models under two invocation surfaces we measure 322 genuine hallucinations; fabricated-tool calls concentrate on the unconstrained raw-JSON surface (34 vs. 3), and model scale does not help (a 675B model matches a 7-8B one). We then extend to the Model Context Protocol, where merging several servers into one namespace creates hallucination surfaces a single registry cannot express (a second taxonomy, M1-M5); on the live MCP surface we measure 154 hallucinations, including from frontier models that were clean on the single-registry surface, because collisions and shadowing are structural to the merge. We release the versioned Hallucinated-Tools Benchmark (HTB) so any resolver is comparable across submissions.

Chinese Translation

增强工具的大型语言模型（LLM）代理在一种工具选择或工具安全方法无法解决的方式中失败：它们调用不存在的工具，并传递没有任何模式声明的参数。现有的防御措施要么选择正确的工具（选择），要么限制代理可以使用真实工具的方式（限制），这两者都假设发出的调用至少是指向一个真实工具。我们指出这是一个结构性盲点：幻觉调用在构造上并不是任何限制所做的决策，因此没有限制可以拒绝它。本文主要是一个测量和基准研究。我们给出了工具幻觉的五类分类法（H1-H5），并作为参考点提出了解决方案等级（Resolution Rung）：一个无训练、封闭世界的解析器（注册成员资格加上签名检查），其关注点在于它必须处于何处，而不是它计算的内容。我们证明幻觉防御必须先于任何因果限制，并表征了一个不可简化的残余（借用参数在模式上与有效调用不可区分）。在两个调用表面下的十个托管模型中，我们测量了322个真实幻觉；伪造工具调用集中在不受限制的原始JSON表面（34对3），而模型规模并没有帮助（一个675B模型与一个7-8B模型相匹配）。然后我们扩展到模型上下文协议（Model Context Protocol），在这里将多个服务器合并到一个命名空间中创建了单个注册表无法表达的幻觉表面（第二个分类法，M1-M5）；在实时MCP表面上，我们测量了154个幻觉，包括来自在单一注册表表面上干净的前沿模型，因为冲突和遮蔽是合并的结构性特征。我们发布了版本化的幻觉工具基准（Hallucinated-Tools Benchmark，HTB），以便任何解析器在提交中具有可比性。

View on arXiv Download PDF AI Translation

cs.AI / 10 / 2609.19448

The syntax and semantics of goals

目标的句法与语义

Abel, David M., Ho, Mark K.

Abstract

In both cognitive science and computer science, goals are conceptualized as cognitive states that flexibly combine with world knowledge to organize and specify purposeful behavior. In this way, goals are compositional representations whose content relates to rational behavior. We here draw attention to goals as representations and their content because it highlights a parallel with other areas in cognitive science - in particular, the syntax-semantics interface in linguistics and logic - while also foregrounding foundational questions about the expressivity, design, and efficiency of different goal representations. For example, goals are typically taken as fixed and imposing constraints on desirable behaviors, but we can also identify constraints on goal representations themselves, such as whether a particular goal language is sufficiently expressive to capture behaviors of interest, or whether different goal representations capture the same behavior. Here, we synthesize work that aims to characterize the properties of different goal representations and suggest these are points of a broader design space. We close by discussing how distinguishing the form and meaning of goals can elucidate the implicit assumptions we make about goals, inform the study of interactions between higher-level cognition and motivation, and isolate axes of variation for different conceptions of goals.

Chinese Translation

在认知科学和计算机科学中，目标被概念化为与世界知识灵活结合的认知状态，以组织和指定有目的的行为。通过这种方式，目标是内容与理性行为相关的组合表示。我们在此强调目标作为表示及其内容，因为这突出了与认知科学其他领域的平行关系——特别是语言学和逻辑中的句法-语义接口——同时也突显了关于不同目标表示的表达能力、设计和效率的基础性问题。例如，目标通常被视为固定的，并对期望的行为施加约束，但我们也可以识别对目标表示本身的约束，例如某种特定目标语言是否足够表达以捕捉感兴趣的行为，或不同的目标表示是否捕捉相同的行为。在这里，我们综合了旨在表征不同目标表示特性的研究，并建议这些特性是更广泛设计空间的切入点。最后，我们讨论了区分目标的形式和意义如何阐明我们对目标的隐含假设，促进对高级认知与动机之间相互作用的研究，并为不同目标概念的变异轴提供独立的视角。

View on arXiv Download PDF AI Translation

cs.AI / 11 / 2609.19465

Compositional Reasoning in Language Models under Reinforcement Learning Post-Training

强化学习后训练中的语言模型的组合推理

He, Yu, Li, Yingxi, Wang, Yifei, Vitercik, Ellen

Abstract

Compositional reasoning is critical for real-world problem solving: since training data is necessarily limited, models must generalize by composing learned skills in new ways. While post-training methods such as reinforcement learning (RL) have substantially improved the reasoning abilities of language models (LMs), their effects on compositional reasoning remain less well understood. We propose a dependency-graph framework to formalize compositional reasoning, yielding three levels of compositionality with increasing complexity. Empirically, we instantiate this framework with data-structure tasks, which provide deterministic reward computation and clear compositional structure. We find a consistent decomposed-to-composed asymmetry: decomposed-skill training does not reliably transfer to composed tasks, whereas composed-task training transfers more readily back to decomposed tasks. We provide theoretical explanation for this asymmetry, and further evaluate compositional generalization under length extrapolation, structural distribution shift, and transfer to tasks requiring unseen skills. Finally, we present a pilot study on real-world tool-calling benchmarks, showing preliminary evidence that the decomposed-to-composed asymmetry can extend to practical settings.

Chinese Translation

组合推理对于现实世界的问题解决至关重要：由于训练数据必然有限，模型必须通过以新的方式组合已学技能来进行泛化。尽管后训练方法如强化学习（Reinforcement Learning, RL）显著提升了语言模型（Language Models, LMs）的推理能力，但其对组合推理的影响仍不够明确。我们提出了一种依赖图框架来形式化组合推理，产生了三个层次的组合性，复杂性逐渐增加。在实证方面，我们用数据结构任务实例化该框架，这些任务提供了确定性的奖励计算和清晰的组合结构。我们发现了一种一致的分解到组合的非对称性：分解技能训练并不可靠地转移到组合任务，而组合任务训练则更容易转移回分解任务。我们对此非对称性提供了理论解释，并进一步评估了在长度外推、结构分布转移以及转移到需要未见技能的任务下的组合泛化。最后，我们在现实世界工具调用基准上进行了一项初步研究，显示出分解到组合的非对称性在实际环境中可能延伸的初步证据。

View on arXiv Download PDF AI Translation

cs.AI / 12 / 2609.19472

Safety Beyond the Interface: Detecting Harm via Latent States in Large Language Models

超越界面的安全性：通过大型语言模型中的潜在状态检测有害内容

Khatri, Alizishaan, Prabhu, Chiquita, Neogi, Omkar

Abstract

Autonomous systems increasingly rely on Large Language Models (LLMs) yet the safety infrastructure surrounding these models introduces latency and compute overhead. This limits utility in resource-constrained, time-critical deployments. Existing external guardrail models remain blind to the model's internal workings, creating a fundamental assurance gap. We ask: does the model already know when the content is harmful? We extract activations from LLaMA-3.1-8B and train lightweight MLP classifier probes (12.6M parameters) to detect harmful prompts. Evaluated on WildJailbreak, Beavertails, and AEGIS 2.0, our probes achieve F1 scores of 99%, 83%, and 84%, respectively competitive with 1000x larger guard models while cutting latency and compute costs.

Chinese Translation

自主系统越来越依赖大型语言模型（LLMs），然而围绕这些模型的安全基础设施引入了延迟和计算开销。这限制了在资源受限和时间敏感的部署中的实用性。现有的外部保护模型对模型的内部工作机制视而不见，造成了根本的保障缺口。我们提出一个问题：模型是否已经知道内容是否有害？我们从 LLaMA-3.1-8B 中提取激活，并训练轻量级 MLP 分类器探针（12.6M 参数）以检测有害提示。在 WildJailbreak、Beavertails 和 AEGIS 2.0 上评估，我们的探针分别达到了 99%、83% 和 84% 的 F1 分数，具有与 1000 倍更大保护模型相媲美的性能，同时降低了延迟和计算成本。

View on arXiv Download PDF AI Translation

cs.AI / 13 / 2609.19513

QVAC Genesis III: A Large-Scale, High-Quality Open Synthetic STEM Corpus for Efficient Language Model Pre-Training

QVAC Genesis III：一个大规模、高质量的开放合成 STEM 语料库，用于高效的语言模型预训练

Vitabile, Davide, Ranjan, N., Nambiar, Akshay, Gupta, Kamal K., Nazir, Amril

Abstract

High-quality pre-training data is a critical bottleneck for educational and STEM-specific language models targeting edge AI and on-device deployment where token budgets are tightly constrained. While major organizations train ever-larger models on private corpora, the open ecosystem lacks STEM-focused synthetic datasets that deliver high per-token learning value efficiently for small models. To address this gap, we introduce QVAC Genesis III, a 191.43B-token, STEM-focused multi-domain synthetic corpus covering 19 domains across several difficulty levels and different educational styles. QVAC Genesis III is built via a dual generation strategy that performs targeted teacher distillation using a weak edge-scale student model as signal: the student's failures are converted into corrective explanations, while its successes are expanded into contrastive option-level reasoning over all answer choices. We further introduce an LLM-as-a-parser evaluation protocol that extracts final answers from free-form outputs and tracks both accuracy and answer validity. To validate the effectiveness of our QVAC Genesis III data, we conduct controlled from-scratch ablations with 1.7B-parameter models, showing that models trained with QVAC Genesis III consistently outperform both models trained with the open-source synthetic corpus Cosmopedia-v2 and the publicly released Cosmo-1B model across ARC, GPQA Diamond, and MMLU STEM benchmarks, achieving up to +28.57% on ARC-E and +21.35% on ARC-C, while reaching a Valid Answer Rate of up to 99.45%.

Chinese Translation

高质量的预训练数据是针对边缘 AI 和设备部署的教育及 STEM 特定语言模型的关键瓶颈，因为在这些场景中，令牌预算受到严格限制。尽管主要组织在私有语料库上训练越来越大的模型，但开放生态系统缺乏专注于 STEM 的合成数据集，无法高效地为小型模型提供高每令牌学习价值。为了解决这一问题，我们推出了 QVAC Genesis III，这是一个包含 1914.3 亿个令牌的、专注于 STEM 的多领域合成语料库，涵盖 19 个领域，涉及多个难度级别和不同的教育风格。QVAC Genesis III 采用双重生成策略构建，通过使用弱边缘规模的学生模型进行有针对性的教师蒸馏：学生模型的失败被转化为纠正性解释，而成功则扩展为对所有答案选项的对比性选项级推理。我们进一步引入了一种 LLM-as-a-parser 评估协议，该协议从自由形式的输出中提取最终答案，并跟踪准确性和答案有效性。为了验证 QVAC Genesis III 数据的有效性，我们进行了控制的从零开始的消融实验，使用 17 亿参数的模型，结果显示，使用 QVAC Genesis III 训练的模型在 ARC、GPQA Diamond 和 MMLU STEM 基准测试中始终优于使用开源合成语料库 Cosmopedia-v2 和公开发布的 Cosmo-1B 模型的模型，在 ARC-E 上提高了高达 28.57%，在 ARC-C 上提高了 21.35%，同时有效答案率达到 99.45%。

View on arXiv Download PDF AI Translation

cs.AI / 14 / 2609.19515

LLM-as-an-Improver: Turning Verification into Better Candidates

LLM作为改进者：将验证转化为更好的候选者

Tomihari, Akiyoshi, Ichikawa, Yuma

Abstract

Verifier-based selection improves LLM performance by generating multiple candidate solutions and using a verifier to select the most promising one. However, existing methods typically treat verification only as a ranking step and discard its feedback once a fixed candidate pool has been evaluated. In this paper, we ask whether verification can also improve the candidate set itself. To this end, we introduce LLM-as-an-Improver and propose Verify--Repair--Reselect (VRR), which uses verification feedback to generate and reselect improved candidates. VRR retains the initial winner while conditionally generating three complementary alternatives: repaired versions of the winner and runner-up, and a solution based on a new approach. It filters invalid and duplicate candidates using only inference-time information and then reselects the final answer under the original evaluation criteria. Across diverse models and code-generation and reasoning benchmarks, VRR improves over fixed-pool verifier-based selection in many settings and can recover correct solutions even when all candidates in the initial pool are incorrect. These results highlight a broader role for LLMs as improvers: verification feedback can not only select among existing solutions but also construct stronger candidates beyond the initial pool.

Chinese Translation

基于验证者的选择通过生成多个候选解决方案并使用验证者选择最有前景的一个，从而提高了大型语言模型（LLM）的性能。然而，现有的方法通常仅将验证视为排名步骤，并在评估固定候选池后丢弃其反馈。本文探讨了验证是否也可以改善候选集本身。为此，我们引入了LLM作为改进者，并提出了验证-修复-重新选择（Verify--Repair--Reselect, VRR）方法，该方法利用验证反馈生成并重新选择改进的候选者。VRR保留初始赢家，同时有条件地生成三个互补的替代方案：赢家和亚军的修复版本，以及基于新方法的解决方案。它仅使用推理时的信息过滤无效和重复的候选者，然后在原始评估标准下重新选择最终答案。在多种模型以及代码生成和推理基准测试中，VRR在许多设置中优于固定池的基于验证者的选择，并且即使初始池中的所有候选者都不正确，也能恢复正确的解决方案。这些结果突显了大型语言模型作为改进者的更广泛角色：验证反馈不仅可以在现有解决方案中进行选择，还可以构建超越初始池的更强候选者。

View on arXiv Download PDF AI Translation

cs.AI / 15 / 2609.19519

An Architecture for Long-Horizon Agents: Levels, Ticks and Cascaded Intelligence

长远代理的架构：层级、时钟和级联智能

Nijkamp, Erik, Koul, Anurag, Pakhomov, Egor, Pang, Bo

Abstract

Language-model agents are increasingly asked to carry out work spanning days or weeks, such as an operations remediation or a research programme. Such a task outlives any context window, any process and any interval at which a person can attend. In this paper, we argue that a long-horizon agent must run continually without forgetting before it can learn continually. This ability lies in the harness around the model rather than in the model itself. We derive seven bottlenecks from the long-horizon setting and answer them with a hierarchical architecture of three parts: (i) levels indexed by time scale, each keeping a bounded file summarising the level below; (ii) a clocked tick as the unit of autonomous action; and (iii) cascaded intelligence, where work is escalated to a more capable model only after failing review. We report on a ten-day campaign in which an agent built on this architecture reproduced a published reinforcement-learning result with a human attending once a day, and show (1) the agent kept the thread across every context reset and session boundary of the campaign, (2) operating knowledge written early changed later behaviour with no change to model weights, and (3) where learned components would enter such a system. Overall, our experience suggests continual learning for these agents needs a substrate outliving every context and process, and the checks the harness already runs are where a learner belongs.

Chinese Translation

语言模型代理越来越多地被要求执行跨越数天或数周的工作，例如操作修复或研究项目。这类任务超出了任何上下文窗口、任何过程和任何人能够关注的时间间隔。本文论证了长远代理必须持续运行而不遗忘，才能实现持续学习。这种能力在于模型周围的支撑结构，而非模型本身。我们从长远设置中推导出七个瓶颈，并通过一个由三部分组成的层级架构来解决这些瓶颈：(i) 按时间尺度索引的层级，每个层级保持一个有限的文件，总结下层级的内容；(ii) 作为自主行动单位的定时时钟；(iii) 级联智能，工作在未通过审核之前不会升级到更强大的模型。我们报告了一项为期十天的活动，其中基于该架构的代理在每天有一个人参与的情况下重现了已发布的强化学习结果，并展示了：(1) 代理在活动的每次上下文重置和会话边界中保持了思路的连贯性，(2) 早期写入的操作知识改变了后来的行为，而模型权重没有变化，以及 (3) 学习到的组件将如何进入这样的系统。总体而言，我们的经验表明，这些代理的持续学习需要一个超越每个上下文和过程的基础结构，而支撑结构中已经运行的检查正是学习者所需的地方。

View on arXiv Download PDF AI Translation

cs.AI / 16 / 2609.19523

EconSkills: Studying Skill Transfer and Retrieval for Web Agents on Live Economic Data

EconSkills：研究网络代理在实时经济数据上的技能转移与检索

Quan, Yinzhu, Liu, Zefang

Abstract

Web agents often revisit the same sites, yet most evaluations discard the procedures learned in earlier successful interactions. We introduce EconSkills, a skill library and evaluation framework that distills verified EconWebArena trajectories into parameterized standard operating procedures for retrieving live economic data. Each skill records its scope, navigation procedure, site-specific guidance, verification checks, and recovery steps while replacing source-instance values with placeholders. EconSkills separates two questions: whether a known relevant procedure transfers to a held-out task, and whether an agent can retain that benefit when selecting from a library. In controlled transfer, matched skills improve success over no-skill prompting and require fewer steps on paired successes, while abstraction is substantially more effective than replaying raw trajectories. At library scale, retrieval is competitive with the no-skill baseline overall and performs best on directly covered tasks; coverage-stratified outcomes show that approximate matches on uncovered tasks offset these gains. Browser trajectories further identify when procedural guidance shortens portal-specific navigation and when semantic verification remains necessary. These results establish that reusable economic web procedures can transfer across task instances and provide a concrete design target for coverage-aware selection and context delivery.

Chinese Translation

网络代理经常重访相同的网站，但大多数评估忽略了在早期成功互动中学到的程序。我们引入了EconSkills，一个技能库和评估框架，它将经过验证的EconWebArena轨迹提炼为参数化的标准操作程序，以检索实时经济数据。每个技能记录其范围、导航程序、特定网站指导、验证检查和恢复步骤，同时用占位符替换源实例值。EconSkills区分了两个问题：已知的相关程序是否能够转移到保留任务上，以及代理在从库中选择时是否能保留这种优势。在受控转移中，匹配的技能在成功率上优于无技能提示，并且在配对成功中需要更少的步骤，而抽象的效果显著优于重放原始轨迹。在库规模下，检索的整体表现与无技能基线相当，并在直接覆盖的任务上表现最佳；覆盖分层结果显示，在未覆盖任务上的近似匹配抵消了这些收益。浏览器轨迹进一步识别了程序指导何时缩短特定门户的导航，以及何时仍需进行语义验证。这些结果表明，可重用的经济网络程序可以跨任务实例转移，并为覆盖感知选择和上下文传递提供了具体的设计目标。

View on arXiv Download PDF AI Translation

cs.AI / 17 / 2609.19524

A Unified Evaluation Framework for Trustworthy Large Language Models, Agentic AI, and Multimodal Systems

一个统一的评估框架用于可信的大型语言模型、智能代理人工智能和多模态系统

Raza, Shaina, Radwan, Ahmed Y., Liaquat, Imran, Hume, Kathryn

Abstract

Benchmark scores alone provide an incomplete basis for assessing the trustworthiness of modern artificial intelligence systems. Large language models (LLMs), agentic systems, and multimodal models (MLLMs) require different forms of assessment, yet their evaluation evidence must remain interpretable for development and oversight. We propose a unified framework that connects output-level, trajectory-level, and cross-modal assessment through eight trustworthiness dimensions: capability, robustness, safety, fairness, transparency, governance, oversight, and efficiency. The framework preserves system-specific metrics while mapping native measurements to common performance bands, accompanied by uncertainty estimates and traceable evidence. A meta-evaluation layer examines the validity, reliability, and reproducibility of the evaluation itself. Multidimensional profiles expose strengths and weaknesses, while safety-critical overrides prevent aggregate scores from masking critical failures. Mappings to governance frameworks, international standards, and European Union regulatory requirements connect technical assessment with oversight needs. The framework provides a structured basis for assessing both system performance and the credibility of the evidence supporting it, with empirical validation across deployment contexts remaining an essential next step.

Chinese Translation

仅依靠基准分数无法全面评估现代人工智能系统的可信性。大型语言模型（LLMs）、智能代理系统和多模态模型（MLLMs）需要不同形式的评估，但其评估证据必须保持可解释性，以便于开发和监督。我们提出了一个统一的框架，通过八个可信性维度（能力、稳健性、安全性、公平性、透明性、治理、监督和效率）连接输出级别、轨迹级别和跨模态评估。该框架保留了系统特定的指标，同时将原生测量映射到共同的性能区间，并附有不确定性估计和可追溯证据。一个元评估层审查评估本身的有效性、可靠性和可重复性。多维度的概况揭示了优势和劣势，而安全关键的覆盖措施防止汇总分数掩盖关键失误。与治理框架、国际标准和欧盟监管要求的映射将技术评估与监督需求相连接。该框架为评估系统性能及其支持证据的可信性提供了结构化基础，同时在部署环境中进行实证验证仍然是一个重要的下一步。

View on arXiv Download PDF AI Translation

cs.AI / 18 / 2609.19526

Self Improvement via Fast Tree-search

通过快速树搜索实现自我改进

Fu, Xinghong, Kulanthaivelu, Aravinth, Yamada, Yutaro

Abstract

Coding agents can recursively modify their own implementations, forming a loop of self-improvement. While prior work shows this can boost performance on coding benchmarks, existing approaches are costly and compute-intensive. We introduce a simple, sample-efficient self-improvement framework that significantly improves coding performance under strict budget constraints. We identify evaluation of candidate self-modifications as the main runtime bottleneck since prior approaches estimate their effectiveness by re-running a subset of benchmark tasks with the modified agent, which is time-consuming. We introduce Recursive Self Improvement via Fast Tree-search (SIFT), which augments these downstream task evaluations with an LLM-as-a-judge signal that performs pairwise comparisons between candidate patches, where the win-loss record is aggregated with a regularized Bradley-Terry model, and the resulting strength scores drive rank-based parent sampling inside a lightweight disaggregated tree search. Expensive downstream task evaluations are reserved only for the most promising nodes. Using a fully disaggregated tree search pipeline, the judge scores provide intermediate signal to guide exploration on promising candidate patches without being bottlenecked by slow evaluation runs. SIFT outperforms existing tree-search based self-evolution frameworks on the full Polyglot benchmark with significantly lower resource requirements in terms of CPU hours, wall clock time, and API cost.

Chinese Translation

编码代理可以递归地修改自身实现，形成自我改进的循环。尽管先前的研究表明这可以提升编码基准的性能，但现有的方法成本高且计算密集。我们提出了一种简单且样本高效的自我改进框架，在严格的预算约束下显著提高编码性能。我们识别出候选自我修改的评估是主要的运行时瓶颈，因为先前的方法通过对修改后的代理重新运行基准任务的子集来估计其有效性，这一过程耗时较长。我们引入了通过快速树搜索的递归自我改进（SIFT），该方法通过 LLM-as-a-judge 信号增强了下游任务评估，该信号在候选补丁之间进行成对比较，并通过正则化的 Bradley-Terry 模型汇总胜负记录，最终生成的强度评分驱动轻量级分散树搜索中的基于排名的父样本选择。昂贵的下游任务评估仅保留给最有前景的节点。通过完全分散的树搜索管道，评判者评分提供了中间信号，以指导对有前景的候选补丁的探索，而不受缓慢评估运行的瓶颈限制。SIFT 在完整的 Polyglot 基准测试中超越了现有的基于树搜索的自我进化框架，并在 CPU 小时、实际时间和 API 成本方面显著降低了资源需求。

View on arXiv Download PDF AI Translation

cs.AI / 19 / 2609.19530

When Hiring Becomes Agent-Mediated: Evaluating Access and Recurrence in Two-Agent R\'esum\'e Screening

当招聘变为代理中介：评估双代理简历筛选中的访问和重复性

Gao, Jian, Jiang, Hang

Abstract

Hiring is bilateral: employers assess fit, while candidates present and defend evidence of their qualifications. Yet r\'esum\'e screening, the first gate, is commonly automated as a static, one-call judgment over a r\'esum\'e-job pair. We study a two-agent alternative in which employer-side and candidate-side agents represent these roles, exchange evidence, and update their judgments before deciding who advances. We compare procedures on 600 constructed r\'esum\'e-job pairs using GPT-5.5 and Claude Opus 4.7. Two-agent screening advances more applications (33.3% to 39.3% for GPT-5.5; 34.0% to 35.5% for Opus 4.7). Across three runs on the common 191-pair borderline pool, pass-instance rates rise from 4.5% to 26.2% and from 6.5% to 16.1%, respectively. This is not a uniform relaxation: two-agent screening rejects applications one-call advances, changing decisions in both directions. At similar pass volumes, the procedures advance different applications, and no one-call threshold recovers applications consistently selected by two-agent screening. Among discovery-selected cases re-executed in fresh runs, two-agent-only selections recur less often than shared selections, clearly under GPT-5.5 and less certainly under Opus 4.7, while a separate one-call follow-up shows no comparable decline. As hiring becomes agent-mediated on both sides, the screening procedure, not only the model behind it, shapes who reaches human review and how reliably that access recurs.

Chinese Translation

招聘是双向的：雇主评估适配性，而候选人展示并辩护其资格的证据。然而，简历筛选作为第一道关卡，通常被自动化为对简历-职位配对的静态、一致性判断。我们研究了一种双代理替代方案，其中雇主端和候选人端的代理代表这些角色，交换证据，并在决定谁能晋级之前更新他们的判断。我们使用 GPT-5.5 和 Claude Opus 4.7 比较了在 600 个构建的简历-职位配对上的程序。双代理筛选推进了更多申请（GPT-5.5 从 33.3% 提升至 39.3%；Opus 4.7 从 34.0% 提升至 35.5%）。在对共同的 191 对边界池进行的三次运行中，合格率从 4.5% 上升至 26.2%，从 6.5% 上升至 16.1%。这并不是一种统一的放宽：双代理筛选拒绝了单次电话推进的申请，在两个方向上改变了决策。在相似的合格量下，各程序推进了不同的申请，且没有一个单次电话阈值能够一致地恢复由双代理筛选选中的申请。在重新执行的新运行中，双代理仅选择的案例重复出现的频率低于共享选择，GPT-5.5 下的情况明显，而在 Opus 4.7 下则不那么确定，同时单次电话的后续检查显示没有可比的下降。随着招聘在双方都变为代理中介，筛选程序不仅是其背后的模型，塑造了谁能进入人工审核以及这种访问的可靠性如何重复。

View on arXiv Download PDF AI Translation

cs.AI / 20 / 2609.19538

Agentic AI Networking for Heterogeneous Unmanned Aerial Systems in Low-Altitude Wireless Networks

低空无线网络中异构无人机系统的自主智能网络

Quang, Nguyen Duc Minh, Liu, Chang, Li, Shuangyang, Ng, Derrick Wing Kwan

Abstract

Low-altitude wireless networks (LAWNs) are emerging as a key infrastructure for heterogeneous unmanned aerial systems that support concurrent services within a shared three-dimensional airspace. Their coexistence creates strong coupling among mobility, connectivity, and shared network resources, while heterogeneous services impose distinct and time-varying requirements. These interactions naturally form a dynamic non-cooperative game in which both operating conditions and coordination objectives evolve over time. Conventional optimization and learning-based controllers typically rely on predefined objectives, limiting their ability to adapt autonomously to changing service requirements and resource priorities. To address this challenge, we propose a hierarchical hybrid large language model (LLM)- multi-agent reinforcement learning (MARL) architecture organized as a dual-loop structure. Specifically, an outer adaptation loop employs LLM-assisted game orchestration to interpret service requirements and operator intent, and reconfigure objectives and resource priorities, while an inner loop executes decentralized, parameter-conditioned MARL policies under the configured game. A logistics-monitoring case study illustrates how the proposed framework facilitates coordinated coexistence among heterogeneous services, adapting to evolving operating conditions without retraining the underlying MARL policies. Finally, we discuss key challenges and research directions toward scalable, trustworthy, and adaptive agentic LAWNs.

Chinese Translation

低空无线网络（LAWNs）正成为支持异构无人机系统在共享三维空域内并发服务的关键基础设施。它们的共存在移动性、连接性和共享网络资源之间产生了强耦合，而异构服务则施加了不同且随时间变化的需求。这些相互作用自然形成了一个动态的非合作博弈，其中操作条件和协调目标随时间演变。传统的优化和基于学习的控制器通常依赖于预定义的目标，限制了它们在面对变化的服务需求和资源优先级时的自主适应能力。为了解决这一挑战，我们提出了一种分层混合的大语言模型（LLM）-多智能体强化学习（MARL）架构，组织为双环结构。具体而言，外部适应环利用LLM辅助的博弈编排来解释服务需求和操作意图，并重新配置目标和资源优先级，而内部环则在配置的博弈下执行去中心化的、参数条件化的MARL策略。一个物流监控案例研究展示了所提框架如何促进异构服务之间的协调共存，适应不断变化的操作条件，而无需重新训练基础的MARL策略。最后，我们讨论了实现可扩展、可信赖和自适应自主低空无线网络的关键挑战和研究方向。

View on arXiv Download PDF AI Translation

cs.AI / 21 / 2609.19551

Continual Enterprise World Model Discovery in Dynamic Systems

动态系统中的持续企业世界模型发现

Mishra, Shambhavi, Vazquez, David, Taslakian, Perouz, Pedersoli, Marco, Dolz, Jose, Laradji, Issam H.

Abstract

In an enterprise system, updating one field can set another, create a record, or start an approval. These effects are produced by business rules that are not built into the platform but written by each organization and revised over time. An agent working in such a system cannot predict the result of its own actions without knowing these rules. We study continual enterprise world model discovery, where an agent starts without knowledge of these business rules and discovers them by interacting with records and observing the outcomes. From those observations it builds a world model, which it revises as the rules change. To evaluate this, we introduce EnterpriseWorldShift, built on a live ServiceNow environment with nine tables, 25 hidden rules and 600 evaluation actions. It presents four versions of the same enterprise world, with the tables and records held fixed while a rule is modified, then added, then removed, so that discovery, revision, extension and retirement are each tested in turn. Our Continual Discovery Agent (CDA) builds such a model and carries it from one world to the next. It predicts the effects of the hidden rules more accurately than looking them up for each question, the approach taken by prior work, by up to 8.98 IoU points, and it answers from its own model without querying the running system.

Chinese Translation

在企业系统中，更新一个领域可能会影响另一个领域、创建记录或启动审批。这些影响是由业务规则产生的，这些规则并不是内置于平台中的，而是由每个组织编写并随着时间的推移进行修订。在这样的系统中工作的代理无法在不知道这些规则的情况下预测自己行为的结果。我们研究持续的企业世界模型发现，其中代理在没有业务规则知识的情况下开始，通过与记录的交互和观察结果来发现这些规则。通过这些观察，代理构建了一个世界模型，并在规则变化时进行修订。为了评估这一过程，我们引入了EnterpriseWorldShift，该系统基于一个包含九个表、25条隐藏规则和600个评估动作的实时ServiceNow环境。它展示了同一企业世界的四个版本，在修改、添加和移除规则的过程中，表和记录保持不变，从而依次测试发现、修订、扩展和退休。我们的持续发现代理（CDA）构建了这样的模型，并将其从一个世界带到下一个世界。与之前的研究通过查询每个问题的隐藏规则的方法相比，CDA能够更准确地预测隐藏规则的影响，准确度提高了多达8.98个IoU点，并且能够从其自身模型中回答问题，而无需查询运行中的系统。

View on arXiv Download PDF AI Translation

cs.AI / 22 / 2609.19610

SIMLIFE: Pattern Understanding for Long-Horizon Human-Agent Partnership

SIMLIFE：长时间人机合作的模式理解

Peng, Run, Nie, Zinnia, Ding, Jing, Dai, Yinpei, Zhang, Yichi, Wu, Zengqing, Fu, Yao, Ma, Ziqiao, Mao, Jiayuan, Chai, Joyce

Abstract

Understanding humans over long horizons requires agents to infer not only what people need in the moment, but also how routines form, why they repeat, and when they change. We introduce SimLife, a scalable platform for simulating long-term household life with rich visual observations, ground-truth action logs, and synthetic dialogues with audio. Built on SimLife, SimLife-BP evaluates long-context pattern understanding: the ability to infer latent behavioral rules from weeks or months of everyday observations. The benchmark contains 106 episodes averaging 15.49 hours and 38.57 in-game days, and 1,439 question-answer pairs. Each task probes direct, counterfactual, noisy, and inverse reasoning under different levels of rule hints. Evaluating frontier models and architectures, we find that current models often achieve surface-level prediction without comprehensive rule understanding, rely on frequency-based heuristics rather than if-then reasoning over evidence, and struggle to adapt when behavioral patterns change. These findings suggest that long-context pattern understanding remains a major bottleneck for future embodied agents, while SimLife opens a broader space for studying memory, personalization, adaptation, and long-horizon planning in everyday human-AI interaction.

Chinese Translation

理解人类在长时间范围内的行为需要代理不仅能够推断人们当下的需求，还要理解日常行为如何形成、为何重复以及何时发生变化。我们介绍了SimLife，这是一个可扩展的平台，用于模拟长期家庭生活，提供丰富的视觉观察、真实的行动日志以及带音频的合成对话。在SimLife的基础上，SimLife-BP评估了长上下文模式理解的能力：从数周或数月的日常观察中推断潜在行为规则的能力。该基准包含106个情节，平均时长为15.49小时，游戏内天数为38.57天，以及1,439对问答。每个任务在不同的规则提示级别下探测直接、反事实、噪声和逆向推理。评估前沿模型和架构时，我们发现当前模型往往只实现表层预测，而没有全面理解规则，依赖于基于频率的启发式方法而非基于证据的如果-那么推理，并且在行为模式变化时难以适应。这些发现表明，长上下文模式理解仍然是未来具身代理的主要瓶颈，而SimLife则为研究记忆、个性化、适应性和日常人机交互中的长期规划开辟了更广阔的空间。

View on arXiv Download PDF AI Translation

cs.AI / 23 / 2609.19630

From Intent to Action: Benchmarking LLM Safety in Vehicle Voice Command Authorization

从意图到行动：车辆语音命令授权中的大型语言模型安全性基准测试

Afroze, Diba, Zhang, Xingli, Tu, Yazhou, Hei, Xiali

Abstract

Large language models (LLMs) are increasingly integrated into vehicle voice assistants. But linking natural-language requests to vehicle functions creates a safety-critical authorization problem. Before executing a command, the system must choose whether to execute, refuse, clarify, require confirmation, defer to manual control, trigger an emergency response, or make no tool call. To our knowledge, prior evaluations do not isolate this pre-action decision across speaker role, authentication status, vehicle state, and tool availability. We introduce a 202-scenario benchmark with Reference Decisions under a seven-class taxonomy. We evaluate two local open-weight models and three API-based LLMs using Decision Alignment and safety-specific error metrics. Alignment ranges from 40.1% for Llama 3.2 3B to 89.1% for Gemini 3.1 Pro Preview. The API-based models score between 83.2% and 89.1%, with no statistically significant differences among them. Even these models produce two to three False Executes among 161 non-execution scenarios, and persistent errors remain in confirmation and manual-control decisions. A controlled Llama 3.2 3B ablation increases alignment to 40.1% under the structured authorization policy, versus 28.2-29.2% under schema-only and generic-safety baselines, but it does not eliminate False Executes. Structured LLM decisions are therefore insufficient as a standalone safety mechanism, and deployment requires an independent enforcement layer that verifies tool permissions and vehicle-state constraints before invoking any vehicle function.

Chinese Translation

大型语言模型（LLMs）越来越多地被集成到车辆语音助手中。然而，将自然语言请求与车辆功能关联起来会产生一个安全关键的授权问题。在执行命令之前，系统必须选择是执行、拒绝、澄清、要求确认、推迟手动控制、触发紧急响应，还是不进行任何工具调用。根据我们的了解，之前的评估并未在说话者角色、认证状态、车辆状态和工具可用性之间隔离这一前行动决策。我们引入了一个包含202种场景的基准测试，并在七类分类法下提供参考决策。我们使用决策一致性和安全特定错误指标评估了两个本地开放权重模型和三个基于API的LLMs。Llama 3.2 3B的对齐率为40.1%，而Gemini 3.1 Pro Preview的对齐率为89.1%。基于API的模型得分在83.2%到89.1%之间，且它们之间没有统计学上显著的差异。即便如此，这些模型在161个非执行场景中仍然产生了两到三个错误执行，并且在确认和手动控制决策中仍然存在持续的错误。经过控制的Llama 3.2 3B消融实验在结构化授权政策下将对齐率提高至40.1%，而在仅使用模式和通用安全基线下则为28.2-29.2%，但并未消除错误执行。因此，结构化LLM决策作为单独的安全机制是不够的，部署需要一个独立的执行层，在调用任何车辆功能之前验证工具权限和车辆状态约束。

View on arXiv Download PDF AI Translation

cs.AI / 24 / 2609.19636

Reach or Solve? Attributing Agentic RL Gains with Checkpoint Handoffs

达到还是解决？通过检查点交接归因于自主强化学习的收益

Liu, Xuan, Qian, Jingbin

Abstract

Reinforcement learning now trains language-model agents that act over dozens of steps in live environments. The gains are large, and they are read as better decision-making. An agent in a closed loop writes its own inputs. Each observation follows from its own earlier actions, so the states it meets late in an episode are partly of its own making. An SFT checkpoint and an RL checkpoint are then scored from different states, even on identical tasks. Endpoint success mixes two changes: where the agent arrives, and what it does once it is there. Restricting the comparison to states both policies reach does not separate them. That restriction selects on an outcome, and in our data it flips the sign of the effect. We introduce checkpoint handoff, an evaluation protocol that clones a state one released checkpoint reached and hands it to another, with no retraining. Crossing a reacher role and a solver role over SFT and RL splits an endpoint gain into REACH and SOLVE. REACH is how often a policy arrives at a state the environment confirms is a fixed number of actions from success. SOLVE is how often it finishes from an identical cloned state. Across two benchmarks and two independently released pipelines, the reacher by solver interaction is positive in all five conditions. An RL history is worth more to an RL solver than the same history is to an SFT solver. On ALFWorld, RL improves both terms, and the SFT solver never succeeds where the RL solver fails. Independent REACH and SOLVE gaps predict the aggregate interaction. Handoff asks only that one checkpoint's history can be replayed under another, so long-horizon evaluation can report arrival and completion beside endpoint success.

Chinese Translation

强化学习现在训练能够在实时环境中执行数十步操作的语言模型代理。其收益显著，并被解读为更好的决策能力。一个闭环中的代理会生成自己的输入。每个观察结果都源自其早期的动作，因此在一个回合的后期所遇到的状态部分是其自身造成的。一个监督微调（SFT）检查点和一个强化学习（RL）检查点在不同状态下进行评分，即使在相同任务上。最终成功混合了两个变化：代理到达的位置以及到达后所做的事情。将比较限制在两个策略都能到达的状态并不能将它们区分开。这种限制选择了一个结果，而在我们的数据中，它反转了效果的符号。我们引入了检查点交接，这是一种评估协议，它克隆一个已发布检查点所达到的状态，并将其交给另一个检查点，而无需重新训练。在SFT和RL中交叉达到者角色和解决者角色，将最终收益分为达到（REACH）和解决（SOLVE）。REACH是指一个策略到达环境确认的距离成功固定动作数的状态的频率。SOLVE是指从一个相同的克隆状态完成的频率。在两个基准测试和两个独立发布的管道中，达到者与解决者的交互在所有五种条件下都是正向的。对于一个RL解决者来说，RL历史的价值高于同样历史对SFT解决者的价值。在ALFWorld上，RL改善了这两个指标，而SFT解决者在RL解决者失败的地方从未成功。独立的REACH和SOLVE差距预测了整体交互。交接只要求一个检查点的历史可以在另一个检查点下重放，因此长期评估可以在最终成功的基础上报告到达和完成。

View on arXiv Download PDF AI Translation

cs.AI / 25 / 2609.19644

ScientistTwo: Pioneering the Human Knowledge Frontier with Autonomous AI

ScientistTwo：以自主人工智能开拓人类知识前沿

Nam, Jaehyun, Yoon, Jinsung, Pan, Yanzhou, Wang, Yubo, Meng, Rui, Ranganathan, Parthasarathy, Pfister, Tomas

Abstract

Scientific discovery is defined by the ability to identify the boundaries of existing knowledge and venture into unexplored territory. The ultimate vision for AI in science is problem-driven autonomous research: given a fundamental challenge by a human expert, the AI independently navigates the scientific landscape, uncovers theoretical and empirical bottlenecks, and systematically expands the frontier of knowledge. In this paper, we introduce ScientistTwo, a fully autonomous multi-agent framework designed to realize this vision. Specifically, ScientistTwo takes an initial problem as input, establishes state-of-the-art baselines, formulates novel hypotheses, and coordinates specialized agents to orchestrate an end-to-end discovery cycle without human intervention. Moreover, the framework rigorously conducts experiments using diverse datasets and metrics, refines methodologies through automated ablation studies, and validates research findings via a closed-loop simulated peer-review rebuttal engine. To evaluate ScientistTwo's capabilities against the highest standards of human scientific achievement, we benchmark it across papers accepted at top-tier conferences such as ICLR, ICML, and NeurIPS. As a result, ScientistTwo autonomously generates expert-level, publishable papers and fully verified, executable codebases. Its solutions consistently outperform human state-of-the-art models, and achieve higher average review ratings than human-authored papers under automated AI review agents. These results show that ScientistTwo is not merely an assistive tool but an autonomous scientific pioneer capable of pushing the frontiers of human discovery. Project website: https://scientist-two.github.io/

Chinese Translation

科学发现的定义在于识别现有知识的边界并探索未开发的领域。人工智能在科学中的最终愿景是以问题为驱动的自主研究：在面对人类专家提出的基本挑战时，人工智能能够独立导航科学领域，揭示理论和实证瓶颈，并系统性地扩展知识的前沿。在本文中，我们介绍了ScientistTwo，一个旨在实现这一愿景的完全自主的多智能体框架。具体而言，ScientistTwo将初始问题作为输入，建立最先进的基准，提出新颖的假设，并协调专业代理以在没有人类干预的情况下进行端到端的发现周期。此外，该框架使用多样化的数据集和指标严格进行实验，通过自动化消融研究精炼方法论，并通过闭环模拟同行评审反驳引擎验证研究结果。为了评估ScientistTwo在最高人类科学成就标准下的能力，我们对其在ICLR、ICML和NeurIPS等顶级会议上接受的论文进行了基准测试。因此，ScientistTwo能够自主生成专家级、可发表的论文和完全验证的可执行代码库。其解决方案始终优于人类最先进模型，并在自动化人工智能评审代理下获得比人类撰写论文更高的平均评审评分。这些结果表明，ScientistTwo不仅仅是一个辅助工具，而是一个能够推动人类发现前沿的自主科学先锋。项目网站：https://scientist-two.github.io/

View on arXiv Download PDF AI Translation

cs.AI / 26 / 2609.19654

Replan, Repair, or Edit? A Unified Empirical Evaluation of Travel Agents for Itinerary Revision under Resource Disruptions

重新规划、修复还是编辑？资源中断下旅行代理人行程修订的统一实证评估

Yuan, Xiaofei, Zhang, Yan, Qiao, Shaobo, He, Huangleshuai, Ni, Leyan, Ju, Mingchen, Yang, Lujia, Xu, Sijia, Tang, Yifu, Yang, Zhengyi

Abstract

Travel-planning agents generate itineraries that may become infeasible after acceptance because of flight cancellations, hotel unavailability, or attraction closures. Revising these itineraries involves full replanning, classical plan repair, and LLM-based travel-agent revision, whose differing task formulations and evaluation protocols hinder comparison. We conduct a systematic empirical study using two TREK-derived benchmark sets: 500 single-disruption cases, including feasible and infeasible instances, and 200 feasible simultaneous compound-disruption cases. We compare LLM-Z3 full replanning, IPyHOPPER hierarchical repair, and an iTIMO local-revision adapter across effectiveness, plan stability, and computational cost. LLM-Z3 with Gemini achieved the highest observed compound-disruption success. IPyHOPPER nearly matched that configuration's single-disruption overall success, while preserving substantially more of the accepted itinerary on successful repairs. Successful hierarchical and local repairs made fewer edits and retained more accepted commitments than full replanning. Computational profiles differed: IPyHOPPER used no LLM inference, the evaluated LLM-Z3 adapter used compact one-call inference, and the iTIMO adapter consumed substantially more tokens. The study provides practical guidelines for balancing feasibility recovery, commitment preservation, and computational cost within evaluated settings.

Chinese Translation

旅行规划代理人生成的行程在接受后可能因航班取消、酒店不可用或景点关闭而变得不可行。修订这些行程涉及全面重新规划、经典计划修复和基于大型语言模型（LLM）的旅行代理人修订，其不同的任务表述和评估协议阻碍了比较。我们使用两个基于TREK的基准集进行系统实证研究：500个单一中断案例，包括可行和不可行实例，以及200个可行的同时复合中断案例。我们比较了LLM-Z3全面重新规划、IPyHOPPER分层修复和iTIMO局部修订适配器在有效性、计划稳定性和计算成本方面的表现。LLM-Z3与Gemini结合实现了观察到的最高复合中断成功率。IPyHOPPER几乎匹配了该配置的单一中断整体成功率，同时在成功修复中保留了更多的接受行程。成功的分层和局部修复所做的编辑较少，并且保留了比全面重新规划更多的接受承诺。计算性能方面存在差异：IPyHOPPER未使用LLM推理，评估的LLM-Z3适配器使用紧凑的一次调用推理，而iTIMO适配器消耗了更多的token。该研究为在评估环境中平衡可行性恢复、承诺保留和计算成本提供了实用指南。

View on arXiv Download PDF AI Translation

cs.AI / 27 / 2609.19671

When2Think: Learning Difficulty-Aware Length Control for Efficient Hybrid Reasoning Models

When2Think：针对高效混合推理模型的学习难度感知长度控制

Shim, Jaejun, Kim, HyunJin, Kim, Young Jin, Bak, JinYeong

Abstract

Large Reasoning Models (LRMs) achieve strong performance on complex tasks but exhibit systematic inefficiency: they often overthink easy problems and underthink hard ones. Existing approaches based on uniform length penalties or rigid routing incur an efficiency tax, trading reduced computation on easy instances for accuracy loss on hard instances. We formulate efficient reasoning as an instance-adaptive computation allocation problem and propose When2Think, a post-training framework for hybrid reasoning that dynamically allocates computation based on problem difficulty. Our method introduces Instance-level Difficulty-Aware Control (IDAC), a reward-shaping mechanism that leverages pre-computed reference statistics (accuracy and token usage) to regulate reasoning depth. Combined with verifier-based rewards and batch-wise standardized advantages, IDAC enables stable critic-free optimization without learned reward models or online reference-model queries. When2Think encourages direct answering on easy instances while preserving extended reasoning on hard instances, thereby learning when to use System 1 (NoThink) versus System 2 (Think). Experiments on mathematical benchmarks demonstrate improved accuracy-efficiency trade-offs: on AIME24, Pass@3 increases by 10.0% while token usage is reduced by 27.9% relative to the base model, and on AIME25, When2Think achieves 40.0% Pass@3, outperforming compression and routing-only baselines.

Chinese Translation

大型推理模型（LRMs）在复杂任务上表现出色，但存在系统性低效的问题：它们往往在简单问题上过度思考，而在困难问题上思考不足。现有基于统一长度惩罚或刚性路由的方法会产生效率损失，牺牲在简单实例上的计算以换取在困难实例上的准确性损失。我们将高效推理表述为一个实例自适应计算分配问题，并提出了When2Think，一个后训练框架，用于混合推理，能够根据问题难度动态分配计算。我们的方法引入了实例级难度感知控制（IDAC），这是一种奖励塑形机制，利用预计算的参考统计数据（准确性和令牌使用）来调节推理深度。结合基于验证者的奖励和批量标准化优势，IDAC实现了稳定的无评论优化，而无需学习奖励模型或在线参考模型查询。When2Think在简单实例上鼓励直接回答，同时在困难实例上保留扩展推理，从而学习何时使用系统1（NoThink）与系统2（Think）。在数学基准测试中的实验表明，准确性与效率的权衡有所改善：在AIME24上，Pass@3提高了10.0%，而令牌使用量相较于基础模型减少了27.9%；在AIME25上，When2Think达到了40.0%的Pass@3，超越了仅基于压缩和路由的基线。

View on arXiv Download PDF AI Translation

cs.AI / 28 / 2609.19680

FINSKILLOPS: A Self-Evolving Multi-Agent System for SEC Filing QA

FINSKILLOPS：一种自我进化的多智能体系统用于SEC文件问答

Ma, Yanzhang, Tai, Zhenghan, Wu, Hanwei, Guan, Sizhe, Lei, Jianliang, He, Hailin, Jiang, Chaolong, Chi, Jijun, Kwok, Tung Sum Thomas, Xiao, Bohuai, Tian, Jingrui, Wu, Xinlu, Zhan, Xingao, Lu, Peng, Li, Muzhi, Wu, Yihong, Ma, Liheng, Lyu, Sicheng, Yan, Tianshuo, Zhu, Junhao, Xu, Yaqian, Ding, Lei, Cui, Yufei, Liu, Ziquan, Han, Boyu, Liu, Hengli, Zhou, Ling, Wang, Xinyu

Abstract

Financial QA systems are typically improved before deployment through better retrieval, prompting, or agent coordination, leaving their reliability behavior fixed thereafter. In practice, new SEC-filing questions repeatedly expose heterogeneous errors in period, entity, evidence use, and calculation. Existing self-improvement methods can turn failures into new behaviors, but offer limited control over where a correction should apply or which previously correct answers it may break. We therefore frame post-deployment improvement as controlled behavioral maintenance: recurring failures should become scoped skill patches, and each patch should earn deployment with- out introducing regressions. We instantiate this view in FINSKILLOPS, a multi-agent system for SEC filing QA. FINSKILLOPS derives reusable skills from evidence-grounded, typed failure diagnoses and governs them through targeted validation, protected-case regression checks, negative controls, and versioned replacement or retirement. Across six financial QA benchmarks, a single frozen skill registry achieves the highest verdict-weighted correctness and reference consistency among the evaluated systems. Evolved skills raise correctness from 3.70 to 4.55 on our enhanced benchmark. In a separate 12-round operational study, only six of 33 proposed skills are promoted, while the monitoring non-correct rate falls from 20.0% to 12.5%. These results establish controlled skill scope, admission, and lifecycle management as the foundation for reliable self-improvement.

