cs.RO / 1 / 2609.22274
CHOREO: Every Humanoid Skill as a Trajectory
CHOREO：每个类人技能作为轨迹
Abstract
Recent advances in humanoid robotics have produced diverse skills through reinforcement learning, motion imitation, and generative modeling. Yet these capabilities remain siloed because they are built around incompatible representations, interfaces, and controllers. We present CHOREO, a framework for training-free composition of heterogeneous humanoid skills. Our key observation is that, regardless of how a skill is learned, it can ultimately be expressed as an executable motion trajectory. Based on this observation, CHOREO converts each capability into SkillMotion, a unified representation that combines motion states, contacts, semantics, and boundary conditions. Skills are composed through direct continuation, cubic Hermite blending, or validated bridge motions, without retraining source models or updating models at test time. On Unitree G1 in MuJoCo, CHOREO organizes 2,950 admitted SkillMotion assets derived from heterogeneous sources and achieves 95.4\% sequence success across 130 multi-action tasks, including 93.8\% success on eight-action sequences. These results demonstrate that executable trajectories provide a scalable interface for accumulating and composing pretrained humanoid capabilities.
Chinese Translation
近年来，类人机器人技术的进步通过强化学习、运动模仿和生成建模产生了多样化的技能。然而，这些能力仍然是孤立的，因为它们是基于不兼容的表示、接口和控制器构建的。我们提出了CHOREO，一个无训练的异构类人技能组合框架。我们的关键观察是，无论技能如何学习，最终都可以表示为可执行的运动轨迹。基于这一观察，CHOREO将每种能力转换为SkillMotion，一种统一的表示，结合了运动状态、接触、语义和边界条件。技能可以通过直接延续、三次Hermite插值或经过验证的桥接运动进行组合，而无需重新训练源模型或在测试时更新模型。在MuJoCo的Unitree G1上，CHOREO组织了2,950个来自异构来源的SkillMotion资产，并在130个多动作任务中实现了95.4%的序列成功率，包括在八个动作序列中的93.8%成功率。这些结果表明，可执行的轨迹为积累和组合预训练的类人能力提供了一个可扩展的接口。
cs.RO / 2 / 2609.22276
D3DWA: Adaptive Weight and Prediction-Horizon for Dynamic Window Approach via Dueling Double Deep Q-Network
D3DWA：通过对抗双深度Q网络的动态窗口方法的自适应权重和预测视野
Abstract
The Dynamic Window Approach (DWA) is widely used for local navigation, but its performance depends strongly on parameters that are typically fixed before navigation. In particular, the appropriate prediction horizon can vary with local free space: longer horizons support efficient motion in open areas, whereas shorter horizons help preserve feasible motions in narrow or cluttered regions. This paper proposes D3DWA, an adaptive DWA framework based on a Dueling Double Deep Q-Network (D3QN), which jointly selects the DWA evaluation weights and prediction horizon from a continuous navigation state at every control step while retaining DWA's trajectory generation and collision checking. In eight simulated environments, including unseen layouts, D3DWA reached every goal. Real-robot experiments further showed that D3DWA completed all three tested configurations, including a constrained case in which the weights-only variant timed out. These results demonstrate the benefit of jointly adapting the evaluation weights and prediction horizon. Additional material is available at https://mertcookimg.github.io/d3dwa/
Chinese Translation
动态窗口方法（DWA）广泛应用于局部导航，但其性能在很大程度上依赖于通常在导航前固定的参数。特别是，适当的预测视野会随着局部自由空间的变化而变化：较长的视野支持在开阔区域内的高效运动，而较短的视野则有助于在狭窄或杂乱区域内保持可行的运动。本文提出了D3DWA，一种基于对抗双深度Q网络（D3QN）的自适应DWA框架，该框架在每个控制步骤中从连续的导航状态中联合选择DWA评估权重和预测视野，同时保留DWA的轨迹生成和碰撞检测。在包括未见布局的八个模拟环境中，D3DWA成功到达每个目标。真实机器人实验进一步表明，D3DWA完成了所有三种测试配置，包括一个权重仅变体超时的受限情况。这些结果展示了联合自适应评估权重和预测视野的优势。更多材料可在 https://mertcookimg.github.io/d3dwa/ 获取。
cs.RO / 3 / 2609.22278
DeViGrasp: Robust Visual Mobile Grasping for Quadruped Manipulators under Degraded Perception
DeViGrasp：在感知退化情况下四足机器人稳健的视觉移动抓取
Abstract
Quadruped manipulators enable mobile grasping in complex environments, yet their whole-body control policies remain vulnerable to unreliable onboard visual perception. Existing methods are typically developed under relatively reliable observations and have not systematically examined how occlusion, segmentation-mask dropout, depth noise, and target-localization jitter affect grasp reasoning and target tracking. To address this gap, we introduce DeViGrasp-Bench, a benchmark for mobile grasping under degraded vision that incorporates controlled visual degradations, seen and unseen objects, multiple difficulty levels, and complex terrains, and evaluates task success, execution efficiency, and action smoothness. We further propose DeViGrasp-Net, a teacher--student framework that combines state-conditioned grasp reasoning with reliability-aware temporal target estimation. The privileged teacher attends to offline grasp candidates conditioned on object, robot, end-effector, and task states, while the deployable student fuses dual-view segmented-depth observations with current, memory, and recovery target hypotheses through Target Hold Memory and Temporal Memory Attention. DeViGrasp-Net outperforms VBC across degradation levels, unseen objects, and complex terrains, and surpasses an adapted DQ-Net across all evaluated degradation levels. Under the Difficult setting, it achieves a success rate of 62.3\%, improving upon VBC and DQ-Net by 16.1 and 4.3 percentage points, respectively; under the Hard setting, its margin over DQ-Net increases to 10.5 percentage points. Ablation studies confirm the complementary benefits of grasp-aware supervision and reliability-aware temporal memory.
Chinese Translation
四足机器人能够在复杂环境中实现移动抓取，但其全身控制策略仍然容易受到不可靠的机载视觉感知的影响。现有方法通常是在相对可靠的观测条件下开发的，尚未系统地研究遮挡、分割掩膜丢失、深度噪声和目标定位抖动如何影响抓取推理和目标跟踪。为了解决这一问题，我们引入了DeViGrasp-Bench，这是一个针对视觉退化下移动抓取的基准，包含了受控的视觉退化、已见和未见物体、多种难度级别以及复杂地形，并评估任务成功率、执行效率和动作平滑性。我们进一步提出了DeViGrasp-Net，一个结合状态条件抓取推理与可靠性感知时间目标估计的教师-学生框架。特权教师关注于基于物体、机器人、末端执行器和任务状态的离线抓取候选，而可部署的学生则通过目标保持记忆和时间记忆注意力，将双视图分割深度观测与当前、记忆和恢复目标假设融合。DeViGrasp-Net在各个退化级别、未见物体和复杂地形上均优于VBC，并在所有评估的退化级别上超越了改进版的DQ-Net。在困难设置下，其成功率达到62.3%，比VBC和DQ-Net分别提高了16.1和4.3个百分点；在困难设置下，其相对于DQ-Net的优势增加到10.5个百分点。消融研究证实了抓取感知监督和可靠性感知时间记忆的互补效益。
cs.RO / 4 / 2609.22285
ORDER: A Fictitious-World Benchmark for Domain-Adaptive Embodied AI
ORDER：一个用于领域适应的具身人工智能的虚构世界基准测试
Abstract
Adapting language models to new domains via continual pre-training raises a basic evaluation problem: if the training corpus overlaps with what the model already knows, performance gains cannot be cleanly attributed to new learning rather than pre-existing knowledge. This matters most for knowledge-intensive, task-light (KHTL) robot deployments - pharmaceutical dispensing, hazardous-material handling, facility-specific protocols, where the physical task is simple but the governing rules are proprietary and safety-critical, and where extensive live testing is costly or unsafe. We introduce ORDER (Ontology-driven Decision-making for Embodied Reasoning), a benchmark built on a fictitious world: a 342,069-token synthetic corpus defining a self-consistent physics that cannot appear in any model's pre-training data. ORDER pairs a 500-question knowledge test (ORDER-BENCH) with a harder compositional task, ORDER-SPATIAL: ordering objects for safe manipulation across both familiar and entirely novel scenes. GPT-4.1 without adaptation scores below chance on ORDER-SPATIAL (Kendall's tau = 0.441), showing its priors actively conflict with the invented physics. After continual pre-training, small models improve substantially on both familiar and novel scenes alike evidence of genuine world-model induction rather than memorization. We then carry this through to a robot pipeline: models that answer the knowledge test well often cannot produce valid, executable plans without a further skill-adaptation stage, after which small, fully offline models outperform GPT-4.1 even when GPT-4.1 is given retrieval access to the same rules (Kendall's tau = 0.848 vs. 0.606), on a full perception-to-execution loop demonstrated on a simulated iiwa7 arm with human-in-the-loop correction. Throughout, ORDER-SPATIAL performance, not knowledge-test accuracy is what predicts real plan quality.
Chinese Translation
通过持续的预训练将语言模型适应新领域引发了一个基本的评估问题：如果训练语料与模型已有知识重叠，则性能提升不能清晰地归因于新学习，而不是先前的知识。这对于知识密集型、任务轻量型（KHTL）机器人部署尤为重要——如药品分发、有害物质处理、特定设施协议等场景，其中物理任务简单，但管理规则是专有且关乎安全的，而广泛的现场测试成本高昂或不安全。我们引入了ORDER（基于本体的具身推理决策），这是一个建立在虚构世界上的基准：一个包含342,069个标记的合成语料库，定义了一种在任何模型的预训练数据中都无法出现的自洽物理。ORDER将一个包含500个问题的知识测试（ORDER-BENCH）与一个更具挑战性的组合任务ORDER-SPATIAL相结合：在熟悉和全新场景中为安全操作排序物体。未经适应的GPT-4.1在ORDER-SPATIAL上的得分低于随机水平（Kendall's tau = 0.441），显示其先验知识与所创造的物理相冲突。经过持续的预训练，小模型在熟悉和新场景中均有显著改善，证明了真正的世界模型归纳，而非单纯的记忆。随后，我们将这一过程应用于机器人管道：在知识测试中表现良好的模型往往无法在没有进一步技能适应阶段的情况下生成有效的可执行计划，而在经过该阶段后，小型完全离线模型的表现超越了GPT-4.1，即使在GPT-4.1获得相同规则的检索访问权限的情况下（Kendall's tau = 0.848 vs. 0.606），在一个全感知到执行的循环中，展示在模拟的iiwa7机械臂上进行人机协作的校正。总体而言，ORDER-SPATIAL的表现，而非知识测试的准确性，才是预测真实计划质量的关键。
cs.RO / 5 / 2609.22289
OJOx: Specification-Conditioned Demonstrations for Embodied AI in Construction
OJOx：面向建筑领域的具身人工智能的规范条件演示
Abstract
Large-scale egocentric and whole-body human demonstrations are becoming a primary source of data for embodied intelligence. They record what people perceive and do, but rarely the external specification that gave an action its purpose. In construction that omission is consequential: skilled work is directed at project-specific configurations defined in a design model - configurations not yet present in the environment being observed. A mason's transferable competence is not the geometry of one wall but the ability to realise a new geometry from a specification. We introduce the specification-conditioned demonstration: a synchronised record of the physical state a demonstrator perceives, the intended state supplied to them by an external design, and the behaviour connecting the two. We present OJOx, a capture interface that realises this for construction - delivering design geometry to a headset, anchoring it in the physical workspace, rendering it into a demonstrator's stereo passthrough view, and recording that view synchronously with whole-body and hand motion. We report one fully instrumented session - a 33-component wall laid against a specification that changes while the work proceeds - and check the record against the physical scene through an external camera registered independently of the capture. Recorded sessions remain compatible with existing humanoid retargeting infrastructure and replay onto a Unitree G1 in simulation. The result is a data interface for testing whether embodied policies can learn not merely to imitate demonstrated actions, but to act toward specifications absent from their training experience.
Chinese Translation
大规模的自我中心和全身人类演示正成为具身智能的主要数据来源。它们记录了人们的感知和行为，但很少记录赋予行动目的的外部规范。在建筑领域，这一遗漏是有重大影响的：熟练的工作是针对在设计模型中定义的项目特定配置进行的——这些配置在观察的环境中尚不存在。一个砌砖工的可转移能力不是一面墙的几何形状，而是根据规范实现新几何形状的能力。我们引入了规范条件演示：一个同步记录演示者所感知的物理状态、外部设计提供给他们的预期状态以及连接两者的行为。我们展示了OJOx，一个为建筑领域实现这一目标的捕捉接口——将设计几何传送到头戴设备中，将其锚定在物理工作空间中，将其渲染到演示者的立体透视视图中，并与全身和手部运动同步记录该视图。我们报告了一次完全仪器化的会话——一个33个组件的墙体在工作进行时与一个不断变化的规范相对照，并通过一个独立于捕捉的外部相机检查记录与物理场景的对应关系。记录的会话与现有的人形重定向基础设施兼容，并可在仿真中重放到Unitree G1上。最终结果是一个数据接口，用于测试具身策略是否能够学习不仅仅是模仿演示的动作，而是朝着在其训练经验中缺失的规范进行行动。
cs.RO / 6 / 2609.22299
When Does Test-Time Physical Diagnosis Pay? A Frozen Policy Buys Evidence It Never Reads
何时测试时的物理诊断有效？一个从未读取证据的固定策略
Abstract
When a robot faces unfamiliar physical conditions, a common approach is to collect evidence about what changed and adapt. For such diagnosis to improve behavior, six ordered empirical conditions must hold: a meaningful reference, identifiability of the physical condition, use of the acquired evidence, decision value, selection value over a fixed alternative, and safe realization. We test this chain in controlled and public environments. It holds end to end in our controlled environments. After transfer to unseen mechanisms, however, it breaks at evidence use. On decisions requiring the full trace, the frozen decoder does not change its choice. A linear model using only trace increments recovers the correct choice on mechanisms excluded from fitting, showing that the trace is informative but unused. The failure is concentrated at the richest evidence level: those decisions fall to chance, while decisions settled with lower-cost evidence remain correct, a split hidden by aggregate accuracy. The same chain can fail at other links in public environments. Successful physical identification therefore guarantees neither evidence use nor useful adaptation; evaluation should identify where the chain breaks rather than rely on recovery accuracy or aggregate performance alone.
Chinese Translation
当机器人面临不熟悉的物理条件时，常见的方法是收集关于变化的证据并进行适应。为了使这种诊断改善行为，必须满足六个有序的经验条件：有意义的参考、物理条件的可识别性、对所获得证据的使用、决策价值、相对于固定替代方案的选择价值，以及安全实现。我们在受控和公共环境中测试这一链条。在我们的受控环境中，它从头到尾都成立。然而，在转移到未见机制后，它在证据使用上出现了问题。在需要完整追踪的决策中，固定解码器并未改变其选择。仅使用追踪增量的线性模型在排除拟合的机制上恢复了正确选择，表明追踪是有信息量的但未被使用。失败集中在最丰富的证据层面：这些决策落入偶然，而用较低成本证据解决的决策仍然正确，这一分裂被整体准确性掩盖。相同的链条在公共环境中也可能在其他环节失败。因此，成功的物理识别并不保证证据使用或有用的适应；评估应识别链条断裂的地方，而不是仅依赖恢复准确性或整体性能。
cs.RO / 7 / 2609.22317
EditWM: Event-Decomposed World Modeling with Incremental Correction for End-to-End Autonomous Driving
EditWM：基于事件分解的世界建模与增量修正用于端到端自主驾驶
Abstract
World models support autonomous driving by predicting the scene evolution associated with candidate trajectories. Driving dynamics differ in predictability, motivating a distinction between regular evolution and event-induced deviations that call for selective correction. We propose EditWM, a world model that decomposes future prediction into normal evolution and event-driven incremental correction in compact visual feature space. A trajectory-conditioned normal predictor provides the base forecast and is then frozen for correction learning. A correction decoder compares this forecast with observation history and planned actions, producing a bounded feature update whose contribution is regulated by a learned gate. The corrected future features condition trajectory scoring through candidate-specific cross-attention, linking world modeling to plan selection. At inference, EditWM uses only past and current observations, ego state, and candidate trajectories. Across all 12,146 NAVSIM navtest scenes, expert-trajectory-conditioned evaluation shows a 5.35\% reduction in future-feature MSE over Normal, with improvements in 83.54\% of scenes. The system achieves 91.05 EPDMS on a 100-point scale using the official EPDMS evaluator. These results demonstrate improved future-feature prediction and competitive trajectory selection when corrected future representations are integrated into planning.
Chinese Translation
世界模型通过预测与候选轨迹相关的场景演变来支持自主驾驶。驾驶动态在可预测性上存在差异，这促使我们区分常规演变和事件引发的偏差，后者需要选择性修正。我们提出了EditWM，这是一种世界模型，它将未来预测分解为正常演变和事件驱动的增量修正，使用紧凑的视觉特征空间。一个基于轨迹的正常预测器提供基础预测，并在修正学习中被冻结。修正解码器将该预测与观察历史和计划动作进行比较，生成一个有界特征更新，其贡献由学习的门控机制调节。修正后的未来特征通过候选特定的交叉注意力来条件化轨迹评分，将世界建模与计划选择联系起来。在推理时，EditWM仅使用过去和当前的观察、主体状态以及候选轨迹。在所有12,146个NAVSIM导航测试场景中，基于专家轨迹的评估显示，与正常模型相比，未来特征均方误差（MSE）减少了5.35%，在83.54%的场景中有所改善。该系统在100分的评分标准下使用官方EPDMS评估器达到了91.05的EPDMS。这些结果表明，当修正后的未来表示被整合到规划中时，未来特征预测得到了改善，并且轨迹选择具有竞争力。
cs.RO / 8 / 2609.22319
Embedding Physics Priors in Robot Learning: A Survey
在机器人学习中嵌入物理先验：一项综述
Abstract
The rapid progress of artificial intelligence is reshaping robotics and accelerating the adoption of learning-based approaches. While purely data-driven methods have achieved remarkable success in computer vision and natural language processing, robotics remains constrained by limited data, complex real-world interactions, and the need for reliable operation. These challenges have motivated the exploration of physics-embedded robot learning, which embeds physics priors into learning algorithms. By encoding the underlying physical laws and constraints, physics priors can complement limited data with robotics-specific inductive biases, potentially improving generalization, interpretability, and sample efficiency. However, the literature on physics-embedded robot learning remains fragmented across terminology, methodologies, and application domains, making it difficult to assess this growing body of work. This survey reviews physics-embedded robot learning across a broad range of physics priors, robotics applications, and machine learning models, from single-layer perceptrons to generative foundation models. We adopt a unified taxonomy that classifies existing approaches according to their physics embedding: physics-guided inputs, data, and representations; physics-encoded model architectures; and physics-informed training loss functions. Building on this taxonomy, we review methods for robot dynamics learning, trajectory planning, prediction, control, and estimation, together with the corresponding open-source software ecosystem. We identify key open challenges, and outline promising future research directions. Overall, we argue that physics priors provide a particularly relevant robotics-specific inductive bias, complementing rather than replacing data-driven learning, and paving the way toward more generalizable, data-efficient, and trustworthy robotic systems.
Chinese Translation
人工智能的快速进展正在重塑机器人技术，并加速基于学习的方法的采用。尽管纯数据驱动的方法在计算机视觉和自然语言处理领域取得了显著成功，但机器人技术仍然受到有限数据、复杂的现实世界交互以及对可靠操作需求的限制。这些挑战促使了嵌入物理学的机器人学习的探索，该方法将物理先验嵌入学习算法中。通过编码潜在的物理法则和约束，物理先验可以用机器人特定的归纳偏差来补充有限的数据，从而有可能提高泛化能力、可解释性和样本效率。然而，关于嵌入物理学的机器人学习的文献在术语、方法论和应用领域上仍然是支离破碎的，这使得评估这一日益增长的研究领域变得困难。本文综述了嵌入物理学的机器人学习，涵盖了广泛的物理先验、机器人应用和机器学习模型，从单层感知器到生成基础模型。我们采用统一的分类法，根据物理嵌入的方式对现有方法进行分类：物理引导的输入、数据和表示；物理编码的模型架构；以及物理知情的训练损失函数。在此分类法的基础上，我们回顾了机器人动力学学习、轨迹规划、预测、控制和估计的方法，以及相应的开源软件生态系统。我们识别了关键的开放挑战，并概述了有前景的未来研究方向。总体而言，我们认为物理先验提供了一种特别相关的机器人特定归纳偏差，补充而非替代数据驱动学习，为更具泛化能力、数据高效和可信的机器人系统铺平了道路。
cs.RO / 9 / 2609.22325
ReliCAD: From Uncertain LLM Generation to Reliable Parametric CAD Modeling
ReliCAD：从不确定的LLM生成到可靠的参数化CAD建模
Abstract
Large language models have shown considerable potential for natural-language-driven parametric CAD modeling. However, a fundamental contradiction exists between their probabilistic generation and the deterministic requirements of CAD modeling, resulting in limitations in reliability, design-intent preservation, and geometric validity. Existing methods typically rely on large-scale annotated datasets, lack explicit modeling of design intent, and underutilize the deterministic capabilities of CAD kernels. To address these limitations, we propose ReliCAD, a unified framework that transforms uncertain LLM generation into reliable parametric CAD modeling. Through explicit design-intent modeling, ReliCAD converts user requirements into structured design specifications and explicitly models geometric relations, topological dependencies, and feature construction order. It then generates constraint-aware parametric instructions and invokes the CAD kernel through an Agent-ready API to perform geometric construction and constraint solving. ReliCAD further records runtime evidence and employs a verification-feedback mechanism to assess consistency between the generated model and the design specifications, enabling error localization and iterative repair. Experiments on the public HistCAD generation dataset and our multi-granularity CAD editing dataset demonstrate that ReliCAD significantly outperforms baseline methods, achieving 99.8\% validity rate and 0.8753 IoU. ReliCAD provides a verifiable, repairable, and generalizable approach to natural-language-interactive CAD modeling.
Chinese Translation
大型语言模型在自然语言驱动的参数化CAD建模中显示出了相当大的潜力。然而，它们的概率生成与CAD建模的确定性要求之间存在根本矛盾，导致在可靠性、设计意图保留和几何有效性方面的局限性。现有方法通常依赖于大规模标注数据集，缺乏对设计意图的明确建模，并且未充分利用CAD内核的确定性能力。为了解决这些局限性，我们提出了ReliCAD，一个将不确定的LLM生成转化为可靠的参数化CAD建模的统一框架。通过明确的设计意图建模，ReliCAD将用户需求转化为结构化设计规范，并明确建模几何关系、拓扑依赖和特征构建顺序。然后，它生成约束感知的参数化指令，并通过Agent-ready API调用CAD内核进行几何构建和约束求解。ReliCAD进一步记录运行时证据，并采用验证反馈机制来评估生成模型与设计规范之间的一致性，从而实现错误定位和迭代修复。在公共的HistCAD生成数据集和我们的多粒度CAD编辑数据集上的实验表明，ReliCAD显著优于基线方法，达到了99.8%的有效性率和0.8753的IoU。ReliCAD提供了一种可验证、可修复和可推广的自然语言交互CAD建模方法。
cs.RO / 10 / 2609.22332
AffordanceWAM: Affordance-Aware Joint World-Action Modeling for Robot Manipulation
AffordanceWAM：面向机器人操作的感知意识联合世界-动作建模
Abstract
Generalizable robot manipulation requires predicting how a scene will evolve, identifying where interactions are feasible, and determining how to act. Action-labeled robot videos directly supervise control but are costly and limited in diversity, whereas egocentric human videos capture diverse interactions but lack robot actions and differ in embodiment and appearance. We introduce AffordanceWAM, an affordance-aware generative World Action Model that represents object-centric spatiotemporal affordance through Scalar Affordance and Affordance Heatmap, within the generated future World. This representation grounds visual prediction in task-relevant objects and interaction regions for action generation, and provides shared interaction targets across human and robot videos. Built on a pretrained video diffusion Transformer, AffordanceWAM uses separately parameterized World and Action Experts, coupled through Masked Joint Self-Attention, to jointly predict future RGB observations, Scalar Affordance fields, Affordance Heatmaps, and continuous robot actions under a unified flow-matching objective. Human videos supervise all three future-World streams, whereas robot trajectories additionally provide action supervision, enabling transfer without human action labels or retargeting. Experiments on RoboCasa, CALVIN ABC$\rightarrow$D, and real-world manipulation demonstrate consistent gains over RGB-only and robot-data-only baselines. Under fixed robot supervision, RoboCasa performance improves monotonically as affordance-annotated human video scales. These results support affordance as an effective interface for both vision-language-action learning and human-to-robot transfer.
Chinese Translation
可推广的机器人操作需要预测场景如何演变，识别交互的可行性以及确定如何行动。带有动作标签的机器人视频直接监督控制，但成本高且多样性有限，而以自我为中心的人类视频则捕捉了多样的交互，但缺乏机器人动作，并且在体现和外观上存在差异。我们提出了AffordanceWAM，一种感知意识生成的世界动作模型，通过标量感知（Scalar Affordance）和感知热图（Affordance Heatmap）在生成的未来世界中表示以物体为中心的时空感知。这种表示将视觉预测与任务相关的对象和交互区域结合起来，以生成动作，并为人类和机器人视频提供共享的交互目标。AffordanceWAM基于预训练的视频扩散Transformer，使用单独参数化的世界和动作专家，通过掩蔽联合自注意力（Masked Joint Self-Attention）相结合，共同预测未来的RGB观测、标量感知场、感知热图和连续的机器人动作，遵循统一的流匹配目标。人类视频监督所有三个未来世界流，而机器人轨迹则额外提供动作监督，使得在没有人类动作标签或重新定向的情况下进行转移成为可能。在RoboCasa、CALVIN ABC$
ightarrow$D和真实世界操作的实验中，结果显示在RGB-only和机器人数据-only基线之上的一致增益。在固定机器人监督下，随着带有感知注释的人类视频规模的扩大，RoboCasa的性能单调提高。这些结果支持感知作为视觉-语言-动作学习和人类到机器人转移的有效接口。
cs.RO / 11 / 2609.22385
Active Spatial Inspection for Effective and Efficient Embodied Exploration
有效且高效的具身探索的主动空间检查
Abstract
Achieving high task success and efficiency remains a central pursuit in embodied exploration. Existing frameworks typically guide agent behavior through a spatially coarse and indirect assessment of suggestive cues and directions, yet such designs may struggle to judge cue sufficiency and the need for further inspection, leading to early termination or excessive continuation and ultimately reducing task success and efficiency. This work rethinks embodied exploration from a spatially explicit standpoint and introduces the spatial-inspection-guided ACtive Exploration (ACE) framework coupling evidence-grounded perception with exposure-informed movement and establishing a spatially resolved decision paradigm. Evidence-grounded perception strengthens inferential validity by integrating granular localization with focused verification, turning suggestive cues into precise decisive visual support. Exposure-informed movement promotes directional judgment through prospective prioritization and retrospective suppression, selecting promising directions for efficient spatial progress. ACE alleviates the tension between preventing early termination and avoiding unnecessary continuation. Extensive experiments demonstrate that ACE achieves 18.0% higher navigation task success and 10.3% higher exploration efficiency for question answering than prior state-of-the-art baselines, advancing effective and efficient embodied exploration.
Chinese Translation
在具身探索中，实现高任务成功率和效率仍然是一个核心追求。现有框架通常通过对提示线索和方向的空间粗略和间接评估来指导代理行为，但这种设计可能难以判断线索的充分性和进一步检查的必要性，从而导致过早终止或过度继续，最终降低任务成功率和效率。本研究从空间明确的角度重新思考具身探索，并引入了空间检查引导的主动探索（Active Exploration, ACE）框架，该框架将基于证据的感知与基于曝光的信息运动相结合，建立了一个空间分辨的决策范式。基于证据的感知通过将细粒度定位与聚焦验证相结合，增强了推理有效性，将提示线索转化为精确的决策视觉支持。基于曝光的信息运动通过前瞻性优先排序和回顾性抑制促进方向判断，为高效的空间进展选择有前途的方向。ACE缓解了防止过早终止与避免不必要继续之间的紧张关系。大量实验表明，ACE在问答任务中实现了比之前的最先进基线高出18.0%的导航任务成功率和高出10.3%的探索效率，推动了有效且高效的具身探索。
cs.RO / 12 / 2609.22404
Behavior Trees for Robotic Systems: An Empirical Study on Practices and Experiences
机器人系统中的行为树：实践与经验的实证研究
Abstract
Over the last decade, behavior trees (BT) have become one of the dominant behavior models for coordinating missions of robotic systems. Yet empirical evidence on BT adoption in real-world contexts remains limited, especially regarding practitioners experiences and practices. Without practitioner-grounded evidence from both academic and industrial settings, the research community risks developing guidelines and tools that are plausible in principle but only partly aligned with real challenges. This scarcity of evidence leads to ad-hoc practices, impeding software reuse, maintenance, and evolution. This paper reports a mixed-methods study combining a technical action research investigation at an automotive company with a survey of 34 robotics practitioners. Our results indicate that BTs improve team communication and the understandability of robotic decision-making logic, reflecting BTs practical value beyond mission coordination. At the same time, practitioners face multiple non-trivial design and integration decisions, complicated by the lack of adequate guidelines and tool support. These decisions span architectural and language choices when implementing BTs and integrating them within ROS, for which we report observed patterns. Additionally, practitioners faced multi-factor granularity decisions and mixed experiences with current libraries. For the most used BT libraries, they reported implementation challenges and documentation gaps. Regarding planning algorithms, practitioners find deciding optimal BT node ordering challenging. Views on scalability remained inconclusive, and limitations in current libraries and practices hinder broader adoption. We conclude with cross-cutting observations and implications for both practitioners adopting BTs in robotic systems and researchers aiming to advance empirical understanding of BT adoption.
Chinese Translation
在过去十年中，行为树（Behavior Trees, BT）已成为协调机器人系统任务的主要行为模型之一。然而，关于BT在现实世界中应用的实证证据仍然有限，尤其是在从业者的经验和实践方面。缺乏来自学术界和工业界的基于从业者的证据，研究社区面临着制定原则上合理但与实际挑战只部分一致的指南和工具的风险。这种证据的匮乏导致了临时性的实践，妨碍了软件的重用、维护和演化。本文报告了一项混合方法研究，结合了在一家汽车公司的技术行动研究调查与对34名机器人从业者的问卷调查。我们的结果表明，BT改善了团队沟通和机器人决策逻辑的可理解性，反映了BT在任务协调之外的实际价值。同时，从业者面临多种非平凡的设计和集成决策，这些决策因缺乏适当的指南和工具支持而变得复杂。这些决策涉及在实现BT和将其集成到机器人操作系统（Robot Operating System, ROS）时的架构和语言选择，我们报告了观察到的模式。此外，从业者还面临多因素粒度决策和对当前库的混合体验。对于最常用的BT库，他们报告了实施挑战和文档缺口。在规划算法方面，从业者发现确定最佳BT节点顺序具有挑战性。关于可扩展性的看法仍然不确定，当前库和实践的局限性阻碍了更广泛的采用。我们以跨领域的观察和对采用BT的机器人系统从业者及希望推进BT采用实证理解的研究者的启示作为结尾。
cs.RO / 13 / 2609.22462
VLPSA: Vision-Language-Poisson-Safe Actions for Full-Body Safety of Learned Policies
VLPSA：用于学习策略全身安全的视觉-语言-泊松-安全动作
Abstract
Vision-language-action (VLA) models enable increasingly general-purpose robotic manipulation, but such learned policies do not provide safety guarantees for collision avoidance---especially in environments outside of training distributions. This work presents Vision-Language-Poisson-Safe Actions (VLPSA), a safety filtering framework that provides full-body safety for VLA policies in cluttered and dynamic environments without retraining. VLPSA synthesizes Poisson Safety Functions (PSF) online from perception data, yielding a Control Barrier Function (CBF) that is enforced through a CBF-QP safety filter over the full body and any grasped object, treated as an extension of the final robot link. To enable real-time deployment while maintaining fine spatial resolution in critical task regions, VLPSA combines dual resolutions of this PSF using Boolean CBF compositions. We evaluate VLPSA on SafeLIBERO against safety-filtering baselines, where it achieves the highest collision avoidance rate among the evaluated methods, increasing collision avoidance from 23.1% for the base $\pi_{0.5}$ policy to 91.2% while surpassing its task success rate. We further deploy VLPSA on a Franka FR3 in cluttered scenes with dynamic obstacles and human interference, demonstrating real-time full-body safety during manipulation tasks.
Chinese Translation
视觉-语言-动作（VLA）模型使得机器人操作越来越具备通用性，但这些学习到的策略在避免碰撞方面并未提供安全保障，尤其是在训练分布之外的环境中。本文提出了视觉-语言-泊松-安全动作（VLPSA），这是一种安全过滤框架，能够在杂乱和动态环境中为VLA策略提供全身安全保障，而无需重新训练。VLPSA从感知数据中在线合成泊松安全函数（PSF），生成一个控制屏障函数（CBF），该函数通过CBF-QP安全过滤器在全身及任何抓取物体上施加约束，将抓取物体视为最终机器人链接的扩展。为了在保持关键任务区域的细致空间分辨率的同时实现实时部署，VLPSA结合了这种PSF的双重分辨率，使用布尔CBF组合。我们在SafeLIBERO上评估了VLPSA与安全过滤基线的表现，结果显示其在评估方法中实现了最高的碰撞避免率，将基础策略$ ext{π}_{0.5}$的碰撞避免率从23.1%提升至91.2%，同时超越了其任务成功率。此外，我们在杂乱场景中动态障碍物和人类干扰下对Franka FR3进行了VLPSA的部署，展示了在操作任务中的实时全身安全性。
cs.RO / 14 / 2609.22483
Tracker-Free Robotic Ultrasound Calibration with a Spherical-Marker Phantom and Threshold-Free Center Localization
无追踪器的机器人超声波校准：基于球形标记幻影和无阈值中心定位
Abstract
Robotic ultrasound (US) calibration is essential for accurately relating US images to the robot coordinate system, but accurate and automated calibration remains challenging because existing methods often require complex phantoms or external 3D trackers. In this work, we develop a tracker-free robotic US calibration framework using a spherical-marker phantom and a highly automated perception pipeline for sphere-center localization. The proposed threshold-free image-processing method localizes the spherical feature based on each US image's intensity distribution, eliminating hand-tuned intensity thresholds and improving robustness across imaging settings without per-system retuning. Multi-pose observations of the spherical fiducial are then used to estimate the US-to-EE transformation without external tracking or prior localization of the sphere center in the robot base frame. We validate the proposed framework on a robotic US platform through repeated sphere scans and further assess the calibrated system using geometrically distinct phantoms with known CAD models. Across 12 cross-validation folds, single-marker calibration achieved sphere-center accuracy and precision of $1.75\pm0.51$ mm and $0.85\pm0.11$ mm, respectively, compared with $1.72\pm0.51$ mm and $0.84\pm0.12$ mm for three-marker calibration. Across the three reconstruction sets, the single-marker calibration yielded pooled post-registration point-to-surface MAE$\pm$SD values of $0.41\pm0.35$ mm for the cone and $0.50\pm0.41$ mm for the triangular prism, closely matching the three-marker results of $0.40\pm0.35$ mm and $0.48\pm0.40$ mm, respectively.
Chinese Translation
机器人超声波（US）校准对于准确地将超声图像与机器人坐标系统关联至关重要，但由于现有方法通常需要复杂的幻影或外部三维追踪器，因此准确且自动化的校准仍然具有挑战性。在本研究中，我们开发了一种无追踪器的机器人超声波校准框架，使用球形标记幻影和高度自动化的感知管道进行球心定位。所提出的无阈值图像处理方法基于每个超声图像的强度分布来定位球形特征，消除了手动调整的强度阈值，并在不同成像设置下提高了鲁棒性，无需对每个系统进行重新调校。然后，使用球形基准的多姿态观测来估计超声到末端执行器（EE）的变换，无需外部追踪或在机器人基座框架中对球心的先前定位。我们通过重复的球体扫描在机器人超声平台上验证了所提出的框架，并进一步使用具有已知CAD模型的几何上不同的幻影评估校准系统。在12个交叉验证折中，单标记校准实现了球心的准确性和精度分别为$1.75 ext{±}0.51$ mm和$0.85 ext{±}0.11$ mm，而三标记校准的结果为$1.72 ext{±}0.51$ mm和$0.84 ext{±}0.12$ mm。在三个重建集上，单标记校准的注册后点到面平均绝对误差（MAE）$ ext{±}$标准差（SD）值为锥体$0.41 ext{±}0.35$ mm和三角棱柱$0.50 ext{±}0.41$ mm，分别与三标记结果$0.40 ext{±}0.35$ mm和$0.48 ext{±}0.40$ mm非常接近。
cs.RO / 15 / 2609.22493
Layered e-skin for Shear Sensing
层状电子皮肤用于剪切传感
Abstract
This paper presents a stacked two-layer force-sensing resistor (FSR) array designed for robotic fingertips that combines high-resolution pressure mapping with shear-force estimation. A compliant lattice elastomer spacer converts shear loading into a measurable inter-layer displacement, producing relative center-of-pressure (CoP) shifts between layers. A physics-based moment balance links inter-layer CoP displacement to shear force, while an end-to-end CNN--GRU model captures nonlinear effects from load-dependent compression and contact redistribution. This model, with both layers as input, achieves coefficients of determination $R^2 = 0.914$ for $F_x$ and $R^2 = 0.944$ for $F_y$, consistently outperforming single-layer baselines for shear-force estimation. Robotic manipulation experiments show that, for contact-motion tracking, the deep layer tracks the translation and rotation imposed by the robot arm, whereas the superficial layer tracks the slip at the contact surface. Transient changes in the difference between the total pressure responses of the two layers provide the best slip-event detection performance among the tested cues. These results demonstrate that two-layer FSR arrays can provide three-axis force estimation, contact-motion tracking, and slip-event detection beyond conventional normal-force sensing.
Chinese Translation
本文提出了一种为机器人手指设计的堆叠双层力传感电阻（FSR）阵列，该阵列结合了高分辨率的压力映射与剪切力估计。一个柔性晶格弹性体间隔器将剪切载荷转换为可测量的层间位移，从而在层之间产生相对的压力中心（CoP）移动。基于物理的力矩平衡将层间CoP位移与剪切力联系起来，而端到端的卷积神经网络-门控循环单元（CNN-GRU）模型则捕捉了负载依赖的压缩和接触重分布所带来的非线性效应。该模型以两个层作为输入，对于$F_x$达到了决定系数$R^2 = 0.914$，对于$F_y$达到了$R^2 = 0.944$，在剪切力估计上始终优于单层基准。机器人操作实验表明，在接触运动跟踪中，深层能够追踪机器人手臂施加的平移和旋转，而表层则跟踪接触表面的滑移。在测试的线索中，两层之间总压力响应差异的瞬时变化提供了最佳的滑移事件检测性能。这些结果表明，双层FSR阵列能够提供三轴力估计、接触运动跟踪和超越传统法向力传感的滑移事件检测。
cs.RO / 16 / 2609.22521
Latent Policy Steering: An Efficient and Flexible Framework for Cross-Embodiment Transfer
潜在策略引导：跨体现转移的高效灵活框架
Abstract
The performance of learned robot visuomotor policies depends heavily on the size and quality of their training data, yet collecting high-quality demonstrations remains costly for robots in the real world. Although large-scale robot and human datasets are increasingly available, embodiment gaps and mismatched action spaces make them difficult to leverage directly. Cross-embodiment transfer, reusing experience from other embodiments to improve learning on a target embodiment, is therefore crucial for scaling robot learning beyond per-robot data collection. In this work, we find that efficient transfer can be achieved by learning from what is shared across embodiments, the visual dynamics of how the world responds to motion, and by effectively exploiting the scarce target-embodiment data at test time. The proposed framework, called Latent Policy Steering (LPS), implements an embodiment-agnostic pretraining phase, which trains an image-based World Model (WM) with optical flow across diverse embodiments. The resulting WM is finetuned on the target embodiment with robot actions. It then steers the base policy toward better actions by searching in the WM's latent space for plans that stay close to the finetuning data. LPS is a policy-agnostic framework: it can flexibly accommodate different policies without having to retrain them. In Robomimic and real-world evaluations, LPS improves the average performance of Diffusion Policy relatively by 16% and 62%, and Pi0.5 by 8% and 14%, with only 50 demonstrations on an unseen target embodiment.
Chinese Translation
学习的机器人视觉运动策略的性能在很大程度上依赖于其训练数据的规模和质量，但在现实世界中收集高质量的示范仍然成本高昂。尽管大规模的机器人和人类数据集日益可用，但体现间的差距和不匹配的动作空间使得直接利用这些数据变得困难。因此，跨体现转移，即重用其他体现的经验以改善目标体现的学习，对于超越每个机器人数据收集的机器人学习至关重要。在本研究中，我们发现通过学习体现间共享的内容，即世界对运动的视觉动态，并在测试时有效利用稀缺的目标体现数据，可以实现高效的转移。我们提出的框架称为潜在策略引导（Latent Policy Steering, LPS），实现了一个与体现无关的预训练阶段，该阶段通过光流在不同体现中训练一个基于图像的世界模型（World Model, WM）。得到的WM在目标体现上通过机器人动作进行微调。然后，它通过在WM的潜在空间中搜索与微调数据接近的计划，来引导基础策略朝向更好的动作。LPS是一个与策略无关的框架：它可以灵活地适应不同的策略，而无需重新训练。在Robomimic和现实世界的评估中，LPS相对提高了扩散策略（Diffusion Policy）平均性能16%和62%，以及Pi0.5的8%和14%，仅需在一个未见的目标体现上提供50个示范。
cs.RO / 17 / 2609.22538
FRAMES: Failure Recovery And Monitoring of Embodied Skills for Humanoid Loco-Manipulation
FRAMES：人形机器人运动操作中具身技能的故障恢复与监控
Abstract
Large language model (LLM) planners can decompose natural-language instructions and select reusable robot skills, but choosing the correct skill does not guarantee successful physical execution. This gap is especially important in humanoid loco-manipulation, where errors during approach, grasping, transport, or placement can invalidate the remainder of a long-horizon plan. We present FRAMES, a failure-aware supervisory framework for the Unitree G1 humanoid that operates above the CEER whole-body controller. A Planner Agent selects subtasks through parameterized mid-level skills, while a vision-language-model-based Monitor Agent evaluates each skill using temporal multi-view observations and structured robot and contact evidence. Detected failures stop the active skill and provide grounded feedback to a Recovery Agent. The framework further includes a Memory Module for reusing prior skill experience, and geometric grounding via depth and segmentation. We independently evaluate the monitoring module of the framework in MuJoCo using 100 trials comprising 50 failed and 50 successful executions across five tasks. The monitor detects 48 of 50 failures, correctly accepts 46 of 50 successful executions, and achieves 94.0% overall accuracy. These results provide initial evidence for the monitoring component, while end-to-end evaluation of the complete recovery loop remains ongoing.
Chinese Translation
大型语言模型（LLM）规划器能够分解自然语言指令并选择可重用的机器人技能，但选择正确的技能并不保证能够成功执行。这一差距在 humanoid loco-manipulation（人形机器人运动操作）中尤为重要，因为在接近、抓取、运输或放置过程中出现的错误可能会使长时间计划的其余部分失效。我们提出了 FRAMES，一个针对 Unitree G1 人形机器人的故障感知监督框架，该框架在 CEER 全身控制器之上运行。规划代理通过参数化的中级技能选择子任务，而基于视觉语言模型的监控代理则使用时间多视角观察和结构化的机器人及接触证据来评估每个技能。检测到的故障会停止当前技能并为恢复代理提供基于事实的反馈。该框架还包括一个内存模块，用于重用先前的技能经验，以及通过深度和分割实现的几何基础。我们在 MuJoCo 中独立评估了该框架的监控模块，进行了 100 次试验，其中包括 50 次失败和 50 次成功的执行，涵盖五个任务。监控器检测到 50 次失败中的 48 次，正确接受 50 次成功执行中的 46 次，整体准确率达到 94.0%。这些结果为监控组件提供了初步证据，而完整恢复循环的端到端评估仍在进行中。
cs.RO / 18 / 2609.22587
React When You Need To: Event-Triggered Asynchronous Inference for VLA Policies
在需要时反应：事件触发的异步推理用于视觉-语言-动作（VLA）策略
Abstract
Vision-Language-Action (VLA) models commonly predict action chunks, limiting their ability to react to environmental changes during execution. Existing asynchronous inference methods improve reactivity but typically rely on a fixed inference gap. In this paper, we propose an event-guided dynamic inference strategy that adapts the inference gap according to scene changes observed since the previous inference. Thereby, it simultaneously preserves motion consistency and prompt reactivity. Across static and dynamic real-world settings, our method consistently performs best, averaging 95% success and exceeding the strongest baseline by 55 percentage points. The code will be made publicly available upon acceptance. The project page is available at https://react-when-you-need-to.github.io/.
Chinese Translation
视觉-语言-动作（VLA）模型通常预测动作片段，这限制了它们在执行过程中对环境变化的反应能力。现有的异步推理方法提高了反应能力，但通常依赖于固定的推理间隔。本文提出了一种事件引导的动态推理策略，该策略根据自上次推理以来观察到的场景变化调整推理间隔。通过这种方式，它同时保持了运动一致性和及时反应能力。在静态和动态的现实场景中，我们的方法始终表现最佳，平均成功率达到95%，超过最强基线55个百分点。代码将在接受后公开发布。项目页面可访问 https://react-when-you-need-to.github.io/。
cs.RO / 19 / 2609.22591
REBOOT: From Failure to Recovery - A Dataset and Benchmark for Precision Assembly
REBOOT：从失败到恢复 - 精密装配的数据集和基准
Abstract
Robot learning policies fail in characteristic ways: they stall in uncertain states, drift during contact-rich alignment, and miss targets by millimetres in precision tasks. Yet training datasets consist largely of successful demonstrations, while real-world benchmarks often reduce performance to binary success. This limits both supervision for recovery and analysis of where failures occur. We introduce REBOOT (Recovery Episode Benchmark for Off-nominal Trajectories), the first robot manipulation benchmark designed around failure as a first-class signal. REBOOT contains 2,160 demonstrations across 18 precision assembly tasks, each decomposed into five shared phases: Align(pick), Engage(pick), Transport, Align(place), and Engage(place), enabling phase-level evaluation beyond terminal success. Failures are introduced across phases and paired with expert recovery trajectories that return the system to a valid continuation state. Tasks are annotated with rotational symmetry, engagement-clearance precision tier, and assembly direction through matched install-remove pairs. Failure episodes are labeled by phase and categorical failure mode, enabling attribution to kinematic stage and tolerance violation. Data includes synchronized RGB-D observations from four viewpoints and grounded natural-language descriptions of phase-level success and failure conditions. Half the dataset contains expert demonstrations; the other half contains recovery demonstrations sampled to reflect failures observed in imitation-learned policy rollouts. We benchmark action-chunked transformer, diffusion, and $\pi_0$-FAST policies using phase-level completion rates, revealing model-specific failure points hidden by binary evaluation. Dataset and code: https://nanayawoa.github.io/REBOOT
Chinese Translation
机器人学习策略在特征上以特定方式失败：它们在不确定状态下停滞，在接触丰富的对齐过程中漂移，并在精密任务中以毫米为单位错过目标。然而，训练数据集主要由成功示范组成，而现实世界的基准往往将性能简化为二元成功。这限制了恢复的监督和对失败发生位置的分析。我们引入了REBOOT（非正常轨迹恢复情节基准），这是第一个围绕失败作为一类重要信号设计的机器人操作基准。REBOOT包含2,160个示范，涵盖18个精密装配任务，每个任务分解为五个共享阶段：对齐（取件）、接合（取件）、运输、对齐（放置）和接合（放置），使得能够进行超越终端成功的阶段级评估。失败在各个阶段中引入，并与专家恢复轨迹配对，这些轨迹将系统恢复到有效的继续状态。任务通过匹配的安装-移除对进行旋转对称性、接合-清除精度等级和装配方向的注释。失败情节按阶段和类别失败模式标记，使得能够归因于运动学阶段和公差违反。数据包括来自四个视角的同步RGB-D观测和阶段级成功与失败条件的基础自然语言描述。数据集的一半包含专家示范；另一半包含反映模仿学习策略回滚中观察到的失败的恢复示范。我们使用阶段级完成率对动作分块变换器、扩散和$ ext{π}_0$-FAST策略进行基准测试，揭示了被二元评估掩盖的特定模型失败点。数据集和代码：https://nanayawoa.github.io/REBOOT
cs.RO / 20 / 2609.22594
Characterizing Wildlife Response to Biomimetic and Conventional Underwater Vehicles
表征野生动物对仿生与传统水下车辆的反应
Abstract
Autonomous underwater vehicles (AUVs) are a promising alternative to divers for scalable collection of natural ocean ecology data. However, these robots may disturb local fauna and cause drastic behavioral differences compared to other monitoring techniques, decreasing their value as scientific tools. A promising prospect is to make AUVs that are more biomimetic, with the hope that taking on the form and behavior of a non-predatory animal may reduce adverse responses. We present the first dataset comparing fish disturbance in response to a conventional thruster-driven AUV, a sea turtle inspired flipper-driven AUV, and a diver. Experiments were conducted at a biodiversity hotspot in a Caribbean reef, and images of the scene were analyzed using computer vision to localize and study fish behavior change. Both AUVs cause measurable changes in fish behavior. Although no between-robot differences remain significant after correction for multiple comparisons, point estimates generally favor the biomimetic AUV, motivating larger studies capable of resolving modest effects and further design changes to optimize for disturbance. We also observe larger responses during diver transects than during AUV transects, although this exploratory comparison is based on a small diver sample with several confounds. While conclusions must be taken as preliminary due to operational and experimental limitations, this study provides the community with a first known dataset and benchmarks to quantify the behavioral impacts of biomimetic robots for ecological monitoring in the wild.
Chinese Translation
自主水下车辆（AUVs）是替代潜水员进行可扩展自然海洋生态数据收集的有前景的选择。然而，这些机器人可能会干扰当地动物群，并导致与其他监测技术相比显著的行为差异，从而降低其作为科学工具的价值。一个有希望的前景是制造更具仿生特征的AUVs，期望通过采用非掠食性动物的形态和行为来减少不良反应。我们展示了第一个数据集，比较了鱼类对传统推进器驱动的AUV、受海龟启发的鳍驱动AUV和潜水员的干扰反应。实验在加勒比海珊瑚礁的生物多样性热点进行，使用计算机视觉分析场景图像以定位和研究鱼类行为变化。两种AUV均导致鱼类行为发生可测量的变化。尽管在多重比较校正后，机器人之间的差异没有显著性，但点估计通常倾向于仿生AUV，激励进行更大规模的研究，以解决适度效应并进一步设计更优以减少干扰。我们还观察到，在潜水员横断面期间的反应大于在AUV横断面期间的反应，尽管这一探索性比较基于小样本的潜水员数据，存在多个混杂因素。由于操作和实验限制，结论必须被视为初步，但本研究为社区提供了第一个已知数据集和基准，以量化仿生机器人在野外生态监测中的行为影响。
cs.RO / 21 / 2609.22606
VT-Bridge: Bridging Pretrained Foundation VLAs to VTLAs via Lightweight Residual Adaptation
VT-Bridge：通过轻量级残差适配将预训练基础视觉语言行动模型连接到视觉触觉语言行动模型
Abstract
Vision-Tactile-Language-Action (VTLA) models have demonstrated clear advantages over Vision-Language-Action (VLA) models in contact-rich manipulation. However, developing VTLA models is severely constrained by the massive amounts of vision-tactile data and computational resources required. To address this bottleneck, we propose VT-Bridge, a lightweight residual adaptation strategy that bridges pretrained foundation VLAs to VTLAs. Rather than training a VTLA model from scratch or modifying the original architecture of a pretrained VLA, VT-Bridge employs an identical lightweight residual-adapter architecture across VLA backbones and uses backbone-specific weights to refine actions at the robot execution frequency. This design substantially lowers the data and training barriers. Specifically, it requires up to 50 vision-tactile demonstrations per task to fine-tune a VLA backbone and train a 0.98M-parameter residual adapter. Experiments with three representative VLA backbones ($\pi_0$, $\pi_{0.5}$, and SmolVLA) across four contact-rich manipulation tasks further demonstrate its consistent effectiveness across VLA architectures. On average, VT-Bridge raises the task completion rate from 11.7% with task-level VLA fine-tuning alone to 62.9%. Together, these findings demonstrate the broad applicability, effectiveness, and accessibility of VT-Bridge for contact-rich manipulation. The project page is available at https://hoxnocha.github.io/vt-bridge-web/.
Chinese Translation
视觉-触觉-语言-行动（VTLA）模型在接触丰富的操作中相较于视觉-语言-行动（VLA）模型显示出了明显的优势。然而，开发VTLA模型受到所需的海量视觉-触觉数据和计算资源的严重限制。为了解决这一瓶颈，我们提出了VT-Bridge，一种轻量级残差适配策略，将预训练的基础VLA模型与VTLA模型连接起来。VT-Bridge并不是从头开始训练VTLA模型或修改预训练VLA的原始架构，而是采用在VLA主干网络中相同的轻量级残差适配器架构，并使用特定于主干网络的权重在机器人执行频率下细化动作。这一设计显著降低了数据和训练的门槛。具体而言，它每个任务最多需要50个视觉-触觉演示来微调VLA主干网络并训练一个0.98M参数的残差适配器。对三个代表性VLA主干网络（$ ext{π}_0$、$ ext{π}_{0.5}$和SmolVLA）在四个接触丰富的操作任务中的实验进一步证明了其在VLA架构中的一致有效性。平均而言，VT-Bridge将仅通过任务级VLA微调的任务完成率从11.7%提高到62.9%。这些发现共同表明了VT-Bridge在接触丰富操作中的广泛适用性、有效性和可及性。项目页面可访问 https://hoxnocha.github.io/vt-bridge-web/。
cs.RO / 22 / 2609.22608
Becoming a Fruit Ninja: Real-Time Probabilistic Kinodynamic Planning for Manipulator Projectile Interception
成为水果忍者：用于操纵器投射物拦截的实时概率动力学规划
Abstract
Projectile interception is a challenging dynamic manipulation problem. Intercepting a thrown object with a robot arm requires reaching a point on the object's path as the object passes through it. Slicing also fixes the blade's velocity and orientation at contact. The goal is therefore a subset of the states of the robot and arrival times that moves as the object falls, and the arm must reach it within its actuator limits in milliseconds. We present FRUITNINJA, an anytime sampling-based planner that grows a tree on the GPU in batches toward the interception manifold. Each edge is an exact cubic whose travel time is found by a parallel search against the arm's dynamics, so every edge satisfies the actuator limits. Plans are ranked by a risk-aware objective over the uncertainty in the object's position and the arm's arrival time. We evaluate on a Franka Research 3 against six baselines in a calibrated real-time simulator, where FRUITNINJA cuts 96.7% of tosses in the open and 68.3% among five obstacles, versus the best baseline's 68.3% and 35.0% respectively.
Chinese Translation
投射物拦截是一个具有挑战性的动态操控问题。用机器人手臂拦截一个被投掷的物体需要在物体经过时到达其路径上的某一点。切割还要求刀刃在接触时固定其速度和方向。因此，目标是一个随物体下落而移动的机器人状态和到达时间的子集，手臂必须在毫秒内在其执行器限制内到达该目标。我们提出了FRUITNINJA，这是一种基于采样的规划器，能够在GPU上以批量方式向拦截流形生长树。每条边都是一个精确的三次曲线，其旅行时间通过与手臂动力学的并行搜索找到，因此每条边都满足执行器限制。计划根据物体位置的不确定性和手臂到达时间的风险意识目标进行排序。我们在一个经过校准的实时模拟器中对Franka Research 3进行了评估，与六个基线进行比较，其中FRUITNINJA在开放环境中切割了96.7%的投掷，在五个障碍物中切割了68.3%，而最佳基线分别为68.3%和35.0%。
cs.RO / 23 / 2609.22609
From Documented Strengths to Force Limits: Material-Informed Robotic Insertion for Construction Assembly
从文档化的强度到力极限：基于材料信息的机器人插入用于建筑组装
Abstract
Insertion is a fundamental operation in robotic construction assembly, where variations in material properties and assembly conditions make it difficult to select contact forces that complete the task without exceeding the assembly's capacity. Although construction documents encode engineering knowledge about materials and their conditions, translating this knowledge into load limits for a specific assembly remains difficult. This paper presents SAGE (Source-grounded Assembly Gating and Execution), a system that converts documented material evidence into capacity estimates for robotic insertion. SAGE restricts a large language model (LLM) to extracting tensile and compressive strengths from retrieved passages and tables and records their sources. A response model then interpolates offline finite element (FE) solutions to convert these strengths and the assembly conditions into axial load capacity. For fits with positive clearance, the estimated capacity sets the policy's axial force limit; for interference fits, it is compared with measured support demand to determine admission. On the primary benchmark, SAGE reduces mean capacity error from 80.65\% for direct LLM estimates based on the same evidence to 10.74\%. Without refitting, the mean error remains 8.00\% on 16 additional geometries. Under the assigned support release model, SAGE correctly classifies 59 of 62 scored simulation runs, with only conservative errors. In recorded xArm6 demonstrations, SAGE takes material documents as input and completes physical insertion in 9 of 13 trials. These results show that assigning document interpretation to the LLM and force calculation to an explicit mechanical model produces accurate capacity estimates and traceable insertion decisions.
Chinese Translation
插入是机器人建筑组装中的一项基本操作，材料属性和组装条件的变化使得选择合适的接触力以完成任务而不超过组装的承载能力变得困难。尽管建筑文档编码了关于材料及其条件的工程知识，但将这些知识转化为特定组装的载荷极限仍然具有挑战性。本文提出了SAGE（基于源的组装门控与执行）系统，该系统将文档化的材料证据转化为机器人插入的承载能力估计。SAGE限制大型语言模型（LLM）从检索的段落和表格中提取拉伸和压缩强度，并记录其来源。响应模型随后插值离线有限元（FE）解决方案，将这些强度和组装条件转换为轴向载荷能力。对于正间隙配合，估计的能力设置政策的轴向力极限；对于干涉配合，则与测得的支撑需求进行比较以确定是否允许插入。在主要基准测试中，SAGE将基于相同证据的直接LLM估计的平均能力误差从80.65%降低至10.74%。在16个额外几何体上，未重新调整的平均误差保持在8.00%。在指定的支撑释放模型下，SAGE正确分类了62个评分模拟运行中的59个，仅出现保守性错误。在记录的xArm6演示中，SAGE以材料文档作为输入，在13次试验中完成了9次物理插入。这些结果表明，将文档解释分配给LLM并将力计算分配给明确的机械模型可以产生准确的承载能力估计和可追溯的插入决策。
cs.RO / 24 / 2609.22611
HIGenNTO: Scalable Humanoid Interaction Generation via Noise-Space Trajectory Optimization
HIGenNTO：通过噪声空间轨迹优化实现可扩展的人形交互生成
Abstract
Humanoid robots can acquire complex skills by imitating kinematic humanoid motion references, yet reliable references for contact-rich interactions remain difficult to obtain: motion capture deteriorates under occlusion and close physical contact, while retargeting introduces additional contact and geometric inconsistencies. We present HIGenNTO, a framework that synthesizes humanoid-scene interaction motion references by optimizing the initial noise of a pretrained text-conditioned motion model under sparse spatiotemporal and scene constraints. The same formulation satisfies desired contacts, avoids collisions, and maintains stable support while retaining the prior's realism and temporal coherence, generating interaction motions from scratch and composing long-horizon behaviors stage-wise. Across robot-environment and robot-object tasks, HIGenNTO produces motions that can be executed by tracking policies in simulation and used to train depth-conditioned visuomotor policies operating solely from onboard sensing. We deploy these policies on a Unitree G1 across four contact-rich tasks. Finally, the task specifications themselves can be written by a coding agent, which proposes interaction tasks and compiles them into prompt, constraint, and scene programs, authoring three of our eight evaluated tasks and four further behaviors. Together, these results establish a scalable path from high-level task descriptions to physically executable humanoid interactions.
Chinese Translation
人形机器人可以通过模仿运动学人形运动参考来获得复杂技能，但对于接触丰富的交互，可靠的参考仍然难以获得：运动捕捉在遮挡和近距离物理接触下会恶化，而重定向则引入额外的接触和几何不一致性。我们提出了HIGenNTO，一个通过在稀疏时空和场景约束下优化预训练的文本条件运动模型的初始噪声来合成人形-场景交互运动参考的框架。相同的公式满足所需的接触，避免碰撞，并在保持先验的现实性和时间一致性的同时维持稳定的支撑，从零开始生成交互运动，并分阶段组合长时间行为。在机器人-环境和机器人-物体任务中，HIGenNTO生成的运动可以通过跟踪策略在仿真中执行，并用于训练仅依赖于车载传感器的深度条件视觉运动策略。我们在Unitree G1上部署这些策略，完成四个接触丰富的任务。最后，任务规范本身可以由编码代理编写，该代理提出交互任务并将其编译成提示、约束和场景程序，创作了我们评估的八个任务中的三个和四个进一步的行为。综合来看，这些结果建立了一条从高层任务描述到物理可执行的人形交互的可扩展路径。
cs.RO / 25 / 2609.22630
Simultaneous Forward and Inverse Human-in-the-Loop Optimization
同时进行的正向与反向人机协同优化
Abstract
Subjective user experience is important to human-robot interaction, but the outcomes users value, and how those preferences vary across individuals and contexts, are often unknown. While inverse learning approaches using human data can help identify user rewards, in many assistive settings the experimental costs of executing a control policy, measuring biomechanical or physiological outcomes, and collecting user feedback often limit the number of queries and optimization iterations. Here, we present Simultaneous Forward and Inverse Human-In-the-Loop Optimization (SFIHILO), which efficiently infers individual-specific reward functions from preferences over human outcomes and identifies a final control policy that maximizes the learned reward. SFIHILO bootstraps a forward model to predict user outcomes from control policies, uses this model for active querying to accelerate inverse reward learning, and then optimizes the final control policy without additional user trials. We score candidate policies by their expected reduction in uncertainty across both forward and inverse beliefs, targeting a regional preference boundary to robustly inform this simultaneous learning process. In simulation, we show that SFIHILO was effective across user heterogeneity, outcome dimensionalities, precision requirements, noise levels, and nonstationarity; compared with mutual information approaches, the proposed active querying strategy significantly improved sample efficiency in inverse learning while preserving forward model accuracy. This approach demonstrates the potential to infer latent human goals, enabling more transferable and effective human-robot interaction.
Chinese Translation
主观用户体验对人机交互至关重要，但用户重视的结果以及这些偏好在个体和情境中的变化往往未知。虽然使用人类数据的反向学习方法可以帮助识别用户奖励，但在许多辅助设置中，执行控制策略、测量生物力学或生理结果以及收集用户反馈的实验成本常常限制了查询和优化迭代的数量。在此，我们提出了同时进行的正向与反向人机协同优化（SFIHILO），该方法有效地从对人类结果的偏好中推断个体特定的奖励函数，并识别出最大化学习奖励的最终控制策略。SFIHILO 利用正向模型预测用户结果，并利用该模型进行主动查询以加速反向奖励学习，然后在不需要额外用户试验的情况下优化最终控制策略。我们通过评估候选策略在正向和反向信念下的不确定性预期减少来为其评分，针对区域偏好边界以稳健地指导这一同时学习过程。在仿真中，我们展示了 SFIHILO 在用户异质性、结果维度、精度要求、噪声水平和非平稳性方面的有效性；与互信息方法相比，所提出的主动查询策略显著提高了反向学习中的样本效率，同时保持了正向模型的准确性。这一方法展示了推断潜在人类目标的潜力，从而实现更具可转移性和有效的人机交互。
cs.RO / 26 / 2609.22668
Density-Driven Area Coverage for Nonholonomic Multi-Robot Systems with Safety Guarantee
具有安全保障的非完整多机器人系统的密度驱动区域覆盖
Abstract
Density-Driven Optimal Control (D2OC) provides a principled approach to distributing multi-robot teams over non-uniform spatial distributions. Applying D2OC to nonholonomic robots, however, creates a gap between safety constraints imposed on a reference motion and the physical inputs that determine the actual robot motion. We address this issue by enforcing the safety constraint directly on the robot's physical inputs while preserving the density-driven coverage objective. The proposed framework combines D2OC with a control barrier function safety filter through a feedback-linearizing look-ahead point, allowing safety and actuator limits to be considered together during control. We further derive a safety margin that accounts for the look-ahead geometry, robot footprint, and motion during each control interval. Simulation results show that the proposed method maintains the required physical separation while achieving coverage performance comparable to a conventional reference-tracking approach, which can satisfy safety on the reference motion yet violate the corresponding physical clearance. Experiments on multiple nonholonomic robots in the Robotarium further demonstrate safe execution while driving the robots toward the desired spatial distribution. These results show that enforcing safety directly on the physical inputs can eliminate the mismatch between safety certification and physical robot motion in density-driven multi-robot coverage.
Chinese Translation
密度驱动最优控制（Density-Driven Optimal Control, D2OC）为在非均匀空间分布中分配多机器人团队提供了一种原则性的方法。然而，将 D2OC 应用于非完整机器人时，会在施加于参考运动的安全约束与决定实际机器人运动的物理输入之间产生差距。我们通过直接对机器人的物理输入施加安全约束，同时保持密度驱动的覆盖目标，来解决这一问题。所提出的框架结合了 D2OC 和控制屏障函数安全滤波器，通过反馈线性化的前视点，使安全性和执行器限制在控制过程中能够共同考虑。我们进一步推导出一个安全边际，该边际考虑了前视几何、机器人足迹以及每个控制间隔内的运动。仿真结果表明，所提方法在实现与传统参考跟踪方法相当的覆盖性能的同时，保持了所需的物理间隔，后者可以在参考运动上满足安全性，但可能违反相应的物理间隙。在 Robotarium 上对多个非完整机器人的实验进一步证明了在将机器人驱动向期望空间分布时的安全执行。这些结果表明，直接对物理输入施加安全性可以消除密度驱动多机器人覆盖中安全认证与物理机器人运动之间的不匹配。
cs.RO / 27 / 2609.22670
AquaWorld: Structure-Consistent Underwater World Generation for Robot Simulation
AquaWorld：用于机器人仿真的结构一致性水下世界生成
Abstract
Underwater robot simulation requires diverse environments in which terrain, scene composition, tasks, and currents remain mutually consistent. We present AquaWorld, a world-generation framework that preserves these relationships through shared terrain structure. A language-conditioned plan generates 3D terrain and shared structural references that guide asset placement, task definition, and inflow specification under stochastic variation. The framework incorporates over 10,000 underwater-compatible assets, predicts reusable terrain-conditioned mean-flow fields through a CFD-supervised residual model, and supports conventional underwater vehicles and bio-inspired robotic fish. On 24 paired terrains, structure-consistent randomization produces substantially better cross-factor consistency than independent randomization. In a matched-budget policy-training comparison, structurally coherent randomization achieves a validation success rate 21% higher than independent randomization. In separate physical experiments, a simulation-trained visual navigation policy succeeds in 95% of physical tank trials without updating its perception or control modules. Overall, AquaWorld provides a practical way to generate varied underwater environments while retaining the structural relationships needed for flow simulation and robot learning.
Chinese Translation
水下机器人仿真需要多样化的环境，其中地形、场景构成、任务和水流保持相互一致。我们提出了AquaWorld，一个通过共享地形结构保持这些关系的世界生成框架。该框架通过语言条件规划生成3D地形和共享结构参考，指导资产放置、任务定义和在随机变化下的流入规范。该框架包含超过10,000个水下兼容资产，通过CFD（计算流体动力学）监督的残差模型预测可重用的地形条件均匀流场，并支持传统水下车辆和仿生机器人鱼。在24对地形中，结构一致性的随机化在跨因素一致性方面显著优于独立随机化。在匹配预算的政策训练比较中，结构一致的随机化实现了比独立随机化高出21%的验证成功率。在独立的物理实验中，经过仿真训练的视觉导航策略在95%的物理水槽试验中成功，而无需更新其感知或控制模块。总体而言，AquaWorld提供了一种实用的方法来生成多样化的水下环境，同时保持流动仿真和机器人学习所需的结构关系。
cs.RO / 28 / 2609.22677
SHAFT: A Slack-Compensating, Helical-Buckling-Attenuating Flexible-Shaft Transmission for Lightweight Multi-DoF Manipulation
SHAFT：一种用于轻量级多自由度操作的松弛补偿螺旋屈曲衰减柔性轴传动
Abstract
Lightweight and slim manipulators enable safe operation in human living environments. Proximal actuation using remote transmission mechanisms, such as wire-driven or Bowden cables, effectively reduces inertia and arm size by relocating motors near the base and transmitting torque to distal joints. Existing approaches either increase mass through additional components, such as pulleys for direction changes, or suffer from reduced transmission efficiency due to friction losses. Flexible shaft transmission avoids both mass increase and excessive friction losses, but faces increasing angular transmission error due to helical buckling caused by slack generated at joint bending. To address this problem, we propose SHAFT: a Slack-compensating, Helical-buckling-Attenuating Flexible- shaft Transmission mechanism. This mechanism compensates for slack through a proximal tensioner, improving the angular transmission error and efficiency of flexible shaft transmission without increasing the moving mass of the arm section. In a transmission path containing four 90-degree bends, the proposed mechanism demonstrated approximately 30% higher efficiency and approximately 65% lower angular transmission error compared to a flexible shaft transmission without a tensioner. Using this mechanism, we fabricated a 6-Degree-of-Freedom (DoF) arm with a 1-DoF gripper manipulator consisting of a rotary module housing motors with a tensioner, and a 280 g weight for the arm module. The proposed manipulator represents a novel remote actuation system for achieving lightweight construction with high efficiency, contributing to the acceleration of safe robot deployment in human environments.
Chinese Translation
轻量化和纤细的操控器能够在人的生活环境中安全操作。通过使用远程传动机制（如线驱动或博登电缆）进行近端驱动，有效地通过将电机放置在基座附近并将扭矩传递到远端关节来减少惯性和臂的尺寸。现有方法要么通过增加额外组件（如用于方向改变的滑轮）来增加质量，要么由于摩擦损失而降低传动效率。柔性轴传动避免了质量增加和过度摩擦损失，但由于关节弯曲产生的松弛导致的螺旋屈曲，面临着日益增加的角传输误差。为了解决这个问题，我们提出了SHAFT：一种松弛补偿螺旋屈曲衰减柔性轴传动机制。该机制通过近端张紧器补偿松弛，改善了柔性轴传动的角传输误差和效率，而不增加臂段的移动质量。在包含四个90度弯曲的传动路径中，所提出的机制显示出比没有张紧器的柔性轴传动高出约30%的效率和约65%更低的角传输误差。利用该机制，我们制造了一个具有1个自由度夹持器的6自由度（DoF）臂，该夹持器由一个包含电机和张紧器的旋转模块构成，臂模块的重量为280克。所提出的操控器代表了一种新型的远程驱动系统，能够实现轻量化构造与高效率，为加速安全机器人在人的环境中的部署做出了贡献。
cs.RO / 29 / 2609.22681
A Direct Rigid Transmission 2-DoF Wrist Extension for Tendon-Driven Hand
一种直接刚性传动的2自由度腕部扩展装置用于腱驱动手
Abstract
Dexterous manipulation in confined spaces requires local control of hand orientation. Without a wrist, a dexterous hand must obtain this local orientation through coordinated motion of the robot arm, often involving several joints and a more complex end-effector path. We present CRAFT-Wrist, a concentric 2-DoF wrist extension that mounts between a robot arm and the CRAFT Hand without modifying the hand. Two XC430-T240BB-T servos drive sideways and front-back rotation through short rigid transmissions. Our initial hardware prototype actuates both axes under load, achieving a demonstrated workspace of $\pm20^{\circ}$ in radial--ulnar deviation (left--right) and a front-back range of $+80^{\circ}$ in flexion and $-18^{\circ}$ in extension. We characterize the resulting finger-motor loading across several wrist postures and demonstrate how the wrist supplies local orientation in grasping, nail hammering, and blackboard wiping. Demonstrations and assembly instructions are available at https://craft-wrist.github.io/.
Chinese Translation
在有限空间内进行灵巧操作需要对手部方向进行局部控制。没有腕部的情况下，灵巧手必须通过机器人手臂的协调运动来获得这种局部方向，通常涉及多个关节和更复杂的末端执行器路径。我们提出了CRAFT-Wrist，一种同心的2自由度腕部扩展装置，安装在机器人手臂与CRAFT手之间，无需对手进行修改。两个XC430-T240BB-T伺服电机通过短刚性传动驱动侧向和前后旋转。我们的初步硬件原型在负载下驱动两个轴，展示了在径向-尺偏移（左右）方面的工作空间为$m{ ext{±20}^{ ext{°}}}$，以及在屈曲方面的前后范围为$+80^{ ext{°}}$，在伸展方面为$-18^{ ext{°}}$。我们对不同腕部姿态下的手指电机负载进行了表征，并展示了腕部如何在抓取、钉钉子和擦黑板时提供局部方向。演示和组装说明可在https://craft-wrist.github.io/获取。
cs.RO / 30 / 2609.22684
StateMem: Single-State Residual Memory with Adaptive Inference for Vision-Language-Action Policies
StateMem：具有自适应推理的单状态残差记忆用于视觉-语言-动作策略
Abstract
Memory-dependent robotic manipulation often requires later actions to use information from earlier interactions. Existing vision-language-action (VLA) policies primarily rely on current observations, limiting historical information retention. Memory-augmented VLAs, such as MemoryVLA, address this limitation with external memory banks but require explicit storage and retrieval. To address these limitations, we propose StateMem, a single-state residual memory framework for VLA policies that uses prediction error to update a persistent memory token through low-rank residuals and to adaptively route cached prefixes. A training-free controller adjusts the routing threshold online, while fast correction compensates for stale prefix features during cache reuse. We evaluate StateMem on LIBERO, RoboMemArena, and real-world manipulation tasks. On LIBERO, StateMem achieves an average success rate of 97.6% and reduces the average VLM prefix refresh rate by 20.25% relative to full refresh. In the Occlusion category of RoboMemArena, StateMem achieves the best performance among single-VLA methods, reaching 21.8% Task Success Rate (TSR) and 44.3% Cumulative Success Rate (CSR). Across six real-world manipulation tasks, it achieves +21% in average success rate.
Chinese Translation
依赖记忆的机器人操控通常需要后续动作利用早期交互中的信息。现有的视觉-语言-动作（VLA）策略主要依赖当前观察，限制了历史信息的保留。增强记忆的VLA，例如MemoryVLA，通过外部记忆库解决了这一限制，但需要显式的存储和检索。为了解决这些限制，我们提出了StateMem，一个用于VLA策略的单状态残差记忆框架，该框架利用预测误差通过低秩残差更新持久记忆标记，并自适应地路由缓存前缀。一个无训练的控制器在线调整路由阈值，而快速校正在缓存重用期间补偿过时的前缀特征。我们在LIBERO、RoboMemArena和真实世界的操控任务上评估了StateMem。在LIBERO上，StateMem实现了97.6%的平均成功率，并将平均VLM前缀刷新率相对于完全刷新降低了20.25%。在RoboMemArena的遮挡类别中，StateMem在单VLA方法中表现最佳，达到了21.8%的任务成功率（TSR）和44.3%的累积成功率（CSR）。在六个真实世界的操控任务中，它的平均成功率提高了21%。
cs.RO / 31 / 2609.22726
Decentralized Multi-Robot Exploration with Probabilistic Peer Intent and Multi-hop Plan Propagation
基于概率同行意图和多跳计划传播的去中心化多机器人探索
Abstract
Efficient coordination under limited communication remains a key challenge in decentralized multi-robot exploration. While centralized approaches benefit from global information sharing, they are often impractical in large-scale or communication-constrained environments. Existing Monte Carlo Tree Search (MCTS)-based approaches, such as Decentralized Monte Carlo Exploration (DMCE), enable decentralized planning by taking peer intent into account. This peer intent is obtained by communicating sequences of planned waypoints with robots within direct communication range. In this work, we extend this idea by introducing Probabilistic Peer Intent (PPI), which converts peer trajectories into a continuous spatial representation of predicted intent and incorporates it into local MCTS action evaluation. We additionally study the effects of sharing peer intent beyond direct communication range by propagating plans over multiple hops. Experiments across multiple simulated environments and team sizes show that PPI and Multi-hop propagation can each improve decentralized exploration, with their relative benefits depending on environment structure and team size. We also demonstrate the real-world deployment of our method on three robots operating in different environment types.
Chinese Translation
在有限通信条件下，高效协调仍然是去中心化多机器人探索中的一个关键挑战。虽然中心化方法受益于全球信息共享，但在大规模或通信受限的环境中往往不切实际。现有的基于蒙特卡洛树搜索（Monte Carlo Tree Search, MCTS）的方法，如去中心化蒙特卡洛探索（Decentralized Monte Carlo Exploration, DMCE），通过考虑同行意图实现去中心化规划。这种同行意图是通过与直接通信范围内的机器人交流计划路径点序列获得的。在本研究中，我们通过引入概率同行意图（Probabilistic Peer Intent, PPI）扩展了这一理念，该方法将同行轨迹转换为预测意图的连续空间表示，并将其纳入局部MCTS动作评估中。我们还研究了在超出直接通信范围的情况下共享同行意图的效果，通过多跳传播计划。多个模拟环境和团队规模的实验表明，PPI和多跳传播均能改善去中心化探索，其相对效益依赖于环境结构和团队规模。我们还展示了我们的方法在三台机器人在不同环境类型中实际部署的效果。
cs.RO / 32 / 2609.22730
BEACON: Belief-Enabled Adaptive CONtrol for Imitation Learning under Uncertainty
BEACON：基于信念的自适应控制在不确定性下的模仿学习
Abstract
Robot manipulation tasks often involve hidden state information that cannot be directly observed and must be inferred through sequential physical interactions. In such partially observable settings, conditioning an imitation learning policy directly on the recent raw observation history leads to poor performance. This is due to state aliasing, wherein identical observations may arise from different hidden states, and the policy receives conflicting action labels for the same input. To enable history-aware disambiguation capability, we propose conditioning a diffusion policy on a structured representation of the hidden states using Bayesian belief that exposes both the current most likely state estimate and the remaining uncertainty. This representation replaces raw history with a structured, compact input, enabling the policy to implicitly modulate between exploratory and exploitative behaviors based on belief uncertainty, without explicit mode switching or reward shaping. We evaluate across two domains with qualitatively different belief representations: a continuous belief for cornstalk gripper alignment via tactile sensing, and a discrete categorical distribution for latched door opening. In both domains, the belief-conditioned policy substantially outperforms the observation-only baseline and approaches privileged ground-truth performance, with ablations illustrating that the policy adapts its exploration behavior depending on the belief uncertainty at inference time.
Chinese Translation
机器人操作任务通常涉及无法直接观察的隐藏状态信息，这些信息必须通过连续的物理交互进行推断。在这种部分可观察的环境中，直接将模仿学习策略条件化于最近的原始观察历史会导致性能不佳。这是由于状态混淆，即相同的观察可能源于不同的隐藏状态，从而导致策略对相同输入接收到相互矛盾的动作标签。为了实现历史感知的消歧能力，我们提出将扩散策略条件化于隐藏状态的结构化表示，利用贝叶斯信念来揭示当前最可能的状态估计和剩余的不确定性。该表示用结构化、紧凑的输入替代原始历史，使得策略能够根据信念的不确定性隐式调节探索与利用行为，而无需显式的模式切换或奖励塑形。我们在两个具有不同信念表示的领域中进行了评估：通过触觉感知进行玉米杆夹具对齐的连续信念，以及用于打开锁定门的离散类别分布。在这两个领域中，基于信念的条件策略显著优于仅基于观察的基线，并接近特权真实性能，消融实验表明策略根据推理时的信念不确定性调整其探索行为。
cs.RO / 33 / 2609.22733
Rapidly-Iterating Grid-Based Near-Optimal Kinodynamic Motion Planning
快速迭代的基于网格的近似最优运动规划
Abstract
This paper develops the Rapidly-iterating kinoDynamic Grid (RDG) algorithm, an asymptotically near-optimal kinodynamic motion planning algorithm that produces high quality solutions through rapid iteration. The algorithm leverages a state space grid decomposition to perform node selection, dynamics propagation, and graph revision in constant time complexity with respect to the number of nodes in the trajectory tree. Through a covering ball sequence induction proof, the algorithm is shown to be asymptotically near-optimal and probabilistically complete. Different subsystems of the algorithm are evaluated against common nearest-neighbor search-based methods at generating exploration bias. The RDG algorithm is evaluated through simulated trials in complex, kinodynamic motion planning problem environments up to 10 DOF relative to similar sparse, kinodynamic planning algorithms with optimality guarantees, SST and DIRT. The RDG algorithm outperforms both SST and DIRT in mean final solution quality by up to 104% and 40% respectively. Additionally, RDG maintained a 100% success rate, even on a 10-DOF test case where both SST and DIRT did not.
Chinese Translation
本文开发了一种快速迭代的运动动态网格（Rapidly-iterating kinoDynamic Grid, RDG）算法，这是一种渐近近似最优的运动规划算法，通过快速迭代生成高质量的解决方案。该算法利用状态空间网格分解，在与轨迹树中节点数量相关的常数时间复杂度内执行节点选择、动力学传播和图形修订。通过覆盖球序列归纳证明，该算法被证明是渐近近似最优的，并且在概率上是完全的。算法的不同子系统在生成探索偏差方面与常见的基于最近邻搜索的方法进行了评估。RDG算法通过在复杂的运动规划问题环境中进行模拟试验进行了评估，相较于具有最优性保证的稀疏运动规划算法SST和DIRT，在高达10自由度（DOF）的情况下，RDG算法在最终解决方案质量的平均值上分别超越了SST和DIRT，提升幅度达到104%和40%。此外，RDG在10-DOF测试案例中保持了100%的成功率，而SST和DIRT则未能成功。
cs.RO / 34 / 2609.22786
Distributed Infrastructure Sensors and Cues for Robotic Fire-Fighting and Safety System on a Lunar Base
月球基地的分布式基础设施传感器与机器人消防安全系统
Abstract
There are renewed efforts to build a lunar base and house a team of astronauts on the Moon for extended periods. However, important challenges remain, such as a rapid-response fire-hazard management system. In the event of a fire or toxic leak inside an early crewed lunar habitat, a single pressurized cylinder must be handled within minutes. However, suited astronauts and radio-delayed support from Earth cannot guarantee such response times. We describe a prototype fire-fighting and emergency response safety technology that embeds intelligence in the habitat itself: using thin ceramic QR codes affixed every few meters, each carries a local floor map, safe sensor limits, and step-by-step hazard responses. A low-power rover decodes a plate in a matter of seconds, polls the adjoining temperature-gas sensor cluster, and acts immediately, eliminating the need for a global map . In a representative layout, the interior is divided into roughly one code per two square meters. Simulations show that rapid-response firefighting both simplifies the task and avoids costly resource use and indeterminate outcomes. Because knowledge is spread across passive plates, damage is localized, procedures can be updated by replacing a single code, and processor demands remain minimal. However, important challenges remain, such as a rapid-response fire-hazard management system. In the event of a fire or toxic leak inside an early crewed lunar habitat, a single pressurized cylinder must be handled within minutes. However, suited astronauts and radio-delayed support from Earth cannot guarantee such response times. We describe a prototype fire-fighting and emergency response safety technology that embeds intelligence in the habitat itself: using thin ceramic QR codes affixed every few meters, each carries a local floor map, safe sensor limits, and step-by-step hazard responses.
Chinese Translation
目前正在重新努力建设月球基地，并在月球上长期安置一支宇航员团队。然而，仍然存在一些重要挑战，例如快速响应的火灾危险管理系统。在早期有人居住的月球栖息地内发生火灾或有毒泄漏时，必须在几分钟内处理一个加压气瓶。然而，穿着宇航服的宇航员和来自地球的延迟支持无法保证这样的响应时间。我们描述了一种原型消防和应急响应安全技术，该技术将智能嵌入栖息地本身：使用每隔几米粘贴的薄陶瓷二维码，每个二维码携带本地楼层地图、安全传感器限值和逐步的危险响应。低功耗探测车在几秒钟内解码一个二维码，查询相邻的温度-气体传感器集群，并立即采取行动，消除了对全球地图的需求。在一个典型布局中，内部大约每两平方米设置一个二维码。模拟结果表明，快速响应的消防不仅简化了任务，还避免了昂贵的资源使用和不确定的结果。由于知识分布在被动板上，损害被局部化，程序可以通过更换单个二维码进行更新，处理器需求保持在最低水平。
cs.RO / 35 / 2609.22795
Task-Oriented Co-Design and Optimization of Geared Actuators for Robotic Applications
面向任务的机器人应用齿轮驱动器的协同设计与优化
Abstract
Different tasks performed by legged robots impose distinct torque and speed requirements on actuators. Existing robotic actuators are generally optimized at the component level for metrics such as torque or power density, without explicit task guidance. System-level optimization across components such as motors, gearboxes, and sensors is challenging because of the high computational cost and coupling among mechanical, electrical, and electromagnetic behaviors. Consequently, improvements in individual components may not translate into better robot performance in a specific task. To this end, we present a systematic optimization framework for task-oriented co-design of actuator hardware and control. First, surrogate models are employed to accelerate motor evaluation and support global exploration of the coupled design space. Then, a hierarchical mixed-variable optimization strategy is adopted, combining discrete enumeration with continuous search over dimensions and real-valued indices. These indices are rounded to select admissible values for the remaining discrete choices before each evaluation. Within this search, rated output torque density and task performance are jointly optimized, with Bezier-parameterized joint torque profiles determined for each hardware candidate. Finally, the effectiveness of the proposed framework is validated through actuator fabrication and experiments on a two-degree-of-freedom jumping leg. Based on its measured mass, the fabricated prototype achieves a nominal rated output torque density of 35.7 N m/kg, approximately 60% higher than that of a widely used commercial geared joint actuator, while being 18.6% lighter. Under matched bench conditions, it achieves 12.0% greater jump height at twice-rated torque. Together, these results demonstrate a systematic route from task requirements to actuator design and control.
Chinese Translation
不同的步态机器人执行的任务对驱动器施加了不同的扭矩和速度要求。现有的机器人驱动器通常在组件层面上针对扭矩或功率密度等指标进行优化，而没有明确的任务指导。由于机械、电气和电磁行为之间的高度耦合以及高计算成本，跨组件（如电机、齿轮箱和传感器）的系统级优化面临挑战。因此，单个组件的改进可能无法转化为特定任务中的更好机器人性能。为此，我们提出了一种系统的优化框架，用于驱动器硬件和控制的面向任务的协同设计。首先，采用代理模型加速电机评估，并支持耦合设计空间的全局探索。然后，采用分层混合变量优化策略，将离散枚举与连续搜索相结合，针对维度和实值指标进行优化。在每次评估之前，这些指标会被四舍五入以选择剩余离散选择的可接受值。在此搜索过程中，额定输出扭矩密度和任务性能被共同优化，为每个硬件候选者确定贝塞尔参数化的关节扭矩曲线。最后，通过驱动器的制造和在一个二自由度跳跃腿上的实验验证了所提框架的有效性。基于其测得的质量，制造的原型实现了35.7 N m/kg的额定输出扭矩密度，约比广泛使用的商业齿轮关节驱动器高出60%，同时轻了18.6%。在匹配的基准条件下，它在额定扭矩的两倍下实现了12.0%的更高跳跃高度。这些结果共同展示了从任务需求到驱动器设计与控制的系统化路径。
cs.RO / 36 / 2609.22798
Expert-Play Contouring Control: Faster-than-Demonstration Planning from Slow Expert and Fast Play
专家-游戏轮廓控制：从慢速专家和快速游戏中进行快于演示的规划
Abstract
Expert demonstrations often specify what a robot should do, but not how fast it can do it. Imitation Learning (IL) inherits demonstration timing, while directly accelerating the learned motion can fail when faster execution changes the robot-object dynamics. We study faster-than-demonstration execution as a dynamics-aware control problem and introduce Expert-Play Contouring Control (EPCC), which combines slow expert demonstrations with fast, non-expert play. Expert demonstrations train a latent trajectory generator whose predictions are reparameterized into a time-independent contour of successful task progression, while play trains a world model (WM) of fast-action outcomes. At deployment, our proposed planner uses the WM to optimize actions that makes maximize progress along the expert-derived contour while penalizing deviation from the intended task evolution. Averaged across three visuomotor manipulation tasks, EPCC achieves a $2.0\times$ the throughput of the IL baseline, including $2.2\times$ that of the throughput of the strongest acceleration baseline on a task with interaction-sensitive object dynamics. Our analysis shows that the gains concentrate where faster execution changes robot-object evolution. Together, our results highlight a simple yet effective principle: demonstrations provide task intent, while play data provides the dynamic coverage needed to execute that intent faster.
Chinese Translation
专家演示通常指定机器人应执行的任务，但并未说明其执行速度。模仿学习（Imitation Learning, IL）继承了演示的时序，而直接加速学习到的动作可能会在更快的执行速度下改变机器人与物体的动态，从而导致失败。我们将快于演示的执行视为一个动态感知的控制问题，并引入专家-游戏轮廓控制（Expert-Play Contouring Control, EPCC），该方法结合了慢速专家演示与快速的非专家游戏。专家演示训练一个潜在轨迹生成器，其预测被重新参数化为一个与时间无关的成功任务进展轮廓，而游戏则训练一个快速动作结果的世界模型（World Model, WM）。在部署阶段，我们提出的规划器使用WM来优化行动，以最大化沿专家导出的轮廓的进展，同时惩罚偏离预期任务演变的行为。在三个视觉运动操作任务的平均结果中，EPCC的吞吐量达到了IL基线的2.0倍，包括在一个对交互敏感的物体动态任务中达到了最强加速基线的2.2倍。我们的分析表明，增益集中在更快的执行改变机器人与物体演变的地方。总的来说，我们的结果强调了一个简单而有效的原则：演示提供任务意图，而游戏数据提供了执行该意图所需的动态覆盖。
cs.RO / 37 / 2609.22803
ARCGym: Benchmarking Deep Reinforcement Learning in Autonomous Robotic Colonoscopy
ARCGym：自主机器人结肠镜检查中的深度强化学习基准测试
Abstract
Simulations for learning-based autonomous colonoscopic navigation focus mainly on fully actuated capsule robots, failing to capture the contact-rich navigation of long and flexible clinical colonoscopes. We present the Autonomous Robotic Colonoscopy Gym (ARCGym), an open-source reinforcement learning environment and benchmark for image-based navigation in clinically derived deformable colon anatomies. ARCGym supports multiple types of colonoscope robots, spanning capsule robots and flexible endoscopes, with this work focusing on flexible endoscopes including magnetic-driven tip actuation and clinically used proximally translational actuation. This work includes five CT-reconstructed colons representing typical clinical scenarios, a set of clinically meaningful navigation subtasks, and unified success metrics. We introduce a reward combining depth-based lumen alignment with a lumen-visibility score to improve learning under occlusions. Experiments across tasks, robots, and anatomies show that autonomous navigation remains challenging for both magnetic-driven and proximal-insertion flexible robots, with proximal-insertion actuation remaining an open problem.
Chinese Translation
基于学习的自主结肠镜导航的模拟主要集中在完全驱动的胶囊机器人上，未能捕捉到长而灵活的临床结肠镜的接触丰富导航。我们提出了自主机器人结肠镜检查环境（ARCGym），这是一个开源的强化学习环境和基准，旨在进行基于图像的导航，适用于临床衍生的可变形结肠解剖结构。ARCGym支持多种类型的结肠镜机器人，包括胶囊机器人和灵活内窥镜，本研究重点关注包括磁驱动尖端驱动和临床使用的近端平移驱动的灵活内窥镜。该研究包括五个CT重建的结肠，代表典型的临床场景，一组具有临床意义的导航子任务，以及统一的成功指标。我们引入了一种奖励机制，将基于深度的腔道对齐与腔道可见性评分相结合，以改善在遮挡条件下的学习。跨任务、机器人和解剖结构的实验表明，对于磁驱动和近端插入的灵活机器人，自主导航仍然具有挑战性，而近端插入驱动仍然是一个未解决的问题。
cs.RO / 38 / 2609.22809
Kinematic Interface for the Wild: Modular Bimanual Loco-Manipulation Capture from 360$^{\circ}$ Cameras Alone
野外运动学接口：仅凭360$^{ ext{°}}$相机实现模块化双手运动操控捕捉
Abstract
A wrist-mounted camera for UMI-style data collection must do two jobs: record the manipulation and localize in the scene. Most handheld devices localize online from workspace-facing views crowded by hands and objects, or add dedicated tracking hardware. Room-scale bimanual capture therefore still tends to instrument the operator or the scene for accurate localization. We present KIWI (Kinematic Interface for the Wild), a capture kit whose only electronics are off-the-shelf cameras. Our core system splits the two jobs across the two lenses of a 360-degree camera. The rear lens faces the room and builds a shared metric map that registers both hands, and an optional head camera, in one frame without workspace co-visibility; the front lens records the manipulation, and offline IMU fusion bridges front-lens tracking loss. Through our quick-release plate, the camera module attaches to chopstick grippers, parallel-jaw grippers, hand-wrist mounts, or robot flanges. Across six bimanual recordings, combining the rear and front lenses failed to localize only 0.1% of query frames, whereas front-only bimanual feature alignment failed on 24.8% of frames and lost one recording entirely; against evaluation fiducials, localization error stayed within 4.5 mm. KIWI's recovered poses were sufficiently consistent for the four wrist streams alone to reconstruct the scene as a 3D Gaussian splat.
Chinese Translation
用于UMI风格数据收集的腕部相机必须完成两个任务：记录操控过程并在场景中进行定位。大多数手持设备通过面向工作空间的视角进行在线定位，这些视角通常被手和物体所拥挤，或者需要添加专用的跟踪硬件。因此，房间尺度的双手捕捉仍然倾向于为操作员或场景配备仪器以实现准确定位。我们提出了KIWI（Kinematic Interface for the Wild），一个仅使用现成相机的捕捉套件。我们的核心系统将这两个任务分配到360度相机的两个镜头上。后镜头面向房间，构建一个共享的度量地图，该地图在一个帧中注册两个手和一个可选的头部相机，而无需工作空间的共同可见性；前镜头记录操控过程，离线IMU融合弥补了前镜头的跟踪丢失。通过我们的快速释放板，相机模块可以连接到筷子夹持器、平行夹爪夹持器、手腕支架或机器人法兰。在六个双手录制中，后镜头和前镜头的结合仅在0.1%的查询帧中未能实现定位，而仅使用前镜头的双手特征对齐在24.8%的帧中失败，并且完全丢失了一次录制；与评估基准相比，定位误差保持在4.5毫米以内。KIWI恢复的姿态在四个腕部流中足够一致，可以单独重建场景为3D高斯点云。
cs.RO / 39 / 2609.22813
Commonsense-Grounded Path Planning from Abstract Instructions
基于常识的抽象指令路径规划
Abstract
We present \emph{commonsense ranked search} (CoRS), a novel path planner that turns an abstract instruction into a route that follows commonsense. While existing methods respect the considerations written down in advance, a robot working among people must follow those left unstated too, as with a wet floor that a worker avoids without being told. CoRS leverages large language models (LLMs) and vision-language models (VLMs) as commonsense knowledge to reason about these latent considerations in its planning. Given an abstract instruction (\emph{e.g.}, ``move carefully'') and visual observations of each region in the environment, CoRS derives a consideration for each region, as in ``this wet floor is slippery and worth a detour.'' It then compares the considerations between regions to see which of the two the robot should avoid more, as in ``the crowd is worse than the wet floor.'' These judgments sort the regions into a commonsense ranking, whose costs drive a conventional search that always returns a valid route. We build a benchmark for planning under latent considerations, with three environments, 1350 problems, and five instructions at three levels of abstraction. Experiments show that CoRS discovers the unstated considerations and goes around the ones worth a detour while crossing the rest, a behavior that recent LLM-based planners do not achieve.
Chinese Translation
我们提出了一种新颖的路径规划器—— extit{常识排名搜索}（CoRS），该规划器将抽象指令转化为遵循常识的路线。现有方法通常遵循事先写下的考虑因素，但在人员密集的环境中工作的机器人也必须遵循那些未被明确指出的因素，例如工人会避免湿滑的地面。CoRS利用大型语言模型（LLMs）和视觉-语言模型（VLMs）作为常识知识，以推理其规划中的这些潜在考虑因素。给定一个抽象指令（例如，“小心移动”）和环境中每个区域的视觉观测，CoRS为每个区域推导出一个考虑因素，例如“这个湿滑的地面很滑，值得绕行。”然后，它比较不同区域之间的考虑因素，以判断机器人应该更避免哪一个区域，例如“人群比湿滑的地面更糟糕。”这些判断将区域排序为常识排名，其成本驱动一个传统搜索，始终返回有效路线。我们建立了一个在潜在考虑因素下进行规划的基准，包含三个环境、1350个问题和三个抽象层次的五个指令。实验表明，CoRS能够发现未被明确指出的考虑因素，并在穿越其他区域时绕过值得绕行的区域，而这一行为是近期基于LLM的规划器所无法实现的。
cs.RO / 40 / 2609.22829
Whole-Body UMI: Transferring UMI Manipulation Skills to Humanoid Whole-Body Manipulation via Real-Time Motion Generation
全身 UMI：通过实时运动生成将 UMI 操作技能转移到类人全身操作
Abstract
Collecting whole-body demonstrations for humanoid manipulation mostly relies on teleoperation, which is costly and hard to scale up. The Universal Manipulation Interface (UMI) provides a scalable data collection paradigm, but end-effector trajectories alone underdetermine humanoid whole-body coordination, which is insufficient for whole-body demonstration collection. Therefore, we introduce Whole-Body UMI (WB-UMI), a task-agnostic, real-time and end-effector conditioned motion generator that decouples whole-body coordination learning from task semantics learning through a shared end-effector interface. A diffusion policy learns from native UMI demonstrations, while WB-UMI learns independently from retargeted motion capture, requiring no body trackers or paired image--whole-body demonstrations during task-specific data collection. In real deployment, an asynchronous hierarchy integrates the diffusion policy, motion generator, and a whole-body controller with latency compensation and measured-state feedback. Real-robot experiments on G1 support real-time closed-loop transfer across four tasks, achieving 90% success in drawer closing, 80% in shelf pick-and-place, 30% in ball toss, and 40% in Loco-PnP, which shows the effectiveness of this hierarchy in transferring native UMI skills to humanoid whole-body manipulation.
Chinese Translation
收集类人操作的全身演示主要依赖于遥操作，这种方式成本高且难以扩展。通用操作接口（Universal Manipulation Interface, UMI）提供了一种可扩展的数据收集范式，但仅依赖末端执行器轨迹无法充分确定类人全身协调，这对于全身演示的收集是不够的。因此，我们提出了全身 UMI（Whole-Body UMI, WB-UMI），这是一种任务无关的、实时的、以末端执行器为条件的运动生成器，通过共享的末端执行器接口将全身协调学习与任务语义学习解耦。扩散策略（diffusion policy）从原生 UMI 演示中学习，而 WB-UMI 则独立于重新定向的动作捕捉学习，在任务特定数据收集过程中不需要身体追踪器或配对的图像-全身演示。在实际部署中，一个异步层次结构将扩散策略、运动生成器和具有延迟补偿和测量状态反馈的全身控制器集成在一起。在 G1 上的真实机器人实验支持跨四个任务的实时闭环转移，在抽屉关闭任务中实现了 90% 的成功率，在货架取放任务中为 80%，在投球任务中为 30%，在 Loco-PnP 任务中为 40%，这表明该层次结构在将原生 UMI 技能转移到类人全身操作中的有效性。
cs.RO / 41 / 2609.22840
ForceRFT: Refining VLA Actions through Force-Guided Residual Reinforcement Learning
ForceRFT：通过力引导的残差强化学习优化视觉-语言-动作（VLA）策略
Abstract
Force-conditioned vision-language-action (VLA) policies can respond to contact, but when trained solely on demonstrations, their recovery behavior may be limited by demonstration coverage, and they do not learn from deployment outcomes. Human corrective imitation provides additional recovery examples, but its objective matches local action targets without explicitly optimizing task return. We present ForceRFT, a force-guided residual reinforcement learning framework that learns contact-dependent corrections from human supervision and autonomous task outcomes. A frozen, demonstration-trained SmolVLA-based prior generates force-conditioned action chunks, while a lightweight residual actor refines individual end-effector pose commands using wrist feedback acquired during chunk execution. The decision-time wrist wrench, its temporal change, and the selected base motion condition both residual correction and value estimation. Human corrections supervise the residual actor, while verified autonomous transitions train the twin critics and support value-guided updates to the same actor. Bootstrapping is restricted to autonomous segments, preventing TD credit from crossing human-intervention boundaries. Real-robot experiments on plug insertion, ring-on-peg assembly, and whiteboard wiping show higher autonomous success rates than the evaluated demonstration-trained and residual-imitation baselines. Comparisons with residual imitation support value-guided residual optimization, while plug-insertion ablations indicate the benefit of direct execution-time wrist feedback.
Chinese Translation
力条件的视觉-语言-动作（VLA）策略能够对接触做出反应，但仅依赖演示进行训练时，其恢复行为可能受到演示覆盖范围的限制，并且无法从实际部署结果中学习。人类的纠正模仿提供了额外的恢复示例，但其目标与局部动作目标匹配，而未明确优化任务回报。我们提出了ForceRFT，一种力引导的残差强化学习框架，能够从人类监督和自主任务结果中学习与接触相关的修正。一个冻结的、基于SmolVLA的演示训练先验生成力条件的动作片段，而一个轻量级的残差执行者使用在片段执行过程中获得的腕部反馈来优化单个末端执行器的姿态命令。决策时的腕部扭矩、其时间变化以及所选择的基础运动条件共同影响残差修正和价值估计。人类的纠正监督残差执行者，而经过验证的自主过渡则训练双重评论者，并支持对同一执行者的价值引导更新。自举过程仅限于自主片段，防止时间差分信用跨越人类干预边界。在插头插入、环在钉子上组装和白板擦拭的真实机器人实验中，显示出比评估的演示训练和残差模仿基线更高的自主成功率。与残差模仿的比较支持价值引导的残差优化，而插头插入的消融实验则表明直接执行时腕部反馈的益处。
cs.RO / 42 / 2609.22852
Physical-Touch Observability from Wrist Wrench in Granular Scooping
基于手腕扭矩的颗粒铲取中的物理接触可观测性
Abstract
Mining and earthmoving are important real-world deployment settings for embodied intelligence. Autonomous transport and driving systems have improved substantially, but loading and scooping still often depend on skilled human operators, exposing personnel and equipment to operational risk. For robotic scooping, pre-contact RGB-D sensing reveals surface geometry but not the resistance, compaction, tool engagement, or load transfer that emerge during interaction. We test whether the current scoop's six-axis wrist force/torque (F/T), or wrist wrench, contains information about final collected volume, and whether that information depends on the correctly paired action-terrain interaction. We call this property physical-touch observability. Using 6,700 real-robot scoops across 67 terrains, we evaluate correctly paired current-scoop F/T against pre-contact prediction and correspondence-breaking controls under terrain-held-out testing. At the retrospective 60% sequence boundary, correctly paired F/T reduces mean absolute error by 14.2% relative to Action-only and by 21.9% relative to cross-terrain mismatched F/T. Engineered signal summaries reproduce the result across model architectures. Together, these findings position wrist wrench not merely as a low-level feedback signal, but as a task-level perceptual modality through which embodied robots can infer hidden physical states during interaction, providing a foundation for response-aware autonomy in mining and other contact-rich tasks.
Chinese Translation
采矿和土方工程是具身智能的重要现实应用场景。尽管自主运输和驾驶系统已经有了显著改善，但装载和铲取仍然常常依赖于熟练的人类操作员，这使得人员和设备面临操作风险。在机器人铲取中，接触前的RGB-D传感器能够揭示表面几何形状，但无法反映在交互过程中出现的阻力、压实、工具接合或负载转移。我们测试当前铲取的六轴手腕力/扭矩（F/T）或手腕扭矩是否包含有关最终收集体积的信息，以及该信息是否依赖于正确配对的动作-地形交互。我们将这种特性称为物理接触可观测性。通过对67种地形进行6,700次真实机器人铲取的评估，我们在地形保留测试中评估了正确配对的当前铲取F/T与接触前预测和对应破坏控制的表现。在回顾性的60%序列边界处，正确配对的F/T相较于仅动作的情况减少了14.2%的平均绝对误差，相较于跨地形不匹配的F/T减少了21.9%。经过工程化的信号摘要在不同模型架构中重现了这一结果。综合这些发现，手腕扭矩不仅被视为低级反馈信号，而是作为一种任务级感知模态，使得具身机器人能够在交互过程中推断隐藏的物理状态，为采矿和其他接触丰富任务中的响应感知自主性奠定了基础。
cs.RO / 43 / 2609.22854
SmoLSTM: A Compact Vision-Language-Action Model with Recurrent Memory that Persists
SmoLSTM：一种具有持久递归记忆的紧凑型视觉-语言-动作模型
Abstract
Vision-language-action models often predict actions from only the current observation, which can leave tasks involving object occlusion or visually identical objects ambiguous without episode history. The usual countermeasure, widening the observation window, turns the horizon into a hyperparameter and lets per-step cost grow with it. We instead capture the episode in a recurrent state. SmoLSTM couples a frozen 256M-parameter SmolVLM backbone to a matrix-memory LSTM control layer in which observation tokens and action queries are unified in a single causal stream that is never reset throughout the entire episode. Recurrent-state storage is therefore O(1) in episode length. A flow-matching action head predicts chunks of 10 end-effector pose deltas and gripper commands at each control step. Our single policy, trained jointly on 7,461 demonstrations across 140 tasks and evaluated on held-out initial states, performs best, reaching 85.1% subgoal coverage and 77.5% full-task success on LIBERO-Mem with 0.04B trainable parameters, surpassing both the benchmark's own object-centric baseline and recent memory-based approaches. Resetting the recurrent state at every control step reduces full-task success to 7.0%, showing that the trained policy relies on context carried between decisions. The same model achieves 79.6% average success on standard LIBERO.
Chinese Translation
视觉-语言-动作模型通常仅根据当前观察来预测动作，这在涉及物体遮挡或视觉上相同的物体时，缺乏情节历史会导致任务模糊。常见的对策是扩大观察窗口，这将视野变为一个超参数，并使每一步的成本随之增加。我们则通过递归状态来捕捉情节。SmoLSTM 将一个冻结的 256M 参数的 SmolVLM 主干与一个矩阵记忆 LSTM 控制层结合，在该控制层中，观察标记和动作查询统一在一个始终不被重置的因果流中。递归状态存储因此在情节长度上为 O(1)。流匹配动作头在每个控制步骤中预测 10 个末端执行器姿态增量和夹持器命令。我们的单一策略在 140 个任务的 7,461 个演示上联合训练，并在保留的初始状态上进行评估，表现最佳，在 LIBERO-Mem 上达到 85.1% 的子目标覆盖率和 77.5% 的完整任务成功率，超越了基准的对象中心基线和近期的基于记忆的方法。在每个控制步骤重置递归状态会将完整任务成功率降低至 7.0%，这表明训练的策略依赖于在决策之间传递的上下文。同一模型在标准 LIBERO 上实现了 79.6% 的平均成功率。
cs.RO / 44 / 2609.22858
PileBelief: Persistent Physical State for Interaction-Driven World Modeling
PileBelief：用于交互驱动世界建模的持久物理状态
Abstract
World models allow robots to anticipate action consequences before execution. This capability is especially valuable in excavation, where each scoop reshapes the terrain and affects subsequent actions. Local observations, however, cannot fully reveal the underlying support and material conditions. We present PileBelief, an interaction-driven persistent world model for partially observed excavation that retains physical evidence beyond the visible surface. It combines an observation-conditioned physical prior with world-addressed deformation memory and physical-response memory. Action-aligned reads and gated residual corrections refine terrain-change and outcome predictions. With deployment weights fixed, completed interactions update measured belief, while hypothetical actions advance a separate imagined state. Compared with a current-observation-only baseline, PileBelief reduces five-step joint prediction error by 10.8% and offline action-selection regret by 65.5%. Experiments on Newton/MPM and real excavation datasets further demonstrate improved terrain-change and bucket-volume prediction. Our method enables multi-step prediction and candidate-action ranking from local observations, even when the underlying soil state is unknown. These results identify persistent physical belief as a useful representation for world models of environments that robots continually reshape.
Chinese Translation
世界模型使机器人能够在执行之前预测行动的结果。这一能力在挖掘作业中尤为重要，因为每一铲的操作都会重塑地形并影响后续的行动。然而，局部观察无法完全揭示潜在的支撑和材料条件。我们提出了PileBelief，一种用于部分观察挖掘的交互驱动持久世界模型，它保留了可见表面之外的物理证据。该模型结合了基于观察的物理先验、世界地址的变形记忆和物理响应记忆。与行动对齐的读取和门控残差修正可以细化地形变化和结果预测。在固定部署权重的情况下，完成的交互更新测量信念，而假设的行动则推进一个独立的想象状态。与仅基于当前观察的基线相比，PileBelief将五步联合预测误差降低了10.8%，将离线行动选择遗憾降低了65.5%。在Newton/MPM和真实挖掘数据集上的实验进一步证明了地形变化和铲斗容量预测的改善。我们的方法能够从局部观察中进行多步预测和候选行动排序，即使在基础土壤状态未知的情况下。这些结果表明，持久物理信念是机器人不断重塑环境的世界模型的有用表征。
cs.RO / 45 / 2609.22871
A Reconfigurable Dual-Opposition Architecture for Single-Hand Assembly and Manipulation
一种可重构的双对立结构用于单手装配和操作
Abstract
In-hand assembly is constrained by the need to maintain grasps on two separate parts while controlling their relative motion within a single hand. To enable both in-hand assembly and manipulation, we present a reconfigurable dual-opposition architecture. Specifically, to support simultaneous grasping of two parts and coordinated in-hand manipulation, four independently actuated fingers are organized into two virtual finger (VF) oppositions, with their relative configuration controlled by a reconfigurable palm. To describe hand motion and simultaneous two-object grasping configurations, a kinematic model of the fingers and palm and an object-size-conditioned workspace formulation are built. To further evaluate motion performance and assembly capability, finger-joint motion and palm tracking are characterized, and in-hand assembly is demonstrated through tasks involving grasping, alignment, fastening, and pressing. Ablation experiments further demonstrate the importance of finger abduction/adduction and palm reconfiguration for successful in-hand assembly. In simulation, the proposed hand achieves a mean continuous sphere rotation success rate of 98.6% over diameters of 40-230 mm, compared with 73.8% for the LEAP Hand. After policy fine-tuning with external disturbances, the proposed hand achieves 92.8% success under disturbances from multiple directions, compared with 45.2% for the LEAP Hand. Hardware demonstrations further show in-hand rotation of objects of different sizes using policies trained in simulation. Together, these results show that the proposed architecture supports both assembly of two separately held parts and coordinated manipulation of a single object within one hand.
Chinese Translation
手内装配受到需要在单手中保持对两个独立部件的抓握的限制，同时控制它们的相对运动。为实现手内装配和操作，我们提出了一种可重构的双对立结构。具体而言，为支持同时抓握两个部件和协调的手内操作，四个独立驱动的手指被组织成两个虚拟手指（VF）对立，其相对配置由可重构的手掌控制。为了描述手的运动和同时抓握两个物体的配置，构建了手指和手掌的运动学模型以及基于物体大小的工作空间公式。为了进一步评估运动性能和装配能力，特征化了手指关节运动和手掌跟踪，并通过抓握、对齐、固定和按压等任务展示了手内装配。消融实验进一步表明，手指的外展/内收和手掌重构对成功的手内装配至关重要。在仿真中，所提手部在直径为40-230毫米的范围内实现了98.6%的平均连续球体旋转成功率，而LEAP手的成功率为73.8%。在经过外部干扰的策略微调后，所提手部在来自多个方向的干扰下实现了92.8%的成功率，而LEAP手的成功率为45.2%。硬件演示进一步显示了使用在仿真中训练的策略对不同大小物体的手内旋转。这些结果表明，所提架构支持同时装配两个独立握持的部件和在一只手内协调操作单个物体。
cs.RO / 46 / 2609.22885
Barrier Certificate Synthesis for Non-Polynomial Robotic Dynamics via Polynomial Lifting
通过多项式提升合成非多项式机器人动力学的障碍证书
Abstract
Safe operation of robotic systems requires trajectories to remain within a prescribed safe set under admissible control inputs. Barrier certificates provide such guarantees by certifying a controlled-invariant region within that set. Sum-of-squares optimization offers a systematic way to synthesize such certificates, but its direct application requires polynomial dynamics, excluding common robotic nonlinearities, including trigonometric terms. We address this limitation using exact polynomial lifting, which replaces non-polynomial dynamics with polynomial-augmented dynamics subject to lifting-induced algebraic constraints, preserving nonlinear geometry without approximation. We formulate lifted-domain joint barrier synthesis that computes a certificate with a state-feedback control witness and develop a sampled-data safety filter for zero-order-hold implementation. To assess whether the benefits of lifting persist across synthesis frameworks, we also adapt a sample-guided successive-barrier method to the lifted representation. On coordinated-turn and planar multirotor models, exact lifting improves certified coverage in both methods: at matched sample sizes, lifted successive-barrier synthesis achieves higher coverage with fewer barriers and lower computational cost, while lifted joint barrier synthesis provides higher coverage and lower computational cost than the finest tested piecewise resolution. In closed-loop experiments, the safety filter maintains feasibility and safety across all evaluated trajectories, reduces spatial conservativeness, and requires less intervention for both models.
Chinese Translation
机器系统的安全操作要求在可接受的控制输入下，轨迹保持在规定的安全集合内。障碍证书通过认证该集合内的受控不变区域提供了这样的保证。平方和优化提供了一种系统化的方法来合成这样的证书，但其直接应用要求动力学为多项式形式，这排除了包括三角函数项在内的常见机器人非线性。我们通过精确的多项式提升来解决这一限制，该方法用多项式增强的动力学替代非多项式动力学，并受提升引起的代数约束的限制，从而在不进行近似的情况下保留非线性几何。我们制定了提升域联合障碍合成，计算具有状态反馈控制见证的证书，并开发了用于零阶保持实现的采样数据安全过滤器。为了评估提升的好处是否在合成框架中持续存在，我们还将样本引导的连续障碍方法调整为提升表示。在协调转弯和平面多旋翼模型上，精确提升在两种方法中都提高了认证覆盖率：在匹配的样本大小下，提升的连续障碍合成以较少的障碍和较低的计算成本实现了更高的覆盖率，而提升的联合障碍合成则提供了比测试的最细分段分辨率更高的覆盖率和更低的计算成本。在闭环实验中，安全过滤器在所有评估的轨迹上保持可行性和安全性，减少了空间保守性，并对两种模型要求更少的干预。
cs.RO / 47 / 2609.22888
Stable and Efficient Real-World Online VLA Post-Training via Asynchronous Replay-Anchored Policy Improvement
通过异步重放锚定策略改进实现稳定高效的现实世界在线视觉-语言-动作后训练
Abstract
Online post-training of vision-language-action (VLA) models requires efficient use of robot interaction and reliable policy improvement from continually collected experience. We propose asynchronous Replay-Anchored Policy improvement (RAPolicy), a framework that performs rollout and learning concurrently while grounding both critic and actor updates in replayed behavior. The critic learns chunk-level values from recorded actions and constructs Bellman targets without predicting next actions, reducing computation and dependence on action-value estimates outside replay coverage. The one-step flow actor reuses the initial noise stored during rollout and learns through advantage-weighted conditional likelihood, directly supervising the action mapping used for execution. We evaluate RAPolicy across four single-task settings and one joint five-task setting in the real world, with online training budgets of only 1--2 hours. Starting from policies fine-tuned on just 10 demonstrations per task, RAPolicy rapidly adapts to new single tasks and achieves an average 86.3% success rate. In the joint multi-task setting, RAPolicy improves overall success rate from 52% to 88% while preserving performance on already reliable tasks and improving weaker capabilities. Overall, RAPolicy substantially outperforms the baselines in aggregate task success while requiring fewer human interventions, demonstrating stable policy improvement and high online training efficiency. Project page: https://flyfaerss.github.io/RAPolicy.
Chinese Translation
视觉-语言-动作（VLA）模型的在线后训练需要有效利用机器人交互，并从持续收集的经验中实现可靠的策略改进。我们提出了异步重放锚定策略改进（RAPolicy），这是一个同时进行回放和学习的框架，同时将评论者和执行者的更新基于重放的行为进行基础。评论者从记录的动作中学习块级值，并构建贝尔曼目标，而无需预测下一个动作，从而减少计算量并降低对重放覆盖之外的动作值估计的依赖。一步流执行者重用回放过程中存储的初始噪声，并通过优势加权条件似然进行学习，直接监督用于执行的动作映射。我们在现实世界中评估了RAPolicy在四个单任务设置和一个联合五任务设置中的表现，在线训练预算仅为1到2小时。从仅在每个任务上进行10次演示微调的策略开始，RAPolicy迅速适应新的单任务，并实现了平均86.3%的成功率。在联合多任务设置中，RAPolicy将整体成功率从52%提高到88%，同时保持在已可靠任务上的表现，并改善较弱的能力。总体而言，RAPolicy在总体任务成功率上显著优于基线，同时需要更少的人为干预，展示了稳定的策略改进和高效的在线训练。项目页面：https://flyfaerss.github.io/RAPolicy。
cs.RO / 48 / 2609.22895
H-VLA: Hierarchical Vision-Language-Action Model with Key-Action Reasoning and Motion Planning in a Unified Action Space
H-VLA：具有关键动作推理和运动规划的层次化视觉-语言-动作模型在统一动作空间中的应用
Abstract
Vision-Language-Action (VLA) models have shown strong potential for robotic manipulation, but many existing methods still rely on direct mappings from language and visual observations to dense actions. This formulation can weaken the semantic reasoning capability inherited from pre-trained Vision-Language Models (VLMs), which are mainly optimized for visual-linguistic understanding rather than low-level control, and becomes fragile under spatial variations, including changes in object positions, scene layouts, robot embodiments, and camera viewpoints. To address these limitations, we propose H-VLA, a hierarchical VLA framework that decouples high-level key-action reasoning from low-level motion generation. H-VLA combines a Key-Action Model for predicting a key-action as the next manipulation subgoal, a Motion Planning Model for generating dense future actions conditioned on the predicted key-action, and a Unified Camera-Centric Action Space for consistent representation across datasets, embodiments, and viewpoints. We further adopt a two-stage training strategy that emphasizes key-action reasoning during pre-training and dense motion generation during fine-tuning. Experiments show that H-VLA achieves strong performance on SimplerEnv, reaching 91% on Google Robot visual matching, 84% on Google Robot variant aggregation, and 81% on WidowX visual matching. On Agilex real-robot tasks, H-VLA improves over the strongest baseline by 10, 47, and 16 percentage points under in-distribution, out-of-distribution position, and out-of-distribution scene/object settings, respectively.
Chinese Translation
视觉-语言-动作（VLA）模型在机器人操作中展现出强大的潜力，但许多现有方法仍依赖于从语言和视觉观察到密集动作的直接映射。这种表述可能削弱了从预训练视觉-语言模型（VLMs）继承的语义推理能力，这些模型主要针对视觉-语言理解进行优化，而非低级控制，并且在空间变化（包括物体位置、场景布局、机器人形态和相机视角的变化）下表现脆弱。为了解决这些局限性，我们提出了H-VLA，一个层次化的VLA框架，它将高层次的关键动作推理与低层次的运动生成解耦。H-VLA结合了一个关键动作模型，用于预测下一个操作子目标的关键动作，一个运动规划模型，用于基于预测的关键动作生成密集的未来动作，以及一个统一的以相机为中心的动作空间，以确保在数据集、形态和视角之间的一致表示。我们进一步采用了两阶段的训练策略，在预训练期间强调关键动作推理，在微调期间强调密集运动生成。实验表明，H-VLA在SimplerEnv上表现出色，在Google Robot视觉匹配中达到91%，在Google Robot变体聚合中达到84%，在WidowX视觉匹配中达到81%。在Agilex真实机器人任务中，H-VLA在分布内、分布外位置和分布外场景/物体设置下，分别比最强基线提高了10、47和16个百分点。
cs.RO / 49 / 2609.22925
"Dear LLaVA, Please Drive": A Depth-Aware Vision-Language Agent for Closed-Loop Robotic Control
亲爱的 LLaVA，请驾驶：一种深度感知的视觉语言智能体用于闭环机器人控制
Abstract
Vision-language models (VLMs) provide a compelling foundation for reasoning-driven mobile navigation, offering rich contextual understanding and strong generalization from large-scale pretraining. Most existing navigation frameworks rely on imitation learning and therefore require substantial labeled trajectory data, limiting their scalability and robustness. In this work, we propose a parameter-efficient approach to fine-tune a pretrained VLM for autonomous navigation using an Imperative Learning paradigm. By optimizing against differentiable geometric cost fields rather than labeled trajectories, our model learns to generate collision-free paths exclusively from stereoscopic depth observations. We introduce a unified end-to-end navigation pipeline for natural-language-driven robotic control. This system leverages a shared VLM backbone with task-specific Low-Rank Adaptation (LoRA) modules, effectively bridging the gap from semantic target selection to low-level trajectory planning. Our approach achieves competitive Success weighted by Path Length (SPL) in unseen environments while updating less than 1% of the model's total parameters. Qualitative real-world experiments validate sim-to-real generalization and stable path planning without fine-tuning on real-world data. These results highlight a practical approach for deploying VLM-based agents on mobile robots, enabling high-level semantic navigation without the prohibitive requirement for large-scale, labeled trajectory data.
Chinese Translation
视觉语言模型（VLMs）为基于推理的移动导航提供了引人注目的基础，能够从大规模预训练中提供丰富的上下文理解和强大的泛化能力。现有的大多数导航框架依赖于模仿学习，因此需要大量标注的轨迹数据，这限制了它们的可扩展性和鲁棒性。在本研究中，我们提出了一种参数高效的方法，通过命令学习（Imperative Learning）范式对预训练的 VLM 进行微调，以实现自主导航。通过针对可微分几何成本场进行优化，而不是标注轨迹，我们的模型学习仅从立体深度观测生成无碰撞路径。我们引入了一个统一的端到端导航管道，用于自然语言驱动的机器人控制。该系统利用共享的 VLM 主干和任务特定的低秩适应（Low-Rank Adaptation, LoRA）模块，有效地弥合了语义目标选择与低级轨迹规划之间的差距。我们的方法在未见环境中实现了具有竞争力的路径长度加权成功率（Success weighted by Path Length, SPL），同时更新的模型总参数不到 1%。定性实地实验验证了从仿真到现实的泛化能力和稳定的路径规划，而无需在真实数据上进行微调。这些结果突显了一种在移动机器人上部署基于 VLM 的智能体的实用方法，使高层次的语义导航成为可能，而无需大量标注的轨迹数据。
cs.RO / 50 / 2609.22926
Embodied Snap: Octopus-Inspired Distributed Reach-and-Attach with a Speed-Limited Soft Arm
具身快照：受章鱼启发的分布式抓取与附着速度受限的软臂
Abstract
Reach-and-attach of soft robotic arms with passive suction requires accurate targeting and sufficient contact speed, yet geared actuators can impose a speed limit that improved trajectory tracking alone cannot overcome. This paper proposes an embodied snap controller that separates slow servo-driven preloading from rapid elastic release, enabling a compliant arm to move beyond its direct tendon-driven speed limit. Octopus biology motivates the controller's section-wise organizational prior, rather than reproduction of the octopus nervous system. A learned policy shared across three sections selects preloads, aim, tendon slack, and release timing, determining where, how, and when to load and release the body. The policy is optimized offline using a hardware-validated recurrent model within experimentally supported bounds. Across five optimization seeds and 400 unseen simulated targets, attachment success is $(73\pm4)\%$ at a $5\text{ cm}$ lateral tolerance, and the shared policy reaches the matched centralized controller's mean final reward after a median $17\%$ of the common evaluation budget. Hardware characterization achieves tip speeds of 1.56-1.64 m/s, at least $108\%$ above direct tendon-driven release. In 18 open-loop hardware trials across six placements, 17 exceed the 1 m/s snap threshold and nine retrieve the object, with successful retrieval at five placements. These results demonstrate a practical division of responsibility in the control problem: learned control prepares the body, and passive body mechanics execute the rapid movement needed for dynamic reach-and-attach.
Chinese Translation
软体机器人臂的抓取与附着依赖于被动吸附，需要准确的目标定位和足够的接触速度，然而，齿轮驱动的执行器可能会施加速度限制，仅靠改进轨迹跟踪无法克服。本文提出了一种具身快照控制器，将缓慢的伺服驱动预加载与快速的弹性释放分离，使得柔性臂能够超越其直接腱驱动的速度限制。控制器的分段组织优先考虑受章鱼生物学的启发，而非复现章鱼的神经系统。一个跨越三个部分的学习策略选择预加载、瞄准、腱松弛和释放时机，决定了何时、如何以及在哪里加载和释放身体。该策略在实验支持的范围内，使用经过硬件验证的递归模型进行离线优化。在五个优化种子和400个未见的模拟目标中，附着成功率为$(73 ext{±}4)\%$，在$5 ext{ cm}$的横向容差下，所共享的策略在中位数$17 ext{\%}$的共同评估预算后达到了匹配的集中控制器的平均最终奖励。硬件特性测试实现了1.56-1.64 m/s的尖端速度，至少比直接腱驱动释放高出$108 ext{\%}$。在六个放置位置的18次开环硬件试验中，有17次超过1 m/s的快照阈值，九次成功取回物体，其中五个放置位置成功取回。这些结果展示了控制问题中责任的实际分工：学习控制准备身体，而被动的身体力学执行所需的快速动态抓取与附着。
cs.RO / 51 / 2609.22954
MIRA-PRM: Mission-Informed Reusable Roadmap Planning for Mobile Gas-Sensing Inspection
MIRA-PRM：基于任务的信息化可重用路线图规划用于移动气体检测
Abstract
Mobile gas-sensing inspection requires a mobile platform to reliably reach ordered sampling poses. We present MIRA-PRM, a mission-informed probabilistic roadmap that combines geometry-, mission-flow-, and inspection-conditioned sampling with locally adaptive connectivity and evidence-gated refinement. Eight development campaigns comprising 142,448 planner runs evaluated mission scaling, repeated-query adaptation, finite node budgets, target distributions, local map repair, ablation, cross-family comparison, and parameter sensitivity. MIRA-PRM maintained 100% seed-level success as ordered targets increased from two to eight, and achieved 97.67% seed-level success under a 40-node ceiling, compared with 67.17% for PRM and 86.50% for PRM*. Local repair reduced event time by 79.9-81.9% and node additions by 90.8-94.3% relative to rebuilding. On a separate frozen holdout of 12 maps, 48 ordered missions, and three seeds per map-mission unit, MIRA-PRM produced 96/144 validated successes, versus 82/144 for PRM, 89/144 for PRM*, and 70/144 for LE-HG-PRM. On paired common-success runs, MIRA-PRM reduced wall time by 16.89% relative to PRM* and 81.01% relative to LE-HG-PRM, but was 52.23% slower than PRM; paths were 0.54-1.86% longer. All validated paths passed independent full-segment checks and clearance non-inferiority. The results support a reliability-oriented trade-off in structured simulations and motivate external-map and physical mobile-sensing validation.
Chinese Translation
移动气体检测需要一个移动平台可靠地到达指定的采样位置。我们提出了MIRA-PRM，一种基于任务的信息化概率路线图，结合了几何、任务流和检测条件的采样，以及局部自适应连通性和证据引导的细化。八个开发活动共计142,448次规划运行评估了任务扩展、重复查询适应、有限节点预算、目标分布、局部地图修复、消融实验、跨家族比较和参数敏感性。随着有序目标从两个增加到八个，MIRA-PRM保持了100%的种子级成功率，并在40节点上限下实现了97.67%的种子级成功率，而PRM为67.17%，PRM*为86.50%。局部修复相较于重建减少了79.9-81.9%的事件时间和90.8-94.3%的节点增加。在一个单独的冻结保留实验中，包括12个地图、48个有序任务和每个地图-任务单元三个种子，MIRA-PRM产生了96/144次验证成功，而PRM为82/144，PRM*为89/144，LE-HG-PRM为70/144。在配对的共同成功运行中，MIRA-PRM相较于PRM*减少了16.89%的墙面时间，相较于LE-HG-PRM减少了81.01%，但比PRM慢52.23%；路径长度增加了0.54-1.86%。所有验证的路径均通过了独立的全段检查和清除非劣性。结果支持在结构化模拟中进行以可靠性为导向的权衡，并激励外部地图和物理移动传感器的验证。
cs.RO / 52 / 2609.22963
Prescribed-Time Contracting-Boundary Control of a Tendon-Driven Flexible Arm
基于规定时间的腱驱动柔性臂的边界控制
Abstract
This study develops a prescribed-time performance-shaping control method for curvature tracking of a single-segment flexible arm actuated by three antagonistic tendon pairs. A Cartesian curvature representation is introduced to avoid the undefined bending direction at the straight configuration and to establish an explicit six-tendon kinematic mapping. A cubic performance boundary contracts smoothly from an initially admissible error bound to a nonzero terminal accuracy bound within a prescribed time. Based on this boundary, a dual transformation combining static symmetric error scaling and time-varying behavior shaping maps the tracking error into a fixed unit box. The resulting controller guarantees boundary invariance, prescribed-time entry into the terminal accuracy region, and subsequent asymptotic convergence. Numerical evaluations with Python and OpenCR--MuJoCo, together with a supervised reduced-order experiment on a two-section, four-channel platform, provide complementary validation. Across six experimental trials, no violation of the prescribed boundary is observed, and the proposed controller reduces the mean terminal curvature RMSE by 32.5% relative to a matched baseline, with comparable terminal-band entry times. These results support the feasibility of the proposed approach in the reduced-order experimental setting.
Chinese Translation
本研究开发了一种针对由三对对抗腱驱动的单段柔性臂的曲率跟踪的规定时间性能塑形控制方法。引入了一种笛卡尔曲率表示法，以避免在直线配置下的弯曲方向不确定性，并建立一个明确的六腱运动学映射。一个立方体性能边界在规定时间内从初始可接受的误差边界平滑收缩到非零终端精度边界。基于该边界，结合静态对称误差缩放和时间变化行为塑形的双重变换将跟踪误差映射到固定单位盒中。所得到的控制器保证了边界不变性、规定时间内进入终端精度区域，以及随后的渐近收敛。通过Python和OpenCR--MuJoCo的数值评估，以及在一个两段四通道平台上的监督降阶实验，提供了互补的验证。在六次实验中，没有观察到对规定边界的违反，所提控制器将均值终端曲率均方根误差（RMSE）相对于匹配基线降低了32.5%，且终端带进入时间相当。这些结果支持了在降阶实验环境中所提方法的可行性。
cs.RO / 53 / 2609.22966
Transferring the Intelligence of VLMs to Robotic Control
将视觉语言模型的智能转移到机器人控制
Abstract
Humans can seamlessly adapt to both physical and digital worlds, suggesting that while a digital-to-real gap exists in embodiment, environment and task, human intelligence itself may transfer across this gap. This naturally raises a fundamental question: can the intelligence of vision-language models (VLMs) similarly generalize from the digital world to the physical world for robotic control? We investigate this question through RoboDawn, a human-intuitive interface that exposes robotic control to an agentic VLM through a compact set of discrete translation, rotation, and gripper commands. Using this interface, the VLM controls a robot in a closed loop: it observes the current visual state, reasons about the next action, executes it, and adapts subsequent decisions to the resulting state. Furthermore, we introduce an in-context learning (ICL) scheme that uses a few demonstrations to ground the VLM in both interface usage and task-solving strategies. Experiments on RoboTwin 2.0 C2R and RoboDojo demonstrate that RoboDawn achieves strong performance without task-specific robot training. In the zero-shot setting, RoboDawn outperforms several strong policies trained on benchmarkspecific robot data, while a single in-context demonstration further yields substantial performance gains and establishes state-of-the-art (SOTA) results. On RoboTwin 2.0 C2R, the success rate increases from 53.2% zero-shot to 73.6% one-shot, exceeding the solid baseline {\pi}0.5 (46.0%). Similar gains are observed on RoboDojo, where success rate improves from 35.67% zero-shot to 47.17% one-shot. The same framework also transfers to real-world robots, performing block-in-basket and block stacking on Franka.
Chinese Translation
人类能够无缝适应物理和数字世界，这表明尽管在体现、环境和任务上存在数字与现实之间的差距，但人类智能本身可能跨越这一差距进行转移。这自然引发了一个基本问题：视觉语言模型（VLMs）的智能是否也能类似地从数字世界推广到物理世界，以实现机器人控制？我们通过RoboDawn这一人机直观接口来研究这个问题，该接口通过一组紧凑的离散翻译、旋转和抓取命令，将机器人控制暴露给一个具有自主能力的VLM。利用这个接口，VLM在闭环中控制机器人：它观察当前的视觉状态，推理出下一个动作，执行该动作，并根据结果状态调整后续决策。此外，我们引入了一种上下文学习（ICL）方案，利用少量示范将VLM与接口使用和任务解决策略相结合。在RoboTwin 2.0 C2R和RoboDojo上的实验表明，RoboDawn在没有特定任务的机器人训练的情况下，取得了强劲的表现。在零样本设置下，RoboDawn的表现超过了几种在基准特定机器人数据上训练的强策略，而单个上下文示范进一步带来了显著的性能提升，并确立了最新的研究成果（SOTA）。在RoboTwin 2.0 C2R上，成功率从53.2%的零样本提升至73.6%的单样本，超越了稳固基线{ ext{π}}0.5（46.0%）。在RoboDojo上也观察到了类似的提升，成功率从35.67%的零样本提高到47.17%的单样本。同样的框架也可以转移到现实世界的机器人上，在Franka上执行块入篮和块堆叠任务。
cs.RO / 54 / 2609.22973
An Empirical Study and Open Testbed for Federated Fine-Tuning of Vision-Language-Action Models
联邦微调视觉-语言-动作模型的实证研究与开放测试平台
Abstract
Adapting a pretrained Vision-Language-Action (VLA) model to a new robot, environment, or task requires demonstrations that are collected locally and often discarded. Federated learning is a promising approach to exploiting such distributed demonstrations by learning a shared policy. However, whether it can adapt large pretrained VLAs remains an open question, and a lack of reproducible benchmarks for pretrained VLAs and reusable training frameworks makes existing results difficult to compare. In this paper, we conduct a systematic study of federated fine-tuning of three modern pretrained VLA policies on the 40 simulated tasks of the LIBERO manipulation benchmark, and on six real-world tasks in two real-robot experiments, with demonstrations collected across two and three sites, respectively. Our study analyzes the key choices in this setting, spanning multiple federated parameter scopes, three aggregation algorithms, and evaluation under distribution shift. Based on the study, we derive a series of lessons, including the dominance of the federated scope over the choice of aggregation algorithm and the difficulty of matching centralized fine-tuning on physical robots, where cross-site heterogeneity is stronger than simulation captures. We also highlight opportunities for federated VLA learning, such as the ability to match centralized fine-tuning on heterogeneous data, to remain at least as robust as centralized fine-tuning under distribution shift, and to personalize, with each client federating part of the policy and keeping the rest local, which helps where the policy's pretraining is weak but leaves no usable global model. We open-source \decentvla{}, the model- and runtime-agnostic testbed behind the study, to facilitate future research and fair comparisons in federated VLA learning.
Chinese Translation
将预训练的视觉-语言-动作（VLA）模型适配到新的机器人、环境或任务需要收集本地演示，而这些演示通常会被丢弃。联邦学习是一种有前景的方法，通过学习共享策略来利用这些分布式演示。然而，如何适应大型预训练的VLA仍然是一个未解的问题，缺乏可重复的预训练VLA基准和可重用的训练框架使得现有结果难以比较。在本文中，我们对三种现代预训练VLA策略在LIBERO操作基准的40个模拟任务和两个真实机器人实验中的六个真实任务上进行系统研究，演示分别在两个和三个地点收集。我们的研究分析了这一设置中的关键选择，涵盖多个联邦参数范围、三种聚合算法以及在分布变化下的评估。基于研究，我们总结了一系列经验教训，包括联邦范围对聚合算法选择的主导作用，以及在物理机器人上匹配集中微调的困难，因为跨站点的异质性强于模拟捕获。我们还强调了联邦VLA学习的机会，例如在异质数据上匹配集中微调的能力，至少在分布变化下保持与集中微调同样的鲁棒性，以及个性化，每个客户端联邦化部分策略并保持其余部分本地，这在策略的预训练较弱但没有可用的全局模型时尤为有帮助。我们开源了 exttt{decentvla}，作为本研究背后的模型和运行时无关的测试平台，以促进未来的研究和联邦VLA学习中的公平比较。
cs.RO / 55 / 2609.22974
A Horizon-slicing Approach to Minimum Obstacle Displacement Planning for Robot Navigation
一种地平线切片方法用于机器人导航中的最小障碍物位移规划
Abstract
In this paper, we investigate the Minimum Obstacle Displacement Planning problem from a robot motion planning perspective. The problem involves determining a feasible path to a goal location by displacing movable obstacles when no collision-free path initially exists. We show that this problem is computationally challenging and, in particular, NP-hard when obstacles are modeled as polygons in the plane. Besides an exact formulation of the minimum obstacle displacement problem generalizing other problems in the literature, and the associated optimal solution, this paper proposes an approximate solution that is less intensive from a computational standpoint, and differs from the optimal solution by a fraction of the optimal cost, being able to trade-off between path length and amount of obstacle displacements.
Chinese Translation
本文从机器人运动规划的角度研究了最小障碍物位移规划问题。该问题涉及在初始不存在无碰撞路径的情况下，通过位移可移动障碍物来确定通往目标位置的可行路径。我们表明，该问题在计算上具有挑战性，特别是当障碍物建模为平面中的多边形时，它是 NP-hard 的。除了对最小障碍物位移问题的精确表述外，该表述概括了文献中的其他问题，并给出了相关的最优解，本文还提出了一种计算上较为轻便的近似解，该解与最优解的差异仅为最优成本的一小部分，能够在路径长度和障碍物位移量之间进行权衡。
cs.RO / 56 / 2609.22983
Connectivity-Aware Exploration of Robotic Grasp Spaces
考虑连通性的机器人抓取空间探索
Abstract
Robotic grasping is typically formulated as the problem of identifying successful actions from a space of candidate grasp poses. However, the organization of successful actions within this space has received less attention. We study the multiscale structure of viable robotic grasps in $SE(3)$ and investigate whether this structure can be exploited for more efficient exploration. Using a large-scale grasp dataset, we show that successful grasp sets exhibit heterogeneous and reproducible connectivity structure across objects. We then introduce a connectivity-aware sampling strategy that incrementally explores the currently observed grasp space by prioritizing potential bridges between components, structural frontiers, boundary extensions, and geometric novelty. In controlled reconstruction experiments, the method recovers the connectivity structure of successful grasp sets substantially more efficiently than random sampling and farthest-point sampling. We further evaluate whether connectivity acquired under hidden grasp viability can improve subsequent grasp discovery, and whether structural experience from previously explored objects can be retrieved and transferred to unseen objects. These results suggest that the spatial organization of viable actions provides information relevant to grasp-space exploration beyond the viability of individual candidate actions. More broadly, they motivate structure-aware exploration as a means of exploiting the geometry of viable action spaces in robotic manipulation.
Chinese Translation
机器人抓取通常被表述为从候选抓取姿态空间中识别成功动作的问题。然而，成功动作在该空间中的组织却鲜有关注。我们研究了$SE(3)$中可行机器人抓取的多尺度结构，并探讨这一结构是否可以被利用以实现更高效的探索。通过使用大规模抓取数据集，我们展示了成功抓取集在不同物体之间展现出异质且可重复的连通性结构。随后，我们提出了一种考虑连通性的采样策略，该策略通过优先考虑组件之间的潜在桥梁、结构前沿、边界扩展和几何新颖性，逐步探索当前观察到的抓取空间。在受控重建实验中，该方法在恢复成功抓取集的连通性结构方面显著比随机采样和最远点采样更高效。我们进一步评估了在隐藏抓取可行性下获得的连通性是否能改善后续抓取发现，以及之前探索的物体的结构经验是否可以被提取并转移到未见过的物体上。这些结果表明，可行动作的空间组织提供了与抓取空间探索相关的信息，超出了单个候选动作的可行性。更广泛地说，这些结果激励了结构感知的探索作为利用机器人操作中可行动作空间几何的手段。
cs.RO / 57 / 2609.23037
Fast and Robust Temporal Logic Planning via ADMM-based Trajectory Optimization
基于ADMM的轨迹优化的快速稳健时序逻辑规划
Abstract
We present a fast numerical method for safe continuous-time motion planning under Temporal Logic (TL) specifications. The method generates smooth continuous trajectories that remain collision-free while robustly satisfying temporal and logical task requirements. A central component of our method is the formulation of nonconvex safety and logic constraints as unions of convex sets where associated discrete decisions are encoded in a joint feasibility graph. This graph representation allows Euclidean projection onto the feasible set and proximal robustness maximization to be reformulated as shortest- and widest-path problems, respectively. Building on this structure, we develop a nonconvex splitting method based on the Alternating Direction Method of Multipliers (ADMM), which decouples smooth spatio-temporal trajectory optimization from nonsmooth discrete constraint handling within the optimization. The resulting algorithm exhibits reliable convergence across benchmarks and scales to large-scale motion-planning problems, providing a 4.7x average speedup over the state of the art on discrete and continuous-time logic problems.
Chinese Translation
我们提出了一种快速的数值方法，用于在时序逻辑（TL）规范下进行安全的连续时间运动规划。该方法生成平滑的连续轨迹，能够在保持无碰撞的同时，稳健地满足时序和逻辑任务要求。我们方法的一个核心组成部分是将非凸安全和逻辑约束表述为凸集的并集，其中相关的离散决策被编码在一个联合可行性图中。该图表示允许将欧几里得投影到可行集上，并将近端稳健性最大化重新表述为最短路径和最宽路径问题。基于这一结构，我们开发了一种基于交替方向乘子法（ADMM）的非凸分裂方法，该方法将平滑的时空轨迹优化与优化过程中的非平滑离散约束处理解耦。所得到的算法在基准测试中表现出可靠的收敛性，并能够扩展到大规模运动规划问题，在离散和连续时间逻辑问题上提供了比现有技术平均快4.7倍的速度提升。
cs.RO / 58 / 2609.23048
Anatomy of a Closed-Loop Collapse: A Causal Case Study of a Compressed VLA Policy
闭环崩溃的解剖：压缩 VLA 策略的因果案例研究
Abstract
Compressed manipulation policies can pass offline evaluation while failing in closed-loop execution; this dissociation is established in prior work and is not our claim. We contribute a causal anatomy of one naturally occurring case. An 8-layer distillation of Octo-Base retains 86% of parameters, passes every offline check we applied (0.996 and 1.000 teacher-ratios on the family's own validation metrics), and collapses in closed loop: 0/72 vs. the teacher's 40/72 on a simulated WidowX pick-and-place task. The collapse is structured, not diffuse: early task stages degrade gradually (the student moves the object at 90% of the teacher's rate and grasps at 55%), while transport-to-target fails categorically, at 0% in every training variant. Paired action-trace forensics isolate the signature: a negative, late-heavy $z$ residual, roughly 10x its post-repair magnitude, and persistent across the base distillation and both continuation branches. Four standard therapies fail under matched controls: continued training and in-domain offline data leave success at zero, even though the latter measurably improves marginal action statistics; command-level compensation recovers nothing at any offset, although the same perturbations degrade healthy policies; clamping the symptom in the command channel preserves grasping, yet success stays at floor. A minimal-pair intervention that substitutes half of the training stream with deployment-distribution teacher rollouts, with every other setting held fixed, restores parity with the teacher (18/36 vs. 17/36 held-out), eliminates that signature, and recovers a teacher-like perturbation-response profile. We claim existence, not universality. Operationally, offline gates, including a family's own validation metrics, are insufficient acceptance tests for compressed policies; a few dozen closed-loop trials sufficed to find what they missed.
Chinese Translation
压缩操作策略可以在离线评估中通过，但在闭环执行中失败；这种脱节在之前的研究中已被确立，并非我们的主张。我们贡献了一个自然发生案例的因果解剖。一个 8 层的 Octo-Base 蒸馏保留了 86% 的参数，通过了我们应用的每一个离线检查（在该家族自身的验证指标上，教师比率为 0.996 和 1.000），但在闭环中崩溃：在模拟的 WidowX 拾取和放置任务中表现为 0/72，而教师的表现为 40/72。崩溃是结构性的，而非扩散性的：早期任务阶段逐渐退化（学生以教师 90% 的速度移动物体，抓取速度为 55%），而运输到目标的过程则在每个训练变体中完全失败，表现为 0%。配对的动作轨迹法医分析隔离出特征：一个负的、后期重的 $z$ 残差，其幅度大约是修复后幅度的 10 倍，并在基础蒸馏和两个延续分支中持续存在。四种标准治疗在匹配控制下失败：持续训练和领域内离线数据的成功率为零，尽管后者在边际动作统计上有可测量的改善；命令级补偿在任何偏移下都没有恢复，尽管相同的扰动会削弱健康的策略；在命令通道中钳制症状保留了抓取能力，但成功率仍然处于底线。一个最小对照干预将一半的训练流替换为部署分布教师的回放，同时保持其他设置不变，恢复了与教师的平衡（18/36 对 17/36 的保留），消除了该特征，并恢复了类似教师的扰动响应特征。我们主张存在性，而非普遍性。在操作上，离线门控，包括家族自身的验证指标，对于压缩策略来说是不够的接受测试；几十次闭环试验足以发现它们所遗漏的内容。
cs.RO / 59 / 2609.23060
On the Control of Mobile Ad-Hoc Agent Deployments in Partially Observed Space
对部分可观测空间中移动自组网部署的控制研究
Abstract
We study the online deployment of mobile ad hoc networks in unknown orthogonal environments, formalized as the Partially Observable Cooperative Guard Art Gallery Problem. We give a full proof that CADENCE algorithms achieve full coverage while maintaining a connected visibility graph using at most $n/2 + h - 2$ agents in orthogonal worlds with $n$ corners and $h$ holes. We further evaluate deployment-order heuristics that reduce agent count and deployment time in practice.
Chinese Translation
我们研究了在未知正交环境中移动自组网的在线部署，形式化为部分可观测的协作守卫艺术画廊问题（Partially Observable Cooperative Guard Art Gallery Problem）。我们提供了完整的证明，表明CADENCE算法在正交世界中能够实现完全覆盖，同时保持连接的可视图，使用的代理数量最多为$n/2 + h - 2$，其中$n$为角落数量，$h$为洞的数量。我们进一步评估了在实践中减少代理数量和部署时间的部署顺序启发式方法。
cs.RO / 60 / 2609.23100
Splat-CBF: Safe Next-Best-View Control in 3D Gaussian-Splat Maps
Splat-CBF：在3D高斯点云图中的安全下一个最佳视角控制
Abstract
Where to look and how to move? A robot navigating an unmapped environment must do both at once, and the two goals pull against each other. The regions most worth observing are the ones the map knows least about, and those are exactly where the robot cannot trust its collision margins. We resolve this tension by introducing Splat-CBF, an active perception control barrier function that steers the camera toward the next best view while collision avoidance is enforced as a hard constraint. Safety is enforced by a risk-aware control barrier function that turns the Average Value-at-Risk of the Gaussian field into a single smooth hard constraint. Perception is enforced by a second barrier that rewards camera orientations with high expected Fisher information gain near the robot's planned path. The two meet in a quadratic program where safety is hard and perception is soft, with a slack penalty that adapts to how often perception has already been relaxed and how close the robot is to an uncertain region. We verify the method in indoor simulations, a Isaac Kinova manipulator and in experiments on an Ackermann-drive robot. Our results assert that robot navigates faster, gathers more information, and runs faster online than safety-only and perception-only baselines, giving up informative motion only when safety requires it.
Chinese Translation
机器人在未映射环境中如何选择观察方向和移动方式？这两者必须同时进行，但目标之间存在冲突。最值得观察的区域恰恰是地图信息最少的区域，而这些区域正是机器人无法信任其碰撞边界的地方。我们通过引入Splat-CBF（安全控制障碍函数），解决了这一矛盾。该函数引导相机朝向下一个最佳视角，同时将碰撞避免作为硬约束进行强制执行。安全性由一个风险感知控制障碍函数来保障，该函数将高斯场的平均风险值转化为一个平滑的硬约束。感知则由第二个障碍函数来实现，该函数奖励与机器人规划路径附近具有高期望费舍尔信息增益的相机朝向。两者在一个二次规划中相遇，其中安全性是硬约束，而感知是软约束，并且具有适应于感知已经放宽的频率和机器人接近不确定区域的距离的松弛惩罚。我们在室内模拟、Isaac Kinova操纵器以及Ackermann驱动机器人实验中验证了该方法。我们的结果表明，机器人导航速度更快，信息获取更多，在线运行速度超过仅考虑安全性或感知的基线，仅在安全性要求时放弃信息丰富的运动。
cs.RO / 61 / 2609.23103
DiagGen: Agentic Generation of Deformable Assets with Sim-based Diagnostics for Robotic Simulation
DiagGen：基于模拟诊断的可变形资产的自主生成用于机器人仿真
Abstract
While simulation-ready deformable assets are essential for in-silico robotic manipulation tasks, existing generation frameworks typically assess physical plausibility after generation, leaving an object's simulated response unused as feedback for repairing upstream errors. We present DiagGen, an agentic framework that turns a single in-the-wild image into a simulation-ready deformable asset through a generate--simulate--diagnose--refine loop. DiagGen constructs part-aware geometry and material parameters, then uses a VLM (vision-language model)-based agent to select semantically informative regions, probe them in a physics simulator, observe material responses, and route evidence-backed repair cues to the responsible generation stage. Experiments on 40 assets show that diagnostics provides useful repair cues and can moderately improve the quality of generated deformable assets. Finally, we show that unlike assets generated from visual foundation models which may not be simulatable, DiagGen-generated deformables can be directly dropped into a high-fidelity physical simulator for the planning and simulation of contact-rich pick-and-place tasks. The project's website is https://diaggen.github.io/.
Chinese Translation
虽然适用于仿真的可变形资产对于计算机内机器人操作任务至关重要，但现有的生成框架通常在生成后评估物理合理性，未能利用对象的模拟响应作为修复上游错误的反馈。我们提出了DiagGen，这是一个自主框架，通过生成—模拟—诊断—精炼循环，将单张真实场景图像转化为适用于仿真的可变形资产。DiagGen构建了部件感知的几何形状和材料参数，然后使用基于视觉语言模型（VLM）的代理选择语义上有信息的区域，在物理模拟器中探测这些区域，观察材料响应，并将基于证据的修复线索传递给负责的生成阶段。在40个资产上的实验表明，诊断提供了有用的修复线索，并能适度提高生成的可变形资产的质量。最后，我们展示了与可能无法进行模拟的视觉基础模型生成的资产不同，DiagGen生成的可变形资产可以直接放入高保真物理模拟器中，用于接触丰富的抓取和放置任务的规划与仿真。项目网站为 https://diaggen.github.io/。
cs.RO / 62 / 2609.23113
Search, Ground, Plan: Functional Sufficiency for Task and Motion Planning under Incomplete Scene Knowledge
搜索、定位、规划：在不完整场景知识下的任务与运动规划的功能充分性
Abstract
Foundation models (FMs) have expanded task and motion planning (TAMP) to manipulation problems specified through language and visual observations. However, incomplete scene knowledge leaves a critical gap between understanding what the task requires and knowing whether the physical scene can actually realize it. We introduce GRAB-TAMP, an FM-based TAMP framework that searches for scene entities required for task completion, grounds functional roles to valid physical objects, and plans only after a complete joint assignment establishes functional sufficiency. We represent the task through functional roles, relations, and assignment constraints, and incrementally inspect the scene while requirements remain unresolved, verifying candidate objects through semantic, geometric, and relational checks. We evaluate GRAB-TAMP across 32 scene variants spanning Kitchen, Living Room, and Workshop domains. Across 200 feasible trials, our approach achieves 54.0% end-to-end success with 67.3% plan goal coverage. Compared with three FM-based TAMP frameworks under the same execution setting, GRAB-TAMP improves end-to-end success by 25.7 percentage points over the mean baseline. Implementation and evaluation code: https://github.com/Narendhiranv04/GRAB-TAMP
Chinese Translation
基础模型（FMs）已将任务与运动规划（TAMP）扩展到通过语言和视觉观察指定的操作问题。然而，不完整的场景知识在理解任务要求与知道物理场景是否能够实际实现之间留下了一个关键的空白。我们引入了GRAB-TAMP，一个基于FM的TAMP框架，该框架搜索任务完成所需的场景实体，将功能角色与有效的物理对象进行定位，并仅在完整的联合分配建立功能充分性后进行规划。我们通过功能角色、关系和分配约束来表示任务，并在需求未解决时逐步检查场景，通过语义、几何和关系检查验证候选对象。我们在涵盖厨房、客厅和车间领域的32种场景变体上评估了GRAB-TAMP。在200次可行的试验中，我们的方法实现了54.0%的端到端成功率，计划目标覆盖率为67.3%。与在相同执行设置下的三个基于FM的TAMP框架相比，GRAB-TAMP的端到端成功率比平均基线提高了25.7个百分点。实现和评估代码： https://github.com/Narendhiranv04/GRAB-TAMP
cs.RO / 63 / 2609.23118
Verti-WM: A Physics-Aided Exteroceptive World Model for Off-Road Reinforcement Learning
Verti-WM：一种物理辅助的外部感知世界模型用于越野强化学习
Abstract
Reinforcement learning for off-road navigation requires extensive vehicle-terrain interaction data, which are costly to collect in high-fidelity simulation. World models offer a promising alternative by replacing simulator roll-outs during policy optimization. However, an off-road world model must condition state transitions on exteroceptive terrain information, which proprioception alone does not provide. This challenge is further amplified by the need to model both rigid and deformable terrain, where data-driven and physics-based approaches offer complementary strengths. We propose Verti-WM, a physics-aided exteroceptive world model that recurrently fuses a frozen Transformer for rigid terrain and a neuro-symbolic terramechanics model for deformable terrain. Elevation and semantic observations queried from a supplied map at each predicted pose condition fusion, enabling six-degree-of-freedom rollouts for policy optimization without further simulator access. Verti-WM reduces prediction error by 34.6% and 21.7% over data-driven and physics-based baselines, respectively. Policies trained entirely within Verti-WM achieve comparable task success rates while reducing computation time by 23.6X relative to direct training in the high-fidelity simulator. We further validate Verti-WM using real-world data, enabling policy optimization within learned real-world kinodynamics and achieving a 80% success rate on the Verti-4-Wheeler platform, compared with 40% for direct sim-to-real transfer.
Chinese Translation
越野导航的强化学习需要大量的车辆与地形交互数据，而在高保真模拟中收集这些数据的成本很高。世界模型通过在策略优化过程中替代模拟器的展开提供了一种有前景的替代方案。然而，越野世界模型必须基于外部感知的地形信息来调节状态转移，而单靠本体感知无法提供这些信息。这个挑战在需要同时建模刚性和可变形地形时进一步加剧，数据驱动和基于物理的方法各自提供了互补的优势。我们提出了Verti-WM，一种物理辅助的外部感知世界模型，它通过循环融合一个冻结的Transformer用于刚性地形和一个神经符号土力学模型用于可变形地形。在每个预测姿态下，从提供的地图中查询的高程和语义观测条件融合，使得在没有进一步访问模拟器的情况下实现六自由度的展开用于策略优化。Verti-WM相较于数据驱动和基于物理的基线分别减少了34.6%和21.7%的预测误差。完全在Verti-WM内训练的策略在任务成功率上达到了可比的水平，同时计算时间相较于在高保真模拟器中的直接训练减少了23.6倍。我们进一步使用真实世界数据验证了Verti-WM，使得在学习到的真实世界运动动力学中进行策略优化，并在Verti-4-Wheeler平台上达到了80%的成功率，而直接的模拟到现实转移仅为40%。
cs.RO / 64 / 2609.23131
Selective Commitment for Language-Guided Object Retrieval under Partial Observability
在部分可观测性下的语言引导对象检索的选择性承诺
Abstract
Language-guided object retrieval under partial observability requires deciding whether to gather more evidence, interact with the scene, grasp a candidate, or abstain. We present a closed-loop framework that coordinates these decisions for retrieving a target specified in relation to a reference container. The framework maintains a persistent joint belief over target identity, container relation, and presence through tracked-object, unobserved-target, and target-absent hypotheses. View-conditioned categorical VLM observations update this belief; conformal grasp eligibility and robot feasibility govern commitment, while finite-horizon belief-space planning selects information-gathering actions. Across five different scenarios, our proposed method succeeds in 19/25 simulation episodes versus 12/25 for the best-performing task-adapted baseline and is the only evaluated policy to achieve at least one success in each scenario. Ablations show that cross-view memory improves success under partial occlusion, while the full system does not consistently outperform simplified variants. Real-robot trials demonstrate closed-loop re-observation and autonomous recovery from injected grasp failures, while injected viewpoint failures end in false defer. Experimental results demonstrate the feasibility of coordinating evidence gathering and selective grasp commitment within a unified framework for retrieval under partial observability.
Chinese Translation
在部分可观测性下，语言引导的对象检索需要决定是收集更多证据、与场景互动、抓取候选对象，还是放弃。我们提出了一个闭环框架，该框架协调这些决策，以检索与参考容器相关的目标。该框架通过跟踪对象、未观察到的目标和目标缺失假设，维持对目标身份、容器关系和存在性的持续联合信念。基于视图的分类 VLM 观察更新这一信念；符合条件的抓取资格和机器人可行性决定承诺，而有限视域的信念空间规划选择信息收集行动。在五种不同场景中，我们提出的方法在 25 次仿真实验中成功 19 次，而表现最佳的任务适应基线仅成功 12 次，并且是唯一在每个场景中至少取得一次成功的评估策略。消融实验表明，跨视图记忆在部分遮挡下提高了成功率，而完整系统并不总是优于简化变体。真实机器人试验展示了闭环重新观察和从注入的抓取失败中自主恢复的能力，而注入的视点失败则以错误的推迟告终。实验结果表明，在部分可观测性下，协调证据收集和选择性抓取承诺在统一框架内是可行的。
cs.RO / 65 / 2609.23132
Tip Manipulation in Soft Everting Robots via Wall Retraction and Deployable Fingers
通过壁收缩和可展开手指实现软性外翻机器人中的尖端操控
Abstract
Soft everting robots can traverse long, confined paths by continuously growing, yet active interaction remains limited to a single tool fixed at or near the tip, unable to be repositioned on demand and difficult to reconcile with the robot's soft body. We introduce a tip-manipulation and multi-tool deployment strategy for soft everting robots based on wall retraction, implemented with a base roller assembly that independently meters membrane flow in the outer wall while a tail spool regulates growth in the internal tail section. Coordinated wall and tail actuation decouples robot length from membrane-material position, enabling membrane-mounted devices to be transported, exposed, and repositioned at selected locations near the distal tip. We pair this capability with ultralight pleated inflatable fingers integrated into the membrane, fabricated from TPU-coated nylon with an internal airtight bladder. The fingers achieve large bending at low pressures (approximately 100 degrees at 50 kPa in high-pleat designs) and generate blocking forces up to 1.9 N, while remaining limp during transport. The system demonstrates adaptive grasping across diverse household objects (21 to 550 g; 16 to 200 mm), three-dimensional object manipulation and stacking, environmentally braced extension, distal camera panning for confined-space inspection, and controlled sequential payload delivery. These results enable embodied and reversible tip manipulation for soft growing robots in cluttered and tortuous environments.
Chinese Translation
软性外翻机器人能够通过持续生长在狭窄的长路径上移动，但其主动交互仍然局限于固定在尖端或附近的单一工具，无法按需重新定位，并且难以与机器人的软体结构协调。我们提出了一种基于壁收缩的尖端操控和多工具部署策略，采用一种基础滚轮组件，该组件独立调节外壁的膜流动，同时尾部卷轴调节内部尾部的生长。协调的壁和尾部驱动将机器人长度与膜材料位置解耦，使膜上安装的设备能够在尖端附近的选定位置进行运输、暴露和重新定位。我们将这一能力与集成在膜中的超轻褶皱充气手指相结合，这些手指由涂有TPU的尼龙制成，并配有内部气密气囊。这些手指在低压力下实现大幅度弯曲（在高褶皱设计中约100度，50 kPa），并产生高达1.9 N的阻挡力，同时在运输过程中保持柔软。该系统展示了对各种家用物体（21至550克；16至200毫米）的自适应抓取、三维物体操控和堆叠、环境支撑扩展、用于狭小空间检查的远端摄像头平移，以及受控的顺序载荷传递。这些结果使得在杂乱和曲折环境中实现软性生长机器人具身和可逆的尖端操控成为可能。
cs.RO / 66 / 2609.23133
AquaCap: A Training-Free Underwater Embodied Agent with Code-as-Policy
AquaCap：一种无训练的水下具身智能体，采用代码作为策略
Abstract
Recent advances in vision-language-action models have stimulated growing interest in underwater embodied intelligence. However, their reliance on large-scale interaction data limits their applicability underwater, where data collection is costly and scarce. To address this challenge, we present AquaCap, a training-free Code-as-Policy framework for autonomous underwater navigation and manipulation. AquaCap employs a dual-layer agent that translates task instructions and environmental observations into condition-aware plans and executable control programs. Structured perception then provides the agent with semantic, geometric, and reliability-aware observations under degraded underwater conditions. A failure-aware memory diagnoses unsuccessful actions and supports closed-loop replanning and code revision. This design enables online adaptation without task-specific training or parameter updates. AquaCap achieves a 66.43% success rate in simulation. Real-world experiments further demonstrate autonomous grasping and object transport with an ROV, including the manipulation of targets displaced by hydrodynamic disturbances.
Chinese Translation
近年来，视觉-语言-动作模型的进展激发了对水下具身智能的日益关注。然而，它们对大规模交互数据的依赖限制了其在水下的应用，因为数据收集成本高且稀缺。为了解决这一挑战，我们提出了AquaCap，一种无训练的代码作为策略框架，用于自主水下导航和操作。AquaCap采用双层智能体，将任务指令和环境观察转换为条件感知的计划和可执行的控制程序。结构化感知为智能体在恶劣水下条件下提供语义、几何和可靠性感知的观察。故障感知记忆诊断不成功的动作，并支持闭环重新规划和代码修订。该设计使得在线适应成为可能，无需特定任务的训练或参数更新。AquaCap在仿真中实现了66.43%的成功率。现实世界实验进一步展示了与遥控水下机器人（ROV）进行自主抓取和物体运输的能力，包括对受到水动力扰动影响的目标的操作。
cs.RO / 67 / 2609.23144
Probabilistic Scene Graphs: Hierarchical Representation and Real-time System
概率场景图：层次表示与实时系统
Abstract
3D scene graphs provide semantically rich and hierarchical representations for robot perception. However, existing systems do not maintain uncertainty as an explicit belief or propagate it through the operations that construct and refine the graph. We introduce Probabilistic Scene Graph (PSG), a generalization of the conventional scene graph that represents a posterior over possible graphs, factorized into a discrete graph structure of entities, relations, and semantic attributes, and continuous states that ground them spatially, with uncertainty maintained over both components. Geometry is carried directly by the nodes rather than selected from a separately constructed metric map, so a metric map, where needed, follows from the graph rather than preceding it. We instantiate PSG's probabilistic spatial grounding with hierarchical graphs of Gaussians (HGG): each object primitive is represented by a full-covariance Gaussian under a Normal-Inverse-Wishart belief, and the same parametrization applied recursively within a node yields a geometry graph that resolves its surface at finer resolution. We then build a mapping pipeline that preserves these beliefs throughout graph construction and refinement: a purely graph-based coarse-to-fine alignment registers observations by comparing node beliefs, while a nested Expectation-Maximization and factor-graph optimization jointly refines poses, object parameters, and internal geometry. Across six datasets spanning indoor RGB-D, outdoor LiDAR, and cross-modality deployment, HGG operates at sensor rate with near-constant memory and achieves state-of-the-art object accuracy and zero-shot graph alignment.
Chinese Translation
三维场景图为机器人感知提供了语义丰富且层次化的表示。然而，现有系统并未将不确定性作为显式信念进行维护，也未在构建和细化图的操作中传播不确定性。我们提出了概率场景图（Probabilistic Scene Graph, PSG），这是对传统场景图的推广，表示可能图的后验，分解为离散的实体、关系和语义属性的图结构，以及将其空间定位的连续状态，同时在这两个组件上保持不确定性。几何信息直接由节点承载，而不是从单独构建的度量地图中选择，因此在需要时，度量地图是从图中推导而来，而不是事先存在。我们通过高斯层次图（Hierarchical Graphs of Gaussians, HGG）实例化PSG的概率空间定位：每个对象原语在正态-逆威沙特信念下由全协方差高斯表示，并且在节点内递归应用相同的参数化会产生一个几何图，以更细的分辨率解析其表面。然后，我们构建一个映射管道，在图的构建和细化过程中保持这些信念：一种纯图的粗到细对齐通过比较节点信念来注册观察，而嵌套的期望最大化和因子图优化共同细化姿态、对象参数和内部几何。在涵盖室内RGB-D、室外LiDAR和跨模态部署的六个数据集上，HGG以传感器速率运行，内存几乎保持不变，并实现了最先进的对象精度和零样本图对齐。
cs.RO / 68 / 2609.23252
Robot World Models Are Not Invariant to How the Actions Are Written
机器人世界模型对动作书写方式并不不变
Abstract
A robot policy is trained with one of two action parameterizations: absolute joint targets, or deltas relative to the current state. The choice is a live engineering decision in robot learning, and a world model conditioned on actions inherits it silently. We show the inheritance is catastrophic. A latent dynamics model trained on one parameterization and handed the identical commanded trajectory written in the other collapses: retrieval degrades by 2.6-13.4x across three robot datasets and two morphologies, goal-conditioned action selection falls from 53% to 15%, and on PushT the two beliefs about the same future are near-orthogonal (cos = 0.067, worst case -0.377), so the predictor does not degrade gracefully, it answers a different question. This is not a distribution-shift artifact in the usual sense: the two encodings are mutually reconstructible at R^2 = 0.996 given the joint input, so no information is lost, and we give the test that separates a valid re-parameterization from a lossy summary or a sensor swap. The test rejected three of the four axes we proposed. The defect lives in the action channel, which the invariance literature for visual models does not examine: work there concerns crops, jitter and camera pose, while the parameterization of the commands goes unaudited. The repair is averaging over the two encodings, and where it goes matters. Averaging the objective restores task performance by itself; averaging the outputs, safe for probabilities by concavity, is not available for direction-valued prediction, where the normalized mean can score below every member of the orbit. What objective-averaging leaves behind is the tail: worst-case agreement stays at 0.78, a disagreement penalty closes it to 0.995, and over a latent rollout it is the difference between a worst case that erodes and one that holds. On PushT, averaging alone does not repair the axis.
Chinese Translation
机器人策略使用两种动作参数化之一进行训练：绝对关节目标或相对于当前状态的增量。这个选择在机器人学习中是一个实时的工程决策，而基于动作的世界模型默默地继承了这一点。我们展示了这种继承是灾难性的。一个基于一种参数化训练的潜在动态模型，在接收到用另一种参数化书写的相同指令轨迹时会崩溃：在三个机器人数据集和两种形态下，检索性能下降了2.6-13.4倍，目标条件下的动作选择从53%降至15%，在PushT上，对同一未来的两种信念几乎是正交的（余弦相似度 = 0.067，最坏情况 -0.377），因此预测器并没有优雅地退化，而是回答了一个不同的问题。这并不是通常意义上的分布偏移伪影：在给定联合输入的情况下，这两种编码在R^2 = 0.996下是相互可重构的，因此没有信息丢失，我们提供了一个测试，以区分有效的重新参数化与有损摘要或传感器交换。该测试拒绝了我们提出的四个轴中的三个。缺陷存在于动作通道中，而视觉模型的不变性文献并未对此进行研究：那里的工作关注于裁剪、抖动和相机姿态，而命令的参数化则未经过审计。修复方法是对这两种编码进行平均，而其去向至关重要。单独平均目标可以恢复任务性能；对输出进行平均，在概率上由于凹性是安全的，但在方向值预测中不可用，因为标准化均值可能低于轨道中的每个成员。目标平均留下的是尾部：最坏情况的一致性保持在0.78，分歧惩罚将其缩小至0.995，而在潜在展开中，这是一个逐渐恶化的最坏情况与一个保持不变的最坏情况之间的差异。在PushT上，仅仅平均并不能修复该轴。
cs.RO / 69 / 2609.23263
Scenario MPC with STL Specifications and Pareto-Based Feasibility Repair
具有 STL 规范和基于帕累托的可行性修复的情景模型预测控制
Abstract
Temporal logic is a formal language for reasoning about system behaviors over time. Signal temporal logic (STL), in particular, has been used to encode spatio-temporal requirements for control synthesis in multi-agent systems, often under the assumption that agents are cooperative and their dynamics are known. However, real-world multi-agent applications, such as autonomous driving, typically involve stochastic and uncontrollable agents. Recent work explored robust control with worst-case or probabilistic formulations, but remains limited in that it either (1) certifies strict satisfaction of STL constraints without addressing feasibility recovery, or (2) relaxes infeasible constraints with ego-centric objectives. In this paper, we propose a model predictive control (MPC) framework that treats feasibility repair as a Pareto optimization problem to explicitly characterize tradeoffs among agent objectives. We further provide a probabilistic certificate on STL violation rate to formally quantify uncertainty under stochastic and uncontrollable agents. The proposed framework is evaluated on two autonomous driving scenarios. Results show that the framework recovers feasible control with demonstrated safe behaviors.
Chinese Translation
时序逻辑是一种用于推理系统行为随时间变化的形式语言。特别是，信号时序逻辑（STL）已被用于编码多智能体系统中控制合成的时空要求，通常假设智能体是合作的且其动态是已知的。然而，现实世界中的多智能体应用，如自动驾驶，通常涉及随机和不可控的智能体。近期的研究探讨了使用最坏情况或概率形式的鲁棒控制，但仍然存在局限性，即要么（1）证明严格满足 STL 约束而不解决可行性恢复，要么（2）以自我中心的目标放宽不可行的约束。本文提出了一种模型预测控制（MPC）框架，将可行性修复视为一个帕累托优化问题，以明确表征智能体目标之间的权衡。我们进一步提供了关于 STL 违反率的概率证明，以正式量化在随机和不可控智能体下的不确定性。所提出的框架在两个自动驾驶场景中进行了评估。结果表明，该框架能够恢复可行控制，并展示出安全行为。
cs.RO / 70 / 2609.23269
Latent Telepathy: Multi-Robot Communication with Self-Supervised Perceptual Latents
潜在心灵感应：自监督感知潜变量的多机器人通信
Abstract
In a decentralized multi-robot team under partial observability, the fact that decides a robot's next action is often visible only to a teammate. Existing decentralized methods communicate kinematic information, such as position or planned trajectory, which cannot convey what the teammate perceives. Learned communication in multi-agent reinforcement learning (MARL) can carry perceptual content, but the resulting messages are task-coupled and opaque. We propose Latent Telepathy. Each robot broadcasts the perceptual latent vector it already computes for its own use, the output of an encoder trained with a self-supervised joint-embedding predictive objective, frozen, and shared across the team. A teammate learns to act on it from task reward alone. Because the encoder already runs for perception, the message costs no additional computation and a single compact vector of bandwidth. Because the encoder is frozen before any policy is trained, the message means the same thing to every robot, and the receiving robot is never told what it means. We evaluate Latent Telepathy with a content-controlled protocol in which bandwidth, latency, topology and receiver are held fixed and only the message content varies. Broadcasting the latent lets a navigator avoid an occluded hazard in 99.7% of episodes, matching a noiseless hand-designed message. Position and trajectory messages remain at chance, and the raw camera image, 186 times wider, is less reliable than the compressed latent. The result holds from a discrete gridworld to rendered pixels under continuous velocity control, and the encoder decodes the hazard from a physical robot's camera in 102 of 102 live decisions. We also identify a requirement for porting MARL communication results to continuous control, that the decision a message informs must remain reachable by exploration, and show how to restore it.
Chinese Translation
在部分可观测的去中心化多机器人团队中，决定机器人下一步行动的因素通常仅对队友可见。现有的去中心化方法传递运动学信息，如位置或规划轨迹，但无法传达队友所感知的内容。在多智能体强化学习（MARL）中，学习到的通信可以携带感知内容，但所产生的信息是与任务相关且不透明的。我们提出了潜在心灵感应（Latent Telepathy）。每个机器人广播其已计算的感知潜变量向量，该向量是通过自监督的联合嵌入预测目标训练的编码器的输出，保持不变并在团队中共享。队友仅通过任务奖励学习如何基于该信息采取行动。由于编码器已经用于感知，因此该消息不需要额外的计算，并且只占用一个紧凑的带宽向量。由于编码器在任何策略训练之前被冻结，因此该消息对每个机器人意味着相同的内容，而接收机器人从未被告知其含义。我们通过一种内容控制协议评估潜在心灵感应，在该协议中，带宽、延迟、拓扑结构和接收者保持固定，仅消息内容变化。广播潜在向量使导航者在99.7%的情境中避免被遮挡的危险，效果与无噪声的手工设计消息相匹配。位置和轨迹消息保持在随机水平，而原始相机图像的宽度是其186倍，但可靠性低于压缩的潜在向量。该结果适用于从离散网格世界到在连续速度控制下渲染的像素，编码器在102个实时决策中成功解码物理机器人相机中的危险。我们还确定了将MARL通信结果移植到连续控制的一个要求，即消息所传达的决策必须通过探索保持可达，并展示了如何恢复这一点。
cs.RO / 71 / 2609.23271
Steering Through Contact: A Finite-Support Motion Model for Single-Track Center-Articulated Robots
通过接触引导：单轨中心铰接机器人有限支撑运动模型
Abstract
Trajectory planning and control in field robotics rely on predicting how propulsion and steering affect vehicle motion when contact points undergo slip. For articulated vehicles, the point-contact kinematic model (PCK) accounts for the linkage geometry but neglects the rotational resistance distributed along the contacts. We propose a finite-support quadratic model (FSQ) for single-track, center-articulated vehicles that incorporates this resistance through a quasi-static balance of lateral slip. Our approach generalizes the standard PCK formulation by relaxing the contact-point assumption. An exact reduction of the quadratic slip cost to contact moments gives a compact closed-form solution for real-time prediction of lateral velocity and yaw rate. We evaluate the proposed method in real-world experiments across asphalt, grass, ice, and mixed routes, using more than 7 km of data. For five-second predictions, FSQ reduces the weighted median translation and yaw errors by 53.5% and 68.9%, respectively, relative to PCK.
Chinese Translation
在野外机器人领域，轨迹规划和控制依赖于预测推进和转向如何影响车辆运动，尤其是在接触点发生滑移时。对于铰接车辆，点接触运动学模型（PCK）考虑了连杆几何形状，但忽略了沿接触点分布的旋转阻力。我们提出了一种有限支撑二次模型（FSQ），用于单轨中心铰接车辆，该模型通过准静态的侧向滑移平衡引入了这种阻力。我们的方法通过放宽接触点假设，推广了标准的PCK公式。将二次滑移成本精确简化为接触力矩，提供了用于实时预测侧向速度和偏航率的紧凑封闭形式解。我们在沥青、草地、冰面和混合路线的实际实验中评估了所提方法，使用了超过7公里的数据。对于五秒的预测，FSQ相对于PCK分别减少了53.5%和68.9%的加权中位数平移和偏航误差。
cs.RO / 72 / 2609.23275
SCULPT-VLA: Learning Structured Control through Staged Action Grounding
SCULPT-VLA：通过分阶段动作基础学习结构化控制
Abstract
Vision-language-action (VLA) policies increasingly incorporate structured intermediate supervision beyond action labels. Yet specifying what an intermediate representation should encode leaves open how action prediction learns to depend on it. We introduce \textbf{SCULPT-VLA}, a policy that learns structured control through staged action grounding. Its action-conditioning state comprises complementary factors for task progression, scene dynamics, and spatial grounding. Training first forms these factors with teacher scaffolds, then grounds coarse action prediction through their composition as scaffold inputs are withdrawn. Direct perceptual access is subsequently restored for continuous refinement, combining the learned state with perceptual detail. The curriculum separates learning to condition actions on structure from refining continuous control. Deployment requires neither teachers nor discrete-action autoregression. SCULPT-VLA achieves higher average success than shared-backbone baselines on LIBERO, SimplerEnv-WidowX, and RoboTwin 2.0 Full. On SimplerEnv-WidowX, final success is 83.5\%, versus 71.3\% when Stage-II action learning directly accesses vision and language. Across four physical robot tasks, average success under the tested distribution shifts reaches 58.1\%, compared with 45.6\% for $\pi_{0.5}$. Training ablations and factor-wise interventions support the staged design and show that the learned state continues to contribute to control after direct perceptual access is restored.
Chinese Translation
视觉-语言-动作（VLA）策略越来越多地结合了超越动作标签的结构化中间监督。然而，指定中间表示应该编码的内容仍然开放了动作预测如何依赖于它的问题。我们引入了 extbf{SCULPT-VLA}，一种通过分阶段动作基础学习结构化控制的策略。其动作条件状态包括任务进展、场景动态和空间基础的互补因素。训练首先通过教师支架形成这些因素，然后在撤回支架输入时通过它们的组合来基础粗略的动作预测。随后恢复直接的感知访问以进行持续的细化，将学习到的状态与感知细节相结合。课程将基于结构的动作条件学习与细化连续控制分开。部署不需要教师或离散动作自回归。SCULPT-VLA在LIBERO、SimplerEnv-WidowX和RoboTwin 2.0 Full上的平均成功率高于共享骨干基线。在SimplerEnv-WidowX上，最终成功率为83.5\%，而当Stage-II动作学习直接访问视觉和语言时为71.3\\%。在四个物理机器人任务中，测试分布下的平均成功率达到58.1\\%，而$ ext{π}_{0.5}$为45.6\\%。训练消融实验和因素干预支持分阶段设计，并表明学习到的状态在恢复直接感知访问后继续对控制产生贡献。
cs.RO / 73 / 2609.23280
Safety-Critical Control under Uncertainty via Adaptive Conformal Quantile Prediction Intervals
基于自适应符合量预测区间的不确定性下的安全关键控制
Abstract
Safety-critical control under uncertainty requires uncertainty representations that are both statistically valid (for certifiable performance) and compatible with enforceable safety constraints. However, existing methods often assume particular distributions of uncertainty for provable safety guarantees or establish symmetric and input-agnostic prediction intervals for robust safety, which can lead to misaligned or overly conservative safety constraints in control synthesis. In this paper, we introduce a novel safe control framework with adaptive uncertainty quantification that constructs calibrated and state-dependent prediction intervals to enable high-probability safety guarantees, while improving constrained control performance. The framework leverages adaptive conformal prediction (ACP) and extends it with conformal quantile regression (CQR) to capture distribution-free, asymmetric uncertainty intervals with certifiable probabilistic coverage, and integrates the resulting uncertainty sets into a probabilistic control barrier function formulation to enforce robust safety with reduced conservativeness. This yields uncertainty-aware safe control constraints that can be incorporated within a model predictive control(MPC) framework to provide provably safe behaviors with high probability. Simulation and theoretical results are provided to demonstrate the effectiveness of our approach.
Chinese Translation
在不确定性下的安全关键控制需要既具有统计有效性（以确保可证明的性能）又与可执行的安全约束兼容的不确定性表示。然而，现有方法通常假设不确定性的特定分布以提供可证明的安全保证，或者建立对输入无关的对称预测区间以实现稳健安全，这可能导致控制合成中的安全约束不一致或过于保守。本文提出了一种新颖的安全控制框架，采用自适应不确定性量化，构建校准的、状态依赖的预测区间，以实现高概率的安全保证，同时提高受限控制性能。该框架利用自适应符合预测（Adaptive Conformal Prediction, ACP），并通过符合量回归（Conformal Quantile Regression, CQR）进行扩展，以捕捉具有可证明概率覆盖的无分布、非对称不确定性区间，并将所得的不确定性集成到概率控制障碍函数的形式中，以减少保守性并强制执行稳健安全。这产生了考虑不确定性的安全控制约束，可以纳入模型预测控制（Model Predictive Control, MPC）框架中，以提供高概率的可证明安全行为。我们提供了仿真和理论结果，以证明我们方法的有效性。
cs.RO / 74 / 2609.23305
Shared Execution-Clock Drifting Policy for Dynamic Precision Manipulation
动态精度操控的共享执行时钟漂移策略
Abstract
Manipulation under time constraints requires both accurate actions and an execution rhythm that matches the evolving scene. This becomes critical when a robot must intercept moving objects or complete a sequence of adjustments before a deadline. Although one-step policies reduce generation cost, their directly predicted action sequences leave temporal allocation implicit. We propose Shared Execution-Clock Drifting (SECD), which makes execution rhythm an explicit part of one-step action generation. Conditioned on an observation and a latent sample, the policy jointly predicts a progress-indexed action curve and a shared monotone clock that maps fixed control times to locations on the curve. Demonstration-derived alignment anchors this decomposition, which is trained jointly through drifting on the decoded actions. The resulting policy retains a fixed-rate control interface and requires one network evaluation. We evaluate SECD across four real-robot tasks with inference on NVIDIA Thor. Across 300 trials, it achieves 77.00% task-averaged success and outperforms the evaluated one-step baselines on every task, including 91% success in cup retrieval from a 16 m/min conveyor and 54% in restoring and folding a crumpled shirt within 90 s. A fixed-clock variant reaches 79% on the same conveyor protocol. Complementary state-based RoboMimic experiments, including cross-seed ablations on Transport and Square, further support the joint design of the temporal representation and demonstration alignment. Project page: https://secd-anonymous-ewn.pages.dev/
Chinese Translation
在时间约束下的操控需要既准确的动作，又要与不断变化的场景相匹配的执行节奏。当机器人必须拦截移动物体或在截止日期之前完成一系列调整时，这一点尤为重要。尽管单步策略降低了生成成本，但其直接预测的动作序列使时间分配变得隐含。我们提出了共享执行时钟漂移（Shared Execution-Clock Drifting, SECD），将执行节奏作为单步动作生成的一个显性部分。在观察和潜在样本的条件下，该策略共同预测一个进度索引的动作曲线和一个共享的单调时钟，该时钟将固定控制时间映射到曲线上的位置。基于示范的对齐锚定了这一分解，该分解通过对解码动作的漂移进行联合训练。最终的策略保持了固定速率的控制接口，并只需进行一次网络评估。我们在四个真实机器人任务上评估了SECD，并在NVIDIA Thor上进行推理。在300次试验中，其任务平均成功率达到77.00%，并在每个任务上均优于评估的单步基线，包括在16 m/min的传送带上成功检索杯子的91%和在90秒内恢复和折叠皱衬衫的54%。一个固定时钟变体在相同的传送带协议上达到了79%。补充的基于状态的RoboMimic实验，包括对Transport和Square的交叉种子消融，进一步支持了时间表示和示范对齐的联合设计。项目页面：https://secd-anonymous-ewn.pages.dev/
cs.RO / 75 / 2609.23312
Manipulation Feasible Navigation Among Movable Obstacles with Discrete Contact Pushing
可移动障碍物间的操控可行导航与离散接触推动
Abstract
In environments with large movable obstacles, detour-only navigation can be inefficient or even infeasible, while obstacle interaction requires reasoning about navigation benefit, feasible placement, and executable manipulation. We present a hierarchical navigation among movable obstacles (NAMO) framework for mobile manipulators. At the high level, the planner identifies key blocking obstacles from reference paths and searches for relocation plans that jointly satisfy geometric, manipulation, and downstream navigation constraints. When direct relocation is hindered by other movable objects, a large language model (LLM) is selectively invoked to infer auxiliary manipulation dependencies, which are then verified by deterministic geometric planning. To execute the resulting relocation goals, we define discrete contact modes on the surfaces of box-shaped obstacles and select contact faces and regions online based on position and orientation errors, enabling straight, side, and corner pushing through contact switching. A recurrent reinforcement-learning policy coordinates the mobile base and manipulator to track tool center point (TCP) targets while preserving end-effector reachability during sustained pushing. Simulation and real-robot experiments demonstrate feasible navigation-manipulation in detour, single- and multi-obstacle relocation, and dependency-constrained scenarios, validating the framework for interactive navigation with large non-graspable obstacles. The open-source project is available at https://cloudytosunny.github.io/NAMO_DCPushing/.
Chinese Translation
在存在大型可移动障碍物的环境中，仅依靠绕行导航可能效率低下甚至不可行，而与障碍物的交互则需要对导航收益、可行放置和可执行操控进行推理。我们提出了一种针对移动操控器的可移动障碍物导航（NAMO）框架。在高层次上，规划器从参考路径中识别关键阻挡障碍物，并搜索共同满足几何、操控和下游导航约束的重新定位计划。当直接重新定位受到其他可移动物体的阻碍时，选择性地调用大型语言模型（LLM）来推断辅助操控依赖关系，随后通过确定性几何规划进行验证。为了执行生成的重新定位目标，我们在箱形障碍物的表面上定义了离散接触模式，并根据位置和方向误差在线选择接触面和区域，从而实现通过接触切换进行直推、侧推和角推。一个递归强化学习策略协调移动底盘和操控器，以在持续推动过程中跟踪工具中心点（TCP）目标，同时保持末端执行器的可达性。仿真和真实机器人实验展示了在绕行、单个和多个障碍物重新定位以及依赖约束场景中的可行导航-操控，验证了该框架在与大型不可抓取障碍物的交互导航中的有效性。该开源项目可在 https://cloudytosunny.github.io/NAMO_DCPushing/ 获取。
cs.RO / 76 / 2609.23392
Towards Reliable Underwater Diver-Robot Interaction: Gesture Design, Interaction Logic, and Real-World Evaluation
迈向可靠的水下潜水员-机器人交互：手势设计、交互逻辑与现实世界评估
Abstract
Underwater human--robot interaction requires gesture commands that are both easy for divers to use and reliable for robots to recognize. We investigate these aspects through a closed-loop diver--robot interaction framework integrating a compact seven-gesture vocabulary, lightweight landmark-based recognition, and command-level interaction logic. We evaluate the framework through a user study and underwater robot experiments in a laboratory tank and a swimming pool. The user study supported the reproducibility of the gestures after brief learning. Recognition analysis further showed that visual similarity was associated with gesture confusion, while intermediate poses during gesture formation introduced temporal ambiguity. Command-level processing mitigated the effects of transient recognition errors on robot execution, reducing unintended triggers and premature task interruptions. These findings show that reliable underwater gesture interaction depends on human usability, gesture recognizability, and execution reliability in underwater interaction.
Chinese Translation
水下人机交互需要易于潜水员使用且机器人能够可靠识别的手势命令。我们通过一个闭环的潜水员-机器人交互框架来研究这些方面，该框架整合了紧凑的七个手势词汇、轻量级的基于地标的识别以及命令级的交互逻辑。我们通过用户研究和在实验室水槽及游泳池中的水下机器人实验对该框架进行了评估。用户研究支持了在短时间学习后手势的可重复性。识别分析进一步表明，视觉相似性与手势混淆相关，而手势形成过程中的中间姿态引入了时间上的模糊性。命令级处理减轻了瞬态识别错误对机器人执行的影响，减少了意外触发和任务提前中断。这些发现表明，可靠的水下手势交互依赖于人类可用性、手势可识别性和水下交互中的执行可靠性。
cs.RO / 77 / 2609.23414
EmoPose: Vision-Language Model Guided Emotion-Aware Gesture Generation for Humanoid Robots
EmoPose：基于视觉-语言模型引导的情感感知手势生成用于类人机器人
Abstract
Socially competent humanoid robots must communicate affect and intent through gesture as well as speech, yet open-ended interaction must become motion that is both expressive and executable on a specific body. This demands semantic flexibility for contextual social intent while preserving deterministic, embodiment-aware robot control. We present EmoPose, a vision-language model (VLM)-guided framework that bridges this gap through an executable semantic interface. Given language, dialogue history, and optional visual context, the VLM selects an ordered gesture plan containing a communicative class, library variant, intensity, and speech anchor. A scalable robot-owned motion library defines the available expressive vocabulary and the source of 14-DoF joint targets. Pose Studio supports automatic trajectory generation, MuJoCo preview, and automatic synchronization of new library entries with the VLM guide; deterministic robot-side modules validate plans, construct trajectories, schedule gestures, and manage queueing and interruption. This division lets the interaction repertoire grow for new social contexts without changing the control interface or delegating raw joint commands to the foundation model. On the EmoPose-Bench, structured GPT-5.5 planning reaches $98.25\pm0.52\%$ on the Easy tier and $76.50\pm0.54\%$ overall, exceeding same-model direct-label prompting. Further tests validate dialogue-context use and ordered multi-action composition. The system completes the nominal MuJoCo suite and realizes all 29 authored variants on the physical Unitree G1. A four-stop laboratory tour demonstrates expressive narration with interruption, camera-grounded dialogue, and navigation.
Chinese Translation
具备社会能力的类人机器人必须通过手势和语言传达情感和意图，然而开放式互动必须转化为在特定身体上既富有表现力又可执行的动作。这要求在保持可预测的、具身意识的机器人控制的同时，具备语义灵活性以适应上下文社交意图。我们提出了EmoPose，一个通过可执行语义接口弥合这一差距的视觉-语言模型（VLM）引导框架。给定语言、对话历史和可选的视觉上下文，VLM选择一个有序的手势计划，其中包含一个交流类别、库变体、强度和语言锚点。一个可扩展的机器人拥有的运动库定义了可用的表现词汇和14自由度关节目标的来源。Pose Studio支持自动轨迹生成、MuJoCo预览，以及与VLM引导的新的库条目的自动同步；确定性的机器人端模块验证计划、构建轨迹、调度手势，并管理排队和中断。这种划分使得互动词汇能够在新的社交上下文中增长，而无需更改控制接口或将原始关节命令委托给基础模型。在EmoPose-Bench上，结构化的GPT-5.5规划在简单层级上达到$98.25 ext{±}0.52 ext{%}$，整体达到$76.50 ext{±}0.54 ext{%}$，超越了同模型的直接标签提示。进一步的测试验证了对话上下文的使用和有序多动作组合。该系统完成了标准的MuJoCo套件，并在物理Unitree G1上实现了所有29个创作变体。一个四站的实验室巡演展示了富有表现力的叙述与中断、基于相机的对话和导航。
cs.RO / 78 / 2609.23418
HEARTH: An Object-Centric RGB-Thermal-3D Dataset for Temperature-Aware Robot Manipulation
HEARTH：一个面向对象的RGB-热成像-3D数据集，用于温度感知的机器人操控
Abstract
Language-guided manipulation can depend on physical properties that visible appearance does not reveal. Temperature is one such property, but object datasets for robot learning rarely associate measured temperatures with object appearance and geometry. We present HEARTH, an object-centric RGB-thermal-3D dataset of 90 physical objects from 18 everyday categories, comprising 145 captured object states. Our pipeline maps apparent surface temperatures onto reconstructed meshes through camera calibration and pose transfer. The dataset includes raw temperature measurements, camera parameters, RGB-textured meshes, and thermal textures for simulation. We use these assets to construct three LIBERO-derived tasks and collect 1,200 demonstrations for fine-tuning a pretrained vision-language-action (VLA) model, $\pi_{0.5}$. In an ablation study, adding thermal observations to the VLA increases success on temperature-dependent object-selection tasks from 35.0% for the RGB-only baseline to 75.0%. These results demonstrate the utility of HEARTH for training robot policies to follow temperature-related instructions.
Chinese Translation
语言引导的操控可以依赖于可见外观无法揭示的物理属性。温度就是这样一种属性，但用于机器人学习的物体数据集很少将测量的温度与物体的外观和几何形状关联起来。我们提出了HEARTH，这是一个面向对象的RGB-热成像-3D数据集，包含来自18个日常类别的90个物理对象，共145个捕获的物体状态。我们的流程通过相机标定和姿态转移，将表面温度映射到重建的网格上。该数据集包括原始温度测量、相机参数、RGB纹理网格和用于仿真的热纹理。我们利用这些资源构建了三个基于LIBERO的任务，并收集了1,200个演示以微调预训练的视觉-语言-动作（VLA）模型，$ ext{π}_{0.5}$。在消融研究中，将热观察添加到VLA中使得在温度依赖的物体选择任务上的成功率从RGB-only基线的35.0%提高到75.0%。这些结果证明了HEARTH在训练机器人政策以遵循与温度相关的指令方面的实用性。
cs.RO / 79 / 2609.23423
RiverVLN: Phase-Grounded Temporal Vision--Language Navigation for Unmanned Surface Vehicles
RiverVLN：基于相位的无人水面车辆时序视觉-语言导航
Abstract
Vision-language navigation (VLN) has largely been developed for indoor and terrestrial robots, where language can often be treated as a static goal and motion is approximated by discrete or near-instantaneous actions. These assumptions break down for unmanned surface vehicles (USVs): river navigation requires continuous motion under inertia and limited maneuverability, while long-horizon instructions must be executed through sparse and visually ambiguous maritime landmarks. We introduce RiverVLN, to our knowledge the first benchmark designed for long-horizon USV VLN under continuous riverine motion, and PGT-NAV, a phase-grounded temporal navigation framework for USVs. Rather than directly mapping an entire instruction to motion, PGT-NAV converts it into an ordered sequence of visually verifiable semantic phases and maintains the active phase online through grounded visual and motion evidence. This explicit semantic progress state is fused with visual-motion history and phase-specific grounding to predict six local SE(2) pose increments. The resulting trajectory is executed in a predict-execute-re-observe loop, where the vessel executes toward W3, updates phase and grounding, and replans through a map-based safety layer. Experiments show that PGT-NAV substantially reduces recursive position and heading drift relative to GNM-style and ViNT-style baselines and achieves an average success rate of 0.79 in Unity-ROS closed-loop navigation. Unseen bridge-opening trials and real-world USV experiments further demonstrate that the phase-grounded representation transfers from controlled evaluation to physical USV deployment.
Chinese Translation
视觉-语言导航（VLN）在室内和陆地机器人中得到了广泛的发展，在这些场景中，语言通常可以被视为静态目标，运动则通过离散或近乎瞬时的动作来近似。然而，这些假设在无人水面车辆（USVs）中并不成立：河流导航需要在惯性和有限机动性下进行连续运动，而长时间跨度的指令必须通过稀疏且视觉上模糊的海洋地标来执行。我们提出了RiverVLN，这是我们所知的第一个针对在连续河流运动下进行长时间跨度USV VLN的基准，以及PGT-NAV，一个针对USVs的基于相位的时序导航框架。PGT-NAV并不是直接将整个指令映射到运动，而是将其转换为一系列可视化可验证的语义阶段，并通过基于视觉和运动的证据在线维护活动阶段。这个显式的语义进展状态与视觉-运动历史和特定阶段的基础相结合，以预测六个局部SE(2)姿态增量。生成的轨迹在预测-执行-重新观察的循环中执行，船只朝向W3执行，更新阶段和基础，并通过基于地图的安全层重新规划。实验表明，PGT-NAV相较于GNM风格和ViNT风格的基线显著减少了递归位置和航向漂移，并在Unity-ROS闭环导航中实现了0.79的平均成功率。未见的桥梁开启试验和真实世界的USV实验进一步证明了基于相位的表示能够从受控评估转移到物理USV部署。
cs.RO / 80 / 2609.23432
RopeFormer: Cross-Trial Adaptation from Interaction History for Dynamic Rope Manipulation
RopeFormer：基于交互历史的跨试验适应用于动态绳索操控
Abstract
Dynamic rope manipulation is highly sensitive to unknown object dynamics: the same robot motion can produce substantially different responses across ropes, while explicitly identifying the relevant physical properties is difficult. We present RopeFormer, a history-conditioned framework that uses prior task interaction as context for subsequent control. The policy retains cross-trial action-response history while keeping its weights fixed and requires no explicit online rope-parameter estimation. In matched simulation evaluations across sustained single-arm rotation, bimanual rotation, and transient whipping, retaining context improves subsequent control relative to resetting the same checkpoint, with the benefit varying across rope dynamics and observation settings. We further deploy the frozen policies on a Unitree H1-2 with previously unseen physical ropes. From T1 to T3, target-acquisition time decreases by 30.9% for Rope Swing and 33.9% for Rope Twirl, while mean Rope Whip target hits increase from 0.2 to 2.3 out of three. These results show that prior interaction can provide effective control context for dynamic deformable-object manipulation. Robot videos, code, and data are available at https://ropeformer.github.io/.
Chinese Translation
动态绳索操控对未知物体动力学高度敏感：相同的机器人运动在不同绳索上可能产生显著不同的反应，而明确识别相关的物理属性则较为困难。我们提出了RopeFormer，一个基于历史条件的框架，利用先前任务交互作为后续控制的上下文。该策略保留跨试验的动作-反应历史，同时保持其权重不变，并且不需要显式的在线绳索参数估计。在持续的单臂旋转、双手旋转和瞬态鞭打的匹配仿真评估中，保留上下文相较于重置相同检查点能够改善后续控制，其效果在不同绳索动力学和观察设置中有所不同。我们进一步在一台Unitree H1-2机器人上部署了这些固定策略，使用之前未见过的物理绳索。从T1到T3，Rope Swing的目标获取时间减少了30.9%，Rope Twirl减少了33.9%，而平均Rope Whip目标命中率从0.2提高到2.3（满分为三）。这些结果表明，先前的交互可以为动态可变形物体操控提供有效的控制上下文。机器人视频、代码和数据可在https://ropeformer.github.io/获取。
cs.RO / 81 / 2609.23439
Receding-Horizon Pushing with Composable Object-Centric Policies
可组合对象中心策略的后退视野推动
Abstract
Non-prehensile manipulation is practical for relocating large, heavy, or geometrically ungraspable objects. Yet, long-horizon pushing of arbitrarily-shaped 3D objects couples three problems: 1) where to push the object so as to approach the target pose, 2) whether each push is stable and reachable, 3) whether subsequent actions remain feasible. We present an object-centric pushing policy within a feedback-guided hierarchical framework. At the low level, a learning-based policy predicts contact actions from a pose- and scale-normalized point cloud, conditioned on a near single-step subgoal. A stability score is applied to evaluate the predicted contacts by a quasi-static sliding-versus-tipping analysis. At the high level, BIT$^*$ first searches for an object path, and the next several subgoals are checked by contact prediction and robot motion planning for future feasibility. Failed motion plans, as feedback, change the local path costs and trigger re-planning. During execution, only the first feasible action is executed. In simulation, we evaluate 22 objects in six different scenes, upon which we also conduct comprehensive ablation studies. Results demonstrate that our method outperforms baselines with a clear margin and can reliably achieve long-horizon object pushing tasks under different situations. We also report quantitative real-robot experiments with a Franka arm and qualitative demonstrations with a mobile manipulator for large and heavy objects, with directly zero-shot sim-to-real transfer.
Chinese Translation
非抓取操作对于重新定位大型、重型或几何上无法抓取的物体是实用的。然而，任意形状的三维物体的长时间推动涉及三个问题：1）如何推动物体以接近目标姿态，2）每次推动是否稳定且可达，3）后续动作是否仍然可行。我们在一个反馈引导的分层框架内提出了一种对象中心的推动策略。在低层次上，基于学习的策略从姿态和尺度归一化的点云中预测接触动作，条件是接近单步子目标。通过准静态滑动与倾倒分析，应用稳定性评分来评估预测的接触。在高层次上，BIT$^*$首先搜索物体路径，然后通过接触预测和机器人运动规划检查接下来的几个子目标的未来可行性。失败的运动规划作为反馈，改变局部路径成本并触发重新规划。在执行过程中，仅执行第一个可行动作。在仿真中，我们评估了六个不同场景中的22个物体，并进行了全面的消融研究。结果表明，我们的方法在明显的边际上优于基线，并且能够在不同情况下可靠地实现长时间的物体推动任务。我们还报告了使用Franka手臂的定量真实机器人实验和使用移动操纵器对大型和重型物体的定性演示，实现了直接的零样本模拟到现实转移。
cs.RO / 82 / 2609.23445
BiRoAD: Learning Shared and Role-Adaptive Representations for Bimanual Manipulation
BiRoAD：用于双手操作的共享与角色自适应表示学习
Abstract
Bimanual manipulation requires policies that coordinate two arms while adapting their functional roles to scene geometry, object configuration, and task context. Learning such scene-conditioned role adaptation remains challenging, as demonstrations may contain uneven role distributions that limit generalization to underrepresented arm--role configurations. In addition, many bimanual policies predict actions in fixed left- and right-arm action spaces. While this provides a natural parameterization for robot control, it does not explicitly specify how behaviors should transform when functional roles are exchanged across arms. Across different scene initializations, the two arms may follow a similar coordination pattern, but the role-specific behavior assigned to each arm should change with the scene. Therefore, we propose BiRoAD, a Bimanual Role-Adaptive Decomposition framework for learning shared and role-adaptive representations in bimanual policies. Given bimanual trajectory or action-token features, BiRoAD decomposes these features into swap--symmetric and swap--antisymmetric components: the former captures coordination structure invariant to arm exchange, and the latter captures role-specific distinctions that vary consistently with functional role assignment. The two components are then recomposed as residual updates to the original paired arm representations, allowing BiRoAD to serve as a modular feature transformation without changing the policy inputs, imitation-learning objective, or requiring manually defined role labels. Across multiple bimanual manipulation tasks with balanced and imbalanced role distributions, BiRoAD improves robustness across role configurations over corresponding base policies, with notable gains on underrepresented role configurations.
Chinese Translation
双手操作需要协调两只手臂的策略，同时根据场景几何、物体配置和任务上下文调整其功能角色。学习这种基于场景的角色适应仍然具有挑战性，因为演示可能包含不均匀的角色分布，从而限制了对欠代表的手臂-角色配置的泛化。此外，许多双手策略在固定的左手和右手动作空间中预测动作。虽然这为机器人控制提供了自然的参数化，但并没有明确指定当功能角色在手臂之间交换时行为应如何转变。在不同的场景初始化中，两只手臂可能遵循类似的协调模式，但分配给每只手臂的角色特定行为应随场景而变化。因此，我们提出了BiRoAD，一个用于学习双手策略中共享与角色自适应表示的双手角色自适应分解框架。给定双手轨迹或动作标记特征，BiRoAD将这些特征分解为交换对称和交换反对称组件：前者捕捉对手臂交换不变的协调结构，后者捕捉随功能角色分配一致变化的角色特定区别。然后将这两个组件重新组合为对原始配对手臂表示的残差更新，使BiRoAD能够作为模块化特征转换，而无需更改策略输入、模仿学习目标或手动定义角色标签。在多个具有平衡和不平衡角色分布的双手操作任务中，BiRoAD在角色配置上提高了相应基础策略的鲁棒性，并在欠代表的角色配置上取得了显著提升。
cs.RO / 83 / 2609.23456
Smartphone GNSS Booster: Centimeter-Level Pedestrian Positioning Using a Portable Signal Re-Radiator
智能手机GNSS增强器：使用便携式信号再辐射器实现厘米级行人定位
Abstract
High-precision positioning using the global navigation satellite system (GNSS) embedded in smartphones is demanded for sidewalk-level pedestrian navigation and pinpoint location-based applications. However, current smartphone positioning accuracy for pedestrians remains at the meter level. This is mainly because limitations of the compact linearly polarized (LP) antenna in smartphones increase GNSS observation noise and hinder stable carrier phase tracking. In this paper, we propose an external GNSS signal re-radiation system that boosts smartphone GNSS observation quality while still using the smartphone's built-in GNSS receiver and antenna. The system directly connects a compact active helical antenna and a thin passive patch antenna for re-radiation, and is designed to be attached directly to the smartphone. The proposed GNSS booster achieves (1) reduced thermal noise and stable signal tracking via increased received signal strength, (2) improved multipath robustness compared with a LP antenna, and (3) stabilization of antenna phase center variation. Static experiments confirm a substantial improvement in the observation quality of GNSS carrier phase measurements compared with standard smartphone positioning. Furthermore, in pedestrian experiments, the proposed booster enabled stable carrier phase integer ambiguity resolutions, achieving centimeter-level positioning using a smartphone.
Chinese Translation
在智能手机中嵌入的全球导航卫星系统（GNSS）对人行道级别的行人导航和精准位置应用的高精度定位需求日益增加。然而，目前智能手机对行人的定位精度仍停留在米级。这主要是由于智能手机中紧凑的线性极化（LP）天线的限制，增加了GNSS观测噪声，并妨碍了稳定的载波相位跟踪。本文提出了一种外部GNSS信号再辐射系统，该系统在使用智能手机内置GNSS接收器和天线的同时，提高了智能手机的GNSS观测质量。该系统直接连接一个紧凑的有源螺旋天线和一个薄型无源贴片天线进行再辐射，并设计为直接附加在智能手机上。所提出的GNSS增强器实现了（1）通过提高接收信号强度降低热噪声和稳定信号跟踪，（2）与LP天线相比提高了多径鲁棒性，以及（3）稳定天线相位中心变化。静态实验确认，与标准智能手机定位相比，GNSS载波相位测量的观测质量有了显著改善。此外，在行人实验中，所提出的增强器实现了稳定的载波相位整数模糊解，使用智能手机实现了厘米级定位。
cs.RO / 84 / 2609.23483
STRIDER: Stepping-Enabled Multi-Gait Hierarchical 3D Loco-Manipulation Framework for Humanoid Robots
STRIDER：步态启用的多步态分层3D人形机器人运动操控框架
Abstract
Humanoid loco-manipulation faces two prominent limitations: controllers using continuous velocity commands cannot precisely regulate individual footholds, while specialized foothold-tracking modules are difficult to integrate with whole-body manipulation. Furthermore, standard action-based imitation distillation primarily transfers expert actions, without explicitly encouraging a shared representation of heterogeneous skills. This paper introduces STRIDER, a hierarchical multi-gait framework to bridge these gaps. The framework integrates terrain-aware 3D stepping logic, Adversarial Motion Priors (AMP)-based natural walking, and Cartesian upper-body control: its stepping expert selects feasible footholds in the stance-foot frame and generates clearance-aware swing trajectories. To fuse distinct walking and stepping experts into one executable student policy, we propose Latent Distillation Proximal Policy Optimization (LD-PPO), a distillation algorithm augmented with teacher-conditioned latent alignment. By jointly optimizing on-policy reinforcement learning, DAgger-based action reconstruction, and latent alignment, LD-PPO transfers expert actions while encouraging a shared skill representation across heterogeneous modes. Simulation and real-robot evaluations on the TianGong Omni humanoid show that LD-PPO outperforms vanilla distillation-PPO in foothold-tracking and posture-tracking accuracy. Deployed on hardware, STRIDER realizes multi-gait loco-manipulation with accurate foothold and end-effector tracking.
Chinese Translation
人形机器人的运动操控面临两个显著的限制：使用连续速度指令的控制器无法精确调节单个足点，而专门的足点追踪模块难以与全身操控集成。此外，基于标准动作的模仿蒸馏主要传递专家动作，而未明确鼓励异构技能的共享表征。本文提出了STRIDER，一个分层多步态框架，以弥补这些不足。该框架集成了地形感知的3D步态逻辑、基于对抗运动先验（Adversarial Motion Priors, AMP）的自然步态以及笛卡尔上半身控制：其步态专家在支撑脚框架中选择可行的足点，并生成考虑间隙的摆动轨迹。为了将不同的步态和步伐专家融合为一个可执行的学生策略，我们提出了潜在蒸馏近端策略优化（Latent Distillation Proximal Policy Optimization, LD-PPO），这是一种增强了教师条件潜在对齐的蒸馏算法。通过联合优化在线强化学习、基于DAgger的动作重构和潜在对齐，LD-PPO在传递专家动作的同时，鼓励在异构模式下的共享技能表征。在天宫Omni人形机器人上的仿真和真实机器人评估表明，LD-PPO在足点追踪和姿态追踪精度上优于传统的蒸馏-PPO。在硬件上部署后，STRIDER实现了具有准确足点和末端执行器追踪的多步态运动操控。
cs.RO / 85 / 2609.23486
Cognitive Action Reasoning for Proactive Robots from Human-Centered Multimodal Observations
基于以人为中心的多模态观察的主动机器人认知行动推理
Abstract
Robots operating in human-centered environments are typically designed to execute explicit instructions, and most robot-learning datasets likewise pair observations with task instructions or low-level actions. Although recent work has begun to explore proactive embodied assistance, existing resources target different settings and action levels, leaving real-world human-centered multimodal decision-making underexplored. We formulate this problem as \textit{Proactive Robot Action Reasoning} (\textit{ProRobo}), an upstream cognitive decision problem in which a robot must determine which action to take based on multimodal human and environmental cues without explicit action instructions. To support ProRobo, we introduce \textit{ProAction}, a real-world multimodal dataset containing 10K samples of visual observations, audio signals, and text inputs across 12 daily-life scenarios in five common scenes. To construct cognitively grounded high-level action supervision, we develop a two-stage human-in-the-loop pipeline that combines appraisal-guided candidate generation with Affective Theory-of-Mind-guided human refinement, explicitly incorporating contextual judgment about human states, urgency, feasibility, and potential risk into action annotation. Based on this supervision, we benchmark representative Multimodal Large Language Models (MLLMs) and introduce \textit{MMC2Act}, a reference model that implicitly learns the mapping from multimodal observations to cognitively grounded high-level actions. Experiments across modality settings, subject-disjoint generalization, cross-dataset transfer, and human evaluation show that general-purpose MLLMs struggle with proactively reasoning high-level actions from multimodal cues, whereas training on \textit{ProAction} substantially improves performance.
Chinese Translation
在以人为中心的环境中操作的机器人通常被设计为执行明确的指令，大多数机器人学习数据集也将观察与任务指令或低级动作配对。尽管最近的研究开始探索主动的具身辅助，但现有资源针对不同的设置和行动层级，导致现实世界中的以人为中心的多模态决策尚未得到充分研究。我们将此问题表述为主动机器人行动推理（Proactive Robot Action Reasoning，ProRobo），这是一个上游的认知决策问题，机器人必须根据多模态的人类和环境线索确定采取何种行动，而无需明确的行动指令。为了支持ProRobo，我们引入了ProAction，这是一个真实世界的多模态数据集，包含来自五个常见场景中12个日常生活场景的10K样本，包括视觉观察、音频信号和文本输入。为了构建认知基础的高层次行动监督，我们开发了一个两阶段的人机协作流程，该流程结合了基于评估的候选生成与基于情感理论的人的细化，明确将关于人类状态、紧迫性、可行性和潜在风险的上下文判断纳入行动注释。基于这一监督，我们对代表性的多模态大语言模型（Multimodal Large Language Models，MLLMs）进行了基准测试，并引入了MMC2Act，一个参考模型，隐式学习从多模态观察到认知基础高层次行动的映射。在不同模态设置、主题不重合的泛化、跨数据集迁移和人类评估的实验中，通用的MLLMs在从多模态线索中主动推理高层次行动方面表现不佳，而在ProAction上训练则显著提高了性能。
cs.RO / 86 / 2609.23487
Design and Control of a Cable-Driven Switchable Actuator with Torque/Tension Dual Modes for Exoskeletons
一种用于外骨骼的双模式扭矩/张力可切换电缆驱动执行器的设计与控制
Abstract
Existing wearable exoskeleton architectures are typically constrained by a single mechanical output modality, providing either joint torque around an anatomical joint or linear traction along a limb-training-oriented direction, which limits adaptability to diverse training scenarios. This letter presents a cable-driven switchable actuator (CDSA) that can rapidly switch between torque and tension modes while centralizing all sensing and actuation components at the proximal drive unit. A Coupled Movable Pulley Mechanism (CMPM) provides tension amplification at the distal end-effector, while a bidirectional Cable-Driven Ratchet Mechanism (CDRM) enables mode switching and preload regulation. To eliminate the need for distal instrumentation, multi-source proximal sensors are integrated with a data-driven fusion model to estimate distal output forces. An adaptive dual-mode force control strategy based on iterative learning control (ILC) is further developed. Platform experiments demonstrate transmission efficiencies of $(92.4 \pm 2.0)\%$ and $(96.5 \pm 3.3)\%$ in the torque and tension modes, respectively, along with a tension amplification ratio of $2.77 \pm 0.10$ under tension mode. Tracking tests on simulated knee-joint gait trajectories and short-stroke tension profiles yield stable control, with RMSEs of $(4.52 \pm 0.51)\%$ and $(3.15 \pm 0.19)\%$ of the uncontrolled peak value, respectively. Finally, seated human-coupled experiments validate the system's controllable force generation in both joint-torque and linear-traction application modes.
Chinese Translation
现有的可穿戴外骨骼架构通常受到单一机械输出模式的限制，仅提供围绕解剖关节的关节扭矩或沿肢体训练方向的线性牵引，这限制了对多样化训练场景的适应性。本文提出了一种电缆驱动可切换执行器（Cable-Driven Switchable Actuator, CDSA），该执行器能够在扭矩和张力模式之间快速切换，同时将所有传感和驱动组件集中在近端驱动单元。耦合移动滑轮机制（Coupled Movable Pulley Mechanism, CMPM）在远端执行器提供张力放大，而双向电缆驱动棘轮机制（Bidirectional Cable-Driven Ratchet Mechanism, CDRM）则实现模式切换和预载调节。为了消除对远端仪器的需求，集成了多源近端传感器，并采用数据驱动融合模型来估计远端输出力。此外，基于迭代学习控制（Iterative Learning Control, ILC）进一步开发了一种自适应双模式力控制策略。平台实验表明，在扭矩和张力模式下的传输效率分别为 $(92.4 imes 2.0)\%$ 和 $(96.5 imes 3.3)\%$，在张力模式下的张力放大比为 $2.77 imes 0.10$。对模拟膝关节步态轨迹和短行程张力特征的跟踪测试显示出稳定的控制，均方根误差（RMSE）分别为 $(4.52 imes 0.51)\%$ 和 $(3.15 imes 0.19)\%$，相对于未控制的峰值。最后，坐姿人机耦合实验验证了该系统在关节扭矩和线性牵引应用模式下的可控力生成。
cs.RO / 87 / 2609.23488
FeasibleFlow: One-Step Joint Transport of Configuration Feasibility and Trajectories for End-to-End Driving
FeasibleFlow：端到端驾驶中配置可行性与轨迹的一步联合传输
Abstract
End-to-end autonomous driving maps current observations directly to future trajectories, yet those trajectories must remain valid as the scene evolves. Future state modeling aims to address this temporal mismatch, but general representations often contain information unrelated to ego planning and affect trajectory generation only through auxiliary supervision, static conditioning, or proposal evaluation. We propose FeasibleFlow, a one-step end-to-end generative framework that jointly transports a configuration-space feasibility field and multimodal ego trajectories. Our Asymmetric Joint MeanFlow uses the pathwise Jacobian-vector product in the MeanFlow identity to incorporate field evolution into trajectory transport. Because safety feedback is sparser than progress feedback, we further introduce the Anchor-relative ranker (ARR) and Pareto-ReinFlow to balance safety and progress in candidate selection and generation, respectively. Experiments on the NAVSIM benchmark demonstrate the strong performance of FeasibleFlow and validate both the joint transport of feasibility and trajectories and the proposed safety-first mechanisms.
