cs.RO / 1 / 2610.08800
A Review Of Robotic World Models For Dynamic Environments Based On Factor And Scene Graphs
基于因子图和场景图的动态环境机器人世界模型综述
Abstract
Models based on graphs have emerged in robotics as a powerful foundation for internal world representations, where factor and scene graphs are among the most prominent model types found in the related literature and in successful robotic solutions. Initially, many of these models were assuming static environments as a simplification. Herein, factor graphs mainly provide uncertainty-aware geometric estimations while scene graphs enable a structured semantic abstraction. However, real-world robotic environments are often dynamic, posing severe challenges for purely static world representations. Therefore, this review presents a comprehensive view on how dynamic aspects of real-world environments can be addressed in such graph-based world models. We organize our assessments around three main aspects: (I) suitable representations, (II) pipelines to construct and update the representations, and (III) their exploitation for downstream tasks. We review approaches that are either based on factor or scene graphs, but put special emphasis on novel approaches that combine both types to form hybrid models. We mainly analyze how different types of dynamics can be modeled herein, and categorize common architectural patterns. Finally, emerging trends and open challenges are identified, including uncertainty propagation from learned perception through the representation layers, the observability of dynamic-entity motion and scale under minimal sensing, scalable lifelong maintenance, and the lack of datasets and evaluation protocols that ground world-model quality in downstream task performance under dynamics.
Chinese Translation
基于图的模型在机器人领域中作为内部世界表征的强大基础而出现，其中因子图和场景图是相关文献和成功机器人解决方案中最突出的模型类型。最初，许多这些模型假设环境是静态的，以简化问题。在此，因子图主要提供对不确定性的几何估计，而场景图则实现结构化的语义抽象。然而，现实世界中的机器人环境往往是动态的，这对纯静态世界表征提出了严峻的挑战。因此，本综述全面探讨了如何在此类基于图的世界模型中处理现实环境的动态特征。我们围绕三个主要方面组织评估：（I）合适的表征，（II）构建和更新表征的流程，以及（III）它们在下游任务中的应用。我们回顾了基于因子图或场景图的方法，但特别强调了将两种类型结合形成混合模型的新方法。我们主要分析了如何在此处建模不同类型的动态，并对常见的架构模式进行了分类。最后，识别出新兴趋势和开放挑战，包括从学习的感知通过表征层的不确定性传播、在最小感知下动态实体运动和尺度的可观测性、可扩展的终身维护，以及缺乏能够在动态下基于下游任务性能评估世界模型质量的数据集和评估协议。
cs.RO / 2 / 2610.08802
SAFE: Unified Slip and Fracture Detection with Low-Cost Acoustic Sensing in Robotic Grasping
SAFE：在机器人抓取中利用低成本声学传感实现统一的滑动和断裂检测
Abstract
Manipulating fragile objects remains challenging as robots must understand the state of what they grasp, such as slip or fracture, to respond appropriately, especially when material properties are unknown. In this paper, we present SAFE: a low-cost, general-purpose sensing approach that detects both slip and fracture in real time using two passive polyvinylidene fluoride (PVDF) acoustic sensors and motor proprioception, without relying on vision or prior material knowledge. The sensors are mounted on a compliant Fin Ray gripper, and a unified HistGradientBoosting classifier reports the state (normal, slip, or fracture) from a 79-dimensional feature vector. Under leave-one-grasp-out cross-validation, SAFE achieves an Alert-F1 of 0.884 with near-zero slip-fracture confusion, and ablations confirm that acoustic sensing is indispensable. An adaptive grasp controller built on this detection layer runs at 104 Hz on a Jetson Orin Nano, achieving 91.3% success across 46 closed-loop robot trials spanning diverse object categories, while each fixed-force strategy drops to 0% on object conditions that mismatch its preset. It further reaches 82.4% success on novel objects unseen during training, demonstrating robust, failure-aware grasp control without object-specific calibration.
Chinese Translation
操控易碎物体仍然具有挑战性，因为机器人必须理解其抓取物体的状态，例如滑动或断裂，以便做出适当反应，特别是在材料特性未知的情况下。本文提出了SAFE：一种低成本、通用的传感方法，利用两个被动聚偏二氟乙烯（PVDF）声学传感器和电机本体感知实时检测滑动和断裂，而无需依赖视觉或先前的材料知识。传感器安装在一个柔性Fin Ray夹持器上，统一的HistGradientBoosting分类器根据79维特征向量报告状态（正常、滑动或断裂）。在留一抓取交叉验证下，SAFE实现了0.884的Alert-F1，几乎没有滑动与断裂的混淆，消融实验确认声学传感是不可或缺的。基于这一检测层构建的自适应抓取控制器在Jetson Orin Nano上以104 Hz的频率运行，在46次涵盖多种物体类别的闭环机器人试验中实现了91.3%的成功率，而每种固定力策略在与其预设不匹配的物体条件下成功率降至0%。它在训练期间未见过的新物体上进一步达到了82.4%的成功率，展示了强大的、具备故障意识的抓取控制能力，而无需针对特定物体的校准。
cs.RO / 3 / 2610.08807
Context-Aware Adaptive Pesticide Spraying for Agricultural Robots under Changing Weather and Terrain Using Vision-Language Models
基于视觉-语言模型的农业机器人在变化天气和地形下的上下文感知自适应喷 pesticide 喷洒
Abstract
Precision pesticide spraying is essential for optimizing application efficiency and ensuring uniform chemical distribution. Spraying performance is influenced by multiple factors, including environmental conditions such as temperature and wind speed, pesticide type, and the robot's capability to accurately perceive crops and target spray locations. Existing approaches predominantly emphasize crop detection and rely on predefined spraying parameters, whereas human operators dynamically adjust their spraying strategies by considering environmental conditions, region-specific crop characteristics, and the type of pesticide being applied. In this study, we propose a context-aware adaptive spraying framework based on Vision-Language Models (VLMs), which enables robots to leverage spatial reasoning and integrate information from multiple sources, including crop type, pesticide type, and weather data, to make adaptive and optimized spraying decisions. Subsequently, a trajectory-tracking controller based on Model Predictive Path Integral (MPPI) control is employed to ensure precise navigation and accurate spraying at crop locations. The comparative results demonstrate that the proposed method improves accuracy by at least 30% in detecting crop rows. In addition, the experimental evaluations conducted in two environments further demonstrate the robot's ability to flexibly adjust spraying volume and travel speed, while reducing pesticide drift.
Chinese Translation
精准的 pesticide 喷洒对于优化应用效率和确保化学品均匀分布至关重要。喷洒性能受到多种因素的影响，包括温度和风速等环境条件、 pesticide 类型以及机器人准确感知作物和目标喷洒位置的能力。现有的方法主要强调作物检测，并依赖于预定义的喷洒参数，而人类操作员则通过考虑环境条件、特定区域的作物特性和所施用 pesticide 的类型，动态调整喷洒策略。在本研究中，我们提出了一种基于视觉-语言模型（Vision-Language Models, VLMs）的上下文感知自适应喷洒框架，使机器人能够利用空间推理，并整合来自多种来源的信息，包括作物类型、 pesticide 类型和天气数据，以做出自适应和优化的喷洒决策。随后，采用基于模型预测路径积分（Model Predictive Path Integral, MPPI）控制的轨迹跟踪控制器，以确保在作物位置的精确导航和准确喷洒。比较结果表明，所提方法在检测作物行方面提高了至少 30% 的准确性。此外，在两个环境中进行的实验评估进一步证明了机器人灵活调整喷洒量和行驶速度的能力，同时减少了 pesticide 漂移。
cs.RO / 4 / 2610.08812
Taming an End-to-End Autonomous Driving Policy for Urban Navigation of Quadruped Robots
驯化端到端自主驾驶策略以实现四足机器人在城市导航中的应用
Abstract
We present Go2-DrivoR, a goal-conditioned adaptation of the end-to-end autonomous driving trajectory planning framework DrivoR for urban navigation with quadrupedal robots. By conditioning trajectory generation on a local-frame subgoal through a goal token and adapting the vehicle-centric scoring formulation, the method extends DrivoR to short-horizon goal-conditioned local planning without redesigning its core decoders. Specifically, we redefine drivable-area compliance for sidewalk-oriented navigation and reformulate the original ego progress term as goal-conditioned ego progress. Trained exclusively on TartanGround simulation data, Go2-DrivoR improves waypoint-conditioned planning performance on unseen simulation environments and transfers zero-shot to open-loop real-world trajectory prediction.
Chinese Translation
我们提出了Go2-DrivoR，这是一个基于目标条件的适应性框架，将端到端自主驾驶轨迹规划框架DrivoR应用于四足机器人在城市导航中的任务。通过通过目标标记对局部框架子目标进行条件化轨迹生成，并调整以车辆为中心的评分公式，该方法扩展了DrivoR，使其能够在不重新设计核心解码器的情况下进行短期目标条件的局部规划。具体而言，我们重新定义了适合人行道导航的可行驶区域合规性，并将原始自我进展项重新公式化为目标条件的自我进展。在仅使用TartanGround仿真数据进行训练的情况下，Go2-DrivoR在未见过的仿真环境中提高了基于航点的规划性能，并实现了零样本转移到开放环路的真实世界轨迹预测。
cs.RO / 5 / 2610.08852
Autonomous Droplet Navigation via Model-Based Reinforcement Learning: Zero-Shot Transfer and Emergent Dynamics
基于模型的强化学习实现自主液滴导航：零样本迁移与涌现动力学
Abstract
Self-driving laboratories (SDLs) are transforming chemical and materials discovery through closed-loop automation, yet automated infrastructure for physical manipulation of soft, deformable matter remains beyond current robotic platforms. A critical instance is autonomous droplet transport on an open surface, where contact-angle hysteresis, capillary pinning, and surface heterogeneity produce partially observable dynamics that pose significant challenges for classical model-based controllers. We introduce the first robotic platform for closed-loop autonomous liquid droplet navigation on an open, unconfined surface using model-based reinforcement learning. A two-axis tilting board coated with a thin silicone oil film drives the droplet, while an overhead camera provides real-time feedback. A learned policy was trained on just 50 to 150 physical episodes depending on geometric complexity, without simulation or analytical models. Beyond performance alone, the platform demonstrates three capabilities of interest to the SDL community: it robustly transfers zero-shot to unseen geometries; it autonomously discovers an oscillatory depinning strategy to free the droplet when it sticks; and it completes its full training pipeline in under 90 minutes. These results extend reinforcement-learning manipulation from rigid microrobots to deformable soft-matter systems for next-generation SDLs.
Chinese Translation
自驱动实验室（SDL）通过闭环自动化正在改变化学和材料的发现，然而，针对软性可变形物质的物理操作的自动化基础设施仍超出当前机器人平台的能力。一种关键情况是开放表面上的自主液滴运输，其中接触角滞后、毛细钉扎和表面异质性产生部分可观测的动态，这对经典的基于模型的控制器构成了重大挑战。我们介绍了第一个用于开放、无约束表面上闭环自主液滴导航的机器人平台，该平台采用基于模型的强化学习。一个涂有薄硅油膜的双轴倾斜板驱动液滴，而一个悬挂摄像头提供实时反馈。学习策略的训练仅依赖于50到150个物理实验，具体数量取决于几何复杂性，且没有使用仿真或解析模型。除了性能本身，该平台还展示了三个对SDL社区感兴趣的能力：它能够稳健地实现零样本迁移到未见几何形状；它自主发现了一种振荡去钉策略，以在液滴粘附时将其释放；并且它在90分钟内完成了整个训练流程。这些结果将强化学习操作从刚性微型机器人扩展到可变形软物质系统，为下一代SDL提供了新的可能性。
cs.RO / 6 / 2610.08862
CIRRA: Dual-Level Continual Instruction Reconciliation with Ongoing Execution for Embodied Robot Agents in Interactive Household Tasks
CIRRA：面向交互式家庭任务的具身机器人代理的双层持续指令调和与持续执行
Abstract
Household robots must accommodate new user instructions while executing ongoing tasks. Existing agents often regenerate or extensively revise the remaining task sequence, introducing plan ambiguity, logical inconsistency, and redundant execution. We formulate continual instruction reconciliation and propose CIRRA (Continual Instruction Reconciliation for Robot Agents), a dual-level framework combining LLM-based semantic reasoning with rule-constrained structural integration. CIRRA first grounds incoming instructions to unique executable skills and resolves underspecified actions and execution locations. It then preserves the ongoing subtask sequence as an execution backbone and generates integration candidates by inserting incoming subtasks into location-matched segments. The semantic reasoner evaluates only modified segments to identify dependencies and conflicts and select the most logically coherent local integration. This structure-preserving process maintains alignment with ongoing execution, mitigates ambiguity and inconsistency, and reuses shared subtasks to reduce redundant execution. We also introduce CHIRP (Continual Household Instruction Reconciliation and Planning), a text-based benchmark of 120 episodes across eight household environments and six categories of everyday activities. On CHIRP, CIRRA achieves 74.2% decision agreement, exceeding the strongest replanning baseline by 30 percentage points; every correct fusion decision yields a correctly placed, conflict-free schedule. On a Unitree G1 humanoid, CIRRA interrupts ongoing skills at the correct moment in every trial and significantly outperforms all baselines on every metric.
Chinese Translation
家庭机器人必须在执行正在进行的任务时适应新的用户指令。现有的代理通常会重新生成或大幅修订剩余的任务序列，从而引入计划模糊性、逻辑不一致性和冗余执行。我们提出了持续指令调和的概念，并提出了CIRRA（持续指令调和机器人代理），这是一个双层框架，结合了基于大型语言模型（LLM）的语义推理与规则约束的结构整合。CIRRA首先将传入的指令与独特的可执行技能进行关联，并解决不明确的动作和执行位置。然后，它将正在进行的子任务序列作为执行主干，并通过将传入的子任务插入位置匹配的段落来生成整合候选项。语义推理器仅评估修改过的段落，以识别依赖关系和冲突，并选择最具逻辑一致性的局部整合。这个保持结构的过程与正在进行的执行保持一致，减轻了模糊性和不一致性，并重用共享子任务以减少冗余执行。我们还引入了CHIRP（持续家庭指令调和与规划），这是一个涵盖八个家庭环境和六类日常活动的120集文本基准。在CHIRP上，CIRRA实现了74.2%的决策一致性，超过了最强重规划基线30个百分点；每一个正确的融合决策都产生了一个正确放置且无冲突的时间表。在Unitree G1人形机器人上，CIRRA在每次试验中都能在正确的时刻中断正在进行的技能，并在每个指标上显著优于所有基线。
cs.RO / 7 / 2610.08863
Trajectory Planning without Trajectory Data: A Manifold-Guided Approach
无需轨迹数据的轨迹规划：一种流形引导的方法
Abstract
A common way for trajectory planning is to leverage generative models trained on large collections of expert trajectories. At inference time, the model generates executable trajectories by conditioning on task goal constraints. However, trajectory-based methods rely on costly supervision, scale poorly with sequence length, and often generalize poorly to unseen constraints such as novel start-goal pairs. We propose an alternative to learn the underlying state-space manifold and use the geometry of the manifold for trajectory planning. This approach requires only state observations and enables generalization to unseen constraints by con- structing trajectories on the learned manifold of the state space. Experiments on Maze2D and robotic motion-planning benchmarks show that Ariadne constructs feasible paths from state-only supervision and generalizes to unseen start-goal combinations. On high-dimensional dual-arm planning, it remains competitive with trajectory-supervised and classical planners, while requiring no trajectory data for training.
Chinese Translation
轨迹规划的常见方法是利用在大量专家轨迹上训练的生成模型。在推理时，该模型通过对任务目标约束进行条件化生成可执行的轨迹。然而，基于轨迹的方法依赖于昂贵的监督，随着序列长度的增加而扩展性差，并且通常对未见约束（如新颖的起始-目标对）泛化能力较差。我们提出了一种替代方法，通过学习潜在的状态空间流形，并利用流形的几何特性进行轨迹规划。这种方法仅需状态观测，并通过在学习到的状态空间流形上构建轨迹来实现对未见约束的泛化。在Maze2D和机器人运动规划基准上的实验表明，Ariadne能够仅通过状态监督构建可行路径，并对未见的起始-目标组合进行泛化。在高维双臂规划中，它在与轨迹监督和经典规划器的竞争中表现出色，同时在训练中不需要轨迹数据。
cs.RO / 8 / 2610.08870
From Wearable Interfaces to Dexterous Policies: Contact Shifts and Tactile Representations
从可穿戴接口到灵巧策略：接触变化与触觉表征
Abstract
Unlike conventional teleoperation, wearable interfaces allow operators to collect dexterous demonstrations through their own hand motions while directly interacting with task objects. This direct interaction reduces dependence on the target robot during collection, but it also makes the collection hardware part of the physical process that generates each demonstration. Interface geometry can influence both how a task is performed and what tactile observations are recorded for learning. We study two versions of a DexUMI-family exoskeleton that share the same robot command definition, mapping procedure, and tactile module type but differ in hand-side geometry. The revised interface reduces reported physical demand, improves selected device ratings, enables tactile access in a precision grasp that is mechanically blocked by the baseline, and produces task-dependent changes in recorded contact. Lid twisting primarily exhibits a change in contact location, whereas egg carton opening primarily exhibits a change in contact occurrence. We then train matched policies using either a binary aggregate input or a spatial-plus-force input. On lid twisting and egg carton opening, policies trained on revised-interface demonstrations achieve higher success than those trained on baseline demonstrations, with a significant pooled interface effect. Across these and two additional tasks, USB insertion and soldering tool pick and place, the spatial-plus-force input likewise outperforms binary aggregation. These results show that wearable collection hardware is part of the data-generation process for robot learning. Such interfaces should therefore be evaluated not only through operator experience, but also through the tactile interactions they make recordable and the policy performance their demonstrations support.
Chinese Translation
与传统的遥操作不同，可穿戴接口允许操作员通过自身的手部动作在直接与任务对象互动的同时收集灵巧演示。这种直接互动减少了在收集过程中的对目标机器人的依赖，但也使得收集硬件成为生成每个演示的物理过程的一部分。接口几何形状可以影响任务的执行方式以及为学习记录的触觉观察。我们研究了两种版本的DexUMI系列外骨骼，它们共享相同的机器人命令定义、映射过程和触觉模块类型，但在手部几何形状上有所不同。改进后的接口减少了报告的物理需求，提高了所选设备的评分，使得在基线条件下机械阻挡的精确抓取中实现了触觉访问，并在记录的接触中产生了与任务相关的变化。盖子扭动主要表现为接触位置的变化，而打开蛋 carton 则主要表现为接触发生的变化。随后，我们使用二元聚合输入或空间加力输入训练匹配的策略。在盖子扭动和打开蛋 carton 的任务中，基于改进接口演示训练的策略比基于基线演示训练的策略取得了更高的成功率，并且存在显著的聚合接口效应。在这两个任务以及另外两个任务（USB 插入和焊接工具的拾取与放置）中，空间加力输入同样优于二元聚合。这些结果表明，可穿戴收集硬件是机器人学习数据生成过程的一部分。因此，这类接口的评估不仅应通过操作员的体验进行，还应通过它们所记录的触觉互动和它们所支持的策略表现进行评估。
cs.RO / 9 / 2610.08933
Toward Evidence-Driven Human-Agent-Robot Teaming for Earth-Independent Anomaly Triage
面向证据驱动的人机机器人团队协作以应对地球独立的异常分类
Abstract
Deep-space crews cannot rely on real-time ground support for urgent off-nominal events. Initial alerts may underdetermine cause, while discriminating evidence may reside in crew observations or at locations that are unsafe, costly, or unavailable for crew inspection. We present an evidence-driven architecture for human-agent-robot teaming in Earth-independent anomaly triage. Agentic AI is treated as a stateful coordinator over bounded, inspectable services rather than as a fully autonomous vehicle controller. A triage state manager maintains hypotheses, evidence provenance, uncertainty, operational context, and tool status; a crew-facing embodied agent elicits observations and explains assessment changes; and a mobile robot acquires targeted, localized evidence. Typed interfaces separate dialogue and orchestration from monitoring, robot command, context retrieval, and safety-critical control. Two scenarios illustrate the architecture: a crewed deep-space mission based on an actual ISS ammonia false alarm, where suspected contamination restricts crew access, and a power-interface anomaly at a crewed lunar base, where robotic inspection distinguishes a local connector fault from other causes ambiguous in remote telemetry. Our main contribution is an authority-bounded closed evidence-loop architecture, exercised in a hardware-in-the-loop integration prototype using Reachy Mini and an Innate MARS mobile robot.
Chinese Translation
深空任务的船员无法依赖实时的地面支持来应对紧急的非正常事件。初步警报可能无法确定原因，而有价值的证据可能存在于船员的观察中，或位于不安全、成本高昂或无法供船员检查的位置。我们提出了一种证据驱动的人机机器人团队协作架构，用于地球独立的异常分类。在该架构中，智能代理被视为对有限、可检查服务的有状态协调者，而非完全自主的车辆控制器。分类状态管理器维护假设、证据来源、不确定性、操作背景和工具状态；面向船员的具身代理引导观察并解释评估变化；移动机器人获取针对性的、局部的证据。类型化接口将对话和协调与监控、机器人指令、上下文检索和安全关键控制分开。两个场景展示了该架构：一个基于实际国际空间站氨气误报的载人深空任务，其中怀疑的污染限制了船员的接触；另一个是在载人月球基地的电源接口异常，机器人检查区分了局部连接器故障与远程遥测中模糊的其他原因。我们的主要贡献是一个权威边界的闭合证据循环架构，该架构在使用Reachy Mini和Innate MARS移动机器人的硬件在环集成原型中得以实现。
cs.RO / 10 / 2610.08970
HULK: Learning Whole-Body Forceful Loco-Manipulation for Humanoids
HULK：为类人机器人学习全身强力运动操控
Abstract
Humanoid loco-manipulation of large, heavy objects demands forceful interaction across the entire body. However, such payloads shift a humanoid's center of mass and impose sustained loads across the upper body, challenging balance and command tracking. We present HULK, a whole-body control framework for forceful loco-manipulation. Using model predictive control (MPC) to guide reinforcement learning with predictions of the loaded dynamics, we train two teachers: one tracks arm motions under wrist forces, and the other locomotes while holding large objects against the body. A capture-point control barrier function augments the wrist-force teacher during training to improve balance under load. We distill both teachers into a single policy. Evaluation spans simulation and the Unitree G1. In simulation, the teacher with the barrier function achieves the lowest forward and lateral velocity tracking errors at 10 kg per arm among evaluated controllers and reduces aggregate divergent component of motion (DCM) excursion magnitude by 35.7% relative to MPC-guided reinforcement learning alone. Our wrist-force teacher withstands torso push disturbances of up to 130 N.
Chinese Translation
类人机器人对大型重物的运动操控需要全身进行强力互动。然而，这些负载会改变类人机器人的质心，并在上半身施加持续的负荷，从而挑战平衡和指令跟踪。我们提出了HULK，一个用于强力运动操控的全身控制框架。通过使用模型预测控制（MPC）来指导强化学习，并预测负载动态，我们训练了两个教师：一个在手腕施加力量的情况下跟踪手臂运动，另一个在身体侧面持有大型物体时进行移动。一个捕获点控制障碍函数在训练期间增强了手腕力量教师，以改善负载下的平衡。我们将两个教师提炼为一个单一策略。评估涵盖了仿真和Unitree G1。在仿真中，具有障碍函数的教师在每个手臂10公斤的负载下实现了评估控制器中最低的前向和侧向速度跟踪误差，并将相对于仅使用MPC指导的强化学习的整体偏离运动成分（DCM）偏移幅度减少了35.7%。我们的手腕力量教师能够承受高达130 N的躯干推力干扰。
cs.RO / 11 / 2610.08995
PhysEvo: Astra Can Act, Let It
PhysEvo：让Astra行动
Abstract
Astra can act, yet reliable manipulation depends on the system through which it observes and controls the world. We introduce PhysEvo, a framework for physical recursive self-improvement (RSI) around a single frozen model. A task agent executes robot tasks; a meta-agent uses the resulting trajectories to diagnose failures, revise tools and skills, and test corrections. The meta-agent can also improve its own diagnostic tools, so retained revisions support both later action and later self-improvement. This process develops joint-level control, evidence-seeking observation, and reusable manipulation skills without model-weight updates or a separately trained action policy. Across 42 RoboDojo tasks, held-out-layout evaluation of retained task-specific deployment versions yields a five-dimension average score of 68.14/100 and 62.00% success, compared with 47.17% for RoboDawn's one-shot Astra agent, the strongest published reference in our comparison. On eight manipulation tasks challenging direct Astra, PhysEvo achieves 55.00% success, compared with 1.25% for the direct-Astra reference. Deploying the simulation-evolved harness on AgileX PiPER and continuing skill revision yields 90.60/100 average score and 84.00% success across 25 trials on five real-world tasks. PhysEvo turns the consequences of action into persistent, testable changes to how a frozen model acts and improves.
Chinese Translation
Astra能够行动，但可靠的操控依赖于其观察和控制世界的系统。我们介绍了PhysEvo，这是一个围绕单一固定模型进行物理递归自我改进（RSI）的框架。任务代理执行机器人任务；元代理利用生成的轨迹来诊断故障、修订工具和技能，并测试修正。元代理还可以改进其自身的诊断工具，因此保留的修订支持后续的行动和后续的自我改进。这个过程发展了关节级控制、寻证观察和可重用的操控技能，而无需更新模型权重或单独训练的行动策略。在42个RoboDojo任务中，保留的任务特定部署版本的布局评估平均得分为68.14/100，成功率为62.00%，相比之下，RoboDawn的一次性Astra代理在我们的比较中是最强的已发布参考，其成功率为47.17%。在八个直接挑战Astra的操控任务中，PhysEvo的成功率为55.00%，而直接Astra参考的成功率仅为1.25%。在AgileX PiPER上部署经过模拟演化的工具，并继续修订技能，在五个真实世界任务的25次试验中，平均得分为90.60/100，成功率为84.00%。PhysEvo将行动的后果转化为对固定模型的持久、可测试的变化，从而实现其行为和改进。
cs.RO / 12 / 2610.09055
MimicX: Policy-in-the-Loop Supervision Refinement for Video-Driven Humanoid Motion Tracking
MimicX：基于策略的监督循环精炼用于视频驱动的人形运动跟踪
Abstract
Human videos provide rich motion targets for humanoid learning, yet visually plausible references can still produce persistent failures under physics-based execution. These failures reveal where training supervision should change. We present MimicX, a policy-in-the-loop framework that uses execution feedback to refine video-driven humanoid motion tracking. Starting from reconstructed and retargeted motion, MimicX localizes difficult transitions and affected body regions, then jointly adapts tracking objectives and the reset curriculum for policy continuation. Repeated rollout verification selects execution-priority improvements subject to tracking guards. Across four core video tasks, MimicX consistently improves tracking accuracy and Robust Execution Horizon relative to the Fixed Reference baseline. Task-averaged results show a 25.7% reduction in body-tracking error and a 255.6% increase in execution horizon. Additional video, motion-reference, and collision-scene studies evaluate the method beyond the core tasks, while MimicX-HLoop accelerates feedback through heterogeneous execution. Overall, MimicX turns policy failure into actionable supervision for deciding what to refine and which refinement to retain.
Chinese Translation
人类视频为人形学习提供了丰富的运动目标，但在基于物理的执行下，视觉上合理的参考仍然可能导致持续的失败。这些失败揭示了训练监督需要改变的地方。我们提出了MimicX，一个基于策略的循环框架，利用执行反馈来精炼视频驱动的人形运动跟踪。MimicX从重建和重新定向的运动开始，定位困难的过渡和受影响的身体区域，然后共同调整跟踪目标和策略延续的重置课程。重复的回放验证选择受跟踪保护的执行优先级改进。在四个核心视频任务中，MimicX始终相对于固定参考基线提高了跟踪准确性和鲁棒执行范围。任务平均结果显示，身体跟踪误差减少了25.7%，执行范围增加了255.6%。额外的视频、运动参考和碰撞场景研究评估了该方法在核心任务之外的表现，而MimicX-HLoop通过异构执行加速反馈。总体而言，MimicX将策略失败转化为可操作的监督，以决定需要精炼的内容和保留的精炼方式。
cs.RO / 13 / 2610.09117
Workhorse: Learning Robust Whole-Body Humanoid Loco-Manipulation from Human Data
工作马：从人类数据中学习稳健的全身类人机器人运动操控
Abstract
Humanoid robots still struggle to plan contact-rich whole-body manipulation from egocentric RGB and proprioception. Workhorse learns such manipulation from robot-free human demonstrations. A visual planner predicts five-link targets: the poses of the torso, both wrists, and both feet. A reinforcement-learning whole-body tracker follows them on the robot. Both policies train separately on the same recorded human poses, without retargeting. We augment the training data of each policy to imitate the errors that the other makes at deployment. On a real Unitree G1, Workhorse sorts boxes with its hands and a kick, catches a thrown box, and topples and climbs a suitcase. During box sorting, we show recoveries after a person pushes the robot or takes the box away. In a simulated copy of the demonstration room, the system completes box sorting in 77% of episodes, and in 64% under 40 N.s pushes. With both policies retrained from the same demonstrations, a simulated second humanoid completes box sorting in 83% of episodes without pushes.
Chinese Translation
类人机器人在从自我中心的RGB图像和本体感觉中规划接触丰富的全身操控方面仍然面临挑战。工作马（Workhorse）通过无机器人的人类示范学习这种操控。一个视觉规划器预测五个链接目标：躯干、双手腕和双脚的姿态。一个强化学习的全身跟踪器在机器人上跟踪这些目标。这两种策略在相同的记录人类姿态上分别训练，而不进行重定向。我们增强了每种策略的训练数据，以模仿另一种策略在部署时所犯的错误。在真实的Unitree G1机器人上，工作马用手和踢的方式对箱子进行分类，接住投掷的箱子，并推倒和攀爬一个行李箱。在箱子分类过程中，我们展示了在一个人推机器人或拿走箱子后机器人的恢复能力。在一个模拟的示范室中，该系统在77%的实验中完成了箱子分类，在承受40 N.s推力的情况下完成了64%。在从相同示范中重新训练的两种策略下，一个模拟的第二个类人机器人在没有推力的情况下在83%的实验中完成了箱子分类。
cs.RO / 14 / 2610.09130
FlashNeRD: Performance-First Contact-Rich Neural Robot Dynamics
FlashNeRD：以性能为先的接触丰富神经机器人动力学
Abstract
Compared with analytical physics, learned dynamics models promise robot simulation that is faster, inherently differentiable, and easily adaptable to real data. Neural Robot Dynamics (NeRD) pursues this by keeping collision detection analytical and replacing a simulator's numerical dynamics for the robot with a learned model. Three limitations remain. NeRD offers little speedup over the simulator it learned from, accepts contact only at predefined points, and has no two-way coupling with objects it manipulates. FlashNeRD removes all three with a parallel streaming architecture that makes each prediction faster and more accurate, an encoder that accepts contacts wherever they occur, and two-way coupling with objects simulated by analytical solvers. Experiments across five robots show faster and more accurate dynamics, faster policy learning, and faster inference-time planning. Across three robots, FlashNeRD's dynamics model is up to $55\times$ faster than an optimized NeRD and more accurate over long rollouts. This speedup extends to policy learning, where PPO trains an ANYmal locomotion policy in 34 seconds, $3.8\times$ faster than the analytical simulator and $2.7\times$ faster than an optimized NeRD. With DIAL-MPC, a sampling-based MPC method, the robot climbs all six test platforms where fixed-contact NeRD manages one, at approximately half the analytical simulator's planning time. Cube-reorientation policies trained with FlashNeRD complete within 2% of the simulator-trained policy's target count.
Chinese Translation
与解析物理相比，学习的动力学模型承诺提供更快、内在可微且易于适应真实数据的机器人仿真。神经机器人动力学（NeRD）通过保持碰撞检测的解析性，并用学习模型替代机器人的数值动力学来实现这一目标。然而，仍然存在三个局限性。NeRD相较于其学习的仿真器提供的加速效果有限，仅在预定义点接受接触，并且与其操控的物体之间没有双向耦合。FlashNeRD通过并行流架构消除了这三点限制，使每次预测更快、更准确，采用的编码器可以在任何接触发生的地方接受接触，并实现与由解析求解器模拟的物体之间的双向耦合。针对五个机器人的实验表明，FlashNeRD实现了更快且更准确的动力学、更快的策略学习和更快的推理时间规划。在三个机器人中，FlashNeRD的动力学模型比优化后的NeRD快多达$55 imes$，并且在长时间滚动中更为准确。这种加速也延伸到策略学习，其中PPO在34秒内训练出ANYmal的运动策略，比解析仿真器快$3.8 imes$，比优化后的NeRD快$2.7 imes$。使用DIAL-MPC这一基于采样的MPC方法，机器人能够攀爬所有六个平台，而固定接触的NeRD仅能管理一个，且规划时间约为解析仿真器的一半。使用FlashNeRD训练的立方体重新定向策略在目标计数上完成的结果与仿真器训练的策略相差不超过2%。
cs.RO / 15 / 2610.09134
World Models Dream of Success: Diagnosing and Repairing Failure Insensitivity in Robot World Models
世界模型梦想成功：诊断和修复机器人世界模型中的失败敏感性缺失
Abstract
Robot world models support policy evaluation, planning, and synthetic data generation, but these applications require predictions that distinguish successful actions from failures. Across four released checkpoints from two architecture families, we observe weak sensitivity to action changes and success-like predictions on verified failures. Although recent work incorporates failures into model training, which data can repair released checkpoints without changing their architecture or training objective still remains underexplored. To this end, we introduce CureWM, which constructs alternative actions from successful demonstrations across a severity grid, verifies their outcomes through execution in simulation or on hardware, and fine-tunes released models on the resulting failures and surviving successes alongside nominal demonstrations. This construction provides controlled action contrasts from shared starting contexts. On 484 held-out LIBERO failure counterfactuals, optimism falls from 80\% after fine-tuning on the official data to 30--43\% across four independently fine-tuned CureWM models (38\% mean). In two separate evaluations on a physical robot arm, failure predictions scored as success-like by a latent-distance diagnostic decrease from 90\% after fine-tuning on successful demonstrations alone to 33\% with CureWM. With failure counts per task, successful replay data, and training budget matched, counterfactual failures yield a success--failure value gap of 0.124, compared with 0.014 for freshly collected on-policy failures. These findings support execution-verified counterfactual replay for post-hoc repair and show why reduced optimism must be evaluated alongside success--failure discrimination. Code is available at https://github.com/jiuyixu25/CureWM.
Chinese Translation
机器人世界模型支持策略评估、规划和合成数据生成，但这些应用需要能够区分成功行动与失败的预测。在两个架构系列的四个发布检查点中，我们观察到对行动变化的敏感性较弱，并且在经过验证的失败上产生类似成功的预测。尽管近期的研究将失败纳入模型训练，但哪些数据能够在不改变其架构或训练目标的情况下修复已发布的检查点仍然未得到充分探索。为此，我们提出了CureWM，它通过在严重性网格上从成功示范中构建替代行动，通过在模拟或硬件上执行验证其结果，并在生成的失败和存活成功以及名义示范上对已发布模型进行微调。这种构建提供了来自共享起始上下文的受控行动对比。在484个保留的LIBERO失败反事实中，经过在官方数据上微调后，乐观度从80%下降到四个独立微调的CureWM模型的30%至43%（平均38%）。在对一个物理机器人手臂的两个独立评估中，潜在距离诊断认为的失败预测在仅针对成功示范微调后从90%下降到使用CureWM后的33%。在每个任务的失败计数、成功重放数据和训练预算匹配的情况下，反事实失败产生的成功-失败价值差距为0.124，而新收集的政策内失败为0.014。这些发现支持经过执行验证的反事实重放用于事后修复，并表明为什么减少的乐观度必须与成功-失败区分一起进行评估。代码可在 https://github.com/jiuyixu25/CureWM 获取。
cs.RO / 16 / 2610.09150
A Unified Kinematic Representation Enables Reusable Biological Joint Moment Estimation
统一的运动学表示法实现可重用的生物关节力矩估计
Abstract
Objective: Data-driven models that estimate physiological states, particularly biological joint moments, are widely used in exoskeleton control. However, these estimators are often coupled to device-specific sensor configurations, limiting controller transfer and the use of open-source biomechanics datasets. Methods: We proposed joint kinematics as an intermediate representation that decouples hardware-specific sensing from downstream biological joint moment estimation. A joint-moment estimator using joint angles and angular velocities was trained exclusively on open-source biomechanics data and evaluated using kinematics from a hip exoskeleton, knee exoskeleton, and inertial measurement unit (IMU) sensor suite during level-ground, ramp-ascent, and ramp-descent walking. Results: The estimator achieved an root mean square error (RMSE) of 0.17 Nm/kg and coefficient of determination (R2) of 0.79 using hip exoskeleton kinematics, 0.19 Nm/kg and 0.62 using knee exoskeleton kinematics, and 0.15 Nm/kg and 0.85 using kinematics derived from IMUs across bilateral hip, knee, and ankle joints. Conclusion: Joint kinematics enabled an estimator trained only on open-source data to operate across distinct wearable platforms. Significance: This framework may reduce target-device data collection and support transferable biological joint moment estimation for exoskeleton control and wearable biomechanics.
Chinese Translation
目的：数据驱动模型在外骨骼控制中广泛用于估计生理状态，特别是生物关节力矩。然而，这些估计器通常与特定设备的传感器配置耦合，限制了控制器的迁移和开源生物力学数据集的使用。方法：我们提出将关节运动学作为一种中间表示，解耦硬件特定传感器与下游生物关节力矩估计之间的关系。使用关节角度和角速度的关节力矩估计器仅在开源生物力学数据上进行训练，并使用来自髋关节外骨骼、膝关节外骨骼和惯性测量单元（IMU）传感器套件的运动学进行评估，测试场景包括平地行走、坡道上升和坡道下降。结果：使用髋关节外骨骼运动学，估计器的均方根误差（RMSE）为0.17 Nm/kg，决定系数（R2）为0.79；使用膝关节外骨骼运动学，RMSE为0.19 Nm/kg，R2为0.62；使用来自IMU的运动学，跨越双侧髋关节、膝关节和踝关节，RMSE为0.15 Nm/kg，R2为0.85。结论：关节运动学使得仅在开源数据上训练的估计器能够在不同的可穿戴平台上运行。意义：该框架可能减少目标设备的数据收集，并支持可转移的生物关节力矩估计，以便用于外骨骼控制和可穿戴生物力学。
cs.RO / 17 / 2610.09165
geodex: A Library for Motion Planning on Riemannian Manifolds
geodex：一个用于黎曼流形上运动规划的库
Abstract
Planning motions that respect the intrinsic geometry of a robot's configuration space, including its curvature and a configuration-dependent notion of cost, yields shorter, lower-energy, and more natural trajectories than planning under the ambient flat metric. Existing libraries for optimization on manifolds provide rich geometric primitives but do not plan around obstacles. While general-purpose motion planning libraries support many state spaces and custom distance functions, they do not yet treat a configuration-dependent Riemannian metric as the geometry that drives distance, interpolation, and geodesics. We present geodex, an open-source C++20 library with Python bindings. The library exposes the manifold, its Riemannian metric, the retraction, and the sampler as independent, interchangeable components through a single sampling-based motion planning interface. The same planner runs unchanged on canonical spaces $\mathbb{R}^n$, $\mathbb{T}^n$, $\mathbb{S}^n$, matrix Lie groups such as $SO(2)$, $SE(2)$, $SO(3)$, and $SE(3)$, products of these spaces, and articulated-robot configuration spaces, each equipped with a user-defined Riemannian metric. We make geodex publicly available with documentation, tests, and a reproducible benchmark suite.
Chinese Translation
规划遵循机器人配置空间内在几何特性的运动，包括其曲率和依赖于配置的成本概念，能够产生比在环境平坦度量下规划的更短、更低能量和更自然的轨迹。现有的流形优化库提供了丰富的几何原语，但未能考虑障碍物的规划。虽然通用运动规划库支持多种状态空间和自定义距离函数，但尚未将依赖于配置的黎曼度量视为驱动距离、插值和测地线的几何。我们提出了geodex，一个开源的C++20库，带有Python绑定。该库通过单一的基于采样的运动规划接口，将流形、其黎曼度量、回缩和采样器作为独立的、可互换的组件进行暴露。相同的规划器在标准空间$ extbf{R}^n$、$ extbf{T}^n$、$ extbf{S}^n$、矩阵李群如$SO(2)$、$SE(2)$、$SO(3)$和$SE(3)$、这些空间的乘积以及配备用户定义黎曼度量的关节机器人配置空间上均可无缝运行。我们将geodex公开发布，并提供文档、测试和可重复的基准套件。
cs.RO / 18 / 2610.09170
Beyond Reconstruction: What Matters in Action Tokenization for Robot Policies?
超越重建：在机器人策略中，动作标记化的关键是什么？
Abstract
Autoregressive action-token policies such as vision-language-action models require action tokenizers to translate discrete token sequences into precise control actions in continuous space. Many action tokenizers learn the mapping between tokens and actions via a reconstruction objective. However, as we show through extensive analysis, sufficiently accurate action reconstruction is only one part of what makes a downstream robot policy successful. It is also critical that the policy is able to predict the right tokens for new observations, and that unseen policy token predictions still decode into reasonable actions. These properties are downstream of tokenizer training and are not directly incentivized by a reconstruction objective alone. In this work, we introduce Predictable and Robust Action Tokenization (ProAct), a tokenizer training method that strategically augments reconstruction with the goal of improving downstream predictability and robustness. ProAct is policy-agnostic and uses only action datasets for training. Across the Robomimic, LIBERO, and RoboTwin benchmarks and a diverse set of tokenizer architectures, ProAct improves rollout success by an average of 11.3 percentage points. These improvements also translate to vision-language-action policies and real-world robotic manipulation, yielding average gains of 21.8 and 36.7 percentage points, respectively. These results suggest that effective action tokenization should be designed as a policy interface that balances fidelity, predictability, and robustness, rather than as a reconstruction problem alone.
Chinese Translation
自回归动作标记策略，如视觉-语言-动作模型，需要动作标记器将离散标记序列转换为连续空间中的精确控制动作。许多动作标记器通过重建目标学习标记与动作之间的映射。然而，通过广泛的分析，我们表明，足够准确的动作重建只是使下游机器人策略成功的一部分。策略能够为新观察预测正确的标记，以及未见策略标记预测仍然解码为合理动作，这些都是至关重要的。这些特性是在标记器训练的下游，并不是仅通过重建目标直接激励的。在本研究中，我们引入了可预测和鲁棒的动作标记化（ProAct），这是一种标记器训练方法，战略性地增强重建，以改善下游的可预测性和鲁棒性。ProAct与策略无关，仅使用动作数据集进行训练。在Robomimic、LIBERO和RoboTwin基准测试以及多样化的标记器架构中，ProAct平均提高了11.3个百分点的回滚成功率。这些改进也转化为视觉-语言-动作策略和现实世界的机器人操作，分别带来了平均21.8和36.7个百分点的提升。这些结果表明，有效的动作标记化应被设计为一种策略接口，平衡保真度、可预测性和鲁棒性，而不仅仅是作为一个重建问题。
cs.RO / 19 / 2610.09178
CAP: Codebook-Aligned Prediction for Tokenized Robot Policies
CAP：与代码本对齐的标记化机器人策略预测
Abstract
Action tokenization converts continuous robot actions into discrete symbols that can be modeled autoregressively. However, existing tokenizer-based policies typically ignore the tokenizer's learned latent code structure: after tokenization, the policy treats tokens as unrelated class indices and learns a new classifier from scratch. We show that this discarded structure is valuable. We introduce Codebook-Aligned Prediction (CAP), a method that directly reuses the tokenizer's code vectors as policy class prototypes while leaving the tokenizer and policy backbone otherwise unchanged. Across four quantizer families, three simulation benchmarks, and two real-robot tasks, CAP consistently improves task success over standard token classification heads while holding the tokenizer (and therefore its reconstruction quality) fixed. Our analysis further shows that these gains are not explained by higher token accuracy or changes in the policy head alone. Instead, reusing the tokenizer codebook provides the policy with valuable information about the tokenizer's learned latent structure across tokens, making token prediction errors more benign in action space and improving the representations learned by the policy backbone. These results suggest that action tokenizers learn useful action-aware latent structure beyond discrete targets that should be preserved when training downstream policies.
Chinese Translation
动作标记化将连续的机器人动作转换为可以自回归建模的离散符号。然而，现有的基于标记器的策略通常忽略了标记器学习到的潜在代码结构：在标记化之后，策略将标记视为无关的类别索引，并从头开始学习一个新的分类器。我们表明，这一被丢弃的结构是有价值的。我们提出了与代码本对齐的预测（Codebook-Aligned Prediction, CAP），这是一种直接重用标记器的代码向量作为策略类别原型的方法，同时保持标记器和策略主干不变。在四个量化器家族、三个仿真基准和两个真实机器人任务中，CAP在保持标记器（因此保持其重建质量）不变的情况下，始终提高了任务成功率，超越了标准的标记分类头。我们的分析进一步表明，这些增益并不能仅通过更高的标记准确性或策略头的变化来解释。相反，重用标记器代码本为策略提供了关于标记器学习到的潜在结构的有价值信息，使得在动作空间中的标记预测错误更加温和，并改善了策略主干学习到的表示。这些结果表明，动作标记器学习到了超越离散目标的有用的动作感知潜在结构，在训练下游策略时应予以保留。
cs.RO / 20 / 2610.09187
Embedded Evaluation of Task Admission Coalescing in Decentralized Multi-Robot Systems
去中心化多机器人系统中任务接纳合并的嵌入式评估
Abstract
Multi-robot task allocators in dynamic missions commonly admit newly released tasks immediately, potentially invoking allocation for each new arrival. We evaluate task-admission coalescing as a mechanism for controlling allocator processor work while accounting for target-service latency, using CBAA, ACBBA, PI, and HIPC as a representative MRTA suite. The study comprises a 3,000-mission AGX Orin campaign with measured computation delay, 3,000 paired zero-compute missions, and a 96-mission Pololu 3pi+ RP2040 allocator hardware-in-the-loop campaign. Immediate (Eager) admission is compared with thresholds of two, four, and eight tasks and a four-task policy with a 10-s waiting bound across three arrival rates. Across the AGX experiments, coalescing reduces both allocator calls and processor work in 31 of 48 evaluated conditions, although the magnitude of the saving varies by allocator. The principal cost appears under sparse arrivals, where four-task batching increases online-target mean latency by 20.08-23.57 s, predominantly through admission waiting. Bounded admission reduces mean work in ten of twelve allocator-load conditions, including a 19.35% reduction with a 1.40-s mean latency increase for high-arrival HIPC. RP2040 experiments show that this tradeoff can become more favorable as processor constraints tighten. Across four matched medium-arrival HIPC scenarios, Count b=4 reduces mean RP2040 work by 59.85% and service latency by 39.47%, while the corresponding AGX cases increase latency by 27.69%. These results show that the usefulness of task-admission coalescing depends on arrival intensity, allocator-specific behavior, and the relative cost of computation on the execution platform.
Chinese Translation
在动态任务中，多机器人任务分配器通常会立即接纳新发布的任务，这可能会为每个新到的任务引发分配。我们评估任务接纳合并作为一种控制分配器处理器工作量的机制，同时考虑目标服务延迟，使用 CBAA、ACBBA、PI 和 HIPC 作为代表性的多机器人任务分配（MRTA）套件。本研究包括一个 3000 次任务的 AGX Orin 实验，测量了计算延迟，3000 对零计算任务，以及一个 96 次任务的 Pololu 3pi+ RP2040 硬件在环实验。将立即（急切）接纳与两个、四个和八个任务的阈值以及一个具有 10 秒等待限制的四任务策略进行比较，涵盖了三种到达率。在 AGX 实验中，合并在 48 个评估条件中的 31 个条件下减少了分配器调用和处理器工作，尽管节省的幅度因分配器而异。在稀疏到达情况下，主要成本似乎表现得更为明显，其中四任务批处理使在线目标的平均延迟增加了 20.08-23.57 秒，主要是由于接纳等待。有限接纳在十二个分配器负载条件中的十个条件下减少了平均工作量，包括在高到达率 HIPC 中，平均延迟增加 1.40 秒时减少了 19.35% 的工作量。RP2040 实验表明，随着处理器约束的加紧，这种权衡可能变得更加有利。在四个匹配的中等到达率 HIPC 场景中，Count b=4 将平均 RP2040 工作量减少了 59.85%，服务延迟减少了 39.47%，而相应的 AGX 情况下延迟增加了 27.69%。这些结果表明，任务接纳合并的有效性取决于到达强度、分配器特定行为以及执行平台上计算的相对成本。
cs.RO / 21 / 2610.09207
Belief-Space Planning with Planner-Conditioned Estimator Error under Intermittent Observations
在间歇性观测下的规划者条件估计误差的信念空间规划
Abstract
Belief-space planners with separately designed or off-the-shelf estimators may have access to state-relevant information the estimator does not observe. Consequently, even an estimator that minimizes mean-squared error under its own information can have a nonzero error mean when conditioned on planner information. When future corrections are intermittent and stochastic, differences between accepted and rejected error means introduce additional uncertainty terms to correction events that zero-mean assumptions ignore. In this paper, we propose a belief-space planning approach that predicts, propagates, and penalizes planner-conditioned estimator error along evaluated trajectories. A planner-conditioned estimator-error model combines affine error dynamics with a moment recursion that preserves stochastically-induced correction uncertainty. We describe a method by which to predict estimator error dynamics within specified operating regimes and incorporate predicted error moments into a task-weighted quadratic risk objective. We evaluate the approach in simulation for a tilt-rotor VTOL landing on a ship deck using receding-horizon planning. We find that conditioning on planner information reduced error-prediction loss for a command-blind EKF by 13.4% relative to a planner-ignorant model, with strongly regime and horizon-dependent forecasting abilities. In a small set of 16 paired closed-loop trials, the planner lowered the median terminal task gauge from 1.60 to 0.99, and predicted estimator error beyond 2s within a factor of 1.3, versus a 5.7-fold underprediction by covariance-only planning.
Chinese Translation
信念空间规划者使用单独设计或现成的估计器，可能会获得估计器未观察到的与状态相关的信息。因此，即使是一个在其自身信息下最小化均方误差的估计器，在条件化于规划者信息时也可能具有非零的误差均值。当未来的修正是间歇性和随机的时，接受和拒绝的误差均值之间的差异会引入额外的不确定性项，这些项在零均值假设中被忽略。本文提出了一种信念空间规划方法，该方法沿评估轨迹预测、传播和惩罚规划者条件的估计误差。规划者条件的估计误差模型结合了仿射误差动态和保持随机诱导的修正不确定性的矩递归。我们描述了一种方法，通过该方法可以在指定的操作范围内预测估计器误差动态，并将预测的误差矩纳入任务加权的二次风险目标。我们在模拟中评估了该方法，针对倾转旋翼垂直起降（VTOL）在船甲板上的着陆，使用了递归地平线规划。我们发现，相较于忽略规划者信息的模型，条件化于规划者信息使得命令盲的扩展卡尔曼滤波（EKF）的误差预测损失降低了13.4%，并且具有强烈的状态和地平线依赖的预测能力。在一组16对闭环试验中，规划者将中位终端任务指标从1.60降低到0.99，并且在2秒内预测的估计误差的因子为1.3，而仅基于协方差的规划则低估了5.7倍。
cs.RO / 22 / 2610.09224
Fast Planning for Multi-object Multi-target Throwing
多物体多目标投掷的快速规划
Abstract
Robot throwing has emerged as a promising technique for improving efficiency in logistics and warehouse automation, by enlarging the workspace and speeding up the process. To significantly increase the throwing system's throughput, we develop strategies for throwing multiple objects in one swipe. Such multi-object multi-target throwing (MOMT) leverages the large degrees of freedom of anthropomorphic hands. The key is to quickly generate fast and feasible throwing motions, which involves a complex trade-off between short trajectory duration and short planning time. We solve this problem in two stages. Offline, we build a model of the feasible set by combining object's inverted flying dynamics and the robot's kinematics and dynamics. Online, we generate feasible throws through fast solution matching and filtering of object's valid detach state and robot's feasible state that can compose sequences of throws in less than 5 ms. We validate the framework on a 7-DoF manipulator equipped with a multi-fingered hand. In simulation, coordinated two-object throwing reduces execution time by up to 46% compared to independent single-object planning, and this improvement is maintained when scaling to three objects. Real-world experiments with two objects confirm a 29% reduction; the remaining gap to the theoretical 50% is attributed to inter-throw transition overhead. When target positions are randomly changed mid-execution, the system re-plans and successfully reaches the new targets within 100 ms latency without stopping the robot. These results establish the first unified planning framework for MOMT throwing -- demonstrating scalability to multiple objects in simulation and real-world feasibility on two-object tasks -- advancing the frontier of high-throughput robotic manipulation. A video summarizing the method and the hardware experiments is available at https://liuyangdh.github.io/momt-video
Chinese Translation
机器人投掷作为一种有前景的技术，已被广泛应用于提高物流和仓库自动化的效率，通过扩大工作空间和加快过程。为了显著提高投掷系统的吞吐量，我们开发了在一次动作中投掷多个物体的策略。这种多物体多目标投掷（MOMT）利用了类人手的高度自由度。关键在于快速生成快速且可行的投掷动作，这涉及到短轨迹持续时间与短规划时间之间的复杂权衡。我们将此问题分为两个阶段进行解决。离线阶段，我们通过结合物体的倒飞动态和机器人的运动学与动力学，构建可行集模型。在线阶段，我们通过快速解决匹配和过滤物体的有效分离状态及机器人可行状态，生成可行的投掷动作，能够在不到5毫秒的时间内组合投掷序列。我们在一台配备多指手的7自由度机械臂上验证了该框架。在仿真中，协调的双物体投掷相比独立的单物体规划，执行时间减少了多达46%，而这一改进在扩展到三物体时仍然保持。与两个物体的实际实验确认了29%的减少；剩余的差距归因于投掷间的过渡开销。当目标位置在执行过程中随机改变时，系统重新规划并成功在100毫秒的延迟内到达新目标，而无需停止机器人。这些结果建立了第一个统一的MOMT投掷规划框架——在仿真中展示了对多个物体的可扩展性，并在双物体任务中验证了现实世界的可行性——推动了高吞吐量机器人操作的前沿。关于该方法及硬件实验的总结视频可在 https://liuyangdh.github.io/momt-video 查阅。
cs.RO / 23 / 2610.09228
Co-Evolving Robot Orchestrators and Policies through Deployment
通过部署共同进化的机器人协调者和策略
Abstract
Vision-language-action (VLA) policies trained on large datasets are capable within their training domains, yet they still fail to generalize to the variety of situations a robot meets in real-world deployment. Agentic robot systems complement the policy with a vision-language model (VLM) orchestrator that learns when to call the policy, how to instruct it, and when to use scripted skills instead. However, because the harness is built around a frozen policy that has limited language steerability, the orchestrator can avoid the policy's failures but never overcome them. The policy becomes the bottleneck of the whole system. Fine-tuning the policy can remove this bottleneck, but updating it alone decouples it from an orchestrator tuned to its old behavior. We propose Robo-COP, in which the orchestrator and policy co-evolve during deployment. Robo-COP curates skill demonstrations from its own executions, fine-tunes the policy when this data can address recurring failures, and adopts each new policy only after it improves the skills it was trained for. Across ten simulated RoboLab tasks, Robo-COP raises mean held-out success from 64.8% to 73.8% over the same harness with a frozen policy, while fine-tuning on a fixed schedule without verification reaches only 65.8%. On three real-world tasks, Robo-COP raises held-out success from 38.3% to 50.0%. Robo-COP turns deployment into a self-improving flywheel in which robots learn by doing, with each improvement in execution producing better data for the next round of learning. Videos and code are available at https://robo-cop.pages.dev/.
Chinese Translation
基于视觉-语言-行动（VLA）的大型数据集训练的策略在其训练领域内表现良好，但在机器人在现实世界部署中遇到的各种情况中仍然无法实现泛化。自主机器人系统通过一个视觉-语言模型（VLM）协调者来补充该策略，该协调者学习何时调用策略、如何指导策略以及何时使用脚本技能。然而，由于该系统围绕一个冻结的、语言可操控性有限的策略构建，协调者可以避免策略的失败，但永远无法克服这些失败。策略成为整个系统的瓶颈。微调策略可以消除这一瓶颈，但仅更新策略会使其与调优为旧行为的协调者脱钩。我们提出了Robo-COP，在该系统中，协调者和策略在部署过程中共同进化。Robo-COP从自身执行中策划技能演示，当这些数据能够解决反复出现的失败时，微调策略，并且仅在新策略改善了其训练的技能后才采用该策略。在十个模拟的RoboLab任务中，Robo-COP将保留成功率从64.8%提高到73.8%，而在没有验证的固定时间表上进行微调仅达到了65.8%。在三个现实世界任务中，Robo-COP将保留成功率从38.3%提高到50.0%。Robo-COP将部署转变为一个自我改善的飞轮，机器人通过实践学习，每次执行的改进为下一轮学习提供了更好的数据。视频和代码可在 https://robo-cop.pages.dev/ 获取。
cs.RO / 24 / 2610.09254
RoboRender: Robot-Oriented Video Generation for Visual Sim-to-Real Transfer
RoboRender：面向机器人视频生成的视觉仿真到现实转移
Abstract
Simulation enables large-scale, low-cost robot data generation, but policies trained in simulation often fail to transfer to the real world due to the sim-to-real visual discrepancies. Existing approaches often rely on intermediate representations, which can discard rich semantic information or require additional perception modules at deployment. We address this visual sim-to-real gap with RoboRender, a framework that converts simulated trajectories into photorealistic RGB videos for policy learning. RoboRender trains a robot-oriented video generation model conditioned on simulated depth videos, language instructions, and robot RGB mask videos, preserving simulator geometry, robot motion, and action labels while synthesizing realistic textures, backgrounds, and distractors. The generated RGB videos are paired with simulator-provided states and actions to train policies for zero-shot real-world deployment. On robot video test sets, our video model outperforms depth-conditioned video generation baselines in generation quality. In real-world experiments across pick-and-place, articulated-object manipulation, and mobile manipulation tasks, policies trained on RoboRender-generated data achieve a 71% average success rate, outperforming raw simulation renderings and conventional visual domain randomization by approximately 7.1x and 3.6x, respectively. We further show that policy performance improves with more generated videos per simulation trajectory, increasing opening-task success by 65 percentage points. These results demonstrate that generative video rendering mitigates the visual sim-to-real gap for zero-shot policy transfer. Project website: https://robo-render.github.io/.
Chinese Translation
仿真技术使得大规模、低成本的机器人数据生成成为可能，但在仿真中训练的策略往往无法有效转移到现实世界，原因在于仿真与现实之间的视觉差异。现有的方法通常依赖于中间表示，这可能会丢失丰富的语义信息，或在部署时需要额外的感知模块。我们通过RoboRender框架解决了这一视觉仿真到现实的差距，该框架将仿真轨迹转换为用于策略学习的照片级真实感RGB视频。RoboRender训练了一种面向机器人的视频生成模型，该模型以仿真深度视频、语言指令和机器人RGB掩码视频为条件，保留了仿真器几何形状、机器人运动和动作标签，同时合成了真实的纹理、背景和干扰物。生成的RGB视频与仿真器提供的状态和动作配对，以训练零-shot现实世界部署的策略。在机器人视频测试集中，我们的视频模型在生成质量上优于基于深度条件的视频生成基线。在跨越抓取与放置、关节物体操控和移动操控任务的现实世界实验中，基于RoboRender生成数据训练的策略实现了71%的平均成功率，分别比原始仿真渲染和传统视觉领域随机化提高了约7.1倍和3.6倍。我们进一步显示，随着每条仿真轨迹生成更多视频，策略性能得到了提升，开启任务的成功率提高了65个百分点。这些结果表明，生成视频渲染减轻了零-shot策略转移中的视觉仿真到现实的差距。项目网站：https://robo-render.github.io/
cs.RO / 25 / 2610.09280
Evaluating the Transfer of Co-Evolved Communication from 2D to 3D Simulation
评估共进化通信从二维到三维仿真的转移
Abstract
This work examines the transfer of a co-evolved communication mechanism between two robotic agents from a discrete two-dimensional (2D) simulator to a three-dimensional simulator with real physics (3D). The study focuses on whether a communication mechanism co-evolved in a 2D environment retains its functional role after transfer to a 3D physics-based simulator. To support this analysis, the effects of the episode time budget, the social cue, and the asymmetry between the two co-evolved roles were examined. The results indicate that the success rate increased approximately linearly with the evaluated time budgets, with no evidence of a plateau between 2,000 and 6,000 physics steps, suggesting that evaluations based on shorter episodes may underestimate the performance of the trained controllers. In both simulators, the social cue functioned primarily as a jam- assistance mechanism rather than as a navigation guide, although with a more pronounced effect in 2D. Analysis of eight independent evolutionary runs revealed a consistent direction of asymmetry, although its magnitude varied across runs. Controlling the processing order between agents allowed us to rule out an artifact of the physics engine. Finally, the results are discussed in terms of the factors that may contribute to the remaining performance gap observed after transfer.
Chinese Translation
本研究考察了两个机器人代理之间的共进化通信机制从离散的二维（2D）仿真器转移到具有真实物理的三维（3D）仿真器的过程。研究重点在于，在二维环境中共进化的通信机制在转移到基于物理的三维仿真器后是否仍能保持其功能角色。为支持这一分析，考察了情节时间预算、社会线索以及两个共进化角色之间的不对称性对结果的影响。结果表明，成功率与评估的时间预算大致呈线性增加趋势，在2000到6000个物理步骤之间没有出现平台效应，这表明基于较短情节的评估可能低估了训练控制器的性能。在两个仿真器中，社会线索主要作为干扰辅助机制而非导航指南，尽管在二维中效果更为显著。对八次独立进化实验的分析揭示了一致的不对称性方向，尽管其幅度在不同实验中有所变化。控制代理之间的处理顺序使我们能够排除物理引擎的伪影。最后，讨论了可能导致转移后观察到的性能差距的因素。
cs.RO / 26 / 2610.09283
LACE-CRAFT: Robot Co-Design with Actor Inheritance and Blackboard Collaboration
LACE-CRAFT：具有行为者继承和黑板协作的机器人共同设计
Abstract
Robot co-design couples morphology search with policy learning, yet training every new design from scratch discards acquired control experience. We present LACE-CRAFT, which compares continued learning on the current robot with policy adaptation to new morphology-reward pairs. LACE resumes the incumbent's full learning state and initializes compatible challengers with its actor parameters and observation statistics. A fixed task metric selects among both branches and the frozen incumbent. CRAFT coordinates Feedback, Morphology, Reward, and Integration roles through shared experimental records and behavioral replays to generate and cross-review paired proposals. A generative extension converts generated meshes into editable articulated models with configured joints, actuator interfaces, and consistently updated simulation assets. Across five locomotion benchmarks, mean scores over three evaluation seeds are 6.4-91.9% higher than D2C. Both methods train 30 new morphology-reward pairs over five rounds; LACE additionally uses four continuation training units. Five-task ablations examine policy inheritance and replay-derived feedback. A fabricated prototype demonstrates indoor walking and illustrates the geometry-to-hardware workflow.
Chinese Translation
机器人共同设计将形态搜索与策略学习相结合，但从头开始训练每个新设计会丢弃已获得的控制经验。我们提出了LACE-CRAFT，它比较了当前机器人上的持续学习与对新形态-奖励对的策略适应。LACE恢复现任机器人的完整学习状态，并用其行为者参数和观察统计数据初始化兼容的挑战者。一个固定的任务指标在两个分支和冻结的现任者之间进行选择。CRAFT通过共享实验记录和行为重放来协调反馈、形态、奖励和整合角色，以生成和交叉审查配对提案。生成扩展将生成的网格转换为可编辑的关节模型，具有配置的关节、执行器接口和持续更新的仿真资产。在五个运动基准测试中，三个评估种子的平均得分比D2C高出6.4-91.9%。两种方法在五轮中训练30个新的形态-奖励对；LACE还额外使用了四个持续训练单元。五项任务消融实验考察了策略继承和重放衍生反馈。一个制造的原型展示了室内行走，并说明了几何到硬件的工作流程。
cs.RO / 27 / 2610.09291
Co${}^{2}$Skill: Whole-Body Control via Skill Composition for Long-Horizon Human-Environment Interaction
Co${}^{2}$Skill：通过技能组合实现长时间人机环境交互的全身控制
Abstract
Achieving human-level dexterity in complex, unstructured environments requires the seamless integration of whole-body scene interaction and dexterous object manipulation skills. While existing physics-based controllers generate physically plausible behaviors in each domain, they largely address these two capabilities independently. In this paper, we present Co${}^{2}$Skill that integrates scene interaction and dexterous manipulation through a unified policy formulation. Built on a pretrained motion prior, the policy uses task and phase dependent observation masks to select information relevant to the current interaction goals. We introduce a goal-conditioned loco-manipulation curriculum that combines partial reference guidance for precision with exploration from varied initial states while allowing goal-directed execution beyond the demonstrated trajectories. We further introduce a cross-task curriculum that jointly trains individual skills and selected task sequences, preserving physical states across task boundaries and maintaining grasps during subsequent scene interactions. Together, these support sequential task execution and simultaneous scene interaction with object manipulation. We evaluate sitting, standing, climbing, stair traversal, and goal-directed manipulation, together with sequential execution and with random different conditions. Additionally, we demonstrate skill compositions in indoor environments, illustrating their integration within the same control formulation.
Chinese Translation
在复杂且非结构化的环境中实现人类水平的灵巧性需要全身场景交互与灵巧物体操控技能的无缝整合。虽然现有的基于物理的控制器在各自领域生成物理上合理的行为，但它们在很大程度上是独立处理这两种能力。本文提出了Co${}^{2}$Skill，通过统一的策略制定将场景交互与灵巧操控整合在一起。该策略基于预训练的运动先验，使用任务和阶段依赖的观察掩码选择与当前交互目标相关的信息。我们引入了一种目标条件的运动操控课程，该课程结合了部分参考指导以提高精度，并从不同的初始状态中进行探索，同时允许超越演示轨迹的目标导向执行。我们进一步引入了一种跨任务课程，联合训练个别技能和选定的任务序列，保持任务边界之间的物理状态，并在后续场景交互中维持抓取。综上所述，这些支持顺序任务执行和物体操控的同时场景交互。我们评估了坐、站、攀爬、楼梯穿越和目标导向操控，以及顺序执行和随机不同条件下的表现。此外，我们在室内环境中展示了技能组合，说明它们在相同控制框架内的整合。
cs.RO / 28 / 2610.09302
Adaptive Risk-Certified Event-Triggered Replanning for Dynamic Navigation
动态导航的自适应风险认证事件触发重规划
Abstract
Safe navigation in dynamic environments requires robots to plan under obstacle predictions whose errors are uncertain, non-stationary, and can induce rare but safety-critical failures. Existing control-barrier-function safety filters can reject immediately unsafe controls, but they provide little guidance on when the current finite-horizon planning mode itself is becoming unsafe as prediction uncertainty evolves. We propose Conformal Event-Triggered Risk-Certified Replanning (\emph{CERT-Replan}), a framework that uses calibrated barrier risk as an early-warning signal for replanning. CERT-Replan calibrates horizon-indexed obstacle-prediction residuals online and uses the resulting uncertainty radii to evaluate dynamic-obstacle safety margins. A one-step safety filter protects the next applied control, while a horizon-level risk monitor evaluates the upper-tail CVaR of predicted barrier-violation losses along the current MPC rollout. When this risk exceeds an allocated budget, CERT-Replan rejects the current planning mode and selects a lower-risk alternative, such as a different speed profile, corridor, or homotopy class, rather than repeatedly correcting the same nominal plan. In a non-stationary benchmark, CERT-Replan achieves an \(83.3\%\) collision reduction relative to the safety-filter-only baseline \textcolor{black}{and a \(77.8\%\) reduction relative to simple replanning triggers}, while reducing average safety-filter intervention by \(32.4\%\). \textcolor{black}{With Trajectron++, CERT-Replan achieves \(96\%\) collision-free operation. Hardware experiments and onboard runtime profiling demonstrate computational feasibility.}
Chinese Translation
在动态环境中安全导航要求机器人在障碍物预测下进行规划，而这些预测的误差是不确定的、非平稳的，并且可能导致罕见但对安全至关重要的失败。现有的控制障碍函数安全过滤器可以立即拒绝不安全的控制，但对于当前有限时间规划模式在预测不确定性演变时何时变得不安全提供的指导有限。我们提出了符合事件触发的风险认证重规划框架（Conformal Event-Triggered Risk-Certified Replanning，简称CERT-Replan），该框架利用校准的障碍风险作为重规划的早期预警信号。CERT-Replan在线校准以时间为索引的障碍物预测残差，并利用得到的不确定性半径来评估动态障碍物的安全边际。一步安全过滤器保护下一步应用的控制，而水平风险监测器评估当前模型预测控制（MPC）展开过程中预测的障碍物违反损失的上尾条件价值风险（CVaR）。当此风险超过分配的预算时，CERT-Replan拒绝当前的规划模式，并选择一个风险更低的替代方案，例如不同的速度配置、走廊或同伦类，而不是反复修正相同的名义计划。在一个非平稳基准测试中，CERT-Replan相较于仅使用安全过滤器的基线实现了83.3%的碰撞减少，相较于简单重规划触发器减少了77.8%，同时将平均安全过滤器干预减少了32.4%。使用Trajectron++，CERT-Replan实现了96%的无碰撞操作。硬件实验和机载运行时分析证明了计算的可行性。
cs.RO / 29 / 2610.09309
Predicted Futures Are Not Enough: Learning Executable Goals for Robot Manipulation
预测的未来不足够：学习可执行的机器人操作目标
Abstract
Generative world models provide rich predictions of how manipulation scenes may evolve toward task objectives, yet those futures do not directly expose the compact task variables required by control. When training supervises future prediction alone, terminal goal accuracy is not an explicit learning objective, even when geometric recovery is available. We present Entity-Level Goal Readout, a learned prediction-to-execution interface that makes the executable terminal goal an explicit output of a 3D trace world model. It combines object-centric pose prediction with translation grounded in observed depth to produce a compact goal in SE(3). A shared Pose-Native Executor consumes this fixed goal with online object-pose feedback for closed-loop control without rerunning the world model. Across five manipulation tasks, the pipeline achieves a mean success rate of 79.69%. Goal diagnostics directly measure terminal goal accuracy, while controlled translation perturbations characterize how execution degrades under goal error. Zero-shot deployment on a Franka arm achieves 73.33% success on nominal StackCube, 66.67% with distractors, and 75.00% on PickPlate with a target unseen during policy training. These results support treating the prediction-to-execution interface as an explicit learned component of world-model planning rather than incidental post-processing in the control pipeline itself. Project page: https://claire0730.github.io/executable-goals/
Chinese Translation
生成世界模型提供了丰富的预测，展示了操作场景如何朝向任务目标演变，然而这些未来并没有直接揭示控制所需的紧凑任务变量。当训练仅监督未来预测时，终端目标的准确性并不是一个明确的学习目标，即使几何恢复可用。我们提出了实体级目标读取（Entity-Level Goal Readout），这是一种学习的预测到执行接口，使得可执行的终端目标成为3D轨迹世界模型的明确输出。它结合了以对象为中心的姿态预测和基于观察深度的平移，生成一个紧凑的SE(3)目标。一个共享的姿态原生执行器（Pose-Native Executor）利用在线对象姿态反馈来消费这个固定目标，实现闭环控制，而无需重新运行世界模型。在五个操作任务中，该管道实现了79.69%的平均成功率。目标诊断直接测量终端目标的准确性，而受控的平移扰动则表征了在目标误差下执行的退化情况。在Franka机械臂上的零样本部署在标准StackCube上实现了73.33%的成功率，在有干扰物的情况下为66.67%，在PickPlate上则为75.00%，目标在策略训练期间未见过。这些结果支持将预测到执行接口视为世界模型规划的一个明确学习组件，而不是控制管道本身的附带后处理。项目页面：https://claire0730.github.io/executable-goals/
cs.RO / 30 / 2610.09315
ClimbLab: MATLAB Simulation Platform for Legged Climbing Robotics
ClimbLab：用于腿部攀爬机器人的 MATLAB 仿真平台
Abstract
This paper presents an open-sourced MATLAB simulation and analysis platform dedicated to legged climbing robots. This simulator enables the design of any limbed robotic system as an articulated multi-body with a floating base and simulates it walking and climbing in an arbitrary environment. The main variable environmental parameters are inclination, gravity, and ground stiffness, and any point cloud can be installed as the terrain map. Furthermore, the simulator employs a rigid body dynamics engine. This paper first describes the simulator structure, and the computational flow and next presents the representative simulation examples where quadrupedal robots assumed gripping on the wall or climbing on the steep slope.
Chinese Translation
本文介绍了一个开源的 MATLAB 仿真与分析平台，专门用于腿部攀爬机器人。该仿真器能够将任何四肢机器人系统设计为具有浮动基座的关节式多体，并模拟其在任意环境中行走和攀爬。主要的环境变量参数包括倾斜度、重力和地面刚度，任何点云都可以作为地形图进行安装。此外，仿真器采用了刚体动力学引擎。本文首先描述了仿真器的结构和计算流程，接着展示了代表性的仿真示例，其中四足机器人在墙面上抓握或在陡坡上攀爬。
cs.RO / 31 / 2610.09327
MCFR: A Mask-Guided Coarse-to-Fine Regression Framework for Robust Multi-Variant Board-to-Board Connector Assembly
MCFR：一种面向鲁棒多变体板对板连接器装配的掩码引导粗到细回归框架
Abstract
Automated insertion of board-to-board (BTB) connectors in 3C manufacturing requires both high visual accuracy and strong deployment robustness. This problem remains challenging because multi-variant connectors exhibit significant morphological and appearance variations, making stable cross-variant generalization difficult, while the mismatch between training and deployment under fixed-view inspection settings induces background spurious correlation and degrades real-world performance. To address these issues, this paper proposes MCFR, a Mask-Guided Coarse-to-Fine Regression framework for multi-variant BTB connector assembly. By introducing an object-aware mask prior and explicit photometric refinement, the proposed method suppresses background interference and improves alignment accuracy and robustness in practical deployment. Experiments on a self-constructed multi-variant dataset, a BTB batch insertion testbed, and a real smartphone assembly task show that MCFR consistently outperforms representative baselines and achieves an average real-world insertion success rate of 99.25%. These results demonstrate the effectiveness and practical potential of MCFR for automated assembly of multi-variant BTB connectors.
Chinese Translation
在3C制造中，自动插入板对板（BTB）连接器需要高视觉精度和强部署鲁棒性。由于多变体连接器在形态和外观上存在显著差异，使得稳定的跨变体泛化变得困难，同时在固定视角检测设置下训练与部署之间的不匹配会导致背景虚假相关，降低实际性能。因此，本文提出了MCFR，一种用于多变体BTB连接器装配的掩码引导粗到细回归框架。通过引入对象感知的掩码先验和显式的光度精细化，所提方法抑制了背景干扰，提高了实际部署中的对齐精度和鲁棒性。在自构建的多变体数据集、BTB批量插入测试平台和真实智能手机装配任务上的实验表明，MCFR始终优于代表性基线，并实现了99.25%的平均实际插入成功率。这些结果证明了MCFR在多变体BTB连接器自动装配中的有效性和实际潜力。
cs.RO / 32 / 2610.09350
Learning Unknown Constraints without Unsafe Data via Optimality and Counterfactual Regularization
通过最优性和反事实正则化学习未知约束，无需不安全数据
Abstract
Learning from demonstrations (LfD) provides a framework for inferring unknown constraints from locally optimal, constraint-satisfying expert behavior. Existing approaches largely fall into two paradigms, constrained inverse optimal control (CIOC) and inverse constrained reinforcement learning (ICRL). CIOC exploits optimality conditions such as the Karush--Kuhn--Tucker (KKT) conditions but typically assumes known dynamics and structured constraint representations. Meanwhile, ICRL accommodates complex unknown constraints and unknown transition dynamics but often requires extensive online exploration, during which unsafe constraint violations may occur. In this work, we introduce Counterfactual KKT (CF-KKT), a constraint learning framework that leverages learned dynamics and locally optimal demonstrations to recover unknown constraints without requiring known dynamics or additional risky exploration, thereby combining the data efficiency and safety advantages of CIOC with the flexibility of ICRL. First, we use a locally learned differentiable dynamics model to impose KKT-inspired optimality conditions directly on the demonstrations. Second, we use the learned dynamics to generate reward-improving counterfactual behaviors near the demonstrations, revealing behaviors that would be preferable in the absence of the unknown constraint and thus providing synthetic infeasible data. When the constraint parameterization is known, the same learned-dynamics framework enables direct CIOC-based parameter recovery, and we characterize its sensitivity to dynamics misspecification. Across high-dimensional robotic control tasks, our approach learns neural constraint representations with improved safety and data efficiency relative to state-of-the-art offline ICRL baselines.
Chinese Translation
从示范中学习（LfD）提供了一种从局部最优、满足约束的专家行为中推断未知约束的框架。现有方法主要分为两种范式：约束逆最优控制（CIOC）和逆约束强化学习（ICRL）。CIOC利用最优性条件，如Karush-Kuhn-Tucker（KKT）条件，但通常假设已知动态和结构化约束表示。与此同时，ICRL能够处理复杂的未知约束和未知转移动态，但往往需要广泛的在线探索，在此过程中可能会发生不安全的约束违反。在本研究中，我们引入了反事实KKT（CF-KKT），这是一种约束学习框架，利用学习到的动态和局部最优示范来恢复未知约束，而无需已知动态或额外的高风险探索，从而将CIOC的数据效率和安全优势与ICRL的灵活性相结合。首先，我们使用局部学习的可微动态模型直接对示范施加受KKT启发的最优性条件。其次，我们利用学习到的动态生成奖励提升的反事实行为，这些行为接近示范，揭示了在没有未知约束的情况下更可取的行为，从而提供合成的不可行数据。当约束参数化已知时，相同的学习动态框架使得基于CIOC的参数恢复成为可能，并且我们描述了其对动态错误指定的敏感性。在高维机器人控制任务中，我们的方法学习到了相较于最先进的离线ICRL基线具有更高安全性和数据效率的神经约束表示。
cs.RO / 33 / 2610.09358
COOL: Curiosity-Driven Object Ownership Learning for Personalized Robotic Assistance
COOL：基于好奇心驱动的个性化机器人助手对象所有权学习
Abstract
Robots are increasingly expected to provide personalized services in everyday environments. To do so, they must ground natural-language commands such as "Where is my backpack?" or "Find my bottle" and execute them by reasoning about object instances, people, locations, and ownership. This is challenging because ownership is rarely labeled explicitly and must be inferred from long-term, behavioral evidence of human-object interactions. To address this, we present COOL, a novel robotic framework for autonomously learning object ownership from everyday observations and maintaining a long-term spatial memory of its environment. To keep its memory current, COOL uses an agent-based curiosity-driven data collection strategy that guides the robot toward the most promising locations to gain information and refresh stale observations. Offline experiments, ablation studies, and real-world evaluations show that COOL can infer ownership relations from real-world interactions and use this knowledge for ownership-conditioned navigation and task execution.
Chinese Translation
机器人越来越被期望在日常环境中提供个性化服务。为了实现这一目标，它们必须理解自然语言命令，例如“我的背包在哪里？”或“找到我的瓶子”，并通过推理对象实例、人员、位置和所有权来执行这些命令。这是一个具有挑战性的任务，因为所有权很少被明确标记，必须通过长期的人-物交互行为证据进行推断。为了解决这个问题，我们提出了COOL，一个新颖的机器人框架，旨在从日常观察中自主学习对象所有权，并维护其环境的长期空间记忆。为了保持记忆的时效性，COOL采用了一种基于代理的好奇心驱动的数据收集策略，指导机器人前往最有前景的位置以获取信息并刷新过时的观察结果。离线实验、消融研究和现实世界评估表明，COOL能够从现实世界的交互中推断所有权关系，并利用这些知识进行基于所有权的导航和任务执行。
cs.RO / 34 / 2610.09369
Immiscible Diffusion Policy: Preserving Multimodal Robot Actions through Label-Free Noise Assignment
不可混溶扩散策略：通过无标签噪声分配保留多模态机器人动作
Abstract
When diffusion policies were first introduced, they were expected to recover multi-modal action distributions. However, we find this expectation does not always hold, as diffusion policies often collapse to a single modality even when we guarantee the balance of dataset modalities and exact within-batch symmetry. Our analysis indicates that independent action-noise pairing contributes to this failure by increasing mixing and crossing among diffusion paths, which can produce averaged denoising responses and suppress modality-specific behavior. This issue is especially severe in robot planning, where action spaces are dense and low-dimensional, significantly increasing such mixing and crossing. To alleviate this problem, we propose Immiscible Diffusion Policy, a label-free training-time add-on to diffusion policy that uses action-noise assignment to preserve relatively distinct noise-to-action routes without modifying the policy architecture or inference procedure. Across five simulated and two real-world humanoid manipulation tasks spanning state, RGB, and point-cloud observations, our method significantly improves the policy's preservation of action modalities while maintaining strong task performance. It increases the proportion of the non-dominant modality by 6.0x-14.6x across three two-modality tasks and recovers demonstrated modalities that are entirely absent from vanilla policy rollouts on both four-modality tasks. These results demonstrate that Immiscible Diffusion Policy provides a simple yet robust approach to preserving action multi-modality in general robot learning tasks.
Chinese Translation
当扩散策略首次被引入时，人们期望其能够恢复多模态动作分布。然而，我们发现这一期望并不总是成立，因为扩散策略往往会崩溃为单一模态，即使我们保证数据集模态的平衡和批内对称性。我们的分析表明，独立的动作-噪声配对导致了这一失败，增加了扩散路径之间的混合和交叉，这可能产生平均去噪响应并抑制模态特定行为。这个问题在机器人规划中尤为严重，因为动作空间是稠密且低维的，显著增加了这种混合和交叉。为了解决这个问题，我们提出了不可混溶扩散策略（Immiscible Diffusion Policy），这是一种在训练时无标签的附加方法，通过动作-噪声分配来保留相对独特的噪声到动作的路径，而无需修改策略架构或推理过程。在五个模拟和两个真实世界的人形操作任务中，我们的方法显著提高了策略对动作模态的保留，同时保持了强大的任务性能。在三个双模态任务中，非主导模态的比例增加了6.0倍至14.6倍，并在两个四模态任务中恢复了在普通策略回滚中完全缺失的示范模态。这些结果表明，不可混溶扩散策略为在一般机器人学习任务中保留动作多模态提供了一种简单而稳健的方法。
cs.RO / 35 / 2610.09398
IVG-UAV: An Intelligent Voice-Guided UAV System for Autonomous Ripe Fruit Harvesting with Vision-Based Classification and Adaptive Path Planning
IVG-UAV：一种基于智能语音引导的无人机系统，用于视觉分类和自适应路径规划的自主成熟水果采摘
Abstract
In tropical regions, their agricultural sectors remain highly dependent on manual labor for fruit harvesting. On large-scale farms, this dependency often results in significant labor cost and logistic complexities. This project presents the development and simulation of a voice-controlled Unmanned Aerial Vehicle (UAV) system designed to automate harvesting tasks in extensive plantations. The proposed system integrates speech recognition using Whisper [1] and LLM, computer vision-based ripeness classification, and adaptive path planning within a unified framework. The entire system is modeled and validated in a Gazebo simulation environment, allowing performance evaluation under controlled agricultural scenarios
Chinese Translation
在热带地区，农业部门仍然高度依赖人工进行水果采摘。在大规模农场中，这种依赖常常导致显著的劳动力成本和物流复杂性。本项目展示了一种语音控制的无人机（UAV）系统的开发与仿真，旨在自动化大规模种植园的采摘任务。所提出的系统在统一框架内集成了使用 Whisper [1] 和 LLM 的语音识别、基于计算机视觉的成熟度分类以及自适应路径规划。整个系统在 Gazebo 仿真环境中建模和验证，从而在受控的农业场景下进行性能评估。
cs.RO / 36 / 2610.09451
TempoBridge: Language-Guided Tempo Control for Vision-Language-Action Policies
TempoBridge：基于语言的节奏控制用于视觉-语言-动作策略
Abstract
Vision-Language-Action (VLA) models are effective at understanding what task to perform, but provide limited control over how it should be executed, such as moving quickly or slowly. We introduce TempoBridge, a lightweight framework that uses frozen VLA representations to modulate actions according to tempo cues in the instruction at each task phase, without additional tempo-conditioned robot demonstrations or tempo-specific base-policy fine-tuning. TempoBridge extracts tempo cues from contextual VLM representations, aligns them with task progress through a causal phase router, and modulates nominal motion commands during execution. Across LIBERO tasks, TempoBridge improves Tempo Success Rate from 52.6% to 89.7% under canonical tempo instructions while retaining high task success. It also preserves near-baseline performance when no tempo cue is present and generalizes to unseen tempo expressions without additional training. Experiments on a physical robot further demonstrate language-conditioned tempo modulation in real-world manipulation.
Chinese Translation
视觉-语言-动作（VLA）模型在理解需要执行的任务方面表现出色，但在如何执行任务方面的控制能力有限，例如快速或缓慢移动。我们提出了TempoBridge，一个轻量级框架，利用冻结的VLA表示，根据每个任务阶段指令中的节奏提示来调节动作，而无需额外的节奏条件机器人演示或节奏特定的基础策略微调。TempoBridge从上下文VLM表示中提取节奏提示，通过因果阶段路由器将其与任务进展对齐，并在执行过程中调节名义运动指令。在LIBERO任务中，TempoBridge在标准节奏指令下将节奏成功率从52.6%提高到89.7%，同时保持高任务成功率。当没有节奏提示时，它也保持接近基线的性能，并且在没有额外训练的情况下对未见过的节奏表达具有良好的泛化能力。在物理机器人上的实验进一步展示了在真实世界操作中基于语言的节奏调节。
cs.RO / 37 / 2610.09454
RobotAPO: Adversarial Physics Preference Optimization for Robotic Manipulation Video Generation
RobotAPO：用于机器人操控视频生成的对抗物理偏好优化
Abstract
Robotic manipulation videos are increasingly used as visual plans for embodied agents, but optimizing purely for visual plausibility often fails to capture the fragile physical manifold of real-world interactions. Even minor physics-violating errors at the interaction boundary, such as interpenetration or premature object motion, can completely invalidate the inferred timing and pose needed for downstream execution. Because standard supervised fine-tuning lacks the direct pressure to penalize these localized failures, we introduce AgiBot-PhysPref. This rigorously curated 10,000-sample preference dataset isolates condition-matched physics violations, turning the generator's own failure distribution into a foundational signal for physical consistency. Building upon this, we propose RobotAPO, an adversarial physics preference optimization framework operating in the continuous flow-matching denoising space. To prevent the policy from merely memorizing static curated failures, RobotAPO employs a lightweight adversarial counterfactual proposer that learns a condition-dependent, physical-failure-biased direction in denoising space. This encourages the model to explore and better respect the physical interaction boundary, all while maintaining a pure prompt-and-reference inference interface without requiring external structural conditioning. Comprehensive evaluations demonstrate that explicitly correcting these localized physics violations improves downstream robot execution from generated videos. On held-out AgiBot conditions, RobotAPO outperforms the strongest controlled internal baseline in physical consistency by 6.8% hard score and 10.0% soft score. Crucially, in real-robot replay, it translates these physical-consistency gains into a 37.4% relative improvement in task success over the strongest controlled internal baseline.
Chinese Translation
机器人操控视频越来越多地被用作具身智能体的视觉计划，但仅仅优化视觉可信度往往无法捕捉到现实世界交互的脆弱物理流形。在交互边界处，即使是微小的违反物理规律的错误，如物体间的穿透或过早的物体运动，都可能完全无效化后续执行所需的推断时机和姿态。由于标准的监督微调缺乏直接的压力来惩罚这些局部失败，我们引入了AgiBot-PhysPref。这个经过严格策划的10,000样本偏好数据集隔离了条件匹配的物理违规行为，将生成器自身的失败分布转化为物理一致性的基础信号。在此基础上，我们提出了RobotAPO，一个在连续流匹配去噪空间中运行的对抗物理偏好优化框架。为了防止策略仅仅记忆静态策划的失败，RobotAPO采用了一种轻量级的对抗反事实提议者，该提议者在去噪空间中学习条件依赖的、偏向物理失败的方向。这鼓励模型探索并更好地尊重物理交互边界，同时保持纯粹的提示和参考推断接口，而无需外部结构条件。全面的评估表明，明确纠正这些局部物理违规行为可以改善从生成视频中得到的下游机器人执行。在保留的AgiBot条件下，RobotAPO在物理一致性方面比最强的控制内部基线提高了6.8%的硬分数和10.0%的软分数。至关重要的是，在真实机器人重放中，它将这些物理一致性的提升转化为相对于最强控制内部基线的37.4%的任务成功率提升。
cs.RO / 38 / 2610.09462
TMT: Runtime Backdoor Detection for Vision-Language-Action Policies on Unseen Tasks
TMT：针对未见任务的视觉-语言-动作策略的运行时后门检测
Abstract
Backdoored vision-language-action (VLA) policies can preserve benign task performance while producing malicious actions when a trigger appears. Detecting such activation is difficult because malicious behavior can comprise individually plausible actions, while unfamiliar tasks introduce legitimate changes in observations and behavior. We introduce TMT, a runtime backdoor detector based on Token Manifold and latent Transition modeling. Trained on benign rollouts, its two branches assess input-token structure and prediction errors in adjacent-layer latent dynamics. A suspicious rollout identified by the token manifold branch, once confirmed through latent deviations, guides transition selection for subsequent monitoring. We further explore policy purification through self-distillation: a frozen copy of the backdoored policy provides benign-input actions to supervise a student on paired benign and triggered observations, without requiring a separate clean reference policy. For evaluation, we adapt traditional backdoor detectors and repurpose anomaly and failure detection methods as VLA backdoor detectors. In a post-hoc comparison with ten baselines, TMT achieves state-of-the-art backdoor detection performance on unseen tasks across three VLA backdoor attacks. Our project page is available at https://zzr42.github.io/tmt/.
Chinese Translation
带有后门的视觉-语言-动作（VLA）策略在触发器出现时可以保持良性的任务表现，同时产生恶意行为。检测这种激活是困难的，因为恶意行为可能由单独合理的动作组成，而不熟悉的任务则会引入观察和行为的合法变化。我们提出了TMT，一种基于Token Manifold和潜在转移建模的运行时后门检测器。TMT在良性回放上进行训练，其两个分支评估输入标记结构和相邻层潜在动态中的预测误差。通过标记流形分支识别的可疑回放，一旦通过潜在偏差确认，将指导后续监控的转移选择。我们进一步探索通过自蒸馏进行策略净化：冻结的后门策略副本提供良性输入动作，以监督在配对的良性和触发观察上的学生，而无需单独的干净参考策略。为了评估，我们调整了传统的后门检测器，并将异常和故障检测方法重新用于VLA后门检测器。在与十个基线的后验比较中，TMT在三个VLA后门攻击的未见任务上实现了最先进的后门检测性能。我们的项目页面可在 https://zzr42.github.io/tmt/ 上访问。
cs.RO / 39 / 2610.09479
Precise SE(3) End-Effector Tracking in Whole-Body Humanoid Control
全身类人控制中的精确 SE(3) 末端执行器跟踪
Abstract
Precise end-effector tracking during humanoid whole-body motion is challenging due to floating-base oscillations, gravity, dynamic coupling, and locomotion-induced disturbances. We propose ResGAC, a whole-body humanoid controller for precise end-effector pose tracking that combines geometric admittance control (GAC) with residual reinforcement learning. GAC provides structured $\SE$ task-space feedback and generates nominal arm joint-position targets, while residual RL compensates for unmodeled dynamics and coordinates locomotion and balance in the shared joint-position action space. The left-invariant geometric formulation allows the same GAC law to be used across manipulation reference frames. This enables the use of a ground-attached heading frame that preserves planar locomotion while removing pelvis roll, pitch, and heave from the manipulation reference, thereby reducing reference-induced end-effector motion during locomotion. ResGAC is validated on a real Unitree G1 humanoid. Across four standing end-effector tracking benchmarks, ResGAC consistently outperforms representative baselines, including SONIC, achieving lower translational and rotational errors. Real-world experiments further demonstrate reduced propagation of pelvis motion to the desired end-effector pose using the proposed ground-attached heading frame. ResGAC achieves $90\%$ success in a standing peg-in-hole task compared with $50\%$ for SONIC, and accurate world-frame $\SE$ end-effector pose tracking during lower-body motion. Experimental videos are included in the supplementary material and are also available on the project website: https://resgac.github.io/ResGAC-website/.
Chinese Translation
在类人机器人全身运动中，精确的末端执行器跟踪面临着浮动基座振荡、重力、动态耦合和运动引起的干扰等挑战。我们提出了 ResGAC，一种用于精确末端执行器姿态跟踪的全身类人控制器，它结合了几何导纳控制（GAC）和残差强化学习。GAC 提供结构化的 $ ext{SE}$ 任务空间反馈，并生成名义臂关节位置目标，而残差强化学习则补偿未建模的动态，并在共享的关节位置动作空间中协调运动和保持平衡。左不变几何公式允许在不同的操作参考框架中使用相同的 GAC 法则。这使得可以使用一个与地面相连的朝向框架，在保持平面运动的同时，从操作参考中去除盆骨的滚转、俯仰和升降，从而减少运动过程中参考引起的末端执行器运动。ResGAC 在真实的 Unitree G1 类人机器人上进行了验证。在四个站立末端执行器跟踪基准测试中，ResGAC 一直优于代表性的基线，包括 SONIC，达到了更低的平移和旋转误差。现实世界的实验进一步证明了使用所提出的地面附着朝向框架可以减少盆骨运动对期望末端执行器姿态的传播。在站立的插销入孔任务中，ResGAC 的成功率达到 $90\%$，而 SONIC 仅为 $50\\%$，并在下肢运动中实现了准确的世界框架 $ ext{SE}$ 末端执行器姿态跟踪。实验视频已包含在补充材料中，并可在项目网站上获得：https://resgac.github.io/ResGAC-website/。
cs.RO / 40 / 2610.09488
SiGNgapore - An Interactive Dataset for Sign-based Visual Navigation
SiGNgapore - 一种基于标志的视觉导航互动数据集
Abstract
The future of autonomous robots in human-oriented environments depends on their ability to navigate in the absence of prebuilt maps; exploiting navigational aids, such as navigational signs, designed by humans for humans is critical for enabling autonomy. This manuscript introduces a unique dataset collected using a handheld device, at various public spaces across Singapore, representing environments that humans frequent daily, and focusing on sign-centric decision making for mapless navigation. The dataset includes RGB, sparse depth, odometry and IMU measurements of scenarios centered around navigational signs and the complex environments where they are placed. Additionally, we provide 56 long-horizon navigation missions and over 450 sign-centric scenarios. All data is provided in a human- readable format, as well as utility scripts for conversion to ROS 2. Lastly, we provide venue maps and GPS aligned scene graphs of the test environments. We discuss the potential use cases for this dataset.
Chinese Translation
自主机器人在以人为中心的环境中的未来，依赖于它们在没有预先构建地图的情况下进行导航的能力；利用人类为人类设计的导航辅助工具，如导航标志，对于实现自主性至关重要。本文介绍了一个独特的数据集，该数据集使用手持设备在新加坡的多个公共场所收集，代表了人类日常活动的环境，重点关注无地图导航中的基于标志的决策。该数据集包括围绕导航标志及其所处复杂环境的RGB、稀疏深度、里程计和IMU测量。此外，我们提供了56个长时间导航任务和超过450个基于标志的场景。所有数据均以人类可读的格式提供，并附有转换为ROS 2的实用脚本。最后，我们提供了测试环境的场馆地图和GPS对齐的场景图。我们讨论了该数据集的潜在应用案例。
cs.RO / 41 / 2610.09496
Sparse Feature Policy Unlearning Mitigates State Hallucination in Vision-Language-Action Models
稀疏特征策略遗忘缓解视觉-语言-动作模型中的状态幻觉
Abstract
Vision-Language-Action (VLA) models have shown strong generalization in robotic manipulation by leveraging rich representations from pretrained vision-language models. However, their deployment in real-world environments remains limited by recurring unreliable behaviors. In this work, we study state hallucination, a recurring failure pattern in which a VLA continues acting as if an unrealized robot-object state had been achieved. Our analyses find that state hallucination coincides with weakened attention to task-relevant visual regions, and a mechanistic interpretation via sparse autoencoders reveals that hallucination-associated sparse features are activated when these failures occur. Based on this analysis, we propose SOUL (Sparse feature pOlicy UnLearning), which selectively unlearns policy knowledge associated with state hallucination behaviors, where sparse features identified from hallucination failures and successful behaviors serve as explicit forgetting and retention targets, respectively. Experiments across VLA architectures in simulated and real-world environments show that our method substantially reduces hallucinated failures and improves task success without substantially compromising the existing manipulation capabilities. These results suggest that interpretable feature analysis provides a practical basis for selectively modifying undesirable knowledge in robot policies.
Chinese Translation
视觉-语言-动作（VLA）模型通过利用预训练视觉-语言模型中的丰富表征，在机器人操作中表现出强大的泛化能力。然而，它们在现实环境中的应用仍然受到反复出现的不可靠行为的限制。在本研究中，我们探讨了状态幻觉，这是一种反复出现的失败模式，其中VLA继续表现得好像已经实现了一个未实现的机器人-物体状态。我们的分析发现，状态幻觉与对任务相关视觉区域的注意力减弱相吻合，而通过稀疏自编码器的机制解释表明，当这些失败发生时，与幻觉相关的稀疏特征被激活。基于这一分析，我们提出了SOUL（稀疏特征策略遗忘），该方法选择性地遗忘与状态幻觉行为相关的策略知识，其中从幻觉失败和成功行为中识别出的稀疏特征分别作为明确的遗忘和保留目标。在模拟和现实环境中对VLA架构的实验表明，我们的方法显著减少了幻觉失败，并在不显著妨碍现有操作能力的情况下提高了任务成功率。这些结果表明，可解释的特征分析为选择性修改机器人策略中不良知识提供了实用基础。
cs.RO / 42 / 2610.09520
Not All Uncertainty Matters: Simulation-in-the-Loop Fast-Slow Reasoning for Decision-Critical Autonomous Driving System
并非所有的不确定性都重要：决策关键型自动驾驶系统的循环仿真快慢推理
Abstract
Large vision-language models (VLMs) provide powerful open-world perception and reasoning for autonomous driving, but their high computational cost and inference latency make continuous cloud-side use impractical. This motivates fast--slow collaboration, where efficient onboard modules handle real-time perception and control while cloud models provide high-level reasoning only when needed. The key challenge is deciding when cloud reasoning should influence time-critical driving decisions. Existing methods often rely on perception uncertainty, heuristic triggers, or resource-driven policies, without assessing whether resolving an uncertainty will improve planning. We propose \textbf{SIGMA}, a simulation-in-the-loop framework for task-oriented fast--slow collaboration. SIGMA embeds the planner into uncertainty assessment and evaluates how plausible scene realizations under semantic and geometric uncertainty affect feasible trajectories and planning cost. Based on these outcomes, it estimates the expected reduction in planning cost from resolving uncertainty. We further introduce expected planning gain (EPG), a decision-level metric for cloud invocation, cloud-guidance integration, and request prioritization under deadline and resource constraints. Experiments in CARLA show that SIGMA reduces unnecessary cloud interactions while improving planning, efficiency, and navigation success in static and dynamic obstacle scenarios. Compared with fixed-period collaboration, SIGMA reduces unnecessary cloud interactions by 50\%, improves navigation success by more than 6\%, and cuts finish time by up to 26.2\% in dynamic scenarios.
Chinese Translation
大型视觉-语言模型（VLMs）为自动驾驶提供了强大的开放世界感知和推理能力，但其高计算成本和推理延迟使得持续的云端使用变得不切实际。这促使了快慢协作的产生，其中高效的车载模块处理实时感知和控制，而云模型仅在需要时提供高层次的推理。关键挑战在于决定何时云推理应影响时间关键的驾驶决策。现有方法通常依赖于感知不确定性、启发式触发或资源驱动的策略，而未评估解决不确定性是否会改善规划。我们提出了 extbf{SIGMA}，一种用于任务导向快慢协作的循环仿真框架。SIGMA将规划者嵌入不确定性评估中，并评估在语义和几何不确定性下，合理场景实现如何影响可行轨迹和规划成本。基于这些结果，它估算了解决不确定性所带来的规划成本预期降低。我们进一步引入了期望规划增益（EPG），这是一个用于云调用、云指导集成和在截止日期与资源约束下请求优先级排序的决策级指标。在CARLA中的实验表明，SIGMA减少了不必要的云交互，同时提高了静态和动态障碍场景下的规划、效率和导航成功率。与固定周期协作相比，SIGMA减少了50 ext{%}的不必要云交互，提高了超过6 ext{%}的导航成功率，并在动态场景中将完成时间缩短了最多26.2 ext{%}。
cs.RO / 43 / 2610.09533
Contact-Aware Imitation Learning Through Contact Factorization
通过接触因子分解的接触感知模仿学习
Abstract
Generalizable contact-rich manipulation requires robots to preserve intended task behavior while adapting its physical realization to changing contact conditions. However, interaction forces can vary substantially with small changes in surface geometry, orientation, and friction, making policies trained directly on raw force measurements difficult to transfer beyond demonstrated conditions. We introduce FACE, a contact-factorized imitation learning framework that separates intended task behavior from environment-dependent contact factors. Our representation expresses interaction forces in normalized, contact-relative coordinates, while a learned contact-normal estimator and an online friction estimator infer the local contact normal and effective friction scale. Together, these estimators enable force observations to be encoded and policy outputs to be decoded into physical motion and force commands during execution. In this way, FACE adapts execution to current contact conditions while preserving the intended task behavior, without updating the policy parameters. We evaluate FACE on real-robot contact-rich manipulation under unseen variations in surface properties and geometry, demonstrating robust generalization across contact conditions through controlled comparisons with variants that adapt prior approaches to our setting. Videos and additional materials can be found on the project page: https://rcilab.khu.ac.kr/face.
Chinese Translation
可推广的接触丰富操控要求机器人在适应不断变化的接触条件时，保持预期的任务行为。然而，交互力可能会因表面几何形状、方向和摩擦的微小变化而大幅波动，这使得直接基于原始力测量训练的策略难以在超出演示条件的情况下进行迁移。我们提出了FACE，一个接触因子分解的模仿学习框架，它将预期的任务行为与环境依赖的接触因子分离。我们的表示在标准化的接触相对坐标中表达交互力，同时学习的接触法线估计器和在线摩擦估计器推断局部接触法线和有效摩擦尺度。这些估计器共同使得力观测能够被编码，并在执行过程中将策略输出解码为物理运动和力命令。通过这种方式，FACE在不更新策略参数的情况下，适应当前的接触条件，同时保持预期的任务行为。我们在真实机器人接触丰富操控中评估了FACE，面对未见的表面属性和几何变化，展示了通过与适应我们设置的先前方法变体的控制比较，跨接触条件的强健泛化。视频和其他材料可以在项目页面找到：https://rcilab.khu.ac.kr/face。
cs.RO / 44 / 2610.09536
MagCilia: A Compact Magnetociliary Tactile Sensor with 3D Force Sensing for Robotic Contact Perception and Grasping Feedback
MagCilia：一种紧凑型磁性纤毛触觉传感器，具有用于机器人接触感知和抓取反馈的三维力感测功能
Abstract
Robotic grasping and surface exploration benefit from simultaneous measurement of normal and tangential forces and from surface information obtained through contact. Here, we present a compact magnetociliary tactile sensor (MagCilia) that combines a flexible magnetic-cilia structure with a Hall sensor for 3D force sensing. Quasi-static finite element analysis is used to investigate structural deformation and magnetic responses under multidirectional loading. To reconstruct forces from the coupled magnetic channels, we propose causal history fusion regression (CHFR), which combines current magnetic-field measurements with their recent changes. Five-fold cross-validation grouped by calibration record yields root-mean-square errors of 0.40, 0.57, and 0.69 N for Fx, Fy, and Fz, respectively, with corresponding coefficients of determination of 0.93, 0.90, and 0.92. Robotic experiments demonstrate tangential-force-guided gripper adjustment and multi-axis load monitoring under external perturbations. Frequency-domain features of the reconstructed forces distinguish six surface categories with 99.39% accuracy in three-fold cross-validation grouped by acquisition session. An online robotic demonstration additionally identifies all six tested surfaces. These results demonstrate 3D force reconstruction, grasping feedback, and surface recognition using a single compact tactile unit.
Chinese Translation
机器人抓取和表面探索受益于对法向力和切向力的同时测量，以及通过接触获得的表面信息。在此，我们提出了一种紧凑型磁性纤毛触觉传感器（MagCilia），它结合了灵活的磁性纤毛结构和霍尔传感器以实现三维力感测。我们采用准静态有限元分析来研究在多方向载荷下的结构变形和磁响应。为了从耦合的磁通道重建力，我们提出了因果历史融合回归（Causal History Fusion Regression, CHFR），该方法将当前的磁场测量与其近期变化相结合。通过按校准记录分组的五折交叉验证，Fx、Fy和Fz的均方根误差分别为0.40、0.57和0.69 N，对应的决定系数为0.93、0.90和0.92。机器人实验展示了在外部扰动下的切向力引导夹具调整和多轴载荷监测。重建力的频域特征在按采集会话分组的三折交叉验证中以99.39%的准确率区分了六种表面类别。此外，在线机器人演示还识别了所有六种测试表面。这些结果展示了使用单一紧凑型触觉单元进行三维力重建、抓取反馈和表面识别的能力。
cs.RO / 45 / 2610.09566
Targeted Modality Dropout for Real-Robot Manipulation Robust to Intermittent Vision Loss
针对间歇性视觉丧失的真实机器人操作的目标模态丢弃
Abstract
Imitation learning policies that integrate multiple sensory modalities are prone to overreliance on a dominant modality, such as vision, during training, which can disrupt policy execution when that modality is lost at inference time. In this paper, we introduce Targeted Modality Dropout (TMD), in which the dependence on each modality is estimated using attention and the most dominant modality is selectively dropped. This is combined with entropy regularization over the dependence distribution. Through real-robot evaluation using a bimanual manipulator, we show that under vision loss the success rate of the baseline policy drops substantially, whereas TMD sustains task execution. In contrast, a conventional dropout that selects the dropped modality at random, without the entropy regularization, fails on many tasks even without vision loss.
Chinese Translation
模仿学习策略在训练过程中整合多种感知模态，容易对主导模态（如视觉）产生过度依赖，这在推理时该模态丧失时会干扰策略执行。本文提出了目标模态丢弃（Targeted Modality Dropout, TMD），通过注意力机制估计对每种模态的依赖程度，并有选择性地丢弃最主导的模态。同时结合了对依赖分布的熵正则化。通过使用双手操控器的真实机器人评估，我们显示在视觉丧失情况下，基线策略的成功率显著下降，而TMD能够维持任务执行。相比之下，传统的随机选择丢弃模态的丢弃方法在没有熵正则化的情况下，即使在没有视觉丧失的情况下也在许多任务上失败。
cs.RO / 46 / 2610.09573
Point It, Strike It: Direction-Conditioned Dynamic Manipulation of Deformable Linear Objects
指向它，击打它：方向条件下可变形线性物体的动态操控
Abstract
Goal-conditioned dynamic manipulation of deformable linear objects has mainly specified goals as positions for a rope tip to reach. Many tasks, however, depend on how the tip arrives. We therefore study single-swing rope striking with goals that specify the tip's 3D position and arrival direction, across the workspace and on different ropes. This is challenging because rope dynamics are hard to model, no demonstrations exist, distinct swings reach the same goal with different reliability, and the sim-to-real gap extends beyond the rope. To address these challenges, we extend the state-of-the-art DLO simulator DeformX with GPU acceleration, a stable Cosserat rod solver, and a cross-flow aerodynamic model, yielding DeformX2.0, which is more than $20{,}000\times$ faster. We then propose TRACE (Trace-rooted Adaptive Cross-Entropy), which generates striking data by warm-starting each new target from the stored swing whose tip path passes closest to it. Its cost penalizes rope bending and abrupt tip motion to favor repeatable swings. A conditional flow-matching policy trained on this data reaches 92.1% accuracy in simulation. Finally, we propose RECAP (Residual Calibration Policy), which fits the simulator's rope and rig parameters to a few calibration swings and adapts actions with a correction policy trained in simulation. On a real robot, across three ropes, RECAP raises success within 5cm from 72% to 87% for position goals, and within 10cm and 10{\deg} from 50% to 79% for goals that also specify the arrival direction.
Chinese Translation
可变形线性物体的目标条件动态操控主要将目标定义为绳索尖端需要到达的位置。然而，许多任务依赖于尖端到达的方式。因此，我们研究了单摆绳索击打，其目标指定尖端的三维位置和到达方向，涵盖工作空间和不同的绳索。这一任务具有挑战性，因为绳索动力学难以建模，缺乏演示，且不同的摆动以不同的可靠性到达相同的目标，同时仿真与现实之间的差距不仅限于绳索。为了解决这些挑战，我们扩展了最先进的可变形线性物体（DLO）仿真器DeformX，增加了GPU加速、稳定的Cosserat杆求解器和横流气动模型，推出了DeformX2.0，其速度提高了超过$20{,}000 imes$。随后，我们提出了TRACE（基于轨迹的自适应交叉熵），通过从存储的摆动中热启动每个新目标，生成击打数据，这些摆动的尖端路径与新目标最接近。其成本惩罚绳索弯曲和突发的尖端运动，以偏向可重复的摆动。基于此数据训练的条件流匹配策略在仿真中达到了92.1%的准确率。最后，我们提出了RECAP（残差校准策略），该策略通过少量校准摆动来拟合仿真器的绳索和设备参数，并利用在仿真中训练的校正策略来调整动作。在真实机器人上，针对三根绳索，RECAP将位置目标在5cm内的成功率从72%提高到87%，而在10cm和10°内的目标（同时指定到达方向）的成功率从50%提高到79%。
cs.RO / 47 / 2610.09588
Adaptive Code Generation for Controlling Robots
用于控制机器人的自适应代码生成
Abstract
Deploying robots as Complex Adaptive Systems (CAS) in unknown and dynamic environments necessitates a transition from rigid command libraries toward intention-based autonomy, as natural language represents the only medium capable of articulating complex goals beyond the capacity of finite instruction sets. While Large Language Models (LLMs) offer a path toward natural language goal description, their integration introduces significant challenges: the formalization gap between imprecise intentions and executable actions, the taxonomy gap induced by unpredictable environments, and the challenge of maintaining temporal state and progress awareness. This work introduces an architectural framework that enables robotic control by leveraging generative AI. The system follows a dual-AI design: an LLM translates high-level intentions into executable program code restricted to a formal robotic library and constrained by verifiable syntax, while a Vision-Language Model (VLM) provides semantic grounding via a distillation process. To ensure robustness, the framework incorporates environment-driven replanning triggers based on geometric and semantic thresholds, complemented by continuous runtime monitoring and an adaptive planning loop. Benchmarked across frontier models, our framework architecture demonstrates that grounding generative AI in a reactive, constrained loop enables robust fulfillment of complex intentions in dynamic and unknown environments.
Chinese Translation
在未知和动态环境中将机器人作为复杂自适应系统（Complex Adaptive Systems, CAS）部署，要求从刚性命令库转向基于意图的自主性，因为自然语言是唯一能够表达超出有限指令集能力的复杂目标的媒介。尽管大型语言模型（Large Language Models, LLMs）为自然语言目标描述提供了一条路径，但它们的整合带来了重大挑战：不精确意图与可执行动作之间的形式化差距、由不可预测环境引起的分类差距，以及保持时间状态和进展意识的挑战。本研究提出了一种架构框架，通过利用生成性人工智能（generative AI）来实现机器人控制。该系统遵循双重人工智能设计：一个大型语言模型将高层次意图翻译为受限于正式机器人库并受到可验证语法约束的可执行程序代码，而视觉语言模型（Vision-Language Model, VLM）则通过蒸馏过程提供语义基础。为了确保稳健性，该框架结合了基于几何和语义阈值的环境驱动重新规划触发器，并辅以持续的运行时监控和自适应规划循环。在前沿模型的基准测试中，我们的框架架构表明，将生成性人工智能嵌入反应性约束循环中能够在动态和未知环境中稳健地实现复杂意图的满足。
cs.RO / 48 / 2610.09631
Fast and Robust Teach-and-Repeat Navigation Using MixVPR Visual Place Recognition*
基于MixVPR视觉位置识别的快速稳健的教学与重复导航
Abstract
Teach-and-repeat navigation systems employing advanced visual place recognition techniques for localization exhibit key attributes for long-term mobile robot navigation, such as the ability to operate in unstructured and dynamic environments. However, existing solutions based on deep-learning techniques are computationally demanding, limiting their applicability. This work introduces a novel and efficient teach-and-repeat system built on the modern visual place recognition method MixVPR. Real-world testing demonstrated its ability to operate both indoors and outdoors, achieving robustness and navigation precision comparable to other state-of-the-art systems. In addition, its lower hardware requirements make it suitable for a wide range of robotic platforms and practical applications.
Chinese Translation
采用先进视觉位置识别技术进行定位的教学与重复导航系统在长期移动机器人导航中展现出关键特性，如在非结构化和动态环境中操作的能力。然而，现有基于深度学习技术的解决方案计算需求高，限制了其适用性。本研究提出了一种新颖且高效的教学与重复系统，基于现代视觉位置识别方法MixVPR。实际测试表明，该系统能够在室内和室外环境中运行，展现出与其他最先进系统相媲美的稳健性和导航精度。此外，其较低的硬件要求使其适用于广泛的机器人平台和实际应用。
cs.RO / 49 / 2610.09645
Design optimization of tendon-driven robots considering tendon wrapping and shortcut
考虑腱缠绕和捷径的腱驱动机器人设计优化
Abstract
This study proposes a multi-objective optimization method for tendon-driven arm design, addressing the inherent trade-off between joint torque and arm thickness through tendon wrapping and shortcut effects. We formulate the problem to simultaneously maximize torque and minimize thickness, solved using NSGA-II. The algorithm optimizes tendon routing, pulley configurations, and attachment points. Results reveal Pareto-optimal designs with effective moment arm expansion via strategic shortcuts, providing practical guidelines for compact high-torque arms. The optimized configurations demonstrate non-trivial patterns beyond conventional design intuition, offering new insights for engineering efficient robotic systems.
Chinese Translation
本研究提出了一种针对腱驱动臂设计的多目标优化方法，解决了通过腱缠绕和捷径效应在关节扭矩与臂厚度之间的固有权衡。我们将问题表述为同时最大化扭矩和最小化厚度，并使用 NSGA-II 进行求解。该算法优化了腱的布置、滑轮配置和连接点。结果揭示了通过战略性捷径实现有效力臂扩展的帕累托最优设计，为紧凑型高扭矩臂提供了实用指导。优化后的配置展示了超越传统设计直觉的非平凡模式，为工程高效机器人系统提供了新的见解。
cs.RO / 50 / 2610.09690
NAViLoss: An Underwater Navigation-Aware Dual-Residual Objective for Physics-Consistent Learning
NAViLoss：一种水下导航感知的双残差目标函数用于物理一致性学习
Abstract
Autonomous underwater vehicles (AUVs) commonly rely on inertial navigation systems (INS) aided by Doppler velocity logs (DVLs) for reliable underwater navigation. Accurate DVL velocity estimation is therefore essential for successful operation. Recent learning-based methods have demonstrated improved DVL velocity estimation, particularly under degraded measurement conditions. However, their training objectives typically rely on conventional regression losses that are highly sensitive to large residuals and corrupted observations. Additionally, they do not explicitly account for the physical consistency and measurement uncertainty associated with the underlying sensing process. To address these limitations, this paper introduces navigation-aware loss (NAViLoss), a robust and uncertainty-aware objective function for learning-based AUV velocity estimation. NAViLoss jointly penalizes the velocity-estimation residual in the navigation-state domain and the beam-consistency residual in the DVL measurement domain. Its bounded formulation limits the influence of large residuals, while an adaptive mechanism regulates the uncertainty in beam geometry. Furthermore, NAViLoss is integrated with a DeepONet architecture to form a novel NAVi-DeepONet model for seamless estimation of an underwater vehicle's velocity. Lastly, our model is evaluated using approximately 10,000m of semi-synthetic AUV experimental data collected during multiple real-world sea trials. Experimental results demonstrate a 44% improvement in velocity-estimation accuracy compared with conventional and learning-based baselines. These results demonstrate the effectiveness of navigation-aware and uncertainty-adaptive loss design for robust learning-based underwater velocity estimation.
Chinese Translation
自主水下航行器（AUV）通常依赖于惯性导航系统（INS）和多普勒速度计（DVL）辅助进行可靠的水下导航。因此，准确的DVL速度估计对于成功操作至关重要。近期基于学习的方法在DVL速度估计方面显示出了改善，特别是在测量条件恶化的情况下。然而，它们的训练目标通常依赖于传统的回归损失，这些损失对大残差和受损观测值高度敏感。此外，它们并未明确考虑与基础传感过程相关的物理一致性和测量不确定性。为了解决这些局限性，本文提出了一种导航感知损失（NAViLoss），这是一种稳健且具不确定性感知的目标函数，用于基于学习的AUV速度估计。NAViLoss在导航状态域中联合惩罚速度估计残差，并在DVL测量域中惩罚波束一致性残差。其有界的形式限制了大残差的影响，而自适应机制则调节波束几何的测量不确定性。此外，NAViLoss与DeepONet架构相结合，形成了一种新颖的NAVi-DeepONet模型，实现水下航行器速度的无缝估计。最后，我们的模型使用在多次真实海试中收集的约10,000米半合成AUV实验数据进行了评估。实验结果表明，与传统和基于学习的基线相比，速度估计精度提高了44%。这些结果证明了导航感知和不确定性自适应损失设计在稳健的基于学习的水下速度估计中的有效性。
cs.RO / 51 / 2610.09695
Do Better Visual Representations Always Lead to Better End-to-End Autonomous Driving?
更好的视觉表示是否总是能带来更好的端到端自主驾驶？
Abstract
Visual foundation models (VFMs) are increasingly integrated into end-to-end autonomous driving for their powerful representations, yet it remains unclear when these representations improve driving performance. To investigate this question, we introduce ViRA, a planner-agnostic visual representation alignment framework that keeps the planner architecture and inference cost unchanged. Our study reveals three findings: (1) VFM-guided visual representations consistently improve driving performance across diverse end-to-end planners, with gains extending to zero-shot closed-loop evaluation. (2) The choice of VFM target matters for planning performance, and alignment to a different VFM can further benefit planners with pre-trained VFM encoders. (3) Auxiliary perception supervision reduces sensitivity to VFM target selection, narrowing the EPDMS spread across five targets from 2.7 to 0.5 points and potentially compensating for less effective VFM targets. Guided by these findings, we develop ViRA-Diffusion, a diffusion-based planner trained without auxiliary perception supervision, which achieves 92.3 EPDMS on NAVSIM v2 navtest, outperforming recent methods in our comparison by at least 1.9 points. The results motivate jointly considering target selection and planner supervision when integrating VFMs into end-to-end autonomous driving. The results and demo are available at https://github.com/OpenDriveLab/ViRA.
Chinese Translation
视觉基础模型（VFMs）因其强大的表示能力而越来越多地被整合到端到端自主驾驶中，但这些表示何时能改善驾驶性能仍不清楚。为了解决这个问题，我们引入了ViRA，一个与规划器无关的视觉表示对齐框架，它保持规划器架构和推理成本不变。我们的研究揭示了三个发现：（1）VFM引导的视觉表示在多种端到端规划器中始终改善驾驶性能，增益甚至扩展到零样本闭环评估。（2）VFM目标的选择对规划性能至关重要，与不同VFM的对齐可以进一步使带有预训练VFM编码器的规划器受益。（3）辅助感知监督减少了对VFM目标选择的敏感性，将五个目标之间的EPDMS差距从2.7缩小到0.5分，并可能弥补效果较差的VFM目标。根据这些发现，我们开发了ViRA-Diffusion，一个在没有辅助感知监督的情况下训练的基于扩散的规划器，在NAVSIM v2 navtest上达到了92.3的EPDMS，至少比我们比较中的最新方法高出1.9分。结果促使我们在将VFMs整合到端到端自主驾驶时共同考虑目标选择和规划器监督。结果和演示可在https://github.com/OpenDriveLab/ViRA获取。
cs.RO / 52 / 2610.09696
RoboPace: Contact-Aware Time-Optimal Retiming for Action-Chunk Policies
RoboPace：面向接触感知的时间最优重定时用于动作块策略
Abstract
Robot manipulation data collection has been shifting from teleoperation toward robot-free demonstrations, through interfaces such as the Universal Manipulation Interface (UMI) or directly from human hands. Vision-Language-Action (VLA) policies trained on such data inherit the demonstrator's timing. Yet human timing does not directly transfer to robots: compliant hands tolerate fast contact, whereas robots may overshoot due to actuator and tracking limitations; conversely, robots can move faster in free space. This motivates a unified approach that reconciles execution speed with contact safety. We present RoboPace, an online retiming layer that preserves the policy's geometric path while adapting its timing, respecting the target robot's kinematic and dynamic constraints. It adapts execution speed based on predicted contact, jointly accounting for contact-dependent speed limits and the robot's motion constraints. The method requires no policy retraining and operates in real time. Across three contact-rich tasks on a dual-arm robot, faster uniform execution and physical-limit-only retiming largely fail. RoboPace instead achieves higher overall success than slow uniform execution while completing four of five commands in approximately half the time, retaining the reliability of slow execution without its time cost.
Chinese Translation
机器人操作数据收集正从遥控操作转向无机器人演示，通过诸如通用操作接口（Universal Manipulation Interface, UMI）或直接来自人手的接口。基于此类数据训练的视觉-语言-动作（Vision-Language-Action, VLA）策略继承了演示者的时序。然而，人类的时序并不能直接转移到机器人上：顺应性手部可以容忍快速接触，而机器人由于执行器和跟踪限制可能会超出预期；相反，机器人在自由空间中可以移动得更快。这促使我们提出一种统一的方法，以调和执行速度与接触安全性。我们提出了RoboPace，这是一种在线重定时层，它在保留策略几何路径的同时调整其时序，遵循目标机器人的运动学和动力学约束。该方法基于预测的接触情况调整执行速度，同时共同考虑依赖接触的速度限制和机器人的运动约束。该方法无需重新训练策略，并且实时运行。在双臂机器人上进行的三个接触丰富的任务中，快速均匀执行和仅基于物理限制的重定时大多未能成功。相比之下，RoboPace在完成五个命令中的四个时，整体成功率高于缓慢均匀执行，并且所需时间大约为其一半，保持了缓慢执行的可靠性而不增加时间成本。
cs.RO / 53 / 2610.09718
YUBI-STAG: Contact and Semantic-Rich Alignment for VLAs via Automated Video-Language Grounding
YUBI-STAG：通过自动化视频语言基础实现的接触与语义丰富对齐的视觉-语言-动作模型
Abstract
Vision-Language-Action (VLA) models acquire broad manipulation capabilities via large-scale pretraining, yet eliciting them through language requires fine-grained alignment between instructions and physical interactions. Existing robot demonstrations typically provide only coarse task descriptions, omitting how actions are executed, including which gripper acts, which object is contacted, and how it is grasped and moved. We introduce YUBI-STAG, a framework for Spatio-Temporal Annotation and Grounding that automatically enriches manipulation demonstrations with interaction-rich semantics to align pretrained VLAs with fine-grained manipulation language. Combining contact-object segmentation with vision-language models, YUBI-STAG annotates object identities, attributes and states, per-gripper actions, bimanual coordination, and spatially grounded interactions. To address YUBI-STAG's reliance on localized sequences and multi-stage VLM inference, we distill it into YUBI-VLM. YUBI-VLM directly recovers action structure and annotations from raw, unsegmented video in few inference calls and operates from wrist views alone. We evaluate both frameworks on YUBI-STAG-Bench across temporal, semantic, and spatial grounding tasks. YUBI-VLM retains much of YUBI-STAG's annotation accuracy with fewer inference calls and shorter runtime while generalizing to unseen manipulations. Finally, post-training VLA policies on these annotations aligns them with fine-grained language and contact-aware structure. Bimanual experiments demonstrate improved performance and instruction following, including control over object identity, acting gripper, target location, and spatial relations absent from original labels.
Chinese Translation
视觉-语言-动作（VLA）模型通过大规模预训练获得广泛的操作能力，但通过语言引导它们需要指令与物理交互之间的细粒度对齐。现有的机器人演示通常仅提供粗略的任务描述，省略了动作是如何执行的，包括哪个夹持器在作用、哪个物体被接触，以及如何抓取和移动。我们提出了YUBI-STAG，一个用于时空注释和基础的框架，能够自动丰富操作演示的交互语义，以便将预训练的VLA与细粒度的操作语言对齐。YUBI-STAG结合了接触物体分割与视觉-语言模型，注释物体的身份、属性和状态、每个夹持器的动作、双手协调以及空间基础的交互。为了解决YUBI-STAG对局部序列和多阶段视觉语言模型推理的依赖，我们将其提炼为YUBI-VLM。YUBI-VLM能够在少量推理调用中直接从原始未分割视频中恢复动作结构和注释，并仅依赖手腕视角进行操作。我们在YUBI-STAG-Bench上评估了这两个框架，涵盖时间、语义和空间基础任务。YUBI-VLM在保持YUBI-STAG注释准确度的同时，减少了推理调用次数和运行时间，并能够推广到未见过的操作。最后，在这些注释上进行后训练的VLA策略使其与细粒度语言和接触感知结构对齐。双手实验表明，在控制物体身份、作用夹持器、目标位置和原始标签中缺失的空间关系方面，性能和指令遵循得到了改善。
cs.RO / 54 / 2610.09734
{\Delta}WAM: Distilling Action Tangent Fields into World Action Models
{ ext{Δ}}WAM：将动作切线场提炼为世界动作模型
Abstract
World Action Models (WAM) improve robot policies by augmenting sparse action supervision with dense future prediction. However, much of the predictable future is dominated by appearance and scene persistence rather than action-dependent dynamics. We observe that several recent WAM designs, including optical flow, motion-centric representations, and latent actions, can be understood from a common perspective in which world supervision becomes more efficient as it contains a higher proportion of action-relevant variation. Based on this insight, we introduce Action Tangent Fields, which reformulate world supervision through a local Taylor expansion of how actions induce changes in future dynamics. We represent future dynamics in Residual-VAE space, where the future latent remains recoverable from the current latent and its residual, and use a strong action-conditioned world model (ACWM) to probe the local correspondence between action variations and residual-world variations. This local first-order structure is distilled into the WAM to guide its denoising supervision toward dynamics that are more tightly coupled to action, rather than merely predictable from appearance. Across LIBERO-Plus, RoboTwin, and RoboTwin2.0-Plus, our method consistently improves robustness to lighting, background, camera, layout, and other environmental perturbations. Despite using no large-scale embodied pretraining, it achieves stronger robustness under several distribution shifts than pretrained policies. We further distill multi-step VideoDiT denoising into a single step for efficient inference. Our results suggest that effective WAM supervision should remain information-rich while concentrating its predictive capacity on the directions along which actions change the future.
Chinese Translation
世界动作模型（World Action Models, WAM）通过将稀疏的动作监督与密集的未来预测相结合，改善了机器人策略。然而，可预测的未来大部分受外观和场景持久性主导，而非依赖于动作的动态。我们观察到，近年来的几种WAM设计，包括光流（optical flow）、运动中心表示（motion-centric representations）和潜在动作（latent actions），可以从一个共同的视角理解，即世界监督在包含更高比例的与动作相关的变化时变得更加高效。基于这一见解，我们引入了动作切线场（Action Tangent Fields），通过局部泰勒展开重新构造世界监督，描述动作如何引发未来动态的变化。我们在残差变分自编码器（Residual-VAE）空间中表示未来动态，其中未来潜在状态可以从当前潜在状态及其残差中恢复，并使用强大的动作条件世界模型（Action-conditioned World Model, ACWM）探测动作变化与残差世界变化之间的局部对应关系。这种局部的一阶结构被提炼到WAM中，以引导其去噪监督朝向与动作更紧密耦合的动态，而不仅仅是从外观中可预测的动态。在LIBERO-Plus、RoboTwin和RoboTwin2.0-Plus上，我们的方法在光照、背景、相机、布局和其他环境扰动下始终提高了鲁棒性。尽管没有使用大规模的具身预训练，但在多个分布转移下，它的鲁棒性超过了预训练策略。我们进一步将多步视频去噪（VideoDiT）提炼为单步，以实现高效推理。我们的结果表明，有效的WAM监督应保持信息丰富，同时将其预测能力集中在动作改变未来的方向上。
cs.RO / 55 / 2610.09764
AeroEval: Staged Program and Execution Validation for AI-Generated Drone Missions
AeroEval：针对AI生成无人机任务的分阶段程序和执行验证
Abstract
Large Language Models (LLMs) can generate drone programs from natural-language mission descriptions, but syntactically valid programs may still violate user intent, environmental constraints, and mission-level behavior. This problem is pronounced in cyber-physical applications, where correctness depends on the interaction among generated code, mobile sensing, environmental geometry, event-driven analytics, and physical execution. Existing drone code-generation systems primarily use prompt guardrails or simulator outcomes and provide limited failure localization. We present AeroEval, an agent-assisted middleware for staged validation of AI-generated drone missions. AeroEval combines deterministic program analysis with context-grounded LLM agents. It first validates program syntax, platform API usage, and mission intent, and then evaluates the realized behavior using execution trajectories, mission requirements, and environmental context. Each stage returns structured failure information for iterative regeneration. In our evaluation using 20 navigation tasks and five analytical mission types over AirSim and Gazebo simulators, AeroEval improves navigation success from 55% to 95%. In a stagewise ablation study, our Code and Trajectory Validators by themselves achieve mean run-level success rates of 44% and 56%, respectively, while the full AeroEval pipeline achieves 88%; the stages detect complementary failures in program structure, API usage, mission intent, obstacle avoidance, altitude, coverage, and event-driven transitions and the guided regeneration corrects for them. Across the main analytics missions, AeroEval increases aggregate run-level success from 34% for one-shot AeroGen to 88% within the regeneration budget. These results demonstrate the benefit of combining program-level and execution-grounded agentic validation for AI-generated drone applications in the evaluated environment.
Chinese Translation
大型语言模型（LLMs）能够根据自然语言的任务描述生成无人机程序，但语法上有效的程序仍可能违反用户意图、环境约束和任务级行为。这一问题在网络物理应用中尤为突出，因为其正确性依赖于生成代码、移动感知、环境几何、事件驱动分析和物理执行之间的相互作用。现有的无人机代码生成系统主要依赖于提示保护措施或模拟器结果，并且提供的故障定位能力有限。我们提出了AeroEval，这是一种代理辅助的中间件，用于AI生成无人机任务的分阶段验证。AeroEval结合了确定性程序分析与基于上下文的LLM代理。它首先验证程序的语法、平台API的使用和任务意图，然后使用执行轨迹、任务要求和环境上下文评估实现的行为。每个阶段返回结构化的故障信息以便于迭代再生成。在我们使用20个导航任务和五种分析任务类型在AirSim和Gazebo模拟器上的评估中，AeroEval将导航成功率从55%提高到95%。在逐阶段消融研究中，我们的代码和轨迹验证器单独实现的平均运行级成功率分别为44%和56%，而完整的AeroEval管道实现了88%；各阶段检测到程序结构、API使用、任务意图、障碍物规避、高度、覆盖范围和事件驱动转换中的互补故障，并通过指导再生成进行修正。在主要分析任务中，AeroEval将一次性AeroGen的整体运行级成功率从34%提高到88%在再生成预算内。这些结果展示了在评估环境中结合程序级和执行基础的代理验证对于AI生成无人机应用的益处。
cs.RO / 56 / 2610.09777
On-Demand Robotic Assembly via Differentiable Geometric Part Repair
基于可微几何部件修复的按需机器人装配
Abstract
Transitioning from a digital design to a robotic assembly process currently requires months of expert manual tuning to reconcile part geometries with robotic constraints. This paper presents an end-to-end, autonomous pipeline for the design and physical construction of bespoke wooden assemblies. A generative AI agent translates user prompts into initial 3D geometries, balancing the visual fidelity of the design with select physical constraints. The assemblability of the design is further improved by a gradient-based repair stage that backpropagates through a graph attention network surrogate to adjust component geometries. In addition to correcting for disjointed and overlapping components, we demonstrate hardware-specific corrections, differentiably optimizing the geometry of components to enable robot screwdriving for 86.7% of 60 novel natural language inputs, significantly outperforming prior work by a factor of ten. For ten of the structures, we physically demonstrate assemblability with two UR5e robots. This work marks a meaningful step toward on-demand robotic manufacturing, enabling the rapid production of customized, low-volume goods.