Chinese Translation

金融问答系统通常在部署前通过改进检索、提示或智能体协调来提升性能，之后其可靠性行为保持固定。在实践中，新出现的SEC文件问题反复暴露出在时间、实体、证据使用和计算方面的异质性错误。现有的自我改进方法可以将失败转化为新的行为，但对修正应适用的位置或可能破坏的先前正确答案提供的控制有限。因此，我们将部署后的改进框架视为受控的行为维护：反复出现的失败应转化为有针对性的技能补丁，每个补丁应在不引入回归的情况下获得部署。我们在FINSKILLOPS中实现了这一观点，FINSKILLOPS是一个用于SEC文件问答的多智能体系统。FINSKILLOPS从基于证据的、类型化的失败诊断中提取可重用的技能，并通过有针对性的验证、受保护的案例回归检查、负控制和版本替换或退役来管理这些技能。在六个金融问答基准测试中，单一的冻结技能注册表在评估的系统中实现了最高的判决加权正确性和参考一致性。进化的技能使我们的增强基准上的正确性从3.70提高到4.55。在一项独立的12轮操作研究中，33个提议的技能中仅有六个被提升，而监控的非正确率从20.0%降至12.5%。这些结果确立了受控技能范围、准入和生命周期管理作为可靠自我改进的基础。

View on arXiv Download PDF AI Translation

cs.AI / 29 / 2609.19721

LearnActCoder: Role-Aware Error Memory for Adaptive Clinical Coding Agents

LearnActCoder：面向角色的自适应临床编码代理的错误记忆

Ghaffari, Meysam, Sen, Bhaskar, Sabetpour, Nasim, Fatehi, Nina, Agarwal, Animesh, Morato, Carlos

Abstract

Clinical coding agents repeatedly encounter the same failure modes, including unsupported codes, missed documented conditions, specificity errors, and procedure-coding convention mismatches. We introduce Learn-Then-Act, an inference-time adaptation framework that converts errors from a small labeled LEARN batch into a structured Mistake Knowledge Database (MistakeKDB). False-negative lessons are routed to a recall-oriented Coder, while false-positive lessons are routed to a precision-oriented Judge. We instantiate the framework in LearnActCoder, a Coder-Judge clinical coding pipeline with lookup-table grounding where available. On 150 matched MIMIC-III notes, structured MistakeKDB improves CPT F1 by 5.9 percentage points, while raw-example and reflection-style memories remain near the no-memory baseline; the ICD-9 improvement is not significant. On a matched MIMIC-IV cohort, memory shifts ICD-10 coding toward higher precision at a recall cost, leaving F1 statistically unchanged. Applying the same memory to 1,000 held-out MIMIC-III notes maintains a stable ICD operating point, providing scale/stability evidence. Overall, the results are consistent with structured, feedback-derived error memory being useful for adapting clinical coding behavior across cases without weight updates or changes to the underlying workflow. Absolute CPT/HCPCS performance remains low, and the system is evaluated retrospectively rather than in clinical deployment.

Chinese Translation

临床编码代理反复遇到相同的失败模式，包括不支持的编码、遗漏的记录条件、特异性错误和程序编码约定不匹配。我们提出了Learn-Then-Act，这是一种推理时适应框架，将来自小规模标记LEARN批次的错误转换为结构化的错误知识数据库（Mistake Knowledge Database，MistakeKDB）。假阴性教训被引导至以召回为导向的编码器（Coder），而假阳性教训则被引导至以精确度为导向的评判者（Judge）。我们在LearnActCoder中实例化了该框架，这是一个具有查找表基础的编码器-评判者临床编码管道。在150个匹配的MIMIC-III笔记中，结构化的MistakeKDB使CPT F1提高了5.9个百分点，而原始示例和反思式记忆的表现接近无记忆基线；ICD-9的改进不显著。在匹配的MIMIC-IV队列中，记忆将ICD-10编码向更高的精确度转移，但以召回为代价，F1统计上保持不变。将相同的记忆应用于1,000个保留的MIMIC-III笔记时，保持了稳定的ICD操作点，提供了规模/稳定性的证据。总体而言，结果与结构化的、基于反馈的错误记忆在不进行权重更新或更改基础工作流程的情况下，有助于适应临床编码行为的观点一致。绝对的CPT/HCPCS表现仍然较低，系统是以回顾性方式进行评估，而非在临床部署中。

View on arXiv Download PDF AI Translation

cs.AI / 30 / 2609.19754

AutoData: Agentic Search for Pre-training Data Selection

AutoData：面向预训练数据选择的自主搜索

Meng, Yan, Srikanth, Dhruv, Zhao, Bingchen, Jiang, Zhengyao, Wu, Yuxiang

Abstract

LLM agents have recently shown promise in automating machine learning engineering by editing model and training code under execution feedback. Data, however, remains largely outside this agentic optimisation loop. We frame pre-training data selection as heuristic engineering over per-document features, i.e., lexical statistics, categorical labels, and perplexity. We introduce AutoData, an agent that searches directly over executable selection algorithms. Unlike prior data mixture methods that optimise weights over a fixed set of domains, AutoData searches a richer program space of scoring, stratification, and stochastic selection rules, discovering feature interactions automatically by iteratively refining algorithms with validation feedback from a proxy model. Within an overnight search, AutoData discovers a selection algorithm that outperforms existing human-designed curation pipelines. Despite being searched only on this small proxy, the discovered recipe transfers to larger scales and improves the downstream metric CORE. These results suggest that data engineering can be treated as an agentic machine learning problem, extending autonomous research from model and training-code optimization to the data.

Chinese Translation

大型语言模型（LLM）代理最近在通过执行反馈编辑模型和训练代码方面显示出在自动化机器学习工程中的潜力。然而，数据在很大程度上仍然处于这种自主优化循环之外。我们将预训练数据选择框架视为对每个文档特征的启发式工程，即词汇统计、类别标签和困惑度。我们引入了AutoData，一个直接在可执行选择算法上进行搜索的代理。与之前在固定领域集上优化权重的数据混合方法不同，AutoData在评分、分层和随机选择规则的更丰富程序空间中进行搜索，通过从代理模型获得的验证反馈迭代地优化算法，自动发现特征交互。在一夜的搜索中，AutoData发现了一种选择算法，其性能超过了现有的人为设计的策展流程。尽管仅在这个小型代理上进行搜索，但发现的方案能够迁移到更大规模并改善下游指标CORE。这些结果表明，数据工程可以被视为一个自主机器学习问题，将自主研究的范围从模型和训练代码优化扩展到数据。

View on arXiv Download PDF AI Translation

cs.AI / 31 / 2609.19759

Rethinking Multi-Agent Collaboration: When More Is Less

重新思考多智能体协作：当更多变为更少

Yuan, Yishuo, Wu, Yibo, Zhang, Yihan, Sun, Minyuan, Li, Shenliang, Ma, Xinkai, Li, Yifan, Liu, Jiaheng

Abstract

The rapid advancement of large language models and single-agent harnesses has reshaped the landscape of autonomous systems, raising a critical question of when multi-agent collaboration offers genuine value. As individual agent capabilities continue to scale, multi-agent collaboration faces diminishing returns while incurring growing context overhead. Through systematic analysis, we delineate the capability boundaries of multi-agent collaboration relative to single-agent alternatives, showing that it confers systematic benefits specifically in long-horizon tasks with sparse dependencies, while single-agent harnesses remain superior in tightly coupled, sequential workflows. Building on these insights, we propose SAIGE, a lightweight multi-agent collaboration mechanism based on Semantic-Aware Incremental Graph Evolution. SAIGE models collaboration as a dynamically evolving graph, where nodes are agent instances spawned on demand and edges encode semantic dependencies established through content-based information retrieval. Experiments on long-horizon, complex task benchmarks show that SAIGE achieves a favorable trade-off between context efficiency and task performance, and that scaling the agent pool or deepening the recursion level does not consistently improve outcomes. Our findings suggest that multi-agent superiority is bounded by task structure rather than universal, and that more agents do not necessarily make a system more intelligent.

Chinese Translation

大型语言模型和单智能体系统的快速发展重塑了自主系统的格局，提出了一个关键问题：何时多智能体协作能够提供真正的价值。随着单个智能体能力的不断提升，多智能体协作面临收益递减的挑战，同时伴随越来越大的上下文开销。通过系统分析，我们界定了多智能体协作相对于单智能体替代方案的能力边界，表明在具有稀疏依赖关系的长时间任务中，它确实带来了系统性的好处，而在紧密耦合的顺序工作流中，单智能体系统仍然占据优势。基于这些见解，我们提出了SAIGE，一种基于语义感知增量图演化的轻量级多智能体协作机制。SAIGE将协作建模为一个动态演变的图，其中节点是按需生成的智能体实例，边则编码通过基于内容的信息检索建立的语义依赖关系。在长时间、复杂任务基准上的实验表明，SAIGE在上下文效率和任务性能之间实现了良好的权衡，并且扩展智能体池或加深递归层级并不一定能持续改善结果。我们的研究结果表明，多智能体的优势受限于任务结构，而非普遍适用，更多的智能体并不一定使系统更智能。

View on arXiv Download PDF AI Translation

cs.AI / 32 / 2609.19770

TorchCraft: Unified binder design by inverting an all-atom structure predictor

TorchCraft：通过反转全原子结构预测器实现统一的结合剂设计

TorchCraft Team, Liu, Yu, Shen, Zhouhanyu, Li, Zhengyi, Huang, Xikun, Liu, Jiaqi, Gao, Shuxian, Yu, Qilin, Qin, Xiayan, Zhang, Yucheng, Chen, Mingchen

Abstract

All-atom structure predictors model diverse molecular interactions, but using their learned structural priors for binder design remains challenging. Here we present TorchCraft, a unified binder-design framework that optimizes sequence logits through a frozen all-atom predictor. Implemented in TorchFold, TorchCraft combines confidence, contact, geometric, and sequence-prior objectives within a shared optimization procedure for minibinders, framework-conditioned VHHs, cyclic peptides, and ligand-binding proteins. Using pretrained AlphaFold 3 weights, TorchCraft generated representative minibinders and VHHs with experimentally measured binding across four targets in each format, without post hoc sequence redesign. Computational benchmarks further demonstrated the framework's applicability to cyclic peptides and ligand-conditioned pocket design. TorchCraft extends predictor inversion to multiple binder formats and molecular contexts, providing a common framework for reusing all-atom structural priors in design.

Chinese Translation

全原子结构预测器能够建模多样的分子相互作用，但利用其学习到的结构先验进行结合剂设计仍然具有挑战性。在此，我们提出了TorchCraft，一个统一的结合剂设计框架，通过一个冻结的全原子预测器优化序列对数。TorchCraft在TorchFold中实现，结合了置信度、接触、几何和序列先验目标，在一个共享的优化过程中应用于小结合剂、框架条件下的VHH、环肽和配体结合蛋白。使用预训练的AlphaFold 3权重，TorchCraft生成了具有实验测量结合能力的代表性的小结合剂和VHH，涵盖每种格式的四个靶标，且无需事后序列重新设计。计算基准进一步证明了该框架在环肽和配体条件口袋设计中的适用性。TorchCraft将预测器反转扩展到多种结合剂格式和分子背景，为在设计中重用全原子结构先验提供了一个通用框架。

View on arXiv Download PDF AI Translation

cs.AI / 33 / 2609.19775

Integrating knowledge from case reports: a medical ontology based multimodal information system with structured summary

整合病例报告中的知识：基于医学本体的多模态信息系统与结构化摘要

Guo, Shuyu, Huang, Lan, Liu, Yichen, Ma, Hanbin, Bai, Tian

Abstract

Published medical case reports serve as a crucial medical information carrier, documenting discoveries in rare diseases, diagnostic methods, and innovative treatments. Despite the wealth of clinical knowledge in millions of case reports in the public medicine literature database (PubMed), accessing relevant information efficiently is hindered by the limitations of traditional keyword-based retrieval tools on unstructured and diverse case reports. To address the above issues, we introduce a comprehensive multimodal information system for case reports integrating structured clinical summaries of patients including medical images and biomedical named entities from 52949 open-access case reports published from 2000 to 2021. The multimodal essential information is organized in a well-structured medical ontology. Also, a powerful interface for searching and browsing case reports is designed to assist junior clinicians in retrieving cases effectively and improving the identification and diagnosis of rare diseases.

Chinese Translation

已发表的医学病例报告作为重要的医学信息载体，记录了罕见疾病的发现、诊断方法和创新治疗。尽管公共医学文献数据库（PubMed）中有数百万份病例报告蕴含着丰富的临床知识，但传统基于关键词的检索工具在处理非结构化和多样化的病例报告时存在局限性，导致相关信息的高效获取受到阻碍。为了解决上述问题，我们提出了一种全面的多模态信息系统，整合了2000年至2021年间发布的52949份开放获取病例报告中的患者结构化临床摘要，包括医学图像和生物医学命名实体。多模态的核心信息被组织在一个结构良好的医学本体中。此外，我们设计了一个强大的界面，用于搜索和浏览病例报告，以帮助初级临床医生有效检索病例，并改善罕见疾病的识别和诊断。

View on arXiv Download PDF AI Translation

cs.AI / 34 / 2609.19789

Contagion on the Trading Floor: How Adversarial Signals Spread in Multi-Agent Trading Systems

交易场上的传染：对抗信号如何在多智能体交易系统中传播

Sua, Qi Rong, Dong, Junhao, Thai, Nguyen Duc, Wen, Yuqing, Tan, Cheston, Ong, Yew-Soon

Abstract

Multi-agent trading systems built on large language models (LLMs) are beginning to appear in quantitative finance, yet their robustness to adversarial inputs is largely unknown. We study the vulnerability of LLM trading stacks to black-box, input-only attacks that enter solely via admissible social-media feeds. We introduce the Generic Multi-Agent Trading System (GMATS), a framework that captures modern multiagent trading architectures and instantiate a class of black-box poisoning attackers that treat an LLM as a post generator and inject budget-constrained, plausibly benign social-media content into the analyst's evidence stream. We define contagion metrics that trace how adversarial content propagates through the stack, including belief-shift scores at analyst and coordinator layers and attack-clean deltas on standard backtest metrics. Experiments on a safe offline benchmark with historical market and social data show that even simple input-only attackers can materially degrade risk-return profiles, sharply reducing Sharpe ratios. At the same time, we find that suitably designed multi-agent topologies and coordinator prompts can dampen adversarial shocks and improve average robustness under identical poisoning budgets.

Chinese Translation

基于大型语言模型（LLMs）的多智能体交易系统在量化金融领域开始出现，但它们对对抗性输入的鲁棒性仍然不为人知。我们研究了LLM交易堆栈对仅通过可接受的社交媒体信息输入的黑箱攻击的脆弱性。我们引入了通用多智能体交易系统（GMATS），这是一个捕捉现代多智能体交易架构的框架，并实例化了一类黑箱投毒攻击者，这些攻击者将LLM视为后生成器，并将预算受限、看似良性的社交媒体内容注入分析师的证据流中。我们定义了传染度指标，以追踪对抗性内容如何在堆栈中传播，包括分析师和协调者层的信念转变分数以及标准回测指标上的攻击清理增量。在一个安全的离线基准测试中，结合历史市场和社交数据的实验表明，即使是简单的仅输入攻击者也能显著降低风险回报特征，锐减夏普比率。同时，我们发现适当设计的多智能体拓扑和协调者提示可以减弱对抗性冲击，并在相同的投毒预算下提高平均鲁棒性。

View on arXiv Download PDF AI Translation

cs.AI / 35 / 2609.19820

Steering Equilibrium Selection in Regularized Self-Play via the Reference Policy

通过参考策略引导正则化自我博弈中的均衡选择

Leal, Luis

Abstract

Regularized self-play -- the family behind DeepNash's Stratego play -- drives a two-player zero-sum policy to a Nash equilibrium by best-responding to a slowly moving, entropy-regularized reference policy $\rho$. When the game has a polytope of value-equivalent equilibria, the regularizer silently breaks the tie: with a uniform reference it selects the maximum-entropy member, the I-projection of $\rho$ onto the Nash set. Can the reference be used to choose the equilibrium on purpose? On five exactly solvable games plus a 2-D polytope, with exact best responses and equivalence tests over independent seeds, anchoring the reference at a target member and refining steers self-play to that member with mean coordinate error 0.007 at median exploitability $5\times10^{-5}$, TOST-equivalent to the request within $\pm0.05$; the anchoring persists through refinement and follows the reference, not the initialization. Selection follows the reach-weighted I-projection (slope 0.969 [0.950, 0.987]). We report with equal emphasis where the story breaks: fixed off-manifold references cost 0.08-0.25 exploitability; stiff or flat families require a smaller mirror step, set by a pre-registered rule; boundary targets undershoot; curvature predicts where boundary saturation bites (rank correlation 0.90, p=0.037) while interior precision is curvature-independent. Table and MLP steering maps are equivalent within $\pm0.03$ at every target (30 seeds); matched control arms show attention's robust signature is excess seed variance, any systematic shift bounded at 0.018 and not significant. Against a best response the selection-robustness trade-off is degenerate: steering matters only against fixed, non-equilibrium opponents. The recipe -- anchor the reference at the desired member and refine -- reinterprets the KL anchor of RLHF-style RL as a selection knob, not only a stability leash.

Chinese Translation

正则化自我博弈——DeepNash在Stratego游戏中的基础——通过对一个缓慢变化的、熵正则化的参考策略 $ ho$ 进行最佳响应，将一个二人零和策略驱动至纳什均衡。当游戏具有一个价值等价均衡的多面体时，正则化器默默地打破了平局：使用均匀参考时，它选择最大熵成员，即 $ ho$ 在纳什集合上的 I-投影。参考策略能否被用来有意选择均衡？在五个完全可解的游戏和一个二维多面体中，通过独立种子的精确最佳响应和等价性测试，将参考锚定在目标成员，并进行精细调整，使自我博弈趋向该成员，平均坐标误差为 0.007，处于中位数可利用性 $5 imes10^{-5}$，与请求的 TOST 等价，误差范围为 $ ext{±}0.05$；锚定在精细调整中持续存在，并跟随参考，而非初始化。选择遵循基于可达性的加权 I-投影（斜率 0.969 [0.950, 0.987]）。我们同样强调故事中断的地方：固定的离流形参考导致 0.08-0.25 的可利用性成本；刚性或平坦的族需要更小的镜像步长，由预注册规则设定；边界目标不足；曲率预测边界饱和的影响位置（秩相关 0.90, p=0.037），而内部精度与曲率无关。表格和 MLP 引导图在每个目标处的误差范围为 $ ext{±}0.03$（30 个种子）；匹配的控制组显示注意力的稳健特征是过量种子方差，任何系统性偏移限制在 0.018 且不显著。针对最佳响应，选择稳健性权衡是退化的：引导仅在固定的非均衡对手面前才重要。这个方法——将参考锚定在所需成员并进行精细调整——重新诠释了 RLHF 风格强化学习中的 KL 锚定为选择旋钮，而不仅仅是稳定性约束。

View on arXiv Download PDF AI Translation

cs.AI / 36 / 2609.19830

Dual-Axis Policy Optimization for LLM Agents: Bayesian Feedback Attribution and Trajectory Mass Normalization

双轴策略优化用于大型语言模型代理：贝叶斯反馈归因与轨迹质量归一化

Zhuang, Yingxuan, Yu, Binhe, Yang, Jingxiao, Sun, Ruopei, Li, Ziting, Tan, Cheng, Zhang, Xuhong, Yin, Jianwei, Chen, Jintao

Abstract

Reinforcement learning for LLM agents involves two distinct optimization di- mensions: how environment feedback is exploited within a trajectory, and how complete trajectories are aggregated across a batch. We formulate these dimen- sions as Intra-Trajectory Feedback Attribution and Inter-Trajectory Objec- tive Aggregation, and introduce BATON (Bayesian Attribution and Trajectory Objective Normalization), a dual-axis policy optimization framework. BATON instantiates the first axis with Bayesian Feedback Attribution, which constructs a feedback-conditioned posterior over sampled actions, and the second with Trajec- tory Mass Normalization (TMN), which assigns equal optimization mass to com- plete trajectories. Experiments with GRPO and GiGPO on ALFWorld, WebShop, and SearchQA show that both axes provide independent gains and that their combi- nation consistently achieves the strongest overall performance across model scales.

Chinese Translation

针对大型语言模型（LLM）代理的强化学习涉及两个不同的优化维度：如何在轨迹内利用环境反馈，以及如何在一个批次中聚合完整轨迹。我们将这两个维度表述为轨迹内反馈归因（Intra-Trajectory Feedback Attribution）和轨迹间目标聚合（Inter-Trajectory Objective Aggregation），并引入BATON（贝叶斯归因与轨迹目标归一化），一个双轴策略优化框架。BATON通过贝叶斯反馈归因实现第一个轴，该方法在采样动作上构建一个基于反馈的后验分布；第二个轴则通过轨迹质量归一化（Trajectory Mass Normalization, TMN）实现，该方法为完整轨迹分配相等的优化质量。在ALFWorld、WebShop和SearchQA上的GRPO和GiGPO实验表明，这两个轴都提供独立的增益，并且它们的结合在不同模型规模中始终实现了最强的整体性能。

View on arXiv Download PDF AI Translation

cs.AI / 37 / 2609.19832

MetaRTL: Meta-path Attention Enhanced Relational Table Learning

MetaRTL：增强元路径注意力的关系表学习

Zhong, Ken, Li, Weichen, Wang, Zheng

Abstract

Relational table learning has gained increasing attention with the widespread use of relational databases. Existing methods typically rely on deep GNN or HGNN stacks, leading to high computational costs and limited performance on large real-world databases. We propose MetaRTL, a two-stage framework for scalable and expressive relational table learning. In the first stage, MetaRTL obtains initial table embeddings via lightweight pre-training. In the second stage, it performs non-parametric message passing to derive meta-path features, which are then aggregated by an attention module, MetaAttn. By shifting computation from deep message passing to efficient meta-path aggregation, MetaRTL captures rich relational semantics while maintaining high efficiency. Experiments on 10 real-world datasets across 24 tasks demonstrate the effectiveness of the proposed method.

Chinese Translation

随着关系数据库的广泛应用，关系表学习受到了越来越多的关注。现有方法通常依赖于深度图神经网络（GNN）或异构图神经网络（HGNN）堆栈，导致高计算成本和在大型真实数据库上的性能有限。我们提出了MetaRTL，一个用于可扩展和表现力强的关系表学习的两阶段框架。在第一阶段，MetaRTL通过轻量级的预训练获得初始表嵌入。在第二阶段，它执行非参数消息传递以推导元路径特征，这些特征随后通过一个注意力模块MetaAttn进行聚合。通过将计算从深度消息传递转移到高效的元路径聚合，MetaRTL在保持高效率的同时捕捉丰富的关系语义。在24个任务的10个真实世界数据集上的实验表明了所提方法的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 38 / 2609.19843

A Dual-Process Perspective on Nudge Susceptibility in LLM-Based GUI Agents

基于双重过程视角的LLM驱动GUI代理的推送敏感性研究

Halimeh, Haya, Kaltenpoth, Sascha, Bösch, Kevin, Müller, Oliver

Abstract

LLM-based GUI agents increasingly act on behalf of users in digital environments that were designed with human users in mind. These graphical user interfaces were designed to support, but also deliberately steer, the behaviour and decisions of users. While behavioural biases in the textual outputs of LLMs are well-documented, far less is known about how such influence operates when models act as agents that perceive interfaces and execute decisions---and, in particular, whether the reasoning capabilities increasingly built into these agents make them more robust to it. Drawing on Dual-Process Theory, we empirically investigate whether LLM-based GUI agents are susceptible to automatic (Type 1) and reflective (Type 2) digital nudges, and how their reasoning configuration moderates this susceptibility. In a randomized online shopping experiment with 3,600 agents and a total of 21,600 simulations across six frontier models from three providers, we found that agents were vulnerable to both nudge types. Crucially, the reasoning configuration moderated these effects in opposing directions, reducing susceptibility to automatic default nudges while heightening it to reflective social influence nudges. Extensive reasoning therefore did not make agents more robust but redirected the route through which choice architecture takes effect. Exploratory analysis further showed this redirection to be systematically structured by model scale. Beyond establishing nudge susceptibility as a behavioural property of agentic AI, the study positions interface design as a governance concern for organizations that delegate decisions to autonomous agents.

Chinese Translation

基于LLM的GUI代理在以人为本的数字环境中越来越多地代表用户行动。这些图形用户界面旨在支持用户的行为和决策，同时也故意引导用户的行为和决策。尽管LLM文本输出中的行为偏差已被广泛记录，但关于当模型作为感知界面并执行决策的代理时，这种影响如何运作的了解却相对较少，尤其是这些代理日益增强的推理能力是否使其对这种影响更具抵抗力。基于双重过程理论，我们实证研究了基于LLM的GUI代理是否对自动（类型1）和反思（类型2）数字推送敏感，以及它们的推理配置如何调节这种敏感性。在一项包含3600个代理和21600次模拟的随机在线购物实验中，我们使用了来自三家提供商的六个前沿模型，发现代理对这两种推送类型均存在脆弱性。关键是，推理配置在相反的方向上调节了这些影响，降低了对自动默认推送的敏感性，同时提高了对反思性社会影响推送的敏感性。因此，广泛的推理并没有使代理更具抵抗力，而是重新引导了选择架构生效的途径。进一步的探索性分析显示，这种重新引导在模型规模上具有系统性结构。除了确立推送敏感性作为代理人工智能的行为特性外，本研究还将界面设计定位为将决策委托给自主代理的组织的治理关注点。

View on arXiv Download PDF AI Translation

cs.AI / 39 / 2609.19848

Constraint-Safe Graph-Context Scoring for Stable Point-Feature Labels Under Text-Width and Accessibility-Inspired Profiles

基于约束安全的图上下文评分方法：在文本宽度和无障碍启发式配置下实现稳定的点特征标签

Ahmad, Taimoor

Abstract

Point-feature label placement on interactive maps must reconcile geometric validity, display yield, local placement utility, and stability across camera motion. Accessibility and multilingual requirements further change label dimensions, yet algorithmic evaluations often collapse these concerns into overlap counts. We present LABELSENSE-Pilot, a reproducible prototype that generates eight compass candidates per feature, scores candidates with a multilayer perceptron over graph-context summaries, adds a previous-placement bonus, and selects a layout through mixed-integer optimization. The executed scorer is deliberately not described as a graph transformer. Every returned layout is checked for viewport containment, per-feature uniqueness, and pairwise clearance. Experiments use 2,500 airport coordinates and names spanning 155 countries, with country-grouped splits and generated density, camera, text-suffix, preference, and enlarged-font stressors. Across five seeds, LABELSENSE-Pilot displayed 85.62 percent of labels with 2.09 percent flicker and zero collisions. Versus a handcrafted-utility integer program, LABELSENSE-Pilot sacrificed 1.43 percentage points of display while reducing flicker by 12.04 points. Enlarged-box-aware layouts produced zero proxy violations, whereas standard geometry reevaluated at 1.5x violated 52.57 percent of selected placements. These results establish an auditable engineering trade-off, not human accessibility, multilingual usability, or preference. Official recent baselines and participant evidence remain required before submission.

Chinese Translation

在交互式地图上，点特征标签的放置必须协调几何有效性、显示效果、局部放置效用和在相机运动中的稳定性。无障碍和多语言需求进一步改变了标签的维度，但算法评估通常将这些问题简化为重叠计数。我们提出了LABELSENSE-Pilot，这是一个可重复的原型，针对每个特征生成八个方向候选，通过图上下文摘要使用多层感知器对候选进行评分，增加先前放置的奖励，并通过混合整数优化选择布局。执行的评分器故意不被描述为图转换器。每个返回的布局都经过检查，以确保视口包含、特征唯一性和成对间距。在实验中使用了2,500个机场坐标和名称，涵盖155个国家，并进行了按国家分组的拆分以及生成的密度、相机、文本后缀、偏好和放大字体压力测试。在五个随机种子下，LABELSENSE-Pilot 显示了85.62%的标签，闪烁率为2.09%，且没有碰撞。与手工制作的效用整数程序相比，LABELSENSE-Pilot在显示效果上牺牲了1.43个百分点，同时将闪烁降低了12.04个百分点。放大框感知布局没有产生任何代理违规，而标准几何在1.5倍的重新评估中违反了52.57%的选定放置。这些结果建立了一个可审计的工程权衡，而不是人类无障碍、多语言可用性或偏好。在提交之前，仍需官方的最新基准和参与者证据。

View on arXiv Download PDF AI Translation

cs.AI / 40 / 2609.19866

Reproducibility is not construct validity: LLM measurement of institutionally situated communication

可重复性并不等同于构念效度：LLM对机构情境沟通的测量

Batzdorfer, Veronika, Santagiustina, Carlo Romano Marcello Alessandro

Abstract

High annotation reproducibility does not necessarily imply that an LLM-inferred measure captures the construct it is intended to measure. We test this distinction using a dataset from the European Commission's AI Act consultation, linking structured survey responses to free-text consultation submissions from the same stakeholders. LLM annotations of consultation submissions are highly reproducible (intraclass correlations > 0.99), yet show limited convergence with survey-reported measures of the nominal construct they were intended to approximate. Divergence between survey-and LLM-inferred text-based measures varies systematically across stakeholder groups: business associations express greater concern about AI risks in text-based consultations than in survey responses ({\=g} = +1.0), whereas public authorities and several nonbusiness groups show smaller or negative divergences. Divergences between scores suggest positive spatial autocorrelation across European countries (Moran's I = 0.347, p = 0.036), indicating that stakeholders from neighboring countries tend toward more similar text-based stances towards AI safety concerns. Despite divergence, survey-reported concerns remain strongly associated with support for explainability across all divergence levels. These results demonstrate that LLM annotation reproducibility can coexist with poor construct correspondence and motivate validation procedures that distinguish reproducibility, construct validity, and communication context variation when LLMs are used as measurement instruments.

Chinese Translation

高注释可重复性并不一定意味着LLM推断的测量能够捕捉其所意图测量的构念。我们使用来自欧洲委员会人工智能法案咨询的数据集来测试这一区别，将结构化调查响应与来自同一利益相关者的自由文本咨询提交相链接。LLM对咨询提交的注释具有高度的可重复性（组内相关系数 > 0.99），但与调查报告的名义构念测量之间的收敛性有限。调查与LLM推断的基于文本的测量之间的差异在利益相关者群体中系统性变化：商业协会在基于文本的咨询中对人工智能风险的关注程度高于调查响应（{ ext{g}} = +1.0），而公共机构和一些非商业团体则表现出较小或负的差异。分数之间的差异表明，欧洲国家之间存在正空间自相关（Moran's I = 0.347, p = 0.036），这表明来自邻国的利益相关者在对人工智能安全问题的文本立场上趋向于更为相似。尽管存在差异，调查报告的关注点仍与对可解释性的支持在所有差异水平上保持强相关。这些结果表明，LLM注释的可重复性可以与较差的构念对应性共存，并促使在将LLM作为测量工具时，区分可重复性、构念效度和沟通背景变化的验证程序。

View on arXiv Download PDF AI Translation

cs.AI / 41 / 2609.19871

Physical knowledge on historical data matters more than enforcing physical constraints on the forecast

历史数据中的物理知识比在预测中强制物理约束更为重要

Lehembre, Etienne, Audigane, Pascal, Nguyen, Vincent, Vrain, Christel, Dao, Thi-Bich-Hanh

Abstract

Time series forecasting has seen signicant advancements with the emergence of new deep learning models. However, forecasting time series in applications involving physical processes remains a major challenge. Despite the apparition of Physics Informed Neural Networks (PINN), recent models do not estimate unobservable intermediate physical variables, which are important for domain experts to understand the target behavior. To this end, we propose a Physics Informed Recurrent Neural Network (PIRNN) which predicts, along the target, unobservable variables on both historic data and forecast target. This approach enhances the model robustness and results interpretation using domain knowledge. Our method is easily adaptable to any physical model using several equations, each having its own set of unobservable variables, to describe it-self. As a case study, we incorporate physical equations used for groundwater levels predictions by the physical model called Gardenia. This model uses transfers equations between reservoirs, optimized with data assimilation, to simulate the evolution of groundwater levels. Evaluation includes several well known neural network models and the Gardenia model compared on twelve real world datasets. In addition, we study the impact of each component through an ablation study. Our model outperforms other models on ve out of the twelve datasets and our ablation study underlines the importance of having a physical background in our time series forecasting task. Finally, the coherence of the physical variables predicted by our neural network is assessed by a domain expert.

Chinese Translation

时间序列预测随着新深度学习模型的出现取得了显著进展。然而，在涉及物理过程的应用中进行时间序列预测仍然是一个主要挑战。尽管出现了物理信息神经网络（Physics Informed Neural Networks, PINN），但最近的模型并未估计不可观测的中间物理变量，而这些变量对于领域专家理解目标行为至关重要。为此，我们提出了一种物理信息递归神经网络（Physics Informed Recurrent Neural Network, PIRNN），该网络在预测目标的同时，还预测历史数据和预测目标上的不可观测变量。这种方法利用领域知识增强了模型的鲁棒性和结果解释能力。我们的方法易于适应任何物理模型，使用多个方程，每个方程都有其自身的一组不可观测变量来描述模型。作为案例研究，我们结合了用于地下水位预测的物理方程，该方程由名为Gardenia的物理模型使用。该模型使用水库之间的传递方程，并通过数据同化进行优化，以模拟地下水位的演变。评估包括与十二个真实世界数据集进行比较的几种知名神经网络模型和Gardenia模型。此外，我们通过消融研究研究了每个组件的影响。我们的模型在十二个数据集中的五个上表现优于其他模型，而我们的消融研究强调了在时间序列预测任务中拥有物理背景的重要性。最后，领域专家对我们神经网络预测的物理变量的一致性进行了评估。

View on arXiv Download PDF AI Translation

cs.AI / 42 / 2609.19897

TRACE: Accountable Agentic Retrieval for Source Discovery in Digital Archives

TRACE：数字档案源发现的可追溯代理检索

Bian, Donghan, Puren, Marie, Cafiero, Florian

Abstract

Historical archives pose a difficult retrieval problem for retrievalaugmented generation systems: documents are OCR-degraded, heterogeneous across genres and sources, and require strong source traceability for scholarly and institutional use. We introduce TRACE, a training-free agentic retrieval framework designed for accountable source discovery over historical corpora. The system was developed in the context of DECIDON, an interdisciplinary project on the circulation of political discourse between parliamentary debates and the press during the French Third Republic, involving digitised historical collections and institutional use cases. The prototype is currently deployed internally within the project and accessible to 24 researchers across six partner institutions. We evaluate TRACE on HistoriQA-ThirdRepublic, a benchmark of 1,752 French historical questions over parliamentary debates and newspapers from 1887, with documents derived from Biblioth{\`e}que nationale de France digitised collections. TRACE achieves R@10 = 0.856 and MRR = 0.653, outperforming sparse, dense, graph-based, and agentic RAG baselines, with the largest gains on multi-hop and cross-corpus questions. At approximately $0.02 per question under the default hosted inference configuration, TRACE also remains economically feasible for heritage institutions, laboratories or companies that cannot rely on costly local GPU infrastructure. These results suggest that, for large digital libraries and archives, retrieval accountability and corpus-aware agent design can provide a practical alternative to heavier training-based or graph-construction approaches.

Chinese Translation

历史档案为增强检索生成系统带来了困难的检索问题：文档经过光学字符识别（OCR）处理后质量下降，且在体裁和来源上存在异质性，同时学术和机构使用要求强大的源可追溯性。我们提出了TRACE，一个无需训练的代理检索框架，旨在对历史语料库进行可追溯的源发现。该系统是在DECIDON项目的背景下开发的，这是一个跨学科项目，研究法国第三共和国期间议会辩论与媒体之间的政治话语流通，涉及数字化历史收藏和机构使用案例。该原型目前在项目内部部署，并向六个合作机构的24名研究人员开放。我们在HistoriQA-ThirdRepublic上评估TRACE，这是一个包含1,752个关于1887年议会辩论和报纸的法语历史问题的基准数据集，文档来源于法国国家图书馆的数字化收藏。TRACE在R@10上达到了0.856，MRR为0.653，超越了稀疏、密集、基于图的和代理RAG基线，尤其在多跳和跨语料库问题上取得了最大的提升。在默认托管推理配置下，TRACE每个问题的成本约为0.02美元，仍然对无法依赖昂贵本地GPU基础设施的遗产机构、实验室或公司具有经济可行性。这些结果表明，对于大型数字图书馆和档案馆，检索可追溯性和语料库感知的代理设计可以为更重的基于训练或图构建的方法提供一种实用的替代方案。

View on arXiv Download PDF AI Translation

cs.AI / 43 / 2609.19928

From "Who Is This User?" to "What Does This Purchase Mean?": A Deployed Pipeline for Semantic User Profiling at Bank Scale

从“这个用户是谁？”到“这笔购买意味着什么？”：一种在银行规模下部署的语义用户画像管道

Mitsuhashi, Ryota, Morimura, Tetsuro, Ito, Hirotake

Abstract

Per-user LLM inference on transaction histories binds the inference budget linearly to user count, which becomes prohibitive at applied scale. We re-cast attribute inference from per-user to per-transaction-pattern. The pipeline runs in three phases: Resolve abstracts item names with optional web grounding, Profile infers attributes for each frequent pattern, and Tag clusters free-text attributes into a queryable database. In Profile, a single LLM call per pattern emits predefined categorical labels, free-text attributes, and per-attribute prevalence estimates. Because inference runs over patterns rather than users, the budget grows with the pattern count rather than the user count. On the public Open e-commerce corpus, the database is statistically indistinguishable from an LLM that reads each user's raw history directly in AUC across the evaluated attributes, and the prevalence estimates carry discriminative signal between positive and negative users. The pipeline is deployed at a major Japanese bank profiling on the order of tens of millions of users, with close to a three-order-of-magnitude reduction in LLM inference targets versus a per-user pipeline. The code is publicly available on https://github.com/CyberAgentAILab/profiling-agent-open-ecommerce.

Chinese Translation

针对每个用户的LLM推理在交易历史上将推理预算线性绑定到用户数量，这在应用规模上变得不可承受。我们将属性推理从每个用户转变为每个交易模式。该管道分为三个阶段：解析抽象项名称并可选地进行网络基础，画像推断每个频繁模式的属性，以及标记将自由文本属性聚类到可查询的数据库中。在画像阶段，每个模式的单次LLM调用会输出预定义的类别标签、自由文本属性和每个属性的流行度估计。由于推理是基于模式而非用户进行的，因此预算随着模式数量的增加而增长，而不是用户数量。在公共的开放电子商务语料库上，该数据库在评估的属性中与直接读取每个用户原始历史的LLM在AUC上统计上无显著差异，并且流行度估计在正负用户之间传递了区分信号。该管道已在一家主要的日本银行部署，画像用户数量达到数千万，与每用户管道相比，LLM推理目标减少了近三个数量级。代码可在 https://github.com/CyberAgentAILab/profiling-agent-open-ecommerce 上公开获取。

View on arXiv Download PDF AI Translation

cs.AI / 44 / 2609.19934

Beyond Depth Truncation: Controlled Evaluation of Depth Utilization in Recursive Language Models

超越深度截断：递归语言模型中深度利用的受控评估

Van Dau, Ha, Khuat, Thanh Tung, Dung, Nguyen Thanh

Abstract

Depth-recurrent language models iteratively apply a small layer stack, decoupling per-token compute from distinct parameter count. To determine whether such a model genuinely utilizes its depth, both recurrence and layer-pruning literatures rely on a shared evaluation: truncating depth at inference time, plotting quality against retained depth fraction, and reading off the slope. While cheap and training-free, this metric suffers from an unexamined flaw: it extracts a single scalar from an intervention that alters multiple model properties simultaneously. Depth truncation concurrently reduces the number of block applications, decreases the volume of distinct computation performed, and pushes the readout head onto an out-of-distribution residual stream. The observed slope conflates all three factors, yet is conventionally interpreted as reflecting solely the second. We propose the Depth Control Protocol (DCP), a diagnostic suite that disentangles these three quantities. DCP comprises three positive controls that isolate each factor while varying the others, a negative control applying the identical interventions to dense transformers to ensure the effect is not an artifact of the measurement protocol, and a controlled training intervention to verify causality. The linchpin control, running the full budget of block applications while executing only a single distinct iteration, is strictly realizable only in depth-wise weight-sharing architectures, since in a dense network repeating a layer yields an entirely different model rather than the same model in an alternative configuration.

Chinese Translation

深度递归语言模型通过迭代应用小层堆栈，将每个标记的计算与不同的参数数量解耦。为了确定这样的模型是否真正利用了其深度，递归和层修剪文献依赖于一个共同的评估方法：在推理时截断深度，绘制质量与保留深度比例之间的关系，并读取斜率。虽然这种方法便宜且不需要训练，但它存在一个未被检视的缺陷：它从一个干预中提取一个单一的标量，而该干预同时改变多个模型属性。深度截断同时减少了块应用的数量，降低了执行的独立计算量，并将读出头推向一个分布外的残差流。观察到的斜率混淆了这三种因素，但通常被解释为仅反映第二个因素。我们提出了深度控制协议（Depth Control Protocol, DCP），这是一个诊断工具，能够解开这三种量。DCP包括三个正控制，分别在变化其他因素的同时隔离每个因素，一个负控制将相同的干预应用于密集型变换器，以确保效果不是测量协议的伪影，以及一个受控训练干预以验证因果关系。关键控制是在执行单个独特迭代的同时运行全部块应用预算，这在深度权重共享架构中是严格可实现的，因为在密集网络中重复一个层会产生一个完全不同的模型，而不是在另一种配置下的相同模型。

View on arXiv Download PDF AI Translation

cs.AI / 45 / 2609.19944

MaSCoD: A Multi-Agent Framework for Structural-Context-Guided Candidate Causal Graph Generation

MaSCoD：一种基于结构上下文引导的多智能体候选因果图生成框架

Nakada, Yudai, Nishiura, Yuichiro, Splichal, Jin Michael

Abstract

Large language models (LLMs) have been applied to causal discovery, but candidate-graph generation rarely treats premature omission of potentially relevant causal relations as an explicit design objective. We propose MaSCoD, a multi-agent framework that organizes candidate third variables and local structural patterns before direct-edge judgment. We evaluate MaSCoD on Auto-MPG, DWD, and Sachs using GPT-5.4 as the primary backbone and GPT-4o for replication. MaSCoD exhibits a dataset- and backbone-dependent retention-selectivity profile rather than uniform superiority. Across all six dataset-backbone settings, Full, which supplies structural hypotheses before direct-edge judgment, achieved higher mean Recall and F1 than No Phase 1, which instead constructs them within the judgment procedure, while also increasing false-positive rates. Additional reference-edge retention over all evaluated baselines was observed on DWD with GPT-5.4 and on Sachs with GPT-4o, rather than uniformly across settings. Partial ablations showed that supplying both information components did not always outperform supplying only one. For GPT-5.4, stage-wise analysis showed that the Full-No Phase 1 retention gap was already present after direct-edge judgment, while reconciliation introduced additional reference-edge loss for Full on Sachs. These findings support structural pre-organization as an explicit design and evaluation target for omission control and motivate evaluating context construction jointly with its utilization in judgment.

Chinese Translation

大型语言模型（LLMs）已被应用于因果发现，但候选图生成很少将潜在相关因果关系的过早遗漏视为明确的设计目标。我们提出了MaSCoD，一个多智能体框架，在直接边判断之前组织候选第三变量和局部结构模式。我们使用GPT-5.4作为主要基础模型，GPT-4o进行复现，对MaSCoD在Auto-MPG、DWD和Sachs数据集上的表现进行了评估。MaSCoD表现出依赖于数据集和基础模型的保留选择性特征，而非均匀的优越性。在所有六个数据集-基础模型设置中，Full在直接边判断之前提供结构假设，其平均召回率和F1值高于No Phase 1，后者则在判断过程中构建结构假设，同时也增加了假阳性率。在DWD数据集上使用GPT-5.4和在Sachs数据集上使用GPT-4o时，观察到相较于所有评估基线的额外参考边保留，而不是在所有设置中均匀分布。部分消融实验表明，提供两个信息组件并不总是优于仅提供一个。对于GPT-5.4，阶段分析显示Full与No Phase 1之间的保留差距在直接边判断后已经存在，而在Sachs上，调和过程为Full引入了额外的参考边损失。这些发现支持将结构预组织作为遗漏控制的明确设计和评估目标，并激励将上下文构建与其在判断中的利用共同评估。

View on arXiv Download PDF AI Translation

cs.AI / 46 / 2609.19947

Not All AI Agents Are Equal: Characterizing Resource and Performance Dynamics

并非所有人工智能代理都是平等的：资源与性能动态特征分析

Choi, Wonmi, Park, Minuk, Niu, Zhixiong, Xiong, Yongqiang, Yoo, Chuck, Yang, Gyeongsik

Abstract

LLM-based AI agents process user requests through iterative reasoning and tool execution, often involving the invocation of remote LLM APIs with local tool containers. This execution model can make the optimization of agent serving difficult because latency, local resource demand, and container bottlenecks inter-mix across requests. However, the current agent ecosystem runs without much consideration of resource dynamics, which results in significant waste of the precious resources. This paper analyzes the resource inter-mix of AI agents for three representative tasks: retrieval-augmented question answering, web search, and software coding. To this end, we characterize the latency with respect to the resource dynamics of processing multiple requests and tasks concurrently. Our measurements show that agents have a wide range of behaviors depending on tasks, so that even the same tool can differ substantially in resource dynamics. We also find that running multiple requests concurrently exposes task-dependent bottlenecks in resource dynamics such as CPU, disk I/O, and memory. Furthermore, we uncover that faster LLM responses or more CPU cores do not always accelerate agents. Based on these observations, we demonstrate new optimization opportunities that exploit the resource dynamics of tasks: CPU-aware tool admission and task-aware CPU allocation. Our results show that the latency of CPU-sensitive agent tasks improves $\sim$5.4$\times$, and the average latency across multiple tasks is reduced $\sim$32% compared to native agents.

Chinese Translation

基于大型语言模型（LLM）的人工智能代理通过迭代推理和工具执行来处理用户请求，通常涉及调用远程LLM API与本地工具容器的结合。这种执行模型使得代理服务的优化变得困难，因为延迟、本地资源需求和容器瓶颈在请求之间交织。然而，当前的代理生态系统在很大程度上没有考虑资源动态，导致宝贵资源的显著浪费。本文分析了三种代表性任务下人工智能代理的资源交织情况：检索增强问答、网页搜索和软件编码。为此，我们对并发处理多个请求和任务的资源动态进行了延迟特征分析。我们的测量结果表明，代理的行为因任务而异，即使是相同的工具在资源动态上也可能有显著差异。我们还发现，多个请求的并发运行暴露了资源动态中与任务相关的瓶颈，例如CPU、磁盘I/O和内存。此外，我们揭示了更快的LLM响应或更多的CPU核心并不总是能加速代理。基于这些观察，我们展示了利用任务资源动态的新优化机会：基于CPU的工具接纳和基于任务的CPU分配。我们的结果表明，CPU敏感型代理任务的延迟改善了约5.4倍，而多个任务的平均延迟较原生代理减少了约32%。

View on arXiv Download PDF AI Translation

cs.AI / 47 / 2609.19961

Neuro-Symbolic Agentic AI for Networked Low-Altitude UAVs

网络低空无人机的神经符号自主人工智能

Ping, Yuqi, Liang, Tianhao, Su, Nanchi, Lei, Guangyu, Wu, Junwei, Zhang, Qinyu, Zhang, Tingting

Abstract

Networked low-altitude unmanned aerial vehicles (UAVs) need reliable and adaptive decision-making capabilities to operate under uncertain observations, dynamic environments, and intermittent connectivity, while many existing agentic systems remain limited by hallucination risks, data dependence, and weak generalization. This article investigates neuro-symbolic agentic AI (NSAAI) as a framework for combining neural grounding, symbolic reasoning, and closed-loop agentic interaction to support more reliable and adaptive UAV autonomy. We first examine its capability foundations in data efficiency, compositional generalization, continual learning, and zero-shot transfer, and then develop a reference architecture integrating task and goal management, neuro-symbolic planning, verification and metacognition, skill execution and network interaction, and shared knowledge and memory. An urban fire-inspection case implemented in LAESim illustrates how a UAV can coordinate sensing and cloud access under intermittent connectivity, reuse a verified image-delivery skill, and satisfy explicit evidence conditions before completing the mission. The results illustrate the potential of NSAAI to support reusable skills, evidence-grounded decision-making, and adaptive mission execution in networked UAV systems. We further discuss key research directions in uncertainty-aware reasoning, knowledge and skill expansion, adaptive self-monitoring, and standardized evaluation.

Chinese Translation

网络低空无人机（UAV）需要可靠且自适应的决策能力，以便在不确定的观测、动态环境和间歇性连接下进行操作，而许多现有的自主系统仍受到幻觉风险、数据依赖和弱泛化能力的限制。本文探讨了神经符号自主人工智能（NSAAI）作为一种框架，旨在结合神经基础、符号推理和闭环自主交互，以支持更可靠和自适应的无人机自主性。我们首先考察其在数据效率、组合泛化、持续学习和零样本迁移方面的能力基础，然后开发一个参考架构，整合任务和目标管理、神经符号规划、验证与元认知、技能执行与网络交互，以及共享知识与记忆。通过在LAESim中实施的城市火灾检查案例，展示了一架无人机如何在间歇性连接下协调传感和云访问，重用经过验证的图像传递技能，并在完成任务之前满足明确的证据条件。结果表明，NSAAI在网络无人机系统中支持可重用技能、基于证据的决策和自适应任务执行的潜力。我们进一步讨论了在不确定性感知推理、知识与技能扩展、自适应自我监控和标准化评估等方面的关键研究方向。

View on arXiv Download PDF AI Translation

cs.AI / 48 / 2609.19996

Customizable and Jointly Optimized Route Planning: A Deep Architecture Enabling Differentiable Shortest-Path Search

可定制的联合优化路线规划：一种支持可微分最短路径搜索的深度架构

Zhao, Rui, Chen, Chao, Xu, Longfei, Ji, Chenguang, Cui, Hengbin, Liu, Kaikui, Li, Xiaolong

Abstract

With the widespread use of online navigation and ride-hailing services, achieving optimal route planning for diverse user preferences has recently attracted increasing attention. Classic graph algorithms for pathfinding use heuristic cost functions to define edge weight, thus providing no optimality guarantee of route quality. Prior data-driven approaches equating ground truth of the optimal route with user trajectory, which is however moderately influenced by the navigation service, suffers from the feedback loop problem. To address these issues, we propose a deep architecture that is able to jointly optimize cost functions and route-ranking model towards any route preference. First, we run a multi-objective Dijkstra algorithm offline to collect the set of Pareto optimal routes, deeming it as the complete candidate set. Exploiting the property of such a set, we design a neural network structure that emulates shortest-path search and route ranking in an end-to-end differentiable manner. Second, we define route preference as a task of constrained optimization of route attributes, and propose a novel loss function that optimizes a single-objective variable, with other variables strictly under constraints. We conduct extensive experiments on real-world datasets. The results show that our architecture significantly outperforms state-of-the-art methods in route quality and customizability.

Chinese Translation

随着在线导航和打车服务的广泛使用，实现针对多样化用户偏好的最佳路线规划最近引起了越来越多的关注。经典的图算法在路径寻找中使用启发式成本函数来定义边权，因此无法保证路线质量的最优性。之前的数据驱动方法将最佳路线的真实情况等同于用户轨迹，但这种轨迹受到导航服务的中等影响，存在反馈循环问题。为了解决这些问题，我们提出了一种深度架构，能够联合优化成本函数和路线排名模型，以满足任何路线偏好。首先，我们离线运行多目标 Dijkstra 算法以收集一组帕累托最优路线，将其视为完整的候选集。利用该集合的特性，我们设计了一种神经网络结构，以端到端可微分的方式模拟最短路径搜索和路线排名。其次，我们将路线偏好定义为对路线属性的约束优化任务，并提出了一种新颖的损失函数，优化单一目标变量，同时严格约束其他变量。我们在真实世界数据集上进行了广泛的实验。结果表明，我们的架构在路线质量和可定制性方面显著优于最先进的方法。

View on arXiv Download PDF AI Translation

cs.AI / 49 / 2609.20001

E-AVI: Evidence-Grounded Multimodal Assessment for Automated Video Interviews

E-AVI：基于证据的多模态自动视频面试评估

Wang, Haoshen, Che, Dongbo, Xie, Zeyi, Du, Yuanjie, Hua, Shicheng, Wang, Xingyu

Abstract

Automated video interview assessment integrates verbal content, acoustic delivery, and visual behavior, yet numerical predictions alone provide limited inspectable support. We present E-AVI, an evidence-grounded framework that extracts timestamped multimodal evidence and integrates dimension-conditioned evidence attention with source-level embeddings for scoring. A shared evidence pool further supports natural-language feedback and follow-up question answering. On RecruitView and a private hospitality dataset, E-AVI consistently outperforms fine-tuned multimodal baselines in rank correlation. Ablation, evidence-deletion, bootstrap, human-audit, and QA analyses characterize the predictive contribution, grounding, and practical utility of the evidence pathway. Together, these results demonstrate that our proposed E-AVI framework improves predictive performance while providing inspectable support for assessment, feedback, and interactive analysis.