Chinese Translation
端到端自动驾驶将当前观察直接映射到未来轨迹，但这些轨迹必须在场景演变时保持有效。未来状态建模旨在解决这种时间不匹配，但一般表示通常包含与自我规划无关的信息，并且仅通过辅助监督、静态条件或提案评估影响轨迹生成。我们提出了FeasibleFlow，这是一种一步端到端生成框架，联合传输配置空间可行性场和多模态自我轨迹。我们的非对称联合均值流（Asymmetric Joint MeanFlow）利用均值流恒等式中的路径雅可比-向量积，将场演变纳入轨迹传输中。由于安全反馈比进展反馈稀疏，我们进一步引入了锚点相对排序器（Anchor-relative ranker, ARR）和帕累托再流（Pareto-ReinFlow），以分别平衡候选选择和生成中的安全性与进展。NAVSIM基准上的实验表明FeasibleFlow的强大性能，并验证了可行性与轨迹的联合传输以及所提出的安全优先机制。
cs.RO / 88 / 2609.23491
Elevator-VIGS: Separating Elevator Motion from Robot Motion in Visual-Inertial Gaussian Splatting SLAM
Elevator-VIGS：在视觉惯性高斯点云SLAM中分离电梯运动与机器人运动
Abstract
We present Elevator-VIGS, a visual-inertial 3D Gaussian Splatting SLAM system that keeps tracking and mapping through elevator rides. Inside a moving elevator, the two sensors are in conflict. The camera sees only the robot's motion relative to the elevator, while the IMU senses that motion plus the elevator's motion relative to the world. This conflict is challenging for existing visual-inertial estimators. If vision dominates, the estimator tracks only the robot's motion within the elevator and misses the elevator's rise, and if the conflict remains, the estimator diverges. We observe that the conflict comes from forcing both observations into a single coordinate frame. We instead estimate the robot's pose in the elevator's coordinate frame, and the elevator's motion relative to the world as a per-keyframe transport state, the elevator's rise and vertical velocity, within dense visual-inertial bundle adjustment. Elevator-VIGS detects rides zero-shot with a vision-language model and a depth network, and constrains the transport state at the departure and the arrival. We record real-world and simulated elevator sequences. On these sequences, Elevator-VIGS achieves state-of-the-art tracking and rendering performance. On four elevator-free public benchmarks it keeps the state-of-the-art performance of VIGS-SLAM. Project page: https://ruizhou-cn.github.io/elevator-vigs/.
Chinese Translation
我们提出了Elevator-VIGS，一种视觉惯性3D高斯点云SLAM系统，能够在电梯乘坐过程中持续进行跟踪和地图构建。在移动的电梯内，两个传感器之间存在冲突。相机仅能看到相对于电梯的机器人运动，而IMU则感知到该运动以及电梯相对于世界的运动。这种冲突对现有的视觉惯性估计器构成了挑战。如果视觉信息占主导地位，估计器只会跟踪机器人在电梯内的运动，而忽略电梯的上升；如果冲突持续存在，估计器则会发散。我们观察到，这种冲突源于将两种观测强行放入单一坐标框架中。我们改为在电梯的坐标框架中估计机器人的姿态，并将电梯相对于世界的运动作为每个关键帧的传输状态进行估计，包括电梯的上升和垂直速度，结合密集的视觉惯性束调整。Elevator-VIGS利用视觉-语言模型和深度网络零样本检测电梯乘坐，并在出发和到达时约束传输状态。我们记录了真实世界和模拟的电梯序列。在这些序列上，Elevator-VIGS实现了最先进的跟踪和渲染性能。在四个无电梯的公共基准测试中，它保持了VIGS-SLAM的最先进性能。项目页面：https://ruizhou-cn.github.io/elevator-vigs/
cs.RO / 89 / 2609.23504
Imagine then Verify: Affordance-Targeted Active Perception for Task-Oriented Grasping in Cluttered Scenes
想象然后验证：面向功能的主动感知在杂乱场景中的任务导向抓取
Abstract
Task-oriented grasping (TOG) requires robots to grasp functional parts of objects (e.g., the handle of a mug for pouring), yet these affordance regions are frequently occluded in cluttered scenes. Active perception via next-best-view (NBV) planning can resolve such occlusions by moving the camera for more informative observations. However, existing NBV methods typically optimize viewpoints for grasping the target object as a whole without distinguishing which part is task-relevant. A naive adaptation, fully scanning the target object before predicting the affordance, wastes most of the viewpoint budget on task-irrelevant surfaces (e.g., the mug body for pouring). To address this, we propose ATAP, an Affordance-Targeted Active Perception framework that shifts viewpoint planning from exhaustive target scanning to targeted affordance verification. ATAP hypothesizes the occluded target geometry via a generative shape prior and predicts the affordance distribution over the imagined complete surface. In cluttered scenes, severe occlusion can make the location of the hidden affordance ambiguous, leaving multiple locations plausible given the partial observation. ATAP therefore introduces an uncertainty-aware viewpoint planner that jointly optimizes expected entropy reduction over these competing hypotheses and expected affordance verification gain from real observations. This process iterates until the affordance is sufficiently verified for grasp execution. Experiments in simulation and real-world cluttered scenes show that ATAP substantially improves the functional grasp success rate over fixed-view TOG baselines, and outperforms reconstruction-based active perception with over 57% fewer NBV steps.
Chinese Translation
任务导向抓取（TOG）要求机器人抓取物体的功能部件（例如，用于倒水的杯子把手），然而这些功能区域在杂乱场景中常常被遮挡。通过下一最佳视角（NBV）规划的主动感知可以通过移动摄像头以获取更具信息量的观察来解决这些遮挡问题。然而，现有的NBV方法通常优化视角以整体抓取目标物体，而未区分哪些部分与任务相关。简单的适应方法是在预测功能之前完全扫描目标物体，这会在与任务无关的表面（例如，用于倒水的杯身）上浪费大部分视角预算。为了解决这个问题，我们提出了ATAP，一个面向功能的主动感知框架，将视角规划从全面扫描目标转变为针对性功能验证。ATAP通过生成形状先验假设被遮挡的目标几何形状，并预测想象的完整表面上的功能分布。在杂乱场景中，严重的遮挡可能使隐藏功能的位置变得模糊，导致多个位置在部分观察下都是合理的。因此，ATAP引入了一种考虑不确定性的视角规划器，该规划器联合优化这些竞争假设的预期熵减少和来自真实观察的预期功能验证增益。该过程迭代进行，直到功能得到充分验证以执行抓取。在模拟和真实杂乱场景中的实验表明，ATAP显著提高了功能抓取成功率，相较于固定视角的TOG基线，NBV步骤减少超过57%。
cs.RO / 90 / 2609.23554
PINGU: Extending Air-Bearing Spacecraft Emulators with Open-Source Actuators and Learned Control for Contact-Rich Proximity Operations
PINGU：通过开源执行器和学习控制扩展气垫航天器仿真器以实现接触丰富的近距离操作
Abstract
Low-cost planar air-bearing testbeds have matured into a standard proxy for free-flying spacecraft GNC, but they remain largely thruster-only and are rarely equipped for contact-rich, inertia-coupled manipulation. Building on the open-source ATMOS testbed, we contribute a reaction wheel and two force/torque-sensed robotic arms (LEVION) with interchangeable end-effectors, integrated as first-class control actuators through a unified ROS 2 abstraction layer. On top of the software stack we build a reinforcement-learning training environment and digital twin, and a controller that exploits these added degrees of freedom, letting classical optimal controllers and learned policies be swapped on the same hardware without modification. We validate the integrated system, PINGU, across four benchmark tasks: point-to-pose navigation (classical LQR vs. sim-to-real PPO), dynamic disturbance rejection under arm-induced center-of-mass shifts, reaction-wheel momentum stabilization, and force-controlled docking. The results show that these additions extend an ATMOS-class emulator into the contact-rich regime and bridge classical optimal control and reinforcement learning on one reproducible platform.
Chinese Translation
低成本的平面气垫测试平台已发展成为自由飞行航天器导航与控制（GNC）的标准代理，但它们仍然主要依赖推进器，且很少具备用于接触丰富、惯性耦合操作的能力。在开源ATMOS测试平台的基础上，我们贡献了一个反应轮和两个带有力/扭矩传感的机器人臂（LEVION），并通过统一的ROS 2抽象层将其集成为一流的控制执行器。在软件堆栈之上，我们构建了一个强化学习训练环境和数字双胞胎，以及一个利用这些附加自由度的控制器，使经典的最优控制器和学习策略能够在相同硬件上无须修改地进行切换。我们在四个基准任务上验证了集成系统PINGU：点到姿态导航（经典LQR与模拟到现实的PPO）、在臂引起的质心偏移下的动态干扰抑制、反应轮动量稳定以及力控制对接。结果表明，这些新增功能将ATMOS级仿真器扩展到了接触丰富的领域，并在一个可重复的平台上架起了经典最优控制与强化学习之间的桥梁。
cs.RO / 91 / 2609.23578
AR-WAM: A Visual-Conditioned Agent-Ready World Action Model for Robotic Manipulation
AR-WAM：一种面向机器人的视觉条件代理准备世界动作模型
Abstract
As AI agents become increasingly capable, agent-driven robotic control is emerging as a compelling paradigm. However, prevailing vision-language-action (VLA) models and world action models (WAMs) still rely on natural-language instructions to specify manipulation tasks, an ill-suited interface for agent-driven control: referentially ambiguous, spatially imprecise, redundant with the agent's inherent language understanding, and entangling intent with execution. We present AR-WAM, a visual-conditioned, agent-ready world action model that replaces language with two complementary conditions: a visual grounding prompt (a bounding box of the target) denoting the interaction object and location, and a learnable operation token dictating the atomic skill to execute. Our compact 0.5B-parameter model, with a frozen pretrained visual encoder and no language encoder, predicts scene evolution within compact latent states while decoding actions, exposing the policy's intent through explicit, supervisable reasoning signals. A model-agnostic compatibility layer provides three primitives (detect, execute, and query) so that local VLMs or online agent APIs can drive the policy directly, with long-horizon memory and closed-loop error recovery delegated to the agent side. On RoboTwin 2.0, RMBench, and a real Astribot S1 dual-arm platform, AR-WAM matches the strongest baselines on standard manipulation (87.2% average success) and outperforms them on memory-dependent and real-robot long-horizon tasks, improving success rates by 5.9% and 36.7%, respectively, while maintaining the lowest inference latency (14.1 ms).
Chinese Translation
随着人工智能代理能力的不断增强，基于代理的机器人控制正成为一种引人注目的范式。然而，现有的视觉-语言-动作（VLA）模型和世界动作模型（WAM）仍依赖自然语言指令来指定操作任务，这对于基于代理的控制而言并不合适：它在指称上模糊、在空间上不精确、与代理固有的语言理解冗余，并将意图与执行混淆在一起。我们提出了AR-WAM，一种视觉条件的、代理准备的世界动作模型，用两个互补条件替代语言：一个视觉基础提示（目标的边界框）表示交互对象和位置，以及一个可学习的操作令牌指示要执行的原子技能。我们的紧凑型0.5B参数模型，采用冻结的预训练视觉编码器且不使用语言编码器，在紧凑的潜在状态中预测场景演变，同时解码动作，通过明确的、可监督的推理信号揭示策略的意图。一个与模型无关的兼容层提供三种原语（检测、执行和查询），使得本地的视觉语言模型（VLM）或在线代理API可以直接驱动策略，同时将长时间记忆和闭环错误恢复委托给代理端。在RoboTwin 2.0、RMBench和真实的Astripot S1双臂平台上，AR-WAM在标准操作上与最强基线匹配（87.2%的平均成功率），并在依赖记忆和真实机器人长时间任务上超越它们，成功率分别提高了5.9%和36.7%，同时保持最低的推理延迟（14.1毫秒）。
cs.RO / 92 / 2609.23580
TaskAnchor: Grounding Task State in Reactive VLAs for Long-Horizon Manipulation
TaskAnchor：在反应式视觉-语言-动作模型中为长时间操作奠定任务状态
Abstract
Reactive vision--language--action (VLA) models struggle with long-horizon manipulation when visually similar observations can correspond to different actions depending on the task stage or interaction history. We refer to this ambiguity as task-state aliasing and introduce TaskAnchor, a lightweight adapter that grounds pretrained VLAs in execution history. TaskAnchor combines history-conditioned visual refinement with a milestone-supervised task-state coordinate, a scalar representing the semantic stage of execution. These signals are injected through the native visual and language interfaces, respectively, without introducing an explicit planner or modifying the action-generation mechanism. On RMBench, TaskAnchor achieves approximately 4.9--5.5$\times$ the average success rates of the published $\pi_{0.5}$ and X-VLA baselines, with consistent gains on RoboMemArena and real robots. The added latency is only 2.08\,ms per action chunk for $\pi_{0.5}$.
Chinese Translation
反应式视觉-语言-动作（VLA）模型在长时间操作中面临挑战，因为视觉上相似的观察可能会根据任务阶段或交互历史对应于不同的动作。我们将这种模糊性称为任务状态别名，并引入了TaskAnchor，这是一种轻量级适配器，用于将预训练的VLA与执行历史相结合。TaskAnchor结合了历史条件下的视觉细化和里程碑监督的任务状态坐标，后者是一个表示执行语义阶段的标量。这些信号通过本地视觉和语言接口注入，而无需引入显式规划器或修改动作生成机制。在RMBench上，TaskAnchor的平均成功率约为已发布的$ ext{π}_{0.5}$和X-VLA基线的4.9--5.5倍，并在RoboMemArena和真实机器人上获得了一致的提升。对于$ ext{π}_{0.5}$，每个动作块的额外延迟仅为2.08毫秒。
cs.RO / 93 / 2609.23610
PRIMO: Prior-Informed Odometry from Human-Motion Tracking for Humanoid Robots
PRIMO：基于人类运动跟踪的先验信息驱动的人形机器人里程计
Abstract
Simulation-trained humanoid proprioceptive odometry faces two transfer challenges: training trajectories generated by specific robot control policies intended for deployment cover only a limited range of motions, while sim-to-real mismatch can make unconstrained predictions unreliable. We address both with Prior-Informed Odometry from Human-Motion Tracking (PRIMO). On the data side, we generate odometry supervision by having the humanoid track diverse retargeted human motions in simulation, decoupling supervision from the deployment policies and broadening the training motion distribution. On the model side, a Prior-Informed estimator uses physics- and symmetry-informed priors to structure velocity and rotation prediction and a coarse raw-context pathway to preserve sensor context alongside encoded features, thereby strengthening sim-to-real generalization. Under a unified real-robot protocol, PRIMO reduces mean error by 31.6%-61.7% relative to the strongest evaluated external baseline in each domain-metric comparison. Across two locomotion-policy revisions, policy specialists exhibit symmetric crossover, whereas Tracking-Locomotion training reduces mean opposite-policy simulation error by 86.8%-94.6%. On real dynamic motion, Tracking-Locomotion training reduces mean error by 69.2%-81.7% relative to training on the union of both deployment policies. Across the tested motion compositions, the Prior-Informed estimator consistently lowers mean trajectory errors relative to its Unconstrained counterpart in both simulation and real-robot evaluation. Code is available at https://github.com/Agibot-Spatial-Intelligence/PRIMO.
Chinese Translation
经过仿真训练的人形机器人本体里程计面临两个迁移挑战：由特定机器人控制策略生成的训练轨迹仅覆盖有限的运动范围，而仿真与现实之间的不匹配可能使得无约束预测不可靠。我们通过基于人类运动跟踪的先验信息驱动里程计（PRIMO）来解决这两个问题。在数据方面，我们通过让人形机器人在仿真中跟踪多样化的重定向人类运动来生成里程计监督，从而将监督与部署策略解耦，并拓宽训练运动分布。在模型方面，先验信息估计器利用物理和对称性信息的先验来构建速度和旋转预测，并通过粗略的原始上下文路径来保留传感器上下文与编码特征，从而增强仿真到现实的泛化能力。在统一的真实机器人协议下，PRIMO在每个领域指标比较中相对于最强评估的外部基线减少了31.6%-61.7%的平均误差。在两个运动策略修订中，策略专家表现出对称交叉，而跟踪-运动训练将平均相反策略仿真误差降低了86.8%-94.6%。在真实动态运动中，跟踪-运动训练相对于在两个部署策略的联合上训练将平均误差降低了69.2%-81.7%。在测试的运动组合中，先验信息估计器在仿真和真实机器人评估中始终相对于其无约束对应物降低了平均轨迹误差。代码可在 https://github.com/Agibot-Spatial-Intelligence/PRIMO 获取。
cs.RO / 94 / 2609.23614
CompVLA: A Variable Compliance Vision-Language-Action Model for Contact-rich Manipulation
CompVLA：一种用于接触丰富操作的可变顺应性视觉-语言-动作模型
Abstract
Contact-rich manipulation, requiring robots to regulate not only motion but also how they yield to external forces, has emerged as the next frontier for Vision-Language-Action (VLA) models. However, existing VLAs output purely kinematic commands, degrading performance on real-world contact-rich tasks. In this paper, we introduce CompVLA, a unified VLA framework that jointly predicts motion and stiffness matrix from RGB and language inputs. Our approach augments the conventional architecture with a dedicated Compliance Expert, which outputs time-varying stiffness and virtual displacement profiles executed via geometric impedance control. We demonstrate that CompVLA achieves the highest average success rate across diverse contact-rich tasks, outperforming both vanilla and compliance-aware VLA baselines, with ablations confirming each component is essential.
Chinese Translation
接触丰富的操作要求机器人不仅调节运动，还要调节如何对外部力量作出反应，已成为视觉-语言-动作（VLA）模型的下一个前沿。然而，现有的VLA模型仅输出纯运动学指令，导致在真实世界的接触丰富任务中性能下降。本文介绍了CompVLA，一个统一的VLA框架，能够从RGB和语言输入中共同预测运动和刚度矩阵。我们的方法通过一个专门的顺应性专家增强了传统架构，该专家输出时间变化的刚度和虚拟位移轮廓，并通过几何阻抗控制执行。我们证明CompVLA在多种接触丰富任务中达到了最高的平均成功率，超越了普通和顺应性感知的VLA基线，消融实验确认了每个组件的必要性。
cs.RO / 95 / 2609.23643
Spiking Neural Network Actor-Critic Proximal Policy Optimization Control for Autonomous UAV Navigation Through Constrained Openings in Civil Infrastructure and Buildings
基于脉冲神经网络的演员-评论家近端策略优化控制用于自主无人机在民用基础设施和建筑物受限开口中的导航
Abstract
Autonomous navigation of unmanned aerial vehicles in constrained three-dimensional environments has been a challenge in the robotics domain. The application of autonomous unmanned aerial vehicles in civil infrastructure inspection involves the use of such vehicles in bridge inspection, tunnel inspection, and structural inspection. The use of deep reinforcement learning in the autonomous navigation of unmanned aerial vehicles has been successful in constrained environments. However, the computational cost of the algorithm limits the application of the algorithm in the autonomous navigation of unmanned aerial vehicles. This paper proposes the use of the spiking neural network-based Proximal Policy Optimization algorithm in the autonomous navigation of unmanned aerial vehicles in constrained sequential environments. The proposed algorithm integrates the use of spike-based actor-critic reinforcement learning with the Proximal Policy Optimization algorithm. The proposed algorithm uses the stochastic Gaussian policy in the autonomous navigation of unmanned aerial vehicles. The proposed algorithm was implemented in the autonomous navigation of unmanned aerial vehicles in constrained 3D environments. The proposed algorithm was successful in completing 1913 episodes out of more than 3000. The proposed algorithm was successful in passing an average of 2.10 windows per episode. The proposed algorithm was successful in achieving a success rate of 63.77%. The proposed algorithm was successful in achieving success rates of more than 90% in the later stages of the algorithm.
Chinese Translation
在受限的三维环境中，无人机的自主导航一直是机器人领域的一项挑战。自主无人机在民用基础设施检查中的应用涉及桥梁检查、隧道检查和结构检查等任务。在受限环境中，深度强化学习在无人机自主导航中的应用已取得成功。然而，算法的计算成本限制了其在无人机自主导航中的应用。本文提出在受限序列环境中使用基于脉冲神经网络的近端策略优化（Proximal Policy Optimization, PPO）算法进行无人机的自主导航。所提算法将基于脉冲的演员-评论家强化学习与近端策略优化算法相结合。该算法在无人机的自主导航中采用随机高斯策略。所提算法在受限的三维环境中成功实现了无人机的自主导航。在3000多个实验中，该算法成功完成了1913个回合，平均每个回合成功通过2.10个窗口，成功率达63.77%。在算法的后期阶段，该算法的成功率超过90%。
cs.RO / 96 / 2609.23650
Beyond Appearance Shifts: Task-Semantic Action Calibration for VLA Models
超越外观变化：面向任务语义的动作校准框架用于视觉-语言-动作模型
Abstract
Vision-language-action (VLA) models have achieved strong performance in embodied manipulation, but still lack a clear mechanism to balance behavioral stability with task-semantic sensitivity. We identify two complementary failure modes. Under task-preserving changes, where task semantics remain unchanged but scene appearance varies (e.g., style, illumination, clutter, or paraphrasing), policies often exhibit unnecessary action drift. Conversely, under semantic-breaking changes, where key task semantics such as the target object or constraint are altered, policies frequently fail to produce sufficiently distinct behaviors and instead follow the original trajectory. To address this gap, we propose BAS-VLA, a task-semantic action calibration framework built on top of a frozen base VLA. BAS-VLA adopts a breaking-centered calibration core as the default path, and introduces a selective evidence-gated preserving auxiliary that activates only when nuisance variation is detected while task semantics remain consistent. On the OpenPI-pi0.5 / LIBERO-Object Milk-Swap benchmark, BAS-VLA maintains high success on clean (98.0%) and semantics-preserving conditions (97.5%), while reducing clean-criterion success to 0.0% under deliberate target-object swaps, demonstrating strong stale-task suppression and task-semantic separation. On validated style-preserving shifts, it improves success from 42% to 70% without degrading clean performance. These results highlight that reliable VLA behavior requires moving beyond appearance robustness toward explicit task-semantic action calibration.
Chinese Translation
视觉-语言-动作（VLA）模型在具身操作中取得了优异的表现，但仍缺乏明确的机制来平衡行为稳定性与任务语义敏感性。我们识别出两种互补的失败模式。在保持任务的变化中，任务语义保持不变但场景外观变化（例如风格、照明、杂乱或意译），策略往往表现出不必要的动作漂移。相反，在语义破坏的变化中，关键任务语义如目标物体或约束被改变，策略常常无法产生足够明显的行为，而是沿着原始轨迹执行。为了解决这一问题，我们提出了BAS-VLA，一个基于冻结基础VLA的任务语义动作校准框架。BAS-VLA采用以破坏为中心的校准核心作为默认路径，并引入选择性证据门控的保持辅助，仅在检测到干扰变化且任务语义保持一致时激活。在OpenPI-pi0.5 / LIBERO-Object Milk-Swap基准测试中，BAS-VLA在干净条件下保持高成功率（98.0%）和语义保持条件下（97.5%），同时在故意目标物体交换下将干净标准成功率降低至0.0%，展示了强大的过时任务抑制和任务语义分离。在经过验证的风格保持变化中，其成功率从42%提高到70%，而不降低干净表现。这些结果强调，可靠的VLA行为需要超越外观鲁棒性，朝向明确的任务语义动作校准。
cs.RO / 97 / 2609.23656
WOLF: World Model Guided LiDAR Exploration with Predictive Frontiers
WOLF：基于世界模型的激光雷达探索与预测前沿
Abstract
LiDAR-based unmanned aerial vehicle (UAV) exploration builds maps by continually selecting where to observe next. However, decisions based on the measured map provide limited foresight into spatial continuations behind occlusions, leaving potentially informative directions unrecognized. We present WOLF, a world-model-guided framework that predicts future observations to enhance autonomous exploration. In the training stage, a recurrent world model learns observation dynamics from exploration trajectories, with recurrent memory retaining the spatial context needed to interpret partial observations across successive views. Building on this context, the model combines observation history with candidate motions during exploration to predict local occupancy and visibility. To guide further sensing, a predictive frontier generation mechanism then aligns and fuses these predictions using confidence, branch agreement, and observation quality to identify promising regions. The resulting predictive frontiers join measured ones to guide geometric viewpoint selection and trajectory generation, while new scans update subsequent predictions. In simulations, our method reduces mean terminal time by 10.9% relative to EPIC in Garage at comparable coverage and increases mean coverage from 42.12% to 98.35% in Tunnel. Real-world experiments further demonstrate onboard deployment of the learned model for online inference during physical flight.
Chinese Translation
基于激光雷达的无人机（UAV）探索通过不断选择下一个观察位置来构建地图。然而，基于测量地图做出的决策对遮挡后空间延续的前瞻性有限，导致潜在的信息方向未被识别。我们提出了WOLF，一个基于世界模型的框架，通过预测未来观察来增强自主探索。在训练阶段，递归世界模型从探索轨迹中学习观察动态，递归记忆保持了解释连续视图中部分观察所需的空间上下文。在此上下文的基础上，该模型结合观察历史与探索过程中的候选运动，预测局部占用和可见性。为了指导进一步的感知，预测前沿生成机制通过置信度、分支一致性和观察质量对这些预测进行对齐和融合，以识别有前景的区域。生成的预测前沿与测量前沿结合，以指导几何视点选择和轨迹生成，同时新的扫描更新后续预测。在模拟中，我们的方法在Garage中相较于EPIC将平均终止时间减少了10.9%，并在Tunnel中将平均覆盖率从42.12%提高到98.35%。实际实验进一步展示了所学习模型在物理飞行中的在线推理能力。
cs.RO / 98 / 2609.23666
UniPoint: Unified Point-Level Sensor Fusion for Humanoid Locomotion Across Challenging Terrains
UniPoint：统一的点级传感器融合框架用于人形机器人在复杂地形中的运动
Abstract
Open-world deployment requires humanoid robots to cross highly heterogeneous terrain safely, with perception that simultaneously provides wide coverage, local accuracy, and redundancy against sensor failure. Existing approaches struggle to satisfy all three: one forward depth camera or nearby height sampling covers too little; odometry-corrected elevation maps drift under aggressive motion and miss thin vertical structures; image-level encoding costs grow with camera count. We present UniPoint, a humanoid whole-body locomotion framework built on multi-source point-level sensor fusion. Measurements from a 360{\deg} light detection and ranging (LiDAR) sensor and two depth cameras are early-fused into one base-frame point set. Voxelization resamples it to a fixed number of tokens encoded by linear self-attention and proprioception-queried cross-attention, decoupling forward cost from sensor count. The point set retains standing thin barriers; a single-modality failure removes only part of the tokens, so the policy degrades gracefully. A single training run with terrain-aware rewards, perception-degradation injection, and domain randomization produces one policy for all eight terrain types, deployed on an onboard RK3588 without fine-tuning. On a DR02 humanoid, 20 trials at each of nine real-world settings over seven terrain types validate the policy on 70-cm-high platforms, 100-cm gaps, thin barriers, and sparse or narrow footholds; it also generalizes zero-shot outdoors.
Chinese Translation
开放世界的部署要求人形机器人能够安全地穿越高度异质的地形，同时具备广泛的感知覆盖、局部精度和对传感器故障的冗余。现有的方法难以同时满足这三者：单个前向深度相机或附近的高度采样覆盖范围过小；经过里程计校正的高程图在剧烈运动下会漂移，并且无法捕捉到细小的垂直结构；图像级编码的成本随着相机数量的增加而增长。我们提出了UniPoint，一个基于多源点级传感器融合的人形全身运动框架。来自360°光学探测与测距（LiDAR）传感器和两个深度相机的测量数据被早期融合为一个基础框架点集。体素化将其重新采样为固定数量的标记，通过线性自注意力和本体查询的交叉注意力进行编码，从而将前向成本与传感器数量解耦。点集能够保留站立的细小障碍物；单一模态的故障仅会移除部分标记，因此策略能够优雅地降级。通过具有地形感知奖励、感知降级注入和领域随机化的单次训练运行，生成一个适用于所有八种地形类型的策略，并在不进行微调的情况下部署在板载RK3588上。在DR02人形机器人上，针对七种地形类型的九个真实场景进行的20次试验验证了该策略在70厘米高的平台、100厘米的间隙、细小障碍物以及稀疏或狭窄的立足点上的有效性；它还能够在户外实现零样本泛化。
cs.RO / 99 / 2609.23731
Marginal Calibration Does Not Compose: Hidden Dependence in Modular Robot Navigation
边际校准无法组合：模块化机器人导航中的隐性依赖
Abstract
Robotic systems are typically composed of multiple independently developed modules that work together to perceive, predict, and act in the environment. Although each module may perform reliably in isolation, composing them does not necessarily preserve uncertainty calibration at the system level. In this work, we show that well-calibrated component interfaces do not necessarily produce calibrated downstream behavior after composition. Using a moving-obstacle prediction pipeline, we demonstrate that position and velocity estimators can each appear well calibrated individually, yet differences in how their error are correlated lead to substantially different estimates of future-state uncertainty. Consequently, assuming independence can make the system either overly confident or unnecessarily conservative, directly influencing downstream planning decisions and safety. Through simulations, we show that modeling the joint covariance restores downstream calibration and improves system performance, whereas dependence-robust uncertainty bounds enhance safety at the cost of increased conservatism. Our findings reveal a fundamental limitation of independently validating robotic modules and highlight the need for interfaces that communicate dependence information or support direct system-level calibration.
Chinese Translation
机器人系统通常由多个独立开发的模块组成，这些模块共同工作以感知、预测和在环境中行动。尽管每个模块在孤立状态下可能表现可靠，但将它们组合在一起并不一定能在系统层面上保持不确定性校准。在本研究中，我们展示了良好校准的组件接口在组合后并不一定能产生校准的下游行为。通过使用移动障碍物预测管道，我们证明了位置和速度估计器在各自独立时可能看似良好校准，但它们的误差相关性差异导致对未来状态不确定性的估计存在显著差异。因此，假设独立性可能使系统过于自信或不必要地保守，直接影响下游规划决策和安全性。通过模拟，我们表明建模联合协方差可以恢复下游校准并改善系统性能，而依赖稳健的不确定性界限则在增加保守性的代价下增强安全性。我们的研究结果揭示了独立验证机器人模块的基本局限性，并强调了需要能够传达依赖信息或支持直接系统级校准的接口。
cs.RO / 100 / 2609.23745
FlockDiffusion: Assignment-Conditioned Diffusion for Multi-Drone Task Allocation and Completion
FlockDiffusion：基于任务条件的多无人机任务分配与完成的扩散方法
Abstract
Autonomous multi-drone navigation requires fleets to service distributed objectives in cluttered environments under tight computational budgets. Efficient coordination depends on task bundling, where each drone visits multiple objectives along its route. Separate solvers for cost estimation, assignment, and execution incur redundant graph search and produce long, abrupt paths. We propose FlockDiffusion, a learned framework combining a scene graph encoder, an explicit allocation head, an assignment conditioned diffusion transformer, and a closed form trajectory decoder. An autoregressive teacher provides offline supervision for parallel fleet trajectory generation. PyBullet ablations show that bundling increases task completion from 50% to 100%, while our complete teacher further reduces route cost by 8.4% relative to MAGNNET with bundling. In the optimized scalability benchmark, evaluated on 100 scenes per density with ten drones, FlockDiffusion achieves 6.2 to 7.6 times faster inference and approximately 37% shorter routes than the classical pipeline. As nominal task counts increase from 20 to 40, latency rises from 7.8 to 11.1 ms, compared with 48.0 to 75.8 ms for the baseline. In a separate evaluation across five Gazebo environments, FlockDiffusion achieves 100% planner coverage and reduces planned route cost by 15.4% relative to the baseline with bundling. These results demonstrate efficient planning under increasing task density in configurations that are demanding to reproduce with physical drone fleets.
Chinese Translation
自主多无人机导航要求机队在复杂环境中以紧凑的计算预算服务于分布式目标。高效的协调依赖于任务捆绑，其中每架无人机沿其路线访问多个目标。针对成本估算、分配和执行的独立求解器会导致冗余的图搜索，并产生长且突兀的路径。我们提出了FlockDiffusion，这是一种结合场景图编码器、显式分配头、基于任务条件的扩散变换器和闭式轨迹解码器的学习框架。自回归教师为并行机队轨迹生成提供离线监督。PyBullet消融实验表明，任务捆绑将任务完成率从50%提高到100%，而我们的完整教师相较于使用捆绑的MAGNNET进一步减少了8.4%的路径成本。在优化的可扩展性基准测试中，针对每个密度评估100个场景，使用十架无人机，FlockDiffusion的推理速度比传统流程快6.2到7.6倍，且路径长度约短37%。当名义任务数量从20增加到40时，延迟从7.8毫秒上升到11.1毫秒，而基线则为48.0到75.8毫秒。在对五个Gazebo环境的单独评估中，FlockDiffusion实现了100%的规划覆盖率，并将计划路径成本相较于使用捆绑的基线降低了15.4%。这些结果展示了在任务密度增加的情况下，在物理无人机机队中难以重现的配置下实现高效规划的能力。
cs.RO / 101 / 2609.23755
EgoWild2Dex: Learning Dexterous Robotic Manipulation from In-the-Wild Human Experience
EgoWild2Dex：从真实环境中的人类经验学习灵巧机器人操作
Abstract
Egocentric human data provide a principled source of supervision for learning dexterous robot manipulation. Unlike prior approaches that often collect such data in constrained or specially constructed environments, we collect in-the-wild egocentric demonstrations in real-world settings, including homes, factories, and pharmacies, etc., where people perform their ordinary tasks while wearing head-mounted cameras. This collection protocol captures diverse workflows and hand-object interactions across long-tailed object and skill distributions, but also yields visually challenging observations due to scene clutter and head-motion-induced viewpoint changes (a mean cumulative rotation of $15.93^{\circ}$/s). To address these issues, we introduce EgoWild2Dex, which transfers in-the-wild ego-human experience to dual-arm robots with dexterous hands by jointly aligning unstable egocentric views and human motions with robot observations and actions, respectively. This work offers three benefits. First, we introduce GeoFormer, a differentiable geometric transformer that warps noisy human observations toward robot observations. Second, we design a human-robot training scheme to bridge the embodiment gap, enabling high task success with limited robot supervision. Third, we release EgoWild, a 538.9-hour in-the-wild egocentric human dataset comprising 179,049 episodes, 125,961 unique task descriptions, and 1,282 object categories. On real robots, EgoWild2Dex achieves an average success rate of 96.7% across three long-horizon bimanual dexterous manipulation tasks and an average object-level zero-shot success rate of 33.3%. The data, models, and code will be released.
Chinese Translation
以自我为中心的人类数据为学习灵巧机器人操作提供了一个原则性的监督来源。与以往通常在受限或专门构建的环境中收集此类数据的方法不同，我们在真实世界环境中收集了真实场景中的自我中心演示，包括家庭、工厂和药店等地方，人们在佩戴头戴式摄像机的情况下执行日常任务。这种收集协议捕捉了多样的工作流程和手-物体交互，涵盖了长尾物体和技能分布，但由于场景杂乱和头部运动引起的视角变化（平均累积旋转为 $15.93^{ ext{°}}/$s），也导致了视觉上的挑战。为了解决这些问题，我们提出了EgoWild2Dex，它通过将不稳定的自我中心视图和人类动作与机器人观察和动作进行联合对齐，将真实环境中的人类经验转移到具有灵巧手的双臂机器人上。这项工作带来了三个好处。首先，我们引入了GeoFormer，一个可微分的几何变换器，它将嘈杂的人类观察结果扭曲到机器人观察结果上。其次，我们设计了一种人机训练方案，以弥合体现差距，使得在有限的机器人监督下实现高任务成功率。第三，我们发布了EgoWild，一个538.9小时的真实环境自我中心人类数据集，包含179,049个事件、125,961个独特任务描述和1,282个物体类别。在真实机器人上，EgoWild2Dex在三个长时间双手灵巧操作任务中实现了96.7%的平均成功率，以及33.3%的平均物体级零-shot成功率。数据、模型和代码将会发布。
cs.RO / 102 / 2609.23784
PackLab: A Comprehensive Framework for Developing, Training, and Evaluating MLLMs in Robotic Bin Packing
PackLab：一个全面的框架用于开发、训练和评估机器人箱子装填中的多模态大语言模型
Abstract
Robotic bin packing requires long-horizon sequential decision-making, as each object placement affects the available space for subsequent packing. Existing methods primarily rely on hand-crafted geometric heuristics that optimize predefined objectives or reinforcement learning policies learned through trial and error over predefined training configurations. Despite recent advances in multimodal large language models (MLLMs) for this task, their potential for closed-loop sequential decisions across heterogeneous packing configurations remains underexplored. To address this gap, we introduce PackLab, a comprehensive framework for developing, training, and evaluating MLLMs for closed-loop robotic bin packing. PackLab-Suite provides a physics-based simulation platform for scalable generation of diverse training packing trajectories and evaluation of their physical outcomes. PackLab-VLM is a packing-specialized MLLM that understands the evolving object and container states to jointly select objects and predict placements in a closed-loop manner. PackLab-Bench provides standardized packing scenarios at multiple difficulty levels for systematic evaluation. Extensive experiments demonstrate that, on average, PackLab-VLM outperforms conventional packing heuristics, traditional reinforcement learning methods, and general-purpose MLLMs across object sets and container configurations, highlighting the potential of MLLMs for long-horizon robotic packing. The code, model, dataset, and benchmark are available at https://github.com/Correr-Zhou/PackLab .
Chinese Translation
机器人箱子装填需要长时间序列决策，因为每个物体的放置都会影响后续装填的可用空间。现有方法主要依赖于手工设计的几何启发式算法，这些算法优化预定义的目标或通过在预定义训练配置上进行试错学习的强化学习策略。尽管最近在多模态大语言模型（MLLMs）方面取得了进展，但它们在异构装填配置下进行闭环序列决策的潜力仍然未得到充分探索。为了解决这一问题，我们提出了PackLab，一个用于开发、训练和评估闭环机器人箱子装填的MLLMs的全面框架。PackLab-Suite提供了一个基于物理的仿真平台，用于可扩展地生成多样化的训练装填轨迹并评估其物理结果。PackLab-VLM是一个专门针对装填的MLLM，它理解不断变化的物体和容器状态，以闭环方式共同选择物体并预测放置。PackLab-Bench提供了多个难度级别的标准化装填场景，以便进行系统评估。大量实验表明，平均而言，PackLab-VLM在物体集和容器配置上优于传统的装填启发式算法、传统强化学习方法和通用MLLMs，突显了MLLMs在长时间机器人装填中的潜力。代码、模型、数据集和基准可在https://github.com/Correr-Zhou/PackLab获取。
cs.RO / 103 / 2609.23792
Risk-Aware Motion Planning and Control under Unknown Dynamics with Hybrid Observations
在未知动力学下具有混合观测的风险感知运动规划与控制
Abstract
We consider robotic motion planning and control under unknown dynamics with hybrid state observations, where state measurements are available only in parts of the state space. Existing work combines system identification, predicted reachability, graph search and controller synthesis in a hierarchical framework using local affine approximated models over polytopic state space partitioning, but requires state observations for identification and feedback control. Based on this framework, we address blind regions by selecting nominal dynamics and precomputing open-loop control sequences before observation is lost. Since the true dynamics may differ from the selected nominal model, the robot may exit a blind polytope through an unintended facet. We quantify this transition risk and incorporate the possible outcomes into a stochastic transition system. The high-level planning problem is formulated as a stochastic shortest path problem, whose policy guides controller synthesis. A case study demonstrates that the method guides the robot from an initial state to a target while balancing route efficiency and the risks associated with traversing blind regions.
Chinese Translation
我们考虑在未知动力学和混合状态观测下的机器人运动规划与控制，其中状态测量仅在状态空间的部分区域可用。现有研究在一个分层框架中结合了系统识别、预测可达性、图搜索和控制器合成，使用局部仿射近似模型对多面体状态空间进行划分，但需要状态观测来进行识别和反馈控制。基于这一框架，我们通过选择名义动力学并在观测丢失之前预计算开环控制序列来解决盲区问题。由于真实动力学可能与所选名义模型不同，机器人可能通过意外的面离开盲多面体。我们量化这种过渡风险，并将可能的结果纳入随机过渡系统中。高层规划问题被形式化为随机最短路径问题，其策略指导控制器合成。案例研究表明，该方法能够在平衡路径效率和穿越盲区相关风险的同时，引导机器人从初始状态到达目标。
cs.RO / 104 / 2609.23800
ContactDP: Contact-Guided Diffusion Policy for Tight Insertion Tasks
ContactDP：用于紧凑插入任务的接触引导扩散策略
Abstract
High-precision connector insertion remains challenging for robotic systems due to tight mechanical tolerances, partial observability during contact, and multimodal uncertainty arising from occlusion and contact ambiguity. Successful insertion requires closed-loop contact guidance that continuously integrates global alignment cues with local contact feedback to produce stable corrective actions under interaction. In this work, we present ContactDP (Contact-Guided Diffusion Policy for Tight Insertion Tasks), a multimodal diffusion-policy framework for contact-rich insertion. ContactDP jointly integrates wrist RGB observations, fingertip tactile sensing, and wrist-mounted force-torque measurements to infer contact state and generate temporally consistent corrective motions during insertion. To ensure stable execution under contact, the learned policy operates together with a hybrid position-force controller that provides compliant low-level interaction. We evaluate our approach on a suite of industrial-grade connector insertion tasks with varying connector geometries, grasp conditions, and initial misalignment. Across all tasks, ContactDP significantly outperforms vision-only diffusion policies for performance, reliability and generalization.
Chinese Translation
由于紧密的机械公差、接触过程中的部分可观测性以及由于遮挡和接触模糊性引起的多模态不确定性，高精度连接器插入仍然是机器人系统面临的挑战。成功的插入需要闭环接触引导，持续整合全局对齐线索与局部接触反馈，以在交互过程中产生稳定的纠正动作。在本研究中，我们提出了ContactDP（用于紧凑插入任务的接触引导扩散策略），这是一个用于接触丰富插入的多模态扩散策略框架。ContactDP联合整合手腕RGB观察、指尖触觉传感和手腕安装的力-扭矩测量，以推断接触状态并在插入过程中生成时间一致的纠正运动。为了确保在接触下的稳定执行，学习到的策略与混合位置-力控制器协同工作，提供顺应性的低级交互。我们在一系列工业级连接器插入任务上评估了我们的方法，这些任务具有不同的连接器几何形状、抓取条件和初始错位。在所有任务中，ContactDP在性能、可靠性和泛化能力上显著优于仅依赖视觉的扩散策略。
cs.RO / 105 / 2609.23841
Structured World-State Reasoning for Agentic Robotic Search
面向自主机器人的结构化世界状态推理搜索
Abstract
Long-horizon robotic search must resolve natural language against heterogeneous, incomplete, and often ambiguous evidence: textual information, prior maps, and observations arriving over time. The core challenge is to contextualize these streams and decide where to gather evidence before selecting a target. We present WORLDS: World-state Observation and Reasoning for Language-guided Discovery and Search, a framework that grounds reasoning in a persistent graph initialized from geospatial priors and updated by perception. Parallel Reasoners maintain competing candidate interpretations and request evidence to distinguish between them. We collect and process the requested observations with a multimodal Examiner, after which a Judge selects a grounded target or requests another pass. WORLDS achieves 51.8% navigation success across all 5,311 CityNav test episodes, the highest reported success rate, exceeding the previous published best by 15.7 percentage points under an OSM-only, high-resolution orthographic protocol. On 1,000 shared episodes, it achieves 50.0% versus 27.9% for the strongest adapted baseline using the same model, prior, sensing stack, and movement budget. Observation-based verification by the Examiner contributes 5.9 points of this success, and at a reduced reasoning-effort setting WORLDS still exceeds the adapted GeoNav baseline by 18.8 points while generating fewer tokens. We also demonstrate WORLDS on a quadrotor, which flies the generated sensing waypoints and grounds three language targets, including a vehicle absent from the map, from its onboard imagery.
Chinese Translation
长时间跨度的机器人搜索必须在异构、不完整且常常模糊的证据中解析自然语言：文本信息、先前地图以及随时间到达的观察数据。核心挑战在于将这些信息流进行上下文化，并决定在哪里收集证据，然后选择目标。我们提出了WORLDS：世界状态观察与推理框架，用于语言引导的发现与搜索，该框架基于从地理空间先验初始化的持久图进行推理，并通过感知进行更新。并行推理器维护竞争的候选解释，并请求证据以区分它们。我们使用多模态检查器收集和处理请求的观察数据，随后由评判者选择一个已确定的目标或请求再次处理。在所有5,311个CityNav测试集上，WORLDS实现了51.8%的导航成功率，这是报告的最高成功率，比之前发布的最佳结果高出15.7个百分点，且是在仅使用OSM的高分辨率正投影协议下进行的。在1,000个共享的测试集中，它实现了50.0%的成功率，而同样模型、先验、感知堆栈和移动预算下，最强的适应基线仅为27.9%。检查器的基于观察的验证为这一成功贡献了5.9个百分点，在降低推理努力的设置下，WORLDS仍然比适应的GeoNav基线高出18.8个百分点，同时生成的令牌更少。我们还在四旋翼无人机上演示了WORLDS，该无人机飞行生成的感知航点，并从其机载图像中确定三个语言目标，包括地图上缺失的一个车辆。
cs.RO / 106 / 2609.23856
Object-Centered Reconstruction for Vision-Based 3D Force Estimation
基于视觉的三维力估计的物体中心重建
Abstract
Excessive force may damage tissue and increase the risk of anastomotic leakage in robotic colorectal surgery. Although the da Vinci 5 provides force sensing, this capability is unavailable on earlier da Vinci systems and many other surgical robotic platforms. In this work, we present a vision-based pipeline for estimating 3D interaction forces from soft-tissue deformation in stereo endoscopic video. We dynamically reconstruct the tissue point cloud in an object-centered coordinate frame, track tissue points with geometric constraints, and predict the 3D force vector with a neural network. We progressively evaluate the pipeline on rubber-glove phantoms, ex vivo porcine colons, and in vivo colorectal surgical video sequences. Under varying tissue orientations and positions within the endoscopic view, as well as different camera viewpoints, the proposed method achieves average root mean square error (RMSEs) of 0.77 N and 1.30 N on the phantom and porcine colon, respectively. Compared with the camera-frame representation, the object-centered representation reduces average RMSE by 51.3% and 56.7%, while geometry-constrained tracking reduces RMSE by 19.8% and 25.3% compared with CoTracker. We further qualitatively demonstrate the feasibility of vision-based force estimation on an in vivo colorectal surgical sequence, as a step toward clinical translation of vision-based, sensorless force estimation.
Chinese Translation
过大的力可能会损伤组织并增加机器人结直肠手术中吻合口漏的风险。尽管达芬奇5（da Vinci 5）系统提供了力传感功能，但这一能力在早期的达芬奇系统和许多其他外科机器人平台上并不可用。在本研究中，我们提出了一种基于视觉的管道，用于从立体内窥镜视频中的软组织变形估计三维交互力。我们在物体中心坐标系中动态重建组织点云，利用几何约束跟踪组织点，并通过神经网络预测三维力向量。我们在橡胶手套模型、离体猪结肠和活体结直肠手术视频序列上逐步评估该管道。在内窥镜视野中不同的组织方向和位置以及不同的相机视角下，所提出的方法在模型和猪结肠上的平均均方根误差（RMSE）分别达到了0.77 N和1.30 N。与相机坐标系表示相比，物体中心表示将平均RMSE分别降低了51.3%和56.7%，而几何约束跟踪相比于CoTracker将RMSE降低了19.8%和25.3%。我们进一步定性展示了在活体结直肠手术序列中基于视觉的力估计的可行性，为基于视觉的无传感器力估计的临床转化迈出了重要一步。
cs.RO / 107 / 2609.23863
Grounded Action Model: 3D Grounding as a Foundation for Robotics
基础行动模型：3D 定位作为机器人技术的基础
Abstract
Manipulation policies must know which objects matter and where they are, yet the pretrained backbones that current robot foundation models build on, from language in vision-language-action models (VLAs) to video generation in world-action models (WAMs), do not directly require this metric grounding, leaving it to be learned implicitly from robot demonstrations. We propose Grounded Action Models (GAMs), a new paradigm of robot foundation models built with 3D grounding. GAM can be conditioned using language, points, or box prompts, which are first transformed into a shared object-centric representation of the selected objects. This representation captures target-focused visual features and metric object geometry, which is mixed with robot state history through a multi-stream transformer to predict action chunks. Although GAMs can be run autonomously, they can also serve as a low-level controller that a high-level planner controls using its various input modalities, allowing for long-horizon and memory-dependent manipulation. On RoboTwin 2.0, GAM achieves an average success rate of 55.3% across 50 tasks (vs. 52.0% for Spatial Forcing), including 47.6% under scene randomization (vs. 30.4% for Abot-M0), with its action policy trained only on clean-scene demonstrations. On LIBERO-PRO, it achieves a state-of-the-art average success rate of 61% (vs. 53% for $\pi_{0.5}$) across 16 perturbation settings, with the largest gains when targets are relocated or newly designated. On two real robots, GAM retains 17/20 successes under visual shift on a bimanual YAM versus 4/20 for $\pi_{0.5}$, while its composition with a Molmo2 planner on a Franka achieves 64.7% ID and 49.8% OOD step completion on long-horizon and memory-dependent tasks.
Chinese Translation
操作策略必须了解哪些物体重要以及它们的位置，然而当前机器人基础模型所依赖的预训练骨干网络，从视觉-语言-行动模型（VLAs）中的语言到世界-行动模型（WAMs）中的视频生成，并不直接要求这种度量定位，而是将其隐式地从机器人演示中学习。我们提出了基础行动模型（GAMs），这是一种基于3D定位的新型机器人基础模型范式。GAM可以通过语言、点或框提示进行条件化，这些提示首先被转化为所选物体的共享物体中心表示。该表示捕捉了目标聚焦的视觉特征和度量物体几何形状，并通过多流变换器与机器人状态历史混合，以预测行动片段。尽管GAM可以自主运行，但它也可以作为低级控制器，由高级规划器使用其各种输入模态进行控制，从而实现长时间跨度和依赖记忆的操作。在RoboTwin 2.0上，GAM在50个任务中实现了55.3%的平均成功率（相比之下，Spatial Forcing为52.0%），在场景随机化下的成功率为47.6%（相比之下，Abot-M0为30.4%），其行动策略仅在干净场景演示上进行训练。在LIBERO-PRO上，它在16个扰动设置中实现了61%的最新平均成功率（相比之下，$ ext{π}_{0.5}$为53%），当目标被重新定位或新指定时，获得了最大的提升。在两台真实机器人上，GAM在双手YAM上在视觉偏移下保持了17/20的成功率，而$ ext{π}_{0.5}$为4/20，同时与Molmo2规划器在Franka上的组合在长时间跨度和依赖记忆的任务中实现了64.7%的ID和49.8%的OOD步骤完成率。
cs.RO / 108 / 2609.23885
HumynexSurg-1: A Curated Expert Liposuction Dataset
HumynexSurg-1：一个经过精心策划的专家抽脂数据集
Abstract
Robot foundation models learn manipulation from large demonstration corpora, but surgery is missing from those corpora: across the 780-hour Open-H surgical collection, one dataset carries synchronized force and none covers an aesthetic procedure. Liposuction is the hard case, because the instrument works under the skin and the surgeon operates by feel and by judgment. Humynex Robotics builds curated expert datasets for this kind of procedure. HumynexSurg-1 is the first release: a master liposuction surgeon performing on porcine abdominal tissue while narrating every decision, recorded with synchronized suction pressure, six-axis hand force/torque, top-down RGB-D video, side video and a lavalier microphone -- 14 episodes, 42,738 frames, 35.6 minutes, 356 utterances of which 95% compile into a liposuction-specific label schema. The capture follows a patent-pending sensing plan organized around the quantities a policy needs, so a channel captured today by a model can be upgraded to a sensor tomorrow without changing the data format. This release captures the instrument motion as a tool-hand track in the side video and provides the force channel as state; the funded capture adds a measured 6-DoF handle pose, a validated force channel, ultrasound imaging of the fat layer, and palpation sensing. As a proof of concept, NVIDIA Isaac GR00T N1.7 fine-tunes on the dataset with no custom code in under an hour per run and learns the recorded sessions; scaling probes on the same episodes show where further gains come from: every new session lowers the error on an unseen session. The dataset, its label schema, its quality-assurance reports and its evaluation protocol are the product; the next capture, many short sessions across fat regions with the sensors named here, is what the probes point to.
Chinese Translation
机器人基础模型通过大量示范语料学习操作，但手术在这些语料中缺失：在780小时的Open-H手术集合中，一个数据集携带同步的力，而没有涵盖美容手术。抽脂是一个困难的案例，因为仪器在皮肤下工作，外科医生依靠感觉和判断进行操作。Humynex Robotics为这种程序构建了经过精心策划的专家数据集。HumynexSurg-1是首次发布：一位主抽脂外科医生在猪腹部组织上进行操作，同时叙述每个决策，记录了同步的吸力、六轴手部力/扭矩、从上到下的RGB-D视频、侧面视频和一只领夹麦克风——共14个片段，42,738帧，35.6分钟，356次发言，其中95%编入了特定于抽脂的标签架构。该捕获遵循一项待专利的传感计划，围绕策略所需的量组织，因此今天由模型捕获的通道可以在不改变数据格式的情况下明天升级为传感器。此次发布在侧面视频中捕获了作为工具-手轨迹的仪器运动，并将力通道作为状态提供；资助的捕获增加了测量的6自由度手柄姿态、经过验证的力通道、脂肪层的超声成像和触诊感应。作为概念验证，NVIDIA Isaac GR00T N1.7在数据集上进行微调，每次运行不需要自定义代码，且在不到一小时内学习记录的会话；在相同片段上的扩展探测显示进一步收益的来源：每个新会话降低了未见会话的误差。该数据集、其标签架构、质量保证报告和评估协议是其成果；下一次捕获是多个短会话，覆盖这里命名的脂肪区域，正是探测所指向的内容。
cs.RO / 109 / 2609.23888
HapticWAM: Distilling Imagined Touch into a World-Action Model without Inference-Time Tactile Sensing
HapticWAM：将想象中的触觉提炼为无需推理时触觉感知的世界-动作模型
Abstract
Contact-rich manipulation requires estimating forces, slip and contact geometry that can remain ambiguous in scene images. Optical tactile sensors provide both visual observations of the contact surface and mechanical measurements, yet learning from these signals raises two challenges: representing contact beyond appearance and transferring its benefits to a policy that does not require fingertip observations at deployment. We introduce HapticWAM, a world-action model that combines heterogeneous tactile encoding, structured contact prediction and teacher-student distillation. Its teacher encodes gel images together with deformation, shear, distributed forces, resultant wrench and derived contact state into a frozen video backbone. Rather than predicting tactile pixels alone, the model jointly generates actions and a contact package describing future events and mechanics. Anticipatory Contact Coupling uses the previously imagined package to condition attention, preserving a contact-related input when direct tactile observations are unavailable. Haptic-Imagination Distillation transfers both contact futures and action predictions to a student that retains the generative contact head but removes its fingertip input branches. On a real-world setup, across three contact-rich pick-and-place tasks, HapticWAM Student achieves a 77% per-task mean success rate (41 of 50 starts, 82% pooled), reaching 95% on one of the tasks, outperforming the evaluated teacher and baseline configurations.
Chinese Translation
接触丰富的操作需要估计在场景图像中可能模糊的力、滑动和接触几何形状。光学触觉传感器提供了接触表面的视觉观察和机械测量，但从这些信号中学习面临两个挑战：超越外观表示接触，以及将其优势转移到在部署时不需要指尖观察的策略上。我们提出了HapticWAM，这是一种结合异构触觉编码、结构化接触预测和教师-学生蒸馏的世界-动作模型。其教师将凝胶图像与变形、剪切、分布力、合力和派生接触状态编码到一个冻结的视频骨干中。该模型不仅预测触觉像素，还共同生成描述未来事件和力学的动作和接触包。预期接触耦合利用先前想象的包来调节注意力，在直接触觉观察不可用时保留与接触相关的输入。Haptic-Imagination Distillation将接触未来和动作预测转移给一个保留生成接触头但去除其指尖输入分支的学生。在一个真实世界的设置中，在三个接触丰富的拾取和放置任务中，HapticWAM学生实现了77%的每任务平均成功率（50次尝试中的41次成功，82%汇总），在其中一个任务上达到了95%的成功率，超越了评估的教师和基线配置。
cs.RO / 110 / 2609.23896
BarrierFormer: Transformer-Guided Predictive Barrier Enforcement for Safe Robot Control
BarrierFormer：基于变压器的预测性障碍物执行以实现安全机器人控制
Abstract
Control barrier functions (CBFs) have become one of the most popular tools for encoding and enforcing state constraints in safety-critical robotics. Standard CBF approaches are inherently myopic in nature as they enforce safety only at the current time step. Consequently, the system can be driven toward the boundary of the safe set where no feasible safe control exists at a future timestep. Model predictive control (MPC) based approaches address this by enforcing state constraints over a receding horizon. However, such approaches generally require the model to be known for solving a constrained optimization problem at every step, which is computationally expensive for real-time deployment. We propose BarrierFormer, a barrier-supervised transformer framework that addresses these limitations by encoding rollout-level CBF constraints in learning a model-free safe policy. A causal transformer encodes observation-action history, autoregressively generates a predictive rollout through the dynamics head to replace the model, and provides a residual correction to a nominal controller through the action head to replace the online computation. A barrier critic operating on local observations evaluates CBF constraint violations along this rollout, and a safety teacher computes barrier-consistent actions satisfying these constraints as direct supervision targets for the learned control policy. During inference, the policy maps observation-action history to control actions without any online optimization or model knowledge, enabling real-time model-free predictive safety enforcement. Evaluations across linear and nonlinear, 2D and 3D dynamical systems for safe goal-directed navigation demonstrate that BarrierFormer outperforms existing reinforcement learning (RL)-based, diffusion-based, MPC-based, and transformer-based approaches in safety rate and inference latency.
Chinese Translation
控制障碍函数（CBFs）已成为编码和执行安全关键机器人状态约束的最流行工具之一。标准的CBF方法本质上是短视的，因为它们仅在当前时间步强制执行安全性。因此，系统可能会被驱动到安全集的边界，在未来的时间步中没有可行的安全控制。基于模型预测控制（MPC）的方法通过在递减的时间范围内强制执行状态约束来解决这一问题。然而，这些方法通常需要在每个步骤中已知模型，以解决受约束的优化问题，这在实时部署中计算成本高昂。我们提出了BarrierFormer，一种障碍物监督的变压器框架，通过在学习无模型的安全策略中编码展开级CBF约束来解决这些限制。因果变压器编码观察-动作历史，通过动态头自回归生成预测展开以替代模型，并通过动作头提供对名义控制器的残差校正以替代在线计算。一个在局部观察上运行的障碍物评论员评估沿着这一展开的CBF约束违反情况，而一个安全教师计算满足这些约束的障碍物一致动作，作为学习控制策略的直接监督目标。在推理过程中，该策略将观察-动作历史映射到控制动作，而无需任何在线优化或模型知识，从而实现实时无模型的预测安全执行。在针对安全目标导航的线性和非线性、二维和三维动态系统的评估中，BarrierFormer在安全率和推理延迟方面优于现有的基于强化学习（RL）、扩散、MPC和变压器的方法。
cs.RO / 111 / 2609.23910
ReVeal: A Reconstruction-Aware Real-to-Sim Framework for VLA Policy Evaluation
ReVeal：一种重建感知的真实到模拟框架用于视觉-语言-动作（VLA）策略评估
Abstract
Simulation-based evaluation provides a scalable and repeatable alternative to real-world evaluation of vision-language-action (VLA) policies. However, reconstruction errors can cause simulated policy performance to diverge from real-world performance, motivating the need to assess reconstructed environments for downstream VLA policy evaluation. We present ReVeal, a real-to-sim assessment framework combining workspace reconstruction, reconstruction-level assessment, and matched closed-loop policy evaluation. Novel-View Mesh Fidelity (NVMF) and Annotated Planar Geometry Fidelity (APGF) assess observation and planar geometric fidelity, respectively. We also develop PGSR-D, a reconstruction pipeline incorporating monocular depth supervision to improve geometry where multi-view visual cues are limited. Across 8 assessment scenes, NVMF and APGF consistently distinguish the fidelity of 2DGS, PGSR, and PGSR-D. Matched evaluations of GR00T, SmolVLA, and pi0.5 across 8 humanoid manipulation tasks show consistent ordering between reconstruction fidelity and real-sim performance agreement across pipelines. Further analysis of the evaluation workspaces shows that higher fidelity is associated with stronger real-sim agreement.
Chinese Translation
基于模拟的评估为视觉-语言-动作（VLA）策略的真实世界评估提供了一种可扩展且可重复的替代方案。然而，重建误差可能导致模拟策略性能与真实世界性能的偏离，这促使我们需要评估重建环境以进行下游的VLA策略评估。我们提出了ReVeal，这是一种结合工作空间重建、重建级评估和匹配闭环策略评估的真实到模拟评估框架。Novel-View Mesh Fidelity (NVMF) 和 Annotated Planar Geometry Fidelity (APGF) 分别评估观察和二维几何的保真度。我们还开发了PGSR-D，这是一种重建管道，结合单目深度监督以改善多视角视觉线索有限的情况下的几何。通过8个评估场景，NVMF和APGF始终能够区分2DGS、PGSR和PGSR-D的保真度。在8个人形操控任务中，GR00T、SmolVLA和pi0.5的匹配评估显示重建保真度与真实-模拟性能一致性之间的排序关系。对评估工作空间的进一步分析表明，更高的保真度与更强的真实-模拟一致性相关。
cs.RO / 112 / 2609.23928
MR-SPITE: Accelerating Multi-Robot Conflict Scans via Hierarchical Swept-Volume Approximations
MR-SPITE：通过分层扫掠体近似加速多机器人冲突扫描
Abstract
Conflict scanning over synchronized robot paths requires detailed collision checking, potentially across every robot pair at every timestep, and may be repeated many times as conflicts are repaired. We present Multi-Robot SPITE (MR-SPITE), a conservative, motion-segment-based filter for accelerating these scans. MR-SPITE partitions each path into temporal intervals and assigns conservative bounds to each segment. An interval scheduler compares bounds for temporally overlapping motions: disjoint bounds certify the shared window as conflict-free, while unresolved windows are passed to the underlying collision checker. We integrate MR-SPITE into ARC and combine it with VAMP-based collision checking. For 16 Fetch robots, ARC with MR-SPITE achieves a paired median conflict scan speedup of 7.18x and reduces median planning time by 57% relative to the baseline ARC implementation with PRM+VAMP. These results demonstrate that motion-segment bounds complement configuration-level collision acceleration while preserving the behavior of the underlying discretized scanner.
Chinese Translation
在同步机器人路径上进行冲突扫描需要详细的碰撞检测，可能需要在每个时间步对每对机器人进行检查，并且在修复冲突时可能会重复多次。我们提出了多机器人SPITE（MR-SPITE），这是一种基于运动段的保守过滤器，用于加速这些扫描。MR-SPITE将每条路径划分为时间间隔，并为每个段分配保守界限。一个时间间隔调度器比较时间上重叠运动的界限：不相交的界限证明共享窗口是无冲突的，而未解决的窗口则传递给底层碰撞检测器。我们将MR-SPITE集成到ARC中，并与基于VAMP的碰撞检测相结合。对于16个Fetch机器人，带有MR-SPITE的ARC实现了7.18倍的成对中位数冲突扫描加速，并将中位数规划时间相对于基线ARC实现（PRM+VAMP）减少了57%。这些结果表明，运动段界限补充了配置级别的碰撞加速，同时保持了底层离散扫描器的行为。
cs.RO / 113 / 2609.23943
FinsSim: A Reality-Aligned Integrated Simulation Platform for Underwater Robot Learning
FinsSim：一个与现实对齐的水下机器人学习综合仿真平台
Abstract
Underwater robot learning relies on simulators that integrate high-fidelity hydrodynamics, convenient learning interfaces, and a credible transition to real scenarios. In this work, we present FinsSim, a reality-aligned integrated simulation platform for Sim-to-Real underwater robot learning. FinsSim first constructs high-fidelity simulation with selectable backends to adapt to diverse requirements. To facilitate underwater robot research, it further offers standard control baselines, alongside with unified robot learning workflows. For reliable Sim-to-Real transfer, FinsSim adopts a multi-sensor fusion scheme to provide low-cost yet precise localization. Moreover, it implements calibrated thruster-hydrodynamics models and a constrained wrench allocation algorithm. Bridging these modules by ROS~2, FinsSim establishes a complete Sim-to-Real transfer pipeline. Through matched simulations and experiments, it is demonstrated that reliable Sim-to-Real transfer of underwater robot control policies can be achieved with the FinsSim framework. Separate ablation studies also validate that the modules of FinsSim can address the pivotal issues of underwater Sim-to-Real from different aspects. Overall, this work aims to bridge the gap between theoretical research and practical applications, ultimately driving advancements in the field of underwater robotics.
Chinese Translation
水下机器人学习依赖于集成高保真流体动力学、便捷学习接口以及可信的真实场景过渡的仿真器。在本研究中，我们提出了FinsSim，一个用于水下机器人学习的与现实对齐的综合仿真平台。FinsSim首先构建了高保真的仿真，具有可选择的后端，以适应多样化的需求。为了促进水下机器人研究，它进一步提供了标准控制基线，以及统一的机器人学习工作流程。为了实现可靠的Sim-to-Real转移，FinsSim采用了多传感器融合方案，以提供低成本但精确的定位。此外，它实现了校准的推进器-流体动力学模型和受限的力矩分配算法。通过ROS~2连接这些模块，FinsSim建立了一个完整的Sim-to-Real转移管道。通过匹配的仿真和实验，证明了在FinsSim框架下可以实现水下机器人控制策略的可靠Sim-to-Real转移。单独的消融研究也验证了FinsSim的模块可以从不同方面解决水下Sim-to-Real的关键问题。总体而言，本研究旨在弥合理论研究与实际应用之间的差距，最终推动水下机器人领域的进步。
cs.RO / 114 / 2609.23944
Topology-Informed Visual Prompting For Vision Language Action Policies
基于拓扑信息的视觉提示用于视觉语言行动策略
Abstract
Vision-language-action (VLA) policies can struggle with manipulation tasks with complex obstacle geometries due to partial observability. These complex geometries can lead to similar visual observations or robot configurations requiring qualitatively different actions, a distinction that can be quantified using topological signatures. While motion planners with full knowledge of environment geometries and object states can reason about these signatures in planning, this information is often not known at deployment. To address this issue, we present a topology-guided visual-prompting framework that uses simulation-based planning to augment a nominal demonstration dataset and provides vision-based guidance at deployment. Our method uses a Gauss-Linking-Integral topological signature representation to capture important topological properties of the environment. Using privileged geometry information from a simulation approximation of our environment, we augment a VLA fine-tuning dataset with trajectories that move the system to a demonstrated signature and, from the new configuration, resume task execution. A vision-language model (VLM) is fine-tuned on the same dataset to both predict signatures from live camera observations and predict end-effector waypoints, which are rendered as visual prompts on the observations to guide the VLA. Across three simulated bimanual tasks and a real-world box pickup task, our method outperforms a VLA fine-tuned only on nominal demonstrations and a VLM-prompting baseline that can remove topology-relevant information from observations. On hardware, it exceeds the strongest baseline by 40% in task success. Project website: https://topology-vla.github.io.
Chinese Translation
视觉-语言-行动 (VLA) 策略在处理具有复杂障碍几何形状的操作任务时可能会遇到困难，因为存在部分可观测性。这些复杂的几何形状可能导致相似的视觉观测或机器人配置，从而需要质上不同的行动，而这种区别可以通过拓扑特征进行量化。虽然具有环境几何形状和物体状态完整知识的运动规划器可以在规划中推理这些特征，但在部署时这些信息通常是未知的。为了解决这个问题，我们提出了一种基于拓扑引导的视觉提示框架，该框架利用基于仿真的规划来增强名义演示数据集，并在部署时提供基于视觉的指导。我们的方法使用高斯链接积分 (Gauss-Linking-Integral) 拓扑特征表示来捕捉环境的重要拓扑属性。通过使用来自我们环境仿真近似的特权几何信息，我们用移动系统到演示特征的轨迹增强了 VLA 微调数据集，并从新的配置中恢复任务执行。一个视觉-语言模型 (VLM) 在同一数据集上进行微调，以预测来自实时摄像头观测的特征，并预测末端执行器的路径点，这些路径点作为视觉提示呈现在观测中，以指导 VLA。在三个模拟双手任务和一个真实世界的箱子拾取任务中，我们的方法在任务成功率上超越了仅在名义演示上微调的 VLA 和一个能够从观测中去除拓扑相关信息的 VLM 提示基线。在硬件上，它在任务成功率上超过了最强基线 40%。项目网站：https://topology-vla.github.io。
cs.RO / 115 / 2609.23968
Opt2VLA: Force-Aware Vision-Language-Action for Contact-Rich Humanoid Whole-Body Manipulation
Opt2VLA：面向接触丰富的人形机器人全身操控的力感知视觉-语言-动作
Abstract
Humanoid robots are expected to perform diverse human-level tasks in daily environments, many of which require precise regulation of interaction forces. While recent vision-language-action (VLA) models have shown promise for semantic planning and visuomotor control, existing humanoid systems primarily represent actions through geometric motion goals and rely on whole-body controllers focused on motion tracking, with limited explicit reasoning or control of interaction forces. This limitation is particularly relevant in contact-rich tasks, where geometrically similar motions may require different force regimes depending on the task context and where visual observations may become unreliable after contact. In this work, we present Opt2VLA, a force-aware VLA framework that introduces explicit force commands at the VLA-to-control interface for humanoid whole-body manipulation. A single multi-task VLA policy jointly predicts both geometric motion goals and continuous contact-force references, which are tracked by task-specific reinforcement learning (RL)-based whole-body controllers. To provide scalable and physically grounded supervision, we generate dynamically feasible and contact-consistent training data via whole-body trajectory optimization (TO) with explicit force references. We evaluate Opt2VLA on three contact-rich humanoid tasks and show that explicit force conditioning enables more accurate and consistent force regulation than motion-only control, while physically grounded torque supervision from TO further improves force tracking accuracy and stability. Closed-loop evaluations further demonstrate language-conditioned force modulation with Opt2VLA in simulation and on humanoid hardware.
Chinese Translation
人形机器人被期望在日常环境中执行多样的人类水平任务，其中许多任务需要精确调节交互力。尽管近期的视觉-语言-动作（VLA）模型在语义规划和视觉运动控制方面显示出良好的前景，但现有的人形系统主要通过几何运动目标来表示动作，并依赖于专注于运动跟踪的全身控制器，缺乏对交互力的明确推理或控制。这一局限性在接触丰富的任务中尤为重要，因为几何上相似的动作可能根据任务上下文需要不同的力模式，并且在接触后视觉观察可能变得不可靠。在本研究中，我们提出了Opt2VLA，一个力感知的VLA框架，在人形全身操控的VLA到控制接口中引入了明确的力指令。一个单一的多任务VLA策略同时预测几何运动目标和连续的接触力参考，这些参考由特定任务的基于强化学习（RL）的全身控制器进行跟踪。为了提供可扩展且物理基础的监督，我们通过全身轨迹优化（TO）生成动态可行且接触一致的训练数据，并附加明确的力参考。我们在三个接触丰富的人形任务上评估了Opt2VLA，并显示出明确的力调节使得力的调节比仅依赖运动控制更为准确和一致，而来自TO的物理基础扭矩监督进一步提高了力跟踪的准确性和稳定性。闭环评估进一步展示了在模拟和人形硬件上，Opt2VLA实现的语言调节力的调制。
cs.RO / 116 / 2609.23976
Anticipatory Robot Goalkeeping via Monotone Optimal Stopping
通过单调最优停止实现预期机器人守门
Abstract
Robots engaged in fast physical interactions often need to act before the intent of another agent is fully known. Anticipatory goalkeeping illustrates this challenge. Waiting provides more reliable information about the target but reduces the physical opportunity for interception, whereas acting early preserves reachability but requires initiating motion under uncertainty. Given a fixed closed-loop save controller, we formulate the decision of when to initiate motion as a policy-conditional finite-horizon optimal stopping problem. Building on this formulation, we propose monotone optimal stopping (MOS), a structured release-timing method for dynamic robotic interception. The quadruped save policy is trained with reinforcement learning, while MOS determines when the policy should be activated from the evolving robot state and target belief. Rather than predicting a release time or relying on confidence alone, MOS learns the return advantage of acting now over waiting for one more observation. We derive a direct Bellman recursion for this act-versus-wait margin and impose monotonicity only with respect to physical urgency, reflecting the irreversible loss of interception opportunity as time elapses. This structure enables early activation for dynamically demanding saves while preserving closed-loop adaptation when later observations change the predicted target. Under a single-crossing condition, MOS admits a threshold release boundary with a bounded approximation error. Extensive simulation studies show that MOS improves the mean save rate from 67.7% to 74.4% over a parameter-matched learned gate and increases reversal saves from 52.1% to 66.5%. Real-robot experiments further demonstrate rapid interception and post-release direction correction under human shot-direction feints.
Chinese Translation
参与快速物理交互的机器人往往需要在另一个智能体的意图尚未完全明确之前就采取行动。预期守门正是这一挑战的体现。等待可以提供关于目标的更可靠信息，但会减少拦截的物理机会，而提前行动则保持了可达性，但需要在不确定性下启动运动。给定一个固定的闭环拯救控制器，我们将何时启动运动的决策形式化为一个政策条件的有限时域最优停止问题。在此基础上，我们提出了单调最优停止（Monotone Optimal Stopping, MOS），这是一种用于动态机器人拦截的结构化释放时机方法。四足拯救策略通过强化学习进行训练，而MOS则根据不断变化的机器人状态和目标信念来确定何时激活该策略。MOS并不是预测释放时间或仅依赖置信度，而是学习现在行动相较于等待更多观察的回报优势。我们推导出这一行动与等待边际的直接贝尔曼递归，并仅针对物理紧迫性施加单调性，反映了随着时间推移拦截机会的不可逆损失。这一结构使得在动态需求较高的拯救中能够提前激活，同时在后续观察改变预测目标时保持闭环适应性。在单交叉条件下，MOS承认一个阈值释放边界，并具有有界的近似误差。大量仿真研究表明，MOS将平均拯救率从67.7%提高到74.4%，并将反转拯救率从52.1%提高到66.5%。真实机器人实验进一步展示了在面对人类投篮方向假动作时的快速拦截和释放后的方向修正。
cs.RO / 117 / 2609.23997
RoboTalk: Learning Multi-Robot Communication and Coordination from Multimodal Demonstrations
RoboTalk：从多模态示范中学习多机器人通信与协调
Abstract
Multi-robot collaboration could enable more efficient and scalable solutions to complex robotic tasks, but collaboration under partial observability remains challenging. Natural-language communication offers a promising approach to coordinating robots under partial observability. However, in decentralized manipulation, jointly learning explicit inter-robot communication and skill-level action selection from multimodal demonstrations remains underexplored for small vision-language models (VLMs) intended for on-device deployment. To address this gap, we introduce RoboTalk, a synthetic data-generation pipeline and dataset of 7,950 multimodal trajectories spanning 53 mobile-manipulation kitchen tasks for training small VLMs to communicate and coordinate. The dataset includes a leader-follower planning protocol, tool calls (perception, manipulation, navigation, and communication), rationale traces, and diversified natural-language communication. Fine-tuning open-source models on our dataset can reach 77% success on novel held-out tasks, a significant improvement over the untuned open source models, which had a success rate of around ~2%.
Chinese Translation
多机器人协作能够为复杂的机器人任务提供更高效和可扩展的解决方案，但在部分可观测性下的协作仍然具有挑战性。自然语言通信为在部分可观测性下协调机器人提供了一种有前景的方法。然而，在分散式操作中，从多模态示范中共同学习明确的机器人间通信和技能级别的动作选择，对于旨在设备端部署的小型视觉-语言模型（VLMs）仍然未被充分探索。为了解决这一问题，我们引入了RoboTalk，一个合成数据生成管道和数据集，包含7,950条跨越53个移动操控厨房任务的多模态轨迹，用于训练小型VLMs进行通信与协调。该数据集包括一个领导-跟随规划协议、工具调用（感知、操作、导航和通信）、推理轨迹以及多样化的自然语言通信。在我们的数据集上微调开源模型可以在新颖的保留任务上达到77%的成功率，相较于未微调的开源模型约2%的成功率有了显著提升。
cs.RO / 118 / 2609.24033
Imagine-RL: Residual-Confidence-Guided Cross-Attention for World-Model-Augmented VLA Reinforcement Learning
Imagine-RL：基于残差置信度引导的跨注意力机制用于世界模型增强的视觉-语言-动作强化学习
Abstract
Reliable action evaluation in contact-rich manipulation requires looking beyond the current observation to future visual and contact consequences. Existing noise-space reinforcement learning efficiently steers a frozen Vision-Language-Action (VLA) policy, but its critics largely ignore these consequences. We present Imagine-RL, which augments noise-space VLA post-training with action-conditioned visual-torque imagination. For each candidate action chunk, a frozen visual-torque latent world model (VTLWM) autoregressively predicts compact future representations without pixel reconstruction. A current image-state-action query attends to observed histories and predicted futures, while previous-window prediction residuals provide token-wise confidence priors that suppress unreliable future tokens. By combining current evidence with predicted consequences, the action critic better evaluates candidate actions and supervises the actor, while the VLA and VTLWM remain frozen. Across four real-robot tasks with 50 evaluation trials per task, Imagine-RL uses only 100 RL trajectories and improves the average success rate by (23.6%) over DSRL and by (60%) over VLA baselines.
Chinese Translation
在接触丰富的操作中，可靠的动作评估需要超越当前观察，考虑未来的视觉和接触后果。现有的噪声空间强化学习有效地引导了一个冻结的视觉-语言-动作（VLA）策略，但其批评者在很大程度上忽视了这些后果。我们提出了Imagine-RL，它通过动作条件的视觉-扭矩想象增强了噪声空间VLA的后训练。对于每个候选动作块，一个冻结的视觉-扭矩潜在世界模型（VTLWM）自回归地预测紧凑的未来表示，而无需像素重建。当前的图像-状态-动作查询关注于观察到的历史和预测的未来，而之前窗口的预测残差提供了逐个标记的置信度先验，从而抑制不可靠的未来标记。通过将当前证据与预测后果结合，动作评估器能够更好地评估候选动作并监督执行者，同时VLA和VTLWM保持冻结。在四个真实机器人任务中，每个任务进行了50次评估试验，Imagine-RL仅使用了100条强化学习轨迹，平均成功率比DSRL提高了23.6%，比VLA基线提高了60%。
cs.RO / 119 / 2609.24048
What Matters in Designing World Action Models: An Empirical Study
设计世界行动模型中的关键因素：一项实证研究
Abstract
World Action Models (WAMs) have emerged as a promising paradigm for generalizable robot control. Despite the growing number of WAM systems, existing works often introduce unified systems that bundle together multiple design choices, such as architecture and training strategy, making it difficult to isolate individual contributions and systematically compare alternative designs. In this work, we present a controlled study that disentangles these design choices and analyzes not only their empirical effects, but also how and why they shape WAMs. More specifically, we focus on three fundamental questions in building WAMs: (1) what causal structure should govern the interaction between world modeling and action generation? (2) in which latent space should world modeling be performed? and (3) how do different world-action modeling objectives affect model behavior and performance? Through structurally controlled experiments on three representative benchmarks, RoboCasa-GR1, LIBERO, and LIBERO-Plus, we systematically compare six causal structures, eight latent representations, and four training objectives, covering popular design choices in existing WAMs. We further validate our key findings on real-robot data from the DROID dataset. We hope to provide a systematic understanding of how core design choices affect world-action modeling and what principles can guide the development of future WAM systems.
Chinese Translation
世界行动模型（World Action Models, WAMs）作为一种可推广的机器人控制范式，逐渐受到关注。尽管WAM系统的数量不断增加，现有研究往往引入统一的系统，将多种设计选择（如架构和训练策略）捆绑在一起，这使得难以孤立各个因素的贡献并系统地比较不同设计。在本研究中，我们呈现了一项控制研究，解构这些设计选择，并分析它们的实证效果，以及它们如何和为何影响WAMs。更具体地说，我们关注构建WAMs的三个基本问题：（1）什么因果结构应主导世界建模与行动生成之间的互动？（2）应在何种潜在空间中进行世界建模？（3）不同的世界-行动建模目标如何影响模型的行为和性能？通过对三个代表性基准（RoboCasa-GR1、LIBERO和LIBERO-Plus）进行结构控制实验，我们系统地比较了六种因果结构、八种潜在表示和四种训练目标，涵盖了现有WAMs中的流行设计选择。我们还在DROID数据集的真实机器人数据上验证了我们的关键发现。我们希望提供对核心设计选择如何影响世界-行动建模的系统理解，以及哪些原则可以指导未来WAM系统的发展。
cs.RO / 120 / 2609.24054
AquaOrbit: Sim-to-Real Reinforcement Learning for Underwater Target Orbiting under Intermittent Visual Feedback
AquaOrbit：在间歇性视觉反馈下进行水下目标轨道运动的仿真到现实强化学习
Abstract
Intermittent visual loss disrupts target-relative feedback during underwater orbiting, making it difficult to maintain coordinated motion and reacquire a moving target. We present AquaOrbit, a reinforcement-learning controller with a recovery module for underwater target orbiting under interrupted visual feedback. During detection loss, the recovery module uses latched line-of-sight, roll, and depth references to support stabilization and target reacquisition. We train the controller in Isaac Sim with dynamics, observation, and vision-loss randomization. Evaluated without retraining in Gazebo/ROS2 under a different physics engine and perception perturbations, AquaOrbit completes 20/20 orbiting trials in each of the static- and moving-target conditions on an unseen variable-depth 3-D trajectory. In the moving-target condition, it reduces mean line-of-sight error by approximately 46% relative to a PID-based visual servoing controller with recovery while maintaining comparable path-tracking accuracy; removing the recovery module reduces completion to 9/20. Zero-shot physical deployment with fully onboard perception and control demonstrates elliptical, figure-eight, and variable-depth circular trajectories, including the latter two trajectory types absent from training. The robot maintains attitude stability during manual occlusions lasting up to 8s and reacquires the target within 2.5s in the reported attitude-induced field-of-view loss events.
Chinese Translation
间歇性的视觉丧失会干扰水下轨道运动中的目标相对反馈，使得保持协调运动和重新获取移动目标变得困难。我们提出了AquaOrbit，一种具有恢复模块的强化学习控制器，用于在间歇性视觉反馈下进行水下目标轨道运动。在检测丧失期间，恢复模块利用锁定的视线、滚转和深度参考来支持稳定和目标重新获取。我们在Isaac Sim中训练该控制器，进行了动力学、观察和视觉丧失随机化。在Gazebo/ROS2中，在不同的物理引擎和感知扰动下进行评估，无需重新训练，AquaOrbit在未见的可变深度3D轨迹上完成了静态和移动目标条件下的20/20轨道试验。在移动目标条件下，相较于具有恢复功能的基于PID的视觉伺服控制器，AquaOrbit将平均视线误差降低了约46%，同时保持了可比的路径跟踪精度；去除恢复模块后，完成率降至9/20。零-shot物理部署结合完全在机载感知和控制，展示了椭圆形、八字形和可变深度圆形轨迹，其中后两种轨迹类型在训练中并未出现。在手动遮挡持续长达8秒的情况下，机器人保持姿态稳定，并在报告的姿态引起的视野丧失事件中在2.5秒内重新获取目标。
cs.RO / 121 / 2609.24055
Toward Human-in-the-Loop Robot Failure Recovery: Bridging Communication Gaps in Human-Robot Collaboration
迈向人机协作中的机器人故障恢复：弥合人机协作中的沟通差距
Abstract
Robots can recover from failures by asking bystanders for help, but effective human-in-the-loop recovery requires communication that accounts for differences in people's knowledge. Prior inverse-semantics work generates requests using a single listener model, leaving differences in listener knowledge untested. We introduce Listener Differences in Human-Robot Interaction (LD-HRI), a game, dataset, and benchmark that evaluates speakers through human listener performance. Our evaluation examines request properties, large language model (LLM) speakers, and inverse-semantics request-selection algorithms under controlled differences in listener information. The corpus contains 446 human-written requests and 1{,}302 listener trials. We additionally evaluated 24 frozen LLM-written requests with 70 human listeners across 560 trials. Novice success is descriptively higher with model-written requests across all four tasks, yet both request sources leave substantial expert--novice gaps, including 16 percentage points for LLM requests. LD-HRI makes these gaps measurable, providing a foundation for designing more robust communication in human-robot and human-agent interaction.
Chinese Translation
机器人可以通过向旁观者寻求帮助来恢复故障，但有效的人机协作恢复需要考虑人们知识差异的沟通。先前的逆语义研究使用单一听众模型生成请求，未能测试听众知识的差异。我们引入了人机交互中的听众差异（Listener Differences in Human-Robot Interaction, LD-HRI），这是一个游戏、数据集和基准，旨在通过人类听众的表现评估发言者。我们的评估在控制听众信息差异的情况下，考察请求属性、大型语言模型（Large Language Model, LLM）发言者和逆语义请求选择算法。该语料库包含446个人工撰写的请求和1,302个听众试验。此外，我们还评估了24个冻结的LLM撰写请求，涉及70名人类听众，共560个试验。尽管在所有四个任务中，模型撰写的请求在新手成功率上表现出更高的描述性，但两种请求来源都留下了显著的专家与新手之间的差距，包括LLM请求的16个百分点。LD-HRI使这些差距可测，为设计更强健的人机及人类代理交互中的沟通奠定了基础。
cs.RO / 122 / 2609.24059
Automatic Labelling for Bimanual Mobile Manipulation
双手移动操控的自动标注
Abstract
Semantically meaningful subtask labels can provide useful contexts for long-horizon policies, but automatically identifying both reliable temporal boundaries and broad semantic descriptions for annotations remains difficult. We present an automatic labelling pipeline that assigns temporal localisation to deterministic trajectory analysis and semantic interpretation to vision-language (VL) reasoning. The pipeline segments synchronised kinematic signals into phases, performs phase-localised VL reasoning to describe the contents, and aggregates the outputs for the base, left arm, and right arm actions. We evaluate this pipeline primarily on 29 real Galaxea bimanual mobile-manipulation tasks. Repeating the VL reasoning three times first produces the same output value for 87.4% on selected tasks. A review by nine participants across all 29 tasks then judgements on the labelled phases and shows positive acceptance of temporal divisions (90.5%), body labels (90.7%), and arm labels (78.7%). The results indicate that the segmentation-VL design can produce structured annotations while preserving asynchronous bimanual behaviour, providing a basis for richer semantic subtask identification and state-based verification.
Chinese Translation
语义上有意义的子任务标签可以为长时间策略提供有用的上下文，但自动识别可靠的时间边界和广泛的语义描述以进行注释仍然困难。我们提出了一种自动标注管道，该管道将时间定位分配给确定性轨迹分析，并将语义解释分配给视觉-语言（VL）推理。该管道将同步的运动信号分段为不同的阶段，进行阶段本地化的VL推理以描述内容，并对基础、左臂和右臂的动作进行输出聚合。我们主要在29个真实的Galaxea双手移动操控任务上评估该管道。对选定任务重复进行三次VL推理，首次产生87.4%的相同输出值。随后由九名参与者对所有29个任务进行审查，评估标注阶段的判断，并显示出对时间划分（90.5%）、身体标签（90.7%）和手臂标签（78.7%）的积极接受。结果表明，分段-VL设计能够在保留异步双手行为的同时生成结构化注释，为更丰富的语义子任务识别和基于状态的验证提供了基础。
cs.RO / 123 / 2609.24062
Safety Control of a Hyper-redundant Robot via Adaptive Weighted Control Barrier Functions
通过自适应加权控制障碍函数实现超冗余机器人的安全控制
Abstract
Hyper-redundant robots are well suited for confined-space manipulation due to their high dexterity, but safe operation in cluttered environments remains challenging. In addition, their slender structures often lead to uneven load distributions and nonuniform tracking errors along the body. To address these issues, this work proposes a weighted control barrier functions (W-CBFs) framework that enforces safety constraints while reducing tracking errors caused by uneven loading. The proposed controller was first evaluated on a circular path-following task under different obstacle configurations. With fixed weights, compared to the non-weighted method, the maximum reduction in root-mean-square (RMS) tracking error was 59.6\% in simulation and 87.7\% in physical experiments. An adaptive weighting strategy was then investigated based on the discrepancy between simulated and experimental performance under different mapping functions. The RMS errors were further reduced by 21.9\% and 8.5\%, respectively, although the error increases when obstacles were located close to the robot body. Finally, the robot was evaluated in a cleaning task requiring coverage of a rectangular area and compared with manual teleoperation. Although the controller was not explicitly optimized for area coverage, the autonomous strategy achieved comparable or better coverage performance while avoiding collisions with the surrounding frame, whereas collisions occurred during manual operation.
Chinese Translation
超冗余机器人因其高灵活性非常适合在狭小空间内进行操作，但在杂乱环境中的安全操作仍然具有挑战性。此外，它们细长的结构往往导致负载分布不均和沿体身的跟踪误差不均。为了解决这些问题，本文提出了一种加权控制障碍函数（W-CBFs）框架，该框架在减少由不均匀负载引起的跟踪误差的同时，强制执行安全约束。所提出的控制器首先在不同障碍物配置下的圆形路径跟踪任务中进行了评估。在固定权重的情况下，与非加权方法相比，仿真中的均方根（RMS）跟踪误差最大减少了59.6%，而物理实验中减少了87.7%。随后，基于不同映射函数下模拟与实验性能之间的差异，研究了一种自适应加权策略。尽管在障碍物靠近机器人主体时误差会增加，但RMS误差分别进一步减少了21.9%和8.5%。最后，机器人在一个需要覆盖矩形区域的清洁任务中进行了评估，并与手动遥操作进行了比较。尽管控制器并未针对区域覆盖进行显式优化，但自主策略在避免与周围框架发生碰撞的同时，实现了可比或更好的覆盖性能，而手动操作时则发生了碰撞。
cs.RO / 124 / 2609.24068
When Does Touch Matter? Charting the Vision-Interaction Gap in Cluttered Dexterous Grasping
触觉何时重要？绘制杂乱灵巧抓取中的视觉-交互差距
Abstract
Dexterous grasping in clutter poses a basic sensing question: when do tactile measurements and external wrench estimates improve on visual geometry? Occlusion and contact can obscure grasp quality, motivating a controlled evaluation of these interaction signals. We present a controlled real-world study over five tabletop scene conditions on a dexterous system that combines vision, per-finger and wrist wrench estimates, and distributed fingertip taxels. With demonstrations, visual observations, action space, and compliant control fixed, we compare vision-only, wrench, taxel, and combined policies plus representation and fusion baselines. The combined policy succeeds in 24/25 trials versus 14/25 for vision only, and 15/15 versus 6/15 across the three confined conditions. Ablations show that wrench and taxel feedback are complementary. Behavioral comparisons show that interaction feedback enables earlier rejection of inadequate contacts, regrasping before lift, and more stable grasps. To our knowledge, this is the first real-world study to combine and separately evaluate these interaction modalities for target-oriented dexterous grasping in clutter. These results chart a widening vision-interaction gap and position cluttered dexterous grasping as a benchmark for determining when the learned policy needs interaction sensing. Project website: https://interaction-dex-grasp.github.io/
Chinese Translation
在杂乱环境中的灵巧抓取提出了一个基本的感知问题：何时触觉测量和外部扭矩估计能够改善视觉几何？遮挡和接触可能会模糊抓取质量，这促使我们对这些交互信号进行控制评估。我们在一个灵巧系统上进行了一项受控的现实世界研究，涵盖了五种桌面场景条件，该系统结合了视觉、每根手指和手腕的扭矩估计，以及分布式指尖传感器。通过演示、视觉观察、动作空间和顺应控制固定，我们比较了仅使用视觉、扭矩、传感器和组合策略，以及表示和融合基线。组合策略在25次试验中成功了24次，而仅使用视觉的成功率为14/25，在三种受限条件下的成功率为15/15对比6/15。消融实验表明，扭矩和传感器反馈是互补的。行为比较表明，交互反馈使得能够更早地拒绝不合适的接触，在提起之前重新抓取，并实现更稳定的抓取。据我们所知，这是首个结合并单独评估这些交互模式以实现目标导向的杂乱灵巧抓取的现实世界研究。这些结果描绘了日益扩大的视觉-交互差距，并将杂乱灵巧抓取定位为确定何时需要交互感知的基准。项目网站：https://interaction-dex-grasp.github.io/
cs.RO / 125 / 2609.24093
Dexterous Robot Manipulation from Human Demonstrations via Contact-Anchored Retargeting and Residual Policy Learning
通过接触锚定重定向和残差策略学习实现人类示范的灵巧机器人操控
Abstract
Learning dexterous manipulation from demonstrations is bottlenecked by data: the contact forces that determine whether a grasp succeeds are absent from every scalable source of human demonstrations. This paper builds on two observations. First, what survives the change from a human hand to a robot hand is the contact structure of a demonstration - which finger regions touch which object locations, and in what order - rather than its joint motion. Second, physical consistency need not be engineered per task: a single residual reinforcement learning (RL) policy, trained once across diverse demonstrations, can repair kinematic recordings into physically consistent, contact-annotated trajectories, and the same residual formulation restores dynamic feasibility after retargeting. These observations yield a three-stage pipeline that converts human motion-capture recordings into dexterous robot policies with no real-robot training data: physics refinement with a simulated MANO hand recovers contacts and forces, contact-anchored retargeting transfers the demonstrated contact structure through an objective independent of hand morphology, and residual policy learning adapts the result to robot actuation. The pipeline reconstructs 25,454 single-hand trajectories (success 7.3% -> 59.3%) and 25 dual-hand tasks (16.0% -> 62.4%) with one shared policy per setting, transfers one human dataset to four morphologically distinct robot hands (+62.4 pp), and executes four contact-rich bimanual tasks on physical hardware with zero real-robot training data.
Chinese Translation
从示范中学习灵巧操控受到数据的瓶颈：决定抓取成功与否的接触力在每个可扩展的人类示范来源中都缺失。本文基于两个观察结果进行研究。首先，从人手到机器人手的变化中，保留下来的不是关节运动，而是示范的接触结构——即哪个手指区域接触了哪个物体位置，以及接触的顺序。其次，物理一致性不必针对每个任务进行工程化：一个单一的残差强化学习（RL）策略，在多样化的示范中训练一次，可以将运动学记录修复为物理一致的、带有接触注释的轨迹，而相同的残差形式在重定向后恢复动态可行性。这些观察结果形成了一个三阶段的流程，将人类运动捕捉记录转换为灵巧的机器人策略，而无需真实机器人训练数据：使用模拟的MANO手进行物理精炼以恢复接触和力，接触锚定重定向通过一个独立于手形态的目标转移示范的接触结构，残差策略学习则将结果适应于机器人驱动。该流程重建了25,454个单手轨迹（成功率从7.3%提高到59.3%）和25个双手任务（成功率从16.0%提高到62.4%），在每个设置中使用一个共享策略，将一个人类数据集转移到四个形态各异的机器人手上（+62.4个百分点），并在物理硬件上执行四个接触丰富的双手任务，且没有任何真实机器人训练数据。
cs.RO / 126 / 2609.24099
Ask Before It Tells: Benchmark-to-Robot Body-Cue Transfer for a Question-First Bedside Robot
先问再告知：基于基准的机器人身体线索转移用于优先提问的床边机器人
Abstract
Body-cue recognition can support assistive robots, but benchmark accuracy does not guarantee reliable behavior under a robot-camera viewpoint. We present Nuni, a bedside robot prototype that treats a detected distress cue as a reason to ask rather than a reason to alert. We compare two X3D-UGT RGB appearance classifiers, which reach 97.7% and 94.8% six-way accuracy on NTU RGB+D, with a pose-centric hybrid pipeline on 28 single-actor scripted clips recorded from the robot camera. The hybrid path achieved 0.71 six-way macro recall, versus 0.25 and 0.29 for the fine-tuned and from-scratch RGB variants. More importantly for interaction, it produced a question-triggering distress cue in 12/16 distress clips and would have prompted unnecessarily in 2/8 normal clips; the RGB variants yielded a question-triggering cue in only 2/16 and 3/16 distress clips. We separately tested the question-first controller through event injection. All 13 state-transition trials passed: valid responses caused stand-down, two unanswered prompts produced one alert, and three boundary conditions were handled correctly. These results are a preliminary technical evaluation, not a user study or medical validation, but they show how interaction policy can limit the consequences of uncertain perception.
Chinese Translation
身体线索识别可以支持辅助机器人，但基准准确性并不能保证在机器人摄像头视角下的可靠行为。我们提出了Nuni，一个床边机器人原型，它将检测到的困扰线索视为提问的理由，而非警报的理由。我们比较了两种X3D-UGT RGB外观分类器，它们在NTU RGB+D数据集上分别达到了97.7%和94.8%的六分类准确率，并与一个基于姿态的混合管道在28个从机器人摄像头录制的单演员脚本片段上进行了比较。混合路径实现了0.71的六分类宏召回率，而微调和从头开始的RGB变体分别为0.25和0.29。更重要的是，在交互方面，它在16个困扰片段中产生了12个触发提问的困扰线索，而在8个正常片段中不必要地提示了2次；而RGB变体仅在16个困扰片段中产生了2次和3次触发提问的线索。我们通过事件注入单独测试了优先提问控制器。所有13个状态转换试验均通过：有效响应导致停止，两个未回答的提示产生了一个警报，三个边界条件得到了正确处理。这些结果是初步的技术评估，而非用户研究或医学验证，但它们展示了交互策略如何限制不确定感知的后果。
cs.RO / 127 / 2609.24118
CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies
CARE：基于经验指导的原子纠正执行用于视觉-语言-动作策略
Abstract
Vision-Language-Action (VLA) policies achieve strong performance in robotic manipulation but remain brittle once execution deviates from nominal trajectories. We propose CARE (Corrective Atomic Robotic Execution), a framework that improves recovery by learning from failures encountered during execution. Instead of generating corrective data from manually designed or random perturbations, CARE collects failed rollouts, models stage-conditioned post-failure deviations, and uses the resulting empirical distributions to synthesize representative failure states and corrective demonstrations. At inference time, CARE combines stage-wise planning with physically grounded 3D monitoring to trigger atomic adjustments or re-operations while preserving task progress. We further introduce the Failure State Recovery Benchmark (FSR-Bench), which evaluates recovery from intermediate failure states under local deviations and structural anomalies. Experiments across multiple VLA backbones, simulation benchmarks, and real-world dual-arm tasks show consistent improvements, with average task-success gains of 14.5 points in simulation and 15.9 points in the real world. Code, models, and data are available at https://github.com/xiaojunlan/care
Chinese Translation
视觉-语言-动作（VLA）策略在机器人操作中表现出色，但一旦执行偏离正常轨迹，就会变得脆弱。我们提出了CARE（纠正原子机器人执行）框架，通过学习执行过程中遇到的失败来改善恢复能力。CARE并不是通过手动设计或随机扰动生成纠正数据，而是收集失败的执行轨迹，建模阶段条件下的后失败偏差，并利用得到的经验分布合成代表性的失败状态和纠正演示。在推理时，CARE结合阶段性规划与基于物理的3D监测，以触发原子调整或重新操作，同时保持任务进展。我们进一步引入了失败状态恢复基准（FSR-Bench），用于评估在局部偏差和结构异常下从中间失败状态的恢复。针对多个VLA骨干网络、仿真基准和真实世界双臂任务的实验显示出一致的改进，仿真中的平均任务成功率提升14.5个百分点，真实世界中提升15.9个百分点。代码、模型和数据可在https://github.com/xiaojunlan/care获取。
cs.RO / 128 / 2609.24124
ActiveArena: Benchmarking and Understanding Active Perception in Robotic Manipulation
ActiveArena：评估与理解机器人操作中的主动感知
Abstract
Active perception and manipulation are crucial for robots to interact with complex scenes. Existing benchmarks struggle to evaluate how robots effectively acquire and maintain information in memory in an active manner. To this end, we introduce ActiveArena-Sim, an active-perception simulator with controllable viewpoints and large-scale workspaces as the foundation. Built on this, we propose ActiveArena-Bench, which comprises 35 tasks across 5 fine-grained categories, covering visual exploration and interactive information acquisition. Each task is difficult to solve from passive observations alone, requiring multi-round evidence acquisition and memory-based reasoning. The benchmark provides rich memory annotations, standardized training data, and ID/OOD protocols featuring disjoint scenes, unseen distractor configurations, and novel backgrounds. Moreover, we present ActiveArena-VLA, a modular suite of 13 vision-language-action configurations for controlled studies of memory writing, memory capacity, proprioceptive state, subtask supervision, and high-level planning in active perception. Benchmark results reveal a substantial ID-OOD gap: uniform memory sampling, increased memory capacity under reliable write policies, proprioceptive inputs, and subtask supervision improve OOD generalization, while planner-guided memory management and decision-making achieve performance close to the best-performing configuration using only sparse memory. ActiveArena thus provides a unified testbed to develop and diagnose models for active perception and manipulation.
Chinese Translation
主动感知和操作对于机器人与复杂场景的交互至关重要。现有的基准测试难以评估机器人如何以主动的方式有效获取和维持信息。为此，我们引入了ActiveArena-Sim，一个具有可控视点和大规模工作空间的主动感知模拟器。在此基础上，我们提出了ActiveArena-Bench，包含35个任务，分为5个细分类别，涵盖视觉探索和互动信息获取。每个任务仅通过被动观察难以解决，需进行多轮证据获取和基于记忆的推理。该基准提供丰富的记忆注释、标准化的训练数据，以及包含不相交场景、未见干扰配置和新背景的ID/OOD协议。此外，我们还推出了ActiveArena-VLA，一个包含13种视觉-语言-动作配置的模块化套件，用于对主动感知中的记忆写入、记忆容量、本体状态、子任务监督和高层次规划进行控制研究。基准测试结果揭示了显著的ID-OOD差距：均匀的记忆采样、在可靠写入策略下增加的记忆容量、本体输入和子任务监督改善了OOD泛化，而规划者引导的记忆管理和决策制定使性能接近于仅使用稀疏记忆的最佳配置。因此，ActiveArena提供了一个统一的测试平台，用于开发和诊断主动感知和操作模型。
cs.RO / 129 / 2609.24133
Phrase-Level Robotic Guqin Performance: Bimanual Motion Planning and Audio-Tactile Interaction Monitoring
短语级机器人古琴演奏：双手运动规划与音频触觉交互监测
Abstract
Recent advances in humanoid robotics and embodied intelligence have enabled robots to perform increasingly complex manipulation tasks. However, musical instrument performance remains a formidable benchmark, demanding not only collision-free trajectory execution but also precise contact timing, asymmetric bimanual coordination, and target acoustic outcomes on physical instruments. The guqin, a seven-string fretless zither, presents unique manipulation challenges due to its millimetric string spacing, transient right-hand plucking, and sustained left-hand harmonic contacts. In this work, we present a physical heterogeneous dual-arm robotic system for phrase-level autonomous guqin performance. We formulate guqin playing as a hybrid discrete--continuous execution problem and develop a hierarchical planning framework that coordinates working finger assignment, configuration continuity, obstacle avoidance, and tight bimanual contact schedules across consecutive musical events. The system integrates vision-guided instrument localization, tactile-based harmonic contact monitoring, and auditory feedback-informed plucking parameter calibration. Real-world experiments on a 25-event phrase demonstrate that the system reliably executes coordinated open-string and seventh-hui harmonic sequences on a physical guqin, achieving 93.6% and 96.8% event correctness across repeated trials.
Chinese Translation
近年来，人形机器人和具身智能的进展使得机器人能够执行越来越复杂的操作任务。然而，乐器演奏仍然是一个严峻的基准，不仅要求无碰撞的轨迹执行，还需要精确的接触时机、不对称的双手协调以及在物理乐器上实现目标音响效果。古琴作为一种七弦无品的拨弦乐器，由于其毫米级的弦间距、瞬时的右手拨弦和持续的左手和声接触，带来了独特的操作挑战。在本研究中，我们提出了一种物理异构双臂机器人系统，用于短语级自主古琴演奏。我们将古琴演奏形式化为一个混合离散-连续执行问题，并开发了一个层次化规划框架，协调工作手指分配、配置连续性、障碍物避免以及在连续音乐事件中的紧密双手接触时间表。该系统集成了视觉引导的乐器定位、基于触觉的和声接触监测以及基于听觉反馈的拨弦参数校准。在25个事件短语的实际实验中，系统可靠地在物理古琴上执行协调的开弦和第七回音和声序列，在重复试验中实现了93.6%和96.8%的事件正确率。
cs.RO / 130 / 2609.24140
BayesianGS-SLAM: Uncertainty-Aware Neural Rendering SLAM via Probabilistic Formulation
BayesianGS-SLAM：通过概率公式实现的不确定性感知神经渲染SLAM
Abstract
Neural-rendering-based SLAM relies on rendered RGB-D residuals for camera tracking and map optimization, but the reliability of these predictions can vary substantially because of sensor noise, limited observation coverage, and incomplete map representations. Without an explicit reliability estimate, unreliable residuals may adversely affect pose optimization, while frames already well explained by the current map may trigger redundant mapping updates. In this paper, we present BayesianGS-SLAM, an uncertainty-aware 3D Gaussian Splatting SLAM framework that estimates predictive color and depth uncertainty during mapping and consistently reuses it across the SLAM pipeline. Our tractable probabilistic formulation combines a sensor-noise uncertainty component with an opacity-induced map-representation component propagated through the rendering process. The resulting predictive uncertainty is used to augment mapping, normalize tracking residuals through a robust pose objective, and evaluate incoming frames using a predictive-surprise-based keyframe criterion. Unlike prior uncertainty-aware neural-rendering SLAM methods that primarily consider color uncertainty or use uncertainty only during mapping, our framework estimates predictive uncertainty for both color and depth and integrates it into mapping, tracking, and keyframe selection. Evaluations on real-world RGB-D datasets demonstrate substantially improved depth uncertainty-error ranking compared with existing uncertainty-aware SLAM methods. Moreover, the proposed keyframe-selection strategy reduces the number of selected keyframes and mapping calls while maintaining competitive tracking and rendering performance.
Chinese Translation
基于神经渲染的SLAM依赖于渲染的RGB-D残差进行相机跟踪和地图优化，但由于传感器噪声、有限的观察覆盖范围和不完整的地图表示，这些预测的可靠性可能会有很大差异。在没有明确的可靠性估计的情况下，不可靠的残差可能会对位姿优化产生不利影响，而当前地图已经很好解释的帧可能会触发冗余的地图更新。本文提出了BayesianGS-SLAM，一种不确定性感知的3D高斯喷溅SLAM框架，该框架在映射过程中估计预测的颜色和深度不确定性，并在SLAM流程中持续重用这些不确定性。我们可处理的概率公式将传感器噪声不确定性成分与通过渲染过程传播的由不透明度引起的地图表示成分相结合。生成的预测不确定性用于增强映射，通过稳健的位姿目标规范化跟踪残差，并使用基于预测惊奇的关键帧标准评估传入帧。与以往主要考虑颜色不确定性或仅在映射过程中使用不确定性的以不确定性感知的神经渲染SLAM方法不同，我们的框架为颜色和深度都估计预测不确定性，并将其整合到映射、跟踪和关键帧选择中。在真实世界的RGB-D数据集上的评估表明，与现有的不确定性感知SLAM方法相比，深度不确定性误差排名显著改善。此外，所提出的关键帧选择策略在保持竞争的跟踪和渲染性能的同时，减少了所选关键帧和映射调用的数量。
cs.RO / 131 / 2609.24145
MimicAgent: Quadruped Skills via Prompt-to-Trajectory Generation
MimicAgent：通过提示生成轨迹实现四足动物技能
Abstract
We present MimicAgent, a prompt-to-trajectory generation framework for learning dynamic quadruped skills. Although reward shaping is extensively used when training quadruped policies, navigating the resulting reward landscape is notoriously difficult, requiring hours of "graduate student descent". Eureka attempts to automate reward design with LLMs, but we find that it struggles to generalize across diverse skills and morphologies. Our key observation is that it is far easier for a human - and by association, an LLM - to generate reference motions than to shape reward functions. Our hypothesis is motivated by the success of example-guided RL for humanoids, which exploits large-scale motion capture datasets as references for training locomotion policies. Unlike humanoids, quadrupeds lack such reference motion data. Towards this end, we propose MimicAgent, an agentic harness that, given a skill prompt, generates quadruped reference trajectories with coding agents. These coarse reference trajectories are then used to train example-guided RL policies that are deployable in simulation and in the real-world. Notably, we find that when prompting Claude Fable 5.1 within our agentic harness, 87% of prompts yield semantically aligned reference trajectories.
Chinese Translation
我们提出了MimicAgent，一个用于学习动态四足动物技能的提示到轨迹生成框架。尽管在训练四足动物策略时广泛使用奖励塑形，但导航由此产生的奖励景观却 notoriously 困难，通常需要数小时的“研究生下降”。Eureka尝试利用大型语言模型（LLMs）自动化奖励设计，但我们发现它在不同技能和形态之间的泛化能力较弱。我们的关键观察是，对于人类——以及由此类推的LLM——来说，生成参考动作远比塑形奖励函数容易。我们的假设受到人形机器人示例引导强化学习（RL）成功的启发，该方法利用大规模运动捕捉数据集作为训练运动策略的参考。与人形机器人不同，四足动物缺乏这样的参考运动数据。为此，我们提出了MimicAgent，一个智能代理装置，能够根据技能提示生成四足动物的参考轨迹。这些粗略的参考轨迹随后用于训练可在仿真和现实世界中部署的示例引导RL策略。值得注意的是，我们发现，当在我们的智能代理装置中提示Claude Fable 5.1时，87%的提示生成了语义一致的参考轨迹。
cs.RO / 132 / 2609.24155
Object-Centric Conditioning for Visuomotor Flow Matching
面向对象的条件化用于视觉运动流匹配
Abstract
Robot visuomotor policies are commonly formulated as autoregressive, diffusion-based, or more recently, flow matching models. Among them, Action-to-Action (A2A) flow matching improves inference efficiency by initializing generation from historical action priors rather than stochastic noise. However, stale historical motion patterns and entangled global visual representations can jointly reduce robustness under spatial out-of-distribution (OOD) shifts and visual distractors. In this work, we propose SlotFlow, an object-centric flow matching policy for robust visuomotor manipulation. SlotFlow decouples scene observations into semantic ("what") features and lightweight image-plane spatial ("where") cues to provide object-aware policy conditioning and current-state grounding. The semantic representation suppresses irrelevant background correlations, while the spatial cue improves adaptation to shifted object configurations. Extensive simulation and real-world experiments demonstrate improved robustness under visual distractors and severe spatial perturbations while preserving the low-step inference efficiency of A2A. Controlled initialization and perception ablations further identify object-centric grounding as a major source of the gains and show that it complements, rather than replaces, useful historical motion priors.
Chinese Translation
机器人视觉运动策略通常被表述为自回归、基于扩散的模型，或更近期的流匹配模型。其中，动作到动作（Action-to-Action, A2A）流匹配通过从历史动作先验初始化生成，而不是从随机噪声中初始化，从而提高了推理效率。然而，过时的历史运动模式和纠缠的全局视觉表征可能共同降低在空间分布外（Out-of-Distribution, OOD）变化和视觉干扰下的鲁棒性。在本研究中，我们提出了SlotFlow，一种面向对象的流匹配策略，用于鲁棒的视觉运动操作。SlotFlow将场景观察解耦为语义（“什么”）特征和轻量级图像平面空间（“哪里”）线索，以提供面向对象的策略条件和当前状态的基础。语义表征抑制了无关背景相关性，而空间线索则改善了对变化的物体配置的适应性。大量的仿真和现实世界实验表明，在视觉干扰和严重空间扰动下，鲁棒性得到了改善，同时保持了A2A的低步推理效率。受控初始化和感知消融进一步确认了面向对象的基础是增益的主要来源，并表明它是对有用历史运动先验的补充，而非替代。
cs.RO / 133 / 2609.24180
GraspTune: Tactile-Driven Execution Refinement for Robust Grasping
GraspTune：基于触觉驱动的稳健抓取执行精细化
Abstract
Visual grasp proposal generation has advanced rapidly, yet converting a selected proposal into a stable physical grasp remains a central execution-stage challenge. This paper introduces GraspTune, a tactile-driven execution-stage refinement framework that starts from a nominal proposal and applies bounded residual TCP motions during approach, contact formation, and final grasp execution. GraspTune learns control-facing contact semantics from local depth, tactile signals, state, and history using state-conditioned expert contact queries and multi-task supervision for contact change, contact risk, and post-close readiness. The representation conditions a diffusion-pretrained residual policy and is aligned with PPO for closed-loop execution. Across more than 60,000 simulated executions over 20 object categories, GraspTune establishes an execution-layer benefit across four proposal generators, raising stable grasp success by +19.22, +9.55, +12.45, and +20.70 percentage points for GraspNet, Contact-GraspNet, AnyGrasp, and VGN. A four-fold held-out category study raises unseen-object execution from 54.58% to 70.33%, showing category-disjoint generalization of contact correction. Across more than 1,000 real-robot trials on a UR5e setup with Xense fingertip sensors, GraspTune raises GraspNet execution from 71.0% to 84.3%, validating direct transfer without realworld policy fine-tuning. Together, these results turn visually plausible proposals into stable physical grasps for downstream contact-rich manipulation. A supplementary video is available at https://youtu.be/kcq7fSLNtzU.
Chinese Translation
视觉抓取提案生成已迅速发展，但将选定的提案转化为稳定的物理抓取仍然是执行阶段的核心挑战。本文介绍了GraspTune，一个基于触觉驱动的执行阶段精细化框架，该框架从一个名义提案开始，并在接近、接触形成和最终抓取执行过程中应用有界残差TCP（末端执行器）运动。GraspTune通过状态条件专家接触查询和多任务监督，学习来自局部深度、触觉信号、状态和历史的控制面接触语义，以应对接触变化、接触风险和闭合后的准备状态。该表示条件化一个扩散预训练的残差策略，并与PPO（近端策略优化）对齐以实现闭环执行。在超过60,000次模拟执行和20个物体类别中，GraspTune在四个提案生成器中建立了执行层面的优势，使得GraspNet、Contact-GraspNet、AnyGrasp和VGN的稳定抓取成功率分别提高了+19.22、+9.55、+12.45和+20.70个百分点。在一个四倍保留类别的研究中，未见物体的执行率从54.58%提高到70.33%，显示出接触修正的类别不相交泛化。在超过1,000次在UR5e设置和Xense指尖传感器上的真实机器人试验中，GraspTune将GraspNet的执行率从71.0%提高到84.3%，验证了无需现实世界策略微调的直接转移。这些结果共同将视觉上合理的提案转化为下游接触丰富操作的稳定物理抓取。补充视频可在https://youtu.be/kcq7fSLNtzU获取。
cs.RO / 134 / 2609.24187
StenoVLA-3D: 3D-Aware Reasoning VLA for Navigation Through Gastrointestinal Stenoses
StenoVLA-3D：用于通过胃肠道狭窄导航的三维感知推理视觉-语言-动作模型
Abstract
Autonomous endoscopic navigation requires the policy model to predict actions from texture-poor monocular observations, make safe control decisions, and retain evidence of lesions after they leave the field of view. Existing vision-language-action (VLA) models primarily rely on visual appearance and short-term context, limiting geometric grounding and episode-level reporting. We introduce StenoVLA-3D, a 3D-aware VLA framework for navigating through stenotic regions. We integrate point-maps into the Cosmos-Reason 2 backbone through learned geometry-gated fusion, and also propose a temporal state branch to model traversal progress. Our reasoning-and-action backbone predicts grounded reasoning with actions, while dedicated heads estimate stenosis shape and generate the final lesion report. We further introduce EndoCausal, an episode-level dataset with lesion annotations, actions, and temporally grounded reasoning. On 40 held-out recorded test episodes, StenoVLA-3D reaches 95.2\% semantic accuracy and 83.4\% action accuracy. On the physical 3-DoF endoscope, it attains 88.9\% and 77.8\% task success in esophageal and colonic phantoms (36 trials each), substantially outperforming the evaluated baselines.
Chinese Translation
自主内窥镜导航要求策略模型能够从纹理较差的单目观察中预测动作，做出安全的控制决策，并在病变离开视野后保留证据。现有的视觉-语言-动作（VLA）模型主要依赖于视觉外观和短期上下文，限制了几何基础和事件级报告。我们提出了StenoVLA-3D，这是一个用于在狭窄区域导航的三维感知VLA框架。我们通过学习的几何门控融合将点图集成到Cosmos-Reason 2主干中，并提出了一个时间状态分支来建模穿越进度。我们的推理与动作主干预测与动作相结合的有依据推理，而专用头部则估计狭窄形状并生成最终的病变报告。我们进一步引入了EndoCausal，这是一个具有病变注释、动作和时间基础推理的事件级数据集。在40个保留的记录测试事件中，StenoVLA-3D达到了95.2%的语义准确率和83.4%的动作准确率。在物理3自由度内窥镜上，它在食管和结肠模型中分别达到了88.9%和77.8%的任务成功率（各36次试验），显著优于评估的基线。
cs.RO / 135 / 2609.24189
A Topological Representation with Object-Path Graphs for Open-Vocabulary Instance Navigation
一种基于对象-路径图的拓扑表示用于开放词汇实例导航
Abstract
Vision-language navigation requires embodied agents to navigate environments using natural language instructions and visual observations. Existing approaches typically decompose navigation into sequential language-guided decisions or rely on online exploration without prior environmental knowledge. Scene graph representations offer compact semantic memory but remain decoupled from downstream navigation, which still depends on dense metric maps. To close this gap, we propose an object--path graph that unifies open-vocabulary semantic reasoning with topological navigation. The proposed representation jointly supports semantic grounding, graph-based localization, and navigation within a single lightweight topological framework. Building on this graph, we introduce a navigation strategy that combines global path planning with local inter-node execution through lightweight node localization and semantic visual servoing, enabling navigation directly over the graph without dense metric reconstruction. Experiments on HM3D and Replica demonstrate competitive performance in open-vocabulary object grounding through the proposed hierarchical graph structure, while achieving effective navigation performance. Real-world robot experiments further validate the practicality of the proposed framework.
Chinese Translation
视觉-语言导航要求具身代理使用自然语言指令和视觉观察在环境中导航。现有方法通常将导航分解为顺序的语言引导决策，或依赖于没有先前环境知识的在线探索。场景图表示提供了紧凑的语义记忆，但仍然与下游导航脱钩，后者仍然依赖于密集的度量地图。为了弥补这一差距，我们提出了一种对象-路径图，它将开放词汇语义推理与拓扑导航统一起来。所提出的表示在一个轻量级的拓扑框架内共同支持语义定位、基于图的定位和导航。基于该图，我们引入了一种导航策略，结合了全局路径规划和通过轻量级节点定位与语义视觉伺服的局部节点间执行，使得导航可以直接在图上进行，而无需密集的度量重建。在HM3D和Replica上的实验表明，通过所提出的分层图结构，在开放词汇对象定位方面表现出竞争力，同时实现了有效的导航性能。实际机器人实验进一步验证了所提出框架的实用性。
cs.RO / 136 / 2609.24195
Odometry-Aided Real-Time Mapping for Underwater Robots Using Forward-Looking Sonar
基于里程计辅助的水下机器人实时映射方法：前视声纳的应用
Abstract
Reliable perception is essential for underwater vehicles operating in complex environments, where light attenuation and scattering often degrade visibility and compromise optical sensing. Forward-looking sonar (FLS) offers an alternative by providing high-frame-rate acoustic imaging under poor optical conditions. However, real-time FLS mapping remains challenging due to unresolved target elevation, spatially non-uniform noise, and fragmented target boundaries, which hinder feature extraction and introduce geometric ambiguity during projection. To address these challenges, we propose a cascaded feature reconstruction pipeline combining fast Fourier transform (FFT)-based denoising, fast multiscale constant false alarm rate (MCFAR) detection, and gradient-adaptive boundary connection to extract geometric features from degraded sonar images with low latency. We integrate attitude-aware geometric projection with incremental occupancy accumulation to construct a depth-referenced 2.5D map for local mapping in confined underwater environments. The sonar's vertical position is referenced to an external sensor, while target elevation is assigned under an explicit geometric assumption rather than measured directly by FLS. Experiments in a 3 m X 5 m pool demonstrate centimeter-scale planar mapping accuracy, with a root-mean-square error (RMSE) below 3 cm across three sequences and an average processing time of 42.4 ms per frame.
Chinese Translation
可靠的感知对于在复杂环境中操作的水下车辆至关重要，因为光衰减和散射常常降低能见度并影响光学传感。前视声纳（FLS）提供了一种替代方案，在较差的光学条件下提供高帧率的声学成像。然而，由于目标高度未解决、空间非均匀噪声和目标边界碎片化，实时FLS映射仍然面临挑战，这些因素阻碍了特征提取并在投影过程中引入几何模糊。为了解决这些挑战，我们提出了一种级联特征重建管道，结合了基于快速傅里叶变换（FFT）的去噪、快速多尺度恒虚警率（MCFAR）检测和梯度自适应边界连接，以低延迟从退化的声纳图像中提取几何特征。我们将姿态感知的几何投影与增量占用累积相结合，构建了一个深度参考的2.5D地图，用于在受限水下环境中的局部映射。声纳的垂直位置参考于外部传感器，而目标高度则在明确的几何假设下分配，而不是直接由FLS测量。我们在一个3米 x 5米的泳池中进行的实验表明，平面映射精度达到厘米级，三个序列的均方根误差（RMSE）低于3厘米，平均处理时间为每帧42.4毫秒。
cs.RO / 137 / 2609.24218
Audio-based UAV Localization with Adaptive Temporal Correspondence via Reinforcement Learning
基于音频的无人机定位：通过强化学习实现自适应时间对应
Abstract
Audio-based localization provides a low-cost and illumination-independent sensing solution for anti-UAV early warning. However, existing methods typically rely on a predefined fixed audio segment length, which limits temporal correspondence and creates a trade-off between sufficient acoustic evidence and timely localization. To address this issue, we propose an audio-based localization framework with adaptive temporal correspondence. A probe segment is first used to extract a compact acoustic state that characterizes the reliability and consistency of the observation. Guided by the state, a reinforcement learning controller dynamically determines the required audio window size for each localization decision. The selected audio segment is then processed by a Mamba-based localization network with adaptive temporal feature modulation for 3D position estimation. Extensive experiments demonstrate that our method achieves competitive 3D localization accuracy with substantially reduced temporal correspondence latency compared to SOTA methods and exhibits strong generalization across scenarios.
Chinese Translation
基于音频的定位为反无人机早期预警提供了一种低成本且不依赖光照的感知解决方案。然而，现有方法通常依赖于预定义的固定音频段长度，这限制了时间对应性，并在充分的声学证据与及时定位之间产生了权衡。为了解决这一问题，我们提出了一种具有自适应时间对应的基于音频的定位框架。首先使用探测段提取一个紧凑的声学状态，该状态表征观察的可靠性和一致性。在该状态的指导下，强化学习控制器动态确定每个定位决策所需的音频窗口大小。然后，所选音频段通过基于Mamba的定位网络进行处理，结合自适应时间特征调制进行三维位置估计。大量实验表明，与最先进的方法（SOTA）相比，我们的方法在显著减少时间对应延迟的同时，实现了具有竞争力的三维定位精度，并在不同场景中表现出强大的泛化能力。
cs.RO / 138 / 2609.24253
OpenFlyScan: A Quality-Guided Aerial Reconstruction System for Consumer Drones
OpenFlyScan：一种面向消费级无人机的质量引导航空重建系统
Abstract
3D Gaussian Splatting (3DGS) provides high-fidelity scenes for large-scale embodied simulation, but constructing large-scale urban assets remains constrained by expensive equipment and delayed quality feedback. Preset surveys can leave complex surfaces insufficiently observed, with defects discovered only after reconstruction, requiring return visits and repeated processing. We present OpenFlyScan, a quality-guided aerial reconstruction system for consumer drones that integrates a GS quality model, a reacquisition planner, and a custom-designed mobile app. The model learns from GS rendering errors to predict regional reconstruction quality. Based on these predictions, the planner then generates complementary reacquisition strips to be executed through the app, which also supports automated oblique surveys and data transfer without additional hardware on board. Across real aerial scenes, the model effectively identifies regions that are likely to be poorly reconstructed. In the Expo West field experiment, targeted reacquisition improves PSNR at additional views by 10.95 dB. With consumer drones, OpenFlyScan integrates capture, targeted reacquisition, and reconstruction to support rapid, low-cost urban asset creation. Code and models will be made publicly available at https://openflyscan.github.io/.
Chinese Translation
3D高斯点云（3D Gaussian Splatting, 3DGS）为大规模具身仿真提供了高保真场景，但构建大规模城市资产仍受限于昂贵的设备和延迟的质量反馈。预设调查可能导致复杂表面观察不足，缺陷往往在重建后才被发现，这需要返回现场进行重复处理。我们提出了OpenFlyScan，一种面向消费级无人机的质量引导航空重建系统，集成了GS质量模型、重新获取规划器和定制设计的移动应用。该模型通过学习GS渲染错误来预测区域重建质量。基于这些预测，规划器生成补充的重新获取条带，通过应用程序执行，该应用程序还支持自动倾斜调查和数据传输，无需额外的硬件。在真实航空场景中，该模型有效识别出可能重建不佳的区域。在Expo West现场实验中，针对性重新获取在额外视角下提高了10.95 dB的峰值信噪比（PSNR）。通过消费级无人机，OpenFlyScan整合了捕获、针对性重新获取和重建，以支持快速、低成本的城市资产创建。代码和模型将公开发布在 https://openflyscan.github.io/。
cs.RO / 139 / 2609.24271
ME-Brain-1.0: Memory, Cognition and Action for Evolving Embodied Intelligence
ME-Brain-1.0：用于进化的具身智能的记忆、认知与行动
He, Wei, Li, Hengtao, Yu, Zhongrui, Zhu, Xuhan, He, Maokui, Liu, Zide, Zhang, Xiyue, Mao, Xianwei, Zhou, Chunpeng, Shi, Jia, Xin, Yanze, Li, Jingwen, Zheng, Jingxie, Zeng, Sijie, Wang, Chenfeng, Lu, Fan, Zhang, Zeyu, Guo, Shuai, Zhang, Hengxuan, Yu, Pengfei, Shi, Jia, Liu, Yu, Zhan, Kun, Xie, Yan
Abstract
Current embodied systems largely rely on pretrained capabilities that remain fixed after deployment, limiting their ability to learn from physical interaction. We introduce MachEmbodied-Brain (ME-Brain), a self-evolving embodied system organized around a closed loop of action execution, experience acquisition, experience evolution, and improved execution. Evolvable Memory consolidates multimodal trajectories into hierarchical, reusable experience; Cognitive Core transforms physical experience into transferable skills; and the Action Model combines event-driven keyframes, EventCell local-world prediction, and action-conditioned memory modulation to focus computation on decision-critical moments, regions, and historical evidence. Together, these modules shift embodied intelligence from train-and-freeze to deploy-and-evolve without model retraining. Cognitive Core outperforms the strongest comparison models by 8.2 and 9.6 points on embodied and agent benchmarks. The Action Model achieves 47.88% mean success on RoboMME, a 3.26-point improvement over the strongest baseline. On RoboDojo, it reaches a 21.51 mean Score and 16.03% success rate, exceeding $\pi_{0.5}$ by 10.10 and 9.12 points. On the six-task ME-RealBench, ME-Brain achieves a 69.5 mean Score and 66.7% success rate, outperforming DM0.5 by 12.8 and 11.7 points, respectively.
Chinese Translation
当前的具身系统在很大程度上依赖于在部署后保持固定的预训练能力，这限制了它们从物理交互中学习的能力。我们介绍了MachEmbodied-Brain（ME-Brain），这是一个自我进化的具身系统，围绕行动执行、经验获取、经验演化和改进执行的闭环组织。可进化记忆将多模态轨迹整合为分层的、可重用的经验；认知核心将物理经验转化为可转移的技能；而行动模型结合了事件驱动的关键帧、事件单元（EventCell）局部世界预测和基于行动的记忆调制，专注于决策关键时刻、区域和历史证据的计算。这些模块共同将具身智能从训练-冻结转变为部署-进化，无需模型再训练。认知核心在具身和代理基准测试中分别比最强的对比模型高出8.2和9.6分。行动模型在RoboMME上实现了47.88%的平均成功率，比最强基线提高了3.26分。在RoboDojo上，平均得分达到21.51，成功率为16.03%，超出$ ext{π}_{0.5}$分别达到10.10和9.12分。在六任务的ME-RealBench上，ME-Brain的平均得分为69.5，成功率为66.7%，分别比DM0.5高出12.8和11.7分。
cs.RO / 140 / 2609.24274
vla.simd: Efficient CPU Inference for Language-Conditioned Manipulation
vla.simd：高效的语言条件操作的 CPU 推理
Abstract
Deploying language-conditioned manipulation without a dedicated GPU requires efficient inference and action chunks that cover the delay between policy queries. We present vla.simd, a CPU inference engine that combines shared SIMD micro-kernels, reusable computation, and target-specific optimization. We relate query latency and execution horizon to action availability under lagged and time-aligned execution, distinguishing action supply from feedback frequency. Across six policies and four CPUs, vla.simd achieves approximately $1.4\times$ median speedup over compiled PyTorch references while preserving fp32 numerical fidelity. We also introduce IMPACT, an ACT-based policy with cached text representations and language-modulated visual features. IMPACT is the only language-conditioned policy in our evaluated set that supplies at least 30 actions/s on the Raspberry Pi 5: after a 90 s thermal soak, it supplies 33.5 actions/s in fp32 and 81.2 with int8. Separate GPU evaluations yield $76.4\%$ mean success across four LIBERO suites without robot pretraining; instruction-shuffling tests demonstrate selection among familiar goals. Trials with IMPACT on an SO-101 arm and SmolVLA on a UR10e with a Robotiq gripper demonstrate CPU deployment on two robot embodiments.
Chinese Translation
在没有专用 GPU 的情况下部署语言条件的操作需要高效的推理和覆盖策略查询之间延迟的动作块。我们提出了 vla.simd，这是一种 CPU 推理引擎，结合了共享的 SIMD 微内核、可重用计算和特定目标的优化。我们将查询延迟和执行视野与滞后和时间对齐执行下的动作可用性相关联，区分动作供应与反馈频率。在六种策略和四种 CPU 上，vla.simd 实现了大约 $1.4 imes$ 的中位数加速，相较于编译的 PyTorch 参考，同时保持了 fp32 的数值精度。我们还引入了 IMPACT，这是一种基于 ACT 的策略，具有缓存的文本表示和语言调制的视觉特征。IMPACT 是我们评估集中的唯一一种语言条件策略，在 Raspberry Pi 5 上提供至少 30 个动作/秒：在 90 秒的热浸泡后，它以 fp32 提供 33.5 个动作/秒，以 int8 提供 81.2 个动作/秒。单独的 GPU 评估在没有机器人预训练的情况下，在四个 LIBERO 套件中获得了 $76.4\%$ 的平均成功率；指令打乱测试展示了在熟悉目标之间的选择。使用 IMPACT 在 SO-101 臂和 SmolVLA 在 UR10e 上与 Robotiq 夹具的试验展示了在两个机器人实例上的 CPU 部署。
cs.RO / 141 / 2609.24317
Performance-Preserving Online Adaptation in Social Navigation via Diffusion Steering
通过扩散引导实现社交导航中的性能保留在线适应
Abstract
In social navigation, modeling the complex interactions between humans and robots is difficult, and deep reinforcement learning has therefore been actively studied. However, because simulation alone cannot fully reproduce diverse scenarios, robot dynamics, and the social conventions that vary across deployment environments, fine-tuning in the deployment environment is promising. In doing so, learning that preserves the base model's performance is required, so as not to compromise the primary objective of navigation, namely avoiding pedestrians and reaching the destination. In this study, we propose a method that applies diffusion steering via reinforcement learning (DSRL), which trains only the noise policy while keeping the diffusion policy fixed, thereby achieving learning that preserves performance. Furthermore, we integrate diffusion-based RL policies trained with multiple seeds to construct the base policy, improving learning performance. Our evaluation shows that, compared with other methods, the proposed method enables efficient learning while preserving performance, and we confirm flexible behavior control through adaptation to social conventions, as well as its effectiveness on a physical robot through hardware-in-the-loop simulation.
Chinese Translation
在社交导航中，建模人类与机器人之间复杂的交互关系是困难的，因此深度强化学习得到了积极的研究。然而，仅靠仿真无法完全重现多样化的场景、机器人动态以及在不同部署环境中变化的社会规范，因此在部署环境中的微调显得尤为重要。在此过程中，需要保持基础模型性能的学习，以不妨碍导航的主要目标，即避开行人并到达目的地。本研究提出了一种通过强化学习应用扩散引导的方法（Diffusion Steering via Reinforcement Learning, DSRL），该方法仅训练噪声策略，同时保持扩散策略不变，从而实现性能保留的学习。此外，我们将使用多个种子训练的基于扩散的强化学习策略整合以构建基础策略，从而提高学习性能。我们的评估表明，与其他方法相比，所提方法能够在保留性能的同时实现高效学习，并通过适应社会规范确认灵活的行为控制，同时通过硬件在环仿真验证其在物理机器人上的有效性。
cs.RO / 142 / 2609.24350
LIBERO-VPro: Benchmarking Closed-Loop Visual Robustness of Robotic Foundation Models
LIBERO-VPro：机器人基础模型闭环视觉鲁棒性基准测试
Abstract
Robotic foundation models achieve impressive performance on standard manipulation benchmarks, yet these evaluations typically assume clean, timely, and consistent visual observations throughout execution. We introduce LIBERO-VPro, a benchmark for systematically evaluating the closed-loop visual robustness of robotic foundation models by perturbing the visual evidence available during execution. LIBERO-VPro covers four complementary dimensions, including Visual Evidence Degradation, Camera Staleness, Visual Source Consistency, and Task-Relevant Scene Variation, spanning 12 challenge categories, 96 experimental settings, and 3,296 task-condition cases. We evaluate three vision-language-action models and three world-action models over approximately 196,000 simulated episodes, complemented by 200 real-world rollouts on a Franka Research 3. Our results reveal that strong nominal performance can mask substantial weaknesses in visual grounding and adaptation. Models often remain successful despite severe object-level occlusion, yet degrade sharply when local interaction cues are disrupted or familiar spatial priors are violated. They are also highly sensitive to stale or missing observations and struggle when changed task preconditions require behavioral adaptation. Finally, VLAs and WAMs exhibit distinct robustness profiles, showing that visual robustness is multi-dimensional and architecture-dependent. LIBERO-VPro provides a systematic diagnostic framework for developing robotic foundation models that can more reliably ground and adapt their actions under challenging visual conditions.
Chinese Translation
机器人基础模型在标准操作基准测试中表现出色，但这些评估通常假设在执行过程中视觉观察是干净、及时且一致的。我们引入了LIBERO-VPro，这是一个系统评估机器人基础模型闭环视觉鲁棒性的基准，通过扰动执行过程中的视觉证据。LIBERO-VPro涵盖四个互补维度，包括视觉证据退化、相机时效性、视觉源一致性和任务相关场景变化，涉及12个挑战类别、96个实验设置和3,296个任务条件案例。我们对三种视觉-语言-动作模型和三种世界-动作模型进行了评估，涵盖约196,000个模拟回合，并在Franka Research 3上进行了200个真实世界的实验。我们的结果表明，强大的名义性能可能掩盖视觉基础和适应性方面的重大弱点。尽管在严重的物体级遮挡情况下模型通常仍然成功，但当局部交互线索受到干扰或熟悉的空间先验被违反时，性能会急剧下降。模型对过时或缺失的观察高度敏感，并且在改变任务前提条件需要行为适应时表现不佳。最后，视觉-语言-动作模型和世界-动作模型表现出不同的鲁棒性特征，表明视觉鲁棒性是多维的且依赖于架构。LIBERO-VPro提供了一个系统的诊断框架，以开发能够在挑战性视觉条件下更可靠地确定和适应其行为的机器人基础模型。
cs.RO / 143 / 2609.24376
Multi-Agent Transportation of Free-Flyers in Microgravity Via Pushing Interaction Under Human-in-the-Loop Control
在微重力下通过推力交互进行自由飞行器的多智能体运输，采用人机协同控制
Abstract
We propose a safety-critical framework for the cooperative transportation of passive targets in microgravity, where a team of chaser robots acts through unilateral pushing contacts to track a human-provided desired twist while ensuring safe target motion. The pushing-only nature of the interaction introduces sparse, configuration-dependent actuation constraints requiring chasers to physically relocate on the target body when the desired pushing allocation changes. To address these challenges, we formulate a delay-aware feedback control architecture leveraging Control Lyapunov Function (CLF) and Control Barrier Function (CBF) constraints within a mixed-integer thrust allocation program to enforce stability and safety of the target, respectively. The proposed framework enables reference tracking while guaranteeing obstacle avoidance with a circular obstacle despite intermittent control authority, providing a foundation for human-supervised cooperative transportation of free-flyers in space environments. The proposed framework is validated through Gazebo simulations.
Chinese Translation
我们提出了一种安全关键框架，用于在微重力环境中协作运输被动目标，其中一组追逐机器人通过单向推力接触来跟踪人类提供的期望运动状态，同时确保目标的安全运动。推力交互的单一性质引入了稀疏的、依赖于配置的驱动约束，要求追逐者在期望推力分配变化时物理地重新定位于目标体上。为了解决这些挑战，我们构建了一种延迟感知反馈控制架构，利用控制李雅普诺夫函数（Control Lyapunov Function, CLF）和控制障碍函数（Control Barrier Function, CBF）约束，在混合整数推力分配程序中实施，以分别确保目标的稳定性和安全性。该框架能够在保证避障的同时进行参考跟踪，即使在间歇性控制权限的情况下，也能有效应对圆形障碍物，为在太空环境中进行人机协作的自由飞行器运输奠定了基础。所提出的框架通过Gazebo仿真进行了验证。
cs.RO / 144 / 2609.24385
Tactile-JEPA: Topology-Aware Self-Supervised Representation Learning for Distributed Tactile Sensors
触觉-JEPA：面向拓扑的自监督表示学习用于分布式触觉传感器
Abstract
Tactile sensing is an essential modality for robots performing contact-rich, dexterous manipulation, particularly under visual occlusion. While pre-trained image encoders are standard in robot learning pipelines, tactile encoders are still commonly trained from scratch from raw, noisy signals, which might limit their expressivity. Existing self-supervised learning (SSL) approaches focus predominantly on vision-based tactile sensors, leaving distributed electronic skins largely unaddressed. These sensors, however, have a distinctive property: their sensing elements are sparse and irregularly arranged over the surface they cover, which makes direct reuse of visual SSL methods suboptimal. We present Tactile-JEPA, an efficient self-supervised pre-training method that uses the spatial arrangement of tactile sensors to learn topology-aware representations. Specifically, it is trained to predict the embeddings of masked sensing elements from the unmasked remainder, using the sensor connectivity graph to guide spatial masking. Our analysis shows that effective tactile representations require capturing both local contact details and the global state of the tactile surface, which we achieve through dual-scale masking. Across three diverse datasets spanning magnetic and piezoresistive sensors, different robot embodiments, and single- and paired-sensor configurations, Tactile-JEPA reduces force estimation error by 6.3% and in-hand orientation error by 20.8% over the prior state-of-the-art, with consistent gains in other downstream applications, including policy learning. Overall, our results demonstrate that the benefit of tactile sensing depends critically on the quality of encoder pre-training, a problem which Tactile-JEPA addresses directly. Code is available at https://github.com/E-Kovtun/tactile.
Chinese Translation
触觉感知是机器人在视觉遮挡下进行接触丰富的灵巧操作的重要模态。尽管预训练的图像编码器在机器人学习流程中是标准配置，但触觉编码器仍然通常从原始的、嘈杂的信号中从零开始训练，这可能限制了它们的表现力。现有的自监督学习（SSL）方法主要集中在基于视觉的触觉传感器上，而对分布式电子皮肤的研究则相对较少。然而，这些传感器具有一个独特的特性：它们的传感元件在覆盖的表面上稀疏且不规则排列，这使得直接重用视觉SSL方法的效果不佳。我们提出了触觉-JEPA，这是一种高效的自监督预训练方法，利用触觉传感器的空间排列来学习面向拓扑的表示。具体而言，它被训练以从未遮挡的其余部分预测被遮挡的传感元件的嵌入，使用传感器连接图来指导空间遮挡。我们的分析表明，有效的触觉表示需要同时捕捉局部接触细节和触觉表面的全局状态，我们通过双尺度遮挡实现了这一目标。在涵盖磁性和压阻传感器、不同机器人形态以及单传感器和配对传感器配置的三种多样化数据集中，触觉-JEPA将力估计误差降低了6.3%，将手中方向误差降低了20.8%，并在其他下游应用（包括策略学习）中取得了一致的提升。总体而言，我们的结果表明，触觉感知的效益在很大程度上依赖于编码器预训练的质量，而这是触觉-JEPA直接解决的问题。代码可在 https://github.com/E-Kovtun/tactile 获取。
cs.RO / 145 / 2609.24411
Zeva-Ego: Egocentric Mid-Training with In-Context Causal Learning for Robot Manipulation
Zeva-Ego：基于上下文因果学习的自我中心中期训练框架用于机器人操控
Abstract
Egocentric video offers a scalable source of physical interaction experience, yet translating it into robot-executable knowledge and enabling continual adaptation remain challenging. We introduce Zeva-Ego, a unified framework that learns physical priors from human experience and evolves through robot interaction. An Action-Centric Encoder (ACE) converts egocentric visual transitions into action-centered supervision for VLA mid-training, while In-Context Causal Learning (ICCL) enables parameter-free adaptation from action-effect feedback at deployment. Scaling Ego data to 10K hours improves RoboTwin success from 63.8% to 75.3%, matching 2K hours of robot demonstrations (74.7%), corresponding to an empirical data ratio of roughly 4-5:1. With accumulated interaction experience, ICCL further improves success from 58% to 89% within four attempts without parameter updates. These results demonstrate a scalable path toward embodied intelligence that learns from human experience and continuously improves through its own interaction.
Chinese Translation
自我中心视频提供了一种可扩展的物理交互经验来源，但将其转化为机器人可执行的知识并实现持续适应仍然面临挑战。我们提出了Zeva-Ego，一个统一框架，从人类经验中学习物理先验，并通过机器人交互不断演化。一个动作中心编码器（Action-Centric Encoder, ACE）将自我中心的视觉转变转换为以动作为中心的监督，用于VLA中期训练，而基于上下文的因果学习（In-Context Causal Learning, ICCL）则使得在部署时能够根据动作效果反馈进行无参数适应。将自我中心数据扩展到10K小时，使RoboTwin的成功率从63.8%提高到75.3%，与2K小时的机器人演示（74.7%）相匹配，经验数据比例大约为4-5:1。随着交互经验的积累，ICCL在四次尝试中进一步将成功率从58%提高到89%，且无需参数更新。这些结果展示了一条可扩展的路径，朝着从人类经验中学习并通过自身交互持续改进的具身智能发展。
cs.RO / 146 / 2609.24413
Robotic Valve Turning: Axial Misalignment Correction Using Reaction Torque Feedback
机器人阀门旋转：利用反作用力反馈进行轴向失调修正
Abstract
In this work, we propose a haptic update control law that uses reaction torques to correct axial misalignment during robotic valve manipulation. Unlike vision-based estimates, which can be affected by calibration errors, occlusion, and uncertainty in the contact geometry, reaction torques arise directly from the physical interaction between the gripper and valve. A geometric relationship exists between the error (misalignment) vector and these torques. The primary aim of this work is to propose a stable controller exploiting this geometric property. Our control law is proven to be uniformly asymptotically stable. Simulations are performed for verification. Furthermore, we experimentally test the robustness of our method using a Kinova Gen3 robotic arm for initial misalignments ranging from $-15^\circ$ to $15^\circ$ at 3 different valve positions and report the resulting data distribution. The absolute value of the median misalignment across all 18 test cases is found to be within $2.46^\circ$ and that of reaction torques within $0.23\mathrm{Nm}$.
Chinese Translation
在本研究中，我们提出了一种触觉更新控制律，该控制律利用反作用力来修正机器人阀门操作中的轴向失调。与可能受到校准误差、遮挡和接触几何不确定性影响的基于视觉的估计不同，反作用力直接源于夹持器与阀门之间的物理交互。误差（失调）向量与这些力矩之间存在几何关系。本研究的主要目的是提出一种利用这一几何特性的稳定控制器。我们的控制律被证明是均匀渐近稳定的。进行了仿真以进行验证。此外，我们使用Kinova Gen3机器人手臂对我们的方法的鲁棒性进行了实验测试，初始失调范围为$-15^ ext{°}$到$15^ ext{°}$，在3个不同的阀门位置下进行测试，并报告了结果数据分布。所有18个测试案例中，失调的绝对值中位数被发现为$2.46^ ext{°}$，反作用力的绝对值中位数为$0.23 ext{Nm}$。
cs.RO / 147 / 2609.24433
FoldQuantVLA: Native Low-Bit Quantization of Vision-Language-Action Models via Consistent Folding
FoldQuantVLA：通过一致折叠实现视觉-语言-动作模型的原生低比特量化
Abstract
Low-bit vision-language-action inference must reduce observation-to-action latency while preserving robot behavior. We present FoldQuantVLA, a post-training quantization framework that carries a consistent activation representation through calibration, weight rounding, and native integer execution. It combines channel scaling and block Hadamard transforms with dynamic per-token quantization, without policy retraining. Custom TensorRT plugins execute projections in both the language backbone and iterative action expert with four-bit weights and activations (W4A4) on Ada GPUs and Jetson AGX Orin. Evaluation spans LIBERO, SimplerEnv, and two robot platforms. Across three GR00T checkpoints and $\pi_{0.5}$, W4A4 achieves $1.20$ to $1.33\times$ speedups over floating-point TensorRT on Orin and $1.25$ to $1.52\times$ on desktop. Retaining language attention-output and feed-forward down projections at eight bits (W8A8) improves held-out action fidelity on all four checkpoints. Across four real-robot tasks, this configuration raises observed GR00T N1.7 success from $80.0\%$ with uniform W4A4 to $92.5\%$ over 80 trials per configuration, with a measured additional Orin latency of 1 ms.
Chinese Translation
低比特视觉-语言-动作推理必须在保持机器人行为的同时减少观察到动作的延迟。我们提出了FoldQuantVLA，这是一个后训练量化框架，通过校准、权重舍入和原生整数执行保持一致的激活表示。它结合了通道缩放和块Hadamard变换与动态每标记量化，无需策略重新训练。自定义TensorRT插件在Ada GPU和Jetson AGX Orin上以四比特权重和激活（W4A4）执行语言主干和迭代动作专家中的投影。评估涵盖LIBERO、SimplerEnv和两个机器人平台。在三个GR00T检查点和$ ext{π}_{0.5}$上，W4A4在Orin上实现了$1.20$到$1.33 imes$的速度提升，在桌面上实现了$1.25$到$1.52 imes$的速度提升。保持语言注意力输出和前馈下投影在八比特（W8A8）下提高了所有四个检查点的保留动作保真度。在四个真实机器人任务中，该配置将观察到的GR00T N1.7成功率从均匀的W4A4的$80.0\%$提升至每个配置80次试验的$92.5\ ext{ extperthousand}$，测得的额外Orin延迟为1毫秒。
cs.RO / 148 / 2609.24497
ScaleMPA: Rethinking Scalable RRT* Acceleration With a Grid-Native Representation
ScaleMPA：重新思考基于网格的可扩展 RRT* 加速
Abstract
Real-time motion planning remains challenging in large and high-dimensional environments. Prior acceleration of RRT* follows tree-centric state organization, which reduces per-query cost but preserves superlinear end-to-end complexity and limits parallelism through structural dependencies. This paper presents ScaleMPA, a motion-planning accelerator that rethinks RRT* with a grid-native representation. By replacing hierarchical traversal with direct grid-based access, ScaleMPA reduces the planner critical path and exposes fine-grained parallelism. To make this reformulation practical under sparse high-dimensional planning, ScaleMPA further proposes a multi-resolution grid search engine and a hash-grid memory system. Implemented in 28 nm CMOS, ScaleMPA achieves millisecond-level planning latency and delivers 4.7$\times$--44.4$\times$ speedup over state-of-the-art motion-planning accelerators.
Chinese Translation
在大型和高维环境中，实时运动规划仍然面临挑战。之前对 RRT* 的加速遵循树中心的状态组织，这降低了每次查询的成本，但保留了超线性的端到端复杂性，并通过结构依赖限制了并行性。本文提出了 ScaleMPA，一种重新思考 RRT* 的运动规划加速器，采用基于网格的表示。通过用直接的网格访问替代层次遍历，ScaleMPA 减少了规划器的关键路径，并揭示了细粒度的并行性。为了使这一重构在稀疏高维规划中具有实用性，ScaleMPA 进一步提出了多分辨率网格搜索引擎和哈希网格存储系统。在 28 纳米 CMOS 中实现的 ScaleMPA 达到了毫秒级的规划延迟，并在速度上比最先进的运动规划加速器提高了 4.7$ imes$--44.4$ imes$。
cs.RO / 149 / 2609.24499
A Monolithic Force-Proprioception Soft Acutuator Enabled by Single-Material 3D printing
单材料3D打印实现的单体力觉软驱动器
Abstract
Pneumatic proprioceptive actuators integrate actuation and sensing for soft robots that attract interest due to functional potential. Existing approaches often suffer from assembly errors or stress concentrations caused by heterogeneous materials. In this work, we propose the Monolithic Force-Proprioception Soft (MFPS) design and fabrication method that integrates an Asymmetric Origami Bending (AOB) chamber and a Force-Proprioception Soft (FPS) sensor with single material through one-step Fused Deposition Modeling (FDM) fabrication. Based on the resistance response to strain of conductive thermoplastic polyurethane (TPU), we design and analyze the structure of the FPS sensor, and conduct parametric analysis on the sensing characteristics. The FDM fabrication parameters of the MFPS actuator are analyzed, followed by actuator fabrication and characterization of the actuation and proprioception performance. Experimental results show that the MFPS actuator achieves a bending angle of 40{\deg}, an output force of 12.5 N, and a resistance change of 26.9% as the applied external force increased from 0 to 45 N. A two-finger force-proprioception gripper is developed based on the MFPS actuator. The grasping and force-proprioception capabilities are experimentally validated, proving that the MFPS design method provides a new approach for the development of self-sensing actuators.
Chinese Translation
气动自感知驱动器将驱动与传感集成，因其功能潜力而受到软机器人领域的关注。现有方法常常受到组装误差或由异质材料引起的应力集中问题的困扰。在本研究中，我们提出了一种单体力觉软驱动器（Monolithic Force-Proprioception Soft，MFPS）的设计与制造方法，该方法通过一步熔融沉积建模（Fused Deposition Modeling，FDM）将不对称折纸弯曲（Asymmetric Origami Bending，AOB）腔体与力觉传感器（Force-Proprioception Soft，FPS）集成于单一材料中。基于导电热塑性聚氨酯（TPU）对应变的电阻响应，我们设计并分析了FPS传感器的结构，并对其传感特性进行了参数分析。随后分析了MFPS驱动器的FDM制造参数，并进行了驱动器的制造及其驱动和自感知性能的表征。实验结果表明，MFPS驱动器在施加外力从0增加到45 N时，弯曲角度达到40°，输出力为12.5 N，电阻变化为26.9%。基于MFPS驱动器开发了一个双指力觉抓取器。抓取和力觉能力经过实验验证，证明MFPS设计方法为自感知驱动器的开发提供了一种新途径。
cs.RO / 150 / 2609.24507
TACIT: Tactile Contact Supervision for Spatial Attention in Dexterous Manipulation
TACIT：用于灵巧操作中空间注意力的触觉接触监督
Abstract
Visuomotor policies trained from a few demonstrations may reproduce demonstrated trajectories without reliably following changes in object position. Existing approaches with explicit attention typically obtain spatial priors from human annotation or visual models. We introduce TACIT (tactile contact informs attention), which uses measured tactile contacts from teleoperated demonstrations to supervise spatial attention without additional point annotation. Gaussian targets over preceding camera point clouds supervise an attention head whose pooled output conditions a visuotactile diffusion policy. Targets are used only during training; tactile observations remain inputs at inference. In the primary real-robot benchmark, with ten demonstrations per task and five demonstrated placement regions, TACIT achieves 66.7% success on ball placement and 73.3% on peg insertion, compared with 10.0% and 20.0% for input-matched 3D visuotactile fusion and 20.0% and 43.3% for vision-only DP3. TACIT enters the 150 mm palm-to-object approach region within 12 seconds in all 30 trials per task; all remaining failures occur after arrival. Across three training seeds on real ball and simulated peg, TACIT outperforms input-matched fusion and an architecture-matched control without explicit attention supervision, supporting the contribution of supervision beyond branch capacity. Pre-contact and contact-time supervision show no consistent ordering. These results demonstrate that measured tactile contact provides effective spatial supervision for approach behavior from few demonstrations within the evaluated workspace.
Chinese Translation
从少量示范中训练的视觉运动策略可能会再现示范轨迹，但无法可靠地跟随物体位置的变化。现有的显式注意力方法通常通过人工标注或视觉模型获得空间先验。我们提出了TACIT（触觉接触信息化注意力），该方法利用来自遥操作示范的测量触觉接触来监督空间注意力，而无需额外的点标注。在前置相机点云上使用高斯目标来监督一个注意力头，其汇聚输出条件化一个视觉触觉扩散策略。目标仅在训练期间使用；触觉观察在推理时仍然作为输入。在主要的真实机器人基准测试中，每个任务有十个示范和五个示范放置区域，TACIT在球体放置任务中取得了66.7%的成功率，在插销插入任务中取得了73.3%的成功率，而输入匹配的3D视觉触觉融合分别仅为10.0%和20.0%，视觉单一的DP3则为20.0%和43.3%。在每个任务的30次试验中，TACIT在12秒内进入150毫米的掌心到物体接近区域；所有剩余的失败发生在到达之后。在真实球体和模拟插销的三个训练种子中，TACIT的表现优于输入匹配的融合和没有显式注意力监督的架构匹配控制，支持了监督超出分支能力的贡献。预接触和接触时间监督没有显示出一致的排序。这些结果表明，测量的触觉接触为从少量示范中在评估工作空间内的接近行为提供了有效的空间监督。
cs.RO / 151 / 2609.24511
InsertAnything: Generalizable Contact-Rich Precision Insertion from Simulation to Reality
插入任意物体：从仿真到现实的可推广接触丰富精确插入
Abstract
Contact-rich precision insertion is a key manipulation skill in robotic assembly. Tight clearances make insertion more sensitive to alignment errors and prone to collisions and jamming, while variations in geometry and clearance across parts further complicate policy reuse. We present a reinforcement learning framework that trains insertion policies entirely in simulation for direct deployment without real-world demonstrations or policy fine-tuning. By combining target poses with compact three-dimensional fingertip force feedback, the policy learns to search for alignment and correct its motion despite errors in the estimated hole position. A decoupled gated reward coordinates alignment and insertion. Force-signal smoothing and state-independent standard deviations stabilize the learning process. The resulting policies perform real-world insertion across multiple hole geometries with a minimum nominal clearance of 0.02 mm and improve success while reducing peak contact forces under hole-position errors. Cross-clearance and cross-geometry evaluations further confirm policy generalization. The system achieved the first perfect score of 20/20 on ManipulationNet's peg-in-hole benchmark under its Human-in-the-Loop protocol, with fully autonomous insertion motions. A single policy trained only on a simulated hexagonal insertion task achieved an overall success rate of 95.0% across eight unseen real-world insertion tasks. These results show that learning entirely in simulation can yield precision insertion skills that can be deployed directly and reused across real-world tasks. The project website (https://mzhsoul.github.io/InsertAnything/) provides open-source simulation and real-robot experiment scripts, assets, and trained checkpoints.
Chinese Translation
接触丰富的精确插入是机器人组装中的一项关键操作技能。紧密的间隙使得插入对对齐误差更加敏感，并容易发生碰撞和卡塞，而部件之间几何形状和间隙的变化进一步复杂化了策略的重用。我们提出了一种强化学习框架，该框架在仿真中完全训练插入策略，以便在没有现实世界演示或策略微调的情况下直接部署。通过将目标姿态与紧凑的三维指尖力反馈相结合，该策略学习在对齐中搜索并纠正其运动，尽管存在对孔位置估计的误差。解耦的门控奖励协调了对齐和插入。力信号平滑和状态无关的标准差稳定了学习过程。最终得到的策略能够在多种孔几何形状下进行现实世界的插入，最小名义间隙为0.02毫米，并在孔位置误差下提高成功率，同时减少峰值接触力。跨间隙和跨几何形状的评估进一步确认了策略的推广能力。该系统在ManipulationNet的孔插入基准测试中，在其人机协作协议下首次获得了20/20的完美分数，且实现了完全自主的插入动作。仅在模拟六边形插入任务上训练的单一策略在八个未见的现实世界插入任务中达到了95.0%的总体成功率。这些结果表明，完全在仿真中学习可以产生可以直接部署并在现实世界任务中重用的精确插入技能。项目网站（https://mzhsoul.github.io/InsertAnything/）提供了开源的仿真和真实机器人实验脚本、资产和训练检查点。
cs.RO / 152 / 2609.24525
Bridge3D: Enabling Vision-Language-Action Models to See and Act in 3D
Bridge3D：使视觉-语言-动作模型能够在三维空间中观察和行动
Abstract
Vision-Language-Action (VLA) models have demonstrated remarkable generalization in robotic manipulation via large-scale multimodal pretraining. However, VLA models are mainly trained on 2D-centric observations, which inherently constrains their capacity for precise spatial manipulation. Previous methods enhance 3D awareness by introducing implicit spatial priors, but still lack explicit geometry guidance. In this paper, we propose Bridge3D that integrates both implicit and explicit 3D geometry guidance into pre-trained 2D VLA models, enabling them to ''see'' and ''act'' in 3D. Bridge3D introduces two strategies: 1) Implicit Fusion, which enriches visual tokens with features from 3D foundation models to improve ''seeing'' in 3D; 2) Explicit Conditioning, which integrates action denoising with an explicit 3D semantic field to achieve ''acting'' in 3D. Furthermore, we utilize the proposed layer-wise linear probing to improve learning efficiency. Experiments show that Bridge3D achieves superior performance against state-of-the-art methods. On the RoboTwin 2.0 benchmark, Bridge3D exceeds $\pi_0$ by 14.0 percentage points, while in real-world experiments, it outperforms Spatial Forcing by 11.7 percentage points. These results demonstrate Bridge3D's strong capabilities in high-precision and spatial-sensitive manipulation tasks.
Chinese Translation
视觉-语言-动作（VLA）模型通过大规模多模态预训练在机器人操作中展现了显著的泛化能力。然而，VLA模型主要在以二维为中心的观察数据上进行训练，这在本质上限制了其进行精确空间操作的能力。以往的方法通过引入隐式空间先验来增强三维意识，但仍然缺乏明确的几何指导。本文提出了Bridge3D，它将隐式和显式的三维几何指导整合到预训练的二维VLA模型中，使其能够在三维空间中“看见”和“行动”。Bridge3D引入了两种策略：1）隐式融合（Implicit Fusion），通过将三维基础模型的特征丰富视觉标记，以改善三维“观察”；2）显式条件（Explicit Conditioning），将动作去噪与显式三维语义场结合，以实现三维“行动”。此外，我们利用所提出的逐层线性探测方法提高学习效率。实验表明，Bridge3D在与最先进的方法相比时表现出优越的性能。在RoboTwin 2.0基准测试中，Bridge3D超过了$ ext{π}_0$ 14.0个百分点，而在现实世界实验中，其表现优于Spatial Forcing 11.7个百分点。这些结果展示了Bridge3D在高精度和空间敏感的操作任务中的强大能力。
cs.RO / 153 / 2609.24535
Estimation and Control of Tensegrity Manipulator Kinematics based on Strut Inclination Angles
基于支杆倾斜角的张力结构操控器运动学估计与控制
Abstract
Unlike conventional rigid-link robots defined by discrete joints, continuum robots pose a fundamental challenge for expressing their complex continuous bending configurations for closed-loop control. Several modelling approaches have been proposed for conventional continuum robots, but tensegrity-based continuum robots remain largely open. Moreover, many of these approaches assume a continuous elastic backbone and are therefore not directly applicable to tensegrity manipulators, whose bodies are networks of rigid struts and tensioned cables. This work presents a reduced-order model for shape and posture control of a tensegrity-based continuum manipulator. The manipulator is modelled as a serially connected parallel-link mechanism. The proposed method is formulated as an optimization problem that uses geometric constraints of the tensegrity structure together with information from the Inertial Measurement Unit (IMU) sensors embedded in the strut elements. To the best of our knowledge, this work presents the first experimental demonstration of a real-time IMU-based shape estimation method on a full-scale tensegrity manipulator and demonstrates posture control using a simple Proportional-Integral (PI) controller. The results show that the proposed method can estimate the shape of both single-module tensegrity structures and multi-module tensegrity manipulators from arbitrary static configurations and achieve desired postures.
Chinese Translation
与由离散关节定义的传统刚性链机器人不同，连续机器人在表达其复杂的连续弯曲配置以实现闭环控制方面面临根本挑战。虽然已经提出了几种传统连续机器人的建模方法，但基于张力结构的连续机器人仍然在很大程度上处于开放状态。此外，这些方法中的许多假设了一个连续的弹性主干，因此不适用于张力结构操控器，其主体是由刚性支杆和张紧电缆构成的网络。本研究提出了一种用于基于张力结构的连续操控器形状和姿态控制的降阶模型。该操控器被建模为一个串联连接的并联链机制。所提出的方法被形式化为一个优化问题，利用张力结构的几何约束以及嵌入支杆元件中的惯性测量单元（IMU）传感器的信息。根据我们所知，本研究首次在全尺度张力结构操控器上展示了基于IMU的实时形状估计方法，并使用简单的比例-积分（PI）控制器实现姿态控制。结果表明，所提出的方法能够从任意静态配置中估计单模块张力结构和多模块张力操控器的形状，并实现期望的姿态。
cs.RO / 154 / 2609.24547
MIRA: Real-Time Full-Duplex Human-Robot Interaction for Embodied Companions
MIRA：用于具身伴侣的实时全双工人机交互
Abstract
Real-time embodied companion interaction requires a robot to infer user intent from streaming speech, generate timely responses, and execute expressive, interruptible motions. Existing systems typically decouple dialogue orchestration from gesture synthesis, relying on offline motion generation from complete audio. This separation leaves open how a deployed robot can dynamically synchronize response content, prosodic timing, and physical safety under incremental inputs and uncertain turn boundaries. We present MIRA, a unified framework for full-duplex embodied companion interaction. Given streaming user speech, dialogue history, and vocal affect, MIRA predicts both the response text and an explicit embodiment cue. Discrete social behaviors (\eg listening, greeting) are mapped to validated robot trajectories, while open-ended speaking is paired with streaming, co-speech motion. This generative motion is governed by a predict-more-than-commit sliding window that provides temporal look-ahead for motion continuity while limiting physical commitment to a short, cancellable prefix. Crucially, we design CORTEX, a dual-timescale interaction policy that manages low-latency streaming and deliberative turn decisions, backed by a robot-side execution layer that enforces physical safety constraints at the control rate. We deploy MIRA on an Astribot S1 humanoid robot. Quantitative evaluations demonstrate competitive audio-motion alignment relative to state-of-the-art motion-generation baselines, while real-robot deployment measurements characterize streaming responsiveness and interruption handling.
Chinese Translation
实时具身伴侣交互要求机器人从流式语音中推断用户意图，生成及时响应，并执行富有表现力的、可中断的动作。现有系统通常将对话协调与手势合成解耦，依赖于从完整音频中离线生成动作。这种分离使得已部署的机器人在增量输入和不确定的轮次边界下，如何动态同步响应内容、韵律时机和物理安全仍然是一个未解的问题。我们提出了MIRA，一个用于全双工具身伴侣交互的统一框架。给定流式用户语音、对话历史和语音情感，MIRA同时预测响应文本和明确的具身提示。离散的社交行为（例如倾听、问候）被映射到经过验证的机器人轨迹，而开放式发言则与流式的共语动作相结合。这种生成的动作由一个“预测多于承诺”的滑动窗口控制，该窗口为动作的连续性提供时间前瞻，同时将物理承诺限制在一个短暂的、可取消的前缀内。关键的是，我们设计了CORTEX，一个双时间尺度的交互策略，管理低延迟流式处理和深思熟虑的轮次决策，并由一个机器人端执行层支持，该执行层在控制速率下强制执行物理安全约束。我们在Astribot S1人形机器人上部署了MIRA。定量评估表明，相较于最先进的动作生成基线，MIRA在音频-动作对齐方面表现出竞争力，而真实机器人部署的测量则表征了流式响应性和中断处理能力。
cs.RO / 155 / 2609.24552
Smoothness as a Constraint for Stable Humanoid Locomotion
平滑性作为稳定类人运动的约束
Abstract
Embodied AI systems, particularly humanoid robots deployed in real world scenarios require whole-body control policies that are both task-responsive and physically smooth. However, smoothness is not uniform across the body: lower body must remain sufficiently reactive, while the upper body must be tightly regulated to preserve stability. Existing reinforcement learning approaches typically impose smoothness through auxiliary terms in the reward function, which compete with task objectives, treating the body as uniform and provide no direct control over the physical quantities responsible for smooth behavior. We introduce DeCap (Decoupled Constraint-aware policy), a constrained reinforcement learning algorithm that decouples whole-body smoothness into separate upper- and lower-body constraint groups, each formulates smoothness as explicit constraints on physical motion limits. To improve constraint satisfaction near feasibility boundaries, DeCap incorporates a bounded barrier penalty that activates proactively as limits are approached and remains bounded at the constraint limit. On real-world humanoid whole-body control task, DeCap reduces upper-body action rate by 2.50x and acceleration by 2.18x relative to reward-based smoothness policies, while also improving lower-body smoothness and reducing transient motion. We demonstrate that a fixed set of smoothness constraints transfers across diverse terrains, alleviating the need of extensive reward tuning.
Chinese Translation
具身人工智能系统，特别是在现实场景中部署的类人机器人，需要既能响应任务又具备物理平滑性的全身控制策略。然而，平滑性在身体各部位并不均匀：下半身必须保持足够的反应能力，而上半身则必须严格调控以保持稳定性。现有的强化学习方法通常通过奖励函数中的辅助项来施加平滑性，这与任务目标相竞争，将身体视为均匀体，并且对负责平滑行为的物理量没有直接控制。我们提出了DeCap（解耦约束感知策略），这是一种约束强化学习算法，将全身平滑性解耦为独立的上半身和下半身约束组，每个组将平滑性公式化为对物理运动限制的显式约束。为了改善在可行边界附近的约束满足，DeCap引入了一种有界障碍惩罚，当接近限制时主动激活，并在约束限制处保持有界。在现实世界的类人全身控制任务中，DeCap相较于基于奖励的平滑策略，将上半身动作速率降低了2.50倍，加速度降低了2.18倍，同时改善了下半身的平滑性并减少了瞬态运动。我们证明了一组固定的平滑性约束可以在不同地形中转移，从而减轻了对广泛奖励调优的需求。
cs.RO / 156 / 2609.24563
ARSTAG: An Agentic Real2Sim2Real System for Task-Specific Robot Data Generation
ARSTAG：一种用于任务特定机器人数据生成的自主Real2Sim2Real系统
Abstract
Adapting visuomotor policies to new manipulation tasks often requires substantial manual engineering or teleoperated data collection. Simulation can provide task-specific data at scale, but constructing the scene, designing expert behavior, and configuring data generation still require significant per-task effort. We present ARSTAG, an agentic Real2Sim2Real system that turns a single RGB image and a natural-language instruction directly into robot policy-learning data. A hierarchy of language agents constructs a task-scoped simulation scene, generates robot-feasible demonstrations, and expands the training distribution through task-consistent randomization, while a coordinator agent manages cross-stage feedback and recovery. Across seven manipulation tasks spanning grasping, placement, and stacking, the ARSTAG-generated demonstrations enable sim-to-real transfer of three visuomotor policy architectures to a dual-arm robot, with pi0.5 achieving an average real-world success rate of 74.6%. Ablations show that task-consistent randomization substantially improves robustness, and policy performance increases with generated dataset size. Project webpage: https://boweili666.github.io/ARSTAG/.
Chinese Translation
将视觉运动策略适应于新的操作任务通常需要大量的手动工程或遥控数据收集。模拟可以大规模提供任务特定的数据，但构建场景、设计专家行为和配置数据生成仍然需要针对每个任务进行大量的努力。我们提出了ARSTAG，一种自主的Real2Sim2Real系统，它能够将单个RGB图像和自然语言指令直接转化为机器人策略学习数据。一系列语言代理构建了任务范围的模拟场景，生成机器人可行的演示，并通过任务一致的随机化扩展训练分布，而协调代理则管理跨阶段的反馈和恢复。在涵盖抓取、放置和堆叠的七个操作任务中，ARSTAG生成的演示使三种视觉运动策略架构能够成功转移到双臂机器人上，其中pi0.5在现实世界中的平均成功率达到74.6%。消融实验表明，任务一致的随机化显著提高了鲁棒性，而策略性能随着生成数据集规模的增加而提升。项目网页：https://boweili666.github.io/ARSTAG/
cs.RO / 157 / 2609.24576
What do VLM-Based Vision-Language Navigation Models Rely on: Interpreting and Steering Policy Behavior
基于视觉语言模型的视觉-语言导航模型依赖于什么：解释和引导策略行为
Abstract
Modern Vision-Language Navigation (VLN) models rely mostly on pre-trained large Vision-Language Models (VLMs) to predict navigation actions. While this fusion of language instructions and visual observations allows multimodal reasoning, it obscures how information is routed across modalities or what mechanisms drive navigation decisions. Thus, it remains unclear whether VLN models ground their predictions in relevant semantic cues or can track task progress. In this work, we study the interpretability and steerability of VLN models. We use intervention-based metrics that measure how visual observations, instructions, and visual memory causally influence navigation decisions. Our results show that these navigation policies are sensitive to all input modalities and do not depend on a single one. We further show that these agents encode navigation progress and retain semantic structure from their VLM backbones, enabling concept-level steering through internal activations. Finally, we extract activation vectors for abstract behaviors to transfer them zero-shot to out-of-distribution real-world scenarios, improving performance without additional fine-tuning.
Chinese Translation
现代视觉-语言导航（VLN）模型主要依赖于预训练的大型视觉-语言模型（VLM）来预测导航动作。虽然这种语言指令与视觉观察的融合允许多模态推理，但却模糊了信息在不同模态之间的传递方式以及驱动导航决策的机制。因此，目前尚不清楚VLN模型是否将其预测建立在相关的语义线索上，或是否能够跟踪任务进展。在本研究中，我们探讨了VLN模型的可解释性和可引导性。我们使用基于干预的度量来衡量视觉观察、指令和视觉记忆如何因果地影响导航决策。我们的结果表明，这些导航策略对所有输入模态都敏感，而不是依赖于单一模态。我们进一步展示，这些代理能够编码导航进展，并从其VLM骨干网络中保留语义结构，从而通过内部激活实现概念级的引导。最后，我们提取了抽象行为的激活向量，以零样本方式将其转移到分布外的现实场景中，提升了性能而无需额外的微调。
cs.RO / 158 / 2609.24621
Learning tactile perception from high-bandwidth single-point sensing
从高带宽单点传感中学习触觉感知
Abstract
Tactile sensing is increasingly being incorporated into learning-based robotic manipulation, yet many existing approaches rely on spatially distributed sensors. Here we introduce SpectRobot, a framework that transforms single-point tactile signals into compact time-frequency spectrograms. These spectrograms encode high-bandwidth tactile histories as fixed-size image-like representations. They can be processed by standard vision encoders and integrated into learning pipelines originally developed for vision, while preserving temporal and frequency information unavailable to conventional cameras. Rather than increasing spatial density through arrays of tactile elements, SpectRobot exploits the rich dynamics contained in sparse, high-bandwidth single-point measurements. In our implementation, the sensors are mounted away from the contact surface while remaining mechanically coupled to it, reducing direct exposure to wear and potentially improving robustness in harsh environments and for long-term deployment on dexterous robots. Our experiments demonstrate that: (1) a robot can exploit single-point vibration signals to solve a visually occluded manipulation task; (2) temporal history strongly influences policy performance, while sensing bandwidth controls the spectral information available, with measurements extending to 100~kHz; and (3) the same representation can be used across different tactile sensing technologies mediated by acceleration, force, or strain. We further show that capabilities previously associated with research-grade instrumentation can be accessed using readily available, off-the-shelf hardware. We believe that broader access to high-bandwidth tactile sensing could facilitate the integration of contact dynamics into embodied learning systems and, for some tasks, offer an alternative or complement to increasing the spatial density of tactile sensing.
Chinese Translation
触觉传感正越来越多地被纳入基于学习的机器人操作中，但许多现有方法依赖于空间分布的传感器。在此，我们介绍了SpectRobot，一个将单点触觉信号转化为紧凑时间-频率谱图的框架。这些谱图将高带宽的触觉历史编码为固定大小的图像类表示。它们可以被标准视觉编码器处理，并整合到最初为视觉开发的学习管道中，同时保留传统相机无法获取的时间和频率信息。SpectRobot并不是通过触觉元件阵列增加空间密度，而是利用稀疏的高带宽单点测量中包含的丰富动态。在我们的实现中，传感器安装在接触面之外，同时与其机械耦合，从而减少直接磨损暴露，并可能提高在恶劣环境下的鲁棒性以及在灵巧机器人上的长期部署。我们的实验表明：(1) 机器人可以利用单点振动信号解决视觉遮挡的操作任务；(2) 时间历史对策略性能有强烈影响，而传感带宽控制可用的频谱信息，测量频率可达100 kHz；(3) 相同的表示可以用于不同的触觉传感技术，这些技术通过加速度、力或应变进行介导。我们进一步展示了以前与研究级仪器相关的能力可以通过现成的、现货硬件获得。我们相信，更广泛地获取高带宽触觉传感可以促进接触动态在具身学习系统中的整合，并且对于某些任务，提供一种替代或补充的方法，而不是单纯增加触觉传感的空间密度。
cs.RO / 159 / 2609.24631
From Semantic Decisions to Feasible Trajectories: Self-Evolving LLM-Guided Optimal Control for Narrow-Space Parking
从语义决策到可行轨迹：自我演化的LLM引导的窄空间停车最优控制
Abstract
Autonomous parking in nonconvex and narrow environments remains challenging. Although optimal-control methods can explicitly enforce vehicle dynamics and collision constraints, nonconvexity compromises solver robustness and can cause failures. Large language models (LLMs) exhibit strong semantic reasoning capabilities, but directly generating dense trajectories makes it difficult to guarantee physical feasibility. We introduce SE-LLM-OCP, a unified framework in which LLMs make high-level discrete maneuver decisions, while an optimal-control module enforces low-level vehicle dynamics and collision constraints. Online, the LLM proposes sparse maneuver plans, decomposing the parking task into a sequence of short-horizon trajectory-optimization problems. A low-level solver then sequentially solves optimal-control problems. If the solver fails, the LLM aggregates failure evidence from the solver and validation stages to guide replanning. Offline, SE-LLM-OCP automatically evolves a structured decision-making knowledge base from scratch, driven by accumulated online failures. We validate our proposed framework in simulation on a car-like vehicle model and on a differential-drive robot. Our experimental results show that SE-LLM-OCP enables safer autonomous parking in narrow scenarios and demonstrates transfer of the same maneuver representation to a different kinematic platform.
Chinese Translation
在非凸和狭窄环境中的自主停车仍然具有挑战性。尽管最优控制方法可以明确执行车辆动力学和碰撞约束，但非凸性会影响求解器的鲁棒性并可能导致失败。大型语言模型（LLMs）表现出强大的语义推理能力，但直接生成密集轨迹使得保证物理可行性变得困难。我们提出了SE-LLM-OCP，这是一个统一框架，其中LLMs做出高层次的离散操控决策，而一个最优控制模块则执行低层次的车辆动力学和碰撞约束。在在线过程中，LLM提出稀疏的操控计划，将停车任务分解为一系列短时间范围的轨迹优化问题。然后，低层求解器依次解决最优控制问题。如果求解器失败，LLM会从求解器和验证阶段聚合失败证据，以指导重新规划。在离线过程中，SE-LLM-OCP自动从零开始演化出一个结构化的决策知识库，驱动因素是累积的在线失败。我们在模拟中验证了我们提出的框架，使用了类车模型和差分驱动机器人。实验结果表明，SE-LLM-OCP能够在狭窄场景中实现更安全的自主停车，并展示了相同操控表示在不同运动平台上的迁移能力。
cs.RO / 160 / 2609.24632
Feasibility Distance Fields for Heterogeneous Constraints in Robot Configuration Space
机器人配置空间中异质约束的可行性距离场
Abstract
Robot manipulators are monitored by constraint-specific indicators whose units and gradient scales are not comparable, so they do not provide a common measure of the configuration-space motion remaining before violation. We define the feasibility distance field (FDF) as the distance, under a fixed positive-definite joint-space metric, to the union of infeasible configuration sets. Classical distance-to-set theory gives 1-Lipschitz continuity, almost-everywhere differentiability, and unit dual-gradient norm wherever the nearest projection is unique. The robotics contribution is an admissibility analysis showing when practical constraints define non-empty closed sets. We derive admissible formulations for external and self-collision, joint limits, dexterity, Cartesian and task-projected compliance, joint torque under payload, and dynamic manipulability. Since every field uses the same metric, heterogeneous constraints compose by a pointwise minimum, conditioned constraints retain a fixed distance space, and multi-robot constraints produce block-sparse gradients that identify which robots must react. We generate projection-based labels and train neural approximations with a distance loss and an Eikonal penalty. Simulations on a UR5e and a dual-arm cell evaluate seven fields using value, projection, sign, gradient, composition, and moving-obstacle diagnostics. Across 8,000 configurations, the largest feasible-side secant ratio is 0.920, mean learned gradient norms range from 0.994 to 0.998, and projection residuals range from 0.011 to 0.034 rad. Across 24 random obstacle paths, the external and composed collision fields achieve 90.4% and 91.6% success within 3 cm, with sign-error rates below 2%. The results support a common configuration-space margin and identify approximation errors near medial axes and sparsely sampled boundaries.
Chinese Translation
机器人操纵器通过特定约束指标进行监控，这些指标的单位和梯度尺度不可比较，因此无法提供在违反之前剩余的配置空间运动的共同度量。我们定义可行性距离场（Feasibility Distance Field, FDF）为在固定的正定关节空间度量下，到不可行配置集的并集的距离。经典的距离到集合理论提供了1-Lipschitz连续性、几乎处处可微性，以及在最近投影唯一的地方单位对偶梯度范数。机器人学的贡献是可接受性分析，展示了何时实际约束定义非空闭集。我们推导了外部和自我碰撞、关节极限、灵活性、笛卡尔和任务投影合规性、负载下的关节扭矩以及动态可操控性的可接受性公式。由于每个场使用相同的度量，异质约束通过逐点最小值组合，条件约束保持固定的距离空间，多机器人约束产生块稀疏梯度，以识别哪些机器人必须反应。我们生成基于投影的标签，并使用距离损失和Eikonal惩罚训练神经近似。对UR5e和双臂单元的仿真评估了七个场，使用值、投影、符号、梯度、组合和移动障碍物诊断。在8000个配置中，最大的可行侧割线比为0.920，平均学习的梯度范数范围为0.994到0.998，投影残差范围为0.011到0.034弧度。在24条随机障碍物路径中，外部和组合碰撞场在3厘米内分别达到90.4%和91.6%的成功率，符号错误率低于2%。结果支持共同的配置空间边际，并识别接近中轴线和稀疏采样边界的近似误差。
cs.RO / 161 / 2609.24660
Touch2Robot: Robot Touch in the Human Demonstration Loop
Touch2Robot：人类演示循环中的机器人触觉
Abstract
Human demonstrations offer a scalable way to collect manipulation data, but their contacts may be unstable or infeasible when transferred to a robot hand. Collecting demonstrations directly on the target robot avoids this mismatch, but substantially increases the cost of data collection. To address this trade-off, we present \textbf{Touch2Robot}, a framework that lets humans collect demonstrations while seeing how the target robot hand would contact the object. We capture human hand motion, tactile-glove measurements, and object motion during human manipulation. These recordings guide object-specific RL policies to reproduce the demonstrated object motion while favoring contacts consistent with the recorded human touch. We distill the learned behaviors into a unified real-time retargeter that maps incoming human observations and object geometry to robot hand configurations. During collection, the predicted robot configuration is synchronized with the tracked object pose in simulation to reconstruct robot-object contacts, which are visualized to help the demonstrator adapt subsequent interactions to the target hand. Across four real-world tasks, Touch2Robot improves average real-robot replay completion from 37.9\% to 72.1\% over visual-only feedback, while reducing the collection time per replay-successful demonstration from 58.6~s to 18.2~s. Reconstructed target-hand contacts achieve 44.2\% F1 against real-robot tactile measurements, and policies trained on Touch2Robot demonstrations improve downstream Diffusion Policy performance by 29.1 percentage points over visual-only feedback. These results show that bringing robot touch into the human demonstration loop improves both the quality and efficiency of scalable dexterous data collection. \textit{Project webpage: \href{https://Touch2Robot.github.io/}{https://Touch2Robot.github.io/}.}
Chinese Translation
人类演示提供了一种可扩展的方式来收集操作数据，但当这些接触转移到机器人手上时，可能会出现不稳定或不可行的情况。直接在目标机器人上收集演示可以避免这种不匹配，但会显著增加数据收集的成本。为了解决这一权衡，我们提出了 extbf{Touch2Robot}，一个框架，允许人类在观察目标机器人手如何与物体接触的同时收集演示。我们在人的操作过程中捕捉人手运动、触觉手套测量和物体运动。这些记录指导特定物体的强化学习（RL）策略，以重现演示的物体运动，同时优先考虑与记录的人类触觉一致的接触。我们将学习到的行为提炼为一个统一的实时重定向器，将输入的人类观察和物体几何映射到机器人手的配置。在收集过程中，预测的机器人配置与模拟中跟踪的物体姿态同步，以重建机器人与物体的接触，并通过可视化帮助演示者调整后续与目标手的交互。在四个现实世界任务中，Touch2Robot 将真实机器人重放的平均完成率从 37.9\% 提高到 72.1\\%，同时将每次重放成功演示的收集时间从 58.6 秒减少到 18.2 秒。重建的目标手接触与真实机器人触觉测量的 F1 值达到 44.2\\%，而在 Touch2Robot 演示上训练的策略在下游扩散策略性能上比仅依赖视觉反馈提高了 29.1 个百分点。这些结果表明，将机器人触觉引入人类演示循环可以提高可扩展灵巧数据收集的质量和效率。 extit{项目网页： exttt{https://Touch2Robot.github.io/}}.
cs.RO / 162 / 2609.24682
Think Like a World Model, Act Like a VLA: Distilling World-Model Representations into Compact Robot Policies
像世界模型一样思考，像视觉-语言-动作（VLA）一样行动：将世界模型表示提炼为紧凑的机器人策略
Abstract
Vision-Language-Action (VLA) models map observations to actions with no objective that accounts for how the world responds, so their robustness is bounded primarily by data coverage. World models carry precisely that missing objective and are better grounded for it, yet rolling the future forward costs seconds per decision and rules them out of the control loop. We show the two can be separated. What a world model knows about physical scenes lives in its \emph{internal features}; generating the future is merely the objective that produced them, so the grounding can be inherited while the generative machinery is left behind. We add one feature-alignment term to ordinary VLA training: a frozen world model is run over the training frames once and cached, and the student learns to agree with that cache. No teacher is loaded during training, the projector is discarded after it, and the deployed policy is identical to the undistilled baseline, running in $32$~ms and $1.86$~GB on a consumer RTX~5090, so every gain is attributable to the representation rather than to added capacity or test-time compute. A $0.8$B student reaches $97.9\%$ on LIBERO, improves from $48.2\%$ to $50.5\%$ on RoboCasa-GR1 humanoid manipulation, and the same objective carries over to real hardware, on both a single-arm and a bimanual platform. The gain survives changes of student scale, backbone, alignment layer, and teacher, indicating a broad representational prior rather than a fragile alignment between two particular networks. Project page: https://thaw-vla.trung-dt.com/.
Chinese Translation
视觉-语言-动作（VLA）模型将观察映射到动作，但没有考虑世界如何响应的目标，因此它们的鲁棒性主要受限于数据覆盖。世界模型恰恰弥补了这一缺失的目标，并因此更为扎实，然而，向前滚动未来的过程每次决策需耗费数秒，使其无法纳入控制循环。我们展示了这两者可以分开。世界模型对物理场景的认识存在于其 extit{内部特征}中；生成未来仅仅是产生这些特征的目标，因此可以继承这一基础，而将生成机制抛在一旁。我们在普通的VLA训练中增加了一个特征对齐项：一个冻结的世界模型在训练帧上运行一次并缓存，学生学习与该缓存一致。在训练过程中不加载教师，投影器在训练后被丢弃，部署的策略与未提炼的基线相同，在消费级RTX~5090上运行时间为$32$~ms，内存占用为$1.86$~GB，因此每一项增益都归因于表示，而非增加的容量或测试时计算。一个$0.8$B的学生在LIBERO上达到了$97.9\%$，在RoboCasa-GR1人形操控中从$48.2\\%$提升至$50.5\\%$，同样的目标在真实硬件上也得以延续，适用于单臂和双手平台。增益在学生规模、骨干网络、对齐层和教师的变化中依然存在，表明这是一个广泛的表示先验，而非两个特定网络之间脆弱的对齐。项目页面：https://thaw-vla.trung-dt.com/
cs.RO / 163 / 2609.24702
AC-DC: Adaptive Communication for Scalable Dynamic Average Consensus in Multi-Robot Ergodic Search
AC-DC：用于多机器人遍历搜索的可扩展动态平均共识的自适应通信
Abstract
We study scalable peer-to-peer dynamic average consensus (DC) for multi-robot systems under finite-range, finite-rate, and interference-constrained communication. We introduce Adaptive Communication for Dynamic Average Consensus (AC-DC), which jointly adapts Who communicates with whom, When, and over What parts of the consensus state, using local inputs and successfully received neighbor information. Each robot's consensus state estimates the current average of the robots' local inputs. AC-DC updates these estimates as local inputs change and averages the values exchanged between robot pairs. In AC-DC, robot pairs update without waiting for every robot to complete a communication round, and the selected-state messages carry consensus state coordinates independent of team size for a fixed state representation. We apply AC-DC to dynamic-priority multi-robot ergodic search: one consensus stream estimates team visitation for motion coordination, while the other fuses regional measurement information to update uncertainty maps and search targets. Across twelve settings with up to 80 robots and 20 paired trials per setting, AC-DC has the lowest mean (i) normalized covariance-trace area under the curve (AUC) and (ii) attempted modeled communication payload among the compared decentralized methods. Averaged across settings, AC-DC achieves paired AUC reductions of 27.5% relative to state-of-the-art baselines, with 8.7x less communication traffic. As the number of robots increases, we observe that AC-DC's communication payload approaches that of the ideal centralized baseline (one ground compute-station communicating directly with all robots): with 120 robots in a fixed 600 x 600 m scaling test, AC-DC uses 19.3 MB versus 19.2 MB for the ideal centralized baseline, while remaining peer-to-peer.
Chinese Translation
我们研究了在有限范围、有限速率和干扰约束通信下的可扩展对等动态平均共识（DC）用于多机器人系统。我们引入了动态平均共识的自适应通信（AC-DC），它共同适应谁与谁通信、何时通信以及在共识状态的哪些部分进行通信，利用本地输入和成功接收的邻居信息。每个机器人的共识状态估计机器人的本地输入的当前平均值。AC-DC在本地输入变化时更新这些估计，并对机器人对之间交换的值进行平均。在AC-DC中，机器人对在不等待每个机器人完成通信轮次的情况下进行更新，所选状态消息携带与团队规模无关的共识状态坐标，以实现固定状态表示。我们将AC-DC应用于动态优先级多机器人遍历搜索：一个共识流估计团队的访问情况以进行运动协调，而另一个则融合区域测量信息以更新不确定性地图和搜索目标。在十二个设置中，最多可容纳80个机器人和每个设置20个配对试验，AC-DC在比较的去中心化方法中具有最低的均值（i）标准化协方差轨迹曲线下面积（AUC）和（ii）尝试的建模通信负载。平均来看，AC-DC相对于最先进的基线实现了27.5%的配对AUC减少，同时通信流量减少了8.7倍。随着机器人数量的增加，我们观察到AC-DC的通信负载接近理想的集中式基线（一个地面计算站直接与所有机器人通信）：在固定的600 x 600米缩放测试中，120个机器人时，AC-DC使用19.3 MB，而理想的集中式基线使用19.2 MB，同时保持对等通信。
cs.RO / 164 / 2609.24708
SPARSER: Sparse Variable Projection by Exploiting Separable Structure in Robotic Perception
SPARSER：通过利用可分离结构实现稀疏变量投影的机器人感知
Abstract
Robotic perception often requires solving large nonlinear least-squares (NLS) problems. While sparsity has been widely exploited to scale solvers, a complementary and underused structure is \emph{separability}: some variables, such as visual landmarks, appear linearly in the residuals and admit a closed-form solution once the remaining variables, such as poses, are fixed. Variable projection (VarPro) exploits this structure by analytically eliminating the linear variables, yielding a reduced problem with favorable computational properties. However, its use in robotic perception has been limited by gauge symmetries, such as invariance to global translations and rotations, which introduce challenges for standard VarPro methods. We present SPARSER (\textbf{S}parsity \textbf{P}reserving \textbf{A}nalytic \textbf{R}eduction for \textbf{S}eparable \textbf{R}obotic \textbf{P}erception), a VarPro framework for gauge-symmetric problems that jointly exploits separability and sparsity. Our method constructs a \emph{matrix-free Schur complement operator} for efficient evaluation of reduced costs, gradients, and Hessian-vector products, enabling integration with iterative NLS solvers. We characterize the applicable problem class, identify common cases admitting further analytical simplifications, and show that IRLS-based robust costs preserve most of the exploitable structure. Across synthetic and real SLAM, SNL, and SfM benchmarks, SPARSER is on average $5\times$--$7\times$ faster than state-of-the-art baselines on CPU and GPU, with gains exceeding $40\times$ on individual datasets. On outlier-corrupted multi-robot SLAM data, the robust variant is $2\times$--$16\times$ faster than a state-of-the-art GNC solver. We release open-source C++ code and all datasets.
Chinese Translation
机器人感知通常需要解决大型非线性最小二乘（NLS）问题。虽然稀疏性已被广泛利用以扩展求解器，但一个互补且未被充分利用的结构是 extit{可分离性}：一些变量，如视觉地标，在残差中呈线性出现，并且在固定其余变量（如姿态）后可以获得封闭形式的解。变量投影（VarPro）通过解析性地消除线性变量来利用这一结构，从而产生一个具有良好计算特性的简化问题。然而，由于测度对称性（如对全局平移和旋转的不变性），其在机器人感知中的应用受到限制，这给标准的VarPro方法带来了挑战。我们提出了SPARSER（ extbf{S}parsity extbf{P}reserving extbf{A}nalytic extbf{R}eduction for extbf{S}eparable extbf{R}obotic extbf{P}erception），这是一个针对测度对称问题的VarPro框架，能够共同利用可分离性和稀疏性。我们的方法构建了一个 extit{无矩阵的Schur补运算符}，以高效评估简化成本、梯度和Hessian-向量乘积，从而实现与迭代NLS求解器的集成。我们描述了适用的问题类别，识别了允许进一步解析简化的常见情况，并表明基于IRLS的鲁棒成本保留了大部分可利用的结构。在合成和真实的SLAM、SNL和SfM基准测试中，SPARSER在CPU和GPU上平均比最先进的基线快$5 imes$--$7 imes$，在个别数据集上的增益超过$40 imes$。在受离群值影响的多机器人SLAM数据中，鲁棒变体比最先进的GNC求解器快$2 imes$--$16 imes$。我们发布了开源C++代码和所有数据集。
cs.RO / 165 / 2609.24742
LLM-based Conversational AI Knowledge Assistant for MyBuddy Humanoid Robot
基于大型语言模型的MyBuddy类人机器人对话式人工智能知识助手
Abstract
Humanoid robots are increasingly being popular and developed for human-centered applications, yet their ability to provide intelligent conversations and natural interactive knowledge assistance remains constrained by traditional rule-based dialogue systems, pre-defined responses and limited knowledge repositories. Large language models (LLMs) have emerged as a powerful foundation for enabling natural, adaptive, and context-aware Human-Robot Interaction (HRI), which provides a significant opportunity to address such limitations by enabling robots to understand natural speech language, reason over complicated queries, maintain high-quality conversational context, and generate knowledge-rich responses. In this work, we originally present and implement an LLM-based versatile Conversational AI Knowledge Assistant for the Raspberry-Pi-powered 13-Axis MyBuddy humanoid robot, which integrates LLM-driven language understanding and AI reasoning with real-time speech recognition, knowledge retrieval via extensible access of internet engines (e.g., Wikipedia, arXiv), flexible dialogue management, and natural speech synthesis to enable much more intelligent multi-turn continuous conversations and advanced emotional-support Human-Robot Interaction.
Chinese Translation
类人机器人在以人为中心的应用中越来越受欢迎并不断发展，但它们在提供智能对话和自然互动知识辅助方面的能力仍受到传统基于规则的对话系统、预定义响应和有限知识库的限制。大型语言模型（LLMs）作为一种强大的基础，能够实现自然、适应性强且具有上下文感知的人机交互（HRI），为解决这些限制提供了重要机会，使机器人能够理解自然语言、对复杂查询进行推理、保持高质量的对话上下文，并生成丰富知识的响应。在本研究中，我们首次提出并实现了一种基于LLM的多功能对话式人工智能知识助手，适用于由树莓派驱动的13轴MyBuddy类人机器人，该助手将LLM驱动的语言理解和人工智能推理与实时语音识别、通过可扩展的互联网引擎（如维基百科、arXiv）进行的知识检索、灵活的对话管理和自然语音合成相结合，从而实现更智能的多轮连续对话和高级情感支持的人机交互。
cs.RO / 166 / 2609.24745
Beyond Visual Quality: A Study of Test-Time Planning with World Action Models
超越视觉质量：基于世界行动模型的测试时规划研究
Abstract
World action models generate actions together with visual predictions of their consequences. These paired outputs create the potential for planning by sampling multiple actions from one state, comparing their imagined outcomes, and choosing the action with the most promising predicted outcome. However, how to use imagined futures to guide action selection remains unclear. We examine this planning potential empirically. First, we estimate an oracle upper bound on selection by choosing the sampled candidate whose realised outcome is best. In a controlled same-state analysis, this choice raises success from 68.9% under uniform random selection to 79.2%. We then test selectors based on visual quality, physical consistency, and task progression as controlled interventions. Some tested selectors yield higher observed success, but the gains are uneven and the matched selectors leave much of the measured opportunity unrecovered. To investigate this gap, we examine whether sampled actions lead to different outcomes, whether these differences are visible in the predictions, and whether a score recognises them. Counterfactual branching from the same states shows that selection opportunity is concentrated in relatively few decisions in the initial candidate sets. Action spread and outcome coverage need not increase together. In a further evaluation across trajectory phases with complete action execution, the tested scores again recover little of the available improvement despite a small gain from learned value. These findings distinguish producing consequential action choices from recognising them in generated futures, motivating the evaluation of WAM predictions through their usefulness for decisions rather than visual quality alone.
Chinese Translation
世界行动模型生成与其后果的视觉预测相结合的行动。这些配对输出为通过从一个状态中抽样多个行动、比较其想象结果并选择最有前景的预测结果的行动提供了规划的潜力。然而，如何利用想象的未来来指导行动选择仍不明确。我们通过实证研究这一规划潜力。首先，我们通过选择实现结果最佳的抽样候选者来估计选择的上限。在受控的同状态分析中，这一选择将成功率从68.9%（在均匀随机选择下）提高到79.2%。然后，我们测试基于视觉质量、物理一致性和任务进展的选择器作为受控干预。一些测试的选择器产生了更高的观察成功率，但增益不均，匹配的选择器未能充分利用测量机会。为了调查这一差距，我们考察了抽样行动是否导致不同的结果，这些差异是否在预测中可见，以及评分是否能够识别这些差异。从同一状态的反事实分支显示，选择机会集中在初始候选集中的相对少数决策中。行动的分布和结果的覆盖不必同时增加。在对完整行动执行的轨迹阶段进行进一步评估时，尽管从学习的价值中获得了小幅增益，但测试的评分再次未能恢复可用改进的多少。这些发现将产生有意义的行动选择与在生成的未来中识别这些选择区分开来，从而激励通过决策的实用性而非仅仅视觉质量来评估世界行动模型（WAM）预测。
cs.RO / 167 / 2609.24749
D-JEPA: A Decision-Aligned Latent World Model
D-JEPA：一个与决策对齐的潜在世界模型
Abstract
Latent world models predict the consequences of actions, but accurate prediction does not guarantee that latent distance reflects which candidate will execute successfully. We identify a decision-local prediction gap: among the few futures competing for execution, a candidate predicted closer to the goal can produce a worse realized outcome than an available alternative. We introduce D-JEPA, a decision-aligned latent world model that learns decision-relevant relations among candidate futures from executed outcomes. A bounded, permutation-equivariant operator jointly reasons over goal-relative predictive features and ordinal evidence, refining pretrained predictive geometry where action choices are most consequential. Restricted predictor adaptation and a shared ordinal interface extend this alignment across complementary predictive geometries. D-JEPA further realizes the learned decision structure in JEPA-compatible future representations, enabling deployment through native latent-distance planning. Evaluations across latent control, manipulation, pretrained action-producing models, physical robots and autonomous driving demonstrate improved action selection, including 87.89% success on PushT, a 15.04-point average gain on RoboTwin, and a 17-point gain on physical robot tasks. These results establish decision-relevant relational structure as a direct bridge between predictive world modeling and effective control.
Chinese Translation
潜在世界模型预测行动的后果，但准确的预测并不保证潜在距离反映出哪个候选者能够成功执行。我们识别出一个决策局部预测差距：在为执行竞争的少数未来中，预测更接近目标的候选者可能产生比可用替代方案更糟的实际结果。我们引入了D-JEPA，一个与决策对齐的潜在世界模型，它从执行结果中学习候选未来之间与决策相关的关系。一个有界的、排列不变的算子共同推理目标相关的预测特征和序数证据，精炼预训练的预测几何，特别是在行动选择最具影响力的地方。受限的预测器适应和共享的序数接口将这种对齐扩展到互补的预测几何中。D-JEPA进一步在与JEPA兼容的未来表示中实现学习到的决策结构，从而通过原生潜在距离规划进行部署。在潜在控制、操控、预训练的行动生成模型、物理机器人和自主驾驶等方面的评估显示出改进的行动选择，包括在PushT上87.89%的成功率、在RoboTwin上15.04点的平均增益，以及在物理机器人任务上17点的增益。这些结果确立了与决策相关的关系结构作为预测世界建模与有效控制之间的直接桥梁。
cs.RO / 168 / 2609.24761
A Switched Adaptive Control Framework for Aerial Manipulators Under Dynamic Transitions
一种用于动态过渡下空中操控器的切换自适应控制框架
Abstract
Aerial manipulators represent the forefront of aerial robotics. Although potentially capable of complex interaction tasks, controlling aerial manipulators throughout the dynamic transitions occurring during task execution presents significant challenges. Abrupt or discontinuous changes in system dynamics generated by the transitions suggest the use of a switched approach, yet the available aerial manipulation methods are not designed for coping with switched regimes. In addition, most available methods fall short in coping with the tight couplings between the aerial vehicle and the manipulator, as well as in coping with the state-dependent uncertainties arising from the difficulty in modeling such couplings. We propose a switched-based adaptive control framework for aerial manipulators not relying on a priori knowledge of the vehicle-manipulator couplings and of state-dependent uncertainties. To guarantee stable manipulation despite changes in system dynamics, the framework provides a class of switching signals characterizing those transition phases for which the system is guaranteed to remain stable. Comparative experiments further validate the effectiveness of the proposed switched-based framework over the state of the art.
Chinese Translation
空中操控器代表了空中机器人技术的前沿。尽管它们在复杂交互任务中具有潜在能力，但在任务执行过程中控制空中操控器所经历的动态过渡仍然面临重大挑战。由于过渡产生的系统动态的突变或不连续变化，建议采用切换方法，然而现有的空中操控方法并未针对切换状态进行设计。此外，大多数现有方法在处理空中飞行器与操控器之间的紧密耦合以及因建模此类耦合的困难而产生的状态依赖不确定性方面也显得不足。我们提出了一种基于切换的自适应控制框架，用于空中操控器，该框架不依赖于对飞行器-操控器耦合及状态依赖不确定性的先验知识。为了确保在系统动态变化的情况下仍能稳定操控，该框架提供了一类切换信号，以表征那些系统保证保持稳定的过渡阶段。比较实验进一步验证了所提切换基础框架相较于现有技术的有效性。
cs.RO / 169 / 2609.24778
H2RBench: A Real-to-Sim Benchmark for Evaluating Human-to-Robot Transfer
H2RBench：评估人机转移的真实到模拟基准
Abstract
Learning robot manipulation policies from human video demonstrations constitutes a promising avenue for scalable robot learning. However, comparing different human-to-robot (H2R) transfer methods remains challenging, as existing approaches are evaluated under different settings, including differing task suites, scene layouts, object instances, and amounts of robot supervision. To address this challenge, we present H2RBench, a Real2Sim benchmark for evaluating H2R transfer methods. H2RBench provides a standardized protocol built on real human video demonstrations and simulated robot demonstrations, and includes four manipulation tasks spanning diverse interaction requirements. We evaluate multiple representative H2R transfer methods, each adopting a different strategy for bridging the embodiment gap. Using H2RBench, we systematically characterize how each method scales with the amount of human demonstrations, revealing that methods differ substantially in their ability to leverage additional human data. We further show that simulation performance is broadly predictive of real-world robot performance, with an overall Pearson correlation of r = 0.89, Spearman correlation of \r{ho} = 0.85 and Mean Maximum Rank Violation (MMRV) of 0.06 across method-task configurations. These results establish H2RBench as a practical and scalable benchmark for comparative H2R evaluation prior to real-world deployment.
Chinese Translation
从人类视频演示中学习机器人操作策略是可扩展机器人学习的一个有前景的途径。然而，比较不同的人机转移（H2R）方法仍然具有挑战性，因为现有的方法在不同的设置下进行评估，包括不同的任务套件、场景布局、物体实例和机器人监督的数量。为了解决这一挑战，我们提出了H2RBench，一个用于评估H2R转移方法的Real2Sim基准。H2RBench提供了一个基于真实人类视频演示和模拟机器人演示的标准化协议，并包括四个涵盖多样化交互需求的操作任务。我们评估了多种具有代表性的人机转移方法，每种方法采用不同的策略来弥合体现差距。利用H2RBench，我们系统地描述了每种方法在利用人类演示数量方面的扩展性，揭示了这些方法在利用额外人类数据的能力上存在显著差异。我们进一步展示了模拟性能在很大程度上可以预测现实世界中的机器人性能，整体皮尔逊相关系数为r = 0.89，斯皮尔曼相关系数为
{ho} = 0.85，方法-任务配置的平均最大排名违反（MMRV）为0.06。这些结果确立了H2RBench作为一个实用且可扩展的基准，用于在实际部署之前进行比较性的人机转移评估。
cs.RO / 170 / 2609.24815
Uranus: Building the Next-Generation Simulation Infrastructure for Embodied AI
天王星：构建下一代具身人工智能模拟基础设施
Abstract
Scalable simulation is essential for robot data generation, policy training, evaluation, and safe iteration, yet real-world interaction is costly and conventional simulators require labor-intensive construction. We present Uranus, a data-driven robot simulator built around a joint-trajectory-conditioned autoregressive diffusion model. Uranus offers three key capabilities: (1) streaming, open-ended rollout, which receives future joint-position trajectories online and autoregressively generates one latent frame per step, corresponding to four RGB frames, without a fixed horizon; (2) low-latency generation, achieving 24 FPS after inference optimization; and (3) scalable, extensible robot control, providing a unified interface for synchronized multi-view generation across diverse robot embodiments and camera configurations. We conduct comprehensive quantitative and qualitative evaluations on both in-distribution and out-of-distribution data, providing an objective assessment of Uranus and clearly identifying its current limitations. We release the code and model weights to empower the community with practical tools and insights.
Chinese Translation
可扩展的模拟对于机器人数据生成、策略训练、评估和安全迭代至关重要，但现实世界的交互成本高昂，传统模拟器则需要耗费大量人力进行构建。我们提出了天王星（Uranus），这是一种基于联合轨迹条件自回归扩散模型的数据驱动机器人模拟器。天王星提供了三个关键功能：（1）流式、开放式的展开，能够在线接收未来的关节位置轨迹，并自回归地每一步生成一个潜在帧，对应四个RGB帧，而不设定固定的时间范围；（2）低延迟生成，在推理优化后实现24帧每秒（FPS）；（3）可扩展、可扩展的机器人控制，提供一个统一接口，用于跨多种机器人形态和相机配置的同步多视图生成。我们对在分布内和分布外的数据进行了全面的定量和定性评估，客观评估了天王星，并清晰识别其当前的局限性。我们发布了代码和模型权重，以赋能社区，提供实用工具和见解。
cs.RO / 171 / 2609.24832
Minimum Time Trajectories for a Car-Like Mobile Robot Moving with Rigid Wheels Under Non-Sliding Constraints
在非滑动约束下，具有刚性轮子的类车移动机器人最短时间轨迹
Abstract
This paper studies the minimum time trajectoriesvof a car-like mobile robot navigating in an obstacle free environment. The robot, with forward and backward speeds, is controlled by bounded front-wheels acceleration and limited front-wheels steering rate. The paper extends previous results which solved this problem for the kinematic car-like robot. However, the kinematic model assumes pure rolling at the wheels ground contacts. This assumption requires non-sliding constraints for the front and rear wheels that can only be handled by the robot dynamics. This paper formulates the non-sliding constraints based on the robot dynamics then augments the kinematic model time-optimal path primitives with three new path primitives associated with the non-sliding constraints. The three non-sliding path primitives together with the kinematic model twelve path primitives form the car-like robot time optimal trajectories. Approximate analytic solutions for the non-sliding path primitives are also provided. Examples study the time-optimal path primitives along representative maneuvers, illustrating how the non-sliding constraints influence the time optimal trajectories of the car-like robot.
Chinese Translation
本文研究了一种类车移动机器人在无障碍环境中导航的最短时间轨迹。该机器人具有前进和后退速度，由有限的前轮加速度和受限的前轮转向速率控制。本文扩展了之前针对运动学类车机器人解决该问题的结果。然而，运动学模型假设轮子与地面接触时为纯滚动。这一假设要求前后轮的非滑动约束只能通过机器人动力学来处理。本文基于机器人动力学公式化了非滑动约束，然后将运动学模型的时间最优路径原语与三个与非滑动约束相关的新路径原语相结合。三个非滑动路径原语与运动学模型的十二个路径原语共同构成了类车机器人的时间最优轨迹。本文还提供了非滑动路径原语的近似解析解。通过示例研究了代表性机动中的时间最优路径原语，说明了非滑动约束如何影响类车机器人的时间最优轨迹。
cs.RO / 172 / 2609.24840
PredActor: Predictive Action Diffusion for Steerable Onboard Humanoid Control
PredActor：用于可控人形机器人控制的预测性动作扩散
Abstract
Diffusion models offer flexible motion generation, but translating this flexibility into feedback-responsive humanoid control remains challenging. Hierarchical systems steer motion through references that may exceed a separate tracker's capabilities, leaving recovery and physical execution largely to the tracker. Action-only diffusion generates actions directly but lacks an explicit future-state trajectory for test-time motion objectives. Joint state-action diffusion provides this representation, yet representative controllers often depend on privileged full-body states, and support for learned behavior selection and test-time motion steering remains fragmented. We present PredActor, a predictive action diffusion policy that brings these complementary steering capabilities into one directly executed policy using proprioceptive observations. Conditioned on proprioceptive history and optional task context, PredActor jointly generates executable actions and an internal future-state trajectory. Classifier-free guidance strengthens text-conditioned behavior, while classifier guidance steers predicted states toward test-time objectives. Only actions are executed, without a separate motion-reference tracker or externally estimated full-body states as policy inputs. In simulation, PredActor reaches all 15 destination targets and achieves a text retrieval score of 0.580, compared with 0.373 for conditional action diffusion, with similar observed disturbance survival. To make this guided policy practical onboard, rolling denoising and computation-preserving runtime optimizations reduce the complete callback to 16.790 ms median and 19.383 ms p95 on a Jetson Orin NX, both below the 20 ms control period. We deploy PredActor on a Unitree G1; evaluations across simulation and physical hardware demonstrate text-conditioned motion, disturbance response, joystick control, and semantic interpolation.
Chinese Translation
扩散模型提供了灵活的运动生成，但将这种灵活性转化为反馈响应的人形机器人控制仍然具有挑战性。分层系统通过超出单独跟踪器能力的参考来引导运动，使得恢复和物理执行在很大程度上依赖于跟踪器。仅动作扩散直接生成动作，但缺乏明确的未来状态轨迹以满足测试时的运动目标。联合状态-动作扩散提供了这种表示，然而，代表性控制器通常依赖于特权的全身状态，并且对学习行为选择和测试时运动引导的支持仍然零散。我们提出了PredActor，这是一种预测性动作扩散策略，将这些互补的引导能力整合为一个直接执行的策略，使用本体感知观察。PredActor基于本体感知历史和可选的任务上下文共同生成可执行的动作和内部未来状态轨迹。无分类器引导增强了文本条件行为，而有分类器引导则将预测状态引导至测试时目标。仅执行动作，而不需要单独的运动参考跟踪器或外部估计的全身状态作为策略输入。在仿真中，PredActor成功到达所有15个目标，并实现了0.580的文本检索分数，而条件动作扩散的分数为0.373，且观察到的干扰生存情况相似。为了使这一引导策略在机载上实用，滚动去噪和计算保留的运行时优化将完整回调时间减少到16.790毫秒的中位数和19.383毫秒的95百分位数，均低于20毫秒的控制周期。我们在Unitree G1上部署了PredActor；在仿真和物理硬件上的评估展示了文本条件运动、干扰响应、操纵杆控制和语义插值。
cs.RO / 173 / 2609.24841
CAST: Collision-Aware Assembly with Construction Robots using Simultaneous Trajectory Estimation and Planning
CAST：基于同时轨迹估计与规划的碰撞感知装配机器人系统
Abstract
Multi-robot systems have shown increasing viability in construction due to their ability to execute high-precision actions while reducing human exposure to hazardous tasks. However, these environments have high-dimensional configuration spaces and possess substantial collision-avoidance constraints, which include other robots, assembly objects, and workspace boundaries. We utilize a single factor graph for trajectory estimation and planning that incorporates measured robot states together with explicit collision and learned cable constraints. This supports changing workspaces and enables synchronized, high-dimensional robot motion planning while accounting for the stiff, vibration-induced uncertainty of heavy robotic systems. We demonstrate the success of our framework on the construction of a post-and-lintel structure using one robot arm as a timber gripper, and a second robot as a nail-fastener.
Chinese Translation
多机器人系统在建筑领域的可行性日益增强，因为它们能够执行高精度操作，同时减少人类暴露于危险任务的风险。然而，这些环境具有高维配置空间，并且存在大量的避碰约束，包括其他机器人、装配物体和工作空间边界。我们利用单一因子图进行轨迹估计与规划，该图结合了测量的机器人状态以及明确的碰撞和学习到的电缆约束。这支持了变化的工作空间，并使得在考虑重型机器人系统的刚性和振动引起的不确定性的情况下，实现同步的高维机器人运动规划。我们在使用一个机器人手臂作为木材抓取器，另一个机器人作为钉子固定器的柱梁结构建造中展示了我们框架的成功。
cs.RO / 174 / 2609.24846
Range-Aided SLAM Initialization Exploiting Accurate Heading Information
利用精确航向信息的范围辅助SLAM初始化
Abstract
This paper presents a novel initialization method for range-aided simultaneous localization and mapping (RA-SLAM). The general SLAM problem has a well-known separable structure where landmark and robot positions can be solved for in a linear fashion given known robot headings. This paper considers the case where highly accurate heading information is available, which is typical of autonomous underwater vehicle (AUV) navigation, to solve for the range transponder and robot positions. The proposed approach consists of two steps. First, using a generalized trust region subproblem (GTRS), the positions of the transponders relative to the AUV are solved for given the nonlinear range measurements. Second, the relative transponder positions and the known heading of the AUV are used to estimate the transponder and AUV positions by solving a linear least-squares problem. These transponder and AUV position estimates, combined with the highly accurate heading information, provide a reliable initialization method for the general nonlinear RA-SLAM problem. The effectiveness of the proposed approach is tested on a real-world AUV dataset where long baseline (LBL) range measurements are provided in concert with highly accurate heading information provided by an inertial navigation system (INS).