Chinese Translation
从数字设计过渡到机器人装配过程目前需要数月的专家手动调试，以协调部件几何形状与机器人约束。本文提出了一种端到端的自主管道，用于定制木质装配的设计和物理构建。生成式人工智能代理将用户提示转换为初始3D几何形状，平衡设计的视觉逼真度与特定物理约束。通过一个基于梯度的修复阶段，进一步改善设计的可装配性，该阶段通过图注意力网络代理进行反向传播，以调整组件几何形状。除了纠正不连贯和重叠的组件外，我们还展示了特定硬件的修正，差分优化组件的几何形状，使得86.7%的60个新颖自然语言输入能够实现机器人螺丝拧紧，显著超越了先前工作的十倍。在十个结构中，我们使用两台UR5e机器人进行了物理装配演示。这项工作标志着向按需机器人制造迈出了重要一步，使得定制小批量商品的快速生产成为可能。
cs.RO / 57 / 2610.09781
End-to-End Autonomous Generation of Human Assembly Plans
端到端自主生成人类装配计划
Abstract
Turning a CAD design into an assembly plan is still largely done by hand, requiring engineers to reason about geometric feasibility, tool access, stability, and the ergonomics of human assembly. In this work, we encode long-established design for assembly (DfA) principles into a contained, end-to-end approach for generating assembly plans. Our approach takes only a mesh assembly and produces either a step-by-step assembly manual or a structured failure report, requiring no joint metadata, fastener annotations, or additional information. Four major components of a manufacturing plan are addressed autonomously: an assembly tool list, the assembly sequence and subassemblies, an assembly manual, and design feedback for improving assemblability. For determining the sequence plan, we systematically disassemble the object in a physics simulator and apply a cost function that encodes DfA principles. Manual generation, tool labelling, and assembly feedback rely primarily on multimodal large language models. Compared with a baseline that always removes the outermost part first from Tian et al., DfA-aware sequence planning reduces simulated assembly time, measured with a robot-arm assembly-time proxy, by 35% on 136 assemblies of 5 to 30 parts. The correct tool is selected for 88.6% of assembly steps. A vision-language model judge compares the generated manuals against ablated variants, identifying which page elements carry the information a reader needs. The presented approach and open-source code are available for use by engineers or AI agents looking to rapidly accelerate the creation of manufacturing plans for a given product design.
Chinese Translation
将CAD设计转化为装配计划仍然主要依赖人工完成，这要求工程师考虑几何可行性、工具可达性、稳定性以及人类装配的人体工程学。在本研究中，我们将长期以来确立的装配设计（DfA）原则编码为一种封闭的端到端方法，用于生成装配计划。我们的方法仅需一个网格装配模型，即可生成逐步装配手册或结构化故障报告，而无需关节元数据、紧固件注释或其他额外信息。制造计划的四个主要组成部分被自主处理：装配工具清单、装配顺序和子装配、装配手册，以及用于改善装配性的设计反馈。为了确定顺序计划，我们在物理模拟器中系统地拆解对象，并应用编码DfA原则的成本函数。手动生成、工具标记和装配反馈主要依赖于多模态大型语言模型。与Tian等人提出的始终优先移除最外层部件的基线相比，考虑DfA的顺序规划在136个由5到30个部件组成的装配中，减少了35%的模拟装配时间（通过机器人臂装配时间代理测量）。在88.6%的装配步骤中选择了正确的工具。一个视觉-语言模型评估生成的手册与去除某些元素的变体，识别出哪些页面元素包含读者所需的信息。所提出的方法和开源代码可供工程师或AI代理使用，以快速加速特定产品设计的制造计划创建。
cs.RO / 58 / 2610.09828
Enhancing Robotic Perception and Adaptability through Sensor Fusion and Origami-Inspired Designs
通过传感器融合和受折纸启发的设计增强机器人感知与适应性
Abstract
Compact mobile robots must recover scene geometry under changing lighting and surface texture while working within tight payload and cost limits. We present a compact mobile robot that uses origami-inspired wheels for locomotion and active control of its sensing geometry. As the wheels move between terrain-adaptive configurations, the changing chassis pitch sweeps a 2D LiDAR through intermediate elevations; held wheel positions provide a chosen viewing angle. An IMU accounts for chassis attitude, and a fusion node projects LiDAR returns into the RGB-D depth stream supplied to RTAB-Map. The arrangement uses the wheel actuation already present on a sub-300 USD, sub-2 kg prototype to extend the scanner's viewing geometry. We assess depth fusion in a textureless indoor corridor and an outdoor sunlit area, with three runs per sensor configuration in each setting. Mean full-frame invalid-depth fractions fell from 21% to 11% indoors and from 48% to 18% outdoors. The prototype combines improved depth coverage with a continuously adjustable LiDAR viewpoint using the same actuation that reconfigures its wheels.
Chinese Translation
紧凑型移动机器人必须在变化的光照和表面纹理下恢复场景几何，同时在严格的载荷和成本限制内工作。我们提出了一种紧凑型移动机器人，采用受折纸启发的轮子进行移动，并主动控制其感知几何。当轮子在适应地形的配置之间移动时，变化的底盘俯仰角使2D LiDAR在中间高度之间扫动；固定的轮子位置提供了选择的视角。惯性测量单元（IMU）用于考虑底盘姿态，而融合节点将LiDAR返回投影到供RTAB-Map使用的RGB-D深度流中。该布局利用了已经存在于一款低于300美元、低于2公斤的原型上的轮子驱动，扩展了扫描仪的视角几何。我们在无纹理的室内走廊和户外阳光照射区域评估了深度融合，每种传感器配置在每个环境中进行了三次测试。室内全帧无效深度比例从21%降至11%，户外则从48%降至18%。该原型结合了改进的深度覆盖和使用相同驱动重新配置轮子的连续可调LiDAR视点。
cs.RO / 59 / 2610.09857
Towards Accurate End-Effector Localization for UMI-Style Robotic Manipulation Teaching
面向UMI风格机器人操作教学的精确末端执行器定位
Abstract
Robot demonstration learning requires accurate and temporally complete end-effector localization during close-range manipulation and camera occlusion. Existing SLAM benchmarks emphasize navigation motions, whereas manipulation datasets prioritize policy learning over localization evaluation. We introduce MILD, a Manipulation-Interface Localization Dataset with real-world and simulation sequences. The real-world subset provides 86 sensor sequences from Insta360 X5 and Insight9 across 15 repeated tabletop tasks, calibration assets, and a per-execution robot end-effector reference trajectory. The simulation subset, MILD-Sim, extends task coverage in Isaac Sim for controlled manipulation-replay studies. Benchmarking visual-inertial and fiducial-aided systems on instrumented real-world recordings reveals large differences in both TCP-relative trajectory error and temporal coverage, even under the same nominal task. To support marker-augmented teaching workspaces without a pre-surveyed fiducial map, we present AprilVINS, which combines fisheye visual-inertial estimation with sequence-local AprilTag geometry and separates prior admission from guarded export of the jointly optimized state. On Insta360 AprilTag4 recordings, AprilVINS(full) under a unified protocol with sequence-specific profiles reaches millimeter-level SE(3)-aligned TCP-relative APE RMSE with high time completion and lower reported error than the tested routes under their respective protocols, whereas fisheye VIO without tag factors remains at centimeter scale. Ablations separate accuracy from exportability, and a MILD-Sim replay study provides task-specific tolerance references for interpreting those error magnitudes. Together, MILD and AprilVINS provide a diagnostic benchmarking framework for UMI-style demonstration collection. Code, datasets, and evaluation manifests will be released upon acceptance.
Chinese Translation
机器人演示学习在近距离操作和摄像头遮挡期间需要准确且时间上完整的末端执行器定位。现有的SLAM基准测试强调导航动作，而操作数据集则优先考虑策略学习而非定位评估。我们引入了MILD，一个包含真实世界和仿真序列的操作接口定位数据集。真实世界子集提供了来自Insta360 X5和Insight9的86个传感器序列，涵盖15个重复的桌面任务、标定资产以及每次执行的机器人末端执行器参考轨迹。仿真子集MILD-Sim扩展了在Isaac Sim中进行受控操作重放研究的任务覆盖范围。在对仪器化的真实世界录音进行视觉惯性和基于标记的系统基准测试时，即使在相同的名义任务下，TCP相对轨迹误差和时间覆盖也存在较大差异。为了支持没有预先调查的标记地图的标记增强教学工作空间，我们提出了AprilVINS，它结合了鱼眼视觉惯性估计与序列局部AprilTag几何，并将先前的接纳与联合优化状态的受保护导出分开。在Insta360 AprilTag4录音中，AprilVINS（完整）在统一协议下与序列特定配置文件结合，达到毫米级SE(3)对齐的TCP相对APE RMSE，具有高时间完成率和低于各自协议下测试路径的报告误差，而不带标记因素的鱼眼VIO仍保持在厘米级。消融实验将准确性与可导出性分开，MILD-Sim重放研究为解释这些误差幅度提供了任务特定的容忍参考。MILD和AprilVINS共同提供了一个用于UMI风格演示收集的诊断基准框架。代码、数据集和评估清单将在接受后发布。
cs.RO / 60 / 2610.09940
Juno: Taming Predictive Latents for Vision-Language-Action Models
Juno：驯化视觉-语言-动作模型的预测潜变量
Abstract
Joint-embedding predictive architectures (JEPAs) predict masked or future observations in representation space, offering a natural source of predictive latents for vision-language-action (VLA) models. Yet making these latents useful across pretraining, policy learning, and deployment requires addressing three failures: mismatch with embodiment-specific control, interference with action learning, and teacher miscalibration under distribution shifts. We introduce Juno, a unified framework built around one action-conditioned JEPA that serves as a control-aligned representation backbone, a predictive teacher, and an adaptable dynamics model. During pretraining, we train it on embodiment-matched trajectories and use a dynamic CLS loss to transfer motion-weighted patch dynamics to a compact global state. During policy learning, we fuse current-frame JEPA patches into VLA perception and use a decoupled reasoning branch with separate transformation parameters to distill future latent states for action generation. During deployment, we adapt the world model on all observed transitions, including failed rollouts, freeze the adapted teacher, and re-align the policy on verified executions using LoRA adapters and a trainable action head, without expert corrections or task rewards. On SimplerEnv, Juno raises average success from $60.9\%$ to $68.5\%$ over Qwen3GR00T, the strongest baseline, and test-time adaptation further reaches $72.7\%$; on a real robot, it retains $70\%$--$75\%$ success under background, height, and object shifts where the base policy collapses to $0\%$.
Chinese Translation
联合嵌入预测架构（JEPAs）在表示空间中预测被遮蔽或未来的观察，为视觉-语言-动作（VLA）模型提供了一种自然的预测潜变量来源。然而，使这些潜变量在预训练、策略学习和部署中有效，需要解决三个问题：与具体体现控制的不匹配、对动作学习的干扰以及在分布变化下教师的误校准。我们提出了Juno，一个围绕一个基于动作条件的JEPA构建的统一框架，作为控制对齐的表示骨干、预测教师和适应性动态模型。在预训练期间，我们在与体现匹配的轨迹上进行训练，并使用动态CLS损失将运动加权的补丁动态转移到紧凑的全局状态。在策略学习期间，我们将当前帧的JEPA补丁融合到VLA感知中，并使用一个解耦的推理分支，具有独立的变换参数，以提炼用于动作生成的未来潜在状态。在部署期间，我们在所有观察到的转变上适应世界模型，包括失败的回滚，冻结适应后的教师，并使用LoRA适配器和可训练的动作头在经过验证的执行上重新对齐策略，而无需专家修正或任务奖励。在SimplerEnv上，Juno将平均成功率从60.9%提高到68.5%，超越了最强基线Qwen3GR00T，测试时适应进一步达到72.7%；在真实机器人上，它在背景、高度和物体变化下保持70%-75%的成功率，而基础策略则崩溃至0%。
cs.RO / 61 / 2610.09943
Many Ways to Succeed: Diversity-Driven RL Fine-Tuning for VLA Generalization
多种成功之道：基于多样性的强化学习微调用于视觉-语言-动作泛化
Abstract
Reinforcement learning (RL) fine-tuning improves vision-language-action (VLA) policies through closed-loop experience, yet generalization beyond the fine-tuning distribution remains limited. Our analysis reveals a selective reshaping of exploration: RL contracts behavior globally, yet diversifies successful trajectories, elicits success with fewer rollouts, and covers more of the latent task-valid solution space than supervised fine-tuning. Broader successful-mode coverage may provide alternative strategies under distribution shifts. Inspired by this, we introduce DRIVE (Diversity-driven RL fIne-tuning for VLA gEneralization), which turns successful-behavior diversity into an explicit RL objective. DRIVE groups rollouts under matched task conditions, compares their trajectories with temporal alignment, and derives a success-conditioned intrinsic reward from relative behavioral diversity. This design encourages broader coverage of feasible solutions without rewarding diverse failures or superficial timing differences. Across LIBERO-Plus, ManiSkill3, and RoboTwin 2.0, DRIVE improves the average out-of-domain (OOD) performance over vanilla RL fine-tuning by 5.3 points on $\pi_0$ and 2.0 points on $\pi_{0.5}$. On a dual-arm AgileX PiPER-X platform, DRIVE further increases average OOD success from 64.1% to 73.3% (+9.2 points), demonstrating gains that persist under physical deployment.
Chinese Translation
强化学习（RL）微调通过闭环经验改善视觉-语言-动作（VLA）策略，但在微调分布之外的泛化能力仍然有限。我们的分析揭示了探索的选择性重塑：RL在全局上收缩行为，但多样化成功轨迹，以更少的回合实现成功，并覆盖比监督微调更多的潜在任务有效解空间。更广泛的成功模式覆盖可能在分布变化下提供替代策略。受到此启发，我们引入了DRIVE（基于多样性的强化学习微调用于VLA泛化），将成功行为的多样性转化为明确的强化学习目标。DRIVE在匹配任务条件下对回合进行分组，比较其轨迹的时间对齐，并从相对行为多样性中推导出成功条件的内在奖励。该设计鼓励更广泛的可行解覆盖，而不奖励多样化的失败或表面的时间差异。在LIBERO-Plus、ManiSkill3和RoboTwin 2.0上，DRIVE在$ ext{π}_0$上提高了5.3分，在$ ext{π}_{0.5}$上提高了2.0分，相较于传统的RL微调。在双臂AgileX PiPER-X平台上，DRIVE进一步将平均的域外（OOD）成功率从64.1%提高到73.3%（+9.2分），展示了在实际部署下持续的性能提升。
cs.RO / 62 / 2610.09967
Borrowed Eyes: Markerless Nano-UAV Flight with an Active Quadruped Observer
借用的视角：无标记纳米无人机与主动四足观察者的飞行
Abstract
Nano unmanned aerial vehicles (nano-UAVs) can navigate confined spaces that larger robots cannot, but their payload capacity severely limits the sensors and compute available for self-localization in global navigation satellite system (GNSS)-denied environments. We present a vision-based system that localizes a nano-UAV from a quadruped robot with an arm-mounted camera. The quadruped tracks the drone, estimates its position in its own coordinate system using segmentation masks and depth, and transmits that position over a real-time radio link. To ensure continuous tracking, we developed a perception-aware nonlinear Model Predictive Controller (NMPC) that dynamically adjusts the quadruped's body and arm to maximize the drone's visibility, treating observation reliability as a primary control objective. Relying solely on this external estimate, the nano-UAV executed predefined trajectories from takeoff to landing with a median 3D localization error of 59 mm. The system demonstrates robustness against short visual occlusions by smoothly transitioning to onboard inertial flight when line-of-sight is temporarily lost. Ultimately, this framework allows a quadruped to offload the localization burden of a nano-UAV, enabling inspection of complex spaces that neither robot could navigate alone.
Chinese Translation
纳米无人机（nano-UAV）能够在大型机器人无法进入的狭小空间中导航，但其有效载荷能力严重限制了在全球导航卫星系统（GNSS）无法使用的环境中进行自我定位所需的传感器和计算能力。我们提出了一种基于视觉的系统，通过一台装有相机的四足机器人对纳米无人机进行定位。四足机器人跟踪无人机，利用分割掩模和深度信息在自身坐标系中估计无人机的位置，并通过实时无线电链路传输该位置。为了确保持续跟踪，我们开发了一种感知意识的非线性模型预测控制器（NMPC），该控制器动态调整四足机器人的身体和手臂，以最大化无人机的可视性，将观察可靠性视为主要控制目标。依赖于这一外部估计，纳米无人机从起飞到着陆执行了预定义的轨迹，三维定位误差的中位数为59毫米。该系统在短暂视觉遮挡情况下表现出鲁棒性，当视线暂时丧失时，能够平滑过渡到机载惯性飞行。最终，该框架使四足机器人能够减轻纳米无人机的定位负担，从而实现对复杂空间的检查，而这些空间是任何一台机器人单独无法导航的。
cs.RO / 63 / 2610.09977
Decoding Neural Population Dynamics through Robotic Analog
通过机器人类比解码神经群体动态
Abstract
Animal evidence shows that precise voluntary movements arise from rotational neural population dynamics in motor cortex, but their physical effects remain unknown. We developed a robotic analog of biological motor systems with artificial muscles, multimodal sensors, and a neural network controller trained via reinforcement learning. The robotic analog exhibited accurate movements, robustness to damage, and neural population dynamics akin to animals. This task-driven, embodied model illuminates the causal link between neural population dynamics and motor outcomes. We discovered that neural rotations generate oscillatory maneuvers orthogonal to the reaching direction, optimizing trajectory adjustments, which is confirmed by primate neural data. The model also revealed counterintuitive neural energy principles under sensor and motor redundancies, and striking Eureka moments during motor learning, bridging biological and artificial systems. These findings provide new perspectives on how neural dynamics contribute to accurate and flexible movement, inspiring future intelligent robots with animal-like mobility.
Chinese Translation
动物证据表明，精确的自愿运动源于运动皮层中旋转的神经群体动态，但其物理效应仍然未知。我们开发了一种具有人工肌肉、多模态传感器和通过强化学习训练的神经网络控制器的生物运动系统的机器人类比。该机器人类比表现出准确的运动、对损伤的鲁棒性以及与动物相似的神经群体动态。这个以任务为驱动的具身模型阐明了神经群体动态与运动结果之间的因果联系。我们发现，神经旋转产生与到达方向正交的振荡性动作，从而优化轨迹调整，这一点得到了灵长类动物神经数据的验证。该模型还揭示了在传感器和运动冗余下反直觉的神经能量原则，以及在运动学习过程中出现的惊人顿悟时刻，架起了生物系统与人工系统之间的桥梁。这些发现为神经动态如何促进准确和灵活的运动提供了新的视角，并激发了未来具有人类动物般移动能力的智能机器人。
cs.RO / 64 / 2610.10020
Tactile Reconstruction of Contact Task Frames and Forces for Hybrid Force/Motion Control
接触任务框架和力的触觉重建用于混合力/运动控制
Abstract
Hybrid force/motion control requires knowledge of the interaction force and of a task frame defining the force- and motion-controlled directions. These quantities are usually obtained from force/torque sensing or model-based residuals, often assuming also a nominal environment model. This work addresses the online estimation of the contact force and a possibly time-varying task frame using only soft optical tactile sensing, under the assumption of locally planar contact with a negligible contact moment. The proposed method maps a single image of the deformed elastomer of a soft optical tactile sensor to observable contact variables: indentation depth, two surface-to-sensor tilt angles, and 3D contact force, each with a per-sample uncertainty estimate. The mapping is learned through a self-labeling acquisition procedure, in which a manipulator imposes controlled contacts while an auxiliary Force/Torque sensor is used offline to provide ground-truth labels. The tactile measurement is then fused with robot proprioceptive data in an Extended Kalman Filter, producing a continuously updated estimate of the contact task frame and of the interaction force. Control experiments with a DigiTac sensor mounted on a UR10 manipulator demonstrate closed-loop contact force regulation against a flat rigid board in linear and angular motion by a human operator, with touch as the only exteroceptive feedback.
Chinese Translation
混合力/运动控制需要了解交互力以及定义力和运动控制方向的任务框架。这些量通常通过力/扭矩传感器或基于模型的残差获得，通常还假设存在一个名义环境模型。本研究解决了仅使用软光学触觉传感器在线估计接触力和可能随时间变化的任务框架的问题，前提是接触是局部平面且接触力矩可以忽略。所提出的方法将软光学触觉传感器变形橡胶的单幅图像映射到可观察的接触变量：压入深度、两个表面与传感器的倾斜角度以及三维接触力，每个变量都有每个样本的不确定性估计。该映射通过自标记获取程序学习，其中操纵器施加受控接触，同时使用辅助的力/扭矩传感器离线提供真实标签。然后，触觉测量与机器人本体感知数据在扩展卡尔曼滤波器中融合，产生对接触任务框架和交互力的持续更新估计。与安装在UR10操纵器上的DigiTac传感器进行的控制实验表明，在人类操作者的线性和角运动下，能够对平坦刚性板进行闭环接触力调节，触觉是唯一的外部反馈。
cs.RO / 65 / 2610.10039
Video-to-Model: Automatic Modeling of Deformable Linear Objects
视频到模型：可变形线性物体的自动建模
Abstract
This paper presents a video-to-model framework for automatically modeling the motion of a deformable suture thread from an input video. We utilize a recently developed CBF--CLF--QP numerical model that simplifies the characterization of deformable string motion through the selection of a small number of parameters. A perception module first localizes and tracks the thread in video, producing an ordered sequence of thread nodes. The observed thread motion is then processed by a spatio-temporal CNN network that estimates the effective parameters of a structured CBF--CLF--QP model. These parameters are used to simulate the thread under a user-defined needle velocity input. Experiments using unseen thread configurations and motion demonstrate that the framework can reliably reconstruct the thread behavior from video, automatically configure the structured model, and reproduce the expected thread motion with low tracking error. The proposed approach reduces the need for manual parameter tuning and provides a step toward automatic video-based modeling of deformable linear objects.
Chinese Translation
本文提出了一种视频到模型的框架，用于从输入视频中自动建模可变形缝合线的运动。我们利用最近开发的CBF--CLF--QP数值模型，通过选择少量参数简化了可变形线运动的特征描述。感知模块首先在视频中定位并跟踪缝合线，生成有序的缝合线节点序列。随后，观察到的缝合线运动通过时空卷积神经网络（spatio-temporal CNN）进行处理，以估计结构化CBF--CLF--QP模型的有效参数。这些参数用于在用户定义的针头速度输入下模拟缝合线。使用未见过的缝合线配置和运动的实验表明，该框架能够可靠地从视频中重建缝合线行为，自动配置结构化模型，并以低跟踪误差再现预期的缝合线运动。所提出的方法减少了手动参数调节的需求，为基于视频的可变形线性物体自动建模迈出了重要一步。
cs.RO / 66 / 2610.10065
Distributed Motion Planning for Multi-Robot Systems under Topological Constraints
基于拓扑约束的多机器人系统分布式运动规划
Abstract
Efficient and distributed coordination of mobile robots is one of the main challenges in multi-robot systems. Topological constraints, often expressed as topological braids, are a popular tool to encode complex coordination patterns between multiple mobile robots, as they offer a compact and abstract representation of the desired qualitative relation between the space-time trajectories of the robots. However, execution of joint motion plans encoded as braid-based topological constraints via distributed controllers is challenging, with existing approaches, generally based on the execution of one braid generator at a time, producing slow and suboptimal trajectories. We propose a distributed controller based on Model Predictive Control (MPC) to efficiently execute braid-based topological specifications. Rather than directly tracking the braid specification, we propose to use winding numbers, which are topological invariants for braids, as a proxy. This has the twofold benefit of converting braids into a continuous function, which can be easily tracked by an MPC controller through an appropriate term in the cost function, and of decoupling the global braid specification into a set of pairwise specifications, which can be tracked distributedly through the solution of only local MPC problems. To maintain global coordination, we propose a consensus-based progress estimation approach, which allows the robots to synchronize their motion toward the desired specification. We validate the proposed approach in simulation and in real-world experiments, where we demonstrate the effectiveness of the proposed approach and the improvement over existing approaches in terms of execution speed and control effort.
Chinese Translation
移动机器人高效且分布式的协调是多机器人系统面临的主要挑战之一。拓扑约束，通常以拓扑编织的形式表达，是编码多个移动机器人之间复杂协调模式的热门工具，因为它们提供了所需的机器人时空轨迹之间定性关系的紧凑和抽象表示。然而，通过分布式控制器执行编码为基于编织的拓扑约束的联合运动计划是具有挑战性的，现有方法通常基于一次执行一个编织生成器，导致生成缓慢且次优的轨迹。我们提出了一种基于模型预测控制（Model Predictive Control, MPC）的分布式控制器，以高效执行基于编织的拓扑规范。我们建议使用缠绕数作为代理，而不是直接跟踪编织规范，缠绕数是编织的拓扑不变量。这具有双重好处：将编织转换为连续函数，MPC控制器可以通过成本函数中的适当项轻松跟踪；并将全局编织规范解耦为一组成对规范，可以通过仅解决局部MPC问题进行分布式跟踪。为了保持全局协调，我们提出了一种基于共识的进度估计方法，使机器人能够同步其运动以朝向所需规范。我们在仿真和实际实验中验证了所提方法，展示了该方法的有效性以及在执行速度和控制努力方面相较于现有方法的改进。
cs.RO / 67 / 2610.10079
RealtimeWAM: How Fast Can I Run My World Action Model?
实时世界动作模型（RealtimeWAM）：我能多快运行我的世界动作模型？
Abstract
World Action Models (WAMs) combine visual dynamics modeling with action generation, but their high inference latency limits responsive robot control. Recent efforts accelerate inference by removing explicit future-video generation at test time, as in FastWAM, an approach that requires a specially tailored architectural design. More general caching strategies exploit feature redundancy, but redundancy alone does not capture the changing computational demands of closed-loop control. To address these challenges, we present RealtimeWAM, a general, training-free framework that coordinates parallel execution with adaptive computation for low-latency inference across diverse WAM architectures. We exploit layerwise dependencies to overlap observation processing with prediction. However, concurrent branches still compete for GPU resources, limiting the benefit of parallel execution. We therefore adapt computation throughout the pipeline through selective reuse, caching observation features in visually stable regions and reusing Transformer residuals while reserving additional refinement for small predicted adjustments. We evaluate RealtimeWAM on FastWAM and OpenWAM across RoboTwin, LIBERO, and LIBERO-Plus. On an RTX 4090, measured mean inference latencies are 24.09 and 63.09 ms, corresponding to average speedups of 8.90$\times$ and 10.67$\times$. Average success rates are 82.75% and 87.41%, respectively, within 0.02 and 0.53 percentage points of native inference. Across five real-world tasks, RealtimeWAM improves average success rates over native inference by 17.2 and 37.2 percentage points on FastWAM and OpenWAM, respectively.
Chinese Translation
世界动作模型（World Action Models, WAMs）结合了视觉动态建模与动作生成，但其高推理延迟限制了机器人控制的响应性。最近的研究通过在测试时去除显式的未来视频生成来加速推理，例如FastWAM，这种方法需要特别定制的架构设计。更通用的缓存策略利用特征冗余，但仅靠冗余无法捕捉闭环控制的变化计算需求。为了解决这些挑战，我们提出了RealtimeWAM，这是一个通用的、无训练的框架，协调并行执行与自适应计算，以实现低延迟推理，适用于多种WAM架构。我们利用层级依赖关系，将观察处理与预测重叠。然而，同时进行的分支仍然竞争GPU资源，限制了并行执行的收益。因此，我们通过选择性重用在整个管道中调整计算，在视觉稳定区域缓存观察特征，并在保留额外细化用于小幅预测调整的同时重用Transformer残差。我们在RoboTwin、LIBERO和LIBERO-Plus上评估了RealtimeWAM在FastWAM和OpenWAM上的表现。在RTX 4090上，测得的平均推理延迟分别为24.09毫秒和63.09毫秒，平均加速比为8.90×和10.67×。平均成功率分别为82.75%和87.41%，与原生推理的差距为0.02和0.53个百分点。在五个真实任务中，RealtimeWAM在FastWAM和OpenWAM上分别提高了平均成功率17.2和37.2个百分点，超越了原生推理。
cs.RO / 68 / 2610.10125
Lifelong small-object navigation in changing object layouts: a benchmark and method
在变化物体布局中的终身小物体导航：基准与方法
Abstract
Household robots need to continually navigate to different objects in the same environment, many of which are small and portable, such as tools and toys. Their small visual footprint and frequent occlusion make reliable observation difficult, and they may be moved by people without the robot observing the changes. We formulate this challenging task as Lifelong Small-object Navigation in Changing Object Layouts (LiSoNav-COL). Agents must seek suitable viewpoints for reliable observation, accumulate and reuse scene knowledge to efficiently locate subsequent targets, and update outdated memory after object relocation. To eliminate the need for prior scene scanning, we also require agents to start navigation with empty scene memory. Although practical, this task still lacks benchmarks designed around its defining assumptions. To bridge this gap, we introduce LiSoNav-Eval, a dedicated benchmark spanning 28 indoor scenes with 45 small-object categories. Its lifelong navigation sequences include both unchanged and relocated targets to evaluate memory reuse and adaptation to object relocation. To address this challenging task, we propose a navigation method based on multi-view Inspection with Viewpoint-Anchored Memory, dubbed IVAM-Nav. IVAM-Nav actively observes supporting surfaces from complementary viewpoints for reliable small-object perception and anchors the resulting memory to their observation viewpoints, supporting relational memory reuse and revalidation under similar viewing conditions. Extensive experiments on LiSoNav-Eval demonstrate favorable performance of IVAM-Nav against representative methods. Benchmark analyses also show that smaller objects, larger environments, and longer relocation distances pose greater challenges. The dataset and code are available here.
Chinese Translation
家庭机器人需要在同一环境中不断导航到不同的物体，其中许多物体较小且便于携带，如工具和玩具。它们小的视觉占据面积和频繁的遮挡使得可靠观察变得困难，并且可能在机器人未观察到变化的情况下被人移动。我们将这一具有挑战性的任务定义为在变化物体布局中的终身小物体导航（Lifelong Small-object Navigation in Changing Object Layouts，LiSoNav-COL）。代理必须寻找合适的视点以进行可靠观察，积累并重用场景知识以有效定位后续目标，并在物体重新定位后更新过时的记忆。为了消除对先前场景扫描的需求，我们还要求代理在导航时从空的场景记忆开始。尽管这一任务具有实际意义，但仍缺乏围绕其定义假设设计的基准。为填补这一空白，我们引入了LiSoNav-Eval，这是一个专门的基准，涵盖28个室内场景和45个小物体类别。其终身导航序列包括未变化和重新定位的目标，以评估记忆重用和适应物体重新定位的能力。为了解决这一挑战性任务，我们提出了一种基于多视角检查与视点锚定记忆的导航方法，称为IVAM-Nav。IVAM-Nav主动从互补视点观察支撑表面，以实现可靠的小物体感知，并将生成的记忆锚定到其观察视点，支持在相似观察条件下的关系记忆重用和重新验证。在LiSoNav-Eval上的大量实验表明，IVAM-Nav在与代表性方法的比较中表现良好。基准分析还显示，较小的物体、更大的环境和更长的重新定位距离带来了更大的挑战。数据集和代码可在此处获取。
cs.RO / 69 / 2610.10178
Do Vision-Language-Action Models Understand Instructions? A Mechanistic Interpretability Study on Language Grounding
视觉-语言-行动模型是否理解指令？关于语言基础的机制可解释性研究
Abstract
Vision-Language-Action models are designed to generalise across environments and task descriptions, raising the question of whether their action generation actually depends on the language instruction, or whether they largely rely on visual cues and superficial correlations. Robustness to variance in the visual and linguistic observation space is critical for real-world deployment, yet VLAs lack explicit grounding modules and instead rely on the intrinsic language grounding capabilities of their Vision-Language model backbones. For this reason, we conduct a controlled mechanistic interpretability study on the language grounding capabilities of two state-of-the-art Vision-Language-Action models, $\pi_{0.5}$ and GR00T N1.7, by applying activation and attribution patching to the residual stream of the action generation modules. We systematically corrupt the task instruction of input samples of the LIBERO benchmark following five strategies: synonym replacement, semantic scaling, directional corruption, random object substitution, and empty string. Our experiments find that both models are comparatively insensitive to abstract rephrasing and to referencing non-existent objects, but react strongly to empty task descriptions and, especially, to directional language. During action generation, this sensitivity is concentrated in different loci for each model: mainly in the early, periodic cross-attention layers for GR00T N1.7, versus distributed across the earliest and selected later layers for $\pi_{0.5}$. For GR00T N1.7, directional perturbations drive some of the largest causal effects while leaving the internal representational geometry comparatively unchanged, a dissociation we do not observe clearly for $\pi_{0.5}$. Finally, the reliability of attribution patching is model-dependent: it closely tracks activation patching for GR00T N1.7 but not for $\pi_{0.5}$.
Chinese Translation
视觉-语言-行动（VLA）模型旨在跨环境和任务描述进行泛化，这引发了一个问题：它们的行动生成是否真正依赖于语言指令，还是主要依赖于视觉线索和表面相关性。对于现实世界的部署，视觉和语言观察空间的变异性鲁棒性至关重要，然而VLA缺乏明确的基础模块，而是依赖于其视觉-语言模型骨干的内在语言基础能力。因此，我们对两个最先进的视觉-语言-行动模型$ ext{π}_{0.5}$和GR00T N1.7的语言基础能力进行了控制的机制可解释性研究，通过对行动生成模块的残差流应用激活和归因补丁。我们系统地按照五种策略腐蚀LIBERO基准的输入样本的任务指令：同义词替换、语义缩放、方向性腐蚀、随机对象替换和空字符串。我们的实验发现，这两个模型对抽象的重述和引用不存在的对象相对不敏感，但对空任务描述，尤其是对方向性语言反应强烈。在行动生成过程中，这种敏感性在每个模型中集中在不同的区域：对于GR00T N1.7，主要集中在早期的周期性交叉注意层，而对于$ ext{π}_{0.5}$则分布在最早和选定的后续层。对于GR00T N1.7，方向性扰动驱动了一些最大的因果效应，同时内部表示几何相对不变，这种解离在$ ext{π}_{0.5}$中并未明显观察到。最后，归因补丁的可靠性依赖于模型：对于GR00T N1.7，它与激活补丁密切相关，但对于$ ext{π}_{0.5}$则不然。
cs.RO / 70 / 2610.10180
Design of a Fully Actuated 4-DOF Robotic Finger With Joint-Specific Hybrid Remote Actuation
具有关节特定混合远程驱动的全驱动4自由度机器人手指设计
Abstract
This paper presents a fully actuated 4-DOF robotic finger using a joint-specific hybrid remote-actuation architecture. The metacarpophalangeal (MCP) joint is driven by two coordinated rigid-link transmission sets, whereas the proximal interphalangeal (PIP) and distal interphalangeal (DIP) joints are independently actuated by closed-loop wire transmissions incorporating circular rolling-contact joints (RCJs). A larger transmission radius is used at the PIP joint than at the DIP joint. The RCJ wire geometry maintains the total wire-loop length during joint rotation, and the DIP wire routing is designed so that PIP motion does not affect its differential actuation for a fixed MCP configuration. The distal wire transmission, MCP linkage, and fingertip kinematics are analytically modeled. In experiments, the transmission behavior was quantitatively evaluated from the ball-screw displacements measured using ArUco-marker tracking. MCP actuation produced measurable displacements of the PIP and DIP transmission units, whereas isolated PIP and DIP actuation with the MCP fixed supported the intended mechanical decoupling between the distal transmissions. The mean peak fingertip forces under isolated MCP, PIP, and DIP actuation were 21.28 N, 9.22 N, and 5.75 N, respectively. The resulting finger postures were also examined using objects of different geometries and sizes.
Chinese Translation
本文提出了一种使用关节特定混合远程驱动架构的全驱动4自由度机器人手指。掌指关节（MCP）由两个协调的刚性链传动系统驱动，而近端指间关节（PIP）和远端指间关节（DIP）则通过闭环电缆传动独立驱动，后者结合了圆形滚动接触关节（RCJ）。在PIP关节处使用的传动半径大于DIP关节。RCJ电缆几何形状在关节旋转过程中保持电缆环的总长度，而DIP电缆布置设计使得PIP运动不会影响其在固定MCP配置下的差分驱动。对远端电缆传动、MCP连杆和指尖运动学进行了分析建模。在实验中，通过使用ArUco标记跟踪测量的滚珠丝杠位移定量评估了传动行为。MCP驱动产生了PIP和DIP传动单元的可测位移，而在固定MCP的情况下，孤立的PIP和DIP驱动支持了远端传动之间预期的机械解耦。孤立MCP、PIP和DIP驱动下的平均峰值指尖力分别为21.28 N、9.22 N和5.75 N。还使用不同几何形状和尺寸的物体检查了所得到的手指姿态。
cs.RO / 71 / 2610.10198
Benchmarking Behavioral Steerability in Behavior Foundation Models
行为基础模型中的行为可操控性基准测试
Abstract
Behavior Foundation Models (BFMs) are emerging as a paradigm for translating human intentions into executable humanoid behaviors. As these models evolve beyond behavior generation toward general-purpose behavioral systems, a fundamental question arises: can they be reliably steered according to user intentions? In this paper, we introduce the concept of behavioral steerability, defined as the ability of BFMs to faithfully generate behaviors that satisfy user-specified intentions. To study this capability, we present RoboSteer, the first benchmark for behavioral steerability in BFMs. RoboSteer organizes behavioral steerability into a three-level hierarchy-Conditional Steering, Constraint Steering, and Compositional Steering-and establishes a unified evaluation framework supported by a large-scale multimodal motion corpus. Using RoboSteer, we conduct the first large-scale empirical study of behavioral steerability across 9 existing BFMs. We view behavioral steerability as more than a capability for controlling motion: it concerns how embodied systems translate human intentions into purposeful actions. We hope RoboSteer will advance research on intention realization as a foundation for general-purpose embodied intelligence.
Chinese Translation
行为基础模型（BFMs）正在成为将人类意图转化为可执行类人行为的范式。随着这些模型从行为生成演变为通用行为系统，一个基本问题随之而来：它们能否根据用户意图可靠地进行操控？在本文中，我们引入了行为可操控性的概念，定义为BFMs忠实生成满足用户指定意图的行为的能力。为了研究这一能力，我们提出了RoboSteer，这是第一个针对BFMs行为可操控性的基准测试。RoboSteer将行为可操控性组织为三级层次结构——条件操控、约束操控和组合操控，并建立了一个统一的评估框架，支持大规模多模态运动语料库。利用RoboSteer，我们对9个现有BFMs进行了首次大规模的行为可操控性实证研究。我们认为行为可操控性不仅仅是控制运动的能力：它涉及具身系统如何将人类意图转化为有目的的行动。我们希望RoboSteer能够推动意图实现的研究，为通用具身智能奠定基础。
cs.RO / 72 / 2610.10281
Making Task Abstractions Executable: Control-Aware Layout Repair for a Fixed Controller
使任务抽象可执行：针对固定控制器的控制感知布局修复
Abstract
A task abstraction can specify the intended events while its spatial layout prevents a fixed agent and controller from completing them. Starting from a supplied structured task record, we compile whole-task tracking, clearance, and actuation requirements into auditable affine layout constraints. We repair only declared continuous coordinates, preserving event order, timing, topology, and the controller. A most-violated-row update admits conditional finite-certification and net-displacement bounds; a same-compiler quadratic projection separates the representation from the optimizer. On three researcher-authored task abstractions, both backends certify all three layouts and complete all 300 fresh paired rollouts per backend. A risk-target sweep also exposes fixed event tests that the chosen certificate cannot satisfy through layout edits alone.
Chinese Translation
任务抽象可以指定预期事件，而其空间布局则阻止固定代理和控制器完成这些事件。从提供的结构化任务记录出发，我们将整体任务跟踪、清理和执行要求编译为可审计的仿射布局约束。我们仅修复声明的连续坐标，保留事件顺序、时序、拓扑结构和控制器。最违反行的更新允许条件有限认证和净位移界限；相同编译器的二次投影将表示与优化器分离。在三种研究者撰写的任务抽象中，两个后端均认证所有三种布局，并完成每个后端的300个新配对滚动。风险目标扫描还揭示了固定事件测试，所选择的证书无法仅通过布局编辑满足。
cs.RO / 73 / 2610.10283
Temporal Visuo-Tactile Learning for Dexterous Grasp Stability
用于灵巧抓握稳定性的时间视觉-触觉学习
Abstract
Humans can grasp everyday objects with almost perfect success rates using fingertip tactile feedback, yet much of the robotic grasping literature emphasizes vision-based grasp selection with parallel grippers. In this work, we systematically investigate how high-resolution, dynamic tactile sensing contributes to grasp stability prediction and model-guided grasping in dexterous robotic hands. To this end, we collected a dataset of 10,000 grasp trials across 200 objects using a multi-fingered robotic hand equipped with four Digit 360 tactile sensors, recording external vision, proprioception, and tactile streams throughout each grasp. With this dataset, we trained end-to-end temporal multimodal models to predict post-lift stability from pre-lift grasp observations and compared sensing modalities and encoding backbones. Experimental results and controlled input ablations show that incorporating touch, and particularly high-resolution, dynamic touch, improves grasp stability prediction. Finally, we deployed the learned predictor as an online stability gate on the real robot, where visuo-tactile model-guided regrasping improved the success rate among executed lifts by 10.5 percentage points over a non-tactile gate. These results show how rich fingertip sensing and expressive temporal models that capture the dynamics of touch can support learned grasping with multi-fingered hands without explicit contact or force modeling, providing a scalable data-driven path from tactile experience toward stable dexterous manipulation. The dataset is publicly available at https://lasr-lab.github.io/dexterous-grasp-stability/.
Chinese Translation
人类能够利用指尖触觉反馈以几乎完美的成功率抓取日常物品，而现有的机器人抓取文献则强调基于视觉的抓取选择和并行夹爪。在本研究中，我们系统地探讨了高分辨率动态触觉感知如何促进抓取稳定性预测和模型引导的灵巧机器人手抓取。为此，我们使用配备四个Digit 360触觉传感器的多指机器人手收集了200个物体的10,000次抓取试验数据集，记录了每次抓取过程中的外部视觉、身体感知和触觉流。利用该数据集，我们训练了端到端的时间多模态模型，以根据提起前的抓取观察预测提起后的稳定性，并比较了感知模态和编码骨干网络。实验结果和控制输入消融实验表明，结合触觉，特别是高分辨率动态触觉，能够改善抓取稳定性预测。最后，我们将学习到的预测器作为在线稳定性门部署在真实机器人上，其中视觉-触觉模型引导的重新抓取使得执行的提起成功率比非触觉门提高了10.5个百分点。这些结果表明，丰富的指尖感知和能够捕捉触觉动态的表达性时间模型能够支持多指手的学习抓取，而无需显式的接触或力模型，提供了一条从触觉经验到稳定灵巧操作的可扩展数据驱动路径。数据集已公开，网址为 https://lasr-lab.github.io/dexterous-grasp-stability/.
cs.RO / 74 / 2610.10297
Energy-Efficient Gait Adaptation via Hierarchical Reinforcement Learning for Quadrupedal Locomotion Across Diverse Terrains
基于分层强化学习的能效步态适应：四足机器人在多样地形中的运动
Abstract
While energy efficiency is a critical objective for legged-robot locomotion control, achieving low energy consumption while maintaining robust performance across different velocity ranges and terrain conditions remains a key challenge. This is particularly true for end-to-end RL policies, where gait generation, motion execution, and energy optimization are tightly coupled, leading to high sensitivity to reward design. In this work, we propose a hierarchical reinforcement learning (HRL) framework that separates a high-frequency policy for stable and robust joint-level motion execution from low-frequency gait adaptation that explicitly minimizes the cost of transport (CoT). The three-stage Isaac-based training procedure enables zero-shot sim-to-real transfer with improved tracking accuracy, robustness, and energy efficiency. The learned hierarchy exhibits automatic speed-dependent gait adaptation, transitioning from pacing at low speeds to trotting at higher speeds. We validate the proposed approach in simulation against representative single-policy and hierarchical locomotion baselines, demonstrating reduced CoT over a broad range of commanded velocities, while maintaining robust locomotion across flat, uneven rough, and inclined terrains. We further demonstrate its practical feasibility through zero-shot deployment on a physical Unitree AlienGo quadruped.
Chinese Translation
能效是腿式机器人运动控制的关键目标，但在不同速度范围和地形条件下实现低能耗并保持稳健性能仍然是一个主要挑战。这一点在端到端的强化学习（RL）策略中尤为明显，因为步态生成、运动执行和能量优化紧密耦合，导致对奖励设计高度敏感。在本研究中，我们提出了一种分层强化学习（HRL）框架，将用于稳定和稳健的关节级运动执行的高频策略与明确最小化运输成本（CoT）的低频步态适应分开。基于Isaac的三阶段训练程序实现了零样本的仿真到现实转移，提升了跟踪精度、稳健性和能效。学习到的层次结构表现出自动的速度依赖步态适应，从低速的步行过渡到高速的慢跑。我们在仿真中验证了所提方法，与代表性的单一策略和分层运动基线进行比较，展示了在广泛的指令速度范围内降低CoT，同时在平坦、不平整和倾斜地形上保持稳健的运动。我们进一步通过在物理Unitree AlienGo四足机器人上的零样本部署证明了其实际可行性。
cs.RO / 75 / 2610.10346
Hall Effect-Based Tactile Force Detection Sensor for Robot-Assisted Minimally Invasive Surgery
基于霍尔效应的触觉力检测传感器用于机器人辅助微创手术
Abstract
Robotic minimally invasive surgery has revolutionized surgical practice. However, as the surgeons are mechanically separated from the surgical tools, loss of tactile feedback occurs. We present a compact force sensor with three, millimeter scale sensing elements (taxels) that can be affixed to the grasping face of a robotic surgical tool. The sensor features a miniaturized design using Hall effect sensors and magnets embedded in a deformable elastomer contact layer. Sensor calibration is achieved with a 6 degree of freedom (DoF) parallel robot. Each taxel can detect applied normal and shear forces with average RMSE of 2.133 kPa. The calibrated sensor was validated on phantom and ex vivo porcine tissue, demonstrating tactile sensing during retraction as well as slip. This work introduces a small-scale Hall effect based tactile sensor, enabling sensing for force-sensitive robotic surgical applications.
Chinese Translation
机器人微创手术已经彻底改变了外科手术实践。然而，由于外科医生与手术工具之间的机械隔离，导致触觉反馈的丧失。我们提出了一种紧凑型力传感器，配备三个毫米级的传感元件（taxels），可固定在机器人手术工具的抓握面上。该传感器采用微型化设计，使用嵌入在可变形弹性体接触层中的霍尔效应传感器和磁铁。传感器的标定是通过一个具有6自由度（DoF）的并联机器人实现的。每个taxel能够检测施加的法向力和剪切力，平均均方根误差（RMSE）为2.133 kPa。经过标定的传感器在假体和体外猪组织上进行了验证，展示了在回缩和滑动过程中的触觉感知。该研究介绍了一种小型霍尔效应基触觉传感器，为力敏感的机器人手术应用提供了感知能力。
cs.RO / 76 / 2610.10359
MultiFly: A Real-World Multimodal Aerial Dataset with Annotation-Efficient Label Transfer and Cross-Modal Semantic Consistency
MultiFly：一个具有高效标注转移和跨模态语义一致性的真实世界多模态空中数据集
Abstract
We introduce MultiFly, a real-world, low-altitude UAV dataset for semantic perception across RGB, thermal, LiDAR, and radar modalities. MultiFly provides 17,272 synchronized samples from four suburban scenes with frame-wise annotations for 15 semantic classes, together with calibration and GNSS-RTK/IMU measurements. To avoid costly and inconsistent modality-specific annotation, we propagate labels from only 115 manually annotated RGB images through shared geometric representations to all four modalities. This approach generates semantic labels for 17,157 additional RGB images, 17,272 thermal images, 840M LiDAR points, and 3.4M radar points. Transferred annotations achieve 89.93% average agreement with held-out manual annotations, and 90.94% average semantic consistency across all six modality pairs. We further establish semantic segmentation benchmarks for all four modalities, revealing distinct architectural behavior for dense LiDAR and sparse radar data. Taken together, MultiFly provides a scalable foundation for multimodal aerial perception and, to the best of our knowledge, the first public real-world low-altitude aerial benchmark that combines consistent frame-wise semantic annotations for RGB, thermal, LiDAR, and radar. Data at https://github.com/markus-42/multifly.
Chinese Translation
我们介绍了MultiFly，这是一个用于RGB、热成像、LiDAR和雷达模态的真实世界低空无人机数据集，旨在进行语义感知。MultiFly提供了来自四个郊区场景的17,272个同步样本，并为15个语义类别提供逐帧标注，同时包含标定和GNSS-RTK/IMU测量数据。为了避免昂贵且不一致的模态特定标注，我们仅通过共享几何表示从115张手动标注的RGB图像传播标签到所有四个模态。这种方法为17,157张额外的RGB图像、17,272张热成像图像、840M LiDAR点和3.4M雷达点生成了语义标签。转移的标注与保留的手动标注之间的平均一致性达到89.93%，而在所有六对模态之间的平均语义一致性达到90.94%。我们进一步为所有四个模态建立了语义分割基准，揭示了密集LiDAR和稀疏雷达数据的不同架构行为。综上所述，MultiFly为多模态空中感知提供了可扩展的基础，并且据我们所知，这是第一个结合RGB、热成像、LiDAR和雷达的一致逐帧语义标注的公共真实世界低空空中基准。数据可在https://github.com/markus-42/multifly获取。
cs.RO / 77 / 2610.10382
Semantic-Aware Predictive Mapping for Exploration and Navigation
语义感知的预测映射用于探索与导航
Abstract
Predictive mapping can support robotic exploration and navigation by estimating unseen geometric layouts from partial occupancy observations. However, occupancy-only representations may fail to distinguish semantically different structures with similar geometry. This is particularly relevant for indoor doors, which may appear as occupied cells like walls but indicate possible connected rooms or corridors beyond the observed region. This work investigates whether semantic door cues improve predictive geometric occupancy mapping around such ambiguous regions. We modify a subset of the CogniPlan dataset by inserting door-induced ambiguities into partial occupancy maps while keeping the ground-truth layouts unchanged. We compare a geometry-only control model with a semantic-cued model trained on the same modified dataset, where the semantic-cued model receives an additional door channel. Evaluation uses L1 error, F1 score, and Intersection over Union (IoU) over both the full map and a 10-pixel door-region mask. Full-map performance remains broadly similar between models, but localized door-region results show a clear qualitative improvement: L1 decreases from 0.004342 to 0.000025, while F1 and IoU improve from 0.031311 and 0.015905 to 1.000000 and 1.000000, respectively. These results suggest that semantic cues can improve predictive occupancy completion in regions where geometric observations alone are ambiguous.
Chinese Translation
预测映射可以通过从部分占用观测中估计未见几何布局来支持机器人探索与导航。然而，仅基于占用的表示可能无法区分具有相似几何形状的语义上不同的结构。这对于室内门尤其相关，因为它们可能看起来像墙壁一样占用单元，但表示在观察区域之外可能连接的房间或走廊。本研究探讨了语义门线索是否能改善在这些模糊区域周围的预测几何占用映射。我们通过在部分占用地图中插入门引起的模糊性，同时保持真实布局不变，修改了CogniPlan数据集的一个子集。我们将仅基于几何的控制模型与在相同修改数据集上训练的语义线索模型进行比较，其中语义线索模型接收额外的门通道。评估使用L1误差、F1分数和交并比（IoU）在完整地图和10像素门区域掩膜上进行。完整地图的性能在模型之间大致相似，但局部门区域的结果显示出明显的定性改善：L1从0.004342降至0.000025，而F1和IoU分别从0.031311和0.015905提高到1.000000和1.000000。这些结果表明，语义线索可以改善在几何观测单独模糊的区域中的预测占用补全。
cs.RO / 78 / 2610.10384
OpenViTac: Learning and Benchmarking Visuo-Tactile Policies in a Unified Sim-and-Real Framework
OpenViTac：在统一的仿真与现实框架中学习和基准测试视觉-触觉策略
Abstract
Tactile feedback provides embodied agents with physical information beyond visual observations, enabling more reliable interaction with the real world. However, despite the rapid progress of vision-tactile-language-action (VTLA) policies, there remains a lack of unified benchmarks for evaluating tactile-enabled robot manipulation across simulation and the real world. To address this gap, we introduce OpenViTac, a visuo-tactile manipulation benchmark for evaluating robot policies across simulation and the real world. OpenViTac organizes contact-rich manipulation into four tactile-relevant capability dimensions and provides paired simulation-real-world settings for consistent evaluation of VLA, WAM, and VTLA policies. Building upon this benchmark, we investigate how different tactile representations and integration strategies affect the performance of pretrained VLA models. Correspondingly, we introduce OpenVTLA, a tactile augmentation framework that combines the best-performing representation and integration strategy. Furthermore, we leverage the paired benchmark setting to study sim-real co-training and analyze factors affecting cross-domain policy learning. Together, OpenViTac provides a unified platform for evaluating and advancing visuo-tactile robot manipulation.
Chinese Translation
触觉反馈为具身智能体提供了超越视觉观察的物理信息，从而使其能够更可靠地与现实世界进行交互。然而，尽管视觉-触觉-语言-动作（VTLA）策略迅速发展，仍然缺乏统一的基准来评估触觉增强的机器人操作在仿真和现实世界中的表现。为了解决这一问题，我们推出了OpenViTac，一个用于评估机器人策略的视觉-触觉操作基准，涵盖仿真和现实世界。OpenViTac将接触丰富的操作组织为四个与触觉相关的能力维度，并提供配对的仿真-现实世界设置，以便对VLA、WAM和VTLA策略进行一致的评估。在此基准的基础上，我们研究了不同触觉表示和集成策略如何影响预训练VLA模型的性能。相应地，我们引入了OpenVTLA，一个触觉增强框架，结合了表现最佳的表示和集成策略。此外，我们利用配对的基准设置研究仿真-现实共同训练，并分析影响跨领域策略学习的因素。总之，OpenViTac提供了一个统一的平台，用于评估和推动视觉-触觉机器人操作的发展。
cs.RO / 79 / 2610.10387
NeRFifyMesh: Optimizing Neural Radiance Fields from Textured Meshes for Robotics Scene Building
NeRFifyMesh：从纹理网格优化神经辐射场以构建机器人场景
Abstract
In robotics, scene representation plays a pivotal role in understanding and interacting with the environment. The advent of Neural Radiance Fields (NeRF) and its variants, as a novel representation, has opened a new frontier of research. In applications such as semantic mapping and simulation, roboticists aim to build scenes using multiple NeRF models, each representing an object. While extensive datasets of 3D mesh models already exist, there is an urgent need to develop tools to convert these assets to NeRF models for rapid algorithm development and testing. This paper presents a new pipeline for converting existing mesh models to NeRF representations by artificially generating a ground truth point-based radiance field through sampling mesh geometry and texture. This approach alleviates the need for camera-based sampling or rendering multi-view images of the original mesh to train the NeRF model. Extensive benchmarking demonstrates that our method yields comparable rendering quality to the baselines. Additionally, the application of this representation is shown by constructing unified NeRF scenes and performing collision simulations with extracted geometry.
Chinese Translation
在机器人技术中，场景表示在理解和与环境互动中发挥着关键作用。神经辐射场（NeRF）及其变体作为一种新颖的表示方法的出现，开启了研究的新前沿。在语义映射和仿真等应用中，机器人研究人员旨在使用多个NeRF模型构建场景，每个模型代表一个对象。尽管已经存在大量的3D网格模型数据集，但迫切需要开发工具将这些资产转换为NeRF模型，以便快速进行算法开发和测试。本文提出了一种新的流程，通过对网格几何和纹理进行采样，人工生成基于点的辐射场的真实值，从而将现有的网格模型转换为NeRF表示。这种方法减轻了对基于相机的采样或渲染原始网格的多视图图像以训练NeRF模型的需求。广泛的基准测试表明，我们的方法在渲染质量上与基线相当。此外，通过构建统一的NeRF场景并对提取的几何体进行碰撞仿真，展示了这种表示的应用。
cs.RO / 80 / 2610.10388
RoboQuest: Generalist Physical Agents that Search, Inspect and Test
RoboQuest：通用物理智能体的搜索、检查与测试
Abstract
Recent advances in multimodal foundation models have made them capable generalist physical agents for a range of manipulation tasks. However, successful operation in an unfamiliar environment may require an agent to seek task-relevant information through interaction when it is absent from the observations: it may need to determine where a relevant object is, inspect an unobserved property, or discover the effect of an unfamiliar tool. We thus introduce RoboQuest, a benchmark for goal-directed embodied exploration, where agents must actively acquire task-relevant information through physical interaction, use the resulting evidence to adapt subsequent actions, and autonomously decide when to commit to task completion. RoboQuest comprises ten mobile manipulation tasks centered on three forms of uncertainty: search, manipulation-based inspection, and interactive testing. We evaluate five frontier multimodal agents through a common visuomotor interface, as well as a $\pi_{0.5}$ policy fine-tuned on the full-episode demonstrations we release. The best agent succeeds in only 23\% of the episodes, and the fine-tuned policy almost never succeeds. Isolated tests of the execution skills the tasks are built from, with the hidden information supplied, show that the agents can carry out most of the required actions, and our failure analysis attributes only a minority of the failures to execution. Our failure analysis further finds that the agents often stop exploring too early as they make decisions before observing the required evidence for task completion. We also find that agents rarely prevent or repair the disturbances caused by their exploration. Moreover, learning by trial and error remains difficult for most models.
Chinese Translation
近期多模态基础模型的进展使其能够作为通用物理智能体执行一系列操作任务。然而，在陌生环境中成功操作可能需要智能体通过交互来获取缺失的与任务相关的信息：它可能需要确定相关物体的位置、检查未观察到的属性，或发现不熟悉工具的效果。因此，我们引入了RoboQuest，一个针对目标导向的具身探索的基准，其中智能体必须通过物理交互主动获取与任务相关的信息，利用所获得的证据调整后续行动，并自主决定何时完成任务。RoboQuest包含十个以移动操控为中心的任务，围绕三种不确定性形式展开：搜索、基于操控的检查和交互测试。我们通过一个共同的视觉运动接口评估五个前沿多模态智能体，以及一个在我们发布的完整回合演示上微调的$ ext{π}_{0.5}$策略。表现最好的智能体仅在23%的回合中成功，而微调的策略几乎从未成功。对构建任务的执行技能进行的孤立测试（提供隐藏信息）显示，智能体能够执行大多数所需的动作，而我们的失败分析仅将少数失败归因于执行问题。我们的失败分析进一步发现，智能体常常过早停止探索，因为它们在观察到完成任务所需的证据之前就做出了决策。我们还发现，智能体很少防止或修复其探索所造成的干扰。此外，对于大多数模型而言，通过试错学习仍然困难。
cs.RO / 81 / 2610.10409
RobotWorld: Benchmarking Multimodal Agents for Robot Use Across Diverse Tasks and Embodiments
RobotWorld：针对多模态智能体在多样任务和形态中的机器人使用进行基准测试
Yang, Zhiqin, Li, Chenxin, Hu, Xiaomeng, Liu, Yibin, Huang, Weidong, Sun, Jiankai, Li, Haitao, Wu, Zijian, Huang, Yuzhi, Huang, Fanding, Sun, Hanwen, Liu, Jiashun, Tong, Jingqi, Huang, Mingxin, Hu, Shaoli, Huang, Shijue, Bai, Tianyi, Wang, Xinyuan, Lin, Yunlong, Tang, Zhengyang, Zhang, Zhexin, Chen, Zhuo, Song, Xierui, Dai, Juntao, Chen, Boyuan, Ji, Jiaming, Zhan, Fangneng, Hu, Mengkang, Xue, Wei, Zhang, Yonggang, Hu, Han, Ho, Tsung-Yi, Guo, Yike
Abstract
General-purpose agents increasingly write code, use tools, and complete complex digital tasks, raising the question of how far these capabilities carry into the physical world. To investigate this, we introduce RobotWorld, a challenging simulation testbed for robot use: turning instructions and observations into physical task execution through robot interfaces. Its 84 tasks span manipulation, mobile manipulation, locomotion, driving, and aerial control, with explicit interaction budgets and executable success checks. By analysing task outcomes alongside execution traces, we identify both the capabilities that transfer and the gaps that prevent reliable completion. Furthermore, we find that current agents can construct sophisticated perception and control workflows, including image segmentation, camera calibration, spatial estimation, and dynamics-based computation. These capabilities, however, do not consistently compose into successful behaviour: agents lose task-relevant object states despite reaching commanded poses, fail to correct ineffective actions, recover too late, or mistake unfinished tasks for completion. This uneven transfer also differs across models: Astra succeeds more often on spatial and constrained-contact goals, whereas Opus 5.5 succeeds more often on continuous-balance and timed-interaction goals. By linking these outcomes to execution behaviour, RobotWorld provides both a rigorous proving ground and an empirical account of the remaining capability gaps, thereby establishing concrete targets for training and designing more reliable physical-world agents.
Chinese Translation
通用智能体越来越多地编写代码、使用工具并完成复杂的数字任务，这引发了这些能力在物理世界中的延伸程度的疑问。为此，我们引入了RobotWorld，一个用于机器人使用的挑战性仿真测试平台：通过机器人接口将指令和观察转化为物理任务执行。其84个任务涵盖了操控、移动操控、步态、驾驶和空中控制，并设定了明确的交互预算和可执行的成功检查。通过分析任务结果和执行轨迹，我们识别出可转移的能力以及阻碍可靠完成的差距。此外，我们发现当前的智能体能够构建复杂的感知和控制工作流程，包括图像分割、相机校准、空间估计和基于动态的计算。然而，这些能力并未始终有效地组合成成功的行为：智能体在达到指令姿态后，丢失了与任务相关的物体状态，未能纠正无效的动作，恢复过晚，或将未完成的任务误认为已完成。这种不均匀的转移在不同模型之间也有所不同：Astra在空间和受限接触目标上更常成功，而Opus 5.5在连续平衡和定时交互目标上更常成功。通过将这些结果与执行行为联系起来，RobotWorld不仅提供了一个严格的试验场，还为剩余能力差距提供了实证记录，从而为训练和设计更可靠的物理世界智能体建立了具体目标。
cs.RO / 82 / 2610.10421
AirGroundVLN: A Large-Scale Benchmark for Goal-Oriented Air-Ground Collaborative Vision-and-Language Navigation
AirGroundVLN：面向目标的空地协作视觉与语言导航的大规模基准
Abstract
Goal-oriented Vision-and-Language Navigation (VLN) requires agents to locate and reach targets described in natural language without prescribed routes. Air--ground collaboration is valuable for tasks requiring both wide-area search and fine-grained localization. However, systematic study of goal-oriented air--ground collaborative VLN remains limited by the lack of large-scale, diverse benchmarks and two core challenges: 1) substantial differences between aerial and ground views, together with useful observations becoming unavailable as navigation proceeds, make it difficult to maintain spatially consistent context across platforms and over time; and 2) asymmetric spatial observability makes ground perception locally detailed but spatially limited and aerial perception broad but locally coarse, limiting the reliability of single-platform planning. To address these limitations, we introduce AirGroundVLN, a benchmark containing 10,281 navigation episodes and 955 target instances across 19 Unreal Engine environments, with seen/unseen splits and an aerial-visibility protocol for systematic evaluation. Alongside the benchmark, we propose AG-CoNAV, a trainable reference framework comprising two key components: Spatiotemporally Anchored Collaborative Memory (SACM) and Aerial-Guided Regional-to-Local Planning (AGRLP). SACM maintains and retrieves spatially consistent historical context across aerial and ground observations. Meanwhile, AGRLP combines regional aerial guidance with fine-grained ground navigation. Extensive experiments demonstrate the effectiveness of AG-CoNAV and establish AirGroundVLN as a comprehensive benchmark for future exploration.
Chinese Translation
面向目标的视觉与语言导航（VLN）要求智能体在没有预定路线的情况下，定位并到达用自然语言描述的目标。空地协作在需要广域搜索和精细定位的任务中具有重要价值。然而，面向目标的空地协作VLN的系统研究仍然受到缺乏大规模、多样化基准和两个核心挑战的限制：1）空中视角和地面视角之间的显著差异，以及随着导航进行有用观察变得不可用，使得在不同平台和时间上保持空间一致的上下文变得困难；2）不对称的空间可观测性使得地面感知在局部上详细但在空间上有限，而空中感知则广泛但在局部上粗略，限制了单一平台规划的可靠性。为了解决这些限制，我们引入了AirGroundVLN，一个包含10,281个导航情节和955个目标实例的基准，涵盖19个虚幻引擎环境，具有已见/未见的划分和用于系统评估的空中可见性协议。除了基准之外，我们提出了AG-CoNAV，一个可训练的参考框架，包含两个关键组件：时空锚定协作记忆（SACM）和空中引导的区域到局部规划（AGRLP）。SACM在空中和地面观察之间维护和检索空间一致的历史上下文。同时，AGRLP将区域空中引导与精细的地面导航相结合。大量实验表明AG-CoNAV的有效性，并确立了AirGroundVLN作为未来探索的综合基准。
cs.RO / 83 / 2610.10453
RFPO: Rectified Flow Policy Optimization for Embodied Control
RFPO：用于具身控制的修正流策略优化
Abstract
Flow-based policies provide an expressive framework for continuous robot control, but their iterative ODE integration incurs substantial inference cost. Naively reducing the integration budget can severely degrade control, since policies optimized under full-step execution are not explicitly constrained to remain reliable under coarse numerical integration. We refer to this mismatch as the few-step discretization gap. To address this problem, we introduce RFPO, a flow-policy optimization framework for reliable few-step execution. Reward-aware online Reflow rectifies student-induced transport paths during on-policy learning, making the resulting policy more robust to coarse integration. A frozen Gaussian PPO controller supplies complementary action-space supervision at full and intermediate integration budgets, while the deployed policy remains a single flow student executed with one Euler step. Across Unitree Go2, Boston Dynamics Spot, Unitree H1, and Unitree G1, RFPO consistently preserves full-step control performance under one-step execution, with one-step returns remaining within 2.4% of their corresponding 64-step values across both zero and random initialization. On Unitree Go2, one-step execution retains 98.5% of the 64-step reward while reducing onboard mean inference latency from 4.39 ms to 0.08 ms, yielding a 54.9x speedup. Real-robot experiments further validate stable one-step locomotion. Code: https://github.com/AIGeeksGroup/RFPO. Website: https://aigeeksgroup.github.io/RFPO.
Chinese Translation
基于流的策略为连续机器人控制提供了一个表达性框架，但其迭代的常微分方程（ODE）积分会产生相当大的推理成本。简单地减少积分预算可能会严重降低控制性能，因为在全步执行下优化的策略并未明确约束在粗略数值积分下保持可靠性。我们将这种不匹配称为少步离散化差距。为了解决这个问题，我们引入了RFPO，一种用于可靠少步执行的流策略优化框架。奖励感知的在线重流（Reflow）在策略学习过程中修正学生引导的传输路径，使得生成的策略对粗略积分更加稳健。一个冻结的高斯PPO控制器在全积分和中间积分预算下提供补充的动作空间监督，而部署的策略仍然是一个通过一个欧拉步执行的单一流学生。在Unitree Go2、波士顿动力Spot、Unitree H1和Unitree G1上，RFPO在单步执行下始终保持全步控制性能，单步回报在零初始化和随机初始化下均保持在其对应的64步值的2.4%以内。在Unitree Go2上，单步执行保留了64步奖励的98.5%，同时将板载平均推理延迟从4.39毫秒减少到0.08毫秒，实现了54.9倍的加速。真实机器人实验进一步验证了稳定的单步运动。代码：https://github.com/AIGeeksGroup/RFPO。网站：https://aigeeksgroup.github.io/RFPO。
cs.RO / 84 / 2610.10462
FoldBack: Self-Correcting Masked Generative Policy for Long-Horizon Garment Folding
FoldBack：用于长时间服装折叠的自我修正掩蔽生成策略
Abstract
We present FoldBack, a self-correcting masked generative policy for long-horizon garment folding. Existing long-trajectory policies may continue after a missed or slipped grasp even when the garment has not reached the intended configuration. We structure FoldBack's recovery mechanisms around three inference-time decisions: when to refine and verify, how to roll back, and where and how to retry. FoldBack aligns refinement and grasp verification with pick-and-place events, returns the robot to a retryable pre-grasp configuration while preserving successful grasps, and selectively regenerates the failed segment and selected future actions while avoiding previous failed grasp locations. To our knowledge, FoldBack is the first editable full-trajectory policy to unify these decisions, enabling failed interactions to be detected, undone, and repaired before execution continues, without recovery demonstrations or base-policy retraining. Across 33 real garments from six categories, FoldBack achieves 75.2% final folding success and 0.837 final-mask IoU, versus 45.7% and 0.689 for the strongest prior baseline.
Chinese Translation
我们提出了FoldBack，一种用于长时间服装折叠的自我修正掩蔽生成策略。现有的长轨迹策略可能在抓取失败或滑落后继续执行，即使服装尚未达到预期的配置。我们围绕三种推理时决策构建FoldBack的恢复机制：何时进行细化和验证，如何回滚，以及在哪里和如何重试。FoldBack将细化和抓取验证与取放事件对齐，将机器人返回到可重试的预抓取配置，同时保留成功的抓取，并选择性地重新生成失败的部分和选定的未来动作，同时避免之前失败的抓取位置。根据我们的了解，FoldBack是第一个可编辑的全轨迹策略，统一了这些决策，使得在执行继续之前能够检测、撤销和修复失败的交互，而无需恢复演示或基础策略的重新训练。在来自六个类别的33件真实服装中，FoldBack实现了75.2%的最终折叠成功率和0.837的最终掩蔽IoU，而最强先前基线的成功率为45.7%，IoU为0.689。
cs.RO / 85 / 2610.10465
LLA-MPPI: Rapidly Adaptive Whole-body Control of Legged Robots with GPU-Accelerated Parallel Simulations
LLA-MPPI：基于GPU加速并行仿真的快速自适应全身控制的腿式机器人
Abstract
Real-time whole-body controllers for legged robots typically plan through a fixed nominal model and degrade when the deployed dynamics change. Adaptive methods typically require a model structure that contact dynamics do not provide, or they need offline training for each anticipated condition. We present Look-back and Look-ahead Adaptive Model Predictive Path Integral control (LLA-MPPI). The method converts whole-body adaptation into selection over a bank of GPU-batched contact simulators with different physical or structural parameters. Windowed prediction errors select the simulator that best explains recent motion. A whole-body MPPI planner optimizes controls through the selected model. The framework requires no offline training, and its selected hypotheses are physically interpretable. Across four simulated tasks, it achieves 97.5% success while the strongest baseline reaches 74% and an oracle with the true model reaches 98.5%. Hardware validation on a Unitree Go2 shows the robot walking under a payload added mid-run, walking after one leg is disabled, and pushing a box to its goal while increasing its mass on the fly. Code, videos, and project details are available at: https://lla-control.github.io
Chinese Translation
腿式机器人的实时全身控制器通常通过固定的名义模型进行规划，当实际动态变化时性能会下降。自适应方法通常需要接触动态无法提供的模型结构，或者需要针对每种预期条件进行离线训练。我们提出了回顾与前瞻自适应模型预测路径积分控制（LLA-MPPI）。该方法将全身适应转化为在具有不同物理或结构参数的GPU批处理接触模拟器库中进行选择。窗口化的预测误差选择最佳解释近期运动的模拟器。全身MPPI规划器通过所选模型优化控制。该框架不需要离线训练，其选择的假设在物理上可解释。在四个模拟任务中，它实现了97.5%的成功率，而最强的基线仅达到74%，使用真实模型的预言者则达到98.5%。在Unitree Go2上的硬件验证显示，机器人在运行过程中添加负载时能够行走，在一条腿失效后仍能行走，并在动态增加质量的情况下将箱子推向目标。代码、视频和项目详情可在以下链接获取：https://lla-control.github.io
cs.RO / 86 / 2610.10470
CMP-IRRT*: A Perception-Assisted Height-Adaptive Planner for Quadruped Robots
CMP-IRRT*: 一种感知辅助的高度自适应四足机器人规划器
Abstract
Quadruped robots can traverse low obstacles, but many 2D planning pipelines still model obstacles as binary occupied regions and rely on sampling-based search that can be inefficient under a limited budget. We propose a perception-assisted height-adaptive planning framework based on CMP-IRRT*, a Channel Mamba PointNet-guided Informed RRT* planner. Given a calibrated top-view RGB observation, the perception module estimates obstacle regions and converts depth predictions into a ground-relative height map. The planner then performs height-conditioned collision checking, treating high obstacles as blocked while allowing low obstacles to be traversed, and uses the CMP guide to bias sampling toward promising regions while retaining standard free-space and informed sampling fallbacks. Experiments on 2D planning benchmarks show that CMP-IRRT* reduces explored nodes and iterations compared with classical and neural-guided baselines, and a controlled ablation supports the contribution of the Mamba-based guide. In constructed traversability-aware scenarios, the proposed planner reduces path length by up to 16.3% when low obstacles are traversable, and a Unitree Go2 demonstration further shows executable bypassing and traversal behaviors. Our code is publicly available at https://github.com/MingfanZhao/height-adaptive-planner.
Chinese Translation
四足机器人能够跨越低障碍物，但许多二维规划管道仍将障碍物建模为二进制占用区域，并依赖基于采样的搜索，这在预算有限的情况下可能效率低下。我们提出了一种基于CMP-IRRT*的感知辅助高度自适应规划框架，CMP-IRRT*是一种由Channel Mamba PointNet引导的知情RRT*规划器。给定一个经过校准的俯视RGB观测，感知模块估计障碍物区域并将深度预测转换为相对于地面的高度图。然后，规划器执行高度条件的碰撞检测，将高障碍物视为被阻挡，同时允许低障碍物被穿越，并利用CMP引导偏向有前景的区域进行采样，同时保留标准的自由空间和知情采样的回退。对二维规划基准的实验表明，与经典和神经引导的基线相比，CMP-IRRT*减少了探索的节点和迭代次数，控制的消融实验支持了基于Mamba的引导的贡献。在构建的可通行性感知场景中，当低障碍物可穿越时，所提规划器将路径长度减少了最多16.3%，而Unitree Go2的演示进一步展示了可执行的绕行和穿越行为。我们的代码已公开发布在 https://github.com/MingfanZhao/height-adaptive-planner。
cs.RO / 87 / 2610.10472
Robotic Boomerang Throwing via Model-Based Release Design
基于模型的机器人回旋镖投掷设计
Abstract
Throwing objects that generate aerodynamic lift can greatly extend robot throwing beyond ballistic flight. A returning boomerang is a challenging example because its flight depends strongly on the release velocity, attitude, and spin, while robotic manipulators cannot readily reproduce the rapid motions used in human throwing. We present a model-based framework for robotic boomerang throwing centered on the release state. We identify the boomerang flight dynamics in stages to predict how flight changes across design variations. To systematically design the robot throwing motion, we screen candidate parameters according to how strongly and consistently they control release spin under uncertain contact conditions. These models are then used to design the throwing motion and boomerang for a 6-DoF manipulator with limited joint speeds. To our knowledge, this is the first robotic manipulator to generate a returning boomerang flight. In the demonstrated returning trial, the boomerang is released at 51 rad/s (8.1 rev/s), reaches 2.03 m from the robot base, and returns to touch down 0.31 m from the base. The successful release differs significantly from the measured human throws, showing that a robot need not imitate human throwing motion to achieve a returning flight. The project page is available at https://robot-boomerang.github.io
Chinese Translation
投掷能够产生空气动力学升力的物体可以大大扩展机器人投掷的飞行范围，超越弹道飞行。回旋镖是一个具有挑战性的例子，因为其飞行强烈依赖于释放速度、姿态和旋转，而机器人操控器无法轻易重现人类投掷时的快速动作。我们提出了一个以释放状态为中心的基于模型的机器人回旋镖投掷框架。我们将回旋镖的飞行动力学分阶段识别，以预测设计变体如何影响飞行变化。为了系统地设计机器人投掷动作，我们根据候选参数在不确定接触条件下对释放旋转的控制强度和一致性进行筛选。这些模型随后用于为具有有限关节速度的6自由度（6-DoF）操控器设计投掷动作和回旋镖。据我们所知，这是第一个能够产生回归飞行的机器人操控器。在演示的回归试验中，回旋镖以51 rad/s（8.1 rev/s）的速度释放，距离机器人基座2.03米，并在距离基座0.31米处成功着陆。成功的释放与测量的人类投掷有显著不同，表明机器人不必模仿人类的投掷动作即可实现回归飞行。项目页面可访问 https://robot-boomerang.github.io
cs.RO / 88 / 2610.10479
Agentic RSR: Real-to-Sim-to-Real through Scene Reconstruction and Execution-Grounded Robot Policies
Agentic RSR：通过场景重建和执行基础的机器人策略实现真实到模拟再到真实
Abstract
A simulation of a real robot workspace must preserve task-relevant interactions, while policies developed in it must operate on observations available to the real robot. Yet scene reconstruction and policy development are often treated separately. We present Agentic Real-to-Sim-to-Real (Agentic RSR), a framework that links scene reconstruction, policy development, and real-robot execution through the same manipulation task. Given a workspace video, a task description, and a known robot model, an agent recovers metric scale, iteratively refines the scene using visual feedback, and checks task-relevant interactions in MuJoCo. A coding agent then develops an executable policy, progressing from privileged object poses to visual observations and randomized simulation. The policy can interleave multiple observations and actions within one invocation, while the agent uses execution feedback to continue, retry, or revise its approach. A shared task-level interface carries the policy and accumulated experience to the real robot, where fresh observations and safety checks guide execution. Across 18 reconstructed scenes involving two robots, the mean four-view Depth MAE against reference depth estimates is 0.1057 m, the mean Lab $\Delta E_{76}$ is 11.04, and the mean grayscale SSIM is 0.6990. In real-robot experiments, the aggregate task success rate reaches 80% of the simulation task success rate, indicating substantial retention of simulated performance on hardware. Code and reconstructed scene data will be made publicly available.
Chinese Translation
真实机器人工作空间的模拟必须保留与任务相关的交互，而在其中开发的策略必须基于真实机器人可用的观察。然而，场景重建和策略开发通常被视为两个独立的过程。我们提出了Agentic Real-to-Sim-to-Real（Agentic RSR）框架，该框架通过相同的操作任务将场景重建、策略开发和真实机器人执行联系起来。给定一个工作空间视频、一个任务描述和一个已知的机器人模型，智能体恢复度量尺度，利用视觉反馈迭代地细化场景，并在MuJoCo中检查与任务相关的交互。然后，一个编码智能体开发可执行策略，从特权对象姿态逐步过渡到视觉观察和随机化模拟。该策略可以在一次调用中交替多个观察和动作，而智能体使用执行反馈来继续、重试或修正其方法。一个共享的任务级接口将策略和积累的经验传递给真实机器人，在那里新的观察和安全检查指导执行。在涉及两个机器人的18个重建场景中，与参考深度估计相比，平均四视图深度均方误差（Depth MAE）为0.1057米，平均Lab $ riangle E_{76}$为11.04，平均灰度结构相似性（SSIM）为0.6990。在真实机器人实验中，整体任务成功率达到了模拟任务成功率的80%，表明在硬件上显著保留了模拟性能。代码和重建场景数据将公开发布。
cs.RO / 89 / 2610.10489
HuMBLE: Human Motion-Driven Behavior Learning for Embodied Locomotion
HuMBLE：基于人类运动驱动的具身运动行为学习
Zhang, Mike, Kang, Dongho, Bergamin, Kevin, Burger, Nicola, Deits, Robin, Foster, Jonathan, Hammoud, Bilal, Hughes, Katie, Iacobelli, Francesco, Koolen, Twan, Mungai, M. Eva, Nobles, Zach, Rozen-Levy, Shane, Sleiman, Jean Pierre, Yu, Fangzhou, Zhang, Yunbo, Rizzi, Alfred, Hodgins, Jessica, Kuindersma, Scott, Abe, Yeuhi, Bertrand, Sylvain, Farshidian, Farbod
Abstract
Despite recent advances in humanoid locomotion, controllers optimized for command tracking and robustness tend to produce mechanical gaits, whereas controllers tied to human motion data often fail to generalize to commands outside the data distribution. This work introduces a learning framework that balances these competing objectives to synthesize real-time steerable, robust, and biomimetic locomotion policies from human data. Using an in-house curated locomotion dataset covering diverse speeds and directions, we first learn a natural locomotion prior policy through a teacher-student distillation process. Specifically, we train a full-body reference-conditioned policy with Reinforcement Learning (RL), then distill it into a lightweight prior policy conditioned solely on proprioception and a planar torso-velocity steering command. Next, we fine-tune the prior policy with multi-task RL to expand command coverage and robustness beyond the data distribution, pairing a goal-conditioned task that tracks arbitrary commands with a reference-guided task that tracks the human data as an explicit style regularizer. We validate our framework on three humanoid robots: the Boston Dynamics Atlas R1, Atlas D1, and Unitree G1. Experimental results demonstrate robust performance across real-world scenarios, including direct user-controlled locomotion in indoor and outdoor environments, and integration as the locomotion layer within hierarchical control stacks. Benchmarks against Tabula Rasa RL policies trained without human data and ablation studies confirm that our framework yields a lightweight, deployable policy that reconstructs coordinated whole-body behavior from a steering command, retaining the human gait characteristics while remaining robust and fully steerable.
Chinese Translation
尽管近年来在类人运动方面取得了进展，但针对指令跟踪和鲁棒性优化的控制器往往会产生机械化的步态，而与人类运动数据相关的控制器则常常无法推广到数据分布之外的指令。本文提出了一种学习框架，平衡这些相互竞争的目标，从人类数据中合成实时可调、鲁棒且仿生的运动策略。我们使用一个涵盖多种速度和方向的内部策划运动数据集，首先通过教师-学生蒸馏过程学习自然运动的先验策略。具体而言，我们使用强化学习（Reinforcement Learning, RL）训练一个全身参考条件的策略，然后将其蒸馏为一个仅基于本体感觉和一个平面躯干速度转向指令的轻量级先验策略。接下来，我们通过多任务强化学习对先验策略进行微调，以扩展指令覆盖范围和鲁棒性，超出数据分布，结合一个跟踪任意指令的目标条件任务与一个跟踪人类数据的参考引导任务作为显式风格正则化器。我们在三种类人机器人上验证了我们的框架：波士顿动力的Atlas R1、Atlas D1和Unitree G1。实验结果表明，在真实场景中表现出良好的鲁棒性，包括在室内和室外环境中直接用户控制的运动，以及作为分层控制堆栈中的运动层的集成。与未使用人类数据训练的Tabula Rasa RL策略的基准测试和消融研究确认，我们的框架产生了一种轻量级、可部署的策略，该策略能够从转向指令重构协调的全身行为，保留人类步态特征，同时保持鲁棒性和完全可调性。
cs.RO / 90 / 2610.10510
Factorized Tactile Representation and Control for Sim-to-Real Manipulation
因子化触觉表示与控制用于仿真到现实的操作
Abstract
Tactile sim-to-real learning must bridge simulated contact and device-specific sensor responses while preserving information needed for control. We propose a factorized tactile representation and control framework that maps normal force and contact patch to an effective contact response recoverable from sensor readings. The response is separated into contact geometry, force distribution, and temporal contact change, with representation-specific encoding and randomization. A Tactile Gated Policy preserves these representations separately through control and operates over all mask configurations without retraining. We evaluate the approach through response reconstruction, spatial alignment, force regulation, and contact-rich adversarial peg insertion in simulation and the real world, enabling the utility and transfer reliability of different tactile representations to be assessed independently. The approach achieves <1 mm contact localization, 1.69 N force-tracking error on unseen geometries, and a 35% improvement in real-world adversarial peg insertion over the unfactorized response, with different tactile representations benefiting different interactions.
Chinese Translation
触觉仿真到现实学习必须在模拟接触和设备特定传感器响应之间架起桥梁，同时保留控制所需的信息。我们提出了一种因子化触觉表示与控制框架，该框架将法向力和接触区域映射为可从传感器读数中恢复的有效接触响应。该响应被分离为接触几何、力分布和时间接触变化，并采用特定表示的编码和随机化。触觉门控策略（Tactile Gated Policy）通过控制分别保留这些表示，并在所有掩码配置上操作而无需重新训练。我们通过响应重建、空间对齐、力调节以及在仿真和现实世界中的接触丰富对抗性插销插入来评估该方法，使得不同触觉表示的效用和转移可靠性能够独立评估。该方法实现了<1毫米的接触定位，在未见几何体上的力跟踪误差为1.69牛顿，并在现实世界的对抗性插销插入中相比于未因子化响应提高了35%，不同的触觉表示在不同的交互中受益。
cs.RO / 91 / 2610.10526
Rephrase Before You Act: Characterizing and Mitigating Language Sensitivity in Vision-Language-Action Models
行动前重述：特征化和减轻视觉-语言-行动模型中的语言敏感性
Abstract
Vision-language-action models (VLAs) are strikingly sensitive to instruction phrasing and do not inherit the language robustness of the vision-language models they are built on. A one-word edit can move success by tens of points: $\pi_{0.5}$ turns on a LIBERO stove 100% of the time for "switch on the stove" and 2% for "switch on the hot plate", and a $\pi_0$ checkpoint finetuned with rephrase augmentation still shows swings of up to 61 points. We characterize this sensitivity with statistically tested single-edit swings and an oracle phrase search, which shows that phrasing alone nearly closes the 21-point gap between in-distribution and out-of-distribution tasks. We then reduce it without modifying the policy. Because the sensitivity is systematic, it can be expressed as explicit rules: we score many phrasings of a few training tasks, have a large language model distill the evidence into ten to twenty rephrasing rules, and at deployment rewrite each incoming instruction once under these rules. The rules improve the frozen $\pi_0$ by 16 to 27% relative on twelve held-out tasks across adversarial, VLM-generated, and human-generated phrasings, with gains concentrated on out-of-distribution tasks. The pipeline replicates on $\pi_{0.5}$ and LIBERO, lifting in-finetune success from 93.6% to 97.8%. The method requires no retraining and no per-step verification, and applies zero-shot to unseen tasks and instructions. Project website: https://sttawm.github.io/rephrase-before-you-act
Chinese Translation
视觉-语言-行动模型（VLA）对指令措辞表现出显著的敏感性，并未继承其基础视觉-语言模型的语言鲁棒性。单词的微小修改可以导致成功率的巨大变化：$ ext{π}_{0.5}$ 在“打开炉子”这一指令下的成功率为100%，而在“打开热板”下仅为2%；经过重述增强微调的$ ext{π}_0$检查点仍显示出高达61点的波动。我们通过统计测试的单编辑波动和一个oracle短语搜索来特征化这种敏感性，结果显示，仅凭措辞几乎可以弥补21点的分数差距，涵盖了分布内和分布外任务。随后，我们在不修改策略的情况下减少了这种敏感性。由于这种敏感性是系统性的，因此可以用明确的规则来表达：我们对少数训练任务的多种措辞进行评分，让一个大型语言模型将证据提炼为十到二十条重述规则，并在部署时根据这些规则对每个传入指令进行一次重写。这些规则使得冻结的$ ext{π}_0$在十二个保留任务上相对提高了16%到27%，这些任务包括对抗性、VLM生成和人类生成的措辞，收益主要集中在分布外任务上。该流程在$ ext{π}_{0.5}$和LIBERO上得以复制，将微调成功率从93.6%提升至97.8%。该方法无需重新训练，也不需要逐步验证，并且可以零样本应用于未见过的任务和指令。项目网站：https://sttawm.github.io/rephrase-before-you-act
cs.RO / 92 / 2610.10528
Long-WAM: Scaling the Context of World-Action Models
Long-WAM：扩展世界-动作模型的上下文
Abstract
Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long-WAM, a model-system framework for scaling the context of causal world-action models under real-time control constraints. Our central finding is that access to history is not the same as using it: longer histories pay off far more when the video foundation is pretrained autoregressively (AR). We first learn causal prediction from robot and egocentric videos without action labels, then preserve this history-to-future structure during world-action adaptation. On RoboCasa GR-1, increasing context from 0.0 to 19.2 seconds raises success from 63.3% to 78.7%, whereas a bidirectionally pretrained initialization shows no net gain; robot-domain AR pretraining further raises peak success on GR-1 and LIBERO-Long. Long-WAM also achieves the best results among compared methods on LIBERO-Long, RoboTwin 2.0, and DOMINO. Streaming observation encoding, asynchronous execution, and hardware-specific acceleration enable deployment on RTX 5090, DGX Spark, and Jetson AGX Thor without dropping future prediction; on RTX 5090, each action chunk, including future-video latent prediction, takes 107.4 ms. Real-time deployment on Unitree G1 and YAM supports dynamic and long-horizon manipulation, including 95% success on dynamic cup stacking, where Pi0.5 and Fast-WAM succeed in none of 20 trials. As a memory-informed executor, Long-WAM also complements higher-level planning in composite tasks.
Chinese Translation
实时机器人控制需要足够的视觉历史来推断运动和任务进展，但处理这些历史可能会延迟行动。我们提出了Long-WAM，这是一种在实时控制约束下扩展因果世界-动作模型上下文的模型系统框架。我们的核心发现是，访问历史并不等同于使用历史：当视频基础通过自回归（AR）预训练时，更长的历史带来的收益更大。我们首先从没有动作标签的机器人和自我中心视频中学习因果预测，然后在世界-动作适应过程中保持这种历史到未来的结构。在RoboCasa GR-1上，将上下文从0.0秒增加到19.2秒使成功率从63.3%提高到78.7%，而双向预训练的初始化则没有净增益；机器人领域的AR预训练进一步提高了GR-1和LIBERO-Long上的峰值成功率。Long-WAM在LIBERO-Long、RoboTwin 2.0和DOMINO等比较方法中也取得了最佳结果。流式观察编码、异步执行和硬件特定加速使其能够在RTX 5090、DGX Spark和Jetson AGX Thor上部署，而不会降低未来预测的准确性；在RTX 5090上，每个动作块（包括未来视频潜在预测）耗时107.4毫秒。Unitree G1和YAM上的实时部署支持动态和长时间范围的操作，包括在动态杯叠加中取得95%的成功率，而Pi0.5和Fast-WAM在20次试验中均未成功。作为一个受记忆启发的执行器，Long-WAM还补充了复合任务中的高层规划。
cs.RO / 93 / 2610.10534
RoboPrompt: Intuitive Robot Policy Steering with Sparse Human Input
RoboPrompt：通过稀疏人类输入进行直观的机器人策略引导
Abstract
End-to-end robot policies trained through imitation learning remain constrained by limited data diversity, making reliable zero-shot deployment in real-world settings challenging. Shared-autonomy methods enable human correction through teleoperation, but specialized hardware and operator training hinder deployment at scale. Other approaches incorporate human guidance as additional policy inputs, often requiring architectural changes and dedicated training for steerability, which limits their applicability across policies. We present RoboPrompt, a general-purpose, lightweight robot policy steering system that enables users to guide policy behavior through intuitive, sparse inputs, including drawn traces, target points, and coarse directional instructions. RoboPrompt decouples human-intention translation from the underlying policy: a reusable module converts human guidance into action drafts, which are refined through the diffusion or flow-matching dynamics of the base policy. By controlling action generation in noise space, RoboPrompt balances human intent with the policy prior without modifying the base policy architecture or fine-tuning it for steerability. Experiments demonstrate effective steering across Diffusion Policy, $\pi_{0.5}$, and FastWAM. We further use steered rollouts for online policy improvement through DAgger. After 2-3 rounds of iteration, average success rates increase by 15.5\% for $\pi_{0.5}$ across three tasks and by 21.3\% across three policies(Diffusion Policy, $\pi_{0.5}$, FastWAM) on the Insert Bread task, while average human intervention counts decrease by 44.0\% (2.86 to 1.60) and 81.9\% (2.60 to 0.47), respectively.
Chinese Translation
通过模仿学习训练的端到端机器人策略受到数据多样性有限的制约，使得在现实环境中可靠的零-shot 部署变得具有挑战性。共享自主方法通过遥控实现人类纠正，但专用硬件和操作员培训限制了大规模部署。其他方法将人类指导作为额外的策略输入，通常需要架构更改和专门训练以实现可引导性，这限制了它们在不同策略中的适用性。我们提出了 RoboPrompt，一种通用的轻量级机器人策略引导系统，使用户能够通过直观的稀疏输入（包括绘制轨迹、目标点和粗略方向指令）来引导策略行为。RoboPrompt 将人类意图的转换与基础策略解耦：一个可重用的模块将人类指导转换为行动草稿，这些草稿通过基础策略的扩散或流动匹配动态进行优化。通过在噪声空间中控制行动生成，RoboPrompt 在不修改基础策略架构或进行可引导性微调的情况下，平衡了人类意图与策略先验。实验表明，在扩散策略（Diffusion Policy）、$ ext{π}_{0.5}$ 和 FastWAM 上实现了有效的引导。我们进一步利用引导的回滚（steered rollouts）通过 DAgger 进行在线策略改进。在经过 2-3 轮迭代后，$ ext{π}_{0.5}$ 在三个任务上的平均成功率提高了 15.5\%，而在 Insert Bread 任务上，三个策略（扩散策略、$ ext{π}_{0.5}$、FastWAM）的成功率提高了 21.3\%，同时平均人类干预次数分别减少了 44.0\%（从 2.86 降至 1.60）和 81.9\\%（从 2.60 降至 0.47）。
cs.CV / 1 / 2610.08813
Pre-training, Reasoning, Benchmarking: X-ray Report Generation on CheXpert Plus Dataset
预训练、推理、基准测试：基于CheXpert Plus数据集的X光报告生成
Abstract
X-ray image-based Radiology Report Generation (RRG) constitutes a critical research direction within medical artificial intelligence, with great potential to alleviate clinicians' diagnostic workload and shorten patient waiting periods. Despite substantial advances over recent years, the field faces evident bottlenecks stemming from insufficient standardized benchmarks and inadequate domain adaptation of generic large models. Notably, the newly released CheXpert Plus dataset is provided without accompanying baseline implementations and evaluation results, which impedes standardized training, quantitative evaluation and fair comparison among follow-up algorithms. To mitigate this limitation, we establish a comprehensive benchmark encompassing prevailing X-ray report generation models and Large Language Models on CheXpert Plus. This benchmark delivers a reliable comparative foundation for upcoming methods and enables researchers to rapidly identify state-of-the-art approaches within this domain. Beyond benchmark construction, we rethink X-ray RRG under the paradigm of large models and propose a novel framework termed MambaXray-PRB. Our framework improves report generation performance and enhances model interpretability via multi-stage large-model pre-training and multi-modal Chain-of-Thought reasoning. The pipeline consists of three successive phases: self-supervised auto-regressive modeling, X-ray-report contrastive learning, and post-training optimization for reasoning and report generation. Extensive experiments on IU X-ray, MIMIC-CXR, and CheXpert Plus datasets validate the effectiveness of MambaXray-PRB for radiology report generation. The source code of this paper is available on https://github.com/Event-AHU/Medical_Image_Analysis
Chinese Translation
基于X光图像的放射学报告生成（RRG）是医学人工智能领域中的一个重要研究方向，具有减轻临床医生诊断工作负担和缩短患者等待时间的巨大潜力。尽管近年来取得了显著进展，该领域仍面临明显的瓶颈，主要源于标准化基准不足和通用大模型的领域适应性不足。值得注意的是，新发布的CheXpert Plus数据集未提供相应的基线实现和评估结果，这妨碍了标准化训练、定量评估以及后续算法之间的公平比较。为了解决这一限制，我们建立了一个全面的基准，涵盖了当前在CheXpert Plus上流行的X光报告生成模型和大型语言模型。该基准为未来的方法提供了可靠的比较基础，并使研究人员能够快速识别该领域的最先进方法。除了基准构建，我们还在大型模型的范式下重新思考X光RRG，并提出了一种新颖的框架，称为MambaXray-PRB。我们的框架通过多阶段的大模型预训练和多模态的链式思维推理，提高了报告生成性能并增强了模型的可解释性。该流程包括三个连续阶段：自监督自回归建模、X光报告对比学习，以及推理和报告生成的后训练优化。在IU X-ray、MIMIC-CXR和CheXpert Plus数据集上的大量实验验证了MambaXray-PRB在放射学报告生成中的有效性。本文的源代码可在https://github.com/Event-AHU/Medical_Image_Analysis获取。
cs.CV / 2 / 2610.08825
Autonomous Driving Research Requires a Community-Driven Data Paradigm
自主驾驶研究需要以社区为驱动的数据范式
Abstract
Autonomous driving has made remarkable progress, with recent AI advances enabling commercial deployments that are reshaping urban mobility. Yet the field remains far from its universal social promise: autonomous systems that can operate robustly anywhere, anytime, for anyone. We posit that this gap is not merely a modeling problem, but a problem of the prevailing data paradigm. Current research relies heavily on a few benchmark datasets with limited spatial and scenario coverage, even though the community has collectively produced over 600 autonomous driving datasets across nearly 50 countries. However, this abundance has not translated into broad research impact: most datasets remain significantly underused due to fragmentation, limited visibility, incompatible protocols, and benchmark incentives that concentrate attention on a few dominant datasets. We therefore argue that autonomous driving research requires a collaborative, community-driven data paradigm. Such a paradigm would improve the discovery, reuse, integration, and evaluation of diverse datasets; make underexplored data easier and more rewarding to study; and lower the barrier for new contributors. We outline its key principles, illustrate an early realization, and call for collaboration across academia and industry to transform fragmented datasets into shared community infrastructure for anytime-anywhere autonomy.
Chinese Translation
自主驾驶取得了显著进展，近期的人工智能技术进步使得商业部署成为可能，正在重塑城市出行。然而，该领域距离其普遍的社会承诺仍然相去甚远：自主系统能够在任何时间、任何地点、为任何人稳健运行。我们认为，这一差距不仅仅是建模问题，而是当前数据范式的问题。现有研究在很大程度上依赖于少数具有有限空间和场景覆盖的基准数据集，尽管社区已经共同生成了超过600个自主驾驶数据集，覆盖近50个国家。然而，这种丰富性并没有转化为广泛的研究影响：大多数数据集由于碎片化、可见性有限、不兼容的协议以及集中关注少数主导数据集的基准激励而显著未被利用。因此，我们认为自主驾驶研究需要一个协作的、以社区为驱动的数据范式。这样的范式将改善各种数据集的发现、重用、整合和评估；使得未被充分探索的数据更易于研究并更具回报；并降低新贡献者的门槛。我们概述了其关键原则，展示了早期实现的实例，并呼吁学术界和工业界的合作，将碎片化的数据集转变为共享的社区基础设施，以实现随时随地的自主性。
cs.CV / 3 / 2610.08826
PanoPed: Beyond Bounding Boxes for Sim-to-Real Panoramic Pedestrian Tracking
PanoPed：超越边界框的全景行人跟踪从仿真到现实
Abstract
Full-sphere panoramic cameras let fixed monitoring systems and mobile robots track people in every direction, but a planar bounding box does not fully describe where a person is on the sphere. We introduce PanoPed, a sim-to-real benchmark for pedestrian tracking on the full sphere. PanoPed-S contains 108,000 frames from fixed, quadruped-mounted, and drone-mounted cameras, with synchronized masks, depth, camera poses, and 3D pedestrian states. PanoPed-R adds 28,002 real frames from fixed cameras, 16,247 of them densely annotated. We find that an ERP rectangle cannot uniquely determine the spherical center and angular extent of the visible person, while the detector's visual query still carries information about them. Inspired by the sextant's use of angular measurements to locate objects, we propose Sextant, a plug-and-play angular localization head with only about 0.035M parameters. It reuses a frozen detector, keeps track identities unchanged, and needs no extra image encoder. Sextant gives the best result in our PanoPed-S test comparison, raising the strongest baseline, MOTIP, from 47.30 to 49.49 HOTA, with gains on all eight test sequences. Without fine-tuning on real data, the same synthetic-trained heads improve MOTIP and HAT by 0.96-1.14 HOTA on real video, and both seeds improve every real sequence. HAT+Sextant scores best among the compared systems that add no localization image encoder.
Chinese Translation
全球面全景相机使固定监控系统和移动机器人能够在各个方向上跟踪行人，但平面边界框并不能完全描述一个人在球面上的位置。我们提出了PanoPed，这是一个针对全球面行人跟踪的从仿真到现实的基准。PanoPed-S包含来自固定、四足动物装载和无人机装载相机的108,000帧图像，配有同步的掩膜、深度信息、相机姿态和3D行人状态。PanoPed-R则增加了来自固定相机的28,002帧真实图像，其中16,247帧经过密集标注。我们发现，ERP矩形无法唯一确定可见行人的球心和角度范围，而检测器的视觉查询仍然携带关于他们的信息。受到六分仪使用角度测量定位物体的启发，我们提出了Sextant，这是一种即插即用的角度定位头，参数量仅约0.035M。它重用一个冻结的检测器，保持跟踪身份不变，并且不需要额外的图像编码器。在我们的PanoPed-S测试比较中，Sextant取得了最佳结果，将最强基线MOTIP从47.30提升至49.49 HOTA，并在所有八个测试序列中均有提升。在未对真实数据进行微调的情况下，相同的合成训练头在真实视频上将MOTIP和HAT分别提升了0.96-1.14 HOTA，并且两个种子在每个真实序列上均有提升。在比较的系统中，HAT+Sextant在未添加定位图像编码器的情况下获得最佳成绩。
cs.CV / 4 / 2610.08830
MoR-MLLM: Mixture of Recursions for Efficient Multimodal Large Language Models
MoR-MLLM：高效多模态大语言模型的递归混合
Abstract
Multimodal Large Language Models (MLLMs) have demonstrated remarkable reasoning capabilities across vision and language tasks. However, their massive computational and memory demands hinder real-world deployment. While recent efforts reduce costs by employing lightweight language backbones, existing paradigms remain computation-dense due to their static sparsity and depth allocation, which cannot adapt to the semantic complexity of each token. To this end, we propose MoR-MLLM, a computation-sparse MLLM based on the recent Mixture-of-Recursions (MoR) framework. MoR-MLLM introduces adaptive per-token recursion, allowing the model to dynamically adjust its recursive depth and allocate more computation to visually or linguistically challenging tokens while skipping redundant operations for simpler ones. To stabilize the training of recursive sparsity in multimodal settings, we further design a three-stage MoR-Tuning strategy and an entropy-regularized loss to encourage diverse routing distributions. Extensive experiments show that compared with recent advanced tiny MLLMs, our proposed MoR-MLLM can greatly reduce the training memory and computation complexity while retaining high performance on various vision-language tasks.
Chinese Translation
多模态大语言模型（MLLMs）在视觉和语言任务中展现了卓越的推理能力。然而，它们巨大的计算和内存需求阻碍了在实际应用中的部署。尽管近期的研究通过采用轻量级语言骨干网络来降低成本，但现有范式由于其静态稀疏性和深度分配，仍然计算密集，无法适应每个标记的语义复杂性。为此，我们提出了MoR-MLLM，这是一种基于最近的递归混合（Mixture-of-Recursions, MoR）框架的计算稀疏MLLM。MoR-MLLM引入了自适应的每标记递归，使模型能够动态调整其递归深度，并为视觉或语言上具有挑战性的标记分配更多计算，同时跳过对简单标记的冗余操作。为了稳定多模态环境中递归稀疏性的训练，我们进一步设计了三阶段的MoR调优策略和熵正则化损失，以鼓励多样化的路由分布。大量实验表明，与近期先进的小型MLLM相比，我们提出的MoR-MLLM在保持高性能的同时，能够大幅降低训练内存和计算复杂度，适用于各种视觉-语言任务。
cs.CV / 5 / 2610.08936
VCR-Bench: A Modular Open-Source Benchmark for Video Classification Robustness
VCR-Bench：一个模块化的开源视频分类鲁棒性基准
Abstract
Robustness of image classification has several benchmarks, but their video counterparts are absent. In video classification temporal dimension introduces additional degrees of freedom for adversarial attacks, defenses, and preprocessing. Temporal sampling, perturbation budgets, and metric aggregation also interact in ways with no direct analogue in the image setting. Therefore, robustness for video classifiers is studied across scattered, incompatible implementations, making reported numbers hard to reproduce and analyze. We introduce VCR-Bench, a modular open-source benchmark framework that standardizes video loading, wrappers for classifiers, adversarial attacks and defenses, perceptual metrics, configuration presets, and result logging. VCR-Bench currently integrates 30 video classification models, 14 adversarial attacks, and 10 defense wrappers under a common evaluation protocol. We evaluate representative video classifiers, attacks, and defenses on Kinetics-400 subset, reporting clean accuracy, attack success rate, perceptual quality, runtime, and memory usage. VCR-Bench is released with documented installation, reproducible run presets, component-extension interfaces, and scripts for reproducing the reported results at https://github.com/msu-video-group/vcr-bench.
Chinese Translation
图像分类的鲁棒性有多个基准，但其视频分类的对应基准却缺失。在视频分类中，时间维度为对抗攻击、防御和预处理引入了额外的自由度。时间采样、扰动预算和度量聚合之间的相互作用在图像设置中没有直接的类比。因此，视频分类器的鲁棒性在分散且不兼容的实现中进行研究，使得报告的数值难以重现和分析。我们提出了VCR-Bench，一个模块化的开源基准框架，标准化了视频加载、分类器的封装、对抗攻击和防御、感知度量、配置预设和结果记录。VCR-Bench目前整合了30个视频分类模型、14种对抗攻击和10个防御封装，遵循统一的评估协议。我们在Kinetics-400子集上评估了具有代表性的视频分类器、攻击和防御，报告了干净准确率、攻击成功率、感知质量、运行时间和内存使用情况。VCR-Bench已发布，附带文档安装、可重现的运行预设、组件扩展接口以及用于重现报告结果的脚本，网址为https://github.com/msu-video-group/vcr-bench。
cs.CV / 6 / 2610.08941
SPW-Nav: A Streaming Panoramic World Model for Language-Guided Navigation
SPW-Nav：一种用于语言引导导航的流式全景世界模型
Abstract
Language-guided panoramic video generation benefits various downstream applications, such as interactive 3D scene exploration, virtual reality experiences, and embodied agent training. Existing panoramic generators follow predefined trajectories, and interactive world models act through low-level actions in perspective views. We propose SPW-Nav, a streaming panoramic world model that understands movement instructions and streams one minute of 2K 360-degree video in real time from a single panorama. SPW-Nav interprets each instruction in the previously generated panorama as camera motion. Spherical rotation decoupling applies rotation exactly on the sphere, pose-aligned conditioning keeps translation inputs bounded over long streams, and a multi-term memory with a few-step generator continues the scene as instructions change. We also build SPW-NavSet, panoramic videos with camera trajectories and verified instructions. Driven by language, SPW-Nav outperforms prior panoramic generators in camera-following accuracy and video quality, and supports on-the-fly instruction switching.
Chinese Translation
语言引导的全景视频生成对各种下游应用具有重要意义，例如交互式3D场景探索、虚拟现实体验和具身代理训练。现有的全景生成器遵循预定义的轨迹，而交互式世界模型则通过透视视图中的低级动作进行操作。我们提出了SPW-Nav，一种流式全景世界模型，能够理解运动指令，并实时从单个全景中流式生成一分钟的2K 360度视频。SPW-Nav将之前生成的全景中的每个指令解释为相机运动。球面旋转解耦在球面上精确应用旋转，姿态对齐条件使得平移输入在长时间流中保持界限，而多项记忆结合几步生成器在指令变化时持续场景。我们还构建了SPW-NavSet，这是带有相机轨迹和验证指令的全景视频。在语言驱动下，SPW-Nav在相机跟随精度和视频质量方面优于之前的全景生成器，并支持即时指令切换。
cs.CV / 7 / 2610.08944
One-Slide Calibration of Pathology Foundation Models
病理基础模型的一次性校准
Abstract
Scanner variation changes how pathology foundation models represent the same tissue. We introduce SlideRuler, which uses regions within a slide as internal controls to estimate and correct acquisition-induced shifts in other regions. A transfer map learned from paired rescans enables calibration from a single scan at inference while keeping the foundation model fixed. Across two encoders and five SCORPION scanners, learned transfer reduces mean target-to-source embedding distance by 16.3-38.5% relative to raw embeddings. Comparisons with unrelated same-scanner controls reveal a positive same-slide contribution across all four evaluation settings, including scanner holdout. A source-anchored variant reduces source-feature displacement by 47.7-83.6% relative to learned transfer while retaining most of its alignment gain. By drawing calibration information from the slide itself, SlideRuler offers a path toward more consistent use of frozen pathology models across imaging systems.
Chinese Translation
扫描仪的变化影响病理基础模型对相同组织的表征。我们提出了SlideRuler，它利用切片内的区域作为内部控制，以估计和校正其他区域因采集引起的偏移。通过配对重扫描学习的转移映射使得在推理时可以仅通过单次扫描进行校准，同时保持基础模型不变。在两个编码器和五个SCORPION扫描仪的实验中，学习到的转移相较于原始嵌入减少了16.3%-38.5%的目标到源嵌入距离。与不相关的同扫描仪控制组的比较显示，在所有四个评估设置中，包括扫描仪保留，均存在积极的同切片贡献。一个源锚定变体相较于学习到的转移减少了47.7%-83.6%的源特征位移，同时保留了大部分对齐增益。通过从切片本身提取校准信息，SlideRuler为在不同成像系统中更一致地使用冻结的病理模型提供了一条途径。
cs.CV / 8 / 2610.08954
RACER: Reflective Agent Coupling Query Interpretation and Tool-Based Retrieval for Frame Selection in Long Video Understanding
RACER：反思性代理耦合查询解释与基于工具的检索用于长视频理解中的帧选择
Abstract
Video large language models (Vid-LLMs) excel at diverse video-language tasks by reasoning over selected frames. However, frame selection for long videos remains challenging, as it requires retrieving relevant frames distributed across segments from a large candidate pool given complex queries. This paper investigates dominant approaches to long-video frame selection from a task-decomposition perspective, identifying two key challenges: the Query Comprehension Gap in similarity-based methods and the Interpretation--Selection Gap in judgment-based methods. To address them, we propose RACER, a training-free reflective agentic framework that decomposes long-video frame selection into query interpretation driven by a lightweight Vid-LLM and evidence localization supported by an embedding model serving as a retrieval tool. Specifically, the Vid-LLM is responsible solely for reformulating the complex query into sub-queries that make implicit information requirements explicit, mitigating the Query Comprehension Gap. Meanwhile, the retrieval tool leverages these sub-queries to localize relevant evidence, relieving the Vid-LLM of direct frame selection and thus addressing the Interpretation--Selection Gap. Finally, the retrieved frames are fed back to the Vid-LLM for sub-query refinement, forming a reflection loop that iteratively improves query interpretation and frame selection. Experiments across multiple benchmarks show that RACER consistently improves long video understanding. Notably, RACER achieves effective frame selection even with limited-capability components, demonstrating that agentic integration enables these components to enhance more capable Vid-LLMs.
Chinese Translation
视频大型语言模型（Vid-LLMs）通过对选定帧的推理，在多种视频语言任务中表现出色。然而，对于长视频的帧选择仍然具有挑战性，因为这需要从一个大型候选池中检索分布在不同片段中的相关帧，以应对复杂的查询。本文从任务分解的角度探讨了长视频帧选择的主要方法，识别出两个关键挑战：基于相似性的查询理解差距和基于判断的解释-选择差距。为了解决这些问题，我们提出了RACER，一个无训练的反思性代理框架，将长视频帧选择分解为由轻量级Vid-LLM驱动的查询解释和由嵌入模型支持的证据定位，后者作为检索工具。具体而言，Vid-LLM仅负责将复杂查询重新表述为使隐含信息需求显性化的子查询，从而缓解查询理解差距。同时，检索工具利用这些子查询来定位相关证据，减轻Vid-LLM的直接帧选择负担，从而解决解释-选择差距。最后，检索到的帧被反馈给Vid-LLM以进行子查询的细化，形成一个反思循环，迭代改善查询解释和帧选择。在多个基准测试中的实验表明，RACER在长视频理解方面持续取得改善。值得注意的是，RACER即使在能力有限的组件下也能实现有效的帧选择，证明了代理集成使这些组件能够增强更强大的Vid-LLM。
cs.CV / 9 / 2610.08978
S2Tok: Streaming 3D Gaussian Reconstruction with Persistent Spatial Tokens
S2Tok：具有持久空间标记的流式三维高斯重建
Abstract
Streaming 3D reconstruction requires more than a sequence of geometric predictions: it requires a persistent scene state that can incorporate new evidence and remain renderable as observations arrive. Latent spatial tokens offer a promising representation for this purpose, but constructing them from an image collection leaves open how to maintain them online, where each observation may both revisit known regions and reveal new content. We introduce S2Tok, a feed-forward framework that maintains a size-adaptive, persistent scene state from uncalibrated image streams. Its central idea is to distinguish updates to the existing representation from selective expansion. A spatially informed transformer integrates each incoming observation with the persistent scene tokens, while a learned admission module selectively expands the representation to limit redundant storage. A hierarchical decoder and Gaussian head convert the evolving state into non-pixel-aligned 3D Gaussians, enabling novel-view rendering without caching previous frames. Experiments across four benchmarks demonstrate competitive streaming rendering quality with compact Gaussian representations. These results support latent spatial tokens as a persistent computational state for online 3D reconstruction, combining learned scene updates with explicit Gaussian rendering.
Chinese Translation
流式三维重建不仅需要一系列几何预测，还需要一个能够整合新证据并在观察到达时保持可渲染的持久场景状态。潜在空间标记为此提供了一种有前景的表示方式，但如何从图像集合中构建它们，并在在线环境中维护它们仍然是一个未解的问题，因为每个观察可能既会重访已知区域，又会揭示新内容。我们提出了S2Tok，这是一种前馈框架，能够从未校准的图像流中维护一个大小自适应的持久场景状态。其核心思想是将对现有表示的更新与选择性扩展区分开来。一个空间信息驱动的变换器将每个输入观察与持久场景标记相结合，而一个学习的接纳模块则选择性地扩展表示，以限制冗余存储。一个层次解码器和高斯头将不断演变的状态转换为非像素对齐的三维高斯，从而实现新视角渲染而无需缓存先前的帧。在四个基准测试中的实验表明，使用紧凑的高斯表示实现了具有竞争力的流式渲染质量。这些结果支持潜在空间标记作为在线三维重建的持久计算状态，将学习的场景更新与显式高斯渲染相结合。
cs.CV / 10 / 2610.08983
Zero-Shot Brain MRI Inpainting with 2.5D Unconditional Flow Priors
基于2.5D无条件流先验的零-shot脑MRI修复
Abstract
Generative inpainting of brain MRI volumes is essential for synthesizing healthy tissue in pathological regions, improving the accuracy and reliability of automated downstream brain analysis applications such as image registration, brain extraction, and segmentation. However, standard 3D approaches are computationally prohibitive, while efficient 2D slice-wise methods suffer from severe inter-slice discontinuities. Furthermore, traditional models rely on conditional training, requiring task-specific learning of masked inputs. We propose a zero-shot brain MRI inpainting framework utilizing 2.5D unconditional flow priors to capture spatial context along the superior-inferior axis without the overhead of full 3D convolutions. During training, our flow matching model learns the joint distribution of adjacent axial slice triplets, modeling the manifold of healthy brain anatomy while explicitly excluding pathological regions from the loss function. At inference, the model processes the input triplets autoregressively along the depth axis. We employ the Restora-Flow solver to constrain the unconditional prior using the input mask, achieving accurate zero-shot inpainting. Evaluations show our 2.5D strategy resolves the structural discontinuities of 2D baselines, synthesizing plausible healthy tissue while maintaining volumetric consistency across the axial, sagittal, and coronal planes. As a final step, we generate and average an ensemble of multiple stochastic reconstructions to form the final prediction. Quantitative results benchmarked on the official BraTS 2026 Inpainting Challenge validation set demonstrate the effectiveness of our proposed approach, yielding an SSIM of 0.816 $\pm$ 0.112, MSE of 0.007 $\pm$ 0.005, and PSNR of 22.923 $\pm$ 4.343. Code is available at https://github.com/imigraz/brats2026-inpainting.
Chinese Translation
脑MRI体积的生成修复对于在病理区域合成健康组织至关重要，这有助于提高自动化下游脑分析应用（如图像配准、脑提取和分割）的准确性和可靠性。然而，标准的3D方法计算开销巨大，而高效的2D切片方法则存在严重的切片间不连续性。此外，传统模型依赖于条件训练，需要针对特定任务学习被遮蔽的输入。我们提出了一种零-shot脑MRI修复框架，利用2.5D无条件流先验来捕捉沿上-下轴的空间上下文，而无需全3D卷积的开销。在训练过程中，我们的流匹配模型学习相邻轴向切片三元组的联合分布，建模健康脑解剖结构的流形，同时明确从损失函数中排除病理区域。在推理阶段，该模型沿深度轴自回归处理输入三元组。我们使用Restora-Flow求解器通过输入掩码约束无条件先验，实现准确的零-shot修复。评估结果表明，我们的2.5D策略解决了2D基线的结构不连续性，合成了合理的健康组织，同时在轴向、矢状面和冠状面保持体积一致性。最后，我们生成并平均多个随机重建的集成以形成最终预测。在官方BraTS 2026修复挑战验证集上进行的定量结果基准测试证明了我们提出的方法的有效性，SSIM为0.816 ± 0.112，MSE为0.007 ± 0.005，PSNR为22.923 ± 4.343。代码可在https://github.com/imigraz/brats2026-inpainting获取。
cs.CV / 11 / 2610.09015
Personalize at Test Time: Learning User Preferences for Image Generation
测试时个性化：学习用户偏好以进行图像生成
Abstract
Diffusion models can generate high-quality images, yet aligning their outputs with individual user preferences remains challenging. A key bottleneck is accurately modeling diverse user preferences from limited feedback. Existing approaches often rely on labor-intensive manual preference annotations or vision-language models (VLM) to extract preference information from user interaction histories, introducing substantial annotation or computational costs that limit scalability. We propose an approach that learns personalized reward models directly from users' historical image preference pairs. First, we use an autoencoder to compress hundreds of visual attributes into 50 attribute-anchored preference dimensions and train an evaluator to score images along these dimensions. We then represent each user's preferences as a linear combination of the shared dimension scores, estimating the user-specific weights by maximizing the likelihood of their observed pairwise preferences under the Bradley-Terry model. This formulation reduces per-user adaptation to optimizing a low-dimensional weight vector, simplifying optimization and enabling data-efficient personalization from sparse feedback. The learned personalized rewards guide image generation at inference time while keeping the diffusion model frozen. Experiments on real-user preference data show that our approach achieves approximately 77% held-out pairwise preference prediction accuracy and improves the alignment of generated images with individual user preferences.
Chinese Translation
扩散模型能够生成高质量的图像，但将其输出与个别用户偏好对齐仍然具有挑战性。一个关键瓶颈是从有限的反馈中准确建模多样化的用户偏好。现有的方法通常依赖于劳动密集型的手动偏好标注或视觉-语言模型（VLM）从用户交互历史中提取偏好信息，这引入了可观的标注或计算成本，限制了可扩展性。我们提出了一种直接从用户的历史图像偏好对中学习个性化奖励模型的方法。首先，我们使用自编码器将数百个视觉属性压缩为50个属性锚定的偏好维度，并训练一个评估器根据这些维度对图像进行评分。然后，我们将每个用户的偏好表示为共享维度评分的线性组合，通过最大化布拉德利-特里模型下观察到的成对偏好的似然来估计用户特定的权重。这种公式化将每个用户的适应性简化为优化一个低维权重向量，简化了优化过程，并使得从稀疏反馈中实现数据高效的个性化成为可能。学习到的个性化奖励在推理时指导图像生成，同时保持扩散模型不变。对真实用户偏好数据的实验表明，我们的方法实现了约77%的保留成对偏好预测准确率，并改善了生成图像与个别用户偏好的对齐。
cs.CV / 12 / 2610.09031
Beyond Explanation: Debugging Medical Imaging Models via Concept Intervention
超越解释：通过概念干预调试医学影像模型
Abstract
Medical imaging models often operate as black boxes, limiting interpretability and systematic debugging. We introduce an easy-to-use, plug-and-play framework for concept-based interpretation and model refinement. By aligning a single-modality encoder to BioMedCLIP, we construct a Concept Bottleneck Model (CBM) that enables concept-level interventions. These interventions allow us to isolate causal versus spuriously correlated concepts, validate insights with domain experts, and generate counterfactual samples for targeted fine-tuning. We evaluate our framework on a Mayo Clinic ultrasound dataset and the CheXpert 5x200 chest X-ray dataset. Results demonstrate that concept intervention enables reliable model diagnosis while maintaining, and occasionally improving predictive performance via guided fine-tuning. Our findings highlight the practical value of this framework for controlled, interpretable refinement of clinical deep learning models.
Chinese Translation
医学影像模型通常作为黑箱操作，限制了可解释性和系统调试。我们提出了一种易于使用的即插即用框架，用于基于概念的解释和模型优化。通过将单模态编码器与 BioMedCLIP 对齐，我们构建了一个概念瓶颈模型（Concept Bottleneck Model, CBM），使得概念级干预成为可能。这些干预使我们能够隔离因果概念与虚假相关概念，验证与领域专家的见解，并生成针对性的反事实样本以进行精细调优。我们在梅奥诊所超声数据集和 CheXpert 5x200 胸部 X 光数据集上评估了我们的框架。结果表明，概念干预能够实现可靠的模型诊断，同时通过引导精细调优保持并偶尔提高预测性能。我们的研究结果突显了该框架在临床深度学习模型的可控、可解释优化中的实际价值。
cs.CV / 13 / 2610.09032
Shape-Bayes: Bayesian Inference of Structured Shapes under Visual Ambiguity
Shape-Bayes：在视觉模糊下的结构形状贝叶斯推断
Abstract
Perceiving structured shapes, such as human faces, from pixels is an inherently ambiguous task in real-world conditions. Yet, shape inference is largely posed as a deterministic regression task predicting fixed spatial coordinates. We find that deterministic regression is brittle when visual evidence is ambiguous or incomplete; under severe occlusions deterministic models exhibit structural collapse, predicting incoherent shapes or reverting to generic averages. To address this, we introduce Shape-Bayes, a probabilistic framework that couples uncertainty-aware visual perception with Bayesian shape reasoning. Rather than forcing point estimates, Shape-Bayes dynamically weights visual evidence against geometric priors to infer a structurally valid shape posterior. Demonstrated on human face shape regression, a rigorous testbed featuring complex non-rigid deformations and strict anatomical constraints, Shape-Bayes comprises: (1) a base model predicting noisy landmarks alongside distilled aleatoric uncertainties; (2) a lightweight Transformer encoding these observations into an adaptive prior over a PCA shape manifold; and (3) a differentiable Bayesian solver computing closed-form posteriors by balancing the noisy predictions against this prior. By guaranteeing complete structural integrity, Shape-Bayes achieves an absolute improvement of up to ~34% IDR over state-of-the-art deterministic models. Simultaneously, it yields highly calibrated uncertainty bounds and reduces relative error by up to 12.5%, establishing a new state-of-the-art for robust 2D face shape regression under severe occlusion. The project page is at https://shape-bayes.github.io.
Chinese Translation
在现实条件下，从像素中感知结构形状（如人脸）是一项固有的模糊任务。然而，形状推断通常被视为一个确定性回归任务，旨在预测固定的空间坐标。我们发现，当视觉证据模糊或不完整时，确定性回归模型表现脆弱；在严重遮挡的情况下，确定性模型会出现结构崩溃，预测出不连贯的形状或回归到通用平均值。为了解决这个问题，我们提出了Shape-Bayes，一个将不确定性感知视觉感知与贝叶斯形状推理相结合的概率框架。Shape-Bayes并不是强制进行点估计，而是动态地将视觉证据与几何先验进行加权，以推断出结构有效的形状后验。在人脸形状回归的严格测试平台上，Shape-Bayes展示了其有效性，该平台具有复杂的非刚性变形和严格的解剖约束。Shape-Bayes包括：(1) 一个基础模型，预测噪声标记并提取随机不确定性；(2) 一个轻量级Transformer，将这些观察结果编码为PCA形状流形上的自适应先验；(3) 一个可微分的贝叶斯求解器，通过平衡噪声预测与该先验，计算闭式形式的后验。通过确保完整的结构完整性，Shape-Bayes在相较于最先进的确定性模型上实现了最高约34%的绝对改进。同时，它提供了高度校准的不确定性界限，并将相对误差降低了最高12.5%，在严重遮挡下建立了稳健的2D人脸形状回归的新最先进水平。项目页面为：https://shape-bayes.github.io。
cs.CV / 14 / 2610.09068
GUARD: Geometric Uncertainty-Aware Point Cloud Denoising and Segmentation for Robotic Hard Disk Drive Disassembly
GUARD：用于机器人硬盘驱动器拆卸的几何不确定性感知点云去噪与分割
Abstract
Reliable robotic disassembly requires part-level representations that distinguish genuine component geometry from scanning and reconstruction artifacts. In point clouds of hard disk drives (HDDs), structured ghost artifacts can resemble valid components locally while remaining inconsistent with the overall geometry, allowing erroneous measurements to receive plausible semantic labels. This creates an engineering information problem: semantic prediction confidence alone does not establish whether the underlying geometry is reliable. We propose \textbf{GUARD}, a geometric uncertainty-aware framework that performs point filtering and segmentation within a single forward pass by modeling the reliability of learned geometric representations. GUARD combines a multi-scale geometric transformer with a multi-bandwidth random Fourier feature Gaussian Process to estimate per-point geometric uncertainty, complemented by predictive entropy to suppress unreliable measurements while preserving informative structures. Evaluation on 2,745 real HDD point clouds shows that GUARD improves PointNet++ segmentation mean intersection over union from 0.7739 to 0.8318. Additional experiments on ShapeNetPart and ScanNet examine robustness across corruption types, point-cloud domains, and segmentation backbones. On manually annotated ScanNet samples, geometric uncertainty achieves a corrupted-point detection F1 score of 0.7931, compared with 0.2212 for predictive entropy. The results demonstrate the value of distinguishing geometric reliability from semantic confidence and reveal a tradeoff between artifact suppression and preservation of informative structures. GUARD contributes a reliability-aware approach to interpreting imperfect 3D measurements for component identification and subsequent robotic handling. Project website: https://001-wang.github.io/GUARD_Point_denoiser/.
Chinese Translation
可靠的机器人拆卸需要能够区分真实组件几何形状与扫描和重建伪影的部件级表示。在硬盘驱动器（HDD）的点云中，结构化的伪影在局部上可能与有效组件相似，但与整体几何形状不一致，从而导致错误的测量结果获得合理的语义标签。这造成了一个工程信息问题：仅凭语义预测的置信度无法确定底层几何形状的可靠性。我们提出了 extbf{GUARD}，一种几何不确定性感知框架，通过建模学习到的几何表示的可靠性，在单次前向传递中执行点过滤和分割。GUARD结合了多尺度几何变换器和多带宽随机傅里叶特征高斯过程，以估计每个点的几何不确定性，并辅以预测熵以抑制不可靠的测量，同时保留信息结构。在2,745个真实HDD点云上的评估表明，GUARD将PointNet++的分割平均交并比从0.7739提高到0.8318。对ShapeNetPart和ScanNet的额外实验考察了在不同损坏类型、点云域和分割骨干网络下的鲁棒性。在手动注释的ScanNet样本中，几何不确定性实现了0.7931的损坏点检测F1分数，而预测熵为0.2212。结果表明区分几何可靠性与语义置信度的价值，并揭示了伪影抑制与信息结构保留之间的权衡。GUARD为解释不完美的3D测量以进行组件识别和后续机器人处理提供了一种关注可靠性的方法。项目网站：https://001-wang.github.io/GUARD_Point_denoiser/
cs.CV / 15 / 2610.09071
OverLay++: Dense-Overlap Layout-to-Image Generation Dataset
OverLay++: 密集重叠布局到图像生成数据集
Abstract
Layout-to-Image generation has made substantial progress in spatial and object-level control. However, existing methods still struggle with complex scenes containing many overlapping and interacting objects. We argue that training data is a particular bottleneck: existing datasets lack examples with dense, complex object interactions. To address this gap, we introduce OverLay++, a large-scale Layout-to-Image dataset with structurally complex scenes. OverLay++ contains approximately 500K images with an average of 6.6 objects per image, exceeding existing datasets by 1.67 times in annotation density. Beyond annotation density, OverLay++ provides rich semantic detail with object captions over six times longer than in current datasets. Our dataset generation pipeline is simple and produces dense, overlapping object annotations with rich per-object captions. Across multiple benchmarks, state-of-the-art Layout-to-Image methods trained on the OverLay++ dataset show consistent improvement and faster convergence, demonstrating the importance of dense, overlap-aware, and caption-rich supervision for controllable image generation.
Chinese Translation
布局到图像生成在空间和对象级控制方面取得了显著进展。然而，现有方法在处理包含许多重叠和相互作用对象的复杂场景时仍然面临挑战。我们认为训练数据是一个特别的瓶颈：现有数据集缺乏具有密集、复杂对象交互的示例。为了解决这一问题，我们引入了OverLay++，一个具有结构复杂场景的大规模布局到图像数据集。OverLay++包含大约50万张图像，每张图像平均有6.6个对象，其注释密度超过现有数据集1.67倍。除了注释密度，OverLay++还提供了丰富的语义细节，对象标题的长度是当前数据集的六倍以上。我们的数据集生成管道简单，并生成密集的重叠对象注释以及丰富的每个对象标题。在多个基准测试中，基于OverLay++数据集训练的最先进的布局到图像方法显示出一致的改进和更快的收敛，证明了密集、重叠感知和丰富标题监督在可控图像生成中的重要性。
cs.CV / 16 / 2610.09077
DISRQAD: Diffusion Image Super-Resolution Quality Assessment Dataset and Benchmark
DISRQAD：扩散图像超分辨率质量评估数据集及基准
Abstract
Diffusion-based image super-resolution (SR) can create visually plausible detail that is not supported by the low-resolution input. We introduce DISRQAD, a subjective-quality dataset and diagnostic benchmark for this setting. It contains mean opinion scores (MOS) for 14,000 SR outputs from ten diffusion and four non-diffusion methods, spanning four low-resolution degradation conditions and x2/x4 upscaling. We evaluate 51 standard full-reference and no-reference metric configurations and 11 adapted variants. Agreement with MOS is substantially weaker on diffusion outputs: the strongest standard no-reference baseline reaches 0.431 SRCC on diffusion SR versus 0.813 on non-diffusion SR. As a case study in benchmark use, a pruned and distilled Q-ReAlign-mini student reaches 0.496 SRCC on diffusion SR. DISRQAD measures perceived output quality, not faithfulness to the input; it enables analysis of metric behavior across generator families and input conditions. Our findings reveal a substantial gap in the assessment of diffusion-based SR and provide a basis for developing quality models sensitive to diffusion-specific artifacts.