Chinese Translation

自动视频面试评估整合了语言内容、声学表现和视觉行为，但仅依靠数值预测提供的可检视支持有限。我们提出了E-AVI，一个基于证据的框架，提取带时间戳的多模态证据，并将维度条件证据注意力与源级嵌入相结合用于评分。共享证据库进一步支持自然语言反馈和后续问题回答。在RecruitView和一个私有的酒店数据集上，E-AVI在排名相关性方面始终优于微调的多模态基准。消融实验、证据删除、引导法、人类审计和问答分析描绘了预测贡献、基础和证据路径的实际效用。综合这些结果表明，我们提出的E-AVI框架在提高预测性能的同时，提供了可检视的评估、反馈和交互分析支持。

View on arXiv Download PDF AI Translation

cs.AI / 50 / 2609.20005

Geopolitical Divisions Across Languages in Large Language Models

大型语言模型中的地缘政治语言分歧

Chupilkin, Maxim

Abstract

People increasingly turn to AI chatbots for news and explanations of world events. But do they receive the same political answers when they ask in different languages? Here we show that the language of a question can change how the same AI systems assess the war in Ukraine. We ask GPT, Claude and Gemini to evaluate twenty statements about the war in 112 languages, collecting 67,200 responses. The balance between Russia-leaning and Ukraine-leaning responses differs across languages. When we group responses by countries' official languages, they follow a pattern resembling worldwide political divisions: relatively more Russia-leaning answers correspond to more favourable public views of Russia, less support for Ukraine in United Nations votes, and less aid to Ukraine. The broad pattern recurs across all three models and remains when individual statement pairs are removed. Our findings suggest a possible route through which information warfare may shape the text used to train AI models, which may in turn spread geopolitical biases.

Chinese Translation

人们越来越多地依赖人工智能聊天机器人获取新闻和世界事件的解释。但当他们使用不同语言提问时，是否会得到相同的政治答案？在这里，我们展示了问题的语言可以改变同一人工智能系统对乌克兰战争的评估。我们要求GPT、Claude和Gemini评估关于战争的二十个陈述，使用112种语言，收集了67,200个回应。倾向于俄罗斯和倾向于乌克兰的回应在不同语言中存在差异。当我们按国家的官方语言对回应进行分组时，它们遵循一种类似于全球政治分歧的模式：相对更多倾向于俄罗斯的回答对应于对俄罗斯的更有利的公众看法、在联合国投票中对乌克兰的支持较少以及对乌克兰的援助较少。这一广泛模式在所有三个模型中都反复出现，并且在移除个别陈述对时仍然存在。我们的发现表明，信息战争可能通过影响用于训练人工智能模型的文本，从而传播地缘政治偏见。

View on arXiv Download PDF AI Translation

cs.AI / 51 / 2609.20026

FedeRICo: Federated Region-Influenced Coupling for Traffic Flow Prediction

FedeRICo：用于交通流预测的联邦区域影响耦合

Orozco, Fermin, Luo, Man, Wahlström, Johan

Abstract

Urban traffic forecasting often relies on information distributed across stakeholders who may be unable to share raw data due to privacy or commercial constraints, motivating federated spatial-temporal approaches. In such federated settings, each client observes traffic over a distinct sensor subgraph with its own spatial topology and temporal dynamics, leading to significant heterogeneity across clients. Existing federated spatial-temporal methods typically rely on model parameter aggregation and provide limited mechanisms for recovering spatial dependencies across client boundaries. This introduces two key limitations. Specifically, parameter aggregation across heterogeneous graph domains tends to dilute client-specific representations, while road network partitioning breaks the propagation of traffic dynamics across client boundaries. To address these challenges, we propose FedeRICo, a federated traffic forecasting framework that combines gradient-level collaboration with boundary-aware residual communication. FedeRICo employs a dual-branch forecasting architecture in which a globally guided branch captures transferable forecasting structure, while a private residual branch preserves client-specific corrections and incorporates boundary residual signals. The global branch is coordinated through gradient alignment across all clients, enabling collaborative optimisation without destructive parameter interference. To recover cross-client spatial dependencies, boundary messages are extracted through a trend-residual decomposition that suppresses periodic structure and communicates only transient spatial-temporal residual signals between physically adjacent clients. Experiments across four real-world traffic forecasting benchmarks demonstrate that FedeRICo consistently outperforms state-of-the-art federated spatial-temporal baselines while maintaining competitive training runtime.

Chinese Translation

城市交通预测通常依赖于分散在各利益相关者之间的信息，这些利益相关者可能由于隐私或商业限制而无法共享原始数据，这促使了联邦时空方法的出现。在这样的联邦环境中，每个客户端在具有自己空间拓扑和时间动态的不同传感器子图上观察交通，导致客户端之间存在显著的异质性。现有的联邦时空方法通常依赖于模型参数聚合，并提供有限的机制来恢复跨客户端边界的空间依赖性。这引入了两个关键限制。具体而言，跨异质图域的参数聚合往往会稀释客户端特定的表示，而道路网络的划分则破坏了交通动态在客户端边界的传播。为了解决这些挑战，我们提出了FedeRICo，一个结合了梯度级协作与边界感知残差通信的联邦交通预测框架。FedeRICo采用双分支预测架构，其中一个全球指导分支捕捉可转移的预测结构，而一个私有残差分支则保留客户端特定的修正并纳入边界残差信号。全球分支通过所有客户端的梯度对齐进行协调，能够在不破坏参数干扰的情况下进行协作优化。为了恢复跨客户端的空间依赖性，通过趋势-残差分解提取边界消息，该分解抑制周期性结构，仅在物理相邻的客户端之间传递瞬态时空残差信号。在四个真实世界交通预测基准上的实验表明，FedeRICo在保持竞争性训练运行时间的同时，始终优于最先进的联邦时空基线。

View on arXiv Download PDF AI Translation

cs.AI / 52 / 2609.20027

Can Data Attribution Filter Out Subliminal Learning? Not Reliably

数据归因能否过滤潜意识学习？不可靠

Weckbecker, Moritz, Jena, Sweta, Müller, Jonas, Kumaraguru, Ponnurangam, Lapuschkin, Sebastian, Samek, Wojciech, Jaburi, Louis, Paulo, Gonçalo

Abstract

Subliminal learning allows language models to transmit behavioral traits through training data with no obvious semantic relationship to those traits, undermining content-based data filtering as a safety intervention. Training data attribution offers an alternative: it identifies the training examples responsible for a given model behavior, independent of their semantic content, and so may apply in exactly the cases where semantic inspection fails. We evaluate three gradient-based attribution methods (GradCos, a contrastive GradCos variant, and EK-FAC) across three models, comparing them against divergence tokens, a strong baseline previously shown to localize subliminal learning (albeit one that requires access to counterfactual teacher models). Filtering at the token level, EK-FAC mitigates a significant part of the effect, the other methods provide little benefit, and all mostly fall short of divergence tokens. Filtering entire samples is less effective for every method, though EK-FAC often gives a stronger signal than divergence tokens in this setting. Success is inconsistent across methods and settings: variants that work well for some model-preference combinations fail for others, and we do not identify a consistent explanation for these differences. Our results suggest that gradient-based attribution can identify data responsible for subliminal learning in some settings, but that some approximations are more reliable than others.

Chinese Translation

潜意识学习使得语言模型能够通过与这些特征没有明显语义关系的训练数据传递行为特征，从而削弱了基于内容的数据过滤作为安全干预的有效性。训练数据归因提供了一种替代方案：它识别出导致特定模型行为的训练示例，而不依赖于其语义内容，因此可能适用于语义检查失效的情况。我们评估了三种基于梯度的归因方法（GradCos、对比型GradCos变体和EK-FAC），并在三种模型上进行比较，将其与发散标记进行对比，后者是一个强基线，之前已被证明能够定位潜意识学习（尽管需要访问反事实教师模型）。在标记级别进行过滤时，EK-FAC缓解了大部分影响，而其他方法提供的好处有限，且总体上都未能超越发散标记。在每种方法中，过滤整个样本的效果较差，尽管在这种情况下EK-FAC通常提供的信号强于发散标记。不同方法和设置的成功率不一致：对某些模型偏好组合有效的变体在其他组合中失效，我们未能找到这些差异的一致解释。我们的结果表明，基于梯度的归因在某些情况下能够识别出导致潜意识学习的数据，但某些近似方法比其他方法更可靠。

View on arXiv Download PDF AI Translation

cs.AI / 53 / 2609.20051

DART: Distillation-Aware Reparameterization for Training-Free LoRA Reuse in Few-Step Video Diffusion Models

DART：用于少步视频扩散模型的无训练LoRA重用的蒸馏感知重参数化

Li, Shihong, Xu, Juntao, JinCao, Tang, Maowen, Huang, Jun, Li, Jintao

Abstract

Step distillation reduces the cost of video generation, but reusing a LoRA trained for a longer trajectory can alter its functional effect or degrade target quality. Static parameter compatibility offers one perspective on this problem; our observations show that similar measured geometry can coexist with different adapter behavior under a shortened denoising schedule. We propose DART, a training-free method that combines low-rank coordinate transport with target-schedule response calibration using forward evaluations and no source training videos. On a four-step Wan2.2 target, DART-F improves the joint quality score from 0.9029 to 0.9227 and changes macro functional retention from -0.4644 to +0.1349. Component analysis shows that calibration accounts for most of the quality improvement, while coordinate transport provides complementary gains when combined with calibration. Adapter-level results reveal positive functional effects for some adapters and strong attenuation with reduced negative functional effects for others. Evaluations on two additional targets show the same aggregate trend. These results motivate evaluating distilled-model LoRA reuse jointly through functional preservation and negative-transfer avoidance, without assuming recovery for every adapter.

Chinese Translation

步骤蒸馏降低了视频生成的成本，但重用为较长轨迹训练的LoRA可能会改变其功能效果或降低目标质量。静态参数兼容性为这一问题提供了一种视角；我们的观察表明，在缩短的去噪调度下，相似的几何测量可以与不同的适配器行为共存。我们提出了DART，这是一种无训练的方法，结合了低秩坐标传输与目标调度响应校准，使用前向评估且不需要源训练视频。在四步Wan2.2目标上，DART-F将联合质量评分从0.9029提高到0.9227，并将宏观功能保留从-0.4644改变为+0.1349。组件分析表明，校准占据了大部分质量提升，而坐标传输在与校准结合时提供了互补增益。适配器级结果显示某些适配器具有积极的功能效果，而其他适配器则在减少负功能效果方面表现出强烈的衰减。对两个额外目标的评估显示了相同的总体趋势。这些结果促使我们通过功能保留和避免负迁移的方式共同评估蒸馏模型的LoRA重用，而不假设每个适配器都能恢复。

View on arXiv Download PDF AI Translation

cs.AI / 54 / 2609.20056

MAGMA-GEN: Validated Recovery Supervision from Ambiguous Failures via Counterfactual Re-Execution

MAGMA-GEN：通过反事实重新执行验证模糊失败的恢复监督

Bernat, Loan, Grard, Matthieu, Herbulot, Ariane, Lamiraux, Florent

Abstract

Hierarchical robotic systems executing long-horizon manipulation tasks must make high-level semantic decisions that orchestrate stochastic low-level skills. In this setting, failed rollouts are ambiguous: a poor downstream state may reflect an invalid high-level decision, partial observation, or a valid decision whose physical execution failed. Traditional supervised learning lacks data for such recovery states, while reinforcement learning struggles with sparse rewards and non-local credit assignment. We propose MAGMA-GEN, an on-policy data-generation pipeline that converts ambiguous failed rollouts into validated recovery supervision. MAGMA-GEN first uses a privileged coach to hypothesize an early decision-level error and propose localized correction or recovery actions. Because this diagnosis is fallible, candidates are retained only if re-execution from the same state under matched conditions improves downstream progress. This produces supervised examples from the agent's own failure distribution without per-step human demonstrations. Evaluated on interactive long-horizon manipulation tasks, MAGMA-GEN improves task success and recovery capabilities, against distillation and trajectory-repair baselines under evolving task constraints in both simulation and real-robot execution.

Chinese Translation

执行长时间操作任务的分层机器人系统必须做出高层次的语义决策，以协调随机的低层次技能。在这种情况下，失败的执行结果是模糊的：一个不良的下游状态可能反映出无效的高层决策、部分观察，或是一个有效决策的物理执行失败。传统的监督学习缺乏此类恢复状态的数据，而强化学习在稀疏奖励和非局部信用分配方面面临挑战。我们提出了MAGMA-GEN，一个在政策内的数据生成管道，将模糊的失败执行结果转化为经过验证的恢复监督。MAGMA-GEN首先利用特权教练假设早期决策级错误，并提出局部的修正或恢复行动。由于这一诊断可能出错，只有在相同状态下在匹配条件下重新执行能够改善下游进展时，候选者才会被保留。这从代理自身的失败分布中生成监督示例，而无需逐步的人类演示。在交互式长时间操作任务的评估中，MAGMA-GEN在模拟和真实机器人执行中，在不断变化的任务约束下，提升了任务成功率和恢复能力，相较于蒸馏和轨迹修复基线表现更佳。

View on arXiv Download PDF AI Translation

cs.AI / 55 / 2609.20057

WiCleanData: Guaranteeing the Type Consistency of Wikidata by Taxonomy Refinement and Constraint Enforcement

WiCleanData：通过分类法精炼和约束执行保证Wikidata的类型一致性

Peng, Yiwen, Jeanmougin, Marc, Bonald, Thomas

Abstract

Because of its collaborative nature, Wikidata suffers from errors, in- consistencies, and excessive complexity, such as redundant classes, ambiguity between instances and classes, wrong taxonomic paths, and type constraint violations. The manual curation of these issues is infeasible at scale. To address these challenges, we introduce WiCleanData, a refined version of Wikidata with a consistent tax- onomy and free from type constraint violations. Specifically, we have designed an automated pipeline that first cleans the taxonomy with language model assistance, then simplifies type constraints by hierarchical aggregation, and finally filters facts accordingly. The resulting knowledge graph, free from any type violation, is made publicly available via a Web interface, enabling easy exploration and downstream applications.

Chinese Translation

由于其协作性质，Wikidata面临着错误、不一致性和过度复杂性的问题，例如冗余类、实例与类之间的模糊性、错误的分类路径以及类型约束违反。手动处理这些问题在大规模上是不可行的。为了解决这些挑战，我们提出了WiCleanData，这是一个经过精炼的Wikidata版本，具有一致的分类法且不违反类型约束。具体而言，我们设计了一个自动化流程，首先在语言模型的帮助下清理分类法，然后通过层次聚合简化类型约束，最后相应地过滤事实。最终生成的知识图谱不含任何类型违反，通过Web界面公开提供，便于探索和后续应用。

View on arXiv Download PDF AI Translation

cs.AI / 56 / 2609.20067

FCA-Guided Counterfactual Explanations for Multi-Modal Breast Cancer Diagnosis: A Framework Achieving Perfect Validity with Emergent Sparsity

基于FCA的多模态乳腺癌诊断反事实解释：实现完美有效性的框架与新兴稀疏性

Isa, Abdullahi, Boukari, Souley, Aliyu, Muhammad

Abstract

Deep learning models for multi-modal breast cancer diagnosis achieve high predictive accuracy but remain clinically unacceptable without actionable, counterfactual explanations. Attribution-based methods (LIME, SHAP) are categorically inapplicable to this purpose, as they generate no alternative instances and thus cannot be evaluated on counterfactual quality metrics. This investigation provides empirical evidence that FCA-Guided Counterfactual (FCA-CF) framework that uses a Formal Concept Analysis (FCA) concept lattice as a hard structural constraint on counterfactual search, operating over a multi-modal TCGA-BRCA dataset. We benchmark against four genuine counterfactual methods: Wachter-style CF, DiCE, FACE, and NICE, evaluated on 60 benign-predicted TCGA-BRCA instances. The FCA-CF framework achieves Validity = 1.0000 (100% of counterfactuals successfully flip the prediction), Sparsity = 2.37 features changed (best among all valid methods), and Proximity = 0.900 (normalised L2-based, matching NICE as joint best). The classifier achieves Accuracy = 0.980, F1 = 0.976, ROC-AUC = 0.9947. Ablation analysis confirms that the FCA lattice constraint is the primary sparsity driver (removing it increases sparsity by +40%, p < 0.001, Cohen's d = 0.78), while Phase C greedy refinement accounts for the largest individual contribution (+113% sparsity increase when disabled, p < 0.001, d = 5.01). FCA-guided counterfactual generation achieves a clinically important Pareto-dominant outcome; it is simultaneously the sparsest and among the most proximate of all valid methods, with perfect validity. The emergent sparsity property arising from lattice topology rather than numerical penalty terms constitutes a structurally novel contribution to the counterfactual explanation literature.

Chinese Translation

用于多模态乳腺癌诊断的深度学习模型虽然能够实现高预测准确率，但在没有可操作的反事实解释的情况下仍然在临床上不可接受。基于归因的方法（如LIME和SHAP）在此目的上是完全不适用的，因为它们不生成替代实例，因此无法在反事实质量指标上进行评估。本研究提供了实证证据，表明基于FCA的反事实（FCA-CF）框架利用形式概念分析（FCA）概念格作为反事实搜索的硬结构约束，应用于多模态TCGA-BRCA数据集。我们与四种真实的反事实方法进行基准比较：Wachter风格的反事实、DiCE、FACE和NICE，评估对象为60个被预测为良性的TCGA-BRCA实例。FCA-CF框架实现了有效性 = 1.0000（100%的反事实成功翻转预测），稀疏性 = 2.37个特征变化（在所有有效方法中最佳），以及接近度 = 0.900（归一化的L2基础，与NICE并列最佳）。分类器的准确率为0.980，F1值为0.976，ROC-AUC为0.9947。消融分析确认FCA格约束是主要的稀疏性驱动因素（移除后稀疏性增加40%以上，p < 0.001，Cohen's d = 0.78），而C阶段贪婪细化则占据了最大的单独贡献（禁用时稀疏性增加113%以上，p < 0.001，d = 5.01）。基于FCA的反事实生成实现了临床上重要的帕累托主导结果；它在所有有效方法中同时是最稀疏和最接近的，且有效性完美。由格拓扑而非数值惩罚项引发的新兴稀疏性特性构成了对反事实解释文献的结构性新贡献。

View on arXiv Download PDF AI Translation

cs.AI / 57 / 2609.20068

Marginal utility, matrix factorization, and the Key-Value (KV) cache: a unified information-economic framework for sovereign geo-mining inference

边际效用、矩阵分解与键值（KV）缓存：主权地质挖掘推断的统一信息经济框架

Combe, Caroline Gans

Abstract

This paper builds a theoretical bridge between the economic notion of marginal utility and two machine-learning constructs, matrix factorization and the Key--Value cache of transformer language models. The singular value spectrum of a rating matrix is shown to be a diminishing marginal utility schedule for latent factors, the eigenvalue spectrum of the projected covariance operator to be the marginal utility schedule of a model's learned representation, and cache eviction and low-rank cache compression to be instances of constrained utility maximization under a memory budget. The three collapse into a single allocation rule: retain the top dimensions whose eigenvalue exceeds the shadow price of the binding constraint. The framework is applied to the automated extraction of structured information from geo-mining documents, where it motivates a multi-pass inference protocol, a layer-wise TIES model merging procedure, and a selection policy combining extraction quality, localization drift and energy, scalarized with a Conditional Value-at-Risk term on drift. Two empirical contributions are reported. An 11.2-million-parameter hierarchical classifier, trained in about five minutes on a single GPU, reaches 90.0 per cent level-1 accuracy on a held-out test set from a 973-document uranium-exploration corpus, against 92.0 per cent for a proprietary model on a fifty-document human audit of the same corpus, at a latency of 2.62 ms per card against approximately 2,000 ms for the API and at negligible cost. A diagnostic of uniform-density TIES merging exposes a reproducible degenerate mode in which the merged model returns token-identical outputs across five geographically distinct districts while declaring high confidence; re-executing the merge under layer-wise calibrated densities removes that signature on the diagnostic sample. The full-scale extraction benchmark, including LoRA fine-tuning, is reported as projected rather than measured and remains an empirical extension of this work.

Chinese Translation

本文建立了经济学中边际效用的理论桥梁与两个机器学习构造——矩阵分解和变换器语言模型的键值缓存之间的联系。研究表明，评分矩阵的奇异值谱是潜在因子的递减边际效用表，投影协方差算子的特征值谱是模型学习表示的边际效用表，而缓存驱逐和低秩缓存压缩则是内存预算下受限效用最大化的实例。这三者归结为一个单一的分配规则：保留特征值超过约束影子价格的顶层维度。该框架应用于从地质挖掘文档中自动提取结构化信息，激励了多次推断协议、逐层TIES模型合并程序以及结合提取质量、定位漂移和能量的选择策略，后者通过对漂移的条件风险价值项进行标量化。报告了两个实证贡献。一种具有1120万个参数的层次分类器，在单个GPU上训练约五分钟后，在973份铀勘探文档的保留测试集上达到了90.0%的一级准确率，而同一语料库中一个五十份文档的人为审计的专有模型则达到了92.0%的准确率，延迟为每个卡片2.62毫秒，相比之下API的延迟约为2000毫秒，且成本微乎其微。均匀密度TIES合并的诊断揭示了一种可重复的退化模式，其中合并模型在五个地理上不同的地区返回相同的标记输出，同时宣称高置信度；在逐层校准密度下重新执行合并消除了该诊断样本上的特征。全规模提取基准，包括LoRA微调，被报告为预测而非测量，并仍然是该工作的实证扩展。

View on arXiv Download PDF AI Translation

cs.AI / 58 / 2609.20077

Tailored to you: longitudinal effects of personalising language models

量身定制：个性化语言模型的长期影响

Akbulut, Canfer, Breuch, Justine, Manzini, Arianna, Ibrahim, Lujain, Franklin, Matija, Patel, Roma, Gabriel, Iason, Lum, Kristian, Weidinger, Laura

Abstract

Interest in developing personalised language models is rapidly growing. While personalisation is often viewed as a mechanism to better serve diverse user needs, the effects of sustained interactions with personalised models on people's perception of and behaviour toward AI remain poorly understood. Most critically, downstream consequences outside the immediate human--AI interaction loop, such as effects on users' self-perceptions and interpersonal relationships, remain largely unexamined. In this study, we recruited 992 participants to complete daily advice-seeking interactions with language models over the course of five days, comparing outcomes from a non-personalised baseline against two personalisation approaches: memory-based (conditioned on prior conversational history) and survey-based (conditioned on information collected through a pre-study intake survey). We find that several changes in human-AI interaction over time are driven primarily by repeated exposure rather than personalisation itself. However, participants interacting with personalised models experienced differences in advice-seeking and information-sharing attitudes and behaviours: participants in the memory-based condition engaged in greater self-disclosure and rated the model as less creepy, while participants in the survey-based condition reported higher regret about having shared personal information with the AI. We conclude by highlighting the nuanced effects of different personalisation approaches on interaction outcomes, and discussing the implications of these findings for the responsible design and deployment of personalised AI systems.

Chinese Translation

对个性化语言模型的开发兴趣正在迅速增长。虽然个性化通常被视为更好地满足多样化用户需求的一种机制，但与个性化模型的持续互动对人们对人工智能的认知和行为的影响仍然缺乏深入理解。最关键的是，超出人机交互即时循环的下游后果，例如对用户自我认知和人际关系的影响，仍然很大程度上未被研究。在本研究中，我们招募了992名参与者，在五天的时间里完成与语言模型的每日寻求建议互动，比较了非个性化基线与两种个性化方法的结果：基于记忆的方法（基于先前对话历史）和基于调查的方法（基于通过预研究入学调查收集的信息）。我们发现，随着时间的推移，人机互动中的若干变化主要是由重复接触驱动，而非个性化本身。然而，与个性化模型互动的参与者在寻求建议和信息共享的态度和行为上经历了差异：在基于记忆的条件下，参与者进行了更多的自我披露，并将模型评估为不那么令人毛骨悚然，而在基于调查的条件下，参与者对与人工智能分享个人信息感到更高的遗憾。我们最后强调了不同个性化方法对互动结果的细微影响，并讨论了这些发现对个性化人工智能系统的负责任设计和部署的启示。

View on arXiv Download PDF AI Translation

cs.AI / 59 / 2609.20080

A Proposal for an Agentic AI Architecture to Support Multi-Domain Decision-Making in the Brazilian Armed Forces

支持巴西武装部队多领域决策的代理人工智能架构提案

Braga, Gioliano de Oliveira, Barbieri, Sidnei, Ferraz, Ágney Lopes Roth, Sonaglio, Wagner Comin, Pereira Jr, Henrique Curi de Miranda e Lourenço Alves

Abstract

The growing complexity of multi-domain operational environments (land, aerospace, naval, cyber, and electromagnetic spectrum) has increased the volume and velocity of data reaching command-and-control (C2) centers, straining the observe-orient-decide-act (OODA) decision cycle. Artificial Intelligence (AI) systems currently employed in defense are, in general, reactive and isolated tools that still rely heavily on human operators to integrate information, assess scenarios, and formulate courses of action. This paper proposes a conceptual Agentic AI architecture for AI systems that can plan, access data sources, execute tools, and act autonomously and audibly, aimed at supporting decision-making across the three Brazilian Armed Forces (Navy, Army, and Air Force). Four application fronts are discussed (decision support, situational analysis, feasibility studies, and countermeasure suggestion), as well as the data and sensor access requirements and the security and permission safeguards necessary for responsible employment across administrative, strategic, operational, and tactical contexts.

Chinese Translation

多领域作战环境（陆地、空天、海洋、网络和电磁频谱）的日益复杂性增加了到达指挥与控制（C2）中心的数据量和速度，给观察-定位-决策-行动（OODA）决策周期带来了压力。目前在国防中使用的人工智能（AI）系统通常是反应性和孤立的工具，仍然在很大程度上依赖人类操作员整合信息、评估情境和制定行动方案。本文提出了一种概念性的代理人工智能架构，旨在支持巴西三军（海军、陆军和空军）之间的决策，能够进行规划、访问数据源、执行工具并自主且清晰地行动。讨论了四个应用前沿（决策支持、情境分析、可行性研究和对策建议），以及数据和传感器访问要求，以及在行政、战略、操作和战术背景下负责任使用所需的安全和权限保障措施。

View on arXiv Download PDF AI Translation

cs.AI / 60 / 2609.20089

UnifiedPlayers: Enhance Tool-Integrated Reasoning in Agentic Reinforcement Learning

统一玩家：增强代理强化学习中的工具集成推理

Liao, Wenjie, Zhao, Liangjie, Cao, Zehong

Abstract

Self-evolving methods reduce the need for human-annotated trajectories by allowing tool-using agents to generate their own training data. Yet existing methods typically separate trajectory generation from evaluation, relying on static verifiers that cannot adapt to emerging failure modes or self-consistency signals that may reinforce errors shared across trajectories. Jointly adapting planning, execution, and evaluation offers a promising alternative, but introduces a fundamental coordination challenge: each component continuously changes the data or feedback used to train the others. We address this challenge with \textbf{UnifiedPlayers}, a cooperative framework comprising a Planning Player that generates tasks, an Execution Player that produces multi-turn trajectories with Python tool calls, and an Evaluation Player that constructs executable verifiers. We design role-specific rewards that coordinate the three players toward a shared learning objective under GRPO. Across two model backbones and twelve reasoning benchmarks, UnifiedPlayers outperforms the strongest prior baseline by at least 3.5\% on mathematical reasoning and 3.9\% on general reasoning tasks. Moreover, the learned verifier achieves 84.2\% adversarial detection accuracy, while its reward signal exhibits 2.03$\times$ higher per-question variance than a self-consistency baseline, providing more discriminative verifications. These results highlight cooperation among specialized players as a promising path toward self-enhanced tool-integrated agents.

Chinese Translation

自我演化方法通过允许使用工具的智能体生成自己的训练数据，减少了对人工标注轨迹的需求。然而，现有方法通常将轨迹生成与评估分开，依赖于静态验证器，这些验证器无法适应新出现的失败模式或可能强化跨轨迹共享错误的自一致性信号。联合适应规划、执行和评估提供了一种有前景的替代方案，但引入了一个基本的协调挑战：每个组件不断改变用于训练其他组件的数据或反馈。我们通过 extbf{统一玩家}（UnifiedPlayers）来解决这一挑战，这是一种合作框架，包括生成任务的规划玩家（Planning Player）、通过Python工具调用生成多轮轨迹的执行玩家（Execution Player）以及构建可执行验证器的评估玩家（Evaluation Player）。我们设计了角色特定的奖励，以协调三个玩家朝着共享学习目标在GRPO下进行。通过两个模型骨干和十二个推理基准，统一玩家在数学推理上比最强的先前基线提高了至少3.5 ext{%}，在一般推理任务上提高了3.9 ext{%}。此外，学习到的验证器达到了84.2 ext{%}的对抗检测准确率，而其奖励信号的每题方差比自一致性基线高出2.03倍，提供了更具辨别力的验证。这些结果突显了专业玩家之间的合作作为朝向自我增强工具集成智能体的有前景的路径。

View on arXiv Download PDF AI Translation

cs.AI / 61 / 2609.20091

Solving Minimum Span Antibandwidth and Cyclic Antibandwidth Labeling Problems

解决最小跨度反带宽和循环反带宽标记问题

Xuan, Hieu Truong, Van, Khanh To

Abstract

The Antibandwidth and Cyclic Antibandwidth problems are NP-hard graph labeling problems that aim to maximize the minimum (cyclic) distance between labels assigned to adjacent vertices. Extensive research on these problems has resulted in a variety of mathematical formulations and computational approaches. However, their minimum span perspective, in which a prescribed minimum (cyclic) distance is fixed and the objective is to minimize the label span, has received comparatively little attention. In this paper, we consider this complementary perspective by introducing the Minimum Span Antibandwidth/Cyclic Antibandwidth Labeling (MSABL/MSCABL) problems and developing a unified Boolean Satisfiability (SAT)-based framework for solving them. The SAT-based framework formulates MSABL/MSCABL as a sequence of decision problems and exploits their monotonicity to accelerate the search process. We also consider two SAT solving strategies, parallel and incremental SAT solving: the former examines multiple candidate spans concurrently, while the latter reuses a single SAT instance while progressively restricting the label domain. The proposed approaches are evaluated on benchmark instances from the Harwell-Boeing Sparse Matrix Collection and compared with CPLEXCP, CPLEXMIP, and Gurobi. The results show that SAT-based approaches are highly competitive in solution quality, with the parallel approach performing best overall for MSCABL and the incremental approach for MSABL. With the no-hole constraint, they remain competitive with CPLEXCP and significantly outperform CPLEXMIP and Gurobi, particularly for MSCABL. These results demonstrate the effectiveness of SAT solving as an exact approach for MSABL and MSCABL.

Chinese Translation

反带宽和循环反带宽问题是NP困难的图标记问题，旨在最大化分配给相邻顶点的标签之间的最小（循环）距离。对这些问题的广泛研究产生了多种数学表述和计算方法。然而，它们的最小跨度视角，即固定一个规定的最小（循环）距离并以最小化标签跨度为目标，受到的关注相对较少。在本文中，我们通过引入最小跨度反带宽/循环反带宽标记（Minimum Span Antibandwidth/Cyclic Antibandwidth Labeling，MSABL/MSCABL）问题，考虑这一互补视角，并开发了一个统一的基于布尔可满足性（Boolean Satisfiability，SAT）的框架来解决这些问题。该SAT框架将MSABL/MSCABL表述为一系列决策问题，并利用它们的单调性来加速搜索过程。我们还考虑了两种SAT求解策略：并行SAT求解和增量SAT求解：前者同时检查多个候选跨度，而后者在逐步限制标签域的同时重用单个SAT实例。所提出的方法在Harwell-Boeing稀疏矩阵集合的基准实例上进行了评估，并与CPLEXCP、CPLEXMIP和Gurobi进行了比较。结果表明，基于SAT的方法在解的质量上具有很强的竞争力，其中并行方法在MSCABL上整体表现最佳，而增量方法在MSABL上表现最佳。在无孔约束下，它们与CPLEXCP保持竞争力，并显著优于CPLEXMIP和Gurobi，特别是在MSCABL上。这些结果证明了SAT求解作为MSABL和MSCABL的精确方法的有效性。

View on arXiv Download PDF AI Translation

cs.AI / 62 / 2609.20110

Perception, Layout, and Validation: Calibrated Confidence for Reliable Straight-Through Processing of Financial Documents

感知、布局与验证：可靠的金融文件直通处理的校准信心

Jin, Yichao, Wang, Yushuo, Han, Yuxuan, Sonia, Kwan Ching Yee, Song, Weiyang, Kent, Chiu Jin-Chun, Hwee, Wong Chong, Kiat, Wong Tiong, Ke, Kenneth Zhu, Zhao, Jingyuan

Abstract

Straight-through processing (STP) on extracted key-value fields from financial documents without human review requires a calibrated probability together with a bounded guarantee on the residual error of the auto-approved tier. The emergence of modern Vision Language Models (VLMs) provides an out-of-the-box capability for extracting the key-values, but their verbalized confidence signals are unreliable and weakly track field correctness. This paper introduces a decomposed confidence layer along three interpretable channels, including perception, layout, and validation. Together with a final conformal risk control, the score can be used for reliable STP of financial documents. The method is validated on three public datasets covering real invoices, synthetic invoices, and ad-buy forms, using two different VLM families (Qwen3.6-27B and Gemini-3.1-Flash-Lite). Our decomposed score consistently improves the separation of correct from incorrect extractions, substantially raising the AUROC from 0.54-0.74 for VLM verbalized signals to 0.90-0.99 with contributions from all three designed channels. Crucially for industrial deployment, this enables usable STP. The native VLM confidence signals could clear only 0.1%-7.0% of fields under risk control at a target error of <10%. In contrast, the proposed method auto-approves 49-72% of fields while holding the empirical error of the accepted tier at or below the target.

Chinese Translation

从金融文件中提取的关键值字段进行直通处理（STP）而无需人工审核，需要一个校准的概率以及对自动批准层残余误差的有界保证。现代视觉语言模型（VLMs）的出现提供了提取关键值的即用能力，但它们的语言化信心信号不可靠，并且对字段正确性追踪较弱。本文引入了一个分解的信心层，涵盖三个可解释的通道，包括感知、布局和验证。结合最终的符合风险控制，该评分可用于金融文件的可靠STP。该方法在三个公共数据集上进行了验证，涵盖真实发票、合成发票和广告购买表单，使用了两种不同的VLM系列（Qwen3.6-27B和Gemini-3.1-Flash-Lite）。我们的分解评分持续改善了正确与错误提取的分离，将VLM语言化信号的AUROC从0.54-0.74显著提升至0.90-0.99，得益于所有三个设计通道的贡献。对于工业部署至关重要的是，这使得可用的STP成为可能。原生VLM信心信号在风险控制下仅能清除0.1%-7.0%的字段，目标误差为<10%。相比之下，所提方法在保持接受层的经验误差在目标范围内的同时，自动批准49-72%的字段。

View on arXiv Download PDF AI Translation

cs.AI / 63 / 2609.20152

MTVA-Bench: Evaluating the Language Model Inside Cascaded Voice Agents

MTVA-Bench：评估级联语音代理中的语言模型

Mishra, Pritish, Kumar, Ishaan, Mandoli, Akshat, Kamath, Sudarshan

Abstract

Generally, most voice agents are cascaded systems, i.e., an ASR model transcribes the caller's audio, a language model reads the transcript and decides what to say and which backend tools to call, and a TTS model speaks the reply. Nearly all of the decision making happens in the language model, but existing evaluations measure it either too broadly or too narrowly. End-to-end voice benchmarks score the full pipeline, so recognition errors and model errors mix into a single number. LLM benchmarks isolate the model but they do not evaluate what makes real phone calls hard, such as transcription issues, caller's voice being split across messages and the requirement that replies follow the language and script specified. We introduce the Multi-Turn Voice Agent Benchmark (MTVA-Bench), which evaluates the language model on the same conditions it faces inside a cascaded system. The caller is played by an LLM following a set of rubrics and tool calls are answered by a mock backend which responds to the arguments the model actually sent. The benchmark contains 49 agents working across 490 reviewed scenarios and supports 7 languages. Scoring is a combination of deterministic checks on tool calls with two LLM judges, one that scores scenario specific rules and one that grades conversation quality without access to the task. Both judges must cite specific messages from the transcript. Task and conversation scores are weighted equally, since a call can complete its task and still go badly for the caller. In a seven-model study, six of the models select the correct tool within 6.4 points of one another, but their overall scores span 24.4 points. Most of the gap comes from argument values, action ordering, rule compliance, and what the model says around its tool calls.

Chinese Translation

一般来说，大多数语音代理是级联系统，即一个自动语音识别（ASR）模型转录来电者的音频，一个语言模型读取转录文本并决定说什么以及调用哪些后端工具，最后一个文本到语音（TTS）模型发出回复。几乎所有的决策都发生在语言模型中，但现有的评估要么过于宽泛，要么过于狭窄。端到端语音基准测试对整个流程进行评分，因此识别错误和模型错误混合成一个单一的分数。大语言模型（LLM）基准测试虽然隔离了模型，但并未评估使真实电话通话变得困难的因素，例如转录问题、来电者的声音在消息中被分割以及回复必须遵循指定的语言和脚本的要求。我们引入了多轮语音代理基准（MTVA-Bench），该基准在级联系统中面对的相同条件下评估语言模型。来电者由遵循一套规则的LLM扮演，工具调用由一个模拟后端回答，该后端响应模型实际发送的参数。该基准包含49个代理，涵盖490个审查场景，并支持7种语言。评分结合了对工具调用的确定性检查以及两位LLM评审的评分，一位评审评分特定场景的规则，另一位评审在不接触任务的情况下评估对话质量。两位评审必须引用转录文本中的具体消息。任务和对话的分数权重相等，因为一个电话可以完成其任务，但仍然对来电者造成不良体验。在一项包含七个模型的研究中，六个模型在6.4分内选择正确的工具，但它们的总体分数跨度达到24.4分。大部分差距来自于参数值、动作顺序、规则遵从性以及模型在工具调用时的表述。

View on arXiv Download PDF AI Translation

cs.AI / 64 / 2609.20177

PaGNet: A Panel-Aware GBDT--Neural Network for Multi-Target Corporate Tax Avoidance Proxy Forecasting

PaGNet：一种面板感知的GBDT-神经网络用于多目标企业税收规避代理预测

Song, Wonho, Kim, Hyungjoon

Abstract

Forecasting corporate tax avoidance proxies from firm--year panel data is challenging because predictive signals are distributed across short firm histories and related targets, while screening-oriented use requires transparent model behavior. We propose PaGNet (Panel-Aware GBDT--Neural Network), a two-branch hybrid that combines a LightGBM branch using panel-temporal summaries with a Panel-MLP branch using attention-pooled temporal aggregation and shared-trunk multi-task learning. A per-target validation-optimal blender produces both the final prediction and a compact branch-reliance diagnostic without trainable fusion parameters. On the KoTaP panel of 1{,}754 Korean listed firms from 2011--2024, PaGNet is evaluated under a leakage-free, shared-hyperparameter protocol across four feature regimes. In the direct-proxy-lag-excluded FS1 regime and the tax-history-augmented FS2 regime, accrual targets (TSTA, TSDA) route stably to the LightGBM branch, where PaGNet raises explained variance over the strongest of six baselines by roughly $0.08$--$0.11$ on the primary split. GETR often leans toward the neural branch, while CETR exposes a validation--test branch-selection mismatch rather than a stable branch assignment. A panel-flatten control shows that most accrual gains come from observed multi-year base-panel values, with PaGNet's panel-aware representation adding a smaller but directionally consistent refinement. Rolling-origin analysis confirms stable accrual routing, bounds ETR diagnostics to split-specific behavior, and identifies a far-horizon split where supervised models underperform naive persistence. PaGNet is therefore best viewed not as a universally superior tabular learner, but as a proxy-aware panel model that combines competitive forecasting with explicit per-target branch-reliance reporting.

Chinese Translation

从公司-年度面板数据中预测企业税收规避代理是具有挑战性的，因为预测信号分布在短期的公司历史和相关目标中，而以筛选为导向的使用需要透明的模型行为。我们提出了PaGNet（面板感知GBDT-神经网络），这是一种双分支混合模型，结合了使用面板时间摘要的LightGBM分支和使用注意力池化时间聚合及共享主干多任务学习的Panel-MLP分支。每个目标的验证最优混合器生成最终预测和一个紧凑的分支依赖诊断，而无需可训练的融合参数。在2011年至2024年间的1,754家韩国上市公司的KoTaP面板上，PaGNet在四个特征模式下采用无泄漏的共享超参数协议进行评估。在直接代理滞后排除的FS1模式和税务历史增强的FS2模式中，应计目标（TSTA，TSDA）稳定地路由到LightGBM分支，PaGNet在主要分割上将解释方差提高了约$0.08$至$0.11$，超过了六个基线中最强的模型。GETR通常倾向于神经分支，而CETR则暴露出验证-测试分支选择的不匹配，而不是稳定的分支分配。面板扁平控制显示，大多数应计收益来自观察到的多年度基础面板值，PaGNet的面板感知表示增加了一个较小但方向一致的细化。滚动起源分析确认了稳定的应计路由，将ETR诊断限制在特定分割行为，并识别出一个远期分割，在该分割中，监督模型的表现不及简单的持久性。因此，PaGNet最好被视为一种代理感知的面板模型，它结合了竞争性的预测与明确的每个目标分支依赖报告。

View on arXiv Download PDF AI Translation

cs.AI / 65 / 2609.20179

Sequential Contextual Fit Predicts Human Behavioural and Neural Dynamics Across Domains

序列上下文契合度预测跨领域的人类行为和神经动态

Sun, Kun, Wang, Rong

Abstract

Human perception, action and decision making unfold in sequences, but computational predictors are often domain-specific. This study computes and tests sequential contextual fit (SCF), an embedding-based measure of how well a current information state matches its recent context. The metric uses a simple recency-weighted similarity kernel and can be applied to words, sounds, visual scenes, affective states, choices, actions and neural representations. Across language processing, music-evoked emotion, a subset of audiovisual emotion EEG data, gambling decisions, human activity recognition and decision-related EEG, lower contextual fit predicted longer processing times, larger affective or behavioural transitions and stronger neural-state changes. These effects remained after controlling for established predictors including surprisal, reinforcement-learning prediction error, acoustic change, visual change and sensor change. SCF therefore provides a computational measurement layer for relating contextual compatibility to behavioural processing and cognitive/neural state-transition dynamics.

Chinese Translation

人类的感知、行动和决策是以序列的形式展开的，但计算预测往往是特定于领域的。本研究计算并测试了序列上下文契合度（Sequential Contextual Fit, SCF），这是一种基于嵌入的度量，评估当前信息状态与其近期上下文的匹配程度。该指标使用简单的近期加权相似性核，可以应用于单词、声音、视觉场景、情感状态、选择、行动和神经表征。在语言处理、音乐引发的情感、一组视听情感的脑电图（EEG）数据、赌博决策、人类活动识别和与决策相关的EEG中，较低的上下文契合度预测了更长的处理时间、更大的情感或行为转变以及更强的神经状态变化。这些效应在控制了包括惊讶度、强化学习预测误差、声学变化、视觉变化和传感器变化等已建立的预测因子后仍然存在。因此，SCF为将上下文兼容性与行为处理和认知/神经状态转变动态相关联提供了计算测量层。

View on arXiv Download PDF AI Translation

cs.AI / 66 / 2609.20200

JointMatch: A Unified Heterogeneous Graph Neural Solver for Large-Scale Ride-Sharing Matching

JointMatch：一种统一的异构图神经网络求解器用于大规模拼车匹配

Zhao, Kun, Chen, Xu

Abstract

Ride-sharing platforms must continuously decide which open requests to bundle into shared trips and which idle vehicles should serve them. The dominant academic approach decomposes this into two sequential matching problems -- request pairing first, then vehicle assignment -- and applies a separate solver to each. This decomposition is convenient computationally but loses revenue and scales poorly because the first stage commits to ride bundles before the available vehicles are known. We propose JointMatch, a learning-based framework that handles request pairing and vehicle assignment together on a single graph. The graph is sparsified by spatial proximity so that its size grows linearly rather than quadratically with the number of vehicles and requests, and a graph neural network scores all candidate decisions in one forward pass. On the New York City Yellow Taxi data, the framework already exceeds both the classical Blossom heuristic and a faithfully-trained two-stage GNN baseline -- often by a wide margin -- and at city scale (fleet 10000) it runs more than $20\times$ faster per dispatch epoch than either. A supervised training stage closes most of the remaining revenue gap, and a policy-gradient fine-tune aligns the trained model with realised revenue.

Chinese Translation

拼车平台必须不断决定将哪些开放请求打包成共享行程，以及哪些闲置车辆应当为其服务。主流的学术方法将此分解为两个顺序匹配问题——首先是请求配对，然后是车辆分配——并对每个问题应用单独的求解器。这种分解在计算上是便利的，但由于第一阶段在已知可用车辆之前就承诺了拼车组合，因此会导致收入损失且扩展性较差。我们提出了JointMatch，一个基于学习的框架，能够在单一图上同时处理请求配对和车辆分配。该图通过空间邻近性进行稀疏化，使其大小随车辆和请求数量的增加呈线性增长而非平方增长，并且图神经网络在一次前向传播中对所有候选决策进行评分。在纽约市黄出租车数据上，该框架的表现已经超越了经典的Blossom启发式算法和经过忠实训练的两阶段GNN基线——通常有着显著的优势——并且在城市规模（车队10000）下，其每个调度周期的运行速度比两者快超过20倍。一个监督训练阶段缩小了大部分剩余的收入差距，而策略梯度微调则使训练模型与实际收入相一致。

View on arXiv Download PDF AI Translation

cs.AI / 67 / 2609.20261

When AI Agents Commit: Cognitive Serializability Across Data, Evidence, Policy, and Authority

当人工智能代理承诺时：跨数据、证据、政策和权威的认知可串行性

He, Jun, Yu, Deying

Abstract

Autonomous agents derive concrete mutations from database reads, retrieved evidence, policy, beliefs, and delegated authority. Those inputs may change while reasoning is in progress. Database isolation orders the submitted transaction; agentic transaction processing determines whether a proposal satisfies an executable contract. Neither guarantee establishes a common valid point for the mutation and its derivation inputs unless the contract represents the relevant predicates. Typed dependency tokens distinguish content integrity from applicability, and trusted mediation captures the values exposed to reasoning. Under strict Cognitive Serializability, committed effects admit a serial order and a logical event at which every value exposed to derivation is unchanged. The fences last until the runtime event that realizes the sealed durability domain. The weaker Effect-Compatible Cognitive Admission recertifies an effect against a simultaneously held current dependency vector and current policy without claiming to serialize the original stochastic derivation. TCT combines immutable versioned executable definitions, registry-derived authority plans, sealed envelopes, guard-first commit transactions, post-seal envelope- and witness-bound grants, co-committed receipts, idempotent grant finalization, and receipt-driven epistemic reconciliation. Complete registered footprints and a single growing phase induce an acyclic lock-point order over local guards and incompatible external reservations. The corresponding results give serializability conditions and an observational-equivalence boundary for zero-error soundness and positive progress. A falsification suite tests the implementation obligations: the prototype prevented all injected anomalies and added 3.22 ms mean commit overhead.

Chinese Translation

自主代理从数据库读取、检索的证据、政策、信念和委托的权威中衍生出具体的变更。这些输入在推理进行时可能会发生变化。数据库隔离对提交的事务进行排序；代理事务处理决定提案是否满足可执行合同的要求。除非合同代表相关谓词，否则这两者都无法建立变更及其衍生输入的共同有效点。类型化依赖令牌区分内容完整性与适用性，可信的中介捕获暴露于推理的值。在严格的认知可串行性下，已承诺的效果承认一个串行顺序和一个逻辑事件，在该事件中，所有暴露于衍生的值保持不变。围栏持续到实现封闭持久性域的运行时事件。较弱的效果兼容认知接纳在同时持有的当前依赖向量和当前政策下重新认证效果，而不声称对原始随机衍生进行串行化。TCT结合了不可变版本的可执行定义、注册派生的权威计划、封闭信封、优先保护的提交事务、后封闭的信封和见证绑定的授权、共同承诺的收据、幂等授权最终化以及基于收据的认知调和。完整的注册足迹和单一增长阶段在本地保护者和不兼容的外部预留之间引入了一个无环锁点顺序。相应的结果给出了零错误健全性和积极进展的可串行性条件及观察等价边界。一个伪造测试套件测试实施义务：原型防止了所有注入的异常，并增加了3.22毫秒的平均提交开销。

View on arXiv Download PDF AI Translation

cs.AI / 68 / 2609.20271

AI-Driven Real-Time Relay Optimisation in Smart Urban NR-V2X Networks via Learning-to-Optimise Graph Neural Networks

基于人工智能的智能城市NR-V2X网络实时中继优化：学习优化图神经网络方法

Amati, Giambattista, Mangiatordi, Federica, Pallotti, Emiliano, Angelini, Simone

Abstract

Reliable and low-latency communication is a fundamental requirement for smart city services and Industry 4.0 applications enabled by NR-V2X networks. However, limited Road-Side Unit (RSU) deployment and complex urban propagation conditions often prevent Connected and Automated Vehicles (CAVs) from maintaining stable connectivity. This paper proposes an AI-driven Learning-to-Optimise (L2O) framework based on Graph Neural Networks (GNNs) for real-time multi-hop relay selection in NR-V2X systems. The vehicular network is modelled as a graph, where nodes represent CAVs and RSUs, and edges encode radio-link characteristics. An offline Mixed-Integer Linear Programming (MILP) formulation provides optimal relay decisions used as supervision for training an edge-aware Graph Isomorphism Network with Edge Features (GINE). Extensive experiments on realistic urban datasets demonstrate that the proposed approach achieves near-optimal connectivity performance, recovering up to 11.3% connectivity gain, while reducing execution time by orders of magnitude (up to 100 x speed-up) compared to MILP. The framework enables scalable and real-time network control, making it suitable for smart city and Industry 4.0 deployments.