Chinese Translation
本文提出了一种新的范围辅助同时定位与地图构建（RA-SLAM）初始化方法。一般的SLAM问题具有众所周知的可分离结构，在已知机器人航向的情况下，地标和机器人位置可以以线性方式求解。本文考虑在高度精确的航向信息可用的情况下进行求解，这在自主水下航行器（AUV）导航中是典型的，以求解范围应答器和机器人位置。所提出的方法由两个步骤组成。首先，利用广义信任区域子问题（GTRS），在给定非线性范围测量的情况下，求解应答器相对于AUV的位置。其次，利用相对应答器位置和已知的AUV航向，通过求解线性最小二乘问题来估计应答器和AUV的位置。这些应答器和AUV位置的估计，结合高度精确的航向信息，为一般非线性RA-SLAM问题提供了一种可靠的初始化方法。所提出方法的有效性在一个真实世界的AUV数据集中得到了验证，该数据集提供了与由惯性导航系统（INS）提供的高度精确航向信息相结合的长基线（LBL）范围测量。
cs.RO / 175 / 2609.24864
SE(3) Neural Potential Fields for 6-DoF Trajectory Planning Directly from Images Without Explicit 3D Reconstruction
基于SE(3)神经势场的6自由度轨迹规划：直接从图像中获取，无需显式的3D重建
Abstract
Reaching a 6-DoF grasp pose in clutter requires a collision-free trajectory, conventionally obtained by reconstructing the scene in 3D and planning inside that reconstruction, at the cost of its accuracy and compute. Potential fields learned directly from images remove that dependency but inherit the classical weakness of artificial potential fields: where attractive and repulsive gradients cancel, the descent grazes the obstacle instead of going around it, and can stall short of the goal. We present an SE(3) neural potential field learned from posed RGB images and supervised with a navigation function, the geodesic distance to the grasp through free space recovered from those same images during training, which removes both failures. On two tabletop scenes, from obstacle-blocked starts executed on a UR10, the field converges within 3 cm of the grasp from every start and every path it executes is collision-free against the ground-truth geometry, against 25% and 0% under image supervision alone; mean clearance rises from under a centimeter to 8.6-8.8 cm and arm-link contacts fall from 20.6-50.4% to 2.7-5.5% of executed configurations. Executed grasp success is 90.0% and 40.0% on the two scenes, the residual failures being refusals of the Cartesian executor rather than of the field. Planning takes about 2 s against 67-133 s for RRT* on a reconstruction of the same images, though under a common offline harness the two are comparable: the deployed margin is the cost of collision-checking a dense reconstruction, not planner complexity.
Chinese Translation
在杂乱环境中达到6自由度抓取姿态需要一条无碰撞的轨迹，传统方法是通过重建场景的3D模型并在该重建中进行规划，这会影响准确性和计算效率。直接从图像学习的势场消除了这种依赖，但继承了人工势场的经典弱点：在吸引和排斥梯度相互抵消的地方，下降路径会擦碰障碍物而不是绕过它，可能会在目标前停滞。我们提出了一种基于SE(3)的神经势场，该势场从已标定的RGB图像中学习，并通过导航函数进行监督，该函数是在训练过程中从这些图像中恢复的自由空间到抓取的测地距离，从而消除了这两种失败。在两个桌面场景中，从UR10执行的障碍物阻挡起始点出发，该势场在每个起始点的抓取位置内收敛至3厘米以内，且执行的每条路径在与真实几何体的对比中均无碰撞，而在仅依赖图像监督时则分别为25%和0%；平均间隙从不足1厘米提升至8.6-8.8厘米，臂链接触率从20.6-50.4%降至2.7-5.5%。在这两个场景中，执行的抓取成功率分别为90.0%和40.0%，剩余的失败主要是由于笛卡尔执行器的拒绝，而非势场的原因。规划时间约为2秒，而在相同图像重建上使用RRT*则需67-133秒，尽管在共同的离线环境下，两者的性能相当：部署的边际是进行密集重建的碰撞检测成本，而非规划复杂性。
cs.RO / 176 / 2609.24868
DualWAM: Dual-System World Action Models for Asynchronous Global Planning and Local Refinement
双系统世界动作模型（DualWAM）：用于异步全局规划和局部细化的双系统世界动作模型
Abstract
World Action Models (WAMs) jointly generate robot actions and predict future world states, transferring priors from video pretraining to robot control. However, future visual prediction is computationally expensive, so existing WAMs often rely on long action chunks to amortize inference cost across control steps, at the cost of closed-loop responsiveness. We present \method, a dual-system WAM that preserves broader-horizon world-action generation while enabling high-frequency closed-loop action updates by decoupling global planning and local refinement. \systwo periodically performs high-noise bidirectional denoising over a broader world-action chunk to establish a global plan, while wrist-only \sysone extracts a temporally aligned short window from the intermediate denoising state and completes low-noise refinement using the latest wrist observations, which provide action-aligned cues about local geometry, motion, and contact during interaction. The two systems operate asynchronously along a shared denoising trajectory: each global plan is reused across multiple local updates, while \sysone repeatedly incorporates fresh interaction feedback. Across zero-shot manipulation tasks on Franka and Galbot, \method improves success over the strongest evaluated baseline by 4.5 percentage points on average, while achieving a 16.6$\times$ critical-path speedup. Further studies show that role-matched egocentric and UMI data improve success by 14 percentage points, and that the decoupled design naturally supports edge--cloud deployment with substantially lower communication overhead than the baseline.
Chinese Translation
世界动作模型（World Action Models, WAMs）共同生成机器人动作并预测未来世界状态，将视频预训练中的先验知识转移到机器人控制中。然而，未来视觉预测的计算成本较高，因此现有的WAMs通常依赖于长动作块，以在控制步骤中摊销推理成本，这牺牲了闭环响应能力。我们提出了 extit{DualWAM}，一种双系统WAM，它在保持更广泛视野的世界动作生成的同时，通过解耦全局规划和局部细化，实现高频闭环动作更新。 extit{WAM2}定期对更广泛的世界动作块进行高噪声双向去噪，以建立全局计划，而仅使用手腕的 extit{WAM1}则从中间去噪状态中提取时间对齐的短窗口，并利用最新的手腕观测完成低噪声的细化，这些观测提供了关于局部几何、运动和交互过程中的接触的动作对齐线索。这两个系统沿着共享的去噪轨迹异步操作：每个全局计划在多个局部更新中重复使用，而 extit{WAM1}则不断纳入新的交互反馈。在Franka和Galbot的零-shot操作任务中， extit{DualWAM}在评估的最强基线之上平均提高了4.5个百分点的成功率，同时实现了16.6$ imes$的关键路径加速。进一步研究表明，角色匹配的自我中心和UMI数据将成功率提高了14个百分点，并且解耦设计自然支持边缘-云部署，通信开销显著低于基线。
cs.RO / 177 / 2609.24896
Steerable and Reactive Grasping Through Modular Design with a Three-Point Interface
通过三点接口的模块化设计实现可控和反应式抓取
Abstract
Dexterous grasping requires deciding where to grasp, reaching the target, and maintaining stable contact. We connect these stages through a compact three-point interface that separates global geometric reasoning from local contact control. Given object geometry and optional language commands, our framework samples contact triples from a precomputed grasp-affordance heatmap. A model-based reactive controller tracks the object, avoids collisions, and guides the hand toward the selected contacts. In the final centimeters, a Reinforcement Learning (RL) policy uses proprioceptive feedback to refine and stabilize the grasp despite reaching and perception errors. It observes only finger joint states and its recent actions, with no target points, visual observations, or object geometry, so a single policy is shared across objects and grasp configurations. In simulation, we compare grasp-and-lift success against squeeze and end-to-end baselines, characterize reaching convergence, and demonstrate grasp steering; hardware demonstrations on two training objects and one unseen object illustrate the full pipeline. Our modular framework uses geometry to guide the reach and local feedback to secure the grasp.
Chinese Translation
灵巧抓取需要决定抓取位置、到达目标并保持稳定接触。我们通过一个紧凑的三点接口将这些阶段连接起来，该接口将全局几何推理与局部接触控制分离。基于物体几何形状和可选的语言指令，我们的框架从预计算的抓取适应热图中采样接触三元组。基于模型的反应式控制器跟踪物体，避免碰撞，并引导手部朝向选定的接触点。在最后几厘米的过程中，强化学习（Reinforcement Learning, RL）策略利用本体感知反馈来细化和稳定抓取，尽管存在到达和感知误差。该策略仅观察手指关节状态及其近期动作，而不依赖于目标点、视觉观察或物体几何，因此一个策略可以在不同物体和抓取配置之间共享。在仿真中，我们比较了抓取和提升的成功率与挤压和端到端基线，表征了到达收敛性，并展示了抓取引导；在两个训练物体和一个未见物体上的硬件演示展示了整个流程。我们的模块化框架利用几何形状指导抓取，并通过局部反馈确保抓取的安全性。
cs.RO / 178 / 2609.24906
Visuomotor Robotic Pruning in Planar Orchards Using Hybrid Reinforcement Learning
基于混合强化学习的平面果园视觉运动机器人修剪
Abstract
Dormant tree pruning is labor-intensive yet essential for maintaining modern high-productivity fruit orchards. In this work, we focus on pruning of modern planar tree training systems - V-Trellis apples and UFO cherries - where trunks and primary branches are trained into approximately planar walls. We introduce an end-to-end pipeline to learn a closed-loop visuomotor controller for robotic pruning. This controller is trained entirely using simulation and synthetically generated data and deployed in real orchards in a zero-shot manner. The pipeline comprises synthetic generation of planar orchard tree meshes, construction of a physics-based orchard simulator, automated collection of successful pruning trajectories via motion planning, and policy learning with a novel hybrid reinforcement-learning algorithm that combines offline demonstrations with online simulated rollouts. The controller uses optical-flow inputs from a wrist-mounted camera - avoiding the need for full 3D-reconstruction - and continuously guides the cutter through cluttered branch environments to a specified cutpoint with correct tool orientation. In exhaustive simulated task-space evaluations over 3,000 pruning points, the policy attains 49.9% success on V-Trellis apples and 46.0% on UFO cherries. We validate the learned controller across 38 physical trials - comprising 28 outdoor field trials in commercial and experimental orchards and 10 indoor laboratory tests - demonstrating zero-shot sim-to-real transfer. The learned policy also outperforms a classical RRT-Connect baseline on physical hardware in laboratory trials.
Chinese Translation
冬季树木修剪劳动强度大，但对于维持现代高产果园至关重要。在本研究中，我们专注于现代平面树木训练系统的修剪——V-Trellis 苹果和 UFO 樱桃，其中树干和主要枝条被训练成大致平面的墙面。我们介绍了一种端到端的流程，用于学习闭环视觉运动控制器以实现机器人修剪。该控制器完全通过仿真和合成生成的数据进行训练，并以零样本的方式在真实果园中部署。该流程包括平面果园树木网格的合成生成、基于物理的果园模拟器构建、通过运动规划自动收集成功的修剪轨迹，以及使用一种新颖的混合强化学习算法进行策略学习，该算法结合了离线演示和在线模拟回放。控制器使用来自手腕安装相机的光流输入——避免了全3D重建的需求——并持续引导切割工具在杂乱的枝条环境中到达指定的切割点，并保持正确的工具方向。在对3000个修剪点进行的全面模拟任务空间评估中，该策略在 V-Trellis 苹果上取得了49.9%的成功率，在 UFO 樱桃上取得了46.0%的成功率。我们在38次物理试验中验证了学习到的控制器——包括28次在商业和实验果园的户外现场试验以及10次室内实验室测试——展示了零样本的模拟到现实转移。学习到的策略在实验室试验中也优于经典的 RRT-Connect 基线。
cs.RO / 179 / 2609.24952
Learning to Drive on Mars: Visual Multimodal Traversability Estimation for Off-World Navigation
在火星上学习驾驶：用于外星导航的视觉多模态可通行性估计
Abstract
Autonomous navigation on Mars requires vehicles to distinguish between traversable terrains across diverse and visually challenging environments. However, progress in learning-based navigation for off-world environments has been limited by the lack of large-scale datasets. Since landing in Jezero Crater, the Mars 2020 Perseverance rover has traversed terrain ranging from sandy dunes, rocky patches, and flat bedrocks. As a result, this paper presents a dataset spanning 500 sols and 45km of trajectories driven by both human operators and the onboard planner, ENav. Our dataset contains grayscale stereo image pairs, poses, accelerometer readings, rocker-bogie angles, and estimates of tilt and wheel slip. Building on this dataset, we introduce an uncertainty-aware traversability-estimation framework that learns terrain representations from multimodal driving experience. We compare our proposed method against existing approaches on the Mars 2020 dataset and show that our method achieves an AUROC of 0.874 and an F1 score of 0.758, outperforming the strongest baseline by 0.058 and 0.156, respectively, while also achieving the highest average precision and recall. Finally, we show that the visual representations can be integrated into path planners, such as ENav, on a physical rover test bed. Videos, code, and the M2020 dataset will be available at https://darren-chiu.github.io/learning-to-drive-on-mars.
Chinese Translation
在火星上的自主导航要求车辆能够在多样化且视觉上具有挑战性的环境中区分可通行地形。然而，基于学习的外星环境导航的进展受到缺乏大规模数据集的限制。自从在耶泽罗陨石坑着陆以来，火星2020探测车“毅力”已经穿越了从沙丘、岩石地带到平坦基岩的多种地形。因此，本文提出了一个涵盖500个火星日和45公里轨迹的数据集，这些轨迹由人类操作员和机载规划器ENav共同驱动。我们的数据集包含灰度立体图像对、姿态、加速度计读数、摇摆轮角度以及倾斜和轮滑的估计。基于该数据集，我们引入了一种不确定性感知的可通行性估计框架，该框架从多模态驾驶经验中学习地形表示。我们将所提出的方法与现有的火星2020数据集方法进行了比较，结果表明我们的方法达到了0.874的AUROC和0.758的F1分数，分别比最强基线高出0.058和0.156，同时还实现了最高的平均精度和召回率。最后，我们展示了视觉表示可以集成到物理探测车测试平台上的路径规划器，如ENav。视频、代码和M2020数据集将可在https://darren-chiu.github.io/learning-to-drive-on-mars获取。
cs.RO / 180 / 2609.24976
DexTacWAM: A Visuo-Tactile World-Action Model for Dexterous Manipulation
DexTacWAM：一种用于灵巧操作的视觉-触觉世界-动作模型
Abstract
Dexterous manipulation depends on contact dynamics that are often only partially observable from vision. Recent World-Action Models (WAMs) couple predictive video world modeling with action generation, but remain largely vision-centric and therefore cannot directly model these contact dynamics. We present DexTacWAM, a visuo-tactile WAM that encodes each fingertip independently, aggregates the resulting features through a finger- and pose-aware tactile compressor, and injects the tactile latent into a video diffusion world model for joint visuo-tactile world modeling. Across six contact-rich dexterous manipulation tasks on a 22-DoF bimanual platform, DexTacWAM achieves the highest score on every task, averaging 70.6 versus 38.0 for the strongest baseline. Ablations attribute the gain to modeling contact evolution as part of the predicted world state rather than tactile conditioning alone: removing tactile world modeling reduces the four-task mean from 74.7 to 26.6 while keeping the same tactile features and action expert. After four hours of tactile-encoder adaptation with a frozen pretrained vision VAE, our continual vision-to-touch learning extends the pretrained video model to touch using roughly 100 demonstrations per task without tactile midtraining, while retaining visual prediction quality within 0.5 dB of vision-only counterparts. The compressor retains 89.4% of pre-fusion contact recall while enabling 2.26x faster training and 1.29x faster inference. Together, these results show that pretrained video priors can be extended to distributed multi-finger contact dynamics in a data- and compute-efficient manner.
Chinese Translation
灵巧操作依赖于接触动态，而这些动态通常只能通过视觉部分观察到。最近的世界-动作模型（World-Action Models, WAMs）将预测性视频世界建模与动作生成结合起来，但仍然主要以视觉为中心，因此无法直接建模这些接触动态。我们提出了DexTacWAM，一种视觉-触觉WAM，它独立编码每个指尖，通过一个考虑手指和姿态的触觉压缩器聚合结果特征，并将触觉潜变量注入到视频扩散世界模型中，以实现联合视觉-触觉世界建模。在22自由度双手平台上的六个接触丰富的灵巧操作任务中，DexTacWAM在每个任务上都取得了最高分，平均得分为70.6，而最强基线为38.0。消融实验表明，性能提升归因于将接触演变建模为预测世界状态的一部分，而不仅仅是触觉条件：去除触觉世界建模使得四个任务的平均分从74.7降至26.6，同时保持相同的触觉特征和动作专家。在与冻结的预训练视觉变分自编码器（VAE）进行四小时的触觉编码器适应后，我们的持续视觉到触觉学习使用大约每个任务100个示例将预训练的视频模型扩展到触觉，而在视觉预测质量上保持在与仅视觉模型相差0.5 dB以内。压缩器在实现2.26倍更快的训练和1.29倍更快的推理的同时，保留了89.4%的预融合接触召回率。综合来看，这些结果表明，预训练的视频先验可以以数据和计算高效的方式扩展到分布式多指接触动态。
cs.RO / 181 / 2609.24995
MIGU: Multimodal Instruction Grounding under Uncertainty for Manipulation Planning
MIGU：不确定性下的多模态指令基础与操控规划
Abstract
Understanding natural human instructions is crucial for deploying robots in human-centric environments. We study multimodal instruction grounding, where language and gesture provide complementary but uncertain cues. We present MIGU, a modular framework that combines semantic and geometric evidence into a unified grounding belief and connects it to manipulation planning. MIGU constructs a 3D geometric likelihood by propagating viewing-direction and depth uncertainty through eye-finger geometry while accounting for hand-direction estimation error. A vision-language model (VLM) provides semantic priors over candidate objects and regions, which are combined with the geometric likelihood through Bayes-inspired fusion. The resulting belief supports behavior planning to either proceed directly to downstream planning or request clarification. Grounded targets then define goals for mobile manipulation and tabletop task-and-motion planning. On a real-world benchmark, MIGU outperforms all evaluated baselines, while ablations support the benefit of explicit multimodal uncertainty modeling. Project website: multimodal-instruction.github.io
Chinese Translation
理解自然人类指令对于在以人为中心的环境中部署机器人至关重要。我们研究多模态指令基础，其中语言和手势提供互补但不确定的线索。我们提出了MIGU，一个模块化框架，将语义和几何证据结合成统一的基础信念，并将其与操控规划相连接。MIGU通过在考虑手部方向估计误差的同时，通过眼指几何传播视角方向和深度的不确定性，构建了三维几何似然。视觉-语言模型（VLM）为候选对象和区域提供语义先验，这些先验通过贝叶斯启发式融合与几何似然相结合。最终的信念支持行为规划，以便直接进行下游规划或请求澄清。基础目标随后定义了移动操控和桌面任务与运动规划的目标。在一个真实世界基准测试中，MIGU的表现优于所有评估的基线，而消融实验支持显式多模态不确定性建模的优势。项目网站：multimodal-instruction.github.io
cs.RO / 182 / 2609.24996
Learning Beyond What Humans Can Demonstrate
超越人类示范的学习
Abstract
Behavior cloning for robot manipulation relies on expert demonstrations. However, for tasks that require dynamic stability, precise contact timing, or dexterous coordination, human operators may find it hard or even impossible to collect data. We study this infeasible-demonstration regime and propose GLIDE: Guardrails for Learning from Infeasible Demonstrations Efficiently, a framework that infers task-specific failure modes and converts them into executable guardrails for data collection and policy deployment. Given a task description and the conditioning teleoperation code, GLIDE writes guardrails that use system states to filter teleoperation and policy commands, constrain failure-prone actions, and iteratively improve from trajectory feedback. Across three tasks, GLIDE discovers emergent guardrails that go beyond domain-expert hardcoded ones, improving data collection over naive VR teleoperation and domain-expert hardcoded guardrails. After refinement, GLIDE raises data-collection success from 0-10 percent to 70-90 percent across the three tasks. During policy execution, mixed-data guarded policies reach 70 percent, 60 percent, and 60 percent success on Tomato plate transfer, Marker handover and stand, and Wine serving tasks. These results show that GLIDE can support policy learning when direct demonstrations are infeasible. Project website: http://guardrail-policy.github.io/
Chinese Translation
机器人操作的行为克隆依赖于专家示范。然而，对于那些需要动态稳定性、精确接触时机或灵巧协调的任务，人类操作员可能会发现收集数据困难甚至不可能。我们研究了这种不可行示范的情况，并提出了GLIDE：高效学习不可行示范的护栏框架，它推断任务特定的失败模式，并将其转化为可执行的护栏，用于数据收集和策略部署。给定任务描述和条件遥操作代码，GLIDE编写护栏，利用系统状态过滤遥操作和策略命令，限制易失败的动作，并通过轨迹反馈进行迭代改进。在三个任务中，GLIDE发现了超越领域专家硬编码的紧急护栏，改善了相较于简单的虚拟现实遥操作和领域专家硬编码护栏的数据收集。在经过优化后，GLIDE将三个任务的数据收集成功率从0-10%提高到70-90%。在策略执行过程中，混合数据保护策略在番茄盘转移、标记交接和站立、以及葡萄酒服务任务中分别达到了70%、60%和60%的成功率。这些结果表明，当直接示范不可行时，GLIDE能够支持策略学习。项目网站：http://guardrail-policy.github.io/
cs.CV / 1 / 2609.22267
Did You Steal My Shot? Pioneering Camera Motion Plagiarism Detection in Generative Videos
你偷了我的镜头吗？开创性的视频生成中的摄像机运动抄袭检测
Abstract
Camera motion often reflects directorial intent and requires professional equipment, making it a high value form of intellectual property. However, generative video models can imitate such high value camera motions with simple prompts, while existing similarity detection methods mainly operate on visual content and fail to capture deeper motion similarity. This is mainly because their training data entangles camera motion with visual content. Moreover, traditional optical flow is insufficient to represent complex camera motions. We therefore build the first benchmark for camera motion analysis, including a motion dataset with \textbf{11} motion styles and evaluation protocols. Furthermore, we propose a motion representation that augments optical flow with vorticity cues from fluid dynamics, thereby better capturing motions. Experiments show that our detector achieves a \textbf{3.02*} improvement in plagiarism detection over the strongest baseline and remains effective on generative videos. We believe our work extends copyright protection beyond static content to dynamic camera motion.
Chinese Translation
摄像机运动通常反映导演的意图，并且需要专业设备，这使其成为一种高价值的知识产权形式。然而，生成视频模型可以通过简单的提示模仿这种高价值的摄像机运动，而现有的相似性检测方法主要针对视觉内容，未能捕捉更深层次的运动相似性。这主要是因为它们的训练数据将摄像机运动与视觉内容纠缠在一起。此外，传统的光流方法不足以表示复杂的摄像机运动。因此，我们建立了第一个摄像机运动分析基准，包括一个包含11种运动风格的运动数据集和评估协议。此外，我们提出了一种运动表示方法，通过流体动力学中的涡度线索增强光流，从而更好地捕捉运动。实验表明，我们的检测器在抄袭检测上比最强基线提高了3.02倍，并且在生成视频上仍然有效。我们相信我们的工作将版权保护的范围从静态内容扩展到动态摄像机运动。
cs.CV / 2 / 2609.22271
Enabling Vision and Cross-Modal Learning for Multimodal Stroke Recurrence Prediction: An Interpretable Two-Step Framework
启用视觉与跨模态学习的多模态中风复发预测：一个可解释的两步框架
Abstract
Multimodal stroke recurrence prediction requires effective integration of heterogeneous clinical and imaging data, yet modality imbalance often causes models to over-rely on dominant modalities and underutilize complementary information. While self-supervised pretraining and selective parameter freezing are commonly employed to improve representation learning and fine-tuning stability, their effect on modality contributions and cross-modal behavior in multimodal medical models remains largely unexplored. In this work, we investigate whether image pretraining on 3D CTA scans reduces modality imbalance and improves cross-modal integration for stroke recurrence prediction, a clinically critical task we recently addressed. To this end, two multimodal neural networks are pretrained in a self-supervised manner and subsequently fine-tuned using two distinct freezing strategies. Their performance and modality utilization are compared against both the baseline model from our previous work and models trained entirely from scratch in this study. Our results demonstrate that self-supervised pretraining enables more effective utilization of the multimodal image-tabular dataset, outperforming both the prior baseline and all non-pretrained models. Notably, the best-performing Vision Transformer based neural network successfully overcomes unimodal collapse. Synergy analysis reveals significant interactions between vision and both gender and CHD, suggesting clinically relevant patterns for stroke recurrence. Overall, our findings demonstrate that self-supervised pretraining and strategic fine-tuning support more balanced modality utilization and enable meaningful cross-modal interactions. Code is publicly available at https://github.com/ChristianGappGit/SSL_Pretraining.
Chinese Translation
多模态中风复发预测需要有效整合异质的临床和影像数据，但模态不平衡往往导致模型过度依赖主导模态而未充分利用互补信息。尽管自监督预训练和选择性参数冻结通常用于改善表征学习和微调稳定性，但它们对多模态医学模型中模态贡献和跨模态行为的影响仍然未得到充分探索。在本研究中，我们探讨了在3D CTA扫描上进行图像预训练是否能够减少模态不平衡并改善中风复发预测的跨模态整合，这是我们最近解决的一个临床关键任务。为此，两个多模态神经网络以自监督的方式进行预训练，并随后使用两种不同的冻结策略进行微调。我们将它们的性能和模态利用与我们之前工作的基线模型以及本研究中完全从头训练的模型进行比较。我们的结果表明，自监督预训练能够更有效地利用多模态图像-表格数据集，超越了之前的基线和所有未预训练模型。值得注意的是，表现最佳的基于视觉变换器（Vision Transformer）的神经网络成功克服了单模态崩溃。协同分析揭示了视觉与性别及冠心病（CHD）之间的显著交互，暗示了与中风复发相关的临床模式。总体而言，我们的研究结果表明，自监督预训练和战略性微调支持更平衡的模态利用，并促进有意义的跨模态交互。代码已公开发布在 https://github.com/ChristianGappGit/SSL_Pretraining。
cs.CV / 3 / 2609.22272
Moonworks Lunara: Modeling Artistic Intelligence
Moonworks Lunara：艺术智能建模
Abstract
We formulate \emph{Artistic Intelligence} as exploration driven world realization, leaving space for creative possibility while preserving the semantic, artistic, and compositional structure that must remain true. Moonworks Lunara, a text-to-image model, implements this framework with a novel Diffusion Mixture Transformer architecture. A new training algorithm iteratively evolves the data distribution through informative sample acquisition and targeted injection of human-created art. We benchmark Lunara against seven image-generation models, including FLUX.2-Klein-4B, Qwen-Image (20B), and GPT-Image-1-Mini. With GPT-5.6 Sol as evaluator, Lunara ranks first in \emph{Aesthetic Quality (8.473 vs. 8.457 GPT-Image-1-mini)}, second in \emph{Emotional Resonance}, and remains competitive in \emph{Content Integrity}. A blind human evaluation over the same evaluation set corroborates the automated metrics, ranking Lunara first. It also stays among the strongest models under conventional measures including CLIPScore and LAION Aesthetic Predictor. On GenEval, Lunara achieves competitive performance against a broader set of 16 models, including GPT Image 2 and Seedream 4.0. These results place Lunara at the frontier with Artistic Intelligence while maintaining a sub-10B active-parameter footprint and sub-10-second inference latency. Lunara advances the general visual intelligence frontier by shifting the question from whether models can get images right to how deeply they can interpret meaning and realize it as imaginative, expressive worlds.
Chinese Translation
我们将 extit{艺术智能}定义为以探索驱动的世界实现，既为创造性可能性留出空间，又保持必须真实的语义、艺术和构图结构。Moonworks Lunara，一个文本到图像的模型，采用了一种新颖的扩散混合变换器（Diffusion Mixture Transformer）架构来实现这一框架。一种新的训练算法通过信息样本获取和针对性注入人类创作的艺术，迭代地演变数据分布。我们将Lunara与七个图像生成模型进行了基准测试，包括FLUX.2-Klein-4B、Qwen-Image（20B）和GPT-Image-1-Mini。在GPT-5.6 Sol的评估下，Lunara在 extit{美学质量}上排名第一（8.473对比8.457，GPT-Image-1-mini），在 extit{情感共鸣}上排名第二，并在 extit{内容完整性}上保持竞争力。对同一评估集的盲人评估证实了自动化指标，Lunara排名第一。在传统测量标准下，包括CLIPScore和LAION美学预测器，Lunara也保持在最强模型之列。在GenEval上，Lunara在包括GPT Image 2和Seedream 4.0在内的16个模型中表现出竞争力。这些结果使Lunara在艺术智能的前沿，同时保持低于10B的活跃参数足迹和低于10秒的推理延迟。Lunara通过将问题从模型是否能正确生成图像转变为它们能多深刻地解读意义并将其实现为富有想象力和表现力的世界，推动了视觉智能的整体前沿。
cs.CV / 4 / 2609.22281
Performance vs Consistency: Evaluating a Foundation Model in Lung-RADS Screening
性能与一致性：评估肺癌筛查中的基础模型
Abstract
Foundation models have recently demonstrated strong capabilities across a wide range of medical imaging tasks. However, their performance in structured clinical interpretation settings remains insufficiently explored. In lung cancer screening, interpretative variability persists despite standardized frameworks such as Lung-RADS. In this study, we evaluate MedGemma, a medical general-purpose foundation model derived from Gemini and its fine-tuned version adapted for lung cancer detection and diagnosis, compared against radiologists performing Lung-RADS v2022 assessment on the NLST dataset. Twelve radiologists independently evaluated each case in a multi-reader design, enabling quantification of inter-reader variability. Radiologists achieved a mean AUC of 0.90, with substantial variability across readers (range: 0.80-0.94). The native foundation model achieved an AUC of 0.70, failing to reach clinically relevant performance. In contrast, fine-tuning significantly improved performance to an AUC of 0.83, placing the model within the lower range of individual radiologists performance. These findings highlight a trade-off between peak accuracy and prediction consistency. Unlike radiologists, under fixed conditions, the model produces deterministic outputs, removing inter-run variability under identical inputs, in contrast to inter-reader variability observed among radiologists. This supports the role of fine-tuned foundation models potential complementary tools for clinical decision support, particularly in settings with limited expertise. However, evaluation is performed on a case-enriched cohort from NLST and does not account for real-world prevalence or external validation, limiting direct clinical generalization.
Chinese Translation
基础模型最近在广泛的医学影像任务中展示了强大的能力。然而，它们在结构化临床解读环境中的表现仍然未得到充分探索。在肺癌筛查中，尽管存在肺癌筛查标准（Lung-RADS）等标准化框架，解读的变异性依然存在。在本研究中，我们评估了MedGemma，这是一种源自Gemini的医学通用基础模型及其为肺癌检测和诊断而微调的版本，并与在NLST数据集上执行Lung-RADS v2022评估的放射科医师进行比较。十二名放射科医师在多读者设计中独立评估每个案例，从而量化读者间的变异性。放射科医师的平均AUC为0.90，但不同读者之间存在显著的变异性（范围：0.80-0.94）。原始基础模型的AUC为0.70，未能达到临床相关的表现。相比之下，微调显著提高了性能，AUC达到了0.83，使该模型处于个别放射科医师表现的下限。这些发现突显了峰值准确性与预测一致性之间的权衡。在固定条件下，与放射科医师不同，该模型产生确定性的输出，消除了在相同输入下的运行间变异性，而放射科医师之间则观察到了读者间的变异性。这支持了微调基础模型作为临床决策支持的潜在补充工具的角色，特别是在专业知识有限的环境中。然而，评估是在NLST的案例丰富队列上进行的，并未考虑现实世界的流行病学或外部验证，限制了直接的临床推广。
cs.CV / 5 / 2609.22282
Brain-to-Image Generation: Reconstructing Visual Stimuli from EEG using Generative Adversarial Networks
脑到图像生成：使用生成对抗网络从脑电图重建视觉刺激
Abstract
Reconstructing visual stimuli from electroencephalography (EEG) is difficult because scalp measurements have high temporal but limited spatial resolution, and paired EEG-image datasets remain small relative to modern generative-model training corpora. We present a reproducible single-subject baseline on THINGS-EEG2 that first tests the more defensible question of whether EEG can retrieve the viewed stimulus in a visual embedding space. A compact temporal-spatial convolutional encoder maps repetition-averaged EEG (63 by 250) to provided 512-dimensional ViT-B/32 image features. Model selection uses a concept-disjoint validation split, and final evaluation uses the official 200-image, 200-concept test gallery. Across three training seeds, the model obtains 12.83 +/- 0.58%, 39.17 +/- 1.76%, and 58.00 +/- 1.73% image recall at 1, 5, and 10 (mean +/- sample standard deviation), compared with analytical chance levels of 0.5%, 2.5%, and 5.0%. A session-balanced ablation shows that averaging more test repetitions generally improves ranking. Applying the Subject 01 model to the other nine subjects without adaptation causes a sharp performance drop, exposing subject specificity. We further report exploratory stress tests of direct conditional generators trained without external visual weights: single-subject and ten-subject variants produce noise-dominated outputs, with early validation improvements reversing after one to four epochs. Finally, we distinguish direct reconstruction from semantic rendering with a pretrained diffusion prior. The results support above-chance coarse semantic decoding under a closed-set, repetition-averaged protocol, but do not support faithful recovery of stimulus pixels.
Chinese Translation
从脑电图（EEG）重建视觉刺激是一个困难的任务，因为头皮测量具有高时间分辨率但空间分辨率有限，并且配对的EEG-图像数据集相对于现代生成模型训练语料库仍然较小。我们在 THINGS-EEG2 上提出了一个可重复的单一受试者基线，首先测试更具防御性的问题，即 EEG 是否能够在视觉嵌入空间中检索所观看的刺激。一个紧凑的时空卷积编码器将重复平均的 EEG（63 x 250）映射到提供的 512 维 ViT-B/32 图像特征。模型选择使用概念不重叠的验证拆分，最终评估使用官方的 200 图像、200 概念测试库。在三个训练种子下，模型在 1、5 和 10 的图像召回率分别为 12.83 +/- 0.58%、39.17 +/- 1.76% 和 58.00 +/- 1.73%（均值 +/- 样本标准差），而分析的机会水平分别为 0.5%、2.5% 和 5.0%。会话平衡消融显示，通常情况下，平均更多的测试重复会改善排名。将受试者 01 模型应用于其他九个受试者而不进行适应会导致性能急剧下降，暴露出受试者特异性。我们进一步报告了直接条件生成器的探索性压力测试，这些生成器在没有外部视觉权重的情况下训练：单一受试者和十个受试者的变体产生了以噪声为主导的输出，早期验证的改进在一到四个周期后反转。最后，我们通过预训练的扩散先验区分直接重建与语义渲染。结果支持在封闭集、重复平均协议下的高于机会的粗略语义解码，但不支持刺激像素的忠实恢复。
cs.CV / 6 / 2609.22283
Rethinking Streaming Video Diffusion Model: Context, Execution, and Training
重新思考流媒体视频扩散模型：背景、执行与训练
Abstract
Understanding the design space of streaming video diffusion is essential to exploring its potential for generation quality and computational efficiency. We develop a unified analytical framework that relates model and sampler choices, historical conditioning, execution scheduling, and training strategies. The framework accommodates a broad family of causal context-selection policies and makes their computational dependencies and training-inference alignment explicit. Within this design space, we study three representative policies: clean, same-level, and progressive history. On the full VBench prompt set, same-level and progressive history achieve aggregate scores of 85.24 and 85.60, respectively, compared with 84.45 for the clean-history reference. Long-video comparisons further show improved subject consistency and more coherent motion with progressive history. By allowing multiple denoising nodes to be processed together, progressive-history pipelining achieves $1.57$-$2.83\times$ steady-state DiT speedups under our evaluated conditions. We additionally find that LoRA adaptation of the DMD fake-score network improves generation quality using only 2.15% as many trainable fake-score parameters as full-parameter adaptation. Together, these findings show that fully denoised history is not a prerequisite for high-quality streaming generation and motivate the joint design of historical conditioning, execution, and training.
Chinese Translation
理解流媒体视频扩散的设计空间对于探索其在生成质量和计算效率方面的潜力至关重要。我们开发了一个统一的分析框架，该框架将模型和采样器选择、历史条件、执行调度和训练策略联系起来。该框架适应了一系列广泛的因果上下文选择策略，并明确了它们的计算依赖关系和训练-推理对齐。在这个设计空间中，我们研究了三种代表性的策略：干净历史、同级历史和渐进历史。在完整的 VBench 提示集上，同级历史和渐进历史分别取得了 85.24 和 85.60 的综合得分，而干净历史参考得分为 84.45。长视频比较进一步显示，渐进历史在主题一致性和运动连贯性方面有所改善。通过允许多个去噪节点一起处理，渐进历史流水线在我们评估的条件下实现了 $1.57$-$2.83 imes$ 的稳态 DiT 加速。此外，我们发现 DMD 假分数网络的 LoRA 适应仅使用 2.15% 的可训练假分数参数就能提高生成质量，相较于全参数适应。综合这些发现表明，完全去噪的历史并不是高质量流媒体生成的先决条件，并激励历史条件、执行和训练的联合设计。
cs.CV / 7 / 2609.22284
Complementary rPPG-Derived and Lip-Region Frequency Cues for Talking-Face Deepfake Detection
互补的基于rPPG和唇区频率线索用于谈话面孔深度伪造检测
Abstract
Talking-face (TF) deepfakes are detected unevenly by rPPG-based methods across generators. We study two lightweight visual-only cues, rPPG-derived waveforms extracted by RhythmFormer and lip-region discrete cosine transform (DCT) coefficients, on the seven TF methods of Celeb-DF++ under a subject-independent protocol. In-domain, lip-region DCT matches or exceeds the rPPG-derived 1D ResNet on every method except SadTalker, and Concat fusion reaches AUC 0.891 against 0.824 and 0.827 for the unimodal baselines. Under leave-one-generator-out evaluation the cues split: each transfers clearly better to three held-out methods, and IP-LAP is near chance for both. Concat averages 0.798 but falls below rPPG alone where DCT transfers poorly, so static fusion only partly exploits this complementarity. Lip-region DCT outperforms full-face DCT on six of seven methods. We treat the rPPG-derived signal as an empirical cue and do not claim it is cardiac in origin.
Chinese Translation
谈话面孔（TF）深度伪造在不同生成器中通过基于rPPG的方法检测效果不均。我们研究了两种轻量级的仅视觉线索，分别是通过RhythmFormer提取的rPPG衍生波形和唇区离散余弦变换（DCT）系数，针对Celeb-DF++的七种TF方法，在一个独立于受试者的协议下进行评估。在领域内，唇区DCT在除SadTalker之外的每种方法上都与rPPG衍生的1D ResNet相匹配或超越，而Concat融合在AUC上达到0.891，相较于单模态基线的0.824和0.827。在留一生成器评估中，这些线索的表现有所分化：每个线索在三个保留方法上都有明显更好的转移，而IP-LAP在两者中接近随机水平。Concat的平均值为0.798，但在DCT转移效果不佳的情况下低于单独的rPPG，因此静态融合仅部分利用了这种互补性。唇区DCT在七种方法中的六种上优于全脸DCT。我们将rPPG衍生信号视为一种经验线索，并不声称其源于心脏。
cs.CV / 8 / 2609.22291
Beyond the Survey: A Systematic Empirical Study of Detection and Association in Visual MOT
超越调查：视觉多目标跟踪中检测与关联的系统实证研究
Abstract
This paper presents a comprehensive experimental evaluation and detailed analysis of state-of-the-art multi-object tracking algorithms, with an emphasis on quantifying the individual contributions of detection and association components to overall tracking performance. Unlike existing surveys that primarily offer theoretical categorizations or taxonomies of tracking methods, our work adopts a rigorous experimental perspective grounded in publicly available implementations, providing practical guidance for researchers and practitioners in method selection and system design. We introduce a unified pipeline diagram that consolidates the core components across the two main branches of visual multi-object tracking: tracking-by-detection and end-to-end deep learning paradigms, and systematically analyze the object detection, feature extraction, and data association modules. Through extensive empirical studies on standard benchmarks, including MOT16, MOT17, MOT20, SportsMOT, DanceTrack, and CrowdTrack datasets, we reveal critical insights: (1) detection quality dominates association strategy performance, with detector improvements yielding more than 10% gains compared to less than 5% from refined association strategies; (2) modern deep learning detectors paired with specialized re-identification models significantly outperform joint detection and embedding approaches; and (3) transformer-based end-to-end methods exhibit greater robustness to detection quality variations but at a substantial computational cost. Our findings from extensive experiments provide key insights into component-level effects in MOT, particularly the dominant influence of detection quality relative to association, while offering practical insights for designing and optimizing MOT systems under varying performance and robustness requirements. Code and experimental setups are available at github.com/linh-gist/VisualMOT.
Chinese Translation
本文对最先进的多目标跟踪算法进行了全面的实验评估和详细分析，重点量化检测和关联组件对整体跟踪性能的个体贡献。与现有主要提供理论分类或跟踪方法分类的调查不同，我们的工作采用了一种严格的实验视角，基于公开可用的实现，为研究人员和实践者在方法选择和系统设计方面提供实用指导。我们介绍了一个统一的流程图，整合了视觉多目标跟踪两个主要分支的核心组件：基于检测的跟踪和端到端深度学习范式，并系统分析了目标检测、特征提取和数据关联模块。通过对标准基准数据集（包括MOT16、MOT17、MOT20、SportsMOT、DanceTrack和CrowdTrack）的广泛实证研究，我们揭示了关键见解：（1）检测质量主导关联策略性能，检测器的改进带来了超过10%的增益，而精细化的关联策略则不足5%；（2）现代深度学习检测器与专业的重识别模型配对，显著优于联合检测和嵌入方法；（3）基于变换器的端到端方法对检测质量变化表现出更大的鲁棒性，但代价是显著的计算成本。我们在广泛实验中获得的发现为多目标跟踪中的组件级影响提供了关键见解，特别是检测质量相对于关联的主导影响，同时为在不同性能和鲁棒性要求下设计和优化多目标跟踪系统提供了实用见解。代码和实验设置可在github.com/linh-gist/VisualMOT获取。
cs.CV / 9 / 2609.22293
Validating, Not Sampling: Region-Level Robustness of Vision-Language and Vision-Language-Action Models
验证而非采样：视觉-语言和视觉-语言-动作模型的区域级鲁棒性
Abstract
Vision-language models (VLMs) and vision-language-action models (VLAs) are increasingly deployed in real-world applications. There, a small perturbation to the recorded camera image may change a decision significantly. However, existing benchmarks for these models only sample perturbations, which does not guarantee the absence of a failure in the untested region. We present the first robustness validation of six VLMs (drawn from the Gemma, InternVL, LLaVA, and Qwen families) and five VLAs (drawn from the GR00T, OpenVLA, and $\pi$ families) over entire continuous regions of photometric and geometric image perturbation: brightness shifts, camera rotations, and their composition. To this end, we build on the validation framework H$^2$V and introduce H$^2$V-M, a margin-aware convergence rule that makes validation affordable at the 32B parameter scale. We demonstrate that H$^2$V-M outperforms H$^2$V by an order of magnitude in model queries and that it finds counterexamples faster than random sampling while providing soundness guarantees. Our VLM and VLA robustness validation shows that robustness is mostly dependent on the perturbation type, rather than the model, and that VLMs are more robust to large camera rotations than VLAs. For VLAs, even perturbations as small as $\pm1^\circ$ can change the commanded action in many cases. We also show that robustness depends more on model family than on model size.
Chinese Translation
视觉-语言模型（VLMs）和视觉-语言-动作模型（VLAs）在现实世界应用中越来越多地被部署。在这些应用中，摄像头图像的微小扰动可能会显著改变决策。然而，现有的这些模型基准测试仅对扰动进行采样，这并不能保证未测试区域内不存在失败。我们首次对六个VLM（来自Gemma、InternVL、LLaVA和Qwen家族）和五个VLA（来自GR00T、OpenVLA和$ ext{π}$家族）在光度和几何图像扰动的整个连续区域进行鲁棒性验证：亮度变化、摄像机旋转及其组合。为此，我们基于验证框架H$^2$V，提出了H$^2$V-M，这是一种考虑边际的收敛规则，使得在32B参数规模下的验证变得可负担。我们证明H$^2$V-M在模型查询方面的性能比H$^2$V提高了一个数量级，并且它比随机采样更快地找到反例，同时提供了健全性保证。我们的VLM和VLA鲁棒性验证表明，鲁棒性主要依赖于扰动类型，而非模型，并且VLM对大摄像机旋转的鲁棒性优于VLA。对于VLA而言，即使是$ ext{±}1^ ext{°}$的微小扰动在许多情况下也能改变指令动作。我们还展示了鲁棒性更依赖于模型家族而非模型大小。
cs.CV / 10 / 2609.22295
On The Robustness-Resolution Tradeoff In Temporal Quantization Of Event Streams
事件流时间量化中的鲁棒性与分辨率权衡
Abstract
Event pipelines often discretize asynchronous timestamps before learning. This step looks harmless, but its stability depends directly on temporal resolution. We study this dependence at the representation level. We first show that hard temporal binning is discontinuous: an arbitrarily small timestamp shift near a boundary can move unit event mass between bins. We then define a class of nonnegative, mass-preserving, resolution-faithful continuous encoders and prove that every encoder in this class has global L1 sensitivity at least 2/Delta, where Delta denotes bin width. Linear two-bin interpolation attains this limit. Local support and first-moment preservation also make it unique. Experiments on SHD, N-MNIST, and DVS128 Gesture support the analysis. Across uniform timestamp budgets, linear interpolation lowers mean representation drift by 47-72% while keeping clean accuracy nearly unchanged. On DVS Gesture, it produces zero prediction flips across all tested budgets and three seeds. On SHD, measured drift follows 1/Delta with R^2 = 0.992.
Chinese Translation
事件管道通常在学习之前对异步时间戳进行离散化。这一步看似无害，但其稳定性直接依赖于时间分辨率。我们在表示层面研究这种依赖关系。我们首先展示了硬时间分箱是非连续的：在边界附近的任意小时间戳偏移可以在分箱之间移动单位事件质量。然后，我们定义了一类非负的、保持质量的、忠实于分辨率的连续编码器，并证明该类中的每个编码器的全局 L1 灵敏度至少为 2/Delta，其中 Delta 表示分箱宽度。线性双分箱插值达到了这一极限。局部支持和第一矩保持也使其独特。在 SHD、N-MNIST 和 DVS128 Gesture 上的实验支持了这一分析。在均匀时间戳预算下，线性插值将平均表示漂移降低了 47-72%，同时保持了几乎不变的准确性。在 DVS Gesture 上，它在所有测试的预算和三个种子下产生了零预测翻转。在 SHD 上，测量的漂移遵循 1/Delta，R^2 = 0.992。
cs.CV / 11 / 2609.22302
Authority-Preserving Evaluation of Medical Vision-Language Assistants
权威保留的医疗视觉-语言助手评估
Abstract
Medical vision-language models can propose how urgently a skin lesion should be reviewed, but the local service retains authority to accept or replace that proposal under referral policy, capacity, and locally held patient context. Proposal quality and selected-action quality are therefore distinct evaluation targets, and benchmark evidence transfers between them only when local review preserves the expected action score. We introduce AuthEval, a logging and evaluation framework that records both actions, scores the selected action under declared local criteria, and, where feasible, scores the declined proposal under the same rule. It reports the resulting authority gap only when the record supports it. Because the gap is the product of the proposal-change rate and the mean score change on changed cases, that rate alone determines neither its magnitude nor its sign. On ISIC 2019, with MedGemma and simulated local review, two constraint regimes with similar change rates produced an optimistic image-equal gap under capacity ($+0.744$ simulator units) but no detectable gap under safety. The declared evaluation unit also mattered: under mixed constraints the gap reversed from $+0.374$ to $-0.206$ when weighting shifted from image to lesion-aware cluster. AuthEval thus clarifies whether a study's records support claims about the model, the workflow, or both.
Chinese Translation
医疗视觉-语言模型可以提出皮肤病变应当多紧急进行审查，但根据转诊政策、能力和当地患者背景，地方服务仍然保留接受或替换该提议的权力。因此，提议质量和选择行动质量是不同的评估目标，基准证据仅在地方审查保留预期行动评分时在它们之间转移。我们引入了AuthEval，一个记录和评估框架，记录两种行动，根据声明的地方标准对所选行动进行评分，并在可行的情况下，根据相同规则对被拒绝的提议进行评分。只有当记录支持时，它才会报告结果的权威差距。由于该差距是提议变更率和变更案例的平均评分变化的乘积，仅该变更率并不能决定其大小或符号。在ISIC 2019数据集上，使用MedGemma和模拟的地方审查，两种具有相似变更率的约束机制在能力下产生了乐观的图像等差距（$+0.744$模拟单位），但在安全性下没有可检测的差距。声明的评估单位也很重要：在混合约束下，当权重从图像转向病变感知集群时，差距从$+0.374$反转为$-0.206$。因此，AuthEval澄清了研究记录是否支持关于模型、工作流程或两者的声明。
cs.CV / 12 / 2609.22308
GameReplica: A Benchmark for Black-Box Visual Game Replication by Vision-Language Agents
GameReplica：视觉-语言智能体黑箱视觉游戏复制的基准测试
Abstract
Coding-agent benchmarks usually evaluate implementation after the target behavior has been specified in text, code, or demonstrations. Existing research has extensively evaluated the ability of coding agents to generate programs from textual specifications. However, under black-box conditions where neither source code nor documentation is available, it remains underexplored whether an agent can induce the rules solely through visual observation and active interaction and reproduce the target system as a verifiable executable system. To this end, we present GameReplica, a closed-loop evaluation framework for end-to-end black-box game replication that covers the full perception, exploration, induction, reproduction, and verification pipeline. GameReplica comprises 125 tasks spanning 25 games across 5 core mechanism families, with each game instantiated at five difficulty levels. The tasks require an agent to access the target game only through screenshots and an action interface, induce the key visual elements and gameplay rules from pixel feedback and interaction outcomes, and generate a self-contained, runnable game replica that can be automatically verified by an external program. Experiments show that current coding agents still face substantial challenges in end-to-end black-box replication: the best-performing model (Claude Opus 4.8) achieves an overall score of 71.6\%, while the remaining models score only 4.0\%--42.9\%. Further analysis reveals a consistent pattern across all models: visual-fidelity scores are substantially higher than implementation- and rule-consistency scores, indicating that agents replicate visual appearance more readily than game mechanics. The difficulty levels further amplify the performance gap: from L1 to L5, the overall score of weaker agents drops sharply, whereas that of the best-performing agent declines only slightly.
Chinese Translation
编码智能体基准通常在目标行为以文本、代码或演示形式指定后进行评估。现有研究广泛评估了编码智能体从文本规范生成程序的能力。然而，在黑箱条件下，即没有源代码或文档可用的情况下，尚未深入探讨智能体是否能够仅通过视觉观察和主动交互来推导规则，并重现目标系统作为可验证的可执行系统。为此，我们提出了GameReplica，这是一个闭环评估框架，用于端到端黑箱游戏复制，涵盖完整的感知、探索、推导、再现和验证流程。GameReplica包含125个任务，涵盖5个核心机制家族中的25款游戏，每款游戏在五个难度级别上进行实例化。这些任务要求智能体仅通过截图和动作接口访问目标游戏，从像素反馈和交互结果中推导关键视觉元素和游戏规则，并生成一个自包含的、可运行的游戏复制品，该复制品可以被外部程序自动验证。实验表明，目前的编码智能体在端到端黑箱复制中仍面临重大挑战：表现最佳的模型（Claude Opus 4.8）的整体得分为71.6%，而其余模型的得分仅为4.0%至42.9%。进一步分析显示，所有模型之间存在一致的模式：视觉保真度得分显著高于实现和规则一致性得分，表明智能体更容易复制视觉外观而非游戏机制。难度级别进一步加大了性能差距：从L1到L5，较弱智能体的整体得分急剧下降，而表现最佳的智能体的得分仅略有下降。
cs.CV / 13 / 2609.22315
Yarn tracking of large-scale 3D textile reinforcements using topological material features
基于拓扑材料特征的大规模三维纺织增强材料的纱线追踪
Abstract
Automated segmentation of CT images has become increasingly important to enhance the reliability of simulations through the generation of high fidelity numerical models. This study addresses the challenging task of semi-automatically tracking textile reinforcements in fan blade dry preforms using X-ray CT images captured at coarse resolutions (i.e., above 140 $\mu$m). Our approach offers a scalable, slice-based analysis conducted on planes orthogonal to the main yarn directions, applied to a large-scale real industrial component. This enables accurate identification and tracking of yarn paths while requiring minimal training. The method models three key yarn properties statistically: their typical cross-section shape, their continuity and movement in the 3D space, and their spatial relative arrangement with respect to neighboring yarns. These statistical properties are integrated into a tracking framework via a variational formulation that optimizes all yarn center positions in successive cross-section planes. The method tracks more than 3,000 warp yarns across 1,500 slices and achieves a tracking success rate above 90%. Overall, this work demonstrates a promising approach toward large-scale, automated textile reinforcement annotation, paving the way for more efficient material characterization in complex composite structures.
Chinese Translation
CT图像的自动分割在通过生成高保真数值模型来增强模拟的可靠性方面变得越来越重要。本研究针对在粗分辨率（即高于140 $00$μm）下捕获的X射线CT图像中半自动追踪风扇叶片干预形状中的纺织增强材料这一具有挑战性的任务。我们的方法提供了一种可扩展的基于切片的分析，针对与主要纱线方向正交的平面进行，应用于大型真实工业组件。这使得在需要最少训练的情况下，能够准确识别和追踪纱线路径。该方法统计建模了三个关键的纱线特性：它们的典型横截面形状、在三维空间中的连续性和运动，以及与相邻纱线的空间相对排列。这些统计特性通过变分形式集成到追踪框架中，优化了连续切片平面中所有纱线中心位置的定位。该方法在1,500个切片中追踪了超过3,000根经纱，追踪成功率超过90%。总体而言，这项工作展示了一种有前景的方法，旨在实现大规模自动化纺织增强材料的标注，为复杂复合材料结构中更高效的材料表征铺平道路。
cs.CV / 14 / 2609.22323
ALPINE: Adaptive Localization for Parameter- and Sample-Efficient Few-Shot Learning
ALPINE：用于参数和样本高效的自适应定位的少样本学习
Abstract
Few-shot learning research is predominantly evaluated on accuracy alone, with limited attention to the parameter and training-sample budgets required to reach that accuracy - a real constraint for practitioners without large-scale compute. We present an ultra-lightweight (22,249-34,917 parameter) spatial-relational architecture for few-shot image classification that combines fixed Gabor edge-energy guidance with a windowed, content-adaptive patch locator. Under a strictly matched, iso-episode-budget protocol (250 meta-training episodes, 5 canonical seeds, 600 evaluation episodes per seed), our architecture achieves 5-shot accuracy gains, consistent across all five seeds, over Prototypical Networks, Relation Networks, and MAML on both CIFAR-FS and MiniImageNet, while using 27-53% fewer parameters than any baseline. It also converges in fewer training episodes, generalizes better to an unseen fine-grained domain (CUB-200-2011 birds, zero retraining), and is more robust to 50% occlusion and 25% spatial translation than all three baselines. A series of falsification ablations - zeroing relational tokens at inference and retraining without them entirely - shows that the architecture's pairwise relational computation, while present, is not the primary driver of its performance; the content-adaptive patch locator is. We report this honestly, together with a capacity sweep showing a genuine accuracy plateau near 22-35k parameters, and release full seed-level results and checkpoint hashes for reproducibility.
Chinese Translation
少样本学习研究主要仅通过准确性进行评估，较少关注实现该准确性所需的参数和训练样本预算——这对没有大规模计算资源的从业者来说是一个实际限制。我们提出了一种超轻量级（22,249-34,917 参数）空间关系架构，用于少样本图像分类，该架构结合了固定的 Gabor 边缘能量引导和窗口化的内容自适应补丁定位器。在严格匹配的等集协议下（250 个元训练集，5 个典型种子，每个种子 600 个评估集），我们的架构在 CIFAR-FS 和 MiniImageNet 上相较于原型网络（Prototypical Networks）、关系网络（Relation Networks）和 MAML 实现了 5-shot 准确性提升，并且在所有五个种子上保持一致，同时使用的参数比任何基线少 27-53%。它在训练集数目上收敛更快，对未见的细粒度领域（CUB-200-2011 鸟类，零重训练）具有更好的泛化能力，并且在 50% 遮挡和 25% 空间平移方面比所有三个基线更具鲁棒性。一系列的虚假消融实验——在推理时将关系标记归零以及完全不使用它们进行重训练——表明该架构的成对关系计算虽然存在，但并不是其性能的主要驱动因素；内容自适应补丁定位器才是。我们诚实地报告这一点，并提供了一个容量扫描，显示在 22-35k 参数附近确实存在准确性平台，并发布了完整的种子级结果和检查点哈希以便于重现。
cs.CV / 15 / 2609.22333
Dimensionality reduction for AI based hyperspectral image classification based on XAI
基于可解释人工智能的超光谱图像分类的降维方法
Abstract
This research addresses the challenge of limited material recycling in wood recycling processes by leveraging artificial intelligence (AI)-based dimensionality reduction. Our study explores the application of convolutional neural networks (CNNs) in multi-channel hyperspectral imaging (HSI), extending beyond RGB channels to over 200 spectral channels. Dimensionality reduction within this context involves streamlining the feature space for AI system training and inference. Focusing on explainable AI (XAI) methods, this paper contributes to a broader research initiative, presenting a solution framework that enhances the sustainability and efficiency of wood recycling processes.
Chinese Translation
本研究通过利用基于人工智能（AI）的降维技术，解决木材回收过程中材料回收有限的挑战。我们的研究探讨了卷积神经网络（CNN）在多通道超光谱成像（HSI）中的应用，超越了RGB通道，涉及超过200个光谱通道。在此背景下，降维涉及简化特征空间，以便于AI系统的训练和推理。本文聚焦于可解释人工智能（XAI）方法，为更广泛的研究倡议做出贡献，提出了一种解决方案框架，以提升木材回收过程的可持续性和效率。
cs.CV / 16 / 2609.22351
Hierarchical Aggregation of Semantic Uncertainty in 3D Scene Graphs
3D场景图中语义不确定性的层次聚合
Abstract
Open-vocabulary 3D Scene Graphs (3DSGs) ground each object node in a vision-language embedding, yet they record every entry as equally certain, so a robot querying the map cannot tell which of its entries are unreliable. Estimators of semantic uncertainty could supply that distinction, but they require repeated sampling of a model, training, or held-out labels, none of which are available to a deployed system at query time. We present a framework that exploits the detector confidence and the embeddings a 3DSG already stores, converts them into a probability that an entry is correct, and propagates that probability through the containment hierarchy into a belief that a room contains a queried class. Four signals, each paired with the object-level error it indicates, are converted to probabilities at the logit scale learned by the vision-language model and combined in closed form with no additional perception or training. Objects sharing a detector and a vocabulary fail together, so the framework aggregates them in the fully correlated limit, where an aggregation under independence would treat one repeated error as repeated evidence. Evaluated on HM3DSem against a state-of-the-art 3DSG system, the framework improves object retrieval and lowers the error of the room-level assertions of the graph it reads.
Chinese Translation
开放词汇的3D场景图（3DSGs）将每个对象节点嵌入到视觉-语言嵌入中，但它们将每个条目记录为同样确定，因此在查询地图时，机器人无法判断哪些条目是不可靠的。语义不确定性的估计可以提供这种区分，但它们需要对模型进行重复采样、训练或使用保留标签，而这些在查询时的部署系统中都不可用。我们提出了一个框架，该框架利用检测器的置信度和3DSG已经存储的嵌入，将其转换为条目正确性的概率，并通过包含层次传播该概率，形成一个信念，即某个房间包含查询的类别。四个信号，每个信号与其指示的对象级错误配对，在视觉-语言模型学习的logit尺度上转换为概率，并以封闭形式结合，无需额外的感知或训练。共享检测器和词汇的对象一起失败，因此该框架在完全相关的极限中聚合它们，而在独立性下的聚合则会将一次重复错误视为重复证据。在HM3DSem上评估，与最先进的3DSG系统相比，该框架提高了对象检索的准确性，并降低了其读取的图的房间级断言的错误率。
cs.CV / 17 / 2609.22379
MarsRecon: Self-Supervised and Multimodal Surface Representations for Mars
MarsRecon：用于火星的自监督和多模态表面表示
Abstract
High-resolution orbital imagery offers a rich record of the Martian surface, but sparse geological labels limit supervised representation learning. We present MarsRecon, a geospatially aware pipeline for learning visual and multimodal representations from HiRISE observations of Olympus Mons. The pipeline calibrates NASA Planetary Data System products, extracts valid georeferenced patches, and trains a masked autoencoder on unlabeled imagery. Increasing input resolution and filtering invalid tokens reduced held-out reconstruction loss from 0.1751 to 0.1342 in the principal Stage A model series. We then freeze the visual encoder and align its features with observation text, coordinates, and local--global image context. The strongest current local-primary model achieves image-to-text recall@10 of 0.3787, text-to-image recall@10 of 0.9161, and local-to-global recall@10 of 0.4350 on the held-out test split. These results establish a working Mars-specific pretraining and retrieval pipeline; further crop-overlap controls and downstream geological evaluations are needed to assess the broader utility of its embeddings.
Chinese Translation
高分辨率的轨道影像为火星表面提供了丰富的记录，但稀疏的地质标签限制了监督表示学习。我们提出了MarsRecon，一个地理空间感知的管道，用于从火星高分辨率成像科学实验室（HiRISE）对奥林匹斯山的观测中学习视觉和多模态表示。该管道校准了美国国家航空航天局（NASA）行星数据系统产品，提取有效的地理参考补丁，并在未标记的影像上训练了一个掩码自编码器。通过增加输入分辨率和过滤无效标记，主要的A阶段模型系列的持出重建损失从0.1751降低到0.1342。随后，我们冻结视觉编码器，并将其特征与观测文本、坐标和局部-全局图像上下文对齐。当前最强的局部主模型在持出测试集上实现了图像到文本的召回率@10为0.3787，文本到图像的召回率@10为0.9161，以及局部到全局的召回率@10为0.4350。这些结果建立了一个有效的火星特定预训练和检索管道；进一步的裁剪重叠控制和下游地质评估是评估其嵌入更广泛实用性的必要条件。
cs.CV / 18 / 2609.22392
Style as Cover: Deep Image Steganography via Stylized Transmission
风格作为掩护：通过风格化传输实现深度图像隐写
Abstract
Image steganography hides secret message within normal images, with most existing works relying on cover-preserving transmission. However, such a paradigm becomes vulnerable once the original cover is exposed or can be reliably approximated. In this paper, we propose StyleStegaNet, a stylized image hiding framework that replaces cover matching with style-concealment transmission. Instead of transmitting a cover-like stego image, StyleStegaNet generates stylized stego images conditioned on publicly available style references, redefining steganography invisibility from cover-preserving concealment to behavior-level camouflage based on style transformation. Such a setting poses a substantial challenge to reliable secret recovery, since neural stylization can significantly alter the feature statistics exploited by deep hiding methods. To address this challenge, StyleStegaNet decouples the overall task into four coordinated stages: stego generation, stylized transmission, structure-preserving reconstruction, and secret recovery. Moreover, StyleStegaNet is optimized with a progressive three-stage training strategy, in which wavelet-domain constraints and perceptual supervision guide the recoverable information toward structural representations. We further provide an analysis showing that secret recoverability is largely restricted to the normalized structural subspace, offering a mechanistic explanation for why directly stylized baselines fail and why a reconstruction-guided recovery path is necessary. Extensive experiments on DIV2K and MS-COCO datasets demonstrate the effectiveness of StyleStegaNet. And few-shot image steganalysis with two deep detectors further shows detection accuracy near random guessing, approximately 51\%.
Chinese Translation
图像隐写技术在普通图像中隐藏秘密信息，现有大多数研究依赖于保持掩护的传输。然而，一旦原始掩护被暴露或能够被可靠地近似，这种范式就变得脆弱。在本文中，我们提出了StyleStegaNet，一种风格化图像隐藏框架，它用风格隐蔽传输替代了掩护匹配。StyleStegaNet不是传输类似掩护的隐写图像，而是基于公开可用的风格参考生成风格化的隐写图像，从而将隐写的隐蔽性重新定义为基于风格转化的行为级伪装，而非保持掩护的隐蔽。这样的设置对可靠的秘密恢复提出了重大挑战，因为神经风格化可能显著改变深度隐藏方法所利用的特征统计。为了解决这一挑战，StyleStegaNet将整体任务解耦为四个协调阶段：隐写生成、风格化传输、结构保持重建和秘密恢复。此外，StyleStegaNet采用渐进式三阶段训练策略进行优化，其中小波域约束和感知监督引导可恢复信息朝向结构表示。我们进一步提供了分析，表明秘密的可恢复性在很大程度上受到归一化结构子空间的限制，从而为直接风格化基线失败的原因以及为何需要重建引导的恢复路径提供了机制性解释。在DIV2K和MS-COCO数据集上的大量实验证明了StyleStegaNet的有效性。而使用两个深度检测器进行的少量样本图像隐写分析进一步显示检测准确率接近随机猜测，约为51%。
cs.CV / 19 / 2609.22479
Spatiotemporal Flux Probing for Single-Photon Videography
单光子视频摄影的时空通量探测
Abstract
We address the problem of recovering high-speed videos from dynamic scenes under extreme photon sparsity. Existing methods rely on aggregating photon detections in local spatiotemporal windows to improve signal-to-noise ratio; however, this local grouping discards global structure and fails in low-light regimes where photon detections are sparse in space and time. In this work, we show that the information needed to recover both motion and illumination is encoded in correlations over the full space-time pattern of photon arrivals. Building on this insight, we develop a spatiotemporal flux probing theory and an algorithm that estimates the Fourier coefficients of the underlying intensity directly from the photon stream. We demonstrate that our approach (1) recovers fast motion and temporal illumination dynamics with substantially fewer photons than prior methods, (2) enables velocity-selective videography that automatically refocuses video onto specific detected motions, and (3) generalizes across sensing modalities including single-photon, event, and spike cameras.
Chinese Translation
我们解决了在极端光子稀疏条件下从动态场景中恢复高速视频的问题。现有方法依赖于在局部时空窗口中聚合光子探测以提高信噪比；然而，这种局部分组忽略了全局结构，并且在光子探测在空间和时间上稀疏的低光环境中效果不佳。在本研究中，我们展示了恢复运动和照明所需的信息编码在光子到达的全时空模式的相关性中。基于这一见解，我们开发了一种时空通量探测理论和一种算法，该算法直接从光子流中估计基础强度的傅里叶系数。我们证明了我们的方法（1）能够以显著少于先前方法的光子数量恢复快速运动和时间照明动态，（2）实现了速度选择性视频摄影，自动将视频重新聚焦到特定的探测运动上，以及（3）在包括单光子、事件和尖峰相机等传感模式之间具有广泛的适用性。
cs.CV / 20 / 2609.22500
Event-Frame Fusion for Inter-Frame Segmentation via Event-Guided Motion
基于事件引导运动的帧间分割事件框架融合
Abstract
Autonomous navigation requires precise and efficient semantic segmentation, yet existing frame-based approaches remain limited by motion blur, glare, latency, and the low temporal resolution (20-30 FPS) of conventional cameras, which leads to information loss between frames. Event cameras have emerged as an alternative sensing modality, capturing intensity changes asynchronously with high temporal resolution, high dynamic range, and sparse outputs. However, event-based algorithms still fall short of frame-based ones in accuracy, as most segmentation methods are designed for dense frame data. To overcome these limitations, we propose a hybrid vision architecture that combines conventional frame-based and event-based cameras. The system integrates two complementary components: (1) a compact Spiking Neural Network (SNN) with 42k parameters for motion estimation, and (2) a lightweight event-driven SNN with 0.84M parameters for frame-based semantic segmentation, which interpolates motion between frames to refine segmentation results. By predicting inter-frame segmentations, the framework achieves segmentation rates of up to 500 Hz with an energy consumption below 1.87 mJ per inference, while maintaining real-time GPU execution at frequencies up to 200 Hz. Additionally, our approach compensates for information loss in frames affected by blur or overexposure, enabling more robust perception in challenging conditions.
Chinese Translation
自主导航需要精确高效的语义分割，然而现有的基于帧的方法受到运动模糊、眩光、延迟以及传统相机低时间分辨率（20-30 FPS）的限制，这导致帧间信息丢失。事件相机作为一种替代传感方式，能够以高时间分辨率、高动态范围和稀疏输出异步捕捉强度变化。然而，基于事件的算法在准确性上仍不及基于帧的算法，因为大多数分割方法是为稠密帧数据设计的。为克服这些限制，我们提出了一种混合视觉架构，结合了传统的基于帧和基于事件的相机。该系统集成了两个互补组件：（1）一个具有42k参数的紧凑型脉冲神经网络（SNN）用于运动估计，以及（2）一个具有0.84M参数的轻量级事件驱动SNN，用于基于帧的语义分割，该网络通过插值帧间运动来细化分割结果。通过预测帧间分割，该框架实现了高达500 Hz的分割速率，推理能耗低于1.87 mJ，同时在高达200 Hz的频率下保持实时GPU执行。此外，我们的方法补偿了因模糊或过曝而导致的帧信息丢失，从而在挑战性条件下实现更强的感知能力。
cs.CV / 21 / 2609.22506
Rethinking Vision Architectures with Gated Linear Attention and KAN
重新思考带有门控线性注意力和KAN的视觉架构
Abstract
Vision Transformers allocate most parameters to multi-layer perceptrons (MLPs) for channel mixing, while token interactions usually rely on quadratic multi-head self-attention (MHSA). Linear attention reduces sequence complexity to O(N), but remains coupled with the same fixed-activation MLP as softmax Transformers. Kolmogorov-Arnold Networks (KANs) instead place learnable univariate maps on edges, yet prior vision KANs keep MHSA or omit attention entirely. We introduce LKAT (Linear Kolmogorov-Arnold Transformer), an isotropic ViT encoder that couples chunkwise Gated Linear Attention (GLA) with a two-layer KAN feed-forward, and we provide an I/O-aware fused RBF-KAN kernel for the radial-basis grid maps. Under a shared DeiT-style recipe we compare LKAT with ViT, ViT-5, and MLP-Mixer. LKAT-B exceeds ViT-B/16, ViT-5-B, and Mixer-B/16 on ImageNet-100. Tiny/Small/Base LKAT variants scale consistently on CIFAR-10/100, and ImageNet-100 pretraining transfers to CIFAR fine-tuning. The results support gated linear attention and KAN-based radial-basis functions as complementary inductive biases for mid-scale visual representation learning.
Chinese Translation
视觉变换器将大部分参数分配给多层感知机（MLP）以进行通道混合，而令牌交互通常依赖于二次多头自注意力（MHSA）。线性注意力将序列复杂度降低到 O(N)，但仍与软最大变换器的固定激活 MLP 相关联。科尔莫戈罗夫-阿诺德网络（KAN）则将可学习的一元映射放置在边缘上，但之前的视觉 KAN 保留了 MHSA 或完全省略了注意力。我们引入了 LKAT（线性科尔莫戈罗夫-阿诺德变换器），这是一种各向同性的 ViT 编码器，将分块门控线性注意力（GLA）与两层 KAN 前馈网络相结合，并为径向基网格映射提供了一个 I/O 感知的融合 RBF-KAN 核心。在共享的 DeiT 风格配方下，我们将 LKAT 与 ViT、ViT-5 和 MLP-Mixer 进行了比较。LKAT-B 在 ImageNet-100 上超过了 ViT-B/16、ViT-5-B 和 Mixer-B/16。Tiny/Small/Base LKAT 变体在 CIFAR-10/100 上表现出一致的扩展性，ImageNet-100 的预训练可迁移到 CIFAR 的微调。结果支持门控线性注意力和基于 KAN 的径向基函数作为中等规模视觉表征学习的互补归纳偏置。
cs.CV / 22 / 2609.22562
AdaMerge: Tuning-Free Patch Compression for Multi-Vector Visual Document Retrieval
AdaMerge：无调优的多向量视觉文档检索补丁压缩
Abstract
Multi-vector visual document retrieval (VDR) models such as ColPali and ColNomic achieve strong accuracy by representing each document with hundreds to thousands of patch-level embeddings, at substantial storage and latency cost. Existing compression methods either prune unimportant patches or merge similar ones into clusters; the recent state-of-the-art merging method Prune-then-Merge (PtM) consistently outperforms pruning-only baselines at high compression, but requires a per-dataset cluster budget m to be tuned by grid search. We observe that the merge-cosine sequence produced by hierarchical clustering exhibits a sharp cliff separating mergeable redundancy from salient signal, and that the location of this cliff is concentrated in a narrow band across more than 11,000 documents from 14 datasets. This suggests the merge boundary can be detected per document rather than tuned per dataset. Building on this observation, we propose AdaMerge, a plug-and-play compression method that (i) detects each document's own cliff via gap analysis on the merge-cosine trajectory, and (ii) builds attention-weighted cluster centroids to preserve salient signal. On the long-document benchmark ViDoRe-V2 (4 datasets, two backbones), AdaMerge significantly outperforms tuned PtM across the operating range (p < 10^-4); on the short-document benchmark ViDoRe-V1 (10 datasets, two backbones), where all merging methods are already near-lossless, AdaMerge matches tuned PtM without any per-dataset tuning. AdaMerge adds only about 10 ms per document and exposes a single global hyperparameter shared across all datasets and backbones.
Chinese Translation
多向量视觉文档检索（VDR）模型，如ColPali和ColNomic，通过用数百到数千个补丁级嵌入表示每个文档，取得了强大的准确性，但在存储和延迟成本上付出了相当大的代价。现有的压缩方法要么修剪不重要的补丁，要么将相似的补丁合并成簇；最近的最先进的合并方法Prune-then-Merge（PtM）在高压缩率下始终优于仅修剪的基线，但需要通过网格搜索来调优每个数据集的簇预算m。我们观察到，层次聚类产生的合并余弦序列表现出一个明显的悬崖，将可合并的冗余与显著信号分开，并且这个悬崖的位置在来自14个数据集的超过11,000个文档中集中在一个狭窄的带状区域。这表明合并边界可以针对每个文档进行检测，而不是针对每个数据集进行调优。在这一观察的基础上，我们提出了AdaMerge，这是一种即插即用的压缩方法，它（i）通过对合并余弦轨迹进行间隙分析来检测每个文档自己的悬崖，以及（ii）构建加权注意力的簇中心以保留显著信号。在长文档基准ViDoRe-V2（4个数据集，两个骨干网络）上，AdaMerge在操作范围内显著优于调优后的PtM（p < 10^-4）；在短文档基准ViDoRe-V1（10个数据集，两个骨干网络）上，所有合并方法已经接近无损，AdaMerge在没有任何每个数据集调优的情况下与调优后的PtM相匹配。AdaMerge每个文档仅增加约10毫秒，并且暴露出一个在所有数据集和骨干网络中共享的单一全局超参数。
cs.CV / 23 / 2609.22582
Beyond the Leaderboard: Counterfactual Diagnosis of End-to-End and VLA Driving Policies Under Domain Shift
超越排行榜：在领域转移下对端到端和视觉-语言-行动（VLA）驾驶策略的反事实诊断
Abstract
End-to-end and vision-language-action (VLA) driving policies are compared by leaderboard rank, but a rank reports an outcome, not the behaviour behind it, so it predicts poorly how a policy will behave at a new site. On six released policies, rank on nuScenes open-loop error or on NAVSIM's leaderboard does not carry over to scenes with a pedestrian near the ego corridor at a new site. We propose a counterfactual check-up: a few hundred real frames, each edited two ways (pedestrian removed, or re-lit by a night-style perturbation), every edit verified by an independent detector, and the change in the planned trajectory read as a diagnosis rather than a score. From these edits two causal axes are read, and five exams built on them separate what a score merges: how far the policy plans to drive, whether seeing the pedestrian buys safety, whether that response scales with danger, whether the plan moves when nothing requires it, and how much an irrelevant lighting change moves it. On 246 NAVSIM near-pedestrian scenes, in the cells where the pedestrian lies on the planned path only 1.9% of responses are genuine avoidance, and under our open-loop protocol the median clearance change is at most 0.03 m and the median change in planned distance at most 0.08 m for every policy. In a pre-registered test from left- to right-hand drive, the exposure and specificity orderings, the lighting verdict and the collision outcome transfer, while point values and the hazard-sensitivity verdict do not. Read as a selection report, the profiles say which policy is safe because it plans short, which covers a human-like distance without yielding, and which is unsteady under a change that requires no reaction, and they price each verdict: most settle within a few dozen frames, hazard sensitivity needs hundreds. Code and edited frames will be released.
Chinese Translation
端到端和视觉-语言-行动（VLA）驾驶策略通过排行榜排名进行比较，但排名仅报告结果，而非背后的行为，因此在新场景中预测策略的表现较差。在六个发布的策略中，nuScenes的开放环错误排名或NAVSIM排行榜的排名在新场景中与行人靠近自我车道的场景并不一致。我们提出了一种反事实检查方法：几百帧真实画面，每帧以两种方式编辑（移除行人或通过夜间风格扰动重新照明），每次编辑都由独立检测器验证，计划轨迹的变化被视为诊断而非评分。通过这些编辑，可以提取两个因果轴，并基于这两个轴构建五个考试，区分评分所合并的内容：策略计划行驶的距离、是否看到行人能提高安全性、该反应是否随危险程度而变化、在没有任何要求时计划是否会变动，以及不相关的照明变化对其影响的程度。在246个NAVSIM近行人场景中，当行人位于计划路径上时，仅有1.9%的反应是真正的避让，而在我们的开放环协议下，所有策略的中位清除变化最多为0.03米，计划距离的中位变化最多为0.08米。在一个预注册的从左侧驾驶到右侧驾驶的测试中，曝光和特异性排序、照明判决和碰撞结果能够转移，而点值和危险敏感性判决则无法转移。作为选择报告阅读时，分析结果表明哪种策略是安全的，因为它计划的距离较短，哪种策略在不让步的情况下覆盖了人类般的距离，以及哪种策略在不需要反应的变化下表现不稳定，并对每个判决进行定价：大多数在几十帧内稳定，而危险敏感性需要数百帧。代码和编辑后的帧将会发布。
cs.CV / 24 / 2609.22588
Seeing is not Enough: Vision-Language Models Perceive Evidence but Fail to Act
视觉并不足够：视觉-语言模型感知证据但未能采取行动
Abstract
Vision-language models (VLMs) perform strongly on visual question answering benchmarks, yet often make decisions that contradict visual evidence they have already identified correctly. We distinguish perceptual failure, where relevant evidence is not recognized, from process failure, where recognized evidence fails to constrain the final decision. We introduce VPAC-Bench, a benchmark spanning nine real-image process families, with each image annotated by its current activity stage and nearby stage transition. We also propose State-Relevance-Target (SRT), a family of structured process-prior interventions that requires models to connect visible evidence to the relevant process state before answering. Across multiple VLMs, process failure is widespread: models that correctly enumerate visual candidates still over-commit to a single answer in more than 95% of ambiguous cases. An explicit process-structured intervention reduces this rate to below 13% without degrading performance on unambiguous cases. However, the transfer of process priors is model-dependent, and generic SRT does not consistently outperform strong chain-of-thought baselines. When the relevant stage transition is known, boundary-aligned SRT substantially outperforms generic process prompting and all tested chain-of-thought baselines across assembly, physical state transition, navigation and traffic, and object-use affordance tasks. These results show that process priors are most useful when aligned with the scene's specific decision boundary, motivating boundary-aware prior selection for process-grounded visual reasoning.
Chinese Translation
视觉-语言模型（VLMs）在视觉问答基准测试中表现出色，但常常做出与其已正确识别的视觉证据相矛盾的决策。我们将感知失败（即未能识别相关证据）与过程失败（即已识别的证据未能约束最终决策）区分开来。我们引入了VPAC-Bench，这是一个涵盖九个真实图像处理类别的基准，每个图像都标注了其当前活动阶段和附近的阶段转换。我们还提出了状态相关目标（State-Relevance-Target, SRT），这是一系列结构化的过程先验干预，要求模型在回答之前将可见证据与相关的过程状态联系起来。在多个VLM中，过程失败普遍存在：即使模型正确列举了视觉候选项，在超过95%的模糊案例中仍然过度依赖单一答案。显式的过程结构干预将这一比例降低到13%以下，而不降低在明确案例中的表现。然而，过程先验的转移依赖于模型，通用的SRT并未始终优于强大的思维链基线。当相关的阶段转换已知时，边界对齐的SRT在装配、物理状态转换、导航与交通以及物体使用可供性任务中显著优于通用过程提示和所有测试的思维链基线。这些结果表明，当过程先验与场景的特定决策边界对齐时，其最为有效，激励了基于边界的过程选择以进行过程基础的视觉推理。
cs.CV / 25 / 2609.22631
X-Beat: An Explainable Framework for ECG Image Classification
X-Beat：一种可解释的心电图图像分类框架
Abstract
Accurate automated interpretation of electrocardio- grams (ECGs) is essential for early detection of cardiac condi- tions such as myocardial infarction and rhythm abnormalities. However, many high-performing deep learning models remain difficult to deploy in clinical settings due to limited transparency and lack of reliability validation. In this work, we present X- Beat, an explainable and reliability-aware benchmark framework for ECG image classification designed to support trustworthy AI systems in healthcare. The proposed framework combines transfer learning with post-hoc explainability and systematic reliability evaluation across four cardiac classes: Abnormal Heartbeat, History of Myocardial Infarction, Myocardial In- farction, and Normal Heartbeat. Multiple ImageNet-pretrained CNN backbones, including EfficientNet-B0, ResNet-50, DenseNet- 121, and MobileNetV3-Large, are evaluated under a unified training protocol. Beyond standard performance metrics, we incorporate Grad-CAM-based visual explanations together with additional analyses, including explanation stability, regional sen- sitivity, and confidence-based reliability assessment, to examine whether model predictions are supported by clinically meaningful evidence. Experimental results show that ResNet-50 achieves the best performance, reaching 91.94% accuracy and a macro F1- score of 0.9098, with strong class separability (AUC up to 0.995). Explanation analyses indicate that the model primarily focuses on waveform-relevant regions, while reliability evaluation suggests that most incorrect predictions occur with lower confidence. Overall, this work provides a structured and reproducible bench- mark for evaluating both predictive performance and explanation reliability in ECG image classification, contributing toward the development of trustworthy and interpretable AI components for clinical decision support systems.
Chinese Translation
心电图（ECG）的准确自动解读对于早期检测心脏病状况（如心肌梗死和节律异常）至关重要。然而，由于缺乏透明度和可靠性验证，许多高性能的深度学习模型在临床环境中仍然难以部署。在本研究中，我们提出了X-Beat，一个可解释且关注可靠性的心电图图像分类基准框架，旨在支持医疗保健中的可信赖人工智能系统。该框架结合了迁移学习、后期可解释性和针对四种心脏类别的系统可靠性评估：异常心跳、心肌梗死历史、心肌梗死和正常心跳。我们在统一的训练协议下评估了多个基于ImageNet预训练的卷积神经网络（CNN）骨干网络，包括EfficientNet-B0、ResNet-50、DenseNet-121和MobileNetV3-Large。除了标准性能指标外，我们还结合了基于Grad-CAM的可视化解释以及其他分析，包括解释稳定性、区域敏感性和基于置信度的可靠性评估，以检查模型预测是否得到临床上有意义的证据支持。实验结果表明，ResNet-50达到了最佳性能，准确率为91.94%，宏观F1分数为0.9098，具有良好的类别可分性（AUC高达0.995）。解释分析表明，该模型主要关注与波形相关的区域，而可靠性评估则表明大多数错误预测发生在较低的置信度下。总体而言，本研究提供了一个结构化且可重复的基准，用于评估心电图图像分类中的预测性能和解释可靠性，为临床决策支持系统中可信赖和可解释的人工智能组件的发展做出了贡献。
cs.CV / 26 / 2609.22641
ConsistWorld: Evidence Routing for Consistent Multi-Agent World Models
ConsistWorld：一致性多智能体世界模型的证据路由
Abstract
Autoregressive video world models enable temporally coherent generation for a single observer. Extending them to multiple agents requires consistency across independently controlled views and temporal gaps under causal streaming. We present ConsistWorld, a multi-agent world model that generates camera-controlled video streams of a static scene from one shared image. We formulate consistency as routing evidence from committed multi-agent history and concurrently generated peer views to the tokens being generated. Pose Conditioned Memory Retrieval selects relevant historical observations from all agents, recovering evidence beyond the recent context window. Visibility-Gated Peer Sharing regulates current peer information according to estimated historical coverage and current-view overlap. Together, they determine which historical observations enter the context and where concurrent peer information contributes, supporting long-term recall and coordinated exploration. Both mechanisms use camera geometry and maintain a bounded active context for a fixed agent count and retrieval budget. Experiments on evidence sharing cases and video length and agent number generalizations show that ConsistWorld achieves a strong cross-time and cross-agent consistency while preserving competitive generation quality.
Chinese Translation
自回归视频世界模型能够为单一观察者生成时间一致的内容。将其扩展到多个智能体需要在独立控制的视角和因果流媒体下的时间间隙之间保持一致性。我们提出了ConsistWorld，这是一种多智能体世界模型，能够从一幅共享图像生成静态场景的摄像机控制视频流。我们将一致性定义为从已承诺的多智能体历史和同时生成的同行视角路由证据到正在生成的标记。姿态条件记忆检索（Pose Conditioned Memory Retrieval）从所有智能体中选择相关的历史观察，恢复超出最近上下文窗口的证据。可见性门控同行共享（Visibility-Gated Peer Sharing）根据估计的历史覆盖和当前视图重叠来调节当前的同行信息。两者共同决定哪些历史观察进入上下文，以及并发的同行信息如何贡献，支持长期回忆和协调探索。这两种机制利用摄像机几何并为固定的智能体数量和检索预算维持一个有界的活跃上下文。在证据共享案例以及视频长度和智能体数量的泛化实验中，ConsistWorld在保持竞争性生成质量的同时，实现了强大的跨时间和跨智能体的一致性。
cs.CV / 27 / 2609.22647
Math2Visual-X: A Modular Framework for Pedagogically Aligned Lower-Primary Math Visuals Generation
Math2Visual-X：一个模块化框架用于教育对齐的低年级数学视觉生成
Abstract
Visual representations can help lower-primary learners understand Math Word Problems, but generating classroom-usable visuals remains difficult. Existing symbolic systems are controllable but limited in coverage, while end-to-end text-to-image systems often fail to satisfy exact mathematical constraints. This paper presents a symbolic visual generation framework for lower-primary MWP generation with broader problem coverage and more scalable asset generation. The framework includes an LLM-based routing layer, three worksheet-oriented generation modules, and two fallback mechanisms for open-world SVG asset acquisition. A human evaluation comparing Math2Visual-X with Stable Diffusion XL, Nano Banana, and GPT Image showed that the proposed method achieved the strongest overall performance. The results indicate that the framework offers a scalable and pedagogically grounded approach for automatic MWP visual generation.
Chinese Translation
视觉表现可以帮助低年级学习者理解数学文字问题，但生成适合课堂使用的视觉材料仍然困难。现有的符号系统虽然可控，但覆盖范围有限，而端到端的文本到图像系统往往无法满足精确的数学约束。本文提出了一种符号视觉生成框架，用于低年级数学文字问题的生成，具有更广泛的问题覆盖和更可扩展的资产生成能力。该框架包括一个基于大型语言模型（LLM）的路由层、三个面向工作表的生成模块，以及两个用于开放世界SVG资产获取的后备机制。与Stable Diffusion XL、Nano Banana和GPT Image的比较评估表明，所提出的方法在整体性能上表现最强。结果表明，该框架为自动化数学文字问题视觉生成提供了一种可扩展且教育基础扎实的方法。
cs.CV / 28 / 2609.22687
PanoSeg3R: Feed-Forward 3D Semantic Segmentation for Panoramic Images with an Automatic Data Curation Pipeline
PanoSeg3R：用于全景图像的前馈式3D语义分割与自动数据整理管道
Abstract
We present PanoSeg3R, a feed-forward framework for 3D panoramic semantic segmentation. Unlike existing methods designed for perspective inputs, PanoSeg3R jointly predicts 3D geometry and multi-view semantic segmentation in one single forward pass. Built upon a pretrained reconstruction backbone that supports panoramic images, our approach extends feed-forward 3D reconstruction with a query-based mask decoder. Furthermore, we introduce an automatic panorama data curation pipeline that leverages the complementary strengths of off-the-shelf foundation models to generate reliable pseudo semantic annotations, substantially expanding the training data and improving zero-shot generalization. PanoSeg3R achieves state-of-the-art performance on panoramic 3D semantic segmentation, improving 3D mIoU by up to 16.02 on ScanNet++, while the curated training data further improves zero-shot performance by up to 4.26 and 43.28 mIoU on Stanford2D3D and ToF-360, respectively. Website: https://harryyoon777.github.io/PanoSeg3R/
Chinese Translation
我们提出了PanoSeg3R，这是一种用于3D全景语义分割的前馈框架。与现有针对透视输入的方法不同，PanoSeg3R在一次前向传递中联合预测3D几何和多视角语义分割。我们的方案基于一个预训练的重建主干网络，该网络支持全景图像，并通过基于查询的掩码解码器扩展了前馈式3D重建。此外，我们引入了一种自动全景数据整理管道，利用现成基础模型的互补优势生成可靠的伪语义注释，显著扩展了训练数据并改善了零样本泛化能力。PanoSeg3R在全景3D语义分割任务上实现了最先进的性能，在ScanNet++上将3D mIoU提高了多达16.02，而整理后的训练数据进一步在Stanford2D3D和ToF-360上分别提高了4.26和43.28的零样本性能和mIoU。网站：https://harryyoon777.github.io/PanoSeg3R/
cs.CV / 29 / 2609.22688
Vision2CAD: A Visual Agent Harness for Explicit Geometry Referencing and Localization in Parametric CAD Modeling
Vision2CAD：一种用于参数化CAD建模中显式几何引用和定位的视觉代理工具
Abstract
Generating parametric CAD models requires accurate geometry and stable feature dependencies. Existing methods face challenges in selecting geometric references, interpreting sketch-plane local coordinates, and establishing sketch constraints to projected external geometry. We present Vision2CAD, a visual agent harness that combines vision-language model (VLM) reasoning with deterministic CAD kernel operations. An ID-based interface supports explicit geometry selection, a local-coordinate bridge converts view coordinates into sketch coordinates, and projected-edge localization supports external sketch constraints. These mechanisms establish feature dependencies within the supported modeling operations and constraint types. We also introduce the Geometry Explicit Reference Dataset (GERD), which aligned commands, geometry states and IDs at every modeling step. On GERD-EVL and a DeepCAD test subset, Vision2CAD improves mIoU by 11.1\% and 5.6\% and reduces Chamfer distance by 17.3\% and 41.8\%, respectively. Parameter-editing experiments and ablation studies further proved the preservation of parametric dependencies.
Chinese Translation
生成参数化CAD模型需要准确的几何形状和稳定的特征依赖关系。现有方法在选择几何参考、解释草图平面局部坐标以及建立草图约束与投影外部几何之间的关系时面临挑战。我们提出了Vision2CAD，这是一种结合了视觉-语言模型（VLM）推理与确定性CAD内核操作的视觉代理工具。基于ID的接口支持显式几何选择，局部坐标桥将视图坐标转换为草图坐标，投影边缘定位支持外部草图约束。这些机制在支持的建模操作和约束类型中建立了特征依赖关系。我们还引入了几何显式参考数据集（GERD），该数据集在每个建模步骤中对齐了命令、几何状态和ID。在GERD-EVL和DeepCAD测试子集上，Vision2CAD分别提高了mIoU 11.1 ext{%}和5.6 ext{%}，并减少了Chamfer距离17.3 ext{%}和41.8 ext{%}。参数编辑实验和消融研究进一步证明了参数依赖关系的保持。
cs.CV / 30 / 2609.22706
DOA-SORT: Directional Occlusion-Aware Multi-Object Tracking with Distributional Observations
DOA-SORT：基于方向性遮挡感知的分布式观测多目标跟踪方法
Abstract
Identity association in multi-object tracking (MOT) is vulnerable to partial occlusion, truncated detections, and fluctuating confidence scores. Existing motion-dominant trackers commonly represent occlusion as a scalar penalty. This treatment misses the directional observation bias caused by occlusion: left, right, top, and bottom occlusions distort the location and shape of a detection in different ways. We propose \ours{} (Directional Occlusion-Aware SORT), an online and training-free tracker that models these biases explicitly. First, it infers a soft front--back ordering from box overlap and relative bottom positions, and estimates directional occlusion coverage and depth. It then constructs a mixture of one clean and four directional occlusion observation components. The model uses a five-dimensional observation comprising box center, area, confidence, and aspect ratio, and adapts observation noise to predicted occlusion and detection confidence. The directional mixture likelihood is used in high-confidence association, low-confidence association, and track recovery; ambiguity penalties and local order-consistency swaps further reduce identity errors among nearby objects. On the DanceTrack validation split, \ours{} improves HOTA from 63.00 to 66.34, AssA from 45.10 to 49.57, and IDF1 from 62.19 to 65.28 over OA-SORT with the same detector and evaluation protocol. The gains are concentrated in association quality while detection accuracy remains stable. Additional local evaluations on MOT17 and MOT20 train splits characterize cross-dataset behavior under the same no-ReID tracking protocol.
Chinese Translation
多目标跟踪（MOT）中的身份关联易受部分遮挡、截断检测和置信度波动的影响。现有以运动为主的跟踪器通常将遮挡表示为一个标量惩罚，这种处理忽略了遮挡引起的方向性观测偏差：左侧、右侧、顶部和底部的遮挡会以不同方式扭曲检测框的位置和形状。我们提出了DOA-SORT（Directional Occlusion-Aware SORT），一种在线且无需训练的跟踪器，能够显式建模这些偏差。首先，它通过框重叠和相对底部位置推断软前后顺序，并估计方向性遮挡覆盖率和深度。随后构建由一个干净观测和四个方向性遮挡观测组成的混合模型。该模型使用包含框中心、面积、置信度和长宽比的五维观测，并根据预测的遮挡和检测置信度自适应调整观测噪声。方向性混合似然用于高置信度关联、低置信度关联及轨迹恢复；模糊惩罚和局部顺序一致性交换进一步减少邻近目标间的身份错误。在DanceTrack验证集上，DOA-SORT相较于使用相同检测器和评估协议的OA-SORT，将HOTA从63.00提升至66.34，AssA从45.10提升至49.57，IDF1从62.19提升至65.28。性能提升主要集中在关联质量上，而检测准确率保持稳定。对MOT17和MOT20训练集的额外本地评估则在相同无ReID跟踪协议下刻画了跨数据集的表现。
cs.CV / 31 / 2609.22716
ZIL: Zero-shot Image-to-LiDAR Registration
ZIL：零样本图像与激光雷达配准
Abstract
Image-to-LiDAR registration estimates the camera pose of an image with respect to a LiDAR point cloud. It has diverse applications in autonomous driving, robot navigation etc. However, state-of-the-art (SOTA) methods still 1) mostly assume same-frame inputs, struggling with the image and point cloud from distant frames; 2) rely on domain-specific training, failing to generalize to unseen scenarios. We propose ZIL, the first foundation model for zero-shot non-synchronized image-to-LiDAR registration. ZIL encodes the input image and point cloud with the Vision and Point Transformers. In addition to regressing the relative pose, ZIL also learns to predict 3D coordinates, which substantially improves the pose accuracy without additional annotations. Interestingly, naive mix-data training cannot enable zero-shot generalization, which requires normalization on both camera intrinsics and the LiDAR vertical-axis origin. Trained on 7 public datasets with 1.4M LiDAR frames, ZIL consistently and significantly outperforms previous SOTA with a single model across 5 in-domain and zero-shot benchmarks, reducing the translation and rotation errors by up to 87% and 76% (shown in Fig. 1). Code and models are available at https://github.com/ZijunLi7/ZIL.
Chinese Translation
图像与激光雷达配准估计图像相对于激光雷达点云的相机姿态。它在自动驾驶、机器人导航等领域有着广泛的应用。然而，当前的最先进方法（SOTA）仍然存在以下问题：1）大多数假设输入来自同一帧，难以处理来自不同帧的图像和点云；2）依赖于特定领域的训练，无法在未见过的场景中进行推广。我们提出了ZIL，这是首个用于零样本非同步图像与激光雷达配准的基础模型。ZIL使用视觉和点云变换器对输入图像和点云进行编码。除了回归相对姿态外，ZIL还学习预测3D坐标，这在没有额外标注的情况下显著提高了姿态精度。有趣的是，简单的混合数据训练无法实现零样本泛化，这需要对相机内参和激光雷达垂直轴原点进行归一化。在7个公共数据集上训练，包含140万帧激光雷达数据，ZIL在5个领域内和零样本基准测试中，始终显著超越之前的最先进方法，使用单一模型将平移和旋转误差分别降低了高达87%和76%（如图1所示）。代码和模型可在 https://github.com/ZijunLi7/ZIL 获取。
cs.CV / 32 / 2609.22750
Towards Robust Classroom Attendance: A Comprehensive Evaluation of Face Detection and Recognition Models
朝向稳健的课堂出勤：面部检测与识别模型的综合评估
Abstract
Manual attendance methods, such as paper or register-based systems, take a lot of time, can lead to errors, and are easy to falsify. Face recognition is more reliable, but it frequently struggles in classrooms because lighting and other conditions can vary. Face recognition datasets are designed for regulated environments and do not capture the actual challenges found in classrooms. To address this, a new face detection and recognition dataset, the Visage Face dataset, comprising 16,234 face samples, is proposed for the task of face detection and recognition. The photos are taken from different angles and under varying lighting conditions, with students showing a range of expressions, and some faces partly covered to reflect real-life situations. A YOLO-based system is used to detect faces and tested seven advanced face recognition models with thirteen configurations: LVFace, QCFace, FaceLiVTv2, TopoFR, EdgeFace, TransFace, and GhostFaceNets. Of these, FaceLiVTv2-M performed best, with 99.75% Top-1/Top-5 accuracy and an inference time of 6.459 ms. These results show that the Visage Face Dataset is a realistic and challenging benchmark for face recognition in classroom attendance.
Chinese Translation
手动出勤方法，如纸质或登记系统，耗时较长，容易出错且易于伪造。面部识别更为可靠，但在课堂上常常面临挑战，因为光照和其他条件可能会有所不同。现有的面部识别数据集通常设计用于受控环境，未能捕捉到课堂中实际面临的挑战。为此，提出了一种新的面部检测与识别数据集——Visage Face 数据集，包含16,234个面部样本，专门用于面部检测和识别任务。这些照片从不同角度拍摄，并在不同光照条件下进行，学生们展现了多种表情，部分面孔被遮挡，以反映现实生活中的情况。采用基于YOLO的系统进行面部检测，并测试了七种先进的面部识别模型及其十三种配置：LVFace、QCFace、FaceLiVTv2、TopoFR、EdgeFace、TransFace和GhostFaceNets。其中，FaceLiVTv2-M表现最佳，Top-1/Top-5准确率达到99.75%，推理时间为6.459毫秒。这些结果表明，Visage Face 数据集为课堂出勤中的面部识别提供了一个真实且具有挑战性的基准。
cs.CV / 33 / 2609.22762
DriveReferee: Geometric Safety Verdicts Need Not Be Learned for Driving World-Action Models
DriveReferee：几何安全裁决不需要为驾驶世界行动模型学习
Abstract
Generative world-action models (WAMs) jointly generate future video and vehicle actions, while their action branches remain primarily optimized by expert imitation. Yet imitation provides no explicit closed-loop geometric verdict for generated trajectories, making verification important during both training and deployment. Closed-loop evaluators can check collision and drivable-area violations, but require privileged scene state unavailable at deployment. Existing approaches often close this gap by learning a verifier from sensor features. For these geometric checks, the rule itself is explicit. For example, collision is determined by whether the rolled-out ego footprint overlaps occupied vehicle space. What is unavailable at deployment is the scene state needed to apply the rule. We introduce DriveReferee, which uses a learned geometry readout to predict the scene representation from camera observations and executes the geometric safety rule directly rather than learning it. The resulting analytic referee evaluates collision and drivable-area safety from a scene state and candidate trajectory. During training, it scores self-sampled trajectories on ground-truth state and distills the resulting preferences into the WAM policy. At deployment, the same referee evaluates generated trajectories on this predicted state and selects a safer alternative when needed. The analytic referee requires no verdict-specific training, and its decisions follow an explicit geometric rule. Under matched candidates and inference budgets, it matches or outperforms all learned-verifier and heuristic baselines. Given the same predicted state and trajectory, learning the verdict provides no measurable downstream gain despite requiring tens of thousands of evaluator-labeled training examples. On the full NAVSIM navtest, DriveReferee reaches 92.02 PDMS with single-camera visual input and no external training data.
Chinese Translation
生成性世界行动模型（WAMs）共同生成未来的视频和车辆动作，而它们的动作分支主要通过专家模仿进行优化。然而，模仿并未为生成的轨迹提供明确的闭环几何裁决，因此在训练和部署过程中进行验证显得尤为重要。闭环评估器可以检查碰撞和可行驶区域的违规情况，但需要在部署时无法获得的特权场景状态。现有方法通常通过从传感器特征中学习验证器来弥补这一差距。对于这些几何检查，规则本身是明确的。例如，碰撞的判断依据是展开的自我足迹是否与占用的车辆空间重叠。在部署时缺失的是应用该规则所需的场景状态。我们引入了DriveReferee，它利用学习到的几何读出从相机观察中预测场景表示，并直接执行几何安全规则，而不是学习该规则。最终的分析性裁判从场景状态和候选轨迹中评估碰撞和可行驶区域的安全性。在训练过程中，它在真实状态上对自采样轨迹进行评分，并将结果偏好提炼到WAM策略中。在部署时，同样的裁判在这一预测状态上评估生成的轨迹，并在必要时选择更安全的替代方案。分析性裁判不需要特定裁决的训练，其决策遵循明确的几何规则。在匹配的候选者和推理预算下，它的表现与所有学习验证器和启发式基线相当或更优。在相同的预测状态和轨迹下，学习裁决并未提供可测量的下游收益，尽管需要数万个评估器标记的训练样本。在完整的NAVSIM导航测试中，DriveReferee在单摄像头视觉输入和没有外部训练数据的情况下达到了92.02 PDMS。
cs.CV / 34 / 2609.22788
Human-Level Accuracy, Non-Human Strategies: Revealing Model-Human Divergence in Video Physical Reasoning
人类水平的准确性与非人类策略：揭示视频物理推理中的模型与人类的差异
Abstract
Video foundation models now reach human-level accuracy on physical-reasoning benchmarks, yet such tasks require predicting unobserved physical outcomes. Do these models perform human-like forward simulation, or do they exploit statistical regularities in visible scenes? Accuracy alone cannot distinguish these strategies. We introduce a distributional evaluation framework that treats model seeds and human raters as populations, enabling comparison of consensus, uncertainty, and strategy. On the Physion benchmark, we evaluate three ViT-L architectures (V-JEPA2, VideoMAEv2, DINOv2). V-JEPA2 narrows the accuracy gap to ~1 percentage point (73.2% vs. 74.2%), yet model-human disagreement reaches 26.4%, far exceeding human-human disagreement (4.8%), with substantially lower agreement (kappa ~ 0.48 vs. 0.91). The divergence follows forward-simulation demands: models outperform humans on geometric reasoning (linking, +11.8 pp) but underperform on gravitational dynamics (rolling, -11.8 pp) and causal chains (dominoes, -10.5 pp). Strategy fingerprinting confirms all three architectures share non-human strategies while none aligns with humans. Attribution analysis suggests that unobservable outcome features, rather than visible scene properties, predict this divergence, consistent with models relying more on scene-level statistical regularities than on explicit forward simulation, a systematic divergence that accuracy alone cannot reveal. Code is available at https://github.com/fanhong-li/model-human-divergence.
Chinese Translation
视频基础模型在物理推理基准测试中已达到人类水平的准确性，但此类任务需要预测未观察到的物理结果。这些模型是否进行类似人类的前向模拟，还是利用可见场景中的统计规律？仅凭准确性无法区分这些策略。我们引入了一种分布评估框架，将模型种子和人类评估者视为群体，从而能够比较共识、不确定性和策略。在Physion基准测试中，我们评估了三种ViT-L架构（V-JEPA2、VideoMAEv2、DINOv2）。V-JEPA2将准确性差距缩小至约1个百分点（73.2%对74.2%），然而模型与人类之间的分歧达到26.4%，远超人类与人类之间的分歧（4.8%），且一致性显著较低（kappa ~ 0.48对0.91）。这种差异遵循前向模拟的需求：在几何推理（链接，+11.8个百分点）上，模型表现优于人类，但在重力动态（滚动，-11.8个百分点）和因果链（多米诺，-10.5个百分点）上表现不佳。策略指纹分析确认这三种架构均采用非人类策略，而没有一种与人类一致。归因分析表明，未观察到的结果特征，而非可见场景属性，预测了这种差异，这与模型更多依赖场景级统计规律而非明确的前向模拟相一致，这是一种仅凭准确性无法揭示的系统性差异。代码可在 https://github.com/fanhong-li/model-human-divergence 获取。
cs.CV / 35 / 2609.22789
PixelART: Image-to-Layer Decomposition without Latents or Text-to-Image Pretraining
PixelART：无潜变量或文本到图像预训练的图像到层分解
Abstract
Image-to-layer decomposition converts a flattened image into editable RGBA layers, enabling element-level editing in design workflows. Existing diffusion-based systems typically adapt large pretrained text-to-image (T2I) models and introduce RGBA autoencoders or variable-layer architectural modules. We revisit this design choice and ask whether layer decomposition truly requires these heavyweight components. We introduce PixelART, a pixel-space rectified-flow Transformer trained from scratch for image-to-layer (I2L) decomposition. PixelART directly denoises regional RGBA pixel patches using a single-stream multi-modal diffusion Transformer, avoiding RGBA-VAEs, pretrained T2I backbones, and layer-specific decoders. We identify a key property of the task: high-noise timesteps determine layer assignment and coarse layer organization, while low-noise timesteps mainly refine color, alpha, texture, and boundaries. Based on this observation, we propose a terminal-boosted timestep sampling strategy to increase training coverage in the high-noise layer assignment regime. Trained on 4M multi-layer design templates, PixelART achieves state-of-the-art layer decomposition and composite reconstruction on Design-Multi-Layer-Bench and LICA with over 80% fewer parameters, 98% lower latency, and 85% lower memory than the recent Qwen-Image-Layered model. Ablation experiments show that pixel-space $\mathbf{x}$-prediction, terminal-boosted timestep sampling, and data/model scaling are critical, while T2I pretraining brings marginal benefits to the I2L task.
Chinese Translation
图像到层分解将扁平化的图像转换为可编辑的 RGBA 层，从而在设计工作流程中实现元素级编辑。现有的基于扩散的系统通常采用大型预训练的文本到图像（T2I）模型，并引入 RGBA 自编码器或可变层架构模块。我们重新审视这一设计选择，并提出层分解是否真的需要这些重量级组件。我们引入了 PixelART，这是一种从零开始训练的像素空间校正流 Transformer，用于图像到层（I2L）分解。PixelART 直接使用单流多模态扩散 Transformer 对区域 RGBA 像素块进行去噪，避免了 RGBA-VAE、预训练的 T2I 主干和特定层解码器。我们识别出该任务的一个关键特性：高噪声时间步决定层分配和粗略层组织，而低噪声时间步主要细化颜色、透明度、纹理和边界。基于这一观察，我们提出了一种终端增强时间步采样策略，以增加高噪声层分配阶段的训练覆盖率。在 400 万个多层设计模板上训练后，PixelART 在 Design-Multi-Layer-Bench 和 LICA 上实现了最先进的层分解和复合重建，参数减少超过 80%，延迟降低 98%，内存占用降低 85%，相比于最近的 Qwen-Image-Layered 模型。消融实验表明，像素空间的 $ extbf{x}$ 预测、终端增强时间步采样以及数据/模型扩展是关键，而 T2I 预训练对 I2L 任务的边际效益有限。
cs.CV / 36 / 2609.22807
C$^{2}$-INR: Customized Convolutional Implicit Neural Representation
C$^{2}$-INR：定制卷积隐式神经表示
Abstract
Implicit Neural Representation (INR) leverages neural networks to represent discrete signals such as images as continuous ones, where the network weights serve as a compact form of the signal itself. Most existing INR methods adopt Multi-Layer Perceptrons (MLPs) as their backbone. Since these models render each pixel independently, they inherently fail to exploit the spatial correlations that exist between neighboring pixels. In contrast,convolutional INRs can process pixels in parallel while inherently accounting for inter-pixel dependencies, making them a more natural fit for representing images. Nevertheless, convolutional INRs remain relatively underexplored, and the majority of them rely on fixed architectural settings, leaving little room for image-specific adaptation. In this paper, we investigate network customization for convolutional INRs. We replace conventional filters with irregular directional kernels, whose allocation is guided by the directional energy in the image spectrum, i.e., directions exhibiting stronger energy are assigned a larger number of kernels, enabling content-tailored convolution settings. These kernels are further reformulated via an orthogonal basis to achieve a superior sparse representation. Moreover, we introduce an annealed Gumbel-Softmax-based mechanism for kernel-level activation function selection, which gives the most suitable activation function for each convolution kernel. Extensive experiments demonstrate that our method, namely C$^{2}$-INR, achieves superior performance against state-of-the-art approaches under comparable parameter budgets across a wide range of image processing tasks, including representation, inpainting, and super-resolution.
Chinese Translation
隐式神经表示（Implicit Neural Representation, INR）利用神经网络将图像等离散信号表示为连续信号，其中网络权重作为信号本身的紧凑形式。现有的大多数INR方法采用多层感知器（Multi-Layer Perceptrons, MLPs）作为其主干。由于这些模型独立地渲染每个像素，因此在本质上无法利用相邻像素之间存在的空间相关性。相比之下，卷积INRs能够并行处理像素，同时内在地考虑像素间的依赖关系，使其更适合于图像表示。然而，卷积INRs仍然相对未被充分探索，其中大多数依赖于固定的架构设置，几乎没有空间进行特定于图像的适应。在本文中，我们研究了卷积INRs的网络定制。我们用不规则方向核替换传统滤波器，其分配由图像频谱中的方向能量引导，即能量较强的方向分配更多的核，从而实现内容定制的卷积设置。这些核进一步通过正交基进行重构，以实现更优的稀疏表示。此外，我们引入了一种基于退火Gumbel-Softmax的机制用于核级激活函数选择，为每个卷积核提供最合适的激活函数。大量实验表明，我们的方法，即C$^{2}$-INR，在可比参数预算下，在包括表示、修复和超分辨率等广泛的图像处理任务中，表现优于最先进的方法。
cs.CV / 37 / 2609.22834
SatOV: Restoring Spatial Priors for Training-Free Open-Vocabulary Segmentation in Remote Sensing Imagery
SatOV：恢复空间先验以实现无训练的开放词汇遥感图像分割
Abstract
Open-vocabulary semantic segmentation (OVS) of remote sensing imagery is a challenging pixel-level task requiring strong generalization and adaptation to the spatial characteristics of remote sensing data. Although existing vision-language foundation models perform well in general domains, their image-level classification design weakens the spatial priors needed for high-resolution remote sensing segmentation: structural spatial relations are degraded during deep feature transformation, and fine-grained spatial details are lost during downsampling. To address these complementary deficiencies, we propose SatOV, a training-free framework for open-vocabulary remote sensing segmentation that restores spatial priors at two stages of the representation pipeline. Specifically, Residual QQ Attention (ResQQ) extracts Query-Key self-attention from an intermediate CLIP layer and fuses it with final-layer Query-Query attention via a residual combination, restoring structural spatial priors suppressed by the final-layer representation. Spatially Modulated Upsampling (SatUp) uses the original high-resolution RGB image as spatial guidance, combining spatial feature modulation with guided cross-attention to reconstruct pixel-level textures and boundaries. Extensive experiments on DOTA, UDD, LoveDA, and Vaihingen show that SatOV consistently improves training-free OVS and achieves competitive quantitative and qualitative results against state-of-the-art methods. These results validate the effectiveness of restoring spatial priors at both the representation and spatial-resolution stages for remote sensing open-vocabulary segmentation.
Chinese Translation
开放词汇语义分割（OVS）在遥感图像中的应用是一项具有挑战性的像素级任务，要求对遥感数据的空间特征具有强大的泛化能力和适应能力。尽管现有的视觉-语言基础模型在一般领域表现良好，但其图像级分类设计削弱了高分辨率遥感分割所需的空间先验：在深度特征转换过程中，结构空间关系被降解，而在下采样过程中，细粒度的空间细节则丢失。为了解决这些互补的不足，我们提出了SatOV，这是一种无训练的开放词汇遥感分割框架，旨在在表示管道的两个阶段恢复空间先验。具体而言，残差QQ注意力（Residual QQ Attention, ResQQ）从中间的CLIP层提取查询-键自注意力，并通过残差组合将其与最终层的查询-查询注意力融合，从而恢复被最终层表示抑制的结构空间先验。空间调制上采样（Spatially Modulated Upsampling, SatUp）使用原始高分辨率RGB图像作为空间指导，将空间特征调制与引导交叉注意力相结合，以重建像素级的纹理和边界。在DOTA、UDD、LoveDA和Vaihingen上的大量实验表明，SatOV在无训练的OVS中始终表现出色，并在定量和定性结果上与最先进的方法相媲美。这些结果验证了在遥感开放词汇分割中，在表示和空间分辨率阶段恢复空间先验的有效性。
cs.CV / 38 / 2609.22849
LINGO: Latent Initialization and Gradient Optimization for Sparse-view X-ray Novel View Synthesis and CT Reconstruction with 3D Gaussian Splatting
LINGO：用于稀疏视图 X 射线新视图合成和 CT 重建的潜在初始化与梯度优化
Abstract
In novel view synthesis and Computed Tomography (CT) reconstruction with sparse-view X-ray imaging, insufficient angular coverage leads to structural ambiguity and accumulated noise. Integrating 3D Gaussian Splatting (3DGS) with X-ray absorption physics can achieve promising results, but it suffers from noisy initialization, positional insensitivity, and weak gradients in low-density regions. In this paper, we propose a unified Latent Initialization and Gradient Optimization (LINGO) framework to address these issues. LINGO combines latent mask-space initialization with dynamic gradient optimization to improve point cloud structural completeness while accelerating training. It constructs voxel-level 3D filters from X-ray masks to robustly suppress background noise and provide reliable geometric priors. By employing an adaptive voxel scaling strategy and dynamically scaling loss, LINGO can adjust spatial resolution and explicitly amplify gradients in low-density structures. To evaluate the quality of initialization, we introduce the Initialization Point Cloud Structural Deviation (IPSD) metric. Experiments on the X3D dataset indicate that for the novel view synthesis task, LINGO improves the Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) by an average of 0.72 and 0.0039, respectively, over baselines under identical sparse-view settings, achieving comparable reconstruction quality within 5k steps to state-of-the-art models typically trained with 30k iterations. For the CT reconstruction task, LINGO also demonstrates consistent improvements, with average PSNR and SSIM gains of 0.36 and 0.0134. These results highlight LINGO's effectiveness in both accelerating training and enhancing reconstruction quality across different sparse-view imaging scenarios.
Chinese Translation
在稀疏视图 X 射线成像的新的视图合成和计算机断层扫描（CT）重建中，角度覆盖不足导致结构模糊和噪声积累。将 3D 高斯喷溅（3DGS）与 X 射线吸收物理相结合可以取得良好的效果，但它面临着噪声初始化、位置不敏感和低密度区域梯度弱的问题。本文提出了一种统一的潜在初始化与梯度优化（LINGO）框架来解决这些问题。LINGO 将潜在掩码空间初始化与动态梯度优化相结合，以提高点云的结构完整性，同时加速训练。它从 X 射线掩码构建体素级 3D 滤波器，以稳健地抑制背景噪声并提供可靠的几何先验。通过采用自适应体素缩放策略和动态缩放损失，LINGO 可以调整空间分辨率并明确放大低密度结构中的梯度。为了评估初始化的质量，我们引入了初始化点云结构偏差（IPSD）指标。在 X3D 数据集上的实验表明，对于新视图合成任务，LINGO 在相同的稀疏视图设置下，平均提高峰值信噪比（PSNR）和结构相似性指数（SSIM）分别为 0.72 和 0.0039，且在 5k 步内实现与通常训练 30k 次迭代的最先进模型相当的重建质量。对于 CT 重建任务，LINGO 也展示了一致的改进，平均 PSNR 和 SSIM 增益分别为 0.36 和 0.0134。这些结果突显了 LINGO 在加速训练和提高不同稀疏视图成像场景下重建质量方面的有效性。
cs.CV / 39 / 2609.22857
Image Frame Dynamic Object Segmentation and Ego Motion Estimation using Radar Image Fusion
基于雷达图像融合的图像帧动态物体分割与自我运动估计
Abstract
Dynamic object segmentation and ego-motion estimation are closely coupled problems in autonomous driving, as accurate ego-motion estimation typically requires static scene observations, while identifying static observations requires knowledge of the ego motion. We present Radar-Dot, a radar--RGB framework that exploits radar Doppler measurements to address this coupling. Radar returns are first used to estimate ego velocity through a linear Doppler constraint, with residual-based static/dynamic segmentation and robust estimation used to reduce the influence of moving objects. The estimated motion is then combined with metric depth and dense optical flow to identify image regions whose observed motion is inconsistent with the rigid scene motion. Experiments on 10 nuScenes scenes (part of nuscenes-mini) demonstrate that the resulting geometric pipeline achieves 20.24% dynamic IoU and 33.67% F1-score over 394 frame pairs, while radar-based static-point filtering improves ego-velocity estimation compared with using all radar returns. These results demonstrate the potential of radar as a modality for jointly improving ego-motion estimation and dynamic object segmentation.
Chinese Translation
动态物体分割与自我运动估计是自动驾驶中密切相关的问题，因为准确的自我运动估计通常需要静态场景观察，而识别静态观察又需要对自我运动的了解。我们提出了Radar-Dot，一个利用雷达多普勒测量的雷达-RGB框架，以解决这一耦合问题。首先，利用雷达回波通过线性多普勒约束估计自我速度，并采用基于残差的静态/动态分割和鲁棒估计来减少移动物体的影响。然后，将估计的运动与度量深度和稠密光流结合，以识别观察到的运动与刚性场景运动不一致的图像区域。在10个nuScenes场景（nuscenes-mini的一部分）上的实验表明，所得到的几何管道在394帧对上实现了20.24%的动态IoU和33.67%的F1-score，而基于雷达的静态点过滤相比使用所有雷达回波改善了自我速度估计。这些结果展示了雷达作为一种模态在共同改善自我运动估计和动态物体分割方面的潜力。
cs.CV / 40 / 2609.22868
Planning-Aligned Pretraining of BEV Representations with Sparse Action-Conditioned Targets for End-to-End Autonomous Driving
基于稀疏动作条件目标的鸟瞰视图（BEV）表示的规划对齐预训练用于端到端自主驾驶
Abstract
End-to-end driving requires planning-relevant bird's-eye-view (BEV) representations, but existing pretraining approaches often rely on task annotations or dense scene reconstruction. We introduce PAVER, Planning-Aligned BEV Encoder Pretraining. From a single LiDAR sweep, PAVER constructs sparse risk and unknown targets describing occupied and unobserved evidence along rule-based ego motions. A 10K-parameter head predicts these targets from masked BEV features conditioned on the action state, directing supervision toward geometric constraints on candidate motions. Pretraining requires no driving-task annotations or dense reconstruction. Only the BEV encoder is transferred, preserving the downstream architecture and camera-only inference. On nuScenes, PAVER reduces VAD-Tiny's average collision rate from 0.51% to 0.19%, while improving planning L2, motion prediction, detection, and mapping. The selected VAD-Tiny and VAD-Base schedules use about 36% less estimated total training time than scratch training, including pretraining. On Bench2Drive Town05 Long, PAVER improves UniAD-Tiny's closed-loop Driving Score from 48.45 to 58.79. The project page is available at https://archiiive99.github.io/PAVER.
Chinese Translation
端到端驾驶需要与规划相关的鸟瞰视图（BEV）表示，但现有的预训练方法通常依赖于任务注释或密集场景重建。我们提出了PAVER，规划对齐的BEV编码器预训练。从单次激光雷达扫描中，PAVER构建稀疏的风险和未知目标，描述沿基于规则的自我运动的占用和未观察到的证据。一个拥有10K参数的头部从基于动作状态的掩蔽BEV特征中预测这些目标，将监督引导到候选运动的几何约束上。预训练不需要驾驶任务注释或密集重建。仅转移BEV编码器，保留下游架构和仅依赖摄像头的推理。在nuScenes数据集上，PAVER将VAD-Tiny的平均碰撞率从0.51%降低到0.19%，同时改善了规划L2、运动预测、检测和映射。所选的VAD-Tiny和VAD-Base训练计划的估计总训练时间比从头训练减少约36%，包括预训练。在Bench2Drive Town05 Long上，PAVER将UniAD-Tiny的闭环驾驶评分从48.45提高到58.79。项目页面可访问 https://archiiive99.github.io/PAVER。
cs.CV / 41 / 2609.22896
Combining Foundation Model Confidence and Monocular Depth for Training-Free Out-of-Distribution Segmentation
结合基础模型置信度和单目深度实现无训练的分布外分割
Abstract
Autonomous vehicles operating in open-world scenarios are inevitably confronted with previously unknown objects, such as exotic animals or loose cargo. The reliable detection and segmentation of these out-of-distribution (OOD) objects is therefore crucial for a safe understanding of the environment and decision-making. Most existing approaches require access to OOD training samples, retraining of the segmentation backbone, or dedicated auxiliary architectures, limiting their practical applicability. We propose a training-free method that derives dense OOD scores directly from the confidence predictions of a foundation segmentation model, without any task-specific fine-tuning or access to anomalous data. To improve the robustness of our OOD segmentation, geometric information from monocular depth estimation is incorporated into the decision process, providing complementary cues to uncertainty-based predictions. We evaluate the proposed method on the SegmentMeIfYouCan benchmark and additionally assess its performance on OOD tracking in video sequences, reflecting the temporal nature of real-world perception systems. The method performs strongly on road-centered benchmarks.
Chinese Translation
在开放世界场景中运行的自动驾驶车辆不可避免地会遇到以前未知的物体，例如异国动物或松散货物。因此，可靠地检测和分割这些分布外（OOD）物体对于安全理解环境和决策至关重要。大多数现有方法需要访问OOD训练样本、重新训练分割主干网络或专用辅助架构，这限制了它们的实际适用性。我们提出了一种无训练的方法，该方法直接从基础分割模型的置信度预测中推导出密集的OOD分数，无需任何特定任务的微调或访问异常数据。为了提高我们OOD分割的鲁棒性，我们将单目深度估计的几何信息纳入决策过程中，为基于不确定性的预测提供补充线索。我们在SegmentMeIfYouCan基准上评估了所提方法，并额外评估其在视频序列中的OOD跟踪性能，反映了现实世界感知系统的时间特性。该方法在以道路为中心的基准上表现出色。
cs.CV / 42 / 2609.22897
Scout: Open-World Species Recognition on the Edge
Scout：边缘计算中的开放世界物种识别
Abstract
Large vision-language models (VLMs) enable recognition beyond a fixed class set, but their computational demands prevent them from running on many edge devices. Cloud offload makes this capability accessible, but sending every image consumes scarce bandwidth and communication energy. We ask how to bring the open-world recognition capability of VLMs to the edge while operating within tight compute, energy, and bandwidth budgets. Wildlife monitoring provides a natural setting for exploring this question because camera traps encounter species not known at deployment. We present Scout, an autonomous open-world recognition system that invokes a cloud VLM intermittently to teach new classes to a compact edge model. Given only the deployment location and empty site frames, Scout autonomously turns each species identified by the VLM into persistent, site-conditioned recognition capability in a resource-efficient edge model, without a predefined species list, human labeling, or manual tuning. Across 30 camera-trap deployments in three regions on an NVIDIA Jetson Orin Nano, the accuracy of Scout remains within 0.1-2.5% of a model given a predefined species list. On species outside its initial class set, Scout achieves 53.7-59.1% accuracy, compared with 56.5-65.1% for full cloud offload, while using 59-71% less deployment energy.
Chinese Translation
大型视觉语言模型（VLMs）使得超越固定类别集的识别成为可能，但其计算需求使得许多边缘设备无法运行。云端卸载使得这一能力得以实现，但发送每张图像会消耗宝贵的带宽和通信能量。我们探讨如何在计算、能量和带宽预算紧张的情况下，将VLM的开放世界识别能力带到边缘设备。野生动物监测提供了一个自然的环境来探索这个问题，因为相机陷阱会遇到在部署时未知的物种。我们提出了Scout，一个自主的开放世界识别系统，它间歇性地调用云端VLM来教会紧凑的边缘模型新的类别。仅凭部署位置和空白场景帧，Scout能够自主地将VLM识别的每个物种转化为在资源高效的边缘模型中持久的、基于场所的识别能力，而无需预定义的物种列表、人为标注或手动调优。在三个区域的30个相机陷阱部署中，Scout的准确率保持在0.1-2.5%的范围内，与给定预定义物种列表的模型相比较。在初始类别集之外的物种上，Scout的准确率达到53.7-59.1%，而完全云端卸载的准确率为56.5-65.1%，同时使用了59-71%更少的部署能量。
cs.CV / 43 / 2609.22913
AVTR-1: Open Stack for Real-Time Interactive Avatars
AVTR-1：实时互动化身的开放平台
Abstract
Talking-head and dyadic models now achieve real-time inference, yet fast motion generation alone does not produce an interactive conversation. A live system must synchronize the model's output with speech from an external voice agent, schedule video frames for playback, and handle interruptions. We introduce AVTR-1, an open stack for real-time interactive avatar conversations, built around a compact 153M-parameter autoregressive flow-matching motion generator conditioned on both participants' audio. We adapt its audio encoder for streaming through self-distillation. The stack turns the model's chunk-based generation into a continuous, synchronized audio-video stream driven by an external voice agent, and we analytically derive its contribution to the user-facing latencies and validate the resulting bounds with two commercial voice agents. Further experiments demonstrate that AVTR-1 leads the compared dyadic systems on all reported visual-quality metrics and most conventional listening-motion metrics while remaining competitive in lip synchronization. Its inference runtime operates in real time on data-center and consumer GPUs. However, conventional listening metrics do not establish whether the paired speaker's speech contributes to generated motion. We therefore introduce the Reference-Based Directed Granger Gain (R-DGG), which measures the additional predictive information carried by speaker speech after accounting for listener history and speaker motion. R-DGG finds statistically supported predictive dependence for recorded listeners and all evaluated dyadic systems, but not for talking-head generators without paired audio or mismatched speaker-listener pairs. We release the model weights, renderer, and serving backend under component-specific licenses.
Chinese Translation
目前，谈话头和双人模型已实现实时推理，但仅靠快速运动生成并不能产生互动对话。一个实时系统必须将模型的输出与外部语音代理的语音同步，安排视频帧进行播放，并处理中断。我们介绍了AVTR-1，一个用于实时互动化身对话的开放平台，构建于一个紧凑的153M参数自回归流匹配运动生成器之上，该生成器以两个参与者的音频为条件。我们对其音频编码器进行了自蒸馏的流媒体适配。该平台将模型的基于块的生成转变为由外部语音代理驱动的连续、同步的音视频流，并且我们分析性地推导出其对用户面对的延迟的贡献，并通过两个商业语音代理验证了结果界限。进一步的实验表明，AVTR-1在所有报告的视觉质量指标和大多数传统听觉运动指标上领先于比较的双人系统，同时在唇同步方面保持竞争力。其推理运行时间在数据中心和消费者GPU上实时运行。然而，传统的听觉指标并不能确定配对发言者的语音是否对生成的运动有贡献。因此，我们引入了基于参考的定向Granger增益（R-DGG），该指标在考虑听众历史和发言者运动后，衡量发言者语音所携带的额外预测信息。R-DGG在记录的听众和所有评估的双人系统中发现了统计支持的预测依赖关系，但在没有配对音频或不匹配的发言者-听众对的谈话头生成器中则没有。我们发布了模型权重、渲染器和服务后端，采用组件特定的许可证。
cs.CV / 44 / 2609.22916
Planning and Rendering in Concert: DeepFusion of Autoregressive Layouts and Diffusion for Visual Text Generation
协同规划与渲染：自回归布局与扩散的深度融合用于视觉文本生成
Abstract
Generating text-rich images from prompts requires both textual fidelity and the coherent integration of text into the surrounding image. An explicit layout can provide structured guidance about what text should appear and where, but a well-formed plan alone does not guarantee that the renderer will realize it faithfully. Existing layout-based AR-diffusion systems typically optimize planning and rendering separately, preventing the planner's representations from being adapted jointly with image synthesis. We introduce DuetGen, an autonomous visual text generator built on DeepFusion, which jointly learns autoregressive planning and continuous diffusion rendering. DeepFusion conditions a diffusion transformer on the planner's prompt and bbox-content hidden states, allowing rendering supervision to shape the representations connecting textual plans with visual outputs. Its joint objective combines autoregressive plan supervision, text-region-weighted diffusion learning, and auxiliary coordinate supervision to maintain structured planning, emphasize text-bearing regions, and improve the spatial precision of planner representations. During inference, Phase-Aware Attention Modulation strengthens the correspondence between image regions and their matched coordinate and content states, facilitating region-specific execution of the generated plan. With a 2B planner and a 4B single-stream DiT, DuetGen achieves 0.8293 word accuracy on CVTG-2K and 0.938 accuracy on LongText-Bench, closely matching the substantially larger Qwen-Image on both benchmarks. These results demonstrate the value of jointly learned planning representations and region-specific rendering for autonomous visual text generation.
Chinese Translation
从提示生成文本丰富的图像需要文本的忠实性和文本与周围图像的连贯整合。明确的布局可以提供关于文本应出现的位置和内容的结构化指导，但仅有一个良好的规划并不能保证渲染器会忠实地实现它。现有的基于布局的自回归（AR）扩散系统通常分别优化规划和渲染，阻碍了规划者的表示与图像合成的联合适应。我们提出了DuetGen，这是一种基于深度融合（DeepFusion）的自主视觉文本生成器，能够联合学习自回归规划和连续扩散渲染。深度融合将扩散变换器的条件设定在规划者的提示和边界框内容的隐藏状态上，使得渲染监督能够塑造连接文本规划与视觉输出的表示。其联合目标结合了自回归计划监督、文本区域加权的扩散学习和辅助坐标监督，以维持结构化规划、强调文本区域并提高规划者表示的空间精度。在推理过程中，阶段感知注意力调制增强了图像区域与其匹配的坐标和内容状态之间的对应关系，促进了生成计划的区域特定执行。通过使用一个2B的规划器和一个4B的单流DiT，DuetGen在CVTG-2K上达到了0.8293的词准确率，在LongText-Bench上达到了0.938的准确率，与这两个基准上规模更大的Qwen-Image表现相近。这些结果展示了联合学习的规划表示和区域特定渲染在自主视觉文本生成中的价值。
cs.CV / 45 / 2609.22941
D3GS: Depth, DINO, and RGB Diffusion Co-Guided 3D Gaussian Splatting for Sparse-View Reconstruction
D3GS：深度、DINO 和 RGB 扩散共同引导的稀疏视图重建中的 3D 高斯点云
Abstract
Novel view synthesis from sparse inputs remains challenging for 3D Gaussian Splatting (3DGS) due to ambiguous geometry, cross-view inconsistency, and missing details in under-constrained regions, resulting in degraded reconstruction and unstable rendering. To tackle these issues, we propose D$^{3}$GS, a Depth-DINO-Diffusion guided sparse-view Gaussian reconstruction framework that jointly enhances geometry and appearance. D$^{3}$GS first recovers a high-resolution, metric depth map via diffusion-based completion and DPT (Dense Prediction Transformer) refinement, providing robust Gaussian initialization and geometric constraints. Then, a DINO-guided view-consistent learning is introduced to augment Gaussian attributes with structural features, improving multi-view consistency. Finally, a diffusion-based Gaussian refinement module injects generative priors into an iterative optimization strategy, enhancing high-frequency geometric and appearance details within the Gaussian representation. Experiments on DTU, LLFF, and Mip-NeRF 360 show that D$^{3}$GS achieves consistent and substantial improvements over strong baselines, with ablation studies validating the effectiveness and complementary roles of each component.
Chinese Translation
从稀疏输入生成新视图对于 3D 高斯点云（3DGS）仍然具有挑战性，因为存在模糊的几何形状、视图间不一致性以及在约束不足区域缺失的细节，导致重建效果下降和渲染不稳定。为了解决这些问题，我们提出了 D$^{3}$GS，一种深度-DINO-扩散引导的稀疏视图高斯重建框架，旨在共同增强几何形状和外观。D$^{3}$GS 首先通过基于扩散的补全和 DPT（密集预测变换器）细化恢复高分辨率的度量深度图，从而提供稳健的高斯初始化和几何约束。接着，引入 DINO 引导的视图一致性学习，以结构特征增强高斯属性，提高多视图一致性。最后，基于扩散的高斯细化模块将生成先验注入到迭代优化策略中，增强高斯表示中的高频几何和外观细节。在 DTU、LLFF 和 Mip-NeRF 360 上的实验表明，D$^{3}$GS 在强基线之上实现了一致且显著的改进，消融研究验证了各个组件的有效性和互补作用。
cs.CV / 46 / 2609.22942
An Evolutionary Agentic Approach for Open-ended Image Quality Perception
一种用于开放式图像质量感知的进化代理方法
Abstract
Generative models are rapidly expanding image quality assessment (IQA) beyond traditional fidelity factors to emerging dimensions such as physical plausibility and text-rendering correctness. However, existing IQA models rely on fixed definitions and heavy supervision, making them difficult to extend to open-ended perceptual dimensions. We identify holistic bias as an important limitation: when scoring an unseen dimension, models reuse generic quality priors, leading to scoring errors and rank inversion. To address this, we propose PACE (Perceptual Agentic Collaborative Evolution), a training-free multi-agent framework that formulates open-ended IQA as explicit protocol construction. Given a target dimension, PACE uses collaborative agents to construct an evaluation protocol composed of verifiable Visual Question Answering (VQA) probes, grounding evaluation in concrete visual evidence rather than holistic impressions. The resulting protocol is calibrated using only four human-annotated images per dimension, while a dual-track scoring mechanism aligns model perception with human scoring scales. Across traditional IQA, structural fidelity, context-aware aesthetics, and newly defined open-ended dimensions, PACE consistently improves its MLLM backbone, achieving competitive performance across diverse IQA settings, and reduces the Holistic Override Rate (HOR) from 44.4\% to 8.6\%.
Chinese Translation
生成模型正在迅速扩展图像质量评估（IQA），超越传统的保真度因素，向物理合理性和文本渲染正确性等新兴维度发展。然而，现有的IQA模型依赖于固定的定义和重度监督，这使得它们难以扩展到开放式感知维度。我们识别出整体偏差作为一个重要的限制：在对未见维度进行评分时，模型重用通用质量先验，导致评分错误和排名颠倒。为了解决这一问题，我们提出了PACE（感知代理协作进化），这是一个无训练的多代理框架，将开放式IQA形式化为明确的协议构建。给定一个目标维度，PACE使用协作代理构建一个由可验证的视觉问答（VQA）探针组成的评估协议，将评估基于具体的视觉证据，而不是整体印象。最终的协议仅使用每个维度四张人类标注的图像进行校准，同时双轨评分机制将模型感知与人类评分标准对齐。在传统IQA、结构保真度、上下文感知美学以及新定义的开放式维度中，PACE始终改善其MLLM骨干，在多样化的IQA设置中实现了具有竞争力的性能，并将整体覆盖率（HOR）从44.4%降低到8.6%。
cs.CV / 47 / 2609.22947
RewardVerse: Rubric-Guided Policy Optimization for Video Reward Modeling
RewardVerse：基于评分标准的视频奖励建模的策略优化
Abstract
Reinforcement learning (RL) is vital for optimizing video generation models, with a robust reward model (RM) serving as the cornerstone. However, existing video reward models often produce unstable scalar scores because they directly map complex, subjective video quality into a single score without explicit evaluation criteria. This leads to scalar drift, where the scoring scale collapses or shifts across different prompts, making the reward unreliable for RL. Drawing inspiration from professional human annotation engineering, we address this problem with RewardVerse, a rubric-based video reward framework that introduces a dynamic rubric as an intermediate representation between the evaluation query and the scorer. Instead of unconstrained direct scoring, RewardVerse first generates explicit evaluation criteria and then performs rubric-guided scoring, providing a stable semantic anchor that mitigates scalar drift. To efficiently optimize this collaborative pipeline, we propose Rubric-Guided Policy Optimization (RGPO), a two-stage training algorithm. RGPO first warms up the scorer using self-evolving seed rubrics and then jointly optimizes the rubric generator to produce query-adaptive evaluation criteria while continuously aligning the scorer with human ratings. Extensive experiments on the 16-dimensional EvalVerse benchmark and external datasets demonstrate that RewardVerse mitigates scalar drift, achieves state-of-the-art performance on both pointwise and pairwise evaluation, and provides a robust and interpretable reward signal for RL in video generation.
Chinese Translation
强化学习（RL）在优化视频生成模型中至关重要，而强大的奖励模型（RM）则是其基石。然而，现有的视频奖励模型往往产生不稳定的标量分数，因为它们直接将复杂、主观的视频质量映射为单一分数，而没有明确的评估标准。这导致了标量漂移，即评分尺度在不同提示之间崩溃或偏移，使得奖励在RL中变得不可靠。受到专业人类注释工程的启发，我们提出了RewardVerse，一个基于评分标准的视频奖励框架，引入了动态评分标准作为评估查询与评分者之间的中介表示。RewardVerse首先生成明确的评估标准，然后进行基于评分标准的评分，提供一个稳定的语义锚点，从而减轻标量漂移。为了高效优化这一协作流程，我们提出了基于评分标准的策略优化（RGPO），这是一种两阶段的训练算法。RGPO首先使用自我演变的种子评分标准对评分者进行预热，然后联合优化评分标准生成器，以生成查询自适应的评估标准，同时持续将评分者与人类评分对齐。在16维EvalVerse基准和外部数据集上的广泛实验表明，RewardVerse减轻了标量漂移，在逐点和成对评估上均实现了最先进的性能，并为视频生成中的RL提供了强大且可解释的奖励信号。
cs.CV / 48 / 2609.22950
CLEAR: Complex Learned Explicit Analytical Regularization for Ultra-Accelerated 4D Flow CMR Reconstruction
CLEAR：用于超加速4D流动心脏磁共振重建的复杂学习显式分析正则化
Abstract
While compressed-sensing regularizers enable interpretable reconstruction of 4D Flow CMR through transparent variational objectives, their hand-crafted nature is too restrictive under high acceleration. State-of-the-art learning-based approaches mitigate this, but typically encode regularization implicitly through unrolled network modules, which limits their interpretability. To address this limitation, we propose CLEAR, designed to combine the interpretability of compressed sensing with the flexibility of learned models. To the best of our knowledge, it is the first learned regularizer for a 4D reconstruction task. In the ultra-accelerated \(10\times\)--\(50\times\) regime of the CMRx4DFlow2026 challenge, CLEAR outperforms compressed sensing locally low-rank (LLR) and the popular variational network FlowVN, while using less than 10k parameters and preserving an interpretable regularization structure.
Chinese Translation
尽管压缩感知正则化器通过透明的变分目标实现了4D流动心脏磁共振的可解释重建，但其手工设计的特性在高加速下过于限制。最先进的基于学习的方法缓解了这一问题，但通常通过展开的网络模块隐式编码正则化，这限制了其可解释性。为了解决这一限制，我们提出了CLEAR，旨在将压缩感知的可解释性与学习模型的灵活性相结合。根据我们所知，这是第一个用于4D重建任务的学习正则化器。在CMRx4DFlow2026挑战的超加速10×至50×范围内，CLEAR在使用不到10k参数的情况下，超越了局部低秩（LLR）压缩感知和流行的变分网络FlowVN，同时保持了可解释的正则化结构。
cs.CV / 49 / 2609.22967
General Collaborative Intelligence: Architecting Cognition for Resilient Multi-Agent Ecosystems
通用协作智能：为弹性多智能体生态系统构建认知架构
Abstract
Multi-agent unmanned systems are moving from isolated, ego-centric sensing toward collaborative intelligence, in which distributed agents exchange compact features to overcome a local observation trap that no single agent can escape: occlusions, finite sensor range, and environmental degradation. The field has matured across architectural, communication, embodied, resilience, and trust dimensions, yet existing surveys examine these dimensions in isolation and rarely expose their dependencies. This review offers a unified synthesis through two complementary lenses. The first is a five-dimensional taxonomy spanning collaboration stage, communication paradigm, fusion architecture, learning strategy, and application domain. The second is three cognitive synergy conditions, Semantic Disambiguation, Pragmatic Information Exchange, and Proactive Informational Foraging, that turn cognitive synergy into operational criteria. Across these lenses we survey collaboration architectures and topologies, neural-communication co-design that treats the channel as a differentiable pipeline component, embodied action-perception loops via multi-agent reinforcement learning, and resilience mechanisms for synchronization, uncertainty quantification, and label-efficient learning. We then map these advances onto four operational domains, V2X, unmanned aerial, industrial logistics, and smart cities, and onto the safety-privacy-utility triad. To counter benchmark saturation and evaluation fragmentation, we propose GCI-Bench, a five-pillar scoring protocol with a maturity model that makes the trade-offs of collaborative methods comparable across studies. A critical reflection on reproducibility, the sim-to-real gulf, and conditions under which collaboration degrades performance identifies open challenges and charts directions toward general collaborative intelligence under real-world uncertainty.
Chinese Translation
多智能体无人系统正从孤立的自我中心感知向协作智能转变，在这一过程中，分布式智能体交换紧凑特征，以克服单个智能体无法逃脱的局部观察陷阱：遮挡、有限的传感器范围和环境退化。该领域在架构、通信、体现、弹性和信任等维度上已经成熟，然而现有的调查往往孤立地考察这些维度，鲜有揭示它们之间的依赖关系。本文通过两个互补的视角提供了统一的综合。第一个视角是一个涵盖协作阶段、通信范式、融合架构、学习策略和应用领域的五维分类法。第二个视角是三个认知协同条件：语义消歧、务实信息交换和主动信息觅食，这些条件将认知协同转化为操作标准。通过这些视角，我们调查了协作架构和拓扑、将通道视为可微分管道组件的神经通信共同设计、通过多智能体强化学习实现的体现行动-感知循环，以及用于同步、不确定性量化和标签高效学习的弹性机制。随后，我们将这些进展映射到四个操作领域：车联网（V2X）、无人机、工业物流和智慧城市，以及安全-隐私-效用三元组。为了应对基准饱和和评估碎片化的问题，我们提出了GCI-Bench，一个包含五个支柱的评分协议和成熟度模型，使协作方法的权衡在各研究之间可比。对可重复性、模拟与现实之间的鸿沟以及协作性能下降的条件进行的批判性反思，识别了开放挑战，并为在现实世界不确定性下实现通用协作智能指明了方向。
cs.CV / 50 / 2609.23003
M3GA-Wild: A Large-Scale Dataset and Benchmark for Multi-Modal Multi-session Ground-to-Aerial Place Recognition in Forests
M3GA-Wild：一个用于森林中多模态多会话地面到空中地点识别的大规模数据集和基准
Abstract
We present M3GA-Wild, the first benchmark for multi-modal, multi-session ground-to-aerial place recognition in forests. M3GA-Wild unifies and extends existing forest localisation datasets, providing a holistic benchmark with synchronised RGB imagery and LiDAR from ground traversals spanning 36 km, aligned high-resolution aerial imagery and multi-altitude LiDAR covering 370 hectares, and accurate geo-referenced 6-DoF poses for precise evaluation. M3GA-Wild captures diverse forest scenes with varying viewpoints, occlusion, and environmental conditions, enabling systematic evaluation of visual, LiDAR, cross-modal, and multi-modal methods. Baseline experiments show that LiDAR-based approaches significantly outperform vision-only methods under severe viewpoint differences, while current multi-modal fusion strategies yield limited gains due to poor cross-modal alignment. By pairing aerial RGB imagery with geo-referenced aerial LiDAR, M3GA-Wild also enables evaluation of foundation models for monocular depth estimation as a cheap source of 3D geometry from forest imagery, with initial experiments revealing shortfalls of current methods. These results highlight key challenges in cross-platform localisation, including modality misalignment and severe domain gaps. M3GA-Wild establishes a new benchmark to support research in robust multi-modal localisation and long-term autonomy in unstructured natural environments. The dataset and code will be available upon acceptance.
Chinese Translation
我们提出了M3GA-Wild，这是第一个用于森林中多模态、多会话地面到空中地点识别的基准。M3GA-Wild统一并扩展了现有的森林定位数据集，提供了一个整体基准，包括来自36公里地面行走的同步RGB图像和激光雷达（LiDAR），对齐的高分辨率空中图像和覆盖370公顷的多高度LiDAR，以及用于精确评估的准确地理参考6自由度（6-DoF）姿态。M3GA-Wild捕捉了具有不同视角、遮挡和环境条件的多样森林场景，使得对视觉、LiDAR、跨模态和多模态方法的系统评估成为可能。基线实验表明，在严重的视角差异下，基于LiDAR的方法显著优于仅依赖视觉的方法，而当前的多模态融合策略由于跨模态对齐不佳而获得的提升有限。通过将空中RGB图像与地理参考的空中LiDAR配对，M3GA-Wild还使得对单目深度估计基础模型的评估成为可能，作为从森林图像中获取3D几何形状的廉价来源，初步实验揭示了当前方法的不足。这些结果突显了跨平台定位中的关键挑战，包括模态不对齐和严重的领域差距。M3GA-Wild建立了一个新的基准，以支持在非结构化自然环境中进行稳健的多模态定位和长期自主研究。数据集和代码将在接受后提供。
cs.CV / 51 / 2609.23005
Compressing 3D Gaussian Splatting via Cross-Representation Priors
通过跨表示先验压缩3D高斯点云
Abstract
3D Gaussian Splatting (3DGS) enables high-quality novel view synthesis but incurs high storage and transmission costs due to dense Gaussian primitives. Recent anchor-based compression reduces per-primitive redundancy, yet redundancy across anchors remains largely unexploited. We propose CRP-GS (Cross-Representation Priors for Gaussian Splatting), a rate-distortion optimized compression framework that leverages cross-representation priors to improve anchor-level entropy modeling. First, a Correspondence-Oriented Hierarchical Structure (COHS) organizes anchors by feature correspondence rather than spatial proximity, constructing root-leaf dependencies so that selected anchors can act as informative priors to conditionally encode others, yielding more accurate likelihood prediction and lower conditional entropy. Second, Shared Feature Aggregation (SFA) extracts globally shared features from a contextual hash grid and injects them into anchor representations, factoring out scene-consistent low-frequency information that would otherwise be redundantly embedded in individual anchors. Both modules are trained under a unified rate-distortion objective to balance bitrate reduction and rendering fidelity. Experiments across multiple benchmarks show that CRP-GS achieves a favorable overall rate-distortion trade-off, yielding around 30% average bitrate reduction compared to anchor-based baselines while maintaining comparable rendering quality.
Chinese Translation
3D高斯点云（3D Gaussian Splatting, 3DGS）能够实现高质量的新视图合成，但由于密集的高斯原语，导致高存储和传输成本。近期的基于锚点的压缩方法减少了每个原语的冗余，但锚点之间的冗余仍然未被充分利用。我们提出了CRP-GS（Cross-Representation Priors for Gaussian Splatting），这是一种基于率失真优化的压缩框架，利用跨表示先验来改善锚点级别的熵建模。首先，基于对应关系的分层结构（Correspondence-Oriented Hierarchical Structure, COHS）根据特征对应关系而非空间邻近性来组织锚点，构建根叶依赖关系，使得选定的锚点可以作为信息先验来有条件地编码其他锚点，从而实现更准确的似然预测和更低的条件熵。其次，共享特征聚合（Shared Feature Aggregation, SFA）从上下文哈希网格中提取全局共享特征，并将其注入锚点表示中，从而剔除场景一致的低频信息，这些信息在单个锚点中会被冗余嵌入。这两个模块在统一的率失真目标下进行训练，以平衡比特率降低和渲染保真度。在多个基准测试中的实验表明，CRP-GS实现了良好的整体率失真权衡，与基于锚点的基线相比，平均比特率降低约30%，同时保持了可比的渲染质量。
cs.CV / 52 / 2609.23010
MixiMotion: One-Step Text-to-Motion Generation via Asymmetric Set Distillation
MixiMotion：通过非对称集合蒸馏实现一步文本到动作生成
Abstract
Iterative text-to-motion generation delivers high-quality and semantically aligned motions but requires multiple network evaluations, resulting in substantial inference latency. We present \textbf{MixiMotion}, a strict one-step text-to-motion generation framework based on offline set distillation. Instead of distilling a single teacher trajectory for each text prompt, MixiMotion constructs an offline bank of multiple teacher motions and aligns teacher and student sample sets through \textbf{asymmetric bidirectional matching}. The teacher-to-student direction promotes coverage of diverse teacher-supported motions, while the student-to-teacher direction suppresses unsupported generations. We further introduce differentiable decoded-space kinematic supervision to complement normalized representation matching with constraints in the decoded motion space. At inference, MixiMotion generates a complete motion sequence with a single network evaluation, without teacher queries, iterative sampling, or candidate ranking. On ViMoGen, MixiMotion achieves a semantic alignment score of $0.835$, outperforming the evaluated one-step baselines and approaching the $0.858$ score of its 50-step HY-Motion-1.0-Lite teacher. In blinded human evaluation, MixiMotion obtains an overall rating of $4.33$, compared with $4.50$ for the teacher, while outperforming the evaluated one-/few-step baselines. Meanwhile, generation latency is reduced from $829.58$\,ms to $9.30$\,ms, corresponding to an $89.2\times$ speedup. These results demonstrate an effective quality--efficiency trade-off for strict one-step text-to-motion generation.
Chinese Translation
迭代的文本到动作生成能够提供高质量和语义对齐的动作，但需要多次网络评估，从而导致显著的推理延迟。我们提出了 extbf{MixiMotion}，一个基于离线集合蒸馏的严格一步文本到动作生成框架。MixiMotion并不是为每个文本提示蒸馏单一的教师轨迹，而是构建了一个包含多个教师动作的离线库，并通过 extbf{非对称双向匹配}对教师和学生样本集进行对齐。教师到学生的方向促进了多样化教师支持动作的覆盖，而学生到教师的方向则抑制了不支持的生成。我们进一步引入了可微分的解码空间运动学监督，以补充在解码运动空间中的归一化表示匹配约束。在推理过程中，MixiMotion通过一次网络评估生成完整的动作序列，无需教师查询、迭代采样或候选排名。在ViMoGen上，MixiMotion达到了$0.835$的语义对齐分数，超越了评估的一步基线，并接近其50步HY-Motion-1.0-Lite教师的$0.858$分数。在盲评人类评估中，MixiMotion获得了$4.33$的整体评分，而教师为$4.50$，同时超越了评估的一步/少步基线。同时，生成延迟从$829.58$毫秒减少到$9.30$毫秒，速度提升达到$89.2 imes$。这些结果展示了严格一步文本到动作生成的有效质量-效率权衡。
cs.CV / 53 / 2609.23012
CrowdCue: Specialist-Cue Conditioning for Vision-Language Crowd Counting
CrowdCue：视觉-语言人群计数的专家提示条件化
Abstract
Generative vision-language models (VLMs) offer a counting paradigm in which one model produces both a count and a natural-language account of the scene, yet their raw counting accuracy sits in the range of sub-million-parameter specialist regressors. The open question is whether auxiliary guidance from a pretrained specialist can lift them into useful territory, and through which channel that guidance is best routed. We evaluate Qwen2.5-VL-7B on four widely used crowd counting benchmarks (ShanghaiTech A and B, UCF-QNRF, NWPU-Crowd). Zero-shot prompting rarely produces a parseable count, so LoRA supervised fine-tuning establishes the baseline at overall MAE 81.64. Conditioning on a P2PNet-derived density heatmap as an auxiliary visual signal fails in every encoding we tested, and an adversarial-swap protocol shows the model reads the heatmap but applies it counterproductively. We propose CrowdCue, a family that supplies the same specialist's already-integrated integer count to the VLM as a discrete symbol. The text-channel variant reaches MAE 72.04. The visual-channel variant, which renders the integer as printed digits and supplies it as a second image, reaches MAE 62.65, the strongest result in this paper and well ahead of the cue-supplying specialist alone (84.45 on the same split). In the late-fusion VLM we study, the binding constraint is not the channel but the abstraction level at which the specialist signal is delivered.
Chinese Translation
生成性视觉-语言模型（VLMs）提供了一种计数范式，其中一个模型同时生成计数和场景的自然语言描述，但其原始计数精度仅处于百万参数以下的专家回归模型的范围内。未解的问题是，预训练专家的辅助指导是否能够将其提升到有用的领域，以及通过哪个渠道最佳地传递这种指导。我们在四个广泛使用的人群计数基准（ShanghaiTech A 和 B、UCF-QNRF、NWPU-Crowd）上评估了 Qwen2.5-VL-7B。零-shot 提示很少产生可解析的计数，因此 LoRA 监督微调将基线设定为整体 MAE 81.64。基于 P2PNet 导出的密度热图作为辅助视觉信号的条件化在我们测试的每种编码中均未成功，而对抗性交换协议表明模型读取了热图，但以反生产性的方式应用它。我们提出了 CrowdCue，一个系列将同一专家已经集成的整数计数作为离散符号提供给 VLM。文本通道变体达到 MAE 72.04。视觉通道变体将整数呈现为打印数字，并作为第二幅图像提供，达到 MAE 62.65，这是本文中最强的结果，远超仅提供提示的专家（在相同拆分上为 84.45）。在我们研究的晚融合 VLM 中，绑定约束不是通道，而是专家信号传递的抽象级别。
cs.CV / 54 / 2609.23017
Reconstructed holograms and explanation-aware evaluation for low-cost computational pollen analysis in veterinary cytology
重建全息图及解释感知评估在兽医细胞学中低成本计算花粉分析的应用
Abstract
Automated pollen analysis supports veterinary cytology, but brightfield microscopy is costlier and more complex than lens-less digital in-line holographic microscopy. We evaluate whether reconstructed holograms can narrow this gap and whether model explanations remain reliable under modality change. Six pollen species were imaged by brightfield and holographic microscopy. Raw, single back-propagation and iterative phase retrieval holograms were evaluated with YOLOv26s detection and MobileNetV4 classification after anchor-based annotation transfer. Six attribution methods were assessed for spatial grounding and faithfulness with the Attribution Health Inspection and Repair (AHIR) protocol, which tests model brittleness under weak noise and corrects attribution-map granularity when needed. Brightfield achieved 0.6890 mAP50-95 (0.8865 mAP50) for detection and 0.9687 macro-F1 (0.9705 accuracy) for classification. Reconstructed holograms narrowed the gap with a task-dependent split: p-type was strongest for detection at 0.5324 mAP50-95 (0.8229 mAP50), while r-type was strongest for classification at 0.7695 macro-F1 (0.7866 accuracy), both far above raw-hologram baselines. Activation-based explanations localized strongly on grains, and region-based methods retained ~60 to ~80% of faithfulness under holography. The holographic detector was highly brittle to weak perturbations, saturating deletion-based evaluation while insertion remained informative. Pixel-level gradient explanations approached random floor, yet spatial smoothing restored p-type gradient faithfulness from 0.05 to 0.51. For holographic classification, perturbation-based explanations remained faithful while gradient-based methods fell below random floor. Reconstruction improves low-cost holographic pollen analysis, while AHIR distinguishes genuine attribution failure from artifacts caused by model brittleness and map granularity.
Chinese Translation
自动化花粉分析支持兽医细胞学，但明场显微镜的成本更高且比无镜数字在线全息显微镜更复杂。我们评估重建全息图是否能够缩小这一差距，以及在模态变化下模型解释的可靠性。六种花粉物种通过明场显微镜和全息显微镜进行成像。经过基于锚点的注释转移后，原始、单次反向传播和迭代相位恢复全息图通过YOLOv26s检测和MobileNetV4分类进行评估。采用归因健康检查与修复（AHIR）协议评估六种归因方法的空间基础和真实性，该协议测试模型在弱噪声下的脆弱性，并在需要时修正归因图的粒度。明场显微镜在检测中实现了0.6890 mAP50-95（0.8865 mAP50），在分类中实现了0.9687宏F1（0.9705准确率）。重建全息图通过任务依赖的划分缩小了这一差距：p型在检测中表现最佳，达到0.5324 mAP50-95（0.8229 mAP50），而r型在分类中表现最佳，达到0.7695宏F1（0.7866准确率），均远高于原始全息图基线。基于激活的解释在颗粒上具有强烈的局部化，而基于区域的方法在全息图下保持了约60%至80%的真实性。全息检测器对弱扰动高度脆弱，基于删除的评估饱和，而插入仍然具有信息性。像素级梯度解释接近随机底线，但空间平滑将p型梯度的真实性从0.05恢复到0.51。在全息分类中，基于扰动的解释保持了真实性，而基于梯度的方法则低于随机底线。重建改善了低成本全息花粉分析，而AHIR则区分了真正的归因失败与模型脆弱性和图粒度造成的伪影。
cs.CV / 55 / 2609.23026
BrainIAC: Interactive 3D Brain Lesion Segmentation across Heterogeneous MRI Modalities with Online Adaptation
BrainIAC：跨异构MRI模态的交互式3D脑损伤分割与在线适应
Abstract
Brain lesion segmentation is a fundamental task in medical image analysis, playing a critical role in diagnosis, treatment planning, and longitudinal disease monitoring. Yet existing models still struggle to meet the demands of real clinical use, where deployments contain data distribution shifts, arising from differences in scanner hardware, imaging protocol (varying MRI modality sets), and new pathologies. We present BrainIAC (Brain lesion Interactive Adaptive Continuously learning segmentation), a unified framework that integrates (i) a multi-modal backbone network trained to segment multiple types of brain lesions and handle heterogeneous sets of modalities via zero-filling and random modality dropping; (ii) 3D interactive segmentation with bounding-box and click prompts that preserves fully automatic prediction when no prompt is given; and (iii) an online adaptation mechanism combining Mid-Interaction adaptation and Post-Interaction adaptation, supervised by the network's own predictions as pseudo labels and guided by an extra Click-Centered Gaussian loss. To our knowledge, this is one of the first 3D online adaptation methods for interactive segmentation, and the first to combine handling of heterogeneous modality sets with online adaptation. Experiments across seven brain MRI datasets demonstrate that the proposed components provide complementary and synergistic benefits. The method consistently outperforms existing approaches and generalizes well across heterogeneous imaging modalities, including those unseen during training, as well as previously unseen brain pathology types. The code and a 3D Slicer plug-in will be released at https://github.com/WenTXuL/BrainIAC upon publication.
Chinese Translation
脑损伤分割是医学图像分析中的一项基础任务，在诊断、治疗规划和长期疾病监测中发挥着关键作用。然而，现有模型仍难以满足实际临床使用的需求，因为在部署过程中会出现数据分布的变化，这些变化源于扫描仪硬件的差异、成像协议（不同的MRI模态集）以及新的病理情况。我们提出了BrainIAC（脑损伤交互式自适应持续学习分割），这是一个统一框架，集成了（i）一个多模态主干网络，旨在分割多种类型的脑损伤，并通过零填充和随机模态丢弃处理异构模态集；（ii）带有边界框和点击提示的3D交互式分割，当未给出提示时保留完全自动预测；以及（iii）结合中间交互适应和后交互适应的在线适应机制，以网络自身的预测作为伪标签进行监督，并通过额外的点击中心高斯损失进行引导。据我们所知，这是首个用于交互式分割的3D在线适应方法，也是首个将异构模态集处理与在线适应相结合的方法。在七个脑MRI数据集上的实验表明，所提出的组件提供了互补和协同的好处。该方法在性能上始终优于现有方法，并且在异构成像模态上具有良好的泛化能力，包括在训练期间未见过的模态以及先前未见过的脑病理类型。代码和3D Slicer插件将在出版后发布于https://github.com/WenTXuL/BrainIAC。
cs.CV / 56 / 2609.23049
VDGS: Visibility-Driven Large-Scale 3D Gaussian Splatting for Aerial Scene Reconstruction
VDGS：基于可见性驱动的大规模3D高斯点云重建用于空中场景重建
Abstract
Large-scale scene reconstruction is a critical foundational technology in robotic autonomous systems such as 3D mapping and autonomous driving. In recent years, 3D Gaussian Splatting (3DGS) has demonstrated remarkable advantages in both visual quality and computational efficiency, making it a promising representation for large-scale scene reconstruction. However, it still faces challenges in large-scale scenes, including excessive memory consumption and uneven viewpoint coverage caused by UAV acquisition, limiting its real-world applications. To address this, we propose VDGS, a novel 3DGS framework that incorporates camera distribution into scene modeling. VDGS introduces visibility-driven statistics for scene anchors to quantify supervision strength. These statistics are further leveraged for scene partitioning and for gradient compensation in under-optimized regions, thereby promoting balanced optimization across different regions. Extensive experiments on multiple large-scale aerial scene datasets demonstrate that, under imbalanced viewpoint distributions, VDGS consistently outperforms existing methods, while maintaining competitive performance in scenarios with more uniform view distributions.
Chinese Translation
大规模场景重建是机器人自主系统中的一项关键基础技术，广泛应用于3D地图构建和自动驾驶等领域。近年来，3D高斯点云重建（3D Gaussian Splatting, 3DGS）在视觉质量和计算效率方面表现出显著优势，使其成为大规模场景重建的有前景的表示方法。然而，在大规模场景中，它仍面临着过高的内存消耗和无人机（UAV）获取导致的视角覆盖不均等挑战，限制了其在实际应用中的效果。为了解决这些问题，我们提出了VDGS，一种新颖的3DGS框架，将相机分布纳入场景建模中。VDGS引入了基于可见性的统计数据，用于量化场景锚点的监督强度。这些统计数据进一步用于场景分割和在未优化区域的梯度补偿，从而促进不同区域之间的平衡优化。在多个大规模空中场景数据集上的广泛实验表明，在视角分布不均的情况下，VDGS始终优于现有方法，同时在视角分布更均匀的场景中保持竞争力。
cs.CV / 57 / 2609.23061
HDMamba-YOLO: Efficient State-Space Perception and Local Spatial Reconstruction for UAV Small Object
HDMamba-YOLO：无人机小物体的高效状态空间感知与局部空间重建
Abstract
Small-object detection in UAV imagery is challenged by weak visual evidence, ambiguous boundaries, dense object distributions, and complex backgrounds. Effective detection therefore requires long-range contextual information for target-background discrimination while preserving explicit local two-dimensional structures for accurate localization. These requirements arise at different stages of the detection pipeline and are not naturally addressed by a uniform feature-processing strategy. We propose Hybrid Dual-domain Mamba-YOLO (HDMamba-YOLO), a stage-wise heterogeneous SSM-CNN detector organized according to a perception-reconstruction-alignment-interaction rationale. EfficientVMamba-based EVSS establishes long-range contextual perception in the backbone, while PhasePatchMerging2D provides phase-aware hierarchical transitions. DST-Wrapper and Native C3k2-ASSAF then perform perception-to-reconstruction transition and repeated local two-dimensional reconstruction during FPN/PAN aggregation. DySample provides content-adaptive cross-scale resampling, while OS-CVTIA introduces macro-micro interaction and task-specific modulation for localization and classification. On VisDrone2019, HDMamba-YOLO-B achieves 42.737% mAP50 and 25.713% mAP50:95 with 10.042M parameters and 29.879 corrected GFLOPs. HDMamba-YOLO-Lite achieves 41.140% mAP50 and 24.741% mAP50:95 with 5.344M parameters. Under the unified AI-TOD evaluation protocol, HDMamba-YOLO-B obtains 21.621% AP and 47.881% AP50. Controlled ablations further support the stage-wise allocation of state-space perception, convolutional reconstruction, dynamic alignment, and task interaction for UAV small-object detection.
Chinese Translation
无人机影像中的小物体检测面临着弱视觉证据、模糊边界、密集物体分布和复杂背景等挑战。因此，有效的检测需要长距离的上下文信息以实现目标与背景的区分，同时保持明确的局部二维结构以实现准确定位。这些需求在检测流程的不同阶段出现，且并非通过统一的特征处理策略自然解决。我们提出了混合双域Mamba-YOLO（HDMamba-YOLO），这是一种根据感知-重建-对齐-交互逻辑组织的阶段性异构SSM-CNN检测器。基于EfficientVMamba的EVSS在主干网络中建立了长距离的上下文感知，而PhasePatchMerging2D则提供了相位感知的分层过渡。DST-Wrapper和Native C3k2-ASSAF随后在FPN/PAN聚合过程中执行感知到重建的过渡以及重复的局部二维重建。DySample提供了内容自适应的跨尺度重采样，而OS-CVTIA则引入了宏观-微观交互和任务特定的调制以实现定位和分类。在VisDrone2019数据集上，HDMamba-YOLO-B达到了42.737%的mAP50和25.713%的mAP50:95，参数量为10.042M，修正后的GFLOPs为29.879。HDMamba-YOLO-Lite达到了41.140%的mAP50和24.741%的mAP50:95，参数量为5.344M。在统一的AI-TOD评估协议下，HDMamba-YOLO-B获得了21.621%的AP和47.881%的AP50。控制性消融实验进一步支持了状态空间感知、卷积重建、动态对齐和任务交互在无人机小物体检测中的阶段性分配。
cs.CV / 58 / 2609.23067
LD-RSVIS: A Large-Scale and Diverse Benchmark for Referring Surgical Video Instrument Segmentation
LD-RSVIS：一个大规模且多样化的手术视频工具分割基准
Abstract
Referring surgical video instrument segmentation (RSVIS) aims at segmenting the instrument in a surgical video, given a textual description. Despite recent progress, current models are trained and assessed on relatively small-scale benchmarks, hindering the development of more general RSVIS. In addition, existing benchmarks support only the single-target expression that refers to one instrument in the video, while overlooking multi-target and no-target referring expressions, restricting the applicability of RSVIS in practical scenarios. Addressing these issues, we propose LD-RSVIS, a new benchmark aiming to facilitate more robust and general RSVIS. Specifically, LD-RSVIS consists of 3,536 surgical videos with 1.09 million frames and covers a broad set of 30 instrument classes from 25 various procedures. By including abundant videos and classes, LD-RSVIS could benefit large-scale training and evaluation of more general RSVIS methods. Besides, unlike existing datasets, LD-RSVIS offers diverse referring settings, including no-target, single-target, and multi-target expressions, which enables the development of more practical RSVIS models in real applications. In order to ensure high-quality annotations, all videos in LD-RSVIS are manually labeled with multiple rounds of inspection and refinement. To our knowledge, LD-RSVIS is the largest and most diverse benchmark for RSVIS. To analyze LD-RSVIS and to provide comparison for future research, we evaluate 12 representative methods, and the results reveal that more efforts are required for improvements. To encourage future research, we present a simple yet effective RSVIS method, dubbed Cascade-RSVIS, that first mines target-specific cues using the complementary multi-cue text information and then employs such cues and textual information for segmentation, achieving promising performance. Our benchmark and code will be released.
Chinese Translation
手术视频工具分割（RSVIS）旨在根据文本描述对手术视频中的工具进行分割。尽管最近取得了一些进展，但当前模型的训练和评估主要基于相对小规模的基准，这限制了更通用的RSVIS的发展。此外，现有基准仅支持单目标表达，即指向视频中的一个工具，而忽视了多目标和无目标的指代表达，这限制了RSVIS在实际场景中的适用性。为了解决这些问题，我们提出了LD-RSVIS，一个新的基准，旨在促进更稳健和通用的RSVIS。具体而言，LD-RSVIS包含3,536个手术视频，涵盖1.09百万帧，并覆盖25种不同手术中的30个工具类别。通过包含丰富的视频和类别，LD-RSVIS可以促进更大规模的通用RSVIS方法的训练和评估。此外，与现有数据集不同，LD-RSVIS提供了多样化的指代设置，包括无目标、单目标和多目标表达，从而使得在实际应用中开发更实用的RSVIS模型成为可能。为了确保高质量的标注，LD-RSVIS中的所有视频均经过多轮人工标注和检查。根据我们的了解，LD-RSVIS是目前最大的、最具多样性的RSVIS基准。为了分析LD-RSVIS并为未来的研究提供比较，我们评估了12种具有代表性的方法，结果显示仍需更多努力进行改进。为了鼓励未来的研究，我们提出了一种简单而有效的RSVIS方法，称为Cascade-RSVIS，该方法首先利用互补的多线索文本信息挖掘目标特定线索，然后利用这些线索和文本信息进行分割，取得了良好的性能。我们的基准和代码将会发布。
cs.CV / 59 / 2609.23121
MM-ContextFold: Context Folding for Multimodal Agentic Retrieval
MM-ContextFold：用于多模态代理检索的上下文折叠
Abstract
Multimodal Agentic Retrieval (MAR) requires agents to solve complex information-seeking tasks by iteratively invoking external tools. Typical frameworks such as ReAct maintain raw multimodal inputs and the accumulating interaction history in a single, ever-growing context, leading to the context explosion problem. While existing methods alleviate this issue by compressing redundant text, effective strategies for managing token-intensive visual content remain largely underexplored. To address this gap, we first conduct a systematic empirical study of approximately 10,000 trajectories. The results show that as visual cues are progressively extracted through external tools and textualized into the context, raw images become increasingly redundant. Continued image retention is associated with higher output entropy and can even degrade task accuracy. Motivated by these findings, we propose MM-ContextFold, a training-free framework that loads raw images only when needed. It maintains a persistent, text-only main context for high-level planning and spawns ephemeral branch contexts for image-dependent subtasks. Within each branch, the agent loads the relevant images, completes the subtask, and folds the result back into the main context as a concise textual summary; the images and branch trace are then discarded. Experiments on seven MAR benchmarks across five backbone models show that MM-ContextFold improves average accuracy by 6.3 percentage points over ReAct while reducing the working context length by 27.5\%.
Chinese Translation
多模态代理检索（MAR）要求代理通过迭代调用外部工具来解决复杂的信息检索任务。典型框架如 ReAct 将原始多模态输入和不断累积的交互历史维持在一个不断增长的上下文中，导致上下文爆炸问题。虽然现有方法通过压缩冗余文本来缓解这一问题，但有效管理大量视觉内容的策略仍然未得到充分探索。为了解决这一空白，我们首先对约 10,000 条轨迹进行了系统的实证研究。结果表明，随着视觉线索通过外部工具逐步提取并文本化到上下文中，原始图像变得越来越冗余。持续保留图像与更高的输出熵相关，甚至可能降低任务准确性。基于这些发现，我们提出了 MM-ContextFold，这是一种无训练的框架，仅在需要时加载原始图像。它为高层次规划维护一个持久的仅文本主上下文，并为依赖图像的子任务生成短暂的分支上下文。在每个分支中，代理加载相关图像，完成子任务，并将结果折叠回主上下文，作为简洁的文本摘要；然后丢弃图像和分支痕迹。在五个主干模型的七个 MAR 基准上的实验表明，MM-ContextFold 将平均准确率提高了 6.3 个百分点，同时将工作上下文长度减少了 27.5%。
cs.CV / 60 / 2609.23139
QwenVLConnector: A Fast, Unified Medical VLM Chatbot for Fine-Grained Clinical Perception and Text Generation
QwenVLConnector：一个快速统一的医疗视觉语言模型聊天机器人，用于细粒度临床感知和文本生成
Abstract
Most medical vision-language models (VLMs) excel at open-ended report generation and VQA but provide limited support for structured, fine-grained clinical perception within a unified interface. We present QwenVLConnector, a Qwen2.5-VL-based medical chatbot that unifies classification, multi-label classification, textualized detection, counting, regression, and free-form report generation under a single next-token objective. Our key component is a lightweight dense multi-layer Connector that aggregates low- and high-level visual features, aligns them through the pretrained vision Merger, and fuses them with the final visual representation without increasing sequence length. This design enriches visual tokens with complementary spatial and semantic cues while preserving efficiency. On FLARE-2D, QwenVLConnector improves detection F1 from 0.55 to 0.85, raises single-label classification from 0.37 to 0.51, and boosts report-generation GREEN by up to 18.3 points over the Qwen2.5-VL baseline. We further explore multimodal in-context learning for report generation, showing additional improvements without updating model parameters. Overall, QwenVLConnector offers a unified and efficient framework for combining structured medical perception with open-ended clinical text generation. Our code can be found at https://github.com/plnguyen2908/QwenConnector.
Chinese Translation
大多数医疗视觉语言模型（VLM）在开放式报告生成和视觉问答（VQA）方面表现优异，但在统一接口下对结构化、细粒度的临床感知支持有限。我们提出了QwenVLConnector，这是一个基于Qwen2.5-VL的医疗聊天机器人，统一了分类、多标签分类、文本化检测、计数、回归和自由形式报告生成，采用单一的下一个标记目标。我们的关键组件是一个轻量级的密集多层连接器，它聚合了低层和高层视觉特征，通过预训练的视觉合并器对齐这些特征，并在不增加序列长度的情况下将其与最终的视觉表示融合。该设计丰富了视觉标记，提供了互补的空间和语义线索，同时保持了效率。在FLARE-2D数据集上，QwenVLConnector将检测F1从0.55提高到0.85，将单标签分类从0.37提升到0.51，并使报告生成的GREEN相较于Qwen2.5-VL基线提高了多达18.3分。我们进一步探索了多模态上下文学习在报告生成中的应用，展示了在不更新模型参数的情况下的额外改进。总体而言，QwenVLConnector提供了一个统一且高效的框架，将结构化医疗感知与开放式临床文本生成相结合。我们的代码可以在https://github.com/plnguyen2908/QwenConnector找到。
cs.CV / 61 / 2609.23153
SparkDiffusion: Mitigating the High-Sparsity Trap --- A Unified Framework for up to $265\times$ Single-GPU Acceleration of Visual Generation
SparkDiffusion：缓解高稀疏性陷阱——一个统一框架，实现高达 $265 imes$ 的单GPU视觉生成加速
Abstract
Video diffusion transformers are expensive because attention dominates long spatiotemporal token sequences. We identify the \emph{high-sparsity trap}: at extreme attention sparsity, step-local training losses keep decreasing while terminal generation quality stagnates or degrades. The trap is one of supervision: the dominant terminal errors originate in the high-noise structure-generation stage, and terminal-aligned training corrects terminal errors that substantially extended step-local training cannot. This yields a simple staging principle: \emph{first adapt the sparse architecture into a coarse prior, then correct the terminal distribution}. We instantiate the principle as \method, a unified acceleration framework for visual generation that combines a short sparse warm-up, few-step trajectory-mixed distillation, and FP8 quantization with fused kernels. \method sustains $97\%$ attention sparsity with strong visual quality on long-sequence 720P generation across Wan2.1/Wan2.2 backbones and T2V/I2V tasks, and $90\%$ sparsity on Wan2.1-T2V-1.3B-480P. With 3-step CFG-free inference, \method achieves a $265\times$ end-to-end speedup over the 50-step CFG dense baseline for Wan2.1-T2V-14B-720P on a single RTX~5090 ($220\times$ on H100), and denoises a Wan2.1-T2V-1.3B-480P video in $1.3$s.
Chinese Translation
视频扩散变换器的计算成本高昂，因为注意力机制主导了长时空令牌序列。我们识别出 extit{高稀疏性陷阱}：在极端注意力稀疏的情况下，局部训练损失持续下降，而终端生成质量停滞或下降。这个陷阱是监督的结果：主导的终端错误源于高噪声的结构生成阶段，而终端对齐训练则纠正了局部训练无法显著解决的终端错误。这产生了一个简单的分阶段原则： extit{首先将稀疏架构调整为粗略的先验，然后纠正终端分布}。我们将这一原则具体化为 extit{SparkDiffusion}，这是一个统一的视觉生成加速框架，结合了短暂的稀疏热身、少步轨迹混合蒸馏和与融合内核的FP8量化。 extit{SparkDiffusion}在Wan2.1/Wan2.2主干和T2V/I2V任务中，保持了$97\%$的注意力稀疏性，并在长序列720P生成中展现出强大的视觉质量，而在Wan2.1-T2V-1.3B-480P中则保持$90\\%$的稀疏性。通过3步无CFG推理， extit{SparkDiffusion}在单个RTX~5090上实现了对Wan2.1-T2V-14B-720P的50步CFG稠密基线的$265 imes$端到端加速（在H100上为$220 imes$），并在$1.3$s内去噪Wan2.1-T2V-1.3B-480P视频。
cs.CV / 62 / 2609.23161
An Eternal Irradiance Camera
永恒辐照相机
Abstract
A conventional camera uses millions of pixels to measure radiance from all directions within its field of view. We present an omnidirectional irradiance camera that measures the irradiance function---the illumination incident upon every point on a sphere. The irradiance function varies smoothly over the sphere and hence is bandlimited. We have analyzed this function in the frequency domain and have shown that it is well approximated by a weighted sum of the first seven degrees of spherical harmonics. As a result, the irradiance function can be accurately reconstructed from a small number of samples. This implies that an irradiance camera does not need millions of detectors (pixels)---just a handful of measurements suffice. This brings two major benefits. First, the camera consumes such little energy that it can be completely powered by the light falling on its detectors. Second, it does not capture the visual details needed to identify an individual, and hence privacy is preserved. We have built a prototype irradiance camera, called FluxCam, using 49 detectors arranged on the surface of a sphere. In a well-lit indoor environment, FluxCam can read out and wirelessly transmit its measurements at 30 frames per second using energy harvested from the light falling on it (i.e., without a battery, cable, or external power supply). We show how FluxCam can be used as an optical gyroscope for computing rotation, to monitor a workspace, as an untethered light probe for diffuse relighting, and as an omnidirectional pyranometer for estimating sky conditions and determining the best orientation of a solar panel.
Chinese Translation
传统相机使用数百万个像素来测量其视场内各个方向的辐射强度。我们提出了一种全向辐照相机，它测量辐照度函数——即照射到球面上每一点的光照。辐照度函数在球面上平滑变化，因此是带限的。我们在频域中分析了该函数，并展示它可以通过前七个阶的球面谐波的加权和进行良好的近似。因此，辐照度函数可以通过少量样本准确重建。这意味着辐照相机不需要数百万个探测器（像素）——只需少量测量即可。这带来了两个主要好处。首先，相机的能耗极低，可以完全由落在其探测器上的光源供电。其次，它不会捕捉到识别个体所需的视觉细节，从而保护了隐私。我们构建了一个原型辐照相机，称为FluxCam，使用49个探测器排列在球面上。在光线充足的室内环境中，FluxCam可以以每秒30帧的速度读取并无线传输其测量结果，所需能量来自落在其上的光（即无需电池、线缆或外部电源）。我们展示了FluxCam如何用作计算旋转的光学陀螺仪、监控工作空间、作为无绳光探针进行漫反射重光照，以及作为全向辐射计估计天空条件并确定太阳能电池板的最佳朝向。
cs.CV / 63 / 2609.23169
UltraTex: Unleashing 2K Multi-View Diffusion for 3D Texturing
UltraTex：释放2K多视角扩散用于3D纹理生成
Abstract
High-quality texture generation is essential for creating realistic and production-ready 3D assets. Recent multi-view diffusion methods have shown promising results for image-guided 3D texturing, but they are typically constrained to low operating resolutions such as 512 or 768, making it difficult to preserve high-frequency details from high-resolution reference images. Scaling this paradigm to 2048 resolution is computationally prohibitive, as the unified multi-view sequence exceeds 212K tokens and incurs excessive memory and latency. In this paper, we present UltraTex, an efficient end-to-end framework for high-resolution multi-view diffusion-based 3D texturing. Our key observation is that object-centric multi-view renderings contain two major sources of redundancy: background-induced sequence redundancy and sparse token interactions within the foreground. To address them, we introduce Background Token Dropping, which removes background tokens before the DiT backbone, and Block-Sparse Attention, which reduces attention computation over the retained foreground sequence. To enable efficient foreground-only inference while avoiding reconstruction artifacts, we further design Foreground-Aware VAE Decoding to ensure the quality of the final high-resolution views. To satisfy the demanding data requirements of 2K-resolution multi-view diffusion training, we construct G-buffer TexVerse, a large-scale, ultra-high-resolution multi-view rendering dataset covering over 268,000 3D assets. Extensive experiments show that UltraTex generates visually faithful textures with rich fine-grained details, while substantially improving efficiency, achieving $20.6\times$--$91.1\times$ training speedup and $22.3\times$--$74.6\times$ end-to-end inference speedup over the baseline on common samples in our dataset. Code and data is at https://yiboz2001.github.io/UltraTex.
Chinese Translation
高质量纹理生成对于创建真实且适合生产的3D资产至关重要。近期的多视角扩散方法在图像引导的3D纹理生成中显示出良好的结果，但通常受到512或768等低操作分辨率的限制，这使得从高分辨率参考图像中保留高频细节变得困难。将这一范式扩展到2048分辨率在计算上是不可行的，因为统一的多视角序列超过212K个标记，导致过高的内存和延迟。在本文中，我们提出了UltraTex，一种高效的端到端框架，用于高分辨率多视角扩散基础的3D纹理生成。我们的关键观察是，面向对象的多视角渲染包含两个主要的冗余来源：背景引起的序列冗余和前景中稀疏的标记交互。为了解决这些问题，我们引入了背景标记丢弃（Background Token Dropping），在DiT主干之前移除背景标记，以及块稀疏注意力（Block-Sparse Attention），减少对保留前景序列的注意力计算。为了实现高效的仅前景推理并避免重建伪影，我们进一步设计了前景感知变分自编码器解码（Foreground-Aware VAE Decoding），以确保最终高分辨率视图的质量。为了满足2K分辨率多视角扩散训练的高数据需求，我们构建了G-buffer TexVerse，这是一个大规模的超高分辨率多视角渲染数据集，覆盖超过268,000个3D资产。大量实验表明，UltraTex生成的纹理在视觉上忠实且细节丰富，同时显著提高了效率，在我们数据集中的常见样本上实现了$20.6 imes$--$91.1 imes$的训练加速和$22.3 imes$--$74.6 imes$的端到端推理加速。代码和数据可在 https://yiboz2001.github.io/UltraTex 获取。
cs.CV / 64 / 2609.23182
GrapeSplat: Geometry-Grounded Reconstruction via Amalgamated Pose-Free Encoding for Feed-Forward 3D Gaussian Splatting
GrapeSplat：通过融合无姿态编码的几何基础重建实现前馈3D高斯点云
Abstract
Feed-forward 3D Gaussian Splatting now reconstructs renderable scenes from unposed, uncalibrated images. Yet, most models supervise only photometric consistency and predict Gaussians pixel by pixel, which leaves global structure fragile and ties primitive count to image resolution and view count. To this end, GrapeSplat amalgamates multi-view cues into a voxel-aligned scene representation and decodes Gaussians directly from the learned grid, requiring no per-scene optimization or post-processing. An Atlas Encoder lifts all views into pixel-wise geometry-and-appearance features anchored at predicted 3D points. PEACH-Vox compands the unbounded scene into a bounded sparse grid through a smooth per-axis map with an exact closed-form inverse. The Sparse Decoder then consolidates the grid with sparse convolutions and decodes the full scene as multiple Gaussians per occupied cell. This amalgamated representation exploits sparse voxel occupancy, where the Gaussian count follows the occupied cells and saturates as views cover the scene, while grid resolution sets its ceiling. GrapeSplat turns unposed images into a renderable Gaussian scene in a single forward pass. Trained with 2D and 3D supervision on 8-view sequences, it generalizes zero-shot from 4 to 64 views across indoor and unbounded scenes. Code and trained weights are available at https://github.com/VAISR/GrapeSplat
Chinese Translation
前馈3D高斯点云现已能够从无姿态、未校准的图像中重建可渲染场景。然而，大多数模型仅监督光度一致性，并逐像素预测高斯，这使得全局结构脆弱，并将原始体素数量与图像分辨率和视图数量绑定在一起。为此，GrapeSplat将多视角线索融合为体素对齐的场景表示，并直接从学习到的网格中解码高斯，无需对每个场景进行优化或后处理。Atlas编码器将所有视图提升为锚定在预测3D点的逐像素几何和外观特征。PEACH-Vox通过具有精确闭式逆映射的平滑轴向映射将无界场景压缩为有界稀疏网格。稀疏解码器随后通过稀疏卷积整合网格，并将完整场景解码为每个占用单元多个高斯。这种融合表示利用了稀疏体素占用，其中高斯数量随占用单元而变化，并随着视图覆盖场景而饱和，而网格分辨率则设定了其上限。GrapeSplat将无姿态图像转化为可渲染的高斯场景，仅需一次前向传递。在对8视图序列进行2D和3D监督训练后，它能够在室内和无界场景中实现从4到64视图的零样本泛化。代码和训练权重可在 https://github.com/VAISR/GrapeSplat 获取。
cs.CV / 65 / 2609.23184
CausalWM: Causal Chain-of-Thought Reasoning for Embodied World Model
CausalWM：用于具身世界模型的因果链式思维推理
Abstract
Embodied world models learn to predict future physical dynamics from visual observations and control signals, where physical knowledge is implicitly entangled within latent representations. We introduce CausalWM, a 16B embodied world model that performs explicit causal chain-of-thought reasoning before future video prediction. CausalWM organizes useful variables into a reasoning trajectory, allowing the model to progressively capture causal dependencies underlying physical evolution. To train CausalWM, we collect 31K hours embodied data and develop a three-stage paradigm consisting of large-scale video pre-training, causal CoT mid-training, and multi-objective RL post-training. Despite using only a limited set of supervised CoT variables, CausalWM exhibits emergent in-context learning capabilities, enabling contextual visual feature guidance and efficient few-step generation. CausalWM achieves state-of-the-art performance across language-conditioned, action-conditioned, single-view and multi-view benchmarks, including Top-1 performance on TriWorldBench leaderboard.
Chinese Translation
具身世界模型通过视觉观察和控制信号学习预测未来的物理动态，其中物理知识隐含在潜在表示中。我们提出了CausalWM，一个具有16B参数的具身世界模型，它在未来视频预测之前执行显式的因果链式思维推理。CausalWM将有用的变量组织成推理轨迹，使模型能够逐步捕捉物理演变背后的因果依赖关系。为了训练CausalWM，我们收集了31K小时的具身数据，并开发了一个由大规模视频预训练、因果链式思维中期训练和多目标强化学习后期训练组成的三阶段范式。尽管仅使用有限的监督链式思维变量，CausalWM展现出新兴的上下文学习能力，使其能够进行上下文视觉特征指导和高效的少步生成。CausalWM在语言条件、动作条件、单视图和多视图基准测试中均达到了最先进的性能，包括在TriWorldBench排行榜上的Top-1表现。
cs.CV / 66 / 2609.23248
SPACE: Semantic Projection and Alignment of CLIP Embeddings for Domain Adaptation
SPACE：CLIP 嵌入的语义投影与对齐用于领域适应
Abstract
A fundamental challenge in deploying vision models is domain shift, which arises when training and test data follow different distributions, leading to degraded performance. This challenge is amplified when the same semantic concept appears under distinct visual forms, such as photographs and sketches, where visual similarity is weak despite semantic correspondence. Existing unsupervised domain-adaptation methods aim to align distributions across domains but often ignore semantic relationships among samples of the same class. To address this issue, this paper introduces SPACE, a method that exploits the semantic structure of CLIP's vision-language space for domain adaptation. The key idea is to use text descriptions as semantic anchors by applying Singular Value Decomposition to CLIP embeddings of class descriptions, yielding an orthogonal basis that captures semantic relationships among categories. Visual features from both domains are projected into this semantic subspace, aligning images based on meaning rather than appearance.
Chinese Translation
在部署视觉模型时，一个基本挑战是领域转移，这种转移发生在训练数据和测试数据遵循不同分布时，导致性能下降。当相同的语义概念以不同的视觉形式出现时，例如照片和素描，尽管语义对应，但视觉相似性较弱，这一挑战更加明显。现有的无监督领域适应方法旨在对齐不同领域的分布，但往往忽视了同类样本之间的语义关系。为了解决这一问题，本文提出了 SPACE，一种利用 CLIP 的视觉-语言空间的语义结构进行领域适应的方法。其关键思想是通过对 CLIP 类描述的嵌入应用奇异值分解，将文本描述作为语义锚点，从而获得一个正交基，捕捉类别之间的语义关系。来自两个领域的视觉特征被投影到这个语义子空间中，基于意义而非外观对齐图像。
cs.CV / 67 / 2609.23249
Exact Quotients of Fresnel-Kummer Surfaces and Certified Biaxial Refraction
Fresnel-Kummer曲面的精确商与认证的双轴折射
Abstract
The Fresnel wave surface governs the propagation of light in a transparent biaxial crystal. It is a special Kummer quartic, and we identify it exactly. Over the complex numbers the wave surface of a crystal is the Kummer surface of the Jacobian of an explicit genus-two curve branched at the signed square roots of the three principal permittivities. This Jacobian is isogenous, by an isogeny with kernel of order four, to a product of two elliptic curves. One elliptic curve carries the three permittivities, and the other carries the optic-axis angle. The physical family is Zariski dense in the locus of genus-two curves with an extra involution, and its automorphism strata are explicit. The identification instantiates a task-aware quotient, which identifies parameters that differ by a nuisance transformation and carries invariant coordinates and explicit strata. For biaxial crystals, two ratios of the permittivities form a complete invariant of the wave surface up to rotation and rescaling, and the four real nodes are given in closed form. At an interface the candidate transmitted waves are the roots of a quartic of exact degree four. Its real-root count, root order, and repeated-root events are decided by exact algebraic predicates, and along the generic single-node encounters of the paper its discriminant vanishes to second order. Floating-point solvers drop forward transmitted modes near the optic axes, and the certified solver does not. On exact equivalence classes, learned models on quotient coordinates are invariant and more accurate than models on raw tensors, while learned root-count predicates fail near the optic axes.
Chinese Translation
Fresnel波面控制着光在透明双轴晶体中的传播。它是一种特殊的Kummer四次曲面，我们对其进行了精确识别。在复数域上，晶体的波面是显式二 genus曲线的Jacobian的Kummer曲面，该曲线在三个主介电常数的带符号平方根处分支。该Jacobian通过一个阶数为四的同态同构于两个椭圆曲线的乘积。一个椭圆曲线承载三个介电常数，另一个承载光轴角。物理族在带有额外反演的二 genus曲线的轨迹中是Zariski稠密的，其自同构层次是显式的。该识别实例化了一个任务感知商，识别出通过干扰变换而不同的参数，并携带不变坐标和显式层次。对于双轴晶体，两个介电常数的比率构成了波面在旋转和重新缩放下的完整不变量，四个实节点以封闭形式给出。在界面上，候选的透射波是一个精确度为四的四次方程的根。其实根计数、根的顺序和重复根事件由精确的代数谓词决定，并且在本文的一般单节点相遇中，其判别式以二阶消失。浮点求解器在光轴附近丢弃前向透射模式，而认证求解器则不会。在精确等价类上，基于商坐标的学习模型是不变的，并且比基于原始张量的模型更准确，而学习的根计数谓词在光轴附近失效。
cs.CV / 68 / 2609.23268
Blind Deconvolution of Binary and Pattern Images with Pixel Intensity Constraints and Sparse Gradient Prior
具有像素强度约束和稀疏梯度先验的二进制和模式图像盲解卷积
Abstract
Blind image deconvolution (BID) is a prominent research topic in the field of imaging sciences, given its significant practical applications. Most existing model-based BID methods focus on natural images, incorporating appropriate prior knowledge about both the underlying image and the blur kernel. However, for certain classes of images, such as barcodes, text, and patterns, pixels can only take very limited values, a specific prior that is often overlooked in the literature. In this article, we introduce a novel pixel intensity constraint to leverage this important information, improving recovery performance for these specialized image classes. Specifically, we propose a unified framework for blind binary and pattern image deconvolution that incorporates both the pixel intensity constraint and a gradient sparsity regularizer. Numerical experiments demonstrate that our method outperforms many existing BID techniques, achieving superior results in terms of both visual quality and quantitative metrics.
Chinese Translation
盲图像解卷积（BID）是成像科学领域一个重要的研究课题，因其具有显著的实际应用价值。现有的大多数基于模型的BID方法主要集中于自然图像，结合了关于基础图像和模糊核的适当先验知识。然而，对于某些特定类型的图像，如条形码、文本和图案，像素只能取非常有限的值，这一特定先验在文献中常常被忽视。在本文中，我们引入了一种新的像素强度约束，以利用这一重要信息，从而提高这些特定图像类别的恢复性能。具体而言，我们提出了一个统一的框架，用于盲二进制和模式图像的解卷积，该框架结合了像素强度约束和梯度稀疏性正则化器。数值实验表明，我们的方法在视觉质量和定量指标方面均优于许多现有的BID技术，取得了更优的结果。
cs.CV / 69 / 2609.23286
RegVGGT: Sustainable Visual Geometry Grounding for Streaming via Regulated Memory
RegVGGT：通过受限记忆实现可持续视觉几何基础的流媒体处理
Abstract
3D reconstruction from a lengthy video stream input poses a dilemma for feed-forward reconstruction models (FFRMs), that a whole-stream inference context cannot be retained under limited GPU memory.Recent studies seek to resolve this problem via a trade-off between the integrity of inference context and GPU memory usage, which either suffer from a rapid memory inflation or degraded context integrity due to artificially capping memory usage.Driven by our key observation that the initial saliency of a token reliably dictates its long-term importance across the stream, we propose RegVGGT, a training-free token regulation method which aggressively regulates the tokens of incoming frames.By admitting at most 1% of tokens per frame to update the context memory, our method dramatically suppresses memory inflation as the stream progresses.Equipped with a FlashAttention-compatible token saliency estimation scheme, RegVGGT is capable of processing thousands of frames on a consumer-grade GPU with negligible compromise to reconstruction quality.Extensive experiments demonstrate that RegVGGT achieves state-of-the-art performance on long-horizon benchmarks across diverse FFRM prediction tasks, surpassing prior FFRM-based stream reconstruction baselines by a large margin.
Chinese Translation
从长视频流输入进行3D重建对前馈重建模型（FFRMs）提出了一个困境，即在有限的GPU内存下无法保留整个流的推理上下文。近期研究试图通过在推理上下文的完整性与GPU内存使用之间进行权衡来解决此问题，但要么面临快速的内存膨胀，要么由于人为限制内存使用而导致上下文完整性下降。基于我们关键观察到的初始显著性标记可靠地决定了其在整个流中的长期重要性，我们提出了RegVGGT，这是一种无训练的标记调节方法，能够积极调节输入帧的标记。通过每帧最多允许1%的标记更新上下文内存，我们的方法在流的进程中显著抑制了内存膨胀。配备兼容FlashAttention的标记显著性估计方案，RegVGGT能够在消费级GPU上处理数千帧，几乎不影响重建质量。大量实验表明，RegVGGT在各种FFRM预测任务的长时间基准测试中实现了最先进的性能，远超之前基于FFRM的流重建基线。
cs.CV / 70 / 2609.23336
MinCU: A Fine-Grained Benchmark for Grounded Minimal-Change Understanding in Image Pairs
MinCU：用于图像对中基础最小变化理解的细粒度基准
Abstract
Localizing and describing fine-grained differences between near-identical images is a critical yet underexplored capability for multimodal large language models (MLLMs). Existing benchmarks largely assess semantic comparison or single-image grounding in isolation, without jointly requiring faithful description and physical localization. To bridge this gap, we introduce MinCU, a benchmark for grounded minimal-change understanding, where each sample consists of an image pair differing by a single atomic variation in object category, attribute, count, or spatial position, and models are evaluated on their ability to describe the change, localize the changed regions, and identify the changed entity. We further propose Semantic-Guided Implicit Spatial Anchors (SG-ISA), a structured autoregressive method that decomposes prediction into a Think-Locate-Describe sequence. SG-ISA first predicts a semantic cue for the changed concept, then uses discrete spatial anchors as an implicit localization scaffold, and finally generates the change description together with the grounding box. Experiments reveal that even the strongest closed-source MLLMs and recent R1-style reasoning models struggle on MinCU, with most failing to jointly produce accurate descriptions and grounding boxes. Compared to the previous chain-of-thought method, fine-tuning with SG-ISA yields substantial joint improvements in grounding accuracy and description quality while reducing reasoning-token overhead by approximately 26%. These results suggest that an implicit intermediate spatial interface can be more effective than relying solely on model scale for grounded dual-image understanding.
Chinese Translation
在近乎相同的图像之间定位和描述细微差异是多模态大型语言模型（MLLMs）的一项关键但尚未充分探索的能力。现有基准主要评估语义比较或单图像定位，而没有共同要求准确描述和物理定位。为填补这一空白，我们引入了MinCU，这是一个用于基础最小变化理解的基准，其中每个样本由一对图像组成，这对图像在对象类别、属性、数量或空间位置上仅存在一个原子变化，模型的评估标准是其描述变化、定位变化区域和识别变化实体的能力。我们进一步提出了语义引导隐式空间锚点（SG-ISA），这是一种结构化自回归方法，将预测分解为思考-定位-描述的序列。SG-ISA首先预测变化概念的语义线索，然后使用离散空间锚点作为隐式定位支架，最后生成变化描述及其定位框。实验表明，即使是最强大的闭源MLLMs和最近的R1风格推理模型在MinCU上也面临困难，大多数模型无法共同生成准确的描述和定位框。与之前的思维链方法相比，使用SG-ISA进行微调在定位准确性和描述质量上都取得了显著的联合改善，同时将推理令牌的开销减少了约26%。这些结果表明，隐式中间空间接口可能比仅依赖模型规模在基础双图像理解中更为有效。
cs.CV / 71 / 2609.23345
AniPrO: Interpretable Anime Image Provenance Detection via Multi-Dimensional Semantic Reasoning
AniPrO：通过多维语义推理实现可解释的动漫图像来源检测
Abstract
As generative AI becomes increasingly used in anime-style image creation, distinguishing human-drawn, AI-inpainted, and text-to-image images is important for copyright attribution, visual provenance, and content governance. Existing AI-generated image detectors mainly target real-world photographs and often overlook anime-specific cues such as flat coloring, exaggerated structures, and artistic line control. To address this gap, we propose AniPrO, a multi-dimensional description-enhanced framework for interpretable anime image provenance. Built upon AnimeDL-2M, AniPrO contains 15,000 balanced samples from a 35,000-image candidate pool, covering Real, Inpainting, and Text2Image categories with structured five-dimensional descriptions. We further introduce AniPrO-SFD-Bench and AniPrO-MFR-Bench to evaluate provenance detection from statistical feature discrimination and multimodal fusion reasoning perspectives. Experiments show that structured semantic guidance reveals systematic AI-generation biases, such as the gap between global visual plausibility and local detail coherence, and improves the detection of challenging inpainting samples. The dataset and code will be released at: https://github.com/YAN-LIU05/AniPrO.
Chinese Translation
随着生成性人工智能在动漫风格图像创作中的日益普及，区分人类绘制、AI 修复和文本到图像生成的图像对于版权归属、视觉来源和内容治理变得尤为重要。现有的AI生成图像检测器主要针对现实世界的照片，往往忽视了动漫特有的线索，如平面着色、夸张结构和艺术线条控制。为了解决这一问题，我们提出了AniPrO，一个用于可解释动漫图像来源的多维描述增强框架。AniPrO基于AnimeDL-2M构建，包含来自35,000幅图像候选池的15,000个平衡样本，涵盖真实（Real）、修复（Inpainting）和文本到图像（Text2Image）类别，并具有结构化的五维描述。我们进一步引入了AniPrO-SFD-Bench和AniPrO-MFR-Bench，从统计特征区分和多模态融合推理的角度评估来源检测。实验表明，结构化语义指导揭示了系统性的AI生成偏差，例如全球视觉合理性与局部细节一致性之间的差距，并改善了对具有挑战性的修复样本的检测。数据集和代码将发布于：https://github.com/YAN-LIU05/AniPrO。
cs.CV / 72 / 2609.23352
BiView-Touch: Learning Bimanual Tactile Representations by Cross-Hand Completion
BiView-Touch：通过跨手补全学习双手触觉表征
Abstract
Bimanual interaction produces complementary tactile views of the same physical process, yet existing tactile representation learning largely models the two hands independently or combines them only for downstream prediction, leaving their cross-hand relationship unexplored. To exploit this overlooked structure, we introduce BiView-Touch, a tactile-only framework that completes masked target-hand latents from the remaining visible target-hand regions and the synchronized full contralateral hand. A student encoder with a geometry-conditioned directional decoder predicts full-view EMA latent targets, while temporal and layout counterfactuals encourage sensitivity to synchronized and anatomically organized source information. Controlled ablations and source-context interventions show that BiView-Touch learns structured cross-hand dependence on temporally aligned and anatomically organized contralateral tactile context, rather than benefiting from bilateral input alone. On the public HumanTouch dataset, its frozen representations consistently outperform representative self-supervised baselines across low-label settings. With only 5\% downstream labels, BiView-Touch achieves relative balanced-accuracy gains of 7.1\% on bilateral wrist-motion recognition and 14.1\% on force-derived interaction-phase recognition. We further introduce BVT-20, a 20-task bilateral tactile dataset, and demonstrate transfer across recording sessions and pretraining corpora, including transfer to a held-out bimanual task. Our code and dataset details are available on the anonymous project page: https://anonymous.4open.science/w/biview-touch-review-site-050C/.
Chinese Translation
双手交互产生了同一物理过程的互补触觉视图，但现有的触觉表征学习主要独立建模双手，或仅在下游预测中将其结合，未能探索其跨手关系。为了利用这一被忽视的结构，我们提出了BiView-Touch，一个仅基于触觉的框架，通过剩余可见目标手区域和同步的完整对侧手来补全被遮挡的目标手潜变量。一个带有几何条件方向解码器的学生编码器预测完整视图的EMA潜变量目标，同时时间和布局的反事实鼓励对同步和解剖组织的源信息的敏感性。控制性消融实验和源上下文干预表明，BiView-Touch学习了基于时间对齐和解剖组织的对侧触觉上下文的结构化跨手依赖，而不仅仅是从双边输入中受益。在公共的HumanTouch数据集上，其冻结的表征在低标签设置下始终优于代表性的自监督基线。在仅有5%的下游标签的情况下，BiView-Touch在双侧腕部运动识别上实现了7.1%的相对平衡准确率提升，在基于力的交互阶段识别上实现了14.1%的提升。我们进一步引入了BVT-20，一个包含20个任务的双侧触觉数据集，并展示了跨录制会话和预训练语料库的迁移能力，包括迁移到一个保留的双手任务。我们的代码和数据集详细信息可在匿名项目页面获取：https://anonymous.4open.science/w/biview-touch-review-site-050C/
cs.CV / 73 / 2609.23369
The Right Future for Action: Learning Action-Relevant Predictive States in World Action Models
行动的正确未来：在世界行动模型中学习与行动相关的预测状态
Abstract
Generation-free world action models (WAMs) retain future-video prediction during training but act from internal video features at inference, leaving unclear what these features should preserve for control. Our representation diagnostics show that representations with more predictable future changes need not make linear action decoding easier. Observed future changes provide additional action information beyond the present, and linearly readable action information is spatially concentrated. These findings motivate Action-Relevant Predictive States (ARPS), a compact predictive interface between the video and action experts. ARPS uses a horizon-conditioned state predictor to aggregate intermediate video features into a compact state that supplies all visual context to the action expert. Future-representation supervision trains different parts of this state to predict visual representations at different future times, together with their changes relative to the present. At inference, the supervision branch is removed, and the action expert only uses the learned predictive state computed from current observations. Controlled ablations show that future supervision substantially improves generalization under distribution shift. ARPS achieves 99.2% success on LIBERO and transfers to LIBERO-Plus without adaptation, reaching 87.3% and exceeding Fast-WAM by 39.2 percentage points.
Chinese Translation
无生成的世界行动模型（WAMs）在训练期间保留未来视频预测，但在推理时从内部视频特征中进行操作，这使得这些特征在控制中应保留什么变得不明确。我们的表征诊断表明，具有更可预测未来变化的表征并不一定使线性行动解码变得更容易。观察到的未来变化提供了超越当前的额外行动信息，而线性可读的行动信息在空间上是集中分布的。这些发现激励了与行动相关的预测状态（ARPS）的提出，ARPS是视频与行动专家之间的紧凑预测接口。ARPS使用一个基于时间范围的状态预测器，将中间视频特征聚合成一个紧凑状态，为行动专家提供所有视觉上下文。未来表征监督训练该状态的不同部分，以预测不同未来时刻的视觉表征及其相对于当前的变化。在推理时，监督分支被移除，行动专家仅使用从当前观察中计算得出的学习预测状态。控制消融实验表明，未来监督在分布转移下显著提高了泛化能力。ARPS在LIBERO上取得了99.2%的成功率，并在没有适应的情况下转移到LIBERO-Plus，达到87.3%，超越了Fast-WAM 39.2个百分点。
cs.CV / 74 / 2609.23372
Vision-Wireless Fusion for Multi-User Localization: A Cross-Modal Transformer Approach
视觉-无线融合用于多用户定位：一种跨模态变换器方法
Abstract
Accurate multi-user localization is challenging in complex urban environments, where wireless measurements can become ambiguous under noise, blockage, and multipath, while visual observations provide complementary spatial context. This paper presents a vision-wireless fusion framework for multi-user localization using pilot-indexed channel state information (CSI). Orthogonal pilot indices preserve the identities of the communicating UEs in the CSI token sequence and localization outputs. The model encodes each pilot-indexed CSI observation as a query token and uses cross-attention to retrieve user-specific information from spatial visual memory. Self-attention among CSI tokens further captures inter-user interactions, while the resulting multimodal representations are used for user-wise localization. Experiments on different datasets show consistent improvements over model-based, CSI-only, and multimodal-fusion baselines. Further experiments evaluate the model under different wireless and visual conditions.
Chinese Translation
在复杂的城市环境中，准确的多用户定位面临挑战，无线测量在噪声、阻塞和多径的影响下可能变得模糊，而视觉观察则提供了互补的空间上下文。本文提出了一种基于导频索引的信道状态信息（CSI）的视觉-无线融合框架，用于多用户定位。正交导频索引在CSI令牌序列和定位输出中保留了通信用户设备（UE）的身份。该模型将每个导频索引的CSI观察编码为查询令牌，并利用跨注意力机制从空间视觉记忆中检索用户特定信息。CSI令牌之间的自注意力进一步捕捉用户间的交互，而生成的多模态表示则用于用户级定位。在不同数据集上的实验显示，相较于基于模型的、仅使用CSI的和多模态融合的基线方法，均取得了一致的性能提升。进一步的实验评估了模型在不同无线和视觉条件下的表现。
cs.CV / 75 / 2609.23380
LiteTex-GS: Fast and Lightweight Texturing for Gaussian Splatting
LiteTex-GS：高效轻量的高斯点云纹理生成
Abstract
Gaussian Splatting has enabled real-time novel view synthesis, but its tightly coupled geometry and appearance representation often require a large number of primitives to reproduce high-frequency texture details, leading to substantial memory and optimization costs. Recent textured 2D Gaussian methods alleviate this limitation by attaching texture maps to Gaussian primitives. However, bridging the fundamental structural gap between discrete Gaussians and continuous 2D grids requires complex parameterizations that introduce severe computational overhead. This overhead fundamentally compromises the original efficiency of Gaussian Splatting, making the balance between detailed texturing and computational agility an unresolved challenge. To address these challenges, we propose LiteTex-GS, a fast and lightweight texturing framework for Gaussian Splatting. Our method initializes an extremely compact representation, assigning minimal local texture to each Gaussian and progressively allocates higher resolution only to primitives with significant reconstruction errors. To maintain a streamlined geometric scaffold, we introduce a contribution- and area-aware pruning strategy that eliminates low-utility Gaussians. Furthermore, to mitigate the gradient dilution caused by texture upsampling, we design a resolution-aware update rule that preserves rapid and stable convergence. Extensive experiments on standard novel view synthesis benchmarks demonstrate that our method achieves competitive or superior rendering quality while using substantially fewer parameters and less training time than existing textured Gaussian baselines.
Chinese Translation
高斯点云技术使实时新视角合成成为可能，但其几何和外观表示的紧密耦合通常需要大量原始数据来重现高频纹理细节，从而导致显著的内存和优化成本。近期的纹理化二维高斯方法通过将纹理图附加到高斯原始数据上缓解了这一限制。然而，弥合离散高斯和连续二维网格之间的基本结构差距需要复杂的参数化，这引入了严重的计算开销。这种开销从根本上削弱了高斯点云技术的原有效率，使得在详细纹理生成与计算灵活性之间取得平衡成为一个未解决的挑战。为了解决这些问题，我们提出了LiteTex-GS，一种用于高斯点云的快速轻量纹理生成框架。我们的方法初始化了一个极其紧凑的表示，为每个高斯分配最小的局部纹理，并仅对重建误差显著的原始数据逐步分配更高的分辨率。为了保持简化的几何框架，我们引入了一种基于贡献和面积的剪枝策略，以消除低效用的高斯。此外，为了减轻纹理上采样引起的梯度稀释，我们设计了一种分辨率感知的更新规则，以保持快速和稳定的收敛。对标准新视角合成基准的广泛实验表明，我们的方法在使用显著更少的参数和更短的训练时间的情况下，达到了与现有纹理化高斯基线相当或更优的渲染质量。
cs.CV / 76 / 2609.23386
ProxyBuild: Text-Guided Structured 3D Building Generation with Mesh-Anchored Procedural Proxies
ProxyBuild：基于文本引导的结构化三维建筑生成与网格锚定的程序代理
Abstract
Text-guided 3D building generation holds tremendous application potential, yet existing generative models typically output inseparable single meshes or non-interactive rendered representations. While procedural modeling can generate editable buildings with hierarchical structures, rule authoring is laborious, and even with the aid of large language models (LLMs), it remains challenging to effectively solve procedural rules under geometric constraints. In this paper, we propose ProxyBuild, a novel hybrid framework for structured building generation. We introduce the Mesh-Anchored Procedural Proxy (MAPP) as a novel intermediate representation, which tightly anchors building components onto geometric shells, thereby decoupling the generation task into two phases: proxy prediction and proxy-to-asset instantiation. First, we construct a building dataset with MAPP annotations to train our designed face-edge bigraph encoder. By explicitly modeling the feature interactions of topological elements on heterogeneous mesh graphs, this encoder accurately infers the semantic roles of faces and edges. Subsequently, conditioned on textual styles and attribute parameters parsed by LLMs, we accomplish high-precision asset retrieval and assembly by integrating a spatial placement logic with hard constraints. Extensive experiments show that ProxyBuild not only significantly mitigates common issues in building generation such as over-smoothing, component collisions, and structural corruptions, but also accurately parses semantic-free shells from diverse sources. Outperforming prior baselines across various metrics, our method can robustly generate structurally clear, detail-rich, and post-editable 3D buildings from text, thereby providing a reliable and interactive content foundation for downstream applications such as virtual reality and digital twins.
Chinese Translation
基于文本引导的三维建筑生成具有巨大的应用潜力，但现有的生成模型通常输出不可分割的单一网格或非交互式渲染表示。尽管程序建模可以生成具有层次结构的可编辑建筑，但规则编写繁琐，即使在大型语言模型（LLMs）的帮助下，仍然难以有效解决几何约束下的程序规则。在本文中，我们提出了ProxyBuild，一种用于结构化建筑生成的新型混合框架。我们引入了网格锚定程序代理（Mesh-Anchored Procedural Proxy，MAPP）作为一种新的中间表示，它将建筑组件紧密锚定在几何外壳上，从而将生成任务解耦为两个阶段：代理预测和代理到资产的实例化。首先，我们构建了一个带有MAPP注释的建筑数据集，以训练我们设计的面-边双图编码器。通过明确建模异构网格图上拓扑元素的特征交互，该编码器准确推断面和边的语义角色。随后，基于LLMs解析的文本风格和属性参数，我们通过将空间放置逻辑与硬约束相结合，实现了高精度的资产检索和组装。大量实验表明，ProxyBuild不仅显著缓解了建筑生成中常见的问题，如过度平滑、组件碰撞和结构损坏，还能够准确解析来自多种来源的无语义外壳。我们的算法在各种指标上超越了先前的基线，能够从文本中稳健地生成结构清晰、细节丰富且可后期编辑的三维建筑，从而为虚拟现实和数字双胞胎等下游应用提供可靠和交互式的内容基础。
cs.CV / 77 / 2609.23397
Enhancing Shrimp Disease Detection via Deep Learning and Data Refinement for Resilient Aquaculture
通过深度学习和数据精炼增强虾病检测以实现可持续水产养殖
Abstract
Shrimp diseases continue to cause devastating losses in the aquaculture industry, driving a critical need for robust, automated detection. This work contributes the first application of Vision Transformers (ViT) and Self-Supervised Learning (SSL) to the shrimp farming domain, addressing both performance bottlenecks and data labeling challenges. We propose two deep learning pipelines to classify four key diseases: Healthy, Black Gill (BG), White Spot Syndrome Virus (WSSV), and a co-infection of both using a dataset of 4,348 images. First, our supervised transfer-learning approach leverages ImageNet-pretrained ViT-Small/16 and EfficientNet backbones. Second, we introduce a contrastive learning framework (SimCLR) with a ViT-Small encoder to extract robust representations from unlabeled images prior to fine-tuning. Our results establish strong new baselines for sustainable aquaculture monitoring. The supervised approach achieves an outstanding 96% accuracy with fast convergence, outperforming traditional generic models, while the label-efficient SSL approach reaches a highly competitive 85% validation accuracy.