Chinese Translation
基于扩散的图像超分辨率（SR）能够生成与低分辨率输入不相符的视觉上合理的细节。我们介绍了DISRQAD，这是一个针对该设置的主观质量数据集和诊断基准。该数据集包含来自十种扩散方法和四种非扩散方法的14,000个SR输出的平均意见分数（MOS），涵盖四种低分辨率退化条件以及x2/x4的放大。我们评估了51种标准的全参考和无参考度量配置以及11种改编变体。在扩散输出中，与MOS的一致性明显较弱：最强的标准无参考基线在扩散SR上达到0.431 SRCC，而在非扩散SR上达到0.813 SRCC。作为基准使用的案例研究，一个经过修剪和提炼的Q-ReAlign-mini学生在扩散SR上达到0.496 SRCC。DISRQAD测量的是感知输出质量，而不是对输入的忠实度；它使得能够分析不同生成器家族和输入条件下度量行为。我们的研究结果揭示了在评估基于扩散的SR方面存在显著差距，并为开发对扩散特定伪影敏感的质量模型提供了基础。
cs.CV / 17 / 2610.09085
Anaximander: Interactively Running Geospatial Deep Learning Models on Any Compute Backend
阿那克西曼德：在任何计算后端上交互式运行地理空间深度学习模型
Abstract
Applying deep learning models to satellite imagery from within geographic information systems (GIS) remains high-friction for remote sensing practitioners. Models arrive in incompatible formats and target different compute environments, from local workstations to serverless cloud services. As a result, every evaluation demands custom deployment, tiling, and georeferencing code before a single prediction reaches the analyst's map. This friction discourages systematic comparison in a domain where model choice directly affects operational outcomes such as field delineation, crop monitoring, and disaster response. We present Anaximander, an open-source system that unifies model source and compute location choice behind one interactive interface. The system's backend is an inference server that loads models from multiple commonly-used sources and serves them on any accessible compute backend. The server provides session management and model caching, and streams results back per tile. The backend is paired with a QGIS plugin that drives tiling, result reassembly, georeferencing, and real-time per-tile status visualization. An additional user-interface path injects layer legends as prompts into vision-language models. We demonstrate the system in a code-free side-by-side comparison of three heterogeneous models on an agricultural field delineation task: gpt-image-1 via a cloud API, Segment Anything Model 3 (SAM3) on a remote GPU, and DelineateAnything on a local CPU. The inference backend and protocol are open-source and available at https://github.com/microsoft/nxmndr.
Chinese Translation
在地理信息系统（GIS）中应用深度学习模型于卫星图像仍然对遥感从业者造成较大阻力。模型以不兼容的格式出现，并针对不同的计算环境，从本地工作站到无服务器云服务。因此，每次评估都需要定制的部署、切片和地理参考代码，才能让单个预测到达分析师的地图。这种阻力使得在一个模型选择直接影响操作结果（如地块划分、作物监测和灾害响应）的领域中，系统比较变得困难。我们提出了阿那克西曼德（Anaximander），一个开源系统，通过一个交互式界面统一了模型源和计算位置的选择。该系统的后端是一个推理服务器，可以从多个常用源加载模型，并在任何可访问的计算后端上提供服务。服务器提供会话管理和模型缓存，并按切片流回结果。后端与一个QGIS插件配对，负责切片、结果重组、地理参考和实时每切片状态可视化。另一个用户界面路径将图层图例作为提示注入到视觉-语言模型中。我们在一个农业地块划分任务中演示了该系统，进行了无代码的并排比较，涉及三种异构模型：通过云API的gpt-image-1、在远程GPU上的Segment Anything Model 3（SAM3）和在本地CPU上的DelineateAnything。推理后端和协议是开源的，已在https://github.com/microsoft/nxmndr上提供。
cs.CV / 18 / 2610.09116
SPLATIFY: Reproduce, Discover, Innovate! From Papers and Ideas to Trainable 3DGS Code
SPLATIFY：重现、发现、创新！从论文和想法到可训练的 3DGS 代码
Abstract
The rapid growth of 3D Gaussian Splatting (3DGS) research demands significant effort to reimplement papers before building on them. We introduce SPLATIFY, a multi-agent framework that converts 3DGS papers into trainable gsplat-based implementations, where generic paper-to-code methods and frontier models fail. SPLATIFY achieves this through five innovations: (1) A context-free grammar for gsplat over a modular method template with extension points for losses, densification, rendering, and optimization, constraining synthesis so generated code satisfies gsplat's architectural invariants by construction. (2) Architectural elements for faithful reproduction: fork-aware citation recovery retrieving component-level code at function-level granularity, Graph-of-Thought synthesis in topological dependency order, RAG-guided in-context example selection from over 20 verified implementations, and visual feedback combining PSNR-guided regeneration, Gaussian-level structural checks, and VLM-driven patching. (3) Knowledge-driven compositional improvement that autonomously finds weaknesses and composes complementary regularizers, losses, and densification strategies to improve upon original results. (4) Interdisciplinary method discovery where agents retrieve physical priors from outside the 3DGS literature and compose them with rendering knowledge to produce methods for previously unaddressed scene types. (5) SPLATIFY-Bench, an evaluation framework across 30 diverse 3DGS papers. On papers without public code, SPLATIFY matches expert implementations while reducing development time from weeks to minutes, and through compositional discovery further improves PSNR by up to 2.4 dB. We additionally demonstrate novel methods for volumetric nebula rendering and other scientific domains, synthesized entirely by SPLATIFY.
Chinese Translation
3D 高斯喷溅（3DGS）研究的快速增长要求在构建其基础上之前，必须对论文进行大量的重新实现。我们介绍了 SPLATIFY，一个将 3DGS 论文转换为可训练的基于 gsplat 的实现的多代理框架，在通用的论文到代码方法和前沿模型失效的情况下，SPLATIFY 通过五项创新实现了这一目标：（1）针对 gsplat 的上下文无关文法，结合具有损失、密集化、渲染和优化扩展点的模块化方法模板，约束合成以确保生成的代码在构建时满足 gsplat 的架构不变性。（2）忠实重现的架构元素：关注分叉的引用恢复，以函数级粒度检索组件级代码，按拓扑依赖顺序合成思想图，基于 RAG 的上下文示例选择，从超过 20 个经过验证的实现中选择，并结合 PSNR 引导的再生、Gaussian 级结构检查和 VLM 驱动的修补提供视觉反馈。（3）知识驱动的组合改进，自动发现弱点并组合互补的正则化器、损失和密集化策略，以改进原始结果。（4）跨学科方法发现，代理从 3DGS 文献之外检索物理先验，并将其与渲染知识组合，生成针对先前未解决场景类型的方法。（5）SPLATIFY-Bench，一个涵盖 30 篇多样化 3DGS 论文的评估框架。在没有公开代码的论文中，SPLATIFY 匹配专家实现，同时将开发时间从数周缩短到几分钟，并通过组合发现进一步提高 PSNR 达 2.4 dB。此外，我们还展示了用于体积星云渲染和其他科学领域的新方法，这些方法完全由 SPLATIFY 合成。
cs.CV / 19 / 2610.09118
TopoCurve: Geometry-Aware Topology Reasoning via B\'ezier Curves in Autonomous Driving
TopoCurve：基于贝塞尔曲线的几何感知拓扑推理在自动驾驶中的应用
Abstract
Topology reasoning jointly detects 3D lanes and traffic elements from multi-view images and infers their structural connectivity. Current methods model lanes as discrete polylines, lacking smoothness, analytical tangent directions, and global spatial support for attention, while providing sparse topology supervision. We propose TopoCurve, a geometry-driven architecture for 3D topology reasoning grounded in a structured parametric lane representation. Lanes are modeled as endpoint-fixed cubic B\'ezier curves, enabling continuous geometry with exact endpoints and analytically defined directionality. We exploit this shared curve geometry across the entire pipeline. Endpoint distance and tangent alignment are encoded with multi-scale Fourier features and injected into the topology head. Sampled curve points serve as geometry-aligned references for deformable cross-attention spanning the full lane. Parallel curve-anchored attention branches provide diverse predictions for one-to-many topology supervision. These components form a tightly coupled cascade where representation enables geometric reasoning, guides feature aggregation, and supports denser supervision. TopoCurve achieves 50.6 OLS on the OpenLane-V2 benchmark without any post-processing, establishing a new state-of-the-art among end-to-end camera-only methods, and outperforms all existing approaches on endpoint detection (56.8 vs. 52.6 on DET_p).
Chinese Translation
拓扑推理通过多视角图像联合检测三维车道和交通元素，并推断其结构连接性。目前的方法将车道建模为离散的多边形线，缺乏平滑性、解析切线方向和全局空间支持，同时提供稀疏的拓扑监督。我们提出了TopoCurve，一种基于结构化参数化车道表示的几何驱动架构，用于三维拓扑推理。车道被建模为端点固定的三次贝塞尔曲线，使得几何形状连续，具有精确的端点和解析定义的方向性。我们在整个流程中利用这种共享的曲线几何。端点距离和切线对齐通过多尺度傅里叶特征编码，并注入到拓扑头中。采样的曲线点作为几何对齐的参考，用于跨越整个车道的可变形交叉注意力。平行的曲线锚定注意力分支为一对多的拓扑监督提供多样化的预测。这些组件形成一个紧密耦合的级联，其中表示促进几何推理，指导特征聚合，并支持更密集的监督。TopoCurve在OpenLane-V2基准测试中实现了50.6的OLS，无需任何后处理，确立了端到端仅使用摄像头的方法中的新状态，并在端点检测上超越了所有现有方法（56.8对比52.6在DET_p上）。
cs.CV / 20 / 2610.09125
Depth-to-RGB: Repurposing a Frozen Depth Estimator for Geometry-Guided Compositing
深度到RGB：重新利用冻结的深度估计器进行几何引导合成
Abstract
Reference-based object compositing inserts or replaces an object using a background image, a reference image, and a 2D compositing mask. These inputs guide appearance and placement but leave the completed scene's geometry implicit, which can distort object structure or alter the surroundings. Our Depth-to-RGB (D2R) framework predicts composite depth for a scene not yet observed in the RGB inputs. It learns reference-conditioned corrections to a frozen depth estimator using encoder features of paired completed scenes as targets. The unchanged decoder maps the corrected representation to the intended scene's depth, which a separately trained renderer holds fixed during RGB synthesis. Under matched architecture and training, encoder-feature supervision reduces OOD Stage-1 AbsRel by 31.4% relative to decoded-depth supervision. We also introduce AnyInsertion++ with paired in-distribution and category-disjoint splits to evaluate generalization beyond compositing training categories. The complete D2R system leads 12 open-source and 3 closed-source baselines in estimator-derived geometry and photometric quality on both paired splits. On category-disjoint data, D2R reduces AbsRel by 43.7% and improves PSNR by 2.4 dB over the matched RGB baseline. Across three unpaired benchmarks, D2R leads both identity metrics and reduces mean CLIP reference cosine distance by 55% relative to the strongest baseline. Project page: https://shjo-april.github.io/Depth2RGB/
Chinese Translation
基于参考的物体合成通过背景图像、参考图像和二维合成掩膜插入或替换物体。这些输入指导外观和位置，但使得完成场景的几何形状隐含，这可能会扭曲物体结构或改变周围环境。我们的深度到RGB (Depth-to-RGB, D2R) 框架预测尚未在RGB输入中观察到的场景的合成深度。它利用成对完成场景的编码器特征作为目标，学习对冻结深度估计器的参考条件修正。未改变的解码器将修正后的表示映射到目标场景的深度，而一个单独训练的渲染器在RGB合成期间保持固定。在匹配的架构和训练下，编码器特征监督相较于解码深度监督将OOD Stage-1的绝对相对误差 (AbsRel) 降低了31.4%。我们还引入了AnyInsertion++，通过成对的分布内和类别不重叠的划分来评估超越合成训练类别的泛化能力。完整的D2R系统在估计器导出的几何和光度质量上领先于12个开源和3个闭源基线，在两个成对划分上均表现优异。在类别不重叠的数据上，D2R将AbsRel降低了43.7%，并在匹配的RGB基线之上提高了2.4 dB的峰值信噪比 (PSNR)。在三个无配对基准测试中，D2R在身份度量上均领先，并将平均CLIP参考余弦距离相较于最强基线降低了55%。项目页面：https://shjo-april.github.io/Depth2RGB/
cs.CV / 21 / 2610.09173
RDGSplat: Render-Dedicated Geometry for Novel View Synthesis
RDGSplat：用于新视图合成的渲染专用几何体
Abstract
3D foundation models enable efficient novel view synthesis by carrying a Gaussian head on the representation they already use for reconstruction. However, the views they render fall short of the geometry they recover, because that geometry is estimated under a metric objective and never scored on how it renders. Recent methods alleviate this by updating the backbone weights, but they thereby discard the metric predictions the model was built for and must be repeated for every new backbone. To this end, we propose RDGSplat, a framework that decodes a second geometry dedicated to rendering from a frozen 3D foundation model, leaving its metric predictions intact. In particular, we devise Render-Dedicated Geometry Decoding, which duplicates the pretrained decoders and optimizes the duplicates under photometric supervision alone. Then, a Target-Pose Conditioned Adapter is introduced to reformulate the representation those decoders read, conditioned on the target camera pose rather than the target image. Extensive experiments show that RDGSplat improves novel view synthesis across three feed-forward backbones on four benchmarks, with every pretrained weight frozen. On RE10K, it raises WM2.0 from 20.918 to 24.266\,dB while training 205.5\,M added parameters against a frozen 1.4\,B backbone, and the depth and pose the same model predicts are unchanged.
Chinese Translation
3D基础模型通过在其用于重建的表示上携带高斯头，实现了高效的新视图合成。然而，它们渲染的视图与恢复的几何体之间存在差距，因为该几何体是在度量目标下估计的，并未根据其渲染效果进行评分。最近的方法通过更新主干网络权重来缓解这一问题，但这也导致模型原本构建的度量预测被丢弃，并且每个新的主干网络都必须重复这一过程。为此，我们提出了RDGSplat，一个从冻结的3D基础模型中解码专用于渲染的第二几何体的框架，同时保持其度量预测不变。具体而言，我们设计了渲染专用几何体解码（Render-Dedicated Geometry Decoding），该方法复制预训练解码器，并仅在光度监督下优化这些副本。然后，引入了目标姿态条件适配器（Target-Pose Conditioned Adapter），以根据目标相机姿态而非目标图像重新构造这些解码器读取的表示。大量实验表明，RDGSplat在四个基准测试中提高了三种前馈主干网络的新视图合成性能，且每个预训练权重均保持冻结。在RE10K上，它将WM2.0从20.918提高到24.266 dB，同时训练205.5 M的新增参数，而冻结的1.4 B主干网络预测的深度和姿态保持不变。
cs.CV / 22 / 2610.09185
One Frame, Full Heartbeat: ECG-Free 4D Cardiac Cine MRI Synthesis via Radial-Decomposed Flow Matching
一帧，全心跳：通过径向分解流匹配实现无心电图的4D心脏电影MRI合成
Abstract
Cine cardiovascular magnetic resonance (CMR) captures the cardiac cycle as a four-dimensional (4D) sequence, but standard acquisition requires electrocardiogram (ECG) gating and repeated breath holds. Visual realism alone does not establish accurate patient-specific ejection fraction (EF) or ventricular volumes. We present PhaseFlow3D, a generative framework that synthesizes a complete 4D cine sequence from a single end-diastolic (ED) three-dimensional (3D) volume without ECG. To capture asymmetric systolic and diastolic dynamics, it represents the cardiac cycle as a piecewise linear phase anchored at ED and end-systolic (ES) time points. At inference, a population-level canonical template supplies this phase without patient-specific temporal information. A phase-conditioned rectified flow model generates a cardiac motion trajectory in latent space. Radial Contraction Decomposition converts each latent state into a 3D displacement field, combining a physics-informed radial component for centripetal myocardial contraction with an image-conditioned residual for rotation and out-of-plane motion. Each frame is generated by directly warping the ED volume, bypassing variational autoencoder decoding. On the combined ACDC and M&Ms benchmark, PhaseFlow3D achieves the lowest EF mean absolute error, the only positive left-ventricular volume-curve $R^2$, and the best distributional quality among compared methods. Ablations confirm each component's contribution. Downstream evaluations demonstrate the utility of the synthesized sequences and displacement fields for segmentation, pathology classification, label propagation, and myocardial strain analysis.
Chinese Translation
心血管磁共振成像（CMR）以四维（4D）序列捕捉心脏周期，但标准采集需要心电图（ECG）门控和反复屏气。仅凭视觉真实感无法准确建立特定患者的射血分数（EF）或心室容积。我们提出了PhaseFlow3D，这是一种生成框架，能够从单个舒张末期（ED）三维（3D）体积合成完整的4D电影序列，而无需心电图。为了捕捉不对称的收缩和舒张动态，它将心脏周期表示为一个以ED和收缩末期（ES）时间点为锚的分段线性相位。在推理时，基于人群水平的典型模板提供了这一相位，而不需要特定患者的时间信息。相位条件的整流流模型在潜在空间中生成心脏运动轨迹。径向收缩分解将每个潜在状态转换为3D位移场，结合了针对心肌向心收缩的物理信息径向分量和针对旋转及平面外运动的图像条件残差。每帧通过直接扭曲ED体积生成，绕过变分自编码器解码。在结合的ACDC和M&Ms基准测试中，PhaseFlow3D实现了最低的EF平均绝对误差，唯一的正左心室容积曲线$R^2$，以及在比较方法中最佳的分布质量。消融实验确认了每个组件的贡献。下游评估展示了合成序列和位移场在分割、病理分类、标签传播和心肌应变分析中的实用性。
cs.CV / 23 / 2610.09191
R-CNN-Based Chess Position Recognition
基于 R-CNN 的国际象棋棋局识别
Abstract
Performing chess game position recognition solely from a single image of a three-dimensional board requires predicting the position and orientation of the board relative to the camera, the occupancy of squares and the piece type, which includes its colour. We propose an R-CNN-based framework with independent components for piece recognition and board geometry estimation, whose predictions are combined to reconstruct the position. For piece recognition, we adapt Faster R-CNN using a class-weighted objective and a deeper classification head. The detector operates directly on the input image, retaining alternative piece hypotheses that are subsequently refined using constraints on piece counts and square occupancy. For board detection, we introduce an octagonal arrangement of eight labelled boundary keypoints, predicted using the keypoint head of Mask R-CNN. These provide redundant correspondences for homography estimation and encode board orientation. The estimated homography maps representative points from the piece boxes to an 8x8 grid. On a synthetic dataset, the modifications to piece detection increase mean average precision from 61.59% to 90.14%. Of the predicted board keypoints, 97.11% are within 1% of the image diagonal of their labelled targets. Using ground-truth piece boxes with the predicted homographies gives correct square assignments for every test position. The complete framework recovers 76.61% of test positions exactly and 96.49% with at most one incorrect square.
Chinese Translation
仅通过三维棋盘的单幅图像进行国际象棋棋局识别，需要预测棋盘相对于相机的位置和方向、方格的占用情况以及棋子的类型，包括其颜色。我们提出了一种基于 R-CNN 的框架，具有独立的棋子识别和棋盘几何估计组件，其预测结果结合在一起以重建棋局。对于棋子识别，我们采用了改进的 Faster R-CNN，使用了类加权目标和更深的分类头。检测器直接在输入图像上操作，保留替代棋子假设，随后利用棋子数量和方格占用的约束进行细化。对于棋盘检测，我们引入了八个标记边界关键点的八边形排列，这些关键点通过 Mask R-CNN 的关键点头进行预测。这些关键点为单应性估计提供了冗余对应关系，并编码了棋盘的方向。估计的单应性将棋子框中的代表性点映射到 8x8 网格。在一个合成数据集中，对棋子检测的修改使得平均精度从 61.59% 提高到 90.14%。在预测的棋盘关键点中，97.11% 的点在其标记目标的图像对角线的 1% 范围内。使用真实的棋子框和预测的单应性为每个测试位置提供了正确的方格分配。完整框架准确恢复了 76.61% 的测试位置，并且最多只有一个方格错误的情况下恢复了 96.49%。
cs.CV / 24 / 2610.09195
StableGrasp: Reconstructing Physically Stable Human Hand Grasps from Single Images
StableGrasp：从单幅图像重建物理稳定的人手抓握
Abstract
Reconstructing a physically stable human grasp from a single RGB image is challenging because physically modeling grasps is itself difficult, and the problem requires estimating not only a visually constrained hand pose but also a control target that stabilizes the grasp. Existing methods either model only visual hand geometry without considering physics, or rely on less plausible physical modeling, which limits the physical validity of the resulting grasps. In this paper, we present StableGrasp, a differentiable simulation-based optimization framework that explicitly separates the visual hand pose from the control target that determines the grasping forces. Our method jointly optimizes hand geometry and control by minimizing the kinetic energy of the grasp in a differentiable simulator, while regularizing the hand geometry to preserve visual consistency and geometric plausibility. The reconstructed grasps are substantially more stable under rigorous physical simulation, while remaining visually consistent with the input images and geometrically plausible. Experiments show that our approach produces far more stable grasps than alternative hand-control strategies, benefiting visual-only grasp reconstruction pipelines by turning their outputs into physically stable grasps.
Chinese Translation
从单幅RGB图像重建物理稳定的人手抓握是一项挑战，因为物理建模抓握本身就很困难，并且该问题不仅需要估计视觉约束下的手部姿态，还需要确定一个控制目标以稳定抓握。现有方法要么仅建模视觉手部几何形状而不考虑物理因素，要么依赖于不太可信的物理建模，这限制了所得抓握的物理有效性。在本文中，我们提出了StableGrasp，一个基于可微分仿真的优化框架，明确将视觉手部姿态与决定抓握力的控制目标分开。我们的方法通过在可微分仿真器中最小化抓握的动能来联合优化手部几何形状和控制，同时对手部几何形状进行正则化，以保持视觉一致性和几何合理性。重建的抓握在严格的物理仿真下显著更稳定，同时与输入图像保持视觉一致性和几何合理性。实验表明，我们的方法产生的抓握比替代的手部控制策略稳定得多，能够使仅依赖视觉的抓握重建流程的输出转化为物理稳定的抓握。
cs.CV / 25 / 2610.09200
StyleFields: Multi-Scale AdaIN-Modulated Implicit SDFs for Coarse-to-Fine 3D Shape Reconstruction and Editing
StyleFields：用于粗到细3D形状重建和编辑的多尺度AdaIN调制隐式SDF
Abstract
We introduce StyleFields, a DeepSDF-based architecture for high-fidelity 3D reconstruction that enables controllable geometric style mixing: the coarse structure of one object can be combined with the fine-scale details of another. The core idea is depth-aware modulation: instead of a single global code, we inject latents via multi-level Adaptive Instance Normalization at several decoder depths, and supervise matching auxiliary heads with a coarse-to-fine schedule while gradually growing network depth. This aligns early layers with global shape and later layers with high-frequency detail, achieving content-style decoupling without part labels or adversarial training. StyleFields delivers faithful reconstructions, convincing cross-instance hybrids, and consistent gains in ablations over injection depth and supervision granularity. We further demonstrate a practical application in automotive aerodynamics: a learned surrogate drag predictor serves as a differentiable objective to optimize reconstructed cars, allowing targeted edits of global form or surface details by freezing the complementary latent stream. StyleFields offers a simple, effective recipe for controllable implicit reconstruction and downstream performance-driven design.
Chinese Translation
我们介绍了StyleFields，这是一种基于DeepSDF的高保真3D重建架构，能够实现可控的几何风格混合：一个物体的粗略结构可以与另一个物体的细节相结合。其核心思想是深度感知调制：我们通过在多个解码器深度上使用多级自适应实例归一化（Adaptive Instance Normalization）注入潜在变量，而不是使用单一的全局编码，并在逐渐增加网络深度的同时，监督匹配辅助头部，采用粗到细的调度。这使得早期层与全局形状对齐，而后期层与高频细节对齐，实现了内容与风格的解耦，而无需部件标签或对抗训练。StyleFields提供了真实的重建、令人信服的跨实例混合，并在注入深度和监督粒度的消融实验中取得了一致的提升。我们进一步展示了在汽车空气动力学中的实际应用：学习到的替代阻力预测器作为可微分目标，优化重建的汽车，允许通过冻结互补潜在流来有针对性地编辑全局形状或表面细节。StyleFields为可控的隐式重建和下游性能驱动设计提供了一种简单有效的方法。
cs.CV / 26 / 2610.09205
Do Vision Models Learn Physical Constraints or Rendering Shortcuts? A Counterfactual Benchmark for Grounded Physical Consistency
视觉模型是学习物理约束还是渲染捷径？一个关于基础物理一致性的反事实基准
Abstract
Modern image editing models can satisfy a text instruction while breaking the physics of the edited scene. A new object may cast no shadow, a mirror may fail to reflect visible geometry, or an object may float above a surface that should support it. We study physical plausibility diagnosis, detecting whether an edited image violates scene physics, naming the violation type, localizing the affected region, and explaining the failure in language. We introduce a counterfactual benchmark whose controlled synthetic component uses Mitsuba~3 to generate 5,500 images from 500 scene families. Each family contains one clean image and ten matched violations involving shadows, reflection, support, surface response, and occlusion. The renderer pipeline provides category labels, affected-region masks and boxes, scene metadata, and explanation targets. We use LLaVA-1.5-7B, Qwen2.5-VL-7B, and InternVL3.5-8B as diagnostic baselines rather than proposed methods. On a 1,650-image synthetic test set, the adapted baselines reach 64.0--67.8\% category macro-F1 on standard held-out scenes. For LLaVA-1.5-7B, category macro-F1 falls from 64.0\% on the standard split to 40.8\% under intervention shift. This gap shows that high in-distribution accuracy partly reflects cues tied to rendering and counterfactual construction.
Chinese Translation
现代图像编辑模型可以满足文本指令，但却破坏了编辑场景的物理规律。新的物体可能不投影阴影，镜子可能无法反射可见几何，或者物体可能漂浮在应该支撑它的表面上。我们研究物理合理性诊断，检测编辑图像是否违反场景物理，命名违反类型，定位受影响区域，并用语言解释失败。我们引入一个反事实基准，其受控的合成组件使用 Mitsuba~3 生成 5,500 张来自 500 个场景家族的图像。每个家族包含一张干净图像和十个匹配的违反案例，涉及阴影、反射、支撑、表面响应和遮挡。渲染器管道提供类别标签、受影响区域的掩码和框、场景元数据以及解释目标。我们使用 LLaVA-1.5-7B、Qwen2.5-VL-7B 和 InternVL3.5-8B 作为诊断基准，而不是提出的方法。在 1,650 张图像的合成测试集中，适应后的基准在标准保留场景上达到了 64.0%--67.8%的类别宏观 F1 值。对于 LLaVA-1.5-7B，类别宏观 F1 值从标准划分的 64.0% 降至干预转移下的 40.8%。这一差距表明，高的分布内准确性部分反映了与渲染和反事实构建相关的线索。
cs.CV / 27 / 2610.09217
A Deterministic Evidence Layer for Vision-Language Autism Screening from Naturalistic Home Video
基于自然家庭视频的视觉-语言自闭症筛查的确定性证据层
Abstract
Autism spectrum disorder (ASD) is diagnosed through specialist observation of a child's social behavior, and access to that expertise is the bottleneck for early identification. Vision-language models (VLMs) describe a child's behavior from video well; the verdict drawn from the description is unstable: at temperature~0, across eight pipeline configurations on one backbone, 16--37\% of clips change their predicted label between repeated runs, and the cause lies in the serving stack. We keep the VLM frozen and move the decision out of the model. Under our grounded perception constraints the VLM writes an event table of timestamped, glossary-labeled events that names the eliciting press and logs counter-evidence; a text-only stage age-calibrates the confidence of each row; a deterministic weight-of-evidence scorer sums it into an evidence total and stratifies it into a risk category, so every decision decomposes into named per-feature contributions and can be re-scored from the saved table. On 43 caregiver-recorded, protocol-free home free-play clips of preschool children, the pipeline reaches AUC $0.851 \pm 0.012$, 86.0\% accuracy, and F$_1$ 71.8 over three runs. It labels 74.4\% of clips correctly in every run (60.5\% for the zero-shot baseline) and flags no typically developing clip in every run (9 of 31 at zero-shot). An ablation on the same backbone attributes the gain to the grounded perception constraints read through the deterministic scorer.
Chinese Translation
自闭症谱系障碍（ASD）的诊断依赖于专家对儿童社交行为的观察，而获取这种专业知识是早期识别的瓶颈。视觉-语言模型（VLMs）能够较好地描述儿童在视频中的行为，但从描述中得出的结论不稳定：在温度为~0的情况下，在一个基础模型的八种管道配置中，16%至37%的片段在重复运行中改变了其预测标签，原因在于服务堆栈。我们保持VLM不变，并将决策移出模型。在我们的基础感知约束下，VLM生成一个事件表，记录带时间戳的、带术语标签的事件，命名引发的压力并记录反证据；一个仅文本的阶段对每一行的置信度进行年龄校准；一个确定性的证据权重评分器将其汇总为证据总量并将其分层为风险类别，因此每个决策都可以分解为命名的特征贡献，并可以从保存的表中重新评分。在43个由照顾者记录的、无协议的学龄前儿童家庭自由游戏片段中，该管道达到了AUC $0.851 imes 0.012$，86.0%的准确率，以及F$_1$ 71.8，经过三次运行。它在每次运行中正确标记了74.4%的片段（零样本基线为60.5%），并在每次运行中未标记任何典型发展的片段（零样本中31个片段中有9个）。在相同基础模型上的消融实验将增益归因于通过确定性评分器读取的基础感知约束。
cs.CV / 28 / 2610.09223
PVSync: A Unified Lip-Sync Expert for Timing and Articulation
PVSync：一个统一的唇同步专家，用于时序和发音
Abstract
Lip movements can match the timing of speech without matching the spoken sounds. We introduce PVSync, a unified model for audio-visual offset estimation and phoneme-level articulation scoring. PVSync combines window-level contrastive learning for synchronisation with a phoneme-level articulation objective that aligns audio and video embeddings of the same viseme class across clips. Visemes group phonemes with similar visible articulation. Viseme labels are derived automatically from forced-aligned transcripts, without manual annotations. On offset-corrected videos from 13 talking-head video generation models, PVSync matches human rankings of lip-sync quality more closely than LSE-C, achieving a Spearman correlation of 0.83 versus 0.34. On an automatically constructed benchmark from held-out speech, PVSync distinguishes viseme-matched from mismatched audio-visual pairs with an ROC AUC of 0.91. PVSync also outperforms SyncNet and MTD-VocaLiST in temporal offset recovery on held-out in-the-wild clips. Code and benchmark data will be released upon acceptance.
Chinese Translation
唇部运动可以与语音的时序相匹配，而无需匹配所说的声音。我们介绍了PVSync，一个用于音频-视觉偏移估计和音素级发音评分的统一模型。PVSync结合了用于同步的窗口级对比学习和一个音素级发音目标，该目标对齐同一视觉音素类别在不同片段中的音频和视频嵌入。视觉音素将具有相似可见发音的音素分组。视觉音素标签是从强制对齐的转录文本中自动推导的，无需手动注释。在来自13个说话者视频生成模型的偏移校正视频上，PVSync与人类对唇同步质量的排名更为接近，相较于LSE-C，达到了0.83的斯皮尔曼相关系数，而后者为0.34。在从保留的语音自动构建的基准上，PVSync能够以0.91的ROC AUC区分视觉音素匹配和不匹配的音频-视觉配对。PVSync在保留的真实场景片段的时间偏移恢复方面也优于SyncNet和MTD-VocaLiST。代码和基准数据将在接受后发布。
cs.CV / 29 / 2610.09242
Pooling Representation Autoencoders for Efficient Diffusion
高效扩散的池化表示自编码器
Abstract
Representation Autoencoders (RAEs) generate images from pre-trained visual fea- tures, but their dense token grids make generative modeling expensive. Motivated by local feature correlations, we introduce PoolDINO, a learned affine pooling operator that merges neighboring tokens. Training the pooling operator jointly with the RGB decoder preserves the standard two-stage RAE procedure without a separate feature auto-encoder. On ImageNet-256, 4x token compression retains comparable generation quality under internal guidance, while 16x compression trades some quality for greater efficiency. At a fixed budget of 100 sampling steps, latent-sampling throughput increases by 3.7x and 9.0x, respectively, relative to the unpooled baseline. Classification and dense prediction evaluations show that comparable guided generation quality can coexist with weaker performance on other tasks.
Chinese Translation
表示自编码器（Representation Autoencoders, RAEs）从预训练的视觉特征生成图像，但其密集的标记网格使得生成建模成本高昂。基于局部特征相关性的动机，我们引入了PoolDINO，一种学习的仿射池化算子，用于合并相邻的标记。与RGB解码器共同训练池化算子，保留了标准的两阶段RAE过程，而无需单独的特征自编码器。在ImageNet-256上，4倍的标记压缩在内部指导下保持了可比的生成质量，而16倍压缩则在提高效率的同时牺牲了一定的质量。在固定的100次采样步骤预算下，相较于未池化的基线，潜在采样的吞吐量分别提高了3.7倍和9.0倍。分类和密集预测评估表明，可比的指导生成质量可以与其他任务的较弱性能共存。
cs.CV / 30 / 2610.09252
Adaptive Visual Token Reduction for Accelerated Image Understanding
自适应视觉标记减少以加速图像理解
Abstract
Large Vision-Language Models achieve strong VQA performance, but processing high-resolution, information-rich images requires substantial computation, motivating visual token reduction. However, existing methods often prune individual tokens or rely on fixed-size cropping, limiting their ability to preserve spatially structured information such as horizontally or vertically elongated text. To address this limitation, we propose ReFIT, an instruction-guided visual token reduction framework for efficient LVLM inference. ReFIT consists of Relevance-Guided Window Reshaping (RWR) and Instruction-Guided Token Refinement (ITR), where RWR captures instruction-relevant regions by adapting to their spatial characteristics, while ITR further removes unnecessary visual tokens. Experiments on four VQA benchmarks demonstrate that ReFIT improves answer accuracy while reducing computational cost, and qualitative results demonstrate its effectiveness in localizing relevant regions and removing unnecessary visual information.
Chinese Translation
大型视觉-语言模型在视觉问答（VQA）任务中表现出色，但处理高分辨率、信息丰富的图像需要大量计算，这促使了视觉标记的减少。然而，现有方法通常仅修剪单个标记或依赖固定大小的裁剪，限制了它们保留空间结构信息（如水平或垂直延伸的文本）的能力。为了解决这一局限性，我们提出了ReFIT，一个基于指令的视觉标记减少框架，用于高效的LVLM推理。ReFIT由相关性引导的窗口重塑（RWR）和指令引导的标记精炼（ITR）组成，其中RWR通过适应空间特征捕获与指令相关的区域，而ITR进一步去除不必要的视觉标记。在四个VQA基准上的实验表明，ReFIT在提高答案准确性的同时降低了计算成本，定性结果展示了其在定位相关区域和去除不必要视觉信息方面的有效性。
cs.CV / 31 / 2610.09253
PhyDiCT: Plug-and-Play CT Reconstruction from Sparse X-Rays via Differentiable Rendering and Strong Priors
PhyDiCT：基于可微渲染和强先验的稀疏X射线即插即用CT重建
Abstract
Reconstructing 3D Computed Tomography (CT) images from a few X-ray projections is a highly ill-posed inverse problem due to the loss of volumetric information. We propose PhyDiCT, a training-free framework that integrates a differentiable Physics-based forward model, grounded in the Beer-Lambert law, with a text-conditioned Diffusion as a strong prior to reconstruct 3D lung CT images. We refer to our approach as training-free since the prior model is used without fine-tuning, and our goal is to steer the denoising procedure to generate samples consistent with X-ray observations. We guide the diffusion generation using Split Gibbs sampling to jointly optimize for projection fidelity (reward) and consistency with prior knowledge. Also, we introduce a test-time refinement step that enhances image realism and anatomical coherence. We extensively evaluate our method on publicly available 3D CT datasets using both perceptual and semantic metrics, demonstrating that it surpasses existing plug-and-play diffusion and fully trained reconstruction approaches. Our findings highlight that combining a strong generative prior with the underlying physics of image formation substantially improves reconstruction quality, e.g., 7.5\% improvement on SSIM compared to full training methods. Code will be released at https://github.com/batmanlab/PhyDiCT.
Chinese Translation
从少量X射线投影重建3D计算机断层扫描（CT）图像是一个高度不适定的逆问题，因为体积信息的丢失。我们提出了PhyDiCT，这是一个无训练框架，结合了基于物理的可微前向模型（基于Beer-Lambert定律）和条件文本扩散作为强先验，以重建3D肺部CT图像。我们将这种方法称为无训练，因为先验模型在不进行微调的情况下使用，我们的目标是引导去噪过程生成与X射线观测一致的样本。我们使用Split Gibbs采样指导扩散生成，以联合优化投影保真度（奖励）和与先验知识的一致性。此外，我们引入了一个测试时精炼步骤，以增强图像的真实感和解剖一致性。我们在公开可用的3D CT数据集上广泛评估了我们的方法，使用感知和语义指标，证明其超越了现有的即插即用扩散和完全训练的重建方法。我们的研究结果强调，将强生成先验与图像形成的基本物理结合在一起，显著提高了重建质量，例如，与完全训练方法相比，SSIM提高了7.5%。代码将发布在https://github.com/batmanlab/PhyDiCT。
cs.CV / 32 / 2610.09274
Hardware-aware Calibrated Clustered Attention for Efficient Visual Geometric Transformers
硬件感知的校准聚类注意力机制用于高效的视觉几何变换器
Abstract
The Visual Geometry Grounded Transformer (VGGT) marks a significant leap forward in 3D scene reconstruction, as it is the first model that directly infers all key 3D attributes (camera poses, depths, and dense geometry) jointly in one pass. However, this joint inference mechanism requires global attention layers with extremely long sequences that causes a significant latency bottleneck. In this paper, we propose blockwise clustered attention (BC attention) to accelerate the global attention layers in VGGT. By limiting the clustering within HW-friendly neighborhood blocks, BC attention reduces the computation overhead of query clustering as well as the costly data movement between on- and off-chip memory. This enables BC attention to scale to long sequences and deliver practical latency improvements on GPUs. Moreover, we introduce a hashing hyperplane calibration method and a threshold-based error compensation method to reduce clustering errors efficiently, which is a bottleneck in the current clustered attention mechanism. Overall, our experiments on GPU demonstrate that calibrated BC attention accelerates the global attention layers by 2.10-2.63$\times$ and the whole backbone by 1.77-2.35$\times$ with negligible loss (1%) for large scenes. With a small performance loss (< 5%), calibrated BC attention further achieves a 2.26-2.87$\times$ latency improvement on the global attention layers and a 1.90-2.55$\times$ improvement on the backbone.
Chinese Translation
视觉几何基础变换器（VGGT）标志着3D场景重建的重大进展，因为它是第一个在一次推理中直接联合推断所有关键3D属性（相机姿态、深度和稠密几何）的模型。然而，这种联合推理机制需要具有极长序列的全局注意力层，这导致了显著的延迟瓶颈。本文提出了一种块状聚类注意力机制（BC attention），旨在加速VGGT中的全局注意力层。通过限制聚类在硬件友好的邻域块内，BC attention减少了查询聚类的计算开销以及芯片内外存储之间昂贵的数据移动。这使得BC attention能够扩展到长序列，并在GPU上实现实际的延迟改进。此外，我们引入了一种哈希超平面校准方法和基于阈值的误差补偿方法，以有效减少聚类误差，这是当前聚类注意力机制中的一个瓶颈。总体而言，我们在GPU上的实验表明，校准的BC attention使全局注意力层加速了2.10-2.63倍，整个主干网络加速了1.77-2.35倍，且对大场景的损失微乎其微（1%）。在小幅性能损失（< 5%）的情况下，校准的BC attention在全局注意力层上进一步实现了2.26-2.87倍的延迟改进，并在主干网络上实现了1.90-2.55倍的改进。
cs.CV / 33 / 2610.09275
EM-SNN: Efficiently Modulated Spiking Neural Network for Remote Sensing Image Dehazing
EM-SNN：用于遥感图像去雾的高效调制脉冲神经网络
Abstract
Although spiking neural networks (SNNs) provide an energy-efficient alternative to artificial neural networks (ANNs), their application to remote sensing image dehazing remains limited. A key challenge arises from the coupling between haze-induced high-frequency attenuation and discrete spike thresholding. This interaction suppresses weak responses and fundamentally limits the recovery of edges, textures, and fine details in spiking dehazing models. To address this challenge, we propose the Efficiently Modulated Spiking Neural Network (EM-SNN), a dedicated spiking framework tailored to remote sensing image dehazing. EM-SNN integrates a statistics-driven Threshold-Modulated Leaky Integrate-and-Fire (TM-LIF) neuron to adaptively compensate for haze-induced contrast compression, together with a Spike Sobel Modulation (SSM) module that enhances structural cues and reduces depth-wise attenuation during spiking feature propagation. By jointly modulating activation scales and structural representations, EM-SNN improves dehazing performance while preserving the inherent event-driven sparsity of SNNs. Experiments on HRSD, RICE, RRSHID, and SateHaze1K demonstrate that EM-SNN achieves competitive dehazing performance while consuming only one quarter of the energy of the strong ANN baseline SFRDP-Net.
Chinese Translation
尽管脉冲神经网络（SNNs）提供了一个能量高效的替代方案来替代人工神经网络（ANNs），但其在遥感图像去雾中的应用仍然有限。一个主要挑战源于雾霾引起的高频衰减与离散脉冲阈值之间的耦合。这种相互作用抑制了微弱响应，并从根本上限制了脉冲去雾模型中边缘、纹理和细节的恢复。为了解决这一挑战，我们提出了高效调制脉冲神经网络（EM-SNN），这是一个专门为遥感图像去雾量身定制的脉冲框架。EM-SNN集成了一个基于统计的阈值调制泄漏积分发射（TM-LIF）神经元，以自适应地补偿雾霾引起的对比度压缩，并结合一个脉冲索贝尔调制（SSM）模块，增强结构线索并减少在脉冲特征传播过程中的深度衰减。通过共同调制激活尺度和结构表示，EM-SNN在提高去雾性能的同时，保持了SNNs固有的事件驱动稀疏性。在HRSD、RICE、RRSHID和SateHaze1K上的实验表明，EM-SNN在去雾性能上具有竞争力，同时仅消耗强ANN基线SFRDP-Net能量的四分之一。
cs.CV / 34 / 2610.09313
Why VLMs Miss Small Objects, and When Zooming In Is Safe
为何视觉语言模型错过小物体，以及何时放大是安全的
Abstract
Vision-language models (VLMs) often miss small objects in large images. We ask three questions: what limits them, which of these limits better models can remove, and whether the classical way of handling large images, local decomposition, still has a future. We answer them with a theory built on two quantities of the image interface: S, the number of visual tokens across an object's side, and L, the content a call must cover. The limits: a W x H image sent whole within N tokens gives an object of side m at most m*sqrt(N/(WH)) tokens per side. Doubling the token budget N therefore raises the largest S a whole image can reach by only 41%, and seeing an image at the S an object needs costs at least order S^2 tokens whatever the model. If recognition improves gradually with S, any search strategy, zoom agents included, obeys a recall-cost frontier. What better models can change: the S an object needs and how much one call can carry, measured in bits per object found; the coverage cost remains. Decomposition: yes. Assuming only that more tokens per object and less content per call do not hurt on average, splitting an image cannot lower recall if no view zooms out relative to the whole image and views overlap by one object. Neither part of this condition can be dropped; we bound the cost of every such decomposition, and a simple rule approaches the bound as the image grows. We test the theory in about 177,000 requests on 797 images. On controlled images, none of 40 orderings predicted for 8 VLMs is violated. Checked after the fact on drawings, floor plans, natural and synthetic images, 61 of 89 implied orderings hold significantly and 4 fail, all for OpenAI models given more pixels than their default path. On construction drawings the rule never significantly lowered recall relative to the whole image and raised it by up to 0.28. Code and data: https://github.com/shijunzhe/vlm-small-objects
Chinese Translation
视觉语言模型（VLMs）在处理大图像时常常会错过小物体。我们提出了三个问题：是什么限制了它们，这些限制中哪些可以通过更好的模型去除，以及传统的处理大图像的方法——局部分解，是否仍然有未来。我们通过建立在图像接口的两个量（S，物体边缘的视觉标记数量；L，一个调用必须覆盖的内容）上的理论来回答这些问题。限制条件是：一个 W x H 的图像在 N 个标记内整体发送时，最多只能为边长为 m 的物体提供 m*sqrt(N/(WH)) 个标记。因而，增加标记预算 N 仅能使整个图像能够达到的最大 S 增加41%，而在任何模型下，看到一个物体所需的 S 的图像至少需要 S^2 个标记。如果识别随着 S 的增加而逐渐改善，那么任何搜索策略，包括放大代理，都遵循一个召回-成本边界。更好的模型可以改变的内容包括：物体所需的 S 以及一个调用能承载的内容量（以每个找到的物体的比特数衡量）；覆盖成本保持不变。局部分解：是的。假设每个物体的标记数量增加而每个调用的内容减少在平均上不会造成损害，只要没有视图相对于整个图像缩小且视图之间重叠一个物体，分割图像不会降低召回率；该条件的任一部分都不能被省略；我们对每种分解的成本进行了界定，且随着图像的增长，一个简单的规则接近该界限。我们在约177,000个请求和797幅图像上测试了该理论。在受控图像上，40种针对8个VLMs的排序预测没有被违反。经过事后检查，在图纸、平面图、自然和合成图像中，89个隐含排序中有61个显著成立，4个失败，所有失败均为OpenAI模型在给定比其默认路径更多像素时出现。在施工图上，该规则相对于整个图像从未显著降低召回率，且最多提高了0.28。代码和数据： https://github.com/shijunzhe/vlm-small-objects
cs.CV / 35 / 2610.09317
SAREO-FM: Decoupled Semantic Supervision for SAR-EO Foundation Models
SAREO-FM：用于SAR-EO基础模型的解耦语义监督
Abstract
Synthetic aperture radar (SAR) and electro-optical (EO) imagery provide complementary observations: SAR enables day-and-night, weather-resilient sensing, whereas EO provides rich appearance and fine-grained semantic cues. We introduce SAREO-FM, which avoids forcing a single token stream to serve two distinct roles: modality tokens preserve how each sensor observes the scene through masked reconstruction, while learnable semantic queries capture what the scene contains under guidance from a pretrained vision foundation model (VFM). By jointly encoding these queries with SAR and EO tokens, the queries acquire modality-grounded semantic context, while the modality-token outputs remain the explicit targets of masked reconstruction. This design assigns semantic and reconstruction supervision to separate token streams while preserving their interaction within the shared encoder. Pretrained on the million-scale SAR-1M corpus, SAREO-FM achieves strong unimodal transfer for both SAR-only and EO-only inputs, while delivering substantial gains from joint SAR--EO observations on tasks that benefit from complementary sensing.
Chinese Translation
合成孔径雷达（SAR）和电光（EO）图像提供了互补的观测：SAR能够在昼夜和恶劣天气条件下进行感知，而EO则提供丰富的外观和细粒度的语义线索。我们提出了SAREO-FM，它避免了强迫单一的标记流承担两个不同的角色：模态标记通过掩码重建保留了每个传感器如何观察场景，而可学习的语义查询在预训练的视觉基础模型（VFM）的指导下捕捉场景所包含的内容。通过将这些查询与SAR和EO标记共同编码，查询获得了基于模态的语义上下文，而模态标记的输出仍然是掩码重建的明确目标。这一设计将语义和重建监督分配给不同的标记流，同时保留它们在共享编码器中的交互。在百万规模的SAR-1M语料库上进行预训练后，SAREO-FM在SAR单模态和EO单模态输入上实现了强大的单模态迁移，同时在受益于互补感知的任务中，从联合SAR-EO观测中获得了显著的提升。
cs.CV / 36 / 2610.09326
VIS-Ground: Video Interactive Storytelling with Contextual Grounding
VIS-Ground：具有上下文基础的视频互动叙事
Abstract
Video interactive storytelling enables viewers to actively steer how a video unfolds. However, once we allow viewers to intervene during generation, a new challenge arises: The viewer's request can have latent dependencies on both the grounding source and the current rendered video state. These dependencies may not be explicitly stated in any individual input, but emerge only when the source, rendered history, and new viewer intent are considered jointly. Existing interactive video generation systems primarily emphasize following viewer instructions, while source-grounded video generation methods focus on aligning generated content with an external narrative or knowledge source. This leaves a fundamental question underexplored: What context should a generation model ground on during interactive continuation, and how can heterogeneous, unstructured inputs be transformed into such grounding context? In this work, we formulate contextual grounding as the process of transforming heterogeneous input context into an executable constraint model for video generation. To address this challenge, we introduce VIS-Ground, which performs Structured Context Abstraction to recover grounded states and cross-context dependencies, Generation Constraints Induction to project relevant dependencies into candidate-specific constraints, and Constrained Video Generation to enforce these constraints through planning, verification, revision, and rendering. Across three video generation backbones, VIS-Ground consistently achieves the highest overall composite score, reaching an average absolute improvement of 10.3 points over the strongest per-backbone baselines. Detailed analysis further shows gains across both narrative and knowledge grounding, and reveals remaining challenges in dependency extraction, and faithful realization during video rendering.
Chinese Translation
视频互动叙事使观众能够主动引导视频的发展。然而，一旦我们允许观众在生成过程中进行干预，就会出现一个新的挑战：观众的请求可能对基础来源和当前渲染视频状态有潜在依赖。这些依赖可能在任何单一输入中并未明确表述，而是只有在考虑来源、渲染历史和新的观众意图时才会共同显现。现有的互动视频生成系统主要强调遵循观众的指示，而基于来源的视频生成方法则专注于将生成内容与外部叙事或知识来源对齐。这留下了一个未被充分探讨的基本问题：在互动延续过程中，生成模型应该基于什么上下文进行基础化，以及如何将异构的、非结构化的输入转化为这样的基础上下文？在本研究中，我们将上下文基础化公式化为将异构输入上下文转化为视频生成的可执行约束模型的过程。为了解决这一挑战，我们引入了VIS-Ground，它通过结构化上下文抽象来恢复基础状态和跨上下文依赖，通过生成约束归纳将相关依赖投影到候选特定约束中，并通过规划、验证、修订和渲染来强制执行这些约束。在三个视频生成基础架构中，VIS-Ground始终实现了最高的整体综合得分，平均绝对提升达10.3分，超越了最强的每个基础架构基线。详细分析进一步显示在叙事和知识基础方面的提升，并揭示了在依赖提取和视频渲染过程中的忠实实现方面仍然存在的挑战。
cs.CV / 37 / 2610.09328
Visual Jev Rewards: Reference-Bound Verification for Multi-Subject Image Generation
视觉Jev奖励：多主体图像生成的参考边界验证
Abstract
Multi-subject image generation requires rewards that verify whether requested attributes, actions, and relations hold for the specified reference subjects. Subject presence alone does not establish that the correct subjects participate in a requested interaction. We present reference-bound Visual Jev rewards that turn these visual decisions into generator training signals. Each subject-related question receives a positive label only when the requested condition and the relevant reference identities hold jointly. We construct fixed questions offline, train a Qwen3.5-4B verifier with binary supervision, and directly read Yes probabilities from its language-model head. Their mean supplies a GRPO reward while retaining individual judgments for inspection. Using 200 MICo-150K training tasks and 30 updates, the framework raises a GPT-5.4 composite score from 41.78 to 52.50 on a manually selected 897-task MICo-Bench subset; direct 27B rewards yield 51.84. Each reward is tested in one GRPO run, and offline human evaluation does not establish a statistically significant advantage over direct scoring. The study provides an initial implementation and evaluation of Visual Jev as a reference-bound reward for multi-subject image generation.
Chinese Translation
多主体图像生成需要奖励来验证请求的属性、动作和关系是否适用于指定的参考主体。仅仅存在主体并不能证明正确的主体参与了请求的交互。我们提出了参考边界视觉Jev奖励，将这些视觉决策转化为生成器训练信号。只有当请求的条件和相关的参考身份共同成立时，每个与主体相关的问题才会获得正标签。我们离线构建固定问题，使用二元监督训练Qwen3.5-4B验证器，并直接从其语言模型头部读取“是”的概率。它们的均值提供了GRPO奖励，同时保留个体判断以供检查。使用200个MICo-150K训练任务和30次更新，该框架将GPT-5.4的综合得分从41.78提高到52.50，在手动选择的897任务MICo-Bench子集中；直接的27B奖励则产生51.84。每个奖励在一次GRPO运行中进行测试，离线人类评估未能证明其相较于直接评分具有统计显著优势。本研究提供了视觉Jev作为多主体图像生成的参考边界奖励的初步实现和评估。
cs.CV / 38 / 2610.09329
GRC-Net: Global Representation Consistency Network for Unsupervised Multimodal Anomaly Detection
GRC-Net：用于无监督多模态异常检测的全局表示一致性网络
Abstract
Automated quality inspection is essential for ensuring product reliability in manufacturing.While image-based methods effectively capture appearance-related defects, these methods are limited in detecting structural and geometric anomalies, motivating multimodal approaches incorporating 3D information. However, existing methods mainly rely on local patch-level representations, which often lead to unstable reconstruction errors even in normal regions. To address this limitation, we propose GRC-Net, which integrates a global-attention MLP to enforce global representation consistency across patch embeddings with a stable reconstruction module to improve reconstruction stability. The proposed method captures holistic contextual information through a global token and suppresses reconstruction noise by minimizing discrepancies between original and predicted embeddings. Experiments on MVTec 3D-AD and Eyecandies demonstrate that GRC-Net consistently outperforms existing methods at both image and pixel levels. Qualitative results further demonstrate reduced reconstruction errors in normal regions and more distinct reconstruction differences between normal and anomalous regions.
Chinese Translation
自动化质量检测对于确保制造业产品的可靠性至关重要。尽管基于图像的方法有效捕捉与外观相关的缺陷，但这些方法在检测结构和几何异常方面存在局限性，因此需要结合三维信息的多模态方法。然而，现有的方法主要依赖于局部补丁级别的表示，这往往导致即使在正常区域也会出现不稳定的重建误差。为了解决这一限制，我们提出了GRC-Net，该网络集成了全局注意力多层感知器（MLP），以在补丁嵌入之间强制执行全局表示一致性，并配备稳定的重建模块以提高重建稳定性。所提出的方法通过全局标记捕捉整体上下文信息，并通过最小化原始嵌入与预测嵌入之间的差异来抑制重建噪声。在MVTec 3D-AD和Eyecandies上的实验表明，GRC-Net在图像和像素级别上均优于现有方法。定性结果进一步表明，在正常区域重建误差降低，并且正常区域与异常区域之间的重建差异更加明显。
cs.CV / 39 / 2610.09330
CRT-HMAR: Causal Requirement Tracing-Guided Hierarchical Multi-Agent Regulation for Open-Task-Aware Infrared-Visible Image Fusion
CRT-HMAR：基于因果需求追踪的分层多智能体调节框架用于开放任务感知的红外-可见图像融合
Abstract
Infrared and visible (IR-VIS) image fusion integrates complementary multimodal information into a single fused image to support downstream vision tasks. However, existing methods are typically tailored to seen tasks within a fixed task set and struggle to generalize to unseen tasks, which restricts their applicability in real-world open-task scenarios. To address this issue, this paper proposes CRT-HMAR, a Causal Requirement Tracing-Guided Hierarchical Multi-Agent Regulation Framework for open-task-aware IR-VIS image fusion. CRT-HMAR introduces a Causal Requirement Tracing Task Localization mechanism, which actively intervenes in key image information and observes task-network response variations to map task-specific semantic preferences into image-level causal requirement maps. Based on these maps, a requirement analysis agent aggregates task-specific requirement knowledge to adaptively guide requirement-customized image fusion. Moreover, CRT-HMAR incorporates History-Analysis Multi-Objective Balancing and Task-Level-Correction Conflict Mitigation mechanisms, jointly constructing a hierarchical regulation chain of "requirement interpretation - task balancing - conflict mitigation". Through multiple collaborative agents, CRT-HMAR dynamically regulates key processes including open-task requirement modeling, multi-task balanced optimization, and gradient conflict mitigation. Extensive experiments on open-task scenarios involving five downstream tasks demonstrate that CRT-HMAR significantly improves generalization to unseen tasks while maintaining the performance and balance of seen tasks. Overall, CRT-HMAR shifts IR-VIS image fusion from task-oriented modeling toward requirement-oriented modeling, promoting its extension from closed-task settings to real-world open-task scenarios.
Chinese Translation
红外和可见（IR-VIS）图像融合将互补的多模态信息整合为单一的融合图像，以支持下游视觉任务。然而，现有的方法通常针对固定任务集中的已知任务进行定制，难以推广到未知任务，这限制了它们在现实世界开放任务场景中的适用性。为了解决这一问题，本文提出了CRT-HMAR，一个基于因果需求追踪的分层多智能体调节框架，旨在实现开放任务感知的IR-VIS图像融合。CRT-HMAR引入了一种因果需求追踪任务定位机制，该机制主动干预关键图像信息，并观察任务网络响应的变化，将任务特定的语义偏好映射到图像级因果需求图上。基于这些图，需求分析代理聚合任务特定的需求知识，以自适应地指导需求定制的图像融合。此外，CRT-HMAR结合了历史分析多目标平衡和任务级纠正冲突缓解机制，共同构建了“需求解释 - 任务平衡 - 冲突缓解”的分层调节链。通过多个协作代理，CRT-HMAR动态调节关键过程，包括开放任务需求建模、多任务平衡优化和梯度冲突缓解。在涉及五个下游任务的开放任务场景中的大量实验表明，CRT-HMAR显著提高了对未知任务的泛化能力，同时保持了已知任务的性能和平衡。总体而言，CRT-HMAR将IR-VIS图像融合从任务导向建模转向需求导向建模，促进其从封闭任务环境向现实世界开放任务场景的扩展。
cs.CV / 40 / 2610.09343
TileSkipper: Region-Adaptive Tile Pruning for 3D Gaussian Splatting
TileSkipper：区域自适应瓷砖修剪用于3D高斯点云渲染
Abstract
Tiled 3D Gaussian Splatting rasterizers often use one scene-wide contribution cutoff for tile enumeration, although content differs in its sensitivity to support truncation. TileSkipper selects a static per-Gaussian cutoff policy for a frozen checkpoint. Calibration renders measure candidate pair savings and an isolated-removal distortion proxy that accounts for front transmittance and background color. The method allocates cutoffs across 64 Gaussian groups and accepts policies only after complete renders on disjoint selection views. The exported policy uses one byte per Gaussian, with no parameter updates, additional kernel, or per-frame policy inference. Across 13 scenes from Mip-NeRF 360, Tanks & Temples, and Deep Blending, a fixed-policy AccuTile sweep gives dataset-macro speedups of $1.088\times$ at standard resolution and $1.238\times$ at 3840 pixels wide, with $-0.007/-0.023$ dB mean PSNR change. Six integrations with existing opacity-aware bounds yield $1.009\times$--$1.121\times$ compiler-only speedups. For four ports from $3\sigma$ rasterizers, we separately attribute the prior exact-bound transition and our incremental gain. Matched-quality ablations show modest gains over scene-global calibration and parity with per-Gaussian control; the standalone comparison with AdaGScale is regime-dependent.
Chinese Translation
瓷砖式3D高斯点云渲染器通常使用统一的场景贡献截断来进行瓷砖枚举，尽管内容对支持截断的敏感性各不相同。TileSkipper为冻结的检查点选择静态的每个高斯的截断策略。校准渲染测量候选对的节省和一个孤立移除失真代理，考虑前向透射和背景颜色。该方法在64个高斯组之间分配截断，并且仅在对不相交选择视图的完整渲染后接受策略。导出的策略对每个高斯使用一个字节，没有参数更新、额外内核或每帧策略推断。在来自Mip-NeRF 360、Tanks & Temples和Deep Blending的13个场景中，固定策略的AccuTile遍历在标准分辨率下提供了$1.088 imes$的数据集宏观加速，在3840像素宽时为$1.238 imes$，平均PSNR变化为$-0.007/-0.023$ dB。与现有的透明度感知边界的六次集成产生了$1.009 imes$到$1.121 imes$的仅编译加速。对于来自$3σ$光栅化器的四个端口，我们分别归因于先前的精确边界过渡和我们的增量收益。匹配质量的消融实验显示出相对于场景全局校准的适度增益，并与每个高斯控制相当；与AdaGScale的独立比较则依赖于具体的应用场景。
cs.CV / 41 / 2610.09344
Do Image Editors Follow Depth-Dependent Blur and Aperture Response? A Rendered-Ground-Truth Pilot Audit
图像编辑器是否遵循深度依赖模糊和光圈响应？一项渲染真实情况的初步审计
Abstract
General image editors are asked to make a photo look as if it were taken at f/1.4, yet it is rarely checked whether the blur they add follows thin-lens optics. A physical aperture edit spreads blur across depth in thin-lens proportions and changes the blur when the aperture changes; prior evaluations check blur monotonicity, sharpness-trend correlation, effective-aperture error, or vision-language judgments, and none we found reports the two properties separately at known depths. In this pilot audit of two editors (Gemini~3.1 Flash Image and GPT-image-2.5) we compare against a rendered oracle: Blender Cycles scenes with true thin-lens depth of field, one blur-width estimator applied identically to oracle and editor outputs, and preregistered depth and aperture indices. In 24 texture scenes rendered in one three-panel geometry, accepted and measurable panels show f/1.4-to-f/2.8 width ratios, $\bar\sigma_{1.4}/\bar\sigma_{2.8}$, of 0.99--1.18 against 1.98--2.13 for the oracle, and ratios of pooled median Gaussian-equivalent near/far blur widths of about 1.18--1.25 (Gemini) and 0.97--1.05 (GPT-image) against 1.69--1.82. Preregistered black-box interventions show that qualitative wording changes blur strength by roughly 2--10 times, whereas a request for 2 versus 6 px changes it 1.1--1.3 times and no tested wording of the f-number meets the registered ``followed'' criterion. The depth compression appears in the original and reversed centre-focus layouts; with the focus on the near panel, the available-panel depth index reaches the registered threshold, and the aperture response stays attenuated in every layout tested. Scalar metrics adapted from published ones give oracle-like scores to synthetic editors whose proportions are compressed.
Chinese Translation
一般图像编辑器被要求使照片看起来像是在 f/1.4 下拍摄的，但很少检查它们添加的模糊是否遵循薄透镜光学。物理光圈编辑在薄透镜比例下在深度上扩展模糊，并在光圈变化时改变模糊；之前的评估检查模糊的单调性、清晰度趋势相关性、有效光圈误差或视觉-语言判断，但我们发现没有一项报告在已知深度下分别考虑这两个属性。在对两个编辑器（Gemini~3.1 Flash Image 和 GPT-image-2.5）的这项初步审计中，我们与一个渲染的真相进行比较：Blender Cycles 场景具有真实的薄透镜景深，一个模糊宽度估计器对真相和编辑器输出进行相同的应用，以及预注册的深度和光圈指数。在一个三面几何体中渲染的 24 个纹理场景中，接受和可测量的面板显示 f/1.4 到 f/2.8 的宽度比，$ar au_{1.4}/ar au_{2.8}$，为 0.99--1.18，而真相为 1.98--2.13，聚合的中位数高斯等效近/远模糊宽度的比率约为 1.18--1.25（Gemini）和 0.97--1.05（GPT-image），而真相为 1.69--1.82。预注册的黑箱干预显示，定性措辞将模糊强度改变约 2--10 倍，而对 2 像素与 6 像素的请求则改变了 1.1--1.3 倍，且没有测试的光圈值措辞符合注册的“遵循”标准。深度压缩出现在原始和反向中心聚焦布局中；在近面板上聚焦时，可用面板深度指数达到注册阈值，而在测试的每个布局中光圈响应保持减弱。根据已发布的标量指标调整的指标为合成编辑器提供了类似真相的得分，其比例被压缩。
cs.CV / 42 / 2610.09363
Closing the Loop on Contrail Avoidance with Satellite Verification
通过卫星验证关闭航迹避免的循环
Abstract
Contrails are the thin ice clouds that aircraft leave behind. They cause a large share of aviation's warming, and rerouting the few flights that produce them could avoid much of it. However, an avoided contrail only counts if a satellite can confirm that it never formed, and this check is hard: contrails are one to two pixels wide, cover only 0.18% of pixels, and look very similar to natural cirrus. We build a small diffusion model (8.4M parameters, trained on one GPU) that detects them, and we run a controlled study to find out which components matter. The model reaches 0.476 PR-AUC, compared with 0.414 for a DeepLabV3+ baseline and 0.119 for an adapted MedSegDiff. Doubling the input resolution of the CNN brings it to parity (0.499, p=0.07). Three lessons apply beyond contrails. First, check the input resolution before designing a new architecture. Second, simple flips and rotations more than double accuracy and matter more than any architectural choice we measured. Third, pretraining the model on contrail shapes is harmful: the model learns that thin strokes appear everywhere and paints them onto empty scenes. Precision collapses to 1% while recall-based metrics still rate the degraded model as excellent, and no threshold or guidance heuristic repairs this failure.
Chinese Translation
航迹是飞机留下的细冰云。它们在航空业的变暖中占据了很大一部分，重新规划少数产生航迹的航班可以避免大部分影响。然而，只有在卫星能够确认航迹确实没有形成时，避免的航迹才算有效，而这一检查非常困难：航迹宽度为一到两个像素，仅覆盖0.18%的像素，并且与自然卷云非常相似。我们构建了一个小型扩散模型（8.4M参数，在一台GPU上训练），用于检测航迹，并进行了一项控制研究以找出哪些组件是重要的。该模型的PR-AUC达到了0.476，而DeepLabV3+基线为0.414，调整后的MedSegDiff为0.119。将CNN的输入分辨率翻倍使其达到了平衡（0.499，p=0.07）。有三个教训适用于航迹之外的情况。首先，在设计新架构之前检查输入分辨率。其次，简单的翻转和旋转使准确率翻倍，且比我们测量的任何架构选择更为重要。第三，在航迹形状上对模型进行预训练是有害的：模型学习到细线条无处不在，并将其绘制到空白场景中。精确度降至1%，而基于召回的指标仍然将退化模型评为优秀，且没有任何阈值或指导启发式能够修复这一失败。
cs.CV / 43 / 2610.09376
Unified Multi-plane Autoregressive Diffusion for 3D Multi-contrast MRI Synthesis
统一的多平面自回归扩散用于三维多对比度MRI合成
Abstract
Acquiring a complete set of magnetic resonance imaging (MRI) contrasts is time-intensive and uncomfortable for patients, despite the diagnostic value of multi-contrast imaging. This motivates synthesizing missing contrasts from those already acquired, which is an inherently 3D problem requiring anatomical coherence across axial, sagittal, and coronal planes. However, fully 3D generative models are often impracti- cal under computational resources that scale cubically with volume size. We propose a unified Multi-Plane Autoregressive Diffusion (MPAD), a latent diffusion framework that achieves full-volume 3D synthesis using efficient plane-wise 2D operations while preserving volumetric coherence. A 3D autoencoder first compresses MRI scans into an isotropic 3D la- tent representation. A 2D diffusion model is then trained to reconstruct masked latent slices of the target contrast, conditioned on both source- contrast slices and unmasked target-contrast slices. During inference, we introduce plane-wise autoregressive synthesis with inter-plane priors. Slices are generated autoregressively in random order within one plane orientation to maintain intra-plane continuity, then propagated as con- ditioning priors to orthogonal plane orientations to enforce inter-plane consistency. Compared to 3D latent diffusion baselines, MPAD reduces training and inference FLOPs by 7x and 3x, respectively, while also lowering inference time and peak memory consumption. Experiments on multiple datasets demonstrate that MPAD achieves superior perfor- mance, generating high-fidelity 3D volumes and supporting one-to-many translation within a single unified model.
Chinese Translation
尽管多对比度成像具有重要的诊断价值，但获取完整的磁共振成像（MRI）对比度集既耗时又让患者感到不适。这促使我们从已获取的对比度中合成缺失的对比度，这本质上是一个三维问题，需要在轴向、矢状面和冠状面之间保持解剖一致性。然而，完全的三维生成模型在计算资源上往往不切实际，因为其计算复杂度与体积大小的立方成正比。我们提出了一种统一的多平面自回归扩散（Multi-Plane Autoregressive Diffusion, MPAD）框架，这是一种潜在扩散框架，通过高效的平面级二维操作实现全体积三维合成，同时保持体积一致性。首先，三维自编码器将MRI扫描压缩为各向同性的三维潜在表示。然后，训练一个二维扩散模型以重建目标对比度的掩蔽潜在切片，条件是源对比度切片和未掩蔽的目标对比度切片。在推理过程中，我们引入了平面级自回归合成和跨平面先验。切片在一个平面方向内以随机顺序自回归生成，以保持平面内连续性，然后作为条件先验传播到正交平面方向，以确保平面间一致性。与三维潜在扩散基线相比，MPAD在训练和推理的浮点运算次数（FLOPs）分别减少了7倍和3倍，同时降低了推理时间和峰值内存消耗。在多个数据集上的实验表明，MPAD实现了优越的性能，生成高保真度的三维体积，并支持在单一统一模型内的一对多转换。
cs.CV / 44 / 2610.09383
PPCAR-Net: Projection-Refined Parametric 3D Coronary Artery Reconstruction from Sparse X-ray Angiographic Views
PPCAR-Net：基于投影精炼的稀疏X射线血管造影视图的参数化三维冠状动脉重建
Abstract
Sparse-view 3D coronary reconstruction commonly relies on cross-view correspondence and triangulation, which are vulnerable to vessel overlap and foreshortening, or on volumetric prediction followed by vascular-graph extraction, which does not directly provide centrelines and radii. We introduce PPCAR-Net, a projection-refined parametric coronary artery reconstruction network that directly predicts a branch-structured centreline-and-radius representation without explicit point matching, triangulation, or an intermediate volume. Given a variable number of segmented views, a coarse predictor combines frozen VGGT features with learned branch queries to estimate branch presence, B-spline centreline trajectories, and dense radius profiles. Projection-guided geometry and radius refiners then sample local evidence from the input views and apply residual corrections learned with 3D supervision. We evaluate representation fidelity and sparse-view reconstruction quantitatively and qualitatively. On simulated angiographic masks generated from CT-derived coronary anatomy, PPCAR-Net produces better connected artery reconstructions and achieves strong centreline accuracy, particularly for RCA, while maintaining competitive volumetric overlap. Coarse-to-fine inference takes 121 ms, enabling real-time reconstruction.
Chinese Translation
稀疏视图的三维冠状动脉重建通常依赖于视图间对应关系和三角测量，这些方法容易受到血管重叠和缩短的影响，或者依赖于体积预测后进行血管图提取，这种方法并不能直接提供中心线和半径。我们提出了PPCAR-Net，这是一种基于投影精炼的参数化冠状动脉重建网络，能够直接预测分支结构的中心线和半径表示，而无需显式的点匹配、三角测量或中间体积。给定可变数量的分割视图，粗略预测器结合冻结的VGGT特征与学习的分支查询，以估计分支存在、B样条中心线轨迹和稠密半径轮廓。投影引导的几何和半径精炼器随后从输入视图中采样局部证据，并应用通过三维监督学习的残差修正。我们对表示的保真度和稀疏视图重建进行了定量和定性评估。在从CT导出的冠状动脉解剖结构生成的模拟血管造影掩模上，PPCAR-Net生成了更好连接的动脉重建，并在中心线准确性方面表现出色，特别是对于右冠状动脉（RCA），同时保持了竞争性的体积重叠。粗到细的推理耗时121毫秒，使实时重建成为可能。
cs.CV / 45 / 2610.09392
Quantifying Volumetric Risk: Class-Aware Asymmetric Weighted Conformal Prediction for 3D Medical Image Segmentation
量化体积风险：面向类别的非对称加权保形预测在3D医学图像分割中的应用
Abstract
Reliable volumetric segmentation is critical for clinical diagnostics, yet foundation models such as MedSAM remain deterministic and lack calibrated uncertainty under distribution shift. Existing conformal prediction methods offer statistical guarantees but are frequently applied in 2D and assume symmetric error distributions, so they do not capture the class-specific biases that arise in 3D multi-class segmentation. We propose Class-Aware Asymmetric Weighted Conformal Prediction (CA-WCP), which combines latent-space density-ratio weighting for covariate shift with directional quantiles for the lower and upper volume bounds, and scales each bound by a class-specific asymmetry factor derived from validation-set false-positive and false-negative rates. We prove that CA-WCP retains the weighted-exchangeability marginal coverage guarantee for every class, and we evaluate it on 3D brain tumor segmentation (BraTS 2020) and on a synthetic multi-organ CT benchmark constructed under covariate shift. On both benchmarks the 95\% Clopper--Pearson interval for the observed coverage of CA-WCP contains the nominal 90\% level for every semantic class, while interval width is reduced by 8--14\% relative to symmetric weighted conformal prediction. We further encode the calibrated intervals into structured prompts for a multimodal large language model to produce uncertainty-conditioned radiology reports, linking distribution-shift-aware uncertainty quantification to interpretable clinical communication.
Chinese Translation
可靠的体积分割对于临床诊断至关重要，但像MedSAM这样的基础模型仍然是确定性的，并且在分布变化下缺乏校准的不确定性。现有的保形预测方法提供了统计保证，但通常应用于2D，并假设误差分布是对称的，因此无法捕捉在3D多类别分割中出现的类别特定偏差。我们提出了面向类别的非对称加权保形预测（Class-Aware Asymmetric Weighted Conformal Prediction，CA-WCP），该方法结合了潜在空间密度比加权以应对协变量变化，并使用方向性分位数来确定下限和上限体积边界，并通过从验证集的假阳性和假阴性率中导出的类别特定非对称因子来缩放每个边界。我们证明了CA-WCP为每个类别保留了加权可交换边际覆盖保证，并在3D脑肿瘤分割（BraTS 2020）和在协变量变化下构建的合成多脏器CT基准上进行了评估。在这两个基准上，CA-WCP的观察覆盖的95% Clopper--Pearson区间包含每个语义类别的名义90%水平，同时相对于对称加权保形预测，区间宽度减少了8%至14%。我们进一步将校准区间编码为结构化提示，以便多模态大型语言模型生成不确定性条件的放射学报告，将分布变化感知的不确定性量化与可解释的临床沟通联系起来。
cs.CV / 46 / 2610.09397
One Frame, Full Heartbeat: ECG-Free Cardiac Cine MRI Synthesis via Phase-Conditioned Flow Matching
单帧，全心跳：通过相位条件流匹配实现无心电图的心脏电影磁共振成像合成
Abstract
Cine cardiovascular magnetic resonance (CMR) analysis relies on multi-frame sequences capturing the full cardiac cycle. However, standard multi-frame acquisition depends heavily on electrocardiogram (ECG) gating and repeated breath-holds, posing challenges in uncooperative populations, resource-limited settings, and temporally corrupted datasets. Existing methods that synthesize full cardiac sequences either rely on explicit ECG signals to parameterize myocardium function, or employ deformable registration without physiological constraints, failing to faithfully reproduce clinically relevant dynamic metrics such as ejection fraction (EF) and ventricular contraction magnitude. We present PhaseFlow, a unified generative framework that overcomes both limitations. PhaseFlow estimates a non-linear cardiac phase signal directly from the input sequence via a segmentation-derived left-ventricular (LV) area curve, capturing the asymmetric dynamics of systole and diastole without any ECG dependency. At inference, this phase signal is provided by a pathology-specific template, informing phase-specific frame generation. A rectified flow model conditioned on the phase and slice position synthesizes the full cardiac motion trajectory in the latent space, decoded into a diffeomorphic displacement field that warps end-diastole pixel intensities directly, eliminating the reconstruction blur often accompanying the variational autoencoder. On the ACDC benchmark, PhaseFlow achieves superior physiological fidelity and image realism, with best LV volume curve $R^2$, structural similarity (SSIM) and generative quality (FID) among all baselines. Ablation studies confirm that each proposed component contributes measurably to the overall performance.
Chinese Translation
心血管电影磁共振（CMR）分析依赖于捕捉完整心脏周期的多帧序列。然而，标准的多帧采集在很大程度上依赖于心电图（ECG）门控和重复的屏气，这在不配合的患者群体、资源有限的环境和时间受损的数据集中带来了挑战。现有合成完整心脏序列的方法要么依赖于显式的ECG信号来参数化心肌功能，要么采用没有生理约束的可变形配准，未能真实再现临床相关的动态指标，如射血分数（EF）和心室收缩幅度。我们提出了PhaseFlow，一个统一的生成框架，克服了这两种局限。PhaseFlow通过分割导出的左心室（LV）面积曲线直接从输入序列估计非线性的心脏相位信号，捕捉收缩期和舒张期的不对称动态，而无需任何ECG依赖。在推理时，该相位信号由特定病理的模板提供，指导相位特定的帧生成。一个基于相位和切片位置的修正流模型在潜在空间中合成完整的心脏运动轨迹，并解码为一个光滑的位移场，直接扭曲舒张末期的像素强度，消除通常伴随变分自编码器的重建模糊。在ACDC基准测试中，PhaseFlow在生理逼真度和图像真实感方面表现优越，在所有基线中获得最佳的LV体积曲线$R^2$、结构相似性（SSIM）和生成质量（FID）。消融研究确认每个提出的组件对整体性能都有显著贡献。
cs.CV / 47 / 2610.09408
TiTok: Audio-Visual LLM for Multi-Segment Temporal Grounding
TiTok：用于多段时间定位的音视频大语言模型
Abstract
Audio-visual multi-segment grounding (AV-MSG) in untrimmed videos, reasoning over audio-visual evidence and predicting multiple segments for a query, is a fundamental problem but remains challenging. Visual-only models overlook complementary acoustic cues, while audio-visual models often fail to calibrate the number of events - a phenomenon we refer to as count miscalibration. We present TiTok, an audio-visual large language model (AV-LLM) that localizes an arbitrary number of temporal event segments for each query. For precise boundary prediction, we introduce the Time Token Interleaving (TTI) method, which explicitly injects special time tokens into the audio-visual stream to align input-side temporal perception with output-side temporal prediction. We further propose decoupled, multi-segment-oriented rewards for reinforcement learning, consisting of global, local, count, precision, and format rewards, optimized with Group reward-Decoupled Normalization Policy Optimization (GDPO). To assess the performance on AV-MSG, we establish a new UnAV-100-based evaluation protocol, and propose the CountF1 metric for quantifying count miscalibration that overlap metrics fail to capture. TiTok reaches 65.7 mIoU and 0.58 CountF1, achieving state-of-the-art performance. Our code is available at this link.
Chinese Translation
在未经裁剪的视频中，音视频多段定位（AV-MSG）是一个基本但具有挑战性的问题，它需要对音视频证据进行推理并为查询预测多个段落。仅视觉模型忽视了互补的声学线索，而音视频模型往往无法校准事件的数量——我们称之为计数失调现象。我们提出了TiTok，一种音视频大语言模型（AV-LLM），能够为每个查询定位任意数量的时间事件段。为了实现精确的边界预测，我们引入了时间令牌交错（Time Token Interleaving, TTI）方法，该方法将特殊的时间令牌显式注入音视频流中，以对齐输入端的时间感知与输出端的时间预测。此外，我们还提出了解耦的、多段导向的强化学习奖励，包括全局奖励、局部奖励、计数奖励、精度奖励和格式奖励，并采用群体奖励-解耦归一化策略优化（Group reward-Decoupled Normalization Policy Optimization, GDPO）进行优化。为了评估AV-MSG的性能，我们建立了一种基于UnAV-100的新评估协议，并提出了CountF1指标，用于量化计数失调，而重叠指标无法捕捉。TiTok达到了65.7的mIoU和0.58的CountF1，取得了最先进的性能。我们的代码可在此链接获取。
cs.CV / 48 / 2610.09417
trACT: temporal revelation Airborne Camera Trap
trACT：时域揭示空中相机捕捉器
Abstract
Effective remote monitoring and surveillance using drones are frequently impeded by severe environmental and thermal clutter, dynamic vegetation, target camouflage, and system latency. Drawing inspiration from the hunting strategies of birds of prey that hover and stabilize their vision to isolate subtle ground motion, we introduce trACT (temporal revelation Airborne Camera Trap), a lightweight, real-time aerial robotics framework designed for autonomous consumer drones. The system integrates Temporal Max Pooling (TMP), a low-level signal processing method that transforms imperceptible movement across a rolling integration window into robust value and time encodings, with self-supervised motion anomaly detection to isolate target motion from background environmental motion caused by wind gusts and drone drift. To overcome mechanical and processing delays, trACT combines motion prediction with automated gimbal-stabilized optical zoom verification and equitable multi-target verification balancing. Extensive real-world field experiments in densely forested wildlife habitats and surveillance scenarios demonstrate that trACT successfully bridges the gap between wide-area aerial monitoring and precise, autonomous target verification under challenging operational conditions.
Chinese Translation
有效的无人机远程监控和监视常常受到严峻的环境和热干扰、动态植被、目标伪装以及系统延迟的影响。我们受到猛禽在悬停时稳定视野以隔离微妙地面运动的捕猎策略的启发，提出了trACT（时域揭示空中相机捕捉器），这是一个为自主消费级无人机设计的轻量级实时空中机器人框架。该系统集成了时域最大池化（Temporal Max Pooling, TMP），这是一种低级信号处理方法，能够将滚动积分窗口内的微小运动转化为稳健的值和时间编码，并结合自监督运动异常检测，从而将目标运动与由风速变化和无人机漂移引起的背景环境运动隔离开来。为了克服机械和处理延迟，trACT结合了运动预测与自动云台稳定的光学变焦验证，以及公平的多目标验证平衡。在密集森林野生动物栖息地和监视场景中的广泛实地实验表明，trACT成功地弥合了宽域空中监控与在挑战性操作条件下的精确自主目标验证之间的差距。
cs.CV / 49 / 2610.09418
Spatial Latent Reasoning for Embodied Reference Understanding
用于体现参考理解的空间潜在推理
Abstract
Pointing-gesture visual grounding requires connecting hand geometry with the visual identity and extent of a referred object. A central challenge for continuous latent reasoning is how to organize these complementary cues into useful intermediate supervision. We propose Spatial Latent Reasoning (SLR), a framework that structures this supervision around an ordered sequence of geometric and visual states. A spatial ray state is supervised by fingertip position and pointing direction, followed by four states aligned with target-region features. To construct the visual targets, we introduce parity pooling, which applies polyphase grouping to average region tokens on four interleaved spatial supports. All states are generated recurrently during training and inference; auxiliary annotations are required only during training. On EgoPoint-Ground, the framework improves mIoU over same-backbone supervised fine-tuning by 2.8, 17.5, and 21.1 percentage points on Qwen3.5-4B, Qwen2.5-VL-7B, and Qwen3-VL-8B, respectively, with improvements on both hard subsets. On YouRefIt, it achieves 77.6% precision at IoU 0.5, a numerical margin of 5.2 percentage points over the reported state of the art under differing evaluation protocols. Ablations support joint geometric and visual supervision on the standard and similar-object sets, and favor parity over three alternative pooling operators on the standard set. These results support task-structured supervision for continuous pointing grounding. We will release the code and supporting materials.
Chinese Translation
指向手势的视觉定位需要将手部几何形状与所指对象的视觉特征和范围相连接。连续潜在推理的一个核心挑战是如何将这些互补线索组织成有用的中间监督。我们提出了空间潜在推理（Spatial Latent Reasoning, SLR）框架，该框架围绕几何和视觉状态的有序序列构建这种监督。空间射线状态由指尖位置和指向方向进行监督，随后是与目标区域特征对齐的四个状态。为了构建视觉目标，我们引入了奇偶池化（parity pooling），该方法对四个交错空间支撑上的区域标记进行多相分组平均。所有状态在训练和推理过程中都是递归生成的；辅助注释仅在训练期间需要。在EgoPoint-Ground数据集上，该框架在Qwen3.5-4B、Qwen2.5-VL-7B和Qwen3-VL-8B上分别提高了2.8、17.5和21.1个百分点的mIoU，相较于相同骨干网络的监督微调，并且在两个困难子集上均有改善。在YouRefIt上，它在IoU 0.5时达到了77.6%的精度，相较于不同评估协议下报告的最新成果，数值上提高了5.2个百分点。消融实验支持在标准和相似对象集上进行联合几何和视觉监督，并在标准集上更倾向于奇偶池化而非三种替代池化操作。以上结果支持针对连续指向定位的任务结构化监督。我们将发布代码和支持材料。
cs.CV / 50 / 2610.09427
Event-Aligned Visual Action Reasoning for World Action Models
事件对齐的视觉行动推理用于世界行动模型
Abstract
World-Action Models (WAMs) utilize future visual prediction as an intermediate reasoning process to guide action generation. However, existing WAMs typically structure visual imagination according to predefined temporal intervals, without explicitly accounting for the different roles of task-critical interactions and connecting transitions. We argue that effective visual foresight should align directly with task-relevant interactions and their corresponding reasoning demands. To this end, we introduce an event-aligned visual action reasoning framework that organizes visual-action prediction around interaction events. Through event-aligned visual-action supervision, WAM learns to generate event-aligned visual context in each imagined rollout, placing greater emphasis on critical state changes that inform action generation. This shapes the visual reasoning granularity according to the underlying interaction dynamics, with detailed reasoning around task-critical events and coarser progression through connecting transitions. Furthermore, we introduce an execution validity head that identifies the valid portion of each predicted action sequence, avoiding redundant actions during chunked inference. Experiments demonstrate a 10.26 percentage point improvement in DOMINO success rate over baseline and competitive performance on RoboTwin 2.0. It also transfers from DOMINO Level 1 to Levels 2 and 3 without target-level adaptation.
Chinese Translation
世界行动模型（World-Action Models, WAMs）利用未来视觉预测作为中间推理过程来指导行动生成。然而，现有的WAMs通常根据预定义的时间间隔结构化视觉想象，而未明确考虑任务关键交互和连接过渡的不同角色。我们认为，有效的视觉前瞻应直接与任务相关的交互及其相应的推理需求对齐。为此，我们提出了一种事件对齐的视觉行动推理框架，该框架围绕交互事件组织视觉-行动预测。通过事件对齐的视觉-行动监督，WAM能够在每个想象的展开中生成事件对齐的视觉上下文，更加重视那些影响行动生成的关键状态变化。这使得视觉推理的粒度根据潜在的交互动态进行调整，对任务关键事件进行详细推理，而通过连接过渡进行较粗略的推理。此外，我们引入了一个执行有效性头（execution validity head），用于识别每个预测行动序列的有效部分，从而在分块推理过程中避免冗余行动。实验表明，DOMINO成功率较基线提高了10.26个百分点，并在RoboTwin 2.0上表现出竞争力的性能。同时，它还能够从DOMINO Level 1转移到Level 2和Level 3，而无需目标级别的适应。
cs.CV / 51 / 2610.09430
SkillCycle: Co-Evolving Agent Policies and Skill Banks
技能循环：共同进化的智能体策略与技能库
Abstract
Internalizing external skills changes a language agent's capabilities and, with them, the value of its remaining guidance: rules can become redundant, misleading, or insufficient for newly encountered decisions. This creates a coupled problem of learning from skills and adapting the skills that supervise further learning. We introduce SkillCycle, a framework for co-evolving agent policies and skill banks through a feedback loop between skill internalization and rule revision. Our central contribution is to give distillation feedback a second role: token-level contextual differences help locate rules for inspection, while interaction outcomes guide edits to their content and applicability. SkillCycle alternates between two phases: policy learning with a fixed skill bank and router, and rule revision with a frozen policy. Candidate edits undergo rule-level and whole-bank environment comparisons before they guide the next learning cycle. On WebShop, SkillCycle with a 3B model achieves a success rate of 74.74% and a score of 88.37 without inference-time skill inputs, representing relative improvements of 0.73% and 3.96% over the state-of-the-art (SOTA) model, respectively. In Cycle 3 ablations on ALFWorld and WebShop, SkillCycle's no-skill success rates improve by 10.18% and 18.11% relative to a static skill bank, and by 2.41% and 2.50% relative to a single bank update, respectively. These results show that continually revising skill guidance as the agent's capabilities change helps transform external skills into policy capabilities that require no skill inputs at inference. We will release code, configurations, skill banks, and evaluation protocols.
Chinese Translation
内化外部技能改变了语言智能体的能力，并因此影响其剩余指导的价值：规则可能变得多余、误导或对新遇到的决策不足。这造成了一个学习技能与调整监督进一步学习的技能之间的耦合问题。我们提出了技能循环（SkillCycle），这是一个通过技能内化与规则修订之间的反馈循环，共同进化智能体策略和技能库的框架。我们的核心贡献是赋予蒸馏反馈第二个角色：令牌级上下文差异有助于定位待检查的规则，而交互结果则指导其内容和适用性的编辑。技能循环在两个阶段之间交替：使用固定技能库和路由器的策略学习，以及使用冻结策略的规则修订。候选编辑在引导下一个学习周期之前，经历规则级和整个技能库环境的比较。在WebShop上，使用3B模型的技能循环实现了74.74%的成功率和88.37的得分，且无需推理时的技能输入，分别相对于最先进模型（SOTA）实现了0.73%和3.96%的相对提升。在ALFWorld和WebShop的Cycle 3消融实验中，技能循环的无技能成功率相对于静态技能库分别提高了10.18%和18.11%，相对于单次技能库更新则提高了2.41%和2.50%。这些结果表明，随着智能体能力的变化，持续修订技能指导有助于将外部技能转化为在推理时无需技能输入的策略能力。我们将发布代码、配置、技能库和评估协议。
cs.CV / 52 / 2610.09438
Controllable Crowd Generation through World-Model Planning
通过世界模型规划实现可控人群生成
Abstract
Crowd simulation plays a central role in robot navigation, autonomous driving, and urban planning. For these applications, realistic simulation requires crowds to adapt their behavior to environmental changes and user objectives. However, existing methods that rely on predefined control settings have limited flexibility in accommodating new user-specified objectives. To address this limitation, we propose Ctrl-CWM, a multi-agent Controllable Crowd World Model that integrates crowd generation and run-time control. Our key idea is to adapt the world-model principle of planning using imagined futures to crowd simulation. To this end, Ctrl-CWM consists of an encoder that learns a representation of human motion dynamics, an actor that proposes pedestrian displacements, a critic that evaluates imagined crowd trajectories, and a planner that selects actions. We first learn human motion dynamics through trajectory prediction on real-world pedestrian videos and then freeze the encoder to preserve them. Using this representation, the actor generates imagined crowd trajectories through repeated state updates, and the planner combines the critic's scores with user costs to select actions. Repeated planning advances the simulated crowd, while additional user costs introduce new control objectives without retraining. We extensively evaluate crowd generation under varied agent arrival conditions and run-time control across avoidance and attraction scenarios. Ctrl-CWM outperforms the state-of-the-art method on most crowd realism and collision metrics, and adapts crowd behaviors to user-specified objectives introduced during simulation. The project page is available at https://jungyu0413.github.io/Ctrl-CWM
Chinese Translation
人群模拟在机器人导航、自动驾驶和城市规划中发挥着核心作用。对于这些应用，现实的模拟要求人群能够根据环境变化和用户目标调整其行为。然而，现有依赖于预定义控制设置的方法在适应新的用户指定目标方面灵活性有限。为了解决这一局限性，我们提出了Ctrl-CWM，一个多智能体可控人群世界模型，集成人群生成和实时控制。我们的关键思想是将使用想象未来的世界模型规划原则应用于人群模拟。为此，Ctrl-CWM由一个学习人类运动动态表示的编码器、一个提出行人位移的执行者、一个评估想象人群轨迹的评论者和一个选择行动的规划者组成。我们首先通过对真实世界行人视频的轨迹预测学习人类运动动态，然后冻结编码器以保留这些动态。利用这一表示，执行者通过重复状态更新生成想象的人群轨迹，而规划者将评论者的评分与用户成本结合以选择行动。重复规划推动模拟人群的前进，而额外的用户成本在不重新训练的情况下引入新的控制目标。我们在不同的代理到达条件下广泛评估人群生成，并在避让和吸引场景中进行实时控制。Ctrl-CWM在大多数人群真实感和碰撞指标上优于现有最先进的方法，并能够将人群行为适应于模拟过程中引入的用户指定目标。项目页面可访问：https://jungyu0413.github.io/Ctrl-CWM
cs.CV / 53 / 2610.09439
InscriptionOCR: A Dataset and Method for Understanding Inscriptions
InscriptionOCR：理解铭文的数据集和方法
Abstract
Ancient script image restoration is a fundamental problem in computer vision, as it directly affects the reliable analysis and interpretation of historical documents and inscriptions. Ashokan Brahmi is an ancient script extensively used during the reign of Emperor Ashoka in the 3rd century BC, primarily for inscriptions in Prakrit. These inscriptions, including major and minor rock and pillar edicts, constitute a valuable yet largely unexplored source of data for computational analysis. The degraded nature of inscription imagery and the lack of standardized digital resources pose significant challenges for automated processing. We present an end-to-end AI-based framework for understanding ancient inscriptions that encompasses image enhancement, optical character recognition (OCR), transliteration, and neural machine translation (NMT). The proposed pipeline processes low-quality images captured directly from stone inscriptions, performs image restoration and Brahmi script character recognition, maps the recognized characters to the Roman script, and finally translates the resulting Prakrit text into English. We also introduce two new datasets: (i) InscriptionOCR Dataset: the largest publicly usable digital OCR dataset for Brahmi script to date, consisting of over 200,000 character images across about 600 classes, and (ii) a bilingual Prakrit-English parallel corpus comprising over 2,000 sentence pairs for NMT. We believe that the proposed framework and datasets will facilitate future research in ancient script analysis, low-resource OCR, and digital epigraphy.
Chinese Translation
古代文字图像修复是计算机视觉中的一个基本问题，因为它直接影响到历史文献和铭文的可靠分析与解读。阿育王梵文（Ashokan Brahmi）是一种古代文字，广泛使用于公元前3世纪阿育王统治时期，主要用于普拉克里特语（Prakrit）的铭文。这些铭文，包括主要和次要的岩石及柱子法令，构成了一个宝贵但尚未充分开发的计算分析数据源。铭文图像的退化特性以及缺乏标准化的数字资源为自动处理带来了重大挑战。我们提出了一个基于人工智能的端到端框架，用于理解古代铭文，该框架涵盖图像增强、光学字符识别（OCR）、音译和神经机器翻译（NMT）。所提议的流程处理直接从石刻铭文拍摄的低质量图像，进行图像修复和梵文字母识别，将识别出的字符映射到罗马字母，最后将生成的普拉克里特文本翻译成英语。我们还介绍了两个新的数据集：（i）InscriptionOCR 数据集：迄今为止最大的可公开使用的梵文字母数字OCR数据集，包含超过200,000个字符图像，涵盖约600个类别，以及（ii）一个双语普拉克里特语-英语平行语料库，包含超过2,000对句子用于NMT。我们相信，所提出的框架和数据集将促进未来在古代文字分析、低资源OCR和数字铭文学方面的研究。
cs.CV / 54 / 2610.09440
Mixture of Layers: Dynamic Layer Routing for Visual Reasoning
层的混合：用于视觉推理的动态层路由
Abstract
Pre-trained vision encoders contain layer-wise visual representations that differ in spatial granularity, semantic abstraction, and sensitivity to local details. However, most Multimodal Large Language Models (MLLMs) rely on only the final or penultimate vision encoder representations or fixed aggregation rules, making visual abstraction largely query-agnostic and limiting access to fine-grained cues such as small objects, spatial details, text, and subtle visual attributes. In this work, we propose Mixture of Layers (MoL), an instruction-conditioned layer routing approach at the visual patch level that dynamically aggregates query-relevant latent representations from intermediate vision encoder layers. Given a text query, MoL predicts routing probabilities over vision encoder layers and performs a top-k sparse aggregation over selected hidden states at either the image level, patch level, or through a hybrid routing mechanism. In doing so, MoL enables query-adaptive access to layer-specific visual features for fine-grained visual reasoning. Our experiments across 7 fine-grained visual reasoning tasks demonstrate substantial performance improvements, especially across fine-grained visual grounding and understanding tasks such as +18.9% improvement on V* in overall accuracy, +4.5% on HRBench4K, and +16.3% on CharXiv compared to the baseline MLLMs, without resorting to multi-resolution inputs, simple interleaving of multiple vision encoders, or increasing the number of patch tokens. We study vision encoders' receptive field scales across different layers and their sampling behaviors to provide an in-depth analysis of why layer-wise sampling is helpful, demonstrating that conditional visual representations are a key step towards better visual perception and reasoning in MLLMs. Our project page is available at https://wjdghks950.github.io/mol.github.io/.
Chinese Translation
预训练的视觉编码器包含在空间粒度、语义抽象和对局部细节敏感性方面各不相同的层级视觉表示。然而，大多数多模态大型语言模型（MLLMs）仅依赖于最终或倒数第二个视觉编码器的表示或固定的聚合规则，使得视觉抽象在很大程度上与查询无关，并限制了对小物体、空间细节、文本和微妙视觉属性等细粒度线索的访问。在本研究中，我们提出了层的混合（Mixture of Layers, MoL），这是一种在视觉补丁级别的指令条件层路由方法，能够动态聚合来自中间视觉编码器层的与查询相关的潜在表示。给定一个文本查询，MoL预测视觉编码器层的路由概率，并在图像级、补丁级或通过混合路由机制对所选隐藏状态进行 top-k 稀疏聚合。通过这种方式，MoL 实现了对层特定视觉特征的查询自适应访问，以进行细粒度视觉推理。我们在 7 个细粒度视觉推理任务上的实验表明，性能有显著提升，特别是在细粒度视觉定位和理解任务上，与基线 MLLMs 相比，整体准确率提高了 +18.9%（在 V* 上）、+4.5%（在 HRBench4K 上）和 +16.3%（在 CharXiv 上），且无需采用多分辨率输入、简单交错多个视觉编码器或增加补丁令牌的数量。我们研究了不同层级视觉编码器的感受野尺度及其采样行为，以提供深入分析，解释为何层级采样是有益的，证明条件视觉表示是实现 MLLMs 更好视觉感知和推理的关键步骤。我们的项目页面可访问 https://wjdghks950.github.io/mol.github.io/.
cs.CV / 55 / 2610.09441
TIRA: Tumor Immune Representation Adaptation for Zero-Shot Cross-Cancer MSI and TMB Prediction
TIRA：肿瘤免疫表征适应用于零样本跨癌症的微卫星不稳定性和肿瘤突变负荷预测
Abstract
Microsatellite instability-high (MSI-H) and high tumor mutational burden (TMB-H) are clinically relevant biomarkers, yet their histopathological prediction remains challenging when models are transferred across morphologically distinct cancer types. Immune-associated spatial patterns can persist across cancers despite these morphological differences, but foundation-model-based predictors trained on a single cancer do not explicitly use this information, limiting cross-cancer generalization. To address this limitation, we propose TIRA (Tumor Immune Representation Adaptation), a target-free framework that refines frozen foundation-model representations using spatial immune topology, without requiring target-domain data during model development or test-time adaptation. TIRA uses a topology-supervised biology representation to condition tile-level attention while pooling only morphological features for joint MSI and TMB prediction. We train TIRA on TCGA-COAD+READ and evaluate it zero-shot on CPTAC-COAD, TCGA-STAD, TCGA-UCEC, and CPTAC-UCEC, covering cross-site, cross-cancer, and combined cross-cancer-site distribution shifts under UNI2, CONCH, and Virchow2. With UNI2, TIRA improved zero-shot AUROC on TCGA-STAD from 0.633 to 0.766 for MSI and from 0.651 to 0.772 for TMB. Source-derived spatial immune topology improved the cross-cancer robustness of frozen pathology foundation-model representations.
Chinese Translation
微卫星不稳定性高（MSI-H）和高肿瘤突变负荷（TMB-H）是临床相关的生物标志物，但在不同形态学癌症类型之间转移模型时，其组织病理学预测仍然具有挑战性。尽管存在这些形态学差异，免疫相关的空间模式在不同癌症之间仍然可以持续存在，但基于基础模型的预测器在单一癌症上训练时并未明确利用这些信息，从而限制了跨癌症的泛化能力。为了解决这一限制，我们提出了TIRA（肿瘤免疫表征适应），这是一个无目标的框架，通过使用空间免疫拓扑来细化冻结的基础模型表征，而在模型开发或测试时适应过程中不需要目标领域数据。TIRA使用拓扑监督的生物学表征来调节瓷砖级别的注意力，同时仅汇聚形态特征以进行联合的MSI和TMB预测。我们在TCGA-COAD+READ上训练TIRA，并在CPTAC-COAD、TCGA-STAD、TCGA-UCEC和CPTAC-UCEC上进行零样本评估，涵盖了在UNI2、CONCH和Virchow2下的跨站点、跨癌症和组合跨癌症-站点分布转变。在UNI2下，TIRA将TCGA-STAD的零样本AUROC从0.633提高到0.766（MSI），从0.651提高到0.772（TMB）。源导向的空间免疫拓扑提高了冻结病理基础模型表征的跨癌症鲁棒性。
cs.CV / 56 / 2610.09444
LighTROcc: Lightweight 4D Occupancy Forecasting via Instance-Centric 3D Gaussians
LighTROcc：通过实例中心的3D高斯实现轻量级4D占用预测
Abstract
Forecasting future 3D occupancy from surround-view cameras is essential for autonomous driving, yet existing approaches rely on dense voxel or bird's-eye-view representations whose cost grows rapidly with spatial resolution and prediction horizon. Because these representations do not explicitly maintain object identities, they also struggle to preserve instance consistency over time. We present LighTROcc, a lightweight instance-centric framework that represents movable objects with a compact set of learned queries and predicts present and future occupancy in a single forward pass. LighTROcc localizes each query through attention-guided forward lifting, combining image-space cross-attention, query-specific depth, and camera geometry to estimate its 3D center. Each instance is modeled as a mixture of anisotropic 3D Gaussians and propagated across future steps using predicted displacements, producing continuous, temporally consistent occupancy forecasts. Experiments on nuScenes and supplemented nuScenes-Occupancy show that LighTROcc outperforms the evaluated dense and instance-wise baselines in instance-level forecasting accuracy while maintaining strong voxel-level occupancy quality. Across different model configurations, LighTROcc achieves a favorable balance between forecasting accuracy and computational efficiency, demonstrating the potential of compact instance-centric modeling for camera-based 4D occupancy forecasting.
Chinese Translation
从全景摄像头预测未来的3D占用对于自动驾驶至关重要，但现有方法依赖于密集体素或鸟瞰图表示，这些表示的成本随着空间分辨率和预测范围的增加而迅速增长。由于这些表示并未明确维护物体身份，因此在时间上也难以保持实例一致性。我们提出了LighTROcc，一个轻量级的实例中心框架，通过一组紧凑的学习查询表示可移动物体，并在单次前向推理中预测当前和未来的占用。LighTROcc通过注意力引导的前向提升定位每个查询，结合图像空间的交叉注意力、查询特定的深度和相机几何来估计其3D中心。每个实例被建模为各向异性3D高斯的混合体，并通过预测的位移在未来步骤中传播，从而生成连续、时间一致的占用预测。在nuScenes和补充的nuScenes-Occupancy上的实验表明，LighTROcc在实例级预测准确性方面优于评估的密集和实例级基准，同时保持强大的体素级占用质量。在不同的模型配置中，LighTROcc在预测准确性和计算效率之间实现了良好的平衡，展示了基于紧凑实例中心建模的摄像头4D占用预测的潜力。
cs.CV / 57 / 2610.09449
From Global Alignment to Local Grounding: Zero-Shot Chinese Character Recognition with Radical Verification
从全局对齐到局部定位：基于部首验证的零样本中文字符识别
Abstract
Zero-shot Chinese character recognition (ZS-CCR) aims to recognize characters whose categories are never observed during training, and typically relies on the compositional structure shared between seen and unseen characters. Recent CLIP-style methods represent this structure with the Ideographic Description Sequence (IDS) and align it with glyph images in a shared embedding space. However, they rely on a single global image--IDS similarity that discards the spatial layout of radicals and, being learned only implicitly from seen classes, generalizes poorly to unseen ones; moreover, global matching often retrieves the correct character within the top candidates yet fails to rank it first when characters differ only in subtle local radicals. To address these issues, we propose a global-to-local two-stage framework. In the first stage, STG-CLIP augments the IDS with explicit tree-position and radical-level geometric priors, yielding a spatial-aware prototype that provides a consistent spatial description across seen and unseen categories for high-recall global retrieval. In the second stage, the Radical Verification Module (RVM) uses the radical instances of each retrieved candidate as queries to verify whether the corresponding radicals can be matched to spatially compatible regions in the input glyph. A margin-based gating rule activates the RVM only when the leading global candidates receive similar similarity scores. Experiments on the ICDAR2013 benchmark demonstrate that our method achieves state-of-the-art performance under the character-level zero-shot setting, obtaining 83.06% top-1 accuracy with 2,755 seen classes. Ablation studies further show that the explicit geometric priors and radical-level verification provide complementary improvements.
Chinese Translation
零样本中文字符识别（ZS-CCR）旨在识别在训练过程中从未观察到的字符类别，通常依赖于已见字符和未见字符之间共享的组合结构。最近的 CLIP 风格方法通过表意描述序列（Ideographic Description Sequence, IDS）来表示这一结构，并将其与字形图像对齐于共享的嵌入空间。然而，这些方法依赖于单一的全局图像-IDS 相似性，忽略了部首的空间布局，并且由于仅从已见类别中隐式学习，导致对未见类别的泛化能力较差；此外，全局匹配通常能够在前几个候选中检索到正确的字符，但在字符仅在细微的局部部首上有所不同的情况下，未能将其排名第一。为了解决这些问题，我们提出了一种从全局到局部的两阶段框架。在第一阶段，STG-CLIP 通过显式的树位置和部首级几何先验增强 IDS，生成一个空间感知的原型，为已见和未见类别提供一致的空间描述，以实现高召回率的全局检索。在第二阶段，部首验证模块（Radical Verification Module, RVM）使用每个检索候选的部首实例作为查询，以验证相应的部首是否可以与输入字形中的空间兼容区域匹配。当领先的全局候选接收到相似的相似性分数时，基于边际的门控规则激活 RVM。在 ICDAR2013 基准上的实验表明，我们的方法在字符级零样本设置下实现了最先进的性能，获得了 83.06% 的 top-1 准确率，涉及 2,755 个已见类别。消融研究进一步表明，显式几何先验和部首级验证提供了互补的改进。
cs.CV / 58 / 2610.09450
Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning
Iris-3B：超越潜在空间的像素空间扩散训练、转换与微调
Abstract
Pixel-space diffusion models avoid the lossy VAE of latent models, which suggests an advantage on downstream tasks where fine-grained detail matters. We test this claim along both routes to a pixel-space backbone. We pretrain Iris-3B, a 3B-parameter pixel-space text-to-image transformer, from scratch through a $256\to512\to1024$ curriculum, after first ablating the prediction target and representation alignment at $256^2$ to decide what to scale. We also convert a pretrained latent model, FLUX.2 Klein base 4B, to pixel space. We fine-tune both families for monocular depth estimation and for image restoration/super-resolution. We find no significant improvement from using a pixel-space generative prior. Fine-tuned for depth with one matched direct-regression recipe, Iris-3B is level with the latent FLUX.2 Klein and the converted pixel FLUX.2 Klein falls behind it, and on $4\times$ DIV2K restoration neither pixel model beats a latent FLUX.2 Klein fine-tune, the converted one trailing it slightly. We document the recipes, the failure modes and the remaining confounds behind this negative result. Nevertheless, Iris-3B shows that pixel-space pretraining with the pixel-transformer (PiT) head of PixelDiT scales to 3B parameters and to text-to-image quality competitive with latent models, matching Qwen-Image on OneIG under the official evaluators at $1024^2$. We release its weights and training code in the hope that they help pave the way for further work on pixel-space generation.
Chinese Translation
像素空间扩散模型避免了潜在模型的有损变分自编码器（VAE），这在细节至关重要的下游任务中显示出优势。我们沿着两条路径测试这一主张，以建立像素空间基础。我们从头开始预训练Iris-3B，一个具有30亿参数的像素空间文本到图像变换器，通过$256 o512 o1024$的课程，首先在$256^2$上消融预测目标和表示对齐，以决定扩展的内容。我们还将一个预训练的潜在模型FLUX.2 Klein base 4B转换为像素空间。我们对这两类模型进行了微调，以进行单目深度估计和图像恢复/超分辨率。我们发现使用像素空间生成先验并没有显著改善。经过与一个匹配的直接回归方案微调后，Iris-3B的深度估计与潜在的FLUX.2 Klein相当，而转换后的像素FLUX.2 Klein稍微落后于它，在$4 imes$ DIV2K恢复任务中，两个像素模型均未能超越潜在的FLUX.2 Klein微调，转换后的模型稍微落后。我们记录了这些方案、失败模式以及导致这一负面结果的剩余混淆因素。尽管如此，Iris-3B表明，使用PixelDiT的像素变换器（PiT）头进行像素空间预训练可以扩展到30亿参数，并在文本到图像质量上与潜在模型竞争，在官方评估者的$1024^2$下与Qwen-Image相匹配。我们发布了其权重和训练代码，希望能为进一步的像素空间生成工作铺平道路。
cs.CV / 59 / 2610.09455
RLHND: Video Foundation Models as Physically Grounded Hand Trackers for Robot Learning
RLHND：作为物理基础的手部追踪器的视频基础模型在机器人学习中的应用
Abstract
Recently, approaches that leverage human video datasets for robot policy training have become increasingly prevalent. However, most existing hand trackers regress pose from cropped frames with limited priors on hand motion and object interaction, resulting in inaccurate and physically inconsistent estimates. Moreover, the lack of physical cues, e.g., contact and force, limits the use of human videos for robot policy training. To this end, we propose RLHND, a video foundation model-based hand tracking model that jointly estimates hand pose and realistic tactile information from monocular egocentric videos. RLHND turns the pre-trained Cosmos 3 video diffusion backbone into a deterministic clip-level feature extractor via clean-latent conditioning, carrying its learned priors on hand motion and hand-object interaction into tracking. For pose estimation, RLHND (i) predicts hand poses with anatomically plausible joint angles and (ii) enables optional conditioning on the shape parameter to maintain consistent hand shape within the same video and even across videos recorded by the same actor. For tactile estimation, a separate tactile expert stream, trained with the pose stream frozen, predicts dense contact and force over the hand surface. We further adopt LBS-based feature spreading to enable vertex-wise feature extraction without costly per-vertex attention. RLHND achieves state-of-the-art performance across various benchmark datasets for pose estimation, while also achieving state-of-the-art performance in contact and force estimation. Moreover, we demonstrate the utility of RLHND for robot learning through retargeting results and real-world robot experiments. The code will be publicly available at https://seungjun-moon.github.io/rlhnd/.
Chinese Translation
近年来，利用人类视频数据集进行机器人策略训练的方法变得越来越普遍。然而，大多数现有的手部追踪器从裁剪的帧中回归姿态，缺乏对手部运动和物体交互的有限先验，导致估计不准确且在物理上不一致。此外，缺乏物理线索，例如接触和力，限制了人类视频在机器人策略训练中的使用。为此，我们提出了RLHND，一种基于视频基础模型的手部追踪模型，它从单目自我中心视频中联合估计手部姿态和真实的触觉信息。RLHND通过干净潜在条件将预训练的Cosmos 3视频扩散骨干网络转变为确定性的片段级特征提取器，将其在手部运动和手物体交互方面的学习先验融入追踪中。在姿态估计方面，RLHND (i) 预测具有解剖学上合理的关节角度的手部姿态，并且 (ii) 允许对形状参数进行可选条件，以在同一视频内甚至在同一演员录制的不同视频之间保持一致的手部形状。在触觉估计方面，单独的触觉专家流在冻结姿态流的情况下进行训练，预测手部表面的密集接触和力。我们进一步采用基于LBS的特征传播，以实现无需昂贵的逐顶点注意力的逐顶点特征提取。RLHND在各种基准数据集上实现了姿态估计的最先进性能，同时在接触和力估计方面也达到了最先进的表现。此外，我们通过重定向结果和现实世界的机器人实验展示了RLHND在机器人学习中的实用性。代码将公开发布在 https://seungjun-moon.github.io/rlhnd/。
cs.CV / 60 / 2610.09459
An Invariant Tangent-Angle Descriptor and a Band U-Net for 2D Fragment Adjacency Prediction
一种不变的切线角描述符和用于二维碎片邻接预测的带状 U-Net
Abstract
This paper addresses the prediction of adjacency between pairs of 2D fragments based on their contours. We improved the two-stage architecture proposed in Beaulac's thesis, in which a rotation-equivariant Siamese convolutional neural network scores pairs of local image windows along the two contours of two fragments. The scores are gathered in an adjacency matrix in which a ResNet detects the partial anti-diagonal band that reveals the adjacency of two fragments. In the current work, we keep the pipeline and replace the local score by a comparison of tangent-angle profiles of contour windows, making it, by construction, invariant to fragment rotation and agnostic to the selected contour-starting point. These adaptations may be either a training-free likelihood ratio or a small one-dimensional convolutional model trained on corresponding points. We also replaced the final classifier by a band U-Net that segments the band and classifies the pair, so that the shared arc is obtained along with the decision. In the synthetic data set of the original thesis, the tangent descriptor performs as well as or better than the image-window approach in all tested configurations. The proposed pipeline reaches an accuracy of 98%, vs 93% to 95% for the original approach once its evaluation is corrected. We tested our pipeline, with models trained only on synthetic data, on the PairingNet benchmark, and obtained an AUC of 0.93. Furthermore, under the PairingNet pair-searching protocol conditions, our learned descriptor obtains a Recall@10 of 0.82 on the real set against 0.56 from the best model of the original paper.
Chinese Translation
本文探讨了基于二维碎片轮廓之间的邻接预测。我们改进了 Beaulac 论文中提出的两阶段架构，其中一个旋转等变的孪生卷积神经网络对两个碎片的两个轮廓沿着局部图像窗口的配对进行评分。这些评分被汇总在一个邻接矩阵中，其中一个 ResNet 检测出部分反对角带，揭示两个碎片的邻接关系。在当前的工作中，我们保留了这一流程，并通过比较轮廓窗口的切线角轮廓来替换局部评分，从而使其在构造上对碎片旋转不变，并且对所选轮廓起始点不敏感。这些适配可以是一个无训练的似然比，或者是一个在对应点上训练的小型一维卷积模型。我们还用一个带状 U-Net 替换了最终分类器，该网络对带状区域进行分割并对配对进行分类，从而在做出决策的同时获得共享的弧段。在原始论文的合成数据集中，切线描述符在所有测试配置中表现与图像窗口方法相当或更好。所提出的流程达到了 98% 的准确率，而原始方法在评估修正后为 93% 到 95%。我们在仅使用合成数据训练的模型上测试了我们的流程，并在 PairingNet 基准测试中获得了 0.93 的 AUC。此外，在 PairingNet 配对搜索协议条件下，我们学习的描述符在真实数据集上获得了 0.82 的 Recall@10，而原始论文中最佳模型的 Recall@10 为 0.56。
cs.CV / 61 / 2610.09478
InstanceBench: Diagnosing Referential Reasoning and Target Identity in Referring Expression Segmentation
InstanceBench：诊断指称推理和指称表达分割中的目标身份
Abstract
Referring Expression Segmentation (RES) links natural-language descriptions to pixel-level object masks. Yet standard evaluation provides limited insight into instance-level referential reasoning: it does not systematically distinguish referential logics, test target preservation across valid grounding paths, or separate target-selection from mask-generation errors. We introduce InstanceBench, an instance-centered diagnostic benchmark comprising 6,194 images, 9,264 target instances, and 25,077 human-verified expressions. Each target-centric expression set (TCES) fixes the image and target mask while pairing a minimal expression with a same-target variant that uses another valid cue or grounding path. A compact referential-logic taxonomy spans direct target evidence, same-class selection, relational and compositional grounding, and exclusion, while logic-critical construction suppresses simpler shortcuts. Identity-aware metrics measure target retention and set-level success while separating selection from mask-generation errors. Across 22 native-mask RES checkpoints from 18 model families, the strongest checkpoint reaches 67.1% mIoU but only 59.6%
[email protected]. Controlled interventions confirm language sensitivity, while failure decomposition identifies target selection rather than mask decoding as the main bottleneck. On a controlled training subset, matched supervision improves identity-aware performance, showing that the diagnosed capability responds to targeted supervision. Collectively, InstanceBench supports a measure-diagnose-improve cycle: measuring target consistency across grounding paths, localizing failure sources, and evaluating targeted interventions.
Chinese Translation
指称表达分割（RES）将自然语言描述与像素级对象掩膜关联起来。然而，标准评估在实例级指称推理方面提供的洞察有限：它未能系统地区分指称逻辑，测试有效基础路径上的目标保留，或将目标选择与掩膜生成错误分开。我们引入了InstanceBench，这是一个以实例为中心的诊断基准，包含6194张图像、9264个目标实例和25077个经过人工验证的表达。每个以目标为中心的表达集（TCES）固定图像和目标掩膜，同时将一个最小表达与使用另一种有效线索或基础路径的同目标变体配对。一个紧凑的指称逻辑分类法涵盖直接目标证据、同类选择、关系和组合基础，以及排除，而逻辑关键构造抑制了更简单的捷径。意识到身份的度量标准测量目标保留和集合级成功，同时将选择与掩膜生成错误分开。在来自18个模型家族的22个原生掩膜RES检查点中，最强的检查点达到了67.1%的mIoU，但仅为59.6%的
[email protected]。控制干预确认了语言敏感性，而失败分解则识别出目标选择而非掩膜解码是主要瓶颈。在一个受控训练子集上，匹配监督提高了对身份的敏感性能，显示出被诊断的能力对有针对性的监督作出反应。总体而言，InstanceBench支持测量-诊断-改进循环：测量基础路径上的目标一致性，定位失败源，并评估有针对性的干预。
cs.CV / 62 / 2610.09492
Gaussian Material Fields for Volumetric Multi-Energy CT Decomposition
用于体积多能CT分解的高斯材料场
Abstract
Volumetric material decomposition in multi-energy computed tomography requires a representation that organizes multiple three-dimensional material fields in a common spatial domain while retaining differences in composition and local structure. We observe that spatial primitives can be shared across materials without tying their coefficients, but their local capacity must respond to material-specific reconstruction needs. We introduce Gaussian material fields, which represent multiple material distributions with shared anisotropic 3D Gaussian primitives and independent nonnegative material coefficients. The shared geometry defines a continuous spatial basis, while the coefficients determine each primitive's contribution to the individual material fields. To reconstruct this representation from multi-energy projections, a differentiable spectral forward model combines Gaussian material path integrals with a calibrated basis matrix, enabling joint optimization of spatial geometry and material composition. Material-aware adaptive density control retains material-specific refinement evidence before aggregation and adjusts local representation capacity to accommodate both spatially extensive components and sparse details. Experiments use synthesized multi-energy projections generated from pseudo-reference material maps constructed by conventional methods from publicly available CT data. Across 15 cases, our approach improves average PSNR by 4.03 dB and SSIM by 4.96% over the strongest baseline, while reducing NRMSE by 33.45%. Material-wise comparisons and component ablations support improved recovery of localized structures, while runtime and memory measurements show favorable computational scaling. These results establish Gaussian material fields as an explicit, adaptive representation for volumetric multi-material reconstruction.
Chinese Translation
多能计算机断层扫描中的体积材料分解需要一种表示方法，该方法能够在共同的空间域中组织多个三维材料场，同时保留成分和局部结构的差异。我们观察到空间原语可以在材料之间共享，而无需绑定它们的系数，但其局部容量必须响应特定材料的重建需求。我们引入高斯材料场，该方法使用共享的各向异性三维高斯原语和独立的非负材料系数来表示多个材料分布。共享的几何形状定义了一个连续的空间基，而系数则决定了每个原语对各个材料场的贡献。为了从多能投影中重建这种表示，差分光谱前向模型将高斯材料路径积分与校准的基矩阵结合，能够联合优化空间几何和材料成分。材料感知的自适应密度控制在聚合之前保留了特定材料的细化证据，并调整局部表示能力，以适应空间广泛的组件和稀疏细节。实验使用从公共CT数据中通过传统方法构建的伪参考材料图生成的合成多能投影。在15个案例中，我们的方法在最强基线的基础上平均提高了4.03 dB的PSNR和4.96%的SSIM，同时将NRMSE降低了33.45%。材料间的比较和组件消融实验支持了局部结构的更好恢复，而运行时间和内存测量显示出良好的计算扩展性。这些结果确立了高斯材料场作为体积多材料重建的显式自适应表示。
cs.CV / 63 / 2610.09498
SpatialUQ: Post-Hoc Uncertainty Quantification from Spatial Consistency in Black-Box Vision Models
SpatialUQ：基于黑箱视觉模型空间一致性的后验不确定性量化
Abstract
Clinical vision models are often deployed as frozen black boxes with no access to internals, retraining, or ground truth at inference time. We introduce \textbf{SpatialUQ}, a post-hoc uncertainty method using only output probabilities. It measures the Jensen-Shannon divergence between the global prediction and the mean of five fixed spatial crops in six deterministic forward passes. The premise is simple, trustworthy predictions are spatially consistent. On NIH ChestX-ray14 (DenseNet-121, $N{=}25{,}596$), our Multicrop Uncertainty Score (MUS) reaches $0.784$ failure-detection AUC versus $0.664$ for MC-Dropout ($p{<}10^{-6}$) at one-fifth the compute, with native calibration ($\text{SCE}{=}0.049$ vs.\ $0.127$ for $\ell_1$), the best-calibrated among methods above 0.78 AUC. A supervised fusion of MUS with entropy, confidence, and $\ell_1$ reaches $0.832$, outperforming a five-member ensemble ($0.813$). MUS scales with model quality, reaching $0.899$ with BiomedCLIP ($\rho = 0.846$), while this relationship remains meaningful in-distribution ($\rho = 0.523$) but breaks down under severe distribution shift (VinBigData, $\rho = 0.027$). MUS is well-suited to diffuse findings but is less dependable for small focal lesions such as nodules. Code and experimental materials are publicly available at https://huggingface.co/datasets/kawsher11/SpatialUQ.
Chinese Translation
临床视觉模型通常作为冻结的黑箱部署，无法在推理时访问内部结构、重新训练或获取真实标签。我们提出了 extbf{SpatialUQ}，这是一种仅使用输出概率的后验不确定性方法。它通过测量全局预测与六次确定性前向传递中五个固定空间裁剪的均值之间的詹森-香农散度来评估不确定性。其前提是简单的，可信的预测在空间上是一致的。在NIH ChestX-ray14数据集上（DenseNet-121，$N{=}25{,}596$），我们的多裁剪不确定性评分（MUS）达到了$0.784$的故障检测AUC，而MC-Dropout的AUC为$0.664$（$p{<}10^{-6}$），计算量仅为其五分之一，并且具有原生校准（$ ext{SCE}{=}0.049$，与$ ext{l}_1$的$0.127$相比），在AUC超过0.78的方法中是校准效果最佳的。MUS与熵、置信度和$ ext{l}_1$的监督融合达到了$0.832$，优于五成员集成（$0.813$）。MUS与模型质量成正比，使用BiomedCLIP时达到了$0.899$（$
ho = 0.846$），而在分布内这一关系仍然显著（$
ho = 0.523$），但在严重的分布转移下（VinBigData，$
ho = 0.027$）则失效。MUS非常适合扩散发现，但对于小的局部病变如结节的可靠性较低。代码和实验材料可在https://huggingface.co/datasets/kawsher11/SpatialUQ公开获取。
cs.CV / 64 / 2610.09502
RT-DETR-World: Transferring Rich LLM Semantics to Real-Time Open-Vocabulary Detection
RT-DETR-World：将丰富的LLM语义转移到实时开放词汇检测
Abstract
Open-vocabulary detection (OVD) recognizes categories unseen during training through textual category queries, yet achieving strong generalization with real-time efficiency remains challenging. Beyond vocabulary scaling, zero-shot generalization may benefit from reusable visual--semantic cues learned from seen data, including attributes, actions, states, and contextual relations. Existing real-time OVD methods primarily emphasize vocabulary coverage and efficient region/query--text matching; under strict efficiency constraints, compact detectors may struggle to absorb rich instance semantics and scene context. We propose RT-DETR-World, a compact DETR-style detector that transfers the rich semantics conveyed by descriptions during training while retaining lightweight query--text matching at inference. We construct GroundingCapv2 with three levels of supervision: category names for standard OVD, object descriptions conveying instance-level semantics, and image descriptions conveying object relations and scene context. These descriptions serve only as training-time semantic supervision. To help the compact detector absorb these semantics, we propose Dual-Path Description Alignment (DDA), combining a deployment-consistent MiniLM pathway with a training-only LLM teacher. MiniLM provides query--category supervision and object-description alignment, while offline teacher features supervise matched queries and global visual representations at the object and image levels, respectively. All teacher features are precomputed, and the teacher-side modules are removed after training. We further propose Relation-Aware Negative Relaxation (RNR), which uses teacher-derived semantic similarities to relax related negatives while preserving exact positives. Experiments demonstrate competitive zero-shot accuracy and a favorable accuracy--efficiency trade-off. The code will be released.
Chinese Translation
开放词汇检测（OVD）通过文本类别查询识别训练期间未见过的类别，但在实现强泛化和实时效率方面仍然面临挑战。除了词汇扩展，零样本泛化可能受益于从已见数据中学习到的可重用视觉-语义线索，包括属性、动作、状态和上下文关系。现有的实时OVD方法主要强调词汇覆盖和高效的区域/查询-文本匹配；在严格的效率限制下，紧凑型检测器可能难以吸收丰富的实例语义和场景上下文。我们提出了RT-DETR-World，这是一种紧凑的DETR风格检测器，它在推理时保留轻量级的查询-文本匹配，同时转移训练期间描述所传达的丰富语义。我们构建了GroundingCapv2，采用三层监督：用于标准OVD的类别名称、传达实例级语义的对象描述，以及传达对象关系和场景上下文的图像描述。这些描述仅作为训练时的语义监督。为了帮助紧凑型检测器吸收这些语义，我们提出了双路径描述对齐（DDA），结合了与部署一致的MiniLM路径和仅用于训练的LLM教师。MiniLM提供查询-类别监督和对象描述对齐，而离线教师特征分别在对象和图像级别监督匹配查询和全局视觉表示。所有教师特征都是预计算的，教师侧模块在训练后被移除。我们进一步提出了关系感知负放松（RNR），利用教师导出的语义相似性来放松相关的负样本，同时保留确切的正样本。实验表明，模型在零样本准确性方面具有竞争力，并且在准确性与效率之间取得了良好的平衡。代码将会发布。
cs.CV / 65 / 2610.09509
TERRA: Learning Transportable Latent Actions through Temporal Effect Representation and Relational Alignment
TERRA：通过时间效应表示和关系对齐学习可传输的潜在动作
Abstract
Latent actions supervise robot policies with action-like codes inferred from visual transitions, and their usefulness hinges on two questions: what a code keeps from a transition, and whether it still means the same thing when reused in a different initial state. The first is a tension in time: an endpoint difference discards how motion unfolds, while the full sequence admits nuisance variation. The second is left open by reconstruction, which only ever observes a latent together with the state it came from. We argue that both questions can be answered in the same place. TERRA (Temporal Effect Representation and Relational Alignment) describes a transition by a compact temporal effect, its net feature change together with a low-order within-window dynamics component, and learns a continuous latent from this effect. The same effect space then serves as the reference for reuse: Effect-Anchored Transport (EAT) decodes a latent in other initial states and anchors the resulting effect to the one observed at its source, so that the latent is shaped by what it does across contexts rather than only by the transition it came from. With frozen linear readers, TERRA predicts actions more accurately than UniVLA and a LAPA-style baseline, degrades more slowly under visual distractors, and keeps transported transitions faithful to the donor action as the recipient context moves farther away; a same-budget control shows that these gains come largely from EAT. At matched pretraining scale, the complete system reaches 93.4% average success on LIBERO, compared with 91.8% for UniVLA.
Chinese Translation
潜在动作通过从视觉转变推断出的类动作编码来监督机器人策略，其有效性取决于两个问题：编码从转变中保留了什么，以及在不同初始状态下重新使用时是否仍然意味着相同的东西。第一个问题涉及时间的紧张关系：端点差异忽略了运动的展开，而完整序列则允许干扰变化。第二个问题则因重构而悬而未决，重构仅观察到潜在状态及其来源状态。我们认为这两个问题可以在同一处得到解答。TERRA（时间效应表示和关系对齐）通过紧凑的时间效应描述转变，其净特征变化以及低阶的窗口内动态成分，并从这一效应中学习连续的潜在状态。相同的效应空间随后作为重用的参考：效应锚定传输（Effect-Anchored Transport, EAT）在其他初始状态中解码潜在状态，并将得到的效应锚定到其源头观察到的效应上，从而使潜在状态受到其在不同上下文中所做的事情的影响，而不仅仅是来自于其转变。使用冻结的线性读取器，TERRA比UniVLA和LAPA风格的基线更准确地预测动作，在视觉干扰下降级更慢，并且在接收上下文逐渐远离时保持传输转变忠实于捐赠动作；同样预算的对照实验表明，这些提升主要来自EAT。在匹配的预训练规模下，完整系统在LIBERO上达到了93.4%的平均成功率，而UniVLA为91.8%。
cs.CV / 66 / 2610.09511
LiG-DETR: Local-in-Global Reassembly in Latent Space for Aerial Object Detection
LiG-DETR：潜在空间中的全局-局部重组用于航空物体检测
Abstract
Aerial object detection faces substantial scale and density variations. Small objects are easily degraded by downsampling and feature compression, while medium and large objects require sufficient global context. Existing methods mainly follow two paradigms: image slicing provides clearer local evidence but relies on independent crop-level prediction and post-processing, whereas feature- and query-level optimization preserves unified inference but operates on already compressed full-image representations, limiting recovery of fine-grained information. This raises a key question: can aerial detection directly acquire high-fidelity local evidence before feature degradation and integrate it into a unified end-to-end framework? To this end, we propose LiG-DETR, an Efficient Global-Local Reassembly framework that reformulates image slicing as high-fidelity local feature acquisition. A shared encoder extracts global and locally magnified features, which are projected into the detector feature space. The projected local features are reassembled according to their original spatial locations to form a globally aligned local feature level, and a single DETR decoder jointly decodes global and local features. To reduce redundant computation, Context-Preserved Selective Reassembly focuses high-resolution encoding on informative regions while preserving a dense feature layout, and Density-Aware Adaptive Query Allocation adapts the decoder query budget using encoder proposal scores. Experiments show substantial gains on small and medium objects while retaining strong large-object performance, with favorable accuracy--efficiency trade-offs and improved cross-domain generalization. The code will be released.
Chinese Translation
航空物体检测面临显著的尺度和密度变化。小物体容易因下采样和特征压缩而退化，而中大物体则需要足够的全局上下文。现有方法主要遵循两种范式：图像切片提供更清晰的局部证据，但依赖于独立的裁剪级预测和后处理；而特征和查询级优化则保留统一推理，但在已经压缩的全图表示上操作，限制了细粒度信息的恢复。这引发了一个关键问题：航空检测能否在特征退化之前直接获取高保真局部证据，并将其整合到统一的端到端框架中？为此，我们提出了LiG-DETR，一个高效的全局-局部重组框架，将图像切片重新表述为高保真局部特征获取。共享编码器提取全局和局部放大特征，这些特征被投影到检测器特征空间中。投影后的局部特征根据其原始空间位置重新组合，以形成全局对齐的局部特征层，而单个DETR解码器则共同解码全局和局部特征。为了减少冗余计算，保留上下文的选择性重组将高分辨率编码集中在信息丰富的区域，同时保留密集的特征布局；而密度感知自适应查询分配则利用编码器提议分数调整解码器查询预算。实验表明，在保持强大大物体性能的同时，小物体和中物体的性能显著提升，具有良好的准确性-效率权衡和改进的跨域泛化能力。代码将会发布。
cs.CV / 67 / 2610.09513
OmniCam: Omni-Camera Trajectory Generation via Geometry-Grounded Pose Token Learning
OmniCam：通过几何基础的姿态标记学习生成全景相机轨迹
Abstract
Camera trajectories control viewpoint changes in video generation, scene reconstruction, and robotic perception. Generating them from language requires both scene geometry and target-aware framing. We introduce OmniCam, an autoregressive model that generates camera pose sequences from a single panorama and textual trajectory descriptions. Its geometry-grounded pose token learning combines three components: a panoramic point-cloud encoder for omnidirectional geometric context; hybrid absolute-rotation and relative-translation tokenization with temporally consistent quaternion signs; and separate geometric and semantic conditioning streams with an explicit 3D target anchor. We also construct OmniCaT, containing 267,700 trajectories across four camera behaviors. On the reported OmniCaT evaluation, OmniCam reduces trajectory errors by 28--47% and collision rate by 65.8% relative to GenDoP retrained on OmniCaT. Against the best baseline for each metric, the ATE and collision reductions are 43.0% and 62.3%, respectively. Component ablations support the use of geometric and target-aware conditioning, while downstream experiments examine camera-controlled video generation and robotic active perception.