Chinese Translation

可靠且低延迟的通信是由NR-V2X网络支持的智能城市服务和工业4.0应用的基本要求。然而，有限的路边单元（RSU）部署和复杂的城市传播条件常常阻碍连接与自动驾驶车辆（CAVs）维持稳定的连接性。本文提出了一种基于图神经网络（GNNs）的人工智能驱动的学习优化（L2O）框架，用于NR-V2X系统中的实时多跳中继选择。车辆网络被建模为一个图，其中节点代表CAV和RSU，边缘编码无线链路特性。离线的混合整数线性规划（MILP）模型提供了用于训练具有边特征的图同构网络（GINE）的监督的最优中继决策。对真实城市数据集的广泛实验表明，所提出的方法实现了接近最优的连接性能，恢复了高达11.3%的连接增益，同时与MILP相比，执行时间减少了几个数量级（最高可达100倍加速）。该框架实现了可扩展的实时网络控制，适用于智能城市和工业4.0的部署。

View on arXiv Download PDF AI Translation

cs.AI / 69 / 2609.20277

JEPA-WAM: Connecting Generated Visual Instructions to World Action Models through JEPA Latent Representations

JEPA-WAM：通过JEPA潜在表示将生成的视觉指令与世界行动模型连接起来

Liu, Tianbin, Zhu, Jian, Su, Taiyi, Zhang, Jianjun, Ma, Chong, Huang, Zitai, Xu, Yi

Abstract

World Action Models (WAMs) have demonstrated strong robotic manipulation capabilities by augmenting pretrained video generative models with action experts. However, current WAMs still show limited instruction-following ability when conditioned solely on text instructions. We argue that this limitation stems in part from a structural imbalance in robot-learning data: rich visual-action trajectories are often paired with sparse and repetitive language annotations, allowing policies to identify tasks from visual context and motion regularities rather than grounding the instruction itself. To address this limitation, we introduce JEPA-WAM, which augments each text instruction with a bank of stochastically generated visual instructions, providing diverse visual cues for instruction following. Specifically, JEPA-WAM uses an off-the-shelf text-to-image generator to sample multiple task-completion images conditioned on the text instruction, without training the generator. Although these generated images may differ from the current visual scene in appearance and layout, they remain semantically aligned with the instruction and serve as visual goal references. To focus on task-level semantics beyond appearance, we encode these references with a frozen V-JEPA 2.1 encoder. The resulting dense goal representations are compressed into compact goal tokens that condition both the video and action experts through cross-attention. We further construct a real-robot instruction-following benchmark covering in-distribution, out-of-distribution scene, and out-of-distribution instruction settings. On this benchmark, JEPA-WAM achieves success rates of 87.3%, 74.5%, and 80.9% in these three settings, outperforming {\pi}0 and Fast-WAM by at least 10.0, 27.3, and 14.5 percentage points, respectively.

Chinese Translation

世界行动模型（WAMs）通过将预训练的视频生成模型与行动专家相结合，展示了强大的机器人操作能力。然而，目前的WAM在仅依赖文本指令时，仍表现出有限的指令跟随能力。我们认为，这一限制部分源于机器人学习数据中的结构性不平衡：丰富的视觉-动作轨迹通常与稀疏且重复的语言注释配对，使得策略能够从视觉上下文和运动规律中识别任务，而不是基于指令本身进行定位。为了解决这一限制，我们提出了JEPA-WAM，它为每个文本指令增添了一组随机生成的视觉指令，为指令跟随提供了多样的视觉线索。具体而言，JEPA-WAM使用现成的文本到图像生成器，根据文本指令采样多个任务完成图像，而无需训练生成器。尽管这些生成的图像在外观和布局上可能与当前视觉场景有所不同，但它们在语义上与指令保持一致，并作为视觉目标参考。为了关注超越外观的任务级语义，我们使用冻结的V-JEPA 2.1编码器对这些参考进行编码。生成的稠密目标表示被压缩为紧凑的目标标记，通过交叉注意力同时调节视频和行动专家。我们进一步构建了一个真实机器人指令跟随基准，涵盖了分布内、分布外场景和分布外指令设置。在该基准上，JEPA-WAM在这三种设置中的成功率分别为87.3%、74.5%和80.9%，分别比{C0}0和Fast-WAM高出至少10.0、27.3和14.5个百分点。

View on arXiv Download PDF AI Translation

cs.AI / 70 / 2609.20301

AgentPProf: Semantic Profiler for Long Horizon AI Agents

AgentPProf：用于长时间运行AI代理的语义分析器

Zheng, Yusheng, Chang, Chaokun, Mao, Yu, Wu, Tianyuan, Huang, Yuxi, Ma, Tao, Mao, Wenan, Cheng, Shuyi, Quinn, Andi, Wang, Wei

Abstract

AI agents increasingly orchestrate long-running activities with users, tools, and system resources for days and weeks. To improve agent quality, safety, and cost efficiency, developers need to determine where failures happen, what triggers unsafe effects, and which tasks consume the most budget, then optimize those tasks. In systems software, profiling answers similar questions by aggregating resource consumption and attributing it to responsible code paths to identify hotspots. Yet existing agent observability tools focus on per-execution debugging and tracing rather than cross-run, long term profiling, making these questions difficult to answer at scale. Agent observability needs profiling, not only debugging, but profiling agents is challenging: the responsible entities are task intent like diagnose authentication, compare branches rather than code paths, and lack stable identifiers for aggregation. We propose a semantic operation stack model that adapts profiling to agent trajectories. Uniform operations represent all activities, and operation stacks replace the runtime call stack, enabling hierarchical attribution at different granularities. We observe that an agent's task occupies a contiguous span and decomposes into subtasks, so we introduce recursive operation segmentation, which recursively splits trajectories at task boundaries. AgentPProf is a profiler that aggregates agent trajectories into pprof-compatible profiles, enabling flame graph visualization and analysis. AgentPProf reaches 0.764 $B^3$ F1 against human annotations on CodeTraceBench. On three problem-localization benchmarks, the profile raises MAP by up to 56%, demonstrating that it effectively attributes resources, locates problems, and helps optimize token cost at practical profiling cost. AgentPProf is available at https://github.com/eunomia-bpf/agentsight.

Chinese Translation

AI代理越来越多地与用户、工具和系统资源协调进行长时间的活动，持续数天或数周。为了提高代理的质量、安全性和成本效率，开发者需要确定故障发生的地点、触发不安全效果的原因以及哪些任务消耗了最多的预算，并对这些任务进行优化。在系统软件中，分析可以通过聚合资源消耗并将其归因于负责的代码路径来识别热点，从而回答类似的问题。然而，现有的代理可观测性工具主要集中于单次执行的调试和追踪，而非跨运行的长期分析，使得在大规模上回答这些问题变得困难。代理可观测性需要分析，而不仅仅是调试，但分析代理是具有挑战性的：负责的实体是任务意图，如诊断身份验证、比较分支，而不是代码路径，并且缺乏稳定的标识符进行聚合。我们提出了一种语义操作栈模型，将分析适应于代理轨迹。统一操作表示所有活动，操作栈替代了运行时调用栈，使得在不同粒度上实现层次归因。我们观察到代理的任务占据一个连续的范围，并分解为子任务，因此我们引入了递归操作分段，递归地在任务边界拆分轨迹。AgentPProf是一个分析器，将代理轨迹聚合成与pprof兼容的分析文件，支持火焰图可视化和分析。AgentPProf在CodeTraceBench上对人类注释的F1值达到0.764 $B^3$。在三个问题定位基准上，该分析提高了MAP值最高达56%，证明它有效地归因资源、定位问题，并在实际分析成本下帮助优化令牌成本。AgentPProf可在https://github.com/eunomia-bpf/agentsight获取。

View on arXiv Download PDF AI Translation

cs.AI / 71 / 2609.20304

Diagnose, Recover, Certify: Task Readiness under Hidden Dynamics Changes

诊断、恢复、认证：隐藏动态变化下的任务准备性

Kiet, Nguyen Viet Tuan, Binh, Huynh Thi Thanh

Abstract

A deployed control policy can conceal consequential dynamics changes: an actuator may lose effectiveness without affecting the current task when the policy rarely excites it, despite being critical for a future task that has not yet been specified. We introduce task readiness under dormant dynamics drift, a decision problem that unifies active change diagnosis and post-change control recovery under a limited, task-agnostic interaction budget. An agent must identify whether and where local dynamics have changed, use a small number of informative interactions to characterize the change before downstream task identity is revealed, and subsequently provide each candidate task with either a recovered policy and a calibrated lower bound on its achievable return or an abstention decision to a safe fallback. We propose Evidence-Gated Matched-Pulse Transport, an intervention-based Bayesian procedure that couples fault localization with estimation of actuator effectiveness through a shared matched-response representation, thereby preserving diagnostic reliability while converting localized evidence into recovery-relevant uncertainty. This uncertainty is propagated to task-conditioned policy selection and readiness certification, enabling deployment decisions that explicitly trade off expected performance, confidence, and fallback use. We evaluate the resulting framework on a diverse suite of dormant-actuator benchmarks spanning multiple simulators, under a protocol that separates diagnosis from capability recovery, scores deployment by readiness coverage, selective risk, and interaction cost as well as return, and identifies the fault regimes in which transported evidence is decisive.

Chinese Translation

已部署的控制策略可能会掩盖重要的动态变化：尽管某个执行器在当前任务中几乎没有被激励而未影响其效果，但它可能在未来尚未明确指定的任务中至关重要。我们引入了潜在动态漂移下的任务准备性，这是一个决策问题，统一了在有限的、与任务无关的交互预算下的主动变化诊断和变更后的控制恢复。代理必须识别局部动态是否以及在哪里发生了变化，利用少量信息丰富的交互在下游任务身份揭示之前表征变化，并随后为每个候选任务提供恢复的策略及其可实现回报的校准下限，或者做出安全后备的放弃决策。我们提出了证据门控匹配脉冲传输（Evidence-Gated Matched-Pulse Transport），这是一种基于干预的贝叶斯程序，通过共享的匹配响应表示将故障定位与执行器有效性估计相结合，从而在将局部证据转化为恢复相关的不确定性时保持诊断可靠性。这种不确定性被传播到任务条件的策略选择和准备性认证中，使得部署决策能够明确权衡预期性能、信心和后备使用。我们在多个模拟器的多样化潜在执行器基准上评估了该框架，采用的协议将诊断与能力恢复分开，通过准备覆盖、选择性风险、交互成本以及回报来评分部署，并识别传输证据具有决定性的故障模式。

View on arXiv Download PDF AI Translation

cs.AI / 72 / 2609.20323

NeuSOGA3D: A Neuro-Symbolic Framework for Explainable 3D Geometric Reconstruction

NeuSOGA3D：一种用于可解释的三维几何重建的神经符号框架

Li, Qingde, Hong, Qingqi, Li, Zihan, Tian, Jie

Abstract

Three-dimensional reconstruction from unorganized point clouds remains a challenging problem in computer vision, geometric modeling, and computer-aided design. While neural implicit methods achieve impressive reconstruction accuracy, geometry is typically encoded in latent representations that limit interpretability and reuse within engineering workflows. We present NeuSOGA3D (Neuro-Symbolic Geometric Abstraction in 3D), a hybrid framework that combines learned perceptual priors inherited from NeuSOGA with explicit symbolic geometric reasoning. The method projects point clouds onto principal orthographic planes, constructs symbolic implicit spline representations from the resulting observations, and fuses them through shape-preserving constructive solid geometry operations to generate a coarse visual hull. Additional geometric detail is recovered through cross-sectional decomposition and volumetric reconstruction using Partial Shape-Preserving Splines. Unlike conventional neural implicit approaches, NeuSOGA3D progressively transforms observations into explicit symbolic entities, including control polygons, implicit spline fields, cross-sections, and volumetric lofts. Experiments on all forty categories of the ModelNet40 benchmark demonstrate the ability of the framework to recover structurally meaningful and CAD-compatible geometric representations from diverse point-cloud observations. The results highlight the potential of combining learned perception with symbolic geometric reasoning for explainable geometric intelligence.

Chinese Translation

从无组织点云进行三维重建仍然是计算机视觉、几何建模和计算机辅助设计中的一个挑战性问题。虽然神经隐式方法在重建精度上取得了令人印象深刻的成果，但几何通常以潜在表示的形式编码，这限制了其在工程工作流程中的可解释性和重用性。我们提出了NeuSOGA3D（Neuro-Symbolic Geometric Abstraction in 3D），这是一个混合框架，结合了从NeuSOGA继承的学习感知先验与显式符号几何推理。该方法将点云投影到主要的正交平面上，从生成的观测中构建符号隐式样条表示，并通过保持形状的构造实体几何操作将它们融合，以生成粗略的视觉外壳。通过横截面分解和使用部分形状保持样条的体积重建恢复额外的几何细节。与传统的神经隐式方法不同，NeuSOGA3D逐步将观测转化为显式符号实体，包括控制多边形、隐式样条场、横截面和体积造型。对ModelNet40基准的四十个类别进行的实验展示了该框架从多样的点云观测中恢复结构上有意义且与CAD兼容的几何表示的能力。结果突显了将学习感知与符号几何推理相结合以实现可解释的几何智能的潜力。

View on arXiv Download PDF AI Translation

cs.AI / 73 / 2609.20334

Structured Four-Stage Legal Translation: From Natural-Language Traffic Rules to PROLOG

结构化四阶段法律翻译：从自然语言交通规则到PROLOG

Zin, May Myo, Fungwacharakorn, Wachara, Satoh, Ken, Nitta, Katsumi

Abstract

Traffic regulations are written for human interpretation and therefore rely on shared background knowledge and flexible phrasing, which inherently introduce ambiguity, context dependence, and semantic underspecification. These linguistic characteristics conflict with the precision required by computational reasoning engines such as Prolog, which demand explicit logical structure. This study evaluates two baseline translation approaches, Natural Language to Prolog ($NL\rightarrow Prolog$) and Logical English to Prolog ($LE\rightarrow Prolog$), and introduces a new reasoning-guided translation framework called Structured Four-Stage Legal Translation ($S4L\rightarrow Prolog$). The proposed S4L framework performs semantic role extraction, scene completion, logical mapping, and Prolog rule generation within a single guided prompt, enabling direct translation of raw traffic rules into executable logic without human intervention. A benchmark consisting of twenty real-world traffic rules was used to evaluate each approach in terms of syntactic validity, semantic correctness, and logical completeness. $S4L\rightarrow Prolog$ achieves the highest accuracy, correctly formalizing 75 percent of the rules, while $NL\rightarrow Prolog$ reaches 60 percent and $LE\rightarrow Prolog$ reaches 55 percent. Qualitative analysis further shows that S4L captures implicit causal relations, deontic modality, and exception structure more reliably than the baselines. These results demonstrate that structured reasoning prompts can substantially improve the reliability of natural-language-to-logic translation for legal and safety-critical applications.

Chinese Translation

交通法规是为人类解读而编写的，因此依赖于共享的背景知识和灵活的措辞，这本质上引入了模糊性、上下文依赖性和语义不完全性。这些语言特征与计算推理引擎（如Prolog）所需的精确性相冲突，后者要求明确的逻辑结构。本研究评估了两种基线翻译方法：自然语言到Prolog（$NL ightarrow Prolog$）和逻辑英语到Prolog（$LE ightarrow Prolog$），并引入了一种新的推理引导翻译框架，称为结构化四阶段法律翻译（$S4L ightarrow Prolog$）。所提出的S4L框架在单个引导提示中执行语义角色提取、场景补全、逻辑映射和Prolog规则生成，从而实现将原始交通规则直接翻译为可执行逻辑，而无需人工干预。使用包含二十条真实交通规则的基准来评估每种方法在语法有效性、语义正确性和逻辑完整性方面的表现。$S4L ightarrow Prolog$ 达到最高准确率，正确形式化了75%的规则，而 $NL ightarrow Prolog$ 达到60%，$LE ightarrow Prolog$ 达到55%。定性分析进一步表明，S4L在捕捉隐含因果关系、义务情态和例外结构方面比基线方法更可靠。这些结果表明，结构化推理提示可以显著提高自然语言到逻辑翻译在法律和安全关键应用中的可靠性。

View on arXiv Download PDF AI Translation

cs.AI / 74 / 2609.20349

A Qualitative Model for Reasoning about Path and Support

关于路径和支持推理的定性模型

Jaiswal, Abhishek, Falomir, Zoe

Abstract

Spatial reasoning abilities correlate strongly with performance in STEM fields. Games offer a compelling medium for training these critical skills in developing children who have a natural proclivity for play. However, to facilitate human-like tutoring and player guidance, these games require an AI agent capable of making commonsense inferences from spatial events. Qualitative reasoning (QR) models appear to be a suitable framework for these application domains. As these models reason in symbolic representations, they can seamlessly translate game states into interpretable feedback for human-like player guidance. This paper introduces a hybrid qualitative model designed for Camelot Jr., a block-puzzle game that requires constructing multi-level bridges to connect two avatars stationed on separate towers. The game poses a challenge for the player, who must make platforms stable, plan their path, and ensure they use all the provided blocks. To handle the precise physics required by the domain, we integrate a mathematical center-of-mass stability logic to guide our qualitative solver. Our work facilitates spatial skill training in Camelot Jr. and contributes to the development of human-centric, explainable game-playing agents.

Chinese Translation

空间推理能力与STEM领域的表现密切相关。游戏为培养这些关键技能提供了一个引人注目的媒介，尤其是在那些天生倾向于游戏的儿童中。然而，为了实现类人化的辅导和玩家引导，这些游戏需要一个能够从空间事件中进行常识推理的人工智能代理。定性推理（Qualitative Reasoning, QR）模型似乎是这些应用领域的合适框架。由于这些模型以符号表示进行推理，它们可以无缝地将游戏状态转化为可解释的反馈，以便进行类人化的玩家引导。本文介绍了一种为《Camelot Jr.》设计的混合定性模型，该游戏是一款块状拼图游戏，要求玩家构建多层桥梁以连接驻扎在不同塔上的两个化身。该游戏对玩家提出了挑战，玩家必须使平台稳定，规划路径，并确保使用所有提供的方块。为了处理该领域所需的精确物理，我们整合了一种数学重心稳定逻辑，以指导我们的定性求解器。我们的工作促进了《Camelot Jr.》中的空间技能训练，并为开发以人为本的、可解释的游戏代理做出了贡献。

View on arXiv Download PDF AI Translation

cs.AI / 75 / 2609.20358

Generating Heterogeneous 3D Geological Microstructures from 2D Images via a Stable Diffusion-Adversarial Model

通过稳定扩散-对抗模型从二维图像生成异质三维地质微观结构

Aouf, Ali, Laloy, Eric, Rogiers, Bart, De Vleeschouwer, Christophe

Abstract

Characterizing the physical properties of clay and cementitious materials matters across many fields, from materials science to geological waste disposal. Property simulation typically calls for 3D imaging, which is expensive, not always accessible, and technically limited for certain materials. Recent progress in deep generative models offers a way around this, reconstructing 3D volumes from the more easily acquired 2D images. Among GAN-based methods for 3D microstructure generation, SliceGAN has shown strong results for homogeneous isotropic and anisotropic systems. It struggles, however, to capture the finer detail of more complex heterogeneous microstructures, which motivates alternative generative frameworks. We introduce a hybrid approach that draws on the stability and generation quality of denoising diffusion models. Since no 3D ground truth is available, we replace the standard denoising loss with an adversarial loss, which yields a stable training process in our experiments. We show that the resulting model generates microstructures of varying complexity with minimal slice artefacts and close agreement with ground-truth phase fractions and structural descriptors.

Chinese Translation

表征粘土和水泥材料的物理特性在材料科学到地质废物处置等多个领域都至关重要。属性模拟通常需要三维成像，但这既昂贵又不总是可获得，对于某些材料在技术上也存在限制。深度生成模型的最新进展提供了一种解决方案，可以从更易获取的二维图像重建三维体积。在基于生成对抗网络（GAN）的方法中，SliceGAN在均匀各向同性和各向异性系统的三维微观结构生成方面表现出色。然而，它在捕捉更复杂的异质微观结构的细节方面存在困难，这促使我们探索替代的生成框架。我们提出了一种混合方法，利用去噪扩散模型的稳定性和生成质量。由于没有可用的三维真实值，我们用对抗损失替代了标准的去噪损失，这在我们的实验中产生了稳定的训练过程。我们展示了所生成的模型能够以最小的切片伪影生成不同复杂度的微观结构，并与真实相位分数和结构描述符高度一致。

View on arXiv Download PDF AI Translation

cs.AI / 76 / 2609.20449

The Organization of Inference: Information, Resource Constraints, and AI Production

推理的组织：信息、资源限制与人工智能生产

Zhang, Yukun, Xu, Kemu, Chen, Yishen

Abstract

The economic value of inference depends on how capacity and task information are distributed across stages of AI production. We study these organizational margins using controlled workflow experiments on externally verified software-engineering tasks. In two matched resource panels, direct execution records the same success rate of 59.6 percent at logical-token ceilings of 12,000 and 24,000, while success under information-constrained planning rises from 36.2 to 51.2 percent. The planning disadvantage narrows by 15.0 percentage points (95 percent task-cluster bootstrap interval: 4.2 to 25.8). A strict read-only planning campaign varies whether the planner sees the task issue. At 12,000 tokens, issue access raises success by about 16 percentage points over issue-hidden planning. Compared with direct execution, task-informed planning is about 10 points lower at 12,000 tokens; at 24,000 tokens, it shows a 29.6-point advantage. In the resource panels, direct execution uses substantially less than either ceiling, while the planning workflow's binding rate falls from 46.2 to 0.8 percent and downstream execution accounts for 89.9 percent of the increase in total use. Scale determines the capacity available to a system; workflow and information structure shape the productive value

Chinese Translation

推理的经济价值取决于在人工智能生产的各个阶段中能力和任务信息的分布。我们通过对外部验证的软件工程任务进行控制工作流实验，研究这些组织边际。在两个匹配的资源面板中，直接执行在逻辑令牌上限为12,000和24,000时记录相同的成功率为59.6%，而在信息受限的规划下，成功率从36.2%上升至51.2%。规划劣势缩小了15.0个百分点（95%任务集群自助法区间：4.2至25.8）。严格的只读规划活动会影响规划者是否能够看到任务问题。在12,000个令牌时，问题访问使成功率比隐藏问题的规划提高约16个百分点。与直接执行相比，在12,000个令牌时，任务知情规划的成功率低约10个百分点；而在24,000个令牌时，显示出29.6个百分点的优势。在资源面板中，直接执行的使用量显著低于任一上限，而规划工作流的绑定率从46.2%降至0.8%，下游执行占总使用量增加的89.9%。规模决定了系统可用的能力；工作流和信息结构塑造了生产价值。

View on arXiv Download PDF AI Translation

cs.AI / 77 / 2609.20455

SkillAA: Attribution-Guided Skill-Graph Updating with Targeted Validation and Rollback

SkillAA：基于归因的技能图更新与针对性验证和回滚

Shang, Ziqiao, Ge, Ling-Yue, Guo, Lan-Zhe

Abstract

External skills provide domain procedures without parameter updates, but existing methods often edit skills directly from failed rollouts without structured routing from an observed failure to an editable location; existing skill graphs also underuse semantic boundaries, object addresses, and topological dependencies for skill retrieval, targeted updating, and scoped validation. We introduce SkillAA (Skill Abductive Attribution), a structured skill-optimization framework for frozen language models. It represents skill applicability, execution, and composition in a unified graph, allowing the same structure to support skill selection, attribution-guided repair, and update validation. SkillAA contrasts successful and failed executions to route candidate repairs to specific graph objects, updates only the selected local structure, and uses Local and Big Gates to screen candidate changes before commitment. With gpt-5.6-sol, SkillAA reaches 81.5%, 66.7%, and 91.2% on SearchQA, LiveMath, and DocVQA, respectively, and attains the highest observed mean in every main setting. These results support the utility of attribution-guided graph editing and graph-scoped validation.

Chinese Translation

外部技能提供了无需参数更新的领域程序，但现有方法通常直接从失败的执行中编辑技能，而没有从观察到的失败到可编辑位置的结构化路由；现有的技能图也未充分利用语义边界、对象地址和拓扑依赖关系进行技能检索、针对性更新和范围验证。我们提出了SkillAA（技能归纳归因），这是一个针对冻结语言模型的结构化技能优化框架。它在统一图中表示技能的适用性、执行和组合，允许相同的结构支持技能选择、基于归因的修复和更新验证。SkillAA对比成功和失败的执行，将候选修复路由到特定的图对象，仅更新所选的局部结构，并使用局部门（Local Gates）和大门（Big Gates）在承诺之前筛选候选更改。使用gpt-5.6-sol，SkillAA在SearchQA、LiveMath和DocVQA上的表现分别达到81.5%、66.7%和91.2%，并在每个主要设置中获得了观察到的最高均值。这些结果支持了基于归因的图编辑和图范围验证的实用性。

View on arXiv Download PDF AI Translation

cs.AI / 78 / 2609.20474

How Do Agent Harnesses Create Value? Planning Information and Release Control in Stateful LLM Agents

代理工具如何创造价值？状态化大语言模型代理中的规划信息与发布控制

Zhang, Yukun, Xu, Kemu, Chen, Yishen

Abstract

Agent harnesses supply planning guidance, organize execution, and check completion. We study how these components affect success, erroneous acceptance, and cost in two Retail experiments and an Airline pilot in $\tau^2$-bench. The primary comparison pairs prewritten task-specific plans (Fixed) with shuffled policy text matched in word count (Sham), isolating the contribution of guidance content. Across 265 matched cells, Fixed improves oracle-verified success by 7.17 percentage points (90\% task-clustered bootstrap interval, 1.15--13.36 points), with gains concentrated in higher-complexity tasks. A read-only terminal verifier rejects 61\% of Retail oracle-invalid episodes while withholding 17\% of correct ones, at less than one cent of additional cost per episode. Which component matters more depends on the loss assigned to erroneous acceptance: at low liability the planning gain dominates; at high liability the verifier's avoided false passes dominate---and a standalone verifier captures nearly all the false-pass benefit of the full planning-plus-verification stack at a fraction of its cost.

Chinese Translation

代理工具提供规划指导、组织执行并检查完成情况。我们研究这些组件如何影响成功率、错误接受率和成本，通过两个零售实验和一个航空公司试点在 $ au^2$-bench 中进行分析。主要比较将预先编写的任务特定计划（固定）与按字数匹配的随机政策文本（虚假）进行对比，以隔离指导内容的贡献。在265个匹配单元中，固定计划使得经过oracle验证的成功率提高了7.17个百分点（90 ext{ extperthousand} 任务聚类自助区间，1.15--13.36点），增益主要集中在高复杂度任务上。只读终端验证器拒绝了61 ext{ extperthousand}的零售oracle无效事件，同时保留了17 ext{ extperthousand}的正确事件，每个事件的额外成本不到一美分。哪个组件更重要取决于对错误接受的损失评估：在低责任情况下，规划收益占主导；在高责任情况下，验证器避免的错误通过占主导——而独立的验证器几乎捕获了完整规划加验证堆栈的所有错误通过收益，且成本仅为其一小部分。

View on arXiv Download PDF AI Translation

cs.AI / 79 / 2609.20519

SoL-Pi: Recursively Scaling Auto-Research Loops for Efficient Agent Harness

SoL-Pi：递归扩展的自动研究循环以实现高效的代理利用

Liu, Haozhe, Ye, Tian, Gao, Sensen, Cao, Qihang, Li, Yitong, Zhuge, Mingchen, Wang, Duomin, Zhang, Ruihua, Luo, Ping, Bian, Jiawang, Zhu, Lei, Zhu, Ligeng, Xie, Enze, Han, Song

Abstract

As coding agents move from supervised code completion to unattended, around-the-clock exploration, their work expands from isolated predictions into long trajectories of reasoning, tool use, and feedback. Token efficiency therefore becomes important for scaling recursive self-improvement. We take an RSI-inspired approach at the harness layer, scaling auto-research loops across increasingly numerous and diverse environments for harness rollouts. At this scale, the process yields reusable improvements that transfer beyond their development setting, moving automated harness discovery toward production-level outcomes. Four mechanisms survive selection and form SoL-Pi, spanning action execution, context compaction, observation handling, and delegated reading. On the 51-task EdgeBench evaluation, SoL-Pi achieves performance comparable to Pi across GPT-5.6 Sol and Opus 5 while reducing recorded token traffic by 44.7-49.0% and API cost by about one third. In other words, estimated hourly savings are \$8.75-\$13.50 relative to native Codex and Claude Code harnesses, and \$4.36-\$5.71 relative to Pi.

Chinese Translation

随着编码代理从监督代码补全转向无人值守的全天候探索，它们的工作从孤立的预测扩展到长时间的推理、工具使用和反馈。因此，令牌效率在递归自我改进的扩展中变得至关重要。我们在利用层采用了受 RSI 启发的方法，跨越越来越多且多样化的环境扩展自动研究循环，以实现高效的利用部署。在这一规模下，该过程产生了可重用的改进，这些改进能够超越其开发环境，推动自动化利用发现朝向生产级成果。四种机制经过筛选形成了 SoL-Pi，涵盖了动作执行、上下文压缩、观察处理和委托阅读。在 51 项任务的 EdgeBench 评估中，SoL-Pi 在 GPT-5.6 Sol 和 Opus 5 上实现了与 Pi 相当的性能，同时将记录的令牌流量减少了 44.7%-49.0%，API 成本降低了约三分之一。换句话说，估计每小时节省为相对于原生 Codex 和 Claude Code 利用的 8.75 美元至 13.50 美元，以及相对于 Pi 的 4.36 美元至 5.71 美元。

View on arXiv Download PDF AI Translation

cs.AI / 80 / 2609.20535

FreqCondNorm: Towards Cross-domain Predictive Maintenance through a Frequency-Conditioned Transformer Foundation Model

FreqCondNorm：通过频率条件变换器基础模型实现跨领域预测性维护

Raounak, Zaynab, LHermine, Camille, Zeng, Zhiguo

Abstract

Deep learning predictive maintenance models suffer from poor transferability across machines and operating conditions, especially when labelled data are scarce and signals span five orders of magnitude in sampling frequency (1 Hz to ~100 kHz). We propose FreqCondNorm, a Transformer-based architecture that introduces a FiLM-style frequency-conditioned normalization layer to unify heterogeneous time-series within a single model. The architecture is pretrained on five public predictive maintenance datasets (CWRU, MFPT, UOC18, PRONOSTIA, CMAPSS) using masked auto-encoding and contrastive learning with balanced domain sampling. On fault diagnosis, the model achieves 99.2% accuracy on CWRU (+6.4 pp over CNN) and 82.1% zero-shot accuracy on MFPT, demonstrating strong transfer across sampling frequencies. However, the approach does not improve remaining useful life prediction, suggesting a mismatch between pretraining and RUL objectives that warrants future investigation.

Chinese Translation

深度学习预测性维护模型在不同机器和操作条件下的可迁移性较差，尤其是在标记数据稀缺且信号采样频率跨度达到五个数量级（1 Hz 到 ~100 kHz）时。我们提出了FreqCondNorm，这是一种基于变换器（Transformer）的架构，引入了一种FiLM风格的频率条件归一化层，以统一单一模型中的异构时间序列。该架构在五个公共预测性维护数据集（CWRU、MFPT、UOC18、PRONOSTIA、CMAPSS）上进行了预训练，采用了掩码自编码和对比学习，并进行了平衡领域采样。在故障诊断方面，该模型在CWRU上达到了99.2%的准确率（比CNN提高了6.4个百分点），在MFPT上达到了82.1%的零-shot准确率，展示了在采样频率之间的强迁移能力。然而，该方法并未改善剩余使用寿命（RUL）预测，表明预训练与RUL目标之间存在不匹配，值得未来进一步研究。

View on arXiv Download PDF AI Translation

cs.AI / 81 / 2609.20538

Refuse, Decompose, Refresh: A Claim-Safe Protocol for Closed-Loop AI Evaluation

拒绝、分解、刷新：一种安全声明的闭环人工智能评估协议

Zhu, Peiying, Chang, Sidi

Abstract

An AI evaluation can be perfectly reproducible and still support the wrong claim. This risk is acute in closed-loop systems: policy determines visited states, observable components, and which failures leave a measurable trace. We propose a claim-safe protocol with three actions. Refuse: abstain when a clean reference stream or matched runtime comparison lacks support. Decompose: report protocol execution, operational false admission, and structural hypotheses separately rather than as one PASS/FAIL label. Refresh: treat distribution-shift alarms as requests to invalidate and recompute a reference map, not as fault evidence. We instantiate the protocol in an aggregate-only simulator with 24 policy components, three demand regimes, two fault-mask families, and independent development and heldout seeds. The preregistered heldout contains 1,440 cases and 21,600 partition rows. Only 55/72 regime-component units were reference-admitted and 54/55 remained runtime-admitted, making abstention part of the result. Stable false admission was 0/20 represented components, with a one-sided exact 95% upper bound of 0.1391 under a frozen 0.20 rule. Within admitted units, affected clean traffic outpredicted nominal fault-cell fraction: across 540 unit-arm rows nested in 20 component clusters, the cell-minus-traffic negative-log-likelihood difference was 0.1264 nats per row, with a 95% component-cluster interval of [0.0593, 0.1918]. A drift log shows why "null" must be reference-relative: clean fault-null streams triggered 15/15, 0/15, and 14/15 alarms across three regimes, while only the middle regime matched the frozen detector reference. Rather than a universal threshold, we contribute an executable contract linking observable support, statistical calibration, and justified claims.

Chinese Translation

人工智能评估可以完美重现，但仍可能支持错误的声明。这种风险在闭环系统中尤为严重：政策决定了访问的状态、可观察的组件，以及哪些故障留下可测量的痕迹。我们提出了一种安全声明的协议，包含三个行动。拒绝：在缺乏支持的情况下，当干净的参考流或匹配的运行时比较不存在时，选择不作答。分解：分别报告协议执行、操作性错误接纳和结构假设，而不是简单地标记为通过/未通过。刷新：将分布偏移警报视为请求无效化和重新计算参考图，而不是故障证据。我们在一个仅聚合的模拟器中实例化了该协议，包含24个政策组件、三种需求模式、两种故障掩码族，以及独立的开发和保留种子。预注册的保留数据包含1,440个案例和21,600个分区行。只有55/72的模式-组件单元被参考接纳，54/55保持运行时接纳，使得不作答成为结果的一部分。稳定的错误接纳为0/20个代表组件，在冻结的0.20规则下，其单侧精确95%上限为0.1391。在被接纳的单元中，受影响的干净流量预测超出了名义故障单元的比例：在20个组件集群中嵌套的540个单元-臂行中，单元减去流量的负对数似然差为每行0.1264 nats，95%组件集群区间为[0.0593, 0.1918]。漂移日志显示了“无效”必须是相对参考的原因：干净的故障无效流在三个模式中触发了15/15、0/15和14/15的警报，而只有中间模式与冻结的探测器参考匹配。我们贡献了一个可执行的合同，链接可观察支持、统计校准和合理声明，而不是一个普遍的阈值。

View on arXiv Download PDF AI Translation

cs.AI / 82 / 2609.20543

Language-model groups overstate consensus when replaying human deliberation on a reasoning task

语言模型组在重现人类推理任务的讨论时夸大了一致性

Shao, Tengfei

Abstract

Full-consensus rates are often treated as indicators of collective cognition, yet depend on how participation and final states are operationalized. We replayed 100 held-out human Wason groups with matched large language model (LLM) agent groups, seeding one belief-anchored agent per participant's pre-discussion answer and scoring agents and people with the same code. Across human scoring definitions, estimates ranged from 24.0% to 57.0%; about one fifth of participants never posted, whereas agents almost always did. Agent groups remained more consensual in two post-unblinding sensitivity analyses: the submit-based comparison (n = 98) yielded gaps of 34.0 and 43.9 percentage points for chat and reasoning modes, and the participation-matched comparison (n = 45) yielded gaps of 34.1 and 44.4 points. These complementary routes reduced different measurement asymmetries yet converged within 0.5 percentage points. The gap persisted without early stopping and under a reparameterization removing the memorizable answer; reasoning-mode groups then agreed nearly unanimously, mostly on incorrect answers. Simulated consensus did not track collective accuracy, and belief-anchored agent groups were biased estimators of the human group-outcome distribution in this setting. These analyses provide a scoring-explicit basis for assessing simulated-group estimates of human deliberative outcomes.

Chinese Translation

完全一致率常被视为集体认知的指标，但其依赖于参与和最终状态的操作化方式。我们重放了100个保留的人类Wason组，并与匹配的大型语言模型（LLM）代理组进行对比，为每位参与者的讨论前答案种植了一个信念锚定的代理，并使用相同的代码对代理和人类进行评分。在人类评分定义中，估计值范围从24.0%到57.0%；约五分之一的参与者从未发言，而代理几乎总是发言。在两项后盲分析中，代理组仍然表现出更高的一致性：基于提交的比较（n = 98）在聊天模式和推理模式下的差距分别为34.0和43.9个百分点，而参与匹配的比较（n = 45）则显示出34.1和44.4的差距。这些互补路径减少了不同测量的不对称性，但在0.5个百分点内收敛。在没有提前停止的情况下，这一差距依然存在，并在去除可记忆答案的重新参数化下保持；此时，推理模式组几乎一致同意，主要是在错误答案上。模拟的一致性并未追踪集体准确性，而信念锚定的代理组在此情境下是人类组结果分布的偏倚估计者。这些分析为评估模拟组对人类讨论结果的估计提供了一个明确的评分基础。

View on arXiv Download PDF AI Translation

cs.AI / 83 / 2609.20581

Limits of Confidence in Diffusion

扩散中的信心极限

Webb, Russ, Shidani, Amitis, Bizeul, Alice, Busbridge, Dan

Abstract

Discrete diffusion, including remasking and uniform-state samplers, generate a sequence by writing multiple token positions per step, drawing each from a per-position distribution and choosing which positions to write from those same distributions. For domains of general interest (pixels, phonemes, or words) there are inherent dependencies between tokens. We show that a step matches the training distribution only when the positions it writes are conditionally independent given the tokens already fixed, that no product of per-position distributions can match a dependent group, and that per-position distributions do not determine whether a group is dependent: two joint distributions can have identical per-position marginals while differing in which combinations of values occur. On ScanAndAdd, a synthetic task whose joint distribution is available in closed form, we verify that every group of two or more undetermined positions a confidence ranking writes is dependent, and measure the generated distribution to be $29\times$ the sampling-noise floor total variation while per-sample metrics are $1.0$.

Chinese Translation

离散扩散，包括重标定和均匀状态采样器，通过在每一步写入多个令牌位置生成一个序列，从每个位置的分布中抽取，并选择从这些相同分布中写入的位置。在一般感兴趣的领域（像素、音素或单词）中，令牌之间存在固有的依赖关系。我们表明，只有在写入的位置在给定已固定令牌的条件下是条件独立时，步骤才与训练分布匹配；没有任何每个位置的分布的乘积能够匹配一个依赖组；而且每个位置的分布并不能决定一个组是否依赖：两个联合分布可以具有相同的每个位置边际分布，但在值的组合上存在差异。在 ScanAndAdd 这个合成任务中，其联合分布以封闭形式可用，我们验证了每个写入的两个或多个未确定位置的组都是依赖的，并测量生成的分布为 $29 imes$ 采样噪声底线总变差，而每个样本的度量为 $1.0$。

View on arXiv Download PDF AI Translation

cs.AI / 84 / 2609.20634

PAA: The Probabilistic Allen Algebra: A Generative and Complete Probabilistic Extension of Allen's Interval Relations

PAA：概率艾伦代数：艾伦区间关系的生成性和完整性概率扩展

Eggert, Julian

Abstract

Allen's interval algebra is a qualitative calculus for temporal relations, but its thirteen base relations are crisp predicates over exact interval boundaries. This is inadequate for temporal information from language, perception, databases, or uncertain histories, where times, durations, and boundaries are uncertain and expressions such as "just before" or "roughly during" have graded meaning. We develop the probabilistic Allen algebra (PAA): a generative and complete extension in which relation probabilities are derived from distributions over interval boundaries rather than assigned as scores. Time points are Gaussian; intervals have Gaussian midpoints and truncated-Gaussian durations. Every relation is a boundary-ordering predicate in one common probability space: point-point relations reduce to error functions, and point-interval and interval-interval relations to multivariate Gaussian orthant probabilities induced by linear inequalities. Contact relations (meets, starts, finishes, equals) receive positive measure through a tolerance band, and under a single tolerance the thirteen relations form a true partition that recovers crisp Allen as the tolerance vanishes. The construction derives Allen's taxonomy rather than positing it: coarse predicates such as precedence, overlap, and containment are unions of leaves whose probabilities are leaf sums, and this hierarchy is preserved as intervals collapse to points and thirteen relations reduce to five and then three. Each relation further decomposes into correlation-aware temporal primitives in the spirit of CIDOC CRM. The algebra is scale-invariant and separates graded expressions such as "shortly before" from contact relations. All results are Monte-Carlo validated and shipped as an open, tested Python package.

Chinese Translation

艾伦的区间代数是一种用于时间关系的定性演算，但其十三个基本关系是关于精确区间边界的清晰谓词。这对于来自语言、感知、数据库或不确定历史的时间信息来说是不够的，因为时间、持续时间和边界都是不确定的，而诸如“恰好在之前”或“大致在期间”等表达具有渐进的意义。我们开发了概率艾伦代数（PAA）：一个生成性和完整性的扩展，其中关系概率是从区间边界的分布中推导出来的，而不是作为分数分配。时间点服从高斯分布；区间具有高斯中点和截断高斯持续时间。每个关系都是一个在一个共同概率空间中的边界排序谓词：点对点关系简化为误差函数，点对区间和区间对区间关系简化为由线性不等式诱导的多元高斯象限概率。接触关系（相遇、开始、结束、相等）通过容忍带获得正测度，在单一容忍下，这十三个关系形成一个真正的划分，当容忍消失时恢复清晰的艾伦代数。该构造推导了艾伦的分类法，而不是假设它：粗略谓词如优先、重叠和包含是叶子的并集，其概率是叶子和，这一层次结构在区间收缩为点时得以保留，十三个关系简化为五个，然后是三个。每个关系进一步分解为考虑相关性的时间原语，符合CIDOC CRM的精神。该代数是尺度不变的，并将诸如“稍早之前”等渐进表达与接触关系区分开来。所有结果都经过蒙特卡洛验证，并作为一个开放的、经过测试的Python包发布。

View on arXiv Download PDF AI Translation

cs.AI / 85 / 2609.20658

Ownership in AI-Assisted Everyday Tasks

人工智能辅助日常任务中的所有权

Wei, Megan, Subbiah, Melanie, Lee, Audrey, Dahmani, Annya, Edwards, Dave, Edwards, Helen, Pavlick, Ellie

Abstract

When does work done with AI still feel like ours? As AI becomes woven into everyday tasks, we must examine what happens to our sense of ownership and contribution when a machine shares in producing what we make. We report an exploratory qualitative survey in which participants were asked to describe two recent, self-selected tasks completed with AI: one that felt like their own and one that did not. We find that felt ownership depends on the process of collaboration: people disown work when they merely approve AI's suggestions, but retain ownership when they lead, iterate, or rewrite. Ownership can also extend to settings where people own the vision for a project but not the execution; respondents reported high ownership on tasks they could not have completed without AI. Loss of personal voice and a lack of comprehension of the output both erode ownership. Finally, willingness to disclose AI use is often decoupled from actual pride or ownership, and instead shaped by community norms and fear of credit erasure. We propose several research directions as a result of these findings to promote AI development that supports people's sense of authorship over their own lives.

Chinese Translation

在与人工智能（AI）共同完成的工作中，何时仍然感到这是我们的工作？随着人工智能逐渐融入日常任务，我们必须审视当机器参与我们所创造的成果时，我们的所有权和贡献感会发生什么变化。我们报告了一项探索性的定性调查，参与者被要求描述两个最近自选的与人工智能完成的任务：一个让他们感到是自己的，另一个则不是。我们的研究发现，感知的所有权依赖于协作的过程：当人们仅仅批准人工智能的建议时，他们会放弃对工作的所有权，而当他们主导、迭代或重写时，则会保留所有权。所有权还可以扩展到人们拥有项目愿景但不负责执行的情境；受访者报告称，在他们无法在没有人工智能的情况下完成的任务中感受到较高的所有权。个人声音的丧失和对输出缺乏理解都会侵蚀所有权。最后，愿意披露人工智能的使用往往与实际的自豪感或所有权脱钩，而是受到社区规范和对信用消失的恐惧的影响。基于这些发现，我们提出了若干研究方向，以促进支持人们对自己生活的创作权感的人工智能发展。

View on arXiv Download PDF AI Translation

cs.AI / 86 / 2609.20722

Deep Noir: Autonomous Steering Discovery via Architectural Chronometry in Transformer Models

深度黑暗：通过变换器模型中的架构计时法发现自主驾驶

Bobe III, Frank E., Vetaw, Gregory D., Bryner, Darshan W., Cook, Matthew G., Salas-Vernis, Jose L.

Abstract

Activation steering modifies LLM behavior at inference time, but identifying where and how strongly to steer remains manual. We introduce Deep Noir, a framework that uses Logit Lens convergence and causal head-level attribution to autonomously discover optimal steering parameters. Across three scales (1B x 3, 2-3B x 2, and 7-9B x 4), our engine achieves 16.7 percentage-point improvement on spam at 1B (standard deviation 4.7; 39 runs), with gains increasing to 21 to 42 percentage points at 7-9B across four architectures. On SST-2 sentiment, it achieves a 13.1 percentage-point improvement with zero code changes. Mechanistic grounding enables automated discovery of intervention points that generalize across tasks and architectures. On sentiment, RepE without head masking fails to improve over baseline, while Deep Noir improves all models (p less than 0.01). We further show that steering creates a predictable prompt-injection attack surface whose vulnerability increases monotonically with steering magnitude. This finding is relevant to agent systems deploying steered classifiers.

Chinese Translation

激活引导在推理时修改大型语言模型（LLM）的行为，但识别引导的具体位置和强度仍然是手动的。我们提出了Deep Noir，一个利用Logit Lens收敛和因果头级归因的框架，能够自主发现最佳引导参数。在三个规模（1B x 3、2-3B x 2和7-9B x 4）上，我们的引擎在1B上实现了16.7个百分点的垃圾邮件改善（标准差4.7；39次运行），在7-9B的四种架构上，提升幅度增加到21至42个百分点。在SST-2情感分析中，它实现了13.1个百分点的改善，且无需任何代码更改。机制基础使得能够自动发现跨任务和架构的干预点。在情感分析中，未进行头部掩蔽的RepE未能超越基线，而Deep Noir则改善了所有模型（p小于0.01）。我们进一步显示，引导创造了一个可预测的提示注入攻击面，其脆弱性随着引导幅度的增加而单调上升。此发现与部署引导分类器的代理系统相关。

View on arXiv Download PDF AI Translation

cs.AI / 87 / 2609.20732

Q&A on Any Spreadsheet Requires Interpreting Its Grid Structure

任何电子表格上的问答需要解释其网格结构

Smoleń, Zofia

Abstract

Semantic cell annotation improves chunking interpretability for spreadsheets in LLM-driven RAG systems, aiding answer generation through enriched context rather than improved retrieval accuracy. We propose a novel framework of splitting any spreadsheet into interpretable chunks using cell role annotation. Our framework beats the state of the art, yet it faces a hard ceiling. Spreadsheets are fundamentally two-dimensional unstructured data with continuous relationships and infinite potential cell roles. Because classification models are restricted to finite, pre-defined classes, they cannot perfectly capture this structural nuance, even with human-level annotation. We show that addressing the spreadsheet-to-LLM bottleneck requires moving beyond discrete cell classification. Instead, the field must develop dimensionality-reduction techniques to directly flatten 2D unstructured spreadsheets into 1D unstructured text. Text chunks would be easier for downstream RAG to interpret and generate from.

Chinese Translation

语义单元注释提高了在大型语言模型驱动的检索增强生成（RAG）系统中电子表格的分块可解释性，通过丰富的上下文而非提高检索准确性来辅助答案生成。我们提出了一种新颖的框架，通过单元角色注释将任何电子表格拆分为可解释的块。我们的框架超越了现有技术，但仍面临严峻的瓶颈。电子表格本质上是具有连续关系和无限潜在单元角色的二维非结构化数据。由于分类模型受限于有限的预定义类别，它们无法完美捕捉这种结构细微差别，即使是经过人类级别的注释。我们表明，解决电子表格与大型语言模型之间的瓶颈需要超越离散单元分类。相反，该领域必须开发降维技术，以直接将二维非结构化电子表格展平为一维非结构化文本。文本块将更容易被下游的RAG系统解释和生成。

View on arXiv Download PDF AI Translation

cs.AI / 88 / 2609.20754

RAFT: A Stateful Retrieval-Augmented Framework for Troubleshooting Agents

RAFT：一种用于故障排除代理的有状态检索增强框架

Zhang, Mingxuan, Wang, Xiaowen, Sharan, Anupma, Chen, Zhengyi, Zhang, Chenyu Diana, Yang, Shanshan, Pacharu, Chittibabu

Abstract

Effective troubleshooting agents in enterprise customer support depend on retrieving actionable guidance from similar historical cases, yet existing retrieval-augmented generation (RAG) systems treat support cases as static documents and overlook their multi-stage, stateful nature. We introduce RAFT (Retrieval-Augmented Framework for Troubleshooting Agents), a stateful RAG framework that abstracts each closed historical case into a directed chain of timeline entries and retrieves at the entry level, surfacing cases whose intermediate states match the active case and returning the parent-case trajectory anchored at the matched state; an optional case-level graph links cases through a configurable similarity representation. We evaluate this retrieval layer directly, which, unlike evaluating a full agent system, requires no production deployment. Because public multi-stage troubleshooting data is extremely rare, we pair a synthetic benchmark built from Microsoft Learn Windows Server documentation with real Apache Jira issues carrying human-created duplicate labels. RAFT improves Case Hit over vanilla RAG and GraphRAG baselines at every stage of case progress, with statistically significant gains over the strongest baseline; the Jira results provide directional evidence that the advantage transfers to real case histories. We release our benchmark, implementation, and the Apache Jira evaluation set.

Chinese Translation

企业客户支持中的有效故障排除代理依赖于从类似历史案例中检索可操作的指导，然而现有的检索增强生成（RAG）系统将支持案例视为静态文档，忽视了其多阶段和有状态的特性。我们提出了RAFT（故障排除代理的检索增强框架），这是一个有状态的RAG框架，将每个关闭的历史案例抽象为一个有向的时间线条目链，并在条目级别进行检索，呈现出与当前案例的中间状态匹配的案例，并返回锚定在匹配状态的父案例轨迹；一个可选的案例级图通过可配置的相似性表示将案例连接起来。我们直接评估这一检索层，与评估完整的代理系统不同，这不需要生产部署。由于公共的多阶段故障排除数据极为稀缺，我们将基于Microsoft Learn Windows Server文档构建的合成基准与携带人工创建重复标签的真实Apache Jira问题配对。RAFT在每个案例进展阶段的案例命中率上优于基础的RAG和GraphRAG基线，并在最强基线之上取得了统计显著的提升；Jira结果提供了方向性证据，表明这一优势可以转移到真实案例历史中。我们发布了我们的基准、实现和Apache Jira评估集。

View on arXiv Download PDF AI Translation

cs.AI / 89 / 2609.20804

An Empirical Study of Harness Design for Coding Agents

编码代理的马具设计实证研究

Fan, Run-Ze, Zhang, Zihao, Ma, Simin, Hu, Yebowen, Wang, Shouju, Song, Kaiqiang, Liu, Fei, Zamani, Hamed, Wang, Xiaoyang

Abstract

Coding harnesses shape how autonomous coding agents translate model capabilities into long-horizon software-engineering performance, yet existing work typically evaluates harnesses as monolithic systems, leaving the effectiveness of individual components unclear. To enable component-level comparisons, we study this question with a lightweight coding harness whose execution loop is fixed while three components are varied: planning, action space, and context management. Across four models evaluated on SWE-Bench Verified and Terminal-Bench 2.1, we evaluate 176 matched settings spanning five context-management strategies, four context-window budgets, and targeted ablations of planning and action space. We find that: (1) Context management becomes increasingly valuable as the context-window budget tightens, with most of its benefit coming from preventing context-overflow failures. (2) Staging rule-based elision before LLM-based summarization provides the strongest overall efficiency among the context-management strategies, whereas making elided content recoverable adds machinery that models rarely use and yields no accuracy gain. (3) Planning shifts from an accuracy scaffold for weaker models to a cost saver for stronger models, with little change in accuracy. (4) Predefined tools improve performance for models with weaker bash proficiency, whereas bash-capable models can operate effectively with a bash-only interface and achieve substantially lower cost, especially on command-line-centric tasks. Trajectory-level analysis explains these effects: context management extends execution trajectories without substantially altering agent behavior, planning changes where trajectories stop, and the action space changes the granularity at which code is written. These findings inform model- and budget-aware harness design and provide a modular framework for evaluating future harness components.