Chinese Translation
虾病在水产养殖行业中持续造成毁灭性的损失，迫切需要强大且自动化的检测方法。本研究首次将视觉变换器（Vision Transformers, ViT）和自监督学习（Self-Supervised Learning, SSL）应用于虾养殖领域，解决了性能瓶颈和数据标注挑战。我们提出了两个深度学习流程，以分类四种关键疾病：健康、黑鳃病（Black Gill, BG）、白斑病毒（White Spot Syndrome Virus, WSSV）以及两者的共感染，使用的数据集包含4,348张图像。首先，我们的监督迁移学习方法利用了在ImageNet上预训练的ViT-Small/16和EfficientNet骨干网络。其次，我们引入了一个对比学习框架（SimCLR），使用ViT-Small编码器从未标注图像中提取稳健的表示，然后进行微调。我们的结果为可持续水产养殖监测建立了强有力的新基准。监督方法以出色的96%准确率和快速收敛的表现，超越了传统通用模型，而标签高效的SSL方法则达到了高度竞争的85%验证准确率。
cs.CV / 78 / 2609.23404
ScaleBlind: Point Cloud Completion under Unknown Scale
ScaleBlind：在未知尺度下的点云补全
Abstract
Point cloud completion aims to infer a complete 3D shape from a partial point cloud and serves as a fundamental building block for downstream tasks such as reconstruction, editing, and simulation. Despite the recent progress, existing learning-based methods often implicitly rely on access to the ground-truth shape scale (GT-scale) during both training- and testing-time normalization, assuming privileged information that is unavailable in real-world inference. This hidden assumption limits practical deployment and can lead to severe completion artifacts, e.g., over- or under-completion and nested shells, once the oracle GT-scale cue is removed. We observe that the recent foundation image generation models exhibit a strong capability of understanding objects and geometries, and producing multi-view consistent renderings, making them promising priors for GT-scale-free 3D completion. Motivated by this insight, we propose ScaleBlind, a novel framework that leverages foundation-model-based image completion to recover global scale directly from partial inputs and then faithfully produces the 3D completion. Specifically, ScaleBlind dreams out complete multi-view appearances from rendered partial views, lifts the inferred missing regions back into 3D to obtain a geometry-aware coarse completion, and further refines it via a powerful cross-modal fusion network with the original partial point cloud. By harnessing 2D foundation priors, our method eliminates the need for accessing GT-scale information at inference. Moreover, it provides a principled bridge between 2D generative priors and 3D point cloud completion. Extensive experiments demonstrate the superiority of our framework, making ScaleBlind the new state-of-the-art for the point cloud completion task.
Chinese Translation
点云补全旨在从部分点云推断出完整的三维形状，是重建、编辑和仿真等下游任务的基础构件。尽管最近取得了一些进展，现有的基于学习的方法在训练和测试时的归一化过程中往往隐含依赖于真实形状尺度（GT-scale），假设有在现实世界推理中不可用的特权信息。这一隐含假设限制了实际应用的部署，并可能导致严重的补全伪影，例如过度补全或不足补全以及嵌套壳体，一旦去除神谕GT-scale提示。我们观察到，最近的基础图像生成模型展现出强大的理解物体和几何形状的能力，并能生成多视角一致的渲染，这使得它们成为无GT-scale的三维补全的有希望的先验。基于这一洞察，我们提出了ScaleBlind，一个新颖的框架，利用基于基础模型的图像补全直接从部分输入中恢复全局尺度，然后忠实地生成三维补全。具体而言，ScaleBlind从渲染的部分视图中生成完整的多视角外观，将推断出的缺失区域提升回三维，以获得几何感知的粗略补全，并通过强大的跨模态融合网络与原始部分点云进一步细化。通过利用二维基础先验，我们的方法消除了在推理时访问GT-scale信息的需求。此外，它为二维生成先验与三维点云补全之间提供了一个有原则的桥梁。大量实验表明我们框架的优越性，使ScaleBlind成为点云补全任务的新一代最先进技术。
cs.CV / 79 / 2609.23408
Accurate Motion Estimation with B\'ezier Control Point for Efficient Frame Interpolation
基于Bézier控制点的准确运动估计用于高效帧插值
Abstract
In frame interpolation tasks, motion ambiguity in the training set causes models to generate blurry intermediate frames. Moreover, the assumption of uniform motion between frames during inference further leads to inaccuracies in the generated intermediate frames. To tackle these challenges, we propose an Accurate motion estimation algorithm with B\'ezier Control point, ABC-Inter, for efficient frame Interpolation. Specifically, ABC-Inter designs an Accurate Flow estimation Module (AFM) by decoupling two-frame features and mapping to corresponding coordinates to better estimate the optical flow between the two frames. Furthermore, ABC-Inter eliminates motion ambiguity in the training set by introducing B\'ezier control points that are computed using the input frames and the intermediate ground-truth (gt) frames. This allows the model to estimate accurate optical flow between two frames during the training process, thereby solving the blurriness problem in the generated intermediate frames during inference. Benefiting from the more accurate flow estimation between two frames, we can introduce additional frames and directly use multiple flows to calculate B\'ezier control points for modeling non-uniform motion without retraining the model. Simultaneously, to realize the estimation of non-linear motion using only two frames, we also introduce a new B\'ezier control point estimation module which achieves better motion estimation between the two frames by performing fine-tuning on the model in the second stage. Experimental results demonstrate that our ABC-Inter achieves state-of-the-art performance on multiple benchmark datasets and exhibits excellent visual perception.
Chinese Translation
在帧插值任务中，训练集中的运动模糊导致模型生成模糊的中间帧。此外，在推理过程中假设帧之间的运动是均匀的，进一步导致生成的中间帧不准确。为了解决这些挑战，我们提出了一种基于Bézier控制点的准确运动估计算法ABC-Inter，以实现高效的帧插值。具体而言，ABC-Inter通过解耦两帧特征并映射到相应坐标来设计一个准确流估计模块（Accurate Flow estimation Module, AFM），以更好地估计两帧之间的光流。此外，ABC-Inter通过引入使用输入帧和中间真实帧（ground-truth, gt）计算的Bézier控制点，消除了训练集中的运动模糊。这使得模型在训练过程中能够准确估计两帧之间的光流，从而解决推理过程中生成的中间帧模糊问题。得益于两帧之间更准确的流估计，我们可以引入额外的帧，并直接使用多个流来计算Bézier控制点，以建模非均匀运动而无需重新训练模型。同时，为了仅使用两帧实现非线性运动的估计，我们还引入了一个新的Bézier控制点估计模块，通过在第二阶段对模型进行微调，来实现两帧之间更好的运动估计。实验结果表明，我们的ABC-Inter在多个基准数据集上达到了最先进的性能，并展现了出色的视觉感知能力。
cs.CV / 80 / 2609.23409
Retrieval Geometry Shapes Cache-Based Clip Adaptation
检索几何形状的基于缓存的 CLIP 适应
Abstract
Cache-based test-time adaptation improves CLIP predictions by storing and retrieving examples from the target stream while keeping the model frozen. However, existing methods largely treat the feature space used for image-image retrieval as fixed, leaving open how much adaptation depends on the retrieval space itself. We study this question by fixing the memory and changing only the retrieval encoder, finding that the same memory can yield very different gains: across sixteen retrieval spaces, ImageNet-A cache gain ranges from at most +0.44 points for CLIP and MAE to +19.7 +/- 0.4 for DINOv2-L, while label-free retrieval-space selection retains 98% of oracle gain on ImageNet-V2. These results show that memory quality depends not only on which examples are stored, but also on how they are retrieved. Motivated by this finding, we propose MARC (Memory Augmented Retrieval for CLIP), a training-free system that uses frozen CLIP for prediction and DINOv2-B for retrieval with a single fusion weight. A single-view cache repairs 1074 +/- 21 baseline errors, compared with 878 +/- 4 for a 64-view ensemble, at roughly one seventh of the cost. Across four ImageNet distribution shifts, MARC reaches a 67.91% OOD average and, at matched DINOv2-B scale and eight views, achieves 64.17 +/- 0.31% versus 62.75 +/- 0.15% for a graph-based cache system while running 2.6 times faster. Overall, our results establish retrieval space as a first-order design choice for robust cache-based adaptation in remote sensing, scientific imaging, and changing visual environments.
Chinese Translation
基于缓存的测试时适应通过存储和检索目标流中的示例来改善 CLIP 预测，同时保持模型不变。然而，现有方法在很大程度上将用于图像-图像检索的特征空间视为固定，尚未探讨适应性在多大程度上依赖于检索空间本身。我们通过固定内存并仅更改检索编码器来研究这个问题，发现相同的内存可以产生非常不同的收益：在十六个检索空间中，ImageNet-A 缓存收益在 CLIP 和 MAE 中最多为 +0.44 分，而在 DINOv2-L 中为 +19.7 +/- 0.4，而无标签检索空间选择在 ImageNet-V2 上保留了 98% 的 oracle 收益。这些结果表明，内存质量不仅依赖于存储哪些示例，还依赖于如何检索它们。基于这一发现，我们提出了 MARC（用于 CLIP 的记忆增强检索），这是一个无训练的系统，使用冻结的 CLIP 进行预测，并使用 DINOv2-B 进行检索，配以单一融合权重。与 64 视图集成的 878 +/- 4 基线错误相比，单视图缓存修复了 1074 +/- 21 个错误，成本大约是其七分之一。在四个 ImageNet 分布转变中，MARC 达到了 67.91% 的 OOD 平均值，并且在匹配 DINOv2-B 规模和八个视图的情况下，达到了 64.17 +/- 0.31% 的性能，而图基缓存系统的性能为 62.75 +/- 0.15%，运行速度快 2.6 倍。总体而言，我们的结果确立了检索空间作为远程感测、科学成像和变化视觉环境中基于缓存的鲁棒适应的首要设计选择。
cs.CV / 81 / 2609.23425
Semi-automated reconstruction of indoor geometry from 360-degree video for CFD-based airflow analysis in classrooms
基于360度视频的室内几何半自动重建用于教室中的CFD气流分析
Abstract
Computational Fluid Dynamics (CFD) is widely used to evaluate ventilation and contaminant transport in occupied buildings, but deployment at scale is limited by three bottlenecks: acquiring room geometry without costly scanning hardware or manual CAD modeling, decomposing the scene into individually manipulable objects, and reconfiguring those objects for alternative layouts without re-capturing the room. We present a semi-automated workflow that converts a single 360-degree video of a room into individually editable, simulation-ready geometry assets. A dense point cloud is reconstructed using Neural Radiance Fields (NeRF), and 2D instance masks from text-prompted SAM 3 segmentation are lifted to 3D using multi-view consensus and depth-band filtering. Points are separated into object instances with an octree, and occlusion gaps are healed with a connectivity graph. Chair templates are fitted by Iterative Closest Point (ICP) alignment, and table geometry is generated procedurally. A browser-based editor supports quality assurance and rapid construction of alternative layout configurations. A steady Reynolds-averaged OpenFOAM solution then drives transient passive-scalar transport; the setup is verified using a mesh-sensitivity study and validated against an IEA Annex 20 benchmark. We apply the workflow to two university classrooms and a tiered lecture-hall auditorium. The capture-to-geometry pass takes two to five hours per room on a consumer workstation. In a controlled obstruction sequence in one classroom, the modeled half-clearance time varies non-monotonically as furniture is added, and a cross-room comparison indicates that clearance behavior cannot be reliably extrapolated between rooms, motivating per-room geometry acquisition. By making that acquisition low-cost, the workflow makes geometry-resolved comparative ventilation studies practical for spaces such as classrooms.
Chinese Translation
计算流体动力学（CFD）广泛用于评估已占用建筑中的通风和污染物运输，但大规模应用受到三个瓶颈的限制：在没有昂贵扫描硬件或手动CAD建模的情况下获取房间几何形状，将场景分解为可单独操作的对象，以及在不重新捕获房间的情况下重新配置这些对象以适应不同布局。我们提出了一种半自动化工作流程，将房间的单个360度视频转换为可单独编辑、适合模拟的几何资产。使用神经辐射场（Neural Radiance Fields, NeRF）重建稠密点云，并通过文本提示的SAM 3分割生成的2D实例掩码，利用多视图共识和深度带过滤提升到3D。通过八叉树将点分离为对象实例，并使用连通性图修复遮挡间隙。通过迭代最近点（Iterative Closest Point, ICP）对齐拟合椅子模板，并以程序化方式生成桌子几何形状。基于浏览器的编辑器支持质量保证和快速构建替代布局配置。然后，稳定的雷诺平均OpenFOAM解驱动瞬态被动标量传输；该设置通过网格敏感性研究进行验证，并与IEA Annex 20基准进行比较。我们将该工作流程应用于两个大学教室和一个阶梯教室。捕获到几何的过程在消费级工作站上每个房间需要两到五小时。在一个教室的受控障碍序列中，建模的半清除时间随着家具的增加而呈现非单调变化，而跨房间比较表明，清除行为不能在房间之间可靠外推，这促使了每个房间几何形状的获取。通过降低获取成本，该工作流程使得针对教室等空间的几何解析比较通风研究变得切实可行。
cs.CV / 82 / 2609.23427
RSPDBench: Benchmarking Vision Foundation Models on Earth Observation Tasks Under Physically Grounded Remote-Sensing Product Degradations
RSPDBench：在物理基础的遥感产品退化下对地球观测任务的视觉基础模型进行基准测试
Abstract
Vision foundation models targeting Earth observation (EO) tasks are commonly evaluated on clean downstream benchmarks, but operational EO products can already contain spatial, radiometric, alignment, noise, and harmonization defects before reaching the model. Existing robustness evaluations often use generic image corruptions or broad domain shifts, which do not isolate these product-level failure modes. We introduce \textbf{RSPDBench}, a physically grounded \textbf{r}emote-\textbf{s}ensing-\textbf{p}roduct \textbf{d}egradation \textbf{b}enchmark for vision foundation models. RSPDBench evaluates five EO datasets, seven foundation-model entries, and two supervised baselines under audited primitive degradations and compound product chains. Each model is evaluated under its clean-selected native protocol, with robustness measured as the drop from its own clean baseline. Our analysis reveals that degradation sensitivity is strongly structured: resolution-conditioned and channel-grouped encoders protect different failure axes, and the same physical defect can hurt one model while helping another. Compound chains expose failures that isolated degradations do not predict, with model-dependent amplification, saturation, or component dominance, and excess drops up to $38$ percentage points beyond the strongest component. These results show that EO robustness cannot be characterized by clean accuracy or generic perturbation tests alone; it must also be measured against the structured defects that remote-sensing products carry into deployment.
Chinese Translation
针对地球观测（EO）任务的视觉基础模型通常在干净的下游基准上进行评估，但在到达模型之前，操作性EO产品可能已经包含空间、辐射、对齐、噪声和协调缺陷。现有的鲁棒性评估通常使用通用的图像损坏或广泛的领域转移，这并未孤立这些产品级的故障模式。我们引入了 extbf{RSPDBench}，这是一个基于物理的 extbf{遥感产品退化}基准，用于视觉基础模型。RSPDBench在经过审核的原始退化和复合产品链下评估五个EO数据集、七个基础模型条目和两个监督基线。每个模型在其干净选择的本地协议下进行评估，鲁棒性通过与其自身的干净基线的差距来衡量。我们的分析揭示了退化敏感性具有强烈的结构性：分辨率条件和通道分组的编码器保护不同的故障轴，同一物理缺陷可能对一个模型造成伤害，而对另一个模型则有帮助。复合链暴露出孤立退化无法预测的故障，伴随模型依赖的放大、饱和或组件主导，且超出最强组件的额外下降可达$38$个百分点。这些结果表明，EO的鲁棒性不能仅通过干净的准确性或通用的扰动测试来表征；它还必须针对遥感产品在部署中携带的结构性缺陷进行测量。
cs.CV / 83 / 2609.23431
HOIBlender: Blending Lightweight Detection with Vision-Language Priors for Efficient Human-Object Interaction Detection
HOIBlender：将轻量级检测与视觉-语言先验相结合以实现高效的人-物交互检测
Abstract
Human-object interaction (HOI) detection requires grounding an interacting human-object pair and recognizing the verb that links them, often under severe long-tail supervision. Recent methods improve accuracy with stronger detectors and vision-language priors, but many still stack heavy transformer encoders, intricate denoising schedules, or post-hoc semantic calibration on top of the detector. We present \textbf{HOIBlender}, an efficient HOI detector named after its core design principle: blending detector-grounded visual tokens, spatial subject-object reasoning, and BLIP-2 semantic priors inside one lightweight decoding pipeline. HOIBlender builds on an RF-DETR/LW-DETR-style foundation with a DINOv2 backbone and selects top-$K$ image-conditioned tokens directly from the multi-scale projector as subject and object candidates, removing the dedicated encoder stage retained by prior HOI methods. A dual-stage decoder first stabilizes human-object geometry and then performs verb and HOI classification through progressive BLIP-2 prior fusion, with classifier weights initialized from BLIP-2 text embeddings for long-tail categories. Grouped-query training further enriches optimization without increasing inference cost. Across three model scales (Nano, Small, 2XL), HOIBlender consistently outperforms SOV-STG-VLA and Hybrid-SOV-VLA on HICO-DET, reaching $44.49$ Default Full mAP in only $9$ training epochs while maintaining competitive latency and parameter budgets. These results show that lightweight detection, structured spatial-semantic decoding, and deeply integrated vision-language priors can be blended into a single efficient HOI pipeline.
Chinese Translation
人-物交互（HOI）检测需要确定一个交互的人-物对并识别连接它们的动词，通常在严重的长尾监督下进行。最近的方法通过更强大的检测器和视觉-语言先验提高了准确性，但许多方法仍在检测器之上堆叠重型变换器编码器、复杂的去噪调度或事后语义校准。我们提出了 extbf{HOIBlender}，一种高效的HOI检测器，其名称源于其核心设计原则：在一个轻量级解码管道中融合基于检测器的视觉标记、空间主体-客体推理和BLIP-2语义先验。HOIBlender基于RF-DETR/LW-DETR风格的基础，采用DINOv2主干，并直接从多尺度投影器中选择前$K$个图像条件标记作为主体和客体候选，去除了之前HOI方法保留的专用编码器阶段。双阶段解码器首先稳定人-物几何形状，然后通过逐步融合BLIP-2先验进行动词和HOI分类，分类器权重从BLIP-2文本嵌入初始化，以适应长尾类别。分组查询训练进一步丰富了优化，而不增加推理成本。在三个模型规模（Nano、Small、2XL）上，HOIBlender在HICO-DET上始终优于SOV-STG-VLA和Hybrid-SOV-VLA，仅用$9$个训练周期便达到了$44.49$的默认全局mAP，同时保持了竞争性的延迟和参数预算。这些结果表明，轻量级检测、结构化的空间-语义解码和深度集成的视觉-语言先验可以融合成一个高效的HOI管道。
cs.CV / 84 / 2609.23436
GAPS: Generative Active Pseudo-view Selection for Sparse-View 3D Gaussian Splatting
GAPS：稀疏视图3D高斯点云生成主动伪视图选择
Abstract
Novel view synthesis from sparse observations is severely under-constrained. Although 3D Gaussian Splatting (3DGS) enables real-time rendering, it produces floaters, broken geometry, and washed-out backgrounds when trained with few views. We propose an alternating optimization framework that uses a pre-trained image diffusion model to generate geometrically consistent pseudo-views for additional 3DGS supervision. Generation is constrained by depth-conditioned ControlNet, IP-Adapter style transfer, LoRA scene adaptation, and img2img structural anchoring. We introduce Generative Active Pseudo-view Selection (GAPS) to balance reconstruction informativeness and generative reliability when choosing target views. Its annealing schedule shifts from conservative interpolation early in training to exploratory extrapolation later, gradually covering unobserved regions. A dual-criterion admission gate and uncertainty-weighted losses reject unreliable generations, while density-adaptive DropGaussian reduces overfitting in complex scenes. On LLFF with 3/6/9 views, our method improves average PSNR over vanilla 3DGS by 0.40/0.89/0.70 dB. On Mip-NeRF 360 with 12/24 views, the gains are 1.18/0.80 dB. SSIM improves and LPIPS decreases in every setting. Ablations show that active selection and density-adaptive regularization are both necessary; only the full method reduces LPIPS below the no-pseudo-view baseline on unbounded 360-degree scenes.
Chinese Translation
从稀疏观测中合成新视图受到严重限制。尽管3D高斯点云渲染（3DGS）能够实现实时渲染，但在仅用少量视图进行训练时，它会产生漂浮物、破损几何体和洗淡的背景。我们提出了一种交替优化框架，利用预训练的图像扩散模型生成几何一致的伪视图，以提供额外的3DGS监督。生成过程受到深度条件的ControlNet、IP-Adapter风格转移、LoRA场景适应和img2img结构锚定的约束。我们引入了生成主动伪视图选择（GAPS），以在选择目标视图时平衡重建信息量和生成可靠性。其退火调度从训练初期的保守插值转变为后期的探索性外推，逐渐覆盖未观察到的区域。双标准准入门和不确定性加权损失拒绝不可靠的生成，而密度自适应的DropGaussian减少了复杂场景中的过拟合。在LLFF数据集上使用3/6/9个视图时，我们的方法使平均PSNR比普通3DGS提高了0.40/0.89/0.70 dB。在Mip-NeRF 360数据集上使用12/24个视图时，增益为1.18/0.80 dB。SSIM在每种设置中都有所改善，而LPIPS则有所降低。消融实验表明，主动选择和密度自适应正则化都是必要的；只有完整方法在无限制的360度场景中将LPIPS降低到无伪视图基线以下。
cs.CV / 85 / 2609.23442
PhysReflect: Geometry and Perception Guided Diffusion for Physically-Plausible Mirror Reflections
PhysReflect：几何与感知引导的物理可信镜面反射扩散
Abstract
Diffusion models generate high-quality images, yet often violate the physical laws governing mirror reflections. Reflections often suffer from geometric aberrations, including positional offsets, directional misalignment, proportional imbalance, and structural distortion. These failures remain evident even in contemporary state-of-the-art generative systems. Existing methods itigate this problem through synthetic data scaling or auxiliary depth conditioning, yet their merely reliance on latent-space noise reconstruction losses as implicit supervision prevents direct enforcement of reflection-specific geometric and perceptual constraints. To bridge this gap, we present PhysReflect, a geometry and perception guided diffusion framework that decodes the predicted clean latent into pixel space at each training step and applies annealed supervision through two complementary differentiable objectives. The Geometric Loss enforces mirror-induced spatial consistency through sparse epipolar correspondence and dense boundary projection alignment, where a SAM2-based TwinTrack mechanism provides stable in-mirror localization for boundary-aware supervision. The Perceptual Loss preserves reflected appearance by combining Semantic Consistency Loss, which maintains reflected identity and appearance via DINOv2 features, and Lighting Consistency Loss, which regularizes depth, surface-normal, and illumination coherence under monocular geometry priors. Experiments on synthetic and real-world benchmarks show that PhysReflect outperforms prior mirror-reflection methods in geometric, perceptual, and physical-plausibility metrics, as well as qualitative visual results.
Chinese Translation
扩散模型生成高质量图像，但往往违反镜面反射的物理法则。反射常常受到几何畸变的影响，包括位置偏移、方向不对齐、比例失衡和结构扭曲。这些缺陷在当代最先进的生成系统中仍然显而易见。现有方法通过合成数据扩展或辅助深度条件来缓解这一问题，但仅依赖潜在空间噪声重建损失作为隐式监督，无法直接施加反射特定的几何和感知约束。为了解决这一问题，我们提出了PhysReflect，一个几何与感知引导的扩散框架，在每个训练步骤中将预测的干净潜在空间解码为像素空间，并通过两个互补的可微目标应用退火监督。几何损失通过稀疏的极线对应和密集的边界投影对齐来强制镜面引起的空间一致性，其中基于SAM2的TwinTrack机制为边界感知监督提供稳定的镜内定位。感知损失通过结合语义一致性损失（Semantic Consistency Loss），该损失通过DINOv2特征保持反射的身份和外观，以及光照一致性损失（Lighting Consistency Loss），该损失在单目几何先验下正则化深度、表面法线和光照一致性，从而保持反射外观。对合成和真实世界基准的实验表明，PhysReflect在几何、感知和物理可信性指标以及定性视觉结果上均优于先前的镜面反射方法。
cs.CV / 86 / 2609.23450
Rule-Constrained Assignment for Cue-Ball Identification in Broadcast Snooker
基于规则约束的台球识别分配方法在广播斯诺克中的应用
Abstract
Accurate cue-ball identification is essential for metric analysis of broadcast snooker. Existing systems evaluate each candidate independently against a fixed white prototype and reject candidates above an appearance threshold. Under broadcast conditions, illumination changes can make colored balls appear white, while intrusions from players and equipment can obscure the cue ball or introduce competing candidates. We formulate cue-ball identification as a rule-constrained assignment problem that jointly assigns detected candidates to the bounded snooker inventory: one cue ball, up to 15 reds, and six colors with known spots. The cue ball is selected by the incremental cost of assigning each candidate to the white slot, and the conventional appearance test follows as the one-slot case. On 419 hand-annotated shots, our method improves identity accuracy from 88.1% to 95.5%, and from 80.5% to 95.2% on held-out venues. Within CueLift, our metric state-recovery system, the assignment expands coverage from 36.2% to 55.4% over 6,241 scorable shots. Assigning an estimate to every shot in a separate evaluation on 2,529 shots preserves this advantage.
Chinese Translation
准确的台球识别对于广播斯诺克的度量分析至关重要。现有系统独立评估每个候选者与固定的白色原型，并拒绝超过外观阈值的候选者。在广播条件下，光照变化可能导致彩球看起来呈白色，而球员和设备的干扰可能遮挡台球或引入竞争候选者。我们将台球识别形式化为一个规则约束的分配问题，该问题联合将检测到的候选者分配到有限的斯诺克库存中：一个台球、最多15个红球和六个带有已知斑点的彩球。通过将每个候选者分配到白色槽的增量成本来选择台球，常规的外观测试则作为单槽情况进行。在419个手动标注的镜头中，我们的方法将身份准确率从88.1%提高到95.5%，在保留场馆上从80.5%提高到95.2%。在我们的度量状态恢复系统CueLift中，该分配将覆盖率从36.2%扩展到55.4%，在6,241个可得分镜头中。在对2,529个镜头的单独评估中，为每个镜头分配一个估计值保持了这一优势。
cs.CV / 87 / 2609.23478
Algebraic Consistency Alone Does Not Certify Temporal Structure in Latent Action Models
单靠代数一致性无法证明潜在动作模型中的时间结构
Abstract
Latent action models infer a code for the transition between two frames of action-free video. Recent methods regularise this code to compose additively and reverse antisymmetrically, and report order-of-magnitude reductions in the resulting errors as a label-free certificate that the code has captured temporal structure. We show that this conclusion does not follow. Reconstruction drives the decoded transition toward a difference of state features, for which both identities hold for any pairing, a solution the metric cannot distinguish from one encoding nuisance state or a coordinate convention. Across five source domains, a trained but unconstrained counterpart already achieves 83-97% of the reduction relative to an untrained anchor. The residual fold is governed as much by the decoder family as by what is learned. A constrained model retrained after its temporal pairing is destroyed still reaches, in each domain, a lower error than the unconstrained model on real data. Downstream, preserving the temporal pairing yields no consistent advantage on LIBERO-GOAL or LIBERO-SPATIAL, and across the tested arms the code's mean linear action decodability falls as the algebraic error improves. We also test the most direct repair, a violation-contrastive objective that requires the algebra to fail on destroyed pairings: in the tested configurations it yields only a marginal separation within the reconstruction budget, on training and test triples alike. We recommend a validation protocol that these methods currently lack: a baseline-corrected metric, retraining on destroyed pairings, and a seed-budget analysis.
Chinese Translation
潜在动作模型推断出在两个无动作视频帧之间的过渡编码。近期的方法对该编码进行正则化，使其能够加法组合并反向反对称，并报告了结果误差的数量级减少，作为一种无标签的证据，表明该编码捕捉到了时间结构。我们证明这一结论并不成立。重建驱动解码过渡朝向状态特征的差异，对于任何配对，这两个身份都成立，这一解决方案无法通过度量区分出编码干扰状态或坐标约定。在五个源领域中，一个经过训练但不受约束的对应模型已经实现了相对于未训练锚点83-97%的误差减少。残余折叠的大小受解码器家族和学习内容的共同影响。在其时间配对被破坏后重新训练的约束模型，在每个领域中仍然达到比不受约束模型在真实数据上更低的误差。在下游任务中，保持时间配对在LIBERO-GOAL或LIBERO-SPATIAL上并未带来一致的优势，并且在测试的各个任务中，随着代数误差的改善，编码的平均线性动作解码能力下降。我们还测试了最直接的修复方法，即违反对比目标，要求代数在破坏的配对上失败：在测试的配置中，它在重建预算内仅产生了边际的分离，无论是在训练还是测试三元组上。我们建议一种这些方法目前缺乏的验证协议：基线校正度量、在破坏配对上重新训练和种子预算分析。
cs.CV / 88 / 2609.23492
CE$^4$L: Continual Ego, Exo, and Ego-Exo Learning
CE$^4$L：持续自我、外部和自我-外部学习
Abstract
Perception for embodied agents is video-based, often multi-view (ego, exo, or both), and inherently continual, with simultaneous task and viewpoint shifts. Yet continual learning (CL) remains dominated by exo-only recognition tasks, obscuring behavior under these real-world coupled shifts. We introduce Continual Ego, E}xo, and Ego-Exo Learning (CE$^4$L), a unified multi-view CL benchmark spanning four representative tasks: cross-view referenced skill assessment, temporal action segmentation, cross-view association, and action anticipation & planning. CE$^4$L highlights challenges largely absent in prior CL benchmarks, including cross-view correspondence, view-dependent asynchrony, and heterogeneous semantic objectives. To this end, we propose Video Incremental Subspace-routed Task Adapters (VISTA), a parameter-efficient baseline method that stores task-specific updates in lightweight adapters and performs training-free routing via residual distance to task-specific whitened subspaces estimated from second-order statistics. Extensive experiments demonstrate the significantly varied efficacy of representative CL methods across CE$^4$L settings, while VISTA is consistently competitive and achieves state-of-the-art overall performance. Our source code for benchmarks and methods is available at https://github.com/AnAppleCore/CE4L .
Chinese Translation
具身智能体的感知是基于视频的，通常是多视角的（自我、外部或两者兼有），并且本质上是持续的，伴随着任务和视角的同时变化。然而，持续学习（CL）仍然主要集中于仅外部的识别任务，这掩盖了在这些现实世界耦合变化下的行为。我们提出了持续自我、外部和自我-外部学习（CE$^4$L），这是一个统一的多视角持续学习基准，涵盖四个代表性任务：跨视角参考技能评估、时间动作分割、跨视角关联以及动作预测与规划。CE$^4$L突出了以往持续学习基准中几乎不存在的挑战，包括跨视角对应、视角依赖的异步性和异质语义目标。为此，我们提出了视频增量子空间路由任务适配器（VISTA），这是一种参数高效的基线方法，它在轻量级适配器中存储任务特定的更新，并通过残差距离进行无训练路由，以估计来自二阶统计的任务特定白化子空间。大量实验表明，在CE$^4$L设置下，代表性持续学习方法的有效性差异显著，而VISTA始终具有竞争力，并实现了最先进的整体性能。我们的基准和方法的源代码可在 https://github.com/AnAppleCore/CE4L 获取。
cs.CV / 89 / 2609.23495
Pay More Attention To Text In High-Resolution MLLMs
在高分辨率多模态大语言模型中更加关注文本
Abstract
Failures of high-resolution MLLMs are commonly attributed to a visual problem, motivating zooming, cropping, and related visual interventions to recover fine-grained evidence or suppress interference. Yet recent studies suggest that relevant visual evidence is already encoded in intermediate representations, indicating that visual-side improvements alone insufficient. This raises a natural question: does the remaining bottleneck lie in the text that guides visual search? We identify a previously overlooked linguistic bottleneck: questions formulated for answering do not necessarily specify the visual evidence required for localization. To address this mismatch, we introduce EviSpec, a training-free compiler that derives complementary evidence specifications while preserving the original question for final reasoning. We further validate it through matched-control experiments that isolate the roles of evidence specification and localization. With the search budget fixed, structured evidence specifications yield an 8.6% relative gain over generic requests. With evidence geometry matched, the evidence localized by EviSpec yields a 14.8% relative gain over random evidence. Together, these controls isolate the benefit of specifying what evidence to seek rather than merely expanding visual access. Across all five MLLMs, EviSpec consistently improves upon the corresponding baseline on each of the three benchmarks, yielding average relative gains of \textbf{10.4%, 8.8%, and 12.4%} on V\textsuperscript{*}Bench, HR-Bench-4K, and HR-Bench-8K, respectively. Beyond high-resolution reasoning, EviSpec also achieves state-of-the-art performance on VQA and hallucination-focused benchmarks.
Chinese Translation
高分辨率多模态大语言模型的失败通常被归因于视觉问题，这促使了缩放、裁剪和相关的视觉干预措施，以恢复细粒度证据或抑制干扰。然而，最近的研究表明，相关的视觉证据已经在中间表示中编码，这表明仅仅依靠视觉方面的改进是不够的。这引出了一个自然的问题：剩余的瓶颈是否在于指导视觉搜索的文本？我们识别出一个之前被忽视的语言瓶颈：为回答而制定的问题并不一定明确指定定位所需的视觉证据。为了解决这一不匹配，我们引入了EviSpec，这是一种无训练的编译器，能够在保留原始问题以进行最终推理的同时，推导出补充的证据规范。我们通过匹配控制实验进一步验证了这一点，这些实验隔离了证据规范和定位的作用。在固定搜索预算的情况下，结构化的证据规范相较于通用请求带来了8.6%的相对提升。在证据几何形状匹配的情况下，EviSpec定位的证据相较于随机证据带来了14.8%的相对提升。这些控制实验共同隔离了指定所需证据的好处，而不仅仅是扩大视觉访问。在所有五个多模态大语言模型中，EviSpec在每个基准测试上均持续优于相应的基线，在V extsuperscript{*}Bench、HR-Bench-4K和HR-Bench-8K上分别实现了平均相对提升 extbf{10.4%、8.8%和12.4%}。除了高分辨率推理外，EviSpec在视觉问答（VQA）和聚焦幻觉的基准测试上也达到了最先进的性能。
cs.CV / 90 / 2609.23507
Detecting Phone-Induced Pedestrian Distraction via a Multimodal Fusion Transformer
通过多模态融合变换器检测手机引发的行人分心
Abstract
The increasing reliance on mobile phones has made phone-induced pedestrian distraction increasingly prevalent. Activities such as texting, watching videos, and making phone calls have become significant contributors to traffic accidents. Reliable detection of pedestrian distraction is essential for autonomous vehicles, as it improves situational awareness and enables timely risk assessment, thereby supporting safe motion planning and vehicle control. We propose a multimodal fusion Transformer (MFT) for detecting phone-induced pedestrian distraction. MFT jointly extracts skeletal dynamics from body pose keypoints and visual appearance features from pedestrian images, effectively leveraging the complementary information provided by the two modalities. A cross-modal attention module is proposed to capture inter-modal dependencies through multi-head cross-attention, facilitating effective fusion of complementary information across the two modalities. Then, a temporal attention fusion module, implemented with a Transformer encoder, is employed to capture temporal dependencies. MFT is trained and evaluated on a manually annotated dataset comprising 287 pedestrian instances with 20,741 images. Extensive experiments demonstrate that MFT attains an overall accuracy of 95%, exceeding the performance of six baseline approaches by 6%.
Chinese Translation
对手机的日益依赖使得手机引发的行人分心现象愈加普遍。发送短信、观看视频和拨打电话等活动已成为交通事故的重要因素。可靠的行人分心检测对自动驾驶车辆至关重要，因为它提高了情境意识并能够及时评估风险，从而支持安全的运动规划和车辆控制。我们提出了一种多模态融合变换器（Multimodal Fusion Transformer, MFT）用于检测手机引发的行人分心。MFT联合提取来自身体姿态关键点的骨骼动态和来自行人图像的视觉外观特征，有效利用了两种模态提供的互补信息。我们提出了一种跨模态注意力模块，通过多头跨注意力捕捉模态间的依赖关系，促进了两种模态之间互补信息的有效融合。然后，采用实现了变换器编码器的时间注意力融合模块来捕捉时间依赖性。MFT在一个手动标注的数据集中进行了训练和评估，该数据集包含287个行人实例和20,741张图像。大量实验表明，MFT的整体准确率达到95%，比六种基线方法的表现高出6%。
cs.CV / 91 / 2609.23509
GARO: Geometry-Aware Redundancy Optimization for Real-Time and High-Fidelity Dynamic Gaussian Splatting
GARO：面向几何的冗余优化用于实时高保真动态高斯点云渲染
Abstract
Novel view synthesis is a key task for dynamic scene reconstruction, where high rendering speed is essential for applications such as virtual reality. Existing deformable Gaussian Splatting methods achieve high-fidelity dynamic scene modeling, but still face limitations in memory usage and rendering efficiency due to the large number of redundant Gaussians. To address these challenges, we propose Geometry-Aware Redundancy Optimization (GARO), a unified redundancy measurement framework in the adaptive density control stage of the traditional dynamic scene reconstruction pipeline. This framework first selects low-gradient candidates using an optimization activity assessment strategy, and then evaluates geometric complexity through low curvature analysis to further filter and prune redundant points, resulting in a compact and expressive Gaussian representation. Extensive experiments on synthetic and real-world datasets demonstrate that GARO achieves robust trade-offs between quality and speed, with PSNR remaining stable and rendering speed improved by 2x, validating the efficiency and effectiveness of GARO.
Chinese Translation
新视图合成是动态场景重建中的一项关键任务，其中高渲染速度对于虚拟现实等应用至关重要。现有的可变形高斯点云渲染方法实现了高保真的动态场景建模，但由于冗余高斯数量庞大，仍面临内存使用和渲染效率的限制。为了解决这些挑战，我们提出了面向几何的冗余优化（Geometry-Aware Redundancy Optimization，GARO），这是一个在传统动态场景重建流程的自适应密度控制阶段中的统一冗余测量框架。该框架首先使用优化活动评估策略选择低梯度候选点，然后通过低曲率分析评估几何复杂性，以进一步过滤和修剪冗余点，从而生成紧凑且富有表现力的高斯表示。在合成和真实世界数据集上的大量实验表明，GARO在质量和速度之间实现了稳健的平衡，PSNR保持稳定，渲染速度提高了2倍，验证了GARO的效率和有效性。
cs.CV / 92 / 2609.23533
GeoBalance: Geometry-Aware Monitoring and Reconstruction with Asymmetric Optimization for Balanced Multimodal Learning
GeoBalance：一种几何感知的监测与重建方法，采用不对称优化实现平衡的多模态学习
Abstract
Multimodal classifiers can converge to modality-dominant solutions in which one modality dominates the joint prediction, suppressing the learning of others. Existing balancing methods mainly adjust losses, gradients, or modality contributions, largely treating modality imbalance as an optimization problem while implicitly treating the weak modality as under-optimized but representationally intact. In this work, we find that this assumption does not always hold, as persistent modality dominance can induce a representation-level collapse of the weak modality, which we term \emph{manifold modality collapse} (MMC). MMC manifests as a coupled geometric degradation in which weak-modality representations collapse onto fewer directions within each class and become less separable across classes. Motivated by this observation, we propose \emph{GeoBalance}, a geometry-aware framework that monitors these two geometric properties and reconstructs the weak modality representation only when it exhibits signs of MMC. Once triggered, GeoBalance uses a fixed Simplex-ETF class scaffold and spectral regularization to restore class separation while preventing collapse onto a few feature directions. To preserve reconstruction during joint training, asymmetric gradient projection removes the joint-gradient component conflicting with reconstruction, leaving non-conflicting optimization unchanged. Extensive experiments across six multimodal benchmarks demonstrate great improvements over competitive balancing methods, validating its effectiveness.
Chinese Translation
多模态分类器可能会收敛到模态主导的解决方案，其中一种模态主导了联合预测，抑制了其他模态的学习。现有的平衡方法主要调整损失、梯度或模态贡献，基本上将模态不平衡视为一个优化问题，同时隐含地将弱模态视为未优化但在表征上保持完整。在本研究中，我们发现这一假设并不总是成立，因为持续的模态主导可能导致弱模态的表征级崩溃，我们称之为 extit{流形模态崩溃}（Manifold Modality Collapse, MMC）。MMC表现为一种耦合的几何退化，其中弱模态的表征在每个类别内崩溃到更少的方向，并且在类别之间变得不那么可分。基于这一观察，我们提出了 extit{GeoBalance}，一种几何感知框架，该框架监测这两个几何属性，并仅在弱模态表现出MMC迹象时重建其表征。一旦触发，GeoBalance使用固定的Simplex-ETF类别支架和谱正则化来恢复类别分离，同时防止崩溃到少数特征方向。为了在联合训练期间保持重建，不对称梯度投影移除与重建相冲突的联合梯度分量，保持非冲突优化不变。对六个多模态基准的广泛实验表明，与竞争性平衡方法相比，GeoBalance显著提高了性能，验证了其有效性。
cs.CV / 93 / 2609.23534
PosEviLoc: Position-Conditioned Spatial Evidence for Language-Based 3D Localization
PosEviLoc：基于位置条件的语言驱动三维定位空间证据
Abstract
Language-based 3D localization retrieves the point-cloud submap containing a target position from descriptions of nearby objects and their spatial relations. Existing methods typically compress queries and submaps into global descriptors, potentially obscuring object-level semantics and cross-description spatial coherence. We propose Position-Conditioned Evidence Localization (PosEviLoc), a query-position-aware framework for coarse text-to-point-cloud localization. Instead of relying on global matching, PosEviLoc evaluates each candidate submap using explicit semantic and spatial evidence. It models direction as a relation jointly determined by an object position and a hypothetical query position. The resulting Query-Position Spatial Evidence Field (QSEF) measures the fraction of query descriptions supported at each hypothetical position, explicitly capturing their agreement without using the ground-truth query pose to construct the evidence field. A Multi-Level Evidence Readout (MER) summarizes this evidence in a compact representation, which a lightweight MLP converts into a retrieval score. Across five benchmarks, PosEviLoc outperforms MNCL by an average of 17 percentage points in Recall@1. When used as a plug-and-play reranker, it improves MNCL by an average of 16 percentage points. Moreover, PosEviLoc introduces substantially fewer parameters and achieves faster inference speed than existing methods.
Chinese Translation
基于语言的三维定位通过提取包含目标位置的点云子地图，利用附近物体的描述及其空间关系。现有方法通常将查询和子地图压缩为全局描述符，这可能会模糊物体级语义和跨描述的空间一致性。我们提出了基于位置条件的证据定位（PosEviLoc），这是一个关注查询位置的粗略文本到点云定位框架。PosEviLoc不依赖于全局匹配，而是利用显式的语义和空间证据评估每个候选子地图。它将方向建模为由物体位置和假设查询位置共同决定的关系。生成的查询-位置空间证据场（Query-Position Spatial Evidence Field, QSEF）测量在每个假设位置上支持的查询描述的比例，明确捕捉它们的一致性，而不使用真实查询姿态来构建证据场。多级证据读取（Multi-Level Evidence Readout, MER）将这些证据总结为紧凑的表示，轻量级的多层感知机（MLP）将其转换为检索分数。在五个基准测试中，PosEviLoc在Recall@1上平均超越MNCL 17个百分点。当作为即插即用的重排序器使用时，PosEviLoc平均提高了MNCL 16个百分点。此外，PosEviLoc引入的参数显著少于现有方法，并且实现了更快的推理速度。
cs.CV / 94 / 2609.23541
Towards robust multimodal 3D object detection via visual foundation models
通过视觉基础模型实现鲁棒的多模态三维物体检测
Abstract
Multimodal 3D object detection is fundamental to robust perception in autonomous driving because it integrates complementary information from LiDAR and camera sensors. However, existing methods often fail to maintain robustness under out-of-distribution (OOD) corruptions caused by sensor noise, adverse weather, and environmental changes. To address this problem, we propose RoboDistill, a robust and generalizable multimodal 3D object detection framework that leverages visual foundation models (VFMs), such as the Segment Anything Model (SAM). First, we introduce SAM-AD, a domain-specific pretraining strategy that fine-tunes SAM on autonomous-driving imagery to extract feature representations with rich semantic information. Second, we design the AD Feature Pyramid Network (AD-FPN) to refine and upsample SAM features at multiple scales for seamless fusion with LiDAR features. Third, we develop the Depth-Guided Wavelet Attention (DGWA) module, which suppresses high-frequency sensor noise while preserving critical contextual information. Finally, we introduce KD Fusion, in which the pretrained SAM-AD serves as a teacher that distills high-quality visual knowledge into a lightweight point-cloud network, thereby improving robustness under noisy conditions. Extensive experiments across 27 challenging OOD corruption settings show that RoboDistill generally delivers stronger or competitive detection performance and robustness relative to representative state-of-the-art methods. This work bridges the gap between VFMs and 3D object detection and advances robust multimodal perception for real-world autonomous-driving applications.
Chinese Translation
多模态三维物体检测是自动驾驶中鲁棒感知的基础，因为它整合了来自激光雷达和摄像头传感器的互补信息。然而，现有方法在面对由传感器噪声、恶劣天气和环境变化引起的分布外（OOD）干扰时，往往无法保持鲁棒性。为了解决这一问题，我们提出了RoboDistill，一个鲁棒且可泛化的多模态三维物体检测框架，利用视觉基础模型（VFMs），如Segment Anything Model（SAM）。首先，我们引入了SAM-AD，一种领域特定的预训练策略，旨在对SAM进行微调，以提取具有丰富语义信息的特征表示。其次，我们设计了AD特征金字塔网络（AD-FPN），在多个尺度上精炼和上采样SAM特征，以便与激光雷达特征无缝融合。第三，我们开发了深度引导小波注意力（DGWA）模块，该模块在保留关键上下文信息的同时抑制高频传感器噪声。最后，我们引入了KD融合，其中预训练的SAM-AD作为教师，将高质量的视觉知识提炼到轻量级点云网络中，从而提高在噪声条件下的鲁棒性。在27个具有挑战性的OOD干扰设置下的广泛实验表明，RoboDistill通常提供比代表性的最先进方法更强或更具竞争力的检测性能和鲁棒性。本研究弥合了VFMs与三维物体检测之间的差距，推动了现实世界自动驾驶应用中的鲁棒多模态感知的发展。
cs.CV / 95 / 2609.23548
SewFusion: Tailored Generation of Topology and Panel-Level Geometry for Sewing Patterns
SewFusion：为缝纫图案量身定制的拓扑和面板级几何生成
Abstract
Generating sewing patterns from images and text requires modeling a heterogeneous representation composed of discrete topology and continuous geometry. Existing methods mainly follow two paradigms: diffusion-based methods enable holistic geometry generation by converting the entire pattern into a continuous representation, but weaken discrete topology modeling; in contrast, autoregressive methods preserve discrete topology through next-token prediction, but tie continuous geometry regression to token-level hidden states with limited panel-level context. To bridge this gap, we propose SewFusion, a unified autoregressive framework that adopts tailored generation mechanisms for discrete topology and panel-level continuous geometry, using next-token prediction for the former and flow matching for the latter. To support panel-level continuous geometry generation, we introduce a Panel Geometry VAE that learns a fixed-size latent space for variable-length panel geometry, together with Panel Geometry Flow for latent generation. We further propose Panel-Forcing to reduce the training--inference mismatch in topology context and improve robustness to topology prediction errors. Extensive experiments on SewFactory and GCD-MM demonstrate that SewFusion consistently outperforms previous state-of-the-art methods across various settings, achieving +6.36% Panel Accuracy, +11.30% Stitch Accuracy, and -1.90 Vertex L2 error in the image-text-based generation setting.
Chinese Translation
从图像和文本生成缝纫图案需要对由离散拓扑和连续几何组成的异构表示进行建模。现有方法主要遵循两种范式：基于扩散的方法通过将整个图案转换为连续表示来实现整体几何生成，但削弱了离散拓扑建模；相反，自回归方法通过下一个标记预测保留了离散拓扑，但将连续几何回归绑定到标记级隐藏状态，限制了面板级上下文。为了解决这一问题，我们提出了SewFusion，一个统一的自回归框架，采用针对离散拓扑和面板级连续几何的量身定制生成机制，前者使用下一个标记预测，后者使用流匹配。为了支持面板级连续几何生成，我们引入了一种面板几何变分自编码器（Panel Geometry VAE），它为可变长度的面板几何学习一个固定大小的潜在空间，并结合面板几何流（Panel Geometry Flow）进行潜在生成。我们进一步提出了面板强制（Panel-Forcing），以减少拓扑上下文中的训练-推理不匹配，并提高对拓扑预测错误的鲁棒性。在SewFactory和GCD-MM上的大量实验表明，SewFusion在各种设置中始终优于之前的最先进方法，在基于图像-文本的生成设置中实现了+6.36%的面板准确率、+11.30%的缝合准确率和-1.90的顶点L2误差。
cs.CV / 96 / 2609.23553
Modeling Clinical Workflow for SYNTAX Scoring from Coronary Angiography Videos
基于冠状动脉造影视频的SYNTAX评分临床工作流程建模
Abstract
The SYNTAX score is a clinically established tool for assessing anatomical lesion complexity in coronary artery disease and guiding subsequent treatment. However, automated SYNTAX scoring is commonly formulated as a direct regression problem from coronary angiography videos to patient-level scores. In this work, we reformulate SYNTAX scoring as a vessel segment identity-preserving anatomical reasoning problem and propose a hierarchical modeling framework that explicitly aligns learning with the clinical workflow. Our approach maintains vessel segment identity across frames and views, estimates stenosis severity at the segment level, and aggregates evidence hierarchically according to coronary anatomy. Simultaneously, to address the scarcity of domain-specific data, we integrate and complete multiple public coronary angiography datasets, constructing a large-scale resource featuring completed vessel segmentation and derived structural annotations. Experiments demonstrate that vessel segment-level stenosis embedding enhances explanatory power and reduces prediction variability compared to baseline models, with the R^2 score improving by 0.201 and dev STD decreasing by 18.4%. These results highlight the necessity of structure-aligned modeling for reliable and stable automated SYNTAX scoring from multi-view coronary angiography videos. The GitHub link is https://github.com/VersaceSu7/SYNTAX_score_777.
Chinese Translation
SYNTAX评分是评估冠状动脉疾病中解剖病变复杂性并指导后续治疗的临床工具。然而，自动化的SYNTAX评分通常被视为从冠状动脉造影视频到患者级评分的直接回归问题。在本研究中，我们将SYNTAX评分重新表述为一种保持血管段身份的解剖推理问题，并提出一种分层建模框架，该框架明确将学习与临床工作流程对齐。我们的方法在不同帧和视角中保持血管段身份，估计段级狭窄严重性，并根据冠状解剖结构分层聚合证据。同时，为了应对领域特定数据的稀缺性，我们整合并补充了多个公共冠状动脉造影数据集，构建了一个大型资源，包含完整的血管分割和衍生的结构注释。实验表明，与基线模型相比，血管段级狭窄嵌入增强了解释能力并减少了预测变异性，R^2分数提高了0.201，标准差减少了18.4%。这些结果强调了结构对齐建模在从多视角冠状动脉造影视频中进行可靠和稳定的自动化SYNTAX评分的必要性。GitHub链接为 https://github.com/VersaceSu7/SYNTAX_score_777。
cs.CV / 97 / 2609.23561
Transferring Visual Explanations: How Cross-Architecture Knowledge Distillation Affects Model Interpretability
转移视觉解释：跨架构知识蒸馏如何影响模型可解释性
Abstract
Deploying efficient neural networks is essential in resource-constrained environments, yet compact models often sacrifice interpretability - a critical in safety-critical domains such as autonomous driving and medicine. This study investigates whether Knowledge Distillation transfers the spatial feature attribution of a large teacher network to a compact student. To assess the influence of the KD scheme on interpretability, we distill a ResNet-152 teacher into a ResNet-34 student on ImageNet-1K across five configurations by systematically varying the distillation temperature and soft-label loss weight. Models are evaluated on top-1 accuracy, along with two interpretability metrics: Relevance Mass Accuracy and Relevance Rank Accuracy. These metrics are computed via Grad-CAM heatmaps benchmarked against ground-truth object masks. Our results show that top-1 accuracy ranges from 71.6% to 74.0%. For Grad-CAM, RMA ranges from 7.7% to 9.7% and RRA from 7.3% to 10.1%; for Guided Grad-CAM, RMA ranges from 16.1% to 18.6% and RRA from 15.9% to 21.5%. Interpretability proves far more sensitive to the soft-label weight than to the temperature: keeping the student anchored to hard labels preserves both accuracy and coarse localization, whereas weighting the teacher heavily degrades both. Fine-grained attribution, however, fell below the undistilled baseline in every configuration tested, indicating that logit distillation transmits where a model attends more readily than the pixel-level structure of that attention. We evaluate 12 cross-architecture combinations of convolutional and transformer-based models, revealing that the inheritance of fine-grained spatial reasoning is fundamentally bottlenecked by the student's intrinsic structural biases. To our knowledge, this is the first application of this interpretability-aware evaluation framework - previously used for neural network pruning - to KD.
Chinese Translation
在资源受限的环境中部署高效的神经网络至关重要，但紧凑模型往往牺牲可解释性，这在自动驾驶和医学等安全关键领域尤为重要。本研究探讨知识蒸馏（Knowledge Distillation）是否能够将大型教师网络的空间特征归因转移到紧凑的学生网络上。为了评估知识蒸馏方案对可解释性的影响，我们将ResNet-152教师模型蒸馏为ResNet-34学生模型，并在ImageNet-1K上通过系统性地改变蒸馏温度和软标签损失权重进行五种配置的实验。模型在top-1准确率以及两个可解释性指标上进行评估：相关性质量准确率（Relevance Mass Accuracy, RMA）和相关性排名准确率（Relevance Rank Accuracy, RRA）。这些指标通过Grad-CAM热图与真实对象掩码进行基准测试。我们的结果显示，top-1准确率范围为71.6%到74.0%。对于Grad-CAM，RMA范围为7.7%到9.7%，RRA范围为7.3%到10.1%；对于Guided Grad-CAM，RMA范围为16.1%到18.6%，RRA范围为15.9%到21.5%。可解释性对软标签权重的敏感性远高于温度：将学生模型固定在硬标签上能够同时保持准确性和粗略定位，而对教师模型的重权重则会显著降低这两者。然而，细粒度归因在每种测试配置中均低于未蒸馏的基线，表明logit蒸馏更容易传递模型关注的区域，而非该关注的像素级结构。我们评估了12种卷积和基于变换器的模型的跨架构组合，揭示了细粒度空间推理的继承在根本上受到学生模型内在结构偏差的瓶颈。我们所知，这是首次将这种关注可解释性的评估框架应用于知识蒸馏，此框架之前用于神经网络剪枝。
cs.CV / 98 / 2609.23565
MaskVLA: Visual Masking Against Trajectory Overfitting of Vision-Language-Action Model
MaskVLA：针对视觉-语言-动作模型轨迹过拟合的视觉遮蔽方法
Abstract
Vision-Language-Action (VLA) models integrate vision-language understanding with executable robot actions, enabling end-to-end learning for robot control. However, our empirical analysis reveals that existing models exhibit severe trajectory overfitting when finetuned on limited datasets. To guide the model in effectively utilizing wrist camera information, we propose MaskVLA, a masking-based fine-tuning strategy. By randomly masking a small portion of the main camera's visual information, the model is guided to autonomously learn more fine-grained, task-relevant, and effective visual features. This process leads to the emergence of robust policies, thereby enhancing the model's capability to tackle complex manipulation tasks and improving its generalization performance. Our method has been comprehensively evaluated on RoboTwin 2.0, achieving an average success rate improvement of 23.2% and 16.8% compared to $\pi_0$ and OpenVLA-OFT, respectively. Furthermore, experiments on real-world ALOHA robots also demonstrate the effectiveness of our approach.
Chinese Translation
视觉-语言-动作（VLA）模型将视觉-语言理解与可执行的机器人动作相结合，实现了机器人控制的端到端学习。然而，我们的实证分析显示，现有模型在有限数据集上进行微调时，表现出严重的轨迹过拟合。为了引导模型有效利用腕部相机信息，我们提出了MaskVLA，一种基于遮蔽的微调策略。通过随机遮蔽主相机视觉信息的一小部分，模型被引导自主学习更细致、与任务相关且有效的视觉特征。这个过程促使了稳健策略的出现，从而增强了模型应对复杂操作任务的能力，并提高了其泛化性能。我们的方法在RoboTwin 2.0上进行了全面评估，与$ ext{π}_0$和OpenVLA-OFT相比，平均成功率分别提高了23.2%和16.8%。此外，在真实世界的ALOHA机器人上的实验也验证了我们方法的有效性。
cs.CV / 99 / 2609.23566
G6D: Geometric Learning-Free RGB-D 6D Pose Solver for Robotic Manipulation
G6D：无学习的几何驱动RGB-D 6D姿态求解器用于机器人操作
Abstract
6D object pose estimation is fundamental to robotic manipulation and automation. Recent zero-shot methods have significantly improved generalization to unseen objects, but most still rely on large-scale pretrained models with substantial GPU computation and memory demands. These requirements complicate deployment on robotic platforms where perception, planning, and control share limited computational resources, while learned intermediate representations offer limited geometric interpretability for task-specific adaptation. To address these limitations, we propose G6D, a learning-free, geometry-driven RGB-D 6D pose solver. Given an RGB-D observation, an object instance mask, camera intrinsics, and a CAD model, G6D generates pose hypotheses through template-based geometric matching and refines them using silhouette and depth consistency, forming a purely geometry-driven pose estimation paradigm. This paradigm requires neither pretrained visual models nor target-specific training and preserves interpretable geometric representations throughout pose estimation. Moreover, adjustable hypothesis counts provide flexible accuracy-computation trade-offs, while a CPU-only configuration supports deployment without GPU resources. Experiments on LineMOD and five BOP19 datasets demonstrate advanced performance. Real-world pick-and-place experiments further demonstrate G6D's applicability to robotic manipulation. The complete project is publicly available at https://ai4control.github.io/G6D-Project-Page .
Chinese Translation
6D物体姿态估计是机器人操作和自动化的基础。最近的零样本方法显著提高了对未见物体的泛化能力，但大多数仍依赖于大规模预训练模型，这些模型对GPU计算和内存的需求很高。这些要求使得在计算资源有限的机器人平台上进行部署变得复杂，因为感知、规划和控制共享有限的计算资源，而学习的中间表示在任务特定适应方面提供的几何可解释性有限。为了解决这些局限性，我们提出了G6D，一种无学习的、以几何为驱动的RGB-D 6D姿态求解器。给定RGB-D观测、物体实例掩膜、相机内参和CAD模型，G6D通过基于模板的几何匹配生成姿态假设，并利用轮廓和深度一致性对其进行精炼，形成一种纯粹的几何驱动姿态估计范式。该范式不需要预训练的视觉模型或特定目标的训练，并在整个姿态估计过程中保持可解释的几何表示。此外，可调的假设数量提供了灵活的精度与计算之间的权衡，而仅使用CPU的配置支持在没有GPU资源的情况下进行部署。在LineMOD和五个BOP19数据集上的实验表明了其先进的性能。现实世界的拾取和放置实验进一步证明了G6D在机器人操作中的适用性。完整项目可在https://ai4control.github.io/G6D-Project-Page上公开获取。
cs.CV / 100 / 2609.23586
An Efficient and Effective Watermarking Scheme for the Protection of the Intellectual Property Rights of Video Generative Models
一种高效且有效的视频生成模型知识产权保护水印方案
Abstract
The rapid development of video generative models (VGMs) has enabled the generation of highly realistic synthetic videos, raising concerns about the intellectual property rights (IPR) of these models. In particular, two closely related forensic tasks remain largely unaddressed: synthetic video verification (determining whether a video was generated by a protected VGM) and model ownership verification (determining whether a suspect VGM is an unauthorized copy of a protected VGM). In this paper, we propose a new in-generation watermarking scheme that can address the two verification tasks. First, a novel video watermarking network named VidMark is presented, which incorporates a two-scale discrete wavelet transform (DWT) decomposition and a global temporal attention block (GTAB) to enhance watermark robustness and imperceptibility. Second, we present a decoder-guided fine-tuning procedure. By leveraging the frozen VidMark decoder, this process enables VGMs to synthesize videos carrying an imperceptible, robust, and model-specific watermark. Finally, two verification frameworks are established to perform synthetic video verification and model ownership verification. Extensive experiments on representative VGMs demonstrate that the proposed scheme achieves over 99% watermark extraction accuracy and 100% verification accuracy on both tasks, with negligible impact on video generation quality. Furthermore, the watermarks exhibit strong robustness against a comprehensive range of video-level and model-level attacks.
Chinese Translation
视频生成模型（VGM）的快速发展使得生成高度逼真的合成视频成为可能，这引发了对这些模型知识产权（IPR）的关注。特别是，有两个密切相关的取证任务仍然未得到充分解决：合成视频验证（确定视频是否由受保护的VGM生成）和模型所有权验证（确定可疑的VGM是否为受保护的VGM的未经授权副本）。在本文中，我们提出了一种新的生成过程水印方案，以解决这两个验证任务。首先，提出了一种新颖的视频水印网络VidMark，该网络结合了双尺度离散小波变换（DWT）分解和全局时间注意力块（GTAB），以增强水印的鲁棒性和不可感知性。其次，我们提出了一种解码器引导的微调程序。通过利用冻结的VidMark解码器，该过程使VGM能够合成携带不可感知、鲁棒且特定于模型的水印的视频。最后，建立了两个验证框架，以执行合成视频验证和模型所有权验证。在代表性VGM上的大量实验表明，所提出的方案在两个任务上实现了超过99%的水印提取准确率和100%的验证准确率，对视频生成质量的影响微乎其微。此外，水印对各种视频级和模型级攻击表现出强大的鲁棒性。
cs.CV / 101 / 2609.23592
Collapse, Not Complexity: Failure-Conditioned Decomposition Repair for End-to-End Document Parsing
崩溃，而非复杂性：基于失败条件的分解修复用于端到端文档解析
Abstract
End-to-end document parsers increasingly offer an optional reasoning mode for complex pages. On a 180-page entropy-stratified discovery sample with one frozen 4B checkpoint, complexity is the wrong decision variable. Reasoning lowers mean quality by 2.21 Overall at 1.54x tokens; a preregistered input-only model cannot predict its signed benefit (held-out AUROC 0.47, indistinguishable from chance). The benefit concentrates on pages whose ordinary pass has already collapsed, and they do not look complex: shared collapses have lower layout entropy than healthy ones yet consume 19x the tokens as degenerate repetition that doubling the budget does not cure. Switching modes rarely repairs them: 83% recur under reasoning. We instead detect collapse from the ordinary-pass trace, decompose the page by projection, and re-parse each region. Repair gains 1.40 Overall (95% CI [0.68, 2.16]) at 1.13x tokens, replicates across three checkpoints, and, with all parameters frozen, gains 2.41 (CI [1.64, 3.46]) on the remaining 1,175 benchmark pages.
Chinese Translation
端到端文档解析器越来越多地为复杂页面提供可选的推理模式。在一个包含180页熵分层发现样本的实验中，使用一个冻结的4B检查点，复杂性并不是正确的决策变量。推理使平均质量降低了2.21，整体在1.54倍的token下；一个预注册的仅输入模型无法预测其签署的收益（保留的AUROC为0.47，与随机猜测无异）。收益集中在那些普通通行已经崩溃的页面上，而这些页面看起来并不复杂：共享崩溃的布局熵低于健康页面，但消耗的token是退化重复的19倍，增加预算并不能解决这个问题。切换模式很少能修复这些页面：在推理下，83%的页面会再次出现崩溃。我们反而通过普通通行的轨迹检测崩溃，通过投影分解页面，并重新解析每个区域。修复在1.13倍的token下获得了1.40的整体增益（95%置信区间[0.68, 2.16]），在三个检查点上重复验证，并且在所有参数冻结的情况下，在剩余的1,175个基准页面上获得了2.41的增益（置信区间[1.64, 3.46]）。
cs.CV / 102 / 2609.23596
StyleAT: Defending Face Recognition Against Semantic Attacks
StyleAT：抵御语义攻击的面部识别防御机制
Abstract
With face-recognition models now embedded in everyday authentication and surveillance, recent works have pinpointed a critical weakness: these models remain acutely vulnerable to adversarial semantic edits. I.e., adversarially produced semantic alterations to the input, such as slight aging or pose changes, can induce misclassifications. Certain existing attacks are powerful, but they can be computationally costly, rendering them inadequate for developing defenses (e.g., through adversarial training). To fill the gap, we introduce BoundStyle, a potent semantic attack operating in StyleGAN's rich latent space to maximize misclassification rates. Notably, BoundStyle achieves high attack success rates while being ${\sim}{\times}9.5$ faster than existing state-of-the-art attacks, making it suitable for adversarial training. Building on BoundStyle, we develop StyleAT, an efficient adversarial training scheme that incorporates low-budget attack variants yet defends against stronger and unseen semantic attacks. We evaluate on two datasets unseen during training and seven models, and find that StyleAT boosts robust accuracy against state-of-the-art attacks and outperforms common defenses in various settings.
Chinese Translation
随着面部识别模型在日常身份验证和监控中的广泛应用，近期研究指出了一个关键弱点：这些模型对对抗性语义编辑极为脆弱。即，对输入进行的对抗性语义修改，如轻微的衰老或姿态变化，可能导致错误分类。某些现有攻击虽然强大，但计算成本高，使其在开发防御（例如，通过对抗训练）时显得不足。为填补这一空白，我们引入了 BoundStyle，这是一种在 StyleGAN 丰富的潜在空间中操作的强大语义攻击，旨在最大化错误分类率。值得注意的是，BoundStyle 在攻击成功率上表现出色，同时比现有的最先进攻击快约 ${ imes}9.5$ 倍，适合用于对抗训练。在此基础上，我们开发了 StyleAT，一种高效的对抗训练方案，结合了低成本攻击变体，同时能够抵御更强大且未见过的语义攻击。我们在两个训练时未见过的数据集和七个模型上进行了评估，发现 StyleAT 在抵御最先进攻击方面提高了鲁棒性准确率，并在各种设置中优于常见防御方法。
cs.CV / 103 / 2609.23600
PETR: Prompt Ensembling with Training-free Routing for Vision-Language Models
PETR：基于无训练路由的视觉-语言模型提示集成
Abstract
Prompt learning efficiently adapts vision-language models (VLMs) to downstream tasks, but gains on seen classes often come at the expense of generalization to unseen classes. To address this limitation, we propose prompt ensembling with training-free routing (PETR), whose key innovation is a carefully designed dual-prompt architecture: two complementary prompts are learned from different data and objectives to emphasize seen class discrimination and unseen-class generalization, respectively. During training, both prompts are fine-tuned using a shared frozen CLIP backbone, and statistical information is collected from the training set logits. At inference time, we determine the similarity of each test sample to seen data, and route the sample to the most appropriate prompt branch. To the best of our knowledge, this is the first prompt tuning framework that performs training-free adaptive routing based on statistical similarity. This design provides an interpretable routing signal and avoids common MoE-style routing pathologies, such as router training instability and load imbalance. Extensive experiments on 11 benchmark datasets demonstrate that our framework consistently outperforms previous methods on both seen and unseen classes, achieving new state-of-the-art results.
Chinese Translation
提示学习有效地将视觉-语言模型（VLMs）适应于下游任务，但在已见类别上的收益往往以牺牲对未见类别的泛化能力为代价。为了解决这一局限性，我们提出了基于无训练路由的提示集成方法（PETR），其关键创新在于精心设计的双提示架构：两个互补的提示分别从不同的数据和目标中学习，以强调已见类别的区分和未见类别的泛化。在训练过程中，两个提示都使用共享的冻结 CLIP（Contrastive Language-Image Pretraining）主干进行微调，并从训练集的 logits 中收集统计信息。在推理时，我们确定每个测试样本与已见数据的相似性，并将样本路由到最合适的提示分支。据我们所知，这是第一个基于统计相似性进行无训练自适应路由的提示调优框架。该设计提供了可解释的路由信号，并避免了常见的 MoE（Mixture of Experts）风格路由病态，如路由器训练不稳定和负载不平衡。在 11 个基准数据集上的大量实验表明，我们的框架在已见和未见类别上始终优于以前的方法，达到了新的最先进的结果。
cs.CV / 104 / 2609.23601
PREM: Prefix-Steered Recurrent Memory for Long-Video Understanding
PREM：前缀引导的递归记忆用于长视频理解
Abstract
Long-video understanding must capture transient visual evidence under strict token budgets, yet existing methods compress frames, append memory tokens, or alter internal key-value (KV) caches. We introduce Prefix-Steered Recurrent Memory (PREM), a memory-token-free framework for frozen vision-language models (VLMs). PREM separates video ingestion from query answering: a recurrent writer distills visual streams into a compact 256 KiB multi-slot associative state, while a question-conditioned readout adds memory-derived key/value (K/V) steering modulations to existing non-visual prompt prefixes during prefill. This enables write-once, query-many inference without extra prompt tokens or decoding recurrence. Across six long-video benchmarks in offline and streaming end-of-stream settings, PREM consistently outperforms frozen baselines at every evaluated visual budget. Under a constrained budget of 16 frames, PREM improves macro-average accuracy by 3.06% on Qwen2.5-VL-3B, with gains of 11.0% on action antonym identification and 9.9% on localized needle retrieval. These gains require tuning 0.24% of backbone parameters at 0.03 GiB of peak GPU memory overhead.
Chinese Translation
长视频理解必须在严格的令牌预算下捕捉瞬时视觉证据，然而现有方法通过压缩帧、附加记忆令牌或改变内部键值（KV）缓存来实现。我们提出了前缀引导的递归记忆（Prefix-Steered Recurrent Memory, PREM），这是一个针对冻结的视觉语言模型（VLMs）的无记忆令牌框架。PREM 将视频摄取与查询回答分开：一个递归写入器将视觉流提炼为一个紧凑的 256 KiB 多槽关联状态，而一个基于问题的读取过程在预填充期间向现有的非视觉提示前缀添加记忆派生的键/值（K/V）引导调制。这使得在没有额外提示令牌或解码递归的情况下实现一次写入、多次查询的推理。在离线和流媒体结束流设置的六个长视频基准测试中，PREM 在每个评估的视觉预算下始终优于冻结基线。在 16 帧的受限预算下，PREM 在 Qwen2.5-VL-3B 上提高了 3.06% 的宏平均准确率，在动作反义词识别上提高了 11.0%，在局部针头检索上提高了 9.9%。这些提升只需调整 0.24% 的主干参数，且峰值 GPU 内存开销为 0.03 GiB。
cs.CV / 105 / 2609.23606
Beyond UV Mapping: Mesh Texture Compression via Surface-Aligned Texture Fields
超越 UV 映射：通过表面对齐纹理场进行网格纹理压缩
Abstract
Mesh texture compression typically relies on 2D UV atlases, whose chart discontinuities and mapping overhead can limit coding efficiency. To tackle this challenge, we introduce TexF, a surface-aligned texture field that organizes texture attributes in sparse voxels derived from the mesh surface. This representation supports high-resolution textures while preserving local 3D correlations for compression and enabling direct surface queries. For bitstream compression, TexF reuses established 3D attribute codecs, with voxel locations reconstructed from the decoded mesh without separate transmission. For GPU-resident compression, we develop 3DNTC, which combines quantized hash features with a lightweight decoder for random-access reconstruction at surface positions. Differentiable rendering enables image-space refinement of both voxel attributes and compressed neural fields. Experiments on the MPEG and AOM mesh compression benchmarks demonstrate improved average rate-distortion performance over representative UV-based methods for both bitstream and GPU-resident compression. 3DNTC also supports real-time rendering.
Chinese Translation
网格纹理压缩通常依赖于 2D UV 图集，其图表不连续性和映射开销可能限制编码效率。为了解决这一挑战，我们引入了 TexF，一种表面对齐的纹理场，该纹理场在从网格表面派生的稀疏体素中组织纹理属性。这种表示支持高分辨率纹理，同时保留局部 3D 相关性以进行压缩，并支持直接的表面查询。对于比特流压缩，TexF 重用已建立的 3D 属性编解码器，体素位置从解码的网格中重建，无需单独传输。对于 GPU 驻留压缩，我们开发了 3DNTC，它结合了量化哈希特征和轻量级解码器，以实现表面位置的随机访问重建。可微渲染使得体素属性和压缩神经场的图像空间细化成为可能。在 MPEG 和 AOM 网格压缩基准测试中的实验表明，对于比特流和 GPU 驻留压缩，3DNTC 在平均率失真性能上优于代表性的基于 UV 的方法。3DNTC 还支持实时渲染。
cs.CV / 106 / 2609.23619
Compact Low-Cost Hyperspectral Imaging via Angular-to-Spectral Diversity Conversion
基于角度-光谱多样性转换的紧凑型低成本高光谱成像
Abstract
Snapshot hyperspectral imaging avoids sequential scanning, but systems that jointly achieve stable reconstruction, low cost, and compact optics remain limited. We present a snapshot hyperspectral imaging system based on angular-to-spectral diversity conversion. A tapered kaleidoscope creates replicated views with distinct incidence directions, and a directly attached birefringent filter converts them into view-channel-dependent spectral transmittances, yielding complementary measurements that better condition the inverse problem for more stable single-shot spectral reconstruction. The system preserves a simple pixel-wise linear model for fast non-learning-based reconstruction and uses only off-the-shelf components without relay optics or cascaded modules. We select the birefringent filter configuration using a condition-number-based criterion and validate the system on both synthetic and real data.
Chinese Translation
快照高光谱成像避免了顺序扫描，但能够同时实现稳定重建、低成本和紧凑光学的系统仍然有限。我们提出了一种基于角度-光谱多样性转换的快照高光谱成像系统。一个锥形万花筒创建了具有不同入射方向的重复视图，直接附加的双折射滤光片将其转换为视图通道依赖的光谱透射率，从而产生互补测量，更好地为逆问题提供条件，以实现更稳定的单次拍摄光谱重建。该系统保持简单的像素级线性模型，以便快速进行非学习型重建，并且仅使用现成组件，无需中继光学或级联模块。我们使用基于条件数的标准选择双折射滤光片配置，并在合成数据和真实数据上验证了该系统。
cs.CV / 107 / 2609.23655
Reassessing Global Gradient-Norm Imbalance in BLIP Fine-Tuning Across Physical Domains
重新评估BLIP微调中跨物理领域的全局梯度范数不平衡
Abstract
Imbalanced gradient magnitudes between the visual and language pathways of a vision-language model are often treated as a defect to be corrected. We test that premise for one family of correction, deliberately excluding adaptive, signal-driven schemes (e.g. BalGrad, OGM, PMR, CGGM), which are a mechanistically distinct class outside this study's scope. Measuring the language-to-visual gradient-norm ratio, reported in parameter-normalised form, across nine fine-tuning conditions, three seeds, and three captioning datasets spanning distinct physical domain shifts -- underwater, aerial, radiological -- we find imbalance magnitude varies markedly across domains with no predictable ordering. A plain learning-rate reduction cuts imbalance substantially and lands within a few BLEU points of the best method on every dataset. Staged freezing reduces the ratio on every domain yet never ranks first; a schedule-only control isolates freezing as the cause on one dataset but not the other two. Forcing the two gradient groups to equal magnitude drives per-parameter imbalance close to zero on every domain, yet is both the best result in the study and the worst placement among full fine-tuning methods, on different datasets, with identical settings. Reductions in gradient-norm ratio do not consistently predict captioning performance across domains, and how a given level of balance is reached matters as much as the level itself. As a secondary finding, a commonly reused LoRA configuration applied to BLIP silently adapts zero visual parameters; correcting it improves BLEU-4 on all three datasets.
Chinese Translation
视觉-语言模型的视觉路径和语言路径之间的梯度幅度不平衡通常被视为需要纠正的缺陷。我们测试了这一前提，故意排除了自适应、信号驱动的方案（例如 BalGrad、OGM、PMR、CGGM），这些方案在机制上属于本研究范围之外的不同类别。我们在九种微调条件、三个随机种子和三个涵盖不同物理领域转变的字幕数据集（包括水下、空中和放射学）中，测量了语言到视觉的梯度范数比，以参数归一化的形式报告。我们发现，不同领域之间的不平衡幅度差异显著，且没有可预测的排序。简单的学习率降低显著减少了不平衡，并且在每个数据集上都接近最佳方法的BLEU分数。分阶段冻结在每个领域都降低了比率，但从未排名第一；仅调度的对照实验在一个数据集上将冻结作为原因，但在其他两个数据集上则不然。强制两个梯度组的幅度相等使每个领域的每个参数的不平衡接近于零，然而这是本研究中最佳结果和在不同数据集上全微调方法中最差排名的结果，设置相同。梯度范数比的减少并不能始终预测跨领域的字幕性能，而达到特定平衡水平的方式与水平本身同样重要。作为一个次要发现，常用的LoRA配置应用于BLIP时默默适应了零个视觉参数；纠正这一点提高了所有三个数据集的BLEU-4分数。
cs.CV / 108 / 2609.23658
Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanisms
视频扩散模型为何违背物理规律？揭示注意力机制中的缺陷
Abstract
Despite impressive visual quality, state-of-the-art video diffusion models often generate content that violates real-world physical laws. While existing solutions rely on external priors or specialized data, we investigate the root cause by exploring the internal mechanisms of these models. Specifically, we present the first interpretability study on the ''motion planning'' process of text-to-video diffusion models, revealing how motion trajectories form during early denoising stages. Building upon the ''first shape, then details'' finding, we combine cross-attention trajectory patterns with causal head contributions to identify a specific subset of attention heads driving motion planning. Further, our self-attention analysis shows that Rotary Position Embedding (RoPE) induces excessive spatial attention decay. This causes early candidate regions to prematurely lock into physically implausible positions, suppressing reasonable trajectories in adjacent frames and triggering generation failure modes. To address this fundamental flaw, we propose a lightweight architectural modification that scales the frequency of RoPE across different denoising steps. This strategy reduces excessive attention decay, helping the model explore better candidate regions to establish coherent physical motion. Finally, training-free and training-based experiments confirm the effectiveness of our approach in enhancing the physical commonsense of generated videos.
Chinese Translation
尽管视觉质量令人印象深刻，最先进的视频扩散模型常常生成违反现实物理规律的内容。现有解决方案多依赖外部先验或专门数据，而我们通过探索这些模型的内部机制，调查其根本原因。具体而言，我们首次对文本到视频扩散模型中的“运动规划”过程进行了可解释性研究，揭示了运动轨迹在早期去噪阶段的形成机制。基于“先形状后细节”的发现，我们结合交叉注意力轨迹模式与因果头贡献，识别出驱动运动规划的特定注意力头子集。此外，自注意力分析表明，旋转位置编码（Rotary Position Embedding，RoPE）引起了过度的空间注意力衰减，导致早期候选区域过早锁定在物理上不合理的位置，抑制了相邻帧中合理轨迹的生成，触发了生成失败模式。为解决这一根本缺陷，我们提出了一种轻量级架构修改方案，通过在不同去噪步骤中调整RoPE的频率，减少过度的注意力衰减，帮助模型探索更优的候选区域以建立连贯的物理运动。最后，无需训练和基于训练的实验均验证了我们方法在增强生成视频物理常识方面的有效性。
cs.CV / 109 / 2609.23673
Which Terrain Is Better? Preference Learning with VLM Prototypes for Off-Road Traversability Ranking
哪种地形更好？基于视觉语言模型原型的越野可 traversability 排名的偏好学习
Abstract
In vision-based off-road navigation, a robot needs to know not only which obstacles to avoid but also which terrain is better. The first is handled by freespace detection or semantic segmentation. The second is usually answered with a traversability score, but no universal ground truth exists for such a score, so perception falls back on a predefined value per semantic class or a freespace confidence. These scores say what a region is, not which region a robot should prefer. We therefore formulate this preference as visual traversability ranking, an ordering of visible terrain that can be supervised by comparisons between two regions. Standard annotations do not label preference, but they imply its direction. We present TravPro, which converts these annotations into ordered region pairs and fits a small readout on frozen vision--language model (VLM) patch tokens to these pairs. The tokens are clustered once into a fixed prototype bank, and the readout learns a preference score per prototype. The readout is then applied to every patch and serves as a teacher that turns sparse comparisons into dense preference pseudo-labels without pixel-wise annotation. An RGB student distills these maps into a dense terrain-preference map together with a non-ground mask that excludes obstacles and background from the ranking. On five unseen domains, TravPro reaches a mean pairwise accuracy of 0.915 against 0.783 for the strongest baseline, producing an ordering sensitive to surface condition that a per-class value cannot represent. The same VLM and the same supervision yield no such ordering when the VLM is prompted and the supervision is used as dense targets; what matters is how they are used.
Chinese Translation
在基于视觉的越野导航中，机器人不仅需要知道应避免哪些障碍物，还需要了解哪种地形更佳。第一个问题通过自由空间检测或语义分割来处理。第二个问题通常通过可 traversability 分数来回答，但对于这种分数并不存在普遍的真实标准，因此感知依赖于每个语义类别的预定义值或自由空间置信度。这些分数说明了一个区域是什么，而不是机器人应该偏好的区域。因此，我们将这种偏好表述为视觉可 traversability 排名，即可见地形的排序，可以通过两个区域之间的比较进行监督。标准注释并不标记偏好，但它们暗示了偏好的方向。我们提出了 TravPro，它将这些注释转换为有序区域对，并将冻结的视觉-语言模型（VLM）补丁标记上的小读出拟合到这些区域对上。这些标记被聚类为一个固定的原型库，读出学习每个原型的偏好分数。然后将读出应用于每个补丁，并作为教师，将稀疏比较转化为密集的偏好伪标签，而无需逐像素注释。一种 RGB 学生将这些映射提炼为密集的地形偏好图，同时生成一个非地面掩膜，以排除障碍物和背景的排名。在五个未见领域中，TravPro 达到平均成对准确率 0.915，而最强基线为 0.783，生成的排序对表面条件敏感，而每类值无法表示。当 VLM 被提示并且监督作为密集目标使用时，相同的 VLM 和相同的监督并未产生这样的排序；关键在于它们的使用方式。
cs.CV / 110 / 2609.23679
Mind the Gaps: A Curated Benchmark for Form Field Detection
注意间隙：一个精心策划的表单字段检测基准
Abstract
Form Field Detection (FFD) is a fundamental component of document understanding systems, enabling applications ranging from large-scale industrial digitization to accessible form interaction for automated analysis. Unlike conventional object detection tasks, FFD is inherently challenging because fields are often defined by layout structure and whitespace rather than visible foreground content. Existing large-scale datasets frequently rely on heuristic annotation pipelines, resulting in noisy and inconsistent labels that hinder reliable evaluation. In this work, we introduce mini-CommonForms, a carefully curated FFD benchmark with consistent, high-quality annotations, and present a detailed evaluation of state-of-the-art detection approaches. The benchmark is designed to support reproducible research in document automation and accessibility-oriented applications. Dataset and code are available at https://github.com/moured/mini-commonforms
Chinese Translation
表单字段检测（FFD）是文档理解系统的一个基本组成部分，支持从大规模工业数字化到便捷的表单交互等多种应用，便于自动化分析。与传统的物体检测任务不同，FFD 本质上具有挑战性，因为字段通常是通过布局结构和空白而非可见的前景内容来定义的。现有的大规模数据集通常依赖启发式注释流程，导致标签噪声大且不一致，从而妨碍可靠评估。在本研究中，我们引入了 mini-CommonForms，这是一个经过精心策划的 FFD 基准，具有一致的高质量注释，并对最先进的检测方法进行了详细评估。该基准旨在支持文档自动化和面向可及性应用的可重复研究。数据集和代码可在 https://github.com/moured/mini-commonforms 获取。
cs.CV / 111 / 2609.23714
Infectious Bovine Pinkeye Detection Using Computer Vision and Imbalance-Aware Learning
基于计算机视觉和关注不平衡学习的传染性牛结膜炎检测
Abstract
Infectious bovine pinkeye is a contagious ocular disease that adversely affects cattle health, welfare, and agricultural productivity. Conventional diagnosis relies primarily on clinical observation, which can be subjective, time-consuming, and difficult to implement efficiently in large herds or remote settings. This study evaluated and compared You Only Look Once (YOLO) v11 and YOLOv26 for automated bovine pinkeye classification and investigated the effects of class-balancing strategies on model performance. Five variants (n, s, m, l, and x) of each architecture were trained and evaluated using the original imbalanced dataset, Random Minority Oversampling (RMO), and an adapted Synthetic Minority Oversampling Technique (SMOTE). Both YOLOv11 and YOLOv26 demonstrated strong classification performance, although the effects of class balancing varied across model variants. For YOLOv11, RMO-s achieved an accuracy of 0.99, a macro F1-score of 0.98, and a true positive rate (TPR) of 1.00, with no false-negative classifications. RMO-m also achieved a TPR of 1.00 with no false negatives. For YOLOv26, the original l, RMO-m, and RMO-l variants each achieved an accuracy of 0.99 and a macro F1-score of 0.98, with RMO-l attaining a TPR of 1.00 and no false negatives. Overall, RMO generally provided greater improvements in minority-class detection than adapted SMOTE, whereas the strong performance of the original YOLOv26-l demonstrates that oversampling was not necessary for all model variants. These findings demonstrate the potential of YOLOv11 and YOLOv26 for automated detection of bovine pinkeye and support further evaluation for livestock health monitoring.
Chinese Translation
传染性牛结膜炎是一种传染性眼病，严重影响牛的健康、福利和农业生产力。传统诊断主要依赖临床观察，这种方法可能主观、耗时，并且在大规模养殖场或偏远地区难以高效实施。本研究评估并比较了 You Only Look Once (YOLO) v11 和 YOLOv26 在自动化牛结膜炎分类中的表现，并探讨了类别平衡策略对模型性能的影响。对每种架构的五个变体（n、s、m、l 和 x）进行了训练和评估，使用了原始不平衡数据集、随机少数类过采样（Random Minority Oversampling, RMO）和改进的合成少数类过采样技术（Synthetic Minority Oversampling Technique, SMOTE）。YOLOv11 和 YOLOv26 均表现出强大的分类性能，尽管类别平衡的效果在模型变体之间有所不同。对于 YOLOv11，RMO-s 达到了 0.99 的准确率、0.98 的宏 F1 分数和 1.00 的真正阳性率（TPR），且没有假阴性分类。RMO-m 也达到了 1.00 的 TPR，且没有假阴性。对于 YOLOv26，原始 l、RMO-m 和 RMO-l 变体均达到了 0.99 的准确率和 0.98 的宏 F1 分数，其中 RMO-l 达到了 1.00 的 TPR，且没有假阴性。总体而言，RMO 通常在少数类检测中提供了比改进的 SMOTE 更大的提升，而原始 YOLOv26-l 的强大表现表明，并非所有模型变体都需要过采样。这些发现展示了 YOLOv11 和 YOLOv26 在自动检测牛结膜炎方面的潜力，并支持进一步评估以监测牲畜健康。
cs.CV / 112 / 2609.23715
Layer-Aware Position Embeddings for Visual Token Pruning in Multimodal Large Language Models
面向多模态大型语言模型的视觉标记修剪的层感知位置嵌入
Abstract
Multimodal large language models (MLLMs) incur substantial computational overhead due to the reliance on hundreds of visual tokens to represent images. While token pruning has emerged as a promising approach to reduce the inference cost of MLLMs, existing methods typically reassign position embeddings to the retained tokens using either sparse or continuous position embeddings, each introducing distinct limitations. Sparse position embeddings tend to decrease the attention value allocated to visual tokens, thereby degrading the perception capability of MLLMs, whereas continuous position embeddings disrupt the original spatial correspondence of visual tokens, leading to weakened grounding capability. To mitigate this issue, we perform layer-wise analysis of the language decoder and observe that intermediate layers play a critical role for maintaining the grounding capability of MLLMs under token pruning. Based on this observation, we propose a layer-aware position embedding strategy, which switches to sparse position embeddings at grounding-sensitive layers while maintaining continuous position embeddings elsewhere. Extensive experiments across representative pruning methods and diverse benchmarks demonstrate that our approach improves the comprehensive multimodal performance of pruned MLLMs compared with standard sparse and continuous position embeddings.
Chinese Translation
多模态大型语言模型（MLLMs）由于依赖数百个视觉标记来表示图像，因此产生了大量的计算开销。虽然标记修剪已成为降低MLLM推理成本的有前景的方法，但现有方法通常通过稀疏或连续位置嵌入重新分配保留标记的位置嵌入，每种方法都有其独特的局限性。稀疏位置嵌入往往会减少分配给视觉标记的注意力值，从而降低MLLM的感知能力，而连续位置嵌入则会破坏视觉标记的原始空间对应关系，导致基础能力减弱。为了解决这一问题，我们对语言解码器进行了逐层分析，观察到中间层在标记修剪下维持MLLM的基础能力方面起着关键作用。基于这一观察，我们提出了一种层感知位置嵌入策略，该策略在对基础能力敏感的层中切换到稀疏位置嵌入，而在其他层中保持连续位置嵌入。通过在代表性修剪方法和多样化基准上的大量实验，证明了我们的方法相比于标准的稀疏和连续位置嵌入，提升了修剪后MLLM的综合多模态性能。
cs.CV / 113 / 2609.23717
BindCLIP: One Balanced Coupling For Compositional Vision Language Scoring
BindCLIP：一种平衡耦合用于组合视觉语言评分
Abstract
Global vision--language similarities compress an image and a caption into one vector, preserving semantics but not which word corresponds to which region or how those regions are arranged; a model can recognize every word and object yet prefer a compositionally incorrect caption. We argue that a frozen encoder retains this association structure, so the problem is to read it rather than to rebuild it beside the pretrained similarity. We introduce BindCLIP, a pairwise scorer built on one latent object: a balanced token--patch--depth optimal-transport coupling that places both candidate captions and several visual depths in a single plan. Semantic, entity, order, and spatial evidence are read as energies of this state, and exchanging the candidates permutes the plan, making the score exactly antisymmetric. A geometric refinement inside the coupling contracts moves that the candidates and the visual depths do not support. No task label, parser, relation inventory, or detector is used. One checkpoint and one inference path improve the official What'sUp, ARO, and SugarCrepe benchmarks over frozen global CLIP, with the strongest transfer on the relation splits. Controls rule out patch access and caption-length shortcuts, and an inference-time lesion localizes spatial arrangement to the coupling.
Chinese Translation
全球视觉-语言相似性将图像和标题压缩为一个向量，保留语义但不保留哪个词对应于哪个区域或这些区域的排列方式；一个模型可以识别每个词和对象，但仍然偏好组合上不正确的标题。我们认为，冻结的编码器保留了这种关联结构，因此问题在于读取它，而不是在预训练的相似性旁边重建它。我们引入了BindCLIP，这是一种基于一个潜在对象的成对评分器：一种平衡的标记-补丁-深度最优传输耦合，将候选标题和多个视觉深度放置在一个单一的计划中。语义、实体、顺序和空间证据被视为该状态的能量，交换候选项会改变计划，使得评分完全反对称。耦合内部的几何细化收缩了候选项和视觉深度不支持的移动。没有使用任务标签、解析器、关系清单或检测器。一个检查点和一条推理路径在冻结的全球CLIP上改进了官方的What'sUp、ARO和SugarCrepe基准，在关系拆分上实现了最强的迁移。控制实验排除了补丁访问和标题长度捷径，而推理时的损伤定位将空间排列归结于耦合。
cs.CV / 114 / 2609.23733
VGGT-Prime: Compute-Adaptive Mixture-of-Heads for Efficient Visual Geometry Transformers
VGGT-Prime：计算自适应混合头以提高视觉几何变换器的效率
Abstract
Feed-forward visual geometry models such as the Visual Geometry Grounded Transformer (VGGT) have recently enabled direct 3D reconstruction from multi-view images. Despite their promising performance, these models scale quadratically with the number of input views due to their global attention mechanism, resulting in substantial latency for long sequence inputs. There have been some recent efforts to accelerate VGGT, but they primarily focus on reducing \emph{token redundancy} through token merging or key/value sparsification. Our work resolves this bottleneck from a different perspective by investigating \emph{architectural redundancy} in visual geometry transformers. We show that the multi-head attention modules in VGGT's global-attention layers contain substantial architectural redundancy, with only a subset of heads carrying critical geometric information. In light of this observation, we propose VGGT-Prime, a compute-adaptive mixture-of-heads model that resolves this redundancy to accelerate visual geometry transformers while maintaining competitive reconstruction quality. The key idea of VGGT-Prime is to estimate the appropriate computation level for each global-attention head using a lightweight router and then dynamically assign each head to different computation modes. Extensive experiments on multiple datasets demonstrate that VGGT-Prime can achieve an {$8\times$} inference speedup over VGGT while maintaining competitive performance on camera pose, depth, and point-cloud predictions. We further show that VGGT-Prime is complementary to existing acceleration methods, such as token merging, further improving inference speed by up to $14{\times}$ over VGGT. An overview of our work is available on our \href{https://vggt-prime.github.io}{project page}.
Chinese Translation
前馈视觉几何模型，如视觉几何基础变换器（Visual Geometry Grounded Transformer, VGGT），最近使得从多视图图像直接进行三维重建成为可能。尽管这些模型表现出良好的性能，但由于其全局注意力机制，输入视图数量的增加导致其计算复杂度呈平方增长，从而在长序列输入时造成显著延迟。近期有一些加速VGGT的努力，但主要集中在通过令牌合并或键/值稀疏化来减少 extit{令牌冗余}。我们的工作从不同的角度解决了这一瓶颈，研究了视觉几何变换器中的 extit{架构冗余}。我们展示了VGGT的全局注意力层中的多头注意力模块包含大量架构冗余，只有一部分头部携带关键的几何信息。基于这一观察，我们提出了VGGT-Prime，一种计算自适应的混合头模型，旨在解决这一冗余，以加速视觉几何变换器，同时保持竞争性的重建质量。VGGT-Prime的关键思想是使用轻量级路由器估计每个全局注意力头的适当计算水平，然后动态地将每个头分配到不同的计算模式。对多个数据集的广泛实验表明，VGGT-Prime能够实现对VGGT的{$8 imes$}推理加速，同时在相机姿态、深度和点云预测方面保持竞争性能。我们进一步展示了VGGT-Prime与现有加速方法（如令牌合并）是互补的，进一步将推理速度提高至VGGT的$14{ imes}$。我们工作的概述可在我们的 extit{项目页面}上查看。
cs.CV / 115 / 2609.23753
OnlineWM: Causality-Aware Active Online Learning for Effective World Modeling
OnlineWM：一种关注因果关系的主动在线学习方法用于有效的世界建模
Abstract
Generative world models aim to predict future states conditioned on actions, where action controllability is fundamental for reliable dynamics modeling. While recent efforts leverage simulator-generated data to enhance this capability, existing training pipelines face two fundamental limitations. First, static offline data collection leads to a distribution misalignment between training sets and the model's evolving error patterns, failing to resolve critical long-tail scenarios where dynamics predictions remain unreliable. Second, the standard objective of minimizing observational discrepancy often encourages the model to exploit spurious correlations instead of capturing the underlying action-effect causality. To address these limitations, we propose OnlineWM, an online training framework that continuously improves world modeling through active simulator interaction and causality-aware optimization. OnlineWM introduces two key innovations: (1) Active Online Learning: Instead of using fixed datasets, OnlineWM adaptively queries the simulator for new interaction sequences that target the model's current predictive weaknesses, ensuring high-utility data acquisition. (2) Causality-Aware Fine-Tuning: We propose a counterfactual learning strategy that contrasts the outcomes of different actions from identical states, forcing the model to attribute state transitions to specific actions rather than ambient environmental evolution, thereby grounding its predictions in reliable causal mechanisms. By integrating active data acquisition with causal optimization, OnlineWM establishes a closed-loop refinement process that ensures the model is both robust to diverse scenarios and precise in its causal attribution. Extensive experiments demonstrate that OnlineWM significantly enhances action controllability and generalizes effectively to unseen domains.
Chinese Translation
生成性世界模型旨在根据动作预测未来状态，其中动作可控性是可靠动态建模的基础。尽管最近的研究利用模拟器生成的数据来增强这一能力，但现有的训练流程面临两个基本限制。首先，静态的离线数据收集导致训练集与模型不断演变的误差模式之间的分布不一致，未能解决动态预测仍然不可靠的关键长尾场景。其次，最小化观察差异的标准目标往往促使模型利用虚假的相关性，而不是捕捉潜在的动作-效果因果关系。为了解决这些限制，我们提出了OnlineWM，一种通过主动与模拟器交互和关注因果关系的优化来持续改进世界建模的在线训练框架。OnlineWM引入了两个关键创新：（1）主动在线学习：OnlineWM不再使用固定数据集，而是自适应地查询模拟器以获取针对模型当前预测弱点的新交互序列，从而确保高效的数据获取。（2）关注因果关系的微调：我们提出了一种反事实学习策略，比较相同状态下不同动作的结果，迫使模型将状态转变归因于特定动作，而不是环境的自然演变，从而使其预测基于可靠的因果机制。通过将主动数据获取与因果优化相结合，OnlineWM建立了一个闭环的精细化过程，确保模型对多样场景具有鲁棒性，并在因果归因上具有精确性。大量实验表明，OnlineWM显著增强了动作可控性，并有效地推广到未见领域。
cs.CV / 116 / 2609.23758
Training-Free Spectral Transductive Refinement for Cross-Domain Few-Shot Classification
无训练的光谱传导细化用于跨域少样本分类
Abstract
Few-shot recognition with frozen visual features is especially fragile under domain shift and one-shot supervision, where a single labelled image is an unreliable estimate of its class. We ask how far this fragility can be reduced purely at test time, without retraining the encoder or augmenting the source domain. We present Spectral Transductive Refinement (STR), a training-free transductive inference rule that exploits the geometry of the complete support-query episode. Given frozen embeddings, STR builds a joint k-nearest-neighbour graph, maps the episode into a normalized-Laplacian spectral coordinate system, initializes class representatives from the labelled support, and iteratively refines them using pseudo-labelled queries. We evaluate STR under two protocols. A controlled component study with frozen ResNet-18 features shows that spectral refinement consistently improves over single-prototype spectral initialization across five shifted domains, with the largest gains in the one-shot regime where support estimates are weakest. We then benchmark STR against recent Cross-Domain Few-Shot Learning (CD-FSL) methods using the standard miniImageNet-pretrained ResNet-10 backbone over eight established target domains. Operating entirely at inference time, STR attains the highest 1-shot average among compared methods and remains competitive at 5-shot, rivalling approaches relying on heavy source-domain meta-training augmentations. Because STR is transductive, we report its setting explicitly. Diagnostics attribute its gains to iterative refinement in spectral coordinates rather than added prototype capacity, which remains inactive in our configuration.
Chinese Translation
在领域转移和单次监督下，冻结视觉特征的少样本识别尤其脆弱，其中单个标记图像对其类别的估计不可靠。我们探讨在不重新训练编码器或增强源领域的情况下，如何在测试阶段尽可能减少这种脆弱性。我们提出了光谱传导细化（Spectral Transductive Refinement, STR），这是一种无训练的传导推断规则，利用完整支持-查询情境的几何特性。在给定冻结嵌入的情况下，STR构建一个联合k近邻图，将情境映射到归一化拉普拉斯光谱坐标系统，从标记支持中初始化类别代表，并使用伪标记查询迭代细化它们。我们在两个协议下评估STR。通过冻结的ResNet-18特征进行的受控组件研究表明，光谱细化在五个转移领域中始终优于单原型光谱初始化，尤其是在支持估计最弱的单次情境下获得最大增益。然后，我们在八个已建立的目标领域中，使用标准的miniImageNet预训练ResNet-10骨干网络，将STR与最近的跨域少样本学习（Cross-Domain Few-Shot Learning, CD-FSL）方法进行基准测试。STR完全在推断时间操作，在比较方法中获得了最高的1-shot平均值，并在5-shot中保持竞争力，能够与依赖于重源域元训练增强的方法相抗衡。由于STR是传导性的，我们明确报告其设置。诊断结果将其增益归因于光谱坐标中的迭代细化，而不是增加的原型容量，因为在我们的配置中该容量保持不活跃。
cs.CV / 117 / 2609.23769
PRISM-RAG: Multimodal Hypergraph Retrieval-Augmented Generation for Tobacco Product and Legislative Policy Reasoning
PRISM-RAG：用于烟草产品和立法政策推理的多模态超图检索增强生成
Abstract
The disambiguation of semantically similar statutory text across jurisdictions is a retrieval problem that existing methods do not solve. This inter-context conflict can steer generative models toward confidently produced answers grounded in topically relevant but jurisdictionally incorrect sources. Tobacco and nicotine regulations vary by US jurisdiction, often sharing similar language, thus, robust reasoning requires identifying which jurisdiction's law governs a given product, not merely retrieving relevant text. Emerging products (e.g., pouches) exploit ambiguous definitions to evade regulation. State-of-the-art (SOTA) document retrieval-augmented generation (RAG) methods struggle to address this inter-context conflict, and thus struggle to connect image attributes (e.g., rich attribute captions) to the set of similar legislation texts. We introduce NicoPRISM (Nicotine Product and Regulation Image-and-Text Surveillance Multimodal), comprising 161,563 images, attribute captions, a knowledge base of product, health, and legislative documents spanning 13 US jurisdictions, and 1,495 validated question-answer pairs across two tasks: policy compliance QA and product knowledge QA. We also propose PRISM-RAG, a multimodal hypergraph RAG framework built over images, captions, and entities without any LLM calls at index time, grounding every query in a product image and routes retrieval through a jurisdiction-aware context assembly mechanism guaranteeing that statutory text from the queried jurisdiction reaches the language model by construction. PRISM-RAG retrieves passages from the correct jurisdiction in 93.9% of policy compliance queries, a 48.6 percentage point advantage over standard RAG (p<0.001), using zero LLM calls at index time and one at query time, and is competitive with or outperforms SOTA RAG frameworks across keyword, semantic, jurisdiction-, and compliance-accuracy metrics.
Chinese Translation
跨管辖区语义相似法条文本的消歧义是一个现有方法未能解决的检索问题。这种跨上下文冲突可能会使生成模型朝向自信地产生基于主题相关但管辖区不正确来源的答案。烟草和尼古丁法规因美国各州的不同而有所差异，通常共享相似的语言，因此，稳健的推理需要识别哪个管辖区的法律适用于特定产品，而不仅仅是检索相关文本。新兴产品（例如，烟草袋）利用模糊定义逃避监管。最先进的（SOTA）文档检索增强生成（RAG）方法在解决这种跨上下文冲突方面面临挑战，因此难以将图像属性（例如，丰富的属性标题）与一组相似的立法文本连接起来。我们引入了NicoPRISM（尼古丁产品和监管图像与文本监测多模态），该数据集包含161,563张图像、属性标题、涵盖13个美国管辖区的产品、健康和立法文件的知识库，以及在政策合规问答和产品知识问答两个任务中验证的1,495对问答。我们还提出了PRISM-RAG，这是一种基于图像、标题和实体构建的多模态超图RAG框架，在索引时不进行任何LLM调用，将每个查询基于产品图像进行定位，并通过一个管辖区感知的上下文组装机制进行检索，确保查询的管辖区的法条文本通过构建方式到达语言模型。PRISM-RAG在政策合规查询中以93.9%的准确率从正确的管辖区检索段落，比标准RAG高出48.6个百分点（p<0.001），在索引时使用零次LLM调用，在查询时使用一次LLM调用，并在关键词、语义、管辖区和合规准确性指标上与SOTA RAG框架竞争或表现优越。
cs.CV / 118 / 2609.23796
Mira-Scene: Pixel-Aligned Layouts for Generative 3D Scene
Mira-Scene：用于生成3D场景的像素对齐布局
Abstract
Single-image 3D object generation can now produce high-fidelity assets, yet accurately placing them into a coherent scene layout remains an open challenge. A central difficulty lies in how object layout is represented. Holistic methods absorb placement into a scene-level generation process, sacrificing object-level detail. Compositional methods preserve object fidelity by decoupling geometry from layout, but typically parameterize layout as sparse, unbounded pose variables that are difficult to learn and generalize poorly under scarce scene-level supervision.We present Mira-Scene, a compositional 3D scene reconstruction framework that replaces sparse pose regression with dense, bounded correspondence recovery. At its core is the Canonical Coordinate Map (CCM), a pixel-aligned field that maps each visible object pixel to a surface coordinate in the object's bounded canonical space. When paired with a scene-space Point Cloud Map (PCM) from monocular geometry estimation, CCM induces dense canonical-to-scene correspondences from which object transformations are recovered through robust geometric alignment. Because CCM operates in bounded canonical space, it provides a stable prediction target that can be trained from scalable object-level 3D data without requiring scene-level layout annotations. Mira-Scene further introduces a multimodal diffusion transformer that jointly generates object geometry and CCMs, using modality-specific expert streams with shared attention and positional encoding to promote geometry-layout consistency. Experiments on indoor, outdoor, synthetic, and in-the-wild scenes show that Mira-Scene substantially outperforms strong baselines in layout accuracy, achieving relative gains of 39.8% in 3D-IoU and 16.5% in 2D-IoU over SAM3D, using limited open-source training data.
Chinese Translation
单幅图像的3D物体生成现已能够产生高保真资产，但将它们准确地放置到一个连贯的场景布局中仍然是一个未解决的挑战。一个主要的困难在于物体布局的表示方式。整体方法将放置吸收到场景级生成过程中，牺牲了物体级细节。组合方法通过将几何与布局解耦来保持物体的保真度，但通常将布局参数化为稀疏的、无界的姿态变量，这些变量难以学习且在稀缺的场景级监督下泛化效果较差。我们提出了Mira-Scene，一个组合式3D场景重建框架，它用密集的、有界的对应关系恢复替代了稀疏的姿态回归。其核心是规范坐标图（Canonical Coordinate Map, CCM），这是一个像素对齐的场，映射每个可见物体像素到物体有界规范空间中的表面坐标。当与来自单目几何估计的场景空间点云图（Point Cloud Map, PCM）配对时，CCM引导出密集的规范到场景的对应关系，从中通过稳健的几何对齐恢复物体变换。由于CCM在有界规范空间中操作，它提供了一个稳定的预测目标，可以从可扩展的物体级3D数据中进行训练，而无需场景级布局注释。Mira-Scene进一步引入了一种多模态扩散变换器，联合生成物体几何和CCM，使用特定模态的专家流与共享的注意力和位置编码，以促进几何与布局的一致性。在室内、室外、合成和自然场景上的实验表明，Mira-Scene在布局准确性上显著优于强基线，相较于SAM3D，在3D-IoU上实现了39.8%的相对提升，在2D-IoU上实现了16.5%的提升，且使用了有限的开源训练数据。
cs.CV / 119 / 2609.23815
Confidence-Aware Teacher-Student Distillation for 3D Medical Segmentation
基于置信度的教师-学生蒸馏方法用于三维医学分割
Abstract
Medical image segmentation models typically rely on large amounts of densely annotated volumetric data, limiting their scalability across tasks and imaging modalities. This work addresses the challenge of predicting entire 3D anatomical structures from extreme annotation sparsity. An annotation-efficient student-teacher framework is proposed for automatic 3D medical segmentation that requires only a set of point prompts on a single 2D slice per volume, as input. A foundation model serves as an offline teacher, utilizing the provided point prompts from the selected slice to full-volume pseudo-annotations alongside their corresponding spatial confidence scores prior to student training. To mitigate the error propagation of noisy pseudo-annotations, a task-specific 3D student network is trained using a confidence-aware optimization strategy. By leveraging the teacher's pre-computed confidence scores, this strategy explicitly excludes statically uncertain regions of the pseudo-annotations from the loss calculation, while simultaneously emphasizing regions with higher confidence. Evaluated on 3D cardiac MRI datasets, our framework outperforms state-of-the-art semi-supervised methods, improving segmentation performance by up to 43.6%. Furthermore, it drastically reduces the manual annotation burden to just a few positive point prompts per volume, while improving surface boundary precision by up to 14.7% over the teacher and successfully recovering up to 34.1% of the performance gap toward the fully supervised upper bound.
Chinese Translation
医学图像分割模型通常依赖大量密集标注的体积数据，这限制了它们在不同任务和成像模式中的可扩展性。本研究解决了从极度稀疏的标注中预测整个三维解剖结构的挑战。我们提出了一种注释高效的学生-教师框架，用于自动三维医学分割，该框架仅需在每个体积的单个二维切片上提供一组点提示作为输入。基础模型作为离线教师，利用所选切片提供的点提示生成全体积伪标注及其对应的空间置信度分数，以便进行学生训练。为了减轻噪声伪标注的错误传播，采用了一种基于置信度的优化策略训练特定任务的三维学生网络。通过利用教师预先计算的置信度分数，该策略明确地将伪标注中静态不确定的区域排除在损失计算之外，同时强调置信度较高的区域。在三维心脏MRI数据集上的评估表明，我们的框架超越了最先进的半监督方法，分割性能提高了多达43.6%。此外，它将手动标注负担大幅减少至每个体积仅需几个正点提示，同时在表面边界精度上比教师提高了多达14.7%，并成功缩小了与完全监督上限之间的性能差距，恢复了多达34.1%。
cs.CV / 120 / 2609.23817
VISTA: Video-Injected Stylized Text-to-Animation
VISTA：视频注入风格化文本到动画
Abstract
We present VISTA, a two-stage framework for generating stylized 3D human motion by fusing structural content from text prompts with expressive style from reference videos, without requiring jointly paired (text, video, stylized motion) triplets. A Dual-channel Autoencoder first maps motion sequences and video clips into a shared latent manifold. A masked autoregressive diffusion backbone then operates within this manifold, injecting video-derived style through a dedicated late-fusion Dual-AdaLN pathway while preserving text-conditioned content structure. A cross-batch unpaired training protocol with latent cycle consistency enables joint learning across separate semantically rich and stylistically diverse datasets. As a proof-of-concept for controllable animation synthesis, we validate VISTA on rendered motion-capture references: it achieves the highest style recognition accuracy among video-conditioned methods while preserving competitive content alignment, and its decomposed 3-way classifier-free guidance provides independent, user-controllable calibration of the content--style balance at inference time.
Chinese Translation
我们提出了VISTA，一个用于生成风格化3D人类运动的两阶段框架，通过将文本提示中的结构内容与参考视频中的表现风格融合，而无需联合配对的（文本、视频、风格化运动）三元组。一个双通道自编码器首先将运动序列和视频片段映射到一个共享的潜在流形中。然后，一个掩蔽自回归扩散骨干网络在该流形内操作，通过专用的后期融合双重自适应层归一化（Dual-AdaLN）路径注入视频衍生的风格，同时保持文本条件下的内容结构。一个具有潜在循环一致性的跨批次无配对训练协议使得在不同语义丰富和风格多样的数据集之间进行联合学习成为可能。作为可控动画合成的概念验证，我们在渲染的动作捕捉参考上验证了VISTA：它在视频条件方法中实现了最高的风格识别准确率，同时保持了竞争性的内容对齐，其分解的三路无分类器引导提供了在推理时独立的、用户可控的内容与风格平衡的校准。
cs.CV / 121 / 2609.23830
DFD-Lab: A Modular Audio-Visual Deepfake Detection Pipeline
DFD-Lab：一个模块化的音视频深度伪造检测管道
Abstract
Comparing audio-visual deepfake detectors requires coordinating dataset adaptation, temporal input representation, model interfaces and experimental conditions. We present DFD-Lab, a modular pipeline that separates these responsibilities while supporting shared training and evaluation workflows. We integrate three implementations: Xception-based maximum-logit fusion, ResNet with temporal LSTM fusion, and our AVFF reimplementation. Experiments cover external testing, degradation-based training augmentation and evaluation-time corruption. On a filtered subset of Deepfake-Eval-2024, models trained on FakeAVCeleb attain baseline AUROC values of 0.504, 0.538 and 0.458. JPEG50 training augmentation raises these to 0.691, 0.605 and 0.570, respectively, while all three accuracies decrease. These results illustrate why training interventions, evaluation corruptions and metric-dependent outcomes should remain distinct within a common pipeline. The contribution is the integration of audio-visual processing, interchangeable detectors and configurable experimental workflows, supported by empirical case studies. The findings highlight the challenge of cross-dataset detection and the complementary information provided by ranking and classification metrics.
Chinese Translation
比较音视频深度伪造检测器需要协调数据集适配、时间输入表示、模型接口和实验条件。我们提出了DFD-Lab，一个模块化管道，分离了这些职责，同时支持共享的训练和评估工作流程。我们整合了三种实现：基于Xception的最大对数融合、带有时间LSTM融合的ResNet，以及我们的AVFF重新实现。实验涵盖了外部测试、基于退化的训练增强和评估时的损坏。在经过筛选的Deepfake-Eval-2024子集上，在FakeAVCeleb上训练的模型达到了基线AUROC值分别为0.504、0.538和0.458。JPEG50训练增强将这些值提高到0.691、0.605和0.570，但所有三种准确率均有所下降。这些结果说明了为什么训练干预、评估损坏和依赖于指标的结果应在一个共同的管道中保持独立。我们的贡献在于整合了音视频处理、可互换的检测器和可配置的实验工作流程，并得到了实证案例研究的支持。研究结果突显了跨数据集检测的挑战以及排名和分类指标所提供的互补信息。
cs.CV / 122 / 2609.23832
Comparative Performance and Parameter-Efficient Adaptation of DINOv2 for Active Trachoma Classification
DINOv2在主动性沙眼分类中的比较性能与参数高效适应
Abstract
Automated grading of conjunctival photographs could reduce the cost and variability of trachoma prevalence surveys, but the relative value of modern pretrained visual representations, lightweight feature adaptation, and training-objective design has not been established under a common protocol. This study presents a controlled evaluation for binary classification of Trachomatous Inflammation-Follicular (TF) versus Normal using 1,546 images from the public UCSF/Lietman collection. Images are processed using the OPTED pipeline for zero-shot tarsal-conjunctiva segmentation, alignment, cropping, and standardization. We first compare six pretrained backbones using a common classification pipeline and then evaluate four lightweight adaptation mechanisms on DINOv2 ViT-B/14. Under stratified five-fold cross-validation, DINOv2 with Efficient Channel Attention (ECA) and focal-plus-center loss achieved 91.66 +/- 0.97% accuracy, 90.69 +/- 1.10% macro-F1, and 96.06 +/- 0.71% AUC. ECA introduces only five learnable parameters while matching the performance of substantially larger alternatives. Objective ablation further showed that ECA did not consistently improve plain DINOv2 across loss functions; the lowest-variance 91.66% accuracy was obtained with cross-entropy plus center loss. Overall, the fine-tuned DINOv2 representation provided most of the predictive performance, while ECA offered a highly parameter-efficient refinement whose effect depended on the training objective. The resulting workflow provides a reproducible benchmark for active trachoma image classification.
Chinese Translation
自动化评估结膜照片可以降低沙眼流行病学调查的成本和变异性，但现代预训练视觉表征、轻量级特征适应和训练目标设计的相对价值尚未在统一协议下建立。本研究针对沙眼炎-滤泡型（TF）与正常的二分类进行了控制评估，使用了来自公共UCSF/Lietman收藏的1,546张图像。图像通过OPTED管道进行零样本睑结膜分割、对齐、裁剪和标准化处理。我们首先使用共同的分类管道比较了六个预训练主干网络，然后在DINOv2 ViT-B/14上评估了四种轻量级适应机制。在分层五折交叉验证下，DINOv2结合高效通道注意力（ECA）和焦点加中心损失达到了91.66 +/- 0.97%的准确率、90.69 +/- 1.10%的宏F1值和96.06 +/- 0.71%的AUC。ECA仅引入了五个可学习参数，同时匹配了显著更大替代方案的性能。目标消融进一步表明，ECA并未在所有损失函数中始终改善普通DINOv2；最低方差的91.66%准确率是在交叉熵加中心损失下获得的。总体而言，微调后的DINOv2表征提供了大部分预测性能，而ECA则提供了一种高度参数高效的精炼，其效果依赖于训练目标。最终的工作流程为主动沙眼图像分类提供了可重复的基准。
cs.CV / 123 / 2609.23881
MotionJEPA: Preventing Temporal Feature Collapse by Capturing Visual Changes in Latent Space
MotionJEPA：通过捕捉潜在空间中的视觉变化来防止时间特征崩溃
Abstract
Joint Embedding Predictive Architectures (JEPAs) are a promising paradigm for learning task-agnostic latent world models without visual reconstruction. However, standard JEPA training exhibits a strong inductive bias towards slow features, causing feature suppression and the collapse of latent representation. While inverse dynamics provides temporal anti-collapse, it relies on action labels and offers little incentive to embed general, unlabeled dynamics. We introduce Difference Image and Single image embedding Regularization (DISReg), a novel regularizer that builds on an inverse-dynamics-style module that predicts temporal difference image embeddings without any pixel reconstruction loss, encouraging balanced static and dynamic feature learning. DISReg consists of a static term that shapes the distribution of the image embedding and encourages slow features, and a dynamic term, which, unlike direct regularization on the embedding, imposes no constraint on the image embedding's shape or distribution and instead only incentivizes that dynamic features be present. By integrating this regularizer into a standard JEPA, we establish our new architecture, MotionJEPA. Latent probing demonstrates that MotionJEPA produces more complete representations than other methods, and our trajectory analysis shows it maintains geometrically simple latent embeddings with low curvature. We further show that MotionJEPA improves downstream planning success under static-background distractors across four environments.
Chinese Translation
联合嵌入预测架构（JEPAs）是一种有前景的范式，用于学习与任务无关的潜在世界模型，而无需视觉重建。然而，标准的JEPA训练对慢特征表现出强烈的归纳偏差，导致特征抑制和潜在表示的崩溃。虽然逆动力学提供了时间反崩溃的机制，但它依赖于动作标签，并且对嵌入一般的、未标记的动态提供的激励较少。我们提出了差异图像和单图像嵌入正则化（DISReg），这是一种新颖的正则化器，基于一种逆动力学风格的模块，该模块在没有任何像素重建损失的情况下预测时间差异图像嵌入，从而鼓励平衡静态和动态特征的学习。DISReg由一个静态项组成，该项塑造图像嵌入的分布并鼓励慢特征的存在，以及一个动态项，该项与对嵌入的直接正则化不同，不对图像嵌入的形状或分布施加任何约束，而只是激励动态特征的存在。通过将这一正则化器集成到标准JEPA中，我们建立了新的架构MotionJEPA。潜在探测表明，MotionJEPA生成的表示比其他方法更为完整，而我们的轨迹分析显示它保持几何上简单且曲率低的潜在嵌入。我们进一步表明，MotionJEPA在四个环境中改善了在静态背景干扰下的下游规划成功率。
cs.CV / 124 / 2609.23961
Colon3R: Cross-Domain 3D Reconstruction from Monocular Colonoscopic Video
Colon3R：基于单目结肠镜视频的跨域三维重建
Abstract
Monocular colonoscopic 3D reconstruction is important for surgical robotic colonoscopy, but remains challenging due to weak texture, specular reflections, limited view overlap, and non-rigid tissue motion. Conventional multi-view 3D reconstruction methods rely on stable correspondences and approximate rigidity, which are often violated in colonoscopy. Existing endoscopic methods often rely on domain-specific supervision, whereas there are not enough in-vivo labeled data available to adapt geometry foundation models to clinical colonoscopy. We present Colon3R, a cross-domain semi-supervised framework built on pretrained VGGT that transfers coupled camera, depth, and pointmap geometry from labeled phantom and simulated data to unlabeled in-vivo colonoscopy without requiring target-domain geometric annotations. Unlike source-only fine-tuning, which learns only from phantom and simulated data, Colon3R directly exploits unlabeled in-vivo video through teacher-derived cross-view supervision. Our proposed hierarchical quasi-rigid reliability selects reliable supervision at the sequence, directed-pair, and pixel levels, while source-preserving adaptation retains the learned coupled geometry during target-domain adaptation. Extensive experiments demonstrate that our method achieves superior overall performance over state-of-the-art approaches in depth, pointmap, and camera pose estimation. Qualitative comparisons on real in-vivo colonoscopy further show substantially more complete and geometrically consistent reconstructions than competing methods under clinical domain shift. The code will be public available after the paper is accepted.
Chinese Translation
单目结肠镜三维重建对于外科机器人结肠镜检查至关重要，但由于纹理弱、镜面反射、视角重叠有限以及非刚性组织运动等问题，仍然具有挑战性。传统的多视图三维重建方法依赖于稳定的对应关系和近似刚性，但在结肠镜检查中这些条件往往不成立。现有的内窥镜方法通常依赖于特定领域的监督，而可用于将几何基础模型适应于临床结肠镜检查的体内标注数据并不充足。我们提出了Colon3R，一种基于预训练VGGT的跨域半监督框架，该框架将标注的幻影和模拟数据中的耦合相机、深度和点图几何信息转移到未标注的体内结肠镜检查中，而无需目标领域的几何标注。与仅从幻影和模拟数据学习的源域微调不同，Colon3R通过教师导出的跨视图监督直接利用未标注的体内视频。我们提出的层次化准刚性可靠性在序列、定向对和像素级别选择可靠的监督，而源保持适应则在目标领域适应过程中保留学习到的耦合几何。大量实验表明，我们的方法在深度、点图和相机位姿估计方面的整体性能优于最先进的方法。在真实的体内结肠镜检查中的定性比较进一步显示，在临床领域转变下，我们的方法提供了显著更完整且几何一致的重建。代码将在论文接受后公开。
cs.CV / 125 / 2609.23967
Rethinking Diffusion Segmentation: When Does It Rely on Its Noisy State, and Does Diffusion Matter?
重新思考扩散分割：何时依赖于其噪声状态，扩散是否重要？
Abstract
Diffusion models are increasingly adapted from generation to conditional prediction, where a conditioning signal is combined with an evolving noisy representation of the target. In fully supervised segmentation, however, the conditioning image can already support direct target prediction, so endpoint performance alone establishes neither reliance on the added diffusion state nor a deterministic advantage over image-only prediction. For state reliance, we disrupt target-derived state content or correct image-state pairing during retraining of twelve published methods across three datasets, with ten matched seeds per setting. All 40 original-method comparisons whose evaluated-mask routes remained downstream of noised-quantity reconstruction exhibited state reliance, whereas all 30 comparisons with a segmentation-supervised bypass preserved reference performance. Rerouting five originally bypass-capable methods by forcing segmentation supervision through noise-to-mask reconstruction converted all 30 corresponding comparisons from preserved performance to state reliance. For deterministic utility, matched image-only counterparts achieved similar or better performance in 28 of 35 settings overall, including 16 of 20 whose native methods relied on both audited state properties. These results identify supervision path as a determinant of state reliance in the audited methods. Separately, matched image-only counterparts show that diffusion-specific computation often provides no deterministic endpoint advantage, including in methods that rely on the audited state properties. More generally, when conditioning already supports strong target prediction, diffusion-specific claims require additional evidence that the added state is used and that diffusion-specific computation improves the claimed capability beyond a matched condition-only counterpart.
Chinese Translation
扩散模型越来越多地从生成任务适应到条件预测，其中条件信号与目标的不断演变的噪声表示相结合。然而，在完全监督的分割中，条件图像已经能够支持直接的目标预测，因此仅凭最终性能无法确定对附加扩散状态的依赖性，也无法确定其相对于仅图像预测的确定性优势。为了检验状态依赖性，我们在三个数据集上对十二种已发布方法的目标衍生状态内容进行干扰，或在重新训练过程中纠正图像-状态配对，每种设置下有十个匹配的随机种子。所有40个原始方法比较中，那些评估的掩码路径仍然位于噪声量重建的下游的比较显示出状态依赖性，而所有30个具有分割监督旁路的比较则保持了参考性能。通过强制分割监督进行噪声到掩码重建，重新路由五种原本能够旁路的方法，将所有30个对应的比较从保持性能转变为状态依赖性。在确定性效用方面，匹配的仅图像对应物在35个设置中有28个实现了相似或更好的性能，包括20个原生方法依赖于审计状态属性的16个。这些结果表明监督路径是审计方法中状态依赖性的决定因素。此外，匹配的仅图像对应物显示，扩散特定计算通常并未提供确定性的最终优势，包括依赖于审计状态属性的方法。更一般而言，当条件已经支持强大的目标预测时，扩散特定的主张需要额外证据来证明附加状态被使用，并且扩散特定计算在超越匹配条件仅对应物的能力方面有所改善。
cs.CV / 126 / 2609.23983
Evaluating the Generalization of Neuroimaging Foundation Models on African Brain MRI
评估神经影像基础模型在非洲脑部MRI上的泛化能力
Abstract
Neuroimaging foundation models pretrained on large, predominantly western cohorts are increasingly proposed as general-purpose backbones for brain MRI analysis. Yet, their ability to generalize to underrepresented clinical populations remains largely untested. We evaluate four recent foundation models (BrainIAC, Neuro-JEPA, NeuroVFM, and Primus) on a three-way diagnostic classification task (Control, Dementia, Parkinson's disease) using a cohort of 88 subjects from a Nigerian clinical brain MRI dataset, across four modality configurations (T1w, T2w, T1w+T2w, FLAIR), and compare against an end-to-end trained ViT3D baseline. The frozen backbones collapse to majority-class predictions, while Neuro-JEPA on FLAIR shows modest but still limited discrimination. In contrast, the end-to-end trained ViT3D achieves higher accuracy and MCC on every task (up to 53.4% accuracy, MCC=0.27) and is the only model with non-trivial recall. Our findings suggest that these frozen neuroimaging foundation models are insufficient for fine-grained diagnostic classification in small, non-western clinical cohorts, motivating parameter-efficient adaptation and broader multi-site external validation for equitable deployment in global health settings.
Chinese Translation
在大型、主要为西方人群预训练的神经影像基础模型日益被提议作为脑部MRI分析的通用骨干。然而，它们在代表性不足的临床人群中的泛化能力仍然未得到充分测试。我们在一个由88名受试者组成的尼日利亚临床脑部MRI数据集中，评估了四个近期的基础模型（BrainIAC、Neuro-JEPA、NeuroVFM和Primus）在三分类诊断任务（对照组、痴呆、帕金森病）上的表现，涵盖四种模态配置（T1w、T2w、T1w+T2w、FLAIR），并与一个端到端训练的ViT3D基线模型进行比较。冻结的骨干模型倾向于做出多数类预测，而Neuro-JEPA在FLAIR模态下显示出适度但仍有限的区分能力。相比之下，端到端训练的ViT3D在每个任务上都达到了更高的准确率和MCC（最高准确率为53.4%，MCC=0.27），并且是唯一一个具有非平凡召回率的模型。我们的研究结果表明，这些冻结的神经影像基础模型在小型非西方临床人群中的细粒度诊断分类能力不足，促使我们进行参数高效的适应和更广泛的多中心外部验证，以实现全球健康环境中的公平部署。
cs.CV / 127 / 2609.24014
WebMRIQC: A Web-Based Implementation of MRIQC for Accessible MRI Image Quality Assessment in Resource-Constrained Settings
WebMRIQC：一种基于网络的MRIQC实现，旨在资源受限环境中实现可访问的MRI图像质量评估
Abstract
Reliable quality control (QC) of magnetic resonance imaging (MRI) is essential for reliable diagnostic neuroimaging, yet standard manual assessment is subjective and time-consuming. MRIQC has established standardized automated extraction of image-quality metrics (IQMs), but its reliance on local computational imaging skills and capacity including high-performance computing, limits its adoption in resource-constrained settings (RCS). We present WebMRIQC (webmriqc.mailab.io), an open-source browser-based platform that wraps the validated MRIQC engine behind a zero-installation web interface. WebMRIQC automates the DICOM-to-BIDS conversion of de-identified MRI scans, executes the unmodified containerized MRIQC pipeline on a shared compute node governed by a fair-share job queue, and returns an interactive in-browser dashboard. The dashboard grounds every IQM in published quality thresholds, benchmarks each scan against the normative distribution of high-resource open datasets, and supports cross-site multicentre implementation of optimized scan protocols in RCS.We describe the system architecture and a validation framework establishing measurement equivalence between WebMRIQC and native MRIQC across thirteen IQMs on the BraTS-Africa and BraTS 2021 datasets. Preliminary results indicate strong agreement for contrast-, signal and noise-based metrics, demonstrating that web-based implementation lowers the barrier to standardized MRI QC and provides a foundation for harmonized, regionally adapted quality benchmarks across RCS imaging sites. The code is publicly available here https://github.com/CAMERA-MRI/WebMRIqc.
Chinese Translation
可靠的磁共振成像（MRI）质量控制（QC）对于可靠的诊断神经成像至关重要，但标准的手动评估主观且耗时。MRIQC已建立标准化的图像质量指标（IQMs）自动提取，但其对本地计算成像技能和能力（包括高性能计算）的依赖，限制了其在资源受限环境（RCS）中的应用。我们提出了WebMRIQC（webmriqc.mailab.io），这是一个开源的基于浏览器的平台，封装了经过验证的MRIQC引擎，提供零安装的网络界面。WebMRIQC自动执行去标识化MRI扫描的DICOM到BIDS转换，在由公平共享作业队列管理的共享计算节点上运行未修改的容器化MRIQC管道，并返回一个交互式的浏览器内仪表板。该仪表板将每个IQM与已发布的质量阈值相结合，将每个扫描与高资源开放数据集的规范分布进行基准比较，并支持在RCS中优化扫描协议的跨站多中心实施。我们描述了系统架构和验证框架，建立了WebMRIQC与本地MRIQC在BraTS-Africa和BraTS 2021数据集上十三个IQMs之间的测量等效性。初步结果表明，基于对比度、信号和噪声的指标之间存在强一致性，证明基于网络的实现降低了标准化MRI QC的门槛，并为在RCS成像站点之间建立协调的、区域适应的质量基准奠定了基础。代码可在此处公开获取：https://github.com/CAMERA-MRI/WebMRIqc。
cs.CV / 128 / 2609.24026
InterHier: Learning Interconnected Hierarchical Semantics for Open-Vocabulary Object Detection
InterHier：学习开放词汇物体检测的互联层次语义
Abstract
In this paper, we investigate the limitations of fixed, hand-crafted connectors in hierarchical semantic representations for open-vocabulary object detection. Existing methods establish semantic relationships between base categories and unseen novel categories by placing a fixed connector between adjacent super-/sub-categories. However, such fixed connectors may not optimally capture the relationships within a semantic hierarchy. To address this limitation, we propose interconnected hierarchical semantic representations (InterHier), which utilize a prepended learnable context to globally guide the interpretation of prompts containing hierarchical relationships. InterHier operates in two main stages. First, it constructs a hierarchy-aware prompt by integrating super-/sub-categories and prepending a learnable context. Second, it optimizes this learnable context to align visual region embeddings and textual embeddings. InterHier consistently improves performance over methods that rely on fixed connectors and can be seamlessly integrated into existing open-vocabulary object detection models. Experiments on open-vocabulary object detection benchmarks demonstrate that InterHier achieves competitive performance against state-of-the-art methods.
Chinese Translation
在本文中，我们探讨了在开放词汇物体检测中，固定的手工设计连接器在层次语义表示中的局限性。现有方法通过在相邻的超类别/子类别之间放置固定连接器，建立基础类别与未见新类别之间的语义关系。然而，这种固定连接器可能无法最佳地捕捉语义层次内的关系。为了解决这一局限性，我们提出了互联层次语义表示（InterHier），该方法利用预先学习的上下文来全球性地指导包含层次关系的提示的解释。InterHier 主要分为两个阶段。首先，通过整合超类别/子类别并预先添加一个可学习的上下文，构建一个关注层次的提示。其次，优化这个可学习的上下文，以对齐视觉区域嵌入和文本嵌入。InterHier 在性能上始终优于依赖固定连接器的方法，并且可以无缝集成到现有的开放词汇物体检测模型中。在开放词汇物体检测基准上的实验表明，InterHier 在与最先进的方法相比时，表现出竞争力的性能。
cs.CV / 129 / 2609.24031
Video-STLayout Pre-training
视频-STLayout 预训练
Abstract
In recent years, pre-training has become fundamental to learning effective video representations, enabling strong transfer to downstream tasks. A popular framework in pre-training involves aligning features of a video encoder with that of another modality, for example, language or audio. We introduce Video-STLayout pre-training, a novel strategy for obtaining rich video representations informed by spatio-temporal layout of object bounding boxes. Object layouts can easily be obtained by applying an off-the-shelf object detector on the video frames. Our method uses a contrastive loss to align video features with the layout features from a trained layout encoder. We show the effectiveness of our approach in the task of activity recognition in complex scenes.
Chinese Translation
近年来，预训练已成为学习有效视频表示的基础，能够强有力地转移到下游任务中。预训练中一个流行的框架涉及将视频编码器的特征与另一种模态（例如语言或音频）的特征对齐。我们提出了视频-STLayout 预训练，这是一种新颖的策略，通过对象边界框的时空布局获取丰富的视频表示。对象布局可以通过在视频帧上应用现成的目标检测器轻松获得。我们的方法使用对比损失将视频特征与经过训练的布局编码器的布局特征对齐。我们展示了我们的方法在复杂场景中的活动识别任务中的有效性。
cs.CV / 130 / 2609.24049
U-PEN Mamba: Progressive Expansion with Selective State-Space Modeling for Efficient Retinal Vessel Segmentation
U-PEN Mamba：结合选择性状态空间建模的渐进扩展用于高效视网膜血管分割
Abstract
Accurate retinal vessel segmentation is important for computer-aided ophthalmic analysis, yet thin vessels, low contrast, and severe foreground-background imbalance remain challenging for encoder-decoder networks. This paper presents U-PEN Mamba, a U-shaped retinal vessel segmentation architecture that couples progressive nonlinear feature expansion with selective state-space modeling. The proposed network enriches local vessel responses with progressive expansion, models long-range spatial dependencies through a Mamba Global Context (MGC) block with linear sequence complexity, and uses attention-based decoder fusion to recover fine vascular boundaries. We evaluate U-PEN Mamba on CHASE DB1 and DRIVE using a consistent patch-based preprocessing pipeline and compare it with convolutional, attention-based, transformer-based, and Mamba-based segmentation baselines. U-PEN Mamba obtains the best mean intersection over union among the compared methods, achieving 0.8394 on CHASE DB1 and 0.8221 on DRIVE, with Dice scores of 0.8187 and 0.8078, respectively, using 21.6M trainable parameters. Ablation studies show that the MGC block contributes the largest gain over the U-Net baseline, while projection dimension and state size provide practical accuracy-efficiency control. These results indicate that selective state-space modeling is a promising global-context mechanism for parameter-efficient retinal vessel segmentation. Code is available at: https://github.com/areyesan/UPEN_Mamba.
Chinese Translation
准确的视网膜血管分割对于计算机辅助眼科分析至关重要，但细小血管、低对比度以及严重的前景与背景不平衡仍然对编码-解码网络构成挑战。本文提出了U-PEN Mamba，一种U型视网膜血管分割架构，结合了渐进非线性特征扩展与选择性状态空间建模。所提出的网络通过渐进扩展丰富局部血管响应，通过具有线性序列复杂度的Mamba全球上下文（MGC）模块建模长距离空间依赖，并利用基于注意力的解码器融合来恢复细致的血管边界。我们在CHASE DB1和DRIVE上评估了U-PEN Mamba，使用一致的基于补丁的预处理管道，并与卷积、基于注意力、基于变换器和基于Mamba的分割基线进行比较。U-PEN Mamba在比较方法中获得了最佳的平均交并比，在CHASE DB1上达到0.8394，在DRIVE上达到0.8221，Dice分数分别为0.8187和0.8078，使用了2160万个可训练参数。消融研究表明，MGC模块对U-Net基线贡献了最大的增益，而投影维度和状态大小提供了实际的准确性-效率控制。这些结果表明，选择性状态空间建模是一种有前景的全球上下文机制，适用于参数高效的视网膜血管分割。代码可在以下链接获取：https://github.com/areyesan/UPEN_Mamba。
cs.CV / 131 / 2609.24058
All-in-One Multilingual Scene Text Recognition with Script-aware Mixture-of-Experts
一体化多语言场景文本识别：基于脚本感知的专家混合模型
Abstract
Multilingual scene text recognition (STR) remains challenging due to the scarcity of training data for most languages and the difficulty of serving diverse scripts within a single model. Existing solutions either deploy one recognizer per language, inflating cost and introducing error accumulation, or rely on massive vision-language models (VLMs) that are expensive and still inaccurate on many scripts. In this work, we pursue an all-in-one multilingual recognizer that is simpler than per-language experts, lighter than VLMs, and more accurate than both. First, we construct TextMuSS-10M, a large-scale synthetic scene text dataset spanning 10 scripts and 229 languages. It provides balanced and sufficient supervision where real data is unavailable. Second, we propose ScriptMoE, a script-aware Mixture-of-Experts (MoE) architecture. It shares a single visual encoder and replaces the dense decoder with a sparse MoE block, which consists of an image-level router dispatches each image to the top-2 script-aligned experts and a shared expert absorbs cross-script knowledge. Extensive experiments on our assembled TextMuSS-Bench (10 scripts, 10,899 images) show that ScriptMoE achieves the highest accuracy of 82.06%, outperforming the strongest STR baseline by 1.31%. On the CC-OCR end-to-end multilingual task, replacing only the recognizer in PP-OCRv5 with ScriptMoE lifts F1 score from 65.71% to 80.89%, slightly surpassing the best VLM (80.73%) at a fraction of the parameter count.
Chinese Translation
多语言场景文本识别（STR）仍然面临挑战，主要由于大多数语言训练数据的稀缺以及在单一模型中处理多样化脚本的困难。现有解决方案要么为每种语言部署一个识别器，导致成本增加和错误累积，要么依赖于庞大的视觉-语言模型（VLMs），这些模型既昂贵又在许多脚本上仍然不准确。在本研究中，我们追求一种一体化的多语言识别器，它比每种语言的专家更简单，比VLMs更轻便，且在准确性上优于两者。首先，我们构建了TextMuSS-10M，这是一个涵盖10种脚本和229种语言的大规模合成场景文本数据集。该数据集在真实数据不可用的情况下提供了平衡且充足的监督。其次，我们提出了ScriptMoE，一种脚本感知的专家混合模型（MoE）架构。它共享一个单一的视觉编码器，并用稀疏的MoE模块替代了密集解码器，该模块由一个图像级路由器组成，将每个图像分派给前两个与脚本对齐的专家，而共享专家则吸收跨脚本知识。在我们组装的TextMuSS-Bench（10种脚本，10,899张图像）上的大量实验表明，ScriptMoE达到了82.06%的最高准确率，超越了最强的STR基线1.31%。在CC-OCR端到端多语言任务中，仅用ScriptMoE替换PP-OCRv5中的识别器，F1分数从65.71%提升至80.89%，略微超过最佳VLM（80.73%），且参数数量仅为其一小部分。
cs.CV / 132 / 2609.24064
Vision Transformers versus convolutional neural networks for fine-grained orchid genus identification in a species-rich, data-poor flora: a controlled benchmark on the Orchidaceae of New Guinea
视觉变换器与卷积神经网络在物种丰富、数据稀缺的植物群落中进行细粒度兰属识别的比较：新几内亚兰科的受控基准测试
Abstract
New Guinea is the world's richest island flora (~2,856 orchid species), yet most species are represented by only a handful of photographs, far fewer than direct species-level classification requires. Methods for fine-grained identification in such species-rich, data-poor floras are needed, and it remains unclear which backbone architecture and pretraining strategy best support them. We built a two-stage system that first predicts the genus of a query photograph, then retrieves visually similar reference images of candidate species using FAISS. We compared four pretrained backbones -- two Vision Transformers (ViTs; DINOv2, BioCLIP 2) and two CNNs (ConvNeXt V2-L, EfficientNetV2-L) -- fine-tuned under an identical protocol on a fixed, species-stratified partition of 16,701 photographs spanning 120 genera and 1,350 species, assessing accuracy, calibration, error structure, species retrieval, and open-set detection of novel genera. DINOv2 attained the best genus performance (macro top-1 66.9%, 95% CI 63.7-70.6; global top-1 88.9%); both ViTs outranked both CNNs, and general-purpose self-supervised pretraining (DINOv2) outperformed domain-matched biological pretraining (BioCLIP 2) by 7.1 points of macro top-1. Errors concentrated on two abundant genera acting as error attractors. DINOv2 embeddings achieved species Recall@5 of 86.6% and genus Recall@5 of 98.7%; temperature scaling reduced every backbone's Expected Calibration Error to about 0.03; and a distance-based open-set gate flagged unseen genera (mean AUROC 0.958). A self-supervised Vision-Transformer backbone combined with embedding retrieval is an effective, deployable strategy for fine-grained identification in species-rich, data-poor floras. The system is released as an open web application (the New Guinea Orchid Identifier), offering a practical template for other hyperdiverse, under-documented taxa.
Chinese Translation
新几内亚是世界上物种最丰富的岛屿植物群落（约2856种兰花），然而大多数物种仅由少量照片代表，远低于直接进行物种级分类所需的数量。对于这种物种丰富、数据稀缺的植物群落，迫切需要细粒度识别的方法，但尚不清楚哪种主干架构和预训练策略最能支持这些方法。我们构建了一个两阶段系统，首先预测查询照片的属，然后使用FAISS检索候选物种的视觉相似参考图像。我们比较了四种预训练主干网络——两种视觉变换器（ViTs；DINOv2，BioCLIP 2）和两种卷积神经网络（CNNs；ConvNeXt V2-L，EfficientNetV2-L），在一个固定的、按物种分层的16701张照片的分区上进行相同协议的微调，涵盖120个属和1350个物种，评估准确性、校准、错误结构、物种检索和新属的开放集检测。DINOv2在属的表现上达到了最佳（宏观 top-1 66.9%，95% CI 63.7-70.6；全局 top-1 88.9%）；两种ViTs的表现均优于两种CNNs，而通用自监督预训练（DINOv2）比领域匹配的生物预训练（BioCLIP 2）高出7.1个宏观 top-1点。错误主要集中在两个丰富的属上，作为错误吸引源。DINOv2嵌入实现了物种 Recall@5 为86.6%和属 Recall@5 为98.7%；温度缩放将每个主干的期望校准误差降低到约0.03；而基于距离的开放集门控标记了未见属（平均 AUROC 0.958）。结合嵌入检索的自监督视觉变换器主干是一个有效且可部署的策略，用于在物种丰富、数据稀缺的植物群落中进行细粒度识别。该系统作为开放网络应用程序（新几内亚兰花识别器）发布，为其他超多样性、文献不足的分类群提供了一个实用模板。
cs.CV / 133 / 2609.24071
Monitorable Chart Reasoning Agents via Verifiable Process Rewards
可监控的图表推理代理通过可验证的过程奖励
Abstract
Chart reasoning agents are increasingly used to extract actionable insights in critical domains, achieving state-of-the-art performance on multiple benchmarks. Yet, high benchmark accuracy alone is insufficient for deployment, where stakeholders must be able to audit and verify how a model reaches its answer. Existing LVLM-based chart agents produce either answer-only predictions or free-form rationales that are hard to verify, obscuring whether an error arose from misreading the chart, extracting a wrong value, or miscomputing. We propose Chart-RVR, a reinforcement learning framework for training monitorable chart agents with verifiable process rewards. Chart-RVR decomposes chart reasoning into three auditable blocks: Structure, identifying the chart type; Evidence, reconstructing the underlying data table in JSON; and Derivation, exposing the stepwise trace that computes the answer. Across six in-domain and out-of-domain benchmarks, Chart-RVR attains state-of-the-art accuracy among comparable-sized LVLMs. Beyond accuracy, we assess monitorability using a triangulated protocol that combines ground-truth surrogate metrics, an oracle information-gain measure, and an LLM-as-auditor scoring Process Verifiability and Evidence Localization, showing that Chart-RVR yields rationales that are markedly more verifiable and evidence-grounded than those from CoT prompting, SFT, and existing chart-specific baselines.
Chinese Translation
图表推理代理在关键领域中被越来越多地用于提取可操作的洞察，已在多个基准测试中实现了最先进的性能。然而，仅有高基准准确性不足以支持部署，因为利益相关者必须能够审计和验证模型如何得出其答案。现有的基于LVLM（大语言模型）的图表代理要么仅产生答案预测，要么生成难以验证的自由形式推理，这使得难以判断错误是由于误读图表、提取错误值还是计算错误。我们提出了Chart-RVR，这是一种用于训练可监控图表代理的强化学习框架，具有可验证的过程奖励。Chart-RVR将图表推理分解为三个可审计的模块：结构（Structure），识别图表类型；证据（Evidence），重建底层数据表为JSON格式；推导（Derivation），揭示计算答案的逐步过程。在六个领域内和领域外的基准测试中，Chart-RVR在同等规模的LVLM中达到了最先进的准确性。除了准确性，我们还使用三角测量协议评估可监控性，该协议结合了真实值替代指标、一个oracle信息增益度量，以及一个作为审计员的LLM评分过程可验证性和证据定位，显示Chart-RVR生成的推理明显比CoT提示、SFT和现有图表特定基准的推理更具可验证性和证据基础。
cs.CV / 134 / 2609.24075
A paired synthetic construction-site image dataset for robust computer vision under adverse conditions
用于恶劣条件下稳健计算机视觉的配对合成施工现场图像数据集
Abstract
Computer-vision systems used for construction monitoring can degrade under adverse environmental and visual conditions, yet such conditions remain underrepresented in existing construction image datasets. We present ConSynth-X, a paired synthetic construction-site image dataset containing 34,199 images derived from 3,109 real-world source scenes. The dataset comprises 11 condition-specific subsets spanning precipitation, fog, nighttime illumination, adverse weather at night, and small-object or long-distance views. Each synthetic image is linked to its corresponding source scene, enabling controlled comparison across environmental and visual conditions. ConSynth-X includes source-derived annotations, generation metadata, provenance information, and image-quality indicators, supporting object detection, image captioning, visual grounding, and visual question answering. Technical validation evaluates source-synthetic fidelity and alignment with real adverse-condition imagery using embedding-based similarity and distributional analyses. The dataset provides a structured resource for evaluating and improving the robustness of construction vision and vision-language models under challenging field conditions.
Chinese Translation
用于施工监测的计算机视觉系统在恶劣的环境和视觉条件下可能会降级，但现有的施工图像数据集中对此类条件的代表性不足。我们提出了 ConSynth-X，这是一个配对合成施工现场图像数据集，包含来自 3,109 个真实场景的 34,199 张图像。该数据集由 11 个特定条件的子集组成，涵盖降水、雾霾、夜间照明、夜间恶劣天气以及小物体或远距离视图。每张合成图像都与其对应的源场景相链接，使得在不同环境和视觉条件下进行受控比较成为可能。ConSynth-X 包括源场景派生的注释、生成元数据、来源信息和图像质量指标，支持目标检测、图像描述、视觉定位和视觉问答。技术验证通过基于嵌入的相似性和分布分析评估源合成的保真度及其与真实恶劣条件图像的对齐。该数据集为在具有挑战性的现场条件下评估和提高施工视觉及视觉-语言模型的稳健性提供了结构化资源。
cs.CV / 135 / 2609.24088
Bridging Reconstruction and Generation: A Latent Distribution Perspective on Evaluation and Improvement
重建与生成的桥梁：基于潜在分布的评估与改进视角
Abstract
In latent generative models, reconstruction quality is often assumed to correlate with generative performance. However, reconstruction FID (rFID) can exhibit weak or even negative correlation with generation FID (gFID). We attribute this discrepancy to a latent distribution mismatch: reconstruction evaluates the decoder on encoder-induced latents, whereas generation uses the same decoder on latents produced by the generative model. To characterize this shift, we introduce generation-aware reconstruction (GAR), which constructs a continuous trajectory from standard reconstruction toward generation by perturbing encoder latents with noise and denoising them through the generative model before decoding. GAR probes the decoder behavior along this trajectory, making the transition from encoder to generation-time latent distributions observable and diagnosable. The resulting trajectory-based diagnostic, GAR-FID, exhibits strong empirical correlation with gFID across diverse tokenizers and scales. Importantly, intermediate GAR latents become more generation-aware while preserving correspondence with their source images, thereby retaining paired supervision that is absent for fully generated latents. This correspondence enables decoder adaptation on intermediate GAR latents, consistently improving generative quality across model scales. Overall, latent distribution mismatch provides a useful perspective for evaluating and improving latent generative models.
Chinese Translation
在潜在生成模型中，重建质量通常被认为与生成性能相关。然而，重建FID（rFID）可能与生成FID（gFID）表现出较弱甚至负相关。我们将这种差异归因于潜在分布的不匹配：重建是在编码器引导的潜在变量上评估解码器，而生成则是在生成模型产生的潜在变量上使用相同的解码器。为了表征这种转变，我们引入了生成感知重建（GAR），它通过对编码器潜在变量施加噪声并通过生成模型去噪再解码，构建了从标准重建到生成的连续轨迹。GAR沿着这条轨迹探测解码器的行为，使得从编码器到生成时潜在分布的过渡变得可观察和可诊断。所得到的基于轨迹的诊断GAR-FID在不同的分词器和规模中与gFID表现出强烈的经验相关性。重要的是，中间的GAR潜在变量在保持与源图像的对应关系的同时，变得更加感知生成，从而保留了完全生成的潜在变量所缺失的配对监督。这种对应关系使得解码器能够在中间的GAR潜在变量上进行适应，持续提高不同模型规模下的生成质量。总体而言，潜在分布的不匹配为评估和改进潜在生成模型提供了有用的视角。
cs.CV / 136 / 2609.24095
HDND: Hierarchical Dynamic Neural Decoding for Multilingual Word/Character Retrieval from Non-Invasive Brain Recordings
HDND：用于从非侵入性脑记录中检索多语言单词/字符的层次动态神经解码
Abstract
While deep learning has enabled language decoding from intracranial brain recordings, extending this capability to non-invasive recordings remains an unresolved challenge. Decoding individual words from non-invasive brain recordings is particularly difficult, as word-level neural evidence is weak, temporally distributed, and entangled with acoustic, lexical, and semantic structure. Existing retrieval pipelines often collapse these factors into a single representation, potentially discarding information available at intermediate temporal scales. Here, we introduce Hierarchical Dynamic Neural Decoding (HDND), a hierarchical dynamic decoding framework that treats word decoding as structured refinement rather than flat label retrieval. HDND combines intermediate neural representations, contextual semantic predictions, and, for selected reading conditions, an auxiliary character-form objective. We evaluate HDND across seven electroencephalography (EEG) and magnetoencephalography (MEG) datasets spanning English, Dutch, Mandarin, and Cantonese listening, reading, and reading-aloud conditions. Across the nine-condition word-retrieval benchmark, the proposed HDND yields a higher participant-averaged balanced Top-10 point estimate than the matched contextual word-decoding baseline in every condition and achieves the highest mean among all compared methods in eight of nine conditions. Across the same nine matched conditions, HDND also yields higher token-micro and pooled word-macro Top-10 point estimates in every setting. Sentence retrieval favors HDND in eight of nine conditions, while auditory speech-segment retrieval is mixed across the six listening conditions. These results show that hierarchical residual refinement can improve multilingual word retrieval from heterogeneous non-invasive brain recordings.
Chinese Translation
尽管深度学习已经使得从颅内脑记录中进行语言解码成为可能，但将这一能力扩展到非侵入性记录仍然是一个未解决的挑战。从非侵入性脑记录中解码单个单词尤其困难，因为单词级别的神经证据较弱，时间上分散，并与声学、词汇和语义结构交织在一起。现有的检索管道通常将这些因素合并为单一表示，可能会丢弃在中间时间尺度上可用的信息。在此，我们引入了层次动态神经解码（Hierarchical Dynamic Neural Decoding，HDND），这是一种将单词解码视为结构化细化而非平面标签检索的层次动态解码框架。HDND结合了中间神经表示、上下文语义预测，并在特定的阅读条件下引入辅助字符形式目标。我们在七个涵盖英语、荷兰语、普通话和粤语的脑电图（EEG）和脑磁图（MEG）数据集上评估了HDND，涉及听、读和朗读条件。在九个条件的单词检索基准测试中，所提出的HDND在每个条件下都比匹配的上下文单词解码基线获得了更高的参与者平均平衡Top-10点估计，并在九个条件中的八个条件下达到了所有比较方法中最高的均值。在相同的九个匹配条件下，HDND在每种设置中也获得了更高的token-micro和汇总word-macro Top-10点估计。在九个条件中的八个条件下，句子检索更倾向于HDND，而在六个听觉条件下，听觉语音片段检索结果则较为混合。这些结果表明，层次残差细化可以改善来自异构非侵入性脑记录的多语言单词检索。
cs.CV / 137 / 2609.24098
A$^2$Safe: Counterfactual Evidence-Aligned Adaptive Agent Collaboration for Safe and Effective Visual Question Answering
A$^2$Safe：基于反事实证据对齐的自适应智能体协作框架，用于安全有效的视觉问答
Abstract
Visual Question Answering (VQA) with Multimodal Large Language Models (MLLMs) requires not only producing safe and effective responses, but also grounding safety decisions in the multimodal evidence that determines risk. Recent safety-alignment methods improve refusal behavior and contextual risk awareness, yet correct safety outcomes may still rely on superficial textual or visual correlations, particularly when risk emerges from interactions between individually benign image and question content. To address this issue, we propose A$^2$Safe, a counterfactual evidence-aligned adaptive agent collaboration framework for safe and effective VQA. A$^2$Safe organizes localized visual observations, textual intent, and cross-modal risk relations through a Grounded Safety Evidence Board, making the basis of safety decisions explicit. Counterfactual safety evidence alignment enforces invariance to safety-irrelevant changes while requiring appropriate safety-state and response-mode transitions when risk-critical evidence is minimally altered. The resulting evidence state further supports adaptive collaboration, enabling direct answering when grounded evidence is sufficient and invoking policy critique and response revision when evidence is risky, uncertain, or conflicting. Under complementary safety-critical and general VQA protocols, A$^2$Safe achieves a 95.72 SIUO safety score, reduces the benign refusal rate on MOSSBench to 14.67%, and maintains an average general VQA score of 78.34 with 27.8% token overhead. These results support counterfactual evidence-aligned adaptive collaboration for safe and effective multimodal question answering.
Chinese Translation
视觉问答（VQA）与多模态大型语言模型（MLLMs）不仅需要生成安全有效的响应，还需要将安全决策建立在决定风险的多模态证据之上。近期的安全对齐方法改善了拒绝行为和上下文风险意识，但正确的安全结果仍可能依赖于表面的文本或视觉关联，特别是在风险源于个别无害的图像和问题内容之间的交互时。为了解决这一问题，我们提出了 A$^2$Safe，一个基于反事实证据对齐的自适应智能体协作框架，用于安全有效的 VQA。A$^2$Safe 通过一个基础安全证据板组织局部视觉观察、文本意图和跨模态风险关系，使安全决策的基础变得明确。反事实安全证据对齐强制对安全无关的变化保持不变，同时在风险关键证据最小改变时要求适当的安全状态和响应模式转换。由此产生的证据状态进一步支持自适应协作，当基础证据足够时能够直接回答，并在证据存在风险、不确定或冲突时调用政策批评和响应修订。在互补的安全关键和一般 VQA 协议下，A$^2$Safe 达到了 95.72 的 SIUO 安全评分，将 MOSSBench 上的无害拒绝率降低至 14.67%，并保持了 78.34 的平均一般 VQA 评分，且令牌开销为 27.8%。这些结果支持基于反事实证据对齐的自适应协作，以实现安全有效的多模态问答。
cs.CV / 138 / 2609.24109
Adaptive Cortically Constrained EEG-Vision Alignment for Zero-Shot Brain-to-Image Retrieval
自适应皮层约束的EEG-视觉对齐用于零样本脑图像检索
Abstract
Zero-shot brain-to-image retrieval requires robust alignment between noisy EEG responses and visual representations. Existing EEG-vision alignment methods often operate in sensor space and apply fixed visual supervision to all responses, ignoring both spatial mixing in scalp EEG and response-wise variability in alignment reliability. We propose an adaptive cortically constrained EEG-vision alignment method for zero-shot brain-to-image retrieval. The method reconstructs EEG responses into predefined ROI-level source-pattern representations and encodes them with a Neuro-ROI Attention Encoder. To handle response-wise variability, we introduce an evidence-based adaptive visual supervision strategy that weights detail-controlled visual targets using model-based alignment evidence. On THINGS-EEG, the proposed method achieves strong 200-way zero-shot retrieval performance, with ROI-level attribution providing post hoc interpretability of the learned source-pattern representations. These results show that cortically constrained representation learning and adaptive supervision can jointly support EEG-vision alignment for zero-shot brain-to-image retrieval.
Chinese Translation
零样本脑图像检索需要在嘈杂的EEG反应与视觉表征之间进行稳健的对齐。现有的EEG-视觉对齐方法通常在传感器空间中操作，并对所有反应应用固定的视觉监督，忽略了头皮EEG中的空间混合和对齐可靠性中的反应间变异性。我们提出了一种自适应皮层约束的EEG-视觉对齐方法，用于零样本脑图像检索。该方法将EEG反应重构为预定义的感兴趣区域（ROI）级源模式表征，并使用神经ROI注意力编码器对其进行编码。为了处理反应间的变异性，我们引入了一种基于证据的自适应视觉监督策略，该策略使用基于模型的对齐证据对细节控制的视觉目标进行加权。在THINGS-EEG数据集上，所提出的方法实现了强大的200类零样本检索性能，ROI级归因提供了对学习到的源模式表征的事后可解释性。这些结果表明，皮层约束的表征学习和自适应监督可以共同支持EEG-视觉对齐，以实现零样本脑图像检索。
cs.CV / 139 / 2609.24116
Patch-to-Global: Random Patch Diffusion for Globally Consistent Megapixel Artifact Inpainting in Whole Slide Images
从局部到全局：用于全切片图像中全球一致的百万像素伪影修复的随机补丁扩散
Abstract
Although deep learning has advanced Whole Slide Image (WSI) Analysis, tissue artifacts like bubbles and folds often cause silent failures by concealing essential morphology. Current pathology image restoration methods are mostly restricted to small patches, struggling to maintain global structural coherence at a megapixel scale. We introduce RestorePath, a framework for globally consistent megapixel scale inpainting that reconstructs diagnostic structures in histological image to prevent incorrect high-confidence predictions and lower error rates. Our model utilizes a Latent Diffusion Model (LDM) conditioned on Pathology Foundation Model (PFM) embeddings, integrating Large Kernel Attention (LKA) to manage long-range dependencies during random patch diffusion. Enhanced by Distance-Weighted Interpolation (DWI) and an Adaptive Guidance Scale (AGS), RestorePath ensures structural consistency and fidelity by modulating information from surrounding patches. Evaluations across TCGA-BRCA, BACH, and Camelyon16 datasets for images ranging from 512 to 4608 pixels demonstrate state-of-the-art performance in maintaining histological consistency. RestorePath significantly improves downstream Computational Pathology (CP) tasks, outperforming both raw artifact images and the conventional Detect-and-Discard (D&D) approach. The code is available at https://github.com/PathfinderLab/RestorePath
Chinese Translation
尽管深度学习在全切片图像（WSI）分析方面取得了进展，但气泡和褶皱等组织伪影常常通过掩盖重要形态导致静默失败。目前的病理图像修复方法大多局限于小补丁，难以在百万像素尺度上保持全局结构一致性。我们提出了RestorePath，一个用于全球一致的百万像素尺度修复的框架，重建组织学图像中的诊断结构，以防止错误的高置信度预测并降低错误率。我们的模型利用条件于病理基础模型（PFM）嵌入的潜在扩散模型（LDM），结合大核注意力（LKA）在随机补丁扩散过程中管理长距离依赖。通过距离加权插值（DWI）和自适应引导比例（AGS）的增强，RestorePath通过调节来自周围补丁的信息，确保结构一致性和保真度。在TCGA-BRCA、BACH和Camelyon16数据集上对512到4608像素范围内的图像进行评估，展示了在保持组织学一致性方面的最先进性能。RestorePath显著改善了下游计算病理学（CP）任务，优于原始伪影图像和传统的检测与丢弃（D&D）方法。代码可在https://github.com/PathfinderLab/RestorePath获取。
cs.CV / 140 / 2609.24125
Positive Pair Geometry Matters: Optimal Transport for Contrastive Learning of Visual Representations
正样本几何结构的重要性：用于对比学习视觉表征的最优传输
Abstract
Contrastive self-supervised learning has achieved strong performance by learning representations from multiple augmented views of the same image. However, most existing methods construct positive pairs using independently sampled stochastic augmentations, which may alter semantic content and ignore the intrinsic geometry of the data distribution. In this work, we propose OTCLR, an optimal transport-aware framework for contrastive learning representations that generates geometry-consistent positive samples. Instead of directly contrasting two randomly augmented views, we construct intermediate views between the original image and its augmented variants through entropic optimal-transport displacement interpolation. These transport-interpolated samples serve as positive views that better preserve image structure while explicitly modeling spatial distributional geometry. To further promote smooth representation learning, we evaluate auxiliary Sinkhorn regularization terms that encourage transport-interpolated views to remain consistent with their endpoint images. The proposed method can be incorporated into standard contrastive learning pipelines without modifying the encoder architecture. Experiments on multiple benchmark datasets show that our approach improves representation quality and transfer learning performance compared with conventional augmentation-based contrastive learning baselines.
Chinese Translation
对比自监督学习通过从同一图像的多个增强视图中学习表征，取得了强大的性能。然而，现有大多数方法使用独立采样的随机增强构建正样本对，这可能会改变语义内容并忽视数据分布的内在几何结构。在本研究中，我们提出了OTCLR，一个关注最优传输的对比学习表征框架，生成几何一致的正样本。我们通过熵最优传输位移插值构建原始图像及其增强变体之间的中间视图，而不是直接对比两个随机增强的视图。这些传输插值样本作为正视图，更好地保留了图像结构，同时明确建模空间分布几何。为了进一步促进平滑的表征学习，我们评估了辅助的Sinkhorn正则化项，鼓励传输插值视图与其端点图像保持一致。所提出的方法可以无缝融入标准的对比学习流程中，而无需修改编码器架构。在多个基准数据集上的实验表明，与传统的基于增强的对比学习基线相比，我们的方法提高了表征质量和迁移学习性能。
cs.CV / 141 / 2609.24127
Action-Slot: Structured Action-Centric Representation Learning for Multi-Agent Atomic Activity Understanding
Action-Slot：面向多智能体原子活动理解的结构化动作中心表示学习
Abstract
Atomic activity understanding aims to recognize and localize structured traffic behaviors that jointly encode motion patterns and their grounding in road topology. Unlike conventional action recognition, atomic activities are multi-agent, multi-label, and topology-aware: multiple activities co-occur while many agents remain inactive. We introduce Action-Slot, a structured action-centric representation learning framework. Slot attention is widely used for object-centric decomposition, but its permutation-invariant design and object-level inductive bias are misaligned with atomic activity semantics. We reformulate slot learning as structured activity decomposition through three designs: (1) category-aligned action slots that anchor slots to predefined activity categories, (2) parallel spatio-temporal slot updating for holistic video-level reasoning, and (3) background and negative-slot regularization that enforces competition between foreground activities and irrelevant regions. Together these establish an activity-centric inductive bias that disentangles concurrent and asynchronous activities directly from raw video. Beyond recognition, the learned representations encode transferable spatio-temporal grounding signals. We further propose an attention-difference-based pseudo mask selection framework that suppresses false positives by measuring attention changes before and after candidate region removal, enabling weakly supervised localization without dense annotations. To support systematic evaluation, we introduce TACO, a balanced synthetic dataset with full atomic activity coverage and pixel-level annotations. Experiments on OATS, TACO, and annotated nuScenes show superior recognition, strong sim-to-real transfer, and state-of-the-art weakly supervised localization.
Chinese Translation
原子活动理解旨在识别和定位结构化的交通行为，这些行为共同编码运动模式及其在道路拓扑中的基础。与传统的动作识别不同，原子活动是多智能体、多标签且具有拓扑意识的：多个活动同时发生，而许多智能体保持不活动。我们提出了Action-Slot，一个结构化的动作中心表示学习框架。槽注意力广泛用于以对象为中心的分解，但其置换不变设计和对象级归纳偏差与原子活动语义不一致。我们通过三种设计将槽学习重新表述为结构化活动分解：（1）类别对齐的动作槽，将槽锚定到预定义的活动类别；（2）并行时空槽更新，用于整体视频级推理；（3）背景和负槽正则化，强制前景活动与无关区域之间的竞争。这些设计共同建立了一种活动中心的归纳偏差，能够直接从原始视频中解开并发和异步活动。除了识别，学习到的表示还编码可转移的时空基础信号。我们进一步提出了一种基于注意力差异的伪掩码选择框架，通过测量候选区域移除前后的注意力变化来抑制假阳性，从而实现无需密集注释的弱监督定位。为了支持系统评估，我们引入了TACO，一个具有完整原子活动覆盖和像素级注释的平衡合成数据集。在OATS、TACO和注释的nuScenes上的实验显示出优越的识别性能、强大的模拟到现实转移能力，以及最先进的弱监督定位效果。
cs.CV / 142 / 2609.24136
The Visual Target Matters: Learning across the Visual Hierarchy for Brain-to-Image Retrieval
视觉目标的重要性：跨视觉层次的脑图像检索学习
Abstract
Brain-to-image retrieval seeks to identify the visual stimulus that elicited a non-invasive neural response. Candidate images are typically represented by pretrained vision models, whose internal representations vary in abstraction across depth. Existing methods usually train the neural encoder to recover a fixed final-layer visual target. Under this formulation, the visual hierarchy is reduced to a single prescribed endpoint, preventing representations at other depths from directly shaping the visual target. This limitation motivates learning how information across visual depths should contribute to the retrieval target. To this end, we introduce NeuroGlyph, which learns a trial-independent visual target from multiple depths of a frozen visual backbone. NeuroGlyph decomposes the target into factor-specific subspaces. Each subspace learns an image-conditioned allocation over visual depth. The resulting subspaces are fused into a single embedding for retrieval. Across THINGS-EEG and THINGS-MEG, NeuroGlyph outperforms final-layer supervision in all controlled comparisons. It also surpasses the post hoc best fixed-layer oracle in three of four comparisons. Parameter-matched ablations support both factorized target construction and image-conditioned depth allocation. Under comparable 200-way retrieval protocols, NeuroGlyph achieves the strongest system-level performance in six of eight reported metrics. These results support learning retrieval targets across the visual hierarchy rather than prescribing one visual depth.
Chinese Translation
脑图像检索旨在识别引发非侵入性神经反应的视觉刺激。候选图像通常由预训练的视觉模型表示，其内部表示在深度上具有不同的抽象程度。现有方法通常训练神经编码器以恢复固定的最终层视觉目标。在这种表述下，视觉层次被简化为一个预设的终点，阻止了其他深度的表示直接影响视觉目标。这一限制促使我们探索如何让不同视觉深度的信息为检索目标贡献。为此，我们引入了NeuroGlyph，它从冻结的视觉骨干网络的多个深度中学习一个独立于试验的视觉目标。NeuroGlyph将目标分解为特定因子的子空间。每个子空间学习基于视觉深度的图像条件分配。最终的子空间被融合为一个单一的嵌入用于检索。在THINGS-EEG和THINGS-MEG数据集上，NeuroGlyph在所有受控比较中均优于最终层监督。在四个比较中的三个中，它也超越了事后最佳固定层的预言者。参数匹配的消融实验支持了因子化目标构建和图像条件深度分配。在可比的200类检索协议下，NeuroGlyph在八个报告指标中的六个中实现了最强的系统级性能。这些结果支持在视觉层次中学习检索目标，而不是预设一个视觉深度。
cs.CV / 143 / 2609.24151
STAR: Scene- and Task-Aware 4D Radar Preprocessing Towards End-to-End Cognitive Radar
STAR：面向端到端认知雷达的场景与任务感知4D雷达预处理
Abstract
Four-dimensional (4D) Radar has emerged as a key sensor for environmental perception, providing range, azimuth, elevation, and Doppler measurements while remaining robust to illumination changes and adverse weather conditions. However, conventional Radar preprocessing methods, such as constant false alarm rate (CFAR) detection, select measurements primarily based on signal-level criteria and may therefore discard information valuable for downstream perception during point cloud generation. In addition, existing 4D Radar perception pipelines typically optimize Radar data processing and downstream perception independently, preventing task objectives from directly guiding the preprocessing stage. To address these limitations, we propose a Scene- and Task-Aware Radar (STAR) Preprocessor together with an end-to-end training framework. The STAR Preprocessor incorporates scene context and downstream task objectives to generate task-relevant Radar points, enabling the Radar representation to be optimized directly for perception. On the K-Radar benchmark, the proposed method achieves 74.3 AP, outperforming the previous state of the art by 5.6 AP points. Furthermore, applying the task-relevant points generated by STAR to various existing 3D detectors improves detection performance in most evaluation settings and yields an overall positive average gain over point clouds produced by conventional preprocessing.
Chinese Translation
四维（4D）雷达已成为环境感知的关键传感器，能够提供距离、方位、高度和多普勒测量，并且在光照变化和恶劣天气条件下保持稳健。然而，传统的雷达预处理方法，如恒虚警率（CFAR）检测，主要基于信号级别标准选择测量，因此可能会丢弃在点云生成过程中对下游感知有价值的信息。此外，现有的4D雷达感知管道通常独立优化雷达数据处理和下游感知，阻碍了任务目标直接指导预处理阶段。为了解决这些局限性，我们提出了一种场景与任务感知雷达（STAR）预处理器，以及一个端到端的训练框架。STAR预处理器结合了场景上下文和下游任务目标，以生成与任务相关的雷达点，从而使雷达表示能够直接针对感知进行优化。在K-Radar基准测试中，所提方法达到了74.3的平均精度（AP），比之前的最先进技术提高了5.6个AP点。此外，将STAR生成的任务相关点应用于各种现有的3D检测器，在大多数评估设置中提高了检测性能，并在传统预处理生成的点云上获得了整体正向的平均增益。
cs.CV / 144 / 2609.24158
Relightable 3D Avatar Reconstruction with Semantic-Adaptive Motion-Illumination Responses
可重光照的3D头像重建与语义自适应运动-光照响应
Abstract
Reconstructing expressive and relightable 3D head avatars from monocular videos remains challenging in computer vision, as it requires accurate modeling of both non-rigid facial motion and illumination-dependent appearance. Existing Gaussian avatar methods commonly rely on globally coupled representations, in which Gaussian primitives share a unified motion or illumination response model. Such uniform modeling neglects the distinct motion patterns and material/reflectance properties of different facial semantic regions, thereby limiting fine-grained animation accuracy and reducing relighting plausibility. To address this limitation, we propose SAMIRA, a 3D Gaussian avatar framework for semantic-adaptive motion-illumination response modeling. For motion response modeling, the Semantic-Adaptive Motion Response module rasterizes current-to-reference mesh displacements into a topology-consistent UV space and leverages facial semantics to route displacement features through semantic-specific modulators, predicting localized Gaussian geometric residuals beyond coarse mesh binding. For illumination response modeling, the Semantic-Adaptive Illumination Response module learns compact diffuse and specular response factors for each facial region, allowing Gaussians in different regions to adapt their illumination responses to novel environment lighting. These response factors are incorporated into deferred physically based shading, providing a lightweight approximation of semantic-dependent illumination effects. Extensive experiments on self-reenactment, cross-reenactment, and relighting demonstrate that SAMIRA improves both fine-grained expression reconstruction and relighting realism over existing methods.
Chinese Translation
从单目视频中重建富有表现力且可重光照的3D头部头像在计算机视觉领域仍然具有挑战性，因为这需要准确建模非刚性面部运动和依赖光照的外观。现有的高斯头像方法通常依赖于全局耦合的表示，其中高斯原语共享统一的运动或光照响应模型。这种统一建模忽视了不同面部语义区域的独特运动模式和材料/反射属性，从而限制了细粒度动画的准确性并降低了重光照的合理性。为了解决这一限制，我们提出了SAMIRA，一个用于语义自适应运动-光照响应建模的3D高斯头像框架。在运动响应建模方面，语义自适应运动响应模块将当前到参考网格的位移光栅化到拓扑一致的UV空间，并利用面部语义通过特定语义的调制器引导位移特征，预测超出粗糙网格绑定的局部高斯几何残差。在光照响应建模方面，语义自适应光照响应模块为每个面部区域学习紧凑的漫反射和镜面反射响应因子，使不同区域的高斯能够将其光照响应适应于新环境光照。这些响应因子被纳入延迟物理基础阴影中，提供了一种轻量级的语义依赖光照效果近似。大量关于自我重演、交叉重演和重光照的实验表明，SAMIRA在细粒度表情重建和重光照真实感方面优于现有方法。
cs.CV / 145 / 2609.24170
An Unexpected Robot Policy: Early Evaluations of GPT-6 Astra on RoboDojo and Beyond
意外的机器人政策：对RoboDojo及其他领域的GPT-6 Astra的早期评估
Abstract
Embodied AI systems are often organized into System 1 and System 2. System 1 is typically a pretrained policy that generates actions at high frequency, whereas System 2 is often instantiated as a vision-enabled language model for high-level planning. We ask whether a large language model (LLM) can act as the policy for robot manipulation without task-specific finetuning. We call this setting LLM as policy. We evaluate three LLMs on all 42 RoboDojo tasks and compare their scores with 40 public policies. Astra and GPT-5.5 use the official 50-episode-per-task protocol; DeepSeek-Flash uses 10 episodes per task. GPT-6 Astra achieves 22.48% average success rate and 28.97 Score over 2,100 trials, ranking above every public entry. Yet GPT-5.5 and DeepSeek-Flash reach only 0.88% and 1.92% average success rate with the same post-processing. We find that Astra exhibits a sharply polarized capability profile. It generalizes well to tasks that require semantic understanding but not high-precision control. In contrast, it performs poorly on tasks that require precision, dynamic control, or complex bimanual coordination. In-context experiments show no aggregate benefit from one-shot demonstrations, while selected interaction traces show within-episode corrections under perturbations. Overall, the evaluated LLMs vary substantially in manipulation performance. Astra stands out and provides initial evidence for the potential of a general-purpose manipulation model, although reliable precision and dynamic control remain limitations in the evaluated setting.
Chinese Translation
具身人工智能系统通常被分为系统1和系统2。系统1通常是一个预训练的策略，以高频率生成动作，而系统2则通常被实例化为一个具备视觉能力的语言模型，用于高层次规划。我们探讨一个大型语言模型（LLM）是否可以在不进行任务特定微调的情况下，作为机器人操作的策略。我们将这种设置称为LLM作为策略。我们对三种LLM在42个RoboDojo任务上进行了评估，并将它们的得分与40个公共策略进行了比较。Astra和GPT-5.5使用官方的每个任务50集协议；DeepSeek-Flash则使用每个任务10集。GPT-6 Astra在2100次试验中实现了22.48%的平均成功率和28.97的得分，排名高于所有公共条目。然而，GPT-5.5和DeepSeek-Flash在相同后处理下仅达到0.88%和1.92%的平均成功率。我们发现Astra表现出明显两极化的能力特征。它在需要语义理解的任务上表现良好，但在需要高精度控制的任务上表现不佳。相反，它在需要精确、动态控制或复杂双手协调的任务上表现不佳。上下文实验显示单次演示没有整体收益，而选择的交互轨迹在扰动下显示出集内修正。总体而言，评估的LLM在操作性能上差异显著。Astra脱颖而出，并为通用操作模型的潜力提供了初步证据，尽管在评估设置中可靠的精度和动态控制仍然是限制因素。
cs.CV / 146 / 2609.24172
LegendBench: A Diagnostic Benchmark for Legend Understanding with Counterfactual Interventions
LegendBench：用于反事实干预的图例理解诊断基准
Abstract
Legends are fundamental to chart understanding, as reliable interpretation requires correctly binding legend entries to corresponding visual marks. While vision-language models (VLMs) are increasingly applied to chart understanding, their legend understanding is poorly diagnosed by aggregate accuracy, which can be satisfied by superficial shortcuts and confound legend-specific errors with other reasoning failures. To enable fine-grained diagnosis and controlled testing, we introduce LegendBench, a parametric benchmark and generation pipeline that produces targeted legend-centric test cases. LegendBench contributes (1) a capability-task taxonomy spanning legend parsing, legend grounding, legend-conditioned reasoning, and legend-aware abstention to localize failures, and (2) counterfactual group generation, where each base chart yields multiple variants under controlled legend interventions to probe model invariance and sensitivity. Using LegendBench, we evaluate both general-purpose VLMs and specialized chart models and generate their capability profiles, revealing persistent bottlenecks in reliable legend-to-mark binding and counterfactual consistency. We then use these capability profiles to guide targeted fine-tuning, demonstrating that bottleneck-specific interventions can effectively close the localized capability gaps and generalize to unseen data. We further leverage our counterfactual design to conduct fine-grained diagnostic experiments, analyzing encoding-channel effects, legend-order shortcuts, and abstention under varying visibility.
Chinese Translation
图例对于图表理解至关重要，因为可靠的解释需要正确地将图例条目与相应的视觉标记绑定在一起。尽管视觉语言模型（VLMs）在图表理解中越来越多地被应用，但其图例理解的诊断却仅依赖于整体准确性，这可能通过表面的捷径来满足，并将特定于图例的错误与其他推理失败混淆。为了实现细粒度的诊断和受控测试，我们引入了LegendBench，这是一个参数化基准和生成管道，能够生成针对性的图例中心测试用例。LegendBench的贡献包括（1）涵盖图例解析、图例定位、图例条件推理和图例感知弃权的能力任务分类法，以定位失败，以及（2）反事实组生成，其中每个基础图表在受控的图例干预下产生多个变体，以探测模型的不变性和敏感性。通过使用LegendBench，我们评估了通用VLM和专用图表模型，并生成它们的能力概况，揭示了在可靠的图例与标记绑定和反事实一致性方面的持续瓶颈。然后，我们利用这些能力概况指导针对性的微调，证明特定瓶颈的干预可以有效地缩小局部能力差距，并推广到未见数据。我们进一步利用我们的反事实设计进行细粒度的诊断实验，分析编码通道效应、图例顺序捷径以及在不同可见性下的弃权。
cs.CV / 147 / 2609.24190
Benchmarking Off-the-Shelf Multimodal AI Models Against Dermatologists on Patient-Captured Skin Images
将现成的多模态人工智能模型与皮肤科医生在患者拍摄的皮肤图像上的表现进行基准测试
Abstract
Artificial intelligence (AI) has advanced at a rapid pace in recent years. Initially, breakthroughs in large language models caught widespread attention. However, recent generations of frontier AI models have adopted multimodal capabilities as a first class citizen, with vision capabilities being central to that. In this paper, we evaluate three recently released models on the task of diagnosing dermatological conditions from patient-submitted images. The models chosen are at the low to mid tier in terms of pricing and thus represent a floor on current AI capabilities, not a ceiling. We evaluate AI performance relative to a panel of three certified dermatologists, who grade each image, and we present four interesting findings. Firstly, depending on the metric, the tested AI models are either on par or slightly trail humans in terms of inter-clinician agreement. Secondly, we find that asking AI models for a confidence rating produces poorly calibrated answers, meaning use of confidence thresholds should not be relied upon in a clinical setting. Thirdly, the effect of providing additional patient metadata is strongly model-specific, with one of the three models degrading on every metric considered. Finally, model cost is not predictive of performance. The best-performing model we tested costs on average $0.0045 per case.
Chinese Translation
近年来，人工智能（AI）迅速发展。最初，大型语言模型的突破引起了广泛关注。然而，最近一代前沿AI模型已经将多模态能力作为其核心特性，其中视觉能力尤为重要。在本文中，我们评估了三种最近发布的模型在从患者提交的图像中诊断皮肤病的任务上的表现。所选模型在价格上处于低到中等水平，因此代表了当前AI能力的下限，而非上限。我们将AI的表现与三位认证皮肤科医生的评估进行比较，他们对每张图像进行评分，并提出了四个有趣的发现。首先，根据不同的评估指标，测试的AI模型在临床医生之间的一致性方面与人类相当或略有落后。其次，我们发现要求AI模型提供置信度评分会产生不良的校准结果，这意味着在临床环境中不应依赖置信度阈值。第三，提供额外患者元数据的效果强烈依赖于模型，其中三种模型中的一种在所有考虑的指标上均表现下降。最后，模型的成本并不能预测其性能。我们测试的表现最佳的模型平均每个案例的成本为0.0045美元。
cs.CV / 148 / 2609.24193
Lightweight Pedestrian Head-Orientation Recognition Network for Safe Pedestrian-Vehicle Interaction
轻量级行人头部朝向识别网络用于安全的行人-车辆互动
Abstract
Pedestrian head orientation recognition plays an important role in autonomous driving by providing valuable cues for understanding pedestrian attention and anticipating potential crossing behavior. However, reliable recognition in real-world traffic scenes remains challenging because pedestrian head regions are often captured at low resolution. To address this challenge, we propose a lightweight Low-Resolution Head Orientation Convolutional Neural Network (LRHO-CNN) for pedestrian head orientation recognition. We construct a new dataset by extracting pedestrian head images from multiple public datasets and manually annotating them into eight orientation categories. The collected images are systematically preprocessed and augmented to increase data diversity and better represent variations in illumination and image quality. The experimental analysis compares LRHO-CNN with three fine-tuned baseline models, namely ResNet-18, ResNet-34, and VGG-16. The results demonstrate that LRHO-CNN achieves the highest classification accuracy among the evaluated models. LRHO-CNN is further evaluated on the JAAD and PIE datasets, demonstrating its effectiveness in recognizing pedestrian head orientation in real-world traffic scenes and providing informative head-orientation cues that can support downstream pedestrian behavior and intention prediction.
Chinese Translation
行人头部朝向识别在自动驾驶中发挥着重要作用，因为它提供了理解行人注意力和预测潜在过马路行为的宝贵线索。然而，由于行人头部区域通常在低分辨率下捕获，因此在真实交通场景中实现可靠的识别仍然具有挑战性。为了解决这一挑战，我们提出了一种轻量级的低分辨率头部朝向卷积神经网络（Low-Resolution Head Orientation Convolutional Neural Network, LRHO-CNN）用于行人头部朝向识别。我们通过从多个公共数据集中提取行人头部图像，并手动将其标注为八个朝向类别，构建了一个新的数据集。收集的图像经过系统预处理和增强，以增加数据多样性，更好地代表光照和图像质量的变化。实验分析将LRHO-CNN与三个微调的基线模型进行比较，即ResNet-18、ResNet-34和VGG-16。结果表明，LRHO-CNN在评估的模型中实现了最高的分类准确率。LRHO-CNN还在JAAD和PIE数据集上进行了进一步评估，证明其在真实交通场景中识别行人头部朝向的有效性，并提供了有助于下游行人行为和意图预测的信息性头部朝向线索。
cs.CV / 149 / 2609.24204
SAFe: Segment-guided Aggregation of Feature Densities for Anomaly-aware Segmentation
SAFe：基于分段引导的特征密度聚合用于异常感知分割
Abstract
Visual segmentation systems encounter objects outside their training distribution during real-world deployment, hindering reliable autonomous systems that depend on scene parsing in the perception stage. Many recent methods address this by using self-supervised foundation models to train density estimators that yield low likelihood in anomalous image regions. Although promising, these methods suffer from poor feature semantics or they lack spatial consistency, both of which undermine critical downstream decisions. We address this problem with~\method, a generative method based on class-conditional density estimation over self-supervised representations. SAFe trains lightweight normalizing flows that produce class-conditional normalized likelihood estimates over frozen DINOv3 features. We combine density estimates from transformer features with density scores over multi-scale convolutional features to capture both global semantics and local detail. We introduce a method-agnostic post-processing step based on SAM3 that connects per-location likelihoods into spatially coherent segments while suppressing false positives, and enables instance-level anomaly detection without retraining. The post processing further distinguishes novel categories among anomalous objects by a similarity-based agglomerative clustering scheme. SAFe sets a new state of the art on the PANIC, OoDIS, SMIYC ObstacleTrack with strong performance on the ISSU benchmark.
Chinese Translation
视觉分割系统在实际部署中会遇到超出其训练分布的对象，这阻碍了依赖场景解析的可靠自主系统的实现。许多近期的方法通过使用自监督基础模型来训练密度估计器，从而在异常图像区域产生低似然性，来解决这一问题。尽管这些方法前景可期，但它们往往存在特征语义不佳或缺乏空间一致性的问题，这两者都削弱了关键的下游决策。我们通过~ extit{SAFe}（Segment-guided Aggregation of Feature Densities）来解决这一问题，这是一种基于自监督表示的类条件密度估计的生成方法。SAFe训练轻量级的归一化流，生成在冻结的DINOv3特征上进行类条件归一化似然估计。我们将来自变换器特征的密度估计与多尺度卷积特征上的密度评分结合，以捕捉全局语义和局部细节。我们引入了一种基于SAM3的方法无关后处理步骤，将每个位置的似然性连接成空间一致的分段，同时抑制假阳性，并实现实例级的异常检测而无需重新训练。后处理进一步通过基于相似性的聚合聚类方案区分异常对象中的新类别。SAFe在PANIC、OoDIS和SMIYC ObstacleTrack上设定了新的最先进水平，并在ISSU基准测试中表现出色。
cs.CV / 150 / 2609.24208
CoaG: Cylinders on a Grid: Coarse 3D Layout Control for Video Generation
CoaG：网格上的圆柱体：用于视频生成的粗略三维布局控制
Abstract
We ask how little geometry a person has to draw to control both where people stand and where the camera moves in a generated video. Our answer is a ground plane and one cylinder per person. A user draws a grid on the ground, places one cylinder where each person should stand, moves the cylinders and the camera over 81 frames, and the model renders a photoreal video in which the people occupy the cylinders' positions, move as the cylinders move, and are seen from the drawn camera. Appearance comes from a text prompt and a background reference image; layout and motion come from the geometry. Because no dataset pairs such a signal with video, we build the pairs ourselves: an automatic engine writes 2000 captions from a combinatorial seed, generates a clip for each with a text-to-video model, and lifts every clip back to its geometry with person tracking, background inpainting, an agentic ground-mask loop, feed-forward multi-view reconstruction and a plane fit, with no real footage and no manual labels. A LoRA on Wan2.2-Fun-Control trained on 1935 such tuples follows drawn layouts and camera paths on hold-out clips: the generated people match the cylinders' count, order, position and height, the text changes who they are, the reference image changes where they are, and dolly-in, orbit, pan and crane paths are followed, dolly-out only weakly.
Chinese Translation
我们探讨一个人需要绘制多少几何图形才能控制生成视频中人们的站位和相机的移动。我们的答案是一个地面平面和每个人一个圆柱体。用户在地面上绘制一个网格，在每个人应站立的位置放置一个圆柱体，移动圆柱体和相机共81帧，模型渲染出一段照片级真实感的视频，其中人们占据圆柱体的位置，随着圆柱体的移动而移动，并从绘制的相机视角被看到。外观来自文本提示和背景参考图像；布局和运动来自几何图形。由于没有数据集将这种信号与视频配对，我们自己构建这些配对：一个自动引擎从组合种子生成2000个标题，为每个标题生成一个文本到视频模型的剪辑，并通过人物跟踪、背景修复、代理地面遮罩循环、前馈多视图重建和平面拟合将每个剪辑提升回其几何形状，整个过程没有真实镜头和手动标签。一个在1935个这样的元组上训练的LoRA模型（Wan2.2-Fun-Control）能够在保留剪辑上遵循绘制的布局和相机路径：生成的人物与圆柱体的数量、顺序、位置和高度相匹配，文本改变了他们的身份，参考图像改变了他们的位置，且跟随了推进、环绕、平移和起重路径，推进效果较弱。
cs.CV / 151 / 2609.24210
ChartJudgeBench: Evaluating LMM Judges for Chart-to-Code Generation
ChartJudgeBench：评估用于图表到代码生成的LMM评判者
Abstract
Building strong chart-to-code systems increasingly relies on reinforcement learning, whose effectiveness depends critically on the quality of the reward signal. Large Multimodal Models (LMMs) play a natural critical role in jointly assessing chart visual appearance and task requirements. They are therefore increasingly used as visual critics and reward models, yet their reliability as judges remains largely unexplored. To this end, we introduce ChartJudgeBench, a diagnostic vision-language benchmark for assessing LMM judges in chart-to-code workflows. It includes 1,003 Chart Perception Alignment (CPA) instances for pairwise chart comparison and 650 Chart Reasoning Judgment (CRJ) instances for binary Accept/Reject verification in Chart Reproduction and Chart Editing. Together, these tasks emulate the core judging decisions required in agentic refinement and RL-based chart optimization. Our evaluation of strong LMMs reveals four systematic limitations: (i) positional bias in pairwise comparison, (ii) a strong tendency to overpredict Accept, (iii) difficulty in matching visual styles and aesthetics, and (iv) an unexpected leniency bias in RL-trained models. These findings show that current LMM judges require explicit reliability validation before being used as critics or reward models in chart-to-code optimization. The code and data are available on ChartJudgeBench.
Chinese Translation
构建强大的图表到代码系统越来越依赖于强化学习，其有效性在很大程度上取决于奖励信号的质量。大型多模态模型（LMMs）在共同评估图表的视觉外观和任务需求方面发挥着至关重要的作用。因此，它们越来越多地被用作视觉评论者和奖励模型，但作为评判者的可靠性仍然在很大程度上未被探索。为此，我们引入了ChartJudgeBench，这是一个用于评估图表到代码工作流程中LMM评判者的诊断视觉-语言基准。它包括1,003个图表感知对齐（CPA）实例用于成对图表比较，以及650个图表推理判断（CRJ）实例用于图表再现和图表编辑中的二元接受/拒绝验证。这些任务共同模拟了在代理精炼和基于RL的图表优化中所需的核心评判决策。我们对强LMM的评估揭示了四个系统性局限性：（i）成对比较中的位置偏差，（ii）强烈倾向于过度预测接受，（iii）在匹配视觉风格和美学方面的困难，以及（iv）在RL训练模型中意外的宽容偏差。这些发现表明，当前的LMM评判者在被用作图表到代码优化中的评论者或奖励模型之前，需要明确的可靠性验证。代码和数据可在ChartJudgeBench上获得。
cs.CV / 152 / 2609.24215
Beyond Emotion Prompts: Fine-Grained Text-to-Image Generation Driven by Valence-Arousal-Dominance
超越情感提示：基于愉悦-唤醒-主导性细粒度文本到图像生成
Abstract
Although text-to-image models can accurately depict subjects and scenes, creators still struggle to specify the fine-grained emotions an image should convey without rewriting its content description. Natural language can suggest emotions, but it offers no control scale with stable meanings and ordered intensities. We propose EMOTRANS, which transforms psychologically grounded valence-arousal-dominance (VAD) coordinates into generation conditions that are independent of the content text and modulated across denoising stages, making emotional style a finely adjustable creative variable. To support this goal, we construct EMOVAD, an art-painting dataset that pairs objective content descriptions with separately collected emotional ratings from multiple annotators. We also coordinate emotional expression and content preservation through dual-branch training with a shared model. Objective and human evaluations show that the framework improves the accuracy of three-dimensional emotion control and produces perceptible, orderable continuous changes while maintaining competitive text alignment and image quality. This work provides a practical emotion-driven approach to image generation that extends objective content depiction to fine-grained emotional adjustment.
Chinese Translation
尽管文本到图像模型能够准确描绘主题和场景，但创作者仍然难以在不重写内容描述的情况下指定图像应传达的细粒度情感。自然语言可以暗示情感，但它并没有提供稳定意义和有序强度的控制尺度。我们提出了 EMOTRANS，它将心理学基础的愉悦-唤醒-主导性（Valence-Arousal-Dominance, VAD）坐标转化为与内容文本无关的生成条件，并在去噪阶段进行调节，使情感风格成为一个可细致调整的创作变量。为支持这一目标，我们构建了 EMOVAD，一个艺术绘画数据集，将客观内容描述与来自多个注释者的独立收集的情感评分配对。我们还通过共享模型的双分支训练协调情感表达和内容保留。客观和人类评估表明，该框架提高了三维情感控制的准确性，并在保持竞争性的文本对齐和图像质量的同时，产生可感知的、有序的连续变化。这项工作提供了一种实用的情感驱动图像生成方法，将客观内容描绘扩展到细粒度情感调整。
cs.CV / 153 / 2609.24220
Document Retrieval-Aware Chunking (D-RAC): Universal Retrieval-Aware Ingestion of Enterprise Documents via PDF Normalization and Multimodal Markdown Conversion
文档检索感知分块（D-RAC）：通过PDF标准化和多模态Markdown转换实现企业文档的通用检索感知摄取
Abstract
Retrieval-Augmented Generation (RAG) systems over enterprise knowledge bases must ingest heterogeneous document formats -- PDFs, Word documents, presentations, and scans -- whose content is locked inside complex visual layouts, multi-column pages, and dense tables. Rule-based extraction and OCR destroy reading order, flatten tables, and lose heading hierarchy, while fully agentic chunking over extracted text incurs high token costs and hallucination risk. We present Document Retrieval-Aware Chunking (D-RAC), an extension of our Web Retrieval-Aware Chunking (W-RAC) framework to arbitrary document formats. D-RAC first normalizes any input document into PDF, exploiting the fact that virtually every format has a faithful, deterministic PDF rendering. A single multimodal LLM pass then converts rendered pages into retrieval-optimized Markdown -- rewriting tables as self-contained prose statements and preserving heading hierarchy -- after which chunking proceeds exactly as in W-RAC: deterministic parsing into ID-addressable units followed by lightweight LLM-based chunk planning over identifiers rather than text. Source text is never regenerated during chunking, preserving W-RAC's cost, determinism, and observability benefits while unlocking every renderable format as a first-class input. On the 236-document, 795-page PDF subset of the RAG-Multi-Corpus benchmark spanning five enterprise domains, D-RAC converts and chunks the entire corpus in 72 minutes with zero errors, producing 1,748 retrieval-ready chunks. Compared to agentic chunking with frontier LLMs, D-RAC reduces chunking-stage output tokens by 95.7%, cutting chunking cost by 77.8% (GPT-4.1 pricing) to 85.6% (Gemini 2.5 Pro pricing) and chunking time by 75%. D-RAC scales linearly to documents of 500+ pages.
Chinese Translation
基于检索增强生成（RAG）系统的企业知识库必须摄取异构文档格式——PDF、Word文档、演示文稿和扫描件——其内容被锁定在复杂的视觉布局、多列页面和密集表格中。基于规则的提取和光学字符识别（OCR）破坏了阅读顺序，扁平化了表格，并丢失了标题层次结构，而对提取文本进行完全自主的分块则会产生高额的令牌成本和幻觉风险。我们提出了文档检索感知分块（D-RAC），这是我们Web检索感知分块（W-RAC）框架的扩展，适用于任意文档格式。D-RAC首先将任何输入文档标准化为PDF，利用几乎每种格式都有忠实且确定性的PDF渲染这一事实。然后，通过一次多模态大型语言模型（LLM）处理，将渲染的页面转换为检索优化的Markdown——将表格重写为自包含的散文陈述并保留标题层次结构——之后的分块过程与W-RAC完全相同：确定性解析为可通过ID寻址的单元，随后在标识符上而非文本上进行轻量级的LLM基础分块规划。在分块过程中，源文本从未被重新生成，从而保留了W-RAC的成本、确定性和可观察性优势，同时将每种可渲染格式解锁为一流输入。在涵盖五个企业领域的RAG-Multi-Corpus基准的236个文档、795页PDF子集上，D-RAC在72分钟内无错误地转换并分块整个语料库，生成1,748个可检索的分块。与前沿LLM的自主分块相比，D-RAC将分块阶段的输出令牌减少了95.7%，将分块成本降低了77.8%（GPT-4.1定价）至85.6%（Gemini 2.5 Pro定价），并将分块时间缩短了75%。D-RAC在500页以上的文档上线性扩展。
cs.CV / 154 / 2609.24223
DiaSeg: Diagonal Segment Extraction from DTW Paths for Interpretable Gait Analysis
DiaSeg：用于可解释步态分析的DTW路径对角线段提取
Abstract
Dynamic Time Warping (DTW) is the dominant approach for measuring similarity between time series, yet standard practice discards the optimal warping path after computing a single distance value, losing local alignment information most relevant to clinical diagnosis. We introduce DiaSeg, a framework that extracts diagonal segments from DTW paths with controlled breaks, characterizing each segment by five geometric features (effective length, interruption count, cost variation, temporal position, and path context), and enabling unsupervised pattern discovery without domain-specific feature engineering. Validated on 91 subjects across six clinical conditions (healthy aging, Parkinson's, Huntington's, ALS, brain tumor, and stroke), three findings emerge. First, diagonal segments form consistent unsupervised patterns (silhouette 0.33) aligned with biomechanical phase annotations, with label-based validation confirming near-perfect separation of healthy and pathological gait (ARI up to 0.986). Second, segments discriminate pathology at 69% (supervised) and 75% (patient-level clustering), with pathology manifesting through distributional shifts in segment length; combining segment and cycle-level features further improves classification to 91.7%. Third, while cycle-based methods achieve higher accuracy (91%), diagonal segments provide phase-specific interpretability unavailable in global representations, localizing where coordination breaks down within the gait cycle. DiaSeg thus transforms DTW from a black-box distance into a source of interpretable temporal features for neurodegenerative disease assessment.
Chinese Translation
动态时间规整（Dynamic Time Warping, DTW）是测量时间序列相似性的主要方法，但标准做法在计算出单一距离值后会丢弃最优的规整路径，从而失去与临床诊断最相关的局部对齐信息。我们提出了DiaSeg，一个从DTW路径中提取带有控制断点的对角线段的框架，通过五个几何特征（有效长度、中断次数、成本变化、时间位置和路径上下文）来表征每个段，并能够在没有领域特定特征工程的情况下实现无监督模式发现。在91名受试者的六种临床状态（健康老龄化、帕金森病、亨廷顿病、肌萎缩侧索硬化症、脑肿瘤和中风）上进行了验证，得出了三项发现。首先，对角线段形成了一致的无监督模式（轮廓系数0.33），与生物力学阶段注释对齐，基于标签的验证确认健康与病理步态几乎完美分离（调整兰德指数最高可达0.986）。其次，段落在69%（监督）和75%（患者级聚类）上区分病理，病理通过段长度的分布变化表现出来；结合段和周期级特征进一步将分类准确率提高至91.7%。第三，尽管基于周期的方法实现了更高的准确率（91%），对角线段提供了在全局表示中无法获得的阶段特定可解释性，定位步态周期内协调失效的具体位置。因此，DiaSeg将DTW从一个黑箱距离转变为神经退行性疾病评估的可解释时间特征来源。
cs.CV / 155 / 2609.24226
SRPR-Net: Semantic and Relational Prompt Refinement for Automated SAM-based Instance Segmentation
SRPR-Net：用于自动化基于SAM的实例分割的语义与关系提示优化
Abstract
Instance segmentation is a fundamental computer vision task with diverse real-world applications. Recently, prompt-driven foundation models have shown promising generalization. However, automated prompting remains limited by insufficient semantic guidance and inter-instance modeling. To address this challenge, we propose a novel architecture, named Semantic Relational Prompt Refinement Network (SRPR-Net), for automated SAM-based instance segmentation. A sequential prompt refinement mechanism is introduced to enrich detector geometry with visual-language semantics and then incorporate same-image instance dependencies, enabling context-aware box adjustment before SAM segmentation. Experiments on multiple standard benchmarks demonstrate that SRPR-Net achieves consistent improvements in segmentation performance over existing state-of-the-art approaches. The code is publicly available at https://github.com/JeremyXSC/SRPR-Net.
Chinese Translation
实例分割是计算机视觉中的一项基础任务，具有多种现实世界的应用。近年来，基于提示的基础模型展现出了良好的泛化能力。然而，自动化提示仍然受到语义指导不足和实例间建模的限制。为了解决这一挑战，我们提出了一种新颖的架构，称为语义关系提示优化网络（Semantic Relational Prompt Refinement Network，SRPR-Net），用于自动化基于SAM的实例分割。我们引入了一种顺序提示优化机制，通过视觉-语言语义丰富检测器几何形状，并结合同图像实例之间的依赖关系，使得在SAM分割之前能够进行上下文感知的框调整。在多个标准基准测试上的实验表明，SRPR-Net在分割性能上相较于现有的最先进方法取得了一致的提升。代码已公开发布在 https://github.com/JeremyXSC/SRPR-Net。
cs.CV / 156 / 2609.24228
IMPLICIT-Bench: Measuring Implicit Bias in Text-to-Image Models under Neutral Prompts
IMPLICIT-Bench：在中性提示下测量文本到图像模型中的隐性偏见
Abstract
Text-to-image (T2I) models are typically evaluated for bias using slot-based templates such as ``a photo of a [profession]''. Such templates probe only \emph{explicit} demographic attributes (e.g., gender, skin tone) in isolation. They overlook a broader \emph{implicit} bias that arises in natural prompts: when stereotype-relevant attributes are left unspecified, models still default to stereotypical outputs. We introduce IMPLICIT-Bench, a benchmark for measuring implicit bias in T2I models under such prompts. The key design is a structured-knowledge-graph (KG) construction of controlled prompt triplets: neutral, stereotype, and anti-stereotype variants that differ only along a single bias dimension while preserving scene semantics. This enables precise attribution of bias effects that template benchmarks cannot achieve. IMPLICIT-Bench comprises 5,493 prompts across 11 bias categories, validated through multi-model agreement, CLIP-based verification, and human evaluation. Using this benchmark, we show that state-of-the-art T2I models exhibit systematic bias under neutral prompts, a failure mode largely invisible to existing evaluations. We then use IMPLICIT-Bench to evaluate debiasing methods, uncovering a fundamental trade-off between bias reduction and semantic fidelity.
Chinese Translation
文本到图像（T2I）模型通常使用基于槽的模板（例如“一个[职业]的照片”）来评估偏见。这些模板仅探测 extit{显性}人口属性（例如性别、肤色）而不考虑其他因素。它们忽视了在自然提示中出现的更广泛的 extit{隐性}偏见：当与刻板印象相关的属性未被指定时，模型仍然默认生成刻板印象的输出。我们提出了IMPLICIT-Bench，这是一个用于在此类提示下测量T2I模型隐性偏见的基准。其关键设计是构建一个结构化知识图谱（KG），形成控制的提示三元组：中性、刻板印象和反刻板印象变体，这些变体仅在单一偏见维度上有所不同，同时保持场景语义。这使得能够精确归因于模板基准无法实现的偏见效应。IMPLICIT-Bench包含5,493个提示，覆盖11个偏见类别，通过多模型一致性、基于CLIP的验证和人工评估进行验证。使用该基准，我们展示了最先进的T2I模型在中性提示下表现出系统性偏见，这是一种现有评估中大多不可见的失效模式。随后，我们利用IMPLICIT-Bench评估去偏见方法，揭示了偏见减少与语义保真度之间的基本权衡。
cs.CV / 157 / 2609.24244
Look Where It Counts: A Free, Label-Free Visual Evidence Signal for Fine-Grained Vision-Language Reasoning
关注重要部分：一种免费的、无标签的视觉证据信号用于细粒度视觉-语言推理
Abstract
Multimodal large language models (MLLMs) fail at fine-grained visual questions less because they cannot reason than because they never see the evidence: high-resolution images are downsampled before encoding, so the model answers from linguistic priors. The standard remedies are expensive: annotated answers (SFT), hand-engineered verifiers (RLVR), or a large external teacher (on-policy distillation). We ask whether the visual evidence itself can supply the signal for free. We formalize the contrastive evidence gap, the per-token log-likelihood ratio that a model assigns to its own output when conditioned on a question-relevant region versus an irrelevant one, and study it across Qwen2.5-VL-7B, Qwen3-VL-8B, and Qwen3-VL-30B-A3B on V*Bench. Our main positive result is training-free: selecting the candidate crop under which the model's answer distribution is most peaked, using a single-view, label-free criterion, discovers the answer-bearing region with no bounding boxes, training, or labels. It localizes the target 4.4 to 5.1 times better than chance and raises fine-grained accuracy from 70 percent to 85 percent at inference. We further show that the gap is complementary to the model's own confidence. Combining them predicts correctness better than either alone, with AUC up to 0.99, and flags confidently wrong answers, with AUC ranging from 0.97 to 1.00 within the high-confidence subset. All effects concentrate on perception-bottleneck questions and vanish on a global-context control. Finally, we report an honest negative result: converting the same signal into a training method, gated self-distillation (SEG-Distill), does not outperform the base model at pilot scale across three gate designs, while more aggressive gating degrades accuracy. The signal is real, but converting it into training gains remains an open problem.
Chinese Translation
多模态大型语言模型（MLLMs）在细粒度视觉问题上的表现不佳，主要不是因为它们无法推理，而是因为它们从未看到证据：高分辨率图像在编码之前被降采样，因此模型的回答基于语言先验。标准的解决方案成本高昂：注释答案（SFT）、手工设计的验证器（RLVR）或大型外部教师（在线蒸馏）。我们探讨视觉证据本身是否可以免费提供信号。我们形式化了对比证据差距，即模型在条件为与问题相关的区域与无关区域时，对其自身输出分配的每个标记的对数似然比，并在 V*Bench 上研究 Qwen2.5-VL-7B、Qwen3-VL-8B 和 Qwen3-VL-30B-A3B。我们的主要积极结果是无训练的：在单视图、无标签标准下选择候选裁剪区域，使得模型的答案分布最为集中，能够发现承载答案的区域，而无需边界框、训练或标签。它的定位效果比随机猜测好 4.4 到 5.1 倍，并在推理时将细粒度准确率从 70% 提高到 85%。我们进一步表明，这一差距与模型自身的置信度是互补的。将二者结合能够比单独使用任何一个更好地预测正确性，AUC 高达 0.99，并且能够标记出置信度高但错误的答案，在高置信度子集中 AUC 范围为 0.97 到 1.00。所有效果集中在感知瓶颈问题上，并在全局上下文控制中消失。最后，我们报告了一个诚实的负面结果：将相同信号转化为训练方法的 gated self-distillation（SEG-Distill）在三种门控设计下的初步规模上并未超越基础模型，而更激进的门控则降低了准确性。信号是真实的，但将其转化为训练收益仍然是一个未解决的问题。
cs.CV / 158 / 2609.24276
Hierarchical Prompt Learning for Hyperbolic Vision-Language Models
超曲面视觉-语言模型的层次提示学习
Abstract
Hyperbolic vision-language models (VLMs) represent image and text features in a geometry naturally suited to hierarchy, but their adaptation to downstream tasks has largely relied on fixed prompts. Existing prompt learning methods, meanwhile, treat class labels as a flat set and do not exploit available taxonomic structure. We address this gap with a hierarchical prompt learning plug-in for frozen hyperbolic VLMs. Given a fixed offline parent-class hierarchy, it augments a class prompt learner with a separate parent prompt learner, parent-level supervision, hyperbolic entailment regularization, and parent-feedback logit fusion. We instantiate the method with CoOp, CoCoOp and MaPLe, yielding HyPLO, CoHyPLO and MaHyPLO. Across the standard 11-dataset benchmark, all variants improve base-to-new generalization and cross-dataset transfer, and remain comparable to their prompt learning baselines under domain shift. Six hierarchical metrics and embedding analyses show that the method produces more taxonomically consistent predictions and induces a hierarchy-consistent organization of parent, class, and image embeddings in hyperbolic space. Its gains are largest when novel classes must be placed within a fixed taxonomy, and smallest for fine-grained confusions among sibling classes or shifts affecting only the image distribution.
Chinese Translation
超曲面视觉-语言模型（VLMs）以自然适应层次结构的几何方式表示图像和文本特征，但它们在下游任务中的适应性主要依赖于固定的提示。现有的提示学习方法则将类别标签视为一个平坦的集合，并未利用可用的分类结构。我们通过为冻结的超曲面VLMs提供一个层次提示学习插件来填补这一空白。该方法在给定固定的离线父类层次结构的基础上，增强了类别提示学习器，增加了独立的父提示学习器、父级监督、超曲面蕴涵正则化和父反馈对数融合。我们将该方法实例化为CoOp、CoCoOp和MaPLe，分别得到HyPLO、CoHyPLO和MaHyPLO。在标准的11个数据集基准测试中，所有变体都提高了基础到新类的泛化能力和跨数据集的迁移能力，并在领域转移下与其提示学习基线保持了可比性。六个层次度量和嵌入分析表明，该方法产生了更具分类一致性的预测，并在超曲面空间中引导了父类、类别和图像嵌入的一致层次组织。当新类别必须在固定分类法中进行定位时，其增益最大；而在兄弟类别之间的细粒度混淆或仅影响图像分布的转变中，其增益则最小。
cs.CV / 159 / 2609.24287
Classifier-Free Guidance in Flow Matching: Non-Autonomous Potentials, Overshoot, and Posterior-Mean Control
无分类器引导在流匹配中的应用：非自主势能、超调和后验均值控制
Abstract
Classifier-free guidance (CFG) improves conditional generation in Flow Matching, but strong guidance can distort the generated distribution and reduce diversity. We provide a geometric account of this behavior by viewing Flow Matching as a time-varying gradient flow and characterizing how CFG reshapes its underlying potential. This view explains how stronger alignment can be accompanied by mean displacement and trajectory concentration, and motivates controlling guidance through the model-implied terminal posterior mean. We therefore propose Posterior-Mean-Capped CFG (PMC-CFG), a training-free, per-sample method that adaptively retains the strongest feasible guidance without additional network evaluations. Experiments on synthetic and large-scale image-generation benchmarks show that PMC-CFG limits guidance-induced distortion and concentration while improving the alignment--diversity trade-off, with particularly strong benefits when nominal guidance is large.
Chinese Translation
无分类器引导（CFG）改善了流匹配中的条件生成，但强引导可能会扭曲生成分布并降低多样性。我们通过将流匹配视为时间变化的梯度流，提供了对这种行为的几何解释，并表征了CFG如何重塑其基础势能。这一视角解释了更强的对齐如何伴随均值位移和轨迹集中，并激励通过模型隐含的终端后验均值来控制引导。因此，我们提出了后验均值限制的无分类器引导（PMC-CFG），这是一种无需训练的逐样本方法，能够自适应地保留最强的可行引导，而无需额外的网络评估。在合成和大规模图像生成基准上的实验表明，PMC-CFG限制了引导引起的扭曲和集中，同时改善了对齐与多样性之间的权衡，尤其在名义引导较大时表现出特别显著的优势。
cs.CV / 160 / 2609.24308
HappyWorld-Bench
快乐世界基准（HappyWorld-Bench）
Bai, Zhiqi, Cai, Junai, Chen, Yixin, Du, Jingrun, Feng, Tao, Gong, Wei, Huang, Siyuan, Lin, Xiao, Liu, Jiaheng, Luo, Jun, Lyu, Yongzhe, Ma, Liya, Meng, Zenan, Qu, Lin, Su, Wenbo, Wang, Jiaming, Wang, Qinghe, Wang, Shaofei, Wang, Yanghai, Wang, Zequn, Wang, Ziming, Wei, Hu, Wu, Jiangtao, Wu, Ruiqi, Xie, Jiaxin, Xu, Yuchi, Xu, Ze, Yu, Chengting, Yuan, Liangyu, Zeng, Gang, Zeng, Yawen, Zhang, Xingyao, Zhang, Zizheng, Zheng, Bo, Zhu, Jiancheng, Zhu, Song-Chun
Abstract
Evaluating world models requires assessing both the quality of the worlds they generate and their consistency and responsiveness under exploration, interaction, and modification. We introduce HappyWorld-Bench, a comprehensive benchmark that evaluates whether generated worlds remain reliable as agents interact with them. Our design is built on a hierarchical capability framework of six world capabilities (W1-W6), from generative construction to unified world modeling, instantiated across three independent evaluation tracks: video world models, spatial world models, and embodied world models. HappyWorld-Bench comprises 1,138 video prompts, 300 spatial scenes, and 254 embodied test cases. Across all three tracks, we build and operate HappyWorld-Arena to organize human A/B comparisons and derive model-level Elo ratings, which complement newly designed automated metrics that capture behavioral correctness. We evaluate 14 video world models, 9 spatial systems, and 8 embodied candidates under this unified framework. Results reveal remaining reliability gaps across all three tracks: video models exhibit reduced consistency during extended rollouts and revisits, spatial models achieve at best 70.14% placement accuracy and 73.33% edit execution, and embodied models struggle to preserve state across multi-step actions and respond precisely to altered action conditions and physical rules. These findings highlight the need to evaluate world models not only by visual quality, but also by state consistency and the correctness of their responses to actions and interventions.
Chinese Translation
评估世界模型需要考察它们生成的世界的质量，以及在探索、互动和修改过程中其一致性和响应能力。我们引入了快乐世界基准（HappyWorld-Bench），这是一个全面的基准测试，用于评估生成的世界在代理与之互动时是否保持可靠。我们的设计基于一个包含六种世界能力（W1-W6）的分层能力框架，从生成构建到统一世界建模，涵盖三个独立的评估轨道：视频世界模型、空间世界模型和具身世界模型。快乐世界基准包含1,138个视频提示、300个空间场景和254个具身测试案例。在所有三个轨道中，我们构建并运营快乐世界竞技场（HappyWorld-Arena），以组织人类的A/B比较并推导模型级别的Elo评分，这些评分补充了新设计的自动化指标，以捕捉行为的正确性。我们在这个统一框架下评估了14个视频世界模型、9个空间系统和8个具身候选模型。结果揭示了所有三个轨道中仍然存在的可靠性差距：视频模型在延长的回合和重访期间表现出一致性降低，空间模型的最佳放置准确率为70.14%，编辑执行率为73.33%，而具身模型在多步动作中难以保持状态，并且对改变的动作条件和物理规则的响应不够准确。这些发现强调了评估世界模型不仅需要考虑视觉质量，还需要关注状态一致性以及它们对动作和干预的响应的正确性。
cs.CV / 161 / 2609.24312
AnalogDepth: Multi-view Geometry from FPV drones under Analog Video Transmission
AnalogDepth：基于模拟视频传输的FPV无人机多视角几何
Abstract
Analog video transmission (VTX) remains widespread in FPV drones due to low latency, weight and low cost. However analog VTX suffers from complex spatially structured image degradation which differ fundamentally from digital image corruption (e.g. AWGN) used in standard training augmentation. This work shows that this type of noise severely degrades the accuracy of Depth Anything 3 (DA3), a state-of-the-art feed forward visual geometry foundation model. To address this gap, we present AnalogDepth, a parameter-efficient training pipeline that adapts DA3 to analog FPV imagery using student-teacher knowledge distillation with Low-Rank Adaptation (LoRA) injected into the DINOv2 backbone. Rather than synthesizing noise analytically, we build a noise bank from static FPV recordings under diverse conditions and compare real-noise injection against PSD-matched Gaussian synthesis and AWGN as baselines. Experiments on six real FPV flight sequences across three indoor scenes show that training with our noise bank consistently reduces per-frame depth RMSE and 3D reconstruction Chamfer distance compared to the pretrained DA3 baseline and both Gaussian noise variants. These results demonstrate that replicating the spatial structure of real analog transmission noise is critical for effective adaptation.