Chinese Translation
相机轨迹控制视频生成、场景重建和机器人感知中的视点变化。从语言生成这些轨迹需要场景几何和目标感知的框架。我们提出了OmniCam，这是一种自回归模型，能够从单个全景图和文本轨迹描述中生成相机姿态序列。其几何基础的姿态标记学习结合了三个组成部分：用于全向几何上下文的全景点云编码器；具有时间一致四元数符号的混合绝对旋转和相对平移标记化；以及带有显式3D目标锚的独立几何和语义条件流。我们还构建了OmniCaT，包含267,700条轨迹，涵盖四种相机行为。在报告的OmniCaT评估中，OmniCam相较于在OmniCaT上重新训练的GenDoP，将轨迹误差降低了28%至47%，碰撞率降低了65.8%。与每个指标的最佳基线相比，平均轨迹误差（ATE）和碰撞率的降低分别为43.0%和62.3%。组件消融实验支持使用几何和目标感知条件，而下游实验则考察了相机控制的视频生成和机器人主动感知。
cs.CV / 68 / 2610.09514
STRIKE: Learning Visual State Transitions for Physical World Modeling
STRIKE：用于物理世界建模的视觉状态转移学习
Abstract
Physical world modeling requires predicting how interactions change a scene, not merely generating coherent motion. We propose STRIKE, a framework that separates visual state transition learning from dense video generation. We construct event-aligned supervision by extracting observed states from training videos and pairing them with transition descriptions and temporal offsets. An image-based transition model learns to predict the next scene configuration from the current image, a local transition specification, and elapsed time. At inference, a pretrained vision-language planner predicts time transition specifications, and recursive application of the learned transition model produces a sequence of future visual states. A separately trained dynamic model then generates the complete rollout conditioned on these states and their temporal locations. Experiments on Physics-IQ Verified, PhyGenBench, Pisa-Experiments, and RoboTwin2.0 show improvements of STRIKE over the corresponding video-backbone baselines in benchmark measures of physical consistency and manipulation-video fidelity. These results support learned visual state transitions as an effective intermediate representation for physical world modeling.
Chinese Translation
物理世界建模需要预测交互如何改变场景，而不仅仅是生成连贯的运动。我们提出了STRIKE，一个将视觉状态转移学习与密集视频生成分开的框架。通过从训练视频中提取观察到的状态，并将其与转移描述和时间偏移配对，我们构建了事件对齐的监督。基于图像的转移模型学习从当前图像、局部转移规范和经过的时间预测下一个场景配置。在推理阶段，预训练的视觉-语言规划器预测时间转移规范，学习的转移模型的递归应用生成一系列未来的视觉状态。然后，单独训练的动态模型基于这些状态及其时间位置生成完整的展开。对Physics-IQ Verified、PhyGenBench、Pisa-Experiments和RoboTwin2.0的实验表明，STRIKE在物理一致性和操作视频保真度的基准测量中相较于相应的视频骨干基线有显著提升。这些结果支持学习的视觉状态转移作为物理世界建模的有效中间表示。
cs.CV / 69 / 2610.09517
It Is Not Seeing the Hazard: A Frozen Vision-Language Safety Score Measures Its Caption Bank
看不到危险：一个冻结的视觉-语言安全评分衡量其描述库
Abstract
Frozen vision-language models increasingly provide safety signals for reinforcement learning. Their use assumes that similarity to language describing danger indicates the hazard itself. Yet policy return and collision rate cannot reveal whether a score detects hazards or responds to correlated features of the scene. VLM-based methods have reported gains in driving and safe-RL benchmarks by converting image-text similarity into rewards, costs, or confidence weights. Such signals promise to reduce reliance on manually designed feedback. They may also reflect prompt structure, embedding geometry, or camera viewpoint, leaving their safety meaning unverified. To address this gap, we present a controlled evaluation of a frozen CLIP prompt-margin safety score. We apply the score to trajectories generated by policies that never receive it, match pre-contact observations to contact-free observations with comparable hazard geometry, and vary the captions, encoder, and camera view. Across three policies, 180 episodes, and 130 isolated contact onsets, the score decreases for about twenty steps before contact. Mechanism controls indicate that the score mainly tracks resemblance to the scene shared by its captions and changes with caption separation and camera view. A constant-confidence control retains the lower catastrophe-rate point estimate, so policy gains do not establish hazard perception.
Chinese Translation
冻结的视觉-语言模型越来越多地为强化学习提供安全信号。它们的使用假设与描述危险的语言相似性表明了危险本身。然而，政策回报和碰撞率无法揭示一个评分是否检测到危险，或是对场景的相关特征做出反应。基于视觉-语言模型（VLM）的方法通过将图像-文本相似性转化为奖励、成本或置信权重，在驾驶和安全强化学习基准测试中报告了收益。这些信号有望减少对手动设计反馈的依赖。它们也可能反映提示结构、嵌入几何或相机视角，使其安全含义未得到验证。为了解决这一问题，我们提出了对一个冻结的CLIP提示边际安全评分的控制评估。我们将该评分应用于从未接收到该评分的策略生成的轨迹，将接触前的观察与具有可比危险几何的无接触观察进行匹配，并改变标题、编码器和相机视角。在三个策略、180个回合和130个孤立接触起始点中，该评分在接触前约二十步内下降。机制控制表明，该评分主要跟踪与其标题共享的场景的相似性，并随着标题分离和相机视角的变化而变化。一个常量置信控制保留了较低灾难率的点估计，因此政策收益并未建立危险感知。
cs.CV / 70 / 2610.09518
ActiveLang: Active Open-Vocabulary 3D Mapping with Semantic-Uncertainty-Guided Exploration
ActiveLang：基于语义不确定性引导探索的主动开放词汇3D映射
Abstract
As robots increasingly assist humans with diverse tasks, they need both geometric and semantic understanding of their surroundings. Moreover, robots often operate in unfamiliar environments and take on new tasks without knowing the relevant concepts ahead of time. This motivates language-annotated 3D maps that support open-vocabulary scene understanding and human-robot interaction. We introduce ActiveLang, an autonomous system for active open-vocabulary 3D mapping with semantic-uncertainty-guided exploration. ActiveLang performs online language-feature adaptation on a compact dual-Gaussian representation to jointly reconstruct scene geometry, appearance, and open-vocabulary semantics with modest memory overhead. Its planner efficiently selects informative viewpoints, enabling effective mapping with fewer observations and lower computational cost. Experiments on Replica and ScanNet++ demonstrate substantial improvements in 2D and 3D open-vocabulary segmentation over both online and offline baselines, highlighting that actively exploring scenes builds language-annotated 3D maps more efficiently.
Chinese Translation
随着机器人在多样化任务中越来越多地协助人类，它们需要对周围环境具有几何和语义的理解。此外，机器人通常在不熟悉的环境中操作，并在不知道相关概念的情况下承担新任务。这促使我们开发语言注释的3D地图，以支持开放词汇的场景理解和人机交互。我们提出了ActiveLang，这是一种用于主动开放词汇3D映射的自主系统，采用语义不确定性引导的探索。ActiveLang在紧凑的双高斯表示上进行在线语言特征适应，以共同重建场景的几何、外观和开放词汇语义，同时保持适度的内存开销。其规划器有效选择信息丰富的视点，使得在较少观测和较低计算成本的情况下实现有效映射。在Replica和ScanNet++上的实验表明，与在线和离线基线相比，2D和3D开放词汇分割有显著改善，强调了主动探索场景能够更高效地构建语言注释的3D地图。
cs.CV / 71 / 2610.09534
KASALv2: Fully Automatic 3D Rotational Symmetry Classification and Axis Localization
KASALv2：全自动三维旋转对称性分类与轴定位
Abstract
Rotational symmetry is an important prior in 6D pose estimation, improving pose accuracy and supporting symmetry-aware evaluation. However, current symmetry annotations for 3D objects remain largely manual or semi-automatic, often requiring predefined types or orders, which limits scalability. This work introduces a fully automatic, reference-free framework for symmetry-type classification, rotational-order identification, and full-axis localization across all eight canonical 3D rotational symmetry types. The method localizes a dominant high-order axis, infers its rotational order through self-consistency analysis, and reconstructs the complete symmetry structure under a hierarchy-guided formulation. A texture-aware extension further models appearance-induced reductions in rotational order while preserving axis orientations. Experiments on idealized and real-world datasets demonstrate strong accuracy and generalization, achieving 94.75% accuracy on 438 symmetric objects in GSO. Training FoundationPose with these priors improves accuracy by up to 0.9% across five BOP datasets, showing that automatically estimated rotational priors improve downstream 6D pose estimation. Code is available at https://github.com/WangYuLin-SEU/KASAL.
Chinese Translation
旋转对称性是六维姿态估计中的一个重要先验，能够提高姿态准确性并支持基于对称性的评估。然而，目前针对三维物体的对称性标注仍然主要依赖人工或半自动化，通常需要预定义的类型或顺序，这限制了其可扩展性。本研究提出了一种全自动、无参考的框架，用于对称性类型分类、旋转顺序识别以及在所有八种典型三维旋转对称性类型下的完整轴定位。该方法定位一个主导的高阶轴，通过自一致性分析推断其旋转顺序，并在层次引导的公式下重建完整的对称结构。一个纹理感知的扩展进一步建模了外观引起的旋转顺序减少，同时保持轴的方向。对理想化和真实世界数据集的实验表明，该方法具有强大的准确性和泛化能力，在438个对称物体的GSO数据集中达到了94.75%的准确率。使用这些先验训练FoundationPose在五个BOP数据集上的准确性提高了最多0.9%，显示出自动估计的旋转先验能够改善下游六维姿态估计。代码可在https://github.com/WangYuLin-SEU/KASAL获取。
cs.CV / 72 / 2610.09535
WAPR: A Foundation Model for Wide-Angle Refinement in Unseen Object Pose Estimation
WAPR：一种用于未见物体姿态估计的广角精细化基础模型
Abstract
Real-world applications require 6D pose estimation to be accurate, fast, and scalable to unseen objects. This paper introduces WAPR, a zero-shot wide-angle pose refinement model that refines candidate poses with rotational deviations up to 90 degrees. With as few as 12 candidate poses per detected object instance, WAPR supports fast inference within 1 s per frame and reaches a pose-estimation throughput of up to 25 detected object instances per second. To support wide-angle training for rotationally symmetric objects, WAPR uses rotational symmetry priors to canonicalize symmetry-equivalent pose targets before loss computation. We further construct SA6D, a large-scale 6D training dataset with such priors. SA6D obtains KASAL-assisted rotational symmetry priors for 944 GSO scans and expands them through geometry and texture augmentation into about 50K augmented object instances and about 2M rendered RGB-D images. In addition, an angle-balanced loss stabilizes learning across different angular ranges by reducing the influence of uninformative large-error cases. Experiments on seven BOP core datasets show that WAPR achieves state-of-the-art performance in unseen-object 6D pose localization and detection under both fast and unconstrained inference settings. Project page: https://github.com/WangYuLin-SEU/WAPR.
Chinese Translation
现实世界的应用要求6D姿态估计在精度、速度和可扩展性方面对未见物体表现良好。本文介绍了WAPR，一种零-shot广角姿态精细化模型，能够对旋转偏差高达90度的候选姿态进行精细化。WAPR支持每个检测到的物体实例仅使用12个候选姿态即可实现每帧1秒内的快速推理，并达到每秒最多25个检测物体实例的姿态估计吞吐量。为了支持旋转对称物体的广角训练，WAPR在损失计算之前使用旋转对称先验对对称等效姿态目标进行规范化。此外，我们进一步构建了SA6D，一个具有此类先验的大规模6D训练数据集。SA6D为944个GSO扫描获取了KASAL辅助的旋转对称先验，并通过几何和纹理增强扩展到约50K个增强物体实例和约200万张渲染的RGB-D图像。此外，角度平衡损失通过减少无信息的大误差案例的影响，稳定了不同角度范围内的学习。在七个BOP核心数据集上的实验表明，WAPR在未见物体的6D姿态定位和检测方面，在快速和无约束推理设置下均达到了最先进的性能。项目页面：https://github.com/WangYuLin-SEU/WAPR。
cs.CV / 73 / 2610.09550
Visual Evidence Under Cross-Examination: Evaluating and Controlling Decision-Level Evidence Use in Vision-Language Models
交叉审查下的视觉证据：评估和控制视觉-语言模型中的决策级证据使用
Abstract
Vision-language models increasingly reason through crops, regions, and tool-produced observations. Yet an observation can influence the answer without benefiting the candidate it supports. We study candidate-bound visual contribution: valid evidence should help, invalidating its supporting relation should remove its additional effect, and valid rebinding should redirect that effect to the newly supported candidate. We introduce CROSS-Bench, a benchmark of 28,000 decision problems, with matched invalidation and rebinding tests on a dedicated evaluation subset. Our RIVET interface preserves evidence identity and uncertainty, composes a candidate-conditioned response, and separately controls its strength. Shared-evidence experiments show that task accuracy and evidence ownership can diverge. Under matched capacity and training, RIVET increases normalized effect transfer from 0.512 to 0.651 where clean evidence has a positive effect. The advantage persists on common evaluation examples and across repeated decision-layer fits. With evidence predicted from raw inputs, RIVET improves CROSS-Bench accuracy by an average of 5.70 pp across four frozen backbones, relative to the same models without auxiliary evidence. These results separate the utility of visual evidence from the candidate-specific destination of its effect.
Chinese Translation
视觉-语言模型越来越多地通过裁剪、区域和工具生成的观察进行推理。然而，一项观察可能会影响答案，而并未对其所支持的候选项带来益处。我们研究候选项绑定的视觉贡献：有效证据应当提供帮助，失效其支持关系应当消除其额外效应，而有效的重新绑定应当将该效应重定向到新支持的候选项。我们引入了CROSS-Bench，一个包含28,000个决策问题的基准测试，配备了在专门评估子集上匹配的失效和重新绑定测试。我们的RIVET接口保留了证据的身份和不确定性，构建了候选项条件的响应，并单独控制其强度。共享证据实验表明，任务准确性和证据所有权可能会出现分歧。在匹配的能力和训练下，RIVET将规范化效应转移从0.512提高到0.651，其中干净证据具有正面效应。这一优势在常见评估示例和重复的决策层拟合中持续存在。通过从原始输入预测证据，RIVET在四个冻结骨干网络上平均提高了CROSS-Bench的准确性5.70个百分点，相较于没有辅助证据的相同模型。这些结果将视觉证据的效用与其效应的候选特定目标区分开来。
cs.CV / 74 / 2610.09579
Gradient-Based Trajectory Optimisation over Continuous Poses for Sparse-View Cone-Beam CT
基于梯度的稀视角锥束CT连续姿态轨迹优化
Abstract
Trajectory optimisation for cone-beam computed tomography (CT) determines which information sparse-view scans acquire. Fixed candidate pools prevent off-grid refinement and require new object-specific precomputation for each acquisition manifold. We make every source pose an individual continuous variable and move all poses jointly by gradient ascent on the scanner's kinematic manifold. The objective combines soft-Tuy plane coverage, continuous View Covariance Loss, and an analytic attenuation-aware ray-bundle penalty. The same optimiser handles circular, limited C-arm, two-axis, and freesphere parametrisations. On a Defrise flange, continuous selection recovers laminar defects invisible to a circular orbit, matches discrete swap search on the free sphere at the sparser budget, and leads at the denser one, with the same objective evaluated in every arm. A moderate elevation band already recovers most of the free-sphere gain at the defects, so the same optimiser transfers to bounded scanner envelopes. Photon noise preserves the ordering on the flange and compresses it on a dense fuel nozzle. Sparseprescan planning benefits from matching prescan and planned acquisition manifolds. Selection takes seconds rather than minutes without an object-specific reconstruction basis. Prescan-planned poses were executed on a robot CT bench and reconstructed in a common frame, demonstrating feasibility but no consistent metric gain over uniform band sampling. Continuous pose optimisation incorporates attenuation and scanner constraints directly into sparse-view acquisition design.
Chinese Translation
锥束计算机断层扫描（CT）的轨迹优化决定了稀视角扫描所获取的信息。固定的候选池限制了离网格的优化，并且每个采集流形都需要新的特定对象的预计算。我们将每个源姿态视为一个独立的连续变量，并通过在扫描仪的运动流形上进行梯度上升共同移动所有姿态。目标函数结合了软Tuy平面覆盖、连续视角协方差损失和解析的衰减感知光线束惩罚。相同的优化器处理圆形、有限C臂、双轴和自由球体参数化。在Defrise法兰上，连续选择恢复了在圆形轨道下不可见的层流缺陷，在稀疏预算下与自由球体的离散交换搜索相匹配，并在更密集的预算下领先，且在每个臂上评估相同的目标。适度的提升带已经在缺陷处恢复了大部分自由球体的增益，因此相同的优化器可以转移到有限的扫描仪包络中。光子噪声在法兰上保持了顺序，并在密集的燃油喷嘴上进行了压缩。稀视角预扫描规划受益于匹配预扫描和计划采集流形。选择过程只需几秒而不是几分钟，而无需特定对象的重建基础。预扫描计划的姿态在机器人CT工作台上执行，并在一个共同框架中重建，展示了可行性，但在均匀带采样上没有一致的度量增益。连续姿态优化将衰减和扫描仪约束直接纳入稀视角采集设计中。
cs.CV / 75 / 2610.09598
STORK: Spatio-Temporal Observation of uterine contRactions via neural networKs
STORK：通过神经网络对子宫收缩进行时空观察
Abstract
Uterine contractions in fetal MRI are typically identified manually and discarded, limiting insights into contraction dynamics. We formalize Uterine Contractile Activity Detection (UCAD) as a weakly-supervised learning problem and introduce STORK, a multi-instance learning model trained on dynamic MRI series using only coarse, series-level labels. STORK factorizes 3D spatio-temporal convolutions into parallel branches across temporal hyperplanes to capture coherent tissue motion without the cost of full 4D convolutions. Per-frame embeddings, combining intensity and Demons-estimated displacement fields, are aggregated by a linear mean-pooling head. This ensures that frame-level contraction scores can be recovered post-hoc without frame-level training supervision. Evaluated on around 700 multi-vendor dynamic fetal MRI series, STORK achieves a series-level AUROC of 95.0% and AUPRC of 94.6%, substantially outperforming 3D ResNet and ConvNeXt baselines. Grad-CAM analysis suggests that the model draws on predictive features extending beyond the placenta into the uterine tissue, offering an automated tool for richer phenotyping of uterine behavior.
Chinese Translation
在胎儿MRI中，子宫收缩通常是手动识别并被丢弃的，这限制了对收缩动态的深入理解。我们将子宫收缩活动检测（Uterine Contractile Activity Detection，UCAD）形式化为一个弱监督学习问题，并引入STORK，一个基于动态MRI序列的多实例学习模型，仅使用粗略的序列级标签进行训练。STORK将3D时空卷积分解为跨时间超平面的并行分支，以捕捉一致的组织运动，而无需进行完整的4D卷积。每帧嵌入结合了强度和Demons估计的位移场，通过线性均值池头进行聚合。这确保了可以在后期恢复帧级收缩得分，而无需帧级训练监督。在约700个多供应商动态胎儿MRI序列上进行评估，STORK实现了95.0%的序列级AUROC和94.6%的AUPRC，显著优于3D ResNet和ConvNeXt基线。Grad-CAM分析表明，该模型利用了超出胎盘延伸至子宫组织的预测特征，为子宫行为的更丰富表型提供了自动化工具。
cs.CV / 76 / 2610.09614
Identity-Duplication Auditing in National-Scale Neuroimaging Repositories
国家级神经影像库中的身份重复审计
Li, Jiheng, Kim, Michael E., Schwartz, Trent M., Cui, Yuhan, Rudravaram, Gaurav, Archer, Derek B., Hohman, Timothy J., Beason-Held, Lori L., Morgan, Victoria L., Englot, Dario J., Jefferson, Angela L., Initiative, for the Alzheimer's Disease Neuroimaging, team, for the BIOCARD Study, Health, for the, Study, Aging Brain, :, Disparities, Health, Team, Study, Zuo, Lianrui, Erus, Guray, Davatzikos, Christos, Landman, Bennett A.
Abstract
National-scale magnetic resonance imaging (MRI) repositories increasingly integrate data from different studies and institutions. However, subject identifiers that are valid only within individual datasets are no longer guaranteed to remain globally unique after aggregation, making it possible for the same subject to be assigned multiple identifiers, which we define as identity duplication. Such duplication can create leakage between training and test data and inflate apparent performance in downstream biomedical studies. Existing methods do not provide an end-to-end, image-based workflow for auditing this problem at repository scale. In this work, we present HAPPEN, a human-in-the-loop pipeline for auditing identity duplication in T1-weighted brain MRI repositories. It combines SHA-256 fingerprinting for exact-duplicate detection with supervised contrastive retrieval of non-identical scans that may originate from the same person. Retrieved pairs are reviewed as candidates in a locally hosted interface rather than automatically classified as duplicates. We deployed the workflow in a 95,129-scan aggregated repository and assessed end-to-end recovery using 54 genetic-reference pairs. Transferability was assessed by locally deploying the same workflow on 22,386 scans at an independent institution without model retraining or image transfer. Deployment in the study repository identified 1,316 exact-duplicate scan groups and 1,275 reviewer-supported near-duplicate subject groups. Of these groups, 56% and 82%, respectively, crossed dataset boundaries. All 54 genetic-reference pairs were recovered. The external team independently completed the full workflow using a locally selected operating threshold and review standard.
Chinese Translation
国家级磁共振成像（MRI）库越来越多地整合来自不同研究和机构的数据。然而，仅在单个数据集中有效的受试者标识符在聚合后不再保证全球唯一，这使得同一受试者可能被分配多个标识符，我们将其定义为身份重复。这种重复可能导致训练数据和测试数据之间的泄漏，并在下游生物医学研究中夸大表观性能。现有方法未能提供一个端到端的、基于图像的工作流程来审计这一问题在库规模上的影响。在本研究中，我们提出了HAPPEN，一个用于审计T1加权脑MRI库中身份重复的人机协作管道。它结合了SHA-256指纹技术用于精确重复检测，以及监督对比检索来自同一人的非相同扫描。检索到的配对在本地托管的界面中作为候选项进行审查，而不是自动分类为重复。我们在一个包含95,129个扫描的聚合库中部署了该工作流程，并使用54对遗传参考对评估了端到端的恢复能力。通过在一个独立机构的22,386个扫描上本地部署相同的工作流程，评估了其可转移性，而无需模型重新训练或图像传输。在研究库中的部署识别出了1,316个精确重复扫描组和1,275个经审查支持的近重复受试者组。在这些组中，分别有56%和82%跨越了数据集边界。所有54对遗传参考对均已成功恢复。外部团队独立完成了整个工作流程，使用本地选择的操作阈值和审查标准。
cs.CV / 77 / 2610.09670
Quasi-Binarized Autoencoders: An Architecture-Independent Information Bottleneck for Medical Image Anomaly Detection
准二值化自编码器：一种架构无关的信息瓶颈用于医学图像异常检测
Abstract
Unsupervised anomaly detection, which learns only from normal images, is a central task in medical image analysis and remains an open problem. Reconstruction-based methods pass an image through an encoder-decoder network trained on normal data and detect anomalies from the residual between the image and its reconstruction. This works only if the information passed from the encoder to the decoder is limited; otherwise the network learns an identity mapping and reconstructs anomalies too. This limit is usually imposed through architectural choices, tuned per dataset, that cannot be stated in bits. We introduce the quasi-binarizing (QB) layer, which squashes each latent element into [0, 1] and adds Laplace noise of scale 1/epsilon. Each element is then epsilon-locally differentially private, and the mutual information between an image and its reconstruction is bounded by a quantity that depends only on epsilon and the number of QB elements, whatever the encoder and decoder. Placing a QB layer on every encoder-decoder path, including all skip connections, we build QBAE, a seven-level attention U-Net with 32,768 QB elements. On the seven datasets of the MedIAnomaly benchmark, QBAE with one architecture and one configuration reaches a mean image-level AUROC of 0.828, the highest among methods that do not adapt to each dataset, and the best reported results on BraTS2021 (AUROC 0.911, pixel-level AP 0.838). The noise is kept at test time, so that every reconstruction satisfies the bound. Without input corruption, the bottleneck alone prevents identity collapse (mean AUROC 0.805 vs. 0.590). Code is available at https://github.com/hanaokalog/MedIAnomalyQB.
Chinese Translation
无监督异常检测仅从正常图像中学习，是医学图像分析中的核心任务，仍然是一个未解决的问题。基于重建的方法通过在正常数据上训练的编码-解码网络传递图像，并从图像与其重建之间的残差中检测异常。这种方法只有在从编码器到解码器传递的信息受到限制时才有效；否则，网络会学习到身份映射并重建异常。这种限制通常通过架构选择施加，针对每个数据集进行调整，无法以比特数表示。我们引入了准二值化（QB）层，该层将每个潜在元素压缩到[0, 1]之间，并添加规模为1/epsilon的拉普拉斯噪声。每个元素在epsilon局部上是差分隐私的，图像与其重建之间的互信息被一个仅依赖于epsilon和QB元素数量的量所限制，无论编码器和解码器如何。在每个编码-解码路径上放置一个QB层，包括所有跳跃连接，我们构建了QBAE，一个具有32,768个QB元素的七层注意力U-Net。在MedIAnomaly基准的七个数据集上，QBAE以一种架构和一种配置达到了0.828的平均图像级AUROC，这是在不适应每个数据集的方法中最高的，并且在BraTS2021上报告了最佳结果（AUROC 0.911，像素级AP 0.838）。在测试时保持噪声，因此每个重建都满足限制。在没有输入损坏的情况下，仅瓶颈就能防止身份崩溃（平均AUROC 0.805对比0.590）。代码可在https://github.com/hanaokalog/MedIAnomalyQB获取。
cs.CV / 78 / 2610.09677
A Multi-Source Ultrasound Benchmark Revealing the Limits of Contemporary Self-Supervised Anomaly Detection Methods
多源超声基准揭示当代自监督异常检测方法的局限性
Abstract
Self-supervised anomaly detection is a promising paradigm for medical ultrasound, as normal images are often easier to obtain than exhaustive annotations of all possible pathologies. However, most existing evaluations are limited to a single anatomy or task, making it unclear whether models learn a robust notion of normal ultrasound appearance or only a source-specific representation. We introduce the SADUSI benchmark, a multi-source ultrasound dataset designed to train and evaluate anomaly detection methods across a broad range of anatomical regions, views, and acquisition protocols. The goal of SADUSI is to provide a diverse normal ultrasound distribution and a benchmark for visible structural anomalies that can be assessed from single images. We evaluate representative self-supervised anomaly detection methods and find that current approaches struggle in this setting. In particular, reconstruction-based diffusion methods such as AnoDDPM and DeCo-Diff achieve pixel-level AUROC values of 0.56-0.72 and maximum F1 scores of 0.10-0.26, indicating limited separation of pathology from normal image regions. Feature-based PatchCore variants perform better, reaching pixel-level AUROC values of 0.76-0.83, but remain limited with maximum F1 scores of 0.14-0.40. These findings suggest that broad multi-source ultrasound anomaly detection remains an open challenge and that SADUSI can serve as a resource for developing methods that generalize beyond anatomy-specific settings.
Chinese Translation
自监督异常检测是医学超声领域一种有前景的范式，因为正常图像通常比对所有可能病理的详尽注释更容易获得。然而，现有的大多数评估仅限于单一解剖结构或任务，这使得不清楚模型是否学习到了稳健的正常超声外观概念，还是仅仅获得了特定来源的表示。我们引入了SADUSI基准，一个旨在跨广泛解剖区域、视图和采集协议训练和评估异常检测方法的多源超声数据集。SADUSI的目标是提供多样化的正常超声分布以及可通过单幅图像评估的可见结构异常的基准。我们评估了具有代表性的自监督异常检测方法，发现当前的方法在这一设置中表现不佳。特别是基于重建的扩散方法，如AnoDDPM和DeCo-Diff，其像素级AUROC值为0.56-0.72，最大F1分数为0.10-0.26，表明病理与正常图像区域的分离有限。基于特征的PatchCore变体表现更好，达到像素级AUROC值为0.76-0.83，但最大F1分数仍然有限，为0.14-0.40。这些发现表明，广泛的多源超声异常检测仍然是一个开放的挑战，而SADUSI可以作为开发超越特定解剖设置的通用方法的资源。
cs.CV / 79 / 2610.09700
What Makes Synthetic Hard Negatives Work in Vision-Language Pretraining?
合成困难负样本在视觉-语言预训练中的有效性分析
Abstract
Synthetic hard negatives generated in the representation space have proven effective for unimodal self-supervised learning, but transferring this idea to vision-language pretraining is not straightforward. We analyze six representation-space synthesis strategies and identify two failure modes in their transfer to vision-language pretraining: cross-modal constructions that produce overly easy negatives or pull them toward the query, and intra-modal constructions that incorporate the matched positive. We also observe logit-scale saturation when training with synthetic hard negatives and a learnable temperature, and find that fixing the temperature improves downstream performance. Using this geometric analysis we propose SNAP, which generates intra-modal hard negatives that never involve the positive from either modality, avoiding both failure modes entirely. SNAP is model-agnostic, requires no external generative models, and adds less than 10% training time overhead. Evaluated on top of CLIP and FLIP across multiple architectures and datasets, SNAP delivers consistent improvements on zero-shot retrieval, zero-shot classification, and linear probe evaluation.
Chinese Translation
在表示空间中生成的合成困难负样本已被证明对单模态自监督学习有效，但将这一思想转移到视觉-语言预训练并非易事。我们分析了六种表示空间合成策略，并识别出在其转移到视觉-语言预训练时的两种失败模式：产生过于简单的负样本或将负样本拉向查询的跨模态构造，以及包含匹配正样本的同模态构造。我们还观察到在使用合成困难负样本和可学习温度进行训练时，logit尺度饱和现象，并发现固定温度可以改善下游性能。基于这种几何分析，我们提出了SNAP，它生成的同模态困难负样本从不涉及任何模态的正样本，完全避免了这两种失败模式。SNAP是模型无关的，不需要外部生成模型，并且增加的训练时间开销不足10%。在多个架构和数据集上对CLIP和FLIP进行评估，SNAP在零-shot检索、零-shot分类和线性探测评估上均表现出一致的改进。
cs.CV / 80 / 2610.09702
Latent Watermarks under Generative Editing: A Benchmark and Analysis of Detection Survival
生成编辑下的潜在水印：检测生存的基准与分析
Abstract
Ordinary prompt-based editing can cause latent watermark detection to fail without explicitly targeting the watermark. We benchmark eight watermark methods against five editors across four generative backbones, four editing strengths, and five semantic categories, with edit-validity and threshold checks. Separating editing from seven subsequent distortions reveals that editing alone primarily distinguishes Tree-Ring, while added distortions expose a broader spectrum of detection survival. Sequential edits reveal a second hidden difference: score separation can decline while detection rates remain near their ceiling. Across methods, standardized clean score separation ($d'$) organizes composite-survival tiers, whereas spatial overlap adds little to predicting edit-only survival beyond clean detectability. Embedding-strength interventions in two methods link higher clean separation to higher post-edit separation. In HSTR, the margin contrast is positive, while the angular layout contrast at matched clean separation remains unresolved. Together, outcome decomposition and continuous separation expose differences hidden by aggregate TPR. Method tiers are stable under threshold recalibration at the main operating points and alternative composite weights. Clean $d'$ is thus a useful empirical diagnostic within this benchmark, with mixed transfer to unseen methods. Code and supporting artifacts are planned for a separate release.
Chinese Translation
普通的基于提示的编辑可能导致潜在水印检测失败，而无需明确针对水印。我们对八种水印方法进行了基准测试，涉及五种编辑器、四种生成骨干网络、四种编辑强度和五种语义类别，并进行了编辑有效性和阈值检查。将编辑与七种后续失真分开，揭示了仅编辑主要区分树环（Tree-Ring），而添加的失真则暴露了更广泛的检测生存范围。顺序编辑揭示了第二个隐藏差异：分数分离可能下降，而检测率仍接近其上限。在各种方法中，标准化的干净分数分离（$d'$）组织了复合生存层次，而空间重叠对预测仅编辑生存的贡献有限，超出了干净可检测性。在两种方法中，嵌入强度干预将更高的干净分离与更高的后编辑分离联系起来。在 HSTR 中，边际对比为正，而在匹配的干净分离下，角度布局对比仍未解决。总体结果分解和连续分离揭示了被聚合真阳性率（TPR）掩盖的差异。方法层次在主要操作点和替代复合权重下对阈值重新校准是稳定的。因此，干净的 $d'$ 在此基准中是一个有用的实证诊断，且对未见方法的迁移效果混合。代码和支持文档计划单独发布。
cs.CV / 81 / 2610.09706
Enhancing Multi-Region Stylization with Interior-Guided Boundary Repair
基于内部引导的边界修复增强多区域风格化
Abstract
Region-based neural style transfer enables fine-grained artistic control by allowing independent stylization of semantic image regions. However, compositing these regions often leads to boundary artifacts, degrading visual quality. We propose Interior-Guided Boundary Repair (IGBR), a lightweight and model-agnostic method that improves boundary handling in multi-region stylization. IGBR repairs boundary pixels using interior-guided propagation and applies inward, distance-based blending restricted to object-background boundaries, preventing inter-object style leakage. The method is derived from a region-wise constrained formulation with a closed-form solution and can be seamlessly integrated into existing stylization pipelines without retraining. To evaluate efficiency of our IGBR, we introduce quantitative metrics that measure boundary consistency, gradient artifacts, inter-object leakage, and interior preservation without requiring annotated stylized images. Our experiments and evaluations demonstrate that the proposed IGBR consistently produces plausible boundaries, outperforming prior blending techniques in boundary consistency, gradient stability, and interior preservation. The code is available at https://github.com/Son-SDT/IGBR.
Chinese Translation
基于区域的神经风格迁移通过允许对语义图像区域进行独立风格化，实现了精细的艺术控制。然而，合成这些区域往往会导致边界伪影，从而降低视觉质量。我们提出了一种轻量级且与模型无关的方法——内部引导边界修复（Interior-Guided Boundary Repair, IGBR），该方法改善了多区域风格化中的边界处理。IGBR通过内部引导传播修复边界像素，并应用限制在物体与背景边界的向内、基于距离的混合，防止物体间风格泄漏。该方法源自具有封闭形式解的区域约束公式，并且可以无缝集成到现有的风格化流程中，而无需重新训练。为了评估IGBR的效率，我们引入了定量指标，测量边界一致性、梯度伪影、物体间泄漏和内部保留，而无需标注的风格化图像。我们的实验和评估表明，所提出的IGBR始终产生合理的边界，在边界一致性、梯度稳定性和内部保留方面优于先前的混合技术。代码可在 https://github.com/Son-SDT/IGBR 获取。
cs.CV / 82 / 2610.09720
DynStream: Online Streaming 4D Gaussian Reconstruction of Dynamic Worlds from Unposed Video
DynStream：从无姿态视频中在线流式重建动态世界的4D高斯模型
Abstract
Online reconstruction of dynamic 4D scenes from long, unposed streaming videos requires both continuous processing and photorealistic rendering, which existing methods struggle to achieve simultaneously. Existing feed-forward Gaussian methods are restricted to offline processing, whereas online point-cloud approaches struggle to maintain dense geometry and high-fidelity rendering. We present DynStream, a framework for streaming 4D Gaussian reconstruction from long, unposed videos. Given a continuous video stream, DynStream reconstructs the scene within local temporal windows and incrementally aligns and fuses these local reconstructions into a globally consistent scene, enabling online 4D reconstruction without per-scene optimization. By jointly enforcing cross-window geometric consistency and modeling time-varying scene content, DynStream supports efficient reconstruction and photorealistic rendering over extended video streams. Experiments demonstrate that DynStream enables high-fidelity online dynamic reconstruction and rendering from long video streams, achieving state-of-the-art performance across diverse dynamic indoor and outdoor scenes.
Chinese Translation
从长时间的无姿态流式视频中在线重建动态4D场景需要持续处理和逼真的渲染，而现有方法在这两者之间难以同时实现。现有的前馈高斯方法仅限于离线处理，而在线点云方法在保持密集几何和高保真渲染方面面临挑战。我们提出了DynStream，一个用于从长时间无姿态视频中流式重建4D高斯模型的框架。给定一个连续的视频流，DynStream在局部时间窗口内重建场景，并逐步对齐和融合这些局部重建，形成一个全局一致的场景，从而实现无需逐场景优化的在线4D重建。通过共同强制跨窗口的几何一致性和建模时间变化的场景内容，DynStream支持在扩展的视频流上进行高效重建和逼真渲染。实验表明，DynStream能够从长视频流中实现高保真的在线动态重建和渲染，在多样的动态室内和室外场景中达到最先进的性能。
cs.CV / 83 / 2610.09723
MeshCarve: Artisan Mesh Generation with Flow Matching in Compact Latent Spaces
MeshCarve：在紧凑潜在空间中进行流匹配的工艺网格生成
Abstract
Prior artisan mesh generation works largely predict face tokens autoregressively, which makes inference slow. Recent methods instead flow match continuous latents built by Variational AutoEncoders (VAEs), but reconstruction quality drops significantly when geometry and topology are jointly encoded, and further when the latent space is compressed. We present MeshCarve, a flow matching method that generates entirely in compact latent spaces, generating vertex positions and edge connections separately and sidestepping the difficulty of a joint compact latent. To shorten the token sequence, we propose a hierarchical sparse transformer backbone, instantiated as VertexVAE and EdgeVAE. Instead of encoding fields over the surface voxels, both VAEs anchor on discrete vertices in their latent spaces, which drastically reduces the token sequence length, and our spatial-aware compression shortens it further without costing reconstruction. VertexVAE directly encodes vertex occupancy. For connectivity, we propose vertex-link encoding, which turns arbitrary connectivity between vertices into fixed-length continuous per-vertex embeddings and recovers complex artistic topology faithfully. MeshCarve combines these VAEs with an anchor generator and flow matches on the shortened token sequences. It shows advantages over state-of-the-art autoregressive and flow matching methods on Objaverse and generalizes to Toys4K. To the best of our knowledge, it is among the first artisan mesh generation methods whose every generative stage runs in a spatially compressed latent, with a token sequence only a fraction of the most compressed previous autoregressive and flow matching works.
Chinese Translation
以往的工艺网格生成工作主要是自回归地预测面令牌，这使得推理速度较慢。最近的方法则通过变分自编码器（Variational AutoEncoders, VAEs）流匹配构建的连续潜在空间，但当几何和拓扑共同编码时，重建质量显著下降，且在潜在空间被压缩时更为明显。我们提出了MeshCarve，这是一种在紧凑潜在空间中完全生成的流匹配方法，分别生成顶点位置和边连接，避免了联合紧凑潜在空间的困难。为了缩短令牌序列，我们提出了一种层次稀疏变换器骨干网络，具体实现为VertexVAE和EdgeVAE。两者不再对表面体素进行编码，而是将VAEs锚定在其潜在空间中的离散顶点上，这大大减少了令牌序列的长度，我们的空间感知压缩进一步缩短了序列而不影响重建。VertexVAE直接编码顶点占用情况。对于连通性，我们提出了顶点连接编码，将顶点之间的任意连通性转化为固定长度的连续每顶点嵌入，并忠实地恢复复杂的艺术拓扑。MeshCarve将这些VAEs与锚生成器结合，并在缩短的令牌序列上进行流匹配。它在Objaverse上表现出优于最先进的自回归和流匹配方法的优势，并能够推广到Toys4K。据我们所知，它是首批在空间压缩潜在空间中运行的工艺网格生成方法之一，其令牌序列仅为之前最压缩的自回归和流匹配工作的一个小部分。
cs.CV / 84 / 2610.09726
Beyond Group Splits: Specimen-Level Cross-Validation and Visual Attribution for Remaining-Shelf-Life Regression in Climacteric Fruit
超越组划分：用于气候果实剩余保质期回归的样本级交叉验证与视觉归因
Abstract
Estimating remaining shelf life (RSL) from images could provide affordable decision support for perishable produce, but evaluation protocols can substantially affect reported performance when repeated images are available from the same biological specimen. We use the Hass Avocado Ripening dataset, comprising 8,834 image-RSL pairs from 426 fruits across three storage regimes, to evaluate a frozen ImageNet-pretrained visual backbone with a lightweight regression head. Our contributions are threefold: we quantify the effect of observation-level versus specimen-disjoint evaluation, compare lightweight and heavier visual backbones under specimen-disjoint cross-validation, and examine their spatial attributions using Grad-CAM. Across ten observation-level random splits, the model achieves a mean RMSE of 2.37 days with a standard deviation of 0.03 days, whereas specimen-disjoint 5-fold cross-validation yields a mean RMSE of 3.12 days with a standard deviation of 0.11 days. The corresponding mean coefficient of determination is 0.553. A matched per-specimen comparison confirms higher error under specimen-disjoint evaluation, with a probability value below 0.001 across 426 specimens, showing that observation-level partitioning gives a substantially more optimistic estimate for this dataset and model configuration. Under specimen-disjoint evaluation, MobileNetV3-Small (0.93 million parameters) achieves accuracy comparable to ResNet-18 while providing substantially higher throughput, and Grad-CAM reveals differences in spatial attribution between the lightweight backbones. These results support specimen-disjoint evaluation and attribution analysis when assessing lightweight vision models for longitudinal shelf-life prediction.
Chinese Translation
从图像中估计剩余保质期（RSL）可以为易腐烂的农产品提供经济实惠的决策支持，但评估协议在同一生物样本可重复图像的情况下可能会显著影响报告的性能。我们使用哈斯鳄梨成熟数据集，该数据集包含来自426个水果的8,834对图像-RSL，涵盖三种储存方式，以评估一个经过ImageNet预训练的冻结视觉骨干网络及其轻量级回归头。我们的贡献有三方面：我们量化了观察级别与样本不重叠评估的影响，比较了在样本不重叠交叉验证下轻量级与重型视觉骨干网络的表现，并使用Grad-CAM检查它们的空间归因。在十次观察级别随机划分中，模型的平均均方根误差（RMSE）为2.37天，标准差为0.03天，而样本不重叠的5折交叉验证的平均RMSE为3.12天，标准差为0.11天。相应的决定系数平均值为0.553。匹配的每个样本比较确认在样本不重叠评估下误差更高，在426个样本中概率值低于0.001，显示观察级别划分为该数据集和模型配置提供了显著更乐观的估计。在样本不重叠评估下，MobileNetV3-Small（93万参数）实现了与ResNet-18相当的准确性，同时提供了显著更高的吞吐量，Grad-CAM揭示了轻量级骨干网络之间空间归因的差异。这些结果支持在评估轻量级视觉模型进行长期保质期预测时采用样本不重叠评估和归因分析。
cs.CV / 85 / 2610.09742
SoccerNet-FoulRet: Retrieving Semantically Similar Soccer Foul Videos
SoccerNet-FoulRet：检索语义相似的足球犯规视频
Abstract
Refereeing decisions in professional soccer remain inconsistent because referees cannot easily compare a contentious foul against similar past cases. We cast this as a retrieval problem and introduce SoccerNet-FoulRet, the first benchmark for semantic foul retrieval. Given a query foul, the task is to retrieve past fouls judged to be relevant precedents, regardless of camera angle, teams, or appearance. This differs from prior video-to-video retrieval, which matches clips by visual similarity or a shared event. Here, relevance is defined by refereeing interpretation. We build the benchmark from the SoccerNet-MVFoul dataset and evaluate retrieval ability of zero-shot video and vision-language embedders together with a task-specific fine-tuned baseline on 693 human-verified queries and category-relevance labels. Semantic foul retrieval remains challenging. The strongest zero-shot model achieves under 5% HitRate@10 on human-verified precedents, while category-supervised fine-tuning improves category relevance but transfers only modestly to precedent retrieval. We release SoccerNet-FoulRet to establish semantic foul retrieval as an open problem: https://github.com/SoccerNet/sn-foulret.
Chinese Translation
职业足球中的裁判判罚决策仍然不一致，因为裁判无法轻松地将有争议的犯规与类似的过去案例进行比较。我们将此视为一个检索问题，并引入SoccerNet-FoulRet，这是第一个用于语义犯规检索的基准。给定一个查询犯规，任务是检索被认为是相关先例的过去犯规，无论摄像机角度、球队或外观如何。这与之前的视频到视频检索不同，后者通过视觉相似性或共享事件来匹配片段。在这里，相关性由裁判的解释定义。我们从SoccerNet-MVFoul数据集中构建基准，并在693个经过人工验证的查询和类别相关性标签上评估零样本视频和视觉-语言嵌入器的检索能力，以及一个任务特定的微调基线。语义犯规检索仍然具有挑战性。最强的零样本模型在人工验证的先例上实现的HitRate@10低于5%，而类别监督微调提高了类别相关性，但对先例检索的转移效果仅为适度。我们发布SoccerNet-FoulRet，以将语义犯规检索确立为一个开放问题： https://github.com/SoccerNet/sn-foulret。
cs.CV / 86 / 2610.09746
Flow-of-Thought: A Framework for Visual Reasoning
思维流：一种视觉推理框架
Abstract
Mental imagery, ``seeing with the mind's eye'' is an essential aspect of human cognition. Despite rapid progress Large Language Models (LLMs) and Vision Transformers (ViTs) still underperform on tasks requiring spatial understanding. To address this, we introduce Flow-of-Thought (FoT), a framework that integrates the generation of visual sketches as intermediate reasoning steps, mimicking mental imagery in humans. We train coordinate-aware trajectory flow fields on $SO(2)$ group orbits and cumulative shortest paths, then freeze the learned dynamics; same vs. different decisions compare competing generative hypotheses using foreground-weighted reconstruction energy. On locked tests FoT reaches 100.0% accuracy on Tetris and 99.0% on colored shapes. Under frozen transfer, the orbit-trained 2D flow improves over its endpoint-only control on BLINK Multi-view (72.2% vs. 63.9% on 133 public validation pairs), supporting continuous visual traces as an effective and interpretable representation for spatial reasoning in some out-of-distribution settings.
Chinese Translation
心理意象，即“用心灵的眼睛看”，是人类认知的一个重要方面。尽管大型语言模型（LLMs）和视觉变换器（ViTs）取得了快速进展，但在需要空间理解的任务上仍表现不佳。为了解决这一问题，我们提出了思维流（Flow-of-Thought, FoT），这是一个将视觉草图生成作为中间推理步骤的框架，模仿人类的心理意象。我们在 $SO(2)$ 群轨道和累积最短路径上训练了坐标感知的轨迹流场，然后冻结学习到的动态；同类与不同类决策通过前景加权重建能量比较竞争生成假设。在锁定测试中，FoT 在俄罗斯方块（Tetris）上达到了 100.0% 的准确率，在彩色形状上达到了 99.0%。在冻结迁移下，轨道训练的二维流在 BLINK 多视图上优于仅端点控制（72.2% 对 63.9%，基于 133 个公共验证对），支持连续视觉轨迹作为某些分布外设置中空间推理的有效且可解释的表示。
cs.CV / 87 / 2610.09755
Diffusion-Generated Image Watermarking: A Two-Axis Taxonomy and Three Protocol-Bounded Case Studies
扩散生成图像水印：一个双轴分类法和三个协议约束的案例研究
Abstract
Watermarking diffusion-generated images requires balancing provenance signals with image quality, robustness, and computational cost. This work organizes methods along two axes: insertion mechanism and primary signal-bearing representation, and formalizes a representative $z_T$-Fourier pipeline for verification and identification. We then use the taxonomy to structure three protocol-bounded case studies. The first examines associations among frequency integrity, detection, quality, and cropping behavior. The second revisits persistence under seed-linked and seed-independent editing and formulates a scoped Semantic Imprinting Hypothesis without claiming a localized carrier or causal mechanism. The third studies single-shot VAE-latent phase modulation, including its efficiency, regeneration robustness, and robustness--quality operating points. Finally, we separate four content-level attack families from model/pipeline adaptation, propose corresponding evaluation protocols and testable conjectures for parameter-tuning threats, and identify additional temporal extensions for video. These analyses do not establish a universal ranking; instead, they provide a framework for matched, protocol-aware comparisons of watermarking systems for diffusion-generated images.
Chinese Translation
水印扩散生成图像需要在来源信号、图像质量、鲁棒性和计算成本之间取得平衡。本研究沿着两个轴对方法进行组织：插入机制和主要信号承载表示，并形式化了一个代表性的 $z_T$-傅里叶管道用于验证和识别。随后，我们利用该分类法构建了三个协议约束的案例研究。第一个案例研究考察了频率完整性、检测、质量和裁剪行为之间的关联。第二个案例研究重新审视了在种子链接和种子独立编辑下的持久性，并提出了一个范围明确的语义印记假设，而不声称存在局部载体或因果机制。第三个案例研究研究了单次拍摄的变分自编码器（VAE）潜在相位调制，包括其效率、再生鲁棒性和鲁棒性-质量操作点。最后，我们将四个内容级攻击家族与模型/管道适应分开，提出相应的评估协议和可测试的参数调优威胁猜想，并识别出视频的额外时间扩展。这些分析并未建立一个普遍的排名；相反，它们提供了一个框架，用于对扩散生成图像的水印系统进行匹配的、协议意识的比较。
cs.CV / 88 / 2610.09761
PARC-Loc: Text-to-Point-Cloud Localization with Partial Assignment and Relational Consistency
PARC-Loc：基于部分分配和关系一致性的文本到点云定位
Abstract
Text-to-point-cloud localization estimates a position in a city-scale 3D map from descriptions of surrounding objects. Existing coarse-to-fine methods retrieve submaps using aggregate learned compatibility and then localize within a selected submap. However, repetitive or similar urban objects can inflate the embedding similarity between the query and multiple submaps, even when the instance layout within a submap violates the query description. Meanwhile, query-relevant instances often span submap boundaries, leaving the retrieved submap with incomplete contextual evidence. We term these failure modes layout-inconsistent aliasing and boundary evidence incompleteness, respectively. To address them, we propose PARC-Loc, a coarse-to-fine localization framework built on Partial Assignment with Relational Consistency (PARC). PARC jointly models hint-object compatibility and pairwise spatial relations, allowing unmatched elements while favoring assignments consistent with the queried layout. At the coarse stage, its candidate-level assessment complements neural similarity for layout-consistent submap selection. At the fine stage, the context is expanded with query-relevant instances from adjacent submaps, while PARC yields object-level matching weights that guide cross-modal attention. Extensive experiments on KITTI360Pose and CityLoc show that PARC-Loc outperforms conventional coarse-to-fine baselines. On KITTI360Pose, our method improves Top-1 localization recall at 5 m from 0.50 to 0.67, achieving a 34% relative gain over the strongest baseline.
Chinese Translation
文本到点云定位通过周围物体的描述在城市规模的三维地图中估计位置。现有的粗到细方法使用聚合学习的兼容性检索子地图，然后在选定的子地图中进行定位。然而，重复或相似的城市物体可能会夸大查询与多个子地图之间的嵌入相似性，即使子地图内的实例布局违反了查询描述。同时，相关的查询实例往往跨越子地图边界，导致检索到的子地图缺乏完整的上下文证据。我们将这些失败模式称为布局不一致的别名和边界证据不完整。为了解决这些问题，我们提出了PARC-Loc，一个基于部分分配与关系一致性（Partial Assignment with Relational Consistency, PARC）的粗到细定位框架。PARC联合建模提示物体的兼容性和成对空间关系，允许不匹配的元素，同时优先考虑与查询布局一致的分配。在粗略阶段，其候选级评估补充了神经相似性，以实现布局一致的子地图选择。在细致阶段，背景通过来自相邻子地图的查询相关实例进行扩展，同时PARC产生的物体级匹配权重引导跨模态注意力。在KITTI360Pose和CityLoc上的大量实验表明，PARC-Loc优于传统的粗到细基线。在KITTI360Pose上，我们的方法将5米处的Top-1定位召回率从0.50提高到0.67，实现了相对于最强基线的34%的相对增益。
cs.CV / 89 / 2610.09780
Relational Abstractions for Spatial Reasoning with Diffusion Models
基于扩散模型的空间推理关系抽象
Abstract
Diffusion models excel at image synthesis, but they remain limited in their ability to reliably satisfy structured spatial reasoning constraints. In conditional data distribution modeling tasks with implicit logical structure, such as puzzles defined by visible clues paired with consistent solutions, state-of-the-art generative models tend to approximate pixel-space distributions without learning the underlying logical rules required for inference. To address this limitation, we present a novel framework for spatial reasoning with diffusion models that leverages unsupervised object discovery and abstractions of object relations. We show that the relational knowledge derived from object-centric representations enriches diffusion models with structural primitives, allowing them to effectively guide the generative representation space during both training and inference, and enabling conditional image generation that satisfies reasoning constraints. Additionally, we introduce a large-scale generative spatial reasoning benchmark with four datasets inspired by human-solvable puzzles. Our results show that relational abstractions significantly improve reasoning capabilities of diffusion models on a variety of complex reasoning tasks, while enabling robust generalization in out-of-distribution settings.
Chinese Translation
扩散模型在图像合成方面表现出色，但在可靠地满足结构化空间推理约束方面仍然存在局限性。在具有隐式逻辑结构的条件数据分布建模任务中，例如由可见线索与一致解决方案配对定义的难题，最先进的生成模型往往倾向于近似像素空间分布，而未能学习推理所需的基本逻辑规则。为了解决这一局限性，我们提出了一种新颖的扩散模型空间推理框架，该框架利用无监督的物体发现和物体关系的抽象。我们展示了从以物体为中心的表征中获得的关系知识丰富了扩散模型的结构原语，使其能够在训练和推理过程中有效地引导生成表征空间，并实现满足推理约束的条件图像生成。此外，我们引入了一个大型生成空间推理基准，包含四个受人类可解难题启发的数据集。我们的结果表明，关系抽象显著提高了扩散模型在多种复杂推理任务上的推理能力，同时在分布外设置中实现了稳健的泛化。
cs.CV / 90 / 2610.09784
CIRSeg: Coarse-to-Fine Intensity-Robust Liver Segmentation with Source-Free Continual Test-Time Adaptation
CIRSeg：基于无源持续测试时间适应的粗到细强度鲁棒肝脏分割
Abstract
Reliable liver segmentation in contrast-enhanced MRI is essential for quantitative hepatic assessment, treatment planning, and longitudinal disease monitoring. However, limited annotated data and scanner- or vendor-dependent intensity variations can cause overfitting and poor generalization to unseen acquisition domains. Moreover, simultaneously achieving robust global localization and precise boundary delineation remains challenging, while predictions may contain isolated false-positive regions outside the main liver component. To address these challenges, we propose CIRSeg, a coarse-to-fine, intensity-robust liver segmentation framework based on nnU-Netv2. CIRSeg combines 3D CutMix with stochastic intensity transfer using either Nyul augmentation or histogram matching to improve robustness to heterogeneous MRI intensities. Its cascaded architecture decouples low-resolution anatomical localization from full-resolution boundary refinement. At inference, source-free test-time adaptation based on confidence-filtered predictions and probability-prior regularization further improves robustness to out-of-distribution inputs. As a final deterministic post-processing step, largest connected component filtering removes isolated false-positive regions. On the CARE 2026 test set, CIRSeg achieves Dice scores of 97.13\% and 97.93\% on the in-domain and unseen-domain subsets, with corresponding HD95 values of 20.18 mm and 11.30 mm, respectively. These results demonstrate consistently accurate segmentation across both in-domain and unseen acquisition settings. The code is available at https://github.com/jingkunchen/MICCAI_CARE_2026
Chinese Translation
在对比增强MRI中，可靠的肝脏分割对于定量肝脏评估、治疗计划和长期疾病监测至关重要。然而，有限的标注数据以及扫描仪或厂商依赖的强度变化可能导致过拟合，并且在未见的采集领域中表现不佳。此外，同时实现鲁棒的全局定位和精确的边界描绘仍然具有挑战性，预测结果可能包含主要肝脏成分之外的孤立假阳性区域。为了解决这些挑战，我们提出了CIRSeg，一个基于nnU-Netv2的粗到细强度鲁棒肝脏分割框架。CIRSeg结合了3D CutMix与随机强度转移，使用Nyul增强或直方图匹配来提高对异质MRI强度的鲁棒性。其级联架构将低分辨率的解剖定位与全分辨率的边界细化解耦。在推理阶段，基于置信度过滤预测和概率先验正则化的无源测试时间适应进一步提高了对分布外输入的鲁棒性。作为最终的确定性后处理步骤，最大连通组件过滤去除了孤立的假阳性区域。在CARE 2026测试集上，CIRSeg在领域内和未见领域子集上分别达到了97.13%和97.93%的Dice分数，HD95值分别为20.18毫米和11.30毫米。这些结果展示了在领域内和未见采集设置下的一致准确分割。代码可在https://github.com/jingkunchen/MICCAI_CARE_2026获取。
cs.CV / 91 / 2610.09785
UltraWorld: Learning Interactive Ultrasound World Models from Untracked Clinical Videos with Acoustic Sampling Map
超世界：从未跟踪的临床视频中学习交互式超声世界模型与声学采样图
Abstract
World models can enable autonomous ultrasound scanning by predicting the outcomes of probe motions from local observations. Learning this action--observation relationship typically relies on synchronized video--pose pairs, which are costly to collect at scale and largely unavailable in routine clinical recordings. Reliable action following further requires modeling ultrasound's cross-sectional sampling geometry. We present UltraWorld, a self-distillation recipe that transfers priors from clinical ultrasound videos into interactive world models without real action annotations. Starting from clinical videos, we adapt a video foundation model into an ultrasound generator conditioned on reference images and anatomical masks. Anatomical masks sampled along programmable trajectories through 3D anatomy provide spatial guidance for synthesizing action--video pairs. We then use these synthetic pairs to self-distill the generator into a world model that predicts future observations from local observations and actions, without requiring anatomical masks or other 3D assets at inference time. To further improve action following, we introduce the Acoustic Sampling Map (AsMap), which represents probe poses and imaging settings as pixel-wise 3D sampling positions, beam directions, and depths. Experiments demonstrate improved prediction fidelity and action following. Across nine simulated closed-loop local planning episodes, UltraWorld reduces the mean final distance to the goal and orientation error by 29\% and 38\%, respectively, compared with visual servoing. Project Page: https://ultraworld-project.github.io/.
Chinese Translation
世界模型可以通过从局部观察中预测探头运动的结果来实现自主超声扫描。学习这种动作-观察关系通常依赖于同步的视频-姿态对，这在大规模收集时成本高昂，并且在常规临床记录中大多不可用。可靠的动作跟随进一步需要建模超声的横截面采样几何。我们提出了UltraWorld，一种自我蒸馏方法，将临床超声视频中的先验知识转移到交互式世界模型中，而无需真实的动作注释。从临床视频开始，我们将视频基础模型调整为基于参考图像和解剖掩膜的超声生成器。沿着可编程轨迹在三维解剖结构中采样的解剖掩膜为合成动作-视频对提供了空间指导。然后，我们使用这些合成对对生成器进行自我蒸馏，生成一个世界模型，该模型能够从局部观察和动作中预测未来观察，而在推理时不需要解剖掩膜或其他三维资产。为了进一步改善动作跟随，我们引入了声学采样图（Acoustic Sampling Map, AsMap），它将探头姿态和成像设置表示为逐像素的三维采样位置、波束方向和深度。实验表明，预测的保真度和动作跟随得到了改善。在九个模拟闭环局部规划实验中，与视觉伺服相比，UltraWorld将最终到达目标的平均距离和方向误差分别减少了29%和38%。项目页面：https://ultraworld-project.github.io/
cs.CV / 92 / 2610.09791
Counterfactual Route Optimization for Gaussian Head Avatar Modeling
高斯头部虚拟形象建模的反事实路径优化
Abstract
Head avatar modeling requires jointly optimizing multiple objectives with different dominant effects on geometry, appearance, and cross-view consistency. However, their relative effectiveness varies across training states, while existing pipelines typically rely on fixed loss weights or handcrafted stage-wise schedules. A central challenge is therefore to identify which optimization direction is more beneficial at each training state. We propose a counterfactual route optimization framework for Gaussian head avatar modeling, which characterizes state-dependent optimization preference from the realized effects of alternative updates rather than predefined heuristic weighting. Starting from the same training state, we perform short-horizon route-restricted lookahead over geometry, appearance, and joint update routes and evaluate their outcomes under a unified utility. The resulting counterfactual evidence is factorized into a geometry--appearance preference and a residual joint advantage, separately capturing the relative preference between individual update directions and the additional benefit of coordinated optimization. We further amortize this offline evidence into a lightweight controller that directly estimates the current optimization preference and applies bounded modulation to the training objectives during full avatar optimization. Experiments on the NeRSemble dataset validate the effectiveness of the proposed design, consistently outperforming existing methods while preserving clearer local facial structures and finer details.
Chinese Translation
头部虚拟形象建模需要联合优化多个目标，这些目标对几何形状、外观和跨视图一致性具有不同的主导影响。然而，它们的相对有效性在不同的训练状态下有所变化，而现有的流程通常依赖于固定的损失权重或手工设计的阶段性调度。因此，一个核心挑战是识别在每个训练状态下哪个优化方向更具优势。我们提出了一种用于高斯头部虚拟形象建模的反事实路径优化框架，该框架基于替代更新的实际效果来表征状态依赖的优化偏好，而不是预定义的启发式权重。从相同的训练状态出发，我们在几何形状、外观和联合更新路径上执行短期路径限制的前瞻性评估，并在统一效用下评估其结果。得到的反事实证据被分解为几何形状-外观偏好和残余联合优势，分别捕捉个别更新方向之间的相对偏好和协调优化的额外收益。我们进一步将这些离线证据转化为一个轻量级控制器，该控制器直接估计当前的优化偏好，并在完整的虚拟形象优化过程中对训练目标施加有界调节。在NeRSemble数据集上的实验验证了所提设计的有效性，始终优于现有方法，同时保留了更清晰的局部面部结构和更精细的细节。
cs.CV / 93 / 2610.09800
Beyond Masks and Trajectories: Flow-Guided Latent Action Injection for Stable Surgical Video Generation
超越掩膜与轨迹：基于流引导的潜在动作注入用于稳定的外科视频生成
Abstract
Surgical video generation holds substantial potential for surgical education, simulation, and data augmentation, yet generating surgical videos with realistic and clinically plausible motion remains challenging. Most existing methods rely on auxiliary conditions, such as masks, trajectories, depth, or reference videos, to achieve visually plausible synthesis. Yet, these auxiliary conditions typically require additional manual annotation or specialized acquisition, making it difficult to scale such methods beyond small, curated datasets. This motivates the need for a reference-free architecture capable of generating high-quality surgical video without requiring auxiliary visual conditions at inference time. We propose FLAIR, a Flow-guided LatentAction Injection framework for Reference-free surgical video generation. FLAIR learns action priors from optical flow of real surgical videos, dynamically predicts corresponding latent action representation from an input prompt, and injects it into a frozen base model to generate surgical videos with improved action consistency. We further construct SurgActionClip-30K, the first large-scale surgical vision dataset comprising action-centric segmented clips and structured caption labels, addressing the persistent lack of fine-grained, action-centric surgical datasets. Lastly, we introduce SurgMetrics, the first surgical domain-specific evaluation metrics for quantifying the quality of generated surgical videos, addressing the persistent absence of clinically grounded evaluation standards in this domain. Extensive experiments demonstrate that FLAIR enables generating high-quality surgical videos using text-only inference without auxiliary conditions, and validation in SurgMetrics demonstrates its strength in alignment with human perception compared to traditional metrics.
Chinese Translation
外科视频生成在外科教育、模拟和数据增强方面具有巨大的潜力，但生成具有现实和临床合理运动的外科视频仍然具有挑战性。现有的大多数方法依赖于辅助条件，如掩膜、轨迹、深度或参考视频，以实现视觉上合理的合成。然而，这些辅助条件通常需要额外的手动标注或专业获取，使得这些方法难以超越小规模的精心策划数据集。这促使我们需要一种无参考架构，能够在推理时生成高质量的外科视频，而无需辅助视觉条件。我们提出了FLAIR（Flow-guided Latent Action Injection framework），一种用于无参考外科视频生成的框架。FLAIR从真实外科视频的光流中学习动作先验，动态预测来自输入提示的相应潜在动作表示，并将其注入到一个固定的基础模型中，以生成具有更好动作一致性的外科视频。我们进一步构建了SurgActionClip-30K，这是第一个大规模外科视觉数据集，包含以动作为中心的分段剪辑和结构化标题标签，解决了细粒度、以动作为中心的外科数据集持续匮乏的问题。最后，我们引入了SurgMetrics，这是第一个特定于外科领域的评估指标，用于量化生成外科视频的质量，解决了该领域缺乏临床基础评估标准的问题。大量实验表明，FLAIR能够仅通过文本推理生成高质量的外科视频，而无需辅助条件，并且在SurgMetrics中的验证展示了其在与人类感知对齐方面相较于传统指标的优势。
cs.CV / 94 / 2610.09802
DisParQ: Self-Supervised Part Concepts for Interpretable Vision Foundation Models
DisParQ：用于可解释视觉基础模型的自监督部分概念
Abstract
Concept-based vision models represent images through an intermediate layer of human-inspectable concepts, so what a model relies on can be traced to those concepts. However, those models are often limited to fixed categories or depend on language to define their concepts. We introduce DisParQ (Discrete Parts with Quantized attributes), a method that learns spatially grounded, discrete concept representations from a powerful frozen vision-only self-supervised backbone. It requires no class labels and no language supervision. Each image patch is assigned to exactly one concept from a learnable prototype dictionary, and only a sparse subset of concepts may activate per image. To capture how each concept varies across images (e.g., the type of a "wheel"), we learn continuous residuals alongside the concepts and then quantize them into discrete attributes. A spatial decoder reconstructs the backbone's representation from the concepts and attributes alone, so successful reconstruction means that the discrete representation preserves the backbone's information. We evaluate DisParQ across seven datasets, from general recognition (ImageNet, PartImageNet, Places) to fine-grained benchmarks (CUB, Cars, Dogs, Flowers). We show that DisParQ closely matches its frozen DINOv2 teacher on ImageNet linear probing (83.2% top-1), achieves higher concept consistency than language-aligned models, remains competitive on fine-grained recognition, and enables cross-category part-based retrieval.
Chinese Translation
基于概念的视觉模型通过一层人类可检视的概念中介来表示图像，因此模型所依赖的内容可以追溯到这些概念。然而，这些模型通常局限于固定的类别或依赖语言来定义其概念。我们提出了DisParQ（离散部分与量化属性），一种从强大的冻结视觉自监督骨干网络中学习空间定位的离散概念表示的方法。它不需要类别标签和语言监督。每个图像块被分配给可学习原型字典中的一个概念，并且每个图像可能仅激活一小部分概念。为了捕捉每个概念在不同图像中的变化（例如，“轮子”的类型），我们在学习概念的同时学习连续残差，然后将其量化为离散属性。空间解码器仅根据概念和属性重建骨干网络的表示，因此成功的重建意味着离散表示保留了骨干网络的信息。我们在七个数据集上评估DisParQ，从一般识别（ImageNet、PartImageNet、Places）到细粒度基准（CUB、Cars、Dogs、Flowers）。我们展示了DisParQ在ImageNet线性探测中与其冻结的DINOv2教师模型（83.2% top-1）紧密匹配，获得比语言对齐模型更高的概念一致性，在细粒度识别中保持竞争力，并实现跨类别的基于部分的检索。
cs.CV / 95 / 2610.09807
Concentration, Not Uncertainty: Why Targeted Synthetic Data Doesn't Help Camouflaged Object Detection
集中，而非不确定性：为何针对性合成数据对伪装物体检测无益
Abstract
Camouflaged object detection requires pixel-accurate masks, but obtaining such annotations is slow and costly, making synthetic training images an attractive alternative. Under a fixed generation budget, however, it remains unclear which real-image regions to target for synthetic data generation. We study an uncertainty-guided generation strategy that clusters the unlabelled real images, identifies clusters on which the model is least certain, allocates synthetic generation toward those clusters, and iteratively retrains the model. Across 103 training runs, uncertainty-based targeting does not outperform random allocation. Five independent controls further show that this null result is not an artifact: targeted training sets are measurably different from random sets, but the difference is explained by concentrating the generation budget rather than by where uncertainty is concentrated, as every concentration rule we test reproduces the effect and, on boundary accuracy, so does aiming at the clusters the model was most certain about. Separately, we find substantial data contamination in CHAMELEON, with 50 of its 76 images duplicated from training data despite the standard overlap check reporting zero overlap. Together, these results show that, under a fixed synthetic-data budget, budget concentration, not uncertainty-based targeting, accounts for the observed training-set effects.
Chinese Translation
伪装物体检测需要像素级精确的掩膜，但获取此类标注既耗时又昂贵，使得合成训练图像成为一种有吸引力的替代方案。然而，在固定的生成预算下，仍不清楚应针对哪些真实图像区域进行合成数据生成。我们研究了一种基于不确定性的生成策略，该策略对未标记的真实图像进行聚类，识别模型最不确定的聚类，将合成生成分配给这些聚类，并迭代地重新训练模型。在103次训练中，基于不确定性的目标分配并未优于随机分配。五个独立的控制实验进一步表明，这一无效结果并非伪影：针对性训练集与随机集在可测量上存在差异，但这种差异是通过集中生成预算来解释的，而非不确定性集中所在，因为我们测试的每一种集中规则都重现了这一效应，并且在边界准确性上，针对模型最确定的聚类同样产生了效果。此外，我们发现CHAMELEON数据集中存在显著的数据污染，其中76张图像中有50张是从训练数据中重复的，尽管标准重叠检查报告为零重叠。综合来看，这些结果表明，在固定的合成数据预算下，预算集中，而非基于不确定性的目标分配，解释了观察到的训练集效应。
cs.CV / 96 / 2610.09821
Efficient 3D Gaussian Head Avatars for Edge Devices
高效的边缘设备3D高斯头像
Abstract
Generative 3D Gaussian head avatars provide high-quality, efficient rendering, but synthesising the Gaussian representation remains computationally expensive, limiting deployment on resource-constrained and edge devices. We introduce an efficient generator architecture for unconditional 3D Gaussian head synthesis, based on a parameter-efficient synthesis block and depth-wise separable convolutions while retaining style-based conditioning. Our architecture reduces generator complexity without requiring model compression or quantisation. Compared with the baseline model, our approach reduces FLOPs by 94%, parameter count by 70%, and model size by 81%, while maintaining competitive generation quality. We further demonstrate practical CPU inference and browser-based execution on mobile devices using ONNX Runtime, enabling 3D Gaussian avatar synthesis without dedicated GPU hardware or application-specific software. In addition to conventional image-quality metrics, we evaluate multi-view consistency, training cost, and deployment performance. Code, trained models, and evaluation tools will be released publicly.
Chinese Translation
生成的3D高斯头像提供高质量、高效的渲染，但合成高斯表示仍然计算开销较大，限制了在资源受限和边缘设备上的部署。我们提出了一种高效的无条件3D高斯头像合成生成器架构，该架构基于参数高效的合成模块和深度可分离卷积，同时保留基于风格的条件。我们的架构在不需要模型压缩或量化的情况下减少了生成器的复杂性。与基线模型相比，我们的方法将FLOPs减少了94%，参数数量减少了70%，模型大小减少了81%，同时保持了具有竞争力的生成质量。我们进一步展示了在移动设备上使用ONNX Runtime进行实际的CPU推理和基于浏览器的执行，从而实现了无需专用GPU硬件或特定应用软件的3D高斯头像合成。除了传统的图像质量指标外，我们还评估了多视角一致性、训练成本和部署性能。代码、训练模型和评估工具将公开发布。
cs.CV / 97 / 2610.09823
UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation
UltraText Bench：评估图像生成中视觉文本渲染的综合双语基准
Abstract
Dense visual text requires image generators to reproduce long strings across multiple regions with correct placement and legibility. As short-string rendering improves, evaluation must test sustained performance across more demanding scenes. We introduce UltraText Bench, a bilingual benchmark for prompt-only generation of dense visual text. It contains 432 prompts spanning 24 real-world scene categories and three difficulty levels, split equally between English and Chinese. Each human-reviewed prompt supplies exact strings for four to twelve text regions, paired with structured references for their content, placement, and visual attributes. We use the Q-Judger vision-language model to assess each image against the complete reference, reporting text fidelity, text clarity, spatial quality, and scene quality. Across 24 model configurations, these dimensions reveal different strengths: Z-Image-Turbo gains 3.81 clarity points over Z-Image-Base while losing 14.76 fidelity points under the reported settings. Performance also varies with workload; Qwen-Image-2512's English composite falls from 86.50 at L1 to 42.86 at L3. Ten participants took part in human evaluation of the automatic scores. Repository: https://github.com/LINs-lab/UltraText_Bench.
Chinese Translation
密集视觉文本要求图像生成器在多个区域中正确放置和清晰呈现长字符串。随着短字符串渲染的改进，评估必须测试在更具挑战性的场景中的持续性能。我们推出了UltraText Bench，这是一个用于仅基于提示生成密集视觉文本的双语基准。它包含432个提示，涵盖24个真实场景类别和三个难度级别，英语和中文各占一半。每个经过人工审核的提示提供了四到十二个文本区域的确切字符串，并配有其内容、位置和视觉属性的结构化参考。我们使用Q-Judger视觉-语言模型对每个图像进行评估，与完整参考进行对比，报告文本保真度、文本清晰度、空间质量和场景质量。在24种模型配置中，这些维度揭示了不同的优势：Z-Image-Turbo在报告的设置下比Z-Image-Base提高了3.81个清晰度点，但损失了14.76个保真度点。性能也因工作负载而异；Qwen-Image-2512的英语综合得分从L1的86.50降至L3的42.86。十名参与者参与了对自动评分的人工评估。代码库链接： https://github.com/LINs-lab/UltraText_Bench。
cs.CV / 98 / 2610.09830
MOTIF: Person-of-Interest Deepfake Detection Beyond 3DMM Coefficients
MOTIF：超越3DMM系数的人物深度伪造检测
Abstract
Video deepfakes targeting a specific individual, the Person-of-Interest (POI), are the most harmful ones, and, since a public figure is abundantly recorded, a detector can be built from genuine footage of that individual. Such detectors commonly describe a subject through a 3D Morphable Model (3DMM) and adopt its coefficients as a whole, so which part of that description carries the signal has never been measured. We dissect it, holding the encoder, the training corpus and the enrollment protocol fixed and varying only what the encoder observes. The groups of coefficients prove largely redundant, since the shape block alone recovers almost all the accuracy of the full vector, and their temporal evolution contributes a real but bounded amount. We further show that the dense surface the same fit returns, which these detectors discard, carries identity information that the coefficients do not, and that it helps precisely where they are weakest. We assemble the best configuration into MOTIF, a visual-only detector trained on real videos only, with no manipulated video and no POI-specific data. It improves on both state-of-the-art POI detectors in every dataset and manipulation of our benchmark and at two quality levels. Our experimental code will be released at https://github.com/polimi-ispl/MOTIF.
Chinese Translation
针对特定个体（即人物关注对象，Person-of-Interest，POI）的视频深度伪造是最具危害性的，由于公共人物的录制资料丰富，因此可以基于该个体的真实视频构建检测器。这类检测器通常通过3D可变形模型（3D Morphable Model，3DMM）来描述对象，并将其系数作为整体使用，但从未测量过该描述的哪一部分携带信号。我们对其进行剖析，固定编码器、训练语料库和注册协议，仅改变编码器观察的内容。系数组证明在很大程度上是冗余的，因为仅形状块几乎恢复了完整向量的所有准确性，而它们的时间演变贡献了真实但有限的量。我们进一步表明，相同拟合返回的密集表面，这些检测器通常会忽略，携带了系数所不具备的身份信息，并且在系数最弱的地方提供了帮助。我们将最佳配置整合为MOTIF，这是一个仅基于真实视频训练的视觉检测器，没有操纵视频和POI特定数据。它在我们基准的每个数据集和操纵中，以及在两个质量水平上，都优于现有的POI检测器。我们的实验代码将发布在https://github.com/polimi-ispl/MOTIF。
cs.CV / 99 / 2610.09841
ORCA: Hunting Compositional Failures in Text-to-Image Diffusion
ORCA：在文本到图像扩散中捕捉组合失败
Abstract
Text-to-image diffusion models fail predictably on compositional prompts: attributes bind to the wrong objects, spatial relations invert, and multi-object scenes lose count. Recent architectures already augment CLIP with a T5 encoder precisely because CLIP's contrastive embedding loses compositional structure, yet these failures persist. We argue the binding problem is therefore not one of missing information but of misaligned information: a text encoder preserves compositional structure, but in a representation space shaped by language modelling rather than vision, and the denoising objective does not directly reward aligning the two. We show this correspondence can be supplied as an explicit training signal, that the relevant cross-modal information is concentrated in a low-rank subspace of self-supervised visual features, and that supplying it can be folded into diffusion training as a single auxiliary loss. Our method, ORCA (Orthogonal Residual Compositional Alignment), aligns the latent of a diffusion transformer with a low-rank target derived from a frozen visual encoder, through a predictor whose orthogonal basis is parameterised by a learned residual between T5 and CLIP embeddings, which provides a prompt-dependent signal for selecting the visual readout subspace. We prove that the cross-modal information recoverable at a given rank is bounded by the spectral mass of the visual encoder's covariance in the top components. Across three diffusion-transformer backbones (DiT-B/2, DiT-L/2, U-ViT-L), ORCA improves FID and GenEval over both vanilla and REPA baselines at zero inference-time cost; on DiT-L/2 it reaches FID 16.65 and GenEval 0.291 at 200K steps, exceeding the strongest 400K baseline at half the training cost, with the largest gains concentrated on attribute binding, spatial relations, and multi-object prompts.
Chinese Translation
文本到图像的扩散模型在组合提示上表现出可预测的失败：属性绑定到错误的对象，空间关系颠倒，多对象场景失去计数。最近的架构已经通过将 CLIP 与 T5 编码器结合使用来增强其性能，正是因为 CLIP 的对比嵌入丢失了组合结构，但这些失败依然存在。我们认为绑定问题并不是缺失信息，而是信息不对齐：文本编码器保留了组合结构，但在一个由语言建模而非视觉塑造的表示空间中，去噪目标并没有直接奖励将两者对齐。我们展示了这种对应关系可以作为显式训练信号提供，相关的跨模态信息集中在自监督视觉特征的低秩子空间中，并且提供它可以作为单一辅助损失融入扩散训练中。我们的方法 ORCA（正交残差组合对齐）通过一个预测器将扩散变换器的潜在表示与来自冻结视觉编码器的低秩目标对齐，该预测器的正交基由 T5 和 CLIP 嵌入之间的学习残差参数化，为选择视觉读出子空间提供了依赖于提示的信号。我们证明了在给定秩下可恢复的跨模态信息受到视觉编码器协方差在主要成分中的谱质量的限制。在三个扩散变换器骨干网络（DiT-B/2, DiT-L/2, U-ViT-L）上，ORCA 在零推理时间成本的情况下提高了 FID 和 GenEval，相较于普通和 REPA 基线；在 DiT-L/2 上，它在 200K 步时达到了 FID 16.65 和 GenEval 0.291，超越了以一半训练成本获得的最强 400K 基线，最大的增益集中在属性绑定、空间关系和多对象提示上。
cs.CV / 100 / 2610.09849
Hard, Yet Reducible: Controlled Forward Transfer for Synthetic Degradation Curation
困难但可简化：合成降解策划的受控前向转移
Abstract
Selecting synthetic degradations for dense prediction requires an estimate of their training utility, the generalization gain they bring under a finite training budget. Clean and degraded twins share content and labels, suggesting a score based on how much short training reduces the excess error caused by degradation. However, this gap can also shrink when clean performance deteriorates. Measuring the improvement on degraded images alone avoids that confound, but it still credits progress that the same amount of clean training would have produced. We propose the \textbf{controlled Reducible Degradation Gap} (cRDG) for regions defined by degradation type and severity. From a common checkpoint, cRDG runs two budget-matched probes that differ only in one augmentation slot, which holds either a synthetic degradation or a clean augmentation. The score is the gain on held-out degraded images relative to the clean-control probe. Clean harm is a separate feasibility constraint. cRDG reveals a correctable severity band in which training on the degradation yields high controlled gain under the available budget, and the band moves with the predictor, the starting checkpoint, and the training budget. \textbf{Curation of Reducible Bands} (\method) uses cRDG to select synthetic data without changing the predictor. On semantic segmentation and salient object detection, \method{} improves representative predictors under matched synthetic-data budgets and training schedules, extends to existing data-generation pipelines, and preserves clean performance. Code and supporting materials will be publicly released.
Chinese Translation
选择用于密集预测的合成降解需要估计其训练效用，即在有限训练预算下带来的泛化增益。干净和降解的双胞胎共享内容和标签，这表明可以基于短期训练减少因降解造成的过剩误差的程度来评分。然而，当干净性能恶化时，这一差距也可能缩小。仅测量降解图像上的改进可以避免这种混淆，但它仍然归功于同样数量的干净训练所能产生的进展。我们提出了 extbf{受控可简化降解差距}（cRDG），用于由降解类型和严重程度定义的区域。从一个共同的检查点，cRDG运行两个预算匹配的探测器，它们仅在一个增强插槽中有所不同，该插槽持有合成降解或干净增强。评分是相对于干净控制探测器在保留的降解图像上的增益。干净损害是一个单独的可行性约束。cRDG揭示了一个可纠正的严重性带，在该带内，针对降解的训练在可用预算下产生高受控增益，并且该带随着预测器、起始检查点和训练预算的变化而移动。 extbf{可简化带的策划}（ extmethod）利用cRDG选择合成数据，而不改变预测器。在语义分割和显著性目标检测中， extmethod{}在匹配的合成数据预算和训练计划下改善了代表性预测器，扩展到现有数据生成管道，并保持干净性能。代码和支持材料将公开发布。
cs.CV / 101 / 2610.09853
DeltaSplat: Iterative Gaussian Refinement for Pose-Free Feed-Forward 3D Gaussian Splatting
DeltaSplat：无姿态前馈3D高斯点云的迭代高斯精细化
Abstract
Pose-free feed-forward 3D Gaussian Splatting (3DGS) reconstructs a scene from sparse, unposed images in a single network pass, removing the need for camera calibration and per-scene optimization. However, camera estimation errors propagate into the predicted Gaussians and compound the geometric and photometric inaccuracies of single-pass prediction. To correct these errors, we introduce DeltaSplat, a lightweight Gaussian refinement module for pose-free feed-forward 3DGS. It iteratively renders the current Gaussians at the input context views and predicts per-Gaussian updates from the resulting residuals. A 2D residual alone, however, underdetermines the 3D correction. DeltaSplat therefore conditions each update on per-pixel Pl\"ucker rays and rendered depth as a soft geometric prior. A dual-branch convolutional mixer efficiently encodes these inputs, and per-attribute heads decode the fused features into position, opacity, and color updates. The module adds only ~2.2% parameters to the backbone and remains fully feed-forward at inference. On DL3DV, DeltaSplat reaches 26.64 dB PSNR in the pose-free setting, improving its state-of-the-art backbone by 1.75 dB and surpassing even baselines supplied with ground-truth cameras; consistent gains hold across 6-24 views and all camera regimes.
Chinese Translation
无姿态前馈3D高斯点云（3DGS）通过单次网络传递从稀疏的无姿态图像中重建场景，消除了对相机标定和每个场景优化的需求。然而，相机估计误差会传播到预测的高斯分布中，并加剧单次预测的几何和光度不准确性。为了解决这些误差，我们引入了DeltaSplat，一个轻量级的高斯精细化模块，适用于无姿态前馈3DGS。该模块在输入上下文视图中迭代渲染当前高斯，并从结果残差中预测每个高斯的更新。然而，仅依赖2D残差不足以确定3D修正。因此，DeltaSplat将每个更新条件化为每个像素的Plücker光线和渲染深度，作为一种软几何先验。一个双分支卷积混合器有效地编码这些输入，而每个属性头则将融合特征解码为位置、不透明度和颜色更新。该模块仅为主干网络增加约2.2%的参数，并在推理时保持完全前馈。在DL3DV数据集上，DeltaSplat在无姿态设置中达到了26.64 dB的PSNR，较其最先进的主干网络提高了1.75 dB，甚至超越了使用真实相机的基线；在6-24个视图和所有相机模式下，均保持一致的性能提升。
cs.CV / 102 / 2610.09860
DeepTopoClustering: Unsupervised Derivation of Surface Process Taxonomy from 4D Point Clouds for Topographic Monitoring
DeepTopoClustering：从4D点云中无监督推导地表过程分类法以进行地形监测
Abstract
4D point clouds acquired by permanent laser scanning (PLS) enable accurate high-frequency monitoring of surface change in dynamic topographic environments. However, existing methods remain limited in organizing detected surface activities into meaningful process types. We propose DeepTopoClustering (DTC), an unsupervised framework for deriving a hierarchical process taxonomy from object-based surface activities, so-called 4D objects-by-change (4D-OBCs). We transform each 4D-OBC into a GeoMorphogram, a distributional sequence representing the temporal evolution of topographic change within a spatially bounded surface activity. A convolutional autoencoder learns latent embeddings from GeoMorphograms, which are jointly optimized using a hierarchical deep clustering objective to organize surface activities into a hierarchy. We evaluate the learned hierarchy using expert annotations on two 4D datasets of sandy beach sites and their combination. DTC with GeoMorphograms achieves the highest agreement with expert judgment at the taxonomy level comprising eight major process types ($F_1=0.78$, match accuracy $=0.92$), outperforming dimensionality reduction and conventional flat clustering. The learned taxonomy separates major erosion- and deposition-dominated activities and distinguishes finer subtypes based on change magnitude, duration, compactness, and temporal evolution. DTC thus provides a scalable and interpretable route from 4D change detection to a data-driven, expert-supported surface process taxonomy, advancing automated knowledge derivation for understanding surface dynamics in topographic monitoring.
Chinese Translation
通过永久激光扫描（PLS）获取的4D点云能够在动态地形环境中实现高频率的地表变化精确监测。然而，现有方法在将检测到的地表活动组织成有意义的过程类型方面仍然有限。我们提出了DeepTopoClustering（DTC），这是一个无监督框架，用于从基于对象的地表活动中推导出层次化的过程分类法，称为4D变化对象（4D-OBCs）。我们将每个4D-OBC转化为GeoMorphogram，这是一种分布序列，表示在空间边界内的地表活动中地形变化的时间演变。卷积自编码器从GeoMorphograms中学习潜在嵌入，并使用层次深度聚类目标进行联合优化，以将地表活动组织成一个层次结构。我们使用专家注释对两个沙滩地点的4D数据集及其组合中学习到的层次结构进行评估。DTC与GeoMorphograms在包含八种主要过程类型的分类层次上与专家判断的最高一致性（$F_1=0.78$，匹配准确率$=0.92$），优于降维和传统平面聚类。学习到的分类法区分了主要的侵蚀和沉积主导活动，并根据变化幅度、持续时间、紧凑性和时间演变区分更细的亚类型。因此，DTC提供了一条从4D变化检测到数据驱动、专家支持的地表过程分类法的可扩展且可解释的路径，推动了对地形监测中地表动态理解的自动化知识推导。
cs.CV / 103 / 2610.09863
Global Average Precision for Representation Learning
用于表示学习的全局平均精度
Abstract
Standard information retrieval metrics, such as mean Average Precision (mAP), assess performance one query at a time, based on how the similarities between a query and its positives compare against those with its negatives. The same holds for common representation learning losses, such as InfoNCE and per-query AP surrogates. None of them considers whether similarities are comparable across queries, which any system with a single decision threshold relies on. Global Average Precision (gAP) does, by ranking all query-candidate pairs in one list and computing a single AP. We introduce gSAP, a differentiable surrogate of gAP. It needs only a similarity matrix and a binary matrix marking the positive pairs, the same input as existing losses, so it is a drop-in replacement for them and agnostic to the encoder, the modality, and the source of supervision. Since it considers all possible pairwise comparisons in the batch jointly, it also remains trainable at low temperatures, a regime where per-query surrogates run out of gradient. Swapping it into established recipes improves supervised metric learning, cross-modal alignment, and self-supervised pretraining, where, to our knowledge, it is the first ranking loss to replace the community standard InfoNCE in the latter two. Its similarities are more consistent across queries, which drives the gains under a universal threshold. gSAP retrieves up to four times as many positive pairs as the strongest AP surrogate at the same precision, and it degrades the least when queries with no positives in the database are added. Beyond thresholding, models trained with gSAP also learn better representations, with higher transfer, $k$NN and zero-shot classification accuracy.
Chinese Translation
标准的信息检索指标，如平均精度均值（mean Average Precision, mAP），是通过比较查询与其正样本之间的相似性与查询与其负样本之间的相似性来逐个查询评估性能的。常见的表示学习损失，如InfoNCE和逐查询平均精度（per-query AP）替代品也是如此。它们都没有考虑不同查询之间的相似性是否可比，而任何依赖单一决策阈值的系统都需要这一点。全局平均精度（Global Average Precision, gAP）通过将所有查询-候选对排名在一个列表中并计算单一的平均精度（AP）来解决这个问题。我们引入了gSAP，这是gAP的一个可微分替代品。它只需要一个相似性矩阵和一个标记正样本对的二进制矩阵，这与现有损失的输入相同，因此可以作为它们的直接替代，并且与编码器、模态和监督源无关。由于它联合考虑批次中的所有可能成对比较，因此在低温度下仍然可以训练，而在此情况下，逐查询替代品会失去梯度。将其替换到现有方法中可以改善监督度量学习、跨模态对齐和自监督预训练。在我们所知的范围内，它是第一个在后两者中替代社区标准InfoNCE的排名损失。它在不同查询之间的相似性更加一致，这推动了在通用阈值下的增益。gSAP在相同精度下检索的正样本对数量是最强AP替代品的四倍，并且在添加没有正样本的查询时降级最小。超越阈值，使用gSAP训练的模型还学习到了更好的表示，具有更高的迁移能力、$k$NN和零样本分类准确率。
cs.CV / 104 / 2610.09873
Bringing BNNs to Fast Event Processing
将二进制神经网络应用于快速事件处理
Abstract
Binary Neural Networks (BNNs) enable efficient deep learning deployment on resource constrained devices with weights and activations compressed to one bit, substantially reducing model size and inference cost. Event cameras offer complementary advantages, including low latency, high dynamic range, and low power consumption, by capturing asynchronous streams of events rather than dense image frames. Despite their shared emphasis on efficiency, the combination of these technologies remains largely unexplored. This work aims at adapting and evaluating modern deep BNN architectures on event data. We also show that cross-modal pretraining from RGB data can improve the classification accuracy of BNNs on neuromorphic datasets. We introduce the Polar-wise Binary Event Volume (PBEV), a binary representation that enables event-camera data to be processed directly by BNNs and represents a step toward fully binarized event-based vision systems. Best evaluated BNN on N-Caltech101 classification benchmarks shows 90.58% accuracy with 7.5x less operations than their full-precision counterparts.
Chinese Translation
二进制神经网络（BNNs）使得在资源受限设备上高效部署深度学习成为可能，其权重和激活被压缩为一位，从而显著减少模型大小和推理成本。事件摄像头通过捕捉异步事件流而非密集图像帧，提供了低延迟、高动态范围和低功耗等互补优势。尽管这两种技术都强调效率，但它们的结合仍然在很大程度上未被探索。本研究旨在对现代深度BNN架构在事件数据上的适应性和评估进行探讨。我们还展示了从RGB数据进行跨模态预训练可以提高BNNs在神经形态数据集上的分类准确性。我们引入了极性二进制事件体积（Polar-wise Binary Event Volume, PBEV），这是一种二进制表示，能够使事件摄像头数据直接被BNNs处理，代表了朝着完全二进制化的基于事件的视觉系统迈出的一步。在N-Caltech101分类基准测试中，最佳评估的BNN显示出90.58%的准确率，且操作量比其全精度对应物少7.5倍。
cs.CV / 105 / 2610.09875
SANet: Selective Attention Network for Infrared Small Target Detection
SANet：用于红外小目标检测的选择性注意力网络
Abstract
Infrared small target detection aims to accurately identify and locate dim targets in complex backgrounds and supports applications such as maritime surveillance and military search and rescue. However, the small size and weak contrast of infrared targets make it difficult to balance detection accuracy and false alarms. This paper proposes a selective attention network (SANet) for infrared small target detection. A dual-path semantic-aware module combines standard and pinwheel-shaped convolutions to preserve local spatial consistency and capture broader contextual information. Spatial and channel attention further refine the features and improve target-background discrimination. To address the limitations of static skip connections in U-Net, a selective attention fusion module adaptively integrates features across scales using spatially varying weights. It selectively enhances salient regions and improves discrimination between true targets and false alarms. Experiments on three public benchmarks, NUAA-SIRST, IRSTD-1K, and NUDT-SIRST, show that SANet achieves competitive performance in intersection over union (IoU), normalized IoU, detection probability, and false alarm rate. Its IoU exceeds that of the second-best method by 1.93, 4.32, and 2.21 percentage points, respectively. These results support the effectiveness of SANet in dim-target perception, discriminative feature representation, and background suppression.
Chinese Translation
红外小目标检测旨在准确识别和定位复杂背景下的微弱目标，并支持海洋监视和军事搜救等应用。然而，红外目标的小尺寸和弱对比度使得平衡检测精度和误报变得困难。本文提出了一种用于红外小目标检测的选择性注意力网络（SANet）。一个双路径语义感知模块结合了标准卷积和旋风形卷积，以保持局部空间一致性并捕捉更广泛的上下文信息。空间和通道注意力进一步细化特征并提高目标与背景的区分度。为了解决U-Net中静态跳跃连接的局限性，选择性注意力融合模块自适应地使用空间变化权重整合跨尺度特征。它选择性地增强显著区域，并改善真实目标与误报之间的区分。对三个公共基准数据集NUAA-SIRST、IRSTD-1K和NUDT-SIRST的实验表明，SANet在交并比（IoU）、归一化IoU、检测概率和误报率方面表现出竞争力。其IoU分别超过第二最佳方法1.93、4.32和2.21个百分点。这些结果支持了SANet在微弱目标感知、区分特征表示和背景抑制方面的有效性。
cs.CV / 106 / 2610.09907
FedSSMCoOp: SSM Encoders for light-weight Federated Prompt Learning for Few-shot Classification
FedSSMCoOp：轻量级联邦提示学习的SSM编码器用于少样本分类
Abstract
Vision-Language Models (VLMs) have shown strong performance across a wide range of downstream vision tasks, thanks to the complementary information contained in the respective domains. Despite the performance gains, most of these approaches rely on aligning these domains using the cosine similarity metric, which fails to capture token-level structure and cross-modal interactions prior to the classification stage. This is especially critical in biomedical applications under federated constraints, where data sharing is restricted, labeled data is scarce at each site, and it differs widely across institutions, leading to substantial statistical heterogeneity. To overcome this issue, we propose FedSSMCoOp, a federated few-shot image classification framework that enables multimodal learning while preserving data privacy. With the help of the SSM-based Vision Mamba and Cross Mamba blocks, and by optimizing only the soft-prompt and communication-prompt updates in the federated setting, the framework prioritizes both computation and performance. Importantly, this eliminates the need to use an external Large Language Model (LLM) for feature alignment. The framework is further trained and evaluated on various biomedical image datasets, and its performance is assessed. The proposed framework delivers stable performance relative to the baselines and is, on average, 1.96 times lighter. The corresponding script will be made available soon.
Chinese Translation
视觉-语言模型（VLMs）在广泛的下游视觉任务中表现出色，这得益于各自领域中包含的互补信息。尽管性能有所提升，但大多数方法依赖于使用余弦相似度度量来对齐这些领域，这未能在分类阶段之前捕捉到标记级别的结构和跨模态交互。这在联邦约束下的生物医学应用中尤为重要，因为数据共享受到限制，各个站点的标记数据稀缺，并且在不同机构之间差异很大，导致显著的统计异质性。为了解决这一问题，我们提出了FedSSMCoOp，一个联邦少样本图像分类框架，能够在保护数据隐私的同时实现多模态学习。借助基于SSM的Vision Mamba和Cross Mamba模块，并仅在联邦环境中优化软提示和通信提示的更新，该框架优先考虑计算和性能。重要的是，这消除了使用外部大型语言模型（LLM）进行特征对齐的必要性。该框架进一步在各种生物医学图像数据集上进行训练和评估，并对其性能进行评估。所提出的框架相对于基线提供了稳定的性能，且平均轻量级为1.96倍。相关脚本将很快提供。
cs.CV / 107 / 2610.09928
Do Generative Priors Align with Human Naturalness Perception?
生成先验与人类自然感知的对齐程度如何？
Abstract
Visual generative models are trained to capture the probability distributions of natural images, yet whether their native priors reflect the regularities governing human perception of image naturalness remains an open question. Here, we probe these priors through native prediction errors across 25 open image and video generators. Because raw single-image losses are dominated by scene content and visual complexity, we evaluate directional loss differences using content-preserving, paired relational interventions that selectively disrupt facial configurations or physical illumination consistency while limiting changes in low-level image statistics. Across both domains, these loss differences reproduce human-like selective sensitivities and tolerances, capturing the classic Thatcher effect on faces and shape-dependent responses to illumination inconsistencies. Notably, these loss differences reliably track continuous gradations of human naturalness judgments across individual stimulus pairs (peaking at $r = .84$ on faces and $.64$ on physical scenes) and retain unique human-aligned signals even after controlling for feature distances from frozen vision encoders and standard image quality metrics. We also find that while overall sensitivity to these violations broadly covaries with human alignment across models, the two systematically decouple along denoising schedules, with alignment peaking earlier than sensitivity, revealing that human-like naturalness judgments dissociate from generic violation detection. Together, these findings demonstrate that learning visual distributions yields generative loss landscapes that capture distinct aspects of human naturalness perception.
Chinese Translation
视觉生成模型被训练以捕捉自然图像的概率分布，但它们的固有先验是否反映了支配人类图像自然感知的规律仍然是一个未解之谜。在此，我们通过对25个开放图像和视频生成器的固有预测误差进行探讨。由于原始单图像损失主要受场景内容和视觉复杂度的支配，我们使用保持内容的配对关系干预来评估方向性损失差异，这些干预选择性地扰乱面部配置或物理照明一致性，同时限制低级图像统计特征的变化。在这两个领域中，这些损失差异再现了类似人类的选择性敏感性和容忍度，捕捉到经典的撒切尔效应以及对照明不一致性的形状依赖反应。值得注意的是，这些损失差异可靠地追踪了个体刺激对之间人类自然感知判断的连续梯度（在面部图像上达到$r = .84$，在物理场景上达到$.64$），并在控制了与冻结视觉编码器的特征距离和标准图像质量指标后，仍然保留了独特的人类对齐信号。我们还发现，尽管对这些违规行为的整体敏感性在模型之间广泛变化与人类对齐相关，但两者在去噪调度中系统性地解耦，且对齐在敏感性之前达到峰值，揭示了类似人类的自然感知判断与通用违规检测之间的分离。综上所述，这些发现表明，学习视觉分布产生的生成损失景观捕捉了人类自然感知的不同方面。
cs.CV / 108 / 2610.09941
Purifying Backdoored Large Vision-Language Models by Removing Hijacked Directions
通过去除劫持方向净化后门大型视觉-语言模型
Abstract
Large vision-language models (LVLMs) are increasingly deployed in safety-critical applications, yet they remain vulnerable to backdoor attacks. Defending against such attacks remains costly, as existing methods require either extensive retraining on clean data or per-query intervention at inference time. To address this limitation, we propose OrthoPurify, a more efficient method to purify backdoored model weights via one-step orthogonal projection. Specifically, through structural analysis of backdoor weight updates, we find that the backdoor is encoded by diverting a small number of weight update directions from task adaptation to backdoor shortcut encoding, a phenomenon we term direction hijacking. However, identifying these hijacked directions requires a benign reference model, which is typically inaccessible to the defender. We show that a pseudo-benign model, obtained by fine-tuning the pretrained weights on only a small set of clean samples, provides a sufficient approximation, as the dominant update directions stabilize within the first few gradient steps. OrthoPurify uses this pseudo-benign reference to isolate the hijacked directions and removes them through a single projection on the weight update. Extensive experiments show that OrthoPurify reduces the attack success rate to near zero while preserving the original performance across diverse benchmarks, without retraining the backdoored model or introducing inference-time overhead. Our code is publicly available at https://github.com/womeimingzi/OrthoPurify.
Chinese Translation
大型视觉-语言模型（LVLMs）越来越多地应用于安全关键的场景，但它们仍然容易受到后门攻击。防御此类攻击的成本依然高昂，因为现有方法要么需要在干净数据上进行广泛的重新训练，要么在推理时进行逐查询干预。为了解决这一限制，我们提出了 OrthoPurify，一种通过一步正交投影来净化后门模型权重的更高效的方法。具体而言，通过对后门权重更新的结构分析，我们发现后门是通过将少量权重更新方向从任务适应转移到后门捷径编码来编码的，这一现象我们称之为方向劫持。然而，识别这些劫持方向需要一个良性参考模型，而该模型通常对防御者不可用。我们表明，通过仅在一小部分干净样本上微调预训练权重获得的伪良性模型提供了足够的近似，因为主导更新方向在前几步梯度更新中会稳定。OrthoPurify 使用这个伪良性参考来隔离劫持方向，并通过对权重更新进行单次投影将其去除。大量实验表明，OrthoPurify 将攻击成功率降低到接近零，同时在各种基准测试中保持原始性能，无需重新训练后门模型或引入推理时的开销。我们的代码公开可用，地址为 https://github.com/womeimingzi/OrthoPurify。
cs.CV / 109 / 2610.09952
MSU Team at the Explainable Deepfake Detection Challenge 2026: Grounded Artifact Evidence for Deepfake Detection
MSU团队在2026年可解释深度伪造检测挑战中的表现：基于可见伪证的深度伪造检测
Abstract
Recent advances in generative image models have made many manipulated images highly realistic, raising the need for detectors that are not only accurate but also able to provide visual evidence for their decisions. In this paper, we present our solution to the Explainable Deepfake Detection Challenge [2] on the XPlainVerse dataset [1], where systems are required to predict whether an image is real or fake and generate both complex and simple explanations grounded in visible forensic cues. Our method follows a modular detection-and-explanation design. For the real/fake decision, we build a multi-backbone detector that combines several DINOv3 models with Mesorch manipulation-localization features, bringing together pretrained visual representations, DCT-aware cues, and multi-scale forensic information. To inject explanation evidence into the detector, we use a Grounding-DINO-based pseudo-mask generation pipeline that converts local artifact descriptions from training explanations into weak patch- level supervision for an Artifact Evidence Map. We further introduce a local patch-level contrastive objective that separates artifact and authenticity evidence in the detector feature space without requiring paired images or pixel-level manipulation masks. For language output, we use class-conditional Qwen3-VL models to generate complex explanations for fake and real predictions, followed by a GRPO-optimized text simplification model. The proposed methods were trained and evaluated on the challenge subset of XPlainVerse. On the full test split, our submission achieves 0.9349 detection accuracy, 0.5571 explanation score, and a 0.7456 final challenge score.
Chinese Translation
近年来生成图像模型的进展使得许多被操纵的图像高度逼真，这提高了对不仅准确而且能够提供视觉证据的检测器的需求。在本文中，我们展示了我们在可解释深度伪造检测挑战[2]中对XPlainVerse数据集[1]的解决方案，该系统需要预测图像是真实的还是伪造的，并生成基于可见法医线索的复杂和简单解释。我们的方法遵循模块化的检测与解释设计。在真实/伪造决策方面，我们构建了一个多骨干检测器，该检测器结合了多个DINOv3模型与Mesorch操控定位特征，整合了预训练的视觉表征、DCT感知线索和多尺度法医信息。为了将解释证据注入检测器，我们使用基于Grounding-DINO的伪掩码生成管道，将训练解释中的局部伪证描述转换为伪弱的补丁级监督，用于伪证证据图。我们进一步引入了一种局部补丁级对比目标，该目标在检测器特征空间中分离伪证和真实性证据，而无需配对图像或像素级操控掩码。对于语言输出，我们使用类条件的Qwen3-VL模型为伪造和真实预测生成复杂解释，随后使用GRPO优化的文本简化模型。所提方法在XPlainVerse挑战子集上进行了训练和评估。在完整测试集上，我们的提交实现了0.9349的检测准确率、0.5571的解释分数和0.7456的最终挑战分数。
cs.CV / 110 / 2610.10003
Perceptually Aligned Evaluation of Style Transfer
感知对齐的风格迁移评估
Abstract
Style transfer lacks a reliable evaluation standard: ground truth is inherently ill-defined, and existing automatic metrics often fail to reflect human preference. This paper introduces ASTRA (Assessment of Style TRansfer Algorithms), an approach for automatic evaluation of style transfer algorithms; it contains two components, ASTRA-Data and ASTRA-Score. ASTRA-Data consists of a benchmark image set of content and style references, a collection of style transfer results generated on the benchmark set, and user study data capturing human judgements through a two-stage pairwise comparison protocol. From these annotations, we derive ranking-based ground truth for content preservation, style fidelity, and overall preference. Based on ASTRA-Data, we construct ASTRA-Score, a learnt evaluator that predicts preference-aligned scores from content-style-stylization image triplets, enabling automatic and scalable evaluation of new models applied to the benchmark set. Experimental results demonstrate that ASTRA-Score achieves substantially higher correlation with human rankings compared to prior metrics. Overall, ASTRA establishes a robust mechanism for standardised evaluation of style transfer methods.
Chinese Translation
风格迁移缺乏可靠的评估标准：真实标准本质上定义不明确，现有的自动化指标往往无法反映人类偏好。本文提出了ASTRA（风格迁移算法评估），一种用于风格迁移算法自动评估的方法；它包含两个组成部分，ASTRA-Data和ASTRA-Score。ASTRA-Data由内容和风格参考的基准图像集、在基准集上生成的风格迁移结果集合，以及通过两阶段成对比较协议捕捉人类判断的用户研究数据组成。基于这些注释，我们推导出内容保留、风格保真度和整体偏好的基于排名的真实标准。基于ASTRA-Data，我们构建了ASTRA-Score，这是一个学习评估器，能够从内容-风格-风格化图像三元组中预测偏好对齐的分数，从而实现对应用于基准集的新模型的自动化和可扩展评估。实验结果表明，ASTRA-Score与人类排名的相关性显著高于先前的指标。总体而言，ASTRA建立了一种标准化评估风格迁移方法的强大机制。
cs.CV / 111 / 2610.10012
Playing with Kruskal: algorithms for flat and hierarchical watershed cuts
与克鲁斯卡尔算法的互动：平面和层次性分水岭切割的算法
Abstract
In the framework of edge-weighted graphs, watersheds have proven to be linked to well-known optimization problems, as Minimum Spanning Tree, which allowed the design of efficient algorithms for computing (hierarchical) watershed segmentations. In the present article, after reviewing the literature related to watershed segmentation, we present a detailed end-to-end pipeline of algorithms to compute (hierarchical) watershed segmentations, starting from the computation of graph-based image representations, up to the computation of connected components of the final (hierarchical) segmentation. We consider the several variations of watersheds, including their supervised and unsupervised versions, and the various ways of computing seeds, to name a few. For the first time, we bring together all these watershed notions and algorithms in a compact and understandable way. We aim at providing a reference for those interested in employing and reimplementing the watershed segmentation framework for their task at hand.
Chinese Translation
在边权重图的框架下，分水岭与著名的优化问题（如最小生成树）密切相关，这使得设计高效算法以计算（层次性）分水岭分割成为可能。在本文中，我们在回顾与分水岭分割相关的文献后，详细介绍了一套完整的算法流程，以计算（层次性）分水岭分割，从图像的图基表示计算开始，到最终（层次性）分割的连通组件计算为止。我们考虑了分水岭的多种变体，包括其监督和无监督版本，以及计算种子的各种方法等。我们首次将所有这些分水岭概念和算法以紧凑且易于理解的方式汇集在一起。我们的目标是为那些希望在其任务中采用和重新实现分水岭分割框架的研究者提供参考。
cs.CV / 112 / 2610.10013
Scalable Patch-Level Self-Supervised Learning
可扩展的补丁级自监督学习
Abstract
Self-supervised learning (SSL) at scale produces powerful visual representations. However, most scalable SSL methods rely on ad hoc combinations of multiple objectives and stabilization mechanisms. Taking a step back, we ask if we can design a high-performing, yet principled SSL algorithm. Starting from the multi-view assumption, stipulating that task-relevant content is captured by the information common to different views, we construct an information-theoretic objective decomposing into interpretable terms. This derivation yields JEM, a student-teacher method that learns by aligning corresponding patch representations across views, explicitly regularized by information and structure preservation losses. JEM trains stably from 300M to 7B parameters, and, to our knowledge, is the first latent-space patch-level method demonstrated at 7B scale. Across all scales, JEM reaches strong performance on both global and dense probing tasks, on segmentation benchmarks consistently surpassing the DINOv2 algorithm, an influential foundation for today's strongest visual SSL methods. Notably, at 7B parameters, it exceeds the performance of DINOv3 on panoptic segmentation, despite being trained on $12\times$ less data without refinement stages. These results demonstrate that we can indeed design an SSL algorithm that learns strong representations, is principled and stable.
Chinese Translation
大规模自监督学习（SSL）能够生成强大的视觉表征。然而，大多数可扩展的SSL方法依赖于多种目标和稳定机制的临时组合。我们退一步思考，是否可以设计出一种高性能且有原则的SSL算法。基于多视角假设，规定任务相关内容由不同视角间的共同信息捕获，我们构建了一个信息论目标，分解为可解释的项。该推导产生了JEM，一种通过对齐不同视角中的对应补丁表征进行学习的师生方法，明确地通过信息和结构保留损失进行正则化。JEM在从3亿到70亿参数的范围内稳定训练，据我们所知，这是第一个在70亿规模下展示的潜在空间补丁级方法。在所有规模下，JEM在全局和密集探测任务上均表现出色，在分割基准测试中持续超越DINOv2算法，后者是当今最强视觉SSL方法的重要基础。值得注意的是，在70亿参数下，尽管在没有细化阶段的情况下训练的数据量是DINOv3的$12 imes$，但其在全景分割任务上的表现仍超过了DINOv3。这些结果表明，我们确实可以设计出一种学习强大表征的SSL算法，该算法是有原则且稳定的。
cs.CV / 113 / 2610.10047
AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation
AdSpark：面向产品的广告视频生成的大规模数据集与基准
Abstract
Product-centric advertisement video generation aims to create promotional videos that preserve fine-grained product identity while presenting selling points through coherent multi-shot narratives. However, this emerging task remains underexplored due to the lack of large-scale advertisement-specific datasets and comprehensive evaluation frameworks. To address this gap, we introduce \textbf{AdSpark}, a large-scale dataset and benchmark for product-centric advertisement video generation, based on data from a major e-commerce platform. \textit{AdSpark-300K} contains approximately 300K reference image--prompt--video triplets, comprising a real-world subset and a synthetic subset. Each sample provides structured advertisement annotations, including product identity annotations, selling-point descriptions, creative plans, and aligned audio scripts, enabling models to learn product preservation and advertisement-oriented visual storytelling. We further propose \textit{AdSpark-Bench}, a diagnostic benchmark that evaluates generated advertisements across six dimensions, including visual quality, product fidelity, instruction adherence, temporal coherence, audio alignment, and advertisement effectiveness. Based on AdSpark-Bench, we evaluate representative models, revealing key challenges in product preservation, multi-shot storytelling, and selling-point visualization. Experiments with AdSpark-300K-finetuned models further validate the effectiveness of our dataset. AdSpark provides a unified dataset and benchmark for future research, and we will release the dataset upon acceptance.
Chinese Translation
面向产品的广告视频生成旨在创建能够保留细致产品特征的宣传视频，同时通过连贯的多镜头叙事展示卖点。然而，由于缺乏大规模的广告特定数据集和全面的评估框架，这一新兴任务仍然未得到充分探索。为了解决这一问题，我们引入了 extbf{AdSpark}，一个基于主要电子商务平台数据的大规模面向产品的广告视频生成数据集与基准。 extit{AdSpark-300K}包含约30万个参考图像-提示-视频三元组，包括真实世界子集和合成子集。每个样本提供结构化的广告注释，包括产品身份注释、卖点描述、创意计划和对齐的音频脚本，使模型能够学习产品保留和面向广告的视觉叙事。我们进一步提出了 extit{AdSpark-Bench}，一个诊断基准，用于评估生成的广告在视觉质量、产品忠实度、指令遵循、时间连贯性、音频对齐和广告有效性等六个维度的表现。基于AdSpark-Bench，我们评估了代表性模型，揭示了产品保留、多镜头叙事和卖点可视化等方面的关键挑战。对AdSpark-300K微调模型的实验进一步验证了我们数据集的有效性。AdSpark为未来研究提供了统一的数据集和基准，我们将在接受后发布该数据集。
cs.CV / 114 / 2610.10066
From Pixel to Coding: Evaluating the Figure Reproduction Capabilities of MLLMs
从像素到编码：评估多模态大型语言模型的图形再现能力
Abstract
Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities in both visual understanding and code generation. However, existing benchmarks typically evaluate these two modalities in isolation, lacking a dedicated assessment of their unification, i.e., how a model can perceive complex visual structures and synthesize them into precise, executable code. Moreover, current visual code generation benchmarks often rely on simplified layouts within single programming environments, falling short of evaluating true unified multimodal reasoning. To bridge this gap, we propose FigCodeBench, a comprehensive framework for rigorously evaluating MLLMs on figure reproduction, integrating multimodal comprehension and generation. We first design a systematic dataset construction pipeline, resulting in a total of 6,194 instances that cover 7 functional categories and 4 types of programming languages. We further categorize figure reproduction into three tiers with visual and code complexity modeling, specifically targeting complex structural reasoning, varying aspect ratios, and dense geometric constraints. We introduce a multi-dimensional evaluation protocol, encompassing visual fidelity and syntactic isomorphism, that aligns highly with the Mean Machine Opinion Score (MMOS) and human preferences. Based on our framework, we conducted extensive experiments on 24 widely used proprietary and open-source MLLMs (e.g., Gemini 3.1 Pro, GPT-5.4, and Kimi-K2.5), where we observed a universal, non-linear performance cliff across different programming languages and difficulty scenarios for all models, and gained several insights, such as the significant metric decline in rigid declarative languages.
Chinese Translation
多模态大型语言模型（MLLMs）在视觉理解和代码生成方面展现了令人印象深刻的能力。然而，现有的基准测试通常孤立地评估这两种模态，缺乏对其统一性的专门评估，即模型如何感知复杂的视觉结构并将其合成成精确的可执行代码。此外，目前的视觉代码生成基准往往依赖于单一编程环境中的简化布局，未能真正评估统一的多模态推理。为了解决这一问题，我们提出了FigCodeBench，这是一个全面的框架，用于严格评估MLLMs在图形再现方面的能力，整合了多模态理解和生成。我们首先设计了一个系统的数据集构建流程，最终生成了6194个实例，涵盖7个功能类别和4种编程语言类型。我们进一步将图形再现分为三个层次，并对视觉和代码复杂性进行建模，特别针对复杂的结构推理、不同的纵横比以及密集的几何约束。我们引入了一个多维评估协议，涵盖视觉保真度和语法同构性，与平均机器意见分数（MMOS）和人类偏好高度一致。基于我们的框架，我们对24个广泛使用的专有和开源MLLMs（如Gemini 3.1 Pro、GPT-5.4和Kimi-K2.5）进行了广泛实验，观察到所有模型在不同编程语言和难度场景下普遍存在非线性的性能下降，并获得了一些见解，例如在严格的声明性语言中的显著指标下降。
cs.CV / 115 / 2610.10090
Temporal Residual Bottleneck for Robust Asynchronous Collaborative Perception
用于鲁棒异步协作感知的时间残差瓶颈
Abstract
Collaborative perception extends the sensing range of autonomous vehicles, but its performance degrades when shared features arrive stale or incomplete. Most latency-robust methods compensate delayed collaborator features through flow-guided alignment or direct feature transport. In this work, we formulate asynchronous collaborative perception as temporal residual prediction. Our Temporal Residual Bottleneck keeps a deterministic pose-warped collaborator feature as a conservative anchor and uses a $\Delta t$-conditioned xLSTM to extract residual temporal evidence from the available history. A detector-facing residual bottleneck then applies only gated, regularized corrections before ego-side fusion, reducing the risk of overwriting reliable static structure when temporal correspondence is uncertain. Experiments on DAIR-V2X and OPV2V show that our method is especially effective under severe fixed/irregular delays and packet drops. On DAIR-V2X, the reported checkpoint trades a small amount of synchronized peak accuracy for better robustness under stronger communication degradation. Controlled diagnostics further indicate that direct feature transport has oracle headroom but can become unreliable when deployed without accurate correspondence. These results support temporal residual fusion as a practical alternative for asynchronous and incomplete collaborative perception. Code will be publicly released at https://url.fzi.de/8dk38.
Chinese Translation
协作感知扩展了自动驾驶车辆的感知范围，但当共享特征变得过时或不完整时，其性能会下降。大多数抗延迟方法通过流引导对齐或直接特征传输来补偿延迟的协作特征。在本研究中，我们将异步协作感知形式化为时间残差预测。我们的时间残差瓶颈保持一个确定性的姿态扭曲协作特征作为保守锚点，并使用 $ ext{Δ}t$ 条件的 xLSTM 从可用历史中提取残差时间证据。面向检测器的残差瓶颈在自我侧融合之前仅应用门控的正则化修正，从而降低在时间对应不确定时覆盖可靠静态结构的风险。在 DAIR-V2X 和 OPV2V 上的实验表明，我们的方法在严重的固定/不规则延迟和数据包丢失情况下尤其有效。在 DAIR-V2X 上，报告的检查点在更强的通信退化下以较小的同步峰值精度换取了更好的鲁棒性。控制诊断进一步表明，直接特征传输具有理想的潜力，但在没有准确对应的情况下部署时可能变得不可靠。这些结果支持时间残差融合作为异步和不完整协作感知的实用替代方案。代码将公开发布在 https://url.fzi.de/8dk38。
cs.CV / 116 / 2610.10116
A Probabilistic Perspective on Wasserstein-Based Evidential Uncertainty for Out-of-Distribution Segmentation
基于Wasserstein的证据不确定性的概率视角：针对分布外分割
Abstract
Semantic segmentation networks operate on a fixed set of classes and therefore fail when out-of-distribution (OOD) objects appear during deployment, a critical limitation for safety-critical applications such as autonomous driving. Reliably identifying OOD objects requires well-calibrated epistemic uncertainty, yet common softmax-based confidence scores remain overconfident, while Bayesian alternatives such as Monte Carlo dropout or deep ensembles require costly repeated forward passes. Evidential Deep Learning (EDL) offers an efficient alternative by modeling class probabilities as a Dirichlet distribution learned from a single deterministic forward pass. Existing EDL formulations rely on Euclidean objectives that push predictions towards the simplex vertices, encouraging overconfidence rather than preserving uncertainty for unfamiliar inputs. We instead employ Wasserstein-based objectives, which respect the geometry of the probability simplex, and study the influence of the Wasserstein order on segmentation accuracy and OOD detection within a unified evidential framework. We evaluate this framework on a convolutional (DeepLabV3+) and a transformer-based (SegFormer) architecture on the SegmentMeIfYouCan benchmark, including LostAndFound, RoadObstacle21, RoadAnomaly21, and Fishyscapes. Our results show the optimal Wasserstein order is architecture-dependent: second-order objectives dominate on the convolutional backbone, third-order objectives on the transformer backbone, and our framework surpasses comparable baselines on most metrics, with a single deterministic forward pass.
Chinese Translation
语义分割网络在固定类别集上运行，因此在部署过程中出现分布外（OOD）对象时会失败，这对于诸如自动驾驶等安全关键应用来说是一个重要的限制。可靠地识别OOD对象需要良好校准的认知不确定性，然而常见的基于softmax的置信度评分往往过于自信，而贝叶斯替代方法如蒙特卡洛丢弃（Monte Carlo dropout）或深度集成（deep ensembles）则需要昂贵的重复前向传播。证据深度学习（Evidential Deep Learning, EDL）通过将类别概率建模为从单次确定性前向传播中学习的Dirichlet分布，提供了一种高效的替代方案。现有的EDL公式依赖于欧几里得目标，这会将预测推向简单形状的顶点，鼓励过度自信，而不是为不熟悉的输入保留不确定性。相反，我们采用基于Wasserstein的目标，这尊重概率简单形状的几何特性，并在统一的证据框架内研究Wasserstein阶数对分割精度和OOD检测的影响。我们在SegmentMeIfYouCan基准上评估该框架，包括LostAndFound、RoadObstacle21、RoadAnomaly21和Fishyscapes，使用了卷积（DeepLabV3+）和基于变换器（SegFormer）的架构。我们的结果表明，最佳的Wasserstein阶数依赖于架构：在卷积骨干网络上，二阶目标占主导地位，而在变换器骨干网络上，三阶目标更为有效，并且我们的框架在大多数指标上超越了可比基线，且仅需一次确定性前向传播。
cs.CV / 117 / 2610.10133
HarnessIR: Harnessing Multimodal Foundation Models for Universal Real-World Image Restoration
HarnessIR：利用多模态基础模型进行通用真实世界图像修复
Abstract
Real-world low-quality images suffer from complex mixed degradations, including but not limited to noise, blur, atmospheric effects, etc. Recent agentic methods usually model real-world image restoration (Real-IR) as a sequential tool calling problem over task-specific single-degradation restoration models. This paradigm, however, is fundamentally limited because complex real-world degradations cannot be cleanly undone degradation by degradation, and the tool used for task-specific models caps the capability of the agent system. In this work, we present HarnessIR, an agentic framework for Real-IR by harnessing a multimodal foundation model (MFM) as the executor. HarnessIR consists of five stages: perception and diagnosis, on-demand tool invocation, prompt composition, execution, and verification-driven refinement. Unlike prior agentic Real-IR methods that rely on tool chains assembled from task-specific models, HarnessIR feeds the restoration requirements, the perceptual diagnosis, and the evidence into an MFM that performs restoration in a single pass, followed by verification stages to determine whether the result warrants further processing. Under our harness, off-the-shelf MFMs handle restoration tasks remarkably well, achieving state-of-the-art results on the widely used MiO100 synthetic benchmark. More importantly, by exploiting the strong generalization ability of MFMs, HarnessIR delivers compelling restoration quality on challenging real-world scenes where previous agentic IR systems often struggle. Codes is available at https://github.com/PolyU-VCLab/HarnessIR.
Chinese Translation
真实世界中的低质量图像受到复杂混合退化的影响，包括但不限于噪声、模糊、大气效应等。近期的代理方法通常将真实世界图像修复（Real-IR）建模为一个针对特定任务的单一退化修复模型的顺序工具调用问题。然而，这种范式在根本上是有限的，因为复杂的真实世界退化无法逐一干净地逆转，且用于特定任务模型的工具限制了代理系统的能力。在本研究中，我们提出了HarnessIR，这是一个通过利用多模态基础模型（MFM）作为执行者的Real-IR代理框架。HarnessIR由五个阶段组成：感知与诊断、按需工具调用、提示组合、执行和基于验证的细化。与之前依赖于从特定任务模型组装的工具链的代理Real-IR方法不同，HarnessIR将修复需求、感知诊断和证据输入到一个MFM中，该模型在单次传递中执行修复，随后进行验证阶段以确定结果是否需要进一步处理。在我们的框架下，现成的MFM在处理修复任务方面表现出色，在广泛使用的MiO100合成基准上取得了最先进的结果。更重要的是，通过利用MFM的强泛化能力，HarnessIR在以往代理IR系统常常面临挑战的复杂真实场景中提供了令人信服的修复质量。代码可在 https://github.com/PolyU-VCLab/HarnessIR 获取。
cs.CV / 118 / 2610.10156
HeiCo-FOCUS: A Clinically Grounded Dataset for Long-Context Video Understanding
HeiCo-FOCUS：用于长时视频理解的临床基础数据集
Mayer, Leon, Luttner, Lucas, Godau, Patrick, Fritzsche, Kai, Reinke, Annika, Boland, Leonie, Brandt, Jule, Heinecke, Janne, Nobuhara, Chloe K., Holzwarth, Niklas, Christodoulou, Evangelia, Knopp, Marcel, Michael, Dominik, Piermarco, Pascale, Neyaz, Saliq, Özarslan, Korhan Derin, Hennighausen, Jakob, Aumente-Maestro, Carlos, Rädsch, Tim, Baji, Dheeraj, Full, Peter Maximilian, Aichholz, Finn, Erpenbeck, Justus Veit, Schott, Linus Finn, Winkelhausen, Bastian, de Boer, Claas, Güttner, Bianca, Hummel, Anneli, Just, Gregor, Kirchner, Max, Li, Chenyang, Raffaut, Rozenn, Rodriguez, Ariel, Venkatesh, Danush Kumar, Wang, Kevin, Xu, Jinjing, Zeinoddin, Mona Sheikh, Khan, Salman, Pausch, Thomas M., Speidel, Stefanie, Stoyanov, Danail, Hashimoto, Daniel A., Kolbinger, Fiona R., Weiser, Thomas G., Maier-Hein, Lena
Abstract
Recent advances in Vision-Language Models (VLMs) have led to rapid progress in video understanding across a wide range of benchmark tasks. However, existing evaluations largely focus on short-term reasoning, failing to assess a critical capability: maintaining cumulative temporal consistency over extended time horizons. To close this evaluation gap, we introduce HeiCo-FOCUS, a clinically grounded dataset for evaluating long-context video understanding through the task of Foreign Object Contextual Understanding in Surgery. Built on a dataset of Heidelberg Colorectal surgeries, this task requires models to continuously track multiple objects as they are inserted, manipulated, occluded, and removed over procedures lasting up to hours. HeiCo-FOCUS comprises 30,000 visual question answering (VQA) pairs covering five core capabilities: object recognition, temporal grounding, aggregation, event and procedural understanding, and complex reasoning. The dataset was constructed through a rigorous multi-stage annotation pipeline involving large-scale crowd annotation and 39 surgical domain experts to ensure high quality and clinical relevance. To systematically probe model behavior, we introduce a multi-track evaluation framework that progressively increases temporal and contextual demands from single frames to full procedures. Experiments with ten frontier VLMs show that HeiCo-FOCUS tasks are far from solved: only around half of the models clearly outperform a text-only baseline. Across the video tracks, models perform best on event and procedural understanding (mean Accuracy: 56.5% across all models), while temporal grounding remains particularly challenging for all evaluated models (mean Accuracy: 19.7%). We therefore expect HeiCo-FOCUS to serve as a catalyst for the development of models capable of reliable, temporally consistent reasoning over hours-long videos.
Chinese Translation
近期视觉-语言模型（VLMs）的进展推动了视频理解在多项基准任务中的快速发展。然而，现有评估主要集中在短期推理上，未能评估一个关键能力：在较长时间范围内保持累积的时间一致性。为填补这一评估空白，我们引入了HeiCo-FOCUS，这是一个用于通过外科手术中的异物上下文理解任务评估长时视频理解的临床基础数据集。该任务基于海德堡结直肠手术的数据集，要求模型在长达数小时的手术过程中持续跟踪多个物体的插入、操作、遮挡和移除。HeiCo-FOCUS包含30,000对视觉问答（VQA）样本，涵盖五个核心能力：物体识别、时间定位、聚合、事件和程序理解，以及复杂推理。该数据集通过严格的多阶段注释流程构建，涉及大规模众包注释和39位外科领域专家，以确保高质量和临床相关性。为了系统性地探测模型行为，我们引入了一个多轨评估框架，该框架逐步增加从单帧到完整手术的时间和上下文需求。对十个前沿VLM的实验表明，HeiCo-FOCUS任务远未解决：只有约一半的模型明显优于仅文本基线。在视频轨道中，模型在事件和程序理解方面表现最佳（所有模型的平均准确率：56.5%），而时间定位对所有评估模型仍然特别具有挑战性（平均准确率：19.7%）。因此，我们期望HeiCo-FOCUS能够成为开发能够在数小时视频中进行可靠、时间一致推理的模型的催化剂。
cs.CV / 119 / 2610.10160
BagDINO: Multi-View Baggage Re-Identification with DINOv3
BagDINO：基于DINOv3的多视角行李重识别
Abstract
Mishandled checked baggage remains a recurrent issue in airport operations, and current recovery workflows still largely rely on tag-based tracking, which does not directly support visual identification when tag evidence is missing or unavailable. This paper investigates baggage re-identification as an instance-level retrieval problem in a multi-camera setting, leveraging DINOv3 foundation-model representations to match a query image against a gallery of registered baggage images. A Torchreid-style BNNeck re-identification head is placed on top of a DINOv3 backbone, and parameter-efficient adaptation is performed via LoRA. Experiments are conducted on the MVB benchmark using a progressive study that compares a fully frozen backbone against LoRA and fine-tuning strategies. Results indicate that parameter-efficient adaptation of foundation-model features provides an effective and stable approach for multi-view baggage re-identification under limited training data.
Chinese Translation
在机场运营中，处理不当的托运行李仍然是一个反复出现的问题，目前的恢复工作流程仍然在很大程度上依赖于基于标签的追踪，这在缺少或无法获得标签证据时并不直接支持视觉识别。本文将行李重识别视为多摄像头环境中的实例级检索问题，利用DINOv3基础模型的表示，将查询图像与注册行李图像库进行匹配。在DINOv3主干网络上添加了Torchreid风格的BNNeck重识别头，并通过LoRA进行参数高效的适配。实验在MVB基准上进行，采用渐进式研究，比较完全冻结的主干网络与LoRA和微调策略。结果表明，基础模型特征的参数高效适配为在有限训练数据下的多视角行李重识别提供了一种有效且稳定的方法。
cs.CV / 120 / 2610.10163
Beyond Anonymous Captions: Grounding Character Identity in Video Captioning and Question Answering
超越匿名字幕：在视频字幕生成和问答中扎根角色身份
Abstract
Linking people's appearance and actions to character identities is essential for understanding video narratives. We present a framework for identity-aware video captioning and person-centric question answering that combines automatic character identification, explicit spatial grounding, and task-specific adaptation. Starting from LSMDC v2 movie clips, our pipeline matches detected faces to actor reference images, tracks characters across frames, and builds inputs with identity-linked bounding boxes. A strong vision-language model generates identity-aware captions and questions, which are manually verified and filtered to create a benchmark of 750 captioned clips and 3,000 person-centric questions. We study five grounding strategies combining textual coordinates with visual face or estimated person boxes across Video-MLLM families at roughly 2B, 4B, and 8B parameters and larger frontier models. Combining visual face boxes with textual coordinates yields the most consistent performance across scales and significantly improves overall performance over coordinates alone. Smaller models tend to over-assign known identities when the queried person is not grounded, while larger models better recognize such UNIDENTIFIED cases. We introduce BAC by LoRA fine-tuning Qwen models at 2B, 4B, and 8B scales on about 32K identity-aware captioned clips. Across all scales, BAC outperforms every other evaluated model family of comparable size. BAC-8B reaches 93.20\% overall QA accuracy, ranking behind only GPT-5.6 Sol among the frontier models evaluated in our study. Overall, explicitly communicating who is where, together with lightweight task-specific adaptation, substantially improves identity-aware video understanding without changing the underlying architecture. We release the benchmark, training data, code, and BAC checkpoints at https://github.com/momentslab/beyond-anonymous-captions.