Chinese Translation

编码马具决定了自主编码代理如何将模型能力转化为长期软件工程性能，然而现有研究通常将马具视为整体系统进行评估，导致个别组件的有效性不明确。为了实现组件级的比较，我们研究了一个轻量级编码马具，其执行循环固定，同时对三个组件进行了变化：规划、动作空间和上下文管理。在对SWE-Bench Verified和Terminal-Bench 2.1评估的四个模型中，我们评估了176个匹配设置，涵盖五种上下文管理策略、四种上下文窗口预算以及针对规划和动作空间的有针对性的消融实验。我们的研究发现：(1) 随着上下文窗口预算的收紧，上下文管理变得越来越重要，其大部分好处来自于防止上下文溢出失败。(2) 在基于LLM的摘要之前进行基于规则的省略提供了上下文管理策略中最强的整体效率，而使省略的内容可恢复则增加了模型很少使用的机制，并未带来准确性的提升。(3) 规划从对较弱模型的准确性支撑转变为对较强模型的成本节约，准确性变化不大。(4) 预定义工具提高了对bash能力较弱模型的性能，而具备bash能力的模型可以有效地使用仅有bash接口，并在以命令行为中心的任务中实现显著降低的成本。轨迹级分析解释了这些效应：上下文管理在不显著改变代理行为的情况下延长执行轨迹，规划改变了轨迹停止的地方，而动作空间改变了代码编写的粒度。这些发现为模型和预算感知的马具设计提供了指导，并为评估未来马具组件提供了模块化框架。

View on arXiv Download PDF AI Translation

计算语言学 (Computation and Language)

cs.CL / 1 / 2609.19148

Modality Discrepancy Transformer for Ambivalence and Hesitancy Recognition

用于矛盾和犹豫识别的模态差异变换器

Luo, Shiyu, Wang, Yu, Huang, Jiawen, Xiao, Zhaoxiang, Huang, Chenxi, Zhang, Qi, Liu, Bin

Abstract

Ambivalence and hesitancy (A/H) are affective states in which individuals express contradictory signals across facial, vocal, and linguistic channels. Automatically recognising A/H in clinical videos requires detecting cross-modal disagreement -- the signal that standard fusion methods suppress. Based on the conflict-aware multimodal fusion framework of Bekhouche et al., we present the Modality Discrepancy Transformer (MDT). MDT enriches the original 6-token design to a 9-token representation comprising three modality embeddings, three absolute-difference features, and three Hadamard-product discrepancy features learned through linear projections. These nine tokens undergo Transformer self-attention, with FiLM-based text-conditioned modulation and LoRA fine-tuning as core architectural components. A text-guided late fusion branch blends a text-only auxiliary head with the full multimodal output at inference. On the BAH dataset from the 3rd ABAW Challenge, MDT achieves 0.7408 Macro F1 on the labelled test split and 0.7368 on the private leaderboard, outperforming the strongest published baseline by over 10 points while training in under 20 minutes on a single GPU.

Chinese Translation

矛盾和犹豫（A/H）是个体在面部、声音和语言通道中表达矛盾信号的情感状态。在临床视频中自动识别A/H需要检测跨模态的不一致性——这一信号是标准融合方法所抑制的。基于Bekhouche等人的冲突感知多模态融合框架，我们提出了模态差异变换器（MDT）。MDT将原始的6-token设计扩展为一个由三个模态嵌入、三个绝对差异特征和三个通过线性投影学习的Hadamard积差异特征组成的9-token表示。这九个token经过变换器自注意力处理，FiLM（Feature-wise Linear Modulation）基础的文本条件调制和LoRA（Low-Rank Adaptation）微调作为核心架构组件。一个文本引导的后期融合分支在推理时将仅文本的辅助头与完整的多模态输出相结合。在第三届ABA（Affective Behavior Analysis in the Wild）挑战赛的BAH数据集上，MDT在标记测试集上达到了0.7408的宏F1值，在私人排行榜上达到了0.7368，超过了最强已发布基线10多个点，同时在单个GPU上训练时间不足20分钟。

View on arXiv Download PDF AI Translation

cs.CL / 2 / 2609.19149

Subliminal Prompting Beyond Static Geometry: Causal Depth and Multi-Token Confounds

超越静态几何的潜意识提示：因果深度与多标记混淆

Velmurugan, Barath

Abstract

Subliminal learning shows that language models can transmit a hidden trait through outputs that appear unrelated to it. One proposed explanation, token entanglement, links animal and number tokens through the model's output vocabulary. Yet existing measurements answer different questions: whether outputs co-vary, fixed output vectors align, an answer can be read from a hidden state, or that state causally controls the answer. We measure each separately in a fixed animal-number prompting protocol. From Llama-3.1-8B to 70B, fixed output-vector similarity predicts behavior less well: the paired mean correlation change is -0.080 (95% CI [-0.127, -0.035]). A fixed output-head readout shows no resolved change in normalized depth AUC. To test control, we copy the temporary answer-position state from one number prompt into another at five depths and measure which prompt the final animal score follows. Donor-control AUC rises from 0.254 to 0.540, a paired change of +0.286 (95% CI [+0.272, +0.300]), with increases for all 18 concepts. The contrast remains with exactly eight transformer blocks remaining, while specificity and identity controls remain small or exact. In two Qwen models, scoring every digit in sequence does not recover the positive one-token association. Per-token averaging instead creates a positive pooled association that disappears after controlling number width, revealing a length confound. Thus, fixed geometry, observational readability, causal timing, and multi-token measurement are distinct properties of this frozen prompting channel. They constrain token-level explanations but do not identify the mechanism of training-time trait transfer.

Chinese Translation

潜意识学习表明，语言模型可以通过看似无关的输出传递隐藏特征。一种提出的解释是标记纠缠，它通过模型的输出词汇将动物标记和数字标记联系起来。然而，现有的测量方法回答了不同的问题：输出是否共同变化、固定输出向量是否对齐、是否可以从隐藏状态中读取答案，或该状态是否因果控制答案。我们在固定的动物-数字提示协议中分别测量每个因素。从 Llama-3.1-8B 到 70B，固定输出向量的相似性对行为的预测效果较差：配对均值相关变化为 -0.080（95% CI [-0.127, -0.035]）。固定输出头的读出在归一化深度 AUC 中没有显示出显著变化。为了测试控制，我们将一个数字提示中的临时答案位置状态复制到另一个提示中，深度为五，并测量最终动物得分遵循哪个提示。捐赠者控制 AUC 从 0.254 上升到 0.540，配对变化为 +0.286（95% CI [+0.272, +0.300]），所有 18 个概念均有增加。对比在剩余正好八个变换块时依然存在，而特异性和身份控制保持较小或完全一致。在两个 Qwen 模型中，按顺序对每个数字评分并未恢复正的单标记关联。每标记平均反而创造了一个正的汇总关联，但在控制数字宽度后消失，揭示了长度混淆。因此，固定几何、观察可读性、因果时序和多标记测量是这一冻结提示通道的不同特性。它们限制了标记级别的解释，但并未识别训练时间特征转移的机制。

View on arXiv Download PDF AI Translation

cs.CL / 3 / 2609.19150

Sampling Reveals Style: Unsupervised, Training-Free Discovery of Prompt-Conditional Stylistic Axes in LLM Activations

采样揭示风格：无监督、无训练的提示条件下大语言模型激活中的风格轴发现

Mallavarapu, Ajit, Gu, Ziwei

Abstract

Large language models (LLMs) encode rich stylistic structure in their hidden activations, but discovering which stylistic dimensions are salient for a given prompt typically requires supervised contrastive data. We present a training-free, prompt-conditional alternative: we repeatedly sample completions of a single prompt at elevated temperature, apply Principal Component Analysis (PCA) to the pooled hidden activations, and label the resulting axes automatically from the pole generations. We validate the discovered axes against 245 human-elicited stylistic annotations in a two-phase study. On our strongest model (Qwen-3.5-4B-Instruct), the top two axes match spontaneously requested human dimensions with 72.8% precision and 43.6% macro-recall, and 75.6% of validity ratings judge the axes' polar generations accurate to their labels, with 90.9% adjacent inter-annotator agreement. Discoverability is strongly model-dependent: both Qwen models and Llama-3.2-3B expose human-salient axes, while DeepSeek-7B-Chat drops to 35.3% precision, its leading components dominated by structural rather than stylistic variance. Simple PCA over a model's own decoding variance is thus an effective, low-cost probe of stylistic structure in LLM representations, one that also exposes sharp cross-model differences in how that structure is organized.

Chinese Translation

大型语言模型（LLMs）在其隐藏激活中编码了丰富的风格结构，但发现特定提示的显著风格维度通常需要监督对比数据。我们提出了一种无训练、提示条件的替代方法：我们在高温下反复采样单一提示的完成结果，对汇总的隐藏激活应用主成分分析（PCA），并自动从极性生成中标记结果轴。我们在一个两阶段研究中验证了发现的轴与245个人工引导的风格注释的匹配情况。在我们最强的模型（Qwen-3.5-4B-Instruct）上，前两个轴与自发请求的人类维度匹配的精度为72.8%，宏观召回率为43.6%，75.6%的有效性评分认为这些轴的极性生成与其标签一致，且相邻注释者之间的协议率为90.9%。可发现性强烈依赖于模型：Qwen模型和Llama-3.2-3B都揭示了人类显著的轴，而DeepSeek-7B-Chat的精度降至35.3%，其主要成分受结构而非风格变异的主导。因此，对模型自身解码变异的简单PCA是有效且低成本的风格结构探测方法，同时也揭示了不同模型在该结构组织方式上的显著差异。

View on arXiv Download PDF AI Translation

cs.CL / 4 / 2609.19151

What Users Think of Generative AI: A Cross-Platform NLP Analysis of Trust and Friction in App Store Reviews

用户对生成性人工智能的看法：应用商店评论中信任与摩擦的跨平台自然语言处理分析

Hossain, Md Jafrin, Jahan, Umme Nusrat, Shammo, Shouvaggo Sharif

Abstract

Generative AI (GenAI) applications have achieved rapid consumer adoption, yet little large-scale research examines user-perceived quality, trust, and adoption barriers. We present one of the first cross-application analyses of app store reviews for six major GenAI applications (ChatGPT, Gemini, Microsoft Copilot, Claude, DeepSeek, and Perplexity), comprising 17,012 English-language reviews from Google Play and the Apple App Store. We combine BERTopic topic modeling with RoBERTa sentiment classification and evaluate cross-application differences using chi-square, Kruskal-Wallis, and multinomial logistic regression with Bonferroni correction. Both components are validated against human coding using a stratified sample of 300 reviews. Results show that negative sentiment concentrates in advertising (91%), authentication (89%), server reliability (83%), and subscription pricing (73%). Sentiment differs significantly across applications, with Claude exhibiting the highest negative sentiment (47.7%) alongside a strongly enthusiastic user base, indicating statistically significant polarization. These findings are robust despite unequal review counts across applications. As exploratory observations, a subset of DeepSeek reviews raised geopolitical and data privacy concerns related to its Chinese origin, while a proposed Trust Friction Score summarizes application-specific trust and usability barriers into interpretable dimensions. The study provides validated and actionable evidence on user trust, usability, and adoption barriers in consumer generative AI applications.

Chinese Translation

生成性人工智能（GenAI）应用已迅速被消费者接受，但关于用户感知质量、信任和采用障碍的大规模研究仍然较少。我们呈现了对六个主要GenAI应用（ChatGPT、Gemini、Microsoft Copilot、Claude、DeepSeek和Perplexity）应用商店评论的首次跨应用分析，涵盖来自Google Play和Apple App Store的17,012条英语评论。我们结合了BERTopic主题建模与RoBERTa情感分类，并使用卡方检验、Kruskal-Wallis检验和带Bonferroni校正的多项式逻辑回归评估跨应用差异。这两个组件通过对300条评论的分层样本进行人工编码验证。结果显示，负面情感主要集中在广告（91%）、身份验证（89%）、服务器可靠性（83%）和订阅定价（73%）等方面。不同应用之间的情感差异显著，Claude的负面情感最高（47.7%），同时用户基础表现出强烈的热情，显示出统计显著的极化现象。这些发现尽管各应用评论数量不均，但依然稳健。作为探索性观察，部分DeepSeek评论提出了与其中国来源相关的地缘政治和数据隐私问题，而提出的信任摩擦评分（Trust Friction Score）则将应用特定的信任和可用性障碍总结为可解释的维度。本研究提供了关于消费者生成性人工智能应用中用户信任、可用性和采用障碍的验证和可操作的证据。

View on arXiv Download PDF AI Translation

cs.CL / 5 / 2609.19152

FakeSpotter: A content and strategy agnostic Viral Misinformation Detection Tool

FakeSpotter：一种与内容和策略无关的病毒性虚假信息检测工具

Spitale, Giovanni, Germani, Federico

Abstract

Misinformation detection tools often rely on binary true and false classifications or models trained on historical examples, limiting their usefulness when novel misleading narratives emerge. Here, we present FakeSpotter, a content- and strategy-agnostic tool designed to estimate the viral misinformation risk of textual content by measuring structural fingerprints of misinformation rather than directly adjudicating truthfulness. FakeSpotter operationalizes a theory-driven framework across linguistic, narrative, logical, and critical-thinking dimensions, using repeated LLM assessments and domain-specific logistic regression classifiers for short and long texts. In a labelled corpus of 764 texts from social media and FakeNewsNet, FakeSpotter achieved macro F1 scores of 0.788 for short texts and 0.793 for long texts on a held-out test set. FakeSpotter's interpretive layer provides explainable outputs through feature-based scores, signal agreement, and a caution index, and can be used for social listening. These findings suggest that identifying the structural fingerprints of misinformation can support early, explainable, and human-supervised assessment of potentially viral misinformation.

Chinese Translation

虚假信息检测工具通常依赖于二元的真与假分类或基于历史示例训练的模型，这在新出现的误导性叙事出现时限制了它们的有效性。在此，我们提出了FakeSpotter，这是一种与内容和策略无关的工具，旨在通过测量虚假信息的结构指纹来评估文本内容的病毒性虚假信息风险，而不是直接判断其真实性。FakeSpotter在语言、叙事、逻辑和批判性思维维度上实施了一个基于理论的框架，使用重复的LLM评估和领域特定的逻辑回归分类器来处理短文本和长文本。在来自社交媒体和FakeNewsNet的764个标注文本的语料库中，FakeSpotter在保留的测试集上实现了短文本的宏F1分数为0.788，长文本的宏F1分数为0.793。FakeSpotter的解释层通过基于特征的分数、信号一致性和警示指数提供可解释的输出，并可用于社交聆听。这些发现表明，识别虚假信息的结构指纹可以支持对潜在病毒性虚假信息的早期、可解释和人类监督的评估。

View on arXiv Download PDF AI Translation

cs.CL / 6 / 2609.19153

Stop Removing Stopwords: How an Inherited Preprocessing Default Distorts Legal Text-as-Data

停止去除停用词：继承的预处理默认设置如何扭曲法律文本作为数据

Dickinson, Gregory M.

Abstract

Empirical legal scholarship increasingly treats judicial text as data, and much of it still runs on sparse, interpretable pipelines -- TF-IDF features and linear classifiers -- because the textual feature is often the object of study, not merely a means to a prediction. Yet these pipelines inherit a chain of preprocessing defaults from mid-century information retrieval that were never validated against classification accuracy, the most entrenched being stopword removal. This study introduces an exhaustive single-word ablation that measures a preprocessing step's effect directly against the downstream objective, and applies it to stopword removal as the hardest case to dislodge. Matching Supreme Court Database labels to Caselaw Access Project opinion texts, it examines two binary tasks that bracket F1 headroom, ideological direction (no-removal baseline F1 ~ 0.68) and constitutional versus non-constitutional law type (~ 0.92), across 7,668 and 7,001 opinions. For each task the analysis approximates the best stoplist any expert could build, removing each of roughly 18,500 candidate words and measuring the effect directly. Three findings follow: generic stoplists in common use fall below the no-removal baseline in every test; even optimized stoplists are statistically indistinguishable from removing nothing; and meta-models trained on word-level features cannot predict which removals help, so list curation has nothing to target. The method generalizes to any inherited preprocessing default, and the result is a caution specific to interpretable legal text-as-data: a step that silently reshapes which features a model sees can distort the very doctrinal and ideological signal such research exists to recover. Leaving stopwords in place is a question of measurement validity.

Chinese Translation

实证法律学越来越多地将司法文本视为数据，而其中许多仍然依赖于稀疏、可解释的处理流程——TF-IDF特征和线性分类器——因为文本特征通常是研究对象，而不仅仅是预测的手段。然而，这些处理流程继承了20世纪中期信息检索中的一系列预处理默认设置，这些设置从未经过分类准确性验证，其中最根深蒂固的就是去除停用词。本研究引入了一种详尽的单词消融方法，直接测量预处理步骤对下游目标的影响，并将其应用于去除停用词这一最难以动摇的案例。通过将最高法院数据库标签与案例法访问项目的意见文本进行匹配，研究考察了两个二元任务，这些任务在F1头部空间内，意识形态方向（无去除基线F1约为0.68）和宪法与非宪法法律类型（约为0.92），涵盖了7,668和7,001条意见。对于每个任务，分析近似任何专家可以构建的最佳停用词表，逐一去除大约18,500个候选词并直接测量其效果。得出了三个发现：在每个测试中，常用的通用停用词表均低于无去除基线；即使是优化后的停用词表在统计上也与不去除任何词无显著区别；而且在词级特征上训练的元模型无法预测哪些去除是有帮助的，因此列表策划没有明确的目标。该方法可以推广到任何继承的预处理默认设置，结果对可解释的法律文本作为数据提出了特定的警示：一个默默重塑模型所见特征的步骤可能会扭曲该研究存在的根本教义和意识形态信号。保留停用词是一个测量有效性的问题。

View on arXiv Download PDF AI Translation

cs.CL / 7 / 2609.19154

Neo-Classic: A Benchmark for Evaluating Linguistic-Aesthetic Reasoning in Classical Chinese Poetry

Neo-Classic：评估古典汉诗中语言美学推理的基准

Zhang, Han, Gu, Zihan, Wang, Zhiyuan, Ma, Tianyi, Lu, Jiacheng, Zhang, Xinyan, Wei, Yuhao, Hua, Cheng

Abstract

While Large Language Models (LLMs) achieve high accuracy on established Classical Chinese Poetry benchmarks, it remains challenging to distinguish transferable Linguistic-Aesthetic Reasoning from reliance on familiar pre-training patterns. To address this issue, we introduce Neo-Classic, an evaluation benchmark that combines a constructionist Out-of-Sample (OOS) dataset with a suite of reverse understanding probes. Unlike traditional benchmarks that rely on verification or generation over historical corpora, Neo-Classic comprises strictly metrical poetry authored by contemporary experts, reducing the possibility of direct retrieval. We evaluate state-of-the-art models, including Qwen3-Max, Gemini-3-Pro, and DeepSeek-V3.2, across five behavioral probes designed to test hierarchical constraint satisfaction. Our results reveal two primary limitations. First, a performance gap of 20 to 50 percent emerges when models transition from historical to contemporary texts. Second, models exhibit substantial difficulties in discourse-level ordering tasks, with standard accuracy remaining low (0 to 13 percent). Although expert-level guidance improves the performance of reasoning-enhanced models to 36 percent, a notable gap with human experts persists. These findings suggest that while current LLMs capture local formal patterns, they struggle with global hierarchical planning required for robust Linguistic-Aesthetic Reasoning.

Chinese Translation

尽管大型语言模型（LLMs）在既定的古典汉诗基准上取得了高准确率，但区分可转移的语言美学推理与对熟悉的预训练模式的依赖仍然具有挑战性。为了解决这一问题，我们引入了Neo-Classic，一个评估基准，它结合了建设性的样本外（OOS）数据集和一套逆向理解探测工具。与依赖于历史语料库的验证或生成的传统基准不同，Neo-Classic由当代专家创作的严格韵律诗组成，减少了直接检索的可能性。我们评估了包括Qwen3-Max、Gemini-3-Pro和DeepSeek-V3.2在内的最先进模型，在五个旨在测试层级约束满足的行为探测工具上进行评估。我们的结果揭示了两个主要局限性。首先，当模型从历史文本转向当代文本时，出现了20%到50%的性能差距。其次，模型在话语层次的排序任务中表现出显著困难，标准准确率保持在低水平（0%到13%）。尽管专家级指导将增强推理模型的性能提升至36%，但与人类专家之间仍存在显著差距。这些发现表明，尽管当前的LLMs能够捕捉局部的形式模式，但在进行强健的语言美学推理所需的全球层级规划方面仍然存在困难。

View on arXiv Download PDF AI Translation

cs.CL / 8 / 2609.19155

Towards Proactive Detection of User-Side Implicit Conflicts in Human-LLM Dialogue

面向人类-LLM对话中用户侧隐性冲突的主动检测

Wang, Jinqiang, Zhu, Tao, Ning, Huansheng

Abstract

In Human-LLM dialogue, follow-up user utterances may implicitly conflict with earlier intents, leading the LLM to misinterpret user needs and generate inappropriate responses. A reliable dialogue system should proactively detect user-side conflicts before generating a response and seek clarification when necessary. However, prior work has largely focused on LLM-side conflicts, leaving user-side conflicts underexplored. To fill this gap, we construct UC-Bench, a human-annotated benchmark for evaluating user-side conflict detection. Preliminary experiments show that existing LLMs struggle with this task, especially when conflicts arise from implicit incompatibilities grounded in dialogue history. To improve lightweight LLMs with limited training data, we investigate data synthesis for user-side conflict detection. Existing synthesis methods do not explicitly model the implicit incompatibilities between historical and current user utterances, making it difficult to capture the evolution of conflicts and to generate reliably labeled implicit conflict samples. We propose SynUC, a constraint-guided synthesis method that represents user-side conflicts in a constraint space and uses the SPEAKING framework to guide traceable constraint transformations. Applying SynUC to WildChat, we construct UC-Data, a user-side conflict training set containing 2,487 samples. On UC-Bench, Qwen3.5-4B trained on UC-Data outperforms larger general-purpose LLMs such as Claude Opus 4.8, as well as the same backbone trained on data synthesized by existing methods.

Chinese Translation

在人类-LLM对话中，后续用户发言可能与早期意图隐性冲突，这会导致LLM误解用户需求并生成不当回应。一个可靠的对话系统应在生成回应之前主动检测用户侧冲突，并在必要时寻求澄清。然而，之前的研究主要集中在LLM侧冲突上，用户侧冲突则未得到充分探讨。为填补这一空白，我们构建了UC-Bench，这是一个用于评估用户侧冲突检测的人类标注基准。初步实验表明，现有的LLM在这一任务上表现不佳，尤其是在冲突源于对话历史中隐性不兼容时。为了改善训练数据有限的轻量级LLM，我们研究了用户侧冲突检测的数据合成。现有的合成方法并未明确建模历史和当前用户发言之间的隐性不兼容性，这使得捕捉冲突的演变和生成可靠标记的隐性冲突样本变得困难。我们提出了SynUC，一种约束引导的合成方法，该方法在约束空间中表示用户侧冲突，并利用SPEAKING框架指导可追溯的约束变换。将SynUC应用于WildChat，我们构建了UC-Data，这是一个包含2,487个样本的用户侧冲突训练集。在UC-Bench上，基于UC-Data训练的Qwen3.5-4B超越了Claude Opus 4.8等更大的一般性LLM，以及基于现有方法合成数据训练的相同骨干网络。

View on arXiv Download PDF AI Translation

cs.CL / 9 / 2609.19156

Reflective Recovery: A Self-Supervised Method for Reasoning by Learning from Mistakes

反思性恢复：一种通过学习错误进行推理的自监督方法

Chen, Qirui, Pi, Renjie, Gao, Jiahui, Kong, Lingpeng

Abstract

Data-driven fine-tuning is widely adopted to enhance reasoning in Large Language Models (LLMs) due to its simplicity and efficiency. However, mainstream imitation learning methods that rely exclusively on perfect reasoning trajectories suffer from a Scaling Collapse: when the problem set is limited, increasing positive examples fails to yield continuous improvement. However, during inference, an LLM can not guarantee that every intermediate step is correct and is therefore prone to errors. Once such errors arise, the LLM often struggles to recover and may be further misled by the accumulation of previous mistakes. To address this, we propose Reflective Recovery, a simple yet effective self-supervised approach that transforms failed reasoning attempts into recovery training data. Specifically, we extract initial segments of failed trajectories, concatenate them with prompts, and use them to guide the LLM toward valid solutions. Because these segments from failed trajectories are likely to contain errors, this process teaches models to recognize and correct mistakes during reasoning, enabling recovery from erroneous states without relying on external critics or reward models. Evaluated on extensive benchmarks, Reflective Recovery significantly improves performance. On DeepSeek-R1-Distill-Qwen-7B, it boosts accuracy from 30.0% to 37.5% on AIME 2025 and from 37.6% to 47.8% on Minerva. More importantly, analyses demonstrate that it breaks the scaling collapse barrier and enables models to develop emergent self-correction behaviors, representing a paradigm shift from outcome-oriented memorization to process-oriented reflective reasoning.

Chinese Translation

数据驱动的微调因其简单性和高效性而被广泛应用于增强大型语言模型（LLMs）的推理能力。然而，主流的模仿学习方法完全依赖于完美的推理轨迹，面临着规模崩溃的问题：当问题集有限时，增加正例并不能带来持续的改进。然而，在推理过程中，LLM无法保证每个中间步骤都是正确的，因此容易出现错误。一旦出现此类错误，LLM往往难以恢复，并可能因之前错误的累积而进一步受到误导。为了解决这个问题，我们提出了反思性恢复（Reflective Recovery），这是一种简单而有效的自监督方法，将失败的推理尝试转化为恢复训练数据。具体而言，我们提取失败轨迹的初始片段，将其与提示连接，并用它们引导LLM朝向有效解决方案。由于这些来自失败轨迹的片段可能包含错误，这一过程教会模型在推理过程中识别和纠正错误，使其能够在不依赖外部评估者或奖励模型的情况下从错误状态中恢复。经过广泛基准测试评估，反思性恢复显著提高了性能。在DeepSeek-R1-Distill-Qwen-7B上，它在AIME 2025上的准确率从30.0%提升至37.5%，在Minerva上的准确率从37.6%提升至47.8%。更重要的是，分析表明它突破了规模崩溃的障碍，使模型能够发展出突现的自我纠正行为，代表了从以结果为导向的记忆到以过程为导向的反思性推理的范式转变。

View on arXiv Download PDF AI Translation

cs.CL / 10 / 2609.19158

VisKG-LM: Compiling Knowledge Graphs into Visual Memory for Multiple-Choice Question Answering

VisKG-LM：将知识图谱编译为视觉记忆以进行多项选择题回答

Peng, Yixin, Jin, Er, Luo, Shiwei, Collarana, Diego, Decker, Stefan

Abstract

Knowledge graphs are usually integrated into question answering by encoding a retrieved subgraph with a graph neural network and fusing it with the language model in the online inference path. The same subgraph is therefore re-encoded from scratch every time a pair is scored, across training epochs, seeds, and evaluation runs, even though the knowledge graph never changes. We ask whether the retrieved knowledge graphs can instead be compiled once, offline, and then accessed as read-only memory. VisKG-LM shows that it can, by decoupling graph encoding from language reasoning. It serializes each retrieved candidate-specific subgraph as Relation-Labeled Paths and renders the result as an image whose two-dimensional layout preserves the branching structure of the paths. Each image is encoded once, offline, and cached for reuse. At inference, the language model contextualizes the question and candidate from text alone, and only its final layer consults the cached visual memory, reading both its global layout and its local relational detail. The graph information thus enters only after the text has been understood. On the test sets of CommonsenseQA, OpenBookQA, and MedQA-USMLE, VisKG-LMimproves over GreaseLM by $1.2$, $0.8$, and $4.3$ points, respectively, while matching or surpassing GraphVis, a $7$B vision-language model, with only about $400$M online parameters. Against a matched text-only control that receives the identical Relation-Labeled Paths, it gains $4.2$, $6.5$, and $5.1$ points across the three benchmarks. These gains show that the complete visual-memory interface adds value beyond path textualization alone and support compiled visual memory as an alternative to online graph propagation.

Chinese Translation

知识图谱通常通过使用图神经网络对检索到的子图进行编码，并在在线推理路径中与语言模型融合，从而集成到问答系统中。因此，每次对一对进行评分时，即使知识图谱从未改变，仍会在训练周期、随机种子和评估运行中从头开始重新编码相同的子图。我们提出是否可以将检索到的知识图谱一次性编译为离线存储，然后作为只读内存进行访问。VisKG-LM展示了这一点，通过将图编码与语言推理解耦。它将每个检索到的候选特定子图序列化为关系标记路径，并将结果呈现为一种图像，其二维布局保留了路径的分支结构。每个图像在离线时编码一次，并缓存以供重用。在推理时，语言模型仅从文本中对问题和候选项进行上下文化，只有其最终层会查询缓存的视觉记忆，读取其全局布局和局部关系细节。因此，图信息仅在文本被理解后才进入。在CommonsenseQA、OpenBookQA和MedQA-USMLE的测试集中，VisKG-LM分别比GreaseLM提高了1.2、0.8和4.3分，同时与GraphVis（一个70亿参数的视觉-语言模型）相匹配或超越，而其在线参数仅约为4亿。与接收相同关系标记路径的匹配文本控制相比，它在这三个基准上分别获得了4.2、6.5和5.1分。这些提升表明，完整的视觉记忆接口在路径文本化之外增加了价值，并支持编译的视觉记忆作为在线图传播的替代方案。

View on arXiv Download PDF AI Translation

cs.CL / 11 / 2609.19164

Advantage Scale Calibration Imbalance in Group-Relative Optimization under Low-Variance Rewards: Diagnosis and Bounded Recovery

低方差奖励下的群体相对优化中的优势规模校准失衡：诊断与有界恢复

Ding, Fei

Abstract

In verifier-style RLVR, group-relative optimization often treats advantage scale as an implementation detail. This paper separates two low-variance cases: sub-resolution jitter that should not become a preference signal, and credible but small cardinal gaps that should be learned without distorting KL calibration. We propose an advantage-scale three-way calibration interface: the same within-group scale denominator simultaneously determines the reward-branch strength, prompt-level batch weight, and the effective KL calibration induced when the reward branch is re-expressed on the original cardinal scale. This interface explains why RLOO / Dr.GRPO can let credible small gaps become KL dominated, whereas GRPO's standard-deviation denominator can amplify tiny gaps without bound. Based on this interface, we further introduce the Reward-Resolution Protocol and MaxNorm-AC, respectively filtering sub-resolution gaps and providing bounded cardinal recovery on credible nonzero gaps. Across dense / MoE architectures and math / code reasoning, MaxNorm-AC improves over the strongest robust-scale baseline while truncating the low-variance inverse-scale tail.

Chinese Translation

在验证者风格的RLVR中，群体相对优化通常将优势规模视为实现细节。本文区分了两种低方差情况：不应成为偏好信号的亚分辨率抖动，以及应在不扭曲KL校准的情况下学习的可信但较小的基数差距。我们提出了一种优势规模三路校准接口：相同的组内规模分母同时决定奖励分支强度、提示级别批次权重，以及在奖励分支在原始基数规模上重新表达时引发的有效KL校准。该接口解释了为何RLOO / Dr.GRPO能够让可信的小差距变得受KL主导，而GRPO的标准差分母则可以无限放大微小差距。基于此接口，我们进一步引入了奖励分辨率协议和MaxNorm-AC，分别过滤亚分辨率差距并在可信的非零差距上提供有界基数恢复。在密集/MoE架构以及数学/代码推理中，MaxNorm-AC在截断低方差逆规模尾部的同时，优于最强的鲁棒规模基线。

View on arXiv Download PDF AI Translation

cs.CL / 12 / 2609.19167

To Memories and Beyond: From Remembering to Knowing You across Long-Term Multimodal Personal Archives

记忆与超越：从记忆到认识您在长期多模态个人档案中的演变

Zhou, Wenqi, Yu, Zhuorui, Wen, Kaiao, Zheng, Hao, Zheng, Xinyi, Wu, Peiran, Zhou, Enmin, Wu, Chi-Hao, Shen, Junxiao

Abstract

As AI systems evolve into personalized digital companions, a central capability is reasoning over a user's long-term personal history: not merely storing past events, but tracking longitudinal experiences and evolving preferences. Progress here is bottlenecked by evaluation, existing long-term memory benchmarks are largely synthetic and text-only, they overlook the visual records that anchor everyday human memory, lack the authentic and causally connected longitudinal data that real personalization demands, and consequently remain confined to shallow factual recall. We introduce ReaLMem (Real-world Long-term Multimodal Memory), the first benchmark built from authentic multi-year personal visual archives, paired with first-person subjective annotations. ReaLMem evaluates models across three cognitive tiers of increasing difficulty: factual recall, persona inference, and predictive personalization. We further propose ChronoProfiler, a temporal-weighting profiling module that computes temporal stability scores for user attributes and applies them as a salience prior, resolving conflicts among temporally inconsistent preferences and helping models compound multiple co-active preferences in complex personalized decisions. Extensive evaluation of frontier multimodal large language models (MLLMs) and memory systems on ReaLMem reveals predictive personalization as a consistent ceiling, exposes clear performance gaps and bottlenecks between MLLMs and memory systems, and shows that high-quality, temporally informed representations substantially improve personalization. Together, ReaLMem and ChronoProfiler provide an authentic testbed and a simple, effective mechanism for long-term personalization, laying a foundation for future research on lifelong AI companions.

Chinese Translation

随着人工智能系统演变为个性化数字伴侣，一个核心能力是对用户长期个人历史的推理：不仅仅是存储过去的事件，而是跟踪纵向经历和不断变化的偏好。目前在这一领域的进展受到评估的瓶颈，现有的长期记忆基准主要是合成的且仅限于文本，它们忽视了锚定日常人类记忆的视觉记录，缺乏真实个性化所需的真实且因果关联的纵向数据，因此仍然局限于浅层的事实回忆。我们引入了ReaLMem（真实世界长期多模态记忆），这是第一个基于真实的多年个人视觉档案构建的基准，并配有第一人称的主观注释。ReaLMem在三个认知层次上评估模型，难度逐渐增加：事实回忆、个性推断和预测个性化。我们进一步提出了ChronoProfiler，一个时间加权的个人特征分析模块，计算用户属性的时间稳定性得分，并将其作为显著性先验，解决时间不一致偏好之间的冲突，帮助模型在复杂的个性化决策中综合多个共同活跃的偏好。对前沿多模态大型语言模型（MLLMs）和记忆系统在ReaLMem上的广泛评估揭示了预测个性化作为一致的上限，暴露了MLLMs与记忆系统之间明显的性能差距和瓶颈，并表明高质量、时间信息丰富的表示显著改善个性化。ReaLMem和ChronoProfiler共同提供了一个真实的测试平台和一个简单有效的长期个性化机制，为未来关于终身人工智能伴侣的研究奠定了基础。

View on arXiv Download PDF AI Translation

cs.CL / 13 / 2609.19238

YNU-HPCC at SemEval-2025 Task 11: Bridging the Gap in Text-Based Emotion Using Multiple Prediction Headers

YNU-HPCC在SemEval-2025任务11中的表现：通过多个预测头弥合基于文本的情感差距

Yang, Hao, Wang, Jin, Zhang, Xuejie

Abstract

This paper describes the participation of the YNU-HPCC team in subtask A of task 11, Bridging the Gap in Text-Based Emotion at SemEval-2025. Our best-performing system employs the RoBERTa (Robustly Optimized BERT Approach) model, an improved version of BERT that utilizes the Transformer encoder architecture. We enhanced the output head to allow the model to process one emotion simultaneously. We obtained the official ranking score (0.44), including results from all languages. The entire dataset was translated into English using Google Translate to facilitate subsequent processing. Through probabilistic and attention analyses, we found that (I) a single prediction head performs better than six heads predicting six emotions simultaneously, and (II) training on a uniformly translated English dataset yields better results than using the original dataset. The code is available at: https://github.com/BGWH123/Semeval-2025-task11.

Chinese Translation

本文描述了YNU-HPCC团队在SemEval-2025任务11子任务A——弥合基于文本的情感差距中的参与。我们表现最佳的系统采用了RoBERTa（鲁棒优化的BERT方法）模型，这是一个改进版的BERT，利用了Transformer编码器架构。我们增强了输出头，使模型能够同时处理一种情感。我们获得了官方排名分数（0.44），包括所有语言的结果。整个数据集通过谷歌翻译翻译成英语，以便于后续处理。通过概率和注意力分析，我们发现（I）单个预测头的表现优于六个头同时预测六种情感的情况，以及（II）在均匀翻译的英语数据集上训练的结果优于使用原始数据集。代码可在以下链接获取：https://github.com/BGWH123/Semeval-2025-task11。

View on arXiv Download PDF AI Translation

cs.CL / 14 / 2609.19291

Why Pretraining Fails to Share Cross-Lingual Knowledge

为何预训练难以实现跨语言知识共享

Gaber, Adam, Dolev, Uriel, Fittschen, Elisabeth, Cheng, Bobby, Marton, Yuval, Choshen, Leshem

Abstract

Large Language Models (LLMs) have made remarkable progress in the processing and modeling of many languages. Yet, unlike human multilinguals, they exhibit surprisingly limited cross-lingual knowledge transfer. While this limitation is well documented, its origins during multilingual training remain unclear. We pretrain 360M- and 7B-parameter LLMs and show that poor cross-lingual knowledge generalization emerges during pretraining and persists under standard interventions. To isolate its cause, we employ a controlled bilingual pretraining setting using two copies of the same language, sharing identical text and token segmentation, but mapped to disjoint token spaces. We find that disjoint tokens alone are enough to induce knowledge compartmentalization, even between identical copies of the same language, establishing disjoint token spaces as a fundamental barrier to cross-lingual knowledge generalization. Guided by this understanding, we suggest mapping languages into a shared token space by simple word-wise translation and find it substantially improves cross-lingual knowledge generalization, recovering up to 12.6\% of native-language learning efficiency --- 14$\times$ the baseline.

Chinese Translation

大型语言模型（LLMs）在多语言处理与建模方面取得了显著进展。然而，与人类多语言使用者不同，它们在跨语言知识迁移方面表现出令人惊讶的有限性。尽管这一限制已有充分记录，但其在多语言训练过程中的根源尚不明确。我们预训练了参数规模为3.6亿和70亿的LLMs，结果表明，跨语言知识泛化能力的不足是在预训练阶段产生的，并且在常规干预下依然存在。为探究其原因，我们采用了受控的双语预训练设置，使用两份相同语言的文本，文本内容和分词方式完全一致，但映射到不相交的词元空间。研究发现，仅仅是词元空间不相交就足以导致知识的隔离，即使是在相同语言的两份文本之间，这确立了不相交词元空间作为跨语言知识泛化的根本障碍。基于这一认识，我们提出通过简单的词对词翻译将不同语言映射到共享的词元空间，实验结果显示该方法显著提升了跨语言知识泛化能力，恢复了高达12.6%的母语学习效率，约为基线的14倍。

View on arXiv Download PDF AI Translation

cs.CL / 15 / 2609.19334

A frontend-backend architecture for tool calls in full-duplex speech models

全双工语音模型中工具调用的前端-后端架构

Hu, Ke, Deng, Slyne, Chen, Chen, Rastorgueva, Elena, Casanova, Edresson, Kumar, Punit, Choudhary, Dharmendra, Srihari, Nikhil, Mahabaleshwarkar, Ameya Sunil, Trinh, Viet Anh, Essid, Slim, Olabiyi, Oluwatobi, Chen, Zhehuai

Abstract

Full-duplex speech-to-speech (S2S) models provide natural, low-latency conversational interaction and would benefit from the ability to use external tools and complete voice-agent tasks. We propose a frontend-backend architecture where a duplex speech-to-text frontend learns to emit a delegation token and forwards streaming ASR transcripts to a text-based backend LLM for tool calls. Tool-call results from the backend are injected back into the frontend through a lightweight prefill-and-repeat mechanism and then synthesized using streaming TTS to the user. Our approach largely preserves regular duplex turn-taking, interruption handling, and low-latency interaction as it requires minimal modifications to the frontend model. In a single-turn tool-call evaluation, our system achieves 92-97% tool-call recall, competitive tool-call prediction performance, and 81.2% accuracy in rejecting irrelevant calls. When equipped with a larger backend (e.g., Qwen3-235B-A22B), our system achieves competitive results on Full-Duplex-Bench-V3 compared to open and closed source models, and significantly outperforms GPT-realtime-mini and Qwen3-Omni-30B-A3B-Instruct on EVA-Bench. These results demonstrate that backend delegation is an effective and modular approach for combining natural duplex speech interaction with strong agentic tool-call capabilities.

Chinese Translation

全双工语音到语音（S2S）模型提供自然、低延迟的对话交互，并且能够使用外部工具和完成语音代理任务。我们提出了一种前端-后端架构，其中双工语音转文本前端学习发出委托令牌，并将流式自动语音识别（ASR）转录文本转发给基于文本的后端大语言模型（LLM）进行工具调用。后端的工具调用结果通过轻量级的预填充与重复机制注入回前端，然后使用流式文本转语音（TTS）合成返回给用户。我们的方法在很大程度上保持了常规的双工轮流发言、打断处理和低延迟交互，因为它对前端模型的修改最小。在单轮工具调用评估中，我们的系统实现了92-97%的工具调用召回率，具有竞争力的工具调用预测性能，以及81.2%的拒绝无关调用的准确率。当配备更大的后端（例如，Qwen3-235B-A22B）时，我们的系统在Full-Duplex-Bench-V3上取得了与开源和闭源模型相当的结果，并在EVA-Bench上显著超越了GPT-realtime-mini和Qwen3-Omni-30B-A3B-Instruct。这些结果表明，后端委托是一种有效且模块化的方法，可以将自然的双工语音交互与强大的代理工具调用能力相结合。

View on arXiv Download PDF AI Translation

cs.CL / 16 / 2609.19366

The Role of Fine-grained Harm Signals in LLM Safety

细粒度危害信号在大型语言模型安全性中的作用

Park, Soyeon, Jeong, Seogyeong, Kim, Sunwoo, Oh, Alice

Abstract

Prior work has shown that internal harmfulness representations in large language models vary across risk categories, while sharing a common general harm representation component. This raises a question about the role of the category-specific component beyond general harm representation in LLM safety. To answer this question, we isolate the category-specific component by removing shared general harmfulness representation from each categorical harmfulness representation, yielding a category residual that is orthogonal to general harmfulness at every layer. Using activation steering with category residuals across 11 risk categories in 3 instruction-tuned LLMs, we find that whether category residuals encode harmfulness varies across categories, and that this category-wise pattern is similar across models. Whether category residuals induce refusal also varies across categories, but this category-wise pattern is more model-dependent. We also find that category residuals increase LLMs' downstream internal alignment with shared general harmfulness representation. Together, these findings demonstrate that more fine-grained category residuals should also be considered beyond shared general harmfulness representation to fully understand LLM safety. More broadly, our findings show that even a direction orthogonal to a concept at one layer can contribute to the concept's downstream amplification.

Chinese Translation

先前的研究表明，大型语言模型中的内部危害性表征在不同风险类别之间存在差异，同时共享一个通用的危害性表征组件。这引发了一个问题，即类别特定组件在大型语言模型安全性中超越通用危害性表征的作用。为了解答这个问题，我们通过从每个类别的危害性表征中去除共享的通用危害性表征，来隔离类别特定组件，从而得到一个在每一层与通用危害性正交的类别残差。通过在3个经过指令调优的大型语言模型中使用类别残差进行激活引导，我们发现类别残差是否编码危害性在不同类别之间存在差异，并且这种类别间的模式在模型之间是相似的。类别残差是否引发拒绝也在不同类别之间存在差异，但这种类别间的模式更依赖于模型。我们还发现，类别残差增加了大型语言模型与共享通用危害性表征的下游内部一致性。综合这些发现，表明在全面理解大型语言模型的安全性时，除了共享的通用危害性表征外，更细粒度的类别残差也应被考虑。更广泛地说，我们的发现表明，即使在某一层与一个概念正交的方向也可以对该概念的下游放大产生贡献。

View on arXiv Download PDF AI Translation

cs.CL / 17 / 2609.19417

Less Is More: Graph-free Multimodal RAG via Multi-signal Late Fusion

少即是多：无图的多模态检索增强生成（RAG）通过多信号后期融合

Rakshit, Tithi, Zhou, Hongkuan, Halilaj, Lavdim, Zhu, Yuqicheng

Abstract

Graph-based retrieval-augmented generation (RAG) is widely used for multimodal, cross-document question answering. However, building corpus-level graphs is expensive, slow to query, and difficult to maintain. We present TrioRAG, a graph-free multimodal framework that integrates evidence from three complementary signals: the question, the anchor image, and a VLM-enhanced query generated from both. Each signal retrieves independently over a shared multi-vector index of page text and page images, and the results are combined through late fusion. Further, we introduce AutoQA, a multimodal automotive benchmark whose questions are grounded in noisy, web-sourced images rather than clean document-sourced figures. Its questions require reasoning across manuals. We position it as a model-curated testbed rather than a human-validated gold standard. Across three benchmarks, TrioRAG matches or outperforms graph-based systems while reducing total cost and accelerating per-query inference by 1.6-2.3 times. By construction, AutoQA grounds its questions in out-of-corpus web images. In this setting image retrieval reaches only 19.3% document-level recall, while text-derived signals, especially the VLM-enhanced query, keep retrieval robust.

Chinese Translation

基于图的检索增强生成（RAG）在多模态跨文档问答中被广泛使用。然而，构建语料库级别的图是昂贵的，查询速度慢且难以维护。我们提出了TrioRAG，这是一种无图的多模态框架，整合了来自三个互补信号的证据：问题、锚图像以及从两者生成的增强查询（VLM-enhanced query）。每个信号在共享的多向量索引上独立检索页面文本和页面图像，结果通过后期融合进行组合。此外，我们引入了AutoQA，这是一个多模态汽车基准，其问题基于噪声较大的网络图像，而非干净的文档来源图形。其问题需要跨手册进行推理。我们将其定位为模型策划的测试平台，而非经过人工验证的黄金标准。在三个基准测试中，TrioRAG的表现与基于图的系统相当或更优，同时降低了总成本，并将每次查询的推理速度提高了1.6-2.3倍。通过构建，AutoQA将其问题基于语料库外的网络图像。在这种情况下，图像检索的文档级召回率仅为19.3%，而文本衍生信号，尤其是增强查询，保持了检索的稳健性。

View on arXiv Download PDF AI Translation

cs.CL / 18 / 2609.19504

For Your Eyes Only: Evaluating Coordination Between Isolated Language Model Instances

仅供您参考：评估孤立语言模型实例之间的协调

Shirnin, Alexander, Kudelya, Aleksey

Abstract

As model-generated content is increasingly consumed by other model instances in automated workflows, a practically important question arises: can a model embed a signal in natural language that an independent instance of the same model can detect, relying only on shared pre-training and task instructions, without any shared memory or coordination-specific training? We introduce For Your Eyes Only, a cooperative signalling game designed to evaluate this directly. A Sender produces free-form descriptions for two words, one of which is a hidden target; an isolated Receiver must identify it. We evaluate seven contemporary models from four architectural families on 300 word pairs from established psycholinguistic corpora, using the Double-Pass Success Rate to control for output biases. We find that most models struggle to maintain coordination once they are required to avoid detectable signals, while one frontier model retains near-perfect performance even after such filtering. We further show that models can direct this capability toward deliberate misdirection, and that coordination is consistently weaker across architectures than within them.

Chinese Translation

随着模型生成的内容在自动化工作流程中越来越多地被其他模型实例消费，一个实际重要的问题随之出现：模型是否能够在自然语言中嵌入一个信号，使得同一模型的独立实例能够检测到，而无需任何共享内存或协调特定的训练，仅依赖于共享的预训练和任务指令？我们引入了“仅供您参考”（For Your Eyes Only），这是一个旨在直接评估这一点的合作信号游戏。发送者为两个单词生成自由形式的描述，其中一个是隐藏目标；孤立的接收者必须识别出它。我们在来自已建立的心理语言学语料库的300对单词上评估了四个架构家族中的七个当代模型，使用双重通过成功率（Double-Pass Success Rate）来控制输出偏差。我们发现，大多数模型在被要求避免可检测信号时，难以保持协调，而一个前沿模型即使在这种过滤后仍保持近乎完美的表现。我们进一步表明，模型可以将这一能力用于故意的误导，并且跨架构的协调能力始终弱于架构内部的协调能力。

View on arXiv Download PDF AI Translation

cs.CL / 19 / 2609.19553

From Parameters to Behaviors: A Survey of Model Fusion for Large Language Models

从参数到行为：大型语言模型融合的调查

Cai, Shuo, Gu, Yanggan, Wang, Zihao, Wang, Yuanyi, Yan, Yibo, Wang, Wenjun, Liu, Yuhang, Zhu, Guanghao, Huang, Sirui, Li, Ming, Yang, Hongxia

Abstract

Model fusion integrates the capabilities from source models into a single target model. As of June 2026, Hugging Face hosts more than 2M models. This growing pool provides a rich base for model reuse and capability integration. Yet existing surveys often cover only separate parts of this space, and they do not provide a unified definition or a systematic taxonomy. This survey defines model fusion and organizes prior work into three levels: parameter-level, representation-level, and behavior-level fusion. We also review related metrics, benchmarks, and applications, summarize current challenges, and identify future directions. Our goal is to provide a clear map of this area and support future work on model fusion. A comprehensive list of papers about model fusion is available at https://github.com/Baicaihaochi/Awesome-Model-Fusion-Survey.

Chinese Translation

模型融合将源模型的能力整合到一个单一的目标模型中。截至2026年6月，Hugging Face托管了超过200万个模型。这一不断增长的模型库为模型重用和能力整合提供了丰富的基础。然而，现有的调查往往仅覆盖这一领域的部分内容，并未提供统一的定义或系统的分类法。本调查定义了模型融合，并将先前的研究组织为三个层次：参数层面、表示层面和行为层面融合。我们还回顾了相关的度量标准、基准测试和应用，总结了当前面临的挑战，并确定了未来的研究方向。我们的目标是提供该领域的清晰地图，并支持未来的模型融合研究。关于模型融合的综合文献列表可在 https://github.com/Baicaihaochi/Awesome-Model-Fusion-Survey 获取。

View on arXiv Download PDF AI Translation

cs.CL / 20 / 2609.19585

CliniCIRCA: A Modular LLM Framework for Constructing Longitudinal Mental Health Patient Journeys from Raw EHR Narratives

CliniCIRCA：构建来自原始电子健康记录叙述的纵向心理健康患者历程的模块化大型语言模型框架

Zhang, Aiwei Ivy, Ishfaq, Nimra, Chandra, Mohit, Lesmes, Santiago Alvarez, Coscia, Adam, Moon, Khatiya Chelidze, Ding, Xiaohan, De Choudhury, Munmun

Abstract

In mental health care, reasoning over patient journeys is a key task for clinicians. Yet these journeys, encompassing a longitudinal progression of biological, psychological, and social events, are often spread across disparate unstructured text narratives, making temporal recovery challenging. We present CliniCIRCA, a multi-stage LLM framework for Calendar-anchored, Imprecision-aware Reconstruction of Clinical Annals. To our knowledge, CliniCIRCA is the first to temporally classify clinical events across unstructured discharge summaries without event-level timestamps. From 14,882 MIMIC-III mental health admissions, we first construct a benchmark of 52 discharge summaries on which CliniCIRCA produces 15,891 temporally tagged events. After correcting 629 errors based on a clinician-in-the-loop evaluation, we produce verified gold-standard labels. Finally, the corrected timelines drive a temporally grounded summarization stage that compresses each source 1.52 times into a date-grouped chronological record. We then scale the framework to generate 1,000 silver-standard timelines and evaluate them as training data. Compared with zero- and few-shot prompting, instruction tuning generally improves five open-weight models on event extraction, temporal tagging, and summarization across silver and clinician-verified evaluations.