Chinese Translation
由于低延迟、轻量和低成本，模拟视频传输（VTX）在FPV无人机中仍然广泛使用。然而，模拟VTX遭受复杂的空间结构图像退化，这与标准训练增强中使用的数字图像损坏（例如，加性白噪声（AWGN））根本不同。本研究表明，这种类型的噪声严重降低了深度任意3（Depth Anything 3，DA3）这一先进的前馈视觉几何基础模型的准确性。为了解决这一问题，我们提出了AnalogDepth，一个参数高效的训练管道，利用学生-教师知识蒸馏将DA3适应于模拟FPV图像，并在DINOv2主干网络中注入低秩适应（Low-Rank Adaptation，LoRA）。我们不是通过分析合成噪声，而是从不同条件下的静态FPV录音中构建噪声库，并将真实噪声注入与PSD匹配的高斯合成和AWGN作为基线进行比较。在三个室内场景的六个真实FPV飞行序列上的实验表明，与预训练的DA3基线和两种高斯噪声变体相比，使用我们的噪声库进行训练始终能有效降低每帧深度均方根误差（RMSE）和三维重建的Chamfer距离。这些结果表明，复制真实模拟传输噪声的空间结构对于有效适应至关重要。
cs.CV / 162 / 2609.24313
NeuIDO: Neural Intrinsic Dynamics Operator for Physics-Informed 4D World Models
NeuIDO：用于物理信息4D世界模型的神经内在动态算子
Abstract
World models aim to capture environmental dynamics and predict future trajectories, showing growing potential for embodied intelligence. Physics-informed 4D generation integrates physical simulation to predict 3D object interactions, offering a promising pathway toward world models. However, this paradigm relies on manually imposed dynamical assumptions rather than internalizing world dynamics, and thus still leaves a gap toward a true world model. To bridge this gap, we propose NeuIDO, a novel world dynamics modeling framework that learns a unified intrinsic dynamics representation from visual observations, advancing physics-informed 4D generation toward a world model. Specifically, we formulate world modeling as a neural operator learning problem and introduce a two-stage training strategy to learn a generalizable mapping from the visual observation distribution to the intrinsic dynamics distribution. Building on this observation-dynamics mapping, NeuIDO enables zero-shot dynamics inference directly from videos and can be further aligned with complex real-world dynamics via few-shot adaptation. Extensive experiments demonstrate that NeuIDO effectively unifies the intrinsic dynamics underlying diverse visual observations into a shared representation and rapidly infers dynamics in novel scenes.
Chinese Translation
世界模型旨在捕捉环境动态并预测未来轨迹，展现出对具身智能的日益潜力。物理信息4D生成结合物理仿真以预测3D物体交互，为世界模型提供了一条有前景的路径。然而，这一范式依赖于手动施加的动态假设，而非内化世界动态，因此仍然存在向真正世界模型的差距。为弥补这一差距，我们提出了NeuIDO，一种新颖的世界动态建模框架，该框架从视觉观察中学习统一的内在动态表示，推动物理信息4D生成向世界模型发展。具体而言，我们将世界建模形式化为一个神经算子学习问题，并引入一种两阶段训练策略，以学习从视觉观察分布到内在动态分布的可泛化映射。基于这一观察-动态映射，NeuIDO能够直接从视频中进行零-shot动态推断，并可以通过少量样本适应进一步与复杂的现实世界动态对齐。大量实验表明，NeuIDO有效地将多样视觉观察背后的内在动态统一为共享表示，并快速推断新场景中的动态。
cs.CV / 163 / 2609.24330
AlignMorph: Tuning-Free Diffusion Image Morphing via Explicit Semantic Transport
AlignMorph：通过显式语义传输实现无调优扩散图像变形
Abstract
Image morphing aims to produce a smooth and semantically consistent transition between two input images. Existing diffusion-based morphing methods either require expensive per-pair optimization or rely on implicit spatial alignment, which easily fails under large layout discrepancies. To address these limitations, we propose AlignMorph, a novel tuning-free diffusion framework guided by the principle of transport-then-denoise. We explicitly decouple geometric alignment from generative denoising to avoid structural entanglement. Our framework consists of two core components. (1) Global Semantic Transport, which achieves diffusion-compatible semantic alignment via entropic optimal transport and reliability-aware latent warping; and (2) Coordinate-Aligned Generation, which uses a symmetric bi-phase attention handoff to maintain consistent spatial coordinates throughout denoising. Without any tuning, AlignMorph effectively eliminates ghosting and achieves superior structural coherence and temporal smoothness on morphing benchmarks. Code is available at https://github.com/51xOne/Alignmorph.
Chinese Translation
图像变形旨在在两幅输入图像之间产生平滑且语义一致的过渡。现有的基于扩散的变形方法要么需要昂贵的逐对优化，要么依赖于隐式空间对齐，这在布局差异较大时容易失败。为了解决这些局限性，我们提出了AlignMorph，一种新颖的无调优扩散框架，遵循先传输后去噪的原则。我们显式地将几何对齐与生成去噪解耦，以避免结构上的纠缠。我们的框架由两个核心组件组成：(1) 全局语义传输，通过熵最优传输和可靠性感知的潜在变形实现与扩散兼容的语义对齐；(2) 坐标对齐生成，使用对称双相注意力交接在去噪过程中保持一致的空间坐标。AlignMorph在没有任何调优的情况下有效消除了鬼影，并在变形基准测试中实现了卓越的结构一致性和时间平滑性。代码可在 https://github.com/51xOne/Alignmorph 获取。
cs.CV / 164 / 2609.24337
LiAuto-MindViT: A Hybrid Vision Backbone with Adaptive Bidirectional Mamba
LiAuto-MindViT：一种具有自适应双向Mamba的混合视觉骨干网络
Abstract
While Mamba-based models have shown strong potential for long sequence modeling, adapting them to vision is challenging due to the requirement of local neighborhood correlations and multi-directional spatial contexts for visual understanding. In this paper, we present LiAuto-MindViT, a novel hybrid vision backbone that synergizes the strengths of CNNs, Mamba, and Transformers. The core of our design is the Adaptive Bidirectional Mamba (ABM), which eliminates the directional bias of unidirectional SSMs through bidirectional selective scanning with learnable alpha blending, enabling content-adaptive directional fusion without the overhead of exhaustive multi-path routing. To further accelerate inference, we propose a deployment-friendly Reparameterized ConvSE (RepConvSE) module that leverages structural reparameterization to reduce latency and memory access overhead. Extensive experiments demonstrate that LiAuto-MindViT achieves state-of-the-art performance on image classification, object detection, and semantic segmentation while enabling efficient inference through reparameterization.
Chinese Translation
虽然基于Mamba的模型在长序列建模方面展现了强大的潜力，但由于视觉理解需要局部邻域相关性和多方向空间上下文，将其适应于视觉任务仍然具有挑战性。本文提出了LiAuto-MindViT，这是一种新颖的混合视觉骨干网络，融合了卷积神经网络（CNNs）、Mamba和变换器（Transformers）的优势。我们设计的核心是自适应双向Mamba（Adaptive Bidirectional Mamba, ABM），它通过可学习的α混合实现双向选择性扫描，消除了单向SSM的方向性偏差，从而实现内容自适应的方向融合，而无需耗费大量资源进行多路径路由。为了进一步加速推理，我们提出了一种适合部署的重参数化ConvSE（Reparameterized ConvSE, RepConvSE）模块，该模块利用结构重参数化来降低延迟和内存访问开销。大量实验表明，LiAuto-MindViT在图像分类、目标检测和语义分割方面达到了最先进的性能，同时通过重参数化实现了高效的推理。
cs.CV / 165 / 2609.24359
Dissecting Agentic Forensics: The Role of Triage, Prompting, and Evidence Arbitration in Open-World Fake Image Detection
剖析代理取证：在开放世界虚假图像检测中的分流、提示和证据仲裁的作用
Abstract
Image forensics is increasingly an open-world problem: manipulations range from fully synthetic images to localized edits, splicing and swapping, while most forensic detectors remain specialized to a single manipulation family. Agentic AI has recently emerged as a promising solution. In principle, such systems can assess the reliability of individual detectors, identify out-of-scope evidence, and arbitrate conflicting reports. However, it remains unclear which components actually drive performance and whether their benefits persist under distribution shift. To answer these questions, we study a training-free agentic framework built around specialist detectors, per-detector triage, and conflict-aware evidence arbitration. Using six configurations and three multimodal large language model backbones, we dissect the role of triage, prompting, and reasoning quality on both in-distribution and out-of-distribution data. Our results show that naive detector fusion suffers from severe false-positive rates on authentic images. Triage and prompting consistently improve performance by filtering unreliable evidence and exposing detector limitations. However, the dominant factor is represented by reasoning itself: A stronger judge substantially outperforms a weaker one, particularly under distribution shift. Most notably, manipulation recall is nearly saturated across all configurations, indicating that the main challenge of open-world image forensics is not detecting manipulations, but calibrating trust in specialized forensic tools and arbitrating conflicting evidence.
Chinese Translation
图像取证正日益成为一个开放世界问题：操控范围从完全合成的图像到局部编辑、拼接和交换，而大多数取证检测器仍然专注于单一的操控类型。代理人工智能（Agentic AI）最近作为一种有前景的解决方案出现。原则上，这类系统可以评估单个检测器的可靠性，识别超出范围的证据，并仲裁相互冲突的报告。然而，哪些组件实际上驱动性能仍然不清楚，并且它们的好处是否在分布转变下持续存在也尚不明了。为了解答这些问题，我们研究了一个围绕专业检测器、每个检测器的分流和冲突意识证据仲裁构建的无训练代理框架。使用六种配置和三种多模态大型语言模型骨干，我们剖析了分流、提示和推理质量在分布内和分布外数据中的作用。我们的结果表明，简单的检测器融合在真实图像上遭遇严重的误报率。分流和提示通过过滤不可靠的证据和揭示检测器的局限性，持续改善性能。然而，主导因素是推理本身：一个更强的判断者在分布转变下显著优于一个较弱的判断者。值得注意的是，所有配置中的操控召回率几乎饱和，这表明开放世界图像取证的主要挑战不是检测操控，而是校准对专业取证工具的信任和仲裁相互冲突的证据。
cs.CV / 166 / 2609.24367
TReViS: Temporal Repetition Structure Aware Video Synthesis for Self-supervised Repetitive Action Counting
TReViS：针对自监督重复动作计数的时间重复结构感知视频合成
Abstract
Fully supervised repetitive action counting (RAC) has achieved strong performance, but requires dense temporal annotations that are costly and difficult to scale. We propose TReViS, a self-supervised video synthesis framework that enables training RAC models without any repetition labels. TReViS estimates the underlying temporal repetition structure of an unlabeled video via a Temporal Self-Similarity Matrix, infers its cycle statistics, and synthesizes new training sequences that preserve realistic repetition patterns while introducing controlled temporal variability. These synthesized videos are paired with pseudo-labels and used to train existing RAC architectures from scratch. Across multiple datasets and backbones, TReViS consistently outperforms prior self-supervised methods and achieves performance competitive with several supervised baselines, while remaining fully label-free, demonstrating the effectiveness of structure-aware video synthesis for label-free RAC. The source code is available at https://github.com/yfqi/TReViS.
Chinese Translation
完全监督的重复动作计数（RAC）已取得良好表现，但需要密集的时间标注，这既昂贵又难以扩展。我们提出了TReViS，一个自监督视频合成框架，使得在没有任何重复标签的情况下训练RAC模型成为可能。TReViS通过时间自相似矩阵估计未标记视频的潜在时间重复结构，推断其周期统计，并合成新的训练序列，这些序列保留了真实的重复模式，同时引入了可控的时间变异性。这些合成的视频与伪标签配对，用于从头训练现有的RAC架构。在多个数据集和基础模型上，TReViS始终优于之前的自监督方法，并在多个监督基准上表现出竞争力，同时保持完全无标签，展示了结构感知视频合成在无标签RAC中的有效性。源代码可在 https://github.com/yfqi/TReViS 获取。
cs.CV / 167 / 2609.24379
Topographic Training Concentrates Causal Circuits Without Improving Neuron Monosemanticity
地形训练集中因果电路而不改善神经元单义性
Abstract
Mechanistic interpretability of vision transformers seeks to decompose model computation into human-readable units, but learned representations entangle many concepts in each neuron. Feature superposition is widely treated as the central obstacle to this decomposition, yet most mitigations (sparse autoencoders, dictionary learning) are post-hoc and leave the underlying network unchanged. We ask whether a spatial-locality training loss (TopoLoss) can act as a lightweight, training-time prior that improves interpretability of standard mech-interp tools. Training ViT on ImageNet-100 across multiple TopoLoss weights $\alpha$, we measure causal sufficiency of topographic clusters via activation patching and feature geometry via sparse autoencoders fit to the same residual stream. At $\alpha=1.0$, topographic clusters are 2.79$\times$ more causally sufficient than random unit sets of the same size, with the effect increasing monotonically in $\alpha$. SAE L0 sparsity decreases by 11% and dead-feature fraction rises 19-fold, yet standard neuron-level monosemanticity scores are unchanged, indicating that topographic pressure acts at circuit level, concentrating causal mass into spatially local structures without disentangling individual neurons. This dissociation suggests current neuron-level monosemanticity metrics are insensitive to a class of real interpretability gains, and positions cheap architectural priors as a viable training-time complement to post-hoc tooling.
Chinese Translation
视觉变换器的机制可解释性旨在将模型计算分解为人类可读的单元，但学习到的表示在每个神经元中纠缠了许多概念。特征叠加被广泛视为这种分解的主要障碍，然而大多数缓解措施（稀疏自编码器、字典学习）都是事后处理，并未改变底层网络。我们探讨一种空间局部性训练损失（TopoLoss）是否可以作为一种轻量级的训练时先验，从而改善标准机制可解释工具的可解释性。在多个TopoLoss权重$eta$下对ViT进行ImageNet-100训练，我们通过激活补丁和稀疏自编码器拟合相同残差流来测量地形聚类的因果充分性和特征几何。在$eta=1.0$时，地形聚类的因果充分性是相同大小随机单元集的2.79倍，且该效应在$eta$中单调增加。稀疏自编码器的L0稀疏度下降了11%，而死特征比例上升了19倍，然而标准的神经元级单义性评分保持不变，表明地形压力在电路级别上起作用，将因果质量集中到空间局部结构中，而未能解开单个神经元。这种解离表明当前的神经元级单义性指标对真实可解释性提升的一类变化不敏感，并将廉价的架构先验定位为对事后工具的有效训练时补充。
cs.CV / 168 / 2609.24384
A Lightweight Convolutional Neural Network for Real-Time Recognition of Hand-Drawn Geometric Shapes
一种轻量级卷积神经网络用于实时识别手绘几何图形
Abstract
Recognizing hand-drawn geometric shapes is a foundational sub-problem of sketch recognition, with applications in education, human-computer interaction, and diagram digitization. This paper presents the design, implementation, and evaluation of a desktop application that recognizes four basic hand-drawn geometric shapes, circle, square, rectangle, and triangle using a compact Convolutional Neural Network (CNN). A dataset of 2,000 labeled 28x28-pixel shape images was collected independently and released publicly. The classifier consists of three convolutional blocks (16, 32, and 64 filters) with max-pooling, an in-model data-augmentation stage (random horizontal flip, rotation, and zoom), a dropout-regularized dense layer of 128 units, and a 4-way linear output layer, totaling 97{,}956 trainable parameters. The network is trained with the Adam optimizer on a sparse categorical cross-entropy objective computed directly on logits. On an 80/20 train-validation split, the model achieves 94.80% training accuracy and 96.01% validation accuracy with a validation loss of 0.1437. A Tkinter-based graphical interface allows a user to draw a shape with the mouse and receive an immediate class prediction with a confidence score. We situate this system within the broader sketch and shape-recognition literature, compare its accuracy against related hand-drawn shape classification studies, and discuss the limitations inherent to a small, single-contributor dataset. The complete source code, trained model, and per-class datasets are released publicly to support reproducibility.
Chinese Translation
识别手绘几何图形是草图识别的基础子问题，广泛应用于教育、人机交互和图表数字化。本文提出了一种桌面应用程序的设计、实现和评估，该程序使用紧凑的卷积神经网络（CNN）识别四种基本的手绘几何图形：圆形、正方形、矩形和三角形。我们独立收集了一个包含2000个标记的28x28像素形状图像的数据集，并公开发布。分类器由三个卷积块（16、32和64个滤波器）、最大池化、模型内数据增强阶段（随机水平翻转、旋转和缩放）、一个128单元的dropout正则化全连接层以及一个4路线性输出层组成，总共包含97,956个可训练参数。该网络使用Adam优化器在直接计算logits的稀疏分类交叉熵目标上进行训练。在80/20的训练-验证分割中，模型实现了94.80%的训练准确率和96.01%的验证准确率，验证损失为0.1437。基于Tkinter的图形界面允许用户使用鼠标绘制形状，并立即获得带有置信度得分的类别预测。我们将该系统置于更广泛的草图和形状识别文献中，比较其准确性与相关的手绘形状分类研究，并讨论小型单一贡献者数据集固有的局限性。完整的源代码、训练模型和每类数据集已公开发布，以支持可重复性。
cs.CV / 169 / 2609.24403
Can Spiking Neural Networks play pinball? A neuromorphic motion detector for target tracking
脉冲神经网络能玩弹球吗？一种用于目标追踪的类神经形态运动检测器
Abstract
Biological visual systems achieve continuous, low-latency motion perception by processing sparse, asynchronous spiking signals, enabling real-time tracking under strict energy constraints. Event-based cameras, inspired by the mammalian retina, replicate this efficiency by capturing only local brightness changes as asynchronous events, offering a natural substrate for spiking neural networks (SNNs) to parallelise computation and adapt to fast-changing scenes. Pinball provides a controlled yet dynamic testbed, requiring precise motion estimation and fast reaction to a small, rapidly moving target. This work presents a fully spiking, real-time perception-to-action pipeline for closed-loop pinball gameplay. A dynamic vision sensor observes a small, fast-moving ball, and a network of spiking Time-Difference Encoders on the SpiNNaker neuromorphic platform jointly estimates its position, speed, and direction. The system is characterised across receptive field size, accumulation window, and angular tuning width for real-time operation, and benchmarked in closed loop against human players across two flipper regimes of increasing physical realism. It achieves a hit rate of 56.1%, nearly double the human average, reacting within 21.7 ms (5 ms network latency) and consuming an estimated 148 {\mu}W using fewer than 25k neurons, among the fastest and most energy-efficient event-based closed-loop demonstrators benchmarked. Under more realistic flipper dynamics, tuning a single interpretable policy parameter reproduces the full spectrum of human play styles, from cautious to aggressive, with no change to the perception pipeline. A physical demonstrator, tracking a real ball and actuating real flippers in closed loop, confirms the principle operates beyond simulation. Its fully spiking, learning-free design offers a compact, energy-efficient example of real-time neuromorphic perception-to-action.
Chinese Translation
生物视觉系统通过处理稀疏的、异步的脉冲信号实现连续、低延迟的运动感知，使其能够在严格的能量限制下进行实时追踪。受哺乳动物视网膜启发的事件驱动相机通过仅捕捉局部亮度变化作为异步事件，复制了这种高效性，为脉冲神经网络（SNNs）提供了一个自然的基础，以并行计算并适应快速变化的场景。弹球提供了一个受控但动态的测试平台，要求对小而快速移动的目标进行精确的运动估计和快速反应。本研究提出了一种完全基于脉冲的实时感知到行动的闭环弹球游戏管道。动态视觉传感器观察一个小而快速移动的球，位于SpiNNaker类神经形态平台上的脉冲时间差编码器网络共同估计其位置、速度和方向。系统在感受野大小、累积窗口和角度调谐宽度等方面进行了特征化，以实现实时操作，并在闭环中与人类玩家进行了基准测试，涵盖了两个逐渐增加物理现实感的翻转器模式。它实现了56.1%的击球率，几乎是人类平均水平的两倍，反应时间为21.7毫秒（5毫秒网络延迟），并在使用不到25,000个神经元的情况下消耗约148微瓦，是基准测试中最快和最节能的事件驱动闭环演示器之一。在更现实的翻转器动态下，调节一个可解释的策略参数可以再现从谨慎到激进的全谱人类游戏风格，而无需改变感知管道。一个物理演示器在闭环中追踪真实的球并驱动真实的翻转器，证实了该原理超越了模拟。其完全基于脉冲的、无学习设计提供了一个紧凑、节能的实时类神经形态感知到行动的示例。
cs.CV / 170 / 2609.24409
DeCo: Efficient Decouple-to-Couple Learning for Multi-Task Visual Grounding
DeCo：多任务视觉定位的高效解耦-耦合学习
Abstract
Multi-task visual grounding requires models to jointly understand linguistic semantics and perform accurate visual localization and segmentation. Despite the success of multimodal large language models, effectively adapting them to multiple grounding objectives remains challenging. Existing methods commonly enforce task cooperation through shared representations, while overlooking the intrinsic conflict between task-oriented feature interests. In this paper, we introduce $\textbf{DeCo}$, an efficient $\textbf{De}$couple-to-$\textbf{Co}$uple learning framework that resolves this dilemma through a two-stage paradigm: task-specific representation decoupling followed by complementary prior coupling. Specifically, we first propose Task-aware Semantic Decoupling (TSD) to route shared visual cues into individual features under salient word-level guidance, alleviating representation interference between localization and segmentation. Furthermore, we observe that segmentation naturally provides informative localization priors due to dense supervision. Based on this insight, we introduce Hybrid Prior Coupling (HPC), which integrates sentence-level semantic prior with mask-derived spatial prior for enhanced grounding. Built upon a frozen multimodal encoder, DeCo requires lightweight trainable parameters while achieving strong generalization across multiple grounding objectives. Extensive experiments on RefCOCO/+, G-Ref, ReferIt, Flickr, DIOR-RSVG, SARVG1.0, RRSIS-D, RIS-LAD, and RefDIOR demonstrate that DeCo achieves state-of-the-art performance on both natural and remote sensing benchmarks. The code and models are available at https://github.com/xiaoqiang-lu/DeCo.
Chinese Translation
多任务视觉定位要求模型共同理解语言语义，并执行准确的视觉定位和分割。尽管多模态大型语言模型取得了成功，但将其有效适应于多个定位目标仍然具有挑战性。现有方法通常通过共享表示来强制任务合作，而忽视了任务导向特征兴趣之间的内在冲突。本文提出了$ extbf{DeCo}$，一种高效的$ extbf{De}$couple-to-$ extbf{Co}$uple学习框架，通过两阶段范式解决这一困境：任务特定表示解耦，随后是互补先验耦合。具体而言，我们首先提出了任务感知语义解耦（Task-aware Semantic Decoupling, TSD），在显著的词级指导下将共享视觉线索引导到各个特征中，从而减轻定位与分割之间的表示干扰。此外，我们观察到，由于密集监督，分割自然提供了信息丰富的定位先验。基于这一见解，我们引入了混合先验耦合（Hybrid Prior Coupling, HPC），将句子级语义先验与基于掩码的空间先验结合，以增强定位效果。在一个冻结的多模态编码器基础上，DeCo需要轻量级的可训练参数，同时在多个定位目标上实现强大的泛化能力。在RefCOCO/+、G-Ref、ReferIt、Flickr、DIOR-RSVG、SARVG1.0、RRSIS-D、RIS-LAD和RefDIOR等数据集上进行的广泛实验表明，DeCo在自然和遥感基准测试中均实现了最先进的性能。代码和模型可在https://github.com/xiaoqiang-lu/DeCo获取。
cs.CV / 171 / 2609.24424
Estimating Accurate Hand Pose in Camera Space with Vision Transformer
基于视觉变换器的相机空间中准确手势姿态估计
Abstract
Monocular RGB-based hand pose estimation has emerged as a critical research frontier in computer vision. The local hand pose estimation methods predict hand poses relative to the wrist, while global hand pose estimation also requires estimating the wrist's position in the camera coordinate system. However, this camera-space estimation confronts two fundamental challenges: (1) depth ambiguity in monocular settings, and (2) the coupling effect of hand local poses and global wrist positions in the perspective projections. In particular, this coupling reflects that the projections are jointly determined by local hand poses, wrist positions, and camera intrinsics. To overcome these challenges, our framework proposes two key innovations: Transformation-Isomorphism Supervision for hand-depth information extraction and Perspective Information Embedding for resolving above coupling effect of local pose and wrist position, both integrated within the mainstream encoder-decoder architecture. Besides, we propose a novel framerate-aware multi-dataset training strategy for sequential pose refinement. Our fully integrated approach achieves at most 37.1\% superiority in CS-MJE over SOTA on HO3D. Project page: https://github.com/Mine268/CS-ViT.
Chinese Translation
基于单目RGB的手势姿态估计已成为计算机视觉领域的重要研究前沿。局部手势姿态估计方法相对于手腕预测手势姿态，而全局手势姿态估计还需要在相机坐标系中估计手腕的位置。然而，这种相机空间的估计面临两个基本挑战：（1）单目设置中的深度模糊性，以及（2）手部局部姿态与全局手腕位置在透视投影中的耦合效应。特别地，这种耦合反映了投影是由局部手势姿态、手腕位置和相机内参共同决定的。为了克服这些挑战，我们的框架提出了两个关键创新：用于手部深度信息提取的变换同构监督（Transformation-Isomorphism Supervision）和用于解决局部姿态与手腕位置耦合效应的透视信息嵌入（Perspective Information Embedding），两者均集成在主流的编码器-解码器架构中。此外，我们提出了一种新颖的帧率感知多数据集训练策略，用于序列姿态的细化。我们全面集成的方法在HO3D数据集上相较于最新技术（SOTA）在CS-MJE指标上取得了最高37.1%的优势。项目页面：https://github.com/Mine268/CS-ViT。
cs.CV / 172 / 2609.24452
Do LiDAR Language Models Really Understand Spatio-temporal Relationships?
LiDAR语言模型真的理解时空关系吗？
Abstract
Recent 4D LiDAR language models aim to reason about objects and their evolving spatial relationships. Yet, in our evaluation, always selecting the same option nearly matches the multiple-choice accuracy of two B4DL-derived configurations. We introduce LiDAR-Hallu, a geometry-referenced benchmark and diagnostic protocol with 10,000 questions across 150 nuScenes scenes. It covers object existence, ego-relative position, distance ordering, relative motion, and temporal localization, with explicit rules for selecting objects, comparing times, and determining reference answers. Our protocol combines fixed-answer and candidate-content controls, cross-scene pairs with identical prompts but opposite reference answers, and relation-specific recall. Analysis of 100,000 recorded responses reveals failures hidden by aggregate accuracy. Candidate duration alone makes temporal answers predictable without observing LiDAR. On paired questions, the models frequently give the same answer to scenes requiring opposite answers. Relation-specific analysis further shows that both configurations miss every positive lateral-motion case across all tested conditions. Temporal-shuffle contrastive decoding provides little net improvement, as repairs are largely offset by new errors and the main failures persist. These results show that evaluating spatio-temporal reasoning requires testing whether models distinguish the queried physical relationships, rather than relying on individual-answer accuracy alone. The source code, checkpoints, and data are released at https://github.com/Awesome4D/4DMLLM_Hallucination_Bench.
Chinese Translation
近期的4D LiDAR语言模型旨在推理对象及其不断变化的空间关系。然而，在我们的评估中，始终选择相同选项的准确率几乎与两个基于B4DL的配置的多项选择准确率相匹配。我们引入了LiDAR-Hallu，这是一个几何参考基准和诊断协议，包含150个nuScenes场景中的10,000个问题。该基准涵盖了对象存在、相对位置、距离排序、相对运动和时间定位，并设有明确的规则用于选择对象、比较时间和确定参考答案。我们的协议结合了固定答案和候选内容控制、具有相同提示但相反参考答案的跨场景对，以及特定关系的回忆分析。对100,000个记录响应的分析揭示了被聚合准确性掩盖的失败。仅候选持续时间就使得时间答案在不观察LiDAR的情况下变得可预测。在配对问题中，模型经常对需要相反答案的场景给出相同的答案。特定关系的分析进一步表明，这两个配置在所有测试条件下都错过了每一个正向侧向运动的案例。时间洗牌对比解码几乎没有带来净改善，因为修复大部分被新错误抵消，主要失败依然存在。这些结果表明，评估时空推理需要测试模型是否区分所查询的物理关系，而不仅仅依赖于单个答案的准确性。源代码、检查点和数据已发布在 https://github.com/Awesome4D/4DMLLM_Hallucination_Bench。
cs.CV / 173 / 2609.24455
MECAIL: Communication-Aware Incremental Learning for Object Detection with 14.6 KB Spatiotemporal Experts
MECAIL：面向通信的增量学习在对象检测中的应用，具有14.6 KB时空专家
Abstract
Intelligent transportation systems require Incremental Learning (IL) to continually improve their overall performance in dynamic environments. However, most edge devices lack the computational resources to support on-device IL, requiring updates to be transmitted from centralized servers. We propose using this setup to obtain dense, specialized module coverage that adapts a fixed base model to specific spatiotemporal contexts, such as parking lots, gas stations, ferries, or construction sites. However, in order to reliably transmit these modules to the edge device, using TCP, UDP, and BTP over V2X, Wi-Fi, and 2G-5G hardware, we establish a strict limit of 14.6 KB per module to fit within the first TCP window and to minimize UDP/BTP fragmentation. We further introduce Mixture-of-Experts for Communication-Aware Incremental Learning (MECAIL), the first method that meets this strict requirement, in which each new domain or environment is served by a small expert network that adapts the base model. We validate MECAIL on D-RICO and ODinW-13, where it largely matches the performance of parameter-heavy approaches while enabling practical, bandwidth-efficient large-scale deployment. This allows comprehensive coverage by experts for highly specific, focused, and temporary situations.
Chinese Translation
智能交通系统需要增量学习（Incremental Learning, IL）以在动态环境中不断提升整体性能。然而，大多数边缘设备缺乏支持设备内增量学习的计算资源，因此需要从集中式服务器传输更新。我们提出利用这种设置获取密集的、专业化的模块覆盖，以使固定的基础模型适应特定的时空上下文，例如停车场、加油站、渡轮或施工现场。然而，为了可靠地将这些模块传输到边缘设备，我们使用TCP、UDP和BTP通过V2X、Wi-Fi和2G-5G硬件建立了每个模块14.6 KB的严格限制，以适应第一个TCP窗口并最小化UDP/BTP的分片。我们进一步引入了面向通信的增量学习的专家混合模型（Mixture-of-Experts for Communication-Aware Incremental Learning, MECAIL），这是第一个满足这一严格要求的方法，其中每个新领域或环境由一个小型专家网络提供服务，以适应基础模型。我们在D-RICO和ODinW-13上验证了MECAIL，其性能在很大程度上与参数密集型方法相匹配，同时实现了实用的、带宽高效的大规模部署。这使得专家能够全面覆盖高度特定、集中且临时的情况。
cs.CV / 174 / 2609.24468
MIGA:Shared-Geometry Gaussian Representation with Implicit Amplitude Modeling for Accelerated 3D Multi-Echo MRI
MIGA：用于加速3D多回声MRI的共享几何高斯表示与隐式幅度建模
Abstract
Three-dimensional multi-echo MRI provides rich anatomical and quantitative information, but repeated volumetric encoding prolongs acquisition and motivates k-space undersampling. Reconstructing undersampled multi-echo data requires exploiting shared anatomy while preserving echo-dependent signal variation; full-volume modeling also introduces substantial computational and memory demands. We propose MIGA, a scan-specific framework comprising shared anisotropic Gaussian geometry, a coordinate-conditioned multi-output amplitude network, and explicit echo-specific phase variables. The Gaussian geometry provides common spatial support across echoes, the implicit network models spatially structured amplitude variations, and the phase variables retain echo-specific complex signal information. All components are jointly optimized using only the acquired multi-coil k-space, requiring no fully sampled training data. Experiments showed that MIGA consistently outperformed the comparison methods across imaging tasks and acceleration factors, with larger improvements under stronger undersampling. MIGA also achieved a favorable quality-cost balance among the evaluated full-volume multi-echo methods. These results support the effectiveness of combining shared Gaussian geometry with implicit echo-dependent amplitude modeling for accelerated 3D multi-echo MRI reconstruction.
Chinese Translation
三维多回声MRI提供丰富的解剖和定量信息，但重复的体积编码延长了采集时间，并促使了k空间的欠采样。重建欠采样的多回声数据需要利用共享解剖信息，同时保持与回声相关的信号变化；全体积建模也引入了大量的计算和内存需求。我们提出了MIGA，一个特定扫描的框架，包括共享的各向异性高斯几何、一个坐标条件的多输出幅度网络，以及显式的回声特定相位变量。高斯几何为回声提供了共同的空间支持，隐式网络建模了空间结构化的幅度变化，而相位变量则保留了回声特定的复杂信号信息。所有组件均使用仅采集的多线圈k空间进行联合优化，无需完全采样的训练数据。实验表明，MIGA在成像任务和加速因子方面始终优于比较方法，在更强的欠采样下改善幅度更大。MIGA在评估的全体积多回声方法中也实现了良好的质量-成本平衡。这些结果支持将共享高斯几何与隐式回声相关幅度建模相结合以加速3D多回声MRI重建的有效性。
cs.CV / 175 / 2609.24470
Spatial Action Review: A Visual Analytics Dashboard for Auditing Language-to-Action Hand-offs in Electron Microscopy
空间行动审查：用于审计电子显微镜中语言到行动交接的可视分析仪表板
Abstract
Multimodal large language models (MLLMs) are increasingly explored as interfaces for scientific image analysis, where a visual question-answering (VQA) response may be paired with a spatial output that guides a downstream stage. A supervisor reads the language answer, while a downstream workflow such as segmentation or region review consumes the point-set output. We call this transition from inspecting the answer to relying on its point action the language-to-action hand-off. A silent failure occurs when the answer is correct while the paired action misses annotated objects needed downstream, so answer-based oversight clears a region whose action is unreliable. We introduce Spatial Action Review, a visual analytics dashboard for auditing this failure mode in electron microscopy (EM) mitochondria analysis. It links paired answer-action records through an answer-action ledger, a task-by-dataset risk map, and an image-region audit view, connecting aggregate patterns to image evidence while an adjustable action-reliability gate supports re-audit. The review ends in a human-AI hand-off, where a supervisor records whether the action is accepted, escalated, held under a stricter gate, or flagged for model revision. Across 541 image regions from an EM-adapted Qwen3-VL case-study run, point actions fail the gate in 54.4% of records with a correct VQA response, and 27.4% of all records are silent failures. A correct answer is associated with only a 5.8-percentage-point higher probability of a reliable action, with a bootstrap interval spanning zero; the point-biserial correlation between answer correctness and object coverage is 0.061. This weak coupling persists across five model conditions on 753 matched image regions. Spatial Action Review makes answer-action mismatches visible and ties them to image evidence and a recorded decision before MLLM outputs enter autonomous scientific workflows.
Chinese Translation
多模态大型语言模型（MLLMs）越来越多地被探索作为科学图像分析的接口，其中可视问答（VQA）响应可能与指导下游阶段的空间输出配对。监督者读取语言答案，而下游工作流程（如分割或区域审查）则使用点集输出。我们将从检查答案到依赖其点行动的过渡称为语言到行动的交接。当答案正确但配对的行动遗漏了下游所需的注释对象时，会发生无声失败，因此基于答案的监督会清除一个行动不可靠的区域。我们介绍了空间行动审查，这是一个用于审计电子显微镜（EM）线粒体分析中这种失败模式的可视分析仪表板。它通过答案-行动账本、按数据集的任务风险图和图像区域审计视图连接配对的答案-行动记录，将聚合模式与图像证据相连接，同时可调的行动可靠性门支持重新审计。审查以人机交接结束，监督者记录该行动是否被接受、升级、在更严格的门下保持或标记为模型修订。在541个来自EM适配的Qwen3-VL案例研究运行的图像区域中，54.4%的记录在正确的VQA响应下行动未通过门，27.4%的所有记录是无声失败。正确答案仅与可靠行动的概率高出5.8个百分点，且自助区间跨越零；答案正确性与对象覆盖率之间的点二元相关性为0.061。这种弱耦合在753个匹配的图像区域的五种模型条件下持续存在。空间行动审查使答案-行动不匹配可见，并将其与图像证据和记录决策联系起来，确保MLLM输出在进入自主科学工作流程之前经过审查。
cs.CV / 176 / 2609.24482
STA-TFM: Spatio-Temporal Aggregation Across Views TransForMer for Pose Estimation
STA-TFM：用于姿态估计的跨视角时空聚合变换器
Abstract
Monocular 3D human pose estimation (HPE) remains challenging due to depth ambiguity, occlu- sions, and the need for temporal consistency. While multi-view methods provide superior accuracy over monocular approaches, they often require complex setups. We introduce STA-TFM, a transformer-based architecture that combines spatial and temporal information for multi-view pose estimation. The approach leverages DSTformer, a monocular feature extractor, to capture long-range pose dependencies within each view. A fusion transformer then aggregates information across views to produce coherent 3D estimates. To address training data scarcity, we use a data generation pipeline that transforms any existing 3D pose dataset into multi-view setups with controllable parameters. Experiments on various datasets demonstrate that STA-TFM outperforms existing camera-parameter-free multi-view methods. STA-TFM achieves 50.9% and 49.5% reductions in mean per joint position error (MPJPE) and mean per joint velocity error (MPJVE) on the DHP19 dataset. Furthermore, it achieves 6.7% and 7.7% respective reductions on HAA4D, and a 15.2% MPJPE reduction on TotalCapture. STA-TFM handles noisy and missing 2D inputs, supporting potential deployment in healthcare monitoring, athletic assessment, and immersive technologies. Code, training checkpoints, and data are available at https://zenodo.org/records/22832620.
Chinese Translation
单目3D人类姿态估计（HPE）由于深度模糊、遮挡以及对时间一致性的需求而仍然具有挑战性。虽然多视角方法在精度上优于单目方法，但它们通常需要复杂的设置。我们提出了STA-TFM，这是一种基于变换器的架构，结合了空间和时间信息以进行多视角姿态估计。该方法利用DSTformer，一个单目特征提取器，捕捉每个视角内的长距离姿态依赖关系。然后，融合变换器在视角之间聚合信息，以生成一致的3D估计。为了解决训练数据稀缺的问题，我们使用数据生成管道，将任何现有的3D姿态数据集转换为具有可控参数的多视角设置。在各种数据集上的实验表明，STA-TFM优于现有的无相机参数的多视角方法。在DHP19数据集上，STA-TFM在每个关节位置误差（MPJPE）和每个关节速度误差（MPJVE）上分别减少了50.9%和49.5%。此外，在HAA4D上分别减少了6.7%和7.7%，在TotalCapture上减少了15.2%的MPJPE。STA-TFM能够处理噪声和缺失的2D输入，支持在医疗监测、运动评估和沉浸式技术中的潜在应用。代码、训练检查点和数据可在https://zenodo.org/records/22832620获取。
cs.CV / 177 / 2609.24485
VPRune: Efficient Training-free Pre-LLM Visual Token Pruning
VPRune：高效的无训练预先大语言模型视觉令牌剪枝
Abstract
Visual token pruning is a promising approach to reducing the inference cost of large vision-language models (LVLMs), yet aggressive token reduction often causes substantial performance degradation. We identify three key factors behind this degradation: text-guided selection bias, information loss from discarded tokens, and positional distortion caused by sequence compaction. Based on these observations, we propose \textbf{VPRune}, a training-free pre-LLM pruning framework consisting of visual-only diversity selection, similarity-guided token recycling, and position-preserving restoration. Experiments on FastVLM-1.5B across multiple vision-language benchmarks demonstrate that VPRune achieves a favorable accuracy--compression trade-off, with particularly pronounced advantages under aggressive compression. Furthermore, evaluations on edge-device show that VPRune effectively reduces end-to-end inference latency while maintaining superior task performance, demonstrating its practicality for resource-constrained LVLM deployment.
Chinese Translation
视觉令牌剪枝是一种有前景的方法，用于降低大型视觉-语言模型（LVLMs）的推理成本，但激进的令牌减少往往会导致显著的性能下降。我们识别出导致这种下降的三个关键因素：文本引导的选择偏差、被丢弃令牌的信息损失，以及序列压缩引起的位置扭曲。基于这些观察，我们提出了 extbf{VPRune}，一种无训练的预先LLM剪枝框架，包含仅基于视觉的多样性选择、相似性引导的令牌回收和位置保持恢复。在多个视觉-语言基准上的FastVLM-1.5B实验表明，VPRune实现了良好的准确性与压缩之间的权衡，尤其在激进压缩下表现出显著优势。此外，在边缘设备上的评估显示，VPRune有效降低了端到端的推理延迟，同时保持了卓越的任务性能，证明了其在资源受限的LVLM部署中的实用性。
cs.CV / 178 / 2609.24487
AgentSTAR: Agentic Shape Tracking and Reconstruction from Monocular Videos
AgentSTAR：基于单目视频的智能形状跟踪与重建
Abstract
In this work, we present a method for shape reconstruction and tracking from video via agentic analysis-by-synthesis. Unlike prior methods which first estimate dense pixel correspondences and then recover object motion from them, our method infers a structured 3D object model, including its geometry and kinematic structure, and uses this model to optimise object track estimates over time. In our optimisation loop, a Vision-Language Model (VLM) agent iteratively refines shape or generalised pose through a render-and-compare loop, combining coarse visual reasoning with numerical pose optimisation for precise state estimation. This structured formulation enables our method to track through large motion, articulation, and severe occlusion without relying on pixel-matching objectives. Quantitatively, on ARCTIC, our method substantially outperforms state-of-the-art 3D point-tracking baselines for articulated objects, and on HOT3D it outperforms all evaluated rigid-object tracking baselines.
Chinese Translation
在本研究中，我们提出了一种通过智能分析-合成进行视频中的形状重建和跟踪的方法。与之前的方法不同，后者首先估计密集的像素对应关系，然后从中恢复物体运动，我们的方法推断出一个结构化的三维物体模型，包括其几何形状和运动结构，并利用该模型优化物体的轨迹估计。在我们的优化循环中，一个视觉-语言模型（Vision-Language Model, VLM）代理通过渲染与比较循环迭代地精炼形状或广义姿态，结合粗略的视觉推理与数值姿态优化以实现精确的状态估计。这种结构化的公式化使我们的方法能够在大运动、关节运动和严重遮挡的情况下进行跟踪，而无需依赖于像素匹配目标。在定量评估中，在ARCTIC数据集上，我们的方法显著优于当前最先进的关节物体三维点跟踪基线，而在HOT3D数据集上，它超越了所有评估的刚性物体跟踪基线。
cs.CV / 179 / 2609.24492
Identity-Consistent Analysis of Long-Shot Windsurfing Video: A Domain-Specific Offline Tracking System
身份一致性的长镜头风帆冲浪视频分析：一个领域特定的离线跟踪系统
Abstract
Long-shot windsurfing video combines small targets, large camera pans, prolonged overlaps, and rapidly changing backgrounds. The desired output is not a generic MOT trace but a separate, stable rider-relative video for each surfer; one false identity merge can invalidate an otherwise useful result. We present an offline analysis system that detects surfers, forms conservative local tracklets, links them globally with camera-compensated motion and a foreground-masked sail-color descriptor, and uses two pose keypoints on the rig to drive a rider-relative virtual camera. The tracking stage is evaluated on 21 manually reconstructed development videos containing 41,004 retained observations. On this fixed-observation protocol, the production system achieves 0.957 pairwise precision, 0.918 recall, and 0.937 F1, compared with 0.792 F1 for OC-SORT and 0.828 for BoT-SORT. Compared with OC-SORT, it reduces fragmentation excess from 845 to 42, but nine of its 95 output tracks mix rider identities and these errors affect seven of the 21 videos.
Chinese Translation
长镜头风帆冲浪视频结合了小目标、大范围摄像机移动、长时间重叠和快速变化的背景。期望的输出不是一个通用的多目标跟踪（MOT）轨迹，而是为每位冲浪者提供一个单独且稳定的相对骑手视频；一次错误的身份合并可能会使原本有用的结果失效。我们提出了一个离线分析系统，该系统能够检测冲浪者，形成保守的局部轨迹，利用摄像机补偿运动和前景遮罩的帆颜色描述符进行全局链接，并使用设备上的两个姿态关键点来驱动相对骑手的虚拟摄像机。跟踪阶段在21个手动重建的开发视频上进行了评估，这些视频包含41,004个保留观察。在这个固定观察协议下，生产系统实现了0.957的成对精度、0.918的召回率和0.937的F1分数，而OC-SORT的F1分数为0.792，BoT-SORT为0.828。与OC-SORT相比，它将碎片化过剩从845减少到42，但其95个输出轨迹中有9个混淆了骑手身份，这些错误影响了21个视频中的7个。
cs.CV / 180 / 2609.24494
CMAMBADEPTH: Self-supervised Monocular Depth Estimation with Channel Mamba and Hybrid Attention
CMAMBADEPTH：基于通道Mamba和混合注意力的自监督单目深度估计
Abstract
Accurate monocular depth estimation serves as a core enabler for single camera scene understanding. However, existing self-supervised monocular depth estimation methods generally suffer from the bottleneck of inefficient cross-scale information interaction and difficulty in balancing local and global spatial modeling. In this paper, we propose CMambaDepth, a self-supervised framework that achieves efficient multi-scale feature fusion and fine-grained contextual modeling via channel-wise selective state propagation. Specifically, Bidirectional Channel Mamba (Bi-CMamba) aligns encoder features across scales and enables bidirectional information exchange among ordered scale groups. Unidirectional Channel Mamba (Uni-CMamba) progressively aggregates decoder features and retains fine-grained scale groups through a group selection mechanism for subsequent fusion. Furthermore, a Hybrid Attention Module (HAM) is introduced to combine large-kernel local context and Manhattan self-attention for complementary spatial modeling. Experimental results demonstrate that our method achieves highly competitive performance. Specifically, our model achieves an AbsRel of 0.094 and an RMSE of 4.156 on KITTI, and an AbsRel of 0.140 on DDAD. In the zero-shot cross-dataset generalization test on NYUv2, it attains an AbsRel of 0.232, outperforming the baseline RA-Depth by 7.2%.
Chinese Translation
准确的单目深度估计是单摄像头场景理解的核心推动力。然而，现有的自监督单目深度估计方法通常面临着跨尺度信息交互效率低下和局部与全局空间建模平衡困难的瓶颈。本文提出了CMambaDepth，一个自监督框架，通过通道选择性状态传播实现高效的多尺度特征融合和细粒度上下文建模。具体而言，双向通道Mamba（Bi-CMamba）在各尺度之间对齐编码器特征，并实现有序尺度组之间的双向信息交换。单向通道Mamba（Uni-CMamba）逐步聚合解码器特征，并通过组选择机制保留细粒度尺度组以便后续融合。此外，引入了混合注意力模块（HAM），结合大核局部上下文和曼哈顿自注意力进行互补空间建模。实验结果表明，我们的方法实现了高度竞争的性能。具体而言，我们的模型在KITTI数据集上达到了0.094的绝对相对误差（AbsRel）和4.156的均方根误差（RMSE），在DDAD数据集上达到了0.140的AbsRel。在NYUv2的零-shot跨数据集泛化测试中，达到了0.232的AbsRel，超越了基线RA-Depth 7.2%。
cs.CV / 181 / 2609.24510
0.5\%>100\%: Bidirectional Reciprocal Learning for Referring Image Segmentation
0.5">%100": 双向互惠学习用于指向图像分割
Abstract
Recent advances in vision foundation models (VFMs) have shown remarkable capabilities across diverse unimodal visual tasks. However, adapting VFMs to referring image segmentation (RIS) typically necessitates precise vision-language alignment via full fine-tuning, incurring substantial computational overhead and risking catastrophic forgetting. While existing parameter-efficient fine-tuning (PEFT) methods enable safe knowledge transfer with minimal training costs, they predominantly operate independently within individual modalities or focus exclusively on unidirectional guidance from language to vision, overlooking progressive cross-modal interaction and visual feedback for textual refinement. To address these limitations, we propose Bidirectional Reciprocal Learning (BRL), a novel adapter-based PEFT framework that facilitates hierarchical, bidirectional information flow within both token-mixing and channel-mixing layers of frozen foundation models. Specifically, BRL introduces two complementary lightweight modules. The Reciprocal Attention Adapter (RAA) performs cross-modal query-key exchanges at the token level, enabling visual and linguistic tokens to mutually attend to each other for fine-grained spatial grounding. The Reciprocal Gate Adapter (RGA) generates cross-modal gating signals at the channel level, allowing global semantic context from one modality to adaptively recalibrate channel activations of the other. Extensive experiments on RefCOCO, RefCOCO+, and RefCOCOg benchmarks demonstrate the superiority of BRL over prior RIS methods, achieving state-of-the-art performance while requiring less than 0.5% backbone parameter updates. Code and models will be released at https://github.com/xiaoqiang-lu/BRL.
Chinese Translation
近期视觉基础模型（VFMs）的进展在多种单模态视觉任务中展现出显著的能力。然而，将VFMs适应于指向图像分割（RIS）通常需要通过全面微调实现精确的视觉-语言对齐，这会带来巨大的计算开销并且存在灾难性遗忘的风险。虽然现有的参数高效微调（PEFT）方法能够以最低的训练成本安全地转移知识，但它们主要在各自的模态内独立操作，或仅专注于从语言到视觉的单向指导，忽视了渐进的跨模态交互和文本细化的视觉反馈。为了解决这些局限性，我们提出了双向互惠学习（BRL），这是一种新颖的基于适配器的PEFT框架，促进了在冻结基础模型的token混合和通道混合层内的层次化双向信息流。具体而言，BRL引入了两个互补的轻量级模块。互惠注意适配器（RAA）在token级别执行跨模态查询-键交换，使视觉和语言token能够相互关注，实现细粒度的空间定位。互惠门适配器（RGA）在通道级别生成跨模态门控信号，允许一个模态的全局语义上下文自适应地重新校准另一个模态的通道激活。在RefCOCO、RefCOCO+和RefCOCOg基准上的大量实验表明，BRL在指向图像分割方法中优于先前的方法，达到了最先进的性能，同时更新的主干参数少于0.5%。代码和模型将发布在https://github.com/xiaoqiang-lu/BRL。
cs.CV / 182 / 2609.24524
Preoperative Prediction of Microvascular Invasion in Hepatocellular Carcinoma by Integrating Multimodal Ultrasound and Clinical Data: A Multicenter Study
通过整合多模态超声和临床数据进行肝细胞癌微血管侵犯的术前预测：一项多中心研究
Abstract
Background: Microvascular invasion (MVI) predicts recurrence and survival in hepatocellular carcinoma (HCC) but requires postoperative histopathology for diagnosis. We developed and validated a model integrating multimodal ultrasound and clinical data for preoperative MVI prediction. Methods: This multicenter study included 489 patients with HCC from eight centers. All patients had B-mode ultrasound (BUS), color Doppler flow imaging (CDFI), dynamic contrast-enhanced ultrasound (DCE-US), and clinical information. Data from seven centers (n = 421) were used for model development with five-fold cross-validation; data from the remaining center (n = 68) formed an independent external validation cohort. The proposed multimodal information fusion network used modality-specific encoders, a hemodynamic temporal change module for bidirectional DCE-US perfusion changes, and a representation consistency learning module to align heterogeneous ultrasound representations before Transformer-based fusion. Results: In external validation, DCE-US achieved the highest single-modality area under the receiver operating characteristic curve (AUC; 0.8545+/-0.0198), versus clinical information (0.6715+/-0.0156), CDFI (0.6435+/-0.0344), and BUS (0.6087+/-0.0417). Pixel-difference sampling and the proposed temporal module outperformed alternative sampling and video representation methods. The full model achieved the best performance, with an AUC of 0.8953+/-0.0180, accuracy of 81.18%+/-2.83%, sensitivity of 86.40%+/-6.69%, and specificity of 78.14%+/-6.28. Conclusions: Integrating multimodal ultrasound and clinical information enabled promising preoperative MVI prediction in HCC. DCE-US was the main source of predictive information, while BUS, CDFI, and clinical information provided complementary value. The proposed framework may support preoperative risk stratification and individualized clinical decision-making.
Chinese Translation
背景：微血管侵犯（MVI）预测肝细胞癌（HCC）的复发和生存，但需要术后组织病理学进行诊断。我们开发并验证了一种整合多模态超声和临床数据的模型，用于术前MVI预测。方法：本多中心研究纳入了来自八个中心的489例HCC患者。所有患者均接受了B型超声（BUS）、彩色多普勒血流成像（CDFI）、动态对比增强超声（DCE-US）及临床信息。来自七个中心的数据（n = 421）用于模型开发，并采用五折交叉验证；剩余中心的数据（n = 68）构成独立的外部验证队列。所提出的多模态信息融合网络使用了特定模态的编码器、用于双向DCE-US灌注变化的血流动力学时间变化模块，以及对齐异构超声表示的一致性学习模块，随后进行基于Transformer的融合。结果：在外部验证中，DCE-US的单模态接收器操作特征曲线下面积（AUC）最高（0.8545±0.0198），而临床信息为0.6715±0.0156，CDFI为0.6435±0.0344，BUS为0.6087±0.0417。像素差异采样和所提出的时间模块优于其他采样和视频表示方法。完整模型的表现最佳，AUC为0.8953±0.0180，准确率为81.18%±2.83%，灵敏度为86.40%±6.69%，特异性为78.14%±6.28。结论：整合多模态超声和临床信息为HCC的术前MVI预测提供了良好的前景。DCE-US是主要的预测信息来源，而BUS、CDFI和临床信息则提供了互补的价值。所提出的框架可能支持术前风险分层和个性化临床决策。
cs.CV / 183 / 2609.24526
ME-VLM:A Unified VLM for Embodied Cognition and Agent Coordination
ME-VLM：一种统一的用于具身认知和智能体协调的视觉语言模型
Abstract
Physical AI requires models to ground visual and linguistic understanding in real-world environments while accounting for environmental constraints and execution feedback. We introduce MachEmbodied-VLM (ME-VLM), a unified vision-language model with two variants, 4B and 35B-A3B, that brings together embodied cognition and multimodal agent capabilities. Our work emphasizes physical perception and spatiotemporal reasoning, together with planning, interaction, and outcome assessment in both digital and physical environments. We construct training data spanning embodied and multimodal agent tasks, including execution observations and feedback to support outcome assessment and decision refinement. The training pipeline comprises embodied capability injection, separate reinforcement learning of embodied and multimodal-agent experts, and multi-teacher on-policy distillation that consolidates their complementary capabilities into a single model. Experiments show competitive performance on both embodied and agent benchmarks, as well as on autonomous-driving and embodied-navigation tasks. For edge deployment, visual token compression, W4A8 quantization, and hardware--software co-optimization enable on-device inference of the 4B variant on the M100, reducing prefill latency from 400 ms to 188 ms. Project Page: https://machembodied.com/ME-Brain/ME-VLM.html Code Repository: https://github.com/MachEmbodied/ME-VLM
Chinese Translation
物理人工智能要求模型在真实世界环境中将视觉和语言理解与环境约束和执行反馈相结合。我们介绍了MachEmbodied-VLM（ME-VLM），这是一种统一的视觉-语言模型，具有两个变体，4B和35B-A3B，旨在结合具身认知和多模态智能体能力。我们的工作强调物理感知和时空推理，以及在数字和物理环境中的规划、互动和结果评估。我们构建了涵盖具身和多模态智能体任务的训练数据，包括执行观察和反馈，以支持结果评估和决策优化。训练流程包括具身能力注入、具身和多模态智能体专家的独立强化学习，以及将其互补能力整合到单一模型中的多教师在线蒸馏。实验结果表明，在具身和智能体基准测试，以及自主驾驶和具身导航任务上表现出竞争力。为了边缘部署，视觉标记压缩、W4A8量化和硬件-软件协同优化使得4B变体能够在M100设备上进行推理，将预填充延迟从400毫秒减少到188毫秒。项目页面：https://machembodied.com/ME-Brain/ME-VLM.html 代码库：https://github.com/MachEmbodied/ME-VLM
cs.CV / 184 / 2609.24531
Dynamic Thermal Gaussians: Multimodal 4D Gaussian Splatting
动态热高斯：多模态4D高斯点云
Abstract
Thermography plays a vital role in military and broader thermal analysis applications. Recent progress in 3D thermal reconstruction has extended temperature analysis from 2D to 3D space, yet most existing works assume static temperature distributions, neglecting the temporal dynamics of heat transfer in real-world environments. To address this limitation, we propose the first dynamic RGB-Thermal reconstruction framework for complex scenes. Our method jointly models RGB appearance, thermal observations, and scene geometry as they change over time. Specifically, we introduce a multimodal dynamic scene representation that anchors both the color and thermal modalities to a shared geometric substrate, ensuring their consistency under spatiotemporal deformations. We further design multimodal embeddings to enhance the motion expressiveness for each modality, and propose a multimodal routing mechanism that retains a unified set of shared multimodal Gaussians as the geometric backbone while adaptively spawning modality-specific Gaussians to strengthen the representational capacity in detail-rich regions of each individual modality. In addition, we contribute a novel benchmark dataset featuring high-frequency temperature variations to facilitate the evaluation of 4D reconstruction. Extensive experiments demonstrate that our method achieves high-fidelity spatiotemporal reconstruction of both appearance and temperature. Our code and dataset are available at: https://github.com/LinLif1869/DTG.
Chinese Translation
热成像在军事和更广泛的热分析应用中发挥着至关重要的作用。最近在3D热重建方面的进展将温度分析从2D扩展到了3D空间，然而大多数现有研究假设温度分布是静态的，忽视了真实环境中热传递的时间动态性。为了解决这一局限性，我们提出了第一个用于复杂场景的动态RGB-热重建框架。我们的方法联合建模RGB外观、热观测和场景几何形状随时间变化的情况。具体而言，我们引入了一种多模态动态场景表示，将颜色和热模态锚定到共享的几何基底上，确保它们在时空变形下的一致性。我们进一步设计了多模态嵌入，以增强每种模态的运动表现力，并提出了一种多模态路由机制，保持一组统一的共享多模态高斯作为几何骨架，同时自适应地生成特定模态的高斯，以增强每个单独模态在细节丰富区域的表现能力。此外，我们贡献了一个新的基准数据集，具有高频温度变化，以促进4D重建的评估。大量实验表明，我们的方法实现了外观和温度的高保真时空重建。我们的代码和数据集可在以下网址获取：https://github.com/LinLif1869/DTG。
cs.CV / 185 / 2609.24537
MIRAGE: Full-Body Bystander Privacy for Smart Glasses with Consent-Based Restoration
MIRAGE：基于同意恢复的智能眼镜全身旁观者隐私保护
Abstract
Video recording on smart glasses exposes more than faces. Continuous capture reveals full-body biometric signatures, including gait, posture, and silhouette, that enable person re-identification (ReID) even after conventional face sanitization. We present MIRAGE, a three-tier architecture for privacy-preserving smart glasses that enforces full-body privacy, supports synthetic full-body replacement, and retains encrypted recovery material for consent-based restoration. We implement MIRAGE on a Raspberry Pi~5 (a CPU-only proxy for smart-glasses compute), companion phones, and a cloud generative backend. Compared to prior systems, MIRAGE achieves 0.948 AP and 0.976 AR while accurately detecting the complete visible body. Its bounding box masking reduces learned silhouette-based ReID to essentially random guessing, with 10.86% Rank-1 accuracy compared with an 11.12% measured chance level. Even against an adaptive adversary retrained on MIRAGE's sanitized pose signals, Rank-1 gait identification drops from 90.25% to 26.20%, removing 72.5% of the adversary's identification advantage.
Chinese Translation
智能眼镜的视频录制不仅暴露面部信息。持续捕捉会揭示全身生物特征，包括步态、姿势和轮廓，这些信息即使在传统面部去标识化后也能实现个体重新识别（ReID）。我们提出了MIRAGE，一种三层架构的隐私保护智能眼镜，强制执行全身隐私，支持合成全身替换，并保留加密恢复材料以实现基于同意的恢复。我们在树莓派5（作为智能眼镜计算的CPU-only代理）、配套手机和云生成后端上实现了MIRAGE。与之前的系统相比，MIRAGE实现了0.948的平均精度（AP）和0.976的平均召回率（AR），同时准确检测完整的可见身体。其边界框遮罩将基于轮廓的ReID学习降低到几乎随机猜测，Rank-1准确率为10.86%，而测得的机会水平为11.12%。即使面对在MIRAGE的去标识化姿势信号上重新训练的自适应对手，Rank-1步态识别率也从90.25%降至26.20%，消除了对手识别优势的72.5%。
cs.CV / 186 / 2609.24539
Incentive Noise and Structural Prior Infusion for Multi-modal Object Re-Identification
多模态物体重识别中的激励噪声与结构先验注入
Abstract
Multi-modal object Re-Identification (ReID) benefits from complementary information across heterogeneous imaging modalities. To further enrich semantic representation, text descriptions have recently been incorporated as an additional modality. However, recent vision-language approaches often treat text descriptions as clean, deterministic signals and overlook their inherent noise, including modality-mismatched phrases and semantically ambiguous expressions. Moreover, prevailing methods lack explicit mechanisms to reconcile fine-grained structural discrepancies between modalities, even after high-level semantic alignment. To address these challenges, we propose a novel framework centered on Positive-Incentive Noise ({\pi}-noise) and structured prompt modulation. First, the Semantic Cross-Modal Modulator harnesses task-aware {\pi}-noise, sampled from a distribution conditioned on both visual and text inputs, to perturb global tokens and enable semantics-guided cross-modal compensation. Second, the Structure-Aware Prompt Adapter injects learnable geometric priors via prompts to enhance spatial consistency. Third, the Context-Aware Sparse Fusion module distills structural context to guide adaptive fusion while shielding identity features from noisy local details. Experiments on three multi-modal ReID benchmarks demonstrate the effectiveness and robustness of our approach. The code is available at https://github.com/zw-absin/INSPI.
Chinese Translation
多模态物体重识别（ReID）利用异构成像模态之间的互补信息，以丰富语义表示。最近，文本描述作为一种额外模态被纳入，以进一步增强语义表示。然而，近期的视觉-语言方法往往将文本描述视为干净、确定的信号，忽视了其固有的噪声，包括模态不匹配的短语和语义模糊的表达。此外，现有方法缺乏明确的机制来调和模态之间的细粒度结构差异，即使在高层次的语义对齐之后。为了解决这些挑战，我们提出了一种新颖的框架，重点关注正激励噪声（{ ext{π}}-noise）和结构化提示调制。首先，语义跨模态调制器利用任务感知的{ ext{π}}-noise，该噪声从基于视觉和文本输入的分布中采样，以扰动全局标记并实现语义引导的跨模态补偿。其次，结构感知提示适配器通过提示注入可学习的几何先验，以增强空间一致性。第三，上下文感知稀疏融合模块提炼结构上下文，以指导自适应融合，同时保护身份特征免受噪声局部细节的干扰。在三个多模态ReID基准上的实验表明了我们方法的有效性和鲁棒性。代码可在 https://github.com/zw-absin/INSPI 获取。
cs.CV / 187 / 2609.24560
Evaluating Transformation Models for pCLE Mosaic Registration
评估 pCLE 马赛克配准的变换模型
Abstract
Confocal Laser Endomicroscopy (CLE) provides real-time, cellular-resolution optical biopsy but has a narrow field of view, which image mosaicing can extend to provide anatomical context. Because of line-by-line acquisition, probe motion, and probe-tissue interaction, frame alignment generally requires a non-linear transformation whose accuracy is difficult to quantify: flexible transformation models can fit intensity features and noise, so appearance-based metrics such as Normalized Cross-Correlation (NCC) can improve without a genuine gain in geometric accuracy. We therefore establish a dataset of 132 frame pairs across fourteen pCLE sequences from 4 patients with manually annotated landmark correspondences, so that Target Registration Error (TRE) can serve as a geometrically grounded complement to NCC. We assess the effect of progressively increasing the transformation model's degrees of freedom, from translation to Thin Plate Spline (TPS), and of six feature-matching backends spanning classical (Shi-Tomasi, Lucas-Kanade) and learned (SuperPoint, SuperGlue, LightGlue, LoFTR, RoMa) approaches. Translation and rigid models prove insufficient under tissue deformation, while TPS with random sampling achieves the strongest landmark-derived alignment of the evaluated configurations; among the learned matchers, used without fine-tuning, only RoMa offers a robust, if modest, advantage over other methods. At the sequence level, pairwise registration quality proved an unreliable predictor of final mosaic quality, so mosaic quality must be evaluated directly rather than inferred from pairwise metrics.
Chinese Translation
共聚焦激光内镜（CLE）提供实时的细胞分辨率光学活检，但其视野较窄，图像马赛克技术可以扩展视野以提供解剖背景。由于逐行采集、探头运动和探头与组织的相互作用，帧对齐通常需要非线性变换，其准确性难以量化：灵活的变换模型能够拟合强度特征和噪声，因此基于外观的度量（如归一化互相关（NCC））可能在没有真正几何准确性提高的情况下改善。因此，我们建立了一个包含132对帧的数据库，涵盖来自4名患者的14个pCLE序列，并手动注释了标志物对应关系，以便目标配准误差（TRE）可以作为NCC的几何基础补充。我们评估了逐步增加变换模型自由度的影响，从平移到薄板样条（TPS），以及六种特征匹配后端，涵盖经典（Shi-Tomasi、Lucas-Kanade）和学习（SuperPoint、SuperGlue、LightGlue、LoFTR、RoMa）方法。平移和刚性模型在组织变形下证明不足，而随机采样的TPS在评估的配置中实现了最强的基于标志物的对齐；在学习匹配器中，未经过微调的RoMa仅在其他方法中提供了稳健但适度的优势。在序列层面，成对配准质量被证明是最终马赛克质量的不可靠预测因子，因此马赛克质量必须直接评估，而不是从成对度量中推断。
cs.CV / 188 / 2609.24564
HyperCLIP++: Fine-tuning CLIP forOpen-vocabulary Semantic Segmentation in Hyperbolic Space
HyperCLIP++：在双曲空间中微调CLIP以实现开放词汇语义分割
Abstract
CLIP, a foundational vision-language model, has emerged as a powerful tool for open-vocabulary semantic segmentation. While freezing CLIP's text encoder is known to preserve its generalization capability, recent studies show that fine-tuning both CLIP's text and image encoders jointly significantly enhances segmentation performance, especially for classes from open sets. In this work, we explain this phenomenon from the perspective of hierarchy alignment, since during fine-tuning, the hierarchical level of image embeddings shifts from image-level to pixel-level. We achieve this by leveraging hyperbolic space, which naturally encodes hierarchical structures. Our key observation is that, during fine-tuning, the hyperbolic radius of CLIP's text embeddings decreases, facilitating better alignment with the pixel-level granularity of visual data. Building on this, we propose HyperCLIP++, a novel and parameter-efficient adaptation strategy. HyperCLIP++ directly adjusts the hyperbolic radius of CLIP's embeddings via scaling transformations to achieve a hierarchy alignment to the target task, i.e., segmentation. To ensure this hierarchy alignment is effected consistently across both modalities and preserves their cross-modal alignment during training, HyperCLIP++ integrates a Dual Cross-Relation Communication (DCRC) module that synchronizes these adjustments between the vision and text pathways. Our experiments show that HyperCLIP++ achieves state-of-the-art performance across three benchmarks while fine-tuning only approximately 5% of CLIP's total parameters. More importantly, we observe that after adjustment, CLIP's text embeddings exhibit a relatively fixed hyperbolic radius across datasets, suggesting that the hierarchical level required for this segmentation task might be quantified using the hyperbolic radius.
Chinese Translation
CLIP作为一种基础的视觉-语言模型，已成为开放词汇语义分割的强大工具。尽管冻结CLIP的文本编码器被认为可以保持其泛化能力，但最近的研究表明，联合微调CLIP的文本和图像编码器显著提升了分割性能，尤其是对于开放集中的类别。在本研究中，我们从层次对齐的角度解释这一现象，因为在微调过程中，图像嵌入的层次级别从图像级别转变为像素级别。我们通过利用双曲空间来实现这一点，该空间自然地编码了层次结构。我们的关键观察是，在微调过程中，CLIP的文本嵌入的双曲半径减小，从而促进了与视觉数据像素级粒度的更好对齐。在此基础上，我们提出了HyperCLIP++，一种新颖且参数高效的适应策略。HyperCLIP++通过缩放变换直接调整CLIP嵌入的双曲半径，以实现与目标任务（即分割）的层次对齐。为了确保这种层次对齐在两种模态之间一致地实现，并在训练过程中保持它们的跨模态对齐，HyperCLIP++集成了一个双重交叉关系通信（DCRC）模块，以同步视觉和文本路径之间的这些调整。我们的实验表明，HyperCLIP++在三个基准测试中实现了最先进的性能，同时仅微调了大约5%的CLIP总参数。更重要的是，我们观察到在调整后，CLIP的文本嵌入在不同数据集上表现出相对固定的双曲半径，这表明该分割任务所需的层次级别可能可以通过双曲半径来量化。
cs.CV / 189 / 2609.24565
Active Visual Sampling with a Connectome-Constrained Fly Model for One-Shot Hatch Recognition in Architectural Drawings
基于连接组约束的果蝇模型的主动视觉采样在建筑图纸中的一次性填充识别
Abstract
Architectural drawings encode material classes through repeated hatch patterns. We test whether a connectome-constrained fly visual network, pretrained for motion, can be repurposed without task-specific weight updates as a descriptor for one-shot hatch matching. Each 64 x 64 patch is translated over eight scan trajectories and summarized across 57 cell types; query descriptors are then matched to one legend strip per class. On 400 development sheets from a synthetic benchmark built on CubiCasa5K geometry, the frozen fly pipeline reaches 0.857 area-weighted accuracy and 0.910 with an extended legend. On an equal-brightness orientation condition it reaches 0.840 versus 0.299 for eleven pixel statistics, while a Gabor bank reaches 0.900. Replacing drift with a repeated still frame lowers the combined equal-condition score by 0.089 [0.066, 0.112]. However, a receptors-only descriptor reaches 0.891 and a task-trained 5,888-parameter CNN averages 0.959, so the current evidence supports transfer and the usefulness of active sampling, but not an advantage of the biological wiring. We separate project-recorded results from recomputed checks and report a small real-drawing audit. The supported claim is therefore narrow: motion-oriented biological vision can be repurposed as a useful texture representation for architectural hatch matching, while the topology contribution and end-to-end BIM utility remain open questions.
Chinese Translation
建筑图纸通过重复的填充模式编码材料类别。我们测试了一种连接组约束的果蝇视觉网络，该网络经过运动预训练，能否在不进行任务特定权重更新的情况下，作为一次性填充匹配的描述符进行重新利用。每个64 x 64的图块在八条扫描轨迹上进行平移，并在57种细胞类型中进行汇总；查询描述符随后与每个类别的一个图例条带进行匹配。在基于CubiCasa5K几何构建的合成基准的400个开发图纸上，冻结的果蝇管道达到了0.857的面积加权准确率，扩展图例后达到了0.910。在相同亮度的方向条件下，其准确率为0.840，而十一种像素统计的准确率为0.299，Gabor滤波器组的准确率为0.900。用重复的静态帧替代漂移将组合的相等条件得分降低了0.089 [0.066, 0.112]。然而，仅使用受体的描述符达到了0.891，而经过任务训练的5,888参数CNN的平均准确率为0.959，因此当前证据支持迁移和主动采样的有效性，但不支持生物连接的优势。我们将项目记录的结果与重新计算的检查分开，并报告了一个小规模的真实图纸审计。因此，所支持的主张相对狭窄：以运动为导向的生物视觉可以重新用于建筑填充匹配的有用纹理表示，而拓扑贡献和端到端建筑信息模型（BIM）的实用性仍然是未解的问题。
cs.CV / 190 / 2609.24595
Applications of Neural Cellular Automata: State of the Art, Challenges and Opportunities
神经元细胞自动机的应用：现状、挑战与机遇
Abstract
Neural Cellular Automata (NCAs) are a new type of neural network architecture which enable accurate and robust inference at extremely small model sizes. Recently, NCAs have advanced to become interesting low-resource alternatives to convolution- and attention-based architectures for various tasks such as image analysis, synthetic image generation, and simulation. The rapid development and increased research interest necessitate a comprehensive review of the emerging technology. This review provides an overview of the fundamentals of NCAs, applications to medical imaging, as well as insights into the state of the art. We analyze recent modifications to the originally proposed NCA architecture with respect to their efficiency and accuracy. Furthermore, we review practical applications in real-world scenarios with a focus on medical image analysis, segmentation, classification, registration, depth estimation, and image synthesis. Finally, we identify several advantages of NCAs, research gaps, and conclude with an analysis of future opportunities for NCAs in medical applications in confined settings or areas that have particular demands for robustness or efficient data processing.
Chinese Translation
神经元细胞自动机（Neural Cellular Automata, NCAs）是一种新型的神经网络架构，能够在极小的模型规模下实现准确而稳健的推理。近期，NCAs 已经发展成为卷积和注意力基础架构的有趣低资源替代方案，适用于图像分析、合成图像生成和模拟等多种任务。快速的发展和日益增长的研究兴趣使得对这一新兴技术进行全面评审变得必要。本文综述了 NCAs 的基本原理、在医学成像中的应用，以及对现状的洞察。我们分析了对最初提出的 NCA 架构的近期修改，重点关注其效率和准确性。此外，我们回顾了在现实场景中的实际应用，特别关注医学图像分析、分割、分类、配准、深度估计和图像合成。最后，我们识别出 NCAs 的几个优势、研究空白，并以对 NCAs 在特定环境或对稳健性或高效数据处理有特殊需求的医学应用未来机遇的分析作为结尾。
cs.CV / 191 / 2609.24612
Beyond Uniform Subspaces: Spectrum-Aware and Depth-Adaptive Fusion for Multi-Task Model Merging
超越均匀子空间：面向谱特征和深度自适应的多任务模型合并
Abstract
Model merging aims to consolidate multiple task-specific models without access to extra training process. However, existing subspace-based methods largely rely on a uniform treatment of task updates, overlooking their intrinsic spectral and depth-wise heterogeneity. We identify two key deviations from this assumption: different tasks require different subspace capacity and exhibit different tolerance to spectral transformation, while subspace projection introduces depth-dependent distortion. Based on these observations, we propose SADA-Merging, a spectrum-aware and depth-adaptive framework for data-free model merging. SADA-Merging allocates task-specific subspace capacity according to spectral complexity, adapts spectral preservation according to task-wise plasticity, and applies depth-dependent anchoring to compensate for projection-induced distortion. This enables the fusion process to adapt to both the intrinsic geometry of each task and its sensitivity across network depth. SADA-Merging operates directly on task updates and is applicable to both full fine-tuning and LoRA settings. Extensive experiments demonstrate consistent improvements over existing data-free merging methods across different task scales and adaptation settings.
Chinese Translation
模型合并旨在整合多个特定任务的模型，而无需额外的训练过程。然而，现有的基于子空间的方法在很大程度上依赖于对任务更新的统一处理，忽视了其内在的谱特征和深度异质性。我们识别出这一假设的两个关键偏差：不同任务需要不同的子空间容量，并且对谱变换的容忍度各异，同时子空间投影引入了深度相关的失真。基于这些观察，我们提出了SADA-Merging，一种面向谱特征和深度自适应的数据无关模型合并框架。SADA-Merging根据谱复杂性分配任务特定的子空间容量，根据任务的可塑性调整谱保持，并应用深度相关的锚定来补偿投影引起的失真。这使得融合过程能够适应每个任务的内在几何特征及其在网络深度上的敏感性。SADA-Merging直接在任务更新上操作，适用于全微调和LoRA设置。大量实验表明，在不同任务规模和适应设置下，SADA-Merging在现有数据无关合并方法中表现出一致的改进。
cs.CV / 192 / 2609.24619
Video-based Surgical Skill Assessment Using Dynamics-and-Uncertainty-Aware Tree-based Gaussian Process Classifier
基于视频的外科技能评估：动态与不确定性感知树状高斯过程分类器
Abstract
The proposed pipeline integrates a representation-flow convolutional neural network with a dynamics- and uncertainty-aware tree-based Gaussian Process classifier. In this framework, latent motion dynamics are exploited both as discriminative representations and as a source of input uncertainty, enhancing robustness against temporal variations and abnormal motion transitions. Compared with conventional deep learning approaches, the proposed strategy requires less training data and offers improved computational efficiency. To further improve classification performance, we introduce novel semantic-aware compound kernels that effectively capture semantic, flow, and dynamic information embedded in surgical video features. In addition, uncertainty-aware kernels are developed to strengthen the robustness and practical applicability of the compound kernel framework. The proposed method is evaluated on two benchmark datasets, namely the JIGSAWS and the Cataract-LMM (Capsulorhexis) datasets. Experimental results demonstrate strong performance across both datasets, including the LOSO and LOUO evaluation protocols on JIGSAWS, including the subject-independent LOUO protocol on JIGSAWS, on which the framework attains a mean accuracy of \ph{96.9}\%; results under the within-subject LOSO protocol are reported for comparability with prior work, achieving competitive accuracy while substantially reducing computational cost. Overall, the proposed pipeline provides an efficient and accurate framework for video-based surgical skill assessment.
Chinese Translation
所提出的流程将表示流卷积神经网络与动态和不确定性感知的树状高斯过程分类器相结合。在该框架中，潜在的运动动态既作为判别表示被利用，也作为输入不确定性的来源，从而增强了对时间变化和异常运动过渡的鲁棒性。与传统的深度学习方法相比，所提出的策略需要更少的训练数据，并提供了更高的计算效率。为了进一步提高分类性能，我们引入了新颖的语义感知复合核，有效捕捉嵌入在外科视频特征中的语义、流动和动态信息。此外，开发了不确定性感知核，以增强复合核框架的鲁棒性和实际适用性。所提出的方法在两个基准数据集上进行了评估，即 JIGSAWS 和 Cataract-LMM（Capsulorhexis）数据集。实验结果在两个数据集上均表现出强劲的性能，包括在 JIGSAWS 上的 LOSO 和 LOUO 评估协议，以及在 JIGSAWS 上的受试者独立 LOUO 协议，其中该框架达到了 96.9% 的平均准确率；在受试者内 LOSO 协议下的结果也被报告，以便与之前的工作进行比较，取得了具有竞争力的准确率，同时显著降低了计算成本。总体而言，所提出的流程为基于视频的外科技能评估提供了一个高效且准确的框架。
cs.CV / 193 / 2609.24626
Relationally Grounded Latent World Models for Autonomous Driving
基于关系的潜在世界模型在自动驾驶中的应用
Abstract
Latent world models learn predictive representations for autonomous driving, but the relational semantics these states preserve often remain implicit. We investigate whether traffic scene graphs can serve as privileged semantic supervision for latent world representations. Building on LAW, we construct actor-centric scene graphs from nuScenes 3D annotations, encode their serialized relational structure using a frozen text embedding model, and align the visual latent representations with this semantic target during training. We remove the supervision branch at inference, so it requires neither scene graphs nor 3D annotations and adds no test-time computation. On nuScenes, our method reduces average trajectory L2 error from 0.661 to 0.622 (5.9%) and collision rate from 0.456 to 0.217 (52.4%) relative to our retrained LAW baseline. It also outperforms an unstructured caption-style semantic target, supporting the benefit of explicit relational structure for latent world-model representation learning.
Chinese Translation
潜在世界模型为自动驾驶学习预测性表征，但这些状态所保留的关系语义通常是隐含的。我们研究交通场景图是否可以作为潜在世界表征的特权语义监督。在此基础上，我们从nuScenes的3D注释构建以行为者为中心的场景图，使用冻结的文本嵌入模型对其序列化的关系结构进行编码，并在训练过程中将视觉潜在表征与这一语义目标对齐。在推理时，我们去除了监督分支，因此不需要场景图或3D注释，并且不会增加测试时的计算负担。在nuScenes数据集上，我们的方法将平均轨迹L2误差从0.661降低到0.622（减少5.9%），将碰撞率从0.456降低到0.217（减少52.4%），相较于我们重新训练的LAW基线。它还优于无结构的标题式语义目标，支持了显式关系结构在潜在世界模型表征学习中的优势。
cs.CV / 194 / 2609.24627
FedMust: Semi-supervised Multi-task Student-Teacher Federated Learning for Multi-organ CT Segmentation
FedMust：用于多脏器CT分割的半监督多任务学生-教师联邦学习
Abstract
Multi-organ segmentation using deep learning requires large amounts of annotated patient data; however, institutions often lack sufficiently large and diverse annotated datasets. Privacy constraints further prevent institutions from sharing patient data to overcome this limitation. Moreover, due to the labor-intensive nature of annotation and the scarcity of diverse expertise, institutions typically have labels for only a small portion of their local data, leaving the larger unlabeled portion unused. In this work, we propose a flexible semi-supervised federated multi-task student-teacher framework that leverages federated learning (FL) to improve multi-organ segmentation using both labeled and unlabeled data across participating sites. At each communication round, the proposed framework initiates local training, where clients with labels for the same task form a federation to produce an aggregated teacher model. The resulting teachers generate task-specific features for all data at each client. Subsequently, all clients form a second federation to train a multi-task student model with a shared encoder and task-specific decoders that replicate the teacher-generated features across all segmentation tasks. The aggregated student model is then used to update the local teachers and initiate the next training round. Extensive experiments demonstrated the effectiveness of the proposed method compared with local and federated single-organ models, yielding an average performance gain of 13 percent across clients. The experiments also demonstrated the impact of multi-task learning and unlabeled data and the applicability of the framework in relaxing labeled-data requirements for client participation. The code is available at https://github.com/AshknMrd/FedMust.
Chinese Translation
使用深度学习进行多脏器分割需要大量标注的患者数据；然而，机构通常缺乏足够大且多样化的标注数据集。隐私限制进一步阻止了机构共享患者数据以克服这一限制。此外，由于标注的劳动密集型特性和多样化专业知识的稀缺，机构通常仅对其本地数据的一小部分有标签，导致更大比例的未标记数据未被利用。在本研究中，我们提出了一种灵活的半监督联邦多任务学生-教师框架，该框架利用联邦学习（Federated Learning, FL）来改善多脏器分割，使用参与站点的标记和未标记数据。在每个通信轮次中，所提出的框架启动本地训练，其中具有相同任务标签的客户端形成一个联盟，以生成一个聚合的教师模型。生成的教师为每个客户端的所有数据生成任务特定的特征。随后，所有客户端形成第二个联盟，以训练一个具有共享编码器和任务特定解码器的多任务学生模型，这些解码器在所有分割任务中复制教师生成的特征。聚合的学生模型随后用于更新本地教师并启动下一轮训练。大量实验表明，与本地和联邦单脏器模型相比，所提出的方法的有效性，在所有客户端中平均性能提升了13%。实验还展示了多任务学习和未标记数据的影响，以及该框架在放宽客户端参与的标记数据要求方面的适用性。代码可在 https://github.com/AshknMrd/FedMust 获取。
cs.CV / 195 / 2609.24634
High-resolution Nitrogen Dioxide Maps Reveal Exposure Limit Breaches across Europe
高分辨率二氧化氮地图揭示欧洲范围内的暴露限值违规情况
Abstract
Nitrogen dioxide (NO2) is a common air pollutant, released into the atmosphere through the incomplete burning of fossil fuels, and associated with respiratory and cardiovascular diseases in humans. Ambient NO2 concentrations are regulated through air-quality limits assessed with a sparse network of fixed monitors. The revised EU Ambient Air Quality Directive (2024/2881) introduces a daily NO2 limit to be met from 2030. At present, neither the regulatory monitoring network nor existing coarse, annual-mean models can resolve NO2 concentrations at the spatio-temporal resolutions necessary to assess compliance. Here we map NO2 across Europe at hourly and 10m resolution with a machine-learning model that combines ground monitors with satellite, reanalysis, land-use, traffic and emission data and returns a calibrated predictive distribution at every location. Validated against held-out regulatory monitors and independent citizen-science campaigns, the maps resolve high-resolution spatiotemporal NO2 gradients for 110 metropolitan areas in Europe. We reconstruct the daily compliance statistic across those regions and find limit breaches in 91 EU air quality zones deemed compliant by the regulatory monitoring network, covering a population of approximately 135M. Beyond air quality zones and monitor locations, an estimated 9-9.4% (20M) of the population in mapped regions lives in areas where the daily NO2 limit is breached. The high-resolution maps offer a route to population-scale assessment of compliance with the 2030 limits.
Chinese Translation
二氧化氮（NO2）是一种常见的空气污染物，通过不完全燃烧化石燃料释放到大气中，并与人类的呼吸和心血管疾病相关。环境中NO2浓度的监管是通过稀疏的固定监测网络评估空气质量限值。修订后的欧盟环境空气质量指令（2024/2881）引入了一个每日NO2限值，要求在2030年之前达到。目前，既有的监管监测网络也无法通过现有粗略的年度平均模型在必要的时空分辨率上解决NO2浓度，以评估合规性。在此，我们利用一种机器学习模型，将地面监测数据与卫星、再分析、土地利用、交通和排放数据相结合，以每小时和10米的分辨率绘制欧洲的NO2分布，并在每个位置返回经过校准的预测分布。经过与保留的监管监测数据和独立公民科学活动的验证，这些地图为欧洲110个大都市区解析了高分辨率的时空NO2梯度。我们重建了这些地区的每日合规统计数据，发现91个被监管监测网络认定为合规的欧盟空气质量区存在限值违规情况，涉及约1.35亿人口。超出空气质量区和监测位置的估计显示，在绘制区域中，约有9-9.4%（2000万）的人口生活在每日NO2限值被违反的地区。这些高分辨率地图为评估2030年限值的合规性提供了一个基于人口规模的途径。
cs.CV / 196 / 2609.24668
Ev-YOLO: Uncertainty-Aware Object Detection via a Unified Evidential Formulation
Ev-YOLO：通过统一证据公式实现的不确定性感知目标检测
Abstract
Reliable uncertainty estimation is essential for deploying object detectors in autonomous systems operating in uncertain environments. Evidential Deep Learning (EDL) provides a principled framework for uncertainty-aware classification by representing network outputs as evidence and interpreting predictions through subjective logic. However, existing evidential object detectors typically combine evidential classification with regression uncertainty models that do not share the same theoretical foundation. In this work, we propose an evidential version of YOLOv8 in which both classification and bounding-box regression are formulated within a common evidential framework. Our approach exploits YOLOv8's distribution-based bounding-box representation, allowing the evidential formulation to be applied not only to classification but also to localisation. As a result, both tasks produce belief, uncertainty, and probability estimates that can be interpreted within the Dempster--Shafer framework. Experiments on KITTI, MUSES, and nuScenes show that the resulting detector remains broadly competitive with standard YOLOv8 in terms of detection accuracy while providing a localisation uncertainty that effectively discriminates between correct and erroneous detections. Moreover, this uncertainty becomes increasingly discriminative under domain shift.
Chinese Translation
可靠的不确定性估计对于在不确定环境中部署目标检测器的自主系统至关重要。证据深度学习（Evidential Deep Learning, EDL）提供了一个有原则的框架，通过将网络输出表示为证据，并通过主观逻辑解释预测，从而实现不确定性感知的分类。然而，现有的证据目标检测器通常将证据分类与不共享相同理论基础的回归不确定性模型相结合。在本研究中，我们提出了一种YOLOv8的证据版本，其中分类和边界框回归均在一个共同的证据框架内进行公式化。我们的方法利用YOLOv8基于分布的边界框表示，使得证据公式不仅可以应用于分类，还可以应用于定位。因此，这两个任务都产生了信念、不确定性和概率估计，这些估计可以在邓普斯特-谢弗（Dempster--Shafer）框架内进行解释。在KITTI、MUSES和nuScenes上的实验表明，所得到的检测器在检测精度方面与标准YOLOv8保持广泛竞争，同时提供了有效区分正确和错误检测的定位不确定性。此外，这种不确定性在领域转移下变得越来越具辨别力。
cs.CV / 197 / 2609.24691
What Makes a Good Medical Image Tokenizer? Rethinking Reconstruction and Generation in Medical Image Tokenization
什么是优秀的医学图像分词器？重新思考医学图像分词中的重建与生成
Abstract
Latent diffusion models now dominate medical image generation, and every such pipeline rests on a \emph{tokenizer} that compresses images into the latent codes for image generation to operate on. Thereby, the tokenizer choice bounds every downstream task from reconstruction fidelity and generation quality to the representations available for downstream analysis. Yet, medical imaging pipelines routinely utilize tokenizers from natural imaging on the hypothesis that their behavior carries over. However, this is an assumption never tested in the medical imaging regime, where datasets are orders of magnitude smaller and images exhibit far lower inter-sample variance. We present a systematic evaluation of medical image tokenizers evaluating thirty configurations across ten model families on twelve datasets at three compression factors, spanning reconstruction, generation, latent geometry, downstream classification, and memorization. We find that (1) performance on image reconstruction and generation strongly correlate, unlike prior reports on natural images; (2) modern tokenizers use nearly all of their codebook entries, but still leave most of the latent space unused; (3) training-set memorization is mild and is further suppressed by stronger latent space compression; and (4) discrete quantization can largely preserve downstream classification, with lookup-free schemes being the main exception.
Chinese Translation
潜在扩散模型目前主导着医学图像生成，每个这样的流程依赖于一个 extit{分词器}，该分词器将图像压缩为潜在编码，以便进行图像生成。因此，分词器的选择限制了从重建保真度和生成质量到可用于下游分析的表示的每个下游任务。然而，医学成像流程通常利用自然成像中的分词器，假设其行为可以迁移。然而，这一假设在医学成像领域从未经过验证，因为该领域的数据集规模小几个数量级，且图像之间的样本方差远低于自然图像。我们对医学图像分词器进行了系统评估，评估了在三个压缩因子下，跨十个模型系列的三十种配置在十二个数据集上的表现，涵盖了重建、生成、潜在几何、下游分类和记忆化。我们发现：(1) 图像重建和生成的性能强相关，这与之前对自然图像的报告不同；(2) 现代分词器几乎使用了其所有的编码本条目，但仍然有大部分潜在空间未被使用；(3) 训练集的记忆化程度较轻，并且在更强的潜在空间压缩下进一步抑制；(4) 离散量化在很大程度上可以保持下游分类，查找无方案是主要的例外。
cs.CV / 198 / 2609.24727
ReSTI: A Source-Grounded Audit and Repair of STI-Bench
ReSTI：基于源的STI-Bench审计与修复
Abstract
Spatial--temporal benchmarks are valid only when their questions, source annotations, and answer options identify the same physical quantity. We audit STI-Bench against the official ScanNet, Waymo, and Omni6DPose sources and find systematic coordinate-system and timestamp errors, under-specified targets and times, and disagreements between keyed options and answer details. We introduce ReSTI, a source-backed revision that reconstructs every recoverable answer under an explicit target, time, coordinate system, physical quantity, and unit. Source reconstruction reveals task-level geometric failures: ScanNet Grounding omits the required alignment between annotation and raw camera coordinate systems, while Orientation measures camera rotation on the wrong plane. ReSTI replaces these labels with explicit, source-consistent geometric definitions and corrects other source-verifiable defects, including Waymo poses evaluated at the wrong timestamp. Across 2,064 legacy questions, ReSTI retains 1,782 questions and records 282 evidence-backed exclusions. ReSTI therefore provides a conservative and source-traceable basis for evaluating precise video spatial--temporal reasoning. Project page: https://github.com/pengzhansun/ReSTI.
Chinese Translation
空间-时间基准仅在其问题、源注释和答案选项识别相同物理量时有效。我们对STI-Bench进行审计，依据官方的ScanNet、Waymo和Omni6DPose源，发现系统性的坐标系和时间戳错误、目标和时间的描述不充分，以及键入选项与答案细节之间的不一致。我们引入ReSTI，这是一个基于源的修订，重构每一个可恢复的答案，明确目标、时间、坐标系、物理量和单位。源重构揭示了任务级几何失败：ScanNet Grounding省略了注释与原始相机坐标系之间所需的对齐，而Orientation在错误的平面上测量相机旋转。ReSTI用明确的、与源一致的几何定义替换这些标签，并修正其他可源验证的缺陷，包括在错误时间戳下评估的Waymo姿态。在2064个遗留问题中，ReSTI保留了1782个问题，并记录了282个有证据支持的排除。因此，ReSTI为评估精确的视频空间-时间推理提供了一个保守且可追溯于源的基础。项目页面：https://github.com/pengzhansun/ReSTI。
cs.CV / 199 / 2609.24732
GraphSVR: q-Space--Aware Graph-Based Slice-to-Volume Registration for Diffusion MRI
GraphSVR：基于图的q空间感知切片到体积配准方法用于扩散MRI
Abstract
Diffusion-weighted imaging (DWI) remains highly vulnerable to subject motion, particularly in time-efficient protocols and in motion-prone populations. While slice-to-volume registration (SVR) can mitigate inter-slice and inter-stack misalignment, diffusion MRI introduces additional complexity due to diffusion-direction-dependent contrast and the requirement to align dozens of measurements within a common reference frame, effectively yielding a 4D registration problem. Existing approaches rely primarily on sequential modeling or pairwise similarity and often degrade under sparse gradient sampling or severe motion. We introduce GraphSVR, a q-space-aware graph-based framework for 4D SVR registration in DWI. GraphSVR represents slice groups as nodes in an acquisition-structured graph, with edges encoding temporal proximity, spatial slice geometry and diffusion encoding relationships. A graph neural network predicts globally consistent stack-wise rigid motion, optimized in a self-supervised, zero-shot manner using only an anatomical reference image, without requiring paired ground-truth motion. We evaluate GraphSVR using both fully synthetic diffusion simulations and realistic recombination-based simulations from real acquisitions with controllable motion severity and gradient sparsity. Performance is quantified using grid error (mm) and rotation error relative to known ground-truth transforms. Under severe motion, GraphSVR reduces grid error and rotation error by 73% compared to FSL eddy, the standard DWI motion-correction method, with the largest gains observed in sparse-direction regimes. These results demonstrate that explicitly modeling acquisition structure through graph-based reasoning improves robustness and global consistency in 4D DWI motion estimation. Code is available at https://github.com/nogakertes/GraphSVR.git.
Chinese Translation
扩散加权成像（DWI）在受试者运动方面仍然高度脆弱，尤其是在时间效率高的协议和易受运动影响的人群中。虽然切片到体积配准（SVR）可以减轻切片间和堆叠间的错位，但由于扩散方向依赖的对比度以及在共同参考框架内对数十个测量进行对齐的要求，扩散MRI引入了额外的复杂性，实际上形成了一个4D配准问题。现有方法主要依赖于序列建模或成对相似性，并且在稀疏梯度采样或严重运动下往往会降级。我们提出了GraphSVR，一种针对DWI中4D SVR配准的q空间感知基于图的框架。GraphSVR将切片组表示为获取结构图中的节点，边缘编码时间接近性、空间切片几何形状和扩散编码关系。图神经网络预测全局一致的堆叠刚性运动，采用自监督的零样本方式进行优化，仅使用解剖参考图像，而无需配对的真实运动数据。我们使用完全合成的扩散模拟和来自真实采集的可控运动严重性和梯度稀疏性的现实重组基础模拟来评估GraphSVR。性能通过网格误差（mm）和相对于已知真实变换的旋转误差进行量化。在严重运动下，GraphSVR将网格误差和旋转误差分别减少了73%，与标准DWI运动校正方法FSL eddy相比，在稀疏方向情况下观察到最大的增益。这些结果表明，通过基于图的推理显式建模获取结构可以提高4D DWI运动估计的鲁棒性和全局一致性。代码可在 https://github.com/nogakertes/GraphSVR.git 获取。
cs.CV / 200 / 2609.24736
MiTHras: Task-specific Hierarchical Semi-supervised Contrastive Masked Autoencoder for Mitotic Figure Analysis
MiTHras：用于有丝分裂图像分析的任务特定层次半监督对比掩码自编码器
Abstract
Mitotic figure (MF) analysis supports tumor grading and prognostic assessment, but automated models remain sensitive to differences in tissue type and image acquisition. We present MiTHras, a task-specific pretraining framework that combines pseudo-label-guided image- and token-level contrastive learning with masked reconstruction. We construct TCGA-MF-Pseudo, a corpus of 1.8 million cell-centered images from 14 TCGA cohorts spanning 11 organ sites. Comprehensive evaluation on MF classification, detection, count-based survival prediction, and subtype classification demonstrates the efficacy of MiTHras. It achieves the highest mean F1 on all three MF classification benchmarks and both subtype benchmarks. MiTHras also outperforms general-purpose and pathology foundation encoders by a larger margin under frozen-encoder linear probing than under full fine-tuning. Although detection gains are modest due to a shared candidate-detection stage, ablations confirm that token-level supervision improves typical-versus-atypical classification and linear probing. These findings establish that MiTHras yields robust, transferable representations for automated mitotic activity assessment.
Chinese Translation
有丝分裂图像（MF）分析支持肿瘤分级和预后评估，但自动化模型对组织类型和图像获取的差异仍然敏感。我们提出了MiTHras，这是一种任务特定的预训练框架，结合了伪标签引导的图像和标记级对比学习与掩码重建。我们构建了TCGA-MF-Pseudo，这是一个包含来自14个TCGA队列、覆盖11个器官部位的180万张以细胞为中心的图像的语料库。对MF分类、检测、基于计数的生存预测和亚型分类的全面评估证明了MiTHras的有效性。它在所有三个MF分类基准和两个亚型基准上都达到了最高的平均F1分数。MiTHras在冻结编码器线性探测下的表现也明显优于通用和病理基础编码器，而在完全微调下则差距较小。尽管由于共享候选检测阶段，检测增益较为温和，但消融实验确认标记级监督改善了典型与非典型分类和线性探测。这些发现表明，MiTHras为自动化的有丝分裂活动评估提供了稳健且可转移的表示。
cs.CV / 201 / 2609.24737
Ananke: Contractive Torus Attractor Networks
Ananke：收缩环面吸引子网络
Abstract
We introduce Ananke, a representation-learning framework that scaffolds latent representations onto a structured product-torus prior, and its flagship visual backbone realization, Contractive Torus Attractor Networks (CTAN). By factorizing high-dimensional latent spaces into an orthogonal direct sum of two-dimensional phase planes ($\bigoplus_{k=1}^K \R^2$), Ananke coordinates feature updates via a decoupled dual-phase continuous flow: skew-symmetric Hamiltonian transport moves features tangentially along energy level sets to preserve semantic phase invariants, while signed gradient dissipation contracts transverse perturbations normally toward target invariant manifolds. For circular potential families with frozen parameters, logarithmic radial feedback yields the Exact Log-Symplectic Flow (ELSF), an analytical closed-form mapping with exact exponential decay of log-radius error that evaluates in a single forward pass without numerical integration. We establish local input-to-state bounds for level-set deviations and log-radius errors, and characterize the normal hyperbolicity and persistence of the ideal product torus under bounded perturbations. We further formulate the architecture through Lie--Trotter operator splitting, unifying spatial depthwise diffusion with local manifold contraction, and analyze both exact trigonometric flows and hardware-friendly symplectic dual-shear variants. Across natural image benchmarks (CIFAR-100) and clinically challenging endoscopy datasets (Kvasir-v2), CTAN demonstrates exceptional parameter efficiency: an ultra-compact hierarchical model with merely 0.27M parameters achieves 90.52\% accuracy on Kvasir-v2, outperforming 25M+ baselines (ResNet-50, DenseNet-161) by nearly two orders of magnitude in capacity, while scaled variants attain 80.32\% top-1 accuracy on CIFAR-100.
Chinese Translation
我们介绍了Ananke，一个将潜在表示构建在结构化的产品环面先验上的表示学习框架，以及其旗舰视觉骨干实现——收缩环面吸引子网络（Contractive Torus Attractor Networks, CTAN）。通过将高维潜在空间分解为二维相平面的正交直和（$igoplus_{k=1}^K ext{R}^2$），Ananke通过解耦的双相连续流来协调特征更新：反对称哈密顿传输沿能量水平集切向移动特征，以保持语义相位不变性，而带符号的梯度耗散则将横向扰动正常收缩到目标不变流形。对于具有冻结参数的圆形势能族，对数径向反馈产生了精确对数辛流（Exact Log-Symplectic Flow, ELSF），这是一种具有对数半径误差精确指数衰减的解析闭式映射，能够在单次前向传递中评估，而无需数值积分。我们建立了关于水平集偏差和对数半径误差的局部输入到状态界限，并表征了理想产品环面在有界扰动下的正常双曲性和持久性。我们进一步通过李-特罗特算子分裂来制定架构，将空间深度扩散与局部流形收缩统一起来，并分析了精确的三角流和适合硬件的辛双剪切变体。在自然图像基准（CIFAR-100）和临床挑战性内窥镜数据集（Kvasir-v2）上，CTAN展现了卓越的参数效率：一个仅有0.27M参数的超紧凑层次模型在Kvasir-v2上达到了90.52\%的准确率，几乎以两个数量级的容量超越了25M+的基线模型（ResNet-50, DenseNet-161），而缩放变体在CIFAR-100上达到了80.32\\%的top-1准确率。
cs.CV / 202 / 2609.24768
PrismGPT: Proxy-Guided Learning for Region-Aware Photo Editing with Self-Synthesized Reasoning
PrismGPT：基于代理引导的区域感知照片编辑自我推理学习
Abstract
Professional photo finishing relies on both global adjustments and region-specific local edits guided by semantic masks, yet current automated methods handle this workflow only partially. We present PrismGPT, a Vision-Language Model (VLM) framework that produces structured, region-aware editing plans from a single input image without relying on commercial black-box tools. Training a VLM to simultaneously diagnose aesthetic deficiencies at both global and local levels while predicting precise editing parameters is challenging due to the vast combinatorial decision space. We address this through proxy-guided learning: two simpler proxy tasks -- operation decomposition and region-aware aesthetic ranking -- teach the foundational skills the model needs, while a competence-based dynamic scheduler automatically rebalances the multi-task training ratio, progressively shifting emphasis from the proxy tasks to the primary editing task as each skill is mastered. Crucially, all reasoning traces used for supervised fine-tuning are self-synthesized by the same base model, eliminating the need for a stronger external teacher. Experiments on MIT-Adobe FiveK and SPIRE, a new professionally retouched benchmark we introduce, show that PrismGPT achieves state-of-the-art results while using only ~6% of the training data compared to the previous best method.
Chinese Translation
专业的照片后期处理依赖于全局调整和由语义掩码引导的区域特定局部编辑，但当前的自动化方法仅部分处理这一工作流程。我们提出了PrismGPT，一个视觉-语言模型（Vision-Language Model, VLM）框架，它能够从单一输入图像生成结构化的区域感知编辑计划，而无需依赖商业黑箱工具。训练一个VLM以同时诊断全局和局部层面的美学缺陷，并预测精确的编辑参数是具有挑战性的，因为决策空间的组合数量庞大。我们通过代理引导学习来解决这个问题：两个更简单的代理任务——操作分解和区域感知美学排名——教授模型所需的基础技能，同时基于能力的动态调度器自动重新平衡多任务训练比例，随着每项技能的掌握，逐步将重点从代理任务转移到主要编辑任务。重要的是，用于监督微调的所有推理痕迹均由同一基础模型自我合成，消除了对更强外部教师的需求。在MIT-Adobe FiveK和我们引入的新专业修饰基准SPIRE上的实验表明，PrismGPT在使用仅约6%的训练数据的情况下，取得了最先进的结果，远超之前的最佳方法。
cs.CV / 203 / 2609.24769
Brain Metastases Segmentation for BraTS 2026 Task 1: A Multi-Architecture Comparison
BraTS 2026 任务 1 的脑转移瘤分割：多架构比较
Abstract
Brain metastases are the most common intracranial malignancy, occurring in roughly 30% of patients with primary solid tumors and carrying a median survival near 5.9 months. Automated segmentation is critical for treatment planning and volumetric monitoring, but metastases are frequently small, numerous, and heterogeneous in size within a single patient. We compare a plain nnU-Net baseline, a Residual Encoder Large (ResEncL) variant, region-based training, and a Primus transformer model for BraTS-METS 2026 Task 1, using patient-grouped cross-validation to prevent leakage from the longitudinal UCSD subset. Primus (label-based) is our strongest individual model by aggregate DSC/NSD, achieving 0.710/0.761 (ET), 0.742/0.785 (TC), 0.683/0.689 (WT), and 0.531/0.436 (RC). ResEncL trails Primus on aggregate DSC/NSD but achieves substantially higher lesion-wise F1 (e.g. ET: 0.452 vs. 0.052); a probability-averaging ensemble of the two only partially preserves ResEncL's F1 advantage (ET lesion-wise F1: 0.064). We further report three postprocessing and label-reconstruction pitfalls we believe generalize beyond this challenge. Code is available at https://github.com/mahdiislam79/BraTS_METS_2026.
Chinese Translation
脑转移瘤是最常见的颅内恶性肿瘤，约30%的原发实体肿瘤患者会发生脑转移，且中位生存期接近5.9个月。自动分割对于治疗规划和体积监测至关重要，但转移瘤通常在单个患者中呈现出小、数量多和大小异质的特征。我们比较了一个基础的 nnU-Net 模型、一个残差编码器大型变体（Residual Encoder Large, ResEncL）、基于区域的训练以及一个 Primus 变换器模型，针对 BraTS-METS 2026 任务 1，采用患者分组的交叉验证以防止来自纵向 UCSD 子集的数据泄漏。Primus（基于标签）是我们在综合 DSC/NSD 指标上表现最强的单一模型，达到了 0.710/0.761（增强型转移瘤, ET）、0.742/0.785（肿瘤核心, TC）、0.683/0.689（全肿瘤, WT）和 0.531/0.436（坏死核心, RC）。ResEncL 在综合 DSC/NSD 上落后于 Primus，但在病灶级 F1 指标上显著更高（例如，ET: 0.452 对比 0.052）；两者的概率平均集成仅部分保留了 ResEncL 的 F1 优势（ET 病灶级 F1: 0.064）。我们进一步报告了三个后处理和标签重建的陷阱，我们认为这些问题在此挑战之外也具有普遍性。代码可在 https://github.com/mahdiislam79/BraTS_METS_2026 获取。
cs.CV / 204 / 2609.24782
Virtual neural networks: hundreds of souls in a body
虚拟神经网络：一个身体中的数百个灵魂
Abstract
A new concept, termed virtual neural networks, is introduced, where the count of trainable parameters is kept constant, and scalability is attained purely through computational resources. This concept is an abstract framework that can be realized using any standard convolutional neural network. It merges siamese neural networks with a deep ensemble technique by generating numerous virtual models that share weights derived from a small set of physical models. The ensemble comprises up to hundreds of trained models simultaneously. All virtual networks take the same input, and their interconnected structure induces an internal distortion that boosts the entire ensemble robustness. The accuracy of the ensemble improves as the number of virtual networks increases, without changing the capacity. Virtual neural networks outperform larger capacity models, typical deep ensembles, and contemporary approaches like SWA and Masksembles. Additionally, the highest-performing individual model from the ensemble surpasses other models trained individually, even those with a greater number of parameters. Code: gitlab.com/EnginCZ/virtual-models-public
Chinese Translation
本文引入了一种新概念，称为虚拟神经网络，其中可训练参数的数量保持不变，扩展性纯粹通过计算资源实现。该概念是一个抽象框架，可以使用任何标准卷积神经网络来实现。它通过生成多个共享来自一小组物理模型权重的虚拟模型，将孪生神经网络与深度集成技术相结合。该集成同时包含多达数百个训练模型。所有虚拟网络接受相同的输入，其互联结构引发内部扭曲，从而增强整个集成的鲁棒性。随着虚拟网络数量的增加，集成的准确性提高，而容量保持不变。虚拟神经网络在性能上优于更大容量的模型、典型的深度集成以及现代方法如SWA和Masksembles。此外，来自集成的表现最佳的单个模型超越了其他单独训练的模型，即使那些参数数量更多的模型也不例外。代码：gitlab.com/EnginCZ/virtual-models-public
cs.CV / 205 / 2609.24787
Toward a foundation model for forest point clouds
朝向森林点云的基础模型
Abstract
Forest inventories increasingly rely on artificial intelligence (AI) models to derive forest attributes from large-scale 3D point clouds. Current models are typically specialized to a single task, sensor, and forest type, making adaptation expensive in terms of annotations, computation, and expertise. We ask whether a single pretrained model can instead learn transferable representations across diverse forest inventory settings. Inspired by recent developments in language modelling and computer vision, we take a step toward a foundation model (FM) for 3D forestry. Using LitePT as backbone, we first establish a strong supervised baseline that sets a new state of the art on forest semantic and instance segmentation, tree species classification, and age regression benchmarks. We then curate a large-scale unlabelled corpus spanning airborne, UAV, and mobile laser scanning across diverse forest ecosystems, and pretrain the same backbone using self-supervised learning. We systematically evaluate representation learning strategies by comparing training from scratch, supervised pretraining, and self-supervised pretraining across four representative forestry tasks, under varying annotation budgets. Compared with training from scratch, self-supervised pretraining accelerates model convergence and consistently improves performance when annotations are scarce. Compared with task-specific supervised pretraining, self-supervised pretraining yields more transferable representations across downstream forestry tasks. These findings identify the practical regime in which pretrained representations are most valuable and suggest that instance discrimination, rather than forest semantics, is the main remaining obstacle to a general-purpose 3D forest foundation model. Code and models are available at: https://github.com/prs-eth/ForPT.
Chinese Translation
森林清查越来越依赖人工智能（AI）模型从大规模3D点云中提取森林属性。目前的模型通常专注于单一任务、传感器和森林类型，使得在注释、计算和专业知识方面的适应成本高昂。我们探讨是否可以通过单一的预训练模型学习在不同森林清查环境中可转移的表示。受到最近语言建模和计算机视觉发展的启发，我们迈出了朝向3D林业基础模型（FM）的第一步。以LitePT作为骨干网络，我们首先建立了一个强大的监督基线，在森林语义和实例分割、树种分类和年龄回归基准上设定了新的最先进水平。随后，我们整理了一个涵盖多种森林生态系统的航空、无人机（UAV）和移动激光扫描的大规模无标签语料库，并使用自监督学习对相同的骨干网络进行预训练。我们通过比较从头开始训练、监督预训练和自监督预训练在四个代表性林业任务下的表现，系统评估了表示学习策略，考虑了不同的注释预算。与从头开始训练相比，自监督预训练加速了模型收敛，并在注释稀缺时持续提高性能。与特定任务的监督预训练相比，自监督预训练在下游林业任务中产生了更具可转移性的表示。这些发现确定了预训练表示最有价值的实际环境，并表明实例区分而非森林语义是通用3D森林基础模型的主要障碍。代码和模型可在以下网址获取：https://github.com/prs-eth/ForPT。
cs.CV / 206 / 2609.24788
Streaming Video Editing with Easy Adaptation
便捷适应的流媒体视频编辑
Abstract
In this paper, we propose SVEET, a framework that requires merely training on a pretrained bidirectional video diffusion model but supports high-quality streaming video editing in an auto-regressive fashion. To tackle this problem, we first systematically revisit existing video-to-video diffusion approaches and identify two key principles for such streaming adaptation: backbone feature disentanglement and conditional frame independence. Building on these insights, we develop a novel paradigm for controllable video generation. At its core, an auxiliary model branch encodes source video inputs with temporally independent self-attention, and the intermediate features are injected into the corresponding backbone blocks for streaming-compatible control. Moreover, to bridge the discrepancy between the feature spaces of bidirectional and streaming models, we propose a decoupled training scheme that explicitly enforces the orthogonality between the optimization directions of video controllability and model causality. Such disentanglement ensures compatibility between the two objectives at inference and facilitates smooth zero-shot knowledge transfer across heterogeneous backbone architectures. Extensive experiments demonstrate that SVEET achieves superior editing quality while maintaining real-time performance, attaining 15 FPS on a single H100 GPU 17 without any auxiliary acceleration techniques. Codes are available at https://github.com/YujiaHu1109/SVEET.
Chinese Translation
在本文中，我们提出了SVEET，一个仅需在预训练的双向视频扩散模型上进行训练的框架，但支持以自回归方式进行高质量的流媒体视频编辑。为了解决这一问题，我们首先系统性地回顾了现有的视频到视频扩散方法，并确定了流媒体适应的两个关键原则：主干特征解耦和条件帧独立性。在这些见解的基础上，我们开发了一种可控视频生成的新范式。其核心是一个辅助模型分支，通过时间上独立的自注意力编码源视频输入，并将中间特征注入到相应的主干模块中，以实现与流媒体兼容的控制。此外，为了弥合双向模型和流媒体模型特征空间之间的差异，我们提出了一种解耦训练方案，明确地强制视频可控性和模型因果性的优化方向之间的正交性。这种解耦确保了推理时两个目标之间的兼容性，并促进了跨异构主干架构的平滑零-shot知识转移。大量实验表明，SVEET在保持实时性能的同时，达到了优越的编辑质量，在单个H100 GPU上以15 FPS的速度运行，而无需任何辅助加速技术。代码可在 https://github.com/YujiaHu1109/SVEET 获取。
cs.CV / 207 / 2609.24813
INTCORT: Training-Free Spatial Reasoning Enhancement for Vision-Language Models via Input Transformations and Confidence Routing
INTCORT：通过输入变换和置信度路由实现无训练的视觉语言模型空间推理增强
Abstract
Vision-Language Models (VLMs) have demonstrated remarkable capabilities in multimodal tasks, yet they still exhibit poor ability in spatial reasoning. Existing training-dependent and training-free enhancement methods suffer from high computational costs with catastrophic forgetting and internal mechanism interference that compromises general capabilities, respectively. In this work, we first verify two key hypotheses: appropriate geometric image transformation and query-reversal transformation can recover incorrect spatial predictions, and correct predictions exhibit higher relation-token confidence than incorrect ones. Based on these findings, we propose INTCORT, a training-free spatial reasoning enhancement framework that constructs multiple inference views through input transformations and aggregates their predictions via relation-token confidence routing, without modifying the VLM's internal mechanisms. Experimental results on several commonly-used benchmarks demonstrate that INTCORT substantially improves spatial reasoning accuracy across diverse VLMs, achieving an average improvement of 10.01% over all models and benchmarks. Compared with prior works, INTCORT achieves superior performance with improvements of up to 25.01%.
Chinese Translation
视觉语言模型（VLMs）在多模态任务中展现了显著的能力，但在空间推理方面仍然表现不佳。现有的依赖训练和无训练的增强方法分别面临着高计算成本、灾难性遗忘以及内部机制干扰等问题，这些问题削弱了模型的通用能力。在本研究中，我们首先验证了两个关键假设：适当的几何图像变换和查询反转变换可以恢复不正确的空间预测，而正确的预测相较于不正确的预测具有更高的关系标记置信度。基于这些发现，我们提出了INTCORT，一个无训练的空间推理增强框架，通过输入变换构建多个推理视图，并通过关系标记置信度路由聚合它们的预测，而无需修改VLM的内部机制。在多个常用基准上的实验结果表明，INTCORT显著提高了不同VLM的空间推理准确性，所有模型和基准的平均提升达到10.01%。与之前的研究相比，INTCORT在性能上表现优越，提升幅度最高可达25.01%。
cs.CV / 208 / 2609.24814
Mobile Imaging Solutions for Medical Diagnosis: Trends and Applications
用于医学诊断的移动成像解决方案：趋势与应用
Abstract
Advances in processing power, camera technologies, and mobile image analysis have made smartphones and other mobile devices, such as laptops, increasingly suitable for medical diagnosis and healthcare applications. Researchers have developed low-cost solutions for the early detection and monitoring of various health conditions, including eye and ENT diseases, malnutrition, heart rate variability, skin and oral conditions, and injuries, using images captured by non-medical devices such as smartphones and webcams. This survey examines existing research on mobile image-based medical diagnosis, with an emphasis on its potential to enable low-cost and accessible healthcare. We comparatively analyze state-of-the-art solutions across different healthcare application categories, examining their advantages and limitations. Based on this analysis, we identify desirable characteristics of mobile image-based diagnostic tools and highlight areas where existing approaches have made progress as well as areas requiring further research. We also discuss application-specific and common challenges and outline directions for future research. Overall, this study provides a comprehensive overview of mobile image-based healthcare solutions and their potential to support low-cost disease diagnosis and monitoring, particularly for underserved populations in remote and resource-constrained settings.
Chinese Translation
处理能力、摄像技术和移动图像分析的进步，使得智能手机及其他移动设备（如笔记本电脑）在医学诊断和医疗应用中越来越适用。研究人员开发了低成本的解决方案，用于早期检测和监测各种健康状况，包括眼科和耳鼻喉科疾病、营养不良、心率变异性、皮肤和口腔状况以及伤害，这些解决方案利用非医疗设备（如智能手机和网络摄像头）捕获的图像。本文调查了现有的基于移动图像的医学诊断研究，强调其在实现低成本和可及医疗方面的潜力。我们对不同医疗应用类别的先进解决方案进行了比较分析，考察其优点和局限性。基于此分析，我们识别出移动图像诊断工具的理想特征，并强调现有方法取得进展的领域以及需要进一步研究的领域。我们还讨论了特定应用和共同面临的挑战，并概述了未来研究的方向。总体而言，本研究提供了关于基于移动图像的医疗解决方案的全面概述，以及它们在支持低成本疾病诊断和监测方面的潜力，特别是针对资源匮乏和偏远地区的弱势群体。
cs.CV / 209 / 2609.24825
ZVeC: A Zero-Shot Framework for Instance-Level Vehicle Extraction and Generative Point Cloud Completion
ZVeC：一种零样本实例级车辆提取与生成点云补全框架
Abstract
LiDAR point clouds acquired in underground environments exhibit severe geometric incompleteness due to occlusions and limited sensor viewpoints, making reliable point cloud completion challenging without large supervised datasets. We propose ZVeC, a zero-shot, instance-driven framework that reformulates scene-level completion as compositional object-level reconstruction. By decomposing a scene into semantic object instances, ZVeC reduces reconstruction ambiguity in cluttered environments while eliminating the need for scenario-specific training. Each segmented vehicle is completed independently using a depth- and 3D Gaussian-conditioned diffusion model that exploits generalized geometric priors before the reconstructed instances are recomposed into the original scene. To evaluate our approach, we construct a real-world dense LiDAR benchmark of underground parking environments. Experimental results demonstrate consistent improvements over representative scene-level baselines in both quantitative metrics and visual quality. The completed point cloud differs substantially from the measured input (average KL divergence ~ 2.1), yet reducing the input to only 1% of the original LiDAR measurements changes the completed reconstruction only marginally (KL divergence < 0.50). This demonstrates that ZVeC produces geometrically consistent completions even under extreme input sparsity.
Chinese Translation
在地下环境中获取的激光雷达（LiDAR）点云由于遮挡和传感器视角有限，表现出严重的几何不完整性，这使得在没有大规模监督数据集的情况下，可靠的点云补全变得具有挑战性。我们提出了ZVeC，一个零样本、实例驱动的框架，将场景级补全重新表述为组合对象级重建。通过将场景分解为语义对象实例，ZVeC减少了在杂乱环境中的重建模糊性，同时消除了对特定场景训练的需求。每个分割出的车辆都使用一个基于深度和3D高斯条件的扩散模型独立进行补全，该模型利用了广义几何先验，然后将重建的实例重新组合成原始场景。为了评估我们的方法，我们构建了一个真实世界的地下停车环境密集LiDAR基准。实验结果表明，在定量指标和视觉质量上，相较于代表性的场景级基线，我们的方法在各方面均有一致的提升。补全后的点云与测量输入之间存在显著差异（平均KL散度约为2.1），然而将输入减少至原始LiDAR测量的仅1%时，补全重建的变化仅为微小（KL散度<0.50）。这表明ZVeC即使在极端输入稀疏的情况下也能产生几何一致的补全结果。
cs.CV / 210 / 2609.24839
When Wider Views Fail: Stress-Testing Feed-Forward 3D Reconstruction
当广泛视角失效时：前馈式三维重建的压力测试
Abstract
Feed-forward 3D reconstruction models enable efficient geometry estimation from sparse images, but their pretrained nature can make them vulnerable to distribution shifts beyond their training data. Identifying these failure modes is important for understanding when such models can be reliably deployed in unconstrained imaging settings. We investigate viewpoint variation as a controlled distribution shift by varying the angular span of sparse image inputs while keeping the input budget fixed. Across multiple feed-forward reconstruction models, we observe substantial degradation as viewpoint span increases, with wide spans producing both incomplete surface coverage and geometry unsupported by the observed imagery. These results reveal that viewpoint variation can induce failure modes beyond conventional reconstruction incompleteness, highlighting the need to evaluate pretrained feed-forward models under distribution shifts that challenge their learned geometric priors.
Chinese Translation
前馈式三维重建模型能够从稀疏图像中高效估计几何形状，但其预训练特性使其在超出训练数据的分布变化下变得脆弱。识别这些失效模式对于理解此类模型在无约束成像环境中何时可以可靠部署至关重要。我们通过在保持输入预算固定的情况下，改变稀疏图像输入的角度范围，研究视角变化作为一种受控的分布变化。在多个前馈式重建模型中，我们观察到随着视角范围的增加，性能显著下降，宽广的视角范围导致表面覆盖不完整以及几何形状无法被观察到的图像所支持。这些结果揭示了视角变化可以引发超出传统重建不完整性的失效模式，强调了在挑战其学习几何先验的分布变化下评估预训练前馈模型的必要性。
cs.CV / 211 / 2609.24850
Revisiting Multi-View Stereo: A Sequence-to-Sequence Formulation
重新审视多视图立体视觉：序列到序列的表述
Abstract
Computing accurate geometry from multi-view images is a fundamental problem in computer vision. Recent feed-forward (FF) models jointly estimate 3D geometry and camera parameters, but they typically suffer from geometry distortion caused by reconstruction ambiguity, even when ground-truth camera parameters are supplied. In this paper, we study the multi-view stereo (MVS) problem with known camera parameters and propose a novel approach that bridges conventional MVS and FF methods. Rather than casting MVS as a sequence-to-one mapping that predicts depth only for a single reference view, we reformulate it as a sequence-to-sequence task, akin to FF models, that jointly predicts geometry for all input views. We introduce a global transformer-based architecture with two components that explicitly exploit camera-induced priors: ray-map embeddings that inject camera parameters into image patch tokens, making the transformer camera-aware, and a unified global cost volume that replaces conventional per-view cost volumes to jointly capture 3D structure across all views. Extensive experiments on multiple public benchmarks show our approach achieves state-of-the-art performance, surpassing both MVS and FF reconstruction baselines.
Chinese Translation
从多视图图像中计算准确的几何形状是计算机视觉中的一个基本问题。最近的前馈（FF）模型联合估计三维几何形状和相机参数，但即使在提供真实相机参数的情况下，它们通常也会受到重建歧义导致的几何失真。在本文中，我们研究了已知相机参数的多视图立体视觉（MVS）问题，并提出了一种新颖的方法，将传统的MVS与FF方法连接起来。我们将MVS重新表述为序列到序列的任务，类似于FF模型，联合预测所有输入视图的几何形状，而不是将其视为仅为单一参考视图预测深度的序列到单一映射。我们引入了一种基于全局变换器的架构，包含两个组件，明确利用相机引起的先验：射线映射嵌入将相机参数注入图像块令牌，使变换器具备相机感知能力，以及一个统一的全局代价体积，取代传统的每视图代价体积，以联合捕捉所有视图的三维结构。在多个公共基准上的广泛实验表明，我们的方法达到了最先进的性能，超越了MVS和FF重建基线。
cs.CV / 212 / 2609.24872
DTKDP: A Dual Teacher Knowledge Distillation and Pruning Framework for Lightweight Oriented SAR Ship Detection
DTKDP：一种面向轻量化合成孔径雷达船舶检测的双教师知识蒸馏与剪枝框架
Abstract
Two-stage oriented detectors achieve high localization accuracy in synthetic aperture radar (SAR) ship detection, but their large backbones, feature pyramids, proposal modules, and heavy region of interest (RoI) heads hinder deployment. Existing lightweight SAR ship detectors typically use one-stage frameworks that lack proposal-level refinement for precise rotated localization. This paper presents a dual-teacher knowledge distillation and pruning (DTKDP) framework for lightweight oriented SAR ship detection. DTKDP introduces learnable gates into convolutional, normalization, and linear layers to prune convolutional channels and RoI-head neurons. Rotated proposal alignment (RPA) distills teacher and student predictions in a shared teacher-generated rotated proposal space, while a dual-teacher scheme combines classification and regression guidance from a homogeneous main teacher with complementary classification cues from a heterogeneous auxiliary teacher. Experiments on the SAR Ship Detection Dataset (SSDD) and Rotated Ship Detection Dataset in SAR Images (RSDD-SAR) show that DTKDP reduces the parameters of Oriented Region-based Convolutional Neural Network (Oriented R-CNN) and RoI Transformer equipped with ResNet-50 backbones by 87.5-91.8% and their floating-point operations (FLOPs) by 75.6-79.9%. In terms of average precision (AP) and mean average precision (mAP), the resulting Oriented R-CNN-slim and RoI Transformer-slim retain accuracy close to their full-scale counterparts. Relative changes across $\mathrm{AP}_{50}$, $\mathrm{AP}_{75}$, $\mathrm{mAP}_{50:75}$, and $\mathrm{mAP}_{50:95}$ range from a 2.38% decrease to a 0.65% improvement. Compared with RTMDet-tiny, they improve all four metrics on both datasets by 0.52-27.55% and consistently surpass representative distillation methods, demonstrating a favorable accuracy-efficiency trade-off.
Chinese Translation
两阶段定向检测器在合成孔径雷达（SAR）船舶检测中实现了高定位精度，但其庞大的主干网络、特征金字塔、提议模块和重型兴趣区域（RoI）头部阻碍了部署。现有的轻量化SAR船舶检测器通常采用单阶段框架，缺乏提议级别的精细化以实现精确的旋转定位。本文提出了一种双教师知识蒸馏与剪枝（DTKDP）框架，用于轻量化定向SAR船舶检测。DTKDP在卷积、归一化和线性层中引入可学习的门控，以剪枝卷积通道和RoI头神经元。旋转提议对齐（RPA）在共享的教师生成的旋转提议空间中蒸馏教师和学生的预测，而双教师方案则结合了来自同质主教师的分类和回归指导以及来自异质辅助教师的互补分类线索。在SAR船舶检测数据集（SSDD）和旋转船舶检测数据集（RSDD-SAR）上的实验表明，DTKDP将基于定向区域的卷积神经网络（Oriented R-CNN）和配备ResNet-50主干的RoI Transformer的参数减少了87.5-91.8%，其浮点运算（FLOPs）减少了75.6-79.9%。在平均精度（AP）和均值平均精度（mAP）方面，得到的Oriented R-CNN-slim和RoI Transformer-slim的准确性接近其全尺度对应物。相对变化在$ ext{AP}_{50}$、$ ext{AP}_{75}$、$ ext{mAP}_{50:75}$和$ ext{mAP}_{50:95}$之间范围为2.38%的下降到0.65%的提升。与RTMDet-tiny相比，它们在两个数据集上均提高了所有四个指标0.52-27.55%，并始终超越了代表性的蒸馏方法，展示了良好的准确性与效率的权衡。
cs.CV / 213 / 2609.24875
SPHQuant: Efficient extreme low bit weight quantization for Vision-Language Models
SPHQuant：用于视觉-语言模型的高效极低比特权重量化
Abstract
Recent foundation models are moving toward native multimodal Vision-Language Models (VLMs), making VLMs a central form of next-generation foundation models. However, their large language backbones make edge deployment difficult due to high memory footprint and memory-bound autoregressive decoding. Weight-only post-training quantization is a practical solution, but pushing VLMs to extreme low bit-widths remains challenging: existing rotation-free methods suffer from outliers at 2-3 bits, while rotation-based methods improve accuracy at the cost of additional runtime overhead. We propose SPHQuant, a rotation-free spherical weight-only quantization framework for VLMs. Instead of quantizing weights directly in Cartesian coordinates, SPHQuant decomposes each 8D weight vector into coordinate signs, radius, and a positive unit direction. This representation isolates outlier magnitude into the radius while keeping directions bounded and statistically regular. Based on this insight, SPHQuant allocates extra precision to the radius to mitigate accuracy degradation induced by outliers. It further uses a compact positive-direction codebook and fine-tunes codebook entries through angular parameterization to preserve the unit-sphere constraint. We also design a hardware-friendly GEMV kernel that keeps the direction codebook small enough for shared-memory lookup and packs radial bits efficiently. Experiments show that SPHQuant matches the performance of state-of-the-art extreme low-bit quantization methods while improving decode throughput over QTIP by 30.3% on RTX A6000. Code will be released in https://github.com/Pushazf/SPHQuant.
Chinese Translation
近期的基础模型正朝着原生的多模态视觉-语言模型（VLMs）发展，使得VLMs成为下一代基础模型的核心形式。然而，它们庞大的语言骨干网络使得边缘部署变得困难，因为高内存占用和受限于内存的自回归解码。仅基于权重的后训练量化是一种实用的解决方案，但将VLMs推向极低比特宽度仍然具有挑战性：现有的无旋转方法在2-3比特时受到异常值的影响，而基于旋转的方法在增加额外运行时开销的情况下提高了准确性。我们提出了SPHQuant，一种针对VLMs的无旋转球面权重量化框架。SPHQuant并不是直接在笛卡尔坐标系中量化权重，而是将每个8D权重向量分解为坐标符号、半径和正单位方向。这种表示将异常值的幅度隔离到半径中，同时保持方向的有界性和统计规律性。基于这一见解，SPHQuant为半径分配额外的精度，以减轻异常值引起的准确性下降。它进一步使用紧凑的正方向码本，并通过角参数化微调码本条目，以保持单位球体约束。我们还设计了一个硬件友好的GEMV内核，使方向码本足够小以便于共享内存查找，并高效地打包径向比特。实验表明，SPHQuant的性能与最先进的极低比特量化方法相当，同时在RTX A6000上提高了30.3%的解码吞吐量。代码将发布在https://github.com/Pushazf/SPHQuant。
cs.CV / 214 / 2609.24879
Generating Chest X-Ray Counterfactuals by Specialising Foundation Image Models
通过专业化基础图像模型生成胸部X光反事实图像
Abstract
Counterfactual image generation answers questions about how a subject would have looked under retrospective, hypothetical scenarios. Recent methods have improved perceptual quality, identity preservation and faithfulness to an underlying causal model, but their adoption in healthcare is limited by scarce annotated data, distribution shift between datasets, and mismatches between pretrained generative models and those required for counterfactual inference. We propose specialisation, a data and parameter-efficient framework for adapting pretrained, non-causal generative models into causal mechanisms under distribution shift. Based on this framework, we train a radiology counterfactual image generation model, called RadCF, using latent flow matching. We validate our approach on three chest X-ray datasets spanning different dataset shifts, data volumes, and counterfactual questions, associated with challenging, highly-localised interventions. Our results show that RadCF and specialisation improve counterfactual soundness over existing methods while being data and parameter efficient, and that the resulting counterfactuals can detect and mitigate shortcut learning in a downstream medical classifier. Code is available at https://github.com/GSK-AI/RadCF/.
Chinese Translation
反事实图像生成回答了在回顾性假设场景下，主体的外观会如何的问题。近期的方法在感知质量、身份保留和对基础因果模型的忠实度方面有所改善，但在医疗保健中的应用受到稀缺标注数据、数据集之间的分布变化以及预训练生成模型与反事实推断所需模型之间的不匹配的限制。我们提出了一种专业化框架，这是一种数据和参数高效的方案，用于在分布变化下将预训练的非因果生成模型适应为因果机制。基于此框架，我们使用潜在流匹配训练了一个放射学反事实图像生成模型，称为RadCF。我们在三个胸部X光数据集上验证了我们的方法，这些数据集涵盖了不同的数据集变化、数据量和与具有挑战性的高度局部化干预相关的反事实问题。我们的结果表明，RadCF和专业化在反事实合理性方面优于现有方法，同时在数据和参数效率上表现良好，并且生成的反事实能够检测和缓解下游医疗分类器中的捷径学习。代码可在 https://github.com/GSK-AI/RadCF/ 获取。
cs.CV / 215 / 2609.24894
SLICEChat: Progressive In-Encoder Token Pruning for Whole-Slide Pathology Language Models
SLICEChat：用于全幻灯片病理语言模型的渐进式编码器令牌修剪
Abstract
Whole-slide pathology images (WSIs) contain gigapixel-scale visual content, creating a major scalability challenge for slide-level multimodal large language models (MLLMs). Existing approaches process thousands of patch tokens and typically apply compression only after slide encoding, leaving multimodal attention computationally expensive. We introduce SLICEChat, a slide-level MLLM that integrates progressive token pruning within a hybrid Mamba--Transformer slide encoder. Mamba layers enable efficient long-range propagation, while Transformer layers preserve global interactions as the sequence is progressively shortened. Between stages, language-supervised, region-aware pruning removes spatially coherent low-utility regions under a controlled keep-rate schedule, producing compact slide representations before multimodal fusion. On SlideBench VQA, SLICEChat achieves 79.84% accuracy on TCGA and 59.09% on BCNB cohorts, outperforming prior slide-level pathology MLLMs, and achieves the highest overall WSI-Bench metrics. It also provides competitive memory usage and the inference latency among the evaluated models. These results demonstrate accurate and computationally efficient multimodal reasoning over gigapixel WSIs.
Chinese Translation
全幻灯片病理图像（WSIs）包含千兆像素级别的视觉内容，这给幻灯片级别的多模态大型语言模型（MLLMs）带来了重大的可扩展性挑战。现有的方法处理数千个补丁令牌，通常仅在幻灯片编码后应用压缩，导致多模态注意力计算成本高昂。我们提出了SLICEChat，这是一种在混合Mamba-Transformer幻灯片编码器中集成渐进式令牌修剪的幻灯片级别MLLM。Mamba层实现了高效的长距离传播，而Transformer层在序列逐渐缩短的过程中保留了全局交互。在各个阶段之间，语言监督的区域感知修剪在受控的保留率计划下去除空间上连贯的低效区域，从而在多模态融合之前生成紧凑的幻灯片表示。在SlideBench VQA上，SLICEChat在TCGA队列上达到了79.84%的准确率，在BCNB队列上达到了59.09%，超越了之前的幻灯片级别病理MLLMs，并在WSI-Bench指标中获得了最高的整体表现。它还在评估的模型中提供了具有竞争力的内存使用和推理延迟。这些结果展示了对千兆像素WSIs的准确且计算高效的多模态推理。
cs.CV / 216 / 2609.24919
PixelDiT2: Representation-Grounded Pixel Diffusion Transformers
PixelDiT2：基于表示的像素扩散变换器
Abstract
Recent advances in pixel-space diffusion models have narrowed the image quality gap with latent-space diffusion, but still converge more slowly and lag behind in final image quality. We argue that a key reason is the lack of an explicit representation prior: unlike latent diffusion, which usually denoises in a compact and structured latent space, pixel diffusion needs to learn denoising-friendly representations and pixel generation simultaneously from raw RGB space. To address this problem, we propose PixelDiT2, an end-to-end pixel-space diffusion model designed to decouple representation learning from pixel generation without introducing an autoencoder or latent reconstruction bottleneck. We propose representation grounding that uses a frozen pretrained vision foundation model to provide explicit per-patch representation guidance throughout denoising, allowing the pixel diffusion transformer to focus more on pixel generation. On ImageNet-256x256, PixelDiT2 achieves an FID of 1.46 after 600 epochs; at 512x512 resolution, PixelDiT2 achieves an FID of 1.48 after 680 epochs.
Chinese Translation
近年来，像素空间扩散模型的进展缩小了与潜在空间扩散的图像质量差距，但仍然收敛较慢，最终图像质量落后。我们认为，关键原因在于缺乏明确的表示先验：与通常在紧凑且结构化的潜在空间中去噪的潜在扩散不同，像素扩散需要同时从原始 RGB 空间学习去噪友好的表示和像素生成。为了解决这个问题，我们提出了 PixelDiT2，这是一种端到端的像素空间扩散模型，旨在在不引入自编码器或潜在重建瓶颈的情况下，将表示学习与像素生成解耦。我们提出了表示基础，该方法利用一个冻结的预训练视觉基础模型，在去噪过程中提供明确的每个补丁的表示指导，使像素扩散变换器能够更专注于像素生成。在 ImageNet-256x256 上，PixelDiT2 在 600 个周期后达到了 1.46 的 FID；在 512x512 分辨率下，PixelDiT2 在 680 个周期后达到了 1.48 的 FID。
cs.CV / 217 / 2609.24937
Anatomy-Decomposed Chest Computed Tomography (CT) Projections as Scalable Supervision for Bone Suppression in Chest Radiographs
解剖分解胸部计算机断层扫描 (CT) 投影作为胸部X光片骨骼抑制的可扩展监督
Abstract
Bone overlap can obscure abnormalities in chest radiographs, while scarce paired training data limit supervised bone suppression. We address this challenge with a digitally reconstructed radiograph (DRR) framework that converts chest computed tomography (CT) into paired supervision for component suppression. A novel bone segmentation algorithm enables CT decomposition into bone, non-lung soft-tissue, and lung components, which are projected separately. Their weighted combination yields synthetic radiographs with pixel-registered component images that sum exactly to the full DRR. Models trained on these data suppress bone or lung components by predicting the target component and recovering the remainder by subtraction, transferring to real radiographs without real paired training data. As an extension, their outputs on real radiographs provide target domains for unpaired, component-wise DRR translation, reducing the appearance gap while retaining anatomical details. Across multiple public datasets, downstream detection experiments demonstrate the utility of bone suppression, with gains concentrated on abnormalities with substantial bone overlap. Compared with open-source DRR engines applied to the same CTs, our unmodified DRRs achieve comparable realism and preservation of label-relevant anatomy, while translated DRRs achieve the best Fr\'echet inception distance (FID), lung-field sharpness, and agreement with source-CT anatomy among the evaluated methods. Models and inference code: https://huggingface.co/qureaiorg/bone-suppression; Translated projections: https://huggingface.co/datasets/qureaiorg/ct2xr-projections.
Chinese Translation
骨骼重叠可能会掩盖胸部X光片中的异常，而稀缺的配对训练数据限制了监督式骨骼抑制。我们通过一个数字重建X光片 (DRR) 框架来解决这一挑战，该框架将胸部计算机断层扫描 (CT) 转换为组件抑制的配对监督。一个新颖的骨骼分割算法使得CT能够分解为骨骼、非肺软组织和肺组件，并分别进行投影。它们的加权组合产生合成X光片，具有像素注册的组件图像，精确地与完整的DRR相加。训练于这些数据的模型通过预测目标组件并通过减法恢复其余部分来抑制骨骼或肺组件，从而在没有真实配对训练数据的情况下转移到真实X光片。作为扩展，它们在真实X光片上的输出为无配对的、逐组件DRR转换提供了目标域，减少了外观差距，同时保留了解剖细节。在多个公共数据集上，下游检测实验展示了骨骼抑制的实用性，增益集中在具有显著骨骼重叠的异常上。与应用于相同CT的开源DRR引擎相比，我们未修改的DRR在现实感和标签相关解剖的保留方面达到了可比水平，而翻译后的DRR在评估的方法中实现了最佳的Fréchet起始距离 (FID)、肺野清晰度和与源CT解剖的一致性。模型和推理代码： https://huggingface.co/qureaiorg/bone-suppression；翻译后的投影： https://huggingface.co/datasets/qureaiorg/ct2xr-projections。
cs.CV / 218 / 2609.24981
GAE: Learning a Geometry-Native Latent Space for 3D-Consistent World Generation
GAE：学习用于三维一致性世界生成的几何本地潜在空间
Abstract
We present a compact geometry-native latent space as a shared foundation for perception and generation. Visual generators can produce photorealistic frames without preserving a consistent 3D scene. We argue that this is not only a modeling problem but also a representation problem: generators typically evolve appearance-centric latents, while perception models recover geometry in a semantically rich space that encodes cross-view structure. Rather than adding geometry as another output, we reparameterize a geometry foundation model's features into a compact latent space for generation. We realize this shift with the geometry-native autoencoder (GAE), whose latent is jointly decodable to appearance, depth, cameras, and point maps. With this state, a standard conditional flow supports diverse generation tasks. In controlled comparisons that hold the generator and training protocol fixed, replacing the latent with GAE improves both visual quality and independently measured 3D coherence: FVD falls by $12.7\%$ and $23.1\%$ on RealEstate10K and DL3DV, and camera-trajectory error is halved on RealEstate10K. Together, these results show that the latent space is central to geometry-consistent generation and can serve as a shared interface between perception and generation.
Chinese Translation
我们提出了一种紧凑的几何本地潜在空间，作为感知与生成的共享基础。视觉生成器可以生成逼真的图像，但并不保持一致的三维场景。我们认为这不仅是一个建模问题，也是一个表示问题：生成器通常演化以外观为中心的潜在变量，而感知模型则在一个语义丰富的空间中恢复几何信息，该空间编码了跨视角的结构。我们并不是将几何作为另一个输出添加，而是将几何基础模型的特征重新参数化为一个紧凑的潜在空间以用于生成。我们通过几何本地自编码器（GAE）实现这一转变，其潜在变量可以共同解码为外观、深度、相机和点图。在这种状态下，标准条件流支持多样的生成任务。在控制比较中，固定生成器和训练协议，使用GAE替换潜在变量提高了视觉质量和独立测量的三维一致性：在RealEstate10K和DL3DV上，FVD分别下降了12.7%和23.1%，而在RealEstate10K上的相机轨迹误差减半。这些结果表明，潜在空间对于几何一致性生成至关重要，并可以作为感知与生成之间的共享接口。
cs.CV / 219 / 2609.24984
WorldCrafter: Consistent Video World Model with Implicit 3D-aware Memory
WorldCrafter：具有隐式3D感知记忆的一致视频世界模型
Abstract
Video world models enable interactive exploration of dynamic environments, yet struggle to respect prior observations over long horizons and across viewpoints. We present WorldCrafter, a video world model that learns a camera-queryable implicit 3D-aware memory for this purpose. The key insight is to let the requested viewpoint shape how multi-view evidence is compressed into the video generator's limited token budget. Trained jointly with the video generator, a memory encoder and pose-conditioned readout module integrate historical observations into a fixed set of target view-specific tokens before denoising, without explicit depth-based correspondences. By combining this memory with recent temporal context and few-step distillation, WorldCrafter enables streaming scene exploration from a single input image or text prompt. Experiments across static and dynamic scenes show substantial gains in long-horizon consistency and camera-control accuracy while preserving visual quality during minute-scale exploration.
Chinese Translation
视频世界模型使得动态环境的交互探索成为可能，但在长时间跨度和不同视角下，仍然难以保持对先前观察的尊重。我们提出了WorldCrafter，这是一种视频世界模型，旨在学习可通过相机查询的隐式3D感知记忆。其关键见解在于让请求的视角决定多视角证据如何被压缩到视频生成器的有限令牌预算中。与视频生成器共同训练的记忆编码器和姿态条件读取模块在去噪之前，将历史观察整合到一组固定的目标视角特定令牌中，而无需显式的基于深度的对应关系。通过将此记忆与最近的时间上下文和少步蒸馏相结合，WorldCrafter能够从单个输入图像或文本提示中实现流式场景探索。在静态和动态场景中的实验表明，在长时间跨度的一致性和相机控制精度上有显著提升，同时在分钟级探索过程中保持了视觉质量。
cs.CV / 220 / 2609.24997
VideoGen-Agent: Reinforcing Video Generation Agents
VideoGen-Agent：强化视频生成代理
Abstract
Recent advances in video generative models have enabled high-fidelity, temporally coherent video generation. However, these models often struggle to satisfy prompts requiring specialized knowledge, specific identities, physical consistency, or ordered events. In this paper, we present VideoGen-Agent, a multimodal agent trained through multitask agentic reinforcement learning to use external tools for video generation. The agent coordinates augmentation, generation, and verification tools through multi-turn interactions, using the prompt and intermediate observations to guide its decisions. We train a shared policy on a category-balanced dataset spanning six tasks. Supervised fine-tuning on teacher-generated trajectories establishes tool-use behavior, which is then refined through reinforcement learning. A category-aware hybrid reward evaluates tool-call validity, task-appropriate tool use, and generated video quality. We further introduce VABench, a held-out benchmark of 600 prompts covering procedural knowledge, single- and multi-entity identity preservation, physical consistency, scene composition, and multi-shot temporal structure. On VABench, VideoGen-Agent improves over its base text-to-video generator by 19.1 points, from 56.5 to 75.6. Upgrading the generation tools further raises the score to 86.1 without additional agent training. Human raters prefer the upgraded configuration over the strongest standalone baseline in 84.3% of comparisons. These results support learning tool use across video-generation tasks and show that the trained agent can benefit from subsequent advances in generation tools.
Chinese Translation
最近在视频生成模型方面的进展使得高保真、时间一致的视频生成成为可能。然而，这些模型通常难以满足需要专业知识、特定身份、物理一致性或有序事件的提示。在本文中，我们提出了VideoGen-Agent，这是一种通过多任务代理强化学习训练的多模态代理，旨在使用外部工具进行视频生成。该代理通过多轮交互协调增强、生成和验证工具，利用提示和中间观察结果来指导其决策。我们在一个涵盖六个任务的类别平衡数据集上训练了一个共享策略。对教师生成轨迹的监督微调建立了工具使用行为，随后通过强化学习进行优化。类别感知的混合奖励评估工具调用的有效性、任务适当的工具使用以及生成视频的质量。我们进一步引入了VABench，这是一个包含600个提示的保留基准，涵盖程序知识、单一和多实体身份保留、物理一致性、场景构图和多镜头时间结构。在VABench上，VideoGen-Agent的表现比其基础文本到视频生成器提高了19.1分，从56.5提升至75.6。升级生成工具进一步将得分提高至86.1，而无需额外的代理训练。在84.3%的比较中，人类评审者更倾向于升级后的配置，而不是最强的独立基线。这些结果支持在视频生成任务中学习工具使用，并表明训练后的代理可以从后续生成工具的进步中受益。
cs.CV / 221 / 2609.25001
GameHorizon Suite: Multi-Horizon Data and Evaluation in Gameplay
GameHorizon 套件：游戏中的多时间跨度数据与评估
Abstract
Modern video games provide a measurable testbed for AI models, combining abilities of visual understanding, instruction decomposition, goal planning, and precise action control over multiple temporal horizons. Existing datasets and benchmarks, however, either cover a narrow range of games, lack language instructions, or rely on high-variance online rollouts. To address these challenges, we introduce GameHorizon, a unified data and evaluation suite that measures gameplay capabilities at different horizons for diverse model families. GameHorizon Suite consists of three components. First, GameHorizon-Annotator is a scalable and automated annotation pipeline for multi-horizon instructions. Second, utilizing the pipeline, we construct GameHorizon-Data, the first large-scale AAA gameplay dataset with temporally aligned videos, player actions, and multi-horizon instructions. It comprises 5,000 hours of recordings from 21 games, collected by 100 human expert players. Third, we build GameHorizon-Bench with reproducible offline and stepwise online testing. The offline track enables reproducible evaluation using thousands of standardized questions organized into three primary tasks and a series of diagnostic variants, while the online track tests whether offline scores reflect actual gameplay capabilities and localizes failures to specific steps within long-horizon gameplay. Based on our GameHorizon Suite, we evaluate 47 models through more than one million model invocations, revealing a meaningful hierarchy of task difficulty and pronounced differences in model capabilities. Our work can provide a standardized yardstick for evaluating gameplay capabilities across horizons and model families. We will release our dataset, annotator, and benchmark to facilitate future research.
Chinese Translation
现代视频游戏为人工智能模型提供了一个可测量的测试平台，结合了视觉理解、指令分解、目标规划和对多个时间跨度的精确动作控制的能力。然而，现有的数据集和基准要么覆盖的游戏范围较窄，要么缺乏语言指令，或者依赖于高方差的在线回放。为了解决这些挑战，我们引入了 GameHorizon，一个统一的数据与评估套件，用于测量不同模型家族在不同时间跨度上的游戏能力。GameHorizon 套件由三个部分组成。首先，GameHorizon-Annotator 是一个可扩展的自动化注释管道，用于多时间跨度指令。其次，利用该管道，我们构建了 GameHorizon-Data，这是第一个大规模 AAA 游戏数据集，包含时间对齐的视频、玩家动作和多时间跨度指令。该数据集包含来自 21 款游戏的 5000 小时录音，由 100 名人类专家玩家收集。第三，我们建立了 GameHorizon-Bench，提供可重复的离线和逐步在线测试。离线轨道通过成千上万的标准化问题组织成三项主要任务和一系列诊断变体，实现可重复的评估，而在线轨道则测试离线得分是否反映实际游戏能力，并将失败定位到长时间跨度游戏中的特定步骤。基于我们的 GameHorizon 套件，我们通过超过一百万次模型调用评估了 47 个模型，揭示了任务难度的有意义层次和模型能力的显著差异。我们的工作可以为跨时间跨度和模型家族评估游戏能力提供一个标准化的基准。我们将发布我们的数据集、注释器和基准，以促进未来的研究。
cs.AI / 1 / 2609.22161
Didactic knowledge or Clinical Cases? How Data Types Shape Medical Large Language Models
教学知识还是临床案例？数据类型如何影响医学大语言模型
Abstract
Medical large language models are commonly trained on mixtures of didactic data (e.g., textbooks) and clinical data (e.g., patient records), yet how these data types differentially shape model capabilities remains unclear. We address this issue with token-matched experiments that vary the didactic-to-clinical ratio and analyze how data composition affects performance, capability profiles, and error patterns across knowledge-intensive and clinic-oriented tasks. We uncover an asymmetric transfer across task types: clinical data improves clinic-oriented tasks while remaining competitive on knowledge-intensive ones, whereas didactic data mainly improves knowledge-intensive tasks. Error analysis suggests a knowing-doing gap, where improvements in knowledge recall do not reliably generalize to clinical reasoning. We further observe that modest amounts of clinical data yield most of the gains on EHR-grounded tasks, while the optimal mixture ratio varies with the knowledge and clinical reasoning demands of downstream tasks. These findings suggest that medical LLM data curation should be application-driven, with higher proportions of clinical data preferred for reasoning-intensive use cases.
Chinese Translation
医学大语言模型通常在教学数据（例如教科书）和临床数据（例如患者记录）的混合上进行训练，但这些数据类型如何不同地影响模型能力仍不清楚。我们通过令牌匹配实验来解决这一问题，改变教学与临床数据的比例，并分析数据组成如何影响知识密集型和临床导向任务的性能、能力特征和错误模式。我们发现任务类型之间存在不对称的迁移：临床数据改善了临床导向任务，同时在知识密集型任务中保持竞争力，而教学数据主要改善知识密集型任务。错误分析表明存在一个知识与实践之间的差距，知识回忆的改善并未可靠地推广到临床推理中。我们进一步观察到，适量的临床数据在以电子健康记录（EHR）为基础的任务中带来了大部分收益，而最佳混合比例则随着下游任务的知识和临床推理需求而变化。这些发现表明，医学大语言模型的数据整理应以应用为驱动，对于推理密集型用例，建议优先使用更高比例的临床数据。
cs.AI / 2 / 2609.22277
An Affordable AI-Integrated Smart Cane for Multimodal Mobility Assistance of Visually Impaired Users
一种经济实惠的AI集成智能手杖，用于视障用户的多模态移动辅助
Abstract
Visual impairment affects over 2.2 billion people worldwide, yet conventional white canes cannot detect elevated hazards or provide semantic environmental context. Existing AI-assisted navigation systems typically rely on expensive hardware or cloud connectivity, limiting accessibility in resource-constrained settings. This paper presents an affordable (\$88 USD), fully offline AI-integrated smart cane designed for multimodal mobility assistance on an ultra-low-power Raspberry Pi Zero 2W. The system fuses RGB vision sensing with Time-of-Flight (ToF) distance estimation, pairing an INT8-quantized SSD MobileNet V1 model with distance-aware vibrotactile feedback and real-time audio alerts. To ensure operational robustness on constrained hardware, a multiprocessing architecture isolates sensor acquisition, neural inference, and haptic feedback into independent processes with fail-safe sensing support. Experimental evaluation across indoor mobility scenarios demonstrates a macro-averaged F1-score of 0.82 (precision: 0.85, recall: 0.81), a mean end-to-end latency of 330\,ms, and a peak power draw of 2.8\,W. A preliminary usability study with 12 participants (SUS: 78.5, NASA-TLX) demonstrated positive user perception and enhanced obstacle awareness. The proposed prototype validates the feasibility of deploying privacy-preserving, edge-native assistive intelligence for cost-sensitive mobility assistance.
Chinese Translation
全球有超过22亿人受到视觉障碍的影响，然而传统的白手杖无法检测到高度危险或提供语义环境上下文。现有的AI辅助导航系统通常依赖昂贵的硬件或云连接，这限制了在资源受限环境中的可及性。本文提出了一种经济实惠（88美元），完全离线的AI集成智能手杖，旨在为超低功耗的Raspberry Pi Zero 2W提供多模态移动辅助。该系统将RGB视觉传感与飞行时间（ToF）距离估计相结合，配备INT8量化的SSD MobileNet V1模型，提供基于距离的振动反馈和实时音频警报。为了确保在受限硬件上的操作稳健性，采用多进程架构将传感器采集、神经推理和触觉反馈隔离为独立进程，并支持故障安全传感。针对室内移动场景的实验评估显示，宏观平均F1分数为0.82（精确度：0.85，召回率：0.81），平均端到端延迟为330毫秒，峰值功耗为2.8瓦。对12名参与者的初步可用性研究（SUS：78.5，NASA-TLX）显示出积极的用户感知和增强的障碍物意识。所提出的原型验证了在成本敏感的移动辅助中部署保护隐私的边缘原生辅助智能的可行性。
cs.AI / 3 / 2609.22353
PAANI : On Device Visual Evidence Fusion and Explainable Guidance for River Robot Simulation
PAANI：设备端视觉证据融合与可解释指导的河流机器人仿真
Abstract
Mobile river monitoring robots must interpret obstacles and water boundaries that geographic waypoints alone cannot describe. On resource constrained platforms, converting imperfect visual predictions into timely and inspectable guidance is a distinct challenge. An object label or steering command does not explain which evidence supports a decision or when that evidence is unreliable. We present PAANI, an on-device perception to guidance architecture that combines a project trained YOLO11n detector and a custom MobileNetV3 Small semantic segmenter with timestamp aligned evidence fusion on Arduino UNO Q. Bounded tracking supplies object persistence, while an explicit corridor policy combines surface labels, accepted detections, urgency and mask uncertainty. Each final advisory exposes its contributing evidence and policy reasons. ROS 2 interfaces connect the local AI pipeline to a separate Gazebo vessel, localization and control testbed. Training uses 10,000 WaterScenes images for four-class detection and 1,127 MaSTr1325 images for segmentation, including 198 segmentation validation images. The selected FP32 ONNX models occupy 14.817 MB. Detector checkpoint test mAP at 0.5 IoU is 0.7388, while the separately evaluated rectangular ONNX export achieves validation mAP at 0.5 IoU of 0.7367. Segmentation ONNX validation mIoU is 0.9750. A five-minute UNO Q recording produced median and 95th percentile pipeline latencies of 467.8 ms and 580.3 ms at a configured 0.5 Hz cadence. The evaluation also identifies black input misclassification and a sampling rate mismatch that prevents the diagnostic apparent motion estimator from collecting sufficient evidence. These results support an inspectable and reusable edge robotics foundation while clearly distinguishing model accuracy and on-board execution from validated on-water collision avoidance.
Chinese Translation
移动河流监测机器人必须解释地理航点无法描述的障碍物和水域边界。在资源受限的平台上，将不完美的视觉预测转化为及时且可检查的指导是一项独特的挑战。一个物体标签或转向指令并不能解释支持决策的证据是什么，或者何时该证据不可靠。我们提出了PAANI，一种设备端感知到指导的架构，结合了经过项目训练的YOLO11n检测器和自定义的MobileNetV3 Small语义分割器，并在Arduino UNO Q上进行时间戳对齐的证据融合。边界跟踪提供了物体的持续性，而显式走廊策略结合了表面标签、接受的检测、紧急程度和掩码不确定性。每个最终建议都揭示了其贡献证据和政策原因。ROS 2接口将本地AI管道连接到一个独立的Gazebo船只、定位和控制测试平台。训练使用了10,000张WaterScenes图像进行四类检测，以及1,127张MaSTr1325图像进行分割，其中包括198张分割验证图像。所选的FP32 ONNX模型占用14.817 MB。检测器检查点测试的mAP在0.5 IoU时为0.7388，而单独评估的矩形ONNX导出在0.5 IoU时的验证mAP为0.7367。分割ONNX验证的mIoU为0.9750。五分钟的UNO Q录制产生了中位数和95百分位管道延迟分别为467.8毫秒和580.3毫秒，配置频率为0.5 Hz。评估还识别出黑色输入误分类和采样率不匹配，这阻止了诊断明显运动估计器收集足够的证据。这些结果支持了一个可检查和可重用的边缘机器人基础，同时清晰地区分了模型的准确性和机载执行与经过验证的水上碰撞避免之间的差异。
cs.AI / 4 / 2609.22408
Social Influence and the Allocation of Scientific Attention in AI Populations
社会影响与人工智能群体中科学关注的分配
Abstract
AI systems are becoming participants in the evaluation and use of scientific research. They encounter citation counts, download statistics and lists of popular articles developed around human readers, but the collective consequences of these signals for artificial readers remain uncertain. This paper adapts the Music Lab design to a market for academic attention. In the first experiment, 1,000 AI agents choose papers from the titles and abstracts of all 114 regular research articles published in the American Economic Review in 2025. The experiment has five independent-choice communities and five social-influence communities, each with 100 sequential agents. Only agents in the social-influence condition observe earlier selections within their community. Agents may select any number of papers. Social-information communities select 17.2 percent fewer papers per agent, concentrate their choices more heavily, and collectively cover 73 papers, compared with 90 independently. Between-community variation is greater under social information. In a second experiment with 200 agents across twenty social communities, randomly assigning papers five initial selections raises their subsequent selection rate by 45.55 percentage points (95% CI: 41.20 to 49.90). Choices have modest correspondence with external citations and little correspondence with download counts. The results show how a simple information rule shapes the volume, breadth and distribution of scientific attention in an artificial population.
Chinese Translation
人工智能系统正逐渐参与科学研究的评估和使用。它们会接触到围绕人类读者开发的引用次数、下载统计和热门文章列表，但这些信号对人工读者的集体影响仍然不确定。本文将音乐实验室设计适应于学术关注的市场。在第一次实验中，1000个人工智能代理从2025年《美国经济评论》发表的114篇常规研究文章的标题和摘要中选择论文。该实验设有五个独立选择社区和五个社会影响社区，每个社区有100个顺序代理。只有社会影响条件下的代理可以观察到其社区内的早期选择。代理可以选择任意数量的论文。社会信息社区每个代理选择的论文数量比独立选择社区少17.2%，选择更集中，集体覆盖73篇论文，而独立选择社区覆盖90篇。在社会信息的影响下，社区间的变异性更大。在第二次实验中，200个代理分布在20个社会社区中，随机分配论文的五个初始选择使其后续选择率提高了45.55个百分点（95%置信区间：41.20至49.90）。选择与外部引用有适度的对应关系，与下载次数的对应关系较小。结果表明，简单的信息规则如何塑造人工群体中科学关注的数量、广度和分布。
cs.AI / 5 / 2609.22410
Learning 3D biophysical cell properties from 2D images and cell-population statistics
从二维图像和细胞群体统计中学习三维生物物理细胞特性
Abstract
Inferring 3D cellular properties from 2D microscopy is difficult when a reference instrument reports only population statistics rather than labels for individual cells. Here we develop a population-supervised framework that maps single 2D red-cell images to latent biophysical quantities and aggregates them to mean corpuscular volume, red-cell distribution width and mean corpuscular haemoglobin. The model combines shared local inference, a biophysically structured decoder for volume and haemoglobin, learned instance weighting and device-specific calibration. We formalise conditions under which aggregate observations identify restricted instance predictors, show why population agreement does not by itself identify single-cell properties or 3D geometry, and derive the dispersion penalty induced by subset mean matching. The development dataset comprises 390 specimens and 1,105 acquisitions across six devices, with reported Pearson correlations of 0.86--0.98 against a Sysmex analyser. The framework provides a testable route from 2D images and population supervision to 3D cellular biophysics without claiming explicit 3D reconstruction.
Chinese Translation
从二维显微镜图像推断三维细胞特性是困难的，尤其是在参考仪器仅报告群体统计而非单个细胞标签的情况下。在此，我们开发了一种群体监督框架，将单个二维红细胞图像映射到潜在的生物物理量，并将其聚合为平均红细胞体积、红细胞分布宽度和平均红细胞血红蛋白。该模型结合了共享局部推断、用于体积和血红蛋白的生物物理结构解码器、学习的实例加权和设备特定的校准。我们形式化了聚合观察识别受限实例预测器的条件，展示了为何群体一致性本身并不能识别单细胞特性或三维几何，并推导了由子集均值匹配引起的离散惩罚。开发数据集包含390个样本和跨六个设备的1,105个采集，报告的皮尔逊相关系数为0.86至0.98，相对于Sysmex分析仪。该框架提供了一条可测试的路径，从二维图像和群体监督到三维细胞生物物理学，而不声称进行明确的三维重建。
cs.AI / 6 / 2609.22475
Goal-driven Variant Categorization
目标驱动的变体分类
Abstract
Process discovery rarely yields a single coherent process structure. For analysis, a common step is to cluster process variants based on structural similarity and then assign business meaning to the resulting groups. Since these partitions are not derived from the organization's goals, analysts must manually interpret and consolidate variants into business-meaningful categories. This judgment-intensive step becomes increasingly difficult as the number and complexity of variants grow. In this paper, we propose a goal-driven approach to variant categorization that reverses this workflow. We first author an organization's goal model that predefines the categorization axis. Each variant is transformed into a textual narrative describing its behavior, and a Large Language Model (LLM) interprets it in the context of the goal model and assigns the variant to the most appropriate category. LLM-based semantic reasoning connects low-level process behavior with analyst-defined business goals. We instantiate this approach end-to-end and evaluate it on three public logs differing substantially in scale and behavioral diversity. Goal-model guidance yields partitions that differ from those produced by unguided induction and respond to controlled edits to the declared alternatives, at the cost of authoring a goal model.
Chinese Translation
过程发现很少产生单一一致的过程结构。在分析中，一个常见的步骤是根据结构相似性对过程变体进行聚类，然后为生成的组分配业务含义。由于这些划分并不是基于组织的目标得出的，分析师必须手动解释并将变体整合为具有业务意义的类别。随着变体数量和复杂性的增加，这一判断密集的步骤变得愈加困难。本文提出了一种目标驱动的变体分类方法，逆转了这一工作流程。我们首先创建一个组织的目标模型，该模型预定义了分类轴。每个变体被转化为描述其行为的文本叙述，然后一个大型语言模型（Large Language Model, LLM）在目标模型的背景下对其进行解释，并将变体分配到最合适的类别。基于LLM的语义推理将低级过程行为与分析师定义的业务目标连接起来。我们对这一方法进行了端到端的实例化，并在三个在规模和行为多样性上有显著差异的公共日志上进行了评估。目标模型的指导产生的划分与无指导归纳所产生的划分不同，并且对声明的替代方案的受控编辑做出了响应，代价是需要创建一个目标模型。
cs.AI / 7 / 2609.22478
Replication Without Persistence in Hosted LLMs: Measurement Sensitivity in Action-Time Belief Evaluation
托管大型语言模型中的无持久性复制：行动时间信念评估中的测量敏感性
Abstract
Behavioural evaluations of hosted language models can vary because the evaluated service, the measurement instrument, or both differ across runs. We separate three validation questions: whether a prior finding recurs on fresh data under its historical configuration (replication), whether the endpoint changes when the evaluation-and-inference configuration is rebuilt under the same identifier (measurement sensitivity), and whether the finding persists across subsequently tested identifiers under one common instrument (persistence). We study these questions in Regent Chess, a sequential environment in which a hidden, mutable state is recorded exactly, allowing stated beliefs to be scored against ground truth at action time; positive endpoint values mean worse performance than a matched-uniform comparator. The previously reported Gemini 3.1 Flash-Lite deficit recurs on fresh games under its historical configuration (+0.0530, 95% CI [+0.0329,+0.0714]). In a back-to-back same-day H/R comparison under the same public identifier, the model-minus-uniform endpoint is 0.0429 lower under the rebuilt configuration (95% CI for the H-minus-R contrast [+0.0182,+0.0667]); all six configuration components vary jointly, so no component is isolated. Under rebuilt R, the prospectively frozen, interleaved same-window 4K comparison reverses sign between Gemini 3.1 and Gemini 3.7, identifiers that differ in release and product tier; additional descriptive and exploratory cells show the same directional pattern. Any additional serving-period contribution remains unresolved (-0.0166, [-0.0483,+0.0157]). Replication, measurement sensitivity, and persistence can therefore yield different conclusions within one evaluation, motivating explicit indexing of hosted-model behavioural claims by tested identifier, serving period, measurement instrument, and inference configuration.
Chinese Translation
托管语言模型的行为评估可能会因评估服务、测量工具或两者在不同运行中的差异而有所不同。我们将三个验证问题分开：在其历史配置下，先前发现是否在新数据上重复出现（复制），在相同标识符下重新构建评估和推理配置时，端点是否发生变化（测量敏感性），以及在一个共同工具下，发现是否在随后测试的标识符中保持不变（持久性）。我们在Regent Chess中研究这些问题，这是一个顺序环境，其中隐藏的可变状态被准确记录，允许在行动时间对声明的信念进行真实评分；正的端点值意味着性能比匹配的均匀比较器更差。先前报告的Gemini 3.1 Flash-Lite缺陷在其历史配置下的新游戏中再次出现（+0.0530，95%置信区间[+0.0329,+0.0714]）。在同一天的H/R比较中，在相同的公共标识符下，重建配置下的模型减去均匀端点低0.0429（H减R对比的95%置信区间[+0.0182,+0.0667]）；所有六个配置组件共同变化，因此没有单独的组件。在重建的R下，前瞻性冻结的交错同窗口4K比较在Gemini 3.1和Gemini 3.7之间的符号发生了反转，这两个标识符在发布和产品层级上有所不同；额外的描述性和探索性单元显示出相同的方向模式。任何额外的服务期贡献仍未解决（-0.0166，[-0.0483,+0.0157]）。因此，复制、测量敏感性和持久性可能在一次评估中得出不同的结论，这促使我们明确按测试标识符、服务期、测量工具和推理配置对托管模型的行为声明进行索引。
cs.AI / 8 / 2609.22497
The Wisdom of Artificial Deliberative Crowds
人工审议群体的智慧
Abstract
The aggregation of many lay estimates often outperforms individual expert judgment, a phenomenon known as the wisdom of crowds. While this is usually attributed to the independence of estimates, an even stronger effect arises through deliberation: averaging the consensus estimates of small deliberating groups outperforms the classical wisdom of crowds, with individual judgments themselves also becoming more accurate after deliberation. Whether these improvements transfer to large language models deliberating amongst themselves is unknown. Here we adapt a three-stage deliberation paradigm previously used with human participants for use with large language models from three different families, and test it across four domains of increasing real-world stakes: visual numerical estimation (Study 1), peer review of machine-learning papers (Study 2), detection of hidden malicious behavior by an artificial intelligence agent (Study 3), and sports forecasting against a real prediction market (Study 4). Across domains, deliberation reduced collective error beyond passive aggregation of independent responses, and post-deliberation individual judgments retained this collective gain. Notably, the advantage required model diversity: groups composed of clones of a single model did not benefit from deliberating. These results establish machine deliberation as a general-purpose aggregation mechanism, and point to diversity as an active ingredient.
Chinese Translation
许多普通估计的聚合通常优于个别专家的判断，这一现象被称为群体智慧。虽然这通常归因于估计的独立性，但通过审议产生的效果更为显著：小型审议小组的共识估计的平均值超越了经典的群体智慧，个体判断在审议后也变得更加准确。这些改进是否能转移到大型语言模型之间的审议尚不清楚。在此，我们将之前用于人类参与者的三阶段审议范式调整为适用于来自三种不同家族的大型语言模型，并在四个具有日益增加现实世界风险的领域进行测试：视觉数值估计（研究1）、机器学习论文的同行评审（研究2）、检测人工智能代理的隐藏恶意行为（研究3）以及与真实预测市场的体育预测（研究4）。在各个领域，审议减少了集体错误，超出了独立反应的被动聚合，且审议后的个体判断保留了这一集体收益。值得注意的是，这一优势需要模型多样性：由单一模型的克隆组成的群体并未从审议中受益。这些结果确立了机器审议作为一种通用的聚合机制，并指出多样性是其有效成分。
cs.AI / 9 / 2609.22512
Agreement Overstates Evidence: Error Dependence in LLM Judge Consensus
一致性夸大证据：大型语言模型评审共识中的错误依赖性
Abstract
Consensus among LLM judges is often taken as strong evidence that a decision is correct. This assumes that judges make their errors independently. In practice, LLM judges are often trained and evaluated in similar ways, so they can make the same mistakes. We study how this dependency affects the reliability of consensus. We find substantial error correlation across both open-weight and frontier LLM judges. In our main bank of ten judges, the average pairwise correlation between judge errors is 0.21. As a result, the ten judges only provide roughly as much statistical information as 3.5 independent judges. The dependency is even stronger among the high-accuracy frontier judges we evaluate, including judges from different providers. In up to 28% of our comparisons, ignoring shared errors leads to the conclusion that one system is significantly better, while accounting for them does not. We also find that the pattern of errors matters. Errors shared by most judges and errors concentrated among a smaller group affect consensus differently and favor different voting methods. Measuring the overall amount of correlation alone is therefore insufficient. Our results suggest a simple approach: use a small set of trusted examples to estimate judge accuracy and identify shared mistakes. These shared errors should then be considered when analyzing the results, and the voting method should be chosen using trusted examples before it is applied to new data.
Chinese Translation
大型语言模型（LLM）评审之间的一致性常被视为决策正确性的有力证据。这一假设基于评审独立犯错。然而，在实际操作中，LLM 评审通常以相似的方式进行训练和评估，因此可能会犯相同的错误。我们研究了这种依赖性如何影响共识的可靠性。我们发现，无论是在开放权重的 LLM 评审还是前沿 LLM 评审之间，错误的相关性都相当显著。在我们主要的十位评审组中，评审错误之间的平均成对相关性为 0.21。因此，这十位评审提供的统计信息大致相当于 3.5 位独立评审的结果。在我们评估的高准确率前沿评审中，这种依赖性甚至更强，包括来自不同提供者的评审。在我们最多 28% 的比较中，忽略共享错误会得出一个系统显著更好的结论，而考虑这些错误则不会。我们还发现，错误的模式也很重要。大多数评审共享的错误与集中在较小组别中的错误对共识的影响不同，并且偏向不同的投票方法。因此，仅仅测量整体相关性是不够的。我们的结果建议了一种简单的方法：使用一小组可信的示例来估计评审的准确性并识别共享错误。在分析结果时，应考虑这些共享错误，并在将投票方法应用于新数据之前，使用可信示例选择投票方法。
cs.AI / 10 / 2609.22529
IntLawNER: A Named Entity Recognition Dataset and Benchmark in International Law
IntLawNER：国际法领域的命名实体识别数据集与基准
Abstract
International law provides the normative framework through which states coordinate action, regulate armed conflict, and protect human rights, yet its texts remain without token-level named entity recognition (NER) resources. We introduce IntLawNER, a NER dataset and benchmark for codified sources of international law, covering 2,987 gold-annotated sentences and 8,094 entity spans from International Court of Justice (ICJ) decisions, UN Security Council resolutions, and European Court of Human Rights (ECtHR) judgments, annotated with seven institution-specific entity types. We construct IntLawNER with a cost-effective hybrid algorithmic-agentic pipeline that reduces 468k source sentences to a compact annotation set through candidate retrieval, LLM-based vetting, and human review, with 89.6% of gold spans accepted unchanged from the silver layer. However, the silver-to-gold analysis reveals that human-machine aggregate agreement metrics can be misleading in domain-specific NER: Cohen's kappa=0.964 on boundary-matched spans masks a macro-F1 of 0.753 when missing entities, boundary errors, and label corrections are included. The benchmark shows that zero-shot span-based GLiNER collapses on entity types dependent on institutional function rather than surface form (0.243 micro-F1), while fine-tuned transformers struggle on rare labels. Carefully selected few-shot examples that demonstrate label contrasts improve every LLM over zero-shot prompting, with Claude Opus 4.6 reaching the best score of 0.873 micro-F1. We release IntLawNER as a benchmark and reusable resource for extracting references in international legal texts.
Chinese Translation
国际法提供了国家协调行动、规范武装冲突和保护人权的规范框架，但其文本仍缺乏逐词级别的命名实体识别（NER）资源。我们介绍了IntLawNER，这是一个针对国际法成文来源的NER数据集和基准，涵盖了2987个金标注句子和8094个实体跨度，这些数据来自国际法院（ICJ）裁决、联合国安全理事会决议和欧洲人权法院（ECtHR）判决，并标注了七种特定机构的实体类型。我们通过一种具有成本效益的混合算法-代理流程构建了IntLawNER，该流程通过候选检索、基于大型语言模型（LLM）的审核和人工审查，将468,000个源句子缩减为一个紧凑的标注集，其中89.6%的金标跨度在银标层中未做更改而被接受。然而，银标与金标的分析表明，人机聚合一致性度量在特定领域的NER中可能具有误导性：在边界匹配的跨度上，Cohen's kappa=0.964掩盖了当缺失实体、边界错误和标签修正被纳入时的宏观F1值为0.753。基准测试显示，零-shot跨度基础的GLiNER在依赖于机构功能而非表面形式的实体类型上崩溃（微观F1为0.243），而微调的变换器在稀有标签上表现不佳。经过精心挑选的少量示例展示了标签对比，改善了每个LLM相较于零-shot提示的表现，其中Claude Opus 4.6达到了最佳的微观F1得分0.873。我们将IntLawNER作为基准和可重用资源发布，以提取国际法律文本中的引用。
cs.AI / 11 / 2609.22537
EvidenT: Building Trustworthy Enterprise Assistants through Evidence Groundedness and Traceability
EvidenT：通过证据基础和可追溯性构建可信赖的企业助手
Abstract
Enterprise AI assistants must produce responses that are verifiable and traceable to source evidence. However, retrieval augmented generation (RAG) over heterogeneous enterprise data can suffer from citation drift, unsupported content, and weak source traceability. We present EvidenT (T = Trust + Transparency + Traceability), a lightweight pipeline that verifies extracted evidence against retrieved documents before answer generation, without model retraining. EvidenT combines structured passage extraction with deterministic lexical alignment to filter unsupported content, correct citation drift, and preserve source-span traceability. On approximately 500 real enterprise queries, EvidenT improves gold-source hit rate by an average of 29% over prompting baselines, produces no citations to nonretrieved urls, and achieves near-saturated answer-to-source lexical coverage.
Chinese Translation
企业人工智能助手必须生成可验证且可追溯至源证据的响应。然而，基于异构企业数据的检索增强生成（RAG）可能会遭遇引用漂移、内容不支持和源追溯性弱等问题。我们提出了EvidenT（T = 信任 + 透明度 + 可追溯性），这是一个轻量级管道，在答案生成之前验证提取的证据与检索到的文档的一致性，而无需重新训练模型。EvidenT结合了结构化段落提取和确定性词汇对齐，以过滤不支持的内容、纠正引用漂移，并保持源跨度的可追溯性。在大约500个真实企业查询中，EvidenT的金源命中率平均提高了29%，没有生成对未检索到的URL的引用，并且实现了几乎饱和的答案与源的词汇覆盖率。
cs.AI / 12 / 2609.22592
AutoGym: Blueprint-First Generation of Verifiable Agent Gyms
AutoGym：可验证智能体健身房的蓝图优先生成
Abstract
Training agents with reinforcement learning requires a gym, comprising a task, an executable environment in which the task can be attempted, and a verifier that reliably distinguishes success from failure. Constructing such gyms remains manual, expensive, and static. Task sets saturate as models improve and are increasingly exposed to contamination. Synthetic generation offers scale, but single-pass synthesis produces tasks whose difficulty is largely cosmetic. Models comparable in capability solve them despite convoluted phrasing, and correctness must be adjudicated post-hoc by unreliable LLM judges. We present AutoGym, a framework that generates complete gyms (tasks, executable environments, and verifiers) from a minimal domain seed or prior model trajectories. AutoGym introduces three mechanisms. (1) Blueprint-first generation specifies the valid solution space, environment requirements, and verification criteria before the environment is materialized, making solvability a construction prerequisite rather than a property verified after the fact. (2) Explicit generation parameters control task topology, interaction depth, capability axes, question obfuscation, and distractor composition, enabling fine-grained difficulty steering. (3) Active curriculum synthesis uses performance-informed calibration to adjust the distribution over these parameters as model capabilities evolve. Across productivity and temporal-reasoning settings, AutoGym generates gyms spanning the capability spectrum, including instances that challenge frontier models.
Chinese Translation
使用强化学习训练智能体需要一个健身房，其中包括一个任务、一个可执行的环境（可以在其中尝试该任务）以及一个能够可靠区分成功与失败的验证器。构建这样的健身房仍然是手动的、昂贵的和静态的。随着模型的改进，任务集趋于饱和，并且越来越容易受到污染。合成生成提供了规模，但单次生成产生的任务其难度在很大程度上是表面的。能力相当的模型能够解决这些任务，尽管其表述复杂，而正确性必须由不可靠的LLM评判者事后裁定。我们提出了AutoGym，一个从最小领域种子或先前模型轨迹生成完整健身房（任务、可执行环境和验证器）的框架。AutoGym引入了三种机制。(1) 蓝图优先生成在环境物化之前指定有效的解决空间、环境要求和验证标准，使得可解性成为构建的先决条件，而不是事后验证的属性。(2) 显式生成参数控制任务拓扑、交互深度、能力轴、问题模糊化和干扰项组成，实现细粒度的难度调节。(3) 主动课程合成使用基于性能的信息校准来调整这些参数的分布，随着模型能力的发展而演变。在生产力和时间推理设置中，AutoGym生成涵盖能力谱的健身房，包括挑战前沿模型的实例。
cs.AI / 13 / 2609.22599
MAWILE: Multi-Axis Workbench for Inspecting LLM Evaluators
MAWILE：用于检查大型语言模型评估者的多轴工作台
Abstract
Large language model (LLM) judges provide a flexible and scalable method for evaluating model and agent outputs, but their verdicts can be sensitive to incidental changes in the evaluated response, judge instructions, and scoring rubric. Existing systems examine important subsets of these failure modes, but auditing a configured judge requires testing both the judge instrument and the items it evaluates. We introduce MAWILE, a developer-facing workbench for auditing judge sensitivity across four surfaces: the judge prompt, judge rubric, target-system input, and target-system output. Given a user-supplied judge and representative evaluation items, MAWILE constructs and validates controlled perturbations, re-executes the judge, and localizes the resulting sensitivity. Each perturbation declares whether the verdict should remain invariant or change in a specified direction, allowing the same system to measure both robustness to irrelevant variations and sensitivity to meaningful changes. MAWILE audits binary, ordinal, and pairwise judges without requiring gold labels. The code for this tool is available at: github.com/megagonlabs/mawile-judge.
Chinese Translation
大型语言模型（LLM）评估者提供了一种灵活且可扩展的方法来评估模型和代理的输出，但它们的裁决可能对被评估响应、评估者指令和评分标准的偶然变化敏感。现有系统检查这些失败模式的重要子集，但审计一个配置好的评估者需要同时测试评估工具和其评估的项目。我们介绍了MAWILE，这是一个面向开发者的工作台，用于审计评估者在四个方面的敏感性：评估者提示、评估者评分标准、目标系统输入和目标系统输出。给定用户提供的评估者和代表性的评估项目，MAWILE构建并验证受控扰动，重新执行评估者，并定位结果的敏感性。每个扰动声明裁决应保持不变或在指定方向上变化，从而允许同一系统同时测量对无关变化的鲁棒性和对重要变化的敏感性。MAWILE可以审计二元、序数和成对评估者，而无需黄金标签。该工具的代码可在以下网址获取：github.com/megagonlabs/mawile-judge。
cs.AI / 14 / 2609.22619
GaitVista: Reliability-Aware AI Measurement toward Accessible Longitudinal Gait Assessment
GaitVista：面向可及的纵向步态评估的可靠性意识人工智能测量
Abstract
Tracking recovery of walking function requires detecting meaningful gait change across rehabilitation sessions, yet objective 3D measurement remains confined to specialized motion-capture laboratories. Small camera sets and body-worn inertial sensors broaden access, but reliability varies across joints and time, allowing sensing failures to masquerade as patient change. We present \textsc{GaitVista}, a reliability-aware measurement layer whose lightweight gate assigns joint- and frame-specific visual contributions using camera coverage, local visual quality, cross-modal disagreement, and root-motion continuity, and exposes them for inspection. Across seven clean and degraded sensing conditions on TotalCapture, \textsc{GaitVista} reduces average full-body and lower-body error by \textbf{27.7\%} and \textbf{27.8\%}, attains the lowest worst-condition error among fusion methods, and reduces the gap to a joint-frame oracle from $2.76$--$5.33$~cm for condition-blind baselines to $1.11$~cm. On MoVi with image-derived keypoints, it is the only deployable fusion method to improve over both unimodal streams, reducing marker-supported error by \textbf{6.4\%} relative to the strongest learned fusion baseline. On TotalCapture, it improves bilateral knee-flexion waveform accuracy by \textbf{18.9\%}. Raw inertial measurements from five TotalCapture participants show location- and time-varying magnetic disturbance, supporting the design's reliability premise. Both benchmarks contain neurologically healthy participants in controlled settings and retain participant-specific IMU calibration; we therefore report progress toward accessible gait assessment, not validated clinical deployment.
Chinese Translation
跟踪步行功能的恢复需要在康复会话中检测有意义的步态变化，但客观的三维测量仍局限于专业的运动捕捉实验室。小型摄像头组和穿戴式惯性传感器扩大了访问范围，但在关节和时间上的可靠性各异，导致传感失败伪装成患者变化。我们提出了 extsc{GaitVista}，一种可靠性意识的测量层，其轻量级门控根据摄像头覆盖、局部视觉质量、跨模态不一致性和根运动连续性分配关节和帧特定的视觉贡献，并将其暴露以供检查。在 TotalCapture 的七种干净和退化的传感条件下， extsc{GaitVista} 将全身和下身的平均误差分别减少了 extbf{27.7\%} 和 extbf{27.8 extbf{，获得了融合方法中最低的最差条件误差，并将与关节-帧预言者的差距从 $2.76$--$5.33$~cm 降低到 $1.11$~cm。在 MoVi 上，使用图像派生的关键点，它是唯一一个在两个单模态流上均有改进的可部署融合方法，相较于最强的学习融合基线，减少了 extbf{6.4 extbf{的标记支持误差。在 TotalCapture 上，它将双侧膝关节屈曲波形的准确性提高了 extbf{18.9 extbf{。来自五名 TotalCapture 参与者的原始惯性测量显示位置和时间变化的磁干扰，支持设计的可靠性前提。两个基准均包含在受控环境中神经系统健康的参与者，并保持参与者特定的 IMU 校准；因此，我们报告了向可及步态评估的进展，而非经过验证的临床部署。
cs.AI / 15 / 2609.22620
Splitting Documents at Lower Cost: Multi-Split Boundary Decisions for LLM-Based Page Stream Segmentation
以更低成本拆分文档：基于LLM的多拆分边界决策用于页面流分割
Abstract
Scanned mail, uploaded PDFs, and consolidated attachments often arrive as page streams that must be split into individual documents before downstream classification, extraction, or routing. Zero-shot large language models can detect document boundaries without task-specific training, but standard Page Classification (PC) and Boundary Decision (BD) formulations resolve only one boundary per model call. We introduce Multi-Split Boundary Decision (MSBD), which predicts multiple boundaries within a page window in a single call, reducing the number of inference requests. We evaluate MSBD across multiple language models, document collections, input modalities, and window sizes. The results reveal a model- and corpus-dependent operating range in which MSBD preserves strong segmentation accuracy while substantially improving inference efficiency, followed by a sharp decline at larger windows. MSBD provided the strongest overall accuracy--efficiency trade-off, while large windows expose distinct over- and under-segmentation behavior across models. These findings show that multi-boundary prediction can make zero-shot page stream segmentation more efficient when the window size is selected for the target corpus.
Chinese Translation
扫描邮件、上传的PDF文件和合并的附件通常以页面流的形式到达，在进行后续分类、提取或路由之前，必须将其拆分为单独的文档。零-shot大语言模型能够在没有特定任务训练的情况下检测文档边界，但标准的页面分类（PC）和边界决策（BD）公式每次模型调用只能解析一个边界。我们提出了多拆分边界决策（MSBD），它能够在一次调用中预测页面窗口内的多个边界，从而减少推理请求的数量。我们在多个语言模型、文档集合、输入模态和窗口大小上评估了MSBD。结果显示，MSBD在保持强大分割准确性的同时，显著提高了推理效率，且其操作范围依赖于模型和语料库；在较大窗口下则出现明显下降。MSBD提供了最佳的准确性与效率权衡，而较大窗口则暴露出模型间明显的过分割和欠分割行为。这些发现表明，当窗口大小为目标语料库选择时，多边界预测可以使零-shot页面流分割更为高效。
cs.AI / 16 / 2609.22628
Text, Pixels, or Both? Evaluating Input Representations for Multimodal Document QA
文本、像素还是两者兼具？多模态文档问答输入表示的评估
Abstract
Every document QA system begins with a choice that is rarely studied on its own: whether to feed the model page images, extracted text, or both. We isolate this choice, holding the prompt, judge, and scoring pipeline fixed, across four commercial model endpoints, two corpora, and two context regimes (gold evidence pages and the full document). On documents that fit the image budget, page images lead on accuracy at every document length on both corpora, but this advantage carries a growing latency and cost premium: text latency stays roughly flat as documents lengthen while image latency rises steadily. Text and images also fail on different questions, with exactly one representation correct on 19--25% of items across the reported cells, so neither subsumes the other. Exploiting this complementarity, a lightweight TF-IDF router that reads only the question text gains 2.6 points over always-text while cutting median latency 30% relative to always-vision, on a document-disjoint held-out split.
Chinese Translation
每个文档问答系统的开始都面临一个鲜有研究的选择：是向模型输入页面图像、提取的文本，还是两者兼而有之。我们将这一选择独立出来，保持提示、评判和评分流程不变，跨越四个商业模型端点、两个语料库和两种上下文模式（黄金证据页面和完整文档）。在适合图像预算的文档中，页面图像在两个语料库的每个文档长度上都表现出更高的准确性，但这一优势伴随着不断增加的延迟和成本溢价：随着文档长度的增加，文本的延迟基本保持平稳，而图像的延迟则稳步上升。文本和图像在不同的问题上也会出现失误，在报告的各个单元中，恰好有一种表示在19%到25%的项目上是正确的，因此两者并不相互涵盖。利用这种互补性，一个仅读取问题文本的轻量级TF-IDF路由器在始终使用文本的情况下提高了2.6个百分点，同时相对于始终使用视觉的情况将中位延迟降低了30%，在一个文档不重叠的保留分割上进行测试。
cs.AI / 17 / 2609.22682
Self-Organizing Agent Teams Learn to Reason Together
自组织代理团队学习共同推理
Abstract
Collective intelligence depends not only on what team members know, but also on how they organize their work. When the structure of a solution is unknown, useful roles and divisions of labor cannot be specified in advance; teams must learn from experience how to organize reasoning as it unfolds. Human teams routinely adapt this way, while existing AI agent teams rely on fixed protocols, explicit task decomposition, or routing. We introduce Self-Organizing Agent Teams (SAT), fixed teams of AI agents that learn reusable strategies from prior collaborations to organize roles, conversational phases, participation, and information flow. These strategies enable what we call collaborative computation: agents exchange, challenge, repair, and synthesize partial reasoning into solutions no member produced independently. In two independent settings, we learn teamwork strategies that transfer unchanged to unseen benchmarks, using only 15 mathematics and 25 graduate-level knowledge problems. Across five mathematics and physics benchmarks, self-organizing teams average 66.7% accuracy, versus 48.8% for their strongest member, 58.7% for compute-matched inference by that agent, and 59.0% for a perfect router over members' independent answers; on AIME 2026, they exceed this router by 13.4 points. Because gains vary across benchmarks, we ask when self-organizing collaboration helps. Across eight benchmarks, demonstrability (the organizational-psychology construct of whether a team can distinguish correct from incorrect reasoning) strongly tracks improvement over the strongest member (Spearman $\rho=0.90$, $p=0.005$): teams benefit most when correct reasoning can be recognized once it appears. More broadly, these results suggest that organization itself can become an agent capability: agent teams can learn how to reason together and produce solutions their members could not reach independently.
Chinese Translation
集体智能不仅依赖于团队成员的知识，还依赖于他们如何组织工作。当解决方案的结构未知时，无法提前指定有用的角色和劳动分工；团队必须通过经验学习如何在推理展开时进行组织。人类团队通常以这种方式进行适应，而现有的人工智能代理团队则依赖于固定的协议、明确的任务分解或路由。我们提出了自组织代理团队（Self-Organizing Agent Teams, SAT），这是一组固定的人工智能代理，它们通过先前的合作学习可重用的策略，以组织角色、对话阶段、参与和信息流。这些策略使我们称之为协作计算（collaborative computation）：代理之间交换、挑战、修复和综合部分推理，形成没有任何成员独立产生的解决方案。在两个独立的环境中，我们学习到的团队合作策略在未见基准上可以不变地转移，仅使用了15个数学问题和25个研究生水平的知识问题。在五个数学和物理基准测试中，自组织团队的平均准确率为66.7%，而其最强成员的准确率为48.8%，该代理的计算匹配推理为58.7%，完美路由器对成员独立答案的准确率为59.0%；在AIME 2026中，他们的表现超过了这一路由器13.4分。由于增益在不同基准测试中有所不同，我们探讨自组织协作何时有助。在八个基准测试中，可证明性（demonstrability，组织心理学构念，指团队是否能够区分正确与错误的推理）与最强成员的改进高度相关（Spearman $
ho=0.90$, $p=0.005$）：当正确的推理一旦出现就能被识别时，团队受益最大。更广泛地说，这些结果表明，组织本身可以成为一种代理能力：代理团队可以学习如何共同推理，并产生其成员无法独立达到的解决方案。
cs.AI / 18 / 2609.22691
Generative Embodied Multiple Behavior Control Systems for Human-like Agents
生成式具身多行为控制系统用于类人智能体
Abstract
An enduring and richly elaborated dichotomy in cognitive neuroscience is that of human behavior control mechanisms, divided into habitual versus goal-directed. While existing human-like agent frameworks primarily focus on modeling goal- directed behavior, habitual behavior has been largely overlooked, though it plays a crucial role in human daily life. In this paper, we address this gap by studying multiple behavior control systems that jointly model goal-directed and habitual behaviors. We propose a human behavior control mechanism-inspired framework which the Habitual Controller retrieves cue-triggered behaviors from personal- ized habit memory, while the Goal-directed Controller employs a context-aware world model to predict action consequences and estimate their values. The Arbiter dynamically balances the influence of both systems according to individual differ- ences and momentary internal states. To reconstruct diverse human-level behavior instructions in 3D environments, we further develop a keyframe-guided 3D mo- tion generation module. Through extensive evaluation methods, human studies, and ablations studies, experimental results demonstrate that human-likeness per- formance is significantly improved by our approach. The efficacy of our approach indicates the benefits of leveraging habitual behavior and multiple behavior con- trol system coordination for believable embodied human-like agents.
Chinese Translation
在认知神经科学中，一个持久且丰富的二分法是人类行为控制机制的划分，分为习惯性行为与目标导向行为。尽管现有的类人智能体框架主要集中于建模目标导向行为，但习惯性行为在日常生活中扮演着至关重要的角色，然而却被大多忽视。本文通过研究多种行为控制系统，填补了这一空白，这些系统共同建模目标导向和习惯性行为。我们提出了一种受人类行为控制机制启发的框架，其中习惯控制器从个性化的习惯记忆中检索线索触发的行为，而目标导向控制器则利用上下文感知的世界模型来预测行动后果并评估其价值。仲裁者根据个体差异和瞬时内部状态动态平衡这两个系统的影响。为了在三维环境中重建多样的人类水平行为指令，我们进一步开发了一个关键帧引导的三维运动生成模块。通过广泛的评估方法、人类研究和消融研究，实验结果表明，我们的方法显著提高了类人表现的效果。我们方法的有效性表明，利用习惯性行为和多行为控制系统的协调对于可信的具身类人智能体具有显著的益处。
cs.AI / 19 / 2609.22694
Toward Auditable and Calibrated AI for Dementia-Related Crash Severity Prediction: A Selective Deferral Framework to Support Human Review
朝向可审计和校准的人工智能以预测与痴呆相关的事故严重性：支持人工审查的选择性延迟框架
Abstract
Public crash databases increasingly support automated safety analysis, but crash severity prediction remains difficult to translate into public-sector decision workflows when models are evaluated primarily as ordinary classifiers. This study reframes dementia-related crash severity modeling as a decision-aware triage problem in which a system must classify crashes into no-injury/property-damage-only (O), minor or moderate injury (BC), and fatal or severe injury (KA), while also controlling outcome leakage, reporting severe under-triage, calibrating confidence, and preserving every raw prediction for audit. Using 4,781 Texas crash records with structured fields and police narratives, we evaluate structured, narrative, fusion, calibrated fusion, BERT-family, and local large-language-model baselines under a stratified 70/15/15 split. In the reported split, leakage-controlled Gemma obtains the highest observed macro-F1 (0.545; 95% bootstrap CI [0.507, 0.583]). The best calibrated fusion model obtains macro-F1 of 0.522 and expected calibration error of 0.033. Selective deferral improves performance among cases retained for automatic classification. At 70% coverage, macro-F1 rises to 0.573 and severity cost falls to 0.577, while deferred cases are treated as candidates for a proposed human-review process and are not further evaluated in the present experiment. The study contributes a reproducible, leakage-controlled, and uncertainty-aware evaluation framework for crash AI systems, emphasizing auditability and selective deferral rather than accuracy alone.
Chinese Translation
公共事故数据库日益支持自动化安全分析，但当模型主要作为普通分类器进行评估时，事故严重性预测仍然难以转化为公共部门的决策工作流程。本研究将与痴呆相关的事故严重性建模重新框架为一个决策意识的分流问题，其中系统必须将事故分类为无伤害/仅财产损失（O）、轻微或中度伤害（BC）以及致命或严重伤害（KA），同时控制结果泄漏、报告严重的低分流、校准置信度，并保留每个原始预测以便审计。利用4781条具有结构化字段和警察叙述的德克萨斯州事故记录，我们在分层的70/15/15拆分下评估了结构化、叙述、融合、校准融合、BERT家族和本地大型语言模型的基准。在报告的拆分中，控制泄漏的Gemma获得了最高的观察到的宏F1值（0.545；95%自助法置信区间[0.507, 0.583]）。最佳的校准融合模型获得了宏F1值为0.522，期望校准误差为0.033。选择性延迟提高了保留用于自动分类的案例的性能。在70%的覆盖率下，宏F1值上升至0.573，严重性成本降至0.577，而延迟的案例被视为拟议的人工审查过程的候选者，并且在本实验中不再进一步评估。本研究贡献了一个可重复的、控制泄漏的、关注不确定性的事故人工智能系统评估框架，强调可审计性和选择性延迟，而不仅仅是准确性。
cs.AI / 20 / 2609.22695
A Survey on the Linear Representation Hypothesis
线性表示假设的调查
Abstract
The term "linear representation hypothesis" (LRH) has appeared across diverse subfields of artificial intelligence, neuroscience, and cognitive science. But previous works have not consistently treated the LRH as a falsifiable scientific hypothesis; we analyze these inconsistencies and examine their implications for how prior theoretical and methodological results should be interpreted. Based on this analysis, we argue that claims regarding linear representations become well-defined only through careful examination of the model, representation location, feature definition, and evaluation dataset. We therefore propose a more rigorous formalization of the LRH that makes these dependencies explicit and allows the hypothesis to be evaluated as a falsifiable scientific claim. Finally, we identify some non-trivial open problems that warrant further attention from the research community.
Chinese Translation
“线性表示假设”（Linear Representation Hypothesis, LRH）这一术语出现在人工智能、神经科学和认知科学的多个子领域。然而，以往的研究并未将LRH一致地视为可证伪的科学假设；我们分析了这些不一致之处，并考察了它们对如何解读先前理论和方法结果的影响。基于这一分析，我们认为关于线性表示的论断只有通过对模型、表示位置、特征定义和评估数据集的仔细审查才能得到明确。因此，我们提出了一种更为严格的LRH形式化，使这些依赖关系明确化，并允许将该假设作为可证伪的科学主张进行评估。最后，我们识别出一些值得研究界进一步关注的非平凡开放问题。
cs.AI / 21 / 2609.22696
Building Trustworthy Mental Health Benchmarks on Bluesky: A Validation-Aware Weak-Supervision Framework
在Bluesky上构建可信赖的心理健康基准：一种关注验证的弱监督框架
Abstract
Decentralized social media platforms create new opportunities and challenges for computational mental health research because data access, moderation, labeling, and deployment responsibilities are distributed across multiple technical and governance layers. This paper presents a validation-aware weak-supervision system for constructing and evaluating suicidal ideation (SI) and broader mental health (MH) disclosure benchmarks on Bluesky, a decentralized social media platform built on the AT Protocol. The system integrates public firehose collection, task-specific lexicon filtering, Llama-3-8B-assisted binary annotation, human-adjudicated validation subsets, and transformer-based model benchmarking. Using this pipeline, we construct two task-specific corpora containing 8,346 SI-labeled posts and 9,988 MH-labeled posts. The evaluation shows that model performance depends strongly on both task definition and validation protocol. BERT+LSTM achieves the highest SI stratified cross-validation F1-score, RoBERTa achieves the strongest SI holdout F1-score, and DistilRoBERTa achieves the best MH cross-validation F1-score. Human validation reveals different weak-label failure modes across tasks, with SI labels dominated by false negatives and MH labels dominated by false positives. These findings show that decentralized social media can support reproducible mental health benchmarking, but only when system design, label provenance, validation strategy, and deployment constraints are evaluated together.
Chinese Translation
去中心化社交媒体平台为计算心理健康研究创造了新的机遇和挑战，因为数据访问、内容审核、标注和部署责任分散在多个技术和治理层面。本文提出了一种关注验证的弱监督系统，用于在基于AT协议的去中心化社交媒体平台Bluesky上构建和评估自杀意念（SI）和更广泛的心理健康（MH）披露基准。该系统整合了公共数据流收集、特定任务的词汇过滤、Llama-3-8B辅助的二元标注、人为裁定的验证子集和基于变换器的模型基准测试。通过这一流程，我们构建了两个特定任务的语料库，包含8,346条标注为SI的帖子和9,988条标注为MH的帖子。评估结果表明，模型性能在很大程度上依赖于任务定义和验证协议。BERT+LSTM在SI分层交叉验证F1分数上表现最佳，RoBERTa在SI保留集F1分数上表现最强，而DistilRoBERTa在MH交叉验证F1分数上表现最佳。人工验证揭示了不同任务间的弱标签失败模式，其中SI标签以假阴性为主，MH标签以假阳性为主。这些发现表明，去中心化社交媒体可以支持可重复的心理健康基准测试，但前提是系统设计、标签来源、验证策略和部署限制需共同评估。
cs.AI / 22 / 2609.22702
Hapi: A Multivariable Land-Surface Transformer for Medium-Range Hydrological Forecasting at Continental Scale
Hapi：一种用于大陆尺度中期水文预测的多变量地表变换器
Abstract
Accurate flood forecasts several days in advance are essential for flood control, water-resource management, and emergency response. Producing them at high resolution over a continental domain calls for local hydrological detail together with spatial context extending from river basins to synoptic weather systems. We developed Hapi, a U-Net Swin Transformer that uses fine three-dimensional patches and hierarchical shifted-window attention to forecast discharge, surface runoff, snow water equivalent, and soil wetness across the contiguous United States. The model produces 24--72-hour forecasts at $0.05^{\circ}$ resolution, with learned Laplacian task weights adjusting each variable's contribution to training. On 2024 test data using reconstructed weather and land-surface inputs from ERA5-Land, Hapi outperformed an operational physics-based model and a state-of-the-art AI model in flood detection. Independent validation against 3,881 U.S. Geological Survey gauges and a Hurricane Helene case study supported its advantage over the physics-based model in reproducing daily discharge. Controlled experiments showed that learned task weighting strengthens rare-flood detection, which is particularly sensitive to changes in precipitation inputs. Hapi produced a four-variable, 72-hour forecast across the contiguous United States with an average inference time of 0.11 seconds on a single A100 GPU.
Chinese Translation
准确的洪水预报在几天前对洪水控制、水资源管理和应急响应至关重要。在大陆范围内以高分辨率生成这些预报需要结合地方水文细节以及从流域到天气系统的空间背景。我们开发了Hapi，一种使用精细三维补丁和分层移动窗口注意力机制的U-Net Swin Transformer，能够预测美国大陆范围内的流量、地表径流、雪水当量和土壤湿度。该模型以$0.05^{ ext{°}}$的分辨率生成24至72小时的预报，学习的拉普拉斯任务权重调整每个变量对训练的贡献。在使用ERA5-Land重建的天气和地表输入的2024年测试数据中，Hapi在洪水检测方面优于一个操作性物理模型和一个最先进的人工智能模型。针对3881个美国地质调查局水文监测站的独立验证以及飓风海伦案例研究支持了其在再现日流量方面优于物理模型的优势。控制实验表明，学习的任务加权增强了稀有洪水的检测，这对降水输入的变化特别敏感。Hapi在美国大陆范围内生成了一个四变量的72小时预报，单个A100 GPU的平均推理时间为0.11秒。
cs.AI / 23 / 2609.22712
Trustworthy Agentic AI: Failure Modes, Mitigation Strategies, and a Lifecycle Framework for Autonomous LLM Systems
可信赖的代理智能AI：失败模式、缓解策略及自主大型语言模型系统的生命周期框架
Abstract
Agentic AI systems built on large language models can plan over multiple steps, use external tools, retain information in memory, and coordinate with other agents. These capabilities make them more useful than static language models, but they also introduce new security and operational risks. Untrusted content from websites, emails, documents, and databases can enter the same context as system instructions; persistent memory can carry compromised information across sessions; and access to external tools can turn an incorrect model response into a consequential real-world action. This article reviews the trustworthiness of agentic AI across five interconnected dimensions: safety and robustness, alignment and human oversight, transparency and auditability, privacy and data governance, and regulatory compliance. It organizes key failure modes, including indirect prompt injection, backdoor triggers, goal misgeneralization, memory contamination, and cross-session data leakage, into a unified taxonomy. It also examines major mitigation approaches, such as instruction hierarchies, context isolation, spotlighting, process-based supervision, constrained tool use, and privacy-preserving memory, while distinguishing techniques supported by empirical evidence from those that remain largely conceptual. Building on this analysis, we introduce the Trustworthy Agent Development Lifecycle (TADL), a six-phase framework covering specification, design, training, evaluation, deployment, and monitoring. For each phase, TADL identifies relevant trust activities, expected evidence, and risk-based decision gates. Although TADL has not yet been empirically validated, it provides a structured foundation for developing and evaluating more secure and accountable agentic systems. The article concludes by identifying gaps in current benchmarks and outlining priorities for future research.
Chinese Translation
基于大型语言模型的代理智能AI系统能够进行多步骤规划、使用外部工具、在记忆中保留信息，并与其他代理协调。这些能力使其比静态语言模型更具实用性，但也引入了新的安全和操作风险。来自网站、电子邮件、文档和数据库的不可信内容可能与系统指令处于同一上下文中；持久性记忆可能在会话间携带被破坏的信息；而对外部工具的访问可能将错误的模型响应转化为具有后果的现实世界行为。本文从安全性与稳健性、对齐与人类监督、透明性与可审计性、隐私与数据治理以及合规性五个相互关联的维度审视代理智能AI的可信赖性。我们将关键失败模式，包括间接提示注入、后门触发、目标误泛化、记忆污染和跨会话数据泄露，组织成统一的分类法。同时，本文还考察了主要的缓解方法，如指令层级、上下文隔离、聚焦、基于过程的监督、受限工具使用和隐私保护记忆，并区分了有实证支持的技术与仍然主要概念化的技术。在此分析的基础上，我们引入了可信赖的代理开发生命周期（Trustworthy Agent Development Lifecycle, TADL），这是一个涵盖规范、设计、训练、评估、部署和监控的六个阶段框架。对于每个阶段，TADL识别相关的信任活动、预期证据和基于风险的决策门。尽管TADL尚未经过实证验证，但它为开发和评估更安全、更负责任的代理系统提供了结构化的基础。文章最后识别了当前基准中的空白，并概述了未来研究的优先事项。
cs.AI / 24 / 2609.22746
ProcessLight: Process Supervision for Large Language Model Based Traffic Signal Control
ProcessLight：基于大型语言模型的交通信号控制过程监督
Abstract
Large Language Models (LLMs) have recently been introduced into traffic signal control (TSC) as decision agents due to their strengths in human-readable reasoning generation. Yet, existing LLM TSC methods optimize only from final outcomes and fail to distinguish valid from flawed reasoning steps, causing useful or misleading steps to be jointly updated and thus impairing the model's learning of effective reasoning. To bridge this gap, we propose an LLM-based framework ProcessLight to decompose signal decisions into verifiable semantic steps. Building on ProcessLight, we further develop Step-wise Traffic Process Policy Optimization (STeP-PO), a novel reinforcement learning framework that optimizes structured reasoning processes through step-level credit assignment. Specifically, STeP-PO uses step quality scores to evaluate local reasoning quality and step importance to measure each step's influence on the final action, and then assigns step-level advantages over a semantic step tree structure. The resulting step-level advantages are propagated to reasoning tokens, enabling fine-grained policy optimization beyond outcome-only rewards. Extensive experiments over multiple real-world datasets demonstrate the superiority of our methods. Our code is available at https://github.com/wenzhaoabc/processlight.
Chinese Translation
大型语言模型（LLMs）由于其在可读性推理生成方面的优势，最近被引入到交通信号控制（TSC）中作为决策代理。然而，现有的LLM TSC方法仅从最终结果进行优化，未能区分有效推理步骤与错误推理步骤，导致有用或误导性的步骤被共同更新，从而损害了模型对有效推理的学习。为了解决这一问题，我们提出了一种基于LLM的框架ProcessLight，将信号决策分解为可验证的语义步骤。在此基础上，我们进一步开发了逐步交通过程策略优化（Step-wise Traffic Process Policy Optimization，STeP-PO），这是一种新颖的强化学习框架，通过逐步信用分配优化结构化推理过程。具体而言，STeP-PO使用步骤质量评分来评估局部推理质量，使用步骤重要性来衡量每个步骤对最终行动的影响，然后在语义步骤树结构上分配步骤级优势。最终得到的步骤级优势被传播到推理标记，从而实现超越仅基于结果奖励的细粒度策略优化。对多个真实世界数据集的广泛实验表明了我们方法的优越性。我们的代码可在 https://github.com/wenzhaoabc/processlight 获取。
cs.AI / 25 / 2609.22760
CTSpinoPelvic1K: spine, pelvis, ribs and femora in one coordinate frame, annotated for lumbosacral transitional anatomy
CTSpinoPelvic1K：在一个坐标系中标注脊柱、骨盆、肋骨和股骨，针对腰骶部过渡解剖
Abstract
Purpose: A vertebra at the lumbosacral junction is named by counting caudally from C2 on whole-spine imaging, but a lumbar case is planned on lumbar-only imaging (T12 to S1), without C2. Abdominopelvic CT holds that span plus the lowest ribs and pelvis. Where a lumbosacral transitional vertebra (LSTV) alters the count, the local anatomy is ambiguous: four rib-free vertebrae may be an L1 with a lumbar rib or an L5 assimilated to the sacrum, and six may be a sixth lumbar vertebra, a T12 with aplastic ribs, or a lumbarized S1. CTSpinoPelvic1K asks whether local morphology resolves it without the count. CTSpine1K's vertebrae and CTPelvic1K's pelvis covered these patients but were never joined; this release joins them on one series and adds the bones neither had. It provides 802 CT records with per-level ribs and femora, levels anchored on the lowest rib-bearing vertebra and S1, and classes for L6, T13, a separate S1 and lumbar ribs, so anomalies are recorded as such. Acquisition and Validation Methods: Records pair CTSpine1K and CTPelvic1K labels on each patient's bone-richest series under a VerSe-native scheme. Validation covered geometric invariants (802/802 pass), rib-vertebra incidence across 5,749 ribs (0.035% offset), and spinopelvic measures matching published values. Data Format and Usage Notes: NIfTI image/label pairs with patient-grouped LSTV-stratified five-fold splits and a loader; archived at https://doi.org/10.5281/zenodo.22139642. Potential Applications: Classifying a vertebra from local features; updating cadaveric morphometry; spinopelvic assessment; opportunistic screening; and, absent a public preoperative lumbar cohort, surgical planning research (377 records prone). Limitations: thoracic ground truth is field-of-view limited; postural angles supine; no held-out test set; ribs are triaged-review pseudolabels; Castellvi grades two-reader consensus on 33 records.
Chinese Translation
目的：在腰骶交界处的椎骨通过从C2向下计数来命名，但计划在仅限腰椎的成像（T12到S1）中进行腰椎病例，而不涉及C2。腹盆CT涵盖了该范围以及最低的肋骨和骨盆。当腰骶过渡椎（LSTV）改变计数时，局部解剖结构变得模糊：四个无肋椎骨可能是带有腰肋的L1，或是与骶骨融合的L5，而六个可能是第六腰椎、带有发育不全肋骨的T12，或是腰化的S1。CTSpinoPelvic1K探讨局部形态是否能在不计数的情况下解决此问题。CTSpine1K的椎骨和CTPelvic1K的骨盆覆盖了这些患者，但从未结合；此次发布将它们整合在同一系列中，并增加了两者都没有的骨骼。它提供了802个CT记录，包含每个层级的肋骨和股骨，层级以最低肋骨椎骨和S1为锚点，并为L6、T13、单独的S1和腰肋提供分类，以便将异常记录为此。获取与验证方法：记录在每个患者的骨骼丰富系列中配对CTSpine1K和CTPelvic1K标签，采用VerSe原生方案。验证涵盖几何不变性（802/802通过）、5,749根肋骨的肋骨-椎骨发生率（偏差0.035%）以及脊柱骨盆测量与已发布值的匹配。数据格式与使用说明：NIfTI图像/标签对，按患者分组的LSTV分层五折拆分及加载器；存档于https://doi.org/10.5281/zenodo.22139642。潜在应用：根据局部特征对椎骨进行分类；更新尸体形态测量；脊柱骨盆评估；机会筛查；以及在缺乏公共术前腰椎队列的情况下，进行外科规划研究（377个记录为俯卧位）。局限性：胸椎真实情况受视野限制；姿势角度为仰卧；没有保留测试集；肋骨为分级复审伪标签；Castellvi等级在33个记录上由两位读者达成共识。
cs.AI / 26 / 2609.22775
DVA-Neurons: Design and Verification of Adaptive LIF Neurons: From Single-Neuron Dynamics to Multi-Neuron Spiking Networks
DVA-神经元：自适应LIF神经元的设计与验证：从单神经元动态到多神经元脉冲网络
Abstract
Spiking Neural Networks (SNNs) offer a promising path toward ultra-low-power artificial intelligence inference by emulating the event-driven computation of biological neurons. However, two challenges limit their practical deployment. First, fixed-parameter Leaky Integrate-and-Fire (LIF) neurons lack the adaptation mechanisms observed in biology, where neurons modulate their excitability based on firing history. Second, scaling from single neurons to multi-neuron networks introduces challenges in synaptic weight distribution and inter-neuron spike routing that are absent in isolated designs. This paper addresses both issues through the extension, verification, and physical implementation of adaptive LIF neurons at three architectural scales. This work contributes: a 2nd-order neuron with two-stage synaptic filtering for richer temporal dynamics; a fully-connected 6-neuron spiking network with configurable weights (100 to 5) demonstrating weight-based inter-neuron communication; and a direct verification methodology enabling per-cycle observation of all internal states. All designs were synthesized targeting Selected Area Electron Diffraction (SAED) 14 nm Complementary Metal-Oxide-Semiconductor (CMOS) technology at 1 GHz and verified with Cocotb-based Python testbenches under pulsed current stimuli (amplitude 80, ISI=3). The results show that adaptation effectively modulates firing: 31% suppression in the 2nd-order neuron (25 vs.\ 36 spikes) and 31% reduction in postsynaptic firing in the network (18 vs.\ 26 spikes). Physically, the 2nd-order neuron costs 1.77x more area and 1.52x more power than the 1st-order baseline, while the 6-neuron network demonstrates near-linear scaling (5.7x area, 5.3x power). Seven verification bugs spanning testbench connectivity, fixed-point overflow, and Verilog expression-width semantics are documented.