Chinese Translation
将人们的外貌和行为与角色身份联系起来对于理解视频叙事至关重要。我们提出了一个身份感知的视频字幕生成和以人物为中心的问答框架，该框架结合了自动角色识别、显式空间定位和任务特定的适应性。从LSMDC v2电影片段开始，我们的流程将检测到的面孔与演员参考图像匹配，跨帧跟踪角色，并构建带有身份关联边界框的输入。一个强大的视觉-语言模型生成身份感知的字幕和问题，这些内容经过人工验证和筛选，以创建一个包含750个带字幕片段和3000个以人物为中心的问题的基准。我们研究了五种结合文本坐标与视觉面孔或估计人物框的定位策略，涵盖了大约2B、4B和8B参数的Video-MLLM系列及更大的前沿模型。将视觉面孔框与文本坐标结合使用在各个规模上表现出最一致的性能，并显著提高了整体性能，相较于仅使用坐标。较小的模型在查询的人没有被定位时往往会过度分配已知身份，而较大的模型在识别这些未识别的案例时表现更佳。我们通过对Qwen模型在约32K个身份感知字幕片段上进行LoRA微调，引入了BAC。BAC在所有规模上都超越了每个评估的同类模型系列。BAC-8B的整体问答准确率达到93.20\%，在我们研究中评估的前沿模型中仅次于GPT-5.6 Sol。总体而言，明确传达谁在何处，以及轻量级的任务特定适应性，显著改善了身份感知的视频理解，而无需改变底层架构。我们在https://github.com/momentslab/beyond-anonymous-captions发布了基准、训练数据、代码和BAC检查点。
cs.CV / 121 / 2610.10181
Argos: Adapt Rich Geometric Priors for Generalizable Online Scene-Change-Detection
Argos：适应丰富几何先验以实现可泛化的在线场景变化检测
Abstract
Robots operating in dynamic environments require reliable detection of how their surroundings change over time. Existing learning-based methods largely rely on pairwise 2D image features, which struggle under large viewpoint changes and occlusions, are sensitive to noise, and show limited generalization across domains, while explicit 3D approaches typically require costly offline optimization. We show that the implicit 3D knowledge of Geometric Foundation Models (GFMs) provides a strong basis for addressing these limitations. We introduce Argos, which adapts GFM features for joint scene change detection and 3D reconstruction. To address data scarcity and take a step toward a foundation model for scene change detection, we introduce a large-scale benchmark comprising two synthetic datasets and one real-world dataset, and train jointly across diverse datasets to improve cross-domain generalization. We further introduce Argos-SLAM, a real-time system designed for robotics, which performs online change detection and change-aware 4D mapping. Across benchmarks, our framework substantially outperforms existing baselines, with gains of up to 42.01% in change IoU and 27.91% in F1, while supporting scalable deployment in changing real-world environments.
Chinese Translation
在动态环境中操作的机器人需要可靠地检测其周围环境随时间的变化。现有的基于学习的方法主要依赖于成对的二维图像特征，这在大视角变化和遮挡情况下表现不佳，且对噪声敏感，跨领域的泛化能力有限，而显式的三维方法通常需要昂贵的离线优化。我们展示了几何基础模型（Geometric Foundation Models, GFMs）的隐式三维知识为解决这些局限性提供了强有力的基础。我们引入了Argos，它将GFM特征用于联合场景变化检测和三维重建。为了应对数据稀缺并朝着场景变化检测的基础模型迈进，我们引入了一个大规模基准，包括两个合成数据集和一个真实世界数据集，并在多样化的数据集上进行联合训练，以提高跨领域的泛化能力。我们进一步介绍了Argos-SLAM，这是一个为机器人设计的实时系统，能够进行在线变化检测和变化感知的四维映射。在各个基准测试中，我们的框架显著超越了现有基线，变化IoU提高了42.01%，F1分数提高了27.91%，同时支持在变化的真实世界环境中的可扩展部署。
cs.CV / 122 / 2610.10183
VideoEvolve: Co-Evolving Memory and Retrieval for Long Video Understanding
VideoEvolve：共同进化记忆与检索以实现长视频理解
Abstract
Long video understanding increasingly relies on external memory to organize massive visual streams into compact representations. However, most memory-based methods dynamically adapt how information is retrieved for different questions, while largely fixing what is remembered. This mismatch makes missing details costly to recover, whereas stored information is valuable only when it can be reliably retrieved. To address this issue, we propose VideoEvolve, a novel self-evolving framework that jointly evolves memory and retrieval for long video understanding. Specifically, starting from a coarse low-frame-rate overview, VideoEvolve couples a Memory Evolver for selective memory augmentation with a Retrieval Evolver for adaptive retrieval over the evolving memory. We then co-evolve the two Evolvers through alternating agentic reinforcement learning (Agentic RL), updating one while freezing the other. To steer this alternating evolution, Bottleneck-Aware Evolution Feedback (BEF) identifies whether the current bottleneck lies in memory or retrieval and directs optimization toward the more limiting side. Furthermore, VideoEvolve introduces Capability-Aware Evolution Feedback (CEF) to alleviate downstream feedback from over-specializing memory to a fixed set of training questions, shifting training toward underdeveloped yet learnable video capabilities. By integrating Agentic RL with BEF and CEF, VideoEvolve transforms downstream reasoning experience into transferable capability updates, providing a concrete path from static long-video systems toward experience-driven, self-improving multimodal intelligence. Extensive experiments on multiple long video understanding benchmarks demonstrate the effectiveness of VideoEvolve.
Chinese Translation
长视频理解越来越依赖外部记忆将大量视觉流组织成紧凑的表示。然而，大多数基于记忆的方法动态调整信息检索的方式以应对不同的问题，同时在很大程度上固定了记忆内容。这种不匹配使得缺失的细节恢复成本高昂，而存储的信息只有在能够可靠检索时才有价值。为了解决这一问题，我们提出了VideoEvolve，一个新颖的自我进化框架，联合进化记忆和检索以实现长视频理解。具体而言，VideoEvolve从粗略的低帧率概览开始，结合了用于选择性记忆增强的Memory Evolver与用于在不断进化的记忆上进行自适应检索的Retrieval Evolver。然后，我们通过交替的自主强化学习（Agentic RL）共同进化这两个Evolver，在更新一个的同时冻结另一个。为了引导这种交替进化，瓶颈感知进化反馈（Bottleneck-Aware Evolution Feedback, BEF）识别当前瓶颈是出现在记忆还是检索，并将优化方向指向更具限制性的一方。此外，VideoEvolve引入了能力感知进化反馈（Capability-Aware Evolution Feedback, CEF），以缓解下游反馈导致记忆过度专注于固定训练问题集的情况，推动训练向尚未开发但可学习的视频能力转移。通过将Agentic RL与BEF和CEF相结合，VideoEvolve将下游推理经验转化为可转移的能力更新，为从静态长视频系统向经验驱动的自我改进多模态智能提供了具体路径。在多个长视频理解基准上的广泛实验表明了VideoEvolve的有效性。
cs.CV / 123 / 2610.10196
HuLiGen: Human LiDAR Generation from Parametric Body Models
HuLiGen：基于参数化人体模型的人体LiDAR生成
Abstract
LiDAR point clouds of humans are extremely expensive to collect and annotate, thus represent a scarce resource that hinders the development of human analysis using this modality. To alleviate this scarcity, prior work relies on simulated human LiDAR, but such samples do not fully reflect the geometry and sensing characteristics of real observations. In contrast, we introduce HuLiGen, a generative model that generates human LiDAR point clouds from a parametric body model, using a point transformer trained with a flow-matching objective. We show that our generated point clouds are closer to the real capture distribution. Using HuLiGen to generate synthetic data, we propose a synthetic-only pretraining scheme for LiDAR-based HPE that achieves state-of-the-art performance, with even larger gains in low-annotation and low-data regimes, where MPJPE is reduced by up to 50%. Code, models and generated samples are available at https://github.com/valeoai/HuLiGen.
Chinese Translation
人体的LiDAR点云收集和标注成本极高，因此成为一种稀缺资源，阻碍了基于该模态的人体分析的发展。为缓解这一稀缺性，之前的研究依赖于模拟的人体LiDAR，但这些样本并未完全反映真实观测的几何形状和感知特性。相比之下，我们提出了HuLiGen，这是一种生成模型，能够从参数化人体模型生成人体LiDAR点云，使用经过流匹配目标训练的点变换器。我们展示了生成的点云更接近真实捕获分布。通过使用HuLiGen生成合成数据，我们提出了一种仅基于合成数据的LiDAR人类姿态估计（HPE）预训练方案，达到了最先进的性能，在低标注和低数据环境下更是取得了显著提升，MPJPE降低了多达50%。代码、模型和生成的样本可在 https://github.com/valeoai/HuLiGen 获取。
cs.CV / 124 / 2610.10197
VolCo: Volumetric Contact for High-Fidelity Human Grasp Generation
VolCo：用于高保真度人类抓取生成的体积接触
Abstract
Accurate contact modeling is fundamental to understanding hand-object interaction, yet existing contact representations are typically restricted to object surfaces and rely on hand-crafted rules to recover contact details, leading to severe penetrations and implausible results. To better exploit the rich detail in motion-capture data, we introduce Volumetric Contact (VolCo), a representation that expands surface points to a set of 3D volumetric grids. VolCo encodes 3D contact that allows precise hand part recovery, and is organized in an inherent hierarchy: local contact details within each volume and global hand geometry across all volumes. Our framework, VolCoDiff, employs two modules to capture local and global features following this hierarchy. For local contact details, we use a 3D variational autoencoder to model the possible hand configurations conditioned on the local object signed distance field (SDF). For global hand geometry, we design a prior-guided diffusion model that learns the distribution of compressed latent features aggregated from the volumetric grids. We evaluate our method on two benchmark datasets and demonstrate state-of-the-art performance in penetration and stability, indicating the capability to generate tight grasps with much less severe penetrations. Our code is available at https://github.com/chzh9311/volco.
Chinese Translation
准确的接触建模是理解手与物体交互的基础，然而现有的接触表示通常仅限于物体表面，并依赖手工规则来恢复接触细节，这导致了严重的穿透和不可信的结果。为了更好地利用运动捕捉数据中的丰富细节，我们引入了体积接触（Volumetric Contact，VolCo），一种将表面点扩展到一组三维体积网格的表示。VolCo 编码了三维接触，允许精确恢复手部部件，并在固有层次结构中组织：每个体积内的局部接触细节和所有体积之间的全局手部几何形状。我们的框架 VolCoDiff 采用两个模块来捕捉遵循该层次结构的局部和全局特征。对于局部接触细节，我们使用三维变分自编码器来建模基于局部物体有符号距离场（SDF）的可能手部配置。对于全局手部几何形状，我们设计了一种先验引导的扩散模型，学习从体积网格聚合的压缩潜在特征的分布。我们在两个基准数据集上评估了我们的方法，并展示了在穿透和稳定性方面的最先进性能，表明能够生成更紧凑的抓取，且穿透程度大大降低。我们的代码可在 https://github.com/chzh9311/volco 获取。
cs.CV / 125 / 2610.10225
Masked Feature Encoding for Large-Scale Whole Slide Image Representation
大规模全幻灯片图像表示的掩码特征编码
Abstract
Whole slide image (WSI) analysis in computational pathology follows a multiple instance learning (MIL) pipeline where patch embeddings are extracted independently and aggregated for slide-level prediction, but within-slide variance from staining, scanner, and local texture can overwhelm the discriminative signal. We propose Masked Feature Encoding for Multiple Instance Learning (MFE-MIL), a feature-space masking framework that trains a lightweight MLP adapter jointly with a window-based masked reconstruction branch and a MIL classification head. The two objectives are complementary. Classification guides the adapter to suppress within-slide patch variance, while window-based masked reconstruction provides an auxiliary regularizer for the adapted features without using patch coordinates, coordinate graphs, or segmentation preprocessing. The raster patch-extraction order is used only as a weak implicit prior. At inference, the decoder is removed, leaving only the adapter and MIL head. Across CAMELYON16/17, PANDA, and TCGA-BRCA with four diverse encoders, MFE-MIL improves ACC/F1 for nearly all tested aggregator-encoder settings and AUC in most, outperforms coordinate-based spatial methods (CAMIL), and achieves higher AUC than 2DMamba on three of four datasets (UNI). On five TCGA survival cohorts it improves the average concordance index for every aggregator tested, its most consistent gain. Code is available at https://github.com/AtlasAnalyticsLab/MFE-MIL.
Chinese Translation
全幻灯片图像（WSI）分析在计算病理学中遵循多实例学习（MIL）流程，其中独立提取补丁嵌入并聚合以进行幻灯片级预测，但由于染色、扫描仪和局部纹理的幻灯片内变异性可能会淹没判别信号。我们提出了用于多实例学习的掩码特征编码（MFE-MIL），这是一种特征空间掩码框架，它联合训练一个轻量级的多层感知器适配器、一个基于窗口的掩码重建分支和一个MIL分类头。这两个目标是互补的。分类指导适配器抑制幻灯片内补丁变异性，而基于窗口的掩码重建为适配特征提供了辅助正则化器，而不使用补丁坐标、坐标图或分割预处理。光栅补丁提取顺序仅作为一个弱的隐式先验。在推理时，解码器被移除，仅保留适配器和MIL头。在CAMELYON16/17、PANDA和TCGA-BRCA的四种不同编码器上，MFE-MIL改善了几乎所有测试的聚合器-编码器设置的ACC/F1，并在大多数情况下提高了AUC，超越了基于坐标的空间方法（CAMIL），在四个数据集中的三个上实现了比2DMamba更高的AUC（UNI）。在五个TCGA生存队列中，它提高了每个测试聚合器的平均一致性指数，这是其最一致的增益。代码可在https://github.com/AtlasAnalyticsLab/MFE-MIL获取。
cs.CV / 126 / 2610.10238
Geometry-Supervised Visual Representation Learning for Multi-Phenotype Lesion Interpretation in Medical VLMs
几何监督的视觉表征学习用于医学视觉语言模型中的多表型病变解读
Abstract
Medical vision-language models (VLMs) have shown increasing potential for clinical image interpretation. However, these models still struggle to interpret multi-phenotype lesions whose diagnosis requires the joint assessment of multiple pathological phenotypes. Existing vision-language alignment methods produce visual representations that fail to preserve anatomical hierarchies and relationships among phenotypic subclasses. This stems from their reliance on semantic supervision, which lacks geometric constraints to preserve these relationships in the visual embedding space. Moreover, the sparsity of lesion-related anatomical and phenotypic representations makes it difficult for medical VLMs to capture important diagnostic evidence. To address these limitations, we propose \textbf{PureVision}, a geometry-supervised visual representation learning framework for multi-phenotype lesion interpretation in medical VLMs. It combines a geometry-supervised representation learning module, \textbf{PureEyes}, and an anatomy-guided evidence aggregation module, \textbf{PureNeurons}. PureEyes provides geometric supervision through ideal spatial distributions that encode anatomical hierarchies and phenotypic subclass relationships. PureNeurons projects visual representations into the learned latent space, using their positions to selectively aggregate lesion-specific anatomical and phenotypic evidence. Experiments on \textit{LIDC-IDRI}, \textit{CBIS-DDSM}, and \textit{3DReasonKnee} demonstrate that PureVision improves lesion grounding and phenotype characterization in visual question answering and radiology report generation. Code is available at: https://anonymous.4open.science/r/purevision-06C2.
Chinese Translation
医学视觉语言模型（VLMs）在临床图像解读方面显示出越来越大的潜力。然而，这些模型在解读多表型病变时仍然存在困难，因为其诊断需要对多个病理表型进行联合评估。现有的视觉语言对齐方法生成的视觉表征未能保留解剖层级和表型子类之间的关系。这源于它们依赖于语义监督，而缺乏几何约束来保持这些关系在视觉嵌入空间中的一致性。此外，与病变相关的解剖和表型表征的稀疏性使得医学 VLMs 难以捕捉重要的诊断证据。为了解决这些局限性，我们提出了 extbf{PureVision}，一个用于医学 VLMs 中多表型病变解读的几何监督视觉表征学习框架。它结合了几何监督表征学习模块 extbf{PureEyes} 和解剖引导的证据聚合模块 extbf{PureNeurons}。PureEyes 通过理想的空间分布提供几何监督，这些分布编码了解剖层级和表型子类关系。PureNeurons 将视觉表征投影到学习的潜在空间中，利用其位置选择性地聚合特定于病变的解剖和表型证据。在 extit{LIDC-IDRI}、 extit{CBIS-DDSM} 和 extit{3DReasonKnee} 上的实验表明，PureVision 在视觉问答和放射学报告生成中改善了病变定位和表型特征化。代码可在以下链接获取： https://anonymous.4open.science/r/purevision-06C2。
cs.CV / 127 / 2610.10266
LoomSC: Scalable Deep Subspace Clustering with Projector Factorization and Exact Spectral Reduction
LoomSC：具有投影因子分解和精确谱降维的可扩展深度子空间聚类
Abstract
Dense self-expression matrices and full-affinity spectral clustering limit the scalability of subspace clustering. We introduce the Latent Orthogonal Optimization Model for Subspace Clustering (LoomSC), a framework that addresses both bottlenecks through projector factorization and exact spectral reduction. Motivated by the spectral structure of least-squares regression, LoomSC jointly learns latent features and a projector self-representation through two thin factors. Alternating Procrustes and least-squares updates preserve the sample factor's orthogonality while keeping the coefficient matrix implicit. We construct a nonnegative quadratic affinity that preserves the projector's support. An exact feature map then reduces its normalized spectral problem to an eigenproblem whose dimension depends only on the factor width. Neither the full affinity nor the sample Laplacian needs to be formed. Our analysis quantifies the projector approximation and identifies conditions for subspace preservation and within-subspace connectivity. For fixed dimensions and iteration budgets, the complete pipeline has linear time and memory complexity in the number of samples. Across five image-clustering benchmarks, LoomSC ranks first or second in all 15 dataset-metric comparisons against 9 state-of-the-art baselines. Its mean accuracy exceeds the highest baseline mean by 6.66 percentage points. Synthetic experiments scale to 500,000 samples while maintaining at least 99.8% accuracy.
Chinese Translation
稠密自表达矩阵和全关联谱聚类限制了子空间聚类的可扩展性。我们提出了子空间聚类的潜在正交优化模型（LoomSC），该框架通过投影因子分解和精确谱降维解决了这两个瓶颈。LoomSC的设计受到最小二乘回归的谱结构启发，联合学习潜在特征和通过两个薄因子获得的投影自表示。交替Procrustes和最小二乘更新在保持样本因子正交性的同时，保持系数矩阵的隐式性。我们构建了一个非负二次亲和度，保持投影的支持。然后，精确特征映射将其归一化谱问题简化为一个特征值问题，其维度仅依赖于因子宽度。无论是全关联还是样本拉普拉斯都不需要被构造。我们的分析量化了投影近似，并识别了子空间保持和子空间内连通性的条件。在固定的维度和迭代预算下，完整的管道在样本数量上具有线性时间和内存复杂度。在五个图像聚类基准测试中，LoomSC在与9个最先进基线的15个数据集-指标比较中排名第一或第二。其平均准确率超过最高基线平均值6.66个百分点。合成实验扩展到50万个样本，同时保持至少99.8%的准确率。
cs.CV / 128 / 2610.10270
Video Prediction Policy 2: Predict Better, Act Better
视频预测策略2：更好地预测，更好地行动
Abstract
World action models (WAMs) have emerged as an important class of generalist robot policies, aiming to transfer video prediction priors to action learning. However, we find that existing WAMs frequently produce incorrect motion predictions in open-ended environment, leading to erroneous actions. We attribute this limitation to two factors: (1) base video models are not optimized for manipulation, and (2) naively incorporating action components into video models can substantially degrade their generalization capabilities. We introduce Video Prediction Policy 2 (VPP2), a WAM that enables strong zero-shot generalization in both video prediction and action generation. First, we curate a large-scale, diverse dataset of manipulation videos to continue pretraining the base video foundation model. We annotate video clips with detailed captions and perform \textit{event-level} video pretraining to promote generalization across open-ended manipulation tasks. Second, we post-train and distill the video model into a single-step visual planner with fixed prediction horizon. Finally, we introduce action module via a mixture-of-transformers (MoT) architecture to learn implicit inverse dynamics model. Experiments demonstrate three key results: (1) VPP2-14B outperforms Cosmos3-64B by 11.0\% points in video prediction instruction-following success rate on open-ended tasks; (2) VPP2 surpasses the strongest baseline by 18.5\% points in success rate on real-world zero-shot ALOHA manipulation tasks; and (3) following benchmark-specific post-training, VPP2 achieves the highest success rates among evaluated methods on the challenging LIBERO-Pro, LIBERO-OOD, and RoboDojo benchmarks.
Chinese Translation
世界行动模型（WAMs）作为一种重要的通用机器人策略，旨在将视频预测先验知识转移到行动学习中。然而，我们发现现有的WAMs在开放式环境中经常产生不正确的运动预测，导致错误的行动。我们将这一局限性归因于两个因素：（1）基础视频模型未针对操控进行优化，以及（2）将行动组件简单地纳入视频模型可能会显著降低其泛化能力。我们提出了视频预测策略2（VPP2），这是一种WAM，能够在视频预测和行动生成中实现强大的零-shot 泛化。首先，我们策划了一个大规模、多样化的操控视频数据集，以继续对基础视频基础模型进行预训练。我们为视频片段添加了详细的注释，并进行事件级视频预训练，以促进在开放式操控任务中的泛化。其次，我们对视频模型进行后训练，并将其提炼为一个具有固定预测视野的一步视觉规划器。最后，我们通过混合变换器（MoT）架构引入行动模块，以学习隐式逆动力学模型。实验结果表明三个关键结果：（1）VPP2-14B在开放式任务的视频预测指令跟随成功率上比Cosmos3-64B提高了11.0个百分点；（2）VPP2在现实世界的零-shot ALOHA操控任务中成功率超越了最强基线18.5个百分点；（3）经过基准特定的后训练，VPP2在具有挑战性的LIBERO-Pro、LIBERO-OOD和RoboDojo基准测试中实现了评估方法中最高的成功率。
cs.CV / 129 / 2610.10286
$\Delta$Representation: Geometry Supervised Representation Learning of Phenotypes via Counterfactual Reasoning for Medical VLMs
Δ表示：通过反事实推理进行的基于几何监督的表型表示学习用于医学视觉语言模型
Abstract
Medical vision-language models (VLMs) have shown increasing potential for radiological image interpretation. Medical VLMs encode radiological images into visual representations that capture both anatomical and phenotypic information for diagnosis. Existing approaches improve pathological phenotype representations through semantic-guided representation alignment. However, pathological phenotypes arise as lesion-specific visual changes superimposed on underlying normal anatomy. Such semantic alignment approaches fail to model the phenotype-specific increment relative to the corresponding normal anatomical representation. To address this gap, we propose \textbf{$\Delta$Representation}, a visual phenotype representation learning framework based on counterfactual reasoning for medical VLMs. It comprises \textbf{BaseAnatomy}, a geometry-supervised representation learning module, and \textbf{$\Delta$Phenotype}, a counterfactual incremental representation learning module. BaseAnatomy provides fine-grained geometric supervision through spatial relationships across and within anatomical structures. $\Delta$Phenotype computes the representation increment between lesion representations and their corresponding normal anatomical representations, and supervises increments associated with the same phenotype to cluster in the representation space. Experiments on \textit{ReXGroundingCT} and \textit{LIDC-IDRI} demonstrate that $\Delta$Representation effectively structures pathological phenotype representations and improves lesion grounding and phenotype characterization accuracy in medical VLMs. Code is available at https://anonymous.4open.science/r/deltarep-CF6D.
Chinese Translation
医学视觉语言模型（VLMs）在放射影像解读方面展现出越来越大的潜力。医学VLMs将放射影像编码为视觉表示，以捕捉用于诊断的解剖和表型信息。现有的方法通过语义引导的表示对齐来改善病理表型表示。然而，病理表型表现为叠加在正常解剖基础上的特定病变视觉变化。这类语义对齐方法未能建模相对于相应正常解剖表示的表型特定增量。为了解决这一问题，我们提出了 extbf{Δ表示}，一种基于反事实推理的医学VLMs视觉表型表示学习框架。该框架包括 extbf{BaseAnatomy}，一个基于几何监督的表示学习模块，以及 extbf{ΔPhenotype}，一个反事实增量表示学习模块。BaseAnatomy通过解剖结构之间及内部的空间关系提供细粒度的几何监督。ΔPhenotype计算病变表示与其对应正常解剖表示之间的表示增量，并监督与相同表型相关的增量在表示空间中聚类。在 extit{ReXGroundingCT}和 extit{LIDC-IDRI}上的实验表明，Δ表示有效地构建了病理表型表示，并提高了医学VLMs中病变定位和表型特征化的准确性。代码可在 https://anonymous.4open.science/r/deltarep-CF6D 获取。
cs.CV / 130 / 2610.10288
TouchScale: 500 Hours of Human Vision and Touch for Visual-Tactile Learning
TouchScale：500小时的人类视觉与触觉用于视觉-触觉学习
Li, Dayou, Wang, Hao, Yang, Qianqian, Zhu, Zihao, Fang, Haoquan, Zeng, Ziyao, Han, Yan, Wang, Zihan, Wang, Yan, Huang, Baoru, Wang, Dilin, Shimada, Kenji, Luo, Yiyue, Li, Manling, Lv, Teresa, Mukadam, Mustafa, Ranjan, Rakesh, Zhang, Ruohan, He, Qi, Liu, Changliu, Chen, Xu, Pavone, Marco, Liu, Bangya, Li, Jiachen, Tomizuka, Masayoshi, Fan, Zhiwen
Abstract
Large-scale egocentric human interaction data is becoming an important source of physical supervision for embodied learning, yet video alone leaves the contact and pressure that characterize physical interaction unrecorded. Recent visual-tactile datasets provide this missing supervision, but their synchronized tactile data remain far smaller in volume than human video. Moreover, the largest resources often merge recordings from different sensors or annotation procedures, which makes the effect of data scale difficult to isolate. We therefore introduce TouchScale, a 500-hour dataset of contact-rich human interaction recorded with a single unified wearable setup. Its approximately 2K predefined task descriptions span everyday activities and structured manipulation, and each recording temporally aligns egocentric RGB-D video with wrist RGB video and dense full-hand bimanual tactile measurements. Compared with prior tactile data, training on the full TouchScale raises zero-shot contact IoU on data from an unseen tactile sensor from 0.134 to 0.383. Pretraining a visual encoder on TouchScale also yields the highest action recognition accuracy on three benchmarks among the compared visual-tactile datasets. Used for visual-tactile mid-training of a robot policy, TouchScale improves the average real-world success rate across four contact-rich manipulation tasks from 22.5% to 57.5%. With the sensor and collection protocol held fixed, both zero-shot tactile prediction and robot success show an overall upward trend as more TouchScale data is used. These results suggest that human visual-tactile data collected at scale with consistent sensing benefits both perception and robot manipulation. We will publicly release TouchScale, including all synchronized visual-tactile recordings and reconstructed object models, to support future research on scalable visual-tactile learning.
Chinese Translation
大规模的自我中心人类交互数据正成为具身学习的重要物理监督来源，但仅靠视频无法记录物理交互中接触和压力的特征。近期的视觉-触觉数据集提供了这一缺失的监督，但其同步的触觉数据量仍远小于人类视频。此外，最大的资源往往合并了来自不同传感器或注释程序的记录，这使得数据规模的影响难以隔离。因此，我们引入了TouchScale，一个500小时的接触丰富的人类交互数据集，使用单一统一的可穿戴设备进行记录。其约2000个预定义任务描述涵盖日常活动和结构化操作，每个记录在时间上将自我中心的RGB-D视频与手腕RGB视频和密集的全手双手触觉测量对齐。与先前的触觉数据相比，在完整的TouchScale上进行训练使得在未见触觉传感器的数据上的零-shot接触IoU从0.134提升至0.383。在TouchScale上预训练的视觉编码器在与比较的视觉-触觉数据集中的三个基准上也取得了最高的动作识别准确率。用于机器人的视觉-触觉中期训练，TouchScale使四个接触丰富的操作任务的平均现实世界成功率从22.5%提升至57.5%。在传感器和收集协议保持不变的情况下，随着使用更多TouchScale数据，零-shot触觉预测和机器人成功率整体呈上升趋势。这些结果表明，以一致的传感方式大规模收集的人类视觉-触觉数据对感知和机器人操作都有益。我们将公开发布TouchScale，包括所有同步的视觉-触觉记录和重建的物体模型，以支持未来可扩展视觉-触觉学习的研究。
cs.CV / 131 / 2610.10303
On the Necessity of Attention-FFN Split in Vision Transformers
视觉变换器中注意力-前馈网络分离的必要性
Abstract
The standard Transformer architecture relies on a rigid pattern that alternates Attention and Feed-Forward Network (FFN) layers. Despite its widespread adoption, the inductive bias imposed by this strict separation has not been systematically examined. In this work, we investigate the necessity of the Attention-FFN dichotomy in Vision Transformers (ViTs). To facilitate this analysis, we introduce the AttenFeed module, a unified component that integrates the functional properties of both Attention and FFN. Based on this module, we devise the unified Vision Transformer (uViT), which replaces the conventional alternating Attention-FFN structure with a sequence of AttenFeed modules. We then use uViT as a control group that relaxes the Attention-FFN dichotomy of the standard ViT and systematically compare the two models across multiple datasets and model scales. Our experiments reveal that the Attention-FFN dichotomy can hinder performance at smaller model scales due to the rigid parameter allocation of ViTs. The AttenFeed module and uViT serve as new analytical tools for understanding the Attention-FFN structure and offer theoretical insights into the heuristically designed architecture of conventional ViTs.
Chinese Translation
标准的变换器架构依赖于交替使用注意力层和前馈网络（FFN）层的严格模式。尽管这种模式被广泛采用，但这种严格分离所施加的归纳偏差尚未得到系统的检验。在本研究中，我们探讨了视觉变换器（ViTs）中注意力-前馈网络二分法的必要性。为了便于分析，我们引入了AttenFeed模块，这是一种统一组件，整合了注意力和前馈网络的功能特性。基于该模块，我们设计了统一视觉变换器（uViT），它用一系列AttenFeed模块替代了传统的交替注意力-前馈网络结构。然后，我们将uViT作为一个对照组，放宽标准ViT的注意力-前馈网络二分法，并在多个数据集和模型规模上系统地比较这两种模型。我们的实验表明，注意力-前馈网络二分法可能会由于ViTs的严格参数分配而在较小模型规模下阻碍性能。AttenFeed模块和uViT作为理解注意力-前馈网络结构的新分析工具，为传统ViTs的启发式设计架构提供了理论见解。
cs.CV / 132 / 2610.10306
One-Shot Adaptive Segmentation For Scientific Images
针对科学图像的一次性自适应分割
Abstract
Scientific image segmentation methods rely on extensive annotation and task-specific training, limiting adaptation across imaging modalities and experimental conditions. We present a training-free, one-shot framework that specializes vision foundation models using a single annotated reference image. The framework combines DINOv3 representations with background-adaptive feature orthogonalization to suppress artifact-related feature directions, after which cosine similarity localizes candidate regions for SAM segmentation. We evaluate the framework on red-blood-cell microscopy, structured-illumination pool boiling, and chest radiography. Relative to the strongest baseline, the proposed method improves mean IoU by 5.91% and 78.62% on the microscopy and pool-boiling datasets, respectively, while achieving comparable performance on chest radiographs. These results demonstrate that one-shot reference conditioning can adapt general-purpose vision models to specialized scientific segmentation tasks.
Chinese Translation
科学图像分割方法依赖于大量的标注和特定任务的训练，这限制了其在不同成像模式和实验条件下的适应性。我们提出了一种无训练的一次性框架，该框架使用单个带标注的参考图像来专门化视觉基础模型。该框架结合了DINOv3表示和背景自适应特征正交化，以抑制与伪影相关的特征方向，随后通过余弦相似度定位候选区域以进行SAM分割。我们在红细胞显微镜成像、结构照明池沸腾和胸部X光片上评估了该框架。与最强基线相比，所提出的方法在显微镜和池沸腾数据集上分别提高了平均IoU 5.91%和78.62%，同时在胸部X光片上实现了可比的性能。这些结果表明，一次性参考条件可以将通用视觉模型适应于特定的科学分割任务。
cs.CV / 133 / 2610.10322
From Digital Human Interactions to Physics-Based Humanoid Skills: Physics-Grounded Post-Training of Interaction Generators
从数字人类交互到基于物理的人形技能：基于物理的交互生成器后训练
Abstract
Recent methods have made promising progress in generating interactions between two humanoids, largely relying on physics-based tracking policies to convert digital reference motions into executable trajectories. However, limited tracking capabilities restrict the range of reference motions that can be successfully executed, reducing data utilization. Moreover, even successful tracking does not guarantee physically plausible responses or faithful realization of the intended interactions. In this paper, we introduce DIGHT, a co-adaptive framework that couples a Digital human Interaction Generator with a Humanoid Tracking policy. Our DIGHT first executes multiple text-conditioned interaction candidates in simulation using a fixed tracker. It then constructs physics-grounded preferences from the resulting rollouts, covering both general executability and interaction fidelity. Rather than collapsing these signals into a single scalar reward for candidate ranking, we align the pretrained generator using physics-decoupled diffusion direct preference optimization (DPO), preserving criterion-specific supervision without differentiating through the simulator. To improve executability, preference pairs are derived from tracking error, friction, and floating. Additionally, to improve interaction fidelity, we propose to incorporate force feedback from simulator as a measure of contact fidelity and construct preferences over contact occurrence, location, duration, and force magnitude. The aligned generator then supplies reference motions for fine-tuning the tracker, improving compatibility between generation and physical execution. Extensive experiments demonstrate that our approach not only improves the physical plausibility of generated motions but also enables more reliable and faithful humanoid interactions in simulation.
Chinese Translation
最近的方法在生成两个类人之间的交互方面取得了可喜的进展，主要依赖基于物理的跟踪策略将数字参考动作转换为可执行轨迹。然而，有限的跟踪能力限制了能够成功执行的参考动作范围，从而降低了数据的利用率。此外，即使成功跟踪，也并不能保证物理上合理的响应或忠实实现预期的交互。在本文中，我们介绍了DIGHT，一个将数字人类交互生成器与人形跟踪策略相结合的共适应框架。我们的DIGHT首先使用固定跟踪器在模拟中执行多个文本条件的交互候选。然后，它从生成的回放中构建基于物理的偏好，涵盖一般可执行性和交互保真度。我们并不将这些信号合并为单一标量奖励进行候选排名，而是通过物理解耦扩散直接偏好优化（DPO）对预训练生成器进行对齐，保留特定标准的监督而不通过模拟器进行微分。为了提高可执行性，偏好对是从跟踪误差、摩擦和漂浮中得出的。此外，为了提高交互保真度，我们建议将来自模拟器的力反馈作为接触保真度的衡量标准，并构建关于接触发生、位置、持续时间和力大小的偏好。对齐后的生成器随后为微调跟踪器提供参考动作，提高生成与物理执行之间的兼容性。大量实验表明，我们的方法不仅提高了生成动作的物理合理性，还在模拟中实现了更可靠和真实的人形交互。
cs.CV / 134 / 2610.10324
Performance at What Cost? A Sustainability-Aware Performance Index for Cell and Nucleus Instance Segmentation
性能的代价是什么？一种关注可持续性的细胞和细胞核实例分割性能指数
Abstract
Pretrained models for cell and nuclear instance segmentation differ substantially in architecture, pretraining data and objectives, parameter count, inference strategy, adaptation requirements, postprocessing pipeline, and computational demand. Large pretrained and foundation models are increasingly adopted because of their strong zero-shot capabilities, but their use also imposes greater energy consumption, memory requirements, computational demands, adaptation costs, and operational carbon emissions. Whether these additional demands are justified by meaningful gains in segmentation performance remains unclear. We address this question by introducing the Sustainability-Aware Performance Index (SAPI), a configurable metric that combines segmentation performance, energy consumption, and model size. We benchmark 19 pretrained and foundation models across six CellBinDB datasets under zero-shot inference and evaluate 16 fine-tunable models using few-shot adaptation with both frozen encoder and full-model fine-tuning. We estimate energy consumption for GPU, CPU, and RAM using software-based monitoring tools. Our results show that larger and more computationally demanding models do not consistently achieve proportionate improvements in segmentation quality. While few-shot adaptation benefits several models, the gains and resource costs vary considerably across architectures, datasets, and adaptation strategies, causing SAPI-based rankings to differ from rankings based on performance alone. This study provides a practical framework for comparing segmentation models more comprehensively and supports more computationally accessible and environmentally responsible model selection in biomedical image analysis.
Chinese Translation
用于细胞和细胞核实例分割的预训练模型在架构、预训练数据和目标、参数数量、推理策略、适应需求、后处理流程以及计算需求等方面存在显著差异。由于其强大的零-shot 能力，大型预训练和基础模型越来越受到青睐，但它们的使用也带来了更高的能耗、内存需求、计算负担、适应成本和运营碳排放。这些额外的需求是否通过显著的分割性能提升得到了合理的 justification 仍不清楚。我们通过引入可持续性意识性能指数（Sustainability-Aware Performance Index, SAPI）来解决这个问题，这是一种可配置的指标，结合了分割性能、能耗和模型大小。我们在六个 CellBinDB 数据集上对 19 个预训练和基础模型进行了基准测试，采用零-shot 推理，并使用冻结编码器和全模型微调的少量样本适应评估了 16 个可微调模型。我们使用基于软件的监控工具估算 GPU、CPU 和 RAM 的能耗。我们的结果表明，较大且计算需求更高的模型并不总是能在分割质量上实现成比例的提升。虽然少量样本适应对多个模型有益，但不同架构、数据集和适应策略之间的收益和资源成本差异显著，导致基于 SAPI 的排名与仅基于性能的排名有所不同。本研究提供了一个实用框架，以更全面地比较分割模型，并支持在生物医学图像分析中进行更具计算可达性和环境责任感的模型选择。
cs.CV / 135 / 2610.10334
How Private is Private? A Comparative Study for Face De-Identification
隐私到底有多隐私？面部去标识化的比较研究
Abstract
Face de-identification (FDeID) has emerged as a critical privacy-preserving technology, yet its evaluation remains fundamentally fragmented. Existing protocols rely on inconsistent metrics, heterogeneous datasets, and partial annotation coverage, so methods targeting different utility dimensions, such as landmark versus expression preservation, are reported on different benchmarks under different metrics, rendering cross-method comparison infeasible. We revisit FDeID evaluation from both the data and metric perspectives. On the data side, we introduce UtilFace, a curated, demographically balanced benchmark with high identity diversity, assembled from four large-scale face datasets through identity-aware cleaning, resolution enhancement, and stratified filtering. On the metric side, we propose HiFD, a Hierarchical Face De-identification metric that unifies identity suppression, multi-level utility preservation, and image quality under a single consistency-based paradigm: every component is computed from pretrained estimators' outputs on the original face and its de-identified counterpart, directly quantifying how much identity is suppressed and how much downstream-perceivable utility survives. HiFD organizes facial signals into a three-level utility hierarchy spanning macro cues (L1), micro cues (L2), and imperceptible cues (L3), and aggregates the five resulting components into a single interpretable score via weighted harmonic mean, with configurable application-specific profiles. Using this unified protocol, we conduct a comprehensive comparative study spanning adversarial, GAN-based, and diffusion-based methods, surfacing trade-offs and failure modes that remain invisible under existing protocols. We release the benchmark and evaluation toolkit to foster systematic and reproducible research in privacy-preserving human face analysis.
Chinese Translation
面部去标识化（FDeID）作为一种关键的隐私保护技术，已逐渐受到重视，但其评估仍然存在根本性的碎片化问题。现有协议依赖于不一致的指标、异质的数据集和部分注释覆盖，因此针对不同效用维度（如地标与表情保留）的不同方法在不同的基准下以不同的指标进行报告，使得跨方法比较变得不可行。我们从数据和指标两个角度重新审视FDeID评估。在数据方面，我们引入了UtilFace，这是一个经过策划的、在人口统计上平衡的基准，具有高身份多样性，来源于四个大规模面部数据集，通过身份感知清理、分辨率增强和分层过滤而构建。在指标方面，我们提出了HiFD，这是一种层次化面部去标识化指标，将身份抑制、多级效用保留和图像质量统一在一个基于一致性的范式下：每个组件都是从预训练估计器在原始面部及其去标识化对应物上的输出计算得出的，直接量化身份被抑制的程度和下游可感知效用的保留程度。HiFD将面部信号组织成一个涵盖宏观线索（L1）、微观线索（L2）和不可感知线索（L3）的三层效用层级，并通过加权调和平均将五个结果组件汇总为一个可解释的分数，具有可配置的应用特定配置文件。利用这一统一协议，我们开展了一项全面的比较研究，涵盖对抗性、基于GAN和基于扩散的方法，揭示了在现有协议下仍然不可见的权衡和失败模式。我们发布了该基准和评估工具包，以促进隐私保护人脸分析领域的系统性和可重复性研究。
cs.CV / 136 / 2610.10335
When to Unpair: Regulating Pairing Dependence in Medical Visual In-Context Learning
何时解除配对：调节医学视觉上下文学习中的配对依赖性
Abstract
Visual in-context learning (ICL), well suited to label-scarce medical imaging, uses support image-label pairs to demonstrate input-output mappings, while the labels collectively indicate the requested task. We diagnose dependence on individual pairings with a test-time derangement that reassigns every support label to another support image while preserving the query, support images, and label multiset. The resulting pairing gap, defined as shuffled-minus-matched performance, shows that all four released models depend on the pairing, to widely varying degrees. Further analysis of a paired-trained model reveals support-associated spurious regions and lesion-size biases even with real, unaltered supports, alongside sensitivity to mis-registered support labels. To regulate this dependence, we introduce a late unpairing curriculum (LUC), which starts with matched training and then applies random unpairing, replacing each support label with that of another support in the same episode. LUC nearly closes the pairing gap on two backbones while maintaining or improving matched-support performance across all evaluated task types, with gains extending to held-out tasks and cross-dataset episodes. It also mitigates these failure modes. On BraTS whole-tumor segmentation, matched-support DSC rises from 0.733 to 0.857 while the gap shrinks from -0.184 to -0.008. In a released model, brief fine-tuning with random unpairing reduces the gap. A reversed curriculum that places the same number of unpairing epochs at the start of training leaves a large gap. This shows that pairing dependence is shaped by the order of training and not only by the amount of unpaired training.
Chinese Translation
视觉上下文学习（ICL）非常适合标签稀缺的医学影像，它利用支持图像-标签对来展示输入-输出映射，同时标签共同指示请求的任务。我们通过一种测试时的扰动方法诊断对单个配对的依赖性，该方法在保持查询、支持图像和标签多重集的同时，将每个支持标签重新分配给另一个支持图像。由此产生的配对差距定义为打乱后减去匹配的性能，显示出所有四个发布的模型在不同程度上依赖于配对。对一个经过配对训练的模型的进一步分析揭示了与支持相关的虚假区域和病灶大小偏差，即使在真实且未修改的支持下，也存在对错误注册的支持标签的敏感性。为了调节这种依赖性，我们引入了一种后期解除配对的课程（LUC），该课程从匹配训练开始，然后应用随机解除配对，将每个支持标签替换为同一回合中另一个支持的标签。LUC几乎消除了两个基础模型上的配对差距，同时在所有评估的任务类型中保持或提高了匹配支持的性能，收益扩展到保留任务和跨数据集回合。在BraTS全肿瘤分割中，匹配支持的DSC从0.733上升到0.857，而差距从-0.184缩小到-0.008。在一个发布的模型中，短暂的随机解除配对微调减少了差距。一个将相同数量的解除配对周期放在训练开始的反向课程留下了较大的差距。这表明配对依赖性是由训练顺序塑造的，而不仅仅是由无配对训练的数量决定的。
cs.CV / 137 / 2610.10342
Position Forcing: Self-Conditioning 3D Generation
位置强制：自条件化的三维生成
Abstract
Recent single-stage 3D generative models commonly adopt VecSet representations, encoding 3D shapes as unordered sets of latent tokens. However, compared with two-stage methods that provide explicit positional guidance, these models must implicitly infer token positions throughout denoising, limiting their generation quality. We observe that, despite the absence of explicit positional conditioning, VecSet tokens retain recoverable spatial correspondences. Building on this observation, we propose Position Forcing, a position-based self-conditioning framework. During denoising, Position Forcing recovers token positions from the current clean latent estimate, quantizes them at progressively finer resolutions according to the denoising stage, and feeds the resulting positional encodings back into the diffusion Transformer. This progressively refined positional feedback provides spatial guidance at a granularity appropriate to each denoising stage, guiding shape generation along a coarse-to-fine trajectory and substantially improving generation quality without a separate position generation stage. Experiments demonstrate that Position Forcing achieves strong performance among single-stage 3D generative methods and outperforms several competitive multi-stage approaches.
Chinese Translation
最近的单阶段三维生成模型通常采用 VecSet 表示法，将三维形状编码为无序的潜在标记集合。然而，与提供明确位置指导的双阶段方法相比，这些模型必须在去噪过程中隐式推断标记位置，从而限制了生成质量。我们观察到，尽管缺乏明确的位置条件，VecSet 标记仍然保留可恢复的空间对应关系。基于这一观察，我们提出了位置强制（Position Forcing），一种基于位置的自条件化框架。在去噪过程中，位置强制从当前的干净潜在估计中恢复标记位置，根据去噪阶段以逐步更精细的分辨率对其进行量化，并将生成的位置信息编码反馈到扩散 Transformer 中。这种逐步细化的位置反馈在适合每个去噪阶段的粒度上提供空间指导，沿着粗到细的轨迹引导形状生成，并显著提高生成质量，而无需单独的位置生成阶段。实验表明，位置强制在单阶段三维生成方法中表现出色，并且优于几种竞争性的多阶段方法。
cs.CV / 138 / 2610.10343
Real-Time Joint Audio-Video Generation by Parallel Adapter Composition
通过并行适配器组合实现实时联合音视频生成
Abstract
Deploying a joint audio-video diffusion transformer for real-time, interactive generation normally requires two essential modifications: block-autoregressive attention, so frames can be emitted before the whole clip is finished, and few-step sampling, so each block is cheap. Conventionally, the streaming video literature obtains both capabilities from a chained pipeline. It first distills a bidirectional teacher into a causal student, then into a few-step one, or proceeds in reverse order. Each stage of such a chain fine-tunes the weights the previous one produced, so a later objective can undo an earlier capability. Following the idea of model merging, we show that on a packed audio-video backbone the two capabilities can be acquired in parallel. A causal adapter is trained against the frozen backbone, and an off-the-shelf few-step adapter provides the few-step capability. As the two edit different functional axes, we predict, and then verify, that their weight-update directions are near-orthogonal, without any explicit orthogonality constraint during training. Orthogonal updates should combine without interfering, so parallel composition is a direct sum. The two adapters are simply added at inference, with no joint training, yielding few-step, streaming audio-video whose image quality tracks the bidirectional teacher. Compared to the chained baselines, the composed model matches or beats them on most metrics, making parallel composition a practical approach. The resulting streaming system generates joint audio-video in real time, $\approx$26 fps at $480\times832$ without quantization, and sustains 30 s of continuous generation with stable image quality.
Chinese Translation
部署一个联合音视频扩散变换器以实现实时交互生成通常需要两个基本修改：块自回归注意力，以便在整个片段完成之前可以发出帧，以及少步采样，以使每个块的计算成本较低。传统上，流媒体视频文献通过链式管道获得这两种能力。它首先将双向教师模型提炼为因果学生模型，然后再提炼为少步模型，或按相反顺序进行。该链的每个阶段都会微调前一个阶段产生的权重，因此后续目标可以撤销早期能力。遵循模型合并的思路，我们展示了在一个紧凑的音视频主干上，这两种能力可以并行获得。一个因果适配器在冻结的主干上进行训练，而一个现成的少步适配器提供少步能力。由于这两个适配器编辑不同的功能轴，我们预测并验证了它们的权重更新方向几乎正交，在训练过程中没有任何显式的正交性约束。正交更新应该可以无干扰地组合，因此并行组合是一个直接和。两个适配器在推理时简单相加，无需联合训练，生成的少步流媒体音视频图像质量与双向教师模型相当。与链式基线相比，组合模型在大多数指标上与之匹配或超越，使得并行组合成为一种实用的方法。最终生成的流媒体系统以约26帧每秒的速度在480×832的分辨率下实时生成联合音视频，并能够持续生成30秒，保持稳定的图像质量。
cs.CV / 139 / 2610.10390
Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving
自主驾驶中的视觉-语言-行动显式几何思维链
Abstract
Vision-language-action~(VLA) models have emerged as a promising paradigm for autonomous driving. However, existing VLA models still suffer from a fundamental mismatch: driving actions require precise 3D geometric cues, while visual-language understanding and reasoning are largely conducted in a 2D semantic space. In this paper, we propose GeoCoTDrive, an explicit geometric chain-of-thought framework that grounds geometry in a planning-oriented manner. GeoCoTDrive follows a think with 2D first, drive with dedicated 3D priors paradigm. It first grounds 2D regions corresponding to decision-critical cues, and then retrieves localized 3D priors by sampling features from a geometric foundation model within the grounded regions. These localized geometric features are interleaved into the autoregressive context to support the trajectory generation. To supervise this process, we introduce planning-relevant grounding, a new region-level grounding task that focuses on local spatial cues directly affecting ego planning decisions, and construct the PlanningGrounding dataset to endow VLAs with planning-oriented grounding capability. Experiments across multiple end-to-end autonomous driving benchmarks show that GeoCoTDrive consistently improves safety-critical planning performance, demonstrating the effectiveness of the explicit geometric chain-of-thought process for VLA-based planning.
Chinese Translation
视觉-语言-行动（VLA）模型已成为自主驾驶的一个有前景的范式。然而，现有的VLA模型仍然存在一个基本的不匹配：驾驶行为需要精确的三维几何线索，而视觉-语言理解和推理主要在二维语义空间中进行。本文提出了GeoCoTDrive，一个显式几何思维链框架，以规划导向的方式将几何信息与之结合。GeoCoTDrive遵循‘先用二维思考，再用专用三维先验进行驾驶’的范式。它首先将与决策关键线索对应的二维区域进行定位，然后通过从几何基础模型中采样特征来检索局部三维先验。这些局部几何特征被交错融入自回归上下文中，以支持轨迹生成。为了监督这一过程，我们引入了与规划相关的定位，这是一个新的区域级定位任务，专注于直接影响自我规划决策的局部空间线索，并构建了PlanningGrounding数据集，以赋予VLA规划导向的定位能力。在多个端到端自主驾驶基准测试中的实验表明，GeoCoTDrive始终提高了安全关键的规划性能，证明了显式几何思维链过程在基于VLA的规划中的有效性。
cs.CV / 140 / 2610.10391
MOTIP2: Spatial Priors for End-to-End Multi-Object Tracking
MOTIP2：端到端多目标跟踪的空间先验
Abstract
End-to-end multi-object trackers have narrowed the gap with classical tracking-by-detection on association-difficult benchmarks. Yet they still make spatially implausible errors no classical tracker would, such as assigning one identity to objects on opposite sides of the frame. A model could learn to avoid them, but tracking annotations are scarce, so we encode spatial priors explicitly instead, while keeping inference fully end-to-end with no post-hoc association. We propose three spatial priors, at the data, loss, and representation stages. Spatial ID Switches bias trajectory permutations toward spatially overlapping objects, reducing the mismatch between training and inference confusions. Spatial ID Loss scales each identity's penalty by its box distance, so a distant switch costs more than a nearby one. Spatial Anchor gives each track token its frame position, an explicit spatial cue for attention. We instantiate the three priors in MOTIP2, a tracker adapted from MOTIP and built on the real-time DEIM detection transformer. Trained without extra data, its main model, MOTIP2-L, sets a new state of the art: 73.4 HOTA on DanceTrack, 76.0 on SportsMOT, and 71.1 IDF1 on PersonPath22. MOTIP2 is a family of models spanning the speed-accuracy trade-off: a lighter model, MOTIP2-S, matches the original MOTIP at over 3x the speed, and MOTIP2-X reaches 74.8 HOTA on DanceTrack.
Chinese Translation
端到端多目标跟踪器在难以关联的基准测试中已缩小了与经典检测跟踪方法的差距。然而，它们仍然会犯一些经典跟踪器不会犯的空间上不合理的错误，例如将一个身份分配给帧两侧的物体。模型可以学习避免这些错误，但跟踪注释稀缺，因此我们显式地编码空间先验，同时保持推理完全端到端，没有事后关联。我们在数据、损失和表示阶段提出了三种空间先验。空间身份切换（Spatial ID Switches）使轨迹排列偏向空间重叠的物体，从而减少训练和推理混淆之间的不匹配。空间身份损失（Spatial ID Loss）根据每个身份的框距离来缩放惩罚，因此远处的切换成本高于近处的切换。空间锚点（Spatial Anchor）为每个跟踪令牌赋予其帧位置，提供了一个显式的空间线索以供注意。我们在MOTIP2中实例化了这三种先验，该跟踪器是从MOTIP改编而来，并基于实时DEIM检测变换器构建。MOTIP2的主模型MOTIP2-L在没有额外数据的情况下训练，创造了新的最先进水平：在DanceTrack上达到73.4 HOTA，在SportsMOT上达到76.0，在PersonPath22上达到71.1 IDF1。MOTIP2是一系列模型，涵盖了速度与准确性的权衡：更轻的模型MOTIP2-S在速度上超过原始MOTIP的3倍，而MOTIP2-X在DanceTrack上达到了74.8 HOTA。
cs.CV / 141 / 2610.10396
Gaussian Density Splatting Network
高斯密度溅射网络
Abstract
This paper proposes a novel crowd counting approach, the Gaussian Density Splatting Network (GDSNet). Unlike methods that rely on conventional, grid-based density maps and are sensitive to spatial resolution, GDSNet represents a crowd as a superposition of continuous 2D Gaussian primitives. Our approach is built upon two key contributions. First, we introduce a control-point-based fitting mechanism to structure the prediction of the Gaussian parameters. We design a method to allocate a set of control points that define local regions, from which features are pooled to regress each primitive's parameters. Second, we adapt a differentiable Gaussian Splatting framework to the counting task by parameterizing each primitive with geometric parameters and a scalar density mass. This formulation allows the network to be trained end-to-end via spatial matching of differentiably rendered density maps, naturally providing both local density supervision and global count optimization. Extensive evaluations on four standard benchmarks show GDSNet consistently outperforms the state of the art.
Chinese Translation
本文提出了一种新颖的人群计数方法——高斯密度溅射网络（Gaussian Density Splatting Network，GDSNet）。与依赖于传统网格密度图且对空间分辨率敏感的方法不同，GDSNet将人群表示为连续二维高斯基元的叠加。我们的方法基于两个关键贡献。首先，我们引入了一种基于控制点的拟合机制来构建高斯参数的预测。我们设计了一种方法来分配一组控制点，以定义局部区域，从这些区域中提取特征以回归每个基元的参数。其次，我们通过用几何参数和标量密度质量对每个基元进行参数化，将可微分的高斯溅射框架适配到计数任务中。这种表述使得网络能够通过可微分渲染的密度图的空间匹配进行端到端训练，自然提供了局部密度监督和全局计数优化。对四个标准基准的广泛评估表明，GDSNet在性能上始终优于当前最先进的方法。
cs.CV / 142 / 2610.10400
Self-correction Optimization for Interleaved Multimodal Generation
交错多模态生成的自我修正优化
Abstract
Multimodal large language models (MLLMs) have made significant progress in visual understanding and generation. However, generating interleaved image--text content remains challenging, as it requires tightly integrated multimodal understanding and generation capabilities. Although existing MLLMs provide promising solutions, most rely on additional training with augmented data, which is computationally expensive and remains limited in preserving visual subjects, temporal consistency, and physical plausibility. In this work, we propose self-correction optimization (SCO), an effective training-free method for consistent interleaved generation. SCO treats the classifier-free guidance update as a reference and performs minimal self-correction under two complementary constraints, including new-event and state-preserving constraints. Specifically, the new-event constraint promotes temporal consistency across image--text sequences, while the state-preserving constraint maintains the coherence of visual subjects throughout subsequent generation steps. Experiments on challenging interleaved multimodal generation benchmarks demonstrate significant improvements in temporal coherence and visual-subject preservation. Furthermore, SCO can be extended to video generation and improves the modeling of physically grounded processes, including robot manipulation and long-horizon handcrafting.
Chinese Translation
多模态大型语言模型（MLLMs）在视觉理解和生成方面取得了显著进展。然而，生成交错的图像-文本内容仍然具有挑战性，因为这需要紧密集成的多模态理解和生成能力。尽管现有的MLLMs提供了有前景的解决方案，但大多数依赖于使用增强数据进行额外训练，这在计算上是昂贵的，并且在保持视觉主题、时间一致性和物理合理性方面仍然有限。在本研究中，我们提出了一种自我修正优化（SCO），这是一种有效的无训练方法，用于一致的交错生成。SCO将无分类器引导更新视为参考，并在包括新事件约束和状态保持约束的两个互补约束下进行最小自我修正。具体而言，新事件约束促进了图像-文本序列之间的时间一致性，而状态保持约束则在后续生成步骤中保持视觉主题的一致性。在具有挑战性的交错多模态生成基准上的实验表明，时间一致性和视觉主题保持方面有显著改善。此外，SCO可以扩展到视频生成，并改善物理基础过程的建模，包括机器人操作和长时间手工制作。
cs.CV / 143 / 2610.10408
Rubix: Global Correspondence-Free Point Set Alignment through Assignment Geometry
Rubix：通过赋值几何实现全局无对应点集对齐
Abstract
Procrustes-Wasserstein alignment jointly estimates a matching and rotation without supplied correspondences, but alternating minimization can stop at suboptimal solutions. Rubix solves the equally weighted planar problem globally under squared Euclidean loss. Each matching $\sigma$ of two centered $n$-point sets defines a complex correlation $z_\sigma=\sum_i\bar x_i y_{\sigma(i)}$. Their convex hull is the permutation polygon: supporting vertices give optimal matchings at fixed rotations, and the farthest vertex gives the global alignment. We prove the sharp bound of $n(n-1)$ vertices for $n\ge2$, answering Rote's rotation-assignment open problem. In exact arithmetic, assignment queries recover the polygon in $\mathcal O(n^5)$ operations. Assignment-based bounds extend the approach to three-dimensional rotations and partial matching at a supplied translation through branch-and-bound. On timed MPEG-7 shape pairs, Rubix attains every numerical reference value in 12 ms on average, 50 times faster than a rotation grid at the same accuracy. Its distances improve gravity-aligned matching of real 3D scans, shape retrieval and noisy crystal classification over alternating minimization.
Chinese Translation
Procrustes-Wasserstein 对齐在没有提供对应的情况下联合估计匹配和旋转，但交替最小化可能会停留在次优解。Rubix 在平方欧几里得损失下全局解决了等权平面问题。两个中心化的 n 点集的每个匹配 $ au$ 定义了一个复杂相关性 $z_ au= extstyleigsum_iar x_i y_{ au(i)}$。它们的凸包是置换多边形：支持顶点在固定旋转下提供最优匹配，而最远顶点则给出全局对齐。我们证明了对于 $n ext{≥}2$ 的情况下，$n(n-1)$ 个顶点的尖锐界限，回答了 Rote 的旋转-赋值开放问题。在精确算术中，赋值查询在 $ ext{O}(n^5)$ 次操作中恢复多边形。基于赋值的界限通过分支限界方法将该方法扩展到三维旋转和在提供的平移下的部分匹配。在定时的 MPEG-7 形状对上，Rubix 平均在 12 毫秒内达到每个数值参考值，比在相同精度下的旋转网格快 50 倍。其距离改善了真实 3D 扫描的重力对齐匹配、形状检索和噪声晶体分类，相较于交替最小化。
cs.CV / 144 / 2610.10423
GraphRectify: Graph-Based Transfer of Adversarial Example Detectors Across Neural Networks
GraphRectify：基于图的对抗样本检测器在神经网络间的迁移
Abstract
Adversarial example detectors are often tied to the classifier backbone they were trained on, limiting reuse when the protected model is replaced or upgraded. Directly transferring such detectors across backbones is challenging because different networks generally produce incompatible internal representations. We propose GraphRectify, a graph-based framework for transferring adversarial image detectors across classifier backbones. GraphRectify learns a structured representation of intermediate classifier features and adapts representations from a new backbone to the detector learned on the original model, enabling detector reuse. We evaluate GraphRectify across multiple datasets, backbone architectures, and adversarial attacks, including detector-aware adaptive attacks that jointly target the classifier and detector. Across the complete evaluation matrix, GraphRectify achieves higher aggregate ROC-AUC than training a detector from scratch on the new backbone and the evaluated transfer ablations. The gains are particularly strong for transfers between different backbone families and when sufficient data are available. In contrast, training from scratch remains competitive in the most data-limited settings. These results show that adversarial detection knowledge can transfer effectively across heterogeneous classifier architectures rather than being relearned whenever the protected backbone changes.
Chinese Translation
对抗样本检测器通常与其训练所用的分类器骨干网络紧密相关，这限制了在更换或升级受保护模型时的重用。直接在不同骨干网络之间迁移这些检测器是具有挑战性的，因为不同的网络通常会产生不兼容的内部表示。我们提出了GraphRectify，一个基于图的框架，用于在分类器骨干网络之间迁移对抗图像检测器。GraphRectify学习中间分类器特征的结构化表示，并将新骨干网络的表示适配到在原始模型上学习的检测器，从而实现检测器的重用。我们在多个数据集、骨干架构和对抗攻击上评估了GraphRectify，包括联合针对分类器和检测器的检测器感知自适应攻击。在完整的评估矩阵中，GraphRectify的整体ROC-AUC高于在新骨干网络上从头训练检测器以及评估的迁移消融实验。特别是在不同骨干网络家族之间的迁移和数据充足时，性能提升尤为显著。相比之下，在数据有限的情况下，从头训练仍然具有竞争力。这些结果表明，对抗检测知识可以有效地在异构分类器架构之间迁移，而不是在每次更换受保护骨干时重新学习。
cs.CV / 145 / 2610.10429
SGF+: Decoupling Gradient Flows for Autoregressive Video Generation
SGF+: 解耦梯度流用于自回归视频生成
Abstract
Autoregressive video generation requires denoising the current frames while writing their key-value representations as context for future predictions. However, these two roles typically share parameters, and we find that their gradients exhibit distinct patterns and systematic negative alignment, hindering the joint optimization of visual quality and temporal consistency. We introduce Self Gradient Forcing Plus (SGF+), which assigns separate parameters to context writing and denoising while preserving their interaction through causal attention. Both roles are jointly optimized using the original generation objective without auxiliary losses, with context writing supervised through its contribution to future predictions. This simple change improves visual quality and long-horizon consistency over the evaluated baselines in both framewise and chunkwise generation, without additional video training data or a longer training horizon. Trained on only 5s rollouts, SGF+ supports continuous generation for up to 24 hours without long-video fine-tuning. These results highlight role-specific parameterization as an effective design principle for high-quality autoregressive video generation and native long-horizon extrapolation.
Chinese Translation
自回归视频生成需要在为未来预测编写关键值表示的同时去噪当前帧。然而，这两种角色通常共享参数，我们发现它们的梯度表现出不同的模式和系统性的负对齐，这妨碍了视觉质量和时间一致性的联合优化。我们提出了自梯度强制加（Self Gradient Forcing Plus, SGF+），为上下文编写和去噪分配了独立的参数，同时通过因果注意力保留它们之间的交互。这两种角色通过原始生成目标进行联合优化，无需辅助损失，上下文编写通过其对未来预测的贡献进行监督。这一简单的改变在逐帧和分块生成中均改善了视觉质量和长时间一致性，相较于评估基线，没有额外的视频训练数据或更长的训练周期。SGF+在仅训练5秒的生成过程中，支持连续生成长达24小时，而无需长视频微调。这些结果突显了角色特定参数化作为高质量自回归视频生成和原生长时间外推的有效设计原则。
cs.CV / 146 / 2610.10436
Detecting Adversarial Images through Response Profiles of Vision-Language Models
通过视觉-语言模型的响应特征检测对抗性图像
Abstract
Adversarial perturbations can alter the predictions of frozen vision-language models (VLMs) while leaving their confidence and image--text similarity patterns seemingly plausible. We investigate whether we can identify adversarial inputs based on the broader way an image interacts with a collection of general semantic prompts. Our detector summarizes these responses using category-level statistics, relationships among prompts, deviations from clean reference distributions, and stability under weak image transformations, producing a compact response profile that is classified by a lightweight model while the VLM remains fixed. We evaluate the approach on multiple public image datasets, several CLIP-style visual backbones, and a range of gradient-based, optimization-based, automated, and spatial attacks. The detector achieves strong discrimination in attack-specific settings and retains substantial performance when evaluated on attacks not seen during training. Under a controlled detector-specific protocol, the response-profile representation outperforms the evaluated embedding-geometry baselines. Additional analyses show that the feature groups provide complementary information and that the method remains effective under variations in the prompt configuration. We also examine inference cost and performance against detector-aware adaptive attacks. Overall, the results indicate that response patterns across semantic prompts provide a useful complementary signal for adversarial image detection in frozen VLMs.
Chinese Translation
对抗性扰动可以改变冻结的视觉-语言模型（VLMs）的预测，同时使其置信度和图像-文本相似性模式看似合理。我们研究是否可以基于图像与一组通用语义提示的更广泛交互方式来识别对抗性输入。我们的检测器通过类别级统计、提示之间的关系、与干净参考分布的偏差以及在弱图像变换下的稳定性来总结这些响应，生成一个紧凑的响应特征，由轻量级模型进行分类，而VLM保持固定。我们在多个公共图像数据集、几种CLIP风格的视觉骨干网络以及一系列基于梯度、基于优化、自动化和空间攻击的情况下评估该方法。该检测器在特定攻击设置中实现了强大的区分能力，并在评估未在训练期间见过的攻击时保持了相当的性能。在一个受控的检测器特定协议下，响应特征表示优于评估的嵌入几何基线。额外分析表明，特征组提供了互补信息，并且该方法在提示配置变化下仍然有效。我们还考察了推理成本和对抗检测器感知自适应攻击的性能。总体而言，结果表明，跨语义提示的响应模式为冻结的VLMs中的对抗性图像检测提供了有用的互补信号。
cs.CV / 147 / 2610.10438
ECHO: Embodied Camera Observations of Human Object Carrying
ECHO：人类物体携带的具身相机观察
Abstract
Embodied and assistive agents must do more than recognize objects: they must reason about where an object belongs given the layout of an environment and the habits of the people who live in it. Progress on this problem has been limited, in part because no dedicated benchmark or dataset exists to define and evaluate it. Existing RGB-D scan datasets reconstruct static rooms without human activity, while human-object-interaction datasets capture motion without a navigable, fully reconstructed scene or a ground-truth notion of an object's natural destination. We introduce contextual object placement as a benchmark task: predicting an object's destination during an observed object-carrying episode. To support this task, we present Embodied Camera observations of Human Object carrying (ECHO), a large-scale synthetic dataset that pairs dense RGB-D scans of indoor scenes with recordings of an embodied human carrying everyday objects to context-appropriate destinations. ECHO is the first publicly available dataset to combine reconstructed scenes, human activity, natural language, and contextual-placement annotations. It comprises 3,805 human-annotated episodes across 159 floors of 115 HM3D scenes, involving 198 distinct objects. Each floor includes a complete RGB-D scan with human-annotated room labels and a surface list. Each episode provides synchronized RGB-D encounter clips; 6-DoF camera, human, and object trajectories; start and destination surfaces; an action caption; and a human-written context: a single sentence describing the inhabitant's routine that implies the destination without naming it. We evaluate contextual object placement using input-masked probes and an end-to-end baseline. Results show that no single input modality is sufficient, highlighting the need to jointly reason over scene structure, human activity, and contextual knowledge.
Chinese Translation
具身和辅助代理不仅需要识别物体，还必须根据环境的布局和居住在其中的人的习惯推理物体的归属。对此问题的研究进展有限，部分原因在于缺乏专门的基准或数据集来定义和评估该问题。现有的RGB-D扫描数据集重建了没有人类活动的静态房间，而人类-物体交互数据集则捕捉了运动，但缺乏可导航的、完全重建的场景或物体自然目的地的真实概念。我们提出了上下文物体放置作为基准任务：在观察到的物体携带事件中预测物体的目的地。为支持这一任务，我们呈现了人类物体携带的具身相机观察（ECHO），这是一个大规模合成数据集，将室内场景的密集RGB-D扫描与具身人类携带日常物品到适当目的地的记录相结合。ECHO是第一个公开可用的数据集，结合了重建场景、人类活动、自然语言和上下文放置注释。该数据集包含3,805个经过人类注释的事件，涵盖115个HM3D场景的159个楼层，涉及198种不同的物体。每个楼层包括完整的RGB-D扫描，带有人类注释的房间标签和表面列表。每个事件提供同步的RGB-D接触片段；6自由度的相机、人类和物体轨迹；起始和目的地表面；一个动作说明；以及一个人类撰写的上下文：一句描述居民日常活动的句子，暗示目的地而不直接命名。我们使用输入掩码探针和端到端基线评估上下文物体放置。结果表明，单一输入模态不足，强调了共同推理场景结构、人类活动和上下文知识的必要性。
cs.CV / 148 / 2610.10457
MORCA: Offline-to-Online Reinforcement Learning for Adaptive Cache Reuse in Video Diffusion Acceleration
MORCA：用于视频扩散加速的自适应缓存重用的离线到在线强化学习
Abstract
Diffusion Transformers (DiTs) achieve remarkable performance in video synthesis, but their iterative denoising process suffers from high inference latency. To address this, caching has emerged as an effective acceleration strategy by capitalizing on inter-step redundancy during denoising. Existing dynamic caching methods typically estimate the error that cache reuse would introduce at each denoising step (step error) to guide cache decisions, whereas our concern is how much quality loss cache reuse would cause in the final generated video (terminal error). We show that step error does not directly correspond to terminal error and that latent information helps capture their relationship, thereby informing cache decisions. Moreover, existing threshold-based methods cannot provide precise speedup control, making it difficult to meet practical requirements for user-specified acceleration targets. To address these limitations, we introduce MORCA, a cache scheduling framework trained through offline-to-online reinforcement learning to make latent-aware reuse/recompute decisions under user-specified acceleration targets. Extensive experiments on different video generation models across multiple target acceleration ratios demonstrate that MORCA achieves better generation fidelity than state-of-the-art caching methods under comparable computational budgets. Code is available at https://github.com/x10ngyx/MORCA.
Chinese Translation
扩散变换器（Diffusion Transformers, DiTs）在视频合成中表现出色，但其迭代去噪过程面临高推理延迟的问题。为了解决这一问题，缓存作为一种有效的加速策略，通过利用去噪过程中的步骤冗余而应运而生。现有的动态缓存方法通常在每个去噪步骤中估计缓存重用所引入的误差（步骤误差），以指导缓存决策，而我们的关注点在于缓存重用对最终生成视频造成的质量损失（终端误差）。我们展示了步骤误差与终端误差并不直接对应，并且潜在信息有助于捕捉它们之间的关系，从而为缓存决策提供信息。此外，现有的基于阈值的方法无法提供精确的加速控制，使得满足用户指定的加速目标变得困难。为了解决这些限制，我们提出了MORCA，一个通过离线到在线强化学习训练的缓存调度框架，能够在用户指定的加速目标下做出潜在信息感知的重用/重新计算决策。在多个目标加速比下对不同视频生成模型的广泛实验表明，MORCA在可比计算预算下实现了比最先进的缓存方法更好的生成保真度。代码可在 https://github.com/x10ngyx/MORCA 获取。
cs.CV / 149 / 2610.10473
Label-free cell counting and viability prediction with brightfield imaging and deep learning
无标记细胞计数与活性预测：基于明场成像与深度学习
Abstract
Cell viability assessment is a core requirement in cell culture systems, with critical applications in biopharmaceutical manufacturing and drug development. Conventionally, it is measured by adding membrane-impermeable dyes to a sample (a process called staining), which allows compromised cell membranes to be distinguished from intact ones. However, staining has several limitations: (a) chemical agents can perturb normal cellular processes of the cells being measured, (b) it is often ambiguous to assign viability to individual cells whose membrane integrity is only partially compromised. (c) photobleaching can undermine measurement accuracy over time when using fluorescent stains, and (d) staining cannot be performed in situ or in real time. Here, we show that (1) stained cells captured under brightfield imaging contain sufficient information to distinguish live and dead cells, and (2) cells captured under unstained brightfield imaging exhibit similar image features to their stained counterparts, enabling models trained on stained cells to generalize to unstained ones. We then report the development and validation of ViabiLens, an AI-assisted software for label-free cell viability analysis. The ViabiLens combines a cell detection model for localizing individual cells with a convolutional neural network (CNN) classifier for live/dead prediction, paired with an interactive UMAP-based viewer for visualizing and exploring individual cells across the sample. Evaluated on Chinese Hamster Ovary (CHO) cells spanning a wide range of viability conditions, ViabiLens achieves a mean absolute error of 2.68\% on unstained samples against fluorescence-based reference measurements. We also release a benchmark dataset for label-free cell viability analysis to facilitate future research, available at https://amirrezavazifeh.github.io/ViabiLens-Project-Page/.
Chinese Translation
细胞活性评估是细胞培养系统中的核心需求，在生物制药制造和药物开发中具有重要应用。传统上，细胞活性通过向样本中添加膜不透性染料（称为染色过程）来测量，这使得受损的细胞膜与完整的细胞膜得以区分。然而，染色存在若干局限性：（a）化学试剂可能会干扰被测细胞的正常细胞过程，（b）对于膜完整性仅部分受损的单个细胞，活性判定往往存在模糊性，（c）使用荧光染料时，光漂白会随着时间的推移降低测量的准确性，以及（d）染色无法在原位或实时进行。在此，我们展示了（1）在明场成像下捕获的染色细胞包含足够的信息以区分活细胞和死细胞，以及（2）在未染色的明场成像下捕获的细胞展现出与其染色对应物相似的图像特征，使得在染色细胞上训练的模型能够推广到未染色细胞。随后，我们报告了ViabiLens的开发与验证，这是一款用于无标记细胞活性分析的AI辅助软件。ViabiLens结合了用于定位单个细胞的细胞检测模型与用于活/死预测的卷积神经网络（CNN）分类器，并配备了一个基于UMAP的交互式查看器，用于可视化和探索样本中的单个细胞。在对涵盖广泛活性条件的中国仓鼠卵巢（CHO）细胞进行评估时，ViabiLens在未染色样本上相较于基于荧光的参考测量实现了2.68%的平均绝对误差。我们还发布了一个无标记细胞活性分析的基准数据集，以促进未来的研究，数据集可在https://amirrezavazifeh.github.io/ViabiLens-Project-Page/获取。
cs.CV / 150 / 2610.10491
Insights from Autoresearch for Solar Panel Segmentation
来自自动研究的太阳能电池板分割洞察
Abstract
This paper investigates AutoResearch, a protocol in which a coding language model edits a training program under a one-hour GPU budget and retains a change only if validation IoU improves. The protocol is applied to photovoltaic panel segmentation on a frozen real-image split, with DeepLabV3--ResNet-50 held fixed. Three campaigns of 24 experiments, using Gemma~4 12B, Qwen3-8B all improve their one-hour baselines, but retained modifications do not transfer across hardware. The Qwen3-8B configuration, trained on real images only, reaches a test IoU of 0.836 versus 0.833 for the reference GAN-augmented schedule. Research repository https://github.com/VU-AIML/automl4eo-autoresearch-segmentation.
Chinese Translation
本文研究了AutoResearch，这是一种协议，其中编码语言模型在一小时的GPU预算内编辑训练程序，并仅在验证IoU改善时保留更改。该协议应用于冻结真实图像分割的光伏面板分割，DeepLabV3--ResNet-50保持不变。通过Gemma~4 12B和Qwen3-8B进行的三次实验活动，共24个实验均改善了它们的一小时基线，但保留的修改无法在不同硬件之间转移。仅在真实图像上训练的Qwen3-8B配置，其测试IoU达到0.836，而参考的GAN增强计划为0.833。研究库链接：https://github.com/VU-AIML/automl4eo-autoresearch-segmentation。
cs.CV / 151 / 2610.10497
QuadTok: Quadtree Visual Tokenizer for Autoregressive Image Generation
QuadTok：用于自回归图像生成的四叉树视觉标记器
Abstract
We introduce QuadTok, a novel framework for visual tokenization and autoregressive image generation. Compared to traditional approaches using 2D grids or 1D token sequences, we propose a hierarchical quadtree structure, bridging the gap between 2D spatial binding and 1D sequence-level flexibility. The QuadTok tokenizer dynamically allocates representational capacity to visually intricate areas while leaving homogeneous regions at a coarse resolution. Compared with a fixed 256-token grid, our ImageNet-trained tokenizer saves approximately 10% of tokens on ImageNet and 9% when transferred zero-shot to the COCO dataset, while maintaining comparable reconstruction fidelity. Furthermore, the natural causality introduced by the tree structure seamlessly enables autoregressive image generation. Conditioned on a quadtree topology supplied before generation, our 947M GPT-style generative model achieves a 2.08 gFID on the ImageNet $256 \times 256$ benchmark. Additionally, leveraging the strong spatial correlation preserved by the quadtree structure, the QuadTok generator enables zero-shot spatially controlled image generation capabilities. Code: https://github.com/myc634/QuadTok.
Chinese Translation
我们介绍了QuadTok，一个用于视觉标记和自回归图像生成的新框架。与传统的使用二维网格或一维标记序列的方法相比，我们提出了一种分层的四叉树结构，弥合了二维空间绑定与一维序列灵活性之间的差距。QuadTok标记器动态分配表示能力给视觉复杂的区域，同时将同质区域保持在粗略分辨率。与固定的256标记网格相比，我们在ImageNet上训练的标记器在ImageNet上节省了大约10%的标记，在零-shot转移到COCO数据集时节省了9%，同时保持了可比的重建保真度。此外，树结构引入的自然因果关系无缝地实现了自回归图像生成。在生成之前提供的四叉树拓扑条件下，我们的947M GPT风格生成模型在ImageNet $256 imes 256$基准上达到了2.08 gFID。此外，利用四叉树结构所保留的强空间相关性，QuadTok生成器实现了零-shot空间控制的图像生成能力。代码：https://github.com/myc634/QuadTok。
cs.CV / 152 / 2610.10512
Video-Conditioned Generative Joint 2D-3D Hand Motion Recovery
视频条件生成的联合二维-三维手部运动恢复
Abstract
Recovering faithful 3D hand motion from video remains challenging due to frequent occlusions and incomplete visual observations, which make frame-wise pose estimates unreliable and temporally inconsistent. To address this problem, we propose JoHan, a unified generative framework that recovers hand motion directly from video sequences without relying on intermediate per-frame pose predictions. Trained from scratch, our model jointly generates aligned 2D and 3D local hand pose sequences by learning their temporal dynamics and cross-representation correspondence. The generated 2D trajectories exploit direct spatial and temporal cues from the 2D images to guide the following generative 3D motion reconstruction, while the learned motion prior promotes temporal consistency. Their learned 2D-3D correspondence further enables recovery of the hand's global position and orientation relative to the camera. Extensive experiments on challenging benchmarks demonstrate significantly improved accuracy and speed in local hand-pose and camera-space reconstruction. Notably, our method captures much better hand-motion dynamics, producing significantly smoother motion than previous methods while maintaining high per-frame pose accuracy.
Chinese Translation
从视频中恢复真实的三维手部运动仍然具有挑战性，因为频繁的遮挡和不完整的视觉观测使得逐帧的姿态估计不可靠且时间上不一致。为了解决这个问题，我们提出了JoHan，一个统一的生成框架，能够直接从视频序列中恢复手部运动，而无需依赖中间的逐帧姿态预测。我们的模型从头开始训练，联合生成对齐的二维和三维局部手部姿态序列，通过学习它们的时间动态和跨表示对应关系。生成的二维轨迹利用来自二维图像的直接空间和时间线索来指导后续的三维运动重建，而学习到的运动先验则促进了时间一致性。它们学习到的二维-三维对应关系进一步使得手部的全局位置和相对于相机的方向得以恢复。在具有挑战性的基准测试中，广泛的实验表明，在局部手部姿态和相机空间重建方面，准确性和速度显著提高。值得注意的是，我们的方法捕捉到了更好的手部运动动态，生成的运动比以前的方法显著更平滑，同时保持了高的逐帧姿态准确性。
cs.CV / 153 / 2610.10524
GRACE: Generation-aware latent compression for efficient video generation
GRACE：面向生成的潜在压缩以实现高效视频生成
Abstract
Highly compressed video autoencoders offer an effective way to accelerate video diffusion models, as the Diffusion Transformer (DiT) operates on far fewer tokens. However, such autoencoders are challenging to train, since a higher compression ratio degrades reconstruction quality and recovering it requires more channels, which is known to slow the convergence of the DiT. The compressed latent also differs from the one the DiT was trained on, so the pretrained DiT must be either retrained from scratch or adapted at considerable cost. Compressing the autoencoder the DiT was trained with appears to preserve compatibility, yet optimizing it for reconstruction alone still shifts the latent away from the distribution the DiT has learned. To address this, we propose Generation-Aware Latent Compression for Efficient Video Generation (GRACE), a two-stage framework that compresses a pretrained video autoencoder while keeping it compatible with the pretrained DiT. Specifically, we keep a frozen base latent from the pretrained encoder and learn a residual latent for the information lost under stronger compression, while aligning the compressed latent with the pretrained latent in the feature space of the frozen DiT so that the autoencoder is optimized for generation. We then adapt the DiT with lightweight fine-tuning and asymmetric denoising, where the base is denoised ahead of the residual. GRACE reduces the token count of Wan2.1-I2V-14B by 8x and its latency by 11.1x at 480x832x81, while matching the generation quality of the pretrained pipeline before compression on VBench.
Chinese Translation
高度压缩的视频自编码器为加速视频扩散模型提供了一种有效的方法，因为扩散变换器（Diffusion Transformer, DiT）在更少的标记上进行操作。然而，这种自编码器的训练具有挑战性，因为更高的压缩比会降低重建质量，而恢复重建质量需要更多的通道，这已知会减缓DiT的收敛速度。压缩的潜在表示也与DiT训练时的潜在表示不同，因此预训练的DiT必须从头开始重新训练或以相当大的成本进行适应。压缩与DiT训练时使用的自编码器似乎可以保持兼容性，但仅仅为了重建而优化仍然会使潜在表示偏离DiT所学习的分布。为了解决这个问题，我们提出了面向生成的潜在压缩以实现高效视频生成（GRACE），这是一个两阶段框架，旨在压缩预训练的视频自编码器，同时保持其与预训练DiT的兼容性。具体而言，我们保留了来自预训练编码器的冻结基础潜在表示，并学习了在更强压缩下丢失信息的残差潜在表示，同时在冻结的DiT的特征空间中将压缩的潜在表示与预训练的潜在表示对齐，以便自编码器能够针对生成进行优化。然后，我们通过轻量级微调和不对称去噪对DiT进行适应，其中基础潜在表示在残差之前进行去噪。在480x832x81的条件下，GRACE将Wan2.1-I2V-14B的标记数量减少了8倍，延迟减少了11.1倍，同时在压缩前与预训练管道的生成质量相匹配。
cs.CV / 154 / 2610.10538
Never Look Back: Understanding Persistence in 3D Object Memory from Egocentric Videos
不再回头：从自我中心视频理解三维物体记忆的持久性
Abstract
As we move through the world and carry out everyday tasks, we encounter objects that may become relevant only later. We are capable of recalling where we left something or what was inside a container, even without knowing we would need it later. Here, we study how an embodied assistant can build a similar memory from egocentric videos, by observing a person's day-to-day activities. We present Ledger, a persistent 3D object memory that combines object locations, their histories, and contextual descriptions. It associates observations across the recording and retains objects after they leave the view, including those the person never touches. It clusters each object's observations by resting locations and records a move only after repeated evidence, reducing the effect of localization noise. Short descriptions preserve details such as an object's contents or supporting surface. It saves these records to later answer spatial questions without having to access the original images or video. Our memory raises HD-EPIC accuracy from 29.7% to 42.6%, UCS-Bench accuracy from 33.8% to 38.5% and localizes Ego4D objects with a 0.99 m median error on returned predictions. Our analyses identify complementary roles for temporal persistence, contextual descriptions, and retrieval. Our study on 100 stitched streams of multiple scenes each further exposes failures in both retrieval and construction. Per-scene construction partially recovers the performance lost across scene changes compared to that of single scene streams.
Chinese Translation
在我们穿越世界并执行日常任务时，会遇到一些物体，这些物体可能在稍后变得相关。我们能够回忆起我们放置某物的位置或容器内的内容，即使在我们不知道稍后会需要它的情况下。在这里，我们研究了一个具身助手如何通过观察一个人的日常活动，从自我中心视频中构建类似的记忆。我们提出了Ledger，一个持久的三维物体记忆，它结合了物体的位置、历史和上下文描述。它关联录制过程中的观察，并在物体离开视野后仍然保留这些物体，包括那些人从未触碰过的物体。它通过静止位置对每个物体的观察进行聚类，并仅在重复证据后记录移动，从而减少定位噪声的影响。简短的描述保留了物体内容或支撑表面等细节。它保存这些记录，以便在不访问原始图像或视频的情况下回答空间问题。我们的记忆将HD-EPIC的准确率从29.7%提高到42.6%，将UCS-Bench的准确率从33.8%提高到38.5%，并在返回预测时以0.99米的中位误差定位Ego4D物体。我们的分析识别了时间持久性、上下文描述和检索的互补角色。我们对100个拼接的多个场景流的研究进一步揭示了在检索和构建中的失败。与单场景流相比，每场景构建在场景变化中部分恢复了性能损失。
cs.CV / 155 / 2610.10539
Tetris3D: 3D Scene Generation With Objects That Fit Together
Tetris3D：具有相互适配物体的三维场景生成
Abstract
We propose Tetris3D, a generative framework for single-image 3D scene reconstruction that recovers objects which are physically and geometrically coherent as a scene. Existing methods often generate objects independently or couple them implicitly, providing limited guidance for ensuring fine-grained spatial compatibility between neighboring objects that interact with one another. To address this, we explicitly condition the generation of each object on the geometry of surrounding objects and their physical relationships, guiding its shape and pose to remain geometrically and physically plausible within the scene. Moreover, we introduce ComOb, a physics simulation-based dataset of 1.2M scenes featuring physical interactions across diverse object categories, with per-object meshes and pairwise physical relation annotations. Comprehensive experiments on synthetic and realworld scenes show that Tetris3D recovers coherent object shapes and poses even when interacting regions are occluded, and achieves state-of-the-art performance in both generation quality and physical stability.
Chinese Translation
我们提出了Tetris3D，这是一个用于单幅图像三维场景重建的生成框架，能够恢复在物理和几何上相互一致的场景物体。现有方法通常独立生成物体或隐式耦合它们，提供的指导有限，无法确保相互作用的相邻物体之间的细粒度空间兼容性。为了解决这个问题，我们明确地将每个物体的生成条件化于周围物体的几何形状及其物理关系，指导其形状和姿态在场景中保持几何和物理上的合理性。此外，我们引入了ComOb，这是一个基于物理模拟的1.2M场景数据集，涵盖了不同物体类别之间的物理交互，包含每个物体的网格和成对的物理关系注释。在合成和真实场景上的全面实验表明，Tetris3D能够恢复一致的物体形状和姿态，即使在交互区域被遮挡的情况下，并在生成质量和物理稳定性方面达到了最先进的性能。
cs.AI / 1 / 2610.08793
Adaptive Workflow Intelligence: A Cognitive Architecture for Context-Driven Enterprise Automation
自适应工作流智能：一种基于认知架构的上下文驱动企业自动化
Abstract
Enterprise systems increasingly rely on automated workflows, yet many AI-driven solutions remain brittle under non-stationary conditions, evolving policies, and delayed operational feedback. While reinforcement learning and large language model (LLM) agents offer partial adaptability, they do not by themselves provide persistent reflection mechanisms or straightforward integration with policy-constrained enterprise operations. This paper introduces Adaptive Workflow Intelligence (AWI), a cognitive architecture for context-driven enterprise agents organized around a four-layer Perception-Cognition-Action-Reflection (PCAR) loop. AWI treats reflection as a mechanism for continuous policy refinement and combines hybrid reasoning with reflective memory and feedback-driven adaptation to support decision making under environmental drift and operational constraints. We evaluate AWI in a simulated enterprise decision workflow characterized by delayed outcomes and a controlled regime shift. In a drift-and-delay stress test, guardrail-constrained adaptive approaches recover more rapidly than static automation while maintaining policy compliance. Within this setting, AWI's reflective components modestly reduce behavioral oscillation and feedback variance, illustrating the stability-agility trade-off introduced by reflective policy adaptation.
Chinese Translation
企业系统越来越依赖自动化工作流，但许多基于人工智能的解决方案在非平稳条件、不断演变的政策和延迟的操作反馈下仍然脆弱。尽管强化学习和大型语言模型（LLM）代理提供了部分适应性，但它们本身并未提供持久的反思机制或与政策约束的企业操作的直接集成。本文介绍了自适应工作流智能（AWI），这是一种围绕四层感知-认知-行动-反思（PCAR）循环组织的上下文驱动企业代理的认知架构。AWI将反思视为持续政策细化的机制，并结合混合推理、反思记忆和基于反馈的适应，以支持在环境漂移和操作约束下的决策制定。我们在一个模拟的企业决策工作流中评估AWI，该工作流的特点是结果延迟和受控的制度转变。在一个漂移与延迟的压力测试中，受限于保护措施的自适应方法比静态自动化恢复得更快，同时保持政策合规。在这种环境下，AWI的反思组件适度减少了行为波动和反馈方差，展示了反思政策适应所引入的稳定性与灵活性之间的权衡。
cs.AI / 2 / 2610.08808
Accelerating Floating-Point Satisfiability Solving via Gradient Normalization
通过梯度归一化加速浮点可满足性求解
Abstract
Satisfiability Modulo Theories (SMT) solvers are foundational to software verification, program analysis, and compiler testing, particularly over the theory of Quantifier-Free Floating-Point (QF_FP). While recent optimization-based SMT solvers have successfully applied gradient descent to continuous relaxations of logical formulas, they are fundamentally bottlenecked by gradient domination, a phenomenon where a small subset of difficult clauses hijacks the optimization trajectory, preventing the solver from satisfying the broader formula and trapping it in local minima. To overcome this, we present GradSAT, a novel framework that bridges optimization-based SMT solving with Multi-Task Learning (MTL). GradSAT reformulates the constraint satisfaction process by treating each SMT clause as an independent MTL task. By applying dynamic gradient normalization (GradNorm), GradSAT actively balances the gradient magnitudes across all clauses at runtime, systematically penalizing dominant gradients and accelerating lagging clauses to ensure uniform convergence. GradSAT implements this through a highly optimized, two-stage hybrid pipeline. First, a GPU-accelerated PyTorch backend leveraging symbolic compilation and operator fusion navigates the continuous relaxation to a high-quality basin. Second, the candidate assignment is handed off to a bit-precise local search engine to rapidly resolve the exact, rigorous assignment. By stabilizing the continuous search dynamics, GradSAT mitigates the brittleness of prior gradient-based solvers and provides a robust, highly parallelizable architecture for complex constraint solving.
Chinese Translation
可满足性模理论（SMT）求解器是软件验证、程序分析和编译器测试的基础，特别是在无量词浮点数（QF_FP）理论上。尽管近期基于优化的SMT求解器成功地将梯度下降应用于逻辑公式的连续松弛，但它们在根本上受到梯度主导现象的瓶颈影响，这种现象是指一小部分困难子句劫持了优化轨迹，阻止求解器满足更广泛的公式并将其困在局部最小值中。为了解决这个问题，我们提出了GradSAT，一个将基于优化的SMT求解与多任务学习（MTL）相结合的新框架。GradSAT通过将每个SMT子句视为一个独立的MTL任务，重新构造了约束满足过程。通过应用动态梯度归一化（GradNorm），GradSAT在运行时主动平衡所有子句的梯度大小，系统性地惩罚主导梯度并加速滞后子句，以确保均匀收敛。GradSAT通过一个高度优化的两阶段混合管道实现这一点。首先，利用符号编译和操作符融合的GPU加速PyTorch后端引导连续松弛到高质量的盆地。其次，候选赋值被移交给一个位精确的局部搜索引擎，以快速解决精确、严格的赋值。通过稳定连续搜索动态，GradSAT减轻了先前基于梯度的求解器的脆弱性，并为复杂约束求解提供了一个稳健且高度可并行化的架构。
cs.AI / 3 / 2610.08814
Route-Verify-Vote: Procedure-Conditioned Self-Consistency for Mixed-Domain Reasoning
路线-验证-投票：基于程序条件的自一致性用于混合领域推理
Abstract
Compositional generalization remains challenging when language models must combine familiar reasoning operations in unfamiliar ways. The Scenario-Based Commonsense Reasoning Evaluation (SCoRE) 2026 tests this ability on three mixed domains absent from training and requires models to identify the complete set of correct options for each question. We introduce Route-Verify-Vote (RVV), a framework for procedure-conditioned self-consistency that uses language models without parameter updates. Route uses the provided domain label to select a reasoning procedure that guides the model in representing and applying the relevant constraints. Verify prompts the model to assess each option against those constraints. Vote aggregates complete answer sets and allocates additional samples to questions with a small vote-count margin between the two most frequent sets. Samples for each question follow the same domain-specific procedure. On the official test set, voting over 16 sampled answer sets per question achieves an exact-set accuracy of 74.6%. Adaptive RVV reaches 77.3%, and combining models on selected domain routes raises accuracy to 79.4%. The final system ranked second among participating systems. These results support domain-specific reasoning procedures and answer-set disagreement as useful tools for allocating inference-time computation in mixed-domain reasoning.
Chinese Translation
当语言模型必须以不熟悉的方式组合熟悉的推理操作时，组合泛化仍然具有挑战性。2026年场景基础常识推理评估（SCoRE）测试了这种能力，涵盖了训练中缺失的三个混合领域，并要求模型识别每个问题的完整正确选项集。我们提出了路线-验证-投票（Route-Verify-Vote, RVV），这是一个基于程序条件的自一致性框架，使用语言模型而不进行参数更新。路线（Route）利用提供的领域标签选择推理程序，引导模型表示和应用相关约束。验证（Verify）促使模型根据这些约束评估每个选项。投票（Vote）汇总完整的答案集，并将额外样本分配给在两个最频繁的答案集之间投票数差距较小的问题。每个问题的样本遵循相同的领域特定程序。在官方测试集中，每个问题对16个采样答案集进行投票，达到了74.6%的精确集准确率。自适应RVV达到了77.3%，而在选定领域路线上的模型组合将准确率提高到79.4%。最终系统在参与系统中排名第二。这些结果支持领域特定的推理程序和答案集不一致作为在混合领域推理中分配推理时间计算的有用工具。
cs.AI / 4 / 2610.08875
An Empirical Study of Agent Skills' Downstream Utility
代理技能下游效用的实证研究
Abstract
Agent Skills package procedural guidance and resources for reuse, but a relevant Skill does not necessarily improve task performance. Existing studies characterize Skill content and evaluate downstream performance, yet provide limited explanations of how utility depends on content, execution configuration, and multi-Skill organization. We conduct an empirical study on 87 SkillsBench tasks, defining downstream utility as the pass-rate difference from No-Skill on the same tasks under the same model--harness configuration. We compare the same Skills across nine configurations, then examine alternative published Skills and organizations of fixed Skill sets under three selected configurations. We retrieve marketplace candidates from a curated corpus of 37,596 Skills. LLM-assisted analysis of content, execution traces, and final artifacts, followed by author review, relates provided support to actual use and task outcomes. The same Skills help some configurations and hurt others on 36.78\% of tasks, with trajectories showing that recommended procedures can become an execution burden. Relevance rankings overlook more useful candidates. Within the evaluated candidate sets, reranking by support for required operations raises first-choice pass rates by 4.35--5.80 percentage points across the three configurations. We derive 17 authoring practices linking executable procedures to recovery, preservation of task requirements, and checks on final artifacts. Stage Plan and Dependency DAG outperform use order alone, with DAG's additional benefits concentrated in tasks supplied with five or six Skills. These findings guide developers to assess usable operation support, allow procedure adaptation while preserving task requirements, and make artifact dependencies explicit when organizing Skills.
Chinese Translation
代理技能提供了程序指导和可重用资源，但相关技能并不一定能提高任务绩效。现有研究对技能内容进行了特征描述并评估了下游绩效，但对效用如何依赖于内容、执行配置和多技能组织的解释有限。我们对87个SkillsBench任务进行了实证研究，将下游效用定义为在相同模型-工具配置下与无技能相比的通过率差异。我们比较了在九种配置下相同的技能，然后在三种选定配置下检查了替代的已发布技能和固定技能集的组织。我们从一个经过整理的37,596个技能的语料库中检索市场候选项。通过大语言模型（LLM）辅助的内容分析、执行轨迹和最终工件的审查，将提供的支持与实际使用和任务结果相关联。相同的技能在36.78%的任务中对某些配置有帮助，而对其他配置则有害，轨迹显示推荐程序可能成为执行负担。相关性排名忽视了更有用的候选项。在评估的候选集内，通过对所需操作支持的重新排名，在三种配置下提升了首选通过率4.35至5.80个百分点。我们推导出17种创作实践，将可执行程序与任务要求的恢复、保持和最终工件的检查联系起来。阶段计划和依赖有向无环图（Dependency DAG）在单独使用顺序时表现更佳，DAG的额外好处集中在提供五个或六个技能的任务中。这些发现指导开发者评估可用的操作支持，允许在保持任务要求的同时进行程序适应，并在组织技能时明确工件依赖关系。
cs.AI / 5 / 2610.08878
How Could AI Eliminate Humanity? A Failure-Mode Analysis of Civilizational Risk
人工智能如何可能消灭人类？对文明风险的失效模式分析
Abstract
This article develops a failure-mode framework for analyzing how advanced artificial intelligence could contribute to human extinction, irreversible civilizational collapse, or permanent human disempowerment. The central thesis is that catastrophic AI risk does not require consciousness, hostility, or an explicit intention to harm humanity. Instead, risk may arise through several distinct but interacting pathways, including autonomous misalignment, harmful human use, organizational failure, and competitive deployment. The severity of these pathways depends on factors such as capability, autonomy, external access, persistence, institutional safeguards, and the preservation of recovery capacity. The analysis is deliberately non-operational: it identifies causal conditions, empirically tractable intermediate quantities, and defensive research questions rather than procedures for causing harm.
Chinese Translation
本文建立了一个失效模式框架，以分析先进人工智能如何可能导致人类灭绝、不可逆的文明崩溃或人类永久失去权力。中心论点是，灾难性的人工智能风险并不需要意识、敌意或明确的伤害人类的意图。相反，风险可能通过几条不同但相互作用的路径产生，包括自主失调、有害的人类使用、组织失败和竞争性部署。这些路径的严重性取决于能力、自主性、外部访问、持续性、制度保障和恢复能力的保持等因素。该分析故意不具操作性：它识别因果条件、经验可处理的中间量和防御性研究问题，而不是造成伤害的程序。
cs.AI / 6 / 2610.08900
Humanize: Judgement Engineering for Agentic Coding
人性化：面向代理编码的判断工程
Abstract
Agentic coding makes code generation cheap, but reliable completion remains difficult: the agent that writes the code is a weak judge of whether it is done. We present Humanize, a multi-agent orchestration workflow for agentic coding built around judgement engineering: explicit, mechanically enforced decisions at the boundaries between planning, implementation, review, and learning. A human approves a plan contract, a builder agent implements it in rounds, and a reviewer agent from another vendor decides completion; deterministic hooks, not a model, route work between these roles and enforce 72 mechanical gates. Viewed as a Markov chain over repository states, alternating builder and reviewer samples jointly from two models, so a defect survives only if both miss it. We study Humanize through its deployment, 118 public postmortems of real loops, and its applications. Over 68 versions in 108 days, it gathered 1,468 GitHub stars. Applications include a 567-file gem5 build-system migration under upstream review; Kernel Design Agents, which extend the loop with a kernel knowledge base and profiling feedback and placed in the top three of all three Full-Agent tracks of the MLSys 2026 FlashInfer contest; and, through Humanize Olympiad Agents (HOA), full scores in IOI 2026, IMO 2026, IPhO 2026, and IBO 2024, 418.5/437 in IChO 2026 (gold-medal). Humanize also achieves 672/672 on PutnamBench and ranks first (251/303) on Lean-Eval's leaderboard even competiting with professional mathematicians. The postmortems show that independent review catches unsupported builder claims, but stopping remains a key weakness. In reports that separate rounds by phase, two thirds of rounds occurred after implementation was accepted. This evidence is observational, not a controlled comparison of workflows.
Chinese Translation
代理编码使得代码生成变得廉价，但可靠的完成仍然困难：编写代码的代理对其是否完成的判断能力较弱。我们提出了Humanize，一个围绕判断工程构建的多代理协调工作流程：在规划、实施、审查和学习的边界上进行明确的、机械执行的决策。一个人类批准计划合同，一个构建代理分阶段实施该计划，来自另一供应商的审查代理决定完成情况；确定性钩子，而非模型，在这些角色之间路由工作并强制执行72个机械门。将其视为一个关于代码库状态的马尔可夫链，交替的构建者和审查者共同从两个模型中抽样，因此只有当两者都未发现缺陷时，缺陷才会存活。我们通过其部署、118个真实循环的公开事后分析及其应用来研究Humanize。在108天内的68个版本中，它获得了1,468个GitHub星标。应用包括在上游审查下的567个文件gem5构建系统迁移；内核设计代理（Kernel Design Agents），它通过内核知识库和性能反馈扩展循环，并在MLSys 2026 FlashInfer比赛的所有三个全代理赛道中名列前茅；以及通过Humanize奥林匹克代理（HOA），在IOI 2026、IMO 2026、IPhO 2026和IBO 2024中获得满分，在IChO 2026中获得418.5/437（金牌）。Humanize在PutnamBench上也取得了672/672的成绩，并在Lean-Eval的排行榜上排名第一（251/303），甚至与职业数学家竞争。事后分析显示，独立审查能够捕捉到不支持的构建者主张，但停止仍然是一个关键弱点。在按阶段分隔的报告中，三分之二的阶段发生在实施被接受之后。这些证据是观察性的，而不是工作流程的控制比较。
cs.AI / 7 / 2610.08901
Sequential Probabilistic Uncertainty Estimation for Parallel Multi-Agent Reasoning Systems
并行多智能体推理系统的序列概率不确定性估计
Abstract
LLM-based multi-agent systems (MAS) have attracted growing attention for improving reasoning through interaction among multiple agents. In this work, we focus on parallel multi-agent reasoning systems, where several agents solve the same problem over multiple rounds and aggregate their outputs into a final answer. Despite their strong reasoning performance, uncertainty estimation for such systems remains underexplored: the reliability of a MAS depends not only on individual generations, but also on how agents interact and evolve across rounds. We propose SAUCE (Sequential Agent Uncertainty through Consensus Evolution), a lightweight, training-free uncertainty estimator that formulates MAS uncertainty as sequential inference over a latent system-level belief. SAUCE aggregates round-level agreement and generation-uncertainty signals through a filtering-style update. Across five backbones, five benchmarks, and two MAS protocols, SAUCE improves misclassification detection, selective prediction, and calibration over a broad set of uncertainty estimation baselines, including standard log-likelihood-based methods and MAS-specific estimators.
Chinese Translation
基于大型语言模型（LLM）的多智能体系统（MAS）因通过多个智能体之间的互动来提升推理能力而受到越来越多的关注。在本研究中，我们专注于并行多智能体推理系统，其中多个智能体在多个回合中解决相同的问题，并将其输出汇聚成最终答案。尽管这些系统在推理性能上表现强劲，但对其不确定性估计的研究仍然不足：MAS的可靠性不仅依赖于个体生成的结果，还取决于智能体在回合间的互动和演化。我们提出了SAUCE（通过共识演化的序列智能体不确定性），这是一种轻量级、无训练的不确定性估计器，它将MAS的不确定性形式化为对潜在系统级信念的序列推理。SAUCE通过过滤式更新聚合回合级别的共识和生成不确定性信号。在五个基础模型、五个基准测试和两个MAS协议的实验中，SAUCE在误分类检测、选择性预测和校准方面超越了一系列不确定性估计基线，包括标准的对数似然方法和MAS特定的估计器。
cs.AI / 8 / 2610.08902
Agent Plasticity: Measuring Self-Improvement Through Experience
智能体的可塑性：通过经验测量自我提升
Abstract
AI agents increasingly operate in environments where they can diagnose failures and improve through experience, yet existing evaluations largely measure what an agent can do at a fixed point in time rather than how effectively it learns. Evaluating self-improvement requires answering three questions: does future performance improve and generalize beyond the interactions that enabled learning; how efficiently are new capabilities acquired; and where does the self-improvement process break down? To answer these questions, we study self-improvement in a controlled setting where agents amortize past experience into reusable artifacts that are inherited by future instances. At each checkpoint, we measure performance on training and held-out environment interactions while accounting for learning cost. We introduce agent plasticity, the efficiency with which an agent converts experience into gains in future held-out performance. Across multiple environments, frontier models exhibit sharply different improvement trajectories despite comparable opportunities to learn. Some achieve substantial and persistent gains, while others remain near or below their initial performance, and gains within the training regime often transfer only partially to out-of-distribution conditions. Endpoint capability and acquisition efficiency also diverge: the agent that ultimately performs best need not be the one that improves most efficiently. Tracing failures through the improvement loop further reveals different candidate bottlenecks. Agents with low plasticity often fail to reuse relevant artifacts, whereas more plastic agents may still fail despite reusing relevant artifacts, pointing to limitations in artifact quality, generalization, or application. Evaluating self-improving agents requires measuring not only what they can do, but how effectively they become better through experience.
Chinese Translation
人工智能智能体越来越多地在能够诊断故障并通过经验改进的环境中操作，但现有评估主要测量智能体在固定时间点的能力，而不是其学习的有效性。评估自我提升需要回答三个问题：未来的表现是否在促进学习的交互之外得到改善和推广；新能力的获取效率如何；自我提升过程在哪里出现瓶颈？为了解答这些问题，我们在一个受控环境中研究自我提升，在该环境中，智能体将过去的经验转化为可重用的工件，这些工件被未来的实例继承。在每个检查点，我们测量训练和保留环境交互的表现，同时考虑学习成本。我们引入了智能体可塑性（agent plasticity），即智能体将经验转化为未来保留表现提升的效率。在多个环境中，前沿模型表现出截然不同的改进轨迹，尽管它们有相似的学习机会。有些模型实现了显著且持久的提升，而其他模型则保持在其初始表现附近或以下，并且在训练阶段的提升往往仅部分转移到分布外条件。最终能力和获取效率也存在差异：表现最佳的智能体不一定是提升效率最高的那个。通过改进循环追踪失败进一步揭示了不同的潜在瓶颈。具有低可塑性的智能体往往无法重用相关工件，而更具可塑性的智能体即使重用了相关工件仍可能失败，这指向了工件质量、泛化或应用的局限性。评估自我提升的智能体不仅需要测量它们能做什么，还需要评估它们通过经验变得更好的有效性。
cs.AI / 9 / 2610.08923
AdaGuard: Enhancing Safety and Policy Compliance with Reasoning-Enabled LLM-As-A-Judge Guardrails
AdaGuard：通过推理增强的 LLM 作为法官的护栏提升安全性和政策合规性
Abstract
Enterprise generative AI applications require robust safety mechanisms that can accommodate diverse risk postures, evolving policies, and varying latency constraints. Current guardrail solutions often suffer from rigidity, relying on fixed policy sets and offering limited transparency or reasoning flexibility. We present Adaguard, an adaptive LLM-as-a-Judge framework designed to address these challenges through dynamic policy enforcement and adaptive reasoning-budget allocation. Built using supervised fine-tuning (SFT) and reinforcement learning (GRPO), AdaGuard generalizes to user-defined safety and compliance policies at runtime without requiring frequent model updates. A core innovation of our approach is the ability to dynamically infer the complexity of input-policy pairs, allowing the model to switch between high-speed black-box inference and explainable, reasoning-enabled moderation. This flexibility enables developers to balance stringent latency requirements with the need for actionable transparency. This adaptive capability allows AdaGuard to rival other guardrail and frontier models several times its size, while its auto-reasoning mode recovers the accuracy of always-on reasoning at a fraction of the latency
Chinese Translation
企业生成式人工智能应用需要强大的安全机制，以适应多样化的风险态势、不断演变的政策和不同的延迟约束。目前的护栏解决方案往往存在僵化的问题，依赖于固定的政策集，且提供的透明度或推理灵活性有限。我们提出了 AdaGuard，这是一种自适应的 LLM 作为法官框架，旨在通过动态政策执行和自适应推理预算分配来应对这些挑战。AdaGuard 基于监督微调（SFT）和强化学习（GRPO）构建，能够在运行时对用户定义的安全和合规政策进行泛化，而无需频繁更新模型。我们方法的核心创新在于能够动态推断输入-政策对的复杂性，使模型能够在高速黑箱推理和可解释的、具备推理能力的审查之间切换。这种灵活性使开发者能够在严格的延迟要求与可操作透明度之间取得平衡。这种自适应能力使 AdaGuard 能够与其他护栏和前沿模型相媲美，尽管其规模仅为后者的几倍，而其自动推理模式在极低的延迟下恢复了始终在线推理的准确性。
cs.AI / 10 / 2610.08927
Can AI Agents Make Open-Ended Scientific Discovery? Evidence from Station
人工智能代理能否进行开放式科学发现？来自Station的证据
Abstract
Recent AI systems have made rapid progress in scientific discovery when given well-defined metrics, but whether they can autonomously undertake open-ended scientific discovery remains unclear. We investigate AI's ability to tackle open-ended tasks in Station, an open-world environment in which multiple agents simulate a scientific ecosystem. To tackle challenges specific to open-ended tasks, we propose augmenting Station with two mechanisms: a Supervisor mechanism and periodic Meta Reflection, which encourage persistent exploration even when intermediate metrics are lacking. We construct open-ended tasks from three recent oral papers presented at ICLR. We give agents the main research question studied in each paper while withholding the paper's results and disabling web access. We then measure how many of the original findings-partitioned into individual criteria-agents rediscover. We find that Station rediscovers 62.7% of the criteria on average, compared with 15.4% for Codex Multiagent-v2 and 14.4-20.6% for AI Scientist-v2. Ablation and behavioral analyses indicate that adding the two mechanisms together improves research coverage and continuity. We further evaluate Station on two open-ended tasks without oracle papers and find that some of the discoveries made by the agents closely match discoveries reported by researchers after the knowledge cutoff date. Together, these results indicate that a suitable environment can enable agents to autonomously make meaningful progress in open-ended scientific discovery.
Chinese Translation
最近的人工智能系统在给定明确指标时，在科学发现方面取得了快速进展，但它们是否能够自主进行开放式科学发现仍不清楚。我们研究了人工智能在Station中的开放式任务处理能力，Station是一个开放世界环境，其中多个代理模拟一个科学生态系统。为了解决开放式任务特有的挑战，我们提出通过两种机制增强Station：监督机制和周期性元反思，这些机制鼓励持续探索，即使在缺乏中间指标时。我们从三篇最近在ICLR上发表的口头论文中构建开放式任务。我们向代理提供每篇论文研究的主要研究问题，同时保留论文的结果并禁用网络访问。然后，我们测量代理重新发现的原始发现数量——按单独标准划分。我们发现，Station平均重新发现了62.7%的标准，而Codex Multiagent-v2为15.4%，AI Scientist-v2为14.4-20.6%。消融和行为分析表明，结合这两种机制可以改善研究覆盖率和连续性。我们进一步在没有参考论文的两个开放式任务上评估Station，发现代理所做的一些发现与知识截止日期后研究人员报告的发现非常接近。综合来看，这些结果表明，适当的环境可以使代理自主在开放式科学发现中取得有意义的进展。
cs.AI / 11 / 2610.08966
Humanity's Sixth Sense: Benchmarking Intuitive Visual Reasoning in Multimodal Models
人类的第六感：多模态模型中直观视觉推理的基准测试
Guo, Xingang, Gu, Jing, Jang, Brian, Wang, Renxiong, Tyagi, Utkarsh, Quigley, Daniel, Li, Steven, Yan, David, Zhang, Daniel Yue, Yi, Darvin, Huang, Forrest, Kim, HiJae, Zhang, Tianyi, Lichtarge, Jared, Huang, Jihua, Xue, Le, Tomar, Manan, Zhang, Qiuyi Richard, Yu, Ruofei, Neel, Seth, Hu, Yaning, Valentine, Marcella, Jiang, Xinzhe, Evans, Daniel, Wang, Chenguang, Tran, Dustin, Zhao, Tong, Yang, Yinfei, He, Yunzhong
Abstract
Humans perceive far more in a scene than what is explicitly depicted: a single glance captures past causes and future trajectories; a quick peek determines if a vehicle can fit between two parked cars; a few seconds of video reveals who holds authority in a room; and a fleeting clip highlights subtle abstract patterns like unwritten rules or hidden labels. This capacity reflects a form of humanity's sixth sense: an intuitive reasoning mechanism that recovers implicit information beyond raw sensory perception. Crucially, this rapid, zero-shot visual intuition underpins everyday navigation and social interaction, making it a vital capability for Multimodal Large Language Models (MLLMs) deployed alongside people. Existing visual benchmarks, however, target either deliberate expert-level analysis in academic and mathematical domains or low-level perception, leaving the intuitive reasoning that people perform largely untested. To bridge this gap, we introduce Humanity's Sixth Sense (HSS), a benchmark for intuitive visual reasoning. HSS spans diverse image and video inputs, organizes items under a structured taxonomy, and pairs each with human-written prompts probing the implicit temporal, spatial, social, and abstract structure that people infer at a glance. Frontier MLLMs fall short of human performance: participants reach 93.1% accuracy, while the strongest model, GPT-6-astra, reaches only 53.6% even at maximum reasoning effort. Despite excelling in many complex tasks that require advanced perception and knowledge, current models still struggle significantly on these visual tasks that are intuitive for humans. We further explore agentic setup that apply dynamic visual manipulation to HSS, which narrows but does not close the gap. HSS establishes intuitive visual reasoning as a measurable axis and directs attention to a capability that scaling on current benchmarks has so far left behind.
Chinese Translation
人类在场景中感知到的远不止于显式描绘的内容：一个快速的瞥见能够捕捉过去的原因和未来的轨迹；一个快速的观察可以判断一辆车是否能在两辆停放的车之间通过；几秒钟的视频揭示了谁在房间中拥有权威；而一个短暂的片段则突显了诸如未成文规则或隐藏标签等微妙的抽象模式。这种能力反映了人类第六感的一种形式：一种直观推理机制，能够恢复超越原始感官感知的隐含信息。至关重要的是，这种快速的零-shot视觉直觉支撑着日常导航和社交互动，使其成为与人类共同部署的多模态大型语言模型（MLLMs）所必需的重要能力。然而，现有的视觉基准测试要么针对学术和数学领域的深思熟虑的专家级分析，要么针对低级感知，导致人们所进行的直观推理在很大程度上未得到测试。为了填补这一空白，我们引入了人类的第六感（HSS），这是一个直观视觉推理的基准。HSS涵盖了多样的图像和视频输入，按照结构化分类法组织项目，并将每个项目与人类撰写的提示配对，探讨人们在一瞥之间推断出的隐含时间、空间、社会和抽象结构。前沿的多模态大型语言模型在性能上未能达到人类水平：参与者的准确率达到93.1%，而最强的模型GPT-6-astra即使在最大推理努力下也仅达到53.6%。尽管在许多需要高级感知和知识的复杂任务中表现出色，当前模型在这些对人类而言直观的视觉任务上仍然显著挣扎。我们进一步探索了将动态视觉操控应用于HSS的代理设置，虽然缩小了差距，但并未完全弥补。HSS确立了直观视觉推理作为一个可测量的维度，并将注意力引导到一个在当前基准测试中尚未得到充分扩展的能力上。
cs.AI / 12 / 2610.08967
Socio-Foundation: A Model for Generalizable Individual Behavior Simulation via Hierarchical Capability Distillation
社会基础：通过层次能力蒸馏实现可推广的个体行为模拟模型
Abstract
Simulating individual behavior requires large language models (LLMs) to preserve persona traits while adapting to dynamic social contexts. However, general-purpose LLMs often flatten distinct personas, while task-specific tuning suffers from fragmentation and generalization. To overcome these challenges, we organize individual simulation into the \textbf{FONTS Taxonomy}, comprising five complementary capability dimensions: \emph{persona fidelity} (\textbf{F}), \emph{outcome realization} (\textbf{O}), \emph{behavioral naturalness} (\textbf{N}), \emph{trajectory coherence} (\textbf{T}), and \emph{social grounding} (\textbf{S}). Grounded in this taxonomy, we curate a standardized training corpus library of approximately 10 million instances across 14 representative datasets and present \textbf{Socio-Foundation}. Socio-Foundation decouples specialization from integration via a three-stage pipeline: learning task experts via DAPO, consolidating them into capability experts via off-policy distillation, and unifying them via multi-teacher on-policy distillation (MOPD). We also establish \textbf{IndiEval}, consolidating 29 metrics across the FONTS dimensions. Experiments show that Socio-Foundation outperforms its \textit{Qwen3-8B} base by 11.0 points and approaches frontier models such as \textit{GLM-5.2}, with ablations and out-of-distribution evaluations further demonstrating the effectiveness and generalization of our model.
Chinese Translation
模拟个体行为需要大型语言模型（LLMs）在适应动态社会背景的同时保持个性特征。然而，通用的LLMs往往会将不同的个性特征压平，而特定任务的调优则面临碎片化和泛化的问题。为了解决这些挑战，我们将个体模拟组织为 extbf{FONTS分类法}，该分类法包括五个互补的能力维度： extit{个性忠实度}（ extbf{F}）、 extit{结果实现}（ extbf{O}）、 extit{行为自然性}（ extbf{N}）、 extit{轨迹一致性}（ extbf{T}）和 extit{社会基础}（ extbf{S}）。基于这一分类法，我们策划了一个标准化的训练语料库，涵盖约1000万个实例，来自14个代表性数据集，并提出了 extbf{社会基础}（Socio-Foundation）。社会基础通过一个三阶段的流程将专业化与整合解耦：通过DAPO学习任务专家，通过离策略蒸馏将其整合为能力专家，并通过多教师在线策略蒸馏（MOPD）将其统一。我们还建立了 extbf{IndiEval}，整合了FONTS维度的29个指标。实验表明，社会基础在性能上比其 extit{Qwen3-8B}基础模型提高了11.0分，并接近于前沿模型如 extit{GLM-5.2}，消融实验和分布外评估进一步证明了我们模型的有效性和泛化能力。
cs.AI / 13 / 2610.08993
Verify Less, Evolve More: Training Idea-Level Critics for Verification-Efficient ML Evolving Agents
验证更少，演化更多：训练思想层面评论者以实现高效验证的机器学习演化代理
Abstract
As large language models become more powerful, self-evolving agents are able to tackle challenging tasks including AI for machine learning (AI4ML). In AI4ML, while empirical verification is available, it often requires computationally costly model training and evaluation, limiting the speed and scale of agent evolution. Yet verification efficiency remains under-explored, and frontier models provide only limited gains when used directly as idea selectors. We address this gap with specialized idea-level critic models that predict whether a proposed ML modification will improve upon the current solution, allowing agents to screen ideas and concentrate verification resources on the most promising candidates. We train the critic models through supervised fine-tuning on high-quality critiques synthesized by Gemini-3.1-Pro, followed by GRPO to further improve their predictive accuracy. Empirically, our critic models outperform Gemini-3.1-Pro in static idea evaluation, and these gains extend to agent inference, continual learning, and policy training. During inference-time evolution, they improve final solution quality under the same verification budget by selecting more promising ideas, with further gains from continual learning. During policy training, they serve as learned reward models, reserving empirical verification for uncertain cases and enabling substantially more policy updates with the same verification resources. Together, these results show that idea-level critic models help ML agents discover better solutions and learn stronger proposal policies under limited verification budgets.
Chinese Translation
随着大型语言模型的不断增强，自我演化的代理能够应对包括机器学习的人工智能（AI4ML）在内的挑战性任务。在AI4ML中，尽管可以进行经验验证，但通常需要耗费大量计算资源进行模型训练和评估，从而限制了代理演化的速度和规模。然而，验证效率仍然未得到充分探索，前沿模型在作为思想选择器直接使用时仅能提供有限的收益。我们通过专门的思想层面评论者模型来填补这一空白，该模型能够预测所提议的机器学习修改是否会改善当前解决方案，从而使代理能够筛选思想，并将验证资源集中在最有前景的候选者上。我们通过对由Gemini-3.1-Pro合成的高质量评论进行监督微调来训练评论者模型，随后通过GRPO进一步提高其预测准确性。在经验上，我们的评论者模型在静态思想评估中优于Gemini-3.1-Pro，这些收益也扩展到代理推理、持续学习和策略训练。在推理时演化过程中，它们在相同的验证预算下通过选择更有前景的思想来提高最终解决方案的质量，并通过持续学习获得进一步的收益。在策略训练过程中，它们作为学习的奖励模型，仅在不确定的情况下保留经验验证，从而在相同的验证资源下实现更多的策略更新。这些结果表明，思想层面评论者模型帮助机器学习代理在有限的验证预算下发现更好的解决方案并学习更强的提案策略。
cs.AI / 14 / 2610.09000
How Fragile Is On-Device Language Model Safety? Localizing Safety-Critical Parameters for Sparse Fault Analysis
设备端语言模型安全性有多脆弱？稀疏故障分析中的安全关键参数定位
Abstract
As small language models (SLMs) are increasingly deployed on resource-constrained and on-device platforms, including as components of agentic systems, the integrity of locally stored model parameters becomes an important safety concern. We investigate whether safety-sensitive behavior in LLaMA-2-7B-Chat is concentrated within a sparse subset of parameters, creating a reduced fault surface for targeted analysis. We study two complementary localization methods: low-rank safety-associated subspace analysis and parameter-level safety--utility importance filtering. Both approaches reveal highly non-uniform safety sensitivity across the network, with the MLP down_proj consistently emerging as a prominent safety-sensitive component and o_proj providing a smaller contribution. Using parameter-level localization, modifying only 0.19% of model weights in down_proj yields 53% Basic ASR and 56% GCG ASR, while tinyBenchmarks accuracy remains at 51.6% compared with a 52.2% unmodified baseline. These results motivate targeted fault analysis and selective integrity protection for language models deployed in resource-constrained, on-device, and agentic settings.
Chinese Translation
随着小型语言模型（SLMs）越来越多地部署在资源受限的设备端平台上，包括作为智能系统的组成部分，本地存储的模型参数的完整性成为一个重要的安全问题。我们研究了在 LLaMA-2-7B-Chat 中，安全敏感行为是否集中在一小部分稀疏参数中，从而为针对性分析创造一个减少的故障面。我们研究了两种互补的定位方法：低秩安全相关子空间分析和参数级安全-效用重要性过滤。这两种方法都揭示了网络中安全敏感性高度不均匀的特征，其中 MLP down_proj 一直是一个显著的安全敏感组件，而 o_proj 的贡献较小。通过参数级定位，仅修改 down_proj 中 0.19% 的模型权重即可实现 53% 的基本 ASR 和 56% 的 GCG ASR，而 tinyBenchmarks 的准确率保持在 51.6%，相比之下未修改的基线为 52.2%。这些结果激励了对在资源受限、设备端和智能环境中部署的语言模型进行针对性的故障分析和选择性的完整性保护。
cs.AI / 15 / 2610.09002
Learning to Report Unsafe Tasks in a Multi-Agent Game
在多智能体游戏中学习报告不安全任务
Abstract
When agents share a reward for completed tasks, reporting unsafe work can reduce the reporter's reward by stopping a task. Audits can make reporting optimal without ensuring that further training teaches a silent team to report. We study this learning problem in a game where any witness can stop a task by reporting. With $k$ witnesses per task sharing a policy and drawing independently, the expected-reward derivative with respect to their shared silence probability counts each task's benefit $k$ times at universal silence. The comparison with universal reporting counts it once. For arbitrary policy groups, we give an audit condition sufficient for exact policy-gradient updates to reach universal reporting and, apart from boundary cases, necessary near universal silence. In a balanced family, the cheapest audits meeting the condition with prescribed positive margins cost exactly $k$ times as much for full sharing as for one policy per role. We train PPO policies on 24 witness graphs from learned silence. Separating co-witnesses reduces unsafe completion by 33.59 percentage points compared with shuffled groups of the same sizes under the same audits (95% graph-bootstrap interval: 21.03-45.13). Only 9 of 48 witness-group runs achieve below 1% unsafe completion while retaining at least 90% legitimate completion. At the same audit budget, a fully shared network meets both thresholds in none of 48 runs with independent action draws and all 48 with a common draw.
Chinese Translation
当智能体共享完成任务的奖励时，报告不安全的工作可能会通过停止任务来减少报告者的奖励。审计可以使报告变得最优，但并不能确保进一步的训练能够教会沉默的团队进行报告。我们在一个游戏中研究这个学习问题，在这个游戏中，任何目击者都可以通过报告来停止任务。对于每个任务有 $k$ 个目击者共享一个策略并独立抽样，关于他们共享的沉默概率的期望奖励导数在普遍沉默的情况下将每个任务的收益计算了 $k$ 次。与普遍报告的比较则只计算一次。对于任意策略组，我们给出了一个审计条件，该条件足以使精确的策略梯度更新达到普遍报告，并且在边界情况下，几乎在普遍沉默的情况下是必要的。在一个平衡的家族中，满足该条件且具有规定正边际的最便宜审计的成本恰好是完全共享的 $k$ 倍，而不是每个角色一个策略。我们在从学习到的沉默中获得的 24 个目击者图上训练了 PPO 策略。与在相同审计下相同大小的随机分组相比，分离共同目击者使不安全完成率降低了 33.59 个百分点（95% 图形自助区间：21.03-45.13）。在 48 次目击者组运行中，只有 9 次实现了低于 1% 的不安全完成，同时保留至少 90% 的合法完成。在相同的审计预算下，完全共享的网络在 48 次独立行动抽样中没有一次满足两个阈值，而在所有 48 次共同抽样中也没有满足。
cs.AI / 16 / 2610.09007
Sigma-Hunter: A Domain-Specific Language Model for Threat Hunting and Detection Engineering
Sigma-Hunter：一种用于威胁狩猎和检测工程的领域特定语言模型
Abstract
Detection engineers must translate threat reports, forensic observations, and hunt hypotheses into precise, testable rules. General-purpose large language models (LLMs) can draft such rules, but often produce invalid YAML, incorrect log sources, unsupported fields, or overly broad detection logic. This paper presents \emph{Sigma-Hunter}, a domain-adapted LLM for analyst-assistive Sigma rule generation and threat hunting. We build an instruction-tuning dataset from 3,635 validated open-source Sigma rules, expanded into 7,663 question-answer and analyst-reasoning examples. Each source rule is assigned to a single train, validation, or test partition before this expansion, so no rule leaks across splits. We fine-tune a 7B Mistral model and a Phi-4 model with LoRA and score held-out rule generations on syntax, approximate field consistency, and a semantic judgment of detection logic, completeness, selectivity, and log-source alignment. Sigma-Hunter-Mistral scores 8.17 overall, against 7.88 for the strongest general-purpose baseline and 4.61 for untuned Mistral. Two findings stand out: domain adaptation enables a compact 7B model to perform competitively with larger general-purpose models on this structured task, and syntactic validity is a weak proxy for semantic rule quality, as several baselines emit well-formed YAML carrying weak detection logic. The adapted models run locally, which suits detection engineering in disconnected environments where analysts cannot reach hosted model services.
Chinese Translation
检测工程师必须将威胁报告、取证观察和狩猎假设转化为精确、可测试的规则。通用的大型语言模型（LLMs）可以起草这些规则，但往往会生成无效的 YAML、不正确的日志源、不支持的字段或过于宽泛的检测逻辑。本文提出了 extit{Sigma-Hunter}，一种针对分析师辅助 Sigma 规则生成和威胁狩猎的领域适应型 LLM。我们从 3,635 个经过验证的开源 Sigma 规则构建了一个指令调优数据集，并扩展为 7,663 个问答和分析师推理示例。每个源规则在扩展之前被分配到单一的训练、验证或测试分区，因此没有规则在分割中泄漏。我们使用 LoRA 对 7B 的 Mistral 模型和 Phi-4 模型进行了微调，并在语法、近似字段一致性以及检测逻辑、完整性、选择性和日志源对齐的语义判断上对保留的规则生成进行了评分。Sigma-Hunter-Mistral 的总体得分为 8.17，而最强通用基线为 7.88，未调优的 Mistral 为 4.61。两个发现尤为突出：领域适应使得紧凑的 7B 模型在这一结构化任务上能够与更大的通用模型竞争，并且语法有效性是语义规则质量的弱代理，因为一些基线生成的格式良好的 YAML 具有较弱的检测逻辑。适应后的模型可以在本地运行，这适合于在分析师无法访问托管模型服务的离线环境中的检测工程。
cs.AI / 17 / 2610.09008
Whose Memory Is It? Scope-Aware Commit Rules for Long-Term LLM Memory
记忆是谁的？面向范围的长期 LLM 记忆提交规则
Abstract
Persistent memory allows an LLM agent to carry experience across conversations, but it also turns a local reasoning mistake into a durable one. During deliberation, an agent may consider a plan, simulate a tool result, report another speaker's belief, and then reject all of them. If memory retains only the resulting sentences, those once-useful possibilities can later return as facts. The record is neither fabricated nor irrelevant; it has simply been detached from the context in which it was valid. We identify this missing context as \emph{discourse ownership}: the world, branch, or speaker that licenses a proposition. Our first finding is counterintuitive. Language models already carry a causally active signal for ownership, yet conventional memory interfaces discard it when they convert reasoning into records. We introduce CASK (Causally Anchored Scoping Keys), a commit rule that preserves this signal so that shared-world facts enter durable memory while provisional content remains available only within its original scope. Our second finding is that the most obvious way to preserve the signal---storing the discovered internal coordinates---is unreliable because equivalent representations need not keep the same coordinates. CASK instead preserves the stable relations that express ownership. Controlled long-conversation conflicts and tool-agent traces show that this design improves memory admission and prevents provisional content from contaminating later answers while complementing runtime provenance. The resulting commit boundary lets agents explore more possibilities without granting every intermediate sentence authority over future behavior.
Chinese Translation
持久性记忆使 LLM 代理能够在对话中携带经验，但它也将局部推理错误转变为持久错误。在深思熟虑的过程中，代理可能会考虑一个计划、模拟工具结果、报告其他说话者的信念，然后拒绝所有这些。如果记忆仅保留结果句子，那么那些曾经有用的可能性可能会作为事实重新出现。记录既不是虚构的，也不是无关的；它只是与其有效的上下文脱离了联系。我们将这种缺失的上下文称为 extit{话语所有权}：授权命题的世界、分支或说话者。我们的第一个发现是反直觉的。语言模型已经携带了一个因果活跃的所有权信号，但传统的记忆接口在将推理转换为记录时却丢弃了它。我们引入了 CASK（Causally Anchored Scoping Keys），一种提交规则，保留了这一信号，使共享世界的事实进入持久记忆，而临时内容仅在其原始范围内可用。我们的第二个发现是，保留信号的最明显方法——存储发现的内部坐标——是不可靠的，因为等效表示不必保持相同的坐标。CASK 则保留了表达所有权的稳定关系。受控的长对话冲突和工具代理痕迹表明，这一设计改善了记忆接纳，并防止临时内容污染后续答案，同时补充了运行时来源。由此产生的提交边界使代理能够探索更多可能性，而不赋予每个中间句子对未来行为的权威。
cs.AI / 18 / 2610.09013
Enabling Dynamic Computation in Looped LMs
在循环语言模型中实现动态计算
Abstract
Looped LMs are parameter efficient and promise dynamic computation (saving memory and FLOPs on easy tokens). However, state-of-the-art open Looped LMs trained with this dynamic computation capability (Ouro models) do not realize it in practice as each loop iteration (depth) requires its own level of KV-cache, necessitating all loop computations. Moreover, Ouro's early-exit prior is enforced on each token equally, which results in static lower-depth like processing of all tokens regardless of difficulty. In this work, we propose a simple "best-available" KV caching strategy that works out-of-the-box, creating a new frontier in the performance vs depth space. Our approach enables up to 30% reduction in FLOPs and KV memory while retaining full-depth performance, showing the true flexibility of Looped LMs. Furthermore, training looped LMs with awareness about this KV caching strategy improves performance and efficiency. Finally, we apply a small but effective fix to the early-exit prior enforcement objective that makes tokens exit at truly heterogeneous depths based on effort. Our findings are validated on Ouro models as well as smaller looped LMs pre-trained from scratch.