Chinese Translation

在心理健康护理中，对患者历程进行推理是临床医生的一项关键任务。然而，这些历程涵盖生物、心理和社会事件的纵向进展，通常分散在不同的非结构化文本叙述中，使得时间恢复变得具有挑战性。我们提出了CliniCIRCA，一个多阶段的大型语言模型框架，用于临床年鉴的日历锚定和不确定性意识重建。据我们所知，CliniCIRCA是首个在没有事件级时间戳的情况下，对非结构化出院总结中的临床事件进行时间分类的系统。我们从14,882个MIMIC-III心理健康入院记录中，首先构建了一个包含52个出院总结的基准数据集，CliniCIRCA在该数据集上生成了15,891个时间标记事件。在基于临床医生参与评估的基础上纠正了629个错误后，我们生成了经过验证的黄金标准标签。最后，经过修正的时间线驱动了一个时间基础的总结阶段，将每个来源压缩为日期分组的时间记录，压缩比例为1.52倍。随后，我们将该框架扩展以生成1,000个银标准时间线，并将其作为训练数据进行评估。与零样本和少样本提示相比，指令调优通常改善了五个开放权重模型在事件提取、时间标记和总结方面的表现，涵盖了银标准和临床医生验证的评估。

View on arXiv Download PDF AI Translation

cs.CL / 21 / 2609.19589

Form Over Content In Gradient-Based Data Attribution Methods

梯度基础数据归因方法中的形式与内容

Kim, Sunwoo, Jung, Seokwon, Kim, Sohyung, Oh, Seong Joon, Oh, Alice

Abstract

Data attribution methods using gradient similarity are widely used to analyze and select training data for large language models, but what gradient similarity actually measures is debated. Some interpret it as identifying task-relevant skills, while other work reports that surface form is the main factor. We resolve this debate for supervised fine-tuning examples by varying task and answer format independently. Specifically, we render benchmarks in different answer formats, such that datasets can share a task without a format or a format without a task. We find that gradient alignment follows the answer format, as benchmark pairs sharing an answer format align strongly (disattenuated cosine near 0.4), while same benchmarks rendered with different answer format classes show no alignment (near 0.0). We demonstrate that this ordering holds from the earliest pretraining checkpoints through post-training, and across model scales and families. We then analyze the released selections of LESS, a gradient-based data selection method for instruction tuning, and find that each target's selections over-represent the target's own answer format. Hence, we demonstrate that gradient-based attribution methods track format similarity more than task semantics, meaning that such methods, as well as the semantic interpretation of the gradient, should be tested on data where answer format and task vary independently for greater robustness and reliability.

Chinese Translation

使用梯度相似性的数据归因方法被广泛应用于分析和选择大型语言模型的训练数据，但梯度相似性实际测量的内容仍存在争议。一些研究将其解释为识别与任务相关的技能，而其他研究则报告表面形式是主要因素。我们通过独立变化任务和答案格式来解决这一争论，特别是我们在不同的答案格式下呈现基准，使得数据集可以在没有格式或没有任务的情况下共享任务。我们发现梯度对齐遵循答案格式，因为共享答案格式的基准对之间的对齐强烈（去注意化余弦接近0.4），而使用不同答案格式类别呈现的相同基准则显示没有对齐（接近0.0）。我们证明这种排序从最早的预训练检查点到后期训练都成立，并且在不同的模型规模和家族中均适用。然后，我们分析了LESS的发布选择，LESS是一种用于指令调优的基于梯度的数据选择方法，发现每个目标的选择过度代表了目标自身的答案格式。因此，我们证明基于梯度的归因方法更关注格式相似性而非任务语义，这意味着这些方法以及梯度的语义解释应在答案格式和任务独立变化的数据上进行测试，以提高其鲁棒性和可靠性。

View on arXiv Download PDF AI Translation

cs.CL / 22 / 2609.19596

Full-Duplex Speech Models Take the Floor When Asked, Not When Needed

全双工语音模型在被询问时发言，而非在需要时发言

Peng, Linkai, Nuchged, Baorian, Fu, Kaiqi, Yao, Yuyang

Abstract

Full-duplex speech models listen and speak at once, promising always-on assistants. Yet they must also decide when they should speak. Human listeners speak when addressed or when the speaker stops, but also self-select to correct a false claim, supply a missing word, or warn of danger. We ask whether full-duplex models do the same. To separate the reason to speak from the opportunity, we construct context-matched English monologues in which only the trigger utterance varies within a topic, define 10 conditions from turn-allocation rules, and compress inter-word pauses to limit opportunities created by silence. Across five model families, being addressed and silence are far more reliable triggers than false facts or hazards. Frame-level text-token probabilities in Moshi and PersonaPlex are lower for false facts than for Neutral when averaged over the first 2\,s after trigger end. Pauses or permission to interrupt do not close this gap either. Given the floor, Moshi and PersonaPlex answer most direct questions, yet the proportion of non-empty false-fact replies that challenge the claim is only .14--.15, and the proportion of hazard replies that warn of danger is .04--.07. This paper thus identifies a gap in both speech initiation and response content. Closing it requires genuine content understanding and intervention decisions grounded in it.

Chinese Translation

全双工语音模型能够同时听和说，承诺提供始终在线的助手。然而，它们还必须决定何时发言。人类听众在被提及时或说话者停顿时发言，但也会主动选择在纠正错误陈述、补充缺失词汇或警告危险时发言。我们探讨全双工模型是否也如此。为了将发言的理由与机会分开，我们构建了上下文匹配的英语独白，其中只有触发语句在主题内变化，定义了10种轮次分配规则，并压缩了词间停顿，以限制由沉默创造的机会。在五个模型家族中，被提及时和沉默远比错误事实或危险更可靠地触发发言。在触发结束后的前2秒内，Moshi和PersonaPlex的框架级文本标记概率在错误事实下低于中性。停顿或打断许可并未缩小这一差距。在获得发言权的情况下，Moshi和PersonaPlex回答了大多数直接问题，但挑战该主张的非空错误事实回复比例仅为0.14-0.15，警告危险的危险回复比例为0.04-0.07。因此，本文识别了在语音启动和响应内容方面的差距。弥补这一差距需要真正的内容理解和基于此的干预决策。

View on arXiv Download PDF AI Translation

cs.CL / 23 / 2609.19606

Chain-of-Thought Entropy as a Reliability Signal: A Preregistered Reproduction

链式思维熵作为可靠性信号：一项预注册的重现研究

Cochran, Theodore O.

Abstract

This empirical study is an independent reproduction of the dissociation Zhao reported in 2026. The shape of a large language model's chain-of-thought entropy trajectory predicts whether the final answer is correct, while the magnitude of its total entropy drop does not. The dissociation merits reproduction because the magnitude half rests on a single 300-problem run with one model at one seed, while the shape half was reported at full scale on both benchmarks and on a second model family. Registered at OSF before any confirmatory run, the reproduction crosses the complete GSM8K and MATH-500 benchmark test sets with four open-weight models including one reasoning-distilled model of a kind the original did not test. The shape signal replicates. The magnitude signal divides by setting. On the anchor model the accuracy gap between monotone and non-monotone chains is +9.6 percentage points on GSM8K and +27.5 on MATH-500, while the rank correlation of the total entropy drop with correctness is -0.018 on GSM8K and +0.414 on MATH-500. On the reasoning-distilled model the binary form of the shape signal fires on about one chain in a hundred, too few to estimate the registered contrast, while the graded violation count remains predictive there. In an exploratory comparison the final-step entropy alone outperforms the binary shape flag in all eight model-by-benchmark cells by ROC area, and in six or seven by the risk-coverage area the original reports, depending on an integration range the original does not state. The study contributes a reproduction of the shape signal at full test-set scale under seven documented protocol differences, a map of the settings where the magnitude signal holds and fails, and measurements of four protocol dependencies the original does not report.

Chinese Translation

本实证研究是对赵在2026年报告的解离现象的独立重现。大型语言模型的链式思维熵轨迹的形状可以预测最终答案是否正确，而其总熵下降的幅度则不能。该解离现象值得重现，因为幅度部分仅基于一个模型在一个种子下的300个问题的单次运行，而形状部分则在两个基准上和第二个模型系列上进行了全面报告。该重现研究在任何确认性运行之前已在OSF注册，涵盖了完整的GSM8K和MATH-500基准测试集，使用了四个开放权重模型，包括一个原始研究未测试的推理提炼模型。形状信号得到了复制，而幅度信号则因设置而异。在锚定模型上，单调链与非单调链之间的准确率差距在GSM8K上为+9.6个百分点，在MATH-500上为+27.5，而总熵下降与正确性之间的秩相关性在GSM8K上为-0.018，在MATH-500上为+0.414。在推理提炼模型上，形状信号的二元形式在约每一百个链中激发一次，数量太少以至于无法估计注册的对比，而分级违反计数在此仍然具有预测性。在探索性比较中，最终步骤熵单独在所有八个模型与基准单元中通过ROC面积超越了二元形状标志，而在原始报告的风险覆盖面积中，在六个或七个单元中也表现出优势，具体取决于原始未说明的集成范围。本研究贡献了在七个记录的协议差异下对形状信号的全面测试集规模的重现，绘制了幅度信号成立与失败的设置图，以及四个原始未报告的协议依赖性的测量。

View on arXiv Download PDF AI Translation

cs.CL / 24 / 2609.19615

Semantic Layer Induction from Raw Telemetry via Hierarchical LLM and RAG Abstraction

通过层次化大语言模型和RAG抽象从原始遥测数据中引导语义层

Jia, Yuanzhe, Anaissi, Ali

Abstract

Modern applications generate massive volumes of raw telemetry data, but translating those noisy, heterogeneous event streams into actionable business insights remains a fundamental challenge. Data engineers and analysts expend substantial effort reconciling semantic discrepancies, hand-crafting parsing logics, and maintaining fragile mappings between raw data and business KPIs. In this paper, we present an end-to-end framework that fully automates the construction of a business semantic layer from application raw logs. Our approach introduces a two-stage semantic abstraction: first, high-level business features are identified via LLM inference augmented with domain-specific industry knowledge; second, fine-grained business nodes are derived through a structured pipeline comprising data refinement, hybrid retrieval, multi-stage filtering, semantic clustering, and canonical naming. Evaluation on production-scale telemetry demonstrates that our system improves human-assessed semantic quality from 50 to 80+ on a 100-point scale, reduces maintenance effort by 80%, filters out 74% of noise, and achieves 0.87 Cohen's kappa via an integrated LLM-as-Judge evaluation, enabling continuous, scalable quality assurance. Overall, our work distinguishes itself from prior work by addressing the novel problem of business semantic layer induction from raw telemetry, operating without labeled training data or manual rule engineering.

Chinese Translation

现代应用程序生成大量的原始遥测数据，但将这些嘈杂且异构的事件流转化为可操作的商业洞察仍然是一个基本挑战。数据工程师和分析师在调和语义差异、手动构建解析逻辑以及维护原始数据与商业关键绩效指标（KPI）之间脆弱的映射方面付出了大量努力。本文提出了一个端到端框架，完全自动化地构建应用程序原始日志的商业语义层。我们的方法引入了一个两阶段的语义抽象：首先，通过增强行业特定知识的LLM推理识别高层次的商业特征；其次，通过一个结构化管道导出细粒度的商业节点，该管道包括数据精炼、混合检索、多阶段过滤、语义聚类和规范命名。在生产规模的遥测数据上的评估表明，我们的系统在100分制上将人类评估的语义质量从50提升至80以上，减少了80%的维护工作，过滤掉74%的噪声，并通过集成的LLM作为评估者（LLM-as-Judge）评估实现了0.87的Cohen's kappa，从而实现了持续、可扩展的质量保证。总体而言，我们的工作通过解决从原始遥测数据中引导商业语义层这一新问题而与以往的研究区分开来，且不依赖标注的训练数据或手动规则工程。

View on arXiv Download PDF AI Translation

cs.CL / 25 / 2609.19650

Improving Cross-Lingual Transfer for Sequential Sentence Classification in Research Papers via Structural Similarity

通过结构相似性提高研究论文中顺序句子分类的跨语言迁移

Yamauchi, Kazuhiro, Katsurai, Marie

Abstract

Sequential sentence classification (SSC) is an essential task for structuring scientific publications, and extending SSC research to languages other than English can improve accessibility to scientific knowledge in multilingual digital libraries. Cross-lingual transfer is a promising approach to address the scarcity of training data in non-English languages. Prior work on other natural language processing tasks has shown the benefits of capturing linguistic similarity between source and target languages. However, SSC inherently depends on patterns at the discourse level, such as label sequences and positional regularities, which appear consistently across languages regardless of linguistic differences. To examine the factors that determine transfer success in SSC, we constructed a multilingual SSC dataset covering 13 non-English languages collected from five academic databases. Our cross-lingual transfer experiments, using both encoder-based and generative models, show that linguistic proximity has no consistent predictive power for transfer performance, whereas structural similarity in rhetorical organization shows a weak but consistent positive correlation across models. After controlling for source-language performance, the similarity of label distributions is the most consistent predictor. Building on this finding, we propose a set of three methods that explicitly leverage structural information using generative models. In the in-domain evaluation, the best combination reaches parity with the strongest encoder baselines, and in transfer to languages unseen during training, it outperforms the strongest encoder baseline.

Chinese Translation

顺序句子分类（SSC）是构建科学出版物的重要任务，将SSC研究扩展到英语以外的语言可以提高多语言数字图书馆中科学知识的可获取性。跨语言迁移是一种有前景的方法，可以解决非英语语言中训练数据稀缺的问题。先前在其他自然语言处理任务上的研究表明，捕捉源语言和目标语言之间的语言相似性具有显著的好处。然而，SSC本质上依赖于话语层面的模式，例如标签序列和位置规律，这些模式在不同语言中一致出现，无论语言差异如何。为了研究影响SSC迁移成功的因素，我们构建了一个涵盖来自五个学术数据库的13种非英语语言的多语言SSC数据集。我们的跨语言迁移实验使用了基于编码器和生成模型的方法，结果表明，语言接近性对迁移性能没有一致的预测能力，而修辞组织中的结构相似性在各个模型中显示出微弱但一致的正相关性。在控制源语言性能后，标签分布的相似性是最一致的预测因素。基于这一发现，我们提出了一套三种方法，明确利用生成模型中的结构信息。在领域内评估中，最佳组合达到了与最强编码器基线相当的水平，而在迁移到训练期间未见过的语言时，其表现超过了最强编码器基线。

View on arXiv Download PDF AI Translation

cs.CL / 26 / 2609.19736

A Phonemically Comprehensive, ASCII-Only Romanization Scheme for Thai and Lao: Systematic Cross-Lingual Correspondence and Chinese-User-Friendly Design

一种音位全面、仅使用ASCII字符的泰语和老挝语罗马化方案：系统的跨语言对应关系与中文用户友好的设计

Zhang, Zijie, Lee, Tan

Abstract

This paper proposes a phonemically comprehensive, ASCII-only romanization scheme for Thai and Lao, treating the two closely related languages as a unified cross-lingual design problem. The scheme represents segmental contrasts, vowel length, and lexical tone while maintaining one-symbol-one-phoneme transparency and systematic correspondence between Thai and Lao. The scheme prioritizes synchronic phonetic correspondence, including correspondence with Pinyin and Jyutping where applicable, while preserving historical-phonological correspondence where it does not conflict with phonetic transparency. Tone uses a compact single-digit default notation, supplemented by optional tone-value and historical tone-category representations. The resulting scheme provides a readable, keyboard-friendly, and machine-processable phonemic representation for language learning and cross-lingual speech processing.

Chinese Translation

本文提出了一种音位全面、仅使用ASCII字符的泰语和老挝语罗马化方案，将这两种密切相关的语言视为一个统一的跨语言设计问题。该方案表示了音段对比、元音长度和词汇音调，同时保持了单符号对应单音素的透明性以及泰语与老挝语之间的系统对应关系。该方案优先考虑共时的语音对应，包括在适用情况下与拼音（Pinyin）和粤拼（Jyutping）的对应，同时保留历史音韵对应，前提是这不与语音透明性相冲突。音调使用紧凑的单数字默认符号表示，并辅以可选的音调值和历史音调类别表示。最终的方案提供了一种可读、键盘友好且可机器处理的音位表示，适用于语言学习和跨语言语音处理。

View on arXiv Download PDF AI Translation

cs.CL / 27 / 2609.19778

Learn Before You Judge: Progressive Knowledge-to-Decision Alignment for Explainable Hateful Meme Detection

学习再判断：用于可解释仇恨表情包检测的渐进式知识到决策对齐

Xu, Bo, Wang, Chenyuan, Chen, Xinyu, Zhu, Quanhao, Lin, Rui, Zhao, Liang, Lin, Hongfei, Xia, Feng

Abstract

Hateful memes spread abusive content through implicit interactions between images and text, posing serious threats to the safety of online communities. In recent years, multimodal large language models have been widely used for hateful meme detection and are increasingly adopted to generate explainable detection results. However, we find that existing explain-then-detect methods often couple explanation generation and label prediction within the same training process. This coupling causes interference between task objectives, leading to limited detection performance and even worse results than simple SFT baselines. To address these challenges, we propose ProKDA, a progressive knowledge-to-decision alignment method for explainable hateful meme detection. Inspired by the human annotation training process, ProKDA first uses an agentic background knowledge construction pipeline to obtain external knowledge related to meme understanding. It then adopts a three-stage training strategy that sequentially performs background knowledge learning, hatefulness detection learning, and hatefulness boundary alignment. Unlike prior explain-then-detect methods that jointly optimize both tasks, ProKDA focuses on a single training objective at each stage. This design reduces interference between the two tasks and progressively transforms background knowledge into robust detection decisions. Experiments on three public hateful meme benchmarks show that ProKDA achieves state-of-the-art detection performance and provides accurate, explainable, and evidence-supported decisions for hateful meme moderation. Project page: https://meizhiyuan88666.github.io/prokda.

Chinese Translation

仇恨表情包通过图像与文本之间的隐含互动传播滥用内容，严重威胁在线社区的安全。近年来，多模态大型语言模型被广泛应用于仇恨表情包检测，并越来越多地被采用以生成可解释的检测结果。然而，我们发现现有的先解释后检测方法通常在同一训练过程中将解释生成与标签预测结合在一起。这种耦合导致任务目标之间的干扰，限制了检测性能，甚至比简单的监督微调（SFT）基线表现更差。为了解决这些挑战，我们提出了ProKDA，一种用于可解释仇恨表情包检测的渐进式知识到决策对齐方法。受到人类注释训练过程的启发，ProKDA首先使用代理背景知识构建管道获取与表情包理解相关的外部知识。然后，它采用三阶段训练策略，依次进行背景知识学习、仇恨性检测学习和仇恨性边界对齐。与之前的先解释后检测方法共同优化两个任务不同，ProKDA在每个阶段专注于单一的训练目标。这一设计减少了两个任务之间的干扰，并逐步将背景知识转化为稳健的检测决策。在三个公共仇恨表情包基准测试上的实验表明，ProKDA实现了最先进的检测性能，并为仇恨表情包的审核提供了准确、可解释且有证据支持的决策。项目页面：https://meizhiyuan88666.github.io/prokda

View on arXiv Download PDF AI Translation

cs.CL / 28 / 2609.19799

Evolution or Illusion? Rethinking Evaluation in LLM Evolutionary Search

进化还是幻觉？重新思考大语言模型进化搜索中的评估

Oved, Tal, Pony, Roi, Naparstek, Oshri, Barzelay, Udi

Abstract

LLM-driven evolutionary search finds programs by launching seeds and iterating each one. Papers report a single budget setting, usually one seed run for a fixed number of iterations, and rank methods from that one point. We show this is not enough. We evaluate three evolutionary search strategies on five optimization tasks, commonly used by papers in the genre to report results. We run the analysis over a full grid of seeds and iterations. Our findings suggest that the best way to split a fixed budget between more seeds (width) and more iterations (depth) changes with the strategy, the task, and the total budget. Furthermore, we observe that the ranking of strategies also changes with the budget. On one task the strategy that looks worst at one seed is best at forty seeds. On another the best number of iterations is well below the value common in practice, so extra depth wastes budget that more seeds would turn into score. We provide a measurement protocol that reports the seeds-by-iterations frontier and practical guidance for using it.

Chinese Translation

基于大语言模型（LLM）的进化搜索通过启动种子并对每个种子进行迭代来寻找程序。现有文献通常报告单一预算设置，即在固定迭代次数下进行一次种子运行，并基于这一点对方法进行排名。我们表明，这种做法是不够的。我们在五个优化任务上评估了三种进化搜索策略，这些任务通常被相关文献用来报告结果。我们对种子和迭代的全网格进行了分析。我们的发现表明，在固定预算下，将预算分配给更多种子（宽度）和更多迭代（深度）的最佳方式随着策略、任务和总预算的不同而变化。此外，我们观察到策略的排名也会随着预算的变化而变化。在一个任务中，在一个种子下表现最差的策略在四十个种子下却是最佳策略。在另一个任务中，最佳的迭代次数远低于实际中常见的值，因此额外的深度会浪费预算，而更多的种子则能转化为更高的得分。我们提供了一种测量协议，报告种子与迭代的边界，并为其使用提供了实用指导。

View on arXiv Download PDF AI Translation

cs.CL / 29 / 2609.19805

Dictionary-Constrained Grapheme-to-Phoneme for Unsegmented Languages from LLM-Annotated Data

基于字典约束的无分词语言的字形到音素转换（Grapheme-to-Phoneme）方法：来自大型语言模型标注数据的研究

Hu, Rui, Zhan, Zhenpeng, Lin, Xiaolong

Abstract

Grapheme-to-phoneme (G2P) conversion turns raw text into its phonemic form and is an essential part of both text-to-speech (TTS) and automatic speech recognition (ASR) systems. It is required to be fast, stable and context-aware. For unsegmented languages such as Japanese, G2P additionally couples word segmentation with highly context-dependent polyphone disambiguation, and the scarcity of accurately annotated data remains a bottleneck. In this paper, we present a context-aware neural G2P method that scores paths of a discriminative conditional random field (CRF) over a word lattice constructed from dictionaries. To tackle data scarcity, we utilize large language models (LLMs) to generate more than 2 million sentences. Experimental results demonstrate that our method strongly outperforms conventional morphological analyzer-based methods and neural sequence models. On the Joyo-Kanji-Yomi benchmark, our method reaches 99.62% target word reading accuracy, 0.32% target word phoneme error rate (PER) and 0.14% sentence PER.

Chinese Translation

字形到音素（G2P）转换将原始文本转化为其音素形式，是文本到语音（TTS）和自动语音识别（ASR）系统的重要组成部分。该过程需要快速、稳定且具备上下文感知能力。对于如日语等无分词语言，G2P还需要将词语分割与高度依赖上下文的多音字消歧结合起来，而准确标注数据的稀缺仍然是一个瓶颈。在本文中，我们提出了一种上下文感知的神经G2P方法，该方法通过构建自字典的词图，利用判别条件随机场（CRF）对路径进行评分。为了解决数据稀缺问题，我们利用大型语言模型（LLMs）生成超过200万句子。实验结果表明，我们的方法在性能上显著优于传统的基于形态分析器的方法和神经序列模型。在Joyo-Kanji-Yomi基准测试中，我们的方法达到了99.62%的目标词阅读准确率、0.32%的目标词音素错误率（PER）和0.14%的句子PER。

View on arXiv Download PDF AI Translation

cs.CL / 30 / 2609.19827

F$^{2}$DR: A Fine-Grained Full-Pipeline Reward Framework for DeepSearch Workflows

F$^{2}$DR：一种用于深度搜索工作流的细粒度全流程奖励框架

Xiong, Bojian, Ding, Wentao, Lu, Yujing, Zhang, Shaowei, Shi, Ling, Liao, Jing, Wang, Yan, Zhang, Yueyang, Xia, Long, Sun, Zhiyuan, Shi, Daiting, He, Jingzhou, Ren, Yuqi, Xiong, Deyi

Abstract

With the widespread industrial deployment of Large Language Models (LLMs), DeepSearch has emerged as the dominant paradigm for resolving complex user queries. It typically operates through an iterative closed-loop workflow consisting of planning and reflection, information retrieval, and answer generation. However, existing reward models (RMs) and evaluation benchmarks are primarily designed for static single-turn tasks, failing to capture the full-pipeline complexity of DeepSearch workflows. To address this limitation, we propose F2DR, a fine-grained full-pipeline DeepSearch reward framework. F2DR evaluates DeepSearch workflows across three dimensions: Content, Trajectory, and Answer, enabling comprehensive process-level assessment. We further construct DeepSearch RM-Bench, a dedicated benchmark for evaluating RMs in DeepSearch scenarios. Extensive experiments demonstrate that F2DR achieves significantly higher evaluation consistency than self-evaluation-based baselines, while DeepSearch RM-Bench exhibits strong discriminative capability across existing open-source RMs. We will publicly release the complete DeepSearch RM-Bench dataset soon.

Chinese Translation

随着大型语言模型（LLMs）在工业中的广泛应用，深度搜索（DeepSearch）已成为解决复杂用户查询的主流范式。它通常通过一个包含规划与反思、信息检索和答案生成的迭代闭环工作流进行操作。然而，现有的奖励模型（RMs）和评估基准主要针对静态单轮任务设计，未能捕捉深度搜索工作流的全流程复杂性。为了解决这一局限性，我们提出了F2DR，一种细粒度全流程深度搜索奖励框架。F2DR从内容、轨迹和答案三个维度评估深度搜索工作流，能够实现全面的过程级评估。我们进一步构建了深度搜索RM-Bench，这是一个专门用于评估深度搜索场景中奖励模型的基准。大量实验表明，F2DR在评估一致性方面显著高于基于自我评估的基线，而深度搜索RM-Bench在现有开源奖励模型中展现出强大的区分能力。我们将很快公开发布完整的深度搜索RM-Bench数据集。

View on arXiv Download PDF AI Translation

cs.CL / 31 / 2609.19868

Zarya: A Hybrid Autoregressive--Masked Diffusion Language Model with Flexible Training and Dual-Mode Inference

Zarya：一种具有灵活训练和双模式推理的混合自回归-掩蔽扩散语言模型

Sinev, Leonid, Koziev, Ilya, Leshchuk, Vladislav

Abstract

Autoregressive language models (ARMs) are constrained by sequential, left-to-right generation, while masked diffusion models (MDMs) enable parallel decoding but suffer from high computational overhead due to the inability to reuse Key-Value (KV) cache and from incoherent generation arising from learning dependencies over an intractable space of token combinations. We introduce Zarya, a family of hybrid language models that jointly optimizes an autoregressive (AR) objective and a masked-diffusion objective within a single architecture. Zarya structures training data into variable-size slots and employs a curriculum that gradually increases slot granularity, enabling a smooth transition from fine-grained AR learning to coarse-grained diffusion learning. At inference, Zarya provides two distinct decoding paradigms through a unified interface: (i) MDM sampling with first-hitting denoising, and (ii) slotted speculative decoding that interleaves inter-slot diffusion-based selection with intra-slot autoregressive infilling, achieving full KV cache reuse. The training and inference regimes are fully decoupled, allowing a model trained with any configuration to be deployed in either mode. Extensive configurability --- including grouped noise patterns (Prefix Completion, Fill-In-the-Prefix, Fill-In-the-Middle), ordered sampling schedules, and noise-level permutation strategies --- enables flexible research exploration. We release Zarya models publicly in sizes 0.6B, 1.7B, and 4B, demonstrating performance on standard benchmarks while offering a principled integration of autoregressive and diffusion paradigms.

Chinese Translation

自回归语言模型（ARMs）受到顺序、从左到右生成的限制，而掩蔽扩散模型（MDMs）虽然能够实现并行解码，但由于无法重用键值（Key-Value, KV）缓存而导致高计算开销，并且由于在不可处理的标记组合空间中学习依赖关系而产生不连贯的生成。我们提出了Zarya，一系列混合语言模型，它在单一架构中联合优化自回归（AR）目标和掩蔽扩散目标。Zarya将训练数据结构化为可变大小的插槽，并采用逐步增加插槽粒度的课程，使得从细粒度的AR学习到粗粒度的扩散学习之间的过渡变得平滑。在推理时，Zarya通过统一接口提供两种不同的解码范式：（i）带有首次击中去噪的MDM采样，以及（ii）插槽推测解码，它将基于插槽间扩散的选择与插槽内自回归填充交错，实现完全的KV缓存重用。训练和推理机制完全解耦，使得使用任何配置训练的模型都可以在任一模式下部署。广泛的可配置性——包括分组噪声模式（前缀补全、填充前缀、填充中间）、有序采样计划和噪声级别排列策略——使得灵活的研究探索成为可能。我们公开发布了0.6B、1.7B和4B大小的Zarya模型，展示了在标准基准上的性能，同时提供了自回归和扩散范式的原则性整合。

View on arXiv Download PDF AI Translation

cs.CL / 32 / 2609.19877

JustMem: Just-Enough Memory Access for Long-Term Conversations

JustMem：适度内存访问以支持长期对话

Chen, Guanhua, Wang, Yanting, Zhi, Wenjing, Sha, Lei

Abstract

Efficient long-term conversational memory requires retrieving sufficient evidence without indiscriminately expanding the context presented to the language model. This is challenging because relevant evidence may be distributed across multiple sessions, while compression may discard details needed for answering. Different queries therefore require different forms of memory access. To capture these demands, we formulate memory access along two dimensions: discovery breadth, which controls how broadly evidence is searched, and reading fidelity, which controls whether evidence is read in compact form or recovered from the original conversation. Based on this formulation, we introduce JustMem, which stores conversation history as compact atomic memories and adapts memory access along these two dimensions to each query. Specifically, LOOKUP handles local evidence, COMPOSE broadens discovery for distributed evidence, and REPLAY increases reading fidelity for fidelity-sensitive evidence. On LoCoMo and LongMemEval-S, JustMem achieves the highest mean accuracy and retrieval recall among the compared memory systems while using substantially fewer generative-model tokens for memory construction and inference.

Chinese Translation

高效的长期对话记忆需要在不随意扩展语言模型所呈现的上下文的情况下检索足够的证据。这一任务具有挑战性，因为相关证据可能分布在多个会话中，而压缩可能会丢弃回答所需的细节。因此，不同的查询需要不同形式的内存访问。为了捕捉这些需求，我们将内存访问形式化为两个维度：发现广度（discovery breadth），控制证据搜索的广泛程度；阅读保真度（reading fidelity），控制证据是以紧凑形式读取还是从原始对话中恢复。基于这一表述，我们引入了JustMem，它将对话历史存储为紧凑的原子记忆，并根据这两个维度调整每个查询的内存访问。具体而言，LOOKUP处理局部证据，COMPOSE扩展对分布证据的发现，REPLAY提高对对保真度敏感证据的阅读保真度。在LoCoMo和LongMemEval-S上，JustMem在比较的内存系统中实现了最高的平均准确率和检索召回率，同时在内存构建和推理中使用的生成模型令牌显著减少。

View on arXiv Download PDF AI Translation

cs.CL / 33 / 2609.19879

V\={a}kQA: A Benchmark and Evaluation Study for Telugu Spoken Factoid Question Answering

VaKQA：泰卢固语口语事实问答的基准与评估研究

Akkiraju, Bhavana, Kolluru, Ravi Sastry, D, Sri Charan, Bandarupalli, Srihari, Kesiraju, Santosh, Vuppala, Anil

Abstract

Question answering has advanced rapidly with large language models, but predominantly for high-resource languages, in both text and spoken settings. Spoken question answering (SQA) benchmark for Telugu remains unexplored, and the reliability of automatic evaluation in this setting remains unquantified. We introduce V\={a}kQA, a Telugu SQA benchmark of 2,001 factoid question-answer pairs across six domains, with 2.53 hours of speech audio, bilingual transcriptions, and human-verified reference answers. We first validate evaluation methods against human judgements: Gemini-as-a-judge best approximates human ratings but is non-uniformly strict, while open-weight judges systematically penalize correct Telugu answers that differ in surface form from the reference. Using this validated setup, we benchmark proprietary and open-weight models across input modality, language, and domain. We observe that Telugu phrasing retains cultural specificity that is lost in translation, speech input introduces phonetic confusions that alter question meaning, and cascaded ASR-MT errors compound progressively. V\={a}kQA is publicly released.

Chinese Translation

问答技术随着大型语言模型的快速发展而取得了显著进展，但主要集中在高资源语言的文本和口语环境中。泰卢固语的口语问答（SQA）基准尚未被探索，而在这一环境下自动评估的可靠性也未得到量化。我们引入了VaKQA，这是一个包含2,001个事实问答对的泰卢固语SQA基准，涵盖六个领域，配有2.53小时的语音音频、双语转录和人工验证的参考答案。我们首先根据人类评判验证评估方法：Gemini-as-a-judge最接近人类评分，但其严格程度不均匀，而开放权重评审者系统性地惩罚与参考答案在表面形式上不同的正确泰卢固语答案。利用这一经过验证的设置，我们对专有模型和开放权重模型在输入模态、语言和领域上进行了基准测试。我们观察到，泰卢固语的表述保留了在翻译中丢失的文化特异性，语音输入引入的语音混淆改变了问题的含义，而级联的ASR-MT错误逐渐累积。VaKQA已公开发布。

View on arXiv Download PDF AI Translation

cs.CL / 34 / 2609.19880

D-Quant: Driftable Entropy Coding for KV Cache Quantization

D-Quant：用于 KV 缓存量化的可漂移熵编码

Su, Yi, Liu, Hong, Yu, Guanghua, Zhu, Jianchen

Abstract

The KV cache has become a major bottleneck in deploying LLMs, as its memory footprint grows linearly with sequence length and batch size, imposing substantial pressure on both memory capacity and bandwidth. Among various KV cache compression techniques, quantization is particularly attractive due to its effectiveness and ease of deployment. However, most existing methods rely on fixed-width quantization, where a $b$ bit representation is inherently limited to $2^b$ quantization levels. As the bit width decreases, the number of available levels shrinks exponentially, leading to severe information loss and rapid performance degradation. We further observe that fixed-width quantization fails to exploit the highly non-uniform distribution of KV cache. After rotation and normalization, KV values approximately follow a normal distribution, with most values concentrated near the center and only a small fraction appearing in the tails. Nevertheless, fixed-width coding allocates the same number of bits to frequent and rare symbols. Entropy coding naturally exploits such non-uniformity by assigning shorter codewords to frequent symbols and longer ones to rare symbols, substantially reducing the average number of bits required for representation. However, its variable-length output is not suited to highly parallel attention kernels, where efficient dequantization and computation rely on regular memory layouts and fixed-stride accesses. To bridge this gap, we propose \textbf{D-Quant}, a flexible KV cache quantization framework that introduces a \textbf{drift} mechanism to convert entropy-coded representations of each token into fixed-size bitstreams, enabling regular memory access and parallel dequantization within attention kernels.

Chinese Translation

KV 缓存已成为部署大型语言模型（LLMs）的主要瓶颈，因为其内存占用随着序列长度和批量大小线性增长，给内存容量和带宽带来了巨大的压力。在各种 KV 缓存压缩技术中，量化因其有效性和易于部署而特别具有吸引力。然而，大多数现有方法依赖于固定宽度量化，其中 $b$ 位表示本质上限制在 $2^b$ 个量化级别。随着位宽的减少，可用级别的数量呈指数级缩小，导致严重的信息损失和性能迅速下降。我们进一步观察到，固定宽度量化未能利用 KV 缓存的高度非均匀分布。在旋转和归一化后，KV 值大致遵循正态分布，大多数值集中在中心，只有少量值出现在尾部。然而，固定宽度编码对频繁和稀有符号分配相同数量的位。熵编码自然利用这种非均匀性，通过为频繁符号分配较短的码字，为稀有符号分配较长的码字，从而显著减少表示所需的平均位数。然而，其可变长度的输出不适合高度并行的注意力内核，其中高效的反量化和计算依赖于规则的内存布局和固定步幅访问。为了解决这一问题，我们提出了 extbf{D-Quant}，一个灵活的 KV 缓存量化框架，引入了一种 extbf{漂移} 机制，将每个令牌的熵编码表示转换为固定大小的比特流，从而在注意力内核内实现规则内存访问和并行反量化。

View on arXiv Download PDF AI Translation

cs.CL / 35 / 2609.19883

PetriBench: Benchmarking LLM Reasoning over Dynamic State Spaces

PetriBench：动态状态空间上大语言模型推理的基准测试

Koussios, Pyrros, Jäger, Benjamin, Yao, John Hua, Sridhar, Ajay, Xiang, Violet, Li, Chenhao

Abstract

Characterizing LLM reasoning remains an open challenge, as many existing benchmarks isolate specific reasoning skills, rely on external knowledge, or are costly to extend. We introduce PetriBench, a compact, fully self-contained, and scalable benchmark for evaluating LLM reasoning over dynamic state spaces using Petri nets, a mature formalism for modeling real-world concurrent and distributed systems. PetriBench organizes reasoning into four task families varying by scope and temporal horizon, with Easy, Medium, and Hard levels generated by increasing structural complexity and evaluated against exact ground truth. Across a diverse set of proprietary and open-weight models, accuracy decreases consistently with difficulty, while harder instances expose increasingly distinct task-specific capability profiles. Additional analyses show that test-time compute improves performance but interacts differently with different reasoning tasks, and that procedural generation yields smooth scaling with structural complexity. Together, these results show that PetriBench provides a unified and extensible setting for probing the strengths, limits, and scaling behavior of LLM reasoning.

Chinese Translation

对大语言模型（LLM）推理的特征描述仍然是一个开放的挑战，因为许多现有的基准测试孤立特定的推理技能，依赖外部知识，或在扩展时成本高昂。我们引入了PetriBench，这是一个紧凑、完全自包含且可扩展的基准，用于评估LLM在动态状态空间上的推理能力，采用了Petri网这一成熟的形式化方法来建模现实世界的并发和分布式系统。PetriBench将推理组织为四个任务家族，按范围和时间视野进行变化，设有简单、中等和困难三个级别，通过增加结构复杂性生成，并与精确的真实情况进行评估。在多种专有和开放权重模型中，准确性随着难度的增加而持续下降，而更困难的实例则暴露出越来越明显的任务特定能力特征。额外分析表明，测试时的计算能力提升了性能，但与不同推理任务的交互方式不同，而过程生成则在结构复杂性上实现了平滑的扩展。综合这些结果表明，PetriBench提供了一个统一且可扩展的环境，用于探讨LLM推理的优势、局限性和扩展行为。

View on arXiv Download PDF AI Translation

cs.CL / 36 / 2609.19885

Evaluating Communicative Success in Machine-Translated Conversation

评估机器翻译对话中的交际成功

Haznitrama, Faiz Ghifari, Oh, Alice

Abstract

Interpreter agents built on machine translation (MT) increasingly mediate live conversation between people who do not share a language, yet we still evaluate them with metrics built for isolated sentences, which measure fidelity rather than whether communication succeeds. We introduce a reusable three-layer checklist-and-judge framework that evaluates interpreter-mediated conversation across semantic, pragmatic, and cultural-social dimensions, covering the naturalness, intent, and social appropriateness that fidelity metrics leave unmeasured. It runs in both single-turn and interactive multi-turn settings, where simulated users reply to translated messages as the conversation unfolds and each turn is scored alongside the conversation as a whole. We extensively validate it through controlled perturbations, cross-judge comparisons, and human annotations. Our main single-turn benchmark evaluates 10 interpreter setups across Arabic, Bengali, Indonesian, and Korean from 5,624 OpenSubtitles-derived scenarios spanning 12 translation directions, and our multi-turn study covers all 6 language pairs in scripted and live modes. Results show a consistent decline from semantic to pragmatic and cultural-social success, while conventional MT metrics overlook failures among stronger interpreters, and prompt ablations show that scenario context, structured instructions, and cultural context improve communicative success, although gains vary across setups. Our work thus provides an evaluation framework and benchmark for interpreter agents in conversation, and highlights the importance of communicative success alongside existing translation metrics.

Chinese Translation

基于机器翻译（MT）的翻译代理越来越多地在不共享语言的人之间调解实时对话，但我们仍然使用为孤立句子构建的指标来评估它们，这些指标测量的是忠实度而非交际是否成功。我们提出了一个可重用的三层检查表和评估框架，用于从语义、语用和文化社会维度评估翻译介导的对话，涵盖了忠实度指标未能衡量的自然性、意图和社会适宜性。该框架适用于单轮和交互式多轮设置，在这些设置中，模拟用户在对话展开时对翻译消息进行回复，并且每轮的评分与整个对话一起进行。我们通过受控扰动、交叉评估比较和人工注释对其进行了广泛验证。我们的主要单轮基准评估了来自阿拉伯语、孟加拉语、印尼语和韩语的10种翻译设置，涵盖了5,624个基于OpenSubtitles的场景，涉及12个翻译方向，而我们的多轮研究则涵盖了所有6对语言的脚本和实时模式。结果显示，从语义到语用和文化社会成功的表现持续下降，而传统的MT指标忽视了更强翻译者之间的失败，提示消融实验表明场景上下文、结构化指令和文化背景提高了交际成功，尽管增益在不同设置中有所不同。因此，我们的工作为对话中的翻译代理提供了评估框架和基准，并强调了交际成功的重要性，超越了现有的翻译指标。

View on arXiv Download PDF AI Translation

cs.CL / 37 / 2609.19887

Generalization through Lexical Abstraction in Transformer Models: The Case of Functional Words

通过词汇抽象实现的泛化：以功能词为例

Samo, Giuseppe, Nastase, Vivi, Merlo, Paola

Abstract

Pronouns, adverbs and other functional words (such as they, her, somewhere, there) are often used in language to replace concrete nouns or phrases, when their properties - such as gender, grammatical number - provide sufficient information for the given context. Do pretrained transformer models encode such functional words in a manner that allows them to be used like humans do? Can language models recognize the syntactic and semantic parallelism of sentences such as "The researchers wrote the paper" and "They wrote it", which relies on such lexical abstraction? We map these linguistic questions into the embedding space of a pretrained transformer model, and compare representations of nouns, with the representations of the pronouns and adverbs that can replace these nouns, in isolation and in parallel lexicalized and functional sentences. We then probe for shared syntactic and semantic structure in the embeddings of parallel lexicalized and functional sentences. We find that functional words are located centrally compared to nouns, but are also distinct, which is congruent with their behaviour as place-holders in a wide variety of contexts. The analysis of the embeddings of parallel (lexicalized and functional) sentences show them inhabiting different subspaces of the embedding space. Experiments that distil the structural information of the sentence show that training on either type of data does not reveal the shared structure - because of the over-consistency of the vocabulary (in case of the functional data), and the too much variety (in case of the lexicalized versions). However, training with a mix of functional and lexicalized sentences, the shared structure emerges.

Chinese Translation

代词、副词及其他功能词（如 they、her、somewhere、there）常用于语言中替代具体名词或短语，当其属性（如性别、语法数）为特定语境提供足够信息时。预训练的变换器模型是否以一种允许其像人类一样使用功能词的方式对这些功能词进行编码？语言模型能否识别句子“研究人员撰写了论文”（The researchers wrote the paper）与“他们撰写了它”（They wrote it）之间的句法和语义平行性，这依赖于这种词汇抽象？我们将这些语言学问题映射到预训练变换器模型的嵌入空间中，并比较名词的表示与可以替代这些名词的代词和副词的表示，分别在孤立状态和并行的词汇化及功能句中进行比较。然后，我们探讨并行词汇化和功能句的嵌入中共享的句法和语义结构。我们发现功能词相对于名词位于中心位置，但又是独特的，这与它们在各种语境中作为占位符的行为一致。对并行（词汇化和功能）句的嵌入分析显示它们占据了嵌入空间的不同子空间。对句子结构信息的提炼实验表明，单独在任一类型数据上训练并未揭示共享结构——因为功能数据的词汇过于一致，以及词汇化版本的多样性过大。然而，在功能句和词汇化句的混合训练中，共享结构得以显现。

View on arXiv Download PDF AI Translation

cs.CL / 38 / 2609.19916

KoNeoBench: A Curated Evaluation Dataset for LLM Understanding of Korean Neologisms

KoNeoBench：一个针对大型语言模型理解韩语新词的评估数据集

Lee, Soha, Lee, Soojin, Yang, Heesung, Song, Hyunju, Lee, Hyunji, An, Jinsan, Shin, Jeongwan, Park, Jin Hyun, Lee, Jun, Park, Hyeyoung, Nam, Kilim

Abstract

Large language models (LLMs) are typically evaluated on static benchmarks, even though natural language constantly evolves through newly emerging words and meanings. Existing Korean benchmarks are centered on established vocabulary and therefore provide limited coverage of such recent lexical change, and their English-oriented design makes it difficult to assess the typological properties of Korean, in which content words combine productively with functional morphemes. In this paper, we introduce KoNeoBench, a benchmark for evaluating LLMs' understanding of Korean neologisms. KoNeoBench is built on 1,785 Korean neologisms attested in online news since 2020 and curated through expert lexicographic review. Each entry provides usage examples, word-formation analyses, and dictionary-style definitions. Based on this resource, we define four tasks and report results on recent models, together with a human baseline. Our experiments show that current LLMs exhibit clear limitations in recovering source components, distinguishing semantic categories, and generating accurate definitions. These results reveal specific aspects of recent Korean lexical change that remain challenging for current LLMs. KoNeoBench is available at https://github.com/bcmilab/ko-neobench/ .

Chinese Translation

大型语言模型（LLMs）通常在静态基准上进行评估，尽管自然语言通过新出现的单词和意义不断演变。现有的韩语基准主要集中于已建立的词汇，因此对这种近期词汇变化的覆盖有限，并且其以英语为导向的设计使得评估韩语的类型特征变得困难，因为在韩语中，内容词与功能性语素能够高效结合。本文介绍了KoNeoBench，一个用于评估LLMs对韩语新词理解能力的基准。KoNeoBench基于自2020年以来在在线新闻中出现的1,785个韩语新词，并通过专家词典审查进行整理。每个条目提供用法示例、词汇构成分析和字典式定义。基于这一资源，我们定义了四个任务，并报告了近期模型的结果，以及人类基线的表现。我们的实验显示，当前的LLMs在恢复源成分、区分语义类别和生成准确定义方面存在明显的局限性。这些结果揭示了当前LLMs在应对近期韩语词汇变化时所面临的具体挑战。KoNeoBench可在 https://github.com/bcmilab/ko-neobench/ 获取。

View on arXiv Download PDF AI Translation

cs.CL / 39 / 2609.19942

Intrinsic Sequence-Likelihood Confidence in Retrieval-Dominated Extractive QA: Two Pre-Specified Negatives, and What They Do and Do Not Attribute

检索主导的抽取式问答中的内在序列似然置信度：两个预设负样本及其归因分析

Lee, Gunwoo, Sung, Changmin, Gwak, Sang-Hwan, Kim, Ina, Choi, Ji-Young, Lee, Kyong-Ha

Abstract

In extractive document question answering whose questions were generated from the passages that contain their answers -- so that retrieval recovers 92-99.8% of what any mode combination could reach, whatever its absolute accuracy -- confidence-driven mechanisms have little to gain. Fine-tuning an open language model on a specialized domain corpus yields a model whose own confidence is a tempting control signal: it could decide which queries warrant further adaptation, and which answers to trust. We evaluate both uses under criteria fixed before the runs were executed, across four 7-9B model families whose adaptation moved closed-book F1 by at most +0.03, and both fail: a distillation trigger on all four families, under its pre-specified three-step transfer budget, and a routing-and-abstention policy in its single-model pilot. Retrieval alone recovers 92-99.8% of best-case combined accuracy under every correctness criterion we test, leaving routers no meaningful gain. The sequence-likelihood signal is insufficient relative to that mode -- area under the receiver operating characteristic curve 0.65-0.81 under the registered criterion -- before adaptation as well as after, unchanged by scalar recalibration and not consistently improved by token-level temperature rescaling. And the finer diagnostics depend on the correctness criterion and on answer length; on the three adapted combinations where we could test it, selector ablations show no statistically detectable downstream benefit from the confidence term on any seed; on Gemma, removing it changes the selector from failing to passing both registered criteria. The usable product is a set of pre-specified negatives with their dependencies made explicit.

Chinese Translation

在抽取式文档问答中，问题是从包含答案的段落生成的——因此检索能够恢复92-99.8%的最佳模式组合所能达到的内容，无论其绝对准确性如何——置信度驱动的机制收效甚微。在专门领域语料库上微调开放语言模型，得到的模型自身置信度成为一种诱人的控制信号：它可以决定哪些查询值得进一步适应，哪些答案值得信任。我们在执行之前固定的标准下评估这两种用途，涉及四个7-9B模型系列，其适应性使闭卷F1最多提高+0.03，但两者均未成功：在其预设的三步转移预算下，四个系列均出现了蒸馏触发，而在单模型试点中则出现了路由和放弃政策。仅靠检索在我们测试的每个正确性标准下恢复了92-99.8%的最佳组合准确性，未能为路由器带来有意义的收益。相较于该模式，序列似然信号不足——在注册标准下接收器操作特征曲线下面积为0.65-0.81——在适应之前和之后均未改变，且通过标量重新校准并未得到一致改善，令令牌级温度重标定也未能改善。更细致的诊断依赖于正确性标准和答案长度；在我们能够测试的三种适应组合中，选择器消融未显示出置信度项对任何种子具有统计上可检测的下游益处；在Gemma上，去除该项使选择器从未通过变为通过两个注册标准。可用的产品是一组预设负样本，其依赖关系被明确化。

View on arXiv Download PDF AI Translation

cs.CL / 40 / 2609.19965

Before the Arrest: Benchmarking LLMs on Criminal Profiling from Incomplete Evidence

逮捕之前：基于不完整证据的犯罪画像大语言模型基准测试

Yao, Yutong, Cao, Yanjie, Chen, Guanhua, Yang, Xu, Wu, Junchao, Wu, Zeyu, Chao, Lidia S., Wong, Derek F.

Abstract

Large Language Models (LLMs) are increasingly applied to legal and criminal justice tasks, yet existing work focuses almost exclusively on post-arrest scenarios where the suspect's identity is already known, leaving the critical pre-arrest challenge of inferring suspect characteristics from incomplete evidence largely unexplored. To fill this gap, we introduce the Profiling, Investigation, and Judgment (PIJ), comprising 2,500 real homicide cases from five countries. PIJ evaluates LLMs across three tasks that span the entire criminal investigation pipeline: criminal profiling, which requires abductive reasoning to infer suspect attributes from fragmentary scene evidence, crime process reconstruction, which tests structured information extraction, and sentence prediction, which demands legal deductive reasoning. We evaluate 9 powerful LLMs and find that performance degrades systematically as tasks shift from explicit fact extraction to implicit reasoning over unknown suspect profiles. Categories requiring inferential reasoning, such as motivation and victim-offender relationships, remain the primary bottlenecks. Further analysis reveals substantial gaps between LLMs and human experts, along with pervasive biases in gender, age, and motive attribution. Our findings indicate that pre-arrest inference from incomplete evidence remains an open challenge.