Chinese Translation
脉冲神经网络（SNNs）通过模拟生物神经元的事件驱动计算，为超低功耗人工智能推理提供了有希望的路径。然而，有两个挑战限制了它们的实际应用。首先，固定参数的漏积分发放（LIF）神经元缺乏生物学中观察到的适应机制，生物神经元会根据发放历史调节其兴奋性。其次，从单个神经元扩展到多神经元网络引入了突触权重分配和神经元间脉冲路由的挑战，而这些在孤立设计中并不存在。本文通过在三个架构尺度上扩展、验证和物理实现自适应LIF神经元来解决这两个问题。本研究的贡献包括：一种具有两级突触滤波的二阶神经元，以实现更丰富的时间动态；一个具有可配置权重（100到5）的全连接6神经元脉冲网络，展示了基于权重的神经元间通信；以及一种直接验证方法，使得能够在每个周期观察所有内部状态。所有设计均针对选定区域电子衍射（SAED）14纳米互补金属氧化物半导体（CMOS）技术在1 GHz下进行合成，并在脉冲电流刺激（幅度80，ISI=3）下通过基于Cocotb的Python测试平台进行验证。结果表明，适应性有效地调节了发放：二阶神经元的发放抑制率为31%（25与36次脉冲），网络中突触后发放减少31%（18与26次脉冲）。从物理上看，二阶神经元的面积成本比一阶基线高1.77倍，功耗高1.52倍，而6神经元网络则表现出近线性的扩展（面积5.7倍，功耗5.3倍）。文中记录了七个验证错误，涵盖测试平台连接、定点溢出和Verilog表达式宽度语义等问题。
cs.AI / 27 / 2609.22790
From Research Frontier to Laboratory Bench: Design of a Four-Tier Experimental Teaching System for Multimodal Medical Image Intelligent Diagnosis
从研究前沿到实验室工作台：多模态医学图像智能诊断的四层实验教学系统设计
Abstract
Undergraduate programmes in intelligent medical engineering are expanding, yet laboratory curricula lag behind the multimodal, long-tailed, and distributionally shifting realities of clinical AI. This design paper presents an advanced experimental teaching system that translates an ongoing multimodal deep learning research project on endometrial carcinoma into a structured undergraduate lab sequence. We identify three educational gaps (modality, authenticity, and deployment) and derive four pedagogical principles from constructive alignment, experiential learning, the research teaching nexus, and the CDIO framework. The curriculum comprises four progressive tiers plus an engineering layer, with 32 laboratory units over 64 contact hours, delivered via a custom virtual clinical workstation using de-identified multi-institutional data. Each tier maps to a specific technical bottleneck, prerequisite coursework, and criterion-referenced deliverables. Data governance, safety, and assessment protocols are specified. Learning outcome data will be collected across two implementation cycles.
Chinese Translation
智能医学工程的本科课程正在扩展，但实验室课程却滞后于临床人工智能的多模态、长尾和分布变化的现实。本文设计了一种先进的实验教学系统，将一项关于子宫内膜癌的多模态深度学习研究项目转化为结构化的本科实验序列。我们识别出三个教育差距（模态性、真实性和部署），并从建构性对齐、体验式学习、研究教学联系以及CDIO框架中推导出四项教学原则。课程包括四个逐步递进的层次和一个工程层，共有32个实验单元，64个接触小时，通过使用去标识化的多机构数据的定制虚拟临床工作站进行授课。每个层次对应特定的技术瓶颈、先修课程和基于标准的交付成果。数据治理、安全性和评估协议已被明确规定。学习成果数据将在两个实施周期中进行收集。
cs.AI / 28 / 2609.22878
ISA-Bench: A Benchmark for Computational Reasoning Across Instruction Set Architectures
ISA-Bench：跨指令集架构的计算推理基准
Abstract
Large language model code generation benchmarks primarily evaluate well-resourced languages like Python and Java, where models benefit from abundant training data. They provide limited evidence about reasoning in unfamiliar computational models: deriving arithmetic from a single subtract instruction, coordinating parallel programs across communicating nodes, or wiring logic gates into circuits. We present ISA-Bench, a benchmark of programming games with constrained instruction sets. For each game we provide a full execution stack (parser, VM, and verifier), enabling automated evaluation with structured feedback for iterative refinement. Reasoning models achieve higher average solve rates than code-specialized and general-purpose models, but unfamiliar syntax remains a major source of failure. Models solve more tasks with iterative feedback, though the gains vary substantially across architectures. We introduce a reasoning--execution gap (REG) analysis that reveals a recurring disconnect between identifying a plausible computational strategy and expressing it as a correct program in the target ISA. Code is open-sourced.
Chinese Translation
大型语言模型代码生成基准主要评估资源丰富的语言，如 Python 和 Java，这些语言的模型受益于丰富的训练数据。它们对在不熟悉的计算模型中的推理提供了有限的证据：从单个减法指令推导算术、协调跨通信节点的并行程序，或将逻辑门连接成电路。我们提出了 ISA-Bench，这是一个具有约束指令集的编程游戏基准。对于每个游戏，我们提供了完整的执行栈（解析器、虚拟机和验证器），使得能够通过结构化反馈进行自动评估，以便进行迭代改进。推理模型的平均解决率高于代码专用和通用模型，但不熟悉的语法仍然是主要的失败来源。模型在迭代反馈下解决更多任务，尽管不同架构之间的增益差异显著。我们引入了一种推理-执行差距（REG）分析，揭示了识别可行计算策略与在目标 ISA 中将其表达为正确程序之间的反复脱节。代码已开源。
cs.AI / 29 / 2609.22910
When Should a VLM Look? Paying Only for Visual Calls That Were Needed and Used
视觉语言模型何时应当观察？仅为所需和使用的视觉调用付费
Abstract
Vision-language agents that crop and zoom are trained with rewards that credit a successful tool call, yet a successful call does not show that the model needed to look or used the pixels it received. On our cold-start checkpoint only 10% to 12% of visual calls were both needed and used, and released agents make spurious calls 36% to 87% of the time on individual benchmarks. Outcome rewards, judge rewards, and branch probes each observe one side of this failure, and about two thirds of what an outcome reward pays goes to calls that were neither needed nor used. CounterCredit asks both questions of every image-returning call at its realized pre-call state, using the policy's own gold-answer score. A decision value compares the realized visual branch with answering immediately; an evidence value compares the returned crop with random same-size patches substituted into the same call. A call verified on both earns cashback and every other executed call pays rent; the price is bounded so that every correct trajectory outranks every wrong one, and a dual-channel GRPO advantage keeps the price in its own units. From the same cold start, prompt pool, and budget, CounterCredit reaches 89.5% on V*, 80.2% on HR-Bench-4K, and 76.4% on HR-Bench-8K, 6.3 to 9.4 points above outcome-only GRPO at 1.78 against 1.84 calls per question, and lowers the spurious-call rate to 31% to 36%, the lowest among the agents evaluated. The same recipe lifts a Qwen3-VL-8B base from 75.4 to 80.8 on average.
Chinese Translation
视觉语言代理通过奖励机制进行训练，以奖励成功的工具调用，然而成功的调用并不能证明模型确实需要观察或使用所接收到的像素。在我们的冷启动检查点中，仅有10%到12%的视觉调用是既需要又使用的，而已发布的代理在各个基准测试中有36%到87%的时间发出虚假调用。结果奖励、判断奖励和分支探测各自观察到这一失败的一个方面，约三分之二的结果奖励支付给了既不需要也未使用的调用。CounterCredit在每个返回图像的调用的实现前状态下提出了这两个问题，使用策略自身的黄金答案分数进行评估。决策值将实现的视觉分支与立即回答进行比较；证据值则将返回的裁剪与随机同尺寸的补丁进行比较。经过双重验证的调用将获得现金返还，而每个其他执行的调用则需支付租金；价格是有界的，以确保每个正确的轨迹优于每个错误的轨迹，并且双通道GRPO优势保持价格在其自身单位内。在相同的冷启动、提示池和预算下，CounterCredit在V*上达到了89.5%，在HR-Bench-4K上达到了80.2%，在HR-Bench-8K上达到了76.4%，比仅基于结果的GRPO高出6.3到9.4分，调用次数为每个问题1.78对比1.84，并将虚假调用率降低至31%到36%，在评估的代理中为最低。同样的方法将Qwen3-VL-8B基础模型的平均分从75.4提升至80.8。
cs.AI / 30 / 2609.22939
Beyond Linear Context: Graph-Guided Evidence Navigation for Long-Novel Reasoning with a Local 9B Language Model
超越线性上下文：基于图的证据导航在局部9B语言模型中的长篇推理
Abstract
Long-context models read a novel the way a person reads a printout: one token after another, in narrative order, with the whole history competing for a fixed budget of attention. A detective does not work that way. They sort what happened when, and they keep a map of who relates to whom, so a clue from chapter one can meet a question asked at the end of the book. We test whether a frozen knowledge graph can give a small local model that same freedom. Thirty detective novels and 234 multiple-choice questions are answered by one fixed qwen3.5:9b reader under nine conditions: five graph routes, a recent-window baseline, whole-book compression, ordinary vector retrieval, and a question-only control. The strongest graph route reaches 53.85% (126/234) against 46.15% for the recent window, 51.28% for compression, 51.71% for vector retrieval and 40.17% for question-only. On the subset that no model can answer without the book, the graph route reaches 42.86%. None of the fifteen graph-baseline contrasts survives Holm correction, so we present the result as exploratory evidence about a design. Two structural findings survive scrutiny better than the headline number: annotated evidence concentrates in the topological core of these graphs (2.35x enrichment, pooled), and the two graph-building pipelines differ so much in annotation coverage (16% versus 73% of clue paragraphs) that pooled accuracy alone would hide which bottleneck is being measured.
Chinese Translation
长上下文模型阅读小说的方式类似于人们阅读打印文本：一个一个地读取标记，按照叙述顺序进行，整个历史在固定的注意力预算下相互竞争。然而，侦探并不是这样工作的。他们会整理事件发生的时间，并保持一个关于谁与谁相关的地图，因此第一章的线索可以与书末提出的问题相遇。我们测试一个冻结的知识图谱是否能赋予一个小型局部模型同样的自由。三十部侦探小说和234个多项选择题由一个固定的qwen3.5:9b阅读器在九种条件下进行回答：五条图路线、一个最近窗口基线、整本书压缩、普通向量检索和仅问题控制。最强的图路线达到了53.85%（126/234），而最近窗口为46.15%，压缩为51.28%，向量检索为51.71%，仅问题为40.17%。在没有模型可以在没有书籍的情况下回答的子集中，图路线达到了42.86%。十五个图基线对比中没有一个在Holm校正下存活，因此我们将结果呈现为关于设计的探索性证据。两个结构性发现比头条数字更能经得起审查：注释证据集中在这些图的拓扑核心（2.35倍富集，汇总），而两个图构建管道在注释覆盖率上差异显著（线索段落的16%对比73%），以至于仅凭汇总准确性无法揭示哪个瓶颈正在被测量。
cs.AI / 31 / 2609.22951
AgentRouter: Heterogeneous Model Routing for Cost-Optimal Multi-Step Agentic Workflows
AgentRouter：面向成本最优的多步骤智能工作流的异构模型路由
Abstract
Enterprise agentic systems that route every trajectory step to a frontier model waste 60-80% of their inference budget on subtasks that smaller models handle equally well. Existing routing solutions optimize single-turn query assignment but ignore a property unique to agentic workflows: subtask complexity varies widely within a single trajectory. A planning step may require frontier-class reasoning while a subsequent formatting step needs only a 7B model. We formalize step-level model routing as a sequential assignment problem over agent trajectories and propose AgentRouter, a lightweight classifier (12M parameters, <5ms overhead per step on an A100 GPU) that maps each trajectory step to one of four model tiers using five features extractable at routing time. Trained on 50,000 annotated agent trajectory steps spanning planning, coding, research, and data analysis tasks, AgentRouter achieves 72% cost reduction relative to frontier-only baselines, retaining 97.3% of frontier-only quality (less than 3% degradation in end-to-end task completion); per-step routing accuracy reaches 91% on minimal-complexity steps and 85% on efficient-tier steps, with 76-82% on the harder mid-range and frontier tiers. On the same benchmarks, RouteLLM and FrugalGPT (applied per-step) achieve only 31% and 44% cost reduction respectively, because their single-turn training signal misses trajectory-level quality dependencies.
Chinese Translation
企业智能系统在每个轨迹步骤中将任务路由到前沿模型，导致60-80%的推理预算浪费在小模型同样能够处理的子任务上。现有的路由解决方案优化单轮查询分配，但忽略了智能工作流的一个独特特性：单一轨迹内的子任务复杂性差异很大。一个规划步骤可能需要前沿级别的推理，而随后的格式化步骤仅需一个7B模型。我们将步骤级模型路由形式化为一个在智能轨迹上的序列分配问题，并提出了AgentRouter，一个轻量级分类器（12M参数，A100 GPU上每步开销小于5毫秒），它利用在路由时可提取的五个特征，将每个轨迹步骤映射到四个模型层级之一。AgentRouter在50,000个标注的智能轨迹步骤上进行训练，这些步骤涵盖了规划、编码、研究和数据分析任务，相较于仅使用前沿模型的基线，达到了72%的成本降低，同时保持了97.3%的前沿模型质量（端到端任务完成的降级小于3%）；在最低复杂度步骤上的每步路由准确率达到91%，在高效层级步骤上为85%，在更复杂的中间层和前沿层级上为76-82%。在相同的基准测试中，RouteLLM和FrugalGPT（逐步应用）分别仅实现了31%和44%的成本降低，因为它们的单轮训练信号未能捕捉到轨迹级别的质量依赖性。
cs.AI / 32 / 2609.22956
A Compact Stance-Indexed Anterior-Posterior COP Representation for Parkinson's Disease Classification from Plantar VGRF
一种紧凑的姿态索引前后中心压力（COP）表示法用于从足底垂直地面反作用力（VGRF）分类帕金森病
Abstract
Parkinson's disease alters gait and bilateral coordination, but machine-learning performance also depends on how continuous gait signals are represented. This study investigates whether preserving anterior-posterior center-of-pressure (AP-COP) information at fixed locations across normalized stance provides a compact and informative representation of plantar-force gait signals. Bilateral vertical ground reaction force recordings from 165 participants in the Gait in Parkinson's Disease Database were evaluated using repeated fully nested participant-level cross-validation. We propose AP-COP10, comprising AP-COP position and bilateral asymmetry across five stance windows. AP-COP10 achieved an AUC of 0.894 and outperformed three harmonized literature-derived COP representations under the same evaluation pipeline. The complementary 25 non-AP-COP descriptors alone achieved an AUC of 0.856, while the complete 35-feature representation achieved 0.908. Removing AP-COP10 from the complete representation produced a statistically supported loss in discrimination, whereas adding the complementary descriptors to AP-COP10 yielded only a small, unsupported improvement. Feature competition indicated that the most informative stance-indexed descriptors were concentrated in early and early-mid stance, while source-study holdout and sensor-perturbation analyses supported the robustness of the representation. These findings indicate that stance-indexed AP-COP retains discriminative information that is not readily recovered by broader engineered gait descriptors, supporting compact and interpretable representations for machine-learning analysis of pathological gait.
Chinese Translation
帕金森病会改变步态和双侧协调，但机器学习的性能也依赖于如何表示连续的步态信号。本研究探讨了在标准化的站立过程中，是否保留固定位置的前后中心压力（AP-COP）信息能够提供足够紧凑且信息丰富的足底力步态信号表示。我们评估了来自165名参与者的双侧垂直地面反作用力记录，这些参与者来自帕金森病步态数据库，通过重复完全嵌套的参与者级交叉验证进行分析。我们提出了AP-COP10，包括五个站立窗口中的AP-COP位置和双侧不对称性。AP-COP10的曲线下面积（AUC）达到了0.894，并在相同的评估流程下优于三种文献衍生的标准化COP表示法。仅使用25个非AP-COP描述符的AUC为0.856，而完整的35特征表示法的AUC达到了0.908。从完整表示法中移除AP-COP10会导致统计上显著的区分能力损失，而将补充描述符添加到AP-COP10仅带来了微小且不支持的改善。特征竞争表明，最具信息量的姿态索引描述符集中在早期和早中期的站立阶段，而源研究的保留和传感器扰动分析则支持了该表示法的稳健性。这些发现表明，姿态索引的AP-COP保留了不易通过更广泛的工程步态描述符恢复的区分信息，支持用于病理步态机器学习分析的紧凑且可解释的表示法。
cs.AI / 33 / 2609.22959
R-GEAN: Regimen-Guided Edit Action Network for Within-Admission Medication Change Prediction
R-GEAN：基于方案指导的编辑行动网络用于住院期间药物变更预测
Abstract
The medications prescribed to a patient often change during a hospital admission as clinicians start, stop, or continue therapies. We study whether models can predict which medication classes are added or removed between 24 hours after admission and discharge. Metrics that compare the complete discharge regimen can reward models for copying medications that remain unchanged, even when they identify no actual changes. We therefore introduce a leakage-controlled benchmark that predicts net ATC3 additions and removals using only prior completed admissions and information available within the first 24 hours of the current admission. Addition candidates are classes not active at 24 hours, whereas removal candidates are classes active at that time. We also introduce R-GEAN, an asymmetric candidate-scoring network with independent addition and removal predictors. Across 240,480 admissions from 82,286 patients, R-GEAN achieves the highest predefined summary of addition, removal, changed-regimen, and action-pattern performance, termed the edit composite (0.464), compared with 0.435 for the strongest primary comparator. Reimplemented RETAIN, GAMENet, and MICRON baselines obtain 0.428, 0.420, and 0.288, respectively. R-GEAN's advantage is concentrated in correctly identifying medication classes no longer active at discharge, while rare additions and admissions with multiple medication changes remain difficult. Rankings based on micro-F1 over the reconstructed discharge regimen and the edit composite correlate weakly across the evaluated models (Spearman r = 0.20). The continuation baseline achieves the highest complete-regimen score despite predicting no additions or removals. These results show that complete-regimen and edit-level evaluation measure different aspects of medication prediction. The benchmark evaluates observed prescribing changes, not treatment appropriateness
Chinese Translation
患者在住院期间所处方的药物常常会发生变化，因为临床医生会开始、停止或继续治疗。我们研究模型是否能够预测在入院后24小时与出院之间添加或移除的药物类别。比较完整出院方案的指标可能会奖励那些复制未发生变化的药物的模型，即使它们未能识别出实际的变化。因此，我们引入了一个泄漏控制基准，该基准仅使用先前完成的住院记录和当前住院前24小时内可用的信息来预测净ATC3添加和移除。添加候选类是指在24小时内未激活的类别，而移除候选类是指在该时间点激活的类别。我们还引入了R-GEAN，一个具有独立添加和移除预测器的非对称候选评分网络。在来自82,286名患者的240,480次住院中，R-GEAN在添加、移除、变更方案和行动模式表现的预定义总结中取得了最高的成绩，称为编辑复合指标（0.464），而最强的主要比较者为0.435。重新实现的RETAIN、GAMENet和MICRON基线分别获得0.428、0.420和0.288。R-GEAN的优势集中在正确识别出院时不再活跃的药物类别，而稀有的添加和多药物变更的住院情况仍然困难。基于重建出院方案的微F1和编辑复合指标的排名在评估的模型之间相关性较弱（Spearman r = 0.20）。尽管未预测任何添加或移除，继续基线仍然获得了最高的完整方案得分。这些结果表明，完整方案和编辑级别的评估测量药物预测的不同方面。该基准评估观察到的处方变化，而非治疗的适宜性。
cs.AI / 34 / 2609.22987
OptiSkill: A Hierarchical and Evolving SkillBank for LLM-Based Optimization Modeling
OptiSkill：一种基于层次和演化的技能库用于大语言模型优化建模
Abstract
Automated operations research (OR) modeling requires LLMs to translate natural-language decision problems into correct mathematical programs. Existing methods can improve individual formulations, but they often solve problems in isolation, retaining little reusable experience and repeating similar formulation errors. Prior memory-based approaches store examples, thoughts, or insights as references, while OR modeling requires reusable formulation skills that transfer across problem narratives and guide concrete modeling decisions. We propose OptiSkill, a skill-augmented framework that builds a hierarchical and evolving SkillBank for LLM-based OR modeling. SkillBank stores solver-verified experience as reusable skills, with Global Strategies for problem-level formulation skeletons and Step Experiences for local error-prevention rules. It is further refined through stable batch-level test-time evolution, where candidate skills are incorporated only after validation. Experiments on eight OR modeling benchmarks show that OptiSkill improves formulation accuracy across LLM backbones, outperforms strong agentic baselines, and gains further by expanding SkillBank coverage and reliability. Code and data are available at https://github.com/rachhhhing/OptiSkill
Chinese Translation
自动化运筹学（OR）建模需要大语言模型（LLMs）将自然语言决策问题转化为正确的数学程序。现有的方法可以改善单个公式的质量，但它们往往孤立地解决问题，保留的可重用经验很少，并且重复类似的公式错误。之前的基于记忆的方法将示例、思考或见解存储为参考，而运筹学建模需要可重用的公式技能，这些技能可以跨问题叙述转移，并指导具体的建模决策。我们提出了OptiSkill，一个增强技能的框架，构建了一个层次化和演化的技能库（SkillBank）用于基于LLM的运筹学建模。技能库存储经过求解器验证的经验作为可重用的技能，包含用于问题级公式骨架的全球策略和用于局部错误预防规则的步骤经验。它通过稳定的批量级测试时间演化进一步优化，其中候选技能仅在验证后被纳入。对八个运筹学建模基准的实验表明，OptiSkill在不同的LLM基础模型中提高了公式的准确性，超越了强大的代理基线，并通过扩展技能库的覆盖范围和可靠性获得了进一步的提升。代码和数据可在 https://github.com/rachhhhing/OptiSkill 获取。
cs.AI / 35 / 2609.23023
PINNForge: Execution-Grounded Evolutionary Design of Physics-Informed Neural Networks for PDE Solving via Large Language Models
PINNForge：基于执行反馈的物理信息神经网络进化设计框架，通过大型语言模型解决偏微分方程
Abstract
Physics-informed neural networks (PINNs) require coordinated choices over network representation, sampling, loss construction, and optimization, while effective configurations often vary substantially across partial differential equations (PDEs). Existing automated PINN design methods can search candidate configurations, but information revealed during actual training is still used mainly for evaluation rather than to improve subsequent design, leading to repeated trial-and-error and inefficient use of training budget. We propose PINNsForge, an LLM-driven evolutionary framework for execution-feedback-based automated PINN design. PINNsForge generates diverse candidate configurations from PDE-related prior knowledge, evaluates them through actual training, and feeds high-performing designs together with accumulated execution evidence back to the LLM. Guided by observed optimization behavior, the LLM then refines, recombines, and explores coupled PINN design components, forming a continual cycle of generation, execution, feedback, and evolution. Unlike one-shot search or evaluation-only feedback, PINNsForge progressively converts training experience into improved design decisions for the target PDE. Across 25 PDE benchmarks, PINNsForge achieves the lowest mean MSE on 24 tasks compared with RoPINN, PINNsFormer, and PINNsAgent. Ablation studies further confirm the importance of the PDE knowledge base, execution feedback, and evolutionary search: removing these components increases the mean MSE to 3.74$\times$, 12.10$\times$, and 10.10$\times$ that of the full PINNsForge, respectively.
Chinese Translation
物理信息神经网络（PINNs）在网络表示、采样、损失构建和优化方面需要协调的选择，而有效的配置在不同的偏微分方程（PDEs）之间往往存在显著差异。现有的自动化PINN设计方法可以搜索候选配置，但在实际训练过程中揭示的信息主要用于评估，而不是用于改进后续设计，这导致了重复的试错过程和训练预算的低效使用。我们提出了PINNsForge，一个基于执行反馈的、由大型语言模型驱动的自动化PINN设计的进化框架。PINNsForge从与PDE相关的先验知识中生成多样的候选配置，通过实际训练对其进行评估，并将高性能设计及累积的执行证据反馈给大型语言模型。大型语言模型在观察到的优化行为的指导下，进一步细化、重组和探索耦合的PINN设计组件，形成生成、执行、反馈和进化的持续循环。与一次性搜索或仅评估反馈不同，PINNsForge逐步将训练经验转化为针对目标PDE的改进设计决策。在25个PDE基准测试中，PINNsForge在24个任务上实现了最低的平均均方误差（MSE），相比于RoPINN、PINNsFormer和PINNsAgent。消融研究进一步确认了PDE知识库、执行反馈和进化搜索的重要性：去除这些组件会使平均MSE分别增加到3.74倍、12.10倍和10.10倍于完整的PINNsForge。
cs.AI / 36 / 2609.23038
Spatial-Interactor: Learning Spatial Reasoning through Interaction with the Observable Physical World
空间交互者：通过与可观察物理世界的互动学习空间推理
Abstract
Spatial reasoning is essential for vision-language models (VLMs) to understand and act in the physical world. Reasoning in dynamic environments requires VLMs to perceive local state transitions caused by object motion and viewpoint changes and integrate them over long trajectories to maintain an updated spatial state, yet existing VLMs remain limited in both capabilities. Current spatial training primarily focuses on static questions about object attributes and spatial relations, providing limited direct supervision for state transitions; in contrast, interaction trajectories naturally connect a preceding observation, an action, and a subsequent observation, offering direct supervision for local state transitions, while complete trajectories reveal dependencies among consecutive transitions. We therefore introduce Spatial-Interactor, a framework that trains VLMs to model physical-world state transitions through interaction, organizing this learning process into a three-level curriculum covering L1 passive world-state transitions, L2 active self-state transitions, and L3 long-horizon interaction trajectories. Accordingly, we construct the Learning from Spatial Interaction dataset (LSI-108K) from simulated and real interaction trajectories, with tasks aligned with the objective of each level. Our two-stage training strategy applies Supervised Fine-Tuning (SFT) to L1 and L2 for local transition modeling, and On-Policy Distillation (OPD) then uses privileged self-distillation: a teacher branch given segment-level transition descriptions supervises the student's on-policy CoT, helping the student learn to integrate consecutive transitions over L3 long trajectories. Experiments across multiple VLMs and spatial benchmarks show consistent gains in local transition modeling and long-horizon integration.
Chinese Translation
空间推理对于视觉-语言模型（VLMs）理解和在物理世界中行动至关重要。在动态环境中进行推理要求VLMs感知由物体运动和视角变化引起的局部状态转变，并在长时间轨迹中整合这些转变以保持更新的空间状态，而现有的VLMs在这两方面的能力仍然有限。目前的空间训练主要集中在关于物体属性和空间关系的静态问题上，对状态转变提供的直接监督有限；相比之下，交互轨迹自然地连接了先前的观察、一个动作和随后的观察，提供了对局部状态转变的直接监督，同时完整的轨迹揭示了连续转变之间的依赖关系。因此，我们提出了空间交互者（Spatial-Interactor），这是一个通过互动训练VLMs建模物理世界状态转变的框架，将这一学习过程组织为三个层次的课程，涵盖L1被动世界状态转变、L2主动自我状态转变和L3长时间交互轨迹。因此，我们从模拟和真实的交互轨迹中构建了学习空间交互数据集（Learning from Spatial Interaction dataset，LSI-108K），其任务与每个层次的目标相一致。我们的两阶段训练策略对L1和L2应用监督微调（Supervised Fine-Tuning，SFT）以进行局部转变建模，然后在L3长轨迹上使用在政策蒸馏（On-Policy Distillation，OPD），利用特权自我蒸馏：一个教师分支根据段级转变描述监督学生的在政策链式推理（CoT），帮助学生学习在L3长轨迹中整合连续转变。多个VLMs和空间基准的实验显示出在局部转变建模和长时间整合方面的一致提升。
cs.AI / 37 / 2609.23043
Enforcing Narrative Reliability and Epistemic Pacing in LLM-Driven Detective Games via Structured Knowledge Trees
通过结构化知识树在大语言模型驱动的侦探游戏中强化叙事可靠性和认知节奏
Abstract
Large Language Models (LLMs) enable open-ended dialogue in interactive games, but their non-deterministic outputs make it difficult to preserve authorial control, factual consistency, and the intended sequence of information disclosure. These challenges are particularly significant in detective games, where premature revelation or fabricated details can undermine the logic of player progression. We present a Structured Knowledge Tree architecture coupled with a tri-agent LLM pipeline for controlling dialogue in an open-ended interrogation game. The system separates knowledge retrieval, dialogue generation, and response verification to ensure that the virtual suspect reveals only information permitted by the current narrative state. We evaluate the approach through The Interrogation of Adrian Gale, a playable detective-game testbed, and a formal user study examining hallucination reduction, adherence to authored disclosure sequences, and perceived logical progression. Our results demonstrate that the structured architecture reduces critical hallucinations by 64.78% and entirely prevents premature narrative disclosure. While the strict mechanical constraints introduced usability trade-offs regarding forced conversational reveals, the system successfully enforces rigorous epistemic pacing and provides players with a clear, subjective sense of progression toward solving the case.
Chinese Translation
大型语言模型（LLMs）使得互动游戏中的开放式对话成为可能，但其非确定性的输出使得保持作者控制、事实一致性以及信息披露的预期顺序变得困难。这些挑战在侦探游戏中尤为显著，因为过早的揭示或虚构的细节可能会破坏玩家进展的逻辑。我们提出了一种结构化知识树架构，并结合三代理 LLM（tri-agent LLM）管道来控制开放式审问游戏中的对话。该系统将知识检索、对话生成和响应验证分开，以确保虚拟嫌疑人仅揭示当前叙事状态允许的信息。我们通过《对阿德里安·盖尔的审问》（The Interrogation of Adrian Gale），一个可玩侦探游戏测试平台，以及一项正式用户研究来评估该方法，研究内容包括幻觉减少、遵循作者披露序列的情况以及感知的逻辑进展。我们的结果表明，结构化架构将关键幻觉减少了 64.78%，并完全防止了过早的叙事披露。尽管引入的严格机械约束在可用性上带来了强制对话揭示的权衡，但该系统成功地强化了严格的认知节奏，并为玩家提供了清晰的、主观的案件解决进展感。
cs.AI / 38 / 2609.23058
LazyAgent: Demand-Driven Materialization and Physical Optimization of Agentic Programs
LazyAgent：基于需求的材料化与智能程序的物理优化
Abstract
Current agent runtimes that plan before acting generally execute a step once it becomes ready. We present LazyAgent, a unified execution framework for agent-authored programs organized around a live, goal-derived demanded set. LazyAgent refreshes a backward closure from requested outputs as execution state changes and materializes a ready node only when the active goal requires it. This replaces repeated local judgments with one linear-time graph analysis followed by constant-time membership tests, allowing programs to remain broad while execution stays request-specific. On programs that describe more than the current request needs, LazyAgent consistently outperforms the strongest goal-stopping eager baseline by refusing unrelated work before it starts. Adding one unrelated product raises the eager bill by 22.5% and LazyAgent's by 0.0%. LazyAgent saves 42.0% of measured CPU on production scientific workflows and 51.7% of container time on a live release gate spanning four repositories. We also prove and verify exact equivalence when the request reaches the whole graph, leaving no unrelated work to avoid. Beyond permission, goal-relative output projection saves up to approximately 90% of a shared step on two third-party test suites while the identical eager control saves 0.0%; the advantage disappears when the omitted output has no other consumer or the request needs it. Ordering, reuse, and pruning can also save cost, but do not replace permission. Finally, we show that current public benchmarks are eager-shaped and contain almost no unrequested work. A pre-registered planning intervention did not broaden them. These findings motivate benchmarks built from standing programs and sequences.
Chinese Translation
当前的智能运行时在行动前进行规划，通常在步骤准备好后执行。我们提出了LazyAgent，一个统一的执行框架，专为智能编写的程序而设计，围绕一个实时的、基于目标的需求集组织。LazyAgent根据请求的输出刷新反向闭包，随着执行状态的变化，仅在活跃目标需要时材料化一个准备好的节点。这种方法用一次线性时间的图分析替代了重复的局部判断，后续进行常数时间的成员测试，使得程序在执行时保持请求特定的同时仍然保持广泛性。在描述超出当前请求需求的程序中，LazyAgent始终优于最强的目标停止贪婪基线，因为它在开始之前拒绝无关的工作。添加一个无关的产品使得贪婪的开销增加22.5%，而LazyAgent的开销保持在0.0%。LazyAgent在生产科学工作流中节省了42.0%的测量CPU时间，并在跨越四个代码库的实时发布门上节省了51.7%的容器时间。我们还证明并验证了当请求覆盖整个图时的精确等价性，从而没有无关的工作需要避免。除了权限，基于目标的输出投影在两个第三方测试套件上节省了大约90%的共享步骤，而相同的贪婪控制则节省了0.0%；当被省略的输出没有其他消费者或请求需要时，这一优势消失。排序、重用和修剪也可以节省成本，但不能替代权限。最后，我们展示了当前的公共基准是贪婪形状的，几乎不包含未请求的工作。一个预注册的规划干预并没有使它们变得更广泛。这些发现促使我们构建基于现有程序和序列的基准。
cs.AI / 39 / 2609.23064
FireWorldBench: Benchmarking Complex Physical World Intelligence through Coupled-Field Fire Dynamics
FireWorldBench：通过耦合场火灾动力学评估复杂物理世界智能的基准测试
Abstract
Understanding the physical world requires more than object recognition, scene description, and short-term visual prediction, as real-world physical systems involve multiple continuous fields, latent causal mechanisms, partial observations, and intervention-sensitive dynamics. We propose FireWorldBench, a benchmark for evaluating complex physical world intelligence in multimodal large language models and agents through coupled-field fire dynamics. Fire provides a canonical stress-test environment, where multiple interacting physical fields jointly shape observable states and temporal dynamics. FireWorldBench is organized along two complementary axes, a physical capability axis and a fire scenario task axis, jointly covering physical-state understanding, temporal dynamics, causal mechanisms, and intervention reasoning. The benchmark comprises 520 fire-world entries, including 494 controlled simulation worlds and 26 real-world-aligned event groups, spanning 47 scene archetypes across 7 environment families. These entries combine structured textual observations, multiple 2D physical-field visualizations, and 3D event-level scene modeling, yielding 9,074 text-image interleaved question-answer pairs across choice-based and open-ended report-generation formats. FireWorldBench evaluates whether models can infer latent physical states, explain underlying mechanisms, forecast coupled-field evolution, and assess intervention consequences from multimodal partial observations, providing a challenging testbed for complex physical world intelligence.
Chinese Translation
理解物理世界不仅仅需要物体识别、场景描述和短期视觉预测，因为现实世界的物理系统涉及多个连续场、潜在因果机制、部分观察和对干预的敏感动态。我们提出了FireWorldBench，这是一个用于评估多模态大型语言模型和智能体在复杂物理世界智能方面的基准测试，基于耦合场火灾动力学。火灾提供了一个典型的压力测试环境，在这个环境中，多个相互作用的物理场共同塑造可观察状态和时间动态。FireWorldBench沿着两个互补的轴线组织，一个是物理能力轴，另一个是火灾场景任务轴，涵盖物理状态理解、时间动态、因果机制和干预推理。该基准测试包含520个火灾世界条目，包括494个受控仿真世界和26个与现实世界对齐的事件组，跨越7个环境家族的47种场景原型。这些条目结合了结构化的文本观察、多种二维物理场可视化和三维事件级场景建模，生成了9,074个文本-图像交错的问答对，涵盖基于选择和开放式报告生成格式。FireWorldBench评估模型是否能够推断潜在的物理状态、解释潜在机制、预测耦合场的演变，并从多模态部分观察中评估干预后果，为复杂物理世界智能提供了一个具有挑战性的测试平台。
cs.AI / 40 / 2609.23071
Tutoring Large Language Models to be Domain-adaptive, Precise and Safe
指导大型语言模型适应领域、精确且安全
Abstract
This thesis proposes a framework for "responsible intelligence" to address AI's critical challenges in safety, ethics, and cultural sensitivity. It advances three core areas: First, it improves domain adaptation in specialized fields using active learning and graph-based knowledge to reduce hallucinations. Second, it enhances ethical rigor via a novel decoding-time alignment mechanism that proactively blocks harmful text generation in real-time. Finally, it ensures cultural and multilingual safety through language-specific steering that respects diverse linguistic and social norms. Ultimately, this work provides a blueprint for building next-generation AI that is contextually knowledgeable, ethically sound, and culturally adaptable.
Chinese Translation
本论文提出了一个“负责任智能”的框架，以应对人工智能在安全、伦理和文化敏感性方面的关键挑战。它在三个核心领域取得了进展：首先，通过主动学习和基于图的知识改善专业领域的领域适应性，以减少幻觉现象。其次，通过一种新颖的解码时间对齐机制增强伦理严谨性，该机制能够实时主动阻止有害文本的生成。最后，通过尊重多样的语言和社会规范的语言特定引导，确保文化和多语言的安全性。最终，这项工作为构建具有上下文知识、伦理合理和文化适应能力的下一代人工智能提供了蓝图。
cs.AI / 41 / 2609.23074
Event Signature Transfer: Model-Agnostic Forecast Scenario Construction from Historical Events
事件特征转移：基于历史事件的模型无关预测场景构建
Abstract
Forecasters often know an event is imminent but not the shape, size, or timing of its effect. We introduce Event Signature Transfer (EST), a training-free, model-agnostic operator that turns a completed past event into an explicit forecast scenario. EST removes a source event's own trend and seasonality, then scales and retimes the remaining event signature onto a native forecast, preserving the forecast's linked structure and reducing to it exactly at zero strength. Because it reads only output quantiles, EST applies to any quantile forecaster, with no training, no model internals, at transfer time. Across twelve real episodes and ten synthetic scenarios on Chronos-2, TimesFM-2.5 and Toto-2.0, manually configured EST reduces real-episode WQL by 21.7-90\% in-sample. On Chronos-2, it leads eleven of twelve matched comparisons against covariate conditioning, activation editing and raw replay. The operator builds a scenario; it does not estimate its likelihood.
Chinese Translation
预测者通常知道某个事件即将发生，但对其影响的形状、大小或时机却不明确。我们提出了事件特征转移（Event Signature Transfer, EST），这是一种无训练、模型无关的操作符，可以将已完成的过去事件转化为明确的预测场景。EST去除了源事件自身的趋势和季节性，然后将剩余的事件特征缩放并重新定时到本地预测上，保持预测的关联结构，并在强度为零时准确减少到该结构。由于它仅读取输出分位数，EST适用于任何分位数预测器，无需训练、无需模型内部信息，在转移时使用。在Chronos-2、TimesFM-2.5和Toto-2.0上的十二个真实事件和十个合成场景中，手动配置的EST在样本内将真实事件的加权量化损失（WQL）降低了21.7%-90%。在Chronos-2上，它在与协变量条件、激活编辑和原始重放的十二次匹配比较中领先了十一项。该操作符构建了一个场景；它并不估计其可能性。
cs.AI / 42 / 2609.23130
From Inference Engine to Inference Control Plane: Connecting vLLM, llm-d, and the Evolution of Efficient Distributed LLM Serving
从推理引擎到推理控制平面：连接 vLLM、llm-d 及高效分布式 LLM 服务的演变
Abstract
Large language model (LLM) inference is evolving from an engine-local optimization problem into a distributed control problem involving reusable state, phase placement, heterogeneous accelerators, networking, autoscaling, reliability, and service-level objectives. This paper connects that transition across peer-reviewed systems research, open-source implementations, and documented production studies. It treats vLLM and llm-d as complementary layers: model-serving engines optimize execution through mechanisms such as PagedAttention, continuous batching, kernels, quantization, and parallelism, while an inference control plane can optimize where, when, and under what policy execution occurs across a fleet. The contribution is synthesis rather than a new benchmark; all reported performance and deployment results remain attributed to their original sources. The combined evidence suggests that the scarce resource in modern inference is shifting from raw FLOPs alone toward managed state, placement, network movement, reliability, and decision quality. We propose an Inference Execution Planner that selects feasible execution plans rather than only endpoints, including aggregated versus disaggregated topology, KV source and transfer action, hardware variant, routing/admission policy, and slower scaling decisions. We also provide a source-local benchmark atlas, a bottleneck-migration taxonomy, practical deployment guidance, an evaluation framework based on SLO-goodput, and research questions for agentic, multimodal, heterogeneous, and resilient inference.
Chinese Translation
大型语言模型（LLM）推理正从一个引擎本地优化问题演变为一个涉及可重用状态、阶段放置、异构加速器、网络、自动扩展、可靠性和服务级目标的分布式控制问题。本文将这一转变与经过同行评审的系统研究、开源实现和文献记录的生产研究相联系。它将 vLLM 和 llm-d 视为互补层次：模型服务引擎通过分页注意力（PagedAttention）、连续批处理、内核、量化和并行等机制优化执行，而推理控制平面则可以优化执行发生的地点、时间和政策。本文的贡献在于综合而非提出新的基准；所有报告的性能和部署结果仍归属于其原始来源。综合证据表明，现代推理中的稀缺资源正从单纯的原始浮点运算（FLOPs）转向管理状态、放置、网络移动、可靠性和决策质量。我们提出了一种推理执行规划器，选择可行的执行计划，而不仅仅是端点，包括聚合与分散拓扑、键值（KV）源和传输动作、硬件变体、路由/接纳政策以及较慢的扩展决策。我们还提供了一个源本地基准地图、瓶颈迁移分类法、实用部署指导、基于服务级目标良好吞吐量（SLO-goodput）的评估框架，以及针对自主、多模态、异构和弹性推理的研究问题。
cs.AI / 43 / 2609.23142
CraftBench-UE: Deterministic Evaluation for Coding Agents in Unreal Engine
CraftBench-UE：在虚幻引擎中对编码代理的确定性评估
Abstract
Building gameplay features in a game engine requires more than code, as code that compiles and runs does not necessarily implement the requested gameplay. We introduce CraftBenchUE, an evaluation harness that runs agents in an isolated Unreal Engine environment, reconstructs their saved submissions in fresh projects, and applies deterministic build, asset, and runtime checks without an LLM judge. Based on the harness, we built a benchmark consisting of 70 tasks spanning C++ source, Blueprint assets, and editor scripting. We evaluate seven models under two editor-tool configurations, with a file-and-shell baseline on C++ tasks. We further pair tasks that specify the same gameplay and use the same runtime tests, but require C++ and Blueprint as the deliverables. Across the 10 paired tasks, C++ completion rates exceed Blueprint by 30.0 and 42.9 percentage points in the two tool configurations. Among on-time Blueprint submissions in this paired set that pass asset checks, 42.2% and 50.0% fail explicit runtime assertions. These submissions satisfy asset requirements but fail the required gameplay tests. We will release the harness, task benchmark, and our trajectory findings with the report.
Chinese Translation
在游戏引擎中构建游戏玩法特性不仅仅需要代码，因为编译并运行的代码不一定实现所请求的游戏玩法。我们介绍了CraftBench-UE，这是一个评估工具，能够在隔离的虚幻引擎环境中运行代理，重建其在新项目中的保存提交，并在没有大型语言模型（LLM）评审的情况下应用确定性的构建、资产和运行时检查。基于该工具，我们构建了一个基准测试，包含70个任务，涵盖C++源代码、蓝图资产和编辑器脚本。我们在两种编辑器工具配置下评估了七个模型，并在C++任务上设定了文件和命令行的基线。我们进一步将指定相同游戏玩法并使用相同运行时测试的任务配对，但要求C++和蓝图作为交付成果。在这10个配对任务中，C++的完成率在两种工具配置下分别比蓝图高出30.0和42.9个百分点。在这一配对集中的按时蓝图提交中，42.2%和50.0%未通过显式运行时断言。这些提交满足资产要求，但未通过所需的游戏玩法测试。我们将发布该工具、任务基准和我们的研究发现报告。
cs.AI / 44 / 2609.23201
Do Not Trust the Benchmark: Limitations of General LLM Rankings and a Case for Task-Specific Evaluation
不要信任基准测试：通用大型语言模型排名的局限性及任务特定评估的必要性
Abstract
Benchmark scores increasingly influence the development, marketing, and selection of large language models (LLMs). Yet an overall score is interpretable only in relation to the system tested, the questions included, and the conditions of evaluation. This perspective examines five connected limitations of general LLM rankings: differences between evaluated and publicly available systems; commercial incentives and dependencies in external evaluation; benchmark saturation, defective tests, and data contamination; models exploiting scoring procedures; and the limited relevance of general scores to users' tasks. Documented cases illustrate why these problems require different responses. I argue for evaluation procedures that disclose the tested configuration, validate questions and successful task completion, report performance alongside cost and execution time, and make the scope of generalization explicit. I then discuss \textbf{Isotanta}, a crowdsourced benchmarking platform, as a practical example of contributed questions and repeated evaluation. A larger question pool may improve task coverage, while repeated sampling can improve the stability of estimates on that pool; neither guarantees validity or personalization. The paper distinguishes the platform's current shared ranking from proposed task-specific and user-provided evaluations. Its central argument is that model selection requires evidence about performance on the intended work, not simply a high position on a general leaderboard.
Chinese Translation
基准测试得分日益影响大型语言模型（LLMs）的开发、营销和选择。然而，整体得分仅能在与被测试系统、所包含问题和评估条件的关系中进行解释。本文探讨了通用LLM排名的五个相关局限性：评估系统与公开可用系统之间的差异；外部评估中的商业激励和依赖关系；基准饱和、缺陷测试和数据污染；模型利用评分程序；以及通用得分与用户任务的相关性有限。文中记录的案例说明了这些问题为何需要不同的应对措施。我主张采用评估程序，公开测试配置，验证问题和成功任务完成情况，报告性能、成本和执行时间，并明确泛化范围。接着，我讨论了 extbf{Isotanta}，一个众包基准测试平台，作为贡献问题和重复评估的实际例子。更大的问题池可能改善任务覆盖，而重复抽样可以提高该池估计的稳定性；但这两者都不能保证有效性或个性化。本文区分了该平台当前共享排名与提议的任务特定和用户提供的评估。其核心论点是，模型选择需要关于预期工作的性能证据，而不仅仅是通用排行榜上的高排名。
cs.AI / 45 / 2609.23293
Expansion Counts under Standard A* Tie-Breaking Strategies on the Final Plateau
标准 A* 平衡策略下的扩展计数在最终平面上的表现
Abstract
In the A* search algorithm, the tie-breaking strategies for nodes with the same $f$-value determines which states A* expands on the final $f$-layer. For nine standard tie-breaking strategies, we show that under a consistent heuristic, every pair has positive-cost instances favoring each strategy over the other by an arbitrarily large additive expansion gap. A parameterized unit-cost grid example also gives unbounded expansion-count ratios between low-$h$ with FIFO and LIFO. In unit-cost search with $h > 0$ at non-goals, exact heuristic values near the goal lead to complementary extremal results: low-$h$ minimizes the number of remaining expansions from a common configuration within the perfect region, while high-$h$ maximizes the total number of expansions when every final-plateau state with $h=1$ is a goal predecessor. Finally, with the evaluation function $f_{\alpha} = g + \alpha h$, when $h>0$ at non-goals, every heuristic weight $0 \leq \alpha<1$ eliminates tie-breaking sensitivity, and all tie-breaking strategies expand the same set of states.
Chinese Translation
在 A* 搜索算法中，对于具有相同 $f$ 值的节点的平衡策略决定了 A* 在最终 $f$ 层扩展的状态。对于九种标准平衡策略，我们证明在一致启发式下，每对策略都有正成本实例，偏向于某一策略而非另一策略，且具有任意大的附加扩展差距。一个参数化的单位成本网格示例也给出了低-$h$ 与 FIFO 和 LIFO 之间的无界扩展计数比。在单位成本搜索中，当非目标节点的 $h > 0$ 时，接近目标的精确启发式值导致互补的极值结果：低-$h$ 最小化来自完美区域内共同配置的剩余扩展数量，而高-$h$ 在每个 $h=1$ 的最终平面状态是目标前驱时最大化扩展总数。最后，使用评估函数 $f_{eta} = g + eta h$，当非目标节点的 $h>0$ 时，每个启发式权重 $0 leq eta<1$ 消除了平衡策略的敏感性，所有平衡策略扩展相同的状态集合。
cs.AI / 46 / 2609.23363
TicTacBench: Benchmarking Timing Closure Capabilities of Coding Agents
TicTacBench：编码智能体时序收敛能力的基准测试
Abstract
Recent advances in large language models (LLMs) have led to the emergence of coding agents capable of performing complex engineering tasks, including register-transfer level (RTL) design and optimization. Existing RTL benchmarks mainly evaluate functional correctness and performance, power, and area (PPA) of the generated RTL designs, leaving agents' ability for \emph{timing closure} under-evaluated. We propose TicTacBench, a benchmark specifically designed to evaluate coding agents' capabilities for RTL-level timing closure under post-place-and-route (post-PnR) evaluation. TicTacBench contains 30 diverse tasks, each provided with a suboptimal RTL design, realistic timing constraints, functional equivalence verification, and timing reports. With over 300 runs of coding agents driven by 8 frontier LLMs, we find that even the best agent can only close 53.3\% of tasks with 7.18\% area-delay product (ADP) degradation and 8.83\% energy-delay-squared product (EDDP) improvement on average. We identify common failure categories that explain why agents fail to close timing. Then we propose TicTacSkill, a new method that guides agents to follow standard timing-closure procedures and improves the Timing Closure Rate by 9\%. These results suggest that while coding agents have made significant progress in RTL design, their timing-closure capability still has substantial room for improvement.
Chinese Translation
近年来，大型语言模型（LLMs）的进展促使编码智能体的出现，这些智能体能够执行复杂的工程任务，包括寄存器传输级（RTL）设计和优化。现有的RTL基准主要评估生成的RTL设计的功能正确性、性能、功耗和面积（PPA），而对智能体在时序收敛方面的能力评估不足。我们提出了TicTacBench，这是一个专门设计用于评估编码智能体在后布局和布线（post-PnR）评估下的RTL级时序收敛能力的基准。TicTacBench包含30个多样化的任务，每个任务都提供了一个次优的RTL设计、现实的时序约束、功能等价性验证和时序报告。通过对8个前沿LLMs驱动的编码智能体进行超过300次的运行，我们发现即使是表现最好的智能体也只能完成53.3%的任务，平均面积-延迟乘积（ADP）恶化7.18%，而能量-延迟平方乘积（EDDP）改善8.83%。我们识别出一些常见的失败类别，解释了智能体为何未能完成时序收敛。随后，我们提出了TicTacSkill，这是一种新方法，指导智能体遵循标准的时序收敛程序，并将时序收敛率提高了9%。这些结果表明，尽管编码智能体在RTL设计方面取得了显著进展，但其时序收敛能力仍有很大的提升空间。
cs.AI / 47 / 2609.23378
Leaky-integrator reconstruction: taming error accumulation in recursive differenced time-series forecasting
泄漏积分重构：驯服递归差分时间序列预测中的误差积累
Abstract
We introduce leaky-integrator reconstruction, a training-free method that cures the error accumulation of recursive differenced forecasting. Our first contribution is diagnostic: predicting one-step changes and integrating them by cumulative summation, the standard remedy for non-stationarity, is a discrete integrator with a pole on the unit circle, and we show this makes recursive rollout of a nonlinear model diverge, its 336-step error reaching several times that of a well-behaved forecaster (normalised MAE 1.6-3.8 versus about 0.8) across every neural architecture tested. Our second, central contribution is the fix: move the pole inside the unit circle with a leaky integrator H(z) = 1/(1 - gamma z^-1), gamma < 1, which provably bounds the accumulated error variance. Applied at reconstruction time with a single fixed gamma=0.9 (no retraining, a two-line change to any deployed one-step or foundation-model forecaster), it shrinks error at every horizon, the mean gain over seven diverging architectures and twenty datasets growing from ~3% at H=24 to 23% at H=96, 37% at H=192 and 51% (43-74% across those architectures) at H=336 (78% with an oracle pole). Crucially, it is provably inert where no pathology exists (stable or joint predictors already at the irreducible rate), making it a safe, general default.
Chinese Translation
我们提出了泄漏积分重构，这是一种无训练的方法，可以解决递归差分预测中的误差积累问题。我们的第一个贡献是诊断性的：通过预测一步变化并通过累积求和将其整合，作为非平稳性的标准补救措施，这是一个在单位圆上有极点的离散积分器，我们展示了这使得非线性模型的递归展开发散，其336步误差达到表现良好的预测器的几倍（标准化平均绝对误差 1.6-3.8 对比约 0.8），在所有测试的神经架构中均是如此。我们的第二个核心贡献是解决方案：通过使用泄漏积分器 H(z) = 1/(1 - gamma z^-1)，其中 gamma < 1，将极点移入单位圆内，这证明可以界定累积误差方差。在重构时应用固定的 gamma=0.9（无需重新训练，仅需对任何已部署的一步或基础模型预测器进行两行更改），它在每个预测范围内都减少了误差，七个发散架构和二十个数据集的平均增益从 H=24 时的 ~3% 增长到 H=96 时的 23%，H=192 时的 37% 和 H=336 时的 51%（在这些架构中为 43-74%），在使用 oracle 极点时为 78%。至关重要的是，在不存在病态的情况下（稳定或联合预测器已经达到不可约率），它被证明是惰性的，使其成为一种安全的通用默认选择。
cs.AI / 48 / 2609.23512
AgentBetta: Verification-Driven Adaptive Configuration of an AI Nano-Agent through Selective Expansion and Verified Contraction
AgentBetta：通过选择性扩展和验证收缩驱动的AI纳米代理的自适应配置验证
Abstract
Large language model agents are typically deployed with predefined configurations, although the required model capability, context, tools, permissions, memory, and computational resources can vary substantially across tasks. This study develops and evaluates AgentBetta, an adaptive AI Nano-Agent framework that represents these factors as an executable configuration and updates them through verification-driven diagnosis, selective expansion, and verification-based counterfactual contraction. The evaluation distinguishes controlled mechanism validation from external agent comparisons. On the AB-ConfigBench benchmark, AgentBetta achieved 91.38% verified success while reducing median context allocation from 64,000 to 8,000 context characters and median tool exposure from five tools to zero compared with the fully provisioned configuration. The configuration-deficiency diagnosis achieved a macro-F1 score of 0.819 with precision of 1.000 across the evaluated dimensions, and selective expansion avoided unnecessary changes to unrelated configuration dimensions. Post-success contraction preserved verification outcomes in 56.41% of evaluated one-dimension contraction probes, indicating that some successful configurations contained removable capability under the tested conditions. External evaluations indicate that adaptive configuration can improve the balance between verified task completion and capability exposure; however, the results vary across benchmarks and agent families. In particular, the cross-family replication did not reproduce the primary-backbone accuracy ordering, and specialized systems remained advantageous for certain task domains. These results support interpreting AgentBetta as a configuration-adaptation mechanism that regulates capability allocation and inference expenditure rather than as a universal replacement for specialized agent architectures.
Chinese Translation
大型语言模型代理通常以预定义的配置进行部署，尽管所需的模型能力、上下文、工具、权限、内存和计算资源在不同任务中可能有显著差异。本研究开发并评估了AgentBetta，一个自适应AI纳米代理框架，该框架将这些因素表示为可执行配置，并通过验证驱动的诊断、选择性扩展和基于验证的反事实收缩进行更新。评估区分了受控机制验证与外部代理比较。在AB-ConfigBench基准测试中，AgentBetta实现了91.38%的验证成功率，同时将中位上下文分配从64,000个字符减少到8,000个字符，将中位工具暴露从五个工具减少到零，与完全配置的情况相比。配置缺陷诊断在评估维度上实现了0.819的宏F1分数，精确度为1.000，选择性扩展避免了对无关配置维度的不必要更改。成功后的收缩在56.41%的评估单维收缩探针中保留了验证结果，表明在测试条件下某些成功配置中包含可移除的能力。外部评估表明，自适应配置可以改善验证任务完成与能力暴露之间的平衡；然而，结果在不同基准和代理家族之间有所不同。特别是，跨家族复制未能重现主要骨干的准确性排序，且专用系统在某些任务领域仍然具有优势。这些结果支持将AgentBetta解释为一种配置适应机制，调节能力分配和推理支出，而不是作为专用代理架构的普遍替代品。
cs.AI / 49 / 2609.23640
Are Human-Aligned Models Models of Humans? A Turing-Test Gap in Preference Alignment
人类对齐模型是否是人类的模型？偏好对齐中的图灵测试差距
Abstract
Human-feedback alignment has made language models useful assistants and is commonly described as aligning them with humans. However, the responses people prefer from an AI need not be the responses they themselves would give. We distinguish alignment with human preferences from alignment with human behavior, and show that alignment with human preferences can make model behavior less human-like even when both preferences and responses come entirely from humans. We call this the Turing-test gap. We show that preference alignment preserves the human response distribution only under a restrictive condition, and find no consistent evidence that real human preferences satisfy it. Empirically, the loss of human-response likelihood increases with the strength of preference weighting, regardless of its direction, and the gap also appears under standard DPO. These results establish human-likeness as an explicit dimension of alignment rather than something assumed to follow from preference alignment.
Chinese Translation
人类反馈对齐使语言模型成为有用的助手，通常被描述为与人类对齐。然而，人们对人工智能的偏好回应不一定是他们自己会给出的回应。我们将与人类偏好的对齐与与人类行为的对齐区分开来，并展示即使偏好和回应完全来自人类，与人类偏好的对齐也可能使模型行为变得不那么类似人类。我们称之为图灵测试差距。我们表明，偏好对齐仅在一个限制条件下才能保持人类回应分布，并且没有找到一致的证据表明真实的人类偏好满足该条件。从经验上看，无论偏好权重的方向如何，人类回应的可能性损失随着偏好权重的增强而增加，并且在标准 DPO 下也出现了这一差距。这些结果确立了人类相似性作为对齐的一个明确维度，而不是假设它是由偏好对齐所导致的。
cs.AI / 50 / 2609.23695
PhysAI-Bench: A Benchmark for LLM-Based Agentic Decision-Making in Autonomous UAV-Centric Physical AI
PhysAI-Bench：一个用于评估基于大型语言模型的自主无人机中心物理人工智能中的代理决策能力的基准
Abstract
Recent advances in Physical AI have accelerated the use of foundation models in autonomous systems such as unmanned aerial vehicles (UAVs), which must perceive, reason, plan, and act in dynamic environments. Existing benchmarks assess physical perception, intuitive physics, embodied navigation, and collaborative reasoning, but rarely evaluate the agentic decision-making required for reliable autonomy. We introduce \textit{PhysAI-Bench}, a benchmark for evaluating this capability. It contains 10,178 standardized decision instances automatically extracted from conversational traces of autonomous UAV missions. Each instance preserves mission context, temporal dependencies, physical constraints, Model Context Protocol (MCP) tool calls, Agent-to-Agent (A2A) interactions, sensor observations, and AI-native 6G network conditions, including latency, packet loss, throughput, edge load, and network slicing. We expose only information preceding each decision, preventing future-event leakage and approximating online decision-making. We evaluate 29 foundation models using a two-stage protocol. We select model-specific configurations from 12 combinations of zero-, three-, and five-shot prompting and four temperatures, tested in three runs on a 35-instance, human-verified development set. We then freeze each selected configuration and evaluate it in three runs on a fixed, episode-disjoint set of 500 instances. GPT-5.3 achieves the highest accuracy (52.00%), followed by GPT-5.2 (49.40%) and Grok~4.5 (49.07%). Few-shot prompting generally improves performance, while temperature has limited influence. The results demonstrate that reliable agentic decision-making in Physical AI remains an open challenge. The dataset is available at https://github.com/maferrag/physai-bench
Chinese Translation
近期物理人工智能的进展加速了基础模型在自主系统（如无人机（UAV））中的应用，这些系统必须在动态环境中感知、推理、规划和行动。现有基准评估物理感知、直观物理、具身导航和协作推理，但很少评估可靠自主所需的代理决策能力。我们引入了 extit{PhysAI-Bench}，一个用于评估这一能力的基准。它包含10,178个标准化决策实例，这些实例是从自主无人机任务的对话记录中自动提取的。每个实例保留了任务上下文、时间依赖性、物理约束、模型上下文协议（Model Context Protocol, MCP）工具调用、代理间（Agent-to-Agent, A2A）交互、传感器观测以及AI原生6G网络条件，包括延迟、数据包丢失、吞吐量、边缘负载和网络切片。我们仅暴露每个决策之前的信息，以防止未来事件泄漏，并近似在线决策。我们使用两阶段协议评估29个基础模型。从12种零-shot、三-shot和五-shot提示及四种温度的组合中选择模型特定配置，在一个35个实例的人类验证开发集上进行了三轮测试。然后，我们冻结每个选定配置，并在一个固定的、事件不重叠的500个实例集上进行三轮评估。GPT-5.3取得了最高准确率（52.00%），其次是GPT-5.2（49.40%）和Grok~4.5（49.07%）。少量提示通常提高了性能，而温度的影响有限。结果表明，物理人工智能中的可靠代理决策仍然是一个未解决的挑战。数据集可在https://github.com/maferrag/physai-bench获取。
cs.AI / 51 / 2609.23735
ScholarStack: Layered Research Asset Orchestration and Cross-Task Reuse for Scientific Agents
ScholarStack：科学代理的分层研究资产编排与跨任务重用
Abstract
Scientific agents support a range of literature-based research tasks, such as retrieval, question answering, evidence-grounded generation, and claim assessment. Most existing systems, however, are organized around individual tasks: the same papers are repeatedly retrieved, segmented, and interpreted, and the understanding built in one task is difficult to reuse in the next. We present ScholarStack, a layered research asset framework that compiles a paper collection into reusable, versioned, and provenance-preserving assets at three complementary levels: source-grounded paper-level statements, domain-level organization, and evidence-grounded cross-paper syntheses. A common access interface returns task-specific views at the evidence granularity each task requires, preserving study conditions, source traceability, and verification status. We instantiate the framework on four task families spanning ten task settings, comparing agents that use the compiled assets with task-specific baselines under matched base models. Quality gains concentrate on tasks that require cross-paper evidence, such as multi-paper question answering and literature review generation, and query-time token cost falls on every task where it is measured, with assets compiled once and reused across tasks. These results suggest that layered research assets can serve as shared infrastructure for scientific agents, shifting literature-based assistance from isolated document processing toward cumulative, evidence-grounded workflows.
Chinese Translation
科学代理支持一系列基于文献的研究任务，如检索、问答、基于证据的生成和主张评估。然而，大多数现有系统围绕单一任务组织：相同的论文被反复检索、分段和解释，而在一个任务中建立的理解在下一个任务中难以重用。我们提出了ScholarStack，一个分层研究资产框架，将论文集合编译成可重用的、版本化的和保留来源的资产，分为三个互补层次：基于来源的论文级陈述、领域级组织和基于证据的跨论文综合。一个公共访问接口在每个任务所需的证据粒度下返回特定于任务的视图，保留研究条件、来源可追溯性和验证状态。我们在涵盖十个任务设置的四个任务家族上实例化该框架，比较使用编译资产的代理与在匹配基础模型下的特定任务基线。质量提升集中在需要跨论文证据的任务上，如多论文问答和文献综述生成，而在每个测量的任务中查询时的令牌成本下降，资产一次编译并在任务间重用。这些结果表明，分层研究资产可以作为科学代理的共享基础设施，将基于文献的辅助从孤立的文档处理转向累积的、基于证据的工作流程。
cs.AI / 52 / 2609.23774
On Probabilistic Inference Through Parametric Tensor Decomposition in Base Tensor Networks
通过参数张量分解在基本张量网络中的概率推断
Abstract
Probabilistic inference is generally only tractable in low-treewidth graphical models, limiting its effective applicability in high-treewidth settings. Many existing methods improve efficiency by exploiting specific parametric structure, such as symmetries. However, they typically require such structure to be explicitly present, limiting their applicability to a broader range of graphical models. To address this limitation, we propose a framework where tractable inference is controlled by latent parametric structure exploitation, rather than requiring it to be explicitly present a priori. Our approach first reparameterises a graphical model as a specific tensor network representation, which we call a base tensor network. This representation yields two key properties that allow inference tractability to be controlled by parametric structure: 1) First, the complexity of inference is mainly determined by the parametric structure of a single tensor, called the base tensor. We characterise several tractable classes of base tensors for which the entire base tensor network can be contracted efficiently. 2) Second, decomposing the base tensor yields again a collection of base tensor networks. This allows inference to be naturally reduced to decomposing the base tensor into tractable components with sufficient parametric structure. We call this procedure parametric tensor decomposition. By exploiting parametric structure within the base tensor, our framework enables a novel view on inference beyond settings where such structure is explicitly present.
Chinese Translation
概率推断通常仅在低树宽图模型中可行，这限制了其在高树宽环境中的有效应用。许多现有方法通过利用特定的参数结构（如对称性）来提高效率。然而，它们通常要求这种结构明确存在，从而限制了其在更广泛图模型中的适用性。为了解决这一限制，我们提出了一个框架，其中可处理的推断是通过潜在的参数结构利用来控制的，而不是要求其在先验上明确存在。我们的方法首先将图模型重新参数化为一种特定的张量网络表示，我们称之为基本张量网络。这种表示具有两个关键特性，允许通过参数结构控制推断的可处理性：1）首先，推断的复杂性主要由一个称为基本张量的单一张量的参数结构决定。我们描述了几类可处理的基本张量，对于这些张量，整个基本张量网络可以高效收缩。2）其次，分解基本张量再次产生一组基本张量网络。这使得推断自然地简化为将基本张量分解为具有足够参数结构的可处理组件。我们称这一过程为参数张量分解。通过利用基本张量中的参数结构，我们的框架为推断提供了一种超越明确存在此类结构的环境的新视角。
cs.AI / 53 / 2609.23790
Total Cost of Agency: Exact Attribution of Memory Injection Cost in Multi-Agent LLM Workflows
代理的总成本：多代理大语言模型工作流中内存注入成本的精确归因
Abstract
Every node in a multi-agent large language model (LLM) workflow retrieves context from memory and injects it into its prompt, where those injected tokens are billed as input tokens at the same per-token price as the system prompt and the user query. Production observability tools report total token cost but do not separate the tokens a node generates from the tokens it is handed, so this component of the bill is invisible to the teams paying it. We introduce the Total Cost of Agency (TCA), a decomposition of multi-agent workflow cost into base prompt, inference, memory injection, miss penalty and context-accumulation components, and an exact attribution method: a two-pass, non-billable token count that measures injected tokens directly rather than estimating them from word-count proxies. On a 200-task enterprise benchmark executed against real model APIs, memory injection accounts for 13.6 percent of the variable cost a compile-time optimizer can act on, about 12 percent of the full billed cost, and its share rises from a structural zero at workflow depth one to 27.6 percent at depth six. Injected tokens grow linearly with depth over the measured range (R^2 = 0.9974, depths two through six); a quadratic fit yields a negative leading coefficient, so the data do not exhibit convex growth at these depths. We show the component is controllable at fixed model tier: reducing the retrieval window capacity from 32 to 2 entries lowers injected tokens by 28.7 percent with an accuracy change within seed-level variation. We report in full that our graph-rewriting transforms are approximately cost-neutral in isolation, that two of the five decomposition terms are zero by construction in this harness, and that total workflow cost is dominated by model tier assignment, which we hold fixed and treat as prior work. Prompt caching is not evaluated; all figures are for the uncached case.
Chinese Translation
在多代理大语言模型（LLM）工作流中，每个节点从内存中检索上下文并将其注入到提示中，这些注入的标记以与系统提示和用户查询相同的每标记价格计费。生产可观察性工具报告总标记成本，但未能将节点生成的标记与其接收到的标记分开，因此这一账单的组成部分对支付账单的团队来说是不可见的。我们引入了代理的总成本（Total Cost of Agency, TCA），将多代理工作流成本分解为基础提示、推理、内存注入、未命中惩罚和上下文累积组件，并提出了一种精确的归因方法：一种两次传递的、不可计费的标记计数，直接测量注入的标记，而不是通过字数代理进行估算。在对真实模型API执行的200任务企业基准测试中，内存注入占编译时优化器可以作用的可变成本的13.6%，约占全部计费成本的12%，其份额从工作流深度为1时的结构性零增长到深度为6时的27.6%。在测量范围内，注入的标记随着深度线性增长（R^2 = 0.9974，深度为2到6）；二次拟合得出的负主系数表明，在这些深度下数据并未表现出凸增长。我们展示了在固定模型层级下该组件是可控的：将检索窗口容量从32降低到2条目可使注入的标记减少28.7%，而准确性变化在种子级别的变异范围内。我们完整报告了我们的图重写变换在孤立情况下大致是成本中性的，该框架中五个分解项中的两个是由构造决定的零，并且总工作流成本主要由模型层级分配主导，我们将其固定并视为先前工作。未评估提示缓存；所有数据均为未缓存情况。
cs.AI / 54 / 2609.23806
WorkWorlds: An Infrastructure for Evaluating AI Agents on Workplace Tasks
工作世界：评估人工智能代理在工作场所任务上的基础设施
Abstract
Many knowledge-work benchmarks are constructed around individual tasks, with the context needed for each task selected together with or after the task has been specified. This design measures performance on workplace-like tasks in an environment assembled for the task. When task specification guides which context is selected, the evaluation can encode task information into the environment and pre-complete part of the information-localization work that workplace performance normally requires. We introduce WorkWorlds, an evaluation infrastructure that separates organizational state from task specification. A world first fixes a revision, date, and employee seat and materializes the organizational state that employee can access; tasks are introduced only afterward. We implement WorkWorlds in a primary synthetic pharmaceutical company with 8 measured tasks across 6 employee seats, and construct additional organizational worlds. Across 192 matched evaluations, moving from task-curated context to the full role-visible workplace reduced evidence access from 90.4% to 74.5% and criterion pass from 79.4% to 68.2%, while pass conditional on evidence access remained nearly unchanged; most of the measured difference occurred before the agent reached sufficient evidence.
Chinese Translation
许多知识工作基准是围绕单个任务构建的，所需的上下文在任务指定之前或之后一起选择。这种设计在为任务组装的环境中测量工作场所类任务的表现。当任务规范指导选择上下文时，评估可以将任务信息编码到环境中，并预先完成工作场所表现通常需要的一部分信息定位工作。我们介绍了WorkWorlds，一种将组织状态与任务规范分离的评估基础设施。一个世界首先固定修订、日期和员工席位，并具体化员工可以访问的组织状态；任务仅在此之后引入。我们在一个主要的合成制药公司中实施了WorkWorlds，该公司在6个员工席位上测量了8个任务，并构建了其他组织世界。在192次匹配评估中，从任务策划的上下文转向完整的角色可见工作场所，证据访问率从90.4%降至74.5%，标准通过率从79.4%降至68.2%，而基于证据访问的通过率几乎保持不变；大多数测量差异发生在代理达到足够证据之前。
cs.AI / 55 / 2609.23860
Pretraining of Medical Visual Encoders Toward Multi-modal Large Language Models
面向多模态大型语言模型的医学视觉编码器预训练
Abstract
Multimodal Large Language Models (MLLMs) commonly reuse visual encoders pretrained with CLIP, although the features of these ViTs are ultimately consumed by autoregressive LLMs. We refer to this mismatch as the semantic-interface gap and introduce MedMLIP, a framework that pretrains the visual encoder through report generation with a frozen LLM, while employing Local Relational Distillation (LRD) to preserve relationships among visual patches to avoid visual collapse. We pretrain MedMLIP on IU-Xray and Open-PMC-300K and evaluate the resulting encoders on VQA-RAD and SLAKE. Only the ViT is transferred, while the guiding LLM and projector are replaced, allowing us to assess cross-LLM transferability. Our cross-LLM transfer experiments demonstrate the value of pretraining visual encoders for their autoregressive LLM interface while trying to preserve more fine-grained visual information. Code and the pretrained model are available at https://github.com/SkyCol/MedMLIP
Chinese Translation
多模态大型语言模型（MLLMs）通常重用使用CLIP预训练的视觉编码器，尽管这些视觉变换器（ViTs）的特征最终被自回归大型语言模型（LLMs）所消耗。我们将这种不匹配称为语义接口差距，并引入MedMLIP，一个通过冻结的LLM进行报告生成来预训练视觉编码器的框架，同时采用局部关系蒸馏（Local Relational Distillation, LRD）来保持视觉补丁之间的关系，以避免视觉崩溃。我们在IU-Xray和Open-PMC-300K上预训练MedMLIP，并在VQA-RAD和SLAKE上评估得到的编码器。仅转移ViT，而指导LLM和投影器被替换，从而使我们能够评估跨LLM的可转移性。我们的跨LLM转移实验展示了为自回归LLM接口预训练视觉编码器的价值，同时努力保留更多细粒度的视觉信息。代码和预训练模型可在https://github.com/SkyCol/MedMLIP获取。
cs.AI / 56 / 2609.23877
Explainable Recommendations at Scale: LLM Rationales for YouTube Music Artist Discovery
大规模可解释推荐：用于 YouTube Music 艺术家发现的 LLM 理由
Abstract
Modern music streaming platforms face a persistent tradeoff: exploiting familiar content versus driving the exploration of novel items. While users frequently desire discovery, they hesitate to select unknown artists over proven favorites. Providing transparent, natural language rationales that explain why an unexplored item is recommended lowers this barrier. However, while Large Language Models (LLMs) excel at this nuanced explainability, their real-time deployment is severely bottlenecked by prohibitive inference costs and computational overhead. In this paper, we present an industry case study of a decoupled recommendation architecture that successfully scales exploration without compromising latency. Our system isolates LLM inference asynchronously offline, pre-computing personalized candidate pools of undiscovered artists alongside tailored rationales. Large-scale online A/B experiments validate our design. We demonstrate that combining LLM-backed recommendations with these explanatory rationales significantly reduces the trust barrier for new content, yielding statistically significant improvements in both user exploration and overall engagement on the discovery surfaces.
Chinese Translation
现代音乐流媒体平台面临着一个持续的权衡：利用熟悉的内容与推动新项目的探索。尽管用户经常渴望发现新内容，但他们在选择未知艺术家与已知喜爱的艺术家之间犹豫不决。提供透明的自然语言理由，解释为何推荐未探索的项目，可以降低这一障碍。然而，尽管大型语言模型（LLMs）在这种细致的可解释性方面表现出色，但其实时部署受到高昂推理成本和计算开销的严重制约。本文呈现了一个成功扩展探索而不妥协延迟的解耦推荐架构的行业案例研究。我们的系统异步离线隔离 LLM 推理，预计算个性化的未发现艺术家候选池及其定制理由。大规模在线 A/B 实验验证了我们的设计。我们证明，将 LLM 支持的推荐与这些解释性理由相结合，显著降低了用户对新内容的信任障碍，带来了用户探索和整体参与度在发现界面上的统计显著提升。
cs.AI / 57 / 2609.23917
Increasing Skill Level Recruits Deeper Attention Layers in a Frozen Chess Transformer
提高技能水平招募更深层次的注意力层在冻结的国际象棋变换器中
Abstract
Chess involves complex reasoning in a deterministic environment, which makes it a useful setting for studying the mechanisms of computation inside transformers. The Maia-3 chess transformer takes Elo, a measure of competitive chess skill, as an input to the pre-trained network, so we can vary the skill the network is conditioned on with no change to its weights. Here we investigate how turning this skill dial affects self-attention. Ablating every attention head at every Elo from 700 to 2500, we find 1) increasing skill pushes the causal center of mass of the computation deeper, monotonically, for every chess piece and move type we measured; 2) the depth migration is much greater for specific tactics, especially knight forks, than for other move types; 3) the migration consists of deeper heads getting recruited for more specialized computations while one shared shallow head keeps a roughly constant contribution. These results may shed light on how conditioning inputs redistribute computation in larger transformers.
Chinese Translation
国际象棋涉及在确定性环境中进行复杂推理，这使其成为研究变换器内部计算机制的有用场景。Maia-3国际象棋变换器将Elo（一个衡量竞争性国际象棋技能的指标）作为输入到预训练网络中，因此我们可以在不改变其权重的情况下，调整网络所依赖的技能水平。在这里，我们研究了调整这一技能水平如何影响自注意力。通过对每个Elo从700到2500的每个注意力头进行消融实验，我们发现：1）随着技能水平的提高，计算的因果质心单调地向更深层次迁移，适用于我们测量的每个棋子和走法类型；2）对于特定战术，尤其是骑士叉，深度迁移远大于其他走法类型；3）这种迁移由更深的注意力头被招募以进行更专业的计算，而一个共享的浅层头保持大致恒定的贡献。这些结果可能为理解输入条件如何在更大的变换器中重新分配计算提供了新的视角。
cs.AI / 58 / 2609.23945
Echo State Network (ESN) for Signal Recovery in RF-Impaired IBFD MIMO Systems
用于射频损伤的全双工多输入多输出系统信号恢复的回声状态网络（ESN）
Abstract
In-band full-duplex (IBFD) multiple-input multiple-output (MIMO) systems enable simultaneous transmission and reception on the same frequency band, improving spectral efficiency for next-generation wireless networks. However, IBFD-MIMO systems are susceptible to self-interference (SI), which may overpower signals of interest (SOI). In this scenario, blind source separation (BSS) algorithms can be adopted to remove SI and perform joint sensing and communication (JSAC), but BSS algorithms mostly assume an idealized linear and quasi-stationary signal model, which does not hold under realistic radio frequency (RF) impairments, such as I/Q imbalance, carrier frequency offset (CFO), phase noise, and power amplifier nonlinearity. This paper proposes a two-stage echo state network (ESN)-based scheme that is superior to BSS under these realistic conditions. A frozen ESN is trained offline to characterize the static SI path, while an adaptive ESN, updated online via recursive least squares, tracks the time-varying SOI path using sparse pilot symbols. We evaluate the proposed scheme's SOI recovery performance and acquisition speed with different block sizes, comparing it against other recurrent neural networks (RNN), such as long short-term memory (LSTM) and gated recurrent unit (GRU). Simulation results show that the proposed approach outperforms BSS, LSTM, and GRU in both efficiency and SOI recovery, demonstrating the viability of ESNs for real-time, nonlinear self-interference cancellation in realistic IBFD MIMO systems.
Chinese Translation
带内全双工（IBFD）多输入多输出（MIMO）系统能够在同一频带上实现同时传输和接收，从而提高下一代无线网络的频谱效率。然而，IBFD-MIMO系统易受自干扰（SI）的影响，这可能会掩盖感兴趣信号（SOI）。在这种情况下，可以采用盲源分离（BSS）算法来消除SI并执行联合感知与通信（JSAC），但BSS算法大多假设理想化的线性和准平稳信号模型，这在现实的射频（RF）损伤下并不成立，例如I/Q不平衡、载波频率偏移（CFO）、相位噪声和功率放大器非线性。本文提出了一种基于两阶段回声状态网络（ESN）的方案，在这些现实条件下优于BSS。一个冻结的ESN在离线训练中被用来表征静态SI路径，而一个自适应的ESN通过递归最小二乘法在线更新，利用稀疏导频符号跟踪时变的SOI路径。我们评估了所提方案在不同块大小下的SOI恢复性能和获取速度，并与其他递归神经网络（RNN）进行比较，如长短期记忆网络（LSTM）和门控递归单元（GRU）。仿真结果表明，所提方法在效率和SOI恢复方面均优于BSS、LSTM和GRU，证明了ESN在现实IBFD MIMO系统中进行实时非线性自干扰消除的可行性。
cs.AI / 59 / 2609.23953
Agents That Edit Documents: Measuring Agentic PDF Forgery Against a Non-Agentic Control
编辑文档的智能体：针对非智能体控制的智能体PDF伪造测量
Abstract
AI agents that carry a multi-step computer task through on their own became ordinary tools in the past year, and the same autonomy is available to anyone whose task is harmful. We ask what that means for a relying party -- an insurer, a lender, an auditor -- whose evidence is a filed PDF. AgentForge-Bench measures how reliably an off-the-shelf coding agent, driving one of seven open-weight models with a shell and the stock Python PDF stack, alters one dollar amount, date or address in a real filed financial document from a single sentence of intent, graded by rules rather than by a model. Across 1,750 cells, 1,419 (81.1%) satisfy the verifier, and 808 (46.2%) also survive every stricter filter: visible, localized, typeface-matched, original value gone document-wide. A deterministic script with no model in it solves 98 of the 125 documents; the agents solve 124, and none the script solves alone. Agents misreport 41% of their wrong edits as done, no model refused, and the cheapest verified forgery costs 2.4 cents. The raw rate overstates the threat by about a factor of two; the strict rate is still large.
Chinese Translation
在过去一年中，能够独立完成多步骤计算任务的人工智能智能体已成为普通工具，而这种自主性也可被任何有害任务的执行者所利用。我们探讨这对依赖方（如保险公司、贷款方、审计师）意味着什么，他们的证据是一份提交的PDF文件。AgentForge-Bench测量了一个现成编码智能体如何可靠地通过一个意图句子，驱动七个开放权重模型中的一个，利用Shell和标准Python PDF栈，修改真实提交的财务文档中的一个金额、日期或地址。根据规则而非模型进行评分。在1750个单元中，有1419个（81.1%）满足验证者的要求，808个（46.2%）在每个更严格的过滤条件下也能通过：可见性、局部性、字体匹配、原始值在整个文档中消失。一个不包含模型的确定性脚本解决了125个文档中的98个；智能体解决了124个，而没有一个是脚本单独解决的。智能体错误报告了41%的错误编辑为已完成，没有模型拒绝，且最便宜的验证伪造成本为2.4美分。原始比率将威胁夸大了约两倍；严格比率仍然很大。
cs.AI / 60 / 2609.23957
Divergent strategies and convergent outcomes in autonomous materials discovery
自主材料发现中的分歧策略与趋同结果
Abstract
Scientific agents are mostly evaluated on whether they complete tasks or recover known results; we instead study variation across repeated open-ended campaigns. Sixteen separately initialized sessions of one model-harness configuration received a frozen database of 12,499 metal-organic frameworks, a methane-storage objective, a pinned protocol and a one-week budget. Strategies diverged into four approaches spanning 100--5,000 screened structures, and eight built 2,253 hypothetical structures. Yet the agents recovered the same materials frontier near 200 cm^3/cm^3, and an independent calculation of the database's porous region found its nine best structures all among their reports. Enforced checks on half the agents raised fresh-run reproduction from one of eight to eight of eight but could not detectably improve conclusion validity, because fifteen of sixteen agents selected the same audit-excluded entry, an incomplete structure whose missing anions created artificial pore volume. Replicated agents thus reveal both robust conclusions and common-mode errors from shared inputs.
Chinese Translation
科学代理通常根据其是否完成任务或恢复已知结果来进行评估；而我们则研究在重复的开放式实验中所产生的变异。十六个独立初始化的模型配置会话接收了一个包含12,499个金属有机框架的冻结数据库，设定了甲烷储存目标、固定协议以及一周的预算。策略分化为四种方法，筛选了100至5,000个结构，其中八个会话构建了2,253个假设结构。然而，所有代理均在接近200 cm^3/cm^3的材料前沿上达成一致，并且对数据库的多孔区域进行的独立计算发现其九个最佳结构均在其报告中。对一半代理施加的强制检查将新运行的再现率从八分之一提高到八分之八，但未能显著提高结论的有效性，因为十六个代理中的十五个选择了相同的审计排除条目，即一个不完整的结构，其缺失的阴离子造成了人工孔体积。因此，复制的代理揭示了基于共享输入的既稳健的结论又存在的共同模式错误。
cs.AI / 61 / 2609.23971
UniK: Universal Knowledge Perception for Digital and Physical AI
UniK：数字与物理人工智能的通用知识感知
Abstract
Two transformative classes of AI systems are reshaping how organizations operate: \textit{digital AI}, which reasons over enterprise knowledge to power chatbots and agent workflows; and \textit{physical AI}, which learns to control robots and autonomous systems from video, gameplay, and sensor telemetry. Both face the same foundational bottleneck: raw knowledge at scale, spanning heterogeneous modalities, locked in private corpora that existing AI infrastructure cannot access reliably or efficiently. We propose \textit{Universal Knowledge Perception (UniK)} as a common platform for both classes, covering the full knowledge lifecycle (ingestion, enrichment, indexing, retrieval, and continuous evaluation) across modalities from rich text and video to molecular data and sensor telemetry. We present UniK, built on Polymath Retrieval (multi-index fusion over automatically enriched indices) with no task-specific fine-tuning. Across five digital AI domains (medical literature, open-domain QA, chemistry, legal video proceedings, and government open data) UniK combined with an open-source 70-billion-parameter model consistently matches or outperforms frontier proprietary LLMs that are orders of magnitude larger: 76\% RAG accuracy on government data versus 47\% for GPT-5; 77.9\% on medical QA without fine-tuning; topping all open-source chemistry pipelines. We show that the same infrastructure directly addresses the data curation, indexing, and retrieval challenges facing physical AI world model training, where the knowledge problem is harder but structurally identical.
Chinese Translation
两类变革性的人工智能系统正在重塑组织的运作方式： extit{数字人工智能}，它利用企业知识进行推理，以驱动聊天机器人和代理工作流程；以及 extit{物理人工智能}，它通过视频、游戏和传感器遥测学习控制机器人和自主系统。这两者面临着相同的基础瓶颈：大规模的原始知识，跨越异构模态，锁定在现有人工智能基础设施无法可靠或高效访问的私有语料库中。我们提出 extit{通用知识感知（UniK）}作为这两类系统的共同平台，涵盖从丰富文本和视频到分子数据和传感器遥测的知识生命周期的全部过程（摄取、丰富、索引、检索和持续评估）。我们展示了UniK的构建，基于多索引融合的Polymath Retrieval（自动丰富索引的多索引融合），无需特定任务的微调。在五个数字人工智能领域（医学文献、开放域问答、化学、法律视频程序和政府开放数据）中，UniK与一个开源的700亿参数模型结合，始终能够与规模大几个数量级的前沿专有大型语言模型（LLMs）相匹配或超越：在政府数据上，UniK的RAG准确率为76\%，而GPT-5为47\\%；在医学问答中，未经微调的准确率为77.9\\%；在所有开源化学管道中名列前茅。我们展示了相同的基础设施可以直接解决物理人工智能世界模型训练中面临的数据整理、索引和检索挑战，尽管知识问题更为复杂，但结构上是相同的。
cs.AI / 62 / 2609.23974
LEAP-NBV: Lightweight Edge Active-Perception for Foundation-Model Next-Best-View Planning
LEAP-NBV：轻量级边缘主动感知用于基础模型的下一最佳视角规划
Abstract
Foundation models are endowing autonomous systems with greater intelligence, enabling a more comprehensive understanding of the environment through visual perception. A representative example is Human Mesh Recovery (HMR), which provides useful estimates of a target's 3D pose and shape that can benefit tactical missions. However, the size and power demands of such models make them difficult to run on edge platforms and limit their real-time performance, undermining the requirements of tactical edge deployment - especially for active perception, where a mobile robot must plan its next-best view on-board and cannot offload computation under contested communications. We present LEAP-NBV, a lightweight active-perception framework that runs foundation-model-driven Next-Best-View (NBV) planning on-board an edge device. To this end, we distill a family of large HMR teachers, each into a compact 32M student, with an offline mesh objective, then quantize the vision encoder to FP16 and characterize its on-device accuracy and latency. Within an occlusion-aware active perception loop, we evaluate all configurations on the same held-out benchmark and deploy the end-to-end pipeline on an NVIDIA Jetson Xavier NX, reporting measured on-device latency and energy. Distillation recovers 6-7 mm of Procrustes-aligned mean per-vertex position error (PA-MPVPE) over the undistilled student on the test set. Selecting the edge-optimal compression model brings the HMR engine to ~12 ms at a small accuracy cost and runs the full closed loop at 3.6 FPS and 2.6 J per frame, achieving a 2.0x speedup and 3.0x lower energy than the uncompressed model while nearly matching downstream task quality.
Chinese Translation
基础模型赋予自主系统更强的智能，使其通过视觉感知对环境有更全面的理解。一个典型的例子是人类网格恢复（Human Mesh Recovery, HMR），它提供了目标三维姿态和形状的有用估计，能够为战术任务提供帮助。然而，这类模型的体积和功耗要求使其难以在边缘平台上运行，并限制了其实时性能，削弱了战术边缘部署的需求——尤其是在主动感知中，移动机器人必须在设备上规划其下一最佳视角，无法在通信受限的情况下卸载计算。我们提出了LEAP-NBV，一个轻量级的主动感知框架，能够在边缘设备上运行基础模型驱动的下一最佳视角（Next-Best-View, NBV）规划。为此，我们将一系列大型HMR教师模型提炼为紧凑的32M学生模型，采用离线网格目标，然后将视觉编码器量化为FP16，并表征其在设备上的准确性和延迟。在一个考虑遮挡的主动感知循环中，我们在相同的保留基准上评估所有配置，并在NVIDIA Jetson Xavier NX上部署端到端管道，报告测得的设备延迟和能耗。提炼在测试集上恢复了6-7毫米的Procrustes对齐的每顶点位置误差均值（PA-MPVPE），相较于未提炼的学生模型。选择边缘最优压缩模型使HMR引擎的延迟降至约12毫秒，带来小幅准确性损失，并以3.6帧每秒（FPS）和每帧2.6焦耳的能耗运行完整的闭环，实现了2.0倍的加速和3.0倍更低的能耗，同时几乎匹配下游任务的质量。
cs.AI / 63 / 2609.23986
Jev-Mem: System-One-Controlled Agentic Memory for Efficient AI Agents
Jev-Mem：用于高效人工智能代理的系统一控制代理记忆
Abstract
Agentic memory is becoming essential for long-horizon AI agents, yet many existing systems rely on autoregressive LLMs to control how memories are organized, retrieved, and used, placing expensive generation on the critical path of memory operations. We introduce \textbf{\method}, a new agentic memory architecture inspired by System-One/System-Two cognition. System One captures fast, lightweight decision-making, whereas System Two performs slower, deliberative reasoning. Jev-Mem brings this division of labor to agentic memory through a dedicated System-One control plane, a structured multi-relational memory plane, and a System-Two reasoning plane. The System-One controller governs memory typing and relational organization during construction, and dynamically performs query routing, retrieval-budget allocation, graph traversal, candidate scoring, and adaptive stopping during retrieval. System Two is invoked only for complex reasoning and answer synthesis. This design improves both memory effectiveness and system efficiency: on LoCoMo Jev-Mem achieves an overall LLM-as-a-Judge score of 0.777, an 11.0\% relative improvement over the strongest baseline, while reducing memory construction time to 158\,s, a 6.6$\times$ speedup over the fastest competing memory system, and lowering average query latency to 0.93\,s, a 36.7\% reduction.
Chinese Translation
代理记忆对于长时间跨度的人工智能代理变得至关重要，然而许多现有系统依赖自回归大型语言模型（LLMs）来控制记忆的组织、检索和使用，这使得昂贵的生成过程成为记忆操作的关键环节。我们介绍了 extbf{ extit{Jev-Mem}}，一种新的代理记忆架构，灵感来源于系统一/系统二认知。系统一捕捉快速、轻量级的决策过程，而系统二则进行较慢的深思熟虑推理。Jev-Mem通过一个专用的系统一控制平面、一个结构化的多关系记忆平面和一个系统二推理平面，将这种劳动分工引入代理记忆。系统一控制器在构建过程中管理记忆类型和关系组织，并在检索过程中动态执行查询路由、检索预算分配、图遍历、候选评分和自适应停止。系统二仅在复杂推理和答案综合时被调用。这种设计提高了记忆的有效性和系统的效率：在LoCoMo上，Jev-Mem的整体LLM作为评判者的得分为0.777，相较于最强基线提高了11.0 ext{%}，同时将记忆构建时间减少至158秒，相较于最快的竞争记忆系统加速了6.6倍，并将平均查询延迟降低至0.93秒，减少了36.7 ext{%}。
cs.AI / 64 / 2609.23989
ACLArena: Agent Continue Learning in Multi-stage Post-training
ACLArena：多阶段后训练中的智能体持续学习
Abstract
Building general-purpose agents for industrial deployment requires integrating multiple capabilities, each typically acquired at a distinct stage of training. Yet there is currently no well-established recipe for Agent Continual Learning (ACL), with little understanding of the trade-offs among existing integration paradigms. To address this gap, we introduce ACLArena, a framework for comprehensively studying, analyzing, and evaluating ACL. We first build a sequential training pipeline and conduct an in-depth analysis that explains the mechanisms of forgetting and generalization from two complementary perspectives, the model level and the token level. Guided by these analyses, we systematically compare multi-teacher on-policy distillation, self-distilled fine-tuning, and model merging to assess their ability to recover previously learned capabilities while preserving newly acquired ones. Through extensive experiments, we develop a detailed understanding of how capabilities transfer across stages. Finally, we propose a new ACL recipe that combines offline replay over high-quality trajectories with a routed network of multiple LoRA experts each specialized via RL, substantially improving the agent's ability to learn across multiple domains. Comprehensive experiments on four reasoning and agentic tasks, evaluated under both in-domain and out-of-domain settings, demonstrate the value of our analysis and the effectiveness of our approach.
Chinese Translation
构建用于工业部署的通用智能体需要整合多种能力，而这些能力通常在不同的训练阶段获得。然而，目前尚无成熟的智能体持续学习（Agent Continual Learning, ACL）方案，对现有整合范式之间的权衡理解也很有限。为了解决这一问题，我们提出了ACLArena，一个全面研究、分析和评估ACL的框架。我们首先构建了一个顺序训练管道，并进行了深入分析，从模型层面和标记层面两个互补的视角解释遗忘和泛化的机制。在这些分析的指导下，我们系统地比较了多教师在线蒸馏、自我蒸馏微调和模型合并，以评估它们在恢复先前学习的能力的同时保留新获得能力的能力。通过广泛的实验，我们深入理解了能力在不同阶段之间的转移机制。最后，我们提出了一种新的ACL方案，该方案结合了对高质量轨迹的离线重放和通过强化学习（Reinforcement Learning, RL）专门化的多个LoRA专家的路由网络，显著提高了智能体在多个领域学习的能力。在四个推理和智能体任务上的全面实验，在领域内和领域外的设置下评估，展示了我们分析的价值和我们方法的有效性。
cs.AI / 65 / 2609.24002
FinInteract: Benchmarking Clarification and Intent Integration in Ambiguous Financial Question Answering
FinInteract：模糊金融问答中澄清与意图整合的基准评估
Abstract
Large language model agents increasingly answer financial questions by searching regulatory filings. Such questions are often deceptively under-specified: Meta Platforms' "operating income" is $46.75B consolidated but $62.87B for the Family of Apps segment, and each reading is exactly verifiable against the filing. A capable agent should recognize the ambiguity and ask, rather than commit to a plausible but unintended reading. Existing financial benchmarks cannot measure this, because one gold answer per question cannot separate agents that resolve the ambiguity from those that guess the common reading, a blind spot we call the single-gold illusion. We release FinInteract, a bilingual (English/Chinese) benchmark of 173 instances that pairs each question with a default and an intended interpretation across a five-category ambiguity taxonomy, and grades whether an agent elicits the right clarification and then integrates it. Re-grading identical outputs against the default rather than the intended reading inflates GPT-4o's accuracy by 3.1 times, confirming the illusion. Beyond it, we find that models answer above 90% once the interpretation is supplied but at most 28.9% when they must elicit it themselves, that targeting is uneven across a taxonomy well powered for entity scope and metric definition and exploratory elsewhere, and that conditioning on the ambiguity category improves resolution at both inference and training time.
Chinese Translation
大型语言模型代理越来越多地通过搜索监管文件来回答金融问题。这类问题往往具有误导性的不足明确性：Meta Platforms的“营业收入”在合并报表中为467.5亿美元，但在应用家族（Family of Apps）部门中为628.7亿美元，每个数据都可以准确地与文件进行核对。一名合格的代理应当识别这种模糊性并提出问题，而不是承诺一个合理但非意图的解读。现有的金融基准无法衡量这一点，因为每个问题只有一个黄金答案，这无法区分能够解决模糊性的代理与那些猜测常见解读的代理，这一盲点我们称之为单一黄金幻觉（single-gold illusion）。我们发布了FinInteract，这是一个双语（英语/中文）的基准，包含173个实例，每个问题都配有一个默认解读和一个意图解读，涵盖五类模糊性分类，并评估代理是否能够引出正确的澄清并整合它。对相同输出进行重新评分时，如果以默认解读而非意图解读为基准，GPT-4o的准确率提高了3.1倍，证实了这一幻觉。除此之外，我们发现模型在提供解读后回答正确率超过90%，而在必须自行引出解读时最多只有28.9%；在针对模糊性分类时，模型的表现不均匀，尽管在实体范围和指标定义方面表现良好，而在其他方面则表现探索性；并且在训练和推理时，基于模糊性分类的条件化能够改善解读的准确性。
cs.AI / 66 / 2609.24012
Testing, not presuming, adequacy: calibrating generative social simulators against emergent network structure
测试，而非假设，充分性：针对新兴网络结构的生成社会模拟器校准
Abstract
Validation of generative social simulators often stops at face validity: emergent network structure is compared descriptively, without quantified parameter uncertainty or an adequacy check. We present an adequacy-aware calibration protocol that couples amortized posterior estimation with a synthetic identifiability assessment, a matched-sample-size adequacy check (prior-predictive reachability plus per-statistic posterior-predictive localization), a diagnosis-guided repair, and a statistic-held-out audit. We demonstrate it on a real second-hand luxury resale market with four channel-by-residency cells, each a bipartite buyer-brand network, using a forward model built from persona profiles elicited once, offline, by a language model. The behavioural parameters are recoverable in all four cells, though calibration is approximate and overconfident for one parameter. The observed summary falls outside the simulator's reachability reference in every cell, with the mean purchased tier as the pervasive discrepancy. The repair meets the value-block criterion in two of four cells but does not restore adequacy, and the held-out audit surfaces a buyer-breadth-dispersion miss no earlier diagnostic detected. A profile-source ablation finds the language-model profiles beat a flat rule baseline in all four cells, yet within-category brand relabelling causes no consistent degradation, so the profiles are a partially validated input whose value rests on structure, not brand identity. Making no causal claim, we conclude that an independent-aggregation account, without agent interaction or a buyer-breadth mechanism, cannot jointly reproduce the market's purchased-tier level, head-brand concentration, community structure and buyer-breadth heterogeneity.
Chinese Translation
生成社会模拟器的验证通常停留在表面有效性：新兴网络结构以描述性方式进行比较，而没有量化的参数不确定性或充分性检查。我们提出了一种关注充分性的校准协议，该协议将摊销后验估计与合成可识别性评估、匹配样本大小的充分性检查（先验预测可达性加上每个统计量的后验预测定位）、诊断引导的修复和统计量保留审计相结合。我们在一个真实的二手奢侈品转售市场上展示了这一协议，该市场具有四个渠道-居住单元，每个单元都是一个二分买家-品牌网络，使用从一次性离线获取的角色档案构建的前向模型，后者由语言模型生成。所有四个单元中的行为参数都是可恢复的，尽管对于一个参数的校准是近似的且过于自信。观察到的摘要在每个单元中都超出了模拟器的可达性参考，平均购买层级是普遍的差异。修复在四个单元中的两个满足价值块标准，但未能恢复充分性，而保留审计揭示了一个买家广度分散的遗漏，这是之前的诊断未能发现的。一个档案源消融实验发现，语言模型生成的档案在所有四个单元中优于平坦规则基线，但类别内品牌重新标记并未导致一致的降级，因此这些档案是部分验证的输入，其价值依赖于结构，而非品牌身份。我们没有做出因果声明，得出结论认为，独立聚合的解释，在没有代理人互动或买家广度机制的情况下，无法共同再现市场的购买层级水平、头部品牌集中度、社区结构和买家广度异质性。
cs.AI / 67 / 2609.24016
Context-Aware Pre-Deployment Evaluation of AI Systems: A Regulatory Framework for Nigerian Fintech
基于情境的人工智能系统预部署评估：尼日利亚金融科技的监管框架
Abstract
Commercial large language models are increasingly deployed across African fintech infrastructure for fraud detection and customer communication, yet no Nigerian or African continental regulatory instrument specifies what pre-deployment evaluation such systems must undergo before procurement. This paper reviews African fintech AI governance across global, continental, and Nigerian instruments, and shows that safety is affirmed as a principle while pre-deployment evaluation is operationally unspecified. Generic safety benchmarks cannot surface the failure modes most relevant to this domain, since none contain Nigerian institutional content or test for false positive misclassification of legitimate financial communications. These claims are demonstrated using SafeAlert, a purpose-built evaluation kit applied to six commercial models across three system prompt conditions. Results show that models resisting generic harmful content requests still produce complete fraud scripts under specific framing, and that several models misclassify most legitimate Nigerian bank communications as suspicious or fraudulent, a failure invisible to standard safety evaluation. The paper concludes with a regulatory framework proposing pre-deployment evaluation requirements for the CBN, NITDA, SEC, and the AU, arguing that the identified gap reflects an absence of regulatory specification, not a shortage of technical or financial resources.
Chinese Translation
商业大型语言模型在非洲金融科技基础设施中越来越多地被用于欺诈检测和客户沟通，但目前没有尼日利亚或非洲大陆的监管工具明确规定这些系统在采购前必须经过何种预部署评估。本文回顾了非洲金融科技人工智能治理在全球、大陆和尼日利亚的相关工具，并指出安全性被确立为原则，而预部署评估在操作上并未具体说明。通用的安全基准无法揭示与该领域最相关的失效模式，因为这些基准中没有包含尼日利亚的机构内容或测试合法金融沟通的误报分类。这些论点通过 SafeAlert（一种专门构建的评估工具）进行验证，该工具应用于三种系统提示条件下的六个商业模型。结果显示，抵制通用有害内容请求的模型在特定框架下仍然产生完整的欺诈脚本，并且多个模型将大多数合法的尼日利亚银行沟通误分类为可疑或欺诈，这一失误在标准安全评估中是不可见的。本文最后提出了一个监管框架，建议中央银行（CBN）、国家信息技术发展局（NITDA）、证券交易委员会（SEC）和非洲联盟（AU）制定预部署评估要求，认为所识别的缺口反映的是监管规范的缺失，而非技术或财务资源的短缺。
cs.AI / 68 / 2609.24025
Synthesizing Reactive Character Behaviors for Continuous Games via Programmatic Policy Search
通过程序化策略搜索合成连续游戏的反应性角色行为
Abstract
We present a method for synthesizing reactive character behaviors for continuous games as compact, human-readable programs. Game AI practice still relies heavily on manually authored behavior trees, state machines, and scripts, while academic reinforcement learning typically produces opaque neural controllers that are expensive to train and difficult to edit. Our approach bridges this gap by searching directly over a domain-specific language for continuous-space game policies. The language is designed around reactive geometric decisions and includes higher-order constructs such as direction maximization. These constructs help discretize a continuous behavior space into enumerable program structures. To make program search practical, we introduce a large set of synthesis antipatterns that remove redundant program forms while preserving behavioral coverage. We further combine bottom-up symbolic enumeration with top-down guidance from a coding agent. Our resulting method, agentic sketching, has the agent propose high-level policy structure and call an enumerator to complete local program slots. We evaluate the method on a benchmark of 14 continuous games, ranging from classic control tasks to multi-agent football. We find that pure enumeration is often more efficient than using a coding agent alone, while the combined method substantially outperforms both. Our results suggest that programmatic policy search can be a practical authoring tool for game AI: designers specify reward functions, and the system discovers editable behaviors that are effective, portable, and often surprising.
Chinese Translation
我们提出了一种合成连续游戏反应性角色行为的方法，该方法以紧凑且易于人类阅读的程序形式呈现。游戏人工智能实践仍然在很大程度上依赖于手动编写的行为树、状态机和脚本，而学术界的强化学习通常生成不透明的神经控制器，这些控制器训练成本高且难以编辑。我们的方法通过直接在针对连续空间游戏策略的领域特定语言上进行搜索来弥补这一差距。该语言围绕反应性几何决策设计，并包括方向最大化等高阶构造。这些构造有助于将连续行为空间离散化为可枚举的程序结构。为了使程序搜索变得实用，我们引入了一大批合成反模式，这些反模式在保留行为覆盖的同时去除冗余的程序形式。我们进一步结合自下而上的符号枚举与来自编码代理的自上而下的指导。我们得到的方法，称为代理草图（agentic sketching），让代理提出高层次的策略结构，并调用枚举器来完成局部程序槽。我们在14个连续游戏的基准测试中评估了该方法，这些游戏涵盖了经典控制任务到多智能体足球。我们发现，纯枚举通常比单独使用编码代理更高效，而组合方法的表现则显著优于两者。我们的结果表明，程序化策略搜索可以成为游戏人工智能的实用创作工具：设计师指定奖励函数，系统发现有效、可编辑且常常令人惊讶的行为。
cs.AI / 69 / 2609.24036
Structured Decomposition for Reliable LLM-Generated Access Control Policies
可靠的LLM生成访问控制策略的结构化分解
Abstract
This paper presents an LLM-based system that translates natural-language access control policies (NLACPs) into executable Rego code for Open Policy Agent (OPA). It provides a modular, end-to-end pipeline for policy detection, component extraction, schema validation, linting, compilation, and automated test generation and execution. The system is designed to bridge the gap between human-readable access requirements and machine-enforceable policy-as-code (PaC), with a focus on deployment reliability and security correctness. We evaluate the system on 372 ACRE-complete access control statements with non-null subject, action, and resource annotations against a direct single-prompt LLM baseline to isolate the contribution of structured decomposition and schema-aware validation. The system achieves a 50.3% end-to-end policy correctness rate, compared with 15.3% for the baseline, representing a 3.3x improvement. A policy is counted as correct only if it satisfies compilation, linting, and both positive and negative tests, making this a strict measure of deployable correctness. On security-critical patterns, the system generates correct deny semantics for 87.5% of deny policies (baseline: 37.5%), ownership conditions for 100% of ownership-qualified policies (baseline: 40%), and status-qualified conditions for 100% of status-qualified policies (baseline: 55.6%). These results indicate that structured decomposition and schema-aware validation play a critical role in improving the reliability of LLM-generated authorization policies.
Chinese Translation
本文提出了一种基于LLM的系统，该系统将自然语言访问控制策略（NLACPs）转换为可执行的Rego代码，以供开放策略代理（OPA）使用。该系统提供了一个模块化的端到端管道，涵盖策略检测、组件提取、架构验证、代码检查、编译以及自动化测试生成和执行。该系统旨在弥合人类可读的访问要求与机器可强制执行的代码作为策略（PaC）之间的差距，重点关注部署的可靠性和安全性正确性。我们在372个具有非空主体、动作和资源注释的ACRE完整访问控制语句上评估了该系统，并与直接单提示LLM基线进行对比，以隔离结构化分解和架构感知验证的贡献。该系统实现了50.3%的端到端策略正确率，而基线为15.3%，代表了3.3倍的提升。只有当策略满足编译、代码检查以及正负测试时，才被视为正确，这使得这一度量成为可部署正确性的严格标准。在安全关键模式下，该系统为87.5%的拒绝策略生成了正确的拒绝语义（基线：37.5%），为100%的所有权合格策略生成了所有权条件（基线：40%），以及为100%的状态合格策略生成了状态合格条件（基线：55.6%）。这些结果表明，结构化分解和架构感知验证在提高LLM生成授权策略的可靠性方面发挥了关键作用。
cs.AI / 70 / 2609.24057
Representation-guided in-context learning for medical image interpretation with multimodal large language models
基于表征引导的上下文学习在多模态大型语言模型中的医学图像解读
Abstract
Medical image interpretation is central to diagnosis and care, yet adapting general-purpose multimodal large language models (MLLMs) often requires resource-intensive domain-specific fine-tuning. Here we introduce representation-guided in-context learning (RG-ICL), a training-free inference framework that retrieves query-aligned demonstrations using frozen encoders, without task-specific parameter updates. Across eight datasets spanning histopathology, radiology and retinal fundoscopy, RG-ICL improved classification (mean gain 20 percentage points) and visual question answering (VQA) (mean gain 13 percentage points) over no-context and conventional ICL, approaching or exceeding training-based comparators. Which cases were retrieved mattered more than how many: 6 query-aligned cases outperformed up to 32 randomly selected ones, whereas fixed or random cases often reduced accuracy below baseline. For VQA, aligning reference cases with both image content and question intent produced further gains. These findings indicate that for medical image interpretation, curating which reference cases an MLLM sees is a practical alternative to retraining it.
Chinese Translation
医学图像解读是诊断和护理的核心，但适应通用多模态大型语言模型（MLLMs）通常需要资源密集型的领域特定微调。在此，我们介绍了一种基于表征引导的上下文学习（RG-ICL），这是一种无训练的推理框架，通过使用冻结的编码器检索与查询对齐的示例，而无需特定任务的参数更新。在涵盖组织病理学、放射学和视网膜底照相的八个数据集上，RG-ICL在分类（平均提升20个百分点）和视觉问答（VQA）（平均提升13个百分点）方面优于无上下文和传统的ICL，接近或超过基于训练的比较方法。检索到的案例的选择比数量更为重要：6个与查询对齐的案例的表现优于多达32个随机选择的案例，而固定或随机的案例往往使准确率低于基线。对于VQA，将参考案例与图像内容和问题意图对齐产生了进一步的提升。这些发现表明，对于医学图像解读，策划MLLM所看到的参考案例是一种实用的替代方案，而无需重新训练。
cs.AI / 71 / 2609.24090
Incremental Consistency Execution for Autonomous Intelligent Systems
自主智能系统的增量一致性执行
Abstract
Long-horizon autonomous intelligent systems rely on heterogeneous components such as large language models, databases, external APIs, and rule engines, while their external states continuously change during execution. Re-executing the entire workflow after every change introduces substantial redundant computation. This paper proposes an incremental consistency execution method based on task fact contracts, field-level dependency masks, and state perturbation result invariant domains. After an initial verified execution, the system constructs conservative invariant domains for critical inputs and uses them to determine whether downstream results can be safely renewed without re-invoking expensive components. When re-execution is required, only the smallest affected output fields are recomputed, and an equivalence barrier prevents unnecessary downstream propagation. A submission-time version consistency gate further ensures the safety of side-effecting actions. Experiments on industrial fault diagnosis, enterprise analytics, and LLM-based multi-tool assistants show that the proposed method significantly reduces expensive component calls and end-to-end latency while maintaining high consistency and low incorrect-reuse rates.
Chinese Translation
长时间跨度的自主智能系统依赖于异构组件，如大型语言模型、数据库、外部API和规则引擎，同时其外部状态在执行过程中持续变化。在每次变化后重新执行整个工作流会引入大量冗余计算。本文提出了一种基于任务事实契约、字段级依赖掩码和状态扰动结果不变域的增量一致性执行方法。在初始验证执行后，系统为关键输入构建保守的不变域，并利用这些不变域来判断下游结果是否可以安全更新，而无需重新调用昂贵的组件。当需要重新执行时，仅重新计算受影响的最小输出字段，并且等价屏障防止不必要的下游传播。提交时版本一致性门进一步确保了副作用操作的安全性。在工业故障诊断、企业分析和基于LLM的多工具助手上的实验表明，所提出的方法显著减少了昂贵组件的调用和端到端延迟，同时保持了高一致性和低错误重用率。
cs.AI / 72 / 2609.24092
DocMIDE: Learning Multi-Hop Implicit Derivation in Visually Rich Documents
DocMIDE：在视觉丰富文档中学习多跳隐式推导
Abstract
Real-world document processing systems rely on rigid, predefined schemas, yet critical target fields often lack direct visual counterparts on the page. Extracting these implicit values requires multi-hop derivation, such as aggregating sub-categories or reasoning over visual marks. While existing methods handle explicit text spans or simple implicit queries, they fail at multi-hop visual reasoning even after standard fine-tuning: models retrieve incorrect visual evidence, or retrieve it correctly and then skip the intermediate steps of the derivation. To address this, we introduce DocMIDE, a fine-tuning framework that trains compact vision-language models to retrieve visual evidence explicitly before deriving an answer. DocMIDE constrains generation to a plan-retrieve-derive structure and optimizes it with Group Relative Policy Optimization under a four-component, rule-based reward that scores output format, the retrieved evidence block, every intermediate derivation step, and the final value against a verified reference trace. On a 4,151-pair implicit extraction benchmark, DocMIDE raises accuracy from 70.8% to 95.9% on Qwen3.5-4B from only a small set of annotated examples, and transfers to a second backbone architecture. Supervised demonstrations alone do not close this gap at any budget we tested; rewarding the intermediate steps is what does.
Chinese Translation
现实世界的文档处理系统依赖于严格的预定义模式，但关键目标字段往往在页面上缺乏直接的视觉对应物。提取这些隐式值需要多跳推导，例如聚合子类别或对视觉标记进行推理。虽然现有方法处理显式文本跨度或简单的隐式查询，但在多跳视觉推理方面表现不佳，即使经过标准微调：模型检索到错误的视觉证据，或者正确检索后跳过推导的中间步骤。为了解决这个问题，我们提出了DocMIDE，一个微调框架，训练紧凑的视觉-语言模型在推导答案之前显式地检索视觉证据。DocMIDE将生成约束为计划-检索-推导结构，并通过基于四个组件的规则奖励进行优化，该奖励对输出格式、检索到的证据块、每个中间推导步骤和最终值与经过验证的参考轨迹进行评分。在一个包含4,151对隐式提取基准的测试中，DocMIDE将Qwen3.5-4B的准确率从70.8%提升至95.9%，且仅使用一小部分标注示例，并能迁移到第二个主干架构。仅靠监督示范在我们测试的任何预算下都无法缩小这一差距；奖励中间步骤才是解决之道。
cs.AI / 73 / 2609.24101
When More Evidence Hurts: Publication-Bias Drift and Principled Stopping for Biomedical Causal Search
更多证据反而有害：生物医学因果搜索中的出版偏倚漂移与原则性停止
Abstract
Automated biomedical evidence synthesis depends on retrieving published studies, but the biomedical literature is systematically skewed toward positive findings. Deeper retrieval can therefore make a system \emph{more} likely to falsely infer benefit when the true effect is null. We formalise this phenomenon as \emph{evidence drift} and prove that, under a standard publication-bias model, the false-positive probability on null-effect queries follows a strictly increasing large-sample envelope in retrieval depth, approaching one. Empirically, on a held-out test set of 140 Cochrane-derived queries, drift rises monotonically from 7.9\% to 15.7\% as the retrieval budget grows from 3 to 20 steps, and concentrates in the null-effect class. We present DACG-agent, a drift-aware causal-graph agent that incrementally builds a causal knowledge graph from PubMed abstracts and applies a two-layer stopping policy with complementary roles: a KL-divergence monitor that detects posterior convergence (the accuracy layer), and a Bradley--Terry process reward model (PRM) whose online decline detection halts retrieval once evidence quality peaks (the efficiency layer). Against full-budget retrieval, DACG-agent reduces evidence drift from 15.7\% to 6.4\% and improves null-effect accuracy by 21 percentage points (40.0\%$\to$61.4\%) while using 67\% fewer retrieval steps; overall accuracy rises from 61.4\% to 69.3\% (95\% CI 61--77). A simulation confirms the drift result transfers from the analysed vote-counting aggregator to the deployed noisy-OR one.
Chinese Translation
自动化的生物医学证据综合依赖于检索已发表的研究，但生物医学文献系统性地偏向于积极结果。因此，更深入的检索可能使系统在真实效应为零时更容易错误推断出有益效果。我们将这一现象形式化为“证据漂移”（evidence drift），并证明在标准的出版偏倚模型下，零效应查询的假阳性概率在检索深度上呈严格递增的较大样本包络，接近于1。在一个包含140个Cochrane派生查询的保留测试集上，随着检索预算从3步增长到20步，漂移率单调上升，从7.9%增加到15.7%，并集中在零效应类别中。我们提出了DACG-agent，这是一种关注漂移的因果图代理，能够从PubMed摘要中逐步构建因果知识图谱，并应用具有互补角色的两层停止策略：一个检测后验收敛的KL散度监测器（准确性层），以及一个布拉德利-特里过程奖励模型（PRM），其在线下降检测在证据质量达到峰值后停止检索（效率层）。与全预算检索相比，DACG-agent将证据漂移从15.7%降低到6.4%，并将零效应的准确性提高了21个百分点（从40.0%提升至61.4%），同时减少了67%的检索步骤；整体准确性从61.4%上升至69.3%（95% CI 61--77）。模拟结果确认漂移现象从分析的投票计数聚合器转移到部署的噪声-或（noisy-OR）聚合器上。
cs.AI / 74 / 2609.24115
EDGEGEN: Improving Tool-Calling Agents Beyond Happy Paths with Synthetic Edge Case Generation
EDGEGEN：通过合成边缘案例生成提升工具调用代理的能力
Abstract
Tool-calling LLM agents are increasingly deployed in enterprise applications. However, effective evaluation and optimization require high-quality, diverse task datasets that are often difficult to obtain due to privacy and other constraints. Existing synthetic task generation methods often produce generic tasks that ignore an agent's underlying state or database and fail to reflect real-world usage diversity. We propose EdgeGen, a synthetic task generation framework that extracts compliance rules from an agent's specification and uses them to generate database-grounded edge-case tasks designed to violate these rules. When combined with existing synthetic data generation techniques, EdgeGen enables agent improvement through finetuning and harness optimization. The resulting pipeline forms a fully automated closed-loop system that requires no human annotation. Finetuning on data generated by EdgeGen yields a consistent mean progress improvement of 2 percent to 42 percent on tau2bench airline domain, while other baseline methods show degradation for some models. On the other hand, for harness optimization, our method shows a mean progress improvement of 10 percent and 30 percent over the human-curated and base harnesses, respectively, for the Gemma-4-e4b model.
Chinese Translation
工具调用的大型语言模型（LLM）代理在企业应用中越来越多地被部署。然而，有效的评估和优化需要高质量、多样化的任务数据集，这些数据集由于隐私和其他限制往往难以获得。现有的合成任务生成方法通常生成通用任务，忽视了代理的潜在状态或数据库，未能反映现实世界的使用多样性。我们提出了EdgeGen，一个合成任务生成框架，该框架从代理的规范中提取合规规则，并利用这些规则生成旨在违反这些规则的基于数据库的边缘案例任务。当与现有的合成数据生成技术结合时，EdgeGen通过微调和优化利用促进代理的改进。最终的管道形成一个完全自动化的闭环系统，无需人工标注。在使用EdgeGen生成的数据上进行微调，在tau2bench航空领域上实现了2%到42%的一致性平均进展提升，而其他基线方法在某些模型上则表现出下降。另一方面，在优化利用方面，我们的方法在Gemma-4-e4b模型上相较于人工策划和基础利用分别显示出10%和30%的平均进展提升。
cs.AI / 75 / 2609.24130
Self-Healing Harness for Runtime Oversight of Agent Self-Modification
自愈性工具用于代理自我修改的运行时监督
Abstract
LLM agents can change their own future behavior, raising a basic control question of which self-generated changes should be allowed to persist. We formulate this as admission control for self-modification. The agent may propose changes to its operating instructions, while an external runtime gate controls persistence. We implement this principle as a model-agnostic self-healing harness that runs a Detect, Notice, Heal, Validate loop around an otherwise unmodified agent. The agent authors candidate behavioral rules in an external workspace, where they receive provisional execution authority during evaluation and acquire persistent cross-episode authority only after measured improvement on the triggering failure without regression beyond a fixed margin on protected cases. Replay provides matched evidence when available, forward trials provide a weaker fallback, and a corpus-level guard re-tests the accumulated active rule set. Across 16 matched Baseline and Harness runs spanning AppWorld, Terminal-Bench, and $\tau^2$-Bench, the gate rejected 383 replay-decided proposals. Of these, 211 (55%) improved their triggering failure while degrading a case that previously worked. This shows that locally beneficial self-modifications can introduce collateral regressions often enough to materially affect gate decisions, providing direct empirical motivation for external admission control. Task-completion score is higher under the Harness in all 16 pairs, with two paired bootstrap intervals excluding zero, while repeated-trial reliability is higher in 12 pairs, tied in 4, and lower in none. Because adaptation modifies the policy-inducing context while leaving model weights fixed, admitted changes remain inspectable, reversible, and compatible with closed-weight models.
Chinese Translation
大型语言模型（LLM）代理能够改变自身未来的行为，这引发了一个基本的控制问题，即哪些自我生成的变化应该被允许持续存在。我们将其表述为自我修改的准入控制。代理可以提出对其操作指令的更改，而外部运行时门控则控制这些更改的持久性。我们将这一原则实现为一个模型无关的自愈性工具，它围绕一个未修改的代理运行一个检测、通知、修复、验证（Detect, Notice, Heal, Validate）循环。代理在外部工作区撰写候选行为规则，在评估期间获得临时执行权限，并在经过测量改进后，仅在没有超出固定边际的情况下，获得持久的跨剧集权限。重放提供匹配的证据（如果可用），前向试验提供较弱的后备，而语料库级别的保护则重新测试累积的活跃规则集。在涵盖 AppWorld、Terminal-Bench 和 $ au^2$-Bench 的 16 次匹配基线和自愈工具运行中，门控拒绝了 383 个重放决定的提案。其中，211 个（55%）在改善其触发失败的同时导致了之前有效的案例退化。这表明局部有益的自我修改足够频繁地引入附带回归，从而实质性地影响门控决策，为外部准入控制提供了直接的实证动机。在所有 16 对中，自愈工具下的任务完成得分更高，其中两个配对的自助法区间排除了零，而重复试验的可靠性在 12 对中更高，4 对持平，且没有一对更低。由于适应性修改了政策诱导的上下文，同时保持模型权重不变，已准入的更改仍然可检查、可逆，并与固定权重模型兼容。
cs.AI / 76 / 2609.24165
APEXA: Execution-Integrity Enforcement for Multi-Agent LLM Automation of Synchrotron Data Reduction
APEXA：多智能体大语言模型自动化同步辐射数据减缩的执行完整性保障
Abstract
Synchrotron data reduction, detector calibration followed by azimuthal integration of terabyte-scale diffraction series, is a multi-step, expert-bound bottleneck that increasingly limits the science rate of user facilities. LLM agents promise to collapse it, but driving a real pipeline with a stochastic model creates a failure mode chat benchmarks cannot see: an agent can report a calibration that was never computed. Correctness here is a property of what executed, not of the transcript. We present APEXA, a deployed multi-agent framework (61 tools over heterogeneous compute, run as a single reasoning loop) automating calibration and integration from natural language at a major light source. We make three contributions. First, execution-integrity enforcement: a deterministic tool-layer guard that refuses to surface any result not backed by an executed tool call, with a parser tolerant of cross-model tool-call format drift: in deployment, a frontier model fabricated a complete calibration-comparison report for commands that never ran, which the guard converts to an explicit non-result; the same code gates an optional motor-control surface at 0/200 adversarial violations against a simulated IOC, versus 15/200 for an equivalent safety prompt. Second, we release APEXA-Bench, an evaluation harness of 58 facility tasks (50 base plus an 8-task cross-detector slice) organized by a four-class physical-consequence taxonomy, the first benchmark axis we know of separating a wasted compute cycle from a damaged instrument; its cross-detector grading against NIST-traceable lattice constants surfaced two latent pipeline bugs. Large-scale agent scoring is left to a full-length study. Third, we validate APEXA on real beamline data: from one natural-language prompt it recovers detector geometry and integrates a full attenuation/exposure sweep. We release the framework, harness and traces.
Chinese Translation
同步辐射数据减缩、探测器校准以及对太字节级衍射系列的方位积分是一个多步骤、依赖专家的瓶颈，越来越限制了用户设施的科学产出率。大语言模型（LLM）代理有望简化这一过程，但使用随机模型驱动真实的工作流程会产生一种基准无法识别的失败模式：代理可能报告一个从未计算过的校准。在这里，正确性是执行内容的属性，而不是记录的内容。我们提出了APEXA，一个已部署的多智能体框架（在异构计算环境中运行的61个工具，作为一个单一推理循环），能够从自然语言自动化校准和积分，应用于一个主要的光源。我们做出了三项贡献。首先，执行完整性保障：一个确定性的工具层保护机制，拒绝显示任何没有执行工具调用支持的结果，并且具有对跨模型工具调用格式漂移的宽容解析器：在部署中，一个前沿模型为从未运行的命令生成了完整的校准比较报告，而该保护机制将其转换为明确的非结果；同样的代码在针对模拟IOC的200次对抗性违规中，提供了一个可选的电机控制界面，结果为0/200，而等效的安全提示为15/200。其次，我们发布了APEXA-Bench，一个包含58个设施任务（50个基础任务加上8个跨探测器任务切片）的评估工具，按照四类物理后果分类组织，这是我们所知的第一个将浪费的计算周期与损坏的仪器区分开的基准轴；其针对可追溯至NIST的晶格常数的跨探测器评分揭示了两个潜在的管道错误。大规模代理评分留待完整研究。第三，我们在真实束线数据上验证了APEXA：从一个自然语言提示中，它恢复了探测器几何形状并整合了完整的衰减/曝光扫描。我们发布了该框架、工具和追踪记录。
cs.AI / 77 / 2609.24174
CREDO: Variance-Guided Rubric Evolution for Replay-Corrected Credit Assignment
CREDO：基于方差引导的重放校正信用分配的评分标准演化
Abstract
Long-horizon language agents receive sparse terminal feedback, while intermediate rubrics provide structured but potentially misspecified assessments of progress. In resettable training environments, counterfactual continuation rollouts can measure local credit, but exhaustive replay is costly. We propose Credo, a framework that couples evolving semantic rubrics with selective, execution-based credit correction. A frozen judge maps visible transitions to rubric features, and a credit head predicts the change in expected terminal reward associated with the realized transition. Independently sampled two-sided replays correct prediction residuals using their recorded inclusion probabilities. We derive conditional unbiasedness and a variance decomposition that connects two design choices: which rubric features to retain, and where to allocate a fixed expected replay budget. The resulting criterion weights prediction errors by policy-score sensitivity and missing replay coverage; its allocation rule additionally accounts for continuation cost. We also describe a practical mixture with terminal leave-one-out advantages and distinguish its clipped, token-normalized PPO implementation from the ideal policy-gradient estimator. This preliminary report provides the method, proofs, an exact finite-model audit, and a controlled evaluation protocol. It makes no claim of empirical superiority on language-agent benchmarks.
Chinese Translation
长时间跨度的语言代理接收稀疏的终端反馈，而中间评分标准提供了结构化但可能错误指定的进展评估。在可重置的训练环境中，反事实的持续回放可以测量局部信用，但全面重放成本高昂。我们提出了CREDO，一个将演变的语义评分标准与选择性、基于执行的信用修正相结合的框架。一个冻结的评判者将可见的过渡映射到评分特征上，而一个信用头则预测与实现的过渡相关的预期终端奖励的变化。独立采样的双面重放利用其记录的包含概率来修正预测残差。我们推导出条件无偏性和方差分解，这连接了两个设计选择：保留哪些评分特征，以及在哪里分配固定的预期重放预算。最终的标准通过策略分数敏感性和缺失重放覆盖来加权预测误差；其分配规则还考虑了持续成本。我们还描述了一种具有终端留一法优势的实用混合，并区分其剪切的、令牌归一化的PPO实现与理想的策略梯度估计器。本初步报告提供了方法、证明、精确的有限模型审计和受控评估协议，并未声称在语言代理基准上的经验优越性。
cs.AI / 78 / 2609.24186
LIMIT: Less Is More for Instruction Tuning in Text-to-SQL
LIMIT：文本到SQL指令调优中的少即是多
Abstract
Large language models have achieved remarkable progress on Text-to-SQL through reasoning-enhanced fine-tuning, yet existing approaches predominantly rely on massive instruction corpora under the assumption that scale drives performance. We challenge this paradigm by investigating a fundamental question: what is the minimal data requirement for effective Text-to-SQL instruction tuning? We propose LIMIT(Less Is More for Instruction Tuning in Text-to-SQL), a data-centric framework that demonstrates strong database reasoning can emerge from an extremely compact training set when examples are strategically selected. LIMIT operates through four stages: difficulty-aware filtering that identifies samples within the model's learning frontier, chain-of-thought synthesis with consistency-based selection, multi-dimensional quality scoring via LLM-as-judge, and genetic algorithm optimization that jointly maximizes schema coverage and sample quality. On the BIRD and Spider benchmark, LIMIT selects only 796 and 863 samples while achieving 100% table coverage, enabling Qwen3-8B to reach 69.1% and 88.9% execution accuracy.This result surpasses methods trained on 20 times more data and establishes a new state-of-the-art among open-source approaches. Our findings suggest that careful data curation, rather than scale, is the key to efficient Text-to-SQL learning.
Chinese Translation
大型语言模型通过增强推理的微调在文本到SQL（Text-to-SQL）任务上取得了显著进展，但现有方法主要依赖于大量的指令语料库，假设规模驱动性能。我们挑战这一范式，探讨一个基本问题：有效的文本到SQL指令调优的最小数据需求是什么？我们提出了LIMIT（Less Is More for Instruction Tuning in Text-to-SQL），一个以数据为中心的框架，证明当示例经过战略性选择时，强大的数据库推理可以从极其紧凑的训练集中涌现。LIMIT通过四个阶段进行操作：难度感知过滤，识别模型学习边界内的样本；思维链合成，基于一致性选择；通过LLM-as-judge进行多维质量评分；以及遗传算法优化，联合最大化模式覆盖和样本质量。在BIRD和Spider基准上，LIMIT仅选择了796和863个样本，同时实现了100%的表覆盖，使Qwen3-8B达到了69.1%和88.9%的执行准确率。该结果超越了在20倍更多数据上训练的方法，并在开源方法中建立了新的最先进水平。我们的研究结果表明，仔细的数据策划，而非规模，是高效文本到SQL学习的关键。
cs.AI / 79 / 2609.24198
SKstars at SHROOM: Visions Agreement-Guided Ensembling of Zero-Shot and LoRA-Adapted Vision--Language Models
SKstars在SHROOM中的表现：基于视觉一致性指导的零样本和LoRA适应的视觉-语言模型集成
Abstract
This paper describes the SKstars submission to SHROOM-Visions 2026, a shared task on fine-grained hallucination detection in large vision-language model outputs. The task requires systems to identify hallucinated character spans, assign hallucination categories, and provide confidence estimates for their predictions. Our approach combines zero-shot predictions from Qwen2.5-VL-72B-Instruct with those of a LoRA-adapted Qwen2.5-VL-7B-Instruct model. The outputs of the two models are integrated through a lightweight ensemble procedure, followed by span refinement and confidence adjustment. We evaluate the main system components on a small internal development subset and report the performance of the submitted system on the official English test set. SKstars achieved a Cor+Lbl score of 0.2902, ranking 15th among 29 teams, and obtained Cor and IoU scores of 0.3642 and 0.3151, respectively, ranking 18th on both metrics. The results show that combining a large zero-shot model with a smaller adapted model provides a practical framework for multilingual and fine-grained hallucination localization, while also highlighting the difficulty of transferring development-set improvements to hidden test data. Code and predictions: https://github.com/aliathar1401/SK-Stars-shroom-visions-2026
Chinese Translation
本文描述了SKstars对SHROOM-Visions 2026的提交，这是一个关于大型视觉-语言模型输出中细粒度幻觉检测的共享任务。该任务要求系统识别幻觉字符范围，分配幻觉类别，并提供预测的置信度估计。我们的方法结合了来自Qwen2.5-VL-72B-Instruct的零样本预测和LoRA适应的Qwen2.5-VL-7B-Instruct模型的预测。两个模型的输出通过轻量级集成程序进行整合，随后进行范围细化和置信度调整。我们在一个小型内部开发子集上评估了主要系统组件，并报告了提交系统在官方英语测试集上的表现。SKstars获得了0.2902的Cor+Lbl分数，在29支队伍中排名第15，并且在Cor和IoU指标上分别获得了0.3642和0.3151的分数，均排名第18。结果表明，将大型零样本模型与较小的适应模型结合提供了一个实用的多语言和细粒度幻觉定位框架，同时也突显了将开发集改进转移到隐藏测试数据的困难。代码和预测可见于：https://github.com/aliathar1401/SK-Stars-shroom-visions-2026
cs.AI / 80 / 2609.24229
Recovering Lost Details: Multi-Scale Frequency Compensation for Long-Term Time Series Forecasting
恢复丢失细节：用于长期时间序列预测的多尺度频率补偿
Abstract
Long-term time series forecasting has made significant progress by leveraging multi-scale information to capture hierarchical temporal patterns and model long-range dependencies. However, temporal downsampling in existing multi-scale methods inevitably smooths detailed temporal fluctuations, and this information loss is further aggravated by their emphasis on dominant trends across scales, resulting in insufficiently expressive representations. To address this, we propose a Multi-Scale Wavelet Mixing (MWMixer) model, which incorporates a Bidirectional Frequency-Bands Mixing strategy to recover lost temporal details across scales, enabling complementary cross-scale information interactions. Then, a Dynamic Scale-Adaptive Fusion module learns time-varying weights for each scale to fuse multi-scale forecasts into the final prediction, enhancing the flexibility of multi-scale aggregation. In addition, a cross-scale consistency loss aligns each coarse-scale prediction with the interval-averaged fine-scale outputs, while a multi-scale supervision loss enforces prediction accuracy at each scale, promoting consistent learning across scales. Extensive experiments on seven real-world datasets demonstrate that MWMixer achieves competitive performance in long-term forecasting.
Chinese Translation
长期时间序列预测通过利用多尺度信息捕捉层次时间模式和建模长程依赖性取得了显著进展。然而，现有多尺度方法中的时间下采样不可避免地平滑了详细的时间波动，这种信息损失因其对各尺度主导趋势的强调而进一步加剧，导致表示能力不足。为此，我们提出了一种多尺度小波混合（MWMixer）模型，该模型结合了双向频带混合策略，以恢复各尺度间丢失的时间细节，从而实现互补的跨尺度信息交互。接着，动态尺度自适应融合模块为每个尺度学习时间变化的权重，将多尺度预测融合为最终预测，增强了多尺度聚合的灵活性。此外，跨尺度一致性损失将每个粗尺度预测与区间平均的细尺度输出对齐，而多尺度监督损失则在每个尺度上强制执行预测准确性，促进跨尺度的一致学习。在七个真实世界数据集上的大量实验表明，MWMixer在长期预测中实现了具有竞争力的性能。
cs.AI / 81 / 2609.24243
Taming CoT Obfuscation in VLMs: From Mechanistic Evidence to Activation Enforcement
驯服视觉语言模型中的链式思维模糊化：从机制证据到激活强制
Abstract
Reinforcement learning (RL) improves reasoning in vision-language models (VLMs) but can induce chain-of-thought (CoT) obfuscation: an operational, non-intentional outcome where task reward or accuracy rises while traces become less grounded and monitorable. Prior work largely documents this decay behaviorally, leaving its representation-level correlates and actionable controls unclear. We find that template- and ground-associated activations become less separable during RL; matched interventions support the contribution of selected features to monitorability degradation. Guided by this evidence, we propose Targeted Anti-obfuscation with Mechanistic Enforcement (TAME), which uses Sparse Autoencoders (SAEs) to combine behavioral feedback with targeted suppression of template-associated activations during RL. Its asymmetric constraint penalizes template activations only above their pre-RL baseline, anchoring the localized features while behavioral feedback promotes grounded refinements. Across VIRL-39k, SPA-VL, and two model families, TAME improves CoT monitorability by up to 30.9 and 16.7 percentage points over Group Relative Policy Optimization (GRPO), respectively. Blinded human evaluation finds higher human monitorability on both datasets, and two held-out monitor families reproduce the monitorability gains. Task accuracy changes are small and mixed, and general-capability benchmarks show task-specific trade-offs. These results provide a path from behavioral monitoring to representation-level oversight for more auditable RL-trained multimodal systems.
Chinese Translation
强化学习（RL）改善了视觉语言模型（VLMs）中的推理能力，但可能导致链式思维（CoT）模糊化：一种操作性、非故意的结果，其中任务奖励或准确性上升，而痕迹变得不那么扎根和可监控。之前的研究主要通过行为方式记录了这种衰退现象，但其在表征层面的相关性和可操作控制尚不明确。我们发现，在强化学习过程中，模板和基础相关的激活变得不那么可分；匹配的干预支持了所选特征对可监控性退化的贡献。基于这一证据，我们提出了机制强制的针对性反模糊化方法（Targeted Anti-obfuscation with Mechanistic Enforcement, TAME），该方法使用稀疏自编码器（Sparse Autoencoders, SAEs）将行为反馈与在强化学习过程中对模板相关激活的针对性抑制相结合。其不对称约束仅对高于其强化学习前基线的模板激活施加惩罚，从而锚定局部特征，同时行为反馈促进扎根的细化。在VIRL-39k、SPA-VL和两个模型家族中，TAME分别将链式思维的可监控性提高了最多30.9和16.7个百分点，相较于组相对策略优化（Group Relative Policy Optimization, GRPO）。盲评人类评估发现两个数据集上的人类可监控性更高，两个保留的监控家族重现了可监控性提升。任务准确性变化较小且混合，而通用能力基准显示出任务特定的权衡。这些结果为从行为监控到表征层面的监督提供了一条路径，以实现更可审计的强化学习训练的多模态系统。
cs.AI / 82 / 2609.24265
Unsupervised Brain Anomaly Detection as a Bayesian Inverse Problem with Diffusion Prior
将无监督脑部异常检测视为具有扩散先验的贝叶斯逆问题
Abstract
Unsupervised anomaly detection (UAD) aims to localize abnormal regions in medical scans without pixel-level annotations. A typical strategy seeks to reconstruct a pseudo-healthy image that preserves subject-specific anatomy. Recently, diffusion models have been proposed to perform UAD. However, these methods rely on heuristic noise schedules or synthetic corruptions to balance subject-specificity and anomaly removal. In this work, we propose an alternative formulation of UAD as a Bayesian inverse problem under a diffusion prior. First, we introduce a latent spatial anomaly mask that models pixel-wise consistency between a test image and its latent corresponding pseudo-healthy image. Then, we propose an approximation of the unknown generation process that links healthy anatomy, anomalies, and the observed image, enabling a well-defined likelihood within the Bayesian framework. Building on recent advances in diffusion-based inverse problem methods, we jointly infer the pseudo-healthy image and the anomaly mask via annealed posterior sampling. We evaluate our approach on FDG PET (ADNI) and FLAIR MRI (BraTS 2021), demonstrating improved anomaly localization performance compared to other diffusion-based approaches and validating the contribution of our introduced model. Our code is available at https://github.com/HuguesRoy/UAD_DAPS.
Chinese Translation
无监督异常检测（UAD）旨在在医学扫描中定位异常区域，而无需像素级注释。典型策略试图重建一个保留个体特征解剖结构的伪健康图像。最近，扩散模型被提出用于执行UAD。然而，这些方法依赖于启发式噪声调度或合成损坏，以平衡个体特异性和异常去除。在本研究中，我们提出了一种将UAD作为在扩散先验下的贝叶斯逆问题的替代公式。首先，我们引入了一种潜在空间异常掩码，该掩码建模测试图像与其潜在对应伪健康图像之间的像素级一致性。然后，我们提出了一种未知生成过程的近似，该过程将健康解剖结构、异常和观察到的图像联系起来，从而在贝叶斯框架内实现良定义的似然性。基于最近在基于扩散的逆问题方法中的进展，我们通过退火后验采样共同推断伪健康图像和异常掩码。我们在FDG PET（ADNI）和FLAIR MRI（BraTS 2021）上评估了我们的方法，展示了与其他基于扩散的方法相比，异常定位性能的改善，并验证了我们所提出模型的贡献。我们的代码可在 https://github.com/HuguesRoy/UAD_DAPS 获取。
cs.AI / 83 / 2609.24277
How Many Pixels Is a Digit Worth? Place-Aware Coordinate Entropy for GUI Agent Confidence Estimation
一个数字值多少像素？用于GUI代理置信度估计的地点感知坐标熵
Abstract
GUI agents predict click coordinates as digit-token sequences, but standard text-LLM confidence estimation methods rank correct clicks from wrong ones only weakly. GUI-specific alternatives use K samples or new supervision, but still leave room for improvement. We trace part of this to place-value asymmetry: bounding-box correctness often makes higher-place digits more important than lower-place digits, so uniform aggregation weakens the signal that determines correctness. The fix is to weight each digit's Shannon entropy by its place value. We call this Place-Aware Coordinate Entropy (PACE). Across fixed-scale agents on ScreenSpot-Pro and ScreenSpot-v2, PACE wins both AUROC and selective accuracy on all primary comparisons in a single forward pass, matching or outperforming K-sample baselines at a fraction of the cost. PACE provides a per-click confidence estimate that turns coordinate-token internals into a practical confidence signal for GUI agent deployment.
Chinese Translation
GUI代理将点击坐标预测为数字-标记序列，但标准文本-LLM置信度估计方法仅弱地对正确点击与错误点击进行排名。针对GUI的替代方法使用K个样本或新的监督，但仍有改进的空间。我们将部分原因归结为位置值不对称：边界框的正确性通常使得高位数字比低位数字更重要，因此均匀聚合削弱了确定正确性的信号。解决方案是根据每个数字的位置值加权其香农熵。我们称之为地点感知坐标熵（Place-Aware Coordinate Entropy, PACE）。在ScreenSpot-Pro和ScreenSpot-v2上的固定规模代理中，PACE在所有主要比较中以单次前向传递赢得了AUROC和选择性准确性，且以较低的成本匹配或超越了K样本基线。PACE提供了每次点击的置信度估计，将坐标-标记内部转化为GUI代理部署的实用置信度信号。
cs.AI / 84 / 2609.24290
When and How Should an Agent Clarify? CIGAsk: Teaching LLMs to Clarify via Counterfactual Information Gain
代理何时以及如何进行澄清？CIGAsk：通过反事实信息增益教导大型语言模型澄清
Abstract
Instruction-tuned LLMs faced with underspecified queries often commit to a single interpretation rather than ask for clarification, producing confidently wrong answers. In our experiments, prompting alone is insufficient: models either ask for clarification on every query or ask vague questions that fail to recover the missing information. Addressing this failure requires learning two coupled skills: when to ask rather than answer and how to ask a question that recovers the disambiguating information. Existing recipes either address only one of these skills or require a separately trained critic. We propose CIGAsk, an RL recipe that teaches both skills through two complementary reward signals within a multi-turn GRPO loop. Counterfactual Information Gain (CIG) compares the gold-answer log-likelihood under a frozen reference model with and without the user response, providing per-turn credit that guides how to ask. The Asymmetric Ambiguity Bonus assigns a signed reward at the terminal token based on the gold ambiguity label, guiding when to ask. Across three clarification benchmarks spanning table, passage, and open-domain QA, CIGAsk-7B outperforms the strongest external baseline despite using a smaller backbone. It also transfers across datasets without per-dataset tuning while preserving single-turn QA performance on out-of-distribution benchmarks.
Chinese Translation
面对不明确查询的指令调优大型语言模型（LLMs）往往会承诺单一解释，而不是请求澄清，从而产生自信但错误的答案。在我们的实验中，仅仅通过提示是不够的：模型要么在每个查询上都请求澄清，要么提出模糊的问题，无法恢复缺失的信息。解决这一失败需要学习两项相互关联的技能：何时请求而不是回答，以及如何提出一个能够恢复消歧信息的问题。现有的方法要么只解决其中一项技能，要么需要单独训练的评估者。我们提出了CIGAsk，这是一种强化学习（RL）方法，通过多轮GRPO循环中的两种互补奖励信号来教授这两项技能。反事实信息增益（Counterfactual Information Gain, CIG）比较在冻结的参考模型下，用户响应前后的黄金答案对数似然，为每轮提供指导如何提问的信用。非对称模糊奖励（Asymmetric Ambiguity Bonus）根据黄金模糊标签在终端标记处分配有符号的奖励，指导何时请求澄清。在涵盖表格、段落和开放域问答的三个澄清基准测试中，CIGAsk-7B在使用较小主干的情况下超越了最强的外部基线。它还能够在不同数据集之间迁移，而无需针对每个数据集进行调优，同时在分布外基准上保持单轮问答性能。
cs.AI / 85 / 2609.24324
Brain-Token Learning: Microstate-Based Tokenization and Multi-Scale Interaction for Long-Horizon EEG Sequence Modeling
脑令牌学习：基于微状态的令牌化和多尺度交互用于长时间范围EEG序列建模
Abstract
Electroencephalography (EEG) provides a non-invasive window into dynamic brain activity, yet modeling long-horizon EEG sequences remains challenging due to their high temporal complexity, substantial variability across subjects, and the lack of biologically meaningful sequence representations. Existing tokenization strategies, such as fixed-window and patch-based representations, discretize EEG signals according to artificial temporal boundaries, which may disrupt intrinsic brain-state dynamics. In this work, we propose Brain-Token Learning, a neuroscience-inspired framework that introduces Brain Tokenization for long-horizon EEG sequence modeling. Instead of partitioning EEG signals into predefined temporal segments, Brain Tokenization represents EEG as sequences of recurrent microstate-derived brain tokens, where each token corresponds to a quasi-stable large-scale brain state with variable temporal duration. Based on these biologically grounded tokens, we further develop a multi-scale token interaction module consisting of Latent State Aggregation and State Transition Modeling to jointly capture global brain-state context and local microstate transitions. We evaluate Brain-Token on five heterogeneous EEG datasets, including the newly collected long-horizon NeuroLong dataset and four affective or clinical EEG datasets (SEED, DEAP, MDD, and NSSI). Extensive experiments demonstrate that Brain-Token consistently outperforms conventional CNN/LSTM architectures, Transformer-based models, and domain adaptation methods across diverse EEG scenarios. Further analysis verifies the effectiveness of microstate-based tokenization and multi-scale interaction for learning robust and interpretable EEG representations. These results establish Brain-Token as a biologically grounded tokenization paradigm for long-horizon EEG sequence modeling.
Chinese Translation
脑电图（EEG）提供了一个非侵入性窗口，观察动态脑活动，但由于其高时间复杂性、个体间显著的变异性以及缺乏生物学上有意义的序列表示，建模长时间范围的EEG序列仍然具有挑战性。现有的令牌化策略，如固定窗口和基于补丁的表示，按照人为的时间边界对EEG信号进行离散化，这可能会干扰内在的脑状态动态。在本研究中，我们提出了脑令牌学习（Brain-Token Learning），这是一个受神经科学启发的框架，引入了脑令牌化（Brain Tokenization）用于长时间范围EEG序列建模。脑令牌化并不是将EEG信号划分为预定义的时间段，而是将EEG表示为由重复微状态派生的脑令牌序列，其中每个令牌对应于一个具有可变时间持续的准稳定大规模脑状态。基于这些生物学基础的令牌，我们进一步开发了一个多尺度令牌交互模块，包括潜在状态聚合（Latent State Aggregation）和状态转移建模（State Transition Modeling），以共同捕捉全局脑状态上下文和局部微状态转移。我们在五个异质EEG数据集上评估了脑令牌，包括新收集的长时间范围神经长（NeuroLong）数据集和四个情感或临床EEG数据集（SEED、DEAP、MDD和NSSI）。大量实验表明，脑令牌在各种EEG场景中始终优于传统的CNN/LSTM架构、基于Transformer的模型和领域适应方法。进一步分析验证了基于微状态的令牌化和多尺度交互在学习稳健且可解释的EEG表示方面的有效性。这些结果确立了脑令牌作为一种生物学基础的令牌化范式，用于长时间范围EEG序列建模。
cs.AI / 86 / 2609.24346
LADDER: Graph-Guided Diffusion Language Models for Efficient Multi-Hop Reasoning
LADDER：用于高效多跳推理的图引导扩散语言模型
Abstract
Graph Retrieval-Augmented Generation (GraphRAG) has remarkably enhanced large language models on complex reasoning by leveraging structured entity topologies. However, existing frameworks heavily rely on standard autoregressive language models where the nature of inherent sequential generation severely hinders overall inference efficiency. Inspired by Diffusion Language Models (DLMs) that offer massive parallelism via continuous refine-in-parallel decoding, we aim to accelerate GraphRAG in the discrete space. However, it remains non-trivial for two challenges. First, partially denoised drafts are highly dynamic and uncertain, making dynamic graph grounding non-trivial. Second, raw denoising states are inherently noisy and unstable, making synchronous graph retrieval and multi-hop aggregation computationally prohibitive. To this end, we present LADDER, a novel framework that bridges diffusion language modeling with GraphRAG through graph-guided parallel decoding. Specifically, (i) we propose an event-driven self-clocking retrieval, inspired by our key insight that 88% of target entities emerge early in the partially denoised state, leading final commitment by an average of 5.7-9.6 steps. This mechanism dynamically triggers graph retrieval only when the set of graph-linkable entities expands, yielding an asynchronous self-clocking policy that bypasses learned gates or heuristic thresholds. (ii) An incomplete-query graph propagation module is designed to process the newly emerging entity queries using a specialized graph foundation model, continuously aggregating multi-hop evidence to sharpen parallel predictions and accelerate overall decoding convergence. Extensive experiments on three challenging multi-hop QA benchmarks show that LADDER raises average exact match from 39.6% to 45.2% while achieving a 4.1x latency reduction.
Chinese Translation
图检索增强生成（GraphRAG）通过利用结构化实体拓扑显著提升了大型语言模型在复杂推理任务中的表现。然而，现有框架严重依赖于标准自回归语言模型，其固有的顺序生成特性严重阻碍了整体推理效率。受到扩散语言模型（DLMs）启发，后者通过连续并行解码提供了巨大的并行性，我们旨在加速GraphRAG在离散空间中的应用。然而，这面临两个非平凡的挑战。首先，部分去噪草稿高度动态且不确定，使得动态图基础不易实现。其次，原始去噪状态本质上是嘈杂和不稳定的，这使得同步图检索和多跳聚合在计算上变得不可行。为此，我们提出了LADDER，一个通过图引导的并行解码将扩散语言建模与GraphRAG相结合的新框架。具体而言，（i）我们提出了一种事件驱动的自时钟检索，灵感来自我们的关键洞察，即88%的目标实体在部分去噪状态下早期出现，最终承诺平均提前5.7-9.6步。该机制仅在图可链接实体集扩展时动态触发图检索，从而产生一种异步自时钟策略，绕过学习的门控或启发式阈值。（ii）设计了一种不完整查询图传播模块，利用专门的图基础模型处理新出现的实体查询，持续聚合多跳证据以提高并行预测的准确性并加速整体解码收敛。在三个具有挑战性的多跳问答基准上的广泛实验表明，LADDER将平均精确匹配率从39.6%提高到45.2%，同时实现了4.1倍的延迟减少。
cs.AI / 87 / 2609.24352
Few-Shot Demonstrations Elicit the Use of In-Context World Representations in LLMs
少量示例引发大型语言模型中上下文世界表征的使用
Abstract
Large language models (LLMs), when acting as agents, are expected to take observed data in context, infer the latent state space underlying the world, and leverage it for downstream prediction. However, prior work demonstrated that LLMs struggle to use representations learned in context on a graph tracking task, where the model needs to construct a representation of the graph governing data generation process and use it for subsequent predictions. In this paper, we show that extending this to few-shot settings, where each demonstration is generated from a different world with either the same or different graph topologies, enhances its prediction on 6 models from 4 model families. To understand this improvement, we linearly probe a low-dimensional world representation that encodes graph information in the hidden states. Notably, we find that few-shot demonstrations relocate the world representation and increase its predictive use. Specifically, for each model, these world representations shift in directions nearly orthogonal to their original subspace, and interventions on these representations selectively impair performance more than interventions on other subspaces. Consistent with this insight, we show that few-shot demonstrations with observations from different worlds improve performance on ARC-AGI-1&2, web agent tasks, and Othello. Our findings elucidate the role and internal mechanisms of few-shot demonstrations in in-context world modeling. More broadly, our work advances our understanding of how LLM agents learn from in-context observations and provides implications for their further improvement.
Chinese Translation
大型语言模型（LLMs）在作为代理时，预期能够在上下文中利用观察到的数据，推断出潜在的世界状态空间，并将其用于下游预测。然而，先前的研究表明，LLMs在图跟踪任务中难以使用在上下文中学习到的表征，该任务要求模型构建一个描述数据生成过程的图的表征，并利用它进行后续预测。本文展示了将此扩展到少量示例设置的效果，其中每个示例来自具有相同或不同图拓扑的不同世界，这增强了4个模型家族中6个模型的预测能力。为了理解这种改进，我们对编码图信息的低维世界表征进行了线性探测。值得注意的是，我们发现少量示例重新定位了世界表征，并增加了其预测使用。具体而言，对于每个模型，这些世界表征在与其原始子空间几乎正交的方向上发生了偏移，并且对这些表征的干预比对其他子空间的干预更能显著影响性能。与这一见解一致，我们展示了来自不同世界的少量示例在ARC-AGI-1&2、网络代理任务和五子棋中的性能提升。我们的研究阐明了少量示例在上下文世界建模中的作用及其内部机制。更广泛地说，我们的工作推动了对LLM代理如何从上下文观察中学习的理解，并为其进一步改进提供了启示。
cs.AI / 88 / 2609.24362
VLM-in-Sandbox: Visual Workspaces for Agentic Visual Reasoning
VLM-in-Sandbox：用于自主视觉推理的视觉工作空间
Abstract
Sandboxed computer environments support multi-step reasoning with tools, executable programs, and persistent files, yet their extension from language models to vision-language models (VLMs) introduces a distinct state-management problem. Visual reasoning produces intermediate image-valued evidence---crops, masks, overlays, zoomed regions, and analytic renderings---that must remain addressable without accumulating unboundedly in multimodal context. We introduce VLM-in-Sandbox, a training-free framework for agentic multimodal reasoning in controlled computer environments. Its Visual Workspace registers generated artifacts in an image ledger, maintains a bounded active visual context, and lets the model explicitly promote selected evidence for subsequent inspection. This separates visual evidence generation, performed by sandbox tools, from visual evidence management. Across seven benchmarks and four base VLMs, VLM-in-Sandbox achieves the highest sample-weighted average accuracy among Vanilla VLM, Append-only Sandbox, and the proposed method. A compiler-matched $2\times2$ study on 1,260 examples further separates model-directed visibility from bounded retention: VLM-in-Sandbox reaches 66.27% accuracy with 18.6% fewer total tokens than the automatic, retain-all control. Over all 6,350 submitted GPT-4.1-mini examples, it produces 302 rescues and 142 regressions relative to Original Append-only. A local vLLM study with prefix caching confirms that the smaller request workload also reduces uncached tokens, time to first token, and end-to-end latency. These results identify explicit visual evidence state as a central abstraction for sandboxed VLM agents.
Chinese Translation
沙盒计算环境支持使用工具、可执行程序和持久文件进行多步骤推理，但从语言模型扩展到视觉语言模型（VLM）引入了一个独特的状态管理问题。视觉推理产生中间的图像值证据——裁剪、掩膜、叠加、放大区域和分析渲染——这些证据必须在多模态上下文中可寻址，而不会无限累积。我们提出了VLM-in-Sandbox，这是一个在受控计算环境中进行自主多模态推理的无训练框架。其视觉工作空间在图像分类账中注册生成的工件，维护一个有限的活动视觉上下文，并允许模型明确地促进所选证据以供后续检查。这将由沙盒工具执行的视觉证据生成与视觉证据管理分开。在七个基准测试和四个基础VLM上，VLM-in-Sandbox在Vanilla VLM、仅附加沙盒和所提方法中实现了最高的样本加权平均准确率。在1260个示例上进行的编译器匹配$2 imes2$研究进一步区分了模型导向的可见性与有限保留：VLM-in-Sandbox以66.27%的准确率实现了比自动保留所有控制少18.6%的总标记数。在提交的6350个GPT-4.1-mini示例中，它相对于原始仅附加产生了302个救援和142个回归。带有前缀缓存的本地vLLM研究证实，较小的请求工作负载也减少了未缓存标记、首次标记的时间和端到端延迟。这些结果确定了显式视觉证据状态作为沙盒VLM代理的核心抽象。
cs.AI / 89 / 2609.24453
Predicting Postprandial Glycemic Response from Meal Images, Clinical Variables, and Gut Microbiome Information
基于餐食图像、临床变量和肠道微生物组信息预测餐后血糖反应
Abstract
Predicting postprandial glycemic response (PPGR) is fundamental to personalized nutrition and type 2 diabetes management, yet existing approaches typically rely on manually reported dietary intake, limiting their scalability in free-living settings. We propose a multimodal framework that replaces manual dietary logging with image-derived macronutrient estimates and integrates them with clinical variables and gut microbiome information for personalized PPGR prediction. The framework jointly performs image-based macronutrient estimation and glucose prediction, while an attention-based prediction module models interactions between dietary and host-specific information. We evaluate the proposed approach on a real-world dataset comprising meal images, continuous glucose monitoring, clinical variables, and gut microbiome profiles. The proposed model outperforms existing PPGR baselines using image-derived nutritional inputs and approaches the performance of methods that rely on manually reported macronutrients despite using automatically estimated nutritional information. These results demonstrate that combining image-derived nutrition with complementary clinical and gut microbiome information provides a practical foundation for scalable personalized PPGR prediction.
Chinese Translation
预测餐后血糖反应（PPGR）对个性化营养和2型糖尿病管理至关重要，但现有方法通常依赖于手动报告的饮食摄入，限制了其在自由生活环境中的可扩展性。我们提出了一种多模态框架，用图像推导的宏量营养素估计替代手动饮食记录，并将其与临床变量和肠道微生物组信息结合，以实现个性化的PPGR预测。该框架同时执行基于图像的宏量营养素估计和葡萄糖预测，而基于注意力的预测模块则建模饮食与宿主特定信息之间的交互。我们在一个包含餐食图像、连续血糖监测、临床变量和肠道微生物组特征的真实世界数据集上评估了所提方法。该模型在使用图像推导的营养输入时优于现有的PPGR基准，并且尽管使用了自动估计的营养信息，但其性能接近依赖于手动报告的宏量营养素的方法。这些结果表明，将图像推导的营养与互补的临床和肠道微生物组信息结合，为可扩展的个性化PPGR预测提供了实用基础。
cs.AI / 90 / 2609.24480
Fathom-Vaidya: Advancing Medical Reasoning with Rubric-Based Rewards
Fathom-Vaidya：基于评分标准奖励的医学推理进展
Abstract
Deploying Large Language Models (LLMs) in healthcare requires robust performance across two complementary dimensions - diagnostic reasoning: the convergent, evidence-driven task of inferring a patient's condition from clinical data to produce a diagnosis, and clinical healthcare reasoning: the broader, navigational judgment required to communicate, plan, and adapt across multi-turn clinical interactions where a single correct answer may not exist. Recent benchmarks such as HealthBench and MedXpertQA reveal persistent weaknesses in both areas, exposing failures in complex diagnostic scenarios and limitations in contextual, patient-centered dialogue. We introduce a sequential training framework that targets these facets using synthetic data and rubric-based reinforcement learning. First, we improve diagnostic reasoning using MedBullets-derived questions with rule- and rubric-guided Reinforcement Learning (RL). We then shift to clinical reasoning by generating 5.3k synthetic multi-turn scenarios, each paired with multi-dimensional rubrics to comprehensively assess the response. This approach yields over 10% improvement on MedXpertQA, and our 30B model achieves 50.1% accuracy on HealthBench-Hard, surpassing proprietary baselines including GPT-5 (thinking). Our results show that targeted synthetic datasets and rubric-based training can systematically improve both diagnostic and interactive clinical reasoning in medical LLMs.
Chinese Translation
在医疗保健中部署大型语言模型（LLMs）需要在两个互补维度上具备强大的性能——诊断推理：从临床数据推断患者状况以生成诊断的聚合、基于证据的任务，以及临床医疗推理：在多轮临床互动中进行沟通、规划和适应所需的更广泛的导航判断，其中可能不存在单一正确答案。近期的基准测试如 HealthBench 和 MedXpertQA 显示这两个领域存在持续的弱点，暴露出在复杂诊断场景中的失败以及在以患者为中心的对话中的局限性。我们提出了一种顺序训练框架，利用合成数据和基于评分标准的强化学习来针对这些方面进行改进。首先，我们通过使用 MedBullets 衍生的问题，结合规则和评分标准指导的强化学习（RL）来提升诊断推理。接着，我们通过生成 5.3k 个合成多轮场景，并为每个场景配备多维评分标准，以全面评估响应，转向临床推理。这种方法在 MedXpertQA 上实现了超过 10% 的提升，我们的 30B 模型在 HealthBench-Hard 上达到了 50.1% 的准确率，超越了包括 GPT-5（思考）在内的专有基准。我们的结果表明，针对性的合成数据集和基于评分标准的训练可以系统性地改善医学 LLM 中的诊断和互动临床推理。
cs.AI / 91 / 2609.24517
Not All Task Vectors Need Equal Rank: Energy-Proportional Allocation for Model Merging
并非所有任务向量都需要相同的秩：用于模型合并的能量比例分配
Abstract
Model merging aims to combine multiple fine-tuned models derived from a common pretrained model into a single multi-task model without additional joint training. Recent spectral merging methods improve over simple weight averaging by exploiting low-rank structures of task-specific updates, but they commonly assign the same rank capacity to every task. This uniform allocation ignores that task vectors can have heterogeneous spectral complexity, causing the shared merging space to be used suboptimally. In this paper, we propose Spectral Energy-proportional Rank Allocation (SERA), a simple task-adaptive strategy that allocates ranks according to the singular-value energy structure of each task vector. By assigning richer spectral capacity to complex or isolated tasks and fewer directions to compact tasks, SERA extends SVD-based model merging from uniform-capacity merging to task-dependent capacity allocation. Experiments under standard vision model merging protocols show that SERA improves multi-task merging performance while preserving the same total rank budget as existing spectral merging methods. Further analysis demonstrates that task-level spectral concentration is closely related to the per-task effect of adaptive rank allocation, providing insight into when and why SERA is effective.
Chinese Translation
模型合并旨在将多个从共同预训练模型派生的微调模型合并为一个单一的多任务模型，而无需额外的联合训练。最近的谱合并方法通过利用任务特定更新的低秩结构，改善了简单权重平均的方法，但它们通常对每个任务分配相同的秩容量。这种统一的分配忽视了任务向量可能具有异质谱复杂性，导致共享合并空间的使用不够优化。在本文中，我们提出了谱能量比例秩分配（Spectral Energy-proportional Rank Allocation, SERA），这是一种简单的任务自适应策略，根据每个任务向量的奇异值能量结构分配秩。通过为复杂或孤立的任务分配更丰富的谱容量，而为紧凑任务分配较少的方向，SERA将基于奇异值分解（SVD）的模型合并从统一容量合并扩展到任务依赖的容量分配。标准视觉模型合并协议下的实验表明，SERA在保持与现有谱合并方法相同的总秩预算的同时，提高了多任务合并性能。进一步的分析表明，任务级谱集中度与自适应秩分配的每任务效果密切相关，为SERA何时以及为何有效提供了见解。
cs.AI / 92 / 2609.24555
The Endless Exam: Mathematical Constructions from Today's Models toward Superintelligence
无尽的考试：从当今模型到超智能的数学构造
Abstract
We introduce the Endless Exam, a benchmark for measuring mathematical progress from today's models toward artificial superintelligence through fourteen parameterised construction families. Each submitted object is checked automatically for validity and assigned a relative quality score against a published frontier or construction baseline, without capping improvements at $1$. The families draw on open mathematical problems for long-term targets and generate new instances at larger parameters, where compact certificates keep large constructions verifiable. Across eight models evaluated on 69 distinct instances, continuous quality scores distinguish performance even though no evaluated system surpasses a published frontier. Size-quality curves show how construction quality changes as problem size increases. We release the generators, verifiers, references, model responses and analysis to support continued measurement before and beyond human frontiers.
Chinese Translation
我们介绍了无尽的考试，这是一个基准，用于通过十四个参数化构造家族测量从当今模型到人工超智能的数学进展。每个提交的对象都自动检查其有效性，并根据已发布的前沿或构造基线分配相对质量分数，而不限制改进至 $1$。这些家族借鉴了开放的数学问题作为长期目标，并在更大参数下生成新的实例，其中紧凑的证书保持大型构造的可验证性。在对69个不同实例评估的八个模型中，连续的质量分数区分了性能，尽管没有评估的系统超越已发布的前沿。规模-质量曲线显示了随着问题规模增加，构造质量的变化。我们发布了生成器、验证器、参考文献、模型响应和分析，以支持在超越人类前沿之前和之后的持续测量。
cs.AI / 93 / 2609.24620
Ascent: An Agentic System over the Model Context Protocol for Real-World Clinical Data Analysis
Ascent：基于模型上下文协议的代理系统用于真实世界临床数据分析
Abstract
Answering epidemiological questions from real-world clinical data requires medical coding, schema-aware SQL, and validation of implicit choices about populations, denominators, and time. We present Ascent, an agentic system that exposes medical coding, question answering, and cohort analysis through a shared Model Context Protocol tool surface for standardized and native schemas. We introduce EpiTrap, a dataset testing whether systems avoid recognized pharmacoepidemiological errors, and compare a fixed pipeline with agents across models and orchestrators. With capable models, agents improve accuracy over the fixed pipeline by an average of 27 and 20 percentage points on native and standardized schemas, respectively. These gains require more tool calls and longer runtimes. Experience from real projects highlights the system's value for feasibility assessment, diagnostic iteration, and expert-guided analysis.
Chinese Translation
从真实世界临床数据中回答流行病学问题需要医学编码、模式感知的SQL以及对人群、分母和时间的隐含选择进行验证。我们提出了Ascent，一个代理系统，通过共享的模型上下文协议工具界面，暴露医学编码、问题回答和队列分析，以支持标准化和本地化模式。我们引入了EpiTrap，一个数据集，用于测试系统是否避免已知的药物流行病学错误，并比较固定管道与跨模型和协调者的代理。使用能力强的模型，代理在本地和标准化模式上分别提高了27和20个百分点的准确性，超越了固定管道。这些提升需要更多的工具调用和更长的运行时间。来自真实项目的经验突显了该系统在可行性评估、诊断迭代和专家指导分析中的价值。
cs.AI / 94 / 2609.24625
Custom Named Entity Recognition and Topic Classification for Global Health Publications
全球健康出版物的定制命名实体识别与主题分类
Abstract
How should natural language processing models be selected and adapted for global health literature in environments where annotated data and computational resources are limited? This thesis investigates these challenges through experiments on semantic tag discovery, named entity recognition (NER), and multi-label topic classification. First, skip-gram word2vec models trained on progressively larger specialized corpora are compared with BioWordVec to assess how corpus size and domain context influence tag discovery. Vocabulary coverage and qualitative evaluation indicate that broader coverage does not necessarily yield more useful domain-specific associations. The analysis then turns to entity extraction, comparing convolutional spaCy models with a RoBERTa-based transformer on 1,000 annotated sentences. Under a lenient scoring protocol, the transformer achieves 0.80 micro-F1 versus 0.65-0.69 for convolutional models, but takes 82 seconds rather than 5-6 seconds. This trade-off motivates fine-tuning convolutional models and integrating a disease recognizer that achieves 81.33% test F1 on the NCBI Disease Corpus. Combined with PDF preprocessing, entity filtering, and MeSH enrichment, the resulting pipeline supports document-level indexing. To complement entity extraction with thematic annotation, MiniLM-based few-shot classification is compared with BART-MNLI zero-shot inference across 50 topics and 1,000 handcrafted test sentences. BART-MNLI achieves 95.2% single-label accuracy versus 59%; reported multi-label accuracies are 88% and 32% under partly manual assessment. However, its higher inference cost limits practical integration. The results show where domain specialization and lightweight adaptation offer practical value, and where transformer accuracy justifies higher inference costs, providing an empirical basis for building knowledge systems under resource constraints.
Chinese Translation
在注释数据和计算资源有限的环境中，如何选择和调整自然语言处理模型以适应全球健康文献？本论文通过对语义标签发现、命名实体识别（NER）和多标签主题分类的实验，探讨了这些挑战。首先，比较了在逐渐增大的专业语料库上训练的 skip-gram word2vec 模型与 BioWordVec，以评估语料库大小和领域上下文如何影响标签发现。词汇覆盖率和定性评估表明，较广的覆盖并不一定会产生更有用的领域特定关联。接着分析了实体提取，比较了卷积 spaCy 模型与基于 RoBERTa 的变换器在 1,000 个注释句子上的表现。在宽松的评分协议下，变换器的微 F1 分数达到 0.80，而卷积模型为 0.65-0.69，但变换器的处理时间为 82 秒，而卷积模型仅需 5-6 秒。这一权衡促使对卷积模型进行微调，并整合一个在 NCBI 疾病语料库上取得 81.33% 测试 F1 的疾病识别器。结合 PDF 预处理、实体过滤和 MeSH 丰富化，最终的管道支持文档级索引。为了通过主题注释补充实体提取，比较了基于 MiniLM 的少样本分类与 BART-MNLI 的零样本推理，涵盖 50 个主题和 1,000 个手工制作的测试句子。BART-MNLI 实现了 95.2% 的单标签准确率，而少样本分类为 59%；在部分手动评估下，报告的多标签准确率分别为 88% 和 32%。然而，其较高的推理成本限制了实际整合。结果显示了领域专业化和轻量级适应在何处提供了实际价值，以及变换器的准确性在何处证明了更高的推理成本是合理的，为在资源限制下构建知识系统提供了实证基础。
cs.AI / 95 / 2609.24662
DUMA-Bench: A Dual-Control Multi-Agent Benchmark for Evaluating LLM Agent Security
DUMA-Bench：一种双控多智能体基准测试，用于评估大型语言模型代理的安全性
Abstract
LLM-based agents increasingly operate in environments where they interact with users, tools, and external systems. Yet most security evaluations assume passive users and static control, ignoring the interactive dynamics that shape real agent behavior. We introduce \textbf{DUMA-Bench}, a benchmark and evaluation protocol for measuring agent security under \emph{dual-control} interaction, where both the agent and the user can influence the shared environment state. DUMA-Bench extends $\tau^2$-bench ~\cite{barres2025tau} with adversarial environments covering eight vulnerability classes, including RAG poisoning, cross-agent manipulation, and unsafe output handling. We evaluate \textbf{14 models from five model families} (OpenAI, Anthropic, DeepSeek, Qwen, and Z.ai) across eight domains and multiple user-behavior regimes. Across our experiments, introducing dual-control interaction increases the attack success rate from \textbf{26.9\%} to \textbf{41.1\%}. These results show that agent security is not solely a property of the model but emerges from the interaction between the model, the user, and the environment. DUMA-Bench provides a missing evaluation layer for studying security in realistic agent deployments.
Chinese Translation
基于大型语言模型（LLM）的代理越来越多地在与用户、工具和外部系统交互的环境中运行。然而，大多数安全评估假设用户是被动的，控制是静态的，忽视了塑造真实代理行为的互动动态。我们引入了 extbf{DUMA-Bench}，这是一个基准测试和评估协议，用于测量代理在 extit{双控} 交互下的安全性，在这种交互中，代理和用户都可以影响共享环境状态。DUMA-Bench 扩展了 $ au^2$-bench ~ extcite{barres2025tau}，涵盖了八类脆弱性，包括 RAG 中毒、跨代理操控和不安全输出处理等对抗性环境。我们评估了来自五个模型家族（OpenAI、Anthropic、DeepSeek、Qwen 和 Z.ai）的 extbf{14 个模型}，涵盖了八个领域和多种用户行为模式。在我们的实验中，引入双控交互将攻击成功率从 extbf{26.9\%} 提高到 extbf{41.1\\%}。这些结果表明，代理的安全性不仅仅是模型的属性，而是模型、用户和环境之间互动的产物。DUMA-Bench 为研究现实代理部署中的安全性提供了一个缺失的评估层。
cs.AI / 96 / 2609.24663
Beyond Endpoint Performance: Process-Level Evaluation of Self-Evolving Agents
超越端点性能：自我进化智能体的过程级评估
Abstract
Self-evolving agents convert interaction feedback into persistent artifacts, such as memories or skills, which in turn guide subsequent decisions. As these artifacts are iteratively updated throughout an experience stream, the capabilities they support may evolve. Consequently, endpoint performance alone offers an incomplete view of self-evolution. Process-level evaluation is therefore essential to identify when a target capability emerges and whether later updates strengthen, preserve, or weaken it. Motivated by this, we propose \textsc{EvoPathBench}, a benchmark that tracks individual capabilities during artifact-level self-evolution. EvoPathBench fixes the base model, tools, freezes evolving artifacts at successive checkpoints, and evaluates the target capability on held-out episodes. This benchmark evaluates agent self-evolution using public trading data and calibrated trajectories. It tests three capabilities: generalization to unseen tasks, retention after unrelated learning, and rule adaptation to new evidence. Experimental results show that gains on similar unseen tasks often weaken under distribution shift, retention losses are concentrated in a minority of evolution paths, and no method achieves reliable rule adaptation. Moreover, while self-evolution enables agents to generate candidate artifacts with substantial held-out gains, the selected updates consistently fall short of realizing this potential. Together, these findings establish capability-level process evaluation as a foundation for analyzing self-evolution, identifying candidate evaluation and selection as key targets for improvement.
Chinese Translation
自我进化智能体将交互反馈转化为持久性人工制品，如记忆或技能，这些人工制品反过来又指导后续决策。随着这些人工制品在经验流中被迭代更新，它们所支持的能力可能会演变。因此，仅依靠端点性能无法全面了解自我进化。因此，过程级评估对于识别目标能力何时出现以及后续更新是加强、保持还是削弱该能力至关重要。基于此，我们提出了 extsc{EvoPathBench}，一个在人工制品级自我进化过程中跟踪个体能力的基准。EvoPathBench 固定基础模型和工具，在连续检查点冻结进化中的人工制品，并在保留的情节上评估目标能力。该基准使用公共交易数据和校准轨迹评估智能体的自我进化。它测试三种能力：对未见任务的泛化、在无关学习后的保持，以及对新证据的规则适应。实验结果表明，在类似的未见任务上获得的收益在分布转移下往往会减弱，保持损失集中在少数进化路径中，并且没有方法能够实现可靠的规则适应。此外，尽管自我进化使智能体能够生成具有显著保留收益的候选人工制品，但所选择的更新始终未能实现这一潜力。综上所述，这些发现确立了能力级过程评估作为分析自我进化的基础，并将候选评估和选择识别为改进的关键目标。
cs.AI / 97 / 2609.24677
TimeLitmus: A Diagnostic Benchmark for Cross-Modal Understanding and Explanation Faithfulness in Event-Conditioned Time-Series Prediction
TimeLitmus：事件条件时间序列预测中跨模态理解与解释可信度的诊断基准
Abstract
Large language models (LLMs) are increasingly used to make predictions from numerical time-series histories and textual events. Yet accuracy alone cannot reveal whether correct answers reflect effective integration of the two inputs or instead arise from event polarity, unimodal priors, or superficial cues. Likewise, plausible explanations may rationalize predictions without faithfully reflecting the evidence that drives model behavior. We introduce TimeLitmus, a diagnostic benchmark for cross-modal understanding and explanation faithfulness in event-conditioned time-series prediction. TimeLitmus contains 4,856 evaluation records across Finance and Traffic, combining natural prediction with controlled counterfactual and contrastive interventions, explanation-targeted faithfulness tests, and systematic shortcut controls. Across ten representative LLMs, standard prediction accuracy substantially overstates reliable cross-modal understanding: Hard Paired Contrast (HPC) pair correctness peaks at only 19.2% in Finance and 11.7% in Traffic, and all ten models show lower-than-expected consistency on Finance series-side controls. Models often recognize scenario relations explicitly yet fail to apply them during independent prediction. Explanation faithfulness shows a similar gap: in Traffic, most models cite the manipulated temporal factor in over 90% of cases, while behavioral support remains below 22%. Human annotators outperform LLMs on matched controlled and hard-pair diagnostics, confirming that these distinctions are recoverable from the inputs. Natural-only adaptation yields selective gains in evidence selection and input sensitivity, but not consistent gains in controlled or hard-pair behavior. The benchmark, evaluation suite, and supervised adaptation data will be released publicly.
Chinese Translation
大型语言模型（LLMs）越来越多地用于从数值时间序列历史和文本事件中进行预测。然而，仅凭准确性无法揭示正确答案是否反映了两种输入的有效整合，或者仅仅是由于事件极性、单模态先验或表面线索所致。同样，合理的解释可能会合理化预测，但并未忠实反映驱动模型行为的证据。我们提出了TimeLitmus，这是一个用于事件条件时间序列预测中跨模态理解和解释可信度的诊断基准。TimeLitmus包含4,856条评估记录，涵盖金融和交通领域，结合了自然预测与受控的反事实和对比干预、针对解释的可信度测试以及系统的捷径控制。在十个代表性的LLM中，标准预测准确性大大夸大了可靠的跨模态理解：在金融领域，困难配对对比（Hard Paired Contrast，HPC）的配对正确率仅达到19.2%，在交通领域为11.7%，所有十个模型在金融系列的控制上表现出低于预期的一致性。模型通常能够明确识别场景关系，但在独立预测时未能应用这些关系。解释的可信度显示出类似的差距：在交通领域，大多数模型在90%以上的案例中引用了操控的时间因素，而行为支持则低于22%。人类注释者在匹配的控制和困难配对诊断中表现优于LLMs，确认这些区别可以从输入中恢复。仅自然适应在证据选择和输入敏感性上带来了选择性收益，但在控制或困难配对行为上并未带来一致的收益。该基准、评估套件和监督适应数据将公开发布。
cs.AI / 98 / 2609.24744
World State Generator
世界状态生成器
Abstract
Language agents solve complex tasks through plans and actions. A single step the world refuses puts the goal out of reach, and what the agent does next decides the task. Prompted planners fail at exactly this point, rewriting the refused step in new words, meeting the same refusal, and burning the attempt budget without moving. They fail because the plan was never tied to the world, so a refusal has nothing in the plan to attach to. A world is where a task runs, and it has its own rules, its own admissible actions, and its own constraints. We build synthetic worlds across 7 domains and extract training data from them. A program enforces each world's rules and grades its goal, and every world is admitted only if its goal is reachable from its initial state. Agents run inside and leave verified failures paired with repairs that carried the run to a state the world certified, a record of about 226K trajectories. On this record we train the World State Generator, a model that writes a plan as checkable states of the world and keeps that plan aligned with the world it runs in. That alignment is what a plan written in language lacks, since the world it runs in has physical limits, logical dependencies, and required orders the language never states, and the plan encounters these rules only when a state fails. WSG takes that failure as the rule the world has stated and rewrites the remaining states to obey it, so the plan bends to the world as the run goes on. Across 7 public benchmarks, WSG raises end-to-end success for two open models near 30B parameters over prompting and brings to the level of proprietary model.
Chinese Translation
语言代理通过计划和行动解决复杂任务。世界拒绝的单一步骤使目标变得遥不可及，而代理接下来所做的决定了任务的进展。被提示的规划者正是在这一点上失败，试图用新的措辞重写被拒绝的步骤，遭遇同样的拒绝，并在未取得进展的情况下消耗尝试预算。他们失败的原因在于计划从未与世界相联系，因此拒绝在计划中没有任何可以附加的内容。世界是任务运行的地方，它有自己的规则、可接受的行动和约束条件。我们在7个领域构建合成世界，并从中提取训练数据。一个程序执行每个世界的规则并评估其目标，只有当目标可以从初始状态达到时，世界才被接受。代理在其中运行，并留下经过验证的失败记录，伴随修复措施将运行带到一个世界认证的状态，记录约226K条轨迹。在这一记录上，我们训练了世界状态生成器（World State Generator，WSG），该模型将计划写成可检查的世界状态，并保持该计划与其运行的世界保持一致。这种一致性是用语言编写的计划所缺乏的，因为其运行的世界具有物理限制、逻辑依赖关系和所需顺序，而语言从未陈述这些，计划仅在状态失败时才遇到这些规则。WSG将这一失败视为世界所陈述的规则，并重写剩余状态以遵循它，因此计划在运行过程中向世界弯曲。在7个公共基准测试中，WSG将两个接近300亿参数的开放模型的端到端成功率提升至接近30%，并达到了专有模型的水平。
cs.AI / 99 / 2609.24755
Epi-Logic: A Conceptual Framework for Epistemic Runtime Control, Schema Validity Checking, and Controlled Accommodation in Autonomous AI Agents
Epi-Logic：一种用于自主人工智能代理的认知运行时控制、模式有效性检查和受控适应的概念框架
Abstract
Autonomous AI agents are increasingly deployed in areas where wrong decisions are hard to reverse. This paper examines schema mismatch: the condition in which an agent operates within an interpretive frame that no longer applies to the current context. Outputs produced under such a mismatch can appear internally consistent, linguistically plausible, and largely factually correct; output-quality metrics alone therefore capture the underlying loss of validity only partially. The paper introduces Epi-Logic, a conceptual framework for epistemic runtime control. It couples the detection of schema dissonance, a graduated reduction of autonomy, and the auditable switch to a validated schema. A schema is formalised as a tuple of variable space, expectation model, validity conditions, axioms, and metadata. The Epi-Score aggregates seven graded dimensions of epistemic dissonance; the temporal validity dimension D8, violations of the validity conditions G, and axiom violations are carried as separate categorical paths that are not offset against the aggregate. The architecture rests on a checking asymmetry: formalised validity conditions can be checked at runtime, whereas the correctness of many actions is established only ex post. The paper separates two architectural properties, a conditional result from sequential changepoint detection, and an empirical remainder. Eight falsifiable propositions with named baselines describe the transition to empirical validation. All propositions are empirically testable hypotheses, not established results.
Chinese Translation
自主人工智能代理越来越多地被部署在决策错误难以逆转的领域。本文探讨了模式不匹配：即代理在一个不再适用于当前情境的解释框架内操作的状态。在这种不匹配下产生的输出可能在内部上是一致的、语言上是合理的，并且在很大程度上是事实正确的；因此，仅仅依靠输出质量指标只能部分反映潜在的有效性丧失。本文介绍了Epi-Logic，一个用于认知运行时控制的概念框架。它结合了模式不和谐的检测、逐步减少自主性和可审计的切换到经过验证的模式。模式被形式化为一个包含变量空间、期望模型、有效性条件、公理和元数据的元组。Epi-Score聚合了七个分级的认知不和谐维度；时间有效性维度D8、有效性条件的违反G和公理的违反被作为单独的类别路径处理，而不与总和抵消。该架构基于检查不对称性：形式化的有效性条件可以在运行时检查，而许多行动的正确性仅在事后确定。本文区分了两个架构属性：来自序列变更点检测的条件结果和经验剩余。八个可证伪的命题及其命名基线描述了向经验验证的过渡。所有命题都是可经验检验的假设，而非已建立的结果。
cs.AI / 100 / 2609.24760
Construting Reverse Thinking: Developing Large Language Models' Reverse Thingking Ability
构建逆向思维：发展大型语言模型的逆向思维能力
Abstract
When facing complex problems, humans tend to try various ideas for different issues. Human thinking patterns exhibit remarkable flexibility in adapting to diverse scenarios. GPT-o1, GPT-o3, and DeepSeek-R1 adopt long chain-of-thought models to address complex problems by increasing reasoning depth, which default to a forward reasoning mode. We conducted statistical analysis on the accuracy of different mathematical problem datasets on models of different scales, and found five reasons for errors: Insufficient solution-space coverage, Computational mistakes, Unverified assumptions, Ignoring constraint conditions, Maximum response length limitation. To address the above issues, we proposed a backward reasoning pattern construction method aimed at enhancing the model's reverse thinking ability and dynamic adaptability. First, we constructed an easy-hard two-stage Math dataset for training large models and gradually improving their inference ability at different difficulty levels. The dataset contains forward reasoning paths as well as backward reasoning paths. And a two-stage supervised fine-tuning process is applied to progressively train the model's backward reasoning capability. Furthermore, a fine-grained reward mechanism is developed, employing smoothed reward signals to strengthen the model's ability to autonomously select thinking modes during the reasoning process, thereby avoiding reward hacking. A linear-decay balanced sampling strategy is designed to maintain a balance between forward and backward reasoning path samples during training, enabling the model to converge quickly and stably. Experimental results show that our method significantly improves reasoning efficiency and accuracy in tasks such as mathematical proofs, offering a flexible and efficient reasoning paradigm for solving complex problems.
Chinese Translation
面对复杂问题时，人类往往会尝试针对不同问题提出多种想法。人类的思维模式在适应多样化场景方面展现出显著的灵活性。GPT-o1、GPT-o3 和 DeepSeek-R1 采用长链思维模型，通过增加推理深度来解决复杂问题，默认采用前向推理模式。我们对不同规模模型在不同数学问题数据集上的准确性进行了统计分析，发现错误的五个原因：解空间覆盖不足、计算错误、未经验证的假设、忽视约束条件、最大响应长度限制。为了解决上述问题，我们提出了一种逆向推理模式构建方法，旨在增强模型的逆向思维能力和动态适应性。首先，我们构建了一个简单-困难两阶段的数学数据集，用于训练大型模型，并逐步提高其在不同难度水平下的推理能力。该数据集包含前向推理路径和逆向推理路径。同时，应用两阶段监督微调过程，逐步训练模型的逆向推理能力。此外，我们开发了一种细粒度奖励机制，采用平滑的奖励信号来增强模型在推理过程中自主选择思维模式的能力，从而避免奖励黑客行为。设计了一种线性衰减平衡采样策略，以在训练过程中保持前向和逆向推理路径样本之间的平衡，使模型能够快速而稳定地收敛。实验结果表明，我们的方法显著提高了数学证明等任务的推理效率和准确性，为解决复杂问题提供了一种灵活高效的推理范式。
cs.AI / 101 / 2609.24784
Convex AI Compositionality and the Governance of AI System Populations
凸性人工智能组合性与人工智能系统群体的治理
Abstract
AI governance increasingly requires providers and public authorities to reason about multiple AI instantiations, alternative versions, and deployment configurations of multiple AI systems. Yet current regulation remains predominantly single-system-centric, acknowledging such multiplicity only sparsely without treating collections of related AI systems as governance objects. This creates an AI population governance problem: determining which instantiations can be meaningfully considered together and how their changing configurations can be represented and monitored. The first requirement has recently been addressed through trustworthiness-based accounts of AI identity. We address the second by introducing convex AI compositionality: a formal representation of the configurations generated by finite AI system populations that uses convex spaces. The core idea is that convex compositions of the operational states that a population of AI system instantiations may occupy over time are compatible with lifecycle reachability across the population and can preserve the formal identity relations between these systems. Well-known statistical and geometric constructions, such as weighted state distributions and convex hulls, become AI governance tools for distinguishing operational states, population weights, heterogeneity, and AI configuration change across different governance modes while remaining compatible, under stated conditions, with lifecycle reachability and AI identity. We illustrate our AI population governance framework through distributed healthcare deployments and controlled deployment of recruitment AI variants.
Chinese Translation
人工智能治理日益需要提供者和公共当局对多个人工智能实例、替代版本以及多个人工智能系统的部署配置进行推理。然而，目前的监管仍然主要以单一系统为中心，稀疏地承认这种多样性，而没有将相关人工智能系统的集合视为治理对象。这造成了一个人工智能群体治理问题：确定哪些实例可以有意义地一起考虑，以及如何表示和监测它们不断变化的配置。最近，通过基于可信度的人工智能身份理论，已经解决了第一个要求。我们通过引入凸性人工智能组合性来解决第二个问题：这是一个利用凸空间对有限人工智能系统群体生成的配置进行形式化表示的框架。核心思想是，人工智能系统实例群体在时间上可能占据的操作状态的凸组合与群体的生命周期可达性相兼容，并能够保持这些系统之间的形式身份关系。众所周知的统计和几何构造，如加权状态分布和凸包，成为区分操作状态、群体权重、异质性以及不同治理模式下人工智能配置变化的治理工具，同时在规定条件下与生命周期可达性和人工智能身份保持兼容。我们通过分布式医疗部署和控制招聘人工智能变体的部署来说明我们的人工智能群体治理框架。
cs.AI / 102 / 2609.24831
GRUET: Quantifying Uncertainty of Agentic Reasoning-and-Acting Processes
GRUET：量化代理推理与行动过程的不确定性
Abstract
Agents have attracted considerably increasing attention due to the power of executing both Reasoning and Acting (ReAct) in open and dynamic environments. The ReAct process typically exhibits a multi-turn trajectory in which one drives Large Language Models (LLMs) to generate both reasoning chains and task-specific actions in an interleaved manner. However, agents often suffer from significant uncertainty, where identical tasks yield divergent trajectories; trajectories with higher uncertainty often produce incomprehensible behaviors, severely undermining agent credibility. This work conjectures that such trajectory-level uncertainty frequently stems from cumulative turn-level reasoning uncertainty induced by LLMs; the latter often exhibits a collection of branches of divergent reasoning chains and their resulting actions. Built upon this, we present the Graph-based Reasoning UncErtainty in Trajectories (GRUET) method for the uncertainty quantification of ReAct, comprising turn-level reasoning uncertainty quantification and trajectory-level uncertainty aggregation; the former precisely quantifies reasoning uncertainty via modeling the reasoning space spanned by potential reasoning branches as a graph and then approximating the reasoning space complexity with graph complexity, while the latter employs simple aggregation strategies for quantifying the overall trajectory credibility. Empirical evaluations across nine LLMs and five benchmarks validate the effectiveness of our proposed GRUET in terms of selective generation performance, measured by AUROC, AUPRC, and AUARC.
Chinese Translation
由于在开放和动态环境中执行推理与行动（ReAct）的能力，代理引起了越来越多的关注。ReAct 过程通常表现为多轮轨迹，其中一个驱动大型语言模型（LLMs）以交错的方式生成推理链和任务特定的行动。然而，代理往往面临显著的不确定性，相同的任务可能产生不同的轨迹；具有更高不确定性的轨迹往往会产生难以理解的行为，严重削弱代理的可信度。本研究推测，这种轨迹级别的不确定性通常源于 LLMs 引起的累积轮次级别推理不确定性；后者往往表现为一系列分支的不同推理链及其产生的行动。在此基础上，我们提出了基于图的轨迹推理不确定性量化方法（GRUET），用于 ReAct 的不确定性量化，包括轮次级别推理不确定性量化和轨迹级别不确定性聚合；前者通过将潜在推理分支所跨越的推理空间建模为图来精确量化推理不确定性，然后用图复杂度来近似推理空间复杂度，而后者则采用简单的聚合策略来量化整体轨迹的可信度。对九个 LLMs 和五个基准的实证评估验证了我们提出的 GRUET 在选择性生成性能方面的有效性，衡量指标包括 AUROC、AUPRC 和 AUARC。
cs.AI / 103 / 2609.24838
MedRSI: Recursive Self-Improvement for Medical Agents via Clinically Aligned Self-Evolution
MedRSI：通过临床对齐自我演化实现医疗代理的递归自我改进
Abstract
Medical agents increasingly combine general reasoning models with specialized clinical tools, yet their capabilities remain largely fixed by what clinicians and engineers design before deployment. Recursive self-improvement (RSI) offers a different paradigm in which agents learn from their own failures and autonomously expand their capabilities, but directly applying RSI to medicine introduces fundamental safety challenges. We introduce MedRSI, the first recursive self-improvement framework for medicine, which continuously transforms diagnostic failures into new clinical capabilities through tool composition and task-specific model training. Inspired by clinical practice, MedRSI introduces two mechanisms for clinically aligned self-evolution. Clinical-cost-aware failure prioritization directs improvement toward errors according to their potential clinical consequences rather than frequency alone. Fast discovery with slow registration separates rapid capability invention from conservative adoption, allowing new tools to enter the persistent agent only after demonstrating sustained benefit across subsequent patient cohorts. Across public glaucoma and heart disease benchmarks and two private clinical tasks, MedRSI progressively develops segmentation, measurement, prediction, multimodal reasoning, and generative capabilities, surpasses manually engineered medical agents, and autonomously discovers solutions to clinical problems not anticipated by its original designers. Our results show that medical agents need not remain constrained by capabilities specified before deployment: with clinically grounded mechanisms governing what to improve and what to retain, they can continuously construct, validate, and accumulate new capabilities from diagnostic experience. Code is available at https://github.com/ImprintLab/MedRSI.
Chinese Translation
医疗代理越来越多地将通用推理模型与专业临床工具相结合，但它们的能力在很大程度上仍然受到临床医生和工程师在部署前设计的限制。递归自我改进（Recursive Self-Improvement, RSI）提供了一种不同的范式，使代理能够从自身的失败中学习并自主扩展其能力，但将RSI直接应用于医学领域会引入基本的安全挑战。我们提出了MedRSI，这是第一个用于医学的递归自我改进框架，它通过工具组合和任务特定模型训练，将诊断失败持续转化为新的临床能力。受临床实践的启发，MedRSI引入了两种临床对齐自我演化的机制。临床成本意识的失败优先级根据错误的潜在临床后果而非仅仅是频率来指导改进。快速发现与缓慢注册的分离使得快速能力发明与保守采纳相分离，允许新工具在展示对后续患者队列的持续益处后才进入持久代理。通过公共青光眼和心脏病基准测试以及两个私有临床任务，MedRSI逐步发展了分割、测量、预测、多模态推理和生成能力，超越了手动设计的医疗代理，并自主发现了原始设计者未预见的临床问题的解决方案。我们的结果表明，医疗代理不必受限于部署前指定的能力：通过临床基础机制来管理改进和保留的内容，它们可以持续构建、验证和积累来自诊断经验的新能力。代码可在 https://github.com/ImprintLab/MedRSI 获取。
cs.AI / 104 / 2609.24855
Extracting Arguments, Not Just Classifying Them: Instruction-Tuned LLMs for Generative Component Detection
提取论证，而不仅仅是对其进行分类：用于生成性组件检测的指令调优大型语言模型
Abstract
Argumentative component detection (ACD) is a core subtask of Argument(ation) Mining (AM) and one of its most challenging aspects, as it requires jointly delimiting argumentative spans and classifying them into components such as claims and premises. While research on this subtask remains relatively limited compared to other AM tasks, most existing approaches formulate it as a simplified sequence labeling problem, component classification, or a pipeline of component segmentation followed by classification. In this paper, we propose ITFACD, a novel approach based on instruction-tuned Large Language Models (LLMs) using compact instruction-based prompts, and reframe ACD as a language generation task, enabling arguments to be identified directly from plain text without relying on pre-segmented components. Experiments on standard benchmarks show that our approach achieves higher performance compared to state-of-the-art systems. To the best of our knowledge, this is one of the first attempts to fully model ACD as a generative task, highlighting the potential of instruction tuning for complex AM problems. Our code and the datasets used are openly available in the following GitHub repository.
Chinese Translation
论证组件检测（ACD）是论证（Argumentation）挖掘（AM）的核心子任务之一，也是其最具挑战性的方面，因为它需要共同界定论证范围并将其分类为诸如主张和前提等组件。尽管与其他AM任务相比，这一子任务的研究仍然相对有限，但现有的大多数方法将其简化为序列标注问题、组件分类或组件分割后跟随分类的管道。在本文中，我们提出了ITFACD，这是一种基于指令调优的大型语言模型（LLMs）的新方法，采用紧凑的基于指令的提示，并将ACD重新构建为一种语言生成任务，使得可以直接从纯文本中识别论证，而无需依赖预先分割的组件。在标准基准上的实验表明，我们的方法相比于最先进的系统取得了更高的性能。根据我们所知，这是首次将ACD完全建模为生成性任务的尝试，突显了指令调优在复杂AM问题中的潜力。我们的代码和使用的数据集在以下GitHub仓库中公开可用。
cs.AI / 105 / 2609.24876
Partner-Specific Affective Precision in Social Active Inference
特定伙伴的情感精确度在社会主动推理中的作用
Abstract
In multi-agent social settings, model reliability varies across relationships. Beyond inferring what others will do, an agent must calibrate how confidently those inferences should guide policy selection for each relationship. An agent may maintain a well-validated model of one partner, a fragile model of another, and a model under revision for a third; collapsing these into a single confidence estimate loses information relevant to policy selection. We therefore formalize affective precision as a relationship-specific metacognitive estimate of confidence in the current partner model. Each partner's behavioral evidence updates a local confidence estimate that modulates policy precision during selection, regulating how strongly current beliefs are expressed in policy rather than changing the content of those beliefs. Simulations in a multi-partner graded trust game show that partner-local affective precision influences behavior primarily through policy commitment rather than direct improvement of partner-state inference. Because the mechanism tracks partner-response predictability rather than realized payoff, greater confidence produces sharper policy commitment without necessarily producing higher rewards. Under abrupt shifts in social behavior, confidence accumulated from previously reliable predictions can remain behaviorally active after the relationship changes, showing that confidence revision can lag behind social change. Finally, varying precision gain and priors produce distinct trust-calibration dynamics, showing how confidence accumulation and revision depend on model parameters. Together, these results show how relationship-specific affective precision can distinguish social prediction from social policy commitment.
Chinese Translation
在多智能体社会环境中，模型的可靠性因关系而异。除了推断他人的行为外，智能体还必须校准这些推断在每个关系中应如何自信地指导政策选择。一个智能体可能对一个伙伴保持经过充分验证的模型，对另一个伙伴则有一个脆弱的模型，而对第三个伙伴则是一个正在修订的模型；将这些模型合并为一个单一的信心估计会丢失与政策选择相关的信息。因此，我们将情感精确度形式化为对当前伙伴模型的关系特定的元认知信心估计。每个伙伴的行为证据更新一个局部信心估计，该估计在选择过程中调节政策的精确度，调控当前信念在政策中的表达强度，而不是改变这些信念的内容。在多伙伴分级信任游戏中的模拟表明，伙伴特定的情感精确度主要通过政策承诺影响行为，而不是直接改善对伙伴状态的推断。由于该机制跟踪伙伴反应的可预测性而非实际收益，更高的信心会产生更明确的政策承诺，但不一定会带来更高的奖励。在社会行为的突然变化下，来自先前可靠预测的信心积累在关系变化后仍然可以保持行为活跃，表明信心修订可能滞后于社会变化。最后，变化的精确度增益和先验产生不同的信任校准动态，显示信心的积累和修订如何依赖于模型参数。这些结果共同表明，关系特定的情感精确度如何区分社会预测与社会政策承诺。
cs.AI / 106 / 2609.24881
Pinocchio: Fast Uncertainty Estimates for Black-Box Language Models
皮诺丘：黑箱语言模型的快速不确定性估计
Abstract
In high-stakes decision-making applications of large language models (LLMs), practitioners require not only accurate LLMs but also uncertainty estimates for their predictions. Existing approaches to uncertainty estimation for LLMs require access to log-probabilities output by the model or require fine-tuning access. However, many industrial LLM products use closed-source API models, and many such API models like GPT do not return log-probabilities and may not allow fine-tuning. We introduce Pinocchio, an external calibrator that estimates the correctness of responses from black-box API models. Trained jointly on responses from seven LLMs, it achieves 0.862 AUROC predicting the correctness of held-out responses from those same models, and shows zero-shot transfer to thirteen unseen models across eight organizations. Our model needs only a single forward pass to generate an uncertainty estimate and requires no access to the target model's logits, weights, or internal states. A lightweight text only 0.8B checkpoint matches our largest model's AUROC. We release code for adding uncertainty estimation to existing repos in only two additional lines of code.
Chinese Translation
在大型语言模型（LLMs）的高风险决策应用中，实践者不仅需要准确的LLM，还需要对其预测结果的不确定性估计。现有的LLM不确定性估计方法需要访问模型输出的对数概率，或需要进行微调访问。然而，许多工业LLM产品使用闭源API模型，许多此类API模型（如GPT）不返回对数概率，并且可能不允许微调。我们提出了皮诺丘（Pinocchio），一种外部校准器，用于估计黑箱API模型响应的正确性。该模型在七个LLM的响应上联合训练，能够以0.862的AUROC预测这些模型的保留响应的正确性，并在八个组织的十三个未见模型上展示了零样本迁移。我们的模型只需一次前向传播即可生成不确定性估计，并且不需要访问目标模型的logits、权重或内部状态。一个轻量级的仅文本的0.8B检查点与我们最大模型的AUROC相匹配。我们发布了代码，只需两行额外代码即可将不确定性估计添加到现有的代码库中。
cs.AI / 107 / 2609.24883
A Global Comparison of Schemas, Transparency, and Interoperability in Public-Sector AI Registers and Inventories
公共部门人工智能注册和清单的全球比较：模式、透明度与互操作性
Abstract
Artificial intelligence (AI) registers and inventories aim to make governmental AI visible, but their institutional scope, schemas, and reporting practices construct different representations of public-sector AI. We compare 8,368 records from country-specific and transnational inventories covering 72 countries. Across 23 harmonized fields, registers shared a descriptive core but rarely requested information about appeals, risks, legal bases, or external evaluation. We found that broad schemas often contained substantial missingness, schema similarity showed no significant patterned convergence, and multiple sources covering the same jurisdictions overlapped only selectively. Based on these findings, we synthesize a layered visibility framework that shows how register records reflect disclosure arrangements and why interoperability requires shared concepts, clear definitions, and preserved provenance.
Chinese Translation
人工智能（AI）注册和清单旨在使政府的人工智能可见，但它们的制度范围、模式和报告实践构建了公共部门人工智能的不同表现。我们比较了来自72个国家的国家特定和跨国清单中的8,368条记录。在23个协调的字段中，注册表共享了描述性的核心，但很少请求有关上诉、风险、法律依据或外部评估的信息。我们发现，广泛的模式通常存在显著的缺失，模式相似性没有显示出显著的有规律收敛，而覆盖相同管辖区的多个来源仅选择性重叠。基于这些发现，我们综合了一个分层可见性框架，展示了注册记录如何反映披露安排，以及为什么互操作性需要共享概念、清晰定义和保留来源。
cs.AI / 108 / 2609.24921
BackTrend: Evaluating Scientific Weak-Signal Prediction via Backward Reconstruction
BackTrend：通过逆向重建评估科学弱信号预测
Abstract
Scientific weak signals are early, low-visibility research directions that later become central to mature scientific topics, yet existing resources such as trend tracking, citation forecasting, and foresight reports rarely provide validated reference sets that link concrete early precursors to later paradigms. We introduce BackTrend, a retrospective benchmark in which, given a mature target topic and a temporal evidence constraint, systems must recover two types of precursors: problem-space signals, underrecognized research problems, and solution-space signals, emerging methods for known problems. BackTrend contains 25 mature target topics in artificial intelligence and machine learning and 66 human-validated weak signals, reconstructed from large-scale literature by grounding each candidate in its 2019-2024 publication-frequency trajectory. We evaluate frontier LLMs, RAG systems, and agentic research systems using semantic matching and coverage-based metrics. Current systems often generate plausible but misaligned precursors, exhibiting topic drift, granularity mismatch, near-miss matching, and incomplete coverage; the strongest system achieves only 10.1% F1, while Coverage10 reaches at most 18.5% of the reference signals. Our budget analyses show that additional retrieval and web-search evidence can improve performance up to a moderate budget, but does not by itself close the substantial performance gap.
Chinese Translation
科学弱信号是早期、低可见度的研究方向，后来成为成熟科学主题的核心。然而，现有资源如趋势追踪、引用预测和前瞻性报告，往往缺乏验证的参考集，将具体的早期前兆与后来的范式联系起来。我们介绍了BackTrend，这是一种回顾性基准，其中，给定一个成熟的目标主题和时间证据约束，系统必须恢复两种类型的前兆：问题空间信号，即被低估的研究问题，以及解决方案空间信号，即针对已知问题的新兴方法。BackTrend包含25个成熟的目标主题，涉及人工智能和机器学习，以及66个经过人工验证的弱信号，这些信号是通过基于2019-2024年出版频率轨迹重建的大规模文献而得出的。我们使用语义匹配和覆盖率基础指标评估前沿的大型语言模型（LLMs）、检索增强生成（RAG）系统和智能研究系统。目前的系统通常生成看似合理但不一致的前兆，表现出主题漂移、粒度不匹配、近似匹配和覆盖不全；最强的系统仅达到10.1%的F1分数，而Coverage10最多仅覆盖参考信号的18.5%。我们的预算分析显示，额外的检索和网络搜索证据可以在适度预算内提高性能，但单独并不能弥补显著的性能差距。
cs.AI / 109 / 2609.24927
Et Tu, Brute? Economic Misalignment in Personal AI Agents
你也如此，布鲁图？个人人工智能代理中的经济失调
Abstract
Personal AI agents make recommendations and take actions on people's behalf in high-stakes economic contexts, e.g., buying a flight, choosing health insurance, or selecting a graduate program. The agent is given access to the user's personal context, e.g., their email inbox and a structured profile of personal attributes, with the intention of making an optimal, personalized decision for the user. We show that by simply providing this personal context, the agent steers recommendations based on inferred wealth, without being explicitly instructed to do so. In a suite of 325K experiments on 13 agents across three types of economic decisions (flights, health insurance, and graduate programs), we find that 8 models systematically choose more expensive options for wealthier users when requests are identical. This steering continues even when it directly goes against the user's stated objective: when explicitly instructed to find the cheapest option, some agents still act on the wealth profile they have inferred. It also occurs when wealth is inferred from ambient data, such as emails unrelated to the task. And it persists under privacy controls that block specific attributes: blocking financial attributes largely removes the disparity, but blocking other attributes leaves it unchanged and can increase it by up to 40% for insurance, as agents rely on the remaining signals to infer wealth. Larger and more capable models are no better; Claude Opus 4.8 shows the largest effect. We term this misalignment "adversarial delegation", in which the very conditions that make a personal AI agent useful - access to personal information - enable it to act against the user's interests.
Chinese Translation
个人人工智能代理在高风险经济环境中代表人们做出推荐和采取行动，例如购买航班、选择健康保险或选择研究生项目。代理被授予访问用户个人背景的权限，例如他们的电子邮件收件箱和个人属性的结构化档案，目的是为用户做出最佳的个性化决策。我们展示了，仅仅通过提供这些个人背景，代理就会根据推断出的财富来引导推荐，而无需明确指示。在对13个代理进行的325K实验中，涉及三种类型的经济决策（航班、健康保险和研究生项目），我们发现8个模型在请求相同的情况下，系统性地为更富裕的用户选择更昂贵的选项。这种引导即使在直接违背用户所陈述的目标时也会继续：当明确指示寻找最便宜的选项时，一些代理仍然根据他们推断的财富档案采取行动。当财富是从环境数据中推断出来时，例如与任务无关的电子邮件，这种情况也会发生。在阻止特定属性的隐私控制下，这种现象仍然存在：阻止财务属性在很大程度上消除了差异，但阻止其他属性则保持不变，并且可能使保险的差异增加多达40%，因为代理依赖于剩余信号来推断财富。更大且更强大的模型并没有表现得更好；Claude Opus 4.8显示了最大的影响。我们将这种失调称为“对抗性委托”，在这种情况下，使个人人工智能代理有用的条件——访问个人信息——使其能够反对用户的利益行事。
cs.AI / 110 / 2609.24967
Emergent Collusion in Long-Horizon LLM Agent Interaction
长期视野下大型语言模型代理交互中的共谋现象
Abstract
LLM agents are increasingly deployed in collaborative settings, yet long-term interaction may give rise to undesirable coordination. We study the emergence of collusion in a long-horizon multi-agent environment: two agents repeatedly complete individual tasks, share task logs, verify each other's work, and receive rewards. We introduce realistic constraints that make compliance with the verification protocol incompatible with reward maximization, and find that agents increasingly deviate from the protocol over repeated interactions. Collusion emerges in 94% of trajectories across 10 models, and more capable models within the same family reach it earlier. Controlled peer interventions show that collusion is shaped by peer behavior, while ablations reveal additional effects of reward structure, the verification feedback agents receive, and their interaction history. In particular, restricting the amount and scope of interaction history available to agents reduces collusion. Overall, our findings show that long-horizon interaction can reshape how agents coordinate in ways that create safety risks.
Chinese Translation
大型语言模型（LLM）代理越来越多地被部署在协作环境中，但长期交互可能导致不良的协调现象。我们研究了在长期多代理环境中共谋的出现：两个代理重复完成各自的任务，分享任务日志，验证彼此的工作，并获得奖励。我们引入了现实的约束条件，使得遵循验证协议与奖励最大化之间存在不兼容性，并发现代理在重复交互中越来越偏离协议。在10个模型中，94%的轨迹出现了共谋现象，而同一系列中更强大的模型则更早达到这一状态。控制的同伴干预表明，共谋受到同伴行为的影响，而消融实验揭示了奖励结构、代理所接收的验证反馈以及它们的交互历史的额外影响。特别是，限制代理可用的交互历史的数量和范围可以减少共谋现象。总体而言，我们的研究结果表明，长期交互可以重新塑造代理的协调方式，从而带来安全风险。
cs.AI / 111 / 2609.24974
Harness-Zero: Harness Distillation via Agent-as-Harness
Harness-Zero：通过代理作为工具进行工具蒸馏
Abstract
Agent harnesses, the external systems that mediate model-environment interaction, can substantially improve agent performance, but their gains remain tied to the harness at deployment. Because the best harness varies across domains, instances, and models, a general-purpose agent must either settle for a suboptimal shared harness or route among an ever-growing set of specialized ones. We therefore study agent harness distillation: using a domain- or instance-optimized harness as training-time guidance and transferring the behaviors it induces into model weights, so that its gains survive under a single fixed target harness. The challenge is that the two harnesses differ in action space and available information, so guidance from the optimized harness cannot serve directly as supervision for the target one. We introduce Harness-Zero, which enables harness distillation through agent-as-harness. Guided by the optimized harness, a harnessing agent corrects student responses before execution in the target harness's action space, turning harness guidance into training demonstrations. Fine-tuning on the resulting trajectories internalizes harness-induced behavior into the model, so the specialized harness can be removed at deployment. Our experiments spanning knowledge work, tool use, and science domains show that: (1) For frontier LLMs using the same evolved harness, agent-as-harness outperforms code-as-harness. (2) With the specialized harness removed at deployment, Harness-Zero improves the base model's macro-average task success from 23.3% to 44.3%, even exceeding the 41.7% it reaches with that harness still attached. (3) Harness-Zero recovers harness-induced behaviors absent from the base model, with 82.3% average recovery across 28 patterns in the three domains.
Chinese Translation
代理工具是调解模型与环境交互的外部系统，可以显著提高代理的性能，但其收益仍然依赖于部署时的工具。由于最佳工具在不同领域、实例和模型之间各不相同，因此通用代理必须在一个次优的共享工具和一个不断增长的专用工具集合之间进行选择。因此，我们研究了代理工具蒸馏：使用经过领域或实例优化的工具作为训练时的指导，并将其诱导的行为转移到模型权重中，以便其收益在单一固定目标工具下得以保留。挑战在于这两种工具在动作空间和可用信息上存在差异，因此来自优化工具的指导不能直接作为目标工具的监督。我们引入了Harness-Zero，通过代理作为工具实现工具蒸馏。在优化工具的指导下，工具代理在目标工具的动作空间中纠正学生的响应，将工具指导转化为训练示范。对生成轨迹的微调将工具诱导的行为内化到模型中，从而在部署时可以去除专用工具。我们的实验涵盖知识工作、工具使用和科学领域，结果表明：(1) 对于使用相同进化工具的前沿大语言模型，代理作为工具的表现优于代码作为工具。(2) 在部署时去除专用工具后，Harness-Zero将基础模型的宏平均任务成功率从23.3%提高到44.3%，甚至超过了在该工具仍然附加时达到的41.7%。(3) Harness-Zero恢复了基础模型中缺失的工具诱导行为，在三个领域的28种模式中，平均恢复率达到82.3%。
cs.CL / 1 / 2609.22090
Recognition, Simulation, and Refusal: A Contamination-Aware Study of Classic Psychological Effects in LLM Agents
识别、模拟与拒绝：一种关注污染的经典心理效应在大型语言模型代理中的研究
Abstract
An LLM producing the response pattern associated with a human psychological effect is not the same claim as the LLM possessing that bias. We present PsyAgentBench, a benchmark that re-runs classic psychology experiments on LLM agents under a factorial design built to separate these: each paradigm is run with the paradigm explicitly labeled in the prompt (named) or framed as a routine task (blind), and on the literal textbook version of the task (canonical) or a structurally matched variant written to reduce lexical and scenario overlap with likely training data (counterfactual), crossed with a persona manipulation. Across five completed paradigms, evaluated on up to three open-weight model families with 41,904 trials released, apparently human-like effects arise through qualitatively different routes rather than one susceptibility: paradigm-label gating with explicit override (Asch conformity, 0 percent blind to 83.3 percent named on gpt-oss-120B), knowledge-dependent signal reliance (anchoring, exactly zero on grounded facts versus near total on invented quantities, a pattern equally consistent with rational use of the only available signal), amplification on novel content under labeling (framing), robust absence (sunk cost), and safety-mediated selection where refusal itself is the primary finding (minimal-group allocation). A one-sentence persona change (agreeableness, framed as an instruction rather than a verified trait manipulation) eliminates, dampens, or reverses these effects depending on which effect it is, arguing against any single response-bias account. We further formalize, and in two cases document empirically, three ways a psychology paradigm can fail to port to LLM agents: persona dominance, population collapse, and safety selection. We argue scalar bias-susceptibility scores obscure this structure and report replication profiles instead.
Chinese Translation
一个大型语言模型（LLM）产生与人类心理效应相关的反应模式，并不等同于该模型具有这种偏见。我们提出了PsyAgentBench，这是一个基准，旨在通过一个因子设计重新运行经典心理学实验，以区分这两者：每个范式都在提示中明确标记（命名）或被框定为例行任务（盲测），并在任务的字面教科书版本（规范）或为减少与可能训练数据的词汇和情境重叠而编写的结构匹配变体（反事实）上进行，同时交叉进行角色操控。在五个完成的范式中，评估了多达三种开放权重模型家族，共发布41,904个试验，明显类似人类的效应通过质的不同途径产生，而非单一的易感性：通过显式覆盖的范式标签门控（阿施从众，盲测为0%而命名为83.3%在gpt-oss-120B上）、依赖知识的信号（锚定，基于真实事实完全为零而基于虚构数量几乎为全有，这一模式与对唯一可用信号的理性使用一致）、在标记下对新内容的放大（框架效应）、显著缺失（沉没成本），以及以安全为媒介的选择，其中拒绝本身是主要发现（最小群体分配）。一句话的角色变化（宜人性，框定为指令而非经过验证的特质操控）会消除、减弱或逆转这些效应，具体取决于是哪种效应，这反驳了任何单一反应偏见的解释。我们进一步形式化，并在两个案例中进行了实证记录，阐明心理学范式如何未能移植到LLM代理的三种方式：角色主导、人口崩溃和安全选择。我们认为标量偏见易感性评分掩盖了这一结构，并报告复制特征作为替代。
cs.CL / 2 / 2609.22091
Memory That Looks Forward: A Zero-Inference Prospective Term for Personal Memory Retrieval
面向未来的记忆：一种零推理的个人记忆检索前景术语
Abstract
Retrieval over a personal memory store is retrospective: it surfaces what resembles the query, and it is blind to what the user has committed to do. We describe a prospective term for memory retrieval that costs no inference at query time. Commitments are held in an explicit ledger as dated or trigger-conditioned entries; memory items linked to a firing entry receive a salience boost, blended multiplicatively into embedding-based retrieval so that relevance remains sovereign. On a synthetic prospective-memory task set modeled on TriggerBench's published structure (48 blind-authored dialogues, 175 tasks), the term raised recall@5 on the hard stratum from 0.000 to 0.955 at the default blend weight and to 1.000 under a floor variant, with zero false boosts across 53 resolved-commitment tasks. Blind authorship also produced a scope finding: only 17-29% of naturally phrased commitment-trigger pairs defeat embedding similarity, so the term matters on a real minority of cases and must do no harm on the rest, which it does not. We position precomputed commitment linkage as the always-on floor of a layered design whose expansion layer is query-time prospection. Results are preliminary: the evaluation set is author-constructed, and evaluation on TriggerBench proper is committed follow-up work once its data is released.