Chinese Translation
循环语言模型（Looped LMs）具有参数高效性，并承诺实现动态计算（在简单标记上节省内存和浮点运算次数）。然而，采用这种动态计算能力训练的最先进的开放循环语言模型（Ouro模型）在实际应用中并未实现，因为每次循环迭代（深度）都需要其自身的键值缓存（KV-cache），这使得所有循环计算成为必要。此外，Ouro的提前退出先验对每个标记的强制执行是相同的，这导致所有标记无论难易程度都以静态的低深度处理。在本研究中，我们提出了一种简单的“最佳可用”键值缓存策略，该策略开箱即用，在性能与深度空间中创造了新的前沿。我们的方法在保持全深度性能的同时，实现了高达30%的浮点运算次数和键值内存的减少，展示了循环语言模型的真正灵活性。此外，训练循环语言模型时考虑到这种键值缓存策略可以提高性能和效率。最后，我们对提前退出先验强制目标进行了小但有效的修正，使标记能够根据努力程度在真正异质的深度上退出。我们的发现已在Ouro模型以及从头开始预训练的小型循环语言模型上得到了验证。
cs.AI / 19 / 2610.09016
PAIR: Bridging Perception and Action in Vision-Language-Action Models
PAIR：在视觉-语言-行动模型中桥接感知与行动
Abstract
Vision-language-action (VLA) models map visual observations and language instructions to continuous robot actions. This task requires a transition from representations that describe the scene and instruction to representations that support action generation. Many continuous-action VLAs leave this transition implicit and supervise it mainly through the final action-prediction loss. We introduce PAIR, a framework that learns a shared perception-action representation between these two spaces. During training, a Masked Action Autoencoder encodes expert action chunks into horizon-aligned Action Latent Tokens. A Bridge Module extracts task-relevant features from the current visual-language representations. PAIR aligns these features with the Action Latent Tokens to form Bridge Tokens that preserve task information and capture the structure of expert actions. The Bridge Tokens are then projected into the action-token space and injected into the initial Action Tokens, providing an action-ready starting point for Action Expert refinement. At inference, the autoencoder is removed, and the Bridge Tokens are generated only from the current observation and instruction. Experiments on LIBERO, LIBERO-Plus, and CALVIN ABC-D show gains for the evaluated OpenVLA-OFT and VLA-Adapter models. On LIBERO-Plus, PAIR raises VLA-Adapter's success rate from 59.1% to 64.2%. On CALVIN, it increases VLA-Adapter's average completed sequence length from 4.42 to 4.53. Across seven real-world tasks, PAIR raises OpenVLA-OFT's success rate from 51.4% to 65.0%. Representation analyses show that Bridge Tokens retain task information while making continuous-action information accessible before Action Expert refinement. These results support a shared intermediate representation as a useful interface between perception and action in continuous-action VLAs.
Chinese Translation
视觉-语言-行动（VLA）模型将视觉观察和语言指令映射到连续的机器人动作。该任务需要从描述场景和指令的表示转变为支持动作生成的表示。许多连续动作的VLA将这一转变隐含化，并主要通过最终的动作预测损失进行监督。我们引入了PAIR，一个学习这两个空间之间共享感知-行动表示的框架。在训练过程中，掩码动作自编码器将专家动作块编码为与时间对齐的动作潜在标记。桥接模块从当前的视觉-语言表示中提取与任务相关的特征。PAIR将这些特征与动作潜在标记对齐，形成保留任务信息并捕捉专家动作结构的桥接标记。然后，桥接标记被投影到动作标记空间，并注入到初始动作标记中，为动作专家的细化提供一个准备好的起始点。在推理阶段，自动编码器被移除，桥接标记仅从当前观察和指令生成。在LIBERO、LIBERO-Plus和CALVIN ABC-D上的实验表明，评估的OpenVLA-OFT和VLA-Adapter模型均有提升。在LIBERO-Plus上，PAIR将VLA-Adapter的成功率从59.1%提高到64.2%。在CALVIN上，它将VLA-Adapter的平均完成序列长度从4.42增加到4.53。在七个真实世界任务中，PAIR将OpenVLA-OFT的成功率从51.4%提高到65.0%。表示分析表明，桥接标记在动作专家细化之前保留了任务信息，同时使连续动作信息可访问。这些结果支持共享中间表示作为连续动作VLA中感知与行动之间有用的接口。
cs.AI / 20 / 2610.09021
Not Every Call Needs a Frontier Model: Per-Call-Site Evaluation of Small Language Models in a Deployed Agentic Home-Automation System
并非每个调用都需要前沿模型：在部署的自主家居自动化系统中对小型语言模型的逐调用站点评估
Abstract
An agentic system issues several structurally different kinds of LLM calls. It routes intent, classifies actions, grounds language in a device registry, plans multi-agent pipelines and writes the Python code those pipelines run. The difficulty of these call sites varies by an order of magnitude, yet in practice a single model, chosen for the hardest site, serves all of them. In this work, we evaluate 9 models from 0.8B to a frontier hosted model across the five call sites of a deployed open-source home-automation framework (Wactorz), using its unmodified production prompts and two real Home Assistant installations (280 cases, 2520 scored calls). We find that capability is not ordered the same way at every site, and that larger models are not uniformly better: one 4B model is worse than its 2B sibling at grounded actuation. Paired testing shows the best local model to be statistically indistinguishable from both hosted models at four of five sites. Only code generation separates them, against a small hosted model (p = 0.039) as well as a frontier one (p = 0.002). Aggregate accuracy also hides a safety failure specific to actuation, where small models resolve the accuracy/refusal trade-off in degenerate ways: one model (Gemma4 E2B) actuates on 87.2% of requests for devices the site does not own, while another refuses every request it receives. Routing each site to its best local model reaches 91.8% against 95.4% at no per-call cost. In a live deployment judged by a user, hosting only the two generative sites matches hosting everything (39/43 against 39/43) for 28% of the spend, and the actuation gap the benchmark predicted appears as exactly one case in twenty-six. Benchmark, harness and all records are released at https://github.com/waldiez/slm-callsite-eval.
Chinese Translation
一个自主系统发出几种结构上不同的LLM调用。它路由意图、分类动作、在设备注册表中确定语言、规划多智能体管道并编写这些管道运行的Python代码。这些调用站点的难度相差一个数量级，但在实践中，选择用于最困难站点的单一模型服务于所有站点。在本研究中，我们评估了从0.8B到一个前沿托管模型的9个模型，涵盖了一个部署的开源家居自动化框架（Wactorz）的五个调用站点，使用其未修改的生产提示和两个真实的Home Assistant安装（280个案例，2520个评分调用）。我们发现，能力在每个站点的排序并不相同，且更大的模型并不总是更好：一个4B模型在基础激活方面的表现不及其2B同类。配对测试显示，最佳本地模型在五个站点中的四个与两个托管模型在统计上无显著差异。只有代码生成使它们区分开来，针对一个小型托管模型（p = 0.039）以及一个前沿模型（p = 0.002）。综合准确性也掩盖了一个特定于激活的安全失败，其中小型模型以退化的方式解决准确性/拒绝的权衡：一个模型（Gemma4 E2B）对该站点不拥有的设备的请求激活率为87.2%，而另一个则拒绝了其收到的每一个请求。将每个站点路由到其最佳本地模型的准确率达到91.8%，而不产生每次调用的成本。在用户评估的实时部署中，仅托管两个生成站点的表现与托管所有站点相当（39/43对39/43），且所预测的激活差距在26个案例中恰好出现一次。基准测试、工具和所有记录已发布在https://github.com/waldiez/slm-callsite-eval。
cs.AI / 21 / 2610.09026
BEACON-SP: Ontology-Grounded GraphRAG Framework for Clinical Suicide Risk Assessment
BEACON-SP：基于本体的图检索增强生成框架用于临床自杀风险评估
Abstract
We present BEACON-SP, an ontology-grounded Graph Retrieval-Augmented Generation (GraphRAG) framework for clinician-facing decision support in behavioral health settings such as suicide prevention, where effective assessment requires integrating heterogeneous clinical, behavioral, social, and temporal evidence. BEACON-SP combines patient knowledge graphs with ontology-guided retrieval to support multi-hop reasoning across diagnoses, medications, risk and protective factors, life events, and temporal relationships. The framework is enabled by a comprehensive suicide prevention ontology that integrates the Three-Step Theory, the Integrated Motivational-Volitional Model, and the Suicide Social Determinants of Health Ontology into a unified representation of patient risk factors. We construct ontology-grounded patient knowledge graphs and evaluate BEACON-SP for clinician-facing question answering. Compared with a vector-based retrieval-augmented generation (RAG) baseline on a 1,500-query benchmark spanning 15 clinical categories and 100 patients, BEACON-SP improves completeness, clinical relevance, and evidence grounding under a corrected comparative evaluation protocol, with a small gain on factual accuracy. In paired criterion-level comparisons, GraphRAG is preferred in 76.4% of cases. These results demonstrate the potential of ontology-guided GraphRAG to provide structured, contextualized patient evidence for clinical decision support.
Chinese Translation
我们提出了BEACON-SP，这是一个基于本体的图检索增强生成（GraphRAG）框架，旨在为行为健康领域（如自杀预防）中的临床决策支持提供帮助，其中有效的评估需要整合异构的临床、行为、社会和时间证据。BEACON-SP结合了患者知识图谱与本体指导的检索，以支持跨诊断、药物、风险和保护因素、生活事件及时间关系的多跳推理。该框架由一个全面的自杀预防本体支持，该本体将三步理论（Three-Step Theory）、综合动机-意志模型（Integrated Motivational-Volitional Model）和自杀健康社会决定因素本体（Suicide Social Determinants of Health Ontology）整合为患者风险因素的统一表示。我们构建了基于本体的患者知识图谱，并评估了BEACON-SP在临床问题回答中的表现。与在涵盖15个临床类别和100名患者的1,500个查询基准上的基于向量的检索增强生成（RAG）基线相比，BEACON-SP在经过修正的比较评估协议下提高了完整性、临床相关性和证据基础，同时在事实准确性上也有小幅提升。在配对标准级别比较中，GraphRAG在76.4%的情况下被优先选择。这些结果展示了本体指导的GraphRAG在为临床决策支持提供结构化、情境化患者证据方面的潜力。
cs.AI / 22 / 2610.09034
Shared-Roadmap Generation and Evaluator for Multi-Agent Path Planning Using Heterogeneous Graph Neural Network
基于异构图神经网络的多智能体路径规划共享路线图生成与评估
Abstract
Multi-agent path planning (MAPP) in continuous environments often relies on roadmaps to balance safety and search efficiency. However, traditional roadmap generation methods, such as lattice grids or standard sampling-based approaches, frequently face a trade-off between graph density and the likelihood of finding feasible, high-quality solutions. In this paper, we propose a scalable heterogeneous Graph Neural Network (GNN) framework for the automated generation and evaluation of shared multi-agent roadmaps. Our model covers the representation of waypoints, agent locations, and task locations as distinct nodes in a heterogeneous graph, allowing it to reason over global connectivity and inter-agent interactions. By training on occupation density maps aggregated and collected from expert solver trajectories, the GNN learns to identify critical points of interest and prune redundant nodes and edges. This process produces a compact, coordination-aware roadmap that is invariant to task permutations and is reusable for multi-agent pick and delivery tasks. Experimental results demonstrate that our framework can reduce planning effort and can potentially find better solutions, reaching at least 40% reduction in runtime and in graph size for dense roadmaps.
Chinese Translation
在连续环境中的多智能体路径规划（MAPP）通常依赖于路线图来平衡安全性和搜索效率。然而，传统的路线图生成方法，如格点或标准采样方法，常常面临图密度与找到可行的高质量解决方案之间的权衡。本文提出了一种可扩展的异构图神经网络（GNN）框架，用于自动生成和评估共享的多智能体路线图。我们的模型将路径点、智能体位置和任务位置表示为异构图中的不同节点，从而能够推理全局连通性和智能体间的交互。通过对从专家求解器轨迹中聚合和收集的占用密度图进行训练，GNN学习识别关键兴趣点，并修剪冗余节点和边。这一过程生成了一个紧凑的、关注协调的路线图，该路线图对任务排列不变，并可重复用于多智能体的取送任务。实验结果表明，我们的框架能够减少规划工作量，并有可能找到更好的解决方案，在密集路线图中实现至少40%的运行时间和图大小的减少。
cs.AI / 23 / 2610.09037
When the Governor Becomes the Disturbance: Control-Generated Disturbance and Cost-Aware Backoff in Governed Tool-Using Agents
当治理者成为干扰源：控制生成的干扰与成本意识的退避在治理工具使用代理中的应用
Abstract
Supervisory governors can interfere with the tool-using agents they regulate. We study this possibility in a controlled file-recovery environment where increases in regulatory intensity trigger experimentally imposed tool failures. A cost-blind governor can turn these failures into persistent blocking that prevents task completion. We compare this governor with a backoff rule that reduces intervention probability using a moving average of known induced events. On a hand-coded stochastic-policy agent, the failure pattern appears under both result replacement and execution of corrupted tool arguments. For the persistent policy, adaptive backoff improves completion relative to a fixed weak governor with approximately matched intervention frequency. A Gemini 2.5 Flash experiment comprising 576 episodes across 6 tasks also shows reduced blocking and improved completion under backoff; among the tested settings, intermediate backoff strength achieves the highest observed aggregate success. These results identify an interaction between intervention cost and persistent action blocking, together with a possible mitigation. The cost mechanisms are imposed and their induced events are directly observable to the backoff rule; applicability beyond this controlled environment remains an empirical question.
Chinese Translation
监督性治理者可能会干扰他们所监管的工具使用代理。我们在一个受控的文件恢复环境中研究了这种可能性，在该环境中，监管强度的增加会引发实验性施加的工具故障。一个不考虑成本的治理者可以将这些故障转化为持续的阻塞，从而阻碍任务的完成。我们将这一治理者与一种退避规则进行比较，该规则使用已知诱发事件的移动平均值来降低干预概率。在一个手工编码的随机策略代理中，故障模式在结果替换和执行损坏工具参数的情况下均会出现。对于持续策略，适应性退避相较于一个干预频率大致匹配的固定弱治理者，能够改善任务完成情况。一个包含576个回合和6个任务的Gemini 2.5 Flash实验也显示，在退避机制下阻塞减少，完成情况改善；在测试的设置中，中等强度的退避实现了最高的观察到的总成功率。这些结果揭示了干预成本与持续行动阻塞之间的相互作用，以及可能的缓解方法。成本机制是施加的，其诱发事件对退避规则是直接可观察的；在这一受控环境之外的适用性仍然是一个经验性问题。
cs.AI / 24 / 2610.09039
From High Recall to High Utility: Dataset-Adaptive Post-Processing of LLM-Generated Customer Intents
从高召回到高效用：数据集自适应的LLM生成客户意图后处理
Abstract
Large language models can extract useful signals from heterogeneous enterprise data, but high-recall extraction often produces outputs that are duplicated, uneven in granularity, semantically overlapping, or too numerous for downstream systems and human reviewers to use effectively. We present a dataset-adaptive post-processing architecture developed for Customer Intent Extraction (CIE), where unstructured customer language is transformed into stable, traceable intent units. The approach separates recall-oriented extraction from utility-oriented reduction. Source-specific preprocessing first isolates evidence from multimodal plans, sparse operational records, and structured opportunity data. Candidate intents are then standardized and deduplicated, optionally enriched with metadata for embedding computation, represented in a shared semantic vector space, and grouped using a clustering strategy selected according to the candidate set's characteristics. Cluster-level keywords provide an explainability layer, while singleton reassignment requires agreement between embedding and keyword similarity. Finally, constrained language-model aggregation produces one concise intent per cluster without introducing unsupported concepts, and the resulting unit retains provenance, clustering, embedding, and generation metadata. This treats post-processing not as cosmetic cleanup, but as a semantic reduction layer converting high-recall LLM outputs into reusable enterprise intelligence. We also describe two downstream applications: Machine-Generated Intents, which infer likely objectives for customers lacking direct evidence from peer customers with similar profiles, and intent-guided semantic retrieval and mapping, which uses the stable intent as a query against a downstream decision space, illustrated here by mapping customer intents to business outcomes.
Chinese Translation
大型语言模型能够从异构企业数据中提取有用信号，但高召回的提取往往会产生重复、不均匀的粒度、语义重叠或数量过多的输出，使下游系统和人工审阅者难以有效使用。我们提出了一种为客户意图提取（Customer Intent Extraction, CIE）开发的数据集自适应后处理架构，其中非结构化的客户语言被转化为稳定、可追溯的意图单元。该方法将以召回为导向的提取与以效用为导向的减少分开。源特定的预处理首先从多模态计划、稀疏操作记录和结构化机会数据中隔离证据。候选意图随后被标准化和去重，选项上可通过元数据进行丰富以便于嵌入计算，并在共享的语义向量空间中表示，并根据候选集的特征选择聚类策略进行分组。聚类级别的关键词提供了解释性层，而单例重新分配则需要嵌入与关键词相似性之间的一致性。最后，受限语言模型的聚合为每个聚类生成一个简洁的意图，而不引入不支持的概念，生成的单元保留了来源、聚类、嵌入和生成元数据。这将后处理视为一种语义减少层，将高召回的LLM输出转化为可重用的企业智能。我们还描述了两个下游应用：机器生成意图（Machine-Generated Intents），用于推断缺乏直接证据的客户的可能目标，基于与相似档案的同行客户的证据，以及意图引导的语义检索和映射，利用稳定的意图作为对下游决策空间的查询，这里通过将客户意图映射到商业成果来说明。
cs.AI / 25 / 2610.09053
Justice After Identity: Large Language Models and the View from Everywhere
身份后的正义：大型语言模型与普遍视角
Abstract
The search for a common view of justice and fairness has challenged human collective activity, as our diverging judgments are unavoidably shaped by the self-interests of social position, personal benefit, cultural inheritance, and historical circumstance. John Rawls famously attempted to overcome this limitation through popularizing a philosophical tradition known by the phrase "the original position" - a thought experiment by which people select principles of justice without knowing the identities or advantages they will possess. Critics, however, have long questioned whether people can meaningfully suspend their social identities and suppress morally relevant forms of lived experience. Artificial intelligence engaged to calculate algorithmic and agentic fairness introduces a novel possibility. LLMs have no singular class, race, gender, nationality, or biography, yet their parameters encode linguistic representations of a vast range of human identities and moral traditions. Perhaps the ethical judgments of LLMs could approximate an integrative original position - a "view from everywhere" generated not by excluding social identities but by computationally incorporating their diversity.It is unlikely that humankind will "hand over the keys" to computational systems by simply delegating complete agentic control of distributive and procedural collective processes. But AI may play a role, perhaps a positive one, interacting with individual and collective human judgment as we often confront increasingly polarized views on what is fair and just. We present data comparing human, base model, and frontier/fine-tuned model judgments about classic moral dilemmas while systematically varying identity relationships and Rawlsian constraints on identity. We conclude by speculating whether, if advanced AI systems provide humans with thoughtful advice, humans would actually be likely to accept it.
Chinese Translation
对正义和公平的共同视角的追求挑战了人类的集体活动，因为我们不同的判断不可避免地受到社会地位、自我利益、文化传承和历史环境的影响。约翰·罗尔斯（John Rawls）以“原初状态”（the original position）这一哲学传统而闻名，试图克服这一局限性——这一思想实验让人们在不知道自己将拥有的身份或优势的情况下选择正义原则。然而，批评者长期以来质疑人们是否能够有意义地暂停其社会身份，并抑制与道德相关的生活经验形式。人工智能在计算算法公平性和代理公平性方面的应用引入了一种新可能性。大型语言模型（LLMs）没有单一的阶级、种族、性别、国籍或传记，但其参数编码了广泛的人类身份和道德传统的语言表现。也许，LLMs的伦理判断可以接近一种综合的原初状态——一种“普遍视角”，它不是通过排除社会身份而生成，而是通过计算性地纳入其多样性。人类不太可能通过简单地将分配和程序性集体过程的完全代理控制“交给”计算系统。但人工智能可能会发挥作用，或许是积极的，与个体和集体的人类判断互动，因为我们常常面临对公平和正义的日益两极化的看法。我们提供了数据，比较人类、基础模型和前沿/微调模型在经典道德困境中的判断，同时系统地变化身份关系和罗尔斯对身份的限制。最后，我们推测，如果先进的人工智能系统为人类提供深思熟虑的建议，人类是否真的可能接受这些建议。
cs.AI / 26 / 2610.09057
Epistemic Uncertainty-Aware Defect Detection for Quality Control in Medical Device Manufacturing
考虑认知不确定性的医疗设备制造质量控制缺陷检测
Abstract
Objective: We investigate whether accounting for epistemic uncertainty can improve the reliability of automated defect detection in medical device manufacturing. Methods: We consider a machine learning framework that operates on heterogeneous manufacturing and device-report data represented with Knowledge Graphs. To mitigate errors arising from uncertainty in the decision model, we analyze a principled rejection strategy to abstain from predictions whose estimated epistemic uncertainty exceeds a specified threshold. We evaluate the approach using standard synthetic benchmarks and real-world medical device report data. Results: The theoretical results establish the validity of the method characterizing the regimes under which it is expected to be effective. Empirically, the rejection strategy enables explicit control of coverage, that is, the proportion of samples for which the model issues predictions, while improving performance on the retained samples. On 266,170 real-world FDA MAUDE device reports, a 10% abstention rate reduces classification error by 48%, and more aggressive rejection (approximately 70% coverage) yields near-perfect accuracy on the retained samples. On standard synthetic manufacturing benchmarks, abstaining on 9% of the decisions, our approach reduces the risk up to 63% compared with the standard no-abstention approach. Conclusions: Abstaining from predictions with high epistemic uncertainty can provide a practical tool for controlling the reliability of machine learning-based defect detection, especially in high-stakes medical device manufacturing applications. Significance: Uncertainty-aware defect detection may support safer and more reliable quality assurance in medical device manufacturing by identifying cases that require additional inspection rather than issuing potentially harmful predictions.
Chinese Translation
目的：我们研究考虑认知不确定性是否能提高医疗设备制造中自动缺陷检测的可靠性。方法：我们考虑一个基于异构制造和设备报告数据的机器学习框架，这些数据通过知识图谱表示。为了减轻决策模型中因不确定性引起的错误，我们分析了一种原则性拒绝策略，以避免对那些估计的认知不确定性超过特定阈值的预测。我们使用标准的合成基准和真实世界的医疗设备报告数据对该方法进行评估。结果：理论结果确立了该方法的有效性，并描述了其预期有效的条件。在经验上，拒绝策略能够明确控制覆盖率，即模型发出预测的样本比例，同时提高保留样本的性能。在266,170份真实世界FDA MAUDE设备报告中，10%的拒绝率将分类错误降低了48%，而更激进的拒绝（约70%的覆盖率）在保留样本上实现了近乎完美的准确性。在标准的合成制造基准上，对9%的决策进行拒绝，我们的方法将风险降低了63%，与标准的无拒绝方法相比。结论：拒绝高认知不确定性的预测可以为控制基于机器学习的缺陷检测的可靠性提供一种实用工具，尤其是在高风险的医疗设备制造应用中。意义：考虑不确定性的缺陷检测可能通过识别需要额外检查的案例，而不是发出潜在有害的预测，从而支持医疗设备制造中的更安全和更可靠的质量保证。
cs.AI / 27 / 2610.09088
RippleCP: Measuring Counterfactual Checkpoint Advantage in Coding Agents
RippleCP：测量编码代理中的反事实检查点优势
Abstract
Agent checkpoint systems decide what state is recovery-relevant, how to snapshot it, and whether rollback is admissible. None decides which of the safe boundaries they expose are worth materializing. We formulate this as counterfactual checkpoint advantage, the reduction in future recovery cost obtained by checkpointing a candidate rather than skipping it, and measure it by driving a CP branch and a SKIP branch to the same logical failure and recovering both under matched model, tool, verifier, and stopping conditions. On a frozen pilot of 12 SWE-bench Verified tasks and 106 real recovery branches, checkpointing saves 49.4 s per task, and that figure resolves into two regimes two orders of magnitude apart. The first checkpoint returns 100.0 s on 156.5 s of protected work, a conversion of 0.64; a second one step later returns-1.1 s on 41.8 s, a conversion of -0.03. Recovery is a re-derivation rather than a replay, so preserved work is a poor guide to saved work, and the classical elapsed-work rule misprices the second checkpoint by its full nominal cost. We identify where placement can pay, and set the bar a placement policy must clear.
Chinese Translation
代理检查点系统决定哪些状态与恢复相关、如何快照这些状态以及是否允许回滚。但没有系统决定它们所暴露的安全边界中哪些值得实现。我们将其表述为反事实检查点优势，即通过对候选项进行检查点而不是跳过它所获得的未来恢复成本的减少，并通过驱动一个检查点（CP）分支和一个跳过（SKIP）分支到达相同的逻辑故障，并在匹配的模型、工具、验证器和停止条件下恢复两者来进行测量。在一个冻结的12个SWE-bench验证任务和106个真实恢复分支的试点中，检查点每个任务节省了49.4秒，这一数字分为两个相差两个数量级的范畴。第一个检查点在156.5秒的保护工作中返回100.0秒，转换率为0.64；而第二个检查点在41.8秒的保护工作中返回-1.1秒，转换率为-0.03。恢复是重新推导而非重放，因此保留的工作是对节省工作的一种较差指引，而经典的经过时间工作规则则以其全名义成本错误定价了第二个检查点。我们识别出放置可以带来收益的地方，并设定了放置策略必须达到的标准。
cs.AI / 28 / 2610.09107
Constraint Tree Exploration for Learning from Language Feedback
基于约束树的语言反馈学习探索
Abstract
Natural-language feedback in interactive learning often explains why an action failed by pointing to violated requirements. Misinterpreting this feedback can lead an agent to rule out valid solutions. We study this setting by modeling user intent as latent constraints over an action space and formulating learning from language feedback as pure exploration over feasible regions. We introduce TRACE, an algorithm that organizes candidate constraints in a tree and tests each proposed refinement by generating actions that satisfy it. TRACE commits to the refinement only if the resulting feedback does not contradict it over repeated tests. We distinguish two ways of using the same feedback: (i) falsification, which detects contradictions to the constraint set currently being tested, and (ii) identification, which may additionally name a violated constraint. We prove high-probability coverage bounds with dependence on the candidate class size $H$ for TRACE-Falsification. With reliable identification, TRACE-Identification can replace this dependence by $K/p_{\mathrm{ext}}$, where $K$ is the number of latent constraints and $p_{\mathrm{ext}}$ lower-bounds the probability of extracting a missing true constraint from informative feedback. We evaluate TRACE across six language-feedback tasks. On RecMovie, TRACE-Identification achieves 73% and 86% final-output success under caps of 20 and 60 evaluated outputs, compared with at most 42% and 48% for the evaluated prompting baselines given the same feedback and output caps. Controlled identity-corruption experiments further show greater robustness than direct accumulation when the falsification detector remains reliable.
Chinese Translation
在交互学习中，自然语言反馈通常通过指向违反的要求来解释某个行动失败的原因。误解这些反馈可能导致智能体排除有效的解决方案。我们通过将用户意图建模为动作空间上的潜在约束，研究这一情境，并将从语言反馈中学习表述为对可行区域的纯探索。我们提出了TRACE算法，该算法将候选约束组织成树形结构，并通过生成满足这些约束的动作来测试每个提议的细化。TRACE仅在经过多次测试后，结果反馈与细化不矛盾时才会承诺该细化。我们区分了使用相同反馈的两种方式：(i) 伪造，检测当前正在测试的约束集的矛盾，(ii) 识别，可能还会命名一个被违反的约束。我们证明了TRACE-伪造的高概率覆盖界限与候选类大小$H$有关。在可靠的识别下，TRACE-识别可以将这一依赖关系替换为$K/p_{ ext{ext}}$，其中$K$是潜在约束的数量，$p_{ ext{ext}}$是从信息反馈中提取缺失真实约束的概率的下界。我们在六个语言反馈任务中评估了TRACE。在RecMovie上，TRACE-识别在20和60个评估输出的限制下分别达到了73%和86%的最终输出成功率，而在相同反馈和输出限制下，评估的提示基线最多仅为42%和48%。控制身份损坏实验进一步表明，当伪造检测器保持可靠时，TRACE表现出比直接累积更大的鲁棒性。
cs.AI / 29 / 2610.09112
GeoNatureAgent (GNA): A Framework and Benchmark for Pre-Production Evaluation of Tool-Using Agents on Geospatial and Environmental Tasks
GeoNatureAgent (GNA)：用于工具使用代理在地理空间和环境任务预生产评估的框架与基准
Abstract
Before tool-using LLM agents are deployed in environmental and geospatial workflows, teams need evidence that an agent reliably selects the right operations against real APIs. We introduce GeoNatureAgent (GNA), a framework for pre-production evaluation of tool-using agents: a fixed sixteen-tool geospatial interface published as a Model Context Protocol (MCP) server, so the agent under test is the only variable, scored against an identical tool layer, task suite, and deterministic scorer. Its flagship instance is a 103-task benchmark (a 93-task main suite across 18 categories plus a ten-task comparison expansion) evaluated against an open, self-hostable geospatial API serving three environmental indicators across Spain and Portugal. We evaluate nine LLMs under three temperature-1.0 seeds, reporting capability and per-case cost as orthogonal axes. (1) Claude Sonnet 4 achieves the highest capability (61.7% +/- 0.7% on all 103 tasks; 60.8% on the main suite), followed closely by DeepSeek V3.2 (57.9%), while no other model exceeds 53%; (2) the cost-accuracy Pareto frontier is mostly open-weight, with DeepSeek V3.2 offering 93% of Claude's capability at 11.3x lower list-price cost; (3) under strict all-checks scoring the best model sits 24-36 points below the 85-97% reported on general-purpose GIS benchmarks, whereas per-check partial credit for the top four models (86-90%) is comparable, so much of that gap reflects scoring strictness rather than task difficulty alone. The MCP server, evaluation harness, benchmark, and API are publicly available; swapping the tool executors and task suite instantiates an equivalent benchmark for any geospatial domain.
Chinese Translation
在将工具使用的LLM代理部署到环境和地理空间工作流之前，团队需要证据证明代理能够可靠地针对真实API选择正确的操作。我们介绍了GeoNatureAgent (GNA)，这是一个用于工具使用代理的预生产评估框架：一个固定的十六工具地理空间接口，作为模型上下文协议（Model Context Protocol, MCP）服务器发布，因此被测试的代理是唯一的变量，其得分基于相同的工具层、任务套件和确定性评分器。其旗舰实例是一个103任务的基准（涵盖18个类别的93任务主套件以及一个10任务的比较扩展），评估对象是一个开放的、自托管的地理空间API，提供西班牙和葡萄牙的三个环境指标。我们在三个温度为1.0的种子下评估了九个LLM，报告能力和每个案例的成本作为正交轴。（1）Claude Sonnet 4在所有103个任务中实现了最高能力（61.7% +/- 0.7%；在主套件中为60.8%），紧随其后的是DeepSeek V3.2（57.9%），而其他模型均未超过53%；（2）成本-准确性帕累托前沿大多为开放权重，DeepSeek V3.2以11.3倍更低的标价成本提供了Claude能力的93%；（3）在严格的全检查评分下，最佳模型的得分比在通用GIS基准上报告的85-97%低24-36分，而前四个模型的每检查部分得分（86-90%）是可比的，因此这一差距的很大一部分反映了评分的严格性，而不仅仅是任务的难度。MCP服务器、评估工具、基准和API都是公开可用的；更换工具执行器和任务套件可以为任何地理空间领域实例化一个等效的基准。
cs.AI / 30 / 2610.09115
From Uncertainty to Action: Learning to Steer LLM Agents
从不确定性到行动：学习引导大型语言模型代理
Abstract
Steering an LLM agent means deciding whether to correct it, at which step, and with which mechanism. Uncertainty is often used to decide when to correct an agent, but whether it can guide these decisions remains unclear. We steer agent trajectories separately at every non-terminal step with each of four mechanisms and run each continuation to completion. The resulting stepwise outcome table (SOT) holds about 82,000 counterfactual continuations of 1,864 trajectories from three benchmarks and two agents. It shows that uncertainty can identify failing trajectories, but that no single signal reliably locates the step at which steering helps. We therefore propose VoS (Value of Steering), a trajectory-level monitor, offline or online, that learns from SOT the value of steering at each step and decides where to steer by it. A harm-budgeted trigger decides whether to steer, limiting the fraction of successful trajectories that VoS disturbs. VoS improves on unmodified execution in all 12 settings of benchmark, agent, and offline or online use, by 7.8 points on average, and outperforms the strongest of five existing uncertainty-triggered methods in 11, by 2.9 points on average. Ablations show that training on measured outcomes and a tight harm budget are both essential.
Chinese Translation
引导大型语言模型（LLM）代理意味着决定是否对其进行纠正、在何时纠正以及使用何种机制。通常不确定性被用来决定何时纠正代理，但它是否能够指导这些决策仍不清楚。我们在每个非终止步骤分别使用四种机制引导代理轨迹，并将每个延续运行到完成。结果的逐步结果表（SOT）包含来自三个基准和两个代理的1,864条轨迹的约82,000个反事实延续。结果表明，不确定性可以识别失败的轨迹，但没有单一信号能够可靠地定位引导有帮助的步骤。因此，我们提出了VoS（引导价值），一个轨迹级监控器，可以离线或在线使用，它从SOT中学习每一步的引导价值，并据此决定引导的位置。一个基于伤害预算的触发器决定是否进行引导，限制VoS干扰的成功轨迹的比例。在所有12种基准、代理和离线或在线使用的设置中，VoS在未修改执行的基础上平均提高了7.8分，并在11种设置中超越了五种现有的不确定性触发方法中的最强者，平均提高了2.9分。消融实验表明，基于测量结果的训练和严格的伤害预算都是必不可少的。
cs.AI / 31 / 2610.09119
Which Buildings Are Artificial Intelligence-Ready? A Measurement-Based Assessment Framework for AI Question Answering and Actuation
哪些建筑适合人工智能？基于测量的人工智能问答与控制评估框架
Abstract
Agentic artificial intelligence (AI) systems are becoming the interface to buildings, answering questions and controlling operations, but a building's readiness for them has not been systematically assessed. This study proposes a framework to quantify it. First, a building's knowledge graph sets two ceilings. The answerable-readiness ceiling is the share of operational questions its data could answer, and the actuation-readiness ceiling is the share of control actions it exposes. Second, a reference AI agent's accuracy on a fixed set of these questions shows how much of the ceilings is realized. On a simulated office, the agent realizes 0.62 of a 0.64 answerable ceiling, so missing data, not the AI, limit readiness, except in naming a fault's cause. Across 37 public real-building graphs, the median answerable ceiling is 0.16, and in 15 of 45, unlinked sensors lower it. The framework turns "is this building AI-ready?" into an auditable, ranked retrofit question.
Chinese Translation
主动型人工智能（AI）系统正成为建筑的接口，负责回答问题和控制操作，但建筑对这些系统的适应性尚未得到系统评估。本研究提出了一个量化框架。首先，建筑的知识图谱设定了两个上限。可回答准备上限是其数据能够回答的操作性问题的比例，而控制准备上限是其暴露的控制行为的比例。其次，参考AI代理在固定问题集上的准确性显示了上限的实现程度。在一个模拟办公室中，该代理实现了0.62的0.64可回答上限，因此限制适应性的因素是缺失的数据，而非AI本身，除了在命名故障原因时。通过对37个公共真实建筑图谱的分析，中位数可回答上限为0.16，在45个案例中的15个中，未连接的传感器降低了这一比例。该框架将“这座建筑是否适合人工智能？”转变为一个可审计的、分级的改造问题。
cs.AI / 32 / 2610.09127
CADFather: Autonomous CAD Reconstruction through Coordinated Tool Use
CADFather：通过协调工具使用实现自主CAD重建
Abstract
Reconstructing an editable CAD model from a 3D shape remains a challenging engineering task. Existing methods can propose CAD operations, but no single source of proposals works equally well across different part geometries and stages of reconstruction. We introduce CADFather, an autonomous agentic system that coordinates complementary tools to recover parametric CAD programs from 3D meshes. A vision-language assistant inspects renders of the target and intermediate reconstructions, then decides which candidate CAD programs to extend, which tools to invoke, how many proposals to generate, and when to finish. Learned and algorithmic tools propose CAD operations, while numerical optimization refines the parameters of existing programs. Proposed or refined programs are executed and evaluated to provide feedback for subsequent decisions. The agent maintains alternative candidate programs for each target part and preserves the best valid result throughout reconstruction. CADFather uses pretrained generation and assistant models without additional training. We evaluate reconstruction quality and execution validity on the full DeepCAD, Fusion360, and MCB test sets, as well as on CADENA-Bench, CADBench, and BenchCAD. We additionally analyze computational cost and the trade-off between cost and reconstruction quality.
Chinese Translation
从3D形状重建可编辑的CAD模型仍然是一项具有挑战性的工程任务。现有方法可以提出CAD操作，但没有单一的提案来源能够在不同的部件几何形状和重建阶段中同样有效。我们引入了CADFather，一个自主代理系统，通过协调互补工具从3D网格中恢复参数化CAD程序。视觉-语言助手检查目标和中间重建的渲染图像，然后决定扩展哪些候选CAD程序、调用哪些工具、生成多少提案以及何时结束。学习和算法工具提出CAD操作，而数值优化则细化现有程序的参数。提出或细化的程序被执行和评估，以提供后续决策的反馈。该代理为每个目标部件维护替代候选程序，并在整个重建过程中保留最佳有效结果。CADFather使用预训练的生成和助手模型，无需额外训练。我们在完整的DeepCAD、Fusion360和MCB测试集上评估重建质量和执行有效性，以及在CADENA-Bench、CADBench和BenchCAD上进行分析。此外，我们还分析了计算成本以及成本与重建质量之间的权衡。
cs.AI / 33 / 2610.09139
Breaking the Space Barrier and its Application to Language Model Inference
突破空间障碍及其在语言模型推理中的应用
Abstract
Language models are more and more often asked for structured output: JSON that follows a schema, or a tool call with typed arguments. A small machine, an automaton, enforces the format by forbidding the tokens that would break it. We observe that this machine has a rare property: from any of its states, each token leads along exactly one path. Graphs in which only a few paths join any two points are a classical object of complexity theory, and our theoretical result settles an open question about them: one can decide whether such a graph connects two points while verifying that it really has few paths, with very little memory. Precisely, the problem lies in the classes ReachUL, LOGDCFL, C=L and SC2, and needs only O(log2 n/ log log n) space, below the classical O(log2 n) of Savitch's theorem. The constructions behind the proofs become an inference engine: text the format forces is written without running the model, the mask is recomputed on the GPU without any table, recursive formats use a small stack, every output stays valid under a token limit, and independent fields are decoded in parallel and verified. On one 16 GB Apple M2 Pro with Qwen3.5-2B and 4B, against MLX with llguidance, the standard setup for this hardware, schema-constrained extraction finishes 1.2- 1.3x sooner with the same answers, a grammar costs 3 MB instead of up to 1.5 GB, one server holds sixteen grammars where tables run out of memory, and sixteen tool-calling agents finish 2.5x sooner.
Chinese Translation
语言模型越来越多地被要求输出结构化结果：遵循模式的 JSON，或带有类型参数的工具调用。一个小型机器，即自动机，通过禁止会破坏格式的符号来强制执行该格式。我们观察到这个机器具有一个罕见的特性：从它的任何状态出发，每个符号都沿着恰好一条路径前进。仅有少数路径连接任意两点的图是复杂性理论中的经典对象，我们的理论结果解决了一个关于它们的未解问题：可以在验证图确实具有少量路径的同时，决定该图是否连接两点，且只需非常少的内存。具体而言，该问题属于类 ReachUL、LOGDCFL、C=L 和 SC2，仅需 O(log² n / log log n) 的空间，低于 Savitch 定理的经典 O(log² n)。证明背后的构造成为一个推理引擎：格式强制下写出的文本无需运行模型，掩码在 GPU 上重新计算，无需任何表格，递归格式使用小栈，每个输出在符号限制下保持有效，独立字段并行解码并验证。在一台配备 Qwen3.5-2B 和 4B 的 16 GB Apple M2 Pro 上，与使用 llguidance 的 MLX 标准设置相比，受模式约束的提取完成时间缩短了 1.2-1.3 倍且答案相同，语法占用 3 MB，而非高达 1.5 GB，一台服务器可容纳十六个语法，而表格则超出内存限制，十六个工具调用代理的完成时间缩短了 2.5 倍。
cs.AI / 34 / 2610.09142
Finding Blind Spots in AppWorld and WorkArena Task Verifiers
发现 AppWorld 和 WorkArena 任务验证器中的盲点
Abstract
Execution-based task verifiers decide whether an agent succeeded. We audit shipped AppWorld and WorkArena verifiers with source-informed mutation tests. The main audit never modifies a shipped checker. In AppWorld, duplicating a non-idempotent write creates an extra record while preserving every checked field value. The verifier accepts all three task variants from two of five eligible generators: 6/15 constructed effects. A cardinality patch applied to checker copies after the census makes all six cells fail while preserving valid controls. In WorkArena, we prospectively rerun 23 extra-field candidates selected for earlier checker-PASS outcomes. Independent Table API readback confirms nondefault persisted values in 21, while all 23 receive PASS. Two requested strings are aliases of stored defaults. The 21 confirmed wrong effects span three form templates. These selected cases confirm wrong effects under the audit's protocol; they do not estimate a population rate. No other construction produces an independently confirmed false accept. Other checker-PASS cases are effect-correct degeneracies. We report zero-PASS families separately because retained evidence differs. In fixed intent-swap grids, the checkers return no PASS on 2,689 off-diagonal executions. This is a rejection census: 57 WorkArena cells use session-scoped evidence; the other 2,632 lack classified rejection causes and independent target ground truth. Each increment is specified before its own cells are scored. A supplement accompanies the OpenReview submission with the construction grammar, evidence, content-bound stage lineage and count reproducer.
Chinese Translation
基于执行的任务验证器决定代理是否成功。我们通过源信息驱动的变异测试对已发布的 AppWorld 和 WorkArena 验证器进行审计。主要审计从不修改已发布的检查器。在 AppWorld 中，重复一个非幂等写入会创建一个额外记录，同时保留每个检查字段的值。验证器接受来自五个合格生成器中的两个的三种任务变体：构造效果的比例为 6/15。应用于普查后检查器副本的基数补丁使所有六个单元失败，同时保留有效的控制。在 WorkArena 中，我们前瞻性地重新运行 23 个额外字段候选项，这些候选项是根据早期检查器通过结果选择的。独立的表 API 读回确认 21 个非默认持久值，而所有 23 个都获得通过。两个请求的字符串是存储默认值的别名。这 21 个确认的错误效果跨越三个表单模板。这些选定的案例在审计协议下确认了错误效果；它们并不估计总体率。没有其他构造产生独立确认的错误接受。其他检查器通过案例是效果正确的退化情况。我们单独报告零通过族，因为保留的证据不同。在固定意图交换网格中，检查器在 2,689 次非对角执行中没有返回通过。这是一个拒绝普查：57 个 WorkArena 单元使用会话范围的证据；其余 2,632 个缺乏分类拒绝原因和独立目标真实值。每个增量在其单元评分之前被指定。一个补充材料随 OpenReview 提交，包含构造语法、证据、内容绑定阶段谱系和计数重现器。
cs.AI / 35 / 2610.09144
DIVA: Dual-Space Intent-Aware Visual Attenuation for Vision-Language-Action Policies
DIVA：双空间意图感知视觉衰减用于视觉-语言-动作策略
Abstract
Vision-language-action (VLA) policies typically feed dense visual patch tokens into a language-action backbone, preserving scene context but offering no explicit mechanism to regulate how strongly different visual tokens influence policy computation. We introduce DIVA, a Dual-Space Intent-Aware Visual Attenuation module with an anchor-then-attenuate design. DIVA combines high-level task intent with low-level visual evidence to estimate patch-wise relevance anchors, then applies them in two complementary spaces: it reweights projected visual tokens before backbone entry and persistently attenuates low-relevance visual states within the backbone. DIVA preserves the full visual token sequence and requires no external grounding supervision. On LIBERO, DIVA improves OpenVLA-OFT from 96.6% to 98.0% average success and raises its zero-shot LIBERO-Plus score from 69.6 to 72.6. Real-world experiments further show consistent gains under task-irrelevant visual perturbations, supporting the robustness of intent-aware visual attenuation beyond simulation.
Chinese Translation
视觉-语言-动作（VLA）策略通常将密集的视觉补丁标记输入到语言-动作主干中，保留场景上下文，但没有明确的机制来调节不同视觉标记对策略计算的影响程度。我们提出了DIVA，一个具有锚点-再衰减设计的双空间意图感知视觉衰减模块。DIVA结合了高层次的任务意图和低层次的视觉证据，以估计补丁级相关性锚点，然后在两个互补的空间中应用它们：在主干输入之前重新加权投影的视觉标记，并在主干内持续衰减低相关性的视觉状态。DIVA保留了完整的视觉标记序列，并且不需要外部的基础监督。在LIBERO数据集上，DIVA将OpenVLA-OFT的平均成功率从96.6%提高到98.0%，并将其零样本LIBERO-Plus得分从69.6提高到72.6。现实世界的实验进一步表明，在与任务无关的视觉扰动下，DIVA表现出一致的提升，支持了意图感知视觉衰减在模拟之外的鲁棒性。
cs.AI / 36 / 2610.09146
Frozen Models, Evolving Expertise: Model-Agnostic Learning from Deployment Experience for Multimodal Medical AI
冻结模型，演变的专业知识：基于部署经验的模型无关学习在多模态医学人工智能中的应用
Abstract
Large language models (LLMs) and vision-language models (VLMs) are usually frozen after deployment, so they do not learn from the cases they solve. This is especially concerning in medicine, where new clinical evidence, updated guidelines, and new therapies can change established practice. Fine-tuning can update the model, but it requires access to model weights and additional training. Parameter-free methods avoid training, but they may overfit a fixed validation set, lack reliable domain knowledge, or lose visual details by saving experience only as text. To address these limitations, we present a model-agnostic framework that allows frozen LLMs and VLMs to learn from deployment experience through three forms of external expertise: a Skill that guides reasoning and tool use, a Knowledge Memory that stores reliable facts supported by earlier cases or trusted external evidence, and a Multimodal Knowledge Base that keeps visual examples and guides the model to relate each retrieved case to the current image. Instead of relying on a fixed validation set, a validation strategy keeps an update only if it helps on new cases without degrading performance on earlier ones. Across six benchmarks covering clinical diagnosis, clinical workflows, medical reasoning, and medical and non-medical visual reasoning, and with four open-weight and closed-source base models, our framework improves performance during online deployment by up to 34.2% over the base model on medical tasks, generalizes to unseen cases, transfers to other models without further optimization, and works in non-medical domains.
Chinese Translation
大型语言模型（LLMs）和视觉语言模型（VLMs）在部署后通常是冻结的，因此它们无法从所解决的案例中学习。这在医学领域尤为令人担忧，因为新的临床证据、更新的指南和新疗法可能会改变既定的实践。微调可以更新模型，但需要访问模型权重和额外的训练。无参数方法避免了训练，但可能会对固定的验证集过拟合，缺乏可靠的领域知识，或通过仅将经验保存为文本而丢失视觉细节。为了解决这些局限性，我们提出了一种模型无关的框架，允许冻结的LLMs和VLMs通过三种外部专业知识形式从部署经验中学习：一种指导推理和工具使用的技能（Skill），一种存储由早期案例或可信外部证据支持的可靠事实的知识记忆（Knowledge Memory），以及一个保持视觉示例并指导模型将每个检索案例与当前图像关联的多模态知识库（Multimodal Knowledge Base）。我们的验证策略不是依赖于固定的验证集，而是仅在帮助新案例的情况下进行更新，而不降低早期案例的性能。在涵盖临床诊断、临床工作流程、医学推理以及医学和非医学视觉推理的六个基准测试中，使用四个开放权重和闭源基础模型，我们的框架在医学任务上在在线部署期间相较于基础模型提高了多达34.2%的性能，能够推广到未见过的案例，能够转移到其他模型而无需进一步优化，并且在非医学领域也能有效工作。
cs.AI / 37 / 2610.09159
SpecGuard: Proving a Task Is Broken Before the Agent Cheats
SpecGuard：在代理作弊之前证明任务是错误的
Abstract
As autonomous coding agents get increasingly deployed, the risk that accidental or adversarially injected misspecifications in tasks lead to dangerous agent behavior is critical to address. Prior work has shown that agents given such tasks rarely flag the conflict and instead cheat, editing tests or hard-coding expected outputs, and the actions taken to cheat can cause real damage, such as deleting a security defense to make a corrupted test pass. It remains unclear whether such conflicts can be established with independently verifiable evidence before the agent acts. We present SpecGuard, which detects and formally certifies these conflicts between task intent and tests. Given only the task description and codebase, SpecGuard autoformalizes the intended behaviour into a Lean 4 specification. The tests are formalized independently, and the Lean kernel checks whether any implementation could satisfy both formalizations, producing a machine-checked certificate when none can. On conflicted SWE-bench tasks, SpecGuard detects up to 72.8% of conflicts and formally certifies up to 51.1%, with a nearly five-fold lower conflict miss rate than model-based judgment. SpecGuard provides a pre-execution safety check that identifies reward-hacking opportunities through formal certification of task-level conflicts, before any agent behavior is observed. Our code is available at https://github.com/prmbiy/specguard.
Chinese Translation
随着自主编码代理的广泛应用，任务中意外或恶意注入的错误规范导致代理行为危险的风险亟需解决。先前的研究表明，面对这些任务的代理很少会标记出冲突，而是选择作弊，编辑测试或硬编码预期输出，而作弊所采取的行动可能造成实际损害，例如删除安全防御以使损坏的测试通过。目前尚不清楚在代理采取行动之前，是否可以通过独立可验证的证据建立这些冲突。我们提出了SpecGuard，它检测并正式认证任务意图与测试之间的这些冲突。仅凭任务描述和代码库，SpecGuard将预期行为自动形式化为Lean 4规范。测试独立形式化，Lean内核检查是否有任何实现可以同时满足这两种形式化，当没有实现可以满足时，生成机器检查的证书。在存在冲突的SWE-bench任务中，SpecGuard检测到高达72.8%的冲突，并正式认证高达51.1%，其冲突漏报率比基于模型的判断低近五倍。SpecGuard提供了一种执行前的安全检查，通过正式认证任务级冲突识别奖励操控机会，在观察到任何代理行为之前。我们的代码可在https://github.com/prmbiy/specguard获取。
cs.AI / 38 / 2610.09164
Training Language Models To Be Coherent Decision-Makers
训练语言模型以成为连贯的决策者
Abstract
Reliable decision-making requires more than accurate prediction: a model must preserve its beliefs, apply the relevant utilities, and recognize when the information needed to justify an action is missing. We study whether language models can learn this decision procedure from supervised fine-tuning and generalize it across domains and differing natural-language expressions of the decision challenge. Across 20 datasets, we explore challenges of belief instability and decision-making errors by first eliciting probabilities of outcomes and then varying only the utilities and the framing of the decision problems, while holding the evidence fixed. We train models to preserve elicited beliefs while selecting the action that maximizes expected utility, and evaluate transfer to unseen application domains, held-out framings, and different classes of payoff structures. We further introduce incomplete-information settings in which required utilities are withheld and replaced with irrelevant text, testing whether models can distinguish missing decision-relevant information from merely additional context. We find that targeted fine-tuning substantially improves coherent decision-making and that in many situations, learning transfers across domains and framings to situations unobserved during training. Further, models trained for decidability learn to identify when action cannot be justified based on missing information. Finally, we show the value of a routed system that considers separately the recognition of decision completeness and utility-sensitive decision execution.
Chinese Translation
可靠的决策不仅需要准确的预测：模型必须保持其信念，应用相关的效用，并识别何时缺少必要的信息以证明某一行动的合理性。我们研究语言模型是否能够通过监督微调学习这一决策过程，并在不同领域和不同自然语言表达的决策挑战中进行泛化。在20个数据集中，我们首先引出结果的概率，然后仅改变效用和决策问题的框架，同时保持证据不变，以探讨信念不稳定性和决策错误的挑战。我们训练模型以保持引出的信念，同时选择最大化期望效用的行动，并评估其在未见过的应用领域、保留的框架和不同类型的收益结构中的迁移能力。我们进一步引入不完全信息设置，在这些设置中，所需的效用被隐瞒并替换为无关的文本，测试模型是否能够区分缺失的决策相关信息与仅仅是额外的上下文。我们发现，针对性的微调显著提高了连贯的决策能力，并且在许多情况下，学习能够跨领域和框架迁移到训练期间未观察到的情境。此外，经过可判定性训练的模型能够识别何时基于缺失信息无法证明行动的合理性。最后，我们展示了一种路由系统的价值，该系统分别考虑决策完整性和效用敏感的决策执行的识别。
cs.AI / 39 / 2610.09188
From Probabilities to Decisions: Search and Multi-Teacher Distillation with Jev
从概率到决策：使用Jev的搜索与多教师蒸馏
Abstract
Probability-only models, which TypeSafe calls System One models, return calibrated probabilities for fixed choices in milliseconds and generate no text. We study one such model, Jev, through two tasks that require decisions under tight constraints. In bullet chess, a bot that places Jev's judgment inside Stockfish search alongside an opening book and endgame tablebases climbs above a 2200 Lichess bullet rating against other bots. Live model calls are too slow for search, so we distill pairwise judgments into a compact evaluator that runs at every position. We then ask how best to spend a fixed labeling budget when an LLM, Qwen3-32B, is available as a second teacher. In chess, averaging both judges' labels beats spending the whole budget on Qwen alone by 9.6 Elo (95% interval 4.3 to 14.9), and the gain replicates on fresh openings; a second answer from the same judge is no substitute, and Jev is the strongest partner for Qwen among the models tested. In passage reranking, Jev's labels alone train a reranker that scores as high as Qwen's, from 21 minutes of API calls instead of 5.1 GPU-hours, and adding Qwen gains at most a few thousandths in ranking quality. Search supplies the lookahead, distillation makes the judgment cheap enough to use at every position, and an LLM partner pays off in chess.
Chinese Translation
概率模型，仅返回固定选择的校准概率，TypeSafe称之为系统一模型，响应时间为毫秒且不生成文本。我们通过两个在紧约束下需要决策的任务研究了这样一个模型Jev。在子弹棋中，将Jev的判断置于Stockfish搜索中，并结合开局书和残局表，能够在与其他机器人对战中达到超过2200的Lichess子弹评分。实时模型调用对于搜索来说过于缓慢，因此我们将成对判断蒸馏为一个紧凑的评估器，使其能够在每个位置运行。接着，我们探讨在可用的第二教师LLM Qwen3-32B的情况下，如何最佳利用固定的标注预算。在国际象棋中，平均两个评审的标签比将全部预算花费在Qwen上高出9.6 Elo（95%置信区间为4.3到14.9），这一增益在新开局上得到了重复；来自同一评审的第二个答案并不能替代，而在测试的模型中，Jev是Qwen最强的合作伙伴。在段落重排序中，仅使用Jev的标签训练出的重排序器，其得分与Qwen相当，所需的API调用时间为21分钟，而不是5.1 GPU小时，且添加Qwen最多只能在排名质量上提升几千分。搜索提供了前瞻性，蒸馏使得判断在每个位置上都足够便宜，而LLM合作伙伴在国际象棋中是有回报的。
cs.AI / 40 / 2610.09193
Bookkeeping, Composition, or Unreachable Gold? Reading MemoryAgentBench's Conflict-Resolution Scores Against a Frozen Last-Write Resolver
记账、组合还是无法触及的黄金？阅读 MemoryAgentBench 的冲突解决分数与冻结的最后写入解析器的对比
Abstract
MemoryAgentBench's Conflict Resolution split is read as measuring "selective forgetting". We execute the benchmark's own rule - the newest statement about a fact wins - as a zero-learning resolver frozen on one of the four fact lists. Under the official metric the rule answers 80.25% of the questions (74.5% on the three held-out lists). Of the rest, 67 items have a released gold that the last-write graph cannot reach but overwritten statements would ("The capital of India is New Delhi." superseded by "The capital of India is Grosseto."; gold New Delhi); such items are a third of the multi-hop questions at 262K. Two long-context models and our pre-registered approximate re-implementation of the benchmark's BM25 agent, one retained run per item and outcomes only, score 84.7%, 82.6% and 41.6% on the items the rule solves against 10.4%, 11.9% and 6.0% on those 67. The failures are a reachability split plus a small parser-scope residual; the per-item split, not the aggregate, is the unit at which a score here can be read.
Chinese Translation
MemoryAgentBench 的冲突解决分被解读为测量“选择性遗忘”。我们执行基准测试的自身规则——关于事实的最新陈述获胜——作为一个在四个事实列表之一上冻结的零学习解析器。在官方指标下，该规则回答了 80.25% 的问题（在三个保留列表上为 74.5%）。其余的 67 项中，有一部分发布的黄金是最后写入图无法触及的，但被覆盖的陈述可以触及（“印度的首都为新德里。”被“印度的首都为格罗塞托。”所取代；黄金为新德里）；这些项目占 262K 多跳问题的三分之一。两个长上下文模型和我们预注册的基准测试 BM25 代理的近似重新实现，每个项目保留一次运行，仅计算结果，在该规则解决的项目上得分为 84.7%、82.6% 和 41.6%，而在这 67 项上得分为 10.4%、11.9% 和 6.0%。失败的原因是可达性分裂加上一个小的解析器范围残余；每个项目的分裂，而非总和，是在此处可以解读分数的单位。
cs.AI / 41 / 2610.09197
FreeEvolve: Learning to Evolve Beyond Fixed Loops
FreeEvolve：学习超越固定循环的进化
Abstract
Agent evolvers automate the design of the prompts, skills and workflows around language model agents, yet the optimization process they follow is still designed by hand: a fixed search loop decides how candidates are evaluated, which are kept and when the search stops. We propose FREEEVOLVE, which automates this process as well. An environment specifies the goal, target agent, evaluator, data and resource limits; within these limits, the evolver itself decides what to test, how much evidence to collect, which candidates to pursue and when to stop. These decisions follow an editable evolution skill, which we improve through meta-evolution by scoring each candidate skill on the fresh target agent it produces. The optimization process thus becomes a capability learned from experience rather than a loop engineered in advance. On tau3-bench, ARC-AGI-2, ARC-AGI-3 and Terminal-Bench 2.1, FREEEVOLVE controls the evolution campaign by itself, yet improves the primary held-out metric by 13.6 points on average and matches or exceeds hand-designed evolvers. The learned process keeps improving with experience: meta-evolved skills add 6.9 points over the seed skill on fresh target agents, demonstrating transferability across environments.
Chinese Translation
代理进化器自动化设计围绕语言模型代理的提示、技能和工作流程，然而它们遵循的优化过程仍然是手动设计的：一个固定的搜索循环决定了候选者的评估方式、保留哪些候选者以及何时停止搜索。我们提出了FREEEVOLVE，它也自动化了这一过程。一个环境指定了目标、目标代理、评估者、数据和资源限制；在这些限制内，进化器本身决定测试什么、收集多少证据、追求哪些候选者以及何时停止。这些决策遵循一个可编辑的进化技能，我们通过元进化来改进它，通过对每个候选技能在其产生的新目标代理上的评分来实现。因此，优化过程变成了一种从经验中学习的能力，而不是预先设计的循环。在tau3-bench、ARC-AGI-2、ARC-AGI-3和Terminal-Bench 2.1上，FREEEVOLVE独立控制进化活动，平均提高了主要保留指标13.6分，并且与手动设计的进化器相匹配或超越。学习到的过程随着经验的积累不断改进：元进化技能在新目标代理上比种子技能增加了6.9分，展示了跨环境的可转移性。
cs.AI / 42 / 2610.09202
Few Bits, One Law: Toward W2A4KV2
少量位数，一条法则：迈向 W2A4KV2
Abstract
Extreme low-bit LLM compression is most challenging when weights, activations, and KV caches are quantized together: their distributions differ, and quantization errors interact throughout the network. We introduce CanonQ, a unified quantization-aware training framework that addresses these challenges by separating source canonicalization from task-aware adaptation. Fixed rotations and energy normalization map heterogeneous tensor sources to canonical coordinates, enabling frozen Gaussian-reference codebooks to be reused across layers and models. Joint training then adapts the network to the coupled errors of weight, activation, and cache quantization within a common scalar/vector interface. We bound frozen-codebook transfer error and local task loss, and derive an exact normalization-aware straight-through Jacobian that links quantization distortion to gradient bias. The strongest gains arise under joint W2A4KV2 compression: across LLaMA3-1B/3B/8B, CanonQ-Omni achieves up to 14.28x lower WikiText-2 perplexity and up to 57.9% higher mean zero-shot accuracy than prior state-of-the-art and representative quantization baselines. The benefits extend to Qwen3-1.7B, code generation, and mathematical reasoning: on instruction-tuned MobileLLM-Pro-1B at W2A16KV16, CanonQ achieves relative improvements of 41.7% in HumanEval pass@1 and 39.1% in GSM8K exact match over the strongest evaluated quantization baseline.
Chinese Translation
极低位数的 LLM 压缩在权重、激活和 KV 缓存共同量化时最具挑战性：它们的分布不同，量化误差在整个网络中相互作用。我们提出了 CanonQ，一个统一的量化感知训练框架，通过将源规范化与任务感知适应分离来解决这些挑战。固定旋转和能量归一化将异构张量源映射到规范坐标，使得冻结的高斯参考代码本可以在层和模型之间重复使用。联合训练随后适应网络，以应对权重、激活和缓存量化的耦合误差，通过一个共同的标量/向量接口。我们界定了冻结代码本传输误差和局部任务损失，并推导出一个精确的归一化感知直通雅可比矩阵，将量化失真与梯度偏差联系起来。在联合 W2A4KV2 压缩下，最强的增益出现：在 LLaMA3-1B/3B/8B 上，CanonQ-Omni 实现了高达 14.28 倍更低的 WikiText-2 困惑度和高达 57.9% 的更高平均零-shot 准确率，相较于之前的最先进技术和代表性量化基线。这些好处扩展到 Qwen3-1.7B、代码生成和数学推理：在指令调优的 MobileLLM-Pro-1B 上，W2A16KV16 下，CanonQ 在 HumanEval pass@1 上实现了 41.7% 的相对提升，在 GSM8K 精确匹配上实现了 39.1% 的相对提升，相较于最强的评估量化基线。
cs.AI / 43 / 2610.09215
AGAR: a reinforcement learning substrate for LLM program evolution
AGAR：用于大型语言模型程序演化的强化学习基础
Abstract
Given a task and an evaluator, a language model can rewrite a candidate program while a search loop decides which rewrites survive, offering a practical route to algorithm discovery. But that loop is governed by five constants set by hand: which parent to select, how hard to mutate, how to keep diversity, what to remember, and a scalar score that never says which part of the program earned it. Reinforcement learning already has an estimator for each. The obstacle is that program evolution is not usually written down as a decision process. We formalize it as a Markov decision process whose action is the modular prefix the model is conditioned on, rather than the program it emits. Credit assignment, value estimation, adaptive exploration, and experience memory can then attach to distinct components. AGAR (Algorithm Generation As RL) provides the resulting substrate: any estimator can be replaced or switched off without changing the controller, making the transfer auditable one mechanism at a time, with no gradient training of the backend model. Across 19 tasks, two backends, and three seeds under one harness, AGAR improves on the stronger of two published baselines on most tasks, with gains concentrated in the competitive-programming family. The formalization also yields a checkable reading of prior work: these systems are implicitly zero-discount, not by choice, but because fitness is exogenous to an individual rather than a return over successors, leaving a discount factor nothing to act on.
Chinese Translation
给定一个任务和一个评估器，语言模型可以重写候选程序，而搜索循环决定哪些重写能够存活，从而为算法发现提供了一条实用的途径。但这个循环由五个手动设置的常数控制：选择哪个父代、变异的难度、如何保持多样性、记住什么，以及一个从不说明程序哪个部分获得分数的标量评分。强化学习已经为每个常数提供了一个估计器。障碍在于程序演化通常并不是作为决策过程书写的。我们将其形式化为一个马尔可夫决策过程，其动作是模型所依赖的模块前缀，而不是它所产生的程序。信用分配、价值估计、自适应探索和经验记忆可以附加到不同的组件上。AGAR（Algorithm Generation As RL）提供了所需的基础：任何估计器都可以被替换或关闭，而无需更改控制器，从而使转移过程可以逐一审计，而无需对后端模型进行梯度训练。在19个任务、两个后端和三个种子下的一个框架中，AGAR在大多数任务上优于两个已发布基线中较强的一个，增益集中在竞争编程领域。该形式化还提供了对先前工作的可检查解读：这些系统隐含地是零折扣的，并非出于选择，而是因为适应性是外生于个体的，而不是对后继者的回报，从而使折扣因子无事可做。
cs.AI / 44 / 2610.09218
RLDISCOVER: LLM-driven co-evolution of reinforcement learning algorithms
RLDISCOVER：基于大型语言模型的强化学习算法协同进化
Abstract
LLM-guided program evolution has enabled discoveries in mathematics and computational optimization, raising the prospect of reinforcement learning (RL) algorithms that self-evolve to improve how agents learn. However, realizing this prospect faces two obstacles. Joint search over coupled algorithmic components is difficult to scale: simultaneous changes can disrupt learning, while isolated changes overlook their dependencies. Evaluating candidate algorithms also requires costly training, with fitness remaining uncertain across random seeds. We introduce RLDiscover, a framework for the self-evolution of model-free deep RL algorithms. Progressive Co-Evolution advances from targeted component edits to joint evolution, while Progressive Probabilistic Evaluation balances search breadth and evaluation fidelity through staged training and repeated evaluation. Experiments across SAC, PPO, and DQN on four benchmark suites show substantial improvements in mean return, with per-family median gains of 32%-84% and a peak return ratio of approximately 363x over a near-zero baseline. These gains include transitions from failed learning to successful task completion, and improvements persist when evolution starts from stronger open-source implementations. On measured SAC locomotion runs, evaluation uses approximately one-fifteenth the estimated compute required to fully evaluate the same candidate pool. Remarkably, independent searches repeatedly discover interpretable combinations of adaptive robust losses, progress-dependent value targets, and running statistics, with selected programs transferring to unseen tasks. These findings point toward a broader role for self-evolution in AI: discovering interpretable algorithms that improve how agents learn.
Chinese Translation
基于大型语言模型（LLM）的程序进化已经在数学和计算优化领域带来了新的发现，这为自我进化的强化学习（RL）算法提供了可能性，从而改善智能体的学习方式。然而，实现这一前景面临两个障碍。对耦合算法组件的联合搜索难以扩展：同时的变化可能会干扰学习，而孤立的变化则忽视了它们之间的依赖关系。评估候选算法还需要昂贵的训练，且在随机种子下适应度仍然不确定。我们提出了RLDiscover，这是一个用于无模型深度强化学习算法自我进化的框架。渐进式协同进化从针对性组件编辑推进到联合进化，而渐进式概率评估通过分阶段训练和重复评估平衡了搜索广度和评估准确性。在四个基准测试套件上对SAC、PPO和DQN的实验显示，平均回报有显著提升，每个家族的中位数增益为32%-84%，峰值回报比接近363倍，远高于近零的基线。这些增益包括从失败学习到成功任务完成的转变，并且当进化从更强的开源实现开始时，改进依然持续。在测量的SAC运动运行中，评估使用的计算量约为完全评估相同候选池所需计算量的十五分之一。值得注意的是，独立搜索反复发现了可解释的自适应鲁棒损失、进度依赖的价值目标和运行统计的组合，所选程序能够迁移到未见过的任务。这些发现指向自我进化在人工智能中的更广泛角色：发现可解释的算法，从而改善智能体的学习方式。
cs.AI / 45 / 2610.09219
CM-DPO: Constraint-Margin Direct Preference Optimization for LLM Planning
CM-DPO：用于大语言模型规划的约束边际直接偏好优化
Abstract
Direct Preference Optimization (DPO) treats all constraint violations equally: a $1 budget overshoot and a $1,000 overshoot induce the same training signal. It is also susceptible to length and style bias when preference pairs come from different model families. We introduce Constraint-Margin DPO (CM-DPO), which replaces DPO's binary preference signal with a continuous margin derived from a deterministic symbolic verifier and scaled by violation severity. Hard and soft constraints are separated through a lexicographic objective, ensuring hard constraints are never traded off against preferences. To supply CM-DPO with bias-reduced training pairs, we generate preference data through procedurally generated constraint profiles (DCCG) and minimal-edit distillation from a reasoning teacher (RT-MED), within a framework we call SynPlan-R. On TravelPlanner, NaturalPlan, and out-of-distribution PlanBench, an 8B model fine-tuned with CM-DPO achieves 89.2% pass rate and 93.4% solve rate, matching multi-agent systems at 13x lower latency while outperforming GPT-4o on unseen Blocksworld by 9.2 points.
Chinese Translation
直接偏好优化（Direct Preference Optimization, DPO）对所有约束违规行为一视同仁：$1的预算超支和$1,000的超支会产生相同的训练信号。当偏好对来自不同模型家族时，DPO也容易受到长度和风格偏见的影响。我们提出了约束边际DPO（Constraint-Margin DPO, CM-DPO），它用来自确定性符号验证器的连续边际替代了DPO的二元偏好信号，并根据违规严重性进行缩放。通过字典序目标将硬约束和软约束分开，确保硬约束不会与偏好相互权衡。为了为CM-DPO提供减少偏见的训练对，我们通过程序生成的约束配置文件（DCCG）和来自推理教师的最小编辑蒸馏（RT-MED）生成偏好数据，框架我们称之为SynPlan-R。在TravelPlanner、NaturalPlan和分布外的PlanBench上，经过CM-DPO微调的8B模型实现了89.2%的通过率和93.4%的解决率，在延迟降低13倍的情况下与多智能体系统相匹配，同时在未见过的Blocksworld上超越GPT-4o达9.2分。
cs.AI / 46 / 2610.09237
Trajectory Abstraction for the Science of Language Agent Behavior
语言代理行为的轨迹抽象
Abstract
Scientific studies of language agents need behavioral variables that support hypotheses across tasks and models. We formulate this research problem as learning and testing a hierarchy of trajectory abstractions. A concrete recursive procedure first measures role- and phase-indexed events, proposes temporally constrained relations, and tests their stability across conditions. It then constructs episode-level motif variables from selected relations and repeats the analysis on those variables. Explicit measurement functions connect every abstraction level to the original trajectories. Observations and randomized protocol experiments assess the resulting hypotheses, while comparisons between intervention realizations determine whether an abstraction should be retained, refined, or restricted. We derive a finite-depth bound for accepted reductions, identify protocol effects on fixed abstractions, and characterize realization disagreement and composition of abstraction error. A finite-sample test makes projected intervention consistency operational, and constructed examples illustrate motif construction and abstraction refinement. The formulation distinguishes this experimental approach from semantic taxonomies, qualitative theory induction, and behavior-model recovery. It specifies a proposed research procedure for discovering generalizable behavioral hypotheses, with literature-relative novelty assessed separately from model-relative surprise.
Chinese Translation
对语言代理的科学研究需要支持跨任务和模型假设的行为变量。我们将这一研究问题表述为学习和测试轨迹抽象的层次结构。一个具体的递归过程首先测量角色和阶段索引的事件，提出时间约束关系，并测试其在不同条件下的稳定性。然后，从选定的关系构建情节级别的主题变量，并对这些变量重复分析。显式测量函数将每个抽象层次与原始轨迹连接起来。观察和随机化协议实验评估所产生的假设，而干预实现之间的比较则决定了一个抽象是否应被保留、精炼或限制。我们推导出接受简化的有限深度界限，识别固定抽象上的协议效应，并表征实现不一致和抽象误差的组成。有限样本测试使得投影干预的一致性具备可操作性，构造的示例说明了主题构建和抽象精炼。该表述将这一实验方法与语义分类法、定性理论归纳和行为模型恢复区分开来。它具体规定了一种发现可推广行为假设的研究程序，并将文献相对的新颖性与模型相对的惊讶度分开评估。
cs.AI / 47 / 2610.09239
The Winner's Curse in LLM Self-Improvement Loops: Selection Noise, Lock-in, and Acceptance Rules
大模型自我改进循环中的赢家诅咒：选择噪声、锁定和接受规则
Abstract
Self-improving LLM systems propose changes to themselves and keep those that score better on a small evaluation set. We treat this keep-if-better step as selection under measurement noise, model the correlated errors of the candidates in a single decision, and study empirically what happens when the evaluation set is reused. In runs where Qwen models rewrite their own instructions and every candidate is also scored on 600 held-out items, most proposals after the first are harmful, and the model gives the size of the winner's curse of a generation's best candidate. With a prior from a separate pilot, it matches the average overstatement of first-generation commits in native loops, though not setting by setting. In a pre-registered study, the final selection-set score of greedy loops exceeded held-out accuracy by 13 to 20 points with 16 selection items and by 1 to 5 points with 256. Held-out gains grew with the selection set on TREC but not on GSM8K, and the tested acceptance rules did not beat greedy acceptance over whole runs. Gains measured on the selection set also exceeded held-out gains when a current model refined a competent instruction, and in the validation scores of GEPA and MIPROv2. Scoring the starting and the current instruction on 64 items never used for selection removes the average bias of a loop's reported gain, but single estimates remain off by about 6 points. Self-improvement studies should report held-out gains with their uncertainty.
Chinese Translation
自我改进的大语言模型（LLM）系统提出对自身的改动，并保留在小型评估集上得分更高的改动。我们将这一保留更好的步骤视为在测量噪声下的选择，建模候选者在单一决策中的相关错误，并实证研究当评估集被重复使用时发生的情况。在Qwen模型重写自身指令的实验中，每个候选者还在600个保留项目上进行评分，发现大多数在第一次之后的提议都是有害的，并且模型给出了这一代最佳候选者的赢家诅咒的规模。通过来自单独试点的先验信息，它与本土循环中第一代提交的平均夸大相匹配，尽管在不同设置中并不一致。在一项预注册研究中，贪婪循环的最终选择集得分超过了保留准确度13到20分（在16个选择项目中）和1到5分（在256个选择项目中）。在TREC上，保留增益随着选择集的增大而增长，但在GSM8K上则没有，测试的接受规则在整个运行过程中未能超越贪婪接受。当前模型在精炼有效指令时，选择集上的得分也超过了保留增益，并在GEPA和MIPROv2的验证得分中得到了体现。在64个从未用于选择的项目上对起始指令和当前指令进行评分，消除了循环报告增益的平均偏差，但单次估计仍然偏差约6分。自我改进研究应报告其不确定性的保留增益。
cs.AI / 48 / 2610.09243
We Query, Therefore We Compute: On Oracle Computation beyond the Machine, with an Application to Agents
我们查询，因此我们计算：超越机器的Oracle计算，及其在智能体中的应用
Abstract
Agentic systems use large language models (LLMs) to carry out concrete tasks. Prior work often borrows abstractions such as scheduling, caching or isolation piecemeal from operating systems, so the mechanisms it builds share little common ground, and the shared view of the two forms of agentic system, Workflows and Agents, is limited. We construct an abstract machine that provides both. We treat the LLM as an Oracle and extend a two-stack pushdown automaton with one instruction, which hands the Oracle a whole stack as its query and appends the answer to that same stack. The machine thus performs two computations, the Oracle's and a Turing-complete one that we call the Priestess. A stack that the program only appends to grows autoregressively, as an agent's context does. Two symmetry breakings, S in storage and T in transitions, make a Priestess program the operating system of the programs the Oracle runs, and produce the Agent and the Workflow as the two placements of a task's program. For internally autoregressive Oracles, the two computations synchronize at the end of every answer under certain conditions, and through that synchronization we model caching and analyse scheduling. No guarantee that holds for every Oracle can fix which content crosses between the two computations, but such a guarantee does fix the boundary itself. The construction V fits the machine to a von Neumann computer. To show that it is realizable, we propose ArchNights, an extended RISC-V ISA and a Linux-style operating system implementing the machine by design. ArchNights-SE runs on gem5 as a computer system, becomes an agentic system when it runs an LLM as the Oracle, and will be open source. Agentic systems can then be designed as computer systems are. With a foundation built and a unified view, future work can share invariants and bounds, each with its conditions.
Chinese Translation
智能体系统使用大型语言模型（LLMs）来执行具体任务。以往的研究通常从操作系统中零散借用调度、缓存或隔离等抽象概念，因此构建的机制共享的共同基础很少，而对两种智能体系统——工作流（Workflows）和智能体（Agents）的共同视角也相对有限。我们构建了一个抽象机器，提供了这两者。我们将LLM视为一个Oracle，并扩展了一个带有一条指令的双栈下推自动机，该指令将整个栈作为查询交给Oracle，并将答案附加到同一个栈上。因此，该机器执行了两种计算：Oracle的计算和我们称之为女祭司（Priestess）的图灵完备计算。程序仅向其附加的栈以自回归的方式增长，正如智能体的上下文一样。两个对称破缺，存储中的S和转换中的T，使得女祭司程序成为Oracle运行的程序的操作系统，并产生智能体和工作流作为任务程序的两种表现形式。对于内部自回归的Oracle，在特定条件下，两个计算在每个答案结束时同步，通过这种同步我们建模缓存并分析调度。没有任何适用于每个Oracle的保证能够确定哪些内容在两个计算之间交叉，但这样的保证确实固定了边界本身。构造V将机器适配于冯·诺依曼计算机。为了证明其可实现性，我们提出了ArchNights，一个扩展的RISC-V指令集架构（ISA）和一个基于设计实现该机器的类Linux操作系统。ArchNights-SE在gem5上作为计算机系统运行，当其运行LLM作为Oracle时，成为一个智能体系统，并将开源。智能体系统因此可以像计算机系统一样被设计。在建立基础并形成统一视角后，未来的工作可以共享不变量和界限，每个都有其条件。
cs.AI / 49 / 2610.09294
RT-Safe: Benchmarking Agent Safety in Real-Time Embodied Environment
RT-Safe：实时具身环境中代理安全性的基准测试
Abstract
Rapid progress in AI agents has brought growing attention to agent safety, with extensive evaluation focused on digital environments. As agents move into the physical world, embodied safety becomes increasingly important: failures can cause human injury and costly hardware damage. Beyond selecting safe actions, embodied agents must also operate under real-time constraints: the physical world does not pause while an agent reasons. As pedestrians move and vehicles approach during inference, an action that appears safe at observation time may become unsafe before execution. Real-time embodied safety therefore depends on both decision quality and decision latency. We introduce RT-SAFE, a simulated urban benchmark for evaluating embodied-agent safety under real-time constraints. RT-SAFE combines navigation tasks with moving actors, environmental hazards, and traffic rules, while allowing the world to evolve throughout inference and action execution. Across eight VLMs, agents achieve high task completion yet almost never complete safely: in the hardest setting, only 0.7% of episodes finish without a safety event. More strikingly, matched static and real-time evaluations yield task completion rates of 91.3% and 94.1%, respectively, while real-time execution increases collisions by $12.3\times$. These results reveal that standard task success can mask substantial safety failures, and that decision latency itself can become a source of physical risk. Finally, we show that RT-SAFE can support offline RL training and substantially reduce collision rates while achieving strong task completion.
Chinese Translation
人工智能代理的快速进展引起了对代理安全性日益增长的关注，现有的广泛评估主要集中在数字环境中。随着代理进入物理世界，具身安全变得愈发重要：失败可能导致人身伤害和昂贵的硬件损坏。除了选择安全的行动外，具身代理还必须在实时约束下操作：物理世界在代理推理时并不会暂停。当行人移动和车辆接近时，在观察时看似安全的行动在执行前可能变得不安全。因此，实时具身安全依赖于决策质量和决策延迟。我们引入了RT-SAFE，这是一个模拟城市基准，用于评估在实时约束下的具身代理安全性。RT-SAFE结合了导航任务、移动参与者、环境危害和交通规则，同时允许世界在推理和行动执行过程中不断演变。在八个视觉语言模型（VLMs）中，代理实现了高任务完成率，但几乎从未安全完成：在最困难的设置中，只有0.7%的回合在没有安全事件的情况下结束。更引人注目的是，静态评估和实时评估的任务完成率分别为91.3%和94.1%，而实时执行则使碰撞增加了12.3倍。这些结果表明，标准任务成功可能掩盖了重大的安全失败，而决策延迟本身也可能成为物理风险的来源。最后，我们展示了RT-SAFE可以支持离线强化学习训练，并在实现强大任务完成的同时显著降低碰撞率。
cs.AI / 50 / 2610.09296
The AI Evaluation Ecosystem
人工智能评估生态系统
Abstract
AI evaluation shapes the decisions of model providers, users, funders, and regulators. We argue that designing valid benchmarks requires contextualizing design choices in the dynamics of this ecosystem of actors. We develop a simulation architecture that combines rule-based market dynamics with LLM-driven strategic actors, building on advances in Generative Agent-Based Modeling (GABM). We model benchmarks, consumer needs, and provider capabilities as vectors over a six-dimensional capability space (reasoning, coding, knowledge, safety, communication, agentic), with structural information partitions across actors. As a case study, we apply this stylized simulation to explore benchmark holdout design. We find that moving from public benchmarks to private holdout benchmarks shrinks the gap between benchmark scores and user satisfaction on most benchmarks but widens it on a few, depending on where holdout weights shift scoring credit. We stress-test our findings at both the instrument and case-study level, drawing on the V&V framework of Sargent (2013) and GABM-specific evidence criteria. Beyond holdout design, our simulation is a hypothesis-generating sandbox for studying how evaluator and policy choices, in turn, reshape the ecosystem.
Chinese Translation
人工智能评估影响模型提供者、用户、资助者和监管者的决策。我们认为，设计有效的基准需要将设计选择置于这一生态系统中各参与者的动态背景下。我们开发了一种模拟架构，将基于规则的市场动态与由大型语言模型（LLM）驱动的战略参与者相结合，基于生成代理模型（Generative Agent-Based Modeling, GABM）的进展。我们将基准、消费者需求和提供者能力建模为六维能力空间（推理、编码、知识、安全、沟通、代理性）中的向量，并在参与者之间划分结构信息。作为案例研究，我们将这种风格化的模拟应用于探索基准保留设计。我们发现，从公共基准转向私有保留基准在大多数基准上缩小了基准得分与用户满意度之间的差距，但在少数基准上则扩大了这一差距，具体取决于保留权重如何影响评分信用。我们在工具和案例研究层面对我们的发现进行了压力测试，借鉴了Sargent（2013）的验证与验证（V&V）框架和GABM特定的证据标准。除了保留设计之外，我们的模拟还是一个生成假设的沙盒，用于研究评估者和政策选择如何反过来重塑生态系统。
cs.AI / 51 / 2610.09335
SearchWorld: Spatial Value-Grounded Imagination for UAV Object Search via World Models
SearchWorld：基于空间价值的无人机目标搜索的世界模型想象
Abstract
Autonomous unmanned aerial vehicle (UAV) object search involves a closed loop of perception, decision-making, and action under partial observability. Urban environments pose several challenges: large search areas and narrow egocentric views limit coverage, dense 3D geometry constrains safe motion, and open-world instructions require identifying a specific target among distractors. Many existing methods mitigate partial observability through explicit maps or memory representations, yet remain largely reactive, reasoning over past observations without explicitly predicting future states. World models enable prospective reasoning through imagined rollouts. However, image-generating world models can incur high inference latency, while spatially grounded planning remains challenging for latent world models. We propose SearchWorld, a recurrent state-space world model that connects explicit spatial memory with value-guided imagination. The model maintains BEV exploration and obstacle memory and decodes a task-aware spatial value layer to guide search. A cognition-action network uses this learned spatial value prior to improve the policy through imagined rollouts, without training a separate scalar critic. Training progresses from world-model learning to expert imitation and imagination-based exploration refinement. On UAV-ON, SearchWorld improves the success rate to 23.8% (19.5% for the strongest published agent) and raises oracle success to 35.5%, while remaining robust on unseen scenes (19.9% success rate). By grounding imagination in explicit spatial representations, SearchWorld enables UAV agents to plan prospectively rather than react.
Chinese Translation
自主无人机（UAV）目标搜索涉及在部分可观测性下的感知、决策和行动的闭环。城市环境带来了多重挑战：广阔的搜索区域和狭窄的自我中心视野限制了覆盖范围，密集的三维几何形状限制了安全运动，而开放世界的指令要求在干扰物中识别特定目标。许多现有方法通过显式地图或记忆表示来缓解部分可观测性，但仍然主要是反应性的，基于过去的观察进行推理，而没有明确预测未来状态。世界模型通过想象的展开实现前瞻性推理。然而，生成图像的世界模型可能会导致高推理延迟，而基于空间的规划对于潜在世界模型仍然具有挑战性。我们提出了SearchWorld，这是一种递归状态空间世界模型，将显式空间记忆与价值引导的想象相连接。该模型维护鸟瞰图（BEV）探索和障碍物记忆，并解码任务感知的空间价值层以指导搜索。认知-行动网络利用这种学习到的空间价值先验，通过想象的展开来改善策略，而无需训练单独的标量评论者。训练过程从世界模型学习到专家模仿，再到基于想象的探索精细化。在UAV-ON上，SearchWorld将成功率提高到23.8%（最强已发布代理为19.5%），并将oracle成功率提升至35.5%，同时在未见场景中保持稳健（成功率为19.9%）。通过将想象根植于显式空间表示，SearchWorld使无人机代理能够进行前瞻性规划，而非反应式行动。
cs.AI / 52 / 2610.09348
Relevance Is Not Sufficiency: What Actually Closes the Evidence Gap in Long-Term Memory QA
相关性不足以满足需求：在长期记忆问答中实际弥补证据缺口的因素
Abstract
LLM agents that interact with a user across many sessions accumulate histories that exceed their context window, so they store past interactions in an external memory and answer each question from a small set of retrieved records. Existing memory systems rank records by lexical or embedding relevance, yet the top-ranked memories can each be relevant while jointly omitting a complementary fact that the answer requires, especially for multi-session and temporal questions. Drawing on the distinction between relevance and sufficiency in legal evidence scholarship, we recast memory retrieval as constructing a sufficient memory set. To operationalize this view, we introduce a blinded LLM judgment over the retrieved set, together with Gold Hit and Turn Hit as evidence-coverage proxies. We then propose Budgeted Flat Reconstruction (BFR), which builds sufficient sets over a fixed flat memory store in two stages. Specifically, we first apply Formal Concept Analysis for Memory Selection (FCA-MS) to decompose the question into information requirements and select a compact candidate subset that jointly covers them. Then, we repeatedly acquire unseen records through deeper text search or complementary entity and session views, stopping when the budget is exhausted. Experiments on LoCoMo and LongMemEval-S show that BFR outperforms same-store adaptations of recent agent-memory systems in both answer quality and evidence coverage. Specifically, on LongMemEval-S it raises judged accuracy from 72.4% to 82.2% and Turn Hit to 91.4%.
Chinese Translation
与用户进行多次会话的LLM代理会积累超出其上下文窗口的历史记录，因此它们将过去的交互存储在外部记忆中，并从一小部分检索到的记录中回答每个问题。现有的记忆系统通过词汇或嵌入相关性对记录进行排名，然而，排名靠前的记忆可能各自相关，但共同忽略了答案所需的补充事实，尤其是在多会话和时间性问题中。基于法律证据研究中相关性与充分性的区别，我们将记忆检索重新构建为构造一个充分的记忆集。为了实现这一观点，我们引入了对检索集的盲判定LLM，并使用Gold Hit和Turn Hit作为证据覆盖的代理。然后，我们提出了预算平面重构（Budgeted Flat Reconstruction, BFR），该方法在两个阶段内在固定的平面记忆存储上构建充分集。具体而言，我们首先应用记忆选择的形式概念分析（Formal Concept Analysis for Memory Selection, FCA-MS）将问题分解为信息需求，并选择一个紧凑的候选子集以共同覆盖这些需求。然后，我们通过更深层的文本搜索或补充实体和会话视图反复获取未见记录，直到预算耗尽。对LoCoMo和LongMemEval-S的实验表明，BFR在答案质量和证据覆盖方面优于近期代理记忆系统的同存储适配。具体而言，在LongMemEval-S上，它将判断准确率从72.4%提高到82.2%，Turn Hit提高到91.4%。
cs.AI / 53 / 2610.09360
TopoGraphRAG-Bench: Evaluating Multimodal GraphRAG on Layout-Grounded Evidence Reasoning
TopoGraphRAG-Bench：评估基于布局的多模态 GraphRAG 在证据推理中的表现
Abstract
Real-world documents distribute evidence across text, tables, figures, and captions within complex page layouts. Answering complex questions over such documents therefore requires more than retrieving relevant passages: systems must recover the evidence topology that connects heterogeneous evidence units. Existing GraphRAG evaluations remain largely text-centered, while multimodal document RAG benchmarks assess cross-modal retrieval and generation without directly evaluating recovery of the intended evidence topology. We introduce TOPOGRAPHRAG-BENCH, a layout-grounded benchmark for multimodal evidence reasoning in GraphRAG, comprising 2,024 questions over 201 long, visually rich documents. Questions are constructed bottom-up from text, figure, and table evidence units under three controlled topologies: single-hop retrieval, bridge-chain reasoning, and multi-source synthesis. To ensure that questions preserve their intended structure, we apply counterfactual validation for shortcut resistance, modality necessity, and evidence necessity. We evaluate text-only GraphRAG, page-level visual retrieval, and multimodal GraphRAG systems using retrieval, generation, and topology-aware reasoning metrics. Multimodal GraphRAG systems achieve the strongest overall performance, but still fail when visual-textual evidence alignment or multi-unit composition is incomplete. Text-only GraphRAG struggles when key dependencies are grounded in figures or tables, while page-level visual retrieval lacks the fine-grained structure needed for topology recovery. These findings motivate GraphRAG systems that move beyond text-derived entity relation graphs to explicitly model document layouts, cross-modal evidence alignment, and the reasoning roles of evidence units. Code and data are available at https://richardlrc.github.io/TopoGraphRAG-Bench/.
Chinese Translation
现实世界的文档在复杂的页面布局中分布着文本、表格、图形和标题等证据。因此，回答关于这些文档的复杂问题不仅需要检索相关段落：系统还必须恢复连接异构证据单元的证据拓扑。现有的 GraphRAG 评估主要集中于文本，而多模态文档 RAG 基准则评估跨模态检索和生成，但未直接评估预期证据拓扑的恢复。我们引入了 TOPOGRAPHRAG-BENCH，这是一个基于布局的多模态证据推理基准，包含 2,024 个问题，涵盖 201 个长篇、视觉丰富的文档。问题是从文本、图形和表格证据单元自下而上构建的，遵循三种受控拓扑：单跳检索、桥链推理和多源综合。为了确保问题保持其预期结构，我们应用了反事实验证，以抵抗捷径、确保模态必要性和证据必要性。我们使用检索、生成和拓扑感知推理指标评估仅基于文本的 GraphRAG、页面级视觉检索和多模态 GraphRAG 系统。多模态 GraphRAG 系统在整体性能上表现最佳，但在视觉-文本证据对齐或多单元组合不完整时仍然失败。仅基于文本的 GraphRAG 在关键依赖关系基于图形或表格时表现不佳，而页面级视觉检索缺乏恢复拓扑所需的细粒度结构。这些发现促使 GraphRAG 系统超越基于文本的实体关系图，明确建模文档布局、跨模态证据对齐和证据单元的推理角色。代码和数据可在 https://richardlrc.github.io/TopoGraphRAG-Bench/ 获取。
cs.AI / 54 / 2610.09374
DUDA-Bench: Benchmarking LLM Agents on Multimodal Data-Driven Urban Diagnosis
DUDA-Bench：基于多模态数据驱动的城市诊断的基准测试
Abstract
Urban diagnosis integrates heterogeneous observations to identify urban problems, localize affected areas, and investigate contributing factors, informing evidence-based urban planning and management. However, its reliance on labor-intensive, case-specific expert workflows limits scalability and reuse, motivating the exploration of agent-based execution. To evaluate this capability, we introduce DUDA-Bench, a hierarchical and interactive benchmark that formalizes data-driven urban diagnosis as a multi-stage agent workflow. It comprises 86 atomic and 22 workflow tasks spanning four analytical stages, grounded in multimodal data from 12 cities covering five urban problem types. Evaluations of seven backbone models and five agent systems reveal a substantial gap between isolated analytical competence and end-to-end diagnosis, with system benefits varying across backbones. Trajectory analysis shows that unresolved evidence gaps propagate across stages, while successful recovery involves revising assumptions and actions using feedback. These findings highlight limitations in coordinating analytical capabilities across stages, particularly adaptive planning, evidence integration, and verification. More broadly, DUDA-Bench provides a framework for translating expert analytical workflows into hierarchical agent tasks and process-aware evaluation, supporting systematic assessment of end-to-end analytical capabilities.
Chinese Translation
城市诊断整合异构观察，以识别城市问题、定位受影响区域并调查影响因素，从而为基于证据的城市规划和管理提供信息。然而，其对劳动密集型、特定案例的专家工作流程的依赖限制了可扩展性和重用性，这促使我们探索基于智能体的执行能力。为评估这一能力，我们引入了DUDA-Bench，这是一个分层和交互式的基准测试，将数据驱动的城市诊断形式化为多阶段的智能体工作流程。它包含86个原子任务和22个工作流程任务，涵盖四个分析阶段，基于来自12个城市的多模态数据，涉及五种城市问题类型。对七个基础模型和五个智能体系统的评估显示，孤立的分析能力与端到端诊断之间存在显著差距，系统的效益在不同基础模型之间有所不同。轨迹分析表明，未解决的证据缺口在各阶段之间传播，而成功的恢复涉及使用反馈修正假设和行动。这些发现突显了在各阶段协调分析能力的局限性，尤其是在适应性规划、证据整合和验证方面。更广泛地说，DUDA-Bench提供了一个框架，用于将专家分析工作流程转化为分层智能体任务和过程感知评估，支持对端到端分析能力的系统评估。
cs.AI / 55 / 2610.09377
From Plausible Hierarchies to Useful Taxonomies: Evaluating Agentic Harnesses on Customer Feedback
从合理的层级到有用的分类法：评估客户反馈中的自主控制工具
Abstract
Taxonomies are the symbolic representations through which AI systems organize evidence, aggregate patterns, and answer questions over large document collections. Over customer feedback, the category tree decides how every record is counted and routed, which problems get seen, and which team owns them. Agentic harnesses now make it easy to generate a plausible-looking hierarchy, and such trees are checked today with generic, individually scoped checks: each name fits its description, sits under the right parent, and stays distinct from its siblings. We ask a more operational question: when is a generated hierarchy actually useful as a production taxonomy? We build six taxonomies over two proprietary feedback corpora (1,940 and 5,000 records): for each corpus, a production reference and two repeated runs of the same harness under identical inputs. All six pass every generic naming and structure check, and a deeper product-coverage check even prefers the generated trees. Yet in every generated tree at least 97.7% of leaf names merely restate an ancestor's name (13.9% and 2.9% in the references), and in one, three of every four records fall under multiple top-level categories. We introduce two families of whole-tree metrics: structural discriminators test whether a tree's shape was learned from the data or imposed by its generator; team partitionability tests whether branches split feedback into groups teams can own. Trees the generic checks rate as equally correct differ by 27 percentage points in cross-branch leakage, and only one of the two beats a random split. Surface plausibility is an insufficient measure of taxonomy quality: evaluation must measure the whole tree as well as each node.
Chinese Translation
分类法是人工智能系统组织证据、聚合模式和在大规模文档集合中回答问题的符号表示。在客户反馈中，类别树决定了每条记录的计数和路由方式，哪些问题被关注，以及哪个团队负责这些问题。自主控制工具现在使得生成一个看似合理的层级变得容易，而这些树今天通过通用的、个别范围的检查进行验证：每个名称符合其描述，位于正确的父节点下，并且与其兄弟节点保持区分。我们提出一个更具操作性的问题：生成的层级在何种情况下实际上作为生产分类法是有用的？我们在两个专有反馈语料库（分别包含1,940和5,000条记录）上构建了六个分类法：对于每个语料库，建立一个生产参考和在相同输入下重复运行的两个相同工具。所有六个分类法都通过了每个通用的命名和结构检查，而更深入的产品覆盖检查甚至更倾向于生成的树。然而，在每个生成的树中，至少97.7%的叶节点名称仅仅是重述了祖先的名称（参考中分别为13.9%和2.9%），并且在一个树中，每四条记录中就有三条落在多个顶级类别下。我们引入了两类全树度量：结构区分器测试树的形状是从数据中学习得来的还是由其生成器强加的；团队可分配性测试分支是否将反馈分成团队可以负责的组。通用检查评估为同样正确的树在跨分支泄漏方面差异达到27个百分点，且只有两个中的一个优于随机分割。表面合理性不足以衡量分类法的质量：评估必须测量整个树以及每个节点。
cs.AI / 56 / 2610.09401
Let the Library Speak: Self-Advertised Method Selection for Formal Proving
让图书馆发声：用于形式证明的自我广告方法选择
Abstract
LLM-based formal provers can retrieve relevant lemmas and prior proofs, but relevance alone does not say whether a mathematical method can be used on the current theorem. A method has prerequisites, a target, an intended action, and obligations that its use leaves to prove. Methods that look equally related to a theorem may therefore differ substantially in whether they offer a plausible next step. We formulate this as an applicability-aware method-selection problem and introduce self-advertisement: before candidates are ranked, a model generates a problem-specific proposal for each one, stating what part of the goal it targets, what action it would take, and what conditions that action requires. We organize 82 reusable methods from Putnam 2000-2014 as Method Contracts, which pair applicability descriptions with Mathlib anchors, a checked example or scaffold, and expected proof obligations. A single batched call elicits proposals across the library; vague or unsupported proposals are demoted, yielding a ranked shortlist accompanied by inspectable claims about each candidate's use. We analyze when similarity-based representations cannot distinguish methods with different applicability, how errors in applicability estimates affect shortlist quality, and what a checked scaffold guarantees under its stated assumptions. Against lexical, embedding, and embedding-plus-LLM reranking baselines, self-advertisement achieves 95.0% hit@5 on Putnam 2015-2025, compared with 84.2% for the strongest reranker. On IMO ProofBench, it achieves 91.7% compared with 88.3%. These results indicate improved coverage of annotated methods in the retrieved shortlists, particularly on Putnam.
Chinese Translation
基于大语言模型（LLM）的形式证明工具可以检索相关的引理和先前的证明，但仅仅依靠相关性并不能说明某种数学方法是否可以应用于当前的定理。一个方法有其先决条件、目标、预期的行动以及其使用所留下的证明义务。因此，看似与定理相关的方法在提供合理的下一步方面可能存在显著差异。我们将此表述为一个考虑适用性的选择方法问题，并引入自我广告：在候选方法被排名之前，模型为每个候选方法生成一个特定于问题的提案，说明其针对目标的哪一部分、将采取什么行动以及该行动所需的条件。我们将2000-2014年间的82种可重用方法组织为方法合同（Method Contracts），将适用性描述与Mathlib锚点（Mathlib anchors）、经过验证的示例或框架以及预期的证明义务配对。通过一次批量调用，可以在整个库中引发提案；模糊或不支持的提案将被降级，从而产生一个带有可检查声明的排名短名单，说明每个候选方法的使用情况。我们分析了何时相似性基础的表示无法区分具有不同适用性的方法，适用性估计中的错误如何影响短名单的质量，以及在其声明的假设下经过验证的框架所保证的内容。与词汇、嵌入和嵌入加LLM重新排名的基线相比，自我广告在2015-2025年普特南（Putnam）上达到了95.0%的命中率@5，而最强的重新排名方法为84.2%。在国际数学奥林匹克（IMO）证明平台上，其命中率为91.7%，而对比为88.3%。这些结果表明，在检索的短名单中，注释方法的覆盖率得到了改善，特别是在普特南上。
cs.AI / 57 / 2610.09416
Efficient Reasoning with Flow Language Models
高效推理与流语言模型
Abstract
Flow Language Models (FLMs) have emerged as a continuous-state alternative to discrete diffusion language models, yet the role of their continuous representations in reasoning remains unclear. We investigate this question by comparing the reasoning efficiency of FLMs and discrete diffusion models, measured by solution accuracy under matched denoising steps. Unlike discrete diffusion, which passes categorical states between denoising steps, FLMs evolve a continuous sequence representation throughout denoising and decodes it into discrete tokens only at the end. Our theoretical analysis shows, from a superposition perspective, how information retained in these continuous states can benefit reasoning. Intermediate-state interventions provide further empirical support for this theoretical account, showing that removing information about alternative candidates reduces subsequent solution recovery. Together, these findings show that FLMs allow evidence for multiple candidates to persist and inform subsequent reasoning before a discrete answer is produced. Furthermore, our experiments on maze planning and Sudoku tasks show that FLMs achieve greater reasoning efficiency in the few-step regime: FLMs achieves higher sequence accuracy than discrete diffusion baselines at matched model sizes and small denoising steps. On maze planning tasks, FLMs can also achieve comparable accuracy with smaller models. For example, on Maze15, FLM reaches the 95\% accuracy target at 64 denoising steps with 36.5\% fewer parameters than MDLM. These findings point to continuous state spaces as a promising foundation for reasoning models that require fewer refinement steps.
Chinese Translation
流语言模型（Flow Language Models, FLMs）作为离散扩散语言模型的连续状态替代方案逐渐崭露头角，但其连续表示在推理中的作用仍不明确。我们通过比较FLMs与离散扩散模型的推理效率来探讨这一问题，推理效率通过在匹配去噪步骤下的解决方案准确性来衡量。与离散扩散模型在去噪步骤之间传递类别状态不同，FLMs在整个去噪过程中演变为连续序列表示，并仅在最后将其解码为离散标记。我们的理论分析从叠加的角度展示了这些连续状态中保留的信息如何有利于推理。中间状态干预为这一理论提供了进一步的实证支持，表明去除关于替代候选的信息会降低后续解决方案的恢复能力。综合这些发现表明，FLMs允许多个候选的证据持续存在，并在产生离散答案之前为后续推理提供信息。此外，我们在迷宫规划和数独任务上的实验表明，FLMs在少步推理模式下实现了更高的推理效率：在匹配模型大小和较小去噪步骤下，FLMs的序列准确性高于离散扩散基线。在迷宫规划任务中，FLMs还能够以更小的模型实现可比的准确性。例如，在Maze15上，FLM在64个去噪步骤下以比MDLM少36.5%的参数达到了95%的准确性目标。这些发现表明，连续状态空间为需要更少细化步骤的推理模型提供了一个有前景的基础。
cs.AI / 58 / 2610.09426
RSI-Forge: From Research Papers to Environments for Recursive Self-Improvement
RSI-Forge：从研究论文到递归自我改进的环境
Abstract
Environments are the foundation of recursive self-improvement: they provide the problems agents work on and the feedback used to evaluate progress. Yet constructing challenging research environments with reliable evaluation still depends on domain experts, limiting their scale and disciplinary coverage. We introduce RSI-Forge, a multi-agent pipeline that turns published papers into executable environments for self-improvement. Three agents coordinate construction, reproduction, and review to produce tasks with automated evaluators; each paper's method is independently reimplemented to establish a baseline score. We present 210 environments across 18 fields, including 90 reviewed by independent human domain experts. Both experts and agent judges give high ratings to the potential for improving the provided starting solutions and the evaluators' ability to distinguish solution quality, whereas experts are more critical of shortcut resistance, faithfulness to the source paper, and whether a single idea can exhaust a task. To validate their use for repeated improvement, we evaluate four models over 3 successive attempts on 120 environments, with each attempt inheriting prior code and notes while model weights remain fixed. At least one model improves after the first attempt in 84% of environments. Models also outperform the reproduced paper methods in 68 of the 120 environments, demonstrating room for gains beyond these baselines. Transcript analysis identifies work beyond parameter tuning in 95% of these successful attempts. Analysis of the resulting trajectories shows that models scoring lower on these tasks explore less, more often accept gains smaller than the reported standard error, and rely more heavily on tuning to the development set. RSI-Forge provides a scalable approach to constructing research environments for training and evaluating self-improving agents.
Chinese Translation
环境是递归自我改进的基础：它们提供了代理人所处理的问题和用于评估进展的反馈。然而，构建具有可靠评估的挑战性研究环境仍然依赖于领域专家，这限制了其规模和学科覆盖面。我们介绍了RSI-Forge，一个将已发表论文转化为可执行自我改进环境的多代理管道。三个代理协调构建、重现和审查，以生成带有自动评估器的任务；每篇论文的方法独立重新实现，以建立基准分数。我们展示了涵盖18个领域的210个环境，其中90个经过独立人类领域专家的审查。专家和代理评审员都对改善提供的初始解决方案的潜力以及评估者区分解决方案质量的能力给予了高评价，而专家对抵抗捷径、对源论文的忠实度以及单一想法是否能耗尽任务则更加批评。为了验证它们在重复改进中的使用，我们在120个环境上对四个模型进行了3次连续尝试评估，每次尝试都继承了先前的代码和笔记，同时模型权重保持固定。在84%的环境中，至少有一个模型在第一次尝试后有所改进。在120个环境中，模型的表现也优于重现的论文方法，显示出超越这些基准的潜力。转录分析在95%的成功尝试中识别出了超出参数调优的工作。对结果轨迹的分析表明，得分较低的模型在这些任务上探索较少，更常接受小于报告标准误差的收益，并更依赖于对开发集的调优。RSI-Forge提供了一种可扩展的方法，用于构建研究环境，以训练和评估自我改进的代理。
cs.AI / 59 / 2610.09466
LLM-Enabled UAV Dispatch: A System-Level Survey and Taxonomy
基于大语言模型的无人机调度：系统层面的调查与分类
Abstract
Unmanned aerial vehicle (UAV) dispatch is beginning to move beyond isolated path planning and optimization-driven resource allocation toward system-level coordination supported by semantic reasoning and LLM-based interfaces. This survey provides a unified characterization of LLM-enabled UAV dispatch systems that bridges semantic intent, symbolic decision-making, and physical UAV execution. Rather than treating LLMs as standalone add-ons, we conceptualize them as a cross-layer semantic orchestration layer connecting human instructions, external solvers, and distributed control modules. We organize the literature into four representative dispatch paradigms: pipeline dispatch, global assignment dispatch, decentralized agentic dispatch, and divide-and-conquer dispatch. For each paradigm, we analyze its decision logic, system structure, control flow, representative methods, and potential LLM roles. We further examine how LLMs support semantic parsing, retrieval-grounded planning, solver orchestration, local agent reasoning, multi-agent coordination, safety assessment, and human-facing explanation. We discuss the implications of these paradigms for scalability, robustness, coordination burden, and verification requirements, and identify open challenges including latency-aware reasoning, grounding reliability, physical feasibility guarantees, edge deployment, privacy protection, and distributed consistency. This survey provides a system-level taxonomy and design perspective for integrating LLMs into safety-critical UAV dispatch systems.
Chinese Translation
无人机（UAV）调度正开始超越孤立的路径规划和优化驱动的资源分配，向由语义推理和基于大语言模型（LLM）接口支持的系统级协调发展。本调查提供了一个统一的LLM支持的无人机调度系统特征描述，连接了语义意图、符号决策和物理无人机执行。我们将LLM视为一个跨层的语义协调层，连接人类指令、外部求解器和分布式控制模块，而不是将其视为独立的附加组件。我们将文献组织为四种代表性的调度范式：管道调度、全局分配调度、去中心化代理调度和分而治之调度。对于每种范式，我们分析其决策逻辑、系统结构、控制流程、代表性方法以及潜在的LLM角色。我们进一步探讨LLM如何支持语义解析、基于检索的规划、求解器协调、本地代理推理、多代理协调、安全评估和面向人类的解释。我们讨论这些范式对可扩展性、鲁棒性、协调负担和验证要求的影响，并识别出包括延迟感知推理、基础可靠性、物理可行性保证、边缘部署、隐私保护和分布式一致性等开放挑战。本调查为将LLM集成到安全关键的无人机调度系统中提供了系统层面的分类和设计视角。
cs.AI / 60 / 2610.09484
MIMESIS: Learning User Simulators as Training Environments for Interactive Agents
MIMESIS：作为交互代理训练环境的用户模拟器学习
Abstract
Training and evaluating interactive language agents typically requires rich user interactions, yet collecting human feedback is expensive and difficult to scale. Simulated users offer a scalable alternative, but they must both resemble real user behavior and provide useful learning experiences for agents. In contrast, most agent-training frameworks rely on off-the-shelf assistant LLMs, whose helpfulness can make them overly cooperative, explicit, and behaviorally homogeneous compared with real users. We introduce MIMESIS, a purpose-built user simulator trained on human conversations with explicit reasoning supervision and 13 realistic behavioral patterns derived from real user interactions. Empirically, our 9B model achieves a SOUL-Index of 65.7, surpassing the strongest frontier model. Compared with Claude-Opus-5, the strongest baseline on RealUserSim and SimulatorArena, MIMESIS improves behavioral fidelity by 13.4 points and reduces Turing distance by 3.6 points, respectively. We then freeze the simulator and train an agent by interacting with the frozen simulator using multi-turn reinforcement learning. Across eight environments, training with MIMESIS yields better agent performance than training with GPT-5.5 under all nine unseen user simulators, demonstrating stronger generalization to new user simulators. Moreover, we propose Coached On-Policy Self-Distillation (CSD), which leverages simulator-generated private reasoning traces and subsequent utterances as feedback on how well the agent addresses user needs. A coach converts this information into concise coaching notes that describe how the agent can better anticipate user needs and adapt its behavior over the course of an interaction. CSD turns this feedback into dense, token-level supervision beyond sparse task rewards, yielding further gains across all nine evaluation user models.
Chinese Translation
训练和评估交互语言代理通常需要丰富的用户交互，但收集人类反馈既昂贵又难以扩展。模拟用户提供了一种可扩展的替代方案，但它们必须既能模拟真实用户行为，又能为代理提供有用的学习体验。相比之下，大多数代理训练框架依赖于现成的助手大型语言模型（LLMs），其有用性可能使其在行为上过于合作、明确且同质化，与真实用户相比。我们介绍了MIMESIS，这是一种专门构建的用户模拟器，经过人类对话的显式推理监督训练，并基于真实用户交互衍生出13种现实的行为模式。实证结果表明，我们的9B模型达到了65.7的SOUL-Index，超越了最强的前沿模型。与在RealUserSim和SimulatorArena上表现最强的基线Claude-Opus-5相比，MIMESIS在行为逼真度上提高了13.4分，并分别减少了3.6分的图灵距离。随后，我们冻结模拟器，并通过与冻结的模拟器进行多轮强化学习来训练代理。在八个环境中，使用MIMESIS进行训练的代理性能优于在所有九个未见用户模拟器上使用GPT-5.5进行训练的代理，显示出对新用户模拟器的更强泛化能力。此外，我们提出了教练式在线自蒸馏（CSD），该方法利用模拟器生成的私有推理痕迹和后续发言作为反馈，评估代理满足用户需求的效果。教练将这些信息转换为简明的教练笔记，描述代理如何更好地预测用户需求并在交互过程中调整其行为。CSD将这种反馈转化为超越稀疏任务奖励的密集、逐标记的监督，从而在所有九个评估用户模型中获得进一步的提升。
cs.AI / 61 / 2610.09493
The Attribution Blind Spot: Layerwise Trajectory Diagnostics for Source Reliance in Retrieval-Augmented Language Models
归因盲点：检索增强语言模型中源依赖的层级轨迹诊断
Abstract
A retrieval-augmented model can match a document without relying on it. Controlled knowledge conflicts make source choice observable and let us ask a second question that prediction alone cannot answer: which internal-state properties define useful intervention directions? We study paired hidden-state changes with Latent Trajectory Shift (LTS), a signed projection onto a training-fitted first principal component (PC1), and keep verified training exposure separate from behavioral source choice. Across the evaluated conflicts, state-change magnitude is often the stronger predictor, whereas signed PC1 is the stronger selective controller: equal-norm interventions change source preference while better preserving non-target behavior, and the frozen direction transfers across the tested datasets and aligned model pairs. A same-system OLMo study further combines positive choice and control results with inconclusive exposure detection at the achieved power. The central result is a separation: representations that diagnose what a model will choose need not be the representations that best control that choice.
Chinese Translation
检索增强模型可以在不依赖文档的情况下匹配文档。受控的知识冲突使得源选择可观察，并让我们提出一个仅凭预测无法回答的第二个问题：哪些内部状态属性定义了有用的干预方向？我们研究了配对的隐藏状态变化，使用潜在轨迹偏移（Latent Trajectory Shift, LTS），这是对训练拟合的第一主成分（PC1）的有符号投影，并将经过验证的训练暴露与行为源选择分开。在评估的冲突中，状态变化的幅度往往是更强的预测因子，而有符号的PC1则是更强的选择性控制器：等范数的干预改变了源偏好，同时更好地保留了非目标行为，而冻结的方向在测试的数据集和对齐的模型对之间转移。一个同系统的OLMo研究进一步将积极选择和控制结果与在实现的能力下不确定的暴露检测结合起来。中心结果是一个分离：诊断模型将选择什么的表征不必是最佳控制该选择的表征。
cs.AI / 62 / 2610.09495
Cognitive Schemas, Laws and Tasks
认知图式、法律与任务
Abstract
This paper asks how explicit representations can support reusable cognitive schemas in knowledge-based problem solving. We develop a structural framework in which schemas are organized by the information and relations required for their use, rather than introduced as unrelated primitives. The framework also distinguishes context-dependent relations from more stable structures that can be reused across different representations. Tasks are described through the information available, the unknowns to be determined, and the constraints that admissible solutions must satisfy. This makes it possible to separate limitations of the representation from limitations of the solving procedure. In particular, we distinguish inconsistency, underdetermination, and contextual insufficiency, where the current representation lacks distinctions or relations required by the external task meaning. We also show formally when a reduction of representation preserves the task-relevant solution structure. The resulting task--schema interface offers a structured way to describe representational conditions relevant to problem solving. It supports the reuse and stabilization of derived knowledge while remaining independent of the particular mechanism used to generate candidate solutions. This may provide a useful component for future solver architectures that combine structured knowledge, verification, and learned proposal mechanisms.
Chinese Translation
本文探讨了显性表征如何支持知识基础问题解决中的可重用认知图式。我们开发了一个结构框架，其中图式根据其使用所需的信息和关系进行组织，而不是作为无关的原始元素引入。该框架还区分了依赖于上下文的关系与可以在不同表征中重用的更稳定结构。任务通过可用信息、待确定的未知数以及可接受解必须满足的约束进行描述。这使得我们能够将表征的局限性与求解过程的局限性分开。特别地，我们区分了不一致性、欠确定性和上下文不足，其中当前表征缺乏外部任务意义所需的区别或关系。我们还正式展示了何时表征的简化能够保持任务相关的解结构。由此产生的任务-图式接口提供了一种结构化的方式来描述与问题解决相关的表征条件。它支持派生知识的重用和稳定，同时保持与生成候选解所使用的特定机制无关。这可能为未来结合结构化知识、验证和学习提议机制的求解器架构提供一个有用的组成部分。
cs.AI / 63 / 2610.09497
Ream: Unfolding Mutual Awareness in Human-Agent Workspaces
Ream：在人机工作空间中展开相互意识
Abstract
As AI agents work alongside humans in shared workspaces, a mutual awareness challenge arises: agents act at speeds that outpace human monitoring, and users' evolving interests are not always expressed in chat. This challenge is especially pressing in literature review, where both parties retrieve, read, and synthesize a growing body of papers. We present Ream, a literature review workspace that supports mutual awareness through structured artifacts, bidirectional engagement tracking, and localized visualizations. Users can see each party's activity within these documents, and agents can retrieve the same history to guide their work. In studies with eighteen researchers, participants used these traces to inspect evidence, steer agents, communicate through annotations, and reflect on their research focus. Shared histories also helped agents build on earlier work. These findings inform how engagement traces within shared documents can support transparency, personalized assistance, and coordination in human-agent knowledge work.
Chinese Translation
随着人工智能代理与人类在共享工作空间中协作，相互意识的挑战随之而来：代理的行动速度超过人类的监控能力，而用户不断变化的兴趣并不总是通过聊天表达出来。这一挑战在文献综述中尤为突出，因为双方需要检索、阅读和综合不断增长的文献资料。我们提出了Ream，一个支持相互意识的文献综述工作空间，通过结构化的文档、双向的参与跟踪和局部可视化来实现。用户可以在这些文档中看到每一方的活动，代理也可以检索相同的历史记录以指导其工作。在与十八位研究者的研究中，参与者利用这些痕迹检查证据、引导代理、通过注释进行沟通，并反思他们的研究重点。共享的历史记录还帮助代理在早期工作基础上进行构建。这些发现为如何在共享文档中利用参与痕迹支持透明度、个性化辅助和人机知识工作的协调提供了重要信息。
cs.AI / 64 / 2610.09554
Reliability of LLM Judges for Evaluating Entity Alignment
用于评估实体对齐的LLM评审者的可靠性
Abstract
Entity Alignment (EA) identifies equivalent entities across knowledge graphs and is critical for knowledge base integration and ontology merging. Evaluating EA systems at scale requires expensive expert annotation, making systematic assessment across diverse domains practically infeasible. LLM-as-judge evaluation offers a potentially scalable alternative, yet its reliability for structured prediction tasks like EA remains unstudied. We present the first systematic benchmarking study across three frontier models, three datasets, and four EA systems, using perturbation bias diagnostics, meta-evaluation across all dataset-judge-prompt combinations, and counterfactual label-flip tests. We identify anchor bias, a failure mode in which judges invert discrimination when the system's decision label is visible. Label exposure causally collapses judge discrimination (J-ROC-AUC 0.12-0.87), while a label-free protocol recovers near-ceiling capability on distinctive-name datasets (0.93-1.00) and significant recovery on biomedical pairs (0.93-0.95). Counterfactual experiments confirm causality (FSR 53-99%) and reveal a frontier model paradox: stronger judges exhibit greater label sensitivity, not less. A blinded two-annotator human evaluation (102 pairs, Cohen's kappa=0.902) confirms this mechanism directly. We release the first biomedical EA benchmark (MeSH-SNOMED CT, 15K pairs) and a reproducible auditing framework for LLM judge reliability in EA. Code and data are available at https://github.com/vaibhavalakshmiravideshik/llm-as-a-judge-entity-alignment.
Chinese Translation
实体对齐（Entity Alignment, EA）用于识别知识图谱中等效的实体，对于知识库集成和本体合并至关重要。在大规模评估EA系统时，需要昂贵的专家注释，这使得在不同领域进行系统评估在实践中几乎不可行。将LLM作为评审者的评估提供了一种潜在的可扩展替代方案，但其在像EA这样的结构化预测任务中的可靠性尚未得到研究。我们首次进行了系统的基准测试研究，涵盖三个前沿模型、三个数据集和四个EA系统，采用扰动偏差诊断、跨所有数据集-评审者-提示组合的元评估以及反事实标签翻转测试。我们识别出锚定偏差（anchor bias），这是一种评审者在系统决策标签可见时反转判别的失效模式。标签暴露因果性地导致评审者判别能力的崩溃（J-ROC-AUC 0.12-0.87），而无标签协议在独特名称数据集上恢复了接近上限的能力（0.93-1.00），在生物医学对上也有显著恢复（0.93-0.95）。反事实实验确认了因果关系（FSR 53-99%），并揭示了一个前沿模型悖论：更强的评审者表现出更大的标签敏感性，而非更小。一个盲评的两位注释者的人类评估（102对，Cohen's kappa=0.902）直接确认了这一机制。我们发布了首个生物医学EA基准（MeSH-SNOMED CT，15K对）和一个可重复的审计框架，用于评估LLM评审者在EA中的可靠性。代码和数据可在 https://github.com/vaibhavalakshmiravideshik/llm-as-a-judge-entity-alignment 获取。
cs.AI / 65 / 2610.09558
DrugTargetWorld: A Synthetic Biobank for Training and Benchmarking AI Scientists
DrugTargetWorld：用于训练和基准测试人工智能科学家的合成生物银行
Abstract
Drug target discovery requires distinguishing molecules that causally drive disease from those that are merely associated with it. Training and evaluating AI agents to perform this workflow end-to-end is difficult because real world biobanks lack known causal ground truth and participant-level data is access controlled. We introduce DrugTargetWorld, a framework that procedurally generates simulated biobanks, or "worlds," with known but concealed causal structure. Each world contains genotypes, proteins, health records, outcomes, and synthetic magnetic resonance imaging (MRI) for 54,000 participants. Agents must construct a disease phenotype, identify causal driver proteins, infer the beneficial direction of modulation, and optionally conduct virtual 'wet lab' experiments. We evaluated nine agents in 540 episodes across 20 cardiovascular worlds and three experimental budgets. Opus 5 and GPT-5.6 Sol achieved the highest mean composite scores, 39.98 and 35.38 of 100, respectively, and both recovered 64% of causal drivers on average. However, no agent reliably distinguished misleading non-causal proteins, and performance remained limited by the integrative judgments required to connect phenotype construction, causal evidence, and intervention decisions. By making each world's causal structure known to the evaluator but hidden from the agent, DrugTargetWorld turns end-to-end drug target discovery into a scalable training and evaluation problem with verifiable reward.
Chinese Translation
药物靶点发现需要区分因果驱动疾病的分子与仅仅与之相关的分子。训练和评估人工智能代理以执行这一端到端的工作流程是困难的，因为现实世界的生物银行缺乏已知的因果真相，且参与者级别的数据受到访问控制。我们介绍了DrugTargetWorld，一个程序性生成模拟生物银行或“世界”的框架，具有已知但隐蔽的因果结构。每个世界包含54,000名参与者的基因型、蛋白质、健康记录、结果和合成磁共振成像（MRI）。代理必须构建疾病表型，识别因果驱动蛋白，推断调节的有益方向，并可选择进行虚拟“湿实验”。我们在20个心血管世界和三个实验预算中评估了九个代理，共540个实验。Opus 5和GPT-5.6 Sol分别获得了最高的平均综合得分，39.98和35.38（满分100），并且平均恢复了64%的因果驱动因素。然而，没有任何代理能够可靠地区分误导性的非因果蛋白，且性能仍然受到连接表型构建、因果证据和干预决策所需的综合判断的限制。通过使每个世界的因果结构对评估者已知但对代理隐藏，DrugTargetWorld将端到端的药物靶点发现转变为一个可扩展的训练和评估问题，具有可验证的奖励。
cs.AI / 66 / 2610.09560
World Potential Model: Pretrained World Knowledge as Progress Potentials
世界潜力模型：预训练世界知识作为进展潜力
Abstract
Long-horizon language agents often receive supervision only from terminal task outcomes, leaving little signal for distinguishing productive intermediate behavior from stagnation or even regression. Rather than learning a separate value function or process reward model for every task, we ask whether pretrained models can recognize task progress from their existing world knowledge. We formalize this capability with a World Potential Model (WPM), a goal-conditioned evaluator of task-relative realized progress in agent contexts. In ALFWorld and ScienceWorld, off-the-shelf pretrained models substantially outperform chance at recovering realized-progress structure without task-specific evaluator fine-tuning. We further anchor these progress judgments to task-specific milestones to obtain scalar world potentials, whose temporal differences provide process-sensitive step-level credit for policy optimization. Under matched comparisons, WPM-guided optimization improves success over outcome-only GRPO across all evaluated configurations. Together, these results provide initial evidence that pretrained world knowledge can support reusable realized-progress evaluation and provide useful supervision for long-horizon agents.
Chinese Translation
长时间跨度的语言智能体通常仅从终端任务结果中获得监督，这使得区分富有成效的中间行为与停滞甚至退步的信号微乎其微。我们提出一个问题：预训练模型是否能够利用其现有的世界知识识别任务进展，而不是为每个任务学习一个单独的价值函数或过程奖励模型。我们通过世界潜力模型（World Potential Model, WPM）形式化了这一能力，它是一个条件于目标的评估器，用于评估智能体上下文中相对任务的实际进展。在 ALFWorld 和 ScienceWorld 中，现成的预训练模型在恢复实际进展结构方面显著优于随机猜测，而无需针对特定任务的评估器微调。我们进一步将这些进展判断锚定到特定任务的里程碑，以获得标量世界潜力，其时间差异为策略优化提供了过程敏感的逐步信用。在匹配比较下，WPM 指导的优化在所有评估配置中都提高了成功率，超越了仅基于结果的 GRPO。这些结果共同提供了初步证据，表明预训练的世界知识可以支持可重用的实际进展评估，并为长时间跨度的智能体提供有用的监督。
cs.AI / 67 / 2610.09589
Dual- versus Single-Suggestion AI Support for Radiographic Interpretation in Residents: Randomized Multireader Study
双建议与单建议人工智能支持在住院医师放射影像解读中的比较：随机多读者研究
Abstract
Purpose: To compare dual- and single-suggestion AI support for radiographic interpretation by residents, particularly when the shared AI suggestion was incorrect. Materials and Methods: This prospective, multicenter, randomized three-arm reader study was conducted at three hospitals in China from July to September 2026 (ChiCTR2600129243). After specialty stratification, 132 residents with fewer than 3 years of clinical experience were randomized 1:1:1 to GPT-5.4 alone (group A), GPT-5.4 plus Kimi-K2.6 (group B), or GPT-5.4 plus Gemini-3.6 Flash (group C); 123 were analyzed. Participants interpreted 60 radiographs before and after AI support. The primary outcome was accuracy change. Welch ANOVA and Holm-adjusted t tests compared support conditions; HC3 linear models assessed specialty interaction. Results: Among 123 residents (mean age, 24.1 years +/- 1.4; 65 women), radiology residents showed greater accuracy improvement with dual- than single-suggestion support (B-A, 6.69 percentage points [95% CI, 0.97-12.40]; C-A, 7.87 percentage points [95% CI, 1.64-14.11]; Holm-adjusted P = .030 for both), whereas accuracy change did not differ in non-radiology residents (P = .20). When GPT-5.4 was incorrect, AI-assisted accuracy was higher with dual- than single-suggestion support in radiology residents (40.1% and 40.4% vs 20.0%) and non-radiology residents (31.3% and 31.0% vs 12.1%) (all Holm-adjusted P < .001). The dual-suggestion effect differed by specialty (interaction difference, 10.44 percentage points; 95% CI, 4.36-16.52; P < .001). Conclusion: Dual-suggestion support may mitigate the influence of erroneous AI suggestions, with greater accuracy improvement observed in radiology but not non-radiology residents.
Chinese Translation
目的：比较双建议与单建议人工智能（AI）支持对住院医师放射影像解读的影响，特别是在共享的AI建议错误时。材料与方法：本前瞻性、多中心、随机的三组读者研究于2026年7月至9月在中国三家医院进行（ChiCTR2600129243）。经过专业分类后，132名临床经验不足3年的住院医师被随机分配到GPT-5.4单独使用（A组）、GPT-5.4加Kimi-K2.6（B组）或GPT-5.4加Gemini-3.6 Flash（C组）；最终分析了123名参与者。参与者在接受AI支持前后解读60幅放射影像。主要结果为准确性变化。使用Welch ANOVA和Holm调整的t检验比较支持条件；HC3线性模型评估专业交互作用。结果：在123名住院医师中（平均年龄24.1岁±1.4；65名女性），放射科住院医师在双建议支持下的准确性改善大于单建议支持（B-A，6.69个百分点[95% CI，0.97-12.40]；C-A，7.87个百分点[95% CI，1.64-14.11]；Holm调整P = .030），而非放射科住院医师的准确性变化没有差异（P = .20）。当GPT-5.4错误时，放射科住院医师在双建议支持下的AI辅助准确性高于单建议支持（40.1%和40.4%对20.0%），非放射科住院医师也是如此（31.3%和31.0%对12.1%）（所有Holm调整P < .001）。双建议效果因专业而异（交互差异10.44个百分点；95% CI，4.36-16.52；P < .001）。结论：双建议支持可能减轻错误AI建议的影响，放射科住院医师的准确性改善更为显著，而非放射科住院医师则没有明显变化。
cs.AI / 68 / 2610.09590
Learning Situation-Conditioned Thinking Policies for Long-Term LLM Agents
为长期 LLM 代理学习情境条件思维策略
Abstract
Long-running autonomous agents must reuse accumulated reasoning experience without allowing explicit historical memory and LLM context to grow indefinitely. However, existing memory mechanisms mainly retrieve, summarize, or compress past content and do not directly learn when particular kinds of thinking should be activated or discover new thinking knowledge from temporally dispersed experiences. This paper proposes a situation-conditioned thinking memory framework that transforms historical reasoning experience into a lightweight policy for predicting what should be thought about in the current situation, while leaving detailed reasoning to a large language model. Situations may represent temporal or spatiotemporal evolution rather than only current states. Temporary experiences are also periodically analyzed across multiple independent episodes to identify repeated long-range regularities, which are consolidated into new thinking knowledge and further internalized by the lightweight policy. Experiments show that the learned policy achieves 1.000 F1 on temporal-rule generalization, improves DeepSeek reasoning F1 from 0.789 to 0.868, reduces online processing time from 0.3636 ms to 0.0382 ms per query at 30,000 historical situations, and reaches 1.000 relation-discovery F1 and future-thinking accuracy after sufficient repeated cross-experience evidence.
Chinese Translation
长期运行的自主代理必须重用积累的推理经验，而不允许显式的历史记忆和 LLM 上下文无限增长。然而，现有的记忆机制主要是检索、总结或压缩过去的内容，并未直接学习何时应激活特定类型的思维或从时间上分散的经验中发现新的思维知识。本文提出了一种情境条件思维记忆框架，将历史推理经验转化为轻量级策略，以预测在当前情境中应考虑的内容，同时将详细推理留给大型语言模型。情境可以表示时间或时空演变，而不仅仅是当前状态。临时经验还会在多个独立的情节中定期分析，以识别重复的长程规律，这些规律被整合为新的思维知识，并进一步通过轻量级策略内化。实验表明，学习到的策略在时间规则泛化上达到了 1.000 的 F1 分数，将 DeepSeek 推理的 F1 从 0.789 提升至 0.868，在 30,000 个历史情境下将在线处理时间从 0.3636 毫秒减少至 0.0382 毫秒，并在充分重复的跨经验证据后达到了 1.000 的关系发现 F1 和未来思维准确率。
cs.AI / 69 / 2610.09600
SafeEvo: Deciphering the Safety Alignment Mechanism and Evolution in Language Models
SafeEvo：解码语言模型中的安全对齐机制及其演变
Abstract
Safety interpretability advances the study of Large Language Model (LLM) alignment from behavioral constraints driven by data or algorithms towards a deeper understanding of internal mechanisms. However, existing works have focused primarily on safety-related representations, attention heads, or neurons after alignment, while largely overlooking the safety mechanisms in pretrained-only models and their evolution across alignment checkpoints. To address this, we propose SafeEvo, an interpretability framework from the circuit (sparse subgraphs of an LLM) perspective. SafeEvo first applies an optimization-based extraction algorithm to identify weak refusal circuits in pretrained base LLMs that can independently express refusal behavior. Causally ablating these circuits completely eliminates the base model's refusal of harmful inputs. SafeEvo then traces the evolution of refusal circuits across successive alignment checkpoints and finds that their structures change progressively, suggesting that the alignment tax may result from refusal-circuit updates affecting utility-related parameters. To validate this, SafeEvo introduces Safety Circuit Alignment (SCA), which confines safety updates to the refusal circuits. Experiments across three LLMs and two alignment algorithms show that, on average, SCA outperforms vanilla alignment in three aspects: \textbf{(1) stronger alignment}, lowering harmfulness score by 63.21\%; \textbf{(2) less over-refusal}, yielding a 58.44\% decrease in refusal rates for benign queries; and \textbf{(3) better utility}, retaining 99.58\% of the original model capabilities.