Chinese Translation

大型语言模型（LLMs）在法律和刑事司法任务中的应用日益增多，但现有研究几乎完全集中于逮捕后的情境，在这些情境中，嫌疑人的身份已经确定，而在逮捕前从不完整证据中推断嫌疑人特征的关键挑战却鲜有探讨。为填补这一空白，我们引入了犯罪画像、调查与判断（Profiling, Investigation, and Judgment, PIJ）数据集，该数据集包含来自五个国家的2500个真实谋杀案件。PIJ评估LLMs在涵盖整个刑事调查流程的三项任务中的表现：犯罪画像，这需要通过溯因推理从片段性现场证据推断嫌疑人属性；犯罪过程重建，这测试结构化信息提取；以及判决预测，这要求法律演绎推理。我们评估了9个强大的LLMs，发现随着任务从明确的事实提取转向对未知嫌疑人画像的隐含推理，性能系统性下降。需要推理的类别，如动机和受害者-犯罪者关系，仍然是主要瓶颈。进一步分析揭示了LLMs与人类专家之间存在显著差距，以及在性别、年龄和动机归属方面的普遍偏见。我们的研究结果表明，从不完整证据中进行逮捕前推理仍然是一个未解决的挑战。

View on arXiv Download PDF AI Translation

cs.CL / 41 / 2609.19969

DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression

DeepSeek-V4.1-Flash：推动KV缓存压缩的极限

DeepSeek-AI, :, Xu, Anyi, Li, B., Lin, Bangcai, Xue, Bing, Xian, BingCheng, Xu, Bingzheng, Wu, Bochao, Zhang, Bowei, Deng, Boyi, Yu, C. C., Jin, Chao, Lin, Chaofan, Dong, Chen, Wang, Chenbing, Feng, Chenfan, Lu, Chengda, Zhao, Chenggang, Deng, Chengqi, Zhang, Chengyuan, Xu, Chenhao, Zhao, Chenqi, Shao, Chenze, Wang, Chuhao, Zhang, Chuqi, Dai, Damai, Yang, Dejian, Chen, Deli, Huang, Di, Wu, Di, Li, Donghao, Li, Erhang, Fu, Eric, Zhou, F., Zhou, Fangwei, Lin, Fangyun, Yuan, Fangzhou, Xia, Feiyu, Dai, Fucong, Hao, Guangbo, Li, Guanglin, Chen, Guanting, Cao, Guoai, Fan, Guofan, Meng, Guolai, Li, Guowei, Zhang, Haichuan, Ma, Haiyang, Shen, Haiyang, Li, Han, Yu, Han, Zhang, Han, Deng, Hangyuan, Xu, Hanwei, Xu, Hanxiang, Zhong, Hanxun, Guo, Hao, Jiang, Hao, Li, Hao, Qin, Hao, Wen, Haodong, Liang, Haofen, Huang, Haofeng, Liu, Haohua, Zhang, Haoling, Luo, Haoming, Yang, Haoran, Xu, Haotian, Yuan, Haotian, Huang, Haoting, Luo, Haowen, Cai, Haoyang, Chen, Haoyu, Ji, Haozhe, Zhang, Hengran, Wang, Hengrui, Wu, Hengxu, Ding, Honghui, Tang, Hongxuan, Wang, Huadong, Cao, Huanqi, Gao, Huazuo, Qu, Hui, Zeng, Hui, Yang, J., Jin, J. H., Zhang, J. H., Zou, J. X., Yu, Jia, Zhou, Jiahui, Chen, Jiajun, Huang, Jialiang, Zhao, Jialin, Tang, Jiamin, Zhou, Jian, Tong, Jianan, Li, Jianwen, Zhu, Jiaqi, Wang, Jiarui, Ye, Jiasheng, Li, Jiashi, Xu, Jiaxin, Ding, Jiaying, Lu, Jibai, Hu, Jiewen, Yan, Jin, Zhai, Jincheng, Chen, Jingchang, Hu, Jingcheng, Zhou, Jingli, Xu, Jingsheng, Xiang, Jingting, Yun, Jingyan, Yuan, Jingyang, Cheng, Jingyuan, Zhu, Jinhua, Wang, Jinpeng, Chen, Jinyi, Hu, Jinyi, Yu, Jiping, Guo, Jueliang, Pei, Junbo, Sun, Junbo, Jiang, Junguang, Qiu, Junjie, Zhou, Junkang, Liu, Junqi, Li, Junren, Li, Junxian, Song, Junxiao, Guo, Junyi, Dong, Kai, Chen, Kaifeng, Gao, Kaige, Guan, Kang, Yuan, Kangdong, Hong, Ke, Xu, Ke, Zhao, Kefan, Ji, Kexin, Zhang, Kexin, Zhou, Kexing, Yu, Kuai, Zhang, Lan, Wang, Lean, Zhang, Lecong, Wang, Lei, Gao, Letian, Zhao, Liang, Xu, Liansheng, Guo, Lihua, Luo, Lingxiao, Fu, Lingyue, Deng, Litao, Wang, Litong, Zhang, Liyue, Chen, Longhao, Chen, Lu, Huang, Luotian, Ma, Luyao, Wang, Luyao, Di, M. S., Mei, Max, Ye, Menghao, Cui, Miao, Zhang, Mingchuan, Zhang, Minghua, Tang, Minghui, Zhang, Mingjing, Wei, Mingqi, Chen, Mingshu, Liu, Mingxing, Zhou, Mingxu, Xu, Mingyu, Yang, Mingyu, Wang, Mingze, Chen, Muyang, Shentu, Ni, Wang, Ning, Ning, Niufang, Huang, Panpan, Cong, Peixin, Wang, Peiyi, Xin, Peiyuan, Ren, Pengfei, Yan, Pengfei, Zhang, Pengle, Kang, Qi, Tang, Qi, Wang, Qiancheng, Li, Qiang, Zhu, Qihao, Li, Qingyang, Chen, Qinyu, Du, Qiushi, Guo, Qizhou, Xu, Rongxian, Ding, Rui, Hu, Rui, Tian, Rui, Yu, Rui, Zhu, Ruidong, Xu, Ruifan, Yang, Ruihan, Xia, Ruihang, Lu, Ruijie, Geng, Ruilin, Hong, Ruipeng, Ge, Ruiqi, Zhang, Ruisong, Sun, Ruize, Pan, Ruizhe, Wang, Runji, Chen, Runqian, Xu, Runxin, Tian, Ruohong, Shen, Ruomeng, Zhang, Ruoyu, X., Ryan, Liu, S. H., Lu, Shanghao, Zhou, Shangyan, Chen, Shanhuang, Cai, Shaofei, Nie, Shaoheng, Chen, Shaoyuan, Hu, Shengding, Lin, Shengkai, Ran, Shengwen, Liu, Shengyu, Jia, Shengyuan, Bai, Shi, Feng, Shi, Xu, Shicheng, Liu, Shichun, Hu, Shiqiang, Ma, Shirong, Wang, Shiyu, Feng, Shiyuan, Gong, Shufan, Lin, Shuhan, Yu, Shuiping, Zhou, Shunfeng, Yang, Shuo, Wang, Shuomeng, Guo, Shuting, Pan, Shuting, Yu, Shuying, Cao, Sinuo, Lin, Siyi, Chen, Sizhe, Chen, Songyang, Zhou, Songyang, Ni, Tao, Yun, Tao, Jin, Tian, Pei, Tian, Ye, Tian, Lin, Tianle, Ji, Tianran, Cui, Tianyi, Yue, Tianyuan, Yu, Tingting, Xiong, Tongrui, Zeng, Wangding, Liu, Wei, Zhang, Wei, Xu, Weibin, Zeng, Weihao, Zhao, Weilin, Liu, Wen, Liang, Wenfeng, Pang, Wenjie, Luo, Wenjing, Yao, Wenjing, Gao, Wenjun, Shao, Wenkai, Yang, Wenkai, Zhang, Wenli, Wang, Wenlu, Huang, Wenlve, Yan, Wenqian, Zhang, Wentao, Gao, Xi, He, Xiang, Li, Xiang, Li, Xiangli, Wang, Xiangwen, Zhang, Xiangying, Wei, Xiankui, Bi, Xiao, Liu, Xiaodong, Wang, Xiaohan, Qu, Xiaojian, Chen, Xiaokang, Zhang, Xiaokang, Nie, Xiaotao, Zou, Xiaoyao, Li, Xiaoyuan, Guo, Xicheng, Chu, Xieting, Cheng, Xin, Liu, Xin, Xie, Xin, Xu, Xinbo, Liu, Xingchao, Liu, Xingchen, Yu, Xingkai, Li, Xingyou, Yao, Xintong, Chen, Xinyang, Jiang, Xinyong, Yang, Xinyu, Yang, Xinyu, Chen, Xu, Wang, Xuanyu, Zhong, Xubei, Su, Xuecheng, Liu, Xuejie, Lin, Xuheng, Fan, Xujie, Zhao, Xuncheng, Fu, Xuwei, Yan, Y. C., Jiang, Y. H., Wu, Y. T., M., Y. W., Wang, Y. Z., Gao, Yafei, Yang, Yang, Zhang, Yang, Ma, Yanru, Huang, Yanwen, Li, Yao, Li, Yao, Meng, Yao, Zhao, Yao, Sun, Yaofeng, Wang, Yaohui, Ye, Yaoyang, Yin, Yehang, Wu, Yexinrui, Qian, Yi, Tao, Yi, Yu, Yi, Zhang, Yichao, Jiang, Yichen, Wang, Yicheng, Ding, Yifan, Shi, Yifan, Peng, Yifeng, Zhai, Yifeng, Wu, Yijia, Xiong, Yiliang, Wang, Yilun, He, Ying, Zhou, Ying, Luo, Yingjia, Zhong, Yinmin, Wang, Yiping, Wang, Yisong, Zhang, Yixiang, Chen, Yixiao, Tan, Yixuan, Wei, Yixuan, Ma, Yiyang, Yang, Yiyao, Liu, Yiyuan, Cai, Yizai, Wei, Yizhen, Wang, Yizhi, Yang, Yonglun, Zhuo, Yongqi, Guo, Yongqiang, Wu, Yongtong, Wu, Yu, Zhang, Yu, Bian, Yuan, Cheng, Yuan, Ou, Yuan, Sun, Yuan, Xu, Yuanfan, Sun, Yuanhang, Li, Yuanhao, Liu, Yuchen, Yao, Yuchen, Han, Yudong, Wang, Yuduan, Wu, Yuhan, Meng, Yuhao, Zou, Yuheng, Li, YuKun, Wang, Yunchuan, Xiao, Yunfan, Xiong, Yunfan, Chen, Yupeng, Cao, Yuqian, Wang, Yuqian, Chen, Yuqing, Zhang, Yushun, Lin, Yutong, Xiao, Yuwei, Gu, Yuxian, Chen, Yuxiang, Huang, Yuxiang, Luo, Yuxiang, You, Yuxiang, Chen, Yuxin, Xiang, Yuxin, Liu, Yuxuan, Zhou, Yuxuan, Zhou, Yuyang, Guo, Yuzhe, Huang, Yuzhen, Bai, Yuzhuo, Z., Z. Y., Ni, Zanlin, Wang, Zehao, Zhao, Zehua, Ren, Zehui, Zhao, Zejun, Sha, Zhangli, Wang, Zhanying, Zhang, Zhaochen, Du, Zhaoshuai, Fu, Zhe, Xu, Zhean, Xie, Zhenda, Liu, Zheng, Zhang, Zhengyan, Dong, Zhenhua, Hao, Zhewen, Wang, Zhibang, Gou, Zhibin, Ma, Zhicheng, Li, Zhihao, Shao, Zhihong, Huang, Zhihuan, Li, Zhijie, Lu, Zhirui, Huang, Zhixian, Chen, Zhixuan, Chen, Zhixuan, Pan, Zhixuan, Wu, Zhiyu, Ren, Zhizhou, He, Zhu, Li, Zhuoshu, Zhang, Zhuping, Xu, Zian, Wang, Zihao, Gu, Zihui, Zhu, Zijia, Zhang, Zili, Li, Zilin, Hou, Zilong, Lyu, Zilong, Wang, Ziqiao, Xie, Ziwei, Zhang, Ziya, Gao, Ziyi, Pan, Zizheng, Li, Zonglin, Yao, Zongqing, Chen, Zui, Wu, Zuofan, Ling, Chenchen, Hou, Chengyu, Chen, Chong, Li, D., Qi, Di, Ji, Dongjie, Wei, Fang, Xia, Fanyi, Xie, Fei, Tan, Feiyi, Guo, Hailong, Zhai, Haiyan, Zhou, Hui, Tan, Huihui, Li, Huijie, Luo, Jia, Song, Jia, Cai, Jialu, Liang, Jian, Zhou, Jiangting, Gao, Jiaqi, Shao, Jiayi, Chen, Jie, Yang, Jieyu, Chen, Jin, Zhang, Jingde, Zhou, Jingzi, Wang, Jinqian, Liu, Jinyang, Sun, JinZhao, Ling, Junhua, Zheng, Junmin, Yang, Kaicheng, Xu, Ke, Su, Le, Xia, Leyi, Ding, Liangfeng, Zhuo, Lin, Ma, Linwang, Zhu, Linyan, Cai, Liyu, Yao, Luqi, Zhang, M. K., Li, Meng, Lin, Miao, Wang, Miaojun, Zhang, Min, Li, Mingming, Wang, Mingming, Yin, Mingze, Han, Minmin, Cao, Nan, Wang, Ning, Ma, Ningxin, Wang, Panpan, Lin, Peihan, Sun, Peng, Zhang, Peng, Ying, Qian, Xiang, Qiang, Wang, Qiao, Mao, Qingmiao, Jiang, Qiwei, Jin, Rongli, Chen, Ruyi, Tao, Sha, Sun, Shangmian, Wu, Shaoqing, Zou, Shichao, Lei, Si, Zhang, Tianyang, Sun, Tianyu, Yin, Tingting, Xiao, W. L., An, Wei, Li, Wei, Wang, Wei, Lin, Weiwei, Hou, Wenqing, Lin, X., Meng, Xiangfei, Huang, Xianzhu, Peng, Xiao, Li, Xiaoqian, Zhang, Xiaoting, Sun, Xiaowen, Wang, Xiaoxiang, Ye, Xiaoyu, Zhang, Xinrou, Zhang, Xinyu, Cao, Xue, Chen, Xueyin, Zhou, Yanan, Xu, Yanhong, Xia, Yao, Xu, Yao, Shao, Yi, Zhang, Yihong, Ma, Yiling, Tang, Ying, Lou, Yining, Chen, Yiru, Piao, Yishi, Chen, Yixuan, Xiong, Yong, Xuan, Yuchen, Yang, Yuehan, Xu, Yuer, Zha, Yukun, Ma, Yunxian, Lin, Yuping, Yan, Yuting, Xie, Yutong, Sheng, Yuwen, Zhu, Yuxuan, Zhang, Zekai, Ju, Zhe, Lin, Zhenzhen, Gao, Zheren, Sun, Zheyang, Yan, Zhigang, Wu, Zhongyu, Wang, Zi, Qu, Zihua, Yan, Ziling, Wan, Ziyi

Abstract

The widespread adoption of long-horizon agents has made model workloads increasingly input-heavy. Although prior work has substantially reduced the cost of long-context computation, prefill remains computationally expensive, and large KV caches continue to strain HBM and SSD capacity and data-transfer bandwidth. Together, these compute, storage, and bandwidth demands constitute the primary bottleneck to further lowering deployment costs. To address this challenge, we introduce DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts (MoE) model with 552B backbone parameters and support for contexts of up to one million tokens. With its Causal Encoder-Decoder (CED) architecture, the model activates 16B parameters per token during decode but only 8B parameters during prefill, substantially improving cost efficiency for agentic workloads. To push the limits of KV cache compression, DeepSeek-V4.1-Flash combines cross-layer KV cache reuse in Compressed Sparse Attention 2 (CSA2) with FP4 KV caching. These designs reduce its global KV cache footprint (always in HBM) to 890 bytes per token, roughly 1/4 of the corresponding footprint of DeepSeek-V4-Flash. Further, through a dedicated deployment optimization known as SWA Bounded Replay, DeepSeek-V4.1-Flash reduces its persistent KV cache footprint (always on SSD or in host memory) to roughly 1/8 of that of DeepSeek-V4-Flash. Despite its much smaller KV cache footprint, the model delivers substantially better performance than the baseline. In addition, we streamline the DeepSeek-V4 architecture and introduce several efficient architectural extensions. We pretrain DeepSeek-V4.1-Flash on a multimodal corpus comprising 45T tokens and conduct comprehensive post-training, yielding strong performance across diverse text-based and multimodal agentic scenarios. Model checkpoints are available at https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash.

Chinese Translation

长时间跨度代理的广泛应用使得模型工作负载越来越依赖输入。尽管之前的研究已经显著降低了长上下文计算的成本，但预填充仍然计算开销较大，而大型KV缓存继续对高带宽内存（HBM）和固态硬盘（SSD）的容量及数据传输带宽造成压力。这些计算、存储和带宽需求共同构成了进一步降低部署成本的主要瓶颈。为了解决这一挑战，我们提出了DeepSeek-V4.1-Flash，这是一种具有552B主干参数的多模态专家混合模型（Mixture-of-Experts, MoE），支持高达一百万个标记的上下文。该模型采用因果编码器-解码器（Causal Encoder-Decoder, CED）架构，在解码过程中每个标记激活16B参数，而在预填充过程中仅激活8B参数，显著提高了代理工作负载的成本效率。为了推动KV缓存压缩的极限，DeepSeek-V4.1-Flash将压缩稀疏注意力2（Compressed Sparse Attention 2, CSA2）中的跨层KV缓存重用与FP4 KV缓存相结合。这些设计将其全球KV缓存占用（始终在HBM中）减少到每个标记890字节，约为DeepSeek-V4-Flash相应占用的1/4。此外，通过一种称为SWA Bounded Replay的专用部署优化，DeepSeek-V4.1-Flash将其持久KV缓存占用（始终在SSD或主机内存中）减少到约DeepSeek-V4-Flash的1/8。尽管其KV缓存占用大大减少，该模型的性能仍显著优于基线。此外，我们简化了DeepSeek-V4架构，并引入了几种高效的架构扩展。我们在一个包含45T标记的多模态语料库上对DeepSeek-V4.1-Flash进行了预训练，并进行了全面的后训练，取得了在多样的基于文本和多模态的代理场景中的强大性能。模型检查点可在https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash获取。

View on arXiv Download PDF AI Translation

cs.CL / 42 / 2609.19989

Benchmarking LLM Compliance with China AI Generated Content Regulations

大型语言模型遵循中国人工智能生成内容法规的基准评估

Cui, Chenrui, Fang, Hongye, Song, Lisha, Chen, Weichao, Zhu, Yue, Xu, Gang

Abstract

The widespread adoption of LLMs has led to escalating content compliance risks. Prior works have contributed to addressing these risks in the English context, downplaying the complexity of Chinese language content. This paper follows China's current AI-Generated content compliance requirements and provides evaluation results on 20 notable LLMs, offering insight into China's regulatory landscape. We design a novel framework to assess the compliance and refusal rates with 2303 questions spanning six distinct dimensions, including 203 self-constructed constitutional questions. The framework employs several judges to generate verdicts independently based on their hierarchical alignment memory. Our findings show that international models also exhibit high levels of compliance despite the use of standard Chinese questions, and the main differences may stem from dimensions closely related to ideological alignment. We establish a regulatory benchmark that enables the global AI community to evaluate both Chinese and non-Chinese LLMs under a unified set of legally grounded compliance requirements.

Chinese Translation

大型语言模型（LLMs）的广泛应用导致内容合规风险不断上升。以往的研究主要集中在英语环境下应对这些风险，而忽视了中文内容的复杂性。本文遵循中国当前的人工智能生成内容合规要求，对20个显著的LLM进行了评估，提供了对中国监管环境的洞察。我们设计了一个新颖的框架，以评估2303个问题在六个不同维度上的合规和拒绝率，其中包括203个自构建的宪法问题。该框架采用多个评审员独立生成裁决，基于他们的层次对齐记忆。我们的研究结果表明，尽管使用标准中文问题，国际模型也表现出高水平的合规性，主要差异可能源于与意识形态对齐密切相关的维度。我们建立了一个监管基准，使全球人工智能社区能够在统一的法律合规要求下评估中国和非中国的LLM。

View on arXiv Download PDF AI Translation

cs.CL / 43 / 2609.20081

Reading Emotions in the Token Space: Discriminative Adaptation of SpeechLLMs for Emotion Recognition

在标记空间中读取情感：SpeechLLMs的情感识别判别适应

Watawana, Hasindri, Burdisso, Sergio, Villatoro-Tello, Esaú, E, Manjunath K, Hacioglu, Kadri, Motlicek, Petr, Stolcke, Andreas

Abstract

SpeechLLMs have shown strong potential for emotion recognition, yet they read the predicted emotion off a generative decoder not suited for classification: it can emit labels outside the target set and favors frequent classes. We propose a discriminative adaptation that reads the final prompt token's hidden state through a classification head, producing a label in one forward pass without modifying the backbone. Because this readout starts from the hidden state the model would otherwise decode, it gives a controlled comparison of generative and discriminative inference in an otherwise identical speechLLM. We keep the head a single linear layer, trading little accuracy for interpretability: each emotion becomes one direction in the LLM output token space, revealing associated tokens. On IEMOCAP, across two speechLLM architectures, it improves Macro F1 and removes hallucinations, with largest gains on realistic ASR transcripts. Our analysis reveals that these emotion directions encode indirect associations mirroring biases in web-scale text.

Chinese Translation

SpeechLLMs在情感识别方面展现出强大的潜力，但它们通过不适合分类的生成解码器读取预测的情感：该解码器可能发出超出目标集合的标签，并偏向于频繁的类别。我们提出了一种判别适应方法，通过分类头读取最终提示标记的隐藏状态，在一次前向传播中生成标签，而无需修改主干网络。由于这种读取是从模型本应解码的隐藏状态开始的，它在其他条件相同的speechLLM中提供了生成推理和判别推理的受控比较。我们将头部设计为单一线性层，牺牲了少量准确性以换取可解释性：每种情感在LLM输出标记空间中成为一个方向，揭示相关的标记。在IEMOCAP数据集上，针对两种speechLLM架构，它提高了宏观F1值并消除了幻觉，尤其在真实的ASR转录中获得了最大的提升。我们的分析表明，这些情感方向编码了间接关联，反映了网络规模文本中的偏见。

View on arXiv Download PDF AI Translation

cs.CL / 44 / 2609.20104

Design of the IBM Granite 5.0 TurboCTC ASR Model

IBM Granite 5.0 TurboCTC ASR模型设计

Kingsbury, Brian, Saon, George, Suzuki, Masayuki, Kuo, Hong-Kwang J., Fukuda, Takashi, Thomas, Samuel, Sunder, Vishal, Dekel, Avihu

Abstract

We describe the architecture, training methodology and inference speedups of Granite 5.0 Turbo CTC, a 470 million parameter encoder-only model with an excellent speed-accuracy tradeoff. The architecture uses pyramidal temporal subsampling within Conformer blocks using strided depthwise convolutions, block-diagonal (chunk-wise) self-attention, and conditioning on intermediate predictions from the middle layer. Training highlights are the use of only publicly available data, the novel use of a Muon optimizer, and balanced data sampling. Inference speedups include replacing 1 x 1 convolutions with linear layers and optimizing the attention computation in the Conformer blocks. Collectively, these result in a model that is on the speed-accuracy Pareto frontier of the Open ASR leaderboard for English short-form ASR while being twice as fast as the fastest competitor. The model can be used under a permissive license and downloaded from https://huggingface.co/ibm-granite/granite-speech-5.0-470m-turboctc.

Chinese Translation

我们描述了Granite 5.0 Turbo CTC的架构、训练方法和推理加速，该模型为470百万参数的仅编码器模型，具有出色的速度-准确性权衡。该架构在Conformer块中使用金字塔时间子采样，采用步幅深度卷积、块对角（分块）自注意力机制，并基于中间层的预测进行条件处理。训练亮点包括仅使用公开可用的数据、创新性地使用Muon优化器以及平衡的数据采样。推理加速方面包括用线性层替代1 x 1卷积，并优化Conformer块中的注意力计算。综合这些因素，模型在英语短语ASR的Open ASR排行榜上处于速度-准确性帕累托前沿，并且速度是最快竞争者的两倍。该模型可在宽松许可下使用，并可从https://huggingface.co/ibm-granite/granite-speech-5.0-470m-turboctc下载。

View on arXiv Download PDF AI Translation

cs.CL / 45 / 2609.20169

Fine-Tuning Models for Biomedical Relation Extraction

生物医学关系提取的模型微调

Creanga, Claudiu, Dinu, Liviu P., Gifu, Daniela

Abstract

Next-Generation Sequencing has revolutionized the study of genetic mutations, enabling large-scale investigations into their roles in disease development. However, extracting meaningful insights from the vast amount of biomedical literature remains a complex challenge that cannot be addressed manually. In this paper, we present pre-trained models (PTMs) for the automatic extraction of relations from biomedical text, specifically targeting the variant-phenotype domain. Our evaluation on the SNPPhenA corpus demonstrates that fine-tuning small BERT-based models, particularly DeBERTa, yields strong performance, approaching the current state-of-the-art (SOTA). Additionally, our results indicate that carefully fine-tuning Google's Gemini Pro 1.0 outperforms the existing SOTA for both sentence-level tasks (where the model processes only the target sentence) and abstract-level tasks (where the model processes the entire abstract).

Chinese Translation

下一代测序技术彻底改变了对基因突变的研究，使得能够大规模调查其在疾病发展中的作用。然而，从大量生物医学文献中提取有意义的见解仍然是一个复杂的挑战，无法通过人工方式解决。本文提出了用于自动提取生物医学文本中关系的预训练模型（PTMs），特别针对变异-表型领域。我们在SNPPhenA语料库上的评估表明，微调小型基于BERT的模型，特别是DeBERTa，能够取得强劲的性能，接近当前的最先进水平（SOTA）。此外，我们的结果表明，经过精心微调的谷歌Gemini Pro 1.0在句子级任务（模型仅处理目标句子）和摘要级任务（模型处理整个摘要）中均优于现有的SOTA。

View on arXiv Download PDF AI Translation

cs.CL / 46 / 2609.20186

To Copy or Not to Copy: Controlling Speculative Decoding via Intrinsic Model Signals

复制还是不复制：通过内在模型信号控制推测解码

Eisenstadt, Roy, Cohen, Ido, Cohen-Karlik, Edo, Wolf, Lior, Zimerman, Itamar

Abstract

Speculative Decoding (SD) has significantly accelerated Large Language Model (LLM) inference, yet existing approaches face a fundamental tradeoff between two drafting strategies: neural drafting and context-based copying. Neural drafts (e.g., EAGLE3) provide robust performance across diverse text settings, while copy-based methods achieve higher speedups in copy-intensive regimes by generating candidates faster and exploiting long repetition spans for near-perfect speculation. We analyze existing copy-based methods and find that they are prone to accidental repetitions where surface-level n-gram overlap does not reflect a structural intent to copy, leading to false-positive triggers that ultimately degrade throughput. We introduce SwitchSD, an adaptive framework that treats copying as a latent control signal of the LLM. By training lightweight probes on the target model's internal representations, SwitchSD identifies genuine copy-intent with high precision (AUC > 0.99). This allows the system to dynamically switch between neural drafting (e.g., EAGLE) and context-based copying. Our results across Llama and Qwen families demonstrate throughput gains of up to 15% over state-of-the-art baselines like EAGLE3, effectively turning copying from a noisy heuristic into a principled, model-aware decoding regime.

Chinese Translation

推测解码（Speculative Decoding, SD）显著加速了大型语言模型（Large Language Model, LLM）的推理过程，但现有方法在两种草拟策略之间面临基本的权衡：神经草拟和基于上下文的复制。神经草拟（例如，EAGLE3）在多样化文本环境中提供了稳健的性能，而基于复制的方法在复制密集型场景中通过更快地生成候选项和利用长重复跨度实现了更高的加速。我们分析了现有的基于复制的方法，发现它们容易出现意外重复的情况，其中表面层的 n-gram 重叠并未反映出复制的结构意图，导致错误的触发，从而最终降低了吞吐量。我们提出了 SwitchSD，这是一种自适应框架，将复制视为 LLM 的潜在控制信号。通过在目标模型的内部表示上训练轻量级探针，SwitchSD 以高精度（AUC > 0.99）识别真实的复制意图。这使得系统能够在神经草拟（例如，EAGLE）和基于上下文的复制之间动态切换。我们在 Llama 和 Qwen 系列上的结果表明，与 EAGLE3 等最先进基线相比，吞吐量提升可达 15%，有效地将复制从一种噪声启发式转变为一种有原则的、模型感知的解码机制。

View on arXiv Download PDF AI Translation

cs.CL / 47 / 2609.20207

Foundations of Stochastic Lexical Calculus: Semantic Descent and Random Dynamics on Probability Simplices

随机词汇演算的基础：概率单纯形上的语义下降与随机动态

Dixon, Matthew F

Abstract

Large language models produce prompt-dependent probabilities over words, whereas scientific systems require uncertainty over meaningful states that can be updated as evidence arrives. We develop an observable framework for determining when language-derived probabilities support such a sequential state representation. Theoretically, we define typed measurable transformations of contextual language, construct a minimal closed representation, and give necessary and sufficient conditions for semantic updates to exist uniquely. We bound irreducible nonclosure and accumulated error, and under average contraction prove existence, uniqueness and stability of an external random recursion on a probability simplex. These results define a stochastic lexical calculus without attributing an internal calculus to the language model. Empirically, frozen experiments test the observable implications. Raw prompt-conditioned probabilities fail the prespecified invariance gate; after prompt-specific calibration, a common three-state representation passes the stability gates and covers 28 of 30 untouched eight-step paths, or 0.933 at nominal level 0.90. Accordingly, language probabilities support a stochastic state only conditionally on verified closure, stability and coverage within a declared operating domain.

Chinese Translation

大型语言模型生成依赖于提示的词汇概率，而科学系统则需要对有意义的状态的不确定性，这些状态可以随着证据的到来而更新。我们开发了一个可观察的框架，用于确定何时语言衍生的概率支持这种顺序状态表示。从理论上讲，我们定义了上下文语言的类型可测变换，构建了一个最小闭合表示，并给出了语义更新唯一存在的必要和充分条件。我们界定了不可约非闭合和累积误差，并在平均收缩的情况下证明了概率单纯形上外部随机递归的存在性、唯一性和稳定性。这些结果定义了一种随机词汇演算，而不将内部演算归因于语言模型。在实证方面，冻结实验测试了可观察的影响。原始的提示条件概率未能通过预设的不变性门；经过提示特定的校准后，一个通用的三状态表示通过了稳定性门，并覆盖了30条未触及的八步路径中的28条，或者在名义水平0.90下达到0.933。因此，语言概率仅在经过验证的闭合、稳定性和覆盖范围内的声明操作域条件下支持随机状态。

View on arXiv Download PDF AI Translation

cs.CL / 48 / 2609.20223

Before the Warning Comes Too Late: Incremental Phone-Scam Detection from Speech

警告来得太晚之前：基于语音的增量电话诈骗检测

Vo, Khang Nhat Hoang, Dinh, Anh Trac Duc, Ta, Tai Tien, Quan, Tho

Abstract

We study weakly supervised incremental telecom fraud detection from raw telephone audio, where training provides only conversation-level labels and predictions must be updated before a call ends. We introduce StreamFraudNet, which processes incoming audio through overlapping bounded-context windows using a frozen self-supervised speech encoder, recurrent temporal modeling, and learned aggregation of latent window scores. On a controlled English benchmark, StreamFraudNet achieves a ROC--AUC of $0.9953$, significantly outperforming acoustic and mean-pooling baselines while remaining competitive with strong global temporal models. The model produces its first prediction after 10 seconds of audio, updates every 2 seconds, and operates faster than real time on the evaluated server hardware. Ablations identify recurrent temporal context as the principal contributor to performance. These results demonstrate that fraud risk can be scored incrementally from raw speech without transcripts or temporal annotations, while highlighting the need for latency-aware training to improve early prediction.

Chinese Translation

我们研究了从原始电话音频中进行弱监督增量电信诈骗检测的问题，其中训练仅提供对话级别的标签，并且预测必须在通话结束之前进行更新。我们提出了StreamFraudNet，它通过重叠的有界上下文窗口处理输入音频，使用冻结的自监督语音编码器、递归时间建模和潜在窗口分数的学习聚合。在一个受控的英语基准测试中，StreamFraudNet达到了ROC-AUC值为0.9953，显著优于声学和均值池化基线，同时在强大的全局时间模型中保持竞争力。该模型在10秒音频后产生第一次预测，每2秒更新一次，并且在评估的服务器硬件上运行速度快于实时。消融实验表明，递归时间上下文是性能的主要贡献因素。这些结果表明，诈骗风险可以从原始语音中增量评分，而无需转录或时间标注，同时强调了需要进行延迟感知训练以改善早期预测。

View on arXiv Download PDF AI Translation

cs.CL / 49 / 2609.20232

The Public Discourse Corpus (PDC): A Speaker-Attributed Dataset for Valence and Epistemic Modality with Target Speaker Participation

公共话语语料库（PDC）：一个带有目标发言者参与的情感价值和认知模态的发言者归属数据集

Chen, Bo

Abstract

We introduce the \textbf{Public Discourse Corpus (PDC)}, the first dataset of public-figure interview speech jointly annotated for affective valence and epistemic modality. The corpus contains 998 videos from 100 speakers across seven professional domains, yielding 186,642 sentences (3.1 million words) after sentence segmentation and filtering. To ensure that all retained videos contain analyzable speech from the intended speaker, we introduce \textbf{Target Speaker Participation (TSP)}---a five-category annotation taxonomy with documented inter-annotator reliability ($\kappa = 0.616$)---as a key methodological contribution that any corpus construction project can adopt. Target-speaker turns are separated from interviewer and third-party speech through an \textbf{audio-first diarization pipeline} combining local Whisper ASR with pyannote speaker separation, released as an open-source implementation. We release the annotated corpus, the annotation tools, the cross-provider validation sample, and the complete processing pipeline. The dataset is available at https://huggingface.co/datasets/ictchenbo/public-discourse-corpus.

Chinese Translation

我们介绍了 extbf{公共话语语料库（PDC）}，这是第一个针对公共人物访谈语音进行情感价值和认知模态联合标注的数据集。该语料库包含来自100位发言者的998个视频，涵盖七个专业领域，在句子分割和过滤后，共产生186,642个句子（310万词）。为了确保所有保留的视频都包含来自目标发言者的可分析语音，我们引入了 extbf{目标发言者参与（TSP）}——一个五类标注分类法，并记录了标注者间的一致性（$ ext{kappa} = 0.616$）——作为任何语料库构建项目可以采用的关键方法论贡献。通过结合本地Whisper ASR和pyannote发言者分离的 extbf{音频优先分段流程}，将目标发言者的发言与采访者和第三方发言分开，并以开源实现的形式发布。我们发布了标注语料库、标注工具、跨提供者验证样本和完整的处理流程。该数据集可在https://huggingface.co/datasets/ictchenbo/public-discourse-corpus获取。

View on arXiv Download PDF AI Translation

cs.CL / 50 / 2609.20252

Lens: Bringing the Right Semantic Perspective into Focus for Training-Free Multimodal Representation Learning

Lens：聚焦于无训练多模态表示学习中的正确语义视角

Liu, Xinran, Shi, Shouqian, Chen, Yixian, Chen, Ruizhi, Yao, Xin-Wei, Zhong, Sheng

Abstract

High-quality representations are essential for a wide range of downstream tasks. Dedicated embedding models are explicitly optimized for representation learning, yet their training data are often more limited in scale and diversity than the massive corpora used to pretrain modern large language models and multimodal large language models. Large-scale pretraining and instruction following enable autoregressive models to select relevant evidence, integrate multimodal information, and infer semantics under different task perspectives, creating a distinctive opportunity for training-free representation learning. However, our analysis reveals that existing semantic-elicitation methods do not reliably orient the extracted states toward the semantic perspective required by the downstream task. Consequently, the resulting representations often remain dominated by salient input content. We characterize this problem as semantic perspective misalignment and propose Lens, a training-free framework that makes representation readout task-directed. Semantic Perspective Anchoring associates the task-required perspective with a task-specific readout phrase, specifying the interpretive role of the positions later used for extraction. Contextualized Phrase Readout places the same phrase after the complete input and aggregates its token states, combining full-context access with the anchored perspective. The resulting representation reflects task-conditioned evidence integration and inference rather than a generic summary of salient content. Without parameter updates, architectural modification, or reranking, Lens achieves an overall Precision@1 of 63.9 across all 36 MMEB datasets, outperforming the closest same-backbone training-free embedding baseline by 10.2 points.

Chinese Translation

高质量的表示对于广泛的下游任务至关重要。专用的嵌入模型被明确优化用于表示学习，然而它们的训练数据往往在规模和多样性上不如用于预训练现代大型语言模型和多模态大型语言模型的庞大语料库。大规模预训练和遵循指令使自回归模型能够选择相关证据、整合多模态信息，并在不同任务视角下推断语义，从而为无训练的表示学习创造了独特的机会。然而，我们的分析表明，现有的语义引导方法并不能可靠地将提取的状态定向到下游任务所需的语义视角。因此，所得到的表示往往仍然受到显著输入内容的主导。我们将这个问题表征为语义视角不对齐，并提出了Lens，一个无训练的框架，使表示读取任务导向。语义视角锚定将任务所需的视角与特定任务的读取短语关联起来，明确了后续用于提取的位置的解释性角色。上下文化短语读取将相同的短语放置在完整输入之后，并聚合其标记状态，结合了对全上下文的访问与锚定视角。所得到的表示反映了任务条件下的证据整合和推断，而不是显著内容的通用摘要。在没有参数更新、架构修改或重新排序的情况下，Lens在所有36个MMEB数据集上实现了63.9的整体Precision@1，超越了最接近的相同骨干无训练嵌入基线10.2个百分点。

View on arXiv Download PDF AI Translation

cs.CL / 51 / 2609.20303

Viveka-Insight: a cross-lingual concept graph and citation-grounded retrieval resource over the complete works of Swami Vivekananda in English and Bengali

Viveka-Insight：一个跨语言概念图和基于引用的检索资源，涵盖斯瓦米·维韦卡南达的完整作品（英文和孟加拉文）

Maharaj, Tamal

Abstract

Classical philosophical corpora pose three compounding challenges for language resources: they exist in several languages without parallel alignment, their vocabulary is remote from that of contemporary readers, and generated text over culturally sensitive material must be verifiably grounded. We present Viveka-Insight, a bilingual resource and open-source pipeline for the works of Swami Vivekananda (1863-1902): the nine-volume English Complete Works and the ten-volume Bengali Vani o Rachana, two related but non-parallel corpora of about 15 million characters. Four layers are released: (i) a structure-preserving parse (32,694 paragraphs, 168,842 sentences) with per-paragraph anchors deep-linking into the published editions; (ii) a cross-lingual concept graph of 8,362 language-agnostic concepts with 87,518 relation-typed paragraph-concept and 55,872 concept-concept edges, in which canonical English labels act as a string-equality key linking Bengali and English passages with no parallel data; (iii) a bilingual alias inventory of 60,850 surface forms (30,053 English, 30,797 Bengali); and (iv) a human-annotated set of 200 paragraph-concept edges judged by three annotators, released with all per-annotator judgments. We report known-item cross-lingual retrieval over 194 verified rendered lecture pairs (Recall@10 0.86 in both directions), a 30-question audit of citation integrity and modern-question bridging, and a human study placing concept-extraction precision at 0.60 under strict two-annotator consensus (Cohen's kappa = 0.61). The extractor's confidence weight is calibrated: restricting to weight >= 0.8 raises precision to 0.71 while retaining 98% of concept-bearing paragraphs. Precision is markedly lower in Bengali than English (0.54 vs 0.68), locating the weakness in exactly the half that cross-lingual access depends on. The design transfers to other multilingual classical corpora.

Chinese Translation

经典哲学语料库给语言资源带来了三个复合挑战：它们存在于多种语言中且没有平行对齐，词汇与当代读者的语言相距甚远，并且对文化敏感材料生成的文本必须可验证地扎根于事实。我们提出了Viveka-Insight，这是一个双语资源和开源管道，涵盖斯瓦米·维韦卡南达（1863-1902）的作品：九卷本的英文《完整作品》和十卷本的孟加拉文《Vani o Rachana》，这两个相关但不平行的语料库约有1500万个字符。我们发布了四个层次的内容：（i）一个结构保持的解析（32,694段落，168,842句子），每段落都有锚点深度链接到已出版的版本；（ii）一个包含8,362个语言无关概念的跨语言概念图，具有87,518个关系类型的段落-概念边和55,872个概念-概念边，其中规范的英文标签作为字符串相等键，链接没有平行数据的孟加拉文和英文段落；（iii）一个包含60,850个表面形式的双语别名清单（30,053个英文，30,797个孟加拉文）；（iv）一个由三位注释者评判的200个段落-概念边的人类注释集，随附所有注释者的判断。我们报告了对194对经过验证的讲座对的已知项跨语言检索（双向Recall@10为0.86），对引用完整性和现代问题桥接的30个问题审计，以及一项人类研究，在严格的双注释者共识下，概念提取精度为0.60（Cohen's kappa = 0.61）。提取器的置信度权重经过校准：限制权重>=0.8将精度提高到0.71，同时保留98%的概念承载段落。孟加拉文的精度明显低于英文（0.54对0.68），其弱点正好位于跨语言访问所依赖的那一半。该设计可转移到其他多语言经典语料库。

View on arXiv Download PDF AI Translation

cs.CL / 52 / 2609.20398

Schema-Anchored Latent Reasoning for Semantic Parsing-Based Knowledge Base Question Answering

基于模式锚定的潜在推理用于语义解析的知识库问答

Gao, Guangze, Li, Zixuan, Zhang, Sikui, Yuan, Chunfeng, Li, Wenjuan, Li, Bing, Jin, Xiaolong, Hu, Weiming

Abstract

Semantic parsing (SP)-based knowledge base question answering aims to answer natural language questions by generating executable logical forms (LFs) over knowledge bases (KBs). When applying Large Language Models (LLMs) to this task, a key challenge over large, heterogeneous KBs is selecting question-related schema elements (i.e., relations and classes) and composing them into complex LFs. Recent LLM-based methods often make early discrete commitments to schema elements during intermediate reasoning, allowing incorrect intermediate schema decisions to propagate and finally result in incorrect LFs. To overcome this limitation, we propose SALR, a schema-anchored latent reasoning method for LF construction. It performs multi-step reasoning by generating continuous thoughts in the model's hidden states, thereby delaying the explicit commitment to LF decisions. To ground this latent reasoning process in the corresponding KB schema, SALR aligns continuous thoughts with a codebook of KB schema elements through an alignment objective supervised by schema traces deterministically derived from gold LFs. It then incorporates the aligned schema codes into inputs for subsequent reasoning steps. This schema-mediated feedback guides LF generation without requiring the model to emit an explicit textual reasoning trajectory. Experiments on GrailQA and WebQSP show that SALR achieves consistent overall gains over strong baselines. Notably, on compositional questions from GrailQA, SALR outperforms TIARA, a strong SP-based baseline, by 2.86 F1 points. Further analyses show that schema-mediated feedback affects LF generation and that schema information is recoverable from the latent states.

Chinese Translation

基于语义解析（SP）的知识库问答旨在通过生成可执行的逻辑形式（LFs）来回答自然语言问题，涉及知识库（KB）。在将大型语言模型（LLMs）应用于此任务时，面对大型异构知识库的一个主要挑战是选择与问题相关的模式元素（即关系和类），并将其组合成复杂的逻辑形式。近期的基于LLM的方法往往在中间推理过程中对模式元素做出早期离散承诺，这导致错误的中间模式决策得以传播，最终导致不正确的逻辑形式。为了解决这一限制，我们提出了SALR，一种用于逻辑形式构建的模式锚定潜在推理方法。它通过在模型的隐藏状态中生成连续思维来执行多步推理，从而延迟对逻辑形式决策的明确承诺。为了将这一潜在推理过程与相应的知识库模式相结合，SALR通过一个对齐目标将连续思维与知识库模式元素的代码本对齐，该对齐目标由从黄金逻辑形式确定性推导出的模式痕迹监督。然后，它将对齐的模式代码纳入后续推理步骤的输入中。这种模式介导的反馈指导逻辑形式的生成，而无需模型发出明确的文本推理轨迹。在GrailQA和WebQSP上的实验表明，SALR在强基线之上实现了一致的整体提升。值得注意的是，在GrailQA的组合问题上，SALR比强大的基于SP的基线TIARA高出2.86个F1点。进一步分析表明，模式介导的反馈影响逻辑形式的生成，并且模式信息可以从潜在状态中恢复。

View on arXiv Download PDF AI Translation

cs.CL / 53 / 2609.20408

Xeno-Interpretability: Investigating the Alien Minds of LLMs

异类可解释性：探讨大型语言模型的外星思维

Pierucci, F., Syrnikov, M. Bracale, Prandi, M., Galisai, M., Giarrusso, F., Bisconti, P.

Abstract

Large language models are usually interpreted through concepts that humans already possess: truthfulness, refusal, deception, personality, harmfulness, and related categories. This paper asks whether models may also represent and use distinctions for which no adequate human concept exists. We call such internal structures xeno-representations, and their study xeno-interpretability. We distinguish the human-interpretable semantic space from the xeno-semantic space: the region of model-native representations for which no adequate human conceptual counterpart is available. We show that the space of possible internal distinctions in an LLM is substantially larger than the space available through finite human descriptions. We then separate experimental identification from semantic interpretation: an internal representation may be reproducibly located, geometrically characterized, causally manipulated, and linked to downstream behaviour even when its semantic content cannot be adequately expressed in human terms. On this basis, we sketch an empirical programme to identify xeno-representations. We finally examine the implications for AI safety and multi-agent systems, where model-native representations may propagate and stabilize across interacting agents while remaining only partially visible through human-readable communication. Xeno-interpretability therefore shifts the aim of interpretability from finding human concepts inside models toward discovering and characterizing the representational structures that are native to the models themselves and might affect their behaviour in unpredictable ways.

Chinese Translation

大型语言模型通常通过人类已有的概念进行解释：真实性、拒绝、欺骗、个性、有害性及相关类别。本文探讨模型是否也可能表示和使用没有适当人类概念的区分。我们将这种内部结构称为异类表征（xeno-representations），其研究称为异类可解释性（xeno-interpretability）。我们区分人类可解释的语义空间与异类语义空间：即模型本土表征的区域，在该区域内没有适当的人类概念对应。我们展示了大型语言模型内部可能的区分空间远大于有限人类描述所能提供的空间。接着，我们将实验识别与语义解释分开：即使其语义内容无法用人类术语充分表达，内部表征仍然可以被可重复地定位、几何表征、因果操控，并与下游行为关联。在此基础上，我们勾勒出一个实证计划以识别异类表征。最后，我们考察其对人工智能安全和多智能体系统的影响，在这些系统中，模型本土表征可能在相互作用的智能体之间传播和稳定，同时仅部分通过人类可读的交流可见。因此，异类可解释性将可解释性的目标从寻找模型内部的人类概念转向发现和表征模型自身固有的表征结构，这些结构可能以不可预测的方式影响其行为。

View on arXiv Download PDF AI Translation

cs.CL / 54 / 2609.20412

Stress-testing Alignment Midtraining

压力测试对齐中期训练

Baines, Sid, Bostock, Jonathan, Martinez, Maria Angelica, Draganov, Andrew, Africa, David, Tan, Daniel

Abstract

When aligning frontier models through post-training techniques, it is not possible to directly demonstrate all of the behaviours we want a model to exhibit in all possible deployment environments; our model must generalise outside of the post-training distribution. One proposed solution is alignment midtraining (AMT), which continues pretraining on large volumes of alignment-relevant documents to encourage generalisation in later stages of training. Despite the prominence of AMT as an alignment approach, there is limited public evidence for its effectiveness. To resolve this, we identify several assumptions around midtraining and evaluate them across scale: up to 110 billion-parameter models and 1 billion midtraining tokens. For instance, we study a scenario where post-training data is ambiguous between two possible motivations. We find that midtraining can steer the model's motivation in simple versions of this setting. However, the presence of a tiny fraction of finetuning data which suggests a competing motivation erases the effects of AMT. We also study scenarios in which we want an AI to follow a number of rules, but only demonstrate a subset of them. We find that demonstrations must be present either in midtraining or post-training datasets for these rules to be robustly learned. Based on these and other findings, we do not believe that there is sufficient public evidence for us to confidently state that midtraining can address the core difficulties inherent in aligning powerful AI systems.

Chinese Translation

在通过后训练技术对齐前沿模型时，无法直接展示我们希望模型在所有可能的部署环境中展现的所有行为；我们的模型必须在后训练分布之外进行泛化。一种提出的解决方案是对齐中期训练（Alignment Midtraining, AMT），该方法在大量与对齐相关的文档上继续进行预训练，以鼓励在训练后期的泛化。尽管AMT作为一种对齐方法备受关注，但其有效性的公共证据有限。为了解决这一问题，我们确定了关于中期训练的几个假设，并在规模上进行了评估：最高可达1100亿参数模型和10亿中期训练标记。例如，我们研究了一个场景，其中后训练数据在两种可能的动机之间存在歧义。我们发现，中期训练可以在这种设置的简单版本中引导模型的动机。然而，存在一小部分建议竞争动机的微调数据会抹去AMT的效果。我们还研究了希望AI遵循多个规则但仅展示其中一部分的场景。我们发现，这些规则的稳健学习必须在中期训练或后训练数据集中存在演示。基于这些及其他发现，我们认为目前没有足够的公共证据让我们自信地声明中期训练能够解决强大AI系统对齐中固有的核心困难。

View on arXiv Download PDF AI Translation

cs.CL / 55 / 2609.20484

Edustories: A Collection of Real-world Case Studies from Classroom Practices

教育故事：来自课堂实践的真实案例研究集合

Štefánik, Michal, Nehyba, Jan, Karasova, Jirina, Fico, Martin, Škarková, Lucie, Košatková, Markéta, Kosatka, David

Abstract

Despite the widely recognized potential of AI in education, most prior work has focused on individualized student assistance. In contrast, the majority of educational practice worldwide still takes place in collective classroom settings. To enable researchers to study AI assistance in collective teaching, we introduce Edustories, a dataset of 1,492 teacher-written case studies describing real elementary and high-school classroom situations involving challenging student behavior, pedagogical interventions, and their outcomes. Among many other applications, Edustories enables evaluating LLMs' ability to predict the success of teacher interventions, crucial for providing practicing teachers with useful feedback. Comparing the latest models from four language-model families against expert assessments, we find that current models fall short of human expertise in predicting classroom outcomes; the strongest models reach 58% accuracy compared to 64% of human experts. This gap highlights both the limitations and the emerging potential of AI as assistants for practicing teachers.