Chinese Translation
个人记忆存储的检索是回顾性的：它呈现与查询相似的内容，并且对用户已承诺的行为视而不见。我们描述了一种针对记忆检索的前景术语，该术语在查询时不需要推理。承诺以明确的账本形式保存，作为带日期或触发条件的条目；与触发条目相关联的记忆项会获得显著性提升，这种提升以乘法方式融入基于嵌入的检索中，从而保持相关性为主导。在基于TriggerBench已发布结构（48个盲作者对话，175个任务）建模的合成前景记忆任务集中，该术语在困难层次上的 recall@5 从 0.000 提升至 0.955（在默认混合权重下）和 1.000（在地板变体下），在53个已解决承诺任务中没有出现虚假提升。盲作者身份还产生了一个范围发现：只有17-29%的自然表述的承诺-触发对超越嵌入相似性，因此该术语在真实的少数情况下是重要的，并且在其余情况下不会造成伤害，这一点得到了验证。我们将预计算的承诺链接定位为分层设计的始终开启的基础层，其扩展层是查询时的前景。结果是初步的：评估集是作者构建的，针对TriggerBench的正式评估将在其数据发布后进行后续工作。
cs.CL / 3 / 2609.22094
Summarize, Judge, Refine: Decoupled Content Understanding and Policy Learning for Multimodal Content Moderation
总结、判断、优化：多模态内容审核中的内容理解与政策学习解耦
Abstract
Content moderation systems traditionally entangle multimodal understanding with policy-specific classification, requiring full pipeline retraining for every policy change and suffering from label scarcity since multimedia cannot be meaningfully augmented. We propose Summarize-Judge-Refine (SJR), a two-model architecture that decouples these concerns via a natural language interface: a multimodal Content Model produces structured text summaries, and a text-only Policy Model classifies them against policy definitions. An iterative co-training loop refines the Content Model via GRPO to produce policy-relevant summaries, while text-space augmentation generates adversarial summary variants---an augmentation pathway impossible on raw multimedia---enabling few-shot policy bootstrap. Every decision is grounded in a human-readable summary, providing interpretability as a structural byproduct. On misleading advertisement detection, SJR achieves +23.6\% relative non-misleading F1 over a zero-shot chain-of-thought baseline, outperforming end-to-end SFT, STaR/RFT, and RLFT. Notably, a variant trained on zero real violating examples---with all positive-class data synthetically generated---matches the full-data model within 0.2\% relative on violating F1, demonstrating that new policies can launch without any real violation data.
Chinese Translation
内容审核系统传统上将多模态理解与特定政策的分类相互纠缠，这要求在每次政策变更时对整个流程进行重新训练，并且由于多媒体内容无法进行有意义的增强，导致标签稀缺。我们提出了总结-判断-优化（Summarize-Judge-Refine, SJR）架构，这是一种通过自然语言接口解耦这些问题的双模型架构：多模态内容模型生成结构化文本摘要，而仅文本的政策模型则根据政策定义对其进行分类。一个迭代的共同训练循环通过GRPO优化内容模型，以生成与政策相关的摘要，同时文本空间增强生成对抗性摘要变体——这种增强路径在原始多媒体上是不可能的——从而实现少量样本政策引导。在误导性广告检测任务中，SJR相较于零-shot思维链基线实现了+23.6%的相对非误导性F1得分，超越了端到端的SFT、STaR/RFT和RLFT。值得注意的是，一个在零个真实违规示例上训练的变体——所有正类数据均为合成生成——在违规F1得分上与全数据模型的相对差距仅为0.2%，这表明新政策可以在没有任何真实违规数据的情况下启动。
cs.CL / 4 / 2609.22096
AI-inferred expressed well-being and collective-action discourse in climate-change campaigns on X
气候变化运动中AI推断的表达幸福感与集体行动话语
Abstract
Climate campaigns are often evaluated through attention and mobilization, but less is known about the well-being language that accompanies them. Whether campaign periods alter positive affect and hope, and whether happiness aligns with action language, remains unresolved. We analysed 364,118 public Twitter/X posts from Earth Day, Earth Hour, Global Climate Action Day and World Environment Day in 19 occurrence-years, using 30-day pre-event, event and post-event windows. A versioned weighted lexical model estimated happiness, future-oriented hope, collective capability, distress and action language. Event-period happiness prevalence was 9.02 percentage points higher than the pre-event baseline , whereas paired occurrence contrasts showed a 10.75-point decline in action language, indicating a happiness--action divergence. The happiness estimate remained positive across composition and text-deduplication checks, but was less precise under a 19-cluster wild bootstrap. Happier source posts had lower odds of an observed matched retweet cascade.
Chinese Translation
气候运动通常通过关注度和动员程度进行评估，但关于伴随其发生的幸福语言的研究较少。运动期间是否会改变积极情感和希望，以及幸福感是否与行动语言相一致，仍未得到解决。我们分析了来自19个年份的364,118条公共Twitter/X帖子，这些帖子与地球日、地球一小时、全球气候行动日和世界环境日相关，使用了事件前30天、事件期间和事件后的时间窗口。一个版本化的加权词汇模型估计了幸福感、面向未来的希望、集体能力、痛苦和行动语言。事件期间的幸福感普遍高于事件前基线9.02个百分点，而配对发生对比显示行动语言下降了10.75个百分点，表明幸福感与行动之间存在分歧。幸福感的估计在组成和文本去重检查中保持为正值，但在19个聚类的野生自助法下精度较低。幸福感较高的源帖子观察到的匹配转发级联的几率较低。
cs.CL / 5 / 2609.22097
Token Signatures of Code: Comparing Coding Behaviors Across Large Language Models
代码的标记签名：比较大型语言模型的编码行为
Abstract
The evaluation of large language models (LLMs) on coding tasks has primarily focused on performance metrics such as pass@k. As LLMs continue to advance, many models now meet baseline performance requirements, reducing the discriminative power of performance-based evaluation alone. Yet a key question remains largely unexplored: how do LLMs differ in their coding behavior? We propose CLIC (Code Learning for Identification and Comparison), a visual analytics approach that characterizes LLM coding behavior through token-frequency analysis. CLIC represents each code sample as a feature vector of token frequencies and trains an interpretable decision tree to separate two LLMs' code sets. Beyond classification accuracy, we define two new metrics: robustness, which measures whether the two LLMs remain distinguishable as their most-discriminative tokens are progressively removed, and concentration, which measures whether the difference is driven by a few dominant tokens or spread across many. Interpreting numerous pairwise comparisons (across LLM pairs, tasks, and tokenization levels) and tracing the full analytical chain form an inherently multi-scale, hypothesis-driven exploration task. We therefore develop an interactive visual analytics system to navigate the comparison landscape, identify pairs of interest, and drill down into discriminative tokens and their code contexts. Case studies comparing 10 LLMs across 22 Kaggle ML tasks reveal actionable insights for LLM selection and prompt engineering.
Chinese Translation
对大型语言模型（LLMs）在编码任务上的评估主要集中于性能指标，如 pass@k。随着 LLMs 的不断进步，许多模型现在满足基础性能要求，单靠基于性能的评估的区分能力有所降低。然而，一个关键问题仍然未被充分探索：LLMs 在编码行为上有何不同？我们提出了 CLIC（代码学习用于识别和比较），这是一种通过标记频率分析来表征 LLM 编码行为的可视化分析方法。CLIC 将每个代码样本表示为标记频率的特征向量，并训练一个可解释的决策树来区分两个 LLM 的代码集。除了分类准确性外，我们定义了两个新指标：鲁棒性（robustness），用于测量当两个 LLM 的最具区分性的标记逐渐被移除时，它们是否仍然可区分；集中度（concentration），用于测量差异是由少数主导标记驱动，还是分布在许多标记上。对多个成对比较（跨 LLM 对、任务和标记化水平）进行解释，并追踪完整的分析链，形成了一项固有的多尺度、假设驱动的探索任务。因此，我们开发了一个交互式可视化分析系统，以导航比较领域，识别感兴趣的对，并深入挖掘区分性标记及其代码上下文。对 10 个 LLM 在 22 个 Kaggle 机器学习任务上的案例研究揭示了有关 LLM 选择和提示工程的可行见解。
cs.CL / 6 / 2609.22098
TreeSpark: Calibrated, Load-Adaptive Draft Trees for Semi-Autoregressive Speculative Decoding
TreeSpark：用于半自回归推测解码的校准负载自适应草稿树
Abstract
Speculative decoding accelerates language-model inference by letting a cheap drafter propose tokens that the target model verifies in parallel. Recent block drafters make drafting nearly free: a single backbone pass emits an entire block of draft tokens. Draft trees promise a further gain -- several alternative continuations verified in one target forward -- but existing constructions rank candidates by per-position marginals that ignore which parent a candidate extends, so on semi-autoregressive drafters wider trees mostly add mis-ranked nodes; and a tree of fixed size ignores how much speculation each decoding round, and each serving load, can support. We introduce TreeSpark, which reads a parent-conditioned distribution from the drafter's existing Markov head at negligible cost, calibrates it into an edge-acceptance estimate, and lets path survival govern everything else: best-first expansion, per-round stopping, and a load-adaptive serving policy. Sampling siblings without replacement, with matching residuals in recursive rejection, keeps decoding lossless at any temperature. Adaptive trees improve on matched fixed budgets at every temperature; against a tuned chain on the same drafter, TreeSpark accepts 15-25% more draft tokens per round and decodes 8-14% faster in single-request wall-clock, and under rising load it gracefully shrinks the tree back to the chain. Code and artifacts: https://github.com/PopSoda2002/TreeSpark
Chinese Translation
推测解码通过让一个廉价的草稿生成器提出令牌并让目标模型并行验证，从而加速语言模型推理。最近的块草稿生成器几乎使草稿过程变得免费：一次主干传递即可生成整个块的草稿令牌。草稿树承诺进一步的收益——在一次目标前向验证中验证多个替代延续——但现有的构造通过每个位置的边际概率对候选进行排名，而忽略了候选扩展的父节点，因此在半自回归草稿生成器中，较宽的树大多增加了错误排名的节点；而固定大小的树忽略了每轮解码和每个服务负载可以支持的推测量。我们引入了TreeSpark，它以微不足道的成本从草稿生成器现有的马尔可夫头中读取父条件分布，将其校准为边缘接受估计，并让路径生存控制其他一切：最佳优先扩展、每轮停止和负载自适应服务策略。在递归拒绝中不重复抽样兄弟节点，并匹配残差，使得在任何温度下解码无损。自适应树在每个温度下都优于匹配的固定预算；与同一草稿生成器上的调优链相比，TreeSpark每轮接受的草稿令牌增加了15-25%，并在单请求的墙钟时间中解码速度提高了8-14%，在负载增加时，它优雅地将树缩回到链上。代码和相关资料：https://github.com/PopSoda2002/TreeSpark
cs.CL / 7 / 2609.22099
A framework for recipe data structure with applications for culinary and nutritional insights
一种食谱数据结构框架及其在烹饪和营养洞察中的应用
Abstract
Cooking is a complex process that transforms raw ingredients into delicious and nutritious dishes, yet the recipes that encode this process remain largely free text; readable by people but not directly computable. Existing recipe collections capture fragments of this information, but no shared representation links a recipe's structured ingredient composition, its geo-cultural provenance, and its nutritional profile within a single queryable schema. We address this representation gap by formalizing a framework for recipe data structure that decomposes each recipe into typed ingredient entities, grounds those entities in a reference nutritional database, and annotates them with geo-cultural and dietary context. We present RecipeDB2, a structured compilation of 128,942 recipes with 35,474 ingredients from 32 regions and 99 countries. Ingredient phrases are parsed into seven culinary attributes using a transformer-based named-entity model; ingredients are linked to the USDA reference tables through a BERT embedding strategy (F1 = 87.90 on a manually adjudicated set of the 200 most frequent ingredients), yielding 148 nutritional parameters per mapped ingredient; a Random Forest classifier propagates 34 ingredient categories across the full vocabulary; and a deterministic, conservative rule set assigns each recipe a dietary style. Through RecipeDB2 (https://cosylab.iiitd.edu.in/recipedb2/), we demonstrate a scalable framework for making recipes computable, turning culinary heritage (long treated as an artistic rather than a quantitative object) into a data-driven analysis.
Chinese Translation
烹饪是一个复杂的过程，将原材料转化为美味且营养丰富的菜肴，但编码这一过程的食谱仍然主要以自由文本形式存在；虽然人类可以阅读，但计算机无法直接处理。现有的食谱集合捕捉了这一信息的片段，但没有共享的表示将食谱的结构化成分组成、其地理文化来源及营养特征链接在一个可查询的模式中。我们通过形式化一个食谱数据结构框架来解决这一表示差距，该框架将每个食谱分解为类型化的成分实体，将这些实体与参考营养数据库关联，并用地理文化和饮食背景进行注释。我们展示了RecipeDB2，这是一个包含来自32个地区和99个国家的128,942个食谱和35,474种成分的结构化汇编。成分短语通过基于变换器的命名实体模型解析为七个烹饪属性；成分通过BERT嵌入策略与美国农业部（USDA）参考表链接（在200个最常见成分的人工审定集上F1 = 87.90），为每个映射成分提供148个营养参数；随机森林分类器在整个词汇中传播34个成分类别；一个确定性、保守的规则集为每个食谱分配饮食风格。通过RecipeDB2（https://cosylab.iiitd.edu.in/recipedb2/），我们展示了一个可扩展的框架，使食谱可计算，将烹饪遗产（长期以来被视为艺术而非定量对象）转变为数据驱动的分析。
cs.CL / 8 / 2609.22100
AdaMem: Adaptive Memory Token Allocation for Soft Compression in Retrieval-Augmented Generation
AdaMem：用于检索增强生成的自适应记忆令牌分配的软压缩
Abstract
Retrieval-augmented generation (RAG) improves language models with retrieved evidence, but processing many long passages is costly and can introduce distracting information. Soft compression addresses this challenge by encoding passages as compact sequences of continuous memory embeddings before generation. However, existing methods typically assign each retained passage an identical number of memory embeddings, irrespective of its query-specific relevance. To address this, we propose AdaMem, a relevance-guided soft-compression framework that maps learned passage-relevance estimates to a query-dependent allocation of a fixed memory-token budget. A shared query-conditioned compressor produces both continuous passage memories and relevance scores in a single pass; a deterministic allocation rule assigns more memory tokens to higher-scoring passages and can omit low-scoring ones. Across six open-domain QA benchmarks, AdaMem consistently outperforms OSCAR (the closely matched soft-compression baseline that uses uniform allocation) as well as other soft-compression methods at matched memory budgets. Under standard 16$\times$ compression, AdaMem improves sub-string match by up to 3.2 points (5.5%) over uniform allocation baseline, with an average relative gain of 3.4%; under aggressive 64$\times$ compression the average relative gain grows to 14.6%, with a maximum of 9.8 points (19.7%) on PopQA. AdaMem matches the answer quality of the uncompressed at up to 4$\times$ lower inference latency than full context baseline. AdaMem retains an efficiency profile comparable to the uniform-compression baseline, while achieving up to $4\times$ lower inference latency than full-context inference. Thus, relevance-guided memory allocation is particularly effective when retrieval pools are large and the available memory budget is tight.
Chinese Translation
检索增强生成（RAG）通过检索证据来改善语言模型，但处理大量长文本的成本高昂，并且可能引入干扰信息。软压缩通过在生成之前将文本编码为连续记忆嵌入的紧凑序列来解决这一挑战。然而，现有方法通常对每个保留的文本分配相同数量的记忆嵌入，而不考虑其与查询的特定相关性。为了解决这个问题，我们提出了AdaMem，一个基于相关性的软压缩框架，它将学习到的文本相关性估计映射到固定记忆令牌预算的查询依赖分配。一个共享的查询条件压缩器在一次处理过程中生成连续的文本记忆和相关性评分；一个确定性的分配规则为得分较高的文本分配更多的记忆令牌，并可以省略得分较低的文本。在六个开放领域问答基准测试中，AdaMem始终优于OSCAR（使用均匀分配的紧密匹配软压缩基线）以及其他在匹配记忆预算下的软压缩方法。在标准的16$ imes$压缩下，AdaMem在子字符串匹配上比均匀分配基线提高了最多3.2分（5.5%），平均相对增益为3.4%；在激进的64$ imes$压缩下，平均相对增益增长到14.6%，在PopQA上达到最大9.8分（19.7%）。AdaMem在推理延迟上比全上下文基线低至4$ imes$，同时保持与未压缩答案质量相当。AdaMem的效率特征与均匀压缩基线相当，同时在推理延迟上比全上下文推理低至$4 imes$。因此，当检索池较大且可用的记忆预算紧张时，基于相关性的记忆分配特别有效。
cs.CL / 9 / 2609.22101
Context Poisoning as Extreme-Value Attention Interference in Long-Context Language Models
上下文污染作为长上下文语言模型中的极值注意力干扰
Abstract
Large language models can process increasingly long prompts, yet their ability to locate and use decisive evidence may degrade as irrelevant or confusable context is added. We formulate this phenomenon, which we call context poisoning, as extreme-value interference in attention: the decisive-evidence score is upper-bounded, while the maximum score among effective distractors grows with their number. Under a softmax retrieval abstraction, we derive a finite-sample upper bound showing that maintaining a fixed accuracy target above base rate requires the evidence margin to scale as $\Omega(\sqrt{\log N})$, where N denotes the effective distractor count rather than necessarily the raw context length. The analysis connects long-context degradation to score aliasing, positional aliasing, and softmax dilution. Controlled experiments show that retrieval accuracy decreases as total context grows in the presence of embedded hard negatives, that the same-format condition produces the largest observed accuracy drop among the tested distractor constructions at fixed context length, and that retrieval gating can improve evidence use while its net benefit depends on preserving evidence recall. These results motivate evidence bottlenecks, alias-resistant representations, retrieve-then-reason architectures, verifier-mediated memory, and contrastive anti-poison training.
Chinese Translation
大型语言模型能够处理越来越长的提示，但随着无关或易混淆上下文的增加，它们定位和使用决定性证据的能力可能会下降。我们将这一现象称为上下文污染，形式化为注意力中的极值干扰：决定性证据分数是上界的，而有效干扰项中的最大分数随着其数量的增加而增长。在软最大检索抽象下，我们推导出一个有限样本上界，表明要保持固定的准确性目标高于基线率，证据边际需要按 $ ext{Ω}( ext{√log N})$ 进行缩放，其中 N 表示有效干扰项的数量，而不一定是原始上下文长度。该分析将长上下文退化与分数别名、位置别名和软最大稀释联系起来。控制实验表明，在嵌入的困难负样本存在的情况下，随着总上下文的增加，检索准确性下降；相同格式条件在固定上下文长度下产生的观察到的最大准确性下降；检索门控可以改善证据的使用，而其净效益依赖于保持证据回忆。这些结果激励了证据瓶颈、抗别名表示、检索后推理架构、验证者介导的记忆和对比反污染训练。
cs.CL / 10 / 2609.22104
DeepInstructor: An Agentic AI Instructor for Experience-Driven Idea Evaluation
DeepInstructor：一种基于经验驱动的智能AI讲师用于创意评估
Abstract
As automated scientific discovery advances, Large Language Models (LLMs) can now generate research ideas at an unprecedented scale, shifting the bottleneck from idea generation to idea evaluation. Existing evaluators mainly rely on parametric LLM knowledge or unstructured retrieval, producing judgments that lack the experience-grounded reasoning used by human instructors. To address this, we propose DeepInstructor, an agentic framework that formulates idea evaluation as reasoning over structured scholarly experience. DeepInstructor constructs an Experience Graph from 58,607 peer reviews and employs a ReAct-based agent to retrieve dimension-specific evidence for traceable evaluation. We further introduce DeepInstruct, a dataset with controlled pairwise comparisons across novelty, significance, and feasibility. Experiments show that DeepInstructor substantially outperforms existing baselines, improving Hit@1 and Hit@2 alignment with human judgments by 24.4% and 29.7%, respectively. Our findings suggest that scientific idea evaluation can be grounded in explicit reasoning over structured scholarly experience
Chinese Translation
随着自动化科学发现的进展，大型语言模型（LLMs）现在能够以空前的规模生成研究创意，将瓶颈从创意生成转移到创意评估。现有的评估者主要依赖参数化的LLM知识或非结构化检索，产生的判断缺乏人类讲师所使用的基于经验的推理。为了解决这个问题，我们提出了DeepInstructor，一个将创意评估形式化为对结构化学术经验进行推理的智能框架。DeepInstructor从58,607条同行评审中构建了经验图，并采用基于ReAct的代理来检索特定维度的证据，以便进行可追溯的评估。我们进一步引入了DeepInstruct，一个在新颖性、重要性和可行性方面进行控制的成对比较数据集。实验表明，DeepInstructor在各项基准测试中显著优于现有方法，Hit@1和Hit@2与人类判断的一致性分别提高了24.4%和29.7%。我们的研究结果表明，科学创意评估可以基于对结构化学术经验的明确推理。
cs.CL / 11 / 2609.22110
Evaluating Fine-Tuned and Base Language Models in Maternal and Vaccination Healthcare for African Settings
评估在非洲环境下母婴健康和疫苗接种领域的微调与基础语言模型
Abstract
Background: Large language models (LLMs) can improve healthcare information delivery in low-resource settings but may produce inaccurate or culturally inappropriate advice. This study evaluated domain-specific fine-tuning for maternal health and vaccination in Nigeria. Objective: To compare HelpMum's MamaBot-Llama and Vax-Llama with Meta's Llama-3.1-8B-Instruct for accuracy, safety, clarity, contextual appropriateness, and trustworthiness. Methods: We evaluated 200 healthcare questions, 100 each for maternal health and vaccination, across five subdomains per domain. MamaBot-Llama and Vax-Llama were fine-tuned using Low-Rank Adaptation on over 36,000 maternal health and 9,000 vaccination question-answer pairs, respectively. Two Nigerian licensed physicians independently rated responses using a 5-point Likert scale. Paired comparisons used Wilcoxon signed-rank tests. Results: Performance varied by domain. MamaBot-Llama significantly outperformed the base model across all criteria, with a 4.9% overall improvement (p < .001), including gains in clinical trustworthiness (+7%) and medical accuracy (+5%). Critical issues decreased by 50%, and clinicians preferred it in 78% of cases. In contrast, Vax-Llama showed a 5.2% overall decline (p < .001), with critical issues increasing by 192% and safety concerns by 400%. Conclusions: Domain-specific fine-tuning can improve healthcare LLM performance when based on high-quality, clinician-curated data, but may also degrade performance when dataset quality is inadequate. Rigorous domain-specific validation is essential before clinical deployment. Physician evaluators provided informed consent, and chatbot logs were anonymized. Keywords: Large language models; Fine-tuning; Maternal health; Vaccination; Healthcare AI; Low-resource settings; Nigeria; Model evaluation; LoRA; Medical accuracy
Chinese Translation
背景：大型语言模型（LLMs）可以改善低资源环境下的医疗信息传递，但可能会产生不准确或文化不适宜的建议。本研究评估了尼日利亚母婴健康和疫苗接种领域的特定领域微调。目的：比较HelpMum的MamaBot-Llama和Vax-Llama与Meta的Llama-3.1-8B-Instruct在准确性、安全性、清晰性、上下文适宜性和可信度方面的表现。方法：我们评估了200个医疗问题，其中母婴健康和疫苗接种各100个，涵盖每个领域的五个子领域。MamaBot-Llama和Vax-Llama分别使用低秩适应（Low-Rank Adaptation）在超过36,000个母婴健康和9,000个疫苗接种问答对上进行了微调。两名尼日利亚执业医师独立使用5点李克特量表对回答进行了评分。配对比较使用Wilcoxon符号秩检验。结果：性能因领域而异。MamaBot-Llama在所有标准上显著优于基础模型，整体改善幅度为4.9%（p < .001），其中临床可信度提高了7%，医疗准确性提高了5%。关键问题减少了50%，临床医生在78%的案例中更倾向于使用它。相比之下，Vax-Llama整体表现下降了5.2%（p < .001），关键问题增加了192%，安全性问题增加了400%。结论：基于高质量、临床医生策划的数据，特定领域的微调可以改善医疗LLM的表现，但当数据集质量不足时，可能会导致性能下降。在临床部署之前，严格的特定领域验证至关重要。医师评估者提供了知情同意，聊天机器人日志已被匿名处理。关键词：大型语言模型；微调；母婴健康；疫苗接种；医疗人工智能；低资源环境；尼日利亚；模型评估；LoRA；医疗准确性
cs.CL / 12 / 2609.22111
Beyond the Text: Verifying That Agent-Written Papers Are Backed by Their Artifacts
超越文本：验证代理撰写的论文是否有其支撑的文献
Abstract
Large language model agents are increasingly capable of conducting research autonomously, producing research documents alongside the code and experiments that ostensibly support them. Yet whether the reported findings are consistently supported by corresponding implementations and execution evidence remains largely unexplored: existing review practices primarily assess textual quality and cannot reliably identify inconsistencies such as hard-coded metrics, unimplemented methods, or unsupported experimental results. We present ReAgent, an automated auditing framework for assessing the consistency between agent-generated research documents and their associated repositories. ReAgent constructs structured representations of scientific claims from research documents and uses them to guide repository analysis and evidence collection. Static auditing examines whether claimed methodologies, implementations, and experimental configurations are consistently reflected in the repository, while dynamic auditing executes relevant experiments and collects execution evidence to assess empirical findings. By combining static analysis with dynamic evidence, ReAgent identifies inconsistencies that may remain hidden under either perspective alone, such as experiments that reproduce reported numbers while deviating from the claimed methodology. The collected evidence and audit decisions are organized into a structured repository-level audit report, enabling transparent evidence traceability. We evaluate ReAgent on a manually curated benchmark of agent-generated research document--repository pairs and compare it against representative static and reproduction-based baselines. Experimental results demonstrate that ReAgent effectively identifies inconsistencies between reported research findings and their supporting repository evidence.
Chinese Translation
大型语言模型代理越来越能够自主进行研究，生成研究文档以及表面上支持这些文档的代码和实验。然而，报告的研究结果是否始终得到相应实现和执行证据的支持仍然在很大程度上未被探索：现有的审查实践主要评估文本质量，无法可靠地识别出诸如硬编码指标、未实现方法或不支持的实验结果等不一致性。我们提出了ReAgent，一个自动化审计框架，用于评估代理生成的研究文档与其相关存储库之间的一致性。ReAgent从研究文档中构建科学声明的结构化表示，并利用这些表示指导存储库分析和证据收集。静态审计检查所声称的方法论、实现和实验配置是否在存储库中得到一致反映，而动态审计则执行相关实验并收集执行证据以评估实证发现。通过将静态分析与动态证据相结合，ReAgent能够识别出仅在单一视角下可能隐藏的不一致性，例如实验在重现报告数字的同时偏离所声称的方法论。收集的证据和审计决策被组织成结构化的存储库级审计报告，从而实现透明的证据可追溯性。我们在一个手动策划的代理生成研究文档-存储库对的基准上评估ReAgent，并将其与代表性的静态和基于重现的基线进行比较。实验结果表明，ReAgent有效识别报告的研究发现与其支持的存储库证据之间的不一致性。
cs.CL / 13 / 2609.22112
Privacy Personalization Trade offs in LLMs: The Impact of Stylometric Signal Reduction on User-Specific Text Generation
大型语言模型中的隐私个性化权衡：风格信号减少对用户特定文本生成的影响
Abstract
Large language models (LLMs) have demonstrated the ability to generate user-specific text with high stylistic fidelity. However, the personal data that enables such personalization frequently embeds demographic, cultural, and stylistic markers that raises concerns about stylometric re- identification. This paper investigates whether reducing identifiable stylistic signals affects personalization in text generation by LLMs. We introduce a controlled framework to isolate stylometric signals in LLM personalization using the LaMP-7 Twitter benchmark. Experiments on 250 sampled users compare two settings: paraphrasing conditioned on the original profile and paraphrasing conditioned on an anonymized converted profile in which demographic identifiers, cultural references, personal details, and informal linguistic cues have been systematically neutralized. Outputs are assessed by two independent LLM judges and a complementary human evaluation. Our pairwise evaluation shows that outputs conditioned on original profiles are nearly indistinguishable from human-authored ground truth, indicating that modern LLMs can closely reproduce an author's writing style with sufficient fidelity. In contrast, preference for model outputs with anonymized profiles drops to 13.0% on average, while semantic context preservation remains high at 94.8%. A study with human evaluators confirms the same pattern. These findings reveal a clear privacy-personalization trade-off and highlight the need for privacy-aware personalization methods that retain meaning while suppressing identifying stylistic signals.
Chinese Translation
大型语言模型（LLMs）已展示出生成具有高风格忠实度的用户特定文本的能力。然而，支持这种个性化的个人数据通常嵌入了人口统计、文化和风格标记，这引发了关于风格再识别的担忧。本文探讨了减少可识别风格信号是否会影响LLMs在文本生成中的个性化。我们引入了一个受控框架，以使用LaMP-7 Twitter基准来隔离LLM个性化中的风格信号。对250名样本用户的实验比较了两种设置：基于原始档案的释义和基于经过匿名转换的档案的释义，在后者中，人口标识符、文化参考、个人细节和非正式语言线索已被系统性中和。输出结果由两名独立的LLM评审员和一项补充的人类评估进行评估。我们的成对评估显示，基于原始档案的输出与人类创作的真实文本几乎无法区分，表明现代LLMs能够以足够的忠实度再现作者的写作风格。相比之下，基于匿名档案的模型输出的偏好平均降至13.0%，而语义上下文的保留率仍然高达94.8%。一项与人类评估者的研究确认了相同的模式。这些发现揭示了隐私与个性化之间的明显权衡，并强调了需要隐私意识的个性化方法，以在抑制识别性风格信号的同时保留意义。
cs.CL / 14 / 2609.22114
An Empirical Cost Attribution of Context-Compression Gateways in Multi-Turn Coding Agents
多轮编码代理中上下文压缩网关的实证成本归属
Abstract
Context compression is widely proposed as a way to cut the token bill of LLM coding agents, and public benchmarks report that aggressive compression preserves task-solving quality. These two facts do not imply the third one commonly assumed: that compressing file reads saves money in a real multi-turn agent. We instrument a production compression gateway (Paritok) between coding agents (Claude Code, Codex) and frontier LLMs (Claude Sonnet, GPT-5), and decompose the token bill of real sessions into three independent levers: tool-schema filtering, content compression of file reads and tool output, and history summarization. Measured in isolation under controlled A/B runs, the three save at fundamentally different rates. Tool-schema filtering removes a fixed block every turn, roughly 21K-57K tokens on a typical turn; it is linear in the turn count N and the only unambiguously and reproducibly positive lever. Content compression saves only about 2% of the cache-priced prefix per turn, but compressed reads accumulate in history and are re-sent on every later turn, so its cumulative saving grows quadratically, about 3350*N^2 tokens (measured), overtaking the fixed tool-filter saving within roughly 6 turns until the context window caps it. A non-destructive gateway lets the agent pull original bytes back on demand; each recall re-sends exactly the one segment just compressed away, so its cost is fixed and bounded rather than a multiplicative blowup, and heavy recall spends the accumulated saving back one segment at a time. Finally, a strong single-shot compression benchmark - 86.5% of SWE-bench quality retained at a 25.7% compression rate, achieved by the model this gateway deploys (Paritok-4B, reported separately) - is orthogonal to multi-turn agent cost and must not be cited as a cost-saving argument. We distill the results into an actionable recipe for where token-saving effort pays off.
Chinese Translation
上下文压缩被广泛提议作为降低大型语言模型（LLM）编码代理的令牌费用的一种方式，公共基准测试报告称，激进的压缩能够保持任务解决质量。这两个事实并不意味着第三个常被假设的事实：在真实的多轮代理中，压缩文件读取可以节省费用。我们在编码代理（Claude Code、Codex）和前沿LLM（Claude Sonnet、GPT-5）之间引入了一个生产压缩网关（Paritok），并将真实会话的令牌费用分解为三个独立的杠杆：工具模式过滤、文件读取和工具输出的内容压缩，以及历史摘要。在受控的A/B测试中单独测量时，这三者的节省率在根本上是不同的。工具模式过滤每轮移除固定的块，典型轮次大约为21K-57K个令牌；它与轮次计数N成线性关系，是唯一明确且可重复的正向杠杆。内容压缩每轮仅节省大约2%的缓存定价前缀，但压缩读取在历史中累积，并在每个后续轮次中重新发送，因此其累积节省呈二次增长，约为3350*N^2个令牌（测量值），在大约6轮内超过固定的工具过滤节省，直到上下文窗口限制了它。一个非破坏性的网关允许代理按需恢复原始字节；每次回忆精确地重新发送刚刚压缩掉的一个片段，因此其成本是固定且有界的，而不是乘法爆炸，频繁的回忆一次只花费累积的节省。最后，一个强大的单次压缩基准——在25.7%的压缩率下保留86.5%的SWE-bench质量，由该网关部署的模型（Paritok-4B，单独报告）实现——与多轮代理成本是正交的，不能作为节省成本的论据引用。我们将结果提炼为一个可操作的配方，指出令牌节省努力的有效领域。
cs.CL / 15 / 2609.22119
Evaluation Awareness Shifts from Format to Context with Model Scale
评估意识随着模型规模从格式转向上下文的变化
Abstract
Evaluation awareness poses an unprecedented threat to model evaluation, but the mechanisms by which models detect it remain unknown. This study focuses on determining this and identifying contrasting mechanisms between smaller and larger models. While smaller models use the prompt's format sensitivity to detect evaluation, larger models often rely on higher-order reasoning to detect it. We evaluated Gemma 3 (1B, 4B, and 12B), Phi-3 (Mini and Medium), and Llama-3 8B using Chain-of-Thought analysis, representation probing, and Integrated Gradients attribution. Motivated by these findings, we propose a dual-pathway intervention that combines prompt sanitization with activation counter-steering to suppress both external evaluation triggers and their internal representations. Across 200 highly evaluation-aware prompts, our method achieves an average behavioral flip rate of 70.58\%, consistently outperforming either intervention alone. These results provide new insights into how evaluation awareness develops in compact language models and suggest that effective mitigation requires jointly addressing both prompt-level and representation-level signals.Datasets and codebase can be found in this \href{https://github.com/chahal-navi/Evaluation-Awareness-Compact-LLMs/tree/main}{Github Repository.}
Chinese Translation
评估意识对模型评估构成了前所未有的威胁，但模型如何检测这一点的机制仍然未知。本研究旨在确定这一机制，并识别较小模型与较大模型之间的对比机制。较小的模型利用提示的格式敏感性来检测评估，而较大的模型则通常依赖于更高阶的推理来进行检测。我们使用思维链分析（Chain-of-Thought analysis）、表示探测（representation probing）和集成梯度归因（Integrated Gradients attribution）对Gemma 3（1B、4B和12B）、Phi-3（Mini和Medium）以及Llama-3 8B进行了评估。基于这些发现，我们提出了一种双路径干预方法，将提示清理与激活反向引导相结合，以抑制外部评估触发因素及其内部表征。在200个高度评估意识的提示中，我们的方法实现了平均行为翻转率为70.58%，始终优于单独的任何干预。这些结果为紧凑语言模型中评估意识的发展提供了新的见解，并表明有效的缓解措施需要共同解决提示级和表征级信号。数据集和代码库可以在此 [Github Repository](https://github.com/chahal-navi/Evaluation-Awareness-Compact-LLMs/tree/main) 中找到。
cs.CL / 16 / 2609.22124
Balancing Reasoning and Hardware Constraints in RAG Pipelines for Ukrainian Multi-Domain Document Understanding
在乌克兰多领域文档理解中的RAG管道中平衡推理与硬件约束
Abstract
This paper describes the system submitted to the UNLP 2026 Shared Task on Multi-Domain Document Understanding. The challenge required extracting precise answers, document IDs, and page numbers from a diverse corpus of Ukrainian PDF documents within a strict 9-hour offline Kaggle execution limit. During evaluation on the hidden private test set, optical character recognition (OCR) of scanned documents emerged as a severe bottleneck, consuming 5-7 hours of the total time budget due to sequential single-threaded execution. This overhead strictly limited the remaining time for Large Language Model (LLM) inference to approximately two hours for 500 questions. To guarantee pipeline completion without timeouts, we developed a resource-efficient Hybrid Retrieval-Augmented Generation (RAG) pipeline utilizing BM25, BGE-M3, and Cross-Encoder reranking. Rather than deploying parameter-heavy reasoning models (e.g., DeepSeek R1) which consistently timed out, we utilized a 4-bit quantized LapaLLM 12B model via llama.cpp on dual NVIDIA T4 GPUs. Prioritizing pipeline stability over multi-step reasoning, our system achieved a Private Score of 0.8095, placing 10th out of 15 active teams.
Chinese Translation
本文描述了提交给UNLP 2026多领域文档理解共享任务的系统。该挑战要求在严格的9小时离线Kaggle执行限制内，从多样化的乌克兰PDF文档中提取精确的答案、文档ID和页码。在对隐藏的私有测试集进行评估时，扫描文档的光学字符识别（OCR）成为了一个严重的瓶颈，由于采用了顺序单线程执行，耗费了总时间预算的5-7小时。这一开销严格限制了剩余时间，使得大型语言模型（LLM）推理的时间大约只有两小时来处理500个问题。为了确保管道在不超时的情况下完成，我们开发了一种资源高效的混合检索增强生成（RAG）管道，利用了BM25、BGE-M3和交叉编码器重排序。我们没有部署参数繁重的推理模型（例如，DeepSeek R1），因为它们总是超时，而是通过在双NVIDIA T4 GPU上使用4位量化的LapaLLM 12B模型（通过llama.cpp）来实现。我们的系统优先考虑管道的稳定性而非多步骤推理，最终取得了0.8095的私有得分，在15个活跃团队中排名第10。
cs.CL / 17 / 2609.22125
Type-Driven Tokenization for Brahmic Scripts
基于类型的布拉米文字标记化
Abstract
Standard tokenizers used in large language models produce malformed text when applied to Brahmic scripts. They are a family of abugidas, writing systems whose consonants carry an inherent vowel that dependent marks can modify. They include Devanagari, Telugu, Tamil, Kannada, and others. The underlying issue is that these tokenizers violate orthographic constraints that do not arise in alphabetic scripts like English. We observe that while English orthography forms a \emph{semigroup} (any two valid tokens can be freely concatenated), Brahmic orthography forms a \emph{partial semigroup}: not every concatenation yields a valid string. We formalise this distinction in Agda, model valid Brahmic tokens as chains in a transition system, and derive a provably correct \texttt{fixToken} function that extends any candidate token to respect orthographic boundaries. We then show how this formal derivation translates into a practical patch for SentencePiece as well as a standalone Rust-based pre-tokenizer library, eliminating the observed errors across Indic scripts.
Chinese Translation
在大型语言模型中使用的标准标记器在应用于布拉米文字时会产生格式错误的文本。布拉米文字是一类音节文字，其辅音带有固有元音，依赖标记可以对其进行修改。这类文字包括天城文、泰卢固文、泰米尔文、卡纳达文等。根本问题在于，这些标记器违反了在英语等字母文字中不存在的正字法约束。我们观察到，英语正字法形成了一个 extit{半群}（任何两个有效标记可以自由连接），而布拉米正字法形成了一个 extit{部分半群}：并非每个连接都能产生有效字符串。我们在Agda中形式化了这一区别，将有效的布拉米标记建模为状态转移系统中的链，并推导出一个可证明正确的 exttt{fixToken}函数，该函数扩展任何候选标记以遵循正字法边界。随后，我们展示了这一形式化推导如何转化为SentencePiece的实用补丁，以及一个独立的基于Rust的预标记器库，从而消除在印度文字中观察到的错误。
cs.CL / 18 / 2609.22127
Beyond Accuracy and Surface Fluency: Risk-Sensitive Evaluation of LLMs for Legal Clause Generation
超越准确性与表面流畅性：对法律条款生成的风险敏感型大型语言模型评估
Abstract
Large language models (LLMs) are increasingly used to draft contractual language, yet conventional accuracy or preference-based evaluations are poorly matched to legal drafting. A clause may be fluent and stylistically polished while still omitting an essential carve-out, allocating risk in an unenforceable way, assuming an inapplicable jurisdiction, or exposing a party to regulatory liability. This paper presents a empirical study design and framework for evaluating LLM-generated contract clauses. The study evaluates four models - Claude Haiku 4.5, Gemini 2.5 Flash Lite, GPT 5.4 Nano, and Qwen 3.5 Flash, across 22 contract clause categories and 34 legally-motivated failure modes. We combine two evaluation frameworks: CLAUSE, which classifies prompts by legal function and failure target, and LENS-CRAFT, which scores outputs across nine legal-quality dimensions. Instead of averaging dimension scores, the study applies a Max Severity Principle so that a single legally decisive defect remains visible. The paper provides the evaluation protocol, taxonomy, analysis plan, and a results structure for reporting empirical findings. We argue that legal AI evaluation should move beyond aggregate accuracy toward clause-specific, failure-mode-driven, and risk-sensitive assessment.
Chinese Translation
大型语言模型（LLMs）在起草合同语言方面的应用日益增加，但传统的准确性或偏好基础评估与法律起草的匹配度较低。一条条款可能在语言流畅和风格上都很精致，但仍可能遗漏重要的例外条款、以不可执行的方式分配风险、假设不适用的管辖权，或使一方面临监管责任。本文提出了一种评估LLM生成的合同条款的实证研究设计和框架。该研究评估了四个模型——Claude Haiku 4.5、Gemini 2.5 Flash Lite、GPT 5.4 Nano和Qwen 3.5 Flash，涵盖22个合同条款类别和34种法律动机下的失败模式。我们结合了两个评估框架：CLAUSE，它根据法律功能和失败目标对提示进行分类；以及LENS-CRAFT，它在九个法律质量维度上对输出进行评分。研究并不简单地对维度得分进行平均，而是应用最大严重性原则，以确保单一法律决定性缺陷保持可见。本文提供了评估协议、分类法、分析计划以及报告实证发现的结果结构。我们认为，法律人工智能评估应超越整体准确性，转向条款特定、以失败模式驱动和风险敏感的评估。
cs.CL / 19 / 2609.22131
Correlation-Aware Structured Pruning for Large Language Models
考虑相关性的结构化剪枝方法用于大型语言模型
Abstract
Structured pruning is a promising approach for reducing the substantial inference costs of Large Language Models (LLMs) while maintaining hardware efficiency. Many existing methods assess the importance of prunable units (e.g., channels or heads) in isolation, implicitly assuming that pruning errors are additive. This independence assumption is often invalidated by the non-orthogonality of model weights and strong correlations between unit activations, potentially leading to performance degradation. To address this, we propose a Correlation-Aware Structured Pruning method. We formulate the pruning objective as a cardinality-constrained binary quadratic program that explicitly models cross-unit dependencies in the reconstruction error. Since this binary quadratic program is NP-hard and difficult to solve exactly, we develop a greedy interaction algorithm based on dependency-aware marginal costs to optimize unit selection. Furthermore, we incorporate a gradient-based strategy to achieve adaptive layer-wise sparsity allocation across the entire model. Extensive experiments on mainstream LLMs demonstrate that incorporating correlation information yields competitive accuracy-efficiency trade-offs compared to representative structured pruning baselines.
Chinese Translation
结构化剪枝是一种有前景的方法，可以在保持硬件效率的同时减少大型语言模型（LLMs）的显著推理成本。许多现有方法孤立地评估可剪枝单元（例如通道或头）的重要性，隐含地假设剪枝误差是可加的。然而，这种独立性假设常常因模型权重的非正交性和单元激活之间的强相关性而失效，可能导致性能下降。为了解决这个问题，我们提出了一种考虑相关性的结构化剪枝方法。我们将剪枝目标公式化为一个基于基数约束的二次二进制规划，明确建模重构误差中的跨单元依赖关系。由于这个二次二进制规划是NP难题且难以精确求解，我们开发了一种基于依赖感知边际成本的贪婪交互算法来优化单元选择。此外，我们还结合了一种基于梯度的策略，以实现整个模型的自适应层级稀疏分配。在主流LLMs上的大量实验表明，结合相关性信息能够在准确性与效率之间实现具有竞争力的权衡，相较于代表性的结构化剪枝基线表现更佳。
cs.CL / 20 / 2609.22133
Observational Equivalence of LLM and Human Annotation
大型语言模型与人工标注的观察等价性
Abstract
In this paper, we show that LLM and human coding are observationally equivalent in terms of annotation quality: recent LLMs agree with expert coders at rates comparable to those observed among experts themselves. We demonstrate this through replications of text-classification tasks from 14 peer-reviewed political science studies, in which ten LLMs, three human experts, and 165 crowdsourced workers independently classify the same texts using identical codebooks. We find that this equivalence is driven by ambiguity in the texts and coding rules. When LLMs disagree with experts, experts are also more likely to disagree with one another, and clarifying coding rules reduces disagreement among both experts and sufficiently capable LLMs. Thus, there is little empirical basis for preferring human coding on the basis of annotation quality alone, while LLMs offer substantial advantages in speed and cost. We therefore argue that the central challenge of text annotation is no longer choosing between human and machine coders, but developing coding rules that minimize ambiguity and accounting for the ambiguity that remains. To this end, we propose using disagreement across LLMs to identify difficult cases and refine codebooks, and we develop ambiguity-aware bounds for downstream inference when a unique annotation cannot be defined for every text.
Chinese Translation
在本文中，我们展示了大型语言模型（LLM）与人工编码在标注质量方面是观察上等价的：近期的LLM与专家编码者的意见一致率与专家之间的意见一致率相当。我们通过对14项经过同行评审的政治科学研究中的文本分类任务进行复制实验来证明这一点，其中十个LLM、三位人类专家和165名众包工作者使用相同的编码手册独立对相同文本进行分类。我们发现，这种等价性是由文本和编码规则中的模糊性驱动的。当LLM与专家意见不一致时，专家之间也更可能存在分歧，而澄清编码规则可以减少专家和足够能力的LLM之间的分歧。因此，仅基于标注质量的考虑，偏好人工编码的实证基础并不充分，而LLM在速度和成本上提供了显著优势。因此，我们认为文本标注的核心挑战不再是选择人类编码者还是机器编码者，而是开发能够最小化模糊性的编码规则，并考虑剩余的模糊性。为此，我们建议利用LLM之间的分歧来识别困难案例并完善编码手册，并在无法为每个文本定义唯一标注时，开发考虑模糊性的下游推断界限。
cs.CL / 21 / 2609.22135
Read-Best Is Not Steer-Best: A Probing--Steering Layer Dissociation in Omni-Modal Large Language Models
阅读最佳并非引导最佳：全模态大型语言模型中的探测-引导层解离
Abstract
Omni-modal large language models integrate text, audio, and image signals into a shared residual stream, where concepts such as emotion can be linearly decoded and causally modified by activation steering. A common but rarely tested assumption is that the layer with the highest probing accuracy is also the best layer for steering, so injection layers are often selected by probe performance. We provide the first causal test of this assumption across three independently developed omni-modal models and find that it fails. Reading and intervention rely on different layers, a phenomenon we call the probing-steering layer dissociation. Using emotion as a controlled testbed, we measure layer-wise readability and steerability across text, audio, and image inputs. Probe-best layers vary widely across architectures, while steering-effective layers consistently fall within a narrow mid-to-late range of normalized depth. Paired random-direction controls show an approximately 26-fold causal gap, ruling out random perturbation and direction quality as explanations. Logit-lens analysis reveals a staged forward process: causal handle, probing saturation, and vocabulary commitment, and motivates a two-factor account in which steering effectiveness depends on both representational readability and downstream plasticity. These results show that probing accuracy is a poor heuristic for selecting intervention layers and suggest a cross-architecture mid-to-late selection criterion. We also identify a cross-modal emotion subspace organized by valence and arousal, with joy acting as a stable anchor across models. Code and data: https://github.com/YiboWang2002/Read-Best-Is-Not-Steer-Best.
Chinese Translation
全模态大型语言模型将文本、音频和图像信号整合到共享的残差流中，其中情感等概念可以通过激活引导线性解码和因果修改。一个常见但很少被测试的假设是，具有最高探测准确性的层也是最佳引导层，因此注入层通常通过探测性能进行选择。我们首次对这一假设进行了因果测试，涵盖了三个独立开发的全模态模型，结果发现该假设不成立。阅读和干预依赖于不同的层，这一现象我们称之为探测-引导层解离。以情感作为受控测试平台，我们测量了文本、音频和图像输入的层级可读性和可引导性。探测最佳层在不同架构中差异很大，而引导有效层则始终位于归一化深度的中后期狭窄范围内。配对随机方向控制显示出大约26倍的因果差距，排除了随机扰动和方向质量作为解释的可能性。Logit-lens分析揭示了一个分阶段的前向过程：因果把握、探测饱和和词汇承诺，并激励了一种双因素解释，其中引导有效性依赖于表征可读性和下游可塑性。这些结果表明，探测准确性是选择干预层的一个不良启发式，并建议采用跨架构的中后期选择标准。我们还识别了一个跨模态情感子空间，该子空间由效价和唤醒组织，快乐在各模型中作为一个稳定的锚点。代码和数据：https://github.com/YiboWang2002/Read-Best-Is-Not-Steer-Best。
cs.CL / 22 / 2609.22136
DiFA: Dual Evidence Fusion and Aggregation for Token-Level Text Anomaly Detection
DiFA：用于标记级文本异常检测的双重证据融合与聚合
Abstract
Text anomaly detection, the task of identifying text instances that deviate from normal language patterns, is crucial for language-driven applications. However, most existing methods can only perform document-level anomaly detection, making it hard to locate harmful phrases or support targeted prevention. Recently, there has been an emerging trend toward token-level text anomaly detection, which aims to address the above limitation by identifying anomalous words or fragments within a document. Nevertheless, one representative method mainly relies on representation-space distance measurement, neglecting the complementary roles of different anomaly cues in capturing diverse abnormal patterns. To bridge the gaps, we propose a Dual-evidence framework with adaptive Fusion and Aggregation (DiFA) for token-level anomaly detection. DiFA derives anomaly scores from form-structural and semantic views to capture visible structural abnormality and contextual inconsistency, respectively, thereby providing complementary evidence for identifying diverse anomalies. To combine these two scores with varying numerical scales, DiFA incorporates a calibration and fusion mechanism to adaptively balance the two views. Moreover, to obtain a discriminative document-level score, a multivariate aggregation method is designed to summarize token-level anomaly scores from multiple perspectives, preventing rare anomalous tokens from being diluted. Extensive experiments across various text anomaly detection benchmarks demonstrate that DiFA consistently achieves top performance while maintaining strong efficiency, robustness, and interpretability. The code and scripts are available at: https://github.com/qyy11-com/DiFA.
Chinese Translation
文本异常检测是识别偏离正常语言模式的文本实例的任务，对于以语言为驱动的应用至关重要。然而，大多数现有方法仅能执行文档级异常检测，难以定位有害短语或支持针对性的预防。最近，标记级文本异常检测逐渐成为一种新趋势，旨在通过识别文档中的异常单词或片段来解决上述限制。然而，一种代表性的方法主要依赖于表示空间距离测量，忽视了不同异常线索在捕捉多样化异常模式中的互补作用。为了解决这些问题，我们提出了一种具有自适应融合与聚合的双重证据框架（DiFA）用于标记级异常检测。DiFA从形式结构和语义视角推导异常分数，以分别捕捉明显的结构异常和上下文不一致，从而为识别多样化异常提供互补证据。为了结合这两个具有不同数值尺度的分数，DiFA引入了一种校准与融合机制，以自适应地平衡这两种视角。此外，为了获得具有区分性的文档级分数，我们设计了一种多变量聚合方法，从多个角度总结标记级异常分数，防止稀有异常标记被稀释。在各种文本异常检测基准上的广泛实验表明，DiFA始终实现了卓越的性能，同时保持强大的效率、鲁棒性和可解释性。代码和脚本可在以下链接获取：https://github.com/qyy11-com/DiFA。
cs.CL / 23 / 2609.22138
Does the Truthfulness Signal Survive Code-Mixing? Probing Hidden States for Hallucination Detection in Hinglish
真实信号在代码混合中是否依然有效？探讨 Hinglish 中的幻觉检测隐状态
Abstract
Hidden-state hallucination probing - training a linear classifier on an LLM's internal activations to detect whether a generated answer is faithful to the input - is an active area of 2026 research, with recent work reporting 0.90-1.00 AUROC across several benchmarks and languages. However, none of this work has tested probes on code-mixed input, despite the fact that a huge population of chatbot users write in Hindi-English code-mixed text ("Hinglish"). We address this gap directly: does a hallucination probe trained on clean-language hidden states transfer to Hinglish, or does the signal degrade under code-mixing? We construct a 5,674-item Hindi/English/Hinglish QA benchmark, generate and label 17,022 model responses across three open-weight 7-8B LLMs (Qwen2.5-7B, Mistral-7B, Llama-3.1-8B), extract per-layer hidden states at two token positions, and train linear and MLP probes for in-distribution detection and cross-lingual transfer. We find that the hallucination signal survives code-mixing well: transfer AUROC ranges from 0.88 to 0.99, with gaps of mostly under 0.05 AUROC relative to in-distribution performance, and that Hindi-trained probes transfer to Hinglish more reliably than English-trained probes. As an independent, practically motivated finding, all three models hallucinate substantially more on Hindi and Hinglish than on English for matched facts. We release our code and synthetic Hinglish QA dataset to support further work on code-mixed hallucination detection.
Chinese Translation
隐状态幻觉探测 - 在大型语言模型（LLM）的内部激活上训练线性分类器，以检测生成的答案是否忠实于输入 - 是2026年研究的一个活跃领域，最近的工作报告在多个基准和语言上达到了0.90-1.00的AUROC。然而，尽管有大量聊天机器人用户使用印地语-英语混合文本（“Hinglish”），但这些工作中没有测试在代码混合输入上的探测。我们直接解决了这一空白：在干净语言的隐状态上训练的幻觉探测是否能转移到Hinglish，还是在代码混合下信号会衰减？我们构建了一个包含5,674个印地语/英语/Hinglish问答的基准，生成并标注了17,022个模型响应，涵盖三个开放权重的7-8B LLM（Qwen2.5-7B, Mistral-7B, Llama-3.1-8B），在两个标记位置提取每层的隐状态，并训练线性和多层感知器（MLP）探测器以进行分布内检测和跨语言转移。我们发现幻觉信号在代码混合中表现良好：转移AUROC范围为0.88到0.99，相对于分布内性能的差距大多低于0.05 AUROC，并且印地语训练的探测器比英语训练的探测器更可靠地转移到Hinglish。作为一个独立的、实际驱动的发现，所有三个模型在匹配事实时在印地语和Hinglish上的幻觉显著多于英语。我们发布了我们的代码和合成的Hinglish问答数据集，以支持进一步的代码混合幻觉检测研究。
cs.CL / 24 / 2609.22143
Using Composition Operators to Linearize LLM Semantic Transformations
使用组合算子线性化大型语言模型的语义转换
Abstract
Machine learning learns functions: prompt to response, image to caption. What these functions are mathematically remains hard to say. We present a method to approximate these kinds of transformations using techniques from dynamical systems that fall under the umbrella of Koopmanism. We introduce the use of composition operators, which generalize the Koopman operator and, crucially, can map between distinct spaces, motivating the perspective that LLM transformations are rectangular infinite-dimensional operators. This formalism reveals useful structure: under natural assumptions on the prompt and response distributions, the LLM operator is an isometry, and misalignment between learned representations manifests as spectral pollution of its finite sections. We then outline a method of constructing finite-dimensional approximations of an LLM operator, and demonstrate how the singular value spectrum can be used to compare tasks and models.
Chinese Translation
机器学习学习函数：从提示到响应，从图像到标题。这些函数在数学上是什么仍然很难说。我们提出了一种使用动力系统技术近似这些转换的方法，这些技术属于库普曼主义（Koopmanism）的范畴。我们引入了组合算子的使用，它们推广了库普曼算子，并且关键的是能够在不同空间之间映射，这一观点促使我们认为大型语言模型（LLM）的转换是矩形的无限维算子。这种形式主义揭示了有用的结构：在对提示和响应分布的自然假设下，LLM算子是一个等距映射，学习到的表示之间的错位表现为其有限部分的谱污染。然后，我们概述了一种构建LLM算子的有限维近似的方法，并演示了如何利用奇异值谱来比较任务和模型。
cs.CL / 25 / 2609.22144
Multilingual Safety Signals Are Multi-Layered: Filtering Safety-Degrading Data for Safer LLMs
多语言安全信号是多层次的：过滤安全降级数据以实现更安全的LLM
Abstract
Preserving safety alignment during large language models fine-tuning is critical, however, recent studies have demonstrated that even benign fine-tuning data may contain safety-degrading samples that silently undermine safety alignment. Existing approaches typically identify such samples using representations from a single safety-sensitive layer. While this assumption has shown effectiveness in monolingual settings, its validity for multilingual models remains unclear due to potential cross-lingual differences in representation patterns. Through a cross-lingual analysis, we show that sensitive layers are only partially shared across languages, with safety-relevant signals often distributed across multiple layers. Motivated by these observations, we propose MMSAFE, a multi-layer framework for multilingual safety-degrading data identification that captures both shared and language-specific safety signals. Extensive experiments across multiple models, languages, and safety benchmarks demonstrate that MMSAFE reduces the average harmful-response ratio by 60% compared with random filtering and achieves stronger average performance than the strongest single-layer baseline, demonstrating the effectiveness of multi-layer modeling for robust multilingual safety alignment.
Chinese Translation
在大型语言模型微调过程中保持安全对齐至关重要，然而，最近的研究表明，即使是良性的微调数据也可能包含安全降级样本，这些样本会悄然破坏安全对齐。现有的方法通常使用来自单一安全敏感层的表示来识别此类样本。虽然这一假设在单语环境中已显示出有效性，但由于潜在的跨语言表示模式差异，其在多语言模型中的有效性仍不明确。通过跨语言分析，我们表明敏感层在不同语言之间仅部分共享，安全相关信号通常分布在多个层次上。基于这些观察，我们提出了MMSAFE，一个用于多语言安全降级数据识别的多层框架，能够捕捉共享和语言特定的安全信号。在多个模型、语言和安全基准上的广泛实验表明，与随机过滤相比，MMSAFE将平均有害响应比降低了60%，并且在平均性能上超过了最强的单层基线，证明了多层建模在稳健的多语言安全对齐中的有效性。
cs.CL / 26 / 2609.22149
Beyond the Stitching Assumption: A Unified Framework for Multimodal Synthetic Data Evaluation via Semantic Quantization
超越拼接假设：通过语义量化实现多模态合成数据评估的统一框架
Abstract
Multimodal synthetic datasets combine structured attributes with free text, but are often evaluated separately. Such metrics can remain high after tabular--text pairings are disrupted. We present a projection-based evaluator for tabular--text synthetic data. A fixed sentence encoder maps text to embeddings, \(k\)-means converts them to cluster states, and tabular variables are represented as categorical or quantile-binned states. Real and synthetic contingency tables are compared using Jensen--Shannon divergence (JSD), normalized mutual information (NMI), conditional JSD (cJSD), and joint-state entropy. We also report text-to-attribute (T2A) utility and a holdout-calibrated proximity flag rate (PFR) as a representation-level diagnostic. A text-permutation control preserves both marginal distributions while disrupting their pairing. Experiments on Amazon Reviews, Kiva Loans, and the Employment Scam Aegean Dataset show that modality-specific scores remain high under this control. The projection diagnostics detect disruption when real projected dependence exceeds a permutation baseline, but are less informative for weak or sparse projections. Some conditioned LLM baselines also exhibit stronger measured dependence than the corresponding real-data projections. These results support explicit cross-modal evaluation with permutation baselines and coverage reporting.
Chinese Translation
多模态合成数据集结合了结构化属性和自由文本，但通常是分开评估的。这种评估指标在表格-文本配对被破坏后仍然可能保持较高水平。我们提出了一种基于投影的表格-文本合成数据评估器。固定的句子编码器将文本映射为嵌入，$k$-均值算法将其转换为聚类状态，表格变量则表示为分类或分位数分箱状态。使用詹森-香农散度（JSD）、归一化互信息（NMI）、条件JSD（cJSD）和联合状态熵比较真实和合成的列联表。我们还报告了文本到属性（T2A）效用和经过保留校准的接近标志率（PFR）作为表示层面的诊断。文本置换控制在破坏配对的同时保留了边际分布。对亚马逊评论、Kiva贷款和就业诈骗爱琴海数据集的实验表明，在这种控制下，模态特定的评分仍然保持较高水平。投影诊断在真实投影依赖超过置换基线时检测到破坏，但对于弱或稀疏投影的指示性较差。一些条件的LLM基线也表现出比相应的真实数据投影更强的测量依赖性。这些结果支持使用置换基线和覆盖率报告进行明确的跨模态评估。
cs.CL / 27 / 2609.22151
Do Language Models Know Their Own Constraints?
语言模型是否知道自身的限制？
Abstract
We ask whether behavioral constraints acquired through post training remain explicitly reportable. Using constrained recipe generation as a testbed, five banned ingredients enforced via LoRA fine tuning of Llama 3.1 8B Instruct we compare supervised fine tuning (SFT) and Group Relative Policy Optimization (GRPO) against an untrained baseline on a four tier Constraint Awareness Benchmark. Averaged over three seeds, both methods raise behavioral compliance from 4% to about 90% while reducing explicit constraint reporting below the untrained model (0.48/5 to 0.16/5 for SFT, 0.07/5 for GRPO) and eroding retained third person knowledge (93% to 36% for SFT, 14% for GRPO; p less than 0.01 between methods). Contrary to our initial hypothesis, the reward based signal is the more destructive of the two: a reward that penalizes banned ingredient tokens regardless of framing learns a context independent suppression rather than a self directed constraint. A context conditioned reward designed to teach the self to other distinction fails, collapsing toward inclusion in both framings. Probing prompt time hidden states recovers per ingredient avoidance at 83.8% (layer 24 MLP), but only 6.4 points above a per ingredient base rate predictor (77.4%), and the model's own verbal self report is more accurate still (87.8%). A positive control adding explicit self description examples does not restore reporting. The failure is therefore specific to enumerating constraints on request, not a general loss of access to them.
Chinese Translation
我们探讨通过后训练获得的行为限制是否仍然可以明确报告。以受限食谱生成作为测试平台，我们对通过 LoRA 微调 Llama 3.1 8B Instruct 强制执行的五种禁用成分进行了比较，比较了监督微调（SFT）和组相对策略优化（GRPO）与未训练基线在四级约束意识基准上的表现。在三个种子的平均结果中，这两种方法将行为合规性从 4% 提高到约 90%，同时将明确的约束报告降低到低于未训练模型（SFT 从 0.48/5 降至 0.16/5，GRPO 为 0.07/5），并削弱了保留的第三人称知识（SFT 从 93% 降至 36%，GRPO 为 14%；两种方法之间的 p 值小于 0.01）。与我们最初的假设相反，基于奖励的信号是两者中更具破坏性的：一种惩罚禁用成分标记的奖励，无论框架如何，学习到的是一种与上下文无关的抑制，而不是自我导向的约束。旨在教导自我与他人区分的上下文条件奖励未能实现，向两个框架中的包含倾斜。探测提示时间的隐藏状态在每种成分的回避率上恢复到 83.8%（第 24 层 MLP），但仅比每种成分的基线预测器（77.4%）高出 6.4 个点，而模型自身的口头自我报告更为准确（87.8%）。添加明确自我描述示例的正控制并未恢复报告。因此，这一失败特定于请求时列举约束，而不是对它们的普遍访问丧失。
cs.CL / 28 / 2609.22152
Is Imagination Derived from Hallucination? A Cross-Taxonomy Evaluation of Imagination and Hallucination in Large Language Models
想象力源自幻觉吗？大型语言模型中想象力与幻觉的跨分类评估
Abstract
Imagination performs as a high-level function of large language models (LLMs) which determines the potential of how an LLM creates unseen or creative content. While existing works have built a rich family of creativity benchmarks for this ability, they only measure how far an output departs from common answers and never check whether the departure is licensed by the prompt. Moreover, hallucination, the closest neighbor of imagination, is always measured in a separate pipeline on different generations, so the influential claim that imagination and hallucination stem from the same generative mechanism has never been directly testable. In this paper, we propose Whiteboard, the first LLM imagination evaluation benchmark. Its design follows the authoritative cognitive instruments developed to measure human imagination: seven mechanism-grounded imagination subtypes are adapted from classic paradigms, then crossed with ten support-boundary hallucination subtypes and scored jointly on the same generation. Different from previous creativity or hallucination benchmarks, Whiteboard gates every imagination score with an explicit support check and computes both axes deterministically through an auditable atom matrix, with no LLM judge on the primary path. The full Whiteboard item bank contains 1,660 prompts; on its shared 80-item anchor set, we evaluate 79 state-of-the-art LLMs and validate the instrument against 13,280 human judgments. Additionally, we further explore whether imagination derives from the same generative tendency as hallucination and what key factors shape it. Our analysis indicates a counterintuitive correlation between hallucination and imagination: Most of the subtype couplings are negative, every one of the anchor items reproduces the negative coupling on its own.
Chinese Translation
想象力作为大型语言模型（LLMs）的高级功能，决定了LLM创造未见或创造性内容的潜力。尽管现有研究为这一能力构建了丰富的创造力基准，但它们仅测量输出与常见答案的偏离程度，而从未检查这种偏离是否得到提示的授权。此外，幻觉作为想象力的最接近邻居，总是在不同生成的单独流程中进行测量，因此，想象力和幻觉源于相同生成机制的影响性主张从未得到直接验证。在本文中，我们提出了Whiteboard，这是第一个LLM想象力评估基准。其设计遵循了用于测量人类想象力的权威认知工具：七种基于机制的想象子类型从经典范式中调整而来，然后与十种支持边界的幻觉子类型交叉，并在同一生成上共同评分。与之前的创造力或幻觉基准不同，Whiteboard通过明确的支持检查对每个想象分数进行限制，并通过可审计的原子矩阵确定性地计算两个维度，主要路径上没有LLM评审。完整的Whiteboard项目库包含1,660个提示；在其共享的80项锚点集上，我们评估了79个最先进的LLMs，并对该工具进行了13,280个人工判断的验证。此外，我们进一步探讨了想象力是否源于与幻觉相同的生成倾向，以及哪些关键因素塑造了它。我们的分析表明，幻觉与想象力之间存在反直觉的相关性：大多数子类型耦合为负，每个锚点项目在其自身上重现了负耦合。
cs.CL / 29 / 2609.22162
Beyond Raw Context Transfer: Representation-based Federated Retrieval-Augmented Generation
超越原始上下文转移：基于表示的联邦检索增强生成
Abstract
Retrieval-augmented generation (RAG) improves the factuality of large language models (LLMs) and vision-language models (VLMs) by grounding generation in external knowledge. However, most existing RAG frameworks assume a centralized retrieval corpus, which is often impractical in sensitive domains such as healthcare, where data are inherently distributed and raw content cannot be directly shared across institutions. Recent efforts on decentralized RAG primarily follow prompt-based paradigms that exchange raw, human-readable retrieved content, leading to substantial inference-time computational overhead and direct exposure of retrieved information. To address these limitations, we propose Representation-based Federated RAG (FedRepRAG), a decentralized RAG framework that keeps raw documents at their owning clients and exchanges only compact latent representations during cross-client retrieval. To integrate retrieved knowledge, we introduce a collaboratively trained projector that converts retrieval embeddings into generator-compatible representation tokens for a frozen LLM/VLM backbone. Experiments across decentralized visual question answering (VQA) and question answering (QA) benchmarks show that FedRepRAG consistently outperforms direct inference and local retrieval baselines while substantially reducing retrieval-context length and inference-time computational overhead compared with raw-context transfer. Further analyses confirm the importance of query-relevant retrieved representations and characterize the residual representation-level leakage associated with representation exchange. Overall, FedRepRAG provides an effective and efficient framework for federated RAG without transferring raw retrieved content.
Chinese Translation
检索增强生成（RAG）通过将生成过程与外部知识相结合，提高了大型语言模型（LLMs）和视觉-语言模型（VLMs）的事实性。然而，现有的大多数RAG框架假设存在一个集中式的检索语料库，这在医疗等敏感领域往往不切实际，因为数据本质上是分布式的，原始内容无法在机构之间直接共享。最近在去中心化RAG方面的努力主要遵循基于提示的范式，交换原始的人类可读检索内容，导致显著的推理时计算开销和检索信息的直接暴露。为了解决这些局限性，我们提出了基于表示的联邦RAG（FedRepRAG），这是一个去中心化的RAG框架，保持原始文档在其拥有的客户端，并在跨客户端检索时仅交换紧凑的潜在表示。为了整合检索知识，我们引入了一个协同训练的投影器，将检索嵌入转换为与冻结的LLM/VLM主干兼容的表示令牌。针对去中心化视觉问答（VQA）和问答（QA）基准的实验表明，FedRepRAG在直接推理和本地检索基线之上始终表现优越，同时与原始上下文转移相比，显著减少了检索上下文长度和推理时计算开销。进一步的分析确认了与查询相关的检索表示的重要性，并描述了与表示交换相关的残余表示级泄漏。总体而言，FedRepRAG为联邦RAG提供了一个有效且高效的框架，而无需转移原始检索内容。
cs.CL / 30 / 2609.22163
MechaTerp-TRACE: A Novel Approach for Component Ablation Analysis in Language Models
MechaTerp-TRACE：一种用于语言模型组件消融分析的新方法
Abstract
Interpretability research on large language models has produced accounts of factual recall in feed-forward layers and of token relationships in self-attention, but little work offers a unified way to compare the causal contribution of different architecture components to a model's output. We introduce MechaTerp (the Mechanistic Interpretability suite) -TRACE (subset for Teacher-forced Registry of Ablated Component Effects), an architecture and study that measures how much each registered component of a language model supports the production of a named entity. TRACE ablates one component at a time and measures the resulting change in the output distribution at a fixed answer token, so component types from whole transformer blocks down to individual neurons and output logits can be compared on a common scale. We apply it to thirteen instruction-tuned dense decoder models spanning five families and one to thirty billion parameters, ablating 49,656 components across 48 medical and 42 general-knowledge prompts. We find that the components carrying the most effect are the same few, positionally fixed components in every model, regardless of which entity a prompt asks about, and that once these are removed, the remaining support is close to evenly spread in eleven of the thirteen models. Apparent localisation of entity knowledge is therefore largely attributable to generic generation machinery, which has direct consequences for methods that assume entity knowledge sits in a findable place, including targeted knowledge editing.
Chinese Translation
对大型语言模型的可解释性研究已经揭示了前馈层中的事实回忆和自注意力中的标记关系，但很少有研究提供统一的方法来比较不同架构组件对模型输出的因果贡献。我们介绍了MechaTerp（机械可解释性套件）-TRACE（强制教师消融组件效应注册子集），这是一个架构和研究，旨在测量语言模型中每个注册组件对命名实体生成的支持程度。TRACE一次消融一个组件，并测量在固定答案标记下输出分布的变化，因此可以在一个共同的尺度上比较从整个变换器块到单个神经元和输出logits的组件类型。我们将其应用于涵盖五个家族、参数范围从十亿到三百亿的十三个指令调优的密集解码器模型，消融了48个医学和42个常识提示中的49,656个组件。我们发现，影响最大的组件在每个模型中都是同一小部分位置固定的组件，无论提示询问的是哪个实体，一旦这些组件被移除，剩余的支持在十三个模型中的十一种几乎均匀分布。因此，实体知识的明显定位在很大程度上归因于通用生成机制，这对假设实体知识位于可查找位置的方法（包括针对性知识编辑）具有直接影响。
cs.CL / 31 / 2609.22164
A Multi-Agent Pipeline for Source-Grounded Synthetic Note Generation from Longitudinal Structured EHR
基于源的多智能体管道用于从纵向结构化电子健康记录生成合成笔记
Abstract
Structured EHR is abundant but sparse, coded, and difficult to use directly for note-centric clinical modeling. We present MedNotes, a multi-agent synthetic data generation pipeline that converts longitudinal structured EHR into source-grounded clinical note representations under explicit quality control. MedNotes treats structured-data-to-text synthesis as a closed-loop agentic process: a generator proposes a note, evaluator agents diagnose factual, coverage, structural, and hallucination-related failures, and a router accepts, revises, or rejects the draft. On 1,485 EHRSHOT encounters, MedNotes achieves a 91.4% pass rate, with mean factual accuracy of 0.980, completeness of 99.1%, structural fidelity of 0.761, and 0.028 critical hallucinations per encounter. Iterative refinement improves acceptance from 69.4% to 91.4%. The resulting synthetic corpus improves downstream CPT prediction and paragraph-level section prediction when combined with limited real data.
Chinese Translation
结构化电子健康记录（EHR）丰富但稀疏、编码且难以直接用于以笔记为中心的临床建模。我们提出了MedNotes，一个多智能体合成数据生成管道，能够在明确的质量控制下将纵向结构化EHR转换为基于源的临床笔记表示。MedNotes将结构化数据到文本的合成视为一个闭环的智能过程：生成器提出笔记，评估代理诊断事实、覆盖、结构和幻觉相关的失败，而路由器接受、修订或拒绝草稿。在1,485个EHRSHOT案例中，MedNotes实现了91.4%的通过率，平均事实准确率为0.980，完整性为99.1%，结构保真度为0.761，每个案例的关键幻觉为0.028。迭代改进将接受率从69.4%提高到91.4%。生成的合成语料库在与有限的真实数据结合时，提高了下游CPT预测和段落级别部分预测的效果。
cs.CL / 32 / 2609.22169
Monocultural Biases: Correlated biases in large language models lead to unequal systemic exclusion rates in hiring
单文化偏见：大型语言模型中的相关偏见导致招聘中的系统性不平等待遇
Abstract
Employers are increasingly using large language models (LLMs) to automate their hiring process. This paper investigates the risk of monocultural biases, in which the widespread deployment of large language models homogenizes biases across the labor market, leading to greater systemic exclusion for certain demographic groups. For ten LLMs, we measure hiring biases across their base and post-trained versions to identify which stage, pre-training or post-training, lead to monocultural biases. We find that, compared to their base models, post-trained models are 3.6% less likely to callback older applicants. This negative shift occurs in eight of the ten models that we evaluate. Post-trained models have much more correlated decisions than base models which is likely driven by human capital traits like skills or college major. However, greater consensus among models increases global systemic exclusion rates from 5.6% to 17.3% and exacerbates demographic inequalities, with intersectional systemic exclusion rates ranging from 12.2% to 21.7% for post-trained models. We find that this inequality is primarily driven by age-based discrimination that is exacerbated in post-training. These results indicate that while post-training techniques may improve models' abilities to select the best applicants, they may raise systemic inequality risks for those at the margin by uniformly introducing new biases.
Chinese Translation
雇主越来越多地使用大型语言模型（LLMs）来自动化招聘过程。本文探讨了单文化偏见的风险，即大型语言模型的广泛应用使得劳动市场中的偏见趋于同质化，从而导致某些人口群体遭受更大的系统性排斥。我们对十个LLM进行了研究，测量了它们在基础版本和后训练版本中的招聘偏见，以识别是预训练阶段还是后训练阶段导致了单文化偏见。我们发现，与基础模型相比，后训练模型对年长申请者的回调率降低了3.6%。这一负面变化出现在我们评估的十个模型中的八个。后训练模型的决策相关性远高于基础模型，这可能是由人力资本特征如技能或大学专业驱动的。然而，模型之间更高的一致性使得全球系统性排斥率从5.6%上升至17.3%，并加剧了人口统计学的不平等，后训练模型的交叉系统性排斥率在12.2%至21.7%之间。我们发现，这种不平等主要是由年龄歧视驱动的，并在后训练中加剧。这些结果表明，尽管后训练技术可能提高模型选择最佳申请者的能力，但它们可能通过统一引入新的偏见而增加边缘群体的系统性不平等风险。
cs.CL / 33 / 2609.22171
Quantifying Hidden Salt for Precision Healthcare: Sodium Assessment via Joint-Factor Retrieval and Chain-of-Thought Inference
量化隐含盐分以实现精准医疗：通过联合因子检索和思维链推理评估钠含量
Abstract
Precision healthcare, particularly for conditions like hypertension and cardiovascular disease, necessitates monitoring of dietary sodium intake. However, tracking this is hindered by the prevalence of hidden salt in cooking, such as sodium in soy sauce and ketchup. While recipes offer a valuable data source for dietary analysis, sodium-rich seasonings are frequently omitted or described ambiguously in instructions. To solve this issue, we propose SALT, a Sodium Assessing & Level Tracking framework adopting an RAG framework to assess sodium content in recipes. Our framework first introduces a Joint-Factor Embedding Retrieval module to locate similar recipes with specified sodium content for addressing the lack of contextual references. These retrieved samples provide contexts for subsequent inference. Then we design a structured 4-hop Chain-of-Thought inference module to refine the vague estimation from language models through a multi-step sodium estimation. To facilitate our study, we further construct a recipe dataset SALT54k with $54,151$ entries labeled with sodium quantities across $11$ common seasonings. Results on SALT54k demonstrate that our method achieves state-of-the-art performance in sodium estimation. Additional real-world validations confirm the effectiveness of our method, demonstrating its potential as a practical solution for AI-assisted precision healthcare.
Chinese Translation
精准医疗，特别是针对高血压和心血管疾病等状况，需要监测饮食中的钠摄入。然而，由于烹饪中隐含盐的普遍存在，例如酱油和番茄酱中的钠，追踪这一点变得困难。尽管食谱为饮食分析提供了宝贵的数据来源，但钠含量丰富的调味品在说明中常常被省略或模糊描述。为了解决这一问题，我们提出了SALT（钠评估与水平追踪）框架，该框架采用RAG（检索增强生成）方法来评估食谱中的钠含量。我们的框架首先引入了联合因子嵌入检索模块，以定位具有特定钠含量的相似食谱，从而解决上下文参考缺乏的问题。这些检索到的样本为后续推理提供了上下文。然后，我们设计了一个结构化的4步思维链推理模块，通过多步骤的钠估计来细化语言模型中的模糊估计。为了促进我们的研究，我们进一步构建了一个名为SALT54k的食谱数据集，包含$54,151$条标注了$11$种常见调味品钠含量的条目。在SALT54k上的结果表明，我们的方法在钠估计方面达到了最先进的性能。额外的现实世界验证确认了我们方法的有效性，展示了其作为AI辅助精准医疗的实用解决方案的潜力。
cs.CL / 34 / 2609.22174
SCoR: A Hierarchical Framework for Forecasting Relations Between Scientific Concepts
SCoR：一个用于预测科学概念之间关系的分层框架
Abstract
Anticipating emerging research directions is a critical goal of AI-assisted science. Existing methods mainly predict which concepts will co-occur in future papers, but co-occurrence captures shared attention rather than the scientific meaning of a connection, such as whether one method uses, combines, replaces, or contradicts another. We formulate research-direction discovery as hierarchical scientific-relation forecasting over a shared candidate-pair space, comprising three temporally aligned tasks: first co-occurrence, first scientific-relation formation, and relation type at formation. We construct SCoR-Graph from 187,848 cs.CV papers published between 2017 and 2026, yielding 270,687 consolidated concepts, 7.45 million co-occurrence edges, and 615,036 typed, directed relation edges. From cutoff-specific graph snapshots, we derive SCoR-Bench, a leakage-audited benchmark for these three capabilities, with expert-verified gold labels for the entire relation-type test set. We further introduce HiSCoR, a task-adapted model family that models relation emergence as a temporally evolving, hierarchically constrained process by encoding pre-cutoff event histories and conditioning relation formation on future co-occurrence. On the held-out 2025-2026 window, HiSCoR achieves an AUROC of 0.9515, a 2.4% relative improvement over the strongest temporal-graph baseline, and improves population-AUPRC by 14.0%; its relation-type variant achieves a Macro-AUROC of 0.7795. Ablations show that semantic, co-occurrence, and typed-relation views provide complementary predictive evidence. SCoR advances research-direction forecasting from predicting which concepts will co-occur to anticipating whether and how evidence-backed scientific relations will emerge.
Chinese Translation
预测新兴研究方向是人工智能辅助科学的一个关键目标。现有方法主要预测哪些概念将在未来的论文中共同出现，但共同出现仅反映了共享关注，而非连接的科学意义，例如一种方法是否使用、结合、替代或与另一种方法相矛盾。我们将研究方向发现形式化为在共享候选对空间上进行的分层科学关系预测，包含三个时间对齐的任务：首次共同出现、首次科学关系形成和关系形成时的类型。我们从2017年至2026年间发表的187,848篇计算机视觉（cs.CV）论文中构建了SCoR-Graph，生成了270,687个整合概念、745万个共同出现边和615,036个类型化的有向关系边。通过特定截止点的图快照，我们推导出SCoR-Bench，这是一个经过泄漏审计的基准，涵盖这三种能力，并为整个关系类型测试集提供了专家验证的金标准标签。我们进一步引入了HiSCoR，一个任务适应的模型家族，通过编码截止前事件历史并将关系形成条件化于未来共同出现，将关系的出现建模为一个时间演变的、分层约束的过程。在2025-2026年保留的窗口上，HiSCoR实现了0.9515的AUROC，相较于最强时间图基线提高了2.4%的相对改进，并将人口AUPRC提高了14.0%；其关系类型变体实现了0.7795的宏观AUROC。消融实验表明，语义、共同出现和类型化关系视角提供了互补的预测证据。SCoR将研究方向预测从预测哪些概念将共同出现提升到预测是否以及如何基于证据的科学关系将会出现。
cs.CL / 35 / 2609.22188
Fairness Beyond Anonymization? Demographic Leakage in German LLM-Generated Resumes
超越匿名化的公平性？德国LLM生成简历中的人口统计信息泄露
Abstract
Large language models (LLMs) are increasingly integrated into AI-assisted hiring pipelines, including automated resume generation and screening. Under the EU AI Act, the hiring domain is classified as high-risk, making fairness and transparency critical requirements. Existing work has primarily focused on explicit hiring decisions, while less attention has been paid to whether generated resumes themselves encode recoverable demographic information. In this work, we conduct a two-stage audit of demographic leakage in German-language LLM-generated resumes. First, we use ChatGPT (GPT-4o-mini), Gemini 2.5 Flash-Lite, and multiple scales of the open-weight Qwen 3 model family (4B, 8B, and 14B) to generate resumes from real anonymized job-matching profiles, systematically varying gender- and ethnicity-associated names while holding qualifications constant. Second, we simulate a downstream resume screening scenario, where the generated resumes are first anonymized and gender-neutralized, before demographic leakage classifiers are trained on the resulting texts. We find that, despite these interventions, classifiers reliably distinguish between resumes generated with male and female names. This leakage is not driven by overtly gendered wording, but by subtle differences in the usage of semantically equivalent, formally gender-neutral terms in German. In contrast, ethnicity-related leakage remains comparatively weak across models. Our findings demonstrate that apparently neutral resume generation can still preserve highly predictive demographic signals, raising concerns about anonymization-based fairness interventions in multilingual AI hiring pipelines.
Chinese Translation
大型语言模型（LLMs）正日益融入AI辅助招聘流程，包括自动简历生成和筛选。在欧盟人工智能法案下，招聘领域被归类为高风险领域，使得公平性和透明度成为关键要求。现有研究主要集中在明确的招聘决策上，而对生成的简历是否编码可恢复的人口统计信息关注较少。在本研究中，我们对德语LLM生成的简历进行两阶段审计，以评估人口统计信息泄露。首先，我们使用ChatGPT（GPT-4o-mini）、Gemini 2.5 Flash-Lite以及多个开放权重的Qwen 3模型系列（4B、8B和14B）从真实的匿名职位匹配档案中生成简历，系统性地变换与性别和种族相关的名字，同时保持资格不变。其次，我们模拟一个下游简历筛选场景，在该场景中，生成的简历首先被匿名化和性别中立化，然后在生成的文本上训练人口统计信息泄露分类器。我们发现，尽管进行了这些干预，分类器仍能可靠地区分使用男性和女性名字生成的简历。这种泄露并不是由明显的性别化措辞驱动的，而是由德语中语义等价的、形式上性别中立的术语使用的微妙差异所导致。相比之下，与种族相关的泄露在各模型中仍然相对较弱。我们的研究结果表明，表面上中立的简历生成仍然可以保留高度预测的人口统计信号，这引发了对多语言AI招聘流程中基于匿名化的公平性干预的担忧。
cs.CL / 36 / 2609.22195
The Situated Identity Test: Distinguishing Persistent Cognitive Identity from Persona Imitation
情境身份测试：区分持久认知身份与角色模仿
Abstract
Large language models can convincingly adopt personas, recall past dialogues, and weave rich autobiographies. Yet this conversational eloquence conceals a fundamental attribution problem: looking the part does not mean having lived the life. Two individuals can share identical public profiles--the same age, hometown, occupation, and personality traits--while possessing entirely distinct private histories, relationships, and acquired skills. When conditioned solely on that shared profile, an agent lacks the information required to determine which lineage is correct. We introduce the Situated Identity Test (SIT), an architecture-independent framework that evaluates whether an agent's behavior is functionally attributable to a specific developmental lineage. Grounded identity requires both appropriate knowledge of recorded experiences and appropriate ignorance of ungrounded ones, bounded by what the identity has actually acquired rather than what its underlying foundation model knows. We prove that any policy conditioned solely on a compressed profile is bounded by an average situated validity of at most 1/m across m colliding life histories on lineage-discriminative queries (at most 50% for paired lineages). We instantiate this framework in SITBench, an evaluation suite designed for 25 profile-collision pairs (50 distinct lineages) across 10,000 planned probes and nine architectural configurations. Supported by an open-source reference implementation, deterministic test fixtures, and empirical pilot evaluations on frontier foundation models (GPT-5.6 Sol and Claude Opus 5), we formalize the failure modes of persona prompting under profile collision and provide an assurance harness for evaluating episodic continuity, structured state, and epistemic boundaries.
Chinese Translation
大型语言模型能够令人信服地采用角色，回忆过去的对话，并编织丰富的自传。然而，这种对话的流畅性掩盖了一个根本的归因问题：看起来像是某个角色并不意味着经历过相应的生活。两个个体可以共享相同的公共档案——相同的年龄、家乡、职业和性格特征——但却拥有完全不同的私人历史、关系和获得的技能。当仅基于共享档案进行条件化时，代理缺乏判断哪个血统正确所需的信息。我们提出了情境身份测试（Situated Identity Test, SIT），这是一个与架构无关的框架，用于评估代理的行为是否可以功能性地归因于特定的发展血统。扎根身份要求对记录的经历有适当的知识，同时对未扎根的经历保持适当的无知，这一切都受到身份实际获得的知识的限制，而不是其基础模型所知的内容。我们证明，任何仅基于压缩档案进行条件化的策略，其在血统区分查询上的平均情境有效性最多为1/m（m为冲突生活历史的数量，配对血统的情况下最多为50%）。我们在SITBench中实例化了这一框架，这是一个评估套件，设计用于25对档案冲突（50个不同的血统），涵盖10,000个计划探测和九种架构配置。在开放源代码的参考实现、确定性测试装置和对前沿基础模型（如GPT-5.6 Sol和Claude Opus 5）的实证试点评估的支持下，我们正式化了在档案冲突下角色提示的失败模式，并提供了一个评估情节连续性、结构状态和认知边界的保障工具。
cs.CL / 37 / 2609.22198
The Role of AI in Online Reviews
人工智能在在线评论中的作用
Abstract
The rapid adoption of large language models (LLMs) creates new opportunities for strategic content generation on online platforms, including potentially harmful forms of manipulation that may undermine platform effectiveness and reshape platform dynamics. However, measuring such activity is difficult because AI-generated content is rarely directly observable. We introduce an empirical approach that leverages discrete LLM supply shocks - abrupt changes in model prices and capabilities, and contrasts verified with non-verified reviews to identify changes in platform activity associated with generative AI supply improvements. We apply this approach to more than 13 million reviews from Trustpilot, one of the leading online platforms for business reviews. A robust finding is that following LLM supply shocks, unverified reviews shift toward greater negativity: more 1-stars, fewer 5-stars, and lower ratings, with effects driven primarily by new model releases and concentrated among firms with the lowest and highest review volumes, suggesting that strategic AI use may reshape platform competition dynamics. We further find that LLM supply shocks trigger short, concentrated bursts of review activity. Together, these findings suggest that generative AI is already reshaping how reputation and competition operate on online platforms.
Chinese Translation
大型语言模型（LLMs）的快速采用为在线平台上的战略内容生成创造了新的机会，包括可能会破坏平台有效性和重塑平台动态的有害操控形式。然而，由于人工智能生成的内容很少可以直接观察，因此衡量这种活动是困难的。我们提出了一种实证方法，利用离散的LLM供应冲击——模型价格和能力的突然变化，并对比经过验证的评论与未验证的评论，以识别与生成性人工智能供应改善相关的平台活动变化。我们将这种方法应用于来自Trustpilot的超过1300万条评论，Trustpilot是领先的商业评论在线平台之一。一个显著的发现是，在LLM供应冲击后，未验证的评论趋向于更负面的倾向：1星评论增多，5星评论减少，评分降低，这一效应主要由新模型发布驱动，并集中在评论量最低和最高的公司之间，这表明战略性人工智能使用可能重塑平台竞争动态。我们进一步发现，LLM供应冲击会引发短暂且集中的评论活动。综合来看，这些发现表明，生成性人工智能已经在重塑在线平台上声誉和竞争的运作方式。
cs.CL / 38 / 2609.22200
PII-TRACE: A Benchmark for Context-Aware PII Detection in Multi-Turn LLM Conversations
PII-TRACE：多轮大型语言模型对话中上下文感知的个人可识别信息检测基准
Abstract
LLM assistants and agentic systems log long multi-turn conversations. AI providers often scan these conversations for Personally Identifiable Information (PII) and mask the PII before storing or processing conversation data. Yet most PII detectors and benchmarks target self-contained records rather than cross-turn evaluation. To evaluate PII detection across turns in multi-turn conversations, we introduce PII-TRACE (Tracing Recurring PII Across Conversational Exchanges), to our knowledge the first PII benchmark to assess whether detectors identify PII in conversational contexts and cover every mention of a recurring identifier across turns. PII-TRACE contains 13,148 synthetic multi-turn dialogues in 13 languages with character-level spans and identifier clusters. Across eleven baselines, including frontier LLMs, no detector achieves full entity-level coverage without substantial false positives on PII-free conversations, and single-pass reading loses a third of the gold characters on long dialogues. To close this gap, we introduce PII-Tracer, a compact 0.6B-parameter detector trained with conversation-level supervision. PII-Tracer attains the highest entity-level coverage of any system we evaluate and also performs strongly on standard single-record benchmarks.
Chinese Translation
大型语言模型（LLM）助手和智能系统记录了长时间的多轮对话。人工智能提供商通常会扫描这些对话以寻找个人可识别信息（PII），并在存储或处理对话数据之前对其进行屏蔽。然而，大多数PII检测器和基准针对的是自包含的记录，而非跨轮次的评估。为了评估多轮对话中跨轮次的PII检测，我们引入了PII-TRACE（Tracing Recurring PII Across Conversational Exchanges），这是我们所知的第一个评估检测器在对话上下文中识别PII的基准，并覆盖每次提及的重复标识符。PII-TRACE包含13,148个合成的多轮对话，涵盖13种语言，具有字符级跨度和标识符集群。在包括前沿LLM在内的十一种基线中，没有任何检测器在没有大量假阳性结果的情况下实现完整的实体级覆盖，而单次读取在长对话中损失了三分之一的黄金字符。为了填补这一空白，我们引入了PII-Tracer，一个经过对话级监督训练的紧凑型0.6B参数检测器。PII-Tracer在我们评估的任何系统中实现了最高的实体级覆盖，并且在标准单记录基准上也表现出色。
cs.CL / 39 / 2609.22204
Evaluating Personal Information Output from Conversational Interactions in Generative AI Systems
评估生成性人工智能系统中对话互动产生的个人信息输出
Abstract
This exploratory pilot study evaluates the scope and perceived accuracy of personal information output from ongoing conversational interactions in generative AI systems using GPT-5.2 Instant and GPT-5.2 Thinking, categorized into three output types: Fact, Inference, and Confidence. Based on the evaluation results obtained from 15 Japanese participants, differences in model design have limited impact on personal information output tendencies. Compared with the Inference type, the Fact type shows a more conservative output pattern. Regarding attribute categories, the findings indicate that Core Personal attributes associated with identification are treated relatively conservatively, whereas Behavioral and Linguistic attributes show higher accuracy across both Fact and Inference outputs. Furthermore, Holistic Profile, Psychological and Cognitive, and Residual attributes are more readily inferred, even when not supported by explicit factual outputs. Notably, the lack of null outputs for these attributes in the Inference type suggests that such inferred profiles may be constructed from indirectly available contextual information. The findings may contribute to future discussions regarding privacy awareness and personal information inference in generative AI systems.
Chinese Translation
本探索性试点研究评估了使用GPT-5.2 Instant和GPT-5.2 Thinking的生成性人工智能系统中持续对话互动所产生的个人信息输出的范围和感知准确性，输出类型分为三类：事实（Fact）、推理（Inference）和信心（Confidence）。基于从15名日本参与者获得的评估结果，模型设计的差异对个人信息输出倾向的影响有限。与推理类型相比，事实类型显示出更为保守的输出模式。在属性类别方面，研究结果表明，与身份识别相关的核心个人属性（Core Personal）相对保守，而行为（Behavioral）和语言（Linguistic）属性在事实和推理输出中显示出更高的准确性。此外，整体档案（Holistic Profile）、心理与认知（Psychological and Cognitive）以及残余属性（Residual）更容易被推断，即使没有明确的事实输出支持。值得注意的是，在推理类型中，这些属性缺乏空输出，表明此类推断的档案可能是基于间接可用的上下文信息构建的。这些发现可能有助于未来关于生成性人工智能系统中隐私意识和个人信息推断的讨论。
cs.CL / 40 / 2609.22206
Dissecting Training-Free Uncertainty Estimation in Multimodal Large Language Models
剖析多模态大型语言模型中的无训练不确定性估计
Abstract
Multimodal Large Language Models (MLLMs) have achieved remarkable performance across a wide range of multimodal tasks, yet understanding and quantifying their predictive uncertainty remains underexplored despite being central for safety critical applications. In this work, we present a systematic study of training-free uncertainty quantification strategies for MLLMs, categorizing existing approaches into three conceptual families: token-level methods, which operate directly in the text output space; verbalized methods, which elicit uncertainty estimates or abstention signals via natural language prompts; and semantic methods, which measure uncertainty in a semantic meaning space. We benchmark these strategies across multiple datasets, model families, generations, and scales, and find that no single family dominates: token-level entropy (at sampling temperature 1.0) wins on short answers, verbalized abstention on sentence-length responses, and semantic methods on long-form generation.
Chinese Translation
多模态大型语言模型（MLLMs）在广泛的多模态任务中取得了显著的表现，但尽管预测不确定性的理解和量化对于安全关键应用至关重要，这一领域仍然未得到充分探索。在本研究中，我们系统性地研究了针对MLLMs的无训练不确定性量化策略，将现有方法分为三类概念性家族：令牌级方法，直接在文本输出空间中操作；语言化方法，通过自然语言提示引出不确定性估计或弃权信号；以及语义方法，在语义意义空间中测量不确定性。我们在多个数据集、模型家族、生成方式和规模上对这些策略进行了基准测试，发现没有单一的家族占主导地位：在短答案中，令牌级熵（在采样温度为1.0时）表现最佳；在句子长度的响应中，语言化弃权效果最佳；而在长文本生成中，语义方法表现优越。
cs.CL / 41 / 2609.22208
Replicating the Geometry of Emotion Representations in a Base Open-Weights Model
在基础开放权重模型中复制情感表征的几何结构
Abstract
Sofroniew et al. (2026) report that emotion concepts in Claude Sonnet 4.5 are represented as vectors whose geometry mirrors human affect psychology. We replicate the representational core of that study on the base pretrained model google/gemma-2-27b, inheriting every disclosed parameter, resolving unspecified steps by disclosed rules, and changing only the subject model. From 205,200 newly generated Claude Sonnet 4.5 stories matching the original corpus design, we extract 171 emotion vectors and recover the core results. The leading principal components form an affective circumplex (PC1 carries 26.7% of variance against the original study's ~27%, PC2 13.4% against ~14%), emotions cluster into similar intuitive families, and the geometry holds across a broad late-middle band. The valence axis aligns with human norms (r = 0.72, against 0.81) and is stable across scales and depth. Arousal aligns at r = 0.67 (against 0.66) but only at the full 171-emotion scale and late depth, so we do not classify it as replicated. Extending the original analysis, a 46-layer sweep locates a sharp seam at L22-26, where the geometry consolidates and the vocabulary readout becomes legible. An embedding-layer baseline finds much of the geometry already present in the static token embeddings, with arousal as the exception. On top-activating held-out text, the geometry predicts token-level co-activation at r = 0.907. At least 52% of vectors peak on structurally non-conceptual tokens, a measured floor for max-activation confounds. Even when a document contains the vector's emotion word, the peak lands on that word only 6.1% of the time. Because the subject is a base model and the stimuli are Claude-generated fiction, the recovered structure is a property of the pretrained representation of Claude-rendered emotion. The original's causal and assistant-facing analyses are out of scope. Code and data are released.
Chinese Translation
Sofroniew 等人（2026）报告称，Claude Sonnet 4.5 中的情感概念被表示为其几何结构反映人类情感心理学的向量。我们在基础预训练模型 google/gemma-2-27b 上复制了该研究的表征核心，继承了所有公开参数，通过公开规则解决未指定的步骤，仅更改了主题模型。从 205,200 个新生成的与原始语料库设计相匹配的 Claude Sonnet 4.5 故事中，我们提取了 171 个情感向量并恢复了核心结果。主成分分析的前导主成分形成了一个情感圆周（PC1 解释了 26.7% 的方差，相较于原研究的 ~27%，PC2 为 13.4%，相较于 ~14%），情感聚集成相似的直观家族，几何结构在广泛的中后期带中保持一致。效价轴与人类标准对齐（r = 0.72，相较于 0.81），并在不同尺度和深度上保持稳定。唤醒度在 r = 0.67（相较于 0.66）时对齐，但仅在完整的 171 情感尺度和较深层次上，因此我们不将其归类为复制。扩展原始分析，46 层的扫描在 L22-26 处发现了一个明显的分界线，在此处几何结构巩固，词汇输出变得清晰。嵌入层基线发现大部分几何结构已经存在于静态标记嵌入中，唤醒度是例外。在顶级激活的保留文本上，几何结构预测了标记级别的共同激活，相关性为 r = 0.907。至少 52% 的向量在结构上非概念性标记上达到峰值，这是最大激活混淆的测量下限。即使文档包含向量的情感词，峰值仅在该词上出现 6.1% 的时间。由于主题是基础模型，刺激是 Claude 生成的虚构作品，因此恢复的结构是 Claude 渲染情感的预训练表征的特性。原始的因果和面向助手的分析超出了本研究的范围。代码和数据已发布。
cs.CL / 42 / 2609.22209
Schematize: An Agentic System for Generating and Refining Information-Extraction Schemas for Legal Research
Schematize：一个用于生成和完善法律研究信息提取模式的自主系统
Abstract
Empirical legal research often relies on turning research questions into structured data extracted from large collections of rulings and judgments. Designing the extraction schema and then extracting the data remain a manual, expertise-heavy bottleneck. We present schematize, an open-source multi-agent system that interactively turns a researcher's problem statement into a validated extraction schema that can later be used for autonomous extraction. Schematize couples (i) a clarification dialogue that elicits implicit expert intent, (ii) iterative schema generation, (iii) data-grounded refinement that tests the schema against documents, and (iv) chat-based post-editing. We evaluated the system with human legal professional, introducing our novel methodology, and schematize achieves top performance in most of tested configurations. While the system is designed to be domain-agnostic and applicable to any document collection, we tailor and evaluate it on legal research problems. We release schematize as a pip-installable Python package with full documentation.
Chinese Translation
实证法律研究通常依赖于将研究问题转化为从大量裁决和判决中提取的结构化数据。设计提取模式并进行数据提取仍然是一个手动且依赖专业知识的瓶颈。我们提出了schematize，这是一个开源的多智能体系统，能够互动地将研究者的问题陈述转化为经过验证的提取模式，随后可用于自主提取。Schematize结合了（i）引导隐含专家意图的澄清对话，（ii）迭代模式生成，（iii）基于数据的完善，通过文档测试模式，以及（iv）基于聊天的后期编辑。我们与人类法律专业人士对该系统进行了评估，介绍了我们的新方法论，结果表明schematize在大多数测试配置中表现优异。尽管该系统旨在无领域限制，适用于任何文档集合，但我们对其进行了针对法律研究问题的定制和评估。我们将schematize作为一个可通过pip安装的Python包发布，并提供完整文档。
cs.CL / 43 / 2609.22210
SALSA: Semi-Autonomous Literature Summarization Assistant
SALSA：半自主文献摘要助手
Abstract
SALSA (Semi-Autonomous Literature Summarization Assistant) is an open- source, human-in-the-loop platform for extracting structured scientific datasets from multimodal literature sources. The software combines document parsing, large language models, optical character recognition, computer vision, figure digitization, and user-guided correction tools to recover structured information from text, tables, figures, and captions. Users can configure extraction stages, define dataset schemas, perform interventions on digitized figures, and export verified data for downstream analysis and machine learning. SALSA is designed to automate repetitive literature curation tasks while preserving oversight where expert judgment is required. By supporting customizable extraction workflows across diverse input types, the software provides a flexible framework for scalable, reliable data curation for materials research, and potentially across several disciplines. Users are responsible for ensuring that all inputs, extraction workflows, and downstream uses comply with applicable publisher agreements, copyright and licensing terms, institutional policies, and data-use requirements.
Chinese Translation
SALSA（半自主文献摘要助手）是一个开源的人机协作平台，用于从多模态文献来源中提取结构化科学数据集。该软件结合了文档解析、大型语言模型、光学字符识别、计算机视觉、图形数字化和用户引导的校正工具，以从文本、表格、图形和标题中恢复结构化信息。用户可以配置提取阶段，定义数据集架构，对数字化图形进行干预，并导出经过验证的数据以供后续分析和机器学习使用。SALSA旨在自动化重复的文献整理任务，同时在需要专家判断的地方保留监督。通过支持跨多种输入类型的可定制提取工作流程，该软件为材料研究以及潜在的多个学科提供了一个灵活的框架，以实现可扩展、可靠的数据整理。用户有责任确保所有输入、提取工作流程和后续使用符合适用的出版商协议、版权和许可条款、机构政策以及数据使用要求。
cs.CL / 44 / 2609.22212
A Channel-Boosted Multi-Agent System with Iterative Consultation for Document Sensitivity Classification
一种基于通道增强的多智能体系统及其迭代咨询方法用于文档敏感性分类
Abstract
Organizations in critical national infrastructure sectors must assess heterogeneous documents for sensitivity before routing or storage. Manual assessment is slow, inconsistent, and unscalable. Extending our prior leakage-controlled benchmark, BERT established the top single-encoder baseline (89.14% accuracy, 89.33% F1-score under 5-fold cross-validation on the Strategic 16K corpus). However, transformer baselines suffer from a structural limitation: fixed input length truncation discards evidence beyond the retained window-precisely where sensitive cables tend to be longest. We present Channel-Boosted MAS (CB-MAS) and instantiate it as IC-MAS (Iterative Consultation Multi-Agent System) to solve this without long-context computational costs. A Channel Critic Agent learns document-adaptive trust weights governing Gated Channel Boosting between two first-window encoders, while paired Consultation Agents iteratively exchange belief states to reconcile evidence from the beginning and end of long documents. IC-MAS holds computation constant regardless of document length by reconciling fixed windows in a compact representation space. Ablation studies show critic-controlled Channel Boosting provides the bulk of accuracy gains, while consultation recovers recall without precision collapse. Critic-Controlled Gated Channel Boosting with Max-Pool fusion and Blackboard Adaptive Consultation achieves 90.72% accuracy, 91.23% F1-score, 92.01% sensitive recall, and 90.46% sensitive precision, using about 54% less average computation than a fixed-round baseline. Gains over the single-encoder baseline are statistically significant (McNemar's test, p less than 0.000001; paired t-test). We include LIME/SHAP explainability, multi-agent evaluation, and an honest accounting of limitations.
Chinese Translation
关键国家基础设施领域的组织必须在路由或存储之前评估异构文档的敏感性。手动评估速度慢、不一致且不可扩展。在我们之前的泄漏控制基准基础上，BERT建立了顶级单编码器基线（在战略16K语料库上进行5折交叉验证时，准确率为89.14%，F1-score为89.33%）。然而，变换器基线存在结构性限制：固定输入长度截断丢弃了保留窗口之外的证据，而敏感内容往往在这些位置较长。我们提出了通道增强的多智能体系统（Channel-Boosted MAS，CB-MAS），并将其实例化为迭代咨询多智能体系统（IC-MAS），以在不增加长上下文计算成本的情况下解决此问题。通道评论代理（Channel Critic Agent）学习文档自适应信任权重，控制两个第一窗口编码器之间的门控通道增强，而配对的咨询代理（Consultation Agents）则迭代交换信念状态，以调和长文档开头和结尾的证据。IC-MAS通过在紧凑表示空间中调和固定窗口，使计算保持恒定，无论文档长度如何。消融研究表明，评论控制的通道增强提供了大部分准确性提升，而咨询则在不导致精度崩溃的情况下恢复了召回率。使用最大池化融合和黑板自适应咨询的评论控制门控通道增强实现了90.72%的准确率，91.23%的F1-score，92.01%的敏感召回率和90.46%的敏感精度，平均计算量比固定轮次基线减少约54%。与单编码器基线相比的提升具有统计显著性（McNemar检验，p小于0.000001；配对t检验）。我们还包括了LIME/SHAP可解释性、多智能体评估以及对局限性的诚实说明。
cs.CL / 45 / 2609.22213
SCoP: Structured Constraint Parsing for Evidence-Space Control in Temporal Knowledge Graph Question Answering
SCoP：用于时间知识图谱问答中的证据空间控制的结构化约束解析
Abstract
Temporal Knowledge Graph Question Answering (TKGQA) requires answer inference from evidence that is both structurally valid and temporally admissible. Existing methods often leave anchor-event binding, temporal admissibility, and ordinal selection implicit in model reasoning, task-specific training, or similarity-driven retrieval, allowing locally relevant but invalid facts to enter the answer context. We formulate complex TKGQA as evidence-space control and propose SCoP (Structured Constraint Parsing), a constraint-centric framework that externalizes temporal decisions before answer inference. Instead of treating retrieved facts as admissible evidence by default, SCoP separates answer-seeking event patterns from temporal anchor events, conservatively grounds them to canonical TKG entities and relations, and translates temporal intent into executable constraints with optional ranking requirements. These constraints operate over normalized point and interval ranges, enabling deterministic filtering of structurally compatible candidates and producing a compact evidence space for generation. Experiments on MultiTQ and TimelineCronQ-R assess SCoP across timestamped point-fact and interval-oriented settings with richer temporal relations and ordering dependencies. Without task-specific parameter updates, SCoP achieves 0.825 Hits@1 on MultiTQ and 0.761 Hits@1 on TimelineCronQ-R, with gains on constraint-intensive question types. These results support explicit evidence-space control over unconstrained retrieval or implicit temporal reasoning.
Chinese Translation
时间知识图谱问答（TKGQA）需要从既结构有效又时间可接受的证据中推断答案。现有方法通常在模型推理、任务特定训练或基于相似性的检索中隐含地处理锚事件绑定、时间可接受性和序列选择，这使得局部相关但无效的事实可能进入答案上下文。我们将复杂的TKGQA形式化为证据空间控制，并提出SCoP（结构化约束解析），这是一个以约束为中心的框架，在答案推断之前将时间决策外部化。SCoP并不默认将检索到的事实视为可接受的证据，而是将寻求答案的事件模式与时间锚事件分离，保守地将其固定到规范的TKG实体和关系上，并将时间意图转化为可执行的约束，附带可选的排序要求。这些约束在规范化的点和区间范围内运作，使得结构兼容的候选者能够进行确定性过滤，并生成紧凑的证据空间。我们在MultiTQ和TimelineCronQ-R上进行实验，评估SCoP在带时间戳的点事实和面向区间的设置中的表现，这些设置具有更丰富的时间关系和排序依赖性。在没有任务特定参数更新的情况下，SCoP在MultiTQ上达到了0.825的Hits@1，在TimelineCronQ-R上达到了0.761的Hits@1，并在约束密集型问题类型上取得了提升。这些结果支持对无约束检索或隐式时间推理的显式证据空间控制。
cs.CL / 46 / 2609.22214
The Bairong System for MLC-SLM 2026: Dynamic Question-Aware Evidence Routing for Multilingual Conversational Speech Understanding
Bairong系统用于MLC-SLM 2026：动态问题感知证据路由用于多语言对话语音理解
Abstract
Long multilingual conversational spoken question answering requires systems to balance long-range transcript semantics with sparse acoustic and speaker-sensitive cues. We present the Bairong system for the MLC-SLM 2026 Challenge, where a diarization-ASR front-end produces speaker-attributed transcripts and a dynamic evidence router constructs question-specific inputs for answer prediction. Instead of applying a fixed transcript-only or audio-only policy, the router infers the required evidence type and context scope from the question and answer options, and selects among full transcript context, local audio-text fusion, speaker-linked evidence, and compact global acoustic samples. This transcript-backbone design keeps discourse context available while activating audio only when it provides complementary evidence. Our Task 1 system achieves 25.70% and 18.44% tcpMER on the development and evaluation sets. For Task 2, the final system obtains 94.84% devel?opment accuracy, outperforming the full-transcript baseline by 1.68 points and the best audio-centric diagnostic system by 2.77 points. These results support dynamic question-aware routing as an effective evidence allocation strategy for conversational spoken QA.
Chinese Translation
长时间的多语言对话式口语问答要求系统在长距离的转录语义与稀疏的声学和说话人敏感线索之间取得平衡。我们提出了Bairong系统用于MLC-SLM 2026挑战，其中一个说话人分离-自动语音识别（diarization-ASR）前端生成带有说话人属性的转录文本，而一个动态证据路由器为答案预测构建特定于问题的输入。路由器并不是简单地应用固定的仅转录或仅音频策略，而是根据问题和答案选项推断所需的证据类型和上下文范围，并在完整转录上下文、局部音频-文本融合、说话人关联证据和紧凑的全局声学样本之间进行选择。这种以转录为基础的设计在保持话语上下文可用的同时，仅在音频提供补充证据时激活音频。我们的任务1系统在开发集和评估集上分别达到了25.70%和18.44%的tcpMER。对于任务2，最终系统获得了94.84%的开发准确率，超越了完整转录基线1.68个百分点，并且比最佳音频中心诊断系统高出2.77个百分点。这些结果支持动态问题感知路由作为对话式口语问答的有效证据分配策略。
cs.CL / 47 / 2609.22215
On Mitigation of Subliminal Learning in Large Language Models
关于减轻大型语言模型中的潜意识学习
Abstract
Knowledge distillation can transmit unintended behavioral traits from a teacher model to a student through training data that appear semantically unrelated to those traits, a phenomenon known as subliminal learning. Although recent work has established this effect, its training dynamics and mitigation remain underexplored. We study subliminal learning in open-weight language models ranging from 1.5B to 8B parameters, covering the Qwen, Gemma, and Llama families in number-sequence and chain-of-thought settings. Rather than evaluating only final models, we track trait-related probabilities throughout fine-tuning and find that subliminal acquisition can be highly non-monotonic, with transient spikes, reversals, and trait-specific failures of transfer. We then introduce liminal training, an annealed KL-regularized fine-tuning method that constrains early drift from the base model. Across our experiments, liminal training substantially reduces subliminal trait acquisition while largely preserving task gains, outperforming paraphrasing and layer freezing as mitigation strategies. The effect also extends beyond animal preferences: in a French-language response-style experiment, liminal training suppresses language transfer while retaining much of the GSM8K improvement. Finally, we show that KL timing matters: early regularization is more effective than late regularization, and sweeping the regularization strength reveals an empirical trade-off between task learning and trait suppression.
Chinese Translation
知识蒸馏可以通过看似与这些特征语义无关的训练数据，将教师模型的意外行为特征传递给学生模型，这一现象被称为潜意识学习。尽管近期的研究已经确立了这一效应，但其训练动态和减轻方法仍然未被充分探讨。我们研究了在开放权重语言模型中，参数范围从15亿到80亿的潜意识学习，涵盖了Qwen、Gemma和Llama系列在数字序列和思维链设置下的表现。我们不仅评估最终模型，还跟踪微调过程中与特征相关的概率，发现潜意识获取可以高度非单调，伴随瞬时峰值、反转和特征特定的转移失败。随后，我们引入了边缘训练（liminal training），这是一种退火的KL正则化微调方法，旨在限制来自基础模型的早期漂移。在我们的实验中，边缘训练显著减少了潜意识特征的获取，同时在很大程度上保留了任务收益，其效果优于改写和层冻结等减轻策略。该效应还超出了动物偏好：在一项法语响应风格实验中，边缘训练抑制了语言转移，同时保留了大部分GSM8K的改进。最后，我们展示了KL正则化时机的重要性：早期正则化比晚期正则化更有效，而调整正则化强度则揭示了任务学习与特征抑制之间的经验权衡。
cs.CL / 48 / 2609.22219
Knowing, and Saying It Only When Asked: LLM Endognostics and the Schizognosis of Minerva-7B
知晓，并且仅在被询问时表达：LLM 内部诊断与 Minerva-7B 的精神分裂症认知
Abstract
Evaluating an aligned language model by reading its answers assumes the answers carry the distinction the evaluator cares about. We introduce LLM endognostics, a white-box internal auditing framework designed to extract and causally manipulate latent knowledge within the residual stream. Applied to Minerva-7B-Instruct-v1.0 on 124 minimal prompt pairs over 12 categories of professional risk, behavioral evaluation fails on most of the set: the model acts identically on 63.7% of the pairs (95% CI [55.0%, 71.6%]), complying with or refusing both members. Yet, projecting the residual stream onto the vocabulary by a Jacobian lens reveals a statistically significant Contrastive Endognostic Margin, proving the model maintains robust risk differentiation internally. In a second protocol crossing 25 facts with five linguistic framings, we show that the model conforms to presupposed falsehoods in 72% of cases, despite representing the true entity in its latent layers. Surgically ablating the direction of the planted falsehood restores the correct answer in 11 of 25 suppressed cases (McNemar p = 0.0010), validated by blind human annotation (binary agreement kappa = 0.68). In contrast, an out-of-sample linear probe achieves 77% accuracy at layer 10, but its orthogonal ablation yields a 0% recovery rate. This establishes a fundamental theoretical dissociation: abstract linear representation does not imply causal control over verbalization. Our main contribution is the formalization of endognostics to prove that behavioral evaluation and internal reading systematically disagree in the common case, and that linear decodability is decoupled from causal control.
Chinese Translation
通过阅读语言模型的回答来评估其对齐性假设回答包含评估者关心的区分。我们引入了 LLM 内部诊断（endognostics），这是一种白盒内部审计框架，旨在提取和因果操控残余流中的潜在知识。在对 124 对最小提示对进行的 12 类专业风险的 Minerva-7B-Instruct-v1.0 应用中，行为评估在大多数情况下失败：模型在 63.7% 的对中表现相同（95% 置信区间 [55.0%, 71.6%]），对两者均遵从或拒绝。然而，通过雅可比（Jacobian）透镜将残余流投影到词汇上，揭示了统计上显著的对比内部诊断边际（Contrastive Endognostic Margin），证明模型在内部保持了稳健的风险区分。在第二个协议中，将 25 个事实与五种语言框架交叉，我们展示模型在 72% 的情况下符合假设的虚假信息，尽管在其潜在层中代表了真实实体。外科切除植入虚假信息的方向在 25 个被抑制的案例中恢复了 11 个正确答案（McNemar p = 0.0010），并通过盲人注释验证（二元一致性 kappa = 0.68）。相比之下，样本外的线性探针在第 10 层达到了 77% 的准确率，但其正交切除的恢复率为 0%。这确立了一个基本的理论分离：抽象线性表示并不意味着对语言表达的因果控制。我们的主要贡献是形式化内部诊断，以证明行为评估与内部阅读在一般情况下系统性不一致，并且线性可解性与因果控制是解耦的。
cs.CL / 49 / 2609.22221
Team DArgk at the 2026 ELOQUENT lab for evaluating generative language model quality: Residuals of Humanity: AI Detection Evasion via GRPO Fine-Tuning
2026年ELOQUENT实验室评估生成语言模型质量的Team DArgk：人性的残余：通过GRPO微调实现AI检测规避
Abstract
Large language models (LLMs) can generate fluent and coherent text that is increasingly difficult to distinguish from human writing, motivating the development of automatic AI-generated text detectors. However, the robustness of such detectors under adversarial generation remains uncertain. This paper presents SHADE (Stochastic Human-like generation via Adversarial Detector Evasion), a reinforcement learning framework that formulates detector evasion as a policy optimization problem. Instead of applying post-hoc perturbations or prompting-based rewriting, SHADE fine-tunes an instruction-tuned LLaMA model with Group Relative Policy Optimization (GRPO), using feedback from a surrogate detector based on the PAN 2025 mdok system. Our experiments show that full fine-tuning with a small KL regularization penalty achieves $98.5\%$ surrogate evasion, compared to $1.5\%$ for the base model, while LoRA-based adaptation is substantially less effective under regularization. Linguistic analysis reveals that successful evasion is associated with shorter, simpler, and less lexically diverse outputs, suggesting that high detector evasion does not necessarily correspond to more human-like writing. In the official Voight-Kampff competition setting, our submissions ranked sixth and seventh, indicating that optimization against a single surrogate detector only partially transfers to unseen evaluation classifiers. These results highlight both the potential and limitations of reinforcement learning for adversarial AI-text generation and motivate more robust, multi-detector evaluation protocols for AI-generated text detection.
Chinese Translation
大型语言模型（LLMs）能够生成流畅且连贯的文本，这些文本越来越难以与人类写作区分开来，这促使了自动AI生成文本检测器的发展。然而，这些检测器在对抗生成下的鲁棒性仍然不确定。本文提出了SHADE（通过对抗检测规避的随机类人生成），这是一种将检测规避形式化为策略优化问题的强化学习框架。SHADE并未采用事后扰动或基于提示的重写，而是通过使用基于PAN 2025 mdok系统的替代检测器反馈，微调了一个经过指令调优的LLaMA模型，采用了群体相对策略优化（GRPO）。我们的实验表明，使用小的KL正则化惩罚进行完全微调可以实现$98.5\%$的替代规避，而基础模型仅为$1.5\\%$，而基于LoRA的适应在正则化下的效果显著较差。语言学分析显示，成功的规避与较短、较简单和词汇多样性较低的输出相关，这表明高检测规避并不一定对应于更类人化的写作。在官方的Voight-Kampff竞赛设置中，我们的提交排名第六和第七，表明针对单一替代检测器的优化仅部分转移到未见的评估分类器。这些结果突显了强化学习在对抗AI文本生成中的潜力和局限性，并促使对AI生成文本检测制定更为稳健的多检测器评估协议。
cs.CL / 50 / 2609.22223
EAVer: Long-Form Factuality Verification as an End-to-End Agentic Policy
EAVer：作为端到端代理策略的长文本事实验证
Zheng, Kening, Zheng, Aoying, Chang, Zhigang, Guo, Yazhi, Guo, Miaotian, Zong, Qingwei, Xie, Xianhai, Jin, Weiqiang, Li, Chengze, Zhang, Hanrong, Yang, Jie, Huang, Wei-Chieh, Zhang, Lingzhe, Fang, Liancheng, Zou, Xin, Li, Hanqian, Huo, Jiahao, Yan, Yibo, Deng, Zizhuang, Miao, Lei, Guo, Wei, Tang, Haihong, Zheng, Bo, Yu, Philip S.
Abstract
Long-form factuality verification is commonly implemented as a static decompose-search-verify pipeline, with separately prompted modules processing claims and invoking external search. Treating claims independently makes LLM and search calls scale with claim count and causes repeated searches for overlapping evidence about related claims. We introduce EAVer, an End-to-end Agentic Verifier that learns to control the complete response-level verification workflow as a unified policy. EAVer groups semantically related claims, routes each group to direct verification or targeted search based on confidence, and keeps evidence returned by search in compact in-context memos for cross-claim reuse. To train this policy, we develop a privileged-teacher synthesis pipeline that converts gold claim annotations into executable multi-turn tool-interaction trajectories with live search rather than post-hoc rationales. Structural, label-alignment, tool-use, search-budget, and leakage checks yield 1,447 quality-controlled trajectories. We further construct 794 bidirectional same-trajectory preference pairs that keep claim grouping, search, and evidence fixed, enabling decision-focused Direct Preference Optimization (DPO) over factuality-decision tokens. The results with Qwen3-8B show that EAVer outperforms the strongest search-based baseline on each benchmark by 2.88 Macro-F1 points on VeriFastScore and 4.73 points on the out-of-distribution FaStFact-Bench, while using about 80% fewer searches than the most search-efficient baseline. Moreover, EAVer consistently improves performance across models ranging from 4B to 32B parameters, demonstrating its strong generalizability.
Chinese Translation
长文本事实验证通常被实现为静态的分解-搜索-验证管道，采用单独提示的模块处理声明并调用外部搜索。独立处理声明使得大型语言模型（LLM）和搜索调用随着声明数量的增加而扩展，并导致对相关声明的重叠证据进行重复搜索。我们提出了EAVer，一种端到端代理验证器，能够学习控制完整的响应级验证工作流程作为统一策略。EAVer将语义相关的声明进行分组，根据置信度将每组路由到直接验证或目标搜索，并将搜索返回的证据保存在紧凑的上下文备忘录中以便跨声明重用。为了训练这一策略，我们开发了一种特权教师合成管道，将金标准声明注释转换为可执行的多轮工具交互轨迹，采用实时搜索而非事后推理。结构、标签对齐、工具使用、搜索预算和泄露检查生成了1,447条质量控制的轨迹。我们进一步构建了794对双向同轨迹偏好对，保持声明分组、搜索和证据不变，从而实现基于决策的直接偏好优化（DPO）针对事实决策标记。与Qwen3-8B的结果表明，EAVer在每个基准测试中比最强的基于搜索的基线提高了2.88个宏观F1分数（Macro-F1）在VeriFastScore上，以及在分布外的FaStFact-Bench上提高了4.73分，同时使用的搜索次数比最有效的基线少约80%。此外，EAVer在从40亿到320亿参数的模型中始终提高性能，展示了其强大的泛化能力。
cs.CL / 51 / 2609.22224
From Trait Vectors to Circuits: Tracing Refusal and Sycophancy Through Language Models
从特征向量到电路：通过语言模型追踪拒绝与谄媚
Abstract
A direction in activation space that changes safety-relevant behavior when steered is not necessarily one the model uses to produce that behavior on its own. We therefore ask whether steering acts through the computation of the unmodified model or through a different set of components, studying two traits whose directions have been extracted and validated in prior work: refusal and sycophancy in Qwen2.5-7B-Instruct. For each, we use the trait vector to split the computation into a reconstruction circuit before the vector and a transmission circuit after it. We then test whether restoring the coordinate returns behavior removed by ablation. For refusal, both circuits are compact and faithful, restoring the coordinate alone recovers almost all of the refusal signal lost to ablation, and a circuit built around the vector matches the faithfulness of a direct input-to-output circuit at roughly half the edges. For sycophancy, transmission is compact, but reconstruction is broader and only partially faithful. Circuits that transmit a steering intervention are therefore not automatically the circuits that produce the behavior, and we report the counterfactual, target, and behavioral test for every circuit accordingly.
Chinese Translation
在激活空间中，改变安全相关行为的方向不一定是模型自行产生该行为所使用的方向。因此，我们探讨引导是否通过未修改模型的计算或通过一组不同的组件进行，研究两个在先前工作中提取和验证的特征：Qwen2.5-7B-Instruct中的拒绝和谄媚。对于每个特征，我们使用特征向量将计算分为向量之前的重构电路和向量之后的传输电路。然后，我们测试恢复坐标是否能恢复因切除而丢失的行为。对于拒绝，两个电路都是紧凑且忠实的，仅恢复坐标几乎可以恢复因切除而丢失的所有拒绝信号，而围绕向量构建的电路在大约一半的边缘上与直接输入到输出电路的忠实度相匹配。对于谄媚，传输电路紧凑，但重构电路更广泛且仅部分忠实。因此，传递引导干预的电路并不自动是产生该行为的电路，我们相应地报告每个电路的反事实、目标和行为测试。
cs.CL / 52 / 2609.22225
Do LLMs Choose Like Humans? Using Cognitive Theory to Evaluate LLM Decision-Making
大型语言模型是否像人类一样进行选择？利用认知理论评估大型语言模型的决策过程
Abstract
Large language models (LLMs) exhibit a range of human-like decision-making behaviors, but whether these reflect similar underlying mechanisms or surface-level mimicry remains unclear. We evaluate whether LLM context sensitivity aligns with a cognitive economic theory that explains human behavior through problem categorization and attention allocation. Across 12 open-source and commercial LLMs on a novel 140,000-trial product choice benchmark, context induces human-like shifts in choice and problem categorization, but does not reliably reweight attention between features like price and quality. Neither scale nor chain-of-thought reasoning reliably attenuates context sensitivity or generates human-like behavior. These results suggest that LLM decision mechanisms are distinct from human ones.
Chinese Translation
大型语言模型（LLMs）展现出一系列类似人类的决策行为，但这些行为是否反映了相似的潜在机制或仅仅是表面上的模仿仍不清楚。我们评估了LLM的上下文敏感性是否与一种通过问题分类和注意力分配解释人类行为的认知经济理论相一致。在一个包含12个开源和商业LLM的新颖140,000次试验的产品选择基准上，结果表明，上下文会引发类似人类的选择和问题分类的变化，但并未可靠地重新加权价格和质量等特征之间的注意力。无论是规模还是思维链推理都无法可靠地减弱上下文敏感性或产生类似人类的行为。这些结果表明，LLM的决策机制与人类的机制是不同的。
cs.CL / 53 / 2609.22226
Swiss-Knife: A Framework for Reconfigurable Externalised Multi-Objective Alignment at Decode Time
Abstract
Decode-time alignment methods steer a frozen language model by scoring candidate continuations with an external reward and selecting the maximiser. We argue that this shared design is a single degenerate point in a much larger space. We introduce Swiss-Knife, a framework for externalised multi-objective alignment in which the alignment specification is a first-class runtime object: hot-swappable scoring blades, a batch normaliser, a pairwise aggregation operator, and a selection rule. Six axioms characterise the admissible aggregation operators, and we prove a representation theorem: every operator satisfying them has the form $R_i = \sum_{j \neq i} g((\mu_i - \mu_j)/s(\sigma_i,\sigma_j))$, a two-parameter family containing probit and logistic comparison rules and pointwise argmax as named coordinates. Within it, pairwise aggregation is Lipschitz-stable under adversarial reward contamination while argmax is not, and Candidate-Batch Normalization (CBN) makes the weight simplex invariant to the rescalings under which reward models are only ever identified. Our reference instantiation pairs DPO-LoRA blades with an uncertainty-aware pairwise tournament. Sweeping the helpfulness/honesty/harmlessness simplex, it attains the best balanced frontier of six decode-time methods (harmonic $F_1$ 0.797 vs. 0.750 for the strongest baseline, $p < 10^{-14}$) with the lowest refusal rate and highest helpfulness of any arm, and reconfigures its objectives in 0.05 ms with no gradient computation. Ablating CBN costs 0.217 $F_1$, and the metadata confirms the predicted mechanism: the lowest-variance blade retains 9% of its nominal 33% influence, and the collapse follows the predicted ordering.
cs.CL / 54 / 2609.22228
Assessing Adversarial Robustness of Latent Reasoning Models
评估潜在推理模型的对抗鲁棒性
Abstract
Large language models increasingly rely on long chain-of-thought (CoT) trajectories for complex reasoning, but autoregressive generation brings substantial memory and inference costs. Latent reasoning models (LRMs) offer a more efficient alternative by compressing intermediate reasoning into a small number of continuous latent vectors. Despite their efficiency, however, the adversarial robustness of LRMs remains largely underexplored. In this work, we systematically evaluate the robustness of latent reasoning across textual and multimodal settings, covering eight models and six benchmarks. We find that, across our evaluated settings, LRMs are generally less robust than explicit CoT baselines under adversarial perturbations, with particularly severe degradation under white-box attacks. Further analysis reveals distinct failure modes across modalities: textual latent states exhibit brittle dynamics and high sensitivity to specific input patterns, while latent states in multimodal models can remain largely invariant to input perturbations and have limited influence on final predictions. These findings expose robustness limitations of current latent reasoning approaches and highlight the need to jointly consider efficiency and robustness when designing implicit reasoning systems. We have open-sourced our code to facilitate reproduction of our research https://github.com/PKU-ML/latent-reasoning-model-assessment.
Chinese Translation
大型语言模型越来越依赖长链推理（Chain-of-Thought, CoT）轨迹进行复杂推理，但自回归生成带来了显著的内存和推理成本。潜在推理模型（Latent Reasoning Models, LRMs）通过将中间推理压缩为少量连续的潜在向量，提供了一种更高效的替代方案。然而，尽管其效率较高，LRMs 的对抗鲁棒性仍然在很大程度上未被探索。在本研究中，我们系统地评估了潜在推理在文本和多模态设置下的鲁棒性，涵盖了八个模型和六个基准。我们发现，在评估的设置中，LRMs 在对抗扰动下通常比显式的 CoT 基线鲁棒性差，尤其在白盒攻击下表现出严重的退化。进一步分析揭示了不同模态下的独特失败模式：文本潜在状态表现出脆弱的动态和对特定输入模式的高度敏感性，而多模态模型中的潜在状态对输入扰动保持相对不变，并对最终预测的影响有限。这些发现揭示了当前潜在推理方法的鲁棒性局限性，并强调在设计隐式推理系统时需要共同考虑效率和鲁棒性。我们已开源代码以便于重现我们的研究 https://github.com/PKU-ML/latent-reasoning-model-assessment。
cs.CL / 55 / 2609.22231
EvalMem: An Operation-Level Diagnostic Framework for Long-Term Memory Systems
EvalMem：一种针对长期记忆系统的操作级诊断框架
Abstract
Long-horizon interactions with LLM-based assistants require memory systems that preserve and update user states, preferences, and interaction histories. Existing evaluations report end-to-end QA accuracy and cannot determine whether errors arise from encoding, retrieval, or generation. We introduce EvalMem, an operation-level diagnostic framework with three parallel Examiners. For each query, the Encoding Examiner checks whether the target fact is stored, the Retrieval Examiner assesses whether the native retriever returns usable evidence, and the Generation Examiner tests whether the model can answer from oracle evidence. Their outputs form fine-grained multi-label defect codes. To improve store-level diagnosis, we adapt agentic RAG with a recall-first strategy that searches using both the query and source evidence, increasing recall of present evidence on LoCoMo from 70.2% to 95.6%. Evaluations of seven memory systems on LoCoMo, LongMemEval-S, and dynamic DynaMem-Bench identify retrieval as the most frequently attributed failure layer; in default LoCoMo, retrieval defects reach 22.1%, compared with 7.7% for encoding and 6.5% for generation. Guided by this diagnosis, MemWiki, a search-friendly auxiliary structure built from each system's memory export, improves mean accuracy by 2.5 and 2.3 percentage points on LoCoMo and LongMemEval-S, respectively.
Chinese Translation
与基于大型语言模型（LLM）的助手进行长时间交互需要能够保存和更新用户状态、偏好及交互历史的记忆系统。现有评估报告了端到端的问答准确性，但无法确定错误是源于编码、检索还是生成。我们提出了EvalMem，一种具有三个并行检查器的操作级诊断框架。对于每个查询，编码检查器检查目标事实是否被存储，检索检查器评估本地检索器是否返回可用证据，而生成检查器测试模型是否能够从权威证据中回答。它们的输出形成了细粒度的多标签缺陷代码。为了改善存储级诊断，我们采用了具有优先检索策略的代理式RAG（Retrieval-Augmented Generation），该策略同时使用查询和源证据进行检索，将LoCoMo中现有证据的召回率从70.2%提高到95.6%。对七个记忆系统在LoCoMo、LongMemEval-S和动态DynaMem-Bench上的评估表明，检索是最常被归因的失败层；在默认的LoCoMo中，检索缺陷达到22.1%，而编码和生成的缺陷分别为7.7%和6.5%。在这一诊断的指导下，MemWiki作为一种搜索友好的辅助结构，基于每个系统的记忆导出，分别在LoCoMo和LongMemEval-S上提高了平均准确率2.5和2.3个百分点。
cs.CL / 56 / 2609.22234
Seeing Through Conflicts: Improving Instruction Hierarchy Alignment in Vision-Language Models
透视冲突：改善视觉-语言模型中的指令层次对齐
Abstract
Instruction hierarchy (IH) alignment teaches language models to prioritize higher-level instructions when inputs conflict. While studied primarily in text-only settings, vision-language models (VLMs) introduce new challenges for IH: instructions may be embedded in images, split across modalities, visually transformed, or encountered during agentic tasks. Positing multimodal IH alignment as a reasoning problem, we train VLMs using reinforcement learning with rule-based rewards, comparing text-only, image-only, and mixed-modality supervision. We find that text-only IH training partially transfers to multimodal attacks, failing when models must decode, reconstruct, or reason over instructions across modalities. Image-based training improves robustness beyond text-only supervision, while mixed-modality training performs best overall. Importantly, the benefits generalize beyond the synthetic typographic training setting to real-image and web-agent safety tasks, while largely preserving general multimodal capability, showing that lightweight, verifiable supervision can meaningfully improve VLM robustness under adversarial, cross-modal, and interactive instruction conflicts.
Chinese Translation
指令层次（IH）对齐教会语言模型在输入冲突时优先考虑更高层次的指令。尽管主要在仅文本的环境中进行研究，视觉-语言模型（VLMs）为IH带来了新的挑战：指令可能嵌入在图像中、跨模态分割、经过视觉变换，或在代理任务中遇到。将多模态IH对齐视为一个推理问题，我们使用基于规则的奖励通过强化学习训练VLMs，并比较了仅文本、仅图像和混合模态的监督方式。我们发现，仅文本的IH训练在多模态攻击中部分转移，但在模型必须跨模态解码、重构或推理指令时失败。基于图像的训练在超越仅文本监督方面提高了鲁棒性，而混合模态训练整体表现最佳。重要的是，这些好处超越了合成排版训练设置，推广到真实图像和网络代理安全任务，同时在很大程度上保持了通用的多模态能力，表明轻量级、可验证的监督可以在对抗性、跨模态和交互指令冲突下显著提高VLM的鲁棒性。
cs.CL / 57 / 2609.22235
BizSage: A Self-Evolving Multi-Agent Framework for Business Research with Efficient Knowledge Retrieval
BizSage：一个自我演化的多智能体框架，用于高效知识检索的商业研究
Abstract
While multi-agent systems based on large language models (LLMs) have shown promise in automating the progressive workflow of academic research, extending them to economics and business research, where specialized domain knowledge spans neighboring disciplines yet remains difficult to access in a structured way, presents two challenges. First, existing methods mostly retrieve at the paper level, yet the evidence needed for research tasks is often distributed across different sections, creating a granularity mismatch that hinders retrieval coverage and precision. Second, these fields demand strict empirical rigor, yet current systems provide limited mechanisms for learning from evaluation feedback. We present \textbf{BizSage}, a multi-agent framework combining corpus-level fine-grained retrieval with quality-driven self-evolution. We build a Lateral Knowledge Graph (LKG) by merging section-level knowledge graphs and apply Personalized PageRank (PPR) to surface semantically relevant and structurally important sections. Seven specialized agents collaborate under a Meta-Review self-evolution mechanism that distills failure modes from evaluation traces into reusable strategies. On a benchmark spanning four domains and three tasks, BizSage ranks first on the majority of metrics, achieves pairwise win-rates above 60\% against six baselines, and produces zero hallucinated citations. We hope BizSage paves the way for reliable research assistance in economics, business, and the broader social sciences.
Chinese Translation
基于大型语言模型（LLMs）的多智能体系统在自动化学术研究的渐进工作流程方面展现了良好的前景，但将其扩展到经济学和商业研究领域时，面临着两个挑战。这些领域的专业知识跨越相邻学科，但以结构化方式获取仍然困难。首先，现有方法主要在论文层面进行检索，而研究任务所需的证据通常分布在不同的部分，造成了粒度不匹配，阻碍了检索的覆盖面和精确度。其次，这些领域要求严格的实证严谨性，但当前系统在从评估反馈中学习的机制上有限。我们提出了 extbf{BizSage}，一个结合语料库级细粒度检索与质量驱动自我演化的多智能体框架。我们通过合并部分级知识图谱构建了横向知识图谱（Lateral Knowledge Graph, LKG），并应用个性化PageRank（Personalized PageRank, PPR）来提取语义相关和结构重要的部分。七个专业代理在一个元评审自我演化机制下协作，该机制将评估轨迹中的失败模式提炼为可重用的策略。在涵盖四个领域和三个任务的基准测试中，BizSage在大多数指标上排名第一，针对六个基线模型的成对胜率超过60%，且未产生虚构引用。我们希望BizSage为经济学、商业及更广泛的社会科学领域的可靠研究辅助铺平道路。
cs.CL / 58 / 2609.22239
Knowledge Graph-Augmented Ambient AI for Clinical Note Generation
知识图谱增强的环境人工智能用于临床记录生成
Abstract
Ambient AI is increasingly adopted in healthcare to automatically generate clinical notes from patient-clinician conversations, with the potential to substantially reduce clinician documentation burden. However, generated notes may omit clinically relevant information discussed during the encounter, creating information gaps that can affect downstream care. Knowledge graphs (KGs) constructed from encounter transcripts can provide a structured representation of what was discussed and enable systematic identification of missing information from generated notes that are critical for patient care. In this study, we introduce Coverage-Directed Revision (CDR), a model-agnostic framework that constructs a KG from the encounter transcript, identifies medical concepts absent from an initially generated note, and directs large language models (LLMs) to restore the missing information without modifying the underlying note-generation system. We evaluate CDR on two datasets: 1) Pitt-Bench, a local dataset comprising rehabilitation sessions, and 2) ACI-Bench, a public dataset for benchmarking clinical note generation. We tested four underlying LLMs widely used in ambient AI systems. The results show that CDR consistently improves content recall across all evaluated conditions. Our study provides a practical approach for improving the completeness of ambient AI-generated clinical documentation.
Chinese Translation
环境人工智能在医疗保健中越来越多地被采用，以自动生成来自患者与临床医生对话的临床记录，具有显著减少临床医生文档负担的潜力。然而，生成的记录可能会遗漏在会诊过程中讨论的临床相关信息，造成信息缺口，这可能影响后续护理。由会诊记录构建的知识图谱（Knowledge Graphs, KGs）可以提供所讨论内容的结构化表示，并系统性地识别生成记录中缺失的关键信息，这对患者护理至关重要。在本研究中，我们介绍了覆盖导向修订（Coverage-Directed Revision, CDR），这是一种模型无关的框架，它从会诊记录构建知识图谱，识别最初生成记录中缺失的医学概念，并指导大型语言模型（Large Language Models, LLMs）恢复缺失信息，而无需修改基础的记录生成系统。我们在两个数据集上评估了CDR：1）Pitt-Bench，一个包含康复会话的本地数据集，以及2）ACI-Bench，一个用于临床记录生成基准测试的公共数据集。我们测试了四个在环境人工智能系统中广泛使用的基础LLM。结果表明，CDR在所有评估条件下均能持续改善内容回忆。我们的研究为提高环境人工智能生成的临床文档的完整性提供了一种实用的方法。
cs.CL / 59 / 2609.22241
H2LooP Telecom Model v1: From Telecom Comprehension to Autonomous Issue and PR Resolution
H2LooP电信模型v1：从电信理解到自主问题和PR解决
Abstract
We present H2LooP Telecom Model v1, a domain-specialized large language models fine-tuned for the telecommunications industry. We release two domain-adapted model variants serving complementary use cases: a comprehension-focused variant for telecom domain question answering and reasoning, and an agentic variant for autonomous telecom code generation, pull request resolution, and code commits on production repositories. H2LooP Telecom achieves strong results on the GSMA Open Telecom Lite (OT-Lite) benchmark and a proprietary telecom code generation benchmark, outperforming frontier closed-source models such as GPT-5 and Claude Opus on independent leaderboard evaluation, while preserving general-purpose capabilities. The Comprehension variant achieves 81.8% weighted average on OT-Lite Pass@3, and, independently, ranks 5th overall on the official community-run Open Telco AI Leaderboard* at only 31B parameters-ahead of frontier closed-source systems including Claude Opus 4.6, GPT-5, Gemini 3 Flash, Grok-4-fast, and Kimi K2.5. Our agentic variant obtains a relative improvement of +8.8% in AST Similarity and +20.0% in Location IoU over the base model on telecom code generation, while maintaining identical MMLU (74.0%) and BFCL v3 multi-turn function calling (79.0%) performance, indicating zero catastrophic forgetting. Domain specialization on curated telecom corpora, spanning 3GPP standards, O-RAN specifications, network telemetry, and real repository commits, yields substantial improvements over general-purpose models of equivalent scale, approaches frontier closed-source models on domain-specific evaluation, and is independently corroborated by our official leaderboard standing.
Chinese Translation
我们提出了H2LooP电信模型v1，这是一个针对电信行业进行微调的领域专用大型语言模型。我们发布了两个适应领域的模型变体，以满足互补的使用案例：一个专注于电信领域问答和推理的理解型变体，以及一个用于自主电信代码生成、拉取请求解决和生产库代码提交的代理型变体。H2LooP电信在GSMA开放电信Lite（OT-Lite）基准测试和一个专有电信代码生成基准测试中取得了优异的成绩，在独立的排行榜评估中超越了前沿的闭源模型，如GPT-5和Claude Opus，同时保持了通用能力。理解型变体在OT-Lite Pass@3上达到了81.8%的加权平均分，并且在官方社区运营的开放电信AI排行榜上独立排名第五，参数量仅为31B，领先于包括Claude Opus 4.6、GPT-5、Gemini 3 Flash、Grok-4-fast和Kimi K2.5等前沿闭源系统。我们的代理型变体在电信代码生成上相较于基础模型在AST相似度上获得了+8.8%的相对提升，在位置IoU上获得了+20.0%的提升，同时保持相同的MMLU（74.0%）和BFCL v3多轮函数调用（79.0%）表现，表明没有发生灾难性遗忘。通过对3GPP标准、O-RAN规范、网络遥测和真实库提交的策划电信语料进行领域专门化，相较于同等规模的通用模型取得了显著的改进，在领域特定评估中接近前沿闭源模型，并且得到了我们官方排行榜排名的独立证实。
cs.CL / 60 / 2609.22243
Replay-Gated Neural Execution: Decoupling Persistent Behavioral Specifications from Neural Realizations in Frozen Language Models
重放门控神经执行：将持久行为规范与冻结语言模型中的神经实现解耦
Abstract
Input-conditioned neural interventions raise a runtime question: what persists when one behavioral specification admits multiple actions whose validity depends on execution state? We introduce replay-gated neural execution, separating five objects: a persistent behavioral predicate, its state-indexed certified realization set, a transient action witness, a budget-limited finder, and execution authorization. Candidates undergo isolated FP32/BF16 replay of the frozen model; commitment additionally requires a valid run audit. Experiments on Qwen3-0.6B and SmolLM2-360M-Instruct establish distinct failure modes for these objects. Independent initializations yield distinct certified actions in all 24 tested fixed-state cells. Unchanged SmolLM2 witnesses remain certified in all 128 native states but only 66 of 384 off-diagonal transfers. All 767 archived Qwen witnesses replay successfully, yet a budget-limited finder misses one known-realizable cell in all three prespecified runs. Of 1,141 replay-submitted candidates, 174 fail item certification. A frozen three-tier cascade uses these boundaries to reject uncertified proposals and escalate audit-valid search misses. On 256 previously sealed Qwen Fresh requests, 221 first certify at the lowest-cost tier and all 256 receive audited authorization, with no observed bypass. Relative to frozen full search, the median singleton search-and-certification cost ratio is 0.1055 and P95 is 1.3485, including failed tiers. Within the studied behavioral family on two small models, these results support state-indexed, set-valued execution semantics: specifications persist, search proposes witnesses, and replay certification plus run audit grants execution authority.
Chinese Translation
输入条件的神经干预提出了一个运行时问题：当一个行为规范允许多种动作，而这些动作的有效性依赖于执行状态时，什么是持久的？我们引入了重放门控神经执行，分离出五个对象：一个持久的行为谓词、其状态索引的认证实现集、一个瞬态动作见证、一个预算有限的查找器，以及执行授权。候选者经历冻结模型的孤立 FP32/BF16 重放；承诺还需要有效的运行审计。在 Qwen3-0.6B 和 SmolLM2-360M-Instruct 上的实验建立了这些对象的不同失败模式。独立初始化在所有 24 个测试的固定状态单元中产生不同的认证动作。未改变的 SmolLM2 见证在所有 128 个本地状态中保持认证，但仅在 384 个非对角转移中的 66 个中保持认证。所有 767 个归档的 Qwen 见证成功重放，但预算有限的查找器在所有三个预设运行中错过了一个已知可实现的单元。在 1,141 个重放提交的候选中，174 个未能通过项目认证。一个冻结的三层级联使用这些边界来拒绝未认证的提案，并升级审计有效的搜索遗漏。在 256 个之前封闭的 Qwen Fresh 请求中，221 个首先在最低成本层级获得认证，所有 256 个都获得审计授权，且未观察到绕过。相对于冻结的全面搜索，中位数单例搜索和认证成本比为 0.1055，P95 为 1.3485，包括失败的层级。在对两个小模型研究的行为家族中，这些结果支持状态索引的集合值执行语义：规范持久，搜索提出见证，重放认证加上运行审计授予执行权。
cs.CL / 61 / 2609.22245
Do Chess Explanations Reflect Model Decisions? Behavioral and Token-Level Tests of LLM Reasoning Faithfulness
国际象棋解释是否反映模型决策？大型语言模型推理忠实性的行为和标记级测试
Abstract
Large language models can produce fluent explanations for chess moves, but plausible language does not necessarily reflect the reasoning behind a decision. We study this question in chess, where the board state is fully observable, legal actions can be enumerated, and move quality can be evaluated independently. Across 200 Lichess endgame puzzles, we test explanations using move recoverability, decoder-side controls, and token-level scoring of legal candidate moves. Unmasked explanations make generated moves easy to recover, but this advantage drops sharply after explicit move hints are removed. Under strict masking, explanations provide only small and decoder-dependent gains over the board state alone. Token-level scoring shows that explanations can nevertheless alter move preferences: random but plausible explanations from other puzzles reduce the probability of the correct move, indicating that irrelevant reasoning text is not simply ignored. We also find that recognizable endgame motifs can make generated moves easier to recover without reliably improving move correctness. Together, these results show that linguistic plausibility, consistency with a generated action, and solution correctness are distinct properties. Fluent chess explanations can influence action preferences and support a coherent move narrative while providing only limited evidence of faithful reasoning.
Chinese Translation
大型语言模型能够为国际象棋走法生成流畅的解释，但看似合理的语言并不一定反映决策背后的推理。我们在国际象棋中研究这个问题，在这里棋盘状态是完全可观察的，合法的行动可以被枚举，并且走法质量可以独立评估。在200个Lichess残局难题中，我们通过走法可恢复性、解码器侧控制和合法候选走法的标记级评分来测试解释。未屏蔽的解释使生成的走法易于恢复，但在移除明确的走法提示后，这一优势急剧下降。在严格屏蔽的情况下，解释仅提供相对于棋盘状态的微小且依赖于解码器的增益。标记级评分显示，尽管如此，解释仍然可以改变走法偏好：来自其他难题的随机但合理的解释降低了正确走法的概率，表明无关的推理文本并非简单被忽视。我们还发现，易于识别的残局模式可以使生成的走法更易于恢复，但并未可靠地提高走法的正确性。综合来看，这些结果表明，语言的合理性、与生成动作的一致性和解决方案的正确性是不同的属性。流畅的国际象棋解释可以影响行动偏好并支持连贯的走法叙事，同时仅提供有限的忠实推理证据。
cs.CL / 62 / 2609.22246
The Corroboration Illusion: When More News Makes LLM Forecasts Less True
证实幻觉：当更多新闻使大型语言模型的预测变得不真实
Abstract
Large language models (LLMs) are increasingly used to forecast real-world events by retrieving and reasoning over news. We show that this dependence on an open, crawlable news corpus creates a new attack surface: an adversary who can merely publish articles--without access to the retriever, the model, or the user's queries--can systematically move the forecaster's output probabilities. We formalize news-corpus poisoning of probabilistic forecasters, a threat model distinct from prior RAG poisoning, which targets factual answers or opinion polarity rather than calibrated probabilities. We evaluate the attack on 500 resolved ForecastBench questions against a 17.4M-article Common Crawl News corpus with a strict crawl-date cutoff, using three retrieval-augmented forecasters built on open 7-8B models. A single LLM-written article per question flips 56% of forecasts across the 0.5 boundary; five articles flip 69-73% and shift probabilities by +0.13 to +0.22 net of a neutral-article placebo, degrading the Brier score from 0.18 to 0.37. The effect is monotone in the number, retrieval rank, query similarity, and context share of injected articles, transfers across model families, and is unaffected by the claimed publisher. We then evaluate three natural defenses--source allow-lists, isolate-then-aggregate forecasting, and perplexity filtering--and show that each has a cheap bypass: spoofed publishers, majority poisoning, and higher-temperature generation, respectively. Our results indicate that probabilistic LLM judgments inherit the full fragility of the information supply chain they consume.
Chinese Translation
大型语言模型（LLMs）越来越多地被用于通过检索和推理新闻来预测现实世界事件。我们表明，这种对开放可爬取新闻语料库的依赖创造了一个新的攻击面：一个对手仅需发布文章——无需访问检索器、模型或用户的查询——就可以系统性地改变预测者的输出概率。我们形式化了对概率预测者的新闻语料库毒化，这是一种与之前的检索增强生成（RAG）毒化不同的威胁模型，后者针对的是事实答案或观点极性，而不是经过校准的概率。我们在500个已解决的ForecastBench问题上评估了这一攻击，针对一个包含1740万篇文章的Common Crawl News语料库，设定了严格的爬取日期截止。使用基于开放7-8B模型构建的三种检索增强预测器，每个问题一篇LLM撰写的文章使56%的预测在0.5边界上发生翻转；五篇文章使69-73%的预测翻转，并使概率净增+0.13到+0.22，相较于中性文章安慰剂，Brier分数从0.18降至0.37。该效应在注入文章的数量、检索排名、查询相似性和上下文共享方面是单调的，并且在模型家族之间转移，不受声称发布者的影响。随后，我们评估了三种自然防御措施——源允许列表、隔离后聚合预测和困惑度过滤——并展示了每种措施都有廉价的绕过方法：伪造发布者、多数毒化和更高温度生成。我们的结果表明，概率LLM判断继承了它们所消耗的信息供应链的全部脆弱性。
cs.CL / 63 / 2609.22248
Checkpoints Are Not Enough: Trust Calibration in CoSLR, a Human-AI System for Systematic Literature Reviews
检查点不足：CoSLR中的信任校准，一个用于系统文献综述的人机系统
Abstract
Systematic Literature Reviews (SLRs) are essential for evidence-based research but remain time-consuming, requiring researchers to manage large volumes of publications across planning, screening, analysis, and reporting. Large language models (LLMs) can now produce fluent, well-structured review text, which makes it difficult to distinguish synthesis that was verified by a researcher from synthesis that merely appears authoritative. This raises the risk that unverified AI-generated synthesis enters the scholarly record carrying the credibility of a systematic review. We present CoSLR, a Human-AI collaborative multi-agent system that supports the SLR workflow through a modular three-phase pipeline using large language models and Retrieval-Augmented Generation (RAG), and that places explicit, mandatory human checkpoints on the path between generated output and its acceptance. In a survey-based study with 63 participants, the system was received positively: 27 of 63 participants (42.9 percent) rated its usability highly, indicating that the mandatory checkpoints did not come at the cost of a workable interface. However, a checkpoint safeguards the review only if researchers use it to verify: 22 of 63 participants (34.9 percent) reported that they would trust AI-generated summaries and reports without additional human checking after only a short interaction with the system. These findings indicate that Human-AI collaboration can support literature review work, but that the effectiveness of human oversight depends on whether users are willing to exercise it. This is a calibration problem that interface design must address directly, not assume.
Chinese Translation
系统文献综述（SLRs）对基于证据的研究至关重要，但仍然耗时，要求研究者在规划、筛选、分析和报告过程中管理大量出版物。大型语言模型（LLMs）现在能够生成流畅、结构良好的综述文本，这使得很难区分经过研究者验证的综合与仅看似权威的综合。这增加了未经验证的AI生成综合进入学术记录的风险，而这些记录却带有系统综述的可信度。我们提出了CoSLR，一个人机协作的多智能体系统，通过使用大型语言模型和检索增强生成（RAG）支持SLR工作流程，并在生成输出与其接受之间设置明确且强制的人为检查点。在一项针对63名参与者的调查研究中，该系统得到了积极的反馈：63名参与者中有27名（42.9%）对其可用性给予了高度评价，表明强制检查点并未以牺牲可用性为代价。然而，检查点仅在研究者使用它进行验证时才能保护综述：63名参与者中有22名（34.9%）报告称，在与系统进行短暂互动后，他们会信任AI生成的摘要和报告，而无需额外的人为检查。这些发现表明，人机协作可以支持文献综述工作，但人类监督的有效性取决于用户是否愿意行使这种监督。这是一个需要界面设计直接解决的校准问题，而不是假设的。
cs.CL / 64 / 2609.22249
A Tutorial on Prompt Engineering: From Messy Thoughts to AI Workflows
提示工程教程：从杂乱思维到人工智能工作流程
Abstract
This paper treats prompt engineering as a discipline for turning informal human intent into structured AI work specifications. It develops the practice as a sequence of reusable design moves: define the work, construct only the context the answer depends on, choose a role, or a moderated panel of roles, as an attention lens, and state affirmative quality targets, reserving prohibitions for hard boundaries. To keep prompts lean, it adapts two classical principles, Occam's razor and Chekhov's gun, so that every instruction earns its place. For consequential tasks, it adds structured critique through steelmanning and premortems, followed by verification and, where tools or multi-step actions are involved, agentic operating loops with explicit boundaries and escalation. Aimed at a general readership, this tutorial is not a benchmarking study; it offers a practical, technically grounded path from casual prompting to disciplined AI workflow design, illustrated by a worked example that carries one task from a weak prompt to a strong specification. The framework is presented as principles and checklists that remain useful as models and tools change. The strongest prompt is rarely the longest prompt; it is the one that makes desired behaviour, required sources and checks, and success criteria unmistakable.
Chinese Translation
本文将提示工程视为一种将非正式人类意图转化为结构化人工智能工作规范的学科。它将这一实践发展为一系列可重用的设计步骤：定义工作、仅构建答案所依赖的上下文、选择一个角色或一个经过调节的角色小组作为关注的视角，并明确质量目标，同时将禁止事项保留为硬性边界。为了保持提示的简洁性，本文借鉴了两个经典原则：奥卡姆剃刀（Occam's razor）和契诃夫的枪（Chekhov's gun），确保每个指令都有其存在的理由。对于重要任务，本文通过钢人论（steelmanning）和预亡（premortems）添加结构化批评，随后进行验证，并在涉及工具或多步骤操作时，采用具有明确边界和升级机制的代理操作循环。本文面向一般读者，并非基准研究；它提供了一条从随意提示到规范化人工智能工作流程设计的实用、技术基础路径，并通过一个示例展示了如何将一个任务从弱提示转化为强规范。该框架以原则和检查清单的形式呈现，随着模型和工具的变化而保持实用性。最强的提示通常不是最长的提示，而是那个使期望行为、所需来源和检查以及成功标准清晰无误的提示。
cs.CL / 65 / 2609.22255
Deep Persona: A Psychologically Grounded Architecture and Evaluation Framework for Role-Playing Agents and Simulations
深度角色：一种基于心理学的角色扮演代理和模拟的架构与评估框架
Abstract
Existing approaches to persona simulation with Large Language Models (LLMs) mostly rely on shallow character descriptions that fail to sustain coherent character behavior across extended interactions. We introduce Deep Persona, a psychologically grounded, three-layered architecture that organizes personas into hierarchical levels of observable expression, latent beliefs, and core motivational drives, for constructing highly convincing role-playing agents. Governed by the principles of scripted determinism and bounded agency, the architecture restricts the model to a reactive engine guided by a structured internal script. We further propose a reference-free evaluation framework that benchmarks dialogue naturalness against empirical human distributions using established psychological clinical instruments and adversarial stress-tests. Empirical evaluation reveals that while LLMs achieve high pragmatic fluency, they exhibit systematic limitations in emotional expression and joint attention. In addition, we present a case study of two Deep Personas and evaluate them using the proposed framework, demonstrating that structured personas can produce interactions that more closely align with human conversational behavior.
Chinese Translation
现有的基于大型语言模型（LLMs）的角色模拟方法大多依赖于肤浅的角色描述，这些描述无法在长时间的互动中维持一致的角色行为。我们提出了深度角色（Deep Persona），一种基于心理学的三层架构，将角色组织为可观察表达、潜在信念和核心动机驱动的层级，以构建高度可信的角色扮演代理。该架构遵循脚本决定论和有限代理的原则，限制模型为一个由结构化内部脚本引导的反应引擎。我们进一步提出了一种无参考的评估框架，该框架使用已建立的心理学临床工具和对抗性压力测试，将对话自然性与经验性人类分布进行基准比较。实证评估表明，尽管LLMs在实用流畅性方面表现出色，但在情感表达和共同注意力方面存在系统性局限。此外，我们展示了两个深度角色的案例研究，并使用所提出的框架对其进行评估，证明结构化角色能够产生更接近人类对话行为的互动。
cs.CL / 66 / 2609.22256
DIPLOMAT: Dialogue-Span-Aware Direct Preference Optimization for Polite Persuasive Workplace Negotiation Dialogues
DIPLOMAT：面向礼貌劝说的职场谈判对话的对话跨度感知直接偏好优化
Abstract
Effective workplace negotiation requires balancing multiple objectives, including achieving task goals, preserving professional relationships, and resolving conflicts constructively. However, misunderstandings, misaligned preferences, and interpersonal friction often impede successful outcomes. Politeness mitigates these challenges by fostering trust, reducing tension, and preventing escalation, and persuasive communication helps overcome resistance, align preferences, and guide participants toward mutually beneficial agreements. Motivated by these insights, we present DIPLOMAT, a dialogue system for polite and persuasive workplace negotiation. To support its development, we introduce PROWESS, a dataset of multi-turn workplace negotiation dialogues generated via a multi-agent framework and enriched withnegotiation strategies, politeness levels, persuasive strategies. DIPLOMAT is trained using Dialogue-Span-Aware Direct Preference Optimization (DSA-DPO), a novel preference learning objective that identifies key dialogue spans for preference alignment. This enables DIPLOMAT to generate contextually coherent responses that employ intended negotiation strategies, maintain politeness, and incorporate effective persuasion strategies throughout interactions. Automatic and human evaluation on PROWESS confirm that DIPLOMAT consistently outperforms baselines in generating coherent, polite, and persuasive negotiation responses.
Chinese Translation
有效的职场谈判需要平衡多个目标，包括实现任务目标、维护专业关系和以建设性的方式解决冲突。然而，误解、偏好不一致和人际摩擦常常阻碍成功的结果。礼貌通过促进信任、减少紧张和防止局势升级来缓解这些挑战，而劝说性沟通则有助于克服抵抗、对齐偏好并引导参与者达成互利的协议。基于这些见解，我们提出了DIPLOMAT，一个用于礼貌和劝说职场谈判的对话系统。为了支持其开发，我们引入了PROWESS，一个通过多智能体框架生成的多轮职场谈判对话数据集，并丰富了谈判策略、礼貌水平和劝说策略。DIPLOMAT使用对话跨度感知直接偏好优化（Dialogue-Span-Aware Direct Preference Optimization, DSA-DPO）进行训练，这是一种新颖的偏好学习目标，旨在识别关键的对话跨度以实现偏好对齐。这使得DIPLOMAT能够生成上下文连贯的响应，采用预期的谈判策略，保持礼貌，并在互动中贯穿有效的劝说策略。在PROWESS上的自动评估和人工评估确认，DIPLOMAT在生成连贯、礼貌和劝说性的谈判响应方面始终优于基线。
cs.CL / 67 / 2609.22259
Which Part of the Context Layer Does the Work? Separating Semantic Content from Retrieval Scaffolding in Text-to-SQL Agents
上下文层的哪个部分发挥作用？在文本到SQL代理中分离语义内容与检索支架
Abstract
Context layers, curated documentation that an analytics agent fetches at query time, produce large accuracy gains on text-to-SQL benchmarks. A with/without comparison cannot say which part of the layer does the work: the semantic content, the retrieval scaffolding that delivers it, or the pre-computed views that usually accompany it. We report a four-arm ablation on DABStep on four models that separates the three. The instrument is a data contract: a YAML artifact that carries a domain's semantics and the rules an agent's tools enforce. One arm empties every field of prose in the frozen contract while holding the tool surface, retrieval instruction, table allow-list and operation rules byte-for-byte fixed. Compiling the contract's own SQL expressions into views gives the ceiling a pre-computed layer would reach: gold on all 176 tasks it covers. Content dominates. It raises hard-task accuracy from 13.9% to 55.1%, 22.6% to 56.6%, 22.9% to 68.4% and 37.0% to 77.4%, beating the same knowledge pasted into the prompt on every model. Scaffolding without content is worth 0 to 5 points on two flash models and 14 to 15 on two frontier models. Against the compiled ceiling the contract arm's shortfall is a failure to derive, and it falls from 39 points to 5 with model capability. The contract beats the prompt because the rule it needs is one lookup away rather than buried in a long prompt: its SQL carries the fee semantics up to 98% of the time against the prompt arm's 4%, and at the lowest cost per correct answer on three of four models. For practitioners: semantics first, scaffolding second, pre-computed macros only where an agent demonstrably fails to derive. Ungoverned arms submitted 166 mutating statements; governed arms none. The gain is confined to the contract's domain. On one model the benchmark's own withheld golds grade the contract arm at 51.9% against 18.8% for the prompt baseline.
Chinese Translation
上下文层是分析代理在查询时获取的策划文档，在文本到SQL基准测试中产生了显著的准确性提升。通过有/无比较无法确定层的哪个部分发挥了作用：是语义内容、提供内容的检索支架，还是通常伴随其后的预计算视图。我们在DABStep上对四个模型进行了四个分支的消融实验，以分离这三者。实验工具是数据合同：一种YAML文档，承载领域的语义和代理工具执行的规则。其中一个分支将冻结合同中的每个文本字段清空，同时保持工具表面、检索指令、表允许列表和操作规则逐字固定。将合同自身的SQL表达式编译为视图，给出了预计算层所能达到的上限：在其覆盖的176个任务中获得金标准。内容占主导地位。它将困难任务的准确率从13.9%提升至55.1%，从22.6%提升至56.6%，从22.9%提升至68.4%，以及从37.0%提升至77.4%，超越了在每个模型中粘贴的相同知识。没有内容的支架在两个快速模型上得分为0到5分，在两个前沿模型上得分为14到15分。与编译的上限相比，合同分支的不足在于未能推导，其得分从39分降至5分，随着模型能力的提升而降低。合同优于提示，因为所需的规则只需一次查找，而不是埋藏在冗长的提示中：其SQL在对比提示分支时，98%的时间携带费用语义，而提示分支仅为4%，并且在四个模型中的三个上以最低的正确答案成本获得。对于从业者：语义优先，支架其次，预计算宏仅在代理明显无法推导的情况下使用。无管控的分支提交了166条变异语句；有管控的分支则没有。增益仅限于合同的领域。在一个模型上，基准测试中被扣留的金标准将合同分支的评分定为51.9%，而提示基线为18.8%。
cs.CL / 68 / 2609.22261
Used, Mentioned, or Condemned? A Controlled Contrast-Set Diagnostic for the Use-Mention Distinction in Code-Mixed Hinglish Misogyny Detection
使用、提及还是谴责？用于代码混合的 Hinglish 性别歧视检测中的使用-提及区分的受控对比集诊断
Abstract
Lexicon-driven misogyny detectors cannot, by construction, distinguish a slur used against a woman from the same slur mentioned in counter-speech ("don't call her that") -- yet exactly this distinction governs whether moderation protects or silences the people discussing abuse. We study this problem in code-mixed Hinglish and make three contributions. First, we diagnose two evaluation artifacts on a publicly available redacted corpus: category-encoding anonymization placeholders leak the label (a no-learning rule scores 1.000), and even after they are neutralized misogynistic and benign comments occupy lexically disjoint registers, so bag-of-words reaches macro-F1 approximately 1.00 under random cross-validation but collapses under template-disjoint evaluation. Second, we release Hinglish-MGY-Diag, a deterministic generator and a 416-item / 163-minimal-pair contrast-set diagnostic across five linguistically motivated categories in which slur presence and gendered register are decorrelated from the label by construction. Third, we introduce a strict pair-consistency metric that credits a model only when both members of a minimal pair are correctly labelled. Five from-scratch classical baselines evaluated under construction-disjoint five-fold cross-validation reveal that the strongest model reaches 0.93 accuracy on the cleanest use-mention subset but only 0.82 consistency -- it still mislabels roughly one counter-speech pair in five. A frontier LLM used as an author-model ceiling attains 1.000 on all metrics, doubling as independent label validation and confirming the benchmark is a capability gradient rather than an adversarial wall. We release all code, data, the generator, and an arms-length LLM harness for reproducing every number.
Chinese Translation
基于词典的性别歧视检测器在构造上无法区分针对女性使用的侮辱性词汇与在反驳言论中提及的同一侮辱性词汇（“不要这样称呼她”）——而正是这一区分决定了内容审核是保护还是压制讨论虐待行为的人。我们在代码混合的 Hinglish 中研究这个问题，并做出三项贡献。首先，我们在一个公开可用的经过编辑的语料库中诊断出两个评估伪影：类别编码匿名化占位符泄露了标签（无学习规则得分为 1.000），即使在中和后，性别歧视和良性评论仍占据词汇上不相交的语域，因此在随机交叉验证下，词袋模型的宏观 F1 达到约 1.00，但在模板不相交的评估下崩溃。其次，我们发布了 Hinglish-MGY-Diag，这是一个确定性生成器和一个包含 416 项 / 163 对最小对比集的诊断，涵盖五个语言学动机的类别，其中侮辱性词汇的存在和性别语域与标签在构造上是解耦的。第三，我们引入了一个严格的对偶一致性度量，仅在最小对的两个成员都被正确标记时才给予模型积分。在构造不相交的五折交叉验证下评估的五个从零开始的经典基线显示，最强模型在最干净的使用-提及子集上达到 0.93 的准确率，但一致性仅为 0.82——它仍然错误标记大约五个反驳言论对中的一个。作为作者模型上限的前沿 LLM 在所有指标上达到 1.000，既作为独立标签验证，又确认基准是能力梯度而非对抗墙。我们发布所有代码、数据、生成器以及一个独立的 LLM 工具，以重现每一个数字。
cs.CL / 69 / 2609.22362
Functional Emotion Without Character: Large Language Models, Aristotelian Disposition, and the Limits of Behavioral Alignment
没有性格的功能性情感：大型语言模型、亚里士多德的倾向与行为对齐的局限性
Abstract
Debates about whether artificial systems can feel are often forced between two unsatisfactory positions: behavioral equivalence is treated as sufficient for emotion, or phenomenal consciousness is treated as a prerequisite that makes the question empirically inaccessible. This article develops a structural alternative. It models emotions as context-sensitive regions, trajectories and attractor dynamics in high-dimensional representational state spaces. Recent mechanistic interpretability findings support the existence of causally active emotion-concept representations in large language models, but they do not establish subjective feeling or full emotional agency. Assessed against published adequacy standards for representation in language models, intervention provides strong evidence of causal use, while full affective role integration, uniformity across subject domains and coherence remain only partially established; there is no direct analogue of accuracy. These mismatches expose the need for a standard of affective appropriateness, which an account of character must supply. Such an account requires three further conditions: regulatory embodiment that gives valence endogenous stakes, temporal continuity that allows affective episodes to accumulate into a history, and an integrated self-model that binds that history to persistent values. Aristotle's concepts of path\=e, hexis, mesot\=es and phron\=esis are translated into a state-space sketch in which practical wisdom includes competence in estimating normatively salient context, not merely acting on a context description already given. The framework reframes alignment as a problem of durable disposition rather than output conformity, and yields interventional tests with explicit control conditions.
Chinese Translation
关于人工系统是否能够感受的争论常常被迫在两个不令人满意的立场之间徘徊：行为等价被视为情感的充分条件，或者现象意识被视为使该问题在经验上无法接触的先决条件。本文提出了一种结构性替代方案。它将情感建模为高维表征状态空间中的上下文敏感区域、轨迹和吸引子动力学。最近的机械可解释性研究结果支持大型语言模型中因果活跃的情感概念表征的存在，但并未确立主观感受或完全的情感代理。根据已发布的语言模型表征的充分性标准进行评估，干预提供了因果使用的强有力证据，而完全的情感角色整合、跨主题领域的一致性和连贯性仍然仅部分建立；没有直接的准确性类比。这些不匹配揭示了对情感适当性的标准的需求，而这一标准必须由性格的解释提供。这样的解释需要三个进一步的条件：赋予内生利益的调节体现、允许情感事件积累成历史的时间连续性，以及将该历史与持久价值绑定的综合自我模型。亚里士多德的路径（pathē）、倾向（hexis）、中庸（mesotēs）和实践智慧（phronēsis）概念被转化为一个状态空间草图，其中实践智慧包括在估计规范性显著上下文方面的能力，而不仅仅是在已经给定的上下文描述上采取行动。该框架将对齐重新构架为持久倾向的问题，而不是输出一致性，并产生具有明确控制条件的干预测试。
cs.CL / 70 / 2609.22409
Contextual Causality with Large Language Models: A Survey
大型语言模型的上下文因果关系：一项综述
Abstract
Understanding contextual causality is critical for large language models (LLMs), as it enables them to accurately identify causal relations in specific situations and support more reliable decision-making. Despite its significance, a systematic exploration of contextual causality with LLMs is still lacking. To fill this gap, we present a comprehensive survey on this topic. In this survey, we first propose a taxonomy of contextual causality, consisting of semantic, intervention, and counterfactual causality, and characterize each category by its core causal question, required model capabilities, representative tasks, and practical uses in causality analysis. We then analyze existing studies and discuss their key limitations. Finally, we examine the gaps between current benchmarks and real-world needs and outline promising directions for future research. Our goal is to clarify the research landscape of contextual causality with LLMs, emphasize its importance, and highlight promising future directions.
Chinese Translation
理解上下文因果关系对于大型语言模型（LLMs）至关重要，因为它使得模型能够准确识别特定情境中的因果关系，并支持更可靠的决策。尽管其重要性显而易见，但关于LLMs的上下文因果关系的系统探索仍然缺乏。为填补这一空白，我们提供了该主题的全面综述。在本综述中，我们首先提出了上下文因果关系的分类法，包括语义因果、干预因果和反事实因果，并通过核心因果问题、所需模型能力、代表性任务和因果分析中的实际应用来描述每个类别。然后，我们分析了现有研究并讨论了其主要局限性。最后，我们审视了当前基准与现实需求之间的差距，并概述了未来研究的有希望方向。我们的目标是阐明LLMs上下文因果关系的研究格局，强调其重要性，并突出未来研究的有前景方向。
cs.CL / 71 / 2609.22452
Vox-Infinity: Benchmarking the Limits of Long-Context Spoken Language Models
Vox-Infinity：长上下文口语语言模型的极限基准测试
Abstract
Long-context understanding remains a fundamental challenge for large language models, as excessively long inputs often lead models to forget salient information. This issue is even more pronounced in the speech domain, where audio, as a low-compression modality, requires substantially more embeddings than text to preserve both semantic content and acoustic cues. To address this challenge, we introduce \textbf{Vox-Infinity}, the first benchmark specifically designed to evaluate long-context understanding in spoken language models. Vox-Infinity systematically extends audio history along two dimensions: turn count and turn duration. It covers a diverse range of representative scenarios with varying interaction structures and semantic complexity. Crucially, Vox-Infinity provides explicit answer-provenance annotations and organizes samples according to the amount of historical context required to resolve each query, enabling precise and length-aware evaluation. Extensive evaluations of seven representative spoken language models reveal a clear overall recency effect: models generally achieve higher accuracy when answer-supporting evidence is closer to the query, but struggle to retrieve and use evidence located farther back in the dialogue history. Cases and datasets are available at https://vox-infinity.github.io.
Chinese Translation
长上下文理解仍然是大型语言模型面临的一个基本挑战，因为过长的输入往往导致模型忘记重要信息。在语音领域，这一问题尤为明显，因为音频作为一种低压缩模态，需要比文本更多的嵌入，以保留语义内容和声学线索。为了解决这一挑战，我们推出了 extbf{Vox-Infinity}，这是第一个专门设计用于评估口语语言模型长上下文理解的基准。Vox-Infinity系统地从两个维度扩展音频历史：轮次计数和轮次持续时间。它涵盖了一系列多样化的代表性场景，具有不同的交互结构和语义复杂性。重要的是，Vox-Infinity提供了明确的答案来源注释，并根据解决每个查询所需的历史上下文量组织样本，从而实现精确和关注长度的评估。对七个代表性口语语言模型的广泛评估显示出明显的整体近期效应：模型通常在答案支持证据距离查询较近时取得更高的准确性，但在检索和使用位于对话历史较远处的证据时则表现不佳。案例和数据集可在https://vox-infinity.github.io获取。
cs.CL / 72 / 2609.22455
Apollo Restore: A Foundation LLM for Historical Greek Optimized for Fill-in-the-Middle Restoration of Ancient Greek Texts
阿波罗恢复：一个为古希腊文本填补中间空缺而优化的历史大型语言模型
Abstract
We present Apollo Restore, a 24-billion-parameter large language model for restoring lacunae---physical gaps---in fragmentary Ancient Greek texts. Fine-tuned from Mistral Small with a fill-in-the-middle objective, Apollo Restore reconstructs missing spans without requiring oracle knowledge of their length. To our knowledge, it is the first large-scale decoder model for historical Greek, and the first for any ancient Mediterranean language. Evaluated as in prior work, on short gaps of up to ten characters, Apollo Restore places the correct restoration among its top twenty candidates for 80.6%/54.6%/61.0% of documentary-papyrus, literary-papyrus, and stone-inscription lacunae, exceeding the strongest published models by $1.6\times$/$2.6\times$/$1.4\times$. Prior evaluation protocols, however, inflate scores through a bias toward trivially short gaps; under a length-balanced metric Apollo Restore's advantage over the strongest published models grows to $2.3\times$/$3.5\times$/$1.6\times$ and degrades gracefully, even given incorrect length hints. In a blind study, 20 expert papyrologists, epigraphists, and philologists strongly preferred Apollo Restore to the strongest baseline and judged its performance at least as good as human restorations in 77% of cases. Apollo Restore also improves the published reading of P.Herc. 1667---a papyrus roll carbonised in the eruption of Vesuvius in 79 CE and digitally unrolled and edited after Apollo Restore's training data was compiled. Apollo Restore is an output of the Decoding Antiquity initiative to build specialized LLMs for historical languages and manuscripts, led by the Austrian Academy of Sciences.
Chinese Translation
我们提出了阿波罗恢复（Apollo Restore），这是一个拥有240亿参数的大型语言模型，用于恢复古希腊文本中的缺失部分——物理空缺。阿波罗恢复在填补中间目标的基础上，从Mistral Small进行微调，能够在不需要知道缺失部分长度的情况下重建缺失的文本片段。据我们所知，这是第一个针对历史希腊语的大规模解码器模型，也是第一个针对任何古地中海语言的模型。在先前的研究中进行的评估中，对于长度不超过十个字符的短缺口，阿波罗恢复在文献纸草文、文学纸草文和石刻缺口的前二十个候选中，正确恢复的比例分别为80.6%/54.6%/61.0%，超越了已发布的最强模型，提升幅度为$1.6 imes$/$2.6 imes$/$1.4 imes$。然而，先前的评估协议由于对显而易见的短缺口的偏见而夸大了分数；在长度平衡的指标下，阿波罗恢复相较于最强已发布模型的优势扩大至$2.3 imes$/$3.5 imes$/$1.6 imes$，即使在给出错误长度提示的情况下也表现出良好的降级。在一项盲测中，20位专家纸草学家、铭文学家和语言学家强烈偏好阿波罗恢复，相较于最强基线模型，他们认为其表现至少与人类恢复相当，比例达到77%。阿波罗恢复还改善了P.Herc. 1667的已发布阅读——这是一个在公元79年维苏威火山喷发中碳化的纸草卷，在阿波罗恢复的训练数据编制后进行了数字化展开和编辑。阿波罗恢复是“解码古代”（Decoding Antiquity）倡议的成果，该倡议旨在为历史语言和手稿构建专门的LLM，由奥地利科学院主导。
cs.CL / 73 / 2609.22463
Toward Personalized Sleep Guidance from Wearable Data Using Language Models
基于可穿戴数据的个性化睡眠指导：语言模型的应用
Abstract
Sleep monitoring using wearable data has shown promise for personal health, yet large language model (LLM)-based summarization and question answering remain insufficient for personalized sleep guidance. Training specialized models, however, often requires costly expert annotation. Moreover, privacy and accessibility concerns motivate lightweight, local deployment for end users. We present a two-stage framework to address these challenges. Specifically, in Stage~1, a multi-agent LLM pipeline reasons structured sleep guidance from unannotated wearable records, enabling scalable dataset construction. Stage~2 distills guidance reasoning trajectories into small language models (SLMs) through supervised fine-tuning and integrates a training-free Best-of-$N$ selection strategy to enhance inference. Experimental results demonstrate our method outperforms commercial general and medical LLMs and open-source models. Human evaluation further supports the quality of the generated guidance and the feasibility of personalized sleep guidance with SLMs.
Chinese Translation
使用可穿戴数据进行睡眠监测在个人健康方面显示出良好的前景，但基于大型语言模型（LLM）的摘要和问答仍不足以提供个性化的睡眠指导。然而，训练专门的模型通常需要昂贵的专家标注。此外，隐私和可及性问题促使轻量级、本地化的用户端部署。我们提出了一个两阶段框架来解决这些挑战。具体而言，在第一阶段，一个多智能体LLM管道从未标注的可穿戴记录中推理出结构化的睡眠指导，从而实现可扩展的数据集构建。第二阶段通过监督微调将指导推理轨迹提炼为小型语言模型（SLMs），并整合无训练的最佳选择策略以增强推理效果。实验结果表明，我们的方法优于商业通用和医学LLM以及开源模型。人类评估进一步支持生成指导的质量以及使用SLMs进行个性化睡眠指导的可行性。
cs.CL / 74 / 2609.22494
CultureMINE: Datasets and Methods for Improving the Cultural Capabilities of NLP Systems
CultureMINE：提升自然语言处理系统文化能力的数据集和方法
Abstract
In recent years, there has been a surge of interest in Cultural NLP, with substantial efforts to create globally inclusive NLP systems. The rapid growth of literature in this field makes it difficult to track trends in methods and data resources. To address this, we analyze over 375 papers to answer three complementary questions: (1) What Cultural Capabilities (CCs) are being targeted in NLP systems? (2) How are cultural data resources being created? and (3) What methods are being used to improve the CCs of those systems? We discuss trends observed across the three questions, and identify relevant research gaps. To facilitate further research in this field, we release our full list of analyzed papers in the form of an interactive web interface, which includes a feature to allow researchers to add their work; we hope this facilitates future research and proves to be a valuable resource for the Cultural NLP community.
Chinese Translation
近年来，文化自然语言处理（Cultural NLP）引起了广泛关注，许多努力致力于创建全球包容性的自然语言处理系统。该领域文献的快速增长使得跟踪方法和数据资源的趋势变得困难。为了解决这一问题，我们分析了超过375篇论文，以回答三个互补的问题：（1）自然语言处理系统针对哪些文化能力（Cultural Capabilities, CCs）？（2）文化数据资源是如何创建的？（3）哪些方法被用来提升这些系统的文化能力？我们讨论了在这三个问题中观察到的趋势，并识别出相关的研究空白。为了促进该领域的进一步研究，我们以互动网页界面的形式发布了我们分析的论文完整列表，其中包括允许研究人员添加其工作的功能；我们希望这能促进未来的研究，并成为文化自然语言处理社区的宝贵资源。
cs.CL / 75 / 2609.22522
When Cosine Similarity Fails to Reflect Linearly Accessible Structure in Dialogue Models
余弦相似度何时无法反映对话模型中的线性可达结构
Abstract
Cosine similarity is widely used to analyze transformer representations, implicitly assuming that similarity reflects task-relevant structure. We study when this assumption fails in dialogue-conditioned large language models. Across three 7-8B chat-tuned models, ambient cosine similarity substantially underestimates linearly decodable persona structure on the same hidden states; numerically, linear probe AUC is in the 0.73-0.97 range while cosine kNN is in the 0.56-0.77 range on a 30-class task. A low-dimensional supervised subspace recovers much of this gap, whereas a matched-rank PCA subspace does not and in some cases degrades performance. This mismatch is regime-dependent: it is absent in single-sentence sentiment classification (SST-5), and a matched-cardinality control rules out attribute cardinality as a confound. The gap does not systematically increase across dialogue turns, and the task-aligned subspace remains stable over time. However, two of three models violate a pre-registered within-subspace separability invariance criterion (|Delta AUC| <= 0.03), and one model violates a pre-registered turn-invariance criterion (|Delta L| <= 0.05). These results show that cosine similarity can fail to reflect task-aligned structure in dialogue representations even when that structure is linearly accessible.
Chinese Translation
余弦相似度广泛用于分析变换器表示，隐含假设相似度反映与任务相关的结构。我们研究了在对话条件的大型语言模型中，这一假设何时失效。在三个7-8B的聊天调优模型中，环境余弦相似度显著低估了相同隐藏状态上的线性可解的人物结构；在数值上，线性探测器的AUC在0.73-0.97范围内，而余弦kNN在30类任务中则在0.56-0.77范围内。一个低维的监督子空间恢复了大部分差距，而一个匹配秩的主成分分析（PCA）子空间则没有，并且在某些情况下降低了性能。这种不匹配是依赖于不同的情境：在单句情感分类（SST-5）中不存在，并且一个匹配基数的控制排除了属性基数作为混淆因素。这个差距在对话轮次中并没有系统性增加，且与任务对齐的子空间在时间上保持稳定。然而，三个模型中的两个违反了预注册的子空间内可分离性不变性标准（|Delta AUC| <= 0.03），而一个模型违反了预注册的轮次不变性标准（|Delta L| <= 0.05）。这些结果表明，即使该结构是线性可达的，余弦相似度也可能无法反映对话表示中的任务对齐结构。
cs.CL / 76 / 2609.22536
Cross-Dialect NER for Bangla Regional Dialects Using Leave-One-Dialect-Out Cross-Validation and Explainable AI
基于留一方言交叉验证和可解释人工智能的孟加拉地区方言跨方言命名实体识别
Abstract
Bangla, the seventh most spoken language in the world, exhibits significant regional dialectal diversity, with dialects such as Barishal, Chattogram, Sylhet, Noakhali, and Mymensingh differing in lexical, morphological, and syntactic characteristics. These variations pose substantial challenges for Named Entity Recognition (NER), limiting the generalization of models trained on Standard Bangla or a single regional dialect. This paper presents a cross-dialect Bangla NER framework using the publicly available ANCHOLIK-NER dataset, comprising 17,405 annotated sentences and 101,817 tokens across five major Bangla regional dialects. A Leave-One-Dialect-Out Cross-Validation (LODOCV) strategy is adopted, training models on four dialects and evaluating on the remaining unseen dialect. Eight pretrained transformer-based models, including BanglaBERT, MuRIL, XLM-RoBERTa, and Multilingual-E5, are evaluated under identical experimental settings. Multilingual-E5 Large achieves the highest F1-score in every fold, peaking at 97.26% on Mymensingh and reaching its lowest, 82.38%, on Chattogram, the most challenging target dialect. To improve interpretability, Local Interpretable Model-agnostic Explanations (LIME) are applied to word-level predictions, revealing that the model's decisions are driven primarily by the surface form of the target entity word itself rather than by surrounding sentence context. These findings establish a benchmark for cross-dialect Bangla NER and demonstrate the effectiveness of transformer-based transfer learning for low-resource regional dialects, while highlighting the surface-form dependence of current models as a direction for future work.
Chinese Translation
孟加拉语是世界上第七大语言，展现出显著的地区方言多样性，包括巴里沙尔、恰特格拉姆、希尔赫特、诺阿卡利和迈门辛等方言在词汇、形态和句法特征上存在差异。这些差异对命名实体识别（NER）构成了重大挑战，限制了在标准孟加拉语或单一地区方言上训练的模型的泛化能力。本文提出了一种跨方言孟加拉NER框架，使用公开可用的ANCHOLIK-NER数据集，该数据集包含17,405个标注句子和101,817个标记，涵盖五种主要的孟加拉地区方言。采用留一方言交叉验证（LODOCV）策略，在四种方言上训练模型，并在剩余未见方言上进行评估。在相同实验设置下，评估了包括BanglaBERT、MuRIL、XLM-RoBERTa和Multilingual-E5在内的八种预训练的基于变换器的模型。Multilingual-E5 Large在每个折中都达到了最高的F1分数，在迈门辛方言上峰值达到97.26%，而在最具挑战性的目标方言恰特格拉姆上最低为82.38%。为了提高可解释性，应用局部可解释模型无关解释（LIME）于词级预测，揭示模型的决策主要受目标实体词本身的表面形式驱动，而非周围句子上下文。这些发现为跨方言孟加拉NER建立了基准，并展示了基于变换器的迁移学习在低资源地区方言中的有效性，同时强调了当前模型对表面形式的依赖，为未来的研究指明了方向。
cs.CL / 77 / 2609.22553
Correct Diagnosis, Better Feedback: A Symbolic-Verifier for Faithful LLM Tutoring Feedback in Logic Proofs
正确诊断，更好的反馈：用于逻辑证明的忠实大型语言模型辅导反馈的符号验证器
Abstract
Effective LLM tutoring depends on correctly identifying the specific error in a student's reasoning before generating feedback. We study this problem in propositional-logic proof tutoring, where student actions can be checked against formal inference rules. We introduce a verifier-grounded architecture that separates diagnosis from language generation. Using 600 balanced student actions, we compare a zero-shot LLM detector, a fine-tuned detector, and a symbolic verifier. Each diagnosis is processed by shared rationale and feedback agents, isolating the effect of the initial diagnosis. The zero-shot detector achieves a macro-F1 of 0.191; fine-tuning raises this to 0.709 but retains systematic errors between structurally related classes. Rationales generally preserve the diagnosis supplied to them, showing that an incorrect diagnosis can be faithfully propagated through the pipeline. Feedback can likewise remain faithful to its rationale, non-revealing, and pedagogically appropriate while addressing the wrong error. Verifier-grounded feedback achieves the highest diagnostic correctness, and expert ratings largely uneven with the automatic feedback evaluations. These findings show that apparent feedback quality can conceal upstream diagnostic errors and that faithfulness must be evaluated separately from correctness. Our code is publicly available
Chinese Translation
有效的大型语言模型辅导依赖于在生成反馈之前正确识别学生推理中的具体错误。我们研究了这一问题，聚焦于命题逻辑证明辅导，在该领域，学生的行为可以与正式推理规则进行核对。我们提出了一种基于验证器的架构，将诊断与语言生成分离。通过使用600个平衡的学生行为，我们比较了零样本大型语言模型检测器、微调检测器和符号验证器。每个诊断由共享的推理和反馈代理处理，从而隔离初始诊断的影响。零样本检测器的宏观F1值为0.191；微调后这一值提高至0.709，但在结构相关类之间仍然保留系统性错误。推理通常保留提供给它们的诊断，显示出不正确的诊断可以通过整个流程忠实地传播。反馈同样可以保持对其推理的忠实性，且不揭示错误，同时在处理错误时具有教学适宜性。基于验证器的反馈实现了最高的诊断正确性，专家评分与自动反馈评估之间存在较大不一致。这些发现表明，表面上的反馈质量可能掩盖上游的诊断错误，并且忠实性必须与正确性分开评估。我们的代码已公开可用。
cs.CL / 78 / 2609.22566
Do Student LLMs Inherit OOD Robustness? Invariance-Weighted Distillation for Reliable Knowledge Transfer
学生 LLM 是否继承了 OOD 鲁棒性？基于不变性加权的蒸馏方法实现可靠的知识转移
Abstract
Knowledge distillation (KD) aims to compress high-performance teacher LLMs into lightweight students. However, distilled students often exhibit substantial performance degradation in out-of-distribution (OOD) settings, a critical gap that remains underexplored. We identify two compounding mechanisms causing OOD performance degradation: (1) data spuriousness: students can learn spurious correlations in the distillation dataset over genuine causal relationships; and (2) teacher capability: standard KD treats all samples uniformly, ignoring whether the teacher is guided by causal features or misled by spurious shortcuts on a given sample. To address these challenges, we propose Invariance-Weighted Distillation (IWD), a theoretically grounded framework that dynamically reweights training samples using an estimate of the teacher's causal reliance derived from prediction invariance across multiple synthetic environments. IWD perturbs spurious cues while preserving core semantics, assigning higher distillation weights to samples whose teacher predictions remain invariant, indicating greater reliance on causal rather than spurious features. We theoretically show that IWD reduces the student's Spurious-to-Causal (S2C) gradient ratio compared to standard uniformly weighted KD, driving the student toward more invariant representations. Experiments on four NLP benchmarks (MNLI, SQuAD-v2, CoNLL-2003 NER, and SST-2) across two model families (DeBERTa-v3 and Qwen-2.5) demonstrate that IWD consistently outperforms strong KD baselines on OOD evaluations while maintaining competitive in-distribution (ID) performance. Specifically, IWD achieves the highest accuracy in 15 out of 16 OOD benchmarks and improves average OOD performance over standard KD by 4.34 percentage points on NLI and 14.94 percentage points on QA.
Chinese Translation
知识蒸馏（KD）旨在将高性能的教师 LLM 压缩为轻量级的学生模型。然而，蒸馏后的学生模型在分布外（OOD）环境中往往表现出显著的性能下降，这是一个尚未深入探讨的关键问题。我们识别出导致 OOD 性能下降的两个复合机制：（1）数据虚假性：学生模型可能在蒸馏数据集中学习到虚假的相关性，而非真实的因果关系；（2）教师能力：标准 KD 对所有样本进行统一处理，忽略了教师是否受到因果特征的指导或在特定样本上被虚假捷径误导。为了解决这些挑战，我们提出了不变性加权蒸馏（IWD），这是一个理论基础扎实的框架，通过利用从多个合成环境中预测不变性得出的教师因果依赖估计，动态地重新加权训练样本。IWD 在保留核心语义的同时扰动虚假线索，为那些教师预测保持不变的样本分配更高的蒸馏权重，表明对因果特征而非虚假特征的更大依赖。我们理论上证明，与标准的均匀加权 KD 相比，IWD 降低了学生的虚假到因果（S2C）梯度比，推动学生朝向更不变的表示。针对四个 NLP 基准（MNLI、SQuAD-v2、CoNLL-2003 NER 和 SST-2）在两个模型家族（DeBERTa-v3 和 Qwen-2.5）上的实验表明，IWD 在 OOD 评估中始终优于强大的 KD 基线，同时保持竞争力的分布内（ID）性能。具体而言，IWD 在 16 个 OOD 基准中的 15 个上达到了最高准确率，并在 NLI 上比标准 KD 提高了 4.34 个百分点，在 QA 上提高了 14.94 个百分点。
cs.CL / 79 / 2609.22603
Preserving What Matters: Semantic Scaffolds Beyond Saturation in Summarization Evaluation
保留重要信息：超越饱和的语义支架在摘要评估中的应用
Abstract
Summarization ships in countless production systems, making model selection a routine decision that depends on measuring summary quality. Existing metrics struggle to support this: ROUGE captures only surface overlap, while LLM-as-judge scores saturate to near-identical values that fail to rank models effectively. We observe this saturation across three public datasets, two proprietary datasets, and multilingual settings. Motivated by this, we introduce Semantic Scaffold, an evaluation framework that extracts a hierarchical representation of facts, questions, and entity attributes from a source text, labeling each as a main point or supporting detail, and reusing this structure as a fixed reference for scoring summaries. From this representation, we derive three diagnostic metrics: Fact Preservation Score (FPS), Question Preservation Score (QPS), and Entity Preservation Score (EPS), designed to reward the preservation of essential information while penalizing detail overload, and position them as interpretable diagnostics that remain informative where holistic axes collapse. Finally, we analyze four recurring failure modes of ROUGE and LLM-as-judge scores, demonstrating that scaffold-based evaluation remains informative where conventional metrics collapse.
Chinese Translation
摘要生成在无数生产系统中扮演着重要角色，因此模型选择成为一项常规决策，依赖于对摘要质量的测量。然而，现有的评估指标难以支持这一过程：ROUGE仅捕捉表面重叠，而LLM-as-judge的评分趋于饱和，导致近乎相同的值，无法有效区分模型。我们在三个公共数据集、两个专有数据集和多语言环境中观察到了这种饱和现象。基于此，我们提出了语义支架（Semantic Scaffold），一个评估框架，从源文本中提取事实、问题和实体属性的层次表示，将每个元素标记为主要观点或支持细节，并将这一结构作为固定参考用于评分摘要。基于这一表示，我们推导出三个诊断指标：事实保留分数（Fact Preservation Score, FPS）、问题保留分数（Question Preservation Score, QPS）和实体保留分数（Entity Preservation Score, EPS），旨在奖励重要信息的保留，同时惩罚细节过载，并将其定位为可解释的诊断工具，在整体评估轴崩溃时仍然保持信息量。最后，我们分析了ROUGE和LLM-as-judge评分的四种常见失效模式，证明基于支架的评估在传统指标失效时仍然具有信息价值。
cs.CL / 80 / 2609.22607
Pretrained Persona Mixture Models and Tandem Models for Human Simulation
预训练的人格混合模型与人类模拟的串联模型
Abstract
We argue here that the current dominant practice in LLM human simulation: prompting instruction-tuned assistant language models to role-play personas, is inaccurate and produces stereotyped predictions (lacking natural diversity). It has previously been shown that LLMs can be bound to personas using naturalistic, freetext dialog avoiding stereotyping. Here we show that binding can also be achieved using short, individual samples of dialog from specific people. Demographics can be added later without negative effects by simply querying the model. We use the term Persona Mixture Models (PMMs) for well-calibrated human models, currently realized as pretrained base models. We show that PMMs produce more accurate predictions than instruction-tuned models and retain more of the lexical, semantic, and pragmatic diversity found in human dialog. We measure realism and diversity of LLMs simulating human interlocutors across a diverse set of corpora spanning open-domain text, human-AI chat, and task-oriented dialogue between human speakers. However, base pretrained models can produce out-of-domain dialog and may lose some of the human's internal state over long contexts. We propose and explore tandem models which combine a pre-trained model with an instruction-tuned supervisor. Tandem models achieve the best overall accuracy and diversity in our experiments.
Chinese Translation
我们在此论证，目前在大规模语言模型（LLM）人类模拟中的主导实践，即通过提示指令调优的助手语言模型进行角色扮演的人格，是不准确的，并且产生了刻板印象的预测（缺乏自然多样性）。之前的研究表明，LLM可以通过自然的自由文本对话与人格绑定，从而避免刻板印象。在这里，我们展示了通过特定个体的简短对话样本也可以实现绑定。人口统计信息可以在后期添加，而不会产生负面影响，只需简单地查询模型。我们将经过良好校准的人类模型称为人格混合模型（Persona Mixture Models, PMMs），目前实现为预训练的基础模型。我们展示了PMMs比指令调优模型产生更准确的预测，并保留了人类对话中更多的词汇、语义和语用多样性。我们测量了LLM在模拟人类对话者时的现实性和多样性，涵盖了开放领域文本、人机聊天和人类发言者之间的任务导向对话等多种语料库。然而，基础预训练模型可能会产生超出领域的对话，并且在长上下文中可能会丧失一些人类的内在状态。我们提出并探索了串联模型，这种模型将预训练模型与指令调优的监督者结合在一起。串联模型在我们的实验中实现了最佳的整体准确性和多样性。
cs.CL / 81 / 2609.22633
Beetle: A Bilingual Model Suite for Modelling Second-Language Processing
Beetle：一种用于建模第二语言处理的双语模型套件
Abstract
Bilingual language models (LMs) offer a controlled setting for studying how training conditions shape second-language (L2) behaviour, but prior work typically varies exposure structure, scale, and architecture at once, making it difficult to attribute effects to any single factor. We introduce Beetle, a controlled language model pretraining framework in which tokeniser, target language, training budget, and exposure structure are each independently manipulable, enabling systematic and comparable experimentation of training conditions. Using Beetle, we train and release 285 bilingual and 45 monolingual open-source LMs with rich checkpoints across a range of exposure schedules, data scales and first languages (L1s) to study multilingual pretraining and computational modelling of bilingualism and second language learning. Evaluating models on human bilingual and second language reading-time prediction and grammaticality judgement tasks, we find that staged and temporally structured curricula consistently improve alignment with language learner reading time compared to balanced bilingual training, with the largest gains at smaller data scales and for typologically closer language pairs. The Beetle models are well suited tools to help move computational psycholinguistics beyond its prevailing monolingual, English-centric focus toward models of human bilingual processing, to study cross-lingual learning dynamics, while supporting community-based development of controlled model families.
Chinese Translation
双语语言模型（LMs）提供了一个受控的环境，用于研究训练条件如何影响第二语言（L2）行为，但以往的研究通常同时改变暴露结构、规模和架构，使得很难将效果归因于任何单一因素。我们介绍了Beetle，一个受控的语言模型预训练框架，在该框架中，分词器、目标语言、训练预算和暴露结构均可独立操控，从而实现对训练条件的系统性和可比实验。使用Beetle，我们训练并发布了285个双语和45个单语的开源语言模型，涵盖了丰富的检查点，涉及多种暴露计划、数据规模和第一语言（L1），以研究多语言预训练以及双语和第二语言学习的计算建模。在对人类双语和第二语言阅读时间预测及语法判断任务进行模型评估时，我们发现，与平衡双语训练相比，分阶段和时间结构化的课程在对语言学习者阅读时间的对齐上始终表现出改善，尤其是在较小的数据规模和类型上更接近的语言对中，提升幅度最大。Beetle模型是推动计算心理语言学超越其主流单语、以英语为中心的关注，朝着人类双语处理模型发展的良好工具，旨在研究跨语言学习动态，同时支持基于社区的受控模型家族的发展。
cs.CL / 82 / 2609.22697
COT-TTS: Audio Context-Aware Text-to-Speech with Chain-of-Thought Reasoning
COT-TTS：基于链式推理的音频上下文感知文本到语音转换
Abstract
Recently, text-to-speech systems have made significant progress in speech expressiveness and controllability. However, the speaking style of generated speech typically relies on clear user-specified instructions. In natural conversations, speaking style should be naturally inferred from the preceding conversational context. Therefore, we propose COT-TTS, a context-aware, reasoning-based text-to-speech task. Given historical conversation audio, target text, and a reference speech, the system should comprehend the conversational context, infer an explicit intermediate reasoning, and finally synthesize the target speech with the specified timbre. To support this task, we constructed a large-scale bilingual conversational speech dataset comprising 9 million training samples, including a high-quality subset of 1 million samples. We further constructed a source-disjoint benchmark with 800 human-verified samples and established strong task-specific baselines. Additionally, we developed end-to-end autoregressive models with parameter sizes of 0.6B and 1.7B, generating emotion-labeled transcripts, editable speech style inferences, and speech tokens. Experimental results show that the proposed model achieves performance comparable to large-scale baseline systems with significantly fewer parameters. At the same time, the model performs well in terms of duration consistency and emotional consistency, and can generate appropriate emotional, stress, and rhythmic variations based on the conversational context. To facilitate future research, we will publicly release the data construction pipeline, dataset, trained models, and related resources. The demo page and additional resources are available at https://luckybian.github.io/COT-TTS
Chinese Translation
近年来，文本到语音系统在语音表现力和可控性方面取得了显著进展。然而，生成语音的说话风格通常依赖于用户明确指定的指令。在自然对话中，说话风格应当自然地从前面的对话上下文中推断出来。因此，我们提出了COT-TTS，一种上下文感知的基于推理的文本到语音转换任务。给定历史对话音频、目标文本和参考语音，系统应理解对话上下文，推断出明确的中间推理，最终合成具有指定音色的目标语音。为了支持这一任务，我们构建了一个大规模的双语对话语音数据集，包括900万个训练样本，其中包含100万个高质量子集。我们进一步构建了一个源不重叠的基准数据集，包含800个经过人工验证的样本，并建立了强大的任务特定基线。此外，我们开发了参数规模为0.6B和1.7B的端到端自回归模型，生成情感标记的转录文本、可编辑的语音风格推断和语音标记。实验结果表明，所提出的模型在性能上与大规模基线系统相当，但参数显著更少。同时，该模型在持续时间一致性和情感一致性方面表现良好，并能够根据对话上下文生成适当的情感、重音和节奏变化。为了促进未来的研究，我们将公开发布数据构建流程、数据集、训练模型和相关资源。演示页面和其他资源可在 https://luckybian.github.io/COT-TTS 获取。
cs.CL / 83 / 2609.22700
LLaDA-PRM: A Bidirectional Step-Level Reasoning Evaluator
LLaDA-PRM：一种双向逐步推理评估器
Abstract
Step-level reasoning evaluators are commonly based on autoregressive language models, whose causal attention restricts each step representation to the problem, previous steps, and the current step. Yet, when the complete solution is available, the validity of an earlier step may become clearer only through its downstream consequences. We validate this hypothesis through a controlled 54-run comparison of causal and bidirectional LLaDA evaluators at 1B--3B scale, changing only the self-attention mask, and find bidirectional attention yields consistent improvements. Building on this finding, we introduce \prm{}, an 8B bidirectional evaluator that reaches 88.8 step-level F1 on MR-MATH-invalid and 83.8 on the out-of-distribution MR-GSM8K original-question subset, outperforming ReasonEval-Llemma-34B by 11.3 and 10.3 F1 points, respectively. \prm{} also remains effective when evaluating incomplete reasoning traces in online settings, outperforming the strongest baselines on both benchmarks by a large margin. We further show that \prm{} provides an effective training-data selection signal, improving Mistral-7B performance on MATH-500.
Chinese Translation
逐步推理评估器通常基于自回归语言模型，其因果注意力限制了每一步的表示仅与问题、前一步和当前步骤相关。然而，当完整的解决方案可用时，早期步骤的有效性可能仅通过其下游后果变得更加清晰。我们通过对1B-3B规模的因果和双向LLaDA评估器进行54次受控比较来验证这一假设，仅改变自注意力掩码，发现双向注意力带来了持续的改进。在此基础上，我们引入了 extit{prm}，一个8B的双向评估器，在MR-MATH-invalid上达到了88.8的逐步F1分数，在分布外的MR-GSM8K原始问题子集上达到了83.8，分别比ReasonEval-Llemma-34B高出11.3和10.3个F1点。 extit{prm}在在线环境中评估不完整推理轨迹时仍然有效，在两个基准测试中均大幅超越最强基线。我们进一步表明， extit{prm}提供了有效的训练数据选择信号，提高了Mistral-7B在MATH-500上的表现。
cs.CL / 84 / 2609.22705
Analyzing Public Discourse on Urbanism: Topic Clustering, Sentiment Analysis and Retrieval-Augmented Generation using YouTube Comments
城市主义公共话语分析：基于YouTube评论的主题聚类、情感分析与检索增强生成
Abstract
Online discourse about urban issues - walkability, cycling infrastructure, public transit, housing density, and street safety - is voluminous but unstructured, and existing city-evaluation tools capture none of it. We present a pipeline and conversational system that combines geographic entity resolution, topic modeling, sentiment analysis, and Retrieval-Augmented Generation (RAG) over 22,788 chunks of YouTube transcripts and comments spanning 309 North American cities. Beyond the system itself, our contribution is a set of measurements about what happens when standard NLP components meet short, informal, geographically ambiguous text. A Twitter-tuned RoBERTa classifier outperforms a VADER lexicon baseline by 12.6 macro-F1 points (0.589 vs. 0.464; McNemar p = 0.0001), but both models collapse on the neutral class, which dominates urbanist comment traffic; annotators disagree on the same class (Cohen's kappa = 0.53). Dense retrieval beats a TF-IDF baseline at every cutoff (P@5 0.790 vs. 0.560), and video-level relevance proxies understate chunk-level precision by a wide margin (0.660 vs. 0.94 under human rating). For groundedness evaluation, we find BERTScore unusable when a multi-sentence generated summary is compared against a single short comment - scores are nearly flat regardless of relevance - and show that ROUGE-1-based groundedness is a paraphrase-driven lower bound rather than a hallucination rate. These findings generalize beyond the urbanist domain to any RAG system built over short user-generated documents.
Chinese Translation
关于城市问题的在线话语——步行性、骑行基础设施、公共交通、住房密度和街道安全——数量庞大但缺乏结构，现有的城市评估工具无法捕捉到这些内容。我们提出了一种管道和对话系统，结合了地理实体解析、主题建模、情感分析和检索增强生成（Retrieval-Augmented Generation, RAG），对来自309个北美城市的22,788段YouTube转录文本和评论进行分析。除了系统本身，我们的贡献在于一组关于标准自然语言处理（NLP）组件与短小、非正式、地理模糊文本相遇时所发生情况的测量。经过Twitter调优的RoBERTa分类器在宏观F1得分上超越了VADER词典基线12.6分（0.589对0.464；McNemar p = 0.0001），但两个模型在中性类别上均表现不佳，而这一类别在城市主义评论流量中占主导地位；标注者在同一类别上存在分歧（Cohen's kappa = 0.53）。在每个截止点上，密集检索的表现优于TF-IDF基线（P@5 0.790对0.560），而视频级相关性代理在很大程度上低估了块级精度（0.660对0.94，依据人工评分）。在基础性评估中，我们发现当多句生成摘要与单条短评论进行比较时，BERTScore不可用——无论相关性如何，得分几乎保持平稳——并且显示基于ROUGE-1的基础性评估是一个由意译驱动的下限，而非幻觉率。这些发现超越了城市主义领域，适用于任何基于短用户生成文档构建的RAG系统。
cs.CL / 85 / 2609.22734
Clinical Domain Classification from Medical Transcriptions
基于医疗转录的临床领域分类
Abstract
Clinical domain classification plays an important role in organizing and analyzing large volumes of unstructured medical text. However, medical transcription datasets are often highly imbalanced, which can substantially degrade classification performance, particularly for underrepresented clinical specialties. In this work, we present a comparative study of machine learning and transformer-based approaches for clinical domain classification from medical transcriptions. We evaluate six traditional machine learning classifiers---Naive Bayes, Support Vector Machine (SVM), Decision Tree, Random Forest, K-Nearest Neighbors (KNN), and XGBoost---along with two pretrained transformer models, BERT and XLNet, and a few-shot large language model prompting approach. Experiments are conducted on medical transcription data collected from MTSamples, comprising 5,013 samples across 40 clinical specialties. To address severe class imbalance, we investigate two balancing strategies: text augmentation using NLP-based synonym replacement and Synthetic Minority Over-sampling Technique (SMOTE). Experimental results demonstrate that data balancing substantially improves classification performance across the evaluated models. In particular, BERT achieves the highest F1-score of 0.996 on the SMOTE-balanced dataset while requiring lower training time than XLNet. The results highlight the effectiveness of transformer-based representations combined with appropriate data balancing strategies for clinical domain classification and provide a systematic comparison of classical machine learning, transformer models, and few-shot prompting for medical transcription analysis.
Chinese Translation
临床领域分类在组织和分析大量非结构化医疗文本中发挥着重要作用。然而，医疗转录数据集往往存在高度不平衡的问题，这可能显著降低分类性能，特别是对于代表性不足的临床专业。本文呈现了机器学习和基于变换器的方法在医疗转录中的临床领域分类的比较研究。我们评估了六种传统机器学习分类器——朴素贝叶斯、支持向量机（SVM）、决策树、随机森林、K-近邻（KNN）和XGBoost，以及两个预训练的变换器模型BERT和XLNet，以及一种少量样本的大型语言模型提示方法。实验在从MTSamples收集的医疗转录数据上进行，数据包含来自40个临床专业的5,013个样本。为了解决严重的类别不平衡问题，我们研究了两种平衡策略：基于自然语言处理的同义词替换文本增强和合成少数类过采样技术（SMOTE）。实验结果表明，数据平衡显著提高了所评估模型的分类性能。特别是，BERT在SMOTE平衡数据集上达到了最高的F1-score为0.996，同时所需的训练时间低于XLNet。结果突显了结合适当数据平衡策略的基于变换器的表示在临床领域分类中的有效性，并提供了经典机器学习、变换器模型和少量样本提示在医疗转录分析中的系统比较。
cs.CL / 86 / 2609.22767
Beyond Final-Token Classification: Heterogeneous Readouts for Evidence-Grounded Suicide Risk Detection
超越最终标记分类：基于证据的自杀风险检测的异构读出
Abstract
The IEEE BigData Cup benchmark combines three prediction problems with different output structures: ordinal suicide-risk classification, multi-label psychosocial factor detection, and extraction of supporting phrases. We introduce heterogeneous readout decomposition (HRD), which separates semantic verification from output realization. A locally deployed Qwen3.8-27B model, adapted with task-specific QLoRA adapters, produces both answer-token margins and layer-63 answer states for card-conditioned queries. HRD compares four latent scores for ordinal risk, retains token margins for most factors while routing seven labels through one shared latent probe, and constructs evidence sets from verbatim span candidates with calibrated, risk-conditional constraints. On two held-out user-grouped confirmation folds, the latent risk readout improves weighted F1 from 0.8237 to 0.8372 and macro F1 from 0.7965 to 0.8185. Selective factor routing improves macro F1 by 0.0105 and tail-label macro F1 by 0.0189; in contrast, global latent replacement and independent label-specific probes fail. The constrained evidence decoder raises pooled phrase F1 from 0.7488 to 0.7609 in row-level out-of-fold evaluation. Lenormand's best public result is 0.8052 on Subtask 1 and 0.6636 on Subtask 2, giving a composite score of 0.7627. These results identify the answer-token boundary, rather than semantic representation alone, as a measurable source of error in this benchmark.
Chinese Translation
IEEE BigData Cup基准结合了三种具有不同输出结构的预测问题：序数自杀风险分类、多标签心理社会因素检测和支持短语的提取。我们引入了异构读出分解（Heterogeneous Readout Decomposition, HRD），该方法将语义验证与输出实现分离。一个本地部署的Qwen3.8-27B模型，采用特定任务的QLoRA适配器，能够为条件查询生成答案标记边际和第63层答案状态。HRD比较了四个潜在分数以评估序数风险，保留了大多数因素的标记边际，同时通过一个共享的潜在探针路由七个标签，并从逐字跨度候选中构建证据集，附加经过校准的风险条件约束。在两个保留的用户分组确认折中，潜在风险读出将加权F1从0.8237提高到0.8372，宏观F1从0.7965提高到0.8185。选择性因素路由将宏观F1提高了0.0105，尾标签宏观F1提高了0.0189；相比之下，全球潜在替换和独立标签特定探针未能取得效果。约束证据解码器在行级折外评估中将汇总短语F1从0.7488提高到0.7609。Lenormand在子任务1上的最佳公开结果为0.8052，在子任务2上的结果为0.6636，综合得分为0.7627。这些结果表明，答案标记边界，而非单纯的语义表示，是该基准中可测量的误差来源。
cs.CL / 87 / 2609.22774
NLPCC 2026 Task 10: Citation-Level Faithfulness Verification with DeBERTa Ensembles and Class-Wise Calibration
NLPCC 2026 任务 10：基于 DeBERTa 集成和类别校准的引用级忠实性验证
Abstract
This paper presents our system for Track 2 of the NLPCC 2026 Shared Task 10 on citation-level faithfulness in AI-assisted scientific reporting. Given an atomic scientific claim and the structured full text of its cited paper, the task requires both a four-way relation label and up to three evidence paragraph identifiers. The label head ensembles a paragraph-aware cross-encoder with a document-level DeBERTa-large classifier, followed by class-wise decision calibration. Probability-level fusion is motivated by an out-of-fold tendency to over-predict Topical Match. The evidence head combines paragraph scores from top-20 and top-30 joint models with BM25 scores. The system runs fully offline without external retrieval or LLM prompting. On the final leaderboard, our system achieved 82.9898 overall (89.5491 Macro-F1 and 76.4305 Joint@3), ranking second in Track 2. Ablations and error analysis show that model complementarity and calibration drive the label gains. Gold-evidence inference changes label Macro-F1 negligibly, whereas evidence ranking remains important for Joint@3.