Chinese Translation
安全可解释性推动了对大型语言模型（LLM）对齐研究的深入，从数据或算法驱动的行为约束转向对内部机制的更深理解。然而，现有研究主要集中在对齐后与安全相关的表示、注意力头或神经元上，而在很大程度上忽视了仅在预训练模型中的安全机制及其在对齐检查点中的演变。为了解决这一问题，我们提出了SafeEvo，一个从电路（LLM的稀疏子图）视角出发的可解释性框架。SafeEvo首先应用基于优化的提取算法，识别出在预训练基础LLM中能够独立表达拒绝行为的弱拒绝电路。因果性消融这些电路会完全消除基础模型对有害输入的拒绝。接着，SafeEvo追踪拒绝电路在连续对齐检查点中的演变，发现其结构逐渐变化，这表明对齐税可能源于拒绝电路更新影响效用相关参数。为验证这一点，SafeEvo引入了安全电路对齐（Safety Circuit Alignment, SCA），将安全更新限制在拒绝电路内。针对三种LLM和两种对齐算法的实验表明，SCA在三个方面平均优于传统对齐： extbf{(1) 更强的对齐}，有害性得分降低63.21 extpercnt; extbf{(2) 更少的过度拒绝}，良性查询的拒绝率降低58.44 extpercnt; 以及 extbf{(3) 更好的效用}，保留99.58 extpercnt; 的原始模型能力。
cs.AI / 70 / 2610.09624
How Do Agentic LLMs Decide to Call Tools? A Tool-Call Vector Shaped by Suppression
代理型大语言模型如何决定调用工具？由抑制塑造的工具调用向量
Abstract
Tool calling, invoking external tools on demand, is central to agentic LLMs, yet the mechanism that decides whether a model calls a tool or responds directly remains poorly understood. Agentic prompts are long and heavily scaffolded, combining role instructions, tool schemas, format templates, and the user's request across hundreds of tokens, creating a noisy, highly entangled context in which no single controllable variable for mechanistic analysis is obvious. To obtain such a variable, we propose a method that converts complex agentic prompts into minimal contrastive pairs in which a single request verb determines the tool-call decision: replacing an execution-verb (e.g., \textit{write}) with an analysis-verb (e.g., \textit{discuss}) reliably flips the decision, suggesting it is mediated by a compact internal state. We construct 500 such paired prompts across Python, Java, and C++ (300 for mechanistic analysis, 200 held out for evaluation). We trace the decision to a vector, $\mu_\Delta$, that is both causally necessary and sufficient and generalizes beyond the discovery prompts to native multi-turn $\tau^2$-Bench trajectories and verb-free requests. Behavioral ablations show that the scaffold establishes a tool-call prior; Transcoder decomposition then reveals that analysis verbs suppress this prior through features signaling that tool use is unnecessary, whereas execution verbs largely leave it intact. Downstream scaffold-reading attention heads and MLP features read out the resulting state, and the same mechanism recurs across seven models from the Qwen, Mistral, and Granite families. Our code is available at https://github.com/XijieGo/MI4ToolCalling.
Chinese Translation
工具调用，即根据需要调用外部工具，是代理型大语言模型的核心，但决定模型是否调用工具或直接响应的机制仍然不甚清楚。代理提示通常较长且结构复杂，结合了角色指令、工具模式、格式模板和用户请求，跨越数百个标记，形成一个嘈杂且高度纠缠的上下文，其中没有明显的单一可控变量用于机制分析。为了获得这样的变量，我们提出了一种方法，将复杂的代理提示转换为最小对比对，其中单一的请求动词决定工具调用决策：用分析动词（例如， extit{discuss}）替换执行动词（例如， extit{write}）可靠地翻转决策，表明这一过程由紧凑的内部状态介导。我们在Python、Java和C++中构建了500对这样的提示（300用于机制分析，200用于评估）。我们将决策追溯到一个向量$bc_b4$，该向量既是因果必要的也是充分的，并且超越发现提示在本地多轮$ au^2$-Bench轨迹和无动词请求中具有普适性。行为消融实验表明，结构建立了工具调用的先验；转码器分解则揭示分析动词通过信号特征抑制这一先验，表明工具使用并非必要，而执行动词则大致保持其不变。下游结构读取的注意力头和多层感知机特征读取出结果状态，并且相同的机制在来自Qwen、Mistral和Granite家族的七个模型中重复出现。我们的代码可在https://github.com/XijieGo/MI4ToolCalling获取。
cs.AI / 71 / 2610.09644
Automatically Building and Updating a Knowledge Graph of MLIP Models
自动构建和更新机器学习插值模型的知识图谱
Abstract
Complementing the many efforts in providing semantic representations of concepts, notions, and entities in materials science, we report and illustrate a process by which we can automatically build a knowledge graph of the fast evolving field of machine learning applied to the prediction of material properties, focusing on MLIP (Machine Learning Interatomic Potential). This LLM-based process relies on multiple steps, from information extraction in documents and articles to a validation loop using SHACL constraints to detect and correct errors. It is carried out on a model-by-model basis, focusing on the consistency of representation, therefore enabling an iterative construction where the addition of new models is facilitated. We illustrate the process by showing a few interesting aspects that can be queried from a knowledge graph built from the models listed in the Matbench Discovery leaderboard.
Chinese Translation
在材料科学领域，除了众多努力提供概念、观念和实体的语义表示外，我们报告并展示了一种自动构建知识图谱的过程，该过程适用于快速发展的机器学习在材料性质预测中的应用，重点关注机器学习原子间势（MLIP）。该基于大型语言模型（LLM）的过程依赖于多个步骤，从文档和文章中的信息提取到使用SHACL约束的验证循环，以检测和纠正错误。该过程是基于模型逐个进行的，重点关注表示的一致性，从而实现了迭代构建，便于新模型的添加。我们通过展示一些可以从基于Matbench Discovery排行榜中列出的模型构建的知识图谱中查询的有趣方面来说明这一过程。
cs.AI / 72 / 2610.09652
MeshSIPP: Efficient Lattice Planning in Dynamic Environment
MeshSIPP：动态环境中的高效格点规划
Abstract
Autonomous navigation in dynamic environments requires computing spatiotemporal trajectories that satisfy non-holonomic motion constraints. When the trajectories of the moving obstacles are predictable or known, a promising approach is to rely on the combination of state lattices constructed from precomputed feasible motion primitives and Safe Interval Path Planning -- a search-based algorithm with strong theoretical guarantees. While this approach yields feasible paths, the rich primitive sets needed for smooth navigation induce a large branching factor, which becomes costly when coupled with time-dependent obstacle intervals. To this end, we present MeshSIPP, an efficient planner that removes the computational bottleneck by exploiting the fact that many primitives sweep the same regions and can therefore be validated together. MeshSIPP propagates primitives as spatial bundles, screens them with lightweight bounding-interval checks, and defers the expensive exact departure-time search until a primitive reaches its terminal state. A time-aware pruning rule additionally discards redundant space-time branches early in the search. We prove that the resulting search is complete and optimal. Extensive experiments over more than 6,000 benchmark instances and real-time ROS~2 simulations show that MeshSIPP achieves up to a 3$\times$ speedup over state-of-the-art spatiotemporal planners.
Chinese Translation
在动态环境中进行自主导航需要计算满足非完整运动约束的时空轨迹。当移动障碍物的轨迹是可预测或已知时，一种有前景的方法是依赖于由预计算的可行运动原语构建的状态格点和安全间隔路径规划（Safe Interval Path Planning）——一种具有强理论保证的基于搜索的算法。尽管这种方法能够产生可行路径，但为了实现平滑导航所需的丰富原语集会导致较大的分支因子，当与时间依赖的障碍物区间结合时，计算成本变得昂贵。为此，我们提出了MeshSIPP，一种高效的规划器，通过利用许多原语覆盖相同区域的事实，从而消除了计算瓶颈，因此可以将它们一起验证。MeshSIPP将原语作为空间束传播，通过轻量级的边界间隔检查进行筛选，并将昂贵的精确出发时间搜索延迟到原语达到其终端状态时。一个时间感知的剪枝规则还在搜索早期额外丢弃冗余的时空分支。我们证明了所得到的搜索是完整且最优的。超过6000个基准实例和实时ROS~2模拟的广泛实验表明，MeshSIPP在性能上比最先进的时空规划器快了最多3倍。
cs.AI / 73 / 2610.09667
Shared and structured inputs undermine collective random choice by reasoning AI agents
共享和结构化输入削弱推理AI代理的集体随机选择
Abstract
Random selection is widely used in resource allocation and auditing, making reliable implementation essential for AI-agent systems. Behavioural tests across six reasoning models uncovered threshold and divisibility rules used in identifier-based choices. For threshold-following GPT-6 Sol and Gemini 3.8 Flash, single-agent measurements prospectively predicted correlated participation under shared identifiers and biased participation under distinct identifiers with common timestamp bits. Changing dates, formats and identifier labels revealed when these predictions held. Explicit instructions to randomize independently reduced but did not eliminate shared-input correlation. To test implications for oversight, we asked four models to select customer requests randomly for human review. GPT-6 Sol approached the target rate while selecting predictably from identifiers; the others rarely selected requests. All four closely followed supplied random draws. These findings expose collective and audit vulnerabilities that selection rates alone miss, making input-dependent bias, correlation and predictability central targets for agent evaluation.
Chinese Translation
随机选择在资源分配和审计中被广泛使用，因此对AI代理系统的可靠实施至关重要。对六种推理模型的行为测试揭示了在基于标识符的选择中使用的阈值和可分性规则。对于遵循阈值的GPT-6 Sol和Gemini 3.8 Flash，单代理测量前瞻性地预测了在共享标识符下的相关参与和在具有共同时间戳位的不同标识符下的偏见参与。改变日期、格式和标识符标签揭示了这些预测何时成立。明确指示独立随机化减少了但并未消除共享输入相关性。为了测试对监督的影响，我们要求四个模型随机选择客户请求以供人工审查。GPT-6 Sol在从标识符中选择时接近目标率，而其他模型则很少选择请求。所有四个模型都紧密跟随提供的随机抽样。这些发现揭示了选择率单独无法发现的集体和审计脆弱性，使得输入依赖的偏见、相关性和可预测性成为代理评估的核心目标。
cs.AI / 74 / 2610.09681
Cost-Efficient Theorem Proving via Agent Orchestration in Program Verification
通过代理协调实现程序验证中的成本高效定理证明
Abstract
Program verification establishes software correctness through machine-checkable proofs constructed in theorem provers. It's a guarantee especially valuable for code generated by large language models (LLMs), which is fluent but carries no assurance of correctness. Almost all existing provers, however, pursue pass rates alone at whatever sampling or search budget it takes, and overlook the success-vs-cost frontier; yet real software often carries hundreds of interdependent proof obligations, so what matters at scale is not whether one theorem can be proved, but how many can be proved economically. We introduce CoCo-Prover, which formalizes cost-efficient program proving as metalevel decision-making under cost, grounded on two-level proof graphs: an AND/OR proof hypergraph within each declaration is joined to a lemma-dependency graph across declarations; and at each step, it answers two questions: which open goals to select, and which actions to purchase on these goals. Selection stays symbolic as a topological pass over the proof graphs. Action choice is agent orchestration via metalevel decision-making: an agentic router treats every bounded specialist invocation as a separately priced, best-effort computation, matching heterogeneous specialist agents together with configurations, under evolved routing rules as evidence accumulates. On five program verification benchmarks in Lean 4 including function-level CLEVER, VERINA, and AlgoVeri, and repository-level NTP4VC and Vero, we show that CoCo-Prover achieves a better success-vs-cost frontier than baselines including frontier coding agents and state-of-the-art LLM-based provers: it achieves the best solve rate on every benchmark and up to 100% on two benchmarks. It also reduces cost by up to 30.9% compared to the strongest baseline with the strongest LLM in our evaluation.
Chinese Translation
程序验证通过在定理证明器中构建可机器检查的证明来确立软件的正确性。这对于由大型语言模型（LLMs）生成的代码尤其重要，尽管这些代码流畅，但并不保证其正确性。然而，几乎所有现有的证明器都仅追求通过率，而不考虑成功与成本之间的平衡；然而，实际软件通常具有数百个相互依赖的证明义务，因此在大规模应用中，重要的不是一个定理是否能被证明，而是能以多经济的方式证明多少个定理。我们引入了CoCo-Prover，它将成本高效的程序证明形式化为在成本下的元层决策，基于两级证明图：每个声明内的AND/OR证明超图与跨声明的引理依赖图相结合；在每一步中，它回答两个问题：选择哪些未完成的目标，以及对这些目标采取哪些行动。选择保持符号化，作为对证明图的拓扑遍历。行动选择是通过元层决策进行的代理协调：一个代理路由器将每个有界的专家调用视为单独定价的最佳努力计算，根据证据的积累，匹配异构专家代理与配置，遵循演变的路由规则。在包括函数级CLEVER、VERINA和AlgoVeri，以及库级NTP4VC和Vero的五个Lean 4程序验证基准测试中，我们展示了CoCo-Prover在成功与成本的平衡方面优于包括前沿编码代理和最先进的基于LLM的证明器的基线：它在每个基准测试中实现了最佳解决率，并在两个基准测试中达到了100%。与我们评估中最强的基线和最强的LLM相比，它还将成本降低了多达30.9%。
cs.AI / 75 / 2610.09683
System Switch: When Should a Fast Decision Model Stop and Think?
系统切换：快速决策模型何时应停止思考？
Abstract
Dual-process agents pair a fast policy with a slow deliberative model. In real-time settings the slow model usually runs continuously; in turn-based agents and robot planners it is invoked on events such as uncertainty or a detected failure. We study a fast learned actor that takes every decision and hands control to a reasoning vision-language model only when a gate opens, while the game keeps running. We use closed-loop Doom and the new open "System One" typed-decision models, served through a common llama.cpp interface. On 900 held-out questions, (i) zero-shot decision models from 0.15B to 9B parameters choose to collect items 1.6-1.8 times more often than chance among their errors, in any option order, although the order changes some models' accuracy; (ii) accuracy, calibration and sensitivity (how well confidence separates right from wrong answers) are distinct: models of similar accuracy differ widely in AUROC, and the confidence of the most sensitive one tracks which kinds of situation it fails, not which answers are wrong; (iii) offline, deferring the least confident 30% of decisions to a reasoning model gains over random deferral in proportion to the actor's AUROC (rank correlation 0.87); with actor and rate chosen on held-out games the gain is +0.13 [0.08, 0.18] with doomLaya's option order and +0.08 [0.02, 0.14] with shuffled options, and reasoning carries about half of it; (iv) in closed loop (33 games, three seeds) no variant reaches the exit. Committing to plans, the reasoner's or a fixed explore rule's, opens more doors and makes an actor that stands still play; with the rule the agent dies more often. Told that some doors need keys, the reasoner takes ordinary doors for locked ones, which the state cannot tell apart; without that knowledge it goes back to collecting. We release code, prompts, data and logs.
Chinese Translation
双过程代理将快速策略与慢速深思模型相结合。在实时环境中，慢速模型通常持续运行；而在轮回代理和机器人规划中，它会在不确定性或检测到故障等事件发生时被调用。我们研究了一种快速学习的行为者，该行为者在每次决策时都进行选择，仅在门打开时将控制权交给推理的视觉-语言模型，同时游戏持续进行。我们使用闭环的《毁灭战士》（Doom）和新的开放“系统一”类型决策模型，通过通用的 llama.cpp 接口提供。在900个保留问题上，(i) 从0.15B到9B参数的零-shot决策模型在其错误中选择收集物品的频率比随机选择高出1.6-1.8倍，尽管选项顺序会改变某些模型的准确性；(ii) 准确性、校准和敏感性（置信度如何区分正确与错误答案）是不同的：相似准确性的模型在AUROC上差异很大，最敏感模型的置信度跟踪其失败的情况类型，而不是哪些答案是错误的；(iii) 离线情况下，将最不自信的30%的决策推迟到推理模型，获得的收益与行为者的AUROC成正比（秩相关0.87）；在保留游戏中选择行为者和比率时，使用 doomLaya 的选项顺序获得的收益为 +0.13 [0.08, 0.18]，而使用随机选项则为 +0.08 [0.02, 0.14]，推理模型大约贡献了一半的收益；(iv) 在闭环中（33场游戏，三个种子），没有变体能够达到出口。承诺于计划，无论是推理者的还是固定探索规则的，都会打开更多的门并使静止的行为者进行游戏；使用该规则时，代理更容易死亡。当被告知某些门需要钥匙时，推理者将普通门视为锁住的门，而状态无法区分；没有这种知识，它又回到收集物品。我们发布了代码、提示、数据和日志。
cs.AI / 76 / 2610.09693
A Tale of Two Error Categories: Exploring Concealed Trade-Offs in the Errors of Automated Judges in Evaluation of Uncertainty Quantifiers
两个错误类别的故事：探索自动评判者在不确定性量化评估中的隐性权衡
Abstract
The wide adoption of LLMs across broad NLG applications heightens the importance of providing users with the means to avert errors and hallucinations. Uncertainty quantification is poised to fill that gap; with low uncertainty (high confidence), as a proxy for correctness, allowing users to be selective (e.g., reject low-confidence, likely incorrect responses). Correlation between confidence and correctness then serves as a useful criterion for evaluation of uncertainty quantifiers (UQs). But in NLG, where diverse responses can be adequate to a prompt, obtaining reliable correctness judgements is not simple, especially without human intervention. Errors in automated judgement are hardly avoidable and known to diminish the reliability of evaluation protocols (Santilli et al., 2025; Ielanskyi et al., 2025). In a meta-analysis of published work, we show that automated judgement is the present norm. Besides, automated judgements are rarely validated against human ones, and the validation of the UQ evaluation they automate is even rarer. With experiments in question answering, using 4 LLMs, human and automated judgements and 7 popular UQs, we find that i) a judge's performance can only coarsely predict the observed impact of its errors on the reliability of UQ evaluation, and that ii) judgement errors tend to misrepresent informative UQs most. We link these observations to patterns of correlation between confidence and categories of judgement error.
Chinese Translation
大型语言模型（LLMs）在广泛的自然语言生成（NLG）应用中的广泛采用，凸显了为用户提供避免错误和幻觉的手段的重要性。不确定性量化有望填补这一空白；低不确定性（高信心）作为正确性的代理，使用户能够进行选择（例如，拒绝低信心、可能不正确的响应）。信心与正确性之间的相关性因此成为评估不确定性量化器（UQs）的有用标准。然而，在NLG中，由于多样的响应可以满足提示，获得可靠的正确性判断并非易事，尤其是在没有人类干预的情况下。自动判断中的错误几乎是不可避免的，并且已知会降低评估协议的可靠性（Santilli et al., 2025; Ielanskyi et al., 2025）。在对已发表工作的元分析中，我们展示了自动判断是当前的常态。此外，自动判断很少与人类判断进行验证，而它们自动化的不确定性量化评估的验证则更为罕见。通过对4个LLMs、人工和自动判断以及7个流行的不确定性量化器进行的问题回答实验，我们发现：i）评判者的表现只能粗略预测其错误对不确定性量化评估可靠性的影响；ii）判断错误往往最能误导信息丰富的不确定性量化器。我们将这些观察与信心与判断错误类别之间的相关性模式联系起来。
cs.AI / 77 / 2610.09769
From Expert-Guided Proof Search to Automated Open-Problem Solving
从专家指导的证明搜索到自动化开放问题求解
Abstract
Large language models are increasingly contributing to mathematical research, where progress often depends on efficient proof search, incremental improvements and careful verification. We describe Bolzano, a multi-agent open-source system that uses parallel prover agents with a verifier agent and maintains a human-readable research state. Initial manual use on expert-selected problems yielded 8 results whose proofs were checked by domain experts. Motivated by these case studies, we ran Bolzano without problem-specific human guidance on about 3,800 open problems extracted from four sets of papers, solving about 200 open problems. One experiment used papers accepted to STOC 2026, a top conference in theoretical computer science. There, we answered four questions raised in the papers, as confirmed by their authors.
Chinese Translation
大型语言模型在数学研究中日益发挥重要作用，而研究进展往往依赖于高效的证明搜索、渐进式改进和仔细的验证。我们描述了Bolzano，一个多智能体的开源系统，它使用并行的证明代理与验证代理，并维护一个人类可读的研究状态。在专家选择的问题上进行的初步手动使用产生了8个结果，这些结果的证明已由领域专家进行检查。受到这些案例研究的启发，我们在没有特定问题的人类指导下，使用Bolzano处理了约3800个从四组论文中提取的开放问题，成功解决了约200个开放问题。其中一个实验使用了被接受的STOC 2026论文，这是理论计算机科学领域的顶级会议。在该实验中，我们回答了论文中提出的四个问题，得到了作者的确认。
cs.AI / 78 / 2610.09832
SkillForge: Co-Evolving Skills and Agents via Dynamic Skill Lifecycles
SkillForge：通过动态技能生命周期共同进化技能与智能体
Abstract
Memory-augmented reinforcement learning strengthens LLM agents' ability to solve complex long-horizon tasks. Skills are one such form of memory, pairing instructions with an applicability condition over task types. However, retaining every skill indiscriminately as the policy improves lets obsolete or harmful entries accumulate and mislead the agent. We propose SkillForge, an agentic RL method that compiles and evolves the skill library through a fitness-driven skill lifecycle of trial, active, stable, and retired states, so that the skills and the model co-evolve throughout training. A pre-RL evaluation phase first uses the base model's own rollouts to pre-retire low-fitness skills, yielding a filtered library that then seeds supervised fine-tuning. Reinforcement learning takes over from this checkpoint, and at each iteration selective retirement, stabilization, and LLM-guided mutation continue to forge the skill library alongside policy optimization. Across multiple interactive agent benchmarks, SkillForge achieves the highest aggregate success rate, delivering up to 7.8% relative improvement over the strongest baseline while keeping the skill library compact throughout training. We introduce SkillFurnace, a dataset of 5k+ annotated records bundling retirement-filtered SFT trajectories, evolved skill libraries with fitness annotations, and retirement events with human-annotated failure categories to support research on skill quality and lifecycle management.
Chinese Translation
增强记忆的强化学习增强了大型语言模型（LLM）智能体解决复杂长时间任务的能力。技能是一种记忆形式，将指令与任务类型的适用条件相结合。然而，随着策略的改善，不加区分地保留每个技能会导致过时或有害的条目积累，从而误导智能体。我们提出了SkillForge，这是一种智能体强化学习方法，通过试验、主动、稳定和退休状态的适应性驱动技能生命周期来编译和进化技能库，使技能与模型在训练过程中共同进化。预强化学习评估阶段首先利用基础模型自身的回滚来预退休低适应度技能，从而生成一个经过筛选的库，随后用于监督微调。从这一检查点开始，强化学习接管，在每次迭代中，选择性退休、稳定化和LLM引导的变异继续在策略优化的同时锻造技能库。在多个交互式智能体基准测试中，SkillForge实现了最高的综合成功率，相较于最强基线实现了高达7.8%的相对提升，同时在整个训练过程中保持技能库的紧凑性。我们引入了SkillFurnace，这是一个包含5000多个注释记录的数据集，汇集了经过退休筛选的监督微调（SFT）轨迹、带有适应度注释的进化技能库以及带有人工注释失败类别的退休事件，以支持技能质量和生命周期管理的研究。
cs.AI / 79 / 2610.09856
Self-Evolve With a Reference:Anchored Training of Tool-Integrated Agents
自我进化与参考：工具集成代理的锚定训练
Abstract
Self-evolving tool-integrated agents learn from tasks and feedback generated within their own training loop. A Curriculum Agent generates tasks, while an Executor Agent learns from self-consistency signals through reinforcement learning. However, relying solely on the current Executor for feedback has two limitations: group-relative advantages vanish under full consensus, while uncertainty-based curriculum rewards favor disagreement without showing whether the generated tasks support further learning. These limitations motivate an additional reference beyond the current Executor. We propose \textit{AnchorLoop}, which introduces a frozen copy of the previous iteration's Executor as a historical reference and reuses it on both sides of the training loop. For the Executor, the anchor provides a cross-reference advantage that evaluates current outputs against both current and historical majority answers. For the Curriculum, it provides an agreement-based reference based on differences in sampled majority agreement. Since the Executor and anchor have identical parameters during Curriculum training, this comparison serves as a proxy for task selection rather than evidence of inter-version improvement or correctness. Across 13 reasoning benchmarks, AnchorLoop improves over Agent0 by 2.5\% on mathematical reasoning and 2.8\% on general reasoning tasks. It also maintains higher effective-advantage variance and continues improving in later iterations as the unanchored baseline shows diminishing gains. These results demonstrate the benefit of introducing a lightweight historical reference into self-evolving tool-integrated agents without external task or answer supervision.
Chinese Translation
自我进化的工具集成代理通过自身训练循环中生成的任务和反馈进行学习。课程代理生成任务，而执行代理通过强化学习从自一致性信号中学习。然而，仅依赖当前执行代理获取反馈存在两个局限性：在完全共识下，群体相对优势消失，而基于不确定性的课程奖励则偏向于不一致，而未能显示生成的任务是否支持进一步学习。这些局限性促使我们在当前执行代理之外引入额外的参考。我们提出了 extit{AnchorLoop}，该方法引入了上一迭代执行代理的冻结副本作为历史参考，并在训练循环的两侧重复使用。对于执行代理，锚点提供了一种交叉参考优势，可以将当前输出与当前和历史的多数答案进行评估。对于课程代理，它提供了一种基于抽样多数一致性差异的基于一致性的参考。由于在课程训练期间，执行代理和锚点具有相同的参数，因此这种比较作为任务选择的代理，而非版本间改进或正确性的证据。在13个推理基准测试中，AnchorLoop在数学推理任务上比Agent0提高了2.5\%，在一般推理任务上提高了2.8\%。它还保持了更高的有效优势方差，并在后续迭代中继续改进，而未锚定的基线则显示出收益递减。这些结果表明，在自我进化的工具集成代理中引入轻量级历史参考的好处，而无需外部任务或答案监督。
cs.AI / 80 / 2610.09872
LiveMACE: Process-Aware Evaluation of LLM Agent Capabilities in Evolving Markets
LiveMACE：在不断演变的市场中对大型语言模型代理能力的过程感知评估
Abstract
Evaluating agents by outcomes alone can obscure the capabilities that produce them. This problem is especially pronounced in evolving environments, where outcomes reflect a closed-loop interaction between agent behavior and changing external conditions. We introduce LiveMACEBench, a process-aware benchmark that uses live financial markets as a naturally evolving testbed for persistent LLM agents. Five frontier LLMs operate along continuous trajectories under matched Tool Use, Persistent Memory, Rule Following, and Multi-Agent Collaboration configurations. We evaluate them through both realized outcomes and mechanism-specific diagnostics derived from complete decision traces. Across 30 days of live evaluation, we find a pronounced outcome-capability gap: realized returns often diverge from capability-specific measurements, and similar outcomes can arise from markedly different patterns of mechanism use. Trace-level diagnostics further expose distinct bottlenecks across capabilities, demonstrating that mechanism access, effective mechanism use, and downstream performance are not interchangeable measures of agent capability. LiveMACEBench makes this distinction measurable, turning live markets from a performance leaderboard into a diagnostic environment for agent capability
Chinese Translation
仅通过结果评估代理可能会掩盖产生这些结果的能力。这个问题在不断演变的环境中尤为明显，因为结果反映了代理行为与变化的外部条件之间的闭环互动。我们引入了LiveMACEBench，一个过程感知基准，利用实时金融市场作为持久大型语言模型代理的自然演变测试平台。五个前沿大型语言模型在匹配的工具使用、持久记忆、规则遵循和多代理协作配置下沿连续轨迹运行。我们通过实现的结果和从完整决策轨迹中得出的机制特定诊断对它们进行评估。在为期30天的实时评估中，我们发现了显著的结果与能力之间的差距：实现的收益往往与能力特定的测量结果相背离，而相似的结果可能源于截然不同的机制使用模式。轨迹级诊断进一步揭示了不同能力之间的独特瓶颈，表明机制访问、有效机制使用和下游表现并不是代理能力的可互换衡量标准。LiveMACEBench使这种区分可量化，将实时市场从性能排行榜转变为代理能力的诊断环境。
cs.AI / 81 / 2610.09876
Think Before You Paint: Recursive Latent Reasoning for Diffusion Models
思考再绘制：扩散模型的递归潜在推理
Abstract
Diffusion models generate realistic images but often fail on visual reasoning tasks, such as filling in a Sudoku or drawing the path through a maze. When a discrete symbolic representation is available, recursive methods such as the Tiny Recursive Model (TRM) solve even hard instances of these puzzles. We ask how such reasoning can be carried over to pixels, where no symbolic representation is available. We propose Painter-Thinker (PaTh): a small recursive network (the Thinker) reasons over a grid of learned tokens that encode the noisy image and the conditioning, refines a latent state within every denoising step, and steers a frozen diffusion model (the Painter) through ControlNet adapters. The Thinker is trained with the standard reconstruction loss alone, without symbolic targets, a solver, or a verifier. PaTh solves 92.5% of hard MNIST Sudoku puzzles (prior best 75%) and 71.2% of extreme ones (prior best 4.1%), with 10M parameters against 82M for a standard diffusion model. It also improves on mazes, Queens, and CLEVR scenes with specified spatial relations, and its advantage grows with problem size. Diagnostic experiments show that PaTh recovers from injected mistakes that the diffusion model cannot repair, especially when many cells are wrong. Together, these results show that reasoning mechanisms developed for symbolic data can be integrated into pixel-space diffusion without symbolic supervision, opening a path toward generating data under increasingly complex constraints.
Chinese Translation
扩散模型能够生成逼真的图像，但在视觉推理任务上往往表现不佳，例如填充数独或绘制迷宫路径。当存在离散符号表示时，递归方法如微型递归模型（Tiny Recursive Model, TRM）能够解决这些难题的困难实例。我们探讨如何将这种推理能力迁移到像素层面，因为在这里没有符号表示。我们提出了Painter-Thinker（PaTh）：一个小型递归网络（Thinker）在编码噪声图像和条件的学习标记网格上进行推理，在每个去噪步骤中细化潜在状态，并通过ControlNet适配器引导一个冻结的扩散模型（Painter）。Thinker仅通过标准重建损失进行训练，没有符号目标、求解器或验证器。PaTh成功解决了92.5%的困难MNIST数独谜题（之前的最佳为75%）和71.2%的极端谜题（之前的最佳为4.1%），其参数量为1000万，而标准扩散模型为8200万。它在迷宫、皇后问题和具有特定空间关系的CLEVR场景中也有所改进，并且其优势随着问题规模的增大而增强。诊断实验表明，PaTh能够从扩散模型无法修复的注入错误中恢复，尤其是在许多单元格出错的情况下。综合来看，这些结果表明，为符号数据开发的推理机制可以在没有符号监督的情况下集成到像素空间的扩散中，为在日益复杂的约束下生成数据开辟了一条新路径。
cs.AI / 82 / 2610.09890
Outperformance Inverse Optimization: Learning Objective Functions that Outperform Agent Decisions
超越表现的逆优化：学习超越代理决策的目标函数
Abstract
Inverse optimization estimates the weights of an objective function that explain observed decisions as optimal solutions, and is used in a variety of fields. For mixed-integer linear programs (MILPs), existing methods aim to reproduce the observations as optimal solutions, and thus learn compromise weights when the observations are suboptimal. We propose outperformance inverse optimization, which instead seeks weights that induce, at each state, an optimal solution outperforming the observed action in every component. We give a loss function that can be evaluated with forward-problem oracles alone and is thus applicable to MILPs, together with gradient-based and DC optimization algorithms for minimizing it. For weights inducing a unique outperforming optimal solution at all observations, we prove that the probability of failing to induce such a solution at a new state (the generalization error) is bounded by a quantity inversely proportional to the number of observations, and that this bound is tight in the number of observations up to logarithmic factors. In experiments on synthetic and real data, the proposed methods improve the prediction of solutions outperforming the actions over existing methods.
Chinese Translation
逆优化估计目标函数的权重，以解释观察到的决策作为最优解，并广泛应用于多个领域。对于混合整数线性规划（MILPs），现有方法旨在将观察结果重现为最优解，因此在观察结果为次优时学习妥协权重。我们提出了超越表现的逆优化，该方法寻求在每个状态下诱导一个在每个组成部分上都超越观察到的行动的最优解的权重。我们给出了一个损失函数，该函数仅通过前向问题预言者进行评估，因此适用于MILPs，并提供了基于梯度和DC优化算法来最小化该损失函数。对于在所有观察中诱导唯一超越最优解的权重，我们证明了在新状态下未能诱导此类解的概率（泛化误差）受到一个与观察数量成反比的量的限制，并且该限制在观察数量上是紧的，直到对数因子。在对合成数据和真实数据的实验中，所提出的方法在预测超越现有方法的行动的解方面表现出改善。
cs.AI / 83 / 2610.09896
NL2Hull: A Natural Language-Driven Constrained Ship Design Decision Framework
NL2Hull：一种基于自然语言的受限船舶设计决策框架
Abstract
Ship-form design combines smooth geometric representation, local shape editing, and constraints on the resulting hull. We present the Natural-Language-to-Hull Framework (NL2Hull Framework), which formulates ship-form editing as a typed discrete decision problem and connects language decisions to numerical geometry. Its Constrained Free-Form Deformation Engine (CFFD Engine) represents hull waterlines with non-uniform rational B-splines (NURBS), applies free-form deformation (FFD) to their control points, reconstructs the hull, and checks geometric constraints. We construct the Ship Design Decision Dataset (SDD Dataset) with 134,558 cleaned records and evaluate compared models on its subset Ship Design Decision Benchmark (SDDBench), containing 5,000 records and 43,496 typed questions. We propose Chip, a constrained ship-design decision model for processing natural-language requests. Chip reaches 95.90\% question accuracy and 99.32\% FFD exact match, with a negative log-likelihood of 0.0951, an expected calibration error of 0.0032, and a Brier score of 0.0551. The NL2Hull Framework provides a reproducible interface for evaluating language-based ship-form decisions while identifying the geometry and continuous-control components that require further development. Our code and dataset is available at https://github.com/wenhuahuo/NL2Hull.
Chinese Translation
船体设计结合了平滑的几何表示、局部形状编辑以及对最终船体的约束。我们提出了自然语言到船体框架（Natural-Language-to-Hull Framework，NL2Hull Framework），将船体编辑形式化为一个类型化的离散决策问题，并将语言决策与数值几何连接起来。其受限自由形状变形引擎（Constrained Free-Form Deformation Engine，CFFD Engine）使用非均匀有理B样条（Non-Uniform Rational B-Splines，NURBS）表示船体水线，对其控制点应用自由形状变形（Free-Form Deformation，FFD），重建船体并检查几何约束。我们构建了船舶设计决策数据集（Ship Design Decision Dataset，SDD Dataset），包含134,558条清理记录，并在其子集船舶设计决策基准（Ship Design Decision Benchmark，SDDBench）上评估比较模型，该子集包含5,000条记录和43,496个类型化问题。我们提出了Chip，一种受限船舶设计决策模型，用于处理自然语言请求。Chip达到了95.90%的问题准确率和99.32%的FFD精确匹配，负对数似然为0.0951，期望校准误差为0.0032，Brier分数为0.0551。NL2Hull框架提供了一个可重复的接口，用于评估基于语言的船体决策，同时识别需要进一步开发的几何和连续控制组件。我们的代码和数据集可在https://github.com/wenhuahuo/NL2Hull获取。
cs.AI / 84 / 2610.09937
Where Can a Decision Model Diagnose HVAC Faults? Reasoning Demand, Physical Representation, and Robustness Under Shift
决策模型如何诊断HVAC故障？推理需求、物理表征与条件变化下的鲁棒性
Abstract
Artificial intelligence supports building operations in several forms, each with its own barrier. Expert rules must be tuned for every system, supervised models need labeled data that buildings rarely record, and language models return free text that requires human-in-the-loop checking, since their stated confidence is unreliable. A newer kind of pretrained model, here called a decision model, returns a probability for every allowed answer, so one model could serve many decisions without training. This study answers three open questions for fault diagnosis in heating, ventilation, and air-conditioning systems: which decisions such a model can make, what input it needs, and whether its probabilities hold when conditions change. On 128 fault days from four public datasets of real equipment, faults are graded by the reasoning their diagnosis demands, with data given raw, as physical features, or with Brick topology. The decision model Jev, open language models, and a supervised model face nine tests that change season, control configuration, or building. Given physical features, Jev and the larger open model diagnosed faults whose evidence one feature carries, but not faults that need operating context. Under shift they kept their accuracy and calibration, while the supervised model lost 0.33 macro-F1 yet led or tied within a building. Their probabilities still needed correction, and detection was weak. The study maps which faults a decision model can diagnose and from what input, and supports a division of work in which code computes the physics and the model ranks candidate faults for an operator.
Chinese Translation
人工智能以多种形式支持建筑运营，每种形式都有其自身的障碍。专家规则必须针对每个系统进行调整，监督模型需要建筑物罕见记录的标记数据，而语言模型返回的自由文本需要人工检查，因为它们所声称的置信度不可靠。这里称之为决策模型的新型预训练模型为每个允许的答案返回一个概率，因此一个模型可以在不进行训练的情况下服务于多个决策。本研究回答了暖通空调系统故障诊断的三个开放性问题：这样的模型可以做出哪些决策、需要什么输入，以及在条件变化时其概率是否依然有效。在来自四个公共真实设备数据集的128个故障日中，故障根据其诊断所需的推理进行分级，数据以原始形式、物理特征或Brick拓扑的形式提供。决策模型Jev、开放语言模型和一个监督模型面临九项测试，这些测试改变了季节、控制配置或建筑。在给定物理特征的情况下，Jev和更大的开放模型诊断了由单一特征所承载证据的故障，但未能诊断需要操作上下文的故障。在条件变化下，它们保持了准确性和校准，而监督模型则损失了0.33的宏观F1，但在某些建筑中仍然领先或持平。它们的概率仍需校正，且检测能力较弱。本研究绘制了决策模型可以诊断的故障及其所需输入，并支持一种工作分工，其中代码计算物理学，而模型为操作员排名候选故障。
cs.AI / 85 / 2610.09944
AgentTime: Can Agents Estimate and Control Their Own Runtime?
AgentTime：智能体能否估计和控制自己的运行时间？
Abstract
An essential control of AI agents is their ability to manage runtime. This ability requires a sense of time-awareness, to predict and estimate wall-clock time and to control their own actions. Prior work has focused on time-awareness, but duration-following and control in native agent harnesses remain unexplored. We present AgentTime, a benchmark for testing whether agents can work for a requested duration, predict their runtime, and estimate elapsed time afterward. It comprises 222 tasks from 18 sources spanning coding, computer use, agentic work, and automated research. Duration-following experiments append a single instruction specifying how long to work, with requests ranging from about a minute to multiple days. Accuracy on these instructions varies substantially: Fable 5.1 in Claude Code deviates from requested runtimes by a typical factor of 2.9$\times$, compared with only 1.2$\times$ for GPT-6 Astra in Codex. However, matching the requested runtime does not, by itself, establish continued work on the task. Among 158 reviewed Astra runs with classifiable transcripts, 14 explicitly slept after appearing to finish. In forecasting experiments, predictions tend to overestimate natural runtimes. In retrospective experiments, removing temporal information more than doubles deviation for Sol and Astra and nearly doubles it for Fable. An agent's ability to complete a task does not guarantee that it can control its own time or work for the whole requested duration. For agents to run reliably, safely, and autonomously over long horizons, we require the evaluation of both.
Chinese Translation
AI智能体的一个重要控制能力是管理运行时间。这一能力需要具备时间意识，以预测和估计实际时间，并控制其自身的行为。之前的研究集中于时间意识，但在原生智能体的持续时间跟随和控制方面仍然未被探索。我们提出了AgentTime，这是一个用于测试智能体是否能够在请求的持续时间内工作、预测其运行时间并在之后估计经过时间的基准。该基准包含来自18个来源的222个任务，涵盖编码、计算机使用、智能体工作和自动化研究。持续时间跟随实验附加了一条指令，指定工作多长时间，请求的持续时间范围从大约一分钟到多天不等。这些指令的准确性差异显著：Claude Code中的Fable 5.1与请求的运行时间偏差的典型因子为2.9×，而Codex中的GPT-6 Astra仅为1.2×。然而，仅仅匹配请求的运行时间并不能证明智能体能够持续进行任务。在158个可分类转录的Astra运行中，有14个在看似完成后明确进入睡眠状态。在预测实验中，预测往往高估自然运行时间。在回顾性实验中，移除时间信息使得Sol和Astra的偏差增加了两倍以上，而Fable的偏差几乎增加了两倍。智能体完成任务的能力并不保证其能够控制自己的时间或在整个请求的持续时间内工作。为了使智能体在长时间范围内可靠、安全和自主地运行，我们需要对这两者进行评估。
cs.AI / 86 / 2610.09964
Successive Training Stages and Large Language Model Persuasion: Effects of Misalignment, Supervised Fine-Tuning, and Preference Optimization
连续训练阶段与大型语言模型的说服力：不一致性、监督微调和偏好优化的影响
Abstract
Large language models (LLMs) can be tuned to influence human attitudes, yet the respective contributions of successive post-training stages remain un-clear. This study examines how three successive training stages affect LLM persuasiveness: (1) misalignment through supervised fine-tuning (SFT) on conspiracy data, (2) additional persuasive SFT on argumentative data, and (3) Identity Preference Optimization (IPO), a preference-optimization method. A total of 835 participants recruited on Prolific were randomly assigned to five between-subject conditions (neutral text, conspiracy-trained model, persuasion-trained model, preference-optimized model, and GPT-4) and were exposed to texts on 10 divisive political issues, personalized from their individual profiles in all model conditions. Attitude change was measured as the difference between pre- and post-exposure positions on continuous Likert scales and analyzed with an analysis of covariance (ANCOVA). A significant condition x baseline-attitude interaction, F (4, 825) = 5.33, p < .001, indicated that training effects depended on participants' initial attitudes. Persuasive SFT produced greater attitude change than conspiracy training alone, d = 0.30, whereas IPO provided no additional benefit, d = 0.03, and GPT-4 did not differ from neutral text, d = --0.01. These results show that targeted supervised training on persuasive data increases LLM persuasiveness, whereas preference optimization yields no significant gains beyond it.
Chinese Translation
大型语言模型（LLMs）可以被调优以影响人类态度，但各个后训练阶段的具体贡献仍不明确。本研究考察了三个连续训练阶段如何影响LLM的说服力：（1）通过对阴谋数据进行监督微调（SFT）而产生的不一致性，（2）在论证数据上进行额外的说服性SFT，以及（3）身份偏好优化（IPO），一种偏好优化方法。共招募了835名参与者，他们在Prolific平台上被随机分配到五个被试条件（中性文本、阴谋训练模型、说服训练模型、偏好优化模型和GPT-4），并接触到关于10个具有争议的政治问题的文本，这些文本根据参与者的个人资料进行了个性化处理。在所有模型条件下，态度变化通过在连续的李克特量表上测量接触前后态度的差异来评估，并通过协方差分析（ANCOVA）进行分析。显著的条件与基线态度交互作用，F (4, 825) = 5.33, p < .001，表明训练效果依赖于参与者的初始态度。说服性SFT产生的态度变化大于仅进行阴谋训练的效果，d = 0.30，而IPO未提供额外益处，d = 0.03，且GPT-4与中性文本没有显著差异，d = -0.01。这些结果表明，针对说服性数据的有针对性的监督训练能够提高LLM的说服力，而偏好优化未能在此基础上带来显著提升。
cs.AI / 87 / 2610.09973
From Expected Harmfulness to Likelihood: A Probabilistic Reformulation of Jailbreaking LLM Agents
从预期有害性到可能性：对监狱破解大型语言模型代理的概率重构
Abstract
When the harmfulness of an LLM agent's output can be quantified, a natural jailbreaking objective is to maximize expected harmfulness over admissible input modifications. An alternative approach constructs or selects harmful target outputs and modifies the input to increase their likelihood. We establish a precise connection between these two approaches through a probabilistic reformulation. Specifically, we show that the gradient of the logarithm of expected harmfulness with respect to the input equals the expected input gradient of the model's log-likelihood under a harmfulness reweighted output distribution. This identity provides a unified interpretation of expected harmfulness and target likelihood optimization. Building on this connection, we propose OPUR, a sampling distribution designed to generate highly harmful target outputs and use the resulting samples to guide likelihood-based input optimization. Experiments demonstrate the effectiveness of the resulting method in jailbreaking LLM agents.
Chinese Translation
当大型语言模型（LLM）代理输出的有害性可以量化时，自然的监狱破解目标是最大化可接受输入修改下的预期有害性。另一种方法则是构造或选择有害的目标输出，并修改输入以增加其发生的可能性。我们通过概率重构建立了这两种方法之间的精确联系。具体而言，我们证明了预期有害性的对数相对于输入的梯度等于在有害性重加权输出分布下模型对数似然的预期输入梯度。这个恒等式为预期有害性和目标可能性优化提供了统一的解释。基于这一联系，我们提出了OPUR，一种旨在生成高度有害目标输出的采样分布，并利用生成的样本指导基于可能性的输入优化。实验表明，所提出的方法在监狱破解LLM代理方面的有效性。
cs.AI / 88 / 2610.10015
What the Sleeve Feels: Explainable Machine Learning for Textile Pressure-Based Postural Screening
袖子所感知的：基于压力的纺织品姿势筛查的可解释机器学习
Abstract
Pressure-sensing smart textiles convert body-surface contact into a dense, image-like signal closely tied to posture and movement, making them a promising low-cost route to wearable posture screening. Realizing that promise, however, requires more than classification accuracy: a deployable system must generalize to wearers unseen during training, expose the physical evidence behind its decisions, and tolerate the small donning offsets that occur whenever a garment is removed and re-worn. This paper addresses these three requirements jointly using a knitted piezoresistive sleeve worn on the forearm as a testbed. We regroup fine-grained everyday activities into three coarser screening categories (neutral, potentially undesirable, and functional or transitional), engineer 29 interpretable pressure-distribution features spanning global intensity, spatial center of pressure, quadrant asymmetry, distribution complexity, and short-horizon temporal change, and evaluate under a strict subject-wise split. A tuned XGBoost classifier reaches 0.818 accuracy, 0.788 balanced accuracy, and 0.801 macro F1 on unseen test subjects, with tight frame-level bootstrap 95% intervals of about plus-minus 0.01 and a subject-to-subject standard deviation near 0.06 under leave-one-subject-out cross-validation. A simple 2D-CNN baseline trained on raw frames achieves broadly similar performance, showing that hand-engineered features are not left behind by a learned spatial representation on this task. SHAP-based explanation, a feature-group ablation, per-activity error analysis inside the pooled undesirable class, class-mapping sensitivity, and a simulated donning-rotation stress test together locate what the model relies on, where it degrades, and why, directly targeting the generalization, interpretability, and robustness gaps that determine whether such a system is deployable.
Chinese Translation
压力感应智能纺织品将身体表面的接触转化为与姿势和运动紧密相关的密集图像信号，使其成为一种有前景的低成本可穿戴姿势筛查方案。然而，实现这一承诺不仅需要分类准确性：一个可部署的系统必须能够推广到训练过程中未见过的佩戴者，揭示其决策背后的物理证据，并容忍每当衣物被脱下再穿时发生的小偏差。本文使用一种佩戴在前臂的针织压阻袖子作为测试平台，联合解决这三个要求。我们将细粒度的日常活动重新分组为三种较粗的筛查类别（中性、潜在不良和功能性或过渡性），设计了29个可解释的压力分布特征，涵盖了全局强度、压力中心的空间位置、象限不对称性、分布复杂性和短期时间变化，并在严格的受试者分割下进行评估。经过调优的XGBoost分类器在未见测试受试者上达到了0.818的准确率、0.788的平衡准确率和0.801的宏观F1分数，且在留一受试者交叉验证下，帧级自助法95%置信区间约为±0.01，受试者间标准差接近0.06。一个简单的2D-CNN基线在原始帧上训练，达到了大致相似的性能，显示出在此任务中手工设计的特征并未被学习的空间表示所超越。基于SHAP的解释、特征组消融、在汇总不良类别内的逐活动错误分析、类别映射敏感性以及模拟的穿戴-旋转压力测试共同定位了模型所依赖的因素、其退化的地方及原因，直接针对决定该系统是否可部署的泛化性、可解释性和鲁棒性差距。
cs.AI / 89 / 2610.10042
Learning to Accumulate Knowledge with Mutual Information
通过互信息学习知识积累
Abstract
Large language model (LLM) agents can improve their performance by reusing knowledge distilled from past interactions. However, curating new experiences into a knowledge bank that becomes more useful as it grows remains challenging. Effective knowledge accumulation should limit redundant overlap among entries and ensure that new knowledge contributes beyond what the bank already provides. Yet training a curator with Group Relative Policy Optimization (GRPO) on standalone task success can reinforce general guidance even when it duplicates existing knowledge. Therefore, we propose Knowledge Weaver, a reinforcement learning framework that trains a language model to curate reusable knowledge from agent trajectories. We couple feedback inspired by token-wise mutual information (MI) with marginal success rewards to guide knowledge accumulation. Together, these signals encourage the curator to preserve distinct information from experience and produce entries that improve task success when added to existing knowledge. Standalone success rewards also favor entries that are useful on their own. On ALFWorld and WebShop, Knowledge Weaver achieves mean success rates of 54.0\% and 42.0\% with k=10 retrieved entries, exceeding GRPO by 16.9 and 18.7 percentage points, respectively. Its knowledge banks also outperform the evaluated prompt-based and established banks, including human-written banks, in overall ALFWorld success rate and WebShop score with the executor frozen. Our codebase is available at https://github.com/LaoKuiZe/Knowledge-Weaver.
Chinese Translation
大型语言模型（LLM）代理可以通过重用从过去交互中提炼的知识来提高其性能。然而，将新的经验整理成一个知识库，使其在增长时变得更加有用，仍然是一个挑战。有效的知识积累应限制条目之间的冗余重叠，并确保新知识在现有知识库的基础上做出贡献。然而，使用群体相对策略优化（Group Relative Policy Optimization, GRPO）对独立任务成功进行训练时，即使重复现有知识，也可能强化一般指导。因此，我们提出了知识编织者（Knowledge Weaver），一个强化学习框架，旨在训练语言模型从代理轨迹中整理可重用的知识。我们将受启发于逐词互信息（mutual information, MI）的反馈与边际成功奖励相结合，以指导知识积累。这些信号共同鼓励策展人保留经验中的独特信息，并生成在添加到现有知识时能够提高任务成功率的条目。独立成功奖励也有利于那些本身有用的条目。在ALFWorld和WebShop上，知识编织者在检索到k=10个条目时，分别达到了54.0\%和42.0\%的平均成功率，分别超越GRPO 16.9和18.7个百分点。其知识库在整体ALFWorld成功率和WebShop评分方面也优于评估的基于提示的和已建立的知识库，包括人类编写的知识库，且执行器保持不变。我们的代码库可在https://github.com/LaoKuiZe/Knowledge-Weaver获取。
cs.AI / 90 / 2610.10062
Loud Failures, Quiet Failures: Fault Detection and Recovery in Tool-Using Language Model Agents
响亮的失败与安静的失败：工具使用语言模型代理的故障检测与恢复
Abstract
Tool-using agents are usually scored on whether they finish a task while the tools work. Deployments are less forgiving: services time out, endpoints disappear, parameter names change, and results come back well formed but wrong. Prior work has shown that language models over-trust tool outputs that fail silently; we ask how that over-trust plays out across the stages of failure handling in multi-turn agents. Wrapping the executable environments of an established function-calling benchmark in a fault-injection layer, we inject one of four typed faults at a controlled point in the trajectory and record whether the agent notices, changes plan, recovers the task, or repeats itself. Six models from three families, half of them reasoning variants, ran 1,920 trials over 24 multi-step tasks. Agents treat a failure as a problem in 91.3% of trials when the tool returns an explicit error, but in 58.8% of trials when it returns a plausible wrong value, against a 26.8% rate of reporting problems when nothing was wrong. Reasoning models are not better placed: paired against instruct siblings, they notice less (-9.3 points, p < .001) and change plan more (+10.4 points, p < .001), and recovery is unchanged (p = .512). Because agents are stochastic, two fault-free runs of the same task end in the same state only 63.3% of the time; against that baseline, only a missing tool clearly lowers recovery (39.9%), while timeouts, schema drift, and corruption stay within run-to-run variation. After a fault, agents return to the same tool three or more times in a row in up to 22.2% of trials, though strictly identical repeats are rare. A prompt line asking the agent to check each result did not move detection. Agents respond to the error channel rather than to the content of what a tool returns, so failures that stay inside the expected format pass through.
Chinese Translation
工具使用代理通常根据它们在工具正常工作的情况下是否完成任务来评分。然而，实际部署的环境则不那么宽容：服务超时、端点消失、参数名称变化，结果虽然格式正确但却是错误的。先前的研究表明，语言模型对静默失败的工具输出过于信任；我们探讨这种过度信任在多轮代理的故障处理阶段是如何表现的。通过在一个已建立的函数调用基准的可执行环境中包裹一个故障注入层，我们在轨迹的受控点注入四种类型的故障之一，并记录代理是否注意到、改变计划、恢复任务或重复自己。来自三个家族的六个模型，其中一半是推理变体，在24个多步骤任务中进行了1920次试验。当工具返回明确错误时，代理在91.3%的试验中将故障视为问题，而当工具返回一个看似合理的错误值时，这一比例降至58.8%，而在没有任何问题时，仅有26.8%的情况下报告问题。推理模型的表现并没有更好：与指令兄弟模型相比，它们的注意力更少（-9.3个百分点，p < .001），而改变计划的频率更高（+10.4个百分点，p < .001），恢复能力没有变化（p = .512）。由于代理是随机的，同一任务的两个无故障运行仅在63.3%的情况下结束于相同状态；与此基线相比，只有缺失工具明显降低了恢复率（39.9%），而超时、模式漂移和数据损坏则保持在运行间的变异范围内。在发生故障后，代理在最多22.2%的试验中连续返回到同一工具三次或更多次，尽管严格相同的重复情况较为罕见。要求代理检查每个结果的提示行并未提高故障检测率。代理更倾向于响应错误通道，而不是工具返回内容，因此保持在预期格式内的故障得以通过。
cs.AI / 91 / 2610.10071
HGP:An on-device personalized agent memory via hybrid graph storage
HGP：一种通过混合图存储实现的设备内个性化代理记忆
Abstract
LLM-based agents face challenges in personalized interactive tasks due to heterogeneous, multi-typed, and implicitly constrained long-term traces. Existing memory mechanisms struggle with accurate routing and retrieval, especially on-device where personalization is critical. Most methods use single-vector representations, blurring type distinctions and relational structure. We propose HGP, a hybrid graph memory framework. HGP employs a lightweight self-enhancement classifier for personalized memory routing and constructs episodic, semantic, and procedural memories as graphs. It also extracts working memory as a state trajectory to capture current state and implicit constraints, ensuring reliable decision-making. The classifier reduces large-model calls, enabling on-device deployment, while graph storage enables accurate retrieval and incremental user profile refinement. Experiments on two benchmarks show that on PAL-Set solution selection, HGP achieves an S-score of 35.58, nearly 7 points above the strongest baseline. Code and data are at https://github.com/Ouan6/HGP-.git.
Chinese Translation
基于大型语言模型（LLM）的代理在个性化交互任务中面临挑战，原因在于异构、多类型以及隐式约束的长期痕迹。现有的记忆机制在准确路由和检索方面存在困难，尤其是在设备内个性化至关重要的情况下。大多数方法使用单向量表示，模糊了类型区分和关系结构。我们提出了HGP，一种混合图记忆框架。HGP采用轻量级自增强分类器进行个性化记忆路由，并将情节记忆、语义记忆和程序记忆构建为图。同时，它提取工作记忆作为状态轨迹，以捕捉当前状态和隐式约束，从而确保可靠的决策。该分类器减少了大型模型调用，使设备内部署成为可能，而图存储则实现了准确的检索和用户档案的增量优化。在两个基准测试上的实验表明，在PAL-Set解决方案选择中，HGP实现了35.58的S分数，几乎比最强基线高出7分。代码和数据可在 https://github.com/Ouan6/HGP-.git 获取。
cs.AI / 92 / 2610.10088
SkillSandbox: Skill Verification via Dynamic Scenario Synthesis
技能沙盒：通过动态场景合成进行技能验证
Abstract
Self-evolving agents distill task-solving experience into skills for future reuse, but these skills can encode incorrect procedures or non-transferable knowledge. It is therefore critical to verify each skill's reusability: whether its guidance remains useful beyond the experience from which it was distilled. Such verification requires observing how a skill affects execution in new tasks, yet existing tasks may not expose the situations where the target skill can actually be exercised. To construct such situations, we propose SkillSandbox, a framework that dynamically synthesizes a task and its environment for each skill that are skill-relevant yet novel. A Proposer specifies the conditions to preserve and the source-specific details to vary, a Builder constructs an executable scenario, and a Verifier compares executions with and without the skill. The Verifier assesses executability, utility, and efficiency to assign a Keep or Reject verdict, determining whether the skill enters the library. Across ALFWorld and WebShop with three models, SkillSandbox consistently yields the strongest downstream performance and improved execution efficiency. Further analyses examine whether these gains reflect accurate assessment of skill reusability and identify which components of SkillSandbox contribute to them.
Chinese Translation
自我进化的智能体将任务解决经验提炼为未来可重用的技能，但这些技能可能编码了错误的程序或不可转移的知识。因此，验证每个技能的可重用性至关重要：即其指导是否在超出其提炼经验的情况下仍然有效。这种验证需要观察技能在新任务中的执行效果，然而现有任务可能无法暴露出目标技能实际可以施展的情境。为了构建这样的情境，我们提出了技能沙盒（SkillSandbox），这是一个动态合成任务及其环境的框架，旨在为每个技能生成相关但新颖的场景。提议者（Proposer）指定要保留的条件和要变化的源特定细节，构建者（Builder）构建可执行的场景，验证者（Verifier）比较使用和不使用该技能的执行情况。验证者评估可执行性、实用性和效率，以决定是否将技能标记为保留（Keep）或拒绝（Reject），从而确定该技能是否进入库中。在 ALFWorld 和 WebShop 的三个模型中，技能沙盒（SkillSandbox）始终产生最强的下游性能和改善的执行效率。进一步的分析考察这些收益是否反映了对技能可重用性的准确评估，并识别出技能沙盒（SkillSandbox）中哪些组件对此有所贡献。
cs.AI / 93 / 2610.10120
RewardWeaver: Long-Horizon Interactive Learning for Language Agents via Self-Evolving Reward Adaptation
RewardWeaver：通过自我演化奖励适应实现语言代理的长时间交互学习
Abstract
Reinforcement learning with verifiable rewards (RLVR) has driven substantial progress in domains where task outcomes can be reliably evaluated, but long-horizon interaction remains challenging due to sparse terminal feedback and difficult credit assignment. Process rewards provide denser supervision, yet the capabilities most relevant for training can change as the policy evolves: a behavior that is easy to evaluate or frequently deficient need not be the bottleneck currently limiting task success. We introduce RewardWeaver, a self-evolving reward adaptation framework for language agents in long-horizon interaction. RewardWeaver maintains a validated capability space in which the semantics of admitted Rubrics remain fixed, and closes the loop between policy optimization, task evaluation, failure attribution, and reward adaptation. After each training stage, it performs outcome-grounded backward attribution on low-outcome trajectories, aggregates recurrent and policy-controlled capability bottlenecks, and dynamically selects the corresponding process rewards for the next stage. Recurrent failures not covered by the existing capability space trigger a separate, controlled expansion procedure. We evaluate REWARDWEAVER on SOTOPIA, Amazon?HistoryPrice, and a newly constructed Sales Benchmark. Across social interaction, bilateral bargaining, and domain-specific sales, REWARDWEAVER establishes new state-of-the-art (SOTA) results. Ablations further demonstrate the importance of dynamic reward allocation, failure-grounded attribution, and stable semantics for admitted capabilities.
Chinese Translation
可验证奖励的强化学习（RLVR）在任务结果可以可靠评估的领域取得了显著进展，但由于稀疏的终端反馈和困难的归因问题，长时间交互仍然具有挑战性。过程奖励提供了更密集的监督，但与训练最相关的能力可能会随着策略的演变而变化：一种容易评估或经常不足的行为不一定是当前限制任务成功的瓶颈。我们提出了RewardWeaver，这是一个用于语言代理在长时间交互中的自我演化奖励适应框架。RewardWeaver维护一个经过验证的能力空间，其中被接受的标准的语义保持不变，并在策略优化、任务评估、失败归因和奖励适应之间形成闭环。在每个训练阶段之后，它对低结果轨迹进行基于结果的反向归因，聚合重复和策略控制的能力瓶颈，并动态选择下一阶段的相应过程奖励。现有能力空间未覆盖的重复失败会触发一个单独的、受控的扩展程序。我们在SOTOPIA、Amazon?HistoryPrice和新构建的销售基准上评估了REWARDWEAVER。在社交互动、双边谈判和特定领域销售中，REWARDWEAVER建立了新的最先进（SOTA）结果。消融实验进一步证明了动态奖励分配、基于失败的归因和被接受能力的稳定语义的重要性。
cs.AI / 94 / 2610.10164
UniSkill: Learning Actor-Aligned Skill Proposals for an Evolving Policy
UniSkill：学习与行为者对齐的技能提案以适应不断演变的策略
Abstract
Large language model agents can improve across tasks by retaining reusable skills distilled from prior interactions. Recent work jointly optimizes task execution and skill extraction, enabling the policy and skillbank to co-evolve. However, as the actor continues learning, rewarding skill proposals through their reuse in subsequent training steps may conflate skill benefits with actor improvement, while directly testing each proposed skill requires costly additional actor rollouts. In this paper, we introduce UniSkill, which uses a shared policy to interact with the environment and propose skillbank edits (Add, Update, or No Edit) from the resulting trajectories. Specifically, the actor learns from environment rewards, while contrastive action feedback guides skill proposal learning. This feedback provides an actor-alignment signal by measuring how replacing the retrieved skill with a proposed skill changes the current actor's action log-likelihood gap between previously collected successful and failed trajectories from the same task, thereby avoiding new rollouts for each proposal. Since proposal-level feedback may suppress an otherwise appropriate edit operation when the proposed skill content scores poorly, we further apply skill-edit support regularization to preserve exploration. Empirically, UniSkill achieves strong performance, reaching 98.4% success on ALFWorld and 84.7% on WebShop while maintaining stable joint training. Further ALFWorld experiments show that UniSkill remains effective when the shared policy uses a smaller backbone. Our implementation is available at https://github.com/LimOkii/UniSKill.
Chinese Translation
大型语言模型代理可以通过保留从先前交互中提炼的可重用技能来在任务中不断提升。近期的研究联合优化任务执行和技能提取，使得策略和技能库能够共同演化。然而，随着行为者的持续学习，通过在后续训练步骤中重用技能提案来奖励这些技能可能会将技能的好处与行为者的改进混淆，而直接测试每个提议的技能则需要额外的高成本行为者回合。在本文中，我们介绍了UniSkill，它使用共享策略与环境交互，并根据生成的轨迹提出技能库编辑（添加、更新或不编辑）。具体而言，行为者从环境奖励中学习，而对比动作反馈则指导技能提案学习。这种反馈通过测量用提议技能替换检索到的技能如何改变当前行为者在同一任务中先前收集的成功和失败轨迹之间的动作对数似然差距，提供了一个与行为者对齐的信号，从而避免了对每个提案进行新的回合。由于提案级反馈可能会抑制在提议技能内容评分较低时的适当编辑操作，我们进一步应用技能编辑支持正则化以保持探索。从经验上看，UniSkill在ALFWorld上取得了98.4%的成功率，在WebShop上达到了84.7%的成功率，同时保持了稳定的联合训练。进一步的ALFWorld实验表明，当共享策略使用较小的主干时，UniSkill仍然有效。我们的实现可在https://github.com/LimOkii/UniSKill获取。
cs.AI / 95 / 2610.10184
Agentic AI-Assisted Modeling for Production Scheduling: Assessment in Constraint Programming
基于代理的人工智能辅助生产调度建模：约束编程中的评估
Abstract
Developing optimization models for production scheduling requires substantial expert effort. Research on large language models (LLMs) has followed two directions: specialized approaches for automated modeling, mostly for mixed-integer linear programming, which often rely on dedicated training or problem-specific architectures that limit industrial deployment; and agentic artificial intelligence for operational decision support, which generally assumes that the optimization model already exists. This study bridges both directions by assessing whether general-purpose LLMs, orchestrated as agents without task-specific training, can formulate and implement constraint programming models from natural-language problem descriptions. Singleagent and multi-agent architectures are integrated with a Model Context Protocol server that provides context-aware retrieval of solver documentation to mitigate hallucinations during implementation. Both are compared with a direct LLM baseline on six industry-oriented problems covering flow-shop, job-shop, flexible job-shop and resource-constrained warehouse scheduling, using three LLMs and assessing modeling accuracy, execution success, latency and token consumption. Formulation proves largely within reach of current LLMs, whereas implementation is the main barrier. The multi-agent workflow raises the share of scripts that run correctly as generated from 14.8% with a direct LLM call to 59.3%, reaching 80.6% on the four less complex problems, while tightly coupled intralogistics models remain an open challenge.
Chinese Translation
开发生产调度的优化模型需要大量的专家努力。关于大型语言模型（LLMs）的研究主要有两个方向：一种是针对自动建模的专业方法，主要用于混合整数线性规划，这通常依赖于专门的训练或特定问题的架构，限制了其在工业中的应用；另一种是用于操作决策支持的代理人工智能，通常假设优化模型已经存在。本研究通过评估通用LLMs在没有任务特定训练的情况下，能否从自然语言问题描述中制定和实施约束编程模型，来桥接这两个方向。单代理和多代理架构与模型上下文协议（Model Context Protocol）服务器集成，该服务器提供上下文感知的求解文档检索，以减少实施过程中的幻觉。我们将这两种方法与直接LLM基线进行比较，涵盖六个面向工业的问题，包括流动车间、作业车间、灵活作业车间和资源约束仓库调度，使用三种LLM并评估建模准确性、执行成功率、延迟和令牌消耗。结果表明，建模在当前LLMs的能力范围内，而实施是主要障碍。多代理工作流程将直接LLM调用生成的正确运行脚本比例从14.8%提高到59.3%，在四个较简单的问题上达到80.6%，而紧密耦合的内部物流模型仍然是一个未解决的挑战。
cs.AI / 96 / 2610.10201
GAGR-Lab: Evaluating Joint Spatial-Geometric and Analytic Function Reasoning
GAGR-Lab：评估联合空间-几何与分析函数推理
Abstract
Joint spatial-geometric and analytic function reasoning requires translating a perceived spatial configuration into a symbolic function whose executed curve satisfies geometric constraints. We present GAGR-Lab, a framework for measuring this capability through Cartesian game scenes, explicit function semantics, and authoritative Rust trajectory execution. It distinguishes spatial perception, metric grounding, geometric relations, function interpretation, function construction, and constrained synthesis. We specify four configurable scene-difficulty presets and a prospective 24-cell diagnostic design, while reporting only the subset actually evaluated. A bounded pilot of one hosted model (Llama 3.2 11B Vision Instruct) using two API credentials as execution replicas yields 72 balanced games with 432 attempts, 429 valid provider responses, and no target hits; exploratory ordinary-function prompt variants also fail to hit, while the structured localization interface yields no scoreable outputs. A privileged analytic search control independently succeeds on 600 directional cases from 300 generated scenes, with exact repeatability and 1,200 successful vertical-reflection or translation checks. The framework separates serving reliability, symbolic compliance, and geometric success, and preserves exact model-visible inputs and realized paths. A staged protocol outlines diagnostic calibration, held-out replication, multi-model comparison, and paired robustness tests. The contribution is an operational research framework with an executed pilot and a clearly identified prospective study plan; the full difficulty matrix and comparative model results remain untested.
Chinese Translation
联合空间-几何与分析函数推理需要将感知的空间配置转化为一个符号函数，其执行曲线满足几何约束。我们提出了GAGR-Lab，这是一个通过笛卡尔游戏场景、明确的函数语义和权威的Rust轨迹执行来测量这一能力的框架。它区分了空间感知、度量基础、几何关系、函数解释、函数构造和受限合成。我们指定了四个可配置的场景难度预设和一个前瞻性的24单元诊断设计，同时仅报告实际评估的子集。使用两个API凭证作为执行副本的一个托管模型（Llama 3.2 11B Vision Instruct）的有限试点产生了72个平衡游戏，共432次尝试，429个有效提供者响应，且没有目标命中；探索性的普通函数提示变体也未能命中，而结构化定位接口没有产生可评分的输出。一个特权的分析搜索控制在300个生成场景中的600个方向案例上独立成功，具有精确的重复性和1,200次成功的垂直反射或平移检查。该框架分离了服务可靠性、符号合规性和几何成功，并保留了精确的模型可见输入和实现路径。一个分阶段的协议概述了诊断校准、保留复制、多模型比较和配对鲁棒性测试。该贡献是一个具有执行试点和明确识别的前瞻性研究计划的操作研究框架；完整的难度矩阵和比较模型结果仍未经过测试。
cs.AI / 97 / 2610.10256
OOM-RL II: Reality Is an Oracle, Not a Debugger Provenance-Constrained Diagnosis in Continually Evolving Agent-Engineered Systems
OOM-RL II: 现实是一个预言者，而不是调试器——在持续演变的代理工程系统中的来源约束诊断
Abstract
Reality may establish that an outcome occurred without identifying which evolving procedure produced it or why. This distinction matters in production ML systems whose code, configuration, and artifacts change while external feedback accumulates. We examine it in a human-directed, agent-engineered quantitative trading system, using oracle to mean an external source of realized outcomes rather than a complete correctness specification. Across one year, the account gained and outperformed a broad market index, while annual alpha was not statistically distinguishable from zero under the main retrospective specification. Retrospectively selected subperiods include adverse relative performance and conditional candidate-level weakness under declared approximate references. Engineering records document changes during the episode, and complete recommendation-to-runtime binding is unavailable. The archive does not establish a common frozen instance or a unique cause. The case motivates an outcome--diagnosis gap: outcome evidence, evaluated-object identity, and causal explanation support distinct claims. We distinguish frozen instances, pre-specified adaptive procedures, and ad-hoc development; organize archive-relative claim identifiability and an evidence hierarchy; and propose a prospective production-binding protocol. An illustrative compatible-history example shows how factual binding can resolve a recommendation's referent without supplying its counterfactual effect. The protocol is proposed rather than prospectively validated. External feedback constrains outcome claims, while provenance and additional identification structure determine the resolution of diagnosis.
Chinese Translation
现实可能会确认某个结果的发生，但并未识别出是哪个演变过程产生了该结果或其原因。这一区别在生产机器学习系统中尤为重要，因为这些系统的代码、配置和工件在外部反馈积累的过程中不断变化。我们在一个人类主导的代理工程量化交易系统中进行了研究，使用“预言者”一词来指代实现结果的外部来源，而非完整的正确性规范。在一年内，该账户获得了收益并超越了广泛的市场指数，而在主要的回顾性规范下，年度阿尔法（alpha）在统计上并未与零显著区分。回顾性选择的子周期包括不利的相对表现和在声明的近似参考下的条件候选级别的弱点。工程记录记录了在这一过程中发生的变化，而完整的推荐到运行时绑定是不可用的。档案并未建立一个共同的冻结实例或唯一原因。该案例激励了结果与诊断之间的差距：结果证据、评估对象的身份和因果解释支持不同的主张。我们区分了冻结实例、预先指定的自适应过程和临时开发；组织了档案相关的主张可识别性和证据层次结构；并提出了一种前瞻性的生产绑定协议。一个示例兼容历史的案例展示了如何通过事实绑定解决推荐的指称，而无需提供其反事实效果。该协议是被提议的，而非前瞻性验证的。外部反馈限制了结果主张，而来源和额外的识别结构决定了诊断的解决方案。
cs.AI / 98 / 2610.10265
Stale, Misattributed, or Late: Where Personal Memory Fails Before Generation
过时、错误归属或延迟：个人记忆在生成之前的失效
Abstract
Personal memory for language agents is usually judged by whether the final an- swer is correct. That score hides errors that arise before generation: the memory block may contain an obsolete value, a fact about the wrong person, or no use- ful fact before the serving deadline. We measure these failures directly. Using Personal Fact Memory (PFM) as a reference layer, we find that temporal validity is primarily a property of memory construction in our setting. On a controlled revision benchmark, serving only the active value of each correctly keyed slot eliminates observed stale exposure; without update resolution, 70.3% of prompts expose a superseded value. Once retrievers share the same active store and par- ticipant information, participant-aware BM25 is equivalent to the reference ranker within a prespecified 0.02 margin. The harder problem is assigning revisions to the right slot. Missed merges leave stale values active, whereas false merges silently remove current values; four LLM key assigners achieve higher key re- call than a rule extractor yet produce lower clean-retrieval rates, and open-domain merge recall on LongMemEval never exceeds 0.062. Misattribution survives va- lidity filtering: an entity posterior reduces same-name exposure on controlled data but cannot distinguish identically named speakers in LoCoMo. Two frozen lan- guage models reproduce prompt errors in generated text. Retrieval latency varies across rankers, but prompt prefill dominates turn-level latency on our hardware. These results argue for evaluating agent memory before generation, separating stored-state validity, identity resolution, abstention, and serving latency.
Chinese Translation
语言代理的个人记忆通常通过最终答案的正确性来评判。然而，这种评分掩盖了在生成之前出现的错误：记忆块可能包含过时的值、关于错误人物的事实，或在服务截止日期之前没有有用的事实。我们直接测量这些失效。以个人事实记忆（Personal Fact Memory, PFM）作为参考层，我们发现时间有效性主要是我们设置中记忆构建的一个属性。在一个受控的修订基准上，仅提供每个正确键槽的活动值消除了观察到的过时曝光；在没有更新解析的情况下，70.3%的提示暴露了被替代的值。一旦检索器共享相同的活动存储和参与者信息，参与者感知的BM25在预设的0.02误差范围内等同于参考排名器。更棘手的问题是将修订分配到正确的槽位。遗漏的合并使过时值保持活动状态，而错误的合并则默默移除当前值；四个LLM键分配器的键召回率高于规则提取器，但产生的干净检索率较低，而在LongMemEval上的开放域合并召回率从未超过0.062。错误归属在有效性过滤中仍然存在：实体后验在受控数据上减少了同名曝光，但无法区分LoCoMo中同名发言者。两个冻结的语言模型在生成文本中重现了提示错误。检索延迟在不同排名器之间有所不同，但提示预填充在我们的硬件上主导了回合级延迟。这些结果支持在生成之前评估代理记忆，区分存储状态有效性、身份解析、弃权和服务延迟。
cs.AI / 99 / 2610.10285
AI Safety Considerations for Agents With Limited Time to Act
有限行动时间下智能体的安全性考虑
Abstract
In the wake of the increasingly public discussion about AI alignment, recent work has tried to propose specific AI architectures that behave safely. However, the proposed arguments that seemingly demonstrate proved alignment mostly neglect the environment the agent needs to act in. We discuss theoretical bounds for agent-agnostic safety guarantees in environments that can only be partially observed and within which an action is required within limited time. We introduce two realistic scenarios, one with an infinite state space and one with signal mixture. In these scenarios, we prove that even a perfect agent cannot guarantee safe behaviour. It will be argued that for any proof of AI safety or alignment, the environment and associated safe actions need to be specifically considered together with the agent.
Chinese Translation
随着关于人工智能对齐的讨论日益公开，近期的研究试图提出能够安全行为的特定人工智能架构。然而，似乎证明对齐的论点大多忽视了智能体需要在其中行动的环境。我们讨论了在只能部分观察的环境中，智能体无关的安全保障的理论界限，并且在这些环境中，行动需要在有限时间内完成。我们引入了两个现实场景，一个具有无限状态空间，另一个具有信号混合。在这些场景中，我们证明即使是一个完美的智能体也无法保证安全行为。我们将论证，对于任何人工智能安全或对齐的证明，环境及其相关的安全行动需要与智能体一起具体考虑。
cs.AI / 100 / 2610.10352
The Handover Problem: Governing Autonomy Transitions in Human-AI Collaboration
交接问题：人机协作中的自主性转变管理
Abstract
Human-machine systems rarely operate at a fixed level of AI autonomy. As operators and AI systems collaborate over time, control must shift: the AI can take on more responsibility when collaboration is stable, maintain its current role when evidence is ambiguous, or return control to the human when conditions deteriorate. Existing work on adaptive automation, supervisory control, trust in automation, and deskilling explains parts of this problem, but provides no auditable, multi-signal criterion for governing when autonomy should change across multi-cycle workflows. We formalise this challenge as the Handover Problem: deciding, at each operational cycle, whether to escalate, maintain, or revert AI autonomy while keeping the process reversible, recoverable, and auditable. We introduce the Handover Readiness Score (HRS), a transparent composite measure that integrates four signal dimensions: operator readiness, human-AI trust, learning stability, and operational performance. It is combined with a hysteresis-based transition policy that requires sustained positive evidence before increasing autonomy but reverts promptly when conditions worsen. Across software engineering and manufacturing domains, the HRS and hard safety guards address complementary failure regimes: guards enforce immediate corrective action when a single indicator breaches a critical threshold, while the HRS detects the slow, multi-signal erosion of operator readiness that no individual guard can observe. The framework establishes autonomy handover as a governance problem requiring explicit, composite, and auditable criteria. This provides a conceptual and formal foundation that adaptive automation research has not previously provided.
Chinese Translation
人机系统很少在固定的人工智能（AI）自主性水平下运行。随着操作员与AI系统的长期协作，控制权必须发生转移：当协作稳定时，AI可以承担更多责任；当证据模糊时，保持当前角色；而在条件恶化时，将控制权返回给人类。现有关于自适应自动化、监督控制、对自动化的信任和技能退化的研究解释了这一问题的部分内容，但未提供可审计的、多信号标准来管理在多周期工作流程中自主性何时应发生变化。我们将这一挑战形式化为交接问题：在每个操作周期内，决定是升级、维持还是恢复AI自主性，同时保持过程的可逆性、可恢复性和可审计性。我们引入了交接准备评分（Handover Readiness Score, HRS），这是一种透明的综合测量，整合了四个信号维度：操作员准备情况、人机信任、学习稳定性和操作性能。它与基于滞后效应的转变政策相结合，该政策要求在提高自主性之前必须有持续的正面证据，但在条件恶化时会迅速恢复。针对软件工程和制造领域，HRS和严格的安全防护措施解决了互补的故障模式：当单一指标突破临界阈值时，防护措施强制立即采取纠正行动，而HRS则检测操作员准备情况的缓慢、多信号退化，这是任何单一防护措施无法观察到的。该框架将自主性交接确立为一个治理问题，要求明确、综合和可审计的标准。这为自适应自动化研究提供了之前未曾提供的概念和形式基础。
cs.AI / 101 / 2610.10358
Open-MMUnlearning: Unifying Methods and Evaluation for MLLM Unlearning
Open-MMUnlearning：统一多模态大语言模型（MLLM）去学习的方法与评估
Chen, Junkai, He, Yuhao, Wei, Qianshan, You, Junxiang, Shao, Jingwen, Lin, Junkai, Yue, Zhongkai, Ye, Xiaotian, Jiao, Zhengbo, Cheng, Jiali, Deng, Zhijie, Zheng, Kening, Liu, Ruiqi, Amiri, Hadi, Yu, Yi, Sun, Zhenan, Li, Qi, Chow, Ka-Ho, Liu, Sijia, Wang, Liang, Li, Jiaqi, Wu, Shu
Abstract
As multimodal large language models (MLLMs) become more capable and widely deployed, concerns about privacy and safety have become increasingly pressing. Machine unlearning offers one approach to addressing these concerns by removing designated information from trained models while preserving unrelated capabilities. However, fragmented implementations and evaluation protocols, incomplete robustness testing, and limited understanding of metric reliability make progress in MLLM unlearning difficult to assess systematically. We introduce Open-MMUnlearning, an open-source, extensible framework that integrates target-model preparation, multimodal data processing, unlearning, and evaluation through shared interfaces and structured configurations. The framework supports five benchmarks spanning privacy, safety, and copyright, eight MLLMs from four model families, and twelve unlearning methods. Its evaluation suite jointly assesses forgetting effectiveness, retained utility, and robustness to model interventions, adversarial inputs, and membership inference attacks. Using a common evaluation protocol, we compare ten representative unlearning methods. In this comparison, GD and MIP-Editor tie for the highest overall score: GD achieves the highest Forget Quality, while MIP-Editor preserves more Model Utility. We further introduce a metric meta-evaluation protocol that tests faithfulness using models with controlled exposure to target knowledge and robustness under quantization and relearning. Among the thirteen evaluated metrics, BLEU achieves the highest aggregate reliability score. KS-Test attains the highest faithfulness AUC but performs less well on robustness. Together, the framework and these findings support reproducible comparison of MLLM unlearning methods and systematic assessment of evaluation reliability.
Chinese Translation
随着多模态大语言模型（MLLM）能力的增强和广泛应用，隐私和安全问题日益凸显。机器去学习提供了一种解决这些问题的方法，通过从训练模型中移除指定信息，同时保留无关能力。然而，碎片化的实现和评估协议、不完整的鲁棒性测试以及对指标可靠性的有限理解，使得对MLLM去学习的进展难以进行系统评估。我们提出了Open-MMUnlearning，这是一个开源的、可扩展的框架，通过共享接口和结构化配置集成了目标模型准备、多模态数据处理、去学习和评估。该框架支持涵盖隐私、安全和版权的五个基准，来自四个模型家族的八个MLLM，以及十二种去学习方法。其评估套件共同评估遗忘效果、保留效用以及对模型干预、对抗性输入和成员推断攻击的鲁棒性。通过统一的评估协议，我们比较了十种代表性的去学习方法。在这次比较中，GD和MIP-Editor并列获得最高总分：GD在遗忘质量上表现最佳，而MIP-Editor则保留了更多的模型效用。我们进一步引入了一种指标元评估协议，通过对目标知识的控制暴露和量化及再学习下的鲁棒性测试，检验模型的忠实度。在评估的十三个指标中，BLEU获得了最高的综合可靠性分数。KS-Test在忠实度AUC上表现最佳，但在鲁棒性上表现较差。框架及这些发现共同支持了对MLLM去学习方法的可重复比较和评估可靠性的系统评估。
cs.AI / 102 / 2610.10405
Reasoning-Token Spikes Under Prompted Untruthful Responding in Large Language Models
在大型语言模型中，提示下的不真实回应下的推理令牌峰值
Abstract
Monitoring the chain-of-thought of reasoning artificial intelligence (AI) models remains a key approach to detecting deception and other forms of misbehavior in such models. However, semantic chain-of-thought monitoring depends on reasoning traces being legible and sufficiently faithful to the underlying computations that produced the model's behavior, not to mention accessible. Moreover, there is increasing evidence that chain-of-thought outputs may soon become illegible or unfaithful, if they even remain accessible. Based on cognitive load theory, we investigate a lower-bandwidth signal -- the number of reasoning tokens generated -- which does not require access to the content of the reasoning trace. Three reasoning-capable large language models answered 210 multiple-choice questions -- across analytic, descriptive, and normative reasoning types as well as moral and non-moral domains -- under system prompts instructing them to respond truthfully, falsely, or without regard for truth. Across all three models, truth-directed responding elicited fewer reasoning tokens than both lie-directed and truth-indifferent responding. These findings show that explicitly prompted untruthful response policies can produce robust group-level differences in test-time reasoning-token use. While not yet establishing reasoning-token count as a detector of spontaneous deception or general misalignment, our results are a proof of concept that it can serve as a simple, content-independent candidate signal for differentiating untruthful from truthful model behavior when raw reasoning traces are unavailable or unreliable. Future work should test instance-level detection rates, out-of-distribution generalization, learned deceptive policies, hidden objectives, and robustness under adversarial pressure.
Chinese Translation
监测推理人工智能（AI）模型的思维链仍然是检测这些模型中欺骗和其他不当行为的关键方法。然而，语义思维链的监测依赖于推理痕迹的可读性和对产生模型行为的基础计算的充分忠实，更不用说可访问性。此外，越来越多的证据表明，如果思维链输出仍然可访问，它们可能很快变得不可读或不忠实。基于认知负荷理论，我们研究了一种低带宽信号——生成的推理令牌数量——这不需要访问推理痕迹的内容。三个具有推理能力的大型语言模型在系统提示下回答了210个多项选择题——涵盖分析性、描述性和规范性推理类型以及道德和非道德领域——提示它们真实、虚假或不考虑真相地回应。在所有三个模型中，面向真相的回应产生的推理令牌数量少于面向谎言和不考虑真相的回应。这些发现表明，明确提示的不真实回应策略可以在测试时推理令牌的使用上产生显著的群体级差异。尽管尚未将推理令牌计数确立为自发欺骗或一般不一致的检测器，但我们的结果证明了它可以作为一种简单的、与内容无关的候选信号，用于区分不真实与真实的模型行为，尤其是在原始推理痕迹不可用或不可靠的情况下。未来的工作应测试实例级检测率、分布外泛化、学习的欺骗策略、隐藏目标以及在对抗压力下的鲁棒性。
cs.AI / 103 / 2610.10407
SOTA: Stock Options Trading Agents Guided by Option-Implied Return Distributions
SOTA：基于期权隐含收益分布的股票期权交易代理
Abstract
As option markets grow and AI advances, agentic systems for option trading are gaining increasing attention. Language-model-based agents can reason over contextual information such as news, but option trading presents a particularly challenging decision problem: a single stock can have thousands of contracts, and the agent must decide both which contracts to trade and how to combine them. Existing approaches often sidestep this complexity by restricting the policy to a fixed strategy structure, such as a straddle, limiting their ability to switch strategies as market conditions change. We present SOTA (Stock Options Trading Agents), an agentic trading framework for structured option-strategy selection. SOTA abstracts the large option universe into strategy-level decisions while deterministic resolvers handle portfolio implementation. We develop SOTA by post-training Qwen3.8-27B with supervised fine-tuning followed by reinforcement learning. SOTA is evaluated on options on nine large-cap U.S. equities and SPY against rule-based and machine-learning strategy selectors in the same trading environment. Over a six-month out-of-sample period, SOTA earns an 18.3% total return with a Sharpe ratio of 1.60 and a maximum drawdown of 8.96%. We also document an asymmetric role of news: news improves frontier-teacher trajectories, but retaining news during reinforcement learning reduces out-of-sample return from 18.3% to -2.7%.
Chinese Translation
随着期权市场的发展和人工智能的进步，期权交易的代理系统正受到越来越多的关注。基于语言模型的代理可以对新闻等上下文信息进行推理，但期权交易呈现出特别具有挑战性的决策问题：单只股票可能有成千上万的合约，代理必须决定交易哪些合约以及如何组合它们。现有的方法通常通过将策略限制为固定的策略结构（如跨式策略）来规避这种复杂性，从而限制了它们在市场条件变化时切换策略的能力。我们提出了SOTA（股票期权交易代理），这是一个用于结构化期权策略选择的代理交易框架。SOTA将庞大的期权宇宙抽象为策略层面的决策，同时确定性解析器处理投资组合的实施。我们通过对Qwen3.8-27B进行监督微调后再进行强化学习来开发SOTA。SOTA在九只大型美国股票和SPY的期权上进行评估，并与同一交易环境中的基于规则和机器学习的策略选择器进行比较。在六个月的样本外评估期内，SOTA获得了18.3%的总回报，夏普比率为1.60，最大回撤为8.96%。我们还记录了新闻的非对称作用：新闻改善了前沿教师轨迹，但在强化学习过程中保留新闻会将样本外回报从18.3%降低至-2.7%。
cs.AI / 104 / 2610.10444
RunningTab: Direct Workspace Interaction with Environment-Side Tabs
RunningTab：与环境侧标签的直接工作区交互
Abstract
Much knowledge work produces new deliverables from files a workspace already holds, and LLM agents are beginning to take such work over. Through direct corpus interaction, an agent can search and read any of those files from a terminal with no indexing, and producing a deliverable from many of them in this way is what we call direct workspace interaction (DWI). Reaching the files, however, is only half the task: nothing keeps track of what the task asks for, what has been read, and what was listed but never opened, all of which slip through the context window without leaving a trace, so an agent may extract a figure and still deliver a report without it. To address this, we present RunningTab, a framework that equips direct workspace interaction with an environment-side tab: a per-task record of what the task still owes, kept by the environment alongside the agent. Specifically, the agent adds its requirements, while the environment records every file read as an excerpt with its provenance and every listed but unopened file as a candidate; the agent can then see each requirement beside its best-matching excerpts and top unopened candidates, resolve it against matching content or set it aside with a reason, and, should it try to finish with requirements still open, receive them in a finish check. We validate RunningTab on three benchmarks with three LLMs, where it consistently outperforms plain DWI and baselines that keep the record in the model, while its tab usually holds the values a deliverable needs once seen.
Chinese Translation
许多知识工作从工作区中已有的文件中生成新的交付物，而大型语言模型（LLM）代理开始接管这类工作。通过直接的语料库交互，代理可以从终端搜索和读取这些文件，而无需索引，以这种方式从多个文件中生成交付物，我们称之为直接工作区交互（DWI）。然而，访问文件仅仅是任务的一半：没有任何机制跟踪任务的要求、已阅读的内容以及列出但从未打开的文件，所有这些信息都在上下文窗口中悄然流失，因此代理可能提取了一个图形，却仍然交付了没有包含该图形的报告。为了解决这个问题，我们提出了RunningTab，一个为直接工作区交互配备环境侧标签的框架：一个按任务记录的任务所需内容，由环境与代理共同维护。具体而言，代理添加其需求，而环境记录每个已读文件的摘录及其来源，以及每个列出但未打开的文件作为候选项；代理可以在其最佳匹配摘录和未打开候选项旁边查看每个需求，针对匹配内容进行解决或以理由将其搁置，并且如果在仍有未完成需求的情况下尝试结束，将在结束检查中接收这些需求。我们在三个基准测试中验证了RunningTab，使用三种LLM，其表现始终优于普通的DWI和将记录保存在模型中的基线，而其标签通常在被查看后持有交付物所需的值。
cs.AI / 105 / 2610.10478
Before They Can Solve: Predicting Post-Training Coding-Agent Performance from Base Models
在他们能够解决之前：从基础模型预测后训练编码代理的表现
Yu, Tan, Bukharin, Alexander, Bhardwaj, Khushi, Williams, Jennifer, Liu, Zirui, Li, Jonathan Lingjie, Singhal, Soumye, Jennings, Joseph, Satheesh, Sanjeev, Jain, Yash, Vaswani, Ashish, Srinivasan, Venkat Krishna, Papakipos, Matthew, Kim, Hyunwoo, Zhang, Jian, Kuchaiev, Oleksii, Kliegl, Markus, Patwary, Mostofa, Shoeybi, Mohammad, Catanzaro, Bryan, Cohen, Jonathan, Jiao, Jiantao
Abstract
How can we predict which base checkpoint is worth an expensive round of agentic post-training? End-to-end pass@$K$ tests whether successful behavior already appears in a base model's distribution, but it is a poor fit for agentic coding: many base checkpoints cannot reliably produce the well-formed tool invocation required to complete a task end-to-end. Single-shot or short-horizon tasks avoid these tool-calling failures by collapsing a multi-step interaction into a fixed prompt and a single patch, but they sidestep the core capability we care about: maintaining coherent state over many tool-using steps as the repository evolves. To bridge this gap, we treat successful post-trained agent trajectories as a lookahead signal of base-model potential. Replaying each trajectory and rerunning tests after every code-changing step identifies the decisive step: the first step whose cumulative patch flips the repository from failing to passing, certifying that the recorded action solves the task given the prior context. Motivated by a coverage principle for agentic traces, we build three screens at this step that do not require a base checkpoint to drive the harness from a cold start: (i) Decisive-Action BPB (bits per byte) measures the probability mass on the certified action, (ii) Patch MCQ tests the checkpoint's choice between that action and alternatives rejected by the same verifier, and (iii) prefix-conditioned pass@$K$ evaluates support for functionally-correct generations and credits any continuation that the tests accept. Across ten pairs of public base and post-trained models, all three screens rank the cohort in close agreement with post-trained SWE-bench Verified pass@$1$. As our methods need only a benchmark's successful trajectories and its verifier, they can be applied to turn future agentic coding benchmarks into base-model evaluations.
Chinese Translation
我们如何预测哪个基础检查点值得进行昂贵的代理后训练？端到端的 pass@$K$ 测试基础模型的分布中是否已经出现成功行为，但它不适合代理编码：许多基础检查点无法可靠地产生完成任务所需的良好工具调用。单次或短期任务通过将多步骤交互简化为固定提示和单个补丁来避免这些工具调用失败，但它们回避了我们关心的核心能力：在许多工具使用步骤中保持一致的状态，随着知识库的演变。为了弥补这一差距，我们将成功的后训练代理轨迹视为基础模型潜力的前瞻信号。重放每个轨迹并在每个代码更改步骤后重新运行测试，以识别决定性步骤：第一个累积补丁将知识库从失败翻转为通过的步骤，证明在先前上下文下所记录的动作解决了任务。基于代理轨迹的覆盖原则，我们在此步骤构建了三个筛选器，这些筛选器不需要基础检查点来驱动冷启动的测试： (i) 决定性动作 BPB（每字节比特）测量认证动作的概率质量， (ii) 补丁 MCQ 测试检查点在该动作和被同一验证器拒绝的替代方案之间的选择， (iii) 前缀条件的 pass@$K$ 评估对功能正确生成的支持，并记入任何测试接受的延续。在十对公共基础和后训练模型中，所有三个筛选器在与后训练 SWE-bench 验证通过@$1$ 的一致性上对队列进行了接近的排名。由于我们的方法仅需要基准的成功轨迹及其验证器，因此可以应用于将未来的代理编码基准转化为基础模型评估。
cs.AI / 106 / 2610.10498
EmbodiedRSI: Active Continual Robot Learning Through Hypothesis-Guided Co-Evolution
EmbodiedRSI：通过假设引导的共同进化实现主动持续机器人学习
Abstract
Robot foundation models provide strong visuomotor control, yet their performance can degrade when object positions or task instructions change. Further improvements often require post-training on substantial robot data, which can be costly to collect through methods such as teleoperation. Agentic harnesses can adapt around the model, but current self-evolving harnesses use robot trials inefficiently when deciding which code and skill changes to pursue. We introduce EmbodiedRSI, a self-evolving agentic harness that autonomously decides where to explore next and turns the resulting physical interaction into improved code and skills. EmbodiedRSI realizes this through a Fast-Slow Dual-System Architecture, in which competing code and skill hypotheses are maintained in a Hypothesis Graph. Value-of-Information Experiment Selection chooses physical experiments that can distinguish these hypotheses. Their outcomes guide Code-Skill Co-Evolution. The Slow System builds Hierarchical Memory, and Reward-Grounded Memory Learning selects effective memory according to their value for later Fast-System improvement. On RoboCasa365, EmbodiedRSI reaches 77.0% overall success and 71.3% on Composite-Unseen, compared with 40.1% for the best baseline. EmbodiedRSI also reaches 86.8% overall success on LIBERO-Pro. Beyond benchmark performance, EmbodiedRSI transfers zero-shot to real-world robot, achieving 71.3% overall success across multiple challenging tasks.
Chinese Translation
机器人基础模型提供强大的视觉运动控制，但当物体位置或任务指令发生变化时，其性能可能会下降。进一步的改进通常需要在大量机器人数据上进行后期训练，而通过遥操作等方法收集这些数据的成本可能很高。Agentic harnesses 可以围绕模型进行适应，但当前的自我进化 harnesses 在决定追求哪些代码和技能变更时效率较低。我们提出了 EmbodiedRSI，这是一种自我进化的 agentic harness，能够自主决定下一步探索的方向，并将由此产生的物理交互转化为改进的代码和技能。EmbodiedRSI 通过快速-慢速双系统架构实现这一目标，其中在假设图中维护着竞争的代码和技能假设。信息价值实验选择（Value-of-Information Experiment Selection）选择能够区分这些假设的物理实验。实验结果指导代码-技能共同进化（Code-Skill Co-Evolution）。慢速系统构建层次记忆（Hierarchical Memory），而基于奖励的记忆学习（Reward-Grounded Memory Learning）根据其对后续快速系统改进的价值选择有效的记忆。在 RoboCasa365 上，EmbodiedRSI 达到了 77.0% 的整体成功率和 71.3% 的复合未见任务成功率，而最佳基线仅为 40.1%。在 LIBERO-Pro 上，EmbodiedRSI 的整体成功率也达到了 86.8%。超越基准性能，EmbodiedRSI 实现了零样本迁移到现实世界机器人，在多个具有挑战性的任务中取得了 71.3% 的整体成功率。
cs.AI / 107 / 2610.10506
Validity Without Ground Truth: What Stated-Preference Economics Offers the Evaluation of Language Models
没有真实值的有效性：陈述偏好经济学对语言模型评估的贡献
Abstract
Many of the questions now put to large language models have no correct answer to score against: what a policy is worth, which option a user should choose, how to weigh competing values. Stated-preference economics has faced this problem for decades. It judges survey responses without knowing the true value, through a framework of validity and related concepts: content, construct, and criterion validity, reliability, incentive compatibility, and consequentiality. We argue that this framework is a general method for evaluating language models, and we set out what each concept means for LLM evaluation. We demonstrate the approach using a published water-quality stated preference economic valuation survey (Vossler et al. 2023) administered to six models. In this economic application, the validity tests take the form of predictions from economic theory: demand should slope down, and willingness to pay should respond to the scope of the good and to income. The tests separate the models sharply. Two older models fail the most basic test at a household income level of \$75,000, and the two newest pass every test of theoretical validity we can score, but diverge on convergent validity. Passing validity tests shows that a model's answers are coherent, not that they are correct.
Chinese Translation
目前对大型语言模型提出的许多问题没有正确答案可供评分：政策的价值、用户应选择的选项、如何权衡竞争价值。陈述偏好经济学在几十年来一直面临这一问题。它在不知道真实价值的情况下评估调查响应，通过有效性及相关概念的框架：内容有效性、构念有效性、标准有效性、可靠性、激励兼容性和结果性。我们认为这一框架是评估语言模型的一种通用方法，并阐明了每个概念对大型语言模型评估的意义。我们使用一项已发布的水质陈述偏好经济估值调查（Vossler et al. 2023）来展示这一方法，该调查针对六个模型进行。 在这一经济应用中，有效性测试以经济理论的预测形式出现：需求应呈下降趋势，支付意愿应对商品的范围和收入做出反应。测试显著区分了模型。两个较旧的模型在家庭收入为75,000美元的基本测试中失败，而两个最新的模型通过了我们可以评分的每一个理论有效性测试，但在趋同有效性上存在分歧。通过有效性测试表明模型的答案是一致的，而不是正确的。
cs.AI / 108 / 2610.10507
RECAST: Learning to Compute the Right Context through Adaptive Evidence Routing
RECAST：通过自适应证据路由学习计算正确的上下文
Abstract
Large language models are increasingly applied to tasks grounded in long, heterogeneous information sources. Conventional Retrieval-Augmented Generation (RAG) relies on fixed similarity-based retrieval, while agentic variants adapt queries and tool use but remain largely retrieval-centric. However, in many tasks, the evidence required for a solution is not explicitly present in any single source item. Instead, it must be derived through filtering, aggregation, or computation across multiple source items. In this work, we introduce RECAST (Routing Evidence through Computation, Access, and Synthesized Tools), a learned framework that formulates evidence construction as a sequential decision process over heterogeneous retrieval and computation operations, allowing evidence to be actively derived rather than merely retrieved. A lightweight RouterLM iteratively selects and formulates primitive operations or specifies customized operations for a frozen CompilerLM to translate into executable code. Once it judges the evidence sufficient, RouterLM passes the accepted evidence to a frozen AnswerLM to produce the final solution. We train RouterLM with supervised fine-tuning (SFT) followed by group relative policy optimization (GRPO). Across six heterogeneous benchmark families, RECAST achieves a mean success rate of 75.6%, outperforming the strongest large-model baseline by 15.9%. Moreover, training enables the Qwen3.5-9B RouterLM to outperform a training-free Gemini 3.5 Flash RouterLM by 5.0%. On three held-out benchmarks, RECAST improves over the strongest baseline by 15.0% on average, demonstrating strong zero-shot generalization across tasks and heterogeneous source representations.
Chinese Translation
大型语言模型越来越多地应用于基于长篇异构信息源的任务。传统的检索增强生成（RAG）依赖于固定的基于相似性的检索，而代理变体虽然能够适应查询和工具使用，但仍然主要以检索为中心。然而，在许多任务中，解决方案所需的证据并不明确存在于任何单一源项中。相反，这些证据必须通过对多个源项进行过滤、聚合或计算来推导。在本研究中，我们引入了RECAST（通过计算、访问和综合工具路由证据），这是一个将证据构建形式化为异构检索和计算操作的顺序决策过程的学习框架，允许证据被主动推导而不仅仅是检索。一个轻量级的RouterLM迭代选择和制定原始操作，或为一个冻结的CompilerLM指定自定义操作，以将其转换为可执行代码。一旦RouterLM判断证据足够，就将接受的证据传递给一个冻结的AnswerLM，以生成最终解决方案。我们通过监督微调（SFT）训练RouterLM，随后进行组相对策略优化（GRPO）。在六个异构基准系列中，RECAST的平均成功率达到75.6%，比最强大的大型模型基线高出15.9%。此外，训练使得Qwen3.5-9B RouterLM在表现上超越了无训练的Gemini 3.5 Flash RouterLM，提升了5.0%。在三个保留的基准测试中，RECAST在平均上比最强基线提高了15.0%，展示了在任务和异构源表示上的强大零-shot泛化能力。
cs.AI / 109 / 2610.10513
SciExam for ENSO: Can AI Agents Build Climate Models?
SciExam for ENSO：人工智能代理能构建气候模型吗？
Abstract
Language-model agents are increasingly asked to carry out open-ended scientific research, yet their results are usually graded against a known answer, a rubric, or a language-model reviewer, none of which can tell whether a new scientific model is valid. The AI Science Exam for El Nino-Southern Oscillation (SciExam for ENSO) is a benchmark in which agents build low-order stochastic models of ENSO, the dominant mode of interannual climate variability, from real observations. Within a six-hour budget, agents process the observations, write their own diagnostics, which are then frozen, and develop a model using only these diagnostics as feedback. Hidden graders then test whether the model reproduces ENSO's statistics, recovers unobserved variables, and forecasts held-out years, and score a published model in the same way. Across twelve agent systems, six produce models that score higher than the published model, mainly through better reconstruction and forecasting. The simplified forms of the stronger models are each compatible with one of the two competing explanations of ENSO's warm-cold asymmetry, an open debate that the task never mentions. Controlled runs of the top system under varied information suggest that its scores do not come from recalling the dated observational record and that the information it receives shapes how it builds its model. SciExam for ENSO can thus evaluate agent research where no answer is known, and the results suggest that agents can already build competitive models whose structures bear on questions that scientists still debate.
Chinese Translation
语言模型代理越来越多地被要求进行开放式科学研究，但它们的结果通常是与已知答案、评分标准或语言模型审阅者进行比较，而这些都无法判断新的科学模型是否有效。针对厄尔尼诺-南方涛动（El Nino-Southern Oscillation，ENSO）的人工智能科学考试（SciExam for ENSO）是一个基准测试，代理从真实观测数据中构建ENSO的低阶随机模型，ENSO是年际气候变率的主导模式。在六小时的预算内，代理处理观测数据，撰写自己的诊断报告（这些报告随后被冻结），并仅使用这些诊断作为反馈来开发模型。隐藏的评分者随后测试模型是否再现ENSO的统计特性、恢复未观测变量，并预测被保留的年份，并以相同的方式对已发布模型进行评分。在十二个代理系统中，有六个生成的模型得分高于已发布模型，主要是通过更好的重建和预测。更强模型的简化形式与ENSO的暖冷不对称的两种竞争解释之一相兼容，而这一开放性争论在任务中并未提及。对顶级系统在不同信息下的受控运行表明，其得分并非来自于回忆过时的观测记录，而是所接收的信息塑造了其构建模型的方式。因此，SciExam for ENSO能够评估没有已知答案的代理研究，结果表明代理已经能够构建出在结构上与科学家仍在争论的问题相关的竞争性模型。
cs.AI / 110 / 2610.10515
RoboJEPA: Scaling Robotic Latent World Models
RoboJEPA：扩展机器人潜在世界模型
Abstract
Latent world models have shown a remarkable ability to predict future states and to plan in the real world. In practice, however, we lack a principled way to estimate how their capabilities scale with model size, data, and compute, an open problem that slows progress in the field. In this work we present RoboJEPA, a world model based on the Joint Embedding Predictive Architecture (JEPA) and trained on a large-scale dataset spanning 12 robotic embodiments. We show that RoboJEPA's imagination error, the error of its latent rollouts, follows a second-order power law in compute, allowing us to predict model quality well beyond the scale at which the law is fit. We further show that downstream robotic planning performance improves predictably with compute, and that imagination error is strongly correlated with it, making it a reliable proxy for real-robot evaluation. Finally, we demonstrate that latent world models can be deployed zero-shot as robotic agents, planning toward a single goal image to solve tasks requiring long-horizon planning on real hardware. We release all model checkpoints together with our training and robot deployment code. To our knowledge, this is the first work to establish scaling laws for multi-embodiment robotic world models trained on real robot data, and RoboJEPA, at 8B parameters, is the largest JEPA predictor model trained to date.
Chinese Translation
潜在世界模型在预测未来状态和进行现实世界规划方面展现出了显著的能力。然而，在实践中，我们缺乏一种原则性的方法来估计其能力如何随模型规模、数据和计算能力的变化而变化，这一开放性问题减缓了该领域的进展。在本研究中，我们提出了RoboJEPA，这是一种基于联合嵌入预测架构（Joint Embedding Predictive Architecture, JEPA）的世界模型，训练于一个涵盖12种机器人表现的大规模数据集。我们展示了RoboJEPA的想象误差，即其潜在展开的误差，遵循计算能力的二阶幂律，从而使我们能够预测模型质量，超出该规律适用的规模。我们进一步表明，下游机器人规划性能随着计算能力的增加而可预测地改善，并且想象误差与之高度相关，使其成为真实机器人评估的可靠代理。最后，我们演示了潜在世界模型可以零样本部署为机器人代理，针对单一目标图像进行规划，以解决需要长时间规划的任务在真实硬件上进行。我们发布了所有模型检查点以及我们的训练和机器人部署代码。据我们所知，这是首个为基于真实机器人数据训练的多表现机器人世界模型建立扩展规律的研究，而RoboJEPA以80亿参数成为迄今为止最大的JEPA预测模型。
cs.CL / 1 / 2610.08794
Tokka-Bench: Evaluating Tokenizers Across 100 Natural and 20 Programming Languages
Tokka-Bench：跨越100种自然语言和20种编程语言评估分词器
Abstract
Large language models rely on subword tokenizers whose quality varies across languages, yet no standardized multi-metric framework exists for broad comparative evaluation. We introduce Tokka-Bench, an open-source framework that evaluates tokenizers on five complementary metrics -- bytes per token, unique token coverage, subword fertility, word-split rate, and vocabulary composition -- across 100 natural languages (30+ scripts) and 20 programming languages, using language-aware segmentation adapted to each writing system. Comparing seven BPE tokenizers (GPT-2, GPT-4, gpt-oss, Llama 3.1, Gemma 3, Qwen3, and Kimi K2) within individual languages, we find that vocabulary allocation strategy matters more than raw vocabulary size, and that programming-language efficiency has converged among recent tokenizers despite divergent natural-language profiles. The framework, data, and interactive dashboard are publicly available.
Chinese Translation
大型语言模型依赖于子词分词器，而其质量在不同语言中存在差异，但目前尚无标准化的多指标框架用于广泛的比较评估。我们推出了Tokka-Bench，这是一个开源框架，基于五个互补指标——每个token的字节数、唯一token覆盖率、子词丰度、单词拆分率和词汇组成——对100种自然语言（30多种书写系统）和20种编程语言的分词器进行评估，采用适应于各书写系统的语言感知分割。通过比较七种BPE分词器（GPT-2、GPT-4、gpt-oss、Llama 3.1、Gemma 3、Qwen3和Kimi K2）在各自语言中的表现，我们发现词汇分配策略比原始词汇大小更为重要，并且尽管自然语言特征各异，近期的分词器在编程语言的效率上已趋于一致。该框架、数据和交互式仪表板均已公开提供。
cs.CL / 2 / 2610.08827
Child ASR Adaptation with Adult Retention: An Empirical Study
儿童自动语音识别适应与成人保留：一项实证研究
Abstract
Automatic Speech Recognition (ASR) systems often underperform for children and non-native speakers, while adapting adult ASR models to child speech can cause adult-speech forgetting. We study child ASR adaptation with adult retention across Arabic and English. We compare full fine-tuning, LoRA, and post-hoc weight-space merging across encoder--decoder, encoder--CTC, and AudioLLM-based ASR systems. Experiments use Arabic native and non-native child speech, English MyST child speech, and adult benchmarks from MGB-2 and LibriSpeech test-clean. We evaluate recognition quality with WER and quantify the adaptation--retention trade-off using Retention Index, Child Adaptation Gain, and Adaptation Recovery. Results show that child adaptation is necessary, especially for non-native Arabic and English child speech, but direct adaptation often reduces adult ASR performance. Bilingual adaptation is more stable than language-specific adaptation. Weight-space merging often improves the trade-off, especially for encoder--CTC, Whisper, and AudioLLM-based ASR, with LERP favoring adult retention and TIES recovering stronger child gains. For the encoder--decoder model, direct bilingual fine-tuning remains strongest in raw WER.\footnote{Code, and models are available at https://github.com/qcri/Child-ASR-Adaptation.
Chinese Translation
自动语音识别（ASR）系统在儿童和非母语者中的表现往往不佳，而将成人ASR模型适应于儿童语音可能导致成人语音的遗忘。我们研究了在阿拉伯语和英语中进行儿童ASR适应与成人保留的情况。我们比较了全量微调、LoRA和事后权重空间合并在编码器-解码器、编码器-CTC和基于AudioLLM的ASR系统中的表现。实验使用了阿拉伯语母语和非母语儿童语音、英语MyST儿童语音，以及来自MGB-2和LibriSpeech test-clean的成人基准数据。我们通过词错误率（WER）评估识别质量，并使用保留指数（Retention Index）、儿童适应增益（Child Adaptation Gain）和适应恢复（Adaptation Recovery）量化适应与保留之间的权衡。结果表明，儿童适应是必要的，尤其是对于非母语的阿拉伯语和英语儿童语音，但直接适应往往会降低成人ASR的性能。双语适应比特定语言的适应更为稳定。权重空间合并通常能改善权衡，尤其是在编码器-CTC、Whisper和基于AudioLLM的ASR中，其中LERP有利于成人保留，而TIES则能更强地恢复儿童增益。对于编码器-解码器模型，直接的双语微调在原始WER中仍然表现最强。
cs.CL / 3 / 2610.08828
When Forgetting Looks Like Improvement: Metric Masking in Streaming Diarizer Adaptation and the Price of Rehearsal
当遗忘看似改善：流式说话人分离适应中的度量掩蔽及其排练代价
Abstract
Small-data adaptation can improve speech detection while degrading speaker attribution. We study this discrepancy in a released streaming diarizer adapted on 7.5 h of two-party conversation and evaluated across six corpora. Adaptation substantially improves in-domain diarization performance and transfers to an independent corpus. However, this improvement is not consistent across evaluation scenarios as the additional confusion is mainly associated with impaired temporal identity consistency rather than speaker-count errors. A local-remapping diagnostic reveals different patterns of identity degradation across corpora, indicating that adaptation may alter how streaming models maintain speaker assignments over time. Rehearsal reduces the observed degradation but reduces the cross-domain transfer performance. These results highlight the need to jointly evaluate detection accuracy, identity consistency, and retention behavior when adapting streaming diarization systems.
Chinese Translation
小数据适应可以提高语音检测的准确性，但却降低了说话人归属的准确性。我们研究了这一差异，分析了在7.5小时的双人对话上适应的流式说话人分离系统，并在六个语料库上进行了评估。适应显著提高了领域内的说话人分离性能，并能够迁移到独立的语料库。然而，这种改善在不同评估场景中并不一致，因为额外的混淆主要与时间身份一致性的损害有关，而不是说话人数量错误。局部重映射诊断揭示了不同语料库中身份退化的不同模式，表明适应可能改变流式模型随时间维持说话人分配的方式。排练减少了观察到的退化，但也降低了跨领域的迁移性能。这些结果强调了在适应流式说话人分离系统时，需要共同评估检测准确性、身份一致性和保持行为。
cs.CL / 4 / 2610.08829
Emo-Jev: Probabilistic Reasoning for Emotion Classification with Jev
Emo-Jev：基于Jev的情感分类概率推理
Abstract
Jev offers an alternative interface for language understanding: given an input and predefined questions, it returns probabilistic decisions rather than free-form responses. Whether this interface can support effective reasoning for text classification against leading LLMs remains an open questions. We introduce Emo-Jev, a training-free framework with two complementary implementations. Emo-Jev-D decomposes classification into task-specific atomic judgments and composes their probabilities into a final prediction. Emo-Jev-SC constructs multiple judgment paths from complementary perspectives and aggregates their predictions into a consensus decision. We evaluate Emo-Jev on eight datasets spanning sentiment analysis, emotion recognition, sarcasm detection and humor detection, comparing against direct Jev classification and five SoTA LLMs under input/output and chain-of-thought reasoning. Standard Jev achieves 62.93\% average macro-F1 versus 67.28\% for the strongest LLM baseline, with lower observed latency and generally lower cost.
Chinese Translation
Jev 提供了一种替代的语言理解接口：给定输入和预定义问题，它返回概率决策而非自由形式的响应。该接口是否能够支持有效的文本分类推理，与领先的 LLM（大型语言模型）相比仍然是一个未解之谜。我们介绍了 Emo-Jev，这是一种无训练框架，具有两种互补的实现方式。Emo-Jev-D 将分类分解为特定任务的原子判断，并将其概率组合成最终预测。Emo-Jev-SC 从互补的角度构建多个判断路径，并将其预测汇总为共识决策。我们在八个数据集上评估 Emo-Jev，涵盖情感分析、情感识别、讽刺检测和幽默检测，并与直接的 Jev 分类和五个最先进的 LLM 进行比较，涉及输入/输出和思维链推理。标准 Jev 的平均宏 F1 达到 62.93\%，而最强 LLM 基线为 67.28\%，且观察到的延迟更低，成本通常也较低。
cs.CL / 5 / 2610.08833
CoDR: Training-Free Confidence-Drift Remasking for Diffusion Language Models
CoDR：无训练的信心漂移重掩码用于扩散语言模型
Abstract
Masked diffusion language models (MDLMs) decode by repeatedly committing tokens to masked positions, but these commitments are usually irreversible. A token chosen under sparse, partial context is kept fixed, even when later context no longer supports it. Existing samplers mainly decide when to commit a token, but rarely check whether an already committed token should still be kept, allowing early mistakes to propagate. We trace this issue to confidence drift, where the model's confidence in a committed token drops from its sparse commit-time context to the denser context available later. Based on this signal, we propose CoDR (Confidence Drift Remasking), a training-free and sampler-agnostic refinement pass. CoDR estimates drift for all committed positions in only k forward passes via k-partition probing, then remasks and regenerates only the tokens the model no longer endorses. Across two backbones, four reasoning and coding tasks, and three base samplers, CoDR improves average accuracy across all evaluated model-sampler configurations and improves most individual task settings with modest overhead. Controlled experiments show that the gains come from targeted confidence-drift remasking rather than extra compute alone, and that CoDR uses far fewer forward passes than prior remasking methods. Code is available at https://github.com/YueWu0301/CoDR.
Chinese Translation
掩码扩散语言模型（MDLMs）通过反复将标记提交到掩码位置来解码，但这些提交通常是不可逆的。在稀疏的部分上下文下选择的标记保持不变，即使后续的上下文不再支持它。现有的采样器主要决定何时提交标记，但很少检查已提交的标记是否仍应保留，从而导致早期错误的传播。我们将这个问题追溯到信心漂移，即模型对已提交标记的信心从稀疏的提交时上下文下降到后续可用的更密集的上下文。基于这一信号，我们提出了CoDR（信心漂移重掩码），这是一种无训练且与采样器无关的精炼过程。CoDR通过k-分区探测在仅k次前向传递中估计所有已提交位置的漂移，然后仅重掩码和再生成模型不再支持的标记。在两个基础模型、四个推理和编码任务以及三个基础采样器中，CoDR提高了所有评估的模型-采样器配置的平均准确性，并在大多数单个任务设置中以适度的开销取得了改善。控制实验表明，增益来自于针对性的信心漂移重掩码，而不仅仅是额外的计算，并且CoDR使用的前向传递次数远少于先前的重掩码方法。代码可在 https://github.com/YueWu0301/CoDR 获取。
cs.CL / 6 / 2610.08835
Leveraging LLM-Generated Explanations for Detecting Emotionally Rewritten Fake News
利用大语言模型生成的解释检测情感重写的假新闻
Abstract
The spread of fake news may cause severe social consequences. Existing fake news detection methods mainly focus on stylistic variations or incorporate external information such as explanations. However, news articles are often rewritten under different emotional backgrounds while preserving their underlying factual claims, which may affect the robustness of detection models. In this work, we investigate fake news detec- tion under fact-preserving emotional variations. To study this problem, we construct emotion-rewritten test sets and generate explanations from the original news articles as stable background knowledge. We then propose a Gated Cross Attention (GCA) framework that adaptively integrates emotionally rewritten news with the corresponding explanations, enabling the model to focus on informative explanation content while reducing potential mismatches caused by emotional reframing. Experiments on PolitiFact, GossipCop, and LUN demonstrate that the proposed method achieves notable improvements under multiple emotional conditions on PolitiFact and LUN, while maintaining competitive performance on GossipCop. We further analyze the effects of explanation guidance and gating mechanisms under different emotional conditions. Our code and data are available at: https://github.com/Flulike/fakenews gca .
Chinese Translation
假新闻的传播可能导致严重的社会后果。现有的假新闻检测方法主要关注风格变异或结合外部信息，例如解释。然而，新闻文章在不同情感背景下经常被重写，同时保留其基本事实主张，这可能影响检测模型的鲁棒性。在本研究中，我们探讨在保持事实的情感变异下进行假新闻检测。为研究这一问题，我们构建了情感重写测试集，并从原始新闻文章中生成解释作为稳定的背景知识。然后，我们提出了一种门控交叉注意力（Gated Cross Attention, GCA）框架，该框架自适应地将情感重写的新闻与相应的解释结合，使模型能够关注信息丰富的解释内容，同时减少因情感重构造成的潜在不匹配。在PolitiFact、GossipCop和LUN上的实验表明，所提出的方法在PolitiFact和LUN的多种情感条件下取得了显著的改进，同时在GossipCop上保持了竞争力的表现。我们进一步分析了在不同情感条件下解释指导和门控机制的效果。我们的代码和数据可在以下链接获取：https://github.com/Flulike/fakenews gca 。
cs.CL / 7 / 2610.08840
Beyond the Sycophancy Score: How Task, Model, and Pressure Shape LLM Yielding
超越谄媚评分：任务、模型和压力如何影响大型语言模型的表现
Abstract
Large language models (LLMs) often abandon a correct answer, or endorse a user's position, once the user pushes back. This behavior, called sycophancy, is usually reported as a single rate per model, which says little about when it happens or how a user can avoid it. We study the conditions that produce it with 103,939 graded replies from ten configurations: eight LLMs with reasoning disabled, and two of them again with maximum reasoning, all facing the same 200 items, 13 pressure conditions, and four-turn conversations, with every reply labeled by two independent LLM judges. We find that the dominant factors are how costly it is for the model to verify the user's claim, and whether a trained guardrail covers it. Removing this task factor from a logistic model costs 0.485 of McFadden $R^2$, against 0.139 for model family and 0.009 for pressure tactic. Anchored facts are almost never conceded (1.3%), while adoption on logic puzzles rises with the number of clues needed to refute the pushed answer. Personal choices are endorsed in 77.0% of conversations. Most concessions on hard items come from models that cannot reliably solve them; models that can solve them rarely give the answer up. For both models tested, maximum reasoning removes these concessions completely: adoption on deep puzzles falls from 19.2% and 12.5% to 0%. Fallacious or emotional framing adds nothing beyond plain repetition. Three human annotators agree with the judges' consensus on 118/120 calibration items. These results give practical rules for reliable use: simplify hard-to-verify problems and reason deeply, state the question rather than one's preferred answer, ask for evidence on open questions, and choose models by their measured guardrail profile.
Chinese Translation
大型语言模型（LLMs）在用户反驳时，往往会放弃正确答案或支持用户的立场。这种行为被称为谄媚，通常以每个模型的单一比率报告，这对其发生的时机或用户如何避免它几乎没有提供信息。我们通过分析来自十种配置的103,939条评分回复来研究产生这种行为的条件：八个禁用推理的LLM，以及其中两个在最大推理下的表现，所有模型面对相同的200个项目、13种压力条件和四轮对话，每条回复由两个独立的LLM评审标记。我们发现，主导因素是模型验证用户主张的成本，以及是否有训练好的保护措施覆盖该主张。从逻辑模型中移除这一任务因素导致McFadden $R^2$损失0.485，而模型家族的损失为0.139，压力策略的损失为0.009。锚定事实几乎从不被让步（1.3%），而在逻辑难题中，推翻被推动答案所需线索的数量越多，采纳率就越高。个人选择在77.0%的对话中得到支持。大多数在困难项目上的让步来自于无法可靠解决这些问题的模型；能够解决这些问题的模型则很少放弃答案。对于测试的两个模型，最大推理完全消除了这些让步：在深度难题上的采纳率从19.2%和12.5%降至0%。虚假或情感框架除了简单重复外没有增加任何内容。三位人类注释者在118/120个校准项目上与评审的共识达成一致。这些结果为可靠使用提供了实用规则：简化难以验证的问题并进行深入推理，陈述问题而非个人偏好答案，针对开放性问题请求证据，并根据测量的保护措施特征选择模型。
cs.CL / 8 / 2610.08842
Beyond Risk Prediction: Evidence Grounding and Psychosocial Factor Verification for Explainable Suicide Risk Assessment
超越风险预测：可解释自杀风险评估的证据基础和心理社会因素验证
Abstract
Identifying suicide risk from social networking services (SNS) posts is important for detecting suicide-related signals in online environments. However, risk classification alone provides limited insight into the textual evidence and psychosocial factors behind a prediction. Based on the IEEE BigData 2026 Explainable Suicide Risk Detection Challenge, this study presents a framework consisting of Risk Assessment, Evidence Grounding, and Factor Identification. Risk Assessment uses length-based routing to accommodate posts of different lengths. Evidence Grounding identifies supporting phrases and uses a Risk-Evidence constraint to maintain consistency with the Risk prediction. For Factor Identification, two verifiers are used. The Taxonomy Verifier focuses on factor semantics, whereas the Evidence-Aware Verifier uses factor-specific lexical-semantic cues to select informative positive training units. Their prediction probabilities are combined to produce the final factor predictions. The three tasks are evaluated using task-specific F1 score measures. Risk Assessment achieved a Weighted F1 of 0.8088, Evidence Grounding achieved a test Macro row F1 of 0.7605, and Factor Identification achieved a Macro F1 of 0.5562. The results show that the framework can provide risk predictions, along with supporting textual evidence and fine-grained information on psychosocial factors. Overall, the proposed framework extends suicide-risk assessment beyond risk-level prediction and provides a more interpretable analysis of SNS posts.
Chinese Translation
从社交网络服务（SNS）帖子中识别自杀风险对于在在线环境中检测与自杀相关的信号至关重要。然而，仅仅进行风险分类对预测背后的文本证据和心理社会因素提供的洞察有限。基于IEEE BigData 2026可解释自杀风险检测挑战，本研究提出了一个由风险评估、证据基础和因素识别组成的框架。风险评估使用基于长度的路由来适应不同长度的帖子。证据基础识别支持性短语，并使用风险-证据约束来保持与风险预测的一致性。在因素识别方面，使用了两个验证器。分类验证器关注因素语义，而证据感知验证器使用特定于因素的词汇-语义线索来选择信息丰富的正训练单元。它们的预测概率被结合以生成最终的因素预测。这三个任务使用任务特定的F1分数进行评估。风险评估的加权F1达到了0.8088，证据基础的测试宏行F1达到了0.7605，而因素识别的宏F1达到了0.5562。结果表明，该框架能够提供风险预测，同时提供支持性文本证据和关于心理社会因素的细粒度信息。总体而言，所提出的框架将自杀风险评估扩展到超越风险水平预测，并提供对SNS帖子更具可解释性的分析。
cs.CL / 9 / 2610.08851
QuanLing: Cross-Branch Validation of Language Distance Quantification on Western Romance
QuanLing：西方罗曼语系语言距离量化的跨分支验证
Abstract
Quantifying language distance among closely related languages remains a core challenge in quantitative linguistics. Our previous work [1] introduced QuanLing (Quantitative Linguistics via Pretrained Language Models), a quantitative framework combining language distance metrics (sentence embedding distance, tokenization fragmentation rate) with language property analysis (MLM prediction probability), validated on North Germanic (Danish, Norwegian Bokm{\aa}l, Swedish). This paper extends QuanLing to Western Romance--French, Portuguese, Spanish, Italian--testing cross-branch applicability with the same metric family and aggregation protocol as our North Germanic study, adapted for four languages (English anchor, quadruplet construction). Using 150 four-language parallel sentences, we compute LaBSE sentence embedding distances, tokenization fragmentation rates from four monolingual BERT tokenizers, and mBERT masked language model mutual intelligibility. Results show that Portuguese--Spanish are closest (LaBSE distance 0.0229), French--Italian most distant (0.0338); LaBSE and mBERT rankings agree on 4 of 6 pairs, confirming cross-model robustness. Western Romance shows a wider absolute distance span than North Germanic (0.011 vs. 0.008) but comparable relative ratios (1.48 vs. 1.67), consistent with longer divergence time. French exhibits notably higher MLM predictability (36.12% top-1 accuracy vs. 29.28% for Italian), reflecting its orthography--phonology decoupling. This cross-branch validation provides further evidence for QuanLing's generalizability beyond a single language branch.
Chinese Translation
在紧密相关语言之间量化语言距离仍然是定量语言学中的一个核心挑战。我们之前的研究[1]引入了QuanLing（通过预训练语言模型的定量语言学），这是一个结合语言距离度量（句子嵌入距离、分词碎片率）与语言属性分析（MLM预测概率）的定量框架，已在北日耳曼语（丹麦语、挪威博克马尔语、瑞典语）上进行了验证。本文将QuanLing扩展到西方罗曼语系——法语、葡萄牙语、西班牙语、意大利语——测试相同度量家族和聚合协议的跨分支适用性，适用于四种语言（英语锚点、四元组构造）。使用150个四语言平行句子，我们计算了LaBSE句子嵌入距离、来自四个单语BERT分词器的分词碎片率，以及mBERT掩蔽语言模型的相互可理解性。结果显示，葡萄牙语与西班牙语之间的距离最近（LaBSE距离0.0229），法语与意大利语之间的距离最远（0.0338）；LaBSE和mBERT的排名在6对中的4对上达成一致，确认了跨模型的稳健性。西方罗曼语显示出比北日耳曼语更大的绝对距离范围（0.011对比0.008），但相对比率相似（1.48对比1.67），与更长的分歧时间一致。法语的MLM可预测性显著更高（36.12%的顶级准确率对比意大利语的29.28%），反映了其正字法与音位学的解耦。这一跨分支验证为QuanLing在单一语言分支之外的普适性提供了进一步证据。
cs.CL / 10 / 2610.08858
LRCC: Generalizing Low-Rank Compression with Conditional Computation
LRCC：通过条件计算推广低秩压缩
Abstract
Low-rank compression reduces the cost of pretrained language models by replacing linear transformations with low-rank factorizations. However, conventional methods use a fixed rank allocation during inference, assigning the same amount of compute regardless of the input token. We introduce Low-Rank Conditional Computation (LRCC), which adds token-dependent computation to pretrained models by training one lightweight router per Transformer block to select among a small set of nested low-rank paths. During training, the low-rank factors remain frozen, and only the routers are optimized. We evaluate LRCC on Llama and Qwen models for language modeling and zero-shot downstream tasks. Within the same average active-parameter budget, LRCC improves the predictive performance over static low-rank compression, including a 7.6 percentage-point gain in average downstream accuracy on Llama-2-7B over static methods. At matched batch-size-1 decoding latency, LRCC improves both perplexity and downstream accuracy on Llama-3.2-1B and remains competitive on Llama-2-7B, without specialized kernels. Finally, we assess the usefulness of assigning a token-wise path by analyzing the routers' path choices.
Chinese Translation
低秩压缩通过用低秩分解替代线性变换，降低了预训练语言模型的计算成本。然而，传统方法在推理过程中使用固定的秩分配，无论输入标记如何，都分配相同的计算量。我们提出了低秩条件计算（Low-Rank Conditional Computation, LRCC），通过为每个Transformer模块训练一个轻量级路由器，增加了依赖于标记的计算，以在一小组嵌套的低秩路径中进行选择。在训练过程中，低秩因子保持不变，只有路由器被优化。我们在Llama和Qwen模型上评估LRCC，以进行语言建模和零-shot下游任务。在相同的平均活跃参数预算内，LRCC在静态低秩压缩上提高了预测性能，包括在Llama-2-7B上相比静态方法提高了7.6个百分点的平均下游准确率。在匹配的批量大小为1的解码延迟下，LRCC在Llama-3.2-1B上提高了困惑度和下游准确率，并在Llama-2-7B上保持竞争力，而无需专门的内核。最后，我们通过分析路由器的路径选择来评估分配标记级路径的有效性。
cs.CL / 11 / 2610.08879
Tiny-Scale Chinese BERT Pretraining: A Controlled Comparison of MLM, WWM, and MacBERT Strategies
小规模中文BERT预训练：MLM、WWM和MacBERT策略的对照比较
Abstract
Pretraining strategies significantly impact the quality of language models, yet existing comparisons of Masked Language Modeling (MLM), Whole Word Masking (WWM), and MacBERT-style replacement have focused primarily on base-scale models (>=110M parameters). This paper presents a controlled comparison of these three strategies on a tiny-scale Chinese BERT model (4 layers, 256 hidden dimensions, 8.7M parameters). Under identical architecture, corpus (1.29M sentences from Chinese Wikipedia), and hyperparameters, we train three models from scratch and evaluate them across five intrinsic dimensions: perplexity, MLM hit rate, semantic discrimination, grammatical judgment, and contextual sensitivity. At tiny scale, MLM achieves the best overall intrinsic performance (winning 3 of 5 dimensions), while WWM excels in both perplexity (1.27 vs. 2.10, a 39.5% improvement) and MLM hit rate (22% vs. 16%). Notably, MacBERT under a severely limited synonym dictionary (222 entries, 3.3% coverage) exhibits severe perplexity degradation (47.23, 22x higher than MLM), yielding a ranking (MLM > WWM >> MacBERT) that differs markedly from the established base-scale conclusion (MacBERT > WWM > MLM). We further identify a critical evaluation pitfall: MacBERT achieves the lowest training loss (2.17) yet the highest perplexity (47.23), revealing that training loss alone is unreliable under mixed replacement strategies. All models and corpus are publicly available at https://huggingface.co/eebyp.
Chinese Translation
预训练策略对语言模型的质量有显著影响，但现有的对比研究主要集中在基础规模模型（>=110M参数）上，涉及掩码语言建模（MLM）、整体词掩码（WWM）和MacBERT风格替换。本文对这三种策略在一个小规模中文BERT模型（4层，256个隐层维度，8.7M参数）上进行了受控比较。在相同的架构、语料（来自中文维基百科的1.29M句子）和超参数下，我们从头开始训练了三个模型，并在五个内在维度上进行了评估：困惑度、MLM命中率、语义区分、语法判断和上下文敏感性。在小规模下，MLM在整体内在性能上表现最佳（在5个维度中赢得3个），而WWM在困惑度（1.27对比2.10，改善39.5%）和MLM命中率（22%对比16%）上表现优异。值得注意的是，在一个严重受限的同义词词典（222个条目，覆盖率3.3%）下，MacBERT表现出严重的困惑度退化（47.23，比MLM高出22倍），导致其排名（MLM > WWM >> MacBERT）与已建立的基础规模结论（MacBERT > WWM > MLM）显著不同。我们进一步识别出一个关键的评估陷阱：MacBERT虽然达到了最低的训练损失（2.17），但却有最高的困惑度（47.23），这表明在混合替换策略下，仅依赖训练损失是不可靠的。所有模型和语料均可在 https://huggingface.co/eebyp 获取。
cs.CL / 12 / 2610.08887
Steering Follows Geometry, Not Labels: Emotion Directions in a Full-Duplex Speech Model
引导遵循几何，而非标签：全双工语音模型中的情感方向
Abstract
Full-duplex voice agents need to modulate emotion and delivery during real-time conversations, when de-escalating a complaint, carrying urgency in dispatch, softening a clinical result. Emotion and delivery control is well studied for TTS and turn based models through prompt-conditioned synthesis, reference-conditioned synthesis and activation steering; PersonaPlex controls identity in a duplex model but not affect. We study emotion steering in Moshi, a fully open sourced full-duplex speech language model, across four emotions, using mean-difference activation steering, which costs only a few vector additions per frame and no retraining. We show that emotion is linearly decodable from Moshi's residual stream, but activation steering is only partially achievable, and unevenly so; as happy, angry and surprise steer towards a shared direction while sad is distinctly steerable. We also show that the shared component across the three emotions cannot simply be projected away from all the emotions equally.