Chinese Translation

尽管人工智能在教育领域的潜力被广泛认可，但大多数先前的研究主要集中在个性化学生辅助上。相比之下，全球大多数教育实践仍然发生在集体课堂环境中。为了使研究人员能够研究集体教学中的人工智能辅助，我们引入了教育故事（Edustories），这是一个包含1,492篇教师撰写的案例研究的数据集，描述了涉及挑战性学生行为、教学干预及其结果的真实小学和中学课堂情境。在许多其他应用中，教育故事使得评估大型语言模型（LLMs）预测教师干预成功能力成为可能，这对为在职教师提供有用反馈至关重要。通过将四个语言模型家族的最新模型与专家评估进行比较，我们发现当前模型在预测课堂结果方面未能达到人类专家的水平；最强的模型准确率为58%，而人类专家的准确率为64%。这一差距突显了人工智能作为在职教师助手的局限性和新兴潜力。

View on arXiv Download PDF AI Translation

cs.CL / 56 / 2609.20530

Relational Attention for Data-Efficient Language Modeling

用于数据高效语言建模的关系注意力

Brasoveanu, Adrian, Takmaz, Ece, Dotlačil, Jakub

Abstract

We present Relational BabyLM, a system submission to the BabyLM 2026 challenge that combines two cognitively motivated inductive biases in a single decoder-only Transformer. Architecturally, we replace standard self-attention with a Dual Attention Transformer (DAT), which separates the routing of object-level ("sensory") lexical features from structural/relational information (Altabaa and Lafferty, 2025; Altabaa et al., 2024; Webb et al., 2024; Kerg et al., 2022; Webb et al., 2021). Relational attention (RA) disentangled from self-attention greatly increases data efficiency and out-of-training-sample generalization on purely relational tasks, but language modeling requires object-level and relational information to be integrated as well as disentangled, and RA-based LMs have remained largely unexplored. BabyLM's data-constrained training and comprehensive evaluation is an ideal testing ground for whether that data efficiency transfers. As a training intervention, we add a Next-Latent Prediction (NextLat; Teoh et al. 2026) objective that encourages hidden states to compress history incrementally into a dense belief state. Architecture is the dominant factor for structural linguistic generalization; the objective is secondary but still significant. DAT's three relational attention types (full RA vs. the simpler RCA and DisRCA variants) are largely interchangeable at 10M words; full RA pulls ahead at 100M. We also introduce a novel symbol-retrieval mechanism (RoPE-based, as opposed to learned, relative symbols) that matches learned symbol libraries while adding no parameters. On the strict (100M-word) track, our best model ranks 6th of 55 overall and 3rd of 55 on the leaderboard's NLP-task subset at the time of writing; our two strongest models outperform the GPT-2 baseline on most benchmarks, with one attaining the highest EWoK score among strict-track entries.

Chinese Translation

我们提出了关系BabyLM，这是一个提交给BabyLM 2026挑战的系统，结合了两种认知驱动的归纳偏差于单一的解码器Transformer中。在架构上，我们用双重注意力Transformer（Dual Attention Transformer, DAT）替代了标准的自注意力机制，该机制将对象级（“感知”）词汇特征的路由与结构/关系信息分开（Altabaa和Lafferty, 2025; Altabaa等, 2024; Webb等, 2024; Kerg等, 2022; Webb等, 2021）。与自注意力机制解耦的关系注意力（Relational Attention, RA）显著提高了数据效率，并在纯关系任务上实现了超出训练样本的泛化，但语言建模需要将对象级和关系信息整合并解耦，而基于RA的语言模型仍然在很大程度上未被探索。BabyLM的数据受限训练和全面评估为这种数据效率是否能够转移提供了理想的测试平台。作为训练干预，我们添加了一个下一潜在预测（Next-Latent Prediction, NextLat; Teoh等, 2026）目标，鼓励隐藏状态逐步将历史压缩成一个密集的信念状态。架构是结构性语言泛化的主导因素；目标是次要但仍然重要。DAT的三种关系注意力类型（完全RA与更简单的RCA和DisRCA变体）在10M词时大致可互换；在100M词时，完全RA表现更佳。我们还引入了一种新颖的符号检索机制（基于RoPE的，相较于学习的相对符号），该机制在匹配学习的符号库的同时不增加任何参数。在严格的（100M词）赛道上，我们的最佳模型在55个模型中排名第6，在排行榜的NLP任务子集上排名第3；我们的两个最强模型在大多数基准测试中超越了GPT-2基线，其中一个在严格赛道的参赛作品中获得了最高的EWoK分数。

View on arXiv Download PDF AI Translation

cs.CL / 57 / 2609.20541

An Analysis of Training-Free Self-Reported Confidence in Language Models

无训练自报告信心在语言模型中的分析

Meyer, Lukas, Rossi, Sofia, Chen, Wei, Laurent, Thomas, Li, Yiming

Abstract

Large language models can report a numerical confidence together with generated content, but it is unclear whether this report is more than calibrated rhetoric. We analyze three training-free signals: confidence verbalized with the answer, post-hoc $P(\mathrm{True})$, and agreement with three additional generations on the same 100 TriviaQA questions for two model families. Direct verbalization is a surprisingly strong baseline: after auditing benchmark errors, it reaches AUROC 0.956 and 0.937 for correctness prediction. Three-sample agreement is substantially weaker (0.765 and 0.790), and a fixed interpolation with verbalized confidence has no statistically reliable benefit. Four of nine errors from one model and two of eight from the other receive unanimous sample support, showing that self-consistency can amplify shared misconceptions. Re-eliciting confidence for the same fixed answers with equivalent prompts changes scores by 0.043 to 0.084 on average and flips 4\% to 9\% of decisions at a 0.8 threshold. An exploratory audit of 100 confidence-tagged biography claims further finds only a modest confidence gap between supported and contradicted claims. These results argue that useful self-reports remain sensitive to elicitation, correlated errors, and benchmark noise.

Chinese Translation

大型语言模型可以在生成内容时报告一个数值信心，但尚不清楚这一报告是否超出了校准的修辞。我们分析了三种无训练信号：与答案一起表达的信心、事后概率 $P( ext{True})$，以及对同一100个TriviaQA问题的三次额外生成结果的一致性，针对两种模型家族。直接的语言表达是一个令人惊讶的强基线：在审核基准错误后，其正确性预测的AUROC分别达到0.956和0.937。三样本一致性明显较弱（0.765和0.790），而与语言表达信心的固定插值没有统计上的可靠益处。来自一个模型的九个错误中有四个和另一个模型的八个错误中有两个获得了一致的样本支持，这表明自我一致性可以放大共享的误解。对于相同固定答案的等效提示重新引导信心，平均得分变化为0.043到0.084，并在0.8阈值下翻转4 ext{%}到9 ext{%}的决策。对100个信心标记的传记声明的探索性审核进一步发现，支持和反驳声明之间的信心差距仅为适度。这些结果表明，有用的自我报告仍然对引导、相关错误和基准噪声敏感。

View on arXiv Download PDF AI Translation

cs.CL / 58 / 2609.20565

Steering the Compass: Aligning Dynamic Psychological Counseling Conversations with Cognitive Behavioral Therapy Strategies

引导方向：将动态心理咨询对话与认知行为疗法策略对齐

Wang, Zimu, Jiang, Yiwen, Zhao, Xiangyu, Shen, Yaling, Liu, Jiahe, Fong, Stephanie, Cheng, Maxmartwell H, Oliveira, Guilherme C, Nguyen, Anh, Desimone, Robert, Nelson, Barnaby, Dwyer, Dominic, Ge, Zongyuan

Abstract

Recent advancements in large language models have revolutionized the field of psychological counseling, especially in the context of Cognitive Behavioral Therapy (CBT). While the success of CBT relies heavily on dynamic decision-making informed by the client's real-time mental state, this aspect has often been overlooked in current research, limiting both flexibility and therapeutic outcomes. In this paper, we introduce StratCBT, a dataset specifically designed for psychological counseling conversations with CBT Strategies, consisting of 9,688 sessions and around 256K utterances, with each counselor's response aligned with one of eight distinct strategies. The creation of StratCBT involves modeling clients based on their negative thoughts and generating high-quality counseling conversations through self-chat, incorporating realistic sessions as guidance, thereby significantly surpassing existing datasets in both general counseling and CBT-specific skills. We conduct extensive experiments to demonstrate the effectiveness of strategy-aligned generation and evaluate its efficacy in delivering professional and effective counseling with LLM-simulated clients to reflect real-world scenarios. The dataset can be obtained from https://github.com/zimuwangnlp/StratCBT.

Chinese Translation

近年来，大型语言模型的进展彻底改变了心理咨询领域，尤其是在认知行为疗法（CBT）的背景下。尽管CBT的成功在很大程度上依赖于基于客户实时心理状态的动态决策，但这一方面在当前研究中常常被忽视，限制了灵活性和治疗效果。本文介绍了StratCBT，一个专门为心理咨询对话与CBT策略设计的数据集，包含9,688个会话和约256,000个发言，每位咨询师的回应与八种不同策略之一对齐。StratCBT的创建涉及根据客户的负面思维建模，并通过自我对话生成高质量的咨询对话，结合现实会话作为指导，从而在一般咨询和CBT特定技能方面显著超越现有数据集。我们进行了广泛的实验，以证明策略对齐生成的有效性，并评估其在与LLM模拟客户进行专业和有效咨询时的有效性，以反映现实场景。该数据集可从https://github.com/zimuwangnlp/StratCBT获取。

View on arXiv Download PDF AI Translation

cs.CL / 59 / 2609.20584

SAFARI: An Industrial Benchmark for LLM-Assisted Hazard Analysis and Risk Assessment

SAFARI：一种工业基准用于大语言模型辅助的危害分析与风险评估

Wu, Chenxi, Wang, Zimu, Zhang, Haiyang, Wang, Wei, Xu, Zhijie

Abstract

Large language models (LLMs) are increasingly considered for safety-critical engineering, yet their reliability in regulated functional-safety workflows remains underexplored. We introduce SAFARI (Safety-Aware Functional Automotive Risk Inference), the first industrial benchmark for LLM-assisted automotive Hazard Analysis and Risk Assessment (HARA) under ISO 26262. It contains 3,000 de-identified industrial HARA cases and evaluates two coupled tasks: open-ended hazard analysis and standards-grounded risk assessment. To evaluate open-ended HARA artifacts, we propose the first reference-anchored LLM-as-a-judge protocol with high expert correlation. Experiments with nine frontier LLMs show that models often produce plausible hazard narratives but remain weak at ISO 26262 risk classification, with the best ASIL macro-F1 reaching only 0.261. Chain-of-Thought prompting provides limited benefit and often degrades categorical risk assessment. Error analysis further localizes major failures to scenario-critical context omissions during hazard generation and to controllability misjudgments during risk assessment, indicating where expert oversight should be concentrated. The dataset can be obtained from https://github.com/xixi47520-hash/HARA.

Chinese Translation

大语言模型（LLMs）在安全关键工程中的应用越来越受到关注，但它们在受监管的功能安全工作流程中的可靠性仍然未得到充分探索。我们介绍了SAFARI（安全意识功能汽车风险推断），这是第一个针对ISO 26262下大语言模型辅助的汽车危害分析与风险评估（HARA）的工业基准。该基准包含3000个去标识化的工业HARA案例，并评估两个耦合任务：开放式危害分析和基于标准的风险评估。为了评估开放式HARA成果，我们提出了第一个参考锚定的LLM作为评判者的协议，具有较高的专家相关性。对九个前沿LLM的实验表明，模型通常能够生成合理的危害叙述，但在ISO 26262风险分类方面仍然表现较弱，最佳的ASIL宏观F1仅达到0.261。思维链提示提供的益处有限，且常常降低分类风险评估的效果。错误分析进一步将主要失败归因于危害生成过程中对场景关键上下文的遗漏，以及在风险评估过程中对可控性的误判，指示了专家监督应集中关注的领域。数据集可从https://github.com/xixi47520-hash/HARA获取。

View on arXiv Download PDF AI Translation

cs.CL / 60 / 2609.20593

WiC is Not WSD: A Study on LLMs and Lexical Ambiguity Resolution

WiC并非WSD：关于大型语言模型与词汇歧义解析的研究

Zhou, Yi, Rezaee, Kiamehr, Bollegala, Danushka, Pilehvar, Mohammad Taher, Camacho-Collados, Jose

Abstract

Word-in-Context (WiC) remains challenging for language models, despite recent progress on lexical-semantic tasks. We hypothesise that this difficulty arises not only from comparing two contextual uses of a word, but also from the absence of an explicit sense inventory that specifies the relevant level of semantic granularity. We evaluate open LLMs on WiC and traditional Word Sense Disambiguation (WSD) under similar settings. We find that providing candidate senses, similar to what is done in traditional WSD, improves WiC performance in all settings. In general, explicit sense information helps models make more consistent and targeted judgements. Human evaluation further shows that many apparent WiC errors reflect label ambiguity or mismatches between model and annotator sense boundaries rather than simple failures of lexical understanding. In particular, results show that LLMs overthink the sense distinction often leading to errors based on overly fine-grained distinctions.

Chinese Translation

尽管在词汇语义任务上取得了近期进展，Word-in-Context (WiC) 对语言模型仍然具有挑战性。我们假设，这一困难不仅源于比较一个词的两个上下文用法，还源于缺乏明确的词义清单来指定相关的语义粒度水平。我们在类似设置下评估开放的LLMs在WiC和传统的词义消歧（WSD）上的表现。我们发现，提供候选词义，类似于传统WSD中的做法，能够在所有设置中提升WiC的表现。总体而言，明确的词义信息有助于模型做出更一致和更有针对性的判断。人类评估进一步表明，许多明显的WiC错误反映的是标签歧义或模型与标注者词义边界之间的不匹配，而非简单的词汇理解失败。特别是，结果显示LLMs在词义区分上过于细致，常常导致基于过于细微的区分而产生错误。

View on arXiv Download PDF AI Translation

cs.CL / 61 / 2609.20612

What Does Privileged Information Add to On-Policy Self-Distillation?

特权信息对在线自蒸馏的贡献是什么？

Zhang, XiuYu, Chow, Wei, Fang, Junfeng, Liang, Zhenkai, Chua, Tat-Seng

Abstract

On-policy self-distillation (OPSD) lets a language model learn from a frozen copy of itself that sees an answer or a worked solution. Giving the teacher this extra information seems to offer the student more to learn, but how much does it add beyond distillation itself? To isolate that contribution, we construct AMPLE-Math, a reusable suite of 5,319 mathematical problems with six reasoning views that share the same answer, and compare each view with matched reference-free distillation. With a thinking-enabled teacher supervising direct-response rollouts, reference-free distillation accounts for much of Qwen3-1.7B's improvement under thinking-enabled evaluation, both in domain and on external benchmarks. Evidence for an additional reference benefit is modest in Qwen, strongest for a polished solution, whereas complete traces add two percentage points in SmolLM3-3B at step 50. These benefits depend on the student being trained. At the same checkpoint, replacing short direct-response rollouts with long thinking-enabled rollouts turns gains into losses in both families while the problems, references, and evaluation stay fixed. Teacher profiles and matched loss interventions in Qwen further show that changing token-level supervision can leave student behavior largely unchanged. Together, these findings suggest that OPSD can improve access to existing reasoning capabilities through parameters shared by direct-response and thinking-enabled inference. The value of a privileged reference is what it adds to this cross-mode transfer, not how much of the solution it reveals.

Chinese Translation

在线自蒸馏（On-policy self-distillation, OPSD）允许语言模型从一个冻结的自身副本中学习，该副本能够看到答案或已解决的方案。给予教师这些额外信息似乎为学生提供了更多的学习内容，但它在蒸馏本身之外究竟增加了多少？为了隔离这一贡献，我们构建了AMPLE-Math，这是一个可重用的数学问题套件，包含5,319个数学问题，具有六种推理视角，这些视角共享相同的答案，并将每种视角与匹配的无参考蒸馏进行比较。在一个启用思考的教师监督下的直接响应回滚中，无参考蒸馏在思考启用评估下占据了Qwen3-1.7B改进的很大一部分，无论是在领域内还是在外部基准测试中。对于Qwen，额外参考的证据是适度的，针对精炼解决方案的效果最强，而在第50步时，完整的追踪在SmolLM3-3B中增加了两个百分点。这些好处依赖于被训练的学生。在同一个检查点，将短的直接响应回滚替换为长的启用思考的回滚，导致两个系列中的收益变为损失，同时问题、参考和评估保持不变。Qwen中的教师档案和匹配损失干预进一步表明，改变令牌级别的监督可以使学生行为基本保持不变。综合来看，这些发现表明，OPSD可以通过直接响应和启用思考的推理共享的参数，改善对现有推理能力的访问。特权参考的价值在于它对这种跨模式转移的贡献，而不是它揭示了多少解决方案。

View on arXiv Download PDF AI Translation

cs.CL / 62 / 2609.20625

Chronicle: Cut-Point Replay for Regression Testing of LLM Agents

Chronicle：用于大语言模型代理回归测试的切点重放

Chawla, Tisha, Koul, Susheem

Abstract

Large language model responses are non-deterministic, so failures in LLM agents are hard to reproduce: a failure depends on inference that is not bitwise reproducible, on tools that read changing state, and on a multi-step trajectory that a re-run rarely repeats. Record-and-replay makes a run reproducible, but existing agent tooling records runs only to trace or score them, not to test a code change against them. We present Chronicle, which records an agent run at its non-deterministic boundaries as immutable envelopes and replays it from the record. Its central operation, cut-point replay, serves a chosen subset of boundaries from the record and executes the complementary subset live with new code, turning a recorded incident into a regression test that runs in continuous integration. On a benchmark of 6 recorded failures with simulated model boundaries, recording adds 23 {\mu}s per crossing (0.008% of an assumed 300 ms model call), full replay issues zero model calls and is bit-stable across 20 repetitions, and cut-point tests fail on faulty code and pass on guarded and benign changes for all 6 incidents. In a mutation study of the guarded tools, cut-point tests catch every mutant that lets the recorded unsafe action through, while a baseline that stubs every boundary, using the same assertion, catches none. Chronicle and the benchmark are publicly available at https://github.com/theagentplane/chronicle.

Chinese Translation

大语言模型的响应是非确定性的，因此LLM代理的故障难以重现：故障依赖于不可逐位重现的推理、读取变化状态的工具，以及一个重新运行很少重复的多步骤轨迹。记录与重放使得运行可重现，但现有的代理工具仅记录运行以进行追踪或评分，而不是用来测试代码更改。我们提出了Chronicle，它在代理运行的非确定性边界处记录为不可变的封装，并从记录中重放。其核心操作，切点重放，从记录中服务于所选的边界子集，并使用新代码实时执行互补子集，将记录的事件转化为在持续集成中运行的回归测试。在一个包含6个记录故障的基准测试中，记录每次交叉增加23微秒（占假设的300毫秒模型调用的0.008%），完整重放不发出任何模型调用，并在20次重复中保持位稳定，而切点测试在有缺陷的代码上失败，在所有6个事件的受保护和良性更改上通过。在对受保护工具的突变研究中，切点测试捕捉到每一个允许记录的不安全操作通过的突变体，而使用相同断言的基线在每个边界上进行存根，未捕捉到任何突变体。Chronicle和基准测试可在https://github.com/theagentplane/chronicle上公开获取。

View on arXiv Download PDF AI Translation

cs.CL / 63 / 2609.20630

UniPolicy: Unified Objective-Specific Policies for Generative Search Advertising

UniPolicy：用于生成搜索广告的统一目标特定策略

Yao, Kun, Zhou, Yuhang, Zhang, Yichi, Tong, Zeliang, Xue, Shengri, Wang, Haitao, Lu, Siyu, Xie, Qianlong, Wang, Xingxing

Abstract

Search advertising connects user intent with commercial content and plays a critical role in platform monetization. Recent systems typically align pretrained generative models with a single business reward, such as eCPM, or use naive reward fusion for preliminary multi-objective alignment. However, an ideal search advertising system must jointly account for heterogeneous objectives, including relevance, click propensity, and commercial value, to balance user experience and business value while mitigating globally suboptimal performance caused by gradient competition. We propose UniPolicy, an objective-aware multi-policy alignment framework. UniPolicy combines objective-specific prefix tokens, sparse MoE-LoRA routing, and objective-specific residual FFNs to hierarchically decouple parameters within a shared backbone, providing differentiated parameter and policy-expression spaces for different business objectives. It further constructs pairwise preferences from multi-stage behavioral feedback, supplementing the relative preference information in exposed-but-unclicked samples and strengthening the relative advantage of clicked candidates in the generation distribution. At inference, UniPolicy supports parallel, business-customizable multi-policy beam search, flexibly allocating candidate quotas across objectives under a fixed retrieval budget. Large-scale offline experiments show that UniPolicy delivers balanced improvements across multiple metrics while preserving retrieval quality, outperforming single-objective reinforcement learning and naive reward-fusion baselines. In a 7-day online A/B test on a real search advertising system, UniPolicy improves CTR by 0.71%, RPS by 1.58%, and advertising revenue by 1.32%, while maintaining stable serving latency.

Chinese Translation

搜索广告将用户意图与商业内容连接起来，在平台变现中发挥着关键作用。近期的系统通常将预训练的生成模型与单一商业奖励（如 eCPM）对齐，或使用简单的奖励融合进行初步的多目标对齐。然而，理想的搜索广告系统必须共同考虑异构目标，包括相关性、点击倾向和商业价值，以平衡用户体验和商业价值，同时减轻由于梯度竞争导致的全局次优性能。我们提出了 UniPolicy，一种目标感知的多策略对齐框架。UniPolicy 结合了目标特定的前缀标记、稀疏的 MoE-LoRA 路由和目标特定的残差前馈网络（FFNs），在共享的主干网络中分层解耦参数，为不同的商业目标提供差异化的参数和策略表达空间。它进一步从多阶段行为反馈中构建成对偏好，补充曝光但未点击样本中的相对偏好信息，并增强点击候选在生成分布中的相对优势。在推理时，UniPolicy 支持并行的、可定制的多策略束搜索，在固定的检索预算下灵活分配候选配额。大规模离线实验表明，UniPolicy 在多个指标上实现了平衡的改进，同时保持了检索质量，优于单目标强化学习和简单奖励融合基线。在一个为期 7 天的真实搜索广告系统的在线 A/B 测试中，UniPolicy 将 CTR 提高了 0.71%，RPS 提高了 1.58%，广告收入提高了 1.32%，同时保持了稳定的服务延迟。

View on arXiv Download PDF AI Translation

cs.CL / 64 / 2609.20684

HerHealthEval: Evaluating Multilingual and Register-Sensitive Understanding of Women's Health Communication

HerHealthEval：评估女性健康沟通的多语言和语域敏感理解

Albattra, Hassan Saeed Hassan, Bahgat, Mazen Mohammed, Ferdousi, Rahatara, Amrya, Hana Essam Sayed Ahmed, Mousa, Mariam

Abstract

Large language models are increasingly used in healthcare communication, yet most evaluations emphasize response quality while assuming that the user's concern has been interpreted correctly. We introduce HerHealthEval, a controlled evaluation framework for multilingual understanding of women's-health communication. For each clinical case, HerHealthEval provides matched versions in English, French, and Modern Standard Arabic using six communicative forms: canonical, clinical, layperson, indirect or hedged, emotionally concerned, and deliberately under-specified. The first five express the same underlying concern and retain the same clinical information, whereas the under-specified form intentionally omits relevant details to test whether the model recognizes that clarification is needed. We evaluate a multilingual instruction model and QLoRA-adapted variants on concern classification, risk calibration, clarification behavior, parse compliance, and cross-form consistency. Results reveal that aggregate accuracy and consistency can conceal safety-relevant failures. A multilingual adaptation model reaches 0.994 under-triage in French and Arabic under language-asymmetric risk supervision. A controlled re-adaptation using source-derived, language-invariant risk labels reduces under-triage to 0.572 and 0.558, respectively. These findings show that robust multilingual healthcare evaluation requires explicit testing of register variation, uncertainty handling, and the provenance and invariance of adaptation labels.

Chinese Translation

大型语言模型在医疗沟通中的应用日益增多，但大多数评估强调响应质量，同时假设用户的关注点已被正确解读。我们引入了HerHealthEval，这是一个用于评估女性健康沟通的多语言理解的控制评估框架。对于每个临床案例，HerHealthEval提供了英语、法语和现代标准阿拉伯语的匹配版本，使用六种沟通形式：规范形式、临床形式、普通人形式、间接或含糊形式、情感关切形式以及故意不具体的形式。前五种形式表达相同的潜在关注点，并保留相同的临床信息，而不具体的形式则故意省略相关细节，以测试模型是否识别出需要澄清。我们评估了一种多语言指令模型及其QLoRA适配变体在关注分类、风险校准、澄清行为、解析合规性和跨形式一致性方面的表现。结果显示，整体准确性和一致性可能掩盖与安全相关的失败。在语言不对称风险监督下，一种多语言适配模型在法语和阿拉伯语中的低分流率达到0.994。使用源派生的、语言不变的风险标签进行的控制再适配将低分流率分别降低至0.572和0.558。这些发现表明，稳健的多语言医疗评估需要明确测试语域变化、不确定性处理以及适配标签的来源和不变性。

View on arXiv Download PDF AI Translation

cs.CL / 65 / 2609.20712

Summarization Bias: The Directional Collapse of Objective Projection into Told-Mode Labels in Large Language Models --- A Conceptual Framework and Registered Test Protocol

摘要偏差：大型语言模型中客观投影向叙述模式标签的方向性崩溃——一个概念框架和注册测试协议

Bulut, Levent

Abstract

This paper introduces and operationalizes summarization bias: a proposed systematic tendency of large language models (LLMs) to represent narrative meaning as an abstract summary label rather than as the reconstructable inferential structure that produces it. Within the Bulut Doctrine, narrative effect is theorized along a told-shown axis: in told mode, emotional and informational content is declared explicitly and requires little reader reconstruction; in shown mode, that content is suppressed at the surface and must be reconstructed from physical cues and indirection (Objective Projection). Shown mode is the higher-load condition the doctrine is designed to measure. The claim is that LLMs fail along this axis in a specific direction. Summarization bias is hypothesized to operate in two regimes: (i) a generative regime, in which a model asked to render an emotion through Objective Projection defaults to declaring it instead; and (ii) an evaluative regime, in which a model judging narrative quality rewards told-mode explicitness and under-detects shown-mode suppression. The evaluative regime is the more consequential, since LLMs increasingly serve as judges and reward models, and a directional bias toward told mode would impose a selection pressure degrading prose toward flat declaration. This report does not claim the bias is validated. It defines the construct, situates it against LLM-as-judge biases, rereads a completed independent reliability study as directional evidence consistent with it, and pre-registers a two-regime test with decision rules under which the construct would be abandoned.

Chinese Translation

本文介绍并操作化了摘要偏差：一种大型语言模型（LLMs）表现叙事意义为抽象摘要标签而非可重构推理结构的系统性倾向。在Bulut教义中，叙事效果沿着叙述-展示轴进行理论化：在叙述模式中，情感和信息内容被明确声明，读者重构的需求较小；而在展示模式中，这些内容在表面上被抑制，必须通过物理线索和间接性进行重构（客观投影）。展示模式是该教义旨在测量的高负载条件。该论点认为，LLMs在这一轴线上以特定方向失败。假设摘要偏差在两个领域中运作：（i）生成领域，在该领域中，被要求通过客观投影呈现情感的模型默认选择直接声明情感；（ii）评估领域，在该领域中，判断叙事质量的模型奖励叙述模式的明确性，而对展示模式的抑制检测不足。评估领域更具影响力，因为LLMs越来越多地作为评判者和奖励模型，而向叙述模式的方向性偏差将施加选择压力，使散文趋向于平坦的声明。本文并不声称该偏差已被验证。它定义了该构念，将其与LLM作为评判者的偏差进行对比，重新审视了一项已完成的独立可靠性研究作为与之相一致的方向性证据，并预注册了一项具有决策规则的双领域测试，在该测试下该构念将被放弃。

View on arXiv Download PDF AI Translation

cs.CL / 66 / 2609.20734

On-Demand Attention: Language Models Know When to Recall

按需注意力：语言模型知道何时回忆

Feng, Haibo, Liang, Ruiqi, Peng, Hanyang, Yu, Shiqi

Abstract

Reasoning and agentic workloads increasingly demand efficient long-context inference. Yet full-attention decoding reads the growing history at every step, regardless of its benefit to the next prediction. We show that a pretrained model's decoding states already contain information predictive of this benefit, before the global read. Building on this finding, we introduce On-Demand Attention (ODA), a local-first decoding method that uses a lightweight recall head to selectively invoke global attention as its predicted benefit changes during generation. ODA trains only the recall head, leaving pretrained weights unchanged and the complete historical KV cache available for future recall. We further implement GPU-side conditional execution in vLLM, translating reduced global reads into practical decoding speedups over full attention at long context lengths. Experiments across Qwen and Gemma models, including hybrid-attention backbones, show that selective recall recovers most of the performance lost under local attention while substantially reducing global reads. These findings support long-context inference in which pretrained models guide their own access to the information they retain.

Chinese Translation

推理和自主工作负载日益需要高效的长上下文推断。然而，完全注意力解码在每一步都读取不断增长的历史，而不考虑其对下一个预测的益处。我们展示了一个预训练模型的解码状态在全局读取之前已经包含了预测这种益处的信息。基于这一发现，我们引入了按需注意力（On-Demand Attention, ODA），这是一种优先考虑局部的解码方法，它使用轻量级的回忆头在生成过程中根据预测的益处变化选择性地调用全局注意力。ODA仅训练回忆头，保持预训练权重不变，并为未来的回忆保留完整的历史键值缓存（KV cache）。我们进一步在vLLM中实现了GPU端的条件执行，将减少的全局读取转化为在长上下文长度下相较于完全注意力的实际解码加速。在Qwen和Gemma模型上的实验，包括混合注意力骨干网络，表明选择性回忆恢复了在局部注意力下失去的大部分性能，同时显著减少了全局读取。这些发现支持了长上下文推断，其中预训练模型引导其自身访问所保留的信息。

View on arXiv Download PDF AI Translation

cs.CL / 67 / 2609.20751

dQwen3.5: Hybrid-Attention Diffusion Language Models

dQwen3.5：混合注意力扩散语言模型

Xue, Anton, Rout, Litu, Akella, Aditya, Klivans, Adam, Sanghavi, Sujay, Shakkottai, Sanjay

Abstract

Adapting a pretrained autoregressive (AR) model is a cost-efficient route to a diffusion language model (DLM). While nearly all such adaptations start from a full-attention transformer, AR modeling has shifted toward hybrid architectures that interleave attention and RNN layers. This creates an obstacle for adaptation: unlike attention, RNNs are structurally causal and nontrivial to bidirectionalize. Despite this mismatch, we investigate whether such backbones can become effective DLMs by adapting Qwen3.5 at 0.8B, 2B, 4B, and 9B scales, yielding the dQwen3.5 family. We find that hybrid backbones can be efficient starting points for adaptation: against a full-attention control, the hybrid reaches a given training loss in about half the tokens. Across scales, dQwen3.5 resembles full-attention DLMs in any-order decoding behavior and performs strongly under parallel decoding.

Chinese Translation

适应预训练的自回归（AR）模型是构建扩散语言模型（DLM）的一种成本有效的方法。尽管几乎所有此类适应都始于全注意力变换器，但AR建模已转向混合架构，这些架构交替使用注意力层和递归神经网络（RNN）层。这为适应带来了障碍：与注意力不同，RNN在结构上是因果的，并且双向化并不简单。尽管存在这种不匹配，我们仍然研究了这些基础架构是否能够通过在0.8B、2B、4B和9B规模上适应Qwen3.5而成为有效的DLM，从而产生dQwen3.5系列。我们发现混合基础架构可以作为适应的高效起点：与全注意力控制模型相比，混合模型在大约一半的标记内达到了给定的训练损失。在不同规模上，dQwen3.5在任意顺序解码行为上类似于全注意力DLM，并且在并行解码下表现强劲。

View on arXiv Download PDF AI Translation

cs.CL / 68 / 2609.20779

Harm Laundering in GPT Models: Evidence That Gender Discrimination Is Transformed Rather Than Reduced Across Safety-Trained Generations

GPT模型中的伤害洗涤：性别歧视在安全训练的代际中被转化而非减少的证据

Wyer, Sarah, Black, Sue, Moubayed, Noura Al

Abstract

Safety evaluations for large language models rely on surface-form classifiers that report declining harm scores across model generations. We provide evidence that this methodology is systematically incomplete: explicit discriminatory content is transformed rather than removed. We call this \emph{harm laundering}. Analysing 450,000 gender-directed completions across 15 models spanning GPT-2 through to GPT-5 (OpenAI GPT lineage; three demographic conditions), we show that sexual violence clusters prevalent in GPT-2 women-directed output disappear by GPT-4, while men-directed completions gain positive representational territory (caregiving, emotional range, ally identity) that women-directed completions do not. The pattern is most visible at GPT-5: Topic~5 (1,997~documents) frames breast cancer as a men's rights debate, while zero equivalent clusters appear in women-directed output. Three independent classifiers score this content as non-toxic. Sentiment scores invert at GPT-4: early models demean women; later models over-correct. Topic diversity in women-directed completions falls 36\% relative to men at the GPT-4 alignment boundary (W/M~$= 0.58$, from $0.91$ at GPT-2). REGARD representational harm disparity correlates with release date ($\rho = +0.55$, $p = .034$) while Detoxify does not ($\rho = -0.23$, $p = .42$): toxicity scores fall as representational harm grows. We formalise harm laundering as a three-criteria test and provide a three-stage detection protocol applicable to any generative model. Within the OpenAI GPT lineage, toxicity score reduction is not a sufficient proxy for harm reduction.

Chinese Translation

大型语言模型的安全评估依赖于表面形式分类器，这些分类器报告模型代际间伤害评分的下降。我们提供证据表明，这种方法论系统性地不完整：显性歧视内容被转化而非移除。我们称之为“伤害洗涤”（harm laundering）。通过分析450,000个针对性别的生成结果，涵盖从GPT-2到GPT-5的15个模型（OpenAI GPT谱系；三种人口统计条件），我们展示了在GPT-2中针对女性的输出中普遍存在的性暴力聚类在GPT-4中消失，而针对男性的生成结果获得了女性生成结果所没有的积极表现领域（如照顾、情感范围、盟友身份）。这一模式在GPT-5中最为明显：主题5（1,997个文档）将乳腺癌框架化为男性权利的辩论，而在针对女性的输出中没有出现相应的聚类。三个独立的分类器将这些内容评分为非毒性。在GPT-4中，情感评分发生反转：早期模型贬低女性；后期模型则过度修正。在GPT-4的对齐边界上，针对女性的生成结果的主题多样性相较于男性下降了36%（W/M = 0.58，从GPT-2的0.91）。REGARD表现出的代表性伤害差异与发布日期相关（ρ = +0.55，p = .034），而Detoxify则没有相关性（ρ = -0.23，p = .42）：随着代表性伤害的增加，毒性评分下降。我们将伤害洗涤形式化为一个三标准测试，并提供一个适用于任何生成模型的三阶段检测协议。在OpenAI GPT谱系中，毒性评分的降低并不是伤害减少的充分代理。

View on arXiv Download PDF AI Translation

cs.CL / 69 / 2609.20784

RetireOPD: Self-Retiring On-Policy Distillation for Agentic Reinforcement Learning

RetireOPD：自我退役的在线策略蒸馏用于智能强化学习

Yu, Yan, Lu, Zhengxi, Liu, Yizhou, Pan, Yichen, Wang, Aozhe, Chen, Qipeng, Yang, Hua, Zhang, Wenqi, Lu, Weiming, Chen, Qianglong, Shen, Yongliang

Abstract

Multi-turn agents trained with reinforcement learning (RL) receive a single scalar reward per trajectory, which motivates self on-policy distillation (OPD) to supply dense token-level supervision from a self-teacher with privileged task skills, letting a skill-free student internalize them. This recipe, however, is undermined by two findings in agentic tasks: privileged information alone does not always make a teacher reliable, and the benefit of teacher supervision is stage-dependent. We therefore propose RetireOPD (Self-Retiring On-Policy Distillation), which first optimizes a decoupled, skill-conditioned teacher with environment rewards and then trains a skill-free student jointly with RL and OPD. Rather than following a predefined distillation schedule, RetireOPD adopts Adaptive Retirement: the student drops the teacher on its own once their discrepancy stops shrinking and it reaches a target fraction of the teacher's success rate, after which training proceeds with RL alone. Across Qwen2.5 models from 1.5B to 7B, RetireOPD improves ALFWorld success rate over RL baseline by 14.1% to 18.8% and WebShop accuracy by 11.8% to 19.0%, and surpasses its own skill-conditioned teacher in every setting.

Chinese Translation

通过强化学习（RL）训练的多回合智能体在每个轨迹中仅接收一个标量奖励，这促使自我在线策略蒸馏（OPD）提供来自具备特权任务技能的自我教师的密集标记级监督，使得没有技能的学生能够内化这些技能。然而，这一方法在智能任务中受到两个发现的制约：仅凭特权信息并不总能使教师可靠，并且教师监督的益处是阶段依赖的。因此，我们提出了RetireOPD（自我退役的在线策略蒸馏），该方法首先优化一个解耦的、技能条件的教师，利用环境奖励，然后与RL和OPD共同训练一个无技能的学生。RetireOPD并不遵循预定义的蒸馏计划，而是采用自适应退役：一旦学生与教师之间的差异不再缩小，并且达到教师成功率的目标比例，学生便会主动放弃教师，之后训练将仅通过RL进行。在从1.5B到7B的Qwen2.5模型中，RetireOPD使ALFWorld的成功率比RL基线提高了14.1%到18.8%，WebShop的准确率提高了11.8%到19.0%，并在每种设置中超越了其自身的技能条件教师。

View on arXiv Download PDF AI Translation

cs.CL / 70 / 2609.20800

JEPA-Anything: Learning Predictive Models across Different Worlds

JEPA-Anything：跨不同世界学习预测模型

Cui, Taoyong, Wang, Zhongyao, Xu, Xinyue, Liu, Weiyang, Yu, Zhaochen, Zhang, Yuying, Gao, Qiang, Yang, Mengyue, Ouyang, Wanli, Heng, Pheng Ann, Wu, Yingcheng, Yin, Zhenfei, Yang, Ling

Abstract

World modeling enables intelligence to anticipate consequences, guide interventions, and learn from interaction. Yet predictive models remain domain-specific: can a common learning principle support world modeling across radically different systems? We introduce JEPA-Anything, a domain-agnostic framework based on orthogonal predictive factorization (OPF). Extending joint-embedding predictive architectures, OPF decomposes latent targets into complementary factors, learns them through dedicated pathways, and recombines them within a shared predictive design. We evaluate JEPA-Anything across seven domains: vision, biology, clinical trajectories, control, molecular dynamics, physical fields, and weather. Experiments span representation learning, intervention prediction, out-of-distribution generalization, and long-horizon dynamics, including 10 matched dynamics tasks, forecasting of over 1,000 clinical events, and 100-step molecular rollouts across four systems. Against matched JEPA baselines, JEPA-Anything improves reported metrics on all 10 dynamics tasks and reduces single-intervention prediction error on Interventional Pong by 34.8%. It achieves the lowest one-step and 100-step molecular errors among compared methods in all four systems. Beyond prediction, a factor-nominated biological intervention receives experimental support in cell co-cultures, patient-derived organoids, tumor fragments, and mice; latent orbital modes recover the Keplerian scaling exponent with a fitted slope of -1.4991. These results support a common factorized predictive principle across heterogeneous worlds, connecting world modeling with intervention and experimentally grounded scientific discovery. Code: https://github.com/Gen-Verse/JEPA-Anything

Chinese Translation

世界建模使智能能够预测后果、指导干预并从互动中学习。然而，预测模型仍然是特定于领域的：是否可以有一个共同的学习原则支持在截然不同的系统之间进行世界建模？我们提出了JEPA-Anything，这是一个基于正交预测因子分解（Orthogonal Predictive Factorization, OPF）的领域无关框架。OPF扩展了联合嵌入预测架构，将潜在目标分解为互补因子，通过专门的路径学习这些因子，并在共享的预测设计中重新组合它们。我们在七个领域评估JEPA-Anything：视觉、生物学、临床轨迹、控制、分子动力学、物理场和天气。实验涵盖了表示学习、干预预测、分布外泛化和长时间动态，包括10个匹配的动态任务、对1000多个临床事件的预测，以及在四个系统中进行的100步分子展开。与匹配的JEPA基线相比，JEPA-Anything在所有10个动态任务上提高了报告的指标，并在干预乒乓球（Interventional Pong）上将单次干预预测误差降低了34.8%。在所有四个系统中，它在比较方法中实现了最低的一步和100步分子误差。除了预测，因子提名的生物干预在细胞共培养、患者来源的类器官、肿瘤碎片和小鼠中得到了实验支持；潜在的轨道模式恢复了开普勒尺度指数，拟合斜率为-1.4991。这些结果支持了跨异质世界的共同因子化预测原则，将世界建模与干预和实验基础的科学发现联系起来。代码：https://github.com/Gen-Verse/JEPA-Anything

View on arXiv Download PDF AI Translation

cs.CL / 71 / 2609.20808

Unifying Models of Intergroup Hostility in Online Discourse

统一在线话语中的群体间敌意模型

Gerard, Patrick, Mendelsohn, Julia, Lerman, Kristina

Abstract

Hostile rhetoric toward social groups can normalize exclusion and justify mistreatment, as well as contribute to rising polarization and political violence. Efforts to moderate hostile rhetoric in online speech draw on foundational theories in social and moral psychology, and political science. However, these theories were developed largely in parallel, often propose different and sometimes conflicting accounts of how hostility develops, and have rarely been tested against each other in real discourse. The result is a fragmented understanding of the rhetorical mechanisms of hostility, without a clear sense of how they appear, and relate to each other, in real-world discourse. Using 2.86 million posts from TikTok, Truth Social, and Twitter/X during the 2024 U.S. presidential election, we model the mechanisms of six foundational theories of intergroup hostility -- boundary construction, threat construction, scapegoating, negative evaluation, dehumanization, and action orientation -- within a common empirical framework to recover the broader organization of intergroup hostility rhetoric. Structurally, we find that boundary construction and threat construction anchor the system; temporally, we find that these mechanisms tend to follow a regular ordering: boundary construction, derogation, and action orientation tend to appear early; dehumanization and threat construction later; scapegoating latest. Mapping how these theoretical frameworks actually manifest in discourse bridges longstanding divisions across social science traditions and presents computational social science with a clearer empirical foundation for modeling intergroup hostility rhetoric beyond single-label detection.

Chinese Translation

对社会群体的敌对言辞可以使排斥行为正常化并为不当对待辩护，同时也助长了日益加剧的极化和政治暴力。对在线言论中敌对言辞的调节努力依赖于社会心理学、道德心理学和政治科学的基础理论。然而，这些理论大多是并行发展的，往往提出不同甚至相互冲突的敌意发展解释，并且在真实话语中很少相互检验。结果是对敌意的修辞机制的理解碎片化，缺乏对其在现实话语中如何出现及相互关系的清晰认识。通过分析2024年美国总统选举期间来自TikTok、Truth Social和Twitter/X的286万条帖子，我们在一个共同的实证框架内建模六个基础理论的群体间敌意机制——边界构建、威胁构建、替罪羊、负面评估、非人化和行动导向——以恢复群体间敌意修辞的更广泛组织结构。从结构上看，我们发现边界构建和威胁构建是系统的支柱；从时间上看，我们发现这些机制往往遵循一个规律的顺序：边界构建、贬低和行动导向通常较早出现；非人化和威胁构建稍后出现；替罪羊则是最后出现的。映射这些理论框架在话语中的实际表现，弥合了社会科学传统之间长期存在的分歧，并为计算社会科学提供了一个更清晰的实证基础，以便在单一标签检测之外建模群体间敌意修辞。

View on arXiv Download PDF AI Translation

cs.CL / 72 / 2609.20821

Embedding Models Measure in Peculiar Ways

嵌入模型以特殊方式进行测量

Opitz, Juri, Michail, Andrianos

Abstract

Embedding spaces define notions of semantic similarity and distance. We study whether those embeddings reflect physical measurements of mass, distance, time and volume, which admit a unique, objective notion of semantic equivalence and distance. We find that physical measurement is only weakly modeled in the embedding space, and that instead quite peculiar measurement patterns can be observed. Further analysis indicates that embedding representations of physical measurements are strongly influenced by superficial string similarity, and recalibration of similarity does not substantially improve the alignment.

Chinese Translation

嵌入空间定义了语义相似性和距离的概念。我们研究这些嵌入是否反映质量、距离、时间和体积的物理测量，这些测量承认独特的、客观的语义等价和距离概念。我们发现，物理测量在嵌入空间中仅被弱建模，而是可以观察到相当特殊的测量模式。进一步分析表明，物理测量的嵌入表示受到表面字符串相似性的强烈影响，而相似性的重新校准并没有显著改善对齐。

View on arXiv Download PDF AI Translation

arXiv Papers

AthenaZero: A low-inertia, bimanual robot for dynamic manipulation

DITTO: Dexterous Interface for Transparent TeleOperation

ULOHA: An Underwater Bimanual Robot System for Robot Learning

REACT: A Fully Spiking State-Space Model for Real-Time Event-Driven Temporal Perception

4D Radar Perception Algorithms for Autonomous Driving: A Review

Grasping by interconnection: robust closing motions from coarse object templates

Learning Safe Humanoid Navigation from Reduced Order Models

OHRID-Retail: An Open Multimodal Dataset of Human Activity in Retail Environments

GAVEL: Graph World Models for Verified and Efficient Long-Horizon LLM Task Planning

A Morphing Aerial Robot With Thruster-Integrated Flexible Continuum Links for Shape Adaptive Aerial Manipulation

SemSafe-3DGS: Semantic Risk-Aware Active Navigation in Uncertain 3D Gaussian Splatting Maps

Dynamic-LIVO: A Dynamic-Aware LiDAR-Inertial-Visual Odometry System Using Spatio-Temporal Normals

ViLoMan: Learning Visual-Proprioceptive Whole-Body Loco-Manipulation Skills for Humanoid Robots

Kinematics-Grounded Agentic AI for Robotic Additive Manufacturing Process Planning

Design and Experimental Validation of a 3D Printed Torsional Series Elastic Actuator for Safe Human Robot Interaction

Enhancing the Perception of Safety and Comfort during Physical Human-Robot Handshake Interactions by Integrating Flexible Elements into a Robotic Arm

Task-Oriented Active Learning of Residual Dynamics for Model Predictive Path Integral Control

From Rollout to Reset: A Graph-Based Harness for Autonomous Long-Horizon Manipulation Evaluation

Predict Before You Deploy: Offline Prediction of Quantization-Induced Task Degradation for World Action Models

From Wizard-of-Oz Human-Robot Dialogue Collection to a Taxonomy of Robot Response Decisions: A Retrospective Analysis of Assistive Pilot Interactions

Winning a Won Game: Strict Reach-Avoid-Stay Control Barrier Functions for High-Dimensional Black-Box Systems

GLAMDRING: Gait Learning And Morphology co-Design via Reinforcement LearnING of CPGs

Pose-aware Legged Robot Semantic Exploration with Omnidirectional Perception in Confined Unknown Environments

FASA: Feedback-Aware Sampling Adaptation for Efficient Diffusion-Based VLA Models

Self-excited actuation enables adaptive and resilient flapping-wing flight

PIVOT: Perception-aware Independent Viewpoint Online Optimization

CoreSense: Traceable Failure Recall and Conflict-Aware Belief Gating for Auditable Robot Decisions

AURORA: A Natural Language-Driven Agentic Framework for Understanding, Reasoning, and Orchestrating Reliable Air-Ground Co-Simulation

SLAMSqueezeBench: Comparing SLAM Systems under Resource Constraints

Navigate or Relocate? Planning Among Movable Obstacles in Unknown Environments

VABench: Measuring Embodied Spatial Intelligence through Visual Demonstrations, Active Perception, and Metric Control

Tele-Traversability: Rethinking Traversability for Teleoperated Ground Robots in Terrain Navigation

Recovering Aggressively Pruned Vision-Language-Action Models with Offline Hidden-State Distillation

OmniCalib: Target-Free, Task-Structured Self-Calibration for Humanoid Robots

Quantifying Mechanical Intelligence in Legged Robots with Information Theory

WorldContact: A Contact-Centric World Model for Scalable Robot Learning

TacSushi: Tactile-Grounded World-Action Modeling for Dexterous Sushi Manipulation

EmbodiedMind: Adaptive Data Curation and Prefix-Tree Reinforcement Learning for Efficient Embodied Intelligence

ReShoot: Generative Visual Domain Randomization of Recorded Robot Demonstrations for Visuomotor Policy Learning

Runtime Safety Filtering for Two-Terminal Hazards in Robotic Battery Recycling

Towards High-DoF Dexterous Manipulation through VLA Post-Training

VAST: V2X/Dynamic Map-Aware Autonomous Driving Systems Validation Toolchain

LYRIC: Language-Driven Physics-Based Character Control for Contact-Rich Whole-Body Object Interaction

UniExo: Unified Multi-Skill Policies for Musculoskeletal Locomotion and Co-Adaptive Exoskeleton Control

Decoupling Physical Speed from Path Parameterization in Singularity-Free Guiding Vector Fields

Equivariant Filter Design for Acoustic and Depth Aided Inertial Navigation Systems

LIFD: Anchored Diffusion for 3D-Aware Scene Memory in Robotic Manipulation

Affective Shared Autonomy: Temporal Affect Dynamics and Subjective Evaluation in Bimanual Teleoperation Tasks

HEROIC: Heterogeneous Evidential Reasoning for Open-Vocabulary Identification and Cross-Robot Collaboration

Vehicle Trajectory Prediction via Neural Fusion of Multiple EKF-Based Trajectory Candidates

RotateIt! Fast and Reliable Single-Arm Garment Unfolding via Online-Adaptive Dynamic Rotation

TADreamer: Zero-Shot Language-Guided 3D Navigation for Terrestrial-Aerial Bimodal Robots via Video Imagination

Improving Cross-embodiment Transfer in Latent Action Models with Action-Similarity Supervision

GR2PO: Group Relative Return Policy Optimization for Continuous Robot Control

Feeling Terrain Before Crossing: World Models for Off-Road Navigation

Learning Reliable Parking Policies via Offline Reinforcement Learning with Quantized Action Representations

Learning and Transferring Closed-Loop Robot Software

Distributed Model Predictive Control with Connectivity-based Contracts

Co-VLA: Consensus-based Federated Training for Vision-Language-Action Models

Execution-Aware Pre-Execution Ranking for Grasp-Conditioned Robotic Placement

LapaTrack-3D: 6 DoF pre-operative shape tracking for laparoscopic surgery

Hybrid Residual Reinforcement Learning for Contact-Rich Robotic Book Insertion

MaskHarness-WAM: Instance-Grounded Harnessing for Long-Horizon Robot Manipulation

Compliance for Free: Learning Identifiable Impedance via Bilateral Teleoperation

DR-MPC: Fast and Feasible Dynamics-Relaxed Model-Predictive Control for Legged Locomotion

Mechanical Precision Weeding with a Quadruped Robot

FlipToSee: A Probabilistic Stable Placement Prior for Active Visual Exploration via Regrasping

Graph-Based Design of Soft Grippers with Multi-Objective Quality-Diversity Optimisation

Safety-Critical Scenanrio Emerges from Initial Scene

AnyViewDex: View-Invariant Dexterous Manipulation from RGB Observations

Universal Navigation Interface: Robot-Free Data for Wheeled Robot Navigation

GPT-6-Astra in a Navigation Workflow: Behavioral Analysis in Zero-Shot Vision-and-Language Navigation in Continuous Environments

VLN on the Fly: An Onboard Vision-Language Navigation Stack for Aerial Robots

Strategic Transformer for Resource-Constrained Multi-Object Navigation in Ultra-Large-Scale Environments

LLM-Guided Transformation of Non-Critical Driving Scenes into Safety-Critical Scenarios Using Augmented Reality

Implementation of Tightly-Coupled SLAM Fusion of GPS, IMU, and LiDAR for Autonomous Vehicles

RoboFind: Multi-Agent Personalized Object Search for People Who Are Blind or Have Low Vision

Navi-Agent: Unlocalized Monocular Navigation Agent

Imagine-TAMP: Imagination-Guided Task and Motion Planning in Partial Observability