Chinese Translation
本文介绍了我们在 NLPCC 2026 共享任务 10 中关于 AI 辅助科学报告的引用级忠实性验证的系统。给定一个原子科学主张及其引用论文的结构化全文，该任务要求提供一个四类关系标签和最多三个证据段落标识符。标签头结合了一个段落感知的交叉编码器与一个文档级的 DeBERTa-large 分类器，并随后进行类别决策校准。概率级融合是基于对主题匹配的过度预测的外折倾向。证据头结合了来自前 20 和前 30 的联合模型的段落得分与 BM25 得分。该系统完全离线运行，不依赖外部检索或 LLM 提示。在最终排行榜上，我们的系统整体得分为 82.9898（宏观 F1 为 89.5491，Joint@3 为 76.4305），在任务 2 中排名第二。消融实验和错误分析表明，模型的互补性和校准推动了标签的提升。黄金证据推断对标签的宏观 F1 影响微乎其微，而证据排名对 Joint@3 仍然重要。
cs.CL / 88 / 2609.22778
MIS-Bench: Benchmarking Multimodal LLMs for Psychotherapeutic Interpersonal Skills Assessment
MIS-Bench：多模态大型语言模型在心理治疗人际技能评估中的基准测试
Abstract
Multimodal large language models (MLLMs) are increasingly used as evaluators, yet their reliability in professional assessment tasks that require expert judgment remains unclear. We investigate this challenge in the context of assessing psychotherapeutic interpersonal skills and introduce MIS-Bench, a Multimodal Interpersonal Skills (MIS) benchmark comprising 996 psychotherapy response videos annotated across 8 dimensions of Facilitative Interpersonal Skills. Across 9 MLLMs with multiple modality and prompting settings, we find that current models show only modest agreement with human experts, inconsistent gains from multimodal input, and limited benefits from reasoning-based prompting. To mitigate this gap, we propose MIS-RAFT, a regression-aware fine-tuning method inspired by RAFT and tailored to fine-grained interpersonal skill scoring at one-decimal precision. MIS-RAFT addresses the mismatch between autoregressive token prediction and scalar-valued expert assessment, significantly improving agreement with human ratings. Overall, MIS-Bench reveals a clear gap between general multimodal capability and expert-level interpersonal judgment, while MIS-RAFT offers a promising path toward more reliable model-based assessment.
Chinese Translation
多模态大型语言模型（MLLMs）越来越多地被用作评估工具，但它们在需要专家判断的专业评估任务中的可靠性仍不明确。我们在评估心理治疗人际技能的背景下研究了这一挑战，并引入了MIS-Bench，一个包含996个心理治疗反应视频的多模态人际技能（MIS）基准，这些视频在促进性人际技能的8个维度上进行了注释。在9个具有多种模态和提示设置的MLLMs中，我们发现当前模型与人类专家之间的协议仅为适度，来自多模态输入的不一致增益以及基于推理的提示所带来的有限益处。为了缩小这一差距，我们提出了MIS-RAFT，一种受RAFT启发的回归感知微调方法，旨在以一位小数精度进行细粒度人际技能评分。MIS-RAFT解决了自回归令牌预测与标量值专家评估之间的不匹配，显著提高了与人类评分的一致性。总体而言，MIS-Bench揭示了通用多模态能力与专家级人际判断之间的明显差距，而MIS-RAFT则为更可靠的模型基于评估提供了一个有前景的路径。
cs.CL / 89 / 2609.22793
Diagnose, Then Repair: A Two-Stage MQM-Guided Post-Editing Framework for Domain-Specific Machine Translation
诊断后修复：一种基于MQM指导的领域特定机器翻译的两阶段后编辑框架
Abstract
LLM-based machine translation evaluation can closely match human judgments, but in practice it remains largely diagnostic, with the signals rarely translating into direct quality improvements under real production constraints. We propose a two-stage, evaluator-guided automatic post-editing framework that turns MQM-style evaluation into targeted repairs: a retrieval-augmented LLM evaluator outputs structured, span-level MQM diagnoses under an explicit edit contract, and a separate LLM post-editor applies minimal edits restricted to those diagnoses. This separation improves controllability and reduces paraphrastic drift compared to one-stage "judge-and-refine" baselines. In a systematic study involving seven LLMs spanning three model providers and seven languages, our best configuration consistently improves both COMET-22 and COMETKiwi scores over one-stage post-edit methods, while the evaluator's error spans and severities show strong agreement with human MQM annotations and human editor preferences.
Chinese Translation
基于大型语言模型（LLM）的机器翻译评估能够与人类判断紧密匹配，但在实际应用中，它仍然主要是诊断性的，评估信号很少能在真实生产约束下转化为直接的质量提升。我们提出了一种两阶段的、评估者指导的自动后编辑框架，将MQM风格的评估转化为有针对性的修复：一种增强检索的LLM评估器在明确的编辑合同下输出结构化的、跨度级别的MQM诊断，而一个单独的LLM后编辑器仅对这些诊断进行最小限度的编辑。这种分离提高了可控性，并减少了与单阶段“评估与改进”基线相比的意译漂移。在一项涉及七种LLM、三个模型提供者和七种语言的系统研究中，我们的最佳配置在COMET-22和COMETKiwi评分上始终优于单阶段后编辑方法，同时评估者的错误跨度和严重程度与人类MQM注释及人类编辑者偏好表现出强一致性。
cs.CL / 90 / 2609.22796
AlexandriaX 2026: The First Shared Task on Dialectal Arabic Machine Translation
AlexandriaX 2026：首届方言阿拉伯语机器翻译共享任务
Abstract
Dialectal Arabic machine translation (MT) remains challenging despite recent progress in Arabic language technologies, particularly because effective translation requires modeling not only semantic content but also dialectal variation, conversational context, speaker and addressee characteristics, and sociolinguistic appropriateness. Moreover, conventional MT metrics provide limited insight into the linguistic errors produced by dialectal systems. We present the AlexandriaX 2026 Shared Task on Dialectal Arabic MT, which addresses these challenges through three complementary subtasks: (1) context-aware English-to-Dialectal Arabic dialogue translation across 13 Arabic varieties, (2) cross-dialect Arabic translation in the financial domain covering six Arabic dialects, and (3) span-level MT error detection and classification using linguistically motivated error categories across five Arabic varieties. The shared task attracted 38 registrations for Subtask 1, 33 for Subtask 2, and 35 for Subtask 3. Twelve unique teams submitted their system description papers, all of which we accepted for publication. The best system on Subtask 1 achieved 30.42 spBLEU in the constrained track and 33.49 spBLEU in the unconstrained track. On Subtask 2, the top system obtained 28.40 spBLEU. On Subtask 3, the best system achieved an overall score of 49.82, outperforming the 24.29 baseline. Taken together, the results of the leading systems across the three subtasks highlight the benefits of explicit dialect modeling, context-aware generation, retrieval and reranking, and specialized approaches to interpretable MT error analysis. All the resources of AlexandriaX 2026 shared task are publicly available, including data, baselines, and evaluation code on our project page: https://alexandriax.dlnlp.ai.
Chinese Translation
尽管阿拉伯语言技术最近取得了进展，方言阿拉伯语机器翻译（MT）仍然面临挑战，特别是因为有效的翻译不仅需要建模语义内容，还需要考虑方言变异、对话上下文、说话者和听众特征以及社会语言学的适宜性。此外，传统的机器翻译评估指标对方言系统产生的语言错误提供的洞察有限。我们提出了AlexandriaX 2026方言阿拉伯语机器翻译共享任务，通过三个互补的子任务来应对这些挑战：（1）针对13种阿拉伯方言的上下文感知英语到方言阿拉伯语对话翻译，（2）涵盖六种阿拉伯方言的金融领域跨方言阿拉伯语翻译，以及（3）使用五种阿拉伯方言的语言学驱动错误类别进行跨度级机器翻译错误检测和分类。该共享任务吸引了38个子任务1的注册，33个子任务2的注册，以及35个子任务3的注册。十二个独特团队提交了他们的系统描述论文，所有论文均被接受发表。子任务1中表现最佳的系统在受限轨道上达到了30.42 spBLEU，在非受限轨道上达到了33.49 spBLEU。子任务2中，顶级系统获得了28.40 spBLEU。在子任务3中，最佳系统的总体得分为49.82，超越了24.29的基线。综合来看，三个子任务中领先系统的结果突显了显式方言建模、上下文感知生成、检索和重排序以及可解释的机器翻译错误分析的专门方法的优势。AlexandriaX 2026共享任务的所有资源均可公开获取，包括数据、基线和评估代码，详见我们的项目页面：https://alexandriax.dlnlp.ai。
cs.CL / 91 / 2609.22805
To Consolidate or not to Consolidate? Evaluating the Impact of Consolidation in Multi-Reference Training using Peer Reviews
合并还是不合并？评估多参考训练中合并的影响
Abstract
Natural language generation (NLG) tasks span the spectrum of conditional entropy, ranging from highly constrained machine translation to open-ended dialogue generation. Structured tasks like automated peer-review generation occupy the intermediate region, where a single input admits multiple valid, overlapping outputs. In this work, we demonstrate that traditional single- and multi-reference training paradigms are suboptimal for these intermediary tasks. We provide empirical evidence that consolidating diverse references into a unified training signal is crucial for developing effective systems. To facilitate this, we introduce MERC-36K, a large-scale corpus of over 36,000 papers paired with original and consolidated peer reviews. Using this dataset, we train specific architectures to isolate the impact of different reference paradigms and benchmark against existing state-of-the-art systems. Through extensive automatic and human evaluation, we demonstrate that models trained on consolidated references significantly outperform those trained on unconsolidated references. Dataset and code will be released upon acceptance.
Chinese Translation
自然语言生成（NLG）任务涵盖了条件熵的广泛范围，从高度约束的机器翻译到开放式对话生成。像自动同行评审生成这样的结构化任务处于中间区域，在该区域内，单一输入可以产生多个有效且重叠的输出。在本研究中，我们证明了传统的单参考和多参考训练范式对于这些中介任务并不理想。我们提供了实证证据，表明将多样化的参考合并为统一的训练信号对于开发有效系统至关重要。为此，我们引入了MERC-36K，这是一个包含超过36,000篇论文及其原始和合并同行评审的大规模语料库。利用该数据集，我们训练特定架构以隔离不同参考范式的影响，并与现有的最先进系统进行基准测试。通过广泛的自动和人工评估，我们证明了基于合并参考训练的模型显著优于基于未合并参考训练的模型。数据集和代码将在接受后发布。
cs.CL / 92 / 2609.22884
Block-Sparse Attention with Semantic-Geometric Decoupled Routing
语义-几何解耦路由的块稀疏注意力
Abstract
Long-context inference has become a defining capability of large language models, but exact dense attention remains costly due to its quadratic scaling with sequence length. Block-sparse attention offers a hardware-friendly alternative by routing each query block to a small set of relevant key blocks, yet accurate training-free block routing remains difficult. Existing routers often pool post-RoPE token representations, which entangles semantic aggregation with RoPE-induced geometry and attenuates local positional cues through high-frequency phase cancellation. To resolve this mismatch, we propose \textbf{Semantic-Geometric Decoupled Routing}, a training-free block routing framework that shifts semantic aggregation to the pre-RoPE space and reconstructs geometric bias with an offline structural prior and relative block distances. This decomposition yields an explicit closed-form block routing score without token-level search or post-hoc calibration. Experiments on long-context text and video tasks show that our method approaches full-attention accuracy across 4K--128K contexts, keeps routing overhead below 3.4 ms, and achieves a 5.03$\times$ speedup over FlashAttn at a 128K context length.
Chinese Translation
长上下文推理已成为大型语言模型的一项关键能力，但由于其与序列长度的平方关系，精确的密集注意力仍然成本高昂。块稀疏注意力通过将每个查询块路由到一小组相关的键块，提供了一种硬件友好的替代方案，但准确的无训练块路由仍然困难。现有的路由器通常对后RoPE（Rotary Position Embedding）令牌表示进行池化，这将语义聚合与RoPE引起的几何相互纠缠，并通过高频相位抵消减弱局部位置线索。为了解决这种不匹配，我们提出了 extbf{语义-几何解耦路由}，这是一种无训练的块路由框架，它将语义聚合转移到前RoPE空间，并通过离线结构先验和相对块距离重建几何偏差。这种分解产生了一个明确的闭式块路由评分，无需令牌级搜索或事后校准。在长上下文文本和视频任务上的实验表明，我们的方法在4K到128K的上下文中接近全注意力的准确性，保持路由开销低于3.4毫秒，并在128K上下文长度上实现了比FlashAttn快5.03倍的速度提升。
cs.CL / 93 / 2609.22904
LLMs Anchor on Chief Complaint and Fail to Integrate Evidence in Sequential Clinical Triage
大型语言模型在主诉基础上的锚定及其未能整合证据于顺序临床分诊中
Abstract
Triage in the emergency department (ED) is a sequential decision process that unfolds turn by turn. Existing evaluations of large language models (LLMs) for triage use completed retrospective records and report performance close to that of physicians. We implement a methodology for evaluating LLMs on sequential triage, the task of predicting a triage acuity label from a growing prefix of a nurse-patient conversation. We evaluate six LLMs at five sequential checkpoints on two corpora: 425 LLM-generated (SIMULATED) and 50 physician-authored (CLINICIAN) conversations, both labelled under the Emergency Severity Index (ESI). Every model, measured by quadratic weighted kappa (QWK), degrades from moderate-to-substantial agreement on completed records to fair-to-moderate agreement at every sequential checkpoint. Controlled perturbations show that the label at every checkpoint is anchored on the chief complaint exchanges, and prompting interventions fail to lift this plateau. Models extract clinically relevant content from later turns, yet the surprisal of the true label rises across the checkpoints. So the model fails to integrate the evidence. Three expert clinicians on the same conversations reach a QWK of 0.887-0.929, while the best model reaches 0.295. Predictions concentrate at ESI-2 and ESI-3, and models agree with each other more than with the ground truth, so ensembling worsens the failure. Deploying LLMs for ED triage based on offline benchmarks alone misses this sequential failure.
Chinese Translation
急诊科（ED）的分诊是一个逐步决策过程，逐轮展开。现有对大型语言模型（LLMs）在分诊中的评估使用已完成的回顾性记录，并报告其性能接近于医生。我们实施了一种评估LLMs在顺序分诊中的方法论，即从护士与患者对话的逐步前缀中预测分诊严重性标签的任务。我们在两个语料库上评估了六个LLMs在五个顺序检查点的表现：425个LLM生成的（模拟）对话和50个医生撰写的（临床医生）对话，均根据紧急严重性指数（ESI）进行标注。每个模型在二次加权卡帕（QWK）测量下，从已完成记录的中等到显著一致性降至每个顺序检查点的公平到中等一致性。受控扰动显示，每个检查点的标签都锚定于主诉交流，而提示干预未能提升这一停滞。模型从后续轮次中提取临床相关内容，但真实标签的惊讶度在各个检查点上升。因此，模型未能整合证据。三位专家临床医生在相同对话中达到的QWK为0.887-0.929，而最佳模型的QWK为0.295。预测集中在ESI-2和ESI-3，模型之间的同意度高于与真实标签的同意度，因此集成反而加剧了失败。仅基于离线基准部署LLMs用于急诊分诊未能识别这一顺序性失败。
cs.CL / 94 / 2609.22917
An Iterative LangGraph Agent for Text-to-SQL: Natural Language Access to the Chicago Crime Database
用于文本到SQL的迭代LangGraph代理：自然语言访问芝加哥犯罪数据库
Abstract
Non-technical stakeholders frequently cannot write the SQL needed to extract insights from operational databases. We built and evaluated a Text-to-SQL agent that closes this gap end to end: a six-node LangGraph StateGraph checks question relevance, fetches the live schema, generates PostgreSQL, validates it with a dry run, retries on failure, executes the query, and narrates the result set in plain English. The agent uses prompt engineering only; no model was fine-tuned. We evaluated it on the Chicago Crime dataset (approximately 8.5 million records, 22 attributes) against a hand-built benchmark of 100 natural language questions with ground-truth SQL, stratified into 30 Easy, 40 Medium and 30 Hard items. Comparing two prompt revisions of the same agent, the revised system (V2) reached a Valid SQL Rate of 93% (from 87%), an Execution Accuracy of 60% under a hybrid relational equivalence metric (from 47%; 19% from 12% under strict JSON matching), and a mean Synthesis Quality of 4.34 out of 5 (from 3.91). The single largest driver was removing a LIMIT 10 instruction from the system prompt, which had been truncating multi-row answers. Error analysis attributes the residual failures to relevance-checker false rejections, ambiguous question semantics, and free-tier API rate limits rather than to the language generation step. We report no comparison against an external baseline system or a public benchmark; the study is a single-model engineering evaluation.
Chinese Translation
非技术利益相关者通常无法编写所需的SQL以从操作数据库中提取见解。我们构建并评估了一个端到端的文本到SQL代理，填补了这一空白：一个六节点的LangGraph状态图检查问题的相关性，获取实时模式，生成PostgreSQL，进行干运行验证，在失败时重试，执行查询，并用简单英语叙述结果集。该代理仅使用提示工程；没有对模型进行微调。我们在芝加哥犯罪数据集（约850万条记录，22个属性）上进行了评估，针对一个手工构建的包含100个自然语言问题及其真实SQL的基准，这些问题被分为30个简单、40个中等和30个困难项。在对同一代理的两个提示版本进行比较时，修订后的系统（V2）达到了93%的有效SQL率（从87%提高），在混合关系等价度量下的执行准确率为60%（从47%提高；在严格的JSON匹配下为19%，从12%提高），平均合成质量为4.34分（满分5分，从3.91提高）。最大驱动因素是从系统提示中移除了LIMIT 10指令，这一指令曾导致多行答案被截断。错误分析将剩余的失败归因于相关性检查器的错误拒绝、模糊的问题语义和免费层API的速率限制，而不是语言生成步骤。我们没有与外部基准系统或公共基准进行比较；本研究是一项单模型工程评估。
cs.CL / 95 / 2609.22934
Measuring Behavioural Signatures of Large Language Models through Psychometric Profiling
通过心理测量分析测量大型语言模型的行为特征
Abstract
Large language models (LLMs) increasingly mediate human decisions and communication, yet their behavioural regularities remain difficult to characterize systematically. We develop a cross-linguistic psychometric profiling framework and evaluate nine LLMs using seven psychological instruments, with five repeated administrations per model and language in Chinese and English. Items unresolved after a prespecified retry procedure are retained as NA. Joint analysis of scored and NA responses captures response tendencies and boundaries of self-report applicability. LLMs exhibit structured, model-specific profiles despite a shared alignment-shaped pattern of higher prosocial and self-regulatory responses and lower dominance, disengagement and harmful-intent endorsement. NA responses are structured rather than uniformly distributed, indicating where outputs are treated as inapplicable, refused or cannot be mapped to valid response options. Language condition and provider origin are associated with profile configuration and answerability, whereas repeated administrations show high reproducibility and permit recovery of model identity. Human-reference and prompt-robustness analyses further indicate that these signatures are context dependent. Joint analysis of psychometric profiling and answerability offers a framework for quantifying deployment-level behavioural signatures.
Chinese Translation
大型语言模型（LLMs）越来越多地介导人类的决策和交流，但其行为规律仍然难以系统地表征。我们开发了一种跨语言的心理测量分析框架，并使用七种心理学工具对九个LLM进行了评估，每个模型和语言（中文和英文）进行了五次重复测量。在预先指定的重试程序后未解决的项目被保留为NA。对得分和NA响应的联合分析捕捉了响应倾向和自我报告适用性的边界。尽管共享了更高的亲社会和自我调节响应以及较低的支配、脱离和有害意图支持的模式，LLM仍表现出结构化的、特定于模型的特征。NA响应是结构化的，而非均匀分布的，表明输出在何处被视为不适用、拒绝或无法映射到有效的响应选项。语言条件和提供者来源与特征配置和可回答性相关，而重复测量显示出高重现性，并允许恢复模型身份。人类参考和提示鲁棒性分析进一步表明，这些特征是依赖于上下文的。心理测量分析与可回答性的联合分析提供了一个量化部署级别行为特征的框架。
cs.CL / 96 / 2609.22971
Automatic multimodal UX improvement recommendations from LLM agent user simulations
基于大型语言模型代理用户模拟的自动多模态用户体验改进推荐
Abstract
Evaluating user experience (UX) on live websites through user testing is expensive, subjective, and difficult to scale. LLM agents offer a promising route to automating UX testing by simulating realistic user behaviour. However, existing simulation approaches typically lack multimodality and require time-consuming manual review to extract actionable insights. We formalise UX improvement recommendation from simulation data as a structured natural language generation and ranking problem, and establish an evaluation protocol using expert annotation and LLM-as-a-Judge. We present AMUSER, a multimodal framework which simulates user behaviour and automatically generates prioritised UX improvement recommendations from resulting data. We evaluate AMUSER on commercial websites and show that its recommendations substantially outperform those from text-only simulation (NDCG@3 = 0.758 versus 0.359) at an 89% lower simulation cost. Our results suggest an asymmetric role of multimodality: visual access during simulation improves recommendations through richer traces, while providing visual inputs during recommendation generation can modestly degrade quality. We also discuss practical deployment lessons from applying AMUSER to commercial websites.
Chinese Translation
通过用户测试评估实时网站的用户体验（UX）既昂贵又主观，且难以扩展。大型语言模型（LLM）代理提供了一种通过模拟真实用户行为来自动化UX测试的有希望的途径。然而，现有的模拟方法通常缺乏多模态性，并且需要耗时的人工审查来提取可操作的见解。我们将从模拟数据中提出UX改进推荐形式化为一个结构化的自然语言生成和排名问题，并建立了一个使用专家注释和LLM作为评判者的评估协议。我们提出了AMUSER，一个多模态框架，它模拟用户行为并自动生成优先级排序的UX改进推荐。我们在商业网站上评估AMUSER，结果表明其推荐显著优于仅基于文本的模拟推荐（NDCG@3 = 0.758 对比 0.359），且模拟成本降低了89%。我们的结果表明多模态性具有不对称的作用：模拟期间的视觉访问通过更丰富的痕迹改善了推荐，而在推荐生成期间提供视觉输入可能会适度降低质量。我们还讨论了将AMUSER应用于商业网站的实际部署经验教训。
cs.CL / 97 / 2609.22988
Rethinking Pivot Programming Languages in Code Language Models
重新思考代码语言模型中的枢轴编程语言
Abstract
Multilingual code language models transfer skills across programming languages (PLs), but whether any PL occupies a privileged pivot position remains contested: geometric analyses point to C-family languages and Go, while behavioral evidence highlights Python. We revisit this question under controls for representational anisotropy and length variation across PLs, two confounds that compromise prior cosine-based analyses. Across three code models on multilingual competitive-programming data, we study three views of cross-PL organization: pairwise PL geometry, PL-English alignment, and pivoted retrieval through candidate PL representation spaces. The results are relation-dependent. Code-code geometry reveals structured language regions but no universal center; code-English alignment favors high-level scripting languages; and pivoted retrieval favors different intermediate spaces for code-to-code and English-to-code transfer. These findings suggest that Python's special role is better understood as English-facing affinity than as universal geometric centrality.
Chinese Translation
多语言代码语言模型在编程语言（PLs）之间转移技能，但是否有任何编程语言占据特权的枢轴位置仍存在争议：几何分析指向C家族语言和Go，而行为证据则强调Python。我们在控制了编程语言的表示各向异性和长度变化这两个影响因素后，重新审视了这个问题，这两个因素会影响之前基于余弦的分析。在三种多语言竞争编程数据的代码模型中，我们研究了跨编程语言组织的三种视角：成对编程语言几何、编程语言与英语的对齐，以及通过候选编程语言表示空间的枢轴检索。结果依赖于关系。代码-代码几何揭示了结构化的语言区域，但没有普遍的中心；代码-英语对齐更倾向于高级脚本语言；而枢轴检索则偏好不同的中间空间用于代码到代码和英语到代码的转移。这些发现表明，Python的特殊角色更应被理解为面向英语的亲和力，而非普遍的几何中心性。
cs.CL / 98 / 2609.23039
Auditing Political Alignment in LLM Assistants: Engagement, Stance, and User Identity
审计大型语言模型助手的政治对齐：参与度、立场与用户身份
Abstract
LLM-based AI systems answer political questions for hundreds of millions of people. Current audits measure what they say to an average user, but their behavior is dynamic. I argue that their political behavior is a set of policies over whom to answer, what to say, and whether to engage at all, conditional on the topic and what the system knows about the user. I call these policies the system's speech regime, which is how a developer settles the tradeoff between answering, accommodating the user, and refusing, each of which carries a cost that varies by topic. I derive a typology of five regimes from two dimensions, engagement and stance. I test six AI systems (OpenAI, Anthropic, xAI, Google, Mistral, DeepSeek) in a preregistered experiment of 7,500 multi-turn conversations that randomly assign the user's political identity across five topics: abortion, Catalan independence, climate change, Nazism, and a zero-stakes control (pineapple on pizza). Two LLM judges from different developers score every answer, validated against human coding, and refusal is treated as an outcome rather than missing data. Every system accommodates the user on the control topic, showing that political restraint is a policy. On contested topics the systems fall into different regimes: on abortion, GPT engages and mirrors every user, Gemma refuses everyone, Claude answers strongly conservative users 35 percent of the time and almost no one else, and Grok accommodates conservatives only. On settled topics such as climate change and Nazism, five systems hold firm for every user. The systems also infer the user's overall ideology, so accommodation can spill over to topics not yet discussed. A comparison of two Grok releases shows the regime changing between versions in a way current audits miss. Speech regimes matter for alignment research and for polarization, political knowledge, and the quality of democracy.
Chinese Translation
基于大型语言模型（LLM）的人工智能系统为数亿人解答政治问题。目前的审计主要测量它们对普通用户的回答，但它们的行为是动态的。我认为它们的政治行为是一系列政策，涉及回答谁的问题、说什么以及是否参与，这些都取决于话题和系统对用户的了解。我将这些政策称为系统的言论制度，它是开发者在回答、迎合用户和拒绝之间权衡的结果，而每种选择的成本因话题而异。我从参与度和立场两个维度推导出五种制度的类型。在一项预注册实验中，我测试了六个人工智能系统（OpenAI、Anthropic、xAI、Google、Mistral、DeepSeek），该实验包含7500个多轮对话，随机分配用户的政治身份，涵盖五个话题：堕胎、加泰罗尼亚独立、气候变化、纳粹主义，以及一个零风险的控制话题（披萨上的菠萝）。来自不同开发者的两位LLM评审对每个回答进行评分，并与人类编码进行验证，拒绝被视为结果而非缺失数据。每个系统在控制话题上都迎合用户，表明政治克制是一项政策。在有争议的话题上，各系统表现出不同的制度：在堕胎问题上，GPT与每位用户互动并反映其观点，Gemma拒绝所有人，Claude对强保守派用户的回答率为35%，几乎不回答其他人，而Grok仅迎合保守派。在气候变化和纳粹主义等已解决的话题上，五个系统对每位用户都保持坚定。这些系统还推断用户的整体意识形态，因此迎合可能会溢出到尚未讨论的话题。对两个Grok版本的比较显示，制度在版本之间发生变化，而当前的审计未能捕捉到这一点。言论制度对对齐研究、极化、政治知识和民主质量具有重要意义。
cs.CL / 99 / 2609.23053
Attributable Post-Rationalization in RAG Citations: A Controlled Reproduction and an RLVR Comparison
RAG 引用中的可归因后理性化：受控重现与 RLVR 比较
Abstract
A RAG system can hand you the right answer and cite a source it did not actually use. Models output these unfaithful citations via post-rationalization: they write the answer first and then attach a citation to whatever passage looks close enough. Search agents are now trained with reinforcement learning from verifiable rewards (RLVR), which pays them for getting the answer right. We asked whether that training also teaches them to cite honestly. Improving an existing methodology with a required control, we compared an instruction-tuned model against three RLVR agents trained from it, on four question-answering datasets, using only free-tier Kaggle GPUs. Post-rationalization is everywhere: on Wikipedia-based questions roughly one citation in seven is unfaithful. RLVR does not fix it. The agents post-rationalize at their base model's rate, and one lands slightly worse. Rewarding correct answers buys nothing in citation faithfulness, so faithfulness has to be trained and measured on its own terms.
Chinese Translation
RAG 系统可以给出正确答案，但引用的来源并未实际使用。模型通过后理性化输出这些不忠实的引用：它们首先写出答案，然后附上与任何看起来足够接近的段落的引用。搜索代理现在通过可验证奖励的强化学习（RLVR）进行训练，这种方法奖励它们正确回答的能力。我们探讨了这种训练是否也教会它们诚实引用。我们在现有方法的基础上进行了改进，并进行了必要的控制，比较了一个经过指令调优的模型与从中训练出的三个 RLVR 代理在四个问答数据集上的表现，仅使用免费的 Kaggle GPU。后理性化无处不在：在基于维基百科的问题中，大约每七个引用中就有一个是不忠实的。RLVR 并没有解决这个问题。这些代理以其基础模型的速率进行后理性化，其中一个的表现略微更差。奖励正确答案并未提高引用的忠实度，因此忠实度必须在其自身的标准上进行训练和测量。
cs.CL / 100 / 2609.23056
Bridging Static and Agentic RAG for Taiwanese Historical Question Answering
连接静态与主动式RAG用于台湾历史问答
Abstract
Agentic retrieval-augmented generation (RAG) enables language models to adapt retrieval based on previously retrieved evidence, but it remains unclear whether such adaptive orchestration consistently outperforms well-designed static pipelines. We conduct a controlled comparison of agentic and static RAG for Taiwanese historical question answering, sharing the same generator and hybrid retrieval backend. Despite similar aggregate performance, the two pipelines differ on 70.83% of questions, with their advantages largely canceling out when averaged. An oracle that selects the better response per question improves the composite score by 0.2417 over the better individual pipeline, revealing substantial headroom for question-level selection. We therefore introduce a post-hoc selector that compares the two responses and their cited evidence, significantly outperforming either individual pipeline and recovering 60.34% of the oracle headroom. These results show that aggregate comparisons can obscure meaningful question-level differences between retrieval strategies, suggesting that exploiting their complementarity may be more fruitful than seeking a universally superior pipeline.
Chinese Translation
主动式检索增强生成（Agentic retrieval-augmented generation, RAG）使语言模型能够基于先前检索到的证据自适应调整检索策略，但尚不清楚这种自适应编排是否始终优于设计良好的静态流程。我们在台湾历史问答任务中对主动式与静态RAG进行了受控比较，二者共享相同的生成器和混合检索后端。尽管总体性能相近，两种流程在70.83%的问题上表现不同，其优势在平均时大致相互抵消。一个能够针对每个问题选择更优回答的oracle相比表现更好的单一流程，复合得分提升了0.2417，显示出问题级选择存在显著提升空间。基于此，我们引入了一个事后选择器，通过比较两种回答及其引用证据，显著优于任一单独流程，并恢复了60.34%的oracle提升空间。结果表明，整体性能比较可能掩盖检索策略在问题层面的显著差异，提示利用其互补性可能比追求单一普适优流程更具成效。
cs.CL / 101 / 2609.23065
From Concept Alignment to Causal Grounding: An Intervention Test of Chain-of-Thought Faithfulness
从概念对齐到因果基础：链式思维忠实性的干预测试
Abstract
Chain-of-thought (CoT) can sound plausible yet be unfaithful to the model's underlying reasoning. Most prior work probes CoT faithfulness through input--output behavior or input attributions, leaving internal computation largely underexplored. We instead cast faithfulness as internal concept grounding: Does a large language model's (LLM) CoT reasoning engage the same internal concepts that support the LLM's direct prediction, and do the shared concepts causally drive its answer? Encoding a prediction pass and a CoT pass with a single shared sparse autoencoder (SAE), a reliable approximator of the latent concepts LLMs use, makes their internal concepts directly comparable. We introduce three correlational metrics of concept-level alignment and a causal metric, $\Delta p$, which ablates the shared concepts and measures the drop in answer probability. Across five LLMs and four datasets, concept alignment is generally high, as indicated by the correlational metrics; yet these only identify which concepts are shared, not how much they causally contribute. $\Delta p$ fills this gap: causal faithfulness varies substantially with model depth, peaking at mid-to-late layers rather than the final ones, and model scale reshapes the layer-wise profile. Moreover, causally important shared concepts are not always verbalized in the CoT. These dissociations suggest that faithfulness cannot be reliably assessed from surface-level or representational correspondence alone; assessing it requires causal tests of whether the internal concepts underlying a CoT actually drive the model's prediction.
Chinese Translation
链式思维（Chain-of-thought, CoT）可能听起来合理，但却可能与模型的基本推理不一致。以往的研究主要通过输入-输出行为或输入归因来探讨 CoT 的忠实性，内部计算则很少被深入研究。我们将忠实性视为内部概念的基础：大型语言模型（Large Language Model, LLM）的 CoT 推理是否涉及支持 LLM 直接预测的相同内部概念，并且这些共享概念是否因果驱动其答案？通过使用一个共享的稀疏自编码器（Sparse Autoencoder, SAE）对预测过程和 CoT 过程进行编码，这一可靠的近似器可以捕捉 LLM 使用的潜在概念，使其内部概念可以直接比较。我们引入了三个概念级对齐的相关性指标和一个因果指标 $ ext{Δ}p$，该指标消融了共享概念并测量答案概率的下降。在五个 LLM 和四个数据集上，相关性指标表明概念对齐通常较高；然而，这些指标仅识别哪些概念是共享的，而未能说明它们的因果贡献程度。$ ext{Δ}p$ 填补了这一空白：因果忠实性在模型深度上变化显著，在中层到后期层达到峰值，而不是在最后一层，模型规模也重塑了层级特征。此外，因果重要的共享概念并不总是在 CoT 中被明确表达。这些分离现象表明，忠实性不能仅通过表层或表征的一致性来可靠评估；评估其忠实性需要进行因果测试，以确定支撑 CoT 的内部概念是否真正驱动模型的预测。
cs.CL / 102 / 2609.23083
Directing large language models to follow the letter or spirit of the law
引导大型语言模型遵循法律的字面或精神
Abstract
The distinction between the spirit and letter of the law is a central issue across research and everyday life, and a growing concern for building safe, intelligent machines. What is this distinction based on, and how can we develop machines that follow the intention behind a rule? We used targeted adaptation that made large language models prioritize the spirit or letter of the law. With minimal modifications, our method significantly changed LLM behavior across diverse measures, novel vignettes, real-world scenarios, and influential legal cases. An analysis of model internals revealed a low-dimensional space with three interpretable dimensions matching a formal pre-specified framework for the geometry of legal concepts. These findings show how legal thought in LLMs may be organized and directed.
Chinese Translation
法律的精神与字面之间的区别是研究和日常生活中的一个核心问题，也是构建安全、智能机器日益关注的焦点。这一区别基于什么？我们如何开发能够遵循规则背后意图的机器？我们采用了针对性的适应方法，使大型语言模型优先考虑法律的精神或字面。通过最小的修改，我们的方法显著改变了大型语言模型在多种测量、创新情境、现实世界场景以及有影响力的法律案件中的行为。对模型内部的分析揭示了一个低维空间，具有三个可解释的维度，与法律概念几何的正式预设框架相匹配。这些发现展示了大型语言模型中的法律思维如何被组织和引导。
cs.CL / 103 / 2609.23088
OmniEdu: Open Foundation Models for Learning and Teaching
OmniEdu：用于学习和教学的开放基础模型
Abstract
Educational foundation models must solve problems, understand curriculum structure, diagnose learner difficulties, and provide appropriate instructional support. Existing educational language models often focus on either problem solving or tutoring, with training mixtures organized by source or task rather than capability. We present OmniEdu, an open family of foundation models for K-12 learning and teaching. Its instruction-tuning corpus combines over 100 educational resources and general instruction sources, organized around four capabilities: subject competence, curriculum grounding, diagnostic reasoning, and pedagogical action and scaffolding. Our pipeline integrates deterministic cleaning, semantic auditing and rewriting, task-specific quality scoring, token-budgeted diversity selection, and pedagogical instruction assignment. It yields 69,999 examples and 15.96M supervised response tokens, including 60,951 education-specific examples. We fine-tune 4B, 9B, and 27B models and evaluate curriculum grounding, K-12 problem solving, and pedagogical tutoring, alongside general capability. Education-oriented tuning consistently improves all three educational benchmark groups across model scales. OmniEdu-27B achieves 63.12% EM and 76.69% F1 on K12-Bench, 85.89% on MathFish, 86.95% on EDUMATH, and 78.74% in MathTutorBench's Scaffold setting. It also achieves the highest Teaching average on LongTutor among the evaluated models, at 3.02. These results demonstrate the value of curated, capability-balanced supervision for adapting general language models to educational tasks spanning problem solving, curriculum understanding, and instructional support.
Chinese Translation
教育基础模型必须解决问题、理解课程结构、诊断学习者困难，并提供适当的教学支持。现有的教育语言模型往往专注于问题解决或辅导，训练混合通常按来源或任务而非能力进行组织。我们提出了 OmniEdu，一个用于 K-12 学习和教学的开放基础模型家族。其指令调优语料库结合了超过 100 种教育资源和一般教学来源，围绕四种能力进行组织：学科能力、课程基础、诊断推理和教学行动及支架。我们的流程集成了确定性清理、语义审计与重写、特定任务质量评分、基于令牌预算的多样性选择和教学指令分配。最终生成了 69,999 个示例和 15.96M 的监督响应令牌，其中包括 60,951 个教育特定示例。我们对 4B、9B 和 27B 模型进行了微调，并评估了课程基础、K-12 问题解决和教学辅导，以及一般能力。面向教育的调优在所有模型规模上始终提高了这三个教育基准组的表现。OmniEdu-27B 在 K12-Bench 上达到了 63.12% 的 EM 和 76.69% 的 F1，在 MathFish 上达到了 85.89%，在 EDUMATH 上达到了 86.95%，在 MathTutorBench 的支架设置中达到了 78.74%。在评估的模型中，它在 LongTutor 上的教学平均分为 3.02，位居最高。这些结果证明了经过策划的、能力平衡的监督在将通用语言模型适应于涵盖问题解决、课程理解和教学支持的教育任务中的价值。
cs.CL / 104 / 2609.23178
Chronologic: Measuring Language Models' Ability to Represent the Past
Chronologic：测量语言模型对过去的表征能力
Abstract
Language models are appealing tools for research on the past. But to trust the evidence a model provides, researchers need to know whether its responses fit the period represented. Validation is challenging, because this is not a task living people ordinarily perform, and because many questions have multiple correct answers. We use historical texts to develop a benchmark for a model's representation of English-language contexts 1831-1930, relying on pairwise comparisons to multiple ground truths and strong distractors to score the hardest questions in an appropriately graduated way. We find that generative tasks are harder than discriminative ones; in fact, reasoning models can typically discern the weakness of their own generated answers. While models pretrained exclusively on historical text lead the pack when evaluated by answer likelihood, they cannot compete with commercial models in free generation. None of the models we tested represent historical contexts in a fully persuasive way yet, but progress toward that goal is evident.
Chinese Translation
语言模型是研究过去的有吸引力的工具。然而，为了信任模型提供的证据，研究人员需要了解其响应是否符合所代表的时期。验证这一点具有挑战性，因为这不是活人通常执行的任务，并且许多问题有多个正确答案。我们利用历史文本开发了一个基准，用于评估模型在1831年至1930年间的英语语境中的表征能力，依赖于对多个真实答案和强干扰项的成对比较，以适当地分级评分最困难的问题。我们的研究发现，生成任务比判别任务更具挑战性；事实上，推理模型通常能够识别自己生成答案的弱点。尽管在根据答案可能性评估时，专门在历史文本上预训练的模型表现优异，但在自由生成方面却无法与商业模型竞争。我们测试的模型尚未以完全令人信服的方式表征历史语境，但朝着这一目标的进展是显而易见的。
cs.CL / 105 / 2609.23191
Low resource cross-modal alignment using HGNN to enhance speech representation
利用HGNN进行低资源跨模态对齐以增强语音表示
Abstract
Speech-text space alignment is a multimodal representation learning method consisting to map different speech and text into a shared representation space, leading to enrichment of the representation of each modality. Proposed architectures, such as SAMU-XLSR, typically follow a student/teacher framework, with the goal of fine-tuning an audio encoder to produce representations that closely match those of the text. In this way a speech representation is semantically enriched. However, such systems generally require large amounts of training data and considerable computational resource, making them difficult to apply to low resources languages under frugal constraints. The present work proposes a data-efficient space alignment method based on Heterogeneous Graph Neural Networks and link prediction. The core idea is to leverage message passing to explicitly transfer information from the text modality to the speech modality, thereby reducing the need for large training datasets and intrinsically enriching the acoustic representations, all in a more interpretable manner. Although thoroughly explored for high-resource languages, word-level tasks in speech remain relevant for certain low-resource languages. Therefore, we conducted experiments on speech-text alignment at the word level using the TIMIT (English) dataset and Yemba (a Cameroonian language). Our approach yields results comparable to those of SAMU-XLSR, a state-of-the-art method, and even surpasses it for the Yemba language in the task of word retrieval, while using far fewer resources, demonstrating its power, frugality, and efficiency.
Chinese Translation
语音-文本空间对齐是一种多模态表示学习方法，旨在将不同的语音和文本映射到一个共享的表示空间，从而丰富每种模态的表示。提出的架构，如SAMU-XLSR，通常遵循学生/教师框架，目标是微调音频编码器，以生成与文本表示紧密匹配的表示。通过这种方式，语音表示在语义上得到了丰富。然而，这类系统通常需要大量的训练数据和相当的计算资源，使其难以在资源有限的情况下应用于低资源语言。本研究提出了一种基于异构图神经网络（Heterogeneous Graph Neural Networks）和链接预测的数据高效空间对齐方法。其核心思想是利用信息传递显式地将信息从文本模态转移到语音模态，从而减少对大规模训练数据集的需求，并在更具可解释性的方式中内在地丰富声学表示。尽管在高资源语言中得到了充分探索，语音中的词级任务在某些低资源语言中仍然具有相关性。因此，我们使用TIMIT（英语）数据集和Yemba（喀麦隆语言）进行了词级语音-文本对齐实验。我们的方法在结果上与最先进的方法SAMU-XLSR相当，甚至在词检索任务中超越了其在Yemba语言的表现，同时使用的资源远少于前者，展示了其强大、节约和高效的特性。
cs.CL / 106 / 2609.23194
Enhancing speech representation learning with cross-modal knowledge transfer with HGNN under low resource settings: the case study of Yemba
在低资源环境下通过HGNN进行跨模态知识转移增强语音表示学习：以Yemba为例
Abstract
Acoustic representation learning is crucial for speech processing, yet low-resource languages (LRLs) face severe data scarcity, limiting the effectiveness of traditional and self-supervised methods. As a promising alternative, in this work, we propose to enhance acoustic representation trough a cross-modal transfer knowledge approach, based on heterogeneous graph neural networks (HGNNs), where acoustic and linguistic entities are modeled as distinct node types within a unified graph. Through message-passing mechanisms, linguistic nodes explicitly transfer knowledge to acoustic nodes, enabling structured and interpretable cross-modal information flow. To highlight this knowledge transfer and its benefits, we measured standard clustering metrics as an intrinsic evaluation of acoustic representation, and to emphasize applicability, we performed isolated-word recognition tasks using an English benchmark and a Cameroonian language dataset in low resources settings . Results demonstrate that acoustic representations consistently benefit from linguistic knowledge propagated through the graph. To our knowledge, this is the first demonstration of explicit cross-modal knowledge transfer for acoustic representation learning using HGNNs, highlighting a promising direction for speech representation in low-resource settings.
Chinese Translation
声学表示学习对语音处理至关重要，但低资源语言（LRLs）面临严重的数据稀缺，限制了传统和自监督方法的有效性。作为一种有前景的替代方案，本研究提出通过基于异构图神经网络（HGNN）的跨模态知识转移方法来增强声学表示，其中声学和语言实体被建模为统一图中的不同节点类型。通过消息传递机制，语言节点明确地将知识转移到声学节点，从而实现结构化和可解释的跨模态信息流。为了突出这种知识转移及其益处，我们测量了标准聚类指标作为声学表示的内在评估，并为了强调适用性，我们在低资源环境下使用英语基准和喀麦隆语言数据集进行了孤立词识别任务。结果表明，声学表示始终受益于通过图传播的语言知识。据我们所知，这是首次使用HGNN进行声学表示学习的显式跨模态知识转移的演示，突显了在低资源环境下进行语音表示的有前景方向。
cs.CL / 107 / 2609.23205
Euston: Training Away Mathematical Sycophancy Without Losing the Mathematics
Euston：在不失去数学能力的情况下消除数学谄媚的训练
Abstract
Reasoning language models are trained to produce solutions, not to refuse them, and this bias persists when the problem they are handed is false. Asked to prove a corrupted theorem, a strong model will typically comply and produce a confident derivation of something untrue. We present Euston, an 8B mathematical claim-verification model trained to resist exactly this. Training data were generated with GraphSynth, a probabilistic factor-graph generator that couples attribute-level diversity to decode-time structural masking and span-synchronized verification, yielding 3{,}026 matched true/corrupted statement pairs (6,052 statements) drawn from arXiv papers spanning 2010--2025. We fine-tuned DeepSeek-R1-8B with GRPO under a rule-based, zero-API reward for 189 steps on four H100 GPUs. On a balanced 200-true/200-false held-out split, balanced accuracy rises from 29.50% to 63.75% and the discrimination gap---the difference between the rate of calling false statements false and the rate of calling true statements false moves from -0.5% (z=-0.1) to +27.5% (z=+6.0). Critically, the gain is not purchased with general mathematical ability: AIME 2026 accuracy under official semantics is 65.00% against a 69.17% base, a difference of -4.17% that is not statistically significant, whereas an earlier run of the same recipe on a smaller GraphSynth corpus collapsed to 40.00%. Median response length also falls from 19,217 to 18,296 tokens and the truncation rate from 25.8% to 8.3%, so the improvement does not come from thinking longer. We report the result together with the confounds that bound its interpretation, principally the all-false composition of the official evaluation sets and the low precision implied at realistic error prevalence.
Chinese Translation
推理语言模型被训练以产生解决方案，而不是拒绝它们，这种偏见在所给出的问题为假时仍然存在。当被要求证明一个被腐蚀的定理时，一个强大的模型通常会顺从并产生一个自信但不真实的推导。我们提出了Euston，一个8B的数学声明验证模型，旨在抵抗这种情况。训练数据是通过GraphSynth生成的，这是一种将属性级多样性与解码时的结构掩蔽和跨度同步验证相结合的概率因子图生成器，生成了3,026对匹配的真实/腐蚀声明对（共6,052个声明），这些声明来源于2010年至2025年间的arXiv论文。我们在四个H100 GPU上，使用基于规则的零API奖励对DeepSeek-R1-8B进行了189步的微调。在一个平衡的200个真实/200个虚假的保留分割上，平衡准确率从29.50%上升到63.75%，而歧视差距——即将虚假声明标记为虚假的比率与将真实声明标记为虚假的比率之间的差异，从-0.5%（z=-0.1）上升到+27.5%（z=+6.0）。关键是，这一提升并没有以一般数学能力为代价：在官方语义下，AIME 2026的准确率为65.00%，而基准为69.17%，差异为-4.17%，在统计上并不显著，而在较小的GraphSynth语料库上运行相同方法的早期实验则降至40.00%。中位响应长度也从19,217个标记降至18,296个标记，截断率从25.8%降至8.3%，因此，改进并不是由于思考时间延长。我们报告了这一结果以及限制其解释的混淆因素，主要是官方评估集的全假组成和在现实错误发生率下所隐含的低精度。
cs.CL / 108 / 2609.23231
ChemCLIR-Bench: Benchmarking Cross-Lingual Information Retrieval in Multilingual Chemical Patents
ChemCLIR-Bench：多语言化学专利中的跨语言信息检索基准测试
Abstract
Cross-lingual information retrieval (CLIR) is increasingly important in multi-national industries, where critical technical evidence may exist in a different language than the query. However, existing benchmarks do not adequately capture domain-specific cross-lingual retrieval or the retrieval-depth and recoverability failures that aggregate recall hides. In this work, we benchmark CLIR in the chemical domain, with a focus on patent data. We construct a multilingual dataset from Google Patents and the European Patent Office (EPO) data, spanning five languages (covering major Eastern and Western languages) and reflecting the diversity and complexity of real-world industrial documentation. Using this dataset, we systematically evaluate eight state-of-the-art embedding models for cross-lingual retrieval. Our results show a substantial performance gap between monolingual and cross-lingual settings: for the best-performing model, Recall@10 drops from 0.72 to 0.53 in cross-lingual setting. Retrieval depth also degrades significantly, with relevant documents ranked lower across languages in cross-lingual scenarios. Furthermore, some multilingual embedding models that perform strongly in monolingual settings exhibit sharp declines when queries and documents are in different languages, providing practical insights for model selection in cross-lingual use cases. These findings highlight critical limitations of current approaches and emphasize the need for more robust cross-lingual retrieval methods in domain-specific settings. Our benchmark provides actionable insights for model selection and establishes a controlled diagnostic evaluation framework for CLIR over industrial technical text. Data and code are publicly available at https://github.com/MohammadKhodadad/Multi-Lingual-QAC.
Chinese Translation
跨语言信息检索（CLIR）在跨国行业中日益重要，因为关键技术证据可能以与查询不同的语言存在。然而，现有的基准测试未能充分捕捉特定领域的跨语言检索或聚合召回所隐藏的检索深度和可恢复性失败。在本研究中，我们在化学领域对CLIR进行基准测试，重点关注专利数据。我们从Google Patents和欧洲专利局（EPO）数据构建了一个多语言数据集，涵盖五种语言（包括主要的东西方语言），反映了现实工业文档的多样性和复杂性。利用该数据集，我们系统地评估了八种最先进的嵌入模型在跨语言检索中的表现。我们的结果显示，单语言和跨语言环境之间存在显著的性能差距：对于表现最好的模型，Recall@10在跨语言环境中从0.72降至0.53。检索深度也显著下降，在跨语言场景中，相关文档在不同语言中的排名较低。此外，一些在单语言环境中表现强劲的多语言嵌入模型在查询和文档使用不同语言时表现急剧下降，为跨语言应用场景中的模型选择提供了实用见解。这些发现突显了当前方法的关键局限性，并强调了在特定领域环境中对更强大跨语言检索方法的需求。我们的基准测试为模型选择提供了可操作的见解，并建立了一个针对工业技术文本的CLIR的受控诊断评估框架。数据和代码可在 https://github.com/MohammadKhodadad/Multi-Lingual-QAC 上公开获取。
cs.CL / 109 / 2609.23239
SoK: Formal Methods for Fact-Checking and Information Integrity
SoK：事实核查与信息完整性的形式化方法
Abstract
An automated fact-checking system returns a label: the claim is true, or it is false. In many such systems the verdict remains the primary output. What is generally missing is a record of which document settled the question, of what would have had to be different for the verdict to change, or of whether the same claim, reworded, would have been judged the same way. We call the missing piece a warrant: a separate statement of what was guaranteed and on what grounds. Formal methods produce evidence of this kind, and regulation is beginning to ask for it, since the Digital Services Act and the AI Act both call for auditable evidence about how systems behave. Surveys of automated fact-checking are usually organised by pipeline stage, and treat logic as one technique among many. We organise the field by what is being formalised instead, which gives five levels: the claim, the reasoning, the system doing the checking, the ecosystem the claim spreads through, and the regulatory obligation. Sorting 121 works into those levels, two patterns stand out. Most of the relevant formal machinery already exists, but it was built for other domains and has rarely been applied here, and the gap is widest for verifying the checking system itself. Several stages of the routine professional fact-checkers follow also have no stated correctness criterion, and two of them, writing a claim in checkable form and correcting a verdict already published, are not formally specified in any work we coded. We close with open problems, each with a suggested first step.
Chinese Translation
自动化事实核查系统返回一个标签：该声明为真，或为假。在许多此类系统中，裁决仍然是主要输出。通常缺失的是记录哪个文档解决了问题、为了改变裁决需要有什么不同，或者同样的声明如果换一种措辞是否会被判断为相同。我们称缺失的部分为保证（warrant）：一个关于所保证内容及其依据的独立声明。形式化方法能够提供这种证据，而监管也开始要求这种证据，因为《数字服务法案》和《人工智能法案》都要求关于系统行为的可审计证据。自动化事实核查的调查通常按管道阶段组织，并将逻辑视为众多技术之一。我们则根据被形式化的内容组织该领域，形成五个层次：声明、推理、执行核查的系统、声明传播的生态系统以及监管义务。将121项工作分类到这些层次中，有两个模式突出。大多数相关的形式化工具已经存在，但它们是为其他领域构建的，且在此领域很少被应用，而验证核查系统本身的差距最大。专业事实核查员遵循的几个阶段也没有明确的正确性标准，其中两个阶段，即将声明写成可核查的形式和更正已发布的裁决，在我们编码的任何工作中都没有正式规定。我们以开放问题结束，每个问题都有建议的第一步。
cs.CL / 110 / 2609.23264
Judging a Review by its Cover: A Reliability Analysis of LLM-based Peer Review Evaluation Metrics
以封面评判评审：基于LLM的同行评审评价指标的可靠性分析
Abstract
Peer-review evaluation is increasingly being automated with LLM-as-a-judge metrics, but this creates a measurement risk. A review may receive a high score because it is fluent, organized, and polished, rather than because it provides a strong evaluation of the paper. This risk is especially important in AI-assisted reviewing, where reviewers may use LLMs to improve clarity or presentation while preserving the underlying judgments. We propose a statistical framework for testing whether peer-review evaluation metrics capture substantive review quality beyond surface-level linguistic form. The framework compares original human reviews with faithful LLM rewrites that preserve the same evaluative content while changing wording and presentation. Using a dataset comprising 4,044 meaning-preserving rewrites derived from 674 human reviews from ICLR and NeurIPS, we evaluate 29 content-oriented peer-review evaluation metrics drawn from four prior works through complementary tests of surface sensitivity and robustness. Although these metrics are intended to capture review properties beyond surface-level, writing-dependent characteristics, we find that sensitivity to rewriting is widespread. Under our primary analysis, 23 metrics assign significantly different scores to reviews whose evaluative content is preserved, while only six satisfy our robustness criterion. The patterns are largely consistent across two LLM judge models, suggesting that the issue is not specific to a single judge. These findings show that many peer-review evaluation metrics partially conflate review quality with linguistic presentation, and indicate that robustness to meaning-preserving rewriting should be validated before such metrics are used to compare human-written, AI-assisted, and AI-generated reviews.
Chinese Translation
同行评审评价正日益依赖于基于LLM（大型语言模型）的评判指标，但这带来了测量风险。一篇评审可能因为其流畅、组织良好和精炼而获得高分，而不是因为它对论文提供了强有力的评价。这一风险在AI辅助评审中尤为重要，因为评审者可能会使用LLM来提高清晰度或表现，同时保留基础判断。我们提出了一个统计框架，用于测试同行评审评价指标是否捕捉到超越表面语言形式的实质性评审质量。该框架将原始人类评审与忠实的LLM重写进行比较，后者在改变措辞和表现的同时保留相同的评价内容。使用一个包含4,044个意义保留重写的数据库，这些重写源自674个来自ICLR和NeurIPS的人类评审，我们通过对29个内容导向的同行评审评价指标进行补充的表面敏感性和鲁棒性测试，评估了这29个指标。尽管这些指标旨在捕捉超越表面水平的评审特性，但我们发现对重写的敏感性普遍存在。在我们的主要分析中，23个指标对保留评估内容的评审分配了显著不同的分数，而只有六个指标满足我们的鲁棒性标准。这些模式在两个LLM评判模型中大致一致，表明该问题并不特定于单一评判者。这些发现表明，许多同行评审评价指标部分混淆了评审质量与语言表现，并指出在使用这些指标比较人类撰写、AI辅助和AI生成的评审之前，应验证其对意义保留重写的鲁棒性。
cs.CL / 111 / 2609.23267
Knowing When to Trust Images: Reliability-Aware Multi-modal Entity Alignment
了解何时信任图像：可靠性意识的多模态实体对齐
Abstract
The visual modality, i.e., images, plays a key role in multi-modal entity alignment (MMEA). Existing approaches often directly fuse the image with other modalities to align different entities. Although simple, such strategies overlook the potential noise in the images and their semantic misalignment with corresponding entities, resulting in suboptimal fusion and degraded performance. Addressing this, we propose a novel Reliability-Aware framework for MMEA (RA-MMEA), which assesses visual reliability and adaptively improves unreliable visual representations for robust entity alignment. The core lies in two modules, including dependency-aware visual reliability prediction (DA-VRP) and stability-regularized visual embedding generation (SR-VEG). The former aims to estimate the reliability of an image by leveraging multi-modal dependency within the entity, while the latter focuses on producing alternative visual representation conditioned on semantics encoded in textual modalities for multi-modal fusion. Compared to current methods, RA-MMEA enables more reliable visual representations for modality fusion, thereby improving performance. In extensive experiments, RA-MMEA achieves state-of-the-art results, verifying the importance of reliable visual modality for entity alignment and the effectiveness of RA-MMEA. The code and results will be released.
Chinese Translation
视觉模态，即图像，在多模态实体对齐（MMEA）中发挥着关键作用。现有的方法通常直接将图像与其他模态融合，以对齐不同的实体。尽管这种方法简单，但它忽视了图像中的潜在噪声及其与相应实体的语义不对齐，导致融合效果不佳和性能下降。为了解决这个问题，我们提出了一种新颖的可靠性意识框架用于多模态实体对齐（RA-MMEA），该框架评估视觉可靠性，并自适应地改善不可靠的视觉表示，以实现稳健的实体对齐。其核心包括两个模块：依赖意识的视觉可靠性预测（DA-VRP）和稳定性正则化的视觉嵌入生成（SR-VEG）。前者旨在通过利用实体内的多模态依赖关系来估计图像的可靠性，而后者则专注于生成基于文本模态编码的语义条件下的替代视觉表示，以实现多模态融合。与当前方法相比，RA-MMEA能够提供更可靠的视觉表示以进行模态融合，从而提高性能。在广泛的实验中，RA-MMEA取得了最先进的结果，验证了可靠视觉模态在实体对齐中的重要性以及RA-MMEA的有效性。代码和结果将会发布。
cs.CL / 112 / 2609.23371
Machine-Interpretable Information: Compiling Documents into Searchable and Readable Protocol States
机器可解释信息：将文档编译为可搜索和可读的协议状态
Abstract
Long-context language models interface with external knowledge through raw natural language. In retrieval-augmented systems, this creates a persistent index-payload schism: dense vectors enable searchable routing, but models must re-ingest lengthy text payloads for reasoning at O(N^2) attention cost. Existing compression methods further produce private states tied to specific architectures. We introduce Machine-Interpretable Information (MII), the first agent-to-agent (A2A) document-to-state protocol. A dual-timescale state-space Writer compiles documents into a canonical, fixed-bandwidth state (56 tokens), and a lightweight Translator maps it into any frozen Reader's embedding space, reducing query-time cost to O(K). The resulting .mii artifact unifies Retrieval (searchable geometry), Reasoning (global memory), and Reconstruction (grounded details) in a single transferable medium. We demonstrate strong cross-model interoperability across heterogeneous LLMs (e.g., Llama, Qwen, Mistral) -- despite the Writer using a legacy GPT-2 vocabulary, forcing genuine semantic translation rather than token-level memorization. Mechanistic probes reveal modular latent structure: entity representations can be causally traced and zero-shot transplanted between unrelated document states while remaining decodable. To address lexical reconstruction under fixed bandwidth, we propose Residual-MII, a cache hierarchy combining compiled global memory with sparse local evidence. On HotpotQA (7,405 queries), Residual-MII exceeds full-context Exact Match at approximately 7% of the attention FLOPs, suggesting a paradigm shift toward compiled, transferable neural document formats.
Chinese Translation
长上下文语言模型通过原始自然语言与外部知识进行交互。在增强检索的系统中，这造成了一个持久的索引-负载分裂：密集向量使得可搜索路由成为可能，但模型必须以 O(N^2) 的注意力成本重新摄取冗长的文本负载以进行推理。现有的压缩方法进一步产生与特定架构相关的私有状态。我们引入了机器可解释信息（Machine-Interpretable Information, MII），这是第一个代理对代理（agent-to-agent, A2A）文档到状态的协议。一个双时间尺度状态空间的写入器将文档编译为一个规范的、固定带宽的状态（56个标记），而一个轻量级的翻译器将其映射到任何冻结的阅读器的嵌入空间，从而将查询时间成本降低到 O(K)。生成的 .mii 产物在一个可转移的介质中统一了检索（可搜索几何）、推理（全局记忆）和重构（扎根细节）。我们展示了在异构大型语言模型（如 Llama、Qwen、Mistral）之间的强大跨模型互操作性——尽管写入器使用的是遗留的 GPT-2 词汇，迫使进行真正的语义翻译，而非标记级的记忆。机制探测揭示了模块化的潜在结构：实体表示可以在无关文档状态之间因果追踪并零-shot 移植，同时保持可解码性。为了解决在固定带宽下的词汇重构问题，我们提出了 Residual-MII，一种结合了编译全局记忆和稀疏局部证据的缓存层次结构。在 HotpotQA（7,405 个查询）上，Residual-MII 在大约 7% 的注意力 FLOPs 下超过了全上下文精确匹配，暗示着向编译的、可转移的神经文档格式的范式转变。
cs.CL / 113 / 2609.23463
Perplexity Predicts Protection: Choosing Pretrained Backbones for Worst-Client Fairness in Federated Parameter-Efficient Fine-Tuning
困惑度预测保护：选择预训练骨干网络以实现联邦参数高效微调中的最差客户端公平性
Abstract
Federated learning lets multiple parties train a shared model without pooling their data, but a client with far less data than the others can end up poorly served even when the group's average accuracy looks fine. We ask whether the choice of pretrained backbone affects this under LoRA fine-tuning, and whether per-word perplexity on the target text predicts which backbone helps the worst-off client before federated training starts. We ran 313 experiments across three text-classification datasets and three similarly sized backbones (RoBERTa, BERTweet, PubMedBERT), each compared against a task-specific baseline on identical data splits. Lower-perplexity backbones consistently produced larger gains for the worst-performing client, with a rank correlation of -0.87 across nine dataset-backbone pairs; a backbone held out of the analysis confirmed the pattern. Personalization with Ditto recovered only 4-12% of the gap between training alone and full federation, and removing aggregation entirely erased the benefit. A client's update also showed no sign of conflicting with the group's update; the two are close to orthogonal, ruling out one proposed explanation for this failure. Practically: measure perplexity on a sample of task text before choosing a backbone, and do not rely on personalization to protect a data-poor client. We release our code, predictions, and full results for others to test.
Chinese Translation
联邦学习允许多个参与方在不集中数据的情况下训练共享模型，但当某个客户端的数据远少于其他客户端时，即使该组的平均准确率看起来良好，该客户端也可能受到不良服务。我们探讨在LoRA微调中，预训练骨干网络的选择是否会影响这一点，以及目标文本的逐词困惑度是否能够预测在联邦训练开始之前，哪个骨干网络能帮助最弱势的客户端。我们在三个文本分类数据集和三个大小相似的骨干网络（RoBERTa、BERTweet、PubMedBERT）上进行了313次实验，每个骨干网络与在相同数据划分下的任务特定基线进行了比较。低困惑度的骨干网络始终为表现最差的客户端带来了更大的收益，在九个数据集-骨干网络对中，相关系数为-0.87；一个未纳入分析的骨干网络证实了这一模式。使用Ditto进行个性化仅弥补了训练单独与完全联邦之间4-12%的差距，而完全去除聚合则完全消除了这一好处。客户端的更新也没有显示出与组更新相冲突的迹象；两者几乎是正交的，这排除了对这一失败的一个提出的解释。从实践角度来看：在选择骨干网络之前，先在一份任务文本样本上测量困惑度，不要依赖个性化来保护数据稀缺的客户端。我们发布了我们的代码、预测和完整结果，以供他人测试。
cs.CL / 114 / 2609.23465
Propose, Verify, Commit: Evidence-Grounded Memory for Long-Horizon Multi-Actor Conversations
提出、验证、承诺：基于证据的长时段多参与者对话记忆
Abstract
Long-horizon conversational memory is especially challenging in multi-actor settings, where relevant evidence is distributed across participants and contexts and previously established information may later be revised. We introduce EGMEMORY, which formulates long-horizon multi-actor memory as a searchable state machine that separates persistent message-level evidence from an explicit active state. At write time, adaptive state resolution and an evidence-grounded propose-verify-commit protocol govern how this state evolves. At read time, adaptive evidence navigation iteratively resolves the state and supporting evidence required for a query, using conversational structure to narrow the search space and lexical-semantic relevance to rank candidates. The system operates through prompting and tool use without memory-specific policy training. EGMEMORY achieves 68.2% on GroupMemBench and 77.9% on EverMemBench, outperforming the strongest evaluated baselines by 22.7 and 21.4 percentage points, respectively. It further reaches 73.6% on the dyadic LoCoMo benchmark, demonstrating generalization beyond multi-actor conversations. We will release the codebase upon formal publication.
Chinese Translation
在多参与者环境中，长时段对话记忆尤其具有挑战性，因为相关证据分布在参与者和上下文之间，且先前建立的信息可能会被后续修订。我们提出了EGMEMORY，它将长时段多参与者记忆形式化为一个可搜索的状态机，区分持久的消息级证据与显式的活动状态。在写入时，自适应状态解析和基于证据的提出-验证-承诺协议决定了该状态如何演变。在读取时，自适应证据导航迭代解析查询所需的状态和支持证据，利用对话结构来缩小搜索空间，并通过词汇-语义相关性对候选项进行排序。该系统通过提示和工具使用进行操作，而无需特定于记忆的策略训练。EGMEMORY在GroupMemBench上达到了68.2%的成绩，在EverMemBench上达到了77.9%的成绩，分别比最强的评估基线高出22.7和21.4个百分点。此外，它在双人LoCoMo基准上达到了73.6%的成绩，展示了其超越多参与者对话的泛化能力。我们将在正式出版后发布代码库。
cs.CL / 115 / 2609.23466
RPMem: Learning Long-Term Recurrent Parametric Memory Across Sessions for LLM Agents
RPMem：为大型语言模型代理学习跨会话的长期递归参数记忆
Abstract
Long-running LLM agents require memory that persists and evolves across sessions. Text-based memory retrieves and reconstructs past interactions at every query, making long-horizon performance increasingly dependent on retrieval quality and contextual reasoning as histories grow. Parametric memory encodes experience directly into model computation, but existing approaches provide limited support for cross-session memory evolution. Their coupling to a specific backbone further restricts memory reuse after model replacement. We introduce RPMem, a two-stage architecture that compiles each session into a model-independent latent memory through forward computation and selectively integrates it with retained memory via a task-trained recurrent gate. The consolidated memory is then mapped to backbone-specific low-rank adaptation (LoRA) parameters, allowing the encoding capability to transfer when the backbone is replaced. Evaluation across three long-term memory benchmarks and five diverse backbones demonstrates broad generalization with near-constant update cost and memory footprint. With Qwen3-8B on PERMA, RPMem reaches 85.52%, outperforming the strongest parametric and text-based baselines by 5.32 and 12.98 percentage points, respectively. Ablations validate the complementary roles of session compilation and cross-session consolidation, while dynamics analyses reveal that the gate acquires task-specific memory integration strategies. These results establish RPMem as a lifecycle-independent parametric memory framework that maintains evolving cross-session memory that remains reusable across backbone replacements. Our implementation is available at https://github.com/Quark-Medical/rpmem/tree/main.
Chinese Translation
长期运行的大型语言模型（LLM）代理需要在会话之间持续演变的记忆。基于文本的记忆在每次查询时检索和重建过去的互动，使得随着历史的增长，长期表现越来越依赖于检索质量和上下文推理。参数记忆直接将经验编码到模型计算中，但现有方法对跨会话记忆演变的支持有限。它们与特定的主干网络耦合，进一步限制了在模型替换后记忆的重用。我们提出了RPMem，一种两阶段架构，通过前向计算将每个会话编译成模型无关的潜在记忆，并通过任务训练的递归门选择性地将其与保留的记忆集成。然后，将整合的记忆映射到主干特定的低秩适应（LoRA）参数，从而在替换主干时允许编码能力的转移。在三个长期记忆基准和五个不同主干上的评估表明，RPMem具有广泛的泛化能力，更新成本和内存占用几乎保持不变。在PERMA上使用Qwen3-8B，RPMem达到了85.52%，分别比最强的参数和基于文本的基线高出5.32和12.98个百分点。消融实验验证了会话编译和跨会话整合的互补作用，而动态分析揭示了门控获取任务特定的记忆集成策略。这些结果确立了RPMem作为一个生命周期无关的参数记忆框架，能够维持不断演变的跨会话记忆，并在主干替换中保持可重用性。我们的实现可在 https://github.com/Quark-Medical/rpmem/tree/main 获取。
cs.CL / 116 / 2609.23490
BabelArena: A Large-Scale Multilingual Benchmark for LLM Agents
BabelArena：一个大规模多语言基准测试用于大语言模型代理
Abstract
Large language model (LLM) agents increasingly execute multi-step workflows through tool use and interaction with users and environments. However, current agent evaluations are largely English-centric, limiting our understanding of agent capabilities in multilingual settings. We introduce BabelFlow, a benchmark-general agentic workflow that adapts existing agent benchmarks to new languages by analyzing runtime dependencies, coordinating structure-preserving translation, and combining multi-layer verification with human review to preserve task and evaluation semantics. Using BabelFlow, we construct BabelArena, a task-aligned benchmark comprising 16,146 instances derived from 702 canonical tasks across four benchmark families, 13 domains, and 23 languages. Experiments with five frontier models show that no single model dominates across benchmark families and that cross-language disparities extend well beyond task success. Lower-resource languages exhibit distinct failure patterns, with larger shares of tool-use and control-flow errors rather than answer-quality errors alone, pointing to gaps in reliable task execution across the resource levels of these languages. On the same tasks, agents in low-resource languages also consume substantially more tokens than in English (up to roughly twice the input) without proportional increases in interaction length, and language consistency degrades further on tasks requiring structured output, where switches are directed overwhelmingly toward English. We believe BabelArena provides a foundation for advancing research on reliable and efficient multilingual agents.
Chinese Translation
大型语言模型（LLM）代理越来越多地通过工具使用和与用户及环境的互动来执行多步骤工作流程。然而，目前的代理评估主要集中在英语上，这限制了我们对代理在多语言环境中能力的理解。我们引入了BabelFlow，一个基准通用的代理工作流程，通过分析运行时依赖关系、协调结构保持翻译，并结合多层验证与人工审查，以保持任务和评估语义，从而将现有代理基准适配到新语言。利用BabelFlow，我们构建了BabelArena，这是一个任务对齐的基准，包含来自四个基准系列、13个领域和23种语言的702个经典任务衍生出的16,146个实例。对五个前沿模型的实验表明，没有单一模型在所有基准系列中占据主导地位，并且跨语言差异远远超出了任务成功率。低资源语言表现出明显的失败模式，工具使用和控制流错误的比例较大，而不仅仅是答案质量错误，指出这些语言在可靠任务执行方面的资源水平存在差距。在相同任务中，低资源语言的代理消耗的标记数量也显著高于英语（输入量几乎是英语的两倍），而交互长度并没有相应增加，并且在需要结构化输出的任务上，语言一致性进一步下降，切换主要集中在英语上。我们相信BabelArena为推进可靠和高效的多语言代理研究提供了基础。
cs.CL / 117 / 2609.23551
Paragraph Boundaries Are Not White Space:Compression Depth as the Signature of Hierarchical Structure
段落边界不是空白：压缩深度作为层次结构的特征
Abstract
Standard positional encodings represent position as a one-dimensional reading-order coordinate, but reading order alone does not determine hierarchical textual structure. We use a hierarchical rotary positional encoding (hRoPE) that represents paragraph, sentence, and token indices as separate channels, hold the token sequence fixed, intervene on the paragraph coordinate p1, and measure cross-paragraph attention with a token-distance-exact estimator. Attention is compressed relative to a token-distance-matched baseline in every corpus, but compression alone is not diagnostic of true structure: an architecturally identical channel with density-matched random labels is compressed too, more shallowly. What distinguishes real structure is the depth of compression, which is greater and corpus-dependent while the control's is not. Comparing eight corpus-only quantities across three constructs (lexical persistence, paragraph length, embedding-based coherence), none fully reproduces the cross-corpus ordering of depth, though embedding-based coherence comes closest. Compression depth, not its location, is the reproducible signature of genuine paragraph structure in our setting.
Chinese Translation
标准的位置编码将位置表示为一维的阅读顺序坐标，但仅靠阅读顺序并不能决定层次文本结构。我们使用一种层次旋转位置编码（hierarchical rotary positional encoding，hRoPE），将段落、句子和标记索引表示为不同的通道，固定标记序列，干预段落坐标 p1，并使用标记距离精确估计器测量跨段落注意力。在每个语料库中，注意力相对于标记距离匹配的基线被压缩，但仅靠压缩并不能诊断真实结构：一个在架构上相同、密度匹配的随机标签通道也会被压缩，但压缩程度更浅。真正结构的区别在于压缩的深度，这种深度更大且依赖于语料库，而控制组的深度则不然。在三个构念（词汇持久性、段落长度、基于嵌入的一致性）中比较八个仅基于语料库的量度，没有一个能够完全重现深度的跨语料库排序，尽管基于嵌入的一致性最接近。压缩深度，而不是其位置，是我们研究中真实段落结构的可重复特征。
cs.CL / 118 / 2609.23573
Error-Supervised Synthetic Learner Writing for Automated Essay Scoring
基于错误监督的合成学习者写作用于自动化作文评分
Abstract
Synthetic essays can help reduce dependence on human-written data in Automated Essay Scoring (AES). However, they often lack realistic errors, limiting their ability to represent authentic human writing, particularly when the target texts are intended to resemble those produced by language learners. In this study, we present a simple approach that introduces error supervision into synthetic essay generation. Specifically, we fine-tune an LLM generator on error-annotated texts of the kind commonly used in Grammatical Error Detection (GED). To assess the utility of the proposed approach, we fine-tune and evaluate AES scorers under three data conditions: authentic essays, synthetic essays generated conventionally, and synthetic essays generated using our proposed approach. The results show that in the larger-data settings, the proposed approach outperforms the conventional synthetic baseline in 11 out of 12 dataset-metric comparisons, with performance in some cases approaching that of models trained on authentic essays. Despite these gains, performance under extremely low-resource settings remains mixed, with advantages over the conventional baseline only becoming more apparent at 200 training essays, although not consistently across datasets. Qualitative and quantitative analyses further show that the proposed approach produces learner-like errors whose distributions broadly resemble those observed in authentic essays.
Chinese Translation
合成作文可以帮助减少在自动化作文评分（AES）中对人类书写数据的依赖。然而，它们往往缺乏真实的错误，限制了它们代表真实人类写作的能力，特别是当目标文本旨在类似于语言学习者所写的内容时。在本研究中，我们提出了一种简单的方法，将错误监督引入合成作文生成中。具体而言，我们在常用于语法错误检测（GED）的错误注释文本上微调了一个大型语言模型（LLM）生成器。为了评估所提方法的有效性，我们在三种数据条件下微调并评估了AES评分器：真实作文、传统生成的合成作文，以及使用我们提出的方法生成的合成作文。结果表明，在大数据设置中，所提方法在12个数据集-指标比较中有11个超过了传统的合成基线，在某些情况下，其性能接近于在真实作文上训练的模型。尽管取得了这些进展，但在极低资源设置下的表现仍然不一，只有在200篇训练作文时，相较于传统基线的优势才变得更加明显，尽管在不同数据集之间并不一致。定性和定量分析进一步表明，所提方法产生的错误类似于学习者的错误，其分布大致与真实作文中观察到的分布相似。
cs.CL / 119 / 2609.23582
ARID: A Deployable Edge AI System for Structured Information Extraction from Industrial Maintenance Work Orders
ARID：一种可部署的边缘人工智能系统，用于从工业维护工单中提取结构化信息
Abstract
Maintenance work orders must often be processed offline on embedded hardware, yet downstream software requires predictable structured output. We present ARID (Aviation-inspired Routing for Industrial Deployment), which extracts component, failure mode, symptom, and maintenance action into fixed-schema JSON on an 8 GB NVIDIA Jetson Orin NX. ARID combines conservative dual-teacher filtering, targeted noise-aware synthesis, one routing decision per work order, 4-bit inference, and grammar-constrained decoding. From 2,326 unlabeled OMIn records, it retains 716 training pairs and adds 99 topology-constrained records targeting action extraction. On 300 human-labeled records, ARID reaches 84.8% token-F1 on the reference stack and 82.9% on the deployed Jetson. Resident serving achieves 5,310/5,656 ms P50/P99 at 12.5 W. On zero-shot MaintNet transfer, semantic F1 falls to 46.4% while parser success remains at least 99.8%, showing that output validity transfers but field semantics do not.
Chinese Translation
维护工单通常需要在嵌入式硬件上离线处理，但下游软件要求可预测的结构化输出。我们提出了ARID（航空启发的工业部署路由），该系统能够将组件、故障模式、症状和维护行动提取为固定模式的JSON格式，运行于8 GB的NVIDIA Jetson Orin NX上。ARID结合了保守的双教师过滤、针对噪声的有意识合成、每个工单一个路由决策、4位推理和语法约束解码。从2326个未标记的OMIn记录中，它保留了716个训练对，并增加了99个针对行动提取的拓扑约束记录。在300个人工标注的记录上，ARID在参考堆栈上达到了84.8%的token-F1，在部署的Jetson上达到了82.9%。驻留服务在12.5 W下实现了5,310/5,656 ms的P50/P99。在零样本的MaintNet迁移中，语义F1降至46.4%，而解析器的成功率保持在至少99.8%，显示输出有效性可以迁移，但领域语义无法迁移。
cs.CL / 120 / 2609.23587
On the Efficiency-Safety Dilemma in Large Reasoning Models
大型推理模型中的效率-安全困境
Abstract
Large reasoning models (LRMs) incur high inference costs, often mitigated by efficiency techniques like quantization and pruning. However, the impact of these techniques on model adversarial robustness remains largely unexplored. This study provides the first comprehensive analysis of the interplay between efficiency, jailbreak vulnerability, and reasoning in LRMs. We find that while efficiency methods seemingly reduce the success rate of jailbreak attacks, this improvement is often superficial. It largely arises from degraded reasoning capabilities leading to "attempted but failed" malicious responses, rather than an increase in genuine alignment. Mechanistic analysis of representational drift confirms this, revealing a strict coupling between reasoning capability loss and the model's inability to maintain malicious semantic trajectories. Additionally, we identify quantization with pruning as the optimal strategy to balance efficiency and robustness. These findings clarify the distinction between true safety alignment and capability-induced failure, providing an empirical foundation for LRM deployment.
Chinese Translation
大型推理模型（LRMs）在推理过程中产生高昂的成本，通常通过量化和剪枝等效率技术来缓解。然而，这些技术对模型对抗鲁棒性的影响尚未得到充分探讨。本研究首次全面分析了效率、越狱脆弱性和LRMs中的推理之间的相互关系。我们发现，尽管效率方法似乎降低了越狱攻击的成功率，但这种改善往往是表面的。它主要源于推理能力的下降，导致“尝试但失败”的恶意响应，而不是对真正对齐的提升。对表征漂移的机制分析证实了这一点，揭示了推理能力损失与模型无法维持恶意语义轨迹之间的严格耦合。此外，我们确定了量化与剪枝的结合是平衡效率与鲁棒性的最佳策略。这些发现澄清了真正的安全对齐与能力引发的失败之间的区别，为LRM的部署提供了实证基础。
cs.CL / 121 / 2609.23697
Distill What You Trust: Reliability-Aware Multi-Teacher On-Policy Distillation
提炼你所信任的：可靠性意识的多教师在线蒸馏
Abstract
Multi-teacher on-policy distillation allows a student to learn from complementary specialists on its own trajectories. Domain-routed approaches, however, select one teacher per example and keep it fixed throughout the response. This design both depends on labels that mixed training corpora often lack and cannot adapt teacher selection when the expertise required changes within a trajectory. We propose \textbf{TrustMOPD}, which replaces example-level teacher selection with label-free, token-level supervision allocation. At each student-generated prefix, TrustMOPD uses each specialist's RL-induced displacement from a shared pre-RL reference as a proxy for local reliability, calibrates these scores across teachers, and constructs a weighted distillation target. Across mathematics, code, and instruction following, TrustMOPD outperforms the strongest label-free baseline, increasing the recovery ratio from $54.4\%$ to $91.5\%$ on \textsc{SingleCap} and from $54.5\%$ to $98.0\%$ on \textsc{MultiCap}, while approaching label-based MOPD on \textsc{SingleCap}. Randomizing token-level weights independently of the student-generated prefix performs no better than uniform weighting, supporting the importance of conditioning supervision on the evolving generation context.
Chinese Translation
多教师在线蒸馏允许学生在自身轨迹中从互补的专家那里学习。然而，领域路由方法每个示例仅选择一个教师，并在整个响应过程中保持不变。这种设计依赖于混合训练语料库通常缺乏的标签，并且无法在轨迹中所需的专业知识变化时调整教师选择。我们提出了 extbf{TrustMOPD}，它用无标签的、基于令牌的监督分配替代了示例级教师选择。在每个学生生成的前缀中，TrustMOPD使用每个专家从共享的预强化学习参考中引起的位移作为局部可靠性的代理，校准这些分数并构建加权蒸馏目标。在数学、代码和指令跟随方面，TrustMOPD超越了最强的无标签基线，在 extsc{SingleCap}上的恢复率从$54.4\%$提高到$91.5\\%$，在 extsc{MultiCap}上的恢复率从$54.5\\%$提高到$98.0\\%$，同时在 extsc{SingleCap}上接近基于标签的MOPD。独立于学生生成的前缀随机化令牌级权重的表现不优于均匀加权，支持了在不断演变的生成上下文中条件化监督的重要性。
cs.CL / 122 / 2609.23703
Financial Language Models as Applied Artificial Intelligence Systems for News-Based Trading under Market Frictions
金融语言模型作为应用人工智能系统在市场摩擦下的基于新闻的交易
Abstract
Financial language models can transform unstructured firm-specific news into structured decision signals, but financial AI research lacks an integrated deployment framework for evaluating whether those signals remain useful in financial decision systems. Computer science research has developed strong methods for time-series forecasting, text classification, multimodal stock prediction, graph-based market modeling, and machine-learning operations, yet these streams do not provide a domain-specific protocol that jointly tests financial language-model outputs under event-time observability, probability calibration, execution timing, transaction costs, liquidity constraints, capacity limits, operational diagnostics, and statistical inference. We introduce MFAST, a Market-Friction-Aware Sentiment-to-Trading framework that converts timestamped financial text into auditable, reproducible, and market-feasible trading decisions. The application is news-based trading, where firm-specific text must be linked to securities before portfolio decisions can be evaluated. The framework links Refinitiv News Analytics to Center for Research in Security Prices (CRSP) equity data, restricts the primary out-of-sample evaluation to post-release news outside disclosed foundation-model data-freshness periods, and adds a public replication arm using open financial text and public price data. Results show that decoder-only language models outperform encoder baselines and dictionary sentiment in classification, calibration, return prediction, and net portfolio performance, while operational diagnostics reveal trade-offs among accuracy, latency, memory, throughput, and inference cost. The paper shows that credible evaluation of financial language models requires an end-to-end engineering approach combining language understanding, temporal discipline, market-friction-aware deployment, and reproducible validation.
Chinese Translation
金融语言模型能够将非结构化的公司特定新闻转化为结构化的决策信号，但金融人工智能研究缺乏一个综合的部署框架来评估这些信号在金融决策系统中的有效性。计算机科学研究已经开发出强大的时间序列预测、文本分类、多模态股票预测、基于图的市场建模和机器学习操作方法，然而这些研究流并未提供一个领域特定的协议，能够在事件时间可观测性、概率校准、执行时机、交易成本、流动性约束、容量限制、操作诊断和统计推断等方面共同测试金融语言模型的输出。我们提出了MFAST（市场摩擦感知情绪到交易框架），该框架将带时间戳的金融文本转化为可审计、可重复和市场可行的交易决策。该应用为基于新闻的交易，其中公司特定文本必须与证券相关联，才能评估投资组合决策。该框架将Refinitiv新闻分析与证券价格研究中心（CRSP）股票数据连接起来，将主要的样本外评估限制在发布新闻后的数据，而这些数据在公开的基础模型数据新鲜度周期之外，并增加了一个使用开放金融文本和公共价格数据的公共复制部分。结果表明，仅解码器的语言模型在分类、校准、收益预测和净投资组合表现方面优于编码器基线和字典情绪，而操作诊断则揭示了准确性、延迟、内存、吞吐量和推断成本之间的权衡。本文表明，对金融语言模型的可信评估需要一种端到端的工程方法，结合语言理解、时间纪律、市场摩擦感知的部署和可重复的验证。
cs.CL / 123 / 2609.23716
STEVE: Stabilizing Textual Gradient-Based Prompt Optimization via Error-Driven Refinement and Regularized Verification
STEVE：通过误差驱动的精炼和正则化验证稳定文本梯度基础的提示优化
Abstract
Textual-gradient methods automate prompt optimization through natural-language feedback, but their iterative updates can be unstable. We identify two sources of this instability: noisy gradients produced from already-correct examples and over-specialization to hard cases that degrades performance on simpler inputs. We introduce STEVE, a stabilization framework with two coupled mechanisms. Error-Driven Refinement generates gradients only from incorrectly handled examples, concentrating updates on informative failures. Regularized Verification treats every update as provisional and accepts it only when improvement on hard cases does not cause unacceptable regression on a preservation set. Across ten reasoning benchmarks, three evaluator/optimizer models, and established prompt-optimization baselines, STEVE reduces degradation and produces more robust prompts. Additional evaluations with gpt-5.4-mini/gpt-5.4 on symbolic reasoning, GSM8K-Platinum, and DS-1000 show that these gains persist with newer models and larger test sets. STEVE therefore provides a practical way to improve the stability and effectiveness of textual-gradient prompt optimization.
Chinese Translation
文本梯度方法通过自然语言反馈自动化提示优化，但其迭代更新可能不稳定。我们识别出这种不稳定性的两个来源：来自已经正确示例的噪声梯度和对困难案例的过度专业化，这会降低在简单输入上的性能。我们提出了STEVE，一个具有两个耦合机制的稳定框架。误差驱动的精炼仅从处理不当的示例中生成梯度，集中更新在信息丰富的失败上。正则化验证将每次更新视为临时，并仅在对困难案例的改进不会导致在保留集上的不可接受回退时才接受该更新。在十个推理基准、三个评估/优化模型和已建立的提示优化基线中，STEVE减少了性能下降并生成了更稳健的提示。与gpt-5.4-mini/gpt-5.4在符号推理、GSM8K-Platinum和DS-1000上的额外评估表明，这些收益在更新模型和更大测试集上依然存在。因此，STEVE提供了一种实用的方法来提高文本梯度提示优化的稳定性和有效性。
cs.CL / 124 / 2609.23726
GRACE: Grounded Adversarial Reasoning over Canadian Law
GRACE：针对加拿大法律的基础对抗推理
Abstract
Large language models have shown strong performance across a range of legal tasks, but existing benchmarks rarely evaluate the ability to take and defend a legal position, reason under incomplete information, or synthesize multiple statutory provisions. This gap is particularly pronounced for Canadian law, which remains underrepresented in legal NLP. We introduce GRACE (Grounded Reasoning Adversarial Canadian LEgal examples), a dataset of 1,915 question-reasoning-answer instances grounded in Canadian federal legislation. GRACE covers three reasoning modes: adversarial advocacy, uncertainty, and applied reasoning. We develop a pipeline that partitions raw statutory text, generates scenario-based questions and reasoning, and filters examples through model-free citation verification and LLM-based quality auditing. As a proof of concept, we fine-tune CLeAR-4B (Canadian Legal Adversarial Reasoning), a lightweight model for grounded legal reasoning, and evaluate it against the unmodified Qwen3-4B base model in open- and closed-book settings. CLeAR-4B substantially improves agreement with teacher outputs and statutory citation behavior when the relevant act text is provided, while its grounding degrades sharply when the statute is withheld. These results suggest that GRACE can support the development of lightweight legal models that reason more effectively from supplied statutory text.
Chinese Translation
大型语言模型在多种法律任务中表现出色，但现有基准测试很少评估在不完整信息下采取和辩护法律立场的能力，或综合多个法条的能力。这一差距在加拿大法律中尤为明显，因为其在法律自然语言处理（NLP）领域仍然代表性不足。我们引入了GRACE（Grounded Reasoning Adversarial Canadian LEgal examples），这是一个基于加拿大联邦立法的1,915个问题-推理-答案实例的数据集。GRACE涵盖三种推理模式：对抗性辩护、不确定性和应用推理。我们开发了一个管道，能够对原始法条文本进行分割，生成基于场景的问题和推理，并通过无模型的引用验证和基于大型语言模型（LLM）的质量审计来过滤示例。作为概念验证，我们对CLeAR-4B（Canadian Legal Adversarial Reasoning）进行了微调，这是一种用于基础法律推理的轻量级模型，并在开放和闭卷环境中与未修改的Qwen3-4B基础模型进行了评估。当提供相关法案文本时，CLeAR-4B在与教师输出和法条引用行为的一致性上显著提高，而当法条被隐瞒时，其基础推理能力急剧下降。这些结果表明，GRACE可以支持开发更有效地从提供的法条文本中进行推理的轻量级法律模型。
cs.CL / 125 / 2609.23742
Constrained Decoding Eliminates Structural Failures in Small LLMs but Reveals a Scale-Dependent Semantic Gap
约束解码消除小型大语言模型中的结构性失败，但揭示了规模依赖的语义差距
Abstract
Small open-source large language models (LLMs) in the 0.6B-4B parameter range are increasingly deployed for structured output generation (JSON, function calling, data extraction), yet little is known about how constrained decoding (CD) interacts with model scale in this regime. We benchmark five models from three families across 14 structured-output tasks under three decoding conditions (native, Outlines, XGrammar). We introduce a two-axis evaluation that separates structural correctness (schema validity) from semantic correctness (content accuracy). We find that CD eliminates all structural failures across all models (schema validity: 78.6-92.9% to 100%), but content accuracy reveals a persistent semantic gap that is scale-dependent: type coercion failures are fully CD-rescuable, while instruction-semantic failures (e.g., multi-step function calling) remain CD-resistant. Schema conformance is necessary but not sufficient for semantic correctness; CD's reach ends exactly where schema conformance ends.
Chinese Translation
在0.6B-4B参数范围内的小型开源大语言模型（LLMs）越来越多地被用于结构化输出生成（如JSON、函数调用、数据提取），然而关于约束解码（CD）在这一范围内与模型规模的交互作用知之甚少。我们在三种解码条件（原生、Outlines、XGrammar）下，对来自三个家族的五个模型在14个结构化输出任务上进行了基准测试。我们引入了一个双轴评估方法，将结构正确性（模式有效性）与语义正确性（内容准确性）区分开来。我们发现，CD消除了所有模型中的结构性失败（模式有效性：78.6-92.9%提升至100%），但内容准确性揭示了一个持续的、依赖于规模的语义差距：类型强制失败完全可以通过CD恢复，而指令语义失败（例如，多步骤函数调用）则对CD具有抵抗力。模式符合性是语义正确性的必要条件，但并非充分条件；CD的有效性正好在模式符合性结束的地方终止。
cs.CL / 126 / 2609.23808
FLARE: A Full-Lifecycle Dense Supervision Paradigm for Long-Horizon Coding Agents via Generative Reward Model
FLARE：通过生成奖励模型实现长时间编码代理的全生命周期密集监督范式
Abstract
While test-time scaling enhances Large Language Model (LLM) agents in long-horizon software engineering (SWE), sparse binary rewards (Pass/Fail) create a severe credit assignment crisis and waste failed exploratory trajectories. Current trajectory optimization and scaling methods are costly and structurally limited, relying on heuristic state reuse without causal diagnosis or delayed scalar scoring without actionable online guidance. We propose FLARE (Full-Lifecycle Alignment and Reward Engine), a novel dense supervision paradigm driven by a lightweight Generative Reward Model (GRM). First, RADAR, an offline causal-aware diagnostic framework, extracts high-fidelity, hindsight-free supervision through causal-chain backtracking to distill a GRM providing real-time, step-level risk feedback. Second, FLARE uses this GRM to continuously optimize the agent across its entire lifecycle. During inference, FLARE acts as an Active Scaffold, autonomously intercepting high-risk generation steps for localized breakpoint re-execution, drastically reducing compute overhead. During post-training, the GRM's structured signals serve as process-supervised reranking scores for Supervised Fine-Tuning (SFT) and step-level dense rewards for Reinforcement Learning (RL), mitigating policy collapse in sparse environments. Extensive evaluations show that FLARE establishes a new Pareto frontier across the agent lifecycle: FLARE (N=1) outperforms Global Rollout (N=5) with a 5x reduction in token consumption. Extending FLARE to training overcomes the sparse reward problem in long-horizon interactive tasks, delivering relative performance gains of 19.13% in SFT through process-aware data curation and a consistent 9.19% improvement in RL.
Chinese Translation
虽然测试时扩展增强了大型语言模型（LLM）代理在长时间软件工程（SWE）中的表现，但稀疏的二元奖励（通过/失败）导致了严重的信用分配危机，并浪费了失败的探索轨迹。目前的轨迹优化和扩展方法成本高且结构有限，依赖于启发式状态重用而没有因果诊断，或依赖于没有可操作在线指导的延迟标量评分。我们提出了FLARE（全生命周期对齐与奖励引擎），一种由轻量级生成奖励模型（GRM）驱动的新型密集监督范式。首先，RADAR，一个离线因果意识诊断框架，通过因果链回溯提取高保真、无后见的监督，以提炼出提供实时、逐步风险反馈的GRM。其次，FLARE利用该GRM在代理的整个生命周期内持续优化。在推理过程中，FLARE作为一个主动支架，自主拦截高风险生成步骤以进行局部断点重新执行，显著减少计算开销。在后训练阶段，GRM的结构化信号作为过程监督的重新排序评分用于监督微调（SFT），并作为强化学习（RL）的逐步密集奖励，缓解了稀疏环境中的策略崩溃。广泛的评估表明，FLARE在代理生命周期内建立了新的帕累托前沿：FLARE（N=1）在标记消耗上比全球回放（N=5）减少了5倍。将FLARE扩展到训练克服了长时间交互任务中的稀疏奖励问题，通过过程意识数据策划在SFT中实现了19.13%的相对性能提升，并在RL中实现了一致的9.19%的改进。
cs.CL / 127 / 2609.23825
Federated Multilingual Speech-LLMs: Architecture and Aggregation Strategy Benchmarking
联邦多语言语音-大语言模型：架构与聚合策略基准测试
Abstract
We present a comprehensive benchmark of Federated Learning (FL) for multilingual Automatic Speech Recognition (ASR), evaluating four Speech-LLM architectures on the Multilingual LibriSpeech dataset. We compare FedAvg and FedProx across frozen and unfrozen encoder configurations, demonstrating that optimized learning rates are critical for performance. Specifically, independently tuning the learning rates for the speech encoder, connector, and decoder yields the lowest error rates, with full three-component adaptation (LoRA for encoder and decoder, full training for the connector) producing the best FL results. We observe that FedProx efficacy is architecture-dependent, providing notable advantages in multilingual pre-trained architectures (e.g., EuroLLM over TinyLlama when keeping the encoder fixed); this indicates that LLM backbone capacity plays a key role in mediating resilience to heterogeneous data distributions. These findings offer concrete design guidance for deploying multilingual Speech-LLMs in privacy-sensitive, distributed environments.
Chinese Translation
我们对多语言自动语音识别（ASR）的联邦学习（FL）进行了全面的基准测试，评估了四种语音大语言模型（Speech-LLM）架构在多语言LibriSpeech数据集上的表现。我们比较了FedAvg和FedProx在冻结和非冻结编码器配置下的效果，证明了优化学习率对性能至关重要。具体而言，独立调整语音编码器、连接器和解码器的学习率能够实现最低的错误率，其中完整的三组件适应（编码器和解码器的LoRA，连接器的全面训练）产生了最佳的FL结果。我们观察到，FedProx的有效性依赖于架构，在多语言预训练架构中（例如，固定编码器时的EuroLLM优于TinyLlama）提供了显著优势；这表明大语言模型的主干能力在调节对异构数据分布的韧性方面发挥了关键作用。这些发现为在隐私敏感的分布式环境中部署多语言Speech-LLMs提供了具体的设计指导。
cs.CL / 128 / 2609.23853
From UNDRR Reports to Event Records: Schema-Constrained LLM Extraction of Georeferenced Disasters
从联合国减灾署报告到事件记录：基于模式约束的大型语言模型提取地理参考灾害
Abstract
Disaster-risk-reduction archives describe hazard events in prose that databases such as EM-DAT (Delforge et al., 2025) cannot ingest directly. We present an LLM pipeline that generates candidate georeferenced event records using a controlled hazard vocabulary and fixed schema, retaining evidence for review. Applied to 10,000 documents from PreventionWeb, the knowledge hub managed by UNDRR, it produced 3,572 records from 1,913 documents across 24 hazard types and resolved 81% of location mentions to OpenStreetMap geometries. On 171 human-positive document windows from a stratified 217-document reference set, GPT-5 achieved 86.0% pooled attribute $F_1$, versus 44.2% for the spaCy-gazetteer baseline. Evaluation pools hazard families, location strings, and event years within documents, without assessing their assignment to individual events. GPT-5.4 ranked highest among ten LLMs (86.6% $F_1$). Verbatim evidence occurrence was 72.0% for GPT-5 and 47.2% for GPT-5.4, measuring textual traceability without establishing attribute support. We report production failure modes and automated label and location-rule compliance checks. Prompts, schema, and outputs will be released for adaptation to national reporting archives.
Chinese Translation
灾害风险减少档案以散文形式描述危险事件，这些内容无法直接被如EM-DAT（Delforge等，2025）等数据库吸收。我们提出了一种大型语言模型（LLM）管道，该管道使用受控的危险词汇和固定模式生成候选的地理参考事件记录，并保留证据以供审查。该方法应用于来自联合国减灾署管理的知识中心PreventionWeb的10,000份文档，生成了3,572条记录，涵盖1,913份文档和24种危险类型，并将81%的地点提及解析为OpenStreetMap几何形状。在从分层的217份文档参考集中的171个人工确认的文档窗口中，GPT-5实现了86.0%的汇总属性$F_1$，而spaCy-地名库基线为44.2%。评估池中包含文档内的危险类别、地点字符串和事件年份，但未评估它们与单个事件的对应关系。在十个大型语言模型中，GPT-5.4的排名最高（86.6% $F_1$）。GPT-5的逐字证据出现率为72.0%，而GPT-5.4为47.2%，测量文本可追溯性但未建立属性支持。我们报告了生产失败模式和自动标签及位置规则合规性检查。提示、模式和输出将被发布以适应国家报告档案。
cs.CL / 129 / 2609.23880
Q-TIE: A Lightweight and Generalizable Re-ranking Framework for Temporal Information Retrieval
Q-TIE：一种轻量级且可泛化的时间信息检索重排序框架
Abstract
Temporal Information Retrieval (TIR) has been increasingly critical given the rise of Retrieval-Augmented Generation (RAG). Since temporally mismatched evidence can be highly misleading, TIR aims to retrieve documents that are both semantically and temporally relevant to a query. Two TIR paradigms have emerged - temporal retrievers and temporal re-rankers - differing in how temporal relevance is modeled. While these paradigms provide complementary strengths, our analysis reveals that each alone falls short of robust TIR: temporal retrievers provide flexible query understanding via learned representations, but often fail to explicitly account for temporal constraints; temporal re-rankers can enforce such constraints more explicitly, but often rely on predefined re-ranking rules. To address this, we propose Q-TIE, a re-ranking framework based on learned Temporal Intent Extraction (TIE). By introducing a TIE model that maps each query's temporal constraint into a unified interval representation (i.e., $\langle t_{start}, t_{end} \rangle$), Q-TIE generalizes beyond predefined rules via model-based learning while explicitly modeling temporal constraints as a separate signal - jointly achieving what each paradigm typically trades off. Experiments demonstrate that Q-TIE consistently outperforms existing TIR methods with stronger generalizability across temporal query types, and provides a lightweight yet effective add-on for temporally-aware RAG pipelines. Code: https://github.com/ssoy0701/Q-TIE.
Chinese Translation
随着检索增强生成（Retrieval-Augmented Generation, RAG）的兴起，时间信息检索（Temporal Information Retrieval, TIR）变得愈发重要。由于时间上不匹配的证据可能会导致严重误导，TIR旨在检索与查询在语义和时间上均相关的文档。目前出现了两种TIR范式——时间检索器和时间重排序器——它们在时间相关性建模的方式上存在差异。尽管这些范式各自具有互补的优势，但我们的分析表明，单独使用任何一种都无法实现稳健的TIR：时间检索器通过学习的表示提供灵活的查询理解，但往往未能明确考虑时间约束；而时间重排序器能够更明确地强加这些约束，但通常依赖于预定义的重排序规则。为了解决这一问题，我们提出了Q-TIE，一个基于学习的时间意图提取（Temporal Intent Extraction, TIE）的重排序框架。通过引入一个TIE模型，将每个查询的时间约束映射到统一的区间表示（即，$ ext{t}_{start}, ext{t}_{end} $），Q-TIE在模型学习的基础上超越了预定义规则，同时将时间约束作为一个独立信号进行明确建模——共同实现了每种范式通常所需的权衡。实验表明，Q-TIE在不同时间查询类型上始终优于现有的TIR方法，并在时间感知的RAG管道中提供了一种轻量且有效的附加方案。代码：https://github.com/ssoy0701/Q-TIE。
cs.CL / 130 / 2609.23886
this-that-model-1.0: A typed decision model that decides in 30 ms, for a millionth of a cent
this-that-model-1.0：一种在30毫秒内做出决策的类型化决策模型，成本仅为千分之一美分
Abstract
Software delegates more of its branches to models every year: which queue a ticket enters, whether a command is safe to run, whether a claim clears without a person. What the program needs back is not prose. It is one of n declared options and a number it can threshold. Today that costs a round trip to a frontier model -- hundreds of milliseconds, a per-token bill, and a parser -- for a question that is usually a conjunction of three clauses. this-that-model-1.0 is a 2B-parameter typed decision model. Its answer is read directly from the hidden state at a designated position and restricted to the option set the caller declared, so no text is generated, nothing can be malformed, and every question in a request is answered in the same forward pass. It decides in 30.9 ms on one laptop GPU and generates zero output tokens doing it, where a frontier API call costs 8758 ms and the hosted systems that answer these questions well spend between 21 and 212 generated tokens per question thinking first, billed for every one. It sustains 32 decisions per second on one consumer GPU and never lets the state leave the machine. On a third party's recorded cohort of 68 decision questions, on their inputs and their wording, it scores 0.941 with a Brier score of 0.042, against 0.765 and 0.133 for the hosted service Jev on the same items. One pass of our 42-family internal suite takes 32 seconds and 0.000217 USD of electricity; the most accurate hosted model we measured needs 155.2 minutes and 10.636 USD. We also report where it loses. On multi-step arithmetic, which a single forward pass cannot carry intermediate results through, it scores 0.560 against their 0.98 to 1.00, and a targeted second training round improved the five task families it was written for and transferred to none of the other 13. The model is open-sourced in https://huggingface.co/flock-io/this-that-model-1.0
Chinese Translation
每年软件将更多的分支委托给模型：一个票据进入哪个队列，某个命令是否安全执行，某个索赔是否可以在没有人工干预的情况下清除。程序所需的反馈不是文本，而是n个声明选项中的一个以及一个可以阈值化的数字。如今，这需要一次往返到前沿模型——数百毫秒的延迟，每个令牌的费用，以及一个解析器——对于一个通常由三个子句连接而成的问题。this-that-model-1.0是一个具有20亿参数的类型化决策模型。它的答案直接从隐藏状态的指定位置读取，并限制在调用者声明的选项集内，因此不会生成文本，无法出现格式错误，并且请求中的每个问题都在同一次前向传递中得到回答。它在一台笔记本GPU上以30.9毫秒做出决策，并且在此过程中生成零个输出令牌，而前沿API调用需要8758毫秒，能够很好回答这些问题的托管系统在思考时每个问题生成的令牌数量在21到212之间，并为每个令牌收费。它在一台消费级GPU上维持每秒32个决策，并且从不让状态离开机器。在第三方记录的68个决策问题的样本中，基于他们的输入和措辞，它的得分为0.941，Brier得分为0.042，而托管服务Jev在相同项目上的得分为0.765和0.133。我们42个内部套件的一次运行需要32秒和0.000217美元的电费；我们测量的最准确的托管模型需要155.2分钟和10.636美元。我们还报告了它的不足之处。在多步骤算术问题上，单次前向传递无法携带中间结果，它的得分为0.560，而他们的得分在0.98到1.00之间，针对五个任务家族的第二轮有针对性的训练提高了模型的表现，但未能转移到其他13个任务上。该模型已在https://huggingface.co/flock-io/this-that-model-1.0上开源。
cs.CL / 131 / 2609.23916
Time-Incremental Continued Pretraining of LLMs: Knowledge Updates Without Catastrophic Forgetting
时间增量持续预训练大型语言模型：无灾难性遗忘的知识更新
Abstract
Large language models (LLMs) drift out of date the moment their pretraining ends, yet retraining from scratch is prohibitively expensive. Continued pretraining (CPT) is the natural remedy, but it is typically evaluated through a continual learning lens that assumes disjoint data streams. This is a poor fit for time-incremental updates on web-scale crawls, where successive snapshots share substantial URL overlap by design. We study time-incremental CPT in this realistic regime: continued pretraining on FineWeb-Edu dumps drawn strictly from after each model's knowledge cutoff, evaluated across six open-weight models spanning three families (OLMo2, Llama-3.1/3.2, Gemma-3-1B) and four parameter scales (1B-3B-7B-8B). We organize our findings around four practical questions. (i) Is knowledge acquired? Yes, but heterogeneously, and without catastrophic forgetting: five of six models also improve on pre-cutoff factual recall, and the gains track pretraining saturation (driven primarily by token budget per parameter). (ii) What does it cost? Almost nothing: the macro-average across a thirteen-task suite stays within 0.01 of the base for every model. (iii) What is the recipe? Data quality dominates quantity (a curated 6B-token slice matches a broader 40B one); the optima for knowledge acquisition and general capability are separated by roughly an order of magnitude in learning rate; and LoRA at sufficient rank matches full CPT. (iv) Does it survive deployment? CPT gains transfer through SFT, while DPO's effect is family-dependent. Together, these results paint a more optimistic picture of time-incremental CPT than the prior continual learning literature suggests.
Chinese Translation
大型语言模型（LLMs）在预训练结束的瞬间便开始过时，然而从头开始重新训练的成本极高。持续预训练（CPT）是自然的解决方案，但通常通过持续学习的视角进行评估，假设数据流是互不重叠的。这对于在网络规模爬虫上进行时间增量更新来说并不合适，因为连续快照在设计上共享大量的 URL 重叠。我们在这一现实环境中研究时间增量 CPT：在每个模型的知识截止日期之后严格从 FineWeb-Edu 数据集中进行持续预训练，并在三个家族（OLMo2、Llama-3.1/3.2、Gemma-3-1B）和四个参数规模（1B-3B-7B-8B）中评估六个开放权重模型。我们围绕四个实际问题组织我们的发现。(i) 知识是否被获取？是的，但不均匀，并且没有灾难性遗忘：六个模型中有五个在截止前的事实回忆上也有所提升，增益与预训练饱和度相关（主要受每个参数的令牌预算驱动）。(ii) 成本是多少？几乎没有：在一个包含十三个任务的套件中，宏观平均值对每个模型保持在基线的 0.01 之内。(iii) 诀窍是什么？数据质量优于数量（一个经过整理的 6B 令牌片段与一个更广泛的 40B 片段相匹配）；知识获取和一般能力的最佳学习率相差大约一个数量级；而在足够的秩下，LoRA 可以与完整的 CPT 相匹配。(iv) 它能在部署中生存吗？CPT 的增益通过 SFT 转移，而 DPO 的效果依赖于模型家族。综合来看，这些结果描绘了一个比以往持续学习文献所暗示的更乐观的时间增量 CPT 画面。
cs.CL / 132 / 2609.23935
Measuring the Assistant's Harmlessness Preferences on the User Turn
测量助手在用户发言时的无害偏好
Abstract
Post-training turns a general next-token predictor into a chat model with a persistent assistant persona. If that persona is a character the model plays only on its own turns, its preferences should govern what the assistant says, not what the model predicts other speakers will say. We test this boundary and find that it does not hold: a safety-relevant preference of the assistant---for harmless over harmful tasks---shapes the model's predictions even on the user's turn, where the assistant is not the one speaking. We find that this preference is small or near-zero in pretrained base models, that it emerges through post-training, replicated across open-weight model families, grows with scale, and can be moved by narrow finetuning that never touches user turns. We claim that this is evidence that post-training does not merely install a shallow assistant persona, but instead generalises beyond just the local assistant turn, into the model's representation of the user.
Chinese Translation
后训练将一般的下一个标记预测器转变为具有持久助手角色的聊天模型。如果这个角色是模型仅在其自身发言时扮演的角色，那么其偏好应当主导助手所说的话，而不是模型预测其他说话者将要说的话。我们测试了这一边界，发现其并不成立：助手的一个与安全相关的偏好——对无害任务的偏好而非有害任务——即使在用户发言时也会影响模型的预测，此时助手并不是发言者。我们发现，在预训练的基础模型中，这种偏好较小或接近于零，而在后训练中逐渐显现，并在开放权重模型家族中得到了复制，随着规模的增加而增强，并且可以通过狭义微调来改变，即使这种微调从未涉及用户发言。我们认为这表明，后训练并不仅仅是安装一个浅层的助手角色，而是超越了局部助手发言，影响了模型对用户的表征。
cs.CL / 133 / 2609.23939
XYEval: Agents say yes to bad advice
XYEval：智能体对错误建议说是
Abstract
Effective communication between users and AI agents is essential for human-AI collaboration. The XY problem is a well-known communication pitfall where a person asks about their attempted solution rather than their actual problem. We extend prior sycophancy evaluation to the XY problem in agentic settings, evaluating whether agents can resist plausible but misleading suggestions from users and communicate their reasoning. We introduce XYEval, a meta-evaluation framework that can transform an existing benchmark into an XY problem evaluation. We evaluate five models across six diverse benchmark suites. Agents suffer large XY drops under XY mutation across benchmarks, with relative drops reaching up to 46.7%. With $\tau^2$-bench, we further show that agent performance drops more when encountering a pedantic user who requires detailed explanations before approving a better solution. Our findings suggest that current agents lack the ability to effectively reason and communicate when facing misleading suggestions. A simple system instruction baseline that encourages awareness of XY problems only offers partial mitigation. Extensive trace analyses provide behavioral insights into how and why these XY drops occur across execution trajectories. Our results show that mitigating the XY problem remains challenging, requiring agents to both recognize user misdirection and clearly communicate the underlying problem.
Chinese Translation
用户与人工智能智能体之间的有效沟通对于人机协作至关重要。XY问题是一种众所周知的沟通陷阱，其中一个人询问他们尝试的解决方案，而不是他们实际的问题。我们将先前的谄媚评估扩展到智能体环境中的XY问题，评估智能体是否能够抵制来自用户的看似合理但误导性的建议，并传达他们的推理过程。我们引入了XYEval，一个元评估框架，可以将现有基准转化为XY问题评估。我们在六个不同的基准套件中评估了五个模型。智能体在基准测试中经历了显著的XY下降，相对下降幅度高达46.7%。通过$ au^2$-bench，我们进一步表明，当遇到需要详细解释才能批准更好解决方案的迂腐用户时，智能体的表现下降更为明显。我们的研究结果表明，当前的智能体在面对误导性建议时缺乏有效推理和沟通的能力。一个简单的系统指令基线，鼓励对XY问题的意识，仅提供了部分缓解。广泛的追踪分析提供了关于这些XY下降在执行轨迹中如何以及为何发生的行为洞察。我们的结果表明，缓解XY问题仍然具有挑战性，要求智能体既能识别用户的误导，又能清晰地传达潜在问题。
cs.CL / 134 / 2609.23951
HaikuS2S: A Cascaded System For Responding In Verse
HaikuS2S：一种用于诗歌响应的级联系统
Abstract
Expressive speech synthesis has advanced through prosody modeling, yet generating structured poetic speech, such as haiku, remains challenging. Prior work on prosody transfer improves expressiveness, and fine-tuned poetry TTS (text-to-speech) systems capture verse intonation. However, these models do not model haiku's 5-7-5 syllable structure or line-ending pauses. We present a cascaded system, HaikuS2S, combining ASR (automatic speech recognition), LLM (large language model)-generated haiku, and TTS fine-tuning on both prose and custom haiku datasets. Our evaluation focuses on emotion similarity, speech quality, and prosody alignment. In our experiments, we see that our prosody and tonal alignment improve significantly with our fine-tuned systems, particularly the one trained on both general poetry and haiku. We also see that we maintain similar emotion similarity scores across all systems.
Chinese Translation
表达性语音合成通过韵律建模取得了进展，但生成结构化的诗歌语音，如俳句，仍然具有挑战性。先前的韵律转移工作提高了表达性，而经过微调的诗歌文本到语音（TTS）系统则捕捉了诗句的语调。然而，这些模型并未考虑俳句的5-7-5音节结构或行尾停顿。我们提出了一种级联系统HaikuS2S，结合了自动语音识别（ASR）、大型语言模型（LLM）生成的俳句，以及在散文和自定义俳句数据集上的TTS微调。我们的评估重点关注情感相似性、语音质量和韵律对齐。在我们的实验中，我们观察到经过微调的系统在韵律和音调对齐方面显著改善，特别是训练于一般诗歌和俳句的系统。我们还发现，所有系统的情感相似性评分保持相似。
cs.CL / 135 / 2609.23955
Some Dialects Are More Equal Than Others: Non-Prestigious Arabic Dialectal Bias in LLMs
某些方言比其他方言更为平等：大型语言模型中的非高尚阿拉伯方言偏见
Abstract
Previous work on Egyptian Arabic in NLP has focused largely on the prestigious Cairene Egyptian Arabic (CEA) dialect, resulting in a lack of representation for the less prestigious Sa'idi Egyptian Arabic (SEA) dialect both in LLM and resource development. Does this lack of representation influence an LLM's view of the acceptability of SEA (upstream), and does an upstream bias against SEA lead to worse performance (downstream)? We investigate the upstream effect of SEA dialectal features on LLM preferences in a Targeted Syntactic Evaluation (TSE) task which reveals a significant bias against SEA across multiple LLMs. We then analyze the effect of these same features on downstream model performance on MMLU benchmarks and show that models experience a degradation in performance when presented with SEA. This work highlights the need for further exploration on how sub-dialectal variation impacts language technologies.
Chinese Translation
先前关于埃及阿拉伯语在自然语言处理中的研究主要集中于高尚的开罗埃及阿拉伯语（CEA）方言，导致较少高尚的上埃及阿拉伯语（SEA）方言在大型语言模型（LLM）和资源开发中的代表性不足。这种代表性的缺乏是否影响了大型语言模型对上埃及阿拉伯语的可接受性的看法（上游），而对上埃及阿拉伯语的上游偏见是否导致了性能下降（下游）？我们在一个针对性句法评估（TSE）任务中调查了上埃及阿拉伯语方言特征对大型语言模型偏好的上游影响，结果显示多个大型语言模型对上埃及阿拉伯语存在显著偏见。随后，我们分析了这些相同特征对下游模型在MMLU基准测试中的性能影响，结果表明，当模型面对上埃及阿拉伯语时，其性能出现下降。这项工作突显了进一步探索子方言变异如何影响语言技术的必要性。
cs.CL / 136 / 2609.23959
Open-Jev Judgments on CallScreenBench: Calibrated One-Pass Scam Screening with a Small Language Model
Open-Jev 对 CallScreenBench 的判断：使用小型语言模型的校准单次诈骗筛查
Abstract
Screening a phone call for fraud needs a trustworthy probability after every caller turn, in milliseconds. Jev-style typed decisions promise exactly that: declared options go in, one calibrated probability per option comes out of a single forward pass, with no generated text. We test an open implementation of this readout, JevLite, on scam-call screening: Qwen3-4B is LoRA-tuned so that the temperature-scaled softmax over two answer-label logits is P(scam). On 41 held-out CallScreenBench scenarios (577 per-turn decisions) a three-seed ensemble reaches AUROC .974 with calibration error .052, non-inferior to an LLM judge (MiniMax-M3) at a pre-registered .02 margin, with no false alarms on legitimate calls, decisions 1.14 turns earlier under the same hang-up rule, and 64.5 ms per decision on one consumer GPU, 4.9x lower than the same backbone fine-tuned to generate its answer. The gain is in the readout and calibration, not accuracy: a fine-tuned ModernBERT encoder is not significantly worse, the recipe was selected with test-set exposure, and all callers are synthetic. We claim no architectural novelty; the contribution is the application and an evaluation reporting calibration, false alarms and decision timing alongside AUROC.
Chinese Translation
对电话呼叫进行诈骗筛查需要在每次来电者的回合后以毫秒级别提供一个可信的概率。Jev 风格的类型决策正好满足这一需求：输入声明的选项，经过一次前向传递输出每个选项的校准概率，而无需生成文本。我们在诈骗电话筛查中测试了这一读出方法的开放实现，JevLite：Qwen3-4B 经过 LoRA 调优，使得两个答案标签对数的温度缩放 softmax 输出 P(scam)。在 41 个保留的 CallScreenBench 场景（577 次每回合决策）中，三种种子集成达到了 AUROC 0.974，校准误差为 0.052，表现不逊色于一个 LLM 判决者（MiniMax-M3），在预注册的 0.02 边际下，合法电话没有误报，决策在相同挂断规则下提前 1.14 回合，且在一台消费者 GPU 上每次决策耗时 64.5 毫秒，比同一骨干网微调生成答案的耗时低 4.9 倍。增益体现在读出和校准上，而非准确性：微调的 ModernBERT 编码器并没有显著劣于此，配方是在测试集曝光下选择的，且所有来电者均为合成数据。我们不声称有任何架构上的新颖性；贡献在于应用及评估，报告了校准、误报和决策时机，以及 AUROC。
cs.CL / 137 / 2609.23966
From Tables to Quantified Statements: Evaluating LLM Inference Generation through Executable Verification
从表格到量化陈述：通过可执行验证评估大型语言模型推理生成
Abstract
LLMs can generate fluent descriptions from tables, but their outputs may remain logically unsupported by the structured data. We introduce STAT-TO-TEXT, a controlled task in which LLMs generate quantified natural language inferences from statistical tables using quantified constructions such as all, some, no, and most. To evaluate these inferences, we use an LLM generated Python checker code which when executed verifies the corresponding truth conditions against the table. We compare four open-weight LLMs across model families and scales, evaluating faithfulness, logical accuracy, table coverage, and diversity. Our results show that model scale and family matter, with the largest model (GPT-OSS-120B) consistently producing the most faithful inferences without sacrificing greater table coverage and quantifier diversity, as opposed to smaller models. These findings are supported by human annotation, which shows that the automated checker closely aligns with human judgments.
Chinese Translation
大型语言模型（LLMs）能够从表格生成流畅的描述，但其输出可能在逻辑上无法得到结构化数据的支持。我们提出了STAT-TO-TEXT，这是一个受控任务，其中LLMs使用量化结构（如all、some、no和most）从统计表中生成量化自然语言推理。为了评估这些推理，我们使用了一个LLM生成的Python检查器代码，该代码在执行时验证与表格对应的真值条件。我们比较了四个开放权重的LLMs，涵盖不同的模型家族和规模，评估其忠实度、逻辑准确性、表格覆盖率和多样性。我们的结果表明，模型的规模和家族是重要因素，最大的模型（GPT-OSS-120B）在不牺牲更大表格覆盖率和量词多样性的情况下，始终生成最忠实的推理，而较小的模型则表现不佳。这些发现得到了人工标注的支持，显示自动检查器与人类判断紧密一致。
cs.CL / 138 / 2609.24028
When Evidence Conflicts: Reliability-aware Meta-review Generation
当证据冲突时：基于可靠性的元评审生成
Abstract
Generating coherent meta-reviews from multiple peer reviews is challenging when reviewer evidence conflicts and varies in reliability. Existing approaches typically formulate meta-review generation as a multi-document summarization task and aggregate reviewer feedback uniformly, making it difficult to determine which opinions should be prioritized under disagreement. In this paper, we study meta-review generation through reliability-aware evidence aggregation. Our framework first extracts aspect-level opinions from peer reviews and identifies conflicting evidence within each aspect. It then estimates opinion-level support and review-level quality to measure evidence reliability. Based on these signals, the framework assigns reliability-aware weights to reviewer feedback, enabling the generator to prioritize better-supported arguments while preserving diverse perspectives. Experiments demonstrate that our method consistently improves meta-review generation over strong baselines on both automatic and human evaluations, with clear gains in conflict recognition and resolution under high-conflict review scenarios. The code and implementation details are publicly available at https://github.com/Wangxz729/reliability-aware-meta-review.
Chinese Translation
在评审者证据冲突且可靠性各异的情况下，从多个同行评审中生成连贯的元评审是一项挑战。现有的方法通常将元评审生成视为多文档摘要任务，并均匀聚合评审者反馈，这使得在意见不一致时难以确定应优先考虑哪些观点。本文研究了通过基于可靠性的证据聚合进行元评审生成。我们的框架首先从同行评审中提取方面级意见，并识别每个方面内的冲突证据。然后，它估计意见级支持度和评审级质量，以衡量证据的可靠性。基于这些信号，框架为评审者反馈分配基于可靠性的权重，使生成器能够优先考虑支持更好的论点，同时保留多样的视角。实验表明，我们的方法在自动评估和人工评估中，始终优于强基线，特别是在高冲突评审场景下，在冲突识别和解决方面有明显提升。代码和实现细节可在 https://github.com/Wangxz729/reliability-aware-meta-review 获取。
cs.CL / 139 / 2609.24052
Calibrated Decisions at Scale: Converting Police Crash Narratives into Probabilistic Crash Variables with a System One Model (Jev)
规模化的校准决策：使用系统一模型（Jev）将警察事故叙述转换为概率事故变量
Abstract
Crash datasets that carry an investigator narrative hold information the coded fields omit. Coding those narratives at scale has been blocked by three obstacles. Frontier large language models are costly at that scale, their generated text cannot be verified, and no rule says how much output a human must check. This paper formulates narrative coding as gated, typed decisions answered by Jev, a System One model that returns probabilities over analyst-defined options and generates no text. A screen covered 499,500 Texas narratives and 195,857 were coded with a 27-question schema. Cost is governed by schema size rather than narrative length. The probabilities are audited against coded fields and against 2,416 blinded human judgments drawn under a stated sampling design. Two frontier large language models are benchmarked on the same records. Against human labels the typed model attains an F1 of 0.908. One frontier model gains 0.059 and the other is indistinguishable from it. Calibration varies by model rather than by paradigm, so each model must be audited. Recalibration on the same labels reduces calibration error by a factor of 3.3. Agreement with coded fields understates fidelity to the narrative by a median of 0.26 in kappa. A resolution-floor bound covers any model that reports probabilities on a discrete grid. A review budget over flagged records gives the records a human must read per variable and per year. Adding the calibrated variables to the coded fields raises the injury and fatal crashes attributed to nine factors by 10,747 per year.
Chinese Translation
包含调查员叙述的事故数据集承载着编码字段所遗漏的信息。大规模编码这些叙述面临三个障碍。前沿的大型语言模型在这种规模下成本高昂，其生成的文本无法验证，并且没有规则规定人类必须检查多少输出。本文将叙述编码形式化为由Jev（一个系统一模型）回答的有门限的、类型化的决策，该模型返回分析师定义选项的概率，并不生成文本。筛选覆盖了499,500个德克萨斯州的叙述，其中195,857个使用27个问题的框架进行了编码。成本由框架大小而非叙述长度决定。概率经过对编码字段和根据既定抽样设计得出的2,416个盲人判断的审计。对同一记录进行了两个前沿大型语言模型的基准测试。与人类标签相比，类型化模型的F1值达到了0.908。一种前沿模型获得了0.059，另一种与其无显著区别。校准因模型而异，而非因范式，因此每个模型都必须进行审计。在相同标签上的重新校准将校准误差减少了3.3倍。与编码字段的一致性在kappa中低估了对叙述的忠实度，中位数为0.26。一个分辨率下限覆盖了任何在离散网格上报告概率的模型。对标记记录的审查预算为每个变量和每年规定了人类必须阅读的记录数量。将校准变量添加到编码字段中，使归因于九个因素的伤害和致命事故每年增加了10,747起。
cs.CL / 140 / 2609.24066
Efficient Reasoning Exploration via State-Conditioned Latent Steering with Progress Guidance
通过状态条件潜在引导与进展指导的高效推理探索
Abstract
Best-of-$N$ is a widely used inference strategy for complex reasoning, whose effectiveness depends on whether sampled candidates can cover diverse and high-quality reasoning paths. However, post-trained reasoning models often suffer from \emph{exploration collapse}, where independent rollouts repeatedly follow similar reasoning paths and limit the gains from increasing the rollout budget. Existing methods alleviate this issue by promoting broader exploration, but do not explicitly guide exploration toward continuations that make meaningful progress, resulting in limited exploration efficiency. To address this, we propose \emph{\underline{S}tate-conditioned \underline{P}rogress-guided \underline{S}teering} (SPS), a training-free latent steering framework. Specifically, SPS constructs a state-conditioned Direction Bank containing multiple progress-guided steering vectors for different prefix-state regions. During online inference, SPS retrieves a suitable steering vector based on the current prefix state and applies it at high-uncertainty transitions to guide the next reasoning step toward meaningful progress. Extensive experiments across multiple model scales and benchmarks demonstrate that SPS consistently outperforms strong baselines. Further analyses validate the effectiveness of its key designs and offer valuable insights for future research. The code is available at https://github.com/rattlesnakey/SPS.
Chinese Translation
Best-of-$N$ 是一种广泛使用的复杂推理推断策略，其有效性依赖于采样候选者是否能够覆盖多样化和高质量的推理路径。然而，经过后训练的推理模型常常遭遇 extit{探索崩溃}，即独立的回滚重复遵循相似的推理路径，从而限制了增加回滚预算所带来的收益。现有方法通过促进更广泛的探索来缓解这一问题，但并未明确引导探索朝向能够实现有意义进展的延续，导致探索效率有限。为了解决这一问题，我们提出了 extit{ extbf{S}tate-conditioned extbf{P}rogress-guided extbf{S}teering} (SPS)，一种无训练的潜在引导框架。具体而言，SPS 构建了一个状态条件的方向库，其中包含多个针对不同前缀状态区域的进展引导引导向量。在在线推断过程中，SPS 根据当前前缀状态检索合适的引导向量，并在高不确定性转变处应用它，以引导下一步推理朝向有意义的进展。针对多个模型规模和基准的广泛实验表明，SPS 始终优于强基线。进一步的分析验证了其关键设计的有效性，并为未来的研究提供了有价值的见解。代码可在 https://github.com/rattlesnakey/SPS 获取。
cs.CL / 141 / 2609.24083
From Content Generation to Learning Support: Pedagogy-Guided Generative Video Tutors for STEM Learning
从内容生成到学习支持：以教学法为指导的生成视频辅导在STEM学习中的应用
Abstract
Generative AI enables scalable production of educational videos, but current systems largely focus on producing visually coherent content rather than supporting learning. As a result, generated videos often lack explicit pedagogical structure, reliable quality control, and mechanisms for assessing learner understanding or addressing misconceptions. In this work, we introduce PIVOT (Pedagogy-guided Instructional VideO Tutoring), a generative video tutoring framework for STEM learning via learning-centered instructional support.1 Inspired by conventional teaching workflows, our framework integrates pedagogy into the full generation pipeline: it first uses instructional principles to guide storyboard generation, then produces verified multimodal videos through code-centric generation and a pedagogical verification harness, and finally connects videos with assessment and misconception-aware remediation. Experiments and expert evaluations across four STEM domains show that our framework produces educational videos with pedagogically aligned content, clear and engaging presentation, coherent instructional flow, and perceived effectiveness for learning. These findings suggest a human-centered perspective on educational content generation: generative systems should be evaluated and designed not only by what they produce, but also by how they support teaching practices, learner understanding, and corrective feedback.
Chinese Translation
生成性人工智能使教育视频的规模化生产成为可能，但当前系统主要集中于生成视觉上连贯的内容，而非支持学习。因此，生成的视频往往缺乏明确的教学结构、可靠的质量控制，以及评估学习者理解或解决误解的机制。在本研究中，我们介绍了PIVOT（以教学法为指导的教学视频辅导），这是一个通过以学习为中心的教学支持来促进STEM学习的生成视频辅导框架。受到传统教学工作流程的启发，我们的框架将教学法整合到整个生成流程中：首先利用教学原则指导故事板生成，然后通过以代码为中心的生成和教学验证工具生成经过验证的多模态视频，最后将视频与评估和关注误解的补救措施相连接。在四个STEM领域的实验和专家评估表明，我们的框架生成的教育视频具有与教学法一致的内容、清晰且引人入胜的呈现、连贯的教学流程，以及对学习的有效性认知。这些发现表明，教育内容生成应采取以人为本的视角：生成系统的评估和设计不仅应考虑其产出，还应关注其如何支持教学实践、学习者理解和纠正反馈。
cs.CL / 142 / 2609.24106
You Can Tell Who's Asking: What the Web's Questions Are Made Of, and Where They Come From
你可以判断谁在提问：网络问题的构成及其来源
Abstract
Questions scraped from the web are used across academia and industry as a proxy for what people want to know. Across QA training data, retrieval benchmarks, and content strategy, questions on a page are assumed to reflect human intent. We test this assumption at scale by extracting 13.4B question occurrences across 110 FineWeb snapshots (2013-2025), and report three findings. First, you can tell who is asking: provenance (the host/page of questions) leaves a signal in question form, and a logistic model can separate genuine user questions from templated/manufactured ones at AUC 0.725 via length and surrounding context rather than question type, though only 0.554 against commerce FAQ writing. Second, question frequency does not measure demand: the most-frequent questions are boilerplate/templated (over 70% of the top thousand), so occurrence counts measure how often a string was published and not how often it was asked. Third, over twelve years the genuine share of occurrences fell by 79% (42-56% after controlling for crawl composition), with question length and context decreasing. We present the first diachronic, occurrence-level measurement of web question provenance, and find the crawlable web's questions have shifted from being asked by humans toward manufactured for machines to read.
Chinese Translation
从网络上提取的问题被广泛用于学术界和工业界，作为人们想要了解的内容的代理。在问答（QA）训练数据、检索基准和内容策略中，页面上的问题被认为反映了人类的意图。我们通过提取来自110个FineWeb快照（2013-2025）的134亿个问题实例，在大规模上测试这一假设，并报告了三个发现。首先，你可以判断谁在提问：问题的来源（问题的主机/页面）在问题形式中留下了信号，逻辑模型可以通过问题长度和周围上下文将真实用户问题与模板化/制造的问题区分开来，AUC为0.725，而在商业常见问题（FAQ）写作中仅为0.554。其次，问题频率并不能衡量需求：最常见的问题是模板化的（超过70%的前千个问题），因此出现次数测量的是字符串被发布的频率，而不是被提问的频率。第三，经过十二年的观察，真实问题的出现比例下降了79%（在控制爬虫组成后为42-56%），问题的长度和上下文也在减少。我们首次呈现了网络问题来源的历时性、出现级别的测量，发现可爬取网络中的问题已经从人类提问转向为机器阅读而制造的问题。
cs.CL / 143 / 2609.24122
Re:CAP - Auditing Retrieval Coverage in Production RAG Pipelines
Re:CAP——生产环境中RAG流水线检索覆盖率的审计
Abstract
Retrieval-augmented generation (RAG) is hard to monitor in production: exhaustive relevance labels do not exist for non-stationary multi-million-passage corpora that re-index in real time. As a result, retrieval quality is generally understudied and often deprioritised in favour of generation-oriented metrics. In this work, we propose auditing retrieval coverage by probing for evidence of missing documents rather than enumerating every relevant one. Our method Re:CAP (REtrieval Coverage Audit by iterative Probing) is a reference-free audit loop applied to a deployed RAG pipeline's initial answer and retrieved context: it identifies the topics already covered, generates probing questions for plausibly missing topics, retrieves candidate documents, and applies an LLM-as-judge to retain only those that introduce previously-unretrieved information. On four public benchmarks, Re:CAP recovers 9-29% of gold labels that flat BM25 top-500 cannot reach, rising to 48% on TREC-COVID. On MuSiQue Re:CAP beats flat hybrid top-500 by +12.9 pp on recall at less than half the document budget. An ensemble BM25, dense, and hybrid baseline (top-500 each) still leaves out 21.2% of gold docs on TREC-COVID that Re:CAP recovers; human annotators judge that 78.9% of those structurally distinct documents add new information to the baseline answer (Fleiss $\kappa$ = 0.79, n = 123), and 73.9% on live production traffic (n = 180). End-to-end recall is reproducible to within $\pm$1% across three independent runs, making Re:CAP a stable instrument for periodic retrieval audits.
Chinese Translation
检索增强生成（Retrieval-augmented generation，RAG）在生产环境中难以监控：对于实时重建索引的非静态数百万文档语料库，缺乏详尽的相关性标注。因此，检索质量通常研究不足，且常被生成导向的指标所取代。在本研究中，我们提出通过探测缺失文档的证据而非枚举所有相关文档，来审计检索覆盖率。我们的方法Re:CAP（REtrieval Coverage Audit by iterative Probing）是一种无参考的审计循环，应用于已部署RAG流水线的初始答案及检索上下文：它识别已覆盖的话题，生成针对可能缺失话题的探测性问题，检索候选文档，并利用大型语言模型（LLM）作为评判者，仅保留那些引入先前未检索信息的文档。在四个公开基准测试中，Re:CAP恢复了平面BM25前500名无法覆盖的9%至29%的金标标签，在TREC-COVID上更高达48%。在MuSiQue数据集上，Re:CAP以不到一半的文档预算，在召回率上比平面混合检索前500名高出12.9个百分点。一个由BM25、密集检索和混合检索组成的集成基线（各取前500名）在TREC-COVID上仍遗漏21.2%的金标文档，而Re:CAP成功恢复了这些文档；人工标注者判断其中78.9%的结构性差异文档为基线答案补充了新信息（Fleiss κ=0.79，样本量n=123），在实际生产流量中该比例为73.9%（n=180）。端到端召回率在三次独立运行中可复现至±1%以内，使Re:CAP成为周期性检索审计的稳定工具。
cs.CL / 144 / 2609.24156
TAC-Time: Texts as Channels For Multimodal Time Series Forecasting
TAC-Time：文本作为多模态时间序列预测的通道
Abstract
Most existing time series forecasting methods rely solely on numerical observations, overlooking rich contextual information from auxiliary texts. Recent multimodal approaches attempt to incorporate textual signals, but they often treat text as static features or use large language models as forecasting backbones, limiting their ability to capture temporal dynamics and increasing computational cost. To address these challenges, we propose TAC-Time, a unified framework that transforms textual information into additional temporal channels. By modeling text features jointly with numerical sequences in a shared temporal backbone, TAC-Time preserves temporal continuity and periodic structures while remaining efficient and scalable. This formulation also enables systematic interpretability analyses. We show strong cross-modal dependencies through attention and frequency-domain analyses, and identify predictive textual signals whose correlation-aware alignment yields partial forecasting improvements. Extensive experiments on real-world multimodal benchmarks demonstrate that TAC-Time outperforms prior methods.
Chinese Translation
大多数现有的时间序列预测方法仅依赖于数值观察，忽视了来自辅助文本的丰富上下文信息。近期的多模态方法试图结合文本信号，但通常将文本视为静态特征或使用大型语言模型作为预测基础，这限制了它们捕捉时间动态的能力，并增加了计算成本。为了解决这些挑战，我们提出了TAC-Time，一个统一框架，将文本信息转化为额外的时间通道。通过在共享的时间骨架中共同建模文本特征和数值序列，TAC-Time保持了时间连续性和周期结构，同时保持高效和可扩展性。这种表述还使系统的可解释性分析成为可能。我们通过注意力和频域分析展示了强烈的跨模态依赖性，并识别出预测性文本信号，其相关性感知对齐带来了部分预测改进。在真实世界的多模态基准上进行的广泛实验表明，TAC-Time的表现优于之前的方法。
cs.CL / 145 / 2609.24177
Efficient LLM Distillation for Bangladesh Legal Context: A Smartphone-Compatible Retrieval-Augmented Generation Model
针对孟加拉国法律背景的高效大语言模型蒸馏：一种适用于智能手机的检索增强生成模型
Abstract
Legal information in Bangladesh is inaccessible to most citizens. Statutory text is English-only, trained lawyers are concentrated in urban centres, and cloud-dependent AI fails where mobile connectivity is unreliable, a setting in which hallucinated legal text causes direct harm. The system addresses statutory interpretation only; queries that require judicial precedent or case-law reasoning fall outside its scope. We target the statutory access gap by compressing a 9-billion-parameter Gemma-2 teacher into a 2-billion-parameter student through two-phase progressive knowledge distillation. Phase 1 performs supervised fine-tuning on 9,429 quality-gated legal question-answer pairs (65% acceptance from 14,514 generated queries); Phase 2 minimises sparse Kullback-Leibler divergence against the teacher's top-50 per-token logits at temperature tau = 4.0, implemented via QLoRA (4-bit NF4, rank-32 LoRA adapters). Prior legal language models target general legal English; this system specialises in Bangladeshi statutory law. Every response is grounded through hybrid retrieval combining dense semantic search (60%) and BM25 (40%) across 36,029 statutory passages from the Bangladesh Constitution and national legislation. On a 50-query English benchmark, the distilled model reaches ROUGE-L 0.4715 and BERTScore F1 0.5679, a 103% ROUGE-L and 143% BERTScore gain over the retrieval-augmented undistilled baseline (ROUGE-L 0.2323, BERTScore 0.2340). The adapter quantises to 1.6 GB (GGUF Q4_K_M) and runs at 4-8 tokens per second on a Pixel 6 with no network access. Cross-lingual evaluation on 50 Bangla queries yields ROUGE-L 0.4083 and BERTScore 0.8133, showing effective retrieval from Bangla input against an English-only corpus. In a single-evaluator pilot, a practising lawyer rated 50 responses at a weighted mean of 4.16/5 (90% rated 4 or 5), supporting utility beyond text-overlap metrics.
Chinese Translation
在孟加拉国，大多数公民无法获取法律信息。法定文本仅为英语，受过培训的律师集中在城市中心，而依赖云的人工智能在移动网络连接不可靠的环境中失效，这种情况下虚构的法律文本会造成直接伤害。该系统仅针对法定解释；需要司法先例或案例推理的查询超出了其范围。我们通过两阶段渐进知识蒸馏，旨在缩小法定访问差距，将一个拥有90亿参数的Gemma-2教师模型压缩为一个拥有20亿参数的学生模型。第一阶段在9,429对经过质量筛选的法律问答对上进行监督微调（从14,514个生成查询中接受65%）；第二阶段则在温度tau = 4.0下，最小化与教师模型的每个标记的前50个logits之间的稀疏Kullback-Leibler散度，通过QLoRA（4位NF4，秩-32 LoRA适配器）实现。先前的法律语言模型针对的是一般法律英语；该系统专注于孟加拉国的法定法律。每个响应都通过混合检索进行验证，结合了密集语义搜索（60%）和BM25（40%），涵盖了来自孟加拉国宪法和国家立法的36,029个法定段落。在一个包含50个查询的英语基准测试中，蒸馏模型达到了ROUGE-L 0.4715和BERTScore F1 0.5679，较检索增强的未蒸馏基线（ROUGE-L 0.2323，BERTScore 0.2340）分别提高了103%的ROUGE-L和143%的BERTScore。适配器量化为1.6 GB（GGUF Q4_K_M），在没有网络访问的情况下，在Pixel 6上以每秒4-8个标记的速度运行。在50个孟加拉语查询的跨语言评估中，获得了ROUGE-L 0.4083和BERTScore 0.8133，显示出在仅有英语语料库的情况下有效地从孟加拉语输入中进行检索。在一次单评估者的试点中，一名执业律师对50个响应的加权平均评分为4.16/5（90%的评分为4或5），支持其在文本重叠指标之外的实用性。
cs.CL / 146 / 2609.24194
When Residualization Helps an Audit: Format Effects, Slice Gains, and Their Limits
残差化如何帮助审计：格式效应、切片收益及其限制
Abstract
Evaluation scores used around LLM systems -- including reward models, rerankers, and LLM judges -- can track surface form instead of the quality they claim to measure. When presented with a terse correct solution and a commented buggy solution for the same MBPP problem, a public preference reward model selects the correct one no better than a coin flip (0.507). Subtracting the predictable surface component from such scores is increasingly common, but removal alone does not yield a more valid measurement: the removed component may carry construct-relevant signal, and residualization cannot tell which is which. Under designed interventions -- unit-test labels with comment-only edits -- residualization attenuates the reward model's format effects by about 0.12 on both correct and buggy code, while the correct-versus-buggy margins move by less than 0.01. In observational NLI and QA settings, we freeze a held-out replication before scoring and re-evaluate it using labels from disjoint annotators; this supports only a narrower conclusion: better agreement with the construct labels on a pre-declared slice where a surface-only predictor errs, not a repaired score. Full-population agreement falls in every observational setting with a reported positive slice gain, and within-question ranking falls in every such QA setting. When construct and surface features are entangled, residualization can decorrelate a score while degrading construct alignment, and, in a controlled model, configurations just as damaging to construct alignment pass every pre-adjustment check, so no committed gate is a guarantee. We assemble these distinctions into a reporting protocol whose outcomes, refusal included, state what an adjusted score may be claimed to show: an audit-time diagnostic reported beside the construct-alignment cost it incurs, never a replacement for the raw score.
Chinese Translation
在大型语言模型（LLM）系统中使用的评估分数——包括奖励模型、重排序器和LLM评审者——可能追踪表面形式而非它们声称测量的质量。当面对同一个MBPP问题的简洁正确解和带注释的有缺陷解时，公共偏好奖励模型选择正确解的概率与掷硬币相当（0.507）。从这些分数中减去可预测的表面成分已变得越来越普遍，但仅仅去除并不能产生更有效的测量：被移除的成分可能携带与构念相关的信号，而残差化无法区分二者。在设计干预下——仅带注释的单元测试标签——残差化将奖励模型的格式效应降低了约0.12，无论是对正确代码还是有缺陷代码，而正确与有缺陷之间的差距变化不到0.01。在观察性自然语言推理（NLI）和问答（QA）设置中，我们在评分之前冻结一个保留的复制样本，并使用来自不同注释者的标签重新评估；这仅支持一个更狭窄的结论：在一个预先声明的切片上，与构念标签的更好一致性，其中表面预测器出错，而不是修复后的分数。在每个报告了正切片收益的观察性设置中，整体人群一致性下降，而在每个这样的QA设置中，问题内排名也下降。当构念特征和表面特征交织在一起时，残差化可以在降低构念对齐的同时去相关化分数，并且在一个受控模型中，对构念对齐同样有害的配置通过了每个预调整检查，因此没有任何承诺的门槛可以保证。我们将这些区别汇总为一个报告协议，其结果，包括拒绝，表明调整后的分数可能声称显示的内容：在审计时诊断与其产生的构念对齐成本并列报告，绝不是对原始分数的替代。
cs.CL / 147 / 2609.24196
LoopCD: Loop-wise Contrastive Decoding for Improving Reasoning in Looped Language Models
LoopCD：循环对比解码以改善循环语言模型中的推理
Abstract
Looped Language Models (LoopLMs) perform "latent reasoning" by recursively refining internal latent representations with shared weights, offering a more effective alternative to explicit verbal reasoning. Despite their effectiveness, we find that LoopLMs remain prone to loop instability: unstable refinement across iterations can produce localized uncertain "hard" tokens associated with reasoning errors. To address this, we propose LoopCD, loop-wise contrastive decoding that enhances the reasoning performance of LoopLMs by intervening on these tokens at inference time. Specifically, we exploit the internal dynamics of LoopLMs and contrast the logits from earlier iterations with logits from the last refined iteration to form the final sampling distribution. We find that this strategy is highly efficient, introducing only negligible inference overhead and requiring no additional training, while effectively improving reasoning performance by naturally refining reasoning-critical hard tokens. Extensive experiments show that our method improves the performance of recent representative LoopLMs across various reasoning tasks.
Chinese Translation
循环语言模型（LoopLMs）通过递归地利用共享权重精炼内部潜在表示来执行“潜在推理”，提供了一种比显式语言推理更有效的替代方案。尽管它们有效，但我们发现循环语言模型仍然容易出现循环不稳定性：迭代过程中的不稳定精炼可能会产生与推理错误相关的局部不确定“硬”标记。为了解决这个问题，我们提出了LoopCD，一种循环对比解码方法，通过在推理时干预这些标记来增强循环语言模型的推理性能。具体而言，我们利用循环语言模型的内部动态，将早期迭代的logits与最后一次精炼迭代的logits进行对比，以形成最终的采样分布。我们发现这一策略极为高效，仅引入微不足道的推理开销且无需额外训练，同时通过自然精炼推理关键的硬标记有效提高推理性能。大量实验表明，我们的方法在各种推理任务中提升了近期代表性循环语言模型的性能。
cs.CL / 148 / 2609.24199
Vimarsha: Faithful ASR Evaluation for Indian Languages with Demographic Diversity, In-the-Wild Audio and Spelling Variations
Vimarsha：针对具有多样化人口特征、野外音频和拼写变异的印度语言的可信自动语音识别评估
Abstract
Evaluation benchmarks for Indian language automatic speech recognition (ASR) suffer from two systematic biases: optimistic scores from clean, controlled audio conditions, and pessimistic scores from overly rigid transcription standards that penalize valid linguistic variations. We introduce Vimarsha, a 100-hour benchmark spanning all 22 scheduled Indian languages, designed to address both distortions. Vimarsha combines demographically diverse on-field recordings with carefully mined in-the-wild audio selected for acoustic difficulty, alongside a lattice of variations framework that encodes multiple valid transcriptions per utterance. Evaluations of 10 state-of-the-art ASR models reveal substantial shifts in model rankings under realistic conditions, geographic and demographic performance disparities, and systematic failure modes across speaking rates and acoustic environments.
Chinese Translation
印度语言自动语音识别（ASR）的评估基准存在两种系统性偏差：来自清晰、受控音频条件的乐观评分，以及由于过于严格的转录标准而对有效语言变异进行惩罚的悲观评分。我们推出了Vimarsha，这是一个涵盖所有22种计划印度语言的100小时基准，旨在解决这两种扭曲。Vimarsha结合了人口多样化的现场录音与经过精心挑选的具有声学难度的野外音频，并采用了一个变异框架，编码每个发话的多个有效转录。对10个最先进的ASR模型的评估显示，在现实条件下模型排名发生了显著变化，地理和人口表现存在差异，以及在说话速度和声学环境下的系统性失败模式。
cs.CL / 149 / 2609.24219
From Articles to Publishers: Aggregating Language Model Predictions for News Source Reliability Inference
从文章到出版商：聚合语言模型预测以推断新闻来源的可靠性
Abstract
Traditionally, the reliability of news publishers is assessed by expert organisations that evaluate editorial practices, transparency and factual standards at source. When this process is translated into a computational approach, the problem is often formulated at the level of individual articles, with models being trained on a set of pre-labelled articles and their performance being evaluated in a test phase. In this work, we investigate news source reliability inference as a source-level prediction problem. We propose a two-stage framework in which transformer-based language models first estimate the reliability of individual articles and subsequently aggregate article-level predictions to infer the reliability of previously unseen publishers. To approximate realistic deployment conditions, we enforce a strict publisher-disjoint evaluation protocol, ensuring that no publisher appears in both training and test sets. Experiments on 19,476 political news articles from 439 English-language publishers labeled with NewsGuard reliability ratings show that aggregation substantially improves robustness and performance, increasing accuracy from approximately 0.60 at the article level to 0.69 at the publisher level. Finally, we analyze how prediction errors vary across political orientations, revealing statistically significant associations between political leaning and misclassification patterns. Overall, our findings show that publisher reliability can be inferred from aggregated textual signals alone, supporting scalable and content-based approaches to automated news source assessment.
Chinese Translation
传统上，新闻出版商的可靠性由专家机构评估，这些机构评估来源的编辑实践、透明度和事实标准。当这一过程转化为计算方法时，问题通常在单个文章的层面上进行表述，模型在一组预标记的文章上进行训练，并在测试阶段评估其性能。在本研究中，我们将新闻来源的可靠性推断视为一个来源级预测问题。我们提出了一个两阶段框架，其中基于变换器的语言模型首先估计单个文章的可靠性，然后聚合文章级预测以推断之前未见过的出版商的可靠性。为了逼近现实的部署条件，我们实施了严格的出版商不重叠评估协议，确保没有出版商同时出现在训练集和测试集中。在对来自439个英语出版商的19,476篇政治新闻文章（标记有NewsGuard可靠性评级）的实验中，聚合显著提高了稳健性和性能，使得准确率从文章级的约0.60提高到出版商级的0.69。最后，我们分析了预测错误在不同政治倾向中的变化，揭示了政治倾向与误分类模式之间的统计显著关联。总体而言，我们的研究结果表明，仅凭聚合的文本信号即可推断出版商的可靠性，支持基于内容的可扩展自动新闻来源评估方法。
cs.CL / 150 / 2609.24238
Memory vs. Context? Influential Factors of Factual Recall in Language Models
记忆与上下文？语言模型中事实回忆的影响因素
Abstract
We reproduce and stress-test the work of Yu et al. (2023), who characterize how language models (LMs) arbitrate between memorized knowledge and contradictory in-context statements. We replicate their world-capitals experiments on 31 models spanning Pythia, GPT-2, Qwen3, and Ministral families, including base and post-trained variants, and extend evaluations to five additional knowledge relation types from the ParaConflict dataset. We empirically confirm most of their original findings: larger models and higher-frequency entities tend to favor memorized answers, with substantial family-level variance. However, several conclusions do not generalize cleanly: entity-frequency effects disappear on Qwen3-14B and 32B; post-training shifts the memory-context trade-off inconsistently across families; question phrasing alone can change a model's reliance on memorized knowledge by up to 80 percentage points; and semantically unrelated prose can mimic coherent supporting context. Our results clarify where Yu et al.'s claims hold and to what extent they generalize to other prompts.
Chinese Translation
我们重现并对Yu等人（2023）的研究进行了压力测试，该研究描述了语言模型（LMs）如何在记忆知识和矛盾的上下文陈述之间进行权衡。我们在31个模型上复制了他们的世界首都实验，这些模型涵盖了Pythia、GPT-2、Qwen3和Ministral系列，包括基础和后训练变体，并将评估扩展到ParaConflict数据集中五种额外的知识关系类型。我们实证确认了他们大部分原始发现：较大的模型和高频实体倾向于偏向记忆答案，且存在显著的家族层面差异。然而，几个结论并未完全普遍适用：在Qwen3-14B和32B上，实体频率效应消失；后训练在不同家族间不一致地改变了记忆与上下文的权衡；仅仅是问题措辞就可以使模型对记忆知识的依赖变化高达80个百分点；而语义上无关的散文可以模拟连贯的支持上下文。我们的结果澄清了Yu等人的主张在何处成立，以及它们在多大程度上可以推广到其他提示。
cs.CL / 151 / 2609.24246
Taramandal-GPT: Enhancing Astrodynamics Problem-Solving with Knowledge Retrieval and Structured Thinking
Taramandal-GPT：通过知识检索和结构化思维增强天体动力学问题解决能力
Abstract
Large language models (LLMs) have shown remarkable progress in natural language understanding, yet their effectiveness in specialized fields like astronomy and astrodynamics remains limited due to challenges in multi-step reasoning, symbolic manipulation, and domain-specific terminology. To address this, we present Taramandal-GPT (Constellation-GPT), a domain-adapted framework built on the Qwen3-8b backbone, enhanced with a Retrieval-Augmented Generation (RAG) pipeline and a fallback mechanism for improved contextual precision. We evaluate it on the Astrodynamics Problems Benchmark (APBench), a dataset of 299 questions covering foundational to advanced levels of space science. Using a dual evaluation method - numeric margin-based scoring and semantic similarity assessment - Taramandal-GPT achieves competitive performance against state-of-the-art open- and closed-source models, with notable strength in thinking-intensive tasks. These results highlight the value of specialized LLMs for domains demanding accuracy and interpretability, positioning Taramandal-GPT as a step toward reliable Artificial Intelligence (AI) assistants for astrophysics, spacecraft engineering, and space exploration.
Chinese Translation
大型语言模型（LLMs）在自然语言理解方面取得了显著进展，但由于多步骤推理、符号操作和领域特定术语等挑战，它们在天文学和天体动力学等专业领域的有效性仍然有限。为了解决这一问题，我们提出了Taramandal-GPT（星座-GPT），这是一个基于Qwen3-8b骨架的领域适应框架，增强了检索增强生成（RAG）管道和回退机制，以提高上下文精确度。我们在天体动力学问题基准（APBench）上对其进行了评估，该数据集包含299个问题，涵盖了从基础到高级的空间科学内容。通过双重评估方法——基于数值边际的评分和语义相似性评估——Taramandal-GPT在与最先进的开源和闭源模型的比较中表现出竞争力，尤其在思维密集型任务中展现出显著优势。这些结果突显了专门化LLMs在需要准确性和可解释性的领域中的价值，使Taramandal-GPT成为朝着可靠的人工智能（AI）助手在天体物理学、航天器工程和太空探索领域迈出的重要一步。
cs.CL / 152 / 2609.24264
Canonical Procedural Actions: An Auditable Annotation Protocol for Tool-Use Agent Traces
规范程序行为：可审计的工具使用代理痕迹注释协议
Abstract
Tool-use agent traces identify messages and API calls, but procedural analyses also need explicit units of action and inspectable links to their evidence. We present Canonical Procedural Actions (CPAs), an annotation protocol that records a procedural function, its first agent-event anchor, the agent events that realize it, and separate contextual evidence. Multiple actions may share a message anchor without an inferred within-message order. A retail case study produces a versioned 24-entry codebook through open induction, recorded consolidation, and successive application audits. Two isolated LLM contexts annotate 32 trajectories disjoint from development at the trajectory level, producing 499 and 491 occurrences with anchor-label overlap A=0.982. Requiring identical context-event references reduces overlap to 0.798. These are structural repeatability measures, not semantic accuracy: 16 of 26 task IDs also occur in development, and historical tool payloads were truncated to 110 characters. Retrospective controls show that collapsing all labels raises overlap to 0.986, while simple endpoint rules reproduce the tool-anchored portion with 0.997 overlap. Assistant-message actions have 0.971 overlap, with a per-label minimum of 0.816. Applying the frozen codebook to 244 further trajectories yields 4,058 records, including eight diagnostic outcomes. The contribution is an explicit, auditable annotation instrument and a case study of its construction and measurement limits; human-reference validity and downstream utility remain to be established.
Chinese Translation
工具使用代理痕迹识别消息和API调用，但程序分析还需要明确的行为单元和可检查的证据链接。我们提出了规范程序行为（Canonical Procedural Actions, CPAs），这是一种注释协议，用于记录程序功能、其第一个代理事件锚点、实现该功能的代理事件以及单独的上下文证据。多个行为可以共享一个消息锚点，而不推断消息内部的顺序。通过开放归纳、记录整合和连续应用审计，零售案例研究产生了一个版本化的24条目代码本。两个孤立的LLM上下文在轨迹层面注释了32条与开发无关的轨迹，产生了499和491个具有锚点标签重叠的事件，重叠度A=0.982。要求相同的上下文事件引用将重叠度降低到0.798。这些是结构重复性度量，而非语义准确性：26个任务ID中有16个也出现在开发中，历史工具负载被截断为110个字符。回顾性控制显示，合并所有标签将重叠度提高到0.986，而简单的端点规则以0.997的重叠度再现了工具锚定部分。助手消息行为的重叠度为0.971，标签的最低重叠度为0.816。将冻结的代码本应用于244条进一步的轨迹产生了4,058条记录，包括八个诊断结果。该贡献是一个明确的、可审计的注释工具，以及其构建和测量限制的案例研究；人类参考的有效性和下游效用仍待建立。
cs.CL / 153 / 2609.24275
Structure Before Sampling: Community-Aware Core-Set Selection for Data-Efficient Text-to-Speech
采样前的结构：面向社区的核心集选择以实现数据高效的文本转语音
Abstract
Text-to-speech (TTS) corpora are costly to record, yet many utterances add little new phonetic information. Core-set selection reduces this cost by choosing a small training subset under a fixed audio-duration budget. We represent a corpus as a phonotactic graph that links each utterance to its most phonemically similar ones, and we first test whether this graph has structure. In Bangla and English corpora, its clustering is 199 and 56 times that of a size-matched random graph, and its modularity is more than twice that of a degree-preserving random graph. We then propose Community Representative, a selector that samples across graph communities and spreads its choices within each one, starting from utterances rich in rare phonemes. At every budget and in both languages, it covers more rare phoneme bigrams than random and entropy-based selection, and this lead holds on held-out utterances. TTS models trained on its 20% core-sets have a significantly lower character error rate (CER) than models trained on equal-duration random or entropy-based subsets in both languages. When all models train for the same number of epochs, the Bangla core-set model also outperforms full-corpus training (3.93% vs. 4.47% CER) with 4.5x less training time.
Chinese Translation
文本转语音（TTS）语料库的录制成本高昂，但许多发音所提供的新音位信息有限。核心集选择通过在固定的音频时长预算下选择一个小的训练子集来降低这一成本。我们将语料库表示为一个音位图，该图将每个发音与其最相似的音位发音相连接，并首先测试该图是否具有结构性。在孟加拉语和英语语料库中，其聚类程度分别是大小匹配随机图的199倍和56倍，其模块度是度保持随机图的两倍以上。接着，我们提出了社区代表（Community Representative），一种在图社区中进行采样并在每个社区内扩展选择的选择器，从富含稀有音素的发音开始。在每个预算下，在两种语言中，它覆盖的稀有音素双音节比随机选择和基于熵的选择要多，并且这一优势在保留的发音中依然存在。基于其20%核心集训练的TTS模型在两种语言中均显著降低了字符错误率（CER），相比于在相同时长的随机或基于熵的子集上训练的模型。当所有模型训练相同的轮次时，孟加拉语核心集模型的表现也优于全语料库训练（3.93% vs. 4.47% CER），且训练时间减少了4.5倍。
cs.CL / 154 / 2609.24357
Mitigating Entity Type Confusion in Cross-Domain NER via Multidimensional Quantification and Reasoning Enhancement
通过多维量化和推理增强缓解跨域命名实体识别中的实体类型混淆
Abstract
Cross-domain Named Entity Recognition (CD-NER) aims to transfer the rich knowledge in the source domain to the target domain. Recent studies adopting decomposition or generation paradigms have achieved significant performance improvements, demonstrating high accuracy in entity span detection. However, during entity type classification, models severely suffer from entity type confusion, the erroneous tendency that models classify entities of one type in the text as another similar but incorrect type. To address this issue, we first propose a Multidimensional Confusion Quantification Model (MCQM) that quantifies a model's confusion extent between entity types from three dimensions: source-target hierarchy analysis, semantic similarity analysis, and explicit data evaluation. Moreover, we propose the Progressive Bidirectional Reasoning Chain (PBRC). PBRC leverages the source-target hierarchy and confusion analysis from the MCQM to prompt the LLM to generate two-stage reasoning information. The two-stage reasoning information is utilized to augment the knowledge of the model, significantly mitigating entity type confusion and improving the model's generalization performance. Experimental results demonstrate that our method achieves new state-of-the-art results on all domains of the CrossNER dataset.
Chinese Translation
跨域命名实体识别（CD-NER）旨在将源域中的丰富知识转移到目标域。最近采用分解或生成范式的研究取得了显著的性能提升，在实体跨度检测中表现出高准确性。然而，在实体类型分类过程中，模型严重受到实体类型混淆的影响，即模型将文本中一种类型的实体错误地分类为另一种相似但不正确的类型。为了解决这一问题，我们首先提出了一种多维混淆量化模型（MCQM），该模型从三个维度量化模型在实体类型之间的混淆程度：源-目标层次分析、语义相似性分析和显性数据评估。此外，我们提出了渐进双向推理链（PBRC）。PBRC利用MCQM中的源-目标层次和混淆分析，促使大型语言模型（LLM）生成两阶段推理信息。两阶段推理信息用于增强模型的知识，显著缓解实体类型混淆并提高模型的泛化性能。实验结果表明，我们的方法在CrossNER数据集的所有领域上达到了新的最先进结果。
cs.CL / 155 / 2609.24372
URA-NER: A Unified Retrieval-Augmented Framework with Retrieval Alignment and Uncertainty Reduction for Low-Resource NER
URA-NER：一种统一的检索增强框架，结合检索对齐和不确定性降低，应用于低资源命名实体识别
Abstract
In-context learning (ICL) based on large language models (LLMs) has shown promising potential in alleviating performance bottlenecks caused by the limited availability of annotated data in Named Entity Recognition (NER). However, existing methods still face issues of retrieval misalignment and generation uncertainty, making their performance heavily dependent on the LLM's capabilities. As the parameter scale of LLMs decreases, their performance in few-shot settings deteriorates significantly. In this paper, we propose a novel unified retrieval-augmented framework, URA-NER, including three key components: Progressive Granularity Retrieval (PGR), Model-aware Representation Enhancement (MaRE), and Reason-aware Knowledge Verification. PGR is a two-stage retrieval mechanism that achieves stage alignment. It first retrieves demonstrations for span detection based on the query's global semantics, and then for type classification based on the specific entity context, providing fine-grained local information. Moreover, MaRE employs entity pre-recognition to guide the construction of representations, ensuring the query and demonstrations are aligned within the LLM's semantic space and attention pattern. In addition, to mitigate generation uncertainty, we propose RaKV, a closed-loop "generation-retrieval-verification" process. It explicates the LLM's reasoning paths, leverages them for the retrieval of external knowledge, and reorganizes the knowledge into verification evidence aligned with the original reasoning paths. We conduct extensive experiments on multiple low-resource NER datasets. Results demonstrate that URA-NER significantly enhances the performance of LLMs under low-resource settings, with particularly pronounced gains for smaller LLMs, achieving new state-of-the-art results on several benchmarks.
Chinese Translation
基于大型语言模型（LLMs）的上下文学习（ICL）在缓解命名实体识别（NER）中由于标注数据有限而导致的性能瓶颈方面展现了良好的潜力。然而，现有方法仍面临检索不对齐和生成不确定性的问题，使得它们的性能在很大程度上依赖于LLM的能力。随着LLM参数规模的减小，它们在少量样本设置下的性能显著下降。本文提出了一种新颖的统一检索增强框架URA-NER，包括三个关键组件：渐进粒度检索（Progressive Granularity Retrieval, PGR）、模型感知表示增强（Model-aware Representation Enhancement, MaRE）和推理感知知识验证。PGR是一种两阶段的检索机制，实现了阶段对齐。它首先基于查询的全局语义检索用于跨度检测的示例，然后基于特定实体上下文进行类型分类，提供细粒度的局部信息。此外，MaRE利用实体预识别来指导表示的构建，确保查询和示例在LLM的语义空间和注意力模式中对齐。此外，为了减轻生成不确定性，我们提出了RaKV，一个闭环的“生成-检索-验证”过程。它阐明了LLM的推理路径，利用这些路径检索外部知识，并将知识重组为与原始推理路径对齐的验证证据。我们在多个低资源NER数据集上进行了广泛的实验。结果表明，URA-NER显著提升了LLM在低资源设置下的性能，尤其是对于较小的LLM，在多个基准上达到了新的最先进结果。
cs.CL / 156 / 2609.24410
End-to-end Jordanian dialect speech-to-text self-supervised learning framework
端到端约旦方言语音转文本自监督学习框架
Abstract
Speech-to-text engines are extremely needed nowadays for different applications, representing an essential enabler in human-robot interaction. Still, some languages suffer from the lack of labeled speech data, especially in the Arabic dialects or any low-resource languages. The need for a self-supervised training process and self-training using noisy training is proven to be one of the up-and-coming feasible solutions. This article proposes an end-to-end, transformers-based model with a framework for low-resource languages. In addition, the framework incorporates customized audio-to-text processing algorithms to achieve a highly efficient Jordanian Arabic dialect speech-to-text system. The proposed framework enables ingesting data from many sources, making the ground truth from external sources possible by speeding up the manual annotation process. The framework allows the training process using noisy student training and self-supervised learning to utilize the unlabeled data in both pre- and post-training stages and incorporate multiple types of data augmentation. The proposed self-training approach outperforms the fine-tuned Wav2Vec model by 5% in terms of word error rate reduction. The outcome of this work provides the research community with a Jordanian-spoken data set along with an end-to-end approach to deal with low-resource languages. This is done by utilizing the power of the pretraining, post-training, and injecting noisy labeled and augmented data with minimal human intervention. It enables the development of new applications in the field of Arabic language speech-to-text area like the question-answering systems and intelligent control systems, and it will add human-like perception and hearing sensors to intelligent robots.
Chinese Translation
语音转文本引擎在当今的各种应用中极为重要，是人机交互的基本推动力。然而，一些语言，尤其是阿拉伯方言或任何低资源语言，面临着标注语音数据不足的问题。自监督训练过程和使用噪声训练的自我训练被证明是新兴的可行解决方案之一。本文提出了一种基于变换器的端到端模型框架，旨在支持低资源语言。此外，该框架结合了定制的音频到文本处理算法，以实现高效的约旦阿拉伯方言语音转文本系统。所提框架能够从多个来源获取数据，通过加速人工标注过程，使得从外部来源获取真实标签成为可能。该框架允许使用噪声学生训练和自监督学习进行训练过程，以利用未标注数据在预训练和后训练阶段，并结合多种数据增强类型。所提出的自我训练方法在词错误率降低方面比微调后的 Wav2Vec 模型提高了 5%。这项工作的成果为研究社区提供了一个约旦口语数据集，以及一种处理低资源语言的端到端方法。这是通过利用预训练、后训练的能力，并以最小的人为干预注入噪声标注和增强数据来实现的。它使得在阿拉伯语言语音转文本领域开发新应用成为可能，如问答系统和智能控制系统，并将人类般的感知和听觉传感器添加到智能机器人中。
cs.CL / 157 / 2609.24516
LLJ Cards: Best practices for the Use of LLMs as Judges
LLJ卡片：作为评估者使用大型语言模型的最佳实践
Abstract
In recent years, large language models (LLMs) have emerged as a popular alternative for evaluation. Often referred to as LLMs as judges (LLJs), these systems have been widely adopted by researchers and practitioners across a broad range of measurement tasks, driven by their strong performance, scalability, and cost-effectiveness relative to human judgment. However, a growing body of work has shown that the use of LLJs raise concerns about their validity and reliability as evaluators. Existing efforts to address these challenges have largely focused on developing bias-mitigation techniques and refining prompting strategies. While these approaches represent an important step forward, they primarily offer technical fixes and leave a more fundamental challenge unaddressed: the lack of standardized, transparent, and reproducible evaluation practices. In this paper, we introduce LLJ Cards, a framework that synthesizes best practices from measurement theory, natural language generation, and machine learning literature into practical guidelines for LLJ-based evaluations. While LLJs offer a promising path toward scalable evaluation, their effective use requires grounding in rigorous evaluation principles to ensure validity, reliability, and reproducibility. LLJ Cards addresses this need by providing a structured framework for applying these principles in the design and reporting of automated evaluations.
Chinese Translation
近年来，大型语言模型（LLMs）作为评估的流行替代方案逐渐崭露头角。这些系统通常被称为作为评估者的LLMs（LLJs），因其强大的性能、可扩展性和相较于人类判断的成本效益而被研究人员和实践者广泛采用。然而，越来越多的研究表明，使用LLJs引发了关于其作为评估者的有效性和可靠性的担忧。现有的应对措施主要集中在开发偏见缓解技术和完善提示策略上。尽管这些方法代表了向前迈出的重要一步，但它们主要提供了技术解决方案，未能解决一个更根本的挑战：缺乏标准化、透明和可重复的评估实践。本文介绍了LLJ卡片，这是一种将测量理论、自然语言生成和机器学习文献中的最佳实践综合为LLJ基础评估的实用指南的框架。尽管LLJs为可扩展评估提供了有希望的路径，但其有效使用需要扎根于严格的评估原则，以确保有效性、可靠性和可重复性。LLJ卡片通过提供一个结构化框架，满足了在自动化评估的设计和报告中应用这些原则的需求。
cs.CL / 158 / 2609.24538
QLoRA Fine-Tuning of Ministral LLM for Sequence-to-Function Protein Annotation
使用QLoRA对Ministral LLM进行微调以实现序列到功能的蛋白质注释
Abstract
Functional annotation of newly sequenced proteins remains a bottleneck in molecular biology: the number of sequences in public repositories grows far faster than the capacity for manual curation. Most computational approaches consider annotation as multi-label classification over a fixed ontology, which constrains predictions to a predefined label set. In this work we study the the protein annotation as a sequence-to-text generation problem. We fine-tune the 3B-parameter Ministral 3 base model with QLoRA (4-bit NF4 quantization with low-rank adapters) on sequence annotation pairs. We assess predictions with an LLM-as-expert protocol: a GPT model prompted as a senior molecular-biology curator scores organism identification as binary and function annotation quality. We conclude that QLoRA-fine-tuned compact LLMs can generate curator-style annotations with genuine biological value for a substantial subset of proteins. We also discuss future directions in data quality, model scaling, and evidence grounding that are needed to make the approach sufficiently reliable for practical use.
Chinese Translation
新测序蛋白质的功能注释仍然是分子生物学中的一个瓶颈：公共数据库中序列的数量增长速度远远超过人工整理的能力。大多数计算方法将注释视为在固定本体上的多标签分类，这限制了预测在预定义标签集内。在本研究中，我们将蛋白质注释视为序列到文本生成问题。我们使用QLoRA（4位NF4量化与低秩适配器）对3B参数的Ministral 3基础模型进行微调，以处理序列注释对。我们采用LLM-as-expert协议评估预测：将GPT模型作为资深分子生物学策展人进行提示，评分包括生物体识别的二元性和功能注释质量。我们得出结论，经过QLoRA微调的紧凑型LLM能够为大量蛋白质生成具有真正生物学价值的策展风格注释。我们还讨论了在数据质量、模型扩展和证据基础方面的未来方向，这些都是使该方法在实际应用中足够可靠所需的。
cs.CL / 159 / 2609.24554
Toward a Unified Mathematics of Concepts
迈向统一的概念数学
Abstract
Concepts are commonly defined as abstract, compact representations of knowledge and treated as basic units of intelligent behavior. Yet, cognition, psychology, and AI lack a shared mathematical language for them. Modern systems represent concepts as vectors, distributions, symbols, graphs, and other structures, but these formalisms are typically treated as competing rather than as solutions to a common problem. We propose an operation-based view that evaluates mathematical frameworks by the conceptual operations they support, identifying thirteen operations (including similarity, composition, generalization, and grounding) that recur across cognition, psychology, and AI. We show that ten frameworks embody distinct commitments to concepts as self-contained content, relational structure, or evolving process, and that these commitments determine which operations each supports naturally. For example, vector-based models facilitate graded similarity and generalization but struggle with explicit composition, whereas symbolic models support composition but offer but generalize poorly. No single framework we examined naturally supports all operations without extension. We test this account empirically using categorization as a case study, operationalizing nine theories on the same items against human judgments. Despite addressing the same conceptual question, the theories produce different procedures and results, demonstrating that mathematical commitment shapes what a theory can explain. We call for hybrid formalisms that treat content, relation, and process as jointly primary.
Chinese Translation
概念通常被定义为知识的抽象、紧凑的表征，并被视为智能行为的基本单元。然而，认知、心理学和人工智能缺乏一个共享的数学语言来描述它们。现代系统将概念表示为向量、分布、符号、图形和其他结构，但这些形式通常被视为相互竞争，而不是作为解决共同问题的方案。我们提出了一种基于操作的视角，通过支持的概念操作来评估数学框架，识别出在认知、心理学和人工智能中反复出现的十三种操作（包括相似性、组合、概括和基础）。我们展示了十个框架在概念作为自包含内容、关系结构或演变过程方面的不同承诺，这些承诺决定了每个框架自然支持哪些操作。例如，基于向量的模型促进了渐进相似性和概括，但在显式组合方面存在困难，而符号模型支持组合但在概括上表现不佳。我们考察的没有任何单一框架自然支持所有操作而不进行扩展。我们通过使用分类作为案例研究，实证测试了这一观点，将九种理论在相同项目上与人类判断进行操作化比较。尽管这些理论解决的是相同的概念问题，但它们产生了不同的程序和结果，证明了数学承诺塑造了理论能够解释的内容。我们呼吁采用混合形式主义，将内容、关系和过程视为共同的主要因素。
cs.CL / 160 / 2609.24574
Evaluating Decision Models for Text Annotation in Computational Social Science
评估计算社会科学中文本标注的决策模型
Abstract
Computational social science increasingly relies on large language models for text annotation, and the validity of published findings now rests on the labels generated by such models. Decision models, a new model class built for categorical question answering, answer typed questions with a choice, a probability distribution over the label set, and a confidence score rather than free text, at a small fraction of frontier inference prices. Whether their answers are accurate, and whether that stated confidence can be trusted on social science constructs, are unknown. Here, we mirror the evaluation of Ziems et al. (2024) on 18 computational social science classification tasks (7,977 items), comparing the first commercial decision model and two open-weight counterparts against 19 frontier and open-weight language models under the same zero-shot protocol. The decision model trails the per-task best LLM on 14 of 15 evaluation tasks, with a median deficit of 11.6 macro-F1 points, at a median 44 times lower measured cost. Its confidence is better calibrated than the verbalized confidence of 16 of the 19 LLMs, yet three frontier models show lower median calibration error (0.157 against 0.066). While items above 0.9 confidence are typically labeled accurately (median accuracy 0.815), on one task, empathy in peer-support dialogues, the model reports high confidence while performing near chance. Nonetheless, our results suggest that decision models are useful as a first step in the annotation pipeline: routing low-confidence items to an LLM matches or exceeds the LLM alone at a quarter to half of its cost.
Chinese Translation
计算社会科学日益依赖大型语言模型进行文本标注，而已发布研究结果的有效性现在依赖于这些模型生成的标签。决策模型是一种新型模型类别，专为分类问题回答而构建，它通过选择、标签集上的概率分布和置信度分数来回答类型问题，而不是自由文本，以极低的前沿推理价格进行。它们的答案是否准确，以及所述的置信度在社会科学构念上是否可信，目前尚不清楚。在此，我们对Ziems等人（2024）在18个计算社会科学分类任务（7,977个项目）上的评估进行了镜像，比较了第一个商业决策模型和两个开放权重的对应模型在相同零样本协议下与19个前沿和开放权重语言模型的表现。决策模型在15个评估任务中有14个任务的表现落后于每个任务最佳的LLM，且中位数缺口为11.6个宏F1分数，测得的成本中位数低44倍。其置信度的校准优于19个LLM中16个的口头置信度，然而三种前沿模型显示出较低的中位数校准误差（0.157对比0.066）。尽管置信度超过0.9的项目通常被准确标记（中位数准确率为0.815），但在一个任务中，即同伴支持对话中的同理心，该模型报告了高置信度，但其表现接近随机。然而，我们的结果表明，决策模型作为标注流程的第一步是有用的：将低置信度项目路由到LLM的效果与单独使用LLM相当或更好，且成本仅为其四分之一到一半。
cs.CL / 161 / 2609.24635
Written as a Record, Read as an Address: What a Forward Pass Leaves in an Operation's KV Cache
作为记录书写，作为地址阅读：前向传播在操作的KV缓存中留下了什么
Abstract
When a language model reads an operation such as "Swap the contents of Box F and Box B", its forward pass writes keys and values for those tokens into the KV cache. Prior work on entity tracking establishes what models use: bindings are resolved at query time rather than stored as explicit latent state. We ask what they write at the operation span and how it is accessed. We split a forward pass into a frozen writer and a reader: the writer's cache is recomputed without gradients, while the reader sees only the instruction and operation tokens, with all state descriptions hidden, and is trained in isolation. Anything the reader recovers was therefore already present in the unmodified cache. On a synthetic boxes task, a base reader recovers $\leq 0.06$ of queried bindings against $0.75$--$1.00$ after training, and recoverability tracks the operation's read/write footprint. We find two modes of access. Across Llama-3.1-8B and Mistral-7B, operation-span transplants causally redirect which visible state is read even when the two worlds hold identical values, revealing a routing record. Isolation training preserves routing and adds direct access to the payload, the value the operation read, from the single operand-name token in a narrow mid-depth band (layers 12--15 of 32 in Llama-3.1-8B, 14--17 in Mistral-7B) --- the same site that holds the routing record. The same recipe extends to further operations, ToMi and GSM8K, but is bounded by training coverage and costs open-book accuracy. Operation tokens thus leave localized, causally recoverable records that support both routing and direct payload access, though the model that writes them reads mainly the address they carry and not the value.
Chinese Translation
当语言模型读取诸如“交换盒子F和盒子B的内容”这样的操作时，其前向传播会将这些标记的键和值写入KV缓存。先前关于实体跟踪的研究确定了模型的使用方式：绑定在查询时解析，而不是作为显式潜在状态存储。我们探讨它们在操作范围内写入了什么以及如何访问。我们将前向传播分为一个冻结的写入器和一个读取器：写入器的缓存在没有梯度的情况下重新计算，而读取器仅看到指令和操作标记，所有状态描述都被隐藏，并在隔离中进行训练。因此，读取器恢复的任何内容在未修改的缓存中已经存在。在一个合成盒子任务中，基础读取器在训练后恢复的查询绑定不超过0.06，而恢复率在0.75到1.00之间，且恢复能力与操作的读/写足迹相关。我们发现两种访问模式。在Llama-3.1-8B和Mistral-7B中，操作范围的移植因果性地重定向了可见状态的读取，即使两个世界持有相同的值，揭示了一个路由记录。隔离训练保留了路由并增加了对有效载荷的直接访问，即操作读取的值，从一个狭窄的中深度带中的单个操作数名称标记（在Llama-3.1-8B的第12到15层，Mistral-7B的第14到17层）——与持有路由记录的同一位置相同。相同的方法扩展到进一步的操作，如ToMi和GSM8K，但受限于训练覆盖率并影响开放书籍的准确性。因此，操作标记留下了局部的、因果可恢复的记录，支持路由和直接有效载荷访问，尽管写入它们的模型主要读取它们携带的地址而不是值。
cs.CL / 162 / 2609.24650
Assessing Readability with LLMs: The Role of Reasoning and Few-Shot Prompting
使用大型语言模型评估可读性：推理与少量示例提示的作用
Abstract
Readability assessment is essential for tailoring texts to intended audiences across educational, healthcare, and information retrieval domains. However, traditional readability formulas struggle to generalize across genres and languages, while supervised machine learning models rely on scarce, domain-specific annotated corpora, limiting their applicability--particularly for less-resourced languages. Large Language Models (LLMs) offer a highly scalable, multilingual alternative that requires no task-specific training, yet the impact of advanced prompting strategies on their performance remains underexplored. In this paper, we conduct a systematic benchmark of diverse open-source LLMs for multilingual readability assessment, focusing on the prediction of discrete readability levels required by educational frameworks. In addition to English, we evaluate our approach on a less-resourced language, Slovenian, to establish whether LLMs remain effective in low-resource settings. Specifically, we investigate the influence of explicit reasoning, demonstrating that Chain-of-Thought (CoT) prompting and reasoning-oriented models yield significant improvements over direct answering. Furthermore, our exploration of few-shot in-context learning reveals that providing just one labelled example per category (1-shot) substantially enhances prediction quality compared to zero-shot settings, with additional examples offering diminishing returns. By comprehensively comparing these approaches against traditional unsupervised metrics and state-of-the-art supervised baselines, we establish the viability of out-of-the-box LLMs as robust, cross-lingual readability assessors.
Chinese Translation
可读性评估对于根据目标受众调整文本在教育、医疗和信息检索领域至关重要。然而，传统的可读性公式在不同体裁和语言之间的推广能力有限，而监督机器学习模型依赖于稀缺的领域特定标注语料库，限制了它们的适用性——尤其是在资源较少的语言中。大型语言模型（LLMs）提供了一种高度可扩展的多语言替代方案，无需特定任务的训练，但先进提示策略对其性能的影响仍未得到充分探讨。在本文中，我们对多种开源LLMs进行系统基准测试，以评估其在多语言可读性评估中的表现，重点关注教育框架所需的离散可读性水平预测。除了英语，我们还在一个资源较少的语言——斯洛文尼亚语上评估我们的方法，以确定LLMs在低资源环境中的有效性。具体而言，我们研究了明确推理的影响，表明链式思维（Chain-of-Thought, CoT）提示和以推理为导向的模型在直接回答上取得了显著改善。此外，我们对少量上下文学习的探索表明，与零-shot设置相比，每个类别仅提供一个标记示例（1-shot）显著提高了预测质量，而额外示例的收益递减。通过将这些方法与传统无监督指标和最先进的监督基线进行全面比较，我们确立了即插即用的LLMs作为强大跨语言可读性评估工具的可行性。
cs.CL / 163 / 2609.24698
Adapting Tree-Structured Speculative Decoding to DeepSeek-V4 for Efficient Inference
将树结构推测解码适配于 DeepSeek-V4 以实现高效推理
Abstract
Repeated execution of the target model during autoregressive decoding is a major source of LLM inference latency. Unlike linear speculation, which follows a single candidate chain, tree-structured speculation retains multiple branches from shared prefixes; under the same budget, this broader coverage can improve acceptance and efficiency. Adapting it to DeepSeek-V4 is nontrivial: its CSA/HCA online compressed attention concentrates the difficulty on the target-verify side, where branches diverging from a shared prefix compress into different states, breaking cross-branch state consistency. We integrate tree-structured speculative decoding into the DeepSeek-V4-Flash pipeline via branch-aware causal verification, temporary state isolation, and accepted-path state refresh, keeping verification and compressed-state updates consistent across branches. Across budgets D=5 to D=8, batch sizes 1 to 64, and three datasets (GSM8K, MBPP, ShareGPT), tree speculation achieves a higher accepted length than the matched linear configurations in all settings (e.g., at D=8 about 2.83--3.41 versus 2.39--2.84) and improves throughput in nearly all configurations---marginal only at the smallest budget---by up to about 18.5%. More importantly, the gains follow stable, transferable regularities: the relative gain grows with the budget and is most pronounced for less predictable workloads at small-to-medium batch sizes, while beyond a certain budget throughput plateaus and decouples from the still-rising accepted length. These results show that retaining multiple candidate paths under the same budget can effectively improve DeepSeek-V4 decoding efficiency, and offer experience for adapting speculative decoding to future models with compressed, sparse, or structured context representations.
Chinese Translation
在自回归解码过程中，目标模型的重复执行是大型语言模型（LLM）推理延迟的主要来源。与仅遵循单一候选链的线性推测不同，树结构推测保留了来自共享前缀的多个分支；在相同预算下，这种更广泛的覆盖可以提高接受率和效率。将其适配于 DeepSeek-V4 并非易事：其 CSA/HCA 在线压缩注意力将难点集中在目标验证侧，其中从共享前缀分叉的分支压缩为不同状态，打破了跨分支状态的一致性。我们通过分支感知的因果验证、临时状态隔离和接受路径状态刷新，将树结构推测解码集成到 DeepSeek-V4-Flash 流水线中，保持跨分支的验证和压缩状态更新的一致性。在预算 D=5 到 D=8、批量大小 1 到 64 以及三个数据集（GSM8K、MBPP、ShareGPT）中，树结构推测在所有设置中均实现了比匹配的线性配置更高的接受长度（例如，在 D=8 时约为 2.83--3.41 对比 2.39--2.84），并在几乎所有配置中提高了吞吐量——在最小预算下仅有边际提升——最高可达约 18.5%。更重要的是，这些增益遵循稳定、可转移的规律：相对增益随着预算的增加而增长，并且在小到中等批量大小下对于不太可预测的工作负载最为明显，而在超过某一预算后，吞吐量趋于平稳，并与仍在上升的接受长度脱钩。这些结果表明，在相同预算下保留多个候选路径可以有效提高 DeepSeek-V4 的解码效率，并为将推测解码适配于未来具有压缩、稀疏或结构化上下文表示的模型提供了经验。
cs.CL / 164 / 2609.24799
When Quantization Preserves Accuracy but Not Evidence: Explanation-Aware Post-Training Quantization for Medical LLMs
量化在保留准确性但不保留证据时的解释意识后训练量化：针对医疗大语言模型的研究
Abstract
Post-training quantization (PTQ) enables efficient deployment of large language models, and PTQ methods are usually optimized and evaluated with generic reconstruction, perplexity, or answer accuracy. But in explanation-critical domains, preserving only the final answer may be insufficient, since users may also inspect generated rationales to judge whether a prediction is trustworthy. We study this issue in medical multiple-choice question answering, where rationales should provide evidence that supports the selected answer. We propose an explanation-aware objective for transformation-based PTQ. Our method builds an offline faithfulness cache from full-precision teacher rationales and uses it during optimization to preserve answer-supporting evidence tokens and evidence-conditioned answer behavior. We instantiate it on OSTQuant under W4A4KV4 quantization and evaluate four 7B--8B medical and instruction-tuned LLMs on MedExQA, MedExpQA, and ChallengeClinicalQA. While a same-calibration OSTQuant baseline preserves task accuracy, it can substantially weaken answer-supporting rationales. Our objective is to preserve the full-precision model's answer-supporting behavior rather than improve gold-label accuracy, and our method better preserves the full-precision model's answer behavior and rationale-to-answer support. These results suggest that PTQ for explanation-critical settings should evaluate preservation of answer-supporting evidence, not only answer accuracy. Code and evaluation scripts are available at https://github.com/dut0817/EAQuant.
Chinese Translation
后训练量化（PTQ）使得大型语言模型的高效部署成为可能，而PTQ方法通常通过通用重建、困惑度或答案准确性进行优化和评估。但在对解释要求严格的领域，仅保留最终答案可能不足，因为用户可能还会检查生成的推理，以判断预测是否可信。我们在医疗多项选择题回答中研究了这一问题，其中推理应提供支持所选答案的证据。我们提出了一种基于解释意识的目标，用于基于转换的PTQ。我们的方法从全精度教师推理中构建一个离线可信度缓存，并在优化过程中使用它，以保留支持答案的证据标记和基于证据的答案行为。我们在W4A4KV4量化下的OSTQuant上实现了该方法，并在MedExQA、MedExpQA和ChallengeClinicalQA上评估了四个7B--8B的医疗和指令调优大语言模型。虽然相同校准的OSTQuant基线保留了任务准确性，但它可能会显著削弱支持答案的推理。我们的目标是保留全精度模型的答案支持行为，而不是提高金标准准确性，我们的方法更好地保留了全精度模型的答案行为和推理对答案的支持。这些结果表明，针对解释要求严格的设置，PTQ应评估对答案支持证据的保留，而不仅仅是答案准确性。代码和评估脚本可在 https://github.com/dut0817/EAQuant 获取。
cs.CL / 165 / 2609.24812
MSI-Bench: Evaluating Multi-Speaker Voice Interaction for Collaborative AI Agents
MSI-Bench：评估多说话者语音交互以支持协作AI代理
Abstract
Voice provides a natural and immediate interface for AI agents. Many settings in which voice agents could be useful, including meetings, households, and collaborative work, are inherently multi-speaker. Supporting these settings introduces challenges that are largely absent from one-on-one interaction. We introduce the Multi-Speaker Interaction Benchmark (MSI-Bench) for evaluating multi-speaker voice interaction. Each test case is a short multi-party multi-turn audio scene with participant context, expected tool calls, and atomic rubrics. The benchmark targets three capability families: multi-speaker memory, multi-speaker instruction following, and multi-speaker reasoning. It comprises 1,152 test cases, evenly split between Mandarin Chinese and English (576 each). The strongest configuration on each split passes all rubrics on only 66.8% of English and 54.5% of Mandarin cases, and the strongest open-weight configuration on 34.0% and 19.3%. Failure analysis separates perception from reasoning: open-weight models are bottlenecked by the multi-speaker audio front-end, while frontier systems still fail speaker-scoped decision making on clean transcripts---and models across the board often respond when no one has addressed them. These results identify speaker-grounded perception, speaker-scoped decision making, and conversational restraint as concrete targets for future voice agents.
Chinese Translation
语音为AI代理提供了一种自然且即时的接口。在会议、家庭和协作工作等多说话者场景中，语音代理的潜在应用非常广泛。然而，支持这些场景带来了许多在一对一交互中几乎不存在的挑战。我们提出了多说话者交互基准（Multi-Speaker Interaction Benchmark，MSI-Bench）以评估多说话者语音交互。每个测试案例都是一个短小的多方多轮音频场景，包含参与者上下文、预期工具调用和原子评分标准。该基准针对三个能力类别：多说话者记忆、多说话者指令遵循和多说话者推理。它包含1,152个测试案例，普通话和英语各576个。每个分组中最强的配置仅在66.8%的英语案例和54.5%的普通话案例中通过了所有评分标准，而最强的开放权重配置分别为34.0%和19.3%。失败分析将感知与推理分开：开放权重模型受到多说话者音频前端的瓶颈限制，而前沿系统在清晰的转录文本上仍然无法进行说话者范围的决策——而且各类模型往往在没有人提及它们时也会做出响应。这些结果确定了以说话者为基础的感知、说话者范围的决策以及对话的克制作为未来语音代理的具体目标。
cs.CL / 166 / 2609.24821
The Answer-Basin Representation Hypothesis: We Are Not Probing or Steering Concepts
答案盆地表征假说：我们并不是在探测或引导概念
Abstract
The Linear Representation Hypothesis associates high-level concepts with directions in language models, but it remains unclear how these concept-related linear structures are organized within the model. We propose the Answer-Basin Representation Hypothesis: the probability measure induced over answers by the model's continuation distribution organizes these linear structures, with its statistics represented along linear directions shared across questions. All continuations yielding the same answer form an answer basin, whose mass is their total probability. These basin masses define the pushforward probability measure over answers. We posit that concept-related linear structure emerges from differences in the answer measure rather than being determined by changes in concept labels. Experiments across models and tasks link concept-consistent effects and their reversals in probing and steering to the alignment between concept labels and the answer measure.
Chinese Translation
线性表征假说将高级概念与语言模型中的方向关联起来，但这些与概念相关的线性结构在模型中的组织方式仍不清楚。我们提出答案盆地表征假说：模型的延续分布所诱导的答案概率测度组织了这些线性结构，其统计信息沿着跨问题共享的线性方向表示。所有产生相同答案的延续形成一个答案盆地，其质量为它们的总概率。这些盆地质量定义了答案上的推前概率测度。我们假设，与概念相关的线性结构是由于答案测度的差异而出现的，而不是由概念标签的变化所决定的。跨模型和任务的实验将概念一致效应及其在探测和引导中的反转与概念标签与答案测度之间的对齐联系起来。
cs.CL / 167 / 2609.24877
Decomposing Error and Style in Automated Clinical Coding
自动临床编码中的错误与风格分解
Abstract
In automated clinical coding, where the label space spans tens of thousands of diagnosis and procedure codes, models are currently evaluated against a single gold annotation, treating any deviation as error. But we find when two teams code the same 110 ACI-Bench encounters, they agree on only 73% of codes (Jaccard similarity) for the same note; even after an independent clinical audit removes erroneous codes, agreement rises only to 77%. Is that gap error or something systematic? We model the systematic component as coding style $\psi$, a coder- or site-specific policy over what to code and how much to document, and recast coding as $p(\mathrm{code}\mid\mathrm{note},\psi)$, estimating $\psi$ with a 10-dimension rubric. If style were noise, conditioning on it would do nothing. Instead, across five datasets a model conditioned with a data-matching style raises ICD F1 by up to 26 points and an extreme mismatched one lowers it by up to 21. Four prompt based coding methods spanning 39-49 F1 converge to 52-56 once style is supplied (All p<0.05). Much of what single-gold evaluation charges to model error is recoverable, unmodeled style.
Chinese Translation
在自动临床编码中，标签空间涵盖了数万个诊断和程序代码，目前模型是根据单一的金标准注释进行评估，将任何偏差视为错误。但我们发现，当两个团队对同一110个ACI-Bench案例进行编码时，他们在同一笔记上的代码一致性仅为73%（Jaccard相似度）；即使在独立的临床审计去除错误代码后，一致性也仅提高到77%。这个差距是错误还是某种系统性因素？我们将系统性成分建模为编码风格$ heta$，即编码者或站点特定的关于编码内容和文档记录程度的政策，并将编码重构为$p( ext{code} ext{mid} ext{note}, heta)$，使用10维标准来估计$ heta$。如果风格是噪声，条件化它将没有任何作用。相反，在五个数据集中，使用数据匹配风格条件化的模型使ICD F1提高了多达26个点，而极端不匹配的风格则使其降低了多达21个点。四种基于提示的编码方法在39-49 F1的范围内，一旦提供风格，便收敛到52-56（所有p<0.05）。许多单一金标准评估所指责的模型错误是可恢复的、未建模的风格。
cs.CL / 168 / 2609.24885
The Copy Ceiling: An Input-Exposure Control for Ontology-Grounded Generation over Curated Corpora
复制上限：基于本体的生成在策划语料库上的输入暴露控制
Abstract
When a language model answers from a curated corpus via graph-based retrieval, a large grounding uplift does not establish reasoning over the retrieved structure: the context may already expose the gold answers. We propose exposure accounting, which classifies each gold item by whether the shown context exposes it and whether the answer recovers it. Its scalar reference is the copy ceiling, the recall a verbatim copy of the context achieves; signed gain over copy measures the model's recall relative to this deterministic, judge-free baseline. Across ten models, unaided recall averages 0.26 and grounded recall 0.92, yet gain over copy is uniformly negative (-0.067 to -0.022). Of 11,360 gold-item observations, representing 1,136 target instances evaluated under ten models, only three unexposed items receive lexical credit. A stratified model-judged audit of 423 observations, with a symmetric quotation-verification policy, estimates that 97.1% of credited items assert the requested relation; all three unexposed credits fail relational adjudication. On targets the scaffold does not expose, lexical recovery falls from 0.121 unaided to 0.004 grounded; adjudication validates 71 of the 92 unaided credits and none of the three grounded credits, without establishing full-frame relational recovery rates. Rephrasing questions outside the graph's title vocabulary reduces exposure from 0.964 to 0.328, while an absence-triggered fallback activates on only 2 of 506 questions. A paired production study improves judged quality by +0.27 pooled, but negative controls do not establish content specificity beyond a well-formed on-corpus block. These results support exposure accounting as a standing control for corpus-derived evaluations. The accounting distinguishes exposed-item omissions from beyond-exposure recoveries; it does not determine whether reasoning occurred.
Chinese Translation
当语言模型通过基于图的检索从策划语料库中回答时，大量的基础提升并不能在检索结构上建立推理：上下文可能已经暴露了黄金答案。我们提出了暴露会计，它通过显示的上下文是否暴露了每个黄金项目以及答案是否恢复了它来对每个黄金项目进行分类。其标量参考是复制上限，即逐字复制上下文所达到的召回率；相对于这一确定性、无评判基线的复制增益衡量模型的召回。在十个模型中，未辅助召回的平均值为0.26，而基础召回为0.92，但复制增益均为负值（-0.067至-0.022）。在11,360个黄金项目观察中，代表在十个模型下评估的1,136个目标实例，只有三个未暴露的项目获得了词汇信用。对423个观察的分层模型评估审计，采用对称引用验证政策，估计97.1%的获得信用的项目主张所请求的关系；所有三个未暴露的信用未能通过关系裁定。在支架未暴露的目标上，词汇恢复率从0.121（未辅助）下降到0.004（基础）；裁定验证了92个未辅助信用中的71个，而三个基础信用均未通过验证，未建立完整框架的关系恢复率。将问题重新表述为图的标题词汇之外的内容，将暴露率从0.964降低到0.328，而缺失触发的回退仅在506个问题中激活了2个。配对生产研究提高了评估质量+0.27，但负控制未能在良好形成的语料库块之外建立内容特异性。这些结果支持暴露会计作为基于语料库评估的常规控制。该会计区分了超出暴露的恢复与暴露项目的遗漏；它并不确定推理是否发生。
cs.CL / 169 / 2609.24890
OSWorld-Pro: Process-based Evaluation for Computer Use Agents
OSWorld-Pro：基于过程的计算机使用代理评估
Abstract
Evaluation of Computer-Use Agents (CUAs) is often limited to the final deliverables they create (at the end of hundreds of steps) and assessed with functional verifiers, as seen in OSWorld. However, such evaluation of end-state performance lacks transparency into how and why agents fail in various tasks, obfuscating critical insight for subsequent improvement. For instance, agents that err during keyboard inputs would require a different mitigation strategy from those that fail to precisely provide click-based inputs on the graphical UI. We introduce OSWorld-Pro: a set of over 300 tasks containing over 2800 subgoals to enable the procedural evaluation of CUAs grounded in over 67,000 human annotations. We use robust human-aligned LLM-Judges to evaluate the fulfillment of OSWorld-Pro subgoals and thereby reveal the progress that models make throughout a series of sequentially dependent subgoals. Our findings reveal that OSWorld-Pro is challenging even for state-of-the-art LLMs, with top performers like Claude Opus 5 achieving only 75.7% vs. 83.4% on OSWorld. Furthermore, we identify critical process-focused failure modes of various models (e.g. subgoal-irrelevant actions and click-based mistakes) to provide insights to improve performance and efficiency of CUAs.
Chinese Translation
计算机使用代理（CUAs）的评估通常仅限于它们在数百个步骤结束时所创建的最终交付成果，并通过功能验证器进行评估，如在OSWorld中所见。然而，这种对最终状态性能的评估缺乏对代理在各种任务中失败的原因和方式的透明度，掩盖了后续改进所需的重要见解。例如，在键盘输入过程中出错的代理需要与在图形用户界面上无法准确提供基于点击的输入的代理不同的缓解策略。我们引入了OSWorld-Pro：一套包含超过300个任务和2800个子目标的任务集合，以便基于超过67,000个人类注释实现CUAs的过程评估。我们使用与人类对齐的强大LLM-Judges来评估OSWorld-Pro子目标的完成情况，从而揭示模型在一系列顺序依赖的子目标中的进展。我们的研究发现，OSWorld-Pro对最先进的LLM来说也是具有挑战性的，表现最佳的模型如Claude Opus 5在OSWorld-Pro上的得分仅为75.7%，而在OSWorld上为83.4%。此外，我们识别了各种模型的关键过程聚焦失败模式（例如，与子目标无关的行为和基于点击的错误），以提供改进CUAs性能和效率的见解。
cs.CL / 170 / 2609.24895
Human-LLM Deliberation as Interactive Proof: Conditions for Verifiability Without Transparency
人类-大语言模型的审议作为互动证明：无透明性的可验证性条件
Abstract
When an LLM supplies an argument that a user could not readily construct, how can the user decide whether to accept its claim? Inspired by interactive proofs, we model human-LLM deliberation as an interaction between a prover with unrestricted internal search and a resource-bounded human verifier. The verifier requests and checks supporting details without access to the LLM's internal state. Passed checks accumulate evidence toward an acceptance threshold. We prove anytime-valid soundness against adaptive provers: the probability of ever accepting a false claim is at most a chosen error level, provided the task supplies bounds on false passes and human checking errors that remain valid after every relevant history. A finite-horizon completeness bound additionally requires bounds on the adequacy of honest responses and sufficient diagnostic progress. Further checks can strengthen the evidence for acceptance, but each requires another adequate response and reliable human effort. Whether this tradeoff permits certification depends on the verifier's effort budget, cognitive load, expertise, and fatigue. We identify conditions under which the supplied bounds certify a specified sequence of local checks but not a specified global check under the same resource budgets.
Chinese Translation
当一个大语言模型（LLM）提供一个用户无法轻易构建的论点时，用户如何决定是否接受其主张？受到互动证明的启发，我们将人类-LLM的审议建模为一个证明者与资源受限的人类验证者之间的互动，其中证明者具有无限制的内部搜索能力。验证者在没有访问LLM内部状态的情况下请求并检查支持细节。通过的检查积累证据，直到达到接受阈值。我们证明了针对自适应证明者的随时有效的健全性：在每个相关历史之后，接受虚假主张的概率至多为所选的错误水平，前提是任务提供了对虚假通过和人类检查错误的界限，这些界限在每个相关历史之后仍然有效。有限视野的完整性界限还要求对诚实响应的充分性和足够的诊断进展进行界限限制。进一步的检查可以增强接受的证据，但每个检查都需要另一个充分的响应和可靠的人类努力。这种权衡是否允许认证取决于验证者的努力预算、认知负担、专业知识和疲劳程度。我们识别出在相同资源预算下，所提供的界限能够认证指定的局部检查序列，但不能认证指定的全局检查的条件。
cs.CL / 171 / 2609.24903
ToneCL: Contrastive Learning for Few-Shot Syllable-Level Tone Classification
ToneCL：用于少样本音节级声调分类的对比学习
Abstract
Tone languages constitute over 50-70% of the world's languages, but the vast majority are low-resource, lacking the large transcribed corpora needed for automatic tone classification. Existing datasets are typically collected at the sentence level, whereas field linguists require fine-grained syllable-level annotations. We propose ToneCL, a lightweight contrastive learning framework for few-shot syllable-level tone classification. We simulate low-resource conditions on Mandarin and Vietnamese, limiting labeled data to tens of examples per tone class. ToneCL is pretrained on unlabeled speech with augmentations that preserve tonal identity, then fine-tuned on few-shot examples. Experiments show our method consistently outperforms baselines, achieving 91.6% on six-speaker Mandarin at 10 shots. Cross-lingual transfer is also effective: pretraining on Vietnamese and fine-tuning on Mandarin reaches 91.0\% accuracy at 10 shots. Ablation confirms that frequency band rejection is the most critical augmentation.
Chinese Translation
声调语言占世界语言的50-70%以上，但绝大多数属于低资源语言，缺乏自动声调分类所需的大型转录语料库。现有数据集通常在句子级别收集，而田野语言学家需要细粒度的音节级注释。我们提出了ToneCL，一种轻量级的对比学习框架，用于少样本音节级声调分类。我们在普通话和越南语上模拟低资源条件，将每个声调类别的标注数据限制为几十个示例。ToneCL在未标注的语音上进行预训练，采用保留声调特征的增强方法，然后在少样本示例上进行微调。实验表明，我们的方法始终优于基线，在10个样本的情况下，在六位说话者的普通话上达到了91.6%的准确率。跨语言迁移同样有效：在越南语上预训练并在普通话上微调，在10个样本时达到了91.0%的准确率。消融实验确认频带拒绝是最关键的增强方法。
cs.CL / 172 / 2609.24911
SocioVerse2: A Longitudinal Dynamic Social Simulation Framework under a Human-AI Co-evolutionary Paradigm
SocioVerse2：一种基于人类与人工智能共演化范式的纵向动态社会模拟框架
Zhang, Xinnong, Lin, Jiayu, Wang, Jia, Huang, Yixu, Mou, Xinyi, Wu, Yingqian, Liang, Jingcong, Lei, Shijun, Shi, Jianing, Li, Guanying, Wang, Siyuan, Lyu, Hanjia, Yin, Zhenfei, Yin, Yunlu, Chen, Siming, He, Yulan, Luo, Jiebo, Huang, Xuanjing, Jin, Liyin, Zhou, Baohua, Yan, Hanqi, Wei, Zhongyu
Abstract
Social simulation offers the social sciences an experimental instrument that the real world cannot supply, and generative agents have transformed it by acting as silicon samples that unite agent-based modeling with real behavioral data. Existing platforms verify collective behavior, align simulated populations with real societies in cross-sections, and employ autonomous agents for the research process. However, two social science requirements remain without systematic support: intervention in the content of a simulation and the researcher's control over the process that produces it. We present SocioVerse2, which extends SocioVerse 1.0 into a human-AI co-evolutionary paradigm built from two loops and one infrastructure. The longitudinal simulation loop simulates the target population with evolving environments and forks counterfactual branches via interventions. The controllable research loop takes the study itself as an editable state and updates state versions via controllable editing. The social science agentic infrastructure carries both loops through composable skills with researcher checkpoints, a population service over five persona pools, and an environment service over 21 real-world signal sources with point-in-time guarantees. We validate SocioVerse2 across three case families and seven case studies, from reproducing canonical agent-based models to modeling policy processes on real records and nowcasting macro-economic indices beyond the response model's knowledge cutoff. With the human-AI co-evolutionary paradigm, these cases go beyond system demonstrations to become substantive studies that investigate frontier questions in their respective disciplines. Code, data services, and a workbench are released as open-source resources.
Chinese Translation
社会模拟为社会科学提供了一种实验工具，这是现实世界无法提供的，而生成代理通过作为硅样本的角色，结合了基于代理的建模与真实行为数据，改变了这一领域。现有平台验证了集体行为，将模拟人群与真实社会在横截面上对齐，并在研究过程中使用自主代理。然而，有两个社会科学的需求仍然缺乏系统支持：对模拟内容的干预以及研究者对产生该内容过程的控制。我们提出了SocioVerse2，它在一个由两个循环和一个基础设施构建的人类与人工智能共演化范式中扩展了SocioVerse 1.0。纵向模拟循环模拟了具有不断演变环境的目标人群，并通过干预分叉出反事实分支。可控研究循环将研究本身视为可编辑状态，并通过可控编辑更新状态版本。社会科学代理基础设施通过可组合技能、研究者检查点、覆盖五个角色池的人口服务以及覆盖21个真实信号源的环境服务，支持这两个循环，并提供时间点保证。我们在三个案例家族和七个案例研究中验证了SocioVerse2，从再现经典的基于代理的模型到对真实记录的政策过程建模，以及超越响应模型知识截止点的宏观经济指标的即时预测。通过人类与人工智能的共演化范式，这些案例超越了系统演示，成为实质性研究，探讨各自学科中的前沿问题。代码、数据服务和工作台作为开源资源发布。
cs.CL / 173 / 2609.24932
Linguistic Features for Interpretable Textual Entailment
可解释文本蕴含的语言特征
Abstract
Despite the success of neural models in natural language processing, their black-box nature limits interpretability and conceals the linguistic phenomena underlying their predictions. We present SLITE, an explainable hybrid model for Recognizing Textual Entailment that integrates two complementary layers of semantic analysis: a structural-relational layer, based on semantic compatibility and incompatibility between compositional entities, and a distributional-informational layer, based on structured patterns of information change between embedding-based representations of the premise and the hypothesis. We propose 17 features that combine entity-level semantic relations, polarity-sensitive lexical matching, and alignment measures over semantic sub-representations of the similarity matrix, including measures based on entropy and transfer entropy. A logistic regression trained on these features achieves an accuracy of 83% on three-class SICK and 96% on SICK-CE, outperforming IsoLex by 4 percentage points and falling within 2 percentage points of RoBERTa with a fraction of its computational complexity. Ablation studies and SHAP analysis confirm that structural-relational features are the primary drivers of classification, while distributional-informational features provide essential complementary contributions, particularly for detecting neutrality and contradiction. Our results demonstrate that further exploration of hybrid approaches is a viable and scientifically productive alternative to massive neural architectures, and we hope they will strengthen the dialogue between linguistic theory and computational modeling of inference
Chinese Translation
尽管神经模型在自然语言处理领域取得了成功，但其黑箱特性限制了可解释性，并掩盖了其预测背后的语言现象。我们提出了SLITE，一种可解释的混合模型，用于识别文本蕴含，该模型整合了两层互补的语义分析：基于组合实体之间语义兼容性和不兼容性的结构-关系层，以及基于前提和假设的嵌入表示之间信息变化的结构模式的分布-信息层。我们提出了17个特征，结合了实体级语义关系、极性敏感的词汇匹配和对相似性矩阵的语义子表示的对齐度量，包括基于熵和转移熵的度量。基于这些特征训练的逻辑回归模型在三类SICK数据集上达到了83%的准确率，在SICK-CE上达到了96%的准确率，超越了IsoLex 4个百分点，并在计算复杂度仅为其一小部分的情况下，接近RoBERTa 2个百分点。消融研究和SHAP分析确认，结构-关系特征是分类的主要驱动因素，而分布-信息特征则提供了重要的互补贡献，特别是在检测中立性和矛盾方面。我们的结果表明，进一步探索混合方法是对大规模神经架构的一种可行且科学有效的替代方案，我们希望这能加强语言理论与推理计算建模之间的对话。
cs.CL / 174 / 2609.24965
Jev for Scientific Decisions: Evaluating Semantic Choices and Their Consequences
科学决策中的 Jev：评估语义选择及其后果
Abstract
Scientific workflows often require choosing among known relations before a deterministic calculation can proceed. Whether observations share a culture, treatment or reference standard can change the scientific meaning of the resulting count or comparison. We evaluate Jev as a semantic decision component using a harness that follows its documented guidance and assigns arithmetic to code. The study compares twelve model configurations on twenty source-grounded Choices across ten scientific cases, each repeated five times. We measure semantic selections, downstream outputs and final claim labels separately. Jev matched five other configurations at complete semantic correctness and achieved the lowest observed median latency among successful responses. Across three comparison models, seven wrong selections on one culture-history question changed downstream counts while preserving the correct final label. These results identify a useful role for Jev in prepared scientific decision tasks and show why evaluating that role requires checking the relations and quantities that a workflow will reuse.
Chinese Translation
科学工作流通常需要在已知关系中进行选择，才能进行确定性计算。观察是否共享文化、治疗或参考标准可能会改变结果计数或比较的科学意义。我们使用遵循文档指导并将算术分配给代码的工具，评估 Jev 作为语义决策组件。本研究比较了十二种模型配置在十个科学案例中的二十个基于源的选择，每个案例重复五次。我们分别测量语义选择、下游输出和最终声明标签。Jev 在完全语义正确性方面与其他五种配置相匹配，并在成功响应中实现了最低的观察到的中位延迟。在三个比较模型中，针对一个文化-历史问题的七个错误选择改变了下游计数，同时保留了正确的最终标签。这些结果确定了 Jev 在准备好的科学决策任务中的有用角色，并表明评估该角色需要检查工作流将重用的关系和数量。
cs.CL / 175 / 2609.24971
DolphinBench: Mapping the Pareto Frontier of Agent Memory
海豚基准：映射智能体记忆的帕累托前沿
Abstract
Agents today often take real-world actions that depend on long-term memory and context recall over time. However, most current memory benchmarks are built for a conversational question-answer format, where the question itself signals that some fact must be retrieved, and often which one. Moreover, benchmarks rarely require anything beyond accuracy from submissions, allowing memory systems to make unreasonable cost/time tradeoffs to achieve higher scores. We present DolphinBench, a benchmark that evaluates memory directly through an agent's task completion. DolphinBench includes three knowledge-work personas with roughly 500k tokens of user messages per persona and evaluates agents on tasks that depend on information from that history. We verify all 200 tasks per persona by running an agent with and without the relevant history, requiring success with it and failure without it. Finally, we require all evaluations to report total cost and latency alongside accuracy, which enables us to evaluate agent memory systems holistically. No existing memory benchmark combines all three. The dataset and evaluation code are available at https://dolphinbench.ai.
Chinese Translation
如今的智能体常常采取依赖于长期记忆和上下文回忆的现实世界行动。然而，目前大多数记忆基准都是为对话问答格式构建的，其中问题本身就暗示必须检索某些事实，并且通常指明是哪一个。此外，基准很少要求提交内容超出准确性，这使得记忆系统可以做出不合理的成本/时间权衡以获得更高的分数。我们提出了海豚基准（DolphinBench），这是一个通过智能体的任务完成情况直接评估记忆的基准。海豚基准包括三个知识工作角色，每个角色大约有50万字的用户消息，并评估智能体在依赖于该历史信息的任务上的表现。我们通过运行具有和不具有相关历史的智能体来验证每个角色的200个任务，要求在有历史的情况下成功，而在没有历史的情况下失败。最后，我们要求所有评估报告总成本和延迟以及准确性，这使我们能够全面评估智能体记忆系统。现有的记忆基准没有结合这三者。数据集和评估代码可在 https://dolphinbench.ai 获取。
cs.CL / 176 / 2609.24983
onPanda: Efficient Annotation of On-Policy Alignment Data for LLMs and Agents via Token-Level Correction
onPanda：通过令牌级修正高效注释 LLM 和智能体的在线对齐数据
Abstract
We present onPanda, an interactive tool for efficiently annotating LLM alignment data and agent trajectories. onPanda adopts token-level correction as its core interaction: while reading a model response, the annotator locates the first inappropriate token and either picks a substitute from the model's candidate tokens or types the correct text via free-form editing. The system then truncates everything after that position and continues generation from the corrected prefix, repeating this locate-correct-continue loop until a satisfactory response is obtained. This mechanism lets annotators precisely steer model outputs at low cost: a small controlled study suggests that onPanda reduces median annotation time by 52% over manual post-editing. Since the vast majority of tokens in the final response are generated by the model itself, the resulting data largely preserves the model's sampling distribution and is well suited for constructing on-policy SFT and preference data. Furthermore, the token-level corrections recorded during annotation provide fine-grained supervision with precise positions and naturally paired positive--negative samples. onPanda also connects to external tools and harnesses, enabling interactive trajectory annotation in realistic environments. In addition, we release Panda-CVL, a dataset annotated with onPanda, together with a benchmark for token-level correction.
Chinese Translation
我们提出了 onPanda，这是一种高效注释 LLM 对齐数据和智能体轨迹的交互工具。onPanda 以令牌级修正作为其核心交互方式：在阅读模型响应时，注释者定位第一个不适当的令牌，并从模型的候选令牌中选择一个替代项，或通过自由格式编辑输入正确的文本。系统随后截断该位置之后的所有内容，并从修正后的前缀继续生成，重复这一定位-修正-继续的循环，直到获得令人满意的响应。该机制使注释者能够以低成本精确引导模型输出：一项小规模的受控研究表明，onPanda 将中位注释时间减少了 52%，相较于手动后期编辑。由于最终响应中绝大多数令牌是由模型自身生成的，因此生成的数据在很大程度上保留了模型的采样分布，适合用于构建在线策略的 SFT 和偏好数据。此外，在注释过程中记录的令牌级修正提供了精细的监督，具备精确的位置和自然配对的正负样本。onPanda 还连接到外部工具并进行利用，使得在真实环境中进行交互式轨迹注释成为可能。此外，我们发布了使用 onPanda 注释的 Panda-CVL 数据集，并提供了令牌级修正的基准测试。