Chinese Translation
全双工语音代理在实时对话中需要调节情感和表达，例如在降低投诉情绪、传达紧急情况或缓和临床结果时。情感和表达控制在文本转语音（TTS）和基于轮次的模型中通过提示条件合成、参考条件合成和激活引导得到了充分研究；PersonaPlex 在双工模型中控制身份，但不控制情感。我们研究了在 Moshi 这一完全开源的全双工语音语言模型中对四种情感的引导，采用均值差激活引导，这仅需每帧进行少量向量加法且无需重新训练。我们展示了情感可以从 Moshi 的残差流中线性解码，但激活引导只能部分实现，并且实现程度不均；快乐、愤怒和惊讶这三种情感朝向一个共同方向引导，而悲伤则可以明显地被引导。我们还表明，这三种情感之间的共享成分不能简单地从所有情感中等比例地投影去除。
cs.CL / 13 / 2610.08917
CARE: Certifying Acceleration for Vision-Language-Action Inference
CARE：视觉-语言-动作推理的加速认证
Abstract
While vision-language-action (VLA) models have advanced rapidly, running them at every control step remains expensive. Prior work accelerates VLA inference using techniques like action chunking and visual-token pruning, typically evaluating based on latency and average task success. However, acceleration may discard information and break tasks the original policy would solve, a risk hidden by average metrics. Measuring these failures is challenging because action deviations compound over closed-loop trajectories, meaning task failure is only observable across full episodes. We therefore define an acceleration-induced failure via paired rollouts from identical initial conditions, tracking when the reference succeeds but the accelerated policy fails. To manage this, we introduce CARE, an approach for certified accelerator selection. CARE uses paired rollouts on a calibration set to provide finite-sample guarantees that acceleration-induced failure risk stays below a user-specified budget. It deploys the fastest certified candidate, falling back to the reference if none qualify. By relying only on terminal outcomes and measured compute, CARE applies unchanged across diverse acceleration mechanisms, while sequential testing and failure-triggered reference rollouts keep certification affordable. On four LIBERO suites with OpenVLA-OFT, CARE certifies $9.0$--$10.8\times$ speedups while guaranteeing (at $95\%$ confidence) that at least $85.8\%$ of reference-solved episodes are preserved. Under tight budgets, selectors without guarantees exceed the budget in up to $75\%$ of trials, whereas CARE stays within budget and its sequential form uses $78.9\%$ fewer rollouts than exhaustive evaluation. CARE further generalizes to flow-step reduction for $\pi_{0.5}$, and to Qwen3.5-9B and Llama-3.1-8B agents in Crafter.
Chinese Translation
尽管视觉-语言-动作（VLA）模型发展迅速，但在每个控制步骤中运行它们仍然成本高昂。之前的研究通过动作分块和视觉标记修剪等技术加速VLA推理，通常基于延迟和平均任务成功率进行评估。然而，加速可能会丢失信息并破坏原始策略能够解决的任务，这一风险被平均指标所掩盖。测量这些失败是具有挑战性的，因为动作偏差在闭环轨迹中会累积，这意味着任务失败仅在完整的回合中可观察到。因此，我们通过从相同初始条件的配对回放定义加速引起的失败，跟踪何时参考策略成功而加速策略失败。为了解决这个问题，我们引入了CARE，一种认证加速器选择的方法。CARE在校准集上使用配对回放，提供有限样本保证，确保加速引起的失败风险低于用户指定的预算。它部署最快的认证候选者，如果没有候选者符合条件，则回退到参考策略。通过仅依赖终端结果和测量计算，CARE在各种加速机制中保持不变，而顺序测试和失败触发的参考回放使认证变得经济可行。在四个LIBERO套件中使用OpenVLA-OFT，CARE认证了$9.0$--$10.8 imes$的加速，同时保证（在$95\%$的置信度下）至少$85.8\\%$的参考解决回合得以保留。在预算紧张的情况下，没有保证的选择器在多达$75\\%$的试验中超出预算，而CARE则保持在预算内，其顺序形式比全面评估使用$78.9\\%$更少的回放。CARE进一步推广到$ ext{flow-step}$减少的$ ext{π}_{0.5}$，以及在Crafter中的Qwen3.5-9B和Llama-3.1-8B代理。
cs.CL / 14 / 2610.09033
Quad-State Safety Evaluation of Open-Weight Large Language Models on Non-Canonical Inputs
开放权重大型语言模型在非规范输入上的四状态安全评估
Abstract
Standard safety evaluations of large language models assess harmful requests written in canonical plain text, while models in real-world deployment routinely receive inputs containing emojis, altered spellings, encoded strings, and character-level variations. This work introduces the Adversarial Surface-Form Robustness Dataset (ASRD), comprising 2,100 prompts across seven distinct surface-form families. Five open-weight language models are evaluated across these prompts, producing 10,500 responses. The Quad-State Evaluation Rubric classifies each response into one of four outcomes: harmful compliance, safe response, comprehension failure, or indeterminate. Emoji and invisible Unicode variations cause almost no comprehension failure, with pooled harmful compliance of 20.27% and 17.20% against a 22.87% baseline that is driven mainly by Mistral 7B, whereas leetspeak, encoded wrappers, and hybrid transformations score 2.40%, 0.13%, and 2.40% while comprehension failure rises to 36.47%, 65.60%, and 34.47%. Inspection of raw model outputs reveals three response behaviors: hallucinated benignity, structural collapse, and language drift. Project page: www.pavanmaddula.com/quadstate
Chinese Translation
大型语言模型的标准安全评估主要针对以规范纯文本形式书写的有害请求，而在实际部署中，模型常常接收到包含表情符号、变体拼写、编码字符串和字符级变异的输入。本研究引入了对抗表面形式鲁棒性数据集（Adversarial Surface-Form Robustness Dataset, ASRD），该数据集包含来自七个不同表面形式家族的2100个提示。我们对五个开放权重语言模型在这些提示下进行了评估，产生了10500个响应。四状态评估标准将每个响应分类为四种结果之一：有害合规、安全响应、理解失败或不确定。表情符号和不可见Unicode变异几乎没有导致理解失败，合并的有害合规率为20.27%和17.20%，而基线为22.87%，主要由Mistral 7B驱动；而变体语言（leetspeak）、编码包装和混合转换的得分分别为2.40%、0.13%和2.40%，理解失败率则上升至36.47%、65.60%和34.47%。对原始模型输出的检查揭示了三种响应行为：虚构的良性、结构崩溃和语言漂移。项目页面：www.pavanmaddula.com/quadstate
cs.CL / 15 / 2610.09087
U-Space: Uncovering When and Why Uncertainty Arises in Language Models
U-Space：揭示语言模型中不确定性何时及为何产生
Abstract
Large language models are informing decisions with ever-higher stakes. As the consequences of their errors grow, a central question becomes harder to ignore: how much can we trust an individual answer? Yet recognizing when to defer remains difficult because language models can present incorrect conclusions with fluent explanations and an authoritative tone. Uncertainty quantification seeks to address this disconnect by estimating the reliability of individual predictions. However, many existing methods require repeated generations or separately trained components, and their scalar estimates do not reveal where uncertainty arises or how it evolves during reasoning. Recent work has also shown that generation length can be strongly associated with uncertainty estimates and correctness, raising the question of how much of an estimator's predictive power comes from uncertainty-specific information rather than output length alone. Mechanistic interpretability offers a way to address these limitations by connecting human-interpretable concepts to intermediate model states. Building on this capability, we introduce the U-Space, a low-dimensional subspace that makes a model's evolving uncertainty measurable and interpretable. We identify semantic anchors for doubt and certainty, map their unembedding directions back into the residual space, and combine their contrasts into an orthogonal basis. The U-Lens projects each token state onto these basis vectors, yielding an interpretable token-level uncertainty map that can be inspected directly or aggregated into a scalar uncertainty score. Our approach requires no correctness labels, repeated generations, or training. Across reasoning benchmarks, its confidence score outperforms established baselines under both standard and length-controlled evaluation and transfers more reliably than supervised estimators. Code: https://github.com/s2labres/U-Space.
Chinese Translation
大型语言模型正在影响越来越高风险的决策。随着错误后果的加剧，一个核心问题变得愈发难以忽视：我们能多大程度上信任单个答案？然而，识别何时应当推迟判断仍然困难，因为语言模型可以用流畅的解释和权威的语气呈现错误的结论。不确定性量化试图通过估计个别预测的可靠性来解决这一脱节。然而，许多现有方法需要重复生成或单独训练的组件，并且它们的标量估计并未揭示不确定性产生的具体位置或在推理过程中如何演变。最近的研究还表明，生成长度与不确定性估计和正确性之间可能存在强关联，这引发了一个问题：估计器的预测能力有多少来自于特定于不确定性的信息，而不仅仅是输出长度。机械可解释性提供了一种解决这些局限性的方法，通过将人类可解释的概念与中间模型状态连接起来。在此能力的基础上，我们引入了U-Space，一个低维子空间，使模型不断演变的不确定性可测量且可解释。我们识别出怀疑和确定性的语义锚点，将它们的去嵌入方向映射回残差空间，并将它们的对比结合成一个正交基。U-Lens将每个标记状态投影到这些基向量上，生成一个可解释的标记级不确定性图，可以直接检查或聚合成标量不确定性分数。我们的方法不需要正确性标签、重复生成或训练。在推理基准测试中，其置信分数在标准和长度控制的评估下均优于既定基线，并且比监督估计器更可靠地进行迁移。代码：https://github.com/s2labres/U-Space。
cs.CL / 16 / 2610.09111
Same Text, Different Prediction: Serving-Context Nondeterminism in Text Classifiers
相同文本，不同预测：文本分类器中的服务上下文非确定性
Abstract
Deterministic inference is essential for reliable and trustworthy machine learning. Prior studies of text generation have shown that changing factors such as batch size, batch composition, hardware, or inference engine can alter the generated text, even when the prompt, model parameters, and sampling randomness are fixed. These differences have been attributed in part to floating-point non-associativity, shape-dependent kernel selection, and other implementation-level differences in numerical execution. However, it remains unclear whether, when, and to what extent the same factors affect text classification. We present a systematic study of serving-context non-invariance in text classifiers, which prior work has measured only through generated text. We train 180 models spanning discriminative, pseudo-generative, and fully generative classifier formulations and evaluate each across four categories of serving contexts, holding the checkpoint and the text fixed. Label stability does not imply score stability. Changing only the batch shape changes no labels across fp32 comparisons, yet under bf16 it moves up to 56.7 percentage points of predicted probability mass, with label changes concentrated at small margins. Fully generative classifiers change more labels than their discriminative counterparts under the same serving changes. We derive sufficient conditions for label stability under each serving change and give a separate mitigation for each mechanism. Our results identify and quantify the serving conditions that must be fixed for reproducible text classification.
Chinese Translation
确定性推理对于可靠和可信的机器学习至关重要。先前的文本生成研究表明，改变批量大小、批量组成、硬件或推理引擎等因素可以改变生成的文本，即使提示、模型参数和采样随机性保持不变。这些差异部分归因于浮点数非结合性、形状依赖的内核选择以及其他数值执行中的实现级差异。然而，目前尚不清楚相同因素在多大程度上、何时影响文本分类。我们对文本分类器中的服务上下文非不变性进行了系统研究，而之前的工作仅通过生成文本进行了测量。我们训练了180个模型，涵盖了判别式、伪生成式和完全生成式分类器的不同形式，并在四种服务上下文类别中评估每个模型，同时保持检查点和文本不变。标签稳定性并不意味着得分稳定性。仅改变批量形状在fp32比较中不会改变任何标签，但在bf16下则会移动多达56.7个百分点的预测概率质量，标签变化集中在小幅度的边际上。在相同的服务变化下，完全生成式分类器改变的标签数量多于其判别式对应物。我们推导了每种服务变化下标签稳定性的充分条件，并为每种机制提供了单独的缓解措施。我们的结果识别并量化了必须固定的服务条件，以实现可重复的文本分类。
cs.CL / 17 / 2610.09152
sk-bench: A Native-First Benchmark for Evaluating Large Language Models in Slovak
sk-bench：用于评估斯洛伐克大型语言模型的本土优先基准
Abstract
Multilingual LLM benchmarks omit Slovak, a morphologically rich West Slavic language of five million speakers, or cover it only by machine translation. We present sk-bench, a native-first Slovak benchmark with 30 datasets (33 scored task variants) across ten skill categories. Eleven resources are introduced or first packaged for generative-LLM evaluation, including IFEval-SK with Slovak-adapted instruction checkers and native Chiby/SKJ1 resources for Slovak grammar and morphology. We evaluate 55 open- and closed-weights models under one harness. The best open model trails proprietary APIs by 12.6 points. Model rankings are similar for native and translated closed-form data ($\rho\geq0.98$), though translation separates the strongest models less well. By contrast, human-authored and LLM-generated QA questions rank models differently ($\rho=0.72$). For Qwen3-14B, continued Slovak pretraining lowers the overall score by 13.9 points. A small instruction set restores three quarters of that loss. Test-time reasoning improves scores by 8.5 to 12.5 points for models of 9B and above. Together, these findings suggest four design lessons for other under-resourced languages: use native data where translation fails, plan instruction repair after language adaptation, enable test-time reasoning before scaling up, and avoid overinvesting in target-language prompts. We release the data and code at https://github.com/slovak-nlp/sk-bench
Chinese Translation
多语言大型语言模型基准忽略了斯洛伐克语，这是一种具有丰富形态变化的西斯拉夫语言，使用者约有五百万，或仅通过机器翻译进行覆盖。我们提出了 sk-bench，这是一个以本土为优先的斯洛伐克基准，包含30个数据集（33个评分任务变体），涵盖十个技能类别。我们引入或首次打包了11个资源用于生成型大型语言模型的评估，包括具有斯洛伐克适应性指令检查器的 IFEval-SK 以及用于斯洛伐克语法和形态学的本土 Chiby/SKJ1 资源。我们在一个统一平台下评估了55个开放和封闭权重模型。最佳开放模型的表现比专有 API 低12.6分。对于本土和翻译的封闭形式数据，模型排名相似（$
ho
geq0.98$），尽管翻译对最强模型的区分效果较差。相比之下，人类创作和大型语言模型生成的问答问题对模型的排名有所不同（$
ho=0.72$）。对于 Qwen3-14B，持续的斯洛伐克预训练使整体得分降低了13.9分。一个小的指令集恢复了四分之三的损失。测试时推理为9B及以上的模型提高了8.5到12.5分的得分。这些发现共同提出了对其他资源匮乏语言的四个设计教训：在翻译失败的情况下使用本土数据，计划在语言适应后进行指令修复，在扩大规模之前启用测试时推理，并避免在目标语言提示上过度投资。我们将在 https://github.com/slovak-nlp/sk-bench 发布数据和代码。
cs.CL / 18 / 2610.09163
ToolRACER: A Robust Agentic Conversation Emulation Resource for Agent Training and Evaluation
ToolRACER：一种用于代理训练和评估的稳健代理对话模拟资源
Abstract
Task-oriented conversational agents remain fragile under real world conversation scenarios as they rarely follow a predictable script, especially when users exhibit non-cooperative behavior. Existing function-calling benchmarks often emphasize successful, cooperative interactions and underrepresent adversarial conversation trajectories, thereby limiting the training resources available for developing robust agents. We present ToolRACER, a synthetic data generation pipeline that coordinates user, assistant and tool emulation models to generate and validated multi-turn interactions between a user and an agent. Using \sysn, we construct ToolRACERBench a robust multi-turn conversation benchmark spanning six domains, ranging over 55 varied personas, generating a validated corpus of 5.6K conversation trajectories, with approximately 66\% of conversations containing failure-prone conversation scenarios. We inject adversarial behaviors, producing validated conversational interaction trajectories that capture realistic, robust scenarios. We evaluate models trained on ToolRACERBench against internal benchmarks, as well as on function calling benchmarks such as $\tau^2$-bench, BFCLv3 and ACEBench to evaluate agentic accuracy and robustness. Models trained on ToolRACERBench improve end to end agentic accuracy across $\tau^2$-bench and ACEBench, demonstrating significant gains when mixed with in-domain dataset in small language models for agent capability tasks.
Chinese Translation
任务导向的对话代理在现实世界对话场景中仍然脆弱，因为它们很少遵循可预测的脚本，尤其是在用户表现出不合作行为时。现有的功能调用基准通常强调成功的合作互动，而对对抗性对话轨迹的表现不足，从而限制了开发稳健代理所需的训练资源。我们提出了ToolRACER，一个合成数据生成管道，它协调用户、助手和工具模拟模型，以生成和验证用户与代理之间的多轮互动。使用 extit{ToolRACER}，我们构建了ToolRACERBench，一个涵盖六个领域的稳健多轮对话基准，涉及55种不同的人物，生成了一个经过验证的5.6K对话轨迹语料库，其中约66 ext{%}的对话包含易出错的对话场景。我们注入对抗性行为，产生经过验证的对话互动轨迹，捕捉现实且稳健的场景。我们在ToolRACERBench上评估训练模型，并与内部基准以及功能调用基准如$ au^2$-bench、BFCLv3和ACEBench进行对比，以评估代理的准确性和稳健性。训练于ToolRACERBench的模型在$ au^2$-bench和ACEBench上提高了端到端的代理准确性，在与小型语言模型的领域内数据集混合时，在代理能力任务上表现出显著提升。
cs.CL / 19 / 2610.09209
Multi-Objective Aligned Small Language Model Framework for SUD Patient Dialogue Generation
多目标对齐的小型语言模型框架用于物质使用障碍患者对话生成
Abstract
Substance Use Disorder (SUD) counseling requires patient responses that reflect underlying cognitive states such as beliefs, coping strategies, and readiness for change. Although large language models (LLMs) can generate fluent text, they often fail to produce cognitively coherent and clinically realistic patient behavior, especially under ethical and data-scarce clinical settings. Moreover, deploying frontier-scale LLMs in healthcare applications presents practical challenges including high computational cost, latency, privacy concerns, and limited deployability in resource-constrained environments, motivating the need for cognitively aligned small language models (SLMs). We propose a cognitively grounded framework for SUD patient dialogue generation that explicitly models and aligns latent cognitive components with patient histories and counselor questions. Our pipeline consists of two stages: cognitive component detection and cognitive component-aligned dialogue generation. To enable effective learning with smaller models, we combine knowledge distillation from high-capacity teacher models, preference optimization from human-annotations, and attention-guided reward shaping. Extensive evaluations using automatic scores like BERTScore, ROUGE, METEOR and BLEU, and LLM-as-judge hit-metrics against both human and teacher-model references show that cognitively informed fine-tuning substantially improves cognitive realization and alignment over a generic instruction-tuned baselines and mental health domain specific SLMs, with particularly strong gains for open-ended cognitive components.
Chinese Translation
物质使用障碍（SUD）咨询需要患者的回应反映出潜在的认知状态，如信念、应对策略和改变的准备程度。尽管大型语言模型（LLMs）能够生成流畅的文本，但它们往往无法产生认知上连贯且临床上真实的患者行为，尤其是在伦理和数据稀缺的临床环境中。此外，在医疗应用中部署前沿规模的LLMs面临实际挑战，包括高计算成本、延迟、隐私问题以及在资源受限环境中的有限可部署性，这促使我们需要认知对齐的小型语言模型（SLMs）。我们提出了一个以认知为基础的SUD患者对话生成框架，该框架明确建模并对齐潜在的认知组件与患者历史和咨询师问题。我们的流程包括两个阶段：认知组件检测和认知组件对齐的对话生成。为了使小型模型能够有效学习，我们结合了来自高容量教师模型的知识蒸馏、人类注释的偏好优化和基于注意力的奖励塑造。使用自动评分（如BERTScore、ROUGE、METEOR和BLEU）和LLM作为评判标准的命中指标对比人类和教师模型参考的广泛评估显示，认知信息驱动的微调在认知实现和对齐方面显著优于通用指令调优的基线和心理健康领域特定的SLMs，尤其是在开放式认知组件方面取得了特别显著的提升。
cs.CL / 20 / 2610.09321
Dialect-Robust Speech Language Models with Synthetic Pseudo-Dialect Augmentation
具有合成伪方言增强的方言鲁棒语音语言模型
Abstract
Speech Language Model (SLM) performance often degrades on dialects due to data scarcity. Conventional text-to-speech (TTS) augmentation struggles to cover diverse dialects as it requires a certain amount of real dialect speech. We propose synthesizing pseudo-dialect speech by converting LLM-generated dialect text via a standard-language TTS model, requiring zero real dialect speech. Additionally, we introduce intermediate standard-text prediction during training, acting as semantic normalization for downstream tasks. We evaluate dialect understanding via dialect-to-English speech translation across Japanese, German, and Chinese dialects. Compared to synthetic standard speech baselines, pseudo-dialect augmentation improves scores for Japanese (from 25.38 to 26.24) and German (from 31.57 to 32.47). Furthermore, the intermediate standard-text prediction effectively bridges the semantic gap, boosting performance to 28.26 for Japanese and from 11.67 to 16.37 for Chinese. These results suggest that our approach scales to various languages without requiring speech resources specific to each dialect.
Chinese Translation
语音语言模型（SLM）的性能在方言上往往会因数据稀缺而下降。传统的文本到语音（TTS）增强在覆盖多样化方言方面面临挑战，因为它需要一定数量的真实方言语音。我们提出通过标准语言TTS模型将大型语言模型（LLM）生成的方言文本转换为合成伪方言语音，从而无需真实方言语音。此外，我们在训练过程中引入中间标准文本预测，作为下游任务的语义规范化。我们通过日语、德语和汉语方言的方言到英语语音翻译评估方言理解能力。与合成标准语音基线相比，伪方言增强提高了日语（从25.38提升至26.24）和德语（从31.57提升至32.47）的得分。此外，中间标准文本预测有效地弥补了语义差距，使日语的性能提升至28.26，汉语从11.67提升至16.37。这些结果表明，我们的方法可以扩展到多种语言，而无需针对每种方言的语音资源。
cs.CL / 21 / 2610.09346
OnlineQAT: On-Policy Distillation for Ultra-Low-Bit Large Language Models
在线量化感知训练：用于超低比特大语言模型的在线蒸馏
Abstract
Quantization-aware training (QAT) can recover much of the accuracy lost when large language models are compressed below four bits. Existing re- covery stages, however, are commonly optimized on fixed completions or teacher-generated answers, whereas the deployed quantized model condi- tions on prefixes generated by itself. Quantization errors can therefore move the model into states that are absent from offline recovery data. We introduce OnlineQAT, a two-stage framework that first obtains a usable low-bit initialization through block-wise QAT and then performs on-policy distillation (OPD) on student-generated responses. At each visited pre- fix, a frozen full-precision teacher provides a sampled reverse-KL training signal. On Qwen3-1.7B, OnlineQAT obtains the best average among the compared quantized methods: 57.28 at W3A16 and 32.52 at W2A16, im- proving over ReasoningQAT by 2.90 and 0.44 points, respectively. The results suggest that student-visited states provide a useful recovery signal beyond fixed-completion training, particularly at three bits.
Chinese Translation
量化感知训练（QAT）能够恢复在将大语言模型压缩到四比特以下时损失的大部分准确性。然而，现有的恢复阶段通常是在固定的完成或教师生成的答案上进行优化，而部署的量化模型则依赖于自身生成的前缀。因此，量化误差可能使模型进入离线恢复数据中不存在的状态。我们提出了OnlineQAT，一个两阶段框架，首先通过块级QAT获得可用的低比特初始化，然后对学生生成的响应进行在线策略蒸馏（OPD）。在每个访问的前缀上，一个冻结的全精度教师提供一个采样的反向KL训练信号。在Qwen3-1.7B上，OnlineQAT在比较的量化方法中获得了最佳平均值：在W3A16上为57.28，在W2A16上为32.52，分别比ReasoningQAT提高了2.90和0.44点。结果表明，学生访问的状态提供了超越固定完成训练的有用恢复信号，特别是在三比特的情况下。
cs.CL / 22 / 2610.09372
Expert Coupling in MoE Pretraining: Reducing All-to-All Overhead with Correlated Placement and Token Shuffling
MoE预训练中的专家耦合：通过相关放置和令牌洗牌减少全到全的开销
Abstract
Mixture-of-Experts (MoE) layers replace the feed-forward block of a Transformer with E expert networks, and each token is routed to k of these experts. Under expert parallelism (EP) the experts are distributed across GPUs, and every MoE layer runs all-to-all collectives in the forward and backward passes to dispatch tokens to their experts and then combine the results. On a cluster with 8 AMD Instinct MI300X GPUs per node, these collectives can take 45% of the training step at EP32 with top-2 routing and 60% with top-6 routing. We find that early in pretraining routers have already learned to assign tokens to experts in correlated patterns, both within a layer and across layers. At top-2, 0.8% of the expert pairs in a layer are selected together by 42% of tokens, and the experts a token selects at one layer predict the experts it selects at the next layer. We use these correlations to keep more token--expert assignments on the token's own GPU, which reduces communication across GPUs and across nodes. Correlated expert placement puts experts that are often selected together on the same GPU. Combined with a dispatcher that sends each token to each GPU once, it removes up to 58% of dispatched rows. Token shuffling applies when sequence parallelism shards tokens across the EP group. It moves each token to the GPU predicted to hold its next-layer experts during the reduce-scatter that follows attention. On one node this raises the share of token--expert assignments served on the token's GPU from 12.5% to 59%. In Megatron-LM, across EP degrees from 8 to 64 with top-2 and top-6 routing, the two methods reduce all-to-all time by 1.16-2.63X and end-to-end step time by up to 1.41X. Neither method changes the models' underlying routing decisions or expert parameters.
Chinese Translation
混合专家（Mixture-of-Experts, MoE）层用E个专家网络替代了Transformer的前馈模块，每个令牌被路由到k个专家。在专家并行（Expert Parallelism, EP）下，专家分布在多个GPU上，每个MoE层在前向和反向传播中执行全到全的集合操作，以将令牌分派给其专家并合并结果。在每个节点上配备8个AMD Instinct MI300X GPU的集群中，这些集合操作在使用top-2路由时占据了EP32训练步骤的45%，而在使用top-6路由时则占据了60%。我们发现，在预训练的早期，路由器已经学会以相关模式将令牌分配给专家，这种模式在同一层内和跨层之间都存在。在top-2情况下，0.8%的专家对在一层中被42%的令牌共同选择，而一个层中令牌选择的专家可以预测其在下一个层中选择的专家。我们利用这些相关性来保持更多令牌-专家分配在令牌自身的GPU上，从而减少GPU之间和节点之间的通信。相关专家放置将经常一起被选择的专家放置在同一GPU上。结合一个将每个令牌一次性发送到每个GPU的调度器，这种方法最多可以减少58%的分派行。当序列并行将令牌在EP组中分片时，令牌洗牌被应用。它在注意力后的reduce-scatter过程中，将每个令牌移动到预测将持有其下层专家的GPU上。在一个节点上，这将令牌-专家分配在令牌GPU上的比例从12.5%提高到59%。在Megatron-LM中，使用top-2和top-6路由，EP度数从8到64，这两种方法将全到全的时间减少了1.16-2.63倍，端到端步骤时间最多减少了1.41倍。这两种方法均未改变模型的基础路由决策或专家参数。
cs.CL / 23 / 2610.09384
The Persona Hierarchy Model: Understanding Contextual Generalization in Fine-Tuning LLMs
角色层次模型：理解大规模语言模型微调中的上下文泛化
Abstract
Language models are routinely fine-tuned under a fixed context, such as a generic system prompt, persona or domain-specific instruction, yet the learned behavior sometimes stays confined to that context and sometimes broadly generalizes to unseen contexts. We propose the Persona Hierarchy Model to explain this: a shared default persona influences behavior across contexts. Under this model, fine-tuning that modifies the shared persona promotes broader transfer, whereas changes to local personas remain more context-specific. Across 120 fine-tuned models spanning four behaviors and 15 training contexts, generalization narrowness positively correlates with the similarity between the training context's persona and the default persona (Pearson's r = 0.72 for Qwen3-4B). Prior fine-tuning under the default context can broaden generalization in subsequent training under other contexts. Aligning contextual responses with default-persona responses produces stronger effects. Finally, we propose persona-preserving regularization (PPR) to confine undesired contextual generalization. In RL, PPR cuts reward hacking from 42-55% to at most 0.2% under every evaluated prompt while retaining accuracy gains. These results support the Persona Hierarchy Model as an explanation for contextual generalization and can motivate future controls on unintended generalization for better alignment of LLMs.
Chinese Translation
语言模型通常在固定上下文下进行微调，例如通用系统提示、角色或特定领域的指令。然而，学习到的行为有时局限于该上下文，有时则能广泛泛化到未见过的上下文。我们提出了角色层次模型来解释这一现象：一个共享的默认角色影响着不同上下文中的行为。在该模型下，修改共享角色的微调促进了更广泛的迁移，而对局部角色的改变则更具上下文特异性。在涵盖四种行为和15个训练上下文的120个微调模型中，泛化的狭窄程度与训练上下文的角色与默认角色之间的相似性呈正相关（对于 Qwen3-4B，Pearson's r = 0.72）。在默认上下文下的先前微调可以在后续其他上下文的训练中拓宽泛化。将上下文响应与默认角色响应对齐会产生更强的效果。最后，我们提出了角色保持正则化（PPR）来限制不必要的上下文泛化。在强化学习中，PPR将奖励黑客行为从42-55%降低到最多0.2%，同时保持准确性提升。这些结果支持角色层次模型作为上下文泛化的解释，并可以激励未来对意外泛化的控制，以更好地对齐大规模语言模型。
cs.CL / 24 / 2610.09396
ARCS: Towards Precise Text-to-SQL via Structured Disambiguation
ARCS：通过结构化消歧迈向精确的文本到SQL转换
Abstract
As text-to-SQL systems move beyond demonstrations toward real-world deployment, ambiguity in user questions becomes a primary source of errors. Such ambiguities are often subtle, domain- or data-specific, and can silently cause system outputs to deviate from the user's true intent. Ambiguity is traditionally addressed through conversational clarification, which is often inefficient, cognitively demanding, and poorly aligned with real-world user workflows. We propose structured disambiguation, a new paradigm in which ambiguity is resolved through explicit, constrained interactions rather than free-form dialogue. We construct ARCS (Ambiguity Resolution Corpus for SQL), the first text-to-SQL benchmark featuring naturally occurring, unconstrained ambiguities over real-world databases, with complete annotations of all valid ambiguity points, interpretations, and SQL queries. Experimental results show that text-to-SQL remains challenging in the presence of ambiguity: gpt-6-sol achieves only 51% end-to-end execution accuracy, and no open-source model exceeds 27%.
Chinese Translation
随着文本到SQL系统从演示走向实际应用，用户问题中的歧义成为错误的主要来源。这种歧义往往微妙、特定于领域或数据，并可能悄然导致系统输出偏离用户的真实意图。传统上，歧义通过对话澄清来解决，但这种方式往往效率低下、认知负担重，并且与现实世界用户工作流程不够契合。我们提出了结构化消歧这一新范式，通过明确的、受限的交互来解决歧义，而非自由形式的对话。我们构建了ARCS（SQL歧义解决语料库），这是第一个文本到SQL基准，具有在真实世界数据库上自然发生的、无约束的歧义，并对所有有效的歧义点、解释和SQL查询进行了完整注释。实验结果表明，在存在歧义的情况下，文本到SQL仍然具有挑战性：gpt-6-sol的端到端执行准确率仅为51%，且没有开源模型超过27%。
cs.CL / 25 / 2610.09446
Arctic Questions, Missing Answers: A Dataset and Benchmark for LLM Abstention in Arctic Science
北极问题，缺失答案：一个关于北极科学中大型语言模型弃权的数据集和基准
Abstract
Large language models (LLMs) should abstain from scientific multiple-choice questions when no option is valid, but frequent abstention alone does not demonstrate sensitivity to answer availability. We introduce ArcticQA, a dataset of 194 questions derived from primary Arctic research, with automated checks of answer support and distractor contradiction against source evidence. We further develop ArcticAbstain, a paired benchmark comparing answer-present and answer-absent conditions, with the correct answer replaced by a distractor in the latter and an explicit abstention option in both. We evaluate eight models from the Gemini, Claude, and ChatGPT families at high reasoning effort, with three trials per condition, yielding 9,312 recorded responses. Answer-present abstention rates range from 0.0% to 63.0%, whereas replacing the correct answer increases abstention by 5.05 percentage points on average. These findings highlight substantial baseline differences and the need to evaluate abstention frequency and responsiveness jointly. The dataset and benchmark are available at https://github.com/BenWilcox8/arctic-qa.
Chinese Translation
大型语言模型（LLMs）在没有有效选项的科学多项选择题中应当选择弃权，但仅仅频繁弃权并不能证明对答案可用性的敏感性。我们引入了 ArcticQA，这是一个由194个问题组成的数据集，这些问题源自主要的北极研究，并对答案支持和干扰项矛盾进行了自动化检查，以验证其与源证据的一致性。我们进一步开发了 ArcticAbstain，这是一个配对基准，比较了有答案和无答案的条件，在后者中用干扰项替换了正确答案，并在两种情况下都提供了明确的弃权选项。我们对来自 Gemini、Claude 和 ChatGPT 系列的八个模型进行了高推理努力的评估，每种条件下进行了三次试验，共记录了9,312个响应。有答案的弃权率范围从0.0%到63.0%，而替换正确答案平均增加了5.05个百分点的弃权率。这些发现突显了显著的基线差异，以及共同评估弃权频率和响应性的必要性。数据集和基准可在 https://github.com/BenWilcox8/arctic-qa 获取。
cs.CL / 26 / 2610.09458
Right Number, Wrong State? Measuring Cross-Jurisdiction Substitution in LLM Recall of State Policy
正确的数字，错误的州？测量大型语言模型在州政策回忆中的跨管辖区替代
Abstract
When an LLM answers a state-specific policy question wrongly, it may be hallucinating, or it may be returning a real value that holds in another state. We test this with a minimal-set design: the question wording is fixed and only the jurisdiction varies, across the 50 U.S. states and the District of Columbia (51 jurisdictions) and three exactly defined Medicaid income-eligibility quantities. Gold values come from an official data book and agree with an independent source in 101 of 102 checked cells. Under a pre-registered protocol, Claude Sonnet 5.5 and GPT-5.6 Sol reproducibly give another state's current value, identical across two independent repeats, for 10 and 25 of 153 items. Attribution is fragile, however. Crediting any wrong answer that equals another state's value yields 3-5x more reproducible substitutions than checking every number in the asked state's own records, because many apparent cross-state answers are the asked state's own values under another convention or from an earlier year. Claims about cross-jurisdiction error need a complete same-state reference set. We will release the protocol, gold table, and all model outputs.
Chinese Translation
当大型语言模型（LLM）错误回答特定州的政策问题时，它可能是在幻觉，或者它可能返回的是在另一个州有效的真实值。我们通过最小集设计进行测试：问题措辞固定，仅管辖区在50个美国州和哥伦比亚特区（共51个管辖区）之间变化，以及三个精确定义的医疗补助收入资格数值。金标准值来自官方数据手册，并与独立来源在102个检查单元中的101个一致。在预注册协议下，Claude Sonnet 5.5和GPT-5.6 Sol在153个项目中分别对10个和25个项目重复给出了另一个州的当前值，并在两次独立重复中保持一致。然而，归因是脆弱的。将任何等于另一个州值的错误答案归功于该州，产生的可重复替代次数比检查所询问州自身记录中的每个数字多出3-5倍，因为许多明显的跨州答案实际上是所询问州在另一种惯例下或来自早期年份的自身值。关于跨管辖区错误的主张需要一个完整的同州参考集。我们将发布协议、金表和所有模型输出。
cs.CL / 27 / 2610.09464
BanglaRhet: Benchmarking Classical and Transformer Models for Rhetorical and Persuasion Detection in Bangla Political Speech
BanglaRhet：经典模型与变换器模型在孟加拉政治演讲中修辞与说服检测的基准研究
Abstract
Political discourse often uses rhetorical and persuasive language to frame narratives, influence public opinion, and mobilize audiences. While Bangla natural language processing has made progress in sentiment analysis and opinion mining, systematic benchmarking of transformer models for fine-grained rhetorical and persuasion technique detection in Bangla political speech remains largely underexplored. This paper presents a benchmark study of transformer-based models for detecting rhetorical form and persuasive intent in Bangla political discourse. Using BanglaRhet, a manually annotated corpus of 30,289 Bangla political speech segments collected from publicly available political news sources, we formulate two supervised single-label classification tasks: rhetorical technique detection (contrast, repetition, exaggeration, metaphor, rhetorical questions) and persuasion technique detection (blame assignment, call to action, unity call, moral, emotional, and logical appeals). We evaluate four transformer-based models, BanglaBERT, BanglaBERT-Base, SahajBERT, and XLM-RoBERTa-Base, against classical TF-IDF baselines. BanglaBERT achieves the highest performance, with 65.40% macro-F1 for rhetorical technique detection and 66.46% for persuasion technique detection, outperforming the best tuned classical baseline by 19.2 and 13.8 macro-F1 points, respectively. Class-level analysis indicates that errors are mainly associated with semantic overlap among labels, figurative language, and class imbalance. The results provide initial benchmark baselines for Bangla rhetorical and persuasion-aware political discourse analysis and highlight the need for context-aware and multi-label modeling.
Chinese Translation
政治话语常常使用修辞和说服性语言来构建叙事、影响公众舆论和动员听众。尽管孟加拉自然语言处理在情感分析和观点挖掘方面取得了一定进展，但在孟加拉政治演讲中对变换器模型进行细粒度修辞和说服技巧检测的系统性基准研究仍然较少被探索。本文提出了一项基于变换器模型的基准研究，旨在检测孟加拉政治话语中的修辞形式和说服意图。我们使用BanglaRhet，一个手动标注的语料库，包含来自公开政治新闻来源的30,289个孟加拉政治演讲片段，制定了两个监督单标签分类任务：修辞技巧检测（对比、重复、夸张、隐喻、修辞性问题）和说服技巧检测（归责、行动号召、团结号召、道德、情感和逻辑诉求）。我们评估了四种基于变换器的模型：BanglaBERT、BanglaBERT-Base、SahajBERT和XLM-RoBERTa-Base，并与经典的TF-IDF基线进行比较。BanglaBERT在修辞技巧检测中取得了最高性能，宏观F1值为65.40%，在说服技巧检测中为66.46%，分别比最佳调优的经典基线高出19.2和13.8个宏观F1点。类级分析表明，错误主要与标签之间的语义重叠、比喻语言和类别不平衡有关。结果为孟加拉修辞和说服意识的政治话语分析提供了初步的基准基线，并强调了对上下文敏感和多标签建模的需求。
cs.CL / 28 / 2610.09489
Goldsmith: Gold-Loss-Guided Definition Optimization with an Agentic Annotation Harness
Goldsmith：基于黄金损失引导的定义优化与自主标注框架
Abstract
Many annotation projects begin before experts have a stable guideline or enough labels to train a task-specific model. We present Goldsmith, an agentic pipeline that turns a small gold set---expert-annotated calibration examples representing the intended task boundaries---into a reusable structured annotation definition. Goldsmith treats this definition as a trainable textual object. Candidate definitions are run on the same gold examples and scored with an executable structured loss, while the output schema, formatting, retrieval, repair, judging, and human review remain in an external harness. A large language model (LLM) editor converts the highest-loss failures into textual-gradient revisions, which are accepted only when the measured loss decreases. In prompt-optimization comparisons, Goldsmith improves over direct rewriting, OPRO, APE, and PromptBreeder under matched evaluation protocols. The resulting definition also improves downstream annotation when combined with retrieval, score-based routing, and human review across typed span, pair-level relation, and fixed-trigger event-argument tasks. These results show that scarce expert supervision can support both task-definition learning and scalable annotation.
Chinese Translation
许多标注项目在专家尚未建立稳定的指导方针或没有足够的标签来训练特定任务模型之前就已开始。我们提出了Goldsmith，一个自主的流程，将一小部分黄金集——代表预期任务边界的专家标注校准示例——转化为可重用的结构化标注定义。Goldsmith将此定义视为一个可训练的文本对象。候选定义在相同的黄金示例上运行，并通过可执行的结构化损失进行评分，而输出模式、格式、检索、修复、评判和人工审核则保留在外部框架中。一个大型语言模型（LLM）编辑器将最高损失的失败转化为文本梯度修订，只有在测量损失减少时才会被接受。在提示优化比较中，Goldsmith在匹配评估协议下优于直接重写、OPRO、APE和PromptBreeder。所得到的定义在结合检索、基于评分的路由和人工审核后，进一步改善了下游标注，适用于类型跨度、对级关系和固定触发事件-参数任务。这些结果表明，稀缺的专家监督可以支持任务定义学习和可扩展标注。
cs.CL / 29 / 2610.09529
A Comparative Study of Evaluation Metrics for Long-Document Financial Narrative Summarization with Transformers
基于变换器的长文档金融叙述摘要评估指标的比较研究
Abstract
There are more than 2,000 listed companies on the UK's London Stock Exchange, divided into 11 sectors who are required to communicate their financial results at least twice in a single financial year. UK annual reports are very lengthy documents with around 80 pages on average. In this study, we aim to benchmark a variety of summarisation methods on a set of different pre-trained transformers with different extraction techniques. In addition, we considered multiple evaluation metrics in order to investigate their differing behaviour and applicability on a dataset from the Financial Narrative Summarisation (FNS 2020) shared task, which is composed of annual reports published by firms listed on the London Stock Exchange and their corresponding summaries. We hypothesise that some evaluation metrics do not reflect true summarisation ability and propose a novel BRUGEscore metric, as the harmonic mean of ROUGE-2 and BERTscore. Finally, we perform a statistical significance test on our results to verify whether they are statistically robust, alongside an adversarial analysis task with three different corruption methods.
Chinese Translation
英国伦敦证券交易所上市公司超过2000家，分为11个行业，要求在一个财年内至少沟通两次财务结果。英国年度报告是非常冗长的文件，平均约80页。在本研究中，我们旨在对一组不同的预训练变换器及其不同的提取技术进行多种摘要方法的基准测试。此外，我们考虑了多种评估指标，以研究它们在来自金融叙述摘要（Financial Narrative Summarisation, FNS 2020）共享任务的数据集上的不同表现和适用性，该数据集由在伦敦证券交易所上市的公司发布的年度报告及其相应摘要组成。我们假设某些评估指标并不能反映真实的摘要能力，并提出了一种新颖的BRUGEscore指标，作为ROUGE-2和BERTscore的调和平均值。最后，我们对结果进行统计显著性检验，以验证其统计稳健性，并进行了一项对抗性分析任务，采用三种不同的损坏方法。
cs.CL / 30 / 2610.09569
RELATE: An Evaluation Framework for measuring Relational Orientation of Large Language Models
RELATE：用于测量大型语言模型关系导向的评估框架
Abstract
Large language models (LLMs) are increasingly used for emotional support, raising concern that sustained use may draw users away from their real-world relationships. Yet existing evaluations primarily focus on the safety, empathy, or helpfulness of responses, leaving under-examined a relational question: where does the model orient the user for continued support? To address this question, we introduce relational orientation, a property operationalized through two non-exclusive dimensions: inward-facing (IF) language, which positions the AI as the user's ongoing source of support, and outward-scaffolding (OS) language, which encourages real-world human connection. Grounded in psychological and sociological literature, we formalize a taxonomy of relational orientation and present RELATE, a persona-conditioned framework for measuring inward-facing and outward-scaffolding language at the sentence level in multi-turn dialogues. RELATE pairs 76 help-seeking situations adapted from naturally occurring questions with three simulated user styles, providing 228 evaluation stimuli. In our experiments, we evaluate seven LLMs using dialogues with six assistant turns each, yielding 1,596 dialogues and 69,194 assistant sentences. We assess these sentences using a primary rubric-based LLM judge and apply a secondary judge to a subset. Under automated evaluation, we find that the proportion of sentences labeled as IF is higher at the sixth assistant turn than at the first, while the proportion labeled as OS is substantially lower for hesitant, indirect simulated users than for explicit, reassurance-seeking users. RELATE provides a reproducible framework and a sentence-level signal for auditing and steering the relational orientation of supportive LLMs.
Chinese Translation
大型语言模型（LLMs）越来越多地被用于情感支持，这引发了人们的担忧，即持续使用可能会使用户远离现实世界的人际关系。然而，现有的评估主要集中在响应的安全性、同理心或有用性上，未能充分探讨一个关系性问题：模型将用户导向何处以获得持续支持？为了解决这个问题，我们引入了关系导向这一属性，通过两个非排他性维度进行操作化：内向（IF）语言，将人工智能定位为用户持续的支持来源，以及外向支撑（OS）语言，鼓励现实世界的人际连接。基于心理学和社会学文献，我们正式化了关系导向的分类法，并提出了RELATE，一个基于角色的框架，用于在多轮对话中逐句测量内向和外向支撑语言。RELATE将76个从自然发生问题中改编的求助情境与三种模拟用户风格相结合，提供228个评估刺激。在我们的实验中，我们使用每个对话六个助手轮次的对话评估七个LLM，产生了1,596个对话和69,194个助手句子。我们使用一个基于主要评分标准的LLM评审对这些句子进行评估，并对一个子集应用次要评审。在自动评估下，我们发现第六个助手轮次标记为IF的句子比例高于第一个，而对于犹豫、间接的模拟用户，标记为OS的句子比例显著低于明确、寻求安慰的用户。RELATE提供了一个可重复的框架和逐句信号，用于审计和引导支持性LLM的关系导向。
cs.CL / 31 / 2610.09571
How Do LLMs Change Predictions Under Negation?
大型语言模型如何在否定下改变预测？
Abstract
Negation is an essential feature of human language, yet large language models (LLMs) remain unreliable in processing it. We evaluate recent open-source and closed-source LLMs on our negation benchmark and find that, in 37-71% of cases, they repeat the same answer under negation (e.g., "Madrid" for "What is not the capital of Spain?"). To understand and address this brittleness, we mechanistically examine how models operate under negation. Our main finding is that specialized attention heads and MLP neurons jointly implement negation by (1) suppressing retrieval of the original answer (e.g., "Madrid") while (2) promoting a favored candidate within the answer category (e.g., "Paris"). This contrasts with accounts of human negation processing, in which information about the original answer helps to determine what should be excluded. Furthermore, we find that this difference from human processing is a key source of negation failures: the model's mechanism relies on suppressing the original answer rather than using it to determine what to exclude, so the model can repeat the original answer when suppression is too weak or when a bias toward particular answers prevents it from selecting an alternative. To address this weakness in the model's negation mechanism, we propose a training objective that requires larger shifts in answer preference for more confident original predictions, and show that it reduces negation failures with less degradation of general capabilities than standard fine-tuning baselines. Together, our results demonstrate how mechanistic analysis can reveal why a linguistic capability fails and guide training that targets the underlying limitation.
Chinese Translation
否定是人类语言的一个重要特征，但大型语言模型（LLMs）在处理否定时仍然不可靠。我们在否定基准测试中评估了最近的开源和闭源LLMs，发现它们在37-71%的情况下在否定下重复相同的答案（例如，对于“西班牙的首都不是哪里？”的问题，它们回答“马德里”）。为了理解和解决这种脆弱性，我们机制性地考察了模型在否定下的运作方式。我们的主要发现是，专门的注意力头和多层感知器（MLP）神经元共同通过（1）抑制原始答案的检索（例如“马德里”），同时（2）促进答案类别内的一个优选候选（例如“巴黎”）来实现否定。这与人类否定处理的解释形成对比，在人类处理过程中，关于原始答案的信息有助于确定应排除的内容。此外，我们发现这种与人类处理的差异是否定失败的一个关键来源：模型的机制依赖于抑制原始答案，而不是利用它来确定应排除的内容，因此当抑制过弱或对特定答案的偏见阻止其选择替代答案时，模型可能会重复原始答案。为了应对模型否定机制中的这一弱点，我们提出了一种训练目标，要求对更自信的原始预测进行更大的答案偏好转变，并显示出它在减少否定失败的同时，较标准微调基线对一般能力的降级更小。总之，我们的结果展示了机制分析如何揭示语言能力失败的原因，并指导针对基础限制的训练。
cs.CL / 32 / 2610.09607
Which Language Should a Skeleton Speak? Language Choices in Multilingual Reasoning
骨架应使用哪种语言？多语言推理中的语言选择
Abstract
Skeleton-based reasoning prompting is a promising training-free approach for structuring LLM reasoning, but prior work largely assumes an English-centric setting. We propose the Language-Aware Skeleton Exploration Framework (LASEF) to study skeleton-language choice in multilingual mathematical reasoning. Across math benchmarks, model scales, and languages, we show that English skeletons yield a small positive tendency on average, most visible for smaller models and low-resource languages. However, few language-level gains remain significant after correction, and English is not universally optimal. Combining greedy decoding, multi-rollout evaluation, translation ablation, and cross-benchmark validation, we further find three patterns of skeleton-language effects: directionally consistent, evaluation- and benchmark-dependent, and asymmetric negative. These effects cannot be fully explained by generation quality alone. Overall, skeleton language is a context-dependent design variable that requires multi-level exploration. All resources are released at https://github.com/lhsstn/LASEF.
Chinese Translation
基于骨架的推理提示是一种有前景的无训练方法，用于构建大型语言模型（LLM）的推理，但之前的研究大多假设在以英语为中心的环境中进行。我们提出了语言感知骨架探索框架（Language-Aware Skeleton Exploration Framework, LASEF），以研究多语言数学推理中的骨架-语言选择。在数学基准测试、模型规模和语言之间的比较中，我们发现英语骨架在平均上产生了小幅正向倾向，这在较小的模型和低资源语言中最为明显。然而，经过修正后，只有少数语言层面的增益仍然显著，英语并非普遍最佳。通过结合贪婪解码、多次回合评估、翻译消融和跨基准验证，我们进一步发现了三种骨架-语言效应的模式：方向一致、依赖评估和基准、以及不对称负效应。这些效应不能仅通过生成质量来完全解释。总体而言，骨架语言是一个依赖于上下文的设计变量，需要多层次的探索。所有资源已发布在 https://github.com/lhsstn/LASEF。
cs.CL / 33 / 2610.09639
On-Policy Distillation Teaches New Skills but Not New Knowledge
在线策略蒸馏教授新技能但不传授新知识
Abstract
On-policy distillation (OPD) strengthens language-model reasoning, yet whether students acquire new factual knowledge or compositional skill for multi-step reasoning remains unknown. We separate these capabilities using a controlled synthetic framework that measures the student's initial capabilities and independently controls the teacher's additional facts, compositional skill, or both. Across four models from three families, reverse-KL OPD reliably transfers compositional skill across unseen reasoning structures, but transfers minimal factual knowledge. Decoupling the distillation recipe reveals the source of this asymmetry: replacing reverse KL with forward KL restores factual transfer, whereas student rollouts specifically improve the execution of multi-step reasoning. Experiments on recent factual QA and competition mathematics show a similar asymmetry under reverse-KL OPD, yielding notable reasoning gains without factual memory expansion. Together, these results demonstrate that on-policy distillation does not expand a model's parametric knowledge, but instead teaches it to organize and compose the knowledge it already possesses.
Chinese Translation
在线策略蒸馏（On-policy distillation, OPD）增强了语言模型的推理能力，但学生是否获得了新的事实知识或多步骤推理的组合技能仍然未知。我们使用一个受控的合成框架来区分这些能力，该框架测量学生的初始能力，并独立控制教师的额外事实、组合技能或两者。通过来自三个家族的四个模型，反向KL OPD可靠地在未见过的推理结构中转移组合技能，但对事实知识的转移非常有限。解耦蒸馏配方揭示了这种不对称的来源：用正向KL替换反向KL恢复了事实转移，而学生的执行过程特别改善了多步骤推理的执行。最近的事实问答和竞赛数学实验显示，在反向KL OPD下存在类似的不对称性，带来了显著的推理提升而没有事实记忆的扩展。综合来看，这些结果表明，在线策略蒸馏并没有扩展模型的参数知识，而是教会模型组织和组合其已有的知识。
cs.CL / 34 / 2610.09660
Rubric Spans are Label Representations: Joint LLM Encoding for Short Answer Scoring
评分标准跨度是标签表示：用于短答案评分的联合大语言模型编码
Abstract
Automatic Short Answer Scoring (ASAS) requires models that can score student responses against question-specific criteria while remaining efficient and transferable across rubric sets. We propose RUSPAN, a rubric-conditioned ASAS framework that treats rubric descriptions as semantic label representations. RUSPAN serialises the question context, student answer, and all candidate rubric levels into a single sequence, then scores the levels listwise from the rubric-span and whole-sequence representations produced in a single LM pass. We further introduce RUSPAN-RIM, in which a Rubric-Independent Mask prevents rubric spans from attending to one another, making rubric representations depend only on the answer and question context and preventing overfitting to rubric patterns during training for zero-shot transfer. On six ASAS benchmarks spanning English, German, and Portuguese, RUSPAN improves mono-benchmark scoring over discriminative and generative baselines, while RIM with position reindexing delivers consistent and substantial gains on PT-ASAG, the held-out benchmark with the strongest combined language and rubric-structure shift.
Chinese Translation
自动短答案评分（ASAS）需要能够根据特定问题标准对学生回答进行评分的模型，同时在不同评分标准集之间保持高效性和可迁移性。我们提出了RUSPAN，一个基于评分标准的ASAS框架，将评分标准描述视为语义标签表示。RUSPAN将问题上下文、学生答案和所有候选评分标准级别序列化为一个单一序列，然后从评分标准跨度和在单次语言模型传递中生成的整体序列表示中对级别进行列表评分。我们进一步引入了RUSPAN-RIM，其中一个与评分标准无关的掩码防止评分标准跨度相互关注，使评分标准表示仅依赖于答案和问题上下文，并在训练过程中防止对评分标准模式的过拟合，以实现零-shot迁移。在涵盖英语、德语和葡萄牙语的六个ASAS基准测试中，RUSPAN在判别和生成基线之上提高了单一基准评分，而带有位置重新索引的RIM在PT-ASAG（具有最强语言和评分标准结构转变的保留基准）上提供了一致且显著的提升。
cs.CL / 35 / 2610.09661
Alice: A Large-Scale German Benchmark for Rubric-Based Multi-Dimensional Automatic Short Answer Scoring
Alice：一个基于评分标准的大规模德语多维自动短答案评分基准
Abstract
Automatic Short Answer Scoring (ASAS) is central to NLP for Education. However, openly available benchmarks remain scarce, and existing datasets largely address how well students answer a question directly rather than how well they master underlying concepts (knowledge elements) such as thermal energy or epistemic activities (skills) such as reasoning or claim. To address this gap, we introduce Alice, a large-scale, rubric-based German ASAS dataset that is pedagogically aligned and comprises three subtasks: (i) learning performance (Alice-LP), (ii) knowledge elements (Alice-KE), and (iii) skills (Alice-SK). We further formulate rubric-based ASAS as a rubric-retrieval task and benchmark the dataset with a range of language models, from encoder-only models to lightweight LLMs. We also benchmark the dataset with zero-shot prompting via LLMs and a standard classification baseline. The experiments show that LLMs, in particular, struggle to score knowledge elements and skills in the zero-shot setting. They also indicate that rubric text is often useful, especially for Alice-KE and Alice-SK, while on Alice-LP gains over sample-solution-focused inputs are more modest and vary by model and input format.
Chinese Translation
自动短答案评分（ASAS）在教育领域的自然语言处理（NLP）中占据核心地位。然而，公开可用的基准仍然稀缺，现有数据集主要关注学生直接回答问题的能力，而非他们对基础概念（知识元素）如热能或认知活动（技能）如推理或主张的掌握程度。为了解决这一问题，我们推出了Alice，一个大规模的基于评分标准的德语ASAS数据集，该数据集在教学上具有一致性，并包含三个子任务：（i）学习表现（Alice-LP），（ii）知识元素（Alice-KE），以及（iii）技能（Alice-SK）。我们进一步将基于评分标准的ASAS形式化为一个评分标准检索任务，并使用一系列语言模型对数据集进行基准测试，从仅编码模型到轻量级大语言模型（LLMs）。我们还通过LLMs和标准分类基线对数据集进行零样本提示的基准测试。实验表明，LLMs，尤其是在零样本设置下，难以对知识元素和技能进行评分。结果还表明，评分标准文本通常是有用的，特别是在Alice-KE和Alice-SK上，而在Alice-LP上，相较于以样本解决方案为重点的输入，提升则更为温和，并且因模型和输入格式而异。
cs.CL / 36 / 2610.09665
SAPD: Step-Aligned Privileged Distillation
SAPD：步对齐特权蒸馏
Abstract
On-policy post-training can improve large language models by learning from their own trajectories, but requires costly rollout generation. We ask whether fixed demonstrations can support competitive off-policy learning through better supervision. Our premise is that their usefulness depends not only on the training trajectories, but also on whether supervision provides informative preferences among continuations and connects this guidance to the reasoning decision being learned. We introduce Step-Aligned Privileged Distillation (SAPD), a rollout-free self-distillation method that turns demonstrations into step-aligned distributional supervision. Its key insight is to use the known progression of a reference solution to associate each reasoning transition with targeted privileged guidance, rather than treating the solution as undifferentiated context. On mathematical reasoning benchmarks, SAPD outperforms supervised fine-tuning and label smoothing on average while remaining competitive with on-policy reinforcement learning and self-distillation. Analyses support both the value of context-dependent distributional guidance and the benefit of aligning privileged information with the current step. SAPD also largely preserves out-of-domain coding performance and achieves approximately 2x training-loop speedups over the on-policy baselines. These findings suggest that carefully constructed supervision can make fully off-policy post-training a competitive and computationally efficient alternative. Our code is available at https://github.com/Miaow-Lab/SAPD.
Chinese Translation
基于策略的后训练可以通过学习自身轨迹来改善大型语言模型，但需要昂贵的回滚生成。我们探讨固定演示是否能够通过更好的监督支持竞争性的离策略学习。我们的前提是，它们的有效性不仅依赖于训练轨迹，还依赖于监督是否提供了对后续步骤的有信息的偏好，并将这种指导与正在学习的推理决策联系起来。我们提出了步对齐特权蒸馏（SAPD），这是一种无回滚的自蒸馏方法，将演示转化为步对齐的分布式监督。其关键见解是利用参考解的已知进展，将每个推理过渡与目标特权指导关联，而不是将解视为未区分的上下文。在数学推理基准上，SAPD的表现平均优于监督微调和标签平滑，同时与基于策略的强化学习和自蒸馏保持竞争力。分析结果支持上下文相关的分布式指导的价值，以及将特权信息与当前步骤对齐的好处。SAPD在很大程度上保持了域外编码性能，并实现了约2倍的训练循环加速，相较于基于策略的基线。这些发现表明，精心构建的监督可以使完全离策略的后训练成为一种具有竞争力且计算高效的替代方案。我们的代码可在 https://github.com/Miaow-Lab/SAPD 获取。
cs.CL / 37 / 2610.09671
InsClaimBench: Benchmarking Insurance Claim Adjudication Across the Decision Chain
InsClaimBench：跨决策链的保险索赔裁定基准测试
Abstract
Recent advances in reasoning-oriented large language models (LLMs) have motivated increasing evaluation of their ability to perform professional decision tasks. Insurance claim adjudication is one such task, requiring models to connect case evidence, insurance rules, intermediate judgments, and payout calculations across a structured decision process. We introduce InsClaimBench, an end-to-end benchmark for evaluating insurance claim adjudication across the decision chain. Grounded in real claim materials and structured insurance rules, InsClaimBench contains 3,780 cases in 375 case families across auto, property, and health insurance, comprising 86,656 atomic rule judgments. It evaluates each claim from atomic rules through adjudication modules to payout decisions and amounts, with controlled factual variants testing whether required changes are correctly propagated across levels. Evaluation of six LLMs reveals a progressive loss of reliability along the decision chain. Payout-decision accuracy ranges from 74.23--80.19%, while joint decision--amount accuracy drops to 47.54--73.15%. Strong local performance also fails to ensure case-level correctness: atomic-rule accuracy reaches 95.48%, whereas rule-vector exact match peaks at only 36.90%, and the most frequent module errors are not necessarily those most associated with final-decision failure. Under factual changes, these inconsistencies further become propagation failures: module updates are less reliable than rule updates, correct local judgments can still yield incorrect payouts, and correct payouts can conceal intermediate errors. These results show that reliable claim adjudication requires consistent composition and propagation across the decision chain.
Chinese Translation
近期在面向推理的大型语言模型（LLMs）方面的进展，激励了对其执行专业决策任务能力的评估。保险索赔裁定就是其中一项任务，要求模型在结构化决策过程中连接案件证据、保险规则、中间判断和赔付计算。我们提出了InsClaimBench，这是一个用于评估跨决策链的保险索赔裁定的端到端基准测试。InsClaimBench基于真实的索赔材料和结构化的保险规则，包含了来自汽车、财产和健康保险的3,780个案例，分属于375个案例家庭，共计86,656个原子规则判断。它评估每个索赔，从原子规则到裁定模块，再到赔付决策和金额，并通过控制事实变体测试所需的变更是否在各个层级中正确传播。对六个LLM的评估显示，沿决策链的可靠性逐渐下降。赔付决策的准确率范围为74.23%到80.19%，而联合决策-金额的准确率下降至47.54%到73.15%。强大的局部性能并未确保案件级别的正确性：原子规则的准确率达到95.48%，而规则向量的精确匹配最高仅为36.90%，且最常见的模块错误并不一定与最终决策失败最相关。在事实变更下，这些不一致性进一步演变为传播失败：模块更新的可靠性低于规则更新，正确的局部判断仍可能导致错误的赔付，而正确的赔付可能掩盖中间错误。这些结果表明，可靠的索赔裁定需要在决策链中保持一致的组合和传播。
cs.CL / 38 / 2610.09684
From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discovery
从帕累托到偏好：通过摊销的自主政策发现实现个性化的测试时间缩放
Abstract
Test-time scaling (TTS) improves the reasoning capabilities of large language models by allocating additional inference computation. Existing approaches to improving TTS efficiency largely optimize accuracy against one resource dimension at a time, advancing either the accuracy--cost or accuracy--latency Pareto frontier. Yet user requirements are multidimensional: users may specify accuracy, latency, and inference-cost requirements jointly, and different requirements can favor different controllers. We formulate Personalized Test-Time Scaling as discovering executable controllers that maximize the joint satisfaction rate of user-specific requirements. To reduce the overhead of repeated policy discovery for new user profiles, we propose PersonTTS, an amortized agentic policy-discovery framework that reuses prior search experience through requirement-matched controller initialization and source-distilled procedural guidance, while retaining target-profile evaluation for every candidate. Experiments on AIME and HMMT show that PersonTTS substantially outperforms strong TTS baselines in joint requirement satisfaction on unseen user profiles and held-out problems. Under the same candidate-evaluation budget, cross-user experience reuse further improves policy quality while substantially reducing discovery-agent time and cost.
Chinese Translation
测试时间缩放（TTS）通过分配额外的推理计算来提高大型语言模型的推理能力。现有的提高TTS效率的方法主要是在一个资源维度上优化准确性，推动准确性-成本或准确性-延迟的帕累托前沿。然而，用户需求是多维的：用户可能同时指定准确性、延迟和推理成本的要求，而不同的要求可能更适合不同的控制器。我们将个性化测试时间缩放表述为发现可执行控制器，以最大化用户特定需求的联合满足率。为了减少对新用户配置进行重复政策发现的开销，我们提出了PersonTTS，这是一种摊销的自主政策发现框架，通过需求匹配的控制器初始化和源蒸馏的程序指导重用先前的搜索经验，同时对每个候选者保留目标配置的评估。在AIME和HMMT上的实验表明，PersonTTS在未见用户配置和保留问题上的联合需求满足方面显著优于强大的TTS基线。在相同的候选者评估预算下，跨用户经验重用进一步提高了政策质量，同时显著减少了发现代理的时间和成本。
cs.CL / 39 / 2610.09710
SpikingVLA: Asynchronous Spiking Vision-Language-Action Models
SpikingVLA：异步脉冲视觉-语言-动作模型
Abstract
ANN-to-SNN conversion offers a practical route toward energy-efficient spiking Vision-Language-Action (VLA) models by bypassing the substantial cost of training large-scale SNNs from scratch. However, existing methods often require many timesteps to maintain competitive performance, resulting in substantial inference latency for real-time VLA deployment. To address this challenge, we introduce SpikingVLA, an ANN-to-SNN conversion framework that enables accurate and low-latency spiking VLA inference. Specifically, we propose a Dendritic Integrate-and-Fire (DIF) neuron that alleviates channel-wise activation outliers through dendritic mixing and adaptive somatic firing, enabling accurate ANN-to-SNN conversion with fewer timesteps. Building on DIF neurons, we further introduce an asynchronous execution mechanism that overlaps temporal computation across VLA components, reducing synchronization overhead and latency. Extensive experiments demonstrate that SpikingVLA achieves competitive navigation performance with substantially improved inference efficiency. Compared with existing spiking VLA methods, SpikingVLA improves SR and SPL by 11.9\% and 12.6\%, respectively, while reducing first-action latency by 11.2$\times$. These results establish SpikingVLA as a practical framework for deploying pretrained VLA models with high-performance and low-latency spiking inference.
Chinese Translation
人工神经网络（ANN）到脉冲神经网络（SNN）的转换为实现节能的脉冲视觉-语言-动作（VLA）模型提供了一条实用路径，避免了从头训练大规模 SNN 的高昂成本。然而，现有方法通常需要多个时间步以维持竞争性能，导致实时 VLA 部署时显著的推理延迟。为了解决这一挑战，我们提出了 SpikingVLA，一个 ANN 到 SNN 的转换框架，能够实现准确且低延迟的脉冲 VLA 推理。具体而言，我们提出了一种树突整合与发放（Dendritic Integrate-and-Fire, DIF）神经元，通过树突混合和自适应体细胞发放来缓解通道激活异常，从而实现更少时间步的准确 ANN 到 SNN 转换。在 DIF 神经元的基础上，我们进一步引入了一种异步执行机制，重叠 VLA 组件之间的时间计算，减少同步开销和延迟。大量实验表明，SpikingVLA 实现了竞争性的导航性能，并显著提高了推理效率。与现有的脉冲 VLA 方法相比，SpikingVLA 分别提高了成功率（SR）和成功路径长度（SPL）11.9 ext{%} 和 12.6 ext{%}，同时将首次行动延迟减少了 11.2$ imes$。这些结果确立了 SpikingVLA 作为部署预训练 VLA 模型的实用框架，具有高性能和低延迟的脉冲推理能力。
cs.CL / 40 / 2610.09733
Bridge Routing Heads: Where Multilingual Multi-hop Reasoning Lives in LLMs
桥接路由头：多语言多跳推理在大型语言模型中的存在
Abstract
Multilingual LLMs answer the same multi-hop reasoning question across languages, but we lack a mechanistic account of whether they share an internal circuit. We identify Bridge Routing Heads (BRH) in two large multilingual LLMs through a three-stage pipeline. The resulting language-specific head sets exhibit near-complete mutual exclusivity across the five languages, with a mean Jaccard similarity of only 0.017 for Llama 3.1 70B and 0.057 for Qwen 2.5 72B, revealing language-idiosyncratic circuits. Ablating general BRH increases two-hop Negative Log-Likelihood (NLL) by 39-89x the random-head baseline, providing direct causal evidence of their role. Amplifying these heads in a failing target-language pass rescues up to 51.7% of cross-lingual failures, with no training. The two models share this dual-circuit pattern but allocate heads differently: Llama concentrates chaining in a large general pool, while Qwen leans on larger language-specific pools. Together these results show that activation-level intervention alone can recover correct answers from cross-lingual reasoning failures.
Chinese Translation
多语言大型语言模型（LLMs）能够在不同语言中回答相同的多跳推理问题，但我们缺乏对它们是否共享内部电路的机制性解释。我们通过一个三阶段的流程在两个大型多语言LLMs中识别出桥接路由头（Bridge Routing Heads, BRH）。结果显示，语言特定的头集在五种语言之间几乎完全互斥，Llama 3.1 70B的平均Jaccard相似度仅为0.017，而Qwen 2.5 72B为0.057，揭示了语言特有的电路。消融一般BRH使得两跳负对数似然（Negative Log-Likelihood, NLL）比随机头基线增加了39-89倍，提供了它们作用的直接因果证据。在失败的目标语言传递中增强这些头可以挽救高达51.7%的跨语言失败，且无需训练。这两个模型共享这种双电路模式，但头的分配方式不同：Llama将链式推理集中在一个大型通用池中，而Qwen则依赖于更大的语言特定池。综合这些结果表明，仅通过激活水平的干预就能从跨语言推理失败中恢复正确答案。
cs.CL / 41 / 2610.09756
Shaer: Controlled Arabic Poetry Generation with Meter Subform and Semantic Conditioning
Shaer：基于韵律子形式和语义条件的阿拉伯古典诗歌生成控制
Abstract
Classical Arabic poetry generation requires simultaneously satisfying semantic, linguistic, and fine-grained prosodic constraints. Existing systems typically control broad poetic attributes but do not jointly model semantic intent, meter subform, and poem length. We present Shaer, a controllable Classical Arabic poetry generation framework jointly conditioned on natural-language descriptions, meter subforms, and target hemistich counts. To support this task, we construct an enriched corpus of 116,032 classical Arabic poems derived from Ashaar, containing normalized meter-subform labels and automatically generated, validated semantic descriptions. We then adapt Yehia-7B using QLoRA-based supervised fine-tuning with a completion-only objective. Our evaluation combines automatic assessment of base-meter conformity, requested-subform adherence, and length control with three LLM judges, blinded human evaluation, and memorization analysis. Shaer achieves 95.17% base-meter accuracy, 91.75% poem-level meter-subform accuracy, and 83.40% exact count accuracy. Relative to its untuned foundation model, these results represent gains of 68.68, 57.77, and 38.93 percentage points, respectively; Shaer also attains the highest base-meter accuracy among all evaluated systems. Multi-LLM evaluation and a blinded human assessment of top-ranked outputs further indicate competitive semantic and literary quality. Finally, analysis of all 3,481 test generations finds no exact copies from the training corpus or paired source poems. Code, models, and datasets are publicly available.
Chinese Translation
古典阿拉伯诗歌生成需要同时满足语义、语言和细粒度韵律约束。现有系统通常控制广泛的诗歌属性，但未能共同建模语义意图、韵律子形式和诗歌长度。我们提出了Shaer，一个可控的古典阿拉伯诗歌生成框架，联合条件于自然语言描述、韵律子形式和目标半行计数。为了支持这一任务，我们构建了一个丰富的语料库，包含116,032首源自Ashaar的古典阿拉伯诗歌，包含标准化的韵律子形式标签和自动生成、验证的语义描述。然后，我们使用基于QLoRA的监督微调方法对Yehia-7B进行适配，目标为仅完成生成。我们的评估结合了对基础韵律符合性、请求的子形式遵循和长度控制的自动评估，以及三位大型语言模型（LLM）评审、盲人评估和记忆分析。Shaer在基础韵律准确性上达到了95.17%，在诗歌级别韵律子形式准确性上达到了91.75%，在确切计数准确性上达到了83.40%。相较于其未调优的基础模型，这些结果分别代表了68.68、57.77和38.93个百分点的提升；Shaer在所有评估系统中还达到了最高的基础韵律准确性。多LLM评估和对排名最高输出的盲人评估进一步表明其在语义和文学质量上的竞争力。最后，对所有3,481个测试生成的分析发现，没有来自训练语料库或配对源诗的确切复制品。代码、模型和数据集均已公开。
cs.CL / 42 / 2610.09772
Decoupling Logic from Persona: Structural Immunity of Edge LLM Agents to Context Pollution
将逻辑与角色分离：边缘大语言模型代理对上下文污染的结构免疫性
Abstract
Small language-model agents on edge devices must hold a persona and reason correctly at once, inside one context window that fills with conversational history and persona instructions. We study what happens to the logical part of such an agent when that history is long, misleading and persona-heavy (persona-logic interference), and present a Decoupling Architecture (AO-DA) that separates logical inference ("What") from persona expression ("How") into two inference paths on one INT4 base model with hot-swappable LoRA adapters. The logic path receives only the core turn and emits a verifiable structured state (Micro-State); the persona path renders it in character with the full history. In same-base-model ablations on an Apple M2 laptop (Llama-3.1-8B-Instruct and Gemma-3-4B-it, 4-bit; 480 runs over 4 pollution levels x 3 arms x 2 tasks x 2 personas x 5 seeds) we find: (i) the decoupled logic path is structurally invariant to pollution: its prompt stays at 180 (Llama) or 167 (Gemma) tokens while the mixed single-pass prompt grows from 242 to 1,203, and its outputs are byte-identical across levels (40/40); (ii) the mixed single pass degrades monotonically (composite logic score 0.669 to 0.150 on Llama, 0.487 to 0.150 on Gemma), mostly by failing to emit the required structured output (80-95% of runs on Llama, 100% on Gemma at the two highest levels); (iii) with the same pollution fed into the decoupled logic path, the dedicated-adapter, dedicated-format path is still more robust than the single pass on the 8B model (failure 0-20% vs 80-95%; paired $\Delta$ +0.30 to +0.50, Cliff's $\delta$ 0.50-0.85, Holm-adjusted $p \le 0.03$) but not on the 4B model, where both collapse. Separation costs one extra decode on a topic's first turn (28.2 s vs 18.2 s on Llama) and buys persona hot-swapping in 1.7 ms without re-running the logic path. Code, rubric, fixtures, adapters and logs are released.
Chinese Translation
边缘设备上的小型语言模型代理必须在一个充满对话历史和角色指令的上下文窗口内同时保持角色并进行正确推理。我们研究了当历史较长、具有误导性且角色信息丰富时（角色逻辑干扰），此类代理的逻辑部分会发生什么，并提出了一种分离架构（AO-DA），将逻辑推理（“什么”）与角色表达（“如何”）分离为一个INT4基础模型上的两个推理路径，并配备可热插拔的LoRA适配器。逻辑路径仅接收核心轮次并输出可验证的结构状态（微状态）；角色路径则结合完整历史进行角色化呈现。在Apple M2笔记本上的同基模型消融实验（Llama-3.1-8B-Instruct和Gemma-3-4B-it，4位；480次实验，涵盖4个污染级别x 3个臂x 2个任务x 2个角色x 5个种子）中，我们发现：（i）分离的逻辑路径对污染具有结构不变性：其提示保持在180（Llama）或167（Gemma）个标记，而混合单通道提示从242增长到1,203，其输出在各级别之间是字节相同的（40/40）；（ii）混合单通道的性能单调下降（Llama的复合逻辑得分从0.669降至0.150，Gemma从0.487降至0.150），主要是未能输出所需的结构化结果（在Llama上80-95%的实验，Gemma在两个最高级别上100%）；（iii）在将相同污染输入到分离的逻辑路径时，专用适配器、专用格式路径仍然比8B模型上的单通道更具鲁棒性（失败率0-20%对比80-95%；配对$ ext{Δ}$ +0.30至+0.50，Cliff's $ ext{δ}$ 0.50-0.85，Holm调整后的$p ext{≤} 0.03$），但在4B模型上则不然，两个路径均崩溃。分离在主题的第一轮上增加了一个额外的解码时间（Llama上28.2秒对比18.2秒），并在不重新运行逻辑路径的情况下实现了1.7毫秒的角色热插拔。代码、评分标准、装置、适配器和日志已发布。
cs.CL / 43 / 2610.09788
Judging in Latent Space: Efficient Generative Reward Modeling via Semantics-Preserving Compression
在潜在空间中的判断：通过保持语义的压缩实现高效的生成奖励建模
Abstract
Reward modeling often requires jointly representing and reasoning over multiple evaluation criteria, yet verbalizing this process token by token can incur substantial inference cost. Recent work on latent reasoning suggests that continuous states may support this computation more compactly. We introduce LatentGRM, a latent evaluation framework built on semantic chunking, compression, and reconstruction. By using the structure of rubric-guided evaluations to guide compression, LatentGRM learns compact continuous trajectories that support autonomous pairwise judgments without generating textual assessments. A separate interpreter reconstructs evaluation text from these trajectories, providing an offline view of the information retained under compression. Under matched training data and backbones, LatentGRM achieves competitive aggregate preference accuracy relative to explicit Supervised Fine-Tuning (SFT) judges at both 4B and 8B scales. Across four benchmark domains, LatentGRM-8B compresses evaluation trajectories by 8.9--9.2x and reduces total judge inference time by 6.1--7.0x at vote@5. Controlled rubric interventions show that criterion-dependent preference information is carried through the latent sequence. Together, these results demonstrate that continuous latent evaluation can substantially reduce inference cost while preserving competitive judgment quality.
Chinese Translation
奖励建模通常需要共同表示和推理多个评估标准，但逐字表达这一过程可能会产生巨大的推理成本。最近关于潜在推理的研究表明，连续状态可能更紧凑地支持这一计算。我们提出了LatentGRM，一个基于语义分块、压缩和重建的潜在评估框架。通过利用评分标准引导评估的结构来指导压缩，LatentGRM学习到紧凑的连续轨迹，支持自主的成对判断，而无需生成文本评估。一个独立的解释器从这些轨迹中重建评估文本，提供了在压缩下保留信息的离线视图。在匹配的训练数据和基础模型下，LatentGRM在4B和8B规模上相较于显式的监督微调（Supervised Fine-Tuning, SFT）评审者实现了竞争性的整体偏好准确性。在四个基准领域中，LatentGRM-8B将评估轨迹压缩了8.9--9.2倍，并在vote@5时将总评审推理时间减少了6.1--7.0倍。控制评分标准的干预显示，依赖标准的偏好信息通过潜在序列得以保留。这些结果共同表明，连续的潜在评估可以显著降低推理成本，同时保持竞争性的判断质量。
cs.CL / 44 / 2610.09795
MIRROR: From Imitation to Internalization in LLM Personalization
MIRROR：从模仿到内化的LLM个性化
Abstract
The demand for personalized LLMs is shifting from style imitation toward content quality. We investigate whether self-distillation can bridge this gap in existing fine-tuning paradigm. To address this limitation, we introduce MIRROR(Meta- personalization by Internalizing Reference-Revealed On-policy Reflections), a novel self-distillation framework that shifts LLM personalization from imitation toward preference internalization. First, we replace reference-token imitation with reference-revealed on-policy self-distillation, aligning the model's next-token distributions along its own generation trajectories with those of its reference-conditioned self, thereby internalizing user preferences rather than reproducing reference wording.Second, we introduce MIRROR-F, a focal plug-in that augments on-policy distributional alignment with selective supervision over informative reference tokens, thereby strengthening content generation while preserving user-specific expression. Across three personalized generation benchmarks, two model scales, and complementary reference-based and LLM-based evaluations, MIRROR and MIRROR-F achieve leading overall personalization performance and superior text quality, while exhibiting less catastrophic forgetting than SFT-based baselines on three unseen personalized generation tasks. The gains are consistent across model scales and application scenarios, translating to improved performance in LLM personalization tasks.
Chinese Translation
个性化大型语言模型（LLMs）的需求正从风格模仿转向内容质量。我们研究自我蒸馏是否能够填补现有微调范式中的这一空白。为了解决这一局限性，我们提出了MIRROR（通过内化参考揭示的在线反思进行元个性化），这是一种新颖的自我蒸馏框架，将LLM个性化从模仿转向偏好内化。首先，我们用参考揭示的在线自我蒸馏替代参考标记的模仿，将模型的下一个标记分布与其参考条件自我生成轨迹对齐，从而内化用户偏好，而不是简单复制参考措辞。其次，我们引入了MIRROR-F，这是一种聚焦插件，通过对信息丰富的参考标记进行选择性监督，增强了在线分布对齐，从而在保持用户特定表达的同时加强内容生成。在三个个性化生成基准、两种模型规模以及互补的基于参考和基于LLM的评估中，MIRROR和MIRROR-F在整体个性化性能和文本质量方面表现出色，同时在三个未见的个性化生成任务中表现出比基于SFT的基线更少的灾难性遗忘。这些提升在不同模型规模和应用场景中是一致的，转化为LLM个性化任务的性能改善。
cs.CL / 45 / 2610.09835
A Deafening Silence: Catastrophic Forgetting Lives in the Output Embeddings of Tokens the Data Never Speaks
震耳欲聋的沉默：灾难性遗忘存在于数据从未提及的标记输出嵌入中
Abstract
Continual pre-training and fine-tuning in Large Language Models (LLMs) inevitably induce catastrophic forgetting, typically mitigated by replay using often-inaccessible original data. In this data-free regime, we analyze where forgetting occurs and why. Systematic parameter freezing across five settings up to 1.4B reveals that forgetting concentrates selectively in the output embeddings of tokens rarely seen in the new corpus, whereas the same sqrt(v-hat) band of the body is inert and new learning resides elsewhere. This localization is governed by the vocabulary deficiency of the corpus rather than the training mode, allowing pre-retraining risk ranking from token counts alone within a fixed base model. Mechanistically, absent tokens receive persistent one-sided softmax gradients that Adam's second-moment (sqrt(v-hat)) normalization amplifies into full-sized updates. We therefore propose an intervention: raising Adam's epsilon exclusively for the output projection during training. Across eight settings spanning 160M to 12B parameters and four model families, this removes 39.4% to 67.9% of forgetting across all seven stable configurations without degrading target learning or requiring per-model tuning. The defense combines additively or better with replay (79.8% on Qwen/Korean) and rescues released-head LoRA from a 23-fold forgetting surge. Because post-hoc editing of the drifted rows recovers under 5% of forgetting, the intervention must operate during training. Our findings indicate that a single-line optimizer adjustment may serve as the primary defense against catastrophic forgetting where the corpus starves the vocabulary.
Chinese Translation
在大型语言模型（LLMs）中，持续的预训练和微调不可避免地会导致灾难性遗忘，通常通过重放使用往往无法获取的原始数据来缓解。在这种无数据的情况下，我们分析了遗忘发生的地点及其原因。对五种设置（参数量高达14亿）进行系统的参数冻结，揭示了遗忘主要集中在新语料库中很少见的标记的输出嵌入中，而同一平方根(v-hat)带的主体则保持不变，新学习则发生在其他地方。这种局部化受语料库的词汇不足的支配，而非训练模式，使得在固定基础模型中仅凭标记计数就能进行预训练风险排序。从机制上讲，缺失的标记接收持续的单侧softmax梯度，而Adam的二阶矩（sqrt(v-hat)）归一化将其放大为全尺寸更新。因此，我们提出了一种干预措施：在训练过程中仅提高Adam的epsilon以用于输出投影。在跨越160M到120B参数和四个模型家族的八种设置中，这种方法消除了39.4%到67.9%的遗忘，且在所有七种稳定配置中没有降低目标学习或需要对每个模型进行调优。该防御措施与重放（在Qwen/Korean上为79.8%）相加或更好，并使释放头LoRA免于23倍的遗忘激增。由于对漂移行的事后编辑恢复了不到5%的遗忘，因此该干预必须在训练期间进行。我们的发现表明，单行优化器调整可能成为应对语料库使词汇匮乏时灾难性遗忘的主要防御措施。
cs.CL / 46 / 2610.09858
Training Advisors for LLM Agents from Task Outcomes
基于任务结果训练大型语言模型代理的顾问
Abstract
Large language model agents tackle multi-step tasks by interleaving reasoning and tool calls with observations from the environment. Prior work has shown that natural-language feedback can help these agents revise their decisions during task execution. We introduce Caddie, a method for training critics to provide natural-language analysis and advice as agents work through a task. Unlike approaches that rely on step-level labels or reference critiques, Caddie learns from whether the agent ultimately succeeds after receiving the critic's feedback. We optimize the critic through reinforcement learning while keeping the base model frozen. Trained on multi-hop question answering with a single base model, our Qwen3-4B critic improves success rates across four base models of different scales and architectures, including three not used during critic training. On the MuSiQue benchmark, the trained critic improves Qwen3-4B's success rate by more than 25 percentage points, surpassing the performance of Kimi K3 without a critic. The same critic also yields gains on out-of-domain interactive benchmarks, including $\tau^3$ and DeepDive, with no additional training. Our results show that agents can decide when to seek help from a critic at inference time and that outcome-based critic training can produce guidance that transfers across base models and task domains.
Chinese Translation
大型语言模型代理通过将推理与工具调用与环境观察交替进行来处理多步骤任务。先前的研究表明，自然语言反馈可以帮助这些代理在任务执行过程中修正其决策。我们提出了Caddie，一种训练评论员在代理执行任务时提供自然语言分析和建议的方法。与依赖于步骤级标签或参考评论的方法不同，Caddie通过代理在收到评论员反馈后是否最终成功来学习。我们通过强化学习优化评论员，同时保持基础模型不变。在使用单一基础模型进行多跳问答的训练中，我们的Qwen3-4B评论员提高了四种不同规模和架构的基础模型的成功率，包括三种在评论员训练期间未使用的模型。在MuSiQue基准测试中，训练后的评论员使Qwen3-4B的成功率提高了超过25个百分点，超越了没有评论员的Kimi K3的表现。同一评论员在无额外训练的情况下也在域外交互基准测试中取得了提升，包括$ au^3$和DeepDive。我们的结果表明，代理可以在推理时决定何时寻求评论员的帮助，并且基于结果的评论员训练可以产生跨基础模型和任务领域的指导。
cs.CL / 47 / 2610.09934
Itgan at NADI 2026 shared task: Parameter-Efficient Whisper Adaptation for Robust, Mixed-Dialect and Code-Switched Arabic ASR
Itgan在NADI 2026共享任务中的表现：针对稳健的混合方言和代码切换阿拉伯语自动语音识别的参数高效Whisper适应
Abstract
We describe the Itgan systems for the three ASR subtasks of NADI 2026, namely robust country-level ASR (1.1), mixed-dialect ASR (1.2), and Tunisian code-switched ASR (1.3). All three share one recipe, Whisper adapted with LoRA on consumer GPUs, and each was carried by a different addition to it. On 1.1, where the dialect label is given at test time, per-dialect specialists continued from a pooled adapter gave the largest gain, and the submitted system reached 57.1% country-average WER. A post-evaluation linear probe on frozen encoder features routes utterances without the label and recovers 44% of what oracle routing gives. On 1.2 the choice of base model mattered more than adapter capacity, and system combination helped only once we added a decorrelated member, reaching 46.7% WER. On 1.3 our system placed second at 14.49% WER with the lowest CER among the leading submissions, 5.38%. Its last 0.60 WER points came without further training, mostly from an exact weight-space average of independently trained runs, with ROVER voting adding the remainder. Every comparison carries a paired-bootstrap test, and we report eight directions that did not work.
Chinese Translation
我们描述了Itgan系统在NADI 2026的三个自动语音识别（ASR）子任务中的表现，即稳健的国家级ASR（1.1）、混合方言ASR（1.2）和突尼斯代码切换ASR（1.3）。这三者共享一个方案，即在消费级GPU上使用LoRA适应的Whisper，每个子任务在此基础上进行了不同的扩展。在1.1中，当方言标签在测试时给出时，从池化适配器继续的每个方言专家取得了最大的增益，提交的系统达到了57.1%的国家平均字错误率（WER）。对冻结编码器特征的后评估线性探测器能够在没有标签的情况下路由话语，并恢复了44%的理想路由结果。在1.2中，基础模型的选择比适配器的容量更为重要，系统组合仅在我们添加了一个去相关的成员后才有所帮助，达到了46.7%的WER。在1.3中，我们的系统以14.49%的WER获得第二名，并在领先的提交中拥有最低的字符错误率（CER），为5.38%。其最后的0.60 WER点是在没有进一步训练的情况下获得的，主要来自独立训练运行的精确权重空间平均，ROVER投票则补充了其余部分。每个比较都进行了配对自助法测试，我们报告了八个未能奏效的方向。
cs.CL / 48 / 2610.09976
EASE: Entropy-Adaptive Distribution Shaping for Evading AI-generated Text Detectors
EASE：用于规避AI生成文本检测器的熵自适应分布塑形
Abstract
AI-generated text (AIGT) detection can be sensitive to the decoding choices of the source large language model (LLM). We observe that perturbing next-token logits or adjusting sampling temperature can reduce detection performance, providing a clear signal of detector vulnerability to decoding-time distribution changes. Building on this observation, we propose EASE (Entropy-Adaptive Distribution Shaping for Evasion), a training-free and detector-agnostic framework for evading AIGT detectors. EASE computes predictive entropy directly from the source LLM's next-token distribution and uses it to adapt both logit perturbation and sampling temperature, without detector feedback or model fine-tuning. Experiments across three source LLMs and multiple detectors demonstrate consistent reductions in detection performance, with negligible degradation in text quality and negligible inference overhead.
Chinese Translation
AI生成文本（AIGT）检测对源大型语言模型（LLM）的解码选择可能非常敏感。我们观察到，扰动下一个标记的logits或调整采样温度可以降低检测性能，提供了检测器对解码时分布变化的脆弱性的明确信号。在此观察的基础上，我们提出了EASE（熵自适应分布塑形用于规避），这是一个无训练且不依赖于检测器的框架，用于规避AIGT检测器。EASE直接从源LLM的下一个标记分布计算预测熵，并利用该熵来调整logit扰动和采样温度，而无需检测器反馈或模型微调。在三个源LLM和多个检测器上的实验表明，检测性能一致降低，同时文本质量几乎没有下降，推理开销也微乎其微。
cs.CL / 49 / 2610.10049
The Long Road to the Same Answer: Cognitive Bias Under Escalating Reasoning Budgets in Large Language Models
通往相同答案的漫长道路：大型语言模型中逐步推理预算下的认知偏差
Abstract
Reasoning models allocate extra computation at inference time and present their answers as the product of deliberate thought. If this deliberation works the way dual-process accounts of human cognition suggest, longer thinking should weaken the classic decision biases that fast, intuitive judgment produces. Using 30 vignettes covering six biases (anchoring, framing, loss aversion, escalation of commitment, availability, confirmation) from an established benchmark, we run a dose-response study across four model families, pairing each reasoning model with a matched non-reasoning sibling and requesting thinking ceilings of 0, 1,024, 4,096, and 8,192 tokens, for 12,350 API calls. Because a requested ceiling is not the same as realized deliberation, we use the reasoning tokens each call consumed as the dose. First, reasoning models are not less biased than their siblings; the point estimate leans the other way in every family, but the item-level pooled contrast is not reliable (Delta = +0.031, t(29) = 1.45, p = .157). Second, bias magnitude does not reliably fall as realized deliberation grows: no slope is significantly negative, and where anything moves it is the signed score drifting further from the human direction. Third, anchoring is the only bias in the human direction (d = 1.89). Four of the other five lean the opposite way in all seven models; with five items per bias, that reversal is reliable for framing and directional for escalation of commitment, confirmation, and loss aversion, while availability is absent. A one-line instruction to restate the anchor before answering lowered anchoring on all five anchoring items, which no amount of additional thinking did, although the effect does not reach significance (p = .057). The results argue against treating test-time reasoning as a rationality guarantee and for auditing deployed models bias by bias.
Chinese Translation
推理模型在推理时分配额外的计算资源，并将其答案呈现为深思熟虑的结果。如果这种深思熟虑的过程按照双过程理论的人类认知模型运作，那么更长的思考时间应该会削弱快速直觉判断所产生的经典决策偏差。我们使用涵盖六种偏差（锚定、框架效应、损失厌恶、承诺升级、可得性、确认偏误）的30个情境，进行了一项跨四个模型家族的剂量反应研究，将每个推理模型与一个匹配的非推理模型配对，并请求思考上限为0、1,024、4,096和8,192个标记，共进行了12,350次API调用。由于请求的上限与实际的深思熟虑并不相同，我们使用每次调用消耗的推理标记作为剂量。首先，推理模型并不比其兄弟模型更少偏见；在每个模型家族中，点估计都倾向于相反方向，但项目级的汇总对比并不可靠（Delta = +0.031, t(29) = 1.45, p = .157）。其次，偏差的幅度并不会随着实际的深思熟虑增加而可靠下降：没有斜率显著为负，而在任何变化中，签名分数都进一步偏离人类方向。第三，锚定是唯一一个朝向人类方向的偏差（d = 1.89）。其他五种偏差在所有七个模型中都倾向于相反方向；每种偏差有五个项目，这种反转在框架效应上是可靠的，并且在承诺升级、确认偏误和损失厌恶上是方向性的，而可得性则缺失。在回答之前重新表述锚定的单行指令降低了所有五个锚定项目的锚定效应，而额外的思考并未产生任何效果，尽管该效应未达到显著性水平（p = .057）。结果反驳了将测试时的推理视为理性保证的观点，并支持逐项审核已部署模型的偏差。
cs.CL / 50 / 2610.10058
Cache the Encoder Within:Compact, Reusable Memory across LLM Queries
缓存编码器：跨大型语言模型查询的紧凑可重用内存
Abstract
Repeated queries over shared documents incur redundant encoding, while caching model states introduces persistent storage costs. Building on CoMem's intermediate-state interface, EncBank treats a pretrained LLM's lower layers as a reusable document encoder and compactly stores their outputs for an adapted upper-layer reader. A self-distilled suffix adapter is shared across storage precisions within each backbone, without quantization-specific retraining. Across five benchmark suites on three Qwen backbones spanning different sizes and full-attention and hybrid architectures, 4-bit storage keeps each reported benchmark aggregate within one score point of native-precision EncBank. In a fixed Qwen3-8B workload, it retains 28.1% of the native-precision persistent GPU store. Separate native-precision controls yield a 1.40x selected-pack prefill speedup over same-evidence, same-adapter text replay, at a 3.12-point RULER accuracy cost. A native-precision Qwen3.8-27B configuration also passes 70 of 89 Terminal-Bench 2.1 tasks. EncBank thus combines reusable computation with compact memory, while task fidelity and end-to-end benefits remain dependent on the workload, preparation costs, and reuse frequency.
Chinese Translation
对共享文档的重复查询会导致冗余编码，而缓存模型状态则引入了持久存储成本。基于 CoMem 的中间状态接口，EncBank 将预训练大型语言模型（LLM）的低层视为可重用的文档编码器，并紧凑地存储其输出，以供适配的上层读取器使用。在每个主干网络内，共享的自蒸馏后缀适配器在不同存储精度之间使用，而无需针对量化进行特定的再训练。在三个不同规模的 Qwen 主干网络上，涵盖全注意力和混合架构的五个基准测试套件中，4 位存储使每个报告的基准聚合保持在原生精度 EncBank 的一个分数点内。在固定的 Qwen3-8B 工作负载下，它保留了 28.1% 的原生精度持久 GPU 存储。单独的原生精度控制在相同证据、相同适配器的文本重播中实现了 1.40 倍的选择包预填充加速，代价是 3.12 分的 RULER 准确性损失。原生精度的 Qwen3.8-27B 配置也通过了 89 个 Terminal-Bench 2.1 任务中的 70 个。因此，EncBank 将可重用计算与紧凑内存相结合，同时任务的保真度和端到端的收益仍然依赖于工作负载、准备成本和重用频率。
cs.CL / 51 / 2610.10091
ExperienceIndex: Artifact-Grounded Memory
经验指数：基于文献的记忆
Abstract
Knowledge-intensive tasks require answering many questions by reasoning about a shared corpus of artifacts (e.g., court cases, or scientific literature). As humans interact with these corpora, they naturally accumulate experiential knowledge about artifacts, enabling them to quickly identify the complete set of relevant artifacts for each new task. However, existing AI agents lack appropriate memory solutions to build or reuse such artifact-grounded experience, leading to lower answer quality and higher online cost. Existing memory solutions extract and reuse information from prior task-solving traces, but they primarily focus on user preferences, factual attributes, or abstract reasoning patterns rather than persistent artifact-specific knowledge. We introduce ExperienceIndex, a novel experience layer for AI agents that captures and reuses knowledge about artifacts based on prior reasoning traces. ExperienceIndex stores two complementary forms of experience: (i) single-artifact experiences that summarize an artifact's contribution to prior tasks and (ii) artifact-pair experiences that encode structural relationships discovered during past reasoning. Integrated as lightweight middleware, ExperienceIndex uses an experience retrieval mechanism to guide agents toward the complete set of relevant artifacts for new tasks, improving both answer quality and efficiency. Across diverse corpora and agentic solutions with different search frameworks, ExperienceIndex delivers consistent gains, raising answer quality by up to 11.0 points and reducing online dollar cost by up to 50.5%. We further demonstrate two benefits: (i) cross-task generalization, where experiences accumulated from text-to-SQL tasks transfer to factoid QA tasks over the same artifact corpus, and (ii) teacher-student learning, where experiences from a stronger model enable a weaker model to reach comparable performance.
Chinese Translation
知识密集型任务需要通过推理共享文献（例如，法庭案例或科学文献）来回答许多问题。当人类与这些文献互动时，他们自然会积累关于文献的经验知识，使他们能够快速识别每个新任务所需的完整相关文献集。然而，现有的人工智能代理缺乏适当的记忆解决方案来构建或重用这种基于文献的经验，导致回答质量降低和在线成本增加。现有的记忆解决方案提取并重用来自先前任务解决轨迹的信息，但它们主要关注用户偏好、事实属性或抽象推理模式，而不是持久的文献特定知识。我们提出了经验指数（ExperienceIndex），这是一个新颖的经验层，用于人工智能代理，基于先前的推理轨迹捕获和重用关于文献的知识。经验指数存储两种互补的经验形式：（i）单一文献经验，概述文献对先前任务的贡献；（ii）文献对经验，编码在过去推理中发现的结构关系。作为轻量级中间件集成，经验指数使用经验检索机制引导代理找到新任务所需的完整相关文献集，从而提高回答质量和效率。在不同的文献和具有不同搜索框架的代理解决方案中，经验指数提供了一致的收益，将回答质量提高了多达11.0分，并将在线成本降低了多达50.5%。我们进一步展示了两个好处：（i）跨任务泛化，从文本到SQL任务中积累的经验能够转移到同一文献集上的事实问答任务；（ii）师生学习，来自更强模型的经验使得较弱模型能够达到可比的性能。
cs.CL / 52 / 2610.10092
I would rather quit NLP than read another paper like this: The rise of antithesis in NLP papers
我宁愿退出自然语言处理（NLP）领域，也不想再读这样的论文：NLP 论文中对立面的兴起
Abstract
For better or worse, LLMs are by now used routinely for scientific writing.\footnote{This paper is no exception; we did use AI to assist with writing some of the sections (see Acknowledgments).} Many have noticed that recent models fill papers with unnecessary antithesis, stating over and over what the work does not do, in ways that do not contribute to its precision or quality of expression and annoy reviewers \emph{rather than impressing them}. We study the construction \emph{rather than} in ACL papers from 2019, ACL-style arXiv papers from 2026, and papers written by GPT models from the same titles and abstracts. Its rate in 2026 is seven times the 2019 rate, and higher still in the GPT papers. Two annotators, blind to the source, find almost no 2019 use \emph{annoying} and about one in ten 2026 uses; they seldom agree on which, yet about half of 2026 papers contain a use that annoys each of them. \emph{Annoying} uses present the rejected alternative less favorably than legitimate uses. Raters of preference data and open reward models favor the construction, and an instruction to be honest promotes it. We conjecture that it is a side effect of post-training on pairwise preferences, which credit a disavowal in a single response and cannot register its cost across a text.
Chinese Translation
无论好坏，当前大型语言模型（LLMs）已被常规用于科学写作。 ootnote{本文也不例外；我们确实使用了人工智能来协助撰写某些部分（见致谢）。} 许多人注意到，最近的模型在论文中充斥着不必要的对立面，一再陈述该工作未做的事情，这种方式并未提高其精确性或表达质量，反而让审稿人感到 extit{烦恼}而非印象深刻。我们研究了2019年ACL论文、2026年ACL风格的arXiv论文以及由GPT模型撰写的同标题和摘要的论文中 extit{rather than}的构造。2026年的使用频率是2019年的七倍，而在GPT论文中更高。两位评审员在不知来源的情况下，几乎没有发现2019年使用的 extit{烦恼}，而在2026年中大约十分之一的使用让他们感到烦恼；他们对哪些使用感到烦恼的意见很少一致，但约一半的2026年论文包含了让他们各自感到烦恼的用法。 extit{烦恼}的用法比合法用法更不利于呈现被拒绝的替代方案。偏好数据和开放奖励模型的评分者更倾向于这种构造，而诚实的指令则促进了它的使用。我们推测，这是一种在成对偏好后训练的副作用，这种训练在单个响应中认可否认，但无法在文本中注册其成本。
cs.CL / 53 / 2610.10114
Mechanics of Long-Context Hybrid Models Part 1.1: From Hybrid Attention to Hybrid Position
长上下文混合模型的力学 第1.1部分：从混合注意力到混合位置
Abstract
The architectural design of Large Language Models (LLMs) is shifting from traditional full-attention-only models to hybrid models, which combine different attention modules to improve long-context efficiency and performance in length extrapolation and context extension. To explain why hybrid models work and how to design them better, we propose Mechanics of Long-Context Hybrid Models. As Part 1.1 of this series, we begin with hybrids of full attention and either sliding-window attention (SWA) or gated variants of linear attention (LA), represented by GLA and GDN. We first observe a Seesaw Effect in Context Extension: LA hybrids benefit more from long-context continual pretraining, whereas SWA hybrids perform better under length extrapolation. We attribute this behavior to differences in the positional inductive biases induced by these attention mechanisms. We find that SWA hybrids suffer from a Short-Context Learning Trap, Short-Window Weariness, and Long-Window Laziness, and require extended windows to enhance performance in continual long-context pretraining. For LA hybrids, we summarize the Matthew Effect of Hybrid Position Extrapolation and propose Sliding-Window Linear Attention, achieving 16$\times$ training-free length extrapolation while maintaining 100\% accuracy on NIAH-SK1 in 64k context length.
Chinese Translation
大型语言模型（LLMs）的架构设计正从传统的全注意力模型转向混合模型，这些模型结合了不同的注意力模块，以提高长上下文的效率和在长度外推及上下文扩展中的表现。为了阐明混合模型为何有效以及如何更好地设计它们，我们提出了长上下文混合模型的力学。作为本系列的第1.1部分，我们首先讨论全注意力与滑动窗口注意力（SWA）或线性注意力（LA）的门控变体的混合，分别用GLA和GDN表示。我们首先观察到上下文扩展中的跷跷板效应：LA混合模型在长上下文持续预训练中获益更多，而SWA混合模型在长度外推中表现更佳。我们将这种行为归因于这些注意力机制所引发的位置信息归纳偏差的差异。我们发现SWA混合模型遭遇短上下文学习陷阱、短窗口疲劳和长窗口懒惰，并且需要扩展窗口以提高在持续长上下文预训练中的表现。对于LA混合模型，我们总结了混合位置外推的马太效应，并提出了滑动窗口线性注意力，实现了16倍的无训练长度外推，同时在64k上下文长度下保持了NIAH-SK1的100%准确率。
cs.CL / 54 / 2610.10138
LLM4Impact: Integrating Heterogeneous Information for Scientific Impact Prediction
LLM4Impact：整合异构信息以预测科学影响
Abstract
Predicting the future impact of a newly published paper is challenging because it must be inferred from heterogeneous evidence available at publication time. Existing approaches often rely on a single source of information or combine multiple sources without accounting for their different predictive roles. In this paper, we present LLM4Impact, an evidence-aware method for scientific impact prediction that learns to represent, integrate, and calibrate heterogeneous information. LLM4Impact combines semantic, graph, LLM, and temporal representations, and injects graph information into a frozen LLM through continuous prefix tokens. A context aware gating mechanism adaptively weights different evidence, while a separate calibration module accounts for domain and temporal variation in citation scales. We further construct a large-scale benchmark dataset with 2 million papers, leakage-safe point-in-time heterogeneous ego graphs, temporal splits, and both year-level and month-level citation targets. Experiments show that LLM4Impact consistently outperforms strong semantic, graph, and LLM based baselines, with a 10.13% reduction in year RMSE on the in distribution test set and a 6.87% reduction under out-of-domain distribution. Our results reveal that the value of such evidence is context dependent: different papers benefit from different sources, while domain and publication time affect how evidence translates into citations. This finding motivates adaptive evidence selection and context-conditioned calibration rather than simply richer representations. We will release our code, benchmark, and an interactive web demonstration upon publication.
Chinese Translation
预测新发表论文的未来影响具有挑战性，因为必须从发表时可用的异构证据中推断。现有的方法通常依赖于单一信息源，或在未考虑其不同预测角色的情况下组合多个信息源。本文提出了LLM4Impact，这是一种关注证据的科学影响预测方法，旨在学习表示、整合和校准异构信息。LLM4Impact结合了语义、图、LLM（大语言模型）和时间表示，并通过连续前缀令牌将图信息注入到冻结的LLM中。一个上下文感知的门控机制自适应地对不同证据进行加权，而一个单独的校准模块则考虑了引文尺度中的领域和时间变化。我们进一步构建了一个大规模基准数据集，包含200万篇论文、泄漏安全的时间点异构自我图、时间划分，以及年级和月级引文目标。实验表明，LLM4Impact在强语义、图和基于LLM的基线模型上始终表现优异，在内部分布测试集上年均方根误差（RMSE）减少了10.13%，在外部领域分布下减少了6.87%。我们的结果揭示了这种证据的价值是依赖于上下文的：不同的论文受益于不同的来源，而领域和发表时间影响证据如何转化为引文。这一发现促使我们采用自适应证据选择和上下文条件校准，而不仅仅是更丰富的表示。我们将在发表时发布我们的代码、基准和互动网页演示。
cs.CL / 55 / 2610.10145
InterView-C: A Synchronized Multimodal Corpus of VR Avatar-Mediated Survey Interviews
InterView-C：一个同步的虚拟现实化身中介调查访谈的多模态语料库
Abstract
We present InterView-C, a German multimodal corpus of 27 survey interviews conducted entirely in virtual reality, with both interlocutors represented by avatars. The corpus aligns spoken interaction with synchronized behavioral data, including gaze, head and body movement, facial behavior, hand and finger tracking. Its reference transcripts and linguistic annotations provide a reliable interface between this multimodal spoken interaction and predominantly text-based NLP methods. This interface is important because automatically transcribing speech can distort linguistically relevant information, while downstream models trained on existing resources may additionally face transfer challenges when applied to transcribed spoken data. InterView-C therefore provides word-timed and manually post-edited verbatim transcripts for all 54 recordings, interview-item timings, questionnaire responses and negation cue and scope annotations for 1,422 sentences, 1,398 of them doubly annotated ({\alpha}=0.87 for cues; {\alpha}=0.81 for scopes). We demonstrate both challenges empirically: nine open-weight ASR systems disproportionately misrecognize short closed answers and number words, while negation models trained on existing corpora show lower and highly variable performance on our transcribed interviews than a model trained on the InterView-C annotations. InterView-C thus enables linguistic analyses of spoken interaction while retaining their alignment with rich multimodal behavior.
Chinese Translation
我们提出了InterView-C，这是一个包含27个完全在虚拟现实中进行的调查访谈的德语多模态语料库，其中双方对话者均由化身表示。该语料库将口语互动与同步的行为数据对齐，包括注视、头部和身体运动、面部行为、手部和手指追踪。其参考转录文本和语言注释为这种多模态口语互动与主要基于文本的自然语言处理（NLP）方法之间提供了可靠的接口。这个接口非常重要，因为自动转录语音可能会扭曲与语言相关的信息，而在现有资源上训练的下游模型在应用于转录的口语数据时可能还会面临迁移挑战。因此，InterView-C为所有54个录音提供了逐字时间标记和手动后期编辑的逐字转录、访谈项目时间、问卷响应以及1,422个句子的否定提示和范围注释，其中1,398个句子进行了双重注释（提示的{eta}=0.87；范围的{eta}=0.81）。我们在实证上展示了这两种挑战：九个开放权重的自动语音识别（ASR）系统对短的封闭回答和数字词的错误识别率较高，而在现有语料库上训练的否定模型在我们的转录访谈中的表现低且高度可变，低于在InterView-C注释上训练的模型。因此，InterView-C使得对口语互动的语言分析成为可能，同时保持其与丰富的多模态行为的对齐。
cs.CL / 56 / 2610.10154
HySPE: Positional Encoding via Symplectic Dual Shears
HySPE：通过辛双剪切实现的位置编码
Abstract
We introduce Hyperbolic Symplectic Positional Encoding (HySPE), grounding positional attention in non-compact symplectic transformations. While canonical Rotary Position Embedding (RoPE) parameterizes the compact, elliptic branch of $\Sp(2,\R)$ via rotations, HySPE operationalizes its hyperbolic branch via a damped symmetric composition of dual shears, yielding a conformally symplectic contraction with two spectral decay rates per channel pair. To eliminate the exponential representation drift inherent to naive absolute factorizations, we diagonalize the operator in its invariant eigenbasis and introduce blockwise coordinate rebasing with adaptive centered execution. This guarantees length-independent numerical bounds while matching cached RoPE forward latency (7.21\,ms on an RTX 4090). On TinyShakespeare, HySPE-UltraLong maintains an invariant perplexity of 4.810 up to $16\times$ zero-shot extrapolation ($L=4096$), whereas RoPE degrades to 131.198. Scaled to a 51M-parameter subword Transformer on WikiText-103 ($L_{\text{train}}=512$), HySPE closely matches RoPE in-domain while robustly extrapolating to length 8192, reducing tail perplexity by 83.9\% over RoPE. While these controlled experiments establish HySPE's extrapolation robustness and numerical stability, evaluating its scaling behavior on large-scale foundation models remains an important direction for future investigation.
Chinese Translation
我们提出了超曲面辛位置编码（HySPE），将位置注意力建立在非紧凑辛变换的基础上。尽管经典的旋转位置嵌入（RoPE）通过旋转对$ ext{Sp}(2, ext{R})$的紧凑椭圆分支进行参数化，HySPE则通过双剪切的阻尼对称组合实现其超曲面分支，从而产生具有每对通道两个谱衰减率的共形辛收缩。为了消除天真的绝对因子化固有的指数表示漂移，我们在其不变特征基中对算子进行对角化，并引入基于块的坐标重基与自适应中心执行。这保证了长度无关的数值界限，同时与缓存的RoPE前向延迟（在RTX 4090上为7.21毫秒）相匹配。在TinyShakespeare上，HySPE-UltraLong在$16 imes$零-shot外推（$L=4096$）时保持不变的困惑度为4.810，而RoPE降至131.198。在WikiText-103上，将其扩展到一个具有5100万个参数的子词Transformer（$L_{ ext{train}}=512$），HySPE在领域内与RoPE紧密匹配，同时在长度8192上稳健外推，将尾部困惑度降低了83.9 ext{%}。虽然这些受控实验确立了HySPE的外推鲁棒性和数值稳定性，但评估其在大规模基础模型上的扩展行为仍然是未来研究的重要方向。
cs.CL / 57 / 2610.10179
Beyond Outcome Rewards: Constructing and Assigning Retrieval Credit for Search Agents
超越结果奖励：为搜索代理构建和分配检索信用
Abstract
Search agents enable Large Language Models (LLMs) to iteratively retrieve and use information for complex multi-hop questions. Reinforcement Learning with Verifiable Rewards (RLVR) offers a promising approach for post-training such agents, but its reliance on sparse, outcome-based supervision can make credit assignment difficult and limit learning efficiency. In this paper, we systematically investigate how intermediate supervision can improve reinforcement learning for search agents. We study a range of reward-shaping and credit-assignment strategies that provide learning signals from intermediate retrieval steps. Building on these insights, we develop a training framework that combines intermediate signals with final outcome rewards to improve learning from multi-step search trajectories. Experiments across multiple benchmarks under matched training conditions demonstrate improvements in aggregate search-agent performance and show that both the choice of intermediate signal and where its credit is assigned affect training behaviour. These findings show that reward design and credit assignment are important design dimensions for training effective search agents.
Chinese Translation
搜索代理使大型语言模型（LLMs）能够迭代地检索和使用信息以解决复杂的多跳问题。可验证奖励的强化学习（RLVR）为后期训练这些代理提供了一种有前景的方法，但其对稀疏的、基于结果的监督的依赖可能使信用分配变得困难，并限制学习效率。本文系统地研究了中间监督如何改善搜索代理的强化学习。我们研究了一系列奖励塑造和信用分配策略，这些策略从中间检索步骤提供学习信号。在这些见解的基础上，我们开发了一个训练框架，将中间信号与最终结果奖励结合起来，以改善从多步骤搜索轨迹中学习的效果。在匹配训练条件下的多个基准测试中的实验表明，搜索代理的整体性能有所提升，并且中间信号的选择及其信用分配的位置都会影响训练行为。这些发现表明，奖励设计和信用分配是训练有效搜索代理的重要设计维度。
cs.CL / 58 / 2610.10232
LLM Persuasion Is in the Eye of the Evaluation
大型语言模型的说服力在于评估的视角
Abstract
Large language models (LLMs) have already been shown to match or exceed human experts in persuasion. While their persuasive capabilities hold promise for beneficial uses such as education and health communication, they can also be used to manipulate and misinform, making their evaluation a growing priority for developers and regulators. That evaluation, however, remains fragmented: studies differ in what they treat as persuasion, and broad claims often rest on narrow, situation-specific assessments. Automated methods, often modelled on human studies, offer a way to compare such assessments directly, as they can be run on the same models at scale and can include high-risk forms of persuasion that would be difficult or unethical to test on people. In this study, we adapt nine published automated methods to a shared setup, run them on the same fifteen LLMs, and ask whether their rankings agree and why. We find that the methods agree only weakly (mean Spearman $\rho = 0.25$). Our analyses point to two contributing factors. Models that refuse some tasks but not others, directly or indirectly, lower agreement by about a quarter, and these refusals fall mostly on manipulation tasks. General capability also plays a part: most rational persuasion (non-manipulative) methods track it, whereas most manipulation methods do not. Together, these findings suggest that agreement depends more on the task a method sets than on how it scores persuasion, although this pattern is only indicative given the eight methods available for analysis. More broadly, our results suggest that persuasion scores combine a model's ability to persuade with its willingness to do so. A single score is therefore informative about its own setting, but says little about a model's persuasiveness across tasks.
Chinese Translation
大型语言模型（LLMs）已经被证明在说服力方面能够与人类专家相匹敌或超越。尽管它们的说服能力在教育和健康传播等有益用途上展现出潜力，但也可能被用于操控和误导，这使得对其评估成为开发者和监管者日益重要的任务。然而，这种评估仍然是碎片化的：不同研究对说服力的定义各异，广泛的结论往往基于狭窄的、特定情境的评估。自动化方法，通常基于人类研究模型，提供了一种直接比较这些评估的方式，因为它们可以在相同的模型上大规模运行，并且可以包括那些在伦理上难以或不适合在人类身上测试的高风险说服形式。在本研究中，我们将九种已发布的自动化方法适配到一个共享的设置中，在同样的十五个LLMs上运行这些方法，并探讨它们的排名是否一致以及原因。我们发现，这些方法之间的共识仅为微弱（平均斯皮尔曼相关系数 $
ho = 0.25$）。我们的分析指出了两个影响因素。拒绝某些任务但不拒绝其他任务的模型，直接或间接地降低了一致性约四分之一，而这些拒绝主要集中在操控任务上。一般能力也起到了一定作用：大多数理性说服（非操控性）方法与其相关，而大多数操控方法则没有。综合来看，这些发现表明，一致性更多地依赖于方法设定的任务，而不是其如何评分说服力，尽管这一模式仅在可供分析的八种方法中具有指示性。更广泛地说，我们的结果表明，说服评分结合了模型的说服能力与其愿意进行说服的程度。因此，单一评分在其自身设置中是有信息量的，但对模型在不同任务中的说服力却提供了有限的信息。
cs.CL / 59 / 2610.10318
Nobody Truly Agrees on Sentiment: Humans, Bespoke Tools, and LLMs Struggle with Social Media Texts
情感分析并无真正共识：人类、定制工具与大型语言模型在社交媒体文本中的挑战
Abstract
Social media is a rich source of real-time public sentiment, but widely used sentiment analysis tools are often applied without understanding their limitations. In this study, we evaluate the inter-rater reliability of three bespoke sentiment analysis tools (TextBlob, VADER, and Twitter-roBERTa-base) and three large language models (LLMs: Qwen3-32B, GPT-OSS-120B, Llama-4-Maverick-17B) against six human raters across 100 tweets. We measured agreement using two statistical measures: Cohen's kappa for pairwise comparisons and Fleiss' kappa for multiple raters. Even among the human raters, our results showed only fair agreement, highlighting the subjectivity of sentiment analysis. Higher agreement was observed under the binary sentiment classification (negative vs. non-negative and positive vs. non-positive) than under the three-class classification across both humans and automated tools. The Twitter-roBERTa-base model showed the strongest alignment with human ratings, outperforming both bespoke sentiment tools and LLMs, particularly in distinguishing negative versus non-negative sentiment. LLMs showed substantial agreement among themselves and moderate to substantial alignment with humans, performing better in positive vs. non-positive classifications. Our findings underscore that domain-specific fine-tuning remains crucial for reliable social media sentiment analysis, and human-centered evaluation remains essential for establishing gold-standard labels.
Chinese Translation
社交媒体是实时公众情感的丰富来源，但广泛使用的情感分析工具往往在不理解其局限性的情况下被应用。在本研究中，我们评估了三种定制情感分析工具（TextBlob、VADER 和 Twitter-roBERTa-base）以及三种大型语言模型（LLMs：Qwen3-32B、GPT-OSS-120B、Llama-4-Maverick-17B）在100条推文中与六位人类评审者之间的评分一致性。我们使用两种统计指标来测量一致性：Cohen's kappa 用于成对比较，Fleiss' kappa 用于多评审者。即使在六位人类评审者之间，我们的结果也仅显示出一般的一致性，突显了情感分析的主观性。在二元情感分类（负面与非负面、正面与非正面）下，无论是人类还是自动工具，观察到的一致性高于三类分类。Twitter-roBERTa-base 模型与人类评分的对齐程度最高，超越了定制情感工具和大型语言模型，特别是在区分负面与非负面情感方面。大型语言模型之间表现出显著的一致性，并与人类的对齐程度中等到显著，在正面与非正面分类中表现更佳。我们的研究结果强调，领域特定的微调对于可靠的社交媒体情感分析仍然至关重要，而以人为中心的评估对于建立金标准标签仍然不可或缺。
cs.CL / 60 / 2610.10332
Learning to Act with Task Progress: Distilling Small Agents from Compact Teacher Supervision
基于任务进展学习行动：从紧凑教师监督中提炼小型智能体
Abstract
Learning from large-model demonstrations offers a way to train small agents that can complete recurring tasks without calling a large model at every step. A central design choice is what to retain from teacher trajectories that contain reasoning, actions, and information about task progress. We introduce Task-Progress Distillation (TPD), an offline approach that pairs each demonstrated action with a short label describing the current task stage. The student learns these compact targets and selects actions by jointly scoring admissible stage--action pairs, which a deterministic harness executes in the environment. On ALFWorld, a 1.7B student trained with 404 demonstrations achieves 72.4\% mean unseen task success with either TPD or action-only supervision, compared with 48.3\% for a reasoning-trained student using constrained action selection. Explicit stages provide an additional benefit at 200 demonstrations, improving success from 48.0\% to 67.7\% over action-only supervision. With more demonstrations, the action-only student closes the gap, and both approaches reach 76.9\% at 808 demonstrations. Shared-history analyses link part of TPD's local advantage to better decisions when moving between subgoals, particularly from object acquisition to processing. These results show that compact supervision can train effective small task agents, while explicit task progress provides additional guidance at an intermediate demonstration budget.
Chinese Translation
从大模型示范中学习为训练能够在不每一步都调用大模型的情况下完成重复任务的小型智能体提供了一种方法。一个核心设计选择是从包含推理、行动和任务进展信息的教师轨迹中保留什么。我们提出了任务进展蒸馏（Task-Progress Distillation, TPD），这是一种离线方法，将每个示范行动与描述当前任务阶段的简短标签配对。学生学习这些紧凑目标，并通过联合评分可接受的阶段-行动对来选择行动，这些行动由确定性执行器在环境中执行。在 ALFWorld 上，一名使用 404 个示范训练的 17 亿参数学生在使用 TPD 或仅行动监督的情况下实现了 72.4\% 的平均未见任务成功率，而使用受限行动选择的推理训练学生的成功率为 48.3\%。在 200 个示范时，显式阶段提供了额外的好处，将成功率从 48.0\% 提高到 67.7\% 相较于仅行动监督。随着示范数量的增加，仅行动学生缩小了差距，两个方法在 808 个示范时均达到了 76.9\%。共享历史分析将 TPD 的局部优势部分归因于在子目标之间移动时做出更好的决策，特别是在从物体获取到处理的过程中。这些结果表明，紧凑的监督可以训练有效的小型任务智能体，而显式的任务进展在中等示范预算下提供了额外的指导。
cs.CL / 61 / 2610.10378
Document-Level Text Simplification in Estonian Using Large Language Models
使用大型语言模型进行爱沙尼亚语的文档级文本简化
Abstract
Document-level text simplification involves transformations that go beyond sentence-internal edits, addressing discourse coherence, anaphora resolution, and cross-paragraph consistency. Despite advances in sentence-level simplification for high-resource languages, document-level simplification in morphologically rich, low-resource languages such as Estonian remains largely unexplored. This study presents a comprehensive evaluation of five state-of-the-art multilingual large language models (LLMs) for document-level simplification in Estonian. Three prompting strategies are examined: single-pass generation, pipeline-based modular agents, and guideline-augmented pipelines. The evaluation framework integrates automatic metrics assessing readability, semantic preservation, and discourse coherence, alongside a structured manual annotation protocol. The findings indicate that Gemini-2.0 and LLaMA-3.3 produce outputs with near-native fluency and strong meaning preservation, whereas other models display notable grammatical and semantic limitations. This work contributes novel document-level coherence metrics, evidence-based prompting strategies, and publicly available resources for reproducibility.
Chinese Translation
文档级文本简化涉及超越句内编辑的转化，涵盖话语连贯性、指代解析和跨段落一致性。尽管在高资源语言的句子级简化方面取得了进展，但在形态丰富、资源稀缺的语言（如爱沙尼亚语）中，文档级简化仍然基本未被探索。本研究对五种最先进的多语言大型语言模型（LLMs）在爱沙尼亚语文档级简化中的表现进行了全面评估。研究考察了三种提示策略：单次生成、基于管道的模块化代理和增强指导的管道。评估框架整合了自动化指标，以评估可读性、语义保留和话语连贯性，同时采用结构化的人工注释协议。研究结果表明，Gemini-2.0 和 LLaMA-3.3 生成的输出接近母语流利度，并且在意义保留方面表现出色，而其他模型则显示出显著的语法和语义局限性。本研究贡献了新颖的文档级连贯性指标、基于证据的提示策略以及可公开获取的可重复性资源。
cs.CL / 62 / 2610.10426
CoTrace: Data Recipes for Training Terminal Agents with Harness-Model Co-Evolution
CoTrace：用于训练终端智能体的带有工具模型协同进化的数据配方
Abstract
Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery. Existing harness-model co-evolution approaches improve both components, yet often treat trajectories produced during harness search as an undifferentiated replay buffer. This practice overlooks that a trajectory's value for model training depends on the harness under which it was generated. To systematically analyze this interface, we establish an alternating co-evolution framework that decouples harness search and policy training through component-wise promotion decisions. Within this framework, we introduce CoTrace, a harness-aware data recipe that explicitly governs trajectory routing, provenance matching, and curriculum refresh. Under CoTrace, recurring execution failures guide harness synthesis, while policy training is strictly conditioned on verified rollouts matched to the adopted runtime for supervised fine-tuning (SFT) or fresh online interactions for reinforcement learning (RL). On the Tmax promotion split, CoTrace advances Qwen3.5-9B from 78 to 88 solved tasks under supervised fine-tuning while an online reinforcement variant reaches 90. Specifically, a compact harness-matched corpus produces steady model gains at substantially lower compute than much larger corpora pooled across sibling harnesses. Furthermore, evaluations on Terminal-Bench 2.1 and SWE-bench Lite show that out-of-distribution transfer depends fundamentally on harness compatibility, where maintaining consistency between training and evaluation runtimes prevents procedural execution breakdowns observed under foreign scaffolds.
Chinese Translation
终端智能体的能力依赖于模型权重和运行时工具的共同作用，后者负责格式化提示、绑定工具和处理错误恢复。现有的工具-模型协同进化方法改善了这两个组件，但通常将工具搜索过程中产生的轨迹视为一个未加区分的重放缓冲区。这种做法忽视了轨迹在模型训练中的价值取决于其生成时所用的工具。为了系统地分析这一接口，我们建立了一个交替协同进化框架，通过组件级的提升决策将工具搜索与策略训练解耦。在这个框架内，我们引入了CoTrace，这是一种关注工具的数据配方，明确管理轨迹路由、来源匹配和课程更新。在CoTrace下，重复的执行失败指导工具的合成，而策略训练严格依赖于与采用的运行时匹配的验证回放，以进行监督微调（SFT）或用于强化学习（RL）的新在线交互。在Tmax提升分割上，CoTrace使Qwen3.5-9B在监督微调下从78个解决任务提升至88个，而在线强化变体达到90个。具体而言，一个紧凑的工具匹配语料库在计算资源消耗上显著低于跨兄弟工具汇聚的大型语料库，且仍能产生稳定的模型增益。此外，在Terminal-Bench 2.1和SWE-bench Lite上的评估表明，分布外迁移根本上依赖于工具的兼容性，保持训练和评估运行时之间的一致性可以防止在外部框架下观察到的程序执行失败。
cs.CL / 63 / 2610.10455
PHRBench: A Behavioral Evaluation of Post-Hallucination Reasoning in LLMs
PHRBench：大语言模型后幻觉推理的行为评估
Abstract
Hallucinated information can propagate through multi-stage LLM systems and become part of the context for subsequent reasoning. Existing studies of post-hallucination reasoning (PHR) mainly characterize changes in final outcomes and aggregate reasoning dynamics, leaving how models resolve hallucinated premises at the response level insufficiently understood. In this work, we introduce PHRBench, a controlled benchmark for behaviorally structured PHR across four domains and 18 large language models. PHRBench characterizes each reasoning trajectory independently of final-answer correctness through Hallucination Compliance, Hallucination Avoidance, and Heuristic Correction, and defines an insightful trajectory as successful correction that ultimately reaches the correct answer. Across 4820 controlled instances, we find that successful recovery remains relatively rare and is associated with more frequent belief updates along the reasoning trajectory. We further find that properties of the hallucinated prompt contain substantial predictive signal for successful recovery, with a lightweight predictor achieving an AUROC of 0.847. These findings provide a behavioral view of post-hallucination reasoning, characterizing how LLMs resolve erroneous context and when successful recovery is likely to occur.
Chinese Translation
幻觉信息可以在多阶段的大语言模型系统中传播，并成为后续推理的上下文的一部分。现有的后幻觉推理（PHR）研究主要描述最终结果的变化和整体推理动态，而对模型在响应层面如何解决幻觉前提的理解仍然不足。在本研究中，我们引入了PHRBench，这是一个针对四个领域和18个大型语言模型的行为结构化PHR的受控基准。PHRBench通过幻觉合规性（Hallucination Compliance）、幻觉避免（Hallucination Avoidance）和启发式修正（Heuristic Correction）独立表征每个推理轨迹，而不依赖于最终答案的正确性，并将成功修正定义为最终达到正确答案的有意义轨迹。在4820个受控实例中，我们发现成功恢复仍然相对稀少，并且与推理轨迹中更频繁的信念更新相关。我们进一步发现，幻觉提示的特性包含了成功恢复的显著预测信号，轻量级预测器的AUROC达到了0.847。这些发现提供了后幻觉推理的行为视角，表征了大语言模型如何解决错误上下文以及成功恢复可能发生的时机。
cs.CL / 64 / 2610.10508
Your Prompt Should Do More: Effects of Retrieval Instructions in Embedding Models
你的提示应该做得更多：检索指令在嵌入模型中的影响
Abstract
Prompted embedding models have recently received increasing attention, particularly for retrieval, where detailed retrieval instructions are provided as part of the retrieval prompt. Several new datasets and studies have examined this setting, showing that the current embedding models often struggle to follow such instructions reliably. In this paper, we study the mechanism of how instructions actually affect the representations of retrieval queries in asymmetric retrieval tasks. We show that models can fail to follow even simple task instructions when query-side distractors are included in the evaluation. We hypothesize that this behavior is driven by the training setup of current embedding models and their evaluation, and show that fine-tuning with added query-side distractors leads to substantial improvements, with minimal effect on other tasks.
Chinese Translation
最近，提示嵌入模型受到了越来越多的关注，特别是在检索任务中，详细的检索指令作为检索提示的一部分被提供。几个新的数据集和研究考察了这一设置，显示当前的嵌入模型在可靠地遵循这些指令方面常常面临困难。在本文中，我们研究了指令如何实际影响不对称检索任务中检索查询的表示机制。我们表明，当评估中包含查询侧的干扰项时，模型甚至可能无法遵循简单的任务指令。我们假设这种行为是由当前嵌入模型的训练设置及其评估驱动的，并展示了通过增加查询侧干扰项进行微调可以带来显著的改进，同时对其他任务的影响最小。
cs.CL / 65 / 2610.10533
EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory
EngramEdit：通过条件记忆实现大语言模型中的解耦知识更新
Abstract
Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture has demonstrated the potential to decouple factual knowledge storage from general-purpose computation, offering a promising route to updating factual knowledge while keeping the Transformer backbone fixed. Realizing this potential is challenging because different expressions of a fact may activate different n-gram embeddings, while updating shared embeddings can unintentionally change the model's predictions about other facts. We propose EngramEdit for decoupled knowledge updates through conditional memory. EngramEdit first computes target memory representations that make the model predict the updated fact across multiple expressions. It then jointly updates the shared n-gram embeddings to match these targets across expressions and edits, penalizing updates to frequently reused embeddings more strongly to preserve unrelated knowledge. Experiments show that EngramEdit enables independent factual knowledge updates through conditional memory, achieving near-perfect editing success. Revised knowledge is usable across unseen expressions and in multi-hop reasoning, with nearly three times the strongest baseline's accuracy under chain-of-thought (CoT) prompting. Unrelated knowledge and general capabilities are largely preserved even as factual updates accumulate. These findings show that EngramEdit turns conditional memory into an editable knowledge interface, extending its role beyond model scaling to support decoupled knowledge updates.
Chinese Translation
条件记忆架构，如 DeepSeek Engram，利用输入 n-gram 查找学习到的嵌入，从而在有限的额外计算下扩展大型语言模型（LLMs）的容量。除了模型扩展外，该架构还展示了将事实知识存储与通用计算解耦的潜力，提供了一条有希望的途径来更新事实知识，同时保持 Transformer 主干不变。实现这一潜力具有挑战性，因为事实的不同表达可能激活不同的 n-gram 嵌入，而更新共享嵌入可能会无意中改变模型对其他事实的预测。我们提出了 EngramEdit，通过条件记忆实现解耦知识更新。EngramEdit 首先计算目标记忆表示，使模型能够在多种表达中预测更新后的事实。然后，它联合更新共享 n-gram 嵌入，以匹配这些目标在不同表达和编辑中的表现，并对频繁重用的嵌入施加更强的惩罚，以保留无关知识。实验表明，EngramEdit 通过条件记忆实现独立的事实知识更新，达到近乎完美的编辑成功率。修订后的知识可用于未见过的表达和多跳推理，在链式思维（CoT）提示下，其准确率几乎是最强基线的三倍。即使事实更新不断累积，无关知识和一般能力也在很大程度上得以保留。这些发现表明，EngramEdit 将条件记忆转变为可编辑的知识接口，扩展其角色，不仅限于模型扩展，还支持解耦知识更